configuration changes and README

Makefile

@@ -238,8 +238,8 @@ pico/cd/LC89510.o: CFLAGS += -fno-strict-aliasing
 pico/cd/gfx_cd.o: CFLAGS += -fno-strict-aliasing
 ifeq (1,$(use_sh2drc))
 ifneq (,$(findstring -flto,$(CFLAGS)))
-# if using the DRC, memory and sh2soc use a global register variable to avoid
+# if using the DRC, memory and sh2soc directly use the DRC register for SH2 SR
-# saving and reloading the SH2 SR. However, this collides with the use of LTO.
+# to avoid saving and reloading it. However, this collides with the use of LTO.
 pico/32x/memory.o: CFLAGS += -fno-lto
 pico/32x/sh2soc.o: CFLAGS += -fno-lto
 endif

README.md

@@ -0,0 +1,112 @@
+This is my foray into dynamic recompilation using PicoDrive, a
+Megadrive / Genesis / Sega CD / Mega CD / 32X / SMS emulator.
+
+I added support for MIPS (mips32r1) and ARM64 (aarch64) to the recompiler, as
+well as spent much effort to optimize the code generated by the DRC.
+I also optimized SH2 memory access inside the emulator, and did some work on
+M68K/SH2 CPU synchronization to fix some problems and speed up the emulator.
+
+It got a bit out of hand. I ended up doing fixes and optimzations all over the
+place, mainly for 32X and CD, 32X graphics handling, and probably some more,
+see the commit history.
+
+### compiling
+
+I mainly worked with standalone PicoDrive versions as created by configure/make.
+A list of platforms for which this is possible can be obtained with
+
+> configure --help
+
+If you want to build an executable for a unixoid platform not listed in the
+platform list, just use
+
+> configure --platform=generic
+
+If DRC is available for the platform, it should be enabled automatically.
+
+For other platforms using a cross-compiling toolchain I used this,
+assuming $TC points to the appropriate cross compile toolchain directory:
+
+platform|toolchain|configure command
+--------|---------|-----------------
+gp2x,wiz,caanoo|open2x|CROSS_COMPILE=arm-open2x-linux- CFLAGS="-I$TC/gcc-4.1.1-glibc-2.3.6/arm-open2x-linux/include -fno-stack-protector -fno-common -finline-limit=42" LDFLAGS="--sysroot $TC/gcc-4.1.1-glibc-2.3.6/arm-open2x-linux -L$TC/gcc-4.1.1-glibc-2.3.6/arm-open2x-linux/lib" ./configure --platform=gp2x
+gp2x,wiz,caanoo|open2x with ubuntu arm gcc 4.7|CROSS_COMPILE=arm-linux-gnueabi- CFLAGS="-I$TC/gcc-4.1.1-glibc-2.3.6/arm-open2x-linux/include -fno-stack-protector -fno-common -finline-limit=42 -fipa-pta" LDFLAGS="-B$TC/gcc-4.1.1-glibc-2.3.6/lib/gcc/arm-open2x-linux/4.1.1 -B$TC/gcc-4.1.1-glibc-2.3.6/arm-open2x-linux/lib -L$TC/gcc-4.1.1-glibc-2.3.6/arm-open2x-linux/lib" ./configure --platform=gp2x
+opendingux|opendingux|CROSS_COMPILE=mipsel-linux- CFLAGS="-I$TC/usr/include -I$TC/usr/include/SDL -fno-stack-protector -fno-common -finline-limit=42 -fipa-pta" LDFLAGS="--sysroot $TC -L$TC/lib" ./configure --platform=opendingux
+opendingux|opendingux with ubuntu mips gcc 5.4|CROSS_COMPILE=mipsel-linux-gnu- CFLAGS="-I$TC/usr/include -I$TC/usr/include/SDL -fno-stack-protector -fno-common -finline-limit=42 -fipa-pta" LDFLAGS="-B$TC/usr/lib -B$TC/lib -Wl,-rpath-link=$TC/usr/lib -Wl,-rpath-link=$TC/lib" ./configure --platform=opendingux
