Merge remote-tracking branch 'irixxxx/master' into master

.github/workflows/ci.yml

@@ -239,6 +239,7 @@ jobs:
       run: |
         git config --global --add safe.directory $PWD
         ver=$(cut -d'"' -f2 platform/common/version.h)-$(git rev-parse --short HEAD)
+        # toolchain has EGL, but retrofw 2.2 has not, so disable it
         ./configure --platform=retrofw
         make -j$(getconf _NPROCESSORS_ONLN)
         mv PicoDrive.opk PicoDrive-retrofw-$ver.opk

ChangeLog

@@ -1,3 +1,221 @@
+2.05 (2025-05-31)
+  + added lightgun support for SDL platforms, using the host mouse for Menacer, Justifier, and Light Phaser
+  * improved handling of OpenGL (may enable OpenGL video mode on devices with JZ4770 CPUs)
+  * improved SDL Joystick/Pad support (should now work with what devices are supported by SDL)
+  * improved support for Windows platform
+  * slightly improved SDL window resize handling
+  * slightly improved UI for keyboard settings
+  * fixed major VDP window rendering regression (games using a VDP window with color table != 0)
+  * fixed interlaced mode rendering in ARMv7 assembly
+  * fixed VDP sprite parsing for MD (fixes a test case in SpriteMaskingTestRom)
+  * fixed crash in VDP rendering (DMA pointer wraparound)
+  * fixed pads not initialized correctly after state load
+  * fixed 32X VINT timing
+  * improved 32X interrupt handling (needs work, still not fully as the SH7604 handles it)
+  * improved 32X WDT handling
+  * several 32X CPU snychronisation fixes
+
+2.04 (2025-04-01)
+  + added mouse support for SDL platforms, using the host mouse for Sega Mouse, Pico Pen
+  * improved scaling when resizing the window on SDL platforms (Linux, OSX, Windows)
+  + added preselect window sizes for integer scaling for SDL platforms
+  + added fullscreen mode for SDL platforms (well, sort of)
+  * attempted to improve pad detection for ps2
+  + noticeably improved save state accuracy (for retroarch run-ahead)
+  * made the DC filter reacting faster (closer to real MD according to notaz' measurements)
+  * fixed a crash when using MD VDP hscroll in conjunction with 2-cell mode
+  * fixed handling shadow/highlight in MD VDP layer A window (fixes Int. Super Star Soccer)
+  * added toggling the field bit in MD VDP status for interlace modes
+  + added some support for MD VDP non-linear color DAC (Knuckles Chaotix color test)
+  * fixed MD+ playing binary tracks as audio (fixes a bug in DOOM Fusion)
+  * improve MCD CPU sync (fixes Shadow of the Beast)
+  * fixed 32X handling of MD H32 mode (MD and 32X layer have 4 px offset)
+  * fixed GG region code reading (fixes Pop Breaker)
+  * fixed GG screen in 224 and 240 line mode (fixes Micro Machines)
+  * fixed GG/SMS display off after scanline start (fixes Madou I)
+
+2.03 (2025-02-15)
+  + added support for Pico keyboard (thanks @qufb for supplying this, literally years ago. Mea maxima culpa!)
+  + added support for SC-3000 keyboard and cassette drive
+  + added configurable mapping for physical keyboard
+  + added virtual keyboard for devices not having a physical keyboard
+  * improved window handling
+  * fix bug in libretro frontend for Pico pad handling (thanks @theLilaQ)
+  * fixes for PS2 vsync handling (thanks @fjtrujy; this is still not being perfect)
+  * fix a crash for MD when rendering in 2-cell VSRAM mode (thanks @notaz for pointing this out)
+  * more fixes for MD+ (for running Doom 32x Resurrection and Fusion)
+  * fix for 32X screen rendering (for running Sonic Robo Blast 2)
+  * fix handling of the 32X cycles UI option
+  * other minor bug fixes and optimizations
+
+2.02 (2024-12-29)
+  * fixes for MSU and MD+ (for running Doom 32x Resurrection and Fusion)
+  + minimal SMS 3D glasses support
+  * updates to libretro common (apparently fixes some obscure bugs)
+  * some CD fixes
+  * fix a cartridge mapping bug for ARM32 (cyclone)
+  * fixes for 32X watchdog timer handling (for running Sonic Robo Blast 2, thanks @notaz)
+  * other minor bug fixes
+
+2.01 (2024-10-17)
+  + added windows build
+  + added integer scaling for SDL window mode
+  + added support for mode 1
+  + added support for MD+/32X+
+  * improved CD emulation compatibility
+  + added support for J-cart games (with 2 additional on-cart controller ports)
+  + added some more mappings for Mega Drive games
+  + added PS2 support for multitap and up to 4 controllers
+  * fixes for PS2 and PSP (audio, menu)
+  * other bug fixes
+
+2.00 (2024-07-19)
+  * improved system detection by extension
+  * improved save/load (Z80, VDP state)
+  * improved some auto mappings for SMS
+  + added stereo for YM2612 DAC output
+  + updated libchdr for zstd support
+  * fixes for pandora
+  * changed miyoo build to ipk
+  * switched miyoo menu key to RESET
+  + added some more MD cartridges
+  * added more keys for some opendingux devices
+  * several fixes for SH2 DRC
+  * several VDP rendering fixes
+  * a lot of other bug fixes (32X reset, Z80 timing, etc.)
+
+2.00-beta3 (2024-04-28)
+  * bug fixes (VDP reset, Z80 reset)
+  * improvements for Sega/Mega CD
+
+2.00-beta2 (2024-03-26)
+  + substantial overhaul of Sega Pico UI
+    supports png images for pad overlays and storyware, new hotkeys, new option (see below)
+  * improved z80 and 68k timing for md and alike
+  * improved timing for 68k access to z80 bus
+  * improved 68k open bus read
+  + implemented "busy" status bit in ym2612
+  * fixes for big endian CPUs for state save/load
+
+2.00-beta1 (2024-03-03)
+  + added ps2 standalone target (thanks to fjtrujy for the fundamentals)
+  * improved psp support
+  * fixed menu RGB/BGR handling
+  * fixed several Z80 related timing inaccuracies
+  * slightly increased volume for CD audio
+  * other bug fixes
+
+2.00-alpha3 (2024-01-13)
+  * improved pandora layer handling
+  + added handling of "fantom bitmap" (DMA to background color)
+  + added SMS option to select TMS bitmap for legacy gfx modes
+  + added stereo sound support for GG
+  + overall revision of Pico support (correct sound, better handling of pen, better save/load handling)
+
+2.00-alpha2 (2023-12-14)
+  * bug fixes and improvements for psp, miyoo, opendingux, pandora
+  + UI fixes and improvements
+  * other bugfixes
+
+2.00-alpha1 (2023-11-09)
+  + improved MD FM sound (DAC ladder effect/clipping, filtering, native rate)
+  + added support for 32X using MD H32 mode
+  + added support for MSU-MD (cartridge+audio CD for MD and 32X)
+  + added basic MD multiplayer support (team player, 4-way play)
+  + added mapping for some additional Mega Drive games
+  + added support for both Overdrive demos
+  * improved save/load handling (SVP, MD mapping, MCD/32X cycle counter)
+  + added more mappers for SMS cartridges
+  + added some support for GG cartridges running in SMS mode
+  * improved SMS graphics if 1st column is blanked
+  + added missing TMS9918 modes
+  + added SG-1000/SC-3000 support (no kbd yet, upcoming)
+  + added additional handhelds/host systems (miyoo, gkd, rg99, odbeta)
+  * improved handling of files by extension
+  + revised menu (more consistent options handling, option profiles)
+  + added dynarec support for PS3 using libretro
+  * add CI to repo (including artifact uploading for the curious)
+  * improved psp support
+  + revived pandora support
+  * a plethora of bug fixes and minor improvements (also most probably new bugs ;-))
+
+1.99 (2021-11-14)
+  + improved Master System support
+  + added SRAM save support for SMS cartridges
+  + added a lot of mappers for SMS cartridges
+  + added Game Gear support
+  + added Master System TMS9918 modes 0 and 2
+  + added upscaling for improved aspect ratios
+  * improved Sega/Mega CD support
+  + added mapping for some additional Mega Drive games
+  + improvements for libretro support (new menu structure courtesy of jdgleaver)
+  * a load of bug fixes and minor improvements (also most probably new bugs ;-))
+
+1.98 (2021-04-02)
+  + CHD support for CD images
+  + mp3 playback support via minimp3
+  + big endian host platform support
+  + sound low pass filtering for more realistic sound (original HW has analog low pass filter)
+  + menu revision
+  * improvements for 8bit fast rendererer (mainly better PAL support)
+  + basic support for legacy dingux (somewhat limited) and retrofw devices
+  + experimental support for PSP (incomplete)
+  * a load of bug fixes and minor improvements (also most probably new bugs ;-))
+
+1.97 (2020-11-23)
+  + powerpc support for DRC
+  + 2x mode in SDL video overlay mode for better color resolution
+  + FM support for Master System (YM2413) (thanks to @hiroshica for supplying)
+  * bug fixes
+
+1.96 (2020-05-16)
+  + added YM2612 SSG-EG emulation
+  * optimisations for SH2 DRC
+  + added a DC filter to 32X PWM sound to reduce clipping
+  * improved VDP emulation (some Overdrive 2 support)
+  + completely rewrote VDP FIFO handling for improved timing accuracy
+  + revived SVP DRC for ARM
+
+1.95 (2019-11-27)
+  + added RISC-V code emitter for SH2 DRC
+  * SH2 DRC optimisations
+
+1.94 (2019-10-12)
+  + revive Helix Mp3 decoder support for ARM (Helix sources not supplied due to licensing)
+  * update GPH device support to use newer GCC 4.7
+  * ARM specific optimisations (speedup for emulated memory access, 32X draw code, DRC disassemblers)
+  * a lot of SH2 DRC and 32X emulation fixes to improve game compatibility
+  * improved DRC code generation (generated code now noticeably faster)
+  * improved SH2 DRC memory access functions (added 32 bit versions, sign extension)
+  * improved SH2 DRC register caching
+  + added branch caching to SH2 DRC
+  * optimisations to DRC ARM backend (code reordering, literal pool)
+  + added polling/idling loop detection and basic loop optimisation to SH2 DRC
+  + added call/return cache for SH2 DRC
+  + added cache for communication areas between 32X CPUs
+  + added T bit caching to ARM SH2 DRC backend (kept in host CPU flags as long as possible if possible)
+  + added x86_64. MIPS and aarch64 backends to SH2 DRC
+  * 32X DMA emulation optimisation
+  + revised SH2 DRC branch handling and host code block linking
+  + added 32X PWM interrupt optimisation (configurable since it may not work with all games)
+  * improved auto frameskipping
+
+1.93 (2019-01-25)
+  + implemented SSF2 mapper for 32X
+  + revived x86 DRC
+  + added 68K overclocking
+  + improved FAME/C 68K emulation speed
+  + support for SMS cheats
+  * 32X DRC bug fixes and improvements
+  * other bug fixes
+
+1.92 (2017-10-04)
+  + some support for obscure VDP features (128K, debug register)
+  + import SPI EEPROM from GPGX
+  * fixes for FAME/C 68k emulation
+  + added Raspberry Pi 1,2 support
+  + other bug fixes
+
 1.91 (2013-10-12)
   + Added OpenDingux support (Paul Cercueil).
   * Save directory changed to ~/.picodrive/ for generic platform build

Makefile

@@ -44,7 +44,7 @@ endif
 
 ifeq "$(ASAN)" "1"
 	CFLAGS += -fsanitize=address -fsanitize=leak -fsanitize=bounds -fno-omit-frame-pointer -fno-common -O1 -g
-	LDLIBS += -fsanitize=address -fsanitize=leak -fsanitize=bounds -static-libasan
+	LDFLAGS += -fsanitize=address -fsanitize=leak -fsanitize=bounds -static-libasan
 else
 ifeq "$(DEBUG)" "0"
 	CFLAGS += -O3 -DNDEBUG
@@ -173,15 +173,7 @@ use_inputmap ?= 1
 PLATFORM := generic
 endif
 ifeq ("$(PLATFORM)",$(filter "$(PLATFORM)","rpi1" "rpi2"))
-CFLAGS += -DHAVE_GLES -DRASPBERRY
-CFLAGS += -I/opt/vc/include/ -I/opt/vc/include/interface/vcos/pthreads/ -I/opt/vc/include/interface/vmcs_host/linux/
-LDFLAGS += -ldl -lbcm_host -L/opt/vc/lib
-# Stupid renaming occured in latest raspbian...
-ifneq (,$(wildcard /opt/vc/lib/libbrcmGLESv2.so))
-LDFLAGS += -lbrcmEGL -lbrcmGLESv2
-else
-LDFLAGS += -lEGL -lGLESv2 # on raspi GLESv1_CM is included in GLESv2
-endif
+CFLAGS += -DHAVE_GLES -DEGL_NO_X11 -DSDL_BUFFER_3X -DSDL_REDRAW_EVT
 OBJS += platform/linux/emu.o platform/linux/blit.o # FIXME
 OBJS += platform/common/plat_sdl.o platform/common/inputmap_kbd.o
 OBJS += platform/libpicofe/plat_sdl.o platform/libpicofe/in_sdl.o
@@ -189,10 +181,8 @@ OBJS += platform/libpicofe/linux/plat.o
 USE_FRONTEND = 1
 endif
 ifeq "$(PLATFORM)" "generic"
-#ifeq (y,$(shell echo "\#include <GLES/gl.h>" | $(CC) -E -xc - >/dev/null 2>&1 && echo y))
 ifeq "$(HAVE_GLES)" "1"
-CFLAGS += -DHAVE_GLES
-LDFLAGS += -lEGL -lGLESv1_CM
+CFLAGS += -DHAVE_GLES -DEGL_NO_X11
 endif
 CFLAGS += -DSDL_OVERLAY_2X -DSDL_BUFFER_3X -DSDL_REDRAW_EVT
 OBJS += platform/linux/emu.o platform/linux/blit.o # FIXME
@@ -304,7 +294,7 @@ OBJS += platform/common/main.o platform/common/emu.o platform/common/upscale.o \
 OBJS += platform/libpicofe/input.o platform/libpicofe/readpng.o \
 	platform/libpicofe/fonts.o
 ifneq (,$(filter %HAVE_GLES, $(CFLAGS)))
-OBJS += platform/libpicofe/gl.o platform/libpicofe/gl_platform.o
+OBJS += platform/libpicofe/gl.o platform/libpicofe/gl_platform.o platform/libpicofe/gl_loader.o
 endif
 
 # libpicofe - sound

README.md

@@ -35,11 +35,20 @@ the pages can be found in `platform/base_readme.txt`.
 
 On all platforms with physical mouse support, PicoDrive can use real mouse
 input to emulate a Sega Mouse or the Pico Pen. A physical mouse can be operated
-in either a captured or uncaptured state, selectable via the `Capture mouse` hotkey.
-Mouse mode can be activated by setting `mouse` as the input device for one
-of the pads. It depends on the game as to which pad should be used for mouse
+in either a captured or uncaptured state, selectable via the `Capture mouse`
+hotkey. Mouse mode can be activated by setting `mouse` as the input device for
+one of the pads. It depends on the game as to which pad should be used for mouse
 input. More information is located in `platform/base_readme.txt`.
 
