GitHub/Stockfish
1
0
Fork 0
mirror of https://github.com/official-stockfish/Stockfish.git synced 2025-11-01 06:32:32 +00:00
Code Issues Packages Projects Releases Wiki Activity

Remove x86-64-vnni256 target

Browse Source 
When vnni256 was first introduced #3038 it was very slightly faster than vnni512
on some machines. We have since sped up vnni512 significantly (#4796 over 10%
alone, #6139, and probably others I'm forgetting). Since any machine that can
run vnni256 can run vnni512 this arch is no longer useful.

Note, x86-64-avxvnni still covers targets that don't have AVX512 but do have
VNNI on 128- and 256-bit vectors.

closes https://github.com/official-stockfish/Stockfish/pull/6340

No functional change
This commit is contained in:
mstembera
2025-10-04 15:12:25 -07:00
committed by Joost VandeVondele
parent e5c2dc5edd
commit ee243f0fdc
3 changed files with 4 additions and 46 deletions
Download Patch File Download Diff File Expand all files Collapse all files
.github/ci/matrix.json
-19
View File
@@ -61,7 +61,6 @@
"x86-64-bmi2",
"x86-64-avxvnni",
"x86-64-avx512",
"x86-64-vnni256",
"x86-64-vnni512",
"x86-64-avx512icl",
"apple-silicon",
@@ -105,12 +104,6 @@
"os": "macos-14"
}
},
{
"binaries": "x86-64-vnni256",
"config": {
"os": "macos-14"
}
},
{
"binaries": "x86-64-vnni512",
"config": {
@@ -135,12 +128,6 @@
"os": "macos-15-intel"
}
},
{
"binaries": "x86-64-vnni256",
"config": {
"os": "macos-15-intel"
}
},
{
"binaries": "x86-64-vnni512",
"config": {
@@ -189,12 +176,6 @@
"os": "windows-11-arm"
}
},
{
"binaries": "x86-64-vnni256",
"config": {
"os": "windows-11-arm"
}
},
{
"binaries": "x86-64-vnni512",
"config": {
scripts/get_native_properties.sh
+2 -2
View File
@@ -42,7 +42,7 @@ set_arch_x86_64() {
if check_flags 'avx512f' 'avx512cd' 'avx512vl' 'avx512dq' 'avx512bw' 'avx512ifma' 'avx512vbmi' 'avx512vbmi2' 'avx512vpopcntdq' 'avx512bitalg' 'avx512vnni' 'vpclmulqdq' 'gfni' 'vaes'; then
true_arch='x86-64-avx512icl'
elif check_flags 'avx512vnni' 'avx512dq' 'avx512f' 'avx512bw' 'avx512vl'; then
true_arch='x86-64-vnni256'
true_arch='x86-64-vnni512'
elif check_flags 'avx512f' 'avx512bw'; then
true_arch='x86-64-avx512'
elif [ -z "${znver_1_2+1}" ] && check_flags 'bmi2'; then
@@ -83,7 +83,7 @@ case $uname_s in
'x86_64')
flags=$(sysctl -n machdep.cpu.features machdep.cpu.leaf7_features | tr '\n' ' ' | tr '[:upper:]' '[:lower:]' | tr -d '_.')
set_arch_x86_64
if [ "$true_arch" = 'x86-64-vnni256' ] || [ "$true_arch" = 'x86-64-avx512' ]; then
if [ "$true_arch" = 'x86-64-avx512' ]; then
file_arch='x86-64-bmi2'
fi
;;
src/Makefile
+2 -25
View File
@@ -97,7 +97,6 @@ VPATH = syzygy:nnue:nnue/features
# avx2 = yes/no --- -mavx2 --- Use Intel Advanced Vector Extensions 2
# avxvnni = yes/no --- -mavxvnni --- Use Intel Vector Neural Network Instructions AVX
# avx512 = yes/no --- -mavx512bw --- Use Intel Advanced Vector Extensions 512
# vnni256 = yes/no --- -mavx256vnni --- Use Intel Vector Neural Network Instructions 512 with 256bit operands
# vnni512 = yes/no --- -mavx512vnni --- Use Intel Vector Neural Network Instructions 512
# avx512icl = yes/no --- ... multiple ... --- Use All AVX-512 features available on both Intel Ice Lake and AMD Zen 4
# altivec = yes/no --- -maltivec --- Use PowerPC Altivec SIMD extension
@@ -128,7 +127,7 @@ endif
