X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=src%2FMakefile;h=cc9b4201555d4fefb3125c7b6d44ee7855d2d459;hb=f7d1491b3df28bf10faac81e340cb6a22fc5b57b;hp=bf3c3560bbd8efa375d3cc62eac520d173c86813;hpb=0171b506ec96e2b8c8fcc7758cd5624ad26c0552;p=stockfish diff --git a/src/Makefile b/src/Makefile index bf3c3560..cc9b4201 100644 --- a/src/Makefile +++ b/src/Makefile @@ -1,5 +1,5 @@ # Stockfish, a UCI chess playing engine derived from Glaurung 2.1 -# Copyright (C) 2004-2021 The Stockfish developers (see AUTHORS file) +# Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file) # # Stockfish is free software: you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by @@ -19,11 +19,29 @@ ### Section 1. General Configuration ### ========================================================================== +### Establish the operating system name +KERNEL = $(shell uname -s) +ifeq ($(KERNEL),Linux) + OS = $(shell uname -o) +endif + +### Target Windows OS +ifeq ($(OS),Windows_NT) + ifneq ($(COMP),ndk) + target_windows = yes + endif +else ifeq ($(COMP),mingw) + target_windows = yes + ifeq ($(WINE_PATH),) + WINE_PATH = $(shell which wine) + endif +endif + ### Executable name -ifeq ($(COMP),mingw) -EXE = stockfish.exe +ifeq ($(target_windows),yes) + EXE = stockfish.exe else -EXE = stockfish + EXE = stockfish endif ### Installation dir definitions @@ -32,27 +50,21 @@ BINDIR = $(PREFIX)/bin ### Built-in benchmark for pgo-builds ifeq ($(SDE_PATH),) - PGOBENCH = ./$(EXE) bench + PGOBENCH = $(WINE_PATH) ./$(EXE) bench else - PGOBENCH = $(SDE_PATH) -- ./$(EXE) bench + PGOBENCH = $(SDE_PATH) -- $(WINE_PATH) ./$(EXE) bench endif ### Source and object files 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)) VPATH = syzygy:nnue:nnue/features -### Establish the operating system name -KERNEL = $(shell uname -s) -ifeq ($(KERNEL),Linux) - OS = $(shell uname -o) -endif - ### ========================================================================== ### Section 2. High-level Configuration ### ========================================================================== @@ -78,6 +90,7 @@ endif # ssse3 = yes/no --- -mssse3 --- Use Intel Supplemental Streaming SIMD Extensions 3 # sse41 = yes/no --- -msse4.1 --- Use Intel Streaming SIMD Extensions 4.1 # avx2 = yes/no --- -mavx2 --- Use Intel Advanced Vector Extensions 2 +# avxvnni = yes/no --- -mavxvnni --- Use Intel Vector Neural Network Instructions AVX # avx512 = yes/no --- -mavx512bw --- Use Intel Advanced Vector Extensions 512 # vnni256 = yes/no --- -mavx512vnni --- Use Intel Vector Neural Network Instructions 256 # vnni512 = yes/no --- -mavx512vnni --- Use Intel Vector Neural Network Instructions 512 @@ -88,7 +101,7 @@ endif # at the end of the line for flag values. # # Example of use for these flags: -# make build ARCH=x86-64-avx512 debug=on sanitize="address undefined" +# make build ARCH=x86-64-avx512 debug=yes sanitize="address undefined" ### 2.1. General and architecture defaults @@ -100,8 +113,8 @@ endif # explicitly check for the list of supported architectures (as listed with make help), # the user can override with `make ARCH=x86-32-vnni256 SUPPORTED_ARCH=true` ifeq ($(ARCH), $(filter $(ARCH), \ - x86-64-vnni512 x86-64-vnni256 x86-64-avx512 x86-64-bmi2 x86-64-avx2 \ - x86-64-sse41-popcnt x86-64-modern x86-64-ssse3 x86-64-sse3-popcnt \ + x86-64-vnni512 x86-64-vnni256 x86-64-avx512 x86-64-avxvnni x86-64-bmi2 \ + x86-64-avx2 x86-64-sse41-popcnt x86-64-modern x86-64-ssse3 x86-64-sse3-popcnt \ x86-64 x86-32-sse41-popcnt x86-32-sse2 x86-32 ppc-64 ppc-32 e2k \ armv7 armv7-neon armv8 apple-silicon general-64 general-32)) SUPPORTED_ARCH=true @@ -122,10 +135,12 @@ sse2 = no ssse3 = no sse41 = no avx2 = no +avxvnni = no avx512 = no vnni256 = no vnni512 = no neon = no +arm_version = 0 STRIP = strip ### 2.2 Architecture specific @@ -137,7 +152,7 @@ ifeq ($(findstring x86,$(ARCH)),x86) ifeq ($(findstring x86-32,$(ARCH)),x86-32) arch = i386 bits = 32 - sse = yes + sse = no mmx = yes else arch = x86_64 @@ -192,6 +207,17 @@ ifeq ($(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,$(ARCH)),-bmi2) popcnt = yes sse = yes @@ -262,6 +288,7 @@ ifeq ($(ARCH),armv7) arch = armv7 prefetch = yes bits = 32 + arm_version = 7 endif ifeq ($(ARCH),armv7-neon) @@ -270,6 +297,7 @@ ifeq ($(ARCH),armv7-neon) popcnt = yes neon = yes bits = 32 + arm_version = 7 endif ifeq ($(ARCH),armv8) @@ -277,6 +305,7 @@ ifeq ($(ARCH),armv8) prefetch = yes popcnt = yes neon = yes + arm_version = 8 endif ifeq ($(ARCH),apple-silicon) @@ -284,6 +313,7 @@ ifeq ($(ARCH),apple-silicon) prefetch = yes popcnt = yes neon = yes + arm_version = 8 endif ifeq ($(ARCH),ppc-32) @@ -347,29 +377,27 @@ ifeq ($(COMP),gcc) endif endif +ifeq ($(target_windows),yes) + LDFLAGS += -static +endif + ifeq ($(COMP),mingw) comp=mingw - ifeq ($(KERNEL),Linux) - ifeq ($(bits),64) - ifeq ($(shell which x86_64-w64-mingw32-c++-posix),) - CXX=x86_64-w64-mingw32-c++ - else - CXX=x86_64-w64-mingw32-c++-posix - endif + ifeq ($(bits),64) + ifeq ($(shell which x86_64-w64-mingw32-c++-posix 2> /dev/null),) + CXX=x86_64-w64-mingw32-c++ else - ifeq ($(shell which i686-w64-mingw32-c++-posix),) - CXX=i686-w64-mingw32-c++ - else - CXX=i686-w64-mingw32-c++-posix - endif + CXX=x86_64-w64-mingw32-c++-posix endif else - CXX=g++ + ifeq ($(shell which i686-w64-mingw32-c++-posix 2> /dev/null),) + CXX=i686-w64-mingw32-c++ + else + CXX=i686-w64-mingw32-c++-posix + endif endif - - CXXFLAGS += -Wextra -Wshadow - LDFLAGS += -static + CXXFLAGS += -pedantic -Wextra -Wshadow endif ifeq ($(COMP),icc) @@ -381,11 +409,15 @@ endif ifeq ($(COMP),clang) comp=clang CXX=clang++ + ifeq ($(target_windows),yes) + CXX=x86_64-w64-mingw32-clang++ + endif + CXXFLAGS += -pedantic -Wextra -Wshadow - ifneq ($(KERNEL),Darwin) - ifneq ($(KERNEL),OpenBSD) - ifneq ($(KERNEL),FreeBSD) + ifeq ($(filter $(KERNEL),Darwin OpenBSD FreeBSD),) + ifeq ($(target_windows),) + ifneq ($(RTLIB),compiler-rt) LDFLAGS += -latomic