diff --git a/go.mod b/go.mod index a0de8138b..77e64dc41 100644 --- a/go.mod +++ b/go.mod @@ -2,12 +2,12 @@ module github.com/MixinNetwork/mixin go 1.22.6 -replace github.com/dgraph-io/badger/v4 => github.com/MixinNetwork/badger/v4 v4.2.0-F1 +replace github.com/dgraph-io/badger/v4 => github.com/MixinNetwork/badger/v4 v4.3.0-F1 require ( filippo.io/edwards25519 v1.1.0 - github.com/dgraph-io/badger/v4 v4.2.0 - github.com/dgraph-io/ristretto v0.1.1 + github.com/dgraph-io/badger/v4 v4.3.0 + github.com/dgraph-io/ristretto v0.1.2-0.20240116140435-c67e07994f91 github.com/pelletier/go-toml v1.9.5 github.com/quic-go/quic-go v0.46.0 github.com/shopspring/decimal v1.4.0 @@ -25,20 +25,18 @@ require ( github.com/go-task/slim-sprig v0.0.0-20230315185526-52ccab3ef572 // indirect github.com/go-task/slim-sprig/v3 v3.0.0 // indirect github.com/gogo/protobuf v1.3.2 // indirect - github.com/golang/glog v1.2.2 // indirect github.com/golang/protobuf v1.5.4 // indirect - github.com/golang/snappy v0.0.4 // indirect github.com/google/flatbuffers v24.3.25+incompatible // indirect - github.com/google/pprof v0.0.0-20240727154555-813a5fbdbec8 // indirect + github.com/google/pprof v0.0.0-20240829160300-da1f7e9f2b25 // indirect github.com/klauspost/compress v1.17.9 // indirect github.com/klauspost/cpuid/v2 v2.2.8 // indirect - github.com/onsi/ginkgo/v2 v2.20.0 // indirect + github.com/onsi/ginkgo/v2 v2.20.2 // indirect github.com/pkg/errors v0.9.1 // indirect github.com/pmezard/go-difflib v1.0.0 // indirect github.com/russross/blackfriday/v2 v2.1.0 // indirect github.com/xrash/smetrics v0.0.0-20240521201337-686a1a2994c1 // indirect go.uber.org/mock v0.4.0 // indirect - golang.org/x/exp v0.0.0-20240808152545-0cdaa3abc0fa // indirect + golang.org/x/exp v0.0.0-20240823005443-9b4947da3948 // indirect golang.org/x/mod v0.20.0 // indirect golang.org/x/net v0.28.0 // indirect golang.org/x/sync v0.8.0 // indirect diff --git a/go.sum b/go.sum index 40a3ad67d..c8b58859a 100644 --- a/go.sum +++ b/go.sum @@ -1,8 +1,7 @@ filippo.io/edwards25519 v1.1.0 h1:FNf4tywRC1HmFuKW5xopWpigGjJKiJSV0Cqo0cJWDaA= filippo.io/edwards25519 v1.1.0/go.mod h1:BxyFTGdWcka3PhytdK4V28tE5sGfRvvvRV7EaN4VDT4= -github.com/MixinNetwork/badger/v4 v4.2.0-F1 h1:C8K6AYHsfqTPv7eSiSv4wl1awrW2dgkM6EQU8Sbwq7w= -github.com/MixinNetwork/badger/v4 v4.2.0-F1/go.mod h1:eoudq8Vt4sWtlqwVp9XZ1Keww3/rjYZHoffiMV4IEuY= -github.com/cespare/xxhash/v2 v2.1.1/go.mod h1:VGX0DQ3Q6kWi7AoAeZDth3/j3BFtOZR5XLFGgcrjCOs= +github.com/MixinNetwork/badger/v4 v4.3.0-F1 h1:s6/DUM3qFyqPcESwSerpvCmWtpQ2TL8wQt27vLMRE1A= +github.com/MixinNetwork/badger/v4 v4.3.0-F1/go.mod h1:42o0EL86yekTcj8jpQ7sX2s3hE6A8VC57pJBiC3IqdU= github.com/cespare/xxhash/v2 v2.3.0 h1:UL815xU9SqsFlibzuggzjXhog7bL6oX9BbNZnL2UFvs= github.com/cespare/xxhash/v2 v2.3.0/go.mod h1:VGX0DQ3Q6kWi7AoAeZDth3/j3BFtOZR5XLFGgcrjCOs= github.com/cpuguy83/go-md2man/v2 v2.0.4 h1:wfIWP927BUkWJb2NmU/kNDYIBTh/ziUX91+lVfRxZq4= @@ -10,10 +9,8 @@ github.com/cpuguy83/go-md2man/v2 v2.0.4/go.mod h1:tgQtvFlXSQOSOSIRvRPT7W67SCa46t github.com/davecgh/go-spew v1.1.0/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38= github.com/davecgh/go-spew v1.1.1 h1:vj9j/u1bqnvCEfJOwUhtlOARqs3+rkHYY13jYWTU97c= github.com/davecgh/go-spew v1.1.1/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38= -github.com/dgraph-io/ristretto v0.1.1 h1:6CWw5tJNgpegArSHpNHJKldNeq03FQCwYvfMVWajOK8= -github.com/dgraph-io/ristretto v0.1.1/go.mod h1:S1GPSBCYCIhmVNfcth17y2zZtQT6wzkzgwUve0VDWWA= -github.com/dgryski/go-farm v0.0.0-20190423205320-6a90982ecee2/go.mod h1:SqUrOPUnsFjfmXRMNPybcSiG0BgUW2AuFH8PAnS2iTw= -github.com/dustin/go-humanize v1.0.0/go.mod h1:HtrtbFcZ19U5GC7JDqmcUSB87Iq5E25KnS6fMYU6eOk= +github.com/dgraph-io/ristretto v0.1.2-0.20240116140435-c67e07994f91 h1:Pux6+xANi0I7RRo5E1gflI4EZ2yx3BGZ75JkAIvGEOA= +github.com/dgraph-io/ristretto v0.1.2-0.20240116140435-c67e07994f91/go.mod h1:swkazRqnUf1N62d0Nutz7KIj2UKqsm/H8tD0nBJAXqM= github.com/dustin/go-humanize v1.0.1 h1:GzkhY7T5VNhEkwH0PVJgjz+fX1rhBrR7pRT3mDkpeCY= github.com/dustin/go-humanize v1.0.1/go.mod h1:Mu1zIs6XwVuF/gI1OepvI0qD18qycQx+mFykh5fBlto= github.com/go-task/slim-sprig v0.0.0-20230315185526-52ccab3ef572 h1:tfuBGBXKqDEevZMzYi5KSi8KkcZtzBcTgAUUtapy0OI= @@ -22,25 +19,20 @@ github.com/go-task/slim-sprig/v3 v3.0.0 h1:sUs3vkvUymDpBKi3qH1YSqBQk9+9D/8M2mN1v github.com/go-task/slim-sprig/v3 v3.0.0/go.mod h1:W848ghGpv3Qj3dhTPRyJypKRiqCdHZiAzKg9hl15HA8= github.com/gogo/protobuf v1.3.2 h1:Ov1cvc58UF3b5XjBnZv7+opcTcQFZebYjWzi34vdm4Q= github.com/gogo/protobuf v1.3.2/go.mod h1:P1XiOD3dCwIKUDQYPy72D8LYyHL2YPYrpS2s69NZV8Q= -github.com/golang/glog v0.0.0-20160126235308-23def4e6c14b/go.mod h1:SBH7ygxi8pfUlaOkMMuAQtPIUF8ecWP5IEl/CR7VP2Q= -github.com/golang/glog v1.2.2 h1:1+mZ9upx1Dh6FmUTFR1naJ77miKiXgALjWOZ3NVFPmY= -github.com/golang/glog v1.2.2/go.mod h1:6AhwSGph0fcJtXVM/PEHPqZlFeoLxhs7/t5UDAwmO+w= github.com/golang/protobuf v1.5.4 h1:i7eJL8qZTpSEXOPTxNKhASYpMn+8e5Q6AdndVa1dWek= github.com/golang/protobuf v1.5.4/go.mod h1:lnTiLA8Wa4RWRcIUkrtSVa5nRhsEGBg48fD6rSs7xps= -github.com/golang/snappy v0.0.4 h1:yAGX7huGHXlcLOEtBnF4w7FQwA26wojNCwOYAEhLjQM= -github.com/golang/snappy v0.0.4/go.mod h1:/XxbfmMg8lxefKM7IXC3fBNl/7bRcc72aCRzEWrmP2Q= github.com/google/flatbuffers v24.3.25+incompatible h1:CX395cjN9Kke9mmalRoL3d81AtFUxJM+yDthflgJGkI= github.com/google/flatbuffers v24.3.25+incompatible/go.mod h1:1AeVuKshWv4vARoZatz6mlQ0JxURH0Kv5+zNeJKJCa8= -github.com/google/pprof v0.0.0-20240727154555-813a5fbdbec8 h1:FKHo8hFI3A+7w0aUQuYXQ+6EN5stWmeY/AZqtM8xk9k= -github.com/google/pprof v0.0.0-20240727154555-813a5fbdbec8/go.mod h1:K1liHPHnj73Fdn/EKuT8nrFqBihUSKXoLYU0BuatOYo= +github.com/google/pprof v0.0.0-20240829160300-da1f7e9f2b25 h1:sEDPKUw6iPjczdu33njxFjO6tYa9bfc0z/QyB/zSsBw= +github.com/google/pprof v0.0.0-20240829160300-da1f7e9f2b25/go.mod h1:vavhavw2zAxS5dIdcRluK6cSGGPlZynqzFM8NdvU144= github.com/kisielk/errcheck v1.5.0/go.mod h1:pFxgyoBC7bSaBwPgfKdkLd5X25qrDl4LWUI2bnpBCr8= github.com/kisielk/gotool v1.0.0/go.mod h1:XhKaO+MFFWcvkIS/tQcRk01m1F5IRFswLeQ+oQHNcck= github.com/klauspost/compress v1.17.9 h1:6KIumPrER1LHsvBVuDa0r5xaG0Es51mhhB9BQB2qeMA= github.com/klauspost/compress v1.17.9/go.mod h1:Di0epgTjJY877eYKx5yC51cX2A2Vl2ibi7bDH9ttBbw= github.com/klauspost/cpuid/v2 v2.2.8 h1:+StwCXwm9PdpiEkPyzBXIy+M9KUb4ODm0Zarf1kS5BM= github.com/klauspost/cpuid/v2 v2.2.8/go.mod h1:Lcz8mBdAVJIBVzewtcLocK12l3Y+JytZYpaMropDUws= -github.com/onsi/ginkgo/v2 v2.20.0 h1:PE84V2mHqoT1sglvHc8ZdQtPcwmvvt29WLEEO3xmdZw= -github.com/onsi/ginkgo/v2 v2.20.0/go.mod h1:lG9ey2Z29hR41WMVthyJBGUBcBhGOtoPF2VFMvBXFCI= +github.com/onsi/ginkgo/v2 v2.20.2 h1:7NVCeyIWROIAheY21RLS+3j2bb52W0W82tkberYytp4= +github.com/onsi/ginkgo/v2 v2.20.2/go.mod h1:K9gyxPIlb+aIvnZ8bd9Ak+YP18w3APlR+5coaZoE2ag= github.com/pelletier/go-toml v1.9.5 h1:4yBQzkHv+7BHq2PQUZF3Mx0IYxG7LsP222s7Agd3ve8= github.com/pelletier/go-toml v1.9.5/go.mod h1:u1nR/EPcESfeI/szUZKdtJ0xRNbUoANCkoOuaOx1Y+c= github.com/pkg/errors v0.9.1 h1:FEBLx1zS214owpjy7qsBeixbURkuhQAwrK5UwLGTwt4= @@ -54,7 +46,6 @@ github.com/russross/blackfriday/v2 v2.1.0/go.mod h1:+Rmxgy9KzJVeS9/2gXHxylqXiyQD github.com/shopspring/decimal v1.4.0 h1:bxl37RwXBklmTi0C79JfXCEBD1cqqHt0bbgBAGFp81k= github.com/shopspring/decimal v1.4.0/go.mod h1:gawqmDU56v4yIKSwfBSFip1HdCCXN8/+DMd9qYNcwME= github.com/stretchr/objx v0.1.0/go.mod h1:HFkY916IF+rwdDfMAkV7OtwuqBVzrE8GR6GFx+wExME= -github.com/stretchr/testify v1.4.0/go.mod h1:j7eGeouHqKxXV5pUuKE4zz7dFj8WfuZ+81PSLYec5m4= github.com/stretchr/testify v1.6.1/go.mod h1:6Fq8oRcR53rry900zMqJjRRixrwX3KX962/h/Wwjteg= github.com/stretchr/testify v1.9.0 h1:HtqpIVDClZ4nwg75+f6Lvsy/wHu+3BoSGCbBAcpTsTg= github.com/stretchr/testify v1.9.0/go.mod h1:r2ic/lqez/lEtzL7wO/rwa5dbSLXVDPFyf8C91i36aY= @@ -73,8 +64,8 @@ golang.org/x/crypto v0.0.0-20191011191535-87dc89f01550/go.mod h1:yigFU9vqHzYiE8U golang.org/x/crypto v0.0.0-20200622213623-75b288015ac9/go.mod h1:LzIPMQfyMNhhGPhUkYOs5KpL4U8rLKemX1yGLhDgUto= golang.org/x/crypto v0.26.0 h1:RrRspgV4mU+YwB4FYnuBoKsUapNIL5cohGAmSH3azsw= golang.org/x/crypto v0.26.0/go.mod h1:GY7jblb9wI+FOo5y8/S2oY4zWP07AkOJ4+jxCqdqn54= -golang.org/x/exp v0.0.0-20240808152545-0cdaa3abc0fa h1:ELnwvuAXPNtPk1TJRuGkI9fDTwym6AYBu0qzT8AcHdI= -golang.org/x/exp v0.0.0-20240808152545-0cdaa3abc0fa/go.mod h1:akd2r19cwCdwSwWeIdzYQGa/EZZyqcOdwWiwj5L5eKQ= +golang.org/x/exp v0.0.0-20240823005443-9b4947da3948 h1:kx6Ds3MlpiUHKj7syVnbp57++8WpuKPcR5yjLBjvLEA= +golang.org/x/exp v0.0.0-20240823005443-9b4947da3948/go.mod h1:akd2r19cwCdwSwWeIdzYQGa/EZZyqcOdwWiwj5L5eKQ= golang.org/x/mod v0.2.0/go.mod h1:s0Qsj1ACt9ePp/hMypM3fl4fZqREWJwdYDEqhRiZZUA= golang.org/x/mod v0.3.0/go.mod h1:s0Qsj1ACt9ePp/hMypM3fl4fZqREWJwdYDEqhRiZZUA= golang.org/x/mod v0.20.0 h1:utOm6MM3R3dnawAiJgn0y+xvuYRsm1RKM/4giyfDgV0= @@ -93,7 +84,6 @@ golang.org/x/sync v0.8.0/go.mod h1:Czt+wKu1gCyEFDUtn0jG5QVvpJ6rzVqr5aXyt9drQfk= golang.org/x/sys v0.0.0-20190215142949-d0b11bdaac8a/go.mod h1:STP8DvDyc/dI5b8T5hshtkjS+E42TnysNCUPdjciGhY= golang.org/x/sys v0.0.0-20190412213103-97732733099d/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs= golang.org/x/sys v0.0.0-20200930185726-fdedc70b468f/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs= -golang.org/x/sys v0.0.0-20221010170243-090e33056c14/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg= golang.org/x/sys v0.5.0/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg= golang.org/x/sys v0.24.0 h1:Twjiwq9dn6R1fQcyiK+wQyHWfaz/BJB+YIpzU/Cv3Xg= golang.org/x/sys v0.24.0/go.mod h1:/VUhepiaJMQUp4+oa/7Zr1D23ma6VTLIYjOOTFZPUcA= @@ -112,7 +102,6 @@ golang.org/x/xerrors v0.0.0-20200804184101-5ec99f83aff1/go.mod h1:I/5z698sn9Ka8T google.golang.org/protobuf v1.34.2 h1:6xV6lTsCfpGD21XK49h7MhtcApnLqkfYgPcdHftf6hg= google.golang.org/protobuf v1.34.2/go.mod h1:qYOHts0dSfpeUzUFpOMr/WGzszTmLH+DiWniOlNbLDw= gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405/go.mod h1:Co6ibVJAznAaIkqp8huTwlJQCZ016jof/cbN4VW5Yz0= -gopkg.in/yaml.v2 v2.2.2/go.mod h1:hI93XBmqTisBFMUTm0b8Fm+jr3Dg1NNxqwp+5A1VGuI= gopkg.in/yaml.v3 v3.0.0-20200313102051-9f266ea9e77c/go.mod h1:K4uyk7z7BCEPqu6E+C64Yfv1cQ7kz7rIZviUmN+EgEM= gopkg.in/yaml.v3 v3.0.1 h1:fxVm/GzAzEWqLHuvctI91KS9hhNmmWOoWu0XTYJS7CA= gopkg.in/yaml.v3 v3.0.1/go.mod h1:K4uyk7z7BCEPqu6E+C64Yfv1cQ7kz7rIZviUmN+EgEM= diff --git a/kernel/election_test.go b/kernel/election_test.go index d5fc2d307..b49d15cc5 100644 --- a/kernel/election_test.go +++ b/kernel/election_test.go @@ -110,7 +110,7 @@ func setupTestNode(require *require.Assertions, dir string) *Node { gns, err := common.ReadGenesis(dir + "/genesis.json") require.Nil(err) - cache, err := ristretto.NewCache(&ristretto.Config{ + cache, err := ristretto.NewCache(&ristretto.Config[[]byte, any]{ NumCounters: 1e7, // number of keys to track frequency of (10M). MaxCost: 1 << 30, BufferItems: 64, // number of keys per Get buffer. diff --git a/kernel/node.go b/kernel/node.go index 7e2431028..ed42d731d 100644 --- a/kernel/node.go +++ b/kernel/node.go @@ -45,7 +45,7 @@ type Node struct { startAt time.Time networkId crypto.Hash persistStore storage.Store - cacheStore *ristretto.Cache + cacheStore *ristretto.Cache[[]byte, any] custom *config.Custom done chan struct{} @@ -69,7 +69,7 @@ type CNode struct { ConsensusIndex int } -func SetupNode(custom *config.Custom, store storage.Store, cache *ristretto.Cache, gns *common.Genesis) (*Node, error) { +func SetupNode(custom *config.Custom, store storage.Store, cache *ristretto.Cache[[]byte, any], gns *common.Genesis) (*Node, error) { node := &Node{ SyncPoints: &syncMap{mutex: new(sync.RWMutex), m: make(map[crypto.Hash]*p2p.SyncPoint)}, chains: &chainsMap{m: make(map[crypto.Hash]*Chain)}, @@ -428,7 +428,7 @@ func (node *Node) Uptime() time.Duration { return clock.Now().Sub(node.startAt) } -func (node *Node) GetCacheStore() *ristretto.Cache { +func (node *Node) GetCacheStore() *ristretto.Cache[[]byte, any] { return node.cacheStore } diff --git a/main.go b/main.go index c2cfb0184..694bdffde 100644 --- a/main.go +++ b/main.go @@ -712,9 +712,9 @@ func kernelCmd(c *cli.Context) error { return node.Loop() } -func newCache(conf *config.Custom) (*ristretto.Cache, error) { +func newCache(conf *config.Custom) (*ristretto.Cache[[]byte, any], error) { cost := int64(conf.Node.MemoryCacheSize * 1024 * 1024) - return ristretto.NewCache(&ristretto.Config{ + return ristretto.NewCache(&ristretto.Config[[]byte, any]{ NumCounters: cost / 1024 * 10, MaxCost: cost, BufferItems: 64, diff --git a/p2p/handle.go b/p2p/handle.go index 4726a4906..b9630a209 100644 --- a/p2p/handle.go +++ b/p2p/handle.go @@ -64,7 +64,7 @@ type AuthToken struct { } type SyncHandle interface { - GetCacheStore() *ristretto.Cache + GetCacheStore() *ristretto.Cache[[]byte, any] SignData(data []byte) crypto.Signature BuildAuthenticationMessage(relayerId crypto.Hash) []byte AuthenticateAs(recipientId crypto.Hash, msg []byte, timeoutSec int64) (*AuthToken, error) diff --git a/p2p/util.go b/p2p/util.go index 3114eac85..36290a88c 100644 --- a/p2p/util.go +++ b/p2p/util.go @@ -11,7 +11,7 @@ import ( ) type confirmMap struct { - cache *ristretto.Cache + cache *ristretto.Cache[[]byte, any] } func (m *confirmMap) contains(key []byte, duration time.Duration) bool { diff --git a/rpc/consensus_test.go b/rpc/consensus_test.go index 1763c3509..c2a606149 100644 --- a/rpc/consensus_test.go +++ b/rpc/consensus_test.go @@ -1182,8 +1182,8 @@ func electSnapshotNode(nodes []*Node, node *kernel.Node, operation byte, now uin return nil } -func newCache(conf *config.Custom) *ristretto.Cache { - cache, err := ristretto.NewCache(&ristretto.Config{ +func newCache(conf *config.Custom) *ristretto.Cache[[]byte, any] { + cache, err := ristretto.NewCache(&ristretto.Config[[]byte, any]{ NumCounters: 1e7, // number of keys to track frequency of (10M). MaxCost: int64(conf.Node.MemoryCacheSize) * 1024 * 1024, BufferItems: 64, // number of keys per Get buffer. diff --git a/vendor/github.com/dgraph-io/badger/v4/CHANGELOG.md b/vendor/github.com/dgraph-io/badger/v4/CHANGELOG.md index a2ebd9b0b..acd51459a 100644 --- a/vendor/github.com/dgraph-io/badger/v4/CHANGELOG.md +++ b/vendor/github.com/dgraph-io/badger/v4/CHANGELOG.md @@ -12,6 +12,7 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/). ### Fixed +- fix(db): avoid panic in parallel reads after closing DB (#1987) - fix(logging): fix direct access to logger (#1980) - fix(sec): bump google.golang.org/grpc from 1.20.1 to 1.53.0 (#1977) - fix(sec): update gopkg.in/yaml.v2 package (#1969) diff --git a/vendor/github.com/dgraph-io/badger/v4/README.md b/vendor/github.com/dgraph-io/badger/v4/README.md index 0862aa4ca..06e91d5e6 100644 --- a/vendor/github.com/dgraph-io/badger/v4/README.md +++ b/vendor/github.com/dgraph-io/badger/v4/README.md @@ -39,7 +39,7 @@ Please consult the [Changelog] for more detailed information on releases. For more details on our version naming schema please read [Choosing a version](#choosing-a-version). -[Changelog]:https://github.com/dgraph-io/badger/blob/master/CHANGELOG.md +[Changelog]:https://github.com/dgraph-io/badger/blob/main/CHANGELOG.md ## Table of Contents - [BadgerDB](#badgerdb) @@ -62,7 +62,7 @@ For more details on our version naming schema please read [Choosing a version](# ## Getting Started ### Installing -To start using Badger, install Go 1.19 or above. Badger v3 needs go modules. From your project, run the following command +To start using Badger, install Go 1.19 or above. Badger v3 and above needs go modules. From your project, run the following command ```sh $ go get github.com/dgraph-io/badger/v4 @@ -219,6 +219,9 @@ Below is a list of known projects that use Badger: * [Loggie](https://github.com/loggie-io/loggie) - A lightweight, cloud-native data transfer agent and aggregator. * [raft-badger](https://github.com/rfyiamcool/raft-badger) - raft-badger implements LogStore and StableStore Interface of hashcorp/raft. it is used to store raft log and metadata of hashcorp/raft. * [DVID](https://github.com/janelia-flyem/dvid) - A dataservice for branched versioning of a variety of data types. Originally created for large-scale brain reconstructions in Connectomics. +* [KVS](https://github.com/tauraamui/kvs) - A library for making it easy to persist, load and query full structs into BadgerDB, using an ownership hierarchy model. +* [LLS](https://github.com/Boc-chi-no/LLS) - LLS is an efficient URL Shortener that can be used to shorten links and track link usage. Support for BadgerDB and MongoDB. Improved performance by more than 30% when using BadgerDB +* [lakeFS](https://github.com/treeverse/lakeFS) - lakeFS is an open-source data version control that transforms your object storage to Git-like repositories. lakeFS uses BadgerDB for its underlying local metadata KV store implementation. If you are using Badger in a project please send a pull request to add it to the list. diff --git a/vendor/github.com/dgraph-io/badger/v4/db.go b/vendor/github.com/dgraph-io/badger/v4/db.go index d30ac6c3d..ba060cc57 100644 --- a/vendor/github.com/dgraph-io/badger/v4/db.go +++ b/vendor/github.com/dgraph-io/badger/v4/db.go @@ -34,6 +34,7 @@ import ( humanize "github.com/dustin/go-humanize" "github.com/pkg/errors" + "github.com/dgraph-io/badger/v4/fb" "github.com/dgraph-io/badger/v4/options" "github.com/dgraph-io/badger/v4/pb" "github.com/dgraph-io/badger/v4/skl" @@ -123,8 +124,8 @@ type DB struct { pub *publisher registry *KeyRegistry - blockCache *ristretto.Cache - indexCache *ristretto.Cache + blockCache *ristretto.Cache[[]byte, *table.Block] + indexCache *ristretto.Cache[uint64, *fb.TableIndex] allocPool *z.AllocatorPool } @@ -162,10 +163,10 @@ func checkAndSetOptions(opt *Options) error { // the transaction APIs. Transaction batches entries into batches of size opt.maxBatchSize. if opt.ValueThreshold > opt.maxBatchSize { return errors.Errorf("Valuethreshold %d greater than max batch size of %d. Either "+ - "reduce opt.ValueThreshold or increase opt.MaxTableSize.", + "reduce opt.ValueThreshold or increase opt.BaseTableSize.", opt.ValueThreshold, opt.maxBatchSize) } - // ValueLogFileSize should be stricly LESS than 2<<30 otherwise we will + // ValueLogFileSize should be strictly LESS than 2<<30 otherwise we will // overflow the uint32 when we mmap it in OpenMemtable. if !(opt.ValueLogFileSize < 2<<30 && opt.ValueLogFileSize >= 1<<20) { return ErrValueLogSize @@ -274,14 +275,14 @@ func Open(opt Options) (*DB, error) { numInCache = 1 } - config := ristretto.Config{ + config := ristretto.Config[[]byte, *table.Block]{ NumCounters: numInCache * 8, MaxCost: opt.BlockCacheSize, BufferItems: 64, Metrics: true, OnExit: table.BlockEvictHandler, } - db.blockCache, err = ristretto.NewCache(&config) + db.blockCache, err = ristretto.NewCache[[]byte, *table.Block](&config) if err != nil { return nil, y.Wrap(err, "failed to create data cache") } @@ -297,7 +298,7 @@ func Open(opt Options) (*DB, error) { numInCache = 1 } - config := ristretto.Config{ + config := ristretto.Config[uint64, *fb.TableIndex]{ NumCounters: numInCache * 8, MaxCost: opt.IndexCacheSize, BufferItems: 64, @@ -402,7 +403,7 @@ func Open(opt Options) (*DB, error) { return db, nil } -// initBannedNamespaces retrieves the banned namepsaces from the DB and updates in-memory structure. +// initBannedNamespaces retrieves the banned namespaces from the DB and updates in-memory structure. func (db *DB) initBannedNamespaces() error { if db.opt.NamespaceOffset < 0 { return nil @@ -904,7 +905,7 @@ func (db *DB) sendToWriteCh(entries []*Entry) (*request, error) { return nil, ErrTxnTooBig } - // We can only service one request because we need each txn to be stored in a contigous section. + // We can only service one request because we need each txn to be stored in a contiguous section. // Txns should not interleave among other txns or rewrites. req := requestPool.Get().(*request) req.reset() @@ -1617,7 +1618,7 @@ func (db *DB) Flatten(workers int) error { } } if len(levels) <= 1 { - prios := db.lc.pickCompactLevels() + prios := db.lc.pickCompactLevels(nil) if len(prios) == 0 || prios[0].score <= 1.0 { db.opt.Infof("All tables consolidated into one level. Flattening done.\n") return nil @@ -1709,7 +1710,7 @@ func (db *DB) dropAll() (func(), error) { if err != nil { return f, err } - // prepareToDrop will stop all the incomming write and flushes any pending memtables. + // prepareToDrop will stop all the incoming write and flushes any pending memtables. // Before we drop, we'll stop the compaction because anyways all the datas are going to // be deleted. db.stopCompactions() @@ -1752,7 +1753,7 @@ func (db *DB) dropAll() (func(), error) { // DropPrefix would drop all the keys with the provided prefix. It does this in the following way: // - Stop accepting new writes. -// - Stop memtable flushes before acquiring lock. Because we're acquring lock here +// - Stop memtable flushes before acquiring lock. Because we're acquiring lock here // and memtable flush stalls for lock, which leads to deadlock // - Flush out all memtables, skipping over keys with the given prefix, Kp. // - Write out the value log header to memtables when flushing, so we don't accidentally bring Kp diff --git a/vendor/github.com/dgraph-io/badger/v4/iterator.go b/vendor/github.com/dgraph-io/badger/v4/iterator.go index 2f69db6b2..a4a611001 100644 --- a/vendor/github.com/dgraph-io/badger/v4/iterator.go +++ b/vendor/github.com/dgraph-io/badger/v4/iterator.go @@ -589,7 +589,7 @@ func (it *Iterator) Next() { // Set next item to current it.item = it.data.pop() - for it.iitr.Valid() { + for it.iitr.Valid() && hasPrefix(it.iitr, it.opt.Prefix) { if it.parseItem() { // parseItem calls one extra next. // This is used to deal with the complexity of reverse iteration. @@ -725,6 +725,13 @@ func (it *Iterator) fill(item *Item) { } } +func hasPrefix(it y.Iterator, prefix []byte) bool { + if len(prefix) > 0 { + return bytes.HasPrefix(y.ParseKey(it.Key()), prefix) + } + return true +} + func (it *Iterator) prefetch() { prefetchSize := 2 if it.opt.PrefetchValues && it.opt.PrefetchSize > 1 { @@ -734,7 +741,7 @@ func (it *Iterator) prefetch() { i := it.iitr var count int it.item = nil - for i.Valid() { + for i.Valid() && hasPrefix(i, it.opt.Prefix) { if !it.parseItem() { continue } diff --git a/vendor/github.com/dgraph-io/badger/v4/level_handler.go b/vendor/github.com/dgraph-io/badger/v4/level_handler.go index 31673f15b..fc81cc452 100644 --- a/vendor/github.com/dgraph-io/badger/v4/level_handler.go +++ b/vendor/github.com/dgraph-io/badger/v4/level_handler.go @@ -165,8 +165,8 @@ func (s *levelHandler) addTable(t *table.Table) { // sortTables sorts tables of levelHandler based on table.Smallest. // Normally it should be called after all addTable calls. func (s *levelHandler) sortTables() { - s.RLock() - defer s.RUnlock() + s.Lock() + defer s.Unlock() sort.Slice(s.tables, func(i, j int) bool { return y.CompareKeys(s.tables[i].Smallest(), s.tables[j].Smallest()) < 0 diff --git a/vendor/github.com/dgraph-io/badger/v4/levels.go b/vendor/github.com/dgraph-io/badger/v4/levels.go index b87c2430f..fbd0a7b7b 100644 --- a/vendor/github.com/dgraph-io/badger/v4/levels.go +++ b/vendor/github.com/dgraph-io/badger/v4/levels.go @@ -466,8 +466,13 @@ func (s *levelsController) runCompactor(id int, lc *z.Closer) { } return false } + + var priosBuffer []compactionPriority runOnce := func() bool { - prios := s.pickCompactLevels() + prios := s.pickCompactLevels(priosBuffer) + defer func() { + priosBuffer = prios + }() if id == 0 { // Worker ID zero prefers to compact L0 always. prios = moveL0toFront(prios) @@ -529,7 +534,9 @@ func (s *levelsController) lastLevel() *levelHandler { // pickCompactLevel determines which level to compact. // Based on: https://github.com/facebook/rocksdb/wiki/Leveled-Compaction -func (s *levelsController) pickCompactLevels() (prios []compactionPriority) { +// It tries to reuse priosBuffer to reduce memory allocation, +// passing nil is acceptable, then new memory will be allocated. +func (s *levelsController) pickCompactLevels(priosBuffer []compactionPriority) (prios []compactionPriority) { t := s.levelTargets() addPriority := func(level int, score float64) { pri := compactionPriority{ @@ -541,6 +548,12 @@ func (s *levelsController) pickCompactLevels() (prios []compactionPriority) { prios = append(prios, pri) } + // Grow buffer to fit all levels. + if cap(priosBuffer) < len(s.levels) { + priosBuffer = make([]compactionPriority, 0, len(s.levels)) + } + prios = priosBuffer[:0] + // Add L0 priority based on the number of tables. addPriority(0, float64(s.levels[0].numTables())/float64(s.kv.opt.NumLevelZeroTables)) @@ -1690,7 +1703,7 @@ type LevelInfo struct { func (s *levelsController) getLevelInfo() []LevelInfo { t := s.levelTargets() - prios := s.pickCompactLevels() + prios := s.pickCompactLevels(nil) result := make([]LevelInfo, len(s.levels)) for i, l := range s.levels { l.RLock() diff --git a/vendor/github.com/dgraph-io/badger/v4/options.go b/vendor/github.com/dgraph-io/badger/v4/options.go index ac046bc1d..bb6131b30 100644 --- a/vendor/github.com/dgraph-io/badger/v4/options.go +++ b/vendor/github.com/dgraph-io/badger/v4/options.go @@ -173,8 +173,6 @@ func DefaultOptions(path string) Options { // Benchmark code can be found in table/builder_test.go file ZSTDCompressionLevel: 1, - // Nothing to read/write value log using standard File I/O - // MemoryMap to mmap() the value log files // (2^30 - 1)*2 when mmapping < 2^31 - 1, max int32. // -1 so 2*ValueLogFileSize won't overflow on 32-bit systems. ValueLogFileSize: 1<<30 - 1, @@ -463,7 +461,7 @@ func (opt Options) WithLoggingLevel(val loggingLevel) Options { return opt } -// WithBaseTableSize returns a new Options value with MaxTableSize set to the given value. +// WithBaseTableSize returns a new Options value with BaseTableSize set to the given value. // // BaseTableSize sets the maximum size in bytes for LSM table or file in the base level. // diff --git a/vendor/github.com/dgraph-io/badger/v4/table/builder.go b/vendor/github.com/dgraph-io/badger/v4/table/builder.go index 5c9e065e0..bf0ac319a 100644 --- a/vendor/github.com/dgraph-io/badger/v4/table/builder.go +++ b/vendor/github.com/dgraph-io/badger/v4/table/builder.go @@ -25,8 +25,8 @@ import ( "unsafe" "github.com/golang/protobuf/proto" - "github.com/golang/snappy" fbs "github.com/google/flatbuffers/go" + "github.com/klauspost/compress/s2" "github.com/pkg/errors" "github.com/dgraph-io/badger/v4/fb" @@ -159,7 +159,7 @@ func NewTableBuilder(opts Options) *Builder { func maxEncodedLen(ctype options.CompressionType, sz int) int { switch ctype { case options.Snappy: - return snappy.MaxEncodedLen(sz) + return s2.MaxEncodedLen(sz) case options.ZSTD: return y.ZSTDCompressBound(sz) } @@ -523,9 +523,9 @@ func (b *Builder) compressData(data []byte) ([]byte, error) { case options.None: return data, nil case options.Snappy: - sz := snappy.MaxEncodedLen(len(data)) + sz := s2.MaxEncodedLen(len(data)) dst := b.alloc.Allocate(sz) - return snappy.Encode(dst, data), nil + return s2.EncodeSnappy(dst, data), nil case options.ZSTD: sz := y.ZSTDCompressBound(len(data)) dst := b.alloc.Allocate(sz) diff --git a/vendor/github.com/dgraph-io/badger/v4/table/iterator.go b/vendor/github.com/dgraph-io/badger/v4/table/iterator.go index 109856582..d99203e39 100644 --- a/vendor/github.com/dgraph-io/badger/v4/table/iterator.go +++ b/vendor/github.com/dgraph-io/badger/v4/table/iterator.go @@ -34,7 +34,7 @@ type blockIterator struct { key []byte val []byte entryOffsets []uint32 - block *block + block *Block tableID uint64 blockID int @@ -43,7 +43,7 @@ type blockIterator struct { prevOverlap uint16 } -func (itr *blockIterator) setBlock(b *block) { +func (itr *blockIterator) setBlock(b *Block) { // Decrement the ref for the old block. If the old block was compressed, we // might be able to reuse it. itr.block.decrRef() diff --git a/vendor/github.com/dgraph-io/badger/v4/table/table.go b/vendor/github.com/dgraph-io/badger/v4/table/table.go index 0bbc91089..432bec3f2 100644 --- a/vendor/github.com/dgraph-io/badger/v4/table/table.go +++ b/vendor/github.com/dgraph-io/badger/v4/table/table.go @@ -32,7 +32,8 @@ import ( "unsafe" "github.com/golang/protobuf/proto" - "github.com/golang/snappy" + "github.com/klauspost/compress/snappy" + "github.com/klauspost/compress/zstd" "github.com/pkg/errors" "github.com/dgraph-io/badger/v4/fb" @@ -76,8 +77,8 @@ type Options struct { Compression options.CompressionType // Block cache is used to cache decompressed and decrypted blocks. - BlockCache *ristretto.Cache - IndexCache *ristretto.Cache + BlockCache *ristretto.Cache[[]byte, *Block] + IndexCache *ristretto.Cache[uint64, *fb.TableIndex] AllocPool *z.AllocatorPool @@ -177,13 +178,11 @@ func (t *Table) DecrRef() error { } // BlockEvictHandler is used to reuse the byte slice stored in the block on cache eviction. -func BlockEvictHandler(value interface{}) { - if b, ok := value.(*block); ok { - b.decrRef() - } +func BlockEvictHandler(b *Block) { + b.decrRef() } -type block struct { +type Block struct { offset int data []byte checksum []byte @@ -198,7 +197,7 @@ var NumBlocks atomic.Int32 // incrRef increments the ref of a block and return a bool indicating if the // increment was successful. A true value indicates that the block can be used. -func (b *block) incrRef() bool { +func (b *Block) incrRef() bool { for { // We can't blindly add 1 to ref. We need to check whether it has // reached zero first, because if it did, then we should absolutely not @@ -221,7 +220,7 @@ func (b *block) incrRef() bool { } } } -func (b *block) decrRef() { +func (b *Block) decrRef() { if b == nil { return } @@ -241,12 +240,12 @@ func (b *block) decrRef() { } y.AssertTrue(b.ref.Load() >= 0) } -func (b *block) size() int64 { +func (b *Block) size() int64 { return int64(3*intSize /* Size of the offset, entriesIndexStart and chkLen */ + cap(b.data) + cap(b.checksum) + cap(b.entryOffsets)*4) } -func (b *block) verifyCheckSum() error { +func (b *Block) verifyCheckSum() error { cs := &pb.Checksum{} if err := proto.Unmarshal(b.checksum, cs); err != nil { return y.Wrapf(err, "unable to unmarshal checksum for block") @@ -520,7 +519,7 @@ func (t *Table) fetchIndex() *fb.TableIndex { panic("Index Cache must be set for encrypted workloads") } if val, ok := t.opt.IndexCache.Get(t.indexKey()); ok && val != nil { - return val.(*fb.TableIndex) + return val } index, err := t.readTableIndex() @@ -536,7 +535,7 @@ func (t *Table) offsets(ko *fb.BlockOffset, i int) bool { // block function return a new block. Each block holds a ref and the byte // slice stored in the block will be reused when the ref becomes zero. The // caller should release the block by calling block.decrRef() on it. -func (t *Table) block(idx int, useCache bool) (*block, error) { +func (t *Table) block(idx int, useCache bool) (*Block, error) { y.AssertTruef(idx >= 0, "idx=%d", idx) if idx >= t.offsetsLength() { return nil, errors.New("block out of index") @@ -548,15 +547,15 @@ func (t *Table) block(idx int, useCache bool) (*block, error) { // Use the block only if the increment was successful. The block // could get evicted from the cache between the Get() call and the // incrRef() call. - if b := blk.(*block); b.incrRef() { - return b, nil + if blk.incrRef() { + return blk, nil } } } var ko fb.BlockOffset y.AssertTrue(t.offsets(&ko, idx)) - blk := &block{offset: int(ko.Offset())} + blk := &Block{offset: int(ko.Offset())} blk.ref.Store(1) defer blk.decrRef() // Deal with any errors, where blk would not be returned. NumBlocks.Add(1) @@ -794,7 +793,7 @@ func NewFilename(id uint64, dir string) string { } // decompress decompresses the data stored in a block. -func (t *Table) decompress(b *block) error { +func (t *Table) decompress(b *Block) error { var dst []byte var err error @@ -818,6 +817,11 @@ func (t *Table) decompress(b *block) error { } case options.ZSTD: sz := int(float64(t.opt.BlockSize) * 1.2) + // Get frame content size from header. + var hdr zstd.Header + if err := hdr.Decode(b.data); err == nil && hdr.HasFCS && hdr.FrameContentSize < uint64(t.opt.BlockSize*2) { + sz = int(hdr.FrameContentSize) + } dst = z.Calloc(sz, "Table.Decompress") b.data, err = y.ZSTDDecompress(dst, b.data) if err != nil { diff --git a/vendor/github.com/dgraph-io/badger/v4/txn.go b/vendor/github.com/dgraph-io/badger/v4/txn.go index 4a5fe476c..438af8d5d 100644 --- a/vendor/github.com/dgraph-io/badger/v4/txn.go +++ b/vendor/github.com/dgraph-io/badger/v4/txn.go @@ -661,7 +661,9 @@ func (txn *Txn) Commit() error { // txn.conflictKeys can be zero if conflict detection is turned off. So we // should check txn.pendingWrites. if len(txn.pendingWrites) == 0 { - return nil // Nothing to do. + // Discard the transaction so that the read is marked done. + txn.Discard() + return nil } // Precheck before discarding txn. if err := txn.commitPrecheck(); err != nil { @@ -716,6 +718,8 @@ func (txn *Txn) CommitWith(cb func(error)) { // callback might be acquiring the same locks. Instead run the callback // from another goroutine. go runTxnCallback(&txnCb{user: cb, err: nil}) + // Discard the transaction so that the read is marked done. + txn.Discard() return } diff --git a/vendor/github.com/dgraph-io/badger/v4/y/watermark.go b/vendor/github.com/dgraph-io/badger/v4/y/watermark.go index cf2992b8b..7fc0c82c4 100644 --- a/vendor/github.com/dgraph-io/badger/v4/y/watermark.go +++ b/vendor/github.com/dgraph-io/badger/v4/y/watermark.go @@ -228,7 +228,8 @@ func (w *WaterMark) process(closer *z.Closer) { } } } else { - if mark.index > 0 { + // it is possible that mark.index is zero. We need to handle that case as well. + if mark.index > 0 || (mark.index == 0 && len(mark.indices) == 0) { processOne(mark.index, mark.done) } for _, index := range mark.indices { diff --git a/vendor/github.com/dgraph-io/ristretto/.deepsource.toml b/vendor/github.com/dgraph-io/ristretto/.deepsource.toml deleted file mode 100644 index 40609eff3..000000000 --- a/vendor/github.com/dgraph-io/ristretto/.deepsource.toml +++ /dev/null @@ -1,17 +0,0 @@ -version = 1 - -test_patterns = [ - '**/*_test.go' -] - -exclude_patterns = [ - -] - -[[analyzers]] -name = 'go' -enabled = true - - - [analyzers.meta] - import_path = 'github.com/dgraph-io/ristretto' diff --git a/vendor/github.com/dgraph-io/ristretto/.go-version b/vendor/github.com/dgraph-io/ristretto/.go-version index b8f1e3fd3..bc4493477 100644 --- a/vendor/github.com/dgraph-io/ristretto/.go-version +++ b/vendor/github.com/dgraph-io/ristretto/.go-version @@ -1 +1 @@ -1.17.11 +1.19 diff --git a/vendor/github.com/dgraph-io/ristretto/.golangci.yml b/vendor/github.com/dgraph-io/ristretto/.golangci.yml index 7318e9a3b..c342c7aab 100644 --- a/vendor/github.com/dgraph-io/ristretto/.golangci.yml +++ b/vendor/github.com/dgraph-io/ristretto/.golangci.yml @@ -1,23 +1,29 @@ run: - tests: false skip-dirs: - - contrib - - sim + skip-files: linters-settings: lll: line-length: 120 + staticcheck: + checks: + - all + - '-SA1019' # it is okay to use math/rand at times. + gosec: + excludes: + - G404 # it is okay to use math/rand at times. linters: disable-all: true enable: - #- errcheck - #- ineffassign - - gas - #- gofmt - #- golint - #- gosimple - #- govet + - errcheck + - gofmt + - goimports + - gosec + - gosimple + - govet + - ineffassign - lll - #- varcheck - #- unused + - staticcheck + - unconvert + - unused diff --git a/vendor/github.com/dgraph-io/ristretto/CHANGELOG.md b/vendor/github.com/dgraph-io/ristretto/CHANGELOG.md index 3d18e39ed..5e1ced12b 100644 --- a/vendor/github.com/dgraph-io/ristretto/CHANGELOG.md +++ b/vendor/github.com/dgraph-io/ristretto/CHANGELOG.md @@ -172,7 +172,7 @@ improve performance and reduce memory requirements. ### Fixed - Fix the way metrics are handled for deletions. ([#111][]) -- Support nil `*Cache` values in `Clear` and `Close`. ([#119][]) +- Support nil `*Cache` values in `Clear` and `Close`. ([#119][]) - Delete item immediately. ([#113][]) - Remove key from policy after TTL eviction. ([#130][]) diff --git a/vendor/github.com/dgraph-io/ristretto/README.md b/vendor/github.com/dgraph-io/ristretto/README.md index e71ae3df9..8793ad0d4 100644 --- a/vendor/github.com/dgraph-io/ristretto/README.md +++ b/vendor/github.com/dgraph-io/ristretto/README.md @@ -7,8 +7,7 @@ Ristretto is a fast, concurrent cache library built with a focus on performance and correctness. -The motivation to build Ristretto comes from the need for a contention-free -cache in [Dgraph][]. +The motivation to build Ristretto comes from the need for a contention-free cache in [Dgraph][]. [Dgraph]: https://github.com/dgraph-io/dgraph @@ -19,7 +18,7 @@ cache in [Dgraph][]. * **Admission: TinyLFU** - extra performance with little memory overhead (12 bits per counter). * **Fast Throughput** - we use a variety of techniques for managing contention and the result is excellent throughput. * **Cost-Based Eviction** - any large new item deemed valuable can evict multiple smaller items (cost could be anything). -* **Fully Concurrent** - you can use as many goroutines as you want with little throughput degradation. +* **Fully Concurrent** - you can use as many goroutines as you want with little throughput degradation. * **Metrics** - optional performance metrics for throughput, hit ratios, and other stats. * **Simple API** - just figure out your ideal `Config` values and you're off and running. @@ -29,34 +28,41 @@ Ristretto is production-ready. See [Projects using Ristretto](#projects-using-ri ## Table of Contents -* [Usage](#Usage) - * [Example](#Example) - * [Config](#Config) - * [NumCounters](#Config) - * [MaxCost](#Config) - * [BufferItems](#Config) - * [Metrics](#Config) - * [OnEvict](#Config) - * [KeyToHash](#Config) - * [Cost](#Config) -* [Benchmarks](#Benchmarks) - * [Hit Ratios](#Hit-Ratios) - * [Search](#Search) - * [Database](#Database) - * [Looping](#Looping) - * [CODASYL](#CODASYL) - * [Throughput](#Throughput) - * [Mixed](#Mixed) - * [Read](#Read) - * [Write](#Write) -* [Projects using Ristretto](#projects-using-ristretto) -* [FAQ](#FAQ) +- [Ristretto](#ristretto) + - [Features](#features) + - [Status](#status) + - [Table of Contents](#table-of-contents) + - [Usage](#usage) + - [Example](#example) + - [Config](#config) + - [Benchmarks](#benchmarks) + - [Hit Ratios](#hit-ratios) + - [Search](#search) + - [Database](#database) + - [Looping](#looping) + - [CODASYL](#codasyl) + - [Throughput](#throughput) + - [Mixed](#mixed) + - [Read](#read) + - [Write](#write) + - [Projects Using Ristretto](#projects-using-ristretto) + - [FAQ](#faq) + - [How are you achieving this performance? What shortcuts are you taking?](#how-are-you-achieving-this-performance-what-shortcuts-are-you-taking) + - [Is Ristretto distributed?](#is-ristretto-distributed) ## Usage ### Example ```go +package main + +import ( + "fmt" + + "github.com/dgraph-io/ristretto" +) + func main() { cache, err := ristretto.NewCache(&ristretto.Config{ NumCounters: 1e7, // number of keys to track frequency of (10M). @@ -69,46 +75,49 @@ func main() { // set a value with a cost of 1 cache.Set("key", "value", 1) - + // wait for value to pass through buffers - time.Sleep(10 * time.Millisecond) + cache.Wait() + // get value from cache value, found := cache.Get("key") if !found { panic("missing value") } fmt.Println(value) + + // del value from cache cache.Del("key") } ``` ### Config -The `Config` struct is passed to `NewCache` when creating Ristretto instances (see the example above). +The `Config` struct is passed to `NewCache` when creating Ristretto instances (see the example above). **NumCounters** `int64` -NumCounters is the number of 4-bit access counters to keep for admission and eviction. We've seen good performance in setting this to 10x the number of items you expect to keep in the cache when full. +NumCounters is the number of 4-bit access counters to keep for admission and eviction. We've seen good performance in setting this to 10x the number of items you expect to keep in the cache when full. -For example, if you expect each item to have a cost of 1 and MaxCost is 100, set NumCounters to 1,000. Or, if you use variable cost values but expect the cache to hold around 10,000 items when full, set NumCounters to 100,000. The important thing is the *number of unique items* in the full cache, not necessarily the MaxCost value. +For example, if you expect each item to have a cost of 1 and MaxCost is 100, set NumCounters to 1,000. Or, if you use variable cost values but expect the cache to hold around 10,000 items when full, set NumCounters to 100,000. The important thing is the *number of unique items* in the full cache, not necessarily the MaxCost value. **MaxCost** `int64` -MaxCost is how eviction decisions are made. For example, if MaxCost is 100 and a new item with a cost of 1 increases total cache cost to 101, 1 item will be evicted. +MaxCost is how eviction decisions are made. For example, if MaxCost is 100 and a new item with a cost of 1 increases total cache cost to 101, 1 item will be evicted. -MaxCost can also be used to denote the max size in bytes. For example, if MaxCost is 1,000,000 (1MB) and the cache is full with 1,000 1KB items, a new item (that's accepted) would cause 5 1KB items to be evicted. +MaxCost can also be used to denote the max size in bytes. For example, if MaxCost is 1,000,000 (1MB) and the cache is full with 1,000 1KB items, a new item (that's accepted) would cause 5 1KB items to be evicted. -MaxCost could be anything as long as it matches how you're using the cost values when calling Set. +MaxCost could be anything as long as it matches how you're using the cost values when calling Set. **BufferItems** `int64` -BufferItems is the size of the Get buffers. The best value we've found for this is 64. +BufferItems is the size of the Get buffers. The best value we've found for this is 64. If for some reason you see Get performance decreasing with lots of contention (you shouldn't), try increasing this value in increments of 64. This is a fine-tuning mechanism and you probably won't have to touch this. **Metrics** `bool` -Metrics is true when you want real-time logging of a variety of stats. The reason this is a Config flag is because there's a 10% throughput performance overhead. +Metrics is true when you want real-time logging of a variety of stats. The reason this is a Config flag is because there's a 10% throughput performance overhead. **OnEvict** `func(hashes [2]uint64, value interface{}, cost int64)` @@ -213,8 +222,8 @@ Below is a list of known projects that use Ristretto: We go into detail in the [Ristretto blog post](https://blog.dgraph.io/post/introducing-ristretto-high-perf-go-cache/), but in short: our throughput performance can be attributed to a mix of batching and eventual consistency. Our hit ratio performance is mostly due to an excellent [admission policy](https://arxiv.org/abs/1512.00727) and SampledLFU eviction policy. -As for "shortcuts," the only thing Ristretto does that could be construed as one is dropping some Set calls. That means a Set call for a new item (updates are guaranteed) isn't guaranteed to make it into the cache. The new item could be dropped at two points: when passing through the Set buffer or when passing through the admission policy. However, this doesn't affect hit ratios much at all as we expect the most popular items to be Set multiple times and eventually make it in the cache. +As for "shortcuts," the only thing Ristretto does that could be construed as one is dropping some Set calls. That means a Set call for a new item (updates are guaranteed) isn't guaranteed to make it into the cache. The new item could be dropped at two points: when passing through the Set buffer or when passing through the admission policy. However, this doesn't affect hit ratios much at all as we expect the most popular items to be Set multiple times and eventually make it in the cache. ### Is Ristretto distributed? -No, it's just like any other Go library that you can import into your project and use in a single process. +No, it's just like any other Go library that you can import into your project and use in a single process. diff --git a/vendor/github.com/dgraph-io/ristretto/cache.go b/vendor/github.com/dgraph-io/ristretto/cache.go index 7226245bc..398ee1413 100644 --- a/vendor/github.com/dgraph-io/ristretto/cache.go +++ b/vendor/github.com/dgraph-io/ristretto/cache.go @@ -36,40 +36,45 @@ var ( setBufSize = 32 * 1024 ) -type itemCallback func(*Item) +const itemSize = int64(unsafe.Sizeof(storeItem[any]{})) -const itemSize = int64(unsafe.Sizeof(storeItem{})) +func zeroValue[T any]() T { + var zero T + return zero +} + +type Key = z.Key // Cache is a thread-safe implementation of a hashmap with a TinyLFU admission // policy and a Sampled LFU eviction policy. You can use the same Cache instance // from as many goroutines as you want. -type Cache struct { - // store is the central concurrent hashmap where key-value items are stored. - store store - // policy determines what gets let in to the cache and what gets kicked out. - policy policy +type Cache[K Key, V any] struct { + // storedItems is the central concurrent hashmap where key-value items are stored. + storedItems store[V] + // cachePolicy determines what gets let in to the cache and what gets kicked out. + cachePolicy policy[V] // getBuf is a custom ring buffer implementation that gets pushed to when // keys are read. getBuf *ringBuffer // setBuf is a buffer allowing us to batch/drop Sets during times of high // contention. - setBuf chan *Item + setBuf chan *Item[V] // onEvict is called for item evictions. - onEvict itemCallback + onEvict func(*Item[V]) // onReject is called when an item is rejected via admission policy. - onReject itemCallback + onReject func(*Item[V]) // onExit is called whenever a value goes out of scope from the cache. - onExit (func(interface{})) + onExit (func(V)) // KeyToHash function is used to customize the key hashing algorithm. // Each key will be hashed using the provided function. If keyToHash value // is not set, the default keyToHash function is used. - keyToHash func(interface{}) (uint64, uint64) + keyToHash func(K) (uint64, uint64) // stop is used to stop the processItems goroutine. stop chan struct{} // indicates whether cache is closed. isClosed bool // cost calculates cost from a value. - cost func(value interface{}) int64 + cost func(value V) int64 // ignoreInternalCost dictates whether to ignore the cost of internally storing // the item in the cost calculation. ignoreInternalCost bool @@ -81,7 +86,7 @@ type Cache struct { } // Config is passed to NewCache for creating new Cache instances. -type Config struct { +type Config[K Key, V any] struct { // NumCounters determines the number of counters (keys) to keep that hold // access frequency information. It's generally a good idea to have more // counters than the max cache capacity, as this will improve eviction @@ -115,26 +120,28 @@ type Config struct { Metrics bool // OnEvict is called for every eviction and passes the hashed key, value, // and cost to the function. - OnEvict func(item *Item) + OnEvict func(item *Item[V]) // OnReject is called for every rejection done via the policy. - OnReject func(item *Item) + OnReject func(item *Item[V]) // OnExit is called whenever a value is removed from cache. This can be // used to do manual memory deallocation. Would also be called on eviction // and rejection of the value. - OnExit func(val interface{}) + OnExit func(val V) // KeyToHash function is used to customize the key hashing algorithm. // Each key will be hashed using the provided function. If keyToHash value // is not set, the default keyToHash function is used. - KeyToHash func(key interface{}) (uint64, uint64) + KeyToHash func(key K) (uint64, uint64) // Cost evaluates a value and outputs a corresponding cost. This function // is ran after Set is called for a new item or an item update with a cost // param of 0. - Cost func(value interface{}) int64 + Cost func(value V) int64 // IgnoreInternalCost set to true indicates to the cache that the cost of // internally storing the value should be ignored. This is useful when the // cost passed to set is not using bytes as units. Keep in mind that setting // this to true will increase the memory usage. IgnoreInternalCost bool + // TtlTickerDurationInSec set the value of time ticker for cleanup keys on ttl + TtlTickerDurationInSec int64 } type itemFlag byte @@ -146,18 +153,18 @@ const ( ) // Item is passed to setBuf so items can eventually be added to the cache. -type Item struct { +type Item[V any] struct { flag itemFlag Key uint64 Conflict uint64 - Value interface{} + Value V Cost int64 Expiration time.Time wg *sync.WaitGroup } // NewCache returns a new Cache instance and any configuration errors, if any. -func NewCache(config *Config) (*Cache, error) { +func NewCache[K Key, V any](config *Config[K, V]) (*Cache[K, V], error) { switch { case config.NumCounters == 0: return nil, errors.New("NumCounters can't be zero") @@ -165,39 +172,42 @@ func NewCache(config *Config) (*Cache, error) { return nil, errors.New("MaxCost can't be zero") case config.BufferItems == 0: return nil, errors.New("BufferItems can't be zero") + case config.TtlTickerDurationInSec == 0: + config.TtlTickerDurationInSec = bucketDurationSecs } - policy := newPolicy(config.NumCounters, config.MaxCost) - cache := &Cache{ - store: newStore(), - policy: policy, + policy := newPolicy[V](config.NumCounters, config.MaxCost) + cache := &Cache[K, V]{ + storedItems: newStore[V](), + cachePolicy: policy, getBuf: newRingBuffer(policy, config.BufferItems), - setBuf: make(chan *Item, setBufSize), + setBuf: make(chan *Item[V], setBufSize), keyToHash: config.KeyToHash, stop: make(chan struct{}), cost: config.Cost, ignoreInternalCost: config.IgnoreInternalCost, - cleanupTicker: time.NewTicker(time.Duration(bucketDurationSecs) * time.Second / 2), + cleanupTicker: time.NewTicker(time.Duration(config.TtlTickerDurationInSec) * time.Second / 2), } - cache.onExit = func(val interface{}) { - if config.OnExit != nil && val != nil { + cache.onExit = func(val V) { + if config.OnExit != nil { config.OnExit(val) } } - cache.onEvict = func(item *Item) { + cache.onEvict = func(item *Item[V]) { if config.OnEvict != nil { config.OnEvict(item) } cache.onExit(item.Value) } - cache.onReject = func(item *Item) { + cache.onReject = func(item *Item[V]) { if config.OnReject != nil { config.OnReject(item) } cache.onExit(item.Value) } if cache.keyToHash == nil { - cache.keyToHash = z.KeyToHash + cache.keyToHash = z.KeyToHash[K] } + if config.Metrics { cache.collectMetrics() } @@ -208,26 +218,29 @@ func NewCache(config *Config) (*Cache, error) { return cache, nil } -func (c *Cache) Wait() { +// Wait blocks until all buffered writes have been applied. This ensures a call to Set() +// will be visible to future calls to Get(). +func (c *Cache[K, V]) Wait() { if c == nil || c.isClosed { return } wg := &sync.WaitGroup{} wg.Add(1) - c.setBuf <- &Item{wg: wg} + c.setBuf <- &Item[V]{wg: wg} wg.Wait() } // Get returns the value (if any) and a boolean representing whether the // value was found or not. The value can be nil and the boolean can be true at -// the same time. -func (c *Cache) Get(key interface{}) (interface{}, bool) { - if c == nil || c.isClosed || key == nil { - return nil, false +// the same time. Get will not return expired items. +func (c *Cache[K, V]) Get(key K) (V, bool) { + if c == nil || c.isClosed { + return zeroValue[V](), false } keyHash, conflictHash := c.keyToHash(key) + c.getBuf.Push(keyHash) - value, ok := c.store.Get(keyHash, conflictHash) + value, ok := c.storedItems.Get(keyHash, conflictHash) if ok { c.Metrics.add(hit, keyHash, 1) } else { @@ -245,7 +258,7 @@ func (c *Cache) Get(key interface{}) (interface{}, bool) { // To dynamically evaluate the items cost using the Config.Coster function, set // the cost parameter to 0 and Coster will be ran when needed in order to find // the items true cost. -func (c *Cache) Set(key, value interface{}, cost int64) bool { +func (c *Cache[K, V]) Set(key K, value V, cost int64) bool { return c.SetWithTTL(key, value, cost, 0*time.Second) } @@ -253,8 +266,8 @@ func (c *Cache) Set(key, value interface{}, cost int64) bool { // after the specified TTL (time to live) has passed. A zero value means the value never // expires, which is identical to calling Set. A negative value is a no-op and the value // is discarded. -func (c *Cache) SetWithTTL(key, value interface{}, cost int64, ttl time.Duration) bool { - if c == nil || c.isClosed || key == nil { +func (c *Cache[K, V]) SetWithTTL(key K, value V, cost int64, ttl time.Duration) bool { + if c == nil || c.isClosed { return false } @@ -264,14 +277,14 @@ func (c *Cache) SetWithTTL(key, value interface{}, cost int64, ttl time.Duration // No expiration. break case ttl < 0: - // Treat this a a no-op. + // Treat this a no-op. return false default: expiration = time.Now().Add(ttl) } keyHash, conflictHash := c.keyToHash(key) - i := &Item{ + i := &Item[V]{ flag: itemNew, Key: keyHash, Conflict: conflictHash, @@ -281,18 +294,18 @@ func (c *Cache) SetWithTTL(key, value interface{}, cost int64, ttl time.Duration } // cost is eventually updated. The expiration must also be immediately updated // to prevent items from being prematurely removed from the map. - if prev, ok := c.store.Update(i); ok { + if prev, ok := c.storedItems.Update(i); ok { c.onExit(prev) i.flag = itemUpdate } - // Attempt to send item to policy. + // Attempt to send item to cachePolicy. select { case c.setBuf <- i: return true default: if i.flag == itemUpdate { // Return true if this was an update operation since we've already - // updated the store. For all the other operations (set/delete), we + // updated the storedItems. For all the other operations (set/delete), we // return false which means the item was not inserted. return true } @@ -302,19 +315,19 @@ func (c *Cache) SetWithTTL(key, value interface{}, cost int64, ttl time.Duration } // Del deletes the key-value item from the cache if it exists. -func (c *Cache) Del(key interface{}) { - if c == nil || c.isClosed || key == nil { +func (c *Cache[K, V]) Del(key K) { + if c == nil || c.isClosed { return } keyHash, conflictHash := c.keyToHash(key) // Delete immediately. - _, prev := c.store.Del(keyHash, conflictHash) + _, prev := c.storedItems.Del(keyHash, conflictHash) c.onExit(prev) // If we've set an item, it would be applied slightly later. // So we must push the same item to `setBuf` with the deletion flag. // This ensures that if a set is followed by a delete, it will be // applied in the correct order. - c.setBuf <- &Item{ + c.setBuf <- &Item[V]{ flag: itemDelete, Key: keyHash, Conflict: conflictHash, @@ -323,18 +336,18 @@ func (c *Cache) Del(key interface{}) { // GetTTL returns the TTL for the specified key and a bool that is true if the // item was found and is not expired. -func (c *Cache) GetTTL(key interface{}) (time.Duration, bool) { - if c == nil || key == nil { +func (c *Cache[K, V]) GetTTL(key K) (time.Duration, bool) { + if c == nil { return 0, false } keyHash, conflictHash := c.keyToHash(key) - if _, ok := c.store.Get(keyHash, conflictHash); !ok { + if _, ok := c.storedItems.Get(keyHash, conflictHash); !ok { // not found return 0, false } - expiration := c.store.Expiration(keyHash) + expiration := c.storedItems.Expiration(keyHash) if expiration.IsZero() { // found but no expiration return 0, true @@ -349,7 +362,7 @@ func (c *Cache) GetTTL(key interface{}) (time.Duration, bool) { } // Close stops all goroutines and closes all channels. -func (c *Cache) Close() { +func (c *Cache[K, V]) Close() { if c == nil || c.isClosed { return } @@ -359,14 +372,15 @@ func (c *Cache) Close() { c.stop <- struct{}{} close(c.stop) close(c.setBuf) - c.policy.Close() + c.cachePolicy.Close() + c.cleanupTicker.Stop() c.isClosed = true } -// Clear empties the hashmap and zeroes all policy counters. Note that this is +// Clear empties the hashmap and zeroes all cachePolicy counters. Note that this is // not an atomic operation (but that shouldn't be a problem as it's assumed that // Set/Get calls won't be occurring until after this). -func (c *Cache) Clear() { +func (c *Cache[K, V]) Clear() { if c == nil || c.isClosed { return } @@ -383,7 +397,7 @@ loop: continue } if i.flag != itemUpdate { - // In itemUpdate, the value is already set in the store. So, no need to call + // In itemUpdate, the value is already set in the storedItems. So, no need to call // onEvict here. c.onEvict(i) } @@ -392,9 +406,9 @@ loop: } } - // Clear value hashmap and policy data. - c.policy.Clear() - c.store.Clear(c.onEvict) + // Clear value hashmap and cachePolicy data. + c.cachePolicy.Clear() + c.storedItems.Clear(c.onEvict) // Only reset metrics if they're enabled. if c.Metrics != nil { c.Metrics.Clear() @@ -404,23 +418,23 @@ loop: } // MaxCost returns the max cost of the cache. -func (c *Cache) MaxCost() int64 { +func (c *Cache[K, V]) MaxCost() int64 { if c == nil { return 0 } - return c.policy.MaxCost() + return c.cachePolicy.MaxCost() } // UpdateMaxCost updates the maxCost of an existing cache. -func (c *Cache) UpdateMaxCost(maxCost int64) { +func (c *Cache[K, V]) UpdateMaxCost(maxCost int64) { if c == nil { return } - c.policy.UpdateMaxCost(maxCost) + c.cachePolicy.UpdateMaxCost(maxCost) } // processItems is ran by goroutines processing the Set buffer. -func (c *Cache) processItems() { +func (c *Cache[K, V]) processItems() { startTs := make(map[uint64]time.Time) numToKeep := 100000 // TODO: Make this configurable via options. @@ -438,7 +452,7 @@ func (c *Cache) processItems() { } } } - onEvict := func(i *Item) { + onEvict := func(i *Item[V]) { if ts, has := startTs[i.Key]; has { c.Metrics.trackEviction(int64(time.Since(ts) / time.Second)) delete(startTs, i.Key) @@ -466,29 +480,29 @@ func (c *Cache) processItems() { switch i.flag { case itemNew: - victims, added := c.policy.Add(i.Key, i.Cost) + victims, added := c.cachePolicy.Add(i.Key, i.Cost) if added { - c.store.Set(i) + c.storedItems.Set(i) c.Metrics.add(keyAdd, i.Key, 1) trackAdmission(i.Key) } else { c.onReject(i) } for _, victim := range victims { - victim.Conflict, victim.Value = c.store.Del(victim.Key, 0) + victim.Conflict, victim.Value = c.storedItems.Del(victim.Key, 0) onEvict(victim) } case itemUpdate: - c.policy.Update(i.Key, i.Cost) + c.cachePolicy.Update(i.Key, i.Cost) case itemDelete: - c.policy.Del(i.Key) // Deals with metrics updates. - _, val := c.store.Del(i.Key, i.Conflict) + c.cachePolicy.Del(i.Key) // Deals with metrics updates. + _, val := c.storedItems.Del(i.Key, i.Conflict) c.onExit(val) } case <-c.cleanupTicker.C: - c.store.Cleanup(c.policy, onEvict) + c.storedItems.Cleanup(c.cachePolicy, onEvict) case <-c.stop: return } @@ -497,9 +511,9 @@ func (c *Cache) processItems() { // collectMetrics just creates a new *Metrics instance and adds the pointers // to the cache and policy instances. -func (c *Cache) collectMetrics() { +func (c *Cache[K, V]) collectMetrics() { c.Metrics = newMetrics() - c.policy.CollectMetrics(c.Metrics) + c.cachePolicy.CollectMetrics(c.Metrics) } type metricType int diff --git a/vendor/github.com/dgraph-io/ristretto/policy.go b/vendor/github.com/dgraph-io/ristretto/policy.go index bf23f91fd..290488d87 100644 --- a/vendor/github.com/dgraph-io/ristretto/policy.go +++ b/vendor/github.com/dgraph-io/ristretto/policy.go @@ -31,15 +31,14 @@ const ( ) // policy is the interface encapsulating eviction/admission behavior. -// // TODO: remove this interface and just rename defaultPolicy to policy, as we -// are probably only going to use/implement/maintain one policy. -type policy interface { +// are probably only going to use/implement/maintain one policy. +type policy[V any] interface { ringConsumer // Add attempts to Add the key-cost pair to the Policy. It returns a slice // of evicted keys and a bool denoting whether or not the key-cost pair // was added. If it returns true, the key should be stored in cache. - Add(uint64, int64) ([]*Item, bool) + Add(uint64, int64) ([]*Item[V], bool) // Has returns true if the key exists in the Policy. Has(uint64) bool // Del deletes the key from the Policy. @@ -62,11 +61,11 @@ type policy interface { UpdateMaxCost(int64) } -func newPolicy(numCounters, maxCost int64) policy { - return newDefaultPolicy(numCounters, maxCost) +func newPolicy[V any](numCounters, maxCost int64) policy[V] { + return newDefaultPolicy[V](numCounters, maxCost) } -type defaultPolicy struct { +type defaultPolicy[V any] struct { sync.Mutex admit *tinyLFU evict *sampledLFU @@ -76,8 +75,8 @@ type defaultPolicy struct { metrics *Metrics } -func newDefaultPolicy(numCounters, maxCost int64) *defaultPolicy { - p := &defaultPolicy{ +func newDefaultPolicy[V any](numCounters, maxCost int64) *defaultPolicy[V] { + p := &defaultPolicy[V]{ admit: newTinyLFU(numCounters), evict: newSampledLFU(maxCost), itemsCh: make(chan []uint64, 3), @@ -87,7 +86,7 @@ func newDefaultPolicy(numCounters, maxCost int64) *defaultPolicy { return p } -func (p *defaultPolicy) CollectMetrics(metrics *Metrics) { +func (p *defaultPolicy[V]) CollectMetrics(metrics *Metrics) { p.metrics = metrics p.evict.metrics = metrics } @@ -97,7 +96,7 @@ type policyPair struct { cost int64 } -func (p *defaultPolicy) processItems() { +func (p *defaultPolicy[V]) processItems() { for { select { case items := <-p.itemsCh: @@ -110,7 +109,7 @@ func (p *defaultPolicy) processItems() { } } -func (p *defaultPolicy) Push(keys []uint64) bool { +func (p *defaultPolicy[V]) Push(keys []uint64) bool { if p.isClosed { return false } @@ -132,7 +131,7 @@ func (p *defaultPolicy) Push(keys []uint64) bool { // Add decides whether the item with the given key and cost should be accepted by // the policy. It returns the list of victims that have been evicted and a boolean // indicating whether the incoming item should be accepted. -func (p *defaultPolicy) Add(key uint64, cost int64) ([]*Item, bool) { +func (p *defaultPolicy[V]) Add(key uint64, cost int64) ([]*Item[V], bool) { p.Lock() defer p.Unlock() @@ -166,7 +165,7 @@ func (p *defaultPolicy) Add(key uint64, cost int64) ([]*Item, bool) { // O(lg N). sample := make([]*policyPair, 0, lfuSample) // As items are evicted they will be appended to victims. - victims := make([]*Item, 0) + victims := make([]*Item[V], 0) // Delete victims until there's enough space or a minKey is found that has // more hits than incoming item. @@ -196,7 +195,7 @@ func (p *defaultPolicy) Add(key uint64, cost int64) ([]*Item, bool) { sample[minId] = sample[len(sample)-1] sample = sample[:len(sample)-1] // Store victim in evicted victims slice. - victims = append(victims, &Item{ + victims = append(victims, &Item[V]{ Key: minKey, Conflict: 0, Cost: minCost, @@ -208,33 +207,33 @@ func (p *defaultPolicy) Add(key uint64, cost int64) ([]*Item, bool) { return victims, true } -func (p *defaultPolicy) Has(key uint64) bool { +func (p *defaultPolicy[V]) Has(key uint64) bool { p.Lock() _, exists := p.evict.keyCosts[key] p.Unlock() return exists } -func (p *defaultPolicy) Del(key uint64) { +func (p *defaultPolicy[V]) Del(key uint64) { p.Lock() p.evict.del(key) p.Unlock() } -func (p *defaultPolicy) Cap() int64 { +func (p *defaultPolicy[V]) Cap() int64 { p.Lock() - capacity := int64(p.evict.getMaxCost() - p.evict.used) + capacity := p.evict.getMaxCost() - p.evict.used p.Unlock() return capacity } -func (p *defaultPolicy) Update(key uint64, cost int64) { +func (p *defaultPolicy[V]) Update(key uint64, cost int64) { p.Lock() p.evict.updateIfHas(key, cost) p.Unlock() } -func (p *defaultPolicy) Cost(key uint64) int64 { +func (p *defaultPolicy[V]) Cost(key uint64) int64 { p.Lock() if cost, found := p.evict.keyCosts[key]; found { p.Unlock() @@ -244,14 +243,14 @@ func (p *defaultPolicy) Cost(key uint64) int64 { return -1 } -func (p *defaultPolicy) Clear() { +func (p *defaultPolicy[V]) Clear() { p.Lock() p.admit.clear() p.evict.clear() p.Unlock() } -func (p *defaultPolicy) Close() { +func (p *defaultPolicy[V]) Close() { if p.isClosed { return } @@ -263,14 +262,14 @@ func (p *defaultPolicy) Close() { p.isClosed = true } -func (p *defaultPolicy) MaxCost() int64 { +func (p *defaultPolicy[V]) MaxCost() int64 { if p == nil || p.evict == nil { return 0 } return p.evict.getMaxCost() } -func (p *defaultPolicy) UpdateMaxCost(maxCost int64) { +func (p *defaultPolicy[V]) UpdateMaxCost(maxCost int64) { if p == nil || p.evict == nil { return } @@ -346,7 +345,7 @@ func (p *sampledLFU) updateIfHas(key uint64, cost int64) bool { p.metrics.add(keyUpdate, key, 1) if prev > cost { diff := prev - cost - p.metrics.add(costAdd, key, ^uint64(uint64(diff)-1)) + p.metrics.add(costAdd, key, ^(uint64(diff) - 1)) } else if cost > prev { diff := cost - prev p.metrics.add(costAdd, key, uint64(diff)) diff --git a/vendor/github.com/dgraph-io/ristretto/sketch.go b/vendor/github.com/dgraph-io/ristretto/sketch.go index 6368d2bde..196bdfdc4 100644 --- a/vendor/github.com/dgraph-io/ristretto/sketch.go +++ b/vendor/github.com/dgraph-io/ristretto/sketch.go @@ -103,7 +103,7 @@ func newCmRow(numCounters int64) cmRow { } func (r cmRow) get(n uint64) byte { - return byte(r[n/2]>>((n&1)*4)) & 0x0f + return (r[n/2] >> ((n & 1) * 4)) & 0x0f } func (r cmRow) increment(n uint64) { diff --git a/vendor/github.com/dgraph-io/ristretto/store.go b/vendor/github.com/dgraph-io/ristretto/store.go index e42a98b78..3a53d593f 100644 --- a/vendor/github.com/dgraph-io/ristretto/store.go +++ b/vendor/github.com/dgraph-io/ristretto/store.go @@ -22,10 +22,10 @@ import ( ) // TODO: Do we need this to be a separate struct from Item? -type storeItem struct { +type storeItem[V any] struct { key uint64 conflict uint64 - value interface{} + value V expiration time.Time } @@ -35,58 +35,58 @@ type storeItem struct { // in Ristretto. // // Every store is safe for concurrent usage. -type store interface { +type store[V any] interface { // Get returns the value associated with the key parameter. - Get(uint64, uint64) (interface{}, bool) + Get(uint64, uint64) (V, bool) // Expiration returns the expiration time for this key. Expiration(uint64) time.Time // Set adds the key-value pair to the Map or updates the value if it's // already present. The key-value pair is passed as a pointer to an // item object. - Set(*Item) + Set(*Item[V]) // Del deletes the key-value pair from the Map. - Del(uint64, uint64) (uint64, interface{}) + Del(uint64, uint64) (uint64, V) // Update attempts to update the key with a new value and returns true if // successful. - Update(*Item) (interface{}, bool) + Update(*Item[V]) (V, bool) // Cleanup removes items that have an expired TTL. - Cleanup(policy policy, onEvict itemCallback) + Cleanup(policy policy[V], onEvict func(item *Item[V])) // Clear clears all contents of the store. - Clear(onEvict itemCallback) + Clear(onEvict func(item *Item[V])) } // newStore returns the default store implementation. -func newStore() store { - return newShardedMap() +func newStore[V any]() store[V] { + return newShardedMap[V]() } const numShards uint64 = 256 -type shardedMap struct { - shards []*lockedMap - expiryMap *expirationMap +type shardedMap[V any] struct { + shards []*lockedMap[V] + expiryMap *expirationMap[V] } -func newShardedMap() *shardedMap { - sm := &shardedMap{ - shards: make([]*lockedMap, int(numShards)), - expiryMap: newExpirationMap(), +func newShardedMap[V any]() *shardedMap[V] { + sm := &shardedMap[V]{ + shards: make([]*lockedMap[V], int(numShards)), + expiryMap: newExpirationMap[V](), } for i := range sm.shards { - sm.shards[i] = newLockedMap(sm.expiryMap) + sm.shards[i] = newLockedMap[V](sm.expiryMap) } return sm } -func (sm *shardedMap) Get(key, conflict uint64) (interface{}, bool) { +func (sm *shardedMap[V]) Get(key, conflict uint64) (V, bool) { return sm.shards[key%numShards].get(key, conflict) } -func (sm *shardedMap) Expiration(key uint64) time.Time { +func (sm *shardedMap[V]) Expiration(key uint64) time.Time { return sm.shards[key%numShards].Expiration(key) } -func (sm *shardedMap) Set(i *Item) { +func (sm *shardedMap[V]) Set(i *Item[V]) { if i == nil { // If item is nil make this Set a no-op. return @@ -95,62 +95,62 @@ func (sm *shardedMap) Set(i *Item) { sm.shards[i.Key%numShards].Set(i) } -func (sm *shardedMap) Del(key, conflict uint64) (uint64, interface{}) { +func (sm *shardedMap[V]) Del(key, conflict uint64) (uint64, V) { return sm.shards[key%numShards].Del(key, conflict) } -func (sm *shardedMap) Update(newItem *Item) (interface{}, bool) { +func (sm *shardedMap[V]) Update(newItem *Item[V]) (V, bool) { return sm.shards[newItem.Key%numShards].Update(newItem) } -func (sm *shardedMap) Cleanup(policy policy, onEvict itemCallback) { +func (sm *shardedMap[V]) Cleanup(policy policy[V], onEvict func(item *Item[V])) { sm.expiryMap.cleanup(sm, policy, onEvict) } -func (sm *shardedMap) Clear(onEvict itemCallback) { +func (sm *shardedMap[V]) Clear(onEvict func(item *Item[V])) { for i := uint64(0); i < numShards; i++ { sm.shards[i].Clear(onEvict) } } -type lockedMap struct { +type lockedMap[V any] struct { sync.RWMutex - data map[uint64]storeItem - em *expirationMap + data map[uint64]storeItem[V] + em *expirationMap[V] } -func newLockedMap(em *expirationMap) *lockedMap { - return &lockedMap{ - data: make(map[uint64]storeItem), +func newLockedMap[V any](em *expirationMap[V]) *lockedMap[V] { + return &lockedMap[V]{ + data: make(map[uint64]storeItem[V]), em: em, } } -func (m *lockedMap) get(key, conflict uint64) (interface{}, bool) { +func (m *lockedMap[V]) get(key, conflict uint64) (V, bool) { m.RLock() item, ok := m.data[key] m.RUnlock() if !ok { - return nil, false + return zeroValue[V](), false } if conflict != 0 && (conflict != item.conflict) { - return nil, false + return zeroValue[V](), false } // Handle expired items. if !item.expiration.IsZero() && time.Now().After(item.expiration) { - return nil, false + return zeroValue[V](), false } return item.value, true } -func (m *lockedMap) Expiration(key uint64) time.Time { +func (m *lockedMap[V]) Expiration(key uint64) time.Time { m.RLock() defer m.RUnlock() return m.data[key].expiration } -func (m *lockedMap) Set(i *Item) { +func (m *lockedMap[V]) Set(i *Item[V]) { if i == nil { // If the item is nil make this Set a no-op. return @@ -173,7 +173,7 @@ func (m *lockedMap) Set(i *Item) { m.em.add(i.Key, i.Conflict, i.Expiration) } - m.data[i.Key] = storeItem{ + m.data[i.Key] = storeItem[V]{ key: i.Key, conflict: i.Conflict, value: i.Value, @@ -181,16 +181,16 @@ func (m *lockedMap) Set(i *Item) { } } -func (m *lockedMap) Del(key, conflict uint64) (uint64, interface{}) { +func (m *lockedMap[V]) Del(key, conflict uint64) (uint64, V) { m.Lock() item, ok := m.data[key] if !ok { m.Unlock() - return 0, nil + return 0, zeroValue[V]() } if conflict != 0 && (conflict != item.conflict) { m.Unlock() - return 0, nil + return 0, zeroValue[V]() } if !item.expiration.IsZero() { @@ -202,20 +202,20 @@ func (m *lockedMap) Del(key, conflict uint64) (uint64, interface{}) { return item.conflict, item.value } -func (m *lockedMap) Update(newItem *Item) (interface{}, bool) { +func (m *lockedMap[V]) Update(newItem *Item[V]) (V, bool) { m.Lock() item, ok := m.data[newItem.Key] if !ok { m.Unlock() - return nil, false + return zeroValue[V](), false } if newItem.Conflict != 0 && (newItem.Conflict != item.conflict) { m.Unlock() - return nil, false + return zeroValue[V](), false } m.em.update(newItem.Key, newItem.Conflict, item.expiration, newItem.Expiration) - m.data[newItem.Key] = storeItem{ + m.data[newItem.Key] = storeItem[V]{ key: newItem.Key, conflict: newItem.Conflict, value: newItem.Value, @@ -226,9 +226,9 @@ func (m *lockedMap) Update(newItem *Item) (interface{}, bool) { return item.value, true } -func (m *lockedMap) Clear(onEvict itemCallback) { +func (m *lockedMap[V]) Clear(onEvict func(item *Item[V])) { m.Lock() - i := &Item{} + i := &Item[V]{} if onEvict != nil { for _, si := range m.data { i.Key = si.key @@ -237,6 +237,6 @@ func (m *lockedMap) Clear(onEvict itemCallback) { onEvict(i) } } - m.data = make(map[uint64]storeItem) + m.data = make(map[uint64]storeItem[V]) m.Unlock() } diff --git a/vendor/github.com/dgraph-io/ristretto/ttl.go b/vendor/github.com/dgraph-io/ristretto/ttl.go index 337976ad4..b71c1d80d 100644 --- a/vendor/github.com/dgraph-io/ristretto/ttl.go +++ b/vendor/github.com/dgraph-io/ristretto/ttl.go @@ -40,18 +40,18 @@ func cleanupBucket(t time.Time) int64 { type bucket map[uint64]uint64 // expirationMap is a map of bucket number to the corresponding bucket. -type expirationMap struct { +type expirationMap[V any] struct { sync.RWMutex buckets map[int64]bucket } -func newExpirationMap() *expirationMap { - return &expirationMap{ +func newExpirationMap[V any]() *expirationMap[V] { + return &expirationMap[V]{ buckets: make(map[int64]bucket), } } -func (m *expirationMap) add(key, conflict uint64, expiration time.Time) { +func (m *expirationMap[_]) add(key, conflict uint64, expiration time.Time) { if m == nil { return } @@ -73,7 +73,7 @@ func (m *expirationMap) add(key, conflict uint64, expiration time.Time) { b[key] = conflict } -func (m *expirationMap) update(key, conflict uint64, oldExpTime, newExpTime time.Time) { +func (m *expirationMap[_]) update(key, conflict uint64, oldExpTime, newExpTime time.Time) { if m == nil { return } @@ -96,7 +96,7 @@ func (m *expirationMap) update(key, conflict uint64, oldExpTime, newExpTime time newBucket[key] = conflict } -func (m *expirationMap) del(key uint64, expiration time.Time) { +func (m *expirationMap[_]) del(key uint64, expiration time.Time) { if m == nil { return } @@ -114,7 +114,7 @@ func (m *expirationMap) del(key uint64, expiration time.Time) { // cleanup removes all the items in the bucket that was just completed. It deletes // those items from the store, and calls the onEvict function on those items. // This function is meant to be called periodically. -func (m *expirationMap) cleanup(store store, policy policy, onEvict itemCallback) { +func (m *expirationMap[V]) cleanup(store store[V], policy policy[V], onEvict func(item *Item[V])) { if m == nil { return } @@ -127,8 +127,9 @@ func (m *expirationMap) cleanup(store store, policy policy, onEvict itemCallback m.Unlock() for key, conflict := range keys { + expr := store.Expiration(key) // Sanity check. Verify that the store agrees that this key is expired. - if store.Expiration(key).After(now) { + if expr.After(now) { continue } @@ -137,10 +138,11 @@ func (m *expirationMap) cleanup(store store, policy policy, onEvict itemCallback _, value := store.Del(key, conflict) if onEvict != nil { - onEvict(&Item{Key: key, - Conflict: conflict, - Value: value, - Cost: cost, + onEvict(&Item[V]{Key: key, + Conflict: conflict, + Value: value, + Cost: cost, + Expiration: expr, }) } } diff --git a/vendor/github.com/dgraph-io/ristretto/z/README.md b/vendor/github.com/dgraph-io/ristretto/z/README.md index 6d77e146e..ad48a3c17 100644 --- a/vendor/github.com/dgraph-io/ristretto/z/README.md +++ b/vendor/github.com/dgraph-io/ristretto/z/README.md @@ -1,22 +1,22 @@ ## bbloom: a bitset Bloom filter for go/golang === -package implements a fast bloom filter with real 'bitset' and JSONMarshal/JSONUnmarshal to store/reload the Bloom filter. +package implements a fast bloom filter with real 'bitset' and JSONMarshal/JSONUnmarshal to store/reload the Bloom filter. NOTE: the package uses unsafe.Pointer to set and read the bits from the bitset. If you're uncomfortable with using the unsafe package, please consider using my bloom filter package at github.com/AndreasBriese/bloom === -changelog 11/2015: new thread safe methods AddTS(), HasTS(), AddIfNotHasTS() following a suggestion from Srdjan Marinovic (github @a-little-srdjan), who used this to code a bloomfilter cache. +changelog 11/2015: new thread safe methods AddTS(), HasTS(), AddIfNotHasTS() following a suggestion from Srdjan Marinovic (github @a-little-srdjan), who used this to code a bloomfilter cache. -This bloom filter was developed to strengthen a website-log database and was tested and optimized for this log-entry mask: "2014/%02i/%02i %02i:%02i:%02i /info.html". -Nonetheless bbloom should work with any other form of entries. +This bloom filter was developed to strengthen a website-log database and was tested and optimized for this log-entry mask: "2014/%02i/%02i %02i:%02i:%02i /info.html". +Nonetheless bbloom should work with any other form of entries. ~~Hash function is a modified Berkeley DB sdbm hash (to optimize for smaller strings). sdbm http://www.cse.yorku.ca/~oz/hash.html~~ Found sipHash (SipHash-2-4, a fast short-input PRF created by Jean-Philippe Aumasson and Daniel J. Bernstein.) to be about as fast. sipHash had been ported by Dimtry Chestnyk to Go (github.com/dchest/siphash ) -Minimum hashset size is: 512 ([4]uint64; will be set automatically). +Minimum hashset size is: 512 ([4]uint64; will be set automatically). ###install @@ -25,7 +25,7 @@ go get github.com/AndreasBriese/bbloom ``` ###test -+ change to folder ../bbloom ++ change to folder ../bbloom + create wordlist in file "words.txt" (you might use `python permut.py`) + run 'go test -bench=.' within the folder @@ -52,10 +52,10 @@ import ( at your header. In the program use ```go -// create a bloom filter for 65536 items and 1 % wrong-positive ratio +// create a bloom filter for 65536 items and 1 % wrong-positive ratio bf := bbloom.New(float64(1<<16), float64(0.01)) -// or +// or // create a bloom filter with 650000 for 65536 items and 7 locs per hash explicitly // bf = bbloom.New(float64(650000), float64(7)) // or @@ -64,7 +64,7 @@ bf = bbloom.New(650000.0, 7.0) // add one item bf.Add([]byte("butter")) -// Number of elements added is exposed now +// Number of elements added is exposed now // Note: ElemNum will not be included in JSON export (for compatability to older version) nOfElementsInFilter := bf.ElemNum @@ -86,7 +86,7 @@ isNotIn = bf.HasTS([]byte("peanutButter")) // should be false added = bf.AddIfNotHasTS([]byte("butter")) // should be false because 'peanutbutter' is already in the set added = bf.AddIfNotHasTS([]byte("peanutbuTTer")) // should be true because 'penutbuTTer' is new -// convert to JSON ([]byte) +// convert to JSON ([]byte) Json := bf.JSONMarshal() // bloomfilters Mutex is exposed for external un-/locking @@ -95,7 +95,7 @@ bf.Mtx.Lock() Json = bf.JSONMarshal() bf.Mtx.Unlock() -// restore a bloom filter from storage +// restore a bloom filter from storage bfNew := bbloom.JSONUnmarshal(Json) isInNew := bfNew.Has([]byte("butter")) // should be true @@ -105,17 +105,17 @@ isNotInNew := bfNew.Has([]byte("Butter")) // should be false to work with the bloom filter. -### why 'fast'? +### why 'fast'? + +It's about 3 times faster than William Fitzgeralds bitset bloom filter https://github.com/willf/bloom . And it is about so fast as my []bool set variant for Boom filters (see https://github.com/AndreasBriese/bloom ) but having a 8times smaller memory footprint: -It's about 3 times faster than William Fitzgeralds bitset bloom filter https://github.com/willf/bloom . And it is about so fast as my []bool set variant for Boom filters (see https://github.com/AndreasBriese/bloom ) but having a 8times smaller memory footprint: - Bloom filter (filter size 524288, 7 hashlocs) github.com/AndreasBriese/bbloom 'Add' 65536 items (10 repetitions): 6595800 ns (100 ns/op) github.com/AndreasBriese/bbloom 'Has' 65536 items (10 repetitions): 5986600 ns (91 ns/op) github.com/AndreasBriese/bloom 'Add' 65536 items (10 repetitions): 6304684 ns (96 ns/op) github.com/AndreasBriese/bloom 'Has' 65536 items (10 repetitions): 6568663 ns (100 ns/op) - + github.com/willf/bloom 'Add' 65536 items (10 repetitions): 24367224 ns (371 ns/op) github.com/willf/bloom 'Test' 65536 items (10 repetitions): 21881142 ns (333 ns/op) github.com/dataence/bloom/standard 'Add' 65536 items (10 repetitions): 23041644 ns (351 ns/op) @@ -126,4 +126,4 @@ It's about 3 times faster than William Fitzgeralds bitset bloom filter https://g (on MBPro15 OSX10.8.5 i7 4Core 2.4Ghz) -With 32bit bloom filters (bloom32) using modified sdbm, bloom32 does hashing with only 2 bit shifts, one xor and one substraction per byte. smdb is about as fast as fnv64a but gives less collisions with the dataset (see mask above). bloom.New(float64(10 * 1<<16),float64(7)) populated with 1<<16 random items from the dataset (see above) and tested against the rest results in less than 0.05% collisions. +With 32bit bloom filters (bloom32) using modified sdbm, bloom32 does hashing with only 2 bit shifts, one xor and one substraction per byte. smdb is about as fast as fnv64a but gives less collisions with the dataset (see mask above). bloom.New(float64(10 * 1<<16),float64(7)) populated with 1<<16 random items from the dataset (see above) and tested against the rest results in less than 0.05% collisions. diff --git a/vendor/github.com/dgraph-io/ristretto/z/bbloom.go b/vendor/github.com/dgraph-io/ristretto/z/bbloom.go index 37135b012..3e2137ff6 100644 --- a/vendor/github.com/dgraph-io/ristretto/z/bbloom.go +++ b/vendor/github.com/dgraph-io/ristretto/z/bbloom.go @@ -23,10 +23,9 @@ package z import ( "bytes" "encoding/json" + "log" "math" "unsafe" - - "github.com/golang/glog" ) // helper @@ -60,7 +59,7 @@ func NewBloomFilter(params ...float64) (bloomfilter *Bloom) { entries, locs = uint64(params[0]), uint64(params[1]) } } else { - glog.Fatal("usage: New(float64(number_of_entries), float64(number_of_hashlocations))" + + log.Fatal("usage: New(float64(number_of_entries), float64(number_of_hashlocations))" + " i.e. New(float64(1000), float64(3)) or New(float64(number_of_entries)," + " float64(number_of_hashlocations)) i.e. New(float64(1000), float64(0.03))") } @@ -205,7 +204,7 @@ func (bl Bloom) JSONMarshal() []byte { } data, err := json.Marshal(bloomImEx) if err != nil { - glog.Fatal("json.Marshal failed: ", err) + log.Fatal("json.Marshal failed: ", err) } return data } diff --git a/vendor/github.com/dgraph-io/ristretto/z/btree.go b/vendor/github.com/dgraph-io/ristretto/z/btree.go index 12b735bb0..33c3046b9 100644 --- a/vendor/github.com/dgraph-io/ristretto/z/btree.go +++ b/vendor/github.com/dgraph-io/ristretto/z/btree.go @@ -30,6 +30,7 @@ import ( var ( pageSize = os.Getpagesize() maxKeys = (pageSize / 16) - 1 + //nolint:unused oneThird = int(float64(maxKeys) / 3) ) @@ -480,6 +481,8 @@ func (t *Tree) split(pid uint64) node { // shareWithSiblingXXX is unused for now. The idea is to move some keys to // sibling when a node is full. But, I don't see any special benefits in our // access pattern. It doesn't result in better occupancy ratios. +// +//nolint:unused func (t *Tree) shareWithSiblingXXX(n node, idx int) bool { if idx == 0 { return false diff --git a/vendor/github.com/dgraph-io/ristretto/z/buffer.go b/vendor/github.com/dgraph-io/ristretto/z/buffer.go index 5a22de8c7..a662d47cc 100644 --- a/vendor/github.com/dgraph-io/ristretto/z/buffer.go +++ b/vendor/github.com/dgraph-io/ristretto/z/buffer.go @@ -19,12 +19,11 @@ package z import ( "encoding/binary" "fmt" - "io/ioutil" + "log" "os" "sort" "sync/atomic" - "github.com/golang/glog" "github.com/pkg/errors" ) @@ -92,7 +91,7 @@ func NewBufferTmp(dir string, capacity int) (*Buffer, error) { if dir == "" { dir = tmpDir } - file, err := ioutil.TempFile(dir, "buffer") + file, err := os.CreateTemp(dir, "buffer") if err != nil { return nil, err } @@ -200,7 +199,7 @@ func (b *Buffer) Grow(n int) { // If autoMmap gets triggered, copy the slice over to an mmaped file. if b.autoMmapAfter > 0 && b.curSz > b.autoMmapAfter { b.bufType = UseMmap - file, err := ioutil.TempFile(b.autoMmapDir, "") + file, err := os.CreateTemp(b.autoMmapDir, "") if err != nil { panic(err) } @@ -254,8 +253,8 @@ func (b *Buffer) AllocateOffset(n int) int { } func (b *Buffer) writeLen(sz int) { - buf := b.Allocate(4) - binary.BigEndian.PutUint32(buf, uint32(sz)) + buf := b.Allocate(8) + binary.BigEndian.PutUint64(buf, uint64(sz)) } // SliceAllocate would encode the size provided into the buffer, followed by a call to Allocate, @@ -263,7 +262,7 @@ func (b *Buffer) writeLen(sz int) { // this big buffer. // Note that SliceAllocate should NOT be mixed with normal calls to Write. func (b *Buffer) SliceAllocate(sz int) []byte { - b.Grow(4 + sz) + b.Grow(8 + sz) b.writeLen(sz) return b.Allocate(sz) } @@ -281,7 +280,9 @@ func (b *Buffer) SliceIterate(f func(slice []byte) error) error { if b.IsEmpty() { return nil } - slice, next := []byte{}, b.StartOffset() + + next := b.StartOffset() + var slice []byte for next >= 0 { slice, next = b.Slice(next) if len(slice) == 0 { @@ -291,6 +292,7 @@ func (b *Buffer) SliceIterate(f func(slice []byte) error) error { return err } } + return nil } @@ -339,19 +341,19 @@ func (s *sortHelper) sortSmall(start, end int) { }) // Now we iterate over the s.small offsets and copy over the slices. The result is now in order. for _, off := range s.small { - s.tmp.Write(rawSlice(s.b.buf[off:])) + _, _ = s.tmp.Write(rawSlice(s.b.buf[off:])) } assert(end-start == copy(s.b.buf[start:end], s.tmp.Bytes())) } func assert(b bool) { if !b { - glog.Fatalf("%+v", errors.Errorf("Assertion failure")) + log.Fatalf("%+v", errors.Errorf("Assertion failure")) } } func check(err error) { if err != nil { - glog.Fatalf("%+v", err) + log.Fatalf("%+v", err) } } func check2(_ interface{}, err error) { @@ -392,7 +394,7 @@ func (s *sortHelper) merge(left, right []byte, start, end int) { rs = rawSlice(right) // We skip the first 4 bytes in the rawSlice, because that stores the length. - if s.less(ls[4:], rs[4:]) { + if s.less(ls[8:], rs[8:]) { copyLeft() } else { copyRight() @@ -454,7 +456,7 @@ func (b *Buffer) SortSliceBetween(start, end int, less LessFunc) { small: make([]int, 0, 1024), tmp: NewBuffer(szTmp, b.tag), } - defer s.tmp.Release() + defer func() { _ = s.tmp.Release() }() left := offsets[0] for _, off := range offsets[1:] { @@ -465,8 +467,8 @@ func (b *Buffer) SortSliceBetween(start, end int, less LessFunc) { } func rawSlice(buf []byte) []byte { - sz := binary.BigEndian.Uint32(buf) - return buf[:4+int(sz)] + sz := binary.BigEndian.Uint64(buf) + return buf[:8+int(sz)] } // Slice would return the slice written at offset. @@ -475,8 +477,8 @@ func (b *Buffer) Slice(offset int) ([]byte, int) { return nil, -1 } - sz := binary.BigEndian.Uint32(b.buf[offset:]) - start := offset + 4 + sz := binary.BigEndian.Uint64(b.buf[offset:]) + start := offset + 8 next := start + int(sz) res := b.buf[start:next] if next >= int(b.offset) { diff --git a/vendor/github.com/dgraph-io/ristretto/z/calloc_64bit.go b/vendor/github.com/dgraph-io/ristretto/z/calloc_64bit.go index b898248bb..6c02cabd9 100644 --- a/vendor/github.com/dgraph-io/ristretto/z/calloc_64bit.go +++ b/vendor/github.com/dgraph-io/ristretto/z/calloc_64bit.go @@ -2,6 +2,7 @@ // of this source code is governed by a BSD-style license that can be found in // the LICENSE file. +//go:build amd64 || arm64 || arm64be || ppc64 || ppc64le || mips64 || mips64le || riscv64 || s390x || sparc64 // +build amd64 arm64 arm64be ppc64 ppc64le mips64 mips64le riscv64 s390x sparc64 package z diff --git a/vendor/github.com/dgraph-io/ristretto/z/calloc_nojemalloc.go b/vendor/github.com/dgraph-io/ristretto/z/calloc_nojemalloc.go index 93ceedf90..20c9ae359 100644 --- a/vendor/github.com/dgraph-io/ristretto/z/calloc_nojemalloc.go +++ b/vendor/github.com/dgraph-io/ristretto/z/calloc_nojemalloc.go @@ -2,6 +2,7 @@ // of this source code is governed by a BSD-style license that can be found in // the LICENSE file. +//go:build !jemalloc || !cgo // +build !jemalloc !cgo package z @@ -34,4 +35,4 @@ func StatsPrint() { // ReadMemStats doesn't do anything since all the memory is being managed // by the Go runtime. -func ReadMemStats(_ *MemStats) { return } +func ReadMemStats(_ *MemStats) {} diff --git a/vendor/github.com/dgraph-io/ristretto/z/file.go b/vendor/github.com/dgraph-io/ristretto/z/file.go index 880caf0ad..c07949a72 100644 --- a/vendor/github.com/dgraph-io/ristretto/z/file.go +++ b/vendor/github.com/dgraph-io/ristretto/z/file.go @@ -61,7 +61,9 @@ func OpenMmapFileUsing(fd *os.File, sz int, writable bool) (*MmapFile, error) { if fileSize == 0 { dir, _ := filepath.Split(filename) - go SyncDir(dir) + if err := SyncDir(dir); err != nil { + return nil, err + } } return &MmapFile{ Data: buf, diff --git a/vendor/github.com/dgraph-io/ristretto/z/file_default.go b/vendor/github.com/dgraph-io/ristretto/z/file_default.go index d9c0db43e..00e7d0870 100644 --- a/vendor/github.com/dgraph-io/ristretto/z/file_default.go +++ b/vendor/github.com/dgraph-io/ristretto/z/file_default.go @@ -1,3 +1,4 @@ +//go:build !linux // +build !linux /* diff --git a/vendor/github.com/dgraph-io/ristretto/z/flags.go b/vendor/github.com/dgraph-io/ristretto/z/flags.go index a55c474ab..84be7ebb9 100644 --- a/vendor/github.com/dgraph-io/ristretto/z/flags.go +++ b/vendor/github.com/dgraph-io/ristretto/z/flags.go @@ -2,6 +2,7 @@ package z import ( "fmt" + "log" "os" "os/user" "path/filepath" @@ -10,7 +11,6 @@ import ( "strings" "time" - "github.com/golang/glog" "github.com/pkg/errors" ) @@ -109,7 +109,7 @@ type SuperFlag struct { func NewSuperFlag(flag string) *SuperFlag { sf, err := newSuperFlagImpl(flag) if err != nil { - glog.Fatal(err) + log.Fatal(err) } return sf } @@ -136,7 +136,7 @@ func (sf *SuperFlag) String() string { func (sf *SuperFlag) MergeAndCheckDefault(flag string) *SuperFlag { sf, err := sf.mergeAndCheckDefaultImpl(flag) if err != nil { - glog.Fatal(err) + log.Fatal(err) } return sf } @@ -162,7 +162,7 @@ func (sf *SuperFlag) mergeAndCheckDefaultImpl(flag string) (*SuperFlag, error) { } } if numKeys != 0 { - return nil, fmt.Errorf("superflag: found invalid options in flag: %s.\nvalid options: %v", sf, flag) + return nil, fmt.Errorf("superflag: found invalid options: %s.\nvalid options: %v", sf, flag) } for k, v := range src { if _, ok := sf.m[k]; !ok { @@ -207,7 +207,7 @@ func (sf *SuperFlag) GetBool(opt string) bool { err = errors.Wrapf(err, "Unable to parse %s as bool for key: %s. Options: %s\n", val, opt, sf) - glog.Fatalf("%+v", err) + log.Fatalf("%+v", err) } return b } @@ -222,7 +222,7 @@ func (sf *SuperFlag) GetFloat64(opt string) float64 { err = errors.Wrapf(err, "Unable to parse %s as float64 for key: %s. Options: %s\n", val, opt, sf) - glog.Fatalf("%+v", err) + log.Fatalf("%+v", err) } return f } @@ -237,7 +237,7 @@ func (sf *SuperFlag) GetInt64(opt string) int64 { err = errors.Wrapf(err, "Unable to parse %s as int64 for key: %s. Options: %s\n", val, opt, sf) - glog.Fatalf("%+v", err) + log.Fatalf("%+v", err) } return i } @@ -252,7 +252,7 @@ func (sf *SuperFlag) GetUint64(opt string) uint64 { err = errors.Wrapf(err, "Unable to parse %s as uint64 for key: %s. Options: %s\n", val, opt, sf) - glog.Fatalf("%+v", err) + log.Fatalf("%+v", err) } return u } @@ -267,7 +267,7 @@ func (sf *SuperFlag) GetUint32(opt string) uint32 { err = errors.Wrapf(err, "Unable to parse %s as uint32 for key: %s. Options: %s\n", val, opt, sf) - glog.Fatalf("%+v", err) + log.Fatalf("%+v", err) } return uint32(u) } @@ -283,7 +283,7 @@ func (sf *SuperFlag) GetPath(opt string) string { p := sf.GetString(opt) path, err := expandPath(p) if err != nil { - glog.Fatalf("Failed to get path: %+v", err) + log.Fatalf("Failed to get path: %+v", err) } return path } diff --git a/vendor/github.com/dgraph-io/ristretto/z/mmap_unix.go b/vendor/github.com/dgraph-io/ristretto/z/mmap_unix.go index e8b2699cf..629449f9d 100644 --- a/vendor/github.com/dgraph-io/ristretto/z/mmap_unix.go +++ b/vendor/github.com/dgraph-io/ristretto/z/mmap_unix.go @@ -1,4 +1,5 @@ -// +build !windows,!darwin,!plan9,!linux +//go:build !windows && !darwin && !plan9 && !linux && !wasip1 +// +build !windows,!darwin,!plan9,!linux,!wasip1 /* * Copyright 2019 Dgraph Labs, Inc. and Contributors diff --git a/vendor/github.com/dgraph-io/ristretto/z/mmap_wasip1.go b/vendor/github.com/dgraph-io/ristretto/z/mmap_wasip1.go new file mode 100644 index 000000000..94f714845 --- /dev/null +++ b/vendor/github.com/dgraph-io/ristretto/z/mmap_wasip1.go @@ -0,0 +1,40 @@ +//go:build wasip1 + +/* + * Copyright 2023 Dgraph Labs, Inc. and Contributors + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package z + +import ( + "os" + "syscall" +) + +func mmap(fd *os.File, writeable bool, size int64) ([]byte, error) { + return nil, syscall.ENOSYS +} + +func munmap(b []byte) error { + return syscall.ENOSYS +} + +func madvise(b []byte, readahead bool) error { + return syscall.ENOSYS +} + +func msync(b []byte) error { + return syscall.ENOSYS +} diff --git a/vendor/github.com/dgraph-io/ristretto/z/mremap_linux_arm64.go b/vendor/github.com/dgraph-io/ristretto/z/mremap_nosize.go similarity index 95% rename from vendor/github.com/dgraph-io/ristretto/z/mremap_linux_arm64.go rename to vendor/github.com/dgraph-io/ristretto/z/mremap_nosize.go index 09683cdfe..d31009a8b 100644 --- a/vendor/github.com/dgraph-io/ristretto/z/mremap_linux_arm64.go +++ b/vendor/github.com/dgraph-io/ristretto/z/mremap_nosize.go @@ -1,3 +1,7 @@ +//go:build (arm64 || arm) && linux +// +build arm64 arm +// +build linux + /* * Copyright 2020 Dgraph Labs, Inc. and Contributors * diff --git a/vendor/github.com/dgraph-io/ristretto/z/mremap_linux.go b/vendor/github.com/dgraph-io/ristretto/z/mremap_size.go similarity index 95% rename from vendor/github.com/dgraph-io/ristretto/z/mremap_linux.go rename to vendor/github.com/dgraph-io/ristretto/z/mremap_size.go index 225678658..35ff3ef17 100644 --- a/vendor/github.com/dgraph-io/ristretto/z/mremap_linux.go +++ b/vendor/github.com/dgraph-io/ristretto/z/mremap_size.go @@ -1,4 +1,5 @@ -// +build !arm64 +//go:build linux && !arm64 && !arm +// +build linux,!arm64,!arm /* * Copyright 2020 Dgraph Labs, Inc. and Contributors diff --git a/vendor/github.com/dgraph-io/ristretto/z/rtutil.go b/vendor/github.com/dgraph-io/ristretto/z/rtutil.go index 8f317c80d..89e417645 100644 --- a/vendor/github.com/dgraph-io/ristretto/z/rtutil.go +++ b/vendor/github.com/dgraph-io/ristretto/z/rtutil.go @@ -27,10 +27,12 @@ import ( ) // NanoTime returns the current time in nanoseconds from a monotonic clock. +// //go:linkname NanoTime runtime.nanotime func NanoTime() int64 // CPUTicks is a faster alternative to NanoTime to measure time duration. +// //go:linkname CPUTicks runtime.cputicks func CPUTicks() int64 @@ -60,6 +62,7 @@ func MemHashString(str string) uint64 { } // FastRand is a fast thread local random function. +// //go:linkname FastRand runtime.fastrand func FastRand() uint32 diff --git a/vendor/github.com/dgraph-io/ristretto/z/simd/baseline.go b/vendor/github.com/dgraph-io/ristretto/z/simd/baseline.go index 967e3a307..e98f62860 100644 --- a/vendor/github.com/dgraph-io/ristretto/z/simd/baseline.go +++ b/vendor/github.com/dgraph-io/ristretto/z/simd/baseline.go @@ -98,6 +98,7 @@ func Binary(keys []uint64, key uint64) int16 { })) } +//nolint:unused func cmp2_native(twos, pk [2]uint64) int16 { if twos[0] == pk[0] { return 0 @@ -108,6 +109,7 @@ func cmp2_native(twos, pk [2]uint64) int16 { return 2 } +//nolint:unused func cmp4_native(fours, pk [4]uint64) int16 { for i := range fours { if fours[i] >= pk[i] { @@ -117,6 +119,7 @@ func cmp4_native(fours, pk [4]uint64) int16 { return 4 } +//nolint:unused func cmp8_native(a [8]uint64, pk [4]uint64) int16 { for i := range a { if a[i] >= pk[0] { diff --git a/vendor/github.com/dgraph-io/ristretto/z/simd/search.go b/vendor/github.com/dgraph-io/ristretto/z/simd/search.go index b1e639225..f27206163 100644 --- a/vendor/github.com/dgraph-io/ristretto/z/simd/search.go +++ b/vendor/github.com/dgraph-io/ristretto/z/simd/search.go @@ -1,3 +1,4 @@ +//go:build !amd64 // +build !amd64 /* @@ -20,7 +21,7 @@ package simd // Search uses the Clever search to find the correct key. func Search(xs []uint64, k uint64) int16 { - if len(xs) < 8 || (len(xs) % 8 != 0) { + if len(xs) < 8 || (len(xs)%8 != 0) { return Naive(xs, k) } var twos, pk [4]uint64 diff --git a/vendor/github.com/dgraph-io/ristretto/z/z.go b/vendor/github.com/dgraph-io/ristretto/z/z.go index 97455586a..ac424583c 100644 --- a/vendor/github.com/dgraph-io/ristretto/z/z.go +++ b/vendor/github.com/dgraph-io/ristretto/z/z.go @@ -23,19 +23,20 @@ import ( "github.com/cespare/xxhash/v2" ) -// TODO: Figure out a way to re-use memhash for the second uint64 hash, we -// already know that appending bytes isn't reliable for generating a -// second hash (see Ristretto PR #88). -// -// We also know that while the Go runtime has a runtime memhash128 -// function, it's not possible to use it to generate [2]uint64 or -// anything resembling a 128bit hash, even though that's exactly what -// we need in this situation. -func KeyToHash(key interface{}) (uint64, uint64) { - if key == nil { - return 0, 0 - } - switch k := key.(type) { +type Key interface { + uint64 | string | []byte | byte | int | int32 | uint32 | int64 +} + +// TODO: Figure out a way to re-use memhash for the second uint64 hash, +// we already know that appending bytes isn't reliable for generating a +// second hash (see Ristretto PR #88). +// We also know that while the Go runtime has a runtime memhash128 +// function, it's not possible to use it to generate [2]uint64 or +// anything resembling a 128bit hash, even though that's exactly what +// we need in this situation. +func KeyToHash[K Key](key K) (uint64, uint64) { + keyAsAny := any(key) + switch k := keyAsAny.(type) { case uint64: return k, 0 case string: diff --git a/vendor/github.com/golang/glog/LICENSE b/vendor/github.com/golang/glog/LICENSE deleted file mode 100644 index 37ec93a14..000000000 --- a/vendor/github.com/golang/glog/LICENSE +++ /dev/null @@ -1,191 +0,0 @@ -Apache License -Version 2.0, January 2004 -http://www.apache.org/licenses/ - -TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION - -1. Definitions. - -"License" shall mean the terms and conditions for use, reproduction, and -distribution as defined by Sections 1 through 9 of this document. - -"Licensor" shall mean the copyright owner or entity authorized by the copyright -owner that is granting the License. - -"Legal Entity" shall mean the union of the acting entity and all other entities -that control, are controlled by, or are under common control with that entity. -For the purposes of this definition, "control" means (i) the power, direct or -indirect, to cause the direction or management of such entity, whether by -contract or otherwise, or (ii) ownership of fifty percent (50%) or more of the -outstanding shares, or (iii) beneficial ownership of such entity. - -"You" (or "Your") shall mean an individual or Legal Entity exercising -permissions granted by this License. - -"Source" form shall mean the preferred form for making modifications, including -but not limited to software source code, documentation source, and configuration -files. - -"Object" form shall mean any form resulting from mechanical transformation or -translation of a Source form, including but not limited to compiled object code, -generated documentation, and conversions to other media types. - -"Work" shall mean the work of authorship, whether in Source or Object form, made -available under the License, as indicated by a copyright notice that is included -in or attached to the work (an example is provided in the Appendix below). - -"Derivative Works" shall mean any work, whether in Source or Object form, that -is based on (or derived from) the Work and for which the editorial revisions, -annotations, elaborations, or other modifications represent, as a whole, an -original work of authorship. For the purposes of this License, Derivative Works -shall not include works that remain separable from, or merely link (or bind by -name) to the interfaces of, the Work and Derivative Works thereof. - -"Contribution" shall mean any work of authorship, including the original version -of the Work and any modifications or additions to that Work or Derivative Works -thereof, that is intentionally submitted to Licensor for inclusion in the Work -by the copyright owner or by an individual or Legal Entity authorized to submit -on behalf of the copyright owner. For the purposes of this definition, -"submitted" means any form of electronic, verbal, or written communication sent -to the Licensor or its representatives, including but not limited to -communication on electronic mailing lists, source code control systems, and -issue tracking systems that are managed by, or on behalf of, the Licensor for -the purpose of discussing and improving the Work, but excluding communication -that is conspicuously marked or otherwise designated in writing by the copyright -owner as "Not a Contribution." - -"Contributor" shall mean Licensor and any individual or Legal Entity on behalf -of whom a Contribution has been received by Licensor and subsequently -incorporated within the Work. - -2. Grant of Copyright License. - -Subject to the terms and conditions of this License, each Contributor hereby -grants to You a perpetual, worldwide, non-exclusive, no-charge, royalty-free, -irrevocable copyright license to reproduce, prepare Derivative Works of, -publicly display, publicly perform, sublicense, and distribute the Work and such -Derivative Works in Source or Object form. - -3. Grant of Patent License. - -Subject to the terms and conditions of this License, each Contributor hereby -grants to You a perpetual, worldwide, non-exclusive, no-charge, royalty-free, -irrevocable (except as stated in this section) patent license to make, have -made, use, offer to sell, sell, import, and otherwise transfer the Work, where -such license applies only to those patent claims licensable by such Contributor -that are necessarily infringed by their Contribution(s) alone or by combination -of their Contribution(s) with the Work to which such Contribution(s) was -submitted. If You institute patent litigation against any entity (including a -cross-claim or counterclaim in a lawsuit) alleging that the Work or a -Contribution incorporated within the Work constitutes direct or contributory -patent infringement, then any patent licenses granted to You under this License -for that Work shall terminate as of the date such litigation is filed. - -4. Redistribution. - -You may reproduce and distribute copies of the Work or Derivative Works thereof -in any medium, with or without modifications, and in Source or Object form, -provided that You meet the following conditions: - -You must give any other recipients of the Work or Derivative Works a copy of -this License; and -You must cause any modified files to carry prominent notices stating that You -changed the files; and -You must retain, in the Source form of any Derivative Works that You distribute, -all copyright, patent, trademark, and attribution notices from the Source form -of the Work, excluding those notices that do not pertain to any part of the -Derivative Works; and -If the Work includes a "NOTICE" text file as part of its distribution, then any -Derivative Works that You distribute must include a readable copy of the -attribution notices contained within such NOTICE file, excluding those notices -that do not pertain to any part of the Derivative Works, in at least one of the -following places: within a NOTICE text file distributed as part of the -Derivative Works; within the Source form or documentation, if provided along -with the Derivative Works; or, within a display generated by the Derivative -Works, if and wherever such third-party notices normally appear. The contents of -the NOTICE file are for informational purposes only and do not modify the -License. You may add Your own attribution notices within Derivative Works that -You distribute, alongside or as an addendum to the NOTICE text from the Work, -provided that such additional attribution notices cannot be construed as -modifying the License. -You may add Your own copyright statement to Your modifications and may provide -additional or different license terms and conditions for use, reproduction, or -distribution of Your modifications, or for any such Derivative Works as a whole, -provided Your use, reproduction, and distribution of the Work otherwise complies -with the conditions stated in this License. - -5. Submission of Contributions. - -Unless You explicitly state otherwise, any Contribution intentionally submitted -for inclusion in the Work by You to the Licensor shall be under the terms and -conditions of this License, without any additional terms or conditions. -Notwithstanding the above, nothing herein shall supersede or modify the terms of -any separate license agreement you may have executed with Licensor regarding -such Contributions. - -6. Trademarks. - -This License does not grant permission to use the trade names, trademarks, -service marks, or product names of the Licensor, except as required for -reasonable and customary use in describing the origin of the Work and -reproducing the content of the NOTICE file. - -7. Disclaimer of Warranty. - -Unless required by applicable law or agreed to in writing, Licensor provides the -Work (and each Contributor provides its Contributions) on an "AS IS" BASIS, -WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied, -including, without limitation, any warranties or conditions of TITLE, -NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A PARTICULAR PURPOSE. You are -solely responsible for determining the appropriateness of using or -redistributing the Work and assume any risks associated with Your exercise of -permissions under this License. - -8. Limitation of Liability. - -In no event and under no legal theory, whether in tort (including negligence), -contract, or otherwise, unless required by applicable law (such as deliberate -and grossly negligent acts) or agreed to in writing, shall any Contributor be -liable to You for damages, including any direct, indirect, special, incidental, -or consequential damages of any character arising as a result of this License or -out of the use or inability to use the Work (including but not limited to -damages for loss of goodwill, work stoppage, computer failure or malfunction, or -any and all other commercial damages or losses), even if such Contributor has -been advised of the possibility of such damages. - -9. Accepting Warranty or Additional Liability. - -While redistributing the Work or Derivative Works thereof, You may choose to -offer, and charge a fee for, acceptance of support, warranty, indemnity, or -other liability obligations and/or rights consistent with this License. However, -in accepting such obligations, You may act only on Your own behalf and on Your -sole responsibility, not on behalf of any other Contributor, and only if You -agree to indemnify, defend, and hold each Contributor harmless for any liability -incurred by, or claims asserted against, such Contributor by reason of your -accepting any such warranty or additional liability. - -END OF TERMS AND CONDITIONS - -APPENDIX: How to apply the Apache License to your work - -To apply the Apache License to your work, attach the following boilerplate -notice, with the fields enclosed by brackets "[]" replaced with your own -identifying information. (Don't include the brackets!) The text should be -enclosed in the appropriate comment syntax for the file format. We also -recommend that a file or class name and description of purpose be included on -the same "printed page" as the copyright notice for easier identification within -third-party archives. - - Copyright [yyyy] [name of copyright owner] - - Licensed under the Apache License, Version 2.0 (the "License"); - you may not use this file except in compliance with the License. - You may obtain a copy of the License at - - http://www.apache.org/licenses/LICENSE-2.0 - - Unless required by applicable law or agreed to in writing, software - distributed under the License is distributed on an "AS IS" BASIS, - WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - See the License for the specific language governing permissions and - limitations under the License. diff --git a/vendor/github.com/golang/glog/README.md b/vendor/github.com/golang/glog/README.md deleted file mode 100644 index a4f73883b..000000000 --- a/vendor/github.com/golang/glog/README.md +++ /dev/null @@ -1,36 +0,0 @@ -# glog - -[![PkgGoDev](https://pkg.go.dev/badge/github.com/golang/glog)](https://pkg.go.dev/github.com/golang/glog) - -Leveled execution logs for Go. - -This is an efficient pure Go implementation of leveled logs in the -manner of the open source C++ package [_glog_](https://github.com/google/glog). - -By binding methods to booleans it is possible to use the log package without paying the expense of evaluating the arguments to the log. Through the `-vmodule` flag, the package also provides fine-grained -control over logging at the file level. - -The comment from `glog.go` introduces the ideas: - -Package _glog_ implements logging analogous to the Google-internal C++ INFO/ERROR/V setup. It provides the functions Info, Warning, Error, Fatal, plus formatting variants such as Infof. It also provides V-style loggingcontrolled by the `-v` and `-vmodule=file=2` flags. - -Basic examples: - -```go -glog.Info("Prepare to repel boarders") - -glog.Fatalf("Initialization failed: %s", err) -``` - -See the documentation for the V function for an explanation of these examples: - -```go -if glog.V(2) { - glog.Info("Starting transaction...") -} -glog.V(2).Infoln("Processed", nItems, "elements") -``` - -The repository contains an open source version of the log package used inside Google. The master copy of the source lives inside Google, not here. The code in this repo is for export only and is not itself under development. Feature requests will be ignored. - -Send bug reports to golang-nuts@googlegroups.com. diff --git a/vendor/github.com/golang/glog/glog.go b/vendor/github.com/golang/glog/glog.go deleted file mode 100644 index 8c00e737a..000000000 --- a/vendor/github.com/golang/glog/glog.go +++ /dev/null @@ -1,777 +0,0 @@ -// Go support for leveled logs, analogous to https://github.com/google/glog. -// -// Copyright 2023 Google Inc. All Rights Reserved. -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -// Package glog implements logging analogous to the Google-internal C++ INFO/ERROR/V setup. -// It provides functions that have a name matched by regex: -// -// (Info|Warning|Error|Fatal)(Context)?(Depth)?(f)? -// -// If Context is present, function takes context.Context argument. The -// context is used to pass through the Trace Context to log sinks that can make use -// of it. -// It is recommended to use the context variant of the functions over the non-context -// variants if a context is available to make sure the Trace Contexts are present -// in logs. -// -// If Depth is present, this function calls log from a different depth in the call stack. -// This enables a callee to emit logs that use the callsite information of its caller -// or any other callers in the stack. When depth == 0, the original callee's line -// information is emitted. When depth > 0, depth frames are skipped in the call stack -// and the final frame is treated like the original callee to Info. -// -// If 'f' is present, function formats according to a format specifier. -// -// This package also provides V-style logging controlled by the -v and -vmodule=file=2 flags. -// -// Basic examples: -// -// glog.Info("Prepare to repel boarders") -// -// glog.Fatalf("Initialization failed: %s", err) -// -// See the documentation for the V function for an explanation of these examples: -// -// if glog.V(2) { -// glog.Info("Starting transaction...") -// } -// -// glog.V(2).Infoln("Processed", nItems, "elements") -// -// Log output is buffered and written periodically using Flush. Programs -// should call Flush before exiting to guarantee all log output is written. -// -// By default, all log statements write to files in a temporary directory. -// This package provides several flags that modify this behavior. -// As a result, flag.Parse must be called before any logging is done. -// -// -logtostderr=false -// Logs are written to standard error instead of to files. -// -alsologtostderr=false -// Logs are written to standard error as well as to files. -// -stderrthreshold=ERROR -// Log events at or above this severity are logged to standard -// error as well as to files. -// -log_dir="" -// Log files will be written to this directory instead of the -// default temporary directory. -// -// Other flags provide aids to debugging. -// -// -log_backtrace_at="" -// A comma-separated list of file and line numbers holding a logging -// statement, such as -// -log_backtrace_at=gopherflakes.go:234 -// A stack trace will be written to the Info log whenever execution -// hits one of these statements. (Unlike with -vmodule, the ".go" -// must bepresent.) -// -v=0 -// Enable V-leveled logging at the specified level. -// -vmodule="" -// The syntax of the argument is a comma-separated list of pattern=N, -// where pattern is a literal file name (minus the ".go" suffix) or -// "glob" pattern and N is a V level. For instance, -// -vmodule=gopher*=3 -// sets the V level to 3 in all Go files whose names begin with "gopher", -// and -// -vmodule=/path/to/glog/glog_test=1 -// sets the V level to 1 in the Go file /path/to/glog/glog_test.go. -// If a glob pattern contains a slash, it is matched against the full path, -// and the file name. Otherwise, the pattern is -// matched only against the file's basename. When both -vmodule and -v -// are specified, the -vmodule values take precedence for the specified -// modules. -package glog - -// This file contains the parts of the log package that are shared among all -// implementations (file, envelope, and appengine). - -import ( - "bytes" - "context" - "errors" - "fmt" - stdLog "log" - "os" - "reflect" - "runtime" - "runtime/pprof" - "strconv" - "sync" - "sync/atomic" - "time" - - "github.com/golang/glog/internal/logsink" - "github.com/golang/glog/internal/stackdump" -) - -var timeNow = time.Now // Stubbed out for testing. - -// MaxSize is the maximum size of a log file in bytes. -var MaxSize uint64 = 1024 * 1024 * 1800 - -// ErrNoLog is the error we return if no log file has yet been created -// for the specified log type. -var ErrNoLog = errors.New("log file not yet created") - -// OutputStats tracks the number of output lines and bytes written. -type OutputStats struct { - lines int64 - bytes int64 -} - -// Lines returns the number of lines written. -func (s *OutputStats) Lines() int64 { - return atomic.LoadInt64(&s.lines) -} - -// Bytes returns the number of bytes written. -func (s *OutputStats) Bytes() int64 { - return atomic.LoadInt64(&s.bytes) -} - -// Stats tracks the number of lines of output and number of bytes -// per severity level. Values must be read with atomic.LoadInt64. -var Stats struct { - Info, Warning, Error OutputStats -} - -var severityStats = [...]*OutputStats{ - logsink.Info: &Stats.Info, - logsink.Warning: &Stats.Warning, - logsink.Error: &Stats.Error, - logsink.Fatal: nil, -} - -// Level specifies a level of verbosity for V logs. The -v flag is of type -// Level and should be modified only through the flag.Value interface. -type Level int32 - -var metaPool sync.Pool // Pool of *logsink.Meta. - -// metaPoolGet returns a *logsink.Meta from metaPool as both an interface and a -// pointer, allocating a new one if necessary. (Returning the interface value -// directly avoids an allocation if there was an existing pointer in the pool.) -func metaPoolGet() (any, *logsink.Meta) { - if metai := metaPool.Get(); metai != nil { - return metai, metai.(*logsink.Meta) - } - meta := new(logsink.Meta) - return meta, meta -} - -type stack bool - -const ( - noStack = stack(false) - withStack = stack(true) -) - -func appendBacktrace(depth int, format string, args []any) (string, []any) { - // Capture a backtrace as a stackdump.Stack (both text and PC slice). - // Structured log sinks can extract the backtrace in whichever format they - // prefer (PCs or text), and Text sinks will include it as just another part - // of the log message. - // - // Use depth instead of depth+1 so that the backtrace always includes the - // log function itself - otherwise the reason for the trace appearing in the - // log may not be obvious to the reader. - dump := stackdump.Caller(depth) - - // Add an arg and an entry in the format string for the stack dump. - // - // Copy the "args" slice to avoid a rare but serious aliasing bug - // (corrupting the caller's slice if they passed it to a non-Fatal call - // using "..."). - format = format + "\n\n%v\n" - args = append(append([]any(nil), args...), dump) - - return format, args -} - -// logf acts as ctxlogf, but doesn't expect a context. -func logf(depth int, severity logsink.Severity, verbose bool, stack stack, format string, args ...any) { - ctxlogf(nil, depth+1, severity, verbose, stack, format, args...) -} - -// ctxlogf writes a log message for a log function call (or log function wrapper) -// at the given depth in the current goroutine's stack. -func ctxlogf(ctx context.Context, depth int, severity logsink.Severity, verbose bool, stack stack, format string, args ...any) { - now := timeNow() - _, file, line, ok := runtime.Caller(depth + 1) - if !ok { - file = "???" - line = 1 - } - - if stack == withStack || backtraceAt(file, line) { - format, args = appendBacktrace(depth+1, format, args) - } - - metai, meta := metaPoolGet() - *meta = logsink.Meta{ - Context: ctx, - Time: now, - File: file, - Line: line, - Depth: depth + 1, - Severity: severity, - Verbose: verbose, - Thread: int64(pid), - } - sinkf(meta, format, args...) - // Clear pointer fields so they can be garbage collected early. - meta.Context = nil - meta.Stack = nil - metaPool.Put(metai) -} - -func sinkf(meta *logsink.Meta, format string, args ...any) { - meta.Depth++ - n, err := logsink.Printf(meta, format, args...) - if stats := severityStats[meta.Severity]; stats != nil { - atomic.AddInt64(&stats.lines, 1) - atomic.AddInt64(&stats.bytes, int64(n)) - } - - if err != nil { - logsink.Printf(meta, "glog: exiting because of error: %s", err) - sinks.file.Flush() - os.Exit(2) - } -} - -// CopyStandardLogTo arranges for messages written to the Go "log" package's -// default logs to also appear in the Google logs for the named and lower -// severities. Subsequent changes to the standard log's default output location -// or format may break this behavior. -// -// Valid names are "INFO", "WARNING", "ERROR", and "FATAL". If the name is not -// recognized, CopyStandardLogTo panics. -func CopyStandardLogTo(name string) { - sev, err := logsink.ParseSeverity(name) - if err != nil { - panic(fmt.Sprintf("log.CopyStandardLogTo(%q): %v", name, err)) - } - // Set a log format that captures the user's file and line: - // d.go:23: message - stdLog.SetFlags(stdLog.Lshortfile) - stdLog.SetOutput(logBridge(sev)) -} - -// NewStandardLogger returns a Logger that writes to the Google logs for the -// named and lower severities. -// -// Valid names are "INFO", "WARNING", "ERROR", and "FATAL". If the name is not -// recognized, NewStandardLogger panics. -func NewStandardLogger(name string) *stdLog.Logger { - sev, err := logsink.ParseSeverity(name) - if err != nil { - panic(fmt.Sprintf("log.NewStandardLogger(%q): %v", name, err)) - } - return stdLog.New(logBridge(sev), "", stdLog.Lshortfile) -} - -// logBridge provides the Write method that enables CopyStandardLogTo to connect -// Go's standard logs to the logs provided by this package. -type logBridge logsink.Severity - -// Write parses the standard logging line and passes its components to the -// logger for severity(lb). -func (lb logBridge) Write(b []byte) (n int, err error) { - var ( - file = "???" - line = 1 - text string - ) - // Split "d.go:23: message" into "d.go", "23", and "message". - if parts := bytes.SplitN(b, []byte{':'}, 3); len(parts) != 3 || len(parts[0]) < 1 || len(parts[2]) < 1 { - text = fmt.Sprintf("bad log format: %s", b) - } else { - file = string(parts[0]) - text = string(parts[2][1:]) // skip leading space - line, err = strconv.Atoi(string(parts[1])) - if err != nil { - text = fmt.Sprintf("bad line number: %s", b) - line = 1 - } - } - - // The depth below hard-codes details of how stdlog gets here. The alternative would be to walk - // up the stack looking for src/log/log.go but that seems like it would be - // unfortunately slow. - const stdLogDepth = 4 - - metai, meta := metaPoolGet() - *meta = logsink.Meta{ - Time: timeNow(), - File: file, - Line: line, - Depth: stdLogDepth, - Severity: logsink.Severity(lb), - Thread: int64(pid), - } - - format := "%s" - args := []any{text} - if backtraceAt(file, line) { - format, args = appendBacktrace(meta.Depth, format, args) - } - - sinkf(meta, format, args...) - metaPool.Put(metai) - - return len(b), nil -} - -// defaultFormat returns a fmt.Printf format specifier that formats its -// arguments as if they were passed to fmt.Print. -func defaultFormat(args []any) string { - n := len(args) - switch n { - case 0: - return "" - case 1: - return "%v" - } - - b := make([]byte, 0, n*3-1) - wasString := true // Suppress leading space. - for _, arg := range args { - isString := arg != nil && reflect.TypeOf(arg).Kind() == reflect.String - if wasString || isString { - b = append(b, "%v"...) - } else { - b = append(b, " %v"...) - } - wasString = isString - } - return string(b) -} - -// lnFormat returns a fmt.Printf format specifier that formats its arguments -// as if they were passed to fmt.Println. -func lnFormat(args []any) string { - if len(args) == 0 { - return "\n" - } - - b := make([]byte, 0, len(args)*3) - for range args { - b = append(b, "%v "...) - } - b[len(b)-1] = '\n' // Replace the last space with a newline. - return string(b) -} - -// Verbose is a boolean type that implements Infof (like Printf) etc. -// See the documentation of V for more information. -type Verbose bool - -// V reports whether verbosity at the call site is at least the requested level. -// The returned value is a boolean of type Verbose, which implements Info, Infoln -// and Infof. These methods will write to the Info log if called. -// Thus, one may write either -// -// if glog.V(2) { glog.Info("log this") } -// -// or -// -// glog.V(2).Info("log this") -// -// The second form is shorter but the first is cheaper if logging is off because it does -// not evaluate its arguments. -// -// Whether an individual call to V generates a log record depends on the setting of -// the -v and --vmodule flags; both are off by default. If the level in the call to -// V is at most the value of -v, or of -vmodule for the source file containing the -// call, the V call will log. -func V(level Level) Verbose { - return VDepth(1, level) -} - -// VDepth acts as V but uses depth to determine which call frame to check vmodule for. -// VDepth(0, level) is the same as V(level). -func VDepth(depth int, level Level) Verbose { - return Verbose(verboseEnabled(depth+1, level)) -} - -// Info is equivalent to the global Info function, guarded by the value of v. -// See the documentation of V for usage. -func (v Verbose) Info(args ...any) { - v.InfoDepth(1, args...) -} - -// InfoDepth is equivalent to the global InfoDepth function, guarded by the value of v. -// See the documentation of V for usage. -func (v Verbose) InfoDepth(depth int, args ...any) { - if v { - logf(depth+1, logsink.Info, true, noStack, defaultFormat(args), args...) - } -} - -// InfoDepthf is equivalent to the global InfoDepthf function, guarded by the value of v. -// See the documentation of V for usage. -func (v Verbose) InfoDepthf(depth int, format string, args ...any) { - if v { - logf(depth+1, logsink.Info, true, noStack, format, args...) - } -} - -// Infoln is equivalent to the global Infoln function, guarded by the value of v. -// See the documentation of V for usage. -func (v Verbose) Infoln(args ...any) { - if v { - logf(1, logsink.Info, true, noStack, lnFormat(args), args...) - } -} - -// Infof is equivalent to the global Infof function, guarded by the value of v. -// See the documentation of V for usage. -func (v Verbose) Infof(format string, args ...any) { - if v { - logf(1, logsink.Info, true, noStack, format, args...) - } -} - -// InfoContext is equivalent to the global InfoContext function, guarded by the value of v. -// See the documentation of V for usage. -func (v Verbose) InfoContext(ctx context.Context, args ...any) { - v.InfoContextDepth(ctx, 1, args...) -} - -// InfoContextf is equivalent to the global InfoContextf function, guarded by the value of v. -// See the documentation of V for usage. -func (v Verbose) InfoContextf(ctx context.Context, format string, args ...any) { - if v { - ctxlogf(ctx, 1, logsink.Info, true, noStack, format, args...) - } -} - -// InfoContextDepth is equivalent to the global InfoContextDepth function, guarded by the value of v. -// See the documentation of V for usage. -func (v Verbose) InfoContextDepth(ctx context.Context, depth int, args ...any) { - if v { - ctxlogf(ctx, depth+1, logsink.Info, true, noStack, defaultFormat(args), args...) - } -} - -// InfoContextDepthf is equivalent to the global InfoContextDepthf function, guarded by the value of v. -// See the documentation of V for usage. -func (v Verbose) InfoContextDepthf(ctx context.Context, depth int, format string, args ...any) { - if v { - ctxlogf(ctx, depth+1, logsink.Info, true, noStack, format, args...) - } -} - -// Info logs to the INFO log. -// Arguments are handled in the manner of fmt.Print; a newline is appended if missing. -func Info(args ...any) { - InfoDepth(1, args...) -} - -// InfoDepth calls Info from a different depth in the call stack. -// This enables a callee to emit logs that use the callsite information of its caller -// or any other callers in the stack. When depth == 0, the original callee's line -// information is emitted. When depth > 0, depth frames are skipped in the call stack -// and the final frame is treated like the original callee to Info. -func InfoDepth(depth int, args ...any) { - logf(depth+1, logsink.Info, false, noStack, defaultFormat(args), args...) -} - -// InfoDepthf acts as InfoDepth but with format string. -func InfoDepthf(depth int, format string, args ...any) { - logf(depth+1, logsink.Info, false, noStack, format, args...) -} - -// Infoln logs to the INFO log. -// Arguments are handled in the manner of fmt.Println; a newline is appended if missing. -func Infoln(args ...any) { - logf(1, logsink.Info, false, noStack, lnFormat(args), args...) -} - -// Infof logs to the INFO log. -// Arguments are handled in the manner of fmt.Printf; a newline is appended if missing. -func Infof(format string, args ...any) { - logf(1, logsink.Info, false, noStack, format, args...) -} - -// InfoContext is like [Info], but with an extra [context.Context] parameter. The -// context is used to pass the Trace Context to log sinks. -func InfoContext(ctx context.Context, args ...any) { - InfoContextDepth(ctx, 1, args...) -} - -// InfoContextf is like [Infof], but with an extra [context.Context] parameter. The -// context is used to pass the Trace Context to log sinks. -func InfoContextf(ctx context.Context, format string, args ...any) { - ctxlogf(ctx, 1, logsink.Info, false, noStack, format, args...) -} - -// InfoContextDepth is like [InfoDepth], but with an extra [context.Context] parameter. The -// context is used to pass the Trace Context to log sinks. -func InfoContextDepth(ctx context.Context, depth int, args ...any) { - ctxlogf(ctx, depth+1, logsink.Info, false, noStack, defaultFormat(args), args...) -} - -// InfoContextDepthf is like [InfoDepthf], but with an extra [context.Context] parameter. The -// context is used to pass the Trace Context to log sinks. -func InfoContextDepthf(ctx context.Context, depth int, format string, args ...any) { - ctxlogf(ctx, depth+1, logsink.Info, false, noStack, format, args...) -} - -// Warning logs to the WARNING and INFO logs. -// Arguments are handled in the manner of fmt.Print; a newline is appended if missing. -func Warning(args ...any) { - WarningDepth(1, args...) -} - -// WarningDepth acts as Warning but uses depth to determine which call frame to log. -// WarningDepth(0, "msg") is the same as Warning("msg"). -func WarningDepth(depth int, args ...any) { - logf(depth+1, logsink.Warning, false, noStack, defaultFormat(args), args...) -} - -// WarningDepthf acts as Warningf but uses depth to determine which call frame to log. -// WarningDepthf(0, "msg") is the same as Warningf("msg"). -func WarningDepthf(depth int, format string, args ...any) { - logf(depth+1, logsink.Warning, false, noStack, format, args...) -} - -// Warningln logs to the WARNING and INFO logs. -// Arguments are handled in the manner of fmt.Println; a newline is appended if missing. -func Warningln(args ...any) { - logf(1, logsink.Warning, false, noStack, lnFormat(args), args...) -} - -// Warningf logs to the WARNING and INFO logs. -// Arguments are handled in the manner of fmt.Printf; a newline is appended if missing. -func Warningf(format string, args ...any) { - logf(1, logsink.Warning, false, noStack, format, args...) -} - -// WarningContext is like [Warning], but with an extra [context.Context] parameter. The -// context is used to pass the Trace Context to log sinks. -func WarningContext(ctx context.Context, args ...any) { - WarningContextDepth(ctx, 1, args...) -} - -// WarningContextf is like [Warningf], but with an extra [context.Context] parameter. The -// context is used to pass the Trace Context to log sinks. -func WarningContextf(ctx context.Context, format string, args ...any) { - ctxlogf(ctx, 1, logsink.Warning, false, noStack, format, args...) -} - -// WarningContextDepth is like [WarningDepth], but with an extra [context.Context] parameter. The -// context is used to pass the Trace Context to log sinks. -func WarningContextDepth(ctx context.Context, depth int, args ...any) { - ctxlogf(ctx, depth+1, logsink.Warning, false, noStack, defaultFormat(args), args...) -} - -// WarningContextDepthf is like [WarningDepthf], but with an extra [context.Context] parameter. The -// context is used to pass the Trace Context to log sinks. -func WarningContextDepthf(ctx context.Context, depth int, format string, args ...any) { - ctxlogf(ctx, depth+1, logsink.Warning, false, noStack, format, args...) -} - -// Error logs to the ERROR, WARNING, and INFO logs. -// Arguments are handled in the manner of fmt.Print; a newline is appended if missing. -func Error(args ...any) { - ErrorDepth(1, args...) -} - -// ErrorDepth acts as Error but uses depth to determine which call frame to log. -// ErrorDepth(0, "msg") is the same as Error("msg"). -func ErrorDepth(depth int, args ...any) { - logf(depth+1, logsink.Error, false, noStack, defaultFormat(args), args...) -} - -// ErrorDepthf acts as Errorf but uses depth to determine which call frame to log. -// ErrorDepthf(0, "msg") is the same as Errorf("msg"). -func ErrorDepthf(depth int, format string, args ...any) { - logf(depth+1, logsink.Error, false, noStack, format, args...) -} - -// Errorln logs to the ERROR, WARNING, and INFO logs. -// Arguments are handled in the manner of fmt.Println; a newline is appended if missing. -func Errorln(args ...any) { - logf(1, logsink.Error, false, noStack, lnFormat(args), args...) -} - -// Errorf logs to the ERROR, WARNING, and INFO logs. -// Arguments are handled in the manner of fmt.Printf; a newline is appended if missing. -func Errorf(format string, args ...any) { - logf(1, logsink.Error, false, noStack, format, args...) -} - -// ErrorContext is like [Error], but with an extra [context.Context] parameter. The -// context is used to pass the Trace Context to log sinks. -func ErrorContext(ctx context.Context, args ...any) { - ErrorContextDepth(ctx, 1, args...) -} - -// ErrorContextf is like [Errorf], but with an extra [context.Context] parameter. The -// context is used to pass the Trace Context to log sinks. -func ErrorContextf(ctx context.Context, format string, args ...any) { - ctxlogf(ctx, 1, logsink.Error, false, noStack, format, args...) -} - -// ErrorContextDepth is like [ErrorDepth], but with an extra [context.Context] parameter. The -// context is used to pass the Trace Context to log sinks. -func ErrorContextDepth(ctx context.Context, depth int, args ...any) { - ctxlogf(ctx, depth+1, logsink.Error, false, noStack, defaultFormat(args), args...) -} - -// ErrorContextDepthf is like [ErrorDepthf], but with an extra [context.Context] parameter. The -// context is used to pass the Trace Context to log sinks. -func ErrorContextDepthf(ctx context.Context, depth int, format string, args ...any) { - ctxlogf(ctx, depth+1, logsink.Error, false, noStack, format, args...) -} - -func ctxfatalf(ctx context.Context, depth int, format string, args ...any) { - ctxlogf(ctx, depth+1, logsink.Fatal, false, withStack, format, args...) - sinks.file.Flush() - - err := abortProcess() // Should not return. - - // Failed to abort the process using signals. Dump a stack trace and exit. - Errorf("abortProcess returned unexpectedly: %v", err) - sinks.file.Flush() - pprof.Lookup("goroutine").WriteTo(os.Stderr, 1) - os.Exit(2) // Exit with the same code as the default SIGABRT handler. -} - -func fatalf(depth int, format string, args ...any) { - ctxfatalf(nil, depth+1, format, args...) -} - -// Fatal logs to the FATAL, ERROR, WARNING, and INFO logs, -// including a stack trace of all running goroutines, then calls os.Exit(2). -// Arguments are handled in the manner of fmt.Print; a newline is appended if missing. -func Fatal(args ...any) { - FatalDepth(1, args...) -} - -// FatalDepth acts as Fatal but uses depth to determine which call frame to log. -// FatalDepth(0, "msg") is the same as Fatal("msg"). -func FatalDepth(depth int, args ...any) { - fatalf(depth+1, defaultFormat(args), args...) -} - -// FatalDepthf acts as Fatalf but uses depth to determine which call frame to log. -// FatalDepthf(0, "msg") is the same as Fatalf("msg"). -func FatalDepthf(depth int, format string, args ...any) { - fatalf(depth+1, format, args...) -} - -// Fatalln logs to the FATAL, ERROR, WARNING, and INFO logs, -// including a stack trace of all running goroutines, then calls os.Exit(2). -// Arguments are handled in the manner of fmt.Println; a newline is appended if missing. -func Fatalln(args ...any) { - fatalf(1, lnFormat(args), args...) -} - -// Fatalf logs to the FATAL, ERROR, WARNING, and INFO logs, -// including a stack trace of all running goroutines, then calls os.Exit(2). -// Arguments are handled in the manner of fmt.Printf; a newline is appended if missing. -func Fatalf(format string, args ...any) { - fatalf(1, format, args...) -} - -// FatalContext is like [Fatal], but with an extra [context.Context] parameter. The -// context is used to pass the Trace Context to log sinks. -func FatalContext(ctx context.Context, args ...any) { - FatalContextDepth(ctx, 1, args...) -} - -// FatalContextf is like [Fatalf], but with an extra [context.Context] parameter. The -// context is used to pass the Trace Context to log sinks. -func FatalContextf(ctx context.Context, format string, args ...any) { - ctxfatalf(ctx, 1, format, args...) -} - -// FatalContextDepth is like [FatalDepth], but with an extra [context.Context] parameter. The -// context is used to pass the Trace Context to log sinks. -func FatalContextDepth(ctx context.Context, depth int, args ...any) { - ctxfatalf(ctx, depth+1, defaultFormat(args), args...) -} - -// FatalContextDepthf is like [FatalDepthf], but with an extra [context.Context] parameter. -func FatalContextDepthf(ctx context.Context, depth int, format string, args ...any) { - ctxfatalf(ctx, depth+1, format, args...) -} - -func ctxexitf(ctx context.Context, depth int, format string, args ...any) { - ctxlogf(ctx, depth+1, logsink.Fatal, false, noStack, format, args...) - sinks.file.Flush() - os.Exit(1) -} - -func exitf(depth int, format string, args ...any) { - ctxexitf(nil, depth+1, format, args...) -} - -// Exit logs to the FATAL, ERROR, WARNING, and INFO logs, then calls os.Exit(1). -// Arguments are handled in the manner of fmt.Print; a newline is appended if missing. -func Exit(args ...any) { - ExitDepth(1, args...) -} - -// ExitDepth acts as Exit but uses depth to determine which call frame to log. -// ExitDepth(0, "msg") is the same as Exit("msg"). -func ExitDepth(depth int, args ...any) { - exitf(depth+1, defaultFormat(args), args...) -} - -// ExitDepthf acts as Exitf but uses depth to determine which call frame to log. -// ExitDepthf(0, "msg") is the same as Exitf("msg"). -func ExitDepthf(depth int, format string, args ...any) { - exitf(depth+1, format, args...) -} - -// Exitln logs to the FATAL, ERROR, WARNING, and INFO logs, then calls os.Exit(1). -func Exitln(args ...any) { - exitf(1, lnFormat(args), args...) -} - -// Exitf logs to the FATAL, ERROR, WARNING, and INFO logs, then calls os.Exit(1). -// Arguments are handled in the manner of fmt.Printf; a newline is appended if missing. -func Exitf(format string, args ...any) { - exitf(1, format, args...) -} - -// ExitContext is like [Exit], but with an extra [context.Context] parameter. The -// context is used to pass the Trace Context to log sinks. -func ExitContext(ctx context.Context, args ...any) { - ExitContextDepth(ctx, 1, args...) -} - -// ExitContextf is like [Exitf], but with an extra [context.Context] parameter. The -// context is used to pass the Trace Context to log sinks. -func ExitContextf(ctx context.Context, format string, args ...any) { - ctxexitf(ctx, 1, format, args...) -} - -// ExitContextDepth is like [ExitDepth], but with an extra [context.Context] parameter. The -// context is used to pass the Trace Context to log sinks. -func ExitContextDepth(ctx context.Context, depth int, args ...any) { - ctxexitf(ctx, depth+1, defaultFormat(args), args...) -} - -// ExitContextDepthf is like [ExitDepthf], but with an extra [context.Context] parameter. The -// context is used to pass the Trace Context to log sinks. -func ExitContextDepthf(ctx context.Context, depth int, format string, args ...any) { - ctxexitf(ctx, depth+1, format, args...) -} diff --git a/vendor/github.com/golang/glog/glog_file.go b/vendor/github.com/golang/glog/glog_file.go deleted file mode 100644 index 8eb8b08c6..000000000 --- a/vendor/github.com/golang/glog/glog_file.go +++ /dev/null @@ -1,418 +0,0 @@ -// Go support for leveled logs, analogous to https://github.com/google/glog. -// -// Copyright 2023 Google Inc. All Rights Reserved. -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -// File I/O for logs. - -package glog - -import ( - "bufio" - "bytes" - "errors" - "flag" - "fmt" - "io" - "os" - "path/filepath" - "runtime" - "strings" - "sync" - "time" - - "github.com/golang/glog/internal/logsink" -) - -// logDirs lists the candidate directories for new log files. -var logDirs []string - -var ( - // If non-empty, overrides the choice of directory in which to write logs. - // See createLogDirs for the full list of possible destinations. - logDir = flag.String("log_dir", "", "If non-empty, write log files in this directory") - logLink = flag.String("log_link", "", "If non-empty, add symbolic links in this directory to the log files") - logBufLevel = flag.Int("logbuflevel", int(logsink.Info), "Buffer log messages logged at this level or lower"+ - " (-1 means don't buffer; 0 means buffer INFO only; ...). Has limited applicability on non-prod platforms.") -) - -func createLogDirs() { - if *logDir != "" { - logDirs = append(logDirs, *logDir) - } - logDirs = append(logDirs, os.TempDir()) -} - -var ( - pid = os.Getpid() - program = filepath.Base(os.Args[0]) - host = "unknownhost" - userName = "unknownuser" -) - -func init() { - h, err := os.Hostname() - if err == nil { - host = shortHostname(h) - } - - if u := lookupUser(); u != "" { - userName = u - } - // Sanitize userName since it is used to construct file paths. - userName = strings.Map(func(r rune) rune { - switch { - case r >= 'a' && r <= 'z': - case r >= 'A' && r <= 'Z': - case r >= '0' && r <= '9': - default: - return '_' - } - return r - }, userName) -} - -// shortHostname returns its argument, truncating at the first period. -// For instance, given "www.google.com" it returns "www". -func shortHostname(hostname string) string { - if i := strings.Index(hostname, "."); i >= 0 { - return hostname[:i] - } - return hostname -} - -// logName returns a new log file name containing tag, with start time t, and -// the name for the symlink for tag. -func logName(tag string, t time.Time) (name, link string) { - name = fmt.Sprintf("%s.%s.%s.log.%s.%04d%02d%02d-%02d%02d%02d.