dotprod unsupported for apple-silicon

[XInTheDark]

Describe the issue

When building with make build ARCH=apple-silicon on an Apple M1 chip, I get the following error:

/var/folders/5m/v_lfr66s1mg7l9qhv3mjfkfc0000gn/T//ccCMWtGb.s:641:2: error: instruction requires: dotprod
        sdot    v0.4s, v1.16b, v2.16b
        ^
/var/folders/5m/v_lfr66s1mg7l9qhv3mjfkfc0000gn/T//ccCMWtGb.s:783:2: error: instruction requires: dotprod
        sdot    v0.4s, v3.16b, v4.16b
        ^

(etc.)

make[1]: *** [evaluate_nnue.o] Error 1
make: *** [build] Error 2

Expected behavior

Build should complete successfully.

Steps to reproduce

make build ARCH=apple-silicon

Anything else?

This seems to be caused by PR #4400, which claims to add support for apple-silicon dot product instructions, but this is not working properly.

Operating system

MacOS

Stockfish version

472e726

[vondele]

can you provide the output of cat /proc/cpuinfo (if the file exists on MacOS), and the version of the compiler used?

[Disservin]

Cant reproduce on my M1 make -j build ARCH=apple-silicon and cat /proc/cpuinfo doesnt work. But theres this sysctl -a | grep machdep.cpu which doesnt give you much information.

machdep.cpu.cores_per_package: 8
machdep.cpu.core_count: 8
machdep.cpu.logical_per_package: 8
machdep.cpu.thread_count: 8
machdep.cpu.brand_string: Apple M1

[XInTheDark]

can you provide the output of cat /proc/cpuinfo (if the file exists on MacOS), and the version of the compiler used?

The file doesn't exist on my machine either. Compiler version is gcc version 12.2.0 (Homebrew GCC 12.2.0).

[vondele]

actually the error message might be coming from from the assembler or down the toolchain in binutils. Can you see what additional output you get with CXXFLAGS="-v" make build ARCH=apple-silicon

[XInTheDark]

Weirdly enough, after a restart of the machine the issue no longer occurs. I'll close this issue then.

[XInTheDark]

I ran CXXFLAGS="-v" make build ARCH=apple-silicon before the restart though, and part of the output was this:

warning: MPC header version 1.2.1 differs from library version 1.3.1.
GGC heuristics: --param ggc-min-expand=100 --param ggc-min-heapsize=131072
ignoring nonexistent directory "/opt/homebrew/Cellar/gcc/12.2.0/bin/../lib/gcc/current/gcc/aarch64-apple-darwin22/12/../../../../../../aarch64-apple-darwin22/include"
ignoring duplicate directory "/opt/homebrew/Cellar/gcc/12.2.0/bin/../lib/gcc/current/gcc/../../../../lib/gcc/current/gcc/aarch64-apple-darwin22/12/../../../../../../include/c++/12"
ignoring duplicate directory "/opt/homebrew/Cellar/gcc/12.2.0/bin/../lib/gcc/current/gcc/../../../../lib/gcc/current/gcc/aarch64-apple-darwin22/12/../../../../../../include/c++/12/aarch64-apple-darwin22"
ignoring duplicate directory "/opt/homebrew/Cellar/gcc/12.2.0/bin/../lib/gcc/current/gcc/../../../../lib/gcc/current/gcc/aarch64-apple-darwin22/12/../../../../../../include/c++/12/backward"
ignoring duplicate directory "/opt/homebrew/Cellar/gcc/12.2.0/bin/../lib/gcc/current/gcc/../../../../lib/gcc/current/gcc/aarch64-apple-darwin22/12/include"
ignoring nonexistent directory "/Library/Developer/CommandLineTools/SDKs/MacOSX13.sdk/usr/local/include"
ignoring duplicate directory "/opt/homebrew/Cellar/gcc/12.2.0/bin/../lib/gcc/current/gcc/../../../../lib/gcc/current/gcc/aarch64-apple-darwin22/12/include-fixed"
ignoring nonexistent directory "/opt/homebrew/Cellar/gcc/12.2.0/bin/../lib/gcc/current/gcc/../../../../lib/gcc/current/gcc/aarch64-apple-darwin22/12/../../../../../../aarch64-apple-darwin22/include"
ignoring nonexistent directory "/Library/Developer/CommandLineTools/SDKs/MacOSX13.sdk/Library/Frameworks"
#include "..." search starts here:
#include <...> search starts here:
 /opt/homebrew/Cellar/gcc/12.2.0/bin/../lib/gcc/current/gcc/aarch64-apple-darwin22/12/../../../../../../include/c++/12
 /opt/homebrew/Cellar/gcc/12.2.0/bin/../lib/gcc/current/gcc/aarch64-apple-darwin22/12/../../../../../../include/c++/12/aarch64-apple-darwin22
 /opt/homebrew/Cellar/gcc/12.2.0/bin/../lib/gcc/current/gcc/aarch64-apple-darwin22/12/../../../../../../include/c++/12/backward
 /opt/homebrew/Cellar/gcc/12.2.0/bin/../lib/gcc/current/gcc/aarch64-apple-darwin22/12/include
 /opt/homebrew/Cellar/gcc/12.2.0/bin/../lib/gcc/current/gcc/aarch64-apple-darwin22/12/include-fixed
 /Library/Developer/CommandLineTools/SDKs/MacOSX13.sdk/usr/include
 /Library/Developer/CommandLineTools/SDKs/MacOSX13.sdk/System/Library/Frameworks
End of search list.
GNU C++17 (Homebrew GCC 12.2.0) version 12.2.0 (aarch64-apple-darwin22)
    compiled by GNU C version 12.2.0, GMP version 6.2.1, MPFR version 4.1.0-p13, MPC version 1.2.1, isl version isl-0.25-GMP