+gcw0|gcw0|CROSS_COMPILE=mipsel-gcw0-linux-uclibc- CFLAGS="-I$TC/usr/mipsel-gcw0-linux-uclibc/sysroot/usr/include -I$TC/usr/mipsel-gcw0-linux-uclibc/sysroot/usr/include/SDL -fno-stack-protector -fno-common -finline-limit=42 -fipa-pta" LDFLAGS="--sysroot $TC/usr/mipsel-gcw0-linux-uclibc/sysroot" ./configure --platform=gcw0
+
+For gp2x, wiz, and caanoo you may need to compile libpng first, and additionally
+this patch may need to be applied to the cpu/cyclone submodule:
+> diff --git a/OpArith.cpp b/OpArith.cpp
+> index 96c7e0d..09517b8 100644
+> --- a/OpArith.cpp
+> +++ b/OpArith.cpp
+> @@ -425,7 +425,7 @@ int OpAbcd(int op)
+>      ot("  add r1,r1,r0\n");
+>      ot("  add r1,r1,r6\n");
+>      ot("  mov r12,r1\n");
+> -    ot("  addhi r12,#6 ;@ Decimal adjust units\n");
+> +    ot("  addhi r12,r12,#6 ;@ Decimal adjust units\n");
+>      ot("  tst r1,#0x80\n");
+>      ot("  orreq r10,r10,#0x10000000 ;@ Undefined V behavior\n");
+>      ot("  cmp r12,#0x9f\n");
+> @@ -452,7 +452,7 @@ int OpAbcd(int op)
+>      ot("  cmp r1,r12\n");
+>      ot("  orrlt r10,r10,#0x20000000 ;@ C\n");
+>      ot("  cmp r1,#0xff\n");
+> -    ot("  addhi r1,#0xa0\n");
+> +    ot("  addhi r1,r1,#0xa0\n");
+>      ot("  sub r12,r1,r12\n");
+>      ot("  movs r0,r12,lsl #24\n");
+>      ot("  bicmi r10,r10,#0x10000000 ;@ Undefined V behavior part II\n");
+> diff --git a/OpLogic.cpp b/OpLogic.cpp
+> index 012e35a..d40d814 100644
+> --- a/OpLogic.cpp
+> +++ b/OpLogic.cpp
+> @@ -74,12 +74,12 @@ const char *TestCond(int m68k_cc, int invert)
+>        break;
+>      case 0x0e: // gt
+>        ot("  eor r0,r10,r10,lsl #3 ;@ gt: !Z && N == V\n");
+> -      ot("  orrs r0,r10,lsl #1\n");
+> +      ot("  orrs r0,r0,r10,lsl #1\n");
+>        cond="pl", icond="mi";
+>        break;
+>      case 0x0f: // le
+>        ot("  eor r0,r10,r10,lsl #3 ;@ le: Z || N != V\n");
+> -      ot("  orrs r0,r10,lsl #1\n");
+> +      ot("  orrs r0,r0,r10,lsl #1\n");
+>        cond="mi", icond="pl";
+>        break;
+>      default:
+
+After configure, compile with
+
+> make opk # for opendingux and gcw0
+> 
+> make # for anything else
+
+### helix MP3 decoder
+
+For 32 bit ARM platforms, there is the possibility to compile the helix MP3
+decoder into a shared library to be able to use MP3 audio files with CD games.
+The helix source files aren't supplied because of licensing issues. However, if
+you have obtained the sources, put them into the platform/common/helix
+directory, set CROSS to your cross compiler prefix (e.g. arm-linux-gnueabi-)
+and LIBGCC to your cross compiler's libgcc.a
+(e.g. /usr/lib/gcc-cross/arm-linux-gnueabi/4.7/libgcc.a), and compile with
+
+> make -C platform/common/helix CROSS=$CROSS LIBGCC=$LIBGCC
+
+Copy the resulting ${CROSS}helix_mp3.so as libhelix.so to the directory where
+the PicoDrive binary is.
+
+### installing
+
+You need to install the resulting binary onto your device manually.
+For opendingux and gcw0, copy the opk to your SD card.
+For gp2x, wiz and caanoo, the easiest way is to unpack
+[PicoDrive_191.zip](http://notaz.gp2x.de/releases/PicoDrive/PicoDrive_191.zip)
+on you SD card and replace the PicoDrive binary.
+
+Send bug reports, fixes etc to <derkub@gmail.com>
+Kai-Uwe Bloem

