Merge commit '2ce47573b4d3664dca4cbc4354c8c600540d16ad' into cluster

This is the last commit before there are more conflicts.
This commit is contained in:
Steinar H. Gunderson
2025-12-25 17:22:27 +01:00
13 changed files with 242 additions and 159 deletions
+40 -2
View File
@@ -99,6 +99,8 @@ VPATH = syzygy:nnue:nnue/features
# 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
# altivec = yes/no --- -maltivec --- Use PowerPC Altivec SIMD extension
# vsx = yes/no --- -mvsx --- Use POWER VSX SIMD extension
# neon = yes/no --- -DUSE_NEON --- Use ARM SIMD architecture
# dotprod = yes/no --- -DUSE_NEON_DOTPROD --- Use ARM advanced SIMD Int8 dot product instructions
# lsx = yes/no --- -mlsx --- Use Loongson SIMD eXtension
@@ -128,7 +130,7 @@ endif
ifeq ($(ARCH), $(filter $(ARCH), \
x86-64-vnni512 x86-64-vnni256 x86-64-avx512 x86-64-avxvnni x86-64-bmi2 \
x86-64-avx2 x86-64-sse41-popcnt x86-64-modern x86-64-ssse3 x86-64-sse3-popcnt \
x86-64 x86-32-sse41-popcnt x86-32-sse2 x86-32 ppc-64 ppc-32 e2k \
x86-64 x86-32-sse41-popcnt x86-32-sse2 x86-32 ppc-64 ppc-64-altivec ppc-64-vsx ppc-32 e2k \
armv7 armv7-neon armv8 armv8-dotprod apple-silicon general-64 general-32 riscv64 \
loongarch64 loongarch64-lsx loongarch64-lasx))
SUPPORTED_ARCH=true
@@ -153,6 +155,8 @@ avxvnni = no
avx512 = no
vnni256 = no
vnni512 = no
altivec = no
vsx = no
neon = no
mpi = no
dotprod = no
@@ -363,6 +367,20 @@ ifeq ($(ARCH),ppc-64)
prefetch = yes
endif
ifeq ($(ARCH),ppc-64-altivec)
arch = ppc64
popcnt = yes
prefetch = yes
altivec = yes
endif
ifeq ($(ARCH),ppc-64-vsx)
arch = ppc64
popcnt = yes
prefetch = yes
vsx = yes
endif
ifeq ($(findstring e2k,$(ARCH)),e2k)
arch = e2k
mmx = yes
@@ -653,7 +671,7 @@ else
endif
ifeq ($(popcnt),yes)
ifeq ($(arch),$(filter $(arch),ppc64 armv7 armv8 arm64))
ifeq ($(arch),$(filter $(arch),ppc64 ppc64-altivec ppc64-vsx armv7 armv8 arm64))
CXXFLAGS += -DUSE_POPCNT
else
CXXFLAGS += -msse3 -mpopcnt -DUSE_POPCNT
@@ -723,6 +741,20 @@ ifeq ($(mmx),yes)
endif
endif
ifeq ($(altivec),yes)
CXXFLAGS += -maltivec
ifeq ($(COMP),gcc)
CXXFLAGS += -mabi=altivec
endif
endif
ifeq ($(vsx),yes)
CXXFLAGS += -mvsx
ifeq ($(COMP),gcc)
CXXFLAGS += -DNO_WARN_X86_INTRINSICS -DUSE_SSE2
endif
endif
ifeq ($(neon),yes)
CXXFLAGS += -DUSE_NEON=$(arm_version)
ifeq ($(KERNEL),Linux)
@@ -864,6 +896,8 @@ help:
@echo "x86-32-sse2 > x86 32-bit with sse2 support"
@echo "x86-32 > x86 32-bit generic (with mmx compile support)"
@echo "ppc-64 > PPC 64-bit"
@echo "ppc-64-altivec > PPC 64-bit with altivec support"
@echo "ppc-64-vsx > PPC 64-bit with vsx support"
@echo "ppc-32 > PPC 32-bit"
@echo "armv7 > ARMv7 32-bit"
@echo "armv7-neon > ARMv7 32-bit with popcnt and neon"
@@ -999,6 +1033,8 @@ config-sanity: net
@echo "avx512: '$(avx512)'"
@echo "vnni256: '$(vnni256)'"
@echo "vnni512: '$(vnni512)'"
@echo "altivec: '$(altivec)'"
@echo "vsx: '$(vsx)'"
@echo "neon: '$(neon)'"
@echo "mpi: '$(mpi)'"
@echo "dotprod: '$(dotprod)'"
@@ -1033,6 +1069,8 @@ config-sanity: net
@test "$(avx512)" = "yes" || test "$(avx512)" = "no"
@test "$(vnni256)" = "yes" || test "$(vnni256)" = "no"
@test "$(vnni512)" = "yes" || test "$(vnni512)" = "no"
@test "$(altivec)" = "yes" || test "$(altivec)" = "no"
@test "$(vsx)" = "yes" || test "$(vsx)" = "no"
@test "$(neon)" = "yes" || test "$(neon)" = "no"
@test "$(lsx)" = "yes" || test "$(lsx)" = "no"
@test "$(lasx)" = "yes" || test "$(lasx)" = "no"