Correctly describe POPCNT compile
[stockfish] / src / Makefile
index 25bb0f0..bc7dc5c 100644 (file)
@@ -1,8 +1,6 @@
 # Stockfish, a UCI chess playing engine derived from Glaurung 2.1
 # Copyright (C) 2004-2008 Tord Romstad (Glaurung author)
-# Copyright (C) 2008-2010 Marco Costalba, Joona Kiiski, Tord Romstad
-#
-# This file is part of Stockfish.
+# Copyright (C) 2008-2014 Marco Costalba, Joona Kiiski, Tord Romstad
 #
 # Stockfish is free software: you can redistribute it and/or modify
 # it under the terms of the GNU General Public License as published by
 ### Section 1. General Configuration
 ### ==========================================================================
 
+### Establish the operating system name
+UNAME = $(shell uname)
+
 ### Executable name
 EXE = stockfish
 
 ### Installation dir definitions
 PREFIX = /usr/local
+# Haiku has a non-standard filesystem layout
+ifeq ($(UNAME),Haiku)
+       PREFIX=/boot/system/non-packaged
+endif
 BINDIR = $(PREFIX)/bin
 
 ### Built-in benchmark for pgo-builds
-PGOBENCH = ./$(EXE) bench 32 1 10 default depth
+PGOBENCH = ./$(EXE) bench 32 1 1 default time
 
 ### Object files
-OBJS = application.o bitboard.o pawns.o material.o endgame.o evaluate.o main.o \
-       misc.o move.o movegen.o history.o movepick.o search.o position.o \
-       direction.o tt.o uci.o ucioption.o book.o bitbase.o san.o benchmark.o timeman.o
-
+OBJS = benchmark.o bitbase.o bitboard.o endgame.o evaluate.o main.o \
+       material.o misc.o movegen.o movepick.o pawns.o position.o \
+       search.o thread.o timeman.o tt.o uci.o ucioption.o
 
 ### ==========================================================================
 ### Section 2. High-level Configuration
@@ -45,148 +49,94 @@ OBJS = application.o bitboard.o pawns.o material.o endgame.o evaluate.o main.o \
 # flag                --- Comp switch --- Description
 # ----------------------------------------------------------------------------
 #
-# debug = no/yes      --- -DNDEBUG    --- Enable/Disable debug mode
+# debug = yes/no      --- -DNDEBUG         --- Enable/Disable debug mode
 # optimize = yes/no   --- (-O3/-fast etc.) --- Enable/Disable optimizations
-# arch = (name)       --- (-arch)     --- Target architecture
-# os = (name)         ---             --- Target operating system
-# bits = 64/32        --- -DIS_64BIT  --- 64-/32-bit operating system
-# bigendian = no/yes  --- -DBIGENDIAN --- big/little-endian byte order
-# prefetch = no/yes   --- -DUSE_PREFETCH  --- Use prefetch x86 asm-instruction
-# bsfq = no/yes       --- -DUSE_BSFQ  --- Use bsfq x86_64 asm-instruction
-#                                     --- (Works only with GCC and ICC 64-bit)
-# popcnt = no/yes     --- -DUSE_POPCNT --- Use popcnt x86_64 asm-instruction
+# arch = (name)       --- (-arch)          --- Target architecture
+# os = (name)         ---                  --- Target operating system
+# bits = 64/32        --- -DIS_64BIT       --- 64-/32-bit operating system
+# prefetch = yes/no   --- -DUSE_PREFETCH   --- Use prefetch x86 asm-instruction
+# bsfq = yes/no       --- -DUSE_BSFQ       --- Use bsfq x86_64 asm-instruction (only
+#                                              with GCC and ICC 64-bit)
+# popcnt = yes/no     --- -DUSE_POPCNT     --- Use popcnt x86_64 asm-instruction
+# sse = yes/no        --- -msse            --- Use Intel Streaming SIMD Extensions
+# pext = yes/no       --- -DUSE_PEXT       --- Use pext x86_64 asm-instruction
 #
 # 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.
 
