# ssse3 = yes/no --- -mssse3 --- Use Intel Supplemental Streaming SIMD Extensions 3
# sse41 = yes/no --- -msse4.1 --- Use Intel Streaming SIMD Extensions 4.1
# 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 --- -mavx512vnni --- Use Intel Vector Neural Network Instructions 256
# vnni512 = yes/no --- -mavx512vnni --- Use Intel Vector Neural Network Instructions 512
# explicitly check for the list of supported architectures (as listed with make help),
# the user can override with `make ARCH=x86-32-vnni256 SUPPORTED_ARCH=true`
ifeq ($(ARCH), $(filter $(ARCH), \
- x86-64-vnni512 x86-64-vnni256 x86-64-avx512 x86-64-bmi2 x86-64-avx2 \
- x86-64-sse41-popcnt x86-64-modern x86-64-ssse3 x86-64-sse3-popcnt \
+ x86-64-vnni512 x86-64-vnni256 x86-64-avx512 x86-64-avxvnni x86-64-bmi2 \
+ x86-64-avx2 x86-64-sse41-popcnt x86-64-modern x86-64-ssse3 x86-64-sse3-popcnt \
x86-64 x86-32-sse41-popcnt x86-32-sse2 x86-32 ppc-64 ppc-32 e2k \
armv7 armv7-neon armv8 apple-silicon general-64 general-32))
SUPPORTED_ARCH=true
ssse3 = no
sse41 = no
avx2 = no
+avxvnni = no
avx512 = no
vnni256 = no
vnni512 = no
avx2 = yes
endif
+ifeq ($(findstring -avxvnni,$(ARCH)),-avxvnni)
+ popcnt = yes
+ sse = yes
+ sse2 = yes
+ ssse3 = yes
+ sse41 = yes
+ avx2 = yes
+ avxvnni = yes
+ pext = yes
+endif
+
ifeq ($(findstring -bmi2,$(ARCH)),-bmi2)
popcnt = yes
sse = yes
endif
endif
+ifeq ($(avxvnni),yes)
+ CXXFLAGS += -DUSE_VNNI -DUSE_AVXVNNI
+ ifeq ($(comp),$(filter $(comp),gcc clang mingw))
+ CXXFLAGS += -mavxvnni
+ endif
+endif
+
ifeq ($(avx512),yes)
CXXFLAGS += -DUSE_AVX512
ifeq ($(comp),$(filter $(comp),gcc clang mingw))
@echo "x86-64-vnni512 > x86 64-bit with vnni support 512bit wide"
@echo "x86-64-vnni256 > x86 64-bit with vnni support 256bit wide"
@echo "x86-64-avx512 > x86 64-bit with avx512 support"
+ @echo "x86-64-avxvnni > x86 64-bit with avxvnni support"
@echo "x86-64-bmi2 > x86 64-bit with bmi2 support"
@echo "x86-64-avx2 > x86 64-bit with avx2 support"
@echo "x86-64-sse41-popcnt > x86 64-bit with sse41 and popcnt support"
$(MAKE) ARCH=$(ARCH) COMP=$(COMP) $(profile_make)
@echo ""
@echo "Step 2/4. Running benchmark for pgo-build ..."
- $(PGOBENCH) > /dev/null
+ $(PGOBENCH) 2>&1 | tail -n 4
@echo ""
@echo "Step 3/4. Building optimized executable ..."
$(MAKE) ARCH=$(ARCH) COMP=$(COMP) objclean
@echo "ssse3: '$(ssse3)'"
@echo "sse41: '$(sse41)'"
@echo "avx2: '$(avx2)'"
+ @echo "avxvnni: '$(avxvnni)'"
@echo "avx512: '$(avx512)'"
@echo "vnni256: '$(vnni256)'"
@echo "vnni512: '$(vnni512)'"