endif endif @@ -403,8 +435,12 @@ ifeq ($(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 $(arch) + endif XCRUN = xcrun endif @@ -417,11 +453,19 @@ ifeq ($(COMP),ndk) ifeq ($(arch),armv7) CXX=armv7a-linux-androideabi16-clang++ CXXFLAGS += -mthumb -march=armv7-a -mfloat-abi=softfp -mfpu=neon - STRIP=arm-linux-androideabi-strip + ifneq ($(shell which arm-linux-androideabi-strip 2>/dev/null),) + STRIP=arm-linux-androideabi-strip + else + STRIP=llvm-strip + endif endif ifeq ($(arch),armv8) CXX=aarch64-linux-android21-clang++ - STRIP=aarch64-linux-android-strip + ifneq ($(shell which aarch64-linux-android-strip 2>/dev/null),) + STRIP=aarch64-linux-android-strip + else + STRIP=llvm-strip + endif endif LDFLAGS += -static-libstdc++ -pie -lm -latomic endif @@ -435,6 +479,9 @@ else ifeq ($(comp),clang) else profile_make = gcc-profile-make profile_use = gcc-profile-use + ifeq ($(KERNEL),Darwin) + EXTRAPROFILEFLAGS = -fvisibility=hidden + endif endif ### Travis CI script uses COMPILER to overwrite CXX @@ -495,10 +542,16 @@ ifeq ($(optimize),yes) endif endif - ifeq ($(comp),$(filter $(comp),gcc clang icc)) - ifeq ($(KERNEL),Darwin) + ifeq ($(KERNEL),Darwin) + ifeq ($(comp),$(filter $(comp),clang icc)) CXXFLAGS += -mdynamic-no-pic endif + + ifeq ($(comp),gcc) + ifneq ($(arch),arm64) + CXXFLAGS += -mdynamic-no-pic + endif + endif endif ifeq ($(comp),clang) @@ -511,7 +564,7 @@ ifeq ($(bits),64) CXXFLAGS += -DIS_64BIT endif -### 3.5 prefetch +### 3.5 prefetch and popcount ifeq ($(prefetch),yes) ifeq ($(sse),yes) CXXFLAGS += -msse @@ -520,7 +573,6 @@ else CXXFLAGS += -DNO_PREFETCH endif -### 3.6 popcnt ifeq ($(popcnt),yes) ifeq ($(arch),$(filter $(arch),ppc64 armv7 armv8 arm64)) CXXFLAGS += -DUSE_POPCNT @@ -531,6 +583,7 @@ ifeq ($(popcnt),yes) endif endif +### 3.6 SIMD architectures ifeq ($(avx2),yes) CXXFLAGS += -DUSE_AVX2 ifeq ($(comp),$(filter $(comp),gcc clang mingw)) @@ -538,6 +591,13 @@ ifeq ($(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 $(comp),gcc clang mingw)) @@ -588,7 +648,7 @@ ifeq ($(mmx),yes) endif ifeq ($(neon),yes) - CXXFLAGS += -DUSE_NEON + CXXFLAGS += -DUSE_NEON=$(arm_version) ifeq ($(KERNEL),Linux) ifneq ($(COMP),ndk) ifneq ($(arch),armv8) @@ -613,9 +673,7 @@ ifeq ($(optimize),yes) ifeq ($(debug), no) ifeq ($(comp),clang) CXXFLAGS += -flto - ifneq ($(findstring MINGW,$(KERNEL)),) - CXXFLAGS += -fuse-ld=lld - else ifneq ($(findstring MSYS,$(KERNEL)),) + ifeq ($(target_windows),yes) CXXFLAGS += -fuse-ld=lld endif LDFLAGS += $(CXXFLAGS) @@ -626,25 +684,17 @@ ifeq ($(debug), no) ifeq ($(gccisclang),) CXXFLAGS += -flto LDFLAGS += $(CXXFLAGS) -flto=jobserver - ifneq ($(findstring MINGW,$(KERNEL)),) - LDFLAGS += -save-temps - else ifneq ($(findstring MSYS,$(KERNEL)),) - LDFLAGS += -save-temps - endif else CXXFLAGS += -flto LDFLAGS += $(CXXFLAGS) endif -# To use LTO and static linking on windows, the tool chain requires a recent gcc: -# gcc version 10.1 in msys2 or TDM-GCC version 9.2 are known to work, older