warning: MPC header version 1.2.1 differs from library version 1.3.1.
GGC heuristics: --param ggc-min-expand=100 --param ggc-min-heapsize=131072
Compiler executable checksum: 92be565a621eebe8268e8cf75aa62159
COLLECT_GCC_OPTIONS='-v' '-Wall' '-Wcast-qual' '-fno-exceptions' '-std=c++17' '-Wpedantic' '-Wextra' '-Wshadow' '-Wmissing-declarations' '-mabi=darwinpcs' '-D' 'USE_PTHREADS' '-D' 'NDEBUG' '-O3' '-D' 'IS_64BIT' '-D' 'USE_POPCNT' '-D' 'USE_NEON=8' '-march=armv8.2-a+dotprod' '-D' 'USE_NEON_DOTPROD' '-D' 'GIT_SHA="472e726b"' '-D' 'GIT_DATE="20230224"' '-flto' '-flto-partition=one' '-c' '-o' 'evaluate_nnue.o' '-mmacosx-version-min=10.14'  '-nodefaultexport' '-mlittle-endian'
 as -arch arm64 -v -mmacosx-version-min=10.14 -o evaluate_nnue.o /var/folders/5m/v_lfr66s1mg7l9qhv3mjfkfc0000gn/T//ccIVQ4sn.s
Apple clang version 14.0.0 (clang-1400.0.29.202)
Target: arm64-apple-darwin22.2.0
Thread model: posix
InstalledDir: /Library/Developer/CommandLineTools/usr/bin
 "/Library/Developer/CommandLineTools/usr/bin/clang" -cc1as -triple arm64-apple-macosx10.14.0 -filetype obj -main-file-name ccIVQ4sn.s -target-cpu apple-m1 -target-feature +v8.5a -target-feature +fp-armv8 -target-feature +neon -target-feature +crc -target-feature +crypto -target-feature +dotprod -target-feature +fp16fml -target-feature +ras -target-feature +lse -target-feature +rdm -target-feature +rcpc -target-feature +zcm -target-feature +zcz -target-feature +fullfp16 -target-feature +sm4 -target-feature +sha3 -target-feature +sha2 -target-feature +aes -fdebug-compilation-dir=/Users/jerry/Documents/Chess/Stockfish-dev-official/Stockfish/src -dwarf-debug-producer "Apple clang version 14.0.0 (clang-1400.0.29.202)" -dwarf-version=4 -mrelocation-model pic --mrelax-relocations -mllvm -disable-aligned-alloc-awareness=1 -o evaluate_nnue.o /var/folders/5m/v_lfr66s1mg7l9qhv3mjfkfc0000gn/T//ccIVQ4sn.s

/var/folders/5m/v_lfr66s1mg7l9qhv3mjfkfc0000gn/T//ccIVQ4sn.s:641:2: error: instruction requires: dotprod
        sdot    v0.4s, v1.16b, v2.16b

etc.

[XInTheDark]

This issue occurs again with latest master.

[TonHaver]

Can't reproduce on Apple M1 Pro

$ sysctl -a | grep machdep.cpu machdep.cpu.cores_per_package: 8 machdep.cpu.core_count: 8 machdep.cpu.logical_per_package: 8 machdep.cpu.thread_count: 8 machdep.cpu.brand_string: Apple M1 Pro

/Applications/Xcode.app/Contents/Developer/usr/bin/make ARCH=apple-silicon COMP=gcc all g++ -v -Wall -Wcast-qual -fno-exceptions -std=c++17 -pedantic -Wextra -Wshadow -Wmissing-declarations -m64 -mmacosx-version-min=10.14 -arch arm64 -DUSE_PTHREADS -DNDEBUG -O3 -DIS_64BIT -DUSE_POPCNT -DUSE_NEON=8 -march=armv8.2-a+dotprod -DUSE_NEON_DOTPROD -flto=full -c -o misc.o misc.cpp Apple clang version 13.1.6 (clang-1316.0.21.2.5) Target: arm64-apple-darwin21.4.0 Thread model: posix

$ ./stockfish Stockfish dev-20230303-nogit by the Stockfish developers (see AUTHORS file)

[vondele]

I believe this is not a Stockfish issue, but probably some issue with how the toolchain is setup on the user's machine. If anybody wants to help, probably best to have that exchange on discord.

[alryks]

I don't know if this would be helpful, but I had the same issue. I solved it via changing Makefile, in specific, this line: CXXFLAGS += -march=armv8.2-a+dotprod -DUSE_NEON_DOTPROD To this: CXXFLAGS += -march=armv8-a+dotprod -DUSE_NEON_DOTPROD

[vondele]

OK, this latest suggestion might require a closer look, I'm reopening this issue.