+### Sega Menacer and Light Phaser and Konami Justifier
+
+PicoDrive can use a physical mouse for light gun emulation on all platforms
+with mouse support. This can be activated by selecting an input device of
+either `light gun` or `justifier`. The `light gun` setting represents the Sega
+Menacer for the Mega Drive or the Light Phaser for the Master System; it may
+be attached to either pad input port. The `justifier` is only available on
+input 2. Supplementary information is provided in `platform/base_readme.txt`.
+
 ### Sega Pico and SC-3000 Keyboards
 
 PicoDrive provides support for the Pico and SC-3000 keyboards. This can be

configure

@@ -49,8 +49,7 @@ have_armv7=""
 have_arm_oabi=""
 have_arm_neon=""
 have_libavcodec=""
-have_libchdr=""
-have_gles="no"
+have_gles=""
 need_sdl="no"
 need_zlib="no"
 
@@ -78,11 +77,9 @@ set_platform()
   case "$platform" in
   rpi1)
     MFLAGS="-mcpu=arm1176jzf-s -mfpu=vfp"
-    have_gles="yes"
     ;;
   rpi2)
     MFLAGS="-mcpu=cortex-a7 -mfpu=neon"
-    have_gles="yes"
     ;;
   generic)
     MFLAGS=""
@@ -217,12 +214,6 @@ if ! test -f "platform/libpicofe/README"; then
   fail "libpicofe is missing, please run 'git submodule update --init'"
 fi
 
-#if [ "$need_warm" = "yes" ]; then
-#  if ! test -f "frontend/warm/README"; then
-#    fail "wARM is missing, please run 'git submodule init && git submodule update'"
-#  fi
-#fi
-
 if [ -z "$ARCH" ]; then
   ARCH=`$CC $MFLAGS $CFLAGS -dumpmachine | awk -F '-' '{print $1}'`
 fi
@@ -359,13 +350,40 @@ EOF
   compile_object "$@"
 }
 
-check_libchdr()
+check_gles()
 {
+  # check for VideoCore stuff for Raspberry Pi
+  if [ VIDEOCORE != "no" -a -d /opt/vc/include -a -d /opt/vc/lib ]; then
+    CFLAGS="$CFLAGS -I/opt/vc/include -I/opt/vc/include/interface/vcos/pthreads -I/opt/vc/include/interface/vmcs_host/linux"
+    LDLIBS="$LDLIBS -L/opt/vc/lib"
+    if [ -f /opt/vc/lib/libbcm_host.so ]; then
+      LDLIBS="$LDLIBS -lbcm_host"
+    fi
+    VIDEOCORE="yes"
+  fi
+
+  # check for GLES headers
   cat > $TMPC <<EOF
-  #include <libchdr/chd.h>
-  int main (int argc, char *argv[]) { chd_open("", 0, NULL, NULL); }
+  #define EGL_NO_X11
+  #include <GLES/gl.h>
+  #include <EGL/egl.h>
+  int main(void) {
+    return (int)eglGetDisplay( (EGLNativeDisplayType)0 );
+  }
 EOF
-  compile_object "$@"
+  if [ "$VIDEOCORE" = "yes" ] && compile_binary $CFLAGS $LDLIBS -lbrcmEGL -lbrcmGLESv2; then
+    have_gles="yes"
+#   LDLIBS="$LDLIBS -lbrcmEGL -lbrcmGLESv2" # non-standard names in old raspbian...
+  elif [ "$VIDEOCORE" = "yes" ] && compile_binary $CFLAGS $LDLIBS -lEGL -lGLESv2; then
+    have_gles="yes"
+#   LDLIBS="$LDLIBS -lEGL -lGLESv2" # on raspi GLESv1_CM is included in GLESv2
+  elif compile_binary $CFLAGS $LDLIBS -lEGL -lGLESv1_CM; then
+    have_gles="yes"
+#   LDLIBS="$LDLIBS -lEGL -lGLESv1_CM"
+  elif compile_binary $CFLAGS $LDLIBS -lEGL -lGLES_CM; then
+    have_gles="yes"
+#   LDLIBS="$LDLIBS -lEGL -lGLES_CM"
+  fi
 }
 
 test -n "$SDL_CONFIG" || SDL_CONFIG="$(ls $SYSROOT/*bin*/sdl-config 2>/dev/null | grep /bin/sdl-config | head -n 1)"
@@ -375,7 +393,7 @@ test -n "$SDL_CONFIG" || SDL_CONFIG="$(ls $SYSROOT/*/*bin*/sdl-config 2>/dev/nul
 SDLVERSION=sdl && echo $SDL_CONFIG | grep -q sdl2 && SDLVERSION=sdl2
 test -n "$SDL_CONFIG" || SDL_CONFIG=false
 
-MAIN_LDLIBS="$LDLIBS -lm"
+MAIN_LDLIBS="-lm"
 
 check_zlib -lz && MAIN_LDLIBS="$MAIN_LDLIBS -lz" || need_zlib="yes"
 
@@ -393,11 +411,6 @@ y|Y|yes)
 *) have_libavcodec="no" ;;
 esac
 
-#if check_libchdr; then
-#  have_libchdr="yes"
-#  MAIN_LDLIBS="-lchdr $MAIN_LDLIBS"
-#fi
-
 # find what audio support we can compile
 if [ "x$sound_drivers" = "x" ]; then
   if check_oss; then sound_drivers="$sound_drivers oss"; fi
@@ -423,7 +436,7 @@ if [ "$need_sdl" = "yes" ]; then
   [ -x "$SDL_CONFIG" ] || \
     fail "sdl-config is missing; please install libsdl (libsdl1.2-dev)"
   CFLAGS="$CFLAGS `$SDL_CONFIG --cflags`"
-  MAIN_LDLIBS="`$SDL_CONFIG --libs` $MAIN_LDLIBS"
+  LDLIBS="$LDLIBS `$SDL_CONFIG --libs`"
   SYSLIBS="$SYSLIBS -ldl"
   need_dl=yes
   check_sdl `$SDL_CONFIG --libs` || fail "please install libsdl (libsdl1.2-dev)"
@@ -440,6 +453,9 @@ if [ "$need_dl" = "yes" ]; then
   esac
 fi
 
+# check for OpenGLES with SDL
+[ "$need_sdl" = "yes" -a "$have_gles" != "no" ] && check_gles
+
 if check_option -Wno-unused_result; then
   CFLAGS="$CFLAGS -Wno-unused-result"
 fi
@@ -448,17 +464,17 @@ fi
 test "x$have_armv6" != "x" || have_armv6="no"
 test "x$have_armv7" != "x" || have_armv7="no"
 test "x$have_libavcodec" != "x" || have_libavcodec="no"
-test "x$have_libchdr" != "x" || have_libchdr="no"
+test "x$have_gles" != "x" || have_gles="no"
 
 echo "architecture        $ARCH"
 echo "platform            $platform"
 echo "sound drivers       $sound_drivers"
 echo "C compiler          $CC"
 echo "C compiler flags    $MFLAGS $CFLAGS"
-echo "libraries           $MAIN_LDLIBS"
+echo "libraries           $LDLIBS $MAIN_LDLIBS"
 echo "linker flags        $LDFLAGS"
 echo "libavcodec (mp3)    $have_libavcodec"
-#echo "libchdr             $have_libchdr"
+echo "use SDL with GLES   $have_gles"
 # echo "ARMv7 optimizations $have_armv7"
 
 echo "# Automatically generated by configure" > $config_mak
@@ -474,7 +490,7 @@ echo "LD = $LD" >> $config_mak
 echo "CFLAGS += $MFLAGS $CFLAGS" >> $config_mak
 echo "ASFLAGS += $MFLAGS $ASFLAGS" >> $config_mak
 echo "LDFLAGS += $LDFLAGS" >> $config_mak
-echo "LDLIBS += $MAIN_LDLIBS" >> $config_mak
+echo "LDLIBS += $LDLIBS $MAIN_LDLIBS" >> $config_mak
 echo >> $config_mak
 
 echo "ARCH = $ARCH" >> $config_mak
@@ -486,9 +502,6 @@ fi
 if [ "$have_libavcodec" = "yes" ]; then
   echo "HAVE_LIBAVCODEC = 1" >> $config_mak
 fi
-if [ "$have_libchdr" = "yes" ]; then
-  echo "HAVE_LIBCHDR = 1" >> $config_mak
-fi
 if [ "$need_zlib" = "yes" ]; then
   echo "PLATFORM_ZLIB = 1" >> $config_mak
 fi

cpu/sh2/compiler.c

@@ -4651,7 +4651,6 @@ static void REGPARM(2) *sh2_translate(SH2 *sh2, int tcache_id)
           FLUSH_CYCLES(sr);
           rcache_get_reg_arg(0, GET_Rn(), NULL);
           tmp = emit_memhandler_read(0);
-          sr  = rcache_get_reg(SHR_SR, RC_GR_RMW, NULL);
           emith_clr_t_cond(sr);
           emith_cmp_r_imm(tmp, 0);
           emith_set_t_cond(sr, DCOND_EQ);

cpu/sh2/sh2.h

@@ -50,7 +50,6 @@ typedef struct SH2_
 #define SH2_STATE_CPOLL (1 << 2)	// polling comm regs
 #define SH2_STATE_VPOLL (1 << 3)	// polling VDP
 #define SH2_STATE_RPOLL (1 << 4)	// polling address in SDRAM
-#define SH2_TIMER_RUN   (1 << 6)	// SOC WDT timer is running
 #define SH2_IN_DRC      (1 << 7)	// DRC in use
 	unsigned int	state;
 	uint32_t	poll_addr;

pico/32x/32x.c

@@ -123,7 +123,6 @@ void Pico32xStartup(void)
 
   PicoMemSetup32x();
   p32x_pwm_ctl_changed();
-  p32x_timers_recalc();
 
   Pico32x.regs[0] |= P32XS_ADEN;
 
@@ -166,6 +165,7 @@ void p32x_reset_sh2s(void)
   sh2_reset(&ssh2);
   sh2_peripheral_reset(&msh2);
   sh2_peripheral_reset(&ssh2);
+  msh2.state = ssh2.state = 0;
 
   // if we don't have BIOS set, perform it's work here.
   // MSH2
@@ -218,6 +218,7 @@ void Pico32xInit(void)
 void PicoPower32x(void)
 {
   memset(&Pico32x, 0, sizeof(Pico32x));
+  memset(p32x_event_times, 0, sizeof(p32x_event_times));
 
   Pico32x.regs[0] = P32XS_REN|P32XS_nRES; // verified
   Pico32x.regs[0x10/2] = 0xffff;
@@ -240,11 +241,10 @@ void PicoUnload32x(void)
 void PicoReset32x(void)
 {
   if (PicoIn.AHW & PAHW_32X) {
-    p32x_trigger_irq(NULL, SekCyclesDone(), P32XI_VRES);
-    p32x_sh2_poll_event(msh2.poll_addr, &msh2, SH2_IDLE_STATES, SekCyclesDone());
-    p32x_sh2_poll_event(ssh2.poll_addr, &ssh2, SH2_IDLE_STATES, SekCyclesDone());
-    p32x_pwm_ctl_changed();
-    p32x_timers_recalc();
+    p32x_trigger_irq(NULL, Pico.t.m68c_aim, P32XI_VRES);
+    p32x_m68k_poll_event(0, -1);
+    p32x_sh2_poll_event(msh2.poll_addr, &msh2, SH2_IDLE_STATES, Pico.t.m68c_aim);
+    p32x_sh2_poll_event(ssh2.poll_addr, &ssh2, SH2_IDLE_STATES, Pico.t.m68c_aim);
   }
 }
 
@@ -279,7 +279,7 @@ void Pico32xDrawSync(SH2 *sh2)
   // the fast renderer isn't operating on a line-by-line base
   if (sh2 && !(PicoIn.opt & POPT_ALT_RENDERER)) {
     unsigned int cycle = (sh2 ? sh2_cycles_done_m68k(sh2) : SekCyclesDone());
-    int line = ((cycle - Pico.t.m68c_frame_start) * (long long)((1LL<<32)/488.5)) >> 32;
+    int line = DIVQ32(cycle - Pico.t.m68c_frame_start, 488.5);
 
     if (Pico32x.sync_line < line && line < (Pico.video.reg[1] & 8 ? 240 : 224)) {
       // make sure the MD image is also sync'ed to this line for merging
@@ -324,7 +324,7 @@ static void p32x_start_blank(void)
     Pico32xSwapDRAM(Pico32x.pending_fb ^ P32XV_FS);
   }
 
-  p32x_trigger_irq(NULL, Pico.t.m68c_aim, P32XI_VINT);
+  p32x_schedule_vint(NULL, Pico.t.m68c_aim);
   p32x_sh2_poll_event(msh2.poll_addr, &msh2, SH2_STATE_VPOLL, Pico.t.m68c_aim);
   p32x_sh2_poll_event(ssh2.poll_addr, &ssh2, SH2_STATE_VPOLL, Pico.t.m68c_aim);
 }
@@ -336,7 +336,7 @@ static void p32x_end_blank(void)
   if ((Pico32x.vdp_regs[0] & P32XV_Mx) != 0) // no forced blanking
     Pico32x.vdp_regs[0x0a/2] &= ~P32XV_PEN; // no palette access
   if (!(Pico32x.sh2_regs[0] & 0x80)) {
-    // NB must precede VInt per hw manual, min 4 SH-2 cycles to pass Mars Check
+    // TODO must be early by min 4 SH-2 cycles to pass Mars Check?
     Pico32x.hint_counter = (int)(-1.5*0x10);
     p32x_schedule_hint(NULL, Pico.t.m68c_aim);
   }
@@ -349,6 +349,7 @@ void p32x_schedule_hint(SH2 *sh2, unsigned int m68k_cycles)
 {
   // rather rough, 32x hint is useless in practice
   int after;
+
   if (!((Pico32x.sh2irq_mask[0] | Pico32x.sh2irq_mask[1]) & 4))
     return; // nobody cares
   if (!(Pico32x.sh2_regs[0] & 0x80) && (Pico.video.status & PVS_VB2))
@@ -363,6 +364,17 @@ void p32x_schedule_hint(SH2 *sh2, unsigned int m68k_cycles)
     p32x_event_schedule(m68k_cycles, P32X_EVENT_HINT, after);
 }
 
+void p32x_schedule_vint(SH2 *sh2, unsigned int m68k_cycles)
+{
+  // according to 32X hw manual, HInt and VInt are tied to HBLANK and VBLANK.
+  // Moreover, VBLANK is 27 px (224 mclk, 32 68k cycles) after HBLANK.
+  // TODO check if that matches HBLANK/VBLANK on the cartridge slot?
+  if (sh2 != NULL)
+    p32x_event_schedule_sh2(sh2, P32X_EVENT_VINT, 32);
+  else
+    p32x_event_schedule(m68k_cycles, P32X_EVENT_VINT, 32);
+}
+
 /* events */
 static void fillend_event(unsigned int now)
 {
@@ -377,6 +389,21 @@ static void hint_event(unsigned int now)
   p32x_schedule_hint(NULL, now);
 }
 
+static void vint_event(unsigned int now)
+{
+  p32x_trigger_irq(NULL, now, P32XI_VINT);
+}
+
+static void mtimer_event(unsigned int now)
+{
+  p32x_timer_irq(&msh2, now);
+}
+
+static void stimer_event(unsigned int now)
+{
+  p32x_timer_irq(&ssh2, now);
+}
+
 typedef void (event_cb)(unsigned int now);
 
 /* times are in m68k (7.6MHz) cycles */
@@ -386,6 +413,9 @@ static event_cb *p32x_event_cbs[P32X_EVENT_COUNT] = {
   p32x_pwm_irq_event, // P32X_EVENT_PWM
   fillend_event,      // P32X_EVENT_FILLEND
   hint_event,         // P32X_EVENT_HINT
+  vint_event,         // P32X_EVENT_VINT
+  mtimer_event,       // P32X_EVENT_MTIMER
+  stimer_event,       // P32X_EVENT_STIMER
 };
 
 // schedule event at some time 'after', in m68k clocks
@@ -407,6 +437,9 @@ void p32x_event_schedule_sh2(SH2 *sh2, enum p32x_event event, int after)
   unsigned int now = sh2_cycles_done_m68k(sh2);
   int left_to_next;
 
+  if (sh2->state & SH2_IDLE_STATES)
+    now = SekCyclesDone();
+
   p32x_event_schedule(now, event, after);
 
   left_to_next = C_M68K_TO_SH2(sh2, (int)(event_time_next - now));
@@ -483,8 +516,10 @@ void p32x_sync_other_sh2(SH2 *sh2, unsigned int m68k_target)
   int left_to_event;
   int m68k_cycles;
 
-  if (osh2->state & SH2_STATE_RUN)
+  if (osh2->state & SH2_STATE_RUN) {
+    sh2_end_run(sh2, 0);
     return;
+  }
 
   m68k_cycles = m68k_target - osh2->m68krcycles_done;
   if (m68k_cycles < 200)
@@ -512,7 +547,7 @@ void p32x_sync_other_sh2(SH2 *sh2, unsigned int m68k_target)
 }
 
 #define STEP_LS 24
-#define STEP_N 528 // at least one line (488)
+#define STEP_N 192 // NFL; TODO at least a scanline (489) for good performance?
 
 #define sync_sh2s_normal p32x_sync_sh2s
 //#define sync_sh2s_lockstep p32x_sync_sh2s
@@ -520,7 +555,7 @@ void p32x_sync_other_sh2(SH2 *sh2, unsigned int m68k_target)
 /* most timing is in 68k clock */
 void sync_sh2s_normal(unsigned int m68k_target)
 {
-  unsigned int now, target, next, timer_cycles;
+  unsigned int now, target, next;
   int cycles;
 
   elprintf(EL_32X, "sh2 sync to %u", m68k_target);
@@ -533,7 +568,6 @@ void sync_sh2s_normal(unsigned int m68k_target)
   now = msh2.m68krcycles_done;
   if (CYCLES_GT(now, ssh2.m68krcycles_done))
     now = ssh2.m68krcycles_done;
-  timer_cycles = now;
 
   pprof_start(m68k);
   while (CYCLES_GT(m68k_target, now))
@@ -590,20 +624,7 @@ void sync_sh2s_normal(unsigned int m68k_target)
         if (!(ssh2.state & SH2_IDLE_STATES)) 
           now = ssh2.m68krcycles_done;
       }
-      if (CYCLES_GT(now, timer_cycles+STEP_N)) {
-        if  (msh2.state & SH2_TIMER_RUN)
-          p32x_timer_do(&msh2, now - timer_cycles);
-        if  (ssh2.state & SH2_TIMER_RUN)
-          p32x_timer_do(&ssh2, now - timer_cycles);
-        timer_cycles = now;
-      }
     }
-
-    if  (msh2.state & SH2_TIMER_RUN)
-      p32x_timer_do(&msh2, now - timer_cycles);
-    if  (ssh2.state & SH2_TIMER_RUN)
-      p32x_timer_do(&ssh2, now - timer_cycles);
-    timer_cycles = now;
   }
   pprof_end_sub(m68k);
 
@@ -616,9 +637,6 @@ void sync_sh2s_normal(unsigned int m68k_target)
     if (CYCLES_GT(m68k_target, ssh2.m68krcycles_done))
       ssh2.m68krcycles_done = m68k_target;
   }
-
-  // everyone is in sync now
-  Pico32x.comm_dirty = 0;
 }
 
 void sync_sh2s_lockstep(unsigned int m68k_target)
@@ -663,6 +681,9 @@ void PicoFrame32x(void)
     Pico.est.rendstatus |= PDRAW_SYNC_NEEDED;
   PicoFrameHints();
 
