Add pgo builds with GCC and LLVM to Makefile

Makefile

@@ -49,10 +49,17 @@ ifdef CXXFLAGS
 else
   CXXFLAGS:= -O2 -x c++
 endif
+
 CXXFLAGS+= -Wall -Wextra -Wno-unused-parameter -Wno-ignored-qualifiers
+
 ifdef HAVE_GCC
   CXXFLAGS+= -Wno-multichar -Wunused -fno-rtti -Woverloaded-virtual -Wnon-virtual-dtor -std=c++14
 endif
+
+ifdef HAVE_CLANG
+  CXXFLAGS+= -Wno-multichar -Wunused -fno-rtti -Woverloaded-virtual -Wnon-virtual-dtor -std=c++14
+endif
+
 ifdef CLANG_WARNINGS
   CXXFLAGS+= -Weverything -Wno-c++17-extensions -Wno-c++98-compat -Wno-c++98-compat-pedantic \
     -Wno-double-promotion -Wno-switch-enum -Wno-conversion -Wno-covered-switch-default \
@@ -62,23 +69,60 @@ ifdef CLANG_WARNINGS
 endif
 
 ifdef PROFILE
-  PROF:= -g -pg -fprofile-arcs -ftest-coverage
+  PROF:= -pg -fprofile-arcs -ftest-coverage
   CXXFLAGS+= $(PROF)
+endif
+
+ifdef DEBUG
+	CXXFLAGS += -g
 else
   ifdef HAVE_GCC
     CXXFLAGS+= -fomit-frame-pointer
   endif
-endif
 
+	ifdef HAVE_CLANG
+		CXXFLAGS+= -fomit-frame-pointer
+	endif
+endif
 
 #######################################################################
 # Misc stuff - you should never have to edit this                     #
 #######################################################################
 
 EXECUTABLE := stella$(EXEEXT)
+EXECUTABLE_PROFILE_GENERATE := stella-pgo-generate$(EXEEXT)
+EXECUTABLE_PROFILE_USE := stella-pgo$(EXEEXT)
+
+PROFILE_DIR = $(CURDIR)/profile
+PROFILE_OUT = $(PROFILE_DIR)/out
+PROFILE_STAMP = profile.stamp
+
+CXXFLAGS_PROFILE_GENERATE = $(CXXFLAGS)
+CXXFLAGS_PROFILE_USE = $(CXXFLAGS)
+LDFLAGS_PROFILE_GENERATE = $(LDFLAGS)
+STELLA_PROFILE_GENERATE = $(BINARY_LOADER) ./$(EXECUTABLE_PROFILE_GENERATE) -profile \
+	$(PROFILE_DIR)/128.bin:10 \
+	$(PROFILE_DIR)/catharsis_theory.bin:60
+
+ifdef HAVE_CLANG
+	CXXFLAGS_PROFILE_GENERATE += -fprofile-generate=$(PROFILE_OUT)
+	CXXFLAGS_PROFILE_USE += -fprofile-use=$(PROFILE_OUT)
+	LDFLAGS_PROFILE_GENERATE += -fprofile-generate
+	STELLA_PROFILE_GENERATE := \
+		LLVM_PROFILE_FILE="$(PROFILE_OUT)/default.profraw" $(STELLA_PROFILE_GENERATE) && \
+		$(LLVM_PROFDATA) merge -o $(PROFILE_OUT)/default.profdata $(PROFILE_OUT)/default.profraw
+endif
+
+ifdef HAVE_GCC
+	CXXFLAGS_PROFILE_GENERATE += -fprofile-generate -fprofile-dir=$(PROFILE_OUT)
+	CXXFLAGS_PROFILE_USE += -fprofile-use -fprofile-dir=$(PROFILE_OUT)
+	LDFLAGS_PROFILE_GENERATE += -fprofile-generate
+	LDFLAGS_PROFILE_USE += -fprofile-generate
+endif
 
 all: $(EXECUTABLE)
 
+pgo: $(EXECUTABLE_PROFILE_USE)
 
 ######################################################################
 # Various minor settings
@@ -119,57 +163,118 @@ CPPFLAGS:= $(DEFINES) $(INCLUDES)
 DEPDIRS = $(addsuffix /$(DEPDIR),$(MODULE_DIRS))
 DEPFILES =
 