config.dingux

@@ -1,5 +1,5 @@
 # Automatically generated by configure
-# Configured with: './configure' '--platform=generic'
+# Configured with: './configure' '--platform=opendingux'
 CC = mipsel-linux-gcc
 CXX = mipsel-linux-g++
 AS = mipsel-linux-as

config.dingux54

@@ -1,5 +1,5 @@
 # Automatically generated by configure
-# Configured with: './configure' '--platform=generic'
+# Configured with: './configure' '--platform=opendingux'
 CC = mipsel-linux-gnu-gcc
 CXX = mipsel-linux-gnu-g++
 AS = mipsel-linux-gnu-as

config.gcw0

@@ -1,5 +1,5 @@
 # Automatically generated by configure
-# Configured with: './configure' '--platform=generic'
+# Configured with: './configure' '--platform=gcw0'
 CC = mipsel-gcw0-linux-uclibc-gcc
 CXX = mipsel-gcw0-linux-uclibc-g++
 AS = mipsel-gcw0-linux-uclibc-as

config.gp2x

@@ -4,8 +4,8 @@ CC = arm-open2x-linux-gcc
 CXX = arm-open2x-linux-g++
 AS = arm-open2x-linux-as
 STRIP = arm-open2x-linux-strip
-CFLAGS += -I${HOME}/opt/open2x/gcc-4.1.1-glibc-2.3.6/arm-open2x-linux/include -I${HOME}/src/gp2x/armroot/include  -D__GP2X__ -Wno-unused-result
+CFLAGS += -I${HOME}/opt/open2x/gcc-4.1.1-glibc-2.3.6/arm-open2x-linux/include -I${HOME}/src/gp2x/armroot/include  -D__GP2X__
-CFLAGS += -msoft-float -mcpu=arm920t -mtune=arm920t
+CFLAGS += -mfloat-abi=soft -mcpu=arm920t -mtune=arm920t -mno-thumb-interwork -fno-stack-protector -fno-common
 CFLAGS += -finline-limit=42 -fipa-cp -fno-ipa-pure-const
 ASFLAGS += -mcpu=arm920t -mfloat-abi=soft
 LDFLAGS += --sysroot ${HOME}/opt/open2x/gcc-4.1.1-glibc-2.3.6/arm-open2x-linux -L${HOME}/opt/open2x/gcc-4.1.1-glibc-2.3.6/arm-open2x-linux/lib -L${HOME}/src/gp2x/armroot/lib -static

config.gp2x47

@@ -5,8 +5,7 @@ CXX = arm-linux-gnueabi-g++
 AS = arm-linux-gnueabi-as
 STRIP = arm-linux-gnueabi-strip
 CFLAGS += -I${HOME}/opt/open2x/gcc-4.1.1-glibc-2.3.6/arm-open2x-linux/include -I${HOME}/src/gp2x/armroot/include -D__GP2X__ -Wno-unused-result
-CFLAGS += -mabi=apcs-gnu -mfloat-abi=soft -mfpu=fpa -mcpu=arm920t -mtune=arm920t
+CFLAGS += -mabi=apcs-gnu -mfloat-abi=soft -mfpu=fpa -mcpu=arm920t -mtune=arm920t -mno-thumb-interwork -fno-stack-protector -fno-common
-CFLAGS += -mno-thumb-interwork -fno-stack-protector -fno-common
 CFLAGS += -finline-limit=42 -fipa-pta -fno-ipa-sra -fno-ipa-pure-const
 ASFLAGS += -mabi=apcs-gnu -mfloat-abi=soft -mfpu=fpa -mcpu=arm920t
 LDFLAGS += -mabi=apcs-gnu -mfpu=fpa -B${HOME}/opt/open2x/gcc-4.1.1-glibc-2.3.6/lib/gcc/arm-open2x-linux/4.1.1 -B${HOME}/opt/open2x/gcc-4.1.1-glibc-2.3.6/arm-open2x-linux/lib -L${HOME}/opt/open2x/gcc-4.1.1-glibc-2.3.6/arm-open2x-linux/usr/lib -L${HOME}/src/gp2x/armroot/lib -static

