X-Git-Url: https://git.sesse.net/?p=stockfish;a=blobdiff_plain;f=src%2FMakefile;h=1ff03f8391acca25bc4e3c5af9938d3408aa062d;hp=1f8ba4559dce1e5c104c8c5b7fc709953e7b9163;hb=550fed3343089357dc89ecf78ce8eb4b35bcab88;hpb=fbae5614eb1e82bccd37fbcfb0d2ca388b7a9a7d diff --git a/src/Makefile b/src/Makefile index 1f8ba455..1ff03f83 100644 --- a/src/Makefile +++ b/src/Makefile @@ -67,13 +67,16 @@ endif # bits = 64/32 --- -DIS_64BIT --- 64-/32-bit operating system # prefetch = yes/no --- -DUSE_PREFETCH --- Use prefetch asm-instruction # popcnt = yes/no --- -DUSE_POPCNT --- Use popcnt asm-instruction +# pext = yes/no --- -DUSE_PEXT --- Use pext x86_64 asm-instruction # sse = yes/no --- -msse --- Use Intel Streaming SIMD Extensions +# mmx = yes/no --- -mmmx --- Use Intel MMX instructions +# sse2 = yes/no --- -msse2 --- Use Intel Streaming SIMD Extensions 2 # 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 -# pext = yes/no --- -DUSE_PEXT --- Use pext x86_64 asm-instruction # avx512 = yes/no --- -mavx512bw --- Use Intel Advanced Vector Extensions 512 -# vnni = yes/no --- -mavx512vnni --- Use Intel Vector Neural Network Instructions 512 +# vnni256 = yes/no --- -mavx512vnni --- Use Intel Vector Neural Network Instructions 256 +# vnni512 = yes/no --- -mavx512vnni --- Use Intel Vector Neural Network Instructions 512 # neon = yes/no --- -DUSE_NEON --- Use ARM SIMD architecture # # Note that Makefile is space sensitive, so when adding new architectures @@ -83,7 +86,19 @@ endif ### 2.1. General and architecture defaults ifeq ($(ARCH),) - empty_arch = yes + ARCH = x86-64-modern + help_skip_sanity = yes +endif +# explicitly check for the list of supported architectures (as listed with make help), +# the user can override with `make ARCH=x86-32-vnni256 SUPPORTED_ARCH=true` +ifeq ($(ARCH), $(filter $(ARCH), \ + x86-64-vnni512 x86-64-vnni256 x86-64-avx512 x86-64-bmi2 x86-64-avx2 \ + x86-64-sse41-popcnt x86-64-modern x86-64-ssse3 x86-64-sse3-popcnt \ + x86-64 x86-32-sse41-popcnt x86-32-sse2 x86-32 ppc-64 ppc-32 \ + armv7 armv7-neon armv8 apple-silicon general-64 general-32)) + SUPPORTED_ARCH=true +else + SUPPORTED_ARCH=false endif optimize = yes @@ -92,97 +107,97 @@ sanitize = no bits = 64 prefetch = no popcnt = no -mmx = no +pext = no sse = no +mmx = no +sse2 = no ssse3 = no sse41 = no avx2 = no -pext = no avx512 = no -vnni = no +vnni256 = no +vnni512 = no neon = no -ARCH = x86-64-modern STRIP = strip ### 2.2 Architecture specific -ifeq ($(ARCH),general-32) - arch = any - bits = 32 -endif +ifeq ($(findstring x86,$(ARCH)),x86) -ifeq ($(ARCH),x86-32-old) - arch = i386 - bits = 32 -endif +# x86-32/64 -ifeq ($(ARCH),x86-32) +ifeq ($(findstring x86-32,$(ARCH)),x86-32) arch = i386 bits = 32 - prefetch = yes + sse = yes mmx = yes +else + arch = x86_64 sse = yes + sse2 = yes endif -ifeq ($(ARCH),general-64) - arch = any +ifeq ($(findstring -sse,$(ARCH)),-sse) + sse = yes endif -ifeq ($(ARCH),x86-64) - arch = x86_64 - prefetch = yes +ifeq ($(findstring -popcnt,$(ARCH)),-popcnt) + popcnt = yes +endif + +ifeq ($(findstring -mmx,$(ARCH)),-mmx) + mmx = yes +endif + +ifeq ($(findstring -sse2,$(ARCH)),-sse2) sse = yes + sse2 = yes endif -ifeq ($(ARCH),x86-64-sse3-popcnt) - arch = x86_64 - prefetch = yes +ifeq ($(findstring -ssse3,$(ARCH)),-ssse3) sse = yes - popcnt = yes + sse2 = yes + ssse3 = yes endif -ifeq ($(ARCH),x86-64-ssse3) - arch = x86_64 - prefetch = yes +ifeq ($(findstring -sse41,$(ARCH)),-sse41) sse = yes + sse2 = yes ssse3 = yes + sse41 = yes endif -ifeq ($(ARCH),$(filter $(ARCH),x86-64-sse41-popcnt x86-64-modern)) - arch = x86_64 - prefetch = yes +ifeq ($(findstring -modern,$(ARCH)),-modern) popcnt = yes sse = yes + sse2 = yes ssse3 = yes sse41 = yes endif -ifeq ($(ARCH),x86-64-avx2) - arch = x86_64 - prefetch = yes +ifeq ($(findstring -avx2,$(ARCH)),-avx2) popcnt = yes sse = yes + sse2 = yes ssse3 = yes sse41 = yes avx2 = yes endif -ifeq ($(ARCH),x86-64-bmi2) - arch = x86_64 - prefetch = yes +ifeq ($(findstring -bmi2,$(ARCH)),-bmi2) popcnt = yes sse = yes + sse2 = yes ssse3 = yes sse41 = yes avx2 = yes pext = yes endif -ifeq ($(ARCH),x86-64-avx512) - arch = x86_64 - prefetch = yes +ifeq ($(findstring -avx512,$(ARCH)),-avx512) popcnt = yes sse = yes + sse2 = yes ssse3 = yes sse41 = yes avx2 = yes @@ -190,17 +205,49 @@ ifeq ($(ARCH),x86-64-avx512) avx512 = yes endif -ifeq ($(ARCH),x86-64-vnni) - arch = x86_64 - prefetch = yes +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 + sse2 = yes ssse3 = yes sse41 = yes avx2 = yes pext = yes avx512 = yes - vnni = yes + vnni512 = yes +endif + +ifeq ($(sse),yes) + prefetch = yes +endif + +# 64-bit pext is not available on x86-32 +ifeq ($(bits),32) + pext = no +endif + +else + +# all other architectures + +ifeq ($(ARCH),general-32) + arch = any + bits = 32 +endif + +ifeq ($(ARCH),general-64) + arch = any endif ifeq ($(ARCH),armv7) @@ -218,7 +265,7 @@ ifeq ($(ARCH),armv7-neon) endif ifeq ($(ARCH),armv8) - arch = armv8-a + arch = armv8 prefetch = yes popcnt = yes neon = yes @@ -242,6 +289,8 @@ ifeq ($(ARCH),ppc-64) prefetch = yes endif +endif + ### ========================================================================== ### Section 3. Low-level Configuration ### ========================================================================== @@ -260,7 +309,7 @@ ifeq ($(COMP),gcc) CXX=g++ CXXFLAGS += -pedantic -Wextra -Wshadow - ifeq ($(arch),$(filter $(arch),armv7 armv8-a)) + ifeq ($(arch),$(filter $(arch),armv7 armv8)) ifeq ($(OS),Android) CXXFLAGS += -m$(bits) LDFLAGS += -m$(bits) @@ -277,9 +326,6 @@ ifeq ($(COMP),gcc) ifneq ($(KERNEL),Darwin) LDFLAGS += -Wl,--no-as-needed endif - - gccversion = $(shell $(CXX) --version) - gccisclang = $(findstring clang,$(gccversion)) endif ifeq ($(COMP),mingw) @@ -320,9 +366,11 @@ ifeq ($(COMP),clang) ifneq ($(KERNEL),Darwin) ifneq ($(KERNEL),OpenBSD) + ifneq ($(KERNEL),FreeBSD) LDFLAGS += -latomic endif endif + endif ifeq ($(arch),$(filter $(arch),armv7 armv8)) ifeq ($(OS),Android) @@ -335,22 +383,10 @@ ifeq ($(COMP),clang) endif endif -ifeq ($(comp),icc) - profile_make = icc-profile-make - profile_use = icc-profile-use -else -ifeq ($(comp),clang) - profile_make = clang-profile-make - profile_use = clang-profile-use -else - profile_make = gcc-profile-make - profile_use = gcc-profile-use -endif -endif - ifeq ($(KERNEL),Darwin) CXXFLAGS += -arch $(arch) -mmacosx-version-min=10.14 LDFLAGS += -arch $(arch) -mmacosx-version-min=10.14 + XCRUN = xcrun endif # To cross-compile for Android, NDK version r21 or later is recommended. @@ -358,20 +394,30 @@ endif # Currently we don't know how to make PGO builds with the NDK yet. ifeq ($(COMP),ndk) CXXFLAGS += -stdlib=libc++ -fPIE + comp=clang ifeq ($(arch),armv7) - comp=armv7a-linux-androideabi16-clang CXX=armv7a-linux-androideabi16-clang++ CXXFLAGS += -mthumb -march=armv7-a -mfloat-abi=softfp -mfpu=neon STRIP=arm-linux-androideabi-strip endif - ifeq ($(arch),armv8-a) - comp=aarch64-linux-android21-clang + ifeq ($(arch),armv8) CXX=aarch64-linux-android21-clang++ STRIP=aarch64-linux-android-strip endif LDFLAGS += -static-libstdc++ -pie -lm -latomic endif +ifeq ($(comp),icc) + profile_make = icc-profile-make + profile_use = icc-profile-use +else ifeq ($(comp),clang) + profile_make = clang-profile-make + profile_use = clang-profile-use +else + profile_make = gcc-profile-make + profile_use = gcc-profile-use +endif + ### Travis CI script uses COMPILER to overwrite CXX ifdef COMPILER COMPCXX=$(COMPILER) @@ -382,8 +428,19 @@ ifdef COMPCXX CXX=$(COMPCXX) endif +### Sometimes gcc is really clang +ifeq ($(COMP),gcc) + gccversion = $(shell $(CXX) --version) + gccisclang = $(findstring clang,$(gccversion)) + ifneq ($(gccisclang),) + profile_make = clang-profile-make + profile_use = clang-profile-use + endif +endif + ### On mingw use Windows threads, otherwise POSIX ifneq ($(comp),mingw) + CXXFLAGS += -DUSE_PTHREADS # On Android Bionic's C library comes with its own pthread implementation bundled in ifneq ($(OS),Android) # Haiku has pthreads in its libroot, so only link it in on other platforms @@ -424,6 +481,10 @@ ifeq ($(optimize),yes) CXXFLAGS += -mdynamic-no-pic endif endif + + ifeq ($(comp),clang) + CXXFLAGS += -fexperimental-new-pass-manager + endif endif ### 3.4 Bits @@ -442,7 +503,7 @@ endif ### 3.6 popcnt ifeq ($(popcnt),yes) - ifeq ($(arch),$(filter $(arch),ppc64 armv7 armv8-a arm64)) + ifeq ($(arch),$(filter $(arch),ppc64 armv7 armv8 arm64)) CXXFLAGS += -DUSE_POPCNT else ifeq ($(comp),icc) CXXFLAGS += -msse3 -DUSE_POPCNT @@ -466,7 +527,14 @@ ifeq ($(avx512),yes) endif endif -ifeq ($(vnni),yes) +ifeq ($(vnni256),yes) + CXXFLAGS += -DUSE_VNNI + ifeq ($(comp),$(filter $(comp),gcc clang mingw)) + CXXFLAGS += -mavx512f -mavx512bw -mavx512vnni -mavx512dq -mavx512vl -mprefer-vector-width=256 + endif +endif + +ifeq ($(vnni512),yes) CXXFLAGS += -DUSE_VNNI ifeq ($(comp),$(filter $(comp),gcc clang mingw)) CXXFLAGS += -mavx512vnni -mavx512dq -mavx512vl @@ -487,6 +555,13 @@ ifeq ($(ssse3),yes) endif endif +ifeq ($(sse2),yes) + CXXFLAGS += -DUSE_SSE2 + ifeq ($(comp),$(filter $(comp),gcc clang mingw)) + CXXFLAGS += -msse2 + endif +endif + ifeq ($(mmx),yes) CXXFLAGS += -DUSE_MMX ifeq ($(comp),$(filter $(comp),gcc clang mingw)) @@ -498,13 +573,11 @@ ifeq ($(neon),yes) CXXFLAGS += -DUSE_NEON ifeq ($(KERNEL),Linux) ifneq ($(COMP),ndk) + ifneq ($(arch),armv8) CXXFLAGS += -mfpu=neon endif endif -endif - -ifeq ($(arch),x86_64) - CXXFLAGS += -msse2 -DUSE_SSE2 + endif endif ### 3.7 pext @@ -520,11 +593,13 @@ endif ### needs access to the optimization flags. ifeq ($(optimize),yes) ifeq ($(debug), no) - ifeq ($(COMP),ndk) - CXXFLAGS += -flto=thin - LDFLAGS += $(CXXFLAGS) - else ifeq ($(comp),clang) - CXXFLAGS += -flto=thin + ifeq ($(comp),clang) + CXXFLAGS += -flto + ifneq ($(findstring MINGW,$(KERNEL)),) + CXXFLAGS += -fuse-ld=lld + else ifneq ($(findstring MSYS,$(KERNEL)),) + CXXFLAGS += -fuse-ld=lld + endif LDFLAGS += $(CXXFLAGS) # GCC and CLANG use different methods for parallelizing LTO and CLANG pretends to be @@ -539,7 +614,7 @@ ifeq ($(debug), no) LDFLAGS += -save-temps endif else - CXXFLAGS += -flto=thin + CXXFLAGS += -flto LDFLAGS += $(CXXFLAGS) endif @@ -548,10 +623,12 @@ ifeq ($(debug), no) # So, only enable it for a cross from Linux by default. else ifeq ($(comp),mingw) ifeq ($(KERNEL),Linux) + ifneq ($(arch),i386) CXXFLAGS += -flto LDFLAGS += $(CXXFLAGS) -flto=jobserver endif endif + endif endif endif @@ -566,6 +643,7 @@ endif ### Section 4. Public Targets ### ========================================================================== + help: @echo "" @echo "To compile stockfish, type: " @@ -584,7 +662,8 @@ help: @echo "" @echo "Supported archs:" @echo "" - @echo "x86-64-vnni > x86 64-bit with vnni support" + @echo "x86-64-vnni512 > x86 64-bit with vnni support 512bit wide" + @echo "x86-64-vnni256 > x86 64-bit with vnni support 256bit wide" @echo "x86-64-avx512 > x86 64-bit with avx512 support" @echo "x86-64-bmi2 > x86 64-bit with bmi2 support" @echo "x86-64-avx2 > x86 64-bit with avx2 support" @@ -592,9 +671,10 @@ help: @echo "x86-64-modern > common modern CPU, currently x86-64-sse41-popcnt" @echo "x86-64-ssse3 > x86 64-bit with ssse3 support" @echo "x86-64-sse3-popcnt > x86 64-bit with sse3 and popcnt support" - @echo "x86-64 > x86 64-bit generic" - @echo "x86-32 > x86 32-bit (also enables MMX and SSE)" - @echo "x86-32-old > x86 32-bit fall back for old hardware" + @echo "x86-64 > x86 64-bit generic (with sse2 support)" + @echo "x86-32-sse41-popcnt > x86 32-bit with sse41 and popcnt support" + @echo "x86-32-sse2 > x86 32-bit with sse2 support" + @echo "x86-32 > x86 32-bit generic (with mmx and sse support)" @echo "ppc-64 > PPC 64-bit" @echo "ppc-32 > PPC 32-bit" @echo "armv7 > ARMv7 32-bit" @@ -623,10 +703,13 @@ help: @echo "make -j profile-build ARCH=x86-64-bmi2 COMP=gcc COMPCXX=g++-9.0" @echo "make -j build ARCH=x86-64-ssse3 COMP=clang" @echo "" -ifneq ($(empty_arch), yes) - @echo "-------------------------------\n" + @echo "-------------------------------" +ifeq ($(SUPPORTED_ARCH)$(help_skip_sanity), true) @echo "The selected architecture $(ARCH) will enable the following configuration: " @$(MAKE) ARCH=$(ARCH) COMP=$(COMP) config-sanity +else + @echo "Specify a supported architecture with the ARCH option for more details" + @echo "" endif @@ -634,7 +717,7 @@ endif config-sanity icc-profile-use icc-profile-make gcc-profile-use gcc-profile-make \ clang-profile-use clang-profile-make -build: config-sanity +build: net config-sanity $(MAKE) ARCH=$(ARCH) COMP=$(COMP) all profile-build: net config-sanity objclean profileclean @@ -660,18 +743,33 @@ install: -cp $(EXE) $(BINDIR) -strip $(BINDIR)/$(EXE) -#clean all +# clean all clean: objclean profileclean @rm -f .depend *~ core +# evaluation network (nnue) net: - $(eval nnuenet := $(shell grep EvalFile ucioption.cpp | grep Option | sed 's/.*\(nn-[a-z0-9]\{12\}.nnue\).*/\1/')) + $(eval nnuenet := $(shell grep EvalFileDefaultName evaluate.h | grep define | sed 's/.*\(nn-[a-z0-9]\{12\}.nnue\).