X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=src%2FMakefile;h=a9333a22f46b7ac0928acf7ddfed53cd53257bb6;hb=b82d93ece484f833c994b40d9eddd959ba20ef92;hp=bd75ec524827d4c5533408e197115f6cf7e3849d;hpb=4445965f9714402050119f9e6a76c6a8fc4f8d9a;p=stockfish diff --git a/src/Makefile b/src/Makefile index bd75ec52..a9333a22 100644 --- a/src/Makefile +++ b/src/Makefile @@ -41,7 +41,7 @@ endif SRCS = benchmark.cpp bitbase.cpp bitboard.cpp endgame.cpp evaluate.cpp main.cpp \ material.cpp misc.cpp movegen.cpp movepick.cpp pawns.cpp position.cpp psqt.cpp \ search.cpp thread.cpp timeman.cpp tt.cpp uci.cpp ucioption.cpp tune.cpp syzygy/tbprobe.cpp \ - nnue/evaluate_nnue.cpp nnue/features/half_ka_v2.cpp + nnue/evaluate_nnue.cpp nnue/features/half_ka_v2_hm.cpp OBJS = $(notdir $(SRCS:.cpp=.o)) @@ -61,9 +61,11 @@ endif # ---------------------------------------------------------------------------- # # debug = yes/no --- -DNDEBUG --- Enable/Disable debug mode -# sanitize = undefined/thread/no (-fsanitize ) +# sanitize = none/ ... (-fsanitize ) # --- ( undefined ) --- enable undefined behavior checks -# --- ( thread ) --- enable threading error checks +# --- ( thread ) --- enable threading error checks +# --- ( address ) --- enable memory access checks +# --- ...etc... --- see compiler documentation for supported sanitizers # optimize = yes/no --- (-O3/-fast etc.) --- Enable/Disable optimizations # arch = (name) --- (-arch) --- Target architecture # bits = 64/32 --- -DIS_64BIT --- 64-/32-bit operating system @@ -76,6 +78,7 @@ endif # ssse3 = yes/no --- -mssse3 --- Use Intel Supplemental Streaming SIMD Extensions 3 # sse41 = yes/no --- -msse4.1 --- Use Intel Streaming SIMD Extensions 4.1 # avx2 = yes/no --- -mavx2 --- Use Intel Advanced Vector Extensions 2 +# avxvnni = yes/no --- -mavxvnni --- Use Intel Vector Neural Network Instructions AVX # avx512 = yes/no --- -mavx512bw --- Use Intel Advanced Vector Extensions 512 # vnni256 = yes/no --- -mavx512vnni --- Use Intel Vector Neural Network Instructions 256 # vnni512 = yes/no --- -mavx512vnni --- Use Intel Vector Neural Network Instructions 512 @@ -84,6 +87,10 @@ endif # Note that Makefile is space sensitive, so when adding new architectures # or modifying existing flags, you have to make sure there are no extra spaces # at the end of the line for flag values. +# +# Example of use for these flags: +# make build ARCH=x86-64-avx512 debug=yes sanitize="address undefined" + ### 2.1. General and architecture defaults @@ -94,8 +101,8 @@ endif # explicitly check for the list of supported architectures (as listed with make help), # the user can override with `make ARCH=x86-32-vnni256 SUPPORTED_ARCH=true` ifeq ($(ARCH), $(filter $(ARCH), \ - x86-64-vnni512 x86-64-vnni256 x86-64-avx512 x86-64-bmi2 x86-64-avx2 \ - x86-64-sse41-popcnt x86-64-modern x86-64-ssse3 x86-64-sse3-popcnt \ + x86-64-vnni512 x86-64-vnni256 x86-64-avx512 x86-64-avxvnni x86-64-bmi2 \ + x86-64-avx2 x86-64-sse41-popcnt x86-64-modern x86-64-ssse3 x86-64-sse3-popcnt \ x86-64 x86-32-sse41-popcnt x86-32-sse2 x86-32 ppc-64 ppc-32 e2k \ armv7 armv7-neon armv8 apple-silicon general-64 general-32)) SUPPORTED_ARCH=true @@ -105,7 +112,7 @@ endif optimize = yes debug = no -sanitize = no +sanitize = none bits = 64 prefetch = no popcnt = no @@ -116,6 +123,7 @@ sse2 = no ssse3 = no sse41 = no avx2 = no +avxvnni = no avx512 = no vnni256 = no vnni512 = no @@ -186,6 +194,17 @@ ifeq ($(findstring -avx2,$(ARCH)),-avx2) avx2 = yes endif +ifeq ($(findstring -avxvnni,$(ARCH)),-avxvnni) + popcnt = yes + sse = yes + sse2 = yes + ssse3 = yes + sse41 = yes + avx2 = yes + avxvnni = yes + pext = yes +endif + ifeq ($(findstring -bmi2,$(ARCH)),-bmi2) popcnt = yes sse = yes @@ -362,7 +381,7 @@ ifeq ($(COMP),mingw) CXX=g++ endif - CXXFLAGS += -Wextra -Wshadow + CXXFLAGS += -pedantic -Wextra -Wshadow LDFLAGS += -static endif @@ -380,10 +399,12 @@ ifeq ($(COMP),clang) ifneq ($(KERNEL),Darwin) ifneq ($(KERNEL),OpenBSD) ifneq ($(KERNEL),FreeBSD) + ifneq ($(RTLIB),compiler-rt) LDFLAGS += -latomic endif endif endif + endif ifeq ($(arch),$(filter $(arch),armv7 armv8)) ifeq ($(OS),Android) @@ -397,8 +418,12 @@ ifeq ($(COMP),clang) endif ifeq ($(KERNEL),Darwin) - CXXFLAGS += -arch $(arch) -mmacosx-version-min=10.14 - LDFLAGS += -arch $(arch) -mmacosx-version-min=10.14 + CXXFLAGS += -mmacosx-version-min=10.14 + LDFLAGS += -mmacosx-version-min=10.14 + ifneq ($(arch),any) + CXXFLAGS += -arch $(arch) + LDFLAGS += -arch $(arch) + endif XCRUN = xcrun endif @@ -473,9 +498,9 @@ else endif ### 3.2.2 Debugging with undefined behavior sanitizers -ifneq ($(sanitize),no) - CXXFLAGS += -g3 -fsanitize=$(sanitize) - LDFLAGS += -fsanitize=$(sanitize) +ifneq ($(sanitize),none) + CXXFLAGS += -g3 $(addprefix -fsanitize=,$(sanitize)) + LDFLAGS += $(addprefix -fsanitize=,$(sanitize)) endif ### 3.3 Optimization @@ -505,7 +530,7 @@ ifeq ($(bits),64) CXXFLAGS += -DIS_64BIT endif -### 3.5 prefetch +### 3.5 prefetch and popcount ifeq ($(prefetch),yes) ifeq ($(sse),yes) CXXFLAGS += -msse @@ -514,7 +539,6 @@ else CXXFLAGS += -DNO_PREFETCH endif -### 3.6 popcnt ifeq ($(popcnt),yes) ifeq ($(arch),$(filter $(arch),ppc64 armv7 armv8 arm64)) CXXFLAGS += -DUSE_POPCNT @@ -525,6 +549,7 @@ ifeq ($(popcnt),yes) endif endif +### 3.6 SIMD architectures ifeq ($(avx2),yes) CXXFLAGS += -DUSE_AVX2 ifeq ($(comp),$(filter $(comp),gcc clang mingw)) @@ -532,6 +557,13 @@ ifeq ($(avx2),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)) @@ -677,6 +709,7 @@ help: @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" @@ -739,7 +772,7 @@ profile-build: net config-sanity objclean profileclean $(MAKE) ARCH=$(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=$(COMP) objclean @@ -825,6 +858,7 @@ config-sanity: net @echo "ssse3: '$(ssse3)'" @echo "sse41: '$(sse41)'" @echo "avx2: '$(avx2)'" + @echo "avxvnni: '$(avxvnni)'" @echo "avx512: '$(avx512)'" @echo "vnni256: '$(vnni256)'" @echo "vnni512: '$(vnni512)'" @@ -838,7 +872,6 @@ config-sanity: net @echo "Testing config sanity. If this fails, try 'make help' ..." @echo "" @test "$(debug)" = "yes" || test "$(debug)" = "no" - @test "$(sanitize)" = "undefined" || test "$(sanitize)" = "thread" || test "$(sanitize)" = "address" || test "$(sanitize)" = "no" @test "$(optimize)" = "yes" || test "$(optimize)" = "no" @test "$(SUPPORTED_ARCH)" = "true" @test "$(arch)" = "any" || test "$(arch)" = "x86_64" || test "$(arch)" = "i386" || \ @@ -878,14 +911,15 @@ clang-profile-use: all gcc-profile-make: + @mkdir -p profdir $(MAKE) ARCH=$(ARCH) COMP=$(COMP) \ - EXTRACXXFLAGS='-fprofile-generate' \ + EXTRACXXFLAGS='-fprofile-generate=profdir' \ EXTRALDFLAGS='-lgcov' \ all gcc-profile-use: $(MAKE) ARCH=$(ARCH) COMP=$(COMP) \ - EXTRACXXFLAGS='-fprofile-use -fno-peel-loops -fno-tracer' \ + EXTRACXXFLAGS='-fprofile-use=profdir -fno-peel-loops -fno-tracer' \ EXTRALDFLAGS='-lgcov' \ all @@ -900,7 +934,7 @@ icc-profile-use: EXTRACXXFLAGS='-prof_use -prof_dir ./profdir' \ all -.depend: +.depend: $(SRCS) -@$(CXX) $(DEPENDFLAGS) -MM $(SRCS) > $@ 2> /dev/null -include .depend