libretro, more fixes and cleanups for windows and osx

Makefile

@@ -174,6 +174,7 @@ OBJS += platform/libretro/libretro-common/streams/file_stream.o
 OBJS += platform/libretro/libretro-common/streams/file_stream_transforms.o
 OBJS += platform/libretro/libretro-common/vfs/vfs_implementation.o
 endif
+PLATFORM_ZLIB = 1
 endif
 
 ifeq "$(USE_FRONTEND)" "1"

Makefile.libretro

@@ -52,6 +52,8 @@ asm_ym2612 = 0
 asm_misc = 0
 asm_cdmemory = 0
 asm_mix = 0
+asm_32xdraw = 0
+asm_32xmemory = 0
 
 fpic :=
 
@@ -97,8 +99,8 @@ else ifeq ($(platform), osx)
 	endif
 	ifeq ($(arch),ppc)
 		CFLAGS += -DBLARGG_BIG_ENDIAN=1 -D__ppc__ -DFAMEC_NO_GOTOS
-        else
-        use_sh2drc = 1
+	else
+		use_sh2drc = 1
 	endif
 	OSXVER = `sw_vers -productVersion | cut -d. -f 2`
 	OSX_LT_MAVERICKS = `(( $(OSXVER) <= 9)) && echo "YES"`
@@ -120,14 +122,14 @@ else ifeq ($(platform), staticios)
 	CXX    += -miphoneos-version-min=8.0
 	CC_AS  += -miphoneos-version-min=8.0
 	CFLAGS += -miphoneos-version-min=8.0
-	ARCH := arm
+	ARCH := aarch64
 
 	STATIC_LINKING = 1
 	use_cyclone = 0
 	use_fame = 1
 	use_drz80 = 0
 	use_cz80 = 1
-	use_sh2drc = 0
+	use_sh2drc = 1
 	use_svpdrc = 0
 
 # iOS
@@ -142,7 +144,7 @@ else ifneq (,$(findstring ios,$(platform)))
 	ifeq ($(platform),ios-arm64)
 	  CC = clang -arch arm64 -isysroot $(IOSSDK)
 	  CXX = clang++ -arch arm64 -isysroot $(IOSSDK)
-      CFLAGS += -marm -DARM -D__aarch64__=1 
+	  CFLAGS += -marm -DARM -D__aarch64__=1 
 	else
 	  CC = clang -arch armv7 -isysroot $(IOSSDK)
 	  CXX = clang++ -arch armv7 -isysroot $(IOSSDK)
@@ -157,35 +159,32 @@ ifeq ($(platform),$(filter $(platform),ios9 ios-arm64))
 	CXX    += -miphoneos-version-min=8.0
 	CC_AS  += -miphoneos-version-min=8.0
 	CFLAGS += -miphoneos-version-min=8.0
+	ARCH := aarch64
 else
 	CC     += -miphoneos-version-min=5.0
 	CXX    += -miphoneos-version-min=5.0
 	CC_AS  += -miphoneos-version-min=5.0
 	CFLAGS += -miphoneos-version-min=5.0
-endif
 	ARCH := arm
+endif
 
 	use_cyclone = 0
 	use_fame = 1
 	use_drz80 = 0
 	use_cz80 = 1
-	ifeq ($(platform),ios-arm64)
-		use_sh2drc = 0
-		use_svpdrc = 0
-	else
-		use_sh2drc = 1
-		use_svpdrc = 1
-	endif
+	use_sh2drc = 1
+	use_svpdrc = 0
 
 # tvOS
 else ifeq ($(platform), tvos-arm64)
-	ARCH := arm
+	ARCH := aarch64
 	use_cyclone = 0
 	use_fame = 1
 	use_drz80 = 0
 	use_cz80 = 1
-	use_sh2drc = 0
+	use_sh2drc = 1
 	use_svpdrc = 0
+
 	TARGET := $(TARGET_NAME)_libretro_tvos.dylib
 	SHARED := -dynamiclib
 	fpic := -fPIC
@@ -194,6 +193,9 @@ else ifeq ($(platform), tvos-arm64)
 		IOSSDK := $(shell xcodebuild -version -sdk appletvos Path)
 	endif
 	CC_AS = perl ./tools/gas-preprocessor.pl $(CC)
+	CC = clang -arch arm64 -isysroot $(IOSSDK)
+	CXX = clang++ -arch arm64 -isysroot $(IOSSDK)
+	CFLAGS += -marm -DARM -D__aarch64__=1 
 	CFLAGS += -DIOS
 