+  p32x_timer_do(&msh2, Pico.t.m68c_aim);
+  p32x_timer_do(&ssh2, Pico.t.m68c_aim);
+
   elprintf(EL_32X, "poll: %02x %02x %02x",
     Pico32x.emu_flags & 3, msh2.state, ssh2.state);
 }
@@ -692,11 +713,11 @@ void Pico32xStateLoaded(int is_early)
 
   if (CYCLES_GE(sh2s[0].m68krcycles_done - Pico.t.m68c_aim, 500) ||
       CYCLES_GE(sh2s[1].m68krcycles_done - Pico.t.m68c_aim, 500))
-    sh2s[0].m68krcycles_done = sh2s[1].m68krcycles_done = SekCyclesDone();
-  p32x_update_irls(NULL, SekCyclesDone());
-  p32x_timers_recalc();
+    sh2s[0].m68krcycles_done = sh2s[1].m68krcycles_done = Pico.t.m68c_aim;
+  p32x_update_irls(NULL, Pico.t.m68c_aim);
+  sh2_peripheral_state_loaded();
   p32x_pwm_state_loaded();
-  p32x_run_events(SekCyclesDone());
+  p32x_run_events(Pico.t.m68c_aim);
 
   // TODO wakeup CPUs for now. poll detection stuff must go to the save state!
   p32x_m68k_poll_event(0, -1);

pico/32x/draw.c

@@ -129,7 +129,6 @@ void FinalizeLine32xRGB555(int sh, int line, struct PicoEState *est)
   unsigned short *dst = est->DrawLineDest;
   unsigned short *pal = Pico32xMem->pal_native;
   unsigned char  *pmd = est->HighCol + 8;
-  unsigned short *palmd = est->HighPal;
   unsigned short *dram, *p32x;
   unsigned char   mdbg;
   int h32 = !(Pico.video.reg[12] & 0x1);

pico/32x/memory.c

@@ -198,7 +198,7 @@ static NOINLINE u32 sh2_poll_read(u32 a, u32 d, unsigned int cycles, SH2* sh2)
   int hix = (a >> 1) % PFIFO_CNT;
   struct sh2_poll_fifo *fifo = sh2_poll_fifo[hix];
   struct sh2_poll_fifo *p;
-  int cpu = sh2 ? sh2->is_slave : -1;
+  int cpu = 1 << (sh2 ? sh2->is_slave : 2), age = sh2 ? 60 : 360;
   unsigned idx;
 
   a &= ~0x20000000; // ignore writethrough bit
@@ -208,14 +208,14 @@ static NOINLINE u32 sh2_poll_read(u32 a, u32 d, unsigned int cycles, SH2* sh2)
     p = &fifo[idx];
     idx = (idx+1) % PFIFO_SZ;
 
-    if (cpu != p->cpu) {
-      if (CYCLES_GT(cycles, p->cycles+60)) { // ~180 sh2 cycles, Spiderman
+    if (p->cpu & cpu) {
+      if (CYCLES_GT(cycles, p->cycles+age)) { // ~180 sh2 cycles, Spiderman
         // drop older fifo stores that may cause synchronisation problems.
-        p->a = -1;
+        p->cpu &= ~cpu;
       } else if (p->a == a) {
         // replace current data with fifo value and discard fifo entry
         d = p->d;
-        p->a = -1;
+        p->cpu &= ~cpu;
         break;
       }
     }
@@ -228,7 +228,7 @@ static NOINLINE void sh2_poll_write(u32 a, u32 d, unsigned int cycles, SH2 *sh2)
   int hix = (a >> 1) % PFIFO_CNT;
   struct sh2_poll_fifo *fifo = sh2_poll_fifo[hix];
   struct sh2_poll_fifo *q;
-  int cpu = sh2 ? sh2->is_slave : -1;
+  int cpu = 1 << (sh2 ? sh2->is_slave : 2);
   unsigned rd = sh2_poll_rd[hix], wr = sh2_poll_wr[hix];
   unsigned idx, nrd;
 
@@ -238,8 +238,8 @@ static NOINLINE void sh2_poll_write(u32 a, u32 d, unsigned int cycles, SH2 *sh2)
   for (idx = nrd = wr; idx != rd; ) {
     idx = (idx-1) % PFIFO_SZ;
     q = &fifo[idx];
-    if (q->a == a && q->cpu != cpu)	{ q->a = -1; }
-    if (q->a != -1)			{ nrd = idx; }
+    if (q->a == a && (q->cpu & cpu))	{ q->cpu = 0; }
+    if (q->cpu)				{ nrd = idx; }
   }
   rd = nrd;
 
@@ -248,12 +248,12 @@ static NOINLINE void sh2_poll_write(u32 a, u32 d, unsigned int cycles, SH2 *sh2)
   // NB this can take an eternity on m68k: mov.b <addr1.l>,<addr2.l> needs
   // 28 m68k-cycles (~80 sh2-cycles) to complete (observed in Metal Head)
   q = &fifo[(sh2_poll_wr[hix]-1) % PFIFO_SZ];
-  if (rd != wr && q->a == a && !CYCLES_GT(cycles,q->cycles + (cpu<0 ? 30:4))) {
+  if (rd != wr && q->a == a && !CYCLES_GT(cycles,q->cycles + (!sh2 ? 30:4))) {
     q->d = d;
   } else {
     // store write to poll address in fifo
-    fifo[wr] =
-        (struct sh2_poll_fifo){ .cycles = cycles, .a = a, .d = d, .cpu = cpu };
+    fifo[wr] = (struct sh2_poll_fifo)
+        { .cycles = cycles, .a = a, .d = d, .cpu = ~cpu & 7 };
     wr = (wr+1) % PFIFO_SZ;
     if (wr == rd)
       // fifo overflow, discard oldest value
@@ -419,8 +419,10 @@ static void p32x_reg_write8(u32 a, u32 d)
       if ((d ^ r[0]) & ~d & P32XS_ADEN) {
         d |= P32XS_nRES;
         Pico32xShutdown();
-      } else if ((d ^ r[0]) & d & P32XS_nRES)
+      } else if ((d ^ r[0]) & d & P32XS_nRES) {
+        p32x_sync_sh2s(SekCyclesDone());
         p32x_reset_sh2s();
+      }
       REG8IN16(r, 0x01) &= ~(P32XS_nRES|P32XS_ADEN);
       REG8IN16(r, 0x01) |= d & (P32XS_nRES|P32XS_ADEN);
       return;
@@ -587,8 +589,10 @@ static void p32x_reg_write16(u32 a, u32 d)
       if ((d ^ r[0]) & ~d & P32XS_ADEN) {
         d |= P32XS_nRES;
         Pico32xShutdown();
-      } else if ((d ^ r[0]) & d & P32XS_nRES)
+      } else if ((d ^ r[0]) & d & P32XS_nRES) {
+        p32x_sync_sh2s(SekCyclesDone());
         p32x_reset_sh2s();
+      }
       r[0] &= ~(P32XS_FM|P32XS_nRES|P32XS_ADEN);
       r[0] |= d & (P32XS_FM|P32XS_nRES|P32XS_ADEN);
       return;
@@ -758,7 +762,7 @@ static u32 p32x_sh2reg_read16(u32 a, SH2 *sh2)
       cycles = sh2_cycles_done_m68k(sh2);
       sh2s_sync_on_read(sh2, cycles);
       return sh2_poll_read(a, Pico32x.sh2_regs[4 / 2], cycles, sh2);
-    case 0x06/2:
+    case 0x06/2: // DREQ ctrl
       return (r[a / 2] & ~P32XS_FULL) | 0x4000;
     case 0x08/2: // DREQ src
     case 0x0a/2:
@@ -846,7 +850,7 @@ static void p32x_sh2reg_write8(u32 a, u32 d, SH2 *sh2)
         Pico32x.sh2_regs[4 / 2] = d;
         p32x_sh2_poll_event(a, sh2->other_sh2, SH2_STATE_CPOLL, cycles);
         if (p32x_sh2_ready(sh2->other_sh2, cycles+8))
-          sh2_end_run(sh2, 4);
+          p32x_sync_other_sh2(sh2, cycles);
         sh2_poll_write(a & ~1, d, cycles, sh2);
       }
       return;
@@ -873,7 +877,7 @@ static void p32x_sh2reg_write8(u32 a, u32 d, SH2 *sh2)
         p32x_m68k_poll_event(a, P32XF_68KCPOLL);
         p32x_sh2_poll_event(a, sh2->other_sh2, SH2_STATE_CPOLL, cycles);
         if (p32x_sh2_ready(sh2->other_sh2, cycles+8))
-          sh2_end_run(sh2, 0);
+          p32x_sync_other_sh2(sh2, cycles);
         sh2_poll_write(a & ~1, r[a / 2], cycles, sh2);
       }
       return;
@@ -943,6 +947,7 @@ static void p32x_sh2reg_write16(u32 a, u32 d, SH2 *sh2)
       Pico32x.sh2irqi[sh2->is_slave] &= ~P32XI_HINT;
       goto irls;
     case 0x1a/2:
+      p32x_sync_other_sh2(sh2, sh2_cycles_done_m68k(sh2));
       Pico32x.regs[2 / 2] &= ~(1 << sh2->is_slave);
       p32x_update_cmd_irq(sh2, 0);
       return;
@@ -966,7 +971,7 @@ static void p32x_sh2reg_write16(u32 a, u32 d, SH2 *sh2)
         p32x_m68k_poll_event(a, P32XF_68KCPOLL);
         p32x_sh2_poll_event(a, sh2->other_sh2, SH2_STATE_CPOLL, cycles);
         if (p32x_sh2_ready(sh2->other_sh2, cycles+8))
-          sh2_end_run(sh2, 0);
+          p32x_sync_other_sh2(sh2, cycles);
         sh2_poll_write(a, d, cycles, sh2);
       }
       return;
@@ -1636,7 +1641,7 @@ static void sh2_sdram_poll(u32 a, u32 d, SH2 *sh2)
   sh2_poll_write(a, d, cycles, sh2);
   p32x_sh2_poll_event(a, sh2->other_sh2, SH2_STATE_RPOLL, cycles);
   if (p32x_sh2_ready(sh2->other_sh2, cycles+8))
-    sh2_end_run(sh2, 0);
+    p32x_sync_other_sh2(sh2, cycles);
   DRC_RESTORE_SR(sh2);
 }
 

pico/32x/sh2soc.c

@@ -70,7 +70,8 @@ static void dmac_te_irq(SH2 *sh2, struct dma_chan *chan)
                dmac->vcrdma0 : dmac->vcrdma1;
 
   elprintf(EL_32XP, "dmac irq %d %d", level, vector);
-  sh2_internal_irq(sh2, level, vector & 0x7f);
+  if (level)
+    sh2_internal_irq(sh2, level, vector & 0x7f);
 }
 
 static void dmac_transfer_complete(SH2 *sh2, struct dma_chan *chan)
@@ -189,53 +190,78 @@ static void dmac_trigger(SH2 *sh2, struct dma_chan *chan)
     chan->sar, chan->dar, chan->tcr, chan->chcr, sh2->pc);
 }
 
-// FIXME, timer state, in unused peri_regs space to have it in state save/load
-#define timer_cycles(sh2) *((u32 *)&(sh2)->peri_regs[0x1c])
+// wdt timers for the 2 SH2's
+
 static u32 timer_tick_shift[2];
+static u32 timer_last_cycle[2];
 
-// timers
-void p32x_timers_recalc(void)
+void p32x_timer_recalc(SH2 *sh2)
 {
-  int tmp, i;
+  int tmp, i = sh2->is_slave;
+
+  if (PREG8(sh2->peri_regs, 0x80) & 0x20) { // TME
+    // SH2 timer step
+    tmp = PREG8(sh2->peri_regs, 0x80) & 7;
 
-  // SH2 timer step
-  for (i = 0; i < 2; i++) {
-    sh2s[i].state &= ~SH2_TIMER_RUN;
-    if (PREG8(sh2s[i].peri_regs, 0x80) & 0x20) // TME
-      sh2s[i].state |= SH2_TIMER_RUN;
-    tmp = PREG8(sh2s[i].peri_regs, 0x80) & 7;
     // Sclk cycles per timer tick
     if (tmp >= 6) tmp++;
     if (tmp) tmp += 5;
     else tmp = 1;
     timer_tick_shift[i] = tmp;
     elprintf(EL_32XP, "WDT cycles[%d] = %d", i, 1 << tmp);
+
+    tmp = (1 << timer_tick_shift[i]) - Pico32x.wdt_cycle[i]; // current tick
+    if (tmp < 0) tmp = 0; // in case of invalid tick cycle data
+    tmp += (255 - PREG8(sh2->peri_regs, 0x81)) << timer_tick_shift[i];
+
+    if (PREG8(sh2->peri_regs, 0xe3) >> 4)
+      p32x_event_schedule_sh2(sh2, P32X_EVENT_MTIMER+i, DIVQ32(tmp,3));
   }
 }
 
-NOINLINE void p32x_timer_do(SH2 *sh2, unsigned int m68k_slice)
+void p32x_timer_do(SH2 *sh2, unsigned int now)
 {
-  unsigned int cycles = m68k_slice * 3;
+  int i = sh2->is_slave;
   void *pregs = sh2->peri_regs;
-  int cnt; int i = sh2->is_slave;
 
   // WDT timer
-  timer_cycles(sh2) += cycles;
-  cnt = (timer_cycles(sh2) >> timer_tick_shift[i]);
-  if (cnt) {
-    timer_cycles(sh2) -= cnt << timer_tick_shift[i];
+  if (PREG8(sh2->peri_regs, 0x80) & 0x20) { // TME
+    unsigned int cycles = (now - timer_last_cycle[i]) * 3;
+    int cnt;
 
-    cnt += PREG8(pregs, 0x81);
-    if (cnt >= 0x100) {
-      int level = PREG8(pregs, 0xe3) >> 4;
-      int vector = PREG8(pregs, 0xe4) & 0x7f;
-      PREG8(pregs, 0x80) |= 0x80; // WOVF
-      elprintf(EL_32XP, "%csh2 WDT irq (%d, %d)",
-        i ? 's' : 'm', level, vector);
+    cycles += Pico32x.wdt_cycle[i];
+    cnt = (cycles >> timer_tick_shift[i]);
+    cycles -= cnt << timer_tick_shift[i];
+
+    PREG8(pregs, 0x81) += cnt;
+    Pico32x.wdt_cycle[i] = cycles;
+    timer_last_cycle[i] = now;
+  }
+}
+
+void p32x_timer_irq(SH2 *sh2, unsigned int now)
+{
+  void *pregs = sh2->peri_regs;
+
+  if (PREG8(sh2->peri_regs, 0x80) & 0x20) { // TME
+    int i = sh2->is_slave;
+    int after = 256 << timer_tick_shift[i];
+    int level = PREG8(pregs, 0xe3) >> 4;
+    int vector = PREG8(pregs, 0xe4) & 0x7f;
+
+    PREG8(pregs, 0x80) |= 0x80; // WOVF
+    elprintf(EL_32XP, "%csh2 WDT irq (%d, %d)",
+      sh2->is_slave ? 's' : 'm', level, vector);
+
+    if (level) {
       sh2_internal_irq(sh2, level, vector);
-      cnt &= 0xff;
+      p32x_event_schedule_sh2(sh2, P32X_EVENT_MTIMER+i, DIVQ32(after,3));
     }
-    PREG8(pregs, 0x81) = cnt;
+
+    // reset counting to avoid rounding errors
+    PREG8(pregs, 0x81) = 0x00;
+    Pico32x.wdt_cycle[i] = 0;
+    timer_last_cycle[i] = now;
   }
 }
 
@@ -249,6 +275,13 @@ void sh2_peripheral_reset(SH2 *sh2)
   PREG8(sh2->peri_regs, 0x017) = 0xe0; // TOCR
 }
 
+void sh2_peripheral_state_loaded(void)
+{
+  timer_last_cycle[0] = timer_last_cycle[1] = Pico.t.m68c_aim;
+  p32x_timer_recalc(&msh2);
+  p32x_timer_recalc(&ssh2);
+}
+
 // ------------------------------------------------------------------
 // SH2 internal peripheral memhandlers
 // we keep them in little endian format
@@ -260,6 +293,8 @@ u32 REGPARM(2) sh2_peripheral_read8(u32 a, SH2 *sh2)
 
   DRC_SAVE_SR(sh2);
   a &= 0x1ff;
+  if (a == 0x81)
+    p32x_timer_do(sh2, sh2_cycles_done_m68k(sh2));
   d = PREG8(r, a);
 
   elprintf_sh2(sh2, EL_32XP, "peri r8  [%08x]       %02x @%06x",
@@ -279,6 +314,8 @@ u32 REGPARM(2) sh2_peripheral_read16(u32 a, SH2 *sh2)
 
   DRC_SAVE_SR(sh2);
   a &= 0x1fe;
+  if (a == 0x80)
+    p32x_timer_do(sh2, sh2_cycles_done_m68k(sh2));
   d = r[MEM_BE2(a / 2)];
 
   elprintf_sh2(sh2, EL_32XP, "peri r16 [%08x]     %04x @%06x",
@@ -297,6 +334,8 @@ u32 REGPARM(2) sh2_peripheral_read32(u32 a, SH2 *sh2)
 
   DRC_SAVE_SR(sh2);
   a &= 0x1fc;
+  if (a == 0x80)
+    p32x_timer_do(sh2, sh2_cycles_done_m68k(sh2));
   d = sh2->peri_regs[a / 4];
 
   elprintf_sh2(sh2, EL_32XP, "peri r32 [%08x] %08x @%06x",
@@ -335,7 +374,8 @@ static void sci_trigger(SH2 *sh2, u8 *r)
     int vector = PREG8(oregs, 0x64) & 0x7f;
     elprintf_sh2(sh2, EL_32XP, "SCI tx irq (%d, %d)",
       level, vector);
-    sh2_internal_irq(sh2, level, vector);
+    if (level)
+      sh2_internal_irq(sh2, level, vector);
   }
   // TODO: TEIE
   if (PREG8(oregs, 2) & 0x40) { // RIE - rx irq enabled
@@ -343,7 +383,8 @@ static void sci_trigger(SH2 *sh2, u8 *r)
     int vector = PREG8(oregs, 0x63) & 0x7f;
     elprintf_sh2(sh2->other_sh2, EL_32XP, "SCI rx irq (%d, %d)",
       level, vector);
-    sh2_internal_irq(sh2->other_sh2, level, vector);
+    if (level)
+      sh2_internal_irq(sh2->other_sh2, level, vector);
   }
 }
 