%d", - program, - host, - userName, - tag, - t.Year(), - t.Month(), - t.Day(), - t.Hour(), - t.Minute(), - t.Second(), - pid) - return name, program + "." + tag -} - -var onceLogDirs sync.Once - -// create creates a new log file and returns the file and its filename, which -// contains tag ("INFO", "FATAL", etc.) and t. If the file is created -// successfully, create also attempts to update the symlink for that tag, ignoring -// errors. -func create(tag string, t time.Time) (f *os.File, filename string, err error) { - onceLogDirs.Do(createLogDirs) - if len(logDirs) == 0 { - return nil, "", errors.New("log: no log dirs") - } - name, link := logName(tag, t) - var lastErr error - for _, dir := range logDirs { - fname := filepath.Join(dir, name) - f, err := os.Create(fname) - if err == nil { - symlink := filepath.Join(dir, link) - os.Remove(symlink) // ignore err - os.Symlink(name, symlink) // ignore err - if *logLink != "" { - lsymlink := filepath.Join(*logLink, link) - os.Remove(lsymlink) // ignore err - os.Symlink(fname, lsymlink) // ignore err - } - return f, fname, nil - } - lastErr = err - } - return nil, "", fmt.Errorf("log: cannot create log: %v", lastErr) -} - -// flushSyncWriter is the interface satisfied by logging destinations. -type flushSyncWriter interface { - Flush() error - Sync() error - io.Writer - filenames() []string -} - -var sinks struct { - stderr stderrSink - file fileSink -} - -func init() { - // Register stderr first: that way if we crash during file-writing at least - // the log will have gone somewhere. - logsink.TextSinks = append(logsink.TextSinks, &sinks.stderr, &sinks.file) - - sinks.file.flushChan = make(chan logsink.Severity, 1) - go sinks.file.flushDaemon() -} - -// stderrSink is a logsink.Text that writes log entries to stderr -// if they meet certain conditions. -type stderrSink struct { - mu sync.Mutex - w io.Writer // if nil Emit uses os.Stderr directly -} - -// Enabled implements logsink.Text.Enabled. It returns true if any of the -// various stderr flags are enabled for logs of the given severity, if the log -// message is from the standard "log" package, or if google.Init has not yet run -// (and hence file logging is not yet initialized). -func (s *stderrSink) Enabled(m *logsink.Meta) bool { - return toStderr || alsoToStderr || m.Severity >= stderrThreshold.get() -} - -// Emit implements logsink.Text.Emit. -func (s *stderrSink) Emit(m *logsink.Meta, data []byte) (n int, err error) { - s.mu.Lock() - defer s.mu.Unlock() - w := s.w - if w == nil { - w = os.Stderr - } - dn, err := w.Write(data) - n += dn - return n, err -} - -// severityWriters is an array of flushSyncWriter with a value for each -// logsink.Severity. -type severityWriters [4]flushSyncWriter - -// fileSink is a logsink.Text that prints to a set of Google log files. -type fileSink struct { - mu sync.Mutex - // file holds writer for each of the log types. - file severityWriters - flushChan chan logsink.Severity -} - -// Enabled implements logsink.Text.Enabled. It returns true if google.Init -// has run and both --disable_log_to_disk and --logtostderr are false. -func (s *fileSink) Enabled(m *logsink.Meta) bool { - return !toStderr -} - -// Emit implements logsink.Text.Emit -func (s *fileSink) Emit(m *logsink.Meta, data []byte) (n int, err error) { - s.mu.Lock() - defer s.mu.Unlock() - - if err = s.createMissingFiles(m.Severity); err != nil { - return 0, err - } - for sev := m.Severity; sev >= logsink.Info; sev-- { - if _, fErr := s.file[sev].Write(data); fErr != nil && err == nil { - err = fErr // Take the first error. - } - } - n = len(data) - if int(m.Severity) > *logBufLevel { - select { - case s.flushChan <- m.Severity: - default: - } - } - - return n, err -} - -// syncBuffer joins a bufio.Writer to its underlying file, providing access to the -// file's Sync method and providing a wrapper for the Write method that provides log -// file rotation. There are conflicting methods, so the file cannot be embedded. -// s.mu is held for all its methods. -type syncBuffer struct { - sink *fileSink - *bufio.Writer - file *os.File - names []string - sev logsink.Severity - nbytes uint64 // The number of bytes written to this file -} - -func (sb *syncBuffer) Sync() error { - return sb.file.Sync() -} - -func (sb *syncBuffer) Write(p []byte) (n int, err error) { - if sb.nbytes+uint64(len(p)) >= MaxSize { - if err := sb.rotateFile(time.Now()); err != nil { - return 0, err - } - } - n, err = sb.Writer.Write(p) - sb.nbytes += uint64(n) - return n, err -} - -func (sb *syncBuffer) filenames() []string { - return sb.names -} - -const footer = "\nCONTINUED IN NEXT FILE\n" - -// rotateFile closes the syncBuffer's file and starts a new one. -func (sb *syncBuffer) rotateFile(now time.Time) error { - var err error - pn := "" - file, name, err := create(sb.sev.String(), now) - - if sb.file != nil { - // The current log file becomes the previous log at the end of - // this block, so save its name for use in the header of the next - // file. - pn = sb.file.Name() - sb.Flush() - // If there's an existing file, write a footer with the name of - // the next file in the chain, followed by the constant string - // \nCONTINUED IN NEXT FILE\n to make continuation detection simple. - sb.file.Write([]byte("Next log: ")) - sb.file.Write([]byte(name)) - sb.file.Write([]byte(footer)) - sb.file.Close() - } - - sb.file = file - sb.names = append(sb.names, name) - sb.nbytes = 0 - if err != nil { - return err - } - - sb.Writer = bufio.NewWriterSize(sb.file, bufferSize) - - // Write header. - var buf bytes.Buffer - fmt.Fprintf(&buf, "Log file created at: %s\n", now.Format("2006/01/02 15:04:05")) - fmt.Fprintf(&buf, "Running on machine: %s\n", host) - fmt.Fprintf(&buf, "Binary: Built with %s %s for %s/%s\n", runtime.Compiler, runtime.Version(), runtime.GOOS, runtime.GOARCH) - fmt.Fprintf(&buf, "Previous log: %s\n", pn) - fmt.Fprintf(&buf, "Log line format: [IWEF]mmdd hh:mm:ss.uuuuuu threadid file:line] msg\n") - n, err := sb.file.Write(buf.Bytes()) - sb.nbytes += uint64(n) - return err -} - -// bufferSize sizes the buffer associated with each log file. It's large -// so that log records can accumulate without the logging thread blocking -// on disk I/O. The flushDaemon will block instead. -const bufferSize = 256 * 1024 - -// createMissingFiles creates all the log files for severity from infoLog up to -// upTo that have not already been created. -// s.mu is held. -func (s *fileSink) createMissingFiles(upTo logsink.Severity) error { - if s.file[upTo] != nil { - return nil - } - now := time.Now() - // Files are created in increasing severity order, so we can be assured that - // if a high severity logfile exists, then so do all of lower severity. - for sev := logsink.Info; sev <= upTo; sev++ { - if s.file[sev] != nil { - continue - } - sb := &syncBuffer{ - sink: s, - sev: sev, - } - if err := sb.rotateFile(now); err != nil { - return err - } - s.file[sev] = sb - } - return nil -} - -// flushDaemon periodically flushes the log file buffers. -func (s *fileSink) flushDaemon() { - tick := time.NewTicker(30 * time.Second) - defer tick.Stop() - for { - select { - case <-tick.C: - s.Flush() - case sev := <-s.flushChan: - s.flush(sev) - } - } -} - -// Flush flushes all pending log I/O. -func Flush() { - sinks.file.Flush() -} - -// Flush flushes all the logs and attempts to "sync" their data to disk. -func (s *fileSink) Flush() error { - return s.flush(logsink.Info) -} - -// flush flushes all logs of severity threshold or greater. -func (s *fileSink) flush(threshold logsink.Severity) error { - var firstErr error - updateErr := func(err error) { - if err != nil && firstErr == nil { - firstErr = err - } - } - - // Remember where we flushed, so we can call sync without holding - // the lock. - var files []flushSyncWriter - func() { - s.mu.Lock() - defer s.mu.Unlock() - // Flush from fatal down, in case there's trouble flushing. - for sev := logsink.Fatal; sev >= threshold; sev-- { - if file := s.file[sev]; file != nil { - updateErr(file.Flush()) - files = append(files, file) - } - } - }() - - for _, file := range files { - updateErr(file.Sync()) - } - - return firstErr -} - -// Names returns the names of the log files holding the FATAL, ERROR, -// WARNING, or INFO logs. Returns ErrNoLog if the log for the given -// level doesn't exist (e.g. because no messages of that level have been -// written). This may return multiple names if the log type requested -// has rolled over. -func Names(s string) ([]string, error) { - severity, err := logsink.ParseSeverity(s) - if err != nil { - return nil, err - } - - sinks.file.mu.Lock() - defer sinks.file.mu.Unlock() - f := sinks.file.file[severity] - if f == nil { - return nil, ErrNoLog - } - - return f.filenames(), nil -} diff --git a/vendor/github.com/golang/glog/glog_file_linux.go b/vendor/github.com/golang/glog/glog_file_linux.go deleted file mode 100644 index d795092d0..000000000 --- a/vendor/github.com/golang/glog/glog_file_linux.go +++ /dev/null @@ -1,39 +0,0 @@ -// Go support for leveled logs, analogous to https://github.com/google/glog. -// -// Copyright 2023 Google Inc. All Rights Reserved. -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -//go:build linux - -package glog - -import ( - "errors" - "runtime" - "syscall" -) - -// abortProcess attempts to kill the current process in a way that will dump the -// currently-running goroutines someplace useful (like stderr). -// -// It does this by sending SIGABRT to the current thread. -// -// If successful, abortProcess does not return. -func abortProcess() error { - runtime.LockOSThread() - if err := syscall.Tgkill(syscall.Getpid(), syscall.Gettid(), syscall.SIGABRT); err != nil { - return err - } - return errors.New("log: killed current thread with SIGABRT, but still running") -} diff --git a/vendor/github.com/golang/glog/glog_file_nonwindows.go b/vendor/github.com/golang/glog/glog_file_nonwindows.go deleted file mode 100644 index d5cdb793c..000000000 --- a/vendor/github.com/golang/glog/glog_file_nonwindows.go +++ /dev/null @@ -1,12 +0,0 @@ -//go:build !windows - -package glog - -import "os/user" - -func lookupUser() string { - if current, err := user.Current(); err == nil { - return current.Username - } - return "" -} diff --git a/vendor/github.com/golang/glog/glog_file_other.go b/vendor/github.com/golang/glog/glog_file_other.go deleted file mode 100644 index 9540f14fc..000000000 --- a/vendor/github.com/golang/glog/glog_file_other.go +++ /dev/null @@ -1,30 +0,0 @@ -// Go support for leveled logs, analogous to https://github.com/google/glog. -// -// Copyright 2023 Google Inc. All Rights Reserved. -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -//go:build !(unix || windows) - -package glog - -import ( - "fmt" - "runtime" -) - -// abortProcess returns an error on platforms that presumably don't support signals. -func abortProcess() error { - return fmt.Errorf("not sending SIGABRT (%s/%s does not support signals), falling back", runtime.GOOS, runtime.GOARCH) - -} diff --git a/vendor/github.com/golang/glog/glog_file_posix.go b/vendor/github.com/golang/glog/glog_file_posix.go deleted file mode 100644 index c27c7c0e4..000000000 --- a/vendor/github.com/golang/glog/glog_file_posix.go +++ /dev/null @@ -1,53 +0,0 @@ -// Go support for leveled logs, analogous to https://github.com/google/glog. -// -// Copyright 2023 Google Inc. All Rights Reserved. -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -//go:build (unix || windows) && !linux - -package glog - -import ( - "os" - "syscall" - "time" -) - -// abortProcess attempts to kill the current process in a way that will dump the -// currently-running goroutines someplace useful (like stderr). -// -// It does this by sending SIGABRT to the current process. Unfortunately, the -// signal may or may not be delivered to the current thread; in order to do that -// portably, we would need to add a cgo dependency and call pthread_kill. -// -// If successful, abortProcess does not return. -func abortProcess() error { - p, err := os.FindProcess(os.Getpid()) - if err != nil { - return err - } - if err := p.Signal(syscall.SIGABRT); err != nil { - return err - } - - // Sent the signal. Now we wait for it to arrive and any SIGABRT handlers to - // run (and eventually terminate the process themselves). - // - // We could just "select{}" here, but there's an outside chance that would - // trigger the runtime's deadlock detector if there happen not to be any - // background goroutines running. So we'll sleep a while first to give - // the signal some time. - time.Sleep(10 * time.Second) - select {} -} diff --git a/vendor/github.com/golang/glog/glog_file_windows.go b/vendor/github.com/golang/glog/glog_file_windows.go deleted file mode 100644 index a9e4f609d..000000000 --- a/vendor/github.com/golang/glog/glog_file_windows.go +++ /dev/null @@ -1,30 +0,0 @@ -//go:build windows - -package glog - -import ( - "syscall" -) - -// This follows the logic in the standard library's user.Current() function, except -// that it leaves out the potentially expensive calls required to look up the user's -// display name in Active Directory. -func lookupUser() string { - token, err := syscall.OpenCurrentProcessToken() - if err != nil { - return "" - } - defer token.Close() - tokenUser, err := token.GetTokenUser() - if err != nil { - return "" - } - username, _, accountType, err := tokenUser.User.Sid.LookupAccount("") - if err != nil { - return "" - } - if accountType != syscall.SidTypeUser { - return "" - } - return username -} diff --git a/vendor/github.com/golang/glog/glog_flags.go b/vendor/github.com/golang/glog/glog_flags.go deleted file mode 100644 index fa4371afd..000000000 --- a/vendor/github.com/golang/glog/glog_flags.go +++ /dev/null @@ -1,398 +0,0 @@ -// Go support for leveled logs, analogous to https://github.com/google/glog. -// -// Copyright 2023 Google Inc. All Rights Reserved. -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -package glog - -import ( - "bytes" - "errors" - "flag" - "fmt" - "path/filepath" - "runtime" - "strconv" - "strings" - "sync" - "sync/atomic" - - "github.com/golang/glog/internal/logsink" -) - -// modulePat contains a filter for the -vmodule flag. -// It holds a verbosity level and a file pattern to match. -type modulePat struct { - pattern string - literal bool // The pattern is a literal string - full bool // The pattern wants to match the full path - level Level -} - -// match reports whether the file matches the pattern. It uses a string -// comparison if the pattern contains no metacharacters. -func (m *modulePat) match(full, file string) bool { - if m.literal { - if m.full { - return full == m.pattern - } - return file == m.pattern - } - if m.full { - match, _ := filepath.Match(m.pattern, full) - return match - } - match, _ := filepath.Match(m.pattern, file) - return match -} - -// isLiteral reports whether the pattern is a literal string, that is, has no metacharacters -// that require filepath.Match to be called to match the pattern. -func isLiteral(pattern string) bool { - return !strings.ContainsAny(pattern, `\*?[]`) -} - -// isFull reports whether the pattern matches the full file path, that is, -// whether it contains /. -func isFull(pattern string) bool { - return strings.ContainsRune(pattern, '/') -} - -// verboseFlags represents the setting of the -v and -vmodule flags. -type verboseFlags struct { - // moduleLevelCache is a sync.Map storing the -vmodule Level for each V() - // call site, identified by PC. If there is no matching -vmodule filter, - // the cached value is exactly v. moduleLevelCache is replaced with a new - // Map whenever the -vmodule or -v flag changes state. - moduleLevelCache atomic.Value - - // mu guards all fields below. - mu sync.Mutex - - // v stores the value of the -v flag. It may be read safely using - // sync.LoadInt32, but is only modified under mu. - v Level - - // module stores the parsed -vmodule flag. - module []modulePat - - // moduleLength caches len(module). If greater than zero, it - // means vmodule is enabled. It may be read safely using sync.LoadInt32, but - // is only modified under mu. - moduleLength int32 -} - -// NOTE: For compatibility with the open-sourced v1 version of this -// package (github.com/golang/glog) we need to retain that flag.Level -// implements the flag.Value interface. See also go/log-vs-glog. - -// String is part of the flag.Value interface. -func (l *Level) String() string { - return strconv.FormatInt(int64(l.Get().(Level)), 10) -} - -// Get is part of the flag.Value interface. -func (l *Level) Get() any { - if l == &vflags.v { - // l is the value registered for the -v flag. - return Level(atomic.LoadInt32((*int32)(l))) - } - return *l -} - -// Set is part of the flag.Value interface. -func (l *Level) Set(value string) error { - v, err := strconv.Atoi(value) - if err != nil { - return err - } - if l == &vflags.v { - // l is the value registered for the -v flag. - vflags.mu.Lock() - defer vflags.mu.Unlock() - vflags.moduleLevelCache.Store(&sync.Map{}) - atomic.StoreInt32((*int32)(l), int32(v)) - return nil - } - *l = Level(v) - return nil -} - -// vModuleFlag is the flag.Value for the --vmodule flag. -type vModuleFlag struct{ *verboseFlags } - -func (f vModuleFlag) String() string { - // Do not panic on the zero value. - // https://groups.google.com/g/golang-nuts/c/Atlr8uAjn6U/m/iId17Td5BQAJ. - if f.verboseFlags == nil { - return "" - } - f.mu.Lock() - defer f.mu.Unlock() - - var b bytes.Buffer - for i, f := range f.module { - if i > 0 { - b.WriteRune(',') - } - fmt.Fprintf(&b, "%s=%d", f.pattern, f.level) - } - return b.String() -} - -// Get returns nil for this flag type since the struct is not exported. -func (f vModuleFlag) Get() any { return nil } - -var errVmoduleSyntax = errors.New("syntax error: expect comma-separated list of filename=N") - -// Syntax: -vmodule=recordio=2,foo/bar/baz=1,gfs*=3 -func (f vModuleFlag) Set(value string) error { - var filter []modulePat - for _, pat := range strings.Split(value, ",") { - if len(pat) == 0 { - // Empty strings such as from a trailing comma can be ignored. - continue - } - patLev := strings.Split(pat, "=") - if len(patLev) != 2 || len(patLev[0]) == 0 || len(patLev[1]) == 0 { - return errVmoduleSyntax - } - pattern := patLev[0] - v, err := strconv.Atoi(patLev[1]) - if err != nil { - return errors.New("syntax error: expect comma-separated list of filename=N") - } - // TODO: check syntax of filter? - filter = append(filter, modulePat{pattern, isLiteral(pattern), isFull(pattern), Level(v)}) - } - - f.mu.Lock() - defer f.mu.Unlock() - f.module = filter - atomic.StoreInt32((*int32)(&f.moduleLength), int32(len(f.module))) - f.moduleLevelCache.Store(&sync.Map{}) - return nil -} - -func (f *verboseFlags) levelForPC(pc uintptr) Level { - if level, ok := f.moduleLevelCache.Load().(*sync.Map).Load(pc); ok { - return level.(Level) - } - - f.mu.Lock() - defer f.mu.Unlock() - level := Level(f.v) - fn := runtime.FuncForPC(pc) - file, _ := fn.FileLine(pc) - // The file is something like /a/b/c/d.go. We want just the d for - // regular matches, /a/b/c/d for full matches. - file = strings.TrimSuffix(file, ".go") - full := file - if slash := strings.LastIndex(file, "/"); slash >= 0 { - file = file[slash+1:] - } - for _, filter := range f.module { - if filter.match(full, file) { - level = filter.level - break // Use the first matching level. - } - } - f.moduleLevelCache.Load().(*sync.Map).Store(pc, level) - return level -} - -func (f *verboseFlags) enabled(callerDepth int, level Level) bool { - if atomic.LoadInt32(&f.moduleLength) == 0 { - // No vmodule values specified, so compare against v level. - return Level(atomic.LoadInt32((*int32)(&f.v))) >= level - } - - pcs := [1]uintptr{} - if runtime.Callers(callerDepth+2, pcs[:]) < 1 { - return false - } - frame, _ := runtime.CallersFrames(pcs[:]).Next() - return f.levelForPC(frame.Entry) >= level -} - -// traceLocation represents an entry in the -log_backtrace_at flag. -type traceLocation struct { - file string - line int -} - -var errTraceSyntax = errors.New("syntax error: expect file.go:234") - -func parseTraceLocation(value string) (traceLocation, error) { - fields := strings.Split(value, ":") - if len(fields) != 2 { - return traceLocation{}, errTraceSyntax - } - file, lineStr := fields[0], fields[1] - if !strings.Contains(file, ".") { - return traceLocation{}, errTraceSyntax - } - line, err := strconv.Atoi(lineStr) - if err != nil { - return traceLocation{}, errTraceSyntax - } - if line < 0 { - return traceLocation{}, errors.New("negative value for line") - } - return traceLocation{file, line}, nil -} - -// match reports whether the specified file and line matches the trace location. -// The argument file name is the full path, not the basename specified in the flag. -func (t traceLocation) match(file string, line int) bool { - if t.line != line { - return false - } - if i := strings.LastIndex(file, "/"); i >= 0 { - file = file[i+1:] - } - return t.file == file -} - -func (t traceLocation) String() string { - return fmt.Sprintf("%s:%d", t.file, t.line) -} - -// traceLocations represents the -log_backtrace_at flag. -// Syntax: -log_backtrace_at=recordio.go:234,sstable.go:456 -// Note that unlike vmodule the file extension is included here. -type traceLocations struct { - mu sync.Mutex - locsLen int32 // Safe for atomic read without mu. - locs []traceLocation -} - -func (t *traceLocations) String() string { - t.mu.Lock() - defer t.mu.Unlock() - - var buf bytes.Buffer - for i, tl := range t.locs { - if i > 0 { - buf.WriteString(",") - } - buf.WriteString(tl.String()) - } - return buf.String() -} - -// Get always returns nil for this flag type since the struct is not exported -func (t *traceLocations) Get() any { return nil } - -func (t *traceLocations) Set(value string) error { - var locs []traceLocation - for _, s := range strings.Split(value, ",") { - if s == "" { - continue - } - loc, err := parseTraceLocation(s) - if err != nil { - return err - } - locs = append(locs, loc) - } - - t.mu.Lock() - defer t.mu.Unlock() - atomic.StoreInt32(&t.locsLen, int32(len(locs))) - t.locs = locs - return nil -} - -func (t *traceLocations) match(file string, line int) bool { - if atomic.LoadInt32(&t.locsLen) == 0 { - return false - } - - t.mu.Lock() - defer t.mu.Unlock() - for _, tl := range t.locs { - if tl.match(file, line) { - return true - } - } - return false -} - -// severityFlag is an atomic flag.Value implementation for logsink.Severity. -type severityFlag int32 - -func (s *severityFlag) get() logsink.Severity { - return logsink.Severity(atomic.LoadInt32((*int32)(s))) -} -func (s *severityFlag) String() string { return strconv.FormatInt(int64(*s), 10) } -func (s *severityFlag) Get() any { return s.get() } -func (s *severityFlag) Set(value string) error { - threshold, err := logsink.ParseSeverity(value) - if err != nil { - // Not a severity name. Try a raw number. - v, err := strconv.Atoi(value) - if err != nil { - return err - } - threshold = logsink.Severity(v) - if threshold < logsink.Info || threshold > logsink.Fatal { - return fmt.Errorf("Severity %d out of range (min %d, max %d).", v, logsink.Info, logsink.Fatal) - } - } - atomic.StoreInt32((*int32)(s), int32(threshold)) - return nil -} - -var ( - vflags verboseFlags // The -v and -vmodule flags. - - logBacktraceAt traceLocations // The -log_backtrace_at flag. - - // Boolean flags. Not handled atomically because the flag.Value interface - // does not let us avoid the =true, and that shorthand is necessary for - // compatibility. TODO: does this matter enough to fix? Seems unlikely. - toStderr bool // The -logtostderr flag. - alsoToStderr bool // The -alsologtostderr flag. - - stderrThreshold severityFlag // The -stderrthreshold flag. -) - -// verboseEnabled returns whether the caller at the given depth should emit -// verbose logs at the given level, with depth 0 identifying the caller of -// verboseEnabled. -func verboseEnabled(callerDepth int, level Level) bool { - return vflags.enabled(callerDepth+1, level) -} - -// backtraceAt returns whether the logging call at the given function and line -// should also emit a backtrace of the current call stack. -func backtraceAt(file string, line int) bool { - return logBacktraceAt.match(file, line) -} - -func init() { - vflags.moduleLevelCache.Store(&sync.Map{}) - - flag.Var(&vflags.v, "v", "log level for V logs") - flag.Var(vModuleFlag{&vflags}, "vmodule", "comma-separated list of pattern=N settings for file-filtered logging") - - flag.Var(&logBacktraceAt, "log_backtrace_at", "when logging hits line file:N, emit a stack trace") - - stderrThreshold = severityFlag(logsink.Error) - - flag.BoolVar(&toStderr, "logtostderr", false, "log to standard error instead of files") - flag.BoolVar(&alsoToStderr, "alsologtostderr", false, "log to standard error as well as files") - flag.Var(&stderrThreshold, "stderrthreshold", "logs at or above this threshold go to stderr") -} diff --git a/vendor/github.com/golang/glog/internal/logsink/logsink.go b/vendor/github.com/golang/glog/internal/logsink/logsink.go deleted file mode 100644 index 28c38a6ab..000000000 --- a/vendor/github.com/golang/glog/internal/logsink/logsink.go +++ /dev/null @@ -1,393 +0,0 @@ -// Copyright 2023 Google Inc. All Rights Reserved. -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -package logsink - -import ( - "bytes" - "context" - "fmt" - "strconv" - "strings" - "sync" - "time" - - "github.com/golang/glog/internal/stackdump" -) - -// MaxLogMessageLen is the limit on length of a formatted log message, including -// the standard line prefix and trailing newline. -// -// Chosen to match C++ glog. -const MaxLogMessageLen = 15000 - -// A Severity is a severity at which a message can be logged. -type Severity int8 - -// These constants identify the log levels in order of increasing severity. -// A message written to a high-severity log file is also written to each -// lower-severity log file. -const ( - Info Severity = iota - Warning - Error - - // Fatal contains logs written immediately before the process terminates. - // - // Sink implementations should not terminate the process themselves: the log - // package will perform any necessary cleanup and terminate the process as - // appropriate. - Fatal -) - -func (s Severity) String() string { - switch s { - case Info: - return "INFO" - case Warning: - return "WARNING" - case Error: - return "ERROR" - case Fatal: - return "FATAL" - } - return fmt.Sprintf("%T(%d)", s, s) -} - -// ParseSeverity returns the case-insensitive Severity value for the given string. -func ParseSeverity(name string) (Severity, error) { - name = strings.ToUpper(name) - for s := Info; s <= Fatal; s++ { - if s.String() == name { - return s, nil - } - } - return -1, fmt.Errorf("logsink: invalid severity %q", name) -} - -// Meta is metadata about a logging call. -type Meta struct { - // The context with which the log call was made (or nil). If set, the context - // is only valid during the logsink.Structured.Printf call, it should not be - // retained. - Context context.Context - - // Time is the time at which the log call was made. - Time time.Time - - // File is the source file from which the log entry originates. - File string - // Line is the line offset within the source file. - Line int - // Depth is the number of stack frames between the logsink and the log call. - Depth int - - Severity Severity - - // Verbose indicates whether the call was made via "log.V". Log entries below - // the current verbosity threshold are not sent to the sink. - Verbose bool - - // Thread ID. This can be populated with a thread ID from another source, - // such as a system we are importing logs from. In the normal case, this - // will be set to the process ID (PID), since Go doesn't have threads. - Thread int64 - - // Stack trace starting in the logging function. May be nil. - // A logsink should implement the StackWanter interface to request this. - // - // Even if WantStack returns false, this field may be set (e.g. if another - // sink wants a stack trace). - Stack *stackdump.Stack -} - -// Structured is a logging destination that accepts structured data as input. -type Structured interface { - // Printf formats according to a fmt.Printf format specifier and writes a log - // entry. The precise result of formatting depends on the sink, but should - // aim for consistency with fmt.Printf. - // - // Printf returns the number of bytes occupied by the log entry, which - // may not be equal to the total number of bytes written. - // - // Printf returns any error encountered *if* it is severe enough that the log - // package should terminate the process. - // - // The sink must not modify the *Meta parameter, nor reference it after - // Printf has returned: it may be reused in subsequent calls. - Printf(meta *Meta, format string, a ...any) (n int, err error) -} - -// StackWanter can be implemented by a logsink.Structured to indicate that it -// wants a stack trace to accompany at least some of the log messages it receives. -type StackWanter interface { - // WantStack returns true if the sink requires a stack trace for a log message - // with this metadata. - // - // NOTE: Returning true implies that meta.Stack will be non-nil. Returning - // false does NOT imply that meta.Stack will be nil. - WantStack(meta *Meta) bool -} - -// Text is a logging destination that accepts pre-formatted log lines (instead of -// structured data). -type Text interface { - // Enabled returns whether this sink should output messages for the given - // Meta. If the sink returns false for a given Meta, the Printf function will - // not call Emit on it for the corresponding log message. - Enabled(*Meta) bool - - // Emit writes a pre-formatted text log entry (including any applicable - // header) to the log. It returns the number of bytes occupied by the entry - // (which may differ from the length of the passed-in slice). - // - // Emit returns any error encountered *if* it is severe enough that the log - // package should terminate the process. - // - // The sink must not modify the *Meta parameter, nor reference it after - // Printf has returned: it may be reused in subsequent calls. - // - // NOTE: When developing a text sink, keep in mind the surface in which the - // logs will be displayed, and whether it's important that the sink be - // resistent to tampering in the style of b/211428300. Standard text sinks - // (like `stderrSink`) do not protect against this (e.g. by escaping - // characters) because the cases where they would show user-influenced bytes - // are vanishingly small. - Emit(*Meta, []byte) (n int, err error) -} - -// bufs is a pool of *bytes.Buffer used in formatting log entries. -var bufs sync.Pool // Pool of *bytes.Buffer. - -// textPrintf formats a text log entry and emits it to all specified Text sinks. -// -// The returned n is the maximum across all Emit calls. -// The returned err is the first non-nil error encountered. -// Sinks that are disabled by configuration should return (0, nil). -func textPrintf(m *Meta, textSinks []Text, format string, args ...any) (n int, err error) { - // We expect at most file, stderr, and perhaps syslog. If there are more, - // we'll end up allocating - no big deal. - const maxExpectedTextSinks = 3 - var noAllocSinks [maxExpectedTextSinks]Text - - sinks := noAllocSinks[:0] - for _, s := range textSinks { - if s.Enabled(m) { - sinks = append(sinks, s) - } - } - if len(sinks) == 0 && m.Severity != Fatal { - return 0, nil // No TextSinks specified; don't bother formatting. - } - - bufi := bufs.Get() - var buf *bytes.Buffer - if bufi == nil { - buf = bytes.NewBuffer(nil) - bufi = buf - } else { - buf = bufi.(*bytes.Buffer) - buf.Reset() - } - - // Lmmdd hh:mm:ss.uuuuuu PID/GID file:line] - // - // The "PID" entry arguably ought to be TID for consistency with other - // environments, but TID is not meaningful in a Go program due to the - // multiplexing of goroutines across threads. - // - // Avoid Fprintf, for speed. The format is so simple that we can do it quickly by hand. - // It's worth about 3X. Fprintf is hard. - const severityChar = "IWEF" - buf.WriteByte(severityChar[m.Severity]) - - _, month, day := m.Time.Date() - hour, minute, second := m.Time.Clock() - twoDigits(buf, int(month)) - twoDigits(buf, day) - buf.WriteByte(' ') - twoDigits(buf, hour) - buf.WriteByte(':') - twoDigits(buf, minute) - buf.WriteByte(':') - twoDigits(buf, second) - buf.WriteByte('.') - nDigits(buf, 6, uint64(m.Time.Nanosecond()/1000), '0') - buf.WriteByte(' ') - - nDigits(buf, 7, uint64(m.Thread), ' ') - buf.WriteByte(' ') - - { - file := m.File - if i := strings.LastIndex(file, "/"); i >= 0 { - file = file[i+1:] - } - buf.WriteString(file) - } - - buf.WriteByte(':') - { - var tmp [19]byte - buf.Write(strconv.AppendInt(tmp[:0], int64(m.Line), 10)) - } - buf.WriteString("] ") - - msgStart := buf.Len() - fmt.Fprintf(buf, format, args...) - if buf.Len() > MaxLogMessageLen-1 { - buf.Truncate(MaxLogMessageLen - 1) - } - msgEnd := buf.Len() - if b := buf.Bytes(); b[len(b)-1] != '\n' { - buf.WriteByte('\n') - } - - for _, s := range sinks { - sn, sErr := s.Emit(m, buf.Bytes()) - if sn > n { - n = sn - } - if sErr != nil && err == nil { - err = sErr - } - } - - if m.Severity == Fatal { - savedM := *m - fatalMessageStore(savedEntry{ - meta: &savedM, - msg: buf.Bytes()[msgStart:msgEnd], - }) - } else { - bufs.Put(bufi) - } - return n, err -} - -const digits = "0123456789" - -// twoDigits formats a zero-prefixed two-digit integer to buf. -func twoDigits(buf *bytes.Buffer, d int) { - buf.WriteByte(digits[(d/10)%10]) - buf.WriteByte(digits[d%10]) -} - -// nDigits formats an n-digit integer to buf, padding with pad on the left. It -// assumes d != 0. -func nDigits(buf *bytes.Buffer, n int, d uint64, pad byte) { - var tmp [20]byte - - cutoff := len(tmp) - n - j := len(tmp) - 1 - for ; d > 0; j-- { - tmp[j] = digits[d%10] - d /= 10 - } - for ; j >= cutoff; j-- { - tmp[j] = pad - } - j++ - buf.Write(tmp[j:]) -} - -// Printf writes a log entry to all registered TextSinks in this package, then -// to all registered StructuredSinks. -// -// The returned n is the maximum across all Emit and Printf calls. -// The returned err is the first non-nil error encountered. -// Sinks that are disabled by configuration should return (0, nil). -func Printf(m *Meta, format string, args ...any) (n int, err error) { - m.Depth++ - n, err = textPrintf(m, TextSinks, format, args...) - - for _, sink := range StructuredSinks { - // TODO: Support TextSinks that implement StackWanter? - if sw, ok := sink.(StackWanter); ok && sw.WantStack(m) { - if m.Stack == nil { - // First, try to find a stacktrace in args, otherwise generate one. - for _, arg := range args { - if stack, ok := arg.(stackdump.Stack); ok { - m.Stack = &stack - break - } - } - if m.Stack == nil { - stack := stackdump.Caller( /* skipDepth = */ m.Depth) - m.Stack = &stack - } - } - } - sn, sErr := sink.Printf(m, format, args...) - if sn > n { - n = sn - } - if sErr != nil && err == nil { - err = sErr - } - } - return n, err -} - -// The sets of sinks to which logs should be written. -// -// These must only be modified during package init, and are read-only thereafter. -var ( - // StructuredSinks is the set of Structured sink instances to which logs - // should be written. - StructuredSinks []Structured - - // TextSinks is the set of Text sink instances to which logs should be - // written. - // - // These are registered separately from Structured sink implementations to - // avoid the need to repeat the work of formatting a message for each Text - // sink that writes it. The package-level Printf function writes to both sets - // independenty, so a given log destination should only register a Structured - // *or* a Text sink (not both). - TextSinks []Text -) - -type savedEntry struct { - meta *Meta - msg []byte -} - -// StructuredTextWrapper is a Structured sink which forwards logs to a set of Text sinks. -// -// The purpose of this sink is to allow applications to intercept logging calls before they are -// serialized and sent to Text sinks. For example, if one needs to redact PII from logging -// arguments before they reach STDERR, one solution would be to do the redacting in a Structured -// sink that forwards logs to a StructuredTextWrapper instance, and make STDERR a child of that -// StructuredTextWrapper instance. This is how one could set this up in their application: -// -// func init() { -// -// wrapper := logsink.StructuredTextWrapper{TextSinks: logsink.TextSinks} -// // sanitizersink will intercept logs and remove PII -// sanitizer := sanitizersink{Sink: &wrapper} -// logsink.StructuredSinks = append(logsink.StructuredSinks, &sanitizer) -// logsink.TextSinks = nil -// -// } -type StructuredTextWrapper struct { - // TextSinks is the set of Text sinks that should receive logs from this - // StructuredTextWrapper instance. - TextSinks []Text -} - -// Printf forwards logs to all Text sinks registered in the StructuredTextWrapper. -func (w *StructuredTextWrapper) Printf(meta *Meta, format string, args ...any) (n int, err error) { - return textPrintf(meta, w.TextSinks, format, args...) -} diff --git a/vendor/github.com/golang/glog/internal/logsink/logsink_fatal.go b/vendor/github.com/golang/glog/internal/logsink/logsink_fatal.go deleted file mode 100644 index 3dc269abc..000000000 --- a/vendor/github.com/golang/glog/internal/logsink/logsink_fatal.go +++ /dev/null @@ -1,35 +0,0 @@ -package logsink - -import ( - "sync/atomic" - "unsafe" -) - -func fatalMessageStore(e savedEntry) { - // Only put a new one in if we haven't assigned before. - atomic.CompareAndSwapPointer(&fatalMessage, nil, unsafe.Pointer(&e)) -} - -var fatalMessage unsafe.Pointer // savedEntry stored with CompareAndSwapPointer - -// FatalMessage returns the Meta and message contents of the first message -// logged with Fatal severity, or false if none has occurred. -func FatalMessage() (*Meta, []byte, bool) { - e := (*savedEntry)(atomic.LoadPointer(&fatalMessage)) - if e == nil { - return nil, nil, false - } - return e.meta, e.msg, true -} - -// DoNotUseRacyFatalMessage is FatalMessage, but worse. -// -//go:norace -//go:nosplit -func DoNotUseRacyFatalMessage() (*Meta, []byte, bool) { - e := (*savedEntry)(fatalMessage) - if e == nil { - return nil, nil, false - } - return e.meta, e.msg, true -} diff --git a/vendor/github.com/golang/glog/internal/stackdump/stackdump.go b/vendor/github.com/golang/glog/internal/stackdump/stackdump.go deleted file mode 100644 index 3427c9d6b..000000000 --- a/vendor/github.com/golang/glog/internal/stackdump/stackdump.go +++ /dev/null @@ -1,127 +0,0 @@ -// Copyright 2023 Google Inc. All Rights Reserved. -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -// Package stackdump provides wrappers for runtime.Stack and runtime.Callers -// with uniform support for skipping caller frames. -// -// ⚠ Unlike the functions in the runtime package, these may allocate a -// non-trivial quantity of memory: use them with care. ⚠ -package stackdump - -import ( - "bytes" - "runtime" -) - -// runtimeStackSelfFrames is 1 if runtime.Stack includes the call to -// runtime.Stack itself or 0 if it does not. -// -// As of 2016-04-27, the gccgo compiler includes runtime.Stack but the gc -// compiler does not. -var runtimeStackSelfFrames = func() int { - for n := 1 << 10; n < 1<<20; n *= 2 { - buf := make([]byte, n) - n := runtime.Stack(buf, false) - if bytes.Contains(buf[:n], []byte("runtime.Stack")) { - return 1 - } else if n < len(buf) || bytes.Count(buf, []byte("\n")) >= 3 { - return 0 - } - } - return 0 -}() - -// Stack is a stack dump for a single goroutine. -type Stack struct { - // Text is a representation of the stack dump in a human-readable format. - Text []byte - - // PC is a representation of the stack dump using raw program counter values. - PC []uintptr -} - -func (s Stack) String() string { return string(s.Text) } - -// Caller returns the Stack dump for the calling goroutine, starting skipDepth -// frames before the caller of Caller. (Caller(0) provides a dump starting at -// the caller of this function.) -func Caller(skipDepth int) Stack { - return Stack{ - Text: CallerText(skipDepth + 1), - PC: CallerPC(skipDepth + 1), - } -} - -// CallerText returns a textual dump of the stack starting skipDepth frames before -// the caller. (CallerText(0) provides a dump starting at the caller of this -// function.) -func CallerText(skipDepth int) []byte { - for n := 1 << 10; ; n *= 2 { - buf := make([]byte, n) - n := runtime.Stack(buf, false) - if n < len(buf) { - return pruneFrames(skipDepth+1+runtimeStackSelfFrames, buf[:n]) - } - } -} - -// CallerPC returns a dump of the program counters of the stack starting -// skipDepth frames before the caller. (CallerPC(0) provides a dump starting at -// the caller of this function.) -func CallerPC(skipDepth int) []uintptr { - for n := 1 << 8; ; n *= 2 { - buf := make([]uintptr, n) - n := runtime.Callers(skipDepth+2, buf) - if n < len(buf) { - return buf[:n] - } - } -} - -// pruneFrames removes the topmost skipDepth frames of the first goroutine in a -// textual stack dump. It overwrites the passed-in slice. -// -// If there are fewer than skipDepth frames in the first goroutine's stack, -// pruneFrames prunes it to an empty stack and leaves the remaining contents -// intact. -func pruneFrames(skipDepth int, stack []byte) []byte { - headerLen := 0 - for i, c := range stack { - if c == '\n' { - headerLen = i + 1 - break - } - } - if headerLen == 0 { - return stack // No header line - not a well-formed stack trace. - } - - skipLen := headerLen - skipNewlines := skipDepth * 2 - for ; skipLen < len(stack) && skipNewlines > 0; skipLen++ { - c := stack[skipLen] - if c != '\n' { - continue - } - skipNewlines-- - skipLen++ - if skipNewlines == 0 || skipLen == len(stack) || stack[skipLen] == '\n' { - break - } - } - - pruned := stack[skipLen-headerLen:] - copy(pruned, stack[:headerLen]) - return pruned -} diff --git a/vendor/github.com/golang/snappy/README b/vendor/github.com/golang/snappy/README deleted file mode 100644 index cea12879a..000000000 --- a/vendor/github.com/golang/snappy/README +++ /dev/null @@ -1,107 +0,0 @@ -The Snappy compression format in the Go programming language. - -To download and install from source: -$ go get github.com/golang/snappy - -Unless otherwise noted, the Snappy-Go source files are distributed -under the BSD-style license found in the LICENSE file. - - - -Benchmarks. - -The golang/snappy benchmarks include compressing (Z) and decompressing (U) ten -or so files, the same set used by the C++ Snappy code (github.com/google/snappy -and note the "google", not "golang"). On an "Intel(R) Core(TM) i7-3770 CPU @ -3.40GHz", Go's GOARCH=amd64 numbers as of 2016-05-29: - -"go test -test.bench=." - -_UFlat0-8 2.19GB/s ± 0% html -_UFlat1-8 1.41GB/s ± 0% urls -_UFlat2-8 23.5GB/s ± 2% jpg -_UFlat3-8 1.91GB/s ± 0% jpg_200 -_UFlat4-8 14.0GB/s ± 1% pdf -_UFlat5-8 1.97GB/s ± 0% html4 -_UFlat6-8 814MB/s ± 0% txt1 -_UFlat7-8 785MB/s ± 0% txt2 -_UFlat8-8 857MB/s ± 0% txt3 -_UFlat9-8 719MB/s ± 1% txt4 -_UFlat10-8 2.84GB/s ± 0% pb -_UFlat11-8 1.05GB/s ± 0% gaviota - -_ZFlat0-8 1.04GB/s ± 0% html -_ZFlat1-8 534MB/s ± 0% urls -_ZFlat2-8 15.7GB/s ± 1% jpg -_ZFlat3-8 740MB/s ± 3% jpg_200 -_ZFlat4-8 9.20GB/s ± 1% pdf -_ZFlat5-8 991MB/s ± 0% html4 -_ZFlat6-8 379MB/s ± 0% txt1 -_ZFlat7-8 352MB/s ± 0% txt2 -_ZFlat8-8 396MB/s ± 1% txt3 -_ZFlat9-8 327MB/s ± 1% txt4 -_ZFlat10-8 1.33GB/s ± 1% pb -_ZFlat11-8 605MB/s ± 1% gaviota - - - -"go test -test.bench=. -tags=noasm" - -_UFlat0-8 621MB/s ± 2% html -_UFlat1-8 494MB/s ± 1% urls -_UFlat2-8 23.2GB/s ± 1% jpg -_UFlat3-8 1.12GB/s ± 1% jpg_200 -_UFlat4-8 4.35GB/s ± 1% pdf -_UFlat5-8 609MB/s ± 0% html4 -_UFlat6-8 296MB/s ± 0% txt1 -_UFlat7-8 288MB/s ± 0% txt2 -_UFlat8-8 309MB/s ± 1% txt3 -_UFlat9-8 280MB/s ± 1% txt4 -_UFlat10-8 753MB/s ± 0% pb -_UFlat11-8 400MB/s ± 0% gaviota - -_ZFlat0-8 409MB/s ± 1% html -_ZFlat1-8 250MB/s ± 1% urls -_ZFlat2-8 12.3GB/s ± 1% jpg -_ZFlat3-8 132MB/s ± 0% jpg_200 -_ZFlat4-8 2.92GB/s ± 0% pdf -_ZFlat5-8 405MB/s ± 1% html4 -_ZFlat6-8 179MB/s ± 1% txt1 -_ZFlat7-8 170MB/s ± 1% txt2 -_ZFlat8-8 189MB/s ± 1% txt3 -_ZFlat9-8 164MB/s ± 1% txt4 -_ZFlat10-8 479MB/s ± 1% pb -_ZFlat11-8 270MB/s ± 1% gaviota - - - -For comparison (Go's encoded output is byte-for-byte identical to C++'s), here -are the numbers from C++ Snappy's - -make CXXFLAGS="-O2 -DNDEBUG -g" clean snappy_unittest.log && cat snappy_unittest.log - -BM_UFlat/0 2.4GB/s html -BM_UFlat/1 1.4GB/s urls -BM_UFlat/2 21.8GB/s jpg -BM_UFlat/3 1.5GB/s jpg_200 -BM_UFlat/4 13.3GB/s pdf -BM_UFlat/5 2.1GB/s html4 -BM_UFlat/6 1.0GB/s txt1 -BM_UFlat/7 959.4MB/s txt2 -BM_UFlat/8 1.0GB/s txt3 -BM_UFlat/9 864.5MB/s txt4 -BM_UFlat/10 2.9GB/s pb -BM_UFlat/11 1.2GB/s gaviota - -BM_ZFlat/0 944.3MB/s html (22.31 %) -BM_ZFlat/1 501.6MB/s urls (47.78 %) -BM_ZFlat/2 14.3GB/s jpg (99.95 %) -BM_ZFlat/3 538.3MB/s jpg_200 (73.00 %) -BM_ZFlat/4 8.3GB/s pdf (83.30 %) -BM_ZFlat/5 903.5MB/s html4 (22.52 %) -BM_ZFlat/6 336.0MB/s txt1 (57.88 %) -BM_ZFlat/7 312.3MB/s txt2 (61.91 %) -BM_ZFlat/8 353.1MB/s txt3 (54.99 %) -BM_ZFlat/9 289.9MB/s txt4 (66.26 %) -BM_ZFlat/10 1.2GB/s pb (19.68 %) -BM_ZFlat/11 527.4MB/s gaviota (37.72 %) diff --git a/vendor/github.com/golang/snappy/decode.go b/vendor/github.com/golang/snappy/decode.go deleted file mode 100644 index 23c6e26c6..000000000 --- a/vendor/github.com/golang/snappy/decode.go +++ /dev/null @@ -1,264 +0,0 @@ -// Copyright 2011 The Snappy-Go Authors. All rights reserved. -// Use of this source code is governed by a BSD-style -// license that can be found in the LICENSE file. - -package snappy - -import ( - "encoding/binary" - "errors" - "io" -) - -var ( - // ErrCorrupt reports that the input is invalid. - ErrCorrupt = errors.New("snappy: corrupt input") - // ErrTooLarge reports that the uncompressed length is too large. - ErrTooLarge = errors.New("snappy: decoded block is too large") - // ErrUnsupported reports that the input isn't supported. - ErrUnsupported = errors.New("snappy: unsupported input") - - errUnsupportedLiteralLength = errors.New("snappy: unsupported literal length") -) - -// DecodedLen returns the length of the decoded block. -func DecodedLen(src []byte) (int, error) { - v, _, err := decodedLen(src) - return v, err -} - -// decodedLen returns the length of the decoded block and the number of bytes -// that the length header occupied. -func decodedLen(src []byte) (blockLen, headerLen int, err error) { - v, n := binary.Uvarint(src) - if n <= 0 || v > 0xffffffff { - return 0, 0, ErrCorrupt - } - - const wordSize = 32 << (^uint(0) >> 32 & 1) - if wordSize == 32 && v > 0x7fffffff { - return 0, 0, ErrTooLarge - } - return int(v), n, nil -} - -const ( - decodeErrCodeCorrupt = 1 - decodeErrCodeUnsupportedLiteralLength = 2 -) - -// Decode returns the decoded form of src. The returned slice may be a sub- -// slice of dst if dst was large enough to hold the entire decoded block. -// Otherwise, a newly allocated slice will be returned. -// -// The dst and src must not overlap. It is valid to pass a nil dst. -// -// Decode handles the Snappy block format, not the Snappy stream format. -func Decode(dst, src []byte) ([]byte, error) { - dLen, s, err := decodedLen(src) - if err != nil { - return nil, err - } - if dLen <= len(dst) { - dst = dst[:dLen] - } else { - dst = make([]byte, dLen) - } - switch decode(dst, src[s:]) { - case 0: - return dst, nil - case decodeErrCodeUnsupportedLiteralLength: - return nil, errUnsupportedLiteralLength - } - return nil, ErrCorrupt -} - -// NewReader returns a new Reader that decompresses from r, using the framing -// format described at -// https://github.com/google/snappy/blob/master/framing_format.txt -func NewReader(r io.Reader) *Reader { - return &Reader{ - r: r, - decoded: make([]byte, maxBlockSize), - buf: make([]byte, maxEncodedLenOfMaxBlockSize+checksumSize), - } -} - -// Reader is an io.Reader that can read Snappy-compressed bytes. -// -// Reader handles the Snappy stream format, not the Snappy block format. -type Reader struct { - r io.Reader - err error - decoded []byte - buf []byte - // decoded[i:j] contains decoded bytes that have not yet been passed on. - i, j int - readHeader bool -} - -// Reset discards any buffered data, resets all state, and switches the Snappy -// reader to read from r. This permits reusing a Reader rather than allocating -// a new one. -func (r *Reader) Reset(reader io.Reader) { - r.r = reader - r.err = nil - r.i = 0 - r.j = 0 - r.readHeader = false -} - -func (r *Reader) readFull(p []byte, allowEOF bool) (ok bool) { - if _, r.err = io.ReadFull(r.r, p); r.err != nil { - if r.err == io.ErrUnexpectedEOF || (r.err == io.EOF && !allowEOF) { - r.err = ErrCorrupt - } - return false - } - return true -} - -func (r *Reader) fill() error { - for r.i >= r.j { - if !r.readFull(r.buf[:4], true) { - return r.err - } - chunkType := r.buf[0] - if !r.readHeader { - if chunkType != chunkTypeStreamIdentifier { - r.err = ErrCorrupt - return r.err - } - r.readHeader = true - } - chunkLen := int(r.buf[1]) | int(r.buf[2])<<8 | int(r.buf[3])<<16 - if chunkLen > len(r.buf) { - r.err = ErrUnsupported - return r.err - } - - // The chunk types are specified at - // https://github.com/google/snappy/blob/master/framing_format.txt - switch chunkType { - case chunkTypeCompressedData: - // Section 4.2. Compressed data (chunk type 0x00). - if chunkLen < checksumSize { - r.err = ErrCorrupt - return r.err - } - buf := r.buf[:chunkLen] - if !r.readFull(buf, false) { - return r.err - } - checksum := uint32(buf[0]) | uint32(buf[1])<<8 | uint32(buf[2])<<16 | uint32(buf[3])<<24 - buf = buf[checksumSize:] - - n, err := DecodedLen(buf) - if err != nil { - r.err = err - return r.err - } - if n > len(r.decoded) { - r.err = ErrCorrupt - return r.err - } - if _, err := Decode(r.decoded, buf); err != nil { - r.err = err - return r.err - } - if crc(r.decoded[:n]) != checksum { - r.err = ErrCorrupt - return r.err - } - r.i, r.j = 0, n - continue - - case chunkTypeUncompressedData: - // Section 4.3. Uncompressed data (chunk type 0x01). - if chunkLen < checksumSize { - r.err = ErrCorrupt - return r.err - } - buf := r.buf[:checksumSize] - if !r.readFull(buf, false) { - return r.err - } - checksum := uint32(buf[0]) | uint32(buf[1])<<8 | uint32(buf[2])<<16 | uint32(buf[3])<<24 - // Read directly into r.decoded instead of via r.buf. - n := chunkLen - checksumSize - if n > len(r.decoded) { - r.err = ErrCorrupt - return r.err - } - if !r.readFull(r.decoded[:n], false) { - return r.err - } - if crc(r.decoded[:n]) != checksum { - r.err = ErrCorrupt - return r.err - } - r.i, r.j = 0, n - continue - - case chunkTypeStreamIdentifier: - // Section 4.1. Stream identifier (chunk type 0xff). - if chunkLen != len(magicBody) { - r.err = ErrCorrupt - return r.err - } - if !r.readFull(r.buf[:len(magicBody)], false) { - return r.err - } - for i := 0; i < len(magicBody); i++ { - if r.buf[i] != magicBody[i] { - r.err = ErrCorrupt - return r.err - } - } - continue - } - - if chunkType <= 0x7f { - // Section 4.5. Reserved unskippable chunks (chunk types 0x02-0x7f). - r.err = ErrUnsupported - return r.err - } - // Section 4.4 Padding (chunk type 0xfe). - // Section 4.6. Reserved skippable chunks (chunk types 0x80-0xfd). - if !r.readFull(r.buf[:chunkLen], false) { - return r.err - } - } - - return nil -} - -// Read satisfies the io.Reader interface. -func (r *Reader) Read(p []byte) (int, error) { - if r.err != nil { - return 0, r.err - } - - if err := r.fill(); err != nil { - return 0, err - } - - n := copy(p, r.decoded[r.i:r.j]) - r.i += n - return n, nil -} - -// ReadByte satisfies the io.ByteReader interface. -func (r *Reader) ReadByte() (byte, error) { - if r.err != nil { - return 0, r.err - } - - if err := r.fill(); err != nil { - return 0, err - } - - c := r.decoded[r.i] - r.i++ - return c, nil -} diff --git a/vendor/github.com/golang/snappy/decode_other.go b/vendor/github.com/golang/snappy/decode_other.go deleted file mode 100644 index 2f672be55..000000000 --- a/vendor/github.com/golang/snappy/decode_other.go +++ /dev/null @@ -1,115 +0,0 @@ -// Copyright 2016 The Snappy-Go Authors. All rights reserved. -// Use of this source code is governed by a BSD-style -// license that can be found in the LICENSE file. - -// +build !amd64,!arm64 appengine !gc noasm - -package snappy - -// decode writes the decoding of src to dst. It assumes that the varint-encoded -// length of the decompressed bytes has already been read, and that len(dst) -// equals that length. -// -// It returns 0 on success or a decodeErrCodeXxx error code on failure. -func decode(dst, src []byte) int { - var d, s, offset, length int - for s < len(src) { - switch src[s] & 0x03 { - case tagLiteral: - x := uint32(src[s] >> 2) - switch { - case x < 60: - s++ - case x == 60: - s += 2 - if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line. - return decodeErrCodeCorrupt - } - x = uint32(src[s-1]) - case x == 61: - s += 3 - if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line. - return decodeErrCodeCorrupt - } - x = uint32(src[s-2]) | uint32(src[s-1])<<8 - case x == 62: - s += 4 - if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line. - return decodeErrCodeCorrupt - } - x = uint32(src[s-3]) | uint32(src[s-2])<<8 | uint32(src[s-1])<<16 - case x == 63: - s += 5 - if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line. - return decodeErrCodeCorrupt - } - x = uint32(src[s-4]) | uint32(src[s-3])<<8 | uint32(src[s-2])<<16 | uint32(src[s-1])<<24 - } - length = int(x) + 1 - if length <= 0 { - return decodeErrCodeUnsupportedLiteralLength - } - if length > len(dst)-d || length > len(src)-s { - return decodeErrCodeCorrupt - } - copy(dst[d:], src[s:s+length]) - d += length - s += length - continue - - case tagCopy1: - s += 2 - if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line. - return decodeErrCodeCorrupt - } - length = 4 + int(src[s-2])>>2&0x7 - offset = int(uint32(src[s-2])&0xe0<<3 | uint32(src[s-1])) - - case tagCopy2: - s += 3 - if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line. - return decodeErrCodeCorrupt - } - length = 1 + int(src[s-3])>>2 - offset = int(uint32(src[s-2]) | uint32(src[s-1])<<8) - - case tagCopy4: - s += 5 - if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line. - return decodeErrCodeCorrupt - } - length = 1 + int(src[s-5])>>2 - offset = int(uint32(src[s-4]) | uint32(src[s-3])<<8 | uint32(src[s-2])<<16 | uint32(src[s-1])<<24) - } - - if offset <= 0 || d < offset || length > len(dst)-d { - return decodeErrCodeCorrupt - } - // Copy from an earlier sub-slice of dst to a later sub-slice. - // If no overlap, use the built-in copy: - if offset >= length { - copy(dst[d:d+length], dst[d-offset:]) - d += length - continue - } - - // Unlike the built-in copy function, this byte-by-byte copy always runs - // forwards, even if the slices overlap. Conceptually, this is: - // - // d += forwardCopy(dst[d:d+length], dst[d-offset:]) - // - // We align the slices into a and b and show the compiler they are the same size. - // This allows the loop to run without bounds checks. - a := dst[d : d+length] - b := dst[d-offset:] - b = b[:len(a)] - for i := range a { - a[i] = b[i] - } - d += length - } - if d != len(dst) { - return decodeErrCodeCorrupt - } - return 0 -} diff --git a/vendor/github.com/golang/snappy/encode.go b/vendor/github.com/golang/snappy/encode.go deleted file mode 100644 index 7f2365707..000000000 --- a/vendor/github.com/golang/snappy/encode.go +++ /dev/null @@ -1,289 +0,0 @@ -// Copyright 2011 The Snappy-Go Authors. All rights reserved. -// Use of this source code is governed by a BSD-style -// license that can be found in the LICENSE file. - -package snappy - -import ( - "encoding/binary" - "errors" - "io" -) - -// Encode returns the encoded form of src. The returned slice may be a sub- -// slice of dst if dst was large enough to hold the entire encoded block. -// Otherwise, a newly allocated slice will be returned. -// -// The dst and src must not overlap. It is valid to pass a nil dst. -// -// Encode handles the Snappy block format, not the Snappy stream format. -func Encode(dst, src []byte) []byte { - if n := MaxEncodedLen(len(src)); n < 0 { - panic(ErrTooLarge) - } else if len(dst) < n { - dst = make([]byte, n) - } - - // The block starts with the varint-encoded length of the decompressed bytes. - d := binary.PutUvarint(dst, uint64(len(src))) - - for len(src) > 0 { - p := src - src = nil - if len(p) > maxBlockSize { - p, src = p[:maxBlockSize], p[maxBlockSize:] - } - if len(p) < minNonLiteralBlockSize { - d += emitLiteral(dst[d:], p) - } else { - d += encodeBlock(dst[d:], p) - } - } - return dst[:d] -} - -// inputMargin is the minimum number of extra input bytes to keep, inside -// encodeBlock's inner loop. On some architectures, this margin lets us -// implement a fast path for emitLiteral, where the copy of short (<= 16 byte) -// literals can be implemented as a single load to and store from a 16-byte -// register. That literal's actual length can be as short as 1 byte, so this -// can copy up to 15 bytes too much, but that's OK as subsequent iterations of -// the encoding loop will fix up the copy overrun, and this inputMargin ensures -// that we don't overrun the dst and src buffers. -const inputMargin = 16 - 1 - -// minNonLiteralBlockSize is the minimum size of the input to encodeBlock that -// could be encoded with a copy tag. This is the minimum with respect to the -// algorithm used by encodeBlock, not a minimum enforced by the file format. -// -// The encoded output must start with at least a 1 byte literal, as there are -// no previous bytes to copy. A minimal (1 byte) copy after that, generated -// from an emitCopy call in encodeBlock's main loop, would require at least -// another inputMargin bytes, for the reason above: we want any emitLiteral -// calls inside encodeBlock's main loop to use the fast path if possible, which -// requires being able to overrun by inputMargin bytes. Thus, -// minNonLiteralBlockSize equals 1 + 1 + inputMargin. -// -// The C++ code doesn't use this exact threshold, but it could, as discussed at -// https://groups.google.com/d/topic/snappy-compression/oGbhsdIJSJ8/discussion -// The difference between Go (2+inputMargin) and C++ (inputMargin) is purely an -// optimization. It should not affect the encoded form. This is tested by -// TestSameEncodingAsCppShortCopies. -const minNonLiteralBlockSize = 1 + 1 + inputMargin - -// MaxEncodedLen returns the maximum length of a snappy block, given its -// uncompressed length. -// -// It will return a negative value if srcLen is too large to encode. -func MaxEncodedLen(srcLen int) int { - n := uint64(srcLen) - if n > 0xffffffff { - return -1 - } - // Compressed data can be defined as: - // compressed := item* literal* - // item := literal* copy - // - // The trailing literal sequence has a space blowup of at most 62/60 - // since a literal of length 60 needs one tag byte + one extra byte - // for length information. - // - // Item blowup is trickier to measure. Suppose the "copy" op copies - // 4 bytes of data. Because of a special check in the encoding code, - // we produce a 4-byte copy only if the offset is < 65536. Therefore - // the copy op takes 3 bytes to encode, and this type of item leads - // to at most the 62/60 blowup for representing literals. - // - // Suppose the "copy" op copies 5 bytes of data. If the offset is big - // enough, it will take 5 bytes to encode the copy op. Therefore the - // worst case here is a one-byte literal followed by a five-byte copy. - // That is, 6 bytes of input turn into 7 bytes of "compressed" data. - // - // This last factor dominates the blowup, so the final estimate is: - n = 32 + n + n/6 - if n > 0xffffffff { - return -1 - } - return int(n) -} - -var errClosed = errors.New("snappy: Writer is closed") - -// NewWriter returns a new Writer that compresses to w. -// -// The Writer returned does not buffer writes. There is no need to Flush or -// Close such a Writer. -// -// Deprecated: the Writer returned is not suitable for many small writes, only -// for few large writes. Use NewBufferedWriter instead, which is efficient -// regardless of the frequency and shape of the writes, and remember to Close -// that Writer when done. -func NewWriter(w io.Writer) *Writer { - return &Writer{ - w: w, - obuf: make([]byte, obufLen), - } -} - -// NewBufferedWriter returns a new Writer that compresses to w, using the -// framing format described at -// https://github.com/google/snappy/blob/master/framing_format.txt -// -// The Writer returned buffers writes. Users must call Close to guarantee all -// data has been forwarded to the underlying io.Writer. They may also call -// Flush zero or more times before calling Close. -func NewBufferedWriter(w io.Writer) *Writer { - return &Writer{ - w: w, - ibuf: make([]byte, 0, maxBlockSize), - obuf: make([]byte, obufLen), - } -} - -// Writer is an io.Writer that can write Snappy-compressed bytes. -// -// Writer handles the Snappy stream format, not the Snappy block format. -type Writer struct { - w io.Writer - err error - - // ibuf is a buffer for the incoming (uncompressed) bytes. - // - // Its use is optional. For backwards compatibility, Writers created by the - // NewWriter function have ibuf == nil, do not buffer incoming bytes, and - // therefore do not need to be Flush'ed or Close'd. - ibuf []byte - - // obuf is a buffer for the outgoing (compressed) bytes. - obuf []byte - - // wroteStreamHeader is whether we have written the stream header. - wroteStreamHeader bool -} - -// Reset discards the writer's state and switches the Snappy writer to write to -// w. This permits reusing a Writer rather than allocating a new one. -func (w *Writer) Reset(writer io.Writer) { - w.w = writer - w.err = nil - if w.ibuf != nil { - w.ibuf = w.ibuf[:0] - } - w.wroteStreamHeader = false -} - -// Write satisfies the io.Writer interface. -func (w *Writer) Write(p []byte) (nRet int, errRet error) { - if w.ibuf == nil { - // Do not buffer incoming bytes. This does not perform or compress well - // if the caller of Writer.Write writes many small slices. This - // behavior is therefore deprecated, but still supported for backwards - // compatibility with code that doesn't explicitly Flush or Close. - return w.write(p) - } - - // The remainder of this method is based on bufio.Writer.Write from the - // standard library. - - for len(p) > (cap(w.ibuf)-len(w.ibuf)) && w.err == nil { - var n int - if len(w.ibuf) == 0 { - // Large write, empty buffer. - // Write directly from p to avoid copy. - n, _ = w.write(p) - } else { - n = copy(w.ibuf[len(w.ibuf):cap(w.ibuf)], p) - w.ibuf = w.ibuf[:len(w.ibuf)+n] - w.Flush() - } - nRet += n - p = p[n:] - } - if w.err != nil { - return nRet, w.err - } - n := copy(w.ibuf[len(w.ibuf):cap(w.ibuf)], p) - w.ibuf = w.ibuf[:len(w.ibuf)+n] - nRet += n - return nRet, nil -} - -func (w *Writer) write(p []byte) (nRet int, errRet error) { - if w.err != nil { - return 0, w.err - } - for len(p) > 0 { - obufStart := len(magicChunk) - if !w.wroteStreamHeader { - w.wroteStreamHeader = true - copy(w.obuf, magicChunk) - obufStart = 0 - } - - var uncompressed []byte - if len(p) > maxBlockSize { - uncompressed, p = p[:maxBlockSize], p[maxBlockSize:] - } else { - uncompressed, p = p, nil - } - checksum := crc(uncompressed) - - // Compress the buffer, discarding the result if the improvement - // isn't at least 12.5%. - compressed := Encode(w.obuf[obufHeaderLen:], uncompressed) - chunkType := uint8(chunkTypeCompressedData) - chunkLen := 4 + len(compressed) - obufEnd := obufHeaderLen + len(compressed) - if len(compressed) >= len(uncompressed)-len(uncompressed)/8 { - chunkType = chunkTypeUncompressedData - chunkLen = 4 + len(uncompressed) - obufEnd = obufHeaderLen - } - - // Fill in the per-chunk header that comes before the body. - w.obuf[len(magicChunk)+0] = chunkType - w.obuf[len(magicChunk)+1] = uint8(chunkLen >> 0) - w.obuf[len(magicChunk)+2] = uint8(chunkLen >> 8) - w.obuf[len(magicChunk)+3] = uint8(chunkLen >> 16) - w.obuf[len(magicChunk)+4] = uint8(checksum >> 0) - w.obuf[len(magicChunk)+5] = uint8(checksum >> 8) - w.obuf[len(magicChunk)+6] = uint8(checksum >> 16) - w.obuf[len(magicChunk)+7] = uint8(checksum >> 24) - - if _, err := w.w.Write(w.obuf[obufStart:obufEnd]); err != nil { - w.err = err - return nRet, err - } - if chunkType == chunkTypeUncompressedData { - if _, err := w.w.Write(uncompressed); err != nil { - w.err = err - return nRet, err - } - } - nRet += len(uncompressed) - } - return nRet, nil -} - -// Flush flushes the Writer to its underlying io.Writer. -func (w *Writer) Flush() error { - if w.err != nil { - return w.err - } - if len(w.ibuf) == 0 { - return nil - } - w.write(w.ibuf) - w.ibuf = w.ibuf[:0] - return w.err -} - -// Close calls Flush and then closes the Writer. -func (w *Writer) Close() error { - w.Flush() - ret := w.err - if w.err == nil { - w.err = errClosed - } - return ret -} diff --git a/vendor/github.com/golang/snappy/encode_amd64.s b/vendor/github.com/golang/snappy/encode_amd64.s deleted file mode 100644 index adfd979fe..000000000 --- a/vendor/github.com/golang/snappy/encode_amd64.s +++ /dev/null @@ -1,730 +0,0 @@ -// Copyright 2016 The Go Authors. All rights reserved. -// Use of this source code is governed by a BSD-style -// license that can be found in the LICENSE file. - -// +build !appengine -// +build gc -// +build !noasm - -#include "textflag.h" - -// The XXX lines assemble on Go 1.4, 1.5 and 1.7, but not 1.6, due to a -// Go toolchain regression. See https://github.com/golang/go/issues/15426 and -// https://github.com/golang/snappy/issues/29 -// -// As a workaround, the package was built with a known good assembler, and -// those instructions were disassembled by "objdump -d" to yield the -// 4e 0f b7 7c 5c 78 movzwq 0x78(%rsp,%r11,2),%r15 -// style comments, in AT&T asm syntax. Note that rsp here is a physical -// register, not Go/asm's SP pseudo-register (see https://golang.org/doc/asm). -// The instructions were then encoded as "BYTE $0x.." sequences, which assemble -// fine on Go 1.6. - -// The asm code generally follows the pure Go code in encode_other.go, except -// where marked with a "!!!". - -// ---------------------------------------------------------------------------- - -// func emitLiteral(dst, lit []byte) int -// -// All local variables fit into registers. The register allocation: -// - AX len(lit) -// - BX n -// - DX return value -// - DI &dst[i] -// - R10 &lit[0] -// -// The 24 bytes of stack space is to call runtime·memmove. -// -// The unusual register allocation of local variables, such as R10 for the -// source pointer, matches the allocation used at the call site in encodeBlock, -// which makes it easier to manually inline this function. -TEXT ·emitLiteral(SB), NOSPLIT, $24-56 - MOVQ dst_base+0(FP), DI - MOVQ lit_base+24(FP), R10 - MOVQ lit_len+32(FP), AX - MOVQ AX, DX - MOVL AX, BX - SUBL $1, BX - - CMPL BX, $60 - JLT oneByte - CMPL BX, $256 - JLT twoBytes - -threeBytes: - MOVB $0xf4, 0(DI) - MOVW BX, 1(DI) - ADDQ $3, DI - ADDQ $3, DX - JMP memmove - -twoBytes: - MOVB $0xf0, 0(DI) - MOVB BX, 1(DI) - ADDQ $2, DI - ADDQ $2, DX - JMP memmove - -oneByte: - SHLB $2, BX - MOVB BX, 0(DI) - ADDQ $1, DI - ADDQ $1, DX - -memmove: - MOVQ DX, ret+48(FP) - - // copy(dst[i:], lit) - // - // This means calling runtime·memmove(&dst[i], &lit[0], len(lit)), so we push - // DI, R10 and AX as arguments. - MOVQ DI, 0(SP) - MOVQ R10, 8(SP) - MOVQ AX, 16(SP) - CALL runtime·memmove(SB) - RET - -// ---------------------------------------------------------------------------- - -// func emitCopy(dst []byte, offset, length int) int -// -// All local variables fit into registers. The register allocation: -// - AX length -// - SI &dst[0] -// - DI &dst[i] -// - R11 offset -// -// The unusual register allocation of local variables, such as R11 for the -// offset, matches the allocation used at the call site in encodeBlock, which -// makes it easier to manually inline this function. -TEXT ·emitCopy(SB), NOSPLIT, $0-48 - MOVQ dst_base+0(FP), DI - MOVQ DI, SI - MOVQ offset+24(FP), R11 - MOVQ length+32(FP), AX - -loop0: - // for length >= 68 { etc } - CMPL AX, $68 - JLT step1 - - // Emit a length 64 copy, encoded as 3 bytes. - MOVB $0xfe, 0(DI) - MOVW R11, 1(DI) - ADDQ $3, DI - SUBL $64, AX - JMP loop0 - -step1: - // if length > 64 { etc } - CMPL AX, $64 - JLE step2 - - // Emit a length 60 copy, encoded as 3 bytes. - MOVB $0xee, 0(DI) - MOVW R11, 1(DI) - ADDQ $3, DI - SUBL $60, AX - -step2: - // if length >= 12 || offset >= 2048 { goto step3 } - CMPL AX, $12 - JGE step3 - CMPL R11, $2048 - JGE step3 - - // Emit the remaining copy, encoded as 2 bytes. - MOVB R11, 1(DI) - SHRL $8, R11 - SHLB $5, R11 - SUBB $4, AX - SHLB $2, AX - ORB AX, R11 - ORB $1, R11 - MOVB R11, 0(DI) - ADDQ $2, DI - - // Return the number of bytes written. - SUBQ SI, DI - MOVQ DI, ret+40(FP) - RET - -step3: - // Emit the remaining copy, encoded as 3 bytes. - SUBL $1, AX - SHLB $2, AX - ORB $2, AX - MOVB AX, 0(DI) - MOVW R11, 1(DI) - ADDQ $3, DI - - // Return the number of bytes written. - SUBQ SI, DI - MOVQ DI, ret+40(FP) - RET - -// ---------------------------------------------------------------------------- - -// func extendMatch(src []byte, i, j int) int -// -// All local variables fit into registers. The register allocation: -// - DX &src[0] -// - SI &src[j] -// - R13 &src[len(src) - 8] -// - R14 &src[len(src)] -// - R15 &src[i] -// -// The unusual register allocation of local variables, such as R15 for a source -// pointer, matches the allocation used at the call site in encodeBlock, which -// makes it easier to manually inline this function. -TEXT ·extendMatch(SB), NOSPLIT, $0-48 - MOVQ src_base+0(FP), DX - MOVQ src_len+8(FP), R14 - MOVQ i+24(FP), R15 - MOVQ j+32(FP), SI - ADDQ DX, R14 - ADDQ DX, R15 - ADDQ DX, SI - MOVQ R14, R13 - SUBQ $8, R13 - -cmp8: - // As long as we are 8 or more bytes before the end of src, we can load and - // compare 8 bytes at a time. If those 8 bytes are equal, repeat. - CMPQ SI, R13 - JA cmp1 - MOVQ (R15), AX - MOVQ (SI), BX - CMPQ AX, BX - JNE bsf - ADDQ $8, R15 - ADDQ $8, SI - JMP cmp8 - -bsf: - // If those 8 bytes were not equal, XOR the two 8 byte values, and return - // the index of the first byte that differs. The BSF instruction finds the - // least significant 1 bit, the amd64 architecture is little-endian, and - // the shift by 3 converts a bit index to a byte index. - XORQ AX, BX - BSFQ BX, BX - SHRQ $3, BX - ADDQ BX, SI - - // Convert from &src[ret] to ret. - SUBQ DX, SI - MOVQ SI, ret+40(FP) - RET - -cmp1: - // In src's tail, compare 1 byte at a time. - CMPQ SI, R14 - JAE extendMatchEnd - MOVB (R15), AX - MOVB (SI), BX - CMPB AX, BX - JNE extendMatchEnd - ADDQ $1, R15 - ADDQ $1, SI - JMP cmp1 - -extendMatchEnd: - // Convert from &src[ret] to ret. - SUBQ DX, SI - MOVQ SI, ret+40(FP) - RET - -// ---------------------------------------------------------------------------- - -// func encodeBlock(dst, src []byte) (d int) -// -// All local variables fit into registers, other than "var table". The register -// allocation: -// - AX . . -// - BX . . -// - CX 56 shift (note that amd64 shifts by non-immediates must use CX). -// - DX 64 &src[0], tableSize -// - SI 72 &src[s] -// - DI 80 &dst[d] -// - R9 88 sLimit -// - R10 . &src[nextEmit] -// - R11 96 prevHash, currHash, nextHash, offset -// - R12 104 &src[base], skip -// - R13 . &src[nextS], &src[len(src) - 8] -// - R14 . len(src), bytesBetweenHashLookups, &src[len(src)], x -// - R15 112 candidate -// -// The second column (56, 64, etc) is the stack offset to spill the registers -// when calling other functions. We could pack this slightly tighter, but it's -// simpler to have a dedicated spill map independent of the function called. -// -// "var table [maxTableSize]uint16" takes up 32768 bytes of stack space. An -// extra 56 bytes, to call other functions, and an extra 64 bytes, to spill -// local variables (registers) during calls gives 32768 + 56 + 64 = 32888. -TEXT ·encodeBlock(SB), 0, $32888-56 - MOVQ dst_base+0(FP), DI - MOVQ src_base+24(FP), SI - MOVQ src_len+32(FP), R14 - - // shift, tableSize := uint32(32-8), 1<<8 - MOVQ $24, CX - MOVQ $256, DX - -calcShift: - // for ; tableSize < maxTableSize && tableSize < len(src); tableSize *= 2 { - // shift-- - // } - CMPQ DX, $16384 - JGE varTable - CMPQ DX, R14 - JGE varTable - SUBQ $1, CX - SHLQ $1, DX - JMP calcShift - -varTable: - // var table [maxTableSize]uint16 - // - // In the asm code, unlike the Go code, we can zero-initialize only the - // first tableSize elements. Each uint16 element is 2 bytes and each MOVOU - // writes 16 bytes, so we can do only tableSize/8 writes instead of the - // 2048 writes that would zero-initialize all of table's 32768 bytes. - SHRQ $3, DX - LEAQ table-32768(SP), BX - PXOR X0, X0 - -memclr: - MOVOU X0, 0(BX) - ADDQ $16, BX - SUBQ $1, DX - JNZ memclr - - // !!! DX = &src[0] - MOVQ SI, DX - - // sLimit := len(src) - inputMargin - MOVQ R14, R9 - SUBQ $15, R9 - - // !!! Pre-emptively spill CX, DX and R9 to the stack. Their values don't - // change for the rest of the function. - MOVQ CX, 56(SP) - MOVQ DX, 64(SP) - MOVQ R9, 88(SP) - - // nextEmit := 0 - MOVQ DX, R10 - - // s := 1 - ADDQ $1, SI - - // nextHash := hash(load32(src, s), shift) - MOVL 0(SI), R11 - IMULL $0x1e35a7bd, R11 - SHRL CX, R11 - -outer: - // for { etc } - - // skip := 32 - MOVQ $32, R12 - - // nextS := s - MOVQ SI, R13 - - // candidate := 0 - MOVQ $0, R15 - -inner0: - // for { etc } - - // s := nextS - MOVQ R13, SI - - // bytesBetweenHashLookups := skip >> 5 - MOVQ R12, R14 - SHRQ $5, R14 - - // nextS = s + bytesBetweenHashLookups - ADDQ R14, R13 - - // skip += bytesBetweenHashLookups - ADDQ R14, R12 - - // if nextS > sLimit { goto emitRemainder } - MOVQ R13, AX - SUBQ DX, AX - CMPQ AX, R9 - JA emitRemainder - - // candidate = int(table[nextHash]) - // XXX: MOVWQZX table-32768(SP)(R11*2), R15 - // XXX: 4e 0f b7 7c 5c 78 movzwq 0x78(%rsp,%r11,2),%r15 - BYTE $0x4e - BYTE $0x0f - BYTE $0xb7 - BYTE $0x7c - BYTE $0x5c - BYTE $0x78 - - // table[nextHash] = uint16(s) - MOVQ SI, AX - SUBQ DX, AX - - // XXX: MOVW AX, table-32768(SP)(R11*2) - // XXX: 66 42 89 44 5c 78 mov %ax,0x78(%rsp,%r11,2) - BYTE $0x66 - BYTE $0x42 - BYTE $0x89 - BYTE $0x44 - BYTE $0x5c - BYTE $0x78 - - // nextHash = hash(load32(src, nextS), shift) - MOVL 0(R13), R11 - IMULL $0x1e35a7bd, R11 - SHRL CX, R11 - - // if load32(src, s) != load32(src, candidate) { continue } break - MOVL 0(SI), AX - MOVL (DX)(R15*1), BX - CMPL AX, BX - JNE inner0 - -fourByteMatch: - // As per the encode_other.go code: - // - // A 4-byte match has been found. We'll later see etc. - - // !!! Jump to a fast path for short (<= 16 byte) literals. See the comment - // on inputMargin in encode.go. - MOVQ SI, AX - SUBQ R10, AX - CMPQ AX, $16 - JLE emitLiteralFastPath - - // ---------------------------------------- - // Begin inline of the emitLiteral call. - // - // d += emitLiteral(dst[d:], src[nextEmit:s]) - - MOVL AX, BX - SUBL $1, BX - - CMPL BX, $60 - JLT inlineEmitLiteralOneByte - CMPL BX, $256 - JLT inlineEmitLiteralTwoBytes - -inlineEmitLiteralThreeBytes: - MOVB $0xf4, 0(DI) - MOVW BX, 1(DI) - ADDQ $3, DI - JMP inlineEmitLiteralMemmove - -inlineEmitLiteralTwoBytes: - MOVB $0xf0, 0(DI) - MOVB BX, 1(DI) - ADDQ $2, DI - JMP inlineEmitLiteralMemmove - -inlineEmitLiteralOneByte: - SHLB $2, BX - MOVB BX, 0(DI) - ADDQ $1, DI - -inlineEmitLiteralMemmove: - // Spill local variables (registers) onto the stack; call; unspill. - // - // copy(dst[i:], lit) - // - // This means calling runtime·memmove(&dst[i], &lit[0], len(lit)), so we push - // DI, R10 and AX as arguments. - MOVQ DI, 0(SP) - MOVQ R10, 8(SP) - MOVQ AX, 16(SP) - ADDQ AX, DI // Finish the "d +=" part of "d += emitLiteral(etc)". - MOVQ SI, 72(SP) - MOVQ DI, 80(SP) - MOVQ R15, 112(SP) - CALL runtime·memmove(SB) - MOVQ 56(SP), CX - MOVQ 64(SP), DX - MOVQ 72(SP), SI - MOVQ 80(SP), DI - MOVQ 88(SP), R9 - MOVQ 112(SP), R15 - JMP inner1 - -inlineEmitLiteralEnd: - // End inline of the emitLiteral call. - // ---------------------------------------- - -emitLiteralFastPath: - // !!! Emit the 1-byte encoding "uint8(len(lit)-1)<<2". - MOVB AX, BX - SUBB $1, BX - SHLB $2, BX - MOVB BX, (DI) - ADDQ $1, DI - - // !!! Implement the copy from lit to dst as a 16-byte load and store. - // (Encode's documentation says that dst and src must not overlap.) - // - // This always copies 16 bytes, instead of only len(lit) bytes, but that's - // OK. Subsequent iterations will fix up the overrun. - // - // Note that on amd64, it is legal and cheap to issue unaligned 8-byte or - // 16-byte loads and stores. This technique probably wouldn't be as - // effective on architectures that are fussier about alignment. - MOVOU 0(R10), X0 - MOVOU X0, 0(DI) - ADDQ AX, DI - -inner1: - // for { etc } - - // base := s - MOVQ SI, R12 - - // !!! offset := base - candidate - MOVQ R12, R11 - SUBQ R15, R11 - SUBQ DX, R11 - - // ---------------------------------------- - // Begin inline of the extendMatch call. - // - // s = extendMatch(src, candidate+4, s+4) - - // !!! R14 = &src[len(src)] - MOVQ src_len+32(FP), R14 - ADDQ DX, R14 - - // !!! R13 = &src[len(src) - 8] - MOVQ R14, R13 - SUBQ $8, R13 - - // !!! R15 = &src[candidate + 4] - ADDQ $4, R15 - ADDQ DX, R15 - - // !!! s += 4 - ADDQ $4, SI - -inlineExtendMatchCmp8: - // As long as we are 8 or more bytes before the end of src, we can load and - // compare 8 bytes at a time. If those 8 bytes are equal, repeat. - CMPQ SI, R13 - JA inlineExtendMatchCmp1 - MOVQ (R15), AX - MOVQ (SI), BX - CMPQ AX, BX - JNE inlineExtendMatchBSF - ADDQ $8, R15 - ADDQ $8, SI - JMP inlineExtendMatchCmp8 - -inlineExtendMatchBSF: - // If those 8 bytes were not equal, XOR the two 8 byte values, and return - // the index of the first byte that differs. The BSF instruction finds the - // least significant 1 bit, the amd64 architecture is little-endian, and - // the shift by 3 converts a bit index to a byte index. - XORQ AX, BX - BSFQ BX, BX - SHRQ $3, BX - ADDQ BX, SI - JMP inlineExtendMatchEnd - -inlineExtendMatchCmp1: - // In src's tail, compare 1 byte at a time. - CMPQ SI, R14 - JAE inlineExtendMatchEnd - MOVB (R15), AX - MOVB (SI), BX - CMPB AX, BX - JNE inlineExtendMatchEnd - ADDQ $1, R15 - ADDQ $1, SI - JMP inlineExtendMatchCmp1 - -inlineExtendMatchEnd: - // End inline of the extendMatch call. - // ---------------------------------------- - - // ---------------------------------------- - // Begin inline of the emitCopy call. - // - // d += emitCopy(dst[d:], base-candidate, s-base) - - // !!! length := s - base - MOVQ SI, AX - SUBQ R12, AX - -inlineEmitCopyLoop0: - // for length >= 68 { etc } - CMPL AX, $68 - JLT inlineEmitCopyStep1 - - // Emit a length 64 copy, encoded as 3 bytes. - MOVB $0xfe, 0(DI) - MOVW R11, 1(DI) - ADDQ $3, DI - SUBL $64, AX - JMP inlineEmitCopyLoop0 - -inlineEmitCopyStep1: - // if length > 64 { etc } - CMPL AX, $64 - JLE inlineEmitCopyStep2 - - // Emit a length 60 copy, encoded as 3 bytes. - MOVB $0xee, 0(DI) - MOVW R11, 1(DI) - ADDQ $3, DI - SUBL $60, AX - -inlineEmitCopyStep2: - // if length >= 12 || offset >= 2048 { goto inlineEmitCopyStep3 } - CMPL AX, $12 - JGE inlineEmitCopyStep3 - CMPL R11, $2048 - JGE inlineEmitCopyStep3 - - // Emit the remaining copy, encoded as 2 bytes. - MOVB R11, 1(DI) - SHRL $8, R11 - SHLB $5, R11 - SUBB $4, AX - SHLB $2, AX - ORB AX, R11 - ORB $1, R11 - MOVB R11, 0(DI) - ADDQ $2, DI - JMP inlineEmitCopyEnd - -inlineEmitCopyStep3: - // Emit the remaining copy, encoded as 3 bytes. - SUBL $1, AX - SHLB $2, AX - ORB $2, AX - MOVB AX, 0(DI) - MOVW R11, 1(DI) - ADDQ $3, DI - -inlineEmitCopyEnd: - // End inline of the emitCopy call. - // ---------------------------------------- - - // nextEmit = s - MOVQ SI, R10 - - // if s >= sLimit { goto emitRemainder } - MOVQ SI, AX - SUBQ DX, AX - CMPQ AX, R9 - JAE emitRemainder - - // As per the encode_other.go code: - // - // We could immediately etc. - - // x := load64(src, s-1) - MOVQ -1(SI), R14 - - // prevHash := hash(uint32(x>>0), shift) - MOVL R14, R11 - IMULL $0x1e35a7bd, R11 - SHRL CX, R11 - - // table[prevHash] = uint16(s-1) - MOVQ SI, AX - SUBQ DX, AX - SUBQ $1, AX - - // XXX: MOVW AX, table-32768(SP)(R11*2) - // XXX: 66 42 89 44 5c 78 mov %ax,0x78(%rsp,%r11,2) - BYTE $0x66 - BYTE $0x42 - BYTE $0x89 - BYTE $0x44 - BYTE $0x5c - BYTE $0x78 - - // currHash := hash(uint32(x>>8), shift) - SHRQ $8, R14 - MOVL R14, R11 - IMULL $0x1e35a7bd, R11 - SHRL CX, R11 - - // candidate = int(table[currHash]) - // XXX: MOVWQZX table-32768(SP)(R11*2), R15 - // XXX: 4e 0f b7 7c 5c 78 movzwq 0x78(%rsp,%r11,2),%r15 - BYTE $0x4e - BYTE $0x0f - BYTE $0xb7 - BYTE $0x7c - BYTE $0x5c - BYTE $0x78 - - // table[currHash] = uint16(s) - ADDQ $1, AX - - // XXX: MOVW AX, table-32768(SP)(R11*2) - // XXX: 66 42 89 44 5c 78 mov %ax,0x78(%rsp,%r11,2) - BYTE $0x66 - BYTE $0x42 - BYTE $0x89 - BYTE $0x44 - BYTE $0x5c - BYTE $0x78 - - // if uint32(x>>8) == load32(src, candidate) { continue } - MOVL (DX)(R15*1), BX - CMPL R14, BX - JEQ inner1 - - // nextHash = hash(uint32(x>>16), shift) - SHRQ $8, R14 - MOVL R14, R11 - IMULL $0x1e35a7bd, R11 - SHRL CX, R11 - - // s++ - ADDQ $1, SI - - // break out of the inner1 for loop, i.e. continue the outer loop. - JMP outer - -emitRemainder: - // if nextEmit < len(src) { etc } - MOVQ src_len+32(FP), AX - ADDQ DX, AX - CMPQ R10, AX - JEQ encodeBlockEnd - - // d += emitLiteral(dst[d:], src[nextEmit:]) - // - // Push args. - MOVQ DI, 0(SP) - MOVQ $0, 8(SP) // Unnecessary, as the callee ignores it, but conservative. - MOVQ $0, 16(SP) // Unnecessary, as the callee ignores it, but conservative. - MOVQ R10, 24(SP) - SUBQ R10, AX - MOVQ AX, 32(SP) - MOVQ AX, 40(SP) // Unnecessary, as the callee ignores it, but conservative. - - // Spill local variables (registers) onto the stack; call; unspill. - MOVQ DI, 80(SP) - CALL ·emitLiteral(SB) - MOVQ 80(SP), DI - - // Finish the "d +=" part of "d += emitLiteral(etc)". - ADDQ 48(SP), DI - -encodeBlockEnd: - MOVQ dst_base+0(FP), AX - SUBQ AX, DI - MOVQ DI, d+48(FP) - RET diff --git a/vendor/github.com/golang/snappy/encode_arm64.s b/vendor/github.com/golang/snappy/encode_arm64.s deleted file mode 100644 index f8d54adfc..000000000 --- a/vendor/github.com/golang/snappy/encode_arm64.s +++ /dev/null @@ -1,722 +0,0 @@ -// Copyright 2020 The Go Authors. All rights reserved. -// Use of this source code is governed by a BSD-style -// license that can be found in the LICENSE file. - -// +build !appengine -// +build gc -// +build !noasm - -#include "textflag.h" - -// The asm code generally follows the pure Go code in encode_other.go, except -// where marked with a "!!!". - -// ---------------------------------------------------------------------------- - -// func emitLiteral(dst, lit []byte) int -// -// All local variables fit into registers. The register allocation: -// - R3 len(lit) -// - R4 n -// - R6 return value -// - R8 &dst[i] -// - R10 &lit[0] -// -// The 32 bytes of stack space is to call runtime·memmove. -// -// The unusual register allocation of local variables, such as R10 for the -// source pointer, matches the allocation used at the call site in encodeBlock, -// which makes it easier to manually inline this function. -TEXT ·emitLiteral(SB), NOSPLIT, $32-56 - MOVD dst_base+0(FP), R8 - MOVD lit_base+24(FP), R10 - MOVD lit_len+32(FP), R3 - MOVD R3, R6 - MOVW R3, R4 - SUBW $1, R4, R4 - - CMPW $60, R4 - BLT oneByte - CMPW $256, R4 - BLT twoBytes - -threeBytes: - MOVD $0xf4, R2 - MOVB R2, 0(R8) - MOVW R4, 1(R8) - ADD $3, R8, R8 - ADD $3, R6, R6 - B memmove - -twoBytes: - MOVD $0xf0, R2 - MOVB R2, 0(R8) - MOVB R4, 1(R8) - ADD $2, R8, R8 - ADD $2, R6, R6 - B memmove - -oneByte: - LSLW $2, R4, R4 - MOVB R4, 0(R8) - ADD $1, R8, R8 - ADD $1, R6, R6 - -memmove: - MOVD R6, ret+48(FP) - - // copy(dst[i:], lit) - // - // This means calling runtime·memmove(&dst[i], &lit[0], len(lit)), so we push - // R8, R10 and R3 as arguments. - MOVD R8, 8(RSP) - MOVD R10, 16(RSP) - MOVD R3, 24(RSP) - CALL runtime·memmove(SB) - RET - -// ---------------------------------------------------------------------------- - -// func emitCopy(dst []byte, offset, length int) int -// -// All local variables fit into registers. The register allocation: -// - R3 length -// - R7 &dst[0] -// - R8 &dst[i] -// - R11 offset -// -// The unusual register allocation of local variables, such as R11 for the -// offset, matches the allocation used at the call site in encodeBlock, which -// makes it easier to manually inline this function. -TEXT ·emitCopy(SB), NOSPLIT, $0-48 - MOVD dst_base+0(FP), R8 - MOVD R8, R7 - MOVD offset+24(FP), R11 - MOVD length+32(FP), R3 - -loop0: - // for length >= 68 { etc } - CMPW $68, R3 - BLT step1 - - // Emit a length 64 copy, encoded as 3 bytes. - MOVD $0xfe, R2 - MOVB R2, 0(R8) - MOVW R11, 1(R8) - ADD $3, R8, R8 - SUB $64, R3, R3 - B loop0 - -step1: - // if length > 64 { etc } - CMP $64, R3 - BLE step2 - - // Emit a length 60 copy, encoded as 3 bytes. - MOVD $0xee, R2 - MOVB R2, 0(R8) - MOVW R11, 1(R8) - ADD $3, R8, R8 - SUB $60, R3, R3 - -step2: - // if length >= 12 || offset >= 2048 { goto step3 } - CMP $12, R3 - BGE step3 - CMPW $2048, R11 - BGE step3 - - // Emit the remaining copy, encoded as 2 bytes. - MOVB R11, 1(R8) - LSRW $3, R11, R11 - AND $0xe0, R11, R11 - SUB $4, R3, R3 - LSLW $2, R3 - AND $0xff, R3, R3 - ORRW R3, R11, R11 - ORRW $1, R11, R11 - MOVB R11, 0(R8) - ADD $2, R8, R8 - - // Return the number of bytes written. - SUB R7, R8, R8 - MOVD R8, ret+40(FP) - RET - -step3: - // Emit the remaining copy, encoded as 3 bytes. - SUB $1, R3, R3 - AND $0xff, R3, R3 - LSLW $2, R3, R3 - ORRW $2, R3, R3 - MOVB R3, 0(R8) - MOVW R11, 1(R8) - ADD $3, R8, R8 - - // Return the number of bytes written. - SUB R7, R8, R8 - MOVD R8, ret+40(FP) - RET - -// ---------------------------------------------------------------------------- - -// func extendMatch(src []byte, i, j int) int -// -// All local variables fit into registers. The register allocation: -// - R6 &src[0] -// - R7 &src[j] -// - R13 &src[len(src) - 8] -// - R14 &src[len(src)] -// - R15 &src[i] -// -// The unusual register allocation of local variables, such as R15 for a source -// pointer, matches the allocation used at the call site in encodeBlock, which -// makes it easier to manually inline this function. -TEXT ·extendMatch(SB), NOSPLIT, $0-48 - MOVD src_base+0(FP), R6 - MOVD src_len+8(FP), R14 - MOVD i+24(FP), R15 - MOVD j+32(FP), R7 - ADD R6, R14, R14 - ADD R6, R15, R15 - ADD R6, R7, R7 - MOVD R14, R13 - SUB $8, R13, R13 - -cmp8: - // As long as we are 8 or more bytes before the end of src, we can load and - // compare 8 bytes at a time. If those 8 bytes are equal, repeat. - CMP R13, R7 - BHI cmp1 - MOVD (R15), R3 - MOVD (R7), R4 - CMP R4, R3 - BNE bsf - ADD $8, R15, R15 - ADD $8, R7, R7 - B cmp8 - -bsf: - // If those 8 bytes were not equal, XOR the two 8 byte values, and return - // the index of the first byte that differs. - // RBIT reverses the bit order, then CLZ counts the leading zeros, the - // combination of which finds the least significant bit which is set. - // The arm64 architecture is little-endian, and the shift by 3 converts - // a bit index to a byte index. - EOR R3, R4, R4 - RBIT R4, R4 - CLZ R4, R4 - ADD R4>>3, R7, R7 - - // Convert from &src[ret] to ret. - SUB R6, R7, R7 - MOVD R7, ret+40(FP) - RET - -cmp1: - // In src's tail, compare 1 byte at a time. - CMP R7, R14 - BLS extendMatchEnd - MOVB (R15), R3 - MOVB (R7), R4 - CMP R4, R3 - BNE extendMatchEnd - ADD $1, R15, R15 - ADD $1, R7, R7 - B cmp1 - -extendMatchEnd: - // Convert from &src[ret] to ret. - SUB R6, R7, R7 - MOVD R7, ret+40(FP) - RET - -// ---------------------------------------------------------------------------- - -// func encodeBlock(dst, src []byte) (d int) -// -// All local variables fit into registers, other than "var table". The register -// allocation: -// - R3 . . -// - R4 . . -// - R5 64 shift -// - R6 72 &src[0], tableSize -// - R7 80 &src[s] -// - R8 88 &dst[d] -// - R9 96 sLimit -// - R10 . &src[nextEmit] -// - R11 104 prevHash, currHash, nextHash, offset -// - R12 112 &src[base], skip -// - R13 . &src[nextS], &src[len(src) - 8] -// - R14 . len(src), bytesBetweenHashLookups, &src[len(src)], x -// - R15 120 candidate -// - R16 . hash constant, 0x1e35a7bd -// - R17 . &table -// - . 128 table -// -// The second column (64, 72, etc) is the stack offset to spill the registers -// when calling other functions. We could pack this slightly tighter, but it's -// simpler to have a dedicated spill map independent of the function called. -// -// "var table [maxTableSize]uint16" takes up 32768 bytes of stack space. An -// extra 64 bytes, to call other functions, and an extra 64 bytes, to spill -// local variables (registers) during calls gives 32768 + 64 + 64 = 32896. -TEXT ·encodeBlock(SB), 0, $32896-56 - MOVD dst_base+0(FP), R8 - MOVD src_base+24(FP), R7 - MOVD src_len+32(FP), R14 - - // shift, tableSize := uint32(32-8), 1<<8 - MOVD $24, R5 - MOVD $256, R6 - MOVW $0xa7bd, R16 - MOVKW $(0x1e35<<16), R16 - -calcShift: - // for ; tableSize < maxTableSize && tableSize < len(src); tableSize *= 2 { - // shift-- - // } - MOVD $16384, R2 - CMP R2, R6 - BGE varTable - CMP R14, R6 - BGE varTable - SUB $1, R5, R5 - LSL $1, R6, R6 - B calcShift - -varTable: - // var table [maxTableSize]uint16 - // - // In the asm code, unlike the Go code, we can zero-initialize only the - // first tableSize elements. Each uint16 element is 2 bytes and each - // iterations writes 64 bytes, so we can do only tableSize/32 writes - // instead of the 2048 writes that would zero-initialize all of table's - // 32768 bytes. This clear could overrun the first tableSize elements, but - // it won't overrun the allocated stack size. - ADD $128, RSP, R17 - MOVD R17, R4 - - // !!! R6 = &src[tableSize] - ADD R6<<1, R17, R6 - -memclr: - STP.P (ZR, ZR), 64(R4) - STP (ZR, ZR), -48(R4) - STP (ZR, ZR), -32(R4) - STP (ZR, ZR), -16(R4) - CMP R4, R6 - BHI memclr - - // !!! R6 = &src[0] - MOVD R7, R6 - - // sLimit := len(src) - inputMargin - MOVD R14, R9 - SUB $15, R9, R9 - - // !!! Pre-emptively spill R5, R6 and R9 to the stack. Their values don't - // change for the rest of the function. - MOVD R5, 64(RSP) - MOVD R6, 72(RSP) - MOVD R9, 96(RSP) - - // nextEmit := 0 - MOVD R6, R10 - - // s := 1 - ADD $1, R7, R7 - - // nextHash := hash(load32(src, s), shift) - MOVW 0(R7), R11 - MULW R16, R11, R11 - LSRW R5, R11, R11 - -outer: - // for { etc } - - // skip := 32 - MOVD $32, R12 - - // nextS := s - MOVD R7, R13 - - // candidate := 0 - MOVD $0, R15 - -inner0: - // for { etc } - - // s := nextS - MOVD R13, R7 - - // bytesBetweenHashLookups := skip >> 5 - MOVD R12, R14 - LSR $5, R14, R14 - - // nextS = s + bytesBetweenHashLookups - ADD R14, R13, R13 - - // skip += bytesBetweenHashLookups - ADD R14, R12, R12 - - // if nextS > sLimit { goto emitRemainder } - MOVD R13, R3 - SUB R6, R3, R3 - CMP R9, R3 - BHI emitRemainder - - // candidate = int(table[nextHash]) - MOVHU 0(R17)(R11<<1), R15 - - // table[nextHash] = uint16(s) - MOVD R7, R3 - SUB R6, R3, R3 - - MOVH R3, 0(R17)(R11<<1) - - // nextHash = hash(load32(src, nextS), shift) - MOVW 0(R13), R11 - MULW R16, R11 - LSRW R5, R11, R11 - - // if load32(src, s) != load32(src, candidate) { continue } break - MOVW 0(R7), R3 - MOVW (R6)(R15), R4 - CMPW R4, R3 - BNE inner0 - -fourByteMatch: - // As per the encode_other.go code: - // - // A 4-byte match has been found. We'll later see etc. - - // !!! Jump to a fast path for short (<= 16 byte) literals. See the comment - // on inputMargin in encode.go. - MOVD R7, R3 - SUB R10, R3, R3 - CMP $16, R3 - BLE emitLiteralFastPath - - // ---------------------------------------- - // Begin inline of the emitLiteral call. - // - // d += emitLiteral(dst[d:], src[nextEmit:s]) - - MOVW R3, R4 - SUBW $1, R4, R4 - - MOVW $60, R2 - CMPW R2, R4 - BLT inlineEmitLiteralOneByte - MOVW $256, R2 - CMPW R2, R4 - BLT inlineEmitLiteralTwoBytes - -inlineEmitLiteralThreeBytes: - MOVD $0xf4, R1 - MOVB R1, 0(R8) - MOVW R4, 1(R8) - ADD $3, R8, R8 - B inlineEmitLiteralMemmove - -inlineEmitLiteralTwoBytes: - MOVD $0xf0, R1 - MOVB R1, 0(R8) - MOVB R4, 1(R8) - ADD $2, R8, R8 - B inlineEmitLiteralMemmove - -inlineEmitLiteralOneByte: - LSLW $2, R4, R4 - MOVB R4, 0(R8) - ADD $1, R8, R8 - -inlineEmitLiteralMemmove: - // Spill local variables (registers) onto the stack; call; unspill. - // - // copy(dst[i:], lit) - // - // This means calling runtime·memmove(&dst[i], &lit[0], len(lit)), so we push - // R8, R10 and R3 as arguments. - MOVD R8, 8(RSP) - MOVD R10, 16(RSP) - MOVD R3, 24(RSP) - - // Finish the "d +=" part of "d += emitLiteral(etc)". - ADD R3, R8, R8 - MOVD R7, 80(RSP) - MOVD R8, 88(RSP) - MOVD R15, 120(RSP) - CALL runtime·memmove(SB) - MOVD 64(RSP), R5 - MOVD 72(RSP), R6 - MOVD 80(RSP), R7 - MOVD 88(RSP), R8 - MOVD 96(RSP), R9 - MOVD 120(RSP), R15 - ADD $128, RSP, R17 - MOVW $0xa7bd, R16 - MOVKW $(0x1e35<<16), R16 - B inner1 - -inlineEmitLiteralEnd: - // End inline of the emitLiteral call. - // ---------------------------------------- - -emitLiteralFastPath: - // !!! Emit the 1-byte encoding "uint8(len(lit)-1)<<2". - MOVB R3, R4 - SUBW $1, R4, R4 - AND $0xff, R4, R4 - LSLW $2, R4, R4 - MOVB R4, (R8) - ADD $1, R8, R8 - - // !!! Implement the copy from lit to dst as a 16-byte load and store. - // (Encode's documentation says that dst and src must not overlap.) - // - // This always copies 16 bytes, instead of only len(lit) bytes, but that's - // OK. Subsequent iterations will fix up the overrun. - // - // Note that on arm64, it is legal and cheap to issue unaligned 8-byte or - // 16-byte loads and stores. This technique probably wouldn't be as - // effective on architectures that are fussier about alignment. - LDP 0(R10), (R0, R1) - STP (R0, R1), 0(R8) - ADD R3, R8, R8 - -inner1: - // for { etc } - - // base := s - MOVD R7, R12 - - // !!! offset := base - candidate - MOVD R12, R11 - SUB R15, R11, R11 - SUB R6, R11, R11 - - // ---------------------------------------- - // Begin inline of the extendMatch call. - // - // s = extendMatch(src, candidate+4, s+4) - - // !!! R14 = &src[len(src)] - MOVD src_len+32(FP), R14 - ADD R6, R14, R14 - - // !!! R13 = &src[len(src) - 8] - MOVD R14, R13 - SUB $8, R13, R13 - - // !!! R15 = &src[candidate + 4] - ADD $4, R15, R15 - ADD R6, R15, R15 - - // !!! s += 4 - ADD $4, R7, R7 - -inlineExtendMatchCmp8: - // As long as we are 8 or more bytes before the end of src, we can load and - // compare 8 bytes at a time. If those 8 bytes are equal, repeat. - CMP R13, R7 - BHI inlineExtendMatchCmp1 - MOVD (R15), R3 - MOVD (R7), R4 - CMP R4, R3 - BNE inlineExtendMatchBSF - ADD $8, R15, R15 - ADD $8, R7, R7 - B inlineExtendMatchCmp8 - -inlineExtendMatchBSF: - // If those 8 bytes were not equal, XOR the two 8 byte values, and return - // the index of the first byte that differs. - // RBIT reverses the bit order, then CLZ counts the leading zeros, the - // combination of which finds the least significant bit which is set. - // The arm64 architecture is little-endian, and the shift by 3 converts - // a bit index to a byte index. - EOR R3, R4, R4 - RBIT R4, R4 - CLZ R4, R4 - ADD R4>>3, R7, R7 - B inlineExtendMatchEnd - -inlineExtendMatchCmp1: - // In src's tail, compare 1 byte at a time. - CMP R7, R14 - BLS inlineExtendMatchEnd - MOVB (R15), R3 - MOVB (R7), R4 - CMP R4, R3 - BNE inlineExtendMatchEnd - ADD $1, R15, R15 - ADD $1, R7, R7 - B inlineExtendMatchCmp1 - -inlineExtendMatchEnd: - // End inline of the extendMatch call. - // ---------------------------------------- - - // ---------------------------------------- - // Begin inline of the emitCopy call. - // - // d += emitCopy(dst[d:], base-candidate, s-base) - - // !!! length := s - base - MOVD R7, R3 - SUB R12, R3, R3 - -inlineEmitCopyLoop0: - // for length >= 68 { etc } - MOVW $68, R2 - CMPW R2, R3 - BLT inlineEmitCopyStep1 - - // Emit a length 64 copy, encoded as 3 bytes. - MOVD $0xfe, R1 - MOVB R1, 0(R8) - MOVW R11, 1(R8) - ADD $3, R8, R8 - SUBW $64, R3, R3 - B inlineEmitCopyLoop0 - -inlineEmitCopyStep1: - // if length > 64 { etc } - MOVW $64, R2 - CMPW R2, R3 - BLE inlineEmitCopyStep2 - - // Emit a length 60 copy, encoded as 3 bytes. - MOVD $0xee, R1 - MOVB R1, 0(R8) - MOVW R11, 1(R8) - ADD $3, R8, R8 - SUBW $60, R3, R3 - -inlineEmitCopyStep2: - // if length >= 12 || offset >= 2048 { goto inlineEmitCopyStep3 } - MOVW $12, R2 - CMPW R2, R3 - BGE inlineEmitCopyStep3 - MOVW $2048, R2 - CMPW R2, R11 - BGE inlineEmitCopyStep3 - - // Emit the remaining copy, encoded as 2 bytes. - MOVB R11, 1(R8) - LSRW $8, R11, R11 - LSLW $5, R11, R11 - SUBW $4, R3, R3 - AND $0xff, R3, R3 - LSLW $2, R3, R3 - ORRW R3, R11, R11 - ORRW $1, R11, R11 - MOVB R11, 0(R8) - ADD $2, R8, R8 - B inlineEmitCopyEnd - -inlineEmitCopyStep3: - // Emit the remaining copy, encoded as 3 bytes. - SUBW $1, R3, R3 - LSLW $2, R3, R3 - ORRW $2, R3, R3 - MOVB R3, 0(R8) - MOVW R11, 1(R8) - ADD $3, R8, R8 - -inlineEmitCopyEnd: - // End inline of the emitCopy call. - // ---------------------------------------- - - // nextEmit = s - MOVD R7, R10 - - // if s >= sLimit { goto emitRemainder } - MOVD R7, R3 - SUB R6, R3, R3 - CMP R3, R9 - BLS emitRemainder - - // As per the encode_other.go code: - // - // We could immediately etc. - - // x := load64(src, s-1) - MOVD -1(R7), R14 - - // prevHash := hash(uint32(x>>0), shift) - MOVW R14, R11 - MULW R16, R11, R11 - LSRW R5, R11, R11 - - // table[prevHash] = uint16(s-1) - MOVD R7, R3 - SUB R6, R3, R3 - SUB $1, R3, R3 - - MOVHU R3, 0(R17)(R11<<1) - - // currHash := hash(uint32(x>>8), shift) - LSR $8, R14, R14 - MOVW R14, R11 - MULW R16, R11, R11 - LSRW R5, R11, R11 - - // candidate = int(table[currHash]) - MOVHU 0(R17)(R11<<1), R15 - - // table[currHash] = uint16(s) - ADD $1, R3, R3 - MOVHU R3, 0(R17)(R11<<1) - - // if uint32(x>>8) == load32(src, candidate) { continue } - MOVW (R6)(R15), R4 - CMPW R4, R14 - BEQ inner1 - - // nextHash = hash(uint32(x>>16), shift) - LSR $8, R14, R14 - MOVW R14, R11 - MULW R16, R11, R11 - LSRW R5, R11, R11 - - // s++ - ADD $1, R7, R7 - - // break out of the inner1 for loop, i.e. continue the outer loop. - B outer - -emitRemainder: - // if nextEmit < len(src) { etc } - MOVD src_len+32(FP), R3 - ADD R6, R3, R3 - CMP R3, R10 - BEQ encodeBlockEnd - - // d += emitLiteral(dst[d:], src[nextEmit:]) - // - // Push args. - MOVD R8, 8(RSP) - MOVD $0, 16(RSP) // Unnecessary, as the callee ignores it, but conservative. - MOVD $0, 24(RSP) // Unnecessary, as the callee ignores it, but conservative. - MOVD R10, 32(RSP) - SUB R10, R3, R3 - MOVD R3, 40(RSP) - MOVD R3, 48(RSP) // Unnecessary, as the callee ignores it, but conservative. - - // Spill local variables (registers) onto the stack; call; unspill. - MOVD R8, 88(RSP) - CALL ·emitLiteral(SB) - MOVD 88(RSP), R8 - - // Finish the "d +=" part of "d += emitLiteral(etc)". - MOVD 56(RSP), R1 - ADD R1, R8, R8 - -encodeBlockEnd: - MOVD dst_base+0(FP), R3 - SUB R3, R8, R8 - MOVD R8, d+48(FP) - RET diff --git a/vendor/github.com/golang/snappy/encode_asm.go b/vendor/github.com/golang/snappy/encode_asm.go deleted file mode 100644 index 107c1e714..000000000 --- a/vendor/github.com/golang/snappy/encode_asm.go +++ /dev/null @@ -1,30 +0,0 @@ -// Copyright 2016 The Snappy-Go Authors. All rights reserved. -// Use of this source code is governed by a BSD-style -// license that can be found in the LICENSE file. - -// +build !appengine -// +build gc -// +build !noasm -// +build amd64 arm64 - -package snappy - -// emitLiteral has the same semantics as in encode_other.go. -// -//go:noescape -func emitLiteral(dst, lit []byte) int - -// emitCopy has the same semantics as in encode_other.go. -// -//go:noescape -func emitCopy(dst []byte, offset, length int) int - -// extendMatch has the same semantics as in encode_other.go. -// -//go:noescape -func extendMatch(src []byte, i, j int) int - -// encodeBlock has the same semantics as in encode_other.go. -// -//go:noescape -func encodeBlock(dst, src []byte) (d int) diff --git a/vendor/github.com/golang/snappy/encode_other.go b/vendor/github.com/golang/snappy/encode_other.go deleted file mode 100644 index 296d7f0be..000000000 --- a/vendor/github.com/golang/snappy/encode_other.go +++ /dev/null @@ -1,238 +0,0 @@ -// Copyright 2016 The Snappy-Go Authors. All rights reserved. -// Use of this source code is governed by a BSD-style -// license that can be found in the LICENSE file. - -// +build !amd64,!arm64 appengine !gc noasm - -package snappy - -func load32(b []byte, i int) uint32 { - b = b[i : i+4 : len(b)] // Help the compiler eliminate bounds checks on the next line. - return uint32(b[0]) | uint32(b[1])<<8 | uint32(b[2])<<16 | uint32(b[3])<<24 -} - -func load64(b []byte, i int) uint64 { - b = b[i : i+8 : len(b)] // Help the compiler eliminate bounds checks on the next line. - return uint64(b[0]) | uint64(b[1])<<8 | uint64(b[2])<<16 | uint64(b[3])<<24 | - uint64(b[4])<<32 | uint64(b[5])<<40 | uint64(b[6])<<48 | uint64(b[7])<<56 -} - -// emitLiteral writes a literal chunk and returns the number of bytes written. -// -// It assumes that: -// dst is long enough to hold the encoded bytes -// 1 <= len(lit) && len(lit) <= 65536 -func emitLiteral(dst, lit []byte) int { - i, n := 0, uint(len(lit)-1) - switch { - case n < 60: - dst[0] = uint8(n)<<2 | tagLiteral - i = 1 - case n < 1<<8: - dst[0] = 60<<2 | tagLiteral - dst[1] = uint8(n) - i = 2 - default: - dst[0] = 61<<2 | tagLiteral - dst[1] = uint8(n) - dst[2] = uint8(n >> 8) - i = 3 - } - return i + copy(dst[i:], lit) -} - -// emitCopy writes a copy chunk and returns the number of bytes written. -// -// It assumes that: -// dst is long enough to hold the encoded bytes -// 1 <= offset && offset <= 65535 -// 4 <= length && length <= 65535 -func emitCopy(dst []byte, offset, length int) int { - i := 0 - // The maximum length for a single tagCopy1 or tagCopy2 op is 64 bytes. The - // threshold for this loop is a little higher (at 68 = 64 + 4), and the - // length emitted down below is is a little lower (at 60 = 64 - 4), because - // it's shorter to encode a length 67 copy as a length 60 tagCopy2 followed - // by a length 7 tagCopy1 (which encodes as 3+2 bytes) than to encode it as - // a length 64 tagCopy2 followed by a length 3 tagCopy2 (which encodes as - // 3+3 bytes). The magic 4 in the 64±4 is because the minimum length for a - // tagCopy1 op is 4 bytes, which is why a length 3 copy has to be an - // encodes-as-3-bytes tagCopy2 instead of an encodes-as-2-bytes tagCopy1. - for length >= 68 { - // Emit a length 64 copy, encoded as 3 bytes. - dst[i+0] = 63<<2 | tagCopy2 - dst[i+1] = uint8(offset) - dst[i+2] = uint8(offset >> 8) - i += 3 - length -= 64 - } - if length > 64 { - // Emit a length 60 copy, encoded as 3 bytes. - dst[i+0] = 59<<2 | tagCopy2 - dst[i+1] = uint8(offset) - dst[i+2] = uint8(offset >> 8) - i += 3 - length -= 60 - } - if length >= 12 || offset >= 2048 { - // Emit the remaining copy, encoded as 3 bytes. - dst[i+0] = uint8(length-1)<<2 | tagCopy2 - dst[i+1] = uint8(offset) - dst[i+2] = uint8(offset >> 8) - return i + 3 - } - // Emit the remaining copy, encoded as 2 bytes. - dst[i+0] = uint8(offset>>8)<<5 | uint8(length-4)<<2 | tagCopy1 - dst[i+1] = uint8(offset) - return i + 2 -} - -// extendMatch returns the largest k such that k <= len(src) and that -// src[i:i+k-j] and src[j:k] have the same contents. -// -// It assumes that: -// 0 <= i && i < j && j <= len(src) -func extendMatch(src []byte, i, j int) int { - for ; j < len(src) && src[i] == src[j]; i, j = i+1, j+1 { - } - return j -} - -func hash(u, shift uint32) uint32 { - return (u * 0x1e35a7bd) >> shift -} - -// encodeBlock encodes a non-empty src to a guaranteed-large-enough dst. It -// assumes that the varint-encoded length of the decompressed bytes has already -// been written. -// -// It also assumes that: -// len(dst) >= MaxEncodedLen(len(src)) && -// minNonLiteralBlockSize <= len(src) && len(src) <= maxBlockSize -func encodeBlock(dst, src []byte) (d int) { - // Initialize the hash table. Its size ranges from 1<<8 to 1<<14 inclusive. - // The table element type is uint16, as s < sLimit and sLimit < len(src) - // and len(src) <= maxBlockSize and maxBlockSize == 65536. - const ( - maxTableSize = 1 << 14 - // tableMask is redundant, but helps the compiler eliminate bounds - // checks. - tableMask = maxTableSize - 1 - ) - shift := uint32(32 - 8) - for tableSize := 1 << 8; tableSize < maxTableSize && tableSize < len(src); tableSize *= 2 { - shift-- - } - // In Go, all array elements are zero-initialized, so there is no advantage - // to a smaller tableSize per se. However, it matches the C++ algorithm, - // and in the asm versions of this code, we can get away with zeroing only - // the first tableSize elements. - var table [maxTableSize]uint16 - - // sLimit is when to stop looking for offset/length copies. The inputMargin - // lets us use a fast path for emitLiteral in the main loop, while we are - // looking for copies. - sLimit := len(src) - inputMargin - - // nextEmit is where in src the next emitLiteral should start from. - nextEmit := 0 - - // The encoded form must start with a literal, as there are no previous - // bytes to copy, so we start looking for hash matches at s == 1. - s := 1 - nextHash := hash(load32(src, s), shift) - - for { - // Copied from the C++ snappy implementation: - // - // Heuristic match skipping: If 32 bytes are scanned with no matches - // found, start looking only at every other byte. If 32 more bytes are - // scanned (or skipped), look at every third byte, etc.. When a match - // is found, immediately go back to looking at every byte. This is a - // small loss (~5% performance, ~0.1% density) for compressible data - // due to more bookkeeping, but for non-compressible data (such as - // JPEG) it's a huge win since the compressor quickly "realizes" the - // data is incompressible and doesn't bother looking for matches - // everywhere. - // - // The "skip" variable keeps track of how many bytes there are since - // the last match; dividing it by 32 (ie. right-shifting by five) gives - // the number of bytes to move ahead for each iteration. - skip := 32 - - nextS := s - candidate := 0 - for { - s = nextS - bytesBetweenHashLookups := skip >> 5 - nextS = s + bytesBetweenHashLookups - skip += bytesBetweenHashLookups - if nextS > sLimit { - goto emitRemainder - } - candidate = int(table[nextHash&tableMask]) - table[nextHash&tableMask] = uint16(s) - nextHash = hash(load32(src, nextS), shift) - if load32(src, s) == load32(src, candidate) { - break - } - } - - // A 4-byte match has been found. We'll later see if more than 4 bytes - // match. But, prior to the match, src[nextEmit:s] are unmatched. Emit - // them as literal bytes. - d += emitLiteral(dst[d:], src[nextEmit:s]) - - // Call emitCopy, and then see if another emitCopy could be our next - // move. Repeat until we find no match for the input immediately after - // what was consumed by the last emitCopy call. - // - // If we exit this loop normally then we need to call emitLiteral next, - // though we don't yet know how big the literal will be. We handle that - // by proceeding to the next iteration of the main loop. We also can - // exit this loop via goto if we get close to exhausting the input. - for { - // Invariant: we have a 4-byte match at s, and no need to emit any - // literal bytes prior to s. - base := s - - // Extend the 4-byte match as long as possible. - // - // This is an inlined version of: - // s = extendMatch(src, candidate+4, s+4) - s += 4 - for i := candidate + 4; s < len(src) && src[i] == src[s]; i, s = i+1, s+1 { - } - - d += emitCopy(dst[d:], base-candidate, s-base) - nextEmit = s - if s >= sLimit { - goto emitRemainder - } - - // We could immediately start working at s now, but to improve - // compression we first update the hash table at s-1 and at s. If - // another emitCopy is not our next move, also calculate nextHash - // at s+1. At least on GOARCH=amd64, these three hash calculations - // are faster as one load64 call (with some shifts) instead of - // three load32 calls. - x := load64(src, s-1) - prevHash := hash(uint32(x>>0), shift) - table[prevHash&tableMask] = uint16(s - 1) - currHash := hash(uint32(x>>8), shift) - candidate = int(table[currHash&tableMask]) - table[currHash&tableMask] = uint16(s) - if uint32(x>>8) != load32(src, candidate) { - nextHash = hash(uint32(x>>16), shift) - s++ - break - } - } - } - -emitRemainder: - if nextEmit < len(src) { - d += emitLiteral(dst[d:], src[nextEmit:]) - } - return d -} diff --git a/vendor/github.com/google/pprof/profile/encode.go b/vendor/github.com/google/pprof/profile/encode.go index 860bb304c..8ce9d3cf3 100644 --- a/vendor/github.com/google/pprof/profile/encode.go +++ b/vendor/github.com/google/pprof/profile/encode.go @@ -122,6 +122,7 @@ func (p *Profile) preEncode() { } p.defaultSampleTypeX = addString(strings, p.DefaultSampleType) + p.docURLX = addString(strings, p.DocURL) p.stringTable = make([]string, len(strings)) for s, i := range strings { @@ -156,6 +157,7 @@ func (p *Profile) encode(b *buffer) { encodeInt64Opt(b, 12, p.Period) encodeInt64s(b, 13, p.commentX) encodeInt64(b, 14, p.defaultSampleTypeX) + encodeInt64Opt(b, 15, p.docURLX) } var profileDecoder = []decoder{ @@ -237,6 +239,8 @@ var profileDecoder = []decoder{ func(b *buffer, m message) error { return decodeInt64s(b, &m.(*Profile).commentX) }, // int64 defaultSampleType = 14 func(b *buffer, m message) error { return decodeInt64(b, &m.(*Profile).defaultSampleTypeX) }, + // string doc_link = 15; + func(b *buffer, m message) error { return decodeInt64(b, &m.(*Profile).docURLX) }, } // postDecode takes the unexported fields populated by decode (with @@ -384,6 +388,7 @@ func (p *Profile) postDecode() error { p.commentX = nil p.DefaultSampleType, err = getString(p.stringTable, &p.defaultSampleTypeX, err) + p.DocURL, err = getString(p.stringTable, &p.docURLX, err) p.stringTable = nil return err } diff --git a/vendor/github.com/google/pprof/profile/merge.go b/vendor/github.com/google/pprof/profile/merge.go index eee0132e7..ba4d74640 100644 --- a/vendor/github.com/google/pprof/profile/merge.go +++ b/vendor/github.com/google/pprof/profile/merge.go @@ -476,6 +476,7 @@ func combineHeaders(srcs []*Profile) (*Profile, error) { var timeNanos, durationNanos, period int64 var comments []string seenComments := map[string]bool{} + var docURL string var defaultSampleType string for _, s := range srcs { if timeNanos == 0 || s.TimeNanos < timeNanos { @@ -494,6 +495,9 @@ func combineHeaders(srcs []*Profile) (*Profile, error) { if defaultSampleType == "" { defaultSampleType = s.DefaultSampleType } + if docURL == "" { + docURL = s.DocURL + } } p := &Profile{ @@ -509,6 +513,7 @@ func combineHeaders(srcs []*Profile) (*Profile, error) { Comments: comments, DefaultSampleType: defaultSampleType, + DocURL: docURL, } copy(p.SampleType, srcs[0].SampleType) return p, nil diff --git a/vendor/github.com/google/pprof/profile/profile.go b/vendor/github.com/google/pprof/profile/profile.go index 5551eb0bf..0983656c2 100644 --- a/vendor/github.com/google/pprof/profile/profile.go +++ b/vendor/github.com/google/pprof/profile/profile.go @@ -39,6 +39,7 @@ type Profile struct { Location []*Location Function []*Function Comments []string + DocURL string DropFrames string KeepFrames string @@ -53,6 +54,7 @@ type Profile struct { encodeMu sync.Mutex commentX []int64 + docURLX int64 dropFramesX int64 keepFramesX int64 stringTable []string @@ -555,6 +557,9 @@ func (p *Profile) String() string { for _, c := range p.Comments { ss = append(ss, "Comment: "+c) } + if url := p.DocURL; url != "" { + ss = append(ss, fmt.Sprintf("Doc: %s", url)) + } if pt := p.PeriodType; pt != nil { ss = append(ss, fmt.Sprintf("PeriodType: %s %s", pt.Type, pt.Unit)) } diff --git a/vendor/github.com/klauspost/compress/internal/race/norace.go b/vendor/github.com/klauspost/compress/internal/race/norace.go new file mode 100644 index 000000000..affbbbb59 --- /dev/null +++ b/vendor/github.com/klauspost/compress/internal/race/norace.go @@ -0,0 +1,13 @@ +// Copyright 2015 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build !race + +package race + +func ReadSlice[T any](s []T) { +} + +func WriteSlice[T any](s []T) { +} diff --git a/vendor/github.com/klauspost/compress/internal/race/race.go b/vendor/github.com/klauspost/compress/internal/race/race.go new file mode 100644 index 000000000..f5e240dcd --- /dev/null +++ b/vendor/github.com/klauspost/compress/internal/race/race.go @@ -0,0 +1,26 @@ +// Copyright 2015 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build race + +package race + +import ( + "runtime" + "unsafe" +) + +func ReadSlice[T any](s []T) { + if len(s) == 0 { + return + } + runtime.RaceReadRange(unsafe.Pointer(&s[0]), len(s)*int(unsafe.Sizeof(s[0]))) +} + +func WriteSlice[T any](s []T) { + if len(s) == 0 { + return + } + runtime.RaceWriteRange(unsafe.Pointer(&s[0]), len(s)*int(unsafe.Sizeof(s[0]))) +} diff --git a/vendor/github.com/klauspost/compress/s2/.gitignore b/vendor/github.com/klauspost/compress/s2/.gitignore new file mode 100644 index 000000000..3a89c6e3e --- /dev/null +++ b/vendor/github.com/klauspost/compress/s2/.gitignore @@ -0,0 +1,15 @@ +testdata/bench + +# These explicitly listed benchmark data files are for an obsolete version of +# snappy_test.go. +testdata/alice29.txt +testdata/asyoulik.txt +testdata/fireworks.jpeg +testdata/geo.protodata +testdata/html +testdata/html_x_4 +testdata/kppkn.gtb +testdata/lcet10.txt +testdata/paper-100k.pdf +testdata/plrabn12.txt +testdata/urls.10K diff --git a/vendor/github.com/klauspost/compress/s2/LICENSE b/vendor/github.com/klauspost/compress/s2/LICENSE new file mode 100644 index 000000000..1d2d645bd --- /dev/null +++ b/vendor/github.com/klauspost/compress/s2/LICENSE @@ -0,0 +1,28 @@ +Copyright (c) 2011 The Snappy-Go Authors. All rights reserved. +Copyright (c) 2019 Klaus Post. All rights reserved. + +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions are +met: + + * Redistributions of source code must retain the above copyright +notice, this list of conditions and the following disclaimer. + * Redistributions in binary form must reproduce the above +copyright notice, this list of conditions and the following disclaimer +in the documentation and/or other materials provided with the +distribution. + * Neither the name of Google Inc. nor the names of its +contributors may be used to endorse or promote products derived from +this software without specific prior written permission. + +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS +"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT +LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR +A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT +OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, +SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT +LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, +DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY +THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE +OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. diff --git a/vendor/github.com/klauspost/compress/s2/README.md b/vendor/github.com/klauspost/compress/s2/README.md new file mode 100644 index 000000000..8284bb081 --- /dev/null +++ b/vendor/github.com/klauspost/compress/s2/README.md @@ -0,0 +1,1120 @@ +# S2 Compression + +S2 is an extension of [Snappy](https://github.com/google/snappy). + +S2 is aimed for high throughput, which is why it features concurrent compression for bigger payloads. + +Decoding is compatible with Snappy compressed content, but content compressed with S2 cannot be decompressed by Snappy. +This means that S2 can seamlessly replace Snappy without converting compressed content. + +S2 can produce Snappy compatible output, faster and better than Snappy. +If you want full benefit of the changes you should use s2 without Snappy compatibility. + +S2 is designed to have high throughput on content that cannot be compressed. +This is important, so you don't have to worry about spending CPU cycles on already compressed data. + +## Benefits over Snappy + +* Better compression +* Adjustable compression (3 levels) +* Concurrent stream compression +* Faster decompression, even for Snappy compatible content +* Concurrent Snappy/S2 stream decompression +* Skip forward in compressed stream +* Random seeking with indexes +* Compatible with reading Snappy compressed content +* Smaller block size overhead on incompressible blocks +* Block concatenation +* Block Dictionary support +* Uncompressed stream mode +* Automatic stream size padding +* Snappy compatible block compression + +## Drawbacks over Snappy + +* Not optimized for 32 bit systems +* Streams use slightly more memory due to larger blocks and concurrency (configurable) + +# Usage + +Installation: `go get -u github.com/klauspost/compress/s2` + +Full package documentation: + +[![godoc][1]][2] + +[1]: https://godoc.org/github.com/klauspost/compress?status.svg +[2]: https://godoc.org/github.com/klauspost/compress/s2 + +## Compression + +```Go +func EncodeStream(src io.Reader, dst io.Writer) error { + enc := s2.NewWriter(dst) + _, err := io.Copy(enc, src) + if err != nil { + enc.Close() + return err + } + // Blocks until compression is done. + return enc.Close() +} +``` + +You should always call `enc.Close()`, otherwise you will leak resources and your encode will be incomplete. + +For the best throughput, you should attempt to reuse the `Writer` using the `Reset()` method. + +The Writer in S2 is always buffered, therefore `NewBufferedWriter` in Snappy can be replaced with `NewWriter` in S2. +It is possible to flush any buffered data using the `Flush()` method. +This will block until all data sent to the encoder has been written to the output. + +S2 also supports the `io.ReaderFrom` interface, which will consume all input from a reader. + +As a final method to compress data, if you have a single block of data you would like to have encoded as a stream, +a slightly more efficient method is to use the `EncodeBuffer` method. +This will take ownership of the buffer until the stream is closed. + +```Go +func EncodeStream(src []byte, dst io.Writer) error { + enc := s2.NewWriter(dst) + // The encoder owns the buffer until Flush or Close is called. + err := enc.EncodeBuffer(buf) + if err != nil { + enc.Close() + return err + } + // Blocks until compression is done. + return enc.Close() +} +``` + +Each call to `EncodeBuffer` will result in discrete blocks being created without buffering, +so it should only be used a single time per stream. +If you need to write several blocks, you should use the regular io.Writer interface. + + +## Decompression + +```Go +func DecodeStream(src io.Reader, dst io.Writer) error { + dec := s2.NewReader(src) + _, err := io.Copy(dst, dec) + return err +} +``` + +Similar to the Writer, a Reader can be reused using the `Reset` method. + +For the best possible throughput, there is a `EncodeBuffer(buf []byte)` function available. +However, it requires that the provided buffer isn't used after it is handed over to S2 and until the stream is flushed or closed. + +For smaller data blocks, there is also a non-streaming interface: `Encode()`, `EncodeBetter()` and `Decode()`. +Do however note that these functions (similar to Snappy) does not provide validation of data, +so data corruption may be undetected. Stream encoding provides CRC checks of data. + +It is possible to efficiently skip forward in a compressed stream using the `Skip()` method. +For big skips the decompressor is able to skip blocks without decompressing them. + +## Single Blocks + +Similar to Snappy S2 offers single block compression. +Blocks do not offer the same flexibility and safety as streams, +but may be preferable for very small payloads, less than 100K. + +Using a simple `dst := s2.Encode(nil, src)` will compress `src` and return the compressed result. +It is possible to provide a destination buffer. +If the buffer has a capacity of `s2.MaxEncodedLen(len(src))` it will be used. +If not a new will be allocated. + +Alternatively `EncodeBetter`/`EncodeBest` can also be used for better, but slightly slower compression. + +Similarly to decompress a block you can use `dst, err := s2.Decode(nil, src)`. +Again an optional destination buffer can be supplied. +The `s2.DecodedLen(src)` can be used to get the minimum capacity needed. +If that is not satisfied a new buffer will be allocated. + +Block function always operate on a single goroutine since it should only be used for small payloads. + +# Commandline tools + +Some very simply commandline tools are provided; `s2c` for compression and `s2d` for decompression. + +Binaries can be downloaded on the [Releases Page](https://github.com/klauspost/compress/releases). + +Installing then requires Go to be installed. To install them, use: + +`go install github.com/klauspost/compress/s2/cmd/s2c@latest && go install github.com/klauspost/compress/s2/cmd/s2d@latest` + +To build binaries to the current folder use: + +`go build github.com/klauspost/compress/s2/cmd/s2c && go build github.com/klauspost/compress/s2/cmd/s2d` + + +## s2c + +``` +Usage: s2c [options] file1 file2 + +Compresses all files supplied as input separately. +Output files are written as 'filename.ext.s2' or 'filename.ext.snappy'. +By default output files will be overwritten. +Use - as the only file name to read from stdin and write to stdout. + +Wildcards are accepted: testdir/*.txt will compress all files in testdir ending with .txt +Directories can be wildcards as well. testdir/*/*.txt will match testdir/subdir/b.txt + +File names beginning with 'http://' and 'https://' will be downloaded and compressed. +Only http response code 200 is accepted. + +Options: + -bench int + Run benchmark n times. No output will be written + -blocksize string + Max block size. Examples: 64K, 256K, 1M, 4M. Must be power of two and <= 4MB (default "4M") + -c Write all output to stdout. Multiple input files will be concatenated + -cpu int + Compress using this amount of threads (default 32) + -faster + Compress faster, but with a minor compression loss + -help + Display help + -index + Add seek index (default true) + -o string + Write output to another file. Single input file only + -pad string + Pad size to a multiple of this value, Examples: 500, 64K, 256K, 1M, 4M, etc (default "1") + -q Don't write any output to terminal, except errors + -rm + Delete source file(s) after successful compression + -safe + Do not overwrite output files + -slower + Compress more, but a lot slower + -snappy + Generate Snappy compatible output stream + -verify + Verify written files + +``` + +## s2d + +``` +Usage: s2d [options] file1 file2 + +Decompresses all files supplied as input. Input files must end with '.s2' or '.snappy'. +Output file names have the extension removed. By default output files will be overwritten. +Use - as the only file name to read from stdin and write to stdout. + +Wildcards are accepted: testdir/*.txt will compress all files in testdir ending with .txt +Directories can be wildcards as well. testdir/*/*.txt will match testdir/subdir/b.txt + +File names beginning with 'http://' and 'https://' will be downloaded and decompressed. +Extensions on downloaded files are ignored. Only http response code 200 is accepted. + +Options: + -bench int + Run benchmark n times. No output will be written + -c Write all output to stdout. Multiple input files will be concatenated + -help + Display help + -o string + Write output to another file. Single input file only + -offset string + Start at offset. Examples: 92, 64K, 256K, 1M, 4M. Requires Index + -q Don't write any output to terminal, except errors + -rm + Delete source file(s) after successful decompression + -safe + Do not overwrite output files + -tail string + Return last of compressed file. Examples: 92, 64K, 256K, 1M, 4M. Requires Index + -verify + Verify files, but do not write output +``` + +## s2sx: self-extracting archives + +s2sx allows creating self-extracting archives with no dependencies. + +By default, executables are created for the same platforms as the host os, +but this can be overridden with `-os` and `-arch` parameters. + +Extracted files have 0666 permissions, except when untar option used. + +``` +Usage: s2sx [options] file1 file2 + +Compresses all files supplied as input separately. +If files have '.s2' extension they are assumed to be compressed already. +Output files are written as 'filename.s2sx' and with '.exe' for windows targets. +If output is big, an additional file with ".more" is written. This must be included as well. +By default output files will be overwritten. + +Wildcards are accepted: testdir/*.txt will compress all files in testdir ending with .txt +Directories can be wildcards as well. testdir/*/*.txt will match testdir/subdir/b.txt + +Options: + -arch string + Destination architecture (default "amd64") + -c Write all output to stdout. Multiple input files will be concatenated + -cpu int + Compress using this amount of threads (default 32) + -help + Display help + -max string + Maximum executable size. Rest will be written to another file. (default "1G") + -os string + Destination operating system (default "windows") + -q Don't write any output to terminal, except errors + -rm + Delete source file(s) after successful compression + -safe + Do not overwrite output files + -untar + Untar on destination +``` + +Available platforms are: + + * darwin-amd64 + * darwin-arm64 + * linux-amd64 + * linux-arm + * linux-arm64 + * linux-mips64 + * linux-ppc64le + * windows-386 + * windows-amd64 + +By default, there is a size limit of 1GB for the output executable. + +When this is exceeded the remaining file content is written to a file called +output+`.more`. This file must be included for a successful extraction and +placed alongside the executable for a successful extraction. + +This file *must* have the same name as the executable, so if the executable is renamed, +so must the `.more` file. + +This functionality is disabled with stdin/stdout. + +### Self-extracting TAR files + +If you wrap a TAR file you can specify `-untar` to make it untar on the destination host. + +Files are extracted to the current folder with the path specified in the tar file. + +Note that tar files are not validated before they are wrapped. + +For security reasons files that move below the root folder are not allowed. + +# Performance + +This section will focus on comparisons to Snappy. +This package is solely aimed at replacing Snappy as a high speed compression package. +If you are mainly looking for better compression [zstandard](https://github.com/klauspost/compress/tree/master/zstd#zstd) +gives better compression, but typically at speeds slightly below "better" mode in this package. + +Compression is increased compared to Snappy, mostly around 5-20% and the throughput is typically 25-40% increased (single threaded) compared to the Snappy Go implementation. + +Streams are concurrently compressed. The stream will be distributed among all available CPU cores for the best possible throughput. + +A "better" compression mode is also available. This allows to trade a bit of speed for a minor compression gain. +The content compressed in this mode is fully compatible with the standard decoder. + +Snappy vs S2 **compression** speed on 16 core (32 thread) computer, using all threads and a single thread (1 CPU): + +| File | S2 Speed | S2 Throughput | S2 % smaller | S2 "better" | "better" throughput | "better" % smaller | +|---------------------------------------------------------------------------------------------------------|----------|---------------|--------------|-------------|---------------------|--------------------| +| [rawstudio-mint14.tar](https://files.klauspost.com/compress/rawstudio-mint14.7z) | 16.33x | 10556 MB/s | 8.0% | 6.04x | 5252 MB/s | 14.7% | +| (1 CPU) | 1.08x | 940 MB/s | - | 0.46x | 400 MB/s | - | +| [github-june-2days-2019.json](https://files.klauspost.com/compress/github-june-2days-2019.json.zst) | 16.51x | 15224 MB/s | 31.70% | 9.47x | 8734 MB/s | 37.71% | +| (1 CPU) | 1.26x | 1157 MB/s | - | 0.60x | 556 MB/s | - | +| [github-ranks-backup.bin](https://files.klauspost.com/compress/github-ranks-backup.bin.zst) | 15.14x | 12598 MB/s | -5.76% | 6.23x | 5675 MB/s | 3.62% | +| (1 CPU) | 1.02x | 932 MB/s | - | 0.47x | 432 MB/s | - | +| [consensus.db.10gb](https://files.klauspost.com/compress/consensus.db.10gb.zst) | 11.21x | 12116 MB/s | 15.95% | 3.24x | 3500 MB/s | 18.00% | +| (1 CPU) | 1.05x | 1135 MB/s | - | 0.27x | 292 MB/s | - | +| [apache.log](https://files.klauspost.com/compress/apache.log.zst) | 8.55x | 16673 MB/s | 20.54% | 5.85x | 11420 MB/s | 24.97% | +| (1 CPU) | 1.91x | 1771 MB/s | - | 0.53x | 1041 MB/s | - | +| [gob-stream](https://files.klauspost.com/compress/gob-stream.7z) | 15.76x | 14357 MB/s | 24.01% | 8.67x | 7891 MB/s | 33.68% | +| (1 CPU) | 1.17x | 1064 MB/s | - | 0.65x | 595 MB/s | - | +| [10gb.tar](http://mattmahoney.net/dc/10gb.html) | 13.33x | 9835 MB/s | 2.34% | 6.85x | 4863 MB/s | 9.96% | +| (1 CPU) | 0.97x | 689 MB/s | - | 0.55x | 387 MB/s | - | +| sharnd.out.2gb | 9.11x | 13213 MB/s | 0.01% | 1.49x | 9184 MB/s | 0.01% | +| (1 CPU) | 0.88x | 5418 MB/s | - | 0.77x | 5417 MB/s | - | +| [sofia-air-quality-dataset csv](https://files.klauspost.com/compress/sofia-air-quality-dataset.tar.zst) | 22.00x | 11477 MB/s | 18.73% | 11.15x | 5817 MB/s | 27.88% | +| (1 CPU) | 1.23x | 642 MB/s | - | 0.71x | 642 MB/s | - | +| [silesia.tar](http://sun.aei.polsl.pl/~sdeor/corpus/silesia.zip) | 11.23x | 6520 MB/s | 5.9% | 5.35x | 3109 MB/s | 15.88% | +| (1 CPU) | 1.05x | 607 MB/s | - | 0.52x | 304 MB/s | - | +| [enwik9](https://files.klauspost.com/compress/enwik9.zst) | 19.28x | 8440 MB/s | 4.04% | 9.31x | 4076 MB/s | 18.04% | +| (1 CPU) | 1.12x | 488 MB/s | - | 0.57x | 250 MB/s | - | + +### Legend + +* `S2 Speed`: Speed of S2 compared to Snappy, using 16 cores and 1 core. +* `S2 Throughput`: Throughput of S2 in MB/s. +* `S2 % smaller`: How many percent of the Snappy output size is S2 better. +* `S2 "better"`: Speed when enabling "better" compression mode in S2 compared to Snappy. +* `"better" throughput`: Speed when enabling "better" compression mode in S2 compared to Snappy. +* `"better" % smaller`: How many percent of the Snappy output size is S2 better when using "better" compression. + +There is a good speedup across the board when using a single thread and a significant speedup when using multiple threads. + +Machine generated data gets by far the biggest compression boost, with size being reduced by up to 35% of Snappy size. + +The "better" compression mode sees a good improvement in all cases, but usually at a performance cost. + +Incompressible content (`sharnd.out.2gb`, 2GB random data) sees the smallest speedup. +This is likely dominated by synchronization overhead, which is confirmed by the fact that single threaded performance is higher (see above). + +## Decompression + +S2 attempts to create content that is also fast to decompress, except in "better" mode where the smallest representation is used. + +S2 vs Snappy **decompression** speed. Both operating on single core: + +| File | S2 Throughput | vs. Snappy | Better Throughput | vs. Snappy | +|-----------------------------------------------------------------------------------------------------|---------------|------------|-------------------|------------| +| [rawstudio-mint14.tar](https://files.klauspost.com/compress/rawstudio-mint14.7z) | 2117 MB/s | 1.14x | 1738 MB/s | 0.94x | +| [github-june-2days-2019.json](https://files.klauspost.com/compress/github-june-2days-2019.json.zst) | 2401 MB/s | 1.25x | 2307 MB/s | 1.20x | +| [github-ranks-backup.bin](https://files.klauspost.com/compress/github-ranks-backup.bin.zst) | 2075 MB/s | 0.98x | 1764 MB/s | 0.83x | +| [consensus.db.10gb](https://files.klauspost.com/compress/consensus.db.10gb.zst) | 2967 MB/s | 1.05x | 2885 MB/s | 1.02x | +| [adresser.json](https://files.klauspost.com/compress/adresser.json.zst) | 4141 MB/s | 1.07x | 4184 MB/s | 1.08x | +| [gob-stream](https://files.klauspost.com/compress/gob-stream.7z) | 2264 MB/s | 1.12x | 2185 MB/s | 1.08x | +| [10gb.tar](http://mattmahoney.net/dc/10gb.html) | 1525 MB/s | 1.03x | 1347 MB/s | 0.91x | +| sharnd.out.2gb | 3813 MB/s | 0.79x | 3900 MB/s | 0.81x | +| [enwik9](http://mattmahoney.net/dc/textdata.html) | 1246 MB/s | 1.29x | 967 MB/s | 1.00x | +| [silesia.tar](http://sun.aei.polsl.pl/~sdeor/corpus/silesia.zip) | 1433 MB/s | 1.12x | 1203 MB/s | 0.94x | +| [enwik10](https://encode.su/threads/3315-enwik10-benchmark-results) | 1284 MB/s | 1.32x | 1010 MB/s | 1.04x | + +### Legend + +* `S2 Throughput`: Decompression speed of S2 encoded content. +* `Better Throughput`: Decompression speed of S2 "better" encoded content. +* `vs Snappy`: Decompression speed of S2 "better" mode compared to Snappy and absolute speed. + + +While the decompression code hasn't changed, there is a significant speedup in decompression speed. +S2 prefers longer matches and will typically only find matches that are 6 bytes or longer. +While this reduces compression a bit, it improves decompression speed. + +The "better" compression mode will actively look for shorter matches, which is why it has a decompression speed quite similar to Snappy. + +Without assembly decompression is also very fast; single goroutine decompression speed. No assembly: + +| File | S2 Throughput | S2 throughput | +|--------------------------------|---------------|---------------| +| consensus.db.10gb.s2 | 1.84x | 2289.8 MB/s | +| 10gb.tar.s2 | 1.30x | 867.07 MB/s | +| rawstudio-mint14.tar.s2 | 1.66x | 1329.65 MB/s | +| github-june-2days-2019.json.s2 | 2.36x | 1831.59 MB/s | +| github-ranks-backup.bin.s2 | 1.73x | 1390.7 MB/s | +| enwik9.s2 | 1.67x | 681.53 MB/s | +| adresser.json.s2 | 3.41x | 4230.53 MB/s | +| silesia.tar.s2 | 1.52x | 811.58 | + +Even though S2 typically compresses better than Snappy, decompression speed is always better. + +### Concurrent Stream Decompression + +For full stream decompression S2 offers a [DecodeConcurrent](https://pkg.go.dev/github.com/klauspost/compress/s2#Reader.DecodeConcurrent) +that will decode a full stream using multiple goroutines. + +Example scaling, AMD Ryzen 3950X, 16 cores, decompression using `s2d -bench=3 `, best of 3: + +| Input | `-cpu=1` | `-cpu=2` | `-cpu=4` | `-cpu=8` | `-cpu=16` | +|-------------------------------------------|------------|------------|------------|------------|-------------| +| enwik10.snappy | 1098.6MB/s | 1819.8MB/s | 3625.6MB/s | 6910.6MB/s | 10818.2MB/s | +| enwik10.s2 | 1303.5MB/s | 2606.1MB/s | 4847.9MB/s | 8878.4MB/s | 9592.1MB/s | +| sofia-air-quality-dataset.tar.snappy | 1302.0MB/s | 2165.0MB/s | 4244.5MB/s | 8241.0MB/s | 12920.5MB/s | +| sofia-air-quality-dataset.tar.s2 | 1399.2MB/s | 2463.2MB/s | 5196.5MB/s | 9639.8MB/s | 11439.5MB/s | +| sofia-air-quality-dataset.tar.s2 (no asm) | 837.5MB/s | 1652.6MB/s | 3183.6MB/s | 5945.0MB/s | 9620.7MB/s | + +Scaling can be expected to be pretty linear until memory bandwidth is saturated. + +For now the DecodeConcurrent can only be used for full streams without seeking or combining with regular reads. + +## Block compression + + +When compressing blocks no concurrent compression is performed just as Snappy. +This is because blocks are for smaller payloads and generally will not benefit from concurrent compression. + +An important change is that incompressible blocks will not be more than at most 10 bytes bigger than the input. +In rare, worst case scenario Snappy blocks could be significantly bigger than the input. + +### Mixed content blocks + +The most reliable is a wide dataset. +For this we use [`webdevdata.org-2015-01-07-subset`](https://files.klauspost.com/compress/webdevdata.org-2015-01-07-4GB-subset.7z), +53927 files, total input size: 4,014,735,833 bytes. Single goroutine used. + +| * | Input | Output | Reduction | MB/s | +|-------------------|------------|------------|------------|------------| +| S2 | 4014735833 | 1059723369 | 73.60% | **936.73** | +| S2 Better | 4014735833 | 961580539 | 76.05% | 451.10 | +| S2 Best | 4014735833 | 899182886 | **77.60%** | 46.84 | +| Snappy | 4014735833 | 1128706759 | 71.89% | 790.15 | +| S2, Snappy Output | 4014735833 | 1093823291 | 72.75% | 936.60 | +| LZ4 | 4014735833 | 1063768713 | 73.50% | 452.02 | + +S2 delivers both the best single threaded throughput with regular mode and the best compression rate with "best". +"Better" mode provides the same compression speed as LZ4 with better compression ratio. + +When outputting Snappy compatible output it still delivers better throughput (150MB/s more) and better compression. + +As can be seen from the other benchmarks decompression should also be easier on the S2 generated output. + +Though they cannot be compared due to different decompression speeds here are the speed/size comparisons for +other Go compressors: + +| * | Input | Output | Reduction | MB/s | +|-------------------|------------|------------|-----------|--------| +| Zstd Fastest (Go) | 4014735833 | 794608518 | 80.21% | 236.04 | +| Zstd Best (Go) | 4014735833 | 704603356 | 82.45% | 35.63 | +| Deflate (Go) l1 | 4014735833 | 871294239 | 78.30% | 214.04 | +| Deflate (Go) l9 | 4014735833 | 730389060 | 81.81% | 41.17 | + +### Standard block compression + +Benchmarking single block performance is subject to a lot more variation since it only tests a limited number of file patterns. +So individual benchmarks should only be seen as a guideline and the overall picture is more important. + +These micro-benchmarks are with data in cache and trained branch predictors. For a more realistic benchmark see the mixed content above. + +Block compression. Parallel benchmark running on 16 cores, 16 goroutines. + +AMD64 assembly is use for both S2 and Snappy. + +| Absolute Perf | Snappy size | S2 Size | Snappy Speed | S2 Speed | Snappy dec | S2 dec | +|-----------------------|-------------|---------|--------------|-------------|-------------|-------------| +| html | 22843 | 20868 | 16246 MB/s | 18617 MB/s | 40972 MB/s | 49263 MB/s | +| urls.10K | 335492 | 286541 | 7943 MB/s | 10201 MB/s | 22523 MB/s | 26484 MB/s | +| fireworks.jpeg | 123034 | 123100 | 349544 MB/s | 303228 MB/s | 718321 MB/s | 827552 MB/s | +| fireworks.jpeg (200B) | 146 | 155 | 8869 MB/s | 20180 MB/s | 33691 MB/s | 52421 MB/s | +| paper-100k.pdf | 85304 | 84202 | 167546 MB/s | 112988 MB/s | 326905 MB/s | 291944 MB/s | +| html_x_4 | 92234 | 20870 | 15194 MB/s | 54457 MB/s | 30843 MB/s | 32217 MB/s | +| alice29.txt | 88034 | 85934 | 5936 MB/s | 6540 MB/s | 12882 MB/s | 20044 MB/s | +| asyoulik.txt | 77503 | 79575 | 5517 MB/s | 6657 MB/s | 12735 MB/s | 22806 MB/s | +| lcet10.txt | 234661 | 220383 | 6235 MB/s | 6303 MB/s | 14519 MB/s | 18697 MB/s | +| plrabn12.txt | 319267 | 318196 | 5159 MB/s | 6074 MB/s | 11923 MB/s | 19901 MB/s | +| geo.protodata | 23335 | 18606 | 21220 MB/s | 25432 MB/s | 56271 MB/s | 62540 MB/s | +| kppkn.gtb | 69526 | 65019 | 9732 MB/s | 8905 MB/s | 18491 MB/s | 18969 MB/s | +| alice29.txt (128B) | 80 | 82 | 6691 MB/s | 17179 MB/s | 31883 MB/s | 38874 MB/s | +| alice29.txt (1000B) | 774 | 774 | 12204 MB/s | 13273 MB/s | 48056 MB/s | 52341 MB/s | +| alice29.txt (10000B) | 6648 | 6933 | 10044 MB/s | 12824 MB/s | 32378 MB/s | 46322 MB/s | +| alice29.txt (20000B) | 12686 | 13516 | 7733 MB/s | 12160 MB/s | 30566 MB/s | 58969 MB/s | + + +Speed is generally at or above Snappy. Small blocks gets a significant speedup, although at the expense of size. + +Decompression speed is better than Snappy, except in one case. + +Since payloads are very small the variance in terms of size is rather big, so they should only be seen as a general guideline. + +Size is on average around Snappy, but varies on content type. +In cases where compression is worse, it usually is compensated by a speed boost. + + +### Better compression + +Benchmarking single block performance is subject to a lot more variation since it only tests a limited number of file patterns. +So individual benchmarks should only be seen as a guideline and the overall picture is more important. + +| Absolute Perf | Snappy size | Better Size | Snappy Speed | Better Speed | Snappy dec | Better dec | +|-----------------------|-------------|-------------|--------------|--------------|-------------|-------------| +| html | 22843 | 18972 | 16246 MB/s | 8621 MB/s | 40972 MB/s | 40292 MB/s | +| urls.10K | 335492 | 248079 | 7943 MB/s | 5104 MB/s | 22523 MB/s | 20981 MB/s | +| fireworks.jpeg | 123034 | 123100 | 349544 MB/s | 84429 MB/s | 718321 MB/s | 823698 MB/s | +| fireworks.jpeg (200B) | 146 | 149 | 8869 MB/s | 7125 MB/s | 33691 MB/s | 30101 MB/s | +| paper-100k.pdf | 85304 | 82887 | 167546 MB/s | 11087 MB/s | 326905 MB/s | 198869 MB/s | +| html_x_4 | 92234 | 18982 | 15194 MB/s | 29316 MB/s | 30843 MB/s | 30937 MB/s | +| alice29.txt | 88034 | 71611 | 5936 MB/s | 3709 MB/s | 12882 MB/s | 16611 MB/s | +| asyoulik.txt | 77503 | 65941 | 5517 MB/s | 3380 MB/s | 12735 MB/s | 14975 MB/s | +| lcet10.txt | 234661 | 184939 | 6235 MB/s | 3537 MB/s | 14519 MB/s | 16634 MB/s | +| plrabn12.txt | 319267 | 264990 | 5159 MB/s | 2960 MB/s | 11923 MB/s | 13382 MB/s | +| geo.protodata | 23335 | 17689 | 21220 MB/s | 10859 MB/s | 56271 MB/s | 57961 MB/s | +| kppkn.gtb | 69526 | 55398 | 9732 MB/s | 5206 MB/s | 18491 MB/s | 16524 MB/s | +| alice29.txt (128B) | 80 | 78 | 6691 MB/s | 7422 MB/s | 31883 MB/s | 34225 MB/s | +| alice29.txt (1000B) | 774 | 746 | 12204 MB/s | 5734 MB/s | 48056 MB/s | 42068 MB/s | +| alice29.txt (10000B) | 6648 | 6218 | 10044 MB/s | 6055 MB/s | 32378 MB/s | 28813 MB/s | +| alice29.txt (20000B) | 12686 | 11492 | 7733 MB/s | 3143 MB/s | 30566 MB/s | 27315 MB/s | + + +Except for the mostly incompressible JPEG image compression is better and usually in the +double digits in terms of percentage reduction over Snappy. + +The PDF sample shows a significant slowdown compared to Snappy, as this mode tries harder +to compress the data. Very small blocks are also not favorable for better compression, so throughput is way down. + +This mode aims to provide better compression at the expense of performance and achieves that +without a huge performance penalty, except on very small blocks. + +Decompression speed suffers a little compared to the regular S2 mode, +but still manages to be close to Snappy in spite of increased compression. + +# Best compression mode + +S2 offers a "best" compression mode. + +This will compress as much as possible with little regard to CPU usage. + +Mainly for offline compression, but where decompression speed should still +be high and compatible with other S2 compressed data. + +Some examples compared on 16 core CPU, amd64 assembly used: + +``` +* enwik10 +Default... 10000000000 -> 4759950115 [47.60%]; 1.03s, 9263.0MB/s +Better... 10000000000 -> 4084706676 [40.85%]; 2.16s, 4415.4MB/s +Best... 10000000000 -> 3615520079 [36.16%]; 42.259s, 225.7MB/s + +* github-june-2days-2019.json +Default... 6273951764 -> 1041700255 [16.60%]; 431ms, 13882.3MB/s +Better... 6273951764 -> 945841238 [15.08%]; 547ms, 10938.4MB/s +Best... 6273951764 -> 826392576 [13.17%]; 9.455s, 632.8MB/s + +* nyc-taxi-data-10M.csv +Default... 3325605752 -> 1093516949 [32.88%]; 324ms, 9788.7MB/s +Better... 3325605752 -> 885394158 [26.62%]; 491ms, 6459.4MB/s +Best... 3325605752 -> 773681257 [23.26%]; 8.29s, 412.0MB/s + +* 10gb.tar +Default... 10065157632 -> 5915541066 [58.77%]; 1.028s, 9337.4MB/s +Better... 10065157632 -> 5453844650 [54.19%]; 1.597s, 4862.7MB/s +Best... 10065157632 -> 5192495021 [51.59%]; 32.78s, 308.2MB/ + +* consensus.db.10gb +Default... 10737418240 -> 4549762344 [42.37%]; 882ms, 12118.4MB/s +Better... 10737418240 -> 4438535064 [41.34%]; 1.533s, 3500.9MB/s +Best... 10737418240 -> 4210602774 [39.21%]; 42.96s, 254.4MB/s +``` + +Decompression speed should be around the same as using the 'better' compression mode. + +## Dictionaries + +*Note: S2 dictionary compression is currently at an early implementation stage, with no assembly for +neither encoding nor decoding. Performance improvements can be expected in the future.