*/\1/')) @echo "Default net: $(nnuenet)" $(eval nnuedownloadurl := https://tests.stockfishchess.org/api/nn/$(nnuenet)) $(eval curl_or_wget := $(shell if hash curl 2>/dev/null; then echo "curl -skL"; elif hash wget 2>/dev/null; then echo "wget -qO-"; fi)) - @if test -f "$(nnuenet)"; then echo "Already available."; else echo "Downloading $(nnuedownloadurl)"; $(curl_or_wget) $(nnuedownloadurl) > $(nnuenet); fi + @if test -f "$(nnuenet)"; then \ + echo "Already available."; \ + else \ + if [ "x$(curl_or_wget)" = "x" ]; then \ + echo "Automatic download failed: neither curl nor wget is installed. Install one of these tools or download the net manually"; exit 1; \ + else \ + echo "Downloading $(nnuedownloadurl)"; $(curl_or_wget) $(nnuedownloadurl) > $(nnuenet);\ + fi; \ + fi; $(eval shasum_command := $(shell if hash shasum 2>/dev/null; then echo "shasum -a 256 "; elif hash sha256sum 2>/dev/null; then echo "sha256sum "; fi)) - @if [ "$(nnuenet)" != "nn-"`$(shasum_command) $(nnuenet) | cut -c1-12`".nnue" ]; then echo "Failed download or $(nnuenet) corrupted, please delete!"; exit 1; fi + @if [ "x$(shasum_command)" != "x" ]; then \ + if [ "$(nnuenet)" != "nn-"`$(shasum_command) $(nnuenet) | cut -c1-12`".nnue" ]; then \ + echo "Failed download or $(nnuenet) corrupted, please delete!"; exit 1; \ + fi \ + else \ + echo "shasum / sha256sum not found, skipping net validation"; \ + fi # clean binaries and objects objclean: @@ -692,7 +790,7 @@ default: all: $(EXE) .depend -config-sanity: +config-sanity: net @echo "" @echo "Config:" @echo "debug: '$(debug)'" @@ -704,13 +802,16 @@ config-sanity: @echo "os: '$(OS)'" @echo "prefetch: '$(prefetch)'" @echo "popcnt: '$(popcnt)'" + @echo "pext: '$(pext)'" @echo "sse: '$(sse)'" + @echo "mmx: '$(mmx)'" + @echo "sse2: '$(sse2)'" @echo "ssse3: '$(ssse3)'" @echo "sse41: '$(sse41)'" @echo "avx2: '$(avx2)'" - @echo "pext: '$(pext)'" @echo "avx512: '$(avx512)'" - @echo "vnni: '$(vnni)'" + @echo "vnni256: '$(vnni256)'" + @echo "vnni512: '$(vnni512)'" @echo "neon: '$(neon)'" @echo "" @echo "Flags:" @@ -723,19 +824,23 @@ config-sanity: @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" || \ test "$(arch)" = "ppc64" || test "$(arch)" = "ppc" || \ - test "$(arch)" = "armv7" || test "$(arch)" = "armv8-a" || test "$(arch)" = "arm64" + test "$(arch)" = "armv7" || test "$(arch)" = "armv8" || test "$(arch)" = "arm64" @test "$(bits)" = "32" || test "$(bits)" = "64" @test "$(prefetch)" = "yes" || test "$(prefetch)" = "no" @test "$(popcnt)" = "yes" || test "$(popcnt)" = "no" + @test "$(pext)" = "yes" || test "$(pext)" = "no" @test "$(sse)" = "yes" || test "$(sse)" = "no" + @test "$(mmx)" = "yes" || test "$(mmx)" = "no" + @test "$(sse2)" = "yes" || test "$(sse2)" = "no" @test "$(ssse3)" = "yes" || test "$(ssse3)" = "no" @test "$(sse41)" = "yes" || test "$(sse41)" = "no" @test "$(avx2)" = "yes" || test "$(avx2)" = "no" - @test "$(pext)" = "yes" || test "$(pext)" = "no" @test "$(avx512)" = "yes" || test "$(avx512)" = "no" - @test "$(vnni)" = "yes" || test "$(vnni)" = "no" + @test "$(vnni256)" = "yes" || test "$(vnni256)" = "no" + @test "$(vnni512)" = "yes" || test "$(vnni512)" = "no" @test "$(neon)" = "yes" || test "$(neon)" = "no" @test "$(comp)" = "gcc" || test "$(comp)" = "icc" || test "$(comp)" = "mingw" || test "$(comp)" = "clang" \ || test "$(comp)" = "armv7a-linux-androideabi16-clang" || test "$(comp)" = "aarch64-linux-android21-clang" @@ -750,7 +855,7 @@ clang-profile-make: all clang-profile-use: - llvm-profdata merge -output=stockfish.profdata *.profraw + $(XCRUN) llvm-profdata merge -output=stockfish.profdata *.profraw $(MAKE) ARCH=$(ARCH) COMP=$(COMP) \ EXTRACXXFLAGS='-fprofile-instr-use=stockfish.profdata' \ EXTRALDFLAGS='-fprofile-use ' \