X-Git-Url: https://git.sesse.net/?p=stockfish;a=blobdiff_plain;f=src%2FMakefile;h=5c52661b784e45db2539664df810ea2e7773cbec;hp=3e1b7c351bc4dd5e7d90ea25f2c5c3202488d55d;hb=0e89d6e7546d26a19a108d047b489d9ba6f7970c;hpb=e453f09f06f41680ef96f594f593f8de33e62b8f diff --git a/src/Makefile b/src/Makefile index 3e1b7c35..5c52661b 100644 --- a/src/Makefile +++ b/src/Makefile @@ -1,7 +1,5 @@ # Stockfish, a UCI chess playing engine derived from Glaurung 2.1 -# Copyright (C) 2004-2008 Tord Romstad (Glaurung author) -# Copyright (C) 2008-2015 Marco Costalba, Joona Kiiski, Tord Romstad -# Copyright (C) 2015-2019 Marco Costalba, Joona Kiiski, Gary Linscott, Tord Romstad +# Copyright (C) 2004-2021 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 @@ -33,13 +31,17 @@ PREFIX = /usr/local BINDIR = $(PREFIX)/bin ### Built-in benchmark for pgo-builds -PGOBENCH = ./$(EXE) bench +ifeq ($(SDE_PATH),) + PGOBENCH = ./$(EXE) bench +else + PGOBENCH = $(SDE_PATH) -- ./$(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_kp.cpp + nnue/evaluate_nnue.cpp nnue/features/half_ka_v2_hm.cpp OBJS = $(notdir $(SRCS:.cpp=.o)) @@ -59,9 +61,11 @@ endif # ---------------------------------------------------------------------------- # # debug = yes/no --- -DNDEBUG --- Enable/Disable debug mode -# sanitize = undefined/thread/no (-fsanitize ) +# sanitize = none/ ... (-fsanitize ) # --- ( undefined ) --- enable undefined behavior checks -# --- ( thread ) --- enable threading error checks +# --- ( thread ) --- enable threading error checks +# --- ( address ) --- enable memory access checks +# --- ...etc... --- see compiler documentation for supported sanitizers # optimize = yes/no --- (-O3/-fast etc.) --- Enable/Disable optimizations # arch = (name) --- (-arch) --- Target architecture # bits = 64/32 --- -DIS_64BIT --- 64-/32-bit operating system @@ -75,22 +79,39 @@ endif # sse41 = yes/no --- -msse4.1 --- Use Intel Streaming SIMD Extensions 4.1 # avx2 = yes/no --- -mavx2 --- Use Intel Advanced Vector Extensions 2 # avx512 = yes/no --- -mavx512bw --- Use Intel Advanced Vector Extensions 512 -# vnni = yes/no --- -mavx512vnni --- Use Intel Vector Neural Network Instructions 512 +# vnni256 = yes/no --- -mavx512vnni --- Use Intel Vector Neural Network Instructions 256 +# vnni512 = yes/no --- -mavx512vnni --- Use Intel Vector Neural Network Instructions 512 # neon = yes/no --- -DUSE_NEON --- Use ARM SIMD architecture # # Note that Makefile is space sensitive, so when adding new architectures # or modifying existing flags, you have to make sure there are no extra spaces # at the end of the line for flag values. +# +# Example of use for these flags: +# make build ARCH=x86-64-avx512 debug=yes sanitize="address undefined" + ### 2.1. General and architecture defaults ifeq ($(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 $(ARCH), \ + x86-64-vnni512 x86-64-vnni256 x86-64-avx512 x86-64-bmi2 x86-64-avx2 \ + x86-64-sse41-popcnt x86-64-modern x86-64-ssse3 x86-64-sse3-popcnt \ + x86-64 x86-32-sse41-popcnt x86-32-sse2 x86-32 ppc-64 ppc-32 e2k \ + armv7 armv7-neon armv8 apple-silicon general-64 general-32)) + SUPPORTED_ARCH=true +else + SUPPORTED_ARCH=false endif optimize = yes debug = no -sanitize = no +sanitize = none bits = 64 prefetch = no popcnt = no @@ -102,9 +123,9 @@ ssse3 = no sse41 = no avx2 = no avx512 = no -vnni = no +vnni256 = no +vnni512 = no neon = no -ARCH = x86-64-modern STRIP = strip ### 2.2 Architecture specific @@ -192,7 +213,18 @@ ifeq ($(findstring -avx512,$(ARCH)),-avx512) avx512 = yes endif -ifeq ($(findstring -vnni,$(ARCH)),-vnni) +ifeq ($(findstring -vnni256,$(ARCH)),-vnni256) + popcnt = yes + sse = yes + sse2 = yes + ssse3 = yes + sse41 = yes + avx2 = yes + pext = yes + vnni256 = yes +endif + +ifeq ($(findstring -vnni512,$(ARCH)),-vnni512) popcnt = yes sse = yes sse2 = yes @@ -201,7 +233,7 @@ ifeq ($(findstring -vnni,$(ARCH)),-vnni) avx2 = yes pext = yes avx512 = yes - vnni = yes + vnni512 = yes endif ifeq ($(sse),yes) @@ -265,6 +297,17 @@ ifeq ($(ARCH),ppc-64) prefetch = yes endif +ifeq ($(findstring e2k,$(ARCH)),e2k) + arch = e2k + mmx = yes + bits = 64 + sse = yes + sse2 = yes + ssse3 = yes + sse41 = yes + popcnt = yes +endif + endif ### ========================================================================== @@ -325,7 +368,7 @@ ifeq ($(COMP),mingw) CXX=g++ endif - CXXFLAGS += -Wextra -Wshadow + CXXFLAGS += -pedantic -Wextra -Wshadow LDFLAGS += -static endif @@ -342,9 +385,13 @@ 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) @@ -357,22 +404,13 @@ 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 $(arch) -mmacosx-version-min=10.14 + CXXFLAGS += -mmacosx-version-min=10.14 + LDFLAGS += -mmacosx-version-min=10.14 + ifneq ($(arch),any) + CXXFLAGS += -arch $(arch) + LDFLAGS += -arch $(arch) + endif XCRUN = xcrun endif @@ -381,20 +419,30 @@ endif # Currently we don't know how to make PGO builds with the NDK yet. ifeq ($(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) - comp=aarch64-linux-android21-clang 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=$(COMPILER) @@ -437,9 +485,9 @@ else endif ### 3.2.2 Debugging with undefined behavior sanitizers -ifneq ($(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 @@ -458,6 +506,10 @@ ifeq ($(optimize),yes) CXXFLAGS += -mdynamic-no-pic endif endif + + ifeq ($(comp),clang) + CXXFLAGS += -fexperimental-new-pass-manager + endif endif ### 3.4 Bits @@ -465,7 +517,7 @@ ifeq ($(bits),64) CXXFLAGS += -DIS_64BIT endif -### 3.5 prefetch +### 3.5 prefetch and popcount ifeq ($(prefetch),yes) ifeq ($(sse),yes) CXXFLAGS += -msse @@ -474,7 +526,6 @@ else CXXFLAGS += -DNO_PREFETCH endif -### 3.6 popcnt ifeq ($(popcnt),yes) ifeq ($(arch),$(filter $(arch),ppc64 armv7 armv8 arm64)) CXXFLAGS += -DUSE_POPCNT @@ -485,7 +536,7 @@ ifeq ($(popcnt),yes) endif endif - +### 3.6 SIMD architectures ifeq ($(avx2),yes) CXXFLAGS += -DUSE_AVX2 ifeq ($(comp),$(filter $(comp),gcc clang mingw)) @@ -500,7 +551,14 @@ ifeq ($(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 $(comp),gcc clang mingw)) CXXFLAGS += -mavx512vnni -mavx512dq -mavx512vl @@ -559,11 +617,13 @@ endif ### needs access to the optimization flags. ifeq ($(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 += $(CXXFLAGS) # GCC and CLANG use different methods for parallelizing LTO and CLANG pretends to be @@ -578,7 +638,7 @@ ifeq ($(debug), no) LDFLAGS += -save-temps endif else - CXXFLAGS += -flto=thin + CXXFLAGS += -flto LDFLAGS += $(CXXFLAGS) endif @@ -587,10 +647,12 @@ ifeq ($(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 += $(CXXFLAGS) -flto=jobserver endif endif + endif endif endif @@ -605,6 +667,7 @@ endif ### Section 4. Public Targets ### ========================================================================== + help: @echo "" @echo "To compile stockfish, type: " @@ -623,7 +686,8 @@ help: @echo "" @echo "Supported archs:" @echo "" - @echo "x86-64-vnni > x86 64-bit with vnni support" + @echo "x86-64-vnni512 > x86 64-bit with vnni support 512bit wide" + @echo "x86-64-vnni256 > x86 64-bit with vnni support 256bit wide" @echo "x86-64-avx512 > x86 64-bit with avx512 support" @echo "x86-64-bmi2 > x86 64-bit with bmi2 support" @echo "x86-64-avx2 > x86 64-bit with avx2 support" @@ -640,6 +704,7 @@ help: @echo "armv7 > ARMv7 32-bit" @echo "armv7-neon > ARMv7 32-bit with popcnt and neon" @echo "armv8 > ARMv8 64-bit with popcnt and neon" + @echo "e2k > Elbrus 2000" @echo "apple-silicon > Apple silicon ARM64" @echo "general-64 > unspecified 64-bit" @echo "general-32 > unspecified 32-bit" @@ -663,10 +728,13 @@ help: @echo "make -j profile-build ARCH=x86-64-bmi2 COMP=gcc COMPCXX=g++-9.0" @echo "make -j build ARCH=x86-64-ssse3 COMP=clang" @echo "" -ifneq ($(empty_arch), yes) @echo "-------------------------------" +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 @@ -674,7 +742,7 @@ 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 @@ -683,7 +751,7 @@ 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 @@ -700,12 +768,13 @@ 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)) @@ -727,7 +796,6 @@ net: 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 @@ -737,6 +805,9 @@ 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 ./-lstdc++.res default: help @@ -747,7 +818,7 @@ default: all: $(EXE) .depend -config-sanity: +config-sanity: net @echo "" @echo "Config:" @echo "debug: '$(debug)'" @@ -767,7 +838,8 @@ config-sanity: @echo "sse41: '$(sse41)'" @echo "avx2: '$(avx2)'" @echo "avx512: '$(avx512)'" - @echo "vnni: '$(vnni)'" + @echo "vnni256: '$(vnni256)'" + @echo "vnni512: '$(vnni512)'" @echo "neon: '$(neon)'" @echo "" @echo "Flags:" @@ -778,10 +850,10 @@ config-sanity: @echo "Testing config sanity. If this fails, try 'make help' ..." @echo "" @test "$(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 "$(arch)" = "i386" || \ - test "$(arch)" = "ppc64" || test "$(arch)" = "ppc" || \ + test "$(arch)" = "ppc64" || test "$(arch)" = "ppc" || test "$(arch)" = "e2k" || \ test "$(arch)" = "armv7" || test "$(arch)" = "armv8" || test "$(arch)" = "arm64" @test "$(bits)" = "32" || test "$(bits)" = "64" @test "$(prefetch)" = "yes" || test "$(prefetch)" = "no" @@ -794,7 +866,8 @@ config-sanity: @test "$(sse41)" = "yes" || test "$(sse41)" = "no" @test "$(avx2)" = "yes" || test "$(avx2)" = "no" @test "$(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" @@ -816,14 +889,15 @@ 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 @@ -838,7 +912,7 @@ icc-profile-use: EXTRACXXFLAGS='-prof_use -prof_dir ./profdir' \ all -.depend: +.depend: $(SRCS) -@$(CXX) $(DEPENDFLAGS) -MM $(SRCS) > $@ 2> /dev/null -include .depend