Support VNNI on 256bit vectors

[stockfish] / src / Makefile

diff --git a/src/Makefile b/src/Makefile
index 79c7333ac26edc43fe73f3a44c8ad29e86296ca2..228ea851f208154001cd3dcab5926d298ff240cb 100644 (file)
--- a/src/Makefile
+++ b/src/Makefile
@@ -75,7 +75,8 @@ 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
@@ -102,7 +103,8 @@ ssse3 = no
 sse41 = no
 avx2 = no
 avx512 = no
-vnni = no
+vnni256 = no
+vnni512 = no
 neon = no
 ARCH = x86-64-modern
 STRIP = strip
@@ -192,7 +194,18 @@ 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 +214,7 @@ ifeq ($(findstring -vnni,$(ARCH)),-vnni)
        avx2 = yes
        pext = yes
        avx512 = yes
-       vnni = yes
+       vnni512 = yes
 endif
 
 ifeq ($(sse),yes)
@@ -241,7 +254,7 @@ ifeq ($(ARCH),armv7-neon)
 endif
 
 ifeq ($(ARCH),armv8)
-       arch = armv8-a
+       arch = armv8
        prefetch = yes
        popcnt = yes
        neon = yes
@@ -285,7 +298,7 @@ ifeq ($(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)
@@ -302,9 +315,6 @@ ifeq ($(COMP),gcc)
        ifneq ($(KERNEL),Darwin)
           LDFLAGS += -Wl,--no-as-needed
        endif
-
-       gccversion = $(shell $(CXX) --version)
-       gccisclang = $(findstring clang,$(gccversion))
 endif
 
 ifeq ($(COMP),mingw)
@@ -376,6 +386,7 @@ endif
 ifeq ($(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.
@@ -389,7 +400,7 @@ ifeq ($(COMP),ndk)
                CXXFLAGS += -mthumb -march=armv7-a -mfloat-abi=softfp -mfpu=neon
                STRIP=arm-linux-androideabi-strip
        endif
-       ifeq ($(arch),armv8-a)
+       ifeq ($(arch),armv8)
                comp=aarch64-linux-android21-clang
                CXX=aarch64-linux-android21-clang++
                STRIP=aarch64-linux-android-strip
@@ -407,8 +418,19 @@ ifdef COMPCXX
        CXX=$(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
@@ -467,7 +489,7 @@ 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
@@ -491,7 +513,14 @@ ifeq ($(avx512),yes)
        endif
 endif
 
-ifeq ($(vnni),yes)
+ifeq ($(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 $(comp),gcc clang mingw))
                CXXFLAGS += -mavx512vnni -mavx512dq -mavx512vl
@@ -530,9 +559,11 @@ ifeq ($(neon),yes)
        CXXFLAGS += -DUSE_NEON
        ifeq ($(KERNEL),Linux)
        ifneq ($(COMP),ndk)
+       ifneq ($(arch),armv8)
                CXXFLAGS += -mfpu=neon
        endif
        endif
+       endif
 endif
 
 ### 3.7 pext
@@ -612,7 +643,8 @@ 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"
@@ -756,7 +788,8 @@ config-sanity:
        @echo "sse41: '$(sse41)'"
        @echo "avx2: '$(avx2)'"
        @echo "avx512: '$(avx512)'"
-       @echo "vnni: '$(vnni)'"
+       @echo "vnni256: '$(vnni256)'"
+       @echo "vnni512: '$(vnni512)'"
        @echo "neon: '$(neon)'"
        @echo ""
        @echo "Flags:"
@@ -771,7 +804,7 @@ config-sanity:
        @test "$(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"
@@ -783,7 +816,8 @@ 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"
@@ -798,7 +832,7 @@ 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 ' \