# explicitly check for the list of supported architectures (as listed with make help),
# the user can override with `make ARCH=x86-64-avx512icl SUPPORTED_ARCH=true`
ifeq ($(ARCH), $(filter $(ARCH), \
x86-64-avx512icl x86-64-vnni512 x86-64-vnni256 x86-64-avx512 x86-64-avxvnni \
x86-64-avx512icl x86-64-vnni512 x86-64-avx512 x86-64-avxvnni \
x86-64-bmi2 x86-64-avx2 x86-64-sse41-popcnt x86-64-modern x86-64-ssse3 x86-64-sse3-popcnt \
x86-64 x86-32-sse41-popcnt x86-32-sse2 x86-32 ppc-64 ppc-64-altivec ppc-64-vsx ppc-32 e2k \
armv7 armv7-neon armv8 armv8-dotprod apple-silicon general-64 general-32 riscv64 \
@@ -153,7 +152,6 @@ sse41 = no
avx2 = no
avxvnni = no
avx512 = no
vnni256 = no
vnni512 = no
avx512icl = no
altivec = no
@@ -269,17 +267,6 @@ ifeq ($(findstring -avx512,$(ARCH)),-avx512)
avx512 = yes
endif
ifeq ($(findstring -vnni256,$(ARCH)),-vnni256)
popcnt = yes
sse = yes
sse2 = yes
ssse3 = yes
sse41 = yes
avx2 = yes
pext = yes
vnni256 = yes
endif
ifeq ($(findstring -vnni512,$(ARCH)),-vnni512)
popcnt = yes
sse = yes
@@ -724,17 +711,10 @@ ifeq ($(avx512),yes)
endif
endif
ifeq ($(vnni256),yes)
CXXFLAGS += -DUSE_VNNI
ifeq ($(comp),$(filter $(comp),gcc clang mingw icx))
CXXFLAGS += -mavx512f -mavx512bw -mavx512vnni -mavx512dq -mavx512vl -mprefer-vector-width=256
endif
endif
ifeq ($(vnni512),yes)
CXXFLAGS += -DUSE_VNNI
ifeq ($(comp),$(filter $(comp),gcc clang mingw icx))
CXXFLAGS += -mavx512f -mavx512bw -mavx512vnni -mavx512dq -mavx512vl -mprefer-vector-width=512
CXXFLAGS += -mavx512f -mavx512bw -mavx512vnni -mavx512dq -mavx512vl
endif
endif
@@ -905,7 +885,6 @@ help:
echo "native > select the best architecture for the host processor (default)" && \
echo "x86-64-avx512icl > x86 64-bit with minimum avx512 support of Intel Ice Lake or AMD Zen 4" && \
echo "x86-64-vnni512 > x86 64-bit with vnni 512bit support" && \
echo "x86-64-vnni256 > x86 64-bit with vnni 512bit support, limit operands to 256bit wide" && \
echo "x86-64-avx512 > x86 64-bit with avx512 support" && \
echo "x86-64-avxvnni > x86 64-bit with vnni 256bit support" && \
echo "x86-64-bmi2 > x86 64-bit with bmi2 support" && \
@@ -1050,7 +1029,6 @@ config-sanity: net
echo "avx2: '$(avx2)'" && \
echo "avxvnni: '$(avxvnni)'" && \
echo "avx512: '$(avx512)'" && \
echo "vnni256: '$(vnni256)'" && \
echo "vnni512: '$(vnni512)'" && \
echo "avx512icl: '$(avx512icl)'" && \
echo "altivec: '$(altivec)'" && \
@@ -1087,7 +1065,6 @@ config-sanity: net
(test "$(sse41)" = "yes" || test "$(sse41)" = "no") && \
(test "$(avx2)" = "yes" || test "$(avx2)" = "no") && \
(test "$(avx512)" = "yes" || test "$(avx512)" = "no") && \
(test "$(vnni256)" = "yes" || test "$(vnni256)" = "no") && \
(test "$(vnni512)" = "yes" || test "$(vnni512)" = "no") && \
(test "$(avx512icl)" = "yes" || test "$(avx512icl)" = "no") && \
(test "$(altivec)" = "yes" || test "$(altivec)" = "no") && \