-### 2.1. General
-debug = no
+### 2.1. General and architecture defaults
 optimize = yes
+debug = no
+os = any
+bits = 32
+prefetch = no
+bsfq = no
+popcnt = no
+sse = no
+pext = no
 
 ### 2.2 Architecture specific
 
-# General-section
-ifeq ($(ARCH),general-64)
+ifeq ($(ARCH),general-32)
        arch = any
-       os = any
-       bits = 64
-       bigendian = no
-       prefetch = no
-       bsfq = no
-       popcnt = no
 endif
 
-ifeq ($(ARCH),general-32)
-       arch = any
-       os = any
-       bits = 32
-       bigendian = no
-       prefetch = no
-       bsfq = no
-       popcnt = no
+ifeq ($(ARCH),x86-32-old)
+       arch = i386
 endif
 
-ifeq ($(ARCH),bigendian-64)
-       arch = any
-       os = any
-       bits = 64
-       bigendian = yes
-       prefetch = no
-       bsfq = no
-       popcnt = no
+ifeq ($(ARCH),x86-32)
+       arch = i386
+       prefetch = yes
+       sse = yes
 endif
 
-ifeq ($(ARCH),bigendian-32)
+ifeq ($(ARCH),general-64)
        arch = any
-       os = any
-       bits = 32
-       bigendian = yes
-       prefetch = no
-       bsfq = no
-       popcnt = no
+       bits = 64
 endif
 
-# x86-section
 ifeq ($(ARCH),x86-64)
        arch = x86_64
-       os = any
        bits = 64
-       bigendian = no
        prefetch = yes
        bsfq = yes
-       popcnt = no
+       sse = yes
 endif
 
 ifeq ($(ARCH),x86-64-modern)
        arch = x86_64
-       os = any
        bits = 64
-       bigendian = no
        prefetch = yes
        bsfq = yes
        popcnt = yes
+       sse = yes
 endif
 
-ifeq ($(ARCH),x86-32)
-       arch = i386
-       os = any
-       bits = 32
-       bigendian = no
+ifeq ($(ARCH),x86-64-bmi2)
+       arch = x86_64
+       bits = 64
        prefetch = yes
-       bsfq = no
-       popcnt = no
-endif
-
-ifeq ($(ARCH),x86-32-old)
-       arch = i386
-       os = any
-       bits = 32
-       bigendian = no
-       prefetch = no
-       bsfq = no
-       popcnt = no
+       bsfq = yes
+       popcnt = yes
+       sse = yes
+       pext = yes
 endif
 
-# osx-section
-ifeq ($(ARCH),osx-ppc-64)
-       arch = ppc64
-       os = osx
-       bits = 64
-       bigendian = yes
-       prefetch = no
-       bsfq = no
-       popcnt = no
+ifeq ($(ARCH),armv7)
+       arch = armv7
+       prefetch = yes
+       bsfq = yes
 endif
 
-ifeq ($(ARCH),osx-ppc-32)
+ifeq ($(ARCH),ppc-32)
        arch = ppc
-       os = osx
-       bits = 32
-       bigendian = yes
-       prefetch = no
-       bsfq = no
-       popcnt = no
 endif
 
-ifeq ($(ARCH),osx-x86-64)
-       arch = x86_64
-       os = osx
+ifeq ($(ARCH),ppc-64)
+       arch = ppc64
        bits = 64
-       bigendian = no
-       prefetch = yes
-       bsfq = yes
-       popcnt = no
-endif
-
-ifeq ($(ARCH),osx-x86-32)
-       arch = i386
-       os = osx
-       bits = 32
-       bigendian = no
-       prefetch = yes
-       bsfq = no
-       popcnt = no
 endif
 
 
@@ -195,66 +145,72 @@ endif
 ### ==========================================================================
 
 ### 3.1 Selecting compiler (default = gcc)
+
+CXXFLAGS += -Wall -Wcast-qual -fno-exceptions -fno-rtti $(EXTRACXXFLAGS)
+LDFLAGS += $(EXTRALDFLAGS)
+
 ifeq ($(COMP),)
        COMP=gcc
 endif
 