+OBJS_PROFILE_GENERATE=$(OBJS:.o=.pgen.o)
+OBJS_PROFILE_USE=$(OBJS:.o=.pgo.o)
+
 # The build rule for the Stella executable
 $(EXECUTABLE): $(OBJS)
 	$(LD) $(LDFLAGS) $(PRE_OBJS_FLAGS) $+ $(POST_OBJS_FLAGS) $(LIBS) $(PROF) -o $@
 
+$(EXECUTABLE_PROFILE_GENERATE): $(OBJS_PROFILE_GENERATE)
+	$(LD) $(LDFLAGS_PROFILE_GENERATE) $(PRE_OBJS_FLAGS) $+ $(POST_OBJS_FLAGS) $(LIBS) $(PROF) -o $@
+
+$(EXECUTABLE_PROFILE_USE): $(OBJS_PROFILE_USE)
+	$(LD) $(LDFLAGS_PROFILE_USE) $(PRE_OBJS_FLAGS) $+ $(POST_OBJS_FLAGS) $(LIBS) $(PROF) -o $@
+
 distclean: clean
 	$(RM_REC) $(DEPDIRS)
 	$(RM) build.rules config.h config.mak config.log
 
 clean:
-	$(RM) $(OBJS) $(EXECUTABLE)
+	-$(RM) -fr \
+		$(OBJS) $(OBJS_PROFILE_GENERATE) $(OBJS_PROFILE_USE) \
+		$(EXECUTABLE) $(EXECUTABLE_PROFILE_GENERATE) $(EXECUTABLE_PROFILE_USE) \
+		$(PROFILE_OUT) $(PROFILE_STAMP)
 
 .PHONY: all clean dist distclean
 
 .SUFFIXES: .cxx
 
+define create_depdir
+$(MKDIR) $(*D)/$(DEPDIR)
+endef
+
+define merge_dep
+$(ECHO) "$(*D)/" > $(*D)/$(DEPDIR)/$(*F).d
+$(CAT) "$(*D)/$(DEPDIR)/$(*F).d2" >> "$(*D)/$(DEPDIR)/$(*F).d"
+$(RM) "$(*D)/$(DEPDIR)/$(*F).d2"
+endef
 
 ifndef CXX_UPDATE_DEP_FLAG
 # If you use GCC, disable the above and enable this for intelligent
 # dependency tracking.
 CXX_UPDATE_DEP_FLAG = -Wp,-MMD,"$(*D)/$(DEPDIR)/$(*F).d2"
-.cxx.o:
-	$(MKDIR) $(*D)/$(DEPDIR)
-	$(CXX) $(CXX_UPDATE_DEP_FLAG) $(CXXFLAGS) $(CPPFLAGS) -c $(<) -o $*.o
-	$(ECHO) "$(*D)/" > $(*D)/$(DEPDIR)/$(*F).d
-	$(CAT) "$(*D)/$(DEPDIR)/$(*F).d2" >> "$(*D)/$(DEPDIR)/$(*F).d"
-	$(RM) "$(*D)/$(DEPDIR)/$(*F).d2"
 
