FastLanes

.github/workflows/bench-pr.yml

@@ -22,6 +22,8 @@ jobs:
           labels: benchmark
 
       - uses: actions/checkout@v4
+        with:
+          submodules: recursive
 
       - uses: ./.github/actions/setup-zig
       - uses: ./.github/actions/setup-rust

.github/workflows/bench.yml

@@ -16,6 +16,8 @@ jobs:
     if: ${{ github.event_name == 'workflow_dispatch' || (contains(github.event.head_commit.message, '[benchmark]') && github.ref_name == 'develop') }}
     steps:
       - uses: actions/checkout@v4
+        with:
+          submodules: recursive
 
       - uses: ./.github/actions/setup-zig
       - uses: ./.github/actions/setup-rust

.github/workflows/ci.yml

@@ -17,6 +17,8 @@ jobs:
     runs-on: ubuntu-latest-medium
     steps:
       - uses: actions/checkout@v4
+        with:
+          submodules: recursive
       - uses: ./.github/actions/setup-zig
       - uses: ./.github/actions/setup-rust
       - uses: ./.github/actions/setup-python

.gitmodules

@@ -0,0 +1,3 @@
+[submodule "deps/fastlanez"]
+	path = deps/fastlanez
+	url = https://github.com/fulcrum-so/fastlanez.git

Cargo.toml

@@ -3,11 +3,13 @@ members = [
     "bench-vortex",
     "codecz",
     "codecz-sys",
+    "fastlanez-sys",
     "pyvortex",
     "vortex",
     "vortex-alloc",
     "vortex-alp",
     "vortex-dict",
+    "vortex-fastlanes",
     "vortex-ffor",
     "vortex-ree",
     "vortex-roaring",

bench-vortex/.gitignore

@@ -0,0 +1 @@
+data

bench-vortex/Cargo.toml

@@ -11,6 +11,7 @@ arrow-array = "50.0.0"
 vortex = { path = "../vortex" }
 vortex-alp = { path = "../vortex-alp" }
 vortex-dict = { path = "../vortex-dict" }
+vortex-fastlanes = { path = "../vortex-fastlanes" }
 vortex-ffor = { path = "../vortex-ffor" }
 vortex-ree = { path = "../vortex-ree" }
 vortex-roaring = { path = "../vortex-roaring" }

bench-vortex/benches/compress_benchmark.rs

@@ -29,11 +29,10 @@ use vortex::array::{Array, ArrayRef};
 use vortex::compress::CompressCtx;
 use vortex::dtype::DType;
 use vortex::error::{VortexError, VortexResult};
-
 use vortex_bench::enumerate_arrays;
 
 fn download_taxi_data() -> &'static Path {
-    let download_path = Path::new("../../pyspiral/bench/.data/https-d37ci6vzurychx-cloudfront-net-trip-data-yellow-tripdata-2023-11.parquet");
+    let download_path = Path::new("data/yellow-tripdata-2023-11.parquet");
     if download_path.exists() {
         return download_path;
     }

bench-vortex/src/lib.rs

@@ -17,7 +17,7 @@ use itertools::Itertools;
 use vortex::array::Encoding;
 use vortex_alp::ALPEncoding;
 use vortex_dict::DictEncoding;
-use vortex_ffor::FFoREncoding;
+use vortex_fastlanes::{BitPackedEncoding, FoREncoding};
 use vortex_ree::REEEncoding;
 use vortex_roaring::{RoaringBoolEncoding, RoaringIntEncoding};
 use vortex_zigzag::ZigZagEncoding;
@@ -26,7 +26,9 @@ pub fn enumerate_arrays() {
     let encodings: Vec<&dyn Encoding> = vec![
         &ALPEncoding,
         &DictEncoding,
-        &FFoREncoding,
+        &BitPackedEncoding,
+        &FoREncoding,
+        //&FFoREncoding,
         &REEEncoding,
         &RoaringBoolEncoding,
         &RoaringIntEncoding,
@@ -46,15 +48,16 @@ mod test {
     use parquet::arrow::arrow_reader::ParquetRecordBatchReaderBuilder;
     use simplelog::{ColorChoice, Config, TermLogger, TerminalMode};
 
