# 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
+# 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
# at the end of the line for flag values.
### 2.1. General and architecture defaults
+
+ifeq ($(ARCH),)
+ empty_arch = yes
+endif
+
optimize = yes
debug = no
sanitize = no
bits = 64
prefetch = no
popcnt = no
-mmx = no
+pext = no
sse = no
+mmx = no
+sse2 = no
ssse3 = no
sse41 = no
avx2 = no
-pext = no
avx512 = no
+vnni256 = no
+vnni512 = no
neon = no
ARCH = x86-64-modern
+STRIP = strip
### 2.2 Architecture specific
-ifeq ($(ARCH),general-32)
- arch = any
- bits = 32
-endif
-ifeq ($(ARCH),x86-32-old)
- arch = i386
- bits = 32
-endif
+ifeq ($(findstring x86,$(ARCH)),x86)
+
+# x86-32/64
-ifeq ($(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 ($(ARCH),x86-64-modern)
- arch = x86_64
- prefetch = yes
+ifeq ($(findstring -modern,$(ARCH)),-modern)
popcnt = yes
sse = yes
+ sse2 = yes
ssse3 = yes
sse41 = yes
endif
-ifeq ($(ARCH),x86-64-sse41-popcnt)
- arch = x86_64
- prefetch = yes
+ifeq ($(findstring -avx2,$(ARCH)),-avx2)
popcnt = yes
sse = yes
+ sse2 = yes
ssse3 = yes
sse41 = yes
+ avx2 = yes
endif
-ifeq ($(ARCH),x86-64-avx2)
- arch = x86_64
- prefetch = yes
+ifeq ($(findstring -bmi2,$(ARCH)),-bmi2)
popcnt = yes
sse = yes
+ sse2 = yes
ssse3 = yes
sse41 = yes
avx2 = yes
+ pext = yes
endif
-ifeq ($(ARCH),x86-64-bmi2)
- arch = x86_64
- prefetch = yes
+ifeq ($(findstring -avx512,$(ARCH)),-avx512)
popcnt = yes
sse = yes
+ sse2 = yes
ssse3 = yes
sse41 = yes
avx2 = yes
pext = yes
+ avx512 = yes
endif
-ifeq ($(ARCH),x86-64-avx512)
- arch = x86_64
- prefetch = yes
+ifeq ($(findstring -vnni256,$(ARCH)),-vnni256)
+ popcnt = yes
+ sse = yes
+ sse2 = yes
+ ssse3 = yes
+ sse41 = yes
+ avx2 = yes
+ pext = yes
+ vnni256 = yes
+endif
+
+ifeq ($(findstring -vnni512,$(ARCH)),-vnni512)
popcnt = yes
sse = yes
+ sse2 = yes
ssse3 = yes
sse41 = yes
avx2 = yes
pext = yes
avx512 = yes
+ vnni512 = yes
+endif
+
+ifeq ($(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)
bits = 32
endif
+ifeq ($(ARCH),armv7-neon)
+ arch = armv7
+ prefetch = yes
+ popcnt = yes
+ neon = yes
+ bits = 32
+endif
+
ifeq ($(ARCH),armv8)
- arch = armv8-a
+ arch = armv8
prefetch = yes
popcnt = yes
neon = yes
prefetch = yes
endif
+endif
+
### ==========================================================================
### Section 3. Low-level Configuration
### ==========================================================================
CXX=g++
CXXFLAGS += -pedantic -Wextra -Wshadow
- ifeq ($(ARCH),$(filter $(ARCH),armv7 armv8))
+ ifeq ($(arch),$(filter $(arch),armv7 armv8))
ifeq ($(OS),Android)
CXXFLAGS += -m$(bits)
LDFLAGS += -m$(bits)
LDFLAGS += -m$(bits)
endif
+ ifeq ($(arch),$(filter $(arch),armv7))
+ LDFLAGS += -latomic
+ endif
+
ifneq ($(KERNEL),Darwin)
LDFLAGS += -Wl,--no-as-needed
endif
-
- gccversion = $(shell $(CXX) --version)
- gccisclang = $(findstring clang,$(gccversion))
endif
ifeq ($(COMP),mingw)
endif
endif
- ifeq ($(ARCH),$(filter $(ARCH),armv7 armv8))
+ ifeq ($(arch),$(filter $(arch),armv7 armv8))
ifeq ($(OS),Android)
CXXFLAGS += -m$(bits)
LDFLAGS += -m$(bits)
ifeq ($(KERNEL),Darwin)
CXXFLAGS += -arch $(arch) -mmacosx-version-min=10.14
LDFLAGS += -arch $(arch) -mmacosx-version-min=10.14
+ XCRUN = xcrun
+endif
+
+# To cross-compile for Android, NDK version r21 or later is recommended.