-.c.o:
+%.o: %.cxx
-	$(MKDIR) $(*D)/$(DEPDIR)
+	$(create_depdir)
-	$(CC) $(CXX_UPDATE_DEP_FLAG) $(CXXFLAGS) $(CPPFLAGS) -c $(<) -o $*.o
+	$(CXX) $(CXX_UPDATE_DEP_FLAG) $(CXXFLAGS) $(CPPFLAGS) -c $(<) -o $@
-	$(ECHO) "$(*D)/" > $(*D)/$(DEPDIR)/$(*F).d
+	$(merge_dep)
-	$(CAT) "$(*D)/$(DEPDIR)/$(*F).d2" >> "$(*D)/$(DEPDIR)/$(*F).d"
+
-	$(RM) "$(*D)/$(DEPDIR)/$(*F).d2"
+%.o: %.c
+	$(create_depdir)
+	$(CC) $(CXX_UPDATE_DEP_FLAG) $(CXXFLAGS) $(CPPFLAGS) -c $(<) -o $@
+	$(merge_dep)
+
+%.pgen.o: %.cxx
+	$(create_depdir)
+	$(CXX) $(CXX_UPDATE_DEP_FLAG) $(CXXFLAGS_PROFILE_GENERATE) $(CPPFLAGS) -c $(<) -o $@
+	$(merge_dep)
+
+%.pgen.o: %.cxx
+	$(create_depdir)
+	$(CC) $(CXX_UPDATE_DEP_FLAG) $(CXXFLAGS_PROFILE_GENERATE) $(CPPFLAGS) -c $(<) -o $@
+	$(merge_dep)
+
+%.pgo.o: %.cxx $(PROFILE_STAMP)
+	$(create_depdir)
+	$(CXX) $(CXX_UPDATE_DEP_FLAG) $(CXXFLAGS_PROFILE_USE) $(CPPFLAGS) -c $(<) -o $@
+	$(merge_dep)
+
+%.pgo.o: %.cxx $(PROFILE_STAMP)
+	$(create_depdir)
+	$(CC) $(CXX_UPDATE_DEP_FLAG) $(CXXFLAGS_PROFILE_USE) $(CPPFLAGS) -c $(<) -o $@
+	$(merge_dep)
+
 else
 # If you even have GCC 3.x, you can use this build rule, which is safer; the above
 # rule can get you into a bad state if you Ctrl-C at the wrong moment.
 # Also, with this GCC inserts additional dummy rules for the involved headers,
 # which ensures a smooth compilation even if said headers become obsolete.
-.cxx.o:
+%.o: %.cxx
-	$(MKDIR) $(*D)/$(DEPDIR)
+	$(create_depdir)
-	$(CXX) $(CXX_UPDATE_DEP_FLAG) $(CXXFLAGS) $(CPPFLAGS) -c $(<) -o $*.o
+	$(CXX) $(CXX_UPDATE_DEP_FLAG) $(CXXFLAGS) $(CPPFLAGS) -c $(<) -o $@
+
+%.o: %.c
+	$(create_depdir)
+	$(CC) $(CXX_UPDATE_DEP_FLAG) $(CXXFLAGS) $(CPPFLAGS) -c $(<) -o $@
+
+%.pgen.o: %.cxx
+	$(create_depdir)
+	$(CXX) $(CXX_UPDATE_DEP_FLAG) $(CXXFLAGS_PROFILE_GENERATE) $(CPPFLAGS) -c $(<) -o $@
+
+%.pgen.o: %.c
+	$(create_depdir)
+	$(CC) $(CXX_UPDATE_DEP_FLAG) $(CXXFLAGS_PROFILE_GENERATE) $(CPPFLAGS) -c $(<) -o $@
+
+%.pgo.o: %.cxx $(PROFILE_STAMP)
+	$(create_depdir)
+	$(CXX) $(CXX_UPDATE_DEP_FLAG) $(CXXFLAGS_PROFILE_USE) $(CPPFLAGS) -c $(<) -o $@
+
+%.pgo.o: %.c $(PROFILE_STAMP)
+	$(create_depdir)
+	$(CC) $(CXX_UPDATE_DEP_FLAG) $(CXXFLAGS_PROFILE_USE) $(CPPFLAGS) -c $(<) -o $@
 
-.c.o:
-	$(MKDIR) $(*D)/$(DEPDIR)
-	$(CC) $(CXX_UPDATE_DEP_FLAG) $(CXXFLAGS) $(CPPFLAGS) -c $(<) -o $*.o
 endif
 
 # Include the dependency tracking files. We add /dev/null at the end
 # of the list to avoid a warning/error if no .d file exist
 -include $(wildcard $(addsuffix /*.d,$(DEPDIRS))) /dev/null
 
+$(PROFILE_STAMP): $(EXECUTABLE_PROFILE_GENERATE)
+	-rm -fr $(PROFILE_OUT)
+	$(STELLA_PROFILE_GENERATE)
+	touch $(PROFILE_STAMP)
+
 # check if configure has been run or has been changed since last run
 config.mak: $(srcdir)/configure
 	@echo "You need to run ./configure before you can run make"

configure

@@ -24,6 +24,7 @@ _build_joystick=yes
 _build_cheats=yes
 _build_static=no
 _build_profile=no
+_build_debug=no
 
 # more defaults
 _ranlib=ranlib
@@ -204,6 +205,8 @@ Optional Features:
   --disable-static
   --enable-profile       build binary with profiling info [disabled]
   --disable-profile
+	--enable-debug				 build with debugging symbols [disabled]
+	--disable-debug
   --force-builtin-libpng force use of built-in libpng library [auto]
 
