Update cluster branch to latest master

Fixes a few merge conflicts.

Verified equal bench for 1 rank, and expected performance master vs cluster with 2 ranks.

Score of cluster vs master: 196 - 54 - 400  [0.609] 650
Elo difference: 77.1 +/- 16.3, LOS: 100.0 %, DrawRatio: 61.5 %

No functional change.
This commit is contained in:
Joost VandeVondele
2020-06-28 11:45:25 +02:00
38 changed files with 1811 additions and 1172 deletions
+37 -26
View File
@@ -35,10 +35,14 @@ BINDIR = $(PREFIX)/bin
### Built-in benchmark for pgo-builds
PGOBENCH = ./$(EXE) bench
### Object files
OBJS = benchmark.o bitbase.o bitboard.o cluster.o endgame.o evaluate.o main.o \
material.o misc.o movegen.o movepick.o pawns.o position.o psqt.o \
search.o thread.o timeman.o tt.o uci.o ucioption.o syzygy/tbprobe.o
### Source and object files
SRCS = benchmark.cpp bitbase.cpp bitboard.cpp cluster.cpp endgame.cpp evaluate.cpp main.cpp \
material.cpp misc.cpp movegen.cpp movepick.cpp pawns.cpp position.cpp psqt.cpp \
search.cpp thread.cpp timeman.cpp tt.cpp uci.cpp ucioption.cpp tune.cpp syzygy/tbprobe.cpp
OBJS = $(notdir $(SRCS:.cpp=.o))
VPATH = syzygy
### Establish the operating system name
KERNEL = $(shell uname -s)
@@ -50,7 +54,7 @@ endif
### Section 2. High-level Configuration
### ==========================================================================
#
# flag --- Comp switch --- Description
# flag --- Comp switch --- Description
# ----------------------------------------------------------------------------
#
# debug = yes/no --- -DNDEBUG --- Enable/Disable debug mode
@@ -74,7 +78,7 @@ endif
optimize = yes
debug = no
sanitize = no
bits = 32
bits = 64
prefetch = no
popcnt = no
sse = no
@@ -82,36 +86,35 @@ pext = no
mpi = no
### 2.2 Architecture specific
ifeq ($(ARCH),general-32)
arch = any
bits = 32
endif
ifeq ($(ARCH),x86-32-old)
arch = i386
bits = 32
endif
ifeq ($(ARCH),x86-32)
arch = i386
bits = 32
prefetch = yes
sse = yes
endif
ifeq ($(ARCH),general-64)
arch = any
bits = 64
endif
ifeq ($(ARCH),x86-64)
arch = x86_64
bits = 64
prefetch = yes
sse = yes
endif
ifeq ($(ARCH),x86-64-modern)
arch = x86_64
bits = 64
prefetch = yes
popcnt = yes
sse = yes
@@ -119,7 +122,6 @@ endif
ifeq ($(ARCH),x86-64-bmi2)
arch = x86_64
bits = 64
prefetch = yes
popcnt = yes
sse = yes
@@ -129,26 +131,31 @@ endif
ifeq ($(ARCH),armv7)
arch = armv7
prefetch = yes
bits = 32
endif
ifeq ($(ARCH),armv8)
arch = armv8-a
prefetch = yes
popcnt = yes
endif
ifeq ($(ARCH),ppc-32)
arch = ppc
bits = 32
endif
ifeq ($(ARCH),ppc-64)
arch = ppc64
bits = 64
popcnt = yes
prefetch = yes
endif
### ==========================================================================
### Section 3. Low-level configuration
### Section 3. Low-level Configuration
### ==========================================================================
### 3.1 Selecting compiler (default = gcc)
CXXFLAGS += -Wall -Wcast-qual -fno-exceptions -std=c++11 $(EXTRACXXFLAGS)
DEPENDFLAGS += -std=c++11
LDFLAGS += $(EXTRALDFLAGS)
@@ -162,7 +169,7 @@ ifeq ($(COMP),gcc)
CXX=g++
CXXFLAGS += -pedantic -Wextra -Wshadow
ifeq ($(ARCH),armv7)
ifeq ($(ARCH),$(filter $(ARCH),armv7 armv8))
