Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Fix Mhz -> MHz in README, comments and Dhrystone benchmark output #413

Merged
merged 2 commits into from
Jun 21, 2024
Merged

Fix Mhz -> MHz in README, comments and Dhrystone benchmark output #413

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
7beb988
Fix Mhz -> MHz in readme and comments
mrcmry Jun 17, 2024
8968b5a
Fix Mhz -> MHz in Dhrystone benchmark report generation
mrcmry Jun 17, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
140 changes: 70 additions & 70 deletions README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -59,7 +59,7 @@ This repository hosts a RISC-V implementation written in SpinalHDL. Here are som

- RV32I[M][A][F[D]][C] instruction set
- Pipelined from 2 to 5+ stages ([Fetch*X], Decode, Execute, [Memory], [WriteBack])
- 1.44 DMIPS/Mhz --no-inline when nearly all features are enabled (1.57 DMIPS/Mhz when the divider lookup table is enabled)
- 1.44 DMIPS/MHz --no-inline when nearly all features are enabled (1.57 DMIPS/MHz when the divider lookup table is enabled)
- Optimized for FPGA, does not use any vendor specific IP block / primitive
- AXI4, Avalon, wishbone ready
- Optional MUL/DIV extensions
Expand Down Expand Up @@ -97,54 +97,54 @@ dhrystone binaries which fit inside a 4KB I$ and 4KB D$ (I already had this case
The CPU configurations used below can be found in the `src/scala/vexriscv/demo` directory.

```
VexRiscv small (RV32I, 0.52 DMIPS/Mhz, no datapath bypass, no interrupt) ->
Artix 7 -> 243 Mhz 504 LUT 505 FF
Cyclone V -> 174 Mhz 352 ALMs
Cyclone IV -> 179 Mhz 731 LUT 494 FF
iCE40 -> 92 Mhz 1130 LC

VexRiscv small (RV32I, 0.52 DMIPS/Mhz, no datapath bypass) ->
Artix 7 -> 240 Mhz 556 LUT 566 FF
Cyclone V -> 194 Mhz 394 ALMs
Cyclone IV -> 174 Mhz 831 LUT 555 FF
iCE40 -> 85 Mhz 1292 LC

VexRiscv small and productive (RV32I, 0.82 DMIPS/Mhz) ->
Artix 7 -> 232 Mhz 816 LUT 534 FF
Cyclone V -> 155 Mhz 492 ALMs
Cyclone IV -> 155 Mhz 1,111 LUT 530 FF
iCE40 -> 63 Mhz 1596 LC

VexRiscv small and productive with I$ (RV32I, 0.70 DMIPS/Mhz, 4KB-I$) ->
Artix 7 -> 220 Mhz 730 LUT 570 FF
Cyclone V -> 142 Mhz 501 ALMs
Cyclone IV -> 150 Mhz 1,139 LUT 536 FF
iCE40 -> 66 Mhz 1680 LC

VexRiscv full no cache (RV32IM, 1.21 DMIPS/Mhz 2.30 Coremark/Mhz, single cycle barrel shifter, debug module, catch exceptions, static branch) ->
Artix 7 -> 216 Mhz 1418 LUT 949 FF
Cyclone V -> 133 Mhz 933 ALMs
Cyclone IV -> 143 Mhz 2,076 LUT 972 FF

VexRiscv full (RV32IM, 1.21 DMIPS/Mhz 2.30 Coremark/Mhz with cache trashing, 4KB-I$,4KB-D$, single cycle barrel shifter, debug module, catch exceptions, static branch) ->
Artix 7 -> 199 Mhz 1840 LUT 1158 FF
Cyclone V -> 141 Mhz 1,166 ALMs
Cyclone IV -> 131 Mhz 2,407 LUT 1,067 FF

VexRiscv full max perf (HZ*IPC) -> (RV32IM, 1.38 DMIPS/Mhz 2.57 Coremark/Mhz, 8KB-I$,8KB-D$, single cycle barrel shifter, debug module, catch exceptions, dynamic branch prediction in the fetch stage, branch and shift operations done in the Execute stage) ->
Artix 7 -> 200 Mhz 1935 LUT 1216 FF