 Optional Libraries:
@@ -241,6 +244,8 @@ for ac_option in $@; do
       --disable-static)         _build_static=no     ;;
       --enable-profile)         _build_profile=yes   ;;
       --disable-profile)        _build_profile=no    ;;
+			--enable-debug)						_build_debug=yes		 ;;
+			--disable-debug)          _build_debug=false	 ;;
       --force-builtin-libpng)   _libpng=no           ;;
       --with-sdl-prefix=*)
         arg=`echo $ac_option | cut -d '=' -f 2`
@@ -404,6 +409,8 @@ if test "$have_clang" = yes; then
 			cxx_version="$cxx_version, bad"
 			cxx_verc_fail=yes
 		fi
+
+		_make_def_CLANG_WARNINGS='CLANG_WARNINGS = 1'
 	else
     # Need at least version 3.5
 		if [ $clang_major -ge 4 ] || [ $clang_major -eq 3 -a $clang_minor -ge 5 ]; then
@@ -420,9 +427,8 @@ if test "$have_clang" = yes; then
 		fi
 	fi
 	CXXFLAGS="$CXXFLAGS"
-	_make_def_HAVE_GCC3='HAVE_GCC3 = 1'
+	_make_def_HAVE_CLANG='HAVE_CLANG = 1'
 	add_line_to_config_mk 'CXX_UPDATE_DEP_FLAG = -MMD -MF "$(*D)/$(DEPDIR)/$(*F).d" -MQ "$@" -MP'
-	_make_def_HAVE_GCC='HAVE_GCC = 1'
 	echo "$cxx_version"
 
 elif test "$have_gcc" = yes; then
@@ -448,7 +454,6 @@ elif test "$have_gcc" = yes; then
 			;;
 	esac
 	CXXFLAGS="$CXXFLAGS"
-	_make_def_HAVE_GCC3='HAVE_GCC3 = 1'
 	add_line_to_config_mk 'CXX_UPDATE_DEP_FLAG = -MMD -MF "$(*D)/$(DEPDIR)/$(*F).d" -MQ "$@" -MP'
 	_make_def_HAVE_GCC='HAVE_GCC = 1'
 	echo "$cxx_version"
@@ -670,11 +675,20 @@ fi
 if test "$_build_profile" = yes ; then
 	echo_n "   Profiling enabled"
 	echo
+
+	_build_debug=yes
 else
 	echo_n "   Profiling disabled"
 	echo
 fi
 
+if test "$_build_debug" = yes ; then
+	echo_n "   Debug symbols enabled"
+	echo
+else
+	echo_n "   Debug symbols disabled"
+	echo
+fi
 
 #
 # Now, add the appropriate defines/libraries/headers
@@ -774,6 +788,9 @@ if test "$_build_profile" = no ; then
 	_build_profile=
 fi
 
+if test "$_build_debug" = no ; then
+	_build_debug=
+fi
 
 echo "Creating config.mak"
 cat > config.mak << EOF
@@ -795,6 +812,8 @@ ZIP := $_zip
 CP := $_cp
 WINDOWSPATH=$_windowspath
 STRIP := $_strip
+LLVM_PROFDATA := llvm-profdata
+BINARY_LOADER :=
 
 MODULES += $MODULES
 MODULE_DIRS += $MODULE_DIRS
@@ -805,9 +824,10 @@ BINDIR := $_bindir
 DOCDIR := $_docdir
 DATADIR := $_datadir
 PROFILE := $_build_profile
+DEBUG   := $_build_debug
 
 $_make_def_HAVE_GCC
-$_make_def_HAVE_GCC3
+$_make_def_HAVE_CLANG
 $_make_def_CLANG_WARNINGS
 
 INCLUDES += $INCLUDES
@@ -818,12 +838,11 @@ $_config_mk_data
 EOF
 
 # This should be taken care of elsewhere, but I'm not sure where
-rm -f stella-conf*
+rm -f stella-conf stella-conf.cxx
 
 if test "$_host_os" = darwin; then
 	cat <<EOI
 
-WARNING: plain UNIX-style builds on macOS without XCode have degraded functionality
+WARNING: plain UNIX-style builds on macOS without XCode are unsupported. Continue on your own risk...
-and are unsupported. Continue on your own risk...
 EOI
 fi

profile/README.md

@@ -0,0 +1,4 @@
+The ROMS that are used for profiling were created by and are by courtesy of
+
+  * Spiceware (`128.bin`)
+  * Kylearan (Catharsis Theory)