* + +Adding dictionaries allow providing a custom dictionary that will serve as lookup in the beginning of blocks. + +The same dictionary *must* be used for both encoding and decoding. +S2 does not keep track of whether the same dictionary is used, +and using the wrong dictionary will most often not result in an error when decompressing. + +Blocks encoded *without* dictionaries can be decompressed seamlessly *with* a dictionary. +This means it is possible to switch from an encoding without dictionaries to an encoding with dictionaries +and treat the blocks similarly. + +Similar to [zStandard dictionaries](https://github.com/facebook/zstd#the-case-for-small-data-compression), +the same usage scenario applies to S2 dictionaries. + +> Training works if there is some correlation in a family of small data samples. The more data-specific a dictionary is, the more efficient it is (there is no universal dictionary). Hence, deploying one dictionary per type of data will provide the greatest benefits. Dictionary gains are mostly effective in the first few KB. Then, the compression algorithm will gradually use previously decoded content to better compress the rest of the file. + +S2 further limits the dictionary to only be enabled on the first 64KB of a block. +This will remove any negative (speed) impacts of the dictionaries on bigger blocks. + +### Compression + +Using the [github_users_sample_set](https://github.com/facebook/zstd/releases/download/v1.1.3/github_users_sample_set.tar.zst) +and a 64KB dictionary trained with zStandard the following sizes can be achieved. + +| | Default | Better | Best | +|--------------------|------------------|------------------|-----------------------| +| Without Dictionary | 3362023 (44.92%) | 3083163 (41.19%) | 3057944 (40.86%) | +| With Dictionary | 921524 (12.31%) | 873154 (11.67%) | 785503 bytes (10.49%) | + +So for highly repetitive content, this case provides an almost 3x reduction in size. + +For less uniform data we will use the Go source code tree. +Compressing First 64KB of all `.go` files in `go/src`, Go 1.19.5, 8912 files, 51253563 bytes input: + +| | Default | Better | Best | +|--------------------|-------------------|-------------------|-------------------| +| Without Dictionary | 22955767 (44.79%) | 20189613 (39.39% | 19482828 (38.01%) | +| With Dictionary | 19654568 (38.35%) | 16289357 (31.78%) | 15184589 (29.63%) | +| Saving/file | 362 bytes | 428 bytes | 472 bytes | + + +### Creating Dictionaries + +There are no tools to create dictionaries in S2. +However, there are multiple ways to create a useful dictionary: + +#### Using a Sample File + +If your input is very uniform, you can just use a sample file as the dictionary. + +For example in the `github_users_sample_set` above, the average compression only goes up from +10.49% to 11.48% by using the first file as dictionary compared to using a dedicated dictionary. + +```Go + // Read a sample + sample, err := os.ReadFile("sample.json") + + // Create a dictionary. + dict := s2.MakeDict(sample, nil) + + // b := dict.Bytes() will provide a dictionary that can be saved + // and reloaded with s2.NewDict(b). + + // To encode: + encoded := dict.Encode(nil, file) + + // To decode: + decoded, err := dict.Decode(nil, file) +``` + +#### Using Zstandard + +Zstandard dictionaries can easily be converted to S2 dictionaries. + +This can be helpful to generate dictionaries for files that don't have a fixed structure. + + +Example, with training set files placed in `./training-set`: + +`λ zstd -r --train-fastcover training-set/* --maxdict=65536 -o name.dict` + +This will create a dictionary of 64KB, that can be converted to a dictionary like this: + +```Go + // Decode the Zstandard dictionary. + insp, err := zstd.InspectDictionary(zdict) + if err != nil { + panic(err) + } + + // We are only interested in the contents. + // Assume that files start with "// Copyright (c) 2023". + // Search for the longest match for that. + // This may save a few bytes. + dict := s2.MakeDict(insp.Content(), []byte("// Copyright (c) 2023")) + + // b := dict.Bytes() will provide a dictionary that can be saved + // and reloaded with s2.NewDict(b). + + // We can now encode using this dictionary + encodedWithDict := dict.Encode(nil, payload) + + // To decode content: + decoded, err := dict.Decode(nil, encodedWithDict) +``` + +It is recommended to save the dictionary returned by ` b:= dict.Bytes()`, since that will contain only the S2 dictionary. + +This dictionary can later be loaded using `s2.NewDict(b)`. The dictionary then no longer requires `zstd` to be initialized. + +Also note how `s2.MakeDict` allows you to search for a common starting sequence of your files. +This can be omitted, at the expense of a few bytes. + +# Snappy Compatibility + +S2 now offers full compatibility with Snappy. + +This means that the efficient encoders of S2 can be used to generate fully Snappy compatible output. + +There is a [snappy](https://github.com/klauspost/compress/tree/master/snappy) package that can be used by +simply changing imports from `github.com/golang/snappy` to `github.com/klauspost/compress/snappy`. +This uses "better" mode for all operations. +If you would like more control, you can use the s2 package as described below: + +## Blocks + +Snappy compatible blocks can be generated with the S2 encoder. +Compression and speed is typically a bit better `MaxEncodedLen` is also smaller for smaller memory usage. Replace + +| Snappy | S2 replacement | +|---------------------------|-----------------------| +| snappy.Encode(...) | s2.EncodeSnappy(...) | +| snappy.MaxEncodedLen(...) | s2.MaxEncodedLen(...) | + +`s2.EncodeSnappy` can be replaced with `s2.EncodeSnappyBetter` or `s2.EncodeSnappyBest` to get more efficiently compressed snappy compatible output. + +`s2.ConcatBlocks` is compatible with snappy blocks. + +Comparison of [`webdevdata.org-2015-01-07-subset`](https://files.klauspost.com/compress/webdevdata.org-2015-01-07-4GB-subset.7z), +53927 files, total input size: 4,014,735,833 bytes. amd64, single goroutine used: + +| Encoder | Size | MB/s | Reduction | +|-----------------------|------------|------------|------------| +| snappy.Encode | 1128706759 | 725.59 | 71.89% | +| s2.EncodeSnappy | 1093823291 | **899.16** | 72.75% | +| s2.EncodeSnappyBetter | 1001158548 | 578.49 | 75.06% | +| s2.EncodeSnappyBest | 944507998 | 66.00 | **76.47%** | + +## Streams + +For streams, replace `enc = snappy.NewBufferedWriter(w)` with `enc = s2.NewWriter(w, s2.WriterSnappyCompat())`. +All other options are available, but note that block size limit is different for snappy. + +Comparison of different streams, AMD Ryzen 3950x, 16 cores. Size and throughput: + +| File | snappy.NewWriter | S2 Snappy | S2 Snappy, Better | S2 Snappy, Best | +|-----------------------------|--------------------------|---------------------------|--------------------------|-------------------------| +| nyc-taxi-data-10M.csv | 1316042016 - 539.47MB/s | 1307003093 - 10132.73MB/s | 1174534014 - 5002.44MB/s | 1115904679 - 177.97MB/s | +| enwik10 (xml) | 5088294643 - 451.13MB/s | 5175840939 - 9440.69MB/s | 4560784526 - 4487.21MB/s | 4340299103 - 158.92MB/s | +| 10gb.tar (mixed) | 6056946612 - 729.73MB/s | 6208571995 - 9978.05MB/s | 5741646126 - 4919.98MB/s | 5548973895 - 180.44MB/s | +| github-june-2days-2019.json | 1525176492 - 933.00MB/s | 1476519054 - 13150.12MB/s | 1400547532 - 5803.40MB/s | 1321887137 - 204.29MB/s | +| consensus.db.10gb (db) | 5412897703 - 1102.14MB/s | 5354073487 - 13562.91MB/s | 5335069899 - 5294.73MB/s | 5201000954 - 175.72MB/s | + +# Decompression + +All decompression functions map directly to equivalent s2 functions. + +| Snappy | S2 replacement | +|------------------------|--------------------| +| snappy.Decode(...) | s2.Decode(...) | +| snappy.DecodedLen(...) | s2.DecodedLen(...) | +| snappy.NewReader(...) | s2.NewReader(...) | + +Features like [quick forward skipping without decompression](https://pkg.go.dev/github.com/klauspost/compress/s2#Reader.Skip) +are also available for Snappy streams. + +If you know you are only decompressing snappy streams, setting [`ReaderMaxBlockSize(64<<10)`](https://pkg.go.dev/github.com/klauspost/compress/s2#ReaderMaxBlockSize) +on your Reader will reduce memory consumption. + +# Concatenating blocks and streams. + +Concatenating streams will concatenate the output of both without recompressing them. +While this is inefficient in terms of compression it might be usable in certain scenarios. +The 10 byte 'stream identifier' of the second stream can optionally be stripped, but it is not a requirement. + +Blocks can be concatenated using the `ConcatBlocks` function. + +Snappy blocks/streams can safely be concatenated with S2 blocks and streams. +Streams with indexes (see below) will currently not work on concatenated streams. + +# Stream Seek Index + +S2 and Snappy streams can have indexes. These indexes will allow random seeking within the compressed data. + +The index can either be appended to the stream as a skippable block or returned for separate storage. + +When the index is appended to a stream it will be skipped by regular decoders, +so the output remains compatible with other decoders. + +## Creating an Index + +To automatically add an index to a stream, add `WriterAddIndex()` option to your writer. +Then the index will be added to the stream when `Close()` is called. + +``` + // Add Index to stream... + enc := s2.NewWriter(w, s2.WriterAddIndex()) + io.Copy(enc, r) + enc.Close() +``` + +If you want to store the index separately, you can use `CloseIndex()` instead of the regular `Close()`. +This will return the index. Note that `CloseIndex()` should only be called once, and you shouldn't call `Close()`. + +``` + // Get index for separate storage... + enc := s2.NewWriter(w) + io.Copy(enc, r) + index, err := enc.CloseIndex() +``` + +The `index` can then be used needing to read from the stream. +This means the index can be used without needing to seek to the end of the stream +or for manually forwarding streams. See below. + +Finally, an existing S2/Snappy stream can be indexed using the `s2.IndexStream(r io.Reader)` function. + +## Using Indexes + +To use indexes there is a `ReadSeeker(random bool, index []byte) (*ReadSeeker, error)` function available. + +Calling ReadSeeker will return an [io.ReadSeeker](https://pkg.go.dev/io#ReadSeeker) compatible version of the reader. + +If 'random' is specified the returned io.Seeker can be used for random seeking, otherwise only forward seeking is supported. +Enabling random seeking requires the original input to support the [io.Seeker](https://pkg.go.dev/io#Seeker) interface. + +``` + dec := s2.NewReader(r) + rs, err := dec.ReadSeeker(false, nil) + rs.Seek(wantOffset, io.SeekStart) +``` + +Get a seeker to seek forward. Since no index is provided, the index is read from the stream. +This requires that an index was added and that `r` supports the [io.Seeker](https://pkg.go.dev/io#Seeker) interface. + +A custom index can be specified which will be used if supplied. +When using a custom index, it will not be read from the input stream. + +``` + dec := s2.NewReader(r) + rs, err := dec.ReadSeeker(false, index) + rs.Seek(wantOffset, io.SeekStart) +``` + +This will read the index from `index`. Since we specify non-random (forward only) seeking `r` does not have to be an io.Seeker + +``` + dec := s2.NewReader(r) + rs, err := dec.ReadSeeker(true, index) + rs.Seek(wantOffset, io.SeekStart) +``` + +Finally, since we specify that we want to do random seeking `r` must be an io.Seeker. + +The returned [ReadSeeker](https://pkg.go.dev/github.com/klauspost/compress/s2#ReadSeeker) contains a shallow reference to the existing Reader, +meaning changes performed to one is reflected in the other. + +To check if a stream contains an index at the end, the `(*Index).LoadStream(rs io.ReadSeeker) error` can be used. + +## Manually Forwarding Streams + +Indexes can also be read outside the decoder using the [Index](https://pkg.go.dev/github.com/klauspost/compress/s2#Index) type. +This can be used for parsing indexes, either separate or in streams. + +In some cases it may not be possible to serve a seekable stream. +This can for instance be an HTTP stream, where the Range request +is sent at the start of the stream. + +With a little bit of extra code it is still possible to use indexes +to forward to specific offset with a single forward skip. + +It is possible to load the index manually like this: +``` + var index s2.Index + _, err = index.Load(idxBytes) +``` + +This can be used to figure out how much to offset the compressed stream: + +``` + compressedOffset, uncompressedOffset, err := index.Find(wantOffset) +``` + +The `compressedOffset` is the number of bytes that should be skipped +from the beginning of the compressed file. + +The `uncompressedOffset` will then be offset of the uncompressed bytes returned +when decoding from that position. This will always be <= wantOffset. + +When creating a decoder it must be specified that it should *not* expect a stream identifier +at the beginning of the stream. Assuming the io.Reader `r` has been forwarded to `compressedOffset` +we create the decoder like this: + +``` + dec := s2.NewReader(r, s2.ReaderIgnoreStreamIdentifier()) +``` + +We are not completely done. We still need to forward the stream the uncompressed bytes we didn't want. +This is done using the regular "Skip" function: + +``` + err = dec.Skip(wantOffset - uncompressedOffset) +``` + +This will ensure that we are at exactly the offset we want, and reading from `dec` will start at the requested offset. + +# Compact storage + +For compact storage [RemoveIndexHeaders](https://pkg.go.dev/github.com/klauspost/compress/s2#RemoveIndexHeaders) can be used to remove any redundant info from +a serialized index. If you remove the header it must be restored before [Loading](https://pkg.go.dev/github.com/klauspost/compress/s2#Index.Load). + +This is expected to save 20 bytes. These can be restored using [RestoreIndexHeaders](https://pkg.go.dev/github.com/klauspost/compress/s2#RestoreIndexHeaders). This removes a layer of security, but is the most compact representation. Returns nil if headers contains errors. + +## Index Format: + +Each block is structured as a snappy skippable block, with the chunk ID 0x99. + +The block can be read from the front, but contains information so it can be read from the back as well. + +Numbers are stored as fixed size little endian values or [zigzag encoded](https://developers.google.com/protocol-buffers/docs/encoding#signed_integers) [base 128 varints](https://developers.google.com/protocol-buffers/docs/encoding), +with un-encoded value length of 64 bits, unless other limits are specified. + +| Content | Format | +|--------------------------------------|-------------------------------------------------------------------------------------------------------------------------------| +| ID, `[1]byte` | Always 0x99. | +| Data Length, `[3]byte` | 3 byte little-endian length of the chunk in bytes, following this. | +| Header `[6]byte` | Header, must be `[115, 50, 105, 100, 120, 0]` or in text: "s2idx\x00". | +| UncompressedSize, Varint | Total Uncompressed size. | +| CompressedSize, Varint | Total Compressed size if known. Should be -1 if unknown. | +| EstBlockSize, Varint | Block Size, used for guessing uncompressed offsets. Must be >= 0. | +| Entries, Varint | Number of Entries in index, must be < 65536 and >=0. | +| HasUncompressedOffsets `byte` | 0 if no uncompressed offsets are present, 1 if present. Other values are invalid. | +| UncompressedOffsets, [Entries]VarInt | Uncompressed offsets. See below how to decode. | +| CompressedOffsets, [Entries]VarInt | Compressed offsets. See below how to decode. | +| Block Size, `[4]byte` | Little Endian total encoded size (including header and trailer). Can be used for searching backwards to start of block. | +| Trailer `[6]byte` | Trailer, must be `[0, 120, 100, 105, 50, 115]` or in text: "\x00xdi2s". Can be used for identifying block from end of stream. | + +For regular streams the uncompressed offsets are fully predictable, +so `HasUncompressedOffsets` allows to specify that compressed blocks all have +exactly `EstBlockSize` bytes of uncompressed content. + +Entries *must* be in order, starting with the lowest offset, +and there *must* be no uncompressed offset duplicates. +Entries *may* point to the start of a skippable block, +but it is then not allowed to also have an entry for the next block since +that would give an uncompressed offset duplicate. + +There is no requirement for all blocks to be represented in the index. +In fact there is a maximum of 65536 block entries in an index. + +The writer can use any method to reduce the number of entries. +An implicit block start at 0,0 can be assumed. + +### Decoding entries: + +``` +// Read Uncompressed entries. +// Each assumes EstBlockSize delta from previous. +for each entry { + uOff = 0 + if HasUncompressedOffsets == 1 { + uOff = ReadVarInt // Read value from stream + } + + // Except for the first entry, use previous values. + if entryNum == 0 { + entry[entryNum].UncompressedOffset = uOff + continue + } + + // Uncompressed uses previous offset and adds EstBlockSize + entry[entryNum].UncompressedOffset = entry[entryNum-1].UncompressedOffset + EstBlockSize + uOff +} + + +// Guess that the first block will be 50% of uncompressed size. +// Integer truncating division must be used. +CompressGuess := EstBlockSize / 2 + +// Read Compressed entries. +// Each assumes CompressGuess delta from previous. +// CompressGuess is adjusted for each value. +for each entry { + cOff = ReadVarInt // Read value from stream + + // Except for the first entry, use previous values. + if entryNum == 0 { + entry[entryNum].CompressedOffset = cOff + continue + } + + // Compressed uses previous and our estimate. + entry[entryNum].CompressedOffset = entry[entryNum-1].CompressedOffset + CompressGuess + cOff + + // Adjust compressed offset for next loop, integer truncating division must be used. + CompressGuess += cOff/2 +} +``` + +To decode from any given uncompressed offset `(wantOffset)`: + +* Iterate entries until `entry[n].UncompressedOffset > wantOffset`. +* Start decoding from `entry[n-1].CompressedOffset`. +* Discard `entry[n-1].UncompressedOffset - wantOffset` bytes from the decoded stream. + +See [using indexes](https://github.com/klauspost/compress/tree/master/s2#using-indexes) for functions that perform the operations with a simpler interface. + + +# Format Extensions + +* Frame [Stream identifier](https://github.com/google/snappy/blob/master/framing_format.txt#L68) changed from `sNaPpY` to `S2sTwO`. +* [Framed compressed blocks](https://github.com/google/snappy/blob/master/format_description.txt) can be up to 4MB (up from 64KB). +* Compressed blocks can have an offset of `0`, which indicates to repeat the last seen offset. + +Repeat offsets must be encoded as a [2.2.1. Copy with 1-byte offset (01)](https://github.com/google/snappy/blob/master/format_description.txt#L89), where the offset is 0. + +The length is specified by reading the 3-bit length specified in the tag and decode using this table: + +| Length | Actual Length | +|--------|----------------------| +| 0 | 4 | +| 1 | 5 | +| 2 | 6 | +| 3 | 7 | +| 4 | 8 | +| 5 | 8 + read 1 byte | +| 6 | 260 + read 2 bytes | +| 7 | 65540 + read 3 bytes | + +This allows any repeat offset + length to be represented by 2 to 5 bytes. +It also allows to emit matches longer than 64 bytes with one copy + one repeat instead of several 64 byte copies. + +Lengths are stored as little endian values. + +The first copy of a block cannot be a repeat offset and the offset is reset on every block in streams. + +Default streaming block size is 1MB. + +# Dictionary Encoding + +Adding dictionaries allow providing a custom dictionary that will serve as lookup in the beginning of blocks. + +A dictionary provides an initial repeat value that can be used to point to a common header. + +Other than that the dictionary contains values that can be used as back-references. + +Often used data should be placed at the *end* of the dictionary since offsets < 2048 bytes will be smaller. + +## Format + +Dictionary *content* must at least 16 bytes and less or equal to 64KiB (65536 bytes). + +Encoding: `[repeat value (uvarint)][dictionary content...]` + +Before the dictionary content, an unsigned base-128 (uvarint) encoded value specifying the initial repeat offset. +This value is an offset into the dictionary content and not a back-reference offset, +so setting this to 0 will make the repeat value point to the first value of the dictionary. + +The value must be less than the dictionary length-8 + +## Encoding + +From the decoder point of view the dictionary content is seen as preceding the encoded content. + +`[dictionary content][decoded output]` + +Backreferences to the dictionary are encoded as ordinary backreferences that have an offset before the start of the decoded block. + +Matches copying from the dictionary are **not** allowed to cross from the dictionary into the decoded data. +However, if a copy ends at the end of the dictionary the next repeat will point to the start of the decoded buffer, which is allowed. + +The first match can be a repeat value, which will use the repeat offset stored in the dictionary. + +When 64KB (65536 bytes) has been en/decoded it is no longer allowed to reference the dictionary, +neither by a copy nor repeat operations. +If the boundary is crossed while copying from the dictionary, the operation should complete, +but the next instruction is not allowed to reference the dictionary. + +Valid blocks encoded *without* a dictionary can be decoded with any dictionary. +There are no checks whether the supplied dictionary is the correct for a block. +Because of this there is no overhead by using a dictionary. + +## Example + +This is the dictionary content. Elements are separated by `[]`. + +Dictionary: `[0x0a][Yesterday 25 bananas were added to Benjamins brown bag]`. + +Initial repeat offset is set at 10, which is the letter `2`. + +Encoded `[LIT "10"][REPEAT len=10][LIT "hich"][MATCH off=50 len=6][MATCH off=31 len=6][MATCH off=61 len=10]` + +Decoded: `[10][ bananas w][hich][ were ][brown ][were added]` + +Output: `10 bananas which were brown were added` + + +## Streams + +For streams each block can use the dictionary. + +The dictionary cannot not currently be provided on the stream. + + +# LICENSE + +This code is based on the [Snappy-Go](https://github.com/golang/snappy) implementation. + +Use of this source code is governed by a BSD-style license that can be found in the LICENSE file. diff --git a/vendor/github.com/klauspost/compress/s2/decode.go b/vendor/github.com/klauspost/compress/s2/decode.go new file mode 100644 index 000000000..264ffd0a9 --- /dev/null +++ b/vendor/github.com/klauspost/compress/s2/decode.go @@ -0,0 +1,443 @@ +// Copyright 2011 The Snappy-Go Authors. All rights reserved. +// Copyright (c) 2019 Klaus Post. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package s2 + +import ( + "encoding/binary" + "errors" + "fmt" + "strconv" + + "github.com/klauspost/compress/internal/race" +) + +var ( + // ErrCorrupt reports that the input is invalid. + ErrCorrupt = errors.New("s2: corrupt input") + // ErrCRC reports that the input failed CRC validation (streams only) + ErrCRC = errors.New("s2: corrupt input, crc mismatch") + // ErrTooLarge reports that the uncompressed length is too large. + ErrTooLarge = errors.New("s2: decoded block is too large") + // ErrUnsupported reports that the input isn't supported. + ErrUnsupported = errors.New("s2: unsupported input") +) + +// DecodedLen returns the length of the decoded block. +func DecodedLen(src []byte) (int, error) { + v, _, err := decodedLen(src) + return v, err +} + +// decodedLen returns the length of the decoded block and the number of bytes +// that the length header occupied. +func decodedLen(src []byte) (blockLen, headerLen int, err error) { + v, n := binary.Uvarint(src) + if n <= 0 || v > 0xffffffff { + return 0, 0, ErrCorrupt + } + + const wordSize = 32 << (^uint(0) >> 32 & 1) + if wordSize == 32 && v > 0x7fffffff { + return 0, 0, ErrTooLarge + } + return int(v), n, nil +} + +const ( + decodeErrCodeCorrupt = 1 +) + +// Decode returns the decoded form of src. The returned slice may be a sub- +// slice of dst if dst was large enough to hold the entire decoded block. +// Otherwise, a newly allocated slice will be returned. +// +// The dst and src must not overlap. It is valid to pass a nil dst. +func Decode(dst, src []byte) ([]byte, error) { + dLen, s, err := decodedLen(src) + if err != nil { + return nil, err + } + if dLen <= cap(dst) { + dst = dst[:dLen] + } else { + dst = make([]byte, dLen) + } + + race.WriteSlice(dst) + race.ReadSlice(src[s:]) + + if s2Decode(dst, src[s:]) != 0 { + return nil, ErrCorrupt + } + return dst, nil +} + +// s2DecodeDict writes the decoding of src to dst. It assumes that the varint-encoded +// length of the decompressed bytes has already been read, and that len(dst) +// equals that length. +// +// It returns 0 on success or a decodeErrCodeXxx error code on failure. +func s2DecodeDict(dst, src []byte, dict *Dict) int { + if dict == nil { + return s2Decode(dst, src) + } + const debug = false + const debugErrs = debug + + if debug { + fmt.Println("Starting decode, dst len:", len(dst)) + } + var d, s, length int + offset := len(dict.dict) - dict.repeat + + // As long as we can read at least 5 bytes... + for s < len(src)-5 { + // Removing bounds checks is SLOWER, when if doing + // in := src[s:s+5] + // Checked on Go 1.18 + switch src[s] & 0x03 { + case tagLiteral: + x := uint32(src[s] >> 2) + switch { + case x < 60: + s++ + case x == 60: + s += 2 + x = uint32(src[s-1]) + case x == 61: + in := src[s : s+3] + x = uint32(in[1]) | uint32(in[2])<<8 + s += 3 + case x == 62: + in := src[s : s+4] + // Load as 32 bit and shift down. + x = uint32(in[0]) | uint32(in[1])<<8 | uint32(in[2])<<16 | uint32(in[3])<<24 + x >>= 8 + s += 4 + case x == 63: + in := src[s : s+5] + x = uint32(in[1]) | uint32(in[2])<<8 | uint32(in[3])<<16 | uint32(in[4])<<24 + s += 5 + } + length = int(x) + 1 + if debug { + fmt.Println("literals, length:", length, "d-after:", d+length) + } + if length > len(dst)-d || length > len(src)-s || (strconv.IntSize == 32 && length <= 0) { + if debugErrs { + fmt.Println("corrupt literal: length:", length, "d-left:", len(dst)-d, "src-left:", len(src)-s) + } + return decodeErrCodeCorrupt + } + + copy(dst[d:], src[s:s+length]) + d += length + s += length + continue + + case tagCopy1: + s += 2 + toffset := int(uint32(src[s-2])&0xe0<<3 | uint32(src[s-1])) + length = int(src[s-2]) >> 2 & 0x7 + if toffset == 0 { + if debug { + fmt.Print("(repeat) ") + } + // keep last offset + switch length { + case 5: + length = int(src[s]) + 4 + s += 1 + case 6: + in := src[s : s+2] + length = int(uint32(in[0])|(uint32(in[1])<<8)) + (1 << 8) + s += 2 + case 7: + in := src[s : s+3] + length = int((uint32(in[2])<<16)|(uint32(in[1])<<8)|uint32(in[0])) + (1 << 16) + s += 3 + default: // 0-> 4 + } + } else { + offset = toffset + } + length += 4 + case tagCopy2: + in := src[s : s+3] + offset = int(uint32(in[1]) | uint32(in[2])<<8) + length = 1 + int(in[0])>>2 + s += 3 + + case tagCopy4: + in := src[s : s+5] + offset = int(uint32(in[1]) | uint32(in[2])<<8 | uint32(in[3])<<16 | uint32(in[4])<<24) + length = 1 + int(in[0])>>2 + s += 5 + } + + if offset <= 0 || length > len(dst)-d { + if debugErrs { + fmt.Println("match error; offset:", offset, "length:", length, "dst-left:", len(dst)-d) + } + return decodeErrCodeCorrupt + } + + // copy from dict + if d < offset { + if d > MaxDictSrcOffset { + if debugErrs { + fmt.Println("dict after", MaxDictSrcOffset, "d:", d, "offset:", offset, "length:", length) + } + return decodeErrCodeCorrupt + } + startOff := len(dict.dict) - offset + d + if startOff < 0 || startOff+length > len(dict.dict) { + if debugErrs { + fmt.Printf("offset (%d) + length (%d) bigger than dict (%d)\n", offset, length, len(dict.dict)) + } + return decodeErrCodeCorrupt + } + if debug { + fmt.Println("dict copy, length:", length, "offset:", offset, "d-after:", d+length, "dict start offset:", startOff) + } + copy(dst[d:d+length], dict.dict[startOff:]) + d += length + continue + } + + if debug { + fmt.Println("copy, length:", length, "offset:", offset, "d-after:", d+length) + } + + // Copy from an earlier sub-slice of dst to a later sub-slice. + // If no overlap, use the built-in copy: + if offset > length { + copy(dst[d:d+length], dst[d-offset:]) + d += length + continue + } + + // Unlike the built-in copy function, this byte-by-byte copy always runs + // forwards, even if the slices overlap. Conceptually, this is: + // + // d += forwardCopy(dst[d:d+length], dst[d-offset:]) + // + // We align the slices into a and b and show the compiler they are the same size. + // This allows the loop to run without bounds checks. + a := dst[d : d+length] + b := dst[d-offset:] + b = b[:len(a)] + for i := range a { + a[i] = b[i] + } + d += length + } + + // Remaining with extra checks... + for s < len(src) { + switch src[s] & 0x03 { + case tagLiteral: + x := uint32(src[s] >> 2) + switch { + case x < 60: + s++ + case x == 60: + s += 2 + if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line. + if debugErrs { + fmt.Println("src went oob") + } + return decodeErrCodeCorrupt + } + x = uint32(src[s-1]) + case x == 61: + s += 3 + if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line. + if debugErrs { + fmt.Println("src went oob") + } + return decodeErrCodeCorrupt + } + x = uint32(src[s-2]) | uint32(src[s-1])<<8 + case x == 62: + s += 4 + if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line. + if debugErrs { + fmt.Println("src went oob") + } + return decodeErrCodeCorrupt + } + x = uint32(src[s-3]) | uint32(src[s-2])<<8 | uint32(src[s-1])<<16 + case x == 63: + s += 5 + if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line. + if debugErrs { + fmt.Println("src went oob") + } + return decodeErrCodeCorrupt + } + x = uint32(src[s-4]) | uint32(src[s-3])<<8 | uint32(src[s-2])<<16 | uint32(src[s-1])<<24 + } + length = int(x) + 1 + if length > len(dst)-d || length > len(src)-s || (strconv.IntSize == 32 && length <= 0) { + if debugErrs { + fmt.Println("corrupt literal: length:", length, "d-left:", len(dst)-d, "src-left:", len(src)-s) + } + return decodeErrCodeCorrupt + } + if debug { + fmt.Println("literals, length:", length, "d-after:", d+length) + } + + copy(dst[d:], src[s:s+length]) + d += length + s += length + continue + + case tagCopy1: + s += 2 + if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line. + if debugErrs { + fmt.Println("src went oob") + } + return decodeErrCodeCorrupt + } + length = int(src[s-2]) >> 2 & 0x7 + toffset := int(uint32(src[s-2])&0xe0<<3 | uint32(src[s-1])) + if toffset == 0 { + if debug { + fmt.Print("(repeat) ") + } + // keep last offset + switch length { + case 5: + s += 1 + if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line. + if debugErrs { + fmt.Println("src went oob") + } + return decodeErrCodeCorrupt + } + length = int(uint32(src[s-1])) + 4 + case 6: + s += 2 + if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line. + if debugErrs { + fmt.Println("src went oob") + } + return decodeErrCodeCorrupt + } + length = int(uint32(src[s-2])|(uint32(src[s-1])<<8)) + (1 << 8) + case 7: + s += 3 + if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line. + if debugErrs { + fmt.Println("src went oob") + } + return decodeErrCodeCorrupt + } + length = int(uint32(src[s-3])|(uint32(src[s-2])<<8)|(uint32(src[s-1])<<16)) + (1 << 16) + default: // 0-> 4 + } + } else { + offset = toffset + } + length += 4 + case tagCopy2: + s += 3 + if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line. + if debugErrs { + fmt.Println("src went oob") + } + return decodeErrCodeCorrupt + } + length = 1 + int(src[s-3])>>2 + offset = int(uint32(src[s-2]) | uint32(src[s-1])<<8) + + case tagCopy4: + s += 5 + if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line. + if debugErrs { + fmt.Println("src went oob") + } + return decodeErrCodeCorrupt + } + length = 1 + int(src[s-5])>>2 + offset = int(uint32(src[s-4]) | uint32(src[s-3])<<8 | uint32(src[s-2])<<16 | uint32(src[s-1])<<24) + } + + if offset <= 0 || length > len(dst)-d { + if debugErrs { + fmt.Println("match error; offset:", offset, "length:", length, "dst-left:", len(dst)-d) + } + return decodeErrCodeCorrupt + } + + // copy from dict + if d < offset { + if d > MaxDictSrcOffset { + if debugErrs { + fmt.Println("dict after", MaxDictSrcOffset, "d:", d, "offset:", offset, "length:", length) + } + return decodeErrCodeCorrupt + } + rOff := len(dict.dict) - (offset - d) + if debug { + fmt.Println("starting dict entry from dict offset", len(dict.dict)-rOff) + } + if rOff+length > len(dict.dict) { + if debugErrs { + fmt.Println("err: END offset", rOff+length, "bigger than dict", len(dict.dict), "dict offset:", rOff, "length:", length) + } + return decodeErrCodeCorrupt + } + if rOff < 0 { + if debugErrs { + fmt.Println("err: START offset", rOff, "less than 0", len(dict.dict), "dict offset:", rOff, "length:", length) + } + return decodeErrCodeCorrupt + } + copy(dst[d:d+length], dict.dict[rOff:]) + d += length + continue + } + + if debug { + fmt.Println("copy, length:", length, "offset:", offset, "d-after:", d+length) + } + + // Copy from an earlier sub-slice of dst to a later sub-slice. + // If no overlap, use the built-in copy: + if offset > length { + copy(dst[d:d+length], dst[d-offset:]) + d += length + continue + } + + // Unlike the built-in copy function, this byte-by-byte copy always runs + // forwards, even if the slices overlap. Conceptually, this is: + // + // d += forwardCopy(dst[d:d+length], dst[d-offset:]) + // + // We align the slices into a and b and show the compiler they are the same size. + // This allows the loop to run without bounds checks. + a := dst[d : d+length] + b := dst[d-offset:] + b = b[:len(a)] + for i := range a { + a[i] = b[i] + } + d += length + } + + if d != len(dst) { + if debugErrs { + fmt.Println("wanted length", len(dst), "got", d) + } + return decodeErrCodeCorrupt + } + return 0 +} diff --git a/vendor/github.com/golang/snappy/decode_amd64.s b/vendor/github.com/klauspost/compress/s2/decode_amd64.s similarity index 60% rename from vendor/github.com/golang/snappy/decode_amd64.s rename to vendor/github.com/klauspost/compress/s2/decode_amd64.s index e6179f65e..9b105e03c 100644 --- a/vendor/github.com/golang/snappy/decode_amd64.s +++ b/vendor/github.com/klauspost/compress/s2/decode_amd64.s @@ -1,4 +1,5 @@ // Copyright 2016 The Go Authors. All rights reserved. +// Copyright (c) 2019 Klaus Post. All rights reserved. // Use of this source code is governed by a BSD-style // license that can be found in the LICENSE file. @@ -8,6 +9,21 @@ #include "textflag.h" +#define R_TMP0 AX +#define R_TMP1 BX +#define R_LEN CX +#define R_OFF DX +#define R_SRC SI +#define R_DST DI +#define R_DBASE R8 +#define R_DLEN R9 +#define R_DEND R10 +#define R_SBASE R11 +#define R_SLEN R12 +#define R_SEND R13 +#define R_TMP2 R14 +#define R_TMP3 R15 + // The asm code generally follows the pure Go code in decode_other.go, except // where marked with a "!!!". @@ -15,51 +31,52 @@ // // All local variables fit into registers. The non-zero stack size is only to // spill registers and push args when issuing a CALL. The register allocation: -// - AX scratch -// - BX scratch -// - CX length or x -// - DX offset -// - SI &src[s] -// - DI &dst[d] -// + R8 dst_base -// + R9 dst_len -// + R10 dst_base + dst_len -// + R11 src_base -// + R12 src_len -// + R13 src_base + src_len -// - R14 used by doCopy -// - R15 used by doCopy +// - R_TMP0 scratch +// - R_TMP1 scratch +// - R_LEN length or x (shared) +// - R_OFF offset +// - R_SRC &src[s] +// - R_DST &dst[d] +// + R_DBASE dst_base +// + R_DLEN dst_len +// + R_DEND dst_base + dst_len +// + R_SBASE src_base +// + R_SLEN src_len +// + R_SEND src_base + src_len +// - R_TMP2 used by doCopy +// - R_TMP3 used by doCopy // -// The registers R8-R13 (marked with a "+") are set at the start of the +// The registers R_DBASE-R_SEND (marked with a "+") are set at the start of the // function, and after a CALL returns, and are not otherwise modified. // -// The d variable is implicitly DI - R8, and len(dst)-d is R10 - DI. -// The s variable is implicitly SI - R11, and len(src)-s is R13 - SI. -TEXT ·decode(SB), NOSPLIT, $48-56 - // Initialize SI, DI and R8-R13. - MOVQ dst_base+0(FP), R8 - MOVQ dst_len+8(FP), R9 - MOVQ R8, DI - MOVQ R8, R10 - ADDQ R9, R10 - MOVQ src_base+24(FP), R11 - MOVQ src_len+32(FP), R12 - MOVQ R11, SI - MOVQ R11, R13 - ADDQ R12, R13 +// The d variable is implicitly R_DST - R_DBASE, and len(dst)-d is R_DEND - R_DST. +// The s variable is implicitly R_SRC - R_SBASE, and len(src)-s is R_SEND - R_SRC. +TEXT ·s2Decode(SB), NOSPLIT, $48-56 + // Initialize R_SRC, R_DST and R_DBASE-R_SEND. + MOVQ dst_base+0(FP), R_DBASE + MOVQ dst_len+8(FP), R_DLEN + MOVQ R_DBASE, R_DST + MOVQ R_DBASE, R_DEND + ADDQ R_DLEN, R_DEND + MOVQ src_base+24(FP), R_SBASE + MOVQ src_len+32(FP), R_SLEN + MOVQ R_SBASE, R_SRC + MOVQ R_SBASE, R_SEND + ADDQ R_SLEN, R_SEND + XORQ R_OFF, R_OFF loop: // for s < len(src) - CMPQ SI, R13 + CMPQ R_SRC, R_SEND JEQ end - // CX = uint32(src[s]) + // R_LEN = uint32(src[s]) // // switch src[s] & 0x03 - MOVBLZX (SI), CX - MOVL CX, BX - ANDL $3, BX - CMPL BX, $1 + MOVBLZX (R_SRC), R_LEN + MOVL R_LEN, R_TMP1 + ANDL $3, R_TMP1 + CMPL R_TMP1, $1 JAE tagCopy // ---------------------------------------- @@ -68,35 +85,35 @@ loop: // case tagLiteral: // x := uint32(src[s] >> 2) // switch - SHRL $2, CX - CMPL CX, $60 + SHRL $2, R_LEN + CMPL R_LEN, $60 JAE tagLit60Plus // case x < 60: // s++ - INCQ SI + INCQ R_SRC doLit: // This is the end of the inner "switch", when we have a literal tag. // - // We assume that CX == x and x fits in a uint32, where x is the variable + // We assume that R_LEN == x and x fits in a uint32, where x is the variable // used in the pure Go decode_other.go code. // length = int(x) + 1 // // Unlike the pure Go code, we don't need to check if length <= 0 because - // CX can hold 64 bits, so the increment cannot overflow. - INCQ CX + // R_LEN can hold 64 bits, so the increment cannot overflow. + INCQ R_LEN // Prepare to check if copying length bytes will run past the end of dst or // src. // - // AX = len(dst) - d - // BX = len(src) - s - MOVQ R10, AX - SUBQ DI, AX - MOVQ R13, BX - SUBQ SI, BX + // R_TMP0 = len(dst) - d + // R_TMP1 = len(src) - s + MOVQ R_DEND, R_TMP0 + SUBQ R_DST, R_TMP0 + MOVQ R_SEND, R_TMP1 + SUBQ R_SRC, R_TMP1 // !!! Try a faster technique for short (16 or fewer bytes) copies. // @@ -109,11 +126,11 @@ doLit: // is contiguous in memory and so it needs to leave enough source bytes to // read the next tag without refilling buffers, but Go's Decode assumes // contiguousness (the src argument is a []byte). - CMPQ CX, $16 + CMPQ R_LEN, $16 JGT callMemmove - CMPQ AX, $16 + CMPQ R_TMP0, $16 JLT callMemmove - CMPQ BX, $16 + CMPQ R_TMP1, $16 JLT callMemmove // !!! Implement the copy from src to dst as a 16-byte load and store. @@ -127,53 +144,55 @@ doLit: // Note that on amd64, it is legal and cheap to issue unaligned 8-byte or // 16-byte loads and stores. This technique probably wouldn't be as // effective on architectures that are fussier about alignment. - MOVOU 0(SI), X0 - MOVOU X0, 0(DI) + MOVOU 0(R_SRC), X0 + MOVOU X0, 0(R_DST) // d += length // s += length - ADDQ CX, DI - ADDQ CX, SI + ADDQ R_LEN, R_DST + ADDQ R_LEN, R_SRC JMP loop callMemmove: // if length > len(dst)-d || length > len(src)-s { etc } - CMPQ CX, AX + CMPQ R_LEN, R_TMP0 JGT errCorrupt - CMPQ CX, BX + CMPQ R_LEN, R_TMP1 JGT errCorrupt // copy(dst[d:], src[s:s+length]) // // This means calling runtime·memmove(&dst[d], &src[s], length), so we push - // DI, SI and CX as arguments. Coincidentally, we also need to spill those + // R_DST, R_SRC and R_LEN as arguments. Coincidentally, we also need to spill those // three registers to the stack, to save local variables across the CALL. - MOVQ DI, 0(SP) - MOVQ SI, 8(SP) - MOVQ CX, 16(SP) - MOVQ DI, 24(SP) - MOVQ SI, 32(SP) - MOVQ CX, 40(SP) + MOVQ R_DST, 0(SP) + MOVQ R_SRC, 8(SP) + MOVQ R_LEN, 16(SP) + MOVQ R_DST, 24(SP) + MOVQ R_SRC, 32(SP) + MOVQ R_LEN, 40(SP) + MOVQ R_OFF, 48(SP) CALL runtime·memmove(SB) // Restore local variables: unspill registers from the stack and - // re-calculate R8-R13. - MOVQ 24(SP), DI - MOVQ 32(SP), SI - MOVQ 40(SP), CX - MOVQ dst_base+0(FP), R8 - MOVQ dst_len+8(FP), R9 - MOVQ R8, R10 - ADDQ R9, R10 - MOVQ src_base+24(FP), R11 - MOVQ src_len+32(FP), R12 - MOVQ R11, R13 - ADDQ R12, R13 + // re-calculate R_DBASE-R_SEND. + MOVQ 24(SP), R_DST + MOVQ 32(SP), R_SRC + MOVQ 40(SP), R_LEN + MOVQ 48(SP), R_OFF + MOVQ dst_base+0(FP), R_DBASE + MOVQ dst_len+8(FP), R_DLEN + MOVQ R_DBASE, R_DEND + ADDQ R_DLEN, R_DEND + MOVQ src_base+24(FP), R_SBASE + MOVQ src_len+32(FP), R_SLEN + MOVQ R_SBASE, R_SEND + ADDQ R_SLEN, R_SEND // d += length // s += length - ADDQ CX, DI - ADDQ CX, SI + ADDQ R_LEN, R_DST + ADDQ R_LEN, R_SRC JMP loop tagLit60Plus: @@ -182,44 +201,42 @@ tagLit60Plus: // s += x - 58; if uint(s) > uint(len(src)) { etc } // // checks. In the asm version, we code it once instead of once per switch case. - ADDQ CX, SI - SUBQ $58, SI - MOVQ SI, BX - SUBQ R11, BX - CMPQ BX, R12 + ADDQ R_LEN, R_SRC + SUBQ $58, R_SRC + CMPQ R_SRC, R_SEND JA errCorrupt // case x == 60: - CMPL CX, $61 + CMPL R_LEN, $61 JEQ tagLit61 JA tagLit62Plus // x = uint32(src[s-1]) - MOVBLZX -1(SI), CX + MOVBLZX -1(R_SRC), R_LEN JMP doLit tagLit61: // case x == 61: // x = uint32(src[s-2]) | uint32(src[s-1])<<8 - MOVWLZX -2(SI), CX + MOVWLZX -2(R_SRC), R_LEN JMP doLit tagLit62Plus: - CMPL CX, $62 + CMPL R_LEN, $62 JA tagLit63 // case x == 62: // x = uint32(src[s-3]) | uint32(src[s-2])<<8 | uint32(src[s-1])<<16 - MOVWLZX -3(SI), CX - MOVBLZX -1(SI), BX - SHLL $16, BX - ORL BX, CX - JMP doLit + // We read one byte, safe to read one back, since we are just reading tag. + // x = binary.LittleEndian.Uint32(src[s-1:]) >> 8 + MOVL -4(R_SRC), R_LEN + SHRL $8, R_LEN + JMP doLit tagLit63: // case x == 63: // x = uint32(src[s-4]) | uint32(src[s-3])<<8 | uint32(src[s-2])<<16 | uint32(src[s-1])<<24 - MOVL -4(SI), CX + MOVL -4(R_SRC), R_LEN JMP doLit // The code above handles literal tags. @@ -229,103 +246,161 @@ tagLit63: tagCopy4: // case tagCopy4: // s += 5 - ADDQ $5, SI + ADDQ $5, R_SRC // if uint(s) > uint(len(src)) { etc } - MOVQ SI, BX - SUBQ R11, BX - CMPQ BX, R12 + CMPQ R_SRC, R_SEND JA errCorrupt // length = 1 + int(src[s-5])>>2 - SHRQ $2, CX - INCQ CX + SHRQ $2, R_LEN + INCQ R_LEN // offset = int(uint32(src[s-4]) | uint32(src[s-3])<<8 | uint32(src[s-2])<<16 | uint32(src[s-1])<<24) - MOVLQZX -4(SI), DX + MOVLQZX -4(R_SRC), R_OFF JMP doCopy tagCopy2: // case tagCopy2: // s += 3 - ADDQ $3, SI + ADDQ $3, R_SRC // if uint(s) > uint(len(src)) { etc } - MOVQ SI, BX - SUBQ R11, BX - CMPQ BX, R12 + CMPQ R_SRC, R_SEND JA errCorrupt // length = 1 + int(src[s-3])>>2 - SHRQ $2, CX - INCQ CX + SHRQ $2, R_LEN + INCQ R_LEN // offset = int(uint32(src[s-2]) | uint32(src[s-1])<<8) - MOVWQZX -2(SI), DX + MOVWQZX -2(R_SRC), R_OFF JMP doCopy tagCopy: // We have a copy tag. We assume that: - // - BX == src[s] & 0x03 - // - CX == src[s] - CMPQ BX, $2 + // - R_TMP1 == src[s] & 0x03 + // - R_LEN == src[s] + CMPQ R_TMP1, $2 JEQ tagCopy2 JA tagCopy4 // case tagCopy1: // s += 2 - ADDQ $2, SI + ADDQ $2, R_SRC // if uint(s) > uint(len(src)) { etc } - MOVQ SI, BX - SUBQ R11, BX - CMPQ BX, R12 + CMPQ R_SRC, R_SEND JA errCorrupt // offset = int(uint32(src[s-2])&0xe0<<3 | uint32(src[s-1])) - MOVQ CX, DX - ANDQ $0xe0, DX - SHLQ $3, DX - MOVBQZX -1(SI), BX - ORQ BX, DX - // length = 4 + int(src[s-2])>>2&0x7 - SHRQ $2, CX - ANDQ $7, CX - ADDQ $4, CX + MOVBQZX -1(R_SRC), R_TMP1 + MOVQ R_LEN, R_TMP0 + SHRQ $2, R_LEN + ANDQ $0xe0, R_TMP0 + ANDQ $7, R_LEN + SHLQ $3, R_TMP0 + ADDQ $4, R_LEN + ORQ R_TMP1, R_TMP0 + + // check if repeat code, ZF set by ORQ. + JZ repeatCode + + // This is a regular copy, transfer our temporary value to R_OFF (length) + MOVQ R_TMP0, R_OFF + JMP doCopy + +// This is a repeat code. +repeatCode: + // If length < 9, reuse last offset, with the length already calculated. + CMPQ R_LEN, $9 + JL doCopyRepeat + + // Read additional bytes for length. + JE repeatLen1 + + // Rare, so the extra branch shouldn't hurt too much. + CMPQ R_LEN, $10 + JE repeatLen2 + JMP repeatLen3 + +// Read repeat lengths. +repeatLen1: + // s ++ + ADDQ $1, R_SRC + + // if uint(s) > uint(len(src)) { etc } + CMPQ R_SRC, R_SEND + JA errCorrupt + + // length = src[s-1] + 8 + MOVBQZX -1(R_SRC), R_LEN + ADDL $8, R_LEN + JMP doCopyRepeat + +repeatLen2: + // s +=2 + ADDQ $2, R_SRC + + // if uint(s) > uint(len(src)) { etc } + CMPQ R_SRC, R_SEND + JA errCorrupt + + // length = uint32(src[s-2]) | (uint32(src[s-1])<<8) + (1 << 8) + MOVWQZX -2(R_SRC), R_LEN + ADDL $260, R_LEN + JMP doCopyRepeat + +repeatLen3: + // s +=3 + ADDQ $3, R_SRC + + // if uint(s) > uint(len(src)) { etc } + CMPQ R_SRC, R_SEND + JA errCorrupt + + // length = uint32(src[s-3]) | (uint32(src[s-2])<<8) | (uint32(src[s-1])<<16) + (1 << 16) + // Read one byte further back (just part of the tag, shifted out) + MOVL -4(R_SRC), R_LEN + SHRL $8, R_LEN + ADDL $65540, R_LEN + JMP doCopyRepeat doCopy: // This is the end of the outer "switch", when we have a copy tag. // // We assume that: - // - CX == length && CX > 0 - // - DX == offset - - // if offset <= 0 { etc } - CMPQ DX, $0 - JLE errCorrupt + // - R_LEN == length && R_LEN > 0 + // - R_OFF == offset // if d < offset { etc } - MOVQ DI, BX - SUBQ R8, BX - CMPQ BX, DX + MOVQ R_DST, R_TMP1 + SUBQ R_DBASE, R_TMP1 + CMPQ R_TMP1, R_OFF JLT errCorrupt + // Repeat values can skip the test above, since any offset > 0 will be in dst. +doCopyRepeat: + // if offset <= 0 { etc } + CMPQ R_OFF, $0 + JLE errCorrupt + // if length > len(dst)-d { etc } - MOVQ R10, BX - SUBQ DI, BX - CMPQ CX, BX + MOVQ R_DEND, R_TMP1 + SUBQ R_DST, R_TMP1 + CMPQ R_LEN, R_TMP1 JGT errCorrupt // forwardCopy(dst[d:d+length], dst[d-offset:]); d += length // // Set: - // - R14 = len(dst)-d - // - R15 = &dst[d-offset] - MOVQ R10, R14 - SUBQ DI, R14 - MOVQ DI, R15 - SUBQ DX, R15 + // - R_TMP2 = len(dst)-d + // - R_TMP3 = &dst[d-offset] + MOVQ R_DEND, R_TMP2 + SUBQ R_DST, R_TMP2 + MOVQ R_DST, R_TMP3 + SUBQ R_OFF, R_TMP3 // !!! Try a faster technique for short (16 or fewer bytes) forward copies. // @@ -340,17 +415,17 @@ doCopy: // } // copy 16 bytes // d += length - CMPQ CX, $16 + CMPQ R_LEN, $16 JGT slowForwardCopy - CMPQ DX, $8 + CMPQ R_OFF, $8 JLT slowForwardCopy - CMPQ R14, $16 + CMPQ R_TMP2, $16 JLT slowForwardCopy - MOVQ 0(R15), AX - MOVQ AX, 0(DI) - MOVQ 8(R15), BX - MOVQ BX, 8(DI) - ADDQ CX, DI + MOVQ 0(R_TMP3), R_TMP0 + MOVQ R_TMP0, 0(R_DST) + MOVQ 8(R_TMP3), R_TMP1 + MOVQ R_TMP1, 8(R_DST) + ADDQ R_LEN, R_DST JMP loop slowForwardCopy: @@ -402,10 +477,13 @@ slowForwardCopy: // if length > len(dst)-d-10 { // goto verySlowForwardCopy // } - SUBQ $10, R14 - CMPQ CX, R14 + SUBQ $10, R_TMP2 + CMPQ R_LEN, R_TMP2 JGT verySlowForwardCopy + // We want to keep the offset, so we use R_TMP2 from here. + MOVQ R_OFF, R_TMP2 + makeOffsetAtLeast8: // !!! As above, expand the pattern so that offset >= 8 and we can use // 8-byte load/stores. @@ -416,37 +494,37 @@ makeOffsetAtLeast8: // d += offset // offset += offset // // The two previous lines together means that d-offset, and therefore - // // R15, is unchanged. + // // R_TMP3, is unchanged. // } - CMPQ DX, $8 + CMPQ R_TMP2, $8 JGE fixUpSlowForwardCopy - MOVQ (R15), BX - MOVQ BX, (DI) - SUBQ DX, CX - ADDQ DX, DI - ADDQ DX, DX + MOVQ (R_TMP3), R_TMP1 + MOVQ R_TMP1, (R_DST) + SUBQ R_TMP2, R_LEN + ADDQ R_TMP2, R_DST + ADDQ R_TMP2, R_TMP2 JMP makeOffsetAtLeast8 fixUpSlowForwardCopy: - // !!! Add length (which might be negative now) to d (implied by DI being + // !!! Add length (which might be negative now) to d (implied by R_DST being // &dst[d]) so that d ends up at the right place when we jump back to the - // top of the loop. Before we do that, though, we save DI to AX so that, if + // top of the loop. Before we do that, though, we save R_DST to R_TMP0 so that, if // length is positive, copying the remaining length bytes will write to the // right place. - MOVQ DI, AX - ADDQ CX, DI + MOVQ R_DST, R_TMP0 + ADDQ R_LEN, R_DST finishSlowForwardCopy: // !!! Repeat 8-byte load/stores until length <= 0. Ending with a negative // length means that we overrun, but as above, that will be fixed up by // subsequent iterations of the outermost loop. - CMPQ CX, $0 + CMPQ R_LEN, $0 JLE loop - MOVQ (R15), BX - MOVQ BX, (AX) - ADDQ $8, R15 - ADDQ $8, AX - SUBQ $8, CX + MOVQ (R_TMP3), R_TMP1 + MOVQ R_TMP1, (R_TMP0) + ADDQ $8, R_TMP3 + ADDQ $8, R_TMP0 + SUBQ $8, R_LEN JMP finishSlowForwardCopy verySlowForwardCopy: @@ -462,11 +540,11 @@ verySlowForwardCopy: // break // } // } - MOVB (R15), BX - MOVB BX, (DI) - INCQ R15 - INCQ DI - DECQ CX + MOVB (R_TMP3), R_TMP1 + MOVB R_TMP1, (R_DST) + INCQ R_TMP3 + INCQ R_DST + DECQ R_LEN JNZ verySlowForwardCopy JMP loop @@ -477,7 +555,7 @@ end: // This is the end of the "for s < len(src)". // // if d != len(dst) { etc } - CMPQ DI, R10 + CMPQ R_DST, R_DEND JNE errCorrupt // return 0 diff --git a/vendor/github.com/golang/snappy/decode_arm64.s b/vendor/github.com/klauspost/compress/s2/decode_arm64.s similarity index 59% rename from vendor/github.com/golang/snappy/decode_arm64.s rename to vendor/github.com/klauspost/compress/s2/decode_arm64.s index 7a3ead17e..78e463f34 100644 --- a/vendor/github.com/golang/snappy/decode_arm64.s +++ b/vendor/github.com/klauspost/compress/s2/decode_arm64.s @@ -8,6 +8,31 @@ #include "textflag.h" +#define R_TMP0 R2 +#define R_TMP1 R3 +#define R_LEN R4 +#define R_OFF R5 +#define R_SRC R6 +#define R_DST R7 +#define R_DBASE R8 +#define R_DLEN R9 +#define R_DEND R10 +#define R_SBASE R11 +#define R_SLEN R12 +#define R_SEND R13 +#define R_TMP2 R14 +#define R_TMP3 R15 + +// TEST_SRC will check if R_SRC is <= SRC_END +#define TEST_SRC() \ + CMP R_SEND, R_SRC \ + BGT errCorrupt + +// MOVD R_SRC, R_TMP1 +// SUB R_SBASE, R_TMP1, R_TMP1 +// CMP R_SLEN, R_TMP1 +// BGT errCorrupt + // The asm code generally follows the pure Go code in decode_other.go, except // where marked with a "!!!". @@ -15,52 +40,53 @@ // // All local variables fit into registers. The non-zero stack size is only to // spill registers and push args when issuing a CALL. The register allocation: -// - R2 scratch -// - R3 scratch -// - R4 length or x -// - R5 offset -// - R6 &src[s] -// - R7 &dst[d] -// + R8 dst_base -// + R9 dst_len -// + R10 dst_base + dst_len -// + R11 src_base -// + R12 src_len -// + R13 src_base + src_len -// - R14 used by doCopy -// - R15 used by doCopy +// - R_TMP0 scratch +// - R_TMP1 scratch +// - R_LEN length or x +// - R_OFF offset +// - R_SRC &src[s] +// - R_DST &dst[d] +// + R_DBASE dst_base +// + R_DLEN dst_len +// + R_DEND dst_base + dst_len +// + R_SBASE src_base +// + R_SLEN src_len +// + R_SEND src_base + src_len +// - R_TMP2 used by doCopy +// - R_TMP3 used by doCopy // -// The registers R8-R13 (marked with a "+") are set at the start of the +// The registers R_DBASE-R_SEND (marked with a "+") are set at the start of the // function, and after a CALL returns, and are not otherwise modified. // -// The d variable is implicitly R7 - R8, and len(dst)-d is R10 - R7. -// The s variable is implicitly R6 - R11, and len(src)-s is R13 - R6. -TEXT ·decode(SB), NOSPLIT, $56-56 - // Initialize R6, R7 and R8-R13. - MOVD dst_base+0(FP), R8 - MOVD dst_len+8(FP), R9 - MOVD R8, R7 - MOVD R8, R10 - ADD R9, R10, R10 - MOVD src_base+24(FP), R11 - MOVD src_len+32(FP), R12 - MOVD R11, R6 - MOVD R11, R13 - ADD R12, R13, R13 +// The d variable is implicitly R_DST - R_DBASE, and len(dst)-d is R_DEND - R_DST. +// The s variable is implicitly R_SRC - R_SBASE, and len(src)-s is R_SEND - R_SRC. +TEXT ·s2Decode(SB), NOSPLIT, $56-56 + // Initialize R_SRC, R_DST and R_DBASE-R_SEND. + MOVD dst_base+0(FP), R_DBASE + MOVD dst_len+8(FP), R_DLEN + MOVD R_DBASE, R_DST + MOVD R_DBASE, R_DEND + ADD R_DLEN, R_DEND, R_DEND + MOVD src_base+24(FP), R_SBASE + MOVD src_len+32(FP), R_SLEN + MOVD R_SBASE, R_SRC + MOVD R_SBASE, R_SEND + ADD R_SLEN, R_SEND, R_SEND + MOVD $0, R_OFF loop: // for s < len(src) - CMP R13, R6 + CMP R_SEND, R_SRC BEQ end - // R4 = uint32(src[s]) + // R_LEN = uint32(src[s]) // // switch src[s] & 0x03 - MOVBU (R6), R4 - MOVW R4, R3 - ANDW $3, R3 + MOVBU (R_SRC), R_LEN + MOVW R_LEN, R_TMP1 + ANDW $3, R_TMP1 MOVW $1, R1 - CMPW R1, R3 + CMPW R1, R_TMP1 BGE tagCopy // ---------------------------------------- @@ -70,35 +96,35 @@ loop: // x := uint32(src[s] >> 2) // switch MOVW $60, R1 - LSRW $2, R4, R4 - CMPW R4, R1 + LSRW $2, R_LEN, R_LEN + CMPW R_LEN, R1 BLS tagLit60Plus // case x < 60: // s++ - ADD $1, R6, R6 + ADD $1, R_SRC, R_SRC doLit: // This is the end of the inner "switch", when we have a literal tag. // - // We assume that R4 == x and x fits in a uint32, where x is the variable + // We assume that R_LEN == x and x fits in a uint32, where x is the variable // used in the pure Go decode_other.go code. // length = int(x) + 1 // // Unlike the pure Go code, we don't need to check if length <= 0 because - // R4 can hold 64 bits, so the increment cannot overflow. - ADD $1, R4, R4 + // R_LEN can hold 64 bits, so the increment cannot overflow. + ADD $1, R_LEN, R_LEN // Prepare to check if copying length bytes will run past the end of dst or // src. // - // R2 = len(dst) - d - // R3 = len(src) - s - MOVD R10, R2 - SUB R7, R2, R2 - MOVD R13, R3 - SUB R6, R3, R3 + // R_TMP0 = len(dst) - d + // R_TMP1 = len(src) - s + MOVD R_DEND, R_TMP0 + SUB R_DST, R_TMP0, R_TMP0 + MOVD R_SEND, R_TMP1 + SUB R_SRC, R_TMP1, R_TMP1 // !!! Try a faster technique for short (16 or fewer bytes) copies. // @@ -111,11 +137,11 @@ doLit: // is contiguous in memory and so it needs to leave enough source bytes to // read the next tag without refilling buffers, but Go's Decode assumes // contiguousness (the src argument is a []byte). - CMP $16, R4 + CMP $16, R_LEN BGT callMemmove - CMP $16, R2 + CMP $16, R_TMP0 BLT callMemmove - CMP $16, R3 + CMP $16, R_TMP1 BLT callMemmove // !!! Implement the copy from src to dst as a 16-byte load and store. @@ -129,53 +155,55 @@ doLit: // Note that on arm64, it is legal and cheap to issue unaligned 8-byte or // 16-byte loads and stores. This technique probably wouldn't be as // effective on architectures that are fussier about alignment. - LDP 0(R6), (R14, R15) - STP (R14, R15), 0(R7) + LDP 0(R_SRC), (R_TMP2, R_TMP3) + STP (R_TMP2, R_TMP3), 0(R_DST) // d += length // s += length - ADD R4, R7, R7 - ADD R4, R6, R6 + ADD R_LEN, R_DST, R_DST + ADD R_LEN, R_SRC, R_SRC B loop callMemmove: // if length > len(dst)-d || length > len(src)-s { etc } - CMP R2, R4 + CMP R_TMP0, R_LEN BGT errCorrupt - CMP R3, R4 + CMP R_TMP1, R_LEN BGT errCorrupt // copy(dst[d:], src[s:s+length]) // // This means calling runtime·memmove(&dst[d], &src[s], length), so we push - // R7, R6 and R4 as arguments. Coincidentally, we also need to spill those + // R_DST, R_SRC and R_LEN as arguments. Coincidentally, we also need to spill those // three registers to the stack, to save local variables across the CALL. - MOVD R7, 8(RSP) - MOVD R6, 16(RSP) - MOVD R4, 24(RSP) - MOVD R7, 32(RSP) - MOVD R6, 40(RSP) - MOVD R4, 48(RSP) + MOVD R_DST, 8(RSP) + MOVD R_SRC, 16(RSP) + MOVD R_LEN, 24(RSP) + MOVD R_DST, 32(RSP) + MOVD R_SRC, 40(RSP) + MOVD R_LEN, 48(RSP) + MOVD R_OFF, 56(RSP) CALL runtime·memmove(SB) // Restore local variables: unspill registers from the stack and - // re-calculate R8-R13. - MOVD 32(RSP), R7 - MOVD 40(RSP), R6 - MOVD 48(RSP), R4 - MOVD dst_base+0(FP), R8 - MOVD dst_len+8(FP), R9 - MOVD R8, R10 - ADD R9, R10, R10 - MOVD src_base+24(FP), R11 - MOVD src_len+32(FP), R12 - MOVD R11, R13 - ADD R12, R13, R13 + // re-calculate R_DBASE-R_SEND. + MOVD 32(RSP), R_DST + MOVD 40(RSP), R_SRC + MOVD 48(RSP), R_LEN + MOVD 56(RSP), R_OFF + MOVD dst_base+0(FP), R_DBASE + MOVD dst_len+8(FP), R_DLEN + MOVD R_DBASE, R_DEND + ADD R_DLEN, R_DEND, R_DEND + MOVD src_base+24(FP), R_SBASE + MOVD src_len+32(FP), R_SLEN + MOVD R_SBASE, R_SEND + ADD R_SLEN, R_SEND, R_SEND // d += length // s += length - ADD R4, R7, R7 - ADD R4, R6, R6 + ADD R_LEN, R_DST, R_DST + ADD R_LEN, R_SRC, R_SRC B loop tagLit60Plus: @@ -184,44 +212,41 @@ tagLit60Plus: // s += x - 58; if uint(s) > uint(len(src)) { etc } // // checks. In the asm version, we code it once instead of once per switch case. - ADD R4, R6, R6 - SUB $58, R6, R6 - MOVD R6, R3 - SUB R11, R3, R3 - CMP R12, R3 - BGT errCorrupt + ADD R_LEN, R_SRC, R_SRC + SUB $58, R_SRC, R_SRC + TEST_SRC() // case x == 60: MOVW $61, R1 - CMPW R1, R4 + CMPW R1, R_LEN BEQ tagLit61 BGT tagLit62Plus // x = uint32(src[s-1]) - MOVBU -1(R6), R4 + MOVBU -1(R_SRC), R_LEN B doLit tagLit61: // case x == 61: // x = uint32(src[s-2]) | uint32(src[s-1])<<8 - MOVHU -2(R6), R4 + MOVHU -2(R_SRC), R_LEN B doLit tagLit62Plus: - CMPW $62, R4 + CMPW $62, R_LEN BHI tagLit63 // case x == 62: // x = uint32(src[s-3]) | uint32(src[s-2])<<8 | uint32(src[s-1])<<16 - MOVHU -3(R6), R4 - MOVBU -1(R6), R3 - ORR R3<<16, R4 + MOVHU -3(R_SRC), R_LEN + MOVBU -1(R_SRC), R_TMP1 + ORR R_TMP1<<16, R_LEN B doLit tagLit63: // case x == 63: // x = uint32(src[s-4]) | uint32(src[s-3])<<8 | uint32(src[s-2])<<16 | uint32(src[s-1])<<24 - MOVWU -4(R6), R4 + MOVWU -4(R_SRC), R_LEN B doLit // The code above handles literal tags. @@ -231,103 +256,155 @@ tagLit63: tagCopy4: // case tagCopy4: // s += 5 - ADD $5, R6, R6 + ADD $5, R_SRC, R_SRC // if uint(s) > uint(len(src)) { etc } - MOVD R6, R3 - SUB R11, R3, R3 - CMP R12, R3 + MOVD R_SRC, R_TMP1 + SUB R_SBASE, R_TMP1, R_TMP1 + CMP R_SLEN, R_TMP1 BGT errCorrupt // length = 1 + int(src[s-5])>>2 MOVD $1, R1 - ADD R4>>2, R1, R4 + ADD R_LEN>>2, R1, R_LEN // offset = int(uint32(src[s-4]) | uint32(src[s-3])<<8 | uint32(src[s-2])<<16 | uint32(src[s-1])<<24) - MOVWU -4(R6), R5 + MOVWU -4(R_SRC), R_OFF B doCopy tagCopy2: // case tagCopy2: // s += 3 - ADD $3, R6, R6 + ADD $3, R_SRC, R_SRC // if uint(s) > uint(len(src)) { etc } - MOVD R6, R3 - SUB R11, R3, R3 - CMP R12, R3 - BGT errCorrupt + TEST_SRC() // length = 1 + int(src[s-3])>>2 MOVD $1, R1 - ADD R4>>2, R1, R4 + ADD R_LEN>>2, R1, R_LEN // offset = int(uint32(src[s-2]) | uint32(src[s-1])<<8) - MOVHU -2(R6), R5 + MOVHU -2(R_SRC), R_OFF B doCopy tagCopy: // We have a copy tag. We assume that: - // - R3 == src[s] & 0x03 - // - R4 == src[s] - CMP $2, R3 + // - R_TMP1 == src[s] & 0x03 + // - R_LEN == src[s] + CMP $2, R_TMP1 BEQ tagCopy2 BGT tagCopy4 // case tagCopy1: // s += 2 - ADD $2, R6, R6 + ADD $2, R_SRC, R_SRC // if uint(s) > uint(len(src)) { etc } - MOVD R6, R3 - SUB R11, R3, R3 - CMP R12, R3 - BGT errCorrupt + TEST_SRC() // offset = int(uint32(src[s-2])&0xe0<<3 | uint32(src[s-1])) - MOVD R4, R5 - AND $0xe0, R5 - MOVBU -1(R6), R3 - ORR R5<<3, R3, R5 + // Calculate offset in R_TMP0 in case it is a repeat. + MOVD R_LEN, R_TMP0 + AND $0xe0, R_TMP0 + MOVBU -1(R_SRC), R_TMP1 + ORR R_TMP0<<3, R_TMP1, R_TMP0 // length = 4 + int(src[s-2])>>2&0x7 MOVD $7, R1 - AND R4>>2, R1, R4 - ADD $4, R4, R4 + AND R_LEN>>2, R1, R_LEN + ADD $4, R_LEN, R_LEN + + // check if repeat code with offset 0. + CMP $0, R_TMP0 + BEQ repeatCode + + // This is a regular copy, transfer our temporary value to R_OFF (offset) + MOVD R_TMP0, R_OFF + B doCopy + + // This is a repeat code. +repeatCode: + // If length < 9, reuse last offset, with the length already calculated. + CMP $9, R_LEN + BLT doCopyRepeat + BEQ repeatLen1 + CMP $10, R_LEN + BEQ repeatLen2 + +repeatLen3: + // s +=3 + ADD $3, R_SRC, R_SRC + + // if uint(s) > uint(len(src)) { etc } + TEST_SRC() + + // length = uint32(src[s-3]) | (uint32(src[s-2])<<8) | (uint32(src[s-1])<<16) + 65540 + MOVBU -1(R_SRC), R_TMP0 + MOVHU -3(R_SRC), R_LEN + ORR R_TMP0<<16, R_LEN, R_LEN + ADD $65540, R_LEN, R_LEN + B doCopyRepeat + +repeatLen2: + // s +=2 + ADD $2, R_SRC, R_SRC + + // if uint(s) > uint(len(src)) { etc } + TEST_SRC() + + // length = uint32(src[s-2]) | (uint32(src[s-1])<<8) + 260 + MOVHU -2(R_SRC), R_LEN + ADD $260, R_LEN, R_LEN + B doCopyRepeat + +repeatLen1: + // s +=1 + ADD $1, R_SRC, R_SRC + + // if uint(s) > uint(len(src)) { etc } + TEST_SRC() + + // length = src[s-1] + 8 + MOVBU -1(R_SRC), R_LEN + ADD $8, R_LEN, R_LEN + B doCopyRepeat doCopy: // This is the end of the outer "switch", when we have a copy tag. // // We assume that: - // - R4 == length && R4 > 0 - // - R5 == offset - - // if offset <= 0 { etc } - MOVD $0, R1 - CMP R1, R5 - BLE errCorrupt + // - R_LEN == length && R_LEN > 0 + // - R_OFF == offset // if d < offset { etc } - MOVD R7, R3 - SUB R8, R3, R3 - CMP R5, R3 + MOVD R_DST, R_TMP1 + SUB R_DBASE, R_TMP1, R_TMP1 + CMP R_OFF, R_TMP1 BLT errCorrupt + // Repeat values can skip the test above, since any offset > 0 will be in dst. +doCopyRepeat: + + // if offset <= 0 { etc } + CMP $0, R_OFF + BLE errCorrupt + // if length > len(dst)-d { etc } - MOVD R10, R3 - SUB R7, R3, R3 - CMP R3, R4 + MOVD R_DEND, R_TMP1 + SUB R_DST, R_TMP1, R_TMP1 + CMP R_TMP1, R_LEN BGT errCorrupt // forwardCopy(dst[d:d+length], dst[d-offset:]); d += length // // Set: - // - R14 = len(dst)-d - // - R15 = &dst[d-offset] - MOVD R10, R14 - SUB R7, R14, R14 - MOVD R7, R15 - SUB R5, R15, R15 + // - R_TMP2 = len(dst)-d + // - R_TMP3 = &dst[d-offset] + MOVD R_DEND, R_TMP2 + SUB R_DST, R_TMP2, R_TMP2 + MOVD R_DST, R_TMP3 + SUB R_OFF, R_TMP3, R_TMP3 // !!! Try a faster technique for short (16 or fewer bytes) forward copies. // @@ -342,17 +419,17 @@ doCopy: // } // copy 16 bytes // d += length - CMP $16, R4 + CMP $16, R_LEN BGT slowForwardCopy - CMP $8, R5 + CMP $8, R_OFF BLT slowForwardCopy - CMP $16, R14 + CMP $16, R_TMP2 BLT slowForwardCopy - MOVD 0(R15), R2 - MOVD R2, 0(R7) - MOVD 8(R15), R3 - MOVD R3, 8(R7) - ADD R4, R7, R7 + MOVD 0(R_TMP3), R_TMP0 + MOVD R_TMP0, 0(R_DST) + MOVD 8(R_TMP3), R_TMP1 + MOVD R_TMP1, 8(R_DST) + ADD R_LEN, R_DST, R_DST B loop slowForwardCopy: @@ -404,10 +481,13 @@ slowForwardCopy: // if length > len(dst)-d-10 { // goto verySlowForwardCopy // } - SUB $10, R14, R14 - CMP R14, R4 + SUB $10, R_TMP2, R_TMP2 + CMP R_TMP2, R_LEN BGT verySlowForwardCopy + // We want to keep the offset, so we use R_TMP2 from here. + MOVD R_OFF, R_TMP2 + makeOffsetAtLeast8: // !!! As above, expand the pattern so that offset >= 8 and we can use // 8-byte load/stores. @@ -418,38 +498,38 @@ makeOffsetAtLeast8: // d += offset // offset += offset // // The two previous lines together means that d-offset, and therefore - // // R15, is unchanged. + // // R_TMP3, is unchanged. // } - CMP $8, R5 + CMP $8, R_TMP2 BGE fixUpSlowForwardCopy - MOVD (R15), R3 - MOVD R3, (R7) - SUB R5, R4, R4 - ADD R5, R7, R7 - ADD R5, R5, R5 + MOVD (R_TMP3), R_TMP1 + MOVD R_TMP1, (R_DST) + SUB R_TMP2, R_LEN, R_LEN + ADD R_TMP2, R_DST, R_DST + ADD R_TMP2, R_TMP2, R_TMP2 B makeOffsetAtLeast8 fixUpSlowForwardCopy: - // !!! Add length (which might be negative now) to d (implied by R7 being + // !!! Add length (which might be negative now) to d (implied by R_DST being // &dst[d]) so that d ends up at the right place when we jump back to the - // top of the loop. Before we do that, though, we save R7 to R2 so that, if + // top of the loop. Before we do that, though, we save R_DST to R_TMP0 so that, if // length is positive, copying the remaining length bytes will write to the // right place. - MOVD R7, R2 - ADD R4, R7, R7 + MOVD R_DST, R_TMP0 + ADD R_LEN, R_DST, R_DST finishSlowForwardCopy: // !!! Repeat 8-byte load/stores until length <= 0. Ending with a negative // length means that we overrun, but as above, that will be fixed up by // subsequent iterations of the outermost loop. MOVD $0, R1 - CMP R1, R4 + CMP R1, R_LEN BLE loop - MOVD (R15), R3 - MOVD R3, (R2) - ADD $8, R15, R15 - ADD $8, R2, R2 - SUB $8, R4, R4 + MOVD (R_TMP3), R_TMP1 + MOVD R_TMP1, (R_TMP0) + ADD $8, R_TMP3, R_TMP3 + ADD $8, R_TMP0, R_TMP0 + SUB $8, R_LEN, R_LEN B finishSlowForwardCopy verySlowForwardCopy: @@ -465,12 +545,12 @@ verySlowForwardCopy: // break // } // } - MOVB (R15), R3 - MOVB R3, (R7) - ADD $1, R15, R15 - ADD $1, R7, R7 - SUB $1, R4, R4 - CBNZ R4, verySlowForwardCopy + MOVB (R_TMP3), R_TMP1 + MOVB R_TMP1, (R_DST) + ADD $1, R_TMP3, R_TMP3 + ADD $1, R_DST, R_DST + SUB $1, R_LEN, R_LEN + CBNZ R_LEN, verySlowForwardCopy B loop // The code above handles copy tags. @@ -480,7 +560,7 @@ end: // This is the end of the "for s < len(src)". // // if d != len(dst) { etc } - CMP R10, R7 + CMP R_DEND, R_DST BNE errCorrupt // return 0 @@ -489,6 +569,6 @@ end: errCorrupt: // return decodeErrCodeCorrupt - MOVD $1, R2 - MOVD R2, ret+48(FP) + MOVD $1, R_TMP0 + MOVD R_TMP0, ret+48(FP) RET diff --git a/vendor/github.com/golang/snappy/decode_asm.go b/vendor/github.com/klauspost/compress/s2/decode_asm.go similarity index 66% rename from vendor/github.com/golang/snappy/decode_asm.go rename to vendor/github.com/klauspost/compress/s2/decode_asm.go index 7082b3491..cb3576edd 100644 --- a/vendor/github.com/golang/snappy/decode_asm.go +++ b/vendor/github.com/klauspost/compress/s2/decode_asm.go @@ -1,15 +1,17 @@ // Copyright 2016 The Snappy-Go Authors. All rights reserved. +// Copyright (c) 2019 Klaus Post. All rights reserved. // Use of this source code is governed by a BSD-style // license that can be found in the LICENSE file. +//go:build (amd64 || arm64) && !appengine && gc && !noasm +// +build amd64 arm64 // +build !appengine // +build gc // +build !noasm -// +build amd64 arm64 -package snappy +package s2 // decode has the same semantics as in decode_other.go. // //go:noescape -func decode(dst, src []byte) int +func s2Decode(dst, src []byte) int diff --git a/vendor/github.com/klauspost/compress/s2/decode_other.go b/vendor/github.com/klauspost/compress/s2/decode_other.go new file mode 100644 index 000000000..2cb55c2c7 --- /dev/null +++ b/vendor/github.com/klauspost/compress/s2/decode_other.go @@ -0,0 +1,292 @@ +// Copyright 2016 The Snappy-Go Authors. All rights reserved. +// Copyright (c) 2019 Klaus Post. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build (!amd64 && !arm64) || appengine || !gc || noasm +// +build !amd64,!arm64 appengine !gc noasm + +package s2 + +import ( + "fmt" + "strconv" +) + +// decode writes the decoding of src to dst. It assumes that the varint-encoded +// length of the decompressed bytes has already been read, and that len(dst) +// equals that length. +// +// It returns 0 on success or a decodeErrCodeXxx error code on failure. +func s2Decode(dst, src []byte) int { + const debug = false + if debug { + fmt.Println("Starting decode, dst len:", len(dst)) + } + var d, s, length int + offset := 0 + + // As long as we can read at least 5 bytes... + for s < len(src)-5 { + // Removing bounds checks is SLOWER, when if doing + // in := src[s:s+5] + // Checked on Go 1.18 + switch src[s] & 0x03 { + case tagLiteral: + x := uint32(src[s] >> 2) + switch { + case x < 60: + s++ + case x == 60: + s += 2 + x = uint32(src[s-1]) + case x == 61: + in := src[s : s+3] + x = uint32(in[1]) | uint32(in[2])<<8 + s += 3 + case x == 62: + in := src[s : s+4] + // Load as 32 bit and shift down. + x = uint32(in[0]) | uint32(in[1])<<8 | uint32(in[2])<<16 | uint32(in[3])<<24 + x >>= 8 + s += 4 + case x == 63: + in := src[s : s+5] + x = uint32(in[1]) | uint32(in[2])<<8 | uint32(in[3])<<16 | uint32(in[4])<<24 + s += 5 + } + length = int(x) + 1 + if length > len(dst)-d || length > len(src)-s || (strconv.IntSize == 32 && length <= 0) { + if debug { + fmt.Println("corrupt: lit size", length) + } + return decodeErrCodeCorrupt + } + if debug { + fmt.Println("literals, length:", length, "d-after:", d+length) + } + + copy(dst[d:], src[s:s+length]) + d += length + s += length + continue + + case tagCopy1: + s += 2 + toffset := int(uint32(src[s-2])&0xe0<<3 | uint32(src[s-1])) + length = int(src[s-2]) >> 2 & 0x7 + if toffset == 0 { + if debug { + fmt.Print("(repeat) ") + } + // keep last offset + switch length { + case 5: + length = int(src[s]) + 4 + s += 1 + case 6: + in := src[s : s+2] + length = int(uint32(in[0])|(uint32(in[1])<<8)) + (1 << 8) + s += 2 + case 7: + in := src[s : s+3] + length = int((uint32(in[2])<<16)|(uint32(in[1])<<8)|uint32(in[0])) + (1 << 16) + s += 3 + default: // 0-> 4 + } + } else { + offset = toffset + } + length += 4 + case tagCopy2: + in := src[s : s+3] + offset = int(uint32(in[1]) | uint32(in[2])<<8) + length = 1 + int(in[0])>>2 + s += 3 + + case tagCopy4: + in := src[s : s+5] + offset = int(uint32(in[1]) | uint32(in[2])<<8 | uint32(in[3])<<16 | uint32(in[4])<<24) + length = 1 + int(in[0])>>2 + s += 5 + } + + if offset <= 0 || d < offset || length > len(dst)-d { + if debug { + fmt.Println("corrupt: match, length", length, "offset:", offset, "dst avail:", len(dst)-d, "dst pos:", d) + } + + return decodeErrCodeCorrupt + } + + if debug { + fmt.Println("copy, length:", length, "offset:", offset, "d-after:", d+length) + } + + // Copy from an earlier sub-slice of dst to a later sub-slice. + // If no overlap, use the built-in copy: + if offset > length { + copy(dst[d:d+length], dst[d-offset:]) + d += length + continue + } + + // Unlike the built-in copy function, this byte-by-byte copy always runs + // forwards, even if the slices overlap. Conceptually, this is: + // + // d += forwardCopy(dst[d:d+length], dst[d-offset:]) + // + // We align the slices into a and b and show the compiler they are the same size. + // This allows the loop to run without bounds checks. + a := dst[d : d+length] + b := dst[d-offset:] + b = b[:len(a)] + for i := range a { + a[i] = b[i] + } + d += length + } + + // Remaining with extra checks... + for s < len(src) { + switch src[s] & 0x03 { + case tagLiteral: + x := uint32(src[s] >> 2) + switch { + case x < 60: + s++ + case x == 60: + s += 2 + if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line. + return decodeErrCodeCorrupt + } + x = uint32(src[s-1]) + case x == 61: + s += 3 + if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line. + return decodeErrCodeCorrupt + } + x = uint32(src[s-2]) | uint32(src[s-1])<<8 + case x == 62: + s += 4 + if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line. + return decodeErrCodeCorrupt + } + x = uint32(src[s-3]) | uint32(src[s-2])<<8 | uint32(src[s-1])<<16 + case x == 63: + s += 5 + if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line. + return decodeErrCodeCorrupt + } + x = uint32(src[s-4]) | uint32(src[s-3])<<8 | uint32(src[s-2])<<16 | uint32(src[s-1])<<24 + } + length = int(x) + 1 + if length > len(dst)-d || length > len(src)-s || (strconv.IntSize == 32 && length <= 0) { + if debug { + fmt.Println("corrupt: lit size", length) + } + return decodeErrCodeCorrupt + } + if debug { + fmt.Println("literals, length:", length, "d-after:", d+length) + } + + copy(dst[d:], src[s:s+length]) + d += length + s += length + continue + + case tagCopy1: + s += 2 + if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line. + return decodeErrCodeCorrupt + } + length = int(src[s-2]) >> 2 & 0x7 + toffset := int(uint32(src[s-2])&0xe0<<3 | uint32(src[s-1])) + if toffset == 0 { + if debug { + fmt.Print("(repeat) ") + } + // keep last offset + switch length { + case 5: + s += 1 + if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line. + return decodeErrCodeCorrupt + } + length = int(uint32(src[s-1])) + 4 + case 6: + s += 2 + if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line. + return decodeErrCodeCorrupt + } + length = int(uint32(src[s-2])|(uint32(src[s-1])<<8)) + (1 << 8) + case 7: + s += 3 + if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line. + return decodeErrCodeCorrupt + } + length = int(uint32(src[s-3])|(uint32(src[s-2])<<8)|(uint32(src[s-1])<<16)) + (1 << 16) + default: // 0-> 4 + } + } else { + offset = toffset + } + length += 4 + case tagCopy2: + s += 3 + if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line. + return decodeErrCodeCorrupt + } + length = 1 + int(src[s-3])>>2 + offset = int(uint32(src[s-2]) | uint32(src[s-1])<<8) + + case tagCopy4: + s += 5 + if uint(s) > uint(len(src)) { // The uint conversions catch overflow from the previous line. + return decodeErrCodeCorrupt + } + length = 1 + int(src[s-5])>>2 + offset = int(uint32(src[s-4]) | uint32(src[s-3])<<8 | uint32(src[s-2])<<16 | uint32(src[s-1])<<24) + } + + if offset <= 0 || d < offset || length > len(dst)-d { + if debug { + fmt.Println("corrupt: match, length", length, "offset:", offset, "dst avail:", len(dst)-d, "dst pos:", d) + } + return decodeErrCodeCorrupt + } + + if debug { + fmt.Println("copy, length:", length, "offset:", offset, "d-after:", d+length) + } + + // Copy from an earlier sub-slice of dst to a later sub-slice. + // If no overlap, use the built-in copy: + if offset > length { + copy(dst[d:d+length], dst[d-offset:]) + d += length + continue + } + + // Unlike the built-in copy function, this byte-by-byte copy always runs + // forwards, even if the slices overlap. Conceptually, this is: + // + // d += forwardCopy(dst[d:d+length], dst[d-offset:]) + // + // We align the slices into a and b and show the compiler they are the same size. + // This allows the loop to run without bounds checks. + a := dst[d : d+length] + b := dst[d-offset:] + b = b[:len(a)] + for i := range a { + a[i] = b[i] + } + d += length + } + + if d != len(dst) { + return decodeErrCodeCorrupt + } + return 0 +} diff --git a/vendor/github.com/klauspost/compress/s2/dict.go b/vendor/github.com/klauspost/compress/s2/dict.go new file mode 100644 index 000000000..f125ad096 --- /dev/null +++ b/vendor/github.com/klauspost/compress/s2/dict.go @@ -0,0 +1,350 @@ +// Copyright (c) 2022+ Klaus Post. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package s2 + +import ( + "bytes" + "encoding/binary" + "sync" +) + +const ( + // MinDictSize is the minimum dictionary size when repeat has been read. + MinDictSize = 16 + + // MaxDictSize is the maximum dictionary size when repeat has been read. + MaxDictSize = 65536 + + // MaxDictSrcOffset is the maximum offset where a dictionary entry can start. + MaxDictSrcOffset = 65535 +) + +// Dict contains a dictionary that can be used for encoding and decoding s2 +type Dict struct { + dict []byte + repeat int // Repeat as index of dict + + fast, better, best sync.Once + fastTable *[1 << 14]uint16 + + betterTableShort *[1 << 14]uint16 + betterTableLong *[1 << 17]uint16 + + bestTableShort *[1 << 16]uint32 + bestTableLong *[1 << 19]uint32 +} + +// NewDict will read a dictionary. +// It will return nil if the dictionary is invalid. +func NewDict(dict []byte) *Dict { + if len(dict) == 0 { + return nil + } + var d Dict + // Repeat is the first value of the dict + r, n := binary.Uvarint(dict) + if n <= 0 { + return nil + } + dict = dict[n:] + d.dict = dict + if cap(d.dict) < len(d.dict)+16 { + d.dict = append(make([]byte, 0, len(d.dict)+16), d.dict...) + } + if len(dict) < MinDictSize || len(dict) > MaxDictSize { + return nil + } + d.repeat = int(r) + if d.repeat > len(dict) { + return nil + } + return &d +} + +// Bytes will return a serialized version of the dictionary. +// The output can be sent to NewDict. +func (d *Dict) Bytes() []byte { + dst := make([]byte, binary.MaxVarintLen16+len(d.dict)) + return append(dst[:binary.PutUvarint(dst, uint64(d.repeat))], d.dict...) +} + +// MakeDict will create a dictionary. +// 'data' must be at least MinDictSize. +// If data is longer than MaxDictSize only the last MaxDictSize bytes will be used. +// If searchStart is set the start repeat value will be set to the last +// match of this content. +// If no matches are found, it will attempt to find shorter matches. +// This content should match the typical start of a block. +// If at least 4 bytes cannot be matched, repeat is set to start of block. +func MakeDict(data []byte, searchStart []byte) *Dict { + if len(data) == 0 { + return nil + } + if len(data) > MaxDictSize { + data = data[len(data)-MaxDictSize:] + } + var d Dict + dict := data + d.dict = dict + if cap(d.dict) < len(d.dict)+16 { + d.dict = append(make([]byte, 0, len(d.dict)+16), d.dict...) + } + if len(dict) < MinDictSize { + return nil + } + + // Find the longest match possible, last entry if multiple. + for s := len(searchStart); s > 4; s-- { + if idx := bytes.LastIndex(data, searchStart[:s]); idx >= 0 && idx <= len(data)-8 { + d.repeat = idx + break + } + } + + return &d +} + +// MakeDictManual will create a dictionary. +// 'data' must be at least MinDictSize and less than or equal to MaxDictSize. +// A manual first repeat index into data must be provided. +// It must be less than len(data)-8. +func MakeDictManual(data []byte, firstIdx uint16) *Dict { + if len(data) < MinDictSize || int(firstIdx) >= len(data)-8 || len(data) > MaxDictSize { + return nil + } + var d Dict + dict := data + d.dict = dict + if cap(d.dict) < len(d.dict)+16 { + d.dict = append(make([]byte, 0, len(d.dict)+16), d.dict...) + } + + d.repeat = int(firstIdx) + return &d +} + +// Encode returns the encoded form of src. The returned slice may be a sub- +// slice of dst if dst was large enough to hold the entire encoded block. +// Otherwise, a newly allocated slice will be returned. +// +// The dst and src must not overlap. It is valid to pass a nil dst. +// +// The blocks will require the same amount of memory to decode as encoding, +// and does not make for concurrent decoding. +// Also note that blocks do not contain CRC information, so corruption may be undetected. +// +// If you need to encode larger amounts of data, consider using +// the streaming interface which gives all of these features. +func (d *Dict) Encode(dst, src []byte) []byte { + if n := MaxEncodedLen(len(src)); n < 0 { + panic(ErrTooLarge) + } else if cap(dst) < n { + dst = make([]byte, n) + } else { + dst = dst[:n] + } + + // The block starts with the varint-encoded length of the decompressed bytes. + dstP := binary.PutUvarint(dst, uint64(len(src))) + + if len(src) == 0 { + return dst[:dstP] + } + if len(src) < minNonLiteralBlockSize { + dstP += emitLiteral(dst[dstP:], src) + return dst[:dstP] + } + n := encodeBlockDictGo(dst[dstP:], src, d) + if n > 0 { + dstP += n + return dst[:dstP] + } + // Not compressible + dstP += emitLiteral(dst[dstP:], src) + return dst[:dstP] +} + +// EncodeBetter returns the encoded form of src. The returned slice may be a sub- +// slice of dst if dst was large enough to hold the entire encoded block. +// Otherwise, a newly allocated slice will be returned. +// +// EncodeBetter compresses better than Encode but typically with a +// 10-40% speed decrease on both compression and decompression. +// +// The dst and src must not overlap. It is valid to pass a nil dst. +// +// The blocks will require the same amount of memory to decode as encoding, +// and does not make for concurrent decoding. +// Also note that blocks do not contain CRC information, so corruption may be undetected. +// +// If you need to encode larger amounts of data, consider using +// the streaming interface which gives all of these features. +func (d *Dict) EncodeBetter(dst, src []byte) []byte { + if n := MaxEncodedLen(len(src)); n < 0 { + panic(ErrTooLarge) + } else if len(dst) < n { + dst = make([]byte, n) + } + + // The block starts with the varint-encoded length of the decompressed bytes. + dstP := binary.PutUvarint(dst, uint64(len(src))) + + if len(src) == 0 { + return dst[:dstP] + } + if len(src) < minNonLiteralBlockSize { + dstP += emitLiteral(dst[dstP:], src) + return dst[:dstP] + } + n := encodeBlockBetterDict(dst[dstP:], src, d) + if n > 0 { + dstP += n + return dst[:dstP] + } + // Not compressible + dstP += emitLiteral(dst[dstP:], src) + return dst[:dstP] +} + +// EncodeBest returns the encoded form of src. The returned slice may be a sub- +// slice of dst if dst was large enough to hold the entire encoded block. +// Otherwise, a newly allocated slice will be returned. +// +// EncodeBest compresses as good as reasonably possible but with a +// big speed decrease. +// +// The dst and src must not overlap. It is valid to pass a nil dst. +// +// The blocks will require the same amount of memory to decode as encoding, +// and does not make for concurrent decoding. +// Also note that blocks do not contain CRC information, so corruption may be undetected. +// +// If you need to encode larger amounts of data, consider using +// the streaming interface which gives all of these features. +func (d *Dict) EncodeBest(dst, src []byte) []byte { + if n := MaxEncodedLen(len(src)); n < 0 { + panic(ErrTooLarge) + } else if len(dst) < n { + dst = make([]byte, n) + } + + // The block starts with the varint-encoded length of the decompressed bytes. + dstP := binary.PutUvarint(dst, uint64(len(src))) + + if len(src) == 0 { + return dst[:dstP] + } + if len(src) < minNonLiteralBlockSize { + dstP += emitLiteral(dst[dstP:], src) + return dst[:dstP] + } + n := encodeBlockBest(dst[dstP:], src, d) + if n > 0 { + dstP += n + return dst[:dstP] + } + // Not compressible + dstP += emitLiteral(dst[dstP:], src) + return dst[:dstP] +} + +// Decode returns the decoded form of src. The returned slice may be a sub- +// slice of dst if dst was large enough to hold the entire decoded block. +// Otherwise, a newly allocated slice will be returned. +// +// The dst and src must not overlap. It is valid to pass a nil dst. +func (d *Dict) Decode(dst, src []byte) ([]byte, error) { + dLen, s, err := decodedLen(src) + if err != nil { + return nil, err + } + if dLen <= cap(dst) { + dst = dst[:dLen] + } else { + dst = make([]byte, dLen) + } + if s2DecodeDict(dst, src[s:], d) != 0 { + return nil, ErrCorrupt + } + return dst, nil +} + +func (d *Dict) initFast() { + d.fast.Do(func() { + const ( + tableBits = 14 + maxTableSize = 1 << tableBits + ) + + var table [maxTableSize]uint16 + // We stop so any entry of length 8 can always be read. + for i := 0; i < len(d.dict)-8-2; i += 3 { + x0 := load64(d.dict, i) + h0 := hash6(x0, tableBits) + h1 := hash6(x0>>8, tableBits) + h2 := hash6(x0>>16, tableBits) + table[h0] = uint16(i) + table[h1] = uint16(i + 1) + table[h2] = uint16(i + 2) + } + d.fastTable = &table + }) +} + +func (d *Dict) initBetter() { + d.better.Do(func() { + const ( + // Long hash matches. + lTableBits = 17 + maxLTableSize = 1 << lTableBits + + // Short hash matches. + sTableBits = 14 + maxSTableSize = 1 << sTableBits + ) + + var lTable [maxLTableSize]uint16 + var sTable [maxSTableSize]uint16 + + // We stop so any entry of length 8 can always be read. + for i := 0; i < len(d.dict)-8; i++ { + cv := load64(d.dict, i) + lTable[hash7(cv, lTableBits)] = uint16(i) + sTable[hash4(cv, sTableBits)] = uint16(i) + } + d.betterTableShort = &sTable + d.betterTableLong = &lTable + }) +} + +func (d *Dict) initBest() { + d.best.Do(func() { + const ( + // Long hash matches. + lTableBits = 19 + maxLTableSize = 1 << lTableBits + + // Short hash matches. + sTableBits = 16 + maxSTableSize = 1 << sTableBits + ) + + var lTable [maxLTableSize]uint32 + var sTable [maxSTableSize]uint32 + + // We stop so any entry of length 8 can always be read. + for i := 0; i < len(d.dict)-8; i++ { + cv := load64(d.dict, i) + hashL := hash8(cv, lTableBits) + hashS := hash4(cv, sTableBits) + candidateL := lTable[hashL] + candidateS := sTable[hashS] + lTable[hashL] = uint32(i) | candidateL<<16 + sTable[hashS] = uint32(i) | candidateS<<16 + } + d.bestTableShort = &sTable + d.bestTableLong = &lTable + }) +} diff --git a/vendor/github.com/klauspost/compress/s2/encode.go b/vendor/github.com/klauspost/compress/s2/encode.go new file mode 100644 index 000000000..0c9088adf --- /dev/null +++ b/vendor/github.com/klauspost/compress/s2/encode.go @@ -0,0 +1,393 @@ +// Copyright 2011 The Snappy-Go Authors. All rights reserved. +// Copyright (c) 2019 Klaus Post. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package s2 + +import ( + "encoding/binary" + "math" + "math/bits" +) + +// Encode returns the encoded form of src. The returned slice may be a sub- +// slice of dst if dst was large enough to hold the entire encoded block. +// Otherwise, a newly allocated slice will be returned. +// +// The dst and src must not overlap. It is valid to pass a nil dst. +// +// The blocks will require the same amount of memory to decode as encoding, +// and does not make for concurrent decoding. +// Also note that blocks do not contain CRC information, so corruption may be undetected. +// +// If you need to encode larger amounts of data, consider using +// the streaming interface which gives all of these features. +func Encode(dst, src []byte) []byte { + if n := MaxEncodedLen(len(src)); n < 0 { + panic(ErrTooLarge) + } else if cap(dst) < n { + dst = make([]byte, n) + } else { + dst = dst[:n] + } + + // The block starts with the varint-encoded length of the decompressed bytes. + d := binary.PutUvarint(dst, uint64(len(src))) + + if len(src) == 0 { + return dst[:d] + } + if len(src) < minNonLiteralBlockSize { + d += emitLiteral(dst[d:], src) + return dst[:d] + } + n := encodeBlock(dst[d:], src) + if n > 0 { + d += n + return dst[:d] + } + // Not compressible + d += emitLiteral(dst[d:], src) + return dst[:d] +} + +// EstimateBlockSize will perform a very fast compression +// without outputting the result and return the compressed output size. +// The function returns -1 if no improvement could be achieved. +// Using actual compression will most often produce better compression than the estimate. +func EstimateBlockSize(src []byte) (d int) { + if len(src) <= inputMargin || int64(len(src)) > 0xffffffff { + return -1 + } + if len(src) <= 1024 { + d = calcBlockSizeSmall(src) + } else { + d = calcBlockSize(src) + } + + if d == 0 { + return -1 + } + // Size of the varint encoded block size. + d += (bits.Len64(uint64(len(src))) + 7) / 7 + + if d >= len(src) { + return -1 + } + return d +} + +// EncodeBetter returns the encoded form of src. The returned slice may be a sub- +// slice of dst if dst was large enough to hold the entire encoded block. +// Otherwise, a newly allocated slice will be returned. +// +// EncodeBetter compresses better than Encode but typically with a +// 10-40% speed decrease on both compression and decompression. +// +// The dst and src must not overlap. It is valid to pass a nil dst. +// +// The blocks will require the same amount of memory to decode as encoding, +// and does not make for concurrent decoding. +// Also note that blocks do not contain CRC information, so corruption may be undetected. +// +// If you need to encode larger amounts of data, consider using +// the streaming interface which gives all of these features. +func EncodeBetter(dst, src []byte) []byte { + if n := MaxEncodedLen(len(src)); n < 0 { + panic(ErrTooLarge) + } else if len(dst) < n { + dst = make([]byte, n) + } + + // The block starts with the varint-encoded length of the decompressed bytes. + d := binary.PutUvarint(dst, uint64(len(src))) + + if len(src) == 0 { + return dst[:d] + } + if len(src) < minNonLiteralBlockSize { + d += emitLiteral(dst[d:], src) + return dst[:d] + } + n := encodeBlockBetter(dst[d:], src) + if n > 0 { + d += n + return dst[:d] + } + // Not compressible + d += emitLiteral(dst[d:], src) + return dst[:d] +} + +// EncodeBest returns the encoded form of src. The returned slice may be a sub- +// slice of dst if dst was large enough to hold the entire encoded block. +// Otherwise, a newly allocated slice will be returned. +// +// EncodeBest compresses as good as reasonably possible but with a +// big speed decrease. +// +// The dst and src must not overlap. It is valid to pass a nil dst. +// +// The blocks will require the same amount of memory to decode as encoding, +// and does not make for concurrent decoding. +// Also note that blocks do not contain CRC information, so corruption may be undetected. +// +// If you need to encode larger amounts of data, consider using +// the streaming interface which gives all of these features. +func EncodeBest(dst, src []byte) []byte { + if n := MaxEncodedLen(len(src)); n < 0 { + panic(ErrTooLarge) + } else if len(dst) < n { + dst = make([]byte, n) + } + + // The block starts with the varint-encoded length of the decompressed bytes. + d := binary.PutUvarint(dst, uint64(len(src))) + + if len(src) == 0 { + return dst[:d] + } + if len(src) < minNonLiteralBlockSize { + d += emitLiteral(dst[d:], src) + return dst[:d] + } + n := encodeBlockBest(dst[d:], src, nil) + if n > 0 { + d += n + return dst[:d] + } + // Not compressible + d += emitLiteral(dst[d:], src) + return dst[:d] +} + +// EncodeSnappy returns the encoded form of src. The returned slice may be a sub- +// slice of dst if dst was large enough to hold the entire encoded block. +// Otherwise, a newly allocated slice will be returned. +// +// The output is Snappy compatible and will likely decompress faster. +// +// The dst and src must not overlap. It is valid to pass a nil dst. +// +// The blocks will require the same amount of memory to decode as encoding, +// and does not make for concurrent decoding. +// Also note that blocks do not contain CRC information, so corruption may be undetected. +// +// If you need to encode larger amounts of data, consider using +// the streaming interface which gives all of these features. +func EncodeSnappy(dst, src []byte) []byte { + if n := MaxEncodedLen(len(src)); n < 0 { + panic(ErrTooLarge) + } else if cap(dst) < n { + dst = make([]byte, n) + } else { + dst = dst[:n] + } + + // The block starts with the varint-encoded length of the decompressed bytes. + d := binary.PutUvarint(dst, uint64(len(src))) + + if len(src) == 0 { + return dst[:d] + } + if len(src) < minNonLiteralBlockSize { + d += emitLiteral(dst[d:], src) + return dst[:d] + } + + n := encodeBlockSnappy(dst[d:], src) + if n > 0 { + d += n + return dst[:d] + } + // Not compressible + d += emitLiteral(dst[d:], src) + return dst[:d] +} + +// EncodeSnappyBetter returns the encoded form of src. The returned slice may be a sub- +// slice of dst if dst was large enough to hold the entire encoded block. +// Otherwise, a newly allocated slice will be returned. +// +// The output is Snappy compatible and will likely decompress faster. +// +// The dst and src must not overlap. It is valid to pass a nil dst. +// +// The blocks will require the same amount of memory to decode as encoding, +// and does not make for concurrent decoding. +// Also note that blocks do not contain CRC information, so corruption may be undetected. +// +// If you need to encode larger amounts of data, consider using +// the streaming interface which gives all of these features. +func EncodeSnappyBetter(dst, src []byte) []byte { + if n := MaxEncodedLen(len(src)); n < 0 { + panic(ErrTooLarge) + } else if cap(dst) < n { + dst = make([]byte, n) + } else { + dst = dst[:n] + } + + // The block starts with the varint-encoded length of the decompressed bytes. + d := binary.PutUvarint(dst, uint64(len(src))) + + if len(src) == 0 { + return dst[:d] + } + if len(src) < minNonLiteralBlockSize { + d += emitLiteral(dst[d:], src) + return dst[:d] + } + + n := encodeBlockBetterSnappy(dst[d:], src) + if n > 0 { + d += n + return dst[:d] + } + // Not compressible + d += emitLiteral(dst[d:], src) + return dst[:d] +} + +// EncodeSnappyBest returns the encoded form of src. The returned slice may be a sub- +// slice of dst if dst was large enough to hold the entire encoded block. +// Otherwise, a newly allocated slice will be returned. +// +// The output is Snappy compatible and will likely decompress faster. +// +// The dst and src must not overlap. It is valid to pass a nil dst. +// +// The blocks will require the same amount of memory to decode as encoding, +// and does not make for concurrent decoding. +// Also note that blocks do not contain CRC information, so corruption may be undetected. +// +// If you need to encode larger amounts of data, consider using +// the streaming interface which gives all of these features. +func EncodeSnappyBest(dst, src []byte) []byte { + if n := MaxEncodedLen(len(src)); n < 0 { + panic(ErrTooLarge) + } else if cap(dst) < n { + dst = make([]byte, n) + } else { + dst = dst[:n] + } + + // The block starts with the varint-encoded length of the decompressed bytes. + d := binary.PutUvarint(dst, uint64(len(src))) + + if len(src) == 0 { + return dst[:d] + } + if len(src) < minNonLiteralBlockSize { + d += emitLiteral(dst[d:], src) + return dst[:d] + } + + n := encodeBlockBestSnappy(dst[d:], src) + if n > 0 { + d += n + return dst[:d] + } + // Not compressible + d += emitLiteral(dst[d:], src) + return dst[:d] +} + +// ConcatBlocks will concatenate the supplied blocks and append them to the supplied destination. +// If the destination is nil or too small, a new will be allocated. +// The blocks are not validated, so garbage in = garbage out. +// dst may not overlap block data. +// Any data in dst is preserved as is, so it will not be considered a block. +func ConcatBlocks(dst []byte, blocks ...[]byte) ([]byte, error) { + totalSize := uint64(0) + compSize := 0 + for _, b := range blocks { + l, hdr, err := decodedLen(b) + if err != nil { + return nil, err + } + totalSize += uint64(l) + compSize += len(b) - hdr + } + if totalSize == 0 { + dst = append(dst, 0) + return dst, nil + } + if totalSize > math.MaxUint32 { + return nil, ErrTooLarge + } + var tmp [binary.MaxVarintLen32]byte + hdrSize := binary.PutUvarint(tmp[:], totalSize) + wantSize := hdrSize + compSize + + if cap(dst)-len(dst) < wantSize { + dst = append(make([]byte, 0, wantSize+len(dst)), dst...) + } + dst = append(dst, tmp[:hdrSize]...) + for _, b := range blocks { + _, hdr, err := decodedLen(b) + if err != nil { + return nil, err + } + dst = append(dst, b[hdr:]...) + } + return dst, nil +} + +// inputMargin is the minimum number of extra input bytes to keep, inside +// encodeBlock's inner loop. On some architectures, this margin lets us +// implement a fast path for emitLiteral, where the copy of short (<= 16 byte) +// literals can be implemented as a single load to and store from a 16-byte +// register. That literal's actual length can be as short as 1 byte, so this +// can copy up to 15 bytes too much, but that's OK as subsequent iterations of +// the encoding loop will fix up the copy overrun, and this inputMargin ensures +// that we don't overrun the dst and src buffers. +const inputMargin = 8 + +// minNonLiteralBlockSize is the minimum size of the input to encodeBlock that +// will be accepted by the encoder. +const minNonLiteralBlockSize = 32 + +const intReduction = 2 - (1 << (^uint(0) >> 63)) // 1 (32 bits) or 0 (64 bits) + +// MaxBlockSize is the maximum value where MaxEncodedLen will return a valid block size. +// Blocks this big are highly discouraged, though. +// Half the size on 32 bit systems. +const MaxBlockSize = (1<<(32-intReduction) - 1) - binary.MaxVarintLen32 - 5 + +// MaxEncodedLen returns the maximum length of a snappy block, given its +// uncompressed length. +// +// It will return a negative value if srcLen is too large to encode. +// 32 bit platforms will have lower thresholds for rejecting big content. +func MaxEncodedLen(srcLen int) int { + n := uint64(srcLen) + if intReduction == 1 { + // 32 bits + if n > math.MaxInt32 { + // Also includes negative. + return -1 + } + } else if n > 0xffffffff { + // 64 bits + // Also includes negative. + return -1 + } + // Size of the varint encoded block size. + n = n + uint64((bits.Len64(n)+7)/7) + + // Add maximum size of encoding block as literals. + n += uint64(literalExtraSize(int64(srcLen))) + if intReduction == 1 { + // 32 bits + if n > math.MaxInt32 { + return -1 + } + } else if n > 0xffffffff { + // 64 bits + // Also includes negative. + return -1 + } + return int(n) +} diff --git a/vendor/github.com/klauspost/compress/s2/encode_all.go b/vendor/github.com/klauspost/compress/s2/encode_all.go new file mode 100644 index 000000000..997704569 --- /dev/null +++ b/vendor/github.com/klauspost/compress/s2/encode_all.go @@ -0,0 +1,1068 @@ +// Copyright 2016 The Snappy-Go Authors. All rights reserved. +// Copyright (c) 2019 Klaus Post. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package s2 + +import ( + "bytes" + "encoding/binary" + "fmt" + "math/bits" +) + +func load32(b []byte, i int) uint32 { + return binary.LittleEndian.Uint32(b[i:]) +} + +func load64(b []byte, i int) uint64 { + return binary.LittleEndian.Uint64(b[i:]) +} + +// hash6 returns the hash of the lowest 6 bytes of u to fit in a hash table with h bits. +// Preferably h should be a constant and should always be <64. +func hash6(u uint64, h uint8) uint32 { + const prime6bytes = 227718039650203 + return uint32(((u << (64 - 48)) * prime6bytes) >> ((64 - h) & 63)) +} + +func encodeGo(dst, src []byte) []byte { + if n := MaxEncodedLen(len(src)); n < 0 { + panic(ErrTooLarge) + } else if len(dst) < n { + dst = make([]byte, n) + } + + // The block starts with the varint-encoded length of the decompressed bytes. + d := binary.PutUvarint(dst, uint64(len(src))) + + if len(src) == 0 { + return dst[:d] + } + if len(src) < minNonLiteralBlockSize { + d += emitLiteral(dst[d:], src) + return dst[:d] + } + n := encodeBlockGo(dst[d:], src) + if n > 0 { + d += n + return dst[:d] + } + // Not compressible + d += emitLiteral(dst[d:], src) + return dst[:d] +} + +// encodeBlockGo encodes a non-empty src to a guaranteed-large-enough dst. It +// assumes that the varint-encoded length of the decompressed bytes has already +// been written. +// +// It also assumes that: +// +// len(dst) >= MaxEncodedLen(len(src)) && +// minNonLiteralBlockSize <= len(src) && len(src) <= maxBlockSize +func encodeBlockGo(dst, src []byte) (d int) { + // Initialize the hash table. + const ( + tableBits = 14 + maxTableSize = 1 << tableBits + + debug = false + ) + + var table [maxTableSize]uint32 + + // sLimit is when to stop looking for offset/length copies. The inputMargin + // lets us use a fast path for emitLiteral in the main loop, while we are + // looking for copies. + sLimit := len(src) - inputMargin + + // Bail if we can't compress to at least this. + dstLimit := len(src) - len(src)>>5 - 5 + + // nextEmit is where in src the next emitLiteral should start from. + nextEmit := 0 + + // The encoded form must start with a literal, as there are no previous + // bytes to copy, so we start looking for hash matches at s == 1. + s := 1 + cv := load64(src, s) + + // We search for a repeat at -1, but don't output repeats when nextEmit == 0 + repeat := 1 + + for { + candidate := 0 + for { + // Next src position to check + nextS := s + (s-nextEmit)>>6 + 4 + if nextS > sLimit { + goto emitRemainder + } + hash0 := hash6(cv, tableBits) + hash1 := hash6(cv>>8, tableBits) + candidate = int(table[hash0]) + candidate2 := int(table[hash1]) + table[hash0] = uint32(s) + table[hash1] = uint32(s + 1) + hash2 := hash6(cv>>16, tableBits) + + // Check repeat at offset checkRep. + const checkRep = 1 + if uint32(cv>>(checkRep*8)) == load32(src, s-repeat+checkRep) { + base := s + checkRep + // Extend back + for i := base - repeat; base > nextEmit && i > 0 && src[i-1] == src[base-1]; { + i-- + base-- + } + + // Bail if we exceed the maximum size. + if d+(base-nextEmit) > dstLimit { + return 0 + } + + d += emitLiteral(dst[d:], src[nextEmit:base]) + + // Extend forward + candidate := s - repeat + 4 + checkRep + s += 4 + checkRep + for s <= sLimit { + if diff := load64(src, s) ^ load64(src, candidate); diff != 0 { + s += bits.TrailingZeros64(diff) >> 3 + break + } + s += 8 + candidate += 8 + } + if debug { + // Validate match. + if s <= candidate { + panic("s <= candidate") + } + a := src[base:s] + b := src[base-repeat : base-repeat+(s-base)] + if !bytes.Equal(a, b) { + panic("mismatch") + } + } + if nextEmit > 0 { + // same as `add := emitCopy(dst[d:], repeat, s-base)` but skips storing offset. + d += emitRepeat(dst[d:], repeat, s-base) + } else { + // First match, cannot be repeat. + d += emitCopy(dst[d:], repeat, s-base) + } + nextEmit = s + if s >= sLimit { + goto emitRemainder + } + cv = load64(src, s) + continue + } + + if uint32(cv) == load32(src, candidate) { + break + } + candidate = int(table[hash2]) + if uint32(cv>>8) == load32(src, candidate2) { + table[hash2] = uint32(s + 2) + candidate = candidate2 + s++ + break + } + table[hash2] = uint32(s + 2) + if uint32(cv>>16) == load32(src, candidate) { + s += 2 + break + } + + cv = load64(src, nextS) + s = nextS + } + + // Extend backwards. + // The top bytes will be rechecked to get the full match. + for candidate > 0 && s > nextEmit && src[candidate-1] == src[s-1] { + candidate-- + s-- + } + + // Bail if we exceed the maximum size. + if d+(s-nextEmit) > dstLimit { + return 0 + } + + // A 4-byte match has been found. We'll later see if more than 4 bytes + // match. But, prior to the match, src[nextEmit:s] are unmatched. Emit + // them as literal bytes. + + d += emitLiteral(dst[d:], src[nextEmit:s]) + + // Call emitCopy, and then see if another emitCopy could be our next + // move. Repeat until we find no match for the input immediately after + // what was consumed by the last emitCopy call. + // + // If we exit this loop normally then we need to call emitLiteral next, + // though we don't yet know how big the literal will be. We handle that + // by proceeding to the next iteration of the main loop. We also can + // exit this loop via goto if we get close to exhausting the input. + for { + // Invariant: we have a 4-byte match at s, and no need to emit any + // literal bytes prior to s. + base := s + repeat = base - candidate + + // Extend the 4-byte match as long as possible. + s += 4 + candidate += 4 + for s <= len(src)-8 { + if diff := load64(src, s) ^ load64(src, candidate); diff != 0 { + s += bits.TrailingZeros64(diff) >> 3 + break + } + s += 8 + candidate += 8 + } + + d += emitCopy(dst[d:], repeat, s-base) + if debug { + // Validate match. + if s <= candidate { + panic("s <= candidate") + } + a := src[base:s] + b := src[base-repeat : base-repeat+(s-base)] + if !bytes.Equal(a, b) { + panic("mismatch") + } + } + + nextEmit = s + if s >= sLimit { + goto emitRemainder + } + + if d > dstLimit { + // Do we have space for more, if not bail. + return 0 + } + // Check for an immediate match, otherwise start search at s+1 + x := load64(src, s-2) + m2Hash := hash6(x, tableBits) + currHash := hash6(x>>16, tableBits) + candidate = int(table[currHash]) + table[m2Hash] = uint32(s - 2) + table[currHash] = uint32(s) + if debug && s == candidate { + panic("s == candidate") + } + if uint32(x>>16) != load32(src, candidate) { + cv = load64(src, s+1) + s++ + break + } + } + } + +emitRemainder: + if nextEmit < len(src) { + // Bail if we exceed the maximum size. + if d+len(src)-nextEmit > dstLimit { + return 0 + } + d += emitLiteral(dst[d:], src[nextEmit:]) + } + return d +} + +func encodeBlockSnappyGo(dst, src []byte) (d int) { + // Initialize the hash table. + const ( + tableBits = 14 + maxTableSize = 1 << tableBits + ) + + var table [maxTableSize]uint32 + + // sLimit is when to stop looking for offset/length copies. The inputMargin + // lets us use a fast path for emitLiteral in the main loop, while we are + // looking for copies. + sLimit := len(src) - inputMargin + + // Bail if we can't compress to at least this. + dstLimit := len(src) - len(src)>>5 - 5 + + // nextEmit is where in src the next emitLiteral should start from. + nextEmit := 0 + + // The encoded form must start with a literal, as there are no previous + // bytes to copy, so we start looking for hash matches at s == 1. + s := 1 + cv := load64(src, s) + + // We search for a repeat at -1, but don't output repeats when nextEmit == 0 + repeat := 1 + + for { + candidate := 0 + for { + // Next src position to check + nextS := s + (s-nextEmit)>>6 + 4 + if nextS > sLimit { + goto emitRemainder + } + hash0 := hash6(cv, tableBits) + hash1 := hash6(cv>>8, tableBits) + candidate = int(table[hash0]) + candidate2 := int(table[hash1]) + table[hash0] = uint32(s) + table[hash1] = uint32(s + 1) + hash2 := hash6(cv>>16, tableBits) + + // Check repeat at offset checkRep. + const checkRep = 1 + if uint32(cv>>(checkRep*8)) == load32(src, s-repeat+checkRep) { + base := s + checkRep + // Extend back + for i := base - repeat; base > nextEmit && i > 0 && src[i-1] == src[base-1]; { + i-- + base-- + } + // Bail if we exceed the maximum size. + if d+(base-nextEmit) > dstLimit { + return 0 + } + + d += emitLiteral(dst[d:], src[nextEmit:base]) + + // Extend forward + candidate := s - repeat + 4 + checkRep + s += 4 + checkRep + for s <= sLimit { + if diff := load64(src, s) ^ load64(src, candidate); diff != 0 { + s += bits.TrailingZeros64(diff) >> 3 + break + } + s += 8 + candidate += 8 + } + + d += emitCopyNoRepeat(dst[d:], repeat, s-base) + nextEmit = s + if s >= sLimit { + goto emitRemainder + } + + cv = load64(src, s) + continue + } + + if uint32(cv) == load32(src, candidate) { + break + } + candidate = int(table[hash2]) + if uint32(cv>>8) == load32(src, candidate2) { + table[hash2] = uint32(s + 2) + candidate = candidate2 + s++ + break + } + table[hash2] = uint32(s + 2) + if uint32(cv>>16) == load32(src, candidate) { + s += 2 + break + } + + cv = load64(src, nextS) + s = nextS + } + + // Extend backwards + for candidate > 0 && s > nextEmit && src[candidate-1] == src[s-1] { + candidate-- + s-- + } + + // Bail if we exceed the maximum size. + if d+(s-nextEmit) > dstLimit { + return 0 + } + + // A 4-byte match has been found. We'll later see if more than 4 bytes + // match. But, prior to the match, src[nextEmit:s] are unmatched. Emit + // them as literal bytes. + + d += emitLiteral(dst[d:], src[nextEmit:s]) + + // Call emitCopy, and then see if another emitCopy could be our next + // move. Repeat until we find no match for the input immediately after + // what was consumed by the last emitCopy call. + // + // If we exit this loop normally then we need to call emitLiteral next, + // though we don't yet know how big the literal will be. We handle that + // by proceeding to the next iteration of the main loop. We also can + // exit this loop via goto if we get close to exhausting the input. + for { + // Invariant: we have a 4-byte match at s, and no need to emit any + // literal bytes prior to s. + base := s + repeat = base - candidate + + // Extend the 4-byte match as long as possible. + s += 4 + candidate += 4 + for s <= len(src)-8 { + if diff := load64(src, s) ^ load64(src, candidate); diff != 0 { + s += bits.TrailingZeros64(diff) >> 3 + break + } + s += 8 + candidate += 8 + } + + d += emitCopyNoRepeat(dst[d:], repeat, s-base) + if false { + // Validate match. + a := src[base:s] + b := src[base-repeat : base-repeat+(s-base)] + if !bytes.Equal(a, b) { + panic("mismatch") + } + } + + nextEmit = s + if s >= sLimit { + goto emitRemainder + } + + if d > dstLimit { + // Do we have space for more, if not bail. + return 0 + } + // Check for an immediate match, otherwise start search at s+1 + x := load64(src, s-2) + m2Hash := hash6(x, tableBits) + currHash := hash6(x>>16, tableBits) + candidate = int(table[currHash]) + table[m2Hash] = uint32(s - 2) + table[currHash] = uint32(s) + if uint32(x>>16) != load32(src, candidate) { + cv = load64(src, s+1) + s++ + break + } + } + } + +emitRemainder: + if nextEmit < len(src) { + // Bail if we exceed the maximum size. + if d+len(src)-nextEmit > dstLimit { + return 0 + } + d += emitLiteral(dst[d:], src[nextEmit:]) + } + return d +} + +// encodeBlockGo encodes a non-empty src to a guaranteed-large-enough dst. It +// assumes that the varint-encoded length of the decompressed bytes has already +// been written. +// +// It also assumes that: +// +// len(dst) >= MaxEncodedLen(len(src)) && +// minNonLiteralBlockSize <= len(src) && len(src) <= maxBlockSize +func encodeBlockDictGo(dst, src []byte, dict *Dict) (d int) { + // Initialize the hash table. + const ( + tableBits = 14 + maxTableSize = 1 << tableBits + maxAhead = 8 // maximum bytes ahead without checking sLimit + + debug = false + ) + dict.initFast() + + var table [maxTableSize]uint32 + + // sLimit is when to stop looking for offset/length copies. The inputMargin + // lets us use a fast path for emitLiteral in the main loop, while we are + // looking for copies. + sLimit := len(src) - inputMargin + if sLimit > MaxDictSrcOffset-maxAhead { + sLimit = MaxDictSrcOffset - maxAhead + } + + // Bail if we can't compress to at least this. + dstLimit := len(src) - len(src)>>5 - 5 + + // nextEmit is where in src the next emitLiteral should start from. + nextEmit := 0 + + // The encoded form can start with a dict entry (copy or repeat). + s := 0 + + // Convert dict repeat to offset + repeat := len(dict.dict) - dict.repeat + cv := load64(src, 0) + + // While in dict +searchDict: + for { + // Next src position to check + nextS := s + (s-nextEmit)>>6 + 4 + hash0 := hash6(cv, tableBits) + hash1 := hash6(cv>>8, tableBits) + if nextS > sLimit { + if debug { + fmt.Println("slimit reached", s, nextS) + } + break searchDict + } + candidateDict := int(dict.fastTable[hash0]) + candidateDict2 := int(dict.fastTable[hash1]) + candidate2 := int(table[hash1]) + candidate := int(table[hash0]) + table[hash0] = uint32(s) + table[hash1] = uint32(s + 1) + hash2 := hash6(cv>>16, tableBits) + + // Check repeat at offset checkRep. + const checkRep = 1 + + if repeat > s { + candidate := len(dict.dict) - repeat + s + if repeat-s >= 4 && uint32(cv) == load32(dict.dict, candidate) { + // Extend back + base := s + for i := candidate; base > nextEmit && i > 0 && dict.dict[i-1] == src[base-1]; { + i-- + base-- + } + // Bail if we exceed the maximum size. + if d+(base-nextEmit) > dstLimit { + return 0 + } + + d += emitLiteral(dst[d:], src[nextEmit:base]) + if debug && nextEmit != base { + fmt.Println("emitted ", base-nextEmit, "literals") + } + s += 4 + candidate += 4 + for candidate < len(dict.dict)-8 && s <= len(src)-8 { + if diff := load64(src, s) ^ load64(dict.dict, candidate); diff != 0 { + s += bits.TrailingZeros64(diff) >> 3 + break + } + s += 8 + candidate += 8 + } + d += emitRepeat(dst[d:], repeat, s-base) + if debug { + fmt.Println("emitted dict repeat length", s-base, "offset:", repeat, "s:", s) + } + nextEmit = s + if s >= sLimit { + break searchDict + } + cv = load64(src, s) + continue + } + } else if uint32(cv>>(checkRep*8)) == load32(src, s-repeat+checkRep) { + base := s + checkRep + // Extend back + for i := base - repeat; base > nextEmit && i > 0 && src[i-1] == src[base-1]; { + i-- + base-- + } + d += emitLiteral(dst[d:], src[nextEmit:base]) + if debug && nextEmit != base { + fmt.Println("emitted ", base-nextEmit, "literals") + } + + // Extend forward + candidate := s - repeat + 4 + checkRep + s += 4 + checkRep + for s <= sLimit { + if diff := load64(src, s) ^ load64(src, candidate); diff != 0 { + s += bits.TrailingZeros64(diff) >> 3 + break + } + s += 8 + candidate += 8 + } + if debug { + // Validate match. + if s <= candidate { + panic("s <= candidate") + } + a := src[base:s] + b := src[base-repeat : base-repeat+(s-base)] + if !bytes.Equal(a, b) { + panic("mismatch") + } + } + + if nextEmit > 0 { + // same as `add := emitCopy(dst[d:], repeat, s-base)` but skips storing offset. + d += emitRepeat(dst[d:], repeat, s-base) + } else { + // First match, cannot be repeat. + d += emitCopy(dst[d:], repeat, s-base) + } + + nextEmit = s + if s >= sLimit { + break searchDict + } + if debug { + fmt.Println("emitted reg repeat", s-base, "s:", s) + } + cv = load64(src, s) + continue searchDict + } + if s == 0 { + cv = load64(src, nextS) + s = nextS + continue searchDict + } + // Start with table. These matches will always be closer. + if uint32(cv) == load32(src, candidate) { + goto emitMatch + } + candidate = int(table[hash2]) + if uint32(cv>>8) == load32(src, candidate2) { + table[hash2] = uint32(s + 2) + candidate = candidate2 + s++ + goto emitMatch + } + + // Check dict. Dicts have longer offsets, so we want longer matches. + if cv == load64(dict.dict, candidateDict) { + table[hash2] = uint32(s + 2) + goto emitDict + } + + candidateDict = int(dict.fastTable[hash2]) + // Check if upper 7 bytes match + if candidateDict2 >= 1 { + if cv^load64(dict.dict, candidateDict2-1) < (1 << 8) { + table[hash2] = uint32(s + 2) + candidateDict = candidateDict2 + s++ + goto emitDict + } + } + + table[hash2] = uint32(s + 2) + if uint32(cv>>16) == load32(src, candidate) { + s += 2 + goto emitMatch + } + if candidateDict >= 2 { + // Check if upper 6 bytes match + if cv^load64(dict.dict, candidateDict-2) < (1 << 16) { + s += 2 + goto emitDict + } + } + + cv = load64(src, nextS) + s = nextS + continue searchDict + + emitDict: + { + if debug { + if load32(dict.dict, candidateDict) != load32(src, s) { + panic("dict emit mismatch") + } + } + // Extend backwards. + // The top bytes will be rechecked to get the full match. + for candidateDict > 0 && s > nextEmit && dict.dict[candidateDict-1] == src[s-1] { + candidateDict-- + s-- + } + + // Bail if we exceed the maximum size. + if d+(s-nextEmit) > dstLimit { + return 0 + } + + // A 4-byte match has been found. We'll later see if more than 4 bytes + // match. But, prior to the match, src[nextEmit:s] are unmatched. Emit + // them as literal bytes. + + d += emitLiteral(dst[d:], src[nextEmit:s]) + if debug && nextEmit != s { + fmt.Println("emitted ", s-nextEmit, "literals") + } + { + // Invariant: we have a 4-byte match at s, and no need to emit any + // literal bytes prior to s. + base := s + repeat = s + (len(dict.dict)) - candidateDict + + // Extend the 4-byte match as long as possible. + s += 4 + candidateDict += 4 + for s <= len(src)-8 && len(dict.dict)-candidateDict >= 8 { + if diff := load64(src, s) ^ load64(dict.dict, candidateDict); diff != 0 { + s += bits.TrailingZeros64(diff) >> 3 + break + } + s += 8 + candidateDict += 8 + } + + // Matches longer than 64 are split. + if s <= sLimit || s-base < 8 { + d += emitCopy(dst[d:], repeat, s-base) + } else { + // Split to ensure we don't start a copy within next block + d += emitCopy(dst[d:], repeat, 4) + d += emitRepeat(dst[d:], repeat, s-base-4) + } + if false { + // Validate match. + if s <= candidate { + panic("s <= candidate") + } + a := src[base:s] + b := dict.dict[base-repeat : base-repeat+(s-base)] + if !bytes.Equal(a, b) { + panic("mismatch") + } + } + if debug { + fmt.Println("emitted dict copy, length", s-base, "offset:", repeat, "s:", s) + } + nextEmit = s + if s >= sLimit { + break searchDict + } + + if d > dstLimit { + // Do we have space for more, if not bail. + return 0 + } + + // Index and continue loop to try new candidate. + x := load64(src, s-2) + m2Hash := hash6(x, tableBits) + currHash := hash6(x>>8, tableBits) + table[m2Hash] = uint32(s - 2) + table[currHash] = uint32(s - 1) + cv = load64(src, s) + } + continue + } + emitMatch: + + // Extend backwards. + // The top bytes will be rechecked to get the full match. + for candidate > 0 && s > nextEmit && src[candidate-1] == src[s-1] { + candidate-- + s-- + } + + // Bail if we exceed the maximum size. + if d+(s-nextEmit) > dstLimit { + return 0 + } + + // A 4-byte match has been found. We'll later see if more than 4 bytes + // match. But, prior to the match, src[nextEmit:s] are unmatched. Emit + // them as literal bytes. + + d += emitLiteral(dst[d:], src[nextEmit:s]) + if debug && nextEmit != s { + fmt.Println("emitted ", s-nextEmit, "literals") + } + // Call emitCopy, and then see if another emitCopy could be our next + // move. Repeat until we find no match for the input immediately after + // what was consumed by the last emitCopy call. + // + // If we exit this loop normally then we need to call emitLiteral next, + // though we don't yet know how big the literal will be. We handle that + // by proceeding to the next iteration of the main loop. We also can + // exit this loop via goto if we get close to exhausting the input. + for { + // Invariant: we have a 4-byte match at s, and no need to emit any + // literal bytes prior to s. + base := s + repeat = base - candidate + + // Extend the 4-byte match as long as possible. + s += 4 + candidate += 4 + for s <= len(src)-8 { + if diff := load64(src, s) ^ load64(src, candidate); diff != 0 { + s += bits.TrailingZeros64(diff) >> 3 + break + } + s += 8 + candidate += 8 + } + + d += emitCopy(dst[d:], repeat, s-base) + if debug { + // Validate match. + if s <= candidate { + panic("s <= candidate") + } + a := src[base:s] + b := src[base-repeat : base-repeat+(s-base)] + if !bytes.Equal(a, b) { + panic("mismatch") + } + } + if debug { + fmt.Println("emitted src copy, length", s-base, "offset:", repeat, "s:", s) + } + nextEmit = s + if s >= sLimit { + break searchDict + } + + if d > dstLimit { + // Do we have space for more, if not bail. + return 0 + } + // Check for an immediate match, otherwise start search at s+1 + x := load64(src, s-2) + m2Hash := hash6(x, tableBits) + currHash := hash6(x>>16, tableBits) + candidate = int(table[currHash]) + table[m2Hash] = uint32(s - 2) + table[currHash] = uint32(s) + if debug && s == candidate { + panic("s == candidate") + } + if uint32(x>>16) != load32(src, candidate) { + cv = load64(src, s+1) + s++ + break + } + } + } + + // Search without dict: + if repeat > s { + repeat = 0 + } + + // No more dict + sLimit = len(src) - inputMargin + if s >= sLimit { + goto emitRemainder + } + if debug { + fmt.Println("non-dict matching at", s, "repeat:", repeat) + } + cv = load64(src, s) + if debug { + fmt.Println("now", s, "->", sLimit, "out:", d, "left:", len(src)-s, "nextemit:", nextEmit, "dstLimit:", dstLimit, "s:", s) + } + for { + candidate := 0 + for { + // Next src position to check + nextS := s + (s-nextEmit)>>6 + 4 + if nextS > sLimit { + goto emitRemainder + } + hash0 := hash6(cv, tableBits) + hash1 := hash6(cv>>8, tableBits) + candidate = int(table[hash0]) + candidate2 := int(table[hash1]) + table[hash0] = uint32(s) + table[hash1] = uint32(s + 1) + hash2 := hash6(cv>>16, tableBits) + + // Check repeat at offset checkRep. + const checkRep = 1 + if repeat > 0 && uint32(cv>>(checkRep*8)) == load32(src, s-repeat+checkRep) { + base := s + checkRep + // Extend back + for i := base - repeat; base > nextEmit && i > 0 && src[i-1] == src[base-1]; { + i-- + base-- + } + // Bail if we exceed the maximum size. + if d+(base-nextEmit) > dstLimit { + return 0 + } + + d += emitLiteral(dst[d:], src[nextEmit:base]) + if debug && nextEmit != base { + fmt.Println("emitted ", base-nextEmit, "literals") + } + // Extend forward + candidate := s - repeat + 4 + checkRep + s += 4 + checkRep + for s <= sLimit { + if diff := load64(src, s) ^ load64(src, candidate); diff != 0 { + s += bits.TrailingZeros64(diff) >> 3 + break + } + s += 8 + candidate += 8 + } + if debug { + // Validate match. + if s <= candidate { + panic("s <= candidate") + } + a := src[base:s] + b := src[base-repeat : base-repeat+(s-base)] + if !bytes.Equal(a, b) { + panic("mismatch") + } + } + if nextEmit > 0 { + // same as `add := emitCopy(dst[d:], repeat, s-base)` but skips storing offset. + d += emitRepeat(dst[d:], repeat, s-base) + } else { + // First match, cannot be repeat. + d += emitCopy(dst[d:], repeat, s-base) + } + if debug { + fmt.Println("emitted src repeat length", s-base, "offset:", repeat, "s:", s) + } + nextEmit = s + if s >= sLimit { + goto emitRemainder + } + + cv = load64(src, s) + continue + } + + if uint32(cv) == load32(src, candidate) { + break + } + candidate = int(table[hash2]) + if uint32(cv>>8) == load32(src, candidate2) { + table[hash2] = uint32(s + 2) + candidate = candidate2 + s++ + break + } + table[hash2] = uint32(s + 2) + if uint32(cv>>16) == load32(src, candidate) { + s += 2 + break + } + + cv = load64(src, nextS) + s = nextS + } + + // Extend backwards. + // The top bytes will be rechecked to get the full match. + for candidate > 0 && s > nextEmit && src[candidate-1] == src[s-1] { + candidate-- + s-- + } + + // Bail if we exceed the maximum size. + if d+(s-nextEmit) > dstLimit { + return 0 + } + + // A 4-byte match has been found. We'll later see if more than 4 bytes + // match. But, prior to the match, src[nextEmit:s] are unmatched. Emit + // them as literal bytes. + + d += emitLiteral(dst[d:], src[nextEmit:s]) + if debug && nextEmit != s { + fmt.Println("emitted ", s-nextEmit, "literals") + } + // Call emitCopy, and then see if another emitCopy could be our next + // move. Repeat until we find no match for the input immediately after + // what was consumed by the last emitCopy call. + // + // If we exit this loop normally then we need to call emitLiteral next, + // though we don't yet know how big the literal will be. We handle that + // by proceeding to the next iteration of the main loop. We also can + // exit this loop via goto if we get close to exhausting the input. + for { + // Invariant: we have a 4-byte match at s, and no need to emit any + // literal bytes prior to s. + base := s + repeat = base - candidate + + // Extend the 4-byte match as long as possible. + s += 4 + candidate += 4 + for s <= len(src)-8 { + if diff := load64(src, s) ^ load64(src, candidate); diff != 0 { + s += bits.TrailingZeros64(diff) >> 3 + break + } + s += 8 + candidate += 8 + } + + d += emitCopy(dst[d:], repeat, s-base) + if debug { + // Validate match. + if s <= candidate { + panic("s <= candidate") + } + a := src[base:s] + b := src[base-repeat : base-repeat+(s-base)] + if !bytes.Equal(a, b) { + panic("mismatch") + } + } + if debug { + fmt.Println("emitted src copy, length", s-base, "offset:", repeat, "s:", s) + } + nextEmit = s + if s >= sLimit { + goto emitRemainder + } + + if d > dstLimit { + // Do we have space for more, if not bail. + return 0 + } + // Check for an immediate match, otherwise start search at s+1 + x := load64(src, s-2) + m2Hash := hash6(x, tableBits) + currHash := hash6(x>>16, tableBits) + candidate = int(table[currHash]) + table[m2Hash] = uint32(s - 2) + table[currHash] = uint32(s) + if debug && s == candidate { + panic("s == candidate") + } + if uint32(x>>16) != load32(src, candidate) { + cv = load64(src, s+1) + s++ + break + } + } + } + +emitRemainder: + if nextEmit < len(src) { + // Bail if we exceed the maximum size. + if d+len(src)-nextEmit > dstLimit { + return 0 + } + d += emitLiteral(dst[d:], src[nextEmit:]) + if debug && nextEmit != s { + fmt.Println("emitted ", len(src)-nextEmit, "literals") + } + } + return d +} diff --git a/vendor/github.com/klauspost/compress/s2/encode_amd64.go b/vendor/github.com/klauspost/compress/s2/encode_amd64.go new file mode 100644 index 000000000..4f45206a4 --- /dev/null +++ b/vendor/github.com/klauspost/compress/s2/encode_amd64.go @@ -0,0 +1,162 @@ +//go:build !appengine && !noasm && gc +// +build !appengine,!noasm,gc + +package s2 + +import "github.com/klauspost/compress/internal/race" + +const hasAmd64Asm = true + +// encodeBlock encodes a non-empty src to a guaranteed-large-enough dst. It +// assumes that the varint-encoded length of the decompressed bytes has already +// been written. +// +// It also assumes that: +// +// len(dst) >= MaxEncodedLen(len(src)) && +// minNonLiteralBlockSize <= len(src) && len(src) <= maxBlockSize +func encodeBlock(dst, src []byte) (d int) { + race.ReadSlice(src) + race.WriteSlice(dst) + + const ( + // Use 12 bit table when less than... + limit12B = 16 << 10 + // Use 10 bit table when less than... + limit10B = 4 << 10 + // Use 8 bit table when less than... + limit8B = 512 + ) + + if len(src) >= 4<<20 { + return encodeBlockAsm(dst, src) + } + if len(src) >= limit12B { + return encodeBlockAsm4MB(dst, src) + } + if len(src) >= limit10B { + return encodeBlockAsm12B(dst, src) + } + if len(src) >= limit8B { + return encodeBlockAsm10B(dst, src) + } + if len(src) < minNonLiteralBlockSize { + return 0 + } + return encodeBlockAsm8B(dst, src) +} + +// encodeBlockBetter encodes a non-empty src to a guaranteed-large-enough dst. It +// assumes that the varint-encoded length of the decompressed bytes has already +// been written. +// +// It also assumes that: +// +// len(dst) >= MaxEncodedLen(len(src)) && +// minNonLiteralBlockSize <= len(src) && len(src) <= maxBlockSize +func encodeBlockBetter(dst, src []byte) (d int) { + race.ReadSlice(src) + race.WriteSlice(dst) + + const ( + // Use 12 bit table when less than... + limit12B = 16 << 10 + // Use 10 bit table when less than... + limit10B = 4 << 10 + // Use 8 bit table when less than... + limit8B = 512 + ) + + if len(src) > 4<<20 { + return encodeBetterBlockAsm(dst, src) + } + if len(src) >= limit12B { + return encodeBetterBlockAsm4MB(dst, src) + } + if len(src) >= limit10B { + return encodeBetterBlockAsm12B(dst, src) + } + if len(src) >= limit8B { + return encodeBetterBlockAsm10B(dst, src) + } + if len(src) < minNonLiteralBlockSize { + return 0 + } + return encodeBetterBlockAsm8B(dst, src) +} + +// encodeBlockSnappy encodes a non-empty src to a guaranteed-large-enough dst. It +// assumes that the varint-encoded length of the decompressed bytes has already +// been written. +// +// It also assumes that: +// +// len(dst) >= MaxEncodedLen(len(src)) && +// minNonLiteralBlockSize <= len(src) && len(src) <= maxBlockSize +func encodeBlockSnappy(dst, src []byte) (d int) { + race.ReadSlice(src) + race.WriteSlice(dst) + + const ( + // Use 12 bit table when less than... + limit12B = 16 << 10 + // Use 10 bit table when less than... + limit10B = 4 << 10 + // Use 8 bit table when less than... + limit8B = 512 + ) + if len(src) >= 64<<10 { + return encodeSnappyBlockAsm(dst, src) + } + if len(src) >= limit12B { + return encodeSnappyBlockAsm64K(dst, src) + } + if len(src) >= limit10B { + return encodeSnappyBlockAsm12B(dst, src) + } + if len(src) >= limit8B { + return encodeSnappyBlockAsm10B(dst, src) + } + if len(src) < minNonLiteralBlockSize { + return 0 + } + return encodeSnappyBlockAsm8B(dst, src) +} + +// encodeBlockSnappy encodes a non-empty src to a guaranteed-large-enough dst. It +// assumes that the varint-encoded length of the decompressed bytes has already +// been written. +// +// It also assumes that: +// +// len(dst) >= MaxEncodedLen(len(src)) && +// minNonLiteralBlockSize <= len(src) && len(src) <= maxBlockSize +func encodeBlockBetterSnappy(dst, src []byte) (d int) { + race.ReadSlice(src) + race.WriteSlice(dst) + + const ( + // Use 12 bit table when less than... + limit12B = 16 << 10 + // Use 10 bit table when less than... + limit10B = 4 << 10 + // Use 8 bit table when less than... + limit8B = 512 + ) + if len(src) >= 64<<10 { + return encodeSnappyBetterBlockAsm(dst, src) + } + if len(src) >= limit12B { + return encodeSnappyBetterBlockAsm64K(dst, src) + } + if len(src) >= limit10B { + return encodeSnappyBetterBlockAsm12B(dst, src) + } + if len(src) >= limit8B { + return encodeSnappyBetterBlockAsm10B(dst, src) + } + if len(src) < minNonLiteralBlockSize { + return 0 + } + return encodeSnappyBetterBlockAsm8B(dst, src) +} diff --git a/vendor/github.com/klauspost/compress/s2/encode_best.go b/vendor/github.com/klauspost/compress/s2/encode_best.go new file mode 100644 index 000000000..47bac7423 --- /dev/null +++ b/vendor/github.com/klauspost/compress/s2/encode_best.go @@ -0,0 +1,796 @@ +// Copyright 2016 The Snappy-Go Authors. All rights reserved. +// Copyright (c) 2019 Klaus Post. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package s2 + +import ( + "fmt" + "math" + "math/bits" +) + +// encodeBlockBest encodes a non-empty src to a guaranteed-large-enough dst. It +// assumes that the varint-encoded length of the decompressed bytes has already +// been written. +// +// It also assumes that: +// +// len(dst) >= MaxEncodedLen(len(src)) && +// minNonLiteralBlockSize <= len(src) && len(src) <= maxBlockSize +func encodeBlockBest(dst, src []byte, dict *Dict) (d int) { + // Initialize the hash tables. + const ( + // Long hash matches. + lTableBits = 19 + maxLTableSize = 1 << lTableBits + + // Short hash matches. + sTableBits = 16 + maxSTableSize = 1 << sTableBits + + inputMargin = 8 + 2 + + debug = false + ) + + // sLimit is when to stop looking for offset/length copies. The inputMargin + // lets us use a fast path for emitLiteral in the main loop, while we are + // looking for copies. + sLimit := len(src) - inputMargin + if len(src) < minNonLiteralBlockSize { + return 0 + } + sLimitDict := len(src) - inputMargin + if sLimitDict > MaxDictSrcOffset-inputMargin { + sLimitDict = MaxDictSrcOffset - inputMargin + } + + var lTable [maxLTableSize]uint64 + var sTable [maxSTableSize]uint64 + + // Bail if we can't compress to at least this. + dstLimit := len(src) - 5 + + // nextEmit is where in src the next emitLiteral should start from. + nextEmit := 0 + + // The encoded form must start with a literal, as there are no previous + // bytes to copy, so we start looking for hash matches at s == 1. + s := 1 + repeat := 1 + if dict != nil { + dict.initBest() + s = 0 + repeat = len(dict.dict) - dict.repeat + } + cv := load64(src, s) + + // We search for a repeat at -1, but don't output repeats when nextEmit == 0 + const lowbitMask = 0xffffffff + getCur := func(x uint64) int { + return int(x & lowbitMask) + } + getPrev := func(x uint64) int { + return int(x >> 32) + } + const maxSkip = 64 + + for { + type match struct { + offset int + s int + length int + score int + rep, dict bool + } + var best match + for { + // Next src position to check + nextS := (s-nextEmit)>>8 + 1 + if nextS > maxSkip { + nextS = s + maxSkip + } else { + nextS += s + } + if nextS > sLimit { + goto emitRemainder + } + if dict != nil && s >= MaxDictSrcOffset { + dict = nil + if repeat > s { + repeat = math.MinInt32 + } + } + hashL := hash8(cv, lTableBits) + hashS := hash4(cv, sTableBits) + candidateL := lTable[hashL] + candidateS := sTable[hashS] + + score := func(m match) int { + // Matches that are longer forward are penalized since we must emit it as a literal. + score := m.length - m.s + if nextEmit == m.s { + // If we do not have to emit literals, we save 1 byte + score++ + } + offset := m.s - m.offset + if m.rep { + return score - emitRepeatSize(offset, m.length) + } + return score - emitCopySize(offset, m.length) + } + + matchAt := func(offset, s int, first uint32, rep bool) match { + if best.length != 0 && best.s-best.offset == s-offset { + // Don't retest if we have the same offset. + return match{offset: offset, s: s} + } + if load32(src, offset) != first { + return match{offset: offset, s: s} + } + m := match{offset: offset, s: s, length: 4 + offset, rep: rep} + s += 4 + for s < len(src) { + if len(src)-s < 8 { + if src[s] == src[m.length] { + m.length++ + s++ + continue + } + break + } + if diff := load64(src, s) ^ load64(src, m.length); diff != 0 { + m.length += bits.TrailingZeros64(diff) >> 3 + break + } + s += 8 + m.length += 8 + } + m.length -= offset + m.score = score(m) + if m.score <= -m.s { + // Eliminate if no savings, we might find a better one. + m.length = 0 + } + return m + } + matchDict := func(candidate, s int, first uint32, rep bool) match { + if s >= MaxDictSrcOffset { + return match{offset: candidate, s: s} + } + // Calculate offset as if in continuous array with s + offset := -len(dict.dict) + candidate + if best.length != 0 && best.s-best.offset == s-offset && !rep { + // Don't retest if we have the same offset. + return match{offset: offset, s: s} + } + + if load32(dict.dict, candidate) != first { + return match{offset: offset, s: s} + } + m := match{offset: offset, s: s, length: 4 + candidate, rep: rep, dict: true} + s += 4 + if !rep { + for s < sLimitDict && m.length < len(dict.dict) { + if len(src)-s < 8 || len(dict.dict)-m.length < 8 { + if src[s] == dict.dict[m.length] { + m.length++ + s++ + continue + } + break + } + if diff := load64(src, s) ^ load64(dict.dict, m.length); diff != 0 { + m.length += bits.TrailingZeros64(diff) >> 3 + break + } + s += 8 + m.length += 8 + } + } else { + for s < len(src) && m.length < len(dict.dict) { + if len(src)-s < 8 || len(dict.dict)-m.length < 8 { + if src[s] == dict.dict[m.length] { + m.length++ + s++ + continue + } + break + } + if diff := load64(src, s) ^ load64(dict.dict, m.length); diff != 0 { + m.length += bits.TrailingZeros64(diff) >> 3 + break + } + s += 8 + m.length += 8 + } + } + m.length -= candidate + m.score = score(m) + if m.score <= -m.s { + // Eliminate if no savings, we might find a better one. + m.length = 0 + } + return m + } + + bestOf := func(a, b match) match { + if b.length == 0 { + return a + } + if a.length == 0 { + return b + } + as := a.score + b.s + bs := b.score + a.s + if as >= bs { + return a + } + return b + } + + if s > 0 { + best = bestOf(matchAt(getCur(candidateL), s, uint32(cv), false), matchAt(getPrev(candidateL), s, uint32(cv), false)) + best = bestOf(best, matchAt(getCur(candidateS), s, uint32(cv), false)) + best = bestOf(best, matchAt(getPrev(candidateS), s, uint32(cv), false)) + } + if dict != nil { + candidateL := dict.bestTableLong[hashL] + candidateS := dict.bestTableShort[hashS] + best = bestOf(best, matchDict(int(candidateL&0xffff), s, uint32(cv), false)) + best = bestOf(best, matchDict(int(candidateL>>16), s, uint32(cv), false)) + best = bestOf(best, matchDict(int(candidateS&0xffff), s, uint32(cv), false)) + best = bestOf(best, matchDict(int(candidateS>>16), s, uint32(cv), false)) + } + { + if (dict == nil || repeat <= s) && repeat > 0 { + best = bestOf(best, matchAt(s-repeat+1, s+1, uint32(cv>>8), true)) + } else if s-repeat < -4 && dict != nil { + candidate := len(dict.dict) - (repeat - s) + best = bestOf(best, matchDict(candidate, s, uint32(cv), true)) + candidate++ + best = bestOf(best, matchDict(candidate, s+1, uint32(cv>>8), true)) + } + + if best.length > 0 { + hashS := hash4(cv>>8, sTableBits) + // s+1 + nextShort := sTable[hashS] + s := s + 1 + cv := load64(src, s) + hashL := hash8(cv, lTableBits) + nextLong := lTable[hashL] + best = bestOf(best, matchAt(getCur(nextShort), s, uint32(cv), false)) + best = bestOf(best, matchAt(getPrev(nextShort), s, uint32(cv), false)) + best = bestOf(best, matchAt(getCur(nextLong), s, uint32(cv), false)) + best = bestOf(best, matchAt(getPrev(nextLong), s, uint32(cv), false)) + + // Dict at + 1 + if dict != nil { + candidateL := dict.bestTableLong[hashL] + candidateS := dict.bestTableShort[hashS] + + best = bestOf(best, matchDict(int(candidateL&0xffff), s, uint32(cv), false)) + best = bestOf(best, matchDict(int(candidateS&0xffff), s, uint32(cv), false)) + } + + // s+2 + if true { + hashS := hash4(cv>>8, sTableBits) + + nextShort = sTable[hashS] + s++ + cv = load64(src, s) + hashL := hash8(cv, lTableBits) + nextLong = lTable[hashL] + + if (dict == nil || repeat <= s) && repeat > 0 { + // Repeat at + 2 + best = bestOf(best, matchAt(s-repeat, s, uint32(cv), true)) + } else if repeat-s > 4 && dict != nil { + candidate := len(dict.dict) - (repeat - s) + best = bestOf(best, matchDict(candidate, s, uint32(cv), true)) + } + best = bestOf(best, matchAt(getCur(nextShort), s, uint32(cv), false)) + best = bestOf(best, matchAt(getPrev(nextShort), s, uint32(cv), false)) + best = bestOf(best, matchAt(getCur(nextLong), s, uint32(cv), false)) + best = bestOf(best, matchAt(getPrev(nextLong), s, uint32(cv), false)) + + // Dict at +2 + // Very small gain + if dict != nil { + candidateL := dict.bestTableLong[hashL] + candidateS := dict.bestTableShort[hashS] + + best = bestOf(best, matchDict(int(candidateL&0xffff), s, uint32(cv), false)) + best = bestOf(best, matchDict(int(candidateS&0xffff), s, uint32(cv), false)) + } + } + // Search for a match at best match end, see if that is better. + // Allow some bytes at the beginning to mismatch. + // Sweet spot is around 1-2 bytes, but depends on input. + // The skipped bytes are tested in Extend backwards, + // and still picked up as part of the match if they do. + const skipBeginning = 2 + const skipEnd = 1 + if sAt := best.s + best.length - skipEnd; sAt < sLimit { + + sBack := best.s + skipBeginning - skipEnd + backL := best.length - skipBeginning + // Load initial values + cv = load64(src, sBack) + + // Grab candidates... + next := lTable[hash8(load64(src, sAt), lTableBits)] + + if checkAt := getCur(next) - backL; checkAt > 0 { + best = bestOf(best, matchAt(checkAt, sBack, uint32(cv), false)) + } + if checkAt := getPrev(next) - backL; checkAt > 0 { + best = bestOf(best, matchAt(checkAt, sBack, uint32(cv), false)) + } + // Disabled: Extremely small gain + if false { + next = sTable[hash4(load64(src, sAt), sTableBits)] + if checkAt := getCur(next) - backL; checkAt > 0 { + best = bestOf(best, matchAt(checkAt, sBack, uint32(cv), false)) + } + if checkAt := getPrev(next) - backL; checkAt > 0 { + best = bestOf(best, matchAt(checkAt, sBack, uint32(cv), false)) + } + } + } + } + } + + // Update table + lTable[hashL] = uint64(s) | candidateL<<32 + sTable[hashS] = uint64(s) | candidateS<<32 + + if best.length > 0 { + break + } + + cv = load64(src, nextS) + s = nextS + } + + // Extend backwards, not needed for repeats... + s = best.s + if !best.rep && !best.dict { + for best.offset > 0 && s > nextEmit && src[best.offset-1] == src[s-1] { + best.offset-- + best.length++ + s-- + } + } + if false && best.offset >= s { + panic(fmt.Errorf("t %d >= s %d", best.offset, s)) + } + // Bail if we exceed the maximum size. + if d+(s-nextEmit) > dstLimit { + return 0 + } + + base := s + offset := s - best.offset + s += best.length + + if offset > 65535 && s-base <= 5 && !best.rep { + // Bail if the match is equal or worse to the encoding. + s = best.s + 1 + if s >= sLimit { + goto emitRemainder + } + cv = load64(src, s) + continue + } + if debug && nextEmit != base { + fmt.Println("EMIT", base-nextEmit, "literals. base-after:", base) + } + d += emitLiteral(dst[d:], src[nextEmit:base]) + if best.rep { + if nextEmit > 0 || best.dict { + if debug { + fmt.Println("REPEAT, length", best.length, "offset:", offset, "s-after:", s, "dict:", best.dict, "best:", best) + } + // same as `add := emitCopy(dst[d:], repeat, s-base)` but skips storing offset. + d += emitRepeat(dst[d:], offset, best.length) + } else { + // First match without dict cannot be a repeat. + if debug { + fmt.Println("COPY, length", best.length, "offset:", offset, "s-after:", s, "dict:", best.dict, "best:", best) + } + d += emitCopy(dst[d:], offset, best.length) + } + } else { + if debug { + fmt.Println("COPY, length", best.length, "offset:", offset, "s-after:", s, "dict:", best.dict, "best:", best) + } + d += emitCopy(dst[d:], offset, best.length) + } + repeat = offset + + nextEmit = s + if s >= sLimit { + goto emitRemainder + } + + if d > dstLimit { + // Do we have space for more, if not bail. + return 0 + } + // Fill tables... + for i := best.s + 1; i < s; i++ { + cv0 := load64(src, i) + long0 := hash8(cv0, lTableBits) + short0 := hash4(cv0, sTableBits) + lTable[long0] = uint64(i) | lTable[long0]<<32 + sTable[short0] = uint64(i) | sTable[short0]<<32 + } + cv = load64(src, s) + } + +emitRemainder: + if nextEmit < len(src) { + // Bail if we exceed the maximum size. + if d+len(src)-nextEmit > dstLimit { + return 0 + } + if debug && nextEmit != s { + fmt.Println("emitted ", len(src)-nextEmit, "literals") + } + d += emitLiteral(dst[d:], src[nextEmit:]) + } + return d +} + +// encodeBlockBestSnappy encodes a non-empty src to a guaranteed-large-enough dst. It +// assumes that the varint-encoded length of the decompressed bytes has already +// been written. +// +// It also assumes that: +// +// len(dst) >= MaxEncodedLen(len(src)) && +// minNonLiteralBlockSize <= len(src) && len(src) <= maxBlockSize +func encodeBlockBestSnappy(dst, src []byte) (d int) { + // Initialize the hash tables. + const ( + // Long hash matches. + lTableBits = 19 + maxLTableSize = 1 << lTableBits + + // Short hash matches. + sTableBits = 16 + maxSTableSize = 1 << sTableBits + + inputMargin = 8 + 2 + ) + + // sLimit is when to stop looking for offset/length copies. The inputMargin + // lets us use a fast path for emitLiteral in the main loop, while we are + // looking for copies. + sLimit := len(src) - inputMargin + if len(src) < minNonLiteralBlockSize { + return 0 + } + + var lTable [maxLTableSize]uint64 + var sTable [maxSTableSize]uint64 + + // Bail if we can't compress to at least this. + dstLimit := len(src) - 5 + + // nextEmit is where in src the next emitLiteral should start from. + nextEmit := 0 + + // The encoded form must start with a literal, as there are no previous + // bytes to copy, so we start looking for hash matches at s == 1. + s := 1 + cv := load64(src, s) + + // We search for a repeat at -1, but don't output repeats when nextEmit == 0 + repeat := 1 + const lowbitMask = 0xffffffff + getCur := func(x uint64) int { + return int(x & lowbitMask) + } + getPrev := func(x uint64) int { + return int(x >> 32) + } + const maxSkip = 64 + + for { + type match struct { + offset int + s int + length int + score int + } + var best match + for { + // Next src position to check + nextS := (s-nextEmit)>>8 + 1 + if nextS > maxSkip { + nextS = s + maxSkip + } else { + nextS += s + } + if nextS > sLimit { + goto emitRemainder + } + hashL := hash8(cv, lTableBits) + hashS := hash4(cv, sTableBits) + candidateL := lTable[hashL] + candidateS := sTable[hashS] + + score := func(m match) int { + // Matches that are longer forward are penalized since we must emit it as a literal. + score := m.length - m.s + if nextEmit == m.s { + // If we do not have to emit literals, we save 1 byte + score++ + } + offset := m.s - m.offset + + return score - emitCopyNoRepeatSize(offset, m.length) + } + + matchAt := func(offset, s int, first uint32) match { + if best.length != 0 && best.s-best.offset == s-offset { + // Don't retest if we have the same offset. + return match{offset: offset, s: s} + } + if load32(src, offset) != first { + return match{offset: offset, s: s} + } + m := match{offset: offset, s: s, length: 4 + offset} + s += 4 + for s <= sLimit { + if diff := load64(src, s) ^ load64(src, m.length); diff != 0 { + m.length += bits.TrailingZeros64(diff) >> 3 + break + } + s += 8 + m.length += 8 + } + m.length -= offset + m.score = score(m) + if m.score <= -m.s { + // Eliminate if no savings, we might find a better one. + m.length = 0 + } + return m + } + + bestOf := func(a, b match) match { + if b.length == 0 { + return a + } + if a.length == 0 { + return b + } + as := a.score + b.s + bs := b.score + a.s + if as >= bs { + return a + } + return b + } + + best = bestOf(matchAt(getCur(candidateL), s, uint32(cv)), matchAt(getPrev(candidateL), s, uint32(cv))) + best = bestOf(best, matchAt(getCur(candidateS), s, uint32(cv))) + best = bestOf(best, matchAt(getPrev(candidateS), s, uint32(cv))) + + { + best = bestOf(best, matchAt(s-repeat+1, s+1, uint32(cv>>8))) + if best.length > 0 { + // s+1 + nextShort := sTable[hash4(cv>>8, sTableBits)] + s := s + 1 + cv := load64(src, s) + nextLong := lTable[hash8(cv, lTableBits)] + best = bestOf(best, matchAt(getCur(nextShort), s, uint32(cv))) + best = bestOf(best, matchAt(getPrev(nextShort), s, uint32(cv))) + best = bestOf(best, matchAt(getCur(nextLong), s, uint32(cv))) + best = bestOf(best, matchAt(getPrev(nextLong), s, uint32(cv))) + // Repeat at + 2 + best = bestOf(best, matchAt(s-repeat+1, s+1, uint32(cv>>8))) + + // s+2 + if true { + nextShort = sTable[hash4(cv>>8, sTableBits)] + s++ + cv = load64(src, s) + nextLong = lTable[hash8(cv, lTableBits)] + best = bestOf(best, matchAt(getCur(nextShort), s, uint32(cv))) + best = bestOf(best, matchAt(getPrev(nextShort), s, uint32(cv))) + best = bestOf(best, matchAt(getCur(nextLong), s, uint32(cv))) + best = bestOf(best, matchAt(getPrev(nextLong), s, uint32(cv))) + } + // Search for a match at best match end, see if that is better. + if sAt := best.s + best.length; sAt < sLimit { + sBack := best.s + backL := best.length + // Load initial values + cv = load64(src, sBack) + // Search for mismatch + next := lTable[hash8(load64(src, sAt), lTableBits)] + //next := sTable[hash4(load64(src, sAt), sTableBits)] + + if checkAt := getCur(next) - backL; checkAt > 0 { + best = bestOf(best, matchAt(checkAt, sBack, uint32(cv))) + } + if checkAt := getPrev(next) - backL; checkAt > 0 { + best = bestOf(best, matchAt(checkAt, sBack, uint32(cv))) + } + } + } + } + + // Update table + lTable[hashL] = uint64(s) | candidateL<<32 + sTable[hashS] = uint64(s) | candidateS<<32 + + if best.length > 0 { + break + } + + cv = load64(src, nextS) + s = nextS + } + + // Extend backwards, not needed for repeats... + s = best.s + if true { + for best.offset > 0 && s > nextEmit && src[best.offset-1] == src[s-1] { + best.offset-- + best.length++ + s-- + } + } + if false && best.offset >= s { + panic(fmt.Errorf("t %d >= s %d", best.offset, s)) + } + // Bail if we exceed the maximum size. + if d+(s-nextEmit) > dstLimit { + return 0 + } + + base := s + offset := s - best.offset + + s += best.length + + if offset > 65535 && s-base <= 5 { + // Bail if the match is equal or worse to the encoding. + s = best.s + 1 + if s >= sLimit { + goto emitRemainder + } + cv = load64(src, s) + continue + } + d += emitLiteral(dst[d:], src[nextEmit:base]) + d += emitCopyNoRepeat(dst[d:], offset, best.length) + repeat = offset + + nextEmit = s + if s >= sLimit { + goto emitRemainder + } + + if d > dstLimit { + // Do we have space for more, if not bail. + return 0 + } + // Fill tables... + for i := best.s + 1; i < s; i++ { + cv0 := load64(src, i) + long0 := hash8(cv0, lTableBits) + short0 := hash4(cv0, sTableBits) + lTable[long0] = uint64(i) | lTable[long0]<<32 + sTable[short0] = uint64(i) | sTable[short0]<<32 + } + cv = load64(src, s) + } + +emitRemainder: + if nextEmit < len(src) { + // Bail if we exceed the maximum size. + if d+len(src)-nextEmit > dstLimit { + return 0 + } + d += emitLiteral(dst[d:], src[nextEmit:]) + } + return d +} + +// emitCopySize returns the size to encode the offset+length +// +// It assumes that: +// +// 1 <= offset && offset <= math.MaxUint32 +// 4 <= length && length <= 1 << 24 +func emitCopySize(offset, length int) int { + if offset >= 65536 { + i := 0 + if length > 64 { + length -= 64 + if length >= 4 { + // Emit remaining as repeats + return 5 + emitRepeatSize(offset, length) + } + i = 5 + } + if length == 0 { + return i + } + return i + 5 + } + + // Offset no more than 2 bytes. + if length > 64 { + if offset < 2048 { + // Emit 8 bytes, then rest as repeats... + return 2 + emitRepeatSize(offset, length-8) + } + // Emit remaining as repeats, at least 4 bytes remain. + return 3 + emitRepeatSize(offset, length-60) + } + if length >= 12 || offset >= 2048 { + return 3 + } + // Emit the remaining copy, encoded as 2 bytes. + return 2 +} + +// emitCopyNoRepeatSize returns the size to encode the offset+length +// +// It assumes that: +// +// 1 <= offset && offset <= math.MaxUint32 +// 4 <= length && length <= 1 << 24 +func emitCopyNoRepeatSize(offset, length int) int { + if offset >= 65536 { + return 5 + 5*(length/64) + } + + // Offset no more than 2 bytes. + if length > 64 { + // Emit remaining as repeats, at least 4 bytes remain. + return 3 + 3*(length/60) + } + if length >= 12 || offset >= 2048 { + return 3 + } + // Emit the remaining copy, encoded as 2 bytes. + return 2 +} + +// emitRepeatSize returns the number of bytes required to encode a repeat. +// Length must be at least 4 and < 1<<24 +func emitRepeatSize(offset, length int) int { + // Repeat offset, make length cheaper + if length <= 4+4 || (length < 8+4 && offset < 2048) { + return 2 + } + if length < (1<<8)+4+4 { + return 3 + } + if length < (1<<16)+(1<<8)+4 { + return 4 + } + const maxRepeat = (1 << 24) - 1 + length -= (1 << 16) - 4 + left := 0 + if length > maxRepeat { + left = length - maxRepeat + 4 + } + if left > 0 { + return 5 + emitRepeatSize(offset, left) + } + return 5 +} diff --git a/vendor/github.com/klauspost/compress/s2/encode_better.go b/vendor/github.com/klauspost/compress/s2/encode_better.go new file mode 100644 index 000000000..544cb1e17 --- /dev/null +++ b/vendor/github.com/klauspost/compress/s2/encode_better.go @@ -0,0 +1,1106 @@ +// Copyright 2016 The Snappy-Go Authors. All rights reserved. +// Copyright (c) 2019 Klaus Post. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package s2 + +import ( + "bytes" + "fmt" + "math/bits" +) + +// hash4 returns the hash of the lowest 4 bytes of u to fit in a hash table with h bits. +// Preferably h should be a constant and should always be <32. +func hash4(u uint64, h uint8) uint32 { + const prime4bytes = 2654435761 + return (uint32(u) * prime4bytes) >> ((32 - h) & 31) +} + +// hash5 returns the hash of the lowest 5 bytes of u to fit in a hash table with h bits. +// Preferably h should be a constant and should always be <64. +func hash5(u uint64, h uint8) uint32 { + const prime5bytes = 889523592379 + return uint32(((u << (64 - 40)) * prime5bytes) >> ((64 - h) & 63)) +} + +// hash7 returns the hash of the lowest 7 bytes of u to fit in a hash table with h bits. +// Preferably h should be a constant and should always be <64. +func hash7(u uint64, h uint8) uint32 { + const prime7bytes = 58295818150454627 + return uint32(((u << (64 - 56)) * prime7bytes) >> ((64 - h) & 63)) +} + +// hash8 returns the hash of u to fit in a hash table with h bits. +// Preferably h should be a constant and should always be <64. +func hash8(u uint64, h uint8) uint32 { + const prime8bytes = 0xcf1bbcdcb7a56463 + return uint32((u * prime8bytes) >> ((64 - h) & 63)) +} + +// encodeBlockBetter encodes a non-empty src to a guaranteed-large-enough dst. It +// assumes that the varint-encoded length of the decompressed bytes has already +// been written. +// +// It also assumes that: +// +// len(dst) >= MaxEncodedLen(len(src)) && +// minNonLiteralBlockSize <= len(src) && len(src) <= maxBlockSize +func encodeBlockBetterGo(dst, src []byte) (d int) { + // sLimit is when to stop looking for offset/length copies. The inputMargin + // lets us use a fast path for emitLiteral in the main loop, while we are + // looking for copies. + sLimit := len(src) - inputMargin + if len(src) < minNonLiteralBlockSize { + return 0 + } + + // Initialize the hash tables. + const ( + // Long hash matches. + lTableBits = 17 + maxLTableSize = 1 << lTableBits + + // Short hash matches. + sTableBits = 14 + maxSTableSize = 1 << sTableBits + ) + + var lTable [maxLTableSize]uint32 + var sTable [maxSTableSize]uint32 + + // Bail if we can't compress to at least this. + dstLimit := len(src) - len(src)>>5 - 6 + + // nextEmit is where in src the next emitLiteral should start from. + nextEmit := 0 + + // The encoded form must start with a literal, as there are no previous + // bytes to copy, so we start looking for hash matches at s == 1. + s := 1 + cv := load64(src, s) + + // We initialize repeat to 0, so we never match on first attempt + repeat := 0 + + for { + candidateL := 0 + nextS := 0 + for { + // Next src position to check + nextS = s + (s-nextEmit)>>7 + 1 + if nextS > sLimit { + goto emitRemainder + } + hashL := hash7(cv, lTableBits) + hashS := hash4(cv, sTableBits) + candidateL = int(lTable[hashL]) + candidateS := int(sTable[hashS]) + lTable[hashL] = uint32(s) + sTable[hashS] = uint32(s) + + valLong := load64(src, candidateL) + valShort := load64(src, candidateS) + + // If long matches at least 8 bytes, use that. + if cv == valLong { + break + } + if cv == valShort { + candidateL = candidateS + break + } + + // Check repeat at offset checkRep. + const checkRep = 1 + // Minimum length of a repeat. Tested with various values. + // While 4-5 offers improvements in some, 6 reduces + // regressions significantly. + const wantRepeatBytes = 6 + const repeatMask = ((1 << (wantRepeatBytes * 8)) - 1) << (8 * checkRep) + if false && repeat > 0 && cv&repeatMask == load64(src, s-repeat)&repeatMask { + base := s + checkRep + // Extend back + for i := base - repeat; base > nextEmit && i > 0 && src[i-1] == src[base-1]; { + i-- + base-- + } + d += emitLiteral(dst[d:], src[nextEmit:base]) + + // Extend forward + candidate := s - repeat + wantRepeatBytes + checkRep + s += wantRepeatBytes + checkRep + for s < len(src) { + if len(src)-s < 8 { + if src[s] == src[candidate] { + s++ + candidate++ + continue + } + break + } + if diff := load64(src, s) ^ load64(src, candidate); diff != 0 { + s += bits.TrailingZeros64(diff) >> 3 + break + } + s += 8 + candidate += 8 + } + // same as `add := emitCopy(dst[d:], repeat, s-base)` but skips storing offset. + d += emitRepeat(dst[d:], repeat, s-base) + nextEmit = s + if s >= sLimit { + goto emitRemainder + } + // Index in-between + index0 := base + 1 + index1 := s - 2 + + for index0 < index1 { + cv0 := load64(src, index0) + cv1 := load64(src, index1) + lTable[hash7(cv0, lTableBits)] = uint32(index0) + sTable[hash4(cv0>>8, sTableBits)] = uint32(index0 + 1) + + lTable[hash7(cv1, lTableBits)] = uint32(index1) + sTable[hash4(cv1>>8, sTableBits)] = uint32(index1 + 1) + index0 += 2 + index1 -= 2 + } + + cv = load64(src, s) + continue + } + + // Long likely matches 7, so take that. + if uint32(cv) == uint32(valLong) { + break + } + + // Check our short candidate + if uint32(cv) == uint32(valShort) { + // Try a long candidate at s+1 + hashL = hash7(cv>>8, lTableBits) + candidateL = int(lTable[hashL]) + lTable[hashL] = uint32(s + 1) + if uint32(cv>>8) == load32(src, candidateL) { + s++ + break + } + // Use our short candidate. + candidateL = candidateS + break + } + + cv = load64(src, nextS) + s = nextS + } + + // Extend backwards + for candidateL > 0 && s > nextEmit && src[candidateL-1] == src[s-1] { + candidateL-- + s-- + } + + // Bail if we exceed the maximum size. + if d+(s-nextEmit) > dstLimit { + return 0 + } + + base := s + offset := base - candidateL + + // Extend the 4-byte match as long as possible. + s += 4 + candidateL += 4 + for s < len(src) { + if len(src)-s < 8 { + if src[s] == src[candidateL] { + s++ + candidateL++ + continue + } + break + } + if diff := load64(src, s) ^ load64(src, candidateL); diff != 0 { + s += bits.TrailingZeros64(diff) >> 3 + break + } + s += 8 + candidateL += 8 + } + + if offset > 65535 && s-base <= 5 && repeat != offset { + // Bail if the match is equal or worse to the encoding. + s = nextS + 1 + if s >= sLimit { + goto emitRemainder + } + cv = load64(src, s) + continue + } + + d += emitLiteral(dst[d:], src[nextEmit:base]) + if repeat == offset { + d += emitRepeat(dst[d:], offset, s-base) + } else { + d += emitCopy(dst[d:], offset, s-base) + repeat = offset + } + + nextEmit = s + if s >= sLimit { + goto emitRemainder + } + + if d > dstLimit { + // Do we have space for more, if not bail. + return 0 + } + + // Index short & long + index0 := base + 1 + index1 := s - 2 + + cv0 := load64(src, index0) + cv1 := load64(src, index1) + lTable[hash7(cv0, lTableBits)] = uint32(index0) + sTable[hash4(cv0>>8, sTableBits)] = uint32(index0 + 1) + + // lTable could be postponed, but very minor difference. + lTable[hash7(cv1, lTableBits)] = uint32(index1) + sTable[hash4(cv1>>8, sTableBits)] = uint32(index1 + 1) + index0 += 1 + index1 -= 1 + cv = load64(src, s) + + // Index large values sparsely in between. + // We do two starting from different offsets for speed. + index2 := (index0 + index1 + 1) >> 1 + for index2 < index1 { + lTable[hash7(load64(src, index0), lTableBits)] = uint32(index0) + lTable[hash7(load64(src, index2), lTableBits)] = uint32(index2) + index0 += 2 + index2 += 2 + } + } + +emitRemainder: + if nextEmit < len(src) { + // Bail if we exceed the maximum size. + if d+len(src)-nextEmit > dstLimit { + return 0 + } + d += emitLiteral(dst[d:], src[nextEmit:]) + } + return d +} + +// encodeBlockBetterSnappyGo encodes a non-empty src to a guaranteed-large-enough dst. It +// assumes that the varint-encoded length of the decompressed bytes has already +// been written. +// +// It also assumes that: +// +// len(dst) >= MaxEncodedLen(len(src)) && +// minNonLiteralBlockSize <= len(src) && len(src) <= maxBlockSize +func encodeBlockBetterSnappyGo(dst, src []byte) (d int) { + // sLimit is when to stop looking for offset/length copies. The inputMargin + // lets us use a fast path for emitLiteral in the main loop, while we are + // looking for copies. + sLimit := len(src) - inputMargin + if len(src) < minNonLiteralBlockSize { + return 0 + } + + // Initialize the hash tables. + const ( + // Long hash matches. + lTableBits = 16 + maxLTableSize = 1 << lTableBits + + // Short hash matches. + sTableBits = 14 + maxSTableSize = 1 << sTableBits + ) + + var lTable [maxLTableSize]uint32 + var sTable [maxSTableSize]uint32 + + // Bail if we can't compress to at least this. + dstLimit := len(src) - len(src)>>5 - 6 + + // nextEmit is where in src the next emitLiteral should start from. + nextEmit := 0 + + // The encoded form must start with a literal, as there are no previous + // bytes to copy, so we start looking for hash matches at s == 1. + s := 1 + cv := load64(src, s) + + // We initialize repeat to 0, so we never match on first attempt + repeat := 0 + const maxSkip = 100 + + for { + candidateL := 0 + nextS := 0 + for { + // Next src position to check + nextS = (s-nextEmit)>>7 + 1 + if nextS > maxSkip { + nextS = s + maxSkip + } else { + nextS += s + } + + if nextS > sLimit { + goto emitRemainder + } + hashL := hash7(cv, lTableBits) + hashS := hash4(cv, sTableBits) + candidateL = int(lTable[hashL]) + candidateS := int(sTable[hashS]) + lTable[hashL] = uint32(s) + sTable[hashS] = uint32(s) + + if uint32(cv) == load32(src, candidateL) { + break + } + + // Check our short candidate + if uint32(cv) == load32(src, candidateS) { + // Try a long candidate at s+1 + hashL = hash7(cv>>8, lTableBits) + candidateL = int(lTable[hashL]) + lTable[hashL] = uint32(s + 1) + if uint32(cv>>8) == load32(src, candidateL) { + s++ + break + } + // Use our short candidate. + candidateL = candidateS + break + } + + cv = load64(src, nextS) + s = nextS + } + + // Extend backwards + for candidateL > 0 && s > nextEmit && src[candidateL-1] == src[s-1] { + candidateL-- + s-- + } + + // Bail if we exceed the maximum size. + if d+(s-nextEmit) > dstLimit { + return 0 + } + + base := s + offset := base - candidateL + + // Extend the 4-byte match as long as possible. + s += 4 + candidateL += 4 + for s < len(src) { + if len(src)-s < 8 { + if src[s] == src[candidateL] { + s++ + candidateL++ + continue + } + break + } + if diff := load64(src, s) ^ load64(src, candidateL); diff != 0 { + s += bits.TrailingZeros64(diff) >> 3 + break + } + s += 8 + candidateL += 8 + } + + if offset > 65535 && s-base <= 5 && repeat != offset { + // Bail if the match is equal or worse to the encoding. + s = nextS + 1 + if s >= sLimit { + goto emitRemainder + } + cv = load64(src, s) + continue + } + + d += emitLiteral(dst[d:], src[nextEmit:base]) + d += emitCopyNoRepeat(dst[d:], offset, s-base) + repeat = offset + + nextEmit = s + if s >= sLimit { + goto emitRemainder + } + + if d > dstLimit { + // Do we have space for more, if not bail. + return 0 + } + + // Index short & long + index0 := base + 1 + index1 := s - 2 + + cv0 := load64(src, index0) + cv1 := load64(src, index1) + lTable[hash7(cv0, lTableBits)] = uint32(index0) + sTable[hash4(cv0>>8, sTableBits)] = uint32(index0 + 1) + + lTable[hash7(cv1, lTableBits)] = uint32(index1) + sTable[hash4(cv1>>8, sTableBits)] = uint32(index1 + 1) + index0 += 1 + index1 -= 1 + cv = load64(src, s) + + // Index large values sparsely in between. + // We do two starting from different offsets for speed. + index2 := (index0 + index1 + 1) >> 1 + for index2 < index1 { + lTable[hash7(load64(src, index0), lTableBits)] = uint32(index0) + lTable[hash7(load64(src, index2), lTableBits)] = uint32(index2) + index0 += 2 + index2 += 2 + } + } + +emitRemainder: + if nextEmit < len(src) { + // Bail if we exceed the maximum size. + if d+len(src)-nextEmit > dstLimit { + return 0 + } + d += emitLiteral(dst[d:], src[nextEmit:]) + } + return d +} + +// encodeBlockBetterDict encodes a non-empty src to a guaranteed-large-enough dst. It +// assumes that the varint-encoded length of the decompressed bytes has already +// been written. +// +// It also assumes that: +// +// len(dst) >= MaxEncodedLen(len(src)) && +// minNonLiteralBlockSize <= len(src) && len(src) <= maxBlockSize +func encodeBlockBetterDict(dst, src []byte, dict *Dict) (d int) { + // sLimit is when to stop looking for offset/length copies. The inputMargin + // lets us use a fast path for emitLiteral in the main loop, while we are + // looking for copies. + // Initialize the hash tables. + const ( + // Long hash matches. + lTableBits = 17 + maxLTableSize = 1 << lTableBits + + // Short hash matches. + sTableBits = 14 + maxSTableSize = 1 << sTableBits + + maxAhead = 8 // maximum bytes ahead without checking sLimit + + debug = false + ) + + sLimit := len(src) - inputMargin + if sLimit > MaxDictSrcOffset-maxAhead { + sLimit = MaxDictSrcOffset - maxAhead + } + if len(src) < minNonLiteralBlockSize { + return 0 + } + + dict.initBetter() + + var lTable [maxLTableSize]uint32 + var sTable [maxSTableSize]uint32 + + // Bail if we can't compress to at least this. + dstLimit := len(src) - len(src)>>5 - 6 + + // nextEmit is where in src the next emitLiteral should start from. + nextEmit := 0 + + // The encoded form must start with a literal, as there are no previous + // bytes to copy, so we start looking for hash matches at s == 1. + s := 0 + cv := load64(src, s) + + // We initialize repeat to 0, so we never match on first attempt + repeat := len(dict.dict) - dict.repeat + + // While in dict +searchDict: + for { + candidateL := 0 + nextS := 0 + for { + // Next src position to check + nextS = s + (s-nextEmit)>>7 + 1 + if nextS > sLimit { + break searchDict + } + hashL := hash7(cv, lTableBits) + hashS := hash4(cv, sTableBits) + candidateL = int(lTable[hashL]) + candidateS := int(sTable[hashS]) + dictL := int(dict.betterTableLong[hashL]) + dictS := int(dict.betterTableShort[hashS]) + lTable[hashL] = uint32(s) + sTable[hashS] = uint32(s) + + valLong := load64(src, candidateL) + valShort := load64(src, candidateS) + + // If long matches at least 8 bytes, use that. + if s != 0 { + if cv == valLong { + goto emitMatch + } + if cv == valShort { + candidateL = candidateS + goto emitMatch + } + } + + // Check dict repeat. + if repeat >= s+4 { + candidate := len(dict.dict) - repeat + s + if candidate > 0 && uint32(cv) == load32(dict.dict, candidate) { + // Extend back + base := s + for i := candidate; base > nextEmit && i > 0 && dict.dict[i-1] == src[base-1]; { + i-- + base-- + } + d += emitLiteral(dst[d:], src[nextEmit:base]) + if debug && nextEmit != base { + fmt.Println("emitted ", base-nextEmit, "literals") + } + s += 4 + candidate += 4 + for candidate < len(dict.dict)-8 && s <= len(src)-8 { + if diff := load64(src, s) ^ load64(dict.dict, candidate); diff != 0 { + s += bits.TrailingZeros64(diff) >> 3 + break + } + s += 8 + candidate += 8 + } + d += emitRepeat(dst[d:], repeat, s-base) + if debug { + fmt.Println("emitted dict repeat length", s-base, "offset:", repeat, "s:", s) + } + nextEmit = s + if s >= sLimit { + break searchDict + } + // Index in-between + index0 := base + 1 + index1 := s - 2 + + cv = load64(src, s) + for index0 < index1 { + cv0 := load64(src, index0) + cv1 := load64(src, index1) + lTable[hash7(cv0, lTableBits)] = uint32(index0) + sTable[hash4(cv0>>8, sTableBits)] = uint32(index0 + 1) + + lTable[hash7(cv1, lTableBits)] = uint32(index1) + sTable[hash4(cv1>>8, sTableBits)] = uint32(index1 + 1) + index0 += 2 + index1 -= 2 + } + continue + } + } + // Don't try to find match at s==0 + if s == 0 { + cv = load64(src, nextS) + s = nextS + continue + } + + // Long likely matches 7, so take that. + if uint32(cv) == uint32(valLong) { + goto emitMatch + } + + // Long dict... + if uint32(cv) == load32(dict.dict, dictL) { + candidateL = dictL + goto emitDict + } + + // Check our short candidate + if uint32(cv) == uint32(valShort) { + // Try a long candidate at s+1 + hashL = hash7(cv>>8, lTableBits) + candidateL = int(lTable[hashL]) + lTable[hashL] = uint32(s + 1) + if uint32(cv>>8) == load32(src, candidateL) { + s++ + goto emitMatch + } + // Use our short candidate. + candidateL = candidateS + goto emitMatch + } + if uint32(cv) == load32(dict.dict, dictS) { + // Try a long candidate at s+1 + hashL = hash7(cv>>8, lTableBits) + candidateL = int(lTable[hashL]) + lTable[hashL] = uint32(s + 1) + if uint32(cv>>8) == load32(src, candidateL) { + s++ + goto emitMatch + } + candidateL = dictS + goto emitDict + } + cv = load64(src, nextS) + s = nextS + } + emitDict: + { + if debug { + if load32(dict.dict, candidateL) != load32(src, s) { + panic("dict emit mismatch") + } + } + // Extend backwards. + // The top bytes will be rechecked to get the full match. + for candidateL > 0 && s > nextEmit && dict.dict[candidateL-1] == src[s-1] { + candidateL-- + s-- + } + + // Bail if we exceed the maximum size. + if d+(s-nextEmit) > dstLimit { + return 0 + } + + // A 4-byte match has been found. We'll later see if more than 4 bytes + // match. But, prior to the match, src[nextEmit:s] are unmatched. Emit + // them as literal bytes. + + d += emitLiteral(dst[d:], src[nextEmit:s]) + if debug && nextEmit != s { + fmt.Println("emitted ", s-nextEmit, "literals") + } + { + // Invariant: we have a 4-byte match at s, and no need to emit any + // literal bytes prior to s. + base := s + offset := s + (len(dict.dict)) - candidateL + + // Extend the 4-byte match as long as possible. + s += 4 + candidateL += 4 + for s <= len(src)-8 && len(dict.dict)-candidateL >= 8 { + if diff := load64(src, s) ^ load64(dict.dict, candidateL); diff != 0 { + s += bits.TrailingZeros64(diff) >> 3 + break + } + s += 8 + candidateL += 8 + } + + if repeat == offset { + if debug { + fmt.Println("emitted dict repeat, length", s-base, "offset:", offset, "s:", s, "dict offset:", candidateL) + } + d += emitRepeat(dst[d:], offset, s-base) + } else { + if debug { + fmt.Println("emitted dict copy, length", s-base, "offset:", offset, "s:", s, "dict offset:", candidateL) + } + // Matches longer than 64 are split. + if s <= sLimit || s-base < 8 { + d += emitCopy(dst[d:], offset, s-base) + } else { + // Split to ensure we don't start a copy within next block. + d += emitCopy(dst[d:], offset, 4) + d += emitRepeat(dst[d:], offset, s-base-4) + } + repeat = offset + } + if false { + // Validate match. + if s <= candidateL { + panic("s <= candidate") + } + a := src[base:s] + b := dict.dict[base-repeat : base-repeat+(s-base)] + if !bytes.Equal(a, b) { + panic("mismatch") + } + } + + nextEmit = s + if s >= sLimit { + break searchDict + } + + if d > dstLimit { + // Do we have space for more, if not bail. + return 0 + } + + // Index short & long + index0 := base + 1 + index1 := s - 2 + + cv0 := load64(src, index0) + cv1 := load64(src, index1) + lTable[hash7(cv0, lTableBits)] = uint32(index0) + sTable[hash4(cv0>>8, sTableBits)] = uint32(index0 + 1) + + lTable[hash7(cv1, lTableBits)] = uint32(index1) + sTable[hash4(cv1>>8, sTableBits)] = uint32(index1 + 1) + index0 += 1 + index1 -= 1 + cv = load64(src, s) + + // index every second long in between. + for index0 < index1 { + lTable[hash7(load64(src, index0), lTableBits)] = uint32(index0) + lTable[hash7(load64(src, index1), lTableBits)] = uint32(index1) + index0 += 2 + index1 -= 2 + } + } + continue + } + emitMatch: + + // Extend backwards + for candidateL > 0 && s > nextEmit && src[candidateL-1] == src[s-1] { + candidateL-- + s-- + } + + // Bail if we exceed the maximum size. + if d+(s-nextEmit) > dstLimit { + return 0 + } + + base := s + offset := base - candidateL + + // Extend the 4-byte match as long as possible. + s += 4 + candidateL += 4 + for s < len(src) { + if len(src)-s < 8 { + if src[s] == src[candidateL] { + s++ + candidateL++ + continue + } + break + } + if diff := load64(src, s) ^ load64(src, candidateL); diff != 0 { + s += bits.TrailingZeros64(diff) >> 3 + break + } + s += 8 + candidateL += 8 + } + + if offset > 65535 && s-base <= 5 && repeat != offset { + // Bail if the match is equal or worse to the encoding. + s = nextS + 1 + if s >= sLimit { + goto emitRemainder + } + cv = load64(src, s) + continue + } + + d += emitLiteral(dst[d:], src[nextEmit:base]) + if debug && nextEmit != s { + fmt.Println("emitted ", s-nextEmit, "literals") + } + if repeat == offset { + if debug { + fmt.Println("emitted match repeat, length", s-base, "offset:", offset, "s:", s) + } + d += emitRepeat(dst[d:], offset, s-base) + } else { + if debug { + fmt.Println("emitted match copy, length", s-base, "offset:", offset, "s:", s) + } + d += emitCopy(dst[d:], offset, s-base) + repeat = offset + } + + nextEmit = s + if s >= sLimit { + goto emitRemainder + } + + if d > dstLimit { + // Do we have space for more, if not bail. + return 0 + } + + // Index short & long + index0 := base + 1 + index1 := s - 2 + + cv0 := load64(src, index0) + cv1 := load64(src, index1) + lTable[hash7(cv0, lTableBits)] = uint32(index0) + sTable[hash4(cv0>>8, sTableBits)] = uint32(index0 + 1) + + lTable[hash7(cv1, lTableBits)] = uint32(index1) + sTable[hash4(cv1>>8, sTableBits)] = uint32(index1 + 1) + index0 += 1 + index1 -= 1 + cv = load64(src, s) + + // Index large values sparsely in between. + // We do two starting from different offsets for speed. + index2 := (index0 + index1 + 1) >> 1 + for index2 < index1 { + lTable[hash7(load64(src, index0), lTableBits)] = uint32(index0) + lTable[hash7(load64(src, index2), lTableBits)] = uint32(index2) + index0 += 2 + index2 += 2 + } + } + + // Search without dict: + if repeat > s { + repeat = 0 + } + + // No more dict + sLimit = len(src) - inputMargin + if s >= sLimit { + goto emitRemainder + } + cv = load64(src, s) + if debug { + fmt.Println("now", s, "->", sLimit, "out:", d, "left:", len(src)-s, "nextemit:", nextEmit, "dstLimit:", dstLimit, "s:", s) + } + for { + candidateL := 0 + nextS := 0 + for { + // Next src position to check + nextS = s + (s-nextEmit)>>7 + 1 + if nextS > sLimit { + goto emitRemainder + } + hashL := hash7(cv, lTableBits) + hashS := hash4(cv, sTableBits) + candidateL = int(lTable[hashL]) + candidateS := int(sTable[hashS]) + lTable[hashL] = uint32(s) + sTable[hashS] = uint32(s) + + valLong := load64(src, candidateL) + valShort := load64(src, candidateS) + + // If long matches at least 8 bytes, use that. + if cv == valLong { + break + } + if cv == valShort { + candidateL = candidateS + break + } + + // Check repeat at offset checkRep. + const checkRep = 1 + // Minimum length of a repeat. Tested with various values. + // While 4-5 offers improvements in some, 6 reduces + // regressions significantly. + const wantRepeatBytes = 6 + const repeatMask = ((1 << (wantRepeatBytes * 8)) - 1) << (8 * checkRep) + if false && repeat > 0 && cv&repeatMask == load64(src, s-repeat)&repeatMask { + base := s + checkRep + // Extend back + for i := base - repeat; base > nextEmit && i > 0 && src[i-1] == src[base-1]; { + i-- + base-- + } + d += emitLiteral(dst[d:], src[nextEmit:base]) + + // Extend forward + candidate := s - repeat + wantRepeatBytes + checkRep + s += wantRepeatBytes + checkRep + for s < len(src) { + if len(src)-s < 8 { + if src[s] == src[candidate] { + s++ + candidate++ + continue + } + break + } + if diff := load64(src, s) ^ load64(src, candidate); diff != 0 { + s += bits.TrailingZeros64(diff) >> 3 + break + } + s += 8 + candidate += 8 + } + // same as `add := emitCopy(dst[d:], repeat, s-base)` but skips storing offset. + d += emitRepeat(dst[d:], repeat, s-base) + nextEmit = s + if s >= sLimit { + goto emitRemainder + } + // Index in-between + index0 := base + 1 + index1 := s - 2 + + for index0 < index1 { + cv0 := load64(src, index0) + cv1 := load64(src, index1) + lTable[hash7(cv0, lTableBits)] = uint32(index0) + sTable[hash4(cv0>>8, sTableBits)] = uint32(index0 + 1) + + lTable[hash7(cv1, lTableBits)] = uint32(index1) + sTable[hash4(cv1>>8, sTableBits)] = uint32(index1 + 1) + index0 += 2 + index1 -= 2 + } + + cv = load64(src, s) + continue + } + + // Long likely matches 7, so take that. + if uint32(cv) == uint32(valLong) { + break + } + + // Check our short candidate + if uint32(cv) == uint32(valShort) { + // Try a long candidate at s+1 + hashL = hash7(cv>>8, lTableBits) + candidateL = int(lTable[hashL]) + lTable[hashL] = uint32(s + 1) + if uint32(cv>>8) == load32(src, candidateL) { + s++ + break + } + // Use our short candidate. + candidateL = candidateS + break + } + + cv = load64(src, nextS) + s = nextS + } + + // Extend backwards + for candidateL > 0 && s > nextEmit && src[candidateL-1] == src[s-1] { + candidateL-- + s-- + } + + // Bail if we exceed the maximum size. + if d+(s-nextEmit) > dstLimit { + return 0 + } + + base := s + offset := base - candidateL + + // Extend the 4-byte match as long as possible. + s += 4 + candidateL += 4 + for s < len(src) { + if len(src)-s < 8 { + if src[s] == src[candidateL] { + s++ + candidateL++ + continue + } + break + } + if diff := load64(src, s) ^ load64(src, candidateL); diff != 0 { + s += bits.TrailingZeros64(diff) >> 3 + break + } + s += 8 + candidateL += 8 + } + + if offset > 65535 && s-base <= 5 && repeat != offset { + // Bail if the match is equal or worse to the encoding. + s = nextS + 1 + if s >= sLimit { + goto emitRemainder + } + cv = load64(src, s) + continue + } + + d += emitLiteral(dst[d:], src[nextEmit:base]) + if repeat == offset { + d += emitRepeat(dst[d:], offset, s-base) + } else { + d += emitCopy(dst[d:], offset, s-base) + repeat = offset + } + + nextEmit = s + if s >= sLimit { + goto emitRemainder + } + + if d > dstLimit { + // Do we have space for more, if not bail. + return 0 + } + + // Index short & long + index0 := base + 1 + index1 := s - 2 + + cv0 := load64(src, index0) + cv1 := load64(src, index1) + lTable[hash7(cv0, lTableBits)] = uint32(index0) + sTable[hash4(cv0>>8, sTableBits)] = uint32(index0 + 1) + + lTable[hash7(cv1, lTableBits)] = uint32(index1) + sTable[hash4(cv1>>8, sTableBits)] = uint32(index1 + 1) + index0 += 1 + index1 -= 1 + cv = load64(src, s) + + // Index large values sparsely in between. + // We do two starting from different offsets for speed. + index2 := (index0 + index1 + 1) >> 1 + for index2 < index1 { + lTable[hash7(load64(src, index0), lTableBits)] = uint32(index0) + lTable[hash7(load64(src, index2), lTableBits)] = uint32(index2) + index0 += 2 + index2 += 2 + } + } + +emitRemainder: + if nextEmit < len(src) { + // Bail if we exceed the maximum size. + if d+len(src)-nextEmit > dstLimit { + return 0 + } + d += emitLiteral(dst[d:], src[nextEmit:]) + } + return d +} diff --git a/vendor/github.com/klauspost/compress/s2/encode_go.go b/vendor/github.com/klauspost/compress/s2/encode_go.go new file mode 100644 index 000000000..6b393c34d --- /dev/null +++ b/vendor/github.com/klauspost/compress/s2/encode_go.go @@ -0,0 +1,729 @@ +//go:build !amd64 || appengine || !gc || noasm +// +build !amd64 appengine !gc noasm + +package s2 + +import ( + "bytes" + "math/bits" +) + +const hasAmd64Asm = false + +// encodeBlock encodes a non-empty src to a guaranteed-large-enough dst. It +// assumes that the varint-encoded length of the decompressed bytes has already +// been written. +// +// It also assumes that: +// +// len(dst) >= MaxEncodedLen(len(src)) +func encodeBlock(dst, src []byte) (d int) { + if len(src) < minNonLiteralBlockSize { + return 0 + } + return encodeBlockGo(dst, src) +} + +// encodeBlockBetter encodes a non-empty src to a guaranteed-large-enough dst. It +// assumes that the varint-encoded length of the decompressed bytes has already +// been written. +// +// It also assumes that: +// +// len(dst) >= MaxEncodedLen(len(src)) +func encodeBlockBetter(dst, src []byte) (d int) { + return encodeBlockBetterGo(dst, src) +} + +// encodeBlockBetter encodes a non-empty src to a guaranteed-large-enough dst. It +// assumes that the varint-encoded length of the decompressed bytes has already +// been written. +// +// It also assumes that: +// +// len(dst) >= MaxEncodedLen(len(src)) +func encodeBlockBetterSnappy(dst, src []byte) (d int) { + return encodeBlockBetterSnappyGo(dst, src) +} + +// encodeBlock encodes a non-empty src to a guaranteed-large-enough dst. It +// assumes that the varint-encoded length of the decompressed bytes has already +// been written. +// +// It also assumes that: +// +// len(dst) >= MaxEncodedLen(len(src)) +func encodeBlockSnappy(dst, src []byte) (d int) { + if len(src) < minNonLiteralBlockSize { + return 0 + } + return encodeBlockSnappyGo(dst, src) +} + +// emitLiteral writes a literal chunk and returns the number of bytes written. +// +// It assumes that: +// +// dst is long enough to hold the encoded bytes +// 0 <= len(lit) && len(lit) <= math.MaxUint32 +func emitLiteral(dst, lit []byte) int { + if len(lit) == 0 { + return 0 + } + const num = 63<<2 | tagLiteral + i, n := 0, uint(len(lit)-1) + switch { + case n < 60: + dst[0] = uint8(n)<<2 | tagLiteral + i = 1 + case n < 1<<8: + dst[1] = uint8(n) + dst[0] = 60<<2 | tagLiteral + i = 2 + case n < 1<<16: + dst[2] = uint8(n >> 8) + dst[1] = uint8(n) + dst[0] = 61<<2 | tagLiteral + i = 3 + case n < 1<<24: + dst[3] = uint8(n >> 16) + dst[2] = uint8(n >> 8) + dst[1] = uint8(n) + dst[0] = 62<<2 | tagLiteral + i = 4 + default: + dst[4] = uint8(n >> 24) + dst[3] = uint8(n >> 16) + dst[2] = uint8(n >> 8) + dst[1] = uint8(n) + dst[0] = 63<<2 | tagLiteral + i = 5 + } + return i + copy(dst[i:], lit) +} + +// emitRepeat writes a repeat chunk and returns the number of bytes written. +// Length must be at least 4 and < 1<<24 +func emitRepeat(dst []byte, offset, length int) int { + // Repeat offset, make length cheaper + length -= 4 + if length <= 4 { + dst[0] = uint8(length)<<2 | tagCopy1 + dst[1] = 0 + return 2 + } + if length < 8 && offset < 2048 { + // Encode WITH offset + dst[1] = uint8(offset) + dst[0] = uint8(offset>>8)<<5 | uint8(length)<<2 | tagCopy1 + return 2 + } + if length < (1<<8)+4 { + length -= 4 + dst[2] = uint8(length) + dst[1] = 0 + dst[0] = 5<<2 | tagCopy1 + return 3 + } + if length < (1<<16)+(1<<8) { + length -= 1 << 8 + dst[3] = uint8(length >> 8) + dst[2] = uint8(length >> 0) + dst[1] = 0 + dst[0] = 6<<2 | tagCopy1 + return 4 + } + const maxRepeat = (1 << 24) - 1 + length -= 1 << 16 + left := 0 + if length > maxRepeat { + left = length - maxRepeat + 4 + length = maxRepeat - 4 + } + dst[4] = uint8(length >> 16) + dst[3] = uint8(length >> 8) + dst[2] = uint8(length >> 0) + dst[1] = 0 + dst[0] = 7<<2 | tagCopy1 + if left > 0 { + return 5 + emitRepeat(dst[5:], offset, left) + } + return 5 +} + +// emitCopy writes a copy chunk and returns the number of bytes written. +// +// It assumes that: +// +// dst is long enough to hold the encoded bytes +// 1 <= offset && offset <= math.MaxUint32 +// 4 <= length && length <= 1 << 24 +func emitCopy(dst []byte, offset, length int) int { + if offset >= 65536 { + i := 0 + if length > 64 { + // Emit a length 64 copy, encoded as 5 bytes. + dst[4] = uint8(offset >> 24) + dst[3] = uint8(offset >> 16) + dst[2] = uint8(offset >> 8) + dst[1] = uint8(offset) + dst[0] = 63<<2 | tagCopy4 + length -= 64 + if length >= 4 { + // Emit remaining as repeats + return 5 + emitRepeat(dst[5:], offset, length) + } + i = 5 + } + if length == 0 { + return i + } + // Emit a copy, offset encoded as 4 bytes. + dst[i+0] = uint8(length-1)<<2 | tagCopy4 + dst[i+1] = uint8(offset) + dst[i+2] = uint8(offset >> 8) + dst[i+3] = uint8(offset >> 16) + dst[i+4] = uint8(offset >> 24) + return i + 5 + } + + // Offset no more than 2 bytes. + if length > 64 { + off := 3 + if offset < 2048 { + // emit 8 bytes as tagCopy1, rest as repeats. + dst[1] = uint8(offset) + dst[0] = uint8(offset>>8)<<5 | uint8(8-4)<<2 | tagCopy1 + length -= 8 + off = 2 + } else { + // Emit a length 60 copy, encoded as 3 bytes. + // Emit remaining as repeat value (minimum 4 bytes). + dst[2] = uint8(offset >> 8) + dst[1] = uint8(offset) + dst[0] = 59<<2 | tagCopy2 + length -= 60 + } + // Emit remaining as repeats, at least 4 bytes remain. + return off + emitRepeat(dst[off:], offset, length) + } + if length >= 12 || offset >= 2048 { + // Emit the remaining copy, encoded as 3 bytes. + dst[2] = uint8(offset >> 8) + dst[1] = uint8(offset) + dst[0] = uint8(length-1)<<2 | tagCopy2 + return 3 + } + // Emit the remaining copy, encoded as 2 bytes. + dst[1] = uint8(offset) + dst[0] = uint8(offset>>8)<<5 | uint8(length-4)<<2 | tagCopy1 + return 2 +} + +// emitCopyNoRepeat writes a copy chunk and returns the number of bytes written. +// +// It assumes that: +// +// dst is long enough to hold the encoded bytes +// 1 <= offset && offset <= math.MaxUint32 +// 4 <= length && length <= 1 << 24 +func emitCopyNoRepeat(dst []byte, offset, length int) int { + if offset >= 65536 { + i := 0 + if length > 64 { + // Emit a length 64 copy, encoded as 5 bytes. + dst[4] = uint8(offset >> 24) + dst[3] = uint8(offset >> 16) + dst[2] = uint8(offset >> 8) + dst[1] = uint8(offset) + dst[0] = 63<<2 | tagCopy4 + length -= 64 + if length >= 4 { + // Emit remaining as repeats + return 5 + emitCopyNoRepeat(dst[5:], offset, length) + } + i = 5 + } + if length == 0 { + return i + } + // Emit a copy, offset encoded as 4 bytes. + dst[i+0] = uint8(length-1)<<2 | tagCopy4 + dst[i+1] = uint8(offset) + dst[i+2] = uint8(offset >> 8) + dst[i+3] = uint8(offset >> 16) + dst[i+4] = uint8(offset >> 24) + return i + 5 + } + + // Offset no more than 2 bytes. + if length > 64 { + // Emit a length 60 copy, encoded as 3 bytes. + // Emit remaining as repeat value (minimum 4 bytes). + dst[2] = uint8(offset >> 8) + dst[1] = uint8(offset) + dst[0] = 59<<2 | tagCopy2 + length -= 60 + // Emit remaining as repeats, at least 4 bytes remain. + return 3 + emitCopyNoRepeat(dst[3:], offset, length) + } + if length >= 12 || offset >= 2048 { + // Emit the remaining copy, encoded as 3 bytes. + dst[2] = uint8(offset >> 8) + dst[1] = uint8(offset) + dst[0] = uint8(length-1)<<2 | tagCopy2 + return 3 + } + // Emit the remaining copy, encoded as 2 bytes. + dst[1] = uint8(offset) + dst[0] = uint8(offset>>8)<<5 | uint8(length-4)<<2 | tagCopy1 + return 2 +} + +// matchLen returns how many bytes match in a and b +// +// It assumes that: +// +// len(a) <= len(b) +func matchLen(a []byte, b []byte) int { + b = b[:len(a)] + var checked int + if len(a) > 4 { + // Try 4 bytes first + if diff := load32(a, 0) ^ load32(b, 0); diff != 0 { + return bits.TrailingZeros32(diff) >> 3 + } + // Switch to 8 byte matching. + checked = 4 + a = a[4:] + b = b[4:] + for len(a) >= 8 { + b = b[:len(a)] + if diff := load64(a, 0) ^ load64(b, 0); diff != 0 { + return checked + (bits.TrailingZeros64(diff) >> 3) + } + checked += 8 + a = a[8:] + b = b[8:] + } + } + b = b[:len(a)] + for i := range a { + if a[i] != b[i] { + return int(i) + checked + } + } + return len(a) + checked +} + +// input must be > inputMargin +func calcBlockSize(src []byte) (d int) { + // Initialize the hash table. + const ( + tableBits = 13 + maxTableSize = 1 << tableBits + ) + + var table [maxTableSize]uint32 + + // sLimit is when to stop looking for offset/length copies. The inputMargin + // lets us use a fast path for emitLiteral in the main loop, while we are + // looking for copies. + sLimit := len(src) - inputMargin + + // Bail if we can't compress to at least this. + dstLimit := len(src) - len(src)>>5 - 5 + + // nextEmit is where in src the next emitLiteral should start from. + nextEmit := 0 + + // The encoded form must start with a literal, as there are no previous + // bytes to copy, so we start looking for hash matches at s == 1. + s := 1 + cv := load64(src, s) + + // We search for a repeat at -1, but don't output repeats when nextEmit == 0 + repeat := 1 + + for { + candidate := 0 + for { + // Next src position to check + nextS := s + (s-nextEmit)>>6 + 4 + if nextS > sLimit { + goto emitRemainder + } + hash0 := hash6(cv, tableBits) + hash1 := hash6(cv>>8, tableBits) + candidate = int(table[hash0]) + candidate2 := int(table[hash1]) + table[hash0] = uint32(s) + table[hash1] = uint32(s + 1) + hash2 := hash6(cv>>16, tableBits) + + // Check repeat at offset checkRep. + const checkRep = 1 + if uint32(cv>>(checkRep*8)) == load32(src, s-repeat+checkRep) { + base := s + checkRep + // Extend back + for i := base - repeat; base > nextEmit && i > 0 && src[i-1] == src[base-1]; { + i-- + base-- + } + d += emitLiteralSize(src[nextEmit:base]) + + // Extend forward + candidate := s - repeat + 4 + checkRep + s += 4 + checkRep + for s <= sLimit { + if diff := load64(src, s) ^ load64(src, candidate); diff != 0 { + s += bits.TrailingZeros64(diff) >> 3 + break + } + s += 8 + candidate += 8 + } + + d += emitCopyNoRepeatSize(repeat, s-base) + nextEmit = s + if s >= sLimit { + goto emitRemainder + } + + cv = load64(src, s) + continue + } + + if uint32(cv) == load32(src, candidate) { + break + } + candidate = int(table[hash2]) + if uint32(cv>>8) == load32(src, candidate2) { + table[hash2] = uint32(s + 2) + candidate = candidate2 + s++ + break + } + table[hash2] = uint32(s + 2) + if uint32(cv>>16) == load32(src, candidate) { + s += 2 + break + } + + cv = load64(src, nextS) + s = nextS + } + + // Extend backwards + for candidate > 0 && s > nextEmit && src[candidate-1] == src[s-1] { + candidate-- + s-- + } + + // Bail if we exceed the maximum size. + if d+(s-nextEmit) > dstLimit { + return 0 + } + + // A 4-byte match has been found. We'll later see if more than 4 bytes + // match. But, prior to the match, src[nextEmit:s] are unmatched. Emit + // them as literal bytes. + + d += emitLiteralSize(src[nextEmit:s]) + + // Call emitCopy, and then see if another emitCopy could be our next + // move. Repeat until we find no match for the input immediately after + // what was consumed by the last emitCopy call. + // + // If we exit this loop normally then we need to call emitLiteral next, + // though we don't yet know how big the literal will be. We handle that + // by proceeding to the next iteration of the main loop. We also can + // exit this loop via goto if we get close to exhausting the input. + for { + // Invariant: we have a 4-byte match at s, and no need to emit any + // literal bytes prior to s. + base := s + repeat = base - candidate + + // Extend the 4-byte match as long as possible. + s += 4 + candidate += 4 + for s <= len(src)-8 { + if diff := load64(src, s) ^ load64(src, candidate); diff != 0 { + s += bits.TrailingZeros64(diff) >> 3 + break + } + s += 8 + candidate += 8 + } + + d += emitCopyNoRepeatSize(repeat, s-base) + if false { + // Validate match. + a := src[base:s] + b := src[base-repeat : base-repeat+(s-base)] + if !bytes.Equal(a, b) { + panic("mismatch") + } + } + + nextEmit = s + if s >= sLimit { + goto emitRemainder + } + + if d > dstLimit { + // Do we have space for more, if not bail. + return 0 + } + // Check for an immediate match, otherwise start search at s+1 + x := load64(src, s-2) + m2Hash := hash6(x, tableBits) + currHash := hash6(x>>16, tableBits) + candidate = int(table[currHash]) + table[m2Hash] = uint32(s - 2) + table[currHash] = uint32(s) + if uint32(x>>16) != load32(src, candidate) { + cv = load64(src, s+1) + s++ + break + } + } + } + +emitRemainder: + if nextEmit < len(src) { + // Bail if we exceed the maximum size. + if d+len(src)-nextEmit > dstLimit { + return 0 + } + d += emitLiteralSize(src[nextEmit:]) + } + return d +} + +// length must be > inputMargin. +func calcBlockSizeSmall(src []byte) (d int) { + // Initialize the hash table. + const ( + tableBits = 9 + maxTableSize = 1 << tableBits + ) + + var table [maxTableSize]uint32 + + // sLimit is when to stop looking for offset/length copies. The inputMargin + // lets us use a fast path for emitLiteral in the main loop, while we are + // looking for copies. + sLimit := len(src) - inputMargin + + // Bail if we can't compress to at least this. + dstLimit := len(src) - len(src)>>5 - 5 + + // nextEmit is where in src the next emitLiteral should start from. + nextEmit := 0 + + // The encoded form must start with a literal, as there are no previous + // bytes to copy, so we start looking for hash matches at s == 1. + s := 1 + cv := load64(src, s) + + // We search for a repeat at -1, but don't output repeats when nextEmit == 0 + repeat := 1 + + for { + candidate := 0 + for { + // Next src position to check + nextS := s + (s-nextEmit)>>6 + 4 + if nextS > sLimit { + goto emitRemainder + } + hash0 := hash6(cv, tableBits) + hash1 := hash6(cv>>8, tableBits) + candidate = int(table[hash0]) + candidate2 := int(table[hash1]) + table[hash0] = uint32(s) + table[hash1] = uint32(s + 1) + hash2 := hash6(cv>>16, tableBits) + + // Check repeat at offset checkRep. + const checkRep = 1 + if uint32(cv>>(checkRep*8)) == load32(src, s-repeat+checkRep) { + base := s + checkRep + // Extend back + for i := base - repeat; base > nextEmit && i > 0 && src[i-1] == src[base-1]; { + i-- + base-- + } + d += emitLiteralSize(src[nextEmit:base]) + + // Extend forward + candidate := s - repeat + 4 + checkRep + s += 4 + checkRep + for s <= sLimit { + if diff := load64(src, s) ^ load64(src, candidate); diff != 0 { + s += bits.TrailingZeros64(diff) >> 3 + break + } + s += 8 + candidate += 8 + } + + d += emitCopyNoRepeatSize(repeat, s-base) + nextEmit = s + if s >= sLimit { + goto emitRemainder + } + + cv = load64(src, s) + continue + } + + if uint32(cv) == load32(src, candidate) { + break + } + candidate = int(table[hash2]) + if uint32(cv>>8) == load32(src, candidate2) { + table[hash2] = uint32(s + 2) + candidate = candidate2 + s++ + break + } + table[hash2] = uint32(s + 2) + if uint32(cv>>16) == load32(src, candidate) { + s += 2 + break + } + + cv = load64(src, nextS) + s = nextS + } + + // Extend backwards + for candidate > 0 && s > nextEmit && src[candidate-1] == src[s-1] { + candidate-- + s-- + } + + // Bail if we exceed the maximum size. + if d+(s-nextEmit) > dstLimit { + return 0 + } + + // A 4-byte match has been found. We'll later see if more than 4 bytes + // match. But, prior to the match, src[nextEmit:s] are unmatched. Emit + // them as literal bytes. + + d += emitLiteralSize(src[nextEmit:s]) + + // Call emitCopy, and then see if another emitCopy could be our next + // move. Repeat until we find no match for the input immediately after + // what was consumed by the last emitCopy call. + // + // If we exit this loop normally then we need to call emitLiteral next, + // though we don't yet know how big the literal will be. We handle that + // by proceeding to the next iteration of the main loop. We also can + // exit this loop via goto if we get close to exhausting the input. + for { + // Invariant: we have a 4-byte match at s, and no need to emit any + // literal bytes prior to s. + base := s + repeat = base - candidate + + // Extend the 4-byte match as long as possible. + s += 4 + candidate += 4 + for s <= len(src)-8 { + if diff := load64(src, s) ^ load64(src, candidate); diff != 0 { + s += bits.TrailingZeros64(diff) >> 3 + break + } + s += 8 + candidate += 8 + } + + d += emitCopyNoRepeatSize(repeat, s-base) + if false { + // Validate match. + a := src[base:s] + b := src[base-repeat : base-repeat+(s-base)] + if !bytes.Equal(a, b) { + panic("mismatch") + } + } + + nextEmit = s + if s >= sLimit { + goto emitRemainder + } + + if d > dstLimit { + // Do we have space for more, if not bail. + return 0 + } + // Check for an immediate match, otherwise start search at s+1 + x := load64(src, s-2) + m2Hash := hash6(x, tableBits) + currHash := hash6(x>>16, tableBits) + candidate = int(table[currHash]) + table[m2Hash] = uint32(s - 2) + table[currHash] = uint32(s) + if uint32(x>>16) != load32(src, candidate) { + cv = load64(src, s+1) + s++ + break + } + } + } + +emitRemainder: + if nextEmit < len(src) { + // Bail if we exceed the maximum size. + if d+len(src)-nextEmit > dstLimit { + return 0 + } + d += emitLiteralSize(src[nextEmit:]) + } + return d +} + +// emitLiteral writes a literal chunk and returns the number of bytes written. +// +// It assumes that: +// +// dst is long enough to hold the encoded bytes +// 0 <= len(lit) && len(lit) <= math.MaxUint32 +func emitLiteralSize(lit []byte) int { + if len(lit) == 0 { + return 0 + } + switch { + case len(lit) <= 60: + return len(lit) + 1 + case len(lit) <= 1<<8: + return len(lit) + 2 + case len(lit) <= 1<<16: + return len(lit) + 3 + case len(lit) <= 1<<24: + return len(lit) + 4 + default: + return len(lit) + 5 + } +} + +func cvtLZ4BlockAsm(dst []byte, src []byte) (uncompressed int, dstUsed int) { + panic("cvtLZ4BlockAsm should be unreachable") +} + +func cvtLZ4BlockSnappyAsm(dst []byte, src []byte) (uncompressed int, dstUsed int) { + panic("cvtLZ4BlockSnappyAsm should be unreachable") +} + +func cvtLZ4sBlockAsm(dst []byte, src []byte) (uncompressed int, dstUsed int) { + panic("cvtLZ4sBlockAsm should be unreachable") +} + +func cvtLZ4sBlockSnappyAsm(dst []byte, src []byte) (uncompressed int, dstUsed int) { + panic("cvtLZ4sBlockSnappyAsm should be unreachable") +} diff --git a/vendor/github.com/klauspost/compress/s2/encodeblock_amd64.go b/vendor/github.com/klauspost/compress/s2/encodeblock_amd64.go new file mode 100644 index 000000000..297e41501 --- /dev/null +++ b/vendor/github.com/klauspost/compress/s2/encodeblock_amd64.go @@ -0,0 +1,228 @@ +// Code generated by command: go run gen.go -out ../encodeblock_amd64.s -stubs ../encodeblock_amd64.go -pkg=s2. DO NOT EDIT. + +//go:build !appengine && !noasm && gc && !noasm + +package s2 + +func _dummy_() + +// encodeBlockAsm encodes a non-empty src to a guaranteed-large-enough dst. +// Maximum input 4294967295 bytes. +// It assumes that the varint-encoded length of the decompressed bytes has already been written. +// +//go:noescape +func encodeBlockAsm(dst []byte, src []byte) int + +// encodeBlockAsm4MB encodes a non-empty src to a guaranteed-large-enough dst. +// Maximum input 4194304 bytes. +// It assumes that the varint-encoded length of the decompressed bytes has already been written. +// +//go:noescape +func encodeBlockAsm4MB(dst []byte, src []byte) int + +// encodeBlockAsm12B encodes a non-empty src to a guaranteed-large-enough dst. +// Maximum input 16383 bytes. +// It assumes that the varint-encoded length of the decompressed bytes has already been written. +// +//go:noescape +func encodeBlockAsm12B(dst []byte, src []byte) int + +// encodeBlockAsm10B encodes a non-empty src to a guaranteed-large-enough dst. +// Maximum input 4095 bytes. +// It assumes that the varint-encoded length of the decompressed bytes has already been written. +// +//go:noescape +func encodeBlockAsm10B(dst []byte, src []byte) int + +// encodeBlockAsm8B encodes a non-empty src to a guaranteed-large-enough dst. +// Maximum input 511 bytes. +// It assumes that the varint-encoded length of the decompressed bytes has already been written. +// +//go:noescape +func encodeBlockAsm8B(dst []byte, src []byte) int + +// encodeBetterBlockAsm encodes a non-empty src to a guaranteed-large-enough dst. +// Maximum input 4294967295 bytes. +// It assumes that the varint-encoded length of the decompressed bytes has already been written. +// +//go:noescape +func encodeBetterBlockAsm(dst []byte, src []byte) int + +// encodeBetterBlockAsm4MB encodes a non-empty src to a guaranteed-large-enough dst. +// Maximum input 4194304 bytes. +// It assumes that the varint-encoded length of the decompressed bytes has already been written. +// +//go:noescape +func encodeBetterBlockAsm4MB(dst []byte, src []byte) int + +// encodeBetterBlockAsm12B encodes a non-empty src to a guaranteed-large-enough dst. +// Maximum input 16383 bytes. +// It assumes that the varint-encoded length of the decompressed bytes has already been written. +// +//go:noescape +func encodeBetterBlockAsm12B(dst []byte, src []byte) int + +// encodeBetterBlockAsm10B encodes a non-empty src to a guaranteed-large-enough dst. +// Maximum input 4095 bytes. +// It assumes that the varint-encoded length of the decompressed bytes has already been written. +// +//go:noescape +func encodeBetterBlockAsm10B(dst []byte, src []byte) int + +// encodeBetterBlockAsm8B encodes a non-empty src to a guaranteed-large-enough dst. +// Maximum input 511 bytes. +// It assumes that the varint-encoded length of the decompressed bytes has already been written. +// +//go:noescape +func encodeBetterBlockAsm8B(dst []byte, src []byte) int + +// encodeSnappyBlockAsm encodes a non-empty src to a guaranteed-large-enough dst. +// Maximum input 4294967295 bytes. +// It assumes that the varint-encoded length of the decompressed bytes has already been written. +// +//go:noescape +func encodeSnappyBlockAsm(dst []byte, src []byte) int + +// encodeSnappyBlockAsm64K encodes a non-empty src to a guaranteed-large-enough dst. +// Maximum input 65535 bytes. +// It assumes that the varint-encoded length of the decompressed bytes has already been written. +// +//go:noescape +func encodeSnappyBlockAsm64K(dst []byte, src []byte) int + +// encodeSnappyBlockAsm12B encodes a non-empty src to a guaranteed-large-enough dst. +// Maximum input 16383 bytes. +// It assumes that the varint-encoded length of the decompressed bytes has already been written. +// +//go:noescape +func encodeSnappyBlockAsm12B(dst []byte, src []byte) int + +// encodeSnappyBlockAsm10B encodes a non-empty src to a guaranteed-large-enough dst. +// Maximum input 4095 bytes. +// It assumes that the varint-encoded length of the decompressed bytes has already been written. +// +//go:noescape +func encodeSnappyBlockAsm10B(dst []byte, src []byte) int + +// encodeSnappyBlockAsm8B encodes a non-empty src to a guaranteed-large-enough dst. +// Maximum input 511 bytes. +// It assumes that the varint-encoded length of the decompressed bytes has already been written. +// +//go:noescape +func encodeSnappyBlockAsm8B(dst []byte, src []byte) int + +// encodeSnappyBetterBlockAsm encodes a non-empty src to a guaranteed-large-enough dst. +// Maximum input 4294967295 bytes. +// It assumes that the varint-encoded length of the decompressed bytes has already been written. +// +//go:noescape +func encodeSnappyBetterBlockAsm(dst []byte, src []byte) int + +// encodeSnappyBetterBlockAsm64K encodes a non-empty src to a guaranteed-large-enough dst. +// Maximum input 65535 bytes. +// It assumes that the varint-encoded length of the decompressed bytes has already been written. +// +//go:noescape +func encodeSnappyBetterBlockAsm64K(dst []byte, src []byte) int + +// encodeSnappyBetterBlockAsm12B encodes a non-empty src to a guaranteed-large-enough dst. +// Maximum input 16383 bytes. +// It assumes that the varint-encoded length of the decompressed bytes has already been written. +// +//go:noescape +func encodeSnappyBetterBlockAsm12B(dst []byte, src []byte) int + +// encodeSnappyBetterBlockAsm10B encodes a non-empty src to a guaranteed-large-enough dst. +// Maximum input 4095 bytes. +// It assumes that the varint-encoded length of the decompressed bytes has already been written. +// +//go:noescape +func encodeSnappyBetterBlockAsm10B(dst []byte, src []byte) int + +// encodeSnappyBetterBlockAsm8B encodes a non-empty src to a guaranteed-large-enough dst. +// Maximum input 511 bytes. +// It assumes that the varint-encoded length of the decompressed bytes has already been written. +// +//go:noescape +func encodeSnappyBetterBlockAsm8B(dst []byte, src []byte) int + +// calcBlockSize encodes a non-empty src to a guaranteed-large-enough dst. +// Maximum input 4294967295 bytes. +// It assumes that the varint-encoded length of the decompressed bytes has already been written. +// +//go:noescape +func calcBlockSize(src []byte) int + +// calcBlockSizeSmall encodes a non-empty src to a guaranteed-large-enough dst. +// Maximum input 1024 bytes. +// It assumes that the varint-encoded length of the decompressed bytes has already been written. +// +//go:noescape +func calcBlockSizeSmall(src []byte) int + +// emitLiteral writes a literal chunk and returns the number of bytes written. +// +// It assumes that: +// +// dst is long enough to hold the encoded bytes with margin of 0 bytes +// 0 <= len(lit) && len(lit) <= math.MaxUint32 +// +//go:noescape +func emitLiteral(dst []byte, lit []byte) int + +// emitRepeat writes a repeat chunk and returns the number of bytes written. +// Length must be at least 4 and < 1<<32 +// +//go:noescape +func emitRepeat(dst []byte, offset int, length int) int + +// emitCopy writes a copy chunk and returns the number of bytes written. +// +// It assumes that: +// +// dst is long enough to hold the encoded bytes +// 1 <= offset && offset <= math.MaxUint32 +// 4 <= length && length <= 1 << 24 +// +//go:noescape +func emitCopy(dst []byte, offset int, length int) int + +// emitCopyNoRepeat writes a copy chunk and returns the number of bytes written. +// +// It assumes that: +// +// dst is long enough to hold the encoded bytes +// 1 <= offset && offset <= math.MaxUint32 +// 4 <= length && length <= 1 << 24 +// +//go:noescape +func emitCopyNoRepeat(dst []byte, offset int, length int) int + +// matchLen returns how many bytes match in a and b +// +// It assumes that: +// +// len(a) <= len(b) +// +//go:noescape +func matchLen(a []byte, b []byte) int + +// cvtLZ4Block converts an LZ4 block to S2 +// +//go:noescape +func cvtLZ4BlockAsm(dst []byte, src []byte) (uncompressed int, dstUsed int) + +// cvtLZ4sBlock converts an LZ4s block to S2 +// +//go:noescape +func cvtLZ4sBlockAsm(dst []byte, src []byte) (uncompressed int, dstUsed int) + +// cvtLZ4Block converts an LZ4 block to Snappy +// +//go:noescape +func cvtLZ4BlockSnappyAsm(dst []byte, src []byte) (uncompressed int, dstUsed int) + +// cvtLZ4sBlock converts an LZ4s block to Snappy +// +//go:noescape +func cvtLZ4sBlockSnappyAsm(dst []byte, src []byte) (uncompressed int, dstUsed int) diff --git a/vendor/github.com/klauspost/compress/s2/encodeblock_amd64.s b/vendor/github.com/klauspost/compress/s2/encodeblock_amd64.s new file mode 100644 index 000000000..2ff5b3340 --- /dev/null +++ b/vendor/github.com/klauspost/compress/s2/encodeblock_amd64.s @@ -0,0 +1,21277 @@ +// Code generated by command: go run gen.go -out ../encodeblock_amd64.s -stubs ../encodeblock_amd64.go -pkg=s2. DO NOT EDIT. + +//go:build !appengine && !noasm && gc && !noasm + +#include "textflag.h" + +// func _dummy_() +TEXT ·_dummy_(SB), $0 +#ifdef GOAMD64_v4 +#ifndef GOAMD64_v3 +#define GOAMD64_v3 +#endif +#endif + RET + +// func encodeBlockAsm(dst []byte, src []byte) int +// Requires: BMI, SSE2 +TEXT ·encodeBlockAsm(SB), $65560-56 + MOVQ dst_base+0(FP), AX + MOVQ $0x00000200, CX + LEAQ 24(SP), DX + PXOR X0, X0 + +zero_loop_encodeBlockAsm: + MOVOU X0, (DX) + MOVOU X0, 16(DX) + MOVOU X0, 32(DX) + MOVOU X0, 48(DX) + MOVOU X0, 64(DX) + MOVOU X0, 80(DX) + MOVOU X0, 96(DX) + MOVOU X0, 112(DX) + ADDQ $0x80, DX + DECQ CX + JNZ zero_loop_encodeBlockAsm + MOVL $0x00000000, 12(SP) + MOVQ src_len+32(FP), CX + LEAQ -9(CX), DX + LEAQ -8(CX), BX + MOVL BX, 8(SP) + SHRQ $0x05, CX + SUBL CX, DX + LEAQ (AX)(DX*1), DX + MOVQ DX, (SP) + MOVL $0x00000001, CX + MOVL CX, 16(SP) + MOVQ src_base+24(FP), DX + +search_loop_encodeBlockAsm: + MOVL CX, BX + SUBL 12(SP), BX + SHRL $0x06, BX + LEAL 4(CX)(BX*1), BX + CMPL BX, 8(SP) + JAE emit_remainder_encodeBlockAsm + MOVQ (DX)(CX*1), SI + MOVL BX, 20(SP) + MOVQ $0x0000cf1bbcdcbf9b, R8 + MOVQ SI, R9 + MOVQ SI, R10 + SHRQ $0x08, R10 + SHLQ $0x10, R9 + IMULQ R8, R9 + SHRQ $0x32, R9 + SHLQ $0x10, R10 + IMULQ R8, R10 + SHRQ $0x32, R10 + MOVL 24(SP)(R9*4), BX + MOVL 24(SP)(R10*4), DI + MOVL CX, 24(SP)(R9*4) + LEAL 1(CX), R9 + MOVL R9, 24(SP)(R10*4) + MOVQ SI, R9 + SHRQ $0x10, R9 + SHLQ $0x10, R9 + IMULQ R8, R9 + SHRQ $0x32, R9 + MOVL CX, R8 + SUBL 16(SP), R8 + MOVL 1(DX)(R8*1), R10 + MOVQ SI, R8 + SHRQ $0x08, R8 + CMPL R8, R10 + JNE no_repeat_found_encodeBlockAsm + LEAL 1(CX), SI + MOVL 12(SP), DI + MOVL SI, BX + SUBL 16(SP), BX + JZ repeat_extend_back_end_encodeBlockAsm + +repeat_extend_back_loop_encodeBlockAsm: + CMPL SI, DI + JBE repeat_extend_back_end_encodeBlockAsm + MOVB -1(DX)(BX*1), R8 + MOVB -1(DX)(SI*1), R9 + CMPB R8, R9 + JNE repeat_extend_back_end_encodeBlockAsm + LEAL -1(SI), SI + DECL BX + JNZ repeat_extend_back_loop_encodeBlockAsm + +repeat_extend_back_end_encodeBlockAsm: + MOVL SI, BX + SUBL 12(SP), BX + LEAQ 5(AX)(BX*1), BX + CMPQ BX, (SP) + JB repeat_dst_size_check_encodeBlockAsm + MOVQ $0x00000000, ret+48(FP) + RET + +repeat_dst_size_check_encodeBlockAsm: + MOVL 12(SP), BX + CMPL BX, SI + JEQ emit_literal_done_repeat_emit_encodeBlockAsm + MOVL SI, R8 + MOVL SI, 12(SP) + LEAQ (DX)(BX*1), R9 + SUBL BX, R8 + LEAL -1(R8), BX + CMPL BX, $0x3c + JB one_byte_repeat_emit_encodeBlockAsm + CMPL BX, $0x00000100 + JB two_bytes_repeat_emit_encodeBlockAsm + CMPL BX, $0x00010000 + JB three_bytes_repeat_emit_encodeBlockAsm + CMPL BX, $0x01000000 + JB four_bytes_repeat_emit_encodeBlockAsm + MOVB $0xfc, (AX) + MOVL BX, 1(AX) + ADDQ $0x05, AX + JMP memmove_long_repeat_emit_encodeBlockAsm + +four_bytes_repeat_emit_encodeBlockAsm: + MOVL BX, R10 + SHRL $0x10, R10 + MOVB $0xf8, (AX) + MOVW BX, 1(AX) + MOVB R10, 3(AX) + ADDQ $0x04, AX + JMP memmove_long_repeat_emit_encodeBlockAsm + +three_bytes_repeat_emit_encodeBlockAsm: + MOVB $0xf4, (AX) + MOVW BX, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_repeat_emit_encodeBlockAsm + +two_bytes_repeat_emit_encodeBlockAsm: + MOVB $0xf0, (AX) + MOVB BL, 1(AX) + ADDQ $0x02, AX + CMPL BX, $0x40 + JB memmove_repeat_emit_encodeBlockAsm + JMP memmove_long_repeat_emit_encodeBlockAsm + +one_byte_repeat_emit_encodeBlockAsm: + SHLB $0x02, BL + MOVB BL, (AX) + ADDQ $0x01, AX + +memmove_repeat_emit_encodeBlockAsm: + LEAQ (AX)(R8*1), BX + + // genMemMoveShort + CMPQ R8, $0x08 + JBE emit_lit_memmove_repeat_emit_encodeBlockAsm_memmove_move_8 + CMPQ R8, $0x10 + JBE emit_lit_memmove_repeat_emit_encodeBlockAsm_memmove_move_8through16 + CMPQ R8, $0x20 + JBE emit_lit_memmove_repeat_emit_encodeBlockAsm_memmove_move_17through32 + JMP emit_lit_memmove_repeat_emit_encodeBlockAsm_memmove_move_33through64 + +emit_lit_memmove_repeat_emit_encodeBlockAsm_memmove_move_8: + MOVQ (R9), R10 + MOVQ R10, (AX) + JMP memmove_end_copy_repeat_emit_encodeBlockAsm + +emit_lit_memmove_repeat_emit_encodeBlockAsm_memmove_move_8through16: + MOVQ (R9), R10 + MOVQ -8(R9)(R8*1), R9 + MOVQ R10, (AX) + MOVQ R9, -8(AX)(R8*1) + JMP memmove_end_copy_repeat_emit_encodeBlockAsm + +emit_lit_memmove_repeat_emit_encodeBlockAsm_memmove_move_17through32: + MOVOU (R9), X0 + MOVOU -16(R9)(R8*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(R8*1) + JMP memmove_end_copy_repeat_emit_encodeBlockAsm + +emit_lit_memmove_repeat_emit_encodeBlockAsm_memmove_move_33through64: + MOVOU (R9), X0 + MOVOU 16(R9), X1 + MOVOU -32(R9)(R8*1), X2 + MOVOU -16(R9)(R8*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + +memmove_end_copy_repeat_emit_encodeBlockAsm: + MOVQ BX, AX + JMP emit_literal_done_repeat_emit_encodeBlockAsm + +memmove_long_repeat_emit_encodeBlockAsm: + LEAQ (AX)(R8*1), BX + + // genMemMoveLong + MOVOU (R9), X0 + MOVOU 16(R9), X1 + MOVOU -32(R9)(R8*1), X2 + MOVOU -16(R9)(R8*1), X3 + MOVQ R8, R11 + SHRQ $0x05, R11 + MOVQ AX, R10 + ANDL $0x0000001f, R10 + MOVQ $0x00000040, R12 + SUBQ R10, R12 + DECQ R11 + JA emit_lit_memmove_long_repeat_emit_encodeBlockAsmlarge_forward_sse_loop_32 + LEAQ -32(R9)(R12*1), R10 + LEAQ -32(AX)(R12*1), R13 + +emit_lit_memmove_long_repeat_emit_encodeBlockAsmlarge_big_loop_back: + MOVOU (R10), X4 + MOVOU 16(R10), X5 + MOVOA X4, (R13) + MOVOA X5, 16(R13) + ADDQ $0x20, R13 + ADDQ $0x20, R10 + ADDQ $0x20, R12 + DECQ R11 + JNA emit_lit_memmove_long_repeat_emit_encodeBlockAsmlarge_big_loop_back + +emit_lit_memmove_long_repeat_emit_encodeBlockAsmlarge_forward_sse_loop_32: + MOVOU -32(R9)(R12*1), X4 + MOVOU -16(R9)(R12*1), X5 + MOVOA X4, -32(AX)(R12*1) + MOVOA X5, -16(AX)(R12*1) + ADDQ $0x20, R12 + CMPQ R8, R12 + JAE emit_lit_memmove_long_repeat_emit_encodeBlockAsmlarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + MOVQ BX, AX + +emit_literal_done_repeat_emit_encodeBlockAsm: + ADDL $0x05, CX + MOVL CX, BX + SUBL 16(SP), BX + MOVQ src_len+32(FP), R8 + SUBL CX, R8 + LEAQ (DX)(CX*1), R9 + LEAQ (DX)(BX*1), BX + + // matchLen + XORL R11, R11 + +matchlen_loopback_16_repeat_extend_encodeBlockAsm: + CMPL R8, $0x10 + JB matchlen_match8_repeat_extend_encodeBlockAsm + MOVQ (R9)(R11*1), R10 + MOVQ 8(R9)(R11*1), R12 + XORQ (BX)(R11*1), R10 + JNZ matchlen_bsf_8_repeat_extend_encodeBlockAsm + XORQ 8(BX)(R11*1), R12 + JNZ matchlen_bsf_16repeat_extend_encodeBlockAsm + LEAL -16(R8), R8 + LEAL 16(R11), R11 + JMP matchlen_loopback_16_repeat_extend_encodeBlockAsm + +matchlen_bsf_16repeat_extend_encodeBlockAsm: +#ifdef GOAMD64_v3 + TZCNTQ R12, R12 + +#else + BSFQ R12, R12 + +#endif + SARQ $0x03, R12 + LEAL 8(R11)(R12*1), R11 + JMP repeat_extend_forward_end_encodeBlockAsm + +matchlen_match8_repeat_extend_encodeBlockAsm: + CMPL R8, $0x08 + JB matchlen_match4_repeat_extend_encodeBlockAsm + MOVQ (R9)(R11*1), R10 + XORQ (BX)(R11*1), R10 + JNZ matchlen_bsf_8_repeat_extend_encodeBlockAsm + LEAL -8(R8), R8 + LEAL 8(R11), R11 + JMP matchlen_match4_repeat_extend_encodeBlockAsm + +matchlen_bsf_8_repeat_extend_encodeBlockAsm: +#ifdef GOAMD64_v3 + TZCNTQ R10, R10 + +#else + BSFQ R10, R10 + +#endif + SARQ $0x03, R10 + LEAL (R11)(R10*1), R11 + JMP repeat_extend_forward_end_encodeBlockAsm + +matchlen_match4_repeat_extend_encodeBlockAsm: + CMPL R8, $0x04 + JB matchlen_match2_repeat_extend_encodeBlockAsm + MOVL (R9)(R11*1), R10 + CMPL (BX)(R11*1), R10 + JNE matchlen_match2_repeat_extend_encodeBlockAsm + LEAL -4(R8), R8 + LEAL 4(R11), R11 + +matchlen_match2_repeat_extend_encodeBlockAsm: + CMPL R8, $0x01 + JE matchlen_match1_repeat_extend_encodeBlockAsm + JB repeat_extend_forward_end_encodeBlockAsm + MOVW (R9)(R11*1), R10 + CMPW (BX)(R11*1), R10 + JNE matchlen_match1_repeat_extend_encodeBlockAsm + LEAL 2(R11), R11 + SUBL $0x02, R8 + JZ repeat_extend_forward_end_encodeBlockAsm + +matchlen_match1_repeat_extend_encodeBlockAsm: + MOVB (R9)(R11*1), R10 + CMPB (BX)(R11*1), R10 + JNE repeat_extend_forward_end_encodeBlockAsm + LEAL 1(R11), R11 + +repeat_extend_forward_end_encodeBlockAsm: + ADDL R11, CX + MOVL CX, BX + SUBL SI, BX + MOVL 16(SP), SI + TESTL DI, DI + JZ repeat_as_copy_encodeBlockAsm + + // emitRepeat +emit_repeat_again_match_repeat_encodeBlockAsm: + MOVL BX, DI + LEAL -4(BX), BX + CMPL DI, $0x08 + JBE repeat_two_match_repeat_encodeBlockAsm + CMPL DI, $0x0c + JAE cant_repeat_two_offset_match_repeat_encodeBlockAsm + CMPL SI, $0x00000800 + JB repeat_two_offset_match_repeat_encodeBlockAsm + +cant_repeat_two_offset_match_repeat_encodeBlockAsm: + CMPL BX, $0x00000104 + JB repeat_three_match_repeat_encodeBlockAsm + CMPL BX, $0x00010100 + JB repeat_four_match_repeat_encodeBlockAsm + CMPL BX, $0x0100ffff + JB repeat_five_match_repeat_encodeBlockAsm + LEAL -16842747(BX), BX + MOVL $0xfffb001d, (AX) + MOVB $0xff, 4(AX) + ADDQ $0x05, AX + JMP emit_repeat_again_match_repeat_encodeBlockAsm + +repeat_five_match_repeat_encodeBlockAsm: + LEAL -65536(BX), BX + MOVL BX, SI + MOVW $0x001d, (AX) + MOVW BX, 2(AX) + SARL $0x10, SI + MOVB SI, 4(AX) + ADDQ $0x05, AX + JMP repeat_end_emit_encodeBlockAsm + +repeat_four_match_repeat_encodeBlockAsm: + LEAL -256(BX), BX + MOVW $0x0019, (AX) + MOVW BX, 2(AX) + ADDQ $0x04, AX + JMP repeat_end_emit_encodeBlockAsm + +repeat_three_match_repeat_encodeBlockAsm: + LEAL -4(BX), BX + MOVW $0x0015, (AX) + MOVB BL, 2(AX) + ADDQ $0x03, AX + JMP repeat_end_emit_encodeBlockAsm + +repeat_two_match_repeat_encodeBlockAsm: + SHLL $0x02, BX + ORL $0x01, BX + MOVW BX, (AX) + ADDQ $0x02, AX + JMP repeat_end_emit_encodeBlockAsm + +repeat_two_offset_match_repeat_encodeBlockAsm: + XORQ DI, DI + LEAL 1(DI)(BX*4), BX + MOVB SI, 1(AX) + SARL $0x08, SI + SHLL $0x05, SI + ORL SI, BX + MOVB BL, (AX) + ADDQ $0x02, AX + JMP repeat_end_emit_encodeBlockAsm + +repeat_as_copy_encodeBlockAsm: + // emitCopy + CMPL SI, $0x00010000 + JB two_byte_offset_repeat_as_copy_encodeBlockAsm + CMPL BX, $0x40 + JBE four_bytes_remain_repeat_as_copy_encodeBlockAsm + MOVB $0xff, (AX) + MOVL SI, 1(AX) + LEAL -64(BX), BX + ADDQ $0x05, AX + CMPL BX, $0x04 + JB four_bytes_remain_repeat_as_copy_encodeBlockAsm + + // emitRepeat +emit_repeat_again_repeat_as_copy_encodeBlockAsm_emit_copy: + MOVL BX, DI + LEAL -4(BX), BX + CMPL DI, $0x08 + JBE repeat_two_repeat_as_copy_encodeBlockAsm_emit_copy + CMPL DI, $0x0c + JAE cant_repeat_two_offset_repeat_as_copy_encodeBlockAsm_emit_copy + CMPL SI, $0x00000800 + JB repeat_two_offset_repeat_as_copy_encodeBlockAsm_emit_copy + +cant_repeat_two_offset_repeat_as_copy_encodeBlockAsm_emit_copy: + CMPL BX, $0x00000104 + JB repeat_three_repeat_as_copy_encodeBlockAsm_emit_copy + CMPL BX, $0x00010100 + JB repeat_four_repeat_as_copy_encodeBlockAsm_emit_copy + CMPL BX, $0x0100ffff + JB repeat_five_repeat_as_copy_encodeBlockAsm_emit_copy + LEAL -16842747(BX), BX + MOVL $0xfffb001d, (AX) + MOVB $0xff, 4(AX) + ADDQ $0x05, AX + JMP emit_repeat_again_repeat_as_copy_encodeBlockAsm_emit_copy + +repeat_five_repeat_as_copy_encodeBlockAsm_emit_copy: + LEAL -65536(BX), BX + MOVL BX, SI + MOVW $0x001d, (AX) + MOVW BX, 2(AX) + SARL $0x10, SI + MOVB SI, 4(AX) + ADDQ $0x05, AX + JMP repeat_end_emit_encodeBlockAsm + +repeat_four_repeat_as_copy_encodeBlockAsm_emit_copy: + LEAL -256(BX), BX + MOVW $0x0019, (AX) + MOVW BX, 2(AX) + ADDQ $0x04, AX + JMP repeat_end_emit_encodeBlockAsm + +repeat_three_repeat_as_copy_encodeBlockAsm_emit_copy: + LEAL -4(BX), BX + MOVW $0x0015, (AX) + MOVB BL, 2(AX) + ADDQ $0x03, AX + JMP repeat_end_emit_encodeBlockAsm + +repeat_two_repeat_as_copy_encodeBlockAsm_emit_copy: + SHLL $0x02, BX + ORL $0x01, BX + MOVW BX, (AX) + ADDQ $0x02, AX + JMP repeat_end_emit_encodeBlockAsm + +repeat_two_offset_repeat_as_copy_encodeBlockAsm_emit_copy: + XORQ DI, DI + LEAL 1(DI)(BX*4), BX + MOVB SI, 1(AX) + SARL $0x08, SI + SHLL $0x05, SI + ORL SI, BX + MOVB BL, (AX) + ADDQ $0x02, AX + JMP repeat_end_emit_encodeBlockAsm + +four_bytes_remain_repeat_as_copy_encodeBlockAsm: + TESTL BX, BX + JZ repeat_end_emit_encodeBlockAsm + XORL DI, DI + LEAL -1(DI)(BX*4), BX + MOVB BL, (AX) + MOVL SI, 1(AX) + ADDQ $0x05, AX + JMP repeat_end_emit_encodeBlockAsm + +two_byte_offset_repeat_as_copy_encodeBlockAsm: + CMPL BX, $0x40 + JBE two_byte_offset_short_repeat_as_copy_encodeBlockAsm + CMPL SI, $0x00000800 + JAE long_offset_short_repeat_as_copy_encodeBlockAsm + MOVL $0x00000001, DI + LEAL 16(DI), DI + MOVB SI, 1(AX) + MOVL SI, R8 + SHRL $0x08, R8 + SHLL $0x05, R8 + ORL R8, DI + MOVB DI, (AX) + ADDQ $0x02, AX + SUBL $0x08, BX + + // emitRepeat + LEAL -4(BX), BX + JMP cant_repeat_two_offset_repeat_as_copy_encodeBlockAsm_emit_copy_short_2b + +emit_repeat_again_repeat_as_copy_encodeBlockAsm_emit_copy_short_2b: + MOVL BX, DI + LEAL -4(BX), BX + CMPL DI, $0x08 + JBE repeat_two_repeat_as_copy_encodeBlockAsm_emit_copy_short_2b + CMPL DI, $0x0c + JAE cant_repeat_two_offset_repeat_as_copy_encodeBlockAsm_emit_copy_short_2b + CMPL SI, $0x00000800 + JB repeat_two_offset_repeat_as_copy_encodeBlockAsm_emit_copy_short_2b + +cant_repeat_two_offset_repeat_as_copy_encodeBlockAsm_emit_copy_short_2b: + CMPL BX, $0x00000104 + JB repeat_three_repeat_as_copy_encodeBlockAsm_emit_copy_short_2b + CMPL BX, $0x00010100 + JB repeat_four_repeat_as_copy_encodeBlockAsm_emit_copy_short_2b + CMPL BX, $0x0100ffff + JB repeat_five_repeat_as_copy_encodeBlockAsm_emit_copy_short_2b + LEAL -16842747(BX), BX + MOVL $0xfffb001d, (AX) + MOVB $0xff, 4(AX) + ADDQ $0x05, AX + JMP emit_repeat_again_repeat_as_copy_encodeBlockAsm_emit_copy_short_2b + +repeat_five_repeat_as_copy_encodeBlockAsm_emit_copy_short_2b: + LEAL -65536(BX), BX + MOVL BX, SI + MOVW $0x001d, (AX) + MOVW BX, 2(AX) + SARL $0x10, SI + MOVB SI, 4(AX) + ADDQ $0x05, AX + JMP repeat_end_emit_encodeBlockAsm + +repeat_four_repeat_as_copy_encodeBlockAsm_emit_copy_short_2b: + LEAL -256(BX), BX + MOVW $0x0019, (AX) + MOVW BX, 2(AX) + ADDQ $0x04, AX + JMP repeat_end_emit_encodeBlockAsm + +repeat_three_repeat_as_copy_encodeBlockAsm_emit_copy_short_2b: + LEAL -4(BX), BX + MOVW $0x0015, (AX) + MOVB BL, 2(AX) + ADDQ $0x03, AX + JMP repeat_end_emit_encodeBlockAsm + +repeat_two_repeat_as_copy_encodeBlockAsm_emit_copy_short_2b: + SHLL $0x02, BX + ORL $0x01, BX + MOVW BX, (AX) + ADDQ $0x02, AX + JMP repeat_end_emit_encodeBlockAsm + +repeat_two_offset_repeat_as_copy_encodeBlockAsm_emit_copy_short_2b: + XORQ DI, DI + LEAL 1(DI)(BX*4), BX + MOVB SI, 1(AX) + SARL $0x08, SI + SHLL $0x05, SI + ORL SI, BX + MOVB BL, (AX) + ADDQ $0x02, AX + JMP repeat_end_emit_encodeBlockAsm + +long_offset_short_repeat_as_copy_encodeBlockAsm: + MOVB $0xee, (AX) + MOVW SI, 1(AX) + LEAL -60(BX), BX + ADDQ $0x03, AX + + // emitRepeat +emit_repeat_again_repeat_as_copy_encodeBlockAsm_emit_copy_short: + MOVL BX, DI + LEAL -4(BX), BX + CMPL DI, $0x08 + JBE repeat_two_repeat_as_copy_encodeBlockAsm_emit_copy_short + CMPL DI, $0x0c + JAE cant_repeat_two_offset_repeat_as_copy_encodeBlockAsm_emit_copy_short + CMPL SI, $0x00000800 + JB repeat_two_offset_repeat_as_copy_encodeBlockAsm_emit_copy_short + +cant_repeat_two_offset_repeat_as_copy_encodeBlockAsm_emit_copy_short: + CMPL BX, $0x00000104 + JB repeat_three_repeat_as_copy_encodeBlockAsm_emit_copy_short + CMPL BX, $0x00010100 + JB repeat_four_repeat_as_copy_encodeBlockAsm_emit_copy_short + CMPL BX, $0x0100ffff + JB repeat_five_repeat_as_copy_encodeBlockAsm_emit_copy_short + LEAL -16842747(BX), BX + MOVL $0xfffb001d, (AX) + MOVB $0xff, 4(AX) + ADDQ $0x05, AX + JMP emit_repeat_again_repeat_as_copy_encodeBlockAsm_emit_copy_short + +repeat_five_repeat_as_copy_encodeBlockAsm_emit_copy_short: + LEAL -65536(BX), BX + MOVL BX, SI + MOVW $0x001d, (AX) + MOVW BX, 2(AX) + SARL $0x10, SI + MOVB SI, 4(AX) + ADDQ $0x05, AX + JMP repeat_end_emit_encodeBlockAsm + +repeat_four_repeat_as_copy_encodeBlockAsm_emit_copy_short: + LEAL -256(BX), BX + MOVW $0x0019, (AX) + MOVW BX, 2(AX) + ADDQ $0x04, AX + JMP repeat_end_emit_encodeBlockAsm + +repeat_three_repeat_as_copy_encodeBlockAsm_emit_copy_short: + LEAL -4(BX), BX + MOVW $0x0015, (AX) + MOVB BL, 2(AX) + ADDQ $0x03, AX + JMP repeat_end_emit_encodeBlockAsm + +repeat_two_repeat_as_copy_encodeBlockAsm_emit_copy_short: + SHLL $0x02, BX + ORL $0x01, BX + MOVW BX, (AX) + ADDQ $0x02, AX + JMP repeat_end_emit_encodeBlockAsm + +repeat_two_offset_repeat_as_copy_encodeBlockAsm_emit_copy_short: + XORQ DI, DI + LEAL 1(DI)(BX*4), BX + MOVB SI, 1(AX) + SARL $0x08, SI + SHLL $0x05, SI + ORL SI, BX + MOVB BL, (AX) + ADDQ $0x02, AX + JMP repeat_end_emit_encodeBlockAsm + +two_byte_offset_short_repeat_as_copy_encodeBlockAsm: + MOVL BX, DI + SHLL $0x02, DI + CMPL BX, $0x0c + JAE emit_copy_three_repeat_as_copy_encodeBlockAsm + CMPL SI, $0x00000800 + JAE emit_copy_three_repeat_as_copy_encodeBlockAsm + LEAL -15(DI), DI + MOVB SI, 1(AX) + SHRL $0x08, SI + SHLL $0x05, SI + ORL SI, DI + MOVB DI, (AX) + ADDQ $0x02, AX + JMP repeat_end_emit_encodeBlockAsm + +emit_copy_three_repeat_as_copy_encodeBlockAsm: + LEAL -2(DI), DI + MOVB DI, (AX) + MOVW SI, 1(AX) + ADDQ $0x03, AX + +repeat_end_emit_encodeBlockAsm: + MOVL CX, 12(SP) + JMP search_loop_encodeBlockAsm + +no_repeat_found_encodeBlockAsm: + CMPL (DX)(BX*1), SI + JEQ candidate_match_encodeBlockAsm + SHRQ $0x08, SI + MOVL 24(SP)(R9*4), BX + LEAL 2(CX), R8 + CMPL (DX)(DI*1), SI + JEQ candidate2_match_encodeBlockAsm + MOVL R8, 24(SP)(R9*4) + SHRQ $0x08, SI + CMPL (DX)(BX*1), SI + JEQ candidate3_match_encodeBlockAsm + MOVL 20(SP), CX + JMP search_loop_encodeBlockAsm + +candidate3_match_encodeBlockAsm: + ADDL $0x02, CX + JMP candidate_match_encodeBlockAsm + +candidate2_match_encodeBlockAsm: + MOVL R8, 24(SP)(R9*4) + INCL CX + MOVL DI, BX + +candidate_match_encodeBlockAsm: + MOVL 12(SP), SI + TESTL BX, BX + JZ match_extend_back_end_encodeBlockAsm + +match_extend_back_loop_encodeBlockAsm: + CMPL CX, SI + JBE match_extend_back_end_encodeBlockAsm + MOVB -1(DX)(BX*1), DI + MOVB -1(DX)(CX*1), R8 + CMPB DI, R8 + JNE match_extend_back_end_encodeBlockAsm + LEAL -1(CX), CX + DECL BX + JZ match_extend_back_end_encodeBlockAsm + JMP match_extend_back_loop_encodeBlockAsm + +match_extend_back_end_encodeBlockAsm: + MOVL CX, SI + SUBL 12(SP), SI + LEAQ 5(AX)(SI*1), SI + CMPQ SI, (SP) + JB match_dst_size_check_encodeBlockAsm + MOVQ $0x00000000, ret+48(FP) + RET + +match_dst_size_check_encodeBlockAsm: + MOVL CX, SI + MOVL 12(SP), DI + CMPL DI, SI + JEQ emit_literal_done_match_emit_encodeBlockAsm + MOVL SI, R8 + MOVL SI, 12(SP) + LEAQ (DX)(DI*1), SI + SUBL DI, R8 + LEAL -1(R8), DI + CMPL DI, $0x3c + JB one_byte_match_emit_encodeBlockAsm + CMPL DI, $0x00000100 + JB two_bytes_match_emit_encodeBlockAsm + CMPL DI, $0x00010000 + JB three_bytes_match_emit_encodeBlockAsm + CMPL DI, $0x01000000 + JB four_bytes_match_emit_encodeBlockAsm + MOVB $0xfc, (AX) + MOVL DI, 1(AX) + ADDQ $0x05, AX + JMP memmove_long_match_emit_encodeBlockAsm + +four_bytes_match_emit_encodeBlockAsm: + MOVL DI, R9 + SHRL $0x10, R9 + MOVB $0xf8, (AX) + MOVW DI, 1(AX) + MOVB R9, 3(AX) + ADDQ $0x04, AX + JMP memmove_long_match_emit_encodeBlockAsm + +three_bytes_match_emit_encodeBlockAsm: + MOVB $0xf4, (AX) + MOVW DI, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_match_emit_encodeBlockAsm + +two_bytes_match_emit_encodeBlockAsm: + MOVB $0xf0, (AX) + MOVB DI, 1(AX) + ADDQ $0x02, AX + CMPL DI, $0x40 + JB memmove_match_emit_encodeBlockAsm + JMP memmove_long_match_emit_encodeBlockAsm + +one_byte_match_emit_encodeBlockAsm: + SHLB $0x02, DI + MOVB DI, (AX) + ADDQ $0x01, AX + +memmove_match_emit_encodeBlockAsm: + LEAQ (AX)(R8*1), DI + + // genMemMoveShort + CMPQ R8, $0x08 + JBE emit_lit_memmove_match_emit_encodeBlockAsm_memmove_move_8 + CMPQ R8, $0x10 + JBE emit_lit_memmove_match_emit_encodeBlockAsm_memmove_move_8through16 + CMPQ R8, $0x20 + JBE emit_lit_memmove_match_emit_encodeBlockAsm_memmove_move_17through32 + JMP emit_lit_memmove_match_emit_encodeBlockAsm_memmove_move_33through64 + +emit_lit_memmove_match_emit_encodeBlockAsm_memmove_move_8: + MOVQ (SI), R9 + MOVQ R9, (AX) + JMP memmove_end_copy_match_emit_encodeBlockAsm + +emit_lit_memmove_match_emit_encodeBlockAsm_memmove_move_8through16: + MOVQ (SI), R9 + MOVQ -8(SI)(R8*1), SI + MOVQ R9, (AX) + MOVQ SI, -8(AX)(R8*1) + JMP memmove_end_copy_match_emit_encodeBlockAsm + +emit_lit_memmove_match_emit_encodeBlockAsm_memmove_move_17through32: + MOVOU (SI), X0 + MOVOU -16(SI)(R8*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(R8*1) + JMP memmove_end_copy_match_emit_encodeBlockAsm + +emit_lit_memmove_match_emit_encodeBlockAsm_memmove_move_33through64: + MOVOU (SI), X0 + MOVOU 16(SI), X1 + MOVOU -32(SI)(R8*1), X2 + MOVOU -16(SI)(R8*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + +memmove_end_copy_match_emit_encodeBlockAsm: + MOVQ DI, AX + JMP emit_literal_done_match_emit_encodeBlockAsm + +memmove_long_match_emit_encodeBlockAsm: + LEAQ (AX)(R8*1), DI + + // genMemMoveLong + MOVOU (SI), X0 + MOVOU 16(SI), X1 + MOVOU -32(SI)(R8*1), X2 + MOVOU -16(SI)(R8*1), X3 + MOVQ R8, R10 + SHRQ $0x05, R10 + MOVQ AX, R9 + ANDL $0x0000001f, R9 + MOVQ $0x00000040, R11 + SUBQ R9, R11 + DECQ R10 + JA emit_lit_memmove_long_match_emit_encodeBlockAsmlarge_forward_sse_loop_32 + LEAQ -32(SI)(R11*1), R9 + LEAQ -32(AX)(R11*1), R12 + +emit_lit_memmove_long_match_emit_encodeBlockAsmlarge_big_loop_back: + MOVOU (R9), X4 + MOVOU 16(R9), X5 + MOVOA X4, (R12) + MOVOA X5, 16(R12) + ADDQ $0x20, R12 + ADDQ $0x20, R9 + ADDQ $0x20, R11 + DECQ R10 + JNA emit_lit_memmove_long_match_emit_encodeBlockAsmlarge_big_loop_back + +emit_lit_memmove_long_match_emit_encodeBlockAsmlarge_forward_sse_loop_32: + MOVOU -32(SI)(R11*1), X4 + MOVOU -16(SI)(R11*1), X5 + MOVOA X4, -32(AX)(R11*1) + MOVOA X5, -16(AX)(R11*1) + ADDQ $0x20, R11 + CMPQ R8, R11 + JAE emit_lit_memmove_long_match_emit_encodeBlockAsmlarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + MOVQ DI, AX + +emit_literal_done_match_emit_encodeBlockAsm: +match_nolit_loop_encodeBlockAsm: + MOVL CX, SI + SUBL BX, SI + MOVL SI, 16(SP) + ADDL $0x04, CX + ADDL $0x04, BX + MOVQ src_len+32(FP), SI + SUBL CX, SI + LEAQ (DX)(CX*1), DI + LEAQ (DX)(BX*1), BX + + // matchLen + XORL R9, R9 + +matchlen_loopback_16_match_nolit_encodeBlockAsm: + CMPL SI, $0x10 + JB matchlen_match8_match_nolit_encodeBlockAsm + MOVQ (DI)(R9*1), R8 + MOVQ 8(DI)(R9*1), R10 + XORQ (BX)(R9*1), R8 + JNZ matchlen_bsf_8_match_nolit_encodeBlockAsm + XORQ 8(BX)(R9*1), R10 + JNZ matchlen_bsf_16match_nolit_encodeBlockAsm + LEAL -16(SI), SI + LEAL 16(R9), R9 + JMP matchlen_loopback_16_match_nolit_encodeBlockAsm + +matchlen_bsf_16match_nolit_encodeBlockAsm: +#ifdef GOAMD64_v3 + TZCNTQ R10, R10 + +#else + BSFQ R10, R10 + +#endif + SARQ $0x03, R10 + LEAL 8(R9)(R10*1), R9 + JMP match_nolit_end_encodeBlockAsm + +matchlen_match8_match_nolit_encodeBlockAsm: + CMPL SI, $0x08 + JB matchlen_match4_match_nolit_encodeBlockAsm + MOVQ (DI)(R9*1), R8 + XORQ (BX)(R9*1), R8 + JNZ matchlen_bsf_8_match_nolit_encodeBlockAsm + LEAL -8(SI), SI + LEAL 8(R9), R9 + JMP matchlen_match4_match_nolit_encodeBlockAsm + +matchlen_bsf_8_match_nolit_encodeBlockAsm: +#ifdef GOAMD64_v3 + TZCNTQ R8, R8 + +#else + BSFQ R8, R8 + +#endif + SARQ $0x03, R8 + LEAL (R9)(R8*1), R9 + JMP match_nolit_end_encodeBlockAsm + +matchlen_match4_match_nolit_encodeBlockAsm: + CMPL SI, $0x04 + JB matchlen_match2_match_nolit_encodeBlockAsm + MOVL (DI)(R9*1), R8 + CMPL (BX)(R9*1), R8 + JNE matchlen_match2_match_nolit_encodeBlockAsm + LEAL -4(SI), SI + LEAL 4(R9), R9 + +matchlen_match2_match_nolit_encodeBlockAsm: + CMPL SI, $0x01 + JE matchlen_match1_match_nolit_encodeBlockAsm + JB match_nolit_end_encodeBlockAsm + MOVW (DI)(R9*1), R8 + CMPW (BX)(R9*1), R8 + JNE matchlen_match1_match_nolit_encodeBlockAsm + LEAL 2(R9), R9 + SUBL $0x02, SI + JZ match_nolit_end_encodeBlockAsm + +matchlen_match1_match_nolit_encodeBlockAsm: + MOVB (DI)(R9*1), R8 + CMPB (BX)(R9*1), R8 + JNE match_nolit_end_encodeBlockAsm + LEAL 1(R9), R9 + +match_nolit_end_encodeBlockAsm: + ADDL R9, CX + MOVL 16(SP), BX + ADDL $0x04, R9 + MOVL CX, 12(SP) + + // emitCopy + CMPL BX, $0x00010000 + JB two_byte_offset_match_nolit_encodeBlockAsm + CMPL R9, $0x40 + JBE four_bytes_remain_match_nolit_encodeBlockAsm + MOVB $0xff, (AX) + MOVL BX, 1(AX) + LEAL -64(R9), R9 + ADDQ $0x05, AX + CMPL R9, $0x04 + JB four_bytes_remain_match_nolit_encodeBlockAsm + + // emitRepeat +emit_repeat_again_match_nolit_encodeBlockAsm_emit_copy: + MOVL R9, SI + LEAL -4(R9), R9 + CMPL SI, $0x08 + JBE repeat_two_match_nolit_encodeBlockAsm_emit_copy + CMPL SI, $0x0c + JAE cant_repeat_two_offset_match_nolit_encodeBlockAsm_emit_copy + CMPL BX, $0x00000800 + JB repeat_two_offset_match_nolit_encodeBlockAsm_emit_copy + +cant_repeat_two_offset_match_nolit_encodeBlockAsm_emit_copy: + CMPL R9, $0x00000104 + JB repeat_three_match_nolit_encodeBlockAsm_emit_copy + CMPL R9, $0x00010100 + JB repeat_four_match_nolit_encodeBlockAsm_emit_copy + CMPL R9, $0x0100ffff + JB repeat_five_match_nolit_encodeBlockAsm_emit_copy + LEAL -16842747(R9), R9 + MOVL $0xfffb001d, (AX) + MOVB $0xff, 4(AX) + ADDQ $0x05, AX + JMP emit_repeat_again_match_nolit_encodeBlockAsm_emit_copy + +repeat_five_match_nolit_encodeBlockAsm_emit_copy: + LEAL -65536(R9), R9 + MOVL R9, BX + MOVW $0x001d, (AX) + MOVW R9, 2(AX) + SARL $0x10, BX + MOVB BL, 4(AX) + ADDQ $0x05, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm + +repeat_four_match_nolit_encodeBlockAsm_emit_copy: + LEAL -256(R9), R9 + MOVW $0x0019, (AX) + MOVW R9, 2(AX) + ADDQ $0x04, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm + +repeat_three_match_nolit_encodeBlockAsm_emit_copy: + LEAL -4(R9), R9 + MOVW $0x0015, (AX) + MOVB R9, 2(AX) + ADDQ $0x03, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm + +repeat_two_match_nolit_encodeBlockAsm_emit_copy: + SHLL $0x02, R9 + ORL $0x01, R9 + MOVW R9, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm + +repeat_two_offset_match_nolit_encodeBlockAsm_emit_copy: + XORQ SI, SI + LEAL 1(SI)(R9*4), R9 + MOVB BL, 1(AX) + SARL $0x08, BX + SHLL $0x05, BX + ORL BX, R9 + MOVB R9, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm + +four_bytes_remain_match_nolit_encodeBlockAsm: + TESTL R9, R9 + JZ match_nolit_emitcopy_end_encodeBlockAsm + XORL SI, SI + LEAL -1(SI)(R9*4), R9 + MOVB R9, (AX) + MOVL BX, 1(AX) + ADDQ $0x05, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm + +two_byte_offset_match_nolit_encodeBlockAsm: + CMPL R9, $0x40 + JBE two_byte_offset_short_match_nolit_encodeBlockAsm + CMPL BX, $0x00000800 + JAE long_offset_short_match_nolit_encodeBlockAsm + MOVL $0x00000001, SI + LEAL 16(SI), SI + MOVB BL, 1(AX) + MOVL BX, DI + SHRL $0x08, DI + SHLL $0x05, DI + ORL DI, SI + MOVB SI, (AX) + ADDQ $0x02, AX + SUBL $0x08, R9 + + // emitRepeat + LEAL -4(R9), R9 + JMP cant_repeat_two_offset_match_nolit_encodeBlockAsm_emit_copy_short_2b + +emit_repeat_again_match_nolit_encodeBlockAsm_emit_copy_short_2b: + MOVL R9, SI + LEAL -4(R9), R9 + CMPL SI, $0x08 + JBE repeat_two_match_nolit_encodeBlockAsm_emit_copy_short_2b + CMPL SI, $0x0c + JAE cant_repeat_two_offset_match_nolit_encodeBlockAsm_emit_copy_short_2b + CMPL BX, $0x00000800 + JB repeat_two_offset_match_nolit_encodeBlockAsm_emit_copy_short_2b + +cant_repeat_two_offset_match_nolit_encodeBlockAsm_emit_copy_short_2b: + CMPL R9, $0x00000104 + JB repeat_three_match_nolit_encodeBlockAsm_emit_copy_short_2b + CMPL R9, $0x00010100 + JB repeat_four_match_nolit_encodeBlockAsm_emit_copy_short_2b + CMPL R9, $0x0100ffff + JB repeat_five_match_nolit_encodeBlockAsm_emit_copy_short_2b + LEAL -16842747(R9), R9 + MOVL $0xfffb001d, (AX) + MOVB $0xff, 4(AX) + ADDQ $0x05, AX + JMP emit_repeat_again_match_nolit_encodeBlockAsm_emit_copy_short_2b + +repeat_five_match_nolit_encodeBlockAsm_emit_copy_short_2b: + LEAL -65536(R9), R9 + MOVL R9, BX + MOVW $0x001d, (AX) + MOVW R9, 2(AX) + SARL $0x10, BX + MOVB BL, 4(AX) + ADDQ $0x05, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm + +repeat_four_match_nolit_encodeBlockAsm_emit_copy_short_2b: + LEAL -256(R9), R9 + MOVW $0x0019, (AX) + MOVW R9, 2(AX) + ADDQ $0x04, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm + +repeat_three_match_nolit_encodeBlockAsm_emit_copy_short_2b: + LEAL -4(R9), R9 + MOVW $0x0015, (AX) + MOVB R9, 2(AX) + ADDQ $0x03, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm + +repeat_two_match_nolit_encodeBlockAsm_emit_copy_short_2b: + SHLL $0x02, R9 + ORL $0x01, R9 + MOVW R9, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm + +repeat_two_offset_match_nolit_encodeBlockAsm_emit_copy_short_2b: + XORQ SI, SI + LEAL 1(SI)(R9*4), R9 + MOVB BL, 1(AX) + SARL $0x08, BX + SHLL $0x05, BX + ORL BX, R9 + MOVB R9, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm + +long_offset_short_match_nolit_encodeBlockAsm: + MOVB $0xee, (AX) + MOVW BX, 1(AX) + LEAL -60(R9), R9 + ADDQ $0x03, AX + + // emitRepeat +emit_repeat_again_match_nolit_encodeBlockAsm_emit_copy_short: + MOVL R9, SI + LEAL -4(R9), R9 + CMPL SI, $0x08 + JBE repeat_two_match_nolit_encodeBlockAsm_emit_copy_short + CMPL SI, $0x0c + JAE cant_repeat_two_offset_match_nolit_encodeBlockAsm_emit_copy_short + CMPL BX, $0x00000800 + JB repeat_two_offset_match_nolit_encodeBlockAsm_emit_copy_short + +cant_repeat_two_offset_match_nolit_encodeBlockAsm_emit_copy_short: + CMPL R9, $0x00000104 + JB repeat_three_match_nolit_encodeBlockAsm_emit_copy_short + CMPL R9, $0x00010100 + JB repeat_four_match_nolit_encodeBlockAsm_emit_copy_short + CMPL R9, $0x0100ffff + JB repeat_five_match_nolit_encodeBlockAsm_emit_copy_short + LEAL -16842747(R9), R9 + MOVL $0xfffb001d, (AX) + MOVB $0xff, 4(AX) + ADDQ $0x05, AX + JMP emit_repeat_again_match_nolit_encodeBlockAsm_emit_copy_short + +repeat_five_match_nolit_encodeBlockAsm_emit_copy_short: + LEAL -65536(R9), R9 + MOVL R9, BX + MOVW $0x001d, (AX) + MOVW R9, 2(AX) + SARL $0x10, BX + MOVB BL, 4(AX) + ADDQ $0x05, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm + +repeat_four_match_nolit_encodeBlockAsm_emit_copy_short: + LEAL -256(R9), R9 + MOVW $0x0019, (AX) + MOVW R9, 2(AX) + ADDQ $0x04, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm + +repeat_three_match_nolit_encodeBlockAsm_emit_copy_short: + LEAL -4(R9), R9 + MOVW $0x0015, (AX) + MOVB R9, 2(AX) + ADDQ $0x03, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm + +repeat_two_match_nolit_encodeBlockAsm_emit_copy_short: + SHLL $0x02, R9 + ORL $0x01, R9 + MOVW R9, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm + +repeat_two_offset_match_nolit_encodeBlockAsm_emit_copy_short: + XORQ SI, SI + LEAL 1(SI)(R9*4), R9 + MOVB BL, 1(AX) + SARL $0x08, BX + SHLL $0x05, BX + ORL BX, R9 + MOVB R9, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm + +two_byte_offset_short_match_nolit_encodeBlockAsm: + MOVL R9, SI + SHLL $0x02, SI + CMPL R9, $0x0c + JAE emit_copy_three_match_nolit_encodeBlockAsm + CMPL BX, $0x00000800 + JAE emit_copy_three_match_nolit_encodeBlockAsm + LEAL -15(SI), SI + MOVB BL, 1(AX) + SHRL $0x08, BX + SHLL $0x05, BX + ORL BX, SI + MOVB SI, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm + +emit_copy_three_match_nolit_encodeBlockAsm: + LEAL -2(SI), SI + MOVB SI, (AX) + MOVW BX, 1(AX) + ADDQ $0x03, AX + +match_nolit_emitcopy_end_encodeBlockAsm: + CMPL CX, 8(SP) + JAE emit_remainder_encodeBlockAsm + MOVQ -2(DX)(CX*1), SI + CMPQ AX, (SP) + JB match_nolit_dst_ok_encodeBlockAsm + MOVQ $0x00000000, ret+48(FP) + RET + +match_nolit_dst_ok_encodeBlockAsm: + MOVQ $0x0000cf1bbcdcbf9b, R8 + MOVQ SI, DI + SHRQ $0x10, SI + MOVQ SI, BX + SHLQ $0x10, DI + IMULQ R8, DI + SHRQ $0x32, DI + SHLQ $0x10, BX + IMULQ R8, BX + SHRQ $0x32, BX + LEAL -2(CX), R8 + LEAQ 24(SP)(BX*4), R9 + MOVL (R9), BX + MOVL R8, 24(SP)(DI*4) + MOVL CX, (R9) + CMPL (DX)(BX*1), SI + JEQ match_nolit_loop_encodeBlockAsm + INCL CX + JMP search_loop_encodeBlockAsm + +emit_remainder_encodeBlockAsm: + MOVQ src_len+32(FP), CX + SUBL 12(SP), CX + LEAQ 5(AX)(CX*1), CX + CMPQ CX, (SP) + JB emit_remainder_ok_encodeBlockAsm + MOVQ $0x00000000, ret+48(FP) + RET + +emit_remainder_ok_encodeBlockAsm: + MOVQ src_len+32(FP), CX + MOVL 12(SP), BX + CMPL BX, CX + JEQ emit_literal_done_emit_remainder_encodeBlockAsm + MOVL CX, SI + MOVL CX, 12(SP) + LEAQ (DX)(BX*1), CX + SUBL BX, SI + LEAL -1(SI), DX + CMPL DX, $0x3c + JB one_byte_emit_remainder_encodeBlockAsm + CMPL DX, $0x00000100 + JB two_bytes_emit_remainder_encodeBlockAsm + CMPL DX, $0x00010000 + JB three_bytes_emit_remainder_encodeBlockAsm + CMPL DX, $0x01000000 + JB four_bytes_emit_remainder_encodeBlockAsm + MOVB $0xfc, (AX) + MOVL DX, 1(AX) + ADDQ $0x05, AX + JMP memmove_long_emit_remainder_encodeBlockAsm + +four_bytes_emit_remainder_encodeBlockAsm: + MOVL DX, BX + SHRL $0x10, BX + MOVB $0xf8, (AX) + MOVW DX, 1(AX) + MOVB BL, 3(AX) + ADDQ $0x04, AX + JMP memmove_long_emit_remainder_encodeBlockAsm + +three_bytes_emit_remainder_encodeBlockAsm: + MOVB $0xf4, (AX) + MOVW DX, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_emit_remainder_encodeBlockAsm + +two_bytes_emit_remainder_encodeBlockAsm: + MOVB $0xf0, (AX) + MOVB DL, 1(AX) + ADDQ $0x02, AX + CMPL DX, $0x40 + JB memmove_emit_remainder_encodeBlockAsm + JMP memmove_long_emit_remainder_encodeBlockAsm + +one_byte_emit_remainder_encodeBlockAsm: + SHLB $0x02, DL + MOVB DL, (AX) + ADDQ $0x01, AX + +memmove_emit_remainder_encodeBlockAsm: + LEAQ (AX)(SI*1), DX + MOVL SI, BX + + // genMemMoveShort + CMPQ BX, $0x03 + JB emit_lit_memmove_emit_remainder_encodeBlockAsm_memmove_move_1or2 + JE emit_lit_memmove_emit_remainder_encodeBlockAsm_memmove_move_3 + CMPQ BX, $0x08 + JB emit_lit_memmove_emit_remainder_encodeBlockAsm_memmove_move_4through7 + CMPQ BX, $0x10 + JBE emit_lit_memmove_emit_remainder_encodeBlockAsm_memmove_move_8through16 + CMPQ BX, $0x20 + JBE emit_lit_memmove_emit_remainder_encodeBlockAsm_memmove_move_17through32 + JMP emit_lit_memmove_emit_remainder_encodeBlockAsm_memmove_move_33through64 + +emit_lit_memmove_emit_remainder_encodeBlockAsm_memmove_move_1or2: + MOVB (CX), SI + MOVB -1(CX)(BX*1), CL + MOVB SI, (AX) + MOVB CL, -1(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeBlockAsm + +emit_lit_memmove_emit_remainder_encodeBlockAsm_memmove_move_3: + MOVW (CX), SI + MOVB 2(CX), CL + MOVW SI, (AX) + MOVB CL, 2(AX) + JMP memmove_end_copy_emit_remainder_encodeBlockAsm + +emit_lit_memmove_emit_remainder_encodeBlockAsm_memmove_move_4through7: + MOVL (CX), SI + MOVL -4(CX)(BX*1), CX + MOVL SI, (AX) + MOVL CX, -4(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeBlockAsm + +emit_lit_memmove_emit_remainder_encodeBlockAsm_memmove_move_8through16: + MOVQ (CX), SI + MOVQ -8(CX)(BX*1), CX + MOVQ SI, (AX) + MOVQ CX, -8(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeBlockAsm + +emit_lit_memmove_emit_remainder_encodeBlockAsm_memmove_move_17through32: + MOVOU (CX), X0 + MOVOU -16(CX)(BX*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeBlockAsm + +emit_lit_memmove_emit_remainder_encodeBlockAsm_memmove_move_33through64: + MOVOU (CX), X0 + MOVOU 16(CX), X1 + MOVOU -32(CX)(BX*1), X2 + MOVOU -16(CX)(BX*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(BX*1) + MOVOU X3, -16(AX)(BX*1) + +memmove_end_copy_emit_remainder_encodeBlockAsm: + MOVQ DX, AX + JMP emit_literal_done_emit_remainder_encodeBlockAsm + +memmove_long_emit_remainder_encodeBlockAsm: + LEAQ (AX)(SI*1), DX + MOVL SI, BX + + // genMemMoveLong + MOVOU (CX), X0 + MOVOU 16(CX), X1 + MOVOU -32(CX)(BX*1), X2 + MOVOU -16(CX)(BX*1), X3 + MOVQ BX, DI + SHRQ $0x05, DI + MOVQ AX, SI + ANDL $0x0000001f, SI + MOVQ $0x00000040, R8 + SUBQ SI, R8 + DECQ DI + JA emit_lit_memmove_long_emit_remainder_encodeBlockAsmlarge_forward_sse_loop_32 + LEAQ -32(CX)(R8*1), SI + LEAQ -32(AX)(R8*1), R9 + +emit_lit_memmove_long_emit_remainder_encodeBlockAsmlarge_big_loop_back: + MOVOU (SI), X4 + MOVOU 16(SI), X5 + MOVOA X4, (R9) + MOVOA X5, 16(R9) + ADDQ $0x20, R9 + ADDQ $0x20, SI + ADDQ $0x20, R8 + DECQ DI + JNA emit_lit_memmove_long_emit_remainder_encodeBlockAsmlarge_big_loop_back + +emit_lit_memmove_long_emit_remainder_encodeBlockAsmlarge_forward_sse_loop_32: + MOVOU -32(CX)(R8*1), X4 + MOVOU -16(CX)(R8*1), X5 + MOVOA X4, -32(AX)(R8*1) + MOVOA X5, -16(AX)(R8*1) + ADDQ $0x20, R8 + CMPQ BX, R8 + JAE emit_lit_memmove_long_emit_remainder_encodeBlockAsmlarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(BX*1) + MOVOU X3, -16(AX)(BX*1) + MOVQ DX, AX + +emit_literal_done_emit_remainder_encodeBlockAsm: + MOVQ dst_base+0(FP), CX + SUBQ CX, AX + MOVQ AX, ret+48(FP) + RET + +// func encodeBlockAsm4MB(dst []byte, src []byte) int +// Requires: BMI, SSE2 +TEXT ·encodeBlockAsm4MB(SB), $65560-56 + MOVQ dst_base+0(FP), AX + MOVQ $0x00000200, CX + LEAQ 24(SP), DX + PXOR X0, X0 + +zero_loop_encodeBlockAsm4MB: + MOVOU X0, (DX) + MOVOU X0, 16(DX) + MOVOU X0, 32(DX) + MOVOU X0, 48(DX) + MOVOU X0, 64(DX) + MOVOU X0, 80(DX) + MOVOU X0, 96(DX) + MOVOU X0, 112(DX) + ADDQ $0x80, DX + DECQ CX + JNZ zero_loop_encodeBlockAsm4MB + MOVL $0x00000000, 12(SP) + MOVQ src_len+32(FP), CX + LEAQ -9(CX), DX + LEAQ -8(CX), BX + MOVL BX, 8(SP) + SHRQ $0x05, CX + SUBL CX, DX + LEAQ (AX)(DX*1), DX + MOVQ DX, (SP) + MOVL $0x00000001, CX + MOVL CX, 16(SP) + MOVQ src_base+24(FP), DX + +search_loop_encodeBlockAsm4MB: + MOVL CX, BX + SUBL 12(SP), BX + SHRL $0x06, BX + LEAL 4(CX)(BX*1), BX + CMPL BX, 8(SP) + JAE emit_remainder_encodeBlockAsm4MB + MOVQ (DX)(CX*1), SI + MOVL BX, 20(SP) + MOVQ $0x0000cf1bbcdcbf9b, R8 + MOVQ SI, R9 + MOVQ SI, R10 + SHRQ $0x08, R10 + SHLQ $0x10, R9 + IMULQ R8, R9 + SHRQ $0x32, R9 + SHLQ $0x10, R10 + IMULQ R8, R10 + SHRQ $0x32, R10 + MOVL 24(SP)(R9*4), BX + MOVL 24(SP)(R10*4), DI + MOVL CX, 24(SP)(R9*4) + LEAL 1(CX), R9 + MOVL R9, 24(SP)(R10*4) + MOVQ SI, R9 + SHRQ $0x10, R9 + SHLQ $0x10, R9 + IMULQ R8, R9 + SHRQ $0x32, R9 + MOVL CX, R8 + SUBL 16(SP), R8 + MOVL 1(DX)(R8*1), R10 + MOVQ SI, R8 + SHRQ $0x08, R8 + CMPL R8, R10 + JNE no_repeat_found_encodeBlockAsm4MB + LEAL 1(CX), SI + MOVL 12(SP), DI + MOVL SI, BX + SUBL 16(SP), BX + JZ repeat_extend_back_end_encodeBlockAsm4MB + +repeat_extend_back_loop_encodeBlockAsm4MB: + CMPL SI, DI + JBE repeat_extend_back_end_encodeBlockAsm4MB + MOVB -1(DX)(BX*1), R8 + MOVB -1(DX)(SI*1), R9 + CMPB R8, R9 + JNE repeat_extend_back_end_encodeBlockAsm4MB + LEAL -1(SI), SI + DECL BX + JNZ repeat_extend_back_loop_encodeBlockAsm4MB + +repeat_extend_back_end_encodeBlockAsm4MB: + MOVL SI, BX + SUBL 12(SP), BX + LEAQ 4(AX)(BX*1), BX + CMPQ BX, (SP) + JB repeat_dst_size_check_encodeBlockAsm4MB + MOVQ $0x00000000, ret+48(FP) + RET + +repeat_dst_size_check_encodeBlockAsm4MB: + MOVL 12(SP), BX + CMPL BX, SI + JEQ emit_literal_done_repeat_emit_encodeBlockAsm4MB + MOVL SI, R8 + MOVL SI, 12(SP) + LEAQ (DX)(BX*1), R9 + SUBL BX, R8 + LEAL -1(R8), BX + CMPL BX, $0x3c + JB one_byte_repeat_emit_encodeBlockAsm4MB + CMPL BX, $0x00000100 + JB two_bytes_repeat_emit_encodeBlockAsm4MB + CMPL BX, $0x00010000 + JB three_bytes_repeat_emit_encodeBlockAsm4MB + MOVL BX, R10 + SHRL $0x10, R10 + MOVB $0xf8, (AX) + MOVW BX, 1(AX) + MOVB R10, 3(AX) + ADDQ $0x04, AX + JMP memmove_long_repeat_emit_encodeBlockAsm4MB + +three_bytes_repeat_emit_encodeBlockAsm4MB: + MOVB $0xf4, (AX) + MOVW BX, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_repeat_emit_encodeBlockAsm4MB + +two_bytes_repeat_emit_encodeBlockAsm4MB: + MOVB $0xf0, (AX) + MOVB BL, 1(AX) + ADDQ $0x02, AX + CMPL BX, $0x40 + JB memmove_repeat_emit_encodeBlockAsm4MB + JMP memmove_long_repeat_emit_encodeBlockAsm4MB + +one_byte_repeat_emit_encodeBlockAsm4MB: + SHLB $0x02, BL + MOVB BL, (AX) + ADDQ $0x01, AX + +memmove_repeat_emit_encodeBlockAsm4MB: + LEAQ (AX)(R8*1), BX + + // genMemMoveShort + CMPQ R8, $0x08 + JBE emit_lit_memmove_repeat_emit_encodeBlockAsm4MB_memmove_move_8 + CMPQ R8, $0x10 + JBE emit_lit_memmove_repeat_emit_encodeBlockAsm4MB_memmove_move_8through16 + CMPQ R8, $0x20 + JBE emit_lit_memmove_repeat_emit_encodeBlockAsm4MB_memmove_move_17through32 + JMP emit_lit_memmove_repeat_emit_encodeBlockAsm4MB_memmove_move_33through64 + +emit_lit_memmove_repeat_emit_encodeBlockAsm4MB_memmove_move_8: + MOVQ (R9), R10 + MOVQ R10, (AX) + JMP memmove_end_copy_repeat_emit_encodeBlockAsm4MB + +emit_lit_memmove_repeat_emit_encodeBlockAsm4MB_memmove_move_8through16: + MOVQ (R9), R10 + MOVQ -8(R9)(R8*1), R9 + MOVQ R10, (AX) + MOVQ R9, -8(AX)(R8*1) + JMP memmove_end_copy_repeat_emit_encodeBlockAsm4MB + +emit_lit_memmove_repeat_emit_encodeBlockAsm4MB_memmove_move_17through32: + MOVOU (R9), X0 + MOVOU -16(R9)(R8*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(R8*1) + JMP memmove_end_copy_repeat_emit_encodeBlockAsm4MB + +emit_lit_memmove_repeat_emit_encodeBlockAsm4MB_memmove_move_33through64: + MOVOU (R9), X0 + MOVOU 16(R9), X1 + MOVOU -32(R9)(R8*1), X2 + MOVOU -16(R9)(R8*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + +memmove_end_copy_repeat_emit_encodeBlockAsm4MB: + MOVQ BX, AX + JMP emit_literal_done_repeat_emit_encodeBlockAsm4MB + +memmove_long_repeat_emit_encodeBlockAsm4MB: + LEAQ (AX)(R8*1), BX + + // genMemMoveLong + MOVOU (R9), X0 + MOVOU 16(R9), X1 + MOVOU -32(R9)(R8*1), X2 + MOVOU -16(R9)(R8*1), X3 + MOVQ R8, R11 + SHRQ $0x05, R11 + MOVQ AX, R10 + ANDL $0x0000001f, R10 + MOVQ $0x00000040, R12 + SUBQ R10, R12 + DECQ R11 + JA emit_lit_memmove_long_repeat_emit_encodeBlockAsm4MBlarge_forward_sse_loop_32 + LEAQ -32(R9)(R12*1), R10 + LEAQ -32(AX)(R12*1), R13 + +emit_lit_memmove_long_repeat_emit_encodeBlockAsm4MBlarge_big_loop_back: + MOVOU (R10), X4 + MOVOU 16(R10), X5 + MOVOA X4, (R13) + MOVOA X5, 16(R13) + ADDQ $0x20, R13 + ADDQ $0x20, R10 + ADDQ $0x20, R12 + DECQ R11 + JNA emit_lit_memmove_long_repeat_emit_encodeBlockAsm4MBlarge_big_loop_back + +emit_lit_memmove_long_repeat_emit_encodeBlockAsm4MBlarge_forward_sse_loop_32: + MOVOU -32(R9)(R12*1), X4 + MOVOU -16(R9)(R12*1), X5 + MOVOA X4, -32(AX)(R12*1) + MOVOA X5, -16(AX)(R12*1) + ADDQ $0x20, R12 + CMPQ R8, R12 + JAE emit_lit_memmove_long_repeat_emit_encodeBlockAsm4MBlarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + MOVQ BX, AX + +emit_literal_done_repeat_emit_encodeBlockAsm4MB: + ADDL $0x05, CX + MOVL CX, BX + SUBL 16(SP), BX + MOVQ src_len+32(FP), R8 + SUBL CX, R8 + LEAQ (DX)(CX*1), R9 + LEAQ (DX)(BX*1), BX + + // matchLen + XORL R11, R11 + +matchlen_loopback_16_repeat_extend_encodeBlockAsm4MB: + CMPL R8, $0x10 + JB matchlen_match8_repeat_extend_encodeBlockAsm4MB + MOVQ (R9)(R11*1), R10 + MOVQ 8(R9)(R11*1), R12 + XORQ (BX)(R11*1), R10 + JNZ matchlen_bsf_8_repeat_extend_encodeBlockAsm4MB + XORQ 8(BX)(R11*1), R12 + JNZ matchlen_bsf_16repeat_extend_encodeBlockAsm4MB + LEAL -16(R8), R8 + LEAL 16(R11), R11 + JMP matchlen_loopback_16_repeat_extend_encodeBlockAsm4MB + +matchlen_bsf_16repeat_extend_encodeBlockAsm4MB: +#ifdef GOAMD64_v3 + TZCNTQ R12, R12 + +#else + BSFQ R12, R12 + +#endif + SARQ $0x03, R12 + LEAL 8(R11)(R12*1), R11 + JMP repeat_extend_forward_end_encodeBlockAsm4MB + +matchlen_match8_repeat_extend_encodeBlockAsm4MB: + CMPL R8, $0x08 + JB matchlen_match4_repeat_extend_encodeBlockAsm4MB + MOVQ (R9)(R11*1), R10 + XORQ (BX)(R11*1), R10 + JNZ matchlen_bsf_8_repeat_extend_encodeBlockAsm4MB + LEAL -8(R8), R8 + LEAL 8(R11), R11 + JMP matchlen_match4_repeat_extend_encodeBlockAsm4MB + +matchlen_bsf_8_repeat_extend_encodeBlockAsm4MB: +#ifdef GOAMD64_v3 + TZCNTQ R10, R10 + +#else + BSFQ R10, R10 + +#endif + SARQ $0x03, R10 + LEAL (R11)(R10*1), R11 + JMP repeat_extend_forward_end_encodeBlockAsm4MB + +matchlen_match4_repeat_extend_encodeBlockAsm4MB: + CMPL R8, $0x04 + JB matchlen_match2_repeat_extend_encodeBlockAsm4MB + MOVL (R9)(R11*1), R10 + CMPL (BX)(R11*1), R10 + JNE matchlen_match2_repeat_extend_encodeBlockAsm4MB + LEAL -4(R8), R8 + LEAL 4(R11), R11 + +matchlen_match2_repeat_extend_encodeBlockAsm4MB: + CMPL R8, $0x01 + JE matchlen_match1_repeat_extend_encodeBlockAsm4MB + JB repeat_extend_forward_end_encodeBlockAsm4MB + MOVW (R9)(R11*1), R10 + CMPW (BX)(R11*1), R10 + JNE matchlen_match1_repeat_extend_encodeBlockAsm4MB + LEAL 2(R11), R11 + SUBL $0x02, R8 + JZ repeat_extend_forward_end_encodeBlockAsm4MB + +matchlen_match1_repeat_extend_encodeBlockAsm4MB: + MOVB (R9)(R11*1), R10 + CMPB (BX)(R11*1), R10 + JNE repeat_extend_forward_end_encodeBlockAsm4MB + LEAL 1(R11), R11 + +repeat_extend_forward_end_encodeBlockAsm4MB: + ADDL R11, CX + MOVL CX, BX + SUBL SI, BX + MOVL 16(SP), SI + TESTL DI, DI + JZ repeat_as_copy_encodeBlockAsm4MB + + // emitRepeat + MOVL BX, DI + LEAL -4(BX), BX + CMPL DI, $0x08 + JBE repeat_two_match_repeat_encodeBlockAsm4MB + CMPL DI, $0x0c + JAE cant_repeat_two_offset_match_repeat_encodeBlockAsm4MB + CMPL SI, $0x00000800 + JB repeat_two_offset_match_repeat_encodeBlockAsm4MB + +cant_repeat_two_offset_match_repeat_encodeBlockAsm4MB: + CMPL BX, $0x00000104 + JB repeat_three_match_repeat_encodeBlockAsm4MB + CMPL BX, $0x00010100 + JB repeat_four_match_repeat_encodeBlockAsm4MB + LEAL -65536(BX), BX + MOVL BX, SI + MOVW $0x001d, (AX) + MOVW BX, 2(AX) + SARL $0x10, SI + MOVB SI, 4(AX) + ADDQ $0x05, AX + JMP repeat_end_emit_encodeBlockAsm4MB + +repeat_four_match_repeat_encodeBlockAsm4MB: + LEAL -256(BX), BX + MOVW $0x0019, (AX) + MOVW BX, 2(AX) + ADDQ $0x04, AX + JMP repeat_end_emit_encodeBlockAsm4MB + +repeat_three_match_repeat_encodeBlockAsm4MB: + LEAL -4(BX), BX + MOVW $0x0015, (AX) + MOVB BL, 2(AX) + ADDQ $0x03, AX + JMP repeat_end_emit_encodeBlockAsm4MB + +repeat_two_match_repeat_encodeBlockAsm4MB: + SHLL $0x02, BX + ORL $0x01, BX + MOVW BX, (AX) + ADDQ $0x02, AX + JMP repeat_end_emit_encodeBlockAsm4MB + +repeat_two_offset_match_repeat_encodeBlockAsm4MB: + XORQ DI, DI + LEAL 1(DI)(BX*4), BX + MOVB SI, 1(AX) + SARL $0x08, SI + SHLL $0x05, SI + ORL SI, BX + MOVB BL, (AX) + ADDQ $0x02, AX + JMP repeat_end_emit_encodeBlockAsm4MB + +repeat_as_copy_encodeBlockAsm4MB: + // emitCopy + CMPL SI, $0x00010000 + JB two_byte_offset_repeat_as_copy_encodeBlockAsm4MB + CMPL BX, $0x40 + JBE four_bytes_remain_repeat_as_copy_encodeBlockAsm4MB + MOVB $0xff, (AX) + MOVL SI, 1(AX) + LEAL -64(BX), BX + ADDQ $0x05, AX + CMPL BX, $0x04 + JB four_bytes_remain_repeat_as_copy_encodeBlockAsm4MB + + // emitRepeat + MOVL BX, DI + LEAL -4(BX), BX + CMPL DI, $0x08 + JBE repeat_two_repeat_as_copy_encodeBlockAsm4MB_emit_copy + CMPL DI, $0x0c + JAE cant_repeat_two_offset_repeat_as_copy_encodeBlockAsm4MB_emit_copy + CMPL SI, $0x00000800 + JB repeat_two_offset_repeat_as_copy_encodeBlockAsm4MB_emit_copy + +cant_repeat_two_offset_repeat_as_copy_encodeBlockAsm4MB_emit_copy: + CMPL BX, $0x00000104 + JB repeat_three_repeat_as_copy_encodeBlockAsm4MB_emit_copy + CMPL BX, $0x00010100 + JB repeat_four_repeat_as_copy_encodeBlockAsm4MB_emit_copy + LEAL -65536(BX), BX + MOVL BX, SI + MOVW $0x001d, (AX) + MOVW BX, 2(AX) + SARL $0x10, SI + MOVB SI, 4(AX) + ADDQ $0x05, AX + JMP repeat_end_emit_encodeBlockAsm4MB + +repeat_four_repeat_as_copy_encodeBlockAsm4MB_emit_copy: + LEAL -256(BX), BX + MOVW $0x0019, (AX) + MOVW BX, 2(AX) + ADDQ $0x04, AX + JMP repeat_end_emit_encodeBlockAsm4MB + +repeat_three_repeat_as_copy_encodeBlockAsm4MB_emit_copy: + LEAL -4(BX), BX + MOVW $0x0015, (AX) + MOVB BL, 2(AX) + ADDQ $0x03, AX + JMP repeat_end_emit_encodeBlockAsm4MB + +repeat_two_repeat_as_copy_encodeBlockAsm4MB_emit_copy: + SHLL $0x02, BX + ORL $0x01, BX + MOVW BX, (AX) + ADDQ $0x02, AX + JMP repeat_end_emit_encodeBlockAsm4MB + +repeat_two_offset_repeat_as_copy_encodeBlockAsm4MB_emit_copy: + XORQ DI, DI + LEAL 1(DI)(BX*4), BX + MOVB SI, 1(AX) + SARL $0x08, SI + SHLL $0x05, SI + ORL SI, BX + MOVB BL, (AX) + ADDQ $0x02, AX + JMP repeat_end_emit_encodeBlockAsm4MB + +four_bytes_remain_repeat_as_copy_encodeBlockAsm4MB: + TESTL BX, BX + JZ repeat_end_emit_encodeBlockAsm4MB + XORL DI, DI + LEAL -1(DI)(BX*4), BX + MOVB BL, (AX) + MOVL SI, 1(AX) + ADDQ $0x05, AX + JMP repeat_end_emit_encodeBlockAsm4MB + +two_byte_offset_repeat_as_copy_encodeBlockAsm4MB: + CMPL BX, $0x40 + JBE two_byte_offset_short_repeat_as_copy_encodeBlockAsm4MB + CMPL SI, $0x00000800 + JAE long_offset_short_repeat_as_copy_encodeBlockAsm4MB + MOVL $0x00000001, DI + LEAL 16(DI), DI + MOVB SI, 1(AX) + SHRL $0x08, SI + SHLL $0x05, SI + ORL SI, DI + MOVB DI, (AX) + ADDQ $0x02, AX + SUBL $0x08, BX + + // emitRepeat + LEAL -4(BX), BX + JMP cant_repeat_two_offset_repeat_as_copy_encodeBlockAsm4MB_emit_copy_short_2b + MOVL BX, DI + LEAL -4(BX), BX + CMPL DI, $0x08 + JBE repeat_two_repeat_as_copy_encodeBlockAsm4MB_emit_copy_short_2b + CMPL DI, $0x0c + JAE cant_repeat_two_offset_repeat_as_copy_encodeBlockAsm4MB_emit_copy_short_2b + CMPL SI, $0x00000800 + JB repeat_two_offset_repeat_as_copy_encodeBlockAsm4MB_emit_copy_short_2b + +cant_repeat_two_offset_repeat_as_copy_encodeBlockAsm4MB_emit_copy_short_2b: + CMPL BX, $0x00000104 + JB repeat_three_repeat_as_copy_encodeBlockAsm4MB_emit_copy_short_2b + CMPL BX, $0x00010100 + JB repeat_four_repeat_as_copy_encodeBlockAsm4MB_emit_copy_short_2b + LEAL -65536(BX), BX + MOVL BX, SI + MOVW $0x001d, (AX) + MOVW BX, 2(AX) + SARL $0x10, SI + MOVB SI, 4(AX) + ADDQ $0x05, AX + JMP repeat_end_emit_encodeBlockAsm4MB + +repeat_four_repeat_as_copy_encodeBlockAsm4MB_emit_copy_short_2b: + LEAL -256(BX), BX + MOVW $0x0019, (AX) + MOVW BX, 2(AX) + ADDQ $0x04, AX + JMP repeat_end_emit_encodeBlockAsm4MB + +repeat_three_repeat_as_copy_encodeBlockAsm4MB_emit_copy_short_2b: + LEAL -4(BX), BX + MOVW $0x0015, (AX) + MOVB BL, 2(AX) + ADDQ $0x03, AX + JMP repeat_end_emit_encodeBlockAsm4MB + +repeat_two_repeat_as_copy_encodeBlockAsm4MB_emit_copy_short_2b: + SHLL $0x02, BX + ORL $0x01, BX + MOVW BX, (AX) + ADDQ $0x02, AX + JMP repeat_end_emit_encodeBlockAsm4MB + +repeat_two_offset_repeat_as_copy_encodeBlockAsm4MB_emit_copy_short_2b: + XORQ DI, DI + LEAL 1(DI)(BX*4), BX + MOVB SI, 1(AX) + SARL $0x08, SI + SHLL $0x05, SI + ORL SI, BX + MOVB BL, (AX) + ADDQ $0x02, AX + JMP repeat_end_emit_encodeBlockAsm4MB + +long_offset_short_repeat_as_copy_encodeBlockAsm4MB: + MOVB $0xee, (AX) + MOVW SI, 1(AX) + LEAL -60(BX), BX + ADDQ $0x03, AX + + // emitRepeat + MOVL BX, DI + LEAL -4(BX), BX + CMPL DI, $0x08 + JBE repeat_two_repeat_as_copy_encodeBlockAsm4MB_emit_copy_short + CMPL DI, $0x0c + JAE cant_repeat_two_offset_repeat_as_copy_encodeBlockAsm4MB_emit_copy_short + CMPL SI, $0x00000800 + JB repeat_two_offset_repeat_as_copy_encodeBlockAsm4MB_emit_copy_short + +cant_repeat_two_offset_repeat_as_copy_encodeBlockAsm4MB_emit_copy_short: + CMPL BX, $0x00000104 + JB repeat_three_repeat_as_copy_encodeBlockAsm4MB_emit_copy_short + CMPL BX, $0x00010100 + JB repeat_four_repeat_as_copy_encodeBlockAsm4MB_emit_copy_short + LEAL -65536(BX), BX + MOVL BX, SI + MOVW $0x001d, (AX) + MOVW BX, 2(AX) + SARL $0x10, SI + MOVB SI, 4(AX) + ADDQ $0x05, AX + JMP repeat_end_emit_encodeBlockAsm4MB + +repeat_four_repeat_as_copy_encodeBlockAsm4MB_emit_copy_short: + LEAL -256(BX), BX + MOVW $0x0019, (AX) + MOVW BX, 2(AX) + ADDQ $0x04, AX + JMP repeat_end_emit_encodeBlockAsm4MB + +repeat_three_repeat_as_copy_encodeBlockAsm4MB_emit_copy_short: + LEAL -4(BX), BX + MOVW $0x0015, (AX) + MOVB BL, 2(AX) + ADDQ $0x03, AX + JMP repeat_end_emit_encodeBlockAsm4MB + +repeat_two_repeat_as_copy_encodeBlockAsm4MB_emit_copy_short: + SHLL $0x02, BX + ORL $0x01, BX + MOVW BX, (AX) + ADDQ $0x02, AX + JMP repeat_end_emit_encodeBlockAsm4MB + +repeat_two_offset_repeat_as_copy_encodeBlockAsm4MB_emit_copy_short: + XORQ DI, DI + LEAL 1(DI)(BX*4), BX + MOVB SI, 1(AX) + SARL $0x08, SI + SHLL $0x05, SI + ORL SI, BX + MOVB BL, (AX) + ADDQ $0x02, AX + JMP repeat_end_emit_encodeBlockAsm4MB + +two_byte_offset_short_repeat_as_copy_encodeBlockAsm4MB: + MOVL BX, DI + SHLL $0x02, DI + CMPL BX, $0x0c + JAE emit_copy_three_repeat_as_copy_encodeBlockAsm4MB + CMPL SI, $0x00000800 + JAE emit_copy_three_repeat_as_copy_encodeBlockAsm4MB + LEAL -15(DI), DI + MOVB SI, 1(AX) + SHRL $0x08, SI + SHLL $0x05, SI + ORL SI, DI + MOVB DI, (AX) + ADDQ $0x02, AX + JMP repeat_end_emit_encodeBlockAsm4MB + +emit_copy_three_repeat_as_copy_encodeBlockAsm4MB: + LEAL -2(DI), DI + MOVB DI, (AX) + MOVW SI, 1(AX) + ADDQ $0x03, AX + +repeat_end_emit_encodeBlockAsm4MB: + MOVL CX, 12(SP) + JMP search_loop_encodeBlockAsm4MB + +no_repeat_found_encodeBlockAsm4MB: + CMPL (DX)(BX*1), SI + JEQ candidate_match_encodeBlockAsm4MB + SHRQ $0x08, SI + MOVL 24(SP)(R9*4), BX + LEAL 2(CX), R8 + CMPL (DX)(DI*1), SI + JEQ candidate2_match_encodeBlockAsm4MB + MOVL R8, 24(SP)(R9*4) + SHRQ $0x08, SI + CMPL (DX)(BX*1), SI + JEQ candidate3_match_encodeBlockAsm4MB + MOVL 20(SP), CX + JMP search_loop_encodeBlockAsm4MB + +candidate3_match_encodeBlockAsm4MB: + ADDL $0x02, CX + JMP candidate_match_encodeBlockAsm4MB + +candidate2_match_encodeBlockAsm4MB: + MOVL R8, 24(SP)(R9*4) + INCL CX + MOVL DI, BX + +candidate_match_encodeBlockAsm4MB: + MOVL 12(SP), SI + TESTL BX, BX + JZ match_extend_back_end_encodeBlockAsm4MB + +match_extend_back_loop_encodeBlockAsm4MB: + CMPL CX, SI + JBE match_extend_back_end_encodeBlockAsm4MB + MOVB -1(DX)(BX*1), DI + MOVB -1(DX)(CX*1), R8 + CMPB DI, R8 + JNE match_extend_back_end_encodeBlockAsm4MB + LEAL -1(CX), CX + DECL BX + JZ match_extend_back_end_encodeBlockAsm4MB + JMP match_extend_back_loop_encodeBlockAsm4MB + +match_extend_back_end_encodeBlockAsm4MB: + MOVL CX, SI + SUBL 12(SP), SI + LEAQ 4(AX)(SI*1), SI + CMPQ SI, (SP) + JB match_dst_size_check_encodeBlockAsm4MB + MOVQ $0x00000000, ret+48(FP) + RET + +match_dst_size_check_encodeBlockAsm4MB: + MOVL CX, SI + MOVL 12(SP), DI + CMPL DI, SI + JEQ emit_literal_done_match_emit_encodeBlockAsm4MB + MOVL SI, R8 + MOVL SI, 12(SP) + LEAQ (DX)(DI*1), SI + SUBL DI, R8 + LEAL -1(R8), DI + CMPL DI, $0x3c + JB one_byte_match_emit_encodeBlockAsm4MB + CMPL DI, $0x00000100 + JB two_bytes_match_emit_encodeBlockAsm4MB + CMPL DI, $0x00010000 + JB three_bytes_match_emit_encodeBlockAsm4MB + MOVL DI, R9 + SHRL $0x10, R9 + MOVB $0xf8, (AX) + MOVW DI, 1(AX) + MOVB R9, 3(AX) + ADDQ $0x04, AX + JMP memmove_long_match_emit_encodeBlockAsm4MB + +three_bytes_match_emit_encodeBlockAsm4MB: + MOVB $0xf4, (AX) + MOVW DI, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_match_emit_encodeBlockAsm4MB + +two_bytes_match_emit_encodeBlockAsm4MB: + MOVB $0xf0, (AX) + MOVB DI, 1(AX) + ADDQ $0x02, AX + CMPL DI, $0x40 + JB memmove_match_emit_encodeBlockAsm4MB + JMP memmove_long_match_emit_encodeBlockAsm4MB + +one_byte_match_emit_encodeBlockAsm4MB: + SHLB $0x02, DI + MOVB DI, (AX) + ADDQ $0x01, AX + +memmove_match_emit_encodeBlockAsm4MB: + LEAQ (AX)(R8*1), DI + + // genMemMoveShort + CMPQ R8, $0x08 + JBE emit_lit_memmove_match_emit_encodeBlockAsm4MB_memmove_move_8 + CMPQ R8, $0x10 + JBE emit_lit_memmove_match_emit_encodeBlockAsm4MB_memmove_move_8through16 + CMPQ R8, $0x20 + JBE emit_lit_memmove_match_emit_encodeBlockAsm4MB_memmove_move_17through32 + JMP emit_lit_memmove_match_emit_encodeBlockAsm4MB_memmove_move_33through64 + +emit_lit_memmove_match_emit_encodeBlockAsm4MB_memmove_move_8: + MOVQ (SI), R9 + MOVQ R9, (AX) + JMP memmove_end_copy_match_emit_encodeBlockAsm4MB + +emit_lit_memmove_match_emit_encodeBlockAsm4MB_memmove_move_8through16: + MOVQ (SI), R9 + MOVQ -8(SI)(R8*1), SI + MOVQ R9, (AX) + MOVQ SI, -8(AX)(R8*1) + JMP memmove_end_copy_match_emit_encodeBlockAsm4MB + +emit_lit_memmove_match_emit_encodeBlockAsm4MB_memmove_move_17through32: + MOVOU (SI), X0 + MOVOU -16(SI)(R8*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(R8*1) + JMP memmove_end_copy_match_emit_encodeBlockAsm4MB + +emit_lit_memmove_match_emit_encodeBlockAsm4MB_memmove_move_33through64: + MOVOU (SI), X0 + MOVOU 16(SI), X1 + MOVOU -32(SI)(R8*1), X2 + MOVOU -16(SI)(R8*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + +memmove_end_copy_match_emit_encodeBlockAsm4MB: + MOVQ DI, AX + JMP emit_literal_done_match_emit_encodeBlockAsm4MB + +memmove_long_match_emit_encodeBlockAsm4MB: + LEAQ (AX)(R8*1), DI + + // genMemMoveLong + MOVOU (SI), X0 + MOVOU 16(SI), X1 + MOVOU -32(SI)(R8*1), X2 + MOVOU -16(SI)(R8*1), X3 + MOVQ R8, R10 + SHRQ $0x05, R10 + MOVQ AX, R9 + ANDL $0x0000001f, R9 + MOVQ $0x00000040, R11 + SUBQ R9, R11 + DECQ R10 + JA emit_lit_memmove_long_match_emit_encodeBlockAsm4MBlarge_forward_sse_loop_32 + LEAQ -32(SI)(R11*1), R9 + LEAQ -32(AX)(R11*1), R12 + +emit_lit_memmove_long_match_emit_encodeBlockAsm4MBlarge_big_loop_back: + MOVOU (R9), X4 + MOVOU 16(R9), X5 + MOVOA X4, (R12) + MOVOA X5, 16(R12) + ADDQ $0x20, R12 + ADDQ $0x20, R9 + ADDQ $0x20, R11 + DECQ R10 + JNA emit_lit_memmove_long_match_emit_encodeBlockAsm4MBlarge_big_loop_back + +emit_lit_memmove_long_match_emit_encodeBlockAsm4MBlarge_forward_sse_loop_32: + MOVOU -32(SI)(R11*1), X4 + MOVOU -16(SI)(R11*1), X5 + MOVOA X4, -32(AX)(R11*1) + MOVOA X5, -16(AX)(R11*1) + ADDQ $0x20, R11 + CMPQ R8, R11 + JAE emit_lit_memmove_long_match_emit_encodeBlockAsm4MBlarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + MOVQ DI, AX + +emit_literal_done_match_emit_encodeBlockAsm4MB: +match_nolit_loop_encodeBlockAsm4MB: + MOVL CX, SI + SUBL BX, SI + MOVL SI, 16(SP) + ADDL $0x04, CX + ADDL $0x04, BX + MOVQ src_len+32(FP), SI + SUBL CX, SI + LEAQ (DX)(CX*1), DI + LEAQ (DX)(BX*1), BX + + // matchLen + XORL R9, R9 + +matchlen_loopback_16_match_nolit_encodeBlockAsm4MB: + CMPL SI, $0x10 + JB matchlen_match8_match_nolit_encodeBlockAsm4MB + MOVQ (DI)(R9*1), R8 + MOVQ 8(DI)(R9*1), R10 + XORQ (BX)(R9*1), R8 + JNZ matchlen_bsf_8_match_nolit_encodeBlockAsm4MB + XORQ 8(BX)(R9*1), R10 + JNZ matchlen_bsf_16match_nolit_encodeBlockAsm4MB + LEAL -16(SI), SI + LEAL 16(R9), R9 + JMP matchlen_loopback_16_match_nolit_encodeBlockAsm4MB + +matchlen_bsf_16match_nolit_encodeBlockAsm4MB: +#ifdef GOAMD64_v3 + TZCNTQ R10, R10 + +#else + BSFQ R10, R10 + +#endif + SARQ $0x03, R10 + LEAL 8(R9)(R10*1), R9 + JMP match_nolit_end_encodeBlockAsm4MB + +matchlen_match8_match_nolit_encodeBlockAsm4MB: + CMPL SI, $0x08 + JB matchlen_match4_match_nolit_encodeBlockAsm4MB + MOVQ (DI)(R9*1), R8 + XORQ (BX)(R9*1), R8 + JNZ matchlen_bsf_8_match_nolit_encodeBlockAsm4MB + LEAL -8(SI), SI + LEAL 8(R9), R9 + JMP matchlen_match4_match_nolit_encodeBlockAsm4MB + +matchlen_bsf_8_match_nolit_encodeBlockAsm4MB: +#ifdef GOAMD64_v3 + TZCNTQ R8, R8 + +#else + BSFQ R8, R8 + +#endif + SARQ $0x03, R8 + LEAL (R9)(R8*1), R9 + JMP match_nolit_end_encodeBlockAsm4MB + +matchlen_match4_match_nolit_encodeBlockAsm4MB: + CMPL SI, $0x04 + JB matchlen_match2_match_nolit_encodeBlockAsm4MB + MOVL (DI)(R9*1), R8 + CMPL (BX)(R9*1), R8 + JNE matchlen_match2_match_nolit_encodeBlockAsm4MB + LEAL -4(SI), SI + LEAL 4(R9), R9 + +matchlen_match2_match_nolit_encodeBlockAsm4MB: + CMPL SI, $0x01 + JE matchlen_match1_match_nolit_encodeBlockAsm4MB + JB match_nolit_end_encodeBlockAsm4MB + MOVW (DI)(R9*1), R8 + CMPW (BX)(R9*1), R8 + JNE matchlen_match1_match_nolit_encodeBlockAsm4MB + LEAL 2(R9), R9 + SUBL $0x02, SI + JZ match_nolit_end_encodeBlockAsm4MB + +matchlen_match1_match_nolit_encodeBlockAsm4MB: + MOVB (DI)(R9*1), R8 + CMPB (BX)(R9*1), R8 + JNE match_nolit_end_encodeBlockAsm4MB + LEAL 1(R9), R9 + +match_nolit_end_encodeBlockAsm4MB: + ADDL R9, CX + MOVL 16(SP), BX + ADDL $0x04, R9 + MOVL CX, 12(SP) + + // emitCopy + CMPL BX, $0x00010000 + JB two_byte_offset_match_nolit_encodeBlockAsm4MB + CMPL R9, $0x40 + JBE four_bytes_remain_match_nolit_encodeBlockAsm4MB + MOVB $0xff, (AX) + MOVL BX, 1(AX) + LEAL -64(R9), R9 + ADDQ $0x05, AX + CMPL R9, $0x04 + JB four_bytes_remain_match_nolit_encodeBlockAsm4MB + + // emitRepeat + MOVL R9, SI + LEAL -4(R9), R9 + CMPL SI, $0x08 + JBE repeat_two_match_nolit_encodeBlockAsm4MB_emit_copy + CMPL SI, $0x0c + JAE cant_repeat_two_offset_match_nolit_encodeBlockAsm4MB_emit_copy + CMPL BX, $0x00000800 + JB repeat_two_offset_match_nolit_encodeBlockAsm4MB_emit_copy + +cant_repeat_two_offset_match_nolit_encodeBlockAsm4MB_emit_copy: + CMPL R9, $0x00000104 + JB repeat_three_match_nolit_encodeBlockAsm4MB_emit_copy + CMPL R9, $0x00010100 + JB repeat_four_match_nolit_encodeBlockAsm4MB_emit_copy + LEAL -65536(R9), R9 + MOVL R9, BX + MOVW $0x001d, (AX) + MOVW R9, 2(AX) + SARL $0x10, BX + MOVB BL, 4(AX) + ADDQ $0x05, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm4MB + +repeat_four_match_nolit_encodeBlockAsm4MB_emit_copy: + LEAL -256(R9), R9 + MOVW $0x0019, (AX) + MOVW R9, 2(AX) + ADDQ $0x04, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm4MB + +repeat_three_match_nolit_encodeBlockAsm4MB_emit_copy: + LEAL -4(R9), R9 + MOVW $0x0015, (AX) + MOVB R9, 2(AX) + ADDQ $0x03, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm4MB + +repeat_two_match_nolit_encodeBlockAsm4MB_emit_copy: + SHLL $0x02, R9 + ORL $0x01, R9 + MOVW R9, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm4MB + +repeat_two_offset_match_nolit_encodeBlockAsm4MB_emit_copy: + XORQ SI, SI + LEAL 1(SI)(R9*4), R9 + MOVB BL, 1(AX) + SARL $0x08, BX + SHLL $0x05, BX + ORL BX, R9 + MOVB R9, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm4MB + +four_bytes_remain_match_nolit_encodeBlockAsm4MB: + TESTL R9, R9 + JZ match_nolit_emitcopy_end_encodeBlockAsm4MB + XORL SI, SI + LEAL -1(SI)(R9*4), R9 + MOVB R9, (AX) + MOVL BX, 1(AX) + ADDQ $0x05, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm4MB + +two_byte_offset_match_nolit_encodeBlockAsm4MB: + CMPL R9, $0x40 + JBE two_byte_offset_short_match_nolit_encodeBlockAsm4MB + CMPL BX, $0x00000800 + JAE long_offset_short_match_nolit_encodeBlockAsm4MB + MOVL $0x00000001, SI + LEAL 16(SI), SI + MOVB BL, 1(AX) + SHRL $0x08, BX + SHLL $0x05, BX + ORL BX, SI + MOVB SI, (AX) + ADDQ $0x02, AX + SUBL $0x08, R9 + + // emitRepeat + LEAL -4(R9), R9 + JMP cant_repeat_two_offset_match_nolit_encodeBlockAsm4MB_emit_copy_short_2b + MOVL R9, SI + LEAL -4(R9), R9 + CMPL SI, $0x08 + JBE repeat_two_match_nolit_encodeBlockAsm4MB_emit_copy_short_2b + CMPL SI, $0x0c + JAE cant_repeat_two_offset_match_nolit_encodeBlockAsm4MB_emit_copy_short_2b + CMPL BX, $0x00000800 + JB repeat_two_offset_match_nolit_encodeBlockAsm4MB_emit_copy_short_2b + +cant_repeat_two_offset_match_nolit_encodeBlockAsm4MB_emit_copy_short_2b: + CMPL R9, $0x00000104 + JB repeat_three_match_nolit_encodeBlockAsm4MB_emit_copy_short_2b + CMPL R9, $0x00010100 + JB repeat_four_match_nolit_encodeBlockAsm4MB_emit_copy_short_2b + LEAL -65536(R9), R9 + MOVL R9, BX + MOVW $0x001d, (AX) + MOVW R9, 2(AX) + SARL $0x10, BX + MOVB BL, 4(AX) + ADDQ $0x05, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm4MB + +repeat_four_match_nolit_encodeBlockAsm4MB_emit_copy_short_2b: + LEAL -256(R9), R9 + MOVW $0x0019, (AX) + MOVW R9, 2(AX) + ADDQ $0x04, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm4MB + +repeat_three_match_nolit_encodeBlockAsm4MB_emit_copy_short_2b: + LEAL -4(R9), R9 + MOVW $0x0015, (AX) + MOVB R9, 2(AX) + ADDQ $0x03, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm4MB + +repeat_two_match_nolit_encodeBlockAsm4MB_emit_copy_short_2b: + SHLL $0x02, R9 + ORL $0x01, R9 + MOVW R9, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm4MB + +repeat_two_offset_match_nolit_encodeBlockAsm4MB_emit_copy_short_2b: + XORQ SI, SI + LEAL 1(SI)(R9*4), R9 + MOVB BL, 1(AX) + SARL $0x08, BX + SHLL $0x05, BX + ORL BX, R9 + MOVB R9, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm4MB + +long_offset_short_match_nolit_encodeBlockAsm4MB: + MOVB $0xee, (AX) + MOVW BX, 1(AX) + LEAL -60(R9), R9 + ADDQ $0x03, AX + + // emitRepeat + MOVL R9, SI + LEAL -4(R9), R9 + CMPL SI, $0x08 + JBE repeat_two_match_nolit_encodeBlockAsm4MB_emit_copy_short + CMPL SI, $0x0c + JAE cant_repeat_two_offset_match_nolit_encodeBlockAsm4MB_emit_copy_short + CMPL BX, $0x00000800 + JB repeat_two_offset_match_nolit_encodeBlockAsm4MB_emit_copy_short + +cant_repeat_two_offset_match_nolit_encodeBlockAsm4MB_emit_copy_short: + CMPL R9, $0x00000104 + JB repeat_three_match_nolit_encodeBlockAsm4MB_emit_copy_short + CMPL R9, $0x00010100 + JB repeat_four_match_nolit_encodeBlockAsm4MB_emit_copy_short + LEAL -65536(R9), R9 + MOVL R9, BX + MOVW $0x001d, (AX) + MOVW R9, 2(AX) + SARL $0x10, BX + MOVB BL, 4(AX) + ADDQ $0x05, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm4MB + +repeat_four_match_nolit_encodeBlockAsm4MB_emit_copy_short: + LEAL -256(R9), R9 + MOVW $0x0019, (AX) + MOVW R9, 2(AX) + ADDQ $0x04, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm4MB + +repeat_three_match_nolit_encodeBlockAsm4MB_emit_copy_short: + LEAL -4(R9), R9 + MOVW $0x0015, (AX) + MOVB R9, 2(AX) + ADDQ $0x03, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm4MB + +repeat_two_match_nolit_encodeBlockAsm4MB_emit_copy_short: + SHLL $0x02, R9 + ORL $0x01, R9 + MOVW R9, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm4MB + +repeat_two_offset_match_nolit_encodeBlockAsm4MB_emit_copy_short: + XORQ SI, SI + LEAL 1(SI)(R9*4), R9 + MOVB BL, 1(AX) + SARL $0x08, BX + SHLL $0x05, BX + ORL BX, R9 + MOVB R9, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm4MB + +two_byte_offset_short_match_nolit_encodeBlockAsm4MB: + MOVL R9, SI + SHLL $0x02, SI + CMPL R9, $0x0c + JAE emit_copy_three_match_nolit_encodeBlockAsm4MB + CMPL BX, $0x00000800 + JAE emit_copy_three_match_nolit_encodeBlockAsm4MB + LEAL -15(SI), SI + MOVB BL, 1(AX) + SHRL $0x08, BX + SHLL $0x05, BX + ORL BX, SI + MOVB SI, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm4MB + +emit_copy_three_match_nolit_encodeBlockAsm4MB: + LEAL -2(SI), SI + MOVB SI, (AX) + MOVW BX, 1(AX) + ADDQ $0x03, AX + +match_nolit_emitcopy_end_encodeBlockAsm4MB: + CMPL CX, 8(SP) + JAE emit_remainder_encodeBlockAsm4MB + MOVQ -2(DX)(CX*1), SI + CMPQ AX, (SP) + JB match_nolit_dst_ok_encodeBlockAsm4MB + MOVQ $0x00000000, ret+48(FP) + RET + +match_nolit_dst_ok_encodeBlockAsm4MB: + MOVQ $0x0000cf1bbcdcbf9b, R8 + MOVQ SI, DI + SHRQ $0x10, SI + MOVQ SI, BX + SHLQ $0x10, DI + IMULQ R8, DI + SHRQ $0x32, DI + SHLQ $0x10, BX + IMULQ R8, BX + SHRQ $0x32, BX + LEAL -2(CX), R8 + LEAQ 24(SP)(BX*4), R9 + MOVL (R9), BX + MOVL R8, 24(SP)(DI*4) + MOVL CX, (R9) + CMPL (DX)(BX*1), SI + JEQ match_nolit_loop_encodeBlockAsm4MB + INCL CX + JMP search_loop_encodeBlockAsm4MB + +emit_remainder_encodeBlockAsm4MB: + MOVQ src_len+32(FP), CX + SUBL 12(SP), CX + LEAQ 4(AX)(CX*1), CX + CMPQ CX, (SP) + JB emit_remainder_ok_encodeBlockAsm4MB + MOVQ $0x00000000, ret+48(FP) + RET + +emit_remainder_ok_encodeBlockAsm4MB: + MOVQ src_len+32(FP), CX + MOVL 12(SP), BX + CMPL BX, CX + JEQ emit_literal_done_emit_remainder_encodeBlockAsm4MB + MOVL CX, SI + MOVL CX, 12(SP) + LEAQ (DX)(BX*1), CX + SUBL BX, SI + LEAL -1(SI), DX + CMPL DX, $0x3c + JB one_byte_emit_remainder_encodeBlockAsm4MB + CMPL DX, $0x00000100 + JB two_bytes_emit_remainder_encodeBlockAsm4MB + CMPL DX, $0x00010000 + JB three_bytes_emit_remainder_encodeBlockAsm4MB + MOVL DX, BX + SHRL $0x10, BX + MOVB $0xf8, (AX) + MOVW DX, 1(AX) + MOVB BL, 3(AX) + ADDQ $0x04, AX + JMP memmove_long_emit_remainder_encodeBlockAsm4MB + +three_bytes_emit_remainder_encodeBlockAsm4MB: + MOVB $0xf4, (AX) + MOVW DX, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_emit_remainder_encodeBlockAsm4MB + +two_bytes_emit_remainder_encodeBlockAsm4MB: + MOVB $0xf0, (AX) + MOVB DL, 1(AX) + ADDQ $0x02, AX + CMPL DX, $0x40 + JB memmove_emit_remainder_encodeBlockAsm4MB + JMP memmove_long_emit_remainder_encodeBlockAsm4MB + +one_byte_emit_remainder_encodeBlockAsm4MB: + SHLB $0x02, DL + MOVB DL, (AX) + ADDQ $0x01, AX + +memmove_emit_remainder_encodeBlockAsm4MB: + LEAQ (AX)(SI*1), DX + MOVL SI, BX + + // genMemMoveShort + CMPQ BX, $0x03 + JB emit_lit_memmove_emit_remainder_encodeBlockAsm4MB_memmove_move_1or2 + JE emit_lit_memmove_emit_remainder_encodeBlockAsm4MB_memmove_move_3 + CMPQ BX, $0x08 + JB emit_lit_memmove_emit_remainder_encodeBlockAsm4MB_memmove_move_4through7 + CMPQ BX, $0x10 + JBE emit_lit_memmove_emit_remainder_encodeBlockAsm4MB_memmove_move_8through16 + CMPQ BX, $0x20 + JBE emit_lit_memmove_emit_remainder_encodeBlockAsm4MB_memmove_move_17through32 + JMP emit_lit_memmove_emit_remainder_encodeBlockAsm4MB_memmove_move_33through64 + +emit_lit_memmove_emit_remainder_encodeBlockAsm4MB_memmove_move_1or2: + MOVB (CX), SI + MOVB -1(CX)(BX*1), CL + MOVB SI, (AX) + MOVB CL, -1(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeBlockAsm4MB + +emit_lit_memmove_emit_remainder_encodeBlockAsm4MB_memmove_move_3: + MOVW (CX), SI + MOVB 2(CX), CL + MOVW SI, (AX) + MOVB CL, 2(AX) + JMP memmove_end_copy_emit_remainder_encodeBlockAsm4MB + +emit_lit_memmove_emit_remainder_encodeBlockAsm4MB_memmove_move_4through7: + MOVL (CX), SI + MOVL -4(CX)(BX*1), CX + MOVL SI, (AX) + MOVL CX, -4(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeBlockAsm4MB + +emit_lit_memmove_emit_remainder_encodeBlockAsm4MB_memmove_move_8through16: + MOVQ (CX), SI + MOVQ -8(CX)(BX*1), CX + MOVQ SI, (AX) + MOVQ CX, -8(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeBlockAsm4MB + +emit_lit_memmove_emit_remainder_encodeBlockAsm4MB_memmove_move_17through32: + MOVOU (CX), X0 + MOVOU -16(CX)(BX*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeBlockAsm4MB + +emit_lit_memmove_emit_remainder_encodeBlockAsm4MB_memmove_move_33through64: + MOVOU (CX), X0 + MOVOU 16(CX), X1 + MOVOU -32(CX)(BX*1), X2 + MOVOU -16(CX)(BX*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(BX*1) + MOVOU X3, -16(AX)(BX*1) + +memmove_end_copy_emit_remainder_encodeBlockAsm4MB: + MOVQ DX, AX + JMP emit_literal_done_emit_remainder_encodeBlockAsm4MB + +memmove_long_emit_remainder_encodeBlockAsm4MB: + LEAQ (AX)(SI*1), DX + MOVL SI, BX + + // genMemMoveLong + MOVOU (CX), X0 + MOVOU 16(CX), X1 + MOVOU -32(CX)(BX*1), X2 + MOVOU -16(CX)(BX*1), X3 + MOVQ BX, DI + SHRQ $0x05, DI + MOVQ AX, SI + ANDL $0x0000001f, SI + MOVQ $0x00000040, R8 + SUBQ SI, R8 + DECQ DI + JA emit_lit_memmove_long_emit_remainder_encodeBlockAsm4MBlarge_forward_sse_loop_32 + LEAQ -32(CX)(R8*1), SI + LEAQ -32(AX)(R8*1), R9 + +emit_lit_memmove_long_emit_remainder_encodeBlockAsm4MBlarge_big_loop_back: + MOVOU (SI), X4 + MOVOU 16(SI), X5 + MOVOA X4, (R9) + MOVOA X5, 16(R9) + ADDQ $0x20, R9 + ADDQ $0x20, SI + ADDQ $0x20, R8 + DECQ DI + JNA emit_lit_memmove_long_emit_remainder_encodeBlockAsm4MBlarge_big_loop_back + +emit_lit_memmove_long_emit_remainder_encodeBlockAsm4MBlarge_forward_sse_loop_32: + MOVOU -32(CX)(R8*1), X4 + MOVOU -16(CX)(R8*1), X5 + MOVOA X4, -32(AX)(R8*1) + MOVOA X5, -16(AX)(R8*1) + ADDQ $0x20, R8 + CMPQ BX, R8 + JAE emit_lit_memmove_long_emit_remainder_encodeBlockAsm4MBlarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(BX*1) + MOVOU X3, -16(AX)(BX*1) + MOVQ DX, AX + +emit_literal_done_emit_remainder_encodeBlockAsm4MB: + MOVQ dst_base+0(FP), CX + SUBQ CX, AX + MOVQ AX, ret+48(FP) + RET + +// func encodeBlockAsm12B(dst []byte, src []byte) int +// Requires: BMI, SSE2 +TEXT ·encodeBlockAsm12B(SB), $16408-56 + MOVQ dst_base+0(FP), AX + MOVQ $0x00000080, CX + LEAQ 24(SP), DX + PXOR X0, X0 + +zero_loop_encodeBlockAsm12B: + MOVOU X0, (DX) + MOVOU X0, 16(DX) + MOVOU X0, 32(DX) + MOVOU X0, 48(DX) + MOVOU X0, 64(DX) + MOVOU X0, 80(DX) + MOVOU X0, 96(DX) + MOVOU X0, 112(DX) + ADDQ $0x80, DX + DECQ CX + JNZ zero_loop_encodeBlockAsm12B + MOVL $0x00000000, 12(SP) + MOVQ src_len+32(FP), CX + LEAQ -9(CX), DX + LEAQ -8(CX), BX + MOVL BX, 8(SP) + SHRQ $0x05, CX + SUBL CX, DX + LEAQ (AX)(DX*1), DX + MOVQ DX, (SP) + MOVL $0x00000001, CX + MOVL CX, 16(SP) + MOVQ src_base+24(FP), DX + +search_loop_encodeBlockAsm12B: + MOVL CX, BX + SUBL 12(SP), BX + SHRL $0x05, BX + LEAL 4(CX)(BX*1), BX + CMPL BX, 8(SP) + JAE emit_remainder_encodeBlockAsm12B + MOVQ (DX)(CX*1), SI + MOVL BX, 20(SP) + MOVQ $0x000000cf1bbcdcbb, R8 + MOVQ SI, R9 + MOVQ SI, R10 + SHRQ $0x08, R10 + SHLQ $0x18, R9 + IMULQ R8, R9 + SHRQ $0x34, R9 + SHLQ $0x18, R10 + IMULQ R8, R10 + SHRQ $0x34, R10 + MOVL 24(SP)(R9*4), BX + MOVL 24(SP)(R10*4), DI + MOVL CX, 24(SP)(R9*4) + LEAL 1(CX), R9 + MOVL R9, 24(SP)(R10*4) + MOVQ SI, R9 + SHRQ $0x10, R9 + SHLQ $0x18, R9 + IMULQ R8, R9 + SHRQ $0x34, R9 + MOVL CX, R8 + SUBL 16(SP), R8 + MOVL 1(DX)(R8*1), R10 + MOVQ SI, R8 + SHRQ $0x08, R8 + CMPL R8, R10 + JNE no_repeat_found_encodeBlockAsm12B + LEAL 1(CX), SI + MOVL 12(SP), DI + MOVL SI, BX + SUBL 16(SP), BX + JZ repeat_extend_back_end_encodeBlockAsm12B + +repeat_extend_back_loop_encodeBlockAsm12B: + CMPL SI, DI + JBE repeat_extend_back_end_encodeBlockAsm12B + MOVB -1(DX)(BX*1), R8 + MOVB -1(DX)(SI*1), R9 + CMPB R8, R9 + JNE repeat_extend_back_end_encodeBlockAsm12B + LEAL -1(SI), SI + DECL BX + JNZ repeat_extend_back_loop_encodeBlockAsm12B + +repeat_extend_back_end_encodeBlockAsm12B: + MOVL SI, BX + SUBL 12(SP), BX + LEAQ 3(AX)(BX*1), BX + CMPQ BX, (SP) + JB repeat_dst_size_check_encodeBlockAsm12B + MOVQ $0x00000000, ret+48(FP) + RET + +repeat_dst_size_check_encodeBlockAsm12B: + MOVL 12(SP), BX + CMPL BX, SI + JEQ emit_literal_done_repeat_emit_encodeBlockAsm12B + MOVL SI, R8 + MOVL SI, 12(SP) + LEAQ (DX)(BX*1), R9 + SUBL BX, R8 + LEAL -1(R8), BX + CMPL BX, $0x3c + JB one_byte_repeat_emit_encodeBlockAsm12B + CMPL BX, $0x00000100 + JB two_bytes_repeat_emit_encodeBlockAsm12B + JB three_bytes_repeat_emit_encodeBlockAsm12B + +three_bytes_repeat_emit_encodeBlockAsm12B: + MOVB $0xf4, (AX) + MOVW BX, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_repeat_emit_encodeBlockAsm12B + +two_bytes_repeat_emit_encodeBlockAsm12B: + MOVB $0xf0, (AX) + MOVB BL, 1(AX) + ADDQ $0x02, AX + CMPL BX, $0x40 + JB memmove_repeat_emit_encodeBlockAsm12B + JMP memmove_long_repeat_emit_encodeBlockAsm12B + +one_byte_repeat_emit_encodeBlockAsm12B: + SHLB $0x02, BL + MOVB BL, (AX) + ADDQ $0x01, AX + +memmove_repeat_emit_encodeBlockAsm12B: + LEAQ (AX)(R8*1), BX + + // genMemMoveShort + CMPQ R8, $0x08 + JBE emit_lit_memmove_repeat_emit_encodeBlockAsm12B_memmove_move_8 + CMPQ R8, $0x10 + JBE emit_lit_memmove_repeat_emit_encodeBlockAsm12B_memmove_move_8through16 + CMPQ R8, $0x20 + JBE emit_lit_memmove_repeat_emit_encodeBlockAsm12B_memmove_move_17through32 + JMP emit_lit_memmove_repeat_emit_encodeBlockAsm12B_memmove_move_33through64 + +emit_lit_memmove_repeat_emit_encodeBlockAsm12B_memmove_move_8: + MOVQ (R9), R10 + MOVQ R10, (AX) + JMP memmove_end_copy_repeat_emit_encodeBlockAsm12B + +emit_lit_memmove_repeat_emit_encodeBlockAsm12B_memmove_move_8through16: + MOVQ (R9), R10 + MOVQ -8(R9)(R8*1), R9 + MOVQ R10, (AX) + MOVQ R9, -8(AX)(R8*1) + JMP memmove_end_copy_repeat_emit_encodeBlockAsm12B + +emit_lit_memmove_repeat_emit_encodeBlockAsm12B_memmove_move_17through32: + MOVOU (R9), X0 + MOVOU -16(R9)(R8*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(R8*1) + JMP memmove_end_copy_repeat_emit_encodeBlockAsm12B + +emit_lit_memmove_repeat_emit_encodeBlockAsm12B_memmove_move_33through64: + MOVOU (R9), X0 + MOVOU 16(R9), X1 + MOVOU -32(R9)(R8*1), X2 + MOVOU -16(R9)(R8*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + +memmove_end_copy_repeat_emit_encodeBlockAsm12B: + MOVQ BX, AX + JMP emit_literal_done_repeat_emit_encodeBlockAsm12B + +memmove_long_repeat_emit_encodeBlockAsm12B: + LEAQ (AX)(R8*1), BX + + // genMemMoveLong + MOVOU (R9), X0 + MOVOU 16(R9), X1 + MOVOU -32(R9)(R8*1), X2 + MOVOU -16(R9)(R8*1), X3 + MOVQ R8, R11 + SHRQ $0x05, R11 + MOVQ AX, R10 + ANDL $0x0000001f, R10 + MOVQ $0x00000040, R12 + SUBQ R10, R12 + DECQ R11 + JA emit_lit_memmove_long_repeat_emit_encodeBlockAsm12Blarge_forward_sse_loop_32 + LEAQ -32(R9)(R12*1), R10 + LEAQ -32(AX)(R12*1), R13 + +emit_lit_memmove_long_repeat_emit_encodeBlockAsm12Blarge_big_loop_back: + MOVOU (R10), X4 + MOVOU 16(R10), X5 + MOVOA X4, (R13) + MOVOA X5, 16(R13) + ADDQ $0x20, R13 + ADDQ $0x20, R10 + ADDQ $0x20, R12 + DECQ R11 + JNA emit_lit_memmove_long_repeat_emit_encodeBlockAsm12Blarge_big_loop_back + +emit_lit_memmove_long_repeat_emit_encodeBlockAsm12Blarge_forward_sse_loop_32: + MOVOU -32(R9)(R12*1), X4 + MOVOU -16(R9)(R12*1), X5 + MOVOA X4, -32(AX)(R12*1) + MOVOA X5, -16(AX)(R12*1) + ADDQ $0x20, R12 + CMPQ R8, R12 + JAE emit_lit_memmove_long_repeat_emit_encodeBlockAsm12Blarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + MOVQ BX, AX + +emit_literal_done_repeat_emit_encodeBlockAsm12B: + ADDL $0x05, CX + MOVL CX, BX + SUBL 16(SP), BX + MOVQ src_len+32(FP), R8 + SUBL CX, R8 + LEAQ (DX)(CX*1), R9 + LEAQ (DX)(BX*1), BX + + // matchLen + XORL R11, R11 + +matchlen_loopback_16_repeat_extend_encodeBlockAsm12B: + CMPL R8, $0x10 + JB matchlen_match8_repeat_extend_encodeBlockAsm12B + MOVQ (R9)(R11*1), R10 + MOVQ 8(R9)(R11*1), R12 + XORQ (BX)(R11*1), R10 + JNZ matchlen_bsf_8_repeat_extend_encodeBlockAsm12B + XORQ 8(BX)(R11*1), R12 + JNZ matchlen_bsf_16repeat_extend_encodeBlockAsm12B + LEAL -16(R8), R8 + LEAL 16(R11), R11 + JMP matchlen_loopback_16_repeat_extend_encodeBlockAsm12B + +matchlen_bsf_16repeat_extend_encodeBlockAsm12B: +#ifdef GOAMD64_v3 + TZCNTQ R12, R12 + +#else + BSFQ R12, R12 + +#endif + SARQ $0x03, R12 + LEAL 8(R11)(R12*1), R11 + JMP repeat_extend_forward_end_encodeBlockAsm12B + +matchlen_match8_repeat_extend_encodeBlockAsm12B: + CMPL R8, $0x08 + JB matchlen_match4_repeat_extend_encodeBlockAsm12B + MOVQ (R9)(R11*1), R10 + XORQ (BX)(R11*1), R10 + JNZ matchlen_bsf_8_repeat_extend_encodeBlockAsm12B + LEAL -8(R8), R8 + LEAL 8(R11), R11 + JMP matchlen_match4_repeat_extend_encodeBlockAsm12B + +matchlen_bsf_8_repeat_extend_encodeBlockAsm12B: +#ifdef GOAMD64_v3 + TZCNTQ R10, R10 + +#else + BSFQ R10, R10 + +#endif + SARQ $0x03, R10 + LEAL (R11)(R10*1), R11 + JMP repeat_extend_forward_end_encodeBlockAsm12B + +matchlen_match4_repeat_extend_encodeBlockAsm12B: + CMPL R8, $0x04 + JB matchlen_match2_repeat_extend_encodeBlockAsm12B + MOVL (R9)(R11*1), R10 + CMPL (BX)(R11*1), R10 + JNE matchlen_match2_repeat_extend_encodeBlockAsm12B + LEAL -4(R8), R8 + LEAL 4(R11), R11 + +matchlen_match2_repeat_extend_encodeBlockAsm12B: + CMPL R8, $0x01 + JE matchlen_match1_repeat_extend_encodeBlockAsm12B + JB repeat_extend_forward_end_encodeBlockAsm12B + MOVW (R9)(R11*1), R10 + CMPW (BX)(R11*1), R10 + JNE matchlen_match1_repeat_extend_encodeBlockAsm12B + LEAL 2(R11), R11 + SUBL $0x02, R8 + JZ repeat_extend_forward_end_encodeBlockAsm12B + +matchlen_match1_repeat_extend_encodeBlockAsm12B: + MOVB (R9)(R11*1), R10 + CMPB (BX)(R11*1), R10 + JNE repeat_extend_forward_end_encodeBlockAsm12B + LEAL 1(R11), R11 + +repeat_extend_forward_end_encodeBlockAsm12B: + ADDL R11, CX + MOVL CX, BX + SUBL SI, BX + MOVL 16(SP), SI + TESTL DI, DI + JZ repeat_as_copy_encodeBlockAsm12B + + // emitRepeat + MOVL BX, DI + LEAL -4(BX), BX + CMPL DI, $0x08 + JBE repeat_two_match_repeat_encodeBlockAsm12B + CMPL DI, $0x0c + JAE cant_repeat_two_offset_match_repeat_encodeBlockAsm12B + CMPL SI, $0x00000800 + JB repeat_two_offset_match_repeat_encodeBlockAsm12B + +cant_repeat_two_offset_match_repeat_encodeBlockAsm12B: + CMPL BX, $0x00000104 + JB repeat_three_match_repeat_encodeBlockAsm12B + LEAL -256(BX), BX + MOVW $0x0019, (AX) + MOVW BX, 2(AX) + ADDQ $0x04, AX + JMP repeat_end_emit_encodeBlockAsm12B + +repeat_three_match_repeat_encodeBlockAsm12B: + LEAL -4(BX), BX + MOVW $0x0015, (AX) + MOVB BL, 2(AX) + ADDQ $0x03, AX + JMP repeat_end_emit_encodeBlockAsm12B + +repeat_two_match_repeat_encodeBlockAsm12B: + SHLL $0x02, BX + ORL $0x01, BX + MOVW BX, (AX) + ADDQ $0x02, AX + JMP repeat_end_emit_encodeBlockAsm12B + +repeat_two_offset_match_repeat_encodeBlockAsm12B: + XORQ DI, DI + LEAL 1(DI)(BX*4), BX + MOVB SI, 1(AX) + SARL $0x08, SI + SHLL $0x05, SI + ORL SI, BX + MOVB BL, (AX) + ADDQ $0x02, AX + JMP repeat_end_emit_encodeBlockAsm12B + +repeat_as_copy_encodeBlockAsm12B: + // emitCopy + CMPL BX, $0x40 + JBE two_byte_offset_short_repeat_as_copy_encodeBlockAsm12B + CMPL SI, $0x00000800 + JAE long_offset_short_repeat_as_copy_encodeBlockAsm12B + MOVL $0x00000001, DI + LEAL 16(DI), DI + MOVB SI, 1(AX) + SHRL $0x08, SI + SHLL $0x05, SI + ORL SI, DI + MOVB DI, (AX) + ADDQ $0x02, AX + SUBL $0x08, BX + + // emitRepeat + LEAL -4(BX), BX + JMP cant_repeat_two_offset_repeat_as_copy_encodeBlockAsm12B_emit_copy_short_2b + MOVL BX, DI + LEAL -4(BX), BX + CMPL DI, $0x08 + JBE repeat_two_repeat_as_copy_encodeBlockAsm12B_emit_copy_short_2b + CMPL DI, $0x0c + JAE cant_repeat_two_offset_repeat_as_copy_encodeBlockAsm12B_emit_copy_short_2b + CMPL SI, $0x00000800 + JB repeat_two_offset_repeat_as_copy_encodeBlockAsm12B_emit_copy_short_2b + +cant_repeat_two_offset_repeat_as_copy_encodeBlockAsm12B_emit_copy_short_2b: + CMPL BX, $0x00000104 + JB repeat_three_repeat_as_copy_encodeBlockAsm12B_emit_copy_short_2b + LEAL -256(BX), BX + MOVW $0x0019, (AX) + MOVW BX, 2(AX) + ADDQ $0x04, AX + JMP repeat_end_emit_encodeBlockAsm12B + +repeat_three_repeat_as_copy_encodeBlockAsm12B_emit_copy_short_2b: + LEAL -4(BX), BX + MOVW $0x0015, (AX) + MOVB BL, 2(AX) + ADDQ $0x03, AX + JMP repeat_end_emit_encodeBlockAsm12B + +repeat_two_repeat_as_copy_encodeBlockAsm12B_emit_copy_short_2b: + SHLL $0x02, BX + ORL $0x01, BX + MOVW BX, (AX) + ADDQ $0x02, AX + JMP repeat_end_emit_encodeBlockAsm12B + +repeat_two_offset_repeat_as_copy_encodeBlockAsm12B_emit_copy_short_2b: + XORQ DI, DI + LEAL 1(DI)(BX*4), BX + MOVB SI, 1(AX) + SARL $0x08, SI + SHLL $0x05, SI + ORL SI, BX + MOVB BL, (AX) + ADDQ $0x02, AX + JMP repeat_end_emit_encodeBlockAsm12B + +long_offset_short_repeat_as_copy_encodeBlockAsm12B: + MOVB $0xee, (AX) + MOVW SI, 1(AX) + LEAL -60(BX), BX + ADDQ $0x03, AX + + // emitRepeat + MOVL BX, DI + LEAL -4(BX), BX + CMPL DI, $0x08 + JBE repeat_two_repeat_as_copy_encodeBlockAsm12B_emit_copy_short + CMPL DI, $0x0c + JAE cant_repeat_two_offset_repeat_as_copy_encodeBlockAsm12B_emit_copy_short + CMPL SI, $0x00000800 + JB repeat_two_offset_repeat_as_copy_encodeBlockAsm12B_emit_copy_short + +cant_repeat_two_offset_repeat_as_copy_encodeBlockAsm12B_emit_copy_short: + CMPL BX, $0x00000104 + JB repeat_three_repeat_as_copy_encodeBlockAsm12B_emit_copy_short + LEAL -256(BX), BX + MOVW $0x0019, (AX) + MOVW BX, 2(AX) + ADDQ $0x04, AX + JMP repeat_end_emit_encodeBlockAsm12B + +repeat_three_repeat_as_copy_encodeBlockAsm12B_emit_copy_short: + LEAL -4(BX), BX + MOVW $0x0015, (AX) + MOVB BL, 2(AX) + ADDQ $0x03, AX + JMP repeat_end_emit_encodeBlockAsm12B + +repeat_two_repeat_as_copy_encodeBlockAsm12B_emit_copy_short: + SHLL $0x02, BX + ORL $0x01, BX + MOVW BX, (AX) + ADDQ $0x02, AX + JMP repeat_end_emit_encodeBlockAsm12B + +repeat_two_offset_repeat_as_copy_encodeBlockAsm12B_emit_copy_short: + XORQ DI, DI + LEAL 1(DI)(BX*4), BX + MOVB SI, 1(AX) + SARL $0x08, SI + SHLL $0x05, SI + ORL SI, BX + MOVB BL, (AX) + ADDQ $0x02, AX + JMP repeat_end_emit_encodeBlockAsm12B + +two_byte_offset_short_repeat_as_copy_encodeBlockAsm12B: + MOVL BX, DI + SHLL $0x02, DI + CMPL BX, $0x0c + JAE emit_copy_three_repeat_as_copy_encodeBlockAsm12B + CMPL SI, $0x00000800 + JAE emit_copy_three_repeat_as_copy_encodeBlockAsm12B + LEAL -15(DI), DI + MOVB SI, 1(AX) + SHRL $0x08, SI + SHLL $0x05, SI + ORL SI, DI + MOVB DI, (AX) + ADDQ $0x02, AX + JMP repeat_end_emit_encodeBlockAsm12B + +emit_copy_three_repeat_as_copy_encodeBlockAsm12B: + LEAL -2(DI), DI + MOVB DI, (AX) + MOVW SI, 1(AX) + ADDQ $0x03, AX + +repeat_end_emit_encodeBlockAsm12B: + MOVL CX, 12(SP) + JMP search_loop_encodeBlockAsm12B + +no_repeat_found_encodeBlockAsm12B: + CMPL (DX)(BX*1), SI + JEQ candidate_match_encodeBlockAsm12B + SHRQ $0x08, SI + MOVL 24(SP)(R9*4), BX + LEAL 2(CX), R8 + CMPL (DX)(DI*1), SI + JEQ candidate2_match_encodeBlockAsm12B + MOVL R8, 24(SP)(R9*4) + SHRQ $0x08, SI + CMPL (DX)(BX*1), SI + JEQ candidate3_match_encodeBlockAsm12B + MOVL 20(SP), CX + JMP search_loop_encodeBlockAsm12B + +candidate3_match_encodeBlockAsm12B: + ADDL $0x02, CX + JMP candidate_match_encodeBlockAsm12B + +candidate2_match_encodeBlockAsm12B: + MOVL R8, 24(SP)(R9*4) + INCL CX + MOVL DI, BX + +candidate_match_encodeBlockAsm12B: + MOVL 12(SP), SI + TESTL BX, BX + JZ match_extend_back_end_encodeBlockAsm12B + +match_extend_back_loop_encodeBlockAsm12B: + CMPL CX, SI + JBE match_extend_back_end_encodeBlockAsm12B + MOVB -1(DX)(BX*1), DI + MOVB -1(DX)(CX*1), R8 + CMPB DI, R8 + JNE match_extend_back_end_encodeBlockAsm12B + LEAL -1(CX), CX + DECL BX + JZ match_extend_back_end_encodeBlockAsm12B + JMP match_extend_back_loop_encodeBlockAsm12B + +match_extend_back_end_encodeBlockAsm12B: + MOVL CX, SI + SUBL 12(SP), SI + LEAQ 3(AX)(SI*1), SI + CMPQ SI, (SP) + JB match_dst_size_check_encodeBlockAsm12B + MOVQ $0x00000000, ret+48(FP) + RET + +match_dst_size_check_encodeBlockAsm12B: + MOVL CX, SI + MOVL 12(SP), DI + CMPL DI, SI + JEQ emit_literal_done_match_emit_encodeBlockAsm12B + MOVL SI, R8 + MOVL SI, 12(SP) + LEAQ (DX)(DI*1), SI + SUBL DI, R8 + LEAL -1(R8), DI + CMPL DI, $0x3c + JB one_byte_match_emit_encodeBlockAsm12B + CMPL DI, $0x00000100 + JB two_bytes_match_emit_encodeBlockAsm12B + JB three_bytes_match_emit_encodeBlockAsm12B + +three_bytes_match_emit_encodeBlockAsm12B: + MOVB $0xf4, (AX) + MOVW DI, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_match_emit_encodeBlockAsm12B + +two_bytes_match_emit_encodeBlockAsm12B: + MOVB $0xf0, (AX) + MOVB DI, 1(AX) + ADDQ $0x02, AX + CMPL DI, $0x40 + JB memmove_match_emit_encodeBlockAsm12B + JMP memmove_long_match_emit_encodeBlockAsm12B + +one_byte_match_emit_encodeBlockAsm12B: + SHLB $0x02, DI + MOVB DI, (AX) + ADDQ $0x01, AX + +memmove_match_emit_encodeBlockAsm12B: + LEAQ (AX)(R8*1), DI + + // genMemMoveShort + CMPQ R8, $0x08 + JBE emit_lit_memmove_match_emit_encodeBlockAsm12B_memmove_move_8 + CMPQ R8, $0x10 + JBE emit_lit_memmove_match_emit_encodeBlockAsm12B_memmove_move_8through16 + CMPQ R8, $0x20 + JBE emit_lit_memmove_match_emit_encodeBlockAsm12B_memmove_move_17through32 + JMP emit_lit_memmove_match_emit_encodeBlockAsm12B_memmove_move_33through64 + +emit_lit_memmove_match_emit_encodeBlockAsm12B_memmove_move_8: + MOVQ (SI), R9 + MOVQ R9, (AX) + JMP memmove_end_copy_match_emit_encodeBlockAsm12B + +emit_lit_memmove_match_emit_encodeBlockAsm12B_memmove_move_8through16: + MOVQ (SI), R9 + MOVQ -8(SI)(R8*1), SI + MOVQ R9, (AX) + MOVQ SI, -8(AX)(R8*1) + JMP memmove_end_copy_match_emit_encodeBlockAsm12B + +emit_lit_memmove_match_emit_encodeBlockAsm12B_memmove_move_17through32: + MOVOU (SI), X0 + MOVOU -16(SI)(R8*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(R8*1) + JMP