X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;ds=sidebyside;f=src%2FMakefile;h=38f607cb24f58fdc2719197be82b82be87e01b04;hb=cd0b8b4cf28208fffef931322749205a0ddc6066;hp=a48e7dcb3ccd481ae6868b8e430d15ba00756c41;hpb=cb0504028e8830dbc71be53cbd701d78c3d562a1;p=stockfish

diff --git a/src/Makefile b/src/Makefile
index a48e7dcb..38f607cb 100644
--- a/src/Makefile
+++ b/src/Makefile
@@ -73,6 +73,7 @@ endif
 # 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
+# vnni = 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
@@ -80,22 +81,30 @@ endif
 # 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
 sse = no
 ssse3 = no
 sse41 = no
 avx2 = no
 pext = no
 avx512 = no
+vnni = no
 neon = no
 ARCH = x86-64-modern
 
 ### 2.2 Architecture specific
+
 ifeq ($(ARCH),general-32)
 	arch = any
 	bits = 32
@@ -110,6 +119,7 @@ ifeq ($(ARCH),x86-32)
 	arch = i386
 	bits = 32
 	prefetch = yes
+	mmx = yes
 	sse = yes
 endif
 
@@ -137,7 +147,7 @@ ifeq ($(ARCH),x86-64-ssse3)
 	ssse3 = yes
 endif
 
-ifeq ($(ARCH),x86-64-modern)
+ifeq ($(ARCH),$(filter $(ARCH),x86-64-sse41-popcnt x86-64-modern))
 	arch = x86_64
 	prefetch = yes
 	popcnt = yes
@@ -146,16 +156,17 @@ ifeq ($(ARCH),x86-64-modern)
 	sse41 = yes
 endif
 
-ifeq ($(ARCH),x86-64-sse41-popcnt)
+ifeq ($(ARCH),x86-64-avx2)
 	arch = x86_64
 	prefetch = yes
 	popcnt = yes
 	sse = yes
 	ssse3 = yes
 	sse41 = yes
+	avx2 = yes
 endif
 
-ifeq ($(ARCH),x86-64-avx2)
+ifeq ($(ARCH),x86-64-bmi2)
 	arch = x86_64
 	prefetch = yes
 	popcnt = yes
@@ -163,9 +174,10 @@ ifeq ($(ARCH),x86-64-avx2)
 	ssse3 = yes
 	sse41 = yes
 	avx2 = yes
+	pext = yes
 endif
 
-ifeq ($(ARCH),x86-64-bmi2)
+ifeq ($(ARCH),x86-64-avx512)
 	arch = x86_64
 	prefetch = yes
 	popcnt = yes
@@ -174,9 +186,10 @@ ifeq ($(ARCH),x86-64-bmi2)
 	sse41 = yes
 	avx2 = yes
 	pext = yes
+	avx512 = yes
 endif
 
-ifeq ($(ARCH),x86-64-avx512)
+ifeq ($(ARCH),x86-64-vnni)
 	arch = x86_64
 	prefetch = yes
 	popcnt = yes
@@ -186,6 +199,7 @@ ifeq ($(ARCH),x86-64-avx512)
 	avx2 = yes
 	pext = yes
 	avx512 = yes
+	vnni = yes
 endif
 
 ifeq ($(ARCH),armv7)
@@ -250,7 +264,7 @@ ifeq ($(COMP),gcc)
 	ifneq ($(KERNEL),Darwin)
 	   LDFLAGS += -Wl,--no-as-needed
 	endif
-	
+
 	gccversion = $(shell $(CXX) --version)
 	gccisclang = $(findstring clang,$(gccversion))
 endif
@@ -414,7 +428,14 @@ endif
 ifeq ($(avx512),yes)
 	CXXFLAGS += -DUSE_AVX512
 	ifeq ($(comp),$(filter $(comp),gcc clang mingw))
-		CXXFLAGS += -mavx512bw
+		CXXFLAGS += -mavx512f -mavx512bw
+	endif
+endif
+
+ifeq ($(vnni),yes)
+	CXXFLAGS += -DUSE_VNNI
+	ifeq ($(comp),$(filter $(comp),gcc clang mingw))
+		CXXFLAGS += -mavx512vnni -mavx512dq -mavx512vl
 	endif
 endif
 