might not. -# So, only enable it for a cross from Linux by default. +# To use LTO and static linking on Windows, +# the tool chain requires gcc version 10.1 or later. else ifeq ($(comp),mingw) - ifeq ($(KERNEL),Linux) ifneq ($(arch),i386) CXXFLAGS += -flto - LDFLAGS += $(CXXFLAGS) -flto=jobserver - endif + LDFLAGS += $(CXXFLAGS) -save-temps endif endif endif @@ -683,6 +733,7 @@ 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" @@ -745,7 +796,7 @@ 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 @@ -760,7 +811,7 @@ strip: install: -mkdir -p -m 755 $(BINDIR) -cp $(EXE) $(BINDIR) - -strip $(BINDIR)/$(EXE) + $(STRIP) $(BINDIR)/$(EXE) # clean all clean: objclean profileclean @@ -773,34 +824,35 @@ net: $(eval nnuedownloadurl := https://tests.stockfishchess.org/api/nn/$(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 \ - 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) > $(nnuenet);\ - fi; \ - fi; + echo "Already available."; \ + else \ + if [ "x$(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 := $(shell if hash shasum 2>/dev/null; then echo "shasum -a 256 "; elif hash sha256sum 2>/dev/null; then echo "sha256sum "; fi)) @if [ "x$(shasum_command)" != "x" ]; then \ - if [ "$(nnuenet)" != "nn-"`$(shasum_command) $(nnuenet) | cut -c1-12`".nnue" ]; then \ - echo "Failed download or $(nnuenet) corrupted, please delete!"; exit 1; \ - fi \ - else \ - echo "shasum / sha256sum not found, skipping net validation"; \ - fi + if [ "$(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: - @rm -f $(EXE) *.o ./syzygy/*.o ./nnue/*.o ./nnue/features/*.o + @rm -f stockfish stockfish.exe *.o ./syzygy/*.o ./nnue/*.o ./nnue/features/*.o # clean auxiliary profiling files profileclean: @rm -rf profdir @rm -f bench.txt *.gcda *.gcno ./syzygy/*.gcda ./nnue/*.gcda ./nnue/features/*.gcda *.s @rm -f stockfish.profdata *.profraw - @rm -f stockfish.exe.lto_wrapper_args - @rm -f stockfish.exe.ltrans.out + @rm -f stockfish.*args* + @rm -f stockfish.*lt* + @rm -f stockfish.res @rm -f ./-lstdc++.res default: @@ -831,10 +883,12 @@ config-sanity: net @echo "ssse3: '$(ssse3)'" @echo "sse41: '$(sse41)'" @echo "avx2: '$(avx2)'" + @echo "avxvnni: '$(avxvnni)'" @echo "avx512: '$(avx512)'" @echo "vnni256: '$(vnni256)'" @echo "vnni512: '$(vnni512)'" @echo "neon: '$(neon)'" + @echo "arm_version: '$(arm_version)'" @echo "" @echo "Flags:" @echo "CXX: $(CXX)" @@ -886,12 +940,14 @@ gcc-profile-make: @mkdir -p profdir $(MAKE) ARCH=$(ARCH) COMP=$(COMP) \ EXTRACXXFLAGS='-fprofile-generate=profdir' \ + EXTRACXXFLAGS+=$(EXTRAPROFILEFLAGS) \ EXTRALDFLAGS='-lgcov' \ all gcc-profile-use: $(MAKE) ARCH=$(ARCH) COMP=$(COMP) \ EXTRACXXFLAGS='-fprofile-use=profdir -fno-peel-loops -fno-tracer' \ + EXTRACXXFLAGS+=$(EXTRAPROFILEFLAGS) \ EXTRALDFLAGS='-lgcov' \ all @@ -906,7 +962,7 @@ icc-profile-use: EXTRACXXFLAGS='-prof_use -prof_dir ./profdir' \ all -.depend: +.depend: $(SRCS) -@$(CXX) $(DEPENDFLAGS) -MM $(SRCS) > $@ 2> /dev/null -include .depend