configure

@@ -22,6 +22,13 @@ compile_binary()
   $c >> config.log 2>&1
 }
 
+check_option()
+{
+  echo 'void test(void) { }' >$TMPC
+  compile_object $1 || return 1
+  return 0
+}
+
 check_define()
 {
   $CC -E -dD $CFLAGS pico/arm_features.h | grep -q $1 || return 1
@@ -31,17 +38,18 @@ check_define()
 # setting options to "yes" or "no" will make that choice default,
 # "" means "autodetect".
 
-platform_list="generic pandora gp2x opendingux rpi1 rpi2"
+platform_list="generic pandora gp2x wiz caanoo opendingux gcw0 rpi1 rpi2"
 platform="generic"
 sound_driver_list="oss alsa sdl"
 sound_drivers=""
 have_armv5=""
 have_armv6=""
 have_armv7=""
+have_arm_oabi=""
 have_arm_neon=""
 have_libavcodec=""
 need_sdl="no"
-need_xlib="no"
+need_zlib="no"
 # these are for known platforms
 optimize_cortexa8="no"
 optimize_cortexa7="no"
@@ -54,7 +62,7 @@ CC="${CC-${CROSS_COMPILE}gcc}"
 CXX="${CXX-${CROSS_COMPILE}g++}"
 AS="${AS-${CROSS_COMPILE}as}"
 STRIP="${STRIP-${CROSS_COMPILE}strip}"
-test -n "$SDL_CONFIG" || SDL_CONFIG="`$CC --print-sysroot 2> /dev/null || true`/usr/bin/sdl-config"
+test -n "$SDL_CONFIG" || SDL_CONFIG="`$CC $CFLAGS $LDFLAGS --print-sysroot 2> /dev/null || true`/usr/bin/sdl-config"
 MAIN_LDLIBS="$LDLIBS -lm"
 config_mak="config.mak"
 
@@ -78,23 +86,27 @@ set_platform()
     ;;
   generic)
     ;;
-  opendingux)
+  opendingux | gcw0)
     sound_drivers="sdl"
+    # both are really an opendingux
+    platform="opendingux"
     ;;
   pandora)
     sound_drivers="oss alsa"
     optimize_cortexa8="yes"
     have_arm_neon="yes"
     ;;
-  gp2x)
+  gp2x | wiz | caanoo)
     sound_drivers="oss"
     optimize_arm920="yes"
+    # compile for OABI if toolchain provides it (faster code on caanoo)
+    have_arm_oabi="yes"
+    # always use static linking, since caanoo doesn't have OABI libs. Moreover,
+    # dynamic linking slows Wiz 1-10%, and libm on F100 isn't compatible
+    LDFLAGS="$LDFLAGS -static"
+    # unified binary for all of them
     CFLAGS="$CFLAGS -D__GP2X__"
-    if [ "$CROSS_COMPILE" = "arm-linux-" ]; then
+    platform="gp2x"
-      # still using static, dynamic linking slows Wiz 1-10%
-      # also libm on F100 is not compatible
-      MAIN_LDLIBS="$MAIN_LDLIBS -static"
-    fi
     ;;
   *)
     fail "unsupported platform: $platform"
@@ -147,18 +159,11 @@ fi
 #  fi
 #fi
 
-# basic compiler test
-cat > $TMPC <<EOF
-int main(void) { return 0; }
-EOF
-if ! compile_binary; then
-  fail "compiler test failed, please check config.log"
-fi
-
 if [ -z "$ARCH" ]; then
   ARCH=`$CC -dumpmachine | awk -F '-' '{print $1}'`
 fi
 