memmove_end_copy_match_emit_encodeBlockAsm12B + +emit_lit_memmove_match_emit_encodeBlockAsm12B_memmove_move_33through64: + MOVOU (SI), X0 + MOVOU 16(SI), X1 + MOVOU -32(SI)(R8*1), X2 + MOVOU -16(SI)(R8*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + +memmove_end_copy_match_emit_encodeBlockAsm12B: + MOVQ DI, AX + JMP emit_literal_done_match_emit_encodeBlockAsm12B + +memmove_long_match_emit_encodeBlockAsm12B: + LEAQ (AX)(R8*1), DI + + // genMemMoveLong + MOVOU (SI), X0 + MOVOU 16(SI), X1 + MOVOU -32(SI)(R8*1), X2 + MOVOU -16(SI)(R8*1), X3 + MOVQ R8, R10 + SHRQ $0x05, R10 + MOVQ AX, R9 + ANDL $0x0000001f, R9 + MOVQ $0x00000040, R11 + SUBQ R9, R11 + DECQ R10 + JA emit_lit_memmove_long_match_emit_encodeBlockAsm12Blarge_forward_sse_loop_32 + LEAQ -32(SI)(R11*1), R9 + LEAQ -32(AX)(R11*1), R12 + +emit_lit_memmove_long_match_emit_encodeBlockAsm12Blarge_big_loop_back: + MOVOU (R9), X4 + MOVOU 16(R9), X5 + MOVOA X4, (R12) + MOVOA X5, 16(R12) + ADDQ $0x20, R12 + ADDQ $0x20, R9 + ADDQ $0x20, R11 + DECQ R10 + JNA emit_lit_memmove_long_match_emit_encodeBlockAsm12Blarge_big_loop_back + +emit_lit_memmove_long_match_emit_encodeBlockAsm12Blarge_forward_sse_loop_32: + MOVOU -32(SI)(R11*1), X4 + MOVOU -16(SI)(R11*1), X5 + MOVOA X4, -32(AX)(R11*1) + MOVOA X5, -16(AX)(R11*1) + ADDQ $0x20, R11 + CMPQ R8, R11 + JAE emit_lit_memmove_long_match_emit_encodeBlockAsm12Blarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + MOVQ DI, AX + +emit_literal_done_match_emit_encodeBlockAsm12B: +match_nolit_loop_encodeBlockAsm12B: + MOVL CX, SI + SUBL BX, SI + MOVL SI, 16(SP) + ADDL $0x04, CX + ADDL $0x04, BX + MOVQ src_len+32(FP), SI + SUBL CX, SI + LEAQ (DX)(CX*1), DI + LEAQ (DX)(BX*1), BX + + // matchLen + XORL R9, R9 + +matchlen_loopback_16_match_nolit_encodeBlockAsm12B: + CMPL SI, $0x10 + JB matchlen_match8_match_nolit_encodeBlockAsm12B + MOVQ (DI)(R9*1), R8 + MOVQ 8(DI)(R9*1), R10 + XORQ (BX)(R9*1), R8 + JNZ matchlen_bsf_8_match_nolit_encodeBlockAsm12B + XORQ 8(BX)(R9*1), R10 + JNZ matchlen_bsf_16match_nolit_encodeBlockAsm12B + LEAL -16(SI), SI + LEAL 16(R9), R9 + JMP matchlen_loopback_16_match_nolit_encodeBlockAsm12B + +matchlen_bsf_16match_nolit_encodeBlockAsm12B: +#ifdef GOAMD64_v3 + TZCNTQ R10, R10 + +#else + BSFQ R10, R10 + +#endif + SARQ $0x03, R10 + LEAL 8(R9)(R10*1), R9 + JMP match_nolit_end_encodeBlockAsm12B + +matchlen_match8_match_nolit_encodeBlockAsm12B: + CMPL SI, $0x08 + JB matchlen_match4_match_nolit_encodeBlockAsm12B + MOVQ (DI)(R9*1), R8 + XORQ (BX)(R9*1), R8 + JNZ matchlen_bsf_8_match_nolit_encodeBlockAsm12B + LEAL -8(SI), SI + LEAL 8(R9), R9 + JMP matchlen_match4_match_nolit_encodeBlockAsm12B + +matchlen_bsf_8_match_nolit_encodeBlockAsm12B: +#ifdef GOAMD64_v3 + TZCNTQ R8, R8 + +#else + BSFQ R8, R8 + +#endif + SARQ $0x03, R8 + LEAL (R9)(R8*1), R9 + JMP match_nolit_end_encodeBlockAsm12B + +matchlen_match4_match_nolit_encodeBlockAsm12B: + CMPL SI, $0x04 + JB matchlen_match2_match_nolit_encodeBlockAsm12B + MOVL (DI)(R9*1), R8 + CMPL (BX)(R9*1), R8 + JNE matchlen_match2_match_nolit_encodeBlockAsm12B + LEAL -4(SI), SI + LEAL 4(R9), R9 + +matchlen_match2_match_nolit_encodeBlockAsm12B: + CMPL SI, $0x01 + JE matchlen_match1_match_nolit_encodeBlockAsm12B + JB match_nolit_end_encodeBlockAsm12B + MOVW (DI)(R9*1), R8 + CMPW (BX)(R9*1), R8 + JNE matchlen_match1_match_nolit_encodeBlockAsm12B + LEAL 2(R9), R9 + SUBL $0x02, SI + JZ match_nolit_end_encodeBlockAsm12B + +matchlen_match1_match_nolit_encodeBlockAsm12B: + MOVB (DI)(R9*1), R8 + CMPB (BX)(R9*1), R8 + JNE match_nolit_end_encodeBlockAsm12B + LEAL 1(R9), R9 + +match_nolit_end_encodeBlockAsm12B: + ADDL R9, CX + MOVL 16(SP), BX + ADDL $0x04, R9 + MOVL CX, 12(SP) + + // emitCopy + CMPL R9, $0x40 + JBE two_byte_offset_short_match_nolit_encodeBlockAsm12B + CMPL BX, $0x00000800 + JAE long_offset_short_match_nolit_encodeBlockAsm12B + MOVL $0x00000001, SI + LEAL 16(SI), SI + MOVB BL, 1(AX) + SHRL $0x08, BX + SHLL $0x05, BX + ORL BX, SI + MOVB SI, (AX) + ADDQ $0x02, AX + SUBL $0x08, R9 + + // emitRepeat + LEAL -4(R9), R9 + JMP cant_repeat_two_offset_match_nolit_encodeBlockAsm12B_emit_copy_short_2b + MOVL R9, SI + LEAL -4(R9), R9 + CMPL SI, $0x08 + JBE repeat_two_match_nolit_encodeBlockAsm12B_emit_copy_short_2b + CMPL SI, $0x0c + JAE cant_repeat_two_offset_match_nolit_encodeBlockAsm12B_emit_copy_short_2b + CMPL BX, $0x00000800 + JB repeat_two_offset_match_nolit_encodeBlockAsm12B_emit_copy_short_2b + +cant_repeat_two_offset_match_nolit_encodeBlockAsm12B_emit_copy_short_2b: + CMPL R9, $0x00000104 + JB repeat_three_match_nolit_encodeBlockAsm12B_emit_copy_short_2b + LEAL -256(R9), R9 + MOVW $0x0019, (AX) + MOVW R9, 2(AX) + ADDQ $0x04, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm12B + +repeat_three_match_nolit_encodeBlockAsm12B_emit_copy_short_2b: + LEAL -4(R9), R9 + MOVW $0x0015, (AX) + MOVB R9, 2(AX) + ADDQ $0x03, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm12B + +repeat_two_match_nolit_encodeBlockAsm12B_emit_copy_short_2b: + SHLL $0x02, R9 + ORL $0x01, R9 + MOVW R9, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm12B + +repeat_two_offset_match_nolit_encodeBlockAsm12B_emit_copy_short_2b: + XORQ SI, SI + LEAL 1(SI)(R9*4), R9 + MOVB BL, 1(AX) + SARL $0x08, BX + SHLL $0x05, BX + ORL BX, R9 + MOVB R9, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm12B + +long_offset_short_match_nolit_encodeBlockAsm12B: + MOVB $0xee, (AX) + MOVW BX, 1(AX) + LEAL -60(R9), R9 + ADDQ $0x03, AX + + // emitRepeat + MOVL R9, SI + LEAL -4(R9), R9 + CMPL SI, $0x08 + JBE repeat_two_match_nolit_encodeBlockAsm12B_emit_copy_short + CMPL SI, $0x0c + JAE cant_repeat_two_offset_match_nolit_encodeBlockAsm12B_emit_copy_short + CMPL BX, $0x00000800 + JB repeat_two_offset_match_nolit_encodeBlockAsm12B_emit_copy_short + +cant_repeat_two_offset_match_nolit_encodeBlockAsm12B_emit_copy_short: + CMPL R9, $0x00000104 + JB repeat_three_match_nolit_encodeBlockAsm12B_emit_copy_short + LEAL -256(R9), R9 + MOVW $0x0019, (AX) + MOVW R9, 2(AX) + ADDQ $0x04, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm12B + +repeat_three_match_nolit_encodeBlockAsm12B_emit_copy_short: + LEAL -4(R9), R9 + MOVW $0x0015, (AX) + MOVB R9, 2(AX) + ADDQ $0x03, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm12B + +repeat_two_match_nolit_encodeBlockAsm12B_emit_copy_short: + SHLL $0x02, R9 + ORL $0x01, R9 + MOVW R9, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm12B + +repeat_two_offset_match_nolit_encodeBlockAsm12B_emit_copy_short: + XORQ SI, SI + LEAL 1(SI)(R9*4), R9 + MOVB BL, 1(AX) + SARL $0x08, BX + SHLL $0x05, BX + ORL BX, R9 + MOVB R9, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm12B + +two_byte_offset_short_match_nolit_encodeBlockAsm12B: + MOVL R9, SI + SHLL $0x02, SI + CMPL R9, $0x0c + JAE emit_copy_three_match_nolit_encodeBlockAsm12B + CMPL BX, $0x00000800 + JAE emit_copy_three_match_nolit_encodeBlockAsm12B + LEAL -15(SI), SI + MOVB BL, 1(AX) + SHRL $0x08, BX + SHLL $0x05, BX + ORL BX, SI + MOVB SI, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm12B + +emit_copy_three_match_nolit_encodeBlockAsm12B: + LEAL -2(SI), SI + MOVB SI, (AX) + MOVW BX, 1(AX) + ADDQ $0x03, AX + +match_nolit_emitcopy_end_encodeBlockAsm12B: + CMPL CX, 8(SP) + JAE emit_remainder_encodeBlockAsm12B + MOVQ -2(DX)(CX*1), SI + CMPQ AX, (SP) + JB match_nolit_dst_ok_encodeBlockAsm12B + MOVQ $0x00000000, ret+48(FP) + RET + +match_nolit_dst_ok_encodeBlockAsm12B: + MOVQ $0x000000cf1bbcdcbb, R8 + MOVQ SI, DI + SHRQ $0x10, SI + MOVQ SI, BX + SHLQ $0x18, DI + IMULQ R8, DI + SHRQ $0x34, DI + SHLQ $0x18, BX + IMULQ R8, BX + SHRQ $0x34, BX + LEAL -2(CX), R8 + LEAQ 24(SP)(BX*4), R9 + MOVL (R9), BX + MOVL R8, 24(SP)(DI*4) + MOVL CX, (R9) + CMPL (DX)(BX*1), SI + JEQ match_nolit_loop_encodeBlockAsm12B + INCL CX + JMP search_loop_encodeBlockAsm12B + +emit_remainder_encodeBlockAsm12B: + MOVQ src_len+32(FP), CX + SUBL 12(SP), CX + LEAQ 3(AX)(CX*1), CX + CMPQ CX, (SP) + JB emit_remainder_ok_encodeBlockAsm12B + MOVQ $0x00000000, ret+48(FP) + RET + +emit_remainder_ok_encodeBlockAsm12B: + MOVQ src_len+32(FP), CX + MOVL 12(SP), BX + CMPL BX, CX + JEQ emit_literal_done_emit_remainder_encodeBlockAsm12B + MOVL CX, SI + MOVL CX, 12(SP) + LEAQ (DX)(BX*1), CX + SUBL BX, SI + LEAL -1(SI), DX + CMPL DX, $0x3c + JB one_byte_emit_remainder_encodeBlockAsm12B + CMPL DX, $0x00000100 + JB two_bytes_emit_remainder_encodeBlockAsm12B + JB three_bytes_emit_remainder_encodeBlockAsm12B + +three_bytes_emit_remainder_encodeBlockAsm12B: + MOVB $0xf4, (AX) + MOVW DX, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_emit_remainder_encodeBlockAsm12B + +two_bytes_emit_remainder_encodeBlockAsm12B: + MOVB $0xf0, (AX) + MOVB DL, 1(AX) + ADDQ $0x02, AX + CMPL DX, $0x40 + JB memmove_emit_remainder_encodeBlockAsm12B + JMP memmove_long_emit_remainder_encodeBlockAsm12B + +one_byte_emit_remainder_encodeBlockAsm12B: + SHLB $0x02, DL + MOVB DL, (AX) + ADDQ $0x01, AX + +memmove_emit_remainder_encodeBlockAsm12B: + LEAQ (AX)(SI*1), DX + MOVL SI, BX + + // genMemMoveShort + CMPQ BX, $0x03 + JB emit_lit_memmove_emit_remainder_encodeBlockAsm12B_memmove_move_1or2 + JE emit_lit_memmove_emit_remainder_encodeBlockAsm12B_memmove_move_3 + CMPQ BX, $0x08 + JB emit_lit_memmove_emit_remainder_encodeBlockAsm12B_memmove_move_4through7 + CMPQ BX, $0x10 + JBE emit_lit_memmove_emit_remainder_encodeBlockAsm12B_memmove_move_8through16 + CMPQ BX, $0x20 + JBE emit_lit_memmove_emit_remainder_encodeBlockAsm12B_memmove_move_17through32 + JMP emit_lit_memmove_emit_remainder_encodeBlockAsm12B_memmove_move_33through64 + +emit_lit_memmove_emit_remainder_encodeBlockAsm12B_memmove_move_1or2: + MOVB (CX), SI + MOVB -1(CX)(BX*1), CL + MOVB SI, (AX) + MOVB CL, -1(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeBlockAsm12B + +emit_lit_memmove_emit_remainder_encodeBlockAsm12B_memmove_move_3: + MOVW (CX), SI + MOVB 2(CX), CL + MOVW SI, (AX) + MOVB CL, 2(AX) + JMP memmove_end_copy_emit_remainder_encodeBlockAsm12B + +emit_lit_memmove_emit_remainder_encodeBlockAsm12B_memmove_move_4through7: + MOVL (CX), SI + MOVL -4(CX)(BX*1), CX + MOVL SI, (AX) + MOVL CX, -4(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeBlockAsm12B + +emit_lit_memmove_emit_remainder_encodeBlockAsm12B_memmove_move_8through16: + MOVQ (CX), SI + MOVQ -8(CX)(BX*1), CX + MOVQ SI, (AX) + MOVQ CX, -8(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeBlockAsm12B + +emit_lit_memmove_emit_remainder_encodeBlockAsm12B_memmove_move_17through32: + MOVOU (CX), X0 + MOVOU -16(CX)(BX*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeBlockAsm12B + +emit_lit_memmove_emit_remainder_encodeBlockAsm12B_memmove_move_33through64: + MOVOU (CX), X0 + MOVOU 16(CX), X1 + MOVOU -32(CX)(BX*1), X2 + MOVOU -16(CX)(BX*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(BX*1) + MOVOU X3, -16(AX)(BX*1) + +memmove_end_copy_emit_remainder_encodeBlockAsm12B: + MOVQ DX, AX + JMP emit_literal_done_emit_remainder_encodeBlockAsm12B + +memmove_long_emit_remainder_encodeBlockAsm12B: + LEAQ (AX)(SI*1), DX + MOVL SI, BX + + // genMemMoveLong + MOVOU (CX), X0 + MOVOU 16(CX), X1 + MOVOU -32(CX)(BX*1), X2 + MOVOU -16(CX)(BX*1), X3 + MOVQ BX, DI + SHRQ $0x05, DI + MOVQ AX, SI + ANDL $0x0000001f, SI + MOVQ $0x00000040, R8 + SUBQ SI, R8 + DECQ DI + JA emit_lit_memmove_long_emit_remainder_encodeBlockAsm12Blarge_forward_sse_loop_32 + LEAQ -32(CX)(R8*1), SI + LEAQ -32(AX)(R8*1), R9 + +emit_lit_memmove_long_emit_remainder_encodeBlockAsm12Blarge_big_loop_back: + MOVOU (SI), X4 + MOVOU 16(SI), X5 + MOVOA X4, (R9) + MOVOA X5, 16(R9) + ADDQ $0x20, R9 + ADDQ $0x20, SI + ADDQ $0x20, R8 + DECQ DI + JNA emit_lit_memmove_long_emit_remainder_encodeBlockAsm12Blarge_big_loop_back + +emit_lit_memmove_long_emit_remainder_encodeBlockAsm12Blarge_forward_sse_loop_32: + MOVOU -32(CX)(R8*1), X4 + MOVOU -16(CX)(R8*1), X5 + MOVOA X4, -32(AX)(R8*1) + MOVOA X5, -16(AX)(R8*1) + ADDQ $0x20, R8 + CMPQ BX, R8 + JAE emit_lit_memmove_long_emit_remainder_encodeBlockAsm12Blarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(BX*1) + MOVOU X3, -16(AX)(BX*1) + MOVQ DX, AX + +emit_literal_done_emit_remainder_encodeBlockAsm12B: + MOVQ dst_base+0(FP), CX + SUBQ CX, AX + MOVQ AX, ret+48(FP) + RET + +// func encodeBlockAsm10B(dst []byte, src []byte) int +// Requires: BMI, SSE2 +TEXT ·encodeBlockAsm10B(SB), $4120-56 + MOVQ dst_base+0(FP), AX + MOVQ $0x00000020, CX + LEAQ 24(SP), DX + PXOR X0, X0 + +zero_loop_encodeBlockAsm10B: + MOVOU X0, (DX) + MOVOU X0, 16(DX) + MOVOU X0, 32(DX) + MOVOU X0, 48(DX) + MOVOU X0, 64(DX) + MOVOU X0, 80(DX) + MOVOU X0, 96(DX) + MOVOU X0, 112(DX) + ADDQ $0x80, DX + DECQ CX + JNZ zero_loop_encodeBlockAsm10B + MOVL $0x00000000, 12(SP) + MOVQ src_len+32(FP), CX + LEAQ -9(CX), DX + LEAQ -8(CX), BX + MOVL BX, 8(SP) + SHRQ $0x05, CX + SUBL CX, DX + LEAQ (AX)(DX*1), DX + MOVQ DX, (SP) + MOVL $0x00000001, CX + MOVL CX, 16(SP) + MOVQ src_base+24(FP), DX + +search_loop_encodeBlockAsm10B: + MOVL CX, BX + SUBL 12(SP), BX + SHRL $0x05, BX + LEAL 4(CX)(BX*1), BX + CMPL BX, 8(SP) + JAE emit_remainder_encodeBlockAsm10B + MOVQ (DX)(CX*1), SI + MOVL BX, 20(SP) + MOVQ $0x9e3779b1, R8 + MOVQ SI, R9 + MOVQ SI, R10 + SHRQ $0x08, R10 + SHLQ $0x20, R9 + IMULQ R8, R9 + SHRQ $0x36, R9 + SHLQ $0x20, R10 + IMULQ R8, R10 + SHRQ $0x36, R10 + MOVL 24(SP)(R9*4), BX + MOVL 24(SP)(R10*4), DI + MOVL CX, 24(SP)(R9*4) + LEAL 1(CX), R9 + MOVL R9, 24(SP)(R10*4) + MOVQ SI, R9 + SHRQ $0x10, R9 + SHLQ $0x20, R9 + IMULQ R8, R9 + SHRQ $0x36, R9 + MOVL CX, R8 + SUBL 16(SP), R8 + MOVL 1(DX)(R8*1), R10 + MOVQ SI, R8 + SHRQ $0x08, R8 + CMPL R8, R10 + JNE no_repeat_found_encodeBlockAsm10B + LEAL 1(CX), SI + MOVL 12(SP), DI + MOVL SI, BX + SUBL 16(SP), BX + JZ repeat_extend_back_end_encodeBlockAsm10B + +repeat_extend_back_loop_encodeBlockAsm10B: + CMPL SI, DI + JBE repeat_extend_back_end_encodeBlockAsm10B + MOVB -1(DX)(BX*1), R8 + MOVB -1(DX)(SI*1), R9 + CMPB R8, R9 + JNE repeat_extend_back_end_encodeBlockAsm10B + LEAL -1(SI), SI + DECL BX + JNZ repeat_extend_back_loop_encodeBlockAsm10B + +repeat_extend_back_end_encodeBlockAsm10B: + MOVL SI, BX + SUBL 12(SP), BX + LEAQ 3(AX)(BX*1), BX + CMPQ BX, (SP) + JB repeat_dst_size_check_encodeBlockAsm10B + MOVQ $0x00000000, ret+48(FP) + RET + +repeat_dst_size_check_encodeBlockAsm10B: + MOVL 12(SP), BX + CMPL BX, SI + JEQ emit_literal_done_repeat_emit_encodeBlockAsm10B + MOVL SI, R8 + MOVL SI, 12(SP) + LEAQ (DX)(BX*1), R9 + SUBL BX, R8 + LEAL -1(R8), BX + CMPL BX, $0x3c + JB one_byte_repeat_emit_encodeBlockAsm10B + CMPL BX, $0x00000100 + JB two_bytes_repeat_emit_encodeBlockAsm10B + JB three_bytes_repeat_emit_encodeBlockAsm10B + +three_bytes_repeat_emit_encodeBlockAsm10B: + MOVB $0xf4, (AX) + MOVW BX, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_repeat_emit_encodeBlockAsm10B + +two_bytes_repeat_emit_encodeBlockAsm10B: + MOVB $0xf0, (AX) + MOVB BL, 1(AX) + ADDQ $0x02, AX + CMPL BX, $0x40 + JB memmove_repeat_emit_encodeBlockAsm10B + JMP memmove_long_repeat_emit_encodeBlockAsm10B + +one_byte_repeat_emit_encodeBlockAsm10B: + SHLB $0x02, BL + MOVB BL, (AX) + ADDQ $0x01, AX + +memmove_repeat_emit_encodeBlockAsm10B: + LEAQ (AX)(R8*1), BX + + // genMemMoveShort + CMPQ R8, $0x08 + JBE emit_lit_memmove_repeat_emit_encodeBlockAsm10B_memmove_move_8 + CMPQ R8, $0x10 + JBE emit_lit_memmove_repeat_emit_encodeBlockAsm10B_memmove_move_8through16 + CMPQ R8, $0x20 + JBE emit_lit_memmove_repeat_emit_encodeBlockAsm10B_memmove_move_17through32 + JMP emit_lit_memmove_repeat_emit_encodeBlockAsm10B_memmove_move_33through64 + +emit_lit_memmove_repeat_emit_encodeBlockAsm10B_memmove_move_8: + MOVQ (R9), R10 + MOVQ R10, (AX) + JMP memmove_end_copy_repeat_emit_encodeBlockAsm10B + +emit_lit_memmove_repeat_emit_encodeBlockAsm10B_memmove_move_8through16: + MOVQ (R9), R10 + MOVQ -8(R9)(R8*1), R9 + MOVQ R10, (AX) + MOVQ R9, -8(AX)(R8*1) + JMP memmove_end_copy_repeat_emit_encodeBlockAsm10B + +emit_lit_memmove_repeat_emit_encodeBlockAsm10B_memmove_move_17through32: + MOVOU (R9), X0 + MOVOU -16(R9)(R8*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(R8*1) + JMP memmove_end_copy_repeat_emit_encodeBlockAsm10B + +emit_lit_memmove_repeat_emit_encodeBlockAsm10B_memmove_move_33through64: + MOVOU (R9), X0 + MOVOU 16(R9), X1 + MOVOU -32(R9)(R8*1), X2 + MOVOU -16(R9)(R8*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + +memmove_end_copy_repeat_emit_encodeBlockAsm10B: + MOVQ BX, AX + JMP emit_literal_done_repeat_emit_encodeBlockAsm10B + +memmove_long_repeat_emit_encodeBlockAsm10B: + LEAQ (AX)(R8*1), BX + + // genMemMoveLong + MOVOU (R9), X0 + MOVOU 16(R9), X1 + MOVOU -32(R9)(R8*1), X2 + MOVOU -16(R9)(R8*1), X3 + MOVQ R8, R11 + SHRQ $0x05, R11 + MOVQ AX, R10 + ANDL $0x0000001f, R10 + MOVQ $0x00000040, R12 + SUBQ R10, R12 + DECQ R11 + JA emit_lit_memmove_long_repeat_emit_encodeBlockAsm10Blarge_forward_sse_loop_32 + LEAQ -32(R9)(R12*1), R10 + LEAQ -32(AX)(R12*1), R13 + +emit_lit_memmove_long_repeat_emit_encodeBlockAsm10Blarge_big_loop_back: + MOVOU (R10), X4 + MOVOU 16(R10), X5 + MOVOA X4, (R13) + MOVOA X5, 16(R13) + ADDQ $0x20, R13 + ADDQ $0x20, R10 + ADDQ $0x20, R12 + DECQ R11 + JNA emit_lit_memmove_long_repeat_emit_encodeBlockAsm10Blarge_big_loop_back + +emit_lit_memmove_long_repeat_emit_encodeBlockAsm10Blarge_forward_sse_loop_32: + MOVOU -32(R9)(R12*1), X4 + MOVOU -16(R9)(R12*1), X5 + MOVOA X4, -32(AX)(R12*1) + MOVOA X5, -16(AX)(R12*1) + ADDQ $0x20, R12 + CMPQ R8, R12 + JAE emit_lit_memmove_long_repeat_emit_encodeBlockAsm10Blarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + MOVQ BX, AX + +emit_literal_done_repeat_emit_encodeBlockAsm10B: + ADDL $0x05, CX + MOVL CX, BX + SUBL 16(SP), BX + MOVQ src_len+32(FP), R8 + SUBL CX, R8 + LEAQ (DX)(CX*1), R9 + LEAQ (DX)(BX*1), BX + + // matchLen + XORL R11, R11 + +matchlen_loopback_16_repeat_extend_encodeBlockAsm10B: + CMPL R8, $0x10 + JB matchlen_match8_repeat_extend_encodeBlockAsm10B + MOVQ (R9)(R11*1), R10 + MOVQ 8(R9)(R11*1), R12 + XORQ (BX)(R11*1), R10 + JNZ matchlen_bsf_8_repeat_extend_encodeBlockAsm10B + XORQ 8(BX)(R11*1), R12 + JNZ matchlen_bsf_16repeat_extend_encodeBlockAsm10B + LEAL -16(R8), R8 + LEAL 16(R11), R11 + JMP matchlen_loopback_16_repeat_extend_encodeBlockAsm10B + +matchlen_bsf_16repeat_extend_encodeBlockAsm10B: +#ifdef GOAMD64_v3 + TZCNTQ R12, R12 + +#else + BSFQ R12, R12 + +#endif + SARQ $0x03, R12 + LEAL 8(R11)(R12*1), R11 + JMP repeat_extend_forward_end_encodeBlockAsm10B + +matchlen_match8_repeat_extend_encodeBlockAsm10B: + CMPL R8, $0x08 + JB matchlen_match4_repeat_extend_encodeBlockAsm10B + MOVQ (R9)(R11*1), R10 + XORQ (BX)(R11*1), R10 + JNZ matchlen_bsf_8_repeat_extend_encodeBlockAsm10B + LEAL -8(R8), R8 + LEAL 8(R11), R11 + JMP matchlen_match4_repeat_extend_encodeBlockAsm10B + +matchlen_bsf_8_repeat_extend_encodeBlockAsm10B: +#ifdef GOAMD64_v3 + TZCNTQ R10, R10 + +#else + BSFQ R10, R10 + +#endif + SARQ $0x03, R10 + LEAL (R11)(R10*1), R11 + JMP repeat_extend_forward_end_encodeBlockAsm10B + +matchlen_match4_repeat_extend_encodeBlockAsm10B: + CMPL R8, $0x04 + JB matchlen_match2_repeat_extend_encodeBlockAsm10B + MOVL (R9)(R11*1), R10 + CMPL (BX)(R11*1), R10 + JNE matchlen_match2_repeat_extend_encodeBlockAsm10B + LEAL -4(R8), R8 + LEAL 4(R11), R11 + +matchlen_match2_repeat_extend_encodeBlockAsm10B: + CMPL R8, $0x01 + JE matchlen_match1_repeat_extend_encodeBlockAsm10B + JB repeat_extend_forward_end_encodeBlockAsm10B + MOVW (R9)(R11*1), R10 + CMPW (BX)(R11*1), R10 + JNE matchlen_match1_repeat_extend_encodeBlockAsm10B + LEAL 2(R11), R11 + SUBL $0x02, R8 + JZ repeat_extend_forward_end_encodeBlockAsm10B + +matchlen_match1_repeat_extend_encodeBlockAsm10B: + MOVB (R9)(R11*1), R10 + CMPB (BX)(R11*1), R10 + JNE repeat_extend_forward_end_encodeBlockAsm10B + LEAL 1(R11), R11 + +repeat_extend_forward_end_encodeBlockAsm10B: + ADDL R11, CX + MOVL CX, BX + SUBL SI, BX + MOVL 16(SP), SI + TESTL DI, DI + JZ repeat_as_copy_encodeBlockAsm10B + + // emitRepeat + MOVL BX, DI + LEAL -4(BX), BX + CMPL DI, $0x08 + JBE repeat_two_match_repeat_encodeBlockAsm10B + CMPL DI, $0x0c + JAE cant_repeat_two_offset_match_repeat_encodeBlockAsm10B + CMPL SI, $0x00000800 + JB repeat_two_offset_match_repeat_encodeBlockAsm10B + +cant_repeat_two_offset_match_repeat_encodeBlockAsm10B: + CMPL BX, $0x00000104 + JB repeat_three_match_repeat_encodeBlockAsm10B + LEAL -256(BX), BX + MOVW $0x0019, (AX) + MOVW BX, 2(AX) + ADDQ $0x04, AX + JMP repeat_end_emit_encodeBlockAsm10B + +repeat_three_match_repeat_encodeBlockAsm10B: + LEAL -4(BX), BX + MOVW $0x0015, (AX) + MOVB BL, 2(AX) + ADDQ $0x03, AX + JMP repeat_end_emit_encodeBlockAsm10B + +repeat_two_match_repeat_encodeBlockAsm10B: + SHLL $0x02, BX + ORL $0x01, BX + MOVW BX, (AX) + ADDQ $0x02, AX + JMP repeat_end_emit_encodeBlockAsm10B + +repeat_two_offset_match_repeat_encodeBlockAsm10B: + XORQ DI, DI + LEAL 1(DI)(BX*4), BX + MOVB SI, 1(AX) + SARL $0x08, SI + SHLL $0x05, SI + ORL SI, BX + MOVB BL, (AX) + ADDQ $0x02, AX + JMP repeat_end_emit_encodeBlockAsm10B + +repeat_as_copy_encodeBlockAsm10B: + // emitCopy + CMPL BX, $0x40 + JBE two_byte_offset_short_repeat_as_copy_encodeBlockAsm10B + CMPL SI, $0x00000800 + JAE long_offset_short_repeat_as_copy_encodeBlockAsm10B + MOVL $0x00000001, DI + LEAL 16(DI), DI + MOVB SI, 1(AX) + SHRL $0x08, SI + SHLL $0x05, SI + ORL SI, DI + MOVB DI, (AX) + ADDQ $0x02, AX + SUBL $0x08, BX + + // emitRepeat + LEAL -4(BX), BX + JMP cant_repeat_two_offset_repeat_as_copy_encodeBlockAsm10B_emit_copy_short_2b + MOVL BX, DI + LEAL -4(BX), BX + CMPL DI, $0x08 + JBE repeat_two_repeat_as_copy_encodeBlockAsm10B_emit_copy_short_2b + CMPL DI, $0x0c + JAE cant_repeat_two_offset_repeat_as_copy_encodeBlockAsm10B_emit_copy_short_2b + CMPL SI, $0x00000800 + JB repeat_two_offset_repeat_as_copy_encodeBlockAsm10B_emit_copy_short_2b + +cant_repeat_two_offset_repeat_as_copy_encodeBlockAsm10B_emit_copy_short_2b: + CMPL BX, $0x00000104 + JB repeat_three_repeat_as_copy_encodeBlockAsm10B_emit_copy_short_2b + LEAL -256(BX), BX + MOVW $0x0019, (AX) + MOVW BX, 2(AX) + ADDQ $0x04, AX + JMP repeat_end_emit_encodeBlockAsm10B + +repeat_three_repeat_as_copy_encodeBlockAsm10B_emit_copy_short_2b: + LEAL -4(BX), BX + MOVW $0x0015, (AX) + MOVB BL, 2(AX) + ADDQ $0x03, AX + JMP repeat_end_emit_encodeBlockAsm10B + +repeat_two_repeat_as_copy_encodeBlockAsm10B_emit_copy_short_2b: + SHLL $0x02, BX + ORL $0x01, BX + MOVW BX, (AX) + ADDQ $0x02, AX + JMP repeat_end_emit_encodeBlockAsm10B + +repeat_two_offset_repeat_as_copy_encodeBlockAsm10B_emit_copy_short_2b: + XORQ DI, DI + LEAL 1(DI)(BX*4), BX + MOVB SI, 1(AX) + SARL $0x08, SI + SHLL $0x05, SI + ORL SI, BX + MOVB BL, (AX) + ADDQ $0x02, AX + JMP repeat_end_emit_encodeBlockAsm10B + +long_offset_short_repeat_as_copy_encodeBlockAsm10B: + MOVB $0xee, (AX) + MOVW SI, 1(AX) + LEAL -60(BX), BX + ADDQ $0x03, AX + + // emitRepeat + MOVL BX, DI + LEAL -4(BX), BX + CMPL DI, $0x08 + JBE repeat_two_repeat_as_copy_encodeBlockAsm10B_emit_copy_short + CMPL DI, $0x0c + JAE cant_repeat_two_offset_repeat_as_copy_encodeBlockAsm10B_emit_copy_short + CMPL SI, $0x00000800 + JB repeat_two_offset_repeat_as_copy_encodeBlockAsm10B_emit_copy_short + +cant_repeat_two_offset_repeat_as_copy_encodeBlockAsm10B_emit_copy_short: + CMPL BX, $0x00000104 + JB repeat_three_repeat_as_copy_encodeBlockAsm10B_emit_copy_short + LEAL -256(BX), BX + MOVW $0x0019, (AX) + MOVW BX, 2(AX) + ADDQ $0x04, AX + JMP repeat_end_emit_encodeBlockAsm10B + +repeat_three_repeat_as_copy_encodeBlockAsm10B_emit_copy_short: + LEAL -4(BX), BX + MOVW $0x0015, (AX) + MOVB BL, 2(AX) + ADDQ $0x03, AX + JMP repeat_end_emit_encodeBlockAsm10B + +repeat_two_repeat_as_copy_encodeBlockAsm10B_emit_copy_short: + SHLL $0x02, BX + ORL $0x01, BX + MOVW BX, (AX) + ADDQ $0x02, AX + JMP repeat_end_emit_encodeBlockAsm10B + +repeat_two_offset_repeat_as_copy_encodeBlockAsm10B_emit_copy_short: + XORQ DI, DI + LEAL 1(DI)(BX*4), BX + MOVB SI, 1(AX) + SARL $0x08, SI + SHLL $0x05, SI + ORL SI, BX + MOVB BL, (AX) + ADDQ $0x02, AX + JMP repeat_end_emit_encodeBlockAsm10B + +two_byte_offset_short_repeat_as_copy_encodeBlockAsm10B: + MOVL BX, DI + SHLL $0x02, DI + CMPL BX, $0x0c + JAE emit_copy_three_repeat_as_copy_encodeBlockAsm10B + CMPL SI, $0x00000800 + JAE emit_copy_three_repeat_as_copy_encodeBlockAsm10B + LEAL -15(DI), DI + MOVB SI, 1(AX) + SHRL $0x08, SI + SHLL $0x05, SI + ORL SI, DI + MOVB DI, (AX) + ADDQ $0x02, AX + JMP repeat_end_emit_encodeBlockAsm10B + +emit_copy_three_repeat_as_copy_encodeBlockAsm10B: + LEAL -2(DI), DI + MOVB DI, (AX) + MOVW SI, 1(AX) + ADDQ $0x03, AX + +repeat_end_emit_encodeBlockAsm10B: + MOVL CX, 12(SP) + JMP search_loop_encodeBlockAsm10B + +no_repeat_found_encodeBlockAsm10B: + CMPL (DX)(BX*1), SI + JEQ candidate_match_encodeBlockAsm10B + SHRQ $0x08, SI + MOVL 24(SP)(R9*4), BX + LEAL 2(CX), R8 + CMPL (DX)(DI*1), SI + JEQ candidate2_match_encodeBlockAsm10B + MOVL R8, 24(SP)(R9*4) + SHRQ $0x08, SI + CMPL (DX)(BX*1), SI + JEQ candidate3_match_encodeBlockAsm10B + MOVL 20(SP), CX + JMP search_loop_encodeBlockAsm10B + +candidate3_match_encodeBlockAsm10B: + ADDL $0x02, CX + JMP candidate_match_encodeBlockAsm10B + +candidate2_match_encodeBlockAsm10B: + MOVL R8, 24(SP)(R9*4) + INCL CX + MOVL DI, BX + +candidate_match_encodeBlockAsm10B: + MOVL 12(SP), SI + TESTL BX, BX + JZ match_extend_back_end_encodeBlockAsm10B + +match_extend_back_loop_encodeBlockAsm10B: + CMPL CX, SI + JBE match_extend_back_end_encodeBlockAsm10B + MOVB -1(DX)(BX*1), DI + MOVB -1(DX)(CX*1), R8 + CMPB DI, R8 + JNE match_extend_back_end_encodeBlockAsm10B + LEAL -1(CX), CX + DECL BX + JZ match_extend_back_end_encodeBlockAsm10B + JMP match_extend_back_loop_encodeBlockAsm10B + +match_extend_back_end_encodeBlockAsm10B: + MOVL CX, SI + SUBL 12(SP), SI + LEAQ 3(AX)(SI*1), SI + CMPQ SI, (SP) + JB match_dst_size_check_encodeBlockAsm10B + MOVQ $0x00000000, ret+48(FP) + RET + +match_dst_size_check_encodeBlockAsm10B: + MOVL CX, SI + MOVL 12(SP), DI + CMPL DI, SI + JEQ emit_literal_done_match_emit_encodeBlockAsm10B + MOVL SI, R8 + MOVL SI, 12(SP) + LEAQ (DX)(DI*1), SI + SUBL DI, R8 + LEAL -1(R8), DI + CMPL DI, $0x3c + JB one_byte_match_emit_encodeBlockAsm10B + CMPL DI, $0x00000100 + JB two_bytes_match_emit_encodeBlockAsm10B + JB three_bytes_match_emit_encodeBlockAsm10B + +three_bytes_match_emit_encodeBlockAsm10B: + MOVB $0xf4, (AX) + MOVW DI, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_match_emit_encodeBlockAsm10B + +two_bytes_match_emit_encodeBlockAsm10B: + MOVB $0xf0, (AX) + MOVB DI, 1(AX) + ADDQ $0x02, AX + CMPL DI, $0x40 + JB memmove_match_emit_encodeBlockAsm10B + JMP memmove_long_match_emit_encodeBlockAsm10B + +one_byte_match_emit_encodeBlockAsm10B: + SHLB $0x02, DI + MOVB DI, (AX) + ADDQ $0x01, AX + +memmove_match_emit_encodeBlockAsm10B: + LEAQ (AX)(R8*1), DI + + // genMemMoveShort + CMPQ R8, $0x08 + JBE emit_lit_memmove_match_emit_encodeBlockAsm10B_memmove_move_8 + CMPQ R8, $0x10 + JBE emit_lit_memmove_match_emit_encodeBlockAsm10B_memmove_move_8through16 + CMPQ R8, $0x20 + JBE emit_lit_memmove_match_emit_encodeBlockAsm10B_memmove_move_17through32 + JMP emit_lit_memmove_match_emit_encodeBlockAsm10B_memmove_move_33through64 + +emit_lit_memmove_match_emit_encodeBlockAsm10B_memmove_move_8: + MOVQ (SI), R9 + MOVQ R9, (AX) + JMP memmove_end_copy_match_emit_encodeBlockAsm10B + +emit_lit_memmove_match_emit_encodeBlockAsm10B_memmove_move_8through16: + MOVQ (SI), R9 + MOVQ -8(SI)(R8*1), SI + MOVQ R9, (AX) + MOVQ SI, -8(AX)(R8*1) + JMP memmove_end_copy_match_emit_encodeBlockAsm10B + +emit_lit_memmove_match_emit_encodeBlockAsm10B_memmove_move_17through32: + MOVOU (SI), X0 + MOVOU -16(SI)(R8*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(R8*1) + JMP memmove_end_copy_match_emit_encodeBlockAsm10B + +emit_lit_memmove_match_emit_encodeBlockAsm10B_memmove_move_33through64: + MOVOU (SI), X0 + MOVOU 16(SI), X1 + MOVOU -32(SI)(R8*1), X2 + MOVOU -16(SI)(R8*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + +memmove_end_copy_match_emit_encodeBlockAsm10B: + MOVQ DI, AX + JMP emit_literal_done_match_emit_encodeBlockAsm10B + +memmove_long_match_emit_encodeBlockAsm10B: + LEAQ (AX)(R8*1), DI + + // genMemMoveLong + MOVOU (SI), X0 + MOVOU 16(SI), X1 + MOVOU -32(SI)(R8*1), X2 + MOVOU -16(SI)(R8*1), X3 + MOVQ R8, R10 + SHRQ $0x05, R10 + MOVQ AX, R9 + ANDL $0x0000001f, R9 + MOVQ $0x00000040, R11 + SUBQ R9, R11 + DECQ R10 + JA emit_lit_memmove_long_match_emit_encodeBlockAsm10Blarge_forward_sse_loop_32 + LEAQ -32(SI)(R11*1), R9 + LEAQ -32(AX)(R11*1), R12 + +emit_lit_memmove_long_match_emit_encodeBlockAsm10Blarge_big_loop_back: + MOVOU (R9), X4 + MOVOU 16(R9), X5 + MOVOA X4, (R12) + MOVOA X5, 16(R12) + ADDQ $0x20, R12 + ADDQ $0x20, R9 + ADDQ $0x20, R11 + DECQ R10 + JNA emit_lit_memmove_long_match_emit_encodeBlockAsm10Blarge_big_loop_back + +emit_lit_memmove_long_match_emit_encodeBlockAsm10Blarge_forward_sse_loop_32: + MOVOU -32(SI)(R11*1), X4 + MOVOU -16(SI)(R11*1), X5 + MOVOA X4, -32(AX)(R11*1) + MOVOA X5, -16(AX)(R11*1) + ADDQ $0x20, R11 + CMPQ R8, R11 + JAE emit_lit_memmove_long_match_emit_encodeBlockAsm10Blarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + MOVQ DI, AX + +emit_literal_done_match_emit_encodeBlockAsm10B: +match_nolit_loop_encodeBlockAsm10B: + MOVL CX, SI + SUBL BX, SI + MOVL SI, 16(SP) + ADDL $0x04, CX + ADDL $0x04, BX + MOVQ src_len+32(FP), SI + SUBL CX, SI + LEAQ (DX)(CX*1), DI + LEAQ (DX)(BX*1), BX + + // matchLen + XORL R9, R9 + +matchlen_loopback_16_match_nolit_encodeBlockAsm10B: + CMPL SI, $0x10 + JB matchlen_match8_match_nolit_encodeBlockAsm10B + MOVQ (DI)(R9*1), R8 + MOVQ 8(DI)(R9*1), R10 + XORQ (BX)(R9*1), R8 + JNZ matchlen_bsf_8_match_nolit_encodeBlockAsm10B + XORQ 8(BX)(R9*1), R10 + JNZ matchlen_bsf_16match_nolit_encodeBlockAsm10B + LEAL -16(SI), SI + LEAL 16(R9), R9 + JMP matchlen_loopback_16_match_nolit_encodeBlockAsm10B + +matchlen_bsf_16match_nolit_encodeBlockAsm10B: +#ifdef GOAMD64_v3 + TZCNTQ R10, R10 + +#else + BSFQ R10, R10 + +#endif + SARQ $0x03, R10 + LEAL 8(R9)(R10*1), R9 + JMP match_nolit_end_encodeBlockAsm10B + +matchlen_match8_match_nolit_encodeBlockAsm10B: + CMPL SI, $0x08 + JB matchlen_match4_match_nolit_encodeBlockAsm10B + MOVQ (DI)(R9*1), R8 + XORQ (BX)(R9*1), R8 + JNZ matchlen_bsf_8_match_nolit_encodeBlockAsm10B + LEAL -8(SI), SI + LEAL 8(R9), R9 + JMP matchlen_match4_match_nolit_encodeBlockAsm10B + +matchlen_bsf_8_match_nolit_encodeBlockAsm10B: +#ifdef GOAMD64_v3 + TZCNTQ R8, R8 + +#else + BSFQ R8, R8 + +#endif + SARQ $0x03, R8 + LEAL (R9)(R8*1), R9 + JMP match_nolit_end_encodeBlockAsm10B + +matchlen_match4_match_nolit_encodeBlockAsm10B: + CMPL SI, $0x04 + JB matchlen_match2_match_nolit_encodeBlockAsm10B + MOVL (DI)(R9*1), R8 + CMPL (BX)(R9*1), R8 + JNE matchlen_match2_match_nolit_encodeBlockAsm10B + LEAL -4(SI), SI + LEAL 4(R9), R9 + +matchlen_match2_match_nolit_encodeBlockAsm10B: + CMPL SI, $0x01 + JE matchlen_match1_match_nolit_encodeBlockAsm10B + JB match_nolit_end_encodeBlockAsm10B + MOVW (DI)(R9*1), R8 + CMPW (BX)(R9*1), R8 + JNE matchlen_match1_match_nolit_encodeBlockAsm10B + LEAL 2(R9), R9 + SUBL $0x02, SI + JZ match_nolit_end_encodeBlockAsm10B + +matchlen_match1_match_nolit_encodeBlockAsm10B: + MOVB (DI)(R9*1), R8 + CMPB (BX)(R9*1), R8 + JNE match_nolit_end_encodeBlockAsm10B + LEAL 1(R9), R9 + +match_nolit_end_encodeBlockAsm10B: + ADDL R9, CX + MOVL 16(SP), BX + ADDL $0x04, R9 + MOVL CX, 12(SP) + + // emitCopy + CMPL R9, $0x40 + JBE two_byte_offset_short_match_nolit_encodeBlockAsm10B + CMPL BX, $0x00000800 + JAE long_offset_short_match_nolit_encodeBlockAsm10B + MOVL $0x00000001, SI + LEAL 16(SI), SI + MOVB BL, 1(AX) + SHRL $0x08, BX + SHLL $0x05, BX + ORL BX, SI + MOVB SI, (AX) + ADDQ $0x02, AX + SUBL $0x08, R9 + + // emitRepeat + LEAL -4(R9), R9 + JMP cant_repeat_two_offset_match_nolit_encodeBlockAsm10B_emit_copy_short_2b + MOVL R9, SI + LEAL -4(R9), R9 + CMPL SI, $0x08 + JBE repeat_two_match_nolit_encodeBlockAsm10B_emit_copy_short_2b + CMPL SI, $0x0c + JAE cant_repeat_two_offset_match_nolit_encodeBlockAsm10B_emit_copy_short_2b + CMPL BX, $0x00000800 + JB repeat_two_offset_match_nolit_encodeBlockAsm10B_emit_copy_short_2b + +cant_repeat_two_offset_match_nolit_encodeBlockAsm10B_emit_copy_short_2b: + CMPL R9, $0x00000104 + JB repeat_three_match_nolit_encodeBlockAsm10B_emit_copy_short_2b + LEAL -256(R9), R9 + MOVW $0x0019, (AX) + MOVW R9, 2(AX) + ADDQ $0x04, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm10B + +repeat_three_match_nolit_encodeBlockAsm10B_emit_copy_short_2b: + LEAL -4(R9), R9 + MOVW $0x0015, (AX) + MOVB R9, 2(AX) + ADDQ $0x03, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm10B + +repeat_two_match_nolit_encodeBlockAsm10B_emit_copy_short_2b: + SHLL $0x02, R9 + ORL $0x01, R9 + MOVW R9, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm10B + +repeat_two_offset_match_nolit_encodeBlockAsm10B_emit_copy_short_2b: + XORQ SI, SI + LEAL 1(SI)(R9*4), R9 + MOVB BL, 1(AX) + SARL $0x08, BX + SHLL $0x05, BX + ORL BX, R9 + MOVB R9, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm10B + +long_offset_short_match_nolit_encodeBlockAsm10B: + MOVB $0xee, (AX) + MOVW BX, 1(AX) + LEAL -60(R9), R9 + ADDQ $0x03, AX + + // emitRepeat + MOVL R9, SI + LEAL -4(R9), R9 + CMPL SI, $0x08 + JBE repeat_two_match_nolit_encodeBlockAsm10B_emit_copy_short + CMPL SI, $0x0c + JAE cant_repeat_two_offset_match_nolit_encodeBlockAsm10B_emit_copy_short + CMPL BX, $0x00000800 + JB repeat_two_offset_match_nolit_encodeBlockAsm10B_emit_copy_short + +cant_repeat_two_offset_match_nolit_encodeBlockAsm10B_emit_copy_short: + CMPL R9, $0x00000104 + JB repeat_three_match_nolit_encodeBlockAsm10B_emit_copy_short + LEAL -256(R9), R9 + MOVW $0x0019, (AX) + MOVW R9, 2(AX) + ADDQ $0x04, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm10B + +repeat_three_match_nolit_encodeBlockAsm10B_emit_copy_short: + LEAL -4(R9), R9 + MOVW $0x0015, (AX) + MOVB R9, 2(AX) + ADDQ $0x03, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm10B + +repeat_two_match_nolit_encodeBlockAsm10B_emit_copy_short: + SHLL $0x02, R9 + ORL $0x01, R9 + MOVW R9, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm10B + +repeat_two_offset_match_nolit_encodeBlockAsm10B_emit_copy_short: + XORQ SI, SI + LEAL 1(SI)(R9*4), R9 + MOVB BL, 1(AX) + SARL $0x08, BX + SHLL $0x05, BX + ORL BX, R9 + MOVB R9, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm10B + +two_byte_offset_short_match_nolit_encodeBlockAsm10B: + MOVL R9, SI + SHLL $0x02, SI + CMPL R9, $0x0c + JAE emit_copy_three_match_nolit_encodeBlockAsm10B + CMPL BX, $0x00000800 + JAE emit_copy_three_match_nolit_encodeBlockAsm10B + LEAL -15(SI), SI + MOVB BL, 1(AX) + SHRL $0x08, BX + SHLL $0x05, BX + ORL BX, SI + MOVB SI, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm10B + +emit_copy_three_match_nolit_encodeBlockAsm10B: + LEAL -2(SI), SI + MOVB SI, (AX) + MOVW BX, 1(AX) + ADDQ $0x03, AX + +match_nolit_emitcopy_end_encodeBlockAsm10B: + CMPL CX, 8(SP) + JAE emit_remainder_encodeBlockAsm10B + MOVQ -2(DX)(CX*1), SI + CMPQ AX, (SP) + JB match_nolit_dst_ok_encodeBlockAsm10B + MOVQ $0x00000000, ret+48(FP) + RET + +match_nolit_dst_ok_encodeBlockAsm10B: + MOVQ $0x9e3779b1, R8 + MOVQ SI, DI + SHRQ $0x10, SI + MOVQ SI, BX + SHLQ $0x20, DI + IMULQ R8, DI + SHRQ $0x36, DI + SHLQ $0x20, BX + IMULQ R8, BX + SHRQ $0x36, BX + LEAL -2(CX), R8 + LEAQ 24(SP)(BX*4), R9 + MOVL (R9), BX + MOVL R8, 24(SP)(DI*4) + MOVL CX, (R9) + CMPL (DX)(BX*1), SI + JEQ match_nolit_loop_encodeBlockAsm10B + INCL CX + JMP search_loop_encodeBlockAsm10B + +emit_remainder_encodeBlockAsm10B: + MOVQ src_len+32(FP), CX + SUBL 12(SP), CX + LEAQ 3(AX)(CX*1), CX + CMPQ CX, (SP) + JB emit_remainder_ok_encodeBlockAsm10B + MOVQ $0x00000000, ret+48(FP) + RET + +emit_remainder_ok_encodeBlockAsm10B: + MOVQ src_len+32(FP), CX + MOVL 12(SP), BX + CMPL BX, CX + JEQ emit_literal_done_emit_remainder_encodeBlockAsm10B + MOVL CX, SI + MOVL CX, 12(SP) + LEAQ (DX)(BX*1), CX + SUBL BX, SI + LEAL -1(SI), DX + CMPL DX, $0x3c + JB one_byte_emit_remainder_encodeBlockAsm10B + CMPL DX, $0x00000100 + JB two_bytes_emit_remainder_encodeBlockAsm10B + JB three_bytes_emit_remainder_encodeBlockAsm10B + +three_bytes_emit_remainder_encodeBlockAsm10B: + MOVB $0xf4, (AX) + MOVW DX, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_emit_remainder_encodeBlockAsm10B + +two_bytes_emit_remainder_encodeBlockAsm10B: + MOVB $0xf0, (AX) + MOVB DL, 1(AX) + ADDQ $0x02, AX + CMPL DX, $0x40 + JB memmove_emit_remainder_encodeBlockAsm10B + JMP memmove_long_emit_remainder_encodeBlockAsm10B + +one_byte_emit_remainder_encodeBlockAsm10B: + SHLB $0x02, DL + MOVB DL, (AX) + ADDQ $0x01, AX + +memmove_emit_remainder_encodeBlockAsm10B: + LEAQ (AX)(SI*1), DX + MOVL SI, BX + + // genMemMoveShort + CMPQ BX, $0x03 + JB emit_lit_memmove_emit_remainder_encodeBlockAsm10B_memmove_move_1or2 + JE emit_lit_memmove_emit_remainder_encodeBlockAsm10B_memmove_move_3 + CMPQ BX, $0x08 + JB emit_lit_memmove_emit_remainder_encodeBlockAsm10B_memmove_move_4through7 + CMPQ BX, $0x10 + JBE emit_lit_memmove_emit_remainder_encodeBlockAsm10B_memmove_move_8through16 + CMPQ BX, $0x20 + JBE emit_lit_memmove_emit_remainder_encodeBlockAsm10B_memmove_move_17through32 + JMP emit_lit_memmove_emit_remainder_encodeBlockAsm10B_memmove_move_33through64 + +emit_lit_memmove_emit_remainder_encodeBlockAsm10B_memmove_move_1or2: + MOVB (CX), SI + MOVB -1(CX)(BX*1), CL + MOVB SI, (AX) + MOVB CL, -1(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeBlockAsm10B + +emit_lit_memmove_emit_remainder_encodeBlockAsm10B_memmove_move_3: + MOVW (CX), SI + MOVB 2(CX), CL + MOVW SI, (AX) + MOVB CL, 2(AX) + JMP memmove_end_copy_emit_remainder_encodeBlockAsm10B + +emit_lit_memmove_emit_remainder_encodeBlockAsm10B_memmove_move_4through7: + MOVL (CX), SI + MOVL -4(CX)(BX*1), CX + MOVL SI, (AX) + MOVL CX, -4(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeBlockAsm10B + +emit_lit_memmove_emit_remainder_encodeBlockAsm10B_memmove_move_8through16: + MOVQ (CX), SI + MOVQ -8(CX)(BX*1), CX + MOVQ SI, (AX) + MOVQ CX, -8(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeBlockAsm10B + +emit_lit_memmove_emit_remainder_encodeBlockAsm10B_memmove_move_17through32: + MOVOU (CX), X0 + MOVOU -16(CX)(BX*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeBlockAsm10B + +emit_lit_memmove_emit_remainder_encodeBlockAsm10B_memmove_move_33through64: + MOVOU (CX), X0 + MOVOU 16(CX), X1 + MOVOU -32(CX)(BX*1), X2 + MOVOU -16(CX)(BX*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(BX*1) + MOVOU X3, -16(AX)(BX*1) + +memmove_end_copy_emit_remainder_encodeBlockAsm10B: + MOVQ DX, AX + JMP emit_literal_done_emit_remainder_encodeBlockAsm10B + +memmove_long_emit_remainder_encodeBlockAsm10B: + LEAQ (AX)(SI*1), DX + MOVL SI, BX + + // genMemMoveLong + MOVOU (CX), X0 + MOVOU 16(CX), X1 + MOVOU -32(CX)(BX*1), X2 + MOVOU -16(CX)(BX*1), X3 + MOVQ BX, DI + SHRQ $0x05, DI + MOVQ AX, SI + ANDL $0x0000001f, SI + MOVQ $0x00000040, R8 + SUBQ SI, R8 + DECQ DI + JA emit_lit_memmove_long_emit_remainder_encodeBlockAsm10Blarge_forward_sse_loop_32 + LEAQ -32(CX)(R8*1), SI + LEAQ -32(AX)(R8*1), R9 + +emit_lit_memmove_long_emit_remainder_encodeBlockAsm10Blarge_big_loop_back: + MOVOU (SI), X4 + MOVOU 16(SI), X5 + MOVOA X4, (R9) + MOVOA X5, 16(R9) + ADDQ $0x20, R9 + ADDQ $0x20, SI + ADDQ $0x20, R8 + DECQ DI + JNA emit_lit_memmove_long_emit_remainder_encodeBlockAsm10Blarge_big_loop_back + +emit_lit_memmove_long_emit_remainder_encodeBlockAsm10Blarge_forward_sse_loop_32: + MOVOU -32(CX)(R8*1), X4 + MOVOU -16(CX)(R8*1), X5 + MOVOA X4, -32(AX)(R8*1) + MOVOA X5, -16(AX)(R8*1) + ADDQ $0x20, R8 + CMPQ BX, R8 + JAE emit_lit_memmove_long_emit_remainder_encodeBlockAsm10Blarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(BX*1) + MOVOU X3, -16(AX)(BX*1) + MOVQ DX, AX + +emit_literal_done_emit_remainder_encodeBlockAsm10B: + MOVQ dst_base+0(FP), CX + SUBQ CX, AX + MOVQ AX, ret+48(FP) + RET + +// func encodeBlockAsm8B(dst []byte, src []byte) int +// Requires: BMI, SSE2 +TEXT ·encodeBlockAsm8B(SB), $1048-56 + MOVQ dst_base+0(FP), AX + MOVQ $0x00000008, CX + LEAQ 24(SP), DX + PXOR X0, X0 + +zero_loop_encodeBlockAsm8B: + MOVOU X0, (DX) + MOVOU X0, 16(DX) + MOVOU X0, 32(DX) + MOVOU X0, 48(DX) + MOVOU X0, 64(DX) + MOVOU X0, 80(DX) + MOVOU X0, 96(DX) + MOVOU X0, 112(DX) + ADDQ $0x80, DX + DECQ CX + JNZ zero_loop_encodeBlockAsm8B + MOVL $0x00000000, 12(SP) + MOVQ src_len+32(FP), CX + LEAQ -9(CX), DX + LEAQ -8(CX), BX + MOVL BX, 8(SP) + SHRQ $0x05, CX + SUBL CX, DX + LEAQ (AX)(DX*1), DX + MOVQ DX, (SP) + MOVL $0x00000001, CX + MOVL CX, 16(SP) + MOVQ src_base+24(FP), DX + +search_loop_encodeBlockAsm8B: + MOVL CX, BX + SUBL 12(SP), BX + SHRL $0x04, BX + LEAL 4(CX)(BX*1), BX + CMPL BX, 8(SP) + JAE emit_remainder_encodeBlockAsm8B + MOVQ (DX)(CX*1), SI + MOVL BX, 20(SP) + MOVQ $0x9e3779b1, R8 + MOVQ SI, R9 + MOVQ SI, R10 + SHRQ $0x08, R10 + SHLQ $0x20, R9 + IMULQ R8, R9 + SHRQ $0x38, R9 + SHLQ $0x20, R10 + IMULQ R8, R10 + SHRQ $0x38, R10 + MOVL 24(SP)(R9*4), BX + MOVL 24(SP)(R10*4), DI + MOVL CX, 24(SP)(R9*4) + LEAL 1(CX), R9 + MOVL R9, 24(SP)(R10*4) + MOVQ SI, R9 + SHRQ $0x10, R9 + SHLQ $0x20, R9 + IMULQ R8, R9 + SHRQ $0x38, R9 + MOVL CX, R8 + SUBL 16(SP), R8 + MOVL 1(DX)(R8*1), R10 + MOVQ SI, R8 + SHRQ $0x08, R8 + CMPL R8, R10 + JNE no_repeat_found_encodeBlockAsm8B + LEAL 1(CX), SI + MOVL 12(SP), DI + MOVL SI, BX + SUBL 16(SP), BX + JZ repeat_extend_back_end_encodeBlockAsm8B + +repeat_extend_back_loop_encodeBlockAsm8B: + CMPL SI, DI + JBE repeat_extend_back_end_encodeBlockAsm8B + MOVB -1(DX)(BX*1), R8 + MOVB -1(DX)(SI*1), R9 + CMPB R8, R9 + JNE repeat_extend_back_end_encodeBlockAsm8B + LEAL -1(SI), SI + DECL BX + JNZ repeat_extend_back_loop_encodeBlockAsm8B + +repeat_extend_back_end_encodeBlockAsm8B: + MOVL SI, BX + SUBL 12(SP), BX + LEAQ 3(AX)(BX*1), BX + CMPQ BX, (SP) + JB repeat_dst_size_check_encodeBlockAsm8B + MOVQ $0x00000000, ret+48(FP) + RET + +repeat_dst_size_check_encodeBlockAsm8B: + MOVL 12(SP), BX + CMPL BX, SI + JEQ emit_literal_done_repeat_emit_encodeBlockAsm8B + MOVL SI, R8 + MOVL SI, 12(SP) + LEAQ (DX)(BX*1), R9 + SUBL BX, R8 + LEAL -1(R8), BX + CMPL BX, $0x3c + JB one_byte_repeat_emit_encodeBlockAsm8B + CMPL BX, $0x00000100 + JB two_bytes_repeat_emit_encodeBlockAsm8B + JB three_bytes_repeat_emit_encodeBlockAsm8B + +three_bytes_repeat_emit_encodeBlockAsm8B: + MOVB $0xf4, (AX) + MOVW BX, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_repeat_emit_encodeBlockAsm8B + +two_bytes_repeat_emit_encodeBlockAsm8B: + MOVB $0xf0, (AX) + MOVB BL, 1(AX) + ADDQ $0x02, AX + CMPL BX, $0x40 + JB memmove_repeat_emit_encodeBlockAsm8B + JMP memmove_long_repeat_emit_encodeBlockAsm8B + +one_byte_repeat_emit_encodeBlockAsm8B: + SHLB $0x02, BL + MOVB BL, (AX) + ADDQ $0x01, AX + +memmove_repeat_emit_encodeBlockAsm8B: + LEAQ (AX)(R8*1), BX + + // genMemMoveShort + CMPQ R8, $0x08 + JBE emit_lit_memmove_repeat_emit_encodeBlockAsm8B_memmove_move_8 + CMPQ R8, $0x10 + JBE emit_lit_memmove_repeat_emit_encodeBlockAsm8B_memmove_move_8through16 + CMPQ R8, $0x20 + JBE emit_lit_memmove_repeat_emit_encodeBlockAsm8B_memmove_move_17through32 + JMP emit_lit_memmove_repeat_emit_encodeBlockAsm8B_memmove_move_33through64 + +emit_lit_memmove_repeat_emit_encodeBlockAsm8B_memmove_move_8: + MOVQ (R9), R10 + MOVQ R10, (AX) + JMP memmove_end_copy_repeat_emit_encodeBlockAsm8B + +emit_lit_memmove_repeat_emit_encodeBlockAsm8B_memmove_move_8through16: + MOVQ (R9), R10 + MOVQ -8(R9)(R8*1), R9 + MOVQ R10, (AX) + MOVQ R9, -8(AX)(R8*1) + JMP memmove_end_copy_repeat_emit_encodeBlockAsm8B + +emit_lit_memmove_repeat_emit_encodeBlockAsm8B_memmove_move_17through32: + MOVOU (R9), X0 + MOVOU -16(R9)(R8*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(R8*1) + JMP memmove_end_copy_repeat_emit_encodeBlockAsm8B + +emit_lit_memmove_repeat_emit_encodeBlockAsm8B_memmove_move_33through64: + MOVOU (R9), X0 + MOVOU 16(R9), X1 + MOVOU -32(R9)(R8*1), X2 + MOVOU -16(R9)(R8*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + +memmove_end_copy_repeat_emit_encodeBlockAsm8B: + MOVQ BX, AX + JMP emit_literal_done_repeat_emit_encodeBlockAsm8B + +memmove_long_repeat_emit_encodeBlockAsm8B: + LEAQ (AX)(R8*1), BX + + // genMemMoveLong + MOVOU (R9), X0 + MOVOU 16(R9), X1 + MOVOU -32(R9)(R8*1), X2 + MOVOU -16(R9)(R8*1), X3 + MOVQ R8, R11 + SHRQ $0x05, R11 + MOVQ AX, R10 + ANDL $0x0000001f, R10 + MOVQ $0x00000040, R12 + SUBQ R10, R12 + DECQ R11 + JA emit_lit_memmove_long_repeat_emit_encodeBlockAsm8Blarge_forward_sse_loop_32 + LEAQ -32(R9)(R12*1), R10 + LEAQ -32(AX)(R12*1), R13 + +emit_lit_memmove_long_repeat_emit_encodeBlockAsm8Blarge_big_loop_back: + MOVOU (R10), X4 + MOVOU 16(R10), X5 + MOVOA X4, (R13) + MOVOA X5, 16(R13) + ADDQ $0x20, R13 + ADDQ $0x20, R10 + ADDQ $0x20, R12 + DECQ R11 + JNA emit_lit_memmove_long_repeat_emit_encodeBlockAsm8Blarge_big_loop_back + +emit_lit_memmove_long_repeat_emit_encodeBlockAsm8Blarge_forward_sse_loop_32: + MOVOU -32(R9)(R12*1), X4 + MOVOU -16(R9)(R12*1), X5 + MOVOA X4, -32(AX)(R12*1) + MOVOA X5, -16(AX)(R12*1) + ADDQ $0x20, R12 + CMPQ R8, R12 + JAE emit_lit_memmove_long_repeat_emit_encodeBlockAsm8Blarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + MOVQ BX, AX + +emit_literal_done_repeat_emit_encodeBlockAsm8B: + ADDL $0x05, CX + MOVL CX, BX + SUBL 16(SP), BX + MOVQ src_len+32(FP), R8 + SUBL CX, R8 + LEAQ (DX)(CX*1), R9 + LEAQ (DX)(BX*1), BX + + // matchLen + XORL R11, R11 + +matchlen_loopback_16_repeat_extend_encodeBlockAsm8B: + CMPL R8, $0x10 + JB matchlen_match8_repeat_extend_encodeBlockAsm8B + MOVQ (R9)(R11*1), R10 + MOVQ 8(R9)(R11*1), R12 + XORQ (BX)(R11*1), R10 + JNZ matchlen_bsf_8_repeat_extend_encodeBlockAsm8B + XORQ 8(BX)(R11*1), R12 + JNZ matchlen_bsf_16repeat_extend_encodeBlockAsm8B + LEAL -16(R8), R8 + LEAL 16(R11), R11 + JMP matchlen_loopback_16_repeat_extend_encodeBlockAsm8B + +matchlen_bsf_16repeat_extend_encodeBlockAsm8B: +#ifdef GOAMD64_v3 + TZCNTQ R12, R12 + +#else + BSFQ R12, R12 + +#endif + SARQ $0x03, R12 + LEAL 8(R11)(R12*1), R11 + JMP repeat_extend_forward_end_encodeBlockAsm8B + +matchlen_match8_repeat_extend_encodeBlockAsm8B: + CMPL R8, $0x08 + JB matchlen_match4_repeat_extend_encodeBlockAsm8B + MOVQ (R9)(R11*1), R10 + XORQ (BX)(R11*1), R10 + JNZ matchlen_bsf_8_repeat_extend_encodeBlockAsm8B + LEAL -8(R8), R8 + LEAL 8(R11), R11 + JMP matchlen_match4_repeat_extend_encodeBlockAsm8B + +matchlen_bsf_8_repeat_extend_encodeBlockAsm8B: +#ifdef GOAMD64_v3 + TZCNTQ R10, R10 + +#else + BSFQ R10, R10 + +#endif + SARQ $0x03, R10 + LEAL (R11)(R10*1), R11 + JMP repeat_extend_forward_end_encodeBlockAsm8B + +matchlen_match4_repeat_extend_encodeBlockAsm8B: + CMPL R8, $0x04 + JB matchlen_match2_repeat_extend_encodeBlockAsm8B + MOVL (R9)(R11*1), R10 + CMPL (BX)(R11*1), R10 + JNE matchlen_match2_repeat_extend_encodeBlockAsm8B + LEAL -4(R8), R8 + LEAL 4(R11), R11 + +matchlen_match2_repeat_extend_encodeBlockAsm8B: + CMPL R8, $0x01 + JE matchlen_match1_repeat_extend_encodeBlockAsm8B + JB repeat_extend_forward_end_encodeBlockAsm8B + MOVW (R9)(R11*1), R10 + CMPW (BX)(R11*1), R10 + JNE matchlen_match1_repeat_extend_encodeBlockAsm8B + LEAL 2(R11), R11 + SUBL $0x02, R8 + JZ repeat_extend_forward_end_encodeBlockAsm8B + +matchlen_match1_repeat_extend_encodeBlockAsm8B: + MOVB (R9)(R11*1), R10 + CMPB (BX)(R11*1), R10 + JNE repeat_extend_forward_end_encodeBlockAsm8B + LEAL 1(R11), R11 + +repeat_extend_forward_end_encodeBlockAsm8B: + ADDL R11, CX + MOVL CX, BX + SUBL SI, BX + MOVL 16(SP), SI + TESTL DI, DI + JZ repeat_as_copy_encodeBlockAsm8B + + // emitRepeat + MOVL BX, SI + LEAL -4(BX), BX + CMPL SI, $0x08 + JBE repeat_two_match_repeat_encodeBlockAsm8B + CMPL SI, $0x0c + JAE cant_repeat_two_offset_match_repeat_encodeBlockAsm8B + +cant_repeat_two_offset_match_repeat_encodeBlockAsm8B: + CMPL BX, $0x00000104 + JB repeat_three_match_repeat_encodeBlockAsm8B + LEAL -256(BX), BX + MOVW $0x0019, (AX) + MOVW BX, 2(AX) + ADDQ $0x04, AX + JMP repeat_end_emit_encodeBlockAsm8B + +repeat_three_match_repeat_encodeBlockAsm8B: + LEAL -4(BX), BX + MOVW $0x0015, (AX) + MOVB BL, 2(AX) + ADDQ $0x03, AX + JMP repeat_end_emit_encodeBlockAsm8B + +repeat_two_match_repeat_encodeBlockAsm8B: + SHLL $0x02, BX + ORL $0x01, BX + MOVW BX, (AX) + ADDQ $0x02, AX + JMP repeat_end_emit_encodeBlockAsm8B + XORQ DI, DI + LEAL 1(DI)(BX*4), BX + MOVB SI, 1(AX) + SARL $0x08, SI + SHLL $0x05, SI + ORL SI, BX + MOVB BL, (AX) + ADDQ $0x02, AX + JMP repeat_end_emit_encodeBlockAsm8B + +repeat_as_copy_encodeBlockAsm8B: + // emitCopy + CMPL BX, $0x40 + JBE two_byte_offset_short_repeat_as_copy_encodeBlockAsm8B + CMPL SI, $0x00000800 + JAE long_offset_short_repeat_as_copy_encodeBlockAsm8B + MOVL $0x00000001, DI + LEAL 16(DI), DI + MOVB SI, 1(AX) + SHRL $0x08, SI + SHLL $0x05, SI + ORL SI, DI + MOVB DI, (AX) + ADDQ $0x02, AX + SUBL $0x08, BX + + // emitRepeat + LEAL -4(BX), BX + JMP cant_repeat_two_offset_repeat_as_copy_encodeBlockAsm8B_emit_copy_short_2b + MOVL BX, SI + LEAL -4(BX), BX + CMPL SI, $0x08 + JBE repeat_two_repeat_as_copy_encodeBlockAsm8B_emit_copy_short_2b + CMPL SI, $0x0c + JAE cant_repeat_two_offset_repeat_as_copy_encodeBlockAsm8B_emit_copy_short_2b + +cant_repeat_two_offset_repeat_as_copy_encodeBlockAsm8B_emit_copy_short_2b: + CMPL BX, $0x00000104 + JB repeat_three_repeat_as_copy_encodeBlockAsm8B_emit_copy_short_2b + LEAL -256(BX), BX + MOVW $0x0019, (AX) + MOVW BX, 2(AX) + ADDQ $0x04, AX + JMP repeat_end_emit_encodeBlockAsm8B + +repeat_three_repeat_as_copy_encodeBlockAsm8B_emit_copy_short_2b: + LEAL -4(BX), BX + MOVW $0x0015, (AX) + MOVB BL, 2(AX) + ADDQ $0x03, AX + JMP repeat_end_emit_encodeBlockAsm8B + +repeat_two_repeat_as_copy_encodeBlockAsm8B_emit_copy_short_2b: + SHLL $0x02, BX + ORL $0x01, BX + MOVW BX, (AX) + ADDQ $0x02, AX + JMP repeat_end_emit_encodeBlockAsm8B + XORQ DI, DI + LEAL 1(DI)(BX*4), BX + MOVB SI, 1(AX) + SARL $0x08, SI + SHLL $0x05, SI + ORL SI, BX + MOVB BL, (AX) + ADDQ $0x02, AX + JMP repeat_end_emit_encodeBlockAsm8B + +long_offset_short_repeat_as_copy_encodeBlockAsm8B: + MOVB $0xee, (AX) + MOVW SI, 1(AX) + LEAL -60(BX), BX + ADDQ $0x03, AX + + // emitRepeat + MOVL BX, SI + LEAL -4(BX), BX + CMPL SI, $0x08 + JBE repeat_two_repeat_as_copy_encodeBlockAsm8B_emit_copy_short + CMPL SI, $0x0c + JAE cant_repeat_two_offset_repeat_as_copy_encodeBlockAsm8B_emit_copy_short + +cant_repeat_two_offset_repeat_as_copy_encodeBlockAsm8B_emit_copy_short: + CMPL BX, $0x00000104 + JB repeat_three_repeat_as_copy_encodeBlockAsm8B_emit_copy_short + LEAL -256(BX), BX + MOVW $0x0019, (AX) + MOVW BX, 2(AX) + ADDQ $0x04, AX + JMP repeat_end_emit_encodeBlockAsm8B + +repeat_three_repeat_as_copy_encodeBlockAsm8B_emit_copy_short: + LEAL -4(BX), BX + MOVW $0x0015, (AX) + MOVB BL, 2(AX) + ADDQ $0x03, AX + JMP repeat_end_emit_encodeBlockAsm8B + +repeat_two_repeat_as_copy_encodeBlockAsm8B_emit_copy_short: + SHLL $0x02, BX + ORL $0x01, BX + MOVW BX, (AX) + ADDQ $0x02, AX + JMP repeat_end_emit_encodeBlockAsm8B + XORQ DI, DI + LEAL 1(DI)(BX*4), BX + MOVB SI, 1(AX) + SARL $0x08, SI + SHLL $0x05, SI + ORL SI, BX + MOVB BL, (AX) + ADDQ $0x02, AX + JMP repeat_end_emit_encodeBlockAsm8B + +two_byte_offset_short_repeat_as_copy_encodeBlockAsm8B: + MOVL BX, DI + SHLL $0x02, DI + CMPL BX, $0x0c + JAE emit_copy_three_repeat_as_copy_encodeBlockAsm8B + LEAL -15(DI), DI + MOVB SI, 1(AX) + SHRL $0x08, SI + SHLL $0x05, SI + ORL SI, DI + MOVB DI, (AX) + ADDQ $0x02, AX + JMP repeat_end_emit_encodeBlockAsm8B + +emit_copy_three_repeat_as_copy_encodeBlockAsm8B: + LEAL -2(DI), DI + MOVB DI, (AX) + MOVW SI, 1(AX) + ADDQ $0x03, AX + +repeat_end_emit_encodeBlockAsm8B: + MOVL CX, 12(SP) + JMP search_loop_encodeBlockAsm8B + +no_repeat_found_encodeBlockAsm8B: + CMPL (DX)(BX*1), SI + JEQ candidate_match_encodeBlockAsm8B + SHRQ $0x08, SI + MOVL 24(SP)(R9*4), BX + LEAL 2(CX), R8 + CMPL (DX)(DI*1), SI + JEQ candidate2_match_encodeBlockAsm8B + MOVL R8, 24(SP)(R9*4) + SHRQ $0x08, SI + CMPL (DX)(BX*1), SI + JEQ candidate3_match_encodeBlockAsm8B + MOVL 20(SP), CX + JMP search_loop_encodeBlockAsm8B + +candidate3_match_encodeBlockAsm8B: + ADDL $0x02, CX + JMP candidate_match_encodeBlockAsm8B + +candidate2_match_encodeBlockAsm8B: + MOVL R8, 24(SP)(R9*4) + INCL CX + MOVL DI, BX + +candidate_match_encodeBlockAsm8B: + MOVL 12(SP), SI + TESTL BX, BX + JZ match_extend_back_end_encodeBlockAsm8B + +match_extend_back_loop_encodeBlockAsm8B: + CMPL CX, SI + JBE match_extend_back_end_encodeBlockAsm8B + MOVB -1(DX)(BX*1), DI + MOVB -1(DX)(CX*1), R8 + CMPB DI, R8 + JNE match_extend_back_end_encodeBlockAsm8B + LEAL -1(CX), CX + DECL BX + JZ match_extend_back_end_encodeBlockAsm8B + JMP match_extend_back_loop_encodeBlockAsm8B + +match_extend_back_end_encodeBlockAsm8B: + MOVL CX, SI + SUBL 12(SP), SI + LEAQ 3(AX)(SI*1), SI + CMPQ SI, (SP) + JB match_dst_size_check_encodeBlockAsm8B + MOVQ $0x00000000, ret+48(FP) + RET + +match_dst_size_check_encodeBlockAsm8B: + MOVL CX, SI + MOVL 12(SP), DI + CMPL DI, SI + JEQ emit_literal_done_match_emit_encodeBlockAsm8B + MOVL SI, R8 + MOVL SI, 12(SP) + LEAQ (DX)(DI*1), SI + SUBL DI, R8 + LEAL -1(R8), DI + CMPL DI, $0x3c + JB one_byte_match_emit_encodeBlockAsm8B + CMPL DI, $0x00000100 + JB two_bytes_match_emit_encodeBlockAsm8B + JB three_bytes_match_emit_encodeBlockAsm8B + +three_bytes_match_emit_encodeBlockAsm8B: + MOVB $0xf4, (AX) + MOVW DI, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_match_emit_encodeBlockAsm8B + +two_bytes_match_emit_encodeBlockAsm8B: + MOVB $0xf0, (AX) + MOVB DI, 1(AX) + ADDQ $0x02, AX + CMPL DI, $0x40 + JB memmove_match_emit_encodeBlockAsm8B + JMP memmove_long_match_emit_encodeBlockAsm8B + +one_byte_match_emit_encodeBlockAsm8B: + SHLB $0x02, DI + MOVB DI, (AX) + ADDQ $0x01, AX + +memmove_match_emit_encodeBlockAsm8B: + LEAQ (AX)(R8*1), DI + + // genMemMoveShort + CMPQ R8, $0x08 + JBE emit_lit_memmove_match_emit_encodeBlockAsm8B_memmove_move_8 + CMPQ R8, $0x10 + JBE emit_lit_memmove_match_emit_encodeBlockAsm8B_memmove_move_8through16 + CMPQ R8, $0x20 + JBE emit_lit_memmove_match_emit_encodeBlockAsm8B_memmove_move_17through32 + JMP emit_lit_memmove_match_emit_encodeBlockAsm8B_memmove_move_33through64 + +emit_lit_memmove_match_emit_encodeBlockAsm8B_memmove_move_8: + MOVQ (SI), R9 + MOVQ R9, (AX) + JMP memmove_end_copy_match_emit_encodeBlockAsm8B + +emit_lit_memmove_match_emit_encodeBlockAsm8B_memmove_move_8through16: + MOVQ (SI), R9 + MOVQ -8(SI)(R8*1), SI + MOVQ R9, (AX) + MOVQ SI, -8(AX)(R8*1) + JMP memmove_end_copy_match_emit_encodeBlockAsm8B + +emit_lit_memmove_match_emit_encodeBlockAsm8B_memmove_move_17through32: + MOVOU (SI), X0 + MOVOU -16(SI)(R8*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(R8*1) + JMP memmove_end_copy_match_emit_encodeBlockAsm8B + +emit_lit_memmove_match_emit_encodeBlockAsm8B_memmove_move_33through64: + MOVOU (SI), X0 + MOVOU 16(SI), X1 + MOVOU -32(SI)(R8*1), X2 + MOVOU -16(SI)(R8*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + +memmove_end_copy_match_emit_encodeBlockAsm8B: + MOVQ DI, AX + JMP emit_literal_done_match_emit_encodeBlockAsm8B + +memmove_long_match_emit_encodeBlockAsm8B: + LEAQ (AX)(R8*1), DI + + // genMemMoveLong + MOVOU (SI), X0 + MOVOU 16(SI), X1 + MOVOU -32(SI)(R8*1), X2 + MOVOU -16(SI)(R8*1), X3 + MOVQ R8, R10 + SHRQ $0x05, R10 + MOVQ AX, R9 + ANDL $0x0000001f, R9 + MOVQ $0x00000040, R11 + SUBQ R9, R11 + DECQ R10 + JA emit_lit_memmove_long_match_emit_encodeBlockAsm8Blarge_forward_sse_loop_32 + LEAQ -32(SI)(R11*1), R9 + LEAQ -32(AX)(R11*1), R12 + +emit_lit_memmove_long_match_emit_encodeBlockAsm8Blarge_big_loop_back: + MOVOU (R9), X4 + MOVOU 16(R9), X5 + MOVOA X4, (R12) + MOVOA X5, 16(R12) + ADDQ $0x20, R12 + ADDQ $0x20, R9 + ADDQ $0x20, R11 + DECQ R10 + JNA emit_lit_memmove_long_match_emit_encodeBlockAsm8Blarge_big_loop_back + +emit_lit_memmove_long_match_emit_encodeBlockAsm8Blarge_forward_sse_loop_32: + MOVOU -32(SI)(R11*1), X4 + MOVOU -16(SI)(R11*1), X5 + MOVOA X4, -32(AX)(R11*1) + MOVOA X5, -16(AX)(R11*1) + ADDQ $0x20, R11 + CMPQ R8, R11 + JAE emit_lit_memmove_long_match_emit_encodeBlockAsm8Blarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + MOVQ DI, AX + +emit_literal_done_match_emit_encodeBlockAsm8B: +match_nolit_loop_encodeBlockAsm8B: + MOVL CX, SI + SUBL BX, SI + MOVL SI, 16(SP) + ADDL $0x04, CX + ADDL $0x04, BX + MOVQ src_len+32(FP), SI + SUBL CX, SI + LEAQ (DX)(CX*1), DI + LEAQ (DX)(BX*1), BX + + // matchLen + XORL R9, R9 + +matchlen_loopback_16_match_nolit_encodeBlockAsm8B: + CMPL SI, $0x10 + JB matchlen_match8_match_nolit_encodeBlockAsm8B + MOVQ (DI)(R9*1), R8 + MOVQ 8(DI)(R9*1), R10 + XORQ (BX)(R9*1), R8 + JNZ matchlen_bsf_8_match_nolit_encodeBlockAsm8B + XORQ 8(BX)(R9*1), R10 + JNZ matchlen_bsf_16match_nolit_encodeBlockAsm8B + LEAL -16(SI), SI + LEAL 16(R9), R9 + JMP matchlen_loopback_16_match_nolit_encodeBlockAsm8B + +matchlen_bsf_16match_nolit_encodeBlockAsm8B: +#ifdef GOAMD64_v3 + TZCNTQ R10, R10 + +#else + BSFQ R10, R10 + +#endif + SARQ $0x03, R10 + LEAL 8(R9)(R10*1), R9 + JMP match_nolit_end_encodeBlockAsm8B + +matchlen_match8_match_nolit_encodeBlockAsm8B: + CMPL SI, $0x08 + JB matchlen_match4_match_nolit_encodeBlockAsm8B + MOVQ (DI)(R9*1), R8 + XORQ (BX)(R9*1), R8 + JNZ matchlen_bsf_8_match_nolit_encodeBlockAsm8B + LEAL -8(SI), SI + LEAL 8(R9), R9 + JMP matchlen_match4_match_nolit_encodeBlockAsm8B + +matchlen_bsf_8_match_nolit_encodeBlockAsm8B: +#ifdef GOAMD64_v3 + TZCNTQ R8, R8 + +#else + BSFQ R8, R8 + +#endif + SARQ $0x03, R8 + LEAL (R9)(R8*1), R9 + JMP match_nolit_end_encodeBlockAsm8B + +matchlen_match4_match_nolit_encodeBlockAsm8B: + CMPL SI, $0x04 + JB matchlen_match2_match_nolit_encodeBlockAsm8B + MOVL (DI)(R9*1), R8 + CMPL (BX)(R9*1), R8 + JNE matchlen_match2_match_nolit_encodeBlockAsm8B + LEAL -4(SI), SI + LEAL 4(R9), R9 + +matchlen_match2_match_nolit_encodeBlockAsm8B: + CMPL SI, $0x01 + JE matchlen_match1_match_nolit_encodeBlockAsm8B + JB match_nolit_end_encodeBlockAsm8B + MOVW (DI)(R9*1), R8 + CMPW (BX)(R9*1), R8 + JNE matchlen_match1_match_nolit_encodeBlockAsm8B + LEAL 2(R9), R9 + SUBL $0x02, SI + JZ match_nolit_end_encodeBlockAsm8B + +matchlen_match1_match_nolit_encodeBlockAsm8B: + MOVB (DI)(R9*1), R8 + CMPB (BX)(R9*1), R8 + JNE match_nolit_end_encodeBlockAsm8B + LEAL 1(R9), R9 + +match_nolit_end_encodeBlockAsm8B: + ADDL R9, CX + MOVL 16(SP), BX + ADDL $0x04, R9 + MOVL CX, 12(SP) + + // emitCopy + CMPL R9, $0x40 + JBE two_byte_offset_short_match_nolit_encodeBlockAsm8B + CMPL BX, $0x00000800 + JAE long_offset_short_match_nolit_encodeBlockAsm8B + MOVL $0x00000001, SI + LEAL 16(SI), SI + MOVB BL, 1(AX) + SHRL $0x08, BX + SHLL $0x05, BX + ORL BX, SI + MOVB SI, (AX) + ADDQ $0x02, AX + SUBL $0x08, R9 + + // emitRepeat + LEAL -4(R9), R9 + JMP cant_repeat_two_offset_match_nolit_encodeBlockAsm8B_emit_copy_short_2b + MOVL R9, BX + LEAL -4(R9), R9 + CMPL BX, $0x08 + JBE repeat_two_match_nolit_encodeBlockAsm8B_emit_copy_short_2b + CMPL BX, $0x0c + JAE cant_repeat_two_offset_match_nolit_encodeBlockAsm8B_emit_copy_short_2b + +cant_repeat_two_offset_match_nolit_encodeBlockAsm8B_emit_copy_short_2b: + CMPL R9, $0x00000104 + JB repeat_three_match_nolit_encodeBlockAsm8B_emit_copy_short_2b + LEAL -256(R9), R9 + MOVW $0x0019, (AX) + MOVW R9, 2(AX) + ADDQ $0x04, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm8B + +repeat_three_match_nolit_encodeBlockAsm8B_emit_copy_short_2b: + LEAL -4(R9), R9 + MOVW $0x0015, (AX) + MOVB R9, 2(AX) + ADDQ $0x03, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm8B + +repeat_two_match_nolit_encodeBlockAsm8B_emit_copy_short_2b: + SHLL $0x02, R9 + ORL $0x01, R9 + MOVW R9, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm8B + XORQ SI, SI + LEAL 1(SI)(R9*4), R9 + MOVB BL, 1(AX) + SARL $0x08, BX + SHLL $0x05, BX + ORL BX, R9 + MOVB R9, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm8B + +long_offset_short_match_nolit_encodeBlockAsm8B: + MOVB $0xee, (AX) + MOVW BX, 1(AX) + LEAL -60(R9), R9 + ADDQ $0x03, AX + + // emitRepeat + MOVL R9, BX + LEAL -4(R9), R9 + CMPL BX, $0x08 + JBE repeat_two_match_nolit_encodeBlockAsm8B_emit_copy_short + CMPL BX, $0x0c + JAE cant_repeat_two_offset_match_nolit_encodeBlockAsm8B_emit_copy_short + +cant_repeat_two_offset_match_nolit_encodeBlockAsm8B_emit_copy_short: + CMPL R9, $0x00000104 + JB repeat_three_match_nolit_encodeBlockAsm8B_emit_copy_short + LEAL -256(R9), R9 + MOVW $0x0019, (AX) + MOVW R9, 2(AX) + ADDQ $0x04, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm8B + +repeat_three_match_nolit_encodeBlockAsm8B_emit_copy_short: + LEAL -4(R9), R9 + MOVW $0x0015, (AX) + MOVB R9, 2(AX) + ADDQ $0x03, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm8B + +repeat_two_match_nolit_encodeBlockAsm8B_emit_copy_short: + SHLL $0x02, R9 + ORL $0x01, R9 + MOVW R9, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm8B + XORQ SI, SI + LEAL 1(SI)(R9*4), R9 + MOVB BL, 1(AX) + SARL $0x08, BX + SHLL $0x05, BX + ORL BX, R9 + MOVB R9, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm8B + +two_byte_offset_short_match_nolit_encodeBlockAsm8B: + MOVL R9, SI + SHLL $0x02, SI + CMPL R9, $0x0c + JAE emit_copy_three_match_nolit_encodeBlockAsm8B + LEAL -15(SI), SI + MOVB BL, 1(AX) + SHRL $0x08, BX + SHLL $0x05, BX + ORL BX, SI + MOVB SI, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBlockAsm8B + +emit_copy_three_match_nolit_encodeBlockAsm8B: + LEAL -2(SI), SI + MOVB SI, (AX) + MOVW BX, 1(AX) + ADDQ $0x03, AX + +match_nolit_emitcopy_end_encodeBlockAsm8B: + CMPL CX, 8(SP) + JAE emit_remainder_encodeBlockAsm8B + MOVQ -2(DX)(CX*1), SI + CMPQ AX, (SP) + JB match_nolit_dst_ok_encodeBlockAsm8B + MOVQ $0x00000000, ret+48(FP) + RET + +match_nolit_dst_ok_encodeBlockAsm8B: + MOVQ $0x9e3779b1, R8 + MOVQ SI, DI + SHRQ $0x10, SI + MOVQ SI, BX + SHLQ $0x20, DI + IMULQ R8, DI + SHRQ $0x38, DI + SHLQ $0x20, BX + IMULQ R8, BX + SHRQ $0x38, BX + LEAL -2(CX), R8 + LEAQ 24(SP)(BX*4), R9 + MOVL (R9), BX + MOVL R8, 24(SP)(DI*4) + MOVL CX, (R9) + CMPL (DX)(BX*1), SI + JEQ match_nolit_loop_encodeBlockAsm8B + INCL CX + JMP search_loop_encodeBlockAsm8B + +emit_remainder_encodeBlockAsm8B: + MOVQ src_len+32(FP), CX + SUBL 12(SP), CX + LEAQ 3(AX)(CX*1), CX + CMPQ CX, (SP) + JB emit_remainder_ok_encodeBlockAsm8B + MOVQ $0x00000000, ret+48(FP) + RET + +emit_remainder_ok_encodeBlockAsm8B: + MOVQ src_len+32(FP), CX + MOVL 12(SP), BX + CMPL BX, CX + JEQ emit_literal_done_emit_remainder_encodeBlockAsm8B + MOVL CX, SI + MOVL CX, 12(SP) + LEAQ (DX)(BX*1), CX + SUBL BX, SI + LEAL -1(SI), DX + CMPL DX, $0x3c + JB one_byte_emit_remainder_encodeBlockAsm8B + CMPL DX, $0x00000100 + JB two_bytes_emit_remainder_encodeBlockAsm8B + JB three_bytes_emit_remainder_encodeBlockAsm8B + +three_bytes_emit_remainder_encodeBlockAsm8B: + MOVB $0xf4, (AX) + MOVW DX, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_emit_remainder_encodeBlockAsm8B + +two_bytes_emit_remainder_encodeBlockAsm8B: + MOVB $0xf0, (AX) + MOVB DL, 1(AX) + ADDQ $0x02, AX + CMPL DX, $0x40 + JB memmove_emit_remainder_encodeBlockAsm8B + JMP memmove_long_emit_remainder_encodeBlockAsm8B + +one_byte_emit_remainder_encodeBlockAsm8B: + SHLB $0x02, DL + MOVB DL, (AX) + ADDQ $0x01, AX + +memmove_emit_remainder_encodeBlockAsm8B: + LEAQ (AX)(SI*1), DX + MOVL SI, BX + + // genMemMoveShort + CMPQ BX, $0x03 + JB emit_lit_memmove_emit_remainder_encodeBlockAsm8B_memmove_move_1or2 + JE emit_lit_memmove_emit_remainder_encodeBlockAsm8B_memmove_move_3 + CMPQ BX, $0x08 + JB emit_lit_memmove_emit_remainder_encodeBlockAsm8B_memmove_move_4through7 + CMPQ BX, $0x10 + JBE emit_lit_memmove_emit_remainder_encodeBlockAsm8B_memmove_move_8through16 + CMPQ BX, $0x20 + JBE emit_lit_memmove_emit_remainder_encodeBlockAsm8B_memmove_move_17through32 + JMP emit_lit_memmove_emit_remainder_encodeBlockAsm8B_memmove_move_33through64 + +emit_lit_memmove_emit_remainder_encodeBlockAsm8B_memmove_move_1or2: + MOVB (CX), SI + MOVB -1(CX)(BX*1), CL + MOVB SI, (AX) + MOVB CL, -1(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeBlockAsm8B + +emit_lit_memmove_emit_remainder_encodeBlockAsm8B_memmove_move_3: + MOVW (CX), SI + MOVB 2(CX), CL + MOVW SI, (AX) + MOVB CL, 2(AX) + JMP memmove_end_copy_emit_remainder_encodeBlockAsm8B + +emit_lit_memmove_emit_remainder_encodeBlockAsm8B_memmove_move_4through7: + MOVL (CX), SI + MOVL -4(CX)(BX*1), CX + MOVL SI, (AX) + MOVL CX, -4(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeBlockAsm8B + +emit_lit_memmove_emit_remainder_encodeBlockAsm8B_memmove_move_8through16: + MOVQ (CX), SI + MOVQ -8(CX)(BX*1), CX + MOVQ SI, (AX) + MOVQ CX, -8(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeBlockAsm8B + +emit_lit_memmove_emit_remainder_encodeBlockAsm8B_memmove_move_17through32: + MOVOU (CX), X0 + MOVOU -16(CX)(BX*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeBlockAsm8B + +emit_lit_memmove_emit_remainder_encodeBlockAsm8B_memmove_move_33through64: + MOVOU (CX), X0 + MOVOU 16(CX), X1 + MOVOU -32(CX)(BX*1), X2 + MOVOU -16(CX)(BX*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(BX*1) + MOVOU X3, -16(AX)(BX*1) + +memmove_end_copy_emit_remainder_encodeBlockAsm8B: + MOVQ DX, AX + JMP emit_literal_done_emit_remainder_encodeBlockAsm8B + +memmove_long_emit_remainder_encodeBlockAsm8B: + LEAQ (AX)(SI*1), DX + MOVL SI, BX + + // genMemMoveLong + MOVOU (CX), X0 + MOVOU 16(CX), X1 + MOVOU -32(CX)(BX*1), X2 + MOVOU -16(CX)(BX*1), X3 + MOVQ BX, DI + SHRQ $0x05, DI + MOVQ AX, SI + ANDL $0x0000001f, SI + MOVQ $0x00000040, R8 + SUBQ SI, R8 + DECQ DI + JA emit_lit_memmove_long_emit_remainder_encodeBlockAsm8Blarge_forward_sse_loop_32 + LEAQ -32(CX)(R8*1), SI + LEAQ -32(AX)(R8*1), R9 + +emit_lit_memmove_long_emit_remainder_encodeBlockAsm8Blarge_big_loop_back: + MOVOU (SI), X4 + MOVOU 16(SI), X5 + MOVOA X4, (R9) + MOVOA X5, 16(R9) + ADDQ $0x20, R9 + ADDQ $0x20, SI + ADDQ $0x20, R8 + DECQ DI + JNA emit_lit_memmove_long_emit_remainder_encodeBlockAsm8Blarge_big_loop_back + +emit_lit_memmove_long_emit_remainder_encodeBlockAsm8Blarge_forward_sse_loop_32: + MOVOU -32(CX)(R8*1), X4 + MOVOU -16(CX)(R8*1), X5 + MOVOA X4, -32(AX)(R8*1) + MOVOA X5, -16(AX)(R8*1) + ADDQ $0x20, R8 + CMPQ BX, R8 + JAE emit_lit_memmove_long_emit_remainder_encodeBlockAsm8Blarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(BX*1) + MOVOU X3, -16(AX)(BX*1) + MOVQ DX, AX + +emit_literal_done_emit_remainder_encodeBlockAsm8B: + MOVQ dst_base+0(FP), CX + SUBQ CX, AX + MOVQ AX, ret+48(FP) + RET + +// func encodeBetterBlockAsm(dst []byte, src []byte) int +// Requires: BMI, SSE2 +TEXT ·encodeBetterBlockAsm(SB), $589848-56 + MOVQ dst_base+0(FP), AX + MOVQ $0x00001200, CX + LEAQ 24(SP), DX + PXOR X0, X0 + +zero_loop_encodeBetterBlockAsm: + MOVOU X0, (DX) + MOVOU X0, 16(DX) + MOVOU X0, 32(DX) + MOVOU X0, 48(DX) + MOVOU X0, 64(DX) + MOVOU X0, 80(DX) + MOVOU X0, 96(DX) + MOVOU X0, 112(DX) + ADDQ $0x80, DX + DECQ CX + JNZ zero_loop_encodeBetterBlockAsm + MOVL $0x00000000, 12(SP) + MOVQ src_len+32(FP), CX + LEAQ -6(CX), DX + LEAQ -8(CX), BX + MOVL BX, 8(SP) + SHRQ $0x05, CX + SUBL CX, DX + LEAQ (AX)(DX*1), DX + MOVQ DX, (SP) + MOVL $0x00000001, CX + MOVL $0x00000000, 16(SP) + MOVQ src_base+24(FP), DX + +search_loop_encodeBetterBlockAsm: + MOVL CX, BX + SUBL 12(SP), BX + SHRL $0x07, BX + CMPL BX, $0x63 + JBE check_maxskip_ok_encodeBetterBlockAsm + LEAL 100(CX), BX + JMP check_maxskip_cont_encodeBetterBlockAsm + +check_maxskip_ok_encodeBetterBlockAsm: + LEAL 1(CX)(BX*1), BX + +check_maxskip_cont_encodeBetterBlockAsm: + CMPL BX, 8(SP) + JAE emit_remainder_encodeBetterBlockAsm + MOVQ (DX)(CX*1), SI + MOVL BX, 20(SP) + MOVQ $0x00cf1bbcdcbfa563, R8 + MOVQ $0x9e3779b1, BX + MOVQ SI, R9 + MOVQ SI, R10 + SHLQ $0x08, R9 + IMULQ R8, R9 + SHRQ $0x2f, R9 + SHLQ $0x20, R10 + IMULQ BX, R10 + SHRQ $0x32, R10 + MOVL 24(SP)(R9*4), BX + MOVL 524312(SP)(R10*4), DI + MOVL CX, 24(SP)(R9*4) + MOVL CX, 524312(SP)(R10*4) + MOVQ (DX)(BX*1), R9 + MOVQ (DX)(DI*1), R10 + CMPQ R9, SI + JEQ candidate_match_encodeBetterBlockAsm + CMPQ R10, SI + JNE no_short_found_encodeBetterBlockAsm + MOVL DI, BX + JMP candidate_match_encodeBetterBlockAsm + +no_short_found_encodeBetterBlockAsm: + CMPL R9, SI + JEQ candidate_match_encodeBetterBlockAsm + CMPL R10, SI + JEQ candidateS_match_encodeBetterBlockAsm + MOVL 20(SP), CX + JMP search_loop_encodeBetterBlockAsm + +candidateS_match_encodeBetterBlockAsm: + SHRQ $0x08, SI + MOVQ SI, R9 + SHLQ $0x08, R9 + IMULQ R8, R9 + SHRQ $0x2f, R9 + MOVL 24(SP)(R9*4), BX + INCL CX + MOVL CX, 24(SP)(R9*4) + CMPL (DX)(BX*1), SI + JEQ candidate_match_encodeBetterBlockAsm + DECL CX + MOVL DI, BX + +candidate_match_encodeBetterBlockAsm: + MOVL 12(SP), SI + TESTL BX, BX + JZ match_extend_back_end_encodeBetterBlockAsm + +match_extend_back_loop_encodeBetterBlockAsm: + CMPL CX, SI + JBE match_extend_back_end_encodeBetterBlockAsm + MOVB -1(DX)(BX*1), DI + MOVB -1(DX)(CX*1), R8 + CMPB DI, R8 + JNE match_extend_back_end_encodeBetterBlockAsm + LEAL -1(CX), CX + DECL BX + JZ match_extend_back_end_encodeBetterBlockAsm + JMP match_extend_back_loop_encodeBetterBlockAsm + +match_extend_back_end_encodeBetterBlockAsm: + MOVL CX, SI + SUBL 12(SP), SI + LEAQ 5(AX)(SI*1), SI + CMPQ SI, (SP) + JB match_dst_size_check_encodeBetterBlockAsm + MOVQ $0x00000000, ret+48(FP) + RET + +match_dst_size_check_encodeBetterBlockAsm: + MOVL CX, SI + ADDL $0x04, CX + ADDL $0x04, BX + MOVQ src_len+32(FP), DI + SUBL CX, DI + LEAQ (DX)(CX*1), R8 + LEAQ (DX)(BX*1), R9 + + // matchLen + XORL R11, R11 + +matchlen_loopback_16_match_nolit_encodeBetterBlockAsm: + CMPL DI, $0x10 + JB matchlen_match8_match_nolit_encodeBetterBlockAsm + MOVQ (R8)(R11*1), R10 + MOVQ 8(R8)(R11*1), R12 + XORQ (R9)(R11*1), R10 + JNZ matchlen_bsf_8_match_nolit_encodeBetterBlockAsm + XORQ 8(R9)(R11*1), R12 + JNZ matchlen_bsf_16match_nolit_encodeBetterBlockAsm + LEAL -16(DI), DI + LEAL 16(R11), R11 + JMP matchlen_loopback_16_match_nolit_encodeBetterBlockAsm + +matchlen_bsf_16match_nolit_encodeBetterBlockAsm: +#ifdef GOAMD64_v3 + TZCNTQ R12, R12 + +#else + BSFQ R12, R12 + +#endif + SARQ $0x03, R12 + LEAL 8(R11)(R12*1), R11 + JMP match_nolit_end_encodeBetterBlockAsm + +matchlen_match8_match_nolit_encodeBetterBlockAsm: + CMPL DI, $0x08 + JB matchlen_match4_match_nolit_encodeBetterBlockAsm + MOVQ (R8)(R11*1), R10 + XORQ (R9)(R11*1), R10 + JNZ matchlen_bsf_8_match_nolit_encodeBetterBlockAsm + LEAL -8(DI), DI + LEAL 8(R11), R11 + JMP matchlen_match4_match_nolit_encodeBetterBlockAsm + +matchlen_bsf_8_match_nolit_encodeBetterBlockAsm: +#ifdef GOAMD64_v3 + TZCNTQ R10, R10 + +#else + BSFQ R10, R10 + +#endif + SARQ $0x03, R10 + LEAL (R11)(R10*1), R11 + JMP match_nolit_end_encodeBetterBlockAsm + +matchlen_match4_match_nolit_encodeBetterBlockAsm: + CMPL DI, $0x04 + JB matchlen_match2_match_nolit_encodeBetterBlockAsm + MOVL (R8)(R11*1), R10 + CMPL (R9)(R11*1), R10 + JNE matchlen_match2_match_nolit_encodeBetterBlockAsm + LEAL -4(DI), DI + LEAL 4(R11), R11 + +matchlen_match2_match_nolit_encodeBetterBlockAsm: + CMPL DI, $0x01 + JE matchlen_match1_match_nolit_encodeBetterBlockAsm + JB match_nolit_end_encodeBetterBlockAsm + MOVW (R8)(R11*1), R10 + CMPW (R9)(R11*1), R10 + JNE matchlen_match1_match_nolit_encodeBetterBlockAsm + LEAL 2(R11), R11 + SUBL $0x02, DI + JZ match_nolit_end_encodeBetterBlockAsm + +matchlen_match1_match_nolit_encodeBetterBlockAsm: + MOVB (R8)(R11*1), R10 + CMPB (R9)(R11*1), R10 + JNE match_nolit_end_encodeBetterBlockAsm + LEAL 1(R11), R11 + +match_nolit_end_encodeBetterBlockAsm: + MOVL CX, DI + SUBL BX, DI + + // Check if repeat + CMPL 16(SP), DI + JEQ match_is_repeat_encodeBetterBlockAsm + CMPL R11, $0x01 + JA match_length_ok_encodeBetterBlockAsm + CMPL DI, $0x0000ffff + JBE match_length_ok_encodeBetterBlockAsm + MOVL 20(SP), CX + INCL CX + JMP search_loop_encodeBetterBlockAsm + +match_length_ok_encodeBetterBlockAsm: + MOVL DI, 16(SP) + MOVL 12(SP), BX + CMPL BX, SI + JEQ emit_literal_done_match_emit_encodeBetterBlockAsm + MOVL SI, R8 + MOVL SI, 12(SP) + LEAQ (DX)(BX*1), R9 + SUBL BX, R8 + LEAL -1(R8), BX + CMPL BX, $0x3c + JB one_byte_match_emit_encodeBetterBlockAsm + CMPL BX, $0x00000100 + JB two_bytes_match_emit_encodeBetterBlockAsm + CMPL BX, $0x00010000 + JB three_bytes_match_emit_encodeBetterBlockAsm + CMPL BX, $0x01000000 + JB four_bytes_match_emit_encodeBetterBlockAsm + MOVB $0xfc, (AX) + MOVL BX, 1(AX) + ADDQ $0x05, AX + JMP memmove_long_match_emit_encodeBetterBlockAsm + +four_bytes_match_emit_encodeBetterBlockAsm: + MOVL BX, R10 + SHRL $0x10, R10 + MOVB $0xf8, (AX) + MOVW BX, 1(AX) + MOVB R10, 3(AX) + ADDQ $0x04, AX + JMP memmove_long_match_emit_encodeBetterBlockAsm + +three_bytes_match_emit_encodeBetterBlockAsm: + MOVB $0xf4, (AX) + MOVW BX, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_match_emit_encodeBetterBlockAsm + +two_bytes_match_emit_encodeBetterBlockAsm: + MOVB $0xf0, (AX) + MOVB BL, 1(AX) + ADDQ $0x02, AX + CMPL BX, $0x40 + JB memmove_match_emit_encodeBetterBlockAsm + JMP memmove_long_match_emit_encodeBetterBlockAsm + +one_byte_match_emit_encodeBetterBlockAsm: + SHLB $0x02, BL + MOVB BL, (AX) + ADDQ $0x01, AX + +memmove_match_emit_encodeBetterBlockAsm: + LEAQ (AX)(R8*1), BX + + // genMemMoveShort + CMPQ R8, $0x04 + JBE emit_lit_memmove_match_emit_encodeBetterBlockAsm_memmove_move_4 + CMPQ R8, $0x08 + JB emit_lit_memmove_match_emit_encodeBetterBlockAsm_memmove_move_4through7 + CMPQ R8, $0x10 + JBE emit_lit_memmove_match_emit_encodeBetterBlockAsm_memmove_move_8through16 + CMPQ R8, $0x20 + JBE emit_lit_memmove_match_emit_encodeBetterBlockAsm_memmove_move_17through32 + JMP emit_lit_memmove_match_emit_encodeBetterBlockAsm_memmove_move_33through64 + +emit_lit_memmove_match_emit_encodeBetterBlockAsm_memmove_move_4: + MOVL (R9), R10 + MOVL R10, (AX) + JMP memmove_end_copy_match_emit_encodeBetterBlockAsm + +emit_lit_memmove_match_emit_encodeBetterBlockAsm_memmove_move_4through7: + MOVL (R9), R10 + MOVL -4(R9)(R8*1), R9 + MOVL R10, (AX) + MOVL R9, -4(AX)(R8*1) + JMP memmove_end_copy_match_emit_encodeBetterBlockAsm + +emit_lit_memmove_match_emit_encodeBetterBlockAsm_memmove_move_8through16: + MOVQ (R9), R10 + MOVQ -8(R9)(R8*1), R9 + MOVQ R10, (AX) + MOVQ R9, -8(AX)(R8*1) + JMP memmove_end_copy_match_emit_encodeBetterBlockAsm + +emit_lit_memmove_match_emit_encodeBetterBlockAsm_memmove_move_17through32: + MOVOU (R9), X0 + MOVOU -16(R9)(R8*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(R8*1) + JMP memmove_end_copy_match_emit_encodeBetterBlockAsm + +emit_lit_memmove_match_emit_encodeBetterBlockAsm_memmove_move_33through64: + MOVOU (R9), X0 + MOVOU 16(R9), X1 + MOVOU -32(R9)(R8*1), X2 + MOVOU -16(R9)(R8*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + +memmove_end_copy_match_emit_encodeBetterBlockAsm: + MOVQ BX, AX + JMP emit_literal_done_match_emit_encodeBetterBlockAsm + +memmove_long_match_emit_encodeBetterBlockAsm: + LEAQ (AX)(R8*1), BX + + // genMemMoveLong + MOVOU (R9), X0 + MOVOU 16(R9), X1 + MOVOU -32(R9)(R8*1), X2 + MOVOU -16(R9)(R8*1), X3 + MOVQ R8, R12 + SHRQ $0x05, R12 + MOVQ AX, R10 + ANDL $0x0000001f, R10 + MOVQ $0x00000040, R13 + SUBQ R10, R13 + DECQ R12 + JA emit_lit_memmove_long_match_emit_encodeBetterBlockAsmlarge_forward_sse_loop_32 + LEAQ -32(R9)(R13*1), R10 + LEAQ -32(AX)(R13*1), R14 + +emit_lit_memmove_long_match_emit_encodeBetterBlockAsmlarge_big_loop_back: + MOVOU (R10), X4 + MOVOU 16(R10), X5 + MOVOA X4, (R14) + MOVOA X5, 16(R14) + ADDQ $0x20, R14 + ADDQ $0x20, R10 + ADDQ $0x20, R13 + DECQ R12 + JNA emit_lit_memmove_long_match_emit_encodeBetterBlockAsmlarge_big_loop_back + +emit_lit_memmove_long_match_emit_encodeBetterBlockAsmlarge_forward_sse_loop_32: + MOVOU -32(R9)(R13*1), X4 + MOVOU -16(R9)(R13*1), X5 + MOVOA X4, -32(AX)(R13*1) + MOVOA X5, -16(AX)(R13*1) + ADDQ $0x20, R13 + CMPQ R8, R13 + JAE emit_lit_memmove_long_match_emit_encodeBetterBlockAsmlarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + MOVQ BX, AX + +emit_literal_done_match_emit_encodeBetterBlockAsm: + ADDL R11, CX + ADDL $0x04, R11 + MOVL CX, 12(SP) + + // emitCopy + CMPL DI, $0x00010000 + JB two_byte_offset_match_nolit_encodeBetterBlockAsm + CMPL R11, $0x40 + JBE four_bytes_remain_match_nolit_encodeBetterBlockAsm + MOVB $0xff, (AX) + MOVL DI, 1(AX) + LEAL -64(R11), R11 + ADDQ $0x05, AX + CMPL R11, $0x04 + JB four_bytes_remain_match_nolit_encodeBetterBlockAsm + + // emitRepeat +emit_repeat_again_match_nolit_encodeBetterBlockAsm_emit_copy: + MOVL R11, BX + LEAL -4(R11), R11 + CMPL BX, $0x08 + JBE repeat_two_match_nolit_encodeBetterBlockAsm_emit_copy + CMPL BX, $0x0c + JAE cant_repeat_two_offset_match_nolit_encodeBetterBlockAsm_emit_copy + CMPL DI, $0x00000800 + JB repeat_two_offset_match_nolit_encodeBetterBlockAsm_emit_copy + +cant_repeat_two_offset_match_nolit_encodeBetterBlockAsm_emit_copy: + CMPL R11, $0x00000104 + JB repeat_three_match_nolit_encodeBetterBlockAsm_emit_copy + CMPL R11, $0x00010100 + JB repeat_four_match_nolit_encodeBetterBlockAsm_emit_copy + CMPL R11, $0x0100ffff + JB repeat_five_match_nolit_encodeBetterBlockAsm_emit_copy + LEAL -16842747(R11), R11 + MOVL $0xfffb001d, (AX) + MOVB $0xff, 4(AX) + ADDQ $0x05, AX + JMP emit_repeat_again_match_nolit_encodeBetterBlockAsm_emit_copy + +repeat_five_match_nolit_encodeBetterBlockAsm_emit_copy: + LEAL -65536(R11), R11 + MOVL R11, DI + MOVW $0x001d, (AX) + MOVW R11, 2(AX) + SARL $0x10, DI + MOVB DI, 4(AX) + ADDQ $0x05, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm + +repeat_four_match_nolit_encodeBetterBlockAsm_emit_copy: + LEAL -256(R11), R11 + MOVW $0x0019, (AX) + MOVW R11, 2(AX) + ADDQ $0x04, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm + +repeat_three_match_nolit_encodeBetterBlockAsm_emit_copy: + LEAL -4(R11), R11 + MOVW $0x0015, (AX) + MOVB R11, 2(AX) + ADDQ $0x03, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm + +repeat_two_match_nolit_encodeBetterBlockAsm_emit_copy: + SHLL $0x02, R11 + ORL $0x01, R11 + MOVW R11, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm + +repeat_two_offset_match_nolit_encodeBetterBlockAsm_emit_copy: + XORQ BX, BX + LEAL 1(BX)(R11*4), R11 + MOVB DI, 1(AX) + SARL $0x08, DI + SHLL $0x05, DI + ORL DI, R11 + MOVB R11, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm + +four_bytes_remain_match_nolit_encodeBetterBlockAsm: + TESTL R11, R11 + JZ match_nolit_emitcopy_end_encodeBetterBlockAsm + XORL BX, BX + LEAL -1(BX)(R11*4), R11 + MOVB R11, (AX) + MOVL DI, 1(AX) + ADDQ $0x05, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm + +two_byte_offset_match_nolit_encodeBetterBlockAsm: + CMPL R11, $0x40 + JBE two_byte_offset_short_match_nolit_encodeBetterBlockAsm + CMPL DI, $0x00000800 + JAE long_offset_short_match_nolit_encodeBetterBlockAsm + MOVL $0x00000001, BX + LEAL 16(BX), BX + MOVB DI, 1(AX) + MOVL DI, R8 + SHRL $0x08, R8 + SHLL $0x05, R8 + ORL R8, BX + MOVB BL, (AX) + ADDQ $0x02, AX + SUBL $0x08, R11 + + // emitRepeat + LEAL -4(R11), R11 + JMP cant_repeat_two_offset_match_nolit_encodeBetterBlockAsm_emit_copy_short_2b + +emit_repeat_again_match_nolit_encodeBetterBlockAsm_emit_copy_short_2b: + MOVL R11, BX + LEAL -4(R11), R11 + CMPL BX, $0x08 + JBE repeat_two_match_nolit_encodeBetterBlockAsm_emit_copy_short_2b + CMPL BX, $0x0c + JAE cant_repeat_two_offset_match_nolit_encodeBetterBlockAsm_emit_copy_short_2b + CMPL DI, $0x00000800 + JB repeat_two_offset_match_nolit_encodeBetterBlockAsm_emit_copy_short_2b + +cant_repeat_two_offset_match_nolit_encodeBetterBlockAsm_emit_copy_short_2b: + CMPL R11, $0x00000104 + JB repeat_three_match_nolit_encodeBetterBlockAsm_emit_copy_short_2b + CMPL R11, $0x00010100 + JB repeat_four_match_nolit_encodeBetterBlockAsm_emit_copy_short_2b + CMPL R11, $0x0100ffff + JB repeat_five_match_nolit_encodeBetterBlockAsm_emit_copy_short_2b + LEAL -16842747(R11), R11 + MOVL $0xfffb001d, (AX) + MOVB $0xff, 4(AX) + ADDQ $0x05, AX + JMP emit_repeat_again_match_nolit_encodeBetterBlockAsm_emit_copy_short_2b + +repeat_five_match_nolit_encodeBetterBlockAsm_emit_copy_short_2b: + LEAL -65536(R11), R11 + MOVL R11, DI + MOVW $0x001d, (AX) + MOVW R11, 2(AX) + SARL $0x10, DI + MOVB DI, 4(AX) + ADDQ $0x05, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm + +repeat_four_match_nolit_encodeBetterBlockAsm_emit_copy_short_2b: + LEAL -256(R11), R11 + MOVW $0x0019, (AX) + MOVW R11, 2(AX) + ADDQ $0x04, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm + +repeat_three_match_nolit_encodeBetterBlockAsm_emit_copy_short_2b: + LEAL -4(R11), R11 + MOVW $0x0015, (AX) + MOVB R11, 2(AX) + ADDQ $0x03, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm + +repeat_two_match_nolit_encodeBetterBlockAsm_emit_copy_short_2b: + SHLL $0x02, R11 + ORL $0x01, R11 + MOVW R11, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm + +repeat_two_offset_match_nolit_encodeBetterBlockAsm_emit_copy_short_2b: + XORQ BX, BX + LEAL 1(BX)(R11*4), R11 + MOVB DI, 1(AX) + SARL $0x08, DI + SHLL $0x05, DI + ORL DI, R11 + MOVB R11, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm + +long_offset_short_match_nolit_encodeBetterBlockAsm: + MOVB $0xee, (AX) + MOVW DI, 1(AX) + LEAL -60(R11), R11 + ADDQ $0x03, AX + + // emitRepeat +emit_repeat_again_match_nolit_encodeBetterBlockAsm_emit_copy_short: + MOVL R11, BX + LEAL -4(R11), R11 + CMPL BX, $0x08 + JBE repeat_two_match_nolit_encodeBetterBlockAsm_emit_copy_short + CMPL BX, $0x0c + JAE cant_repeat_two_offset_match_nolit_encodeBetterBlockAsm_emit_copy_short + CMPL DI, $0x00000800 + JB repeat_two_offset_match_nolit_encodeBetterBlockAsm_emit_copy_short + +cant_repeat_two_offset_match_nolit_encodeBetterBlockAsm_emit_copy_short: + CMPL R11, $0x00000104 + JB repeat_three_match_nolit_encodeBetterBlockAsm_emit_copy_short + CMPL R11, $0x00010100 + JB repeat_four_match_nolit_encodeBetterBlockAsm_emit_copy_short + CMPL R11, $0x0100ffff + JB repeat_five_match_nolit_encodeBetterBlockAsm_emit_copy_short + LEAL -16842747(R11), R11 + MOVL $0xfffb001d, (AX) + MOVB $0xff, 4(AX) + ADDQ $0x05, AX + JMP emit_repeat_again_match_nolit_encodeBetterBlockAsm_emit_copy_short + +repeat_five_match_nolit_encodeBetterBlockAsm_emit_copy_short: + LEAL -65536(R11), R11 + MOVL R11, DI + MOVW $0x001d, (AX) + MOVW R11, 2(AX) + SARL $0x10, DI + MOVB DI, 4(AX) + ADDQ $0x05, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm + +repeat_four_match_nolit_encodeBetterBlockAsm_emit_copy_short: + LEAL -256(R11), R11 + MOVW $0x0019, (AX) + MOVW R11, 2(AX) + ADDQ $0x04, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm + +repeat_three_match_nolit_encodeBetterBlockAsm_emit_copy_short: + LEAL -4(R11), R11 + MOVW $0x0015, (AX) + MOVB R11, 2(AX) + ADDQ $0x03, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm + +repeat_two_match_nolit_encodeBetterBlockAsm_emit_copy_short: + SHLL $0x02, R11 + ORL $0x01, R11 + MOVW R11, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm + +repeat_two_offset_match_nolit_encodeBetterBlockAsm_emit_copy_short: + XORQ BX, BX + LEAL 1(BX)(R11*4), R11 + MOVB DI, 1(AX) + SARL $0x08, DI + SHLL $0x05, DI + ORL DI, R11 + MOVB R11, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm + +two_byte_offset_short_match_nolit_encodeBetterBlockAsm: + MOVL R11, BX + SHLL $0x02, BX + CMPL R11, $0x0c + JAE emit_copy_three_match_nolit_encodeBetterBlockAsm + CMPL DI, $0x00000800 + JAE emit_copy_three_match_nolit_encodeBetterBlockAsm + LEAL -15(BX), BX + MOVB DI, 1(AX) + SHRL $0x08, DI + SHLL $0x05, DI + ORL DI, BX + MOVB BL, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm + +emit_copy_three_match_nolit_encodeBetterBlockAsm: + LEAL -2(BX), BX + MOVB BL, (AX) + MOVW DI, 1(AX) + ADDQ $0x03, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm + +match_is_repeat_encodeBetterBlockAsm: + MOVL 12(SP), BX + CMPL BX, SI + JEQ emit_literal_done_match_emit_repeat_encodeBetterBlockAsm + MOVL SI, R8 + MOVL SI, 12(SP) + LEAQ (DX)(BX*1), R9 + SUBL BX, R8 + LEAL -1(R8), BX + CMPL BX, $0x3c + JB one_byte_match_emit_repeat_encodeBetterBlockAsm + CMPL BX, $0x00000100 + JB two_bytes_match_emit_repeat_encodeBetterBlockAsm + CMPL BX, $0x00010000 + JB three_bytes_match_emit_repeat_encodeBetterBlockAsm + CMPL BX, $0x01000000 + JB four_bytes_match_emit_repeat_encodeBetterBlockAsm + MOVB $0xfc, (AX) + MOVL BX, 1(AX) + ADDQ $0x05, AX + JMP memmove_long_match_emit_repeat_encodeBetterBlockAsm + +four_bytes_match_emit_repeat_encodeBetterBlockAsm: + MOVL BX, R10 + SHRL $0x10, R10 + MOVB $0xf8, (AX) + MOVW BX, 1(AX) + MOVB R10, 3(AX) + ADDQ $0x04, AX + JMP memmove_long_match_emit_repeat_encodeBetterBlockAsm + +three_bytes_match_emit_repeat_encodeBetterBlockAsm: + MOVB $0xf4, (AX) + MOVW BX, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_match_emit_repeat_encodeBetterBlockAsm + +two_bytes_match_emit_repeat_encodeBetterBlockAsm: + MOVB $0xf0, (AX) + MOVB BL, 1(AX) + ADDQ $0x02, AX + CMPL BX, $0x40 + JB memmove_match_emit_repeat_encodeBetterBlockAsm + JMP memmove_long_match_emit_repeat_encodeBetterBlockAsm + +one_byte_match_emit_repeat_encodeBetterBlockAsm: + SHLB $0x02, BL + MOVB BL, (AX) + ADDQ $0x01, AX + +memmove_match_emit_repeat_encodeBetterBlockAsm: + LEAQ (AX)(R8*1), BX + + // genMemMoveShort + CMPQ R8, $0x04 + JBE emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm_memmove_move_4 + CMPQ R8, $0x08 + JB emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm_memmove_move_4through7 + CMPQ R8, $0x10 + JBE emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm_memmove_move_8through16 + CMPQ R8, $0x20 + JBE emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm_memmove_move_17through32 + JMP emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm_memmove_move_33through64 + +emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm_memmove_move_4: + MOVL (R9), R10 + MOVL R10, (AX) + JMP memmove_end_copy_match_emit_repeat_encodeBetterBlockAsm + +emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm_memmove_move_4through7: + MOVL (R9), R10 + MOVL -4(R9)(R8*1), R9 + MOVL R10, (AX) + MOVL R9, -4(AX)(R8*1) + JMP memmove_end_copy_match_emit_repeat_encodeBetterBlockAsm + +emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm_memmove_move_8through16: + MOVQ (R9), R10 + MOVQ -8(R9)(R8*1), R9 + MOVQ R10, (AX) + MOVQ R9, -8(AX)(R8*1) + JMP memmove_end_copy_match_emit_repeat_encodeBetterBlockAsm + +emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm_memmove_move_17through32: + MOVOU (R9), X0 + MOVOU -16(R9)(R8*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(R8*1) + JMP memmove_end_copy_match_emit_repeat_encodeBetterBlockAsm + +emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm_memmove_move_33through64: + MOVOU (R9), X0 + MOVOU 16(R9), X1 + MOVOU -32(R9)(R8*1), X2 + MOVOU -16(R9)(R8*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + +memmove_end_copy_match_emit_repeat_encodeBetterBlockAsm: + MOVQ BX, AX + JMP emit_literal_done_match_emit_repeat_encodeBetterBlockAsm + +memmove_long_match_emit_repeat_encodeBetterBlockAsm: + LEAQ (AX)(R8*1), BX + + // genMemMoveLong + MOVOU (R9), X0 + MOVOU 16(R9), X1 + MOVOU -32(R9)(R8*1), X2 + MOVOU -16(R9)(R8*1), X3 + MOVQ R8, R12 + SHRQ $0x05, R12 + MOVQ AX, R10 + ANDL $0x0000001f, R10 + MOVQ $0x00000040, R13 + SUBQ R10, R13 + DECQ R12 + JA emit_lit_memmove_long_match_emit_repeat_encodeBetterBlockAsmlarge_forward_sse_loop_32 + LEAQ -32(R9)(R13*1), R10 + LEAQ -32(AX)(R13*1), R14 + +emit_lit_memmove_long_match_emit_repeat_encodeBetterBlockAsmlarge_big_loop_back: + MOVOU (R10), X4 + MOVOU 16(R10), X5 + MOVOA X4, (R14) + MOVOA X5, 16(R14) + ADDQ $0x20, R14 + ADDQ $0x20, R10 + ADDQ $0x20, R13 + DECQ R12 + JNA emit_lit_memmove_long_match_emit_repeat_encodeBetterBlockAsmlarge_big_loop_back + +emit_lit_memmove_long_match_emit_repeat_encodeBetterBlockAsmlarge_forward_sse_loop_32: + MOVOU -32(R9)(R13*1), X4 + MOVOU -16(R9)(R13*1), X5 + MOVOA X4, -32(AX)(R13*1) + MOVOA X5, -16(AX)(R13*1) + ADDQ $0x20, R13 + CMPQ R8, R13 + JAE emit_lit_memmove_long_match_emit_repeat_encodeBetterBlockAsmlarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + MOVQ BX, AX + +emit_literal_done_match_emit_repeat_encodeBetterBlockAsm: + ADDL R11, CX + ADDL $0x04, R11 + MOVL CX, 12(SP) + + // emitRepeat +emit_repeat_again_match_nolit_repeat_encodeBetterBlockAsm: + MOVL R11, BX + LEAL -4(R11), R11 + CMPL BX, $0x08 + JBE repeat_two_match_nolit_repeat_encodeBetterBlockAsm + CMPL BX, $0x0c + JAE cant_repeat_two_offset_match_nolit_repeat_encodeBetterBlockAsm + CMPL DI, $0x00000800 + JB repeat_two_offset_match_nolit_repeat_encodeBetterBlockAsm + +cant_repeat_two_offset_match_nolit_repeat_encodeBetterBlockAsm: + CMPL R11, $0x00000104 + JB repeat_three_match_nolit_repeat_encodeBetterBlockAsm + CMPL R11, $0x00010100 + JB repeat_four_match_nolit_repeat_encodeBetterBlockAsm + CMPL R11, $0x0100ffff + JB repeat_five_match_nolit_repeat_encodeBetterBlockAsm + LEAL -16842747(R11), R11 + MOVL $0xfffb001d, (AX) + MOVB $0xff, 4(AX) + ADDQ $0x05, AX + JMP emit_repeat_again_match_nolit_repeat_encodeBetterBlockAsm + +repeat_five_match_nolit_repeat_encodeBetterBlockAsm: + LEAL -65536(R11), R11 + MOVL R11, DI + MOVW $0x001d, (AX) + MOVW R11, 2(AX) + SARL $0x10, DI + MOVB DI, 4(AX) + ADDQ $0x05, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm + +repeat_four_match_nolit_repeat_encodeBetterBlockAsm: + LEAL -256(R11), R11 + MOVW $0x0019, (AX) + MOVW R11, 2(AX) + ADDQ $0x04, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm + +repeat_three_match_nolit_repeat_encodeBetterBlockAsm: + LEAL -4(R11), R11 + MOVW $0x0015, (AX) + MOVB R11, 2(AX) + ADDQ $0x03, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm + +repeat_two_match_nolit_repeat_encodeBetterBlockAsm: + SHLL $0x02, R11 + ORL $0x01, R11 + MOVW R11, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm + +repeat_two_offset_match_nolit_repeat_encodeBetterBlockAsm: + XORQ BX, BX + LEAL 1(BX)(R11*4), R11 + MOVB DI, 1(AX) + SARL $0x08, DI + SHLL $0x05, DI + ORL DI, R11 + MOVB R11, (AX) + ADDQ $0x02, AX + +match_nolit_emitcopy_end_encodeBetterBlockAsm: + CMPL CX, 8(SP) + JAE emit_remainder_encodeBetterBlockAsm + CMPQ AX, (SP) + JB match_nolit_dst_ok_encodeBetterBlockAsm + MOVQ $0x00000000, ret+48(FP) + RET + +match_nolit_dst_ok_encodeBetterBlockAsm: + MOVQ $0x00cf1bbcdcbfa563, BX + MOVQ $0x9e3779b1, DI + LEAQ 1(SI), SI + LEAQ -2(CX), R8 + MOVQ (DX)(SI*1), R9 + MOVQ 1(DX)(SI*1), R10 + MOVQ (DX)(R8*1), R11 + MOVQ 1(DX)(R8*1), R12 + SHLQ $0x08, R9 + IMULQ BX, R9 + SHRQ $0x2f, R9 + SHLQ $0x20, R10 + IMULQ DI, R10 + SHRQ $0x32, R10 + SHLQ $0x08, R11 + IMULQ BX, R11 + SHRQ $0x2f, R11 + SHLQ $0x20, R12 + IMULQ DI, R12 + SHRQ $0x32, R12 + LEAQ 1(SI), DI + LEAQ 1(R8), R13 + MOVL SI, 24(SP)(R9*4) + MOVL R8, 24(SP)(R11*4) + MOVL DI, 524312(SP)(R10*4) + MOVL R13, 524312(SP)(R12*4) + LEAQ 1(R8)(SI*1), DI + SHRQ $0x01, DI + ADDQ $0x01, SI + SUBQ $0x01, R8 + +index_loop_encodeBetterBlockAsm: + CMPQ DI, R8 + JAE search_loop_encodeBetterBlockAsm + MOVQ (DX)(SI*1), R9 + MOVQ (DX)(DI*1), R10 + SHLQ $0x08, R9 + IMULQ BX, R9 + SHRQ $0x2f, R9 + SHLQ $0x08, R10 + IMULQ BX, R10 + SHRQ $0x2f, R10 + MOVL SI, 24(SP)(R9*4) + MOVL DI, 24(SP)(R10*4) + ADDQ $0x02, SI + ADDQ $0x02, DI + JMP index_loop_encodeBetterBlockAsm + +emit_remainder_encodeBetterBlockAsm: + MOVQ src_len+32(FP), CX + SUBL 12(SP), CX + LEAQ 5(AX)(CX*1), CX + CMPQ CX, (SP) + JB emit_remainder_ok_encodeBetterBlockAsm + MOVQ $0x00000000, ret+48(FP) + RET + +emit_remainder_ok_encodeBetterBlockAsm: + MOVQ src_len+32(FP), CX + MOVL 12(SP), BX + CMPL BX, CX + JEQ emit_literal_done_emit_remainder_encodeBetterBlockAsm + MOVL CX, SI + MOVL CX, 12(SP) + LEAQ (DX)(BX*1), CX + SUBL BX, SI + LEAL -1(SI), DX + CMPL DX, $0x3c + JB one_byte_emit_remainder_encodeBetterBlockAsm + CMPL DX, $0x00000100 + JB two_bytes_emit_remainder_encodeBetterBlockAsm + CMPL DX, $0x00010000 + JB three_bytes_emit_remainder_encodeBetterBlockAsm + CMPL DX, $0x01000000 + JB four_bytes_emit_remainder_encodeBetterBlockAsm + MOVB $0xfc, (AX) + MOVL DX, 1(AX) + ADDQ $0x05, AX + JMP memmove_long_emit_remainder_encodeBetterBlockAsm + +four_bytes_emit_remainder_encodeBetterBlockAsm: + MOVL DX, BX + SHRL $0x10, BX + MOVB $0xf8, (AX) + MOVW DX, 1(AX) + MOVB BL, 3(AX) + ADDQ $0x04, AX + JMP memmove_long_emit_remainder_encodeBetterBlockAsm + +three_bytes_emit_remainder_encodeBetterBlockAsm: + MOVB $0xf4, (AX) + MOVW DX, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_emit_remainder_encodeBetterBlockAsm + +two_bytes_emit_remainder_encodeBetterBlockAsm: + MOVB $0xf0, (AX) + MOVB DL, 1(AX) + ADDQ $0x02, AX + CMPL DX, $0x40 + JB memmove_emit_remainder_encodeBetterBlockAsm + JMP memmove_long_emit_remainder_encodeBetterBlockAsm + +one_byte_emit_remainder_encodeBetterBlockAsm: + SHLB $0x02, DL + MOVB DL, (AX) + ADDQ $0x01, AX + +memmove_emit_remainder_encodeBetterBlockAsm: + LEAQ (AX)(SI*1), DX + MOVL SI, BX + + // genMemMoveShort + CMPQ BX, $0x03 + JB emit_lit_memmove_emit_remainder_encodeBetterBlockAsm_memmove_move_1or2 + JE emit_lit_memmove_emit_remainder_encodeBetterBlockAsm_memmove_move_3 + CMPQ BX, $0x08 + JB emit_lit_memmove_emit_remainder_encodeBetterBlockAsm_memmove_move_4through7 + CMPQ BX, $0x10 + JBE emit_lit_memmove_emit_remainder_encodeBetterBlockAsm_memmove_move_8through16 + CMPQ BX, $0x20 + JBE emit_lit_memmove_emit_remainder_encodeBetterBlockAsm_memmove_move_17through32 + JMP emit_lit_memmove_emit_remainder_encodeBetterBlockAsm_memmove_move_33through64 + +emit_lit_memmove_emit_remainder_encodeBetterBlockAsm_memmove_move_1or2: + MOVB (CX), SI + MOVB -1(CX)(BX*1), CL + MOVB SI, (AX) + MOVB CL, -1(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeBetterBlockAsm + +emit_lit_memmove_emit_remainder_encodeBetterBlockAsm_memmove_move_3: + MOVW (CX), SI + MOVB 2(CX), CL + MOVW SI, (AX) + MOVB CL, 2(AX) + JMP memmove_end_copy_emit_remainder_encodeBetterBlockAsm + +emit_lit_memmove_emit_remainder_encodeBetterBlockAsm_memmove_move_4through7: + MOVL (CX), SI + MOVL -4(CX)(BX*1), CX + MOVL SI, (AX) + MOVL CX, -4(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeBetterBlockAsm + +emit_lit_memmove_emit_remainder_encodeBetterBlockAsm_memmove_move_8through16: + MOVQ (CX), SI + MOVQ -8(CX)(BX*1), CX + MOVQ SI, (AX) + MOVQ CX, -8(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeBetterBlockAsm + +emit_lit_memmove_emit_remainder_encodeBetterBlockAsm_memmove_move_17through32: + MOVOU (CX), X0 + MOVOU -16(CX)(BX*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeBetterBlockAsm + +emit_lit_memmove_emit_remainder_encodeBetterBlockAsm_memmove_move_33through64: + MOVOU (CX), X0 + MOVOU 16(CX), X1 + MOVOU -32(CX)(BX*1), X2 + MOVOU -16(CX)(BX*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(BX*1) + MOVOU X3, -16(AX)(BX*1) + +memmove_end_copy_emit_remainder_encodeBetterBlockAsm: + MOVQ DX, AX + JMP emit_literal_done_emit_remainder_encodeBetterBlockAsm + +memmove_long_emit_remainder_encodeBetterBlockAsm: + LEAQ (AX)(SI*1), DX + MOVL SI, BX + + // genMemMoveLong + MOVOU (CX), X0 + MOVOU 16(CX), X1 + MOVOU -32(CX)(BX*1), X2 + MOVOU -16(CX)(BX*1), X3 + MOVQ BX, DI + SHRQ $0x05, DI + MOVQ AX, SI + ANDL $0x0000001f, SI + MOVQ $0x00000040, R8 + SUBQ SI, R8 + DECQ DI + JA emit_lit_memmove_long_emit_remainder_encodeBetterBlockAsmlarge_forward_sse_loop_32 + LEAQ -32(CX)(R8*1), SI + LEAQ -32(AX)(R8*1), R9 + +emit_lit_memmove_long_emit_remainder_encodeBetterBlockAsmlarge_big_loop_back: + MOVOU (SI), X4 + MOVOU 16(SI), X5 + MOVOA X4, (R9) + MOVOA X5, 16(R9) + ADDQ $0x20, R9 + ADDQ $0x20, SI + ADDQ $0x20, R8 + DECQ DI + JNA emit_lit_memmove_long_emit_remainder_encodeBetterBlockAsmlarge_big_loop_back + +emit_lit_memmove_long_emit_remainder_encodeBetterBlockAsmlarge_forward_sse_loop_32: + MOVOU -32(CX)(R8*1), X4 + MOVOU -16(CX)(R8*1), X5 + MOVOA X4, -32(AX)(R8*1) + MOVOA X5, -16(AX)(R8*1) + ADDQ $0x20, R8 + CMPQ BX, R8 + JAE emit_lit_memmove_long_emit_remainder_encodeBetterBlockAsmlarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(BX*1) + MOVOU X3, -16(AX)(BX*1) + MOVQ DX, AX + +emit_literal_done_emit_remainder_encodeBetterBlockAsm: + MOVQ dst_base+0(FP), CX + SUBQ CX, AX + MOVQ AX, ret+48(FP) + RET + +// func encodeBetterBlockAsm4MB(dst []byte, src []byte) int +// Requires: BMI, SSE2 +TEXT ·encodeBetterBlockAsm4MB(SB), $589848-56 + MOVQ dst_base+0(FP), AX + MOVQ $0x00001200, CX + LEAQ 24(SP), DX + PXOR X0, X0 + +zero_loop_encodeBetterBlockAsm4MB: + MOVOU X0, (DX) + MOVOU X0, 16(DX) + MOVOU X0, 32(DX) + MOVOU X0, 48(DX) + MOVOU X0, 64(DX) + MOVOU X0, 80(DX) + MOVOU X0, 96(DX) + MOVOU X0, 112(DX) + ADDQ $0x80, DX + DECQ CX + JNZ zero_loop_encodeBetterBlockAsm4MB + MOVL $0x00000000, 12(SP) + MOVQ src_len+32(FP), CX + LEAQ -6(CX), DX + LEAQ -8(CX), BX + MOVL BX, 8(SP) + SHRQ $0x05, CX + SUBL CX, DX + LEAQ (AX)(DX*1), DX + MOVQ DX, (SP) + MOVL $0x00000001, CX + MOVL $0x00000000, 16(SP) + MOVQ src_base+24(FP), DX + +search_loop_encodeBetterBlockAsm4MB: + MOVL CX, BX + SUBL 12(SP), BX + SHRL $0x07, BX + CMPL BX, $0x63 + JBE check_maxskip_ok_encodeBetterBlockAsm4MB + LEAL 100(CX), BX + JMP check_maxskip_cont_encodeBetterBlockAsm4MB + +check_maxskip_ok_encodeBetterBlockAsm4MB: + LEAL 1(CX)(BX*1), BX + +check_maxskip_cont_encodeBetterBlockAsm4MB: + CMPL BX, 8(SP) + JAE emit_remainder_encodeBetterBlockAsm4MB + MOVQ (DX)(CX*1), SI + MOVL BX, 20(SP) + MOVQ $0x00cf1bbcdcbfa563, R8 + MOVQ $0x9e3779b1, BX + MOVQ SI, R9 + MOVQ SI, R10 + SHLQ $0x08, R9 + IMULQ R8, R9 + SHRQ $0x2f, R9 + SHLQ $0x20, R10 + IMULQ BX, R10 + SHRQ $0x32, R10 + MOVL 24(SP)(R9*4), BX + MOVL 524312(SP)(R10*4), DI + MOVL CX, 24(SP)(R9*4) + MOVL CX, 524312(SP)(R10*4) + MOVQ (DX)(BX*1), R9 + MOVQ (DX)(DI*1), R10 + CMPQ R9, SI + JEQ candidate_match_encodeBetterBlockAsm4MB + CMPQ R10, SI + JNE no_short_found_encodeBetterBlockAsm4MB + MOVL DI, BX + JMP candidate_match_encodeBetterBlockAsm4MB + +no_short_found_encodeBetterBlockAsm4MB: + CMPL R9, SI + JEQ candidate_match_encodeBetterBlockAsm4MB + CMPL R10, SI + JEQ candidateS_match_encodeBetterBlockAsm4MB + MOVL 20(SP), CX + JMP search_loop_encodeBetterBlockAsm4MB + +candidateS_match_encodeBetterBlockAsm4MB: + SHRQ $0x08, SI + MOVQ SI, R9 + SHLQ $0x08, R9 + IMULQ R8, R9 + SHRQ $0x2f, R9 + MOVL 24(SP)(R9*4), BX + INCL CX + MOVL CX, 24(SP)(R9*4) + CMPL (DX)(BX*1), SI + JEQ candidate_match_encodeBetterBlockAsm4MB + DECL CX + MOVL DI, BX + +candidate_match_encodeBetterBlockAsm4MB: + MOVL 12(SP), SI + TESTL BX, BX + JZ match_extend_back_end_encodeBetterBlockAsm4MB + +match_extend_back_loop_encodeBetterBlockAsm4MB: + CMPL CX, SI + JBE match_extend_back_end_encodeBetterBlockAsm4MB + MOVB -1(DX)(BX*1), DI + MOVB -1(DX)(CX*1), R8 + CMPB DI, R8 + JNE match_extend_back_end_encodeBetterBlockAsm4MB + LEAL -1(CX), CX + DECL BX + JZ match_extend_back_end_encodeBetterBlockAsm4MB + JMP match_extend_back_loop_encodeBetterBlockAsm4MB + +match_extend_back_end_encodeBetterBlockAsm4MB: + MOVL CX, SI + SUBL 12(SP), SI + LEAQ 4(AX)(SI*1), SI + CMPQ SI, (SP) + JB match_dst_size_check_encodeBetterBlockAsm4MB + MOVQ $0x00000000, ret+48(FP) + RET + +match_dst_size_check_encodeBetterBlockAsm4MB: + MOVL CX, SI + ADDL $0x04, CX + ADDL $0x04, BX + MOVQ src_len+32(FP), DI + SUBL CX, DI + LEAQ (DX)(CX*1), R8 + LEAQ (DX)(BX*1), R9 + + // matchLen + XORL R11, R11 + +matchlen_loopback_16_match_nolit_encodeBetterBlockAsm4MB: + CMPL DI, $0x10 + JB matchlen_match8_match_nolit_encodeBetterBlockAsm4MB + MOVQ (R8)(R11*1), R10 + MOVQ 8(R8)(R11*1), R12 + XORQ (R9)(R11*1), R10 + JNZ matchlen_bsf_8_match_nolit_encodeBetterBlockAsm4MB + XORQ 8(R9)(R11*1), R12 + JNZ matchlen_bsf_16match_nolit_encodeBetterBlockAsm4MB + LEAL -16(DI), DI + LEAL 16(R11), R11 + JMP matchlen_loopback_16_match_nolit_encodeBetterBlockAsm4MB + +matchlen_bsf_16match_nolit_encodeBetterBlockAsm4MB: +#ifdef GOAMD64_v3 + TZCNTQ R12, R12 + +#else + BSFQ R12, R12 + +#endif + SARQ $0x03, R12 + LEAL 8(R11)(R12*1), R11 + JMP match_nolit_end_encodeBetterBlockAsm4MB + +matchlen_match8_match_nolit_encodeBetterBlockAsm4MB: + CMPL DI, $0x08 + JB matchlen_match4_match_nolit_encodeBetterBlockAsm4MB + MOVQ (R8)(R11*1), R10 + XORQ (R9)(R11*1), R10 + JNZ matchlen_bsf_8_match_nolit_encodeBetterBlockAsm4MB + LEAL -8(DI), DI + LEAL 8(R11), R11 + JMP matchlen_match4_match_nolit_encodeBetterBlockAsm4MB + +matchlen_bsf_8_match_nolit_encodeBetterBlockAsm4MB: +#ifdef GOAMD64_v3 + TZCNTQ R10, R10 + +#else + BSFQ R10, R10 + +#endif + SARQ $0x03, R10 + LEAL (R11)(R10*1), R11 + JMP match_nolit_end_encodeBetterBlockAsm4MB + +matchlen_match4_match_nolit_encodeBetterBlockAsm4MB: + CMPL DI, $0x04 + JB matchlen_match2_match_nolit_encodeBetterBlockAsm4MB + MOVL (R8)(R11*1), R10 + CMPL (R9)(R11*1), R10 + JNE matchlen_match2_match_nolit_encodeBetterBlockAsm4MB + LEAL -4(DI), DI + LEAL 4(R11), R11 + +matchlen_match2_match_nolit_encodeBetterBlockAsm4MB: + CMPL DI, $0x01 + JE matchlen_match1_match_nolit_encodeBetterBlockAsm4MB + JB match_nolit_end_encodeBetterBlockAsm4MB + MOVW (R8)(R11*1), R10 + CMPW (R9)(R11*1), R10 + JNE matchlen_match1_match_nolit_encodeBetterBlockAsm4MB + LEAL 2(R11), R11 + SUBL $0x02, DI + JZ match_nolit_end_encodeBetterBlockAsm4MB + +matchlen_match1_match_nolit_encodeBetterBlockAsm4MB: + MOVB (R8)(R11*1), R10 + CMPB (R9)(R11*1), R10 + JNE match_nolit_end_encodeBetterBlockAsm4MB + LEAL 1(R11), R11 + +match_nolit_end_encodeBetterBlockAsm4MB: + MOVL CX, DI + SUBL BX, DI + + // Check if repeat + CMPL 16(SP), DI + JEQ match_is_repeat_encodeBetterBlockAsm4MB + CMPL R11, $0x01 + JA match_length_ok_encodeBetterBlockAsm4MB + CMPL DI, $0x0000ffff + JBE match_length_ok_encodeBetterBlockAsm4MB + MOVL 20(SP), CX + INCL CX + JMP search_loop_encodeBetterBlockAsm4MB + +match_length_ok_encodeBetterBlockAsm4MB: + MOVL DI, 16(SP) + MOVL 12(SP), BX + CMPL BX, SI + JEQ emit_literal_done_match_emit_encodeBetterBlockAsm4MB + MOVL SI, R8 + MOVL SI, 12(SP) + LEAQ (DX)(BX*1), R9 + SUBL BX, R8 + LEAL -1(R8), BX + CMPL BX, $0x3c + JB one_byte_match_emit_encodeBetterBlockAsm4MB + CMPL BX, $0x00000100 + JB two_bytes_match_emit_encodeBetterBlockAsm4MB + CMPL BX, $0x00010000 + JB three_bytes_match_emit_encodeBetterBlockAsm4MB + MOVL BX, R10 + SHRL $0x10, R10 + MOVB $0xf8, (AX) + MOVW BX, 1(AX) + MOVB R10, 3(AX) + ADDQ $0x04, AX + JMP memmove_long_match_emit_encodeBetterBlockAsm4MB + +three_bytes_match_emit_encodeBetterBlockAsm4MB: + MOVB $0xf4, (AX) + MOVW BX, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_match_emit_encodeBetterBlockAsm4MB + +two_bytes_match_emit_encodeBetterBlockAsm4MB: + MOVB $0xf0, (AX) + MOVB BL, 1(AX) + ADDQ $0x02, AX + CMPL BX, $0x40 + JB memmove_match_emit_encodeBetterBlockAsm4MB + JMP memmove_long_match_emit_encodeBetterBlockAsm4MB + +one_byte_match_emit_encodeBetterBlockAsm4MB: + SHLB $0x02, BL + MOVB BL, (AX) + ADDQ $0x01, AX + +memmove_match_emit_encodeBetterBlockAsm4MB: + LEAQ (AX)(R8*1), BX + + // genMemMoveShort + CMPQ R8, $0x04 + JBE emit_lit_memmove_match_emit_encodeBetterBlockAsm4MB_memmove_move_4 + CMPQ R8, $0x08 + JB emit_lit_memmove_match_emit_encodeBetterBlockAsm4MB_memmove_move_4through7 + CMPQ R8, $0x10 + JBE emit_lit_memmove_match_emit_encodeBetterBlockAsm4MB_memmove_move_8through16 + CMPQ R8, $0x20 + JBE emit_lit_memmove_match_emit_encodeBetterBlockAsm4MB_memmove_move_17through32 + JMP emit_lit_memmove_match_emit_encodeBetterBlockAsm4MB_memmove_move_33through64 + +emit_lit_memmove_match_emit_encodeBetterBlockAsm4MB_memmove_move_4: + MOVL (R9), R10 + MOVL R10, (AX) + JMP memmove_end_copy_match_emit_encodeBetterBlockAsm4MB + +emit_lit_memmove_match_emit_encodeBetterBlockAsm4MB_memmove_move_4through7: + MOVL (R9), R10 + MOVL -4(R9)(R8*1), R9 + MOVL R10, (AX) + MOVL R9, -4(AX)(R8*1) + JMP memmove_end_copy_match_emit_encodeBetterBlockAsm4MB + +emit_lit_memmove_match_emit_encodeBetterBlockAsm4MB_memmove_move_8through16: + MOVQ (R9), R10 + MOVQ -8(R9)(R8*1), R9 + MOVQ R10, (AX) + MOVQ R9, -8(AX)(R8*1) + JMP memmove_end_copy_match_emit_encodeBetterBlockAsm4MB + +emit_lit_memmove_match_emit_encodeBetterBlockAsm4MB_memmove_move_17through32: + MOVOU (R9), X0 + MOVOU -16(R9)(R8*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(R8*1) + JMP memmove_end_copy_match_emit_encodeBetterBlockAsm4MB + +emit_lit_memmove_match_emit_encodeBetterBlockAsm4MB_memmove_move_33through64: + MOVOU (R9), X0 + MOVOU 16(R9), X1 + MOVOU -32(R9)(R8*1), X2 + MOVOU -16(R9)(R8*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + +memmove_end_copy_match_emit_encodeBetterBlockAsm4MB: + MOVQ BX, AX + JMP emit_literal_done_match_emit_encodeBetterBlockAsm4MB + +memmove_long_match_emit_encodeBetterBlockAsm4MB: + LEAQ (AX)(R8*1), BX + + // genMemMoveLong + MOVOU (R9), X0 + MOVOU 16(R9), X1 + MOVOU -32(R9)(R8*1), X2 + MOVOU -16(R9)(R8*1), X3 + MOVQ R8, R12 + SHRQ $0x05, R12 + MOVQ AX, R10 + ANDL $0x0000001f, R10 + MOVQ $0x00000040, R13 + SUBQ R10, R13 + DECQ R12 + JA emit_lit_memmove_long_match_emit_encodeBetterBlockAsm4MBlarge_forward_sse_loop_32 + LEAQ -32(R9)(R13*1), R10 + LEAQ -32(AX)(R13*1), R14 + +emit_lit_memmove_long_match_emit_encodeBetterBlockAsm4MBlarge_big_loop_back: + MOVOU (R10), X4 + MOVOU 16(R10), X5 + MOVOA X4, (R14) + MOVOA X5, 16(R14) + ADDQ $0x20, R14 + ADDQ $0x20, R10 + ADDQ $0x20, R13 + DECQ R12 + JNA emit_lit_memmove_long_match_emit_encodeBetterBlockAsm4MBlarge_big_loop_back + +emit_lit_memmove_long_match_emit_encodeBetterBlockAsm4MBlarge_forward_sse_loop_32: + MOVOU -32(R9)(R13*1), X4 + MOVOU -16(R9)(R13*1), X5 + MOVOA X4, -32(AX)(R13*1) + MOVOA X5, -16(AX)(R13*1) + ADDQ $0x20, R13 + CMPQ R8, R13 + JAE emit_lit_memmove_long_match_emit_encodeBetterBlockAsm4MBlarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + MOVQ BX, AX + +emit_literal_done_match_emit_encodeBetterBlockAsm4MB: + ADDL R11, CX + ADDL $0x04, R11 + MOVL CX, 12(SP) + + // emitCopy + CMPL DI, $0x00010000 + JB two_byte_offset_match_nolit_encodeBetterBlockAsm4MB + CMPL R11, $0x40 + JBE four_bytes_remain_match_nolit_encodeBetterBlockAsm4MB + MOVB $0xff, (AX) + MOVL DI, 1(AX) + LEAL -64(R11), R11 + ADDQ $0x05, AX + CMPL R11, $0x04 + JB four_bytes_remain_match_nolit_encodeBetterBlockAsm4MB + + // emitRepeat + MOVL R11, BX + LEAL -4(R11), R11 + CMPL BX, $0x08 + JBE repeat_two_match_nolit_encodeBetterBlockAsm4MB_emit_copy + CMPL BX, $0x0c + JAE cant_repeat_two_offset_match_nolit_encodeBetterBlockAsm4MB_emit_copy + CMPL DI, $0x00000800 + JB repeat_two_offset_match_nolit_encodeBetterBlockAsm4MB_emit_copy + +cant_repeat_two_offset_match_nolit_encodeBetterBlockAsm4MB_emit_copy: + CMPL R11, $0x00000104 + JB repeat_three_match_nolit_encodeBetterBlockAsm4MB_emit_copy + CMPL R11, $0x00010100 + JB repeat_four_match_nolit_encodeBetterBlockAsm4MB_emit_copy + LEAL -65536(R11), R11 + MOVL R11, DI + MOVW $0x001d, (AX) + MOVW R11, 2(AX) + SARL $0x10, DI + MOVB DI, 4(AX) + ADDQ $0x05, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm4MB + +repeat_four_match_nolit_encodeBetterBlockAsm4MB_emit_copy: + LEAL -256(R11), R11 + MOVW $0x0019, (AX) + MOVW R11, 2(AX) + ADDQ $0x04, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm4MB + +repeat_three_match_nolit_encodeBetterBlockAsm4MB_emit_copy: + LEAL -4(R11), R11 + MOVW $0x0015, (AX) + MOVB R11, 2(AX) + ADDQ $0x03, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm4MB + +repeat_two_match_nolit_encodeBetterBlockAsm4MB_emit_copy: + SHLL $0x02, R11 + ORL $0x01, R11 + MOVW R11, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm4MB + +repeat_two_offset_match_nolit_encodeBetterBlockAsm4MB_emit_copy: + XORQ BX, BX + LEAL 1(BX)(R11*4), R11 + MOVB DI, 1(AX) + SARL $0x08, DI + SHLL $0x05, DI + ORL DI, R11 + MOVB R11, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm4MB + +four_bytes_remain_match_nolit_encodeBetterBlockAsm4MB: + TESTL R11, R11 + JZ match_nolit_emitcopy_end_encodeBetterBlockAsm4MB + XORL BX, BX + LEAL -1(BX)(R11*4), R11 + MOVB R11, (AX) + MOVL DI, 1(AX) + ADDQ $0x05, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm4MB + +two_byte_offset_match_nolit_encodeBetterBlockAsm4MB: + CMPL R11, $0x40 + JBE two_byte_offset_short_match_nolit_encodeBetterBlockAsm4MB + CMPL DI, $0x00000800 + JAE long_offset_short_match_nolit_encodeBetterBlockAsm4MB + MOVL $0x00000001, BX + LEAL 16(BX), BX + MOVB DI, 1(AX) + SHRL $0x08, DI + SHLL $0x05, DI + ORL DI, BX + MOVB BL, (AX) + ADDQ $0x02, AX + SUBL $0x08, R11 + + // emitRepeat + LEAL -4(R11), R11 + JMP cant_repeat_two_offset_match_nolit_encodeBetterBlockAsm4MB_emit_copy_short_2b + MOVL R11, BX + LEAL -4(R11), R11 + CMPL BX, $0x08 + JBE repeat_two_match_nolit_encodeBetterBlockAsm4MB_emit_copy_short_2b + CMPL BX, $0x0c + JAE cant_repeat_two_offset_match_nolit_encodeBetterBlockAsm4MB_emit_copy_short_2b + CMPL DI, $0x00000800 + JB repeat_two_offset_match_nolit_encodeBetterBlockAsm4MB_emit_copy_short_2b + +cant_repeat_two_offset_match_nolit_encodeBetterBlockAsm4MB_emit_copy_short_2b: + CMPL R11, $0x00000104 + JB repeat_three_match_nolit_encodeBetterBlockAsm4MB_emit_copy_short_2b + CMPL R11, $0x00010100 + JB repeat_four_match_nolit_encodeBetterBlockAsm4MB_emit_copy_short_2b + LEAL -65536(R11), R11 + MOVL R11, DI + MOVW $0x001d, (AX) + MOVW R11, 2(AX) + SARL $0x10, DI + MOVB DI, 4(AX) + ADDQ $0x05, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm4MB + +repeat_four_match_nolit_encodeBetterBlockAsm4MB_emit_copy_short_2b: + LEAL -256(R11), R11 + MOVW $0x0019, (AX) + MOVW R11, 2(AX) + ADDQ $0x04, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm4MB + +repeat_three_match_nolit_encodeBetterBlockAsm4MB_emit_copy_short_2b: + LEAL -4(R11), R11 + MOVW $0x0015, (AX) + MOVB R11, 2(AX) + ADDQ $0x03, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm4MB + +repeat_two_match_nolit_encodeBetterBlockAsm4MB_emit_copy_short_2b: + SHLL $0x02, R11 + ORL $0x01, R11 + MOVW R11, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm4MB + +repeat_two_offset_match_nolit_encodeBetterBlockAsm4MB_emit_copy_short_2b: + XORQ BX, BX + LEAL 1(BX)(R11*4), R11 + MOVB DI, 1(AX) + SARL $0x08, DI + SHLL $0x05, DI + ORL DI, R11 + MOVB R11, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm4MB + +long_offset_short_match_nolit_encodeBetterBlockAsm4MB: + MOVB $0xee, (AX) + MOVW DI, 1(AX) + LEAL -60(R11), R11 + ADDQ $0x03, AX + + // emitRepeat + MOVL R11, BX + LEAL -4(R11), R11 + CMPL BX, $0x08 + JBE repeat_two_match_nolit_encodeBetterBlockAsm4MB_emit_copy_short + CMPL BX, $0x0c + JAE cant_repeat_two_offset_match_nolit_encodeBetterBlockAsm4MB_emit_copy_short + CMPL DI, $0x00000800 + JB repeat_two_offset_match_nolit_encodeBetterBlockAsm4MB_emit_copy_short + +cant_repeat_two_offset_match_nolit_encodeBetterBlockAsm4MB_emit_copy_short: + CMPL R11, $0x00000104 + JB repeat_three_match_nolit_encodeBetterBlockAsm4MB_emit_copy_short + CMPL R11, $0x00010100 + JB repeat_four_match_nolit_encodeBetterBlockAsm4MB_emit_copy_short + LEAL -65536(R11), R11 + MOVL R11, DI + MOVW $0x001d, (AX) + MOVW R11, 2(AX) + SARL $0x10, DI + MOVB DI, 4(AX) + ADDQ $0x05, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm4MB + +repeat_four_match_nolit_encodeBetterBlockAsm4MB_emit_copy_short: + LEAL -256(R11), R11 + MOVW $0x0019, (AX) + MOVW R11, 2(AX) + ADDQ $0x04, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm4MB + +repeat_three_match_nolit_encodeBetterBlockAsm4MB_emit_copy_short: + LEAL -4(R11), R11 + MOVW $0x0015, (AX) + MOVB R11, 2(AX) + ADDQ $0x03, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm4MB + +repeat_two_match_nolit_encodeBetterBlockAsm4MB_emit_copy_short: + SHLL $0x02, R11 + ORL $0x01, R11 + MOVW R11, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm4MB + +repeat_two_offset_match_nolit_encodeBetterBlockAsm4MB_emit_copy_short: + XORQ BX, BX + LEAL 1(BX)(R11*4), R11 + MOVB DI, 1(AX) + SARL $0x08, DI + SHLL $0x05, DI + ORL DI, R11 + MOVB R11, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm4MB + +two_byte_offset_short_match_nolit_encodeBetterBlockAsm4MB: + MOVL R11, BX + SHLL $0x02, BX + CMPL R11, $0x0c + JAE emit_copy_three_match_nolit_encodeBetterBlockAsm4MB + CMPL DI, $0x00000800 + JAE emit_copy_three_match_nolit_encodeBetterBlockAsm4MB + LEAL -15(BX), BX + MOVB DI, 1(AX) + SHRL $0x08, DI + SHLL $0x05, DI + ORL DI, BX + MOVB BL, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm4MB + +emit_copy_three_match_nolit_encodeBetterBlockAsm4MB: + LEAL -2(BX), BX + MOVB BL, (AX) + MOVW DI, 1(AX) + ADDQ $0x03, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm4MB + +match_is_repeat_encodeBetterBlockAsm4MB: + MOVL 12(SP), BX + CMPL BX, SI + JEQ emit_literal_done_match_emit_repeat_encodeBetterBlockAsm4MB + MOVL SI, R8 + MOVL SI, 12(SP) + LEAQ (DX)(BX*1), R9 + SUBL BX, R8 + LEAL -1(R8), BX + CMPL BX, $0x3c + JB one_byte_match_emit_repeat_encodeBetterBlockAsm4MB + CMPL BX, $0x00000100 + JB two_bytes_match_emit_repeat_encodeBetterBlockAsm4MB + CMPL BX, $0x00010000 + JB three_bytes_match_emit_repeat_encodeBetterBlockAsm4MB + MOVL BX, R10 + SHRL $0x10, R10 + MOVB $0xf8, (AX) + MOVW BX, 1(AX) + MOVB R10, 3(AX) + ADDQ $0x04, AX + JMP memmove_long_match_emit_repeat_encodeBetterBlockAsm4MB + +three_bytes_match_emit_repeat_encodeBetterBlockAsm4MB: + MOVB $0xf4, (AX) + MOVW BX, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_match_emit_repeat_encodeBetterBlockAsm4MB + +two_bytes_match_emit_repeat_encodeBetterBlockAsm4MB: + MOVB $0xf0, (AX) + MOVB BL, 1(AX) + ADDQ $0x02, AX + CMPL BX, $0x40 + JB memmove_match_emit_repeat_encodeBetterBlockAsm4MB + JMP memmove_long_match_emit_repeat_encodeBetterBlockAsm4MB + +one_byte_match_emit_repeat_encodeBetterBlockAsm4MB: + SHLB $0x02, BL + MOVB BL, (AX) + ADDQ $0x01, AX + +memmove_match_emit_repeat_encodeBetterBlockAsm4MB: + LEAQ (AX)(R8*1), BX + + // genMemMoveShort + CMPQ R8, $0x04 + JBE emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm4MB_memmove_move_4 + CMPQ R8, $0x08 + JB emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm4MB_memmove_move_4through7 + CMPQ R8, $0x10 + JBE emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm4MB_memmove_move_8through16 + CMPQ R8, $0x20 + JBE emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm4MB_memmove_move_17through32 + JMP emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm4MB_memmove_move_33through64 + +emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm4MB_memmove_move_4: + MOVL (R9), R10 + MOVL R10, (AX) + JMP memmove_end_copy_match_emit_repeat_encodeBetterBlockAsm4MB + +emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm4MB_memmove_move_4through7: + MOVL (R9), R10 + MOVL -4(R9)(R8*1), R9 + MOVL R10, (AX) + MOVL R9, -4(AX)(R8*1) + JMP memmove_end_copy_match_emit_repeat_encodeBetterBlockAsm4MB + +emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm4MB_memmove_move_8through16: + MOVQ (R9), R10 + MOVQ -8(R9)(R8*1), R9 + MOVQ R10, (AX) + MOVQ R9, -8(AX)(R8*1) + JMP memmove_end_copy_match_emit_repeat_encodeBetterBlockAsm4MB + +emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm4MB_memmove_move_17through32: + MOVOU (R9), X0 + MOVOU -16(R9)(R8*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(R8*1) + JMP memmove_end_copy_match_emit_repeat_encodeBetterBlockAsm4MB + +emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm4MB_memmove_move_33through64: + MOVOU (R9), X0 + MOVOU 16(R9), X1 + MOVOU -32(R9)(R8*1), X2 + MOVOU -16(R9)(R8*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + +memmove_end_copy_match_emit_repeat_encodeBetterBlockAsm4MB: + MOVQ BX, AX + JMP emit_literal_done_match_emit_repeat_encodeBetterBlockAsm4MB + +memmove_long_match_emit_repeat_encodeBetterBlockAsm4MB: + LEAQ (AX)(R8*1), BX + + // genMemMoveLong + MOVOU (R9), X0 + MOVOU 16(R9), X1 + MOVOU -32(R9)(R8*1), X2 + MOVOU -16(R9)(R8*1), X3 + MOVQ R8, R12 + SHRQ $0x05, R12 + MOVQ AX, R10 + ANDL $0x0000001f, R10 + MOVQ $0x00000040, R13 + SUBQ R10, R13 + DECQ R12 + JA emit_lit_memmove_long_match_emit_repeat_encodeBetterBlockAsm4MBlarge_forward_sse_loop_32 + LEAQ -32(R9)(R13*1), R10 + LEAQ -32(AX)(R13*1), R14 + +emit_lit_memmove_long_match_emit_repeat_encodeBetterBlockAsm4MBlarge_big_loop_back: + MOVOU (R10), X4 + MOVOU 16(R10), X5 + MOVOA X4, (R14) + MOVOA X5, 16(R14) + ADDQ $0x20, R14 + ADDQ $0x20, R10 + ADDQ $0x20, R13 + DECQ R12 + JNA emit_lit_memmove_long_match_emit_repeat_encodeBetterBlockAsm4MBlarge_big_loop_back + +emit_lit_memmove_long_match_emit_repeat_encodeBetterBlockAsm4MBlarge_forward_sse_loop_32: + MOVOU -32(R9)(R13*1), X4 + MOVOU -16(R9)(R13*1), X5 + MOVOA X4, -32(AX)(R13*1) + MOVOA X5, -16(AX)(R13*1) + ADDQ $0x20, R13 + CMPQ R8, R13 + JAE emit_lit_memmove_long_match_emit_repeat_encodeBetterBlockAsm4MBlarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + MOVQ BX, AX + +emit_literal_done_match_emit_repeat_encodeBetterBlockAsm4MB: + ADDL R11, CX + ADDL $0x04, R11 + MOVL CX, 12(SP) + + // emitRepeat + MOVL R11, BX + LEAL -4(R11), R11 + CMPL BX, $0x08 + JBE repeat_two_match_nolit_repeat_encodeBetterBlockAsm4MB + CMPL BX, $0x0c + JAE cant_repeat_two_offset_match_nolit_repeat_encodeBetterBlockAsm4MB + CMPL DI, $0x00000800 + JB repeat_two_offset_match_nolit_repeat_encodeBetterBlockAsm4MB + +cant_repeat_two_offset_match_nolit_repeat_encodeBetterBlockAsm4MB: + CMPL R11, $0x00000104 + JB repeat_three_match_nolit_repeat_encodeBetterBlockAsm4MB + CMPL R11, $0x00010100 + JB repeat_four_match_nolit_repeat_encodeBetterBlockAsm4MB + LEAL -65536(R11), R11 + MOVL R11, DI + MOVW $0x001d, (AX) + MOVW R11, 2(AX) + SARL $0x10, DI + MOVB DI, 4(AX) + ADDQ $0x05, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm4MB + +repeat_four_match_nolit_repeat_encodeBetterBlockAsm4MB: + LEAL -256(R11), R11 + MOVW $0x0019, (AX) + MOVW R11, 2(AX) + ADDQ $0x04, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm4MB + +repeat_three_match_nolit_repeat_encodeBetterBlockAsm4MB: + LEAL -4(R11), R11 + MOVW $0x0015, (AX) + MOVB R11, 2(AX) + ADDQ $0x03, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm4MB + +repeat_two_match_nolit_repeat_encodeBetterBlockAsm4MB: + SHLL $0x02, R11 + ORL $0x01, R11 + MOVW R11, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm4MB + +repeat_two_offset_match_nolit_repeat_encodeBetterBlockAsm4MB: + XORQ BX, BX + LEAL 1(BX)(R11*4), R11 + MOVB DI, 1(AX) + SARL $0x08, DI + SHLL $0x05, DI + ORL DI, R11 + MOVB R11, (AX) + ADDQ $0x02, AX + +match_nolit_emitcopy_end_encodeBetterBlockAsm4MB: + CMPL CX, 8(SP) + JAE emit_remainder_encodeBetterBlockAsm4MB + CMPQ AX, (SP) + JB match_nolit_dst_ok_encodeBetterBlockAsm4MB + MOVQ $0x00000000, ret+48(FP) + RET + +match_nolit_dst_ok_encodeBetterBlockAsm4MB: + MOVQ $0x00cf1bbcdcbfa563, BX + MOVQ $0x9e3779b1, DI + LEAQ 1(SI), SI + LEAQ -2(CX), R8 + MOVQ (DX)(SI*1), R9 + MOVQ 1(DX)(SI*1), R10 + MOVQ (DX)(R8*1), R11 + MOVQ 1(DX)(R8*1), R12 + SHLQ $0x08, R9 + IMULQ BX, R9 + SHRQ $0x2f, R9 + SHLQ $0x20, R10 + IMULQ DI, R10 + SHRQ $0x32, R10 + SHLQ $0x08, R11 + IMULQ BX, R11 + SHRQ $0x2f, R11 + SHLQ $0x20, R12 + IMULQ DI, R12 + SHRQ $0x32, R12 + LEAQ 1(SI), DI + LEAQ 1(R8), R13 + MOVL SI, 24(SP)(R9*4) + MOVL R8, 24(SP)(R11*4) + MOVL DI, 524312(SP)(R10*4) + MOVL R13, 524312(SP)(R12*4) + LEAQ 1(R8)(SI*1), DI + SHRQ $0x01, DI + ADDQ $0x01, SI + SUBQ $0x01, R8 + +index_loop_encodeBetterBlockAsm4MB: + CMPQ DI, R8 + JAE search_loop_encodeBetterBlockAsm4MB + MOVQ (DX)(SI*1), R9 + MOVQ (DX)(DI*1), R10 + SHLQ $0x08, R9 + IMULQ BX, R9 + SHRQ $0x2f, R9 + SHLQ $0x08, R10 + IMULQ BX, R10 + SHRQ $0x2f, R10 + MOVL SI, 24(SP)(R9*4) + MOVL DI, 24(SP)(R10*4) + ADDQ $0x02, SI + ADDQ $0x02, DI + JMP index_loop_encodeBetterBlockAsm4MB + +emit_remainder_encodeBetterBlockAsm4MB: + MOVQ src_len+32(FP), CX + SUBL 12(SP), CX + LEAQ 4(AX)(CX*1), CX + CMPQ CX, (SP) + JB emit_remainder_ok_encodeBetterBlockAsm4MB + MOVQ $0x00000000, ret+48(FP) + RET + +emit_remainder_ok_encodeBetterBlockAsm4MB: + MOVQ src_len+32(FP), CX + MOVL 12(SP), BX + CMPL BX, CX + JEQ emit_literal_done_emit_remainder_encodeBetterBlockAsm4MB + MOVL CX, SI + MOVL CX, 12(SP) + LEAQ (DX)(BX*1), CX + SUBL BX, SI + LEAL -1(SI), DX + CMPL DX, $0x3c + JB one_byte_emit_remainder_encodeBetterBlockAsm4MB + CMPL DX, $0x00000100 + JB two_bytes_emit_remainder_encodeBetterBlockAsm4MB + CMPL DX, $0x00010000 + JB three_bytes_emit_remainder_encodeBetterBlockAsm4MB + MOVL DX, BX + SHRL $0x10, BX + MOVB $0xf8, (AX) + MOVW DX, 1(AX) + MOVB BL, 3(AX) + ADDQ $0x04, AX + JMP memmove_long_emit_remainder_encodeBetterBlockAsm4MB + +three_bytes_emit_remainder_encodeBetterBlockAsm4MB: + MOVB $0xf4, (AX) + MOVW DX, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_emit_remainder_encodeBetterBlockAsm4MB + +two_bytes_emit_remainder_encodeBetterBlockAsm4MB: + MOVB $0xf0, (AX) + MOVB DL, 1(AX) + ADDQ $0x02, AX + CMPL DX, $0x40 + JB memmove_emit_remainder_encodeBetterBlockAsm4MB + JMP memmove_long_emit_remainder_encodeBetterBlockAsm4MB + +one_byte_emit_remainder_encodeBetterBlockAsm4MB: + SHLB $0x02, DL + MOVB DL, (AX) + ADDQ $0x01, AX + +memmove_emit_remainder_encodeBetterBlockAsm4MB: + LEAQ (AX)(SI*1), DX + MOVL SI, BX + + // genMemMoveShort + CMPQ BX, $0x03 + JB emit_lit_memmove_emit_remainder_encodeBetterBlockAsm4MB_memmove_move_1or2 + JE emit_lit_memmove_emit_remainder_encodeBetterBlockAsm4MB_memmove_move_3 + CMPQ BX, $0x08 + JB emit_lit_memmove_emit_remainder_encodeBetterBlockAsm4MB_memmove_move_4through7 + CMPQ BX, $0x10 + JBE emit_lit_memmove_emit_remainder_encodeBetterBlockAsm4MB_memmove_move_8through16 + CMPQ BX, $0x20 + JBE emit_lit_memmove_emit_remainder_encodeBetterBlockAsm4MB_memmove_move_17through32 + JMP emit_lit_memmove_emit_remainder_encodeBetterBlockAsm4MB_memmove_move_33through64 + +emit_lit_memmove_emit_remainder_encodeBetterBlockAsm4MB_memmove_move_1or2: + MOVB (CX), SI + MOVB -1(CX)(BX*1), CL + MOVB SI, (AX) + MOVB CL, -1(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeBetterBlockAsm4MB + +emit_lit_memmove_emit_remainder_encodeBetterBlockAsm4MB_memmove_move_3: + MOVW (CX), SI + MOVB 2(CX), CL + MOVW SI, (AX) + MOVB CL, 2(AX) + JMP memmove_end_copy_emit_remainder_encodeBetterBlockAsm4MB + +emit_lit_memmove_emit_remainder_encodeBetterBlockAsm4MB_memmove_move_4through7: + MOVL (CX), SI + MOVL -4(CX)(BX*1), CX + MOVL SI, (AX) + MOVL CX, -4(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeBetterBlockAsm4MB + +emit_lit_memmove_emit_remainder_encodeBetterBlockAsm4MB_memmove_move_8through16: + MOVQ (CX), SI + MOVQ -8(CX)(BX*1), CX + MOVQ SI, (AX) + MOVQ CX, -8(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeBetterBlockAsm4MB + +emit_lit_memmove_emit_remainder_encodeBetterBlockAsm4MB_memmove_move_17through32: + MOVOU (CX), X0 + MOVOU -16(CX)(BX*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeBetterBlockAsm4MB + +emit_lit_memmove_emit_remainder_encodeBetterBlockAsm4MB_memmove_move_33through64: + MOVOU (CX), X0 + MOVOU 16(CX), X1 + MOVOU -32(CX)(BX*1), X2 + MOVOU -16(CX)(BX*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(BX*1) + MOVOU X3, -16(AX)(BX*1) + +memmove_end_copy_emit_remainder_encodeBetterBlockAsm4MB: + MOVQ DX, AX + JMP emit_literal_done_emit_remainder_encodeBetterBlockAsm4MB + +memmove_long_emit_remainder_encodeBetterBlockAsm4MB: + LEAQ (AX)(SI*1), DX + MOVL SI, BX + + // genMemMoveLong + MOVOU (CX), X0 + MOVOU 16(CX), X1 + MOVOU -32(CX)(BX*1), X2 + MOVOU -16(CX)(BX*1), X3 + MOVQ BX, DI + SHRQ $0x05, DI + MOVQ AX, SI + ANDL $0x0000001f, SI + MOVQ $0x00000040, R8 + SUBQ SI, R8 + DECQ DI + JA emit_lit_memmove_long_emit_remainder_encodeBetterBlockAsm4MBlarge_forward_sse_loop_32 + LEAQ -32(CX)(R8*1), SI + LEAQ -32(AX)(R8*1), R9 + +emit_lit_memmove_long_emit_remainder_encodeBetterBlockAsm4MBlarge_big_loop_back: + MOVOU (SI), X4 + MOVOU 16(SI), X5 + MOVOA X4, (R9) + MOVOA X5, 16(R9) + ADDQ $0x20, R9 + ADDQ $0x20, SI + ADDQ $0x20, R8 + DECQ DI + JNA emit_lit_memmove_long_emit_remainder_encodeBetterBlockAsm4MBlarge_big_loop_back + +emit_lit_memmove_long_emit_remainder_encodeBetterBlockAsm4MBlarge_forward_sse_loop_32: + MOVOU -32(CX)(R8*1), X4 + MOVOU -16(CX)(R8*1), X5 + MOVOA X4, -32(AX)(R8*1) + MOVOA X5, -16(AX)(R8*1) + ADDQ $0x20, R8 + CMPQ BX, R8 + JAE emit_lit_memmove_long_emit_remainder_encodeBetterBlockAsm4MBlarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(BX*1) + MOVOU X3, -16(AX)(BX*1) + MOVQ DX, AX + +emit_literal_done_emit_remainder_encodeBetterBlockAsm4MB: + MOVQ dst_base+0(FP), CX + SUBQ CX, AX + MOVQ AX, ret+48(FP) + RET + +// func encodeBetterBlockAsm12B(dst []byte, src []byte) int +// Requires: BMI, SSE2 +TEXT ·encodeBetterBlockAsm12B(SB), $81944-56 + MOVQ dst_base+0(FP), AX + MOVQ $0x00000280, CX + LEAQ 24(SP), DX + PXOR X0, X0 + +zero_loop_encodeBetterBlockAsm12B: + MOVOU X0, (DX) + MOVOU X0, 16(DX) + MOVOU X0, 32(DX) + MOVOU X0, 48(DX) + MOVOU X0, 64(DX) + MOVOU X0, 80(DX) + MOVOU X0, 96(DX) + MOVOU X0, 112(DX) + ADDQ $0x80, DX + DECQ CX + JNZ zero_loop_encodeBetterBlockAsm12B + MOVL $0x00000000, 12(SP) + MOVQ src_len+32(FP), CX + LEAQ -6(CX), DX + LEAQ -8(CX), BX + MOVL BX, 8(SP) + SHRQ $0x05, CX + SUBL CX, DX + LEAQ (AX)(DX*1), DX + MOVQ DX, (SP) + MOVL $0x00000001, CX + MOVL $0x00000000, 16(SP) + MOVQ src_base+24(FP), DX + +search_loop_encodeBetterBlockAsm12B: + MOVL CX, BX + SUBL 12(SP), BX + SHRL $0x06, BX + LEAL 1(CX)(BX*1), BX + CMPL BX, 8(SP) + JAE emit_remainder_encodeBetterBlockAsm12B + MOVQ (DX)(CX*1), SI + MOVL BX, 20(SP) + MOVQ $0x0000cf1bbcdcbf9b, R8 + MOVQ $0x9e3779b1, BX + MOVQ SI, R9 + MOVQ SI, R10 + SHLQ $0x10, R9 + IMULQ R8, R9 + SHRQ $0x32, R9 + SHLQ $0x20, R10 + IMULQ BX, R10 + SHRQ $0x34, R10 + MOVL 24(SP)(R9*4), BX + MOVL 65560(SP)(R10*4), DI + MOVL CX, 24(SP)(R9*4) + MOVL CX, 65560(SP)(R10*4) + MOVQ (DX)(BX*1), R9 + MOVQ (DX)(DI*1), R10 + CMPQ R9, SI + JEQ candidate_match_encodeBetterBlockAsm12B + CMPQ R10, SI + JNE no_short_found_encodeBetterBlockAsm12B + MOVL DI, BX + JMP candidate_match_encodeBetterBlockAsm12B + +no_short_found_encodeBetterBlockAsm12B: + CMPL R9, SI + JEQ candidate_match_encodeBetterBlockAsm12B + CMPL R10, SI + JEQ candidateS_match_encodeBetterBlockAsm12B + MOVL 20(SP), CX + JMP search_loop_encodeBetterBlockAsm12B + +candidateS_match_encodeBetterBlockAsm12B: + SHRQ $0x08, SI + MOVQ SI, R9 + SHLQ $0x10, R9 + IMULQ R8, R9 + SHRQ $0x32, R9 + MOVL 24(SP)(R9*4), BX + INCL CX + MOVL CX, 24(SP)(R9*4) + CMPL (DX)(BX*1), SI + JEQ candidate_match_encodeBetterBlockAsm12B + DECL CX + MOVL DI, BX + +candidate_match_encodeBetterBlockAsm12B: + MOVL 12(SP), SI + TESTL BX, BX + JZ match_extend_back_end_encodeBetterBlockAsm12B + +match_extend_back_loop_encodeBetterBlockAsm12B: + CMPL CX, SI + JBE match_extend_back_end_encodeBetterBlockAsm12B + MOVB -1(DX)(BX*1), DI + MOVB -1(DX)(CX*1), R8 + CMPB DI, R8 + JNE match_extend_back_end_encodeBetterBlockAsm12B + LEAL -1(CX), CX + DECL BX + JZ match_extend_back_end_encodeBetterBlockAsm12B + JMP match_extend_back_loop_encodeBetterBlockAsm12B + +match_extend_back_end_encodeBetterBlockAsm12B: + MOVL CX, SI + SUBL 12(SP), SI + LEAQ 3(AX)(SI*1), SI + CMPQ SI, (SP) + JB match_dst_size_check_encodeBetterBlockAsm12B + MOVQ $0x00000000, ret+48(FP) + RET + +match_dst_size_check_encodeBetterBlockAsm12B: + MOVL CX, SI + ADDL $0x04, CX + ADDL $0x04, BX + MOVQ src_len+32(FP), DI + SUBL CX, DI + LEAQ (DX)(CX*1), R8 + LEAQ (DX)(BX*1), R9 + + // matchLen + XORL R11, R11 + +matchlen_loopback_16_match_nolit_encodeBetterBlockAsm12B: + CMPL DI, $0x10 + JB matchlen_match8_match_nolit_encodeBetterBlockAsm12B + MOVQ (R8)(R11*1), R10 + MOVQ 8(R8)(R11*1), R12 + XORQ (R9)(R11*1), R10 + JNZ matchlen_bsf_8_match_nolit_encodeBetterBlockAsm12B + XORQ 8(R9)(R11*1), R12 + JNZ matchlen_bsf_16match_nolit_encodeBetterBlockAsm12B + LEAL -16(DI), DI + LEAL 16(R11), R11 + JMP matchlen_loopback_16_match_nolit_encodeBetterBlockAsm12B + +matchlen_bsf_16match_nolit_encodeBetterBlockAsm12B: +#ifdef GOAMD64_v3 + TZCNTQ R12, R12 + +#else + BSFQ R12, R12 + +#endif + SARQ $0x03, R12 + LEAL 8(R11)(R12*1), R11 + JMP match_nolit_end_encodeBetterBlockAsm12B + +matchlen_match8_match_nolit_encodeBetterBlockAsm12B: + CMPL DI, $0x08 + JB matchlen_match4_match_nolit_encodeBetterBlockAsm12B + MOVQ (R8)(R11*1), R10 + XORQ (R9)(R11*1), R10 + JNZ matchlen_bsf_8_match_nolit_encodeBetterBlockAsm12B + LEAL -8(DI), DI + LEAL 8(R11), R11 + JMP matchlen_match4_match_nolit_encodeBetterBlockAsm12B + +matchlen_bsf_8_match_nolit_encodeBetterBlockAsm12B: +#ifdef GOAMD64_v3 + TZCNTQ R10, R10 + +#else + BSFQ R10, R10 + +#endif + SARQ $0x03, R10 + LEAL (R11)(R10*1), R11 + JMP match_nolit_end_encodeBetterBlockAsm12B + +matchlen_match4_match_nolit_encodeBetterBlockAsm12B: + CMPL DI, $0x04 + JB matchlen_match2_match_nolit_encodeBetterBlockAsm12B + MOVL (R8)(R11*1), R10 + CMPL (R9)(R11*1), R10 + JNE matchlen_match2_match_nolit_encodeBetterBlockAsm12B + LEAL -4(DI), DI + LEAL 4(R11), R11 + +matchlen_match2_match_nolit_encodeBetterBlockAsm12B: + CMPL DI, $0x01 + JE matchlen_match1_match_nolit_encodeBetterBlockAsm12B + JB match_nolit_end_encodeBetterBlockAsm12B + MOVW (R8)(R11*1), R10 + CMPW (R9)(R11*1), R10 + JNE matchlen_match1_match_nolit_encodeBetterBlockAsm12B + LEAL 2(R11), R11 + SUBL $0x02, DI + JZ match_nolit_end_encodeBetterBlockAsm12B + +matchlen_match1_match_nolit_encodeBetterBlockAsm12B: + MOVB (R8)(R11*1), R10 + CMPB (R9)(R11*1), R10 + JNE match_nolit_end_encodeBetterBlockAsm12B + LEAL 1(R11), R11 + +match_nolit_end_encodeBetterBlockAsm12B: + MOVL CX, DI + SUBL BX, DI + + // Check if repeat + CMPL 16(SP), DI + JEQ match_is_repeat_encodeBetterBlockAsm12B + MOVL DI, 16(SP) + MOVL 12(SP), BX + CMPL BX, SI + JEQ emit_literal_done_match_emit_encodeBetterBlockAsm12B + MOVL SI, R8 + MOVL SI, 12(SP) + LEAQ (DX)(BX*1), R9 + SUBL BX, R8 + LEAL -1(R8), BX + CMPL BX, $0x3c + JB one_byte_match_emit_encodeBetterBlockAsm12B + CMPL BX, $0x00000100 + JB two_bytes_match_emit_encodeBetterBlockAsm12B + JB three_bytes_match_emit_encodeBetterBlockAsm12B + +three_bytes_match_emit_encodeBetterBlockAsm12B: + MOVB $0xf4, (AX) + MOVW BX, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_match_emit_encodeBetterBlockAsm12B + +two_bytes_match_emit_encodeBetterBlockAsm12B: + MOVB $0xf0, (AX) + MOVB BL, 1(AX) + ADDQ $0x02, AX + CMPL BX, $0x40 + JB memmove_match_emit_encodeBetterBlockAsm12B + JMP memmove_long_match_emit_encodeBetterBlockAsm12B + +one_byte_match_emit_encodeBetterBlockAsm12B: + SHLB $0x02, BL + MOVB BL, (AX) + ADDQ $0x01, AX + +memmove_match_emit_encodeBetterBlockAsm12B: + LEAQ (AX)(R8*1), BX + + // genMemMoveShort + CMPQ R8, $0x04 + JBE emit_lit_memmove_match_emit_encodeBetterBlockAsm12B_memmove_move_4 + CMPQ R8, $0x08 + JB emit_lit_memmove_match_emit_encodeBetterBlockAsm12B_memmove_move_4through7 + CMPQ R8, $0x10 + JBE emit_lit_memmove_match_emit_encodeBetterBlockAsm12B_memmove_move_8through16 + CMPQ R8, $0x20 + JBE emit_lit_memmove_match_emit_encodeBetterBlockAsm12B_memmove_move_17through32 + JMP emit_lit_memmove_match_emit_encodeBetterBlockAsm12B_memmove_move_33through64 + +emit_lit_memmove_match_emit_encodeBetterBlockAsm12B_memmove_move_4: + MOVL (R9), R10 + MOVL R10, (AX) + JMP memmove_end_copy_match_emit_encodeBetterBlockAsm12B + +emit_lit_memmove_match_emit_encodeBetterBlockAsm12B_memmove_move_4through7: + MOVL (R9), R10 + MOVL -4(R9)(R8*1), R9 + MOVL R10, (AX) + MOVL R9, -4(AX)(R8*1) + JMP memmove_end_copy_match_emit_encodeBetterBlockAsm12B + +emit_lit_memmove_match_emit_encodeBetterBlockAsm12B_memmove_move_8through16: + MOVQ (R9), R10 + MOVQ -8(R9)(R8*1), R9 + MOVQ R10, (AX) + MOVQ R9, -8(AX)(R8*1) + JMP memmove_end_copy_match_emit_encodeBetterBlockAsm12B + +emit_lit_memmove_match_emit_encodeBetterBlockAsm12B_memmove_move_17through32: + MOVOU (R9), X0 + MOVOU -16(R9)(R8*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(R8*1) + JMP memmove_end_copy_match_emit_encodeBetterBlockAsm12B + +emit_lit_memmove_match_emit_encodeBetterBlockAsm12B_memmove_move_33through64: + MOVOU (R9), X0 + MOVOU 16(R9), X1 + MOVOU -32(R9)(R8*1), X2 + MOVOU -16(R9)(R8*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + +memmove_end_copy_match_emit_encodeBetterBlockAsm12B: + MOVQ BX, AX + JMP emit_literal_done_match_emit_encodeBetterBlockAsm12B + +memmove_long_match_emit_encodeBetterBlockAsm12B: + LEAQ (AX)(R8*1), BX + + // genMemMoveLong + MOVOU (R9), X0 + MOVOU 16(R9), X1 + MOVOU -32(R9)(R8*1), X2 + MOVOU -16(R9)(R8*1), X3 + MOVQ R8, R12 + SHRQ $0x05, R12 + MOVQ AX, R10 + ANDL $0x0000001f, R10 + MOVQ $0x00000040, R13 + SUBQ R10, R13 + DECQ R12 + JA emit_lit_memmove_long_match_emit_encodeBetterBlockAsm12Blarge_forward_sse_loop_32 + LEAQ -32(R9)(R13*1), R10 + LEAQ -32(AX)(R13*1), R14 + +emit_lit_memmove_long_match_emit_encodeBetterBlockAsm12Blarge_big_loop_back: + MOVOU (R10), X4 + MOVOU 16(R10), X5 + MOVOA X4, (R14) + MOVOA X5, 16(R14) + ADDQ $0x20, R14 + ADDQ $0x20, R10 + ADDQ $0x20, R13 + DECQ R12 + JNA emit_lit_memmove_long_match_emit_encodeBetterBlockAsm12Blarge_big_loop_back + +emit_lit_memmove_long_match_emit_encodeBetterBlockAsm12Blarge_forward_sse_loop_32: + MOVOU -32(R9)(R13*1), X4 + MOVOU -16(R9)(R13*1), X5 + MOVOA X4, -32(AX)(R13*1) + MOVOA X5, -16(AX)(R13*1) + ADDQ $0x20, R13 + CMPQ R8, R13 + JAE emit_lit_memmove_long_match_emit_encodeBetterBlockAsm12Blarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + MOVQ BX, AX + +emit_literal_done_match_emit_encodeBetterBlockAsm12B: + ADDL R11, CX + ADDL $0x04, R11 + MOVL CX, 12(SP) + + // emitCopy + CMPL R11, $0x40 + JBE two_byte_offset_short_match_nolit_encodeBetterBlockAsm12B + CMPL DI, $0x00000800 + JAE long_offset_short_match_nolit_encodeBetterBlockAsm12B + MOVL $0x00000001, BX + LEAL 16(BX), BX + MOVB DI, 1(AX) + SHRL $0x08, DI + SHLL $0x05, DI + ORL DI, BX + MOVB BL, (AX) + ADDQ $0x02, AX + SUBL $0x08, R11 + + // emitRepeat + LEAL -4(R11), R11 + JMP cant_repeat_two_offset_match_nolit_encodeBetterBlockAsm12B_emit_copy_short_2b + MOVL R11, BX + LEAL -4(R11), R11 + CMPL BX, $0x08 + JBE repeat_two_match_nolit_encodeBetterBlockAsm12B_emit_copy_short_2b + CMPL BX, $0x0c + JAE cant_repeat_two_offset_match_nolit_encodeBetterBlockAsm12B_emit_copy_short_2b + CMPL DI, $0x00000800 + JB repeat_two_offset_match_nolit_encodeBetterBlockAsm12B_emit_copy_short_2b + +cant_repeat_two_offset_match_nolit_encodeBetterBlockAsm12B_emit_copy_short_2b: + CMPL R11, $0x00000104 + JB repeat_three_match_nolit_encodeBetterBlockAsm12B_emit_copy_short_2b + LEAL -256(R11), R11 + MOVW $0x0019, (AX) + MOVW R11, 2(AX) + ADDQ $0x04, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm12B + +repeat_three_match_nolit_encodeBetterBlockAsm12B_emit_copy_short_2b: + LEAL -4(R11), R11 + MOVW $0x0015, (AX) + MOVB R11, 2(AX) + ADDQ $0x03, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm12B + +repeat_two_match_nolit_encodeBetterBlockAsm12B_emit_copy_short_2b: + SHLL $0x02, R11 + ORL $0x01, R11 + MOVW R11, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm12B + +repeat_two_offset_match_nolit_encodeBetterBlockAsm12B_emit_copy_short_2b: + XORQ BX, BX + LEAL 1(BX)(R11*4), R11 + MOVB DI, 1(AX) + SARL $0x08, DI + SHLL $0x05, DI + ORL DI, R11 + MOVB R11, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm12B + +long_offset_short_match_nolit_encodeBetterBlockAsm12B: + MOVB $0xee, (AX) + MOVW DI, 1(AX) + LEAL -60(R11), R11 + ADDQ $0x03, AX + + // emitRepeat + MOVL R11, BX + LEAL -4(R11), R11 + CMPL BX, $0x08 + JBE repeat_two_match_nolit_encodeBetterBlockAsm12B_emit_copy_short + CMPL BX, $0x0c + JAE cant_repeat_two_offset_match_nolit_encodeBetterBlockAsm12B_emit_copy_short + CMPL DI, $0x00000800 + JB repeat_two_offset_match_nolit_encodeBetterBlockAsm12B_emit_copy_short + +cant_repeat_two_offset_match_nolit_encodeBetterBlockAsm12B_emit_copy_short: + CMPL R11, $0x00000104 + JB repeat_three_match_nolit_encodeBetterBlockAsm12B_emit_copy_short + LEAL -256(R11), R11 + MOVW $0x0019, (AX) + MOVW R11, 2(AX) + ADDQ $0x04, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm12B + +repeat_three_match_nolit_encodeBetterBlockAsm12B_emit_copy_short: + LEAL -4(R11), R11 + MOVW $0x0015, (AX) + MOVB R11, 2(AX) + ADDQ $0x03, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm12B + +repeat_two_match_nolit_encodeBetterBlockAsm12B_emit_copy_short: + SHLL $0x02, R11 + ORL $0x01, R11 + MOVW R11, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm12B + +repeat_two_offset_match_nolit_encodeBetterBlockAsm12B_emit_copy_short: + XORQ BX, BX + LEAL 1(BX)(R11*4), R11 + MOVB DI, 1(AX) + SARL $0x08, DI + SHLL $0x05, DI + ORL DI, R11 + MOVB R11, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm12B + +two_byte_offset_short_match_nolit_encodeBetterBlockAsm12B: + MOVL R11, BX + SHLL $0x02, BX + CMPL R11, $0x0c + JAE emit_copy_three_match_nolit_encodeBetterBlockAsm12B + CMPL DI, $0x00000800 + JAE emit_copy_three_match_nolit_encodeBetterBlockAsm12B + LEAL -15(BX), BX + MOVB DI, 1(AX) + SHRL $0x08, DI + SHLL $0x05, DI + ORL DI, BX + MOVB BL, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm12B + +emit_copy_three_match_nolit_encodeBetterBlockAsm12B: + LEAL -2(BX), BX + MOVB BL, (AX) + MOVW DI, 1(AX) + ADDQ $0x03, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm12B + +match_is_repeat_encodeBetterBlockAsm12B: + MOVL 12(SP), BX + CMPL BX, SI + JEQ emit_literal_done_match_emit_repeat_encodeBetterBlockAsm12B + MOVL SI, R8 + MOVL SI, 12(SP) + LEAQ (DX)(BX*1), R9 + SUBL BX, R8 + LEAL -1(R8), BX + CMPL BX, $0x3c + JB one_byte_match_emit_repeat_encodeBetterBlockAsm12B + CMPL BX, $0x00000100 + JB two_bytes_match_emit_repeat_encodeBetterBlockAsm12B + JB three_bytes_match_emit_repeat_encodeBetterBlockAsm12B + +three_bytes_match_emit_repeat_encodeBetterBlockAsm12B: + MOVB $0xf4, (AX) + MOVW BX, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_match_emit_repeat_encodeBetterBlockAsm12B + +two_bytes_match_emit_repeat_encodeBetterBlockAsm12B: + MOVB $0xf0, (AX) + MOVB BL, 1(AX) + ADDQ $0x02, AX + CMPL BX, $0x40 + JB memmove_match_emit_repeat_encodeBetterBlockAsm12B + JMP memmove_long_match_emit_repeat_encodeBetterBlockAsm12B + +one_byte_match_emit_repeat_encodeBetterBlockAsm12B: + SHLB $0x02, BL + MOVB BL, (AX) + ADDQ $0x01, AX + +memmove_match_emit_repeat_encodeBetterBlockAsm12B: + LEAQ (AX)(R8*1), BX + + // genMemMoveShort + CMPQ R8, $0x04 + JBE emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm12B_memmove_move_4 + CMPQ R8, $0x08 + JB emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm12B_memmove_move_4through7 + CMPQ R8, $0x10 + JBE emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm12B_memmove_move_8through16 + CMPQ R8, $0x20 + JBE emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm12B_memmove_move_17through32 + JMP emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm12B_memmove_move_33through64 + +emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm12B_memmove_move_4: + MOVL (R9), R10 + MOVL R10, (AX) + JMP memmove_end_copy_match_emit_repeat_encodeBetterBlockAsm12B + +emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm12B_memmove_move_4through7: + MOVL (R9), R10 + MOVL -4(R9)(R8*1), R9 + MOVL R10, (AX) + MOVL R9, -4(AX)(R8*1) + JMP memmove_end_copy_match_emit_repeat_encodeBetterBlockAsm12B + +emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm12B_memmove_move_8through16: + MOVQ (R9), R10 + MOVQ -8(R9)(R8*1), R9 + MOVQ R10, (AX) + MOVQ R9, -8(AX)(R8*1) + JMP memmove_end_copy_match_emit_repeat_encodeBetterBlockAsm12B + +emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm12B_memmove_move_17through32: + MOVOU (R9), X0 + MOVOU -16(R9)(R8*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(R8*1) + JMP memmove_end_copy_match_emit_repeat_encodeBetterBlockAsm12B + +emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm12B_memmove_move_33through64: + MOVOU (R9), X0 + MOVOU 16(R9), X1 + MOVOU -32(R9)(R8*1), X2 + MOVOU -16(R9)(R8*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + +memmove_end_copy_match_emit_repeat_encodeBetterBlockAsm12B: + MOVQ BX, AX + JMP emit_literal_done_match_emit_repeat_encodeBetterBlockAsm12B + +memmove_long_match_emit_repeat_encodeBetterBlockAsm12B: + LEAQ (AX)(R8*1), BX + + // genMemMoveLong + MOVOU (R9), X0 + MOVOU 16(R9), X1 + MOVOU -32(R9)(R8*1), X2 + MOVOU -16(R9)(R8*1), X3 + MOVQ R8, R12 + SHRQ $0x05, R12 + MOVQ AX, R10 + ANDL $0x0000001f, R10 + MOVQ $0x00000040, R13 + SUBQ R10, R13 + DECQ R12 + JA emit_lit_memmove_long_match_emit_repeat_encodeBetterBlockAsm12Blarge_forward_sse_loop_32 + LEAQ -32(R9)(R13*1), R10 + LEAQ -32(AX)(R13*1), R14 + +emit_lit_memmove_long_match_emit_repeat_encodeBetterBlockAsm12Blarge_big_loop_back: + MOVOU (R10), X4 + MOVOU 16(R10), X5 + MOVOA X4, (R14) + MOVOA X5, 16(R14) + ADDQ $0x20, R14 + ADDQ $0x20, R10 + ADDQ $0x20, R13 + DECQ R12 + JNA emit_lit_memmove_long_match_emit_repeat_encodeBetterBlockAsm12Blarge_big_loop_back + +emit_lit_memmove_long_match_emit_repeat_encodeBetterBlockAsm12Blarge_forward_sse_loop_32: + MOVOU -32(R9)(R13*1), X4 + MOVOU -16(R9)(R13*1), X5 + MOVOA X4, -32(AX)(R13*1) + MOVOA X5, -16(AX)(R13*1) + ADDQ $0x20, R13 + CMPQ R8, R13 + JAE emit_lit_memmove_long_match_emit_repeat_encodeBetterBlockAsm12Blarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + MOVQ BX, AX + +emit_literal_done_match_emit_repeat_encodeBetterBlockAsm12B: + ADDL R11, CX + ADDL $0x04, R11 + MOVL CX, 12(SP) + + // emitRepeat + MOVL R11, BX + LEAL -4(R11), R11 + CMPL BX, $0x08 + JBE repeat_two_match_nolit_repeat_encodeBetterBlockAsm12B + CMPL BX, $0x0c + JAE cant_repeat_two_offset_match_nolit_repeat_encodeBetterBlockAsm12B + CMPL DI, $0x00000800 + JB repeat_two_offset_match_nolit_repeat_encodeBetterBlockAsm12B + +cant_repeat_two_offset_match_nolit_repeat_encodeBetterBlockAsm12B: + CMPL R11, $0x00000104 + JB repeat_three_match_nolit_repeat_encodeBetterBlockAsm12B + LEAL -256(R11), R11 + MOVW $0x0019, (AX) + MOVW R11, 2(AX) + ADDQ $0x04, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm12B + +repeat_three_match_nolit_repeat_encodeBetterBlockAsm12B: + LEAL -4(R11), R11 + MOVW $0x0015, (AX) + MOVB R11, 2(AX) + ADDQ $0x03, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm12B + +repeat_two_match_nolit_repeat_encodeBetterBlockAsm12B: + SHLL $0x02, R11 + ORL $0x01, R11 + MOVW R11, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm12B + +repeat_two_offset_match_nolit_repeat_encodeBetterBlockAsm12B: + XORQ BX, BX + LEAL 1(BX)(R11*4), R11 + MOVB DI, 1(AX) + SARL $0x08, DI + SHLL $0x05, DI + ORL DI, R11 + MOVB R11, (AX) + ADDQ $0x02, AX + +match_nolit_emitcopy_end_encodeBetterBlockAsm12B: + CMPL CX, 8(SP) + JAE emit_remainder_encodeBetterBlockAsm12B + CMPQ AX, (SP) + JB match_nolit_dst_ok_encodeBetterBlockAsm12B + MOVQ $0x00000000, ret+48(FP) + RET + +match_nolit_dst_ok_encodeBetterBlockAsm12B: + MOVQ $0x0000cf1bbcdcbf9b, BX + MOVQ $0x9e3779b1, DI + LEAQ 1(SI), SI + LEAQ -2(CX), R8 + MOVQ (DX)(SI*1), R9 + MOVQ 1(DX)(SI*1), R10 + MOVQ (DX)(R8*1), R11 + MOVQ 1(DX)(R8*1), R12 + SHLQ $0x10, R9 + IMULQ BX, R9 + SHRQ $0x32, R9 + SHLQ $0x20, R10 + IMULQ DI, R10 + SHRQ $0x34, R10 + SHLQ $0x10, R11 + IMULQ BX, R11 + SHRQ $0x32, R11 + SHLQ $0x20, R12 + IMULQ DI, R12 + SHRQ $0x34, R12 + LEAQ 1(SI), DI + LEAQ 1(R8), R13 + MOVL SI, 24(SP)(R9*4) + MOVL R8, 24(SP)(R11*4) + MOVL DI, 65560(SP)(R10*4) + MOVL R13, 65560(SP)(R12*4) + LEAQ 1(R8)(SI*1), DI + SHRQ $0x01, DI + ADDQ $0x01, SI + SUBQ $0x01, R8 + +index_loop_encodeBetterBlockAsm12B: + CMPQ DI, R8 + JAE search_loop_encodeBetterBlockAsm12B + MOVQ (DX)(SI*1), R9 + MOVQ (DX)(DI*1), R10 + SHLQ $0x10, R9 + IMULQ BX, R9 + SHRQ $0x32, R9 + SHLQ $0x10, R10 + IMULQ BX, R10 + SHRQ $0x32, R10 + MOVL SI, 24(SP)(R9*4) + MOVL DI, 24(SP)(R10*4) + ADDQ $0x02, SI + ADDQ $0x02, DI + JMP index_loop_encodeBetterBlockAsm12B + +emit_remainder_encodeBetterBlockAsm12B: + MOVQ src_len+32(FP), CX + SUBL 12(SP), CX + LEAQ 3(AX)(CX*1), CX + CMPQ CX, (SP) + JB emit_remainder_ok_encodeBetterBlockAsm12B + MOVQ $0x00000000, ret+48(FP) + RET + +emit_remainder_ok_encodeBetterBlockAsm12B: + MOVQ src_len+32(FP), CX + MOVL 12(SP), BX + CMPL BX, CX + JEQ emit_literal_done_emit_remainder_encodeBetterBlockAsm12B + MOVL CX, SI + MOVL CX, 12(SP) + LEAQ (DX)(BX*1), CX + SUBL BX, SI + LEAL -1(SI), DX + CMPL DX, $0x3c + JB one_byte_emit_remainder_encodeBetterBlockAsm12B + CMPL DX, $0x00000100 + JB two_bytes_emit_remainder_encodeBetterBlockAsm12B + JB three_bytes_emit_remainder_encodeBetterBlockAsm12B + +three_bytes_emit_remainder_encodeBetterBlockAsm12B: + MOVB $0xf4, (AX) + MOVW DX, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_emit_remainder_encodeBetterBlockAsm12B + +two_bytes_emit_remainder_encodeBetterBlockAsm12B: + MOVB $0xf0, (AX) + MOVB DL, 1(AX) + ADDQ $0x02, AX + CMPL DX, $0x40 + JB memmove_emit_remainder_encodeBetterBlockAsm12B + JMP memmove_long_emit_remainder_encodeBetterBlockAsm12B + +one_byte_emit_remainder_encodeBetterBlockAsm12B: + SHLB $0x02, DL + MOVB DL, (AX) + ADDQ $0x01, AX + +memmove_emit_remainder_encodeBetterBlockAsm12B: + LEAQ (AX)(SI*1), DX + MOVL SI, BX + + // genMemMoveShort + CMPQ BX, $0x03 + JB emit_lit_memmove_emit_remainder_encodeBetterBlockAsm12B_memmove_move_1or2 + JE emit_lit_memmove_emit_remainder_encodeBetterBlockAsm12B_memmove_move_3 + CMPQ BX, $0x08 + JB emit_lit_memmove_emit_remainder_encodeBetterBlockAsm12B_memmove_move_4through7 + CMPQ BX, $0x10 + JBE emit_lit_memmove_emit_remainder_encodeBetterBlockAsm12B_memmove_move_8through16 + CMPQ BX, $0x20 + JBE emit_lit_memmove_emit_remainder_encodeBetterBlockAsm12B_memmove_move_17through32 + JMP emit_lit_memmove_emit_remainder_encodeBetterBlockAsm12B_memmove_move_33through64 + +emit_lit_memmove_emit_remainder_encodeBetterBlockAsm12B_memmove_move_1or2: + MOVB (CX), SI + MOVB -1(CX)(BX*1), CL + MOVB SI, (AX) + MOVB CL, -1(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeBetterBlockAsm12B + +emit_lit_memmove_emit_remainder_encodeBetterBlockAsm12B_memmove_move_3: + MOVW (CX), SI + MOVB 2(CX), CL + MOVW SI, (AX) + MOVB CL, 2(AX) + JMP memmove_end_copy_emit_remainder_encodeBetterBlockAsm12B + +emit_lit_memmove_emit_remainder_encodeBetterBlockAsm12B_memmove_move_4through7: + MOVL (CX), SI + MOVL -4(CX)(BX*1), CX + MOVL SI, (AX) + MOVL CX, -4(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeBetterBlockAsm12B + +emit_lit_memmove_emit_remainder_encodeBetterBlockAsm12B_memmove_move_8through16: + MOVQ (CX), SI + MOVQ -8(CX)(BX*1), CX + MOVQ SI, (AX) + MOVQ CX, -8(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeBetterBlockAsm12B + +emit_lit_memmove_emit_remainder_encodeBetterBlockAsm12B_memmove_move_17through32: + MOVOU (CX), X0 + MOVOU -16(CX)(BX*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeBetterBlockAsm12B + +emit_lit_memmove_emit_remainder_encodeBetterBlockAsm12B_memmove_move_33through64: + MOVOU (CX), X0 + MOVOU 16(CX), X1 + MOVOU -32(CX)(BX*1), X2 + MOVOU -16(CX)(BX*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(BX*1) + MOVOU X3, -16(AX)(BX*1) + +memmove_end_copy_emit_remainder_encodeBetterBlockAsm12B: + MOVQ DX, AX + JMP emit_literal_done_emit_remainder_encodeBetterBlockAsm12B + +memmove_long_emit_remainder_encodeBetterBlockAsm12B: + LEAQ (AX)(SI*1), DX + MOVL SI, BX + + // genMemMoveLong + MOVOU (CX), X0 + MOVOU 16(CX), X1 + MOVOU -32(CX)(BX*1), X2 + MOVOU -16(CX)(BX*1), X3 + MOVQ BX, DI + SHRQ $0x05, DI + MOVQ AX, SI + ANDL $0x0000001f, SI + MOVQ $0x00000040, R8 + SUBQ SI, R8 + DECQ DI + JA emit_lit_memmove_long_emit_remainder_encodeBetterBlockAsm12Blarge_forward_sse_loop_32 + LEAQ -32(CX)(R8*1), SI + LEAQ -32(AX)(R8*1), R9 + +emit_lit_memmove_long_emit_remainder_encodeBetterBlockAsm12Blarge_big_loop_back: + MOVOU (SI), X4 + MOVOU 16(SI), X5 + MOVOA X4, (R9) + MOVOA X5, 16(R9) + ADDQ $0x20, R9 + ADDQ $0x20, SI + ADDQ $0x20, R8 + DECQ DI + JNA emit_lit_memmove_long_emit_remainder_encodeBetterBlockAsm12Blarge_big_loop_back + +emit_lit_memmove_long_emit_remainder_encodeBetterBlockAsm12Blarge_forward_sse_loop_32: + MOVOU -32(CX)(R8*1), X4 + MOVOU -16(CX)(R8*1), X5 + MOVOA X4, -32(AX)(R8*1) + MOVOA X5, -16(AX)(R8*1) + ADDQ $0x20, R8 + CMPQ BX, R8 + JAE emit_lit_memmove_long_emit_remainder_encodeBetterBlockAsm12Blarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(BX*1) + MOVOU X3, -16(AX)(BX*1) + MOVQ DX, AX + +emit_literal_done_emit_remainder_encodeBetterBlockAsm12B: + MOVQ dst_base+0(FP), CX + SUBQ CX, AX + MOVQ AX, ret+48(FP) + RET + +// func encodeBetterBlockAsm10B(dst []byte, src []byte) int +// Requires: BMI, SSE2 +TEXT ·encodeBetterBlockAsm10B(SB), $20504-56 + MOVQ dst_base+0(FP), AX + MOVQ $0x000000a0, CX + LEAQ 24(SP), DX + PXOR X0, X0 + +zero_loop_encodeBetterBlockAsm10B: + MOVOU X0, (DX) + MOVOU X0, 16(DX) + MOVOU X0, 32(DX) + MOVOU X0, 48(DX) + MOVOU X0, 64(DX) + MOVOU X0, 80(DX) + MOVOU X0, 96(DX) + MOVOU X0, 112(DX) + ADDQ $0x80, DX + DECQ CX + JNZ zero_loop_encodeBetterBlockAsm10B + MOVL $0x00000000, 12(SP) + MOVQ src_len+32(FP), CX + LEAQ -6(CX), DX + LEAQ -8(CX), BX + MOVL BX, 8(SP) + SHRQ $0x05, CX + SUBL CX, DX + LEAQ (AX)(DX*1), DX + MOVQ DX, (SP) + MOVL $0x00000001, CX + MOVL $0x00000000, 16(SP) + MOVQ src_base+24(FP), DX + +search_loop_encodeBetterBlockAsm10B: + MOVL CX, BX + SUBL 12(SP), BX + SHRL $0x05, BX + LEAL 1(CX)(BX*1), BX + CMPL BX, 8(SP) + JAE emit_remainder_encodeBetterBlockAsm10B + MOVQ (DX)(CX*1), SI + MOVL BX, 20(SP) + MOVQ $0x0000cf1bbcdcbf9b, R8 + MOVQ $0x9e3779b1, BX + MOVQ SI, R9 + MOVQ SI, R10 + SHLQ $0x10, R9 + IMULQ R8, R9 + SHRQ $0x34, R9 + SHLQ $0x20, R10 + IMULQ BX, R10 + SHRQ $0x36, R10 + MOVL 24(SP)(R9*4), BX + MOVL 16408(SP)(R10*4), DI + MOVL CX, 24(SP)(R9*4) + MOVL CX, 16408(SP)(R10*4) + MOVQ (DX)(BX*1), R9 + MOVQ (DX)(DI*1), R10 + CMPQ R9, SI + JEQ candidate_match_encodeBetterBlockAsm10B + CMPQ R10, SI + JNE no_short_found_encodeBetterBlockAsm10B + MOVL DI, BX + JMP candidate_match_encodeBetterBlockAsm10B + +no_short_found_encodeBetterBlockAsm10B: + CMPL R9, SI + JEQ candidate_match_encodeBetterBlockAsm10B + CMPL R10, SI + JEQ candidateS_match_encodeBetterBlockAsm10B + MOVL 20(SP), CX + JMP search_loop_encodeBetterBlockAsm10B + +candidateS_match_encodeBetterBlockAsm10B: + SHRQ $0x08, SI + MOVQ SI, R9 + SHLQ $0x10, R9 + IMULQ R8, R9 + SHRQ $0x34, R9 + MOVL 24(SP)(R9*4), BX + INCL CX + MOVL CX, 24(SP)(R9*4) + CMPL (DX)(BX*1), SI + JEQ candidate_match_encodeBetterBlockAsm10B + DECL CX + MOVL DI, BX + +candidate_match_encodeBetterBlockAsm10B: + MOVL 12(SP), SI + TESTL BX, BX + JZ match_extend_back_end_encodeBetterBlockAsm10B + +match_extend_back_loop_encodeBetterBlockAsm10B: + CMPL CX, SI + JBE match_extend_back_end_encodeBetterBlockAsm10B + MOVB -1(DX)(BX*1), DI + MOVB -1(DX)(CX*1), R8 + CMPB DI, R8 + JNE match_extend_back_end_encodeBetterBlockAsm10B + LEAL -1(CX), CX + DECL BX + JZ match_extend_back_end_encodeBetterBlockAsm10B + JMP match_extend_back_loop_encodeBetterBlockAsm10B + +match_extend_back_end_encodeBetterBlockAsm10B: + MOVL CX, SI + SUBL 12(SP), SI + LEAQ 3(AX)(SI*1), SI + CMPQ SI, (SP) + JB match_dst_size_check_encodeBetterBlockAsm10B + MOVQ $0x00000000, ret+48(FP) + RET + +match_dst_size_check_encodeBetterBlockAsm10B: + MOVL CX, SI + ADDL $0x04, CX + ADDL $0x04, BX + MOVQ src_len+32(FP), DI + SUBL CX, DI + LEAQ (DX)(CX*1), R8 + LEAQ (DX)(BX*1), R9 + + // matchLen + XORL R11, R11 + +matchlen_loopback_16_match_nolit_encodeBetterBlockAsm10B: + CMPL DI, $0x10 + JB matchlen_match8_match_nolit_encodeBetterBlockAsm10B + MOVQ (R8)(R11*1), R10 + MOVQ 8(R8)(R11*1), R12 + XORQ (R9)(R11*1), R10 + JNZ matchlen_bsf_8_match_nolit_encodeBetterBlockAsm10B + XORQ 8(R9)(R11*1), R12 + JNZ matchlen_bsf_16match_nolit_encodeBetterBlockAsm10B + LEAL -16(DI), DI + LEAL 16(R11), R11 + JMP matchlen_loopback_16_match_nolit_encodeBetterBlockAsm10B + +matchlen_bsf_16match_nolit_encodeBetterBlockAsm10B: +#ifdef GOAMD64_v3 + TZCNTQ R12, R12 + +#else + BSFQ R12, R12 + +#endif + SARQ $0x03, R12 + LEAL 8(R11)(R12*1), R11 + JMP match_nolit_end_encodeBetterBlockAsm10B + +matchlen_match8_match_nolit_encodeBetterBlockAsm10B: + CMPL DI, $0x08 + JB matchlen_match4_match_nolit_encodeBetterBlockAsm10B + MOVQ (R8)(R11*1), R10 + XORQ (R9)(R11*1), R10 + JNZ matchlen_bsf_8_match_nolit_encodeBetterBlockAsm10B + LEAL -8(DI), DI + LEAL 8(R11), R11 + JMP matchlen_match4_match_nolit_encodeBetterBlockAsm10B + +matchlen_bsf_8_match_nolit_encodeBetterBlockAsm10B: +#ifdef GOAMD64_v3 + TZCNTQ R10, R10 + +#else + BSFQ R10, R10 + +#endif + SARQ $0x03, R10 + LEAL (R11)(R10*1), R11 + JMP match_nolit_end_encodeBetterBlockAsm10B + +matchlen_match4_match_nolit_encodeBetterBlockAsm10B: + CMPL DI, $0x04 + JB matchlen_match2_match_nolit_encodeBetterBlockAsm10B + MOVL (R8)(R11*1), R10 + CMPL (R9)(R11*1), R10 + JNE matchlen_match2_match_nolit_encodeBetterBlockAsm10B + LEAL -4(DI), DI + LEAL 4(R11), R11 + +matchlen_match2_match_nolit_encodeBetterBlockAsm10B: + CMPL DI, $0x01 + JE matchlen_match1_match_nolit_encodeBetterBlockAsm10B + JB match_nolit_end_encodeBetterBlockAsm10B + MOVW (R8)(R11*1), R10 + CMPW (R9)(R11*1), R10 + JNE matchlen_match1_match_nolit_encodeBetterBlockAsm10B + LEAL 2(R11), R11 + SUBL $0x02, DI + JZ match_nolit_end_encodeBetterBlockAsm10B + +matchlen_match1_match_nolit_encodeBetterBlockAsm10B: + MOVB (R8)(R11*1), R10 + CMPB (R9)(R11*1), R10 + JNE match_nolit_end_encodeBetterBlockAsm10B + LEAL 1(R11), R11 + +match_nolit_end_encodeBetterBlockAsm10B: + MOVL CX, DI + SUBL BX, DI + + // Check if repeat + CMPL 16(SP), DI + JEQ match_is_repeat_encodeBetterBlockAsm10B + MOVL DI, 16(SP) + MOVL 12(SP), BX + CMPL BX, SI + JEQ emit_literal_done_match_emit_encodeBetterBlockAsm10B + MOVL SI, R8 + MOVL SI, 12(SP) + LEAQ (DX)(BX*1), R9 + SUBL BX, R8 + LEAL -1(R8), BX + CMPL BX, $0x3c + JB one_byte_match_emit_encodeBetterBlockAsm10B + CMPL BX, $0x00000100 + JB two_bytes_match_emit_encodeBetterBlockAsm10B + JB three_bytes_match_emit_encodeBetterBlockAsm10B + +three_bytes_match_emit_encodeBetterBlockAsm10B: + MOVB $0xf4, (AX) + MOVW BX, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_match_emit_encodeBetterBlockAsm10B + +two_bytes_match_emit_encodeBetterBlockAsm10B: + MOVB $0xf0, (AX) + MOVB BL, 1(AX) + ADDQ $0x02, AX + CMPL BX, $0x40 + JB memmove_match_emit_encodeBetterBlockAsm10B + JMP memmove_long_match_emit_encodeBetterBlockAsm10B + +one_byte_match_emit_encodeBetterBlockAsm10B: + SHLB $0x02, BL + MOVB BL, (AX) + ADDQ $0x01, AX + +memmove_match_emit_encodeBetterBlockAsm10B: + LEAQ (AX)(R8*1), BX + + // genMemMoveShort + CMPQ R8, $0x04 + JBE emit_lit_memmove_match_emit_encodeBetterBlockAsm10B_memmove_move_4 + CMPQ R8, $0x08 + JB emit_lit_memmove_match_emit_encodeBetterBlockAsm10B_memmove_move_4through7 + CMPQ R8, $0x10 + JBE emit_lit_memmove_match_emit_encodeBetterBlockAsm10B_memmove_move_8through16 + CMPQ R8, $0x20 + JBE emit_lit_memmove_match_emit_encodeBetterBlockAsm10B_memmove_move_17through32 + JMP emit_lit_memmove_match_emit_encodeBetterBlockAsm10B_memmove_move_33through64 + +emit_lit_memmove_match_emit_encodeBetterBlockAsm10B_memmove_move_4: + MOVL (R9), R10 + MOVL R10, (AX) + JMP memmove_end_copy_match_emit_encodeBetterBlockAsm10B + +emit_lit_memmove_match_emit_encodeBetterBlockAsm10B_memmove_move_4through7: + MOVL (R9), R10 + MOVL -4(R9)(R8*1), R9 + MOVL R10, (AX) + MOVL R9, -4(AX)(R8*1) + JMP memmove_end_copy_match_emit_encodeBetterBlockAsm10B + +emit_lit_memmove_match_emit_encodeBetterBlockAsm10B_memmove_move_8through16: + MOVQ (R9), R10 + MOVQ -8(R9)(R8*1), R9 + MOVQ R10, (AX) + MOVQ R9, -8(AX)(R8*1) + JMP memmove_end_copy_match_emit_encodeBetterBlockAsm10B + +emit_lit_memmove_match_emit_encodeBetterBlockAsm10B_memmove_move_17through32: + MOVOU (R9), X0 + MOVOU -16(R9)(R8*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(R8*1) + JMP memmove_end_copy_match_emit_encodeBetterBlockAsm10B + +emit_lit_memmove_match_emit_encodeBetterBlockAsm10B_memmove_move_33through64: + MOVOU (R9), X0 + MOVOU 16(R9), X1 + MOVOU -32(R9)(R8*1), X2 + MOVOU -16(R9)(R8*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + +memmove_end_copy_match_emit_encodeBetterBlockAsm10B: + MOVQ BX, AX + JMP emit_literal_done_match_emit_encodeBetterBlockAsm10B + +memmove_long_match_emit_encodeBetterBlockAsm10B: + LEAQ (AX)(R8*1), BX + + // genMemMoveLong + MOVOU (R9), X0 + MOVOU 16(R9), X1 + MOVOU -32(R9)(R8*1), X2 + MOVOU -16(R9)(R8*1), X3 + MOVQ R8, R12 + SHRQ $0x05, R12 + MOVQ AX, R10 + ANDL $0x0000001f, R10 + MOVQ $0x00000040, R13 + SUBQ R10, R13 + DECQ R12 + JA emit_lit_memmove_long_match_emit_encodeBetterBlockAsm10Blarge_forward_sse_loop_32 + LEAQ -32(R9)(R13*1), R10 + LEAQ -32(AX)(R13*1), R14 + +emit_lit_memmove_long_match_emit_encodeBetterBlockAsm10Blarge_big_loop_back: + MOVOU (R10), X4 + MOVOU 16(R10), X5 + MOVOA X4, (R14) + MOVOA X5, 16(R14) + ADDQ $0x20, R14 + ADDQ $0x20, R10 + ADDQ $0x20, R13 + DECQ R12 + JNA emit_lit_memmove_long_match_emit_encodeBetterBlockAsm10Blarge_big_loop_back + +emit_lit_memmove_long_match_emit_encodeBetterBlockAsm10Blarge_forward_sse_loop_32: + MOVOU -32(R9)(R13*1), X4 + MOVOU -16(R9)(R13*1), X5 + MOVOA X4, -32(AX)(R13*1) + MOVOA X5, -16(AX)(R13*1) + ADDQ $0x20, R13 + CMPQ R8, R13 + JAE emit_lit_memmove_long_match_emit_encodeBetterBlockAsm10Blarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + MOVQ BX, AX + +emit_literal_done_match_emit_encodeBetterBlockAsm10B: + ADDL R11, CX + ADDL $0x04, R11 + MOVL CX, 12(SP) + + // emitCopy + CMPL R11, $0x40 + JBE two_byte_offset_short_match_nolit_encodeBetterBlockAsm10B + CMPL DI, $0x00000800 + JAE long_offset_short_match_nolit_encodeBetterBlockAsm10B + MOVL $0x00000001, BX + LEAL 16(BX), BX + MOVB DI, 1(AX) + SHRL $0x08, DI + SHLL $0x05, DI + ORL DI, BX + MOVB BL, (AX) + ADDQ $0x02, AX + SUBL $0x08, R11 + + // emitRepeat + LEAL -4(R11), R11 + JMP cant_repeat_two_offset_match_nolit_encodeBetterBlockAsm10B_emit_copy_short_2b + MOVL R11, BX + LEAL -4(R11), R11 + CMPL BX, $0x08 + JBE repeat_two_match_nolit_encodeBetterBlockAsm10B_emit_copy_short_2b + CMPL BX, $0x0c + JAE cant_repeat_two_offset_match_nolit_encodeBetterBlockAsm10B_emit_copy_short_2b + CMPL DI, $0x00000800 + JB repeat_two_offset_match_nolit_encodeBetterBlockAsm10B_emit_copy_short_2b + +cant_repeat_two_offset_match_nolit_encodeBetterBlockAsm10B_emit_copy_short_2b: + CMPL R11, $0x00000104 + JB repeat_three_match_nolit_encodeBetterBlockAsm10B_emit_copy_short_2b + LEAL -256(R11), R11 + MOVW $0x0019, (AX) + MOVW R11, 2(AX) + ADDQ $0x04, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm10B + +repeat_three_match_nolit_encodeBetterBlockAsm10B_emit_copy_short_2b: + LEAL -4(R11), R11 + MOVW $0x0015, (AX) + MOVB R11, 2(AX) + ADDQ $0x03, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm10B + +repeat_two_match_nolit_encodeBetterBlockAsm10B_emit_copy_short_2b: + SHLL $0x02, R11 + ORL $0x01, R11 + MOVW R11, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm10B + +repeat_two_offset_match_nolit_encodeBetterBlockAsm10B_emit_copy_short_2b: + XORQ BX, BX + LEAL 1(BX)(R11*4), R11 + MOVB DI, 1(AX) + SARL $0x08, DI + SHLL $0x05, DI + ORL DI, R11 + MOVB R11, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm10B + +long_offset_short_match_nolit_encodeBetterBlockAsm10B: + MOVB $0xee, (AX) + MOVW DI, 1(AX) + LEAL -60(R11), R11 + ADDQ $0x03, AX + + // emitRepeat + MOVL R11, BX + LEAL -4(R11), R11 + CMPL BX, $0x08 + JBE repeat_two_match_nolit_encodeBetterBlockAsm10B_emit_copy_short + CMPL BX, $0x0c + JAE cant_repeat_two_offset_match_nolit_encodeBetterBlockAsm10B_emit_copy_short + CMPL DI, $0x00000800 + JB repeat_two_offset_match_nolit_encodeBetterBlockAsm10B_emit_copy_short + +cant_repeat_two_offset_match_nolit_encodeBetterBlockAsm10B_emit_copy_short: + CMPL R11, $0x00000104 + JB repeat_three_match_nolit_encodeBetterBlockAsm10B_emit_copy_short + LEAL -256(R11), R11 + MOVW $0x0019, (AX) + MOVW R11, 2(AX) + ADDQ $0x04, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm10B + +repeat_three_match_nolit_encodeBetterBlockAsm10B_emit_copy_short: + LEAL -4(R11), R11 + MOVW $0x0015, (AX) + MOVB R11, 2(AX) + ADDQ $0x03, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm10B + +repeat_two_match_nolit_encodeBetterBlockAsm10B_emit_copy_short: + SHLL $0x02, R11 + ORL $0x01, R11 + MOVW R11, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm10B + +repeat_two_offset_match_nolit_encodeBetterBlockAsm10B_emit_copy_short: + XORQ BX, BX + LEAL 1(BX)(R11*4), R11 + MOVB DI, 1(AX) + SARL $0x08, DI + SHLL $0x05, DI + ORL DI, R11 + MOVB R11, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm10B + +two_byte_offset_short_match_nolit_encodeBetterBlockAsm10B: + MOVL R11, BX + SHLL $0x02, BX + CMPL R11, $0x0c + JAE emit_copy_three_match_nolit_encodeBetterBlockAsm10B + CMPL DI, $0x00000800 + JAE emit_copy_three_match_nolit_encodeBetterBlockAsm10B + LEAL -15(BX), BX + MOVB DI, 1(AX) + SHRL $0x08, DI + SHLL $0x05, DI + ORL DI, BX + MOVB BL, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm10B + +emit_copy_three_match_nolit_encodeBetterBlockAsm10B: + LEAL -2(BX), BX + MOVB BL, (AX) + MOVW DI, 1(AX) + ADDQ $0x03, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm10B + +match_is_repeat_encodeBetterBlockAsm10B: + MOVL 12(SP), BX + CMPL BX, SI + JEQ emit_literal_done_match_emit_repeat_encodeBetterBlockAsm10B + MOVL SI, R8 + MOVL SI, 12(SP) + LEAQ (DX)(BX*1), R9 + SUBL BX, R8 + LEAL -1(R8), BX + CMPL BX, $0x3c + JB one_byte_match_emit_repeat_encodeBetterBlockAsm10B + CMPL BX, $0x00000100 + JB two_bytes_match_emit_repeat_encodeBetterBlockAsm10B + JB three_bytes_match_emit_repeat_encodeBetterBlockAsm10B + +three_bytes_match_emit_repeat_encodeBetterBlockAsm10B: + MOVB $0xf4, (AX) + MOVW BX, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_match_emit_repeat_encodeBetterBlockAsm10B + +two_bytes_match_emit_repeat_encodeBetterBlockAsm10B: + MOVB $0xf0, (AX) + MOVB BL, 1(AX) + ADDQ $0x02, AX + CMPL BX, $0x40 + JB memmove_match_emit_repeat_encodeBetterBlockAsm10B + JMP memmove_long_match_emit_repeat_encodeBetterBlockAsm10B + +one_byte_match_emit_repeat_encodeBetterBlockAsm10B: + SHLB $0x02, BL + MOVB BL, (AX) + ADDQ $0x01, AX + +memmove_match_emit_repeat_encodeBetterBlockAsm10B: + LEAQ (AX)(R8*1), BX + + // genMemMoveShort + CMPQ R8, $0x04 + JBE emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm10B_memmove_move_4 + CMPQ R8, $0x08 + JB emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm10B_memmove_move_4through7 + CMPQ R8, $0x10 + JBE emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm10B_memmove_move_8through16 + CMPQ R8, $0x20 + JBE emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm10B_memmove_move_17through32 + JMP emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm10B_memmove_move_33through64 + +emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm10B_memmove_move_4: + MOVL (R9), R10 + MOVL R10, (AX) + JMP memmove_end_copy_match_emit_repeat_encodeBetterBlockAsm10B + +emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm10B_memmove_move_4through7: + MOVL (R9), R10 + MOVL -4(R9)(R8*1), R9 + MOVL R10, (AX) + MOVL R9, -4(AX)(R8*1) + JMP memmove_end_copy_match_emit_repeat_encodeBetterBlockAsm10B + +emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm10B_memmove_move_8through16: + MOVQ (R9), R10 + MOVQ -8(R9)(R8*1), R9 + MOVQ R10, (AX) + MOVQ R9, -8(AX)(R8*1) + JMP memmove_end_copy_match_emit_repeat_encodeBetterBlockAsm10B + +emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm10B_memmove_move_17through32: + MOVOU (R9), X0 + MOVOU -16(R9)(R8*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(R8*1) + JMP memmove_end_copy_match_emit_repeat_encodeBetterBlockAsm10B + +emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm10B_memmove_move_33through64: + MOVOU (R9), X0 + MOVOU 16(R9), X1 + MOVOU -32(R9)(R8*1), X2 + MOVOU -16(R9)(R8*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + +memmove_end_copy_match_emit_repeat_encodeBetterBlockAsm10B: + MOVQ BX, AX + JMP emit_literal_done_match_emit_repeat_encodeBetterBlockAsm10B + +memmove_long_match_emit_repeat_encodeBetterBlockAsm10B: + LEAQ (AX)(R8*1), BX + + // genMemMoveLong + MOVOU (R9), X0 + MOVOU 16(R9), X1 + MOVOU -32(R9)(R8*1), X2 + MOVOU -16(R9)(R8*1), X3 + MOVQ R8, R12 + SHRQ $0x05, R12 + MOVQ AX, R10 + ANDL $0x0000001f, R10 + MOVQ $0x00000040, R13 + SUBQ R10, R13 + DECQ R12 + JA emit_lit_memmove_long_match_emit_repeat_encodeBetterBlockAsm10Blarge_forward_sse_loop_32 + LEAQ -32(R9)(R13*1), R10 + LEAQ -32(AX)(R13*1), R14 + +emit_lit_memmove_long_match_emit_repeat_encodeBetterBlockAsm10Blarge_big_loop_back: + MOVOU (R10), X4 + MOVOU 16(R10), X5 + MOVOA X4, (R14) + MOVOA X5, 16(R14) + ADDQ $0x20, R14 + ADDQ $0x20, R10 + ADDQ $0x20, R13 + DECQ R12 + JNA emit_lit_memmove_long_match_emit_repeat_encodeBetterBlockAsm10Blarge_big_loop_back + +emit_lit_memmove_long_match_emit_repeat_encodeBetterBlockAsm10Blarge_forward_sse_loop_32: + MOVOU -32(R9)(R13*1), X4 + MOVOU -16(R9)(R13*1), X5 + MOVOA X4, -32(AX)(R13*1) + MOVOA X5, -16(AX)(R13*1) + ADDQ $0x20, R13 + CMPQ R8, R13 + JAE emit_lit_memmove_long_match_emit_repeat_encodeBetterBlockAsm10Blarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + MOVQ BX, AX + +emit_literal_done_match_emit_repeat_encodeBetterBlockAsm10B: + ADDL R11, CX + ADDL $0x04, R11 + MOVL CX, 12(SP) + + // emitRepeat + MOVL R11, BX + LEAL -4(R11), R11 + CMPL BX, $0x08 + JBE repeat_two_match_nolit_repeat_encodeBetterBlockAsm10B + CMPL BX, $0x0c + JAE cant_repeat_two_offset_match_nolit_repeat_encodeBetterBlockAsm10B + CMPL DI, $0x00000800 + JB repeat_two_offset_match_nolit_repeat_encodeBetterBlockAsm10B + +cant_repeat_two_offset_match_nolit_repeat_encodeBetterBlockAsm10B: + CMPL R11, $0x00000104 + JB repeat_three_match_nolit_repeat_encodeBetterBlockAsm10B + LEAL -256(R11), R11 + MOVW $0x0019, (AX) + MOVW R11, 2(AX) + ADDQ $0x04, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm10B + +repeat_three_match_nolit_repeat_encodeBetterBlockAsm10B: + LEAL -4(R11), R11 + MOVW $0x0015, (AX) + MOVB R11, 2(AX) + ADDQ $0x03, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm10B + +repeat_two_match_nolit_repeat_encodeBetterBlockAsm10B: + SHLL $0x02, R11 + ORL $0x01, R11 + MOVW R11, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm10B + +repeat_two_offset_match_nolit_repeat_encodeBetterBlockAsm10B: + XORQ BX, BX + LEAL 1(BX)(R11*4), R11 + MOVB DI, 1(AX) + SARL $0x08, DI + SHLL $0x05, DI + ORL DI, R11 + MOVB R11, (AX) + ADDQ $0x02, AX + +match_nolit_emitcopy_end_encodeBetterBlockAsm10B: + CMPL CX, 8(SP) + JAE emit_remainder_encodeBetterBlockAsm10B + CMPQ AX, (SP) + JB match_nolit_dst_ok_encodeBetterBlockAsm10B + MOVQ $0x00000000, ret+48(FP) + RET + +match_nolit_dst_ok_encodeBetterBlockAsm10B: + MOVQ $0x0000cf1bbcdcbf9b, BX + MOVQ $0x9e3779b1, DI + LEAQ 1(SI), SI + LEAQ -2(CX), R8 + MOVQ (DX)(SI*1), R9 + MOVQ 1(DX)(SI*1), R10 + MOVQ (DX)(R8*1), R11 + MOVQ 1(DX)(R8*1), R12 + SHLQ $0x10, R9 + IMULQ BX, R9 + SHRQ $0x34, R9 + SHLQ $0x20, R10 + IMULQ DI, R10 + SHRQ $0x36, R10 + SHLQ $0x10, R11 + IMULQ BX, R11 + SHRQ $0x34, R11 + SHLQ $0x20, R12 + IMULQ DI, R12 + SHRQ $0x36, R12 + LEAQ 1(SI), DI + LEAQ 1(R8), R13 + MOVL SI, 24(SP)(R9*4) + MOVL R8, 24(SP)(R11*4) + MOVL DI, 16408(SP)(R10*4) + MOVL R13, 16408(SP)(R12*4) + LEAQ 1(R8)(SI*1), DI + SHRQ $0x01, DI + ADDQ $0x01, SI + SUBQ $0x01, R8 + +index_loop_encodeBetterBlockAsm10B: + CMPQ DI, R8 + JAE search_loop_encodeBetterBlockAsm10B + MOVQ (DX)(SI*1), R9 + MOVQ (DX)(DI*1), R10 + SHLQ $0x10, R9 + IMULQ BX, R9 + SHRQ $0x34, R9 + SHLQ $0x10, R10 + IMULQ BX, R10 + SHRQ $0x34, R10 + MOVL SI, 24(SP)(R9*4) + MOVL DI, 24(SP)(R10*4) + ADDQ $0x02, SI + ADDQ $0x02, DI + JMP index_loop_encodeBetterBlockAsm10B + +emit_remainder_encodeBetterBlockAsm10B: + MOVQ src_len+32(FP), CX + SUBL 12(SP), CX + LEAQ 3(AX)(CX*1), CX + CMPQ CX, (SP) + JB emit_remainder_ok_encodeBetterBlockAsm10B + MOVQ $0x00000000, ret+48(FP) + RET + +emit_remainder_ok_encodeBetterBlockAsm10B: + MOVQ src_len+32(FP), CX + MOVL 12(SP), BX + CMPL BX, CX + JEQ emit_literal_done_emit_remainder_encodeBetterBlockAsm10B + MOVL CX, SI + MOVL CX, 12(SP) + LEAQ (DX)(BX*1), CX + SUBL BX, SI + LEAL -1(SI), DX + CMPL DX, $0x3c + JB one_byte_emit_remainder_encodeBetterBlockAsm10B + CMPL DX, $0x00000100 + JB two_bytes_emit_remainder_encodeBetterBlockAsm10B + JB three_bytes_emit_remainder_encodeBetterBlockAsm10B + +three_bytes_emit_remainder_encodeBetterBlockAsm10B: + MOVB $0xf4, (AX) + MOVW DX, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_emit_remainder_encodeBetterBlockAsm10B + +two_bytes_emit_remainder_encodeBetterBlockAsm10B: + MOVB $0xf0, (AX) + MOVB DL, 1(AX) + ADDQ $0x02, AX + CMPL DX, $0x40 + JB memmove_emit_remainder_encodeBetterBlockAsm10B + JMP memmove_long_emit_remainder_encodeBetterBlockAsm10B + +one_byte_emit_remainder_encodeBetterBlockAsm10B: + SHLB $0x02, DL + MOVB DL, (AX) + ADDQ $0x01, AX + +memmove_emit_remainder_encodeBetterBlockAsm10B: + LEAQ (AX)(SI*1), DX + MOVL SI, BX + + // genMemMoveShort + CMPQ BX, $0x03 + JB emit_lit_memmove_emit_remainder_encodeBetterBlockAsm10B_memmove_move_1or2 + JE emit_lit_memmove_emit_remainder_encodeBetterBlockAsm10B_memmove_move_3 + CMPQ BX, $0x08 + JB emit_lit_memmove_emit_remainder_encodeBetterBlockAsm10B_memmove_move_4through7 + CMPQ BX, $0x10 + JBE emit_lit_memmove_emit_remainder_encodeBetterBlockAsm10B_memmove_move_8through16 + CMPQ BX, $0x20 + JBE emit_lit_memmove_emit_remainder_encodeBetterBlockAsm10B_memmove_move_17through32 + JMP emit_lit_memmove_emit_remainder_encodeBetterBlockAsm10B_memmove_move_33through64 + +emit_lit_memmove_emit_remainder_encodeBetterBlockAsm10B_memmove_move_1or2: + MOVB (CX), SI + MOVB -1(CX)(BX*1), CL + MOVB SI, (AX) + MOVB CL, -1(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeBetterBlockAsm10B + +emit_lit_memmove_emit_remainder_encodeBetterBlockAsm10B_memmove_move_3: + MOVW (CX), SI + MOVB 2(CX), CL + MOVW SI, (AX) + MOVB CL, 2(AX) + JMP memmove_end_copy_emit_remainder_encodeBetterBlockAsm10B + +emit_lit_memmove_emit_remainder_encodeBetterBlockAsm10B_memmove_move_4through7: + MOVL (CX), SI + MOVL -4(CX)(BX*1), CX + MOVL SI, (AX) + MOVL CX, -4(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeBetterBlockAsm10B + +emit_lit_memmove_emit_remainder_encodeBetterBlockAsm10B_memmove_move_8through16: + MOVQ (CX), SI + MOVQ -8(CX)(BX*1), CX + MOVQ SI, (AX) + MOVQ CX, -8(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeBetterBlockAsm10B + +emit_lit_memmove_emit_remainder_encodeBetterBlockAsm10B_memmove_move_17through32: + MOVOU (CX), X0 + MOVOU -16(CX)(BX*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeBetterBlockAsm10B + +emit_lit_memmove_emit_remainder_encodeBetterBlockAsm10B_memmove_move_33through64: + MOVOU (CX), X0 + MOVOU 16(CX), X1 + MOVOU -32(CX)(BX*1), X2 + MOVOU -16(CX)(BX*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(BX*1) + MOVOU X3, -16(AX)(BX*1) + +memmove_end_copy_emit_remainder_encodeBetterBlockAsm10B: + MOVQ DX, AX + JMP emit_literal_done_emit_remainder_encodeBetterBlockAsm10B + +memmove_long_emit_remainder_encodeBetterBlockAsm10B: + LEAQ (AX)(SI*1), DX + MOVL SI, BX + + // genMemMoveLong + MOVOU (CX), X0 + MOVOU 16(CX), X1 + MOVOU -32(CX)(BX*1), X2 + MOVOU -16(CX)(BX*1), X3 + MOVQ BX, DI + SHRQ $0x05, DI + MOVQ AX, SI + ANDL $0x0000001f, SI + MOVQ $0x00000040, R8 + SUBQ SI, R8 + DECQ DI + JA emit_lit_memmove_long_emit_remainder_encodeBetterBlockAsm10Blarge_forward_sse_loop_32 + LEAQ -32(CX)(R8*1), SI + LEAQ -32(AX)(R8*1), R9 + +emit_lit_memmove_long_emit_remainder_encodeBetterBlockAsm10Blarge_big_loop_back: + MOVOU (SI), X4 + MOVOU 16(SI), X5 + MOVOA X4, (R9) + MOVOA X5, 16(R9) + ADDQ $0x20, R9 + ADDQ $0x20, SI + ADDQ $0x20, R8 + DECQ DI + JNA emit_lit_memmove_long_emit_remainder_encodeBetterBlockAsm10Blarge_big_loop_back + +emit_lit_memmove_long_emit_remainder_encodeBetterBlockAsm10Blarge_forward_sse_loop_32: + MOVOU -32(CX)(R8*1), X4 + MOVOU -16(CX)(R8*1), X5 + MOVOA X4, -32(AX)(R8*1) + MOVOA X5, -16(AX)(R8*1) + ADDQ $0x20, R8 + CMPQ BX, R8 + JAE emit_lit_memmove_long_emit_remainder_encodeBetterBlockAsm10Blarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(BX*1) + MOVOU X3, -16(AX)(BX*1) + MOVQ DX, AX + +emit_literal_done_emit_remainder_encodeBetterBlockAsm10B: + MOVQ dst_base+0(FP), CX + SUBQ CX, AX + MOVQ AX, ret+48(FP) + RET + +// func encodeBetterBlockAsm8B(dst []byte, src []byte) int +// Requires: BMI, SSE2 +TEXT ·encodeBetterBlockAsm8B(SB), $5144-56 + MOVQ dst_base+0(FP), AX + MOVQ $0x00000028, CX + LEAQ 24(SP), DX + PXOR X0, X0 + +zero_loop_encodeBetterBlockAsm8B: + MOVOU X0, (DX) + MOVOU X0, 16(DX) + MOVOU X0, 32(DX) + MOVOU X0, 48(DX) + MOVOU X0, 64(DX) + MOVOU X0, 80(DX) + MOVOU X0, 96(DX) + MOVOU X0, 112(DX) + ADDQ $0x80, DX + DECQ CX + JNZ zero_loop_encodeBetterBlockAsm8B + MOVL $0x00000000, 12(SP) + MOVQ src_len+32(FP), CX + LEAQ -6(CX), DX + LEAQ -8(CX), BX + MOVL BX, 8(SP) + SHRQ $0x05, CX + SUBL CX, DX + LEAQ (AX)(DX*1), DX + MOVQ DX, (SP) + MOVL $0x00000001, CX + MOVL $0x00000000, 16(SP) + MOVQ src_base+24(FP), DX + +search_loop_encodeBetterBlockAsm8B: + MOVL CX, BX + SUBL 12(SP), BX + SHRL $0x04, BX + LEAL 1(CX)(BX*1), BX + CMPL BX, 8(SP) + JAE emit_remainder_encodeBetterBlockAsm8B + MOVQ (DX)(CX*1), SI + MOVL BX, 20(SP) + MOVQ $0x0000cf1bbcdcbf9b, R8 + MOVQ $0x9e3779b1, BX + MOVQ SI, R9 + MOVQ SI, R10 + SHLQ $0x10, R9 + IMULQ R8, R9 + SHRQ $0x36, R9 + SHLQ $0x20, R10 + IMULQ BX, R10 + SHRQ $0x38, R10 + MOVL 24(SP)(R9*4), BX + MOVL 4120(SP)(R10*4), DI + MOVL CX, 24(SP)(R9*4) + MOVL CX, 4120(SP)(R10*4) + MOVQ (DX)(BX*1), R9 + MOVQ (DX)(DI*1), R10 + CMPQ R9, SI + JEQ candidate_match_encodeBetterBlockAsm8B + CMPQ R10, SI + JNE no_short_found_encodeBetterBlockAsm8B + MOVL DI, BX + JMP candidate_match_encodeBetterBlockAsm8B + +no_short_found_encodeBetterBlockAsm8B: + CMPL R9, SI + JEQ candidate_match_encodeBetterBlockAsm8B + CMPL R10, SI + JEQ candidateS_match_encodeBetterBlockAsm8B + MOVL 20(SP), CX + JMP search_loop_encodeBetterBlockAsm8B + +candidateS_match_encodeBetterBlockAsm8B: + SHRQ $0x08, SI + MOVQ SI, R9 + SHLQ $0x10, R9 + IMULQ R8, R9 + SHRQ $0x36, R9 + MOVL 24(SP)(R9*4), BX + INCL CX + MOVL CX, 24(SP)(R9*4) + CMPL (DX)(BX*1), SI + JEQ candidate_match_encodeBetterBlockAsm8B + DECL CX + MOVL DI, BX + +candidate_match_encodeBetterBlockAsm8B: + MOVL 12(SP), SI + TESTL BX, BX + JZ match_extend_back_end_encodeBetterBlockAsm8B + +match_extend_back_loop_encodeBetterBlockAsm8B: + CMPL CX, SI + JBE match_extend_back_end_encodeBetterBlockAsm8B + MOVB -1(DX)(BX*1), DI + MOVB -1(DX)(CX*1), R8 + CMPB DI, R8 + JNE match_extend_back_end_encodeBetterBlockAsm8B + LEAL -1(CX), CX + DECL BX + JZ match_extend_back_end_encodeBetterBlockAsm8B + JMP match_extend_back_loop_encodeBetterBlockAsm8B + +match_extend_back_end_encodeBetterBlockAsm8B: + MOVL CX, SI + SUBL 12(SP), SI + LEAQ 3(AX)(SI*1), SI + CMPQ SI, (SP) + JB match_dst_size_check_encodeBetterBlockAsm8B + MOVQ $0x00000000, ret+48(FP) + RET + +match_dst_size_check_encodeBetterBlockAsm8B: + MOVL CX, SI + ADDL $0x04, CX + ADDL $0x04, BX + MOVQ src_len+32(FP), DI + SUBL CX, DI + LEAQ (DX)(CX*1), R8 + LEAQ (DX)(BX*1), R9 + + // matchLen + XORL R11, R11 + +matchlen_loopback_16_match_nolit_encodeBetterBlockAsm8B: + CMPL DI, $0x10 + JB matchlen_match8_match_nolit_encodeBetterBlockAsm8B + MOVQ (R8)(R11*1), R10 + MOVQ 8(R8)(R11*1), R12 + XORQ (R9)(R11*1), R10 + JNZ matchlen_bsf_8_match_nolit_encodeBetterBlockAsm8B + XORQ 8(R9)(R11*1), R12 + JNZ matchlen_bsf_16match_nolit_encodeBetterBlockAsm8B + LEAL -16(DI), DI + LEAL 16(R11), R11 + JMP matchlen_loopback_16_match_nolit_encodeBetterBlockAsm8B + +matchlen_bsf_16match_nolit_encodeBetterBlockAsm8B: +#ifdef GOAMD64_v3 + TZCNTQ R12, R12 + +#else + BSFQ R12, R12 + +#endif + SARQ $0x03, R12 + LEAL 8(R11)(R12*1), R11 + JMP match_nolit_end_encodeBetterBlockAsm8B + +matchlen_match8_match_nolit_encodeBetterBlockAsm8B: + CMPL DI, $0x08 + JB matchlen_match4_match_nolit_encodeBetterBlockAsm8B + MOVQ (R8)(R11*1), R10 + XORQ (R9)(R11*1), R10 + JNZ matchlen_bsf_8_match_nolit_encodeBetterBlockAsm8B + LEAL -8(DI), DI + LEAL 8(R11), R11 + JMP matchlen_match4_match_nolit_encodeBetterBlockAsm8B + +matchlen_bsf_8_match_nolit_encodeBetterBlockAsm8B: +#ifdef GOAMD64_v3 + TZCNTQ R10, R10 + +#else + BSFQ R10, R10 + +#endif + SARQ $0x03, R10 + LEAL (R11)(R10*1), R11 + JMP match_nolit_end_encodeBetterBlockAsm8B + +matchlen_match4_match_nolit_encodeBetterBlockAsm8B: + CMPL DI, $0x04 + JB matchlen_match2_match_nolit_encodeBetterBlockAsm8B + MOVL (R8)(R11*1), R10 + CMPL (R9)(R11*1), R10 + JNE matchlen_match2_match_nolit_encodeBetterBlockAsm8B + LEAL -4(DI), DI + LEAL 4(R11), R11 + +matchlen_match2_match_nolit_encodeBetterBlockAsm8B: + CMPL DI, $0x01 + JE matchlen_match1_match_nolit_encodeBetterBlockAsm8B + JB match_nolit_end_encodeBetterBlockAsm8B + MOVW (R8)(R11*1), R10 + CMPW (R9)(R11*1), R10 + JNE matchlen_match1_match_nolit_encodeBetterBlockAsm8B + LEAL 2(R11), R11 + SUBL $0x02, DI + JZ match_nolit_end_encodeBetterBlockAsm8B + +matchlen_match1_match_nolit_encodeBetterBlockAsm8B: + MOVB (R8)(R11*1), R10 + CMPB (R9)(R11*1), R10 + JNE match_nolit_end_encodeBetterBlockAsm8B + LEAL 1(R11), R11 + +match_nolit_end_encodeBetterBlockAsm8B: + MOVL CX, DI + SUBL BX, DI + + // Check if repeat + CMPL 16(SP), DI + JEQ match_is_repeat_encodeBetterBlockAsm8B + MOVL DI, 16(SP) + MOVL 12(SP), BX + CMPL BX, SI + JEQ emit_literal_done_match_emit_encodeBetterBlockAsm8B + MOVL SI, R8 + MOVL SI, 12(SP) + LEAQ (DX)(BX*1), R9 + SUBL BX, R8 + LEAL -1(R8), BX + CMPL BX, $0x3c + JB one_byte_match_emit_encodeBetterBlockAsm8B + CMPL BX, $0x00000100 + JB two_bytes_match_emit_encodeBetterBlockAsm8B + JB three_bytes_match_emit_encodeBetterBlockAsm8B + +three_bytes_match_emit_encodeBetterBlockAsm8B: + MOVB $0xf4, (AX) + MOVW BX, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_match_emit_encodeBetterBlockAsm8B + +two_bytes_match_emit_encodeBetterBlockAsm8B: + MOVB $0xf0, (AX) + MOVB BL, 1(AX) + ADDQ $0x02, AX + CMPL BX, $0x40 + JB memmove_match_emit_encodeBetterBlockAsm8B + JMP memmove_long_match_emit_encodeBetterBlockAsm8B + +one_byte_match_emit_encodeBetterBlockAsm8B: + SHLB $0x02, BL + MOVB BL, (AX) + ADDQ $0x01, AX + +memmove_match_emit_encodeBetterBlockAsm8B: + LEAQ (AX)(R8*1), BX + + // genMemMoveShort + CMPQ R8, $0x04 + JBE emit_lit_memmove_match_emit_encodeBetterBlockAsm8B_memmove_move_4 + CMPQ R8, $0x08 + JB emit_lit_memmove_match_emit_encodeBetterBlockAsm8B_memmove_move_4through7 + CMPQ R8, $0x10 + JBE emit_lit_memmove_match_emit_encodeBetterBlockAsm8B_memmove_move_8through16 + CMPQ R8, $0x20 + JBE emit_lit_memmove_match_emit_encodeBetterBlockAsm8B_memmove_move_17through32 + JMP emit_lit_memmove_match_emit_encodeBetterBlockAsm8B_memmove_move_33through64 + +emit_lit_memmove_match_emit_encodeBetterBlockAsm8B_memmove_move_4: + MOVL (R9), R10 + MOVL R10, (AX) + JMP memmove_end_copy_match_emit_encodeBetterBlockAsm8B + +emit_lit_memmove_match_emit_encodeBetterBlockAsm8B_memmove_move_4through7: + MOVL (R9), R10 + MOVL -4(R9)(R8*1), R9 + MOVL R10, (AX) + MOVL R9, -4(AX)(R8*1) + JMP memmove_end_copy_match_emit_encodeBetterBlockAsm8B + +emit_lit_memmove_match_emit_encodeBetterBlockAsm8B_memmove_move_8through16: + MOVQ (R9), R10 + MOVQ -8(R9)(R8*1), R9 + MOVQ R10, (AX) + MOVQ R9, -8(AX)(R8*1) + JMP memmove_end_copy_match_emit_encodeBetterBlockAsm8B + +emit_lit_memmove_match_emit_encodeBetterBlockAsm8B_memmove_move_17through32: + MOVOU (R9), X0 + MOVOU -16(R9)(R8*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(R8*1) + JMP memmove_end_copy_match_emit_encodeBetterBlockAsm8B + +emit_lit_memmove_match_emit_encodeBetterBlockAsm8B_memmove_move_33through64: + MOVOU (R9), X0 + MOVOU 16(R9), X1 + MOVOU -32(R9)(R8*1), X2 + MOVOU -16(R9)(R8*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + +memmove_end_copy_match_emit_encodeBetterBlockAsm8B: + MOVQ BX, AX + JMP emit_literal_done_match_emit_encodeBetterBlockAsm8B + +memmove_long_match_emit_encodeBetterBlockAsm8B: + LEAQ (AX)(R8*1), BX + + // genMemMoveLong + MOVOU (R9), X0 + MOVOU 16(R9), X1 + MOVOU -32(R9)(R8*1), X2 + MOVOU -16(R9)(R8*1), X3 + MOVQ R8, R12 + SHRQ $0x05, R12 + MOVQ AX, R10 + ANDL $0x0000001f, R10 + MOVQ $0x00000040, R13 + SUBQ R10, R13 + DECQ R12 + JA emit_lit_memmove_long_match_emit_encodeBetterBlockAsm8Blarge_forward_sse_loop_32 + LEAQ -32(R9)(R13*1), R10 + LEAQ -32(AX)(R13*1), R14 + +emit_lit_memmove_long_match_emit_encodeBetterBlockAsm8Blarge_big_loop_back: + MOVOU (R10), X4 + MOVOU 16(R10), X5 + MOVOA X4, (R14) + MOVOA X5, 16(R14) + ADDQ $0x20, R14 + ADDQ $0x20, R10 + ADDQ $0x20, R13 + DECQ R12 + JNA emit_lit_memmove_long_match_emit_encodeBetterBlockAsm8Blarge_big_loop_back + +emit_lit_memmove_long_match_emit_encodeBetterBlockAsm8Blarge_forward_sse_loop_32: + MOVOU -32(R9)(R13*1), X4 + MOVOU -16(R9)(R13*1), X5 + MOVOA X4, -32(AX)(R13*1) + MOVOA X5, -16(AX)(R13*1) + ADDQ $0x20, R13 + CMPQ R8, R13 + JAE emit_lit_memmove_long_match_emit_encodeBetterBlockAsm8Blarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + MOVQ BX, AX + +emit_literal_done_match_emit_encodeBetterBlockAsm8B: + ADDL R11, CX + ADDL $0x04, R11 + MOVL CX, 12(SP) + + // emitCopy + CMPL R11, $0x40 + JBE two_byte_offset_short_match_nolit_encodeBetterBlockAsm8B + CMPL DI, $0x00000800 + JAE long_offset_short_match_nolit_encodeBetterBlockAsm8B + MOVL $0x00000001, BX + LEAL 16(BX), BX + MOVB DI, 1(AX) + SHRL $0x08, DI + SHLL $0x05, DI + ORL DI, BX + MOVB BL, (AX) + ADDQ $0x02, AX + SUBL $0x08, R11 + + // emitRepeat + LEAL -4(R11), R11 + JMP cant_repeat_two_offset_match_nolit_encodeBetterBlockAsm8B_emit_copy_short_2b + MOVL R11, BX + LEAL -4(R11), R11 + CMPL BX, $0x08 + JBE repeat_two_match_nolit_encodeBetterBlockAsm8B_emit_copy_short_2b + CMPL BX, $0x0c + JAE cant_repeat_two_offset_match_nolit_encodeBetterBlockAsm8B_emit_copy_short_2b + +cant_repeat_two_offset_match_nolit_encodeBetterBlockAsm8B_emit_copy_short_2b: + CMPL R11, $0x00000104 + JB repeat_three_match_nolit_encodeBetterBlockAsm8B_emit_copy_short_2b + LEAL -256(R11), R11 + MOVW $0x0019, (AX) + MOVW R11, 2(AX) + ADDQ $0x04, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm8B + +repeat_three_match_nolit_encodeBetterBlockAsm8B_emit_copy_short_2b: + LEAL -4(R11), R11 + MOVW $0x0015, (AX) + MOVB R11, 2(AX) + ADDQ $0x03, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm8B + +repeat_two_match_nolit_encodeBetterBlockAsm8B_emit_copy_short_2b: + SHLL $0x02, R11 + ORL $0x01, R11 + MOVW R11, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm8B + XORQ BX, BX + LEAL 1(BX)(R11*4), R11 + MOVB DI, 1(AX) + SARL $0x08, DI + SHLL $0x05, DI + ORL DI, R11 + MOVB R11, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm8B + +long_offset_short_match_nolit_encodeBetterBlockAsm8B: + MOVB $0xee, (AX) + MOVW DI, 1(AX) + LEAL -60(R11), R11 + ADDQ $0x03, AX + + // emitRepeat + MOVL R11, BX + LEAL -4(R11), R11 + CMPL BX, $0x08 + JBE repeat_two_match_nolit_encodeBetterBlockAsm8B_emit_copy_short + CMPL BX, $0x0c + JAE cant_repeat_two_offset_match_nolit_encodeBetterBlockAsm8B_emit_copy_short + +cant_repeat_two_offset_match_nolit_encodeBetterBlockAsm8B_emit_copy_short: + CMPL R11, $0x00000104 + JB repeat_three_match_nolit_encodeBetterBlockAsm8B_emit_copy_short + LEAL -256(R11), R11 + MOVW $0x0019, (AX) + MOVW R11, 2(AX) + ADDQ $0x04, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm8B + +repeat_three_match_nolit_encodeBetterBlockAsm8B_emit_copy_short: + LEAL -4(R11), R11 + MOVW $0x0015, (AX) + MOVB R11, 2(AX) + ADDQ $0x03, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm8B + +repeat_two_match_nolit_encodeBetterBlockAsm8B_emit_copy_short: + SHLL $0x02, R11 + ORL $0x01, R11 + MOVW R11, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm8B + XORQ BX, BX + LEAL 1(BX)(R11*4), R11 + MOVB DI, 1(AX) + SARL $0x08, DI + SHLL $0x05, DI + ORL DI, R11 + MOVB R11, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm8B + +two_byte_offset_short_match_nolit_encodeBetterBlockAsm8B: + MOVL R11, BX + SHLL $0x02, BX + CMPL R11, $0x0c + JAE emit_copy_three_match_nolit_encodeBetterBlockAsm8B + LEAL -15(BX), BX + MOVB DI, 1(AX) + SHRL $0x08, DI + SHLL $0x05, DI + ORL DI, BX + MOVB BL, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm8B + +emit_copy_three_match_nolit_encodeBetterBlockAsm8B: + LEAL -2(BX), BX + MOVB BL, (AX) + MOVW DI, 1(AX) + ADDQ $0x03, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm8B + +match_is_repeat_encodeBetterBlockAsm8B: + MOVL 12(SP), BX + CMPL BX, SI + JEQ emit_literal_done_match_emit_repeat_encodeBetterBlockAsm8B + MOVL SI, DI + MOVL SI, 12(SP) + LEAQ (DX)(BX*1), R8 + SUBL BX, DI + LEAL -1(DI), BX + CMPL BX, $0x3c + JB one_byte_match_emit_repeat_encodeBetterBlockAsm8B + CMPL BX, $0x00000100 + JB two_bytes_match_emit_repeat_encodeBetterBlockAsm8B + JB three_bytes_match_emit_repeat_encodeBetterBlockAsm8B + +three_bytes_match_emit_repeat_encodeBetterBlockAsm8B: + MOVB $0xf4, (AX) + MOVW BX, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_match_emit_repeat_encodeBetterBlockAsm8B + +two_bytes_match_emit_repeat_encodeBetterBlockAsm8B: + MOVB $0xf0, (AX) + MOVB BL, 1(AX) + ADDQ $0x02, AX + CMPL BX, $0x40 + JB memmove_match_emit_repeat_encodeBetterBlockAsm8B + JMP memmove_long_match_emit_repeat_encodeBetterBlockAsm8B + +one_byte_match_emit_repeat_encodeBetterBlockAsm8B: + SHLB $0x02, BL + MOVB BL, (AX) + ADDQ $0x01, AX + +memmove_match_emit_repeat_encodeBetterBlockAsm8B: + LEAQ (AX)(DI*1), BX + + // genMemMoveShort + CMPQ DI, $0x04 + JBE emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm8B_memmove_move_4 + CMPQ DI, $0x08 + JB emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm8B_memmove_move_4through7 + CMPQ DI, $0x10 + JBE emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm8B_memmove_move_8through16 + CMPQ DI, $0x20 + JBE emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm8B_memmove_move_17through32 + JMP emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm8B_memmove_move_33through64 + +emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm8B_memmove_move_4: + MOVL (R8), R9 + MOVL R9, (AX) + JMP memmove_end_copy_match_emit_repeat_encodeBetterBlockAsm8B + +emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm8B_memmove_move_4through7: + MOVL (R8), R9 + MOVL -4(R8)(DI*1), R8 + MOVL R9, (AX) + MOVL R8, -4(AX)(DI*1) + JMP memmove_end_copy_match_emit_repeat_encodeBetterBlockAsm8B + +emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm8B_memmove_move_8through16: + MOVQ (R8), R9 + MOVQ -8(R8)(DI*1), R8 + MOVQ R9, (AX) + MOVQ R8, -8(AX)(DI*1) + JMP memmove_end_copy_match_emit_repeat_encodeBetterBlockAsm8B + +emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm8B_memmove_move_17through32: + MOVOU (R8), X0 + MOVOU -16(R8)(DI*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(DI*1) + JMP memmove_end_copy_match_emit_repeat_encodeBetterBlockAsm8B + +emit_lit_memmove_match_emit_repeat_encodeBetterBlockAsm8B_memmove_move_33through64: + MOVOU (R8), X0 + MOVOU 16(R8), X1 + MOVOU -32(R8)(DI*1), X2 + MOVOU -16(R8)(DI*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(DI*1) + MOVOU X3, -16(AX)(DI*1) + +memmove_end_copy_match_emit_repeat_encodeBetterBlockAsm8B: + MOVQ BX, AX + JMP emit_literal_done_match_emit_repeat_encodeBetterBlockAsm8B + +memmove_long_match_emit_repeat_encodeBetterBlockAsm8B: + LEAQ (AX)(DI*1), BX + + // genMemMoveLong + MOVOU (R8), X0 + MOVOU 16(R8), X1 + MOVOU -32(R8)(DI*1), X2 + MOVOU -16(R8)(DI*1), X3 + MOVQ DI, R10 + SHRQ $0x05, R10 + MOVQ AX, R9 + ANDL $0x0000001f, R9 + MOVQ $0x00000040, R12 + SUBQ R9, R12 + DECQ R10 + JA emit_lit_memmove_long_match_emit_repeat_encodeBetterBlockAsm8Blarge_forward_sse_loop_32 + LEAQ -32(R8)(R12*1), R9 + LEAQ -32(AX)(R12*1), R13 + +emit_lit_memmove_long_match_emit_repeat_encodeBetterBlockAsm8Blarge_big_loop_back: + MOVOU (R9), X4 + MOVOU 16(R9), X5 + MOVOA X4, (R13) + MOVOA X5, 16(R13) + ADDQ $0x20, R13 + ADDQ $0x20, R9 + ADDQ $0x20, R12 + DECQ R10 + JNA emit_lit_memmove_long_match_emit_repeat_encodeBetterBlockAsm8Blarge_big_loop_back + +emit_lit_memmove_long_match_emit_repeat_encodeBetterBlockAsm8Blarge_forward_sse_loop_32: + MOVOU -32(R8)(R12*1), X4 + MOVOU -16(R8)(R12*1), X5 + MOVOA X4, -32(AX)(R12*1) + MOVOA X5, -16(AX)(R12*1) + ADDQ $0x20, R12 + CMPQ DI, R12 + JAE emit_lit_memmove_long_match_emit_repeat_encodeBetterBlockAsm8Blarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(DI*1) + MOVOU X3, -16(AX)(DI*1) + MOVQ BX, AX + +emit_literal_done_match_emit_repeat_encodeBetterBlockAsm8B: + ADDL R11, CX + ADDL $0x04, R11 + MOVL CX, 12(SP) + + // emitRepeat + MOVL R11, BX + LEAL -4(R11), R11 + CMPL BX, $0x08 + JBE repeat_two_match_nolit_repeat_encodeBetterBlockAsm8B + CMPL BX, $0x0c + JAE cant_repeat_two_offset_match_nolit_repeat_encodeBetterBlockAsm8B + +cant_repeat_two_offset_match_nolit_repeat_encodeBetterBlockAsm8B: + CMPL R11, $0x00000104 + JB repeat_three_match_nolit_repeat_encodeBetterBlockAsm8B + LEAL -256(R11), R11 + MOVW $0x0019, (AX) + MOVW R11, 2(AX) + ADDQ $0x04, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm8B + +repeat_three_match_nolit_repeat_encodeBetterBlockAsm8B: + LEAL -4(R11), R11 + MOVW $0x0015, (AX) + MOVB R11, 2(AX) + ADDQ $0x03, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm8B + +repeat_two_match_nolit_repeat_encodeBetterBlockAsm8B: + SHLL $0x02, R11 + ORL $0x01, R11 + MOVW R11, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeBetterBlockAsm8B + XORQ BX, BX + LEAL 1(BX)(R11*4), R11 + MOVB DI, 1(AX) + SARL $0x08, DI + SHLL $0x05, DI + ORL DI, R11 + MOVB R11, (AX) + ADDQ $0x02, AX + +match_nolit_emitcopy_end_encodeBetterBlockAsm8B: + CMPL CX, 8(SP) + JAE emit_remainder_encodeBetterBlockAsm8B + CMPQ AX, (SP) + JB match_nolit_dst_ok_encodeBetterBlockAsm8B + MOVQ $0x00000000, ret+48(FP) + RET + +match_nolit_dst_ok_encodeBetterBlockAsm8B: + MOVQ $0x0000cf1bbcdcbf9b, BX + MOVQ $0x9e3779b1, DI + LEAQ 1(SI), SI + LEAQ -2(CX), R8 + MOVQ (DX)(SI*1), R9 + MOVQ 1(DX)(SI*1), R10 + MOVQ (DX)(R8*1), R11 + MOVQ 1(DX)(R8*1), R12 + SHLQ $0x10, R9 + IMULQ BX, R9 + SHRQ $0x36, R9 + SHLQ $0x20, R10 + IMULQ DI, R10 + SHRQ $0x38, R10 + SHLQ $0x10, R11 + IMULQ BX, R11 + SHRQ $0x36, R11 + SHLQ $0x20, R12 + IMULQ DI, R12 + SHRQ $0x38, R12 + LEAQ 1(SI), DI + LEAQ 1(R8), R13 + MOVL SI, 24(SP)(R9*4) + MOVL R8, 24(SP)(R11*4) + MOVL DI, 4120(SP)(R10*4) + MOVL R13, 4120(SP)(R12*4) + LEAQ 1(R8)(SI*1), DI + SHRQ $0x01, DI + ADDQ $0x01, SI + SUBQ $0x01, R8 + +index_loop_encodeBetterBlockAsm8B: + CMPQ DI, R8 + JAE search_loop_encodeBetterBlockAsm8B + MOVQ (DX)(SI*1), R9 + MOVQ (DX)(DI*1), R10 + SHLQ $0x10, R9 + IMULQ BX, R9 + SHRQ $0x36, R9 + SHLQ $0x10, R10 + IMULQ BX, R10 + SHRQ $0x36, R10 + MOVL SI, 24(SP)(R9*4) + MOVL DI, 24(SP)(R10*4) + ADDQ $0x02, SI + ADDQ $0x02, DI + JMP index_loop_encodeBetterBlockAsm8B + +emit_remainder_encodeBetterBlockAsm8B: + MOVQ src_len+32(FP), CX + SUBL 12(SP), CX + LEAQ 3(AX)(CX*1), CX + CMPQ CX, (SP) + JB emit_remainder_ok_encodeBetterBlockAsm8B + MOVQ $0x00000000, ret+48(FP) + RET + +emit_remainder_ok_encodeBetterBlockAsm8B: + MOVQ src_len+32(FP), CX + MOVL 12(SP), BX + CMPL BX, CX + JEQ emit_literal_done_emit_remainder_encodeBetterBlockAsm8B + MOVL CX, SI + MOVL CX, 12(SP) + LEAQ (DX)(BX*1), CX + SUBL BX, SI + LEAL -1(SI), DX + CMPL DX, $0x3c + JB one_byte_emit_remainder_encodeBetterBlockAsm8B + CMPL DX, $0x00000100 + JB two_bytes_emit_remainder_encodeBetterBlockAsm8B + JB three_bytes_emit_remainder_encodeBetterBlockAsm8B + +three_bytes_emit_remainder_encodeBetterBlockAsm8B: + MOVB $0xf4, (AX) + MOVW DX, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_emit_remainder_encodeBetterBlockAsm8B + +two_bytes_emit_remainder_encodeBetterBlockAsm8B: + MOVB $0xf0, (AX) + MOVB DL, 1(AX) + ADDQ $0x02, AX + CMPL DX, $0x40 + JB memmove_emit_remainder_encodeBetterBlockAsm8B + JMP memmove_long_emit_remainder_encodeBetterBlockAsm8B + +one_byte_emit_remainder_encodeBetterBlockAsm8B: + SHLB $0x02, DL + MOVB DL, (AX) + ADDQ $0x01, AX + +memmove_emit_remainder_encodeBetterBlockAsm8B: + LEAQ (AX)(SI*1), DX + MOVL SI, BX + + // genMemMoveShort + CMPQ BX, $0x03 + JB emit_lit_memmove_emit_remainder_encodeBetterBlockAsm8B_memmove_move_1or2 + JE emit_lit_memmove_emit_remainder_encodeBetterBlockAsm8B_memmove_move_3 + CMPQ BX, $0x08 + JB emit_lit_memmove_emit_remainder_encodeBetterBlockAsm8B_memmove_move_4through7 + CMPQ BX, $0x10 + JBE emit_lit_memmove_emit_remainder_encodeBetterBlockAsm8B_memmove_move_8through16 + CMPQ BX, $0x20 + JBE emit_lit_memmove_emit_remainder_encodeBetterBlockAsm8B_memmove_move_17through32 + JMP emit_lit_memmove_emit_remainder_encodeBetterBlockAsm8B_memmove_move_33through64 + +emit_lit_memmove_emit_remainder_encodeBetterBlockAsm8B_memmove_move_1or2: + MOVB (CX), SI + MOVB -1(CX)(BX*1), CL + MOVB SI, (AX) + MOVB CL, -1(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeBetterBlockAsm8B + +emit_lit_memmove_emit_remainder_encodeBetterBlockAsm8B_memmove_move_3: + MOVW (CX), SI + MOVB 2(CX), CL + MOVW SI, (AX) + MOVB CL, 2(AX) + JMP memmove_end_copy_emit_remainder_encodeBetterBlockAsm8B + +emit_lit_memmove_emit_remainder_encodeBetterBlockAsm8B_memmove_move_4through7: + MOVL (CX), SI + MOVL -4(CX)(BX*1), CX + MOVL SI, (AX) + MOVL CX, -4(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeBetterBlockAsm8B + +emit_lit_memmove_emit_remainder_encodeBetterBlockAsm8B_memmove_move_8through16: + MOVQ (CX), SI + MOVQ -8(CX)(BX*1), CX + MOVQ SI, (AX) + MOVQ CX, -8(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeBetterBlockAsm8B + +emit_lit_memmove_emit_remainder_encodeBetterBlockAsm8B_memmove_move_17through32: + MOVOU (CX), X0 + MOVOU -16(CX)(BX*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeBetterBlockAsm8B + +emit_lit_memmove_emit_remainder_encodeBetterBlockAsm8B_memmove_move_33through64: + MOVOU (CX), X0 + MOVOU 16(CX), X1 + MOVOU -32(CX)(BX*1), X2 + MOVOU -16(CX)(BX*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(BX*1) + MOVOU X3, -16(AX)(BX*1) + +memmove_end_copy_emit_remainder_encodeBetterBlockAsm8B: + MOVQ DX, AX + JMP emit_literal_done_emit_remainder_encodeBetterBlockAsm8B + +memmove_long_emit_remainder_encodeBetterBlockAsm8B: + LEAQ (AX)(SI*1), DX + MOVL SI, BX + + // genMemMoveLong + MOVOU (CX), X0 + MOVOU 16(CX), X1 + MOVOU -32(CX)(BX*1), X2 + MOVOU -16(CX)(BX*1), X3 + MOVQ BX, DI + SHRQ $0x05, DI + MOVQ AX, SI + ANDL $0x0000001f, SI + MOVQ $0x00000040, R8 + SUBQ SI, R8 + DECQ DI + JA emit_lit_memmove_long_emit_remainder_encodeBetterBlockAsm8Blarge_forward_sse_loop_32 + LEAQ -32(CX)(R8*1), SI + LEAQ -32(AX)(R8*1), R9 + +emit_lit_memmove_long_emit_remainder_encodeBetterBlockAsm8Blarge_big_loop_back: + MOVOU (SI), X4 + MOVOU 16(SI), X5 + MOVOA X4, (R9) + MOVOA X5, 16(R9) + ADDQ $0x20, R9 + ADDQ $0x20, SI + ADDQ $0x20, R8 + DECQ DI + JNA emit_lit_memmove_long_emit_remainder_encodeBetterBlockAsm8Blarge_big_loop_back + +emit_lit_memmove_long_emit_remainder_encodeBetterBlockAsm8Blarge_forward_sse_loop_32: + MOVOU -32(CX)(R8*1), X4 + MOVOU -16(CX)(R8*1), X5 + MOVOA X4, -32(AX)(R8*1) + MOVOA X5, -16(AX)(R8*1) + ADDQ $0x20, R8 + CMPQ BX, R8 + JAE emit_lit_memmove_long_emit_remainder_encodeBetterBlockAsm8Blarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(BX*1) + MOVOU X3, -16(AX)(BX*1) + MOVQ DX, AX + +emit_literal_done_emit_remainder_encodeBetterBlockAsm8B: + MOVQ dst_base+0(FP), CX + SUBQ CX, AX + MOVQ AX, ret+48(FP) + RET + +// func encodeSnappyBlockAsm(dst []byte, src []byte) int +// Requires: BMI, SSE2 +TEXT ·encodeSnappyBlockAsm(SB), $65560-56 + MOVQ dst_base+0(FP), AX + MOVQ $0x00000200, CX + LEAQ 24(SP), DX + PXOR X0, X0 + +zero_loop_encodeSnappyBlockAsm: + MOVOU X0, (DX) + MOVOU X0, 16(DX) + MOVOU X0, 32(DX) + MOVOU X0, 48(DX) + MOVOU X0, 64(DX) + MOVOU X0, 80(DX) + MOVOU X0, 96(DX) + MOVOU X0, 112(DX) + ADDQ $0x80, DX + DECQ CX + JNZ zero_loop_encodeSnappyBlockAsm + MOVL $0x00000000, 12(SP) + MOVQ src_len+32(FP), CX + LEAQ -9(CX), DX + LEAQ -8(CX), BX + MOVL BX, 8(SP) + SHRQ $0x05, CX + SUBL CX, DX + LEAQ (AX)(DX*1), DX + MOVQ DX, (SP) + MOVL $0x00000001, CX + MOVL CX, 16(SP) + MOVQ src_base+24(FP), DX + +search_loop_encodeSnappyBlockAsm: + MOVL CX, BX + SUBL 12(SP), BX + SHRL $0x06, BX + LEAL 4(CX)(BX*1), BX + CMPL BX, 8(SP) + JAE emit_remainder_encodeSnappyBlockAsm + MOVQ (DX)(CX*1), SI + MOVL BX, 20(SP) + MOVQ $0x0000cf1bbcdcbf9b, R8 + MOVQ SI, R9 + MOVQ SI, R10 + SHRQ $0x08, R10 + SHLQ $0x10, R9 + IMULQ R8, R9 + SHRQ $0x32, R9 + SHLQ $0x10, R10 + IMULQ R8, R10 + SHRQ $0x32, R10 + MOVL 24(SP)(R9*4), BX + MOVL 24(SP)(R10*4), DI + MOVL CX, 24(SP)(R9*4) + LEAL 1(CX), R9 + MOVL R9, 24(SP)(R10*4) + MOVQ SI, R9 + SHRQ $0x10, R9 + SHLQ $0x10, R9 + IMULQ R8, R9 + SHRQ $0x32, R9 + MOVL CX, R8 + SUBL 16(SP), R8 + MOVL 1(DX)(R8*1), R10 + MOVQ SI, R8 + SHRQ $0x08, R8 + CMPL R8, R10 + JNE no_repeat_found_encodeSnappyBlockAsm + LEAL 1(CX), SI + MOVL 12(SP), BX + MOVL SI, DI + SUBL 16(SP), DI + JZ repeat_extend_back_end_encodeSnappyBlockAsm + +repeat_extend_back_loop_encodeSnappyBlockAsm: + CMPL SI, BX + JBE repeat_extend_back_end_encodeSnappyBlockAsm + MOVB -1(DX)(DI*1), R8 + MOVB -1(DX)(SI*1), R9 + CMPB R8, R9 + JNE repeat_extend_back_end_encodeSnappyBlockAsm + LEAL -1(SI), SI + DECL DI + JNZ repeat_extend_back_loop_encodeSnappyBlockAsm + +repeat_extend_back_end_encodeSnappyBlockAsm: + MOVL SI, BX + SUBL 12(SP), BX + LEAQ 5(AX)(BX*1), BX + CMPQ BX, (SP) + JB repeat_dst_size_check_encodeSnappyBlockAsm + MOVQ $0x00000000, ret+48(FP) + RET + +repeat_dst_size_check_encodeSnappyBlockAsm: + MOVL 12(SP), BX + CMPL BX, SI + JEQ emit_literal_done_repeat_emit_encodeSnappyBlockAsm + MOVL SI, DI + MOVL SI, 12(SP) + LEAQ (DX)(BX*1), R8 + SUBL BX, DI + LEAL -1(DI), BX + CMPL BX, $0x3c + JB one_byte_repeat_emit_encodeSnappyBlockAsm + CMPL BX, $0x00000100 + JB two_bytes_repeat_emit_encodeSnappyBlockAsm + CMPL BX, $0x00010000 + JB three_bytes_repeat_emit_encodeSnappyBlockAsm + CMPL BX, $0x01000000 + JB four_bytes_repeat_emit_encodeSnappyBlockAsm + MOVB $0xfc, (AX) + MOVL BX, 1(AX) + ADDQ $0x05, AX + JMP memmove_long_repeat_emit_encodeSnappyBlockAsm + +four_bytes_repeat_emit_encodeSnappyBlockAsm: + MOVL BX, R9 + SHRL $0x10, R9 + MOVB $0xf8, (AX) + MOVW BX, 1(AX) + MOVB R9, 3(AX) + ADDQ $0x04, AX + JMP memmove_long_repeat_emit_encodeSnappyBlockAsm + +three_bytes_repeat_emit_encodeSnappyBlockAsm: + MOVB $0xf4, (AX) + MOVW BX, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_repeat_emit_encodeSnappyBlockAsm + +two_bytes_repeat_emit_encodeSnappyBlockAsm: + MOVB $0xf0, (AX) + MOVB BL, 1(AX) + ADDQ $0x02, AX + CMPL BX, $0x40 + JB memmove_repeat_emit_encodeSnappyBlockAsm + JMP memmove_long_repeat_emit_encodeSnappyBlockAsm + +one_byte_repeat_emit_encodeSnappyBlockAsm: + SHLB $0x02, BL + MOVB BL, (AX) + ADDQ $0x01, AX + +memmove_repeat_emit_encodeSnappyBlockAsm: + LEAQ (AX)(DI*1), BX + + // genMemMoveShort + CMPQ DI, $0x08 + JBE emit_lit_memmove_repeat_emit_encodeSnappyBlockAsm_memmove_move_8 + CMPQ DI, $0x10 + JBE emit_lit_memmove_repeat_emit_encodeSnappyBlockAsm_memmove_move_8through16 + CMPQ DI, $0x20 + JBE emit_lit_memmove_repeat_emit_encodeSnappyBlockAsm_memmove_move_17through32 + JMP emit_lit_memmove_repeat_emit_encodeSnappyBlockAsm_memmove_move_33through64 + +emit_lit_memmove_repeat_emit_encodeSnappyBlockAsm_memmove_move_8: + MOVQ (R8), R9 + MOVQ R9, (AX) + JMP memmove_end_copy_repeat_emit_encodeSnappyBlockAsm + +emit_lit_memmove_repeat_emit_encodeSnappyBlockAsm_memmove_move_8through16: + MOVQ (R8), R9 + MOVQ -8(R8)(DI*1), R8 + MOVQ R9, (AX) + MOVQ R8, -8(AX)(DI*1) + JMP memmove_end_copy_repeat_emit_encodeSnappyBlockAsm + +emit_lit_memmove_repeat_emit_encodeSnappyBlockAsm_memmove_move_17through32: + MOVOU (R8), X0 + MOVOU -16(R8)(DI*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(DI*1) + JMP memmove_end_copy_repeat_emit_encodeSnappyBlockAsm + +emit_lit_memmove_repeat_emit_encodeSnappyBlockAsm_memmove_move_33through64: + MOVOU (R8), X0 + MOVOU 16(R8), X1 + MOVOU -32(R8)(DI*1), X2 + MOVOU -16(R8)(DI*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(DI*1) + MOVOU X3, -16(AX)(DI*1) + +memmove_end_copy_repeat_emit_encodeSnappyBlockAsm: + MOVQ BX, AX + JMP emit_literal_done_repeat_emit_encodeSnappyBlockAsm + +memmove_long_repeat_emit_encodeSnappyBlockAsm: + LEAQ (AX)(DI*1), BX + + // genMemMoveLong + MOVOU (R8), X0 + MOVOU 16(R8), X1 + MOVOU -32(R8)(DI*1), X2 + MOVOU -16(R8)(DI*1), X3 + MOVQ DI, R10 + SHRQ $0x05, R10 + MOVQ AX, R9 + ANDL $0x0000001f, R9 + MOVQ $0x00000040, R11 + SUBQ R9, R11 + DECQ R10 + JA emit_lit_memmove_long_repeat_emit_encodeSnappyBlockAsmlarge_forward_sse_loop_32 + LEAQ -32(R8)(R11*1), R9 + LEAQ -32(AX)(R11*1), R12 + +emit_lit_memmove_long_repeat_emit_encodeSnappyBlockAsmlarge_big_loop_back: + MOVOU (R9), X4 + MOVOU 16(R9), X5 + MOVOA X4, (R12) + MOVOA X5, 16(R12) + ADDQ $0x20, R12 + ADDQ $0x20, R9 + ADDQ $0x20, R11 + DECQ R10 + JNA emit_lit_memmove_long_repeat_emit_encodeSnappyBlockAsmlarge_big_loop_back + +emit_lit_memmove_long_repeat_emit_encodeSnappyBlockAsmlarge_forward_sse_loop_32: + MOVOU -32(R8)(R11*1), X4 + MOVOU -16(R8)(R11*1), X5 + MOVOA X4, -32(AX)(R11*1) + MOVOA X5, -16(AX)(R11*1) + ADDQ $0x20, R11 + CMPQ DI, R11 + JAE emit_lit_memmove_long_repeat_emit_encodeSnappyBlockAsmlarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(DI*1) + MOVOU X3, -16(AX)(DI*1) + MOVQ BX, AX + +emit_literal_done_repeat_emit_encodeSnappyBlockAsm: + ADDL $0x05, CX + MOVL CX, BX + SUBL 16(SP), BX + MOVQ src_len+32(FP), DI + SUBL CX, DI + LEAQ (DX)(CX*1), R8 + LEAQ (DX)(BX*1), BX + + // matchLen + XORL R10, R10 + +matchlen_loopback_16_repeat_extend_encodeSnappyBlockAsm: + CMPL DI, $0x10 + JB matchlen_match8_repeat_extend_encodeSnappyBlockAsm + MOVQ (R8)(R10*1), R9 + MOVQ 8(R8)(R10*1), R11 + XORQ (BX)(R10*1), R9 + JNZ matchlen_bsf_8_repeat_extend_encodeSnappyBlockAsm + XORQ 8(BX)(R10*1), R11 + JNZ matchlen_bsf_16repeat_extend_encodeSnappyBlockAsm + LEAL -16(DI), DI + LEAL 16(R10), R10 + JMP matchlen_loopback_16_repeat_extend_encodeSnappyBlockAsm + +matchlen_bsf_16repeat_extend_encodeSnappyBlockAsm: +#ifdef GOAMD64_v3 + TZCNTQ R11, R11 + +#else + BSFQ R11, R11 + +#endif + SARQ $0x03, R11 + LEAL 8(R10)(R11*1), R10 + JMP repeat_extend_forward_end_encodeSnappyBlockAsm + +matchlen_match8_repeat_extend_encodeSnappyBlockAsm: + CMPL DI, $0x08 + JB matchlen_match4_repeat_extend_encodeSnappyBlockAsm + MOVQ (R8)(R10*1), R9 + XORQ (BX)(R10*1), R9 + JNZ matchlen_bsf_8_repeat_extend_encodeSnappyBlockAsm + LEAL -8(DI), DI + LEAL 8(R10), R10 + JMP matchlen_match4_repeat_extend_encodeSnappyBlockAsm + +matchlen_bsf_8_repeat_extend_encodeSnappyBlockAsm: +#ifdef GOAMD64_v3 + TZCNTQ R9, R9 + +#else + BSFQ R9, R9 + +#endif + SARQ $0x03, R9 + LEAL (R10)(R9*1), R10 + JMP repeat_extend_forward_end_encodeSnappyBlockAsm + +matchlen_match4_repeat_extend_encodeSnappyBlockAsm: + CMPL DI, $0x04 + JB matchlen_match2_repeat_extend_encodeSnappyBlockAsm + MOVL (R8)(R10*1), R9 + CMPL (BX)(R10*1), R9 + JNE matchlen_match2_repeat_extend_encodeSnappyBlockAsm + LEAL -4(DI), DI + LEAL 4(R10), R10 + +matchlen_match2_repeat_extend_encodeSnappyBlockAsm: + CMPL DI, $0x01 + JE matchlen_match1_repeat_extend_encodeSnappyBlockAsm + JB repeat_extend_forward_end_encodeSnappyBlockAsm + MOVW (R8)(R10*1), R9 + CMPW (BX)(R10*1), R9 + JNE matchlen_match1_repeat_extend_encodeSnappyBlockAsm + LEAL 2(R10), R10 + SUBL $0x02, DI + JZ repeat_extend_forward_end_encodeSnappyBlockAsm + +matchlen_match1_repeat_extend_encodeSnappyBlockAsm: + MOVB (R8)(R10*1), R9 + CMPB (BX)(R10*1), R9 + JNE repeat_extend_forward_end_encodeSnappyBlockAsm + LEAL 1(R10), R10 + +repeat_extend_forward_end_encodeSnappyBlockAsm: + ADDL R10, CX + MOVL CX, BX + SUBL SI, BX + MOVL 16(SP), SI + + // emitCopy + CMPL SI, $0x00010000 + JB two_byte_offset_repeat_as_copy_encodeSnappyBlockAsm + +four_bytes_loop_back_repeat_as_copy_encodeSnappyBlockAsm: + CMPL BX, $0x40 + JBE four_bytes_remain_repeat_as_copy_encodeSnappyBlockAsm + MOVB $0xff, (AX) + MOVL SI, 1(AX) + LEAL -64(BX), BX + ADDQ $0x05, AX + CMPL BX, $0x04 + JB four_bytes_remain_repeat_as_copy_encodeSnappyBlockAsm + JMP four_bytes_loop_back_repeat_as_copy_encodeSnappyBlockAsm + +four_bytes_remain_repeat_as_copy_encodeSnappyBlockAsm: + TESTL BX, BX + JZ repeat_end_emit_encodeSnappyBlockAsm + XORL DI, DI + LEAL -1(DI)(BX*4), BX + MOVB BL, (AX) + MOVL SI, 1(AX) + ADDQ $0x05, AX + JMP repeat_end_emit_encodeSnappyBlockAsm + +two_byte_offset_repeat_as_copy_encodeSnappyBlockAsm: + CMPL BX, $0x40 + JBE two_byte_offset_short_repeat_as_copy_encodeSnappyBlockAsm + MOVB $0xee, (AX) + MOVW SI, 1(AX) + LEAL -60(BX), BX + ADDQ $0x03, AX + JMP two_byte_offset_repeat_as_copy_encodeSnappyBlockAsm + +two_byte_offset_short_repeat_as_copy_encodeSnappyBlockAsm: + MOVL BX, DI + SHLL $0x02, DI + CMPL BX, $0x0c + JAE emit_copy_three_repeat_as_copy_encodeSnappyBlockAsm + CMPL SI, $0x00000800 + JAE emit_copy_three_repeat_as_copy_encodeSnappyBlockAsm + LEAL -15(DI), DI + MOVB SI, 1(AX) + SHRL $0x08, SI + SHLL $0x05, SI + ORL SI, DI + MOVB DI, (AX) + ADDQ $0x02, AX + JMP repeat_end_emit_encodeSnappyBlockAsm + +emit_copy_three_repeat_as_copy_encodeSnappyBlockAsm: + LEAL -2(DI), DI + MOVB DI, (AX) + MOVW SI, 1(AX) + ADDQ $0x03, AX + +repeat_end_emit_encodeSnappyBlockAsm: + MOVL CX, 12(SP) + JMP search_loop_encodeSnappyBlockAsm + +no_repeat_found_encodeSnappyBlockAsm: + CMPL (DX)(BX*1), SI + JEQ candidate_match_encodeSnappyBlockAsm + SHRQ $0x08, SI + MOVL 24(SP)(R9*4), BX + LEAL 2(CX), R8 + CMPL (DX)(DI*1), SI + JEQ candidate2_match_encodeSnappyBlockAsm + MOVL R8, 24(SP)(R9*4) + SHRQ $0x08, SI + CMPL (DX)(BX*1), SI + JEQ candidate3_match_encodeSnappyBlockAsm + MOVL 20(SP), CX + JMP search_loop_encodeSnappyBlockAsm + +candidate3_match_encodeSnappyBlockAsm: + ADDL $0x02, CX + JMP candidate_match_encodeSnappyBlockAsm + +candidate2_match_encodeSnappyBlockAsm: + MOVL R8, 24(SP)(R9*4) + INCL CX + MOVL DI, BX + +candidate_match_encodeSnappyBlockAsm: + MOVL 12(SP), SI + TESTL BX, BX + JZ match_extend_back_end_encodeSnappyBlockAsm + +match_extend_back_loop_encodeSnappyBlockAsm: + CMPL CX, SI + JBE match_extend_back_end_encodeSnappyBlockAsm + MOVB -1(DX)(BX*1), DI + MOVB -1(DX)(CX*1), R8 + CMPB DI, R8 + JNE match_extend_back_end_encodeSnappyBlockAsm + LEAL -1(CX), CX + DECL BX + JZ match_extend_back_end_encodeSnappyBlockAsm + JMP match_extend_back_loop_encodeSnappyBlockAsm + +match_extend_back_end_encodeSnappyBlockAsm: + MOVL CX, SI + SUBL 12(SP), SI + LEAQ 5(AX)(SI*1), SI + CMPQ SI, (SP) + JB match_dst_size_check_encodeSnappyBlockAsm + MOVQ $0x00000000, ret+48(FP) + RET + +match_dst_size_check_encodeSnappyBlockAsm: + MOVL CX, SI + MOVL 12(SP), DI + CMPL DI, SI + JEQ emit_literal_done_match_emit_encodeSnappyBlockAsm + MOVL SI, R8 + MOVL SI, 12(SP) + LEAQ (DX)(DI*1), SI + SUBL DI, R8 + LEAL -1(R8), DI + CMPL DI, $0x3c + JB one_byte_match_emit_encodeSnappyBlockAsm + CMPL DI, $0x00000100 + JB two_bytes_match_emit_encodeSnappyBlockAsm + CMPL DI, $0x00010000 + JB three_bytes_match_emit_encodeSnappyBlockAsm + CMPL DI, $0x01000000 + JB four_bytes_match_emit_encodeSnappyBlockAsm + MOVB $0xfc, (AX) + MOVL DI, 1(AX) + ADDQ $0x05, AX + JMP memmove_long_match_emit_encodeSnappyBlockAsm + +four_bytes_match_emit_encodeSnappyBlockAsm: + MOVL DI, R9 + SHRL $0x10, R9 + MOVB $0xf8, (AX) + MOVW DI, 1(AX) + MOVB R9, 3(AX) + ADDQ $0x04, AX + JMP memmove_long_match_emit_encodeSnappyBlockAsm + +three_bytes_match_emit_encodeSnappyBlockAsm: + MOVB $0xf4, (AX) + MOVW DI, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_match_emit_encodeSnappyBlockAsm + +two_bytes_match_emit_encodeSnappyBlockAsm: + MOVB $0xf0, (AX) + MOVB DI, 1(AX) + ADDQ $0x02, AX + CMPL DI, $0x40 + JB memmove_match_emit_encodeSnappyBlockAsm + JMP memmove_long_match_emit_encodeSnappyBlockAsm + +one_byte_match_emit_encodeSnappyBlockAsm: + SHLB $0x02, DI + MOVB DI, (AX) + ADDQ $0x01, AX + +memmove_match_emit_encodeSnappyBlockAsm: + LEAQ (AX)(R8*1), DI + + // genMemMoveShort + CMPQ R8, $0x08 + JBE emit_lit_memmove_match_emit_encodeSnappyBlockAsm_memmove_move_8 + CMPQ R8, $0x10 + JBE emit_lit_memmove_match_emit_encodeSnappyBlockAsm_memmove_move_8through16 + CMPQ R8, $0x20 + JBE emit_lit_memmove_match_emit_encodeSnappyBlockAsm_memmove_move_17through32 + JMP emit_lit_memmove_match_emit_encodeSnappyBlockAsm_memmove_move_33through64 + +emit_lit_memmove_match_emit_encodeSnappyBlockAsm_memmove_move_8: + MOVQ (SI), R9 + MOVQ R9, (AX) + JMP memmove_end_copy_match_emit_encodeSnappyBlockAsm + +emit_lit_memmove_match_emit_encodeSnappyBlockAsm_memmove_move_8through16: + MOVQ (SI), R9 + MOVQ -8(SI)(R8*1), SI + MOVQ R9, (AX) + MOVQ SI, -8(AX)(R8*1) + JMP memmove_end_copy_match_emit_encodeSnappyBlockAsm + +emit_lit_memmove_match_emit_encodeSnappyBlockAsm_memmove_move_17through32: + MOVOU (SI), X0 + MOVOU -16(SI)(R8*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(R8*1) + JMP memmove_end_copy_match_emit_encodeSnappyBlockAsm + +emit_lit_memmove_match_emit_encodeSnappyBlockAsm_memmove_move_33through64: + MOVOU (SI), X0 + MOVOU 16(SI), X1 + MOVOU -32(SI)(R8*1), X2 + MOVOU -16(SI)(R8*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + +memmove_end_copy_match_emit_encodeSnappyBlockAsm: + MOVQ DI, AX + JMP emit_literal_done_match_emit_encodeSnappyBlockAsm + +memmove_long_match_emit_encodeSnappyBlockAsm: + LEAQ (AX)(R8*1), DI + + // genMemMoveLong + MOVOU (SI), X0 + MOVOU 16(SI), X1 + MOVOU -32(SI)(R8*1), X2 + MOVOU -16(SI)(R8*1), X3 + MOVQ R8, R10 + SHRQ $0x05, R10 + MOVQ AX, R9 + ANDL $0x0000001f, R9 + MOVQ $0x00000040, R11 + SUBQ R9, R11 + DECQ R10 + JA emit_lit_memmove_long_match_emit_encodeSnappyBlockAsmlarge_forward_sse_loop_32 + LEAQ -32(SI)(R11*1), R9 + LEAQ -32(AX)(R11*1), R12 + +emit_lit_memmove_long_match_emit_encodeSnappyBlockAsmlarge_big_loop_back: + MOVOU (R9), X4 + MOVOU 16(R9), X5 + MOVOA X4, (R12) + MOVOA X5, 16(R12) + ADDQ $0x20, R12 + ADDQ $0x20, R9 + ADDQ $0x20, R11 + DECQ R10 + JNA emit_lit_memmove_long_match_emit_encodeSnappyBlockAsmlarge_big_loop_back + +emit_lit_memmove_long_match_emit_encodeSnappyBlockAsmlarge_forward_sse_loop_32: + MOVOU -32(SI)(R11*1), X4 + MOVOU -16(SI)(R11*1), X5 + MOVOA X4, -32(AX)(R11*1) + MOVOA X5, -16(AX)(R11*1) + ADDQ $0x20, R11 + CMPQ R8, R11 + JAE emit_lit_memmove_long_match_emit_encodeSnappyBlockAsmlarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + MOVQ DI, AX + +emit_literal_done_match_emit_encodeSnappyBlockAsm: +match_nolit_loop_encodeSnappyBlockAsm: + MOVL CX, SI + SUBL BX, SI + MOVL SI, 16(SP) + ADDL $0x04, CX + ADDL $0x04, BX + MOVQ src_len+32(FP), SI + SUBL CX, SI + LEAQ (DX)(CX*1), DI + LEAQ (DX)(BX*1), BX + + // matchLen + XORL R9, R9 + +matchlen_loopback_16_match_nolit_encodeSnappyBlockAsm: + CMPL SI, $0x10 + JB matchlen_match8_match_nolit_encodeSnappyBlockAsm + MOVQ (DI)(R9*1), R8 + MOVQ 8(DI)(R9*1), R10 + XORQ (BX)(R9*1), R8 + JNZ matchlen_bsf_8_match_nolit_encodeSnappyBlockAsm + XORQ 8(BX)(R9*1), R10 + JNZ matchlen_bsf_16match_nolit_encodeSnappyBlockAsm + LEAL -16(SI), SI + LEAL 16(R9), R9 + JMP matchlen_loopback_16_match_nolit_encodeSnappyBlockAsm + +matchlen_bsf_16match_nolit_encodeSnappyBlockAsm: +#ifdef GOAMD64_v3 + TZCNTQ R10, R10 + +#else + BSFQ R10, R10 + +#endif + SARQ $0x03, R10 + LEAL 8(R9)(R10*1), R9 + JMP match_nolit_end_encodeSnappyBlockAsm + +matchlen_match8_match_nolit_encodeSnappyBlockAsm: + CMPL SI, $0x08 + JB matchlen_match4_match_nolit_encodeSnappyBlockAsm + MOVQ (DI)(R9*1), R8 + XORQ (BX)(R9*1), R8 + JNZ matchlen_bsf_8_match_nolit_encodeSnappyBlockAsm + LEAL -8(SI), SI + LEAL 8(R9), R9 + JMP matchlen_match4_match_nolit_encodeSnappyBlockAsm + +matchlen_bsf_8_match_nolit_encodeSnappyBlockAsm: +#ifdef GOAMD64_v3 + TZCNTQ R8, R8 + +#else + BSFQ R8, R8 + +#endif + SARQ $0x03, R8 + LEAL (R9)(R8*1), R9 + JMP match_nolit_end_encodeSnappyBlockAsm + +matchlen_match4_match_nolit_encodeSnappyBlockAsm: + CMPL SI, $0x04 + JB matchlen_match2_match_nolit_encodeSnappyBlockAsm + MOVL (DI)(R9*1), R8 + CMPL (BX)(R9*1), R8 + JNE matchlen_match2_match_nolit_encodeSnappyBlockAsm + LEAL -4(SI), SI + LEAL 4(R9), R9 + +matchlen_match2_match_nolit_encodeSnappyBlockAsm: + CMPL SI, $0x01 + JE matchlen_match1_match_nolit_encodeSnappyBlockAsm + JB match_nolit_end_encodeSnappyBlockAsm + MOVW (DI)(R9*1), R8 + CMPW (BX)(R9*1), R8 + JNE matchlen_match1_match_nolit_encodeSnappyBlockAsm + LEAL 2(R9), R9 + SUBL $0x02, SI + JZ match_nolit_end_encodeSnappyBlockAsm + +matchlen_match1_match_nolit_encodeSnappyBlockAsm: + MOVB (DI)(R9*1), R8 + CMPB (BX)(R9*1), R8 + JNE match_nolit_end_encodeSnappyBlockAsm + LEAL 1(R9), R9 + +match_nolit_end_encodeSnappyBlockAsm: + ADDL R9, CX + MOVL 16(SP), BX + ADDL $0x04, R9 + MOVL CX, 12(SP) + + // emitCopy + CMPL BX, $0x00010000 + JB two_byte_offset_match_nolit_encodeSnappyBlockAsm + +four_bytes_loop_back_match_nolit_encodeSnappyBlockAsm: + CMPL R9, $0x40 + JBE four_bytes_remain_match_nolit_encodeSnappyBlockAsm + MOVB $0xff, (AX) + MOVL BX, 1(AX) + LEAL -64(R9), R9 + ADDQ $0x05, AX + CMPL R9, $0x04 + JB four_bytes_remain_match_nolit_encodeSnappyBlockAsm + JMP four_bytes_loop_back_match_nolit_encodeSnappyBlockAsm + +four_bytes_remain_match_nolit_encodeSnappyBlockAsm: + TESTL R9, R9 + JZ match_nolit_emitcopy_end_encodeSnappyBlockAsm + XORL SI, SI + LEAL -1(SI)(R9*4), R9 + MOVB R9, (AX) + MOVL BX, 1(AX) + ADDQ $0x05, AX + JMP match_nolit_emitcopy_end_encodeSnappyBlockAsm + +two_byte_offset_match_nolit_encodeSnappyBlockAsm: + CMPL R9, $0x40 + JBE two_byte_offset_short_match_nolit_encodeSnappyBlockAsm + MOVB $0xee, (AX) + MOVW BX, 1(AX) + LEAL -60(R9), R9 + ADDQ $0x03, AX + JMP two_byte_offset_match_nolit_encodeSnappyBlockAsm + +two_byte_offset_short_match_nolit_encodeSnappyBlockAsm: + MOVL R9, SI + SHLL $0x02, SI + CMPL R9, $0x0c + JAE emit_copy_three_match_nolit_encodeSnappyBlockAsm + CMPL BX, $0x00000800 + JAE emit_copy_three_match_nolit_encodeSnappyBlockAsm + LEAL -15(SI), SI + MOVB BL, 1(AX) + SHRL $0x08, BX + SHLL $0x05, BX + ORL BX, SI + MOVB SI, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeSnappyBlockAsm + +emit_copy_three_match_nolit_encodeSnappyBlockAsm: + LEAL -2(SI), SI + MOVB SI, (AX) + MOVW BX, 1(AX) + ADDQ $0x03, AX + +match_nolit_emitcopy_end_encodeSnappyBlockAsm: + CMPL CX, 8(SP) + JAE emit_remainder_encodeSnappyBlockAsm + MOVQ -2(DX)(CX*1), SI + CMPQ AX, (SP) + JB match_nolit_dst_ok_encodeSnappyBlockAsm + MOVQ $0x00000000, ret+48(FP) + RET + +match_nolit_dst_ok_encodeSnappyBlockAsm: + MOVQ $0x0000cf1bbcdcbf9b, R8 + MOVQ SI, DI + SHRQ $0x10, SI + MOVQ SI, BX + SHLQ $0x10, DI + IMULQ R8, DI + SHRQ $0x32, DI + SHLQ $0x10, BX + IMULQ R8, BX + SHRQ $0x32, BX + LEAL -2(CX), R8 + LEAQ 24(SP)(BX*4), R9 + MOVL (R9), BX + MOVL R8, 24(SP)(DI*4) + MOVL CX, (R9) + CMPL (DX)(BX*1), SI + JEQ match_nolit_loop_encodeSnappyBlockAsm + INCL CX + JMP search_loop_encodeSnappyBlockAsm + +emit_remainder_encodeSnappyBlockAsm: + MOVQ src_len+32(FP), CX + SUBL 12(SP), CX + LEAQ 5(AX)(CX*1), CX + CMPQ CX, (SP) + JB emit_remainder_ok_encodeSnappyBlockAsm + MOVQ $0x00000000, ret+48(FP) + RET + +emit_remainder_ok_encodeSnappyBlockAsm: + MOVQ src_len+32(FP), CX + MOVL 12(SP), BX + CMPL BX, CX + JEQ emit_literal_done_emit_remainder_encodeSnappyBlockAsm + MOVL CX, SI + MOVL CX, 12(SP) + LEAQ (DX)(BX*1), CX + SUBL BX, SI + LEAL -1(SI), DX + CMPL DX, $0x3c + JB one_byte_emit_remainder_encodeSnappyBlockAsm + CMPL DX, $0x00000100 + JB two_bytes_emit_remainder_encodeSnappyBlockAsm + CMPL DX, $0x00010000 + JB three_bytes_emit_remainder_encodeSnappyBlockAsm + CMPL DX, $0x01000000 + JB four_bytes_emit_remainder_encodeSnappyBlockAsm + MOVB $0xfc, (AX) + MOVL DX, 1(AX) + ADDQ $0x05, AX + JMP memmove_long_emit_remainder_encodeSnappyBlockAsm + +four_bytes_emit_remainder_encodeSnappyBlockAsm: + MOVL DX, BX + SHRL $0x10, BX + MOVB $0xf8, (AX) + MOVW DX, 1(AX) + MOVB BL, 3(AX) + ADDQ $0x04, AX + JMP memmove_long_emit_remainder_encodeSnappyBlockAsm + +three_bytes_emit_remainder_encodeSnappyBlockAsm: + MOVB $0xf4, (AX) + MOVW DX, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_emit_remainder_encodeSnappyBlockAsm + +two_bytes_emit_remainder_encodeSnappyBlockAsm: + MOVB $0xf0, (AX) + MOVB DL, 1(AX) + ADDQ $0x02, AX + CMPL DX, $0x40 + JB memmove_emit_remainder_encodeSnappyBlockAsm + JMP memmove_long_emit_remainder_encodeSnappyBlockAsm + +one_byte_emit_remainder_encodeSnappyBlockAsm: + SHLB $0x02, DL + MOVB DL, (AX) + ADDQ $0x01, AX + +memmove_emit_remainder_encodeSnappyBlockAsm: + LEAQ (AX)(SI*1), DX + MOVL SI, BX + + // genMemMoveShort + CMPQ BX, $0x03 + JB emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm_memmove_move_1or2 + JE emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm_memmove_move_3 + CMPQ BX, $0x08 + JB emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm_memmove_move_4through7 + CMPQ BX, $0x10 + JBE emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm_memmove_move_8through16 + CMPQ BX, $0x20 + JBE emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm_memmove_move_17through32 + JMP emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm_memmove_move_33through64 + +emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm_memmove_move_1or2: + MOVB (CX), SI + MOVB -1(CX)(BX*1), CL + MOVB SI, (AX) + MOVB CL, -1(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeSnappyBlockAsm + +emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm_memmove_move_3: + MOVW (CX), SI + MOVB 2(CX), CL + MOVW SI, (AX) + MOVB CL, 2(AX) + JMP memmove_end_copy_emit_remainder_encodeSnappyBlockAsm + +emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm_memmove_move_4through7: + MOVL (CX), SI + MOVL -4(CX)(BX*1), CX + MOVL SI, (AX) + MOVL CX, -4(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeSnappyBlockAsm + +emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm_memmove_move_8through16: + MOVQ (CX), SI + MOVQ -8(CX)(BX*1), CX + MOVQ SI, (AX) + MOVQ CX, -8(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeSnappyBlockAsm + +emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm_memmove_move_17through32: + MOVOU (CX), X0 + MOVOU -16(CX)(BX*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeSnappyBlockAsm + +emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm_memmove_move_33through64: + MOVOU (CX), X0 + MOVOU 16(CX), X1 + MOVOU -32(CX)(BX*1), X2 + MOVOU -16(CX)(BX*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(BX*1) + MOVOU X3, -16(AX)(BX*1) + +memmove_end_copy_emit_remainder_encodeSnappyBlockAsm: + MOVQ DX, AX + JMP emit_literal_done_emit_remainder_encodeSnappyBlockAsm + +memmove_long_emit_remainder_encodeSnappyBlockAsm: + LEAQ (AX)(SI*1), DX + MOVL SI, BX + + // genMemMoveLong + MOVOU (CX), X0 + MOVOU 16(CX), X1 + MOVOU -32(CX)(BX*1), X2 + MOVOU -16(CX)(BX*1), X3 + MOVQ BX, DI + SHRQ $0x05, DI + MOVQ AX, SI + ANDL $0x0000001f, SI + MOVQ $0x00000040, R8 + SUBQ SI, R8 + DECQ DI + JA emit_lit_memmove_long_emit_remainder_encodeSnappyBlockAsmlarge_forward_sse_loop_32 + LEAQ -32(CX)(R8*1), SI + LEAQ -32(AX)(R8*1), R9 + +emit_lit_memmove_long_emit_remainder_encodeSnappyBlockAsmlarge_big_loop_back: + MOVOU (SI), X4 + MOVOU 16(SI), X5 + MOVOA X4, (R9) + MOVOA X5, 16(R9) + ADDQ $0x20, R9 + ADDQ $0x20, SI + ADDQ $0x20, R8 + DECQ DI + JNA emit_lit_memmove_long_emit_remainder_encodeSnappyBlockAsmlarge_big_loop_back + +emit_lit_memmove_long_emit_remainder_encodeSnappyBlockAsmlarge_forward_sse_loop_32: + MOVOU -32(CX)(R8*1), X4 + MOVOU -16(CX)(R8*1), X5 + MOVOA X4, -32(AX)(R8*1) + MOVOA X5, -16(AX)(R8*1) + ADDQ $0x20, R8 + CMPQ BX, R8 + JAE emit_lit_memmove_long_emit_remainder_encodeSnappyBlockAsmlarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(BX*1) + MOVOU X3, -16(AX)(BX*1) + MOVQ DX, AX + +emit_literal_done_emit_remainder_encodeSnappyBlockAsm: + MOVQ dst_base+0(FP), CX + SUBQ CX, AX + MOVQ AX, ret+48(FP) + RET + +// func encodeSnappyBlockAsm64K(dst []byte, src []byte) int +// Requires: BMI, SSE2 +TEXT ·encodeSnappyBlockAsm64K(SB), $65560-56 + MOVQ dst_base+0(FP), AX + MOVQ $0x00000200, CX + LEAQ 24(SP), DX + PXOR X0, X0 + +zero_loop_encodeSnappyBlockAsm64K: + MOVOU X0, (DX) + MOVOU X0, 16(DX) + MOVOU X0, 32(DX) + MOVOU X0, 48(DX) + MOVOU X0, 64(DX) + MOVOU X0, 80(DX) + MOVOU X0, 96(DX) + MOVOU X0, 112(DX) + ADDQ $0x80, DX + DECQ CX + JNZ zero_loop_encodeSnappyBlockAsm64K + MOVL $0x00000000, 12(SP) + MOVQ src_len+32(FP), CX + LEAQ -9(CX), DX + LEAQ -8(CX), BX + MOVL BX, 8(SP) + SHRQ $0x05, CX + SUBL CX, DX + LEAQ (AX)(DX*1), DX + MOVQ DX, (SP) + MOVL $0x00000001, CX + MOVL CX, 16(SP) + MOVQ src_base+24(FP), DX + +search_loop_encodeSnappyBlockAsm64K: + MOVL CX, BX + SUBL 12(SP), BX + SHRL $0x06, BX + LEAL 4(CX)(BX*1), BX + CMPL BX, 8(SP) + JAE emit_remainder_encodeSnappyBlockAsm64K + MOVQ (DX)(CX*1), SI + MOVL BX, 20(SP) + MOVQ $0x0000cf1bbcdcbf9b, R8 + MOVQ SI, R9 + MOVQ SI, R10 + SHRQ $0x08, R10 + SHLQ $0x10, R9 + IMULQ R8, R9 + SHRQ $0x32, R9 + SHLQ $0x10, R10 + IMULQ R8, R10 + SHRQ $0x32, R10 + MOVL 24(SP)(R9*4), BX + MOVL 24(SP)(R10*4), DI + MOVL CX, 24(SP)(R9*4) + LEAL 1(CX), R9 + MOVL R9, 24(SP)(R10*4) + MOVQ SI, R9 + SHRQ $0x10, R9 + SHLQ $0x10, R9 + IMULQ R8, R9 + SHRQ $0x32, R9 + MOVL CX, R8 + SUBL 16(SP), R8 + MOVL 1(DX)(R8*1), R10 + MOVQ SI, R8 + SHRQ $0x08, R8 + CMPL R8, R10 + JNE no_repeat_found_encodeSnappyBlockAsm64K + LEAL 1(CX), SI + MOVL 12(SP), BX + MOVL SI, DI + SUBL 16(SP), DI + JZ repeat_extend_back_end_encodeSnappyBlockAsm64K + +repeat_extend_back_loop_encodeSnappyBlockAsm64K: + CMPL SI, BX + JBE repeat_extend_back_end_encodeSnappyBlockAsm64K + MOVB -1(DX)(DI*1), R8 + MOVB -1(DX)(SI*1), R9 + CMPB R8, R9 + JNE repeat_extend_back_end_encodeSnappyBlockAsm64K + LEAL -1(SI), SI + DECL DI + JNZ repeat_extend_back_loop_encodeSnappyBlockAsm64K + +repeat_extend_back_end_encodeSnappyBlockAsm64K: + MOVL SI, BX + SUBL 12(SP), BX + LEAQ 3(AX)(BX*1), BX + CMPQ BX, (SP) + JB repeat_dst_size_check_encodeSnappyBlockAsm64K + MOVQ $0x00000000, ret+48(FP) + RET + +repeat_dst_size_check_encodeSnappyBlockAsm64K: + MOVL 12(SP), BX + CMPL BX, SI + JEQ emit_literal_done_repeat_emit_encodeSnappyBlockAsm64K + MOVL SI, DI + MOVL SI, 12(SP) + LEAQ (DX)(BX*1), R8 + SUBL BX, DI + LEAL -1(DI), BX + CMPL BX, $0x3c + JB one_byte_repeat_emit_encodeSnappyBlockAsm64K + CMPL BX, $0x00000100 + JB two_bytes_repeat_emit_encodeSnappyBlockAsm64K + JB three_bytes_repeat_emit_encodeSnappyBlockAsm64K + +three_bytes_repeat_emit_encodeSnappyBlockAsm64K: + MOVB $0xf4, (AX) + MOVW BX, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_repeat_emit_encodeSnappyBlockAsm64K + +two_bytes_repeat_emit_encodeSnappyBlockAsm64K: + MOVB $0xf0, (AX) + MOVB BL, 1(AX) + ADDQ $0x02, AX + CMPL BX, $0x40 + JB memmove_repeat_emit_encodeSnappyBlockAsm64K + JMP memmove_long_repeat_emit_encodeSnappyBlockAsm64K + +one_byte_repeat_emit_encodeSnappyBlockAsm64K: + SHLB $0x02, BL + MOVB BL, (AX) + ADDQ $0x01, AX + +memmove_repeat_emit_encodeSnappyBlockAsm64K: + LEAQ (AX)(DI*1), BX + + // genMemMoveShort + CMPQ DI, $0x08 + JBE emit_lit_memmove_repeat_emit_encodeSnappyBlockAsm64K_memmove_move_8 + CMPQ DI, $0x10 + JBE emit_lit_memmove_repeat_emit_encodeSnappyBlockAsm64K_memmove_move_8through16 + CMPQ DI, $0x20 + JBE emit_lit_memmove_repeat_emit_encodeSnappyBlockAsm64K_memmove_move_17through32 + JMP emit_lit_memmove_repeat_emit_encodeSnappyBlockAsm64K_memmove_move_33through64 + +emit_lit_memmove_repeat_emit_encodeSnappyBlockAsm64K_memmove_move_8: + MOVQ (R8), R9 + MOVQ R9, (AX) + JMP memmove_end_copy_repeat_emit_encodeSnappyBlockAsm64K + +emit_lit_memmove_repeat_emit_encodeSnappyBlockAsm64K_memmove_move_8through16: + MOVQ (R8), R9 + MOVQ -8(R8)(DI*1), R8 + MOVQ R9, (AX) + MOVQ R8, -8(AX)(DI*1) + JMP memmove_end_copy_repeat_emit_encodeSnappyBlockAsm64K + +emit_lit_memmove_repeat_emit_encodeSnappyBlockAsm64K_memmove_move_17through32: + MOVOU (R8), X0 + MOVOU -16(R8)(DI*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(DI*1) + JMP memmove_end_copy_repeat_emit_encodeSnappyBlockAsm64K + +emit_lit_memmove_repeat_emit_encodeSnappyBlockAsm64K_memmove_move_33through64: + MOVOU (R8), X0 + MOVOU 16(R8), X1 + MOVOU -32(R8)(DI*1), X2 + MOVOU -16(R8)(DI*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(DI*1) + MOVOU X3, -16(AX)(DI*1) + +memmove_end_copy_repeat_emit_encodeSnappyBlockAsm64K: + MOVQ BX, AX + JMP emit_literal_done_repeat_emit_encodeSnappyBlockAsm64K + +memmove_long_repeat_emit_encodeSnappyBlockAsm64K: + LEAQ (AX)(DI*1), BX + + // genMemMoveLong + MOVOU (R8), X0 + MOVOU 16(R8), X1 + MOVOU -32(R8)(DI*1), X2 + MOVOU -16(R8)(DI*1), X3 + MOVQ DI, R10 + SHRQ $0x05, R10 + MOVQ AX, R9 + ANDL $0x0000001f, R9 + MOVQ $0x00000040, R11 + SUBQ R9, R11 + DECQ R10 + JA emit_lit_memmove_long_repeat_emit_encodeSnappyBlockAsm64Klarge_forward_sse_loop_32 + LEAQ -32(R8)(R11*1), R9 + LEAQ -32(AX)(R11*1), R12 + +emit_lit_memmove_long_repeat_emit_encodeSnappyBlockAsm64Klarge_big_loop_back: + MOVOU (R9), X4 + MOVOU 16(R9), X5 + MOVOA X4, (R12) + MOVOA X5, 16(R12) + ADDQ $0x20, R12 + ADDQ $0x20, R9 + ADDQ $0x20, R11 + DECQ R10 + JNA emit_lit_memmove_long_repeat_emit_encodeSnappyBlockAsm64Klarge_big_loop_back + +emit_lit_memmove_long_repeat_emit_encodeSnappyBlockAsm64Klarge_forward_sse_loop_32: + MOVOU -32(R8)(R11*1), X4 + MOVOU -16(R8)(R11*1), X5 + MOVOA X4, -32(AX)(R11*1) + MOVOA X5, -16(AX)(R11*1) + ADDQ $0x20, R11 + CMPQ DI, R11 + JAE emit_lit_memmove_long_repeat_emit_encodeSnappyBlockAsm64Klarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(DI*1) + MOVOU X3, -16(AX)(DI*1) + MOVQ BX, AX + +emit_literal_done_repeat_emit_encodeSnappyBlockAsm64K: + ADDL $0x05, CX + MOVL CX, BX + SUBL 16(SP), BX + MOVQ src_len+32(FP), DI + SUBL CX, DI + LEAQ (DX)(CX*1), R8 + LEAQ (DX)(BX*1), BX + + // matchLen + XORL R10, R10 + +matchlen_loopback_16_repeat_extend_encodeSnappyBlockAsm64K: + CMPL DI, $0x10 + JB matchlen_match8_repeat_extend_encodeSnappyBlockAsm64K + MOVQ (R8)(R10*1), R9 + MOVQ 8(R8)(R10*1), R11 + XORQ (BX)(R10*1), R9 + JNZ matchlen_bsf_8_repeat_extend_encodeSnappyBlockAsm64K + XORQ 8(BX)(R10*1), R11 + JNZ matchlen_bsf_16repeat_extend_encodeSnappyBlockAsm64K + LEAL -16(DI), DI + LEAL 16(R10), R10 + JMP matchlen_loopback_16_repeat_extend_encodeSnappyBlockAsm64K + +matchlen_bsf_16repeat_extend_encodeSnappyBlockAsm64K: +#ifdef GOAMD64_v3 + TZCNTQ R11, R11 + +#else + BSFQ R11, R11 + +#endif + SARQ $0x03, R11 + LEAL 8(R10)(R11*1), R10 + JMP repeat_extend_forward_end_encodeSnappyBlockAsm64K + +matchlen_match8_repeat_extend_encodeSnappyBlockAsm64K: + CMPL DI, $0x08 + JB matchlen_match4_repeat_extend_encodeSnappyBlockAsm64K + MOVQ (R8)(R10*1), R9 + XORQ (BX)(R10*1), R9 + JNZ matchlen_bsf_8_repeat_extend_encodeSnappyBlockAsm64K + LEAL -8(DI), DI + LEAL 8(R10), R10 + JMP matchlen_match4_repeat_extend_encodeSnappyBlockAsm64K + +matchlen_bsf_8_repeat_extend_encodeSnappyBlockAsm64K: +#ifdef GOAMD64_v3 + TZCNTQ R9, R9 + +#else + BSFQ R9, R9 + +#endif + SARQ $0x03, R9 + LEAL (R10)(R9*1), R10 + JMP repeat_extend_forward_end_encodeSnappyBlockAsm64K + +matchlen_match4_repeat_extend_encodeSnappyBlockAsm64K: + CMPL DI, $0x04 + JB matchlen_match2_repeat_extend_encodeSnappyBlockAsm64K + MOVL (R8)(R10*1), R9 + CMPL (BX)(R10*1), R9 + JNE matchlen_match2_repeat_extend_encodeSnappyBlockAsm64K + LEAL -4(DI), DI + LEAL 4(R10), R10 + +matchlen_match2_repeat_extend_encodeSnappyBlockAsm64K: + CMPL DI, $0x01 + JE matchlen_match1_repeat_extend_encodeSnappyBlockAsm64K + JB repeat_extend_forward_end_encodeSnappyBlockAsm64K + MOVW (R8)(R10*1), R9 + CMPW (BX)(R10*1), R9 + JNE matchlen_match1_repeat_extend_encodeSnappyBlockAsm64K + LEAL 2(R10), R10 + SUBL $0x02, DI + JZ repeat_extend_forward_end_encodeSnappyBlockAsm64K + +matchlen_match1_repeat_extend_encodeSnappyBlockAsm64K: + MOVB (R8)(R10*1), R9 + CMPB (BX)(R10*1), R9 + JNE repeat_extend_forward_end_encodeSnappyBlockAsm64K + LEAL 1(R10), R10 + +repeat_extend_forward_end_encodeSnappyBlockAsm64K: + ADDL R10, CX + MOVL CX, BX + SUBL SI, BX + MOVL 16(SP), SI + + // emitCopy +two_byte_offset_repeat_as_copy_encodeSnappyBlockAsm64K: + CMPL BX, $0x40 + JBE two_byte_offset_short_repeat_as_copy_encodeSnappyBlockAsm64K + MOVB $0xee, (AX) + MOVW SI, 1(AX) + LEAL -60(BX), BX + ADDQ $0x03, AX + JMP two_byte_offset_repeat_as_copy_encodeSnappyBlockAsm64K + +two_byte_offset_short_repeat_as_copy_encodeSnappyBlockAsm64K: + MOVL BX, DI + SHLL $0x02, DI + CMPL BX, $0x0c + JAE emit_copy_three_repeat_as_copy_encodeSnappyBlockAsm64K + CMPL SI, $0x00000800 + JAE emit_copy_three_repeat_as_copy_encodeSnappyBlockAsm64K + LEAL -15(DI), DI + MOVB SI, 1(AX) + SHRL $0x08, SI + SHLL $0x05, SI + ORL SI, DI + MOVB DI, (AX) + ADDQ $0x02, AX + JMP repeat_end_emit_encodeSnappyBlockAsm64K + +emit_copy_three_repeat_as_copy_encodeSnappyBlockAsm64K: + LEAL -2(DI), DI + MOVB DI, (AX) + MOVW SI, 1(AX) + ADDQ $0x03, AX + +repeat_end_emit_encodeSnappyBlockAsm64K: + MOVL CX, 12(SP) + JMP search_loop_encodeSnappyBlockAsm64K + +no_repeat_found_encodeSnappyBlockAsm64K: + CMPL (DX)(BX*1), SI + JEQ candidate_match_encodeSnappyBlockAsm64K + SHRQ $0x08, SI + MOVL 24(SP)(R9*4), BX + LEAL 2(CX), R8 + CMPL (DX)(DI*1), SI + JEQ candidate2_match_encodeSnappyBlockAsm64K + MOVL R8, 24(SP)(R9*4) + SHRQ $0x08, SI + CMPL (DX)(BX*1), SI + JEQ candidate3_match_encodeSnappyBlockAsm64K + MOVL 20(SP), CX + JMP search_loop_encodeSnappyBlockAsm64K + +candidate3_match_encodeSnappyBlockAsm64K: + ADDL $0x02, CX + JMP candidate_match_encodeSnappyBlockAsm64K + +candidate2_match_encodeSnappyBlockAsm64K: + MOVL R8, 24(SP)(R9*4) + INCL CX + MOVL DI, BX + +candidate_match_encodeSnappyBlockAsm64K: + MOVL 12(SP), SI + TESTL BX, BX + JZ match_extend_back_end_encodeSnappyBlockAsm64K + +match_extend_back_loop_encodeSnappyBlockAsm64K: + CMPL CX, SI + JBE match_extend_back_end_encodeSnappyBlockAsm64K + MOVB -1(DX)(BX*1), DI + MOVB -1(DX)(CX*1), R8 + CMPB DI, R8 + JNE match_extend_back_end_encodeSnappyBlockAsm64K + LEAL -1(CX), CX + DECL BX + JZ match_extend_back_end_encodeSnappyBlockAsm64K + JMP match_extend_back_loop_encodeSnappyBlockAsm64K + +match_extend_back_end_encodeSnappyBlockAsm64K: + MOVL CX, SI + SUBL 12(SP), SI + LEAQ 3(AX)(SI*1), SI + CMPQ SI, (SP) + JB match_dst_size_check_encodeSnappyBlockAsm64K + MOVQ $0x00000000, ret+48(FP) + RET + +match_dst_size_check_encodeSnappyBlockAsm64K: + MOVL CX, SI + MOVL 12(SP), DI + CMPL DI, SI + JEQ emit_literal_done_match_emit_encodeSnappyBlockAsm64K + MOVL SI, R8 + MOVL SI, 12(SP) + LEAQ (DX)(DI*1), SI + SUBL DI, R8 + LEAL -1(R8), DI + CMPL DI, $0x3c + JB one_byte_match_emit_encodeSnappyBlockAsm64K + CMPL DI, $0x00000100 + JB two_bytes_match_emit_encodeSnappyBlockAsm64K + JB three_bytes_match_emit_encodeSnappyBlockAsm64K + +three_bytes_match_emit_encodeSnappyBlockAsm64K: + MOVB $0xf4, (AX) + MOVW DI, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_match_emit_encodeSnappyBlockAsm64K + +two_bytes_match_emit_encodeSnappyBlockAsm64K: + MOVB $0xf0, (AX) + MOVB DI, 1(AX) + ADDQ $0x02, AX + CMPL DI, $0x40 + JB memmove_match_emit_encodeSnappyBlockAsm64K + JMP memmove_long_match_emit_encodeSnappyBlockAsm64K + +one_byte_match_emit_encodeSnappyBlockAsm64K: + SHLB $0x02, DI + MOVB DI, (AX) + ADDQ $0x01, AX + +memmove_match_emit_encodeSnappyBlockAsm64K: + LEAQ (AX)(R8*1), DI + + // genMemMoveShort + CMPQ R8, $0x08 + JBE emit_lit_memmove_match_emit_encodeSnappyBlockAsm64K_memmove_move_8 + CMPQ R8, $0x10 + JBE emit_lit_memmove_match_emit_encodeSnappyBlockAsm64K_memmove_move_8through16 + CMPQ R8, $0x20 + JBE emit_lit_memmove_match_emit_encodeSnappyBlockAsm64K_memmove_move_17through32 + JMP emit_lit_memmove_match_emit_encodeSnappyBlockAsm64K_memmove_move_33through64 + +emit_lit_memmove_match_emit_encodeSnappyBlockAsm64K_memmove_move_8: + MOVQ (SI), R9 + MOVQ R9, (AX) + JMP memmove_end_copy_match_emit_encodeSnappyBlockAsm64K + +emit_lit_memmove_match_emit_encodeSnappyBlockAsm64K_memmove_move_8through16: + MOVQ (SI), R9 + MOVQ -8(SI)(R8*1), SI + MOVQ R9, (AX) + MOVQ SI, -8(AX)(R8*1) + JMP memmove_end_copy_match_emit_encodeSnappyBlockAsm64K + +emit_lit_memmove_match_emit_encodeSnappyBlockAsm64K_memmove_move_17through32: + MOVOU (SI), X0 + MOVOU -16(SI)(R8*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(R8*1) + JMP memmove_end_copy_match_emit_encodeSnappyBlockAsm64K + +emit_lit_memmove_match_emit_encodeSnappyBlockAsm64K_memmove_move_33through64: + MOVOU (SI), X0 + MOVOU 16(SI), X1 + MOVOU -32(SI)(R8*1), X2 + MOVOU -16(SI)(R8*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + +memmove_end_copy_match_emit_encodeSnappyBlockAsm64K: + MOVQ DI, AX + JMP emit_literal_done_match_emit_encodeSnappyBlockAsm64K + +memmove_long_match_emit_encodeSnappyBlockAsm64K: + LEAQ (AX)(R8*1), DI + + // genMemMoveLong + MOVOU (SI), X0 + MOVOU 16(SI), X1 + MOVOU -32(SI)(R8*1), X2 + MOVOU -16(SI)(R8*1), X3 + MOVQ R8, R10 + SHRQ $0x05, R10 + MOVQ AX, R9 + ANDL $0x0000001f, R9 + MOVQ $0x00000040, R11 + SUBQ R9, R11 + DECQ R10 + JA emit_lit_memmove_long_match_emit_encodeSnappyBlockAsm64Klarge_forward_sse_loop_32 + LEAQ -32(SI)(R11*1), R9 + LEAQ -32(AX)(R11*1), R12 + +emit_lit_memmove_long_match_emit_encodeSnappyBlockAsm64Klarge_big_loop_back: + MOVOU (R9), X4 + MOVOU 16(R9), X5 + MOVOA X4, (R12) + MOVOA X5, 16(R12) + ADDQ $0x20, R12 + ADDQ $0x20, R9 + ADDQ $0x20, R11 + DECQ R10 + JNA emit_lit_memmove_long_match_emit_encodeSnappyBlockAsm64Klarge_big_loop_back + +emit_lit_memmove_long_match_emit_encodeSnappyBlockAsm64Klarge_forward_sse_loop_32: + MOVOU -32(SI)(R11*1), X4 + MOVOU -16(SI)(R11*1), X5 + MOVOA X4, -32(AX)(R11*1) + MOVOA X5, -16(AX)(R11*1) + ADDQ $0x20, R11 + CMPQ R8, R11 + JAE emit_lit_memmove_long_match_emit_encodeSnappyBlockAsm64Klarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + MOVQ DI, AX + +emit_literal_done_match_emit_encodeSnappyBlockAsm64K: +match_nolit_loop_encodeSnappyBlockAsm64K: + MOVL CX, SI + SUBL BX, SI + MOVL SI, 16(SP) + ADDL $0x04, CX + ADDL $0x04, BX + MOVQ src_len+32(FP), SI + SUBL CX, SI + LEAQ (DX)(CX*1), DI + LEAQ (DX)(BX*1), BX + + // matchLen + XORL R9, R9 + +matchlen_loopback_16_match_nolit_encodeSnappyBlockAsm64K: + CMPL SI, $0x10 + JB matchlen_match8_match_nolit_encodeSnappyBlockAsm64K + MOVQ (DI)(R9*1), R8 + MOVQ 8(DI)(R9*1), R10 + XORQ (BX)(R9*1), R8 + JNZ matchlen_bsf_8_match_nolit_encodeSnappyBlockAsm64K + XORQ 8(BX)(R9*1), R10 + JNZ matchlen_bsf_16match_nolit_encodeSnappyBlockAsm64K + LEAL -16(SI), SI + LEAL 16(R9), R9 + JMP matchlen_loopback_16_match_nolit_encodeSnappyBlockAsm64K + +matchlen_bsf_16match_nolit_encodeSnappyBlockAsm64K: +#ifdef GOAMD64_v3 + TZCNTQ R10, R10 + +#else + BSFQ R10, R10 + +#endif + SARQ $0x03, R10 + LEAL 8(R9)(R10*1), R9 + JMP match_nolit_end_encodeSnappyBlockAsm64K + +matchlen_match8_match_nolit_encodeSnappyBlockAsm64K: + CMPL SI, $0x08 + JB matchlen_match4_match_nolit_encodeSnappyBlockAsm64K + MOVQ (DI)(R9*1), R8 + XORQ (BX)(R9*1), R8 + JNZ matchlen_bsf_8_match_nolit_encodeSnappyBlockAsm64K + LEAL -8(SI), SI + LEAL 8(R9), R9 + JMP matchlen_match4_match_nolit_encodeSnappyBlockAsm64K + +matchlen_bsf_8_match_nolit_encodeSnappyBlockAsm64K: +#ifdef GOAMD64_v3 + TZCNTQ R8, R8 + +#else + BSFQ R8, R8 + +#endif + SARQ $0x03, R8 + LEAL (R9)(R8*1), R9 + JMP match_nolit_end_encodeSnappyBlockAsm64K + +matchlen_match4_match_nolit_encodeSnappyBlockAsm64K: + CMPL SI, $0x04 + JB matchlen_match2_match_nolit_encodeSnappyBlockAsm64K + MOVL (DI)(R9*1), R8 + CMPL (BX)(R9*1), R8 + JNE matchlen_match2_match_nolit_encodeSnappyBlockAsm64K + LEAL -4(SI), SI + LEAL 4(R9), R9 + +matchlen_match2_match_nolit_encodeSnappyBlockAsm64K: + CMPL SI, $0x01 + JE matchlen_match1_match_nolit_encodeSnappyBlockAsm64K + JB match_nolit_end_encodeSnappyBlockAsm64K + MOVW (DI)(R9*1), R8 + CMPW (BX)(R9*1), R8 + JNE matchlen_match1_match_nolit_encodeSnappyBlockAsm64K + LEAL 2(R9), R9 + SUBL $0x02, SI + JZ match_nolit_end_encodeSnappyBlockAsm64K + +matchlen_match1_match_nolit_encodeSnappyBlockAsm64K: + MOVB (DI)(R9*1), R8 + CMPB (BX)(R9*1), R8 + JNE match_nolit_end_encodeSnappyBlockAsm64K + LEAL 1(R9), R9 + +match_nolit_end_encodeSnappyBlockAsm64K: + ADDL R9, CX + MOVL 16(SP), BX + ADDL $0x04, R9 + MOVL CX, 12(SP) + + // emitCopy +two_byte_offset_match_nolit_encodeSnappyBlockAsm64K: + CMPL R9, $0x40 + JBE two_byte_offset_short_match_nolit_encodeSnappyBlockAsm64K + MOVB $0xee, (AX) + MOVW BX, 1(AX) + LEAL -60(R9), R9 + ADDQ $0x03, AX + JMP two_byte_offset_match_nolit_encodeSnappyBlockAsm64K + +two_byte_offset_short_match_nolit_encodeSnappyBlockAsm64K: + MOVL R9, SI + SHLL $0x02, SI + CMPL R9, $0x0c + JAE emit_copy_three_match_nolit_encodeSnappyBlockAsm64K + CMPL BX, $0x00000800 + JAE emit_copy_three_match_nolit_encodeSnappyBlockAsm64K + LEAL -15(SI), SI + MOVB BL, 1(AX) + SHRL $0x08, BX + SHLL $0x05, BX + ORL BX, SI + MOVB SI, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeSnappyBlockAsm64K + +emit_copy_three_match_nolit_encodeSnappyBlockAsm64K: + LEAL -2(SI), SI + MOVB SI, (AX) + MOVW BX, 1(AX) + ADDQ $0x03, AX + +match_nolit_emitcopy_end_encodeSnappyBlockAsm64K: + CMPL CX, 8(SP) + JAE emit_remainder_encodeSnappyBlockAsm64K + MOVQ -2(DX)(CX*1), SI + CMPQ AX, (SP) + JB match_nolit_dst_ok_encodeSnappyBlockAsm64K + MOVQ $0x00000000, ret+48(FP) + RET + +match_nolit_dst_ok_encodeSnappyBlockAsm64K: + MOVQ $0x0000cf1bbcdcbf9b, R8 + MOVQ SI, DI + SHRQ $0x10, SI + MOVQ SI, BX + SHLQ $0x10, DI + IMULQ R8, DI + SHRQ $0x32, DI + SHLQ $0x10, BX + IMULQ R8, BX + SHRQ $0x32, BX + LEAL -2(CX), R8 + LEAQ 24(SP)(BX*4), R9 + MOVL (R9), BX + MOVL R8, 24(SP)(DI*4) + MOVL CX, (R9) + CMPL (DX)(BX*1), SI + JEQ match_nolit_loop_encodeSnappyBlockAsm64K + INCL CX + JMP search_loop_encodeSnappyBlockAsm64K + +emit_remainder_encodeSnappyBlockAsm64K: + MOVQ src_len+32(FP), CX + SUBL 12(SP), CX + LEAQ 3(AX)(CX*1), CX + CMPQ CX, (SP) + JB emit_remainder_ok_encodeSnappyBlockAsm64K + MOVQ $0x00000000, ret+48(FP) + RET + +emit_remainder_ok_encodeSnappyBlockAsm64K: + MOVQ src_len+32(FP), CX + MOVL 12(SP), BX + CMPL BX, CX + JEQ emit_literal_done_emit_remainder_encodeSnappyBlockAsm64K + MOVL CX, SI + MOVL CX, 12(SP) + LEAQ (DX)(BX*1), CX + SUBL BX, SI + LEAL -1(SI), DX + CMPL DX, $0x3c + JB one_byte_emit_remainder_encodeSnappyBlockAsm64K + CMPL DX, $0x00000100 + JB two_bytes_emit_remainder_encodeSnappyBlockAsm64K + JB three_bytes_emit_remainder_encodeSnappyBlockAsm64K + +three_bytes_emit_remainder_encodeSnappyBlockAsm64K: + MOVB $0xf4, (AX) + MOVW DX, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_emit_remainder_encodeSnappyBlockAsm64K + +two_bytes_emit_remainder_encodeSnappyBlockAsm64K: + MOVB $0xf0, (AX) + MOVB DL, 1(AX) + ADDQ $0x02, AX + CMPL DX, $0x40 + JB memmove_emit_remainder_encodeSnappyBlockAsm64K + JMP memmove_long_emit_remainder_encodeSnappyBlockAsm64K + +one_byte_emit_remainder_encodeSnappyBlockAsm64K: + SHLB $0x02, DL + MOVB DL, (AX) + ADDQ $0x01, AX + +memmove_emit_remainder_encodeSnappyBlockAsm64K: + LEAQ (AX)(SI*1), DX + MOVL SI, BX + + // genMemMoveShort + CMPQ BX, $0x03 + JB emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm64K_memmove_move_1or2 + JE emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm64K_memmove_move_3 + CMPQ BX, $0x08 + JB emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm64K_memmove_move_4through7 + CMPQ BX, $0x10 + JBE emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm64K_memmove_move_8through16 + CMPQ BX, $0x20 + JBE emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm64K_memmove_move_17through32 + JMP emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm64K_memmove_move_33through64 + +emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm64K_memmove_move_1or2: + MOVB (CX), SI + MOVB -1(CX)(BX*1), CL + MOVB SI, (AX) + MOVB CL, -1(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeSnappyBlockAsm64K + +emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm64K_memmove_move_3: + MOVW (CX), SI + MOVB 2(CX), CL + MOVW SI, (AX) + MOVB CL, 2(AX) + JMP memmove_end_copy_emit_remainder_encodeSnappyBlockAsm64K + +emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm64K_memmove_move_4through7: + MOVL (CX), SI + MOVL -4(CX)(BX*1), CX + MOVL SI, (AX) + MOVL CX, -4(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeSnappyBlockAsm64K + +emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm64K_memmove_move_8through16: + MOVQ (CX), SI + MOVQ -8(CX)(BX*1), CX + MOVQ SI, (AX) + MOVQ CX, -8(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeSnappyBlockAsm64K + +emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm64K_memmove_move_17through32: + MOVOU (CX), X0 + MOVOU -16(CX)(BX*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeSnappyBlockAsm64K + +emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm64K_memmove_move_33through64: + MOVOU (CX), X0 + MOVOU 16(CX), X1 + MOVOU -32(CX)(BX*1), X2 + MOVOU -16(CX)(BX*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(BX*1) + MOVOU X3, -16(AX)(BX*1) + +memmove_end_copy_emit_remainder_encodeSnappyBlockAsm64K: + MOVQ DX, AX + JMP emit_literal_done_emit_remainder_encodeSnappyBlockAsm64K + +memmove_long_emit_remainder_encodeSnappyBlockAsm64K: + LEAQ (AX)(SI*1), DX + MOVL SI, BX + + // genMemMoveLong + MOVOU (CX), X0 + MOVOU 16(CX), X1 + MOVOU -32(CX)(BX*1), X2 + MOVOU -16(CX)(BX*1), X3 + MOVQ BX, DI + SHRQ $0x05, DI + MOVQ AX, SI + ANDL $0x0000001f, SI + MOVQ $0x00000040, R8 + SUBQ SI, R8 + DECQ DI + JA emit_lit_memmove_long_emit_remainder_encodeSnappyBlockAsm64Klarge_forward_sse_loop_32 + LEAQ -32(CX)(R8*1), SI + LEAQ -32(AX)(R8*1), R9 + +emit_lit_memmove_long_emit_remainder_encodeSnappyBlockAsm64Klarge_big_loop_back: + MOVOU (SI), X4 + MOVOU 16(SI), X5 + MOVOA X4, (R9) + MOVOA X5, 16(R9) + ADDQ $0x20, R9 + ADDQ $0x20, SI + ADDQ $0x20, R8 + DECQ DI + JNA emit_lit_memmove_long_emit_remainder_encodeSnappyBlockAsm64Klarge_big_loop_back + +emit_lit_memmove_long_emit_remainder_encodeSnappyBlockAsm64Klarge_forward_sse_loop_32: + MOVOU -32(CX)(R8*1), X4 + MOVOU -16(CX)(R8*1), X5 + MOVOA X4, -32(AX)(R8*1) + MOVOA X5, -16(AX)(R8*1) + ADDQ $0x20, R8 + CMPQ BX, R8 + JAE emit_lit_memmove_long_emit_remainder_encodeSnappyBlockAsm64Klarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(BX*1) + MOVOU X3, -16(AX)(BX*1) + MOVQ DX, AX + +emit_literal_done_emit_remainder_encodeSnappyBlockAsm64K: + MOVQ dst_base+0(FP), CX + SUBQ CX, AX + MOVQ AX, ret+48(FP) + RET + +// func encodeSnappyBlockAsm12B(dst []byte, src []byte) int +// Requires: BMI, SSE2 +TEXT ·encodeSnappyBlockAsm12B(SB), $16408-56 + MOVQ dst_base+0(FP), AX + MOVQ $0x00000080, CX + LEAQ 24(SP), DX + PXOR X0, X0 + +zero_loop_encodeSnappyBlockAsm12B: + MOVOU X0, (DX) + MOVOU X0, 16(DX) + MOVOU X0, 32(DX) + MOVOU X0, 48(DX) + MOVOU X0, 64(DX) + MOVOU X0, 80(DX) + MOVOU X0, 96(DX) + MOVOU X0, 112(DX) + ADDQ $0x80, DX + DECQ CX + JNZ zero_loop_encodeSnappyBlockAsm12B + MOVL $0x00000000, 12(SP) + MOVQ src_len+32(FP), CX + LEAQ -9(CX), DX + LEAQ -8(CX), BX + MOVL BX, 8(SP) + SHRQ $0x05, CX + SUBL CX, DX + LEAQ (AX)(DX*1), DX + MOVQ DX, (SP) + MOVL $0x00000001, CX + MOVL CX, 16(SP) + MOVQ src_base+24(FP), DX + +search_loop_encodeSnappyBlockAsm12B: + MOVL CX, BX + SUBL 12(SP), BX + SHRL $0x05, BX + LEAL 4(CX)(BX*1), BX + CMPL BX, 8(SP) + JAE emit_remainder_encodeSnappyBlockAsm12B + MOVQ (DX)(CX*1), SI + MOVL BX, 20(SP) + MOVQ $0x000000cf1bbcdcbb, R8 + MOVQ SI, R9 + MOVQ SI, R10 + SHRQ $0x08, R10 + SHLQ $0x18, R9 + IMULQ R8, R9 + SHRQ $0x34, R9 + SHLQ $0x18, R10 + IMULQ R8, R10 + SHRQ $0x34, R10 + MOVL 24(SP)(R9*4), BX + MOVL 24(SP)(R10*4), DI + MOVL CX, 24(SP)(R9*4) + LEAL 1(CX), R9 + MOVL R9, 24(SP)(R10*4) + MOVQ SI, R9 + SHRQ $0x10, R9 + SHLQ $0x18, R9 + IMULQ R8, R9 + SHRQ $0x34, R9 + MOVL CX, R8 + SUBL 16(SP), R8 + MOVL 1(DX)(R8*1), R10 + MOVQ SI, R8 + SHRQ $0x08, R8 + CMPL R8, R10 + JNE no_repeat_found_encodeSnappyBlockAsm12B + LEAL 1(CX), SI + MOVL 12(SP), BX + MOVL SI, DI + SUBL 16(SP), DI + JZ repeat_extend_back_end_encodeSnappyBlockAsm12B + +repeat_extend_back_loop_encodeSnappyBlockAsm12B: + CMPL SI, BX + JBE repeat_extend_back_end_encodeSnappyBlockAsm12B + MOVB -1(DX)(DI*1), R8 + MOVB -1(DX)(SI*1), R9 + CMPB R8, R9 + JNE repeat_extend_back_end_encodeSnappyBlockAsm12B + LEAL -1(SI), SI + DECL DI + JNZ repeat_extend_back_loop_encodeSnappyBlockAsm12B + +repeat_extend_back_end_encodeSnappyBlockAsm12B: + MOVL SI, BX + SUBL 12(SP), BX + LEAQ 3(AX)(BX*1), BX + CMPQ BX, (SP) + JB repeat_dst_size_check_encodeSnappyBlockAsm12B + MOVQ $0x00000000, ret+48(FP) + RET + +repeat_dst_size_check_encodeSnappyBlockAsm12B: + MOVL 12(SP), BX + CMPL BX, SI + JEQ emit_literal_done_repeat_emit_encodeSnappyBlockAsm12B + MOVL SI, DI + MOVL SI, 12(SP) + LEAQ (DX)(BX*1), R8 + SUBL BX, DI + LEAL -1(DI), BX + CMPL BX, $0x3c + JB one_byte_repeat_emit_encodeSnappyBlockAsm12B + CMPL BX, $0x00000100 + JB two_bytes_repeat_emit_encodeSnappyBlockAsm12B + JB three_bytes_repeat_emit_encodeSnappyBlockAsm12B + +three_bytes_repeat_emit_encodeSnappyBlockAsm12B: + MOVB $0xf4, (AX) + MOVW BX, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_repeat_emit_encodeSnappyBlockAsm12B + +two_bytes_repeat_emit_encodeSnappyBlockAsm12B: + MOVB $0xf0, (AX) + MOVB BL, 1(AX) + ADDQ $0x02, AX + CMPL BX, $0x40 + JB memmove_repeat_emit_encodeSnappyBlockAsm12B + JMP memmove_long_repeat_emit_encodeSnappyBlockAsm12B + +one_byte_repeat_emit_encodeSnappyBlockAsm12B: + SHLB $0x02, BL + MOVB BL, (AX) + ADDQ $0x01, AX + +memmove_repeat_emit_encodeSnappyBlockAsm12B: + LEAQ (AX)(DI*1), BX + + // genMemMoveShort + CMPQ DI, $0x08 + JBE emit_lit_memmove_repeat_emit_encodeSnappyBlockAsm12B_memmove_move_8 + CMPQ DI, $0x10 + JBE emit_lit_memmove_repeat_emit_encodeSnappyBlockAsm12B_memmove_move_8through16 + CMPQ DI, $0x20 + JBE emit_lit_memmove_repeat_emit_encodeSnappyBlockAsm12B_memmove_move_17through32 + JMP emit_lit_memmove_repeat_emit_encodeSnappyBlockAsm12B_memmove_move_33through64 + +emit_lit_memmove_repeat_emit_encodeSnappyBlockAsm12B_memmove_move_8: + MOVQ (R8), R9 + MOVQ R9, (AX) + JMP memmove_end_copy_repeat_emit_encodeSnappyBlockAsm12B + +emit_lit_memmove_repeat_emit_encodeSnappyBlockAsm12B_memmove_move_8through16: + MOVQ (R8), R9 + MOVQ -8(R8)(DI*1), R8 + MOVQ R9, (AX) + MOVQ R8, -8(AX)(DI*1) + JMP memmove_end_copy_repeat_emit_encodeSnappyBlockAsm12B + +emit_lit_memmove_repeat_emit_encodeSnappyBlockAsm12B_memmove_move_17through32: + MOVOU (R8), X0 + MOVOU -16(R8)(DI*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(DI*1) + JMP memmove_end_copy_repeat_emit_encodeSnappyBlockAsm12B + +emit_lit_memmove_repeat_emit_encodeSnappyBlockAsm12B_memmove_move_33through64: + MOVOU (R8), X0 + MOVOU 16(R8), X1 + MOVOU -32(R8)(DI*1), X2 + MOVOU -16(R8)(DI*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(DI*1) + MOVOU X3, -16(AX)(DI*1) + +memmove_end_copy_repeat_emit_encodeSnappyBlockAsm12B: + MOVQ BX, AX + JMP emit_literal_done_repeat_emit_encodeSnappyBlockAsm12B + +memmove_long_repeat_emit_encodeSnappyBlockAsm12B: + LEAQ (AX)(DI*1), BX + + // genMemMoveLong + MOVOU (R8), X0 + MOVOU 16(R8), X1 + MOVOU -32(R8)(DI*1), X2 + MOVOU -16(R8)(DI*1), X3 + MOVQ DI, R10 + SHRQ $0x05, R10 + MOVQ AX, R9 + ANDL $0x0000001f, R9 + MOVQ $0x00000040, R11 + SUBQ R9, R11 + DECQ R10 + JA emit_lit_memmove_long_repeat_emit_encodeSnappyBlockAsm12Blarge_forward_sse_loop_32 + LEAQ -32(R8)(R11*1), R9 + LEAQ -32(AX)(R11*1), R12 + +emit_lit_memmove_long_repeat_emit_encodeSnappyBlockAsm12Blarge_big_loop_back: + MOVOU (R9), X4 + MOVOU 16(R9), X5 + MOVOA X4, (R12) + MOVOA X5, 16(R12) + ADDQ $0x20, R12 + ADDQ $0x20, R9 + ADDQ $0x20, R11 + DECQ R10 + JNA emit_lit_memmove_long_repeat_emit_encodeSnappyBlockAsm12Blarge_big_loop_back + +emit_lit_memmove_long_repeat_emit_encodeSnappyBlockAsm12Blarge_forward_sse_loop_32: + MOVOU -32(R8)(R11*1), X4 + MOVOU -16(R8)(R11*1), X5 + MOVOA X4, -32(AX)(R11*1) + MOVOA X5, -16(AX)(R11*1) + ADDQ $0x20, R11 + CMPQ DI, R11 + JAE emit_lit_memmove_long_repeat_emit_encodeSnappyBlockAsm12Blarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(DI*1) + MOVOU X3, -16(AX)(DI*1) + MOVQ BX, AX + +emit_literal_done_repeat_emit_encodeSnappyBlockAsm12B: + ADDL $0x05, CX + MOVL CX, BX + SUBL 16(SP), BX + MOVQ src_len+32(FP), DI + SUBL CX, DI + LEAQ (DX)(CX*1), R8 + LEAQ (DX)(BX*1), BX + + // matchLen + XORL R10, R10 + +matchlen_loopback_16_repeat_extend_encodeSnappyBlockAsm12B: + CMPL DI, $0x10 + JB matchlen_match8_repeat_extend_encodeSnappyBlockAsm12B + MOVQ (R8)(R10*1), R9 + MOVQ 8(R8)(R10*1), R11 + XORQ (BX)(R10*1), R9 + JNZ matchlen_bsf_8_repeat_extend_encodeSnappyBlockAsm12B + XORQ 8(BX)(R10*1), R11 + JNZ matchlen_bsf_16repeat_extend_encodeSnappyBlockAsm12B + LEAL -16(DI), DI + LEAL 16(R10), R10 + JMP matchlen_loopback_16_repeat_extend_encodeSnappyBlockAsm12B + +matchlen_bsf_16repeat_extend_encodeSnappyBlockAsm12B: +#ifdef GOAMD64_v3 + TZCNTQ R11, R11 + +#else + BSFQ R11, R11 + +#endif + SARQ $0x03, R11 + LEAL 8(R10)(R11*1), R10 + JMP repeat_extend_forward_end_encodeSnappyBlockAsm12B + +matchlen_match8_repeat_extend_encodeSnappyBlockAsm12B: + CMPL DI, $0x08 + JB matchlen_match4_repeat_extend_encodeSnappyBlockAsm12B + MOVQ (R8)(R10*1), R9 + XORQ (BX)(R10*1), R9 + JNZ matchlen_bsf_8_repeat_extend_encodeSnappyBlockAsm12B + LEAL -8(DI), DI + LEAL 8(R10), R10 + JMP matchlen_match4_repeat_extend_encodeSnappyBlockAsm12B + +matchlen_bsf_8_repeat_extend_encodeSnappyBlockAsm12B: +#ifdef GOAMD64_v3 + TZCNTQ R9, R9 + +#else + BSFQ R9, R9 + +#endif + SARQ $0x03, R9 + LEAL (R10)(R9*1), R10 + JMP repeat_extend_forward_end_encodeSnappyBlockAsm12B + +matchlen_match4_repeat_extend_encodeSnappyBlockAsm12B: + CMPL DI, $0x04 + JB matchlen_match2_repeat_extend_encodeSnappyBlockAsm12B + MOVL (R8)(R10*1), R9 + CMPL (BX)(R10*1), R9 + JNE matchlen_match2_repeat_extend_encodeSnappyBlockAsm12B + LEAL -4(DI), DI + LEAL 4(R10), R10 + +matchlen_match2_repeat_extend_encodeSnappyBlockAsm12B: + CMPL DI, $0x01 + JE matchlen_match1_repeat_extend_encodeSnappyBlockAsm12B + JB repeat_extend_forward_end_encodeSnappyBlockAsm12B + MOVW (R8)(R10*1), R9 + CMPW (BX)(R10*1), R9 + JNE matchlen_match1_repeat_extend_encodeSnappyBlockAsm12B + LEAL 2(R10), R10 + SUBL $0x02, DI + JZ repeat_extend_forward_end_encodeSnappyBlockAsm12B + +matchlen_match1_repeat_extend_encodeSnappyBlockAsm12B: + MOVB (R8)(R10*1), R9 + CMPB (BX)(R10*1), R9 + JNE repeat_extend_forward_end_encodeSnappyBlockAsm12B + LEAL 1(R10), R10 + +repeat_extend_forward_end_encodeSnappyBlockAsm12B: + ADDL R10, CX + MOVL CX, BX + SUBL SI, BX + MOVL 16(SP), SI + + // emitCopy +two_byte_offset_repeat_as_copy_encodeSnappyBlockAsm12B: + CMPL BX, $0x40 + JBE two_byte_offset_short_repeat_as_copy_encodeSnappyBlockAsm12B + MOVB $0xee, (AX) + MOVW SI, 1(AX) + LEAL -60(BX), BX + ADDQ $0x03, AX + JMP two_byte_offset_repeat_as_copy_encodeSnappyBlockAsm12B + +two_byte_offset_short_repeat_as_copy_encodeSnappyBlockAsm12B: + MOVL BX, DI + SHLL $0x02, DI + CMPL BX, $0x0c + JAE emit_copy_three_repeat_as_copy_encodeSnappyBlockAsm12B + CMPL SI, $0x00000800 + JAE emit_copy_three_repeat_as_copy_encodeSnappyBlockAsm12B + LEAL -15(DI), DI + MOVB SI, 1(AX) + SHRL $0x08, SI + SHLL $0x05, SI + ORL SI, DI + MOVB DI, (AX) + ADDQ $0x02, AX + JMP repeat_end_emit_encodeSnappyBlockAsm12B + +emit_copy_three_repeat_as_copy_encodeSnappyBlockAsm12B: + LEAL -2(DI), DI + MOVB DI, (AX) + MOVW SI, 1(AX) + ADDQ $0x03, AX + +repeat_end_emit_encodeSnappyBlockAsm12B: + MOVL CX, 12(SP) + JMP search_loop_encodeSnappyBlockAsm12B + +no_repeat_found_encodeSnappyBlockAsm12B: + CMPL (DX)(BX*1), SI + JEQ candidate_match_encodeSnappyBlockAsm12B + SHRQ $0x08, SI + MOVL 24(SP)(R9*4), BX + LEAL 2(CX), R8 + CMPL (DX)(DI*1), SI + JEQ candidate2_match_encodeSnappyBlockAsm12B + MOVL R8, 24(SP)(R9*4) + SHRQ $0x08, SI + CMPL (DX)(BX*1), SI + JEQ candidate3_match_encodeSnappyBlockAsm12B + MOVL 20(SP), CX + JMP search_loop_encodeSnappyBlockAsm12B + +candidate3_match_encodeSnappyBlockAsm12B: + ADDL $0x02, CX + JMP candidate_match_encodeSnappyBlockAsm12B + +candidate2_match_encodeSnappyBlockAsm12B: + MOVL R8, 24(SP)(R9*4) + INCL CX + MOVL DI, BX + +candidate_match_encodeSnappyBlockAsm12B: + MOVL 12(SP), SI + TESTL BX, BX + JZ match_extend_back_end_encodeSnappyBlockAsm12B + +match_extend_back_loop_encodeSnappyBlockAsm12B: + CMPL CX, SI + JBE match_extend_back_end_encodeSnappyBlockAsm12B + MOVB -1(DX)(BX*1), DI + MOVB -1(DX)(CX*1), R8 + CMPB DI, R8 + JNE match_extend_back_end_encodeSnappyBlockAsm12B + LEAL -1(CX), CX + DECL BX + JZ match_extend_back_end_encodeSnappyBlockAsm12B + JMP match_extend_back_loop_encodeSnappyBlockAsm12B + +match_extend_back_end_encodeSnappyBlockAsm12B: + MOVL CX, SI + SUBL 12(SP), SI + LEAQ 3(AX)(SI*1), SI + CMPQ SI, (SP) + JB match_dst_size_check_encodeSnappyBlockAsm12B + MOVQ $0x00000000, ret+48(FP) + RET + +match_dst_size_check_encodeSnappyBlockAsm12B: + MOVL CX, SI + MOVL 12(SP), DI + CMPL DI, SI + JEQ emit_literal_done_match_emit_encodeSnappyBlockAsm12B + MOVL SI, R8 + MOVL SI, 12(SP) + LEAQ (DX)(DI*1), SI + SUBL DI, R8 + LEAL -1(R8), DI + CMPL DI, $0x3c + JB one_byte_match_emit_encodeSnappyBlockAsm12B + CMPL DI, $0x00000100 + JB two_bytes_match_emit_encodeSnappyBlockAsm12B + JB three_bytes_match_emit_encodeSnappyBlockAsm12B + +three_bytes_match_emit_encodeSnappyBlockAsm12B: + MOVB $0xf4, (AX) + MOVW DI, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_match_emit_encodeSnappyBlockAsm12B + +two_bytes_match_emit_encodeSnappyBlockAsm12B: + MOVB $0xf0, (AX) + MOVB DI, 1(AX) + ADDQ $0x02, AX + CMPL DI, $0x40 + JB memmove_match_emit_encodeSnappyBlockAsm12B + JMP memmove_long_match_emit_encodeSnappyBlockAsm12B + +one_byte_match_emit_encodeSnappyBlockAsm12B: + SHLB $0x02, DI + MOVB DI, (AX) + ADDQ $0x01, AX + +memmove_match_emit_encodeSnappyBlockAsm12B: + LEAQ (AX)(R8*1), DI + + // genMemMoveShort + CMPQ R8, $0x08 + JBE emit_lit_memmove_match_emit_encodeSnappyBlockAsm12B_memmove_move_8 + CMPQ R8, $0x10 + JBE emit_lit_memmove_match_emit_encodeSnappyBlockAsm12B_memmove_move_8through16 + CMPQ R8, $0x20 + JBE emit_lit_memmove_match_emit_encodeSnappyBlockAsm12B_memmove_move_17through32 + JMP emit_lit_memmove_match_emit_encodeSnappyBlockAsm12B_memmove_move_33through64 + +emit_lit_memmove_match_emit_encodeSnappyBlockAsm12B_memmove_move_8: + MOVQ (SI), R9 + MOVQ R9, (AX) + JMP memmove_end_copy_match_emit_encodeSnappyBlockAsm12B + +emit_lit_memmove_match_emit_encodeSnappyBlockAsm12B_memmove_move_8through16: + MOVQ (SI), R9 + MOVQ -8(SI)(R8*1), SI + MOVQ R9, (AX) + MOVQ SI, -8(AX)(R8*1) + JMP memmove_end_copy_match_emit_encodeSnappyBlockAsm12B + +emit_lit_memmove_match_emit_encodeSnappyBlockAsm12B_memmove_move_17through32: + MOVOU (SI), X0 + MOVOU -16(SI)(R8*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(R8*1) + JMP memmove_end_copy_match_emit_encodeSnappyBlockAsm12B + +emit_lit_memmove_match_emit_encodeSnappyBlockAsm12B_memmove_move_33through64: + MOVOU (SI), X0 + MOVOU 16(SI), X1 + MOVOU -32(SI)(R8*1), X2 + MOVOU -16(SI)(R8*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + +memmove_end_copy_match_emit_encodeSnappyBlockAsm12B: + MOVQ DI, AX + JMP emit_literal_done_match_emit_encodeSnappyBlockAsm12B + +memmove_long_match_emit_encodeSnappyBlockAsm12B: + LEAQ (AX)(R8*1), DI + + // genMemMoveLong + MOVOU (SI), X0 + MOVOU 16(SI), X1 + MOVOU -32(SI)(R8*1), X2 + MOVOU -16(SI)(R8*1), X3 + MOVQ R8, R10 + SHRQ $0x05, R10 + MOVQ AX, R9 + ANDL $0x0000001f, R9 + MOVQ $0x00000040, R11 + SUBQ R9, R11 + DECQ R10 + JA emit_lit_memmove_long_match_emit_encodeSnappyBlockAsm12Blarge_forward_sse_loop_32 + LEAQ -32(SI)(R11*1), R9 + LEAQ -32(AX)(R11*1), R12 + +emit_lit_memmove_long_match_emit_encodeSnappyBlockAsm12Blarge_big_loop_back: + MOVOU (R9), X4 + MOVOU 16(R9), X5 + MOVOA X4, (R12) + MOVOA X5, 16(R12) + ADDQ $0x20, R12 + ADDQ $0x20, R9 + ADDQ $0x20, R11 + DECQ R10 + JNA emit_lit_memmove_long_match_emit_encodeSnappyBlockAsm12Blarge_big_loop_back + +emit_lit_memmove_long_match_emit_encodeSnappyBlockAsm12Blarge_forward_sse_loop_32: + MOVOU -32(SI)(R11*1), X4 + MOVOU -16(SI)(R11*1), X5 + MOVOA X4, -32(AX)(R11*1) + MOVOA X5, -16(AX)(R11*1) + ADDQ $0x20, R11 + CMPQ R8, R11 + JAE emit_lit_memmove_long_match_emit_encodeSnappyBlockAsm12Blarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + MOVQ DI, AX + +emit_literal_done_match_emit_encodeSnappyBlockAsm12B: +match_nolit_loop_encodeSnappyBlockAsm12B: + MOVL CX, SI + SUBL BX, SI + MOVL SI, 16(SP) + ADDL $0x04, CX + ADDL $0x04, BX + MOVQ src_len+32(FP), SI + SUBL CX, SI + LEAQ (DX)(CX*1), DI + LEAQ (DX)(BX*1), BX + + // matchLen + XORL R9, R9 + +matchlen_loopback_16_match_nolit_encodeSnappyBlockAsm12B: + CMPL SI, $0x10 + JB matchlen_match8_match_nolit_encodeSnappyBlockAsm12B + MOVQ (DI)(R9*1), R8 + MOVQ 8(DI)(R9*1), R10 + XORQ (BX)(R9*1), R8 + JNZ matchlen_bsf_8_match_nolit_encodeSnappyBlockAsm12B + XORQ 8(BX)(R9*1), R10 + JNZ matchlen_bsf_16match_nolit_encodeSnappyBlockAsm12B + LEAL -16(SI), SI + LEAL 16(R9), R9 + JMP matchlen_loopback_16_match_nolit_encodeSnappyBlockAsm12B + +matchlen_bsf_16match_nolit_encodeSnappyBlockAsm12B: +#ifdef GOAMD64_v3 + TZCNTQ R10, R10 + +#else + BSFQ R10, R10 + +#endif + SARQ $0x03, R10 + LEAL 8(R9)(R10*1), R9 + JMP match_nolit_end_encodeSnappyBlockAsm12B + +matchlen_match8_match_nolit_encodeSnappyBlockAsm12B: + CMPL SI, $0x08 + JB matchlen_match4_match_nolit_encodeSnappyBlockAsm12B + MOVQ (DI)(R9*1), R8 + XORQ (BX)(R9*1), R8 + JNZ matchlen_bsf_8_match_nolit_encodeSnappyBlockAsm12B + LEAL -8(SI), SI + LEAL 8(R9), R9 + JMP matchlen_match4_match_nolit_encodeSnappyBlockAsm12B + +matchlen_bsf_8_match_nolit_encodeSnappyBlockAsm12B: +#ifdef GOAMD64_v3 + TZCNTQ R8, R8 + +#else + BSFQ R8, R8 + +#endif + SARQ $0x03, R8 + LEAL (R9)(R8*1), R9 + JMP match_nolit_end_encodeSnappyBlockAsm12B + +matchlen_match4_match_nolit_encodeSnappyBlockAsm12B: + CMPL SI, $0x04 + JB matchlen_match2_match_nolit_encodeSnappyBlockAsm12B + MOVL (DI)(R9*1), R8 + CMPL (BX)(R9*1), R8 + JNE matchlen_match2_match_nolit_encodeSnappyBlockAsm12B + LEAL -4(SI), SI + LEAL 4(R9), R9 + +matchlen_match2_match_nolit_encodeSnappyBlockAsm12B: + CMPL SI, $0x01 + JE matchlen_match1_match_nolit_encodeSnappyBlockAsm12B + JB match_nolit_end_encodeSnappyBlockAsm12B + MOVW (DI)(R9*1), R8 + CMPW (BX)(R9*1), R8 + JNE matchlen_match1_match_nolit_encodeSnappyBlockAsm12B + LEAL 2(R9), R9 + SUBL $0x02, SI + JZ match_nolit_end_encodeSnappyBlockAsm12B + +matchlen_match1_match_nolit_encodeSnappyBlockAsm12B: + MOVB (DI)(R9*1), R8 + CMPB (BX)(R9*1), R8 + JNE match_nolit_end_encodeSnappyBlockAsm12B + LEAL 1(R9), R9 + +match_nolit_end_encodeSnappyBlockAsm12B: + ADDL R9, CX + MOVL 16(SP), BX + ADDL $0x04, R9 + MOVL CX, 12(SP) + + // emitCopy +two_byte_offset_match_nolit_encodeSnappyBlockAsm12B: + CMPL R9, $0x40 + JBE two_byte_offset_short_match_nolit_encodeSnappyBlockAsm12B + MOVB $0xee, (AX) + MOVW BX, 1(AX) + LEAL -60(R9), R9 + ADDQ $0x03, AX + JMP two_byte_offset_match_nolit_encodeSnappyBlockAsm12B + +two_byte_offset_short_match_nolit_encodeSnappyBlockAsm12B: + MOVL R9, SI + SHLL $0x02, SI + CMPL R9, $0x0c + JAE emit_copy_three_match_nolit_encodeSnappyBlockAsm12B + CMPL BX, $0x00000800 + JAE emit_copy_three_match_nolit_encodeSnappyBlockAsm12B + LEAL -15(SI), SI + MOVB BL, 1(AX) + SHRL $0x08, BX + SHLL $0x05, BX + ORL BX, SI + MOVB SI, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeSnappyBlockAsm12B + +emit_copy_three_match_nolit_encodeSnappyBlockAsm12B: + LEAL -2(SI), SI + MOVB SI, (AX) + MOVW BX, 1(AX) + ADDQ $0x03, AX + +match_nolit_emitcopy_end_encodeSnappyBlockAsm12B: + CMPL CX, 8(SP) + JAE emit_remainder_encodeSnappyBlockAsm12B + MOVQ -2(DX)(CX*1), SI + CMPQ AX, (SP) + JB match_nolit_dst_ok_encodeSnappyBlockAsm12B + MOVQ $0x00000000, ret+48(FP) + RET + +match_nolit_dst_ok_encodeSnappyBlockAsm12B: + MOVQ $0x000000cf1bbcdcbb, R8 + MOVQ SI, DI + SHRQ $0x10, SI + MOVQ SI, BX + SHLQ $0x18, DI + IMULQ R8, DI + SHRQ $0x34, DI + SHLQ $0x18, BX + IMULQ R8, BX + SHRQ $0x34, BX + LEAL -2(CX), R8 + LEAQ 24(SP)(BX*4), R9 + MOVL (R9), BX + MOVL R8, 24(SP)(DI*4) + MOVL CX, (R9) + CMPL (DX)(BX*1), SI + JEQ match_nolit_loop_encodeSnappyBlockAsm12B + INCL CX + JMP search_loop_encodeSnappyBlockAsm12B + +emit_remainder_encodeSnappyBlockAsm12B: + MOVQ src_len+32(FP), CX + SUBL 12(SP), CX + LEAQ 3(AX)(CX*1), CX + CMPQ CX, (SP) + JB emit_remainder_ok_encodeSnappyBlockAsm12B + MOVQ $0x00000000, ret+48(FP) + RET + +emit_remainder_ok_encodeSnappyBlockAsm12B: + MOVQ src_len+32(FP), CX + MOVL 12(SP), BX + CMPL BX, CX + JEQ emit_literal_done_emit_remainder_encodeSnappyBlockAsm12B + MOVL CX, SI + MOVL CX, 12(SP) + LEAQ (DX)(BX*1), CX + SUBL BX, SI + LEAL -1(SI), DX + CMPL DX, $0x3c + JB one_byte_emit_remainder_encodeSnappyBlockAsm12B + CMPL DX, $0x00000100 + JB two_bytes_emit_remainder_encodeSnappyBlockAsm12B + JB three_bytes_emit_remainder_encodeSnappyBlockAsm12B + +three_bytes_emit_remainder_encodeSnappyBlockAsm12B: + MOVB $0xf4, (AX) + MOVW DX, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_emit_remainder_encodeSnappyBlockAsm12B + +two_bytes_emit_remainder_encodeSnappyBlockAsm12B: + MOVB $0xf0, (AX) + MOVB DL, 1(AX) + ADDQ $0x02, AX + CMPL DX, $0x40 + JB memmove_emit_remainder_encodeSnappyBlockAsm12B + JMP memmove_long_emit_remainder_encodeSnappyBlockAsm12B + +one_byte_emit_remainder_encodeSnappyBlockAsm12B: + SHLB $0x02, DL + MOVB DL, (AX) + ADDQ $0x01, AX + +memmove_emit_remainder_encodeSnappyBlockAsm12B: + LEAQ (AX)(SI*1), DX + MOVL SI, BX + + // genMemMoveShort + CMPQ BX, $0x03 + JB emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm12B_memmove_move_1or2 + JE emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm12B_memmove_move_3 + CMPQ BX, $0x08 + JB emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm12B_memmove_move_4through7 + CMPQ BX, $0x10 + JBE emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm12B_memmove_move_8through16 + CMPQ BX, $0x20 + JBE emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm12B_memmove_move_17through32 + JMP emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm12B_memmove_move_33through64 + +emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm12B_memmove_move_1or2: + MOVB (CX), SI + MOVB -1(CX)(BX*1), CL + MOVB SI, (AX) + MOVB CL, -1(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeSnappyBlockAsm12B + +emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm12B_memmove_move_3: + MOVW (CX), SI + MOVB 2(CX), CL + MOVW SI, (AX) + MOVB CL, 2(AX) + JMP memmove_end_copy_emit_remainder_encodeSnappyBlockAsm12B + +emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm12B_memmove_move_4through7: + MOVL (CX), SI + MOVL -4(CX)(BX*1), CX + MOVL SI, (AX) + MOVL CX, -4(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeSnappyBlockAsm12B + +emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm12B_memmove_move_8through16: + MOVQ (CX), SI + MOVQ -8(CX)(BX*1), CX + MOVQ SI, (AX) + MOVQ CX, -8(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeSnappyBlockAsm12B + +emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm12B_memmove_move_17through32: + MOVOU (CX), X0 + MOVOU -16(CX)(BX*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeSnappyBlockAsm12B + +emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm12B_memmove_move_33through64: + MOVOU (CX), X0 + MOVOU 16(CX), X1 + MOVOU -32(CX)(BX*1), X2 + MOVOU -16(CX)(BX*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(BX*1) + MOVOU X3, -16(AX)(BX*1) + +memmove_end_copy_emit_remainder_encodeSnappyBlockAsm12B: + MOVQ DX, AX + JMP emit_literal_done_emit_remainder_encodeSnappyBlockAsm12B + +memmove_long_emit_remainder_encodeSnappyBlockAsm12B: + LEAQ (AX)(SI*1), DX + MOVL SI, BX + + // genMemMoveLong + MOVOU (CX), X0 + MOVOU 16(CX), X1 + MOVOU -32(CX)(BX*1), X2 + MOVOU -16(CX)(BX*1), X3 + MOVQ BX, DI + SHRQ $0x05, DI + MOVQ AX, SI + ANDL $0x0000001f, SI + MOVQ $0x00000040, R8 + SUBQ SI, R8 + DECQ DI + JA emit_lit_memmove_long_emit_remainder_encodeSnappyBlockAsm12Blarge_forward_sse_loop_32 + LEAQ -32(CX)(R8*1), SI + LEAQ -32(AX)(R8*1), R9 + +emit_lit_memmove_long_emit_remainder_encodeSnappyBlockAsm12Blarge_big_loop_back: + MOVOU (SI), X4 + MOVOU 16(SI), X5 + MOVOA X4, (R9) + MOVOA X5, 16(R9) + ADDQ $0x20, R9 + ADDQ $0x20, SI + ADDQ $0x20, R8 + DECQ DI + JNA emit_lit_memmove_long_emit_remainder_encodeSnappyBlockAsm12Blarge_big_loop_back + +emit_lit_memmove_long_emit_remainder_encodeSnappyBlockAsm12Blarge_forward_sse_loop_32: + MOVOU -32(CX)(R8*1), X4 + MOVOU -16(CX)(R8*1), X5 + MOVOA X4, -32(AX)(R8*1) + MOVOA X5, -16(AX)(R8*1) + ADDQ $0x20, R8 + CMPQ BX, R8 + JAE emit_lit_memmove_long_emit_remainder_encodeSnappyBlockAsm12Blarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(BX*1) + MOVOU X3, -16(AX)(BX*1) + MOVQ DX, AX + +emit_literal_done_emit_remainder_encodeSnappyBlockAsm12B: + MOVQ dst_base+0(FP), CX + SUBQ CX, AX + MOVQ AX, ret+48(FP) + RET + +// func encodeSnappyBlockAsm10B(dst []byte, src []byte) int +// Requires: BMI, SSE2 +TEXT ·encodeSnappyBlockAsm10B(SB), $4120-56 + MOVQ dst_base+0(FP), AX + MOVQ $0x00000020, CX + LEAQ 24(SP), DX + PXOR X0, X0 + +zero_loop_encodeSnappyBlockAsm10B: + MOVOU X0, (DX) + MOVOU X0, 16(DX) + MOVOU X0, 32(DX) + MOVOU X0, 48(DX) + MOVOU X0, 64(DX) + MOVOU X0, 80(DX) + MOVOU X0, 96(DX) + MOVOU X0, 112(DX) + ADDQ $0x80, DX + DECQ CX + JNZ zero_loop_encodeSnappyBlockAsm10B + MOVL $0x00000000, 12(SP) + MOVQ src_len+32(FP), CX + LEAQ -9(CX), DX + LEAQ -8(CX), BX + MOVL BX, 8(SP) + SHRQ $0x05, CX + SUBL CX, DX + LEAQ (AX)(DX*1), DX + MOVQ DX, (SP) + MOVL $0x00000001, CX + MOVL CX, 16(SP) + MOVQ src_base+24(FP), DX + +search_loop_encodeSnappyBlockAsm10B: + MOVL CX, BX + SUBL 12(SP), BX + SHRL $0x05, BX + LEAL 4(CX)(BX*1), BX + CMPL BX, 8(SP) + JAE emit_remainder_encodeSnappyBlockAsm10B + MOVQ (DX)(CX*1), SI + MOVL BX, 20(SP) + MOVQ $0x9e3779b1, R8 + MOVQ SI, R9 + MOVQ SI, R10 + SHRQ $0x08, R10 + SHLQ $0x20, R9 + IMULQ R8, R9 + SHRQ $0x36, R9 + SHLQ $0x20, R10 + IMULQ R8, R10 + SHRQ $0x36, R10 + MOVL 24(SP)(R9*4), BX + MOVL 24(SP)(R10*4), DI + MOVL CX, 24(SP)(R9*4) + LEAL 1(CX), R9 + MOVL R9, 24(SP)(R10*4) + MOVQ SI, R9 + SHRQ $0x10, R9 + SHLQ $0x20, R9 + IMULQ R8, R9 + SHRQ $0x36, R9 + MOVL CX, R8 + SUBL 16(SP), R8 + MOVL 1(DX)(R8*1), R10 + MOVQ SI, R8 + SHRQ $0x08, R8 + CMPL R8, R10 + JNE no_repeat_found_encodeSnappyBlockAsm10B + LEAL 1(CX), SI + MOVL 12(SP), BX + MOVL SI, DI + SUBL 16(SP), DI + JZ repeat_extend_back_end_encodeSnappyBlockAsm10B + +repeat_extend_back_loop_encodeSnappyBlockAsm10B: + CMPL SI, BX + JBE repeat_extend_back_end_encodeSnappyBlockAsm10B + MOVB -1(DX)(DI*1), R8 + MOVB -1(DX)(SI*1), R9 + CMPB R8, R9 + JNE repeat_extend_back_end_encodeSnappyBlockAsm10B + LEAL -1(SI), SI + DECL DI + JNZ repeat_extend_back_loop_encodeSnappyBlockAsm10B + +repeat_extend_back_end_encodeSnappyBlockAsm10B: + MOVL SI, BX + SUBL 12(SP), BX + LEAQ 3(AX)(BX*1), BX + CMPQ BX, (SP) + JB repeat_dst_size_check_encodeSnappyBlockAsm10B + MOVQ $0x00000000, ret+48(FP) + RET + +repeat_dst_size_check_encodeSnappyBlockAsm10B: + MOVL 12(SP), BX + CMPL BX, SI + JEQ emit_literal_done_repeat_emit_encodeSnappyBlockAsm10B + MOVL SI, DI + MOVL SI, 12(SP) + LEAQ (DX)(BX*1), R8 + SUBL BX, DI + LEAL -1(DI), BX + CMPL BX, $0x3c + JB one_byte_repeat_emit_encodeSnappyBlockAsm10B + CMPL BX, $0x00000100 + JB two_bytes_repeat_emit_encodeSnappyBlockAsm10B + JB three_bytes_repeat_emit_encodeSnappyBlockAsm10B + +three_bytes_repeat_emit_encodeSnappyBlockAsm10B: + MOVB $0xf4, (AX) + MOVW BX, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_repeat_emit_encodeSnappyBlockAsm10B + +two_bytes_repeat_emit_encodeSnappyBlockAsm10B: + MOVB $0xf0, (AX) + MOVB BL, 1(AX) + ADDQ $0x02, AX + CMPL BX, $0x40 + JB memmove_repeat_emit_encodeSnappyBlockAsm10B + JMP memmove_long_repeat_emit_encodeSnappyBlockAsm10B + +one_byte_repeat_emit_encodeSnappyBlockAsm10B: + SHLB $0x02, BL + MOVB BL, (AX) + ADDQ $0x01, AX + +memmove_repeat_emit_encodeSnappyBlockAsm10B: + LEAQ (AX)(DI*1), BX + + // genMemMoveShort + CMPQ DI, $0x08 + JBE emit_lit_memmove_repeat_emit_encodeSnappyBlockAsm10B_memmove_move_8 + CMPQ DI, $0x10 + JBE emit_lit_memmove_repeat_emit_encodeSnappyBlockAsm10B_memmove_move_8through16 + CMPQ DI, $0x20 + JBE emit_lit_memmove_repeat_emit_encodeSnappyBlockAsm10B_memmove_move_17through32 + JMP emit_lit_memmove_repeat_emit_encodeSnappyBlockAsm10B_memmove_move_33through64 + +emit_lit_memmove_repeat_emit_encodeSnappyBlockAsm10B_memmove_move_8: + MOVQ (R8), R9 + MOVQ R9, (AX) + JMP memmove_end_copy_repeat_emit_encodeSnappyBlockAsm10B + +emit_lit_memmove_repeat_emit_encodeSnappyBlockAsm10B_memmove_move_8through16: + MOVQ (R8), R9 + MOVQ -8(R8)(DI*1), R8 + MOVQ R9, (AX) + MOVQ R8, -8(AX)(DI*1) + JMP memmove_end_copy_repeat_emit_encodeSnappyBlockAsm10B + +emit_lit_memmove_repeat_emit_encodeSnappyBlockAsm10B_memmove_move_17through32: + MOVOU (R8), X0 + MOVOU -16(R8)(DI*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(DI*1) + JMP memmove_end_copy_repeat_emit_encodeSnappyBlockAsm10B + +emit_lit_memmove_repeat_emit_encodeSnappyBlockAsm10B_memmove_move_33through64: + MOVOU (R8), X0 + MOVOU 16(R8), X1 + MOVOU -32(R8)(DI*1), X2 + MOVOU -16(R8)(DI*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(DI*1) + MOVOU X3, -16(AX)(DI*1) + +memmove_end_copy_repeat_emit_encodeSnappyBlockAsm10B: + MOVQ BX, AX + JMP emit_literal_done_repeat_emit_encodeSnappyBlockAsm10B + +memmove_long_repeat_emit_encodeSnappyBlockAsm10B: + LEAQ (AX)(DI*1), BX + + // genMemMoveLong + MOVOU (R8), X0 + MOVOU 16(R8), X1 + MOVOU -32(R8)(DI*1), X2 + MOVOU -16(R8)(DI*1), X3 + MOVQ DI, R10 + SHRQ $0x05, R10 + MOVQ AX, R9 + ANDL $0x0000001f, R9 + MOVQ $0x00000040, R11 + SUBQ R9, R11 + DECQ R10 + JA emit_lit_memmove_long_repeat_emit_encodeSnappyBlockAsm10Blarge_forward_sse_loop_32 + LEAQ -32(R8)(R11*1), R9 + LEAQ -32(AX)(R11*1), R12 + +emit_lit_memmove_long_repeat_emit_encodeSnappyBlockAsm10Blarge_big_loop_back: + MOVOU (R9), X4 + MOVOU 16(R9), X5 + MOVOA X4, (R12) + MOVOA X5, 16(R12) + ADDQ $0x20, R12 + ADDQ $0x20, R9 + ADDQ $0x20, R11 + DECQ R10 + JNA emit_lit_memmove_long_repeat_emit_encodeSnappyBlockAsm10Blarge_big_loop_back + +emit_lit_memmove_long_repeat_emit_encodeSnappyBlockAsm10Blarge_forward_sse_loop_32: + MOVOU -32(R8)(R11*1), X4 + MOVOU -16(R8)(R11*1), X5 + MOVOA X4, -32(AX)(R11*1) + MOVOA X5, -16(AX)(R11*1) + ADDQ $0x20, R11 + CMPQ DI, R11 + JAE emit_lit_memmove_long_repeat_emit_encodeSnappyBlockAsm10Blarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(DI*1) + MOVOU X3, -16(AX)(DI*1) + MOVQ BX, AX + +emit_literal_done_repeat_emit_encodeSnappyBlockAsm10B: + ADDL $0x05, CX + MOVL CX, BX + SUBL 16(SP), BX + MOVQ src_len+32(FP), DI + SUBL CX, DI + LEAQ (DX)(CX*1), R8 + LEAQ (DX)(BX*1), BX + + // matchLen + XORL R10, R10 + +matchlen_loopback_16_repeat_extend_encodeSnappyBlockAsm10B: + CMPL DI, $0x10 + JB matchlen_match8_repeat_extend_encodeSnappyBlockAsm10B + MOVQ (R8)(R10*1), R9 + MOVQ 8(R8)(R10*1), R11 + XORQ (BX)(R10*1), R9 + JNZ matchlen_bsf_8_repeat_extend_encodeSnappyBlockAsm10B + XORQ 8(BX)(R10*1), R11 + JNZ matchlen_bsf_16repeat_extend_encodeSnappyBlockAsm10B + LEAL -16(DI), DI + LEAL 16(R10), R10 + JMP matchlen_loopback_16_repeat_extend_encodeSnappyBlockAsm10B + +matchlen_bsf_16repeat_extend_encodeSnappyBlockAsm10B: +#ifdef GOAMD64_v3 + TZCNTQ R11, R11 + +#else + BSFQ R11, R11 + +#endif + SARQ $0x03, R11 + LEAL 8(R10)(R11*1), R10 + JMP repeat_extend_forward_end_encodeSnappyBlockAsm10B + +matchlen_match8_repeat_extend_encodeSnappyBlockAsm10B: + CMPL DI, $0x08 + JB matchlen_match4_repeat_extend_encodeSnappyBlockAsm10B + MOVQ (R8)(R10*1), R9 + XORQ (BX)(R10*1), R9 + JNZ matchlen_bsf_8_repeat_extend_encodeSnappyBlockAsm10B + LEAL -8(DI), DI + LEAL 8(R10), R10 + JMP matchlen_match4_repeat_extend_encodeSnappyBlockAsm10B + +matchlen_bsf_8_repeat_extend_encodeSnappyBlockAsm10B: +#ifdef GOAMD64_v3 + TZCNTQ R9, R9 + +#else + BSFQ R9, R9 + +#endif + SARQ $0x03, R9 + LEAL (R10)(R9*1), R10 + JMP repeat_extend_forward_end_encodeSnappyBlockAsm10B + +matchlen_match4_repeat_extend_encodeSnappyBlockAsm10B: + CMPL DI, $0x04 + JB matchlen_match2_repeat_extend_encodeSnappyBlockAsm10B + MOVL (R8)(R10*1), R9 + CMPL (BX)(R10*1), R9 + JNE matchlen_match2_repeat_extend_encodeSnappyBlockAsm10B + LEAL -4(DI), DI + LEAL 4(R10), R10 + +matchlen_match2_repeat_extend_encodeSnappyBlockAsm10B: + CMPL DI, $0x01 + JE matchlen_match1_repeat_extend_encodeSnappyBlockAsm10B + JB repeat_extend_forward_end_encodeSnappyBlockAsm10B + MOVW (R8)(R10*1), R9 + CMPW (BX)(R10*1), R9 + JNE matchlen_match1_repeat_extend_encodeSnappyBlockAsm10B + LEAL 2(R10), R10 + SUBL $0x02, DI + JZ repeat_extend_forward_end_encodeSnappyBlockAsm10B + +matchlen_match1_repeat_extend_encodeSnappyBlockAsm10B: + MOVB (R8)(R10*1), R9 + CMPB (BX)(R10*1), R9 + JNE repeat_extend_forward_end_encodeSnappyBlockAsm10B + LEAL 1(R10), R10 + +repeat_extend_forward_end_encodeSnappyBlockAsm10B: + ADDL R10, CX + MOVL CX, BX + SUBL SI, BX + MOVL 16(SP), SI + + // emitCopy +two_byte_offset_repeat_as_copy_encodeSnappyBlockAsm10B: + CMPL BX, $0x40 + JBE two_byte_offset_short_repeat_as_copy_encodeSnappyBlockAsm10B + MOVB $0xee, (AX) + MOVW SI, 1(AX) + LEAL -60(BX), BX + ADDQ $0x03, AX + JMP two_byte_offset_repeat_as_copy_encodeSnappyBlockAsm10B + +two_byte_offset_short_repeat_as_copy_encodeSnappyBlockAsm10B: + MOVL BX, DI + SHLL $0x02, DI + CMPL BX, $0x0c + JAE emit_copy_three_repeat_as_copy_encodeSnappyBlockAsm10B + CMPL SI, $0x00000800 + JAE emit_copy_three_repeat_as_copy_encodeSnappyBlockAsm10B + LEAL -15(DI), DI + MOVB SI, 1(AX) + SHRL $0x08, SI + SHLL $0x05, SI + ORL SI, DI + MOVB DI, (AX) + ADDQ $0x02, AX + JMP repeat_end_emit_encodeSnappyBlockAsm10B + +emit_copy_three_repeat_as_copy_encodeSnappyBlockAsm10B: + LEAL -2(DI), DI + MOVB DI, (AX) + MOVW SI, 1(AX) + ADDQ $0x03, AX + +repeat_end_emit_encodeSnappyBlockAsm10B: + MOVL CX, 12(SP) + JMP search_loop_encodeSnappyBlockAsm10B + +no_repeat_found_encodeSnappyBlockAsm10B: + CMPL (DX)(BX*1), SI + JEQ candidate_match_encodeSnappyBlockAsm10B + SHRQ $0x08, SI + MOVL 24(SP)(R9*4), BX + LEAL 2(CX), R8 + CMPL (DX)(DI*1), SI + JEQ candidate2_match_encodeSnappyBlockAsm10B + MOVL R8, 24(SP)(R9*4) + SHRQ $0x08, SI + CMPL (DX)(BX*1), SI + JEQ candidate3_match_encodeSnappyBlockAsm10B + MOVL 20(SP), CX + JMP search_loop_encodeSnappyBlockAsm10B + +candidate3_match_encodeSnappyBlockAsm10B: + ADDL $0x02, CX + JMP candidate_match_encodeSnappyBlockAsm10B + +candidate2_match_encodeSnappyBlockAsm10B: + MOVL R8, 24(SP)(R9*4) + INCL CX + MOVL DI, BX + +candidate_match_encodeSnappyBlockAsm10B: + MOVL 12(SP), SI + TESTL BX, BX + JZ match_extend_back_end_encodeSnappyBlockAsm10B + +match_extend_back_loop_encodeSnappyBlockAsm10B: + CMPL CX, SI + JBE match_extend_back_end_encodeSnappyBlockAsm10B + MOVB -1(DX)(BX*1), DI + MOVB -1(DX)(CX*1), R8 + CMPB DI, R8 + JNE match_extend_back_end_encodeSnappyBlockAsm10B + LEAL -1(CX), CX + DECL BX + JZ match_extend_back_end_encodeSnappyBlockAsm10B + JMP match_extend_back_loop_encodeSnappyBlockAsm10B + +match_extend_back_end_encodeSnappyBlockAsm10B: + MOVL CX, SI + SUBL 12(SP), SI + LEAQ 3(AX)(SI*1), SI + CMPQ SI, (SP) + JB match_dst_size_check_encodeSnappyBlockAsm10B + MOVQ $0x00000000, ret+48(FP) + RET + +match_dst_size_check_encodeSnappyBlockAsm10B: + MOVL CX, SI + MOVL 12(SP), DI + CMPL DI, SI + JEQ emit_literal_done_match_emit_encodeSnappyBlockAsm10B + MOVL SI, R8 + MOVL SI, 12(SP) + LEAQ (DX)(DI*1), SI + SUBL DI, R8 + LEAL -1(R8), DI + CMPL DI, $0x3c + JB one_byte_match_emit_encodeSnappyBlockAsm10B + CMPL DI, $0x00000100 + JB two_bytes_match_emit_encodeSnappyBlockAsm10B + JB three_bytes_match_emit_encodeSnappyBlockAsm10B + +three_bytes_match_emit_encodeSnappyBlockAsm10B: + MOVB $0xf4, (AX) + MOVW DI, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_match_emit_encodeSnappyBlockAsm10B + +two_bytes_match_emit_encodeSnappyBlockAsm10B: + MOVB $0xf0, (AX) + MOVB DI, 1(AX) + ADDQ $0x02, AX + CMPL DI, $0x40 + JB memmove_match_emit_encodeSnappyBlockAsm10B + JMP memmove_long_match_emit_encodeSnappyBlockAsm10B + +one_byte_match_emit_encodeSnappyBlockAsm10B: + SHLB $0x02, DI + MOVB DI, (AX) + ADDQ $0x01, AX + +memmove_match_emit_encodeSnappyBlockAsm10B: + LEAQ (AX)(R8*1), DI + + // genMemMoveShort + CMPQ R8, $0x08 + JBE emit_lit_memmove_match_emit_encodeSnappyBlockAsm10B_memmove_move_8 + CMPQ R8, $0x10 + JBE emit_lit_memmove_match_emit_encodeSnappyBlockAsm10B_memmove_move_8through16 + CMPQ R8, $0x20 + JBE emit_lit_memmove_match_emit_encodeSnappyBlockAsm10B_memmove_move_17through32 + JMP emit_lit_memmove_match_emit_encodeSnappyBlockAsm10B_memmove_move_33through64 + +emit_lit_memmove_match_emit_encodeSnappyBlockAsm10B_memmove_move_8: + MOVQ (SI), R9 + MOVQ R9, (AX) + JMP memmove_end_copy_match_emit_encodeSnappyBlockAsm10B + +emit_lit_memmove_match_emit_encodeSnappyBlockAsm10B_memmove_move_8through16: + MOVQ (SI), R9 + MOVQ -8(SI)(R8*1), SI + MOVQ R9, (AX) + MOVQ SI, -8(AX)(R8*1) + JMP memmove_end_copy_match_emit_encodeSnappyBlockAsm10B + +emit_lit_memmove_match_emit_encodeSnappyBlockAsm10B_memmove_move_17through32: + MOVOU (SI), X0 + MOVOU -16(SI)(R8*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(R8*1) + JMP memmove_end_copy_match_emit_encodeSnappyBlockAsm10B + +emit_lit_memmove_match_emit_encodeSnappyBlockAsm10B_memmove_move_33through64: + MOVOU (SI), X0 + MOVOU 16(SI), X1 + MOVOU -32(SI)(R8*1), X2 + MOVOU -16(SI)(R8*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + +memmove_end_copy_match_emit_encodeSnappyBlockAsm10B: + MOVQ DI, AX + JMP emit_literal_done_match_emit_encodeSnappyBlockAsm10B + +memmove_long_match_emit_encodeSnappyBlockAsm10B: + LEAQ (AX)(R8*1), DI + + // genMemMoveLong + MOVOU (SI), X0 + MOVOU 16(SI), X1 + MOVOU -32(SI)(R8*1), X2 + MOVOU -16(SI)(R8*1), X3 + MOVQ R8, R10 + SHRQ $0x05, R10 + MOVQ AX, R9 + ANDL $0x0000001f, R9 + MOVQ $0x00000040, R11 + SUBQ R9, R11 + DECQ R10 + JA emit_lit_memmove_long_match_emit_encodeSnappyBlockAsm10Blarge_forward_sse_loop_32 + LEAQ -32(SI)(R11*1), R9 + LEAQ -32(AX)(R11*1), R12 + +emit_lit_memmove_long_match_emit_encodeSnappyBlockAsm10Blarge_big_loop_back: + MOVOU (R9), X4 + MOVOU 16(R9), X5 + MOVOA X4, (R12) + MOVOA X5, 16(R12) + ADDQ $0x20, R12 + ADDQ $0x20, R9 + ADDQ $0x20, R11 + DECQ R10 + JNA emit_lit_memmove_long_match_emit_encodeSnappyBlockAsm10Blarge_big_loop_back + +emit_lit_memmove_long_match_emit_encodeSnappyBlockAsm10Blarge_forward_sse_loop_32: + MOVOU -32(SI)(R11*1), X4 + MOVOU -16(SI)(R11*1), X5 + MOVOA X4, -32(AX)(R11*1) + MOVOA X5, -16(AX)(R11*1) + ADDQ $0x20, R11 + CMPQ R8, R11 + JAE emit_lit_memmove_long_match_emit_encodeSnappyBlockAsm10Blarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + MOVQ DI, AX + +emit_literal_done_match_emit_encodeSnappyBlockAsm10B: +match_nolit_loop_encodeSnappyBlockAsm10B: + MOVL CX, SI + SUBL BX, SI + MOVL SI, 16(SP) + ADDL $0x04, CX + ADDL $0x04, BX + MOVQ src_len+32(FP), SI + SUBL CX, SI + LEAQ (DX)(CX*1), DI + LEAQ (DX)(BX*1), BX + + // matchLen + XORL R9, R9 + +matchlen_loopback_16_match_nolit_encodeSnappyBlockAsm10B: + CMPL SI, $0x10 + JB matchlen_match8_match_nolit_encodeSnappyBlockAsm10B + MOVQ (DI)(R9*1), R8 + MOVQ 8(DI)(R9*1), R10 + XORQ (BX)(R9*1), R8 + JNZ matchlen_bsf_8_match_nolit_encodeSnappyBlockAsm10B + XORQ 8(BX)(R9*1), R10 + JNZ matchlen_bsf_16match_nolit_encodeSnappyBlockAsm10B + LEAL -16(SI), SI + LEAL 16(R9), R9 + JMP matchlen_loopback_16_match_nolit_encodeSnappyBlockAsm10B + +matchlen_bsf_16match_nolit_encodeSnappyBlockAsm10B: +#ifdef GOAMD64_v3 + TZCNTQ R10, R10 + +#else + BSFQ R10, R10 + +#endif + SARQ $0x03, R10 + LEAL 8(R9)(R10*1), R9 + JMP match_nolit_end_encodeSnappyBlockAsm10B + +matchlen_match8_match_nolit_encodeSnappyBlockAsm10B: + CMPL SI, $0x08 + JB matchlen_match4_match_nolit_encodeSnappyBlockAsm10B + MOVQ (DI)(R9*1), R8 + XORQ (BX)(R9*1), R8 + JNZ matchlen_bsf_8_match_nolit_encodeSnappyBlockAsm10B + LEAL -8(SI), SI + LEAL 8(R9), R9 + JMP matchlen_match4_match_nolit_encodeSnappyBlockAsm10B + +matchlen_bsf_8_match_nolit_encodeSnappyBlockAsm10B: +#ifdef GOAMD64_v3 + TZCNTQ R8, R8 + +#else + BSFQ R8, R8 + +#endif + SARQ $0x03, R8 + LEAL (R9)(R8*1), R9 + JMP match_nolit_end_encodeSnappyBlockAsm10B + +matchlen_match4_match_nolit_encodeSnappyBlockAsm10B: + CMPL SI, $0x04 + JB matchlen_match2_match_nolit_encodeSnappyBlockAsm10B + MOVL (DI)(R9*1), R8 + CMPL (BX)(R9*1), R8 + JNE matchlen_match2_match_nolit_encodeSnappyBlockAsm10B + LEAL -4(SI), SI + LEAL 4(R9), R9 + +matchlen_match2_match_nolit_encodeSnappyBlockAsm10B: + CMPL SI, $0x01 + JE matchlen_match1_match_nolit_encodeSnappyBlockAsm10B + JB match_nolit_end_encodeSnappyBlockAsm10B + MOVW (DI)(R9*1), R8 + CMPW (BX)(R9*1), R8 + JNE matchlen_match1_match_nolit_encodeSnappyBlockAsm10B + LEAL 2(R9), R9 + SUBL $0x02, SI + JZ match_nolit_end_encodeSnappyBlockAsm10B + +matchlen_match1_match_nolit_encodeSnappyBlockAsm10B: + MOVB (DI)(R9*1), R8 + CMPB (BX)(R9*1), R8 + JNE match_nolit_end_encodeSnappyBlockAsm10B + LEAL 1(R9), R9 + +match_nolit_end_encodeSnappyBlockAsm10B: + ADDL R9, CX + MOVL 16(SP), BX + ADDL $0x04, R9 + MOVL CX, 12(SP) + + // emitCopy +two_byte_offset_match_nolit_encodeSnappyBlockAsm10B: + CMPL R9, $0x40 + JBE two_byte_offset_short_match_nolit_encodeSnappyBlockAsm10B + MOVB $0xee, (AX) + MOVW BX, 1(AX) + LEAL -60(R9), R9 + ADDQ $0x03, AX + JMP two_byte_offset_match_nolit_encodeSnappyBlockAsm10B + +two_byte_offset_short_match_nolit_encodeSnappyBlockAsm10B: + MOVL R9, SI + SHLL $0x02, SI + CMPL R9, $0x0c + JAE emit_copy_three_match_nolit_encodeSnappyBlockAsm10B + CMPL BX, $0x00000800 + JAE emit_copy_three_match_nolit_encodeSnappyBlockAsm10B + LEAL -15(SI), SI + MOVB BL, 1(AX) + SHRL $0x08, BX + SHLL $0x05, BX + ORL BX, SI + MOVB SI, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeSnappyBlockAsm10B + +emit_copy_three_match_nolit_encodeSnappyBlockAsm10B: + LEAL -2(SI), SI + MOVB SI, (AX) + MOVW BX, 1(AX) + ADDQ $0x03, AX + +match_nolit_emitcopy_end_encodeSnappyBlockAsm10B: + CMPL CX, 8(SP) + JAE emit_remainder_encodeSnappyBlockAsm10B + MOVQ -2(DX)(CX*1), SI + CMPQ AX, (SP) + JB match_nolit_dst_ok_encodeSnappyBlockAsm10B + MOVQ $0x00000000, ret+48(FP) + RET + +match_nolit_dst_ok_encodeSnappyBlockAsm10B: + MOVQ $0x9e3779b1, R8 + MOVQ SI, DI + SHRQ $0x10, SI + MOVQ SI, BX + SHLQ $0x20, DI + IMULQ R8, DI + SHRQ $0x36, DI + SHLQ $0x20, BX + IMULQ R8, BX + SHRQ $0x36, BX + LEAL -2(CX), R8 + LEAQ 24(SP)(BX*4), R9 + MOVL (R9), BX + MOVL R8, 24(SP)(DI*4) + MOVL CX, (R9) + CMPL (DX)(BX*1), SI + JEQ match_nolit_loop_encodeSnappyBlockAsm10B + INCL CX + JMP search_loop_encodeSnappyBlockAsm10B + +emit_remainder_encodeSnappyBlockAsm10B: + MOVQ src_len+32(FP), CX + SUBL 12(SP), CX + LEAQ 3(AX)(CX*1), CX + CMPQ CX, (SP) + JB emit_remainder_ok_encodeSnappyBlockAsm10B + MOVQ $0x00000000, ret+48(FP) + RET + +emit_remainder_ok_encodeSnappyBlockAsm10B: + MOVQ src_len+32(FP), CX + MOVL 12(SP), BX + CMPL BX, CX + JEQ emit_literal_done_emit_remainder_encodeSnappyBlockAsm10B + MOVL CX, SI + MOVL CX, 12(SP) + LEAQ (DX)(BX*1), CX + SUBL BX, SI + LEAL -1(SI), DX + CMPL DX, $0x3c + JB one_byte_emit_remainder_encodeSnappyBlockAsm10B + CMPL DX, $0x00000100 + JB two_bytes_emit_remainder_encodeSnappyBlockAsm10B + JB three_bytes_emit_remainder_encodeSnappyBlockAsm10B + +three_bytes_emit_remainder_encodeSnappyBlockAsm10B: + MOVB $0xf4, (AX) + MOVW DX, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_emit_remainder_encodeSnappyBlockAsm10B + +two_bytes_emit_remainder_encodeSnappyBlockAsm10B: + MOVB $0xf0, (AX) + MOVB DL, 1(AX) + ADDQ $0x02, AX + CMPL DX, $0x40 + JB memmove_emit_remainder_encodeSnappyBlockAsm10B + JMP memmove_long_emit_remainder_encodeSnappyBlockAsm10B + +one_byte_emit_remainder_encodeSnappyBlockAsm10B: + SHLB $0x02, DL + MOVB DL, (AX) + ADDQ $0x01, AX + +memmove_emit_remainder_encodeSnappyBlockAsm10B: + LEAQ (AX)(SI*1), DX + MOVL SI, BX + + // genMemMoveShort + CMPQ BX, $0x03 + JB emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm10B_memmove_move_1or2 + JE emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm10B_memmove_move_3 + CMPQ BX, $0x08 + JB emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm10B_memmove_move_4through7 + CMPQ BX, $0x10 + JBE emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm10B_memmove_move_8through16 + CMPQ BX, $0x20 + JBE emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm10B_memmove_move_17through32 + JMP emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm10B_memmove_move_33through64 + +emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm10B_memmove_move_1or2: + MOVB (CX), SI + MOVB -1(CX)(BX*1), CL + MOVB SI, (AX) + MOVB CL, -1(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeSnappyBlockAsm10B + +emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm10B_memmove_move_3: + MOVW (CX), SI + MOVB 2(CX), CL + MOVW SI, (AX) + MOVB CL, 2(AX) + JMP memmove_end_copy_emit_remainder_encodeSnappyBlockAsm10B + +emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm10B_memmove_move_4through7: + MOVL (CX), SI + MOVL -4(CX)(BX*1), CX + MOVL SI, (AX) + MOVL CX, -4(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeSnappyBlockAsm10B + +emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm10B_memmove_move_8through16: + MOVQ (CX), SI + MOVQ -8(CX)(BX*1), CX + MOVQ SI, (AX) + MOVQ CX, -8(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeSnappyBlockAsm10B + +emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm10B_memmove_move_17through32: + MOVOU (CX), X0 + MOVOU -16(CX)(BX*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeSnappyBlockAsm10B + +emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm10B_memmove_move_33through64: + MOVOU (CX), X0 + MOVOU 16(CX), X1 + MOVOU -32(CX)(BX*1), X2 + MOVOU -16(CX)(BX*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(BX*1) + MOVOU X3, -16(AX)(BX*1) + +memmove_end_copy_emit_remainder_encodeSnappyBlockAsm10B: + MOVQ DX, AX + JMP emit_literal_done_emit_remainder_encodeSnappyBlockAsm10B + +memmove_long_emit_remainder_encodeSnappyBlockAsm10B: + LEAQ (AX)(SI*1), DX + MOVL SI, BX + + // genMemMoveLong + MOVOU (CX), X0 + MOVOU 16(CX), X1 + MOVOU -32(CX)(BX*1), X2 + MOVOU -16(CX)(BX*1), X3 + MOVQ BX, DI + SHRQ $0x05, DI + MOVQ AX, SI + ANDL $0x0000001f, SI + MOVQ $0x00000040, R8 + SUBQ SI, R8 + DECQ DI + JA emit_lit_memmove_long_emit_remainder_encodeSnappyBlockAsm10Blarge_forward_sse_loop_32 + LEAQ -32(CX)(R8*1), SI + LEAQ -32(AX)(R8*1), R9 + +emit_lit_memmove_long_emit_remainder_encodeSnappyBlockAsm10Blarge_big_loop_back: + MOVOU (SI), X4 + MOVOU 16(SI), X5 + MOVOA X4, (R9) + MOVOA X5, 16(R9) + ADDQ $0x20, R9 + ADDQ $0x20, SI + ADDQ $0x20, R8 + DECQ DI + JNA emit_lit_memmove_long_emit_remainder_encodeSnappyBlockAsm10Blarge_big_loop_back + +emit_lit_memmove_long_emit_remainder_encodeSnappyBlockAsm10Blarge_forward_sse_loop_32: + MOVOU -32(CX)(R8*1), X4 + MOVOU -16(CX)(R8*1), X5 + MOVOA X4, -32(AX)(R8*1) + MOVOA X5, -16(AX)(R8*1) + ADDQ $0x20, R8 + CMPQ BX, R8 + JAE emit_lit_memmove_long_emit_remainder_encodeSnappyBlockAsm10Blarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(BX*1) + MOVOU X3, -16(AX)(BX*1) + MOVQ DX, AX + +emit_literal_done_emit_remainder_encodeSnappyBlockAsm10B: + MOVQ dst_base+0(FP), CX + SUBQ CX, AX + MOVQ AX, ret+48(FP) + RET + +// func encodeSnappyBlockAsm8B(dst []byte, src []byte) int +// Requires: BMI, SSE2 +TEXT ·encodeSnappyBlockAsm8B(SB), $1048-56 + MOVQ dst_base+0(FP), AX + MOVQ $0x00000008, CX + LEAQ 24(SP), DX + PXOR X0, X0 + +zero_loop_encodeSnappyBlockAsm8B: + MOVOU X0, (DX) + MOVOU X0, 16(DX) + MOVOU X0, 32(DX) + MOVOU X0, 48(DX) + MOVOU X0, 64(DX) + MOVOU X0, 80(DX) + MOVOU X0, 96(DX) + MOVOU X0, 112(DX) + ADDQ $0x80, DX + DECQ CX + JNZ zero_loop_encodeSnappyBlockAsm8B + MOVL $0x00000000, 12(SP) + MOVQ src_len+32(FP), CX + LEAQ -9(CX), DX + LEAQ -8(CX), BX + MOVL BX, 8(SP) + SHRQ $0x05, CX + SUBL CX, DX + LEAQ (AX)(DX*1), DX + MOVQ DX, (SP) + MOVL $0x00000001, CX + MOVL CX, 16(SP) + MOVQ src_base+24(FP), DX + +search_loop_encodeSnappyBlockAsm8B: + MOVL CX, BX + SUBL 12(SP), BX + SHRL $0x04, BX + LEAL 4(CX)(BX*1), BX + CMPL BX, 8(SP) + JAE emit_remainder_encodeSnappyBlockAsm8B + MOVQ (DX)(CX*1), SI + MOVL BX, 20(SP) + MOVQ $0x9e3779b1, R8 + MOVQ SI, R9 + MOVQ SI, R10 + SHRQ $0x08, R10 + SHLQ $0x20, R9 + IMULQ R8, R9 + SHRQ $0x38, R9 + SHLQ $0x20, R10 + IMULQ R8, R10 + SHRQ $0x38, R10 + MOVL 24(SP)(R9*4), BX + MOVL 24(SP)(R10*4), DI + MOVL CX, 24(SP)(R9*4) + LEAL 1(CX), R9 + MOVL R9, 24(SP)(R10*4) + MOVQ SI, R9 + SHRQ $0x10, R9 + SHLQ $0x20, R9 + IMULQ R8, R9 + SHRQ $0x38, R9 + MOVL CX, R8 + SUBL 16(SP), R8 + MOVL 1(DX)(R8*1), R10 + MOVQ SI, R8 + SHRQ $0x08, R8 + CMPL R8, R10 + JNE no_repeat_found_encodeSnappyBlockAsm8B + LEAL 1(CX), SI + MOVL 12(SP), BX + MOVL SI, DI + SUBL 16(SP), DI + JZ repeat_extend_back_end_encodeSnappyBlockAsm8B + +repeat_extend_back_loop_encodeSnappyBlockAsm8B: + CMPL SI, BX + JBE repeat_extend_back_end_encodeSnappyBlockAsm8B + MOVB -1(DX)(DI*1), R8 + MOVB -1(DX)(SI*1), R9 + CMPB R8, R9 + JNE repeat_extend_back_end_encodeSnappyBlockAsm8B + LEAL -1(SI), SI + DECL DI + JNZ repeat_extend_back_loop_encodeSnappyBlockAsm8B + +repeat_extend_back_end_encodeSnappyBlockAsm8B: + MOVL SI, BX + SUBL 12(SP), BX + LEAQ 3(AX)(BX*1), BX + CMPQ BX, (SP) + JB repeat_dst_size_check_encodeSnappyBlockAsm8B + MOVQ $0x00000000, ret+48(FP) + RET + +repeat_dst_size_check_encodeSnappyBlockAsm8B: + MOVL 12(SP), BX + CMPL BX, SI + JEQ emit_literal_done_repeat_emit_encodeSnappyBlockAsm8B + MOVL SI, DI + MOVL SI, 12(SP) + LEAQ (DX)(BX*1), R8 + SUBL BX, DI + LEAL -1(DI), BX + CMPL BX, $0x3c + JB one_byte_repeat_emit_encodeSnappyBlockAsm8B + CMPL BX, $0x00000100 + JB two_bytes_repeat_emit_encodeSnappyBlockAsm8B + JB three_bytes_repeat_emit_encodeSnappyBlockAsm8B + +three_bytes_repeat_emit_encodeSnappyBlockAsm8B: + MOVB $0xf4, (AX) + MOVW BX, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_repeat_emit_encodeSnappyBlockAsm8B + +two_bytes_repeat_emit_encodeSnappyBlockAsm8B: + MOVB $0xf0, (AX) + MOVB BL, 1(AX) + ADDQ $0x02, AX + CMPL BX, $0x40 + JB memmove_repeat_emit_encodeSnappyBlockAsm8B + JMP memmove_long_repeat_emit_encodeSnappyBlockAsm8B + +one_byte_repeat_emit_encodeSnappyBlockAsm8B: + SHLB $0x02, BL + MOVB BL, (AX) + ADDQ $0x01, AX + +memmove_repeat_emit_encodeSnappyBlockAsm8B: + LEAQ (AX)(DI*1), BX + + // genMemMoveShort + CMPQ DI, $0x08 + JBE emit_lit_memmove_repeat_emit_encodeSnappyBlockAsm8B_memmove_move_8 + CMPQ DI, $0x10 + JBE emit_lit_memmove_repeat_emit_encodeSnappyBlockAsm8B_memmove_move_8through16 + CMPQ DI, $0x20 + JBE emit_lit_memmove_repeat_emit_encodeSnappyBlockAsm8B_memmove_move_17through32 + JMP emit_lit_memmove_repeat_emit_encodeSnappyBlockAsm8B_memmove_move_33through64 + +emit_lit_memmove_repeat_emit_encodeSnappyBlockAsm8B_memmove_move_8: + MOVQ (R8), R9 + MOVQ R9, (AX) + JMP memmove_end_copy_repeat_emit_encodeSnappyBlockAsm8B + +emit_lit_memmove_repeat_emit_encodeSnappyBlockAsm8B_memmove_move_8through16: + MOVQ (R8), R9 + MOVQ -8(R8)(DI*1), R8 + MOVQ R9, (AX) + MOVQ R8, -8(AX)(DI*1) + JMP memmove_end_copy_repeat_emit_encodeSnappyBlockAsm8B + +emit_lit_memmove_repeat_emit_encodeSnappyBlockAsm8B_memmove_move_17through32: + MOVOU (R8), X0 + MOVOU -16(R8)(DI*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(DI*1) + JMP memmove_end_copy_repeat_emit_encodeSnappyBlockAsm8B + +emit_lit_memmove_repeat_emit_encodeSnappyBlockAsm8B_memmove_move_33through64: + MOVOU (R8), X0 + MOVOU 16(R8), X1 + MOVOU -32(R8)(DI*1), X2 + MOVOU -16(R8)(DI*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(DI*1) + MOVOU X3, -16(AX)(DI*1) + +memmove_end_copy_repeat_emit_encodeSnappyBlockAsm8B: + MOVQ BX, AX + JMP emit_literal_done_repeat_emit_encodeSnappyBlockAsm8B + +memmove_long_repeat_emit_encodeSnappyBlockAsm8B: + LEAQ (AX)(DI*1), BX + + // genMemMoveLong + MOVOU (R8), X0 + MOVOU 16(R8), X1 + MOVOU -32(R8)(DI*1), X2 + MOVOU -16(R8)(DI*1), X3 + MOVQ DI, R10 + SHRQ $0x05, R10 + MOVQ AX, R9 + ANDL $0x0000001f, R9 + MOVQ $0x00000040, R11 + SUBQ R9, R11 + DECQ R10 + JA emit_lit_memmove_long_repeat_emit_encodeSnappyBlockAsm8Blarge_forward_sse_loop_32 + LEAQ -32(R8)(R11*1), R9 + LEAQ -32(AX)(R11*1), R12 + +emit_lit_memmove_long_repeat_emit_encodeSnappyBlockAsm8Blarge_big_loop_back: + MOVOU (R9), X4 + MOVOU 16(R9), X5 + MOVOA X4, (R12) + MOVOA X5, 16(R12) + ADDQ $0x20, R12 + ADDQ $0x20, R9 + ADDQ $0x20, R11 + DECQ R10 + JNA emit_lit_memmove_long_repeat_emit_encodeSnappyBlockAsm8Blarge_big_loop_back + +emit_lit_memmove_long_repeat_emit_encodeSnappyBlockAsm8Blarge_forward_sse_loop_32: + MOVOU -32(R8)(R11*1), X4 + MOVOU -16(R8)(R11*1), X5 + MOVOA X4, -32(AX)(R11*1) + MOVOA X5, -16(AX)(R11*1) + ADDQ $0x20, R11 + CMPQ DI, R11 + JAE emit_lit_memmove_long_repeat_emit_encodeSnappyBlockAsm8Blarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(DI*1) + MOVOU X3, -16(AX)(DI*1) + MOVQ BX, AX + +emit_literal_done_repeat_emit_encodeSnappyBlockAsm8B: + ADDL $0x05, CX + MOVL CX, BX + SUBL 16(SP), BX + MOVQ src_len+32(FP), DI + SUBL CX, DI + LEAQ (DX)(CX*1), R8 + LEAQ (DX)(BX*1), BX + + // matchLen + XORL R10, R10 + +matchlen_loopback_16_repeat_extend_encodeSnappyBlockAsm8B: + CMPL DI, $0x10 + JB matchlen_match8_repeat_extend_encodeSnappyBlockAsm8B + MOVQ (R8)(R10*1), R9 + MOVQ 8(R8)(R10*1), R11 + XORQ (BX)(R10*1), R9 + JNZ matchlen_bsf_8_repeat_extend_encodeSnappyBlockAsm8B + XORQ 8(BX)(R10*1), R11 + JNZ matchlen_bsf_16repeat_extend_encodeSnappyBlockAsm8B + LEAL -16(DI), DI + LEAL 16(R10), R10 + JMP matchlen_loopback_16_repeat_extend_encodeSnappyBlockAsm8B + +matchlen_bsf_16repeat_extend_encodeSnappyBlockAsm8B: +#ifdef GOAMD64_v3 + TZCNTQ R11, R11 + +#else + BSFQ R11, R11 + +#endif + SARQ $0x03, R11 + LEAL 8(R10)(R11*1), R10 + JMP repeat_extend_forward_end_encodeSnappyBlockAsm8B + +matchlen_match8_repeat_extend_encodeSnappyBlockAsm8B: + CMPL DI, $0x08 + JB matchlen_match4_repeat_extend_encodeSnappyBlockAsm8B + MOVQ (R8)(R10*1), R9 + XORQ (BX)(R10*1), R9 + JNZ matchlen_bsf_8_repeat_extend_encodeSnappyBlockAsm8B + LEAL -8(DI), DI + LEAL 8(R10), R10 + JMP matchlen_match4_repeat_extend_encodeSnappyBlockAsm8B + +matchlen_bsf_8_repeat_extend_encodeSnappyBlockAsm8B: +#ifdef GOAMD64_v3 + TZCNTQ R9, R9 + +#else + BSFQ R9, R9 + +#endif + SARQ $0x03, R9 + LEAL (R10)(R9*1), R10 + JMP repeat_extend_forward_end_encodeSnappyBlockAsm8B + +matchlen_match4_repeat_extend_encodeSnappyBlockAsm8B: + CMPL DI, $0x04 + JB matchlen_match2_repeat_extend_encodeSnappyBlockAsm8B + MOVL (R8)(R10*1), R9 + CMPL (BX)(R10*1), R9 + JNE matchlen_match2_repeat_extend_encodeSnappyBlockAsm8B + LEAL -4(DI), DI + LEAL 4(R10), R10 + +matchlen_match2_repeat_extend_encodeSnappyBlockAsm8B: + CMPL DI, $0x01 + JE matchlen_match1_repeat_extend_encodeSnappyBlockAsm8B + JB repeat_extend_forward_end_encodeSnappyBlockAsm8B + MOVW (R8)(R10*1), R9 + CMPW (BX)(R10*1), R9 + JNE matchlen_match1_repeat_extend_encodeSnappyBlockAsm8B + LEAL 2(R10), R10 + SUBL $0x02, DI + JZ repeat_extend_forward_end_encodeSnappyBlockAsm8B + +matchlen_match1_repeat_extend_encodeSnappyBlockAsm8B: + MOVB (R8)(R10*1), R9 + CMPB (BX)(R10*1), R9 + JNE repeat_extend_forward_end_encodeSnappyBlockAsm8B + LEAL 1(R10), R10 + +repeat_extend_forward_end_encodeSnappyBlockAsm8B: + ADDL R10, CX + MOVL CX, BX + SUBL SI, BX + MOVL 16(SP), SI + + // emitCopy +two_byte_offset_repeat_as_copy_encodeSnappyBlockAsm8B: + CMPL BX, $0x40 + JBE two_byte_offset_short_repeat_as_copy_encodeSnappyBlockAsm8B + MOVB $0xee, (AX) + MOVW SI, 1(AX) + LEAL -60(BX), BX + ADDQ $0x03, AX + JMP two_byte_offset_repeat_as_copy_encodeSnappyBlockAsm8B + +two_byte_offset_short_repeat_as_copy_encodeSnappyBlockAsm8B: + MOVL BX, DI + SHLL $0x02, DI + CMPL BX, $0x0c + JAE emit_copy_three_repeat_as_copy_encodeSnappyBlockAsm8B + LEAL -15(DI), DI + MOVB SI, 1(AX) + SHRL $0x08, SI + SHLL $0x05, SI + ORL SI, DI + MOVB DI, (AX) + ADDQ $0x02, AX + JMP repeat_end_emit_encodeSnappyBlockAsm8B + +emit_copy_three_repeat_as_copy_encodeSnappyBlockAsm8B: + LEAL -2(DI), DI + MOVB DI, (AX) + MOVW SI, 1(AX) + ADDQ $0x03, AX + +repeat_end_emit_encodeSnappyBlockAsm8B: + MOVL CX, 12(SP) + JMP search_loop_encodeSnappyBlockAsm8B + +no_repeat_found_encodeSnappyBlockAsm8B: + CMPL (DX)(BX*1), SI + JEQ candidate_match_encodeSnappyBlockAsm8B + SHRQ $0x08, SI + MOVL 24(SP)(R9*4), BX + LEAL 2(CX), R8 + CMPL (DX)(DI*1), SI + JEQ candidate2_match_encodeSnappyBlockAsm8B + MOVL R8, 24(SP)(R9*4) + SHRQ $0x08, SI + CMPL (DX)(BX*1), SI + JEQ candidate3_match_encodeSnappyBlockAsm8B + MOVL 20(SP), CX + JMP search_loop_encodeSnappyBlockAsm8B + +candidate3_match_encodeSnappyBlockAsm8B: + ADDL $0x02, CX + JMP candidate_match_encodeSnappyBlockAsm8B + +candidate2_match_encodeSnappyBlockAsm8B: + MOVL R8, 24(SP)(R9*4) + INCL CX + MOVL DI, BX + +candidate_match_encodeSnappyBlockAsm8B: + MOVL 12(SP), SI + TESTL BX, BX + JZ match_extend_back_end_encodeSnappyBlockAsm8B + +match_extend_back_loop_encodeSnappyBlockAsm8B: + CMPL CX, SI + JBE match_extend_back_end_encodeSnappyBlockAsm8B + MOVB -1(DX)(BX*1), DI + MOVB -1(DX)(CX*1), R8 + CMPB DI, R8 + JNE match_extend_back_end_encodeSnappyBlockAsm8B + LEAL -1(CX), CX + DECL BX + JZ match_extend_back_end_encodeSnappyBlockAsm8B + JMP match_extend_back_loop_encodeSnappyBlockAsm8B + +match_extend_back_end_encodeSnappyBlockAsm8B: + MOVL CX, SI + SUBL 12(SP), SI + LEAQ 3(AX)(SI*1), SI + CMPQ SI, (SP) + JB match_dst_size_check_encodeSnappyBlockAsm8B + MOVQ $0x00000000, ret+48(FP) + RET + +match_dst_size_check_encodeSnappyBlockAsm8B: + MOVL CX, SI + MOVL 12(SP), DI + CMPL DI, SI + JEQ emit_literal_done_match_emit_encodeSnappyBlockAsm8B + MOVL SI, R8 + MOVL SI, 12(SP) + LEAQ (DX)(DI*1), SI + SUBL DI, R8 + LEAL -1(R8), DI + CMPL DI, $0x3c + JB one_byte_match_emit_encodeSnappyBlockAsm8B + CMPL DI, $0x00000100 + JB two_bytes_match_emit_encodeSnappyBlockAsm8B + JB three_bytes_match_emit_encodeSnappyBlockAsm8B + +three_bytes_match_emit_encodeSnappyBlockAsm8B: + MOVB $0xf4, (AX) + MOVW DI, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_match_emit_encodeSnappyBlockAsm8B + +two_bytes_match_emit_encodeSnappyBlockAsm8B: + MOVB $0xf0, (AX) + MOVB DI, 1(AX) + ADDQ $0x02, AX + CMPL DI, $0x40 + JB memmove_match_emit_encodeSnappyBlockAsm8B + JMP memmove_long_match_emit_encodeSnappyBlockAsm8B + +one_byte_match_emit_encodeSnappyBlockAsm8B: + SHLB $0x02, DI + MOVB DI, (AX) + ADDQ $0x01, AX + +memmove_match_emit_encodeSnappyBlockAsm8B: + LEAQ (AX)(R8*1), DI + + // genMemMoveShort + CMPQ R8, $0x08 + JBE emit_lit_memmove_match_emit_encodeSnappyBlockAsm8B_memmove_move_8 + CMPQ R8, $0x10 + JBE emit_lit_memmove_match_emit_encodeSnappyBlockAsm8B_memmove_move_8through16 + CMPQ R8, $0x20 + JBE emit_lit_memmove_match_emit_encodeSnappyBlockAsm8B_memmove_move_17through32 + JMP emit_lit_memmove_match_emit_encodeSnappyBlockAsm8B_memmove_move_33through64 + +emit_lit_memmove_match_emit_encodeSnappyBlockAsm8B_memmove_move_8: + MOVQ (SI), R9 + MOVQ R9, (AX) + JMP memmove_end_copy_match_emit_encodeSnappyBlockAsm8B + +emit_lit_memmove_match_emit_encodeSnappyBlockAsm8B_memmove_move_8through16: + MOVQ (SI), R9 + MOVQ -8(SI)(R8*1), SI + MOVQ R9, (AX) + MOVQ SI, -8(AX)(R8*1) + JMP memmove_end_copy_match_emit_encodeSnappyBlockAsm8B + +emit_lit_memmove_match_emit_encodeSnappyBlockAsm8B_memmove_move_17through32: + MOVOU (SI), X0 + MOVOU -16(SI)(R8*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(R8*1) + JMP memmove_end_copy_match_emit_encodeSnappyBlockAsm8B + +emit_lit_memmove_match_emit_encodeSnappyBlockAsm8B_memmove_move_33through64: + MOVOU (SI), X0 + MOVOU 16(SI), X1 + MOVOU -32(SI)(R8*1), X2 + MOVOU -16(SI)(R8*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + +memmove_end_copy_match_emit_encodeSnappyBlockAsm8B: + MOVQ DI, AX + JMP emit_literal_done_match_emit_encodeSnappyBlockAsm8B + +memmove_long_match_emit_encodeSnappyBlockAsm8B: + LEAQ (AX)(R8*1), DI + + // genMemMoveLong + MOVOU (SI), X0 + MOVOU 16(SI), X1 + MOVOU -32(SI)(R8*1), X2 + MOVOU -16(SI)(R8*1), X3 + MOVQ R8, R10 + SHRQ $0x05, R10 + MOVQ AX, R9 + ANDL $0x0000001f, R9 + MOVQ $0x00000040, R11 + SUBQ R9, R11 + DECQ R10 + JA emit_lit_memmove_long_match_emit_encodeSnappyBlockAsm8Blarge_forward_sse_loop_32 + LEAQ -32(SI)(R11*1), R9 + LEAQ -32(AX)(R11*1), R12 + +emit_lit_memmove_long_match_emit_encodeSnappyBlockAsm8Blarge_big_loop_back: + MOVOU (R9), X4 + MOVOU 16(R9), X5 + MOVOA X4, (R12) + MOVOA X5, 16(R12) + ADDQ $0x20, R12 + ADDQ $0x20, R9 + ADDQ $0x20, R11 + DECQ R10 + JNA emit_lit_memmove_long_match_emit_encodeSnappyBlockAsm8Blarge_big_loop_back + +emit_lit_memmove_long_match_emit_encodeSnappyBlockAsm8Blarge_forward_sse_loop_32: + MOVOU -32(SI)(R11*1), X4 + MOVOU -16(SI)(R11*1), X5 + MOVOA X4, -32(AX)(R11*1) + MOVOA X5, -16(AX)(R11*1) + ADDQ $0x20, R11 + CMPQ R8, R11 + JAE emit_lit_memmove_long_match_emit_encodeSnappyBlockAsm8Blarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + MOVQ DI, AX + +emit_literal_done_match_emit_encodeSnappyBlockAsm8B: +match_nolit_loop_encodeSnappyBlockAsm8B: + MOVL CX, SI + SUBL BX, SI + MOVL SI, 16(SP) + ADDL $0x04, CX + ADDL $0x04, BX + MOVQ src_len+32(FP), SI + SUBL CX, SI + LEAQ (DX)(CX*1), DI + LEAQ (DX)(BX*1), BX + + // matchLen + XORL R9, R9 + +matchlen_loopback_16_match_nolit_encodeSnappyBlockAsm8B: + CMPL SI, $0x10 + JB matchlen_match8_match_nolit_encodeSnappyBlockAsm8B + MOVQ (DI)(R9*1), R8 + MOVQ 8(DI)(R9*1), R10 + XORQ (BX)(R9*1), R8 + JNZ matchlen_bsf_8_match_nolit_encodeSnappyBlockAsm8B + XORQ 8(BX)(R9*1), R10 + JNZ matchlen_bsf_16match_nolit_encodeSnappyBlockAsm8B + LEAL -16(SI), SI + LEAL 16(R9), R9 + JMP matchlen_loopback_16_match_nolit_encodeSnappyBlockAsm8B + +matchlen_bsf_16match_nolit_encodeSnappyBlockAsm8B: +#ifdef GOAMD64_v3 + TZCNTQ R10, R10 + +#else + BSFQ R10, R10 + +#endif + SARQ $0x03, R10 + LEAL 8(R9)(R10*1), R9 + JMP match_nolit_end_encodeSnappyBlockAsm8B + +matchlen_match8_match_nolit_encodeSnappyBlockAsm8B: + CMPL SI, $0x08 + JB matchlen_match4_match_nolit_encodeSnappyBlockAsm8B + MOVQ (DI)(R9*1), R8 + XORQ (BX)(R9*1), R8 + JNZ matchlen_bsf_8_match_nolit_encodeSnappyBlockAsm8B + LEAL -8(SI), SI + LEAL 8(R9), R9 + JMP matchlen_match4_match_nolit_encodeSnappyBlockAsm8B + +matchlen_bsf_8_match_nolit_encodeSnappyBlockAsm8B: +#ifdef GOAMD64_v3 + TZCNTQ R8, R8 + +#else + BSFQ R8, R8 + +#endif + SARQ $0x03, R8 + LEAL (R9)(R8*1), R9 + JMP match_nolit_end_encodeSnappyBlockAsm8B + +matchlen_match4_match_nolit_encodeSnappyBlockAsm8B: + CMPL SI, $0x04 + JB matchlen_match2_match_nolit_encodeSnappyBlockAsm8B + MOVL (DI)(R9*1), R8 + CMPL (BX)(R9*1), R8 + JNE matchlen_match2_match_nolit_encodeSnappyBlockAsm8B + LEAL -4(SI), SI + LEAL 4(R9), R9 + +matchlen_match2_match_nolit_encodeSnappyBlockAsm8B: + CMPL SI, $0x01 + JE matchlen_match1_match_nolit_encodeSnappyBlockAsm8B + JB match_nolit_end_encodeSnappyBlockAsm8B + MOVW (DI)(R9*1), R8 + CMPW (BX)(R9*1), R8 + JNE matchlen_match1_match_nolit_encodeSnappyBlockAsm8B + LEAL 2(R9), R9 + SUBL $0x02, SI + JZ match_nolit_end_encodeSnappyBlockAsm8B + +matchlen_match1_match_nolit_encodeSnappyBlockAsm8B: + MOVB (DI)(R9*1), R8 + CMPB (BX)(R9*1), R8 + JNE match_nolit_end_encodeSnappyBlockAsm8B + LEAL 1(R9), R9 + +match_nolit_end_encodeSnappyBlockAsm8B: + ADDL R9, CX + MOVL 16(SP), BX + ADDL $0x04, R9 + MOVL CX, 12(SP) + + // emitCopy +two_byte_offset_match_nolit_encodeSnappyBlockAsm8B: + CMPL R9, $0x40 + JBE two_byte_offset_short_match_nolit_encodeSnappyBlockAsm8B + MOVB $0xee, (AX) + MOVW BX, 1(AX) + LEAL -60(R9), R9 + ADDQ $0x03, AX + JMP two_byte_offset_match_nolit_encodeSnappyBlockAsm8B + +two_byte_offset_short_match_nolit_encodeSnappyBlockAsm8B: + MOVL R9, SI + SHLL $0x02, SI + CMPL R9, $0x0c + JAE emit_copy_three_match_nolit_encodeSnappyBlockAsm8B + LEAL -15(SI), SI + MOVB BL, 1(AX) + SHRL $0x08, BX + SHLL $0x05, BX + ORL BX, SI + MOVB SI, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeSnappyBlockAsm8B + +emit_copy_three_match_nolit_encodeSnappyBlockAsm8B: + LEAL -2(SI), SI + MOVB SI, (AX) + MOVW BX, 1(AX) + ADDQ $0x03, AX + +match_nolit_emitcopy_end_encodeSnappyBlockAsm8B: + CMPL CX, 8(SP) + JAE emit_remainder_encodeSnappyBlockAsm8B + MOVQ -2(DX)(CX*1), SI + CMPQ AX, (SP) + JB match_nolit_dst_ok_encodeSnappyBlockAsm8B + MOVQ $0x00000000, ret+48(FP) + RET + +match_nolit_dst_ok_encodeSnappyBlockAsm8B: + MOVQ $0x9e3779b1, R8 + MOVQ SI, DI + SHRQ $0x10, SI + MOVQ SI, BX + SHLQ $0x20, DI + IMULQ R8, DI + SHRQ $0x38, DI + SHLQ $0x20, BX + IMULQ R8, BX + SHRQ $0x38, BX + LEAL -2(CX), R8 + LEAQ 24(SP)(BX*4), R9 + MOVL (R9), BX + MOVL R8, 24(SP)(DI*4) + MOVL CX, (R9) + CMPL (DX)(BX*1), SI + JEQ match_nolit_loop_encodeSnappyBlockAsm8B + INCL CX + JMP search_loop_encodeSnappyBlockAsm8B + +emit_remainder_encodeSnappyBlockAsm8B: + MOVQ src_len+32(FP), CX + SUBL 12(SP), CX + LEAQ 3(AX)(CX*1), CX + CMPQ CX, (SP) + JB emit_remainder_ok_encodeSnappyBlockAsm8B + MOVQ $0x00000000, ret+48(FP) + RET + +emit_remainder_ok_encodeSnappyBlockAsm8B: + MOVQ src_len+32(FP), CX + MOVL 12(SP), BX + CMPL BX, CX + JEQ emit_literal_done_emit_remainder_encodeSnappyBlockAsm8B + MOVL CX, SI + MOVL CX, 12(SP) + LEAQ (DX)(BX*1), CX + SUBL BX, SI + LEAL -1(SI), DX + CMPL DX, $0x3c + JB one_byte_emit_remainder_encodeSnappyBlockAsm8B + CMPL DX, $0x00000100 + JB two_bytes_emit_remainder_encodeSnappyBlockAsm8B + JB three_bytes_emit_remainder_encodeSnappyBlockAsm8B + +three_bytes_emit_remainder_encodeSnappyBlockAsm8B: + MOVB $0xf4, (AX) + MOVW DX, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_emit_remainder_encodeSnappyBlockAsm8B + +two_bytes_emit_remainder_encodeSnappyBlockAsm8B: + MOVB $0xf0, (AX) + MOVB DL, 1(AX) + ADDQ $0x02, AX + CMPL DX, $0x40 + JB memmove_emit_remainder_encodeSnappyBlockAsm8B + JMP memmove_long_emit_remainder_encodeSnappyBlockAsm8B + +one_byte_emit_remainder_encodeSnappyBlockAsm8B: + SHLB $0x02, DL + MOVB DL, (AX) + ADDQ $0x01, AX + +memmove_emit_remainder_encodeSnappyBlockAsm8B: + LEAQ (AX)(SI*1), DX + MOVL SI, BX + + // genMemMoveShort + CMPQ BX, $0x03 + JB emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm8B_memmove_move_1or2 + JE emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm8B_memmove_move_3 + CMPQ BX, $0x08 + JB emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm8B_memmove_move_4through7 + CMPQ BX, $0x10 + JBE emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm8B_memmove_move_8through16 + CMPQ BX, $0x20 + JBE emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm8B_memmove_move_17through32 + JMP emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm8B_memmove_move_33through64 + +emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm8B_memmove_move_1or2: + MOVB (CX), SI + MOVB -1(CX)(BX*1), CL + MOVB SI, (AX) + MOVB CL, -1(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeSnappyBlockAsm8B + +emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm8B_memmove_move_3: + MOVW (CX), SI + MOVB 2(CX), CL + MOVW SI, (AX) + MOVB CL, 2(AX) + JMP memmove_end_copy_emit_remainder_encodeSnappyBlockAsm8B + +emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm8B_memmove_move_4through7: + MOVL (CX), SI + MOVL -4(CX)(BX*1), CX + MOVL SI, (AX) + MOVL CX, -4(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeSnappyBlockAsm8B + +emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm8B_memmove_move_8through16: + MOVQ (CX), SI + MOVQ -8(CX)(BX*1), CX + MOVQ SI, (AX) + MOVQ CX, -8(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeSnappyBlockAsm8B + +emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm8B_memmove_move_17through32: + MOVOU (CX), X0 + MOVOU -16(CX)(BX*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeSnappyBlockAsm8B + +emit_lit_memmove_emit_remainder_encodeSnappyBlockAsm8B_memmove_move_33through64: + MOVOU (CX), X0 + MOVOU 16(CX), X1 + MOVOU -32(CX)(BX*1), X2 + MOVOU -16(CX)(BX*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(BX*1) + MOVOU X3, -16(AX)(BX*1) + +memmove_end_copy_emit_remainder_encodeSnappyBlockAsm8B: + MOVQ DX, AX + JMP emit_literal_done_emit_remainder_encodeSnappyBlockAsm8B + +memmove_long_emit_remainder_encodeSnappyBlockAsm8B: + LEAQ (AX)(SI*1), DX + MOVL SI, BX + + // genMemMoveLong + MOVOU (CX), X0 + MOVOU 16(CX), X1 + MOVOU -32(CX)(BX*1), X2 + MOVOU -16(CX)(BX*1), X3 + MOVQ BX, DI + SHRQ $0x05, DI + MOVQ AX, SI + ANDL $0x0000001f, SI + MOVQ $0x00000040, R8 + SUBQ SI, R8 + DECQ DI + JA emit_lit_memmove_long_emit_remainder_encodeSnappyBlockAsm8Blarge_forward_sse_loop_32 + LEAQ -32(CX)(R8*1), SI + LEAQ -32(AX)(R8*1), R9 + +emit_lit_memmove_long_emit_remainder_encodeSnappyBlockAsm8Blarge_big_loop_back: + MOVOU (SI), X4 + MOVOU 16(SI), X5 + MOVOA X4, (R9) + MOVOA X5, 16(R9) + ADDQ $0x20, R9 + ADDQ $0x20, SI + ADDQ $0x20, R8 + DECQ DI + JNA emit_lit_memmove_long_emit_remainder_encodeSnappyBlockAsm8Blarge_big_loop_back + +emit_lit_memmove_long_emit_remainder_encodeSnappyBlockAsm8Blarge_forward_sse_loop_32: + MOVOU -32(CX)(R8*1), X4 + MOVOU -16(CX)(R8*1), X5 + MOVOA X4, -32(AX)(R8*1) + MOVOA X5, -16(AX)(R8*1) + ADDQ $0x20, R8 + CMPQ BX, R8 + JAE emit_lit_memmove_long_emit_remainder_encodeSnappyBlockAsm8Blarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(BX*1) + MOVOU X3, -16(AX)(BX*1) + MOVQ DX, AX + +emit_literal_done_emit_remainder_encodeSnappyBlockAsm8B: + MOVQ dst_base+0(FP), CX + SUBQ CX, AX + MOVQ AX, ret+48(FP) + RET + +// func encodeSnappyBetterBlockAsm(dst []byte, src []byte) int +// Requires: BMI, SSE2 +TEXT ·encodeSnappyBetterBlockAsm(SB), $589848-56 + MOVQ dst_base+0(FP), AX + MOVQ $0x00001200, CX + LEAQ 24(SP), DX + PXOR X0, X0 + +zero_loop_encodeSnappyBetterBlockAsm: + MOVOU X0, (DX) + MOVOU X0, 16(DX) + MOVOU X0, 32(DX) + MOVOU X0, 48(DX) + MOVOU X0, 64(DX) + MOVOU X0, 80(DX) + MOVOU X0, 96(DX) + MOVOU X0, 112(DX) + ADDQ $0x80, DX + DECQ CX + JNZ zero_loop_encodeSnappyBetterBlockAsm + MOVL $0x00000000, 12(SP) + MOVQ src_len+32(FP), CX + LEAQ -9(CX), DX + LEAQ -8(CX), BX + MOVL BX, 8(SP) + SHRQ $0x05, CX + SUBL CX, DX + LEAQ (AX)(DX*1), DX + MOVQ DX, (SP) + MOVL $0x00000001, CX + MOVL $0x00000000, 16(SP) + MOVQ src_base+24(FP), DX + +search_loop_encodeSnappyBetterBlockAsm: + MOVL CX, BX + SUBL 12(SP), BX + SHRL $0x07, BX + CMPL BX, $0x63 + JBE check_maxskip_ok_encodeSnappyBetterBlockAsm + LEAL 100(CX), BX + JMP check_maxskip_cont_encodeSnappyBetterBlockAsm + +check_maxskip_ok_encodeSnappyBetterBlockAsm: + LEAL 1(CX)(BX*1), BX + +check_maxskip_cont_encodeSnappyBetterBlockAsm: + CMPL BX, 8(SP) + JAE emit_remainder_encodeSnappyBetterBlockAsm + MOVQ (DX)(CX*1), SI + MOVL BX, 20(SP) + MOVQ $0x00cf1bbcdcbfa563, R8 + MOVQ $0x9e3779b1, BX + MOVQ SI, R9 + MOVQ SI, R10 + SHLQ $0x08, R9 + IMULQ R8, R9 + SHRQ $0x2f, R9 + SHLQ $0x20, R10 + IMULQ BX, R10 + SHRQ $0x32, R10 + MOVL 24(SP)(R9*4), BX + MOVL 524312(SP)(R10*4), DI + MOVL CX, 24(SP)(R9*4) + MOVL CX, 524312(SP)(R10*4) + MOVQ (DX)(BX*1), R9 + MOVQ (DX)(DI*1), R10 + CMPQ R9, SI + JEQ candidate_match_encodeSnappyBetterBlockAsm + CMPQ R10, SI + JNE no_short_found_encodeSnappyBetterBlockAsm + MOVL DI, BX + JMP candidate_match_encodeSnappyBetterBlockAsm + +no_short_found_encodeSnappyBetterBlockAsm: + CMPL R9, SI + JEQ candidate_match_encodeSnappyBetterBlockAsm + CMPL R10, SI + JEQ candidateS_match_encodeSnappyBetterBlockAsm + MOVL 20(SP), CX + JMP search_loop_encodeSnappyBetterBlockAsm + +candidateS_match_encodeSnappyBetterBlockAsm: + SHRQ $0x08, SI + MOVQ SI, R9 + SHLQ $0x08, R9 + IMULQ R8, R9 + SHRQ $0x2f, R9 + MOVL 24(SP)(R9*4), BX + INCL CX + MOVL CX, 24(SP)(R9*4) + CMPL (DX)(BX*1), SI + JEQ candidate_match_encodeSnappyBetterBlockAsm + DECL CX + MOVL DI, BX + +candidate_match_encodeSnappyBetterBlockAsm: + MOVL 12(SP), SI + TESTL BX, BX + JZ match_extend_back_end_encodeSnappyBetterBlockAsm + +match_extend_back_loop_encodeSnappyBetterBlockAsm: + CMPL CX, SI + JBE match_extend_back_end_encodeSnappyBetterBlockAsm + MOVB -1(DX)(BX*1), DI + MOVB -1(DX)(CX*1), R8 + CMPB DI, R8 + JNE match_extend_back_end_encodeSnappyBetterBlockAsm + LEAL -1(CX), CX + DECL BX + JZ match_extend_back_end_encodeSnappyBetterBlockAsm + JMP match_extend_back_loop_encodeSnappyBetterBlockAsm + +match_extend_back_end_encodeSnappyBetterBlockAsm: + MOVL CX, SI + SUBL 12(SP), SI + LEAQ 5(AX)(SI*1), SI + CMPQ SI, (SP) + JB match_dst_size_check_encodeSnappyBetterBlockAsm + MOVQ $0x00000000, ret+48(FP) + RET + +match_dst_size_check_encodeSnappyBetterBlockAsm: + MOVL CX, SI + ADDL $0x04, CX + ADDL $0x04, BX + MOVQ src_len+32(FP), DI + SUBL CX, DI + LEAQ (DX)(CX*1), R8 + LEAQ (DX)(BX*1), R9 + + // matchLen + XORL R11, R11 + +matchlen_loopback_16_match_nolit_encodeSnappyBetterBlockAsm: + CMPL DI, $0x10 + JB matchlen_match8_match_nolit_encodeSnappyBetterBlockAsm + MOVQ (R8)(R11*1), R10 + MOVQ 8(R8)(R11*1), R12 + XORQ (R9)(R11*1), R10 + JNZ matchlen_bsf_8_match_nolit_encodeSnappyBetterBlockAsm + XORQ 8(R9)(R11*1), R12 + JNZ matchlen_bsf_16match_nolit_encodeSnappyBetterBlockAsm + LEAL -16(DI), DI + LEAL 16(R11), R11 + JMP matchlen_loopback_16_match_nolit_encodeSnappyBetterBlockAsm + +matchlen_bsf_16match_nolit_encodeSnappyBetterBlockAsm: +#ifdef GOAMD64_v3 + TZCNTQ R12, R12 + +#else + BSFQ R12, R12 + +#endif + SARQ $0x03, R12 + LEAL 8(R11)(R12*1), R11 + JMP match_nolit_end_encodeSnappyBetterBlockAsm + +matchlen_match8_match_nolit_encodeSnappyBetterBlockAsm: + CMPL DI, $0x08 + JB matchlen_match4_match_nolit_encodeSnappyBetterBlockAsm + MOVQ (R8)(R11*1), R10 + XORQ (R9)(R11*1), R10 + JNZ matchlen_bsf_8_match_nolit_encodeSnappyBetterBlockAsm + LEAL -8(DI), DI + LEAL 8(R11), R11 + JMP matchlen_match4_match_nolit_encodeSnappyBetterBlockAsm + +matchlen_bsf_8_match_nolit_encodeSnappyBetterBlockAsm: +#ifdef GOAMD64_v3 + TZCNTQ R10, R10 + +#else + BSFQ R10, R10 + +#endif + SARQ $0x03, R10 + LEAL (R11)(R10*1), R11 + JMP match_nolit_end_encodeSnappyBetterBlockAsm + +matchlen_match4_match_nolit_encodeSnappyBetterBlockAsm: + CMPL DI, $0x04 + JB matchlen_match2_match_nolit_encodeSnappyBetterBlockAsm + MOVL (R8)(R11*1), R10 + CMPL (R9)(R11*1), R10 + JNE matchlen_match2_match_nolit_encodeSnappyBetterBlockAsm + LEAL -4(DI), DI + LEAL 4(R11), R11 + +matchlen_match2_match_nolit_encodeSnappyBetterBlockAsm: + CMPL DI, $0x01 + JE matchlen_match1_match_nolit_encodeSnappyBetterBlockAsm + JB match_nolit_end_encodeSnappyBetterBlockAsm + MOVW (R8)(R11*1), R10 + CMPW (R9)(R11*1), R10 + JNE matchlen_match1_match_nolit_encodeSnappyBetterBlockAsm + LEAL 2(R11), R11 + SUBL $0x02, DI + JZ match_nolit_end_encodeSnappyBetterBlockAsm + +matchlen_match1_match_nolit_encodeSnappyBetterBlockAsm: + MOVB (R8)(R11*1), R10 + CMPB (R9)(R11*1), R10 + JNE match_nolit_end_encodeSnappyBetterBlockAsm + LEAL 1(R11), R11 + +match_nolit_end_encodeSnappyBetterBlockAsm: + MOVL CX, DI + SUBL BX, DI + + // Check if repeat + CMPL R11, $0x01 + JA match_length_ok_encodeSnappyBetterBlockAsm + CMPL DI, $0x0000ffff + JBE match_length_ok_encodeSnappyBetterBlockAsm + MOVL 20(SP), CX + INCL CX + JMP search_loop_encodeSnappyBetterBlockAsm + +match_length_ok_encodeSnappyBetterBlockAsm: + MOVL DI, 16(SP) + MOVL 12(SP), BX + CMPL BX, SI + JEQ emit_literal_done_match_emit_encodeSnappyBetterBlockAsm + MOVL SI, R8 + MOVL SI, 12(SP) + LEAQ (DX)(BX*1), R9 + SUBL BX, R8 + LEAL -1(R8), BX + CMPL BX, $0x3c + JB one_byte_match_emit_encodeSnappyBetterBlockAsm + CMPL BX, $0x00000100 + JB two_bytes_match_emit_encodeSnappyBetterBlockAsm + CMPL BX, $0x00010000 + JB three_bytes_match_emit_encodeSnappyBetterBlockAsm + CMPL BX, $0x01000000 + JB four_bytes_match_emit_encodeSnappyBetterBlockAsm + MOVB $0xfc, (AX) + MOVL BX, 1(AX) + ADDQ $0x05, AX + JMP memmove_long_match_emit_encodeSnappyBetterBlockAsm + +four_bytes_match_emit_encodeSnappyBetterBlockAsm: + MOVL BX, R10 + SHRL $0x10, R10 + MOVB $0xf8, (AX) + MOVW BX, 1(AX) + MOVB R10, 3(AX) + ADDQ $0x04, AX + JMP memmove_long_match_emit_encodeSnappyBetterBlockAsm + +three_bytes_match_emit_encodeSnappyBetterBlockAsm: + MOVB $0xf4, (AX) + MOVW BX, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_match_emit_encodeSnappyBetterBlockAsm + +two_bytes_match_emit_encodeSnappyBetterBlockAsm: + MOVB $0xf0, (AX) + MOVB BL, 1(AX) + ADDQ $0x02, AX + CMPL BX, $0x40 + JB memmove_match_emit_encodeSnappyBetterBlockAsm + JMP memmove_long_match_emit_encodeSnappyBetterBlockAsm + +one_byte_match_emit_encodeSnappyBetterBlockAsm: + SHLB $0x02, BL + MOVB BL, (AX) + ADDQ $0x01, AX + +memmove_match_emit_encodeSnappyBetterBlockAsm: + LEAQ (AX)(R8*1), BX + + // genMemMoveShort + CMPQ R8, $0x08 + JBE emit_lit_memmove_match_emit_encodeSnappyBetterBlockAsm_memmove_move_8 + CMPQ R8, $0x10 + JBE emit_lit_memmove_match_emit_encodeSnappyBetterBlockAsm_memmove_move_8through16 + CMPQ R8, $0x20 + JBE emit_lit_memmove_match_emit_encodeSnappyBetterBlockAsm_memmove_move_17through32 + JMP emit_lit_memmove_match_emit_encodeSnappyBetterBlockAsm_memmove_move_33through64 + +emit_lit_memmove_match_emit_encodeSnappyBetterBlockAsm_memmove_move_8: + MOVQ (R9), R10 + MOVQ R10, (AX) + JMP memmove_end_copy_match_emit_encodeSnappyBetterBlockAsm + +emit_lit_memmove_match_emit_encodeSnappyBetterBlockAsm_memmove_move_8through16: + MOVQ (R9), R10 + MOVQ -8(R9)(R8*1), R9 + MOVQ R10, (AX) + MOVQ R9, -8(AX)(R8*1) + JMP memmove_end_copy_match_emit_encodeSnappyBetterBlockAsm + +emit_lit_memmove_match_emit_encodeSnappyBetterBlockAsm_memmove_move_17through32: + MOVOU (R9), X0 + MOVOU -16(R9)(R8*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(R8*1) + JMP memmove_end_copy_match_emit_encodeSnappyBetterBlockAsm + +emit_lit_memmove_match_emit_encodeSnappyBetterBlockAsm_memmove_move_33through64: + MOVOU (R9), X0 + MOVOU 16(R9), X1 + MOVOU -32(R9)(R8*1), X2 + MOVOU -16(R9)(R8*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + +memmove_end_copy_match_emit_encodeSnappyBetterBlockAsm: + MOVQ BX, AX + JMP emit_literal_done_match_emit_encodeSnappyBetterBlockAsm + +memmove_long_match_emit_encodeSnappyBetterBlockAsm: + LEAQ (AX)(R8*1), BX + + // genMemMoveLong + MOVOU (R9), X0 + MOVOU 16(R9), X1 + MOVOU -32(R9)(R8*1), X2 + MOVOU -16(R9)(R8*1), X3 + MOVQ R8, R12 + SHRQ $0x05, R12 + MOVQ AX, R10 + ANDL $0x0000001f, R10 + MOVQ $0x00000040, R13 + SUBQ R10, R13 + DECQ R12 + JA emit_lit_memmove_long_match_emit_encodeSnappyBetterBlockAsmlarge_forward_sse_loop_32 + LEAQ -32(R9)(R13*1), R10 + LEAQ -32(AX)(R13*1), R14 + +emit_lit_memmove_long_match_emit_encodeSnappyBetterBlockAsmlarge_big_loop_back: + MOVOU (R10), X4 + MOVOU 16(R10), X5 + MOVOA X4, (R14) + MOVOA X5, 16(R14) + ADDQ $0x20, R14 + ADDQ $0x20, R10 + ADDQ $0x20, R13 + DECQ R12 + JNA emit_lit_memmove_long_match_emit_encodeSnappyBetterBlockAsmlarge_big_loop_back + +emit_lit_memmove_long_match_emit_encodeSnappyBetterBlockAsmlarge_forward_sse_loop_32: + MOVOU -32(R9)(R13*1), X4 + MOVOU -16(R9)(R13*1), X5 + MOVOA X4, -32(AX)(R13*1) + MOVOA X5, -16(AX)(R13*1) + ADDQ $0x20, R13 + CMPQ R8, R13 + JAE emit_lit_memmove_long_match_emit_encodeSnappyBetterBlockAsmlarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + MOVQ BX, AX + +emit_literal_done_match_emit_encodeSnappyBetterBlockAsm: + ADDL R11, CX + ADDL $0x04, R11 + MOVL CX, 12(SP) + + // emitCopy + CMPL DI, $0x00010000 + JB two_byte_offset_match_nolit_encodeSnappyBetterBlockAsm + +four_bytes_loop_back_match_nolit_encodeSnappyBetterBlockAsm: + CMPL R11, $0x40 + JBE four_bytes_remain_match_nolit_encodeSnappyBetterBlockAsm + MOVB $0xff, (AX) + MOVL DI, 1(AX) + LEAL -64(R11), R11 + ADDQ $0x05, AX + CMPL R11, $0x04 + JB four_bytes_remain_match_nolit_encodeSnappyBetterBlockAsm + JMP four_bytes_loop_back_match_nolit_encodeSnappyBetterBlockAsm + +four_bytes_remain_match_nolit_encodeSnappyBetterBlockAsm: + TESTL R11, R11 + JZ match_nolit_emitcopy_end_encodeSnappyBetterBlockAsm + XORL BX, BX + LEAL -1(BX)(R11*4), R11 + MOVB R11, (AX) + MOVL DI, 1(AX) + ADDQ $0x05, AX + JMP match_nolit_emitcopy_end_encodeSnappyBetterBlockAsm + +two_byte_offset_match_nolit_encodeSnappyBetterBlockAsm: + CMPL R11, $0x40 + JBE two_byte_offset_short_match_nolit_encodeSnappyBetterBlockAsm + MOVB $0xee, (AX) + MOVW DI, 1(AX) + LEAL -60(R11), R11 + ADDQ $0x03, AX + JMP two_byte_offset_match_nolit_encodeSnappyBetterBlockAsm + +two_byte_offset_short_match_nolit_encodeSnappyBetterBlockAsm: + MOVL R11, BX + SHLL $0x02, BX + CMPL R11, $0x0c + JAE emit_copy_three_match_nolit_encodeSnappyBetterBlockAsm + CMPL DI, $0x00000800 + JAE emit_copy_three_match_nolit_encodeSnappyBetterBlockAsm + LEAL -15(BX), BX + MOVB DI, 1(AX) + SHRL $0x08, DI + SHLL $0x05, DI + ORL DI, BX + MOVB BL, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeSnappyBetterBlockAsm + +emit_copy_three_match_nolit_encodeSnappyBetterBlockAsm: + LEAL -2(BX), BX + MOVB BL, (AX) + MOVW DI, 1(AX) + ADDQ $0x03, AX + +match_nolit_emitcopy_end_encodeSnappyBetterBlockAsm: + CMPL CX, 8(SP) + JAE emit_remainder_encodeSnappyBetterBlockAsm + CMPQ AX, (SP) + JB match_nolit_dst_ok_encodeSnappyBetterBlockAsm + MOVQ $0x00000000, ret+48(FP) + RET + +match_nolit_dst_ok_encodeSnappyBetterBlockAsm: + MOVQ $0x00cf1bbcdcbfa563, BX + MOVQ $0x9e3779b1, DI + LEAQ 1(SI), SI + LEAQ -2(CX), R8 + MOVQ (DX)(SI*1), R9 + MOVQ 1(DX)(SI*1), R10 + MOVQ (DX)(R8*1), R11 + MOVQ 1(DX)(R8*1), R12 + SHLQ $0x08, R9 + IMULQ BX, R9 + SHRQ $0x2f, R9 + SHLQ $0x20, R10 + IMULQ DI, R10 + SHRQ $0x32, R10 + SHLQ $0x08, R11 + IMULQ BX, R11 + SHRQ $0x2f, R11 + SHLQ $0x20, R12 + IMULQ DI, R12 + SHRQ $0x32, R12 + LEAQ 1(SI), DI + LEAQ 1(R8), R13 + MOVL SI, 24(SP)(R9*4) + MOVL R8, 24(SP)(R11*4) + MOVL DI, 524312(SP)(R10*4) + MOVL R13, 524312(SP)(R12*4) + LEAQ 1(R8)(SI*1), DI + SHRQ $0x01, DI + ADDQ $0x01, SI + SUBQ $0x01, R8 + +index_loop_encodeSnappyBetterBlockAsm: + CMPQ DI, R8 + JAE search_loop_encodeSnappyBetterBlockAsm + MOVQ (DX)(SI*1), R9 + MOVQ (DX)(DI*1), R10 + SHLQ $0x08, R9 + IMULQ BX, R9 + SHRQ $0x2f, R9 + SHLQ $0x08, R10 + IMULQ BX, R10 + SHRQ $0x2f, R10 + MOVL SI, 24(SP)(R9*4) + MOVL DI, 24(SP)(R10*4) + ADDQ $0x02, SI + ADDQ $0x02, DI + JMP index_loop_encodeSnappyBetterBlockAsm + +emit_remainder_encodeSnappyBetterBlockAsm: + MOVQ src_len+32(FP), CX + SUBL 12(SP), CX + LEAQ 5(AX)(CX*1), CX + CMPQ CX, (SP) + JB emit_remainder_ok_encodeSnappyBetterBlockAsm + MOVQ $0x00000000, ret+48(FP) + RET + +emit_remainder_ok_encodeSnappyBetterBlockAsm: + MOVQ src_len+32(FP), CX + MOVL 12(SP), BX + CMPL BX, CX + JEQ emit_literal_done_emit_remainder_encodeSnappyBetterBlockAsm + MOVL CX, SI + MOVL CX, 12(SP) + LEAQ (DX)(BX*1), CX + SUBL BX, SI + LEAL -1(SI), DX + CMPL DX, $0x3c + JB one_byte_emit_remainder_encodeSnappyBetterBlockAsm + CMPL DX, $0x00000100 + JB two_bytes_emit_remainder_encodeSnappyBetterBlockAsm + CMPL DX, $0x00010000 + JB three_bytes_emit_remainder_encodeSnappyBetterBlockAsm + CMPL DX, $0x01000000 + JB four_bytes_emit_remainder_encodeSnappyBetterBlockAsm + MOVB $0xfc, (AX) + MOVL DX, 1(AX) + ADDQ $0x05, AX + JMP memmove_long_emit_remainder_encodeSnappyBetterBlockAsm + +four_bytes_emit_remainder_encodeSnappyBetterBlockAsm: + MOVL DX, BX + SHRL $0x10, BX + MOVB $0xf8, (AX) + MOVW DX, 1(AX) + MOVB BL, 3(AX) + ADDQ $0x04, AX + JMP memmove_long_emit_remainder_encodeSnappyBetterBlockAsm + +three_bytes_emit_remainder_encodeSnappyBetterBlockAsm: + MOVB $0xf4, (AX) + MOVW DX, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_emit_remainder_encodeSnappyBetterBlockAsm + +two_bytes_emit_remainder_encodeSnappyBetterBlockAsm: + MOVB $0xf0, (AX) + MOVB DL, 1(AX) + ADDQ $0x02, AX + CMPL DX, $0x40 + JB memmove_emit_remainder_encodeSnappyBetterBlockAsm + JMP memmove_long_emit_remainder_encodeSnappyBetterBlockAsm + +one_byte_emit_remainder_encodeSnappyBetterBlockAsm: + SHLB $0x02, DL + MOVB DL, (AX) + ADDQ $0x01, AX + +memmove_emit_remainder_encodeSnappyBetterBlockAsm: + LEAQ (AX)(SI*1), DX + MOVL SI, BX + + // genMemMoveShort + CMPQ BX, $0x03 + JB emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm_memmove_move_1or2 + JE emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm_memmove_move_3 + CMPQ BX, $0x08 + JB emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm_memmove_move_4through7 + CMPQ BX, $0x10 + JBE emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm_memmove_move_8through16 + CMPQ BX, $0x20 + JBE emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm_memmove_move_17through32 + JMP emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm_memmove_move_33through64 + +emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm_memmove_move_1or2: + MOVB (CX), SI + MOVB -1(CX)(BX*1), CL + MOVB SI, (AX) + MOVB CL, -1(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeSnappyBetterBlockAsm + +emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm_memmove_move_3: + MOVW (CX), SI + MOVB 2(CX), CL + MOVW SI, (AX) + MOVB CL, 2(AX) + JMP memmove_end_copy_emit_remainder_encodeSnappyBetterBlockAsm + +emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm_memmove_move_4through7: + MOVL (CX), SI + MOVL -4(CX)(BX*1), CX + MOVL SI, (AX) + MOVL CX, -4(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeSnappyBetterBlockAsm + +emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm_memmove_move_8through16: + MOVQ (CX), SI + MOVQ -8(CX)(BX*1), CX + MOVQ SI, (AX) + MOVQ CX, -8(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeSnappyBetterBlockAsm + +emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm_memmove_move_17through32: + MOVOU (CX), X0 + MOVOU -16(CX)(BX*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeSnappyBetterBlockAsm + +emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm_memmove_move_33through64: + MOVOU (CX), X0 + MOVOU 16(CX), X1 + MOVOU -32(CX)(BX*1), X2 + MOVOU -16(CX)(BX*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(BX*1) + MOVOU X3, -16(AX)(BX*1) + +memmove_end_copy_emit_remainder_encodeSnappyBetterBlockAsm: + MOVQ DX, AX + JMP emit_literal_done_emit_remainder_encodeSnappyBetterBlockAsm + +memmove_long_emit_remainder_encodeSnappyBetterBlockAsm: + LEAQ (AX)(SI*1), DX + MOVL SI, BX + + // genMemMoveLong + MOVOU (CX), X0 + MOVOU 16(CX), X1 + MOVOU -32(CX)(BX*1), X2 + MOVOU -16(CX)(BX*1), X3 + MOVQ BX, DI + SHRQ $0x05, DI + MOVQ AX, SI + ANDL $0x0000001f, SI + MOVQ $0x00000040, R8 + SUBQ SI, R8 + DECQ DI + JA emit_lit_memmove_long_emit_remainder_encodeSnappyBetterBlockAsmlarge_forward_sse_loop_32 + LEAQ -32(CX)(R8*1), SI + LEAQ -32(AX)(R8*1), R9 + +emit_lit_memmove_long_emit_remainder_encodeSnappyBetterBlockAsmlarge_big_loop_back: + MOVOU (SI), X4 + MOVOU 16(SI), X5 + MOVOA X4, (R9) + MOVOA X5, 16(R9) + ADDQ $0x20, R9 + ADDQ $0x20, SI + ADDQ $0x20, R8 + DECQ DI + JNA emit_lit_memmove_long_emit_remainder_encodeSnappyBetterBlockAsmlarge_big_loop_back + +emit_lit_memmove_long_emit_remainder_encodeSnappyBetterBlockAsmlarge_forward_sse_loop_32: + MOVOU -32(CX)(R8*1), X4 + MOVOU -16(CX)(R8*1), X5 + MOVOA X4, -32(AX)(R8*1) + MOVOA X5, -16(AX)(R8*1) + ADDQ $0x20, R8 + CMPQ BX, R8 + JAE emit_lit_memmove_long_emit_remainder_encodeSnappyBetterBlockAsmlarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(BX*1) + MOVOU X3, -16(AX)(BX*1) + MOVQ DX, AX + +emit_literal_done_emit_remainder_encodeSnappyBetterBlockAsm: + MOVQ dst_base+0(FP), CX + SUBQ CX, AX + MOVQ AX, ret+48(FP) + RET + +// func encodeSnappyBetterBlockAsm64K(dst []byte, src []byte) int +// Requires: BMI, SSE2 +TEXT ·encodeSnappyBetterBlockAsm64K(SB), $327704-56 + MOVQ dst_base+0(FP), AX + MOVQ $0x00000a00, CX + LEAQ 24(SP), DX + PXOR X0, X0 + +zero_loop_encodeSnappyBetterBlockAsm64K: + MOVOU X0, (DX) + MOVOU X0, 16(DX) + MOVOU X0, 32(DX) + MOVOU X0, 48(DX) + MOVOU X0, 64(DX) + MOVOU X0, 80(DX) + MOVOU X0, 96(DX) + MOVOU X0, 112(DX) + ADDQ $0x80, DX + DECQ CX + JNZ zero_loop_encodeSnappyBetterBlockAsm64K + MOVL $0x00000000, 12(SP) + MOVQ src_len+32(FP), CX + LEAQ -9(CX), DX + LEAQ -8(CX), BX + MOVL BX, 8(SP) + SHRQ $0x05, CX + SUBL CX, DX + LEAQ (AX)(DX*1), DX + MOVQ DX, (SP) + MOVL $0x00000001, CX + MOVL $0x00000000, 16(SP) + MOVQ src_base+24(FP), DX + +search_loop_encodeSnappyBetterBlockAsm64K: + MOVL CX, BX + SUBL 12(SP), BX + SHRL $0x07, BX + LEAL 1(CX)(BX*1), BX + CMPL BX, 8(SP) + JAE emit_remainder_encodeSnappyBetterBlockAsm64K + MOVQ (DX)(CX*1), SI + MOVL BX, 20(SP) + MOVQ $0x00cf1bbcdcbfa563, R8 + MOVQ $0x9e3779b1, BX + MOVQ SI, R9 + MOVQ SI, R10 + SHLQ $0x08, R9 + IMULQ R8, R9 + SHRQ $0x30, R9 + SHLQ $0x20, R10 + IMULQ BX, R10 + SHRQ $0x32, R10 + MOVL 24(SP)(R9*4), BX + MOVL 262168(SP)(R10*4), DI + MOVL CX, 24(SP)(R9*4) + MOVL CX, 262168(SP)(R10*4) + MOVQ (DX)(BX*1), R9 + MOVQ (DX)(DI*1), R10 + CMPQ R9, SI + JEQ candidate_match_encodeSnappyBetterBlockAsm64K + CMPQ R10, SI + JNE no_short_found_encodeSnappyBetterBlockAsm64K + MOVL DI, BX + JMP candidate_match_encodeSnappyBetterBlockAsm64K + +no_short_found_encodeSnappyBetterBlockAsm64K: + CMPL R9, SI + JEQ candidate_match_encodeSnappyBetterBlockAsm64K + CMPL R10, SI + JEQ candidateS_match_encodeSnappyBetterBlockAsm64K + MOVL 20(SP), CX + JMP search_loop_encodeSnappyBetterBlockAsm64K + +candidateS_match_encodeSnappyBetterBlockAsm64K: + SHRQ $0x08, SI + MOVQ SI, R9 + SHLQ $0x08, R9 + IMULQ R8, R9 + SHRQ $0x30, R9 + MOVL 24(SP)(R9*4), BX + INCL CX + MOVL CX, 24(SP)(R9*4) + CMPL (DX)(BX*1), SI + JEQ candidate_match_encodeSnappyBetterBlockAsm64K + DECL CX + MOVL DI, BX + +candidate_match_encodeSnappyBetterBlockAsm64K: + MOVL 12(SP), SI + TESTL BX, BX + JZ match_extend_back_end_encodeSnappyBetterBlockAsm64K + +match_extend_back_loop_encodeSnappyBetterBlockAsm64K: + CMPL CX, SI + JBE match_extend_back_end_encodeSnappyBetterBlockAsm64K + MOVB -1(DX)(BX*1), DI + MOVB -1(DX)(CX*1), R8 + CMPB DI, R8 + JNE match_extend_back_end_encodeSnappyBetterBlockAsm64K + LEAL -1(CX), CX + DECL BX + JZ match_extend_back_end_encodeSnappyBetterBlockAsm64K + JMP match_extend_back_loop_encodeSnappyBetterBlockAsm64K + +match_extend_back_end_encodeSnappyBetterBlockAsm64K: + MOVL CX, SI + SUBL 12(SP), SI + LEAQ 3(AX)(SI*1), SI + CMPQ SI, (SP) + JB match_dst_size_check_encodeSnappyBetterBlockAsm64K + MOVQ $0x00000000, ret+48(FP) + RET + +match_dst_size_check_encodeSnappyBetterBlockAsm64K: + MOVL CX, SI + ADDL $0x04, CX + ADDL $0x04, BX + MOVQ src_len+32(FP), DI + SUBL CX, DI + LEAQ (DX)(CX*1), R8 + LEAQ (DX)(BX*1), R9 + + // matchLen + XORL R11, R11 + +matchlen_loopback_16_match_nolit_encodeSnappyBetterBlockAsm64K: + CMPL DI, $0x10 + JB matchlen_match8_match_nolit_encodeSnappyBetterBlockAsm64K + MOVQ (R8)(R11*1), R10 + MOVQ 8(R8)(R11*1), R12 + XORQ (R9)(R11*1), R10 + JNZ matchlen_bsf_8_match_nolit_encodeSnappyBetterBlockAsm64K + XORQ 8(R9)(R11*1), R12 + JNZ matchlen_bsf_16match_nolit_encodeSnappyBetterBlockAsm64K + LEAL -16(DI), DI + LEAL 16(R11), R11 + JMP matchlen_loopback_16_match_nolit_encodeSnappyBetterBlockAsm64K + +matchlen_bsf_16match_nolit_encodeSnappyBetterBlockAsm64K: +#ifdef GOAMD64_v3 + TZCNTQ R12, R12 + +#else + BSFQ R12, R12 + +#endif + SARQ $0x03, R12 + LEAL 8(R11)(R12*1), R11 + JMP match_nolit_end_encodeSnappyBetterBlockAsm64K + +matchlen_match8_match_nolit_encodeSnappyBetterBlockAsm64K: + CMPL DI, $0x08 + JB matchlen_match4_match_nolit_encodeSnappyBetterBlockAsm64K + MOVQ (R8)(R11*1), R10 + XORQ (R9)(R11*1), R10 + JNZ matchlen_bsf_8_match_nolit_encodeSnappyBetterBlockAsm64K + LEAL -8(DI), DI + LEAL 8(R11), R11 + JMP matchlen_match4_match_nolit_encodeSnappyBetterBlockAsm64K + +matchlen_bsf_8_match_nolit_encodeSnappyBetterBlockAsm64K: +#ifdef GOAMD64_v3 + TZCNTQ R10, R10 + +#else + BSFQ R10, R10 + +#endif + SARQ $0x03, R10 + LEAL (R11)(R10*1), R11 + JMP match_nolit_end_encodeSnappyBetterBlockAsm64K + +matchlen_match4_match_nolit_encodeSnappyBetterBlockAsm64K: + CMPL DI, $0x04 + JB matchlen_match2_match_nolit_encodeSnappyBetterBlockAsm64K + MOVL (R8)(R11*1), R10 + CMPL (R9)(R11*1), R10 + JNE matchlen_match2_match_nolit_encodeSnappyBetterBlockAsm64K + LEAL -4(DI), DI + LEAL 4(R11), R11 + +matchlen_match2_match_nolit_encodeSnappyBetterBlockAsm64K: + CMPL DI, $0x01 + JE matchlen_match1_match_nolit_encodeSnappyBetterBlockAsm64K + JB match_nolit_end_encodeSnappyBetterBlockAsm64K + MOVW (R8)(R11*1), R10 + CMPW (R9)(R11*1), R10 + JNE matchlen_match1_match_nolit_encodeSnappyBetterBlockAsm64K + LEAL 2(R11), R11 + SUBL $0x02, DI + JZ match_nolit_end_encodeSnappyBetterBlockAsm64K + +matchlen_match1_match_nolit_encodeSnappyBetterBlockAsm64K: + MOVB (R8)(R11*1), R10 + CMPB (R9)(R11*1), R10 + JNE match_nolit_end_encodeSnappyBetterBlockAsm64K + LEAL 1(R11), R11 + +match_nolit_end_encodeSnappyBetterBlockAsm64K: + MOVL CX, DI + SUBL BX, DI + + // Check if repeat + MOVL DI, 16(SP) + MOVL 12(SP), BX + CMPL BX, SI + JEQ emit_literal_done_match_emit_encodeSnappyBetterBlockAsm64K + MOVL SI, R8 + MOVL SI, 12(SP) + LEAQ (DX)(BX*1), R9 + SUBL BX, R8 + LEAL -1(R8), BX + CMPL BX, $0x3c + JB one_byte_match_emit_encodeSnappyBetterBlockAsm64K + CMPL BX, $0x00000100 + JB two_bytes_match_emit_encodeSnappyBetterBlockAsm64K + JB three_bytes_match_emit_encodeSnappyBetterBlockAsm64K + +three_bytes_match_emit_encodeSnappyBetterBlockAsm64K: + MOVB $0xf4, (AX) + MOVW BX, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_match_emit_encodeSnappyBetterBlockAsm64K + +two_bytes_match_emit_encodeSnappyBetterBlockAsm64K: + MOVB $0xf0, (AX) + MOVB BL, 1(AX) + ADDQ $0x02, AX + CMPL BX, $0x40 + JB memmove_match_emit_encodeSnappyBetterBlockAsm64K + JMP memmove_long_match_emit_encodeSnappyBetterBlockAsm64K + +one_byte_match_emit_encodeSnappyBetterBlockAsm64K: + SHLB $0x02, BL + MOVB BL, (AX) + ADDQ $0x01, AX + +memmove_match_emit_encodeSnappyBetterBlockAsm64K: + LEAQ (AX)(R8*1), BX + + // genMemMoveShort + CMPQ R8, $0x08 + JBE emit_lit_memmove_match_emit_encodeSnappyBetterBlockAsm64K_memmove_move_8 + CMPQ R8, $0x10 + JBE emit_lit_memmove_match_emit_encodeSnappyBetterBlockAsm64K_memmove_move_8through16 + CMPQ R8, $0x20 + JBE emit_lit_memmove_match_emit_encodeSnappyBetterBlockAsm64K_memmove_move_17through32 + JMP emit_lit_memmove_match_emit_encodeSnappyBetterBlockAsm64K_memmove_move_33through64 + +emit_lit_memmove_match_emit_encodeSnappyBetterBlockAsm64K_memmove_move_8: + MOVQ (R9), R10 + MOVQ R10, (AX) + JMP memmove_end_copy_match_emit_encodeSnappyBetterBlockAsm64K + +emit_lit_memmove_match_emit_encodeSnappyBetterBlockAsm64K_memmove_move_8through16: + MOVQ (R9), R10 + MOVQ -8(R9)(R8*1), R9 + MOVQ R10, (AX) + MOVQ R9, -8(AX)(R8*1) + JMP memmove_end_copy_match_emit_encodeSnappyBetterBlockAsm64K + +emit_lit_memmove_match_emit_encodeSnappyBetterBlockAsm64K_memmove_move_17through32: + MOVOU (R9), X0 + MOVOU -16(R9)(R8*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(R8*1) + JMP memmove_end_copy_match_emit_encodeSnappyBetterBlockAsm64K + +emit_lit_memmove_match_emit_encodeSnappyBetterBlockAsm64K_memmove_move_33through64: + MOVOU (R9), X0 + MOVOU 16(R9), X1 + MOVOU -32(R9)(R8*1), X2 + MOVOU -16(R9)(R8*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + +memmove_end_copy_match_emit_encodeSnappyBetterBlockAsm64K: + MOVQ BX, AX + JMP emit_literal_done_match_emit_encodeSnappyBetterBlockAsm64K + +memmove_long_match_emit_encodeSnappyBetterBlockAsm64K: + LEAQ (AX)(R8*1), BX + + // genMemMoveLong + MOVOU (R9), X0 + MOVOU 16(R9), X1 + MOVOU -32(R9)(R8*1), X2 + MOVOU -16(R9)(R8*1), X3 + MOVQ R8, R12 + SHRQ $0x05, R12 + MOVQ AX, R10 + ANDL $0x0000001f, R10 + MOVQ $0x00000040, R13 + SUBQ R10, R13 + DECQ R12 + JA emit_lit_memmove_long_match_emit_encodeSnappyBetterBlockAsm64Klarge_forward_sse_loop_32 + LEAQ -32(R9)(R13*1), R10 + LEAQ -32(AX)(R13*1), R14 + +emit_lit_memmove_long_match_emit_encodeSnappyBetterBlockAsm64Klarge_big_loop_back: + MOVOU (R10), X4 + MOVOU 16(R10), X5 + MOVOA X4, (R14) + MOVOA X5, 16(R14) + ADDQ $0x20, R14 + ADDQ $0x20, R10 + ADDQ $0x20, R13 + DECQ R12 + JNA emit_lit_memmove_long_match_emit_encodeSnappyBetterBlockAsm64Klarge_big_loop_back + +emit_lit_memmove_long_match_emit_encodeSnappyBetterBlockAsm64Klarge_forward_sse_loop_32: + MOVOU -32(R9)(R13*1), X4 + MOVOU -16(R9)(R13*1), X5 + MOVOA X4, -32(AX)(R13*1) + MOVOA X5, -16(AX)(R13*1) + ADDQ $0x20, R13 + CMPQ R8, R13 + JAE emit_lit_memmove_long_match_emit_encodeSnappyBetterBlockAsm64Klarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + MOVQ BX, AX + +emit_literal_done_match_emit_encodeSnappyBetterBlockAsm64K: + ADDL R11, CX + ADDL $0x04, R11 + MOVL CX, 12(SP) + + // emitCopy +two_byte_offset_match_nolit_encodeSnappyBetterBlockAsm64K: + CMPL R11, $0x40 + JBE two_byte_offset_short_match_nolit_encodeSnappyBetterBlockAsm64K + MOVB $0xee, (AX) + MOVW DI, 1(AX) + LEAL -60(R11), R11 + ADDQ $0x03, AX + JMP two_byte_offset_match_nolit_encodeSnappyBetterBlockAsm64K + +two_byte_offset_short_match_nolit_encodeSnappyBetterBlockAsm64K: + MOVL R11, BX + SHLL $0x02, BX + CMPL R11, $0x0c + JAE emit_copy_three_match_nolit_encodeSnappyBetterBlockAsm64K + CMPL DI, $0x00000800 + JAE emit_copy_three_match_nolit_encodeSnappyBetterBlockAsm64K + LEAL -15(BX), BX + MOVB DI, 1(AX) + SHRL $0x08, DI + SHLL $0x05, DI + ORL DI, BX + MOVB BL, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeSnappyBetterBlockAsm64K + +emit_copy_three_match_nolit_encodeSnappyBetterBlockAsm64K: + LEAL -2(BX), BX + MOVB BL, (AX) + MOVW DI, 1(AX) + ADDQ $0x03, AX + +match_nolit_emitcopy_end_encodeSnappyBetterBlockAsm64K: + CMPL CX, 8(SP) + JAE emit_remainder_encodeSnappyBetterBlockAsm64K + CMPQ AX, (SP) + JB match_nolit_dst_ok_encodeSnappyBetterBlockAsm64K + MOVQ $0x00000000, ret+48(FP) + RET + +match_nolit_dst_ok_encodeSnappyBetterBlockAsm64K: + MOVQ $0x00cf1bbcdcbfa563, BX + MOVQ $0x9e3779b1, DI + LEAQ 1(SI), SI + LEAQ -2(CX), R8 + MOVQ (DX)(SI*1), R9 + MOVQ 1(DX)(SI*1), R10 + MOVQ (DX)(R8*1), R11 + MOVQ 1(DX)(R8*1), R12 + SHLQ $0x08, R9 + IMULQ BX, R9 + SHRQ $0x30, R9 + SHLQ $0x20, R10 + IMULQ DI, R10 + SHRQ $0x32, R10 + SHLQ $0x08, R11 + IMULQ BX, R11 + SHRQ $0x30, R11 + SHLQ $0x20, R12 + IMULQ DI, R12 + SHRQ $0x32, R12 + LEAQ 1(SI), DI + LEAQ 1(R8), R13 + MOVL SI, 24(SP)(R9*4) + MOVL R8, 24(SP)(R11*4) + MOVL DI, 262168(SP)(R10*4) + MOVL R13, 262168(SP)(R12*4) + LEAQ 1(R8)(SI*1), DI + SHRQ $0x01, DI + ADDQ $0x01, SI + SUBQ $0x01, R8 + +index_loop_encodeSnappyBetterBlockAsm64K: + CMPQ DI, R8 + JAE search_loop_encodeSnappyBetterBlockAsm64K + MOVQ (DX)(SI*1), R9 + MOVQ (DX)(DI*1), R10 + SHLQ $0x08, R9 + IMULQ BX, R9 + SHRQ $0x30, R9 + SHLQ $0x08, R10 + IMULQ BX, R10 + SHRQ $0x30, R10 + MOVL SI, 24(SP)(R9*4) + MOVL DI, 24(SP)(R10*4) + ADDQ $0x02, SI + ADDQ $0x02, DI + JMP index_loop_encodeSnappyBetterBlockAsm64K + +emit_remainder_encodeSnappyBetterBlockAsm64K: + MOVQ src_len+32(FP), CX + SUBL 12(SP), CX + LEAQ 3(AX)(CX*1), CX + CMPQ CX, (SP) + JB emit_remainder_ok_encodeSnappyBetterBlockAsm64K + MOVQ $0x00000000, ret+48(FP) + RET + +emit_remainder_ok_encodeSnappyBetterBlockAsm64K: + MOVQ src_len+32(FP), CX + MOVL 12(SP), BX + CMPL BX, CX + JEQ emit_literal_done_emit_remainder_encodeSnappyBetterBlockAsm64K + MOVL CX, SI + MOVL CX, 12(SP) + LEAQ (DX)(BX*1), CX + SUBL BX, SI + LEAL -1(SI), DX + CMPL DX, $0x3c + JB one_byte_emit_remainder_encodeSnappyBetterBlockAsm64K + CMPL DX, $0x00000100 + JB two_bytes_emit_remainder_encodeSnappyBetterBlockAsm64K + JB three_bytes_emit_remainder_encodeSnappyBetterBlockAsm64K + +three_bytes_emit_remainder_encodeSnappyBetterBlockAsm64K: + MOVB $0xf4, (AX) + MOVW DX, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_emit_remainder_encodeSnappyBetterBlockAsm64K + +two_bytes_emit_remainder_encodeSnappyBetterBlockAsm64K: + MOVB $0xf0, (AX) + MOVB DL, 1(AX) + ADDQ $0x02, AX + CMPL DX, $0x40 + JB memmove_emit_remainder_encodeSnappyBetterBlockAsm64K + JMP memmove_long_emit_remainder_encodeSnappyBetterBlockAsm64K + +one_byte_emit_remainder_encodeSnappyBetterBlockAsm64K: + SHLB $0x02, DL + MOVB DL, (AX) + ADDQ $0x01, AX + +memmove_emit_remainder_encodeSnappyBetterBlockAsm64K: + LEAQ (AX)(SI*1), DX + MOVL SI, BX + + // genMemMoveShort + CMPQ BX, $0x03 + JB emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm64K_memmove_move_1or2 + JE emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm64K_memmove_move_3 + CMPQ BX, $0x08 + JB emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm64K_memmove_move_4through7 + CMPQ BX, $0x10 + JBE emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm64K_memmove_move_8through16 + CMPQ BX, $0x20 + JBE emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm64K_memmove_move_17through32 + JMP emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm64K_memmove_move_33through64 + +emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm64K_memmove_move_1or2: + MOVB (CX), SI + MOVB -1(CX)(BX*1), CL + MOVB SI, (AX) + MOVB CL, -1(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeSnappyBetterBlockAsm64K + +emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm64K_memmove_move_3: + MOVW (CX), SI + MOVB 2(CX), CL + MOVW SI, (AX) + MOVB CL, 2(AX) + JMP memmove_end_copy_emit_remainder_encodeSnappyBetterBlockAsm64K + +emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm64K_memmove_move_4through7: + MOVL (CX), SI + MOVL -4(CX)(BX*1), CX + MOVL SI, (AX) + MOVL CX, -4(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeSnappyBetterBlockAsm64K + +emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm64K_memmove_move_8through16: + MOVQ (CX), SI + MOVQ -8(CX)(BX*1), CX + MOVQ SI, (AX) + MOVQ CX, -8(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeSnappyBetterBlockAsm64K + +emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm64K_memmove_move_17through32: + MOVOU (CX), X0 + MOVOU -16(CX)(BX*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeSnappyBetterBlockAsm64K + +emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm64K_memmove_move_33through64: + MOVOU (CX), X0 + MOVOU 16(CX), X1 + MOVOU -32(CX)(BX*1), X2 + MOVOU -16(CX)(BX*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(BX*1) + MOVOU X3, -16(AX)(BX*1) + +memmove_end_copy_emit_remainder_encodeSnappyBetterBlockAsm64K: + MOVQ DX, AX + JMP emit_literal_done_emit_remainder_encodeSnappyBetterBlockAsm64K + +memmove_long_emit_remainder_encodeSnappyBetterBlockAsm64K: + LEAQ (AX)(SI*1), DX + MOVL SI, BX + + // genMemMoveLong + MOVOU (CX), X0 + MOVOU 16(CX), X1 + MOVOU -32(CX)(BX*1), X2 + MOVOU -16(CX)(BX*1), X3 + MOVQ BX, DI + SHRQ $0x05, DI + MOVQ AX, SI + ANDL $0x0000001f, SI + MOVQ $0x00000040, R8 + SUBQ SI, R8 + DECQ DI + JA emit_lit_memmove_long_emit_remainder_encodeSnappyBetterBlockAsm64Klarge_forward_sse_loop_32 + LEAQ -32(CX)(R8*1), SI + LEAQ -32(AX)(R8*1), R9 + +emit_lit_memmove_long_emit_remainder_encodeSnappyBetterBlockAsm64Klarge_big_loop_back: + MOVOU (SI), X4 + MOVOU 16(SI), X5 + MOVOA X4, (R9) + MOVOA X5, 16(R9) + ADDQ $0x20, R9 + ADDQ $0x20, SI + ADDQ $0x20, R8 + DECQ DI + JNA emit_lit_memmove_long_emit_remainder_encodeSnappyBetterBlockAsm64Klarge_big_loop_back + +emit_lit_memmove_long_emit_remainder_encodeSnappyBetterBlockAsm64Klarge_forward_sse_loop_32: + MOVOU -32(CX)(R8*1), X4 + MOVOU -16(CX)(R8*1), X5 + MOVOA X4, -32(AX)(R8*1) + MOVOA X5, -16(AX)(R8*1) + ADDQ $0x20, R8 + CMPQ BX, R8 + JAE emit_lit_memmove_long_emit_remainder_encodeSnappyBetterBlockAsm64Klarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(BX*1) + MOVOU X3, -16(AX)(BX*1) + MOVQ DX, AX + +emit_literal_done_emit_remainder_encodeSnappyBetterBlockAsm64K: + MOVQ dst_base+0(FP), CX + SUBQ CX, AX + MOVQ AX, ret+48(FP) + RET + +// func encodeSnappyBetterBlockAsm12B(dst []byte, src []byte) int +// Requires: BMI, SSE2 +TEXT ·encodeSnappyBetterBlockAsm12B(SB), $81944-56 + MOVQ dst_base+0(FP), AX + MOVQ $0x00000280, CX + LEAQ 24(SP), DX + PXOR X0, X0 + +zero_loop_encodeSnappyBetterBlockAsm12B: + MOVOU X0, (DX) + MOVOU X0, 16(DX) + MOVOU X0, 32(DX) + MOVOU X0, 48(DX) + MOVOU X0, 64(DX) + MOVOU X0, 80(DX) + MOVOU X0, 96(DX) + MOVOU X0, 112(DX) + ADDQ $0x80, DX + DECQ CX + JNZ zero_loop_encodeSnappyBetterBlockAsm12B + MOVL $0x00000000, 12(SP) + MOVQ src_len+32(FP), CX + LEAQ -9(CX), DX + LEAQ -8(CX), BX + MOVL BX, 8(SP) + SHRQ $0x05, CX + SUBL CX, DX + LEAQ (AX)(DX*1), DX + MOVQ DX, (SP) + MOVL $0x00000001, CX + MOVL $0x00000000, 16(SP) + MOVQ src_base+24(FP), DX + +search_loop_encodeSnappyBetterBlockAsm12B: + MOVL CX, BX + SUBL 12(SP), BX + SHRL $0x06, BX + LEAL 1(CX)(BX*1), BX + CMPL BX, 8(SP) + JAE emit_remainder_encodeSnappyBetterBlockAsm12B + MOVQ (DX)(CX*1), SI + MOVL BX, 20(SP) + MOVQ $0x0000cf1bbcdcbf9b, R8 + MOVQ $0x9e3779b1, BX + MOVQ SI, R9 + MOVQ SI, R10 + SHLQ $0x10, R9 + IMULQ R8, R9 + SHRQ $0x32, R9 + SHLQ $0x20, R10 + IMULQ BX, R10 + SHRQ $0x34, R10 + MOVL 24(SP)(R9*4), BX + MOVL 65560(SP)(R10*4), DI + MOVL CX, 24(SP)(R9*4) + MOVL CX, 65560(SP)(R10*4) + MOVQ (DX)(BX*1), R9 + MOVQ (DX)(DI*1), R10 + CMPQ R9, SI + JEQ candidate_match_encodeSnappyBetterBlockAsm12B + CMPQ R10, SI + JNE no_short_found_encodeSnappyBetterBlockAsm12B + MOVL DI, BX + JMP candidate_match_encodeSnappyBetterBlockAsm12B + +no_short_found_encodeSnappyBetterBlockAsm12B: + CMPL R9, SI + JEQ candidate_match_encodeSnappyBetterBlockAsm12B + CMPL R10, SI + JEQ candidateS_match_encodeSnappyBetterBlockAsm12B + MOVL 20(SP), CX + JMP search_loop_encodeSnappyBetterBlockAsm12B + +candidateS_match_encodeSnappyBetterBlockAsm12B: + SHRQ $0x08, SI + MOVQ SI, R9 + SHLQ $0x10, R9 + IMULQ R8, R9 + SHRQ $0x32, R9 + MOVL 24(SP)(R9*4), BX + INCL CX + MOVL CX, 24(SP)(R9*4) + CMPL (DX)(BX*1), SI + JEQ candidate_match_encodeSnappyBetterBlockAsm12B + DECL CX + MOVL DI, BX + +candidate_match_encodeSnappyBetterBlockAsm12B: + MOVL 12(SP), SI + TESTL BX, BX + JZ match_extend_back_end_encodeSnappyBetterBlockAsm12B + +match_extend_back_loop_encodeSnappyBetterBlockAsm12B: + CMPL CX, SI + JBE match_extend_back_end_encodeSnappyBetterBlockAsm12B + MOVB -1(DX)(BX*1), DI + MOVB -1(DX)(CX*1), R8 + CMPB DI, R8 + JNE match_extend_back_end_encodeSnappyBetterBlockAsm12B + LEAL -1(CX), CX + DECL BX + JZ match_extend_back_end_encodeSnappyBetterBlockAsm12B + JMP match_extend_back_loop_encodeSnappyBetterBlockAsm12B + +match_extend_back_end_encodeSnappyBetterBlockAsm12B: + MOVL CX, SI + SUBL 12(SP), SI + LEAQ 3(AX)(SI*1), SI + CMPQ SI, (SP) + JB match_dst_size_check_encodeSnappyBetterBlockAsm12B + MOVQ $0x00000000, ret+48(FP) + RET + +match_dst_size_check_encodeSnappyBetterBlockAsm12B: + MOVL CX, SI + ADDL $0x04, CX + ADDL $0x04, BX + MOVQ src_len+32(FP), DI + SUBL CX, DI + LEAQ (DX)(CX*1), R8 + LEAQ (DX)(BX*1), R9 + + // matchLen + XORL R11, R11 + +matchlen_loopback_16_match_nolit_encodeSnappyBetterBlockAsm12B: + CMPL DI, $0x10 + JB matchlen_match8_match_nolit_encodeSnappyBetterBlockAsm12B + MOVQ (R8)(R11*1), R10 + MOVQ 8(R8)(R11*1), R12 + XORQ (R9)(R11*1), R10 + JNZ matchlen_bsf_8_match_nolit_encodeSnappyBetterBlockAsm12B + XORQ 8(R9)(R11*1), R12 + JNZ matchlen_bsf_16match_nolit_encodeSnappyBetterBlockAsm12B + LEAL -16(DI), DI + LEAL 16(R11), R11 + JMP matchlen_loopback_16_match_nolit_encodeSnappyBetterBlockAsm12B + +matchlen_bsf_16match_nolit_encodeSnappyBetterBlockAsm12B: +#ifdef GOAMD64_v3 + TZCNTQ R12, R12 + +#else + BSFQ R12, R12 + +#endif + SARQ $0x03, R12 + LEAL 8(R11)(R12*1), R11 + JMP match_nolit_end_encodeSnappyBetterBlockAsm12B + +matchlen_match8_match_nolit_encodeSnappyBetterBlockAsm12B: + CMPL DI, $0x08 + JB matchlen_match4_match_nolit_encodeSnappyBetterBlockAsm12B + MOVQ (R8)(R11*1), R10 + XORQ (R9)(R11*1), R10 + JNZ matchlen_bsf_8_match_nolit_encodeSnappyBetterBlockAsm12B + LEAL -8(DI), DI + LEAL 8(R11), R11 + JMP matchlen_match4_match_nolit_encodeSnappyBetterBlockAsm12B + +matchlen_bsf_8_match_nolit_encodeSnappyBetterBlockAsm12B: +#ifdef GOAMD64_v3 + TZCNTQ R10, R10 + +#else + BSFQ R10, R10 + +#endif + SARQ $0x03, R10 + LEAL (R11)(R10*1), R11 + JMP match_nolit_end_encodeSnappyBetterBlockAsm12B + +matchlen_match4_match_nolit_encodeSnappyBetterBlockAsm12B: + CMPL DI, $0x04 + JB matchlen_match2_match_nolit_encodeSnappyBetterBlockAsm12B + MOVL (R8)(R11*1), R10 + CMPL (R9)(R11*1), R10 + JNE matchlen_match2_match_nolit_encodeSnappyBetterBlockAsm12B + LEAL -4(DI), DI + LEAL 4(R11), R11 + +matchlen_match2_match_nolit_encodeSnappyBetterBlockAsm12B: + CMPL DI, $0x01 + JE matchlen_match1_match_nolit_encodeSnappyBetterBlockAsm12B + JB match_nolit_end_encodeSnappyBetterBlockAsm12B + MOVW (R8)(R11*1), R10 + CMPW (R9)(R11*1), R10 + JNE matchlen_match1_match_nolit_encodeSnappyBetterBlockAsm12B + LEAL 2(R11), R11 + SUBL $0x02, DI + JZ match_nolit_end_encodeSnappyBetterBlockAsm12B + +matchlen_match1_match_nolit_encodeSnappyBetterBlockAsm12B: + MOVB (R8)(R11*1), R10 + CMPB (R9)(R11*1), R10 + JNE match_nolit_end_encodeSnappyBetterBlockAsm12B + LEAL 1(R11), R11 + +match_nolit_end_encodeSnappyBetterBlockAsm12B: + MOVL CX, DI + SUBL BX, DI + + // Check if repeat + MOVL DI, 16(SP) + MOVL 12(SP), BX + CMPL BX, SI + JEQ emit_literal_done_match_emit_encodeSnappyBetterBlockAsm12B + MOVL SI, R8 + MOVL SI, 12(SP) + LEAQ (DX)(BX*1), R9 + SUBL BX, R8 + LEAL -1(R8), BX + CMPL BX, $0x3c + JB one_byte_match_emit_encodeSnappyBetterBlockAsm12B + CMPL BX, $0x00000100 + JB two_bytes_match_emit_encodeSnappyBetterBlockAsm12B + JB three_bytes_match_emit_encodeSnappyBetterBlockAsm12B + +three_bytes_match_emit_encodeSnappyBetterBlockAsm12B: + MOVB $0xf4, (AX) + MOVW BX, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_match_emit_encodeSnappyBetterBlockAsm12B + +two_bytes_match_emit_encodeSnappyBetterBlockAsm12B: + MOVB $0xf0, (AX) + MOVB BL, 1(AX) + ADDQ $0x02, AX + CMPL BX, $0x40 + JB memmove_match_emit_encodeSnappyBetterBlockAsm12B + JMP memmove_long_match_emit_encodeSnappyBetterBlockAsm12B + +one_byte_match_emit_encodeSnappyBetterBlockAsm12B: + SHLB $0x02, BL + MOVB BL, (AX) + ADDQ $0x01, AX + +memmove_match_emit_encodeSnappyBetterBlockAsm12B: + LEAQ (AX)(R8*1), BX + + // genMemMoveShort + CMPQ R8, $0x08 + JBE emit_lit_memmove_match_emit_encodeSnappyBetterBlockAsm12B_memmove_move_8 + CMPQ R8, $0x10 + JBE emit_lit_memmove_match_emit_encodeSnappyBetterBlockAsm12B_memmove_move_8through16 + CMPQ R8, $0x20 + JBE emit_lit_memmove_match_emit_encodeSnappyBetterBlockAsm12B_memmove_move_17through32 + JMP emit_lit_memmove_match_emit_encodeSnappyBetterBlockAsm12B_memmove_move_33through64 + +emit_lit_memmove_match_emit_encodeSnappyBetterBlockAsm12B_memmove_move_8: + MOVQ (R9), R10 + MOVQ R10, (AX) + JMP memmove_end_copy_match_emit_encodeSnappyBetterBlockAsm12B + +emit_lit_memmove_match_emit_encodeSnappyBetterBlockAsm12B_memmove_move_8through16: + MOVQ (R9), R10 + MOVQ -8(R9)(R8*1), R9 + MOVQ R10, (AX) + MOVQ R9, -8(AX)(R8*1) + JMP memmove_end_copy_match_emit_encodeSnappyBetterBlockAsm12B + +emit_lit_memmove_match_emit_encodeSnappyBetterBlockAsm12B_memmove_move_17through32: + MOVOU (R9), X0 + MOVOU -16(R9)(R8*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(R8*1) + JMP memmove_end_copy_match_emit_encodeSnappyBetterBlockAsm12B + +emit_lit_memmove_match_emit_encodeSnappyBetterBlockAsm12B_memmove_move_33through64: + MOVOU (R9), X0 + MOVOU 16(R9), X1 + MOVOU -32(R9)(R8*1), X2 + MOVOU -16(R9)(R8*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + +memmove_end_copy_match_emit_encodeSnappyBetterBlockAsm12B: + MOVQ BX, AX + JMP emit_literal_done_match_emit_encodeSnappyBetterBlockAsm12B + +memmove_long_match_emit_encodeSnappyBetterBlockAsm12B: + LEAQ (AX)(R8*1), BX + + // genMemMoveLong + MOVOU (R9), X0 + MOVOU 16(R9), X1 + MOVOU -32(R9)(R8*1), X2 + MOVOU -16(R9)(R8*1), X3 + MOVQ R8, R12 + SHRQ $0x05, R12 + MOVQ AX, R10 + ANDL $0x0000001f, R10 + MOVQ $0x00000040, R13 + SUBQ R10, R13 + DECQ R12 + JA emit_lit_memmove_long_match_emit_encodeSnappyBetterBlockAsm12Blarge_forward_sse_loop_32 + LEAQ -32(R9)(R13*1), R10 + LEAQ -32(AX)(R13*1), R14 + +emit_lit_memmove_long_match_emit_encodeSnappyBetterBlockAsm12Blarge_big_loop_back: + MOVOU (R10), X4 + MOVOU 16(R10), X5 + MOVOA X4, (R14) + MOVOA X5, 16(R14) + ADDQ $0x20, R14 + ADDQ $0x20, R10 + ADDQ $0x20, R13 + DECQ R12 + JNA emit_lit_memmove_long_match_emit_encodeSnappyBetterBlockAsm12Blarge_big_loop_back + +emit_lit_memmove_long_match_emit_encodeSnappyBetterBlockAsm12Blarge_forward_sse_loop_32: + MOVOU -32(R9)(R13*1), X4 + MOVOU -16(R9)(R13*1), X5 + MOVOA X4, -32(AX)(R13*1) + MOVOA X5, -16(AX)(R13*1) + ADDQ $0x20, R13 + CMPQ R8, R13 + JAE emit_lit_memmove_long_match_emit_encodeSnappyBetterBlockAsm12Blarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + MOVQ BX, AX + +emit_literal_done_match_emit_encodeSnappyBetterBlockAsm12B: + ADDL R11, CX + ADDL $0x04, R11 + MOVL CX, 12(SP) + + // emitCopy +two_byte_offset_match_nolit_encodeSnappyBetterBlockAsm12B: + CMPL R11, $0x40 + JBE two_byte_offset_short_match_nolit_encodeSnappyBetterBlockAsm12B + MOVB $0xee, (AX) + MOVW DI, 1(AX) + LEAL -60(R11), R11 + ADDQ $0x03, AX + JMP two_byte_offset_match_nolit_encodeSnappyBetterBlockAsm12B + +two_byte_offset_short_match_nolit_encodeSnappyBetterBlockAsm12B: + MOVL R11, BX + SHLL $0x02, BX + CMPL R11, $0x0c + JAE emit_copy_three_match_nolit_encodeSnappyBetterBlockAsm12B + CMPL DI, $0x00000800 + JAE emit_copy_three_match_nolit_encodeSnappyBetterBlockAsm12B + LEAL -15(BX), BX + MOVB DI, 1(AX) + SHRL $0x08, DI + SHLL $0x05, DI + ORL DI, BX + MOVB BL, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeSnappyBetterBlockAsm12B + +emit_copy_three_match_nolit_encodeSnappyBetterBlockAsm12B: + LEAL -2(BX), BX + MOVB BL, (AX) + MOVW DI, 1(AX) + ADDQ $0x03, AX + +match_nolit_emitcopy_end_encodeSnappyBetterBlockAsm12B: + CMPL CX, 8(SP) + JAE emit_remainder_encodeSnappyBetterBlockAsm12B + CMPQ AX, (SP) + JB match_nolit_dst_ok_encodeSnappyBetterBlockAsm12B + MOVQ $0x00000000, ret+48(FP) + RET + +match_nolit_dst_ok_encodeSnappyBetterBlockAsm12B: + MOVQ $0x0000cf1bbcdcbf9b, BX + MOVQ $0x9e3779b1, DI + LEAQ 1(SI), SI + LEAQ -2(CX), R8 + MOVQ (DX)(SI*1), R9 + MOVQ 1(DX)(SI*1), R10 + MOVQ (DX)(R8*1), R11 + MOVQ 1(DX)(R8*1), R12 + SHLQ $0x10, R9 + IMULQ BX, R9 + SHRQ $0x32, R9 + SHLQ $0x20, R10 + IMULQ DI, R10 + SHRQ $0x34, R10 + SHLQ $0x10, R11 + IMULQ BX, R11 + SHRQ $0x32, R11 + SHLQ $0x20, R12 + IMULQ DI, R12 + SHRQ $0x34, R12 + LEAQ 1(SI), DI + LEAQ 1(R8), R13 + MOVL SI, 24(SP)(R9*4) + MOVL R8, 24(SP)(R11*4) + MOVL DI, 65560(SP)(R10*4) + MOVL R13, 65560(SP)(R12*4) + LEAQ 1(R8)(SI*1), DI + SHRQ $0x01, DI + ADDQ $0x01, SI + SUBQ $0x01, R8 + +index_loop_encodeSnappyBetterBlockAsm12B: + CMPQ DI, R8 + JAE search_loop_encodeSnappyBetterBlockAsm12B + MOVQ (DX)(SI*1), R9 + MOVQ (DX)(DI*1), R10 + SHLQ $0x10, R9 + IMULQ BX, R9 + SHRQ $0x32, R9 + SHLQ $0x10, R10 + IMULQ BX, R10 + SHRQ $0x32, R10 + MOVL SI, 24(SP)(R9*4) + MOVL DI, 24(SP)(R10*4) + ADDQ $0x02, SI + ADDQ $0x02, DI + JMP index_loop_encodeSnappyBetterBlockAsm12B + +emit_remainder_encodeSnappyBetterBlockAsm12B: + MOVQ src_len+32(FP), CX + SUBL 12(SP), CX + LEAQ 3(AX)(CX*1), CX + CMPQ CX, (SP) + JB emit_remainder_ok_encodeSnappyBetterBlockAsm12B + MOVQ $0x00000000, ret+48(FP) + RET + +emit_remainder_ok_encodeSnappyBetterBlockAsm12B: + MOVQ src_len+32(FP), CX + MOVL 12(SP), BX + CMPL BX, CX + JEQ emit_literal_done_emit_remainder_encodeSnappyBetterBlockAsm12B + MOVL CX, SI + MOVL CX, 12(SP) + LEAQ (DX)(BX*1), CX + SUBL BX, SI + LEAL -1(SI), DX + CMPL DX, $0x3c + JB one_byte_emit_remainder_encodeSnappyBetterBlockAsm12B + CMPL DX, $0x00000100 + JB two_bytes_emit_remainder_encodeSnappyBetterBlockAsm12B + JB three_bytes_emit_remainder_encodeSnappyBetterBlockAsm12B + +three_bytes_emit_remainder_encodeSnappyBetterBlockAsm12B: + MOVB $0xf4, (AX) + MOVW DX, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_emit_remainder_encodeSnappyBetterBlockAsm12B + +two_bytes_emit_remainder_encodeSnappyBetterBlockAsm12B: + MOVB $0xf0, (AX) + MOVB DL, 1(AX) + ADDQ $0x02, AX + CMPL DX, $0x40 + JB memmove_emit_remainder_encodeSnappyBetterBlockAsm12B + JMP memmove_long_emit_remainder_encodeSnappyBetterBlockAsm12B + +one_byte_emit_remainder_encodeSnappyBetterBlockAsm12B: + SHLB $0x02, DL + MOVB DL, (AX) + ADDQ $0x01, AX + +memmove_emit_remainder_encodeSnappyBetterBlockAsm12B: + LEAQ (AX)(SI*1), DX + MOVL SI, BX + + // genMemMoveShort + CMPQ BX, $0x03 + JB emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm12B_memmove_move_1or2 + JE emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm12B_memmove_move_3 + CMPQ BX, $0x08 + JB emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm12B_memmove_move_4through7 + CMPQ BX, $0x10 + JBE emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm12B_memmove_move_8through16 + CMPQ BX, $0x20 + JBE emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm12B_memmove_move_17through32 + JMP emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm12B_memmove_move_33through64 + +emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm12B_memmove_move_1or2: + MOVB (CX), SI + MOVB -1(CX)(BX*1), CL + MOVB SI, (AX) + MOVB CL, -1(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeSnappyBetterBlockAsm12B + +emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm12B_memmove_move_3: + MOVW (CX), SI + MOVB 2(CX), CL + MOVW SI, (AX) + MOVB CL, 2(AX) + JMP memmove_end_copy_emit_remainder_encodeSnappyBetterBlockAsm12B + +emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm12B_memmove_move_4through7: + MOVL (CX), SI + MOVL -4(CX)(BX*1), CX + MOVL SI, (AX) + MOVL CX, -4(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeSnappyBetterBlockAsm12B + +emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm12B_memmove_move_8through16: + MOVQ (CX), SI + MOVQ -8(CX)(BX*1), CX + MOVQ SI, (AX) + MOVQ CX, -8(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeSnappyBetterBlockAsm12B + +emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm12B_memmove_move_17through32: + MOVOU (CX), X0 + MOVOU -16(CX)(BX*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeSnappyBetterBlockAsm12B + +emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm12B_memmove_move_33through64: + MOVOU (CX), X0 + MOVOU 16(CX), X1 + MOVOU -32(CX)(BX*1), X2 + MOVOU -16(CX)(BX*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(BX*1) + MOVOU X3, -16(AX)(BX*1) + +memmove_end_copy_emit_remainder_encodeSnappyBetterBlockAsm12B: + MOVQ DX, AX + JMP emit_literal_done_emit_remainder_encodeSnappyBetterBlockAsm12B + +memmove_long_emit_remainder_encodeSnappyBetterBlockAsm12B: + LEAQ (AX)(SI*1), DX + MOVL SI, BX + + // genMemMoveLong + MOVOU (CX), X0 + MOVOU 16(CX), X1 + MOVOU -32(CX)(BX*1), X2 + MOVOU -16(CX)(BX*1), X3 + MOVQ BX, DI + SHRQ $0x05, DI + MOVQ AX, SI + ANDL $0x0000001f, SI + MOVQ $0x00000040, R8 + SUBQ SI, R8 + DECQ DI + JA emit_lit_memmove_long_emit_remainder_encodeSnappyBetterBlockAsm12Blarge_forward_sse_loop_32 + LEAQ -32(CX)(R8*1), SI + LEAQ -32(AX)(R8*1), R9 + +emit_lit_memmove_long_emit_remainder_encodeSnappyBetterBlockAsm12Blarge_big_loop_back: + MOVOU (SI), X4 + MOVOU 16(SI), X5 + MOVOA X4, (R9) + MOVOA X5, 16(R9) + ADDQ $0x20, R9 + ADDQ $0x20, SI + ADDQ $0x20, R8 + DECQ DI + JNA emit_lit_memmove_long_emit_remainder_encodeSnappyBetterBlockAsm12Blarge_big_loop_back + +emit_lit_memmove_long_emit_remainder_encodeSnappyBetterBlockAsm12Blarge_forward_sse_loop_32: + MOVOU -32(CX)(R8*1), X4 + MOVOU -16(CX)(R8*1), X5 + MOVOA X4, -32(AX)(R8*1) + MOVOA X5, -16(AX)(R8*1) + ADDQ $0x20, R8 + CMPQ BX, R8 + JAE emit_lit_memmove_long_emit_remainder_encodeSnappyBetterBlockAsm12Blarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(BX*1) + MOVOU X3, -16(AX)(BX*1) + MOVQ DX, AX + +emit_literal_done_emit_remainder_encodeSnappyBetterBlockAsm12B: + MOVQ dst_base+0(FP), CX + SUBQ CX, AX + MOVQ AX, ret+48(FP) + RET + +// func encodeSnappyBetterBlockAsm10B(dst []byte, src []byte) int +// Requires: BMI, SSE2 +TEXT ·encodeSnappyBetterBlockAsm10B(SB), $20504-56 + MOVQ dst_base+0(FP), AX + MOVQ $0x000000a0, CX + LEAQ 24(SP), DX + PXOR X0, X0 + +zero_loop_encodeSnappyBetterBlockAsm10B: + MOVOU X0, (DX) + MOVOU X0, 16(DX) + MOVOU X0, 32(DX) + MOVOU X0, 48(DX) + MOVOU X0, 64(DX) + MOVOU X0, 80(DX) + MOVOU X0, 96(DX) + MOVOU X0, 112(DX) + ADDQ $0x80, DX + DECQ CX + JNZ zero_loop_encodeSnappyBetterBlockAsm10B + MOVL $0x00000000, 12(SP) + MOVQ src_len+32(FP), CX + LEAQ -9(CX), DX + LEAQ -8(CX), BX + MOVL BX, 8(SP) + SHRQ $0x05, CX + SUBL CX, DX + LEAQ (AX)(DX*1), DX + MOVQ DX, (SP) + MOVL $0x00000001, CX + MOVL $0x00000000, 16(SP) + MOVQ src_base+24(FP), DX + +search_loop_encodeSnappyBetterBlockAsm10B: + MOVL CX, BX + SUBL 12(SP), BX + SHRL $0x05, BX + LEAL 1(CX)(BX*1), BX + CMPL BX, 8(SP) + JAE emit_remainder_encodeSnappyBetterBlockAsm10B + MOVQ (DX)(CX*1), SI + MOVL BX, 20(SP) + MOVQ $0x0000cf1bbcdcbf9b, R8 + MOVQ $0x9e3779b1, BX + MOVQ SI, R9 + MOVQ SI, R10 + SHLQ $0x10, R9 + IMULQ R8, R9 + SHRQ $0x34, R9 + SHLQ $0x20, R10 + IMULQ BX, R10 + SHRQ $0x36, R10 + MOVL 24(SP)(R9*4), BX + MOVL 16408(SP)(R10*4), DI + MOVL CX, 24(SP)(R9*4) + MOVL CX, 16408(SP)(R10*4) + MOVQ (DX)(BX*1), R9 + MOVQ (DX)(DI*1), R10 + CMPQ R9, SI + JEQ candidate_match_encodeSnappyBetterBlockAsm10B + CMPQ R10, SI + JNE no_short_found_encodeSnappyBetterBlockAsm10B + MOVL DI, BX + JMP candidate_match_encodeSnappyBetterBlockAsm10B + +no_short_found_encodeSnappyBetterBlockAsm10B: + CMPL R9, SI + JEQ candidate_match_encodeSnappyBetterBlockAsm10B + CMPL R10, SI + JEQ candidateS_match_encodeSnappyBetterBlockAsm10B + MOVL 20(SP), CX + JMP search_loop_encodeSnappyBetterBlockAsm10B + +candidateS_match_encodeSnappyBetterBlockAsm10B: + SHRQ $0x08, SI + MOVQ SI, R9 + SHLQ $0x10, R9 + IMULQ R8, R9 + SHRQ $0x34, R9 + MOVL 24(SP)(R9*4), BX + INCL CX + MOVL CX, 24(SP)(R9*4) + CMPL (DX)(BX*1), SI + JEQ candidate_match_encodeSnappyBetterBlockAsm10B + DECL CX + MOVL DI, BX + +candidate_match_encodeSnappyBetterBlockAsm10B: + MOVL 12(SP), SI + TESTL BX, BX + JZ match_extend_back_end_encodeSnappyBetterBlockAsm10B + +match_extend_back_loop_encodeSnappyBetterBlockAsm10B: + CMPL CX, SI + JBE match_extend_back_end_encodeSnappyBetterBlockAsm10B + MOVB -1(DX)(BX*1), DI + MOVB -1(DX)(CX*1), R8 + CMPB DI, R8 + JNE match_extend_back_end_encodeSnappyBetterBlockAsm10B + LEAL -1(CX), CX + DECL BX + JZ match_extend_back_end_encodeSnappyBetterBlockAsm10B + JMP match_extend_back_loop_encodeSnappyBetterBlockAsm10B + +match_extend_back_end_encodeSnappyBetterBlockAsm10B: + MOVL CX, SI + SUBL 12(SP), SI + LEAQ 3(AX)(SI*1), SI + CMPQ SI, (SP) + JB match_dst_size_check_encodeSnappyBetterBlockAsm10B + MOVQ $0x00000000, ret+48(FP) + RET + +match_dst_size_check_encodeSnappyBetterBlockAsm10B: + MOVL CX, SI + ADDL $0x04, CX + ADDL $0x04, BX + MOVQ src_len+32(FP), DI + SUBL CX, DI + LEAQ (DX)(CX*1), R8 + LEAQ (DX)(BX*1), R9 + + // matchLen + XORL R11, R11 + +matchlen_loopback_16_match_nolit_encodeSnappyBetterBlockAsm10B: + CMPL DI, $0x10 + JB matchlen_match8_match_nolit_encodeSnappyBetterBlockAsm10B + MOVQ (R8)(R11*1), R10 + MOVQ 8(R8)(R11*1), R12 + XORQ (R9)(R11*1), R10 + JNZ matchlen_bsf_8_match_nolit_encodeSnappyBetterBlockAsm10B + XORQ 8(R9)(R11*1), R12 + JNZ matchlen_bsf_16match_nolit_encodeSnappyBetterBlockAsm10B + LEAL -16(DI), DI + LEAL 16(R11), R11 + JMP matchlen_loopback_16_match_nolit_encodeSnappyBetterBlockAsm10B + +matchlen_bsf_16match_nolit_encodeSnappyBetterBlockAsm10B: +#ifdef GOAMD64_v3 + TZCNTQ R12, R12 + +#else + BSFQ R12, R12 + +#endif + SARQ $0x03, R12 + LEAL 8(R11)(R12*1), R11 + JMP match_nolit_end_encodeSnappyBetterBlockAsm10B + +matchlen_match8_match_nolit_encodeSnappyBetterBlockAsm10B: + CMPL DI, $0x08 + JB matchlen_match4_match_nolit_encodeSnappyBetterBlockAsm10B + MOVQ (R8)(R11*1), R10 + XORQ (R9)(R11*1), R10 + JNZ matchlen_bsf_8_match_nolit_encodeSnappyBetterBlockAsm10B + LEAL -8(DI), DI + LEAL 8(R11), R11 + JMP matchlen_match4_match_nolit_encodeSnappyBetterBlockAsm10B + +matchlen_bsf_8_match_nolit_encodeSnappyBetterBlockAsm10B: +#ifdef GOAMD64_v3 + TZCNTQ R10, R10 + +#else + BSFQ R10, R10 + +#endif + SARQ $0x03, R10 + LEAL (R11)(R10*1), R11 + JMP match_nolit_end_encodeSnappyBetterBlockAsm10B + +matchlen_match4_match_nolit_encodeSnappyBetterBlockAsm10B: + CMPL DI, $0x04 + JB matchlen_match2_match_nolit_encodeSnappyBetterBlockAsm10B + MOVL (R8)(R11*1), R10 + CMPL (R9)(R11*1), R10 + JNE matchlen_match2_match_nolit_encodeSnappyBetterBlockAsm10B + LEAL -4(DI), DI + LEAL 4(R11), R11 + +matchlen_match2_match_nolit_encodeSnappyBetterBlockAsm10B: + CMPL DI, $0x01 + JE matchlen_match1_match_nolit_encodeSnappyBetterBlockAsm10B + JB match_nolit_end_encodeSnappyBetterBlockAsm10B + MOVW (R8)(R11*1), R10 + CMPW (R9)(R11*1), R10 + JNE matchlen_match1_match_nolit_encodeSnappyBetterBlockAsm10B + LEAL 2(R11), R11 + SUBL $0x02, DI + JZ match_nolit_end_encodeSnappyBetterBlockAsm10B + +matchlen_match1_match_nolit_encodeSnappyBetterBlockAsm10B: + MOVB (R8)(R11*1), R10 + CMPB (R9)(R11*1), R10 + JNE match_nolit_end_encodeSnappyBetterBlockAsm10B + LEAL 1(R11), R11 + +match_nolit_end_encodeSnappyBetterBlockAsm10B: + MOVL CX, DI + SUBL BX, DI + + // Check if repeat + MOVL DI, 16(SP) + MOVL 12(SP), BX + CMPL BX, SI + JEQ emit_literal_done_match_emit_encodeSnappyBetterBlockAsm10B + MOVL SI, R8 + MOVL SI, 12(SP) + LEAQ (DX)(BX*1), R9 + SUBL BX, R8 + LEAL -1(R8), BX + CMPL BX, $0x3c + JB one_byte_match_emit_encodeSnappyBetterBlockAsm10B + CMPL BX, $0x00000100 + JB two_bytes_match_emit_encodeSnappyBetterBlockAsm10B + JB three_bytes_match_emit_encodeSnappyBetterBlockAsm10B + +three_bytes_match_emit_encodeSnappyBetterBlockAsm10B: + MOVB $0xf4, (AX) + MOVW BX, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_match_emit_encodeSnappyBetterBlockAsm10B + +two_bytes_match_emit_encodeSnappyBetterBlockAsm10B: + MOVB $0xf0, (AX) + MOVB BL, 1(AX) + ADDQ $0x02, AX + CMPL BX, $0x40 + JB memmove_match_emit_encodeSnappyBetterBlockAsm10B + JMP memmove_long_match_emit_encodeSnappyBetterBlockAsm10B + +one_byte_match_emit_encodeSnappyBetterBlockAsm10B: + SHLB $0x02, BL + MOVB BL, (AX) + ADDQ $0x01, AX + +memmove_match_emit_encodeSnappyBetterBlockAsm10B: + LEAQ (AX)(R8*1), BX + + // genMemMoveShort + CMPQ R8, $0x08 + JBE emit_lit_memmove_match_emit_encodeSnappyBetterBlockAsm10B_memmove_move_8 + CMPQ R8, $0x10 + JBE emit_lit_memmove_match_emit_encodeSnappyBetterBlockAsm10B_memmove_move_8through16 + CMPQ R8, $0x20 + JBE emit_lit_memmove_match_emit_encodeSnappyBetterBlockAsm10B_memmove_move_17through32 + JMP emit_lit_memmove_match_emit_encodeSnappyBetterBlockAsm10B_memmove_move_33through64 + +emit_lit_memmove_match_emit_encodeSnappyBetterBlockAsm10B_memmove_move_8: + MOVQ (R9), R10 + MOVQ R10, (AX) + JMP memmove_end_copy_match_emit_encodeSnappyBetterBlockAsm10B + +emit_lit_memmove_match_emit_encodeSnappyBetterBlockAsm10B_memmove_move_8through16: + MOVQ (R9), R10 + MOVQ -8(R9)(R8*1), R9 + MOVQ R10, (AX) + MOVQ R9, -8(AX)(R8*1) + JMP memmove_end_copy_match_emit_encodeSnappyBetterBlockAsm10B + +emit_lit_memmove_match_emit_encodeSnappyBetterBlockAsm10B_memmove_move_17through32: + MOVOU (R9), X0 + MOVOU -16(R9)(R8*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(R8*1) + JMP memmove_end_copy_match_emit_encodeSnappyBetterBlockAsm10B + +emit_lit_memmove_match_emit_encodeSnappyBetterBlockAsm10B_memmove_move_33through64: + MOVOU (R9), X0 + MOVOU 16(R9), X1 + MOVOU -32(R9)(R8*1), X2 + MOVOU -16(R9)(R8*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + +memmove_end_copy_match_emit_encodeSnappyBetterBlockAsm10B: + MOVQ BX, AX + JMP emit_literal_done_match_emit_encodeSnappyBetterBlockAsm10B + +memmove_long_match_emit_encodeSnappyBetterBlockAsm10B: + LEAQ (AX)(R8*1), BX + + // genMemMoveLong + MOVOU (R9), X0 + MOVOU 16(R9), X1 + MOVOU -32(R9)(R8*1), X2 + MOVOU -16(R9)(R8*1), X3 + MOVQ R8, R12 + SHRQ $0x05, R12 + MOVQ AX, R10 + ANDL $0x0000001f, R10 + MOVQ $0x00000040, R13 + SUBQ R10, R13 + DECQ R12 + JA emit_lit_memmove_long_match_emit_encodeSnappyBetterBlockAsm10Blarge_forward_sse_loop_32 + LEAQ -32(R9)(R13*1), R10 + LEAQ -32(AX)(R13*1), R14 + +emit_lit_memmove_long_match_emit_encodeSnappyBetterBlockAsm10Blarge_big_loop_back: + MOVOU (R10), X4 + MOVOU 16(R10), X5 + MOVOA X4, (R14) + MOVOA X5, 16(R14) + ADDQ $0x20, R14 + ADDQ $0x20, R10 + ADDQ $0x20, R13 + DECQ R12 + JNA emit_lit_memmove_long_match_emit_encodeSnappyBetterBlockAsm10Blarge_big_loop_back + +emit_lit_memmove_long_match_emit_encodeSnappyBetterBlockAsm10Blarge_forward_sse_loop_32: + MOVOU -32(R9)(R13*1), X4 + MOVOU -16(R9)(R13*1), X5 + MOVOA X4, -32(AX)(R13*1) + MOVOA X5, -16(AX)(R13*1) + ADDQ $0x20, R13 + CMPQ R8, R13 + JAE emit_lit_memmove_long_match_emit_encodeSnappyBetterBlockAsm10Blarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + MOVQ BX, AX + +emit_literal_done_match_emit_encodeSnappyBetterBlockAsm10B: + ADDL R11, CX + ADDL $0x04, R11 + MOVL CX, 12(SP) + + // emitCopy +two_byte_offset_match_nolit_encodeSnappyBetterBlockAsm10B: + CMPL R11, $0x40 + JBE two_byte_offset_short_match_nolit_encodeSnappyBetterBlockAsm10B + MOVB $0xee, (AX) + MOVW DI, 1(AX) + LEAL -60(R11), R11 + ADDQ $0x03, AX + JMP two_byte_offset_match_nolit_encodeSnappyBetterBlockAsm10B + +two_byte_offset_short_match_nolit_encodeSnappyBetterBlockAsm10B: + MOVL R11, BX + SHLL $0x02, BX + CMPL R11, $0x0c + JAE emit_copy_three_match_nolit_encodeSnappyBetterBlockAsm10B + CMPL DI, $0x00000800 + JAE emit_copy_three_match_nolit_encodeSnappyBetterBlockAsm10B + LEAL -15(BX), BX + MOVB DI, 1(AX) + SHRL $0x08, DI + SHLL $0x05, DI + ORL DI, BX + MOVB BL, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeSnappyBetterBlockAsm10B + +emit_copy_three_match_nolit_encodeSnappyBetterBlockAsm10B: + LEAL -2(BX), BX + MOVB BL, (AX) + MOVW DI, 1(AX) + ADDQ $0x03, AX + +match_nolit_emitcopy_end_encodeSnappyBetterBlockAsm10B: + CMPL CX, 8(SP) + JAE emit_remainder_encodeSnappyBetterBlockAsm10B + CMPQ AX, (SP) + JB match_nolit_dst_ok_encodeSnappyBetterBlockAsm10B + MOVQ $0x00000000, ret+48(FP) + RET + +match_nolit_dst_ok_encodeSnappyBetterBlockAsm10B: + MOVQ $0x0000cf1bbcdcbf9b, BX + MOVQ $0x9e3779b1, DI + LEAQ 1(SI), SI + LEAQ -2(CX), R8 + MOVQ (DX)(SI*1), R9 + MOVQ 1(DX)(SI*1), R10 + MOVQ (DX)(R8*1), R11 + MOVQ 1(DX)(R8*1), R12 + SHLQ $0x10, R9 + IMULQ BX, R9 + SHRQ $0x34, R9 + SHLQ $0x20, R10 + IMULQ DI, R10 + SHRQ $0x36, R10 + SHLQ $0x10, R11 + IMULQ BX, R11 + SHRQ $0x34, R11 + SHLQ $0x20, R12 + IMULQ DI, R12 + SHRQ $0x36, R12 + LEAQ 1(SI), DI + LEAQ 1(R8), R13 + MOVL SI, 24(SP)(R9*4) + MOVL R8, 24(SP)(R11*4) + MOVL DI, 16408(SP)(R10*4) + MOVL R13, 16408(SP)(R12*4) + LEAQ 1(R8)(SI*1), DI + SHRQ $0x01, DI + ADDQ $0x01, SI + SUBQ $0x01, R8 + +index_loop_encodeSnappyBetterBlockAsm10B: + CMPQ DI, R8 + JAE search_loop_encodeSnappyBetterBlockAsm10B + MOVQ (DX)(SI*1), R9 + MOVQ (DX)(DI*1), R10 + SHLQ $0x10, R9 + IMULQ BX, R9 + SHRQ $0x34, R9 + SHLQ $0x10, R10 + IMULQ BX, R10 + SHRQ $0x34, R10 + MOVL SI, 24(SP)(R9*4) + MOVL DI, 24(SP)(R10*4) + ADDQ $0x02, SI + ADDQ $0x02, DI + JMP index_loop_encodeSnappyBetterBlockAsm10B + +emit_remainder_encodeSnappyBetterBlockAsm10B: + MOVQ src_len+32(FP), CX + SUBL 12(SP), CX + LEAQ 3(AX)(CX*1), CX + CMPQ CX, (SP) + JB emit_remainder_ok_encodeSnappyBetterBlockAsm10B + MOVQ $0x00000000, ret+48(FP) + RET + +emit_remainder_ok_encodeSnappyBetterBlockAsm10B: + MOVQ src_len+32(FP), CX + MOVL 12(SP), BX + CMPL BX, CX + JEQ emit_literal_done_emit_remainder_encodeSnappyBetterBlockAsm10B + MOVL CX, SI + MOVL CX, 12(SP) + LEAQ (DX)(BX*1), CX + SUBL BX, SI + LEAL -1(SI), DX + CMPL DX, $0x3c + JB one_byte_emit_remainder_encodeSnappyBetterBlockAsm10B + CMPL DX, $0x00000100 + JB two_bytes_emit_remainder_encodeSnappyBetterBlockAsm10B + JB three_bytes_emit_remainder_encodeSnappyBetterBlockAsm10B + +three_bytes_emit_remainder_encodeSnappyBetterBlockAsm10B: + MOVB $0xf4, (AX) + MOVW DX, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_emit_remainder_encodeSnappyBetterBlockAsm10B + +two_bytes_emit_remainder_encodeSnappyBetterBlockAsm10B: + MOVB $0xf0, (AX) + MOVB DL, 1(AX) + ADDQ $0x02, AX + CMPL DX, $0x40 + JB memmove_emit_remainder_encodeSnappyBetterBlockAsm10B + JMP memmove_long_emit_remainder_encodeSnappyBetterBlockAsm10B + +one_byte_emit_remainder_encodeSnappyBetterBlockAsm10B: + SHLB $0x02, DL + MOVB DL, (AX) + ADDQ $0x01, AX + +memmove_emit_remainder_encodeSnappyBetterBlockAsm10B: + LEAQ (AX)(SI*1), DX + MOVL SI, BX + + // genMemMoveShort + CMPQ BX, $0x03 + JB emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm10B_memmove_move_1or2 + JE emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm10B_memmove_move_3 + CMPQ BX, $0x08 + JB emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm10B_memmove_move_4through7 + CMPQ BX, $0x10 + JBE emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm10B_memmove_move_8through16 + CMPQ BX, $0x20 + JBE emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm10B_memmove_move_17through32 + JMP emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm10B_memmove_move_33through64 + +emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm10B_memmove_move_1or2: + MOVB (CX), SI + MOVB -1(CX)(BX*1), CL + MOVB SI, (AX) + MOVB CL, -1(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeSnappyBetterBlockAsm10B + +emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm10B_memmove_move_3: + MOVW (CX), SI + MOVB 2(CX), CL + MOVW SI, (AX) + MOVB CL, 2(AX) + JMP memmove_end_copy_emit_remainder_encodeSnappyBetterBlockAsm10B + +emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm10B_memmove_move_4through7: + MOVL (CX), SI + MOVL -4(CX)(BX*1), CX + MOVL SI, (AX) + MOVL CX, -4(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeSnappyBetterBlockAsm10B + +emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm10B_memmove_move_8through16: + MOVQ (CX), SI + MOVQ -8(CX)(BX*1), CX + MOVQ SI, (AX) + MOVQ CX, -8(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeSnappyBetterBlockAsm10B + +emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm10B_memmove_move_17through32: + MOVOU (CX), X0 + MOVOU -16(CX)(BX*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeSnappyBetterBlockAsm10B + +emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm10B_memmove_move_33through64: + MOVOU (CX), X0 + MOVOU 16(CX), X1 + MOVOU -32(CX)(BX*1), X2 + MOVOU -16(CX)(BX*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(BX*1) + MOVOU X3, -16(AX)(BX*1) + +memmove_end_copy_emit_remainder_encodeSnappyBetterBlockAsm10B: + MOVQ DX, AX + JMP emit_literal_done_emit_remainder_encodeSnappyBetterBlockAsm10B + +memmove_long_emit_remainder_encodeSnappyBetterBlockAsm10B: + LEAQ (AX)(SI*1), DX + MOVL SI, BX + + // genMemMoveLong + MOVOU (CX), X0 + MOVOU 16(CX), X1 + MOVOU -32(CX)(BX*1), X2 + MOVOU -16(CX)(BX*1), X3 + MOVQ BX, DI + SHRQ $0x05, DI + MOVQ AX, SI + ANDL $0x0000001f, SI + MOVQ $0x00000040, R8 + SUBQ SI, R8 + DECQ DI + JA emit_lit_memmove_long_emit_remainder_encodeSnappyBetterBlockAsm10Blarge_forward_sse_loop_32 + LEAQ -32(CX)(R8*1), SI + LEAQ -32(AX)(R8*1), R9 + +emit_lit_memmove_long_emit_remainder_encodeSnappyBetterBlockAsm10Blarge_big_loop_back: + MOVOU (SI), X4 + MOVOU 16(SI), X5 + MOVOA X4, (R9) + MOVOA X5, 16(R9) + ADDQ $0x20, R9 + ADDQ $0x20, SI + ADDQ $0x20, R8 + DECQ DI + JNA emit_lit_memmove_long_emit_remainder_encodeSnappyBetterBlockAsm10Blarge_big_loop_back + +emit_lit_memmove_long_emit_remainder_encodeSnappyBetterBlockAsm10Blarge_forward_sse_loop_32: + MOVOU -32(CX)(R8*1), X4 + MOVOU -16(CX)(R8*1), X5 + MOVOA X4, -32(AX)(R8*1) + MOVOA X5, -16(AX)(R8*1) + ADDQ $0x20, R8 + CMPQ BX, R8 + JAE emit_lit_memmove_long_emit_remainder_encodeSnappyBetterBlockAsm10Blarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(BX*1) + MOVOU X3, -16(AX)(BX*1) + MOVQ DX, AX + +emit_literal_done_emit_remainder_encodeSnappyBetterBlockAsm10B: + MOVQ dst_base+0(FP), CX + SUBQ CX, AX + MOVQ AX, ret+48(FP) + RET + +// func encodeSnappyBetterBlockAsm8B(dst []byte, src []byte) int +// Requires: BMI, SSE2 +TEXT ·encodeSnappyBetterBlockAsm8B(SB), $5144-56 + MOVQ dst_base+0(FP), AX + MOVQ $0x00000028, CX + LEAQ 24(SP), DX + PXOR X0, X0 + +zero_loop_encodeSnappyBetterBlockAsm8B: + MOVOU X0, (DX) + MOVOU X0, 16(DX) + MOVOU X0, 32(DX) + MOVOU X0, 48(DX) + MOVOU X0, 64(DX) + MOVOU X0, 80(DX) + MOVOU X0, 96(DX) + MOVOU X0, 112(DX) + ADDQ $0x80, DX + DECQ CX + JNZ zero_loop_encodeSnappyBetterBlockAsm8B + MOVL $0x00000000, 12(SP) + MOVQ src_len+32(FP), CX + LEAQ -9(CX), DX + LEAQ -8(CX), BX + MOVL BX, 8(SP) + SHRQ $0x05, CX + SUBL CX, DX + LEAQ (AX)(DX*1), DX + MOVQ DX, (SP) + MOVL $0x00000001, CX + MOVL $0x00000000, 16(SP) + MOVQ src_base+24(FP), DX + +search_loop_encodeSnappyBetterBlockAsm8B: + MOVL CX, BX + SUBL 12(SP), BX + SHRL $0x04, BX + LEAL 1(CX)(BX*1), BX + CMPL BX, 8(SP) + JAE emit_remainder_encodeSnappyBetterBlockAsm8B + MOVQ (DX)(CX*1), SI + MOVL BX, 20(SP) + MOVQ $0x0000cf1bbcdcbf9b, R8 + MOVQ $0x9e3779b1, BX + MOVQ SI, R9 + MOVQ SI, R10 + SHLQ $0x10, R9 + IMULQ R8, R9 + SHRQ $0x36, R9 + SHLQ $0x20, R10 + IMULQ BX, R10 + SHRQ $0x38, R10 + MOVL 24(SP)(R9*4), BX + MOVL 4120(SP)(R10*4), DI + MOVL CX, 24(SP)(R9*4) + MOVL CX, 4120(SP)(R10*4) + MOVQ (DX)(BX*1), R9 + MOVQ (DX)(DI*1), R10 + CMPQ R9, SI + JEQ candidate_match_encodeSnappyBetterBlockAsm8B + CMPQ R10, SI + JNE no_short_found_encodeSnappyBetterBlockAsm8B + MOVL DI, BX + JMP candidate_match_encodeSnappyBetterBlockAsm8B + +no_short_found_encodeSnappyBetterBlockAsm8B: + CMPL R9, SI + JEQ candidate_match_encodeSnappyBetterBlockAsm8B + CMPL R10, SI + JEQ candidateS_match_encodeSnappyBetterBlockAsm8B + MOVL 20(SP), CX + JMP search_loop_encodeSnappyBetterBlockAsm8B + +candidateS_match_encodeSnappyBetterBlockAsm8B: + SHRQ $0x08, SI + MOVQ SI, R9 + SHLQ $0x10, R9 + IMULQ R8, R9 + SHRQ $0x36, R9 + MOVL 24(SP)(R9*4), BX + INCL CX + MOVL CX, 24(SP)(R9*4) + CMPL (DX)(BX*1), SI + JEQ candidate_match_encodeSnappyBetterBlockAsm8B + DECL CX + MOVL DI, BX + +candidate_match_encodeSnappyBetterBlockAsm8B: + MOVL 12(SP), SI + TESTL BX, BX + JZ match_extend_back_end_encodeSnappyBetterBlockAsm8B + +match_extend_back_loop_encodeSnappyBetterBlockAsm8B: + CMPL CX, SI + JBE match_extend_back_end_encodeSnappyBetterBlockAsm8B + MOVB -1(DX)(BX*1), DI + MOVB -1(DX)(CX*1), R8 + CMPB DI, R8 + JNE match_extend_back_end_encodeSnappyBetterBlockAsm8B + LEAL -1(CX), CX + DECL BX + JZ match_extend_back_end_encodeSnappyBetterBlockAsm8B + JMP match_extend_back_loop_encodeSnappyBetterBlockAsm8B + +match_extend_back_end_encodeSnappyBetterBlockAsm8B: + MOVL CX, SI + SUBL 12(SP), SI + LEAQ 3(AX)(SI*1), SI + CMPQ SI, (SP) + JB match_dst_size_check_encodeSnappyBetterBlockAsm8B + MOVQ $0x00000000, ret+48(FP) + RET + +match_dst_size_check_encodeSnappyBetterBlockAsm8B: + MOVL CX, SI + ADDL $0x04, CX + ADDL $0x04, BX + MOVQ src_len+32(FP), DI + SUBL CX, DI + LEAQ (DX)(CX*1), R8 + LEAQ (DX)(BX*1), R9 + + // matchLen + XORL R11, R11 + +matchlen_loopback_16_match_nolit_encodeSnappyBetterBlockAsm8B: + CMPL DI, $0x10 + JB matchlen_match8_match_nolit_encodeSnappyBetterBlockAsm8B + MOVQ (R8)(R11*1), R10 + MOVQ 8(R8)(R11*1), R12 + XORQ (R9)(R11*1), R10 + JNZ matchlen_bsf_8_match_nolit_encodeSnappyBetterBlockAsm8B + XORQ 8(R9)(R11*1), R12 + JNZ matchlen_bsf_16match_nolit_encodeSnappyBetterBlockAsm8B + LEAL -16(DI), DI + LEAL 16(R11), R11 + JMP matchlen_loopback_16_match_nolit_encodeSnappyBetterBlockAsm8B + +matchlen_bsf_16match_nolit_encodeSnappyBetterBlockAsm8B: +#ifdef GOAMD64_v3 + TZCNTQ R12, R12 + +#else + BSFQ R12, R12 + +#endif + SARQ $0x03, R12 + LEAL 8(R11)(R12*1), R11 + JMP match_nolit_end_encodeSnappyBetterBlockAsm8B + +matchlen_match8_match_nolit_encodeSnappyBetterBlockAsm8B: + CMPL DI, $0x08 + JB matchlen_match4_match_nolit_encodeSnappyBetterBlockAsm8B + MOVQ (R8)(R11*1), R10 + XORQ (R9)(R11*1), R10 + JNZ matchlen_bsf_8_match_nolit_encodeSnappyBetterBlockAsm8B + LEAL -8(DI), DI + LEAL 8(R11), R11 + JMP matchlen_match4_match_nolit_encodeSnappyBetterBlockAsm8B + +matchlen_bsf_8_match_nolit_encodeSnappyBetterBlockAsm8B: +#ifdef GOAMD64_v3 + TZCNTQ R10, R10 + +#else + BSFQ R10, R10 + +#endif + SARQ $0x03, R10 + LEAL (R11)(R10*1), R11 + JMP match_nolit_end_encodeSnappyBetterBlockAsm8B + +matchlen_match4_match_nolit_encodeSnappyBetterBlockAsm8B: + CMPL DI, $0x04 + JB matchlen_match2_match_nolit_encodeSnappyBetterBlockAsm8B + MOVL (R8)(R11*1), R10 + CMPL (R9)(R11*1), R10 + JNE matchlen_match2_match_nolit_encodeSnappyBetterBlockAsm8B + LEAL -4(DI), DI + LEAL 4(R11), R11 + +matchlen_match2_match_nolit_encodeSnappyBetterBlockAsm8B: + CMPL DI, $0x01 + JE matchlen_match1_match_nolit_encodeSnappyBetterBlockAsm8B + JB match_nolit_end_encodeSnappyBetterBlockAsm8B + MOVW (R8)(R11*1), R10 + CMPW (R9)(R11*1), R10 + JNE matchlen_match1_match_nolit_encodeSnappyBetterBlockAsm8B + LEAL 2(R11), R11 + SUBL $0x02, DI + JZ match_nolit_end_encodeSnappyBetterBlockAsm8B + +matchlen_match1_match_nolit_encodeSnappyBetterBlockAsm8B: + MOVB (R8)(R11*1), R10 + CMPB (R9)(R11*1), R10 + JNE match_nolit_end_encodeSnappyBetterBlockAsm8B + LEAL 1(R11), R11 + +match_nolit_end_encodeSnappyBetterBlockAsm8B: + MOVL CX, DI + SUBL BX, DI + + // Check if repeat + MOVL DI, 16(SP) + MOVL 12(SP), BX + CMPL BX, SI + JEQ emit_literal_done_match_emit_encodeSnappyBetterBlockAsm8B + MOVL SI, R8 + MOVL SI, 12(SP) + LEAQ (DX)(BX*1), R9 + SUBL BX, R8 + LEAL -1(R8), BX + CMPL BX, $0x3c + JB one_byte_match_emit_encodeSnappyBetterBlockAsm8B + CMPL BX, $0x00000100 + JB two_bytes_match_emit_encodeSnappyBetterBlockAsm8B + JB three_bytes_match_emit_encodeSnappyBetterBlockAsm8B + +three_bytes_match_emit_encodeSnappyBetterBlockAsm8B: + MOVB $0xf4, (AX) + MOVW BX, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_match_emit_encodeSnappyBetterBlockAsm8B + +two_bytes_match_emit_encodeSnappyBetterBlockAsm8B: + MOVB $0xf0, (AX) + MOVB BL, 1(AX) + ADDQ $0x02, AX + CMPL BX, $0x40 + JB memmove_match_emit_encodeSnappyBetterBlockAsm8B + JMP memmove_long_match_emit_encodeSnappyBetterBlockAsm8B + +one_byte_match_emit_encodeSnappyBetterBlockAsm8B: + SHLB $0x02, BL + MOVB BL, (AX) + ADDQ $0x01, AX + +memmove_match_emit_encodeSnappyBetterBlockAsm8B: + LEAQ (AX)(R8*1), BX + + // genMemMoveShort + CMPQ R8, $0x08 + JBE emit_lit_memmove_match_emit_encodeSnappyBetterBlockAsm8B_memmove_move_8 + CMPQ R8, $0x10 + JBE emit_lit_memmove_match_emit_encodeSnappyBetterBlockAsm8B_memmove_move_8through16 + CMPQ R8, $0x20 + JBE emit_lit_memmove_match_emit_encodeSnappyBetterBlockAsm8B_memmove_move_17through32 + JMP emit_lit_memmove_match_emit_encodeSnappyBetterBlockAsm8B_memmove_move_33through64 + +emit_lit_memmove_match_emit_encodeSnappyBetterBlockAsm8B_memmove_move_8: + MOVQ (R9), R10 + MOVQ R10, (AX) + JMP memmove_end_copy_match_emit_encodeSnappyBetterBlockAsm8B + +emit_lit_memmove_match_emit_encodeSnappyBetterBlockAsm8B_memmove_move_8through16: + MOVQ (R9), R10 + MOVQ -8(R9)(R8*1), R9 + MOVQ R10, (AX) + MOVQ R9, -8(AX)(R8*1) + JMP memmove_end_copy_match_emit_encodeSnappyBetterBlockAsm8B + +emit_lit_memmove_match_emit_encodeSnappyBetterBlockAsm8B_memmove_move_17through32: + MOVOU (R9), X0 + MOVOU -16(R9)(R8*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(R8*1) + JMP memmove_end_copy_match_emit_encodeSnappyBetterBlockAsm8B + +emit_lit_memmove_match_emit_encodeSnappyBetterBlockAsm8B_memmove_move_33through64: + MOVOU (R9), X0 + MOVOU 16(R9), X1 + MOVOU -32(R9)(R8*1), X2 + MOVOU -16(R9)(R8*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + +memmove_end_copy_match_emit_encodeSnappyBetterBlockAsm8B: + MOVQ BX, AX + JMP emit_literal_done_match_emit_encodeSnappyBetterBlockAsm8B + +memmove_long_match_emit_encodeSnappyBetterBlockAsm8B: + LEAQ (AX)(R8*1), BX + + // genMemMoveLong + MOVOU (R9), X0 + MOVOU 16(R9), X1 + MOVOU -32(R9)(R8*1), X2 + MOVOU -16(R9)(R8*1), X3 + MOVQ R8, R12 + SHRQ $0x05, R12 + MOVQ AX, R10 + ANDL $0x0000001f, R10 + MOVQ $0x00000040, R13 + SUBQ R10, R13 + DECQ R12 + JA emit_lit_memmove_long_match_emit_encodeSnappyBetterBlockAsm8Blarge_forward_sse_loop_32 + LEAQ -32(R9)(R13*1), R10 + LEAQ -32(AX)(R13*1), R14 + +emit_lit_memmove_long_match_emit_encodeSnappyBetterBlockAsm8Blarge_big_loop_back: + MOVOU (R10), X4 + MOVOU 16(R10), X5 + MOVOA X4, (R14) + MOVOA X5, 16(R14) + ADDQ $0x20, R14 + ADDQ $0x20, R10 + ADDQ $0x20, R13 + DECQ R12 + JNA emit_lit_memmove_long_match_emit_encodeSnappyBetterBlockAsm8Blarge_big_loop_back + +emit_lit_memmove_long_match_emit_encodeSnappyBetterBlockAsm8Blarge_forward_sse_loop_32: + MOVOU -32(R9)(R13*1), X4 + MOVOU -16(R9)(R13*1), X5 + MOVOA X4, -32(AX)(R13*1) + MOVOA X5, -16(AX)(R13*1) + ADDQ $0x20, R13 + CMPQ R8, R13 + JAE emit_lit_memmove_long_match_emit_encodeSnappyBetterBlockAsm8Blarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + MOVQ BX, AX + +emit_literal_done_match_emit_encodeSnappyBetterBlockAsm8B: + ADDL R11, CX + ADDL $0x04, R11 + MOVL CX, 12(SP) + + // emitCopy +two_byte_offset_match_nolit_encodeSnappyBetterBlockAsm8B: + CMPL R11, $0x40 + JBE two_byte_offset_short_match_nolit_encodeSnappyBetterBlockAsm8B + MOVB $0xee, (AX) + MOVW DI, 1(AX) + LEAL -60(R11), R11 + ADDQ $0x03, AX + JMP two_byte_offset_match_nolit_encodeSnappyBetterBlockAsm8B + +two_byte_offset_short_match_nolit_encodeSnappyBetterBlockAsm8B: + MOVL R11, BX + SHLL $0x02, BX + CMPL R11, $0x0c + JAE emit_copy_three_match_nolit_encodeSnappyBetterBlockAsm8B + LEAL -15(BX), BX + MOVB DI, 1(AX) + SHRL $0x08, DI + SHLL $0x05, DI + ORL DI, BX + MOVB BL, (AX) + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_encodeSnappyBetterBlockAsm8B + +emit_copy_three_match_nolit_encodeSnappyBetterBlockAsm8B: + LEAL -2(BX), BX + MOVB BL, (AX) + MOVW DI, 1(AX) + ADDQ $0x03, AX + +match_nolit_emitcopy_end_encodeSnappyBetterBlockAsm8B: + CMPL CX, 8(SP) + JAE emit_remainder_encodeSnappyBetterBlockAsm8B + CMPQ AX, (SP) + JB match_nolit_dst_ok_encodeSnappyBetterBlockAsm8B + MOVQ $0x00000000, ret+48(FP) + RET + +match_nolit_dst_ok_encodeSnappyBetterBlockAsm8B: + MOVQ $0x0000cf1bbcdcbf9b, BX + MOVQ $0x9e3779b1, DI + LEAQ 1(SI), SI + LEAQ -2(CX), R8 + MOVQ (DX)(SI*1), R9 + MOVQ 1(DX)(SI*1), R10 + MOVQ (DX)(R8*1), R11 + MOVQ 1(DX)(R8*1), R12 + SHLQ $0x10, R9 + IMULQ BX, R9 + SHRQ $0x36, R9 + SHLQ $0x20, R10 + IMULQ DI, R10 + SHRQ $0x38, R10 + SHLQ $0x10, R11 + IMULQ BX, R11 + SHRQ $0x36, R11 + SHLQ $0x20, R12 + IMULQ DI, R12 + SHRQ $0x38, R12 + LEAQ 1(SI), DI + LEAQ 1(R8), R13 + MOVL SI, 24(SP)(R9*4) + MOVL R8, 24(SP)(R11*4) + MOVL DI, 4120(SP)(R10*4) + MOVL R13, 4120(SP)(R12*4) + LEAQ 1(R8)(SI*1), DI + SHRQ $0x01, DI + ADDQ $0x01, SI + SUBQ $0x01, R8 + +index_loop_encodeSnappyBetterBlockAsm8B: + CMPQ DI, R8 + JAE search_loop_encodeSnappyBetterBlockAsm8B + MOVQ (DX)(SI*1), R9 + MOVQ (DX)(DI*1), R10 + SHLQ $0x10, R9 + IMULQ BX, R9 + SHRQ $0x36, R9 + SHLQ $0x10, R10 + IMULQ BX, R10 + SHRQ $0x36, R10 + MOVL SI, 24(SP)(R9*4) + MOVL DI, 24(SP)(R10*4) + ADDQ $0x02, SI + ADDQ $0x02, DI + JMP index_loop_encodeSnappyBetterBlockAsm8B + +emit_remainder_encodeSnappyBetterBlockAsm8B: + MOVQ src_len+32(FP), CX + SUBL 12(SP), CX + LEAQ 3(AX)(CX*1), CX + CMPQ CX, (SP) + JB emit_remainder_ok_encodeSnappyBetterBlockAsm8B + MOVQ $0x00000000, ret+48(FP) + RET + +emit_remainder_ok_encodeSnappyBetterBlockAsm8B: + MOVQ src_len+32(FP), CX + MOVL 12(SP), BX + CMPL BX, CX + JEQ emit_literal_done_emit_remainder_encodeSnappyBetterBlockAsm8B + MOVL CX, SI + MOVL CX, 12(SP) + LEAQ (DX)(BX*1), CX + SUBL BX, SI + LEAL -1(SI), DX + CMPL DX, $0x3c + JB one_byte_emit_remainder_encodeSnappyBetterBlockAsm8B + CMPL DX, $0x00000100 + JB two_bytes_emit_remainder_encodeSnappyBetterBlockAsm8B + JB three_bytes_emit_remainder_encodeSnappyBetterBlockAsm8B + +three_bytes_emit_remainder_encodeSnappyBetterBlockAsm8B: + MOVB $0xf4, (AX) + MOVW DX, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_emit_remainder_encodeSnappyBetterBlockAsm8B + +two_bytes_emit_remainder_encodeSnappyBetterBlockAsm8B: + MOVB $0xf0, (AX) + MOVB DL, 1(AX) + ADDQ $0x02, AX + CMPL DX, $0x40 + JB memmove_emit_remainder_encodeSnappyBetterBlockAsm8B + JMP memmove_long_emit_remainder_encodeSnappyBetterBlockAsm8B + +one_byte_emit_remainder_encodeSnappyBetterBlockAsm8B: + SHLB $0x02, DL + MOVB DL, (AX) + ADDQ $0x01, AX + +memmove_emit_remainder_encodeSnappyBetterBlockAsm8B: + LEAQ (AX)(SI*1), DX + MOVL SI, BX + + // genMemMoveShort + CMPQ BX, $0x03 + JB emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm8B_memmove_move_1or2 + JE emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm8B_memmove_move_3 + CMPQ BX, $0x08 + JB emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm8B_memmove_move_4through7 + CMPQ BX, $0x10 + JBE emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm8B_memmove_move_8through16 + CMPQ BX, $0x20 + JBE emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm8B_memmove_move_17through32 + JMP emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm8B_memmove_move_33through64 + +emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm8B_memmove_move_1or2: + MOVB (CX), SI + MOVB -1(CX)(BX*1), CL + MOVB SI, (AX) + MOVB CL, -1(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeSnappyBetterBlockAsm8B + +emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm8B_memmove_move_3: + MOVW (CX), SI + MOVB 2(CX), CL + MOVW SI, (AX) + MOVB CL, 2(AX) + JMP memmove_end_copy_emit_remainder_encodeSnappyBetterBlockAsm8B + +emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm8B_memmove_move_4through7: + MOVL (CX), SI + MOVL -4(CX)(BX*1), CX + MOVL SI, (AX) + MOVL CX, -4(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeSnappyBetterBlockAsm8B + +emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm8B_memmove_move_8through16: + MOVQ (CX), SI + MOVQ -8(CX)(BX*1), CX + MOVQ SI, (AX) + MOVQ CX, -8(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeSnappyBetterBlockAsm8B + +emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm8B_memmove_move_17through32: + MOVOU (CX), X0 + MOVOU -16(CX)(BX*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(BX*1) + JMP memmove_end_copy_emit_remainder_encodeSnappyBetterBlockAsm8B + +emit_lit_memmove_emit_remainder_encodeSnappyBetterBlockAsm8B_memmove_move_33through64: + MOVOU (CX), X0 + MOVOU 16(CX), X1 + MOVOU -32(CX)(BX*1), X2 + MOVOU -16(CX)(BX*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(BX*1) + MOVOU X3, -16(AX)(BX*1) + +memmove_end_copy_emit_remainder_encodeSnappyBetterBlockAsm8B: + MOVQ DX, AX + JMP emit_literal_done_emit_remainder_encodeSnappyBetterBlockAsm8B + +memmove_long_emit_remainder_encodeSnappyBetterBlockAsm8B: + LEAQ (AX)(SI*1), DX + MOVL SI, BX + + // genMemMoveLong + MOVOU (CX), X0 + MOVOU 16(CX), X1 + MOVOU -32(CX)(BX*1), X2 + MOVOU -16(CX)(BX*1), X3 + MOVQ BX, DI + SHRQ $0x05, DI + MOVQ AX, SI + ANDL $0x0000001f, SI + MOVQ $0x00000040, R8 + SUBQ SI, R8 + DECQ DI + JA emit_lit_memmove_long_emit_remainder_encodeSnappyBetterBlockAsm8Blarge_forward_sse_loop_32 + LEAQ -32(CX)(R8*1), SI + LEAQ -32(AX)(R8*1), R9 + +emit_lit_memmove_long_emit_remainder_encodeSnappyBetterBlockAsm8Blarge_big_loop_back: + MOVOU (SI), X4 + MOVOU 16(SI), X5 + MOVOA X4, (R9) + MOVOA X5, 16(R9) + ADDQ $0x20, R9 + ADDQ $0x20, SI + ADDQ $0x20, R8 + DECQ DI + JNA emit_lit_memmove_long_emit_remainder_encodeSnappyBetterBlockAsm8Blarge_big_loop_back + +emit_lit_memmove_long_emit_remainder_encodeSnappyBetterBlockAsm8Blarge_forward_sse_loop_32: + MOVOU -32(CX)(R8*1), X4 + MOVOU -16(CX)(R8*1), X5 + MOVOA X4, -32(AX)(R8*1) + MOVOA X5, -16(AX)(R8*1) + ADDQ $0x20, R8 + CMPQ BX, R8 + JAE emit_lit_memmove_long_emit_remainder_encodeSnappyBetterBlockAsm8Blarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(BX*1) + MOVOU X3, -16(AX)(BX*1) + MOVQ DX, AX + +emit_literal_done_emit_remainder_encodeSnappyBetterBlockAsm8B: + MOVQ dst_base+0(FP), CX + SUBQ CX, AX + MOVQ AX, ret+48(FP) + RET + +// func calcBlockSize(src []byte) int +// Requires: BMI, SSE2 +TEXT ·calcBlockSize(SB), $32792-32 + XORQ AX, AX + MOVQ $0x00000100, CX + LEAQ 24(SP), DX + PXOR X0, X0 + +zero_loop_calcBlockSize: + MOVOU X0, (DX) + MOVOU X0, 16(DX) + MOVOU X0, 32(DX) + MOVOU X0, 48(DX) + MOVOU X0, 64(DX) + MOVOU X0, 80(DX) + MOVOU X0, 96(DX) + MOVOU X0, 112(DX) + ADDQ $0x80, DX + DECQ CX + JNZ zero_loop_calcBlockSize + MOVL $0x00000000, 12(SP) + MOVQ src_len+8(FP), CX + LEAQ -9(CX), DX + LEAQ -8(CX), BX + MOVL BX, 8(SP) + SHRQ $0x05, CX + SUBL CX, DX + LEAQ (AX)(DX*1), DX + MOVQ DX, (SP) + MOVL $0x00000001, CX + MOVL CX, 16(SP) + MOVQ src_base+0(FP), DX + +search_loop_calcBlockSize: + MOVL CX, BX + SUBL 12(SP), BX + SHRL $0x05, BX + LEAL 4(CX)(BX*1), BX + CMPL BX, 8(SP) + JAE emit_remainder_calcBlockSize + MOVQ (DX)(CX*1), SI + MOVL BX, 20(SP) + MOVQ $0x0000cf1bbcdcbf9b, R8 + MOVQ SI, R9 + MOVQ SI, R10 + SHRQ $0x08, R10 + SHLQ $0x10, R9 + IMULQ R8, R9 + SHRQ $0x33, R9 + SHLQ $0x10, R10 + IMULQ R8, R10 + SHRQ $0x33, R10 + MOVL 24(SP)(R9*4), BX + MOVL 24(SP)(R10*4), DI + MOVL CX, 24(SP)(R9*4) + LEAL 1(CX), R9 + MOVL R9, 24(SP)(R10*4) + MOVQ SI, R9 + SHRQ $0x10, R9 + SHLQ $0x10, R9 + IMULQ R8, R9 + SHRQ $0x33, R9 + MOVL CX, R8 + SUBL 16(SP), R8 + MOVL 1(DX)(R8*1), R10 + MOVQ SI, R8 + SHRQ $0x08, R8 + CMPL R8, R10 + JNE no_repeat_found_calcBlockSize + LEAL 1(CX), SI + MOVL 12(SP), BX + MOVL SI, DI + SUBL 16(SP), DI + JZ repeat_extend_back_end_calcBlockSize + +repeat_extend_back_loop_calcBlockSize: + CMPL SI, BX + JBE repeat_extend_back_end_calcBlockSize + MOVB -1(DX)(DI*1), R8 + MOVB -1(DX)(SI*1), R9 + CMPB R8, R9 + JNE repeat_extend_back_end_calcBlockSize + LEAL -1(SI), SI + DECL DI + JNZ repeat_extend_back_loop_calcBlockSize + +repeat_extend_back_end_calcBlockSize: + MOVL SI, BX + SUBL 12(SP), BX + LEAQ 5(AX)(BX*1), BX + CMPQ BX, (SP) + JB repeat_dst_size_check_calcBlockSize + MOVQ $0x00000000, ret+24(FP) + RET + +repeat_dst_size_check_calcBlockSize: + MOVL 12(SP), BX + CMPL BX, SI + JEQ emit_literal_done_repeat_emit_calcBlockSize + MOVL SI, DI + MOVL SI, 12(SP) + LEAQ (DX)(BX*1), R8 + SUBL BX, DI + LEAL -1(DI), BX + CMPL BX, $0x3c + JB one_byte_repeat_emit_calcBlockSize + CMPL BX, $0x00000100 + JB two_bytes_repeat_emit_calcBlockSize + CMPL BX, $0x00010000 + JB three_bytes_repeat_emit_calcBlockSize + CMPL BX, $0x01000000 + JB four_bytes_repeat_emit_calcBlockSize + ADDQ $0x05, AX + JMP memmove_long_repeat_emit_calcBlockSize + +four_bytes_repeat_emit_calcBlockSize: + ADDQ $0x04, AX + JMP memmove_long_repeat_emit_calcBlockSize + +three_bytes_repeat_emit_calcBlockSize: + ADDQ $0x03, AX + JMP memmove_long_repeat_emit_calcBlockSize + +two_bytes_repeat_emit_calcBlockSize: + ADDQ $0x02, AX + CMPL BX, $0x40 + JB memmove_repeat_emit_calcBlockSize + JMP memmove_long_repeat_emit_calcBlockSize + +one_byte_repeat_emit_calcBlockSize: + ADDQ $0x01, AX + +memmove_repeat_emit_calcBlockSize: + LEAQ (AX)(DI*1), AX + JMP emit_literal_done_repeat_emit_calcBlockSize + +memmove_long_repeat_emit_calcBlockSize: + LEAQ (AX)(DI*1), AX + +emit_literal_done_repeat_emit_calcBlockSize: + ADDL $0x05, CX + MOVL CX, BX + SUBL 16(SP), BX + MOVQ src_len+8(FP), DI + SUBL CX, DI + LEAQ (DX)(CX*1), R8 + LEAQ (DX)(BX*1), BX + + // matchLen + XORL R10, R10 + +matchlen_loopback_16_repeat_extend_calcBlockSize: + CMPL DI, $0x10 + JB matchlen_match8_repeat_extend_calcBlockSize + MOVQ (R8)(R10*1), R9 + MOVQ 8(R8)(R10*1), R11 + XORQ (BX)(R10*1), R9 + JNZ matchlen_bsf_8_repeat_extend_calcBlockSize + XORQ 8(BX)(R10*1), R11 + JNZ matchlen_bsf_16repeat_extend_calcBlockSize + LEAL -16(DI), DI + LEAL 16(R10), R10 + JMP matchlen_loopback_16_repeat_extend_calcBlockSize + +matchlen_bsf_16repeat_extend_calcBlockSize: +#ifdef GOAMD64_v3 + TZCNTQ R11, R11 + +#else + BSFQ R11, R11 + +#endif + SARQ $0x03, R11 + LEAL 8(R10)(R11*1), R10 + JMP repeat_extend_forward_end_calcBlockSize + +matchlen_match8_repeat_extend_calcBlockSize: + CMPL DI, $0x08 + JB matchlen_match4_repeat_extend_calcBlockSize + MOVQ (R8)(R10*1), R9 + XORQ (BX)(R10*1), R9 + JNZ matchlen_bsf_8_repeat_extend_calcBlockSize + LEAL -8(DI), DI + LEAL 8(R10), R10 + JMP matchlen_match4_repeat_extend_calcBlockSize + +matchlen_bsf_8_repeat_extend_calcBlockSize: +#ifdef GOAMD64_v3 + TZCNTQ R9, R9 + +#else + BSFQ R9, R9 + +#endif + SARQ $0x03, R9 + LEAL (R10)(R9*1), R10 + JMP repeat_extend_forward_end_calcBlockSize + +matchlen_match4_repeat_extend_calcBlockSize: + CMPL DI, $0x04 + JB matchlen_match2_repeat_extend_calcBlockSize + MOVL (R8)(R10*1), R9 + CMPL (BX)(R10*1), R9 + JNE matchlen_match2_repeat_extend_calcBlockSize + LEAL -4(DI), DI + LEAL 4(R10), R10 + +matchlen_match2_repeat_extend_calcBlockSize: + CMPL DI, $0x01 + JE matchlen_match1_repeat_extend_calcBlockSize + JB repeat_extend_forward_end_calcBlockSize + MOVW (R8)(R10*1), R9 + CMPW (BX)(R10*1), R9 + JNE matchlen_match1_repeat_extend_calcBlockSize + LEAL 2(R10), R10 + SUBL $0x02, DI + JZ repeat_extend_forward_end_calcBlockSize + +matchlen_match1_repeat_extend_calcBlockSize: + MOVB (R8)(R10*1), R9 + CMPB (BX)(R10*1), R9 + JNE repeat_extend_forward_end_calcBlockSize + LEAL 1(R10), R10 + +repeat_extend_forward_end_calcBlockSize: + ADDL R10, CX + MOVL CX, BX + SUBL SI, BX + MOVL 16(SP), SI + + // emitCopy + CMPL SI, $0x00010000 + JB two_byte_offset_repeat_as_copy_calcBlockSize + +four_bytes_loop_back_repeat_as_copy_calcBlockSize: + CMPL BX, $0x40 + JBE four_bytes_remain_repeat_as_copy_calcBlockSize + LEAL -64(BX), BX + ADDQ $0x05, AX + CMPL BX, $0x04 + JB four_bytes_remain_repeat_as_copy_calcBlockSize + JMP four_bytes_loop_back_repeat_as_copy_calcBlockSize + +four_bytes_remain_repeat_as_copy_calcBlockSize: + TESTL BX, BX + JZ repeat_end_emit_calcBlockSize + XORL BX, BX + ADDQ $0x05, AX + JMP repeat_end_emit_calcBlockSize + +two_byte_offset_repeat_as_copy_calcBlockSize: + CMPL BX, $0x40 + JBE two_byte_offset_short_repeat_as_copy_calcBlockSize + LEAL -60(BX), BX + ADDQ $0x03, AX + JMP two_byte_offset_repeat_as_copy_calcBlockSize + +two_byte_offset_short_repeat_as_copy_calcBlockSize: + MOVL BX, DI + SHLL $0x02, DI + CMPL BX, $0x0c + JAE emit_copy_three_repeat_as_copy_calcBlockSize + CMPL SI, $0x00000800 + JAE emit_copy_three_repeat_as_copy_calcBlockSize + ADDQ $0x02, AX + JMP repeat_end_emit_calcBlockSize + +emit_copy_three_repeat_as_copy_calcBlockSize: + ADDQ $0x03, AX + +repeat_end_emit_calcBlockSize: + MOVL CX, 12(SP) + JMP search_loop_calcBlockSize + +no_repeat_found_calcBlockSize: + CMPL (DX)(BX*1), SI + JEQ candidate_match_calcBlockSize + SHRQ $0x08, SI + MOVL 24(SP)(R9*4), BX + LEAL 2(CX), R8 + CMPL (DX)(DI*1), SI + JEQ candidate2_match_calcBlockSize + MOVL R8, 24(SP)(R9*4) + SHRQ $0x08, SI + CMPL (DX)(BX*1), SI + JEQ candidate3_match_calcBlockSize + MOVL 20(SP), CX + JMP search_loop_calcBlockSize + +candidate3_match_calcBlockSize: + ADDL $0x02, CX + JMP candidate_match_calcBlockSize + +candidate2_match_calcBlockSize: + MOVL R8, 24(SP)(R9*4) + INCL CX + MOVL DI, BX + +candidate_match_calcBlockSize: + MOVL 12(SP), SI + TESTL BX, BX + JZ match_extend_back_end_calcBlockSize + +match_extend_back_loop_calcBlockSize: + CMPL CX, SI + JBE match_extend_back_end_calcBlockSize + MOVB -1(DX)(BX*1), DI + MOVB -1(DX)(CX*1), R8 + CMPB DI, R8 + JNE match_extend_back_end_calcBlockSize + LEAL -1(CX), CX + DECL BX + JZ match_extend_back_end_calcBlockSize + JMP match_extend_back_loop_calcBlockSize + +match_extend_back_end_calcBlockSize: + MOVL CX, SI + SUBL 12(SP), SI + LEAQ 5(AX)(SI*1), SI + CMPQ SI, (SP) + JB match_dst_size_check_calcBlockSize + MOVQ $0x00000000, ret+24(FP) + RET + +match_dst_size_check_calcBlockSize: + MOVL CX, SI + MOVL 12(SP), DI + CMPL DI, SI + JEQ emit_literal_done_match_emit_calcBlockSize + MOVL SI, R8 + MOVL SI, 12(SP) + LEAQ (DX)(DI*1), SI + SUBL DI, R8 + LEAL -1(R8), SI + CMPL SI, $0x3c + JB one_byte_match_emit_calcBlockSize + CMPL SI, $0x00000100 + JB two_bytes_match_emit_calcBlockSize + CMPL SI, $0x00010000 + JB three_bytes_match_emit_calcBlockSize + CMPL SI, $0x01000000 + JB four_bytes_match_emit_calcBlockSize + ADDQ $0x05, AX + JMP memmove_long_match_emit_calcBlockSize + +four_bytes_match_emit_calcBlockSize: + ADDQ $0x04, AX + JMP memmove_long_match_emit_calcBlockSize + +three_bytes_match_emit_calcBlockSize: + ADDQ $0x03, AX + JMP memmove_long_match_emit_calcBlockSize + +two_bytes_match_emit_calcBlockSize: + ADDQ $0x02, AX + CMPL SI, $0x40 + JB memmove_match_emit_calcBlockSize + JMP memmove_long_match_emit_calcBlockSize + +one_byte_match_emit_calcBlockSize: + ADDQ $0x01, AX + +memmove_match_emit_calcBlockSize: + LEAQ (AX)(R8*1), AX + JMP emit_literal_done_match_emit_calcBlockSize + +memmove_long_match_emit_calcBlockSize: + LEAQ (AX)(R8*1), AX + +emit_literal_done_match_emit_calcBlockSize: +match_nolit_loop_calcBlockSize: + MOVL CX, SI + SUBL BX, SI + MOVL SI, 16(SP) + ADDL $0x04, CX + ADDL $0x04, BX + MOVQ src_len+8(FP), SI + SUBL CX, SI + LEAQ (DX)(CX*1), DI + LEAQ (DX)(BX*1), BX + + // matchLen + XORL R9, R9 + +matchlen_loopback_16_match_nolit_calcBlockSize: + CMPL SI, $0x10 + JB matchlen_match8_match_nolit_calcBlockSize + MOVQ (DI)(R9*1), R8 + MOVQ 8(DI)(R9*1), R10 + XORQ (BX)(R9*1), R8 + JNZ matchlen_bsf_8_match_nolit_calcBlockSize + XORQ 8(BX)(R9*1), R10 + JNZ matchlen_bsf_16match_nolit_calcBlockSize + LEAL -16(SI), SI + LEAL 16(R9), R9 + JMP matchlen_loopback_16_match_nolit_calcBlockSize + +matchlen_bsf_16match_nolit_calcBlockSize: +#ifdef GOAMD64_v3 + TZCNTQ R10, R10 + +#else + BSFQ R10, R10 + +#endif + SARQ $0x03, R10 + LEAL 8(R9)(R10*1), R9 + JMP match_nolit_end_calcBlockSize + +matchlen_match8_match_nolit_calcBlockSize: + CMPL SI, $0x08 + JB matchlen_match4_match_nolit_calcBlockSize + MOVQ (DI)(R9*1), R8 + XORQ (BX)(R9*1), R8 + JNZ matchlen_bsf_8_match_nolit_calcBlockSize + LEAL -8(SI), SI + LEAL 8(R9), R9 + JMP matchlen_match4_match_nolit_calcBlockSize + +matchlen_bsf_8_match_nolit_calcBlockSize: +#ifdef GOAMD64_v3 + TZCNTQ R8, R8 + +#else + BSFQ R8, R8 + +#endif + SARQ $0x03, R8 + LEAL (R9)(R8*1), R9 + JMP match_nolit_end_calcBlockSize + +matchlen_match4_match_nolit_calcBlockSize: + CMPL SI, $0x04 + JB matchlen_match2_match_nolit_calcBlockSize + MOVL (DI)(R9*1), R8 + CMPL (BX)(R9*1), R8 + JNE matchlen_match2_match_nolit_calcBlockSize + LEAL -4(SI), SI + LEAL 4(R9), R9 + +matchlen_match2_match_nolit_calcBlockSize: + CMPL SI, $0x01 + JE matchlen_match1_match_nolit_calcBlockSize + JB match_nolit_end_calcBlockSize + MOVW (DI)(R9*1), R8 + CMPW (BX)(R9*1), R8 + JNE matchlen_match1_match_nolit_calcBlockSize + LEAL 2(R9), R9 + SUBL $0x02, SI + JZ match_nolit_end_calcBlockSize + +matchlen_match1_match_nolit_calcBlockSize: + MOVB (DI)(R9*1), R8 + CMPB (BX)(R9*1), R8 + JNE match_nolit_end_calcBlockSize + LEAL 1(R9), R9 + +match_nolit_end_calcBlockSize: + ADDL R9, CX + MOVL 16(SP), BX + ADDL $0x04, R9 + MOVL CX, 12(SP) + + // emitCopy + CMPL BX, $0x00010000 + JB two_byte_offset_match_nolit_calcBlockSize + +four_bytes_loop_back_match_nolit_calcBlockSize: + CMPL R9, $0x40 + JBE four_bytes_remain_match_nolit_calcBlockSize + LEAL -64(R9), R9 + ADDQ $0x05, AX + CMPL R9, $0x04 + JB four_bytes_remain_match_nolit_calcBlockSize + JMP four_bytes_loop_back_match_nolit_calcBlockSize + +four_bytes_remain_match_nolit_calcBlockSize: + TESTL R9, R9 + JZ match_nolit_emitcopy_end_calcBlockSize + XORL BX, BX + ADDQ $0x05, AX + JMP match_nolit_emitcopy_end_calcBlockSize + +two_byte_offset_match_nolit_calcBlockSize: + CMPL R9, $0x40 + JBE two_byte_offset_short_match_nolit_calcBlockSize + LEAL -60(R9), R9 + ADDQ $0x03, AX + JMP two_byte_offset_match_nolit_calcBlockSize + +two_byte_offset_short_match_nolit_calcBlockSize: + MOVL R9, SI + SHLL $0x02, SI + CMPL R9, $0x0c + JAE emit_copy_three_match_nolit_calcBlockSize + CMPL BX, $0x00000800 + JAE emit_copy_three_match_nolit_calcBlockSize + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_calcBlockSize + +emit_copy_three_match_nolit_calcBlockSize: + ADDQ $0x03, AX + +match_nolit_emitcopy_end_calcBlockSize: + CMPL CX, 8(SP) + JAE emit_remainder_calcBlockSize + MOVQ -2(DX)(CX*1), SI + CMPQ AX, (SP) + JB match_nolit_dst_ok_calcBlockSize + MOVQ $0x00000000, ret+24(FP) + RET + +match_nolit_dst_ok_calcBlockSize: + MOVQ $0x0000cf1bbcdcbf9b, R8 + MOVQ SI, DI + SHRQ $0x10, SI + MOVQ SI, BX + SHLQ $0x10, DI + IMULQ R8, DI + SHRQ $0x33, DI + SHLQ $0x10, BX + IMULQ R8, BX + SHRQ $0x33, BX + LEAL -2(CX), R8 + LEAQ 24(SP)(BX*4), R9 + MOVL (R9), BX + MOVL R8, 24(SP)(DI*4) + MOVL CX, (R9) + CMPL (DX)(BX*1), SI + JEQ match_nolit_loop_calcBlockSize + INCL CX + JMP search_loop_calcBlockSize + +emit_remainder_calcBlockSize: + MOVQ src_len+8(FP), CX + SUBL 12(SP), CX + LEAQ 5(AX)(CX*1), CX + CMPQ CX, (SP) + JB emit_remainder_ok_calcBlockSize + MOVQ $0x00000000, ret+24(FP) + RET + +emit_remainder_ok_calcBlockSize: + MOVQ src_len+8(FP), CX + MOVL 12(SP), BX + CMPL BX, CX + JEQ emit_literal_done_emit_remainder_calcBlockSize + MOVL CX, SI + MOVL CX, 12(SP) + LEAQ (DX)(BX*1), CX + SUBL BX, SI + LEAL -1(SI), CX + CMPL CX, $0x3c + JB one_byte_emit_remainder_calcBlockSize + CMPL CX, $0x00000100 + JB two_bytes_emit_remainder_calcBlockSize + CMPL CX, $0x00010000 + JB three_bytes_emit_remainder_calcBlockSize + CMPL CX, $0x01000000 + JB four_bytes_emit_remainder_calcBlockSize + ADDQ $0x05, AX + JMP memmove_long_emit_remainder_calcBlockSize + +four_bytes_emit_remainder_calcBlockSize: + ADDQ $0x04, AX + JMP memmove_long_emit_remainder_calcBlockSize + +three_bytes_emit_remainder_calcBlockSize: + ADDQ $0x03, AX + JMP memmove_long_emit_remainder_calcBlockSize + +two_bytes_emit_remainder_calcBlockSize: + ADDQ $0x02, AX + CMPL CX, $0x40 + JB memmove_emit_remainder_calcBlockSize + JMP memmove_long_emit_remainder_calcBlockSize + +one_byte_emit_remainder_calcBlockSize: + ADDQ $0x01, AX + +memmove_emit_remainder_calcBlockSize: + LEAQ (AX)(SI*1), AX + JMP emit_literal_done_emit_remainder_calcBlockSize + +memmove_long_emit_remainder_calcBlockSize: + LEAQ (AX)(SI*1), AX + +emit_literal_done_emit_remainder_calcBlockSize: + MOVQ AX, ret+24(FP) + RET + +// func calcBlockSizeSmall(src []byte) int +// Requires: BMI, SSE2 +TEXT ·calcBlockSizeSmall(SB), $2072-32 + XORQ AX, AX + MOVQ $0x00000010, CX + LEAQ 24(SP), DX + PXOR X0, X0 + +zero_loop_calcBlockSizeSmall: + MOVOU X0, (DX) + MOVOU X0, 16(DX) + MOVOU X0, 32(DX) + MOVOU X0, 48(DX) + MOVOU X0, 64(DX) + MOVOU X0, 80(DX) + MOVOU X0, 96(DX) + MOVOU X0, 112(DX) + ADDQ $0x80, DX + DECQ CX + JNZ zero_loop_calcBlockSizeSmall + MOVL $0x00000000, 12(SP) + MOVQ src_len+8(FP), CX + LEAQ -9(CX), DX + LEAQ -8(CX), BX + MOVL BX, 8(SP) + SHRQ $0x05, CX + SUBL CX, DX + LEAQ (AX)(DX*1), DX + MOVQ DX, (SP) + MOVL $0x00000001, CX + MOVL CX, 16(SP) + MOVQ src_base+0(FP), DX + +search_loop_calcBlockSizeSmall: + MOVL CX, BX + SUBL 12(SP), BX + SHRL $0x04, BX + LEAL 4(CX)(BX*1), BX + CMPL BX, 8(SP) + JAE emit_remainder_calcBlockSizeSmall + MOVQ (DX)(CX*1), SI + MOVL BX, 20(SP) + MOVQ $0x9e3779b1, R8 + MOVQ SI, R9 + MOVQ SI, R10 + SHRQ $0x08, R10 + SHLQ $0x20, R9 + IMULQ R8, R9 + SHRQ $0x37, R9 + SHLQ $0x20, R10 + IMULQ R8, R10 + SHRQ $0x37, R10 + MOVL 24(SP)(R9*4), BX + MOVL 24(SP)(R10*4), DI + MOVL CX, 24(SP)(R9*4) + LEAL 1(CX), R9 + MOVL R9, 24(SP)(R10*4) + MOVQ SI, R9 + SHRQ $0x10, R9 + SHLQ $0x20, R9 + IMULQ R8, R9 + SHRQ $0x37, R9 + MOVL CX, R8 + SUBL 16(SP), R8 + MOVL 1(DX)(R8*1), R10 + MOVQ SI, R8 + SHRQ $0x08, R8 + CMPL R8, R10 + JNE no_repeat_found_calcBlockSizeSmall + LEAL 1(CX), SI + MOVL 12(SP), BX + MOVL SI, DI + SUBL 16(SP), DI + JZ repeat_extend_back_end_calcBlockSizeSmall + +repeat_extend_back_loop_calcBlockSizeSmall: + CMPL SI, BX + JBE repeat_extend_back_end_calcBlockSizeSmall + MOVB -1(DX)(DI*1), R8 + MOVB -1(DX)(SI*1), R9 + CMPB R8, R9 + JNE repeat_extend_back_end_calcBlockSizeSmall + LEAL -1(SI), SI + DECL DI + JNZ repeat_extend_back_loop_calcBlockSizeSmall + +repeat_extend_back_end_calcBlockSizeSmall: + MOVL SI, BX + SUBL 12(SP), BX + LEAQ 3(AX)(BX*1), BX + CMPQ BX, (SP) + JB repeat_dst_size_check_calcBlockSizeSmall + MOVQ $0x00000000, ret+24(FP) + RET + +repeat_dst_size_check_calcBlockSizeSmall: + MOVL 12(SP), BX + CMPL BX, SI + JEQ emit_literal_done_repeat_emit_calcBlockSizeSmall + MOVL SI, DI + MOVL SI, 12(SP) + LEAQ (DX)(BX*1), R8 + SUBL BX, DI + LEAL -1(DI), BX + CMPL BX, $0x3c + JB one_byte_repeat_emit_calcBlockSizeSmall + CMPL BX, $0x00000100 + JB two_bytes_repeat_emit_calcBlockSizeSmall + JB three_bytes_repeat_emit_calcBlockSizeSmall + +three_bytes_repeat_emit_calcBlockSizeSmall: + ADDQ $0x03, AX + JMP memmove_long_repeat_emit_calcBlockSizeSmall + +two_bytes_repeat_emit_calcBlockSizeSmall: + ADDQ $0x02, AX + CMPL BX, $0x40 + JB memmove_repeat_emit_calcBlockSizeSmall + JMP memmove_long_repeat_emit_calcBlockSizeSmall + +one_byte_repeat_emit_calcBlockSizeSmall: + ADDQ $0x01, AX + +memmove_repeat_emit_calcBlockSizeSmall: + LEAQ (AX)(DI*1), AX + JMP emit_literal_done_repeat_emit_calcBlockSizeSmall + +memmove_long_repeat_emit_calcBlockSizeSmall: + LEAQ (AX)(DI*1), AX + +emit_literal_done_repeat_emit_calcBlockSizeSmall: + ADDL $0x05, CX + MOVL CX, BX + SUBL 16(SP), BX + MOVQ src_len+8(FP), DI + SUBL CX, DI + LEAQ (DX)(CX*1), R8 + LEAQ (DX)(BX*1), BX + + // matchLen + XORL R10, R10 + +matchlen_loopback_16_repeat_extend_calcBlockSizeSmall: + CMPL DI, $0x10 + JB matchlen_match8_repeat_extend_calcBlockSizeSmall + MOVQ (R8)(R10*1), R9 + MOVQ 8(R8)(R10*1), R11 + XORQ (BX)(R10*1), R9 + JNZ matchlen_bsf_8_repeat_extend_calcBlockSizeSmall + XORQ 8(BX)(R10*1), R11 + JNZ matchlen_bsf_16repeat_extend_calcBlockSizeSmall + LEAL -16(DI), DI + LEAL 16(R10), R10 + JMP matchlen_loopback_16_repeat_extend_calcBlockSizeSmall + +matchlen_bsf_16repeat_extend_calcBlockSizeSmall: +#ifdef GOAMD64_v3 + TZCNTQ R11, R11 + +#else + BSFQ R11, R11 + +#endif + SARQ $0x03, R11 + LEAL 8(R10)(R11*1), R10 + JMP repeat_extend_forward_end_calcBlockSizeSmall + +matchlen_match8_repeat_extend_calcBlockSizeSmall: + CMPL DI, $0x08 + JB matchlen_match4_repeat_extend_calcBlockSizeSmall + MOVQ (R8)(R10*1), R9 + XORQ (BX)(R10*1), R9 + JNZ matchlen_bsf_8_repeat_extend_calcBlockSizeSmall + LEAL -8(DI), DI + LEAL 8(R10), R10 + JMP matchlen_match4_repeat_extend_calcBlockSizeSmall + +matchlen_bsf_8_repeat_extend_calcBlockSizeSmall: +#ifdef GOAMD64_v3 + TZCNTQ R9, R9 + +#else + BSFQ R9, R9 + +#endif + SARQ $0x03, R9 + LEAL (R10)(R9*1), R10 + JMP repeat_extend_forward_end_calcBlockSizeSmall + +matchlen_match4_repeat_extend_calcBlockSizeSmall: + CMPL DI, $0x04 + JB matchlen_match2_repeat_extend_calcBlockSizeSmall + MOVL (R8)(R10*1), R9 + CMPL (BX)(R10*1), R9 + JNE matchlen_match2_repeat_extend_calcBlockSizeSmall + LEAL -4(DI), DI + LEAL 4(R10), R10 + +matchlen_match2_repeat_extend_calcBlockSizeSmall: + CMPL DI, $0x01 + JE matchlen_match1_repeat_extend_calcBlockSizeSmall + JB repeat_extend_forward_end_calcBlockSizeSmall + MOVW (R8)(R10*1), R9 + CMPW (BX)(R10*1), R9 + JNE matchlen_match1_repeat_extend_calcBlockSizeSmall + LEAL 2(R10), R10 + SUBL $0x02, DI + JZ repeat_extend_forward_end_calcBlockSizeSmall + +matchlen_match1_repeat_extend_calcBlockSizeSmall: + MOVB (R8)(R10*1), R9 + CMPB (BX)(R10*1), R9 + JNE repeat_extend_forward_end_calcBlockSizeSmall + LEAL 1(R10), R10 + +repeat_extend_forward_end_calcBlockSizeSmall: + ADDL R10, CX + MOVL CX, BX + SUBL SI, BX + MOVL 16(SP), SI + + // emitCopy +two_byte_offset_repeat_as_copy_calcBlockSizeSmall: + CMPL BX, $0x40 + JBE two_byte_offset_short_repeat_as_copy_calcBlockSizeSmall + LEAL -60(BX), BX + ADDQ $0x03, AX + JMP two_byte_offset_repeat_as_copy_calcBlockSizeSmall + +two_byte_offset_short_repeat_as_copy_calcBlockSizeSmall: + MOVL BX, SI + SHLL $0x02, SI + CMPL BX, $0x0c + JAE emit_copy_three_repeat_as_copy_calcBlockSizeSmall + ADDQ $0x02, AX + JMP repeat_end_emit_calcBlockSizeSmall + +emit_copy_three_repeat_as_copy_calcBlockSizeSmall: + ADDQ $0x03, AX + +repeat_end_emit_calcBlockSizeSmall: + MOVL CX, 12(SP) + JMP search_loop_calcBlockSizeSmall + +no_repeat_found_calcBlockSizeSmall: + CMPL (DX)(BX*1), SI + JEQ candidate_match_calcBlockSizeSmall + SHRQ $0x08, SI + MOVL 24(SP)(R9*4), BX + LEAL 2(CX), R8 + CMPL (DX)(DI*1), SI + JEQ candidate2_match_calcBlockSizeSmall + MOVL R8, 24(SP)(R9*4) + SHRQ $0x08, SI + CMPL (DX)(BX*1), SI + JEQ candidate3_match_calcBlockSizeSmall + MOVL 20(SP), CX + JMP search_loop_calcBlockSizeSmall + +candidate3_match_calcBlockSizeSmall: + ADDL $0x02, CX + JMP candidate_match_calcBlockSizeSmall + +candidate2_match_calcBlockSizeSmall: + MOVL R8, 24(SP)(R9*4) + INCL CX + MOVL DI, BX + +candidate_match_calcBlockSizeSmall: + MOVL 12(SP), SI + TESTL BX, BX + JZ match_extend_back_end_calcBlockSizeSmall + +match_extend_back_loop_calcBlockSizeSmall: + CMPL CX, SI + JBE match_extend_back_end_calcBlockSizeSmall + MOVB -1(DX)(BX*1), DI + MOVB -1(DX)(CX*1), R8 + CMPB DI, R8 + JNE match_extend_back_end_calcBlockSizeSmall + LEAL -1(CX), CX + DECL BX + JZ match_extend_back_end_calcBlockSizeSmall + JMP match_extend_back_loop_calcBlockSizeSmall + +match_extend_back_end_calcBlockSizeSmall: + MOVL CX, SI + SUBL 12(SP), SI + LEAQ 3(AX)(SI*1), SI + CMPQ SI, (SP) + JB match_dst_size_check_calcBlockSizeSmall + MOVQ $0x00000000, ret+24(FP) + RET + +match_dst_size_check_calcBlockSizeSmall: + MOVL CX, SI + MOVL 12(SP), DI + CMPL DI, SI + JEQ emit_literal_done_match_emit_calcBlockSizeSmall + MOVL SI, R8 + MOVL SI, 12(SP) + LEAQ (DX)(DI*1), SI + SUBL DI, R8 + LEAL -1(R8), SI + CMPL SI, $0x3c + JB one_byte_match_emit_calcBlockSizeSmall + CMPL SI, $0x00000100 + JB two_bytes_match_emit_calcBlockSizeSmall + JB three_bytes_match_emit_calcBlockSizeSmall + +three_bytes_match_emit_calcBlockSizeSmall: + ADDQ $0x03, AX + JMP memmove_long_match_emit_calcBlockSizeSmall + +two_bytes_match_emit_calcBlockSizeSmall: + ADDQ $0x02, AX + CMPL SI, $0x40 + JB memmove_match_emit_calcBlockSizeSmall + JMP memmove_long_match_emit_calcBlockSizeSmall + +one_byte_match_emit_calcBlockSizeSmall: + ADDQ $0x01, AX + +memmove_match_emit_calcBlockSizeSmall: + LEAQ (AX)(R8*1), AX + JMP emit_literal_done_match_emit_calcBlockSizeSmall + +memmove_long_match_emit_calcBlockSizeSmall: + LEAQ (AX)(R8*1), AX + +emit_literal_done_match_emit_calcBlockSizeSmall: +match_nolit_loop_calcBlockSizeSmall: + MOVL CX, SI + SUBL BX, SI + MOVL SI, 16(SP) + ADDL $0x04, CX + ADDL $0x04, BX + MOVQ src_len+8(FP), SI + SUBL CX, SI + LEAQ (DX)(CX*1), DI + LEAQ (DX)(BX*1), BX + + // matchLen + XORL R9, R9 + +matchlen_loopback_16_match_nolit_calcBlockSizeSmall: + CMPL SI, $0x10 + JB matchlen_match8_match_nolit_calcBlockSizeSmall + MOVQ (DI)(R9*1), R8 + MOVQ 8(DI)(R9*1), R10 + XORQ (BX)(R9*1), R8 + JNZ matchlen_bsf_8_match_nolit_calcBlockSizeSmall + XORQ 8(BX)(R9*1), R10 + JNZ matchlen_bsf_16match_nolit_calcBlockSizeSmall + LEAL -16(SI), SI + LEAL 16(R9), R9 + JMP matchlen_loopback_16_match_nolit_calcBlockSizeSmall + +matchlen_bsf_16match_nolit_calcBlockSizeSmall: +#ifdef GOAMD64_v3 + TZCNTQ R10, R10 + +#else + BSFQ R10, R10 + +#endif + SARQ $0x03, R10 + LEAL 8(R9)(R10*1), R9 + JMP match_nolit_end_calcBlockSizeSmall + +matchlen_match8_match_nolit_calcBlockSizeSmall: + CMPL SI, $0x08 + JB matchlen_match4_match_nolit_calcBlockSizeSmall + MOVQ (DI)(R9*1), R8 + XORQ (BX)(R9*1), R8 + JNZ matchlen_bsf_8_match_nolit_calcBlockSizeSmall + LEAL -8(SI), SI + LEAL 8(R9), R9 + JMP matchlen_match4_match_nolit_calcBlockSizeSmall + +matchlen_bsf_8_match_nolit_calcBlockSizeSmall: +#ifdef GOAMD64_v3 + TZCNTQ R8, R8 + +#else + BSFQ R8, R8 + +#endif + SARQ $0x03, R8 + LEAL (R9)(R8*1), R9 + JMP match_nolit_end_calcBlockSizeSmall + +matchlen_match4_match_nolit_calcBlockSizeSmall: + CMPL SI, $0x04 + JB matchlen_match2_match_nolit_calcBlockSizeSmall + MOVL (DI)(R9*1), R8 + CMPL (BX)(R9*1), R8 + JNE matchlen_match2_match_nolit_calcBlockSizeSmall + LEAL -4(SI), SI + LEAL 4(R9), R9 + +matchlen_match2_match_nolit_calcBlockSizeSmall: + CMPL SI, $0x01 + JE matchlen_match1_match_nolit_calcBlockSizeSmall + JB match_nolit_end_calcBlockSizeSmall + MOVW (DI)(R9*1), R8 + CMPW (BX)(R9*1), R8 + JNE matchlen_match1_match_nolit_calcBlockSizeSmall + LEAL 2(R9), R9 + SUBL $0x02, SI + JZ match_nolit_end_calcBlockSizeSmall + +matchlen_match1_match_nolit_calcBlockSizeSmall: + MOVB (DI)(R9*1), R8 + CMPB (BX)(R9*1), R8 + JNE match_nolit_end_calcBlockSizeSmall + LEAL 1(R9), R9 + +match_nolit_end_calcBlockSizeSmall: + ADDL R9, CX + MOVL 16(SP), BX + ADDL $0x04, R9 + MOVL CX, 12(SP) + + // emitCopy +two_byte_offset_match_nolit_calcBlockSizeSmall: + CMPL R9, $0x40 + JBE two_byte_offset_short_match_nolit_calcBlockSizeSmall + LEAL -60(R9), R9 + ADDQ $0x03, AX + JMP two_byte_offset_match_nolit_calcBlockSizeSmall + +two_byte_offset_short_match_nolit_calcBlockSizeSmall: + MOVL R9, BX + SHLL $0x02, BX + CMPL R9, $0x0c + JAE emit_copy_three_match_nolit_calcBlockSizeSmall + ADDQ $0x02, AX + JMP match_nolit_emitcopy_end_calcBlockSizeSmall + +emit_copy_three_match_nolit_calcBlockSizeSmall: + ADDQ $0x03, AX + +match_nolit_emitcopy_end_calcBlockSizeSmall: + CMPL CX, 8(SP) + JAE emit_remainder_calcBlockSizeSmall + MOVQ -2(DX)(CX*1), SI + CMPQ AX, (SP) + JB match_nolit_dst_ok_calcBlockSizeSmall + MOVQ $0x00000000, ret+24(FP) + RET + +match_nolit_dst_ok_calcBlockSizeSmall: + MOVQ $0x9e3779b1, R8 + MOVQ SI, DI + SHRQ $0x10, SI + MOVQ SI, BX + SHLQ $0x20, DI + IMULQ R8, DI + SHRQ $0x37, DI + SHLQ $0x20, BX + IMULQ R8, BX + SHRQ $0x37, BX + LEAL -2(CX), R8 + LEAQ 24(SP)(BX*4), R9 + MOVL (R9), BX + MOVL R8, 24(SP)(DI*4) + MOVL CX, (R9) + CMPL (DX)(BX*1), SI + JEQ match_nolit_loop_calcBlockSizeSmall + INCL CX + JMP search_loop_calcBlockSizeSmall + +emit_remainder_calcBlockSizeSmall: + MOVQ src_len+8(FP), CX + SUBL 12(SP), CX + LEAQ 3(AX)(CX*1), CX + CMPQ CX, (SP) + JB emit_remainder_ok_calcBlockSizeSmall + MOVQ $0x00000000, ret+24(FP) + RET + +emit_remainder_ok_calcBlockSizeSmall: + MOVQ src_len+8(FP), CX + MOVL 12(SP), BX + CMPL BX, CX + JEQ emit_literal_done_emit_remainder_calcBlockSizeSmall + MOVL CX, SI + MOVL CX, 12(SP) + LEAQ (DX)(BX*1), CX + SUBL BX, SI + LEAL -1(SI), CX + CMPL CX, $0x3c + JB one_byte_emit_remainder_calcBlockSizeSmall + CMPL CX, $0x00000100 + JB two_bytes_emit_remainder_calcBlockSizeSmall + JB three_bytes_emit_remainder_calcBlockSizeSmall + +three_bytes_emit_remainder_calcBlockSizeSmall: + ADDQ $0x03, AX + JMP memmove_long_emit_remainder_calcBlockSizeSmall + +two_bytes_emit_remainder_calcBlockSizeSmall: + ADDQ $0x02, AX + CMPL CX, $0x40 + JB memmove_emit_remainder_calcBlockSizeSmall + JMP memmove_long_emit_remainder_calcBlockSizeSmall + +one_byte_emit_remainder_calcBlockSizeSmall: + ADDQ $0x01, AX + +memmove_emit_remainder_calcBlockSizeSmall: + LEAQ (AX)(SI*1), AX + JMP emit_literal_done_emit_remainder_calcBlockSizeSmall + +memmove_long_emit_remainder_calcBlockSizeSmall: + LEAQ (AX)(SI*1), AX + +emit_literal_done_emit_remainder_calcBlockSizeSmall: + MOVQ AX, ret+24(FP) + RET + +// func emitLiteral(dst []byte, lit []byte) int +// Requires: SSE2 +TEXT ·emitLiteral(SB), NOSPLIT, $0-56 + MOVQ lit_len+32(FP), DX + MOVQ dst_base+0(FP), AX + MOVQ lit_base+24(FP), CX + TESTQ DX, DX + JZ emit_literal_end_standalone_skip + MOVL DX, BX + LEAL -1(DX), SI + CMPL SI, $0x3c + JB one_byte_standalone + CMPL SI, $0x00000100 + JB two_bytes_standalone + CMPL SI, $0x00010000 + JB three_bytes_standalone + CMPL SI, $0x01000000 + JB four_bytes_standalone + MOVB $0xfc, (AX) + MOVL SI, 1(AX) + ADDQ $0x05, BX + ADDQ $0x05, AX + JMP memmove_long_standalone + +four_bytes_standalone: + MOVL SI, DI + SHRL $0x10, DI + MOVB $0xf8, (AX) + MOVW SI, 1(AX) + MOVB DI, 3(AX) + ADDQ $0x04, BX + ADDQ $0x04, AX + JMP memmove_long_standalone + +three_bytes_standalone: + MOVB $0xf4, (AX) + MOVW SI, 1(AX) + ADDQ $0x03, BX + ADDQ $0x03, AX + JMP memmove_long_standalone + +two_bytes_standalone: + MOVB $0xf0, (AX) + MOVB SI, 1(AX) + ADDQ $0x02, BX + ADDQ $0x02, AX + CMPL SI, $0x40 + JB memmove_standalone + JMP memmove_long_standalone + +one_byte_standalone: + SHLB $0x02, SI + MOVB SI, (AX) + ADDQ $0x01, BX + ADDQ $0x01, AX + +memmove_standalone: + // genMemMoveShort + CMPQ DX, $0x03 + JB emit_lit_memmove_standalone_memmove_move_1or2 + JE emit_lit_memmove_standalone_memmove_move_3 + CMPQ DX, $0x08 + JB emit_lit_memmove_standalone_memmove_move_4through7 + CMPQ DX, $0x10 + JBE emit_lit_memmove_standalone_memmove_move_8through16 + CMPQ DX, $0x20 + JBE emit_lit_memmove_standalone_memmove_move_17through32 + JMP emit_lit_memmove_standalone_memmove_move_33through64 + +emit_lit_memmove_standalone_memmove_move_1or2: + MOVB (CX), SI + MOVB -1(CX)(DX*1), CL + MOVB SI, (AX) + MOVB CL, -1(AX)(DX*1) + JMP emit_literal_end_standalone + +emit_lit_memmove_standalone_memmove_move_3: + MOVW (CX), SI + MOVB 2(CX), CL + MOVW SI, (AX) + MOVB CL, 2(AX) + JMP emit_literal_end_standalone + +emit_lit_memmove_standalone_memmove_move_4through7: + MOVL (CX), SI + MOVL -4(CX)(DX*1), CX + MOVL SI, (AX) + MOVL CX, -4(AX)(DX*1) + JMP emit_literal_end_standalone + +emit_lit_memmove_standalone_memmove_move_8through16: + MOVQ (CX), SI + MOVQ -8(CX)(DX*1), CX + MOVQ SI, (AX) + MOVQ CX, -8(AX)(DX*1) + JMP emit_literal_end_standalone + +emit_lit_memmove_standalone_memmove_move_17through32: + MOVOU (CX), X0 + MOVOU -16(CX)(DX*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(DX*1) + JMP emit_literal_end_standalone + +emit_lit_memmove_standalone_memmove_move_33through64: + MOVOU (CX), X0 + MOVOU 16(CX), X1 + MOVOU -32(CX)(DX*1), X2 + MOVOU -16(CX)(DX*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(DX*1) + MOVOU X3, -16(AX)(DX*1) + JMP emit_literal_end_standalone + JMP emit_literal_end_standalone + +memmove_long_standalone: + // genMemMoveLong + MOVOU (CX), X0 + MOVOU 16(CX), X1 + MOVOU -32(CX)(DX*1), X2 + MOVOU -16(CX)(DX*1), X3 + MOVQ DX, DI + SHRQ $0x05, DI + MOVQ AX, SI + ANDL $0x0000001f, SI + MOVQ $0x00000040, R8 + SUBQ SI, R8 + DECQ DI + JA emit_lit_memmove_long_standalonelarge_forward_sse_loop_32 + LEAQ -32(CX)(R8*1), SI + LEAQ -32(AX)(R8*1), R9 + +emit_lit_memmove_long_standalonelarge_big_loop_back: + MOVOU (SI), X4 + MOVOU 16(SI), X5 + MOVOA X4, (R9) + MOVOA X5, 16(R9) + ADDQ $0x20, R9 + ADDQ $0x20, SI + ADDQ $0x20, R8 + DECQ DI + JNA emit_lit_memmove_long_standalonelarge_big_loop_back + +emit_lit_memmove_long_standalonelarge_forward_sse_loop_32: + MOVOU -32(CX)(R8*1), X4 + MOVOU -16(CX)(R8*1), X5 + MOVOA X4, -32(AX)(R8*1) + MOVOA X5, -16(AX)(R8*1) + ADDQ $0x20, R8 + CMPQ DX, R8 + JAE emit_lit_memmove_long_standalonelarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(DX*1) + MOVOU X3, -16(AX)(DX*1) + JMP emit_literal_end_standalone + JMP emit_literal_end_standalone + +emit_literal_end_standalone_skip: + XORQ BX, BX + +emit_literal_end_standalone: + MOVQ BX, ret+48(FP) + RET + +// func emitRepeat(dst []byte, offset int, length int) int +TEXT ·emitRepeat(SB), NOSPLIT, $0-48 + XORQ BX, BX + MOVQ dst_base+0(FP), AX + MOVQ offset+24(FP), CX + MOVQ length+32(FP), DX + + // emitRepeat +emit_repeat_again_standalone: + MOVL DX, SI + LEAL -4(DX), DX + CMPL SI, $0x08 + JBE repeat_two_standalone + CMPL SI, $0x0c + JAE cant_repeat_two_offset_standalone + CMPL CX, $0x00000800 + JB repeat_two_offset_standalone + +cant_repeat_two_offset_standalone: + CMPL DX, $0x00000104 + JB repeat_three_standalone + CMPL DX, $0x00010100 + JB repeat_four_standalone + CMPL DX, $0x0100ffff + JB repeat_five_standalone + LEAL -16842747(DX), DX + MOVL $0xfffb001d, (AX) + MOVB $0xff, 4(AX) + ADDQ $0x05, AX + ADDQ $0x05, BX + JMP emit_repeat_again_standalone + +repeat_five_standalone: + LEAL -65536(DX), DX + MOVL DX, CX + MOVW $0x001d, (AX) + MOVW DX, 2(AX) + SARL $0x10, CX + MOVB CL, 4(AX) + ADDQ $0x05, BX + ADDQ $0x05, AX + JMP gen_emit_repeat_end + +repeat_four_standalone: + LEAL -256(DX), DX + MOVW $0x0019, (AX) + MOVW DX, 2(AX) + ADDQ $0x04, BX + ADDQ $0x04, AX + JMP gen_emit_repeat_end + +repeat_three_standalone: + LEAL -4(DX), DX + MOVW $0x0015, (AX) + MOVB DL, 2(AX) + ADDQ $0x03, BX + ADDQ $0x03, AX + JMP gen_emit_repeat_end + +repeat_two_standalone: + SHLL $0x02, DX + ORL $0x01, DX + MOVW DX, (AX) + ADDQ $0x02, BX + ADDQ $0x02, AX + JMP gen_emit_repeat_end + +repeat_two_offset_standalone: + XORQ SI, SI + LEAL 1(SI)(DX*4), DX + MOVB CL, 1(AX) + SARL $0x08, CX + SHLL $0x05, CX + ORL CX, DX + MOVB DL, (AX) + ADDQ $0x02, BX + ADDQ $0x02, AX + +gen_emit_repeat_end: + MOVQ BX, ret+40(FP) + RET + +// func emitCopy(dst []byte, offset int, length int) int +TEXT ·emitCopy(SB), NOSPLIT, $0-48 + XORQ BX, BX + MOVQ dst_base+0(FP), AX + MOVQ offset+24(FP), CX + MOVQ length+32(FP), DX + + // emitCopy + CMPL CX, $0x00010000 + JB two_byte_offset_standalone + CMPL DX, $0x40 + JBE four_bytes_remain_standalone + MOVB $0xff, (AX) + MOVL CX, 1(AX) + LEAL -64(DX), DX + ADDQ $0x05, BX + ADDQ $0x05, AX + CMPL DX, $0x04 + JB four_bytes_remain_standalone + + // emitRepeat +emit_repeat_again_standalone_emit_copy: + MOVL DX, SI + LEAL -4(DX), DX + CMPL SI, $0x08 + JBE repeat_two_standalone_emit_copy + CMPL SI, $0x0c + JAE cant_repeat_two_offset_standalone_emit_copy + CMPL CX, $0x00000800 + JB repeat_two_offset_standalone_emit_copy + +cant_repeat_two_offset_standalone_emit_copy: + CMPL DX, $0x00000104 + JB repeat_three_standalone_emit_copy + CMPL DX, $0x00010100 + JB repeat_four_standalone_emit_copy + CMPL DX, $0x0100ffff + JB repeat_five_standalone_emit_copy + LEAL -16842747(DX), DX + MOVL $0xfffb001d, (AX) + MOVB $0xff, 4(AX) + ADDQ $0x05, AX + ADDQ $0x05, BX + JMP emit_repeat_again_standalone_emit_copy + +repeat_five_standalone_emit_copy: + LEAL -65536(DX), DX + MOVL DX, CX + MOVW $0x001d, (AX) + MOVW DX, 2(AX) + SARL $0x10, CX + MOVB CL, 4(AX) + ADDQ $0x05, BX + ADDQ $0x05, AX + JMP gen_emit_copy_end + +repeat_four_standalone_emit_copy: + LEAL -256(DX), DX + MOVW $0x0019, (AX) + MOVW DX, 2(AX) + ADDQ $0x04, BX + ADDQ $0x04, AX + JMP gen_emit_copy_end + +repeat_three_standalone_emit_copy: + LEAL -4(DX), DX + MOVW $0x0015, (AX) + MOVB DL, 2(AX) + ADDQ $0x03, BX + ADDQ $0x03, AX + JMP gen_emit_copy_end + +repeat_two_standalone_emit_copy: + SHLL $0x02, DX + ORL $0x01, DX + MOVW DX, (AX) + ADDQ $0x02, BX + ADDQ $0x02, AX + JMP gen_emit_copy_end + +repeat_two_offset_standalone_emit_copy: + XORQ SI, SI + LEAL 1(SI)(DX*4), DX + MOVB CL, 1(AX) + SARL $0x08, CX + SHLL $0x05, CX + ORL CX, DX + MOVB DL, (AX) + ADDQ $0x02, BX + ADDQ $0x02, AX + JMP gen_emit_copy_end + +four_bytes_remain_standalone: + TESTL DX, DX + JZ gen_emit_copy_end + XORL SI, SI + LEAL -1(SI)(DX*4), DX + MOVB DL, (AX) + MOVL CX, 1(AX) + ADDQ $0x05, BX + ADDQ $0x05, AX + JMP gen_emit_copy_end + +two_byte_offset_standalone: + CMPL DX, $0x40 + JBE two_byte_offset_short_standalone + CMPL CX, $0x00000800 + JAE long_offset_short_standalone + MOVL $0x00000001, SI + LEAL 16(SI), SI + MOVB CL, 1(AX) + MOVL CX, DI + SHRL $0x08, DI + SHLL $0x05, DI + ORL DI, SI + MOVB SI, (AX) + ADDQ $0x02, BX + ADDQ $0x02, AX + SUBL $0x08, DX + + // emitRepeat + LEAL -4(DX), DX + JMP cant_repeat_two_offset_standalone_emit_copy_short_2b + +emit_repeat_again_standalone_emit_copy_short_2b: + MOVL DX, SI + LEAL -4(DX), DX + CMPL SI, $0x08 + JBE repeat_two_standalone_emit_copy_short_2b + CMPL SI, $0x0c + JAE cant_repeat_two_offset_standalone_emit_copy_short_2b + CMPL CX, $0x00000800 + JB repeat_two_offset_standalone_emit_copy_short_2b + +cant_repeat_two_offset_standalone_emit_copy_short_2b: + CMPL DX, $0x00000104 + JB repeat_three_standalone_emit_copy_short_2b + CMPL DX, $0x00010100 + JB repeat_four_standalone_emit_copy_short_2b + CMPL DX, $0x0100ffff + JB repeat_five_standalone_emit_copy_short_2b + LEAL -16842747(DX), DX + MOVL $0xfffb001d, (AX) + MOVB $0xff, 4(AX) + ADDQ $0x05, AX + ADDQ $0x05, BX + JMP emit_repeat_again_standalone_emit_copy_short_2b + +repeat_five_standalone_emit_copy_short_2b: + LEAL -65536(DX), DX + MOVL DX, CX + MOVW $0x001d, (AX) + MOVW DX, 2(AX) + SARL $0x10, CX + MOVB CL, 4(AX) + ADDQ $0x05, BX + ADDQ $0x05, AX + JMP gen_emit_copy_end + +repeat_four_standalone_emit_copy_short_2b: + LEAL -256(DX), DX + MOVW $0x0019, (AX) + MOVW DX, 2(AX) + ADDQ $0x04, BX + ADDQ $0x04, AX + JMP gen_emit_copy_end + +repeat_three_standalone_emit_copy_short_2b: + LEAL -4(DX), DX + MOVW $0x0015, (AX) + MOVB DL, 2(AX) + ADDQ $0x03, BX + ADDQ $0x03, AX + JMP gen_emit_copy_end + +repeat_two_standalone_emit_copy_short_2b: + SHLL $0x02, DX + ORL $0x01, DX + MOVW DX, (AX) + ADDQ $0x02, BX + ADDQ $0x02, AX + JMP gen_emit_copy_end + +repeat_two_offset_standalone_emit_copy_short_2b: + XORQ SI, SI + LEAL 1(SI)(DX*4), DX + MOVB CL, 1(AX) + SARL $0x08, CX + SHLL $0x05, CX + ORL CX, DX + MOVB DL, (AX) + ADDQ $0x02, BX + ADDQ $0x02, AX + JMP gen_emit_copy_end + +long_offset_short_standalone: + MOVB $0xee, (AX) + MOVW CX, 1(AX) + LEAL -60(DX), DX + ADDQ $0x03, AX + ADDQ $0x03, BX + + // emitRepeat +emit_repeat_again_standalone_emit_copy_short: + MOVL DX, SI + LEAL -4(DX), DX + CMPL SI, $0x08 + JBE repeat_two_standalone_emit_copy_short + CMPL SI, $0x0c + JAE cant_repeat_two_offset_standalone_emit_copy_short + CMPL CX, $0x00000800 + JB repeat_two_offset_standalone_emit_copy_short + +cant_repeat_two_offset_standalone_emit_copy_short: + CMPL DX, $0x00000104 + JB repeat_three_standalone_emit_copy_short + CMPL DX, $0x00010100 + JB repeat_four_standalone_emit_copy_short + CMPL DX, $0x0100ffff + JB repeat_five_standalone_emit_copy_short + LEAL -16842747(DX), DX + MOVL $0xfffb001d, (AX) + MOVB $0xff, 4(AX) + ADDQ $0x05, AX + ADDQ $0x05, BX + JMP emit_repeat_again_standalone_emit_copy_short + +repeat_five_standalone_emit_copy_short: + LEAL -65536(DX), DX + MOVL DX, CX + MOVW $0x001d, (AX) + MOVW DX, 2(AX) + SARL $0x10, CX + MOVB CL, 4(AX) + ADDQ $0x05, BX + ADDQ $0x05, AX + JMP gen_emit_copy_end + +repeat_four_standalone_emit_copy_short: + LEAL -256(DX), DX + MOVW $0x0019, (AX) + MOVW DX, 2(AX) + ADDQ $0x04, BX + ADDQ $0x04, AX + JMP gen_emit_copy_end + +repeat_three_standalone_emit_copy_short: + LEAL -4(DX), DX + MOVW $0x0015, (AX) + MOVB DL, 2(AX) + ADDQ $0x03, BX + ADDQ $0x03, AX + JMP gen_emit_copy_end + +repeat_two_standalone_emit_copy_short: + SHLL $0x02, DX + ORL $0x01, DX + MOVW DX, (AX) + ADDQ $0x02, BX + ADDQ $0x02, AX + JMP gen_emit_copy_end + +repeat_two_offset_standalone_emit_copy_short: + XORQ SI, SI + LEAL 1(SI)(DX*4), DX + MOVB CL, 1(AX) + SARL $0x08, CX + SHLL $0x05, CX + ORL CX, DX + MOVB DL, (AX) + ADDQ $0x02, BX + ADDQ $0x02, AX + JMP gen_emit_copy_end + +two_byte_offset_short_standalone: + MOVL DX, SI + SHLL $0x02, SI + CMPL DX, $0x0c + JAE emit_copy_three_standalone + CMPL CX, $0x00000800 + JAE emit_copy_three_standalone + LEAL -15(SI), SI + MOVB CL, 1(AX) + SHRL $0x08, CX + SHLL $0x05, CX + ORL CX, SI + MOVB SI, (AX) + ADDQ $0x02, BX + ADDQ $0x02, AX + JMP gen_emit_copy_end + +emit_copy_three_standalone: + LEAL -2(SI), SI + MOVB SI, (AX) + MOVW CX, 1(AX) + ADDQ $0x03, BX + ADDQ $0x03, AX + +gen_emit_copy_end: + MOVQ BX, ret+40(FP) + RET + +// func emitCopyNoRepeat(dst []byte, offset int, length int) int +TEXT ·emitCopyNoRepeat(SB), NOSPLIT, $0-48 + XORQ BX, BX + MOVQ dst_base+0(FP), AX + MOVQ offset+24(FP), CX + MOVQ length+32(FP), DX + + // emitCopy + CMPL CX, $0x00010000 + JB two_byte_offset_standalone_snappy + +four_bytes_loop_back_standalone_snappy: + CMPL DX, $0x40 + JBE four_bytes_remain_standalone_snappy + MOVB $0xff, (AX) + MOVL CX, 1(AX) + LEAL -64(DX), DX + ADDQ $0x05, BX + ADDQ $0x05, AX + CMPL DX, $0x04 + JB four_bytes_remain_standalone_snappy + JMP four_bytes_loop_back_standalone_snappy + +four_bytes_remain_standalone_snappy: + TESTL DX, DX + JZ gen_emit_copy_end_snappy + XORL SI, SI + LEAL -1(SI)(DX*4), DX + MOVB DL, (AX) + MOVL CX, 1(AX) + ADDQ $0x05, BX + ADDQ $0x05, AX + JMP gen_emit_copy_end_snappy + +two_byte_offset_standalone_snappy: + CMPL DX, $0x40 + JBE two_byte_offset_short_standalone_snappy + MOVB $0xee, (AX) + MOVW CX, 1(AX) + LEAL -60(DX), DX + ADDQ $0x03, AX + ADDQ $0x03, BX + JMP two_byte_offset_standalone_snappy + +two_byte_offset_short_standalone_snappy: + MOVL DX, SI + SHLL $0x02, SI + CMPL DX, $0x0c + JAE emit_copy_three_standalone_snappy + CMPL CX, $0x00000800 + JAE emit_copy_three_standalone_snappy + LEAL -15(SI), SI + MOVB CL, 1(AX) + SHRL $0x08, CX + SHLL $0x05, CX + ORL CX, SI + MOVB SI, (AX) + ADDQ $0x02, BX + ADDQ $0x02, AX + JMP gen_emit_copy_end_snappy + +emit_copy_three_standalone_snappy: + LEAL -2(SI), SI + MOVB SI, (AX) + MOVW CX, 1(AX) + ADDQ $0x03, BX + ADDQ $0x03, AX + +gen_emit_copy_end_snappy: + MOVQ BX, ret+40(FP) + RET + +// func matchLen(a []byte, b []byte) int +// Requires: BMI +TEXT ·matchLen(SB), NOSPLIT, $0-56 + MOVQ a_base+0(FP), AX + MOVQ b_base+24(FP), CX + MOVQ a_len+8(FP), DX + + // matchLen + XORL SI, SI + +matchlen_loopback_16_standalone: + CMPL DX, $0x10 + JB matchlen_match8_standalone + MOVQ (AX)(SI*1), BX + MOVQ 8(AX)(SI*1), DI + XORQ (CX)(SI*1), BX + JNZ matchlen_bsf_8_standalone + XORQ 8(CX)(SI*1), DI + JNZ matchlen_bsf_16standalone + LEAL -16(DX), DX + LEAL 16(SI), SI + JMP matchlen_loopback_16_standalone + +matchlen_bsf_16standalone: +#ifdef GOAMD64_v3 + TZCNTQ DI, DI + +#else + BSFQ DI, DI + +#endif + SARQ $0x03, DI + LEAL 8(SI)(DI*1), SI + JMP gen_match_len_end + +matchlen_match8_standalone: + CMPL DX, $0x08 + JB matchlen_match4_standalone + MOVQ (AX)(SI*1), BX + XORQ (CX)(SI*1), BX + JNZ matchlen_bsf_8_standalone + LEAL -8(DX), DX + LEAL 8(SI), SI + JMP matchlen_match4_standalone + +matchlen_bsf_8_standalone: +#ifdef GOAMD64_v3 + TZCNTQ BX, BX + +#else + BSFQ BX, BX + +#endif + SARQ $0x03, BX + LEAL (SI)(BX*1), SI + JMP gen_match_len_end + +matchlen_match4_standalone: + CMPL DX, $0x04 + JB matchlen_match2_standalone + MOVL (AX)(SI*1), BX + CMPL (CX)(SI*1), BX + JNE matchlen_match2_standalone + LEAL -4(DX), DX + LEAL 4(SI), SI + +matchlen_match2_standalone: + CMPL DX, $0x01 + JE matchlen_match1_standalone + JB gen_match_len_end + MOVW (AX)(SI*1), BX + CMPW (CX)(SI*1), BX + JNE