Cyclone V -> 130 Mhz 1,166 ALMs
Cyclone IV -> 126 Mhz 2,484 LUT 1,120 FF

VexRiscv full with MMU (RV32IM, 1.24 DMIPS/Mhz 2.35 Coremark/Mhz, with cache trashing, 4KB-I$, 4KB-D$, single cycle barrel shifter, debug module, catch exceptions, dynamic branch, MMU) ->
Artix 7 -> 151 Mhz 2021 LUT 1541 FF
Cyclone V -> 124 Mhz 1,368 ALMs
Cyclone IV -> 128 Mhz 2,826 LUT 1,474 FF

VexRiscv linux balanced (RV32IMA, 1.21 DMIPS/Mhz 2.27 Coremark/Mhz, with cache trashing, 4KB-I$, 4KB-D$, single cycle barrel shifter, catch exceptions, static branch, MMU, Supervisor, Compatible with mainstream linux) ->
Artix 7 -> 180 Mhz 2883 LUT 2130 FF
Cyclone V -> 131 Mhz 1,764 ALMs
Cyclone IV -> 121 Mhz 3,608 LUT 2,082 FF
VexRiscv small (RV32I, 0.52 DMIPS/MHz, no datapath bypass, no interrupt) ->
Artix 7 -> 243 MHz 504 LUT 505 FF
Cyclone V -> 174 MHz 352 ALMs
Cyclone IV -> 179 MHz 731 LUT 494 FF
iCE40 -> 92 MHz 1130 LC

VexRiscv small (RV32I, 0.52 DMIPS/MHz, no datapath bypass) ->
Artix 7 -> 240 MHz 556 LUT 566 FF
Cyclone V -> 194 MHz 394 ALMs
Cyclone IV -> 174 MHz 831 LUT 555 FF
iCE40 -> 85 MHz 1292 LC

VexRiscv small and productive (RV32I, 0.82 DMIPS/MHz) ->
Artix 7 -> 232 MHz 816 LUT 534 FF
Cyclone V -> 155 MHz 492 ALMs
Cyclone IV -> 155 MHz 1,111 LUT 530 FF
iCE40 -> 63 MHz 1596 LC

VexRiscv small and productive with I$ (RV32I, 0.70 DMIPS/MHz, 4KB-I$) ->
Artix 7 -> 220 MHz 730 LUT 570 FF
Cyclone V -> 142 MHz 501 ALMs
Cyclone IV -> 150 MHz 1,139 LUT 536 FF
iCE40 -> 66 MHz 1680 LC

VexRiscv full no cache (RV32IM, 1.21 DMIPS/MHz 2.30 Coremark/MHz, single cycle barrel shifter, debug module, catch exceptions, static branch) ->
Artix 7 -> 216 MHz 1418 LUT 949 FF
Cyclone V -> 133 MHz 933 ALMs
Cyclone IV -> 143 MHz 2,076 LUT 972 FF

VexRiscv full (RV32IM, 1.21 DMIPS/MHz 2.30 Coremark/MHz with cache trashing, 4KB-I$,4KB-D$, single cycle barrel shifter, debug module, catch exceptions, static branch) ->
Artix 7 -> 199 MHz 1840 LUT 1158 FF
Cyclone V -> 141 MHz 1,166 ALMs
Cyclone IV -> 131 MHz 2,407 LUT 1,067 FF

VexRiscv full max perf (HZ*IPC) -> (RV32IM, 1.38 DMIPS/MHz 2.57 Coremark/MHz, 8KB-I$,8KB-D$, single cycle barrel shifter, debug module, catch exceptions, dynamic branch prediction in the fetch stage, branch and shift operations done in the Execute stage) ->
Artix 7 -> 200 MHz 1935 LUT 1216 FF
Cyclone V -> 130 MHz 1,166 ALMs
Cyclone IV -> 126 MHz 2,484 LUT 1,120 FF

VexRiscv full with MMU (RV32IM, 1.24 DMIPS/MHz 2.35 Coremark/MHz, with cache trashing, 4KB-I$, 4KB-D$, single cycle barrel shifter, debug module, catch exceptions, dynamic branch, MMU) ->
Artix 7 -> 151 MHz 2021 LUT 1541 FF
Cyclone V -> 124 MHz 1,368 ALMs
Cyclone IV -> 128 MHz 2,826 LUT 1,474 FF

VexRiscv linux balanced (RV32IMA, 1.21 DMIPS/MHz 2.27 Coremark/MHz, with cache trashing, 4KB-I$, 4KB-D$, single cycle barrel shifter, catch exceptions, static branch, MMU, Supervisor, Compatible with mainstream linux) ->
Artix 7 -> 180 MHz 2883 LUT 2130 FF
Cyclone V -> 131 MHz 1,764 ALMs
Cyclone IV -> 121 MHz 3,608 LUT 2,082 FF
```