+# In earlier NDK versions, you'll need to pass -fno-addrsig if using GNU binutils.
+# Currently we don't know how to make PGO builds with the NDK yet.
+ifeq ($(COMP),ndk)
+ CXXFLAGS += -stdlib=libc++ -fPIE
+ 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
### Travis CI script uses COMPILER to overwrite CXX
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 ($(OS),Android)
# Haiku has pthreads in its libroot, so only link it in on other platforms
ifneq ($(KERNEL),Haiku)
- LDFLAGS += -lpthread
+ ifneq ($(COMP),ndk)
+ LDFLAGS += -lpthread
+ endif
endif
endif
endif
ifeq ($(prefetch),yes)
ifeq ($(sse),yes)
CXXFLAGS += -msse
- DEPENDFLAGS += -msse
endif
else
CXXFLAGS += -DNO_PREFETCH
### 3.6 popcnt
ifeq ($(popcnt),yes)
- ifeq ($(arch),$(filter $(arch),ppc64 armv8-a arm64))
+ ifeq ($(arch),$(filter $(arch),ppc64 armv7 armv8 arm64))
CXXFLAGS += -DUSE_POPCNT
else ifeq ($(comp),icc)
CXXFLAGS += -msse3 -DUSE_POPCNT
endif
endif
+
ifeq ($(avx2),yes)
CXXFLAGS += -DUSE_AVX2
ifeq ($(comp),$(filter $(comp),gcc clang mingw))
endif
endif
+ifeq ($(vnni256),yes)
+ CXXFLAGS += -DUSE_VNNI
+ ifeq ($(comp),$(filter $(comp),gcc clang mingw))
+ CXXFLAGS += -mavx512vnni -mavx512dq -mavx512vl -mprefer-vector-width=256
+ endif
+endif
+
+ifeq ($(vnni512),yes)
+ CXXFLAGS += -DUSE_VNNI
+ ifeq ($(comp),$(filter $(comp),gcc clang mingw))
+ CXXFLAGS += -mavx512vnni -mavx512dq -mavx512vl
+ endif
+endif
+
ifeq ($(sse41),yes)
CXXFLAGS += -DUSE_SSE41
ifeq ($(comp),$(filter $(comp),gcc clang mingw))
endif
endif
+ifeq ($(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 $(comp),gcc clang mingw))
ifeq ($(neon),yes)
CXXFLAGS += -DUSE_NEON
-endif
-
-ifeq ($(arch),x86_64)
- CXXFLAGS += -DUSE_SSE2
+ ifeq ($(KERNEL),Linux)
+ ifneq ($(COMP),ndk)
+ ifneq ($(arch),armv8)
+ CXXFLAGS += -mfpu=neon
+ endif
+ endif
+ endif
endif
### 3.7 pext
### needs access to the optimization flags.
ifeq ($(optimize),yes)
ifeq ($(debug), no)
- ifeq ($(comp),clang)
+ ifeq ($(COMP),ndk)
+ CXXFLAGS += -flto=thin
+ LDFLAGS += $(CXXFLAGS)
+ else ifeq ($(comp),clang)
CXXFLAGS += -flto=thin
LDFLAGS += $(CXXFLAGS)
endif
# To use LTO and static linking on windows, the tool chain requires a recent gcc:
-# gcc version 10.1 in msys2 or TDM-GCC version 9.2 are know to work, older might not.
+# gcc version 10.1 in msys2 or TDM-GCC version 9.2 are known to work, older might not.