@@ -384,6 +425,10 @@ void REGPARM(3) sh2_peripheral_write8(u32 a, u32 d, SH2 *sh2)
     d |= 0xe0;
     PREG8(r, a) = d;
     break;
+  case 0x0e3:
+    p32x_timer_do(sh2, sh2_cycles_done_m68k(sh2));
+    p32x_timer_recalc(sh2);
+    break;
   default:
     if ((a & 0x1c0) == 0x140)
       p32x_sh2_poll_event(a, sh2, SH2_STATE_CPOLL, SekCyclesDone());
@@ -405,13 +450,20 @@ void REGPARM(3) sh2_peripheral_write16(u32 a, u32 d, SH2 *sh2)
   if (a == 0x80) {
     if ((d & 0xff00) == 0xa500) { // WTCSR
       PREG8(r, 0x80) = d;
-      p32x_timers_recalc();
-      timer_cycles(sh2) = 0;
     }
-    if ((d & 0xff00) == 0x5a00) // WTCNT
+    if ((d & 0xff00) == 0x5a00) { // WTCNT
       PREG8(r, 0x81) = d;
+    }
+    // reset to avoid rounding problems
+    Pico32x.wdt_cycle[sh2->is_slave] = 0;
+    timer_last_cycle[sh2->is_slave] = sh2_cycles_done_m68k(sh2);
+    p32x_timer_recalc(sh2);
   } else {
     r[MEM_BE2(a / 2)] = d;
+    if (a == 0x0e2) {
+      p32x_timer_do(sh2, sh2_cycles_done_m68k(sh2));
+      p32x_timer_recalc(sh2);
+    }
     if ((a & 0x1c0) == 0x140)
       p32x_sh2_poll_event(a, sh2, SH2_STATE_CPOLL, SekCyclesDone());
   }

pico/cd/memory.c

@@ -101,7 +101,7 @@ u32 pcd_stopwatch_read(int sub)
 {
   // ugh... stopwatch runs 384 cycles per step, divide by mult with inverse
   u32 d = sub ? SekCyclesDoneS68k() : pcd_cycles_m68k_to_s68k(SekCyclesDone());
-  d = ((d - Pico_mcd->m.stopwatch_base_c) * ((1LL << 32) / 384)) >> 32;
+  d = DIVQ32(d - Pico_mcd->m.stopwatch_base_c, 384);
   return d & 0x0fff;
 }
 

pico/cd/pcm.c

@@ -62,7 +62,7 @@ void pcd_pcm_sync(unsigned int to)
   if ((int)(to - cycles) < 384)
     return;
 
-  steps = (to - cycles) * ((1LL << 32) / 384) >> 32;
+  steps = DIVQ32(to - cycles, 384);
   if (Pico_mcd->pcm_mixpos + steps > PCM_MIXBUF_LEN)
     // shouldn't happen, but occasionally does
     steps = PCM_MIXBUF_LEN - Pico_mcd->pcm_mixpos;

pico/draw.c

@@ -600,7 +600,7 @@ static void DrawWindow(int tstart, int tend, int prio, int sh,
   struct PicoVideo *pvid = &est->Pico->video;
   int tilex,ty,nametab,code,oldcode=-1,blank=-1; // The tile we know is blank
   int yshift,ymask;
-  u32 pack=0;
+  u32 pack=0, pal=0;
   u32 *hc=NULL, lflags=0; // referenced in DrawTile
 
   yshift = 4, ymask = 0x7;
@@ -634,7 +634,7 @@ static void DrawWindow(int tstart, int tend, int prio, int sh,
     sh = (sh ? 0x80 : 0x00); // sh and low prio -> shadow
     for (; tilex < tend; tilex++)
     {
-      int dx, pal=0;
+      int dx;
 
       code = PicoMem.vram[nametab + tilex];
       if ((code>>15) != prio) {
@@ -652,7 +652,7 @@ static void DrawWindow(int tstart, int tend, int prio, int sh,
     sh = lflags; // sh and high prio -> no shadow (lflags to suppress warning)
     for (; tilex < tend; tilex++)
     {
-      int dx, pal=0;
+      int dx;
 
       code = PicoMem.vram[nametab + tilex];
       if((code>>15) != prio) {
@@ -1245,7 +1245,6 @@ static NOINLINE void ParseSprites(int max_lines, int limit)
       {
         unsigned char *p = &HighLnSpr[y][0];
         int cnt = p[0] & 0x7f;
-        if (p[1] & SPRL_MASKED) continue;               // masked?
 
         if (p[3] >= max_line_sprites) continue;         // sprite limit?
         p[3] ++;
@@ -1258,6 +1257,8 @@ static NOINLINE void ParseSprites(int max_lines, int limit)
         }
         p[2] += w;
 
+        if (p[1] & SPRL_MASKED) continue;               // masked?
+
         if (sx == -0x78) {
           if (p[1] & (SPRL_HAVE_X|SPRL_TILE_OVFL))
             p[1] |= SPRL_MASKED; // masked, no more sprites for this line

pico/draw_arm.S

@@ -836,7 +836,9 @@ DrawLayer:
 
     mov     r0, sp
     mov     r1, r9, lsr #29
-    bl      DrawStripInterlace @ struct TileStrip *ts, int plane_sh
+    mov     r2, r9, lsr #8
+    and     r2, r2, #0xff
+    bl      DrawStripInterlace @ struct TileStrip *ts, int plane_sh, int cellskip
 
     add     sp, sp, #6*4
     ldmfd   sp!, {r4-r11,pc}

pico/memory.c

@@ -27,9 +27,9 @@ static void xmap_set(uptr *map, int shift, u32 start_addr, u32 end_addr,
     const void *func_or_mh, int is_func)
 {
 #ifdef __clang__
-  // workaround bug (segfault) in 
+  // workaround bug (segfault) in
   // Apple LLVM version 4.2 (clang-425.0.27) (based on LLVM 3.2svn)
-  volatile 
+  volatile
 #endif
   uptr addr = (uptr)func_or_mh;
   int mask = (1 << shift) - 1;
@@ -236,9 +236,9 @@ static void m68k_unmapped_write16(u32 a, u32 d)
 void m68k_map_unmap(u32 start_addr, u32 end_addr)
 {
 #ifdef __clang__
-  // workaround bug (segfault) in 
+  // workaround bug (segfault) in
   // Apple LLVM version 4.2 (clang-425.0.27) (based on LLVM 3.2svn)
-  volatile 
+  volatile
 #endif
   uptr addr;
   int shift = M68K_MEM_SHIFT;
@@ -457,6 +457,30 @@ static u32 read_pad_mouse(int i, u32 out_bits)
   return value;
 }
 
+static u32 read_pad_menacer(int i, u32 out_bits)
+{
+  u32 pad = PicoIn.padInt[i]; // Get pad MXYZ SACB RLDU; NB not(!) inverted
+  u32 value;
+
+  value = ((pad>>4) & 0x09) | ((pad>>3) & 0x4) | ((pad>>5) & 0x2); // .... SCAB
+
+  value |= 0x40; // TH line is light sensor. Pretend there's always darkness...
+  return value;
+}
+
+static u32 read_pad_justifier(int i, u32 out_bits)
+{
+  u32 pad = ~PicoIn.padInt[i]; // Get inverse of pad MXYZ SACB RLDU
+  u32 value = 0x30; // .111 0000, for ID
+
+  // output data: .tgd ...., t = TH, g = gun select, d = disable
+  if (!(PicoMem.ioports[i+1] & out_bits & 0x70)) // blue gun, enabled
+    value |= ((pad>>6) & 0x03); // .011 ..SA
+
+  value |= (out_bits & 0x40);
+  return value;
+}
+
 static u32 read_nothing(int i, u32 out_bits)
 {
   return 0xff;
@@ -464,6 +488,13 @@ static u32 read_nothing(int i, u32 out_bits)
 
 typedef u32 (port_read_func)(int index, u32 out_bits);
 
+int port_type[3] = {
+  PICO_INPUT_PAD_3BTN,
+  PICO_INPUT_PAD_3BTN,
+  PICO_INPUT_NOTHING
+};
+int port_lightgun;
+
 static port_read_func *port_readers[3] = {
   read_pad_3btn,
   read_pad_3btn,
@@ -472,23 +503,31 @@ static port_read_func *port_readers[3] = {
 
 static int padTHLatency[3];
 static int padTLLatency[3];
+static int padTHTimeout[3];
 
 static NOINLINE u32 port_read(int i)
 {
   u32 data_reg = PicoMem.ioports[i + 1];
   u32 ctrl_reg = PicoMem.ioports[i + 4] | 0x80;
+  u32 cycles = SekCyclesDone();
   u32 in, out;
 
   out = data_reg & ctrl_reg;
 
+  // controllers have a timeout on TH, they reset if not toggled for some time
+  if (CYCLES_GE(cycles, padTHTimeout[i]) && i < 2) {
+    padTHTimeout[i] = cycles + 0x70000000;
+    Pico.m.padTHPhase[i] = 0;
+  }
+
   // pull-ups: should be 0x7f, but Decap Attack has a bug where it temp.
   // disables output before doing TH-low read, so emulate RC filter for TH.
   // Decap Attack reportedly doesn't work on Nomad but works on most
   // other MD revisions (different pull-up strength?).
   u32 mask = 0x3f;
-  if (CYCLES_GE(SekCyclesDone(), padTHLatency[i])) {
+  if (CYCLES_GE(cycles, padTHLatency[i])) {
     mask |= 0x40;
-    padTHLatency[i] = SekCyclesDone();
+    padTHLatency[i] = cycles;
   }
   out |= mask & ~ctrl_reg;
 
@@ -496,14 +535,48 @@ static NOINLINE u32 port_read(int i)
 
   // Sega mouse uses the TL/TR lines for req/ack. For buggy drivers, make sure
   // there's some delay before ack is sent by taking over the new TL line level
-  if (CYCLES_GE(SekCyclesDone(), padTLLatency[i]))
-    padTLLatency[i] = SekCyclesDone();
+  if (CYCLES_GE(cycles, padTLLatency[i]))
+    padTLLatency[i] = cycles;
   else
     in ^= 0x10; // TL
 
   return (in & ~ctrl_reg) | (data_reg & ctrl_reg);
 }
 
+// update ports
+void PicoPortUpdate(void)
+{
+  if (port_lightgun || (Pico.m.hardware & PMS_HW_LG)) {
+    PicoIn.mouseInt[0] += PicoIn.mouse[0] - PicoIn.mouseInt[2];
+    PicoIn.mouseInt[1] += PicoIn.mouse[1] - PicoIn.mouseInt[3];
+    PicoIn.mouseInt[2] = PicoIn.mouse[0];
+    PicoIn.mouseInt[3] = PicoIn.mouse[1];
+    if (PicoIn.mouseInt[0] < 0) PicoIn.mouseInt[0] = 0;
+    if (PicoIn.mouseInt[0] > 320 ) PicoIn.mouseInt[0] = 320;
+    if (PicoIn.mouseInt[1] < 0) PicoIn.mouseInt[1] = 0;
+    if (PicoIn.mouseInt[1] > 240) PicoIn.mouseInt[1] = 240;
+  }
+}
+
+void PicoPortTrigger(void)
+{
+  // TODO mx/my scaling for V28/V30, H32/H40, soft/hardscaling?
+  int mx = PicoIn.mouseInt[0] + PicoIn.gunx;
+  int my = PicoIn.mouseInt[1] + PicoIn.guny;
+
+  if (unlikely(Pico.m.scanline == my) && port_lightgun &&
+      ((PicoMem.ioports[4]|PicoMem.ioports[5]) & 0x80)) {
+    // there's a rather massive delay from VDP, TV, light sensor, infrared
+    // transmission, to the TH pin; it lasts well into the next line.
+    if (port_type[1] == PICO_INPUT_JUSTIFIER) {
+      // justifier should trigger only if gun enabled?
+      if (!(PicoMem.ioports[2] & 0x30))
+        PicoVideoTriggerTH(mx + 66, my);
+    } else
+      PicoVideoTriggerTH(mx + 136, my);
+  }
+}
+
 // pad export for J-Cart
 u32 PicoReadPad(int i, u32 out_bits)
 {
@@ -513,40 +586,31 @@ u32 PicoReadPad(int i, u32 out_bits)
 void PicoSetInputDevice(int port, enum input_device device)
 {
   port_read_func *func;
+  int is_lg = device == PICO_INPUT_LIGHT_GUN || device == PICO_INPUT_JUSTIFIER;
 
   if (port < 0 || port > 2)
     return;
 
-  if (port == 1 && port_readers[0] == read_pad_team)
-    func = read_nothing;
-
-  else switch (device) {
-  case PICO_INPUT_PAD_3BTN:
-    func = read_pad_3btn;
-    break;
-
-  case PICO_INPUT_PAD_6BTN:
-    func = read_pad_6btn;
-    break;
-
-  case PICO_INPUT_PAD_TEAM:
-    func = read_pad_team;
-    break;
-
-  case PICO_INPUT_PAD_4WAY:
-    func = read_pad_4way;
-    break;
-
-  case PICO_INPUT_MOUSE:
-    func = read_pad_mouse;
-    break;
-
-  default:
-    func = read_nothing;
-    break;
+  switch (device) {
+  case PICO_INPUT_PAD_3BTN:   func = read_pad_3btn; break;
+  case PICO_INPUT_PAD_6BTN:   func = read_pad_6btn; break;
+  case PICO_INPUT_PAD_TEAM:   func = read_pad_team; break;
+  case PICO_INPUT_PAD_4WAY:   func = read_pad_4way; break;
+  case PICO_INPUT_MOUSE:      func = read_pad_mouse; break;
+  case PICO_INPUT_LIGHT_GUN:  func = read_pad_menacer; break;
+  case PICO_INPUT_JUSTIFIER:  func = read_pad_justifier; break;
+  default:                    func = read_nothing; break;
   }
 
+  if (port == 1 && port_type[0] == PICO_INPUT_PAD_TEAM)
+    func = read_nothing;
+
+  port_lightgun &= ~(1<<port);
+  port_lightgun |= is_lg<<port;
+  port_type[port] = device;
   port_readers[port] = func;
+
+  padTLLatency[port] = padTHLatency[port] = padTHTimeout[port] = SekCyclesDone();
 }
 
 NOINLINE u32 io_ports_read(u32 a)
@@ -570,18 +634,22 @@ NOINLINE void io_ports_write(u32 a, u32 d)
   // for some controllers, changing TH/TR changes controller state
   if (1 <= a && a <= 2)
   {
-    Pico.m.padDelay[a - 1] = 0;
-    if (port_readers[a - 1] == read_pad_team) {
+    // there's a timeout for toggling the TH line
+    if (CYCLES_GE(SekCyclesDone(), padTHTimeout[a - 1]))
+      Pico.m.padTHPhase[a - 1] = 0;
+    padTHTimeout[a - 1] = SekCyclesDone() + 25*488;
+
+    if (port_type[a - 1] == PICO_INPUT_PAD_TEAM) {
       if (d & 0x40)
         Pico.m.padTHPhase[a - 1] = 0;
       else if ((d^PicoMem.ioports[a]) & 0x60)
         Pico.m.padTHPhase[a - 1]++;
-    } else if (port_readers[0] == read_pad_4way) {
+    } else if (port_type[0] == PICO_INPUT_PAD_4WAY) {
       if (a == 2 && ((PicoMem.ioports[a] ^ d) & 0x70))
         Pico.m.padTHPhase[0] = 0;
       if (a == 1 && !(PicoMem.ioports[a] & 0x40) && (d & 0x40))
         Pico.m.padTHPhase[0]++;
-    } else if (port_readers[a - 1] == read_pad_mouse) {
+    } else if (port_type[a - 1] == PICO_INPUT_MOUSE) {
       if ((d^PicoMem.ioports[a]) & 0x20) {
         if (Pico.m.padTHPhase[a - 1]) { // in readout?
           padTLLatency[a - 1] = SekCyclesDone() + 100;
@@ -597,7 +665,7 @@ NOINLINE void io_ports_write(u32 a, u32 d)
       Pico.m.padTHPhase[a - 1]++;
   }
 
-  // after switching TH to input there's a latency before the pullup value is 
+  // after switching TH to input there's a latency before the pullup value is
   // read back as input (see Decap Attack, not in Samurai Showdown, 32x WWF Raw)
   if (4 <= a && a <= 5) {
     if ((PicoMem.ioports[a] & 0x40) && !(d & 0x40) && !(PicoMem.ioports[a - 3] & 0x40))
@@ -609,6 +677,14 @@ NOINLINE void io_ports_write(u32 a, u32 d)
   PicoMem.ioports[a] = d;
 }
 