The following configuration results in 1.44 DMIPS/MHz:
Expand All @@ -157,7 +157,7 @@ The following configuration results in 1.44 DMIPS/MHz:
- single cycle multiplication with bypassing in the WB stage (late result)
- dynamic branch prediction done in the F stage with a direct mapped target buffer cache (no penalties on correct predictions)

Note that, recently, the capability to remove the Fetch/Memory/WriteBack stage was added to reduce the area of the CPU, which ends up with a smaller CPU and a better DMIPS/Mhz for the small configurations.
Note that, recently, the capability to remove the Fetch/Memory/WriteBack stage was added to reduce the area of the CPU, which ends up with a smaller CPU and a better DMIPS/MHz for the small configurations.

## Dependencies

Expand Down Expand Up @@ -361,9 +361,9 @@ You can find some FPGA projects which instantiate the Briey SoC here (DE1-SoC, D
Here are some measurements of Briey SoC timings and area:

```
Artix 7 -> 181 Mhz 3220 LUT 3181 FF
Cyclone V -> 142 Mhz 2,222 ALMs
Cyclone IV -> 130 Mhz 4,538 LUT 3,211 FF
Artix 7 -> 181 MHz 3220 LUT 3181 FF
Cyclone V -> 142 MHz 2,222 ALMs
Cyclone IV -> 130 MHz 4,538 LUT 3,211 FF
```

## Murax SoC
Expand All @@ -379,8 +379,8 @@ Murax is a very light SoC (it fits in an ICE40 FPGA) which can work without any
- one UART with tx/rx fifo

Depending on the CPU configuration, on the ICE40-hx8k FPGA with icestorm for synthesis, the full SoC has the following area/performance:
- RV32I interlocked stages => 51 Mhz, 2387 LC 0.45 DMIPS/Mhz
- RV32I bypassed stages => 45 Mhz, 2718 LC 0.65 DMIPS/Mhz
- RV32I interlocked stages => 51 MHz, 2387 LC 0.45 DMIPS/MHz
- RV32I bypassed stages => 45 MHz, 2718 LC 0.65 DMIPS/MHz

Its implementation can be found here: `src/main/scala/vexriscv/demo/Murax.scala`.