 # PS3
@@ -206,17 +208,11 @@ else ifeq ($(platform), ps3)
 	NO_MMAP = 1
 	DONT_COMPILE_IN_ZLIB = 1
 
-	asm_memory = 0
-	asm_render = 0
-	asm_ym2612 = 0
-	asm_misc = 0
-	asm_cdpico = 0
-	asm_cdmemory = 0
-	asm_mix = 0
 	use_cyclone = 0
 	use_fame = 1
 	use_drz80 = 0
 	use_cz80 = 1
+	use_sh2drc = 1
 
 # sncps3
 else ifeq ($(platform), sncps3)
@@ -228,17 +224,11 @@ else ifeq ($(platform), sncps3)
 	NO_MMAP = 1
 	DONT_COMPILE_IN_ZLIB = 1
 
-	asm_memory = 0
-	asm_render = 0
-	asm_ym2612 = 0
-	asm_misc = 0
-	asm_cdpico = 0
-	asm_cdmemory = 0
-	asm_mix = 0
 	use_cyclone = 0
 	use_fame = 1
 	use_drz80 = 0
 	use_cz80 = 1
+	use_sh2drc = 1
 
 # Lightweight PS3 Homebrew SDK
 else ifeq ($(platform), psl1ght)
@@ -250,17 +240,11 @@ else ifeq ($(platform), psl1ght)
 	NO_MMAP = 1
 	DONT_COMPILE_IN_ZLIB = 1
 	
-	asm_memory = 0
-	asm_render = 0
-	asm_ym2612 = 0
-	asm_misc = 0
-	asm_cdpico = 0
-	asm_cdmemory = 0
-	asm_mix = 0
 	use_cyclone = 0
 	use_fame = 1
 	use_drz80 = 0
 	use_cz80 = 1
+	use_sh2drc = 1
 
 # PSP
 else ifeq ($(platform), psp1)
@@ -273,17 +257,11 @@ else ifeq ($(platform), psp1)
     NO_MMAP = 1
     DONT_COMPILE_IN_ZLIB = 1
 
-    asm_memory = 0
     asm_render = 1
-    asm_ym2612 = 0
-    asm_misc = 0
-    asm_cdpico = 0
-    asm_cdmemory = 0
-    asm_mix = 0
-    use_cyclone = 0
     use_fame = 1
     use_drz80 = 0
     use_cz80 = 1
+    use_sh2drc = 1
 
 # PS2
 else ifeq ($(platform), ps2)
@@ -297,17 +275,12 @@ else ifeq ($(platform), ps2)
     NO_MMAP = 1
     DONT_COMPILE_IN_ZLIB = 1
 
-    asm_memory = 0
     asm_render = 1
-    asm_ym2612 = 0
-    asm_misc = 0
-    asm_cdpico = 0
-    asm_cdmemory = 0
-    asm_mix = 0
     use_cyclone = 0
     use_fame = 1
     use_drz80 = 0
     use_cz80 = 1
+    use_sh2drc = 1
 
 	OBJS += platform/ps2/asm.o
 
@@ -327,14 +300,6 @@ else ifeq ($(platform), ctr)
    ARCH = arm
    ARM_ASM = 1
 
-   asm_memory = 1
-   asm_render = 1
-   asm_ym2612 = 1
-   asm_misc = 1
-   asm_cdpico = 1
-   asm_cdmemory = 1
-   asm_mix = 1
-
    use_cyclone = 1
    use_fame = 0
    use_drz80 = 1
@@ -357,14 +322,6 @@ else ifeq ($(platform), raspberrypi)
    fpic := -fPIC
    DONT_COMPILE_IN_ZLIB = 1
 
-   asm_memory = 1
-   asm_render = 1
-   asm_ym2612 = 1
-   asm_misc = 1
-   asm_cdpico = 1
-   asm_cdmemory = 1
-   asm_mix = 1
-
    use_cyclone = 1
    use_fame = 0
    use_drz80 = 1
@@ -386,14 +343,8 @@ else ifeq ($(platform), vita)
 	NO_MMAP = 1
 	DONT_COMPILE_IN_ZLIB = 1
 	ARCH = arm
+	ARM_ASM=1
 