+void io_ports_reset(void)
+{
+  int port;
+
+  for (port = 0; port < 3; port++)
+    padTLLatency[port] = padTHLatency[port] = padTHTimeout[port] = SekCyclesDone();
+}
+
 int io_ports_pack(void *buf, size_t size)
 {
   size_t b, i;
@@ -625,6 +701,8 @@ int io_ports_pack(void *buf, size_t size)
     save_u16(buf, &b, PicoIn.padInt[i]);
   for (i = 0; i < 4; i++)
     save_s32(buf, &b, PicoIn.mouseInt[i]);
+  for (i = 0; i < ARRAY_SIZE(padTHTimeout); i++)
+    save_u32(buf, &b, padTHTimeout[i]);
   assert(b <= size);
   return b;
 }
@@ -645,6 +723,15 @@ void io_ports_unpack(const void *buf, size_t size)
     PicoIn.padInt[i] = load_u16(buf, &b);
   for (i = 0; i < 4; i++)
     PicoIn.mouseInt[i] = load_s32(buf, &b);
+
+  // convert pad delay from line count to samples
+  for (i = 0; i < ARRAY_SIZE(padTHTimeout); i++) {
+    if (b < size)
+      padTHTimeout[i] = load_u32(buf, &b);
+    else
+      padTHTimeout[i] = SekCyclesDone() - (25-Pico.m.padDelay[i]) * 488;
+  }
+
   assert(b <= size);
 }
 
@@ -675,7 +762,7 @@ void NOINLINE ctl_write_z80busreq(u32 d)
         pprof_start(m68k);
         PicoSyncZ80(SekCyclesDone());
         pprof_end_sub(m68k);
-        granted = Pico.t.z80c_aim + 6; // M cycle is 3-6 cycles 
+        granted = Pico.t.z80c_aim + 6; // M cycle is 3-6 cycles
         Pico.t.z80_busdelay += (Pico.t.z80c_cnt - granted) << 8;
         Pico.t.z80c_cnt = granted;
       }
@@ -746,7 +833,7 @@ static u32 PicoRead8_sram(u32 a)
   // XXX: this is banking unfriendly
   if (a < Pico.romsize)
     return Pico.rom[MEM_BE2(a)];
-  
+ 
   return m68k_unmapped_read8(a);
 }
 
@@ -835,7 +922,7 @@ static u32 PicoRead8_z80(u32 a)
   if ((a & 0x4000) == 0x0000) {
     d = PicoMem.zram[a & 0x1fff];
   } else if ((a & 0x6000) == 0x4000) // 0x4000-0x5fff
-    d = ym2612_read_local_68k(); 
+    d = ym2612_read_local_68k();
   else {
     elprintf(EL_UIO|EL_ANOMALY, "68k bad read [%06x] @%06x", a, SekPc);
     d = (u8)PicoRead16_floating(a);
@@ -863,8 +950,7 @@ static void PicoWrite8_z80(u32 a, u32 d)
     return;
   }
   if ((a & 0x6000) == 0x4000) { // FM Sound
-    if (PicoIn.opt & POPT_EN_FM)
-      ym2612_write_local(a & 3, d & 0xff, 0);
+    ym2612_write_local(a & 3, d & 0xff, 0);
     return;
   }
   // TODO: probably other VDP access too? Maybe more mirrors?
@@ -1250,7 +1336,7 @@ void ym2612_sync_timers(int z80_cycles, int mode_old, int mode_new)
        * Starting a timer takes place at the next tick, so xcycles needs to be
        * rounded up to that: t = next tick# = (xcycles / TICK_ZCYCLES) + 1
        */
-      unsigned t = ((xcycles * (((1LL<<32)/TIMER_A_TICK_ZCYCLES)+1))>>32) + 1;
+      unsigned t = DIVQ32(xcycles, TIMER_A_TICK_ZCYCLES) + 1;
       Pico.t.timer_a_next_oflow = t*TIMER_A_TICK_ZCYCLES + Pico.t.timer_a_step;
     }
   }
@@ -1273,7 +1359,7 @@ void ym2612_sync_timers(int z80_cycles, int mode_old, int mode_new)
        * reset by loading timer b. The first run of timer b after loading is
        * therefore shorter by up to 15 ticks.
        */
-      unsigned t = ((xcycles * (((1LL<<32)/TIMER_A_TICK_ZCYCLES)+1))>>32) + 1;
+      unsigned t = DIVQ32(xcycles, TIMER_A_TICK_ZCYCLES) + 1;
       int step = Pico.t.timer_b_step - TIMER_A_TICK_ZCYCLES*(t&15);
       Pico.t.timer_b_next_oflow = t*TIMER_A_TICK_ZCYCLES + step;
     }
@@ -1448,9 +1534,9 @@ int ym2612_pack_timers(void *buf, size_t size)
   if (Pico.t.ym2612_busy > 0)
     busy = cycles_z80_to_68k(Pico.t.ym2612_busy);
   if (Pico.t.timer_a_next_oflow != TIMER_NO_OFLOW)
-    tat = ((Pico.t.timer_a_step - Pico.t.timer_a_next_oflow) * ((1LL<<32)/TIMER_A_TICK_ZCYCLES+1))>>16;
+    tat = MULQ((Pico.t.timer_a_step - Pico.t.timer_a_next_oflow), INVQ32(TIMER_A_TICK_ZCYCLES), 16);
   if (Pico.t.timer_b_next_oflow != TIMER_NO_OFLOW)
-    tbt = ((Pico.t.timer_b_step - Pico.t.timer_b_next_oflow) * ((1LL<<32)/TIMER_B_TICK_ZCYCLES+1))>>16;
+    tbt = MULQ((Pico.t.timer_b_step - Pico.t.timer_b_next_oflow), INVQ32(TIMER_B_TICK_ZCYCLES), 16);
   elprintf(EL_YMTIMER, "save: timer a %i/%i", tat >> 16, tac);
   elprintf(EL_YMTIMER, "save: timer b %i/%i", tbt >> 16, tbc);
 
@@ -1609,8 +1695,7 @@ static unsigned char z80_md_bank_read(unsigned short a)
 
 static void z80_md_ym2612_write(unsigned int a, unsigned char data)
 {
-  if (PicoIn.opt & POPT_EN_FM)
-    ym2612_write_local(a, data, 1);
+  ym2612_write_local(a, data, 1);
 }
 
 static void z80_md_vdp_br_write(unsigned int a, unsigned char data)

pico/pico.c

@@ -249,6 +249,8 @@ void PicoLoopPrepare(void)
     PicoMCDPrepare();
   if (PicoIn.AHW & PAHW_32X)
     Pico32xPrepare();
+  if (PicoIn.AHW & PAHW_SMS)
+    PicoPrepareMS();
 }
 
 #include "pico_cmn.c"

pico/pico.h

@@ -78,7 +78,6 @@ extern void *p32x_bios_g, *p32x_bios_m, *p32x_bios_s;
 #define POPT_EN_FM_DAC      (1<<24) //x00 0000
 #define POPT_EN_FM_FILTER   (1<<25)
 #define POPT_EN_KBD         (1<<26)
-#define POPT_EN_MOUSE       (1<<27)
 
 #define PAHW_MCD    (1<<0)
 #define PAHW_32X    (1<<1)
@@ -114,8 +113,9 @@ typedef struct PicoInterface
 	unsigned short AHW;            // active addon hardware: PAHW_* bitfield
 
 	unsigned short kbd;            // SC-3000 or Pico Keyboard
-	int mouse[4];                  // x,y mouse coordinates
-	int mouseInt[4];               // internal copy
+	short mouse[4];                // x,y mouse coordinates
+	short mouseInt[4];             // internal copy
+	short gunx, guny;              // light gun offsets
 
 	unsigned short quirks;         // game-specific quirks: PQUIRK_*
 	unsigned short overclockM68k;  // overclock the emulated 68k, in %
@@ -364,8 +364,11 @@ enum input_device {
   PICO_INPUT_PAD_3BTN,
   PICO_INPUT_PAD_6BTN,
   PICO_INPUT_MOUSE,
+  PICO_INPUT_LIGHT_GUN,
+  PICO_INPUT_JUSTIFIER,
   PICO_INPUT_PAD_TEAM,
   PICO_INPUT_PAD_4WAY,
+  PICO_INPUT_COUNT
 };
 void PicoSetInputDevice(int port, enum input_device device);
 

pico/pico_cmn.c

@@ -10,12 +10,6 @@
 #define CYCLES_M68K_LINE     488 // suitable for both PAL/NTSC
 #define CYCLES_M68K_VINT_LAG 112
 
-// pad delay (for 6 button pads)
-#define PAD_DELAY() { \
-  if(Pico.m.padDelay[0]++ > 25) Pico.m.padTHPhase[0]=0; \
-  if(Pico.m.padDelay[1]++ > 25) Pico.m.padTHPhase[1]=0; \
-}
-
 // CPUS_RUN
 #ifndef CPUS_RUN
 #define CPUS_RUN(m68k_cycles) \
@@ -146,6 +140,7 @@ static int PicoFrameHints(void)
 
   Pico.t.m68c_frame_start = Pico.t.m68c_aim;
   PsndStartFrame();
+  PicoPortUpdate();
 
   hint = pv->hint_cnt;
 
@@ -160,7 +155,7 @@ static int PicoFrameHints(void)
     Pico.m.scanline = y;
     pv->v_counter = PicoVideoGetV(y, 0);
 
-    PAD_DELAY();
+    PicoPortTick();
 
     // H-Interrupts:
     if (--hint < 0)
@@ -217,7 +212,7 @@ static int PicoFrameHints(void)
   pv->v_counter = PicoVideoGetV(y, 0);
 
   memcpy(PicoIn.padInt, PicoIn.pad, sizeof(PicoIn.padInt));
-  PAD_DELAY();
+  PicoPortTick();
 
   // Last H-Int (normally):
   if (--hint < 0)
@@ -288,7 +283,7 @@ static int PicoFrameHints(void)
     Pico.m.scanline = y;
     pv->v_counter = PicoVideoGetV(y, 1);
 
-    PAD_DELAY();
+    PicoPortTick();
 
     if (unlikely(pv->status & PVS_ACTIVE) && --hint < 0)
     {
@@ -324,7 +319,7 @@ static int PicoFrameHints(void)
   Pico.m.scanline = y++;
   pv->v_counter = 0xff;
 
-  PAD_DELAY();
+  PicoPortTick();
 
   if (unlikely(pv->status & PVS_ACTIVE)) {
     if (--hint < 0) {
@@ -361,7 +356,6 @@ static int PicoFrameHints(void)
   return 0;
 }
 
-#undef PAD_DELAY
 #undef CPUS_RUN
 
 // vim:shiftwidth=2:ts=2:expandtab

pico/pico_int.h

@@ -302,6 +302,7 @@ extern SH2 sh2s[2];
 #define PVS_DMAFILL   (1 << 20) // DMA fill is waiting for fill data
 #define PVS_DMABG     (1 << 21) // background DMA operation is running
 #define PVS_FIFORUN   (1 << 22) // FIFO is processing
+#define PVS_HVLATCH   (1 << 23) // next hv read is latched (for lightguns)
 
 struct PicoVideo
 {
@@ -354,6 +355,7 @@ struct PicoMisc
 #define PMS_HW_FM	0x8   // FM sound
 #define PMS_HW_TMS	0x10  // assume TMS9918
 #define PMS_HW_3D	0x20  // 3D glasses
+#define PMS_HW_LG	0x40  // light phaser
 #define PMS_HW_FMUSED	0x80  // FM sound accessed
 
 #define PMS_MAP_AUTO	0
@@ -654,9 +656,7 @@ struct Pico32x
   unsigned int pad[4];
   unsigned int dmac0_fifo_ptr;
   unsigned short vdp_fbcr_fake;
-  unsigned short pad2;
-  unsigned char comm_dirty;
-  unsigned char pad3;            // was comm_dirty_sh2
+  unsigned short wdt_cycle[2];   // wdt clocking (was comm_dirty)
   unsigned char pwm_irq_cnt;
   unsigned char pad1;
   unsigned short pwm_p[2];       // pwm pos in fifo
@@ -757,9 +757,17 @@ u32 PicoRead8_io(u32 a);
 u32 PicoRead16_io(u32 a);
 void PicoWrite8_io(u32 a, u32 d);
 void PicoWrite16_io(u32 a, u32 d);
+void PicoPortUpdate(void);
+void PicoPortTrigger(void);
 u32 PicoReadPad(int i, u32 mask);
+void io_ports_reset(void);
 int io_ports_pack(void *buf, size_t size);
 void io_ports_unpack(const void *buf, size_t size);
+extern int port_type[3];
+extern int port_lightgun;
+
+#define PicoPortTick() if (port_lightgun && \
+            Pico.m.scanline == PicoIn.mouseInt[1]+PicoIn.guny) PicoPortTrigger()
 
 // pico/memory.c
 PICO_INTERNAL void PicoMemSetupPico(void);
@@ -977,6 +985,7 @@ void PicoVideoFIFOMode(int active, int h40);
 int PicoVideoFIFOWrite(int count, int byte_p, unsigned sr_mask, unsigned sr_flags);
 void PicoVideoInit(void);
 void PicoVideoReset(void);
+void PicoVideoTriggerTH(int x, int y);
 void PicoVideoSync(int skip);
 int PicoVideoSave(void *buf);
 void PicoVideoLoad(void *buf, int len);
@@ -1023,6 +1032,7 @@ void PicoPowerMS(void);
 void PicoResetMS(void);
 void PicoMemSetupMS(void);
 void PicoStateLoadedMS(void);
+void PicoPrepareMS(void);
 void PicoFrameMS(void);
 void PicoFrameDrawOnlyMS(void);
 int PicoPlayTape(const char *fname);
@@ -1047,6 +1057,9 @@ enum p32x_event {
   P32X_EVENT_PWM,
   P32X_EVENT_FILLEND,
   P32X_EVENT_HINT,
+  P32X_EVENT_VINT,
+  P32X_EVENT_MTIMER,
+  P32X_EVENT_STIMER,
   P32X_EVENT_COUNT,
 };
 extern unsigned int p32x_event_times[P32X_EVENT_COUNT];
@@ -1070,6 +1083,7 @@ void p32x_reset_sh2s(void);
 void p32x_event_schedule(unsigned int now, enum p32x_event event, int after);
 void p32x_event_schedule_sh2(SH2 *sh2, enum p32x_event event, int after);
 void p32x_schedule_hint(SH2 *sh2, unsigned int m68k_cycles);
+void p32x_schedule_vint(SH2 *sh2, unsigned int m68k_cycles);
 
 #define p32x_sh2_ready(sh2, cycles) \
   (CYCLES_GT(cycles,sh2->m68krcycles_done) && \
@@ -1124,9 +1138,11 @@ void p32x_pwm_state_loaded(void);
 // 32x/sh2soc.c
 void p32x_dreq0_trigger(void);
 void p32x_dreq1_trigger(void);
-void p32x_timers_recalc(void);
-void p32x_timer_do(SH2 *sh2, unsigned int m68k_slice);
+void p32x_timer_recalc(SH2 *sh2);
+void p32x_timer_do(SH2 *sh2, unsigned int now);
+void p32x_timer_irq(SH2 *sh2, unsigned int now);
 void sh2_peripheral_reset(SH2 *sh2);
+void sh2_peripheral_state_loaded(void);
 u32 REGPARM(2) sh2_peripheral_read8(u32 a, SH2 *sh2);
 u32 REGPARM(2) sh2_peripheral_read16(u32 a, SH2 *sh2);
 u32 REGPARM(2) sh2_peripheral_read32(u32 a, SH2 *sh2);

pico/pico_port.h

@@ -80,4 +80,12 @@
 			(((v)<<8)&0xff0000)|(u32)((v)<<24))
 #endif
 
+#define INVQ(n,m)      (long long)((1LL<<(m))/(n)+1) // to replace div by mult with inverse
+#define MULQ(n,d,m)    ((n)*(d) >> (m))
+#define DIVQ(n,d,m)    ((n)*INVQ32(d) >> (m))
+
+#define INVQ32(n)      INVQ(n,32)
+#define MULQ32(n,d)    MULQ(n,d,32)
+#define DIVQ32(n,d)    DIVQ(n,d,32)
+
 #endif // PICO_PORT_INCLUDED

pico/sms.c

@@ -155,8 +155,8 @@ static u8 vdp_hcounter(int cycles)
 {
   // 171 slots per scanline of 228 clocks, counted 0xf4-0x93, 0xe9-0xf3
   // this matches h counter tables in SMSVDPTest:
-  //  hc =   (cycles+2) *   171    /228      -1 + 0xf4;
-  int hc = (((cycles+2) * ((171<<8)/228))>>8)-1 + 0xf4; // Q8 to avoid dividing
+  //  hc =        (cycles+2) * 171 /228  -1 + 0xf4;
+  int hc = DIVQ32((cycles+2) * 171, 228) -1 + 0xf4;
   if (hc > 0x193) hc += 0xe9-0x93-1;
   return hc;
 }
@@ -503,6 +503,20 @@ static unsigned char z80_sms_in(unsigned short a)
             d = (d & ~0x08) | ((Pico.ms.io_ctl >> 3) & 0x08);
           if (Pico.ms.io_ctl & 0x08) d |= 0x80; // TH as input is unconnected
           if (Pico.ms.io_ctl & 0x02) d |= 0x40;
+          if (Pico.m.hardware & PMS_HW_LG) { // light phaser
+            // TODO mx/my scaling is wrong if V28/V30 mode is used?
+            int mx = DIVQ32((PicoIn.mouseInt[0]+PicoIn.gunx) * 256, 320);
+            int my = DIVQ32((PicoIn.mouseInt[1]+PicoIn.guny) * 192, 224);
+            int x = vdp_hcounter(z80_cyclesDone() - Pico.t.z80c_line_start);
+            int dx = 2*x - mx;
+            int dy = Pico.m.scanline - my;
+            int th = 0xc0; // TODO set bits according to port usage
+            d |= th; // TH input, high if no light detected
+            if (dy > -4 && dy < 4 && dx > -40 && dx < 40) {
+              d &= ~th; // TH falling -> save hcounter
+              Pico.ms.vdp_hlatch = (mx >> 1) + 24;
+            }
+          }
         } else {
           int i; // read kbd 4 bits
           kbd_update();
@@ -1223,6 +1237,13 @@ void PicoStateLoadedMS(void)
   memcpy(Pico.ms.carthw, carthw, 16);
 }
 