Expand Down Expand Up @@ -415,17 +415,17 @@ You can find multiple software examples and demos here: https://github.com/Spina
Here are some timing and area measurements of the Murax SoC:

```
Murax interlocked stages (0.45 DMIPS/Mhz, 8 bits GPIO) ->
Artix 7 -> 216 Mhz 1109 LUT 1201 FF
Cyclone V -> 182 Mhz 725 ALMs
Cyclone IV -> 147 Mhz 1,551 LUT 1,223 FF
iCE40 -> 64 Mhz 2422 LC (nextpnr)

MuraxFast bypassed stages (0.65 DMIPS/Mhz, 8 bits GPIO) ->
Artix 7 -> 224 Mhz 1278 LUT 1300 FF
Cyclone V -> 173 Mhz 867 ALMs
Cyclone IV -> 143 Mhz 1,755 LUT 1,258 FF
iCE40 -> 66 Mhz 2799 LC (nextpnr)
Murax interlocked stages (0.45 DMIPS/MHz, 8 bits GPIO) ->
Artix 7 -> 216 MHz 1109 LUT 1201 FF
Cyclone V -> 182 MHz 725 ALMs
Cyclone IV -> 147 MHz 1,551 LUT 1,223 FF
iCE40 -> 64 MHz 2422 LC (nextpnr)

MuraxFast bypassed stages (0.65 DMIPS/MHz, 8 bits GPIO) ->
Artix 7 -> 224 MHz 1278 LUT 1300 FF
Cyclone V -> 173 MHz 867 ALMs
Cyclone IV -> 143 MHz 1,755 LUT 1,258 FF
iCE40 -> 66 MHz 2799 LC (nextpnr)
```

Some scripts to generate the SoC and call the icestorm toolchain can be found here: `scripts/Murax/`
Expand Down Expand Up @@ -814,11 +814,11 @@ Synthesis results of the FPU itself, without the CPU integration, on the fast sp

```
Fpu 32 bits ->
Artix 7 relaxed -> 135 Mhz 1786 LUT 1778 FF
Artix 7 FMax -> 205 Mhz 2101 LUT 1778 FF
Artix 7 relaxed -> 135 MHz 1786 LUT 1778 FF
Artix 7 FMax -> 205 MHz 2101 LUT 1778 FF
Fpu 64/32 bits ->
Artix 7 relaxed -> 101 Mhz 3336 LUT 3033 FF
Artix 7 FMax -> 165 Mhz 3728 LUT 3175 FF
Artix 7 relaxed -> 101 MHz 3336 LUT 3033 FF
Artix 7 FMax -> 165 MHz 3728 LUT 3175 FF
```

Note that if you want to debug FPU code via the openocd_riscv.vexriscv target, you need to use the GDB from :
Expand Down
4 changes: 2 additions & 2 deletions doc/gcdPeripheral/src/main/scala/vexriscv/demo/Murax.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -27,8 +27,8 @@ import vexriscv.periph.tasks.hash._
/** Created by PIC32F_USER on 28/07/2017.
*
* Murax is a very light SoC which could work without any external component.
* - ICE40-hx8k + icestorm => 53 Mhz, 2142 LC
* - 0.37 DMIPS/Mhz
* - ICE40-hx8k + icestorm => 53 MHz, 2142 LC
* - 0.37 DMIPS/MHz
* - 8 kB of on-chip ram
* - JTAG debugger (eclipse/GDB/openocd ready)
* - Interrupt support
Expand Down
4 changes: 2 additions & 2 deletions src/main/scala/vexriscv/demo/Murax.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -22,8 +22,8 @@ import scala.collection.Seq
* Created by PIC32F_USER on 28/07/2017.
*
* Murax is a very light SoC which could work without any external component.
* - ICE40-hx8k + icestorm => 53 Mhz, 2142 LC
* - 0.37 DMIPS/Mhz
* - ICE40-hx8k + icestorm => 53 MHz, 2142 LC
* - 0.37 DMIPS/MHz
* - 8 kB of on-chip ram
* - JTAG debugger (eclipse/GDB/openocd ready)
* - Interrupt support
Expand Down
72 changes: 36 additions & 36 deletions src/main/scala/vexriscv/ip/fpu/FpuCore.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1758,17 +1758,17 @@ object FpuSynthesisBench extends App{
}