-	asm_memory = 1
-	asm_render = 1
-	asm_ym2612 = 1
-	asm_misc = 1
-	asm_cdpico = 1
-	asm_cdmemory = 1
-	asm_mix = 1
 	use_cyclone = 1
 	use_fame = 0
 	use_drz80 = 1
@@ -443,13 +394,14 @@ else ifeq ($(platform), switch)
 else ifeq ($(platform), libnx)
    include $(DEVKITPRO)/libnx/switch_rules
    TARGET := $(TARGET_NAME)_libretro_$(platform).a
-   ARCH := arm64
+   ARCH := aarch64
    CFLAGS += -O3 -fomit-frame-pointer -ffast-math -I$(DEVKITPRO)/libnx/include/ -fPIE -Wl,--allow-multiple-definition
    CFLAGS += -specs=$(DEVKITPRO)/libnx/switch.specs
    CFLAGS += -D__SWITCH__ -DHAVE_LIBNX
    CFLAGS += -DARM -D__aarch64__=1 -march=armv8-a -mtune=cortex-a57 -mtp=soft -ffast-math -mcpu=cortex-a57+crc+fp+simd -ffunction-sections
    CFLAGS += -Ifrontend/switch -ftree-vectorize
    STATIC_LINKING=1
+   use_sh2drc = 1
 
 # QNX
 else ifeq ($(platform), qnx)
@@ -489,7 +441,6 @@ else ifeq ($(platform), classic_armv7_a7)
 	HAVE_NEON = 1
 	ARCH = arm
 	BUILTIN_GPU = neon
-	USE_DYNAREC = 1
 	ifeq ($(shell echo `$(CC) -dumpversion` "< 4.9" | bc -l), 1)
 	  CFLAGS += -march=armv7-a
 	else
@@ -499,6 +450,9 @@ else ifeq ($(platform), classic_armv7_a7)
 	    LDFLAGS += -static-libgcc -static-libstdc++
 	  endif
 	endif
+
+	use_sh2drc = 1
+	use_svpdrc = 1
 	
 # (armv8 a35, hard point, neon based) ###
 # Playstation Classic
@@ -518,19 +472,14 @@ else ifeq ($(platform), classic_armv8_a35)
         CPPFLAGS += $(CFLAGS)
         ASFLAGS += $(CFLAGS)
         HAVE_NEON = 1
-        ARCH = arm
+        ARCH = aarch64
         BUILTIN_GPU = neon
         CFLAGS += -march=armv8-a
         LDFLAGS += -static-libgcc -static-libstdc++
-        asm_memory = 0
-        asm_render = 0
-        asm_ym2612 = 0
-        asm_misc = 0
-        asm_cdpico = 0
-        asm_cdmemory = 0
-        asm_mix = 0
+
+        use_sh2drc = 1
         use_cyclone = 0
-	use_fame = 1
+        use_fame = 1
         use_drz80 = 0
         use_cz80 = 1	
 #######################################
@@ -597,21 +546,14 @@ else ifeq ($(platform), emscripten)
 # GCW0
 else ifeq ($(platform), gcw0)
 	TARGET := $(TARGET_NAME)_libretro.so
-	CC = /opt/gcw0-toolchain/usr/bin/mipsel-linux-gcc
-	AR = /opt/gcw0-toolchain/usr/bin/mipsel-linux-ar
+	CC = mipsel-linux-gcc
+	AR = mipsel-linux-ar
 	SHARED := -shared -nostdlib
         fpic := -fPIC
 	LIBM :=
 	DONT_COMPILE_IN_ZLIB = 1
 	CFLAGS += -ffast-math -march=mips32 -mtune=mips32r2 -mhard-float
 
-	asm_memory = 0
-	asm_render = 0
-	asm_ym2612 = 0
-	asm_misc = 0
-	asm_cdpico = 0
-	asm_cdmemory = 0
-	asm_mix = 0
 	use_cyclone = 0
 	use_fame = 1
 	use_drz80 = 0
@@ -744,6 +686,8 @@ asm_ym2612 = 1
 asm_misc = 1
 asm_cdmemory = 1
 asm_mix = 1
+asm_32xdraw = 1
+asm_32xmemory = 1
 endif
 