+void PicoPrepareMS(void)
+{
+  Pico.m.hardware &= ~PMS_HW_LG;
+  if (port_type[0] == PICO_INPUT_LIGHT_GUN || port_type[1] == PICO_INPUT_LIGHT_GUN)
+    Pico.m.hardware |= PMS_HW_LG;
+}
+
 void PicoFrameMS(void)
 {
   struct PicoVideo *pv = &Pico.video;
@@ -1236,6 +1257,7 @@ void PicoFrameMS(void)
   int y;
 
   PsndStartFrame();
+  PicoPortUpdate();
 
   // for SMS the pause button generates an NMI, for GG ths is not the case
   nmi = (PicoIn.pad[0] >> 7) & 1;
@@ -1367,7 +1389,7 @@ int PicoPlayTape(const char *fname)
   }
 
   pt->cycles_sample = (Pico.m.pal ? OSC_PAL/15 : OSC_NTSC/15) / rate;
-  pt->cycles_mult = (1LL<<32) / pt->cycles_sample;
+  pt->cycles_mult = INVQ32(pt->cycles_sample);
   pt->cycle = Pico.t.z80c_aim;
   pt->phase = Pico.t.z80c_aim;
   pt->pause = 0;
@@ -1410,7 +1432,7 @@ int PicoRecordTape(const char *fname)
   }
 
   pt->cycles_sample = (Pico.m.pal ? OSC_PAL/15 : OSC_NTSC/15) / rate;
-  pt->cycles_mult = (1LL<<32) / pt->cycles_sample;
+  pt->cycles_mult = INVQ32(pt->cycles_sample);
   pt->cycle = Pico.t.z80c_aim;
   pt->phase = Pico.t.z80c_aim;
   pt->pause = 0;

pico/state.c

@@ -429,7 +429,7 @@ static int state_load(void *file)
   int len_check;
   int retval = -1;
   char header[8];
-  int ver, has_32x = 0;
+  int ver, has_32x = 0, has_iov2 = 0;
   int len, len_vdp = 0;
 
   memset(buff_m68k, 0, sizeof(buff_m68k));
@@ -483,6 +483,13 @@ static int state_load(void *file)
       case CHUNK_VDP:     CHECKED_READ2((len_vdp = len), buff_vdp); break;
 
       case CHUNK_IOPORTS: CHECKED_READ_BUFF(PicoMem.ioports); break;
+      case CHUNK_IOPORTSv2:
+        CHECKED_READ(len, buf);
+        io_ports_unpack(buf, len);
+        has_iov2 = 1;
+        break;
+
+
       case CHUNK_PSG:     CHECKED_READ2(28*4, sn76496_regs); break;
       case CHUNK_YM2413:
         CHECKED_READ(len, buf);
@@ -496,7 +503,6 @@ static int state_load(void *file)
         break;
       case CHUNK_FM_TIMERS: CHECKED_READ(len, buf); ym2612_unpack_timers(buf, len); break;
       case CHUNK_FMv3:      CHECKED_READ(len, buf); YM2612PicoStateLoad3(buf, len); break;
-      case CHUNK_IOPORTSv2: CHECKED_READ(len, buf); io_ports_unpack(buf, len); break;
 
       case CHUNK_PICO_PCM:
         CHECKED_READ(len, buf);
@@ -641,6 +647,10 @@ readend:
     Pico.video.status |= (Pico.video.pending_ints << 2) & SR_F;
   }
 
+  // kludge: if there is no new ports data save, they must be initialised
+  if (!has_iov2)
+    io_ports_reset();
+
   Pico.m.dirtyPal = 1;
   retval = 0;
 

pico/videoport.c

@@ -405,7 +405,7 @@ int PicoVideoFIFOHint(void)
   // only need to refresh sprite position if we are synced
   if (Pico.est.DrawScanline == Pico.m.scanline && !(pv->status & SR_VB))
     PicoDrawRefreshSprites();
- 
+
   // if CPU is waiting for the bus, advance CPU and FIFO until bus is free
   if (pv->status & PVS_CPUWR)
     burn = PicoVideoFIFODrain(4, lc, 0);
@@ -603,7 +603,7 @@ static void DmaSlow(int len, u32 source)
           !((source ^ (source + len-1)) & ~mask))
       {
         // most used DMA mode
-        memcpy((char *)r + a, base + (source & mask), len * 2);
+        memcpy((char *)r + (u16)a, base + (source & mask), len * 2);
         a += len * 2;
         break;
       }
@@ -1120,10 +1120,28 @@ PICO_INTERNAL_ASM u32 PicoVideoRead(u32 a)
     unsigned int c;
     u32 d;
 
+    // invalidate cached value if on a new scanline
+    if ((pv->status & PVS_HVLATCH) && pv->hv_latch >> 8 != Pico.m.scanline)
+      pv->status &= ~PVS_HVLATCH;
+
     c = SekCyclesDone() - Pico.t.m68c_line_start;
-    if (pv->reg[0]&2)
-         d = pv->hv_latch;
-    else d = VdpFIFO.fifo_hcounts[c/clkdiv] | (pv->v_counter << 8);
+    if (pv->reg[0]&2) // latched value from TH transition
+      d = pv->hv_latch;
+    else if (pv->status & PVS_HVLATCH) { // cached value from light gun
+      c += 28; // estimate for 68k insn time
+      // return cached value, but add in irq latency; since the irq was fired
+      // at the start of the line, the latency in cycles already is in c.
+      d = pv->hv_latch & 0xff;
+      if (pv->reg[12]&1) { // H40/H32?
+        c = DIVQ32(c * slots40, slcpu); // slots since line start
+        if (d < gapstart40 && d+c >= gapstart40) d += gapend40-gapstart40;
+      } else {
+        c = DIVQ32(c * slots32, slcpu);
+        if (d < gapstart32 && d+c >= gapstart32) d += gapend32-gapstart32;
+      }
+      d = (pv->hv_latch & 0xff00) | ((d+c) & 0x00ff);
+    } else // freerunning
+      d = VdpFIFO.fifo_hcounts[c/clkdiv] | (pv->v_counter << 8);
 
     elprintf(EL_HVCNT, "hv: %02x %02x [%u] @ %06x", d, pv->v_counter, SekCyclesDone(), SekPc);
     return d;
@@ -1193,16 +1211,45 @@ unsigned char PicoVideoRead8HV_L(int is_from_z80)
 
 void PicoVideoReset(void)
 {
-  Pico.video.pending_ints=0;
-  Pico.video.reg[1] &= ~0x40; // TODO verify display disabled after reset
-  Pico.video.reg[10] = 0xff; // HINT is turned off after reset
-  Pico.video.status = 0x3428 | Pico.m.pal; // 'always set' bits | vblank | collision | pal
+  struct PicoVideo *pv = &Pico.video;
+
+  pv->pending_ints=0;
+  pv->reg[1] &= ~0x40; // TODO verify display disabled after reset
+  pv->reg[10] = 0xff; // HINT is turned off after reset
+  pv->status = 0x3428 | Pico.m.pal; // 'always set' bits | vblank | collision | pal
 
   memset(&VdpFIFO, 0, sizeof(VdpFIFO));
   Pico.m.dirtyPal = 1;
 
   PicoDrawBgcDMA(NULL, 0, 0, 0, 0);
-  PicoVideoFIFOMode(Pico.video.reg[1]&0x40, Pico.video.reg[12]&1);
+  PicoVideoFIFOMode(pv->reg[1]&0x40, pv->reg[12]&1);
+}
+
+// Process triggering of the !HL pin at pixel x,y of screen
+void PicoVideoTriggerTH(int x, int y)
+{
+  struct PicoVideo *pv = &Pico.video;
+  int slot;
+
+  // TODO x/y scaling for V28/V30, H32/H40, soft/hardscaling?
+  if (!(pv->reg[12]&1)) { // H32, convert from 320 to 256 px per line
+    slot = DIVQ32(x * 256, 320) / 2;
+    if (slot >= gapstart32) slot += gapend32-gapstart32;
+  } else {
+    slot = x / 2;
+    if (slot >= gapstart40) slot += gapend40-gapstart40;
+  }
+
+  // This is only called at the start of a scanline, hence the irq isn't
+  // cycle correct. There are games reading hv unlatched in the irq handler,
+  // so latch mouse here, and fix the value up when the irq handler reads it.
+  pv->hv_latch = (slot & 0xff) | (y << 8);
+  pv->status |= PVS_HVLATCH;
+
+  // irq here for speed reasons. It's early, but that shouldn't matter much.
+  // TODO is irq really lost if another irq is already pending?
+  if ((Pico.video.reg[11]&8) && SekIrqLevel < 2)
+    SekInterrupt(2);
 }
 
 void PicoVideoCacheSAT(int load)

platform/base_readme.txt

@@ -101,6 +101,7 @@ through the D-pad if the screen has been switched to either pages or pad, or by
 mouse if physical mouse support is activated.
 
 
+#ifdef GENERIC
 How to use the mouse
 --------------------
 
@@ -120,11 +121,44 @@ The physical mouse buttons are mapped as follows:
 
 | Physical Mouse | Sega Mouse |   Pico Pen  |
 |:--------------:|:----------:|:-----------:|
-|   Left Button  |      B     |  Pen Button |
+|  Left Button   |      B     |  Pen Button |
 |  Middle Button |    START   |  Red Button |
 |  Right Button  |      C     | Pen Up/Down |
 
 
+How to use a light gun
+----------------------
+
+PicoDrive additionally supports light guns on systems where mouse support is
+available. It can be activated by selecting `light gun` or `justifier` as the
+input device. General mouse operation is covered in the "How to use the mouse"
+section.
+
+The `light gun` setting auto configures an input device based on the currently
+emulated platform: a Sega Menacer for the Mega Drive (and derivatives) and a
+Light Phaser for the Master System. The `justifier` setting corresponds to a
+Konami Justifier. This is only available on Mega Drive based systems, and can
+only be configured as input device 2.
+
+Selecting any light gun as input enables the light gun configuration submenu.
+This allows for displaying a cursor to indicate the current aiming position,
+as well as the adjustment of an aiming offset to correct for any in-game
+displacements.
+
+The physical mouse buttons are mapped as follows:
+
+| Physical Mouse | Sega Menacer/    |
+|                | Light Phaser/    |
+|                | Konami Justifier |
+|:--------------:|:----------------:|
+|  Left Button   |         A        |
+|  Middle Button |         B        |
+|  Right Button  |       START      |
+
+The pad button mappings may also be used to access other buttons.
+
+
+#endif
 How to use keyboard input
 -------------------------
 

platform/common/config_file.c

@@ -275,6 +275,7 @@ int config_readlrom(const char *fname)
 static int custom_read(menu_entry *me, const char *var, const char *val)
 {
 	char *tmp;
+	int i;
 
 	switch (me->id)
 	{
@@ -364,7 +365,7 @@ static int custom_read(menu_entry *me, const char *var, const char *val)
 			currentConfig.max_skip = atoi(val);
 			return 1;
 
-		case MA_CTRL_KEYBOARD:
+		case MA_OPT_KEYBOARD:
 			currentConfig.keyboard = 0;
 			if (strcasecmp(val, "physical") == 0)
 				currentConfig.keyboard = 2;
@@ -372,6 +373,17 @@ static int custom_read(menu_entry *me, const char *var, const char *val)
 				currentConfig.keyboard = 1;
 			return 1;
 
+		case MA_OPT_INPUT_DEV0:
+			for (i = 0; indev_names[i]; i++)
+				if (strcasecmp(val, indev_names[i]) == 0)
+					currentConfig.input_dev0 = i;
+			return 1;
+		case MA_OPT_INPUT_DEV1:
+			for (i = 0; indev_names[i]; i++)
+				if (strcasecmp(val, indev_names[i]) == 0)
+					currentConfig.input_dev1 = i;
+			return 1;
+
 		/* PSP */
 		case MA_OPT3_VSYNC:
 			// XXX: use enum

platform/common/emu.c

@@ -37,10 +37,8 @@
 
 #ifndef _WIN32
 #define PATH_SEP      "/"
-#define PATH_SEP_C    '/'
 #else
 #define PATH_SEP      "\\"
-#define PATH_SEP_C    '\\'
 #endif
 
 #define STATUS_MSG_TIMEOUT 2000
@@ -134,7 +132,7 @@ static void fname_ext(char *dst, int dstlen, const char *prefix, const char *ext
 	}
 
 	p = fname + strlen(fname) - 1;
-	for (; p >= fname && *p != PATH_SEP_C; p--)
+	for (; p >= fname && *p != *PATH_SEP; p--)
 		;
 	p++;
 	strncpy(dst + prefix_len, p, dstlen - prefix_len - 1);
@@ -745,6 +743,8 @@ int emu_read_config(const char *rom_fname, int no_defaults)
 
 	pemu_validate_config();
 	PicoIn.overclockM68k = currentConfig.overclock_68k;
+	PicoIn.gunx = currentConfig.gunx;
+	PicoIn.guny = currentConfig.guny;
 
 	// some sanity checks
 	if (currentConfig.volume < 0 || currentConfig.volume > 99)
@@ -1202,7 +1202,7 @@ void run_events_pico(unsigned int events)
 		return;
 
 	/* handle other input modes using the pen */
-	if (PicoIn.opt & POPT_EN_MOUSE) {
+	if (currentConfig.EmuOpt & EOPT_MOUSE) {
 		pico_pen_x = PicoIn.mouse[0];
 		pico_pen_y = PicoIn.mouse[1];
 	} else {
@@ -1324,12 +1324,12 @@ static void run_events_ui(unsigned int which)
 	}
 	if (which & PEV_GRAB_INPUT)
 	{
-		if (PicoIn.opt & POPT_EN_MOUSE) {
+		if (currentConfig.EmuOpt & EOPT_MOUSE) {
 			grab_mode = !grab_mode;
-			in_update_analog(0, 2, &mouse_x);
-			in_update_analog(0, 3, &mouse_y);
-			in_update_analog(0, 0, &mouse_x);
-			in_update_analog(0, 1, &mouse_y);
+			in_update_pointer(0, 2, &mouse_x);
+			in_update_pointer(0, 3, &mouse_y);
+			in_update_pointer(0, 0, &mouse_x);
+			in_update_pointer(0, 1, &mouse_y);
 			emu_status_msg("Mouse capture %s", grab_mode ? "on" : "off");
 		} else {
 			grab_mode = 0;
@@ -1356,13 +1356,34 @@ static void run_events_ui(unsigned int which)
 		engineState = PGS_Menu;
 }
 
+static int map_pointer_buttons(int msbtns, int device)
+{
+	int buttons = 0;
+
+	if ((PicoIn.AHW & PAHW_PICO) && device == PICO_INPUT_MOUSE) {
+		if (msbtns & 1) buttons |= 1<<GBTN_C;	// pen button
+		if (msbtns & 2) buttons |= 1<<GBTN_B;	// red button
+		if (msbtns & 4) buttons |= 1<<GBTN_START; // pen up/down
+	} else if ((PicoIn.AHW & PAHW_SMS) || device == PICO_INPUT_MOUSE) {
+		if (msbtns & 1) buttons |= 1<<GBTN_B;	// as Sega Mouse
+		if (msbtns & 2) buttons |= 1<<GBTN_START;
+		if (msbtns & 4) buttons |= 1<<GBTN_C;
+	} else if (device == PICO_INPUT_LIGHT_GUN || device == PICO_INPUT_JUSTIFIER) {
+		if (msbtns & 1) buttons |= 1<<GBTN_A;	// as Sega Menacer
+		if (msbtns & 2) buttons |= 1<<GBTN_B;
+		if (msbtns & 4) buttons |= 1<<GBTN_START;
+	}
+
+	return buttons;
+}
+
 void emu_update_input(void)
 {
 	static int prev_events = 0;
 	int actions[IN_BINDTYPE_COUNT] = { 0, };
 	int actions_kbd[IN_BIND_LAST] = { 0, };
 	int pl_actions[4];
-	int count_kbd = 0, buttons = 0;
+	int count_kbd = 0;
 	int events, i = 0;
 
 	in_update(actions);
@@ -1375,39 +1396,23 @@ void emu_update_input(void)
 	events = actions[IN_BINDTYPE_EMU] & PEV_MASK;
 
 	// update mouse coordinates if there is an emulated mouse
-	if (PicoIn.opt & POPT_EN_MOUSE) {
+	if (currentConfig.EmuOpt & EOPT_MOUSE) {
 		if (!grab_mode) {
-			in_update_analog(0, 0, &PicoIn.mouse[0]);
-			in_update_analog(0, 1, &PicoIn.mouse[1]);
-			// scale mouse coordinates from -1024..1024 to 0..screen_w/h
-			PicoIn.mouse[0] = (PicoIn.mouse[0]+1024) * g_screen_width /2048;
-			PicoIn.mouse[1] = (PicoIn.mouse[1]+1024) * g_screen_height/2048;
+			in_update_pointer(0, 0, &mouse_x);
+			in_update_pointer(0, 1, &mouse_y);
 		} else {
 			int xrel, yrel;
-			in_update_analog(0, 2, &xrel);
-			in_update_analog(0, 3, &yrel);
+			in_update_pointer(0, 2, &xrel);
+			in_update_pointer(0, 3, &yrel);
 			mouse_x += xrel, mouse_y += yrel;
-			// scale mouse coordinates from -1024..1024 to 0..screen_w/h
-			PicoIn.mouse[0] = (mouse_x+1024) * g_screen_width /2048;
-			PicoIn.mouse[1] = (mouse_y+1024) * g_screen_height/2048;
 		}
+		// scale mouse coordinates from -1024..1024 to 0..screen_w/h
+		PicoIn.mouse[0] = (mouse_x+1024) * 320/2048;
+		PicoIn.mouse[1] = (mouse_y+1024) * 240/2048;
 
-		in_update_analog(0, -1, &i); // get mouse buttons, bit 2-0 = RML
-		if (PicoIn.AHW & PAHW_PICO) {
-			// TODO is maintaining 2 different mappings necessary?
-			if (i & 1) buttons |= 1<<GBTN_C;        // pen button
-			if (i & 2) buttons |= 1<<GBTN_B;	// red button
-			if (i & 4) buttons |= 1<<GBTN_START;    // pen up/down
-		} else {
-			if (i & 1) buttons |= 1<<GBTN_B;	// as Sega Mouse
-			if (i & 2) buttons |= 1<<GBTN_START;
-			if (i & 4) buttons |= 1<<GBTN_C;
-		}
-
-		if (currentConfig.input_dev0 == PICO_INPUT_MOUSE)
-			pl_actions[0] |= buttons;
-		if (currentConfig.input_dev1 == PICO_INPUT_MOUSE)
-			pl_actions[1] |= buttons;
+		in_update_pointer(0, -1, &i); // get mouse buttons, bit 2-0 = RML
+		pl_actions[0] |= map_pointer_buttons(i, currentConfig.input_dev0);
+		pl_actions[1] |= map_pointer_buttons(i, currentConfig.input_dev1);
 	}
 
 	if (kbd_mode) {
@@ -1619,8 +1624,10 @@ void emu_sound_wait(void)
 
 static void emu_loop_prep(void)
 {
+	static int pointer[] = { PICO_INPUT_MOUSE, PICO_INPUT_LIGHT_GUN, PICO_INPUT_JUSTIFIER };
 	static int pal_old = -1;
 	static int filter_old = -1;
+	int i;
 
 	if (currentConfig.CPUclock != plat_target_cpu_clock_get())
 		plat_target_cpu_clock_set(currentConfig.CPUclock);
@@ -1646,12 +1653,14 @@ static void emu_loop_prep(void)
 	if (((PicoIn.AHW & PAHW_PICO) || (PicoIn.AHW & PAHW_SC)) && currentConfig.keyboard)
 		PicoIn.opt |= POPT_EN_KBD;
 
-	PicoIn.opt &= ~POPT_EN_MOUSE;
-	if (!(PicoIn.AHW & PAHW_8BIT) && (currentConfig.input_dev0 == PICO_INPUT_MOUSE ||
-					currentConfig.input_dev1 == PICO_INPUT_MOUSE)) {
-		PicoIn.opt |= POPT_EN_MOUSE;
-		plat_grab_cursor(grab_mode);
-	}
+	currentConfig.EmuOpt &= ~EOPT_MOUSE;
+	for (i = 0; i < ARRAY_SIZE(pointer); i++)
+		if ((currentConfig.input_dev0 == pointer[i] ||
+					currentConfig.input_dev1 == pointer[i])) {
+			currentConfig.EmuOpt |= EOPT_MOUSE;
+			plat_grab_cursor(grab_mode);
+			break;
+		}
 
 	pemu_loop_prep();
 }
@@ -1828,7 +1837,7 @@ void emu_loop(void)
 			}
 			if (diff > target_frametime + vsync_delay) {
 				// still too fast
-				plat_wait_till_us(timestamp + (diff - target_frametime));
+				plat_wait_till_us(timestamp + (diff - target_frametime - vsync_delay));
 			}
 		}
 

platform/common/emu.h

@@ -35,6 +35,8 @@ extern int g_screen_ppitch; // pitch in pixels
 #define EOPT_WIZ_TEAR_FIX (1<<19)
 #define EOPT_EXT_FRMLIMIT (1<<20) // no internal frame limiter (limited by snd, etc)
 #define EOPT_PICO_PEN     (1<<21)
+#define EOPT_MOUSE        (1<<22)
+#define EOPT_GUN_CURSOR   (1<<23)
 
 enum {
 	EOPT_SCALE_NONE = 0,
@@ -94,6 +96,7 @@ typedef struct _currentConfig_t {
 	int ghosting;
 	int analog_deadzone;
 	int keyboard;
+	int gunx, guny;
 	int msh2_khz;
 	int ssh2_khz;
 	int overclock_68k;
@@ -196,6 +199,7 @@ void plat_finish(void);
 void plat_show_cursor(int on);
 int plat_grab_cursor(int on);
 int plat_has_wm(void);
+void plat_set_window(int w, int h);
 
 /* used before things blocking for a while (these funcs redraw on return) */
 void plat_status_msg_busy_first(const char *msg);

platform/common/menu_pico.c

@@ -544,13 +544,13 @@ int key_config_kbd_loop(int id, int keys)
 }
 
 
-const char *indev0_names[] = { "none", "3 button pad", "6 button pad", "Mouse", "Team player", "4 way play", NULL };
-const char *indev1_names[] = { "none", "3 button pad", "6 button pad", "Mouse", NULL };
+const char *indev_names[] = { "none", "3 button pad", "6 button pad", "Mouse", "Light gun", "Justifier", "Team player", "4 way play", NULL };
 
 static char h_play12[55];
-static char h_kbd[55];
 static char h_play34[] = "Works only for Mega Drive/CD/32X games having\n"
 				"support for Team player, 4 way play, or J-cart";
+static char h_gun[]    = "Light Phaser for SMS, Menacer for Mega Drive\n"
+				"note: Justifier can only be selected on input 2";
 
 static char player[] = "Player 1";
 
@@ -561,6 +561,7 @@ static const char *mgn_nothing(int id, int *offs)
 
 static int key_config_players(int id, int keys)
 {
+	int x = me_id2offset(e_menu_keyconfig, MA_CTRL_PLAYER1);
 	int pid = player[strlen(player)-1] - '0';
 
 	if (keys & PBTN_LEFT)
@@ -569,7 +570,7 @@ static int key_config_players(int id, int keys)
 		if (++pid > 4) pid = 1;
 
 	player[strlen(player)-1] = pid + '0';
-	e_menu_keyconfig[0].help = (pid >= 3 ? h_play34 : h_play12);
+	e_menu_keyconfig[x].help = (pid >= 3 ? h_play34 : h_play12);
 
 	if (keys & PBTN_MOK)
 		key_config_loop(me_ctrl_actions, array_size(me_ctrl_actions) - 1, pid-1);
@@ -579,10 +580,12 @@ static int key_config_players(int id, int keys)
 
 static const char *mgn_keyboard(int id, int *offs)
 {
-	static char *kbds[] = { "OFF", "virtual", "physical" };
+	static const char *kbds[] = { "OFF", "virtual", "physical" };
+	const char *kbd = kbds[currentConfig.keyboard];
 	if (currentConfig.keyboard < 0 || currentConfig.keyboard > 2)
-		return kbds[0];
-	return kbds[currentConfig.keyboard];
+		kbd = kbds[0];
+	*offs = 3-strlen(kbd);
+	return kbd;
 }
 
 static int key_config_keyboard(int id, int keys)
@@ -601,11 +604,85 @@ static int key_config_keyboard(int id, int keys)
 
 	currentConfig.keyboard = kid;
 
-	e_menu_keyconfig[2].help = (currentConfig.keyboard == 2 ? h_kbd : NULL);
+	me_enable(e_menu_keyconfig, MA_CTRL_KEYBOARD, currentConfig.keyboard == 2);
 
-	if (keys & PBTN_MOK)
-		if (currentConfig.keyboard == 2)
-			key_config_kbd_loop(MA_CTRL_KEYBOARD, 0);
+	return 0;
+}
+
+static const char *mgn_indev(int id, int *offs)
+{
+	const char *dev;
+	if (id == MA_OPT_INPUT_DEV0)
+		dev = indev_names[currentConfig.input_dev0];
+	else	dev = indev_names[currentConfig.input_dev1];
+	*offs = 3-strlen(dev);
+	return dev;
+}
+
+static int mh_indev(int id, int keys)
+{
+	int is0 = id == MA_OPT_INPUT_DEV0;
+	int *indev = is0 ? &currentConfig.input_dev0 : &currentConfig.input_dev1;
+	int x = me_id2offset(e_menu_keyconfig, MA_OPT_INPUT_DEV0);
+
+	if (keys & PBTN_RIGHT) {
+		(*indev) ++;
+#ifdef USE_SDL	// TODO this info should come from platform!
+		// Justifier only in pad port 2
+		if (is0 && *indev == PICO_INPUT_JUSTIFIER)
+			(*indev) ++;
+		// Team/4way only in pad port 1
+		if (!is0 && *indev > PICO_INPUT_JUSTIFIER)
+			(*indev) --;
+#else
+		while (*indev >= PICO_INPUT_MOUSE && *indev <= PICO_INPUT_JUSTIFIER)
+			(*indev) ++;
+		// Team/4way only in pad port 1
+		while (!is0 && *indev >= PICO_INPUT_MOUSE)
+			(*indev) --;
+#endif
+		if (*indev >= PICO_INPUT_COUNT)
+			(*indev) --;
+	}
+	if (keys & PBTN_LEFT) {
+		(*indev) --;
+#ifdef USE_SDL	// TODO this info should come from platform!
+		if (is0 && *indev == PICO_INPUT_JUSTIFIER)
+			(*indev) --;
+#else
+		while (*indev >= PICO_INPUT_MOUSE && *indev <= PICO_INPUT_JUSTIFIER)
+			(*indev) --;
+#endif
+		if (*indev < 0)
+			(*indev) ++;
+	}
+	if (*indev == PICO_INPUT_LIGHT_GUN)
+		e_menu_keyconfig[x + !is0].help = h_gun;
+	else	e_menu_keyconfig[x + !is0].help = NULL;
+
+	x = currentConfig.input_dev0 == PICO_INPUT_LIGHT_GUN ||
+		currentConfig.input_dev1 == PICO_INPUT_LIGHT_GUN ||
+		currentConfig.input_dev1 == PICO_INPUT_JUSTIFIER;
+	me_enable(e_menu_keyconfig, MA_CTRL_LIGHTGUN, x);
+	return 0;
+}
+
+static menu_entry e_menu_gunconfig[] =
+{
+	mee_onoff     ("Gun crosshair",     MA_CTRL_GUN_CURSOR, currentConfig.EmuOpt, EOPT_GUN_CURSOR),
+	mee_range     ("Gun x offset",      MA_CTRL_GUN_XOFFS,currentConfig.gunx, -50, 50),
+	mee_range     ("Gun y offset",      MA_CTRL_GUN_YOFFS,currentConfig.guny, -50, 50),
+	mee_end,
+};
+
+static int gun_config_loop(int id, int keys)
+{
+	static int sel = 0;
+
+	me_loop_d(e_menu_gunconfig, &sel, menu_draw_prep, NULL);
+
+	PicoIn.gunx = currentConfig.gunx;
+	PicoIn.guny = currentConfig.guny;
 
 	return 0;
 }
@@ -614,11 +691,13 @@ static menu_entry e_menu_keyconfig[] =
 {
 	mee_cust_nosave(player,             MA_CTRL_PLAYER1,    key_config_players, mgn_nothing),
 	mee_handler_id("Emulator hotkeys",  MA_CTRL_EMU,        key_config_loop_wrap),
-	mee_cust_h    ("Keyboard",          MA_CTRL_KEYBOARD,   key_config_keyboard, mgn_keyboard, h_kbd),
-	mee_enum      ("Input device 1",    MA_OPT_INPUT_DEV0,  currentConfig.input_dev0, indev0_names),
-	mee_enum      ("Input device 2",    MA_OPT_INPUT_DEV1,  currentConfig.input_dev1, indev1_names),
+	mee_cust_h    ("Keyboard",          MA_OPT_KEYBOARD,    key_config_keyboard, mgn_keyboard, NULL),
+	mee_handler_id("Keys configuration",MA_CTRL_KEYBOARD,   key_config_kbd_loop),
+	mee_cust_h    ("Input device 1",    MA_OPT_INPUT_DEV0,  mh_indev, mgn_indev, NULL),
+	mee_cust_h    ("Input device 2",    MA_OPT_INPUT_DEV1,  mh_indev, mgn_indev, NULL),
 	mee_range     ("Turbo rate",        MA_CTRL_TURBO_RATE, currentConfig.turbo_rate, 1, 30),
 	mee_range     ("Analog deadzone",   MA_CTRL_DEADZONE,   currentConfig.analog_deadzone, 1, 99),
+	mee_handler_id("Gun configuration", MA_CTRL_LIGHTGUN,   gun_config_loop),
 	mee_label     (""),
 	mee_label     ("Input devices:"),
 	mee_label_mk  (MA_CTRL_DEV_FIRST, mgn_dev_name),
@@ -645,11 +724,17 @@ static int menu_loop_keyconfig(int id, int keys)
 	snprintf(l,sizeof(l),"%s", in_get_key_name(-1, -PBTN_LEFT));
 	snprintf(r,sizeof(r),"%s", in_get_key_name(-1, -PBTN_RIGHT));
 	snprintf(h_play12, sizeof(h_play12), "%s/%s - select, %s - configure", l, r, in_get_key_name(-1, -PBTN_MOK));
-	snprintf(h_kbd, sizeof(h_kbd), "%s - configure", in_get_key_name(-1, -PBTN_MOK));
 
 	player[strlen(player)-1] = '1';
-	e_menu_keyconfig[0].help = h_play12;
-	e_menu_keyconfig[2].help = (currentConfig.keyboard == 2 ? h_kbd : NULL);
+	x = me_id2offset(e_menu_keyconfig, MA_CTRL_PLAYER1);
+	e_menu_keyconfig[x].help = h_play12;
+
+	me_enable(e_menu_keyconfig, MA_CTRL_KEYBOARD, currentConfig.keyboard == 2);
+
+	x = currentConfig.input_dev0 == PICO_INPUT_LIGHT_GUN ||
+		currentConfig.input_dev1 == PICO_INPUT_LIGHT_GUN ||
+		currentConfig.input_dev1 == PICO_INPUT_JUSTIFIER;
+	me_enable(e_menu_keyconfig, MA_CTRL_LIGHTGUN, x);
 
 	me_loop_d(e_menu_keyconfig, &sel, menu_draw_prep, NULL);
 
@@ -1346,12 +1431,12 @@ static void draw_frame_credits(void)
 
 static const char credits[] =
 	"PicoDrive v" VERSION "\n"
-	"(c) notaz, 2006-2013; irixxxx, 2018-2024\n\n"
+	"(c) notaz, 2006-2013; irixxxx, 2018-2025\n\n"
 	"Credits:\n"
-	"fDave: initial code\n"
 #ifdef EMU_C68K
-	"      Cyclone 68000 core\n"
+	"fDave: initial code, Cyclone 68000 core\n"
 #else
+	"fDave: initial code\n"
 	"Stef, Chui: FAME/C 68k core\n"
 #endif
 #ifdef _USE_DRZ80
@@ -1365,13 +1450,18 @@ static const char credits[] =
 #ifdef __GP2X__
 	"Squidge: mmuhack\n"
 	"Dzz: ARM940 sample\n"
+#endif
+#if defined __PSP__ || defined __PS2__
+	"fjtrujy: PSP and PS2 platform code\n"
 #endif
 	"\n"
 	"special thanks (for docs, ideas):\n"
 	" Charles MacDonald, Haze,\n"
 	" Stephane Dallongeville,\n"
 	" Lordus, Exophase, Rokas,\n"
-	" Eke, Nemesis, Tasco Deluxe";
+	" Eke, Nemesis, Tasco Deluxe,\n"
+	"special thanks (for testing):\n"
+	" DrUm78, jdgleaver, freq_mod";
 
 static void menu_main_draw_status(void)
 {
@@ -1756,6 +1846,7 @@ static menu_entry *e_menu_table[] =
 	e_menu_sms_options,
 #endif
 	e_menu_keyconfig,
+	e_menu_gunconfig,
 	e_menu_hidden,
 };
 

platform/common/menu_pico.h

@@ -34,6 +34,7 @@ typedef enum
 	MA_OPT_ARM940_SOUND,
 	MA_OPT_INPUT_DEV0,
 	MA_OPT_INPUT_DEV1,
+	MA_OPT_KEYBOARD,
 	MA_OPT_REGION,
 	MA_OPT_CONFIRM_STATES,
 	MA_OPT_SAVE_SLOT,
@@ -110,6 +111,10 @@ typedef enum
 	MA_CTRL_PLAYER4,
 	MA_CTRL_EMU,
 	MA_CTRL_KEYBOARD,
+	MA_CTRL_LIGHTGUN,
+	MA_CTRL_GUN_CURSOR,
+	MA_CTRL_GUN_XOFFS,
+	MA_CTRL_GUN_YOFFS,
 	MA_CTRL_TURBO_RATE,
 	MA_CTRL_DEADZONE,
 	MA_CTRL_DEV_FIRST,
@@ -117,6 +122,8 @@ typedef enum
 	MA_CTRL_DONE,
 } menu_id;
 
+extern const char *indev_names[];
+
 void menu_init(void);
 void menu_loop(void);
 int menu_loop_tray(void);

platform/common/plat_sdl.c

@@ -26,6 +26,8 @@ static void *shadow_fb;
 static int shadow_size;
 static struct area { int w, h; } area;
 