// rotate2_24 ->
// Artix 7 -> 233 Mhz 96 LUT 167 FF
// Artix 7 -> 420 Mhz 86 LUT 229 FF
// Artix 7 -> 233 MHz 96 LUT 167 FF
// Artix 7 -> 420 MHz 86 LUT 229 FF
// rotate2_32 ->
// Artix 7 -> 222 Mhz 108 LUT 238 FF
// Artix 7 -> 399 Mhz 110 LUT 300 FF
// Artix 7 -> 222 MHz 108 LUT 238 FF
// Artix 7 -> 399 MHz 110 LUT 300 FF
// rotate2_52 ->
// Artix 7 -> 195 Mhz 230 LUT 362 FF
// Artix 7 -> 366 Mhz 225 LUT 486 FF
// Artix 7 -> 195 MHz 230 LUT 362 FF
// Artix 7 -> 366 MHz 225 LUT 486 FF
// rotate2_64 ->
// Artix 7 -> 182 Mhz 257 LUT 465 FF
// Artix 7 -> 359 Mhz 266 LUT 591 FF
// Artix 7 -> 182 MHz 257 LUT 465 FF
// Artix 7 -> 359 MHz 266 LUT 591 FF
class Rotate2(width : Int) extends Rtl{
override def getName(): String = "rotate2_" + width
override def getRtlPath(): String = getName() + ".v"
Expand Down Expand Up @@ -1858,56 +1858,56 @@ object FpuSynthesisBench extends App{
}

//Fpu_32 ->
//Artix 7 -> 136 Mhz 1471 LUT 1336 FF
//Artix 7 -> 196 Mhz 1687 LUT 1371 FF
//Artix 7 -> 136 MHz 1471 LUT 1336 FF
//Artix 7 -> 196 MHz 1687 LUT 1371 FF
//Fpu_64 ->
//Artix 7 -> 105 Mhz 2822 LUT 2132 FF
//Artix 7 -> 161 Mhz 3114 LUT 2272 FF
//Artix 7 -> 105 MHz 2822 LUT 2132 FF
//Artix 7 -> 161 MHz 3114 LUT 2272 FF
//
//
//
//Fpu_32 ->
//Artix 7 -> 128 Mhz 1693 LUT 1481 FF
//Artix 7 -> 203 Mhz 1895 LUT 1481 FF
//Artix 7 -> 128 MHz 1693 LUT 1481 FF
//Artix 7 -> 203 MHz 1895 LUT 1481 FF
//Fpu_64 ->
//Artix 7 -> 99 Mhz 3073 LUT 2396 FF
//Artix 7 -> 164 Mhz 3433 LUT 2432 FF
//Artix 7 -> 99 MHz 3073 LUT 2396 FF
//Artix 7 -> 164 MHz 3433 LUT 2432 FF


//Fpu_32 ->
//Artix 7 -> 112 Mhz 1790 LUT 1666 FF
//Artix 7 -> 158 Mhz 1989 LUT 1701 FF
//Artix 7 -> 112 MHz 1790 LUT 1666 FF
//Artix 7 -> 158 MHz 1989 LUT 1701 FF
//Fpu_64 ->
//Artix 7 -> 100 Mhz 3294 LUT 2763 FF
//Artix 7 -> 151 Mhz 3708 LUT 2904 FF
//Artix 7 -> 100 MHz 3294 LUT 2763 FF
//Artix 7 -> 151 MHz 3708 LUT 2904 FF

//Fpu_32 ->
//Artix 7 -> 139 Mhz 1879 LUT 1713 FF
//Artix 7 -> 206 Mhz 2135 LUT 1723 FF
//Artix 7 -> 139 MHz 1879 LUT 1713 FF
//Artix 7 -> 206 MHz 2135 LUT 1723 FF
//Fpu_64 ->
//Artix 7 -> 106 Mhz 3502 LUT 2811 FF
//Artix 7 -> 163 Mhz 3905 LUT 2951 FF
//Artix 7 -> 106 MHz 3502 LUT 2811 FF
//Artix 7 -> 163 MHz 3905 LUT 2951 FF