-    use crate::enumerate_arrays;
     use vortex::array::chunked::ChunkedArray;
     use vortex::array::{Array, ArrayRef};
     use vortex::compress::CompressCtx;
     use vortex::dtype::DType;
     use vortex::error::{VortexError, VortexResult};
 
+    use crate::enumerate_arrays;
+
     pub fn download_taxi_data() -> &'static Path {
-        let download_path = Path::new("../../pyspiral/bench/.data/https-d37ci6vzurychx-cloudfront-net-trip-data-yellow-tripdata-2023-11.parquet");
+        let download_path = Path::new("data/yellow-tripdata-2023-11.parquet");
         if download_path.exists() {
             return download_path;
         }

codecz/src/utils.rs

@@ -14,22 +14,24 @@
 
 extern crate alloc;
 
-use crate::{AlignedVec, ALIGNED_ALLOCATOR};
 use alloc::sync::Arc;
-use arrow_buffer::{BooleanBuffer, Buffer};
 use core::ptr::NonNull;
 
-pub fn into_u32_vec(bb: &BooleanBuffer, cardinality: usize) -> AlignedVec<u32> {
-    let mut vec: AlignedVec<u32> = AlignedVec::with_capacity_in(cardinality, ALIGNED_ALLOCATOR);
+use arrow_buffer::{BooleanBuffer, Buffer};
+
+use crate::{AlignedVec, ALIGNED_ALLOCATOR};
+
+pub fn into_u64_vec(bb: &BooleanBuffer, cardinality: usize) -> AlignedVec<u64> {
+    let mut vec: AlignedVec<u64> = AlignedVec::with_capacity_in(cardinality, ALIGNED_ALLOCATOR);
     if cardinality > 0 {
         for idx in bb.set_indices() {
-            vec.push(idx as u32);
+            vec.push(idx as u64);
         }
     }
     vec
 }
 
-pub fn gather_patches<T: Copy + Sized>(data: &[T], indices: &[u32]) -> AlignedVec<T> {
+pub fn gather_patches<T: Copy + Sized>(data: &[T], indices: &[u64]) -> AlignedVec<T> {
     let mut vec: AlignedVec<T> = AlignedVec::with_capacity_in(indices.len(), ALIGNED_ALLOCATOR);
     for idx in indices {
         vec.push(data[*idx as usize]);

fastlanez-sys/Cargo.toml

@@ -0,0 +1,18 @@
+[package]
+name = "fastlanez-sys"
+version = "0.1.0"
+edition = "2021"
+links = "fastlanez"
+
+[lints]
+workspace = true
+
+[dependencies]
+arrayref = "0.3.7"
+paste = "1.0.14"
+seq-macro = "0.3.5"
+uninit = "0.6.2"
+
+[build-dependencies]
+bindgen = "0.69.1"
+walkdir = "2.4.0"