-ifeq ($(COMP),mingw)
-       comp=mingw
+ifeq ($(COMP),gcc)
+       comp=gcc
        CXX=g++
-       profile_prepare = gcc-profile-prepare
-       profile_make = gcc-profile-make
-       profile_use = gcc-profile-use
-       profile_clean = gcc-profile-clean
+       CXXFLAGS += -ansi -pedantic -Wno-long-long -Wextra -Wshadow
 endif
 
-ifeq ($(COMP),gcc)
-       comp=gcc
+ifeq ($(COMP),mingw)
+       comp=mingw
        CXX=g++
-       profile_prepare = gcc-profile-prepare
-       profile_make = gcc-profile-make
-       profile_use = gcc-profile-use
-       profile_clean = gcc-profile-clean
+       CXXFLAGS += -Wextra -Wshadow
+       LDFLAGS += -static-libstdc++ -static-libgcc
 endif
 
 ifeq ($(COMP),icc)
        comp=icc
        CXX=icpc
-       profile_prepare = icc-profile-prepare
-       profile_make = icc-profile-make
-       profile_use = icc-profile-use
-       profile_clean = icc-profile-clean
-endif
-
-### 3.2 General compiler settings
-CXXFLAGS = -g -Wall -Wcast-qual -fno-exceptions -fno-rtti $(EXTRACXXFLAGS)
-
-ifeq ($(comp),gcc)
-       CXXFLAGS += -ansi -pedantic -Wno-long-long -Wextra
+       CXXFLAGS += -diag-disable 1476,10120 -Wcheck -Wabi -Wdeprecated -strict-ansi
 endif
 
-ifeq ($(comp),mingw)
-       CXXFLAGS += -Wno-long-long -Wextra
+ifeq ($(COMP),clang)
+       comp=clang
+       CXX=clang++
+       CXXFLAGS += -pedantic -Wno-long-long -Wextra -Wshadow
 endif
 
 ifeq ($(comp),icc)
-       CXXFLAGS += -wd383,981,10187,10188,11505,11503 -Wcheck -Wabi -Wdeprecated -strict-ansi
+       profile_prepare = icc-profile-prepare
+       profile_make = icc-profile-make
+       profile_use = icc-profile-use
+       profile_clean = icc-profile-clean
+else
+       profile_prepare = gcc-profile-prepare
+       profile_make = gcc-profile-make
+       profile_use = gcc-profile-use
+       profile_clean = gcc-profile-clean
 endif
 
-ifeq ($(os),osx)
-       CXXFLAGS += -arch $(arch)
+ifeq ($(UNAME),Darwin)
+       CXXFLAGS += -arch $(arch) -mmacosx-version-min=10.6
+       LDFLAGS += -arch $(arch) -mmacosx-version-min=10.6
 endif
 
-### 3.3 General linker settings
-LDFLAGS = -lpthread $(EXTRALDFLAGS)
-
-ifeq ($(os),osx)
-       LDFLAGS += -arch $(arch)
+### On mingw use Windows threads, otherwise POSIX
+ifneq ($(comp),mingw)
+       # On Android Bionic's C library comes with its own pthread implementation bundled in
+       ifneq ($(arch),armv7)
+               # Haiku has pthreads in its libroot, so only link it in on other platforms
+               ifneq ($(UNAME),Haiku)
+                       LDFLAGS += -lpthread
+               endif
+       endif
 endif
 
 ### 3.4 Debugging
 ifeq ($(debug),no)
        CXXFLAGS += -DNDEBUG
+else
+       CXXFLAGS += -g
 endif
 
 ### 3.5 Optimization
@@ -263,7 +219,7 @@ ifeq ($(optimize),yes)
        ifeq ($(comp),gcc)
                CXXFLAGS += -O3
 
-               ifeq ($(os),osx)
+               ifeq ($(UNAME),Darwin)
                        ifeq ($(arch),i386)
                                CXXFLAGS += -mdynamic-no-pic
                        endif
@@ -271,6 +227,10 @@ ifeq ($(optimize),yes)
                                CXXFLAGS += -mdynamic-no-pic
                        endif
                endif
+
+               ifeq ($(arch),armv7)
+                       CXXFLAGS += -fno-gcse -mthumb -march=armv7-a -mfloat-abi=softfp
+               endif
        endif
 
        ifeq ($(comp),mingw)
@@ -278,10 +238,27 @@ ifeq ($(optimize),yes)
        endif
 
        ifeq ($(comp),icc)
-               CXXFLAGS += -fast
+               ifeq ($(UNAME),Darwin)
+                       CXXFLAGS += -fast -mdynamic-no-pic
+               else
+                       CXXFLAGS += -fast
+               endif
+       endif
+
+       ifeq ($(comp),clang)
+               CXXFLAGS += -O3
 