+static int resized;
+
 static struct in_pdata in_sdl_platform_data;
 
 static int hide_cursor;
@@ -68,7 +70,8 @@ int plat_target_init(void)
 		FILE *f = fopen("/proc/device-tree/compatible", "r");
 		if (f) {
 			char buf[10];
-			int c = fread(buf, 1, sizeof(buf), f);
+			int c = fread(buf, 1, sizeof(buf)-1, f);
+			buf[c] = '\0';
 			if (strncmp(buf, "gcw,", 4) == 0)
 				plat_device = "gcw0";
 		}
@@ -204,9 +207,15 @@ static int clear_buf_cnt, clear_stat_cnt;
 
 static void resize_buffers(void)
 {
+	// update pitch as it is needed for resizing, and the menu bg scaler
+	if (plat_sdl_overlay || plat_sdl_gl_active)
+		g_menuscreen_pp = g_menuscreen_w;
+	else
+		g_menuscreen_pp = plat_sdl_screen->pitch / 2;
+
 	// make sure the shadow buffers are big enough in case of resize
-	if (shadow_size < g_menuscreen_w * g_menuscreen_h * 2) {
-		shadow_size = g_menuscreen_w * g_menuscreen_h * 2;
+	if (shadow_size < g_menuscreen_pp * g_menuscreen_h * 2) {
+		shadow_size = g_menuscreen_pp * g_menuscreen_h * 2;
 		shadow_fb = realloc(shadow_fb, shadow_size);
 		g_menubg_ptr = realloc(g_menubg_ptr, shadow_size);
 	}
@@ -214,7 +223,8 @@ static void resize_buffers(void)
 
 void plat_video_set_size(int w, int h)
 {
-	if ((plat_sdl_overlay || plat_sdl_gl_active) && w <= 320 && h <= 240) {
+	if ((plat_sdl_overlay || plat_sdl_gl_active || !plat_sdl_is_windowed()) &&
+			w <= 320 && h <= 240) {
 		// scale to the window, but mind aspect ratio (scaled to 4:3):
 		// w *= win_aspect / 4:3_aspect or h *= 4:3_aspect / win_aspect
 		if (g_menuscreen_w * 3/4 >= g_menuscreen_h)
@@ -233,6 +243,7 @@ void plat_video_set_size(int w, int h)
 				area = (struct area) { g_screen_width,g_screen_height };
 				plat_sdl_change_video_mode(g_screen_width, g_screen_height, 0);
 			}
+			plat_sdl_gl_scaling(currentConfig.filter);
 		}
 		if (plat_sdl_overlay || plat_sdl_gl_active) {
 			// use shadow buffer for overlays
@@ -268,8 +279,6 @@ void plat_video_set_shadow(int w, int h)
 
 void plat_video_flip(void)
 {
-	resize_buffers();
-
 	if (plat_sdl_overlay != NULL) {
 		SDL_Rect dstrect =
 			{ 0, 0, plat_sdl_screen->w, plat_sdl_screen->h };
@@ -296,19 +305,7 @@ void plat_video_flip(void)
 			SDL_UnlockSurface(plat_sdl_screen);
 		SDL_Flip(plat_sdl_screen);
 
-		// take over resized settings for the physical SDL surface
-		if ((plat_sdl_screen->w != g_menuscreen_w ||
-		    plat_sdl_screen->h != g_menuscreen_h)  && plat_sdl_is_windowed() &&
-		    SDL_WM_GrabInput(SDL_GRAB_ON) == SDL_GRAB_ON) {
-			plat_sdl_change_video_mode(g_menuscreen_w, g_menuscreen_h, -1);
-			SDL_WM_GrabInput(SDL_GRAB_OFF);
-			g_menuscreen_pp = plat_sdl_screen->pitch/2;
-
-			// force upper layer to use new dimensions
-			plat_video_set_shadow(g_screen_width, g_screen_height);
-			plat_video_set_buffer(g_screen_ptr);
-			rendstatus_old = -1;
-		} else if (!copy) {
+		if (!copy) {
 			g_screen_ppitch = plat_sdl_screen->pitch/2;
 			g_screen_ptr = plat_sdl_screen->pixels;
 			plat_video_set_buffer(g_screen_ptr);
@@ -323,13 +320,20 @@ void plat_video_flip(void)
 		}
 	}
 
-	// for overlay/gl modes buffer ptr may change on resize
-	if ((plat_sdl_overlay || plat_sdl_gl_active) &&
-	    (g_screen_ptr != shadow_fb || g_screen_ppitch != g_screen_width)) {
-		g_screen_ppitch = g_screen_width;
-		g_screen_ptr = shadow_fb;
+	// take over resized settings for the physical SDL surface
+	if (resized && plat_sdl_is_windowed() &&
+	    SDL_WM_GrabInput(SDL_GRAB_ON) == SDL_GRAB_ON) {
+		plat_sdl_change_video_mode(g_menuscreen_w, g_menuscreen_h, 1);
+		SDL_WM_GrabInput(SDL_GRAB_OFF);
+		resize_buffers();
+
+		// force upper layer to use new dimensions
+		plat_video_set_shadow(g_screen_width, g_screen_height);
 		plat_video_set_buffer(g_screen_ptr);
+		rendstatus_old = -1;
+		resized = 0;
 	}
+
 	if (clear_stat_cnt) {
 		unsigned short *d = (unsigned short *)g_screen_ptr + g_screen_ppitch * g_screen_height;
 		int l = g_screen_ppitch * 8;
@@ -367,17 +371,12 @@ void plat_video_menu_update(void)
 		int w, h;
 		do {
 			w = g_menuscreen_w, h = g_menuscreen_h;
-			plat_sdl_change_video_mode(w, h, -1);
+			plat_sdl_change_video_mode(w, h, 1);
 		} while (w != g_menuscreen_w || h != g_menuscreen_h);
+		plat_sdl_gl_scaling(currentConfig.filter);
 		SDL_WM_GrabInput(SDL_GRAB_OFF);
 	}
 
-	// update pitch as it is needed by the menu bg scaler
-	if (plat_sdl_overlay || plat_sdl_gl_active)
-		g_menuscreen_pp = g_menuscreen_w;
-	else
-		g_menuscreen_pp = plat_sdl_screen->pitch / 2;
-
 	resize_buffers();
 }
 
@@ -385,6 +384,7 @@ void plat_video_menu_enter(int is_rom_loaded)
 {
 	if (SDL_MUSTLOCK(plat_sdl_screen))
 		SDL_UnlockSurface(plat_sdl_screen);
+	plat_sdl_change_video_mode(g_menuscreen_w, g_menuscreen_h, 1);
 }
 
 void plat_video_menu_begin(void)
@@ -437,8 +437,7 @@ void plat_video_loop_prepare(void)
 
 	// take over any new vout settings
 	area = (struct area) { 0, 0 };
-	plat_sdl_change_video_mode(0, 0, 0);
-	resize_buffers();
+//	plat_sdl_change_video_mode(0, 0, 0);
 
 	// switch over to scaled output if available, but keep the aspect ratio
 	if (plat_sdl_overlay || plat_sdl_gl_active)
@@ -470,38 +469,25 @@ int plat_has_wm(void)
 	return plat_sdl_is_windowed();
 }
 
+void plat_set_window(int w, int h)
+{
+	g_menuscreen_w = w;
+	g_menuscreen_h = h;
+	resized = 1;
+}
+
 static void plat_sdl_resize(int w, int h)
 {
 	// take over new settings
 #if defined(__OPENDINGUX__)
-	if (currentConfig.vscaling != EOPT_SCALE_HW &&
-	    plat_sdl_screen->w == 320 && plat_sdl_screen->h == 480) {
-		g_menuscreen_h = 240;
-		g_menuscreen_w = 320;
-	} else
-#endif
-	{
-		g_menuscreen_h = plat_sdl_screen->h;
-		g_menuscreen_w = plat_sdl_screen->w;
-#if 0 // auto resizing may be nice, but creates problems on some SDL platforms
-		if (!plat_sdl_overlay && !plat_sdl_gl_active &&
-		    plat_sdl_is_windowed() && !plat_sdl_is_fullscreen()) {
-			// in SDL window mode, adapt window to integer scaling
-			if (g_menuscreen_w * 3/4 >= g_menuscreen_h)
-				g_menuscreen_w = g_menuscreen_h * 4/3;
-			else
-				g_menuscreen_h = g_menuscreen_w * 3/4;
-			g_menuscreen_w = g_menuscreen_w/320*320;
-			g_menuscreen_h = g_menuscreen_h/240*240;
-			if (g_menuscreen_w == 0) {
-				g_menuscreen_w = 320;
-				g_menuscreen_h = 240;
-			}
-		}
+	if (currentConfig.vscaling != EOPT_SCALE_HW && w == 320 && h == 480)
+		h = 240;
 #endif
+	if (g_menuscreen_w != (w & ~3) || g_menuscreen_h != (h & ~3)){
+		g_menuscreen_h = h & ~3;
+		g_menuscreen_w = w & ~3;
+		resized = 1;
 	}
-
-	rendstatus_old = -1;
 }
 
 static void plat_sdl_quit(void)
@@ -510,6 +496,17 @@ static void plat_sdl_quit(void)
 	exit(1);
 }
 
+static void plat_sdl_handler(void *event_)
+{
+  SDL_Event *event = event_;
+
+  if (event->type == SDL_VIDEORESIZE)
+    // handle resize here to avoid the internal plat_sdl_change_video_mode
+    plat_sdl_resize(event->resize.w, event->resize.h);
+  else
+    plat_sdl_event_handler(event_);
+}
+
 void plat_init(void)
 {
 	int ret;
@@ -530,7 +527,7 @@ void plat_init(void)
 	}
 
 	plat_sdl_quit_cb = plat_sdl_quit;
-	plat_sdl_resize_cb = plat_sdl_resize;
+//	plat_sdl_resize_cb = plat_sdl_resize;
 
 	SDL_WM_SetCaption("PicoDrive " VERSION, NULL);
 
@@ -561,11 +558,11 @@ void plat_init(void)
 	in_sdl_platform_data.joy_map = in_sdl_joy_map,
 	in_sdl_platform_data.key_names = in_sdl_key_names,
 	in_sdl_platform_data.kbd_map = in_sdl_kbd_map,
-	in_sdl_init(&in_sdl_platform_data, plat_sdl_event_handler);
+	in_sdl_init(&in_sdl_platform_data, plat_sdl_handler);
 	in_probe();
 
 	// create an artificial resize event to initialize mouse scaling
-	SDL_Event ev;
+	static SDL_Event ev;
 	ev.resize.type = SDL_VIDEORESIZE;
 	ev.resize.w = g_menuscreen_w;
 	ev.resize.h = g_menuscreen_h;

platform/common/version.h

@@ -1 +1 @@
-#define VERSION "2.04" REVISION
+#define VERSION "2.05" REVISION

platform/linux/emu.c

@@ -228,6 +228,16 @@ void pemu_finalize_frame(const char *fps, const char *notice)
 			draw_pico_ptr();
 	}
 
+	// TODO correct ptr position for hard/soft/no scaling?
+#define is_lightgun(d) (d == PICO_INPUT_LIGHT_GUN || d == PICO_INPUT_JUSTIFIER)
+	if ((currentConfig.EmuOpt & EOPT_GUN_CURSOR) &&
+	    (is_lightgun(currentConfig.input_dev0) || is_lightgun(currentConfig.input_dev1))) {
+		pico_inp_mode = 2;
+		pico_pen_x = PicoPicohw.pen_pos[0] = PicoIn.mouseInt[0];
+		pico_pen_y = PicoPicohw.pen_pos[1] = PicoIn.mouseInt[1];
+		draw_pico_ptr();
+	}
+
 	// draw virtual keyboard on display
 	if (kbd_mode && currentConfig.keyboard == 1 && vkbd)
 		vkbd_draw(vkbd);
@@ -486,7 +496,7 @@ void pemu_loop_prep(void)
 {
 	apply_renderer();
 	plat_video_clear_buffers();
-	plat_show_cursor(!(PicoIn.opt & POPT_EN_MOUSE));
+	plat_show_cursor(!(currentConfig.EmuOpt & EOPT_MOUSE));
 }
 
 void pemu_loop_end(void)

platform/linux/menu.c

@@ -29,8 +29,7 @@ static int mh_windowsize(int id, int keys)
 			if (keys & PBTN_LEFT ) scale--;
 			if (scale <= 0) scale = 1;
 			if (scale > 20) scale = 20;
-			g_menuscreen_w = scale*320;
-			g_menuscreen_h = scale*240;
+			plat_set_window(scale*320, scale*240);
 		}
 		return 0;
 	}

platform/miyoo/pkg.cfg

@@ -1,7 +1,7 @@
 # NOTES:
 ## All variable values should enclosed within double quotes: "<value>"
 ## CONFIGURATION FILE for `gm2xpkg` script version:
-PKGVER="0.7"
+PKGVER="0.9"
 
 # EXEC commands (set to "1" anyone for desired outcome), you can instead use [OPTIONS] of `gm2xpkg`:
 PACKAGE=""

platform/pandora/Makefile

@@ -30,6 +30,7 @@ rel: ../../picodrive PicoDrive.sh picorestore \
 	rm -rf out
 	mkdir out
 	cp -r $^ out/
+	mv out/README.md out/README
 	$(PND_MAKE) -p PicoDrive_$(VER).pnd -d out -x out/PicoDrive.pxml -i out/PicoDrive.png -c
 
 FORCE:

platform/pandora/make_pxml.sh

@@ -7,7 +7,7 @@ test -f $verfile
 major=`head -n 1 $verfile | sed 's/[^0-9]*\([0-9]*\)\.\([0-9]*\).*/\1/g'`
 minor=`head -n 1 $verfile | sed 's/[^0-9]*\([0-9]*\)\.\([0-9]*\).*/\2/g'`
 # lame, I know..
-build=`git describe HEAD | grep -- - | sed -e 's/.*\-\(.*\)\-.*/\1/'`
+build=`git describe --tags HEAD | grep -- - | sed -e 's/.*\-\(.*\)\-.*/\1/'`
 test -n "$build" && build_post="-$build"
 test -n "$build" || build=0
 

platform/ps2/emu.c

@@ -375,7 +375,7 @@ static void sound_deinit(void)
 #define is_16bit_mode() \
 	(currentConfig.renderer == RT_16BIT || (PicoIn.AHW & PAHW_32X) || is_bg_frame)
 
-static int vsync_handler(void)
+static int vsync_handler(int cause)
 {
 	iSignalSema(vsync_sema_id);
 	if (sound_stopped)

platform/win32/plat.c

@@ -126,31 +126,68 @@ int scandir(const char *dir, struct dirent ***namelist, int (*select)(const stru
     char path[MAX_PATH];
     struct dirent **entries = NULL;
     size_t count = 0;
+    int i;
 
-    snprintf(path, sizeof(path), "%s\\*", dir);
-    handle = FindFirstFile(path, &info);
-    if (handle == INVALID_HANDLE_VALUE)
-        return -1;
+    if (dir == NULL || *dir == '\0') {
+        // on top level give a list of available drives
+        DWORD drives = GetLogicalDrives();
+        if (drives == 0) return -1;
 
-    do {
-        struct dirent *entry = (struct dirent *)malloc(sizeof(struct dirent));
-        if (!entry) {
-            free(entries);
-            FindClose(handle);
-            return -1;
+        entries = malloc(sizeof(*entries) * 26);
+        if (!entries) return -1;
+
+        for (i = 0; i < 26; ++i) {
+            int type;
+
+            if (!(drives & (1 << i)))
+                continue;
+
+            snprintf(path, MAX_PATH, "%c:\\", 'A' + i);
+            type = GetDriveTypeA(path);
+            if (type == DRIVE_NO_ROOT_DIR || type == DRIVE_UNKNOWN)
+                continue;
+
+            struct dirent *entry = (struct dirent *)malloc(sizeof(struct dirent));
+            if (!entry)
+                goto err;
+
+            snprintf(entry->d_name, MAX_PATH, "%c:", 'A' + i);
+            entry->d_type = DT_DIR;
+
+            if (!select || select(entry))
+                entries[count++] = entry;
+            else
+                free(entry);
         }
+    } else {
 
-        strcpy(entry->d_name, info.cFileName);
-        entry->d_type = (info.dwFileAttributes & FILE_ATTRIBUTE_DIRECTORY) ? DT_DIR : DT_REG;
+        // create a list of objects in the directory
+        snprintf(path, sizeof(path), "%s\\*", dir);
+        handle = FindFirstFile(path, &info);
+        if (handle == INVALID_HANDLE_VALUE)
+            return -1;
 
-        if (!select || select(entry)) {
-            entries = realloc(entries, (count + 1) * sizeof(struct dirent *));
-            entries[count++] = entry;
-        } else
-            free(entry);
-    } while (FindNextFile(handle, &info));
+        do {
+            struct dirent *entry = (struct dirent *)malloc(sizeof(struct dirent));
+            if (!entry) {
+                FindClose(handle);
+                goto err;
+            }
 
-    FindClose(handle);
+            strcpy(entry->d_name, info.cFileName);
+            entry->d_type = (info.dwFileAttributes & FILE_ATTRIBUTE_DIRECTORY) ? DT_DIR : DT_REG;
+
+            if (!select || select(entry)) {
+                struct dirent **_entries = realloc(entries, (count + 1) * sizeof(struct dirent *));
+                if (!_entries) goto err;
+                entries = _entries;
+                entries[count++] = entry;
+            } else
+                free(entry);
+        } while (FindNextFile(handle, &info));
+
+        FindClose(handle);
+    }
 
     // Sort entries if a comparison function is provided
     if (compar) {
@@ -159,6 +196,15 @@ int scandir(const char *dir, struct dirent ***namelist, int (*select)(const stru
 
     *namelist = entries;
     return count;
+
+err:
+    if (entries) {
+        while (count > 0)
+            if (entries[--count])
+                free(entries[count]);
+        free(entries);
+    }
+    return -1;
 }
 
 int alphasort(const struct dirent **a, const struct dirent **b) {