 CFLAGS += $(fpic)

cpu/drc/emit_x86.c

@@ -733,7 +733,7 @@ enum { xAX = 0, xCX, xDX, xBX, xSP, xBP, xSI, xDI,	// x86-64,i386 common
 	/* mov r <-> [ebp+#offs] */ \
 	if ((offs) == 0) { \
 		emith_deref_modrm(op, 0, r, rs); \
-	} else if (abs(offs) >= 0x80) { \
+	} else if ((s32)(offs) < -0x80 || (s32)(offs) >= 0x80) { \
 		emith_deref_modrm(op, 2, r, rs); \
 		EMIT(offs, u32); \
 	} else { \
@@ -1075,7 +1075,7 @@ enum { xAX = 0, xCX, xDX, xBX, xSP, xBP, xSI, xDI,	// x86-64,i386 common
 #define PARAM_REGS	{ xCX, xDX, xR8, xR9 }
 #define	PRESERVED_REGS	{ xSI, xDI, xR12, xR13, xR14, xR15, xBX, xBP }
 #define TEMPORARY_REGS	{ xAX, xR10, xR11 }
-#define STATIC_SH2_REGS { SHR_SR,xBX , SHR_R(0),xR15 , SH2_R(1),xR14 }
+#define STATIC_SH2_REGS { SHR_SR,xBX , SHR_R(0),xR15 , SHR_R(1),xR14 }
 
 #define host_arg2reg(rd, arg) \
 	switch (arg) { \

cpu/sh2/compiler.c

@@ -1273,24 +1273,23 @@ static void dr_flush_tcache(int tcid)
   blink_free[tcid] = NULL;
   memset(unresolved_links[tcid], 0, sizeof(*unresolved_links[0]) * HASH_TABLE_SIZE(tcid));
   memset(hash_tables[tcid], 0, sizeof(*hash_tables[0]) * HASH_TABLE_SIZE(tcid));
-  if (Pico32xMem->sdram != NULL) {
-    if (tcid == 0) { // ROM, RAM
-      memset(Pico32xMem->drcblk_ram, 0, sizeof(Pico32xMem->drcblk_ram));
-      memset(Pico32xMem->drclit_ram, 0, sizeof(Pico32xMem->drclit_ram));
-      memset(sh2s[0].branch_cache, -1, sizeof(sh2s[0].branch_cache));
-      memset(sh2s[1].branch_cache, -1, sizeof(sh2s[1].branch_cache));
-      memset(sh2s[0].rts_cache, -1, sizeof(sh2s[0].rts_cache));
-      memset(sh2s[1].rts_cache, -1, sizeof(sh2s[1].rts_cache));
-      sh2s[0].rts_cache_idx = sh2s[1].rts_cache_idx = 0;
-    } else {
-      memset(Pico32xMem->drcblk_ram, 0, sizeof(Pico32xMem->drcblk_ram));
-      memset(Pico32xMem->drclit_ram, 0, sizeof(Pico32xMem->drclit_ram));
-      memset(Pico32xMem->drcblk_da[tcid - 1], 0, sizeof(Pico32xMem->drcblk_da[tcid - 1]));
-      memset(Pico32xMem->drclit_da[tcid - 1], 0, sizeof(Pico32xMem->drclit_da[tcid - 1]));
-      memset(sh2s[tcid - 1].branch_cache, -1, sizeof(sh2s[0].branch_cache));
-      memset(sh2s[tcid - 1].rts_cache, -1, sizeof(sh2s[0].rts_cache));
-      sh2s[tcid - 1].rts_cache_idx = 0;
-    }
+
+  if (tcid == 0) { // ROM, RAM
+    memset(Pico32xMem->drcblk_ram, 0, sizeof(Pico32xMem->drcblk_ram));
+    memset(Pico32xMem->drclit_ram, 0, sizeof(Pico32xMem->drclit_ram));
+    memset(sh2s[0].branch_cache, -1, sizeof(sh2s[0].branch_cache));
+    memset(sh2s[1].branch_cache, -1, sizeof(sh2s[1].branch_cache));
+    memset(sh2s[0].rts_cache, -1, sizeof(sh2s[0].rts_cache));
+    memset(sh2s[1].rts_cache, -1, sizeof(sh2s[1].rts_cache));
+    sh2s[0].rts_cache_idx = sh2s[1].rts_cache_idx = 0;
+  } else {
+    memset(Pico32xMem->drcblk_ram, 0, sizeof(Pico32xMem->drcblk_ram));
+    memset(Pico32xMem->drclit_ram, 0, sizeof(Pico32xMem->drclit_ram));
+    memset(Pico32xMem->drcblk_da[tcid - 1], 0, sizeof(Pico32xMem->drcblk_da[tcid - 1]));
+    memset(Pico32xMem->drclit_da[tcid - 1], 0, sizeof(Pico32xMem->drclit_da[tcid - 1]));
+    memset(sh2s[tcid - 1].branch_cache, -1, sizeof(sh2s[0].branch_cache));
+    memset(sh2s[tcid - 1].rts_cache, -1, sizeof(sh2s[0].rts_cache));
+    sh2s[tcid - 1].rts_cache_idx = 0;
   }
 #if (DRC_DEBUG & 4)
   tcache_dsm_ptrs[tcid] = tcache_ring[tcid].base;

