Merge remote-tracking branch 'upstream/master'

[stockfish] / src / Makefile

diff --git a/src/Makefile b/src/Makefile
index cf4f4ecf15c1c4c9e03f9cf1732d5e45cde36c1d..aeb8ba9d29b7d81df4720ebfde628c3b4d7feb38 100644 (file)
--- 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
@@ -81,6 +81,7 @@ 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
@@ -91,7 +92,7 @@ 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
@@ -103,8 +104,8 @@ 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
@@ -125,10 +126,12 @@ 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
@@ -195,6 +198,17 @@ 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
@@ -265,6 +279,7 @@ ifeq ($(ARCH),armv7)
        arch = armv7
        prefetch = yes
        bits = 32
+       arm_version = 7
 endif
 
 ifeq ($(ARCH),armv7-neon)
@@ -273,6 +288,7 @@ ifeq ($(ARCH),armv7-neon)
        popcnt = yes
        neon = yes
        bits = 32
+       arm_version = 7
 endif
 
 ifeq ($(ARCH),armv8)
@@ -280,6 +296,7 @@ ifeq ($(ARCH),armv8)
        prefetch = yes
        popcnt = yes
        neon = yes
+       arm_version = 8
 endif
 
 ifeq ($(ARCH),apple-silicon)
@@ -287,6 +304,7 @@ ifeq ($(ARCH),apple-silicon)
        prefetch = yes
        popcnt = yes
        neon = yes
+       arm_version = 8
 endif
 
 ifeq ($(ARCH),ppc-32)
@@ -389,12 +407,14 @@ ifeq ($(COMP),clang)
        ifneq ($(KERNEL),Darwin)
        ifneq ($(KERNEL),OpenBSD)
        ifneq ($(KERNEL),FreeBSD)
+       ifneq ($(findstring MINGW,$(KERNEL)),MINGW)
        ifneq ($(RTLIB),compiler-rt)
                LDFLAGS += -latomic
        endif
        endif
        endif
        endif
+       endif
 
        ifeq ($(arch),$(filter $(arch),armv7 armv8))
                ifeq ($(OS),Android)
@@ -405,6 +425,11 @@ ifeq ($(COMP),clang)
                CXXFLAGS += -m$(bits)
                LDFLAGS += -m$(bits)
        endif
+
+       ifeq ($(findstring MINGW,$(KERNEL)),MINGW)
+               LDFLAGS += -static
+       endif
+
 endif
 
 ifeq ($(KERNEL),Darwin)
@@ -444,6 +469,9 @@ else ifeq ($(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
@@ -504,11 +532,17 @@ ifeq ($(optimize),yes)
                endif
        endif
 
-       ifeq ($(comp),$(filter $(comp),gcc clang icc))
-               ifeq ($(KERNEL),Darwin)
-                       CXXFLAGS += -mdynamic-no-pic
-               endif
-       endif
+    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 ($(comp),clang)
                CXXFLAGS += -fexperimental-new-pass-manager
@@ -520,7 +554,7 @@ ifeq ($(bits),64)
        CXXFLAGS += -DIS_64BIT
 endif
 
-### 3.5 prefetch
+### 3.5 prefetch and popcount
 ifeq ($(prefetch),yes)
        ifeq ($(sse),yes)
                CXXFLAGS += -msse
@@ -529,7 +563,6 @@ else
        CXXFLAGS += -DNO_PREFETCH
 endif
 
-### 3.6 popcnt
 ifeq ($(popcnt),yes)
        ifeq ($(arch),$(filter $(arch),ppc64 armv7 armv8 arm64))
                CXXFLAGS += -DUSE_POPCNT
@@ -540,6 +573,7 @@ ifeq ($(popcnt),yes)
        endif
 endif
 
+### 3.6 SIMD architectures
 ifeq ($(avx2),yes)
        CXXFLAGS += -DUSE_AVX2
        ifeq ($(comp),$(filter $(comp),gcc clang mingw))
@@ -547,6 +581,13 @@ ifeq ($(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 $(comp),gcc clang mingw))
@@ -597,7 +638,7 @@ ifeq ($(mmx),yes)
 endif
 
 ifeq ($(neon),yes)
-       CXXFLAGS += -DUSE_NEON
+       CXXFLAGS += -DUSE_NEON=$(arm_version)
        ifeq ($(KERNEL),Linux)
        ifneq ($(COMP),ndk)
        ifneq ($(arch),armv8)
@@ -692,6 +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"
@@ -754,7 +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=$(COMP) objclean
@@ -840,10 +882,12 @@ config-sanity: net
        @echo "ssse3: '$(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)"
@@ -895,12 +939,14 @@ 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