-               ifeq ($(os),osx)
-                       CXXFLAGS += -mdynamic-no-pic
+               ifeq ($(UNAME),Darwin)
+                       ifeq ($(pext),no)
+                               CXXFLAGS += -flto
+                               LDFLAGS += $(CXXFLAGS)
+                       endif
+                       ifeq ($(arch),i386)
+                               CXXFLAGS += -mdynamic-no-pic
+                       endif
+                       ifeq ($(arch),x86_64)
+                               CXXFLAGS += -mdynamic-no-pic
+                       endif
                endif
        endif
 endif
@@ -291,36 +268,58 @@ ifeq ($(bits),64)
        CXXFLAGS += -DIS_64BIT
 endif
 
-### 3.7 Endianess
-ifeq ($(bigendian),yes)
-       CXXFLAGS += -DBIGENDIAN
-endif
-
-### 3.8 prefetch
+### 3.7 prefetch
 ifeq ($(prefetch),yes)
-       CXXFLAGS += -msse
-       DEPENDFLAGS += -msse
+       ifeq ($(sse),yes)
+               CXXFLAGS += -msse
+               DEPENDFLAGS += -msse
+       endif
 else
        CXXFLAGS += -DNO_PREFETCH
 endif
 
-### 3.9 bsfq
+### 3.8 bsfq
 ifeq ($(bsfq),yes)
        CXXFLAGS += -DUSE_BSFQ
 endif
 
-### 3.10 popcnt
+### 3.9 popcnt
 ifeq ($(popcnt),yes)
-       CXXFLAGS += -DUSE_POPCNT
+       ifeq ($(comp),icc)
+               CXXFLAGS += -msse3 -DUSE_POPCNT
+       else
+               CXXFLAGS += -msse3 -mpopcnt -DUSE_POPCNT
+       endif
+endif
+
+### 3.10 pext
+ifeq ($(pext),yes)
+       CXXFLAGS += -DUSE_PEXT
+       ifeq ($(comp),$(filter $(comp),gcc clang mingw))
+               CXXFLAGS += -mbmi -mbmi2
+       endif
+endif
+
+### 3.11 Link Time Optimization, it works since gcc 4.5 but not on mingw.
+### This is a mix of compile and link time options because the lto link phase
+### needs access to the optimization flags.
+ifeq ($(comp),gcc)
+       ifeq ($(optimize),yes)
+       ifeq ($(debug),no)
+               GCC_MAJOR := `$(CXX) -dumpversion | cut -f1 -d.`
+               GCC_MINOR := `$(CXX) -dumpversion | cut -f2 -d.`
+               ifeq (1,$(shell expr \( $(GCC_MAJOR) \> 4 \) \| \( $(GCC_MAJOR) \= 4 \& $(GCC_MINOR) \>= 5 \)))
+                       CXXFLAGS += -flto
+                       LDFLAGS += $(CXXFLAGS)
+               endif
+       endif
+       endif
 endif
 
 ### ==========================================================================
 ### Section 4. Public targets
 ### ==========================================================================
 