+# CPU/ABI stuff first, else compile test may fail
 case "$ARCH" in
 arm*)
   # ARM stuff
@@ -206,26 +211,40 @@ arm*)
     have_armv5=`check_define HAVE_ARMV5 && echo yes` || true
   fi
 
-  # automatically set mfpu and mfloat-abi if they are not set
-  if [ "$have_arm_neon" = "yes" ]; then
-    fpu="neon"
-  elif [ "$have_armv6" = "yes" ]; then
-    fpu="vfp"
-  fi
-  if [ "x$fpu" != "x" ]; then
-    echo "$CFLAGS" | grep -q -- '-mfpu=' || CFLAGS="$CFLAGS -mfpu=$fpu"
-    echo "$ASFLAGS" | grep -q -- '-mfpu=' || ASFLAGS="$ASFLAGS -mfpu=$fpu"
-    floatabi_set_by_gcc=`$CC -v 2>&1 | grep -q -- --with-float= && echo yes` || true
-    if [ "$floatabi_set_by_gcc" != "yes" ]; then
-      echo "$CFLAGS" | grep -q -- '-mfloat-abi=' || CFLAGS="$CFLAGS -mfloat-abi=softfp"
-      echo "$ASFLAGS" | grep -q -- '-mfloat-abi=' || ASFLAGS="$ASFLAGS -mfloat-abi=softfp"
-    fi
-  fi
-
   # must disable thumb as recompiler can't handle it
   if check_define __thumb__; then
     CFLAGS="$CFLAGS -marm"
   fi
+  # OABI/EABI selection
+  if [ "$have_arm_oabi" = "yes" ] &&  check_option -mabi=apcs-gnu; then
+    echo "$CFLAGS" | grep -q -- '-mabi=' || CFLAGS="$CFLAGS -mabi=apcs-gnu"
+    echo "$CFLAGS" | grep -q -- '-m\(no-\)*thumb-interwork' || CFLAGS="$CFLAGS -mno-thumb-interwork"
+    echo "$ASFLAGS" | grep -q -- '-mabi=' || ASFLAGS="$ASFLAGS -mabi=apcs-gnu"
+  fi
+
+  # automatically set mfpu and mfloat-abi if they are not set
+  if [ "$have_arm_neon" = "yes" ]; then
+    fpu="neon"
+    abi="hard"
+  elif [ "$have_armv6" = "yes" ]; then
+    fpu="vfp"
+    abi="softfp"
+  elif check_option -mfpu=fpa; then
+    fpu="fpa" # compatibility option for arm-linux-gnueabi-gcc on ubuntu
+    abi="soft"
+  fi
+  if [ "x$fpu" != "x" ]; then
+    echo "$CFLAGS" | grep -q -- '-mfpu=' || CFLAGS="$CFLAGS -mfpu=$fpu"
+    echo "$ASFLAGS" | grep -q -- '-mfpu=' || ASFLAGS="$ASFLAGS -mfpu=$fpu"
+    echo "$CFLAGS" | grep -q -- '-mfloat-abi=' || CFLAGS="$CFLAGS -mfloat-abi=$abi"
+    echo "$ASFLAGS" | grep -q -- '-mfloat-abi=' || ASFLAGS="$ASFLAGS -mfloat-abi=$abi"
+  fi
+
+  # add -ldl for helix support
+  case "$MAIN_LDLIBS" in
+    *"-ldl"*) ;;
+    *) MAIN_LDLIBS="-ldl $MAIN_LDLIBS" ;;
+  esac
 
   # warn about common mistakes
   if [ "$platform" != "gp2x" -a "$have_armv5" != "yes" ]; then
@@ -251,6 +270,14 @@ rpi1 | rpi2 | generic | opendingux)
   ;;
 esac
 
+# basic compiler test
+cat > $TMPC <<EOF
+int main(void) { return 0; }
+EOF
+if ! compile_binary; then
+  fail "compiler test failed, please check config.log"
+fi
+
 # header/library presence tests
 check_zlib()
 {
@@ -308,8 +335,7 @@ EOF
   compile_object "$@"
 }
 
