Skip to content

Commit

Permalink
Update POWER Clang version docs (#5174)
Browse files Browse the repository at this point in the history
Clang prior to 13 no longer fails to build ncnn since #4845.

Clang 18 fixes SSE4.1 translation, which yields a major speedup.

Co-authored-by: Jeremy Rand <jeremyrand@danwin1210.de>
  • Loading branch information
JeremyRand and Jeremy Rand authored Nov 27, 2023
1 parent deae9e6 commit d1f6193
Showing 1 changed file with 2 additions and 2 deletions.
4 changes: 2 additions & 2 deletions docs/how-to-build/how-to-build.md
Original file line number Diff line number Diff line change
Expand Up @@ -92,7 +92,7 @@ For Rasberry Pi 3 on 32bit OS, add `-DCMAKE_TOOLCHAIN_FILE=../toolchains/pi3.too

#### POWER

For POWER9 with Clang 13 or higher:
For POWER9 with Clang:

```shell
cd ncnn
Expand All @@ -102,7 +102,7 @@ cmake -DCMAKE_BUILD_TYPE=Release -DCMAKE_TOOLCHAIN_FILE=../toolchains/power9le-l
make -j$(nproc)
```

Earlier versions of Clang may fail to build ncnn due to [Bug 49864](https://github.com/llvm/llvm-project/issues/49864). To use GCC instead, use the `power9le-linux-gnu-vsx.toolchain.cmake` toolchain file instead. Note that according to benchmarks, Clang appears to produce noticeably faster CPU inference than GCC for POWER9 targets.
To use GCC instead, use the `power9le-linux-gnu-vsx.toolchain.cmake` toolchain file instead. Note that according to benchmarks, Clang appears to produce noticeably faster CPU inference than GCC for POWER9 targets. For fastest inference, use Clang 18 or higher; earlier versions of Clang may have impaired inference speed due to [Bug 49864](https://github.com/llvm/llvm-project/issues/49864) and [Bug 64664](https://github.com/llvm/llvm-project/issues/64664).

For POWER8 instead of POWER9, use the `power8le-linux-gnu-vsx.clang.toolchain.cmake` or `power8le-linux-gnu-vsx.toolchain.cmake` toolchain file instead. POWER8 will be slower than POWER9.

Expand Down

0 comments on commit d1f6193

Please sign in to comment.