-default:
-       $(MAKE) ARCH=$(ARCH) COMP=$(COMP) build
-
 help:
        @echo ""
        @echo "To compile stockfish, type: "
@@ -329,45 +328,43 @@ help:
        @echo ""
        @echo "Supported targets:"
        @echo ""
-       @echo "build                > Build unoptimized version"
-       @echo "profile-build        > Build PGO-optimized version"
-       @echo "popcnt-profile-build > Build PGO-optimized version with optional popcnt-support"
-       @echo "strip                > Strip executable"
-       @echo "install              > Install executable"
-       @echo "clean                > Clean up"
-       @echo "testrun              > Make sample run"
+       @echo "build                   > Standard build"
+       @echo "profile-build           > PGO build"
+       @echo "strip                   > Strip executable"
+       @echo "install                 > Install executable"
+       @echo "clean                   > Clean up"
        @echo ""
        @echo "Supported archs:"
        @echo ""
-       @echo "x86-64               > x86 64-bit"
-       @echo "x86-64-modern        > x86 64-bit with runtime support for popcnt-instruction"
-       @echo "x86-32               > x86 32-bit excluding very old hardware without SSE-support"
-       @echo "x86-32-old           > x86 32-bit including also very old hardware"
-       @echo "osx-ppc-64           > PPC-Mac OS X 64 bit"
-       @echo "osx-ppc-32           > PPC-Mac OS X 32 bit"
-       @echo "osx-x86-64           > x86-Mac OS X 64 bit"
-       @echo "osx-x86-32           > x86-Mac OS X 32 bit"
-       @echo "general-64           > unspecified 64-bit"
-       @echo "general-32           > unspecified 32-bit"
-       @echo "bigendian-64         > unspecified 64-bit with bigendian byte order"
-       @echo "bigendian-32         > unspecified 32-bit with bigendian byte order"
+       @echo "x86-64                  > x86 64-bit"
+       @echo "x86-64-modern           > x86 64-bit with popcnt support"
+       @echo "x86-64-bmi2             > x86 64-bit with pext support"
+       @echo "x86-32                  > x86 32-bit with SSE support"
+       @echo "x86-32-old              > x86 32-bit fall back for old hardware"
+       @echo "ppc-64                  > PPC 64-bit"
+       @echo "ppc-32                  > PPC 32-bit"
+       @echo "armv7                   > ARMv7 32-bit"
+       @echo "general-64              > unspecified 64-bit"
+       @echo "general-32              > unspecified 32-bit"
        @echo ""
-       @echo "Supported comps:"
+       @echo "Supported compilers:"
        @echo ""
-       @echo "gcc                  > Gnu compiler (default)"
-       @echo "icc                  > Intel compiler"
-       @echo "mingw                > Gnu compiler with MinGW under Windows"
+       @echo "gcc                     > Gnu compiler (default)"
+       @echo "mingw                   > Gnu compiler with MinGW under Windows"
+       @echo "clang                   > LLVM Clang compiler"
+       @echo "icc                     > Intel compiler"
        @echo ""
        @echo "Non-standard targets:"
        @echo ""
-       @echo "make hpux           >  Compile for HP-UX. Compiler = aCC"
+       @echo "make hpux               >  Compile for HP-UX. Compiler = aCC"
        @echo ""
        @echo "Examples. If you don't know what to do, you likely want to run: "
        @echo ""
-       @echo "make profile-build ARCH=x86-64    (This is for 64-bit systems)"
-       @echo "make profile-build ARCH=x86-32    (This is for 32-bit systems)"
+       @echo "make build ARCH=x86-64    (This is for 64-bit systems)"
+       @echo "make build ARCH=x86-32    (This is for 32-bit systems)"
        @echo ""
 
+.PHONY: build profile-build
 build:
        $(MAKE) ARCH=$(ARCH) COMP=$(COMP) config-sanity
        $(MAKE) ARCH=$(ARCH) COMP=$(COMP) all
@@ -392,34 +389,6 @@ profile-build:
        @echo "Step 4/4. Deleting profile data ..."
        $(MAKE) ARCH=$(ARCH) COMP=$(COMP) $(profile_clean)
 