# So, only enable it for a cross from Linux by default.
else ifeq ($(comp),mingw)
ifeq ($(KERNEL),Linux)
@echo ""
@echo "Supported targets:"
@echo ""
+ @echo "help > Display architecture details"
@echo "build > Standard build"
- @echo "profile-build > Standard build with PGO"
+ @echo "net > Download the default nnue net"
+ @echo "profile-build > Faster build (with profile-guided optimization)"
@echo "strip > Strip executable"
@echo "install > Install executable"
@echo "clean > Clean up"
- @echo "net > Download the default nnue net"
@echo ""
@echo "Supported archs:"
@echo ""
+ @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"
@echo "x86-64-sse41-popcnt > x86 64-bit with sse41 and popcnt support"
- @echo "x86-64-modern > the same as previous (x86-64-sse41-popcnt)"
+ @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"
- @echo "armv8 > ARMv8 64-bit"
+ @echo "armv7-neon > ARMv7 32-bit with popcnt and neon"
+ @echo "armv8 > ARMv8 64-bit with popcnt and neon"
@echo "apple-silicon > Apple silicon ARM64"
@echo "general-64 > unspecified 64-bit"
@echo "general-32 > unspecified 32-bit"
@echo "mingw > Gnu compiler with MinGW under Windows"
@echo "clang > LLVM Clang compiler"
@echo "icc > Intel compiler"
+ @echo "ndk > Google NDK to cross-compile for Android"
@echo ""
@echo "Simple examples. If you don't know what to do, you likely want to run: "
@echo ""
- @echo "make -j build ARCH=x86-64 (This is for 64-bit systems)"
- @echo "make -j build ARCH=x86-32 (This is for 32-bit systems)"
- @echo ""
- @echo "Advanced examples, for experienced users: "
+ @echo "make -j build ARCH=x86-64 (A portable, slow compile for 64-bit systems)"
+ @echo "make -j build ARCH=x86-32 (A portable, slow compile for 32-bit systems)"
@echo ""
- @echo "make -j build ARCH=x86-64-modern COMP=clang"
- @echo "make -j profile-build ARCH=x86-64-bmi2 COMP=gcc COMPCXX=g++-4.8"
+ @echo "Advanced examples, for experienced users looking for performance: "
@echo ""
- @echo "The selected architecture $(ARCH) enables the following configuration: "
+ @echo "make help ARCH=x86-64-bmi2"
+ @echo "make -j profile-build ARCH=x86-64-bmi2 COMP=gcc COMPCXX=g++-9.0"
+ @echo "make -j build ARCH=x86-64-ssse3 COMP=clang"
@echo ""
+ifneq ($(empty_arch), yes)
+ @echo "-------------------------------"
+ @echo "The selected architecture $(ARCH) will enable the following configuration: "
@$(MAKE) ARCH=$(ARCH) COMP=$(COMP) config-sanity
+endif
.PHONY: help build profile-build strip install clean net objclean profileclean \
$(MAKE) ARCH=$(ARCH) COMP=$(COMP) profileclean
strip:
- strip $(EXE)
+ $(STRIP) $(EXE)
install:
-mkdir -p -m 755 $(BINDIR)
$(eval nnuenet := $(shell grep EvalFile ucioption.cpp | grep Option | 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 := $(shell if hash curl 2>/dev/null; then echo "curl -sL"; elif hash wget 2>/dev/null; then echo "wget -qO-"; fi))
- @if test -f "$(nnuenet)"; then echo "Already available."; else echo "Downloading $(nnuedownloadurl)"; $(curl_or_wget) $(nnuedownloadurl) > $(nnuenet); fi
+ $(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;
+ $(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 $(nnuenet) corrupted, please delete!"; exit 1; \
+ fi \
+ else \
+ echo "shasum / sha256sum not found, skipping net validation"; \
+ fi
+
# clean binaries and objects
objclean:
@echo "os: '$(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: '$(avx512)'"
+ @echo "vnni256: '$(vnni256)'"
+ @echo "vnni512: '$(vnni512)'"
@echo "neon: '$(neon)'"
@echo ""
@echo "Flags:"
@test "$(optimize)" = "yes" || test "$(optimize)" = "no"
@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 "$(avx512)" = "no"
+ @test "$(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)" = "gcc" || test "$(comp)" = "icc" || test "$(comp)" = "mingw" || test "$(comp)" = "clang" \
+ || test "$(comp)" = "armv7a-linux-androideabi16-clang" || test "$(comp)" = "aarch64-linux-android21-clang"
$(EXE): $(OBJS)
+$(CXX) -o $@ $(OBJS) $(LDFLAGS)
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 ' \
projects / stockfish / blobdiff