ifeq ($(OS),Android)
CXXFLAGS += -m$(bits)
LDFLAGS += -m$(bits)
@@ -219,7 +226,7 @@ ifeq ($(COMP),clang)
endif
endif
ifeq ($(ARCH),armv7)
ifeq ($(ARCH),$(filter $(ARCH),armv7 armv8))
ifeq ($(OS),Android)
CXXFLAGS += -m$(bits)
LDFLAGS += -m$(bits)
@@ -317,7 +324,7 @@ endif
### 3.6 popcnt
ifeq ($(popcnt),yes)
ifeq ($(arch),ppc64)
ifeq ($(arch),$(filter $(arch),ppc64 armv8-a))
CXXFLAGS += -DUSE_POPCNT
else ifeq ($(comp),icc)
CXXFLAGS += -msse3 -DUSE_POPCNT
@@ -334,7 +341,7 @@ ifeq ($(pext),yes)
endif
endif
### 3.8 Link Time Optimization, it works since gcc 4.5 but not on mingw under Windows.
### 3.8 Link Time Optimization
### This is a mix of compile and link time options because the lto link phase
### needs access to the optimization flags.
ifeq ($(optimize),yes)
@@ -344,6 +351,9 @@ ifeq ($(debug), no)
LDFLAGS += $(CXXFLAGS)
endif
# To use LTO and static linking on windows, the tool chain requires a recent gcc:
# gcc version 10.1 in msys2 or TDM-GCC version 9.2 are know to work, older might not.
# So, only enable it for a cross from Linux by default.
ifeq ($(comp),mingw)
ifeq ($(KERNEL),Linux)
CXXFLAGS += -flto
@@ -368,7 +378,7 @@ ifneq (,$(findstring mpi, $(CXX)))
endif
### ==========================================================================
### Section 4. Public targets
### Section 4. Public Targets
### ==========================================================================
help:
@@ -395,6 +405,7 @@ help:
@echo "ppc-64 > PPC 64-bit"
@echo "ppc-32 > PPC 32-bit"
@echo "armv7 > ARMv7 32-bit"
@echo "armv8 > ARMv8 64-bit"
@echo "general-64 > unspecified 64-bit"
@echo "general-32 > unspecified 32-bit"
@echo ""
@@ -417,7 +428,7 @@ help:
@echo ""
.PHONY: help build profile-build strip install clean objclean profileclean help \
.PHONY: help build profile-build strip install clean objclean profileclean \
config-sanity icc-profile-use icc-profile-make gcc-profile-use gcc-profile-make \
clang-profile-use clang-profile-make
@@ -458,14 +469,14 @@ objclean:
# clean auxiliary profiling files
profileclean:
@rm -rf profdir
@rm -f bench.txt *.gcda ./syzygy/*.gcda *.gcno ./syzygy/*.gcno
@rm -f bench.txt *.gcda *.gcno
@rm -f stockfish.profdata *.profraw
default:
help
### ==========================================================================
### Section 5. Private targets
### Section 5. Private Targets
### ==========================================================================
all: $(EXE) .depend
@@ -497,7 +508,8 @@ config-sanity:
@test "$(sanitize)" = "undefined" || test "$(sanitize)" = "thread" || test "$(sanitize)" = "address" || test "$(sanitize)" = "no"
@test "$(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)" = "ppc64" || test "$(arch)" = "ppc" || \
test "$(arch)" = "armv7" || test "$(arch)" = "armv8-a"
@test "$(bits)" = "32" || test "$(bits)" = "64"
@test "$(prefetch)" = "yes" || test "$(prefetch)" = "no"
@test "$(popcnt)" = "yes" || test "$(popcnt)" = "no"
@@ -545,7 +557,6 @@ icc-profile-use:
all
.depend:
-@$(CXX) $(DEPENDFLAGS) -MM $(OBJS:.o=.cpp) > $@ 2> /dev/null
-@$(CXX) $(DEPENDFLAGS) -MM $(SRCS) > $@ 2> /dev/null
-include .depend