-popcnt-profile-build:
-       $(MAKE) ARCH=$(ARCH) COMP=$(COMP) config-sanity
-       @echo ""
-       @echo "Step 0/6. Preparing for profile build."
-       $(MAKE) ARCH=$(ARCH) COMP=$(COMP) $(profile_prepare)
-       @echo ""
-       @echo "Step 1/6. Building executable for benchmark (popcnt disabled)..."
-       @touch *.cpp *.h
-       $(MAKE) ARCH=x86-64 COMP=$(COMP) $(profile_make)
-       @echo ""
-       @echo "Step 2/6. Running benchmark for pgo-build (popcnt disabled)..."
-       @$(PGOBENCH) > /dev/null
-       @echo ""
-       @echo "Step 3/6. Building executable for benchmark (popcnt enabled)..."
-       @touch *.cpp *.h
-       $(MAKE) ARCH=x86-64-modern COMP=$(COMP) $(profile_make)
-       @echo ""
-       @echo "Step 4/6. Running benchmark for pgo-build (popcnt enabled)..."
-       @$(PGOBENCH) > /dev/null
-       @echo ""
-       @echo "Step 5/6. Building final executable ..."
-       @touch *.cpp *.h
-       $(MAKE) ARCH=$(ARCH) COMP=$(COMP) $(profile_use)
-       @echo ""
-       @echo "Step 6/6. Deleting profile data ..."
-       $(MAKE) ARCH=$(ARCH) COMP=$(COMP) $(profile_clean)
-       @echo ""
-
 strip:
        strip $(EXE)
 
@@ -431,8 +400,8 @@ install:
 clean:
        $(RM) $(EXE) $(EXE).exe *.o .depend *~ core bench.txt *.gcda
 
-testrun:
-       @$(PGOBENCH)
+default:
+       help
 
 ### ==========================================================================
 ### Section 5. Private targets
@@ -448,10 +417,11 @@ config-sanity:
        @echo "arch: '$(arch)'"
        @echo "os: '$(os)'"
        @echo "bits: '$(bits)'"
-       @echo "bigendian: '$(bigendian)'"
        @echo "prefetch: '$(prefetch)'"
        @echo "bsfq: '$(bsfq)'"
        @echo "popcnt: '$(popcnt)'"
+       @echo "sse: '$(sse)'"
+       @echo "pext: '$(pext)'"
        @echo ""
        @echo "Flags:"
        @echo "CXX: $(CXX)"
@@ -463,14 +433,15 @@ config-sanity:
        @test "$(debug)" = "yes" || test "$(debug)" = "no"
        @test "$(optimize)" = "yes" || test "$(optimize)" = "no"
        @test "$(arch)" = "any" || test "$(arch)" = "x86_64" || test "$(arch)" = "i386" || \
-        test "$(arch)" = "ppc64" || test "$(arch)" = "ppc"
-       @test "$(os)" = "any" || test "$(os)" = "osx"
+        test "$(arch)" = "ppc64" || test "$(arch)" = "ppc" || test "$(arch)" = "armv7"
+       @test "$(os)" = "any"
        @test "$(bits)" = "32" || test "$(bits)" = "64"
-       @test "$(bigendian)" = "yes" || test "$(bigendian)" = "no"
        @test "$(prefetch)" = "yes" || test "$(prefetch)" = "no"
        @test "$(bsfq)" = "yes" || test "$(bsfq)" = "no"
        @test "$(popcnt)" = "yes" || test "$(popcnt)" = "no"
-       @test "$(comp)" = "gcc" || test "$(comp)" = "icc" || test "$(comp)" = "mingw"
+       @test "$(sse)" = "yes" || test "$(sse)" = "no"
+       @test "$(pext)" = "yes" || test "$(pext)" = "no"
+       @test "$(comp)" = "gcc" || test "$(comp)" = "icc" || test "$(comp)" = "mingw" || test "$(comp)" = "clang"
 
 $(EXE): $(OBJS)
        $(CXX) -o $@ $(OBJS) $(LDFLAGS)
@@ -487,6 +458,7 @@ gcc-profile-make:
 gcc-profile-use:
        $(MAKE) ARCH=$(ARCH) COMP=$(COMP) \
        EXTRACXXFLAGS='-fprofile-use' \
+       EXTRALDFLAGS='-lgcov' \
        all
 
 gcc-profile-clean:
@@ -521,7 +493,7 @@ icc-profile-clean:
 
 hpux:
        $(MAKE) \
-       CXX='/opt/aCC/bin/aCC -AA +hpxstd98 -DBIGENDIAN -mt +O3 -DNDEBUG -DNO_PREFETCH' \
+       CXX='/opt/aCC/bin/aCC -AA +hpxstd98 -mt +O3 -DNDEBUG -DNO_PREFETCH' \
        CXXFLAGS="" \
        LDFLAGS="" \
        all