//Fpu_32 ->
//Artix 7 -> 130 Mhz 1889 LUT 1835 FF
//Artix 7 -> 210 Mhz 2131 LUT 1845 FF
//Artix 7 -> 130 MHz 1889 LUT 1835 FF
//Artix 7 -> 210 MHz 2131 LUT 1845 FF
//Fpu_64 ->
//Artix 7 -> 106 Mhz 3322 LUT 3023 FF
//Artix 7 -> 161 Mhz 3675 LUT 3163 FF
//Artix 7 -> 106 MHz 3322 LUT 3023 FF
//Artix 7 -> 161 MHz 3675 LUT 3163 FF

//Fpu_32 ->
//Artix 7 -> 132 Mhz 1891 LUT 1837 FF
//Artix 7 -> 209 Mhz 2132 LUT 1847 FF
//Artix 7 -> 132 MHz 1891 LUT 1837 FF
//Artix 7 -> 209 MHz 2132 LUT 1847 FF
//Fpu_64 ->
//Artix 7 -> 105 Mhz 3348 LUT 3024 FF
//Artix 7 -> 162 Mhz 3712 LUT 3165 FF
//Artix 7 -> 105 MHz 3348 LUT 3024 FF
//Artix 7 -> 162 MHz 3712 LUT 3165 FF

//Fpu_32 ->
//Artix 7 -> 128 Mhz 1796 LUT 1727 FF
//Artix 7 -> 208 Mhz 2049 LUT 1727 FF
//Artix 7 -> 128 MHz 1796 LUT 1727 FF
//Artix 7 -> 208 MHz 2049 LUT 1727 FF
//Fpu_64 ->
//Artix 7 -> 109 Mhz 3417 LUT 2913 FF
//Artix 7 -> 168 Mhz 3844 LUT 3053 FF
//Artix 7 -> 109 MHz 3417 LUT 2913 FF
//Artix 7 -> 168 MHz 3844 LUT 3053 FF

/*
testfloat -tininessafter -all1 > all1.txt
Expand Down
2 changes: 1 addition & 1 deletion src/main/scala/vexriscv/plugin/AesPlugin.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -53,7 +53,7 @@ import vexriscv.{DecoderService, Stageable, VexRiscv}
* - SS specify which byte should be used from RS2 for the processing
*
* In practice the aes-256-cbc performances should improve by a factor 4. See the following results from libopenssl
* from a SoC running linux at 100 Mhz
* from a SoC running linux at 100 MHz
* type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes 16384 bytes
* aes-256-cbc SW 492.58k 700.22k 796.41k 831.49k 830.09k 832.81k
* aes-256 cbc HW 1781.52k 2834.07k 3323.07k 3486.72k 3465.22k 3440.10k
Expand Down
2 changes: 1 addition & 1 deletion src/test/scala/vexriscv/DhrystoneBench.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -45,7 +45,7 @@ class DhrystoneBench extends AnyFunSuite {
val coremarkIterations = intFind.findFirstIn("Iterations \\: (\\d+.?)+".r.findAllIn(str).toList.last).get.toDouble
val coremarkHzs = intFind.findFirstIn("DCLOCKS_PER_SEC=(\\d+.?)+".r.findAllIn(str).toList.last).get.toDouble
val coremarkPerMhz = 1e6 * coremarkIterations / coremarkTicks
report ++= s"$name -> $dmips DMIPS/Mhz $coremarkPerMhz Coremark/Mhz\n"
report ++= s"$name -> $dmips DMIPS/MHz $coremarkPerMhz Coremark/MHz\n"
}

}
Expand Down
Loading