Optimize find_nnz() using AVX512

About a 1% speedup for ARCH x86-64-avx512 and x86-64-vnni512.

Note:  This could be optimized further if we wanted to add an ARCH
supporting VBMI2 which is even more modern than VNNI.

https://en.wikichip.org/wiki/x86/avx512_vbmi2

closes https://github.com/official-stockfish/Stockfish/pull/6139

No functional change
This commit is contained in:
mstembera
2025-07-02 18:41:45 +02:00
committed by Disservin
parent ea85a54fef
commit ce7254b5ea
2 changed files with 36 additions and 2 deletions
+1 -1
View File
@@ -701,7 +701,7 @@ endif
ifeq ($(avx512),yes)
CXXFLAGS += -DUSE_AVX512
ifeq ($(comp),$(filter $(comp),gcc clang mingw icx))
CXXFLAGS += -mavx512f -mavx512bw
CXXFLAGS += -mavx512f -mavx512bw -mavx512dq -mavx512vl
endif
endif