cpu/sh2/compiler.h

@@ -31,7 +31,7 @@ unsigned short scan_block(uint32_t base_pc, int is_slave,
 		unsigned char *op_flags, uint32_t *end_pc,
 		uint32_t *base_literals, uint32_t *end_literals);
 
-#if defined(DRC_SH2) && defined(__GNUC__)
+#if defined(DRC_SH2) && defined(__GNUC__) && !defined(__clang__)
 // direct access to some host CPU registers used by the DRC if gcc is used.
 // XXX MUST match SHR_SR definitions in cpu/drc/emit_*.c; should be moved there
 // XXX yuck, there's no portable way to determine register size. Use long long

cpu/sh2/sh2.h

@@ -106,7 +106,11 @@ int  sh2_execute_interpreter(SH2 *sh2c, int cycles);
 
 static __inline void sh2_execute_prepare(SH2 *sh2, int use_drc)
 {
+#ifdef DRC_SH2
   sh2->run = use_drc ? sh2_execute_drc : sh2_execute_interpreter;
+#else
+  sh2->run = sh2_execute_interpreter;
+#endif
 }
 
 static __inline int sh2_execute(SH2 *sh2, int cycles)

tools/mkoffsets.sh

@@ -8,6 +8,20 @@ CC=${CC:-gcc}
 # endianess of target (automagically determined below)
 ENDIAN=
 
+# don't do this if ELF format isn't used. it doesn't matter since offsets are
+# only needed for the asm parts (currently mips/arm32) and those have ELF
+check_elf ()
+{
+	echo '#include <stdint.h>' >/tmp/getoffs.c
+	echo "const int32_t val = 1;" >>/tmp/getoffs.c
+	$CC $CFLAGS -I .. -c /tmp/getoffs.c -o /tmp/getoffs.o || exit 1
+	if ! command -v readelf >/dev/null || ! file /tmp/getoffs.o | grep -q ELF; then
+		echo "/* mkoffset.sh: no readelf or not ELF, offset table not created */" >$fn
+		echo "WARNING: no readelf or not ELF, offset table not created"
+		exit
+	fi
+}
+
 # compile with target C compiler and extract value from .rodata section
 compile_rodata ()
 {
@@ -49,13 +63,7 @@ get_define () # prefix struct member member...
 fn="${1:-.}/pico_int_offs.h"
 if echo $CFLAGS | grep -qe -flto; then CFLAGS="$CFLAGS -fno-lto"; fi
 
-# don't do this if readelf isn't available. it doesn't matter since offsets are
-# only needed for the asm parts (currently mips/arm32) and those have readelf
-if ! command -v readelf >/dev/null; then
-	echo "/* mkoffset.sh: readelf not found, offset table not created */" >$fn
-	echo "WARNING: readelf not found, offset table not created"
-	exit
-fi
+check_elf
 # determine endianess
 echo '#include <stdint.h>' >/tmp/getoffs.c
 echo "const int32_t val = 1;" >>/tmp/getoffs.c