@@ -432,12 +453,19 @@ ifeq ($(ssse3),yes)
 	endif
 endif
 
+ifeq ($(mmx),yes)
+	CXXFLAGS += -DUSE_MMX
+	ifeq ($(comp),$(filter $(comp),gcc clang mingw))
+		CXXFLAGS += -mmmx
+	endif
+endif
+
 ifeq ($(neon),yes)
 	CXXFLAGS += -DUSE_NEON
 endif
 
 ifeq ($(arch),x86_64)
-	CXXFLAGS += -DUSE_SSE2
+	CXXFLAGS += -msse2 -DUSE_SSE2
 endif
 
 ### 3.7 pext
@@ -463,6 +491,11 @@ 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=thin
 		LDFLAGS += $(CXXFLAGS)
@@ -499,24 +532,26 @@ help:
 	@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-vnni             > x86 64-bit with vnni support"
 	@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 SSE)"
+	@echo "x86-32                  > x86 32-bit (also enables MMX and SSE)"
 	@echo "x86-32-old              > x86 32-bit fall back for old hardware"
 	@echo "ppc-64                  > PPC 64-bit"
 	@echo "ppc-32                  > PPC 32-bit"
@@ -535,17 +570,20 @@ help:
 	@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 "-------------------------------\n"
+	@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 \
@@ -555,7 +593,7 @@ help:
 build: config-sanity
 	$(MAKE) ARCH=$(ARCH) COMP=$(COMP) all
 
-profile-build: config-sanity objclean profileclean
+profile-build: net config-sanity objclean profileclean
 	@echo ""
 	@echo "Step 1/4. Building instrumented executable ..."
 	$(MAKE) ARCH=$(ARCH) COMP=$(COMP) $(profile_make)
@@ -586,8 +624,10 @@ net:
 	$(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))
+	$(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 echo "Downloading $(nnuedownloadurl)"; $(curl_or_wget) $(nnuedownloadurl) > $(nnuenet); fi
+	$(eval shasum_command := $(shell if hash shasum 2>/dev/null; then echo "shasum -a 256 "; elif hash sha256sum 2>/dev/null; then echo "sha256sum "; fi))
+	@if [ "$(nnuenet)" != "nn-"`$(shasum_command) $(nnuenet) | cut -c1-12`".nnue" ]; then echo "Failed download or $(nnuenet) corrupted, please delete!"; exit 1; fi
 
 # clean binaries and objects
 objclean:
@@ -596,7 +636,7 @@ objclean:
 # clean auxiliary profiling files
 profileclean:
 	@rm -rf profdir
-	@rm -f bench.txt *.gcda *.gcno ./syzygy/*.gcda ./nnue/*.gcda ./nnue/features/*.gcda
+	@rm -f bench.txt *.gcda *.gcno ./syzygy/*.gcda ./nnue/*.gcda ./nnue/features/*.gcda *.s
 	@rm -f stockfish.profdata *.profraw
 
 default:
@@ -626,6 +666,7 @@ config-sanity:
 	@echo "avx2: '$(avx2)'"
 	@echo "pext: '$(pext)'"
 	@echo "avx512: '$(avx512)'"
+	@echo "vnni: '$(vnni)'"
 	@echo "neon: '$(neon)'"
 	@echo ""
 	@echo "Flags:"
@@ -650,6 +691,7 @@ config-sanity:
 	@test "$(avx2)" = "yes" || test "$(avx2)" = "no"
 	@test "$(pext)" = "yes" || test "$(pext)" = "no"
 	@test "$(avx512)" = "yes" || test "$(avx512)" = "no"
+	@test "$(vnni)" = "yes" || test "$(vnni)" = "no"
 	@test "$(neon)" = "yes" || test "$(neon)" = "no"
 	@test "$(comp)" = "gcc" || test "$(comp)" = "icc" || test "$(comp)" = "mingw" || test "$(comp)" = "clang"