fastlanez-sys/build.rs

@@ -0,0 +1,121 @@
+use std::env;
+use std::path::{Path, PathBuf};
+use std::process::{Command, Stdio};
+
+use walkdir::WalkDir;
+
+fn main() {
+    let buildrs_dir = PathBuf::from(env::var("CARGO_MANIFEST_DIR").unwrap())
+        .canonicalize()
+        .expect("Failed to canonicalize CARGO_MANIFEST_DIR");
+    let root_dir = buildrs_dir
+        .join("../")
+        .canonicalize()
+        .expect("Failed to canonicalize root dir");
+    let fastlanez_dir = root_dir.join("deps/fastlanez");
+
+    // Tell cargo to tell rustc to link codecz
+    println!(
+        "cargo:rustc-link-search={}",
+        fastlanez_dir.join("zig-out/lib").to_str().unwrap()
+    );
+    println!("cargo:rustc-link-lib=fastlanez");
+
+    rerun_if_changed(&buildrs_dir.join("build.rs"));
+    WalkDir::new(fastlanez_dir.join("src"))
+        .into_iter()
+        .filter_map(|e| e.ok())
+        .for_each(|e| rerun_if_changed(e.path()));
+
+    let zig_opt = get_zig_opt();
+    println!("cargo:info=invoking `zig build` with {}", zig_opt);
+    if !Command::new("zig")
+        .args(["build", "lib"])
+        .arg(zig_opt)
+        .args(["--summary", "all"])
+        .current_dir(fastlanez_dir.clone())
+        .spawn()
+        .expect("Could not invoke `zig build`")
+        .wait()
+        .unwrap()
+        .success()
+    {
+        // Panic if the command was not successful.
+        panic!(
+            "failed to successfully invoke `zig build` in {}",
+            root_dir.to_str().unwrap()
+        );
+    }
+
+    let bindings = bindgen::Builder::default()
+        .header(
+            fastlanez_dir
+                .join("zig-out/include/fastlanez.h")
+                .to_str()
+                .unwrap(),
+        )
+        .clang_args(&[
+            get_zig_include().as_ref(),
+            "-DZIG_TARGET_MAX_INT_ALIGNMENT=16",
+        ])
+        .parse_callbacks(Box::new(bindgen::CargoCallbacks::new()))
+        .allowlist_item("fl_.*")
+        .generate()
+        .expect("Unable to generate bindings");
+
+    // Write the bindings to the $OUT_DIR/bindings.rs file.
+    let out_path = PathBuf::from(env::var("OUT_DIR").unwrap());
+    bindings
+        .write_to_file(out_path.join("bindings.rs"))
+        .expect("Couldn't write bindings!");
+}
+
+fn rerun_if_changed(path: &Path) {
+    println!(
+        "cargo:rerun-if-changed={}",
+        path.canonicalize()
+            .unwrap_or_else(|_| panic!("failed to canonicalize {}", path.to_str().unwrap()))
+            .to_str()
+            .unwrap()
+    );
+}
+
+fn get_zig_opt() -> &'static str {
+    let profile_env = env::var("PROFILE").unwrap();
+    let opt_level_zero = env::var("OPT_LEVEL").unwrap() == "0";
+
+    // based on https://doc.rust-lang.org/cargo/reference/environment-variables.html
+    //
+    // confusingly, the PROFILE env var will be either "debug" or "release" depending on whether the cargo profile
+    // derives from the "dev" or "release" profile, respectively. *facepalm*
+    // so `cargo build` and `cargo test` will be "debug"; `cargo build --release` and `cargo bench` will be "release"
+    //
+    // we also check whether debug_assertions are enabled (to pick a sane value for custom profiles)
+    if profile_env == "debug" || cfg!(debug_assertions) {
+        "-Doptimize=Debug"
+    } else if profile_env == "release" || !opt_level_zero {
+        "-Doptimize=ReleaseSmall"
+    } else {
+        // we're in a custom profile, the opt_level is 0, but debug assertions aren't enabled
+        // pretty weird case, let's default to debug
+        println!(
+            "cargo:warning=unrecognized cargo profile {}, defaulting to `zig build -Doptimize=Debug`", profile_env
+        );
+        "-Doptimize=Debug"
+    }
+}
+
+fn get_zig_include() -> String {
+    String::from_utf8(
+        Command::new("bash")
+            .arg("-c")
+            .arg("zig env | grep lib_dir | awk -F'\"' '{print \"-I\"$4}'")
+            .stdout(Stdio::piped())
+            .output()
+            .expect("Failed to execute command")
+            .stdout,
+    )
+    .expect("Failed to convert command output to string")
+    .trim_end()
+    .to_string()
+}