matchlen_match1_standalone + LEAL 2(SI), SI + SUBL $0x02, DX + JZ gen_match_len_end + +matchlen_match1_standalone: + MOVB (AX)(SI*1), BL + CMPB (CX)(SI*1), BL + JNE gen_match_len_end + LEAL 1(SI), SI + +gen_match_len_end: + MOVQ SI, ret+48(FP) + RET + +// func cvtLZ4BlockAsm(dst []byte, src []byte) (uncompressed int, dstUsed int) +// Requires: SSE2 +TEXT ·cvtLZ4BlockAsm(SB), NOSPLIT, $0-64 + XORQ SI, SI + MOVQ dst_base+0(FP), AX + MOVQ dst_len+8(FP), CX + MOVQ src_base+24(FP), DX + MOVQ src_len+32(FP), BX + LEAQ (DX)(BX*1), BX + LEAQ -10(AX)(CX*1), CX + XORQ DI, DI + +lz4_s2_loop: + CMPQ DX, BX + JAE lz4_s2_corrupt + CMPQ AX, CX + JAE lz4_s2_dstfull + MOVBQZX (DX), R8 + MOVQ R8, R9 + MOVQ R8, R10 + SHRQ $0x04, R9 + ANDQ $0x0f, R10 + CMPQ R8, $0xf0 + JB lz4_s2_ll_end + +lz4_s2_ll_loop: + INCQ DX + CMPQ DX, BX + JAE lz4_s2_corrupt + MOVBQZX (DX), R8 + ADDQ R8, R9 + CMPQ R8, $0xff + JEQ lz4_s2_ll_loop + +lz4_s2_ll_end: + LEAQ (DX)(R9*1), R8 + ADDQ $0x04, R10 + CMPQ R8, BX + JAE lz4_s2_corrupt + INCQ DX + INCQ R8 + TESTQ R9, R9 + JZ lz4_s2_lits_done + LEAQ (AX)(R9*1), R11 + CMPQ R11, CX + JAE lz4_s2_dstfull + ADDQ R9, SI + LEAL -1(R9), R11 + CMPL R11, $0x3c + JB one_byte_lz4_s2 + CMPL R11, $0x00000100 + JB two_bytes_lz4_s2 + CMPL R11, $0x00010000 + JB three_bytes_lz4_s2 + CMPL R11, $0x01000000 + JB four_bytes_lz4_s2 + MOVB $0xfc, (AX) + MOVL R11, 1(AX) + ADDQ $0x05, AX + JMP memmove_long_lz4_s2 + +four_bytes_lz4_s2: + MOVL R11, R12 + SHRL $0x10, R12 + MOVB $0xf8, (AX) + MOVW R11, 1(AX) + MOVB R12, 3(AX) + ADDQ $0x04, AX + JMP memmove_long_lz4_s2 + +three_bytes_lz4_s2: + MOVB $0xf4, (AX) + MOVW R11, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_lz4_s2 + +two_bytes_lz4_s2: + MOVB $0xf0, (AX) + MOVB R11, 1(AX) + ADDQ $0x02, AX + CMPL R11, $0x40 + JB memmove_lz4_s2 + JMP memmove_long_lz4_s2 + +one_byte_lz4_s2: + SHLB $0x02, R11 + MOVB R11, (AX) + ADDQ $0x01, AX + +memmove_lz4_s2: + LEAQ (AX)(R9*1), R11 + + // genMemMoveShort + CMPQ R9, $0x08 + JBE emit_lit_memmove_lz4_s2_memmove_move_8 + CMPQ R9, $0x10 + JBE emit_lit_memmove_lz4_s2_memmove_move_8through16 + CMPQ R9, $0x20 + JBE emit_lit_memmove_lz4_s2_memmove_move_17through32 + JMP emit_lit_memmove_lz4_s2_memmove_move_33through64 + +emit_lit_memmove_lz4_s2_memmove_move_8: + MOVQ (DX), R12 + MOVQ R12, (AX) + JMP memmove_end_copy_lz4_s2 + +emit_lit_memmove_lz4_s2_memmove_move_8through16: + MOVQ (DX), R12 + MOVQ -8(DX)(R9*1), DX + MOVQ R12, (AX) + MOVQ DX, -8(AX)(R9*1) + JMP memmove_end_copy_lz4_s2 + +emit_lit_memmove_lz4_s2_memmove_move_17through32: + MOVOU (DX), X0 + MOVOU -16(DX)(R9*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(R9*1) + JMP memmove_end_copy_lz4_s2 + +emit_lit_memmove_lz4_s2_memmove_move_33through64: + MOVOU (DX), X0 + MOVOU 16(DX), X1 + MOVOU -32(DX)(R9*1), X2 + MOVOU -16(DX)(R9*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R9*1) + MOVOU X3, -16(AX)(R9*1) + +memmove_end_copy_lz4_s2: + MOVQ R11, AX + JMP lz4_s2_lits_emit_done + +memmove_long_lz4_s2: + LEAQ (AX)(R9*1), R11 + + // genMemMoveLong + MOVOU (DX), X0 + MOVOU 16(DX), X1 + MOVOU -32(DX)(R9*1), X2 + MOVOU -16(DX)(R9*1), X3 + MOVQ R9, R13 + SHRQ $0x05, R13 + MOVQ AX, R12 + ANDL $0x0000001f, R12 + MOVQ $0x00000040, R14 + SUBQ R12, R14 + DECQ R13 + JA emit_lit_memmove_long_lz4_s2large_forward_sse_loop_32 + LEAQ -32(DX)(R14*1), R12 + LEAQ -32(AX)(R14*1), R15 + +emit_lit_memmove_long_lz4_s2large_big_loop_back: + MOVOU (R12), X4 + MOVOU 16(R12), X5 + MOVOA X4, (R15) + MOVOA X5, 16(R15) + ADDQ $0x20, R15 + ADDQ $0x20, R12 + ADDQ $0x20, R14 + DECQ R13 + JNA emit_lit_memmove_long_lz4_s2large_big_loop_back + +emit_lit_memmove_long_lz4_s2large_forward_sse_loop_32: + MOVOU -32(DX)(R14*1), X4 + MOVOU -16(DX)(R14*1), X5 + MOVOA X4, -32(AX)(R14*1) + MOVOA X5, -16(AX)(R14*1) + ADDQ $0x20, R14 + CMPQ R9, R14 + JAE emit_lit_memmove_long_lz4_s2large_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R9*1) + MOVOU X3, -16(AX)(R9*1) + MOVQ R11, AX + +lz4_s2_lits_emit_done: + MOVQ R8, DX + +lz4_s2_lits_done: + CMPQ DX, BX + JNE lz4_s2_match + CMPQ R10, $0x04 + JEQ lz4_s2_done + JMP lz4_s2_corrupt + +lz4_s2_match: + LEAQ 2(DX), R8 + CMPQ R8, BX + JAE lz4_s2_corrupt + MOVWQZX (DX), R9 + MOVQ R8, DX + TESTQ R9, R9 + JZ lz4_s2_corrupt + CMPQ R9, SI + JA lz4_s2_corrupt + CMPQ R10, $0x13 + JNE lz4_s2_ml_done + +lz4_s2_ml_loop: + MOVBQZX (DX), R8 + INCQ DX + ADDQ R8, R10 + CMPQ DX, BX + JAE lz4_s2_corrupt + CMPQ R8, $0xff + JEQ lz4_s2_ml_loop + +lz4_s2_ml_done: + ADDQ R10, SI + CMPQ R9, DI + JNE lz4_s2_docopy + + // emitRepeat +emit_repeat_again_lz4_s2: + MOVL R10, R8 + LEAL -4(R10), R10 + CMPL R8, $0x08 + JBE repeat_two_lz4_s2 + CMPL R8, $0x0c + JAE cant_repeat_two_offset_lz4_s2 + CMPL R9, $0x00000800 + JB repeat_two_offset_lz4_s2 + +cant_repeat_two_offset_lz4_s2: + CMPL R10, $0x00000104 + JB repeat_three_lz4_s2 + CMPL R10, $0x00010100 + JB repeat_four_lz4_s2 + CMPL R10, $0x0100ffff + JB repeat_five_lz4_s2 + LEAL -16842747(R10), R10 + MOVL $0xfffb001d, (AX) + MOVB $0xff, 4(AX) + ADDQ $0x05, AX + JMP emit_repeat_again_lz4_s2 + +repeat_five_lz4_s2: + LEAL -65536(R10), R10 + MOVL R10, R9 + MOVW $0x001d, (AX) + MOVW R10, 2(AX) + SARL $0x10, R9 + MOVB R9, 4(AX) + ADDQ $0x05, AX + JMP lz4_s2_loop + +repeat_four_lz4_s2: + LEAL -256(R10), R10 + MOVW $0x0019, (AX) + MOVW R10, 2(AX) + ADDQ $0x04, AX + JMP lz4_s2_loop + +repeat_three_lz4_s2: + LEAL -4(R10), R10 + MOVW $0x0015, (AX) + MOVB R10, 2(AX) + ADDQ $0x03, AX + JMP lz4_s2_loop + +repeat_two_lz4_s2: + SHLL $0x02, R10 + ORL $0x01, R10 + MOVW R10, (AX) + ADDQ $0x02, AX + JMP lz4_s2_loop + +repeat_two_offset_lz4_s2: + XORQ R8, R8 + LEAL 1(R8)(R10*4), R10 + MOVB R9, 1(AX) + SARL $0x08, R9 + SHLL $0x05, R9 + ORL R9, R10 + MOVB R10, (AX) + ADDQ $0x02, AX + JMP lz4_s2_loop + +lz4_s2_docopy: + MOVQ R9, DI + + // emitCopy + CMPL R10, $0x40 + JBE two_byte_offset_short_lz4_s2 + CMPL R9, $0x00000800 + JAE long_offset_short_lz4_s2 + MOVL $0x00000001, R8 + LEAL 16(R8), R8 + MOVB R9, 1(AX) + MOVL R9, R11 + SHRL $0x08, R11 + SHLL $0x05, R11 + ORL R11, R8 + MOVB R8, (AX) + ADDQ $0x02, AX + SUBL $0x08, R10 + + // emitRepeat + LEAL -4(R10), R10 + JMP cant_repeat_two_offset_lz4_s2_emit_copy_short_2b + +emit_repeat_again_lz4_s2_emit_copy_short_2b: + MOVL R10, R8 + LEAL -4(R10), R10 + CMPL R8, $0x08 + JBE repeat_two_lz4_s2_emit_copy_short_2b + CMPL R8, $0x0c + JAE cant_repeat_two_offset_lz4_s2_emit_copy_short_2b + CMPL R9, $0x00000800 + JB repeat_two_offset_lz4_s2_emit_copy_short_2b + +cant_repeat_two_offset_lz4_s2_emit_copy_short_2b: + CMPL R10, $0x00000104 + JB repeat_three_lz4_s2_emit_copy_short_2b + CMPL R10, $0x00010100 + JB repeat_four_lz4_s2_emit_copy_short_2b + CMPL R10, $0x0100ffff + JB repeat_five_lz4_s2_emit_copy_short_2b + LEAL -16842747(R10), R10 + MOVL $0xfffb001d, (AX) + MOVB $0xff, 4(AX) + ADDQ $0x05, AX + JMP emit_repeat_again_lz4_s2_emit_copy_short_2b + +repeat_five_lz4_s2_emit_copy_short_2b: + LEAL -65536(R10), R10 + MOVL R10, R9 + MOVW $0x001d, (AX) + MOVW R10, 2(AX) + SARL $0x10, R9 + MOVB R9, 4(AX) + ADDQ $0x05, AX + JMP lz4_s2_loop + +repeat_four_lz4_s2_emit_copy_short_2b: + LEAL -256(R10), R10 + MOVW $0x0019, (AX) + MOVW R10, 2(AX) + ADDQ $0x04, AX + JMP lz4_s2_loop + +repeat_three_lz4_s2_emit_copy_short_2b: + LEAL -4(R10), R10 + MOVW $0x0015, (AX) + MOVB R10, 2(AX) + ADDQ $0x03, AX + JMP lz4_s2_loop + +repeat_two_lz4_s2_emit_copy_short_2b: + SHLL $0x02, R10 + ORL $0x01, R10 + MOVW R10, (AX) + ADDQ $0x02, AX + JMP lz4_s2_loop + +repeat_two_offset_lz4_s2_emit_copy_short_2b: + XORQ R8, R8 + LEAL 1(R8)(R10*4), R10 + MOVB R9, 1(AX) + SARL $0x08, R9 + SHLL $0x05, R9 + ORL R9, R10 + MOVB R10, (AX) + ADDQ $0x02, AX + JMP lz4_s2_loop + +long_offset_short_lz4_s2: + MOVB $0xee, (AX) + MOVW R9, 1(AX) + LEAL -60(R10), R10 + ADDQ $0x03, AX + + // emitRepeat +emit_repeat_again_lz4_s2_emit_copy_short: + MOVL R10, R8 + LEAL -4(R10), R10 + CMPL R8, $0x08 + JBE repeat_two_lz4_s2_emit_copy_short + CMPL R8, $0x0c + JAE cant_repeat_two_offset_lz4_s2_emit_copy_short + CMPL R9, $0x00000800 + JB repeat_two_offset_lz4_s2_emit_copy_short + +cant_repeat_two_offset_lz4_s2_emit_copy_short: + CMPL R10, $0x00000104 + JB repeat_three_lz4_s2_emit_copy_short + CMPL R10, $0x00010100 + JB repeat_four_lz4_s2_emit_copy_short + CMPL R10, $0x0100ffff + JB repeat_five_lz4_s2_emit_copy_short + LEAL -16842747(R10), R10 + MOVL $0xfffb001d, (AX) + MOVB $0xff, 4(AX) + ADDQ $0x05, AX + JMP emit_repeat_again_lz4_s2_emit_copy_short + +repeat_five_lz4_s2_emit_copy_short: + LEAL -65536(R10), R10 + MOVL R10, R9 + MOVW $0x001d, (AX) + MOVW R10, 2(AX) + SARL $0x10, R9 + MOVB R9, 4(AX) + ADDQ $0x05, AX + JMP lz4_s2_loop + +repeat_four_lz4_s2_emit_copy_short: + LEAL -256(R10), R10 + MOVW $0x0019, (AX) + MOVW R10, 2(AX) + ADDQ $0x04, AX + JMP lz4_s2_loop + +repeat_three_lz4_s2_emit_copy_short: + LEAL -4(R10), R10 + MOVW $0x0015, (AX) + MOVB R10, 2(AX) + ADDQ $0x03, AX + JMP lz4_s2_loop + +repeat_two_lz4_s2_emit_copy_short: + SHLL $0x02, R10 + ORL $0x01, R10 + MOVW R10, (AX) + ADDQ $0x02, AX + JMP lz4_s2_loop + +repeat_two_offset_lz4_s2_emit_copy_short: + XORQ R8, R8 + LEAL 1(R8)(R10*4), R10 + MOVB R9, 1(AX) + SARL $0x08, R9 + SHLL $0x05, R9 + ORL R9, R10 + MOVB R10, (AX) + ADDQ $0x02, AX + JMP lz4_s2_loop + +two_byte_offset_short_lz4_s2: + MOVL R10, R8 + SHLL $0x02, R8 + CMPL R10, $0x0c + JAE emit_copy_three_lz4_s2 + CMPL R9, $0x00000800 + JAE emit_copy_three_lz4_s2 + LEAL -15(R8), R8 + MOVB R9, 1(AX) + SHRL $0x08, R9 + SHLL $0x05, R9 + ORL R9, R8 + MOVB R8, (AX) + ADDQ $0x02, AX + JMP lz4_s2_loop + +emit_copy_three_lz4_s2: + LEAL -2(R8), R8 + MOVB R8, (AX) + MOVW R9, 1(AX) + ADDQ $0x03, AX + JMP lz4_s2_loop + +lz4_s2_done: + MOVQ dst_base+0(FP), CX + SUBQ CX, AX + MOVQ SI, uncompressed+48(FP) + MOVQ AX, dstUsed+56(FP) + RET + +lz4_s2_corrupt: + XORQ AX, AX + LEAQ -1(AX), SI + MOVQ SI, uncompressed+48(FP) + RET + +lz4_s2_dstfull: + XORQ AX, AX + LEAQ -2(AX), SI + MOVQ SI, uncompressed+48(FP) + RET + +// func cvtLZ4sBlockAsm(dst []byte, src []byte) (uncompressed int, dstUsed int) +// Requires: SSE2 +TEXT ·cvtLZ4sBlockAsm(SB), NOSPLIT, $0-64 + XORQ SI, SI + MOVQ dst_base+0(FP), AX + MOVQ dst_len+8(FP), CX + MOVQ src_base+24(FP), DX + MOVQ src_len+32(FP), BX + LEAQ (DX)(BX*1), BX + LEAQ -10(AX)(CX*1), CX + XORQ DI, DI + +lz4s_s2_loop: + CMPQ DX, BX + JAE lz4s_s2_corrupt + CMPQ AX, CX + JAE lz4s_s2_dstfull + MOVBQZX (DX), R8 + MOVQ R8, R9 + MOVQ R8, R10 + SHRQ $0x04, R9 + ANDQ $0x0f, R10 + CMPQ R8, $0xf0 + JB lz4s_s2_ll_end + +lz4s_s2_ll_loop: + INCQ DX + CMPQ DX, BX + JAE lz4s_s2_corrupt + MOVBQZX (DX), R8 + ADDQ R8, R9 + CMPQ R8, $0xff + JEQ lz4s_s2_ll_loop + +lz4s_s2_ll_end: + LEAQ (DX)(R9*1), R8 + ADDQ $0x03, R10 + CMPQ R8, BX + JAE lz4s_s2_corrupt + INCQ DX + INCQ R8 + TESTQ R9, R9 + JZ lz4s_s2_lits_done + LEAQ (AX)(R9*1), R11 + CMPQ R11, CX + JAE lz4s_s2_dstfull + ADDQ R9, SI + LEAL -1(R9), R11 + CMPL R11, $0x3c + JB one_byte_lz4s_s2 + CMPL R11, $0x00000100 + JB two_bytes_lz4s_s2 + CMPL R11, $0x00010000 + JB three_bytes_lz4s_s2 + CMPL R11, $0x01000000 + JB four_bytes_lz4s_s2 + MOVB $0xfc, (AX) + MOVL R11, 1(AX) + ADDQ $0x05, AX + JMP memmove_long_lz4s_s2 + +four_bytes_lz4s_s2: + MOVL R11, R12 + SHRL $0x10, R12 + MOVB $0xf8, (AX) + MOVW R11, 1(AX) + MOVB R12, 3(AX) + ADDQ $0x04, AX + JMP memmove_long_lz4s_s2 + +three_bytes_lz4s_s2: + MOVB $0xf4, (AX) + MOVW R11, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_lz4s_s2 + +two_bytes_lz4s_s2: + MOVB $0xf0, (AX) + MOVB R11, 1(AX) + ADDQ $0x02, AX + CMPL R11, $0x40 + JB memmove_lz4s_s2 + JMP memmove_long_lz4s_s2 + +one_byte_lz4s_s2: + SHLB $0x02, R11 + MOVB R11, (AX) + ADDQ $0x01, AX + +memmove_lz4s_s2: + LEAQ (AX)(R9*1), R11 + + // genMemMoveShort + CMPQ R9, $0x08 + JBE emit_lit_memmove_lz4s_s2_memmove_move_8 + CMPQ R9, $0x10 + JBE emit_lit_memmove_lz4s_s2_memmove_move_8through16 + CMPQ R9, $0x20 + JBE emit_lit_memmove_lz4s_s2_memmove_move_17through32 + JMP emit_lit_memmove_lz4s_s2_memmove_move_33through64 + +emit_lit_memmove_lz4s_s2_memmove_move_8: + MOVQ (DX), R12 + MOVQ R12, (AX) + JMP memmove_end_copy_lz4s_s2 + +emit_lit_memmove_lz4s_s2_memmove_move_8through16: + MOVQ (DX), R12 + MOVQ -8(DX)(R9*1), DX + MOVQ R12, (AX) + MOVQ DX, -8(AX)(R9*1) + JMP memmove_end_copy_lz4s_s2 + +emit_lit_memmove_lz4s_s2_memmove_move_17through32: + MOVOU (DX), X0 + MOVOU -16(DX)(R9*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(R9*1) + JMP memmove_end_copy_lz4s_s2 + +emit_lit_memmove_lz4s_s2_memmove_move_33through64: + MOVOU (DX), X0 + MOVOU 16(DX), X1 + MOVOU -32(DX)(R9*1), X2 + MOVOU -16(DX)(R9*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R9*1) + MOVOU X3, -16(AX)(R9*1) + +memmove_end_copy_lz4s_s2: + MOVQ R11, AX + JMP lz4s_s2_lits_emit_done + +memmove_long_lz4s_s2: + LEAQ (AX)(R9*1), R11 + + // genMemMoveLong + MOVOU (DX), X0 + MOVOU 16(DX), X1 + MOVOU -32(DX)(R9*1), X2 + MOVOU -16(DX)(R9*1), X3 + MOVQ R9, R13 + SHRQ $0x05, R13 + MOVQ AX, R12 + ANDL $0x0000001f, R12 + MOVQ $0x00000040, R14 + SUBQ R12, R14 + DECQ R13 + JA emit_lit_memmove_long_lz4s_s2large_forward_sse_loop_32 + LEAQ -32(DX)(R14*1), R12 + LEAQ -32(AX)(R14*1), R15 + +emit_lit_memmove_long_lz4s_s2large_big_loop_back: + MOVOU (R12), X4 + MOVOU 16(R12), X5 + MOVOA X4, (R15) + MOVOA X5, 16(R15) + ADDQ $0x20, R15 + ADDQ $0x20, R12 + ADDQ $0x20, R14 + DECQ R13 + JNA emit_lit_memmove_long_lz4s_s2large_big_loop_back + +emit_lit_memmove_long_lz4s_s2large_forward_sse_loop_32: + MOVOU -32(DX)(R14*1), X4 + MOVOU -16(DX)(R14*1), X5 + MOVOA X4, -32(AX)(R14*1) + MOVOA X5, -16(AX)(R14*1) + ADDQ $0x20, R14 + CMPQ R9, R14 + JAE emit_lit_memmove_long_lz4s_s2large_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R9*1) + MOVOU X3, -16(AX)(R9*1) + MOVQ R11, AX + +lz4s_s2_lits_emit_done: + MOVQ R8, DX + +lz4s_s2_lits_done: + CMPQ DX, BX + JNE lz4s_s2_match + CMPQ R10, $0x03 + JEQ lz4s_s2_done + JMP lz4s_s2_corrupt + +lz4s_s2_match: + CMPQ R10, $0x03 + JEQ lz4s_s2_loop + LEAQ 2(DX), R8 + CMPQ R8, BX + JAE lz4s_s2_corrupt + MOVWQZX (DX), R9 + MOVQ R8, DX + TESTQ R9, R9 + JZ lz4s_s2_corrupt + CMPQ R9, SI + JA lz4s_s2_corrupt + CMPQ R10, $0x12 + JNE lz4s_s2_ml_done + +lz4s_s2_ml_loop: + MOVBQZX (DX), R8 + INCQ DX + ADDQ R8, R10 + CMPQ DX, BX + JAE lz4s_s2_corrupt + CMPQ R8, $0xff + JEQ lz4s_s2_ml_loop + +lz4s_s2_ml_done: + ADDQ R10, SI + CMPQ R9, DI + JNE lz4s_s2_docopy + + // emitRepeat +emit_repeat_again_lz4_s2: + MOVL R10, R8 + LEAL -4(R10), R10 + CMPL R8, $0x08 + JBE repeat_two_lz4_s2 + CMPL R8, $0x0c + JAE cant_repeat_two_offset_lz4_s2 + CMPL R9, $0x00000800 + JB repeat_two_offset_lz4_s2 + +cant_repeat_two_offset_lz4_s2: + CMPL R10, $0x00000104 + JB repeat_three_lz4_s2 + CMPL R10, $0x00010100 + JB repeat_four_lz4_s2 + CMPL R10, $0x0100ffff + JB repeat_five_lz4_s2 + LEAL -16842747(R10), R10 + MOVL $0xfffb001d, (AX) + MOVB $0xff, 4(AX) + ADDQ $0x05, AX + JMP emit_repeat_again_lz4_s2 + +repeat_five_lz4_s2: + LEAL -65536(R10), R10 + MOVL R10, R9 + MOVW $0x001d, (AX) + MOVW R10, 2(AX) + SARL $0x10, R9 + MOVB R9, 4(AX) + ADDQ $0x05, AX + JMP lz4s_s2_loop + +repeat_four_lz4_s2: + LEAL -256(R10), R10 + MOVW $0x0019, (AX) + MOVW R10, 2(AX) + ADDQ $0x04, AX + JMP lz4s_s2_loop + +repeat_three_lz4_s2: + LEAL -4(R10), R10 + MOVW $0x0015, (AX) + MOVB R10, 2(AX) + ADDQ $0x03, AX + JMP lz4s_s2_loop + +repeat_two_lz4_s2: + SHLL $0x02, R10 + ORL $0x01, R10 + MOVW R10, (AX) + ADDQ $0x02, AX + JMP lz4s_s2_loop + +repeat_two_offset_lz4_s2: + XORQ R8, R8 + LEAL 1(R8)(R10*4), R10 + MOVB R9, 1(AX) + SARL $0x08, R9 + SHLL $0x05, R9 + ORL R9, R10 + MOVB R10, (AX) + ADDQ $0x02, AX + JMP lz4s_s2_loop + +lz4s_s2_docopy: + MOVQ R9, DI + + // emitCopy + CMPL R10, $0x40 + JBE two_byte_offset_short_lz4_s2 + CMPL R9, $0x00000800 + JAE long_offset_short_lz4_s2 + MOVL $0x00000001, R8 + LEAL 16(R8), R8 + MOVB R9, 1(AX) + MOVL R9, R11 + SHRL $0x08, R11 + SHLL $0x05, R11 + ORL R11, R8 + MOVB R8, (AX) + ADDQ $0x02, AX + SUBL $0x08, R10 + + // emitRepeat + LEAL -4(R10), R10 + JMP cant_repeat_two_offset_lz4_s2_emit_copy_short_2b + +emit_repeat_again_lz4_s2_emit_copy_short_2b: + MOVL R10, R8 + LEAL -4(R10), R10 + CMPL R8, $0x08 + JBE repeat_two_lz4_s2_emit_copy_short_2b + CMPL R8, $0x0c + JAE cant_repeat_two_offset_lz4_s2_emit_copy_short_2b + CMPL R9, $0x00000800 + JB repeat_two_offset_lz4_s2_emit_copy_short_2b + +cant_repeat_two_offset_lz4_s2_emit_copy_short_2b: + CMPL R10, $0x00000104 + JB repeat_three_lz4_s2_emit_copy_short_2b + CMPL R10, $0x00010100 + JB repeat_four_lz4_s2_emit_copy_short_2b + CMPL R10, $0x0100ffff + JB repeat_five_lz4_s2_emit_copy_short_2b + LEAL -16842747(R10), R10 + MOVL $0xfffb001d, (AX) + MOVB $0xff, 4(AX) + ADDQ $0x05, AX + JMP emit_repeat_again_lz4_s2_emit_copy_short_2b + +repeat_five_lz4_s2_emit_copy_short_2b: + LEAL -65536(R10), R10 + MOVL R10, R9 + MOVW $0x001d, (AX) + MOVW R10, 2(AX) + SARL $0x10, R9 + MOVB R9, 4(AX) + ADDQ $0x05, AX + JMP lz4s_s2_loop + +repeat_four_lz4_s2_emit_copy_short_2b: + LEAL -256(R10), R10 + MOVW $0x0019, (AX) + MOVW R10, 2(AX) + ADDQ $0x04, AX + JMP lz4s_s2_loop + +repeat_three_lz4_s2_emit_copy_short_2b: + LEAL -4(R10), R10 + MOVW $0x0015, (AX) + MOVB R10, 2(AX) + ADDQ $0x03, AX + JMP lz4s_s2_loop + +repeat_two_lz4_s2_emit_copy_short_2b: + SHLL $0x02, R10 + ORL $0x01, R10 + MOVW R10, (AX) + ADDQ $0x02, AX + JMP lz4s_s2_loop + +repeat_two_offset_lz4_s2_emit_copy_short_2b: + XORQ R8, R8 + LEAL 1(R8)(R10*4), R10 + MOVB R9, 1(AX) + SARL $0x08, R9 + SHLL $0x05, R9 + ORL R9, R10 + MOVB R10, (AX) + ADDQ $0x02, AX + JMP lz4s_s2_loop + +long_offset_short_lz4_s2: + MOVB $0xee, (AX) + MOVW R9, 1(AX) + LEAL -60(R10), R10 + ADDQ $0x03, AX + + // emitRepeat +emit_repeat_again_lz4_s2_emit_copy_short: + MOVL R10, R8 + LEAL -4(R10), R10 + CMPL R8, $0x08 + JBE repeat_two_lz4_s2_emit_copy_short + CMPL R8, $0x0c + JAE cant_repeat_two_offset_lz4_s2_emit_copy_short + CMPL R9, $0x00000800 + JB repeat_two_offset_lz4_s2_emit_copy_short + +cant_repeat_two_offset_lz4_s2_emit_copy_short: + CMPL R10, $0x00000104 + JB repeat_three_lz4_s2_emit_copy_short + CMPL R10, $0x00010100 + JB repeat_four_lz4_s2_emit_copy_short + CMPL R10, $0x0100ffff + JB repeat_five_lz4_s2_emit_copy_short + LEAL -16842747(R10), R10 + MOVL $0xfffb001d, (AX) + MOVB $0xff, 4(AX) + ADDQ $0x05, AX + JMP emit_repeat_again_lz4_s2_emit_copy_short + +repeat_five_lz4_s2_emit_copy_short: + LEAL -65536(R10), R10 + MOVL R10, R9 + MOVW $0x001d, (AX) + MOVW R10, 2(AX) + SARL $0x10, R9 + MOVB R9, 4(AX) + ADDQ $0x05, AX + JMP lz4s_s2_loop + +repeat_four_lz4_s2_emit_copy_short: + LEAL -256(R10), R10 + MOVW $0x0019, (AX) + MOVW R10, 2(AX) + ADDQ $0x04, AX + JMP lz4s_s2_loop + +repeat_three_lz4_s2_emit_copy_short: + LEAL -4(R10), R10 + MOVW $0x0015, (AX) + MOVB R10, 2(AX) + ADDQ $0x03, AX + JMP lz4s_s2_loop + +repeat_two_lz4_s2_emit_copy_short: + SHLL $0x02, R10 + ORL $0x01, R10 + MOVW R10, (AX) + ADDQ $0x02, AX + JMP lz4s_s2_loop + +repeat_two_offset_lz4_s2_emit_copy_short: + XORQ R8, R8 + LEAL 1(R8)(R10*4), R10 + MOVB R9, 1(AX) + SARL $0x08, R9 + SHLL $0x05, R9 + ORL R9, R10 + MOVB R10, (AX) + ADDQ $0x02, AX + JMP lz4s_s2_loop + +two_byte_offset_short_lz4_s2: + MOVL R10, R8 + SHLL $0x02, R8 + CMPL R10, $0x0c + JAE emit_copy_three_lz4_s2 + CMPL R9, $0x00000800 + JAE emit_copy_three_lz4_s2 + LEAL -15(R8), R8 + MOVB R9, 1(AX) + SHRL $0x08, R9 + SHLL $0x05, R9 + ORL R9, R8 + MOVB R8, (AX) + ADDQ $0x02, AX + JMP lz4s_s2_loop + +emit_copy_three_lz4_s2: + LEAL -2(R8), R8 + MOVB R8, (AX) + MOVW R9, 1(AX) + ADDQ $0x03, AX + JMP lz4s_s2_loop + +lz4s_s2_done: + MOVQ dst_base+0(FP), CX + SUBQ CX, AX + MOVQ SI, uncompressed+48(FP) + MOVQ AX, dstUsed+56(FP) + RET + +lz4s_s2_corrupt: + XORQ AX, AX + LEAQ -1(AX), SI + MOVQ SI, uncompressed+48(FP) + RET + +lz4s_s2_dstfull: + XORQ AX, AX + LEAQ -2(AX), SI + MOVQ SI, uncompressed+48(FP) + RET + +// func cvtLZ4BlockSnappyAsm(dst []byte, src []byte) (uncompressed int, dstUsed int) +// Requires: SSE2 +TEXT ·cvtLZ4BlockSnappyAsm(SB), NOSPLIT, $0-64 + XORQ SI, SI + MOVQ dst_base+0(FP), AX + MOVQ dst_len+8(FP), CX + MOVQ src_base+24(FP), DX + MOVQ src_len+32(FP), BX + LEAQ (DX)(BX*1), BX + LEAQ -10(AX)(CX*1), CX + +lz4_snappy_loop: + CMPQ DX, BX + JAE lz4_snappy_corrupt + CMPQ AX, CX + JAE lz4_snappy_dstfull + MOVBQZX (DX), DI + MOVQ DI, R8 + MOVQ DI, R9 + SHRQ $0x04, R8 + ANDQ $0x0f, R9 + CMPQ DI, $0xf0 + JB lz4_snappy_ll_end + +lz4_snappy_ll_loop: + INCQ DX + CMPQ DX, BX + JAE lz4_snappy_corrupt + MOVBQZX (DX), DI + ADDQ DI, R8 + CMPQ DI, $0xff + JEQ lz4_snappy_ll_loop + +lz4_snappy_ll_end: + LEAQ (DX)(R8*1), DI + ADDQ $0x04, R9 + CMPQ DI, BX + JAE lz4_snappy_corrupt + INCQ DX + INCQ DI + TESTQ R8, R8 + JZ lz4_snappy_lits_done + LEAQ (AX)(R8*1), R10 + CMPQ R10, CX + JAE lz4_snappy_dstfull + ADDQ R8, SI + LEAL -1(R8), R10 + CMPL R10, $0x3c + JB one_byte_lz4_snappy + CMPL R10, $0x00000100 + JB two_bytes_lz4_snappy + CMPL R10, $0x00010000 + JB three_bytes_lz4_snappy + CMPL R10, $0x01000000 + JB four_bytes_lz4_snappy + MOVB $0xfc, (AX) + MOVL R10, 1(AX) + ADDQ $0x05, AX + JMP memmove_long_lz4_snappy + +four_bytes_lz4_snappy: + MOVL R10, R11 + SHRL $0x10, R11 + MOVB $0xf8, (AX) + MOVW R10, 1(AX) + MOVB R11, 3(AX) + ADDQ $0x04, AX + JMP memmove_long_lz4_snappy + +three_bytes_lz4_snappy: + MOVB $0xf4, (AX) + MOVW R10, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_lz4_snappy + +two_bytes_lz4_snappy: + MOVB $0xf0, (AX) + MOVB R10, 1(AX) + ADDQ $0x02, AX + CMPL R10, $0x40 + JB memmove_lz4_snappy + JMP memmove_long_lz4_snappy + +one_byte_lz4_snappy: + SHLB $0x02, R10 + MOVB R10, (AX) + ADDQ $0x01, AX + +memmove_lz4_snappy: + LEAQ (AX)(R8*1), R10 + + // genMemMoveShort + CMPQ R8, $0x08 + JBE emit_lit_memmove_lz4_snappy_memmove_move_8 + CMPQ R8, $0x10 + JBE emit_lit_memmove_lz4_snappy_memmove_move_8through16 + CMPQ R8, $0x20 + JBE emit_lit_memmove_lz4_snappy_memmove_move_17through32 + JMP emit_lit_memmove_lz4_snappy_memmove_move_33through64 + +emit_lit_memmove_lz4_snappy_memmove_move_8: + MOVQ (DX), R11 + MOVQ R11, (AX) + JMP memmove_end_copy_lz4_snappy + +emit_lit_memmove_lz4_snappy_memmove_move_8through16: + MOVQ (DX), R11 + MOVQ -8(DX)(R8*1), DX + MOVQ R11, (AX) + MOVQ DX, -8(AX)(R8*1) + JMP memmove_end_copy_lz4_snappy + +emit_lit_memmove_lz4_snappy_memmove_move_17through32: + MOVOU (DX), X0 + MOVOU -16(DX)(R8*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(R8*1) + JMP memmove_end_copy_lz4_snappy + +emit_lit_memmove_lz4_snappy_memmove_move_33through64: + MOVOU (DX), X0 + MOVOU 16(DX), X1 + MOVOU -32(DX)(R8*1), X2 + MOVOU -16(DX)(R8*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + +memmove_end_copy_lz4_snappy: + MOVQ R10, AX + JMP lz4_snappy_lits_emit_done + +memmove_long_lz4_snappy: + LEAQ (AX)(R8*1), R10 + + // genMemMoveLong + MOVOU (DX), X0 + MOVOU 16(DX), X1 + MOVOU -32(DX)(R8*1), X2 + MOVOU -16(DX)(R8*1), X3 + MOVQ R8, R12 + SHRQ $0x05, R12 + MOVQ AX, R11 + ANDL $0x0000001f, R11 + MOVQ $0x00000040, R13 + SUBQ R11, R13 + DECQ R12 + JA emit_lit_memmove_long_lz4_snappylarge_forward_sse_loop_32 + LEAQ -32(DX)(R13*1), R11 + LEAQ -32(AX)(R13*1), R14 + +emit_lit_memmove_long_lz4_snappylarge_big_loop_back: + MOVOU (R11), X4 + MOVOU 16(R11), X5 + MOVOA X4, (R14) + MOVOA X5, 16(R14) + ADDQ $0x20, R14 + ADDQ $0x20, R11 + ADDQ $0x20, R13 + DECQ R12 + JNA emit_lit_memmove_long_lz4_snappylarge_big_loop_back + +emit_lit_memmove_long_lz4_snappylarge_forward_sse_loop_32: + MOVOU -32(DX)(R13*1), X4 + MOVOU -16(DX)(R13*1), X5 + MOVOA X4, -32(AX)(R13*1) + MOVOA X5, -16(AX)(R13*1) + ADDQ $0x20, R13 + CMPQ R8, R13 + JAE emit_lit_memmove_long_lz4_snappylarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + MOVQ R10, AX + +lz4_snappy_lits_emit_done: + MOVQ DI, DX + +lz4_snappy_lits_done: + CMPQ DX, BX + JNE lz4_snappy_match + CMPQ R9, $0x04 + JEQ lz4_snappy_done + JMP lz4_snappy_corrupt + +lz4_snappy_match: + LEAQ 2(DX), DI + CMPQ DI, BX + JAE lz4_snappy_corrupt + MOVWQZX (DX), R8 + MOVQ DI, DX + TESTQ R8, R8 + JZ lz4_snappy_corrupt + CMPQ R8, SI + JA lz4_snappy_corrupt + CMPQ R9, $0x13 + JNE lz4_snappy_ml_done + +lz4_snappy_ml_loop: + MOVBQZX (DX), DI + INCQ DX + ADDQ DI, R9 + CMPQ DX, BX + JAE lz4_snappy_corrupt + CMPQ DI, $0xff + JEQ lz4_snappy_ml_loop + +lz4_snappy_ml_done: + ADDQ R9, SI + + // emitCopy +two_byte_offset_lz4_s2: + CMPL R9, $0x40 + JBE two_byte_offset_short_lz4_s2 + MOVB $0xee, (AX) + MOVW R8, 1(AX) + LEAL -60(R9), R9 + ADDQ $0x03, AX + CMPQ AX, CX + JAE lz4_snappy_loop + JMP two_byte_offset_lz4_s2 + +two_byte_offset_short_lz4_s2: + MOVL R9, DI + SHLL $0x02, DI + CMPL R9, $0x0c + JAE emit_copy_three_lz4_s2 + CMPL R8, $0x00000800 + JAE emit_copy_three_lz4_s2 + LEAL -15(DI), DI + MOVB R8, 1(AX) + SHRL $0x08, R8 + SHLL $0x05, R8 + ORL R8, DI + MOVB DI, (AX) + ADDQ $0x02, AX + JMP lz4_snappy_loop + +emit_copy_three_lz4_s2: + LEAL -2(DI), DI + MOVB DI, (AX) + MOVW R8, 1(AX) + ADDQ $0x03, AX + JMP lz4_snappy_loop + +lz4_snappy_done: + MOVQ dst_base+0(FP), CX + SUBQ CX, AX + MOVQ SI, uncompressed+48(FP) + MOVQ AX, dstUsed+56(FP) + RET + +lz4_snappy_corrupt: + XORQ AX, AX + LEAQ -1(AX), SI + MOVQ SI, uncompressed+48(FP) + RET + +lz4_snappy_dstfull: + XORQ AX, AX + LEAQ -2(AX), SI + MOVQ SI, uncompressed+48(FP) + RET + +// func cvtLZ4sBlockSnappyAsm(dst []byte, src []byte) (uncompressed int, dstUsed int) +// Requires: SSE2 +TEXT ·cvtLZ4sBlockSnappyAsm(SB), NOSPLIT, $0-64 + XORQ SI, SI + MOVQ dst_base+0(FP), AX + MOVQ dst_len+8(FP), CX + MOVQ src_base+24(FP), DX + MOVQ src_len+32(FP), BX + LEAQ (DX)(BX*1), BX + LEAQ -10(AX)(CX*1), CX + +lz4s_snappy_loop: + CMPQ DX, BX + JAE lz4s_snappy_corrupt + CMPQ AX, CX + JAE lz4s_snappy_dstfull + MOVBQZX (DX), DI + MOVQ DI, R8 + MOVQ DI, R9 + SHRQ $0x04, R8 + ANDQ $0x0f, R9 + CMPQ DI, $0xf0 + JB lz4s_snappy_ll_end + +lz4s_snappy_ll_loop: + INCQ DX + CMPQ DX, BX + JAE lz4s_snappy_corrupt + MOVBQZX (DX), DI + ADDQ DI, R8 + CMPQ DI, $0xff + JEQ lz4s_snappy_ll_loop + +lz4s_snappy_ll_end: + LEAQ (DX)(R8*1), DI + ADDQ $0x03, R9 + CMPQ DI, BX + JAE lz4s_snappy_corrupt + INCQ DX + INCQ DI + TESTQ R8, R8 + JZ lz4s_snappy_lits_done + LEAQ (AX)(R8*1), R10 + CMPQ R10, CX + JAE lz4s_snappy_dstfull + ADDQ R8, SI + LEAL -1(R8), R10 + CMPL R10, $0x3c + JB one_byte_lz4s_snappy + CMPL R10, $0x00000100 + JB two_bytes_lz4s_snappy + CMPL R10, $0x00010000 + JB three_bytes_lz4s_snappy + CMPL R10, $0x01000000 + JB four_bytes_lz4s_snappy + MOVB $0xfc, (AX) + MOVL R10, 1(AX) + ADDQ $0x05, AX + JMP memmove_long_lz4s_snappy + +four_bytes_lz4s_snappy: + MOVL R10, R11 + SHRL $0x10, R11 + MOVB $0xf8, (AX) + MOVW R10, 1(AX) + MOVB R11, 3(AX) + ADDQ $0x04, AX + JMP memmove_long_lz4s_snappy + +three_bytes_lz4s_snappy: + MOVB $0xf4, (AX) + MOVW R10, 1(AX) + ADDQ $0x03, AX + JMP memmove_long_lz4s_snappy + +two_bytes_lz4s_snappy: + MOVB $0xf0, (AX) + MOVB R10, 1(AX) + ADDQ $0x02, AX + CMPL R10, $0x40 + JB memmove_lz4s_snappy + JMP memmove_long_lz4s_snappy + +one_byte_lz4s_snappy: + SHLB $0x02, R10 + MOVB R10, (AX) + ADDQ $0x01, AX + +memmove_lz4s_snappy: + LEAQ (AX)(R8*1), R10 + + // genMemMoveShort + CMPQ R8, $0x08 + JBE emit_lit_memmove_lz4s_snappy_memmove_move_8 + CMPQ R8, $0x10 + JBE emit_lit_memmove_lz4s_snappy_memmove_move_8through16 + CMPQ R8, $0x20 + JBE emit_lit_memmove_lz4s_snappy_memmove_move_17through32 + JMP emit_lit_memmove_lz4s_snappy_memmove_move_33through64 + +emit_lit_memmove_lz4s_snappy_memmove_move_8: + MOVQ (DX), R11 + MOVQ R11, (AX) + JMP memmove_end_copy_lz4s_snappy + +emit_lit_memmove_lz4s_snappy_memmove_move_8through16: + MOVQ (DX), R11 + MOVQ -8(DX)(R8*1), DX + MOVQ R11, (AX) + MOVQ DX, -8(AX)(R8*1) + JMP memmove_end_copy_lz4s_snappy + +emit_lit_memmove_lz4s_snappy_memmove_move_17through32: + MOVOU (DX), X0 + MOVOU -16(DX)(R8*1), X1 + MOVOU X0, (AX) + MOVOU X1, -16(AX)(R8*1) + JMP memmove_end_copy_lz4s_snappy + +emit_lit_memmove_lz4s_snappy_memmove_move_33through64: + MOVOU (DX), X0 + MOVOU 16(DX), X1 + MOVOU -32(DX)(R8*1), X2 + MOVOU -16(DX)(R8*1), X3 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + +memmove_end_copy_lz4s_snappy: + MOVQ R10, AX + JMP lz4s_snappy_lits_emit_done + +memmove_long_lz4s_snappy: + LEAQ (AX)(R8*1), R10 + + // genMemMoveLong + MOVOU (DX), X0 + MOVOU 16(DX), X1 + MOVOU -32(DX)(R8*1), X2 + MOVOU -16(DX)(R8*1), X3 + MOVQ R8, R12 + SHRQ $0x05, R12 + MOVQ AX, R11 + ANDL $0x0000001f, R11 + MOVQ $0x00000040, R13 + SUBQ R11, R13 + DECQ R12 + JA emit_lit_memmove_long_lz4s_snappylarge_forward_sse_loop_32 + LEAQ -32(DX)(R13*1), R11 + LEAQ -32(AX)(R13*1), R14 + +emit_lit_memmove_long_lz4s_snappylarge_big_loop_back: + MOVOU (R11), X4 + MOVOU 16(R11), X5 + MOVOA X4, (R14) + MOVOA X5, 16(R14) + ADDQ $0x20, R14 + ADDQ $0x20, R11 + ADDQ $0x20, R13 + DECQ R12 + JNA emit_lit_memmove_long_lz4s_snappylarge_big_loop_back + +emit_lit_memmove_long_lz4s_snappylarge_forward_sse_loop_32: + MOVOU -32(DX)(R13*1), X4 + MOVOU -16(DX)(R13*1), X5 + MOVOA X4, -32(AX)(R13*1) + MOVOA X5, -16(AX)(R13*1) + ADDQ $0x20, R13 + CMPQ R8, R13 + JAE emit_lit_memmove_long_lz4s_snappylarge_forward_sse_loop_32 + MOVOU X0, (AX) + MOVOU X1, 16(AX) + MOVOU X2, -32(AX)(R8*1) + MOVOU X3, -16(AX)(R8*1) + MOVQ R10, AX + +lz4s_snappy_lits_emit_done: + MOVQ DI, DX + +lz4s_snappy_lits_done: + CMPQ DX, BX + JNE lz4s_snappy_match + CMPQ R9, $0x03 + JEQ lz4s_snappy_done + JMP lz4s_snappy_corrupt + +lz4s_snappy_match: + CMPQ R9, $0x03 + JEQ lz4s_snappy_loop + LEAQ 2(DX), DI + CMPQ DI, BX + JAE lz4s_snappy_corrupt + MOVWQZX (DX), R8 + MOVQ DI, DX + TESTQ R8, R8 + JZ lz4s_snappy_corrupt + CMPQ R8, SI + JA lz4s_snappy_corrupt + CMPQ R9, $0x12 + JNE lz4s_snappy_ml_done + +lz4s_snappy_ml_loop: + MOVBQZX (DX), DI + INCQ DX + ADDQ DI, R9 + CMPQ DX, BX + JAE lz4s_snappy_corrupt + CMPQ DI, $0xff + JEQ lz4s_snappy_ml_loop + +lz4s_snappy_ml_done: + ADDQ R9, SI + + // emitCopy +two_byte_offset_lz4_s2: + CMPL R9, $0x40 + JBE two_byte_offset_short_lz4_s2 + MOVB $0xee, (AX) + MOVW R8, 1(AX) + LEAL -60(R9), R9 + ADDQ $0x03, AX + CMPQ AX, CX + JAE lz4s_snappy_loop + JMP two_byte_offset_lz4_s2 + +two_byte_offset_short_lz4_s2: + MOVL R9, DI + SHLL $0x02, DI + CMPL R9, $0x0c + JAE emit_copy_three_lz4_s2 + CMPL R8, $0x00000800 + JAE emit_copy_three_lz4_s2 + LEAL -15(DI), DI + MOVB R8, 1(AX) + SHRL $0x08, R8 + SHLL $0x05, R8 + ORL R8, DI + MOVB DI, (AX) + ADDQ $0x02, AX + JMP lz4s_snappy_loop + +emit_copy_three_lz4_s2: + LEAL -2(DI), DI + MOVB DI, (AX) + MOVW R8, 1(AX) + ADDQ $0x03, AX + JMP lz4s_snappy_loop + +lz4s_snappy_done: + MOVQ dst_base+0(FP), CX + SUBQ CX, AX + MOVQ SI, uncompressed+48(FP) + MOVQ AX, dstUsed+56(FP) + RET + +lz4s_snappy_corrupt: + XORQ AX, AX + LEAQ -1(AX), SI + MOVQ SI, uncompressed+48(FP) + RET + +lz4s_snappy_dstfull: + XORQ AX, AX + LEAQ -2(AX), SI + MOVQ SI, uncompressed+48(FP) + RET diff --git a/vendor/github.com/klauspost/compress/s2/index.go b/vendor/github.com/klauspost/compress/s2/index.go new file mode 100644 index 000000000..4229957b9 --- /dev/null +++ b/vendor/github.com/klauspost/compress/s2/index.go @@ -0,0 +1,602 @@ +// Copyright (c) 2022+ Klaus Post. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package s2 + +import ( + "bytes" + "encoding/binary" + "encoding/json" + "fmt" + "io" + "sort" +) + +const ( + S2IndexHeader = "s2idx\x00" + S2IndexTrailer = "\x00xdi2s" + maxIndexEntries = 1 << 16 + // If distance is less than this, we do not add the entry. + minIndexDist = 1 << 20 +) + +// Index represents an S2/Snappy index. +type Index struct { + TotalUncompressed int64 // Total Uncompressed size if known. Will be -1 if unknown. + TotalCompressed int64 // Total Compressed size if known. Will be -1 if unknown. + info []struct { + compressedOffset int64 + uncompressedOffset int64 + } + estBlockUncomp int64 +} + +func (i *Index) reset(maxBlock int) { + i.estBlockUncomp = int64(maxBlock) + i.TotalCompressed = -1 + i.TotalUncompressed = -1 + if len(i.info) > 0 { + i.info = i.info[:0] + } +} + +// allocInfos will allocate an empty slice of infos. +func (i *Index) allocInfos(n int) { + if n > maxIndexEntries { + panic("n > maxIndexEntries") + } + i.info = make([]struct { + compressedOffset int64 + uncompressedOffset int64 + }, 0, n) +} + +// add an uncompressed and compressed pair. +// Entries must be sent in order. +func (i *Index) add(compressedOffset, uncompressedOffset int64) error { + if i == nil { + return nil + } + lastIdx := len(i.info) - 1 + if lastIdx >= 0 { + latest := i.info[lastIdx] + if latest.uncompressedOffset == uncompressedOffset { + // Uncompressed didn't change, don't add entry, + // but update start index. + latest.compressedOffset = compressedOffset + i.info[lastIdx] = latest + return nil + } + if latest.uncompressedOffset > uncompressedOffset { + return fmt.Errorf("internal error: Earlier uncompressed received (%d > %d)", latest.uncompressedOffset, uncompressedOffset) + } + if latest.compressedOffset > compressedOffset { + return fmt.Errorf("internal error: Earlier compressed received (%d > %d)", latest.uncompressedOffset, uncompressedOffset) + } + if latest.uncompressedOffset+minIndexDist > uncompressedOffset { + // Only add entry if distance is large enough. + return nil + } + } + i.info = append(i.info, struct { + compressedOffset int64 + uncompressedOffset int64 + }{compressedOffset: compressedOffset, uncompressedOffset: uncompressedOffset}) + return nil +} + +// Find the offset at or before the wanted (uncompressed) offset. +// If offset is 0 or positive it is the offset from the beginning of the file. +// If the uncompressed size is known, the offset must be within the file. +// If an offset outside the file is requested io.ErrUnexpectedEOF is returned. +// If the offset is negative, it is interpreted as the distance from the end of the file, +// where -1 represents the last byte. +// If offset from the end of the file is requested, but size is unknown, +// ErrUnsupported will be returned. +func (i *Index) Find(offset int64) (compressedOff, uncompressedOff int64, err error) { + if i.TotalUncompressed < 0 { + return 0, 0, ErrCorrupt + } + if offset < 0 { + offset = i.TotalUncompressed + offset + if offset < 0 { + return 0, 0, io.ErrUnexpectedEOF + } + } + if offset > i.TotalUncompressed { + return 0, 0, io.ErrUnexpectedEOF + } + if len(i.info) > 200 { + n := sort.Search(len(i.info), func(n int) bool { + return i.info[n].uncompressedOffset > offset + }) + if n == 0 { + n = 1 + } + return i.info[n-1].compressedOffset, i.info[n-1].uncompressedOffset, nil + } + for _, info := range i.info { + if info.uncompressedOffset > offset { + break + } + compressedOff = info.compressedOffset + uncompressedOff = info.uncompressedOffset + } + return compressedOff, uncompressedOff, nil +} + +// reduce to stay below maxIndexEntries +func (i *Index) reduce() { + if len(i.info) < maxIndexEntries && i.estBlockUncomp >= minIndexDist { + return + } + + // Algorithm, keep 1, remove removeN entries... + removeN := (len(i.info) + 1) / maxIndexEntries + src := i.info + j := 0 + + // Each block should be at least 1MB, but don't reduce below 1000 entries. + for i.estBlockUncomp*(int64(removeN)+1) < minIndexDist && len(i.info)/(removeN+1) > 1000 { + removeN++ + } + for idx := 0; idx < len(src); idx++ { + i.info[j] = src[idx] + j++ + idx += removeN + } + i.info = i.info[:j] + // Update maxblock estimate. + i.estBlockUncomp += i.estBlockUncomp * int64(removeN) +} + +func (i *Index) appendTo(b []byte, uncompTotal, compTotal int64) []byte { + i.reduce() + var tmp [binary.MaxVarintLen64]byte + + initSize := len(b) + // We make the start a skippable header+size. + b = append(b, ChunkTypeIndex, 0, 0, 0) + b = append(b, []byte(S2IndexHeader)...) + // Total Uncompressed size + n := binary.PutVarint(tmp[:], uncompTotal) + b = append(b, tmp[:n]...) + // Total Compressed size + n = binary.PutVarint(tmp[:], compTotal) + b = append(b, tmp[:n]...) + // Put EstBlockUncomp size + n = binary.PutVarint(tmp[:], i.estBlockUncomp) + b = append(b, tmp[:n]...) + // Put length + n = binary.PutVarint(tmp[:], int64(len(i.info))) + b = append(b, tmp[:n]...) + + // Check if we should add uncompressed offsets + var hasUncompressed byte + for idx, info := range i.info { + if idx == 0 { + if info.uncompressedOffset != 0 { + hasUncompressed = 1 + break + } + continue + } + if info.uncompressedOffset != i.info[idx-1].uncompressedOffset+i.estBlockUncomp { + hasUncompressed = 1 + break + } + } + b = append(b, hasUncompressed) + + // Add each entry + if hasUncompressed == 1 { + for idx, info := range i.info { + uOff := info.uncompressedOffset + if idx > 0 { + prev := i.info[idx-1] + uOff -= prev.uncompressedOffset + (i.estBlockUncomp) + } + n = binary.PutVarint(tmp[:], uOff) + b = append(b, tmp[:n]...) + } + } + + // Initial compressed size estimate. + cPredict := i.estBlockUncomp / 2 + + for idx, info := range i.info { + cOff := info.compressedOffset + if idx > 0 { + prev := i.info[idx-1] + cOff -= prev.compressedOffset + cPredict + // Update compressed size prediction, with half the error. + cPredict += cOff / 2 + } + n = binary.PutVarint(tmp[:], cOff) + b = append(b, tmp[:n]...) + } + + // Add Total Size. + // Stored as fixed size for easier reading. + binary.LittleEndian.PutUint32(tmp[:], uint32(len(b)-initSize+4+len(S2IndexTrailer))) + b = append(b, tmp[:4]...) + // Trailer + b = append(b, []byte(S2IndexTrailer)...) + + // Update size + chunkLen := len(b) - initSize - skippableFrameHeader + b[initSize+1] = uint8(chunkLen >> 0) + b[initSize+2] = uint8(chunkLen >> 8) + b[initSize+3] = uint8(chunkLen >> 16) + //fmt.Printf("chunklen: 0x%x Uncomp:%d, Comp:%d\n", chunkLen, uncompTotal, compTotal) + return b +} + +// Load a binary index. +// A zero value Index can be used or a previous one can be reused. +func (i *Index) Load(b []byte) ([]byte, error) { + if len(b) <= 4+len(S2IndexHeader)+len(S2IndexTrailer) { + return b, io.ErrUnexpectedEOF + } + if b[0] != ChunkTypeIndex { + return b, ErrCorrupt + } + chunkLen := int(b[1]) | int(b[2])<<8 | int(b[3])<<16 + b = b[4:] + + // Validate we have enough... + if len(b) < chunkLen { + return b, io.ErrUnexpectedEOF + } + if !bytes.Equal(b[:len(S2IndexHeader)], []byte(S2IndexHeader)) { + return b, ErrUnsupported + } + b = b[len(S2IndexHeader):] + + // Total Uncompressed + if v, n := binary.Varint(b); n <= 0 || v < 0 { + return b, ErrCorrupt + } else { + i.TotalUncompressed = v + b = b[n:] + } + + // Total Compressed + if v, n := binary.Varint(b); n <= 0 { + return b, ErrCorrupt + } else { + i.TotalCompressed = v + b = b[n:] + } + + // Read EstBlockUncomp + if v, n := binary.Varint(b); n <= 0 { + return b, ErrCorrupt + } else { + if v < 0 { + return b, ErrCorrupt + } + i.estBlockUncomp = v + b = b[n:] + } + + var entries int + if v, n := binary.Varint(b); n <= 0 { + return b, ErrCorrupt + } else { + if v < 0 || v > maxIndexEntries { + return b, ErrCorrupt + } + entries = int(v) + b = b[n:] + } + if cap(i.info) < entries { + i.allocInfos(entries) + } + i.info = i.info[:entries] + + if len(b) < 1 { + return b, io.ErrUnexpectedEOF + } + hasUncompressed := b[0] + b = b[1:] + if hasUncompressed&1 != hasUncompressed { + return b, ErrCorrupt + } + + // Add each uncompressed entry + for idx := range i.info { + var uOff int64 + if hasUncompressed != 0 { + // Load delta + if v, n := binary.Varint(b); n <= 0 { + return b, ErrCorrupt + } else { + uOff = v + b = b[n:] + } + } + + if idx > 0 { + prev := i.info[idx-1].uncompressedOffset + uOff += prev + (i.estBlockUncomp) + if uOff <= prev { + return b, ErrCorrupt + } + } + if uOff < 0 { + return b, ErrCorrupt + } + i.info[idx].uncompressedOffset = uOff + } + + // Initial compressed size estimate. + cPredict := i.estBlockUncomp / 2 + + // Add each compressed entry + for idx := range i.info { + var cOff int64 + if v, n := binary.Varint(b); n <= 0 { + return b, ErrCorrupt + } else { + cOff = v + b = b[n:] + } + + if idx > 0 { + // Update compressed size prediction, with half the error. + cPredictNew := cPredict + cOff/2 + + prev := i.info[idx-1].compressedOffset + cOff += prev + cPredict + if cOff <= prev { + return b, ErrCorrupt + } + cPredict = cPredictNew + } + if cOff < 0 { + return b, ErrCorrupt + } + i.info[idx].compressedOffset = cOff + } + if len(b) < 4+len(S2IndexTrailer) { + return b, io.ErrUnexpectedEOF + } + // Skip size... + b = b[4:] + + // Check trailer... + if !bytes.Equal(b[:len(S2IndexTrailer)], []byte(S2IndexTrailer)) { + return b, ErrCorrupt + } + return b[len(S2IndexTrailer):], nil +} + +// LoadStream will load an index from the end of the supplied stream. +// ErrUnsupported will be returned if the signature cannot be found. +// ErrCorrupt will be returned if unexpected values are found. +// io.ErrUnexpectedEOF is returned if there are too few bytes. +// IO errors are returned as-is. +func (i *Index) LoadStream(rs io.ReadSeeker) error { + // Go to end. + _, err := rs.Seek(-10, io.SeekEnd) + if err != nil { + return err + } + var tmp [10]byte + _, err = io.ReadFull(rs, tmp[:]) + if err != nil { + return err + } + // Check trailer... + if !bytes.Equal(tmp[4:4+len(S2IndexTrailer)], []byte(S2IndexTrailer)) { + return ErrUnsupported + } + sz := binary.LittleEndian.Uint32(tmp[:4]) + if sz > maxChunkSize+skippableFrameHeader { + return ErrCorrupt + } + _, err = rs.Seek(-int64(sz), io.SeekEnd) + if err != nil { + return err + } + + // Read index. + buf := make([]byte, sz) + _, err = io.ReadFull(rs, buf) + if err != nil { + return err + } + _, err = i.Load(buf) + return err +} + +// IndexStream will return an index for a stream. +// The stream structure will be checked, but +// data within blocks is not verified. +// The returned index can either be appended to the end of the stream +// or stored separately. +func IndexStream(r io.Reader) ([]byte, error) { + var i Index + var buf [maxChunkSize]byte + var readHeader bool + for { + _, err := io.ReadFull(r, buf[:4]) + if err != nil { + if err == io.EOF { + return i.appendTo(nil, i.TotalUncompressed, i.TotalCompressed), nil + } + return nil, err + } + // Start of this chunk. + startChunk := i.TotalCompressed + i.TotalCompressed += 4 + + chunkType := buf[0] + if !readHeader { + if chunkType != chunkTypeStreamIdentifier { + return nil, ErrCorrupt + } + readHeader = true + } + chunkLen := int(buf[1]) | int(buf[2])<<8 | int(buf[3])<<16 + if chunkLen < checksumSize { + return nil, ErrCorrupt + } + + i.TotalCompressed += int64(chunkLen) + _, err = io.ReadFull(r, buf[:chunkLen]) + if err != nil { + return nil, io.ErrUnexpectedEOF + } + // The chunk types are specified at + // https://github.com/google/snappy/blob/master/framing_format.txt + switch chunkType { + case chunkTypeCompressedData: + // Section 4.2. Compressed data (chunk type 0x00). + // Skip checksum. + dLen, err := DecodedLen(buf[checksumSize:]) + if err != nil { + return nil, err + } + if dLen > maxBlockSize { + return nil, ErrCorrupt + } + if i.estBlockUncomp == 0 { + // Use first block for estimate... + i.estBlockUncomp = int64(dLen) + } + err = i.add(startChunk, i.TotalUncompressed) + if err != nil { + return nil, err + } + i.TotalUncompressed += int64(dLen) + continue + case chunkTypeUncompressedData: + n2 := chunkLen - checksumSize + if n2 > maxBlockSize { + return nil, ErrCorrupt + } + if i.estBlockUncomp == 0 { + // Use first block for estimate... + i.estBlockUncomp = int64(n2) + } + err = i.add(startChunk, i.TotalUncompressed) + if err != nil { + return nil, err + } + i.TotalUncompressed += int64(n2) + continue + case chunkTypeStreamIdentifier: + // Section 4.1. Stream identifier (chunk type 0xff). + if chunkLen != len(magicBody) { + return nil, ErrCorrupt + } + + if string(buf[:len(magicBody)]) != magicBody { + if string(buf[:len(magicBody)]) != magicBodySnappy { + return nil, ErrCorrupt + } + } + + continue + } + + if chunkType <= 0x7f { + // Section 4.5. Reserved unskippable chunks (chunk types 0x02-0x7f). + return nil, ErrUnsupported + } + if chunkLen > maxChunkSize { + return nil, ErrUnsupported + } + // Section 4.4 Padding (chunk type 0xfe). + // Section 4.6. Reserved skippable chunks (chunk types 0x80-0xfd). + } +} + +// JSON returns the index as JSON text. +func (i *Index) JSON() []byte { + type offset struct { + CompressedOffset int64 `json:"compressed"` + UncompressedOffset int64 `json:"uncompressed"` + } + x := struct { + TotalUncompressed int64 `json:"total_uncompressed"` // Total Uncompressed size if known. Will be -1 if unknown. + TotalCompressed int64 `json:"total_compressed"` // Total Compressed size if known. Will be -1 if unknown. + Offsets []offset `json:"offsets"` + EstBlockUncomp int64 `json:"est_block_uncompressed"` + }{ + TotalUncompressed: i.TotalUncompressed, + TotalCompressed: i.TotalCompressed, + EstBlockUncomp: i.estBlockUncomp, + } + for _, v := range i.info { + x.Offsets = append(x.Offsets, offset{CompressedOffset: v.compressedOffset, UncompressedOffset: v.uncompressedOffset}) + } + b, _ := json.MarshalIndent(x, "", " ") + return b +} + +// RemoveIndexHeaders will trim all headers and trailers from a given index. +// This is expected to save 20 bytes. +// These can be restored using RestoreIndexHeaders. +// This removes a layer of security, but is the most compact representation. +// Returns nil if headers contains errors. +// The returned slice references the provided slice. +func RemoveIndexHeaders(b []byte) []byte { + const save = 4 + len(S2IndexHeader) + len(S2IndexTrailer) + 4 + if len(b) <= save { + return nil + } + if b[0] != ChunkTypeIndex { + return nil + } + chunkLen := int(b[1]) | int(b[2])<<8 | int(b[3])<<16 + b = b[4:] + + // Validate we have enough... + if len(b) < chunkLen { + return nil + } + b = b[:chunkLen] + + if !bytes.Equal(b[:len(S2IndexHeader)], []byte(S2IndexHeader)) { + return nil + } + b = b[len(S2IndexHeader):] + if !bytes.HasSuffix(b, []byte(S2IndexTrailer)) { + return nil + } + b = bytes.TrimSuffix(b, []byte(S2IndexTrailer)) + + if len(b) < 4 { + return nil + } + return b[:len(b)-4] +} + +// RestoreIndexHeaders will index restore headers removed by RemoveIndexHeaders. +// No error checking is performed on the input. +// If a 0 length slice is sent, it is returned without modification. +func RestoreIndexHeaders(in []byte) []byte { + if len(in) == 0 { + return in + } + b := make([]byte, 0, 4+len(S2IndexHeader)+len(in)+len(S2IndexTrailer)+4) + b = append(b, ChunkTypeIndex, 0, 0, 0) + b = append(b, []byte(S2IndexHeader)...) + b = append(b, in...) + + var tmp [4]byte + binary.LittleEndian.PutUint32(tmp[:], uint32(len(b)+4+len(S2IndexTrailer))) + b = append(b, tmp[:4]...) + // Trailer + b = append(b, []byte(S2IndexTrailer)...) + + chunkLen := len(b) - skippableFrameHeader + b[1] = uint8(chunkLen >> 0) + b[2] = uint8(chunkLen >> 8) + b[3] = uint8(chunkLen >> 16) + return b +} diff --git a/vendor/github.com/klauspost/compress/s2/lz4convert.go b/vendor/github.com/klauspost/compress/s2/lz4convert.go new file mode 100644 index 000000000..46ed908e3 --- /dev/null +++ b/vendor/github.com/klauspost/compress/s2/lz4convert.go @@ -0,0 +1,585 @@ +// Copyright (c) 2022 Klaus Post. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package s2 + +import ( + "encoding/binary" + "errors" + "fmt" +) + +// LZ4Converter provides conversion from LZ4 blocks as defined here: +// https://github.com/lz4/lz4/blob/dev/doc/lz4_Block_format.md +type LZ4Converter struct { +} + +// ErrDstTooSmall is returned when provided destination is too small. +var ErrDstTooSmall = errors.New("s2: destination too small") + +// ConvertBlock will convert an LZ4 block and append it as an S2 +// block without block length to dst. +// The uncompressed size is returned as well. +// dst must have capacity to contain the entire compressed block. +func (l *LZ4Converter) ConvertBlock(dst, src []byte) ([]byte, int, error) { + if len(src) == 0 { + return dst, 0, nil + } + const debug = false + const inline = true + const lz4MinMatch = 4 + + s, d := 0, len(dst) + dst = dst[:cap(dst)] + if !debug && hasAmd64Asm { + res, sz := cvtLZ4BlockAsm(dst[d:], src) + if res < 0 { + const ( + errCorrupt = -1 + errDstTooSmall = -2 + ) + switch res { + case errCorrupt: + return nil, 0, ErrCorrupt + case errDstTooSmall: + return nil, 0, ErrDstTooSmall + default: + return nil, 0, fmt.Errorf("unexpected result: %d", res) + } + } + if d+sz > len(dst) { + return nil, 0, ErrDstTooSmall + } + return dst[:d+sz], res, nil + } + + dLimit := len(dst) - 10 + var lastOffset uint16 + var uncompressed int + if debug { + fmt.Printf("convert block start: len(src): %d, len(dst):%d \n", len(src), len(dst)) + } + + for { + if s >= len(src) { + return dst[:d], 0, ErrCorrupt + } + // Read literal info + token := src[s] + ll := int(token >> 4) + ml := int(lz4MinMatch + (token & 0xf)) + + // If upper nibble is 15, literal length is extended + if token >= 0xf0 { + for { + s++ + if s >= len(src) { + if debug { + fmt.Printf("error reading ll: s (%d) >= len(src) (%d)\n", s, len(src)) + } + return dst[:d], 0, ErrCorrupt + } + val := src[s] + ll += int(val) + if val != 255 { + break + } + } + } + // Skip past token + if s+ll >= len(src) { + if debug { + fmt.Printf("error literals: s+ll (%d+%d) >= len(src) (%d)\n", s, ll, len(src)) + } + return nil, 0, ErrCorrupt + } + s++ + if ll > 0 { + if d+ll > dLimit { + return nil, 0, ErrDstTooSmall + } + if debug { + fmt.Printf("emit %d literals\n", ll) + } + d += emitLiteralGo(dst[d:], src[s:s+ll]) + s += ll + uncompressed += ll + } + + // Check if we are done... + if s == len(src) && ml == lz4MinMatch { + break + } + // 2 byte offset + if s >= len(src)-2 { + if debug { + fmt.Printf("s (%d) >= len(src)-2 (%d)", s, len(src)-2) + } + return nil, 0, ErrCorrupt + } + offset := binary.LittleEndian.Uint16(src[s:]) + s += 2 + if offset == 0 { + if debug { + fmt.Printf("error: offset 0, ml: %d, len(src)-s: %d\n", ml, len(src)-s) + } + return nil, 0, ErrCorrupt + } + if int(offset) > uncompressed { + if debug { + fmt.Printf("error: offset (%d)> uncompressed (%d)\n", offset, uncompressed) + } + return nil, 0, ErrCorrupt + } + + if ml == lz4MinMatch+15 { + for { + if s >= len(src) { + if debug { + fmt.Printf("error reading ml: s (%d) >= len(src) (%d)\n", s, len(src)) + } + return nil, 0, ErrCorrupt + } + val := src[s] + s++ + ml += int(val) + if val != 255 { + if s >= len(src) { + if debug { + fmt.Printf("error reading ml: s (%d) >= len(src) (%d)\n", s, len(src)) + } + return nil, 0, ErrCorrupt + } + break + } + } + } + if offset == lastOffset { + if debug { + fmt.Printf("emit repeat, length: %d, offset: %d\n", ml, offset) + } + if !inline { + d += emitRepeat16(dst[d:], offset, ml) + } else { + length := ml + dst := dst[d:] + for len(dst) > 5 { + // Repeat offset, make length cheaper + length -= 4 + if length <= 4 { + dst[0] = uint8(length)<<2 | tagCopy1 + dst[1] = 0 + d += 2 + break + } + if length < 8 && offset < 2048 { + // Encode WITH offset + dst[1] = uint8(offset) + dst[0] = uint8(offset>>8)<<5 | uint8(length)<<2 | tagCopy1 + d += 2 + break + } + if length < (1<<8)+4 { + length -= 4 + dst[2] = uint8(length) + dst[1] = 0 + dst[0] = 5<<2 | tagCopy1 + d += 3 + break + } + if length < (1<<16)+(1<<8) { + length -= 1 << 8 + dst[3] = uint8(length >> 8) + dst[2] = uint8(length >> 0) + dst[1] = 0 + dst[0] = 6<<2 | tagCopy1 + d += 4 + break + } + const maxRepeat = (1 << 24) - 1 + length -= 1 << 16 + left := 0 + if length > maxRepeat { + left = length - maxRepeat + 4 + length = maxRepeat - 4 + } + dst[4] = uint8(length >> 16) + dst[3] = uint8(length >> 8) + dst[2] = uint8(length >> 0) + dst[1] = 0 + dst[0] = 7<<2 | tagCopy1 + if left > 0 { + d += 5 + emitRepeat16(dst[5:], offset, left) + break + } + d += 5 + break + } + } + } else { + if debug { + fmt.Printf("emit copy, length: %d, offset: %d\n", ml, offset) + } + if !inline { + d += emitCopy16(dst[d:], offset, ml) + } else { + length := ml + dst := dst[d:] + for len(dst) > 5 { + // Offset no more than 2 bytes. + if length > 64 { + off := 3 + if offset < 2048 { + // emit 8 bytes as tagCopy1, rest as repeats. + dst[1] = uint8(offset) + dst[0] = uint8(offset>>8)<<5 | uint8(8-4)<<2 | tagCopy1 + length -= 8 + off = 2 + } else { + // Emit a length 60 copy, encoded as 3 bytes. + // Emit remaining as repeat value (minimum 4 bytes). + dst[2] = uint8(offset >> 8) + dst[1] = uint8(offset) + dst[0] = 59<<2 | tagCopy2 + length -= 60 + } + // Emit remaining as repeats, at least 4 bytes remain. + d += off + emitRepeat16(dst[off:], offset, length) + break + } + if length >= 12 || offset >= 2048 { + // Emit the remaining copy, encoded as 3 bytes. + dst[2] = uint8(offset >> 8) + dst[1] = uint8(offset) + dst[0] = uint8(length-1)<<2 | tagCopy2 + d += 3 + break + } + // Emit the remaining copy, encoded as 2 bytes. + dst[1] = uint8(offset) + dst[0] = uint8(offset>>8)<<5 | uint8(length-4)<<2 | tagCopy1 + d += 2 + break + } + } + lastOffset = offset + } + uncompressed += ml + if d > dLimit { + return nil, 0, ErrDstTooSmall + } + } + + return dst[:d], uncompressed, nil +} + +// ConvertBlockSnappy will convert an LZ4 block and append it +// as a Snappy block without block length to dst. +// The uncompressed size is returned as well. +// dst must have capacity to contain the entire compressed block. +func (l *LZ4Converter) ConvertBlockSnappy(dst, src []byte) ([]byte, int, error) { + if len(src) == 0 { + return dst, 0, nil + } + const debug = false + const lz4MinMatch = 4 + + s, d := 0, len(dst) + dst = dst[:cap(dst)] + // Use assembly when possible + if !debug && hasAmd64Asm { + res, sz := cvtLZ4BlockSnappyAsm(dst[d:], src) + if res < 0 { + const ( + errCorrupt = -1 + errDstTooSmall = -2 + ) + switch res { + case errCorrupt: + return nil, 0, ErrCorrupt + case errDstTooSmall: + return nil, 0, ErrDstTooSmall + default: + return nil, 0, fmt.Errorf("unexpected result: %d", res) + } + } + if d+sz > len(dst) { + return nil, 0, ErrDstTooSmall + } + return dst[:d+sz], res, nil + } + + dLimit := len(dst) - 10 + var uncompressed int + if debug { + fmt.Printf("convert block start: len(src): %d, len(dst):%d \n", len(src), len(dst)) + } + + for { + if s >= len(src) { + return nil, 0, ErrCorrupt + } + // Read literal info + token := src[s] + ll := int(token >> 4) + ml := int(lz4MinMatch + (token & 0xf)) + + // If upper nibble is 15, literal length is extended + if token >= 0xf0 { + for { + s++ + if s >= len(src) { + if debug { + fmt.Printf("error reading ll: s (%d) >= len(src) (%d)\n", s, len(src)) + } + return nil, 0, ErrCorrupt + } + val := src[s] + ll += int(val) + if val != 255 { + break + } + } + } + // Skip past token + if s+ll >= len(src) { + if debug { + fmt.Printf("error literals: s+ll (%d+%d) >= len(src) (%d)\n", s, ll, len(src)) + } + return nil, 0, ErrCorrupt + } + s++ + if ll > 0 { + if d+ll > dLimit { + return nil, 0, ErrDstTooSmall + } + if debug { + fmt.Printf("emit %d literals\n", ll) + } + d += emitLiteralGo(dst[d:], src[s:s+ll]) + s += ll + uncompressed += ll + } + + // Check if we are done... + if s == len(src) && ml == lz4MinMatch { + break + } + // 2 byte offset + if s >= len(src)-2 { + if debug { + fmt.Printf("s (%d) >= len(src)-2 (%d)", s, len(src)-2) + } + return nil, 0, ErrCorrupt + } + offset := binary.LittleEndian.Uint16(src[s:]) + s += 2 + if offset == 0 { + if debug { + fmt.Printf("error: offset 0, ml: %d, len(src)-s: %d\n", ml, len(src)-s) + } + return nil, 0, ErrCorrupt + } + if int(offset) > uncompressed { + if debug { + fmt.Printf("error: offset (%d)> uncompressed (%d)\n", offset, uncompressed) + } + return nil, 0, ErrCorrupt + } + + if ml == lz4MinMatch+15 { + for { + if s >= len(src) { + if debug { + fmt.Printf("error reading ml: s (%d) >= len(src) (%d)\n", s, len(src)) + } + return nil, 0, ErrCorrupt + } + val := src[s] + s++ + ml += int(val) + if val != 255 { + if s >= len(src) { + if debug { + fmt.Printf("error reading ml: s (%d) >= len(src) (%d)\n", s, len(src)) + } + return nil, 0, ErrCorrupt + } + break + } + } + } + if debug { + fmt.Printf("emit copy, length: %d, offset: %d\n", ml, offset) + } + length := ml + // d += emitCopyNoRepeat(dst[d:], int(offset), ml) + for length > 0 { + if d >= dLimit { + return nil, 0, ErrDstTooSmall + } + + // Offset no more than 2 bytes. + if length > 64 { + // Emit a length 64 copy, encoded as 3 bytes. + dst[d+2] = uint8(offset >> 8) + dst[d+1] = uint8(offset) + dst[d+0] = 63<<2 | tagCopy2 + length -= 64 + d += 3 + continue + } + if length >= 12 || offset >= 2048 || length < 4 { + // Emit the remaining copy, encoded as 3 bytes. + dst[d+2] = uint8(offset >> 8) + dst[d+1] = uint8(offset) + dst[d+0] = uint8(length-1)<<2 | tagCopy2 + d += 3 + break + } + // Emit the remaining copy, encoded as 2 bytes. + dst[d+1] = uint8(offset) + dst[d+0] = uint8(offset>>8)<<5 | uint8(length-4)<<2 | tagCopy1 + d += 2 + break + } + uncompressed += ml + if d > dLimit { + return nil, 0, ErrDstTooSmall + } + } + + return dst[:d], uncompressed, nil +} + +// emitRepeat writes a repeat chunk and returns the number of bytes written. +// Length must be at least 4 and < 1<<24 +func emitRepeat16(dst []byte, offset uint16, length int) int { + // Repeat offset, make length cheaper + length -= 4 + if length <= 4 { + dst[0] = uint8(length)<<2 | tagCopy1 + dst[1] = 0 + return 2 + } + if length < 8 && offset < 2048 { + // Encode WITH offset + dst[1] = uint8(offset) + dst[0] = uint8(offset>>8)<<5 | uint8(length)<<2 | tagCopy1 + return 2 + } + if length < (1<<8)+4 { + length -= 4 + dst[2] = uint8(length) + dst[1] = 0 + dst[0] = 5<<2 | tagCopy1 + return 3 + } + if length < (1<<16)+(1<<8) { + length -= 1 << 8 + dst[3] = uint8(length >> 8) + dst[2] = uint8(length >> 0) + dst[1] = 0 + dst[0] = 6<<2 | tagCopy1 + return 4 + } + const maxRepeat = (1 << 24) - 1 + length -= 1 << 16 + left := 0 + if length > maxRepeat { + left = length - maxRepeat + 4 + length = maxRepeat - 4 + } + dst[4] = uint8(length >> 16) + dst[3] = uint8(length >> 8) + dst[2] = uint8(length >> 0) + dst[1] = 0 + dst[0] = 7<<2 | tagCopy1 + if left > 0 { + return 5 + emitRepeat16(dst[5:], offset, left) + } + return 5 +} + +// emitCopy writes a copy chunk and returns the number of bytes written. +// +// It assumes that: +// +// dst is long enough to hold the encoded bytes +// 1 <= offset && offset <= math.MaxUint16 +// 4 <= length && length <= math.MaxUint32 +func emitCopy16(dst []byte, offset uint16, length int) int { + // Offset no more than 2 bytes. + if length > 64 { + off := 3 + if offset < 2048 { + // emit 8 bytes as tagCopy1, rest as repeats. + dst[1] = uint8(offset) + dst[0] = uint8(offset>>8)<<5 | uint8(8-4)<<2 | tagCopy1 + length -= 8 + off = 2 + } else { + // Emit a length 60 copy, encoded as 3 bytes. + // Emit remaining as repeat value (minimum 4 bytes). + dst[2] = uint8(offset >> 8) + dst[1] = uint8(offset) + dst[0] = 59<<2 | tagCopy2 + length -= 60 + } + // Emit remaining as repeats, at least 4 bytes remain. + return off + emitRepeat16(dst[off:], offset, length) + } + if length >= 12 || offset >= 2048 { + // Emit the remaining copy, encoded as 3 bytes. + dst[2] = uint8(offset >> 8) + dst[1] = uint8(offset) + dst[0] = uint8(length-1)<<2 | tagCopy2 + return 3 + } + // Emit the remaining copy, encoded as 2 bytes. + dst[1] = uint8(offset) + dst[0] = uint8(offset>>8)<<5 | uint8(length-4)<<2 | tagCopy1 + return 2 +} + +// emitLiteral writes a literal chunk and returns the number of bytes written. +// +// It assumes that: +// +// dst is long enough to hold the encoded bytes +// 0 <= len(lit) && len(lit) <= math.MaxUint32 +func emitLiteralGo(dst, lit []byte) int { + if len(lit) == 0 { + return 0 + } + i, n := 0, uint(len(lit)-1) + switch { + case n < 60: + dst[0] = uint8(n)<<2 | tagLiteral + i = 1 + case n < 1<<8: + dst[1] = uint8(n) + dst[0] = 60<<2 | tagLiteral + i = 2 + case n < 1<<16: + dst[2] = uint8(n >> 8) + dst[1] = uint8(n) + dst[0] = 61<<2 | tagLiteral + i = 3 + case n < 1<<24: + dst[3] = uint8(n >> 16) + dst[2] = uint8(n >> 8) + dst[1] = uint8(n) + dst[0] = 62<<2 | tagLiteral + i = 4 + default: + dst[4] = uint8(n >> 24) + dst[3] = uint8(n >> 16) + dst[2] = uint8(n >> 8) + dst[1] = uint8(n) + dst[0] = 63<<2 | tagLiteral + i = 5 + } + return i + copy(dst[i:], lit) +} diff --git a/vendor/github.com/klauspost/compress/s2/lz4sconvert.go b/vendor/github.com/klauspost/compress/s2/lz4sconvert.go new file mode 100644 index 000000000..000f39719 --- /dev/null +++ b/vendor/github.com/klauspost/compress/s2/lz4sconvert.go @@ -0,0 +1,467 @@ +// Copyright (c) 2022 Klaus Post. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package s2 + +import ( + "encoding/binary" + "fmt" +) + +// LZ4sConverter provides conversion from LZ4s. +// (Intel modified LZ4 Blocks) +// https://cdrdv2-public.intel.com/743912/743912-qat-programmers-guide-v2.0.pdf +// LZ4s is a variant of LZ4 block format. LZ4s should be considered as an intermediate compressed block format. +// The LZ4s format is selected when the application sets the compType to CPA_DC_LZ4S in CpaDcSessionSetupData. +// The LZ4s block returned by the Intel® QAT hardware can be used by an external +// software post-processing to generate other compressed data formats. +// The following table lists the differences between LZ4 and LZ4s block format. LZ4s block format uses +// the same high-level formatting as LZ4 block format with the following encoding changes: +// For Min Match of 4 bytes, Copy length value 1-15 means length 4-18 with 18 bytes adding an extra byte. +// ONLY "Min match of 4 bytes" is supported. +type LZ4sConverter struct { +} + +// ConvertBlock will convert an LZ4s block and append it as an S2 +// block without block length to dst. +// The uncompressed size is returned as well. +// dst must have capacity to contain the entire compressed block. +func (l *LZ4sConverter) ConvertBlock(dst, src []byte) ([]byte, int, error) { + if len(src) == 0 { + return dst, 0, nil + } + const debug = false + const inline = true + const lz4MinMatch = 3 + + s, d := 0, len(dst) + dst = dst[:cap(dst)] + if !debug && hasAmd64Asm { + res, sz := cvtLZ4sBlockAsm(dst[d:], src) + if res < 0 { + const ( + errCorrupt = -1 + errDstTooSmall = -2 + ) + switch res { + case errCorrupt: + return nil, 0, ErrCorrupt + case errDstTooSmall: + return nil, 0, ErrDstTooSmall + default: + return nil, 0, fmt.Errorf("unexpected result: %d", res) + } + } + if d+sz > len(dst) { + return nil, 0, ErrDstTooSmall + } + return dst[:d+sz], res, nil + } + + dLimit := len(dst) - 10 + var lastOffset uint16 + var uncompressed int + if debug { + fmt.Printf("convert block start: len(src): %d, len(dst):%d \n", len(src), len(dst)) + } + + for { + if s >= len(src) { + return dst[:d], 0, ErrCorrupt + } + // Read literal info + token := src[s] + ll := int(token >> 4) + ml := int(lz4MinMatch + (token & 0xf)) + + // If upper nibble is 15, literal length is extended + if token >= 0xf0 { + for { + s++ + if s >= len(src) { + if debug { + fmt.Printf("error reading ll: s (%d) >= len(src) (%d)\n", s, len(src)) + } + return dst[:d], 0, ErrCorrupt + } + val := src[s] + ll += int(val) + if val != 255 { + break + } + } + } + // Skip past token + if s+ll >= len(src) { + if debug { + fmt.Printf("error literals: s+ll (%d+%d) >= len(src) (%d)\n", s, ll, len(src)) + } + return nil, 0, ErrCorrupt + } + s++ + if ll > 0 { + if d+ll > dLimit { + return nil, 0, ErrDstTooSmall + } + if debug { + fmt.Printf("emit %d literals\n", ll) + } + d += emitLiteralGo(dst[d:], src[s:s+ll]) + s += ll + uncompressed += ll + } + + // Check if we are done... + if ml == lz4MinMatch { + if s == len(src) { + break + } + // 0 bytes. + continue + } + // 2 byte offset + if s >= len(src)-2 { + if debug { + fmt.Printf("s (%d) >= len(src)-2 (%d)", s, len(src)-2) + } + return nil, 0, ErrCorrupt + } + offset := binary.LittleEndian.Uint16(src[s:]) + s += 2 + if offset == 0 { + if debug { + fmt.Printf("error: offset 0, ml: %d, len(src)-s: %d\n", ml, len(src)-s) + } + return nil, 0, ErrCorrupt + } + if int(offset) > uncompressed { + if debug { + fmt.Printf("error: offset (%d)> uncompressed (%d)\n", offset, uncompressed) + } + return nil, 0, ErrCorrupt + } + + if ml == lz4MinMatch+15 { + for { + if s >= len(src) { + if debug { + fmt.Printf("error reading ml: s (%d) >= len(src) (%d)\n", s, len(src)) + } + return nil, 0, ErrCorrupt + } + val := src[s] + s++ + ml += int(val) + if val != 255 { + if s >= len(src) { + if debug { + fmt.Printf("error reading ml: s (%d) >= len(src) (%d)\n", s, len(src)) + } + return nil, 0, ErrCorrupt + } + break + } + } + } + if offset == lastOffset { + if debug { + fmt.Printf("emit repeat, length: %d, offset: %d\n", ml, offset) + } + if !inline { + d += emitRepeat16(dst[d:], offset, ml) + } else { + length := ml + dst := dst[d:] + for len(dst) > 5 { + // Repeat offset, make length cheaper + length -= 4 + if length <= 4 { + dst[0] = uint8(length)<<2 | tagCopy1 + dst[1] = 0 + d += 2 + break + } + if length < 8 && offset < 2048 { + // Encode WITH offset + dst[1] = uint8(offset) + dst[0] = uint8(offset>>8)<<5 | uint8(length)<<2 | tagCopy1 + d += 2 + break + } + if length < (1<<8)+4 { + length -= 4 + dst[2] = uint8(length) + dst[1] = 0 + dst[0] = 5<<2 | tagCopy1 + d += 3 + break + } + if length < (1<<16)+(1<<8) { + length -= 1 << 8 + dst[3] = uint8(length >> 8) + dst[2] = uint8(length >> 0) + dst[1] = 0 + dst[0] = 6<<2 | tagCopy1 + d += 4 + break + } + const maxRepeat = (1 << 24) - 1 + length -= 1 << 16 + left := 0 + if length > maxRepeat { + left = length - maxRepeat + 4 + length = maxRepeat - 4 + } + dst[4] = uint8(length >> 16) + dst[3] = uint8(length >> 8) + dst[2] = uint8(length >> 0) + dst[1] = 0 + dst[0] = 7<<2 | tagCopy1 + if left > 0 { + d += 5 + emitRepeat16(dst[5:], offset, left) + break + } + d += 5 + break + } + } + } else { + if debug { + fmt.Printf("emit copy, length: %d, offset: %d\n", ml, offset) + } + if !inline { + d += emitCopy16(dst[d:], offset, ml) + } else { + length := ml + dst := dst[d:] + for len(dst) > 5 { + // Offset no more than 2 bytes. + if length > 64 { + off := 3 + if offset < 2048 { + // emit 8 bytes as tagCopy1, rest as repeats. + dst[1] = uint8(offset) + dst[0] = uint8(offset>>8)<<5 | uint8(8-4)<<2 | tagCopy1 + length -= 8 + off = 2 + } else { + // Emit a length 60 copy, encoded as 3 bytes. + // Emit remaining as repeat value (minimum 4 bytes). + dst[2] = uint8(offset >> 8) + dst[1] = uint8(offset) + dst[0] = 59<<2 | tagCopy2 + length -= 60 + } + // Emit remaining as repeats, at least 4 bytes remain. + d += off + emitRepeat16(dst[off:], offset, length) + break + } + if length >= 12 || offset >= 2048 { + // Emit the remaining copy, encoded as 3 bytes. + dst[2] = uint8(offset >> 8) + dst[1] = uint8(offset) + dst[0] = uint8(length-1)<<2 | tagCopy2 + d += 3 + break + } + // Emit the remaining copy, encoded as 2 bytes. + dst[1] = uint8(offset) + dst[0] = uint8(offset>>8)<<5 | uint8(length-4)<<2 | tagCopy1 + d += 2 + break + } + } + lastOffset = offset + } + uncompressed += ml + if d > dLimit { + return nil, 0, ErrDstTooSmall + } + } + + return dst[:d], uncompressed, nil +} + +// ConvertBlockSnappy will convert an LZ4s block and append it +// as a Snappy block without block length to dst. +// The uncompressed size is returned as well. +// dst must have capacity to contain the entire compressed block. +func (l *LZ4sConverter) ConvertBlockSnappy(dst, src []byte) ([]byte, int, error) { + if len(src) == 0 { + return dst, 0, nil + } + const debug = false + const lz4MinMatch = 3 + + s, d := 0, len(dst) + dst = dst[:cap(dst)] + // Use assembly when possible + if !debug && hasAmd64Asm { + res, sz := cvtLZ4sBlockSnappyAsm(dst[d:], src) + if res < 0 { + const ( + errCorrupt = -1 + errDstTooSmall = -2 + ) + switch res { + case errCorrupt: + return nil, 0, ErrCorrupt + case errDstTooSmall: + return nil, 0, ErrDstTooSmall + default: + return nil, 0, fmt.Errorf("unexpected result: %d", res) + } + } + if d+sz > len(dst) { + return nil, 0, ErrDstTooSmall + } + return dst[:d+sz], res, nil + } + + dLimit := len(dst) - 10 + var uncompressed int + if debug { + fmt.Printf("convert block start: len(src): %d, len(dst):%d \n", len(src), len(dst)) + } + + for { + if s >= len(src) { + return nil, 0, ErrCorrupt + } + // Read literal info + token := src[s] + ll := int(token >> 4) + ml := int(lz4MinMatch + (token & 0xf)) + + // If upper nibble is 15, literal length is extended + if token >= 0xf0 { + for { + s++ + if s >= len(src) { + if debug { + fmt.Printf("error reading ll: s (%d) >= len(src) (%d)\n", s, len(src)) + } + return nil, 0, ErrCorrupt + } + val := src[s] + ll += int(val) + if val != 255 { + break + } + } + } + // Skip past token + if s+ll >= len(src) { + if debug { + fmt.Printf("error literals: s+ll (%d+%d) >= len(src) (%d)\n", s, ll, len(src)) + } + return nil, 0, ErrCorrupt + } + s++ + if ll > 0 { + if d+ll > dLimit { + return nil, 0, ErrDstTooSmall + } + if debug { + fmt.Printf("emit %d literals\n", ll) + } + d += emitLiteralGo(dst[d:], src[s:s+ll]) + s += ll + uncompressed += ll + } + + // Check if we are done... + if ml == lz4MinMatch { + if s == len(src) { + break + } + // 0 bytes. + continue + } + // 2 byte offset + if s >= len(src)-2 { + if debug { + fmt.Printf("s (%d) >= len(src)-2 (%d)", s, len(src)-2) + } + return nil, 0, ErrCorrupt + } + offset := binary.LittleEndian.Uint16(src[s:]) + s += 2 + if offset == 0 { + if debug { + fmt.Printf("error: offset 0, ml: %d, len(src)-s: %d\n", ml, len(src)-s) + } + return nil, 0, ErrCorrupt + } + if int(offset) > uncompressed { + if debug { + fmt.Printf("error: offset (%d)> uncompressed (%d)\n", offset, uncompressed) + } + return nil, 0, ErrCorrupt + } + + if ml == lz4MinMatch+15 { + for { + if s >= len(src) { + if debug { + fmt.Printf("error reading ml: s (%d) >= len(src) (%d)\n", s, len(src)) + } + return nil, 0, ErrCorrupt + } + val := src[s] + s++ + ml += int(val) + if val != 255 { + if s >= len(src) { + if debug { + fmt.Printf("error reading ml: s (%d) >= len(src) (%d)\n", s, len(src)) + } + return nil, 0, ErrCorrupt + } + break + } + } + } + if debug { + fmt.Printf("emit copy, length: %d, offset: %d\n", ml, offset) + } + length := ml + // d += emitCopyNoRepeat(dst[d:], int(offset), ml) + for length > 0 { + if d >= dLimit { + return nil, 0, ErrDstTooSmall + } + + // Offset no more than 2 bytes. + if length > 64 { + // Emit a length 64 copy, encoded as 3 bytes. + dst[d+2] = uint8(offset >> 8) + dst[d+1] = uint8(offset) + dst[d+0] = 63<<2 | tagCopy2 + length -= 64 + d += 3 + continue + } + if length >= 12 || offset >= 2048 || length < 4 { + // Emit the remaining copy, encoded as 3 bytes. + dst[d+2] = uint8(offset >> 8) + dst[d+1] = uint8(offset) + dst[d+0] = uint8(length-1)<<2 | tagCopy2 + d += 3 + break + } + // Emit the remaining copy, encoded as 2 bytes. + dst[d+1] = uint8(offset) + dst[d+0] = uint8(offset>>8)<<5 | uint8(length-4)<<2 | tagCopy1 + d += 2 + break + } + uncompressed += ml + if d > dLimit { + return nil, 0, ErrDstTooSmall + } + } + + return dst[:d], uncompressed, nil +} diff --git a/vendor/github.com/klauspost/compress/s2/reader.go b/vendor/github.com/klauspost/compress/s2/reader.go new file mode 100644 index 000000000..8372d752f --- /dev/null +++ b/vendor/github.com/klauspost/compress/s2/reader.go @@ -0,0 +1,1075 @@ +// Copyright 2011 The Snappy-Go Authors. All rights reserved. +// Copyright (c) 2019+ Klaus Post. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package s2 + +import ( + "errors" + "fmt" + "io" + "io/ioutil" + "math" + "runtime" + "sync" +) + +// ErrCantSeek is returned if the stream cannot be seeked. +type ErrCantSeek struct { + Reason string +} + +// Error returns the error as string. +func (e ErrCantSeek) Error() string { + return fmt.Sprintf("s2: Can't seek because %s", e.Reason) +} + +// NewReader returns a new Reader that decompresses from r, using the framing +// format described at +// https://github.com/google/snappy/blob/master/framing_format.txt with S2 changes. +func NewReader(r io.Reader, opts ...ReaderOption) *Reader { + nr := Reader{ + r: r, + maxBlock: maxBlockSize, + } + for _, opt := range opts { + if err := opt(&nr); err != nil { + nr.err = err + return &nr + } + } + nr.maxBufSize = MaxEncodedLen(nr.maxBlock) + checksumSize + if nr.lazyBuf > 0 { + nr.buf = make([]byte, MaxEncodedLen(nr.lazyBuf)+checksumSize) + } else { + nr.buf = make([]byte, MaxEncodedLen(defaultBlockSize)+checksumSize) + } + nr.readHeader = nr.ignoreStreamID + nr.paramsOK = true + return &nr +} + +// ReaderOption is an option for creating a decoder. +type ReaderOption func(*Reader) error + +// ReaderMaxBlockSize allows to control allocations if the stream +// has been compressed with a smaller WriterBlockSize, or with the default 1MB. +// Blocks must be this size or smaller to decompress, +// otherwise the decoder will return ErrUnsupported. +// +// For streams compressed with Snappy this can safely be set to 64KB (64 << 10). +// +// Default is the maximum limit of 4MB. +func ReaderMaxBlockSize(blockSize int) ReaderOption { + return func(r *Reader) error { + if blockSize > maxBlockSize || blockSize <= 0 { + return errors.New("s2: block size too large. Must be <= 4MB and > 0") + } + if r.lazyBuf == 0 && blockSize < defaultBlockSize { + r.lazyBuf = blockSize + } + r.maxBlock = blockSize + return nil + } +} + +// ReaderAllocBlock allows to control upfront stream allocations +// and not allocate for frames bigger than this initially. +// If frames bigger than this is seen a bigger buffer will be allocated. +// +// Default is 1MB, which is default output size. +func ReaderAllocBlock(blockSize int) ReaderOption { + return func(r *Reader) error { + if blockSize > maxBlockSize || blockSize < 1024 { + return errors.New("s2: invalid ReaderAllocBlock. Must be <= 4MB and >= 1024") + } + r.lazyBuf = blockSize + return nil + } +} + +// ReaderIgnoreStreamIdentifier will make the reader skip the expected +// stream identifier at the beginning of the stream. +// This can be used when serving a stream that has been forwarded to a specific point. +func ReaderIgnoreStreamIdentifier() ReaderOption { + return func(r *Reader) error { + r.ignoreStreamID = true + return nil + } +} + +// ReaderSkippableCB will register a callback for chuncks with the specified ID. +// ID must be a Reserved skippable chunks ID, 0x80-0xfd (inclusive). +// For each chunk with the ID, the callback is called with the content. +// Any returned non-nil error will abort decompression. +// Only one callback per ID is supported, latest sent will be used. +// You can peek the stream, triggering the callback, by doing a Read with a 0 +// byte buffer. +func ReaderSkippableCB(id uint8, fn func(r io.Reader) error) ReaderOption { + return func(r *Reader) error { + if id < 0x80 || id > 0xfd { + return fmt.Errorf("ReaderSkippableCB: Invalid id provided, must be 0x80-0xfd (inclusive)") + } + r.skippableCB[id-0x80] = fn + return nil + } +} + +// ReaderIgnoreCRC will make the reader skip CRC calculation and checks. +func ReaderIgnoreCRC() ReaderOption { + return func(r *Reader) error { + r.ignoreCRC = true + return nil + } +} + +// Reader is an io.Reader that can read Snappy-compressed bytes. +type Reader struct { + r io.Reader + err error + decoded []byte + buf []byte + skippableCB [0xff - 0x80]func(r io.Reader) error + blockStart int64 // Uncompressed offset at start of current. + index *Index + + // decoded[i:j] contains decoded bytes that have not yet been passed on. + i, j int + // maximum block size allowed. + maxBlock int + // maximum expected buffer size. + maxBufSize int + // alloc a buffer this size if > 0. + lazyBuf int + readHeader bool + paramsOK bool + snappyFrame bool + ignoreStreamID bool + ignoreCRC bool +} + +// GetBufferCapacity returns the capacity of the internal buffer. +// This might be useful to know when reusing the same reader in combination +// with the lazy buffer option. +func (r *Reader) GetBufferCapacity() int { + return cap(r.buf) +} + +// ensureBufferSize will ensure that the buffer can take at least n bytes. +// If false is returned the buffer exceeds maximum allowed size. +func (r *Reader) ensureBufferSize(n int) bool { + if n > r.maxBufSize { + r.err = ErrCorrupt + return false + } + if cap(r.buf) >= n { + return true + } + // Realloc buffer. + r.buf = make([]byte, n) + return true +} + +// Reset discards any buffered data, resets all state, and switches the Snappy +// reader to read from r. This permits reusing a Reader rather than allocating +// a new one. +func (r *Reader) Reset(reader io.Reader) { + if !r.paramsOK { + return + } + r.index = nil + r.r = reader + r.err = nil + r.i = 0 + r.j = 0 + r.blockStart = 0 + r.readHeader = r.ignoreStreamID +} + +func (r *Reader) readFull(p []byte, allowEOF bool) (ok bool) { + if _, r.err = io.ReadFull(r.r, p); r.err != nil { + if r.err == io.ErrUnexpectedEOF || (r.err == io.EOF && !allowEOF) { + r.err = ErrCorrupt + } + return false + } + return true +} + +// skippable will skip n bytes. +// If the supplied reader supports seeking that is used. +// tmp is used as a temporary buffer for reading. +// The supplied slice does not need to be the size of the read. +func (r *Reader) skippable(tmp []byte, n int, allowEOF bool, id uint8) (ok bool) { + if id < 0x80 { + r.err = fmt.Errorf("internal error: skippable id < 0x80") + return false + } + if fn := r.skippableCB[id-0x80]; fn != nil { + rd := io.LimitReader(r.r, int64(n)) + r.err = fn(rd) + if r.err != nil { + return false + } + _, r.err = io.CopyBuffer(ioutil.Discard, rd, tmp) + return r.err == nil + } + if rs, ok := r.r.(io.ReadSeeker); ok { + _, err := rs.Seek(int64(n), io.SeekCurrent) + if err == nil { + return true + } + if err == io.ErrUnexpectedEOF || (r.err == io.EOF && !allowEOF) { + r.err = ErrCorrupt + return false + } + } + for n > 0 { + if n < len(tmp) { + tmp = tmp[:n] + } + if _, r.err = io.ReadFull(r.r, tmp); r.err != nil { + if r.err == io.ErrUnexpectedEOF || (r.err == io.EOF && !allowEOF) { + r.err = ErrCorrupt + } + return false + } + n -= len(tmp) + } + return true +} + +// Read satisfies the io.Reader interface. +func (r *Reader) Read(p []byte) (int, error) { + if r.err != nil { + return 0, r.err + } + for { + if r.i < r.j { + n := copy(p, r.decoded[r.i:r.j]) + r.i += n + return n, nil + } + if !r.readFull(r.buf[:4], true) { + return 0, r.err + } + chunkType := r.buf[0] + if !r.readHeader { + if chunkType != chunkTypeStreamIdentifier { + r.err = ErrCorrupt + return 0, r.err + } + r.readHeader = true + } + chunkLen := int(r.buf[1]) | int(r.buf[2])<<8 | int(r.buf[3])<<16 + + // The chunk types are specified at + // https://github.com/google/snappy/blob/master/framing_format.txt + switch chunkType { + case chunkTypeCompressedData: + r.blockStart += int64(r.j) + // Section 4.2. Compressed data (chunk type 0x00). + if chunkLen < checksumSize { + r.err = ErrCorrupt + return 0, r.err + } + if !r.ensureBufferSize(chunkLen) { + if r.err == nil { + r.err = ErrUnsupported + } + return 0, r.err + } + buf := r.buf[:chunkLen] + if !r.readFull(buf, false) { + return 0, r.err + } + checksum := uint32(buf[0]) | uint32(buf[1])<<8 | uint32(buf[2])<<16 | uint32(buf[3])<<24 + buf = buf[checksumSize:] + + n, err := DecodedLen(buf) + if err != nil { + r.err = err + return 0, r.err + } + if r.snappyFrame && n > maxSnappyBlockSize { + r.err = ErrCorrupt + return 0, r.err + } + + if n > len(r.decoded) { + if n > r.maxBlock { + r.err = ErrCorrupt + return 0, r.err + } + r.decoded = make([]byte, n) + } + if _, err := Decode(r.decoded, buf); err != nil { + r.err = err + return 0, r.err + } + if !r.ignoreCRC && crc(r.decoded[:n]) != checksum { + r.err = ErrCRC + return 0, r.err + } + r.i, r.j = 0, n + continue + + case chunkTypeUncompressedData: + r.blockStart += int64(r.j) + // Section 4.3. Uncompressed data (chunk type 0x01). + if chunkLen < checksumSize { + r.err = ErrCorrupt + return 0, r.err + } + if !r.ensureBufferSize(chunkLen) { + if r.err == nil { + r.err = ErrUnsupported + } + return 0, r.err + } + buf := r.buf[:checksumSize] + if !r.readFull(buf, false) { + return 0, r.err + } + checksum := uint32(buf[0]) | uint32(buf[1])<<8 | uint32(buf[2])<<16 | uint32(buf[3])<<24 + // Read directly into r.decoded instead of via r.buf. + n := chunkLen - checksumSize + if r.snappyFrame && n > maxSnappyBlockSize { + r.err = ErrCorrupt + return 0, r.err + } + if n > len(r.decoded) { + if n > r.maxBlock { + r.err = ErrCorrupt + return 0, r.err + } + r.decoded = make([]byte, n) + } + if !r.readFull(r.decoded[:n], false) { + return 0, r.err + } + if !r.ignoreCRC && crc(r.decoded[:n]) != checksum { + r.err = ErrCRC + return 0, r.err + } + r.i, r.j = 0, n + continue + + case chunkTypeStreamIdentifier: + // Section 4.1. Stream identifier (chunk type 0xff). + if chunkLen != len(magicBody) { + r.err = ErrCorrupt + return 0, r.err + } + if !r.readFull(r.buf[:len(magicBody)], false) { + return 0, r.err + } + if string(r.buf[:len(magicBody)]) != magicBody { + if string(r.buf[:len(magicBody)]) != magicBodySnappy { + r.err = ErrCorrupt + return 0, r.err + } else { + r.snappyFrame = true + } + } else { + r.snappyFrame = false + } + continue + } + + if chunkType <= 0x7f { + // Section 4.5. Reserved unskippable chunks (chunk types 0x02-0x7f). + // fmt.Printf("ERR chunktype: 0x%x\n", chunkType) + r.err = ErrUnsupported + return 0, r.err + } + // Section 4.4 Padding (chunk type 0xfe). + // Section 4.6. Reserved skippable chunks (chunk types 0x80-0xfd). + if chunkLen > maxChunkSize { + // fmt.Printf("ERR chunkLen: 0x%x\n", chunkLen) + r.err = ErrUnsupported + return 0, r.err + } + + // fmt.Printf("skippable: ID: 0x%x, len: 0x%x\n", chunkType, chunkLen) + if !r.skippable(r.buf, chunkLen, false, chunkType) { + return 0, r.err + } + } +} + +// DecodeConcurrent will decode the full stream to w. +// This function should not be combined with reading, seeking or other operations. +// Up to 'concurrent' goroutines will be used. +// If <= 0, runtime.NumCPU will be used. +// On success the number of bytes decompressed nil and is returned. +// This is mainly intended for bigger streams. +func (r *Reader) DecodeConcurrent(w io.Writer, concurrent int) (written int64, err error) { + if r.i > 0 || r.j > 0 || r.blockStart > 0 { + return 0, errors.New("DecodeConcurrent called after ") + } + if concurrent <= 0 { + concurrent = runtime.NumCPU() + } + + // Write to output + var errMu sync.Mutex + var aErr error + setErr := func(e error) (ok bool) { + errMu.Lock() + defer errMu.Unlock() + if e == nil { + return aErr == nil + } + if aErr == nil { + aErr = e + } + return false + } + hasErr := func() (ok bool) { + errMu.Lock() + v := aErr != nil + errMu.Unlock() + return v + } + + var aWritten int64 + toRead := make(chan []byte, concurrent) + writtenBlocks := make(chan []byte, concurrent) + queue := make(chan chan []byte, concurrent) + reUse := make(chan chan []byte, concurrent) + for i := 0; i < concurrent; i++ { + toRead <- make([]byte, 0, r.maxBufSize) + writtenBlocks <- make([]byte, 0, r.maxBufSize) + reUse <- make(chan []byte, 1) + } + // Writer + var wg sync.WaitGroup + wg.Add(1) + go func() { + defer wg.Done() + for toWrite := range queue { + entry := <-toWrite + reUse <- toWrite + if hasErr() || entry == nil { + if entry != nil { + writtenBlocks <- entry + } + continue + } + if hasErr() { + writtenBlocks <- entry + continue + } + n, err := w.Write(entry) + want := len(entry) + writtenBlocks <- entry + if err != nil { + setErr(err) + continue + } + if n != want { + setErr(io.ErrShortWrite) + continue + } + aWritten += int64(n) + } + }() + + defer func() { + if r.err != nil { + setErr(r.err) + } else if err != nil { + setErr(err) + } + close(queue) + wg.Wait() + if err == nil { + err = aErr + } + written = aWritten + }() + + // Reader + for !hasErr() { + if !r.readFull(r.buf[:4], true) { + if r.err == io.EOF { + r.err = nil + } + return 0, r.err + } + chunkType := r.buf[0] + if !r.readHeader { + if chunkType != chunkTypeStreamIdentifier { + r.err = ErrCorrupt + return 0, r.err + } + r.readHeader = true + } + chunkLen := int(r.buf[1]) | int(r.buf[2])<<8 | int(r.buf[3])<<16 + + // The chunk types are specified at + // https://github.com/google/snappy/blob/master/framing_format.txt + switch chunkType { + case chunkTypeCompressedData: + r.blockStart += int64(r.j) + // Section 4.2. Compressed data (chunk type 0x00). + if chunkLen < checksumSize { + r.err = ErrCorrupt + return 0, r.err + } + if chunkLen > r.maxBufSize { + r.err = ErrCorrupt + return 0, r.err + } + orgBuf := <-toRead + buf := orgBuf[:chunkLen] + + if !r.readFull(buf, false) { + return 0, r.err + } + + checksum := uint32(buf[0]) | uint32(buf[1])<<8 | uint32(buf[2])<<16 | uint32(buf[3])<<24 + buf = buf[checksumSize:] + + n, err := DecodedLen(buf) + if err != nil { + r.err = err + return 0, r.err + } + if r.snappyFrame && n > maxSnappyBlockSize { + r.err = ErrCorrupt + return 0, r.err + } + + if n > r.maxBlock { + r.err = ErrCorrupt + return 0, r.err + } + wg.Add(1) + + decoded := <-writtenBlocks + entry := <-reUse + queue <- entry + go func() { + defer wg.Done() + decoded = decoded[:n] + _, err := Decode(decoded, buf) + toRead <- orgBuf + if err != nil { + writtenBlocks <- decoded + setErr(err) + entry <- nil + return + } + if !r.ignoreCRC && crc(decoded) != checksum { + writtenBlocks <- decoded + setErr(ErrCRC) + entry <- nil + return + } + entry <- decoded + }() + continue + + case chunkTypeUncompressedData: + + // Section 4.3. Uncompressed data (chunk type 0x01). + if chunkLen < checksumSize { + r.err = ErrCorrupt + return 0, r.err + } + if chunkLen > r.maxBufSize { + r.err = ErrCorrupt + return 0, r.err + } + // Grab write buffer + orgBuf := <-writtenBlocks + buf := orgBuf[:checksumSize] + if !r.readFull(buf, false) { + return 0, r.err + } + checksum := uint32(buf[0]) | uint32(buf[1])<<8 | uint32(buf[2])<<16 | uint32(buf[3])<<24 + // Read content. + n := chunkLen - checksumSize + + if r.snappyFrame && n > maxSnappyBlockSize { + r.err = ErrCorrupt + return 0, r.err + } + if n > r.maxBlock { + r.err = ErrCorrupt + return 0, r.err + } + // Read uncompressed + buf = orgBuf[:n] + if !r.readFull(buf, false) { + return 0, r.err + } + + if !r.ignoreCRC && crc(buf) != checksum { + r.err = ErrCRC + return 0, r.err + } + entry := <-reUse + queue <- entry + entry <- buf + continue + + case chunkTypeStreamIdentifier: + // Section 4.1. Stream identifier (chunk type 0xff). + if chunkLen != len(magicBody) { + r.err = ErrCorrupt + return 0, r.err + } + if !r.readFull(r.buf[:len(magicBody)], false) { + return 0, r.err + } + if string(r.buf[:len(magicBody)]) != magicBody { + if string(r.buf[:len(magicBody)]) != magicBodySnappy { + r.err = ErrCorrupt + return 0, r.err + } else { + r.snappyFrame = true + } + } else { + r.snappyFrame = false + } + continue + } + + if chunkType <= 0x7f { + // Section 4.5. Reserved unskippable chunks (chunk types 0x02-0x7f). + // fmt.Printf("ERR chunktype: 0x%x\n", chunkType) + r.err = ErrUnsupported + return 0, r.err + } + // Section 4.4 Padding (chunk type 0xfe). + // Section 4.6. Reserved skippable chunks (chunk types 0x80-0xfd). + if chunkLen > maxChunkSize { + // fmt.Printf("ERR chunkLen: 0x%x\n", chunkLen) + r.err = ErrUnsupported + return 0, r.err + } + + // fmt.Printf("skippable: ID: 0x%x, len: 0x%x\n", chunkType, chunkLen) + if !r.skippable(r.buf, chunkLen, false, chunkType) { + return 0, r.err + } + } + return 0, r.err +} + +// Skip will skip n bytes forward in the decompressed output. +// For larger skips this consumes less CPU and is faster than reading output and discarding it. +// CRC is not checked on skipped blocks. +// io.ErrUnexpectedEOF is returned if the stream ends before all bytes have been skipped. +// If a decoding error is encountered subsequent calls to Read will also fail. +func (r *Reader) Skip(n int64) error { + if n < 0 { + return errors.New("attempted negative skip") + } + if r.err != nil { + return r.err + } + + for n > 0 { + if r.i < r.j { + // Skip in buffer. + // decoded[i:j] contains decoded bytes that have not yet been passed on. + left := int64(r.j - r.i) + if left >= n { + tmp := int64(r.i) + n + if tmp > math.MaxInt32 { + return errors.New("s2: internal overflow in skip") + } + r.i = int(tmp) + return nil + } + n -= int64(r.j - r.i) + r.i = r.j + } + + // Buffer empty; read blocks until we have content. + if !r.readFull(r.buf[:4], true) { + if r.err == io.EOF { + r.err = io.ErrUnexpectedEOF + } + return r.err + } + chunkType := r.buf[0] + if !r.readHeader { + if chunkType != chunkTypeStreamIdentifier { + r.err = ErrCorrupt + return r.err + } + r.readHeader = true + } + chunkLen := int(r.buf[1]) | int(r.buf[2])<<8 | int(r.buf[3])<<16 + + // The chunk types are specified at + // https://github.com/google/snappy/blob/master/framing_format.txt + switch chunkType { + case chunkTypeCompressedData: + r.blockStart += int64(r.j) + // Section 4.2. Compressed data (chunk type 0x00). + if chunkLen < checksumSize { + r.err = ErrCorrupt + return r.err + } + if !r.ensureBufferSize(chunkLen) { + if r.err == nil { + r.err = ErrUnsupported + } + return r.err + } + buf := r.buf[:chunkLen] + if !r.readFull(buf, false) { + return r.err + } + checksum := uint32(buf[0]) | uint32(buf[1])<<8 | uint32(buf[2])<<16 | uint32(buf[3])<<24 + buf = buf[checksumSize:] + + dLen, err := DecodedLen(buf) + if err != nil { + r.err = err + return r.err + } + if dLen > r.maxBlock { + r.err = ErrCorrupt + return r.err + } + // Check if destination is within this block + if int64(dLen) > n { + if len(r.decoded) < dLen { + r.decoded = make([]byte, dLen) + } + if _, err := Decode(r.decoded, buf); err != nil { + r.err = err + return r.err + } + if crc(r.decoded[:dLen]) != checksum { + r.err = ErrCorrupt + return r.err + } + } else { + // Skip block completely + n -= int64(dLen) + r.blockStart += int64(dLen) + dLen = 0 + } + r.i, r.j = 0, dLen + continue + case chunkTypeUncompressedData: + r.blockStart += int64(r.j) + // Section 4.3. Uncompressed data (chunk type 0x01). + if chunkLen < checksumSize { + r.err = ErrCorrupt + return r.err + } + if !r.ensureBufferSize(chunkLen) { + if r.err != nil { + r.err = ErrUnsupported + } + return r.err + } + buf := r.buf[:checksumSize] + if !r.readFull(buf, false) { + return r.err + } + checksum := uint32(buf[0]) | uint32(buf[1])<<8 | uint32(buf[2])<<16 | uint32(buf[3])<<24 + // Read directly into r.decoded instead of via r.buf. + n2 := chunkLen - checksumSize + if n2 > len(r.decoded) { + if n2 > r.maxBlock { + r.err = ErrCorrupt + return r.err + } + r.decoded = make([]byte, n2) + } + if !r.readFull(r.decoded[:n2], false) { + return r.err + } + if int64(n2) < n { + if crc(r.decoded[:n2]) != checksum { + r.err = ErrCorrupt + return r.err + } + } + r.i, r.j = 0, n2 + continue + case chunkTypeStreamIdentifier: + // Section 4.1. Stream identifier (chunk type 0xff). + if chunkLen != len(magicBody) { + r.err = ErrCorrupt + return r.err + } + if !r.readFull(r.buf[:len(magicBody)], false) { + return r.err + } + if string(r.buf[:len(magicBody)]) != magicBody { + if string(r.buf[:len(magicBody)]) != magicBodySnappy { + r.err = ErrCorrupt + return r.err + } + } + + continue + } + + if chunkType <= 0x7f { + // Section 4.5. Reserved unskippable chunks (chunk types 0x02-0x7f). + r.err = ErrUnsupported + return r.err + } + if chunkLen > maxChunkSize { + r.err = ErrUnsupported + return r.err + } + // Section 4.4 Padding (chunk type 0xfe). + // Section 4.6. Reserved skippable chunks (chunk types 0x80-0xfd). + if !r.skippable(r.buf, chunkLen, false, chunkType) { + return r.err + } + } + return nil +} + +// ReadSeeker provides random or forward seeking in compressed content. +// See Reader.ReadSeeker +type ReadSeeker struct { + *Reader + readAtMu sync.Mutex +} + +// ReadSeeker will return an io.ReadSeeker and io.ReaderAt +// compatible version of the reader. +// If 'random' is specified the returned io.Seeker can be used for +// random seeking, otherwise only forward seeking is supported. +// Enabling random seeking requires the original input to support +// the io.Seeker interface. +// A custom index can be specified which will be used if supplied. +// When using a custom index, it will not be read from the input stream. +// The ReadAt position will affect regular reads and the current position of Seek. +// So using Read after ReadAt will continue from where the ReadAt stopped. +// No functions should be used concurrently. +// The returned ReadSeeker contains a shallow reference to the existing Reader, +// meaning changes performed to one is reflected in the other. +func (r *Reader) ReadSeeker(random bool, index []byte) (*ReadSeeker, error) { + // Read index if provided. + if len(index) != 0 { + if r.index == nil { + r.index = &Index{} + } + if _, err := r.index.Load(index); err != nil { + return nil, ErrCantSeek{Reason: "loading index returned: " + err.Error()} + } + } + + // Check if input is seekable + rs, ok := r.r.(io.ReadSeeker) + if !ok { + if !random { + return &ReadSeeker{Reader: r}, nil + } + return nil, ErrCantSeek{Reason: "input stream isn't seekable"} + } + + if r.index != nil { + // Seekable and index, ok... + return &ReadSeeker{Reader: r}, nil + } + + // Load from stream. + r.index = &Index{} + + // Read current position. + pos, err := rs.Seek(0, io.SeekCurrent) + if err != nil { + return nil, ErrCantSeek{Reason: "seeking input returned: " + err.Error()} + } + err = r.index.LoadStream(rs) + if err != nil { + if err == ErrUnsupported { + // If we don't require random seeking, reset input and return. + if !random { + _, err = rs.Seek(pos, io.SeekStart) + if err != nil { + return nil, ErrCantSeek{Reason: "resetting stream returned: " + err.Error()} + } + r.index = nil + return &ReadSeeker{Reader: r}, nil + } + return nil, ErrCantSeek{Reason: "input stream does not contain an index"} + } + return nil, ErrCantSeek{Reason: "reading index returned: " + err.Error()} + } + + // reset position. + _, err = rs.Seek(pos, io.SeekStart) + if err != nil { + return nil, ErrCantSeek{Reason: "seeking input returned: " + err.Error()} + } + return &ReadSeeker{Reader: r}, nil +} + +// Seek allows seeking in compressed data. +func (r *ReadSeeker) Seek(offset int64, whence int) (int64, error) { + if r.err != nil { + if !errors.Is(r.err, io.EOF) { + return 0, r.err + } + // Reset on EOF + r.err = nil + } + + // Calculate absolute offset. + absOffset := offset + + switch whence { + case io.SeekStart: + case io.SeekCurrent: + absOffset = r.blockStart + int64(r.i) + offset + case io.SeekEnd: + if r.index == nil { + return 0, ErrUnsupported + } + absOffset = r.index.TotalUncompressed + offset + default: + r.err = ErrUnsupported + return 0, r.err + } + + if absOffset < 0 { + return 0, errors.New("seek before start of file") + } + + if !r.readHeader { + // Make sure we read the header. + _, r.err = r.Read([]byte{}) + if r.err != nil { + return 0, r.err + } + } + + // If we are inside current block no need to seek. + // This includes no offset changes. + if absOffset >= r.blockStart && absOffset < r.blockStart+int64(r.j) { + r.i = int(absOffset - r.blockStart) + return r.blockStart + int64(r.i), nil + } + + rs, ok := r.r.(io.ReadSeeker) + if r.index == nil || !ok { + currOffset := r.blockStart + int64(r.i) + if absOffset >= currOffset { + err := r.Skip(absOffset - currOffset) + return r.blockStart + int64(r.i), err + } + return 0, ErrUnsupported + } + + // We can seek and we have an index. + c, u, err := r.index.Find(absOffset) + if err != nil { + return r.blockStart + int64(r.i), err + } + + // Seek to next block + _, err = rs.Seek(c, io.SeekStart) + if err != nil { + return 0, err + } + + r.i = r.j // Remove rest of current block. + r.blockStart = u - int64(r.j) // Adjust current block start for accounting. + if u < absOffset { + // Forward inside block + return absOffset, r.Skip(absOffset - u) + } + if u > absOffset { + return 0, fmt.Errorf("s2 seek: (internal error) u (%d) > absOffset (%d)", u, absOffset) + } + return absOffset, nil +} + +// ReadAt reads len(p) bytes into p starting at offset off in the +// underlying input source. It returns the number of bytes +// read (0 <= n <= len(p)) and any error encountered. +// +// When ReadAt returns n < len(p), it returns a non-nil error +// explaining why more bytes were not returned. In this respect, +// ReadAt is stricter than Read. +// +// Even if ReadAt returns n < len(p), it may use all of p as scratch +// space during the call. If some data is available but not len(p) bytes, +// ReadAt blocks until either all the data is available or an error occurs. +// In this respect ReadAt is different from Read. +// +// If the n = len(p) bytes returned by ReadAt are at the end of the +// input source, ReadAt may return either err == EOF or err == nil. +// +// If ReadAt is reading from an input source with a seek offset, +// ReadAt should not affect nor be affected by the underlying +// seek offset. +// +// Clients of ReadAt can execute parallel ReadAt calls on the +// same input source. This is however not recommended. +func (r *ReadSeeker) ReadAt(p []byte, offset int64) (int, error) { + r.readAtMu.Lock() + defer r.readAtMu.Unlock() + _, err := r.Seek(offset, io.SeekStart) + if err != nil { + return 0, err + } + n := 0 + for n < len(p) { + n2, err := r.Read(p[n:]) + if err != nil { + // This will include io.EOF + return n + n2, err + } + n += n2 + } + return n, nil +} + +// ReadByte satisfies the io.ByteReader interface. +func (r *Reader) ReadByte() (byte, error) { + if r.err != nil { + return 0, r.err + } + if r.i < r.j { + c := r.decoded[r.i] + r.i++ + return c, nil + } + var tmp [1]byte + for i := 0; i < 10; i++ { + n, err := r.Read(tmp[:]) + if err != nil { + return 0, err + } + if n == 1 { + return tmp[0], nil + } + } + return 0, io.ErrNoProgress +} + +// SkippableCB will register a callback for chunks with the specified ID. +// ID must be a Reserved skippable chunks ID, 0x80-0xfd (inclusive). +// For each chunk with the ID, the callback is called with the content. +// Any returned non-nil error will abort decompression. +// Only one callback per ID is supported, latest sent will be used. +// Sending a nil function will disable previous callbacks. +// You can peek the stream, triggering the callback, by doing a Read with a 0 +// byte buffer. +func (r *Reader) SkippableCB(id uint8, fn func(r io.Reader) error) error { + if id < 0x80 || id >= chunkTypePadding { + return fmt.Errorf("ReaderSkippableCB: Invalid id provided, must be 0x80-0xfe (inclusive)") + } + r.skippableCB[id-0x80] = fn + return nil +} diff --git a/vendor/github.com/klauspost/compress/s2/s2.go b/vendor/github.com/klauspost/compress/s2/s2.go new file mode 100644 index 000000000..cbd1ed64d --- /dev/null +++ b/vendor/github.com/klauspost/compress/s2/s2.go @@ -0,0 +1,151 @@ +// Copyright 2011 The Snappy-Go Authors. All rights reserved. +// Copyright (c) 2019 Klaus Post. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Package s2 implements the S2 compression format. +// +// S2 is an extension of Snappy. Similar to Snappy S2 is aimed for high throughput, +// which is why it features concurrent compression for bigger payloads. +// +// Decoding is compatible with Snappy compressed content, +// but content compressed with S2 cannot be decompressed by Snappy. +// +// For more information on Snappy/S2 differences see README in: https://github.com/klauspost/compress/tree/master/s2 +// +// There are actually two S2 formats: block and stream. They are related, +// but different: trying to decompress block-compressed data as a S2 stream +// will fail, and vice versa. The block format is the Decode and Encode +// functions and the stream format is the Reader and Writer types. +// +// A "better" compression option is available. This will trade some compression +// speed +// +// The block format, the more common case, is used when the complete size (the +// number of bytes) of the original data is known upfront, at the time +// compression starts. The stream format, also known as the framing format, is +// for when that isn't always true. +// +// Blocks to not offer much data protection, so it is up to you to +// add data validation of decompressed blocks. +// +// Streams perform CRC validation of the decompressed data. +// Stream compression will also be performed on multiple CPU cores concurrently +// significantly improving throughput. +package s2 + +import ( + "bytes" + "hash/crc32" + + "github.com/klauspost/compress/internal/race" +) + +/* +Each encoded block begins with the varint-encoded length of the decoded data, +followed by a sequence of chunks. Chunks begin and end on byte boundaries. The +first byte of each chunk is broken into its 2 least and 6 most significant bits +called l and m: l ranges in [0, 4) and m ranges in [0, 64). l is the chunk tag. +Zero means a literal tag. All other values mean a copy tag. + +For literal tags: + - If m < 60, the next 1 + m bytes are literal bytes. + - Otherwise, let n be the little-endian unsigned integer denoted by the next + m - 59 bytes. The next 1 + n bytes after that are literal bytes. + +For copy tags, length bytes are copied from offset bytes ago, in the style of +Lempel-Ziv compression algorithms. In particular: + - For l == 1, the offset ranges in [0, 1<<11) and the length in [4, 12). + The length is 4 + the low 3 bits of m. The high 3 bits of m form bits 8-10 + of the offset. The next byte is bits 0-7 of the offset. + - For l == 2, the offset ranges in [0, 1<<16) and the length in [1, 65). + The length is 1 + m. The offset is the little-endian unsigned integer + denoted by the next 2 bytes. + - For l == 3, the offset ranges in [0, 1<<32) and the length in + [1, 65). The length is 1 + m. The offset is the little-endian unsigned + integer denoted by the next 4 bytes. +*/ +const ( + tagLiteral = 0x00 + tagCopy1 = 0x01 + tagCopy2 = 0x02 + tagCopy4 = 0x03 +) + +const ( + checksumSize = 4 + chunkHeaderSize = 4 + magicChunk = "\xff\x06\x00\x00" + magicBody + magicChunkSnappy = "\xff\x06\x00\x00" + magicBodySnappy + magicBodySnappy = "sNaPpY" + magicBody = "S2sTwO" + + // maxBlockSize is the maximum size of the input to encodeBlock. + // + // For the framing format (Writer type instead of Encode function), + // this is the maximum uncompressed size of a block. + maxBlockSize = 4 << 20 + + // minBlockSize is the minimum size of block setting when creating a writer. + minBlockSize = 4 << 10 + + skippableFrameHeader = 4 + maxChunkSize = 1<<24 - 1 // 16777215 + + // Default block size + defaultBlockSize = 1 << 20 + + // maxSnappyBlockSize is the maximum snappy block size. + maxSnappyBlockSize = 1 << 16 + + obufHeaderLen = checksumSize + chunkHeaderSize +) + +const ( + chunkTypeCompressedData = 0x00 + chunkTypeUncompressedData = 0x01 + ChunkTypeIndex = 0x99 + chunkTypePadding = 0xfe + chunkTypeStreamIdentifier = 0xff +) + +var ( + crcTable = crc32.MakeTable(crc32.Castagnoli) + magicChunkSnappyBytes = []byte(magicChunkSnappy) // Can be passed to functions where it escapes. + magicChunkBytes = []byte(magicChunk) // Can be passed to functions where it escapes. +) + +// crc implements the checksum specified in section 3 of +// https://github.com/google/snappy/blob/master/framing_format.txt +func crc(b []byte) uint32 { + race.ReadSlice(b) + + c := crc32.Update(0, crcTable, b) + return c>>15 | c<<17 + 0xa282ead8 +} + +// literalExtraSize returns the extra size of encoding n literals. +// n should be >= 0 and <= math.MaxUint32. +func literalExtraSize(n int64) int64 { + if n == 0 { + return 0 + } + switch { + case n < 60: + return 1 + case n < 1<<8: + return 2 + case n < 1<<16: + return 3 + case n < 1<<24: + return 4 + default: + return 5 + } +} + +type byter interface { + Bytes() []byte +} + +var _ byter = &bytes.Buffer{} diff --git a/vendor/github.com/klauspost/compress/s2/writer.go b/vendor/github.com/klauspost/compress/s2/writer.go new file mode 100644 index 000000000..0a46f2b98 --- /dev/null +++ b/vendor/github.com/klauspost/compress/s2/writer.go @@ -0,0 +1,1039 @@ +// Copyright 2011 The Snappy-Go Authors. All rights reserved. +// Copyright (c) 2019+ Klaus Post. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package s2 + +import ( + "crypto/rand" + "encoding/binary" + "errors" + "fmt" + "io" + "runtime" + "sync" + + "github.com/klauspost/compress/internal/race" +) + +const ( + levelUncompressed = iota + 1 + levelFast + levelBetter + levelBest +) + +// NewWriter returns a new Writer that compresses to w, using the +// framing format described at +// https://github.com/google/snappy/blob/master/framing_format.txt +// +// Users must call Close to guarantee all data has been forwarded to +// the underlying io.Writer and that resources are released. +// They may also call Flush zero or more times before calling Close. +func NewWriter(w io.Writer, opts ...WriterOption) *Writer { + w2 := Writer{ + blockSize: defaultBlockSize, + concurrency: runtime.GOMAXPROCS(0), + randSrc: rand.Reader, + level: levelFast, + } + for _, opt := range opts { + if err := opt(&w2); err != nil { + w2.errState = err + return &w2 + } + } + w2.obufLen = obufHeaderLen + MaxEncodedLen(w2.blockSize) + w2.paramsOK = true + w2.ibuf = make([]byte, 0, w2.blockSize) + w2.buffers.New = func() interface{} { + return make([]byte, w2.obufLen) + } + w2.Reset(w) + return &w2 +} + +// Writer is an io.Writer that can write Snappy-compressed bytes. +type Writer struct { + errMu sync.Mutex + errState error + + // ibuf is a buffer for the incoming (uncompressed) bytes. + ibuf []byte + + blockSize int + obufLen int + concurrency int + written int64 + uncompWritten int64 // Bytes sent to compression + output chan chan result + buffers sync.Pool + pad int + + writer io.Writer + randSrc io.Reader + writerWg sync.WaitGroup + index Index + customEnc func(dst, src []byte) int + + // wroteStreamHeader is whether we have written the stream header. + wroteStreamHeader bool + paramsOK bool + snappy bool + flushOnWrite bool + appendIndex bool + level uint8 +} + +type result struct { + b []byte + // Uncompressed start offset + startOffset int64 +} + +// err returns the previously set error. +// If no error has been set it is set to err if not nil. +func (w *Writer) err(err error) error { + w.errMu.Lock() + errSet := w.errState + if errSet == nil && err != nil { + w.errState = err + errSet = err + } + w.errMu.Unlock() + return errSet +} + +// Reset discards the writer's state and switches the Snappy writer to write to w. +// This permits reusing a Writer rather than allocating a new one. +func (w *Writer) Reset(writer io.Writer) { + if !w.paramsOK { + return + } + // Close previous writer, if any. + if w.output != nil { + close(w.output) + w.writerWg.Wait() + w.output = nil + } + w.errState = nil + w.ibuf = w.ibuf[:0] + w.wroteStreamHeader = false + w.written = 0 + w.writer = writer + w.uncompWritten = 0 + w.index.reset(w.blockSize) + + // If we didn't get a writer, stop here. + if writer == nil { + return + } + // If no concurrency requested, don't spin up writer goroutine. + if w.concurrency == 1 { + return + } + + toWrite := make(chan chan result, w.concurrency) + w.output = toWrite + w.writerWg.Add(1) + + // Start a writer goroutine that will write all output in order. + go func() { + defer w.writerWg.Done() + + // Get a queued write. + for write := range toWrite { + // Wait for the data to be available. + input := <-write + in := input.b + if len(in) > 0 { + if w.err(nil) == nil { + // Don't expose data from previous buffers. + toWrite := in[:len(in):len(in)] + // Write to output. + n, err := writer.Write(toWrite) + if err == nil && n != len(toWrite) { + err = io.ErrShortBuffer + } + _ = w.err(err) + w.err(w.index.add(w.written, input.startOffset)) + w.written += int64(n) + } + } + if cap(in) >= w.obufLen { + w.buffers.Put(in) + } + // close the incoming write request. + // This can be used for synchronizing flushes. + close(write) + } + }() +} + +// Write satisfies the io.Writer interface. +func (w *Writer) Write(p []byte) (nRet int, errRet error) { + if err := w.err(nil); err != nil { + return 0, err + } + if w.flushOnWrite { + return w.write(p) + } + // If we exceed the input buffer size, start writing + for len(p) > (cap(w.ibuf)-len(w.ibuf)) && w.err(nil) == nil { + var n int + if len(w.ibuf) == 0 { + // Large write, empty buffer. + // Write directly from p to avoid copy. + n, _ = w.write(p) + } else { + n = copy(w.ibuf[len(w.ibuf):cap(w.ibuf)], p) + w.ibuf = w.ibuf[:len(w.ibuf)+n] + w.write(w.ibuf) + w.ibuf = w.ibuf[:0] + } + nRet += n + p = p[n:] + } + if err := w.err(nil); err != nil { + return nRet, err + } + // p should always be able to fit into w.ibuf now. + n := copy(w.ibuf[len(w.ibuf):cap(w.ibuf)], p) + w.ibuf = w.ibuf[:len(w.ibuf)+n] + nRet += n + return nRet, nil +} + +// ReadFrom implements the io.ReaderFrom interface. +// Using this is typically more efficient since it avoids a memory copy. +// ReadFrom reads data from r until EOF or error. +// The return value n is the number of bytes read. +// Any error except io.EOF encountered during the read is also returned. +func (w *Writer) ReadFrom(r io.Reader) (n int64, err error) { + if err := w.err(nil); err != nil { + return 0, err + } + if len(w.ibuf) > 0 { + err := w.AsyncFlush() + if err != nil { + return 0, err + } + } + if br, ok := r.(byter); ok { + buf := br.Bytes() + if err := w.EncodeBuffer(buf); err != nil { + return 0, err + } + return int64(len(buf)), w.AsyncFlush() + } + for { + inbuf := w.buffers.Get().([]byte)[:w.blockSize+obufHeaderLen] + n2, err := io.ReadFull(r, inbuf[obufHeaderLen:]) + if err != nil { + if err == io.ErrUnexpectedEOF { + err = io.EOF + } + if err != io.EOF { + return n, w.err(err) + } + } + if n2 == 0 { + if cap(inbuf) >= w.obufLen { + w.buffers.Put(inbuf) + } + break + } + n += int64(n2) + err2 := w.writeFull(inbuf[:n2+obufHeaderLen]) + if w.err(err2) != nil { + break + } + + if err != nil { + // We got EOF and wrote everything + break + } + } + + return n, w.err(nil) +} + +// AddSkippableBlock will add a skippable block to the stream. +// The ID must be 0x80-0xfe (inclusive). +// Length of the skippable block must be <= 16777215 bytes. +func (w *Writer) AddSkippableBlock(id uint8, data []byte) (err error) { + if err := w.err(nil); err != nil { + return err + } + if len(data) == 0 { + return nil + } + if id < 0x80 || id > chunkTypePadding { + return fmt.Errorf("invalid skippable block id %x", id) + } + if len(data) > maxChunkSize { + return fmt.Errorf("skippable block excessed maximum size") + } + var header [4]byte + chunkLen := len(data) + header[0] = id + header[1] = uint8(chunkLen >> 0) + header[2] = uint8(chunkLen >> 8) + header[3] = uint8(chunkLen >> 16) + if w.concurrency == 1 { + write := func(b []byte) error { + n, err := w.writer.Write(b) + if err = w.err(err); err != nil { + return err + } + if n != len(b) { + return w.err(io.ErrShortWrite) + } + w.written += int64(n) + return w.err(nil) + } + if !w.wroteStreamHeader { + w.wroteStreamHeader = true + if w.snappy { + if err := write([]byte(magicChunkSnappy)); err != nil { + return err + } + } else { + if err := write([]byte(magicChunk)); err != nil { + return err + } + } + } + if err := write(header[:]); err != nil { + return err + } + return write(data) + } + + // Create output... + if !w.wroteStreamHeader { + w.wroteStreamHeader = true + hWriter := make(chan result) + w.output <- hWriter + if w.snappy { + hWriter <- result{startOffset: w.uncompWritten, b: magicChunkSnappyBytes} + } else { + hWriter <- result{startOffset: w.uncompWritten, b: magicChunkBytes} + } + } + + // Copy input. + inbuf := w.buffers.Get().([]byte)[:4] + copy(inbuf, header[:]) + inbuf = append(inbuf, data...) + + output := make(chan result, 1) + // Queue output. + w.output <- output + output <- result{startOffset: w.uncompWritten, b: inbuf} + + return nil +} + +// EncodeBuffer will add a buffer to the stream. +// This is the fastest way to encode a stream, +// but the input buffer cannot be written to by the caller +// until Flush or Close has been called when concurrency != 1. +// +// If you cannot control that, use the regular Write function. +// +// Note that input is not buffered. +// This means that each write will result in discrete blocks being created. +// For buffered writes, use the regular Write function. +func (w *Writer) EncodeBuffer(buf []byte) (err error) { + if err := w.err(nil); err != nil { + return err + } + + if w.flushOnWrite { + _, err := w.write(buf) + return err + } + // Flush queued data first. + if len(w.ibuf) > 0 { + err := w.AsyncFlush() + if err != nil { + return err + } + } + if w.concurrency == 1 { + _, err := w.writeSync(buf) + return err + } + + // Spawn goroutine and write block to output channel. + if !w.wroteStreamHeader { + w.wroteStreamHeader = true + hWriter := make(chan result) + w.output <- hWriter + if w.snappy { + hWriter <- result{startOffset: w.uncompWritten, b: magicChunkSnappyBytes} + } else { + hWriter <- result{startOffset: w.uncompWritten, b: magicChunkBytes} + } + } + + for len(buf) > 0 { + // Cut input. + uncompressed := buf + if len(uncompressed) > w.blockSize { + uncompressed = uncompressed[:w.blockSize] + } + buf = buf[len(uncompressed):] + // Get an output buffer. + obuf := w.buffers.Get().([]byte)[:len(uncompressed)+obufHeaderLen] + race.WriteSlice(obuf) + + output := make(chan result) + // Queue output now, so we keep order. + w.output <- output + res := result{ + startOffset: w.uncompWritten, + } + w.uncompWritten += int64(len(uncompressed)) + go func() { + race.ReadSlice(uncompressed) + + checksum := crc(uncompressed) + + // Set to uncompressed. + chunkType := uint8(chunkTypeUncompressedData) + chunkLen := 4 + len(uncompressed) + + // Attempt compressing. + n := binary.PutUvarint(obuf[obufHeaderLen:], uint64(len(uncompressed))) + n2 := w.encodeBlock(obuf[obufHeaderLen+n:], uncompressed) + + // Check if we should use this, or store as uncompressed instead. + if n2 > 0 { + chunkType = uint8(chunkTypeCompressedData) + chunkLen = 4 + n + n2 + obuf = obuf[:obufHeaderLen+n+n2] + } else { + // copy uncompressed + copy(obuf[obufHeaderLen:], uncompressed) + } + + // Fill in the per-chunk header that comes before the body. + obuf[0] = chunkType + obuf[1] = uint8(chunkLen >> 0) + obuf[2] = uint8(chunkLen >> 8) + obuf[3] = uint8(chunkLen >> 16) + obuf[4] = uint8(checksum >> 0) + obuf[5] = uint8(checksum >> 8) + obuf[6] = uint8(checksum >> 16) + obuf[7] = uint8(checksum >> 24) + + // Queue final output. + res.b = obuf + output <- res + }() + } + return nil +} + +func (w *Writer) encodeBlock(obuf, uncompressed []byte) int { + if w.customEnc != nil { + if ret := w.customEnc(obuf, uncompressed); ret >= 0 { + return ret + } + } + if w.snappy { + switch w.level { + case levelFast: + return encodeBlockSnappy(obuf, uncompressed) + case levelBetter: + return encodeBlockBetterSnappy(obuf, uncompressed) + case levelBest: + return encodeBlockBestSnappy(obuf, uncompressed) + } + return 0 + } + switch w.level { + case levelFast: + return encodeBlock(obuf, uncompressed) + case levelBetter: + return encodeBlockBetter(obuf, uncompressed) + case levelBest: + return encodeBlockBest(obuf, uncompressed, nil) + } + return 0 +} + +func (w *Writer) write(p []byte) (nRet int, errRet error) { + if err := w.err(nil); err != nil { + return 0, err + } + if w.concurrency == 1 { + return w.writeSync(p) + } + + // Spawn goroutine and write block to output channel. + for len(p) > 0 { + if !w.wroteStreamHeader { + w.wroteStreamHeader = true + hWriter := make(chan result) + w.output <- hWriter + if w.snappy { + hWriter <- result{startOffset: w.uncompWritten, b: magicChunkSnappyBytes} + } else { + hWriter <- result{startOffset: w.uncompWritten, b: magicChunkBytes} + } + } + + var uncompressed []byte + if len(p) > w.blockSize { + uncompressed, p = p[:w.blockSize], p[w.blockSize:] + } else { + uncompressed, p = p, nil + } + + // Copy input. + // If the block is incompressible, this is used for the result. + inbuf := w.buffers.Get().([]byte)[:len(uncompressed)+obufHeaderLen] + obuf := w.buffers.Get().([]byte)[:w.obufLen] + copy(inbuf[obufHeaderLen:], uncompressed) + uncompressed = inbuf[obufHeaderLen:] + + output := make(chan result) + // Queue output now, so we keep order. + w.output <- output + res := result{ + startOffset: w.uncompWritten, + } + w.uncompWritten += int64(len(uncompressed)) + + go func() { + checksum := crc(uncompressed) + + // Set to uncompressed. + chunkType := uint8(chunkTypeUncompressedData) + chunkLen := 4 + len(uncompressed) + + // Attempt compressing. + n := binary.PutUvarint(obuf[obufHeaderLen:], uint64(len(uncompressed))) + n2 := w.encodeBlock(obuf[obufHeaderLen+n:], uncompressed) + + // Check if we should use this, or store as uncompressed instead. + if n2 > 0 { + chunkType = uint8(chunkTypeCompressedData) + chunkLen = 4 + n + n2 + obuf = obuf[:obufHeaderLen+n+n2] + } else { + // Use input as output. + obuf, inbuf = inbuf, obuf + } + + // Fill in the per-chunk header that comes before the body. + obuf[0] = chunkType + obuf[1] = uint8(chunkLen >> 0) + obuf[2] = uint8(chunkLen >> 8) + obuf[3] = uint8(chunkLen >> 16) + obuf[4] = uint8(checksum >> 0) + obuf[5] = uint8(checksum >> 8) + obuf[6] = uint8(checksum >> 16) + obuf[7] = uint8(checksum >> 24) + + // Queue final output. + res.b = obuf + output <- res + + // Put unused buffer back in pool. + w.buffers.Put(inbuf) + }() + nRet += len(uncompressed) + } + return nRet, nil +} + +// writeFull is a special version of write that will always write the full buffer. +// Data to be compressed should start at offset obufHeaderLen and fill the remainder of the buffer. +// The data will be written as a single block. +// The caller is not allowed to use inbuf after this function has been called. +func (w *Writer) writeFull(inbuf []byte) (errRet error) { + if err := w.err(nil); err != nil { + return err + } + + if w.concurrency == 1 { + _, err := w.writeSync(inbuf[obufHeaderLen:]) + if cap(inbuf) >= w.obufLen { + w.buffers.Put(inbuf) + } + return err + } + + // Spawn goroutine and write block to output channel. + if !w.wroteStreamHeader { + w.wroteStreamHeader = true + hWriter := make(chan result) + w.output <- hWriter + if w.snappy { + hWriter <- result{startOffset: w.uncompWritten, b: magicChunkSnappyBytes} + } else { + hWriter <- result{startOffset: w.uncompWritten, b: magicChunkBytes} + } + } + + // Get an output buffer. + obuf := w.buffers.Get().([]byte)[:w.obufLen] + uncompressed := inbuf[obufHeaderLen:] + + output := make(chan result) + // Queue output now, so we keep order. + w.output <- output + res := result{ + startOffset: w.uncompWritten, + } + w.uncompWritten += int64(len(uncompressed)) + + go func() { + checksum := crc(uncompressed) + + // Set to uncompressed. + chunkType := uint8(chunkTypeUncompressedData) + chunkLen := 4 + len(uncompressed) + + // Attempt compressing. + n := binary.PutUvarint(obuf[obufHeaderLen:], uint64(len(uncompressed))) + n2 := w.encodeBlock(obuf[obufHeaderLen+n:], uncompressed) + + // Check if we should use this, or store as uncompressed instead. + if n2 > 0 { + chunkType = uint8(chunkTypeCompressedData) + chunkLen = 4 + n + n2 + obuf = obuf[:obufHeaderLen+n+n2] + } else { + // Use input as output. + obuf, inbuf = inbuf, obuf + } + + // Fill in the per-chunk header that comes before the body. + obuf[0] = chunkType + obuf[1] = uint8(chunkLen >> 0) + obuf[2] = uint8(chunkLen >> 8) + obuf[3] = uint8(chunkLen >> 16) + obuf[4] = uint8(checksum >> 0) + obuf[5] = uint8(checksum >> 8) + obuf[6] = uint8(checksum >> 16) + obuf[7] = uint8(checksum >> 24) + + // Queue final output. + res.b = obuf + output <- res + + // Put unused buffer back in pool. + w.buffers.Put(inbuf) + }() + return nil +} + +func (w *Writer) writeSync(p []byte) (nRet int, errRet error) { + if err := w.err(nil); err != nil { + return 0, err + } + if !w.wroteStreamHeader { + w.wroteStreamHeader = true + var n int + var err error + if w.snappy { + n, err = w.writer.Write(magicChunkSnappyBytes) + } else { + n, err = w.writer.Write(magicChunkBytes) + } + if err != nil { + return 0, w.err(err) + } + if n != len(magicChunk) { + return 0, w.err(io.ErrShortWrite) + } + w.written += int64(n) + } + + for len(p) > 0 { + var uncompressed []byte + if len(p) > w.blockSize { + uncompressed, p = p[:w.blockSize], p[w.blockSize:] + } else { + uncompressed, p = p, nil + } + + obuf := w.buffers.Get().([]byte)[:w.obufLen] + checksum := crc(uncompressed) + + // Set to uncompressed. + chunkType := uint8(chunkTypeUncompressedData) + chunkLen := 4 + len(uncompressed) + + // Attempt compressing. + n := binary.PutUvarint(obuf[obufHeaderLen:], uint64(len(uncompressed))) + n2 := w.encodeBlock(obuf[obufHeaderLen+n:], uncompressed) + + if n2 > 0 { + chunkType = uint8(chunkTypeCompressedData) + chunkLen = 4 + n + n2 + obuf = obuf[:obufHeaderLen+n+n2] + } else { + obuf = obuf[:8] + } + + // Fill in the per-chunk header that comes before the body. + obuf[0] = chunkType + obuf[1] = uint8(chunkLen >> 0) + obuf[2] = uint8(chunkLen >> 8) + obuf[3] = uint8(chunkLen >> 16) + obuf[4] = uint8(checksum >> 0) + obuf[5] = uint8(checksum >> 8) + obuf[6] = uint8(checksum >> 16) + obuf[7] = uint8(checksum >> 24) + + n, err := w.writer.Write(obuf) + if err != nil { + return 0, w.err(err) + } + if n != len(obuf) { + return 0, w.err(io.ErrShortWrite) + } + w.err(w.index.add(w.written, w.uncompWritten)) + w.written += int64(n) + w.uncompWritten += int64(len(uncompressed)) + + if chunkType == chunkTypeUncompressedData { + // Write uncompressed data. + n, err := w.writer.Write(uncompressed) + if err != nil { + return 0, w.err(err) + } + if n != len(uncompressed) { + return 0, w.err(io.ErrShortWrite) + } + w.written += int64(n) + } + w.buffers.Put(obuf) + // Queue final output. + nRet += len(uncompressed) + } + return nRet, nil +} + +// AsyncFlush writes any buffered bytes to a block and starts compressing it. +// It does not wait for the output has been written as Flush() does. +func (w *Writer) AsyncFlush() error { + if err := w.err(nil); err != nil { + return err + } + + // Queue any data still in input buffer. + if len(w.ibuf) != 0 { + if !w.wroteStreamHeader { + _, err := w.writeSync(w.ibuf) + w.ibuf = w.ibuf[:0] + return w.err(err) + } else { + _, err := w.write(w.ibuf) + w.ibuf = w.ibuf[:0] + err = w.err(err) + if err != nil { + return err + } + } + } + return w.err(nil) +} + +// Flush flushes the Writer to its underlying io.Writer. +// This does not apply padding. +func (w *Writer) Flush() error { + if err := w.AsyncFlush(); err != nil { + return err + } + if w.output == nil { + return w.err(nil) + } + + // Send empty buffer + res := make(chan result) + w.output <- res + // Block until this has been picked up. + res <- result{b: nil, startOffset: w.uncompWritten} + // When it is closed, we have flushed. + <-res + return w.err(nil) +} + +// Close calls Flush and then closes the Writer. +// Calling Close multiple times is ok, +// but calling CloseIndex after this will make it not return the index. +func (w *Writer) Close() error { + _, err := w.closeIndex(w.appendIndex) + return err +} + +// CloseIndex calls Close and returns an index on first call. +// This is not required if you are only adding index to a stream. +func (w *Writer) CloseIndex() ([]byte, error) { + return w.closeIndex(true) +} + +func (w *Writer) closeIndex(idx bool) ([]byte, error) { + err := w.Flush() + if w.output != nil { + close(w.output) + w.writerWg.Wait() + w.output = nil + } + + var index []byte + if w.err(err) == nil && w.writer != nil { + // Create index. + if idx { + compSize := int64(-1) + if w.pad <= 1 { + compSize = w.written + } + index = w.index.appendTo(w.ibuf[:0], w.uncompWritten, compSize) + // Count as written for padding. + if w.appendIndex { + w.written += int64(len(index)) + } + } + + if w.pad > 1 { + tmp := w.ibuf[:0] + if len(index) > 0 { + // Allocate another buffer. + tmp = w.buffers.Get().([]byte)[:0] + defer w.buffers.Put(tmp) + } + add := calcSkippableFrame(w.written, int64(w.pad)) + frame, err := skippableFrame(tmp, add, w.randSrc) + if err = w.err(err); err != nil { + return nil, err + } + n, err2 := w.writer.Write(frame) + if err2 == nil && n != len(frame) { + err2 = io.ErrShortWrite + } + _ = w.err(err2) + } + if len(index) > 0 && w.appendIndex { + n, err2 := w.writer.Write(index) + if err2 == nil && n != len(index) { + err2 = io.ErrShortWrite + } + _ = w.err(err2) + } + } + err = w.err(errClosed) + if err == errClosed { + return index, nil + } + return nil, err +} + +// calcSkippableFrame will return a total size to be added for written +// to be divisible by multiple. +// The value will always be > skippableFrameHeader. +// The function will panic if written < 0 or wantMultiple <= 0. +func calcSkippableFrame(written, wantMultiple int64) int { + if wantMultiple <= 0 { + panic("wantMultiple <= 0") + } + if written < 0 { + panic("written < 0") + } + leftOver := written % wantMultiple + if leftOver == 0 { + return 0 + } + toAdd := wantMultiple - leftOver + for toAdd < skippableFrameHeader { + toAdd += wantMultiple + } + return int(toAdd) +} + +// skippableFrame will add a skippable frame with a total size of bytes. +// total should be >= skippableFrameHeader and < maxBlockSize + skippableFrameHeader +func skippableFrame(dst []byte, total int, r io.Reader) ([]byte, error) { + if total == 0 { + return dst, nil + } + if total < skippableFrameHeader { + return dst, fmt.Errorf("s2: requested skippable frame (%d) < 4", total) + } + if int64(total) >= maxBlockSize+skippableFrameHeader { + return dst, fmt.Errorf("s2: requested skippable frame (%d) >= max 1<<24", total) + } + // Chunk type 0xfe "Section 4.4 Padding (chunk type 0xfe)" + dst = append(dst, chunkTypePadding) + f := uint32(total - skippableFrameHeader) + // Add chunk length. + dst = append(dst, uint8(f), uint8(f>>8), uint8(f>>16)) + // Add data + start := len(dst) + dst = append(dst, make([]byte, f)...) + _, err := io.ReadFull(r, dst[start:]) + return dst, err +} + +var errClosed = errors.New("s2: Writer is closed") + +// WriterOption is an option for creating a encoder. +type WriterOption func(*Writer) error + +// WriterConcurrency will set the concurrency, +// meaning the maximum number of decoders to run concurrently. +// The value supplied must be at least 1. +// By default this will be set to GOMAXPROCS. +func WriterConcurrency(n int) WriterOption { + return func(w *Writer) error { + if n <= 0 { + return errors.New("concurrency must be at least 1") + } + w.concurrency = n + return nil + } +} + +// WriterAddIndex will append an index to the end of a stream +// when it is closed. +func WriterAddIndex() WriterOption { + return func(w *Writer) error { + w.appendIndex = true + return nil + } +} + +// WriterBetterCompression will enable better compression. +// EncodeBetter compresses better than Encode but typically with a +// 10-40% speed decrease on both compression and decompression. +func WriterBetterCompression() WriterOption { + return func(w *Writer) error { + w.level = levelBetter + return nil + } +} + +// WriterBestCompression will enable better compression. +// EncodeBetter compresses better than Encode but typically with a +// big speed decrease on compression. +func WriterBestCompression() WriterOption { + return func(w *Writer) error { + w.level = levelBest + return nil + } +} + +// WriterUncompressed will bypass compression. +// The stream will be written as uncompressed blocks only. +// If concurrency is > 1 CRC and output will still be done async. +func WriterUncompressed() WriterOption { + return func(w *Writer) error { + w.level = levelUncompressed + return nil + } +} + +// WriterBlockSize allows to override the default block size. +// Blocks will be this size or smaller. +// Minimum size is 4KB and maximum size is 4MB. +// +// Bigger blocks may give bigger throughput on systems with many cores, +// and will increase compression slightly, but it will limit the possible +// concurrency for smaller payloads for both encoding and decoding. +// Default block size is 1MB. +// +// When writing Snappy compatible output using WriterSnappyCompat, +// the maximum block size is 64KB. +func WriterBlockSize(n int) WriterOption { + return func(w *Writer) error { + if w.snappy && n > maxSnappyBlockSize || n < minBlockSize { + return errors.New("s2: block size too large. Must be <= 64K and >=4KB on for snappy compatible output") + } + if n > maxBlockSize || n < minBlockSize { + return errors.New("s2: block size too large. Must be <= 4MB and >=4KB") + } + w.blockSize = n + return nil + } +} + +// WriterPadding will add padding to all output so the size will be a multiple of n. +// This can be used to obfuscate the exact output size or make blocks of a certain size. +// The contents will be a skippable frame, so it will be invisible by the decoder. +// n must be > 0 and <= 4MB. +// The padded area will be filled with data from crypto/rand.Reader. +// The padding will be applied whenever Close is called on the writer. +func WriterPadding(n int) WriterOption { + return func(w *Writer) error { + if n <= 0 { + return fmt.Errorf("s2: padding must be at least 1") + } + // No need to waste our time. + if n == 1 { + w.pad = 0 + } + if n > maxBlockSize { + return fmt.Errorf("s2: padding must less than 4MB") + } + w.pad = n + return nil + } +} + +// WriterPaddingSrc will get random data for padding from the supplied source. +// By default crypto/rand is used. +func WriterPaddingSrc(reader io.Reader) WriterOption { + return func(w *Writer) error { + w.randSrc = reader + return nil + } +} + +// WriterSnappyCompat will write snappy compatible output. +// The output can be decompressed using either snappy or s2. +// If block size is more than 64KB it is set to that. +func WriterSnappyCompat() WriterOption { + return func(w *Writer) error { + w.snappy = true + if w.blockSize > 64<<10 { + // We choose 8 bytes less than 64K, since that will make literal emits slightly more effective. + // And allows us to skip some size checks. + w.blockSize = (64 << 10) - 8 + } + return nil + } +} + +// WriterFlushOnWrite will compress blocks on each call to the Write function. +// +// This is quite inefficient as blocks size will depend on the write size. +// +// Use WriterConcurrency(1) to also make sure that output is flushed. +// When Write calls return, otherwise they will be written when compression is done. +func WriterFlushOnWrite() WriterOption { + return func(w *Writer) error { + w.flushOnWrite = true + return nil + } +} + +// WriterCustomEncoder allows to override the encoder for blocks on the stream. +// The function must compress 'src' into 'dst' and return the bytes used in dst as an integer. +// Block size (initial varint) should not be added by the encoder. +// Returning value 0 indicates the block could not be compressed. +// Returning a negative value indicates that compression should be attempted. +// The function should expect to be called concurrently. +func WriterCustomEncoder(fn func(dst, src []byte) int) WriterOption { + return func(w *Writer) error { + w.customEnc = fn + return nil + } +} diff --git a/vendor/github.com/golang/snappy/.gitignore b/vendor/github.com/klauspost/compress/snappy/.gitignore similarity index 100% rename from vendor/github.com/golang/snappy/.gitignore rename to vendor/github.com/klauspost/compress/snappy/.gitignore diff --git a/vendor/github.com/golang/snappy/AUTHORS b/vendor/github.com/klauspost/compress/snappy/AUTHORS similarity index 100% rename from vendor/github.com/golang/snappy/AUTHORS rename to vendor/github.com/klauspost/compress/snappy/AUTHORS diff --git a/vendor/github.com/golang/snappy/CONTRIBUTORS b/vendor/github.com/klauspost/compress/snappy/CONTRIBUTORS similarity index 100% rename from vendor/github.com/golang/snappy/CONTRIBUTORS rename to vendor/github.com/klauspost/compress/snappy/CONTRIBUTORS diff --git a/vendor/github.com/golang/snappy/LICENSE b/vendor/github.com/klauspost/compress/snappy/LICENSE similarity index 100% rename from vendor/github.com/golang/snappy/LICENSE rename to vendor/github.com/klauspost/compress/snappy/LICENSE diff --git a/vendor/github.com/klauspost/compress/snappy/README.md b/vendor/github.com/klauspost/compress/snappy/README.md new file mode 100644 index 000000000..8271bbd09 --- /dev/null +++ b/vendor/github.com/klauspost/compress/snappy/README.md @@ -0,0 +1,17 @@ +# snappy + +The Snappy compression format in the Go programming language. + +This is a drop-in replacement for `github.com/golang/snappy`. + +It provides a full, compatible replacement of the Snappy package by simply changing imports. + +See [Snappy Compatibility](https://github.com/klauspost/compress/tree/master/s2#snappy-compatibility) in the S2 documentation. + +"Better" compression mode is used. For buffered streams concurrent compression is used. + +For more options use the [s2 package](https://pkg.go.dev/github.com/klauspost/compress/s2). + +# usage + +Replace imports `github.com/golang/snappy` with `github.com/klauspost/compress/snappy`. diff --git a/vendor/github.com/klauspost/compress/snappy/decode.go b/vendor/github.com/klauspost/compress/snappy/decode.go new file mode 100644 index 000000000..89f1fa234 --- /dev/null +++ b/vendor/github.com/klauspost/compress/snappy/decode.go @@ -0,0 +1,60 @@ +// Copyright 2011 The Snappy-Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package snappy + +import ( + "io" + + "github.com/klauspost/compress/s2" +) + +var ( + // ErrCorrupt reports that the input is invalid. + ErrCorrupt = s2.ErrCorrupt + // ErrTooLarge reports that the uncompressed length is too large. + ErrTooLarge = s2.ErrTooLarge + // ErrUnsupported reports that the input isn't supported. + ErrUnsupported = s2.ErrUnsupported +) + +const ( + // maxBlockSize is the maximum size of the input to encodeBlock. It is not + // part of the wire format per se, but some parts of the encoder assume + // that an offset fits into a uint16. + // + // Also, for the framing format (Writer type instead of Encode function), + // https://github.com/google/snappy/blob/master/framing_format.txt says + // that "the uncompressed data in a chunk must be no longer than 65536 + // bytes". + maxBlockSize = 65536 +) + +// DecodedLen returns the length of the decoded block. +func DecodedLen(src []byte) (int, error) { + return s2.DecodedLen(src) +} + +// Decode returns the decoded form of src. The returned slice may be a sub- +// slice of dst if dst was large enough to hold the entire decoded block. +// Otherwise, a newly allocated slice will be returned. +// +// The dst and src must not overlap. It is valid to pass a nil dst. +// +// Decode handles the Snappy block format, not the Snappy stream format. +func Decode(dst, src []byte) ([]byte, error) { + return s2.Decode(dst, src) +} + +// NewReader returns a new Reader that decompresses from r, using the framing +// format described at +// https://github.com/google/snappy/blob/master/framing_format.txt +func NewReader(r io.Reader) *Reader { + return s2.NewReader(r, s2.ReaderMaxBlockSize(maxBlockSize)) +} + +// Reader is an io.Reader that can read Snappy-compressed bytes. +// +// Reader handles the Snappy stream format, not the Snappy block format. +type Reader = s2.Reader diff --git a/vendor/github.com/klauspost/compress/snappy/encode.go b/vendor/github.com/klauspost/compress/snappy/encode.go new file mode 100644 index 000000000..e8bd72c18 --- /dev/null +++ b/vendor/github.com/klauspost/compress/snappy/encode.go @@ -0,0 +1,59 @@ +// Copyright 2011 The Snappy-Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package snappy + +import ( + "io" + + "github.com/klauspost/compress/s2" +) + +// Encode returns the encoded form of src. The returned slice may be a sub- +// slice of dst if dst was large enough to hold the entire encoded block. +// Otherwise, a newly allocated slice will be returned. +// +// The dst and src must not overlap. It is valid to pass a nil dst. +// +// Encode handles the Snappy block format, not the Snappy stream format. +func Encode(dst, src []byte) []byte { + return s2.EncodeSnappyBetter(dst, src) +} + +// MaxEncodedLen returns the maximum length of a snappy block, given its +// uncompressed length. +// +// It will return a negative value if srcLen is too large to encode. +func MaxEncodedLen(srcLen int) int { + return s2.MaxEncodedLen(srcLen) +} + +// NewWriter returns a new Writer that compresses to w. +// +// The Writer returned does not buffer writes. There is no need to Flush or +// Close such a Writer. +// +// Deprecated: the Writer returned is not suitable for many small writes, only +// for few large writes. Use NewBufferedWriter instead, which is efficient +// regardless of the frequency and shape of the writes, and remember to Close +// that Writer when done. +func NewWriter(w io.Writer) *Writer { + return s2.NewWriter(w, s2.WriterSnappyCompat(), s2.WriterBetterCompression(), s2.WriterFlushOnWrite(), s2.WriterConcurrency(1)) +} + +// NewBufferedWriter returns a new Writer that compresses to w, using the +// framing format described at +// https://github.com/google/snappy/blob/master/framing_format.txt +// +// The Writer returned buffers writes. Users must call Close to guarantee all +// data has been forwarded to the underlying io.Writer. They may also call +// Flush zero or more times before calling Close. +func NewBufferedWriter(w io.Writer) *Writer { + return s2.NewWriter(w, s2.WriterSnappyCompat(), s2.WriterBetterCompression()) +} + +// Writer is an io.Writer that can write Snappy-compressed bytes. +// +// Writer handles the Snappy stream format, not the Snappy block format. +type Writer = s2.Writer diff --git a/vendor/github.com/golang/snappy/snappy.go b/vendor/github.com/klauspost/compress/snappy/snappy.go similarity index 60% rename from vendor/github.com/golang/snappy/snappy.go rename to vendor/github.com/klauspost/compress/snappy/snappy.go index ece692ea4..398cdc95a 100644 --- a/vendor/github.com/golang/snappy/snappy.go +++ b/vendor/github.com/klauspost/compress/snappy/snappy.go @@ -17,11 +17,7 @@ // // The canonical, C++ implementation is at https://github.com/google/snappy and // it only implements the block format. -package snappy // import "github.com/golang/snappy" - -import ( - "hash/crc32" -) +package snappy /* Each encoded block begins with the varint-encoded length of the decoded data, @@ -48,51 +44,3 @@ Lempel-Ziv compression algorithms. In particular: [1, 65). The length is 1 + m. The offset is the little-endian unsigned integer denoted by the next 4 bytes. */ -const ( - tagLiteral = 0x00 - tagCopy1 = 0x01 - tagCopy2 = 0x02 - tagCopy4 = 0x03 -) - -const ( - checksumSize = 4 - chunkHeaderSize = 4 - magicChunk = "\xff\x06\x00\x00" + magicBody - magicBody = "sNaPpY" - - // maxBlockSize is the maximum size of the input to encodeBlock. It is not - // part of the wire format per se, but some parts of the encoder assume - // that an offset fits into a uint16. - // - // Also, for the framing format (Writer type instead of Encode function), - // https://github.com/google/snappy/blob/master/framing_format.txt says - // that "the uncompressed data in a chunk must be no longer than 65536 - // bytes". - maxBlockSize = 65536 - - // maxEncodedLenOfMaxBlockSize equals MaxEncodedLen(maxBlockSize), but is - // hard coded to be a const instead of a variable, so that obufLen can also - // be a const. Their equivalence is confirmed by - // TestMaxEncodedLenOfMaxBlockSize. - maxEncodedLenOfMaxBlockSize = 76490 - - obufHeaderLen = len(magicChunk) + checksumSize + chunkHeaderSize - obufLen = obufHeaderLen + maxEncodedLenOfMaxBlockSize -) - -const ( - chunkTypeCompressedData = 0x00 - chunkTypeUncompressedData = 0x01 - chunkTypePadding = 0xfe - chunkTypeStreamIdentifier = 0xff -) - -var crcTable = crc32.MakeTable(crc32.Castagnoli) - -// crc implements the checksum specified in section 3 of -// https://github.com/google/snappy/blob/master/framing_format.txt -func crc(b []byte) uint32 { - c := crc32.Update(0, crcTable, b) - return uint32(c>>15|c<<17) + 0xa282ead8 -} diff --git a/vendor/github.com/onsi/ginkgo/v2/types/version.go b/vendor/github.com/onsi/ginkgo/v2/types/version.go index c6af20b68..6dfb25f24 100644 --- a/vendor/github.com/onsi/ginkgo/v2/types/version.go +++ b/vendor/github.com/onsi/ginkgo/v2/types/version.go @@ -1,3 +1,3 @@ package types -const VERSION = "2.20.0" +const VERSION = "2.20.2" diff --git a/vendor/modules.txt b/vendor/modules.txt index 78ec5aa7b..dc3931ae8 100644 --- a/vendor/modules.txt +++ b/vendor/modules.txt @@ -11,8 +11,8 @@ github.com/cpuguy83/go-md2man/v2/md2man # github.com/davecgh/go-spew v1.1.1 ## explicit github.com/davecgh/go-spew/spew -# github.com/dgraph-io/badger/v4 v4.2.0 => github.com/MixinNetwork/badger/v4 v4.2.0-F1 -## explicit; go 1.20 +# github.com/dgraph-io/badger/v4 v4.3.0 => github.com/MixinNetwork/badger/v4 v4.3.0-F1 +## explicit; go 1.22.6 github.com/dgraph-io/badger/v4 github.com/dgraph-io/badger/v4/fb github.com/dgraph-io/badger/v4/options @@ -21,8 +21,8 @@ github.com/dgraph-io/badger/v4/skl github.com/dgraph-io/badger/v4/table github.com/dgraph-io/badger/v4/trie github.com/dgraph-io/badger/v4/y -# github.com/dgraph-io/ristretto v0.1.1 -## explicit; go 1.12 +# github.com/dgraph-io/ristretto v0.1.2-0.20240116140435-c67e07994f91 +## explicit; go 1.19 github.com/dgraph-io/ristretto github.com/dgraph-io/ristretto/z github.com/dgraph-io/ristretto/z/simd @@ -37,22 +37,14 @@ github.com/go-task/slim-sprig/v3 # github.com/gogo/protobuf v1.3.2 ## explicit; go 1.15 github.com/gogo/protobuf/proto -# github.com/golang/glog v1.2.2 -## explicit; go 1.19 -github.com/golang/glog -github.com/golang/glog/internal/logsink -github.com/golang/glog/internal/stackdump # github.com/golang/protobuf v1.5.4 ## explicit; go 1.17 github.com/golang/protobuf/proto -# github.com/golang/snappy v0.0.4 -## explicit -github.com/golang/snappy # github.com/google/flatbuffers v24.3.25+incompatible ## explicit github.com/google/flatbuffers/go -# github.com/google/pprof v0.0.0-20240727154555-813a5fbdbec8 -## explicit; go 1.19 +# github.com/google/pprof v0.0.0-20240829160300-da1f7e9f2b25 +## explicit; go 1.22 github.com/google/pprof/profile # github.com/klauspost/compress v1.17.9 ## explicit; go 1.20 @@ -60,14 +52,17 @@ github.com/klauspost/compress github.com/klauspost/compress/fse github.com/klauspost/compress/huff0 github.com/klauspost/compress/internal/cpuinfo +github.com/klauspost/compress/internal/race github.com/klauspost/compress/internal/snapref +github.com/klauspost/compress/s2 +github.com/klauspost/compress/snappy github.com/klauspost/compress/zstd github.com/klauspost/compress/zstd/internal/xxhash # github.com/klauspost/cpuid/v2 v2.2.8 ## explicit; go 1.15 github.com/klauspost/cpuid/v2 -# github.com/onsi/ginkgo/v2 v2.20.0 -## explicit; go 1.20 +# github.com/onsi/ginkgo/v2 v2.20.2 +## explicit; go 1.22 github.com/onsi/ginkgo/v2/config github.com/onsi/ginkgo/v2/formatter github.com/onsi/ginkgo/v2/ginkgo @@ -150,7 +145,7 @@ golang.org/x/crypto/hkdf golang.org/x/crypto/internal/alias golang.org/x/crypto/internal/poly1305 golang.org/x/crypto/sha3 -# golang.org/x/exp v0.0.0-20240808152545-0cdaa3abc0fa +# golang.org/x/exp v0.0.0-20240823005443-9b4947da3948 ## explicit; go 1.20 golang.org/x/exp/rand # golang.org/x/mod v0.20.0 @@ -225,4 +220,4 @@ google.golang.org/protobuf/types/gofeaturespb # gopkg.in/yaml.v3 v3.0.1 ## explicit gopkg.in/yaml.v3 -# github.com/dgraph-io/badger/v4 => github.com/MixinNetwork/badger/v4 v4.2.0-F1 +# github.com/dgraph-io/badger/v4 => github.com/MixinNetwork/badger/v4 v4.3.0-F1