-MAIN_LDLIBS="$MAIN_LDLIBS -lz"
+check_zlib -lz &&MAIN_LDLIBS="$MAIN_LDLIBS -lz" || need_zlib="yes"
-check_zlib -lz || fail "please install zlib (libz-dev)"
 
 MAIN_LDLIBS="-lpng $MAIN_LDLIBS"
 check_libpng || fail "please install libpng (libpng-dev)"
@@ -352,10 +378,7 @@ if [ "$need_sdl" = "yes" ]; then
   check_sdl `$SDL_CONFIG --libs` || fail "please install libsdl (libsdl1.2-dev)"
 fi
 
-cat > $TMPC <<EOF
+if check_option -Wno-unused_result; then
-void test(void *f, void *d) { fread(d, 1, 1, f); }
-EOF
-if compile_object -Wno-unused-result; then
   CFLAGS="$CFLAGS -Wno-unused-result"
 fi
 
@@ -395,15 +418,18 @@ echo "SOUND_DRIVERS = $sound_drivers" >> $config_mak
 if [ "$have_libavcodec" = "yes" ]; then
   echo "HAVE_LIBAVCODEC = 1" >> $config_mak
 fi
+if [ "$need_zlib" = "yes" ]; then
+  echo "PLATFORM_ZLIB = 1" >> $config_mak
+fi
 
 # GP2X toolchains are too old for UAL asm,
 # so add this here to not litter main Makefile
-if [ "$platform" = "g1p2x" ]; then
+#if [ "$platform" = "gp2x" ]; then
-  echo >> $config_mak
+#  echo >> $config_mak
-  echo "%.o: %.S" >> $config_mak
+#  echo '%.o: %.S' >> $config_mak
-  echo "        $(CC) $(CFLAGS) -E -c $^ -o /tmp/$(notdir $@).s" >> $config_mak
+#  echo '	$(CC) $(CFLAGS) -E -c $^ -o /tmp/$(notdir $@).s' >> $config_mak
-  echo "        $(AS) $(ASFLAGS) /tmp/$(notdir $@).s -o $@" >> $config_mak
+#  echo '	$(AS) $(ASFLAGS) /tmp/$(notdir $@).s -o $@' >> $config_mak
-fi
+#fi
 
 # use pandora's skin (for now)
 test -e skin || ln -s platform/pandora/skin skin

cpu/drc/emit_arm.c

@@ -1,6 +1,7 @@
 /*
  * Basic macros to emit ARM instructions and some utils
  * Copyright (C) 2008,2009,2010 notaz
+ * Copyright (C) 2019 kub
  *
  * This work is licensed under the terms of MAME license.
  * See COPYING file in the top-level directory.

cpu/drc/emit_x86.c

@@ -1,6 +1,7 @@
 /*
  * Basic macros to emit x86 instructions and some utils
  * Copyright (C) 2008,2009,2010 notaz
+ * Copyright (C) 2019 kuv
  *
  * This work is licensed under the terms of MAME license.
  * See COPYING file in the top-level directory.

platform/common/helix/Makefile

@@ -4,6 +4,7 @@ CC = $(CROSS)gcc
 AS = $(CROSS)as
 AR = $(CROSS)ar
 TOOLCHAIN = $(notdir $(CROSS))
+LIBGCC ?= ${HOME}/opt/open2x/gcc-4.1.1-glibc-2.3.6/lib/gcc/arm-open2x-linux/4.1.1/libgcc.a
 
 CFLAGS += -Ipub -O2 -Wall -fstrict-aliasing -ffast-math
 ifneq ($(findstring arm-,$(TOOLCHAIN)),)
@@ -34,7 +35,7 @@ real/arm/asmpoly_gcc.o: real/arm/asmpoly_gcc.s
 
 $(LIB) : $(OBJS)
 	$(AR) r $@ $^
-$(SHLIB) : $(OBJS) /home/build/opt/open2x/gcc-4.1.1-glibc-2.3.6/lib/gcc/arm-open2x-linux/4.1.1/libgcc.a
+$(SHLIB) : $(OBJS) $(LIBGCC)
 	$(CC) -o $@ -nostdlib -shared $(CFLAGS) $^
 
 clean: