Merge git://github.com/notaz/pcsx_rearmed

6 files changed, 205 insertions, 68 deletions

diff --git a/plugins/gpulib/cspace.c b/plugins/gpulib/cspace.c
index 1d5718c..8e3bee9 100644
--- a/plugins/gpulib/cspace.c
+++ b/plugins/gpulib/cspace.c
@@ -1,3 +1,13 @@
+/*
+ * (C) Gražvydas "notaz" Ignotas, 2011,2012
+ *
+ * This work is licensed under the terms of any of these licenses
+ * (at your option):
+ *  - GNU GPL, version 2 or later.
+ *  - GNU LGPL, version 2.1 or later.
+ * See the COPYING file in the top-level directory.
+ */
+
 #include "cspace.h"
 
 /*
@@ -5,6 +15,8 @@
  * in favor of NEON version or platform-specific conversion
  */
 
+#ifndef __ARM_NEON__
+
 void bgr555_to_rgb565(void *dst_, const void *src_, int bytes)
 {
 	const unsigned int *src = src_;
@@ -42,3 +54,92 @@ void bgr888_to_rgb565(void *dst_, const void *src_, int bytes)
 void rgb888_to_rgb565(void *dst, const void *src, int bytes) {}
 void bgr888_to_rgb888(void *dst, const void *src, int bytes) {}
 
+#endif // __ARM_NEON__
+
+/* YUV stuff */
+static int yuv_ry[32], yuv_gy[32], yuv_by[32];
+static unsigned char yuv_u[32 * 2], yuv_v[32 * 2];
+
+void bgr_to_uyvy_init(void)
+{
+  int i, v;
+
+  /* init yuv converter:
+    y0 = (int)((0.299f * r0) + (0.587f * g0) + (0.114f * b0));
+    y1 = (int)((0.299f * r1) + (0.587f * g1) + (0.114f * b1));
+    u = (int)(8 * 0.565f * (b0 - y0)) + 128;
+    v = (int)(8 * 0.713f * (r0 - y0)) + 128;
+  */
+  for (i = 0; i < 32; i++) {
+    yuv_ry[i] = (int)(0.299f * i * 65536.0f + 0.5f);
+    yuv_gy[i] = (int)(0.587f * i * 65536.0f + 0.5f);
+    yuv_by[i] = (int)(0.114f * i * 65536.0f + 0.5f);
+  }
+  for (i = -32; i < 32; i++) {
+    v = (int)(8 * 0.565f * i) + 128;
+    if (v < 0)
+      v = 0;
+    if (v > 255)
+      v = 255;
+    yuv_u[i + 32] = v;
+    v = (int)(8 * 0.713f * i) + 128;
+    if (v < 0)
+      v = 0;
+    if (v > 255)
+      v = 255;
+    yuv_v[i + 32] = v;
+  }
+}
+
+void bgr555_to_uyvy(void *d, const void *s, int pixels)
+{
+  unsigned int *dst = d;
+  const unsigned short *src = s;
+  const unsigned char *yu = yuv_u + 32;
+  const unsigned char *yv = yuv_v + 32;
+  int r0, g0, b0, r1, g1, b1;
+  int y0, y1, u, v;
+
+  for (; pixels > 0; src += 2, dst++, pixels -= 2)
+  {
+    b0 = (src[0] >> 10) & 0x1f;
+    g0 = (src[0] >> 5) & 0x1f;
+    r0 =  src[0] & 0x1f;
+    b1 = (src[1] >> 10) & 0x1f;
+    g1 = (src[1] >> 5) & 0x1f;
+    r1 =  src[1] & 0x1f;
+    y0 = (yuv_ry[r0] + yuv_gy[g0] + yuv_by[b0]) >> 16;
+    y1 = (yuv_ry[r1] + yuv_gy[g1] + yuv_by[b1]) >> 16;
+    u = yu[b0 - y0];
+    v = yv[r0 - y0];
+    // valid Y range seems to be 16..235
+    y0 = 16 + 219 * y0 / 31;
+    y1 = 16 + 219 * y1 / 31;
+
+    *dst = (y1 << 24) | (v << 16) | (y0 << 8) | u;
+  }
+}
+
+void bgr888_to_uyvy(void *d, const void *s, int pixels)
+{
+  unsigned int *dst = d;
+  const unsigned char *src8 = s;
+  const unsigned char *yu = yuv_u + 32;
+  const unsigned char *yv = yuv_v + 32;
+  int r0, g0, b0, r1, g1, b1;
+  int y0, y1, u, v;
+
+  for (; pixels > 0; src8 += 3*2, dst++, pixels -= 2)
+  {
+    r0 = src8[0], g0 = src8[1], b0 = src8[2];
+    r1 = src8[3], g1 = src8[4], b1 = src8[5];
+    y0 = (r0 * 19595 + g0 * 38470 + b0 * 7471) >> 16;
+    y1 = (r1 * 19595 + g1 * 38470 + b1 * 7471) >> 16;
+    u = yu[(b0 - y0) / 8];
+    v = yv[(r0 - y0) / 8];
+    y0 = 16 + 219 * y0 / 255;
+    y1 = 16 + 219 * y1 / 255;
+
+    *dst = (y1 << 24) | (v << 16) | (y0 << 8) | u;
+  }
+}
diff --git a/plugins/gpulib/cspace.h b/plugins/gpulib/cspace.h
index 8c9bcfa..95eae85 100644
--- a/plugins/gpulib/cspace.h
+++ b/plugins/gpulib/cspace.h
@@ -8,6 +8,10 @@ void bgr888_to_rgb888(void *dst, const void *src, int bytes);
 void bgr888_to_rgb565(void *dst, const void *src, int bytes);
 void rgb888_to_rgb565(void *dst, const void *src, int bytes);
 
+void bgr_to_uyvy_init(void);
+void bgr555_to_uyvy(void *d, const void *s, int pixels);
+void bgr888_to_uyvy(void *d, const void *s, int pixels);
+
 #ifdef __cplusplus
 }
 #endif
diff --git a/plugins/gpulib/gpu.c b/plugins/gpulib/gpu.c
index 46e92d1..b300c88 100644
--- a/plugins/gpulib/gpu.c
+++ b/plugins/gpulib/gpu.c
@@ -24,7 +24,7 @@
 //#define log_anomaly gpu_log
 #define log_anomaly(...)
 
-struct psx_gpu gpu __attribute__((aligned(2048)));
+struct psx_gpu gpu;
 
 static noinline int do_cmd_buffer(uint32_t *data, int count);
 static void finish_vram_transfer(int is_read);
@@ -133,6 +133,22 @@ static noinline void get_gpu_info(uint32_t data)
   }
 }
 
+// double, for overdraw guard
+#define VRAM_SIZE (1024 * 512 * 2 * 2)
+
+static int map_vram(void)
+{
+  gpu.vram = gpu.mmap(VRAM_SIZE);
+  if (gpu.vram != NULL) {
+    gpu.vram += 4096 / 2;
+    return 0;
+  }
+  else {
+    fprintf(stderr, "could not map vram, expect crashes\n");
+    return -1;
+  }
+}
+
 long GPUinit(void)
 {
   int ret;
@@ -145,12 +161,26 @@ long GPUinit(void)
   gpu.cmd_len = 0;
   do_reset();
 
+  if (gpu.mmap != NULL) {
+    if (map_vram() != 0)
+      ret = -1;
+  }
   return ret;
 }
 
 long GPUshutdown(void)
 {
-  return vout_finish();
+  long ret;
+
+  renderer_finish();
+  ret = vout_finish();
+  if (gpu.vram != NULL) {
+    gpu.vram -= 4096 / 2;
+    gpu.munmap(gpu.vram, VRAM_SIZE);
+  }
+  gpu.vram = NULL;
+
+  return ret;
 }
 
 void GPUwriteStatus(uint32_t data)
@@ -182,7 +212,7 @@ void GPUwriteStatus(uint32_t data)
       break;
     case 0x05:
       gpu.screen.x = data & 0x3ff;
-      gpu.screen.y = (data >> 10) & 0x3ff;
+      gpu.screen.y = (data >> 10) & 0x1ff;
       if (gpu.frameskip.set) {
         decide_frameskip_allow(gpu.ex_regs[3]);
         if (gpu.frameskip.last_flip_frame != *gpu.state.frame_count) {
@@ -207,6 +237,7 @@ void GPUwriteStatus(uint32_t data)
       gpu.screen.vres = vres[(gpu.status.reg >> 19) & 3];
       update_width();
       update_height();
+      renderer_notify_res_change();
       break;
     default:
       if ((cmd & 0xf0) == 0x10)
@@ -582,13 +613,13 @@ long GPUfreeze(uint32_t type, struct GPUFreeze *freeze)
     case 1: // save
       if (gpu.cmd_len > 0)
         flush_cmd_buffer();
-      memcpy(freeze->psxVRam, gpu.vram, sizeof(gpu.vram));
+      memcpy(freeze->psxVRam, gpu.vram, 1024 * 512 * 2);
       memcpy(freeze->ulControl, gpu.regs, sizeof(gpu.regs));
       memcpy(freeze->ulControl + 0xe0, gpu.ex_regs, sizeof(gpu.ex_regs));
       freeze->ulStatus = gpu.status.reg;
       break;
     case 0: // load
-      memcpy(gpu.vram, freeze->psxVRam, sizeof(gpu.vram));
+      memcpy(gpu.vram, freeze->psxVRam, 1024 * 512 * 2);
       memcpy(gpu.regs, freeze->ulControl, sizeof(gpu.regs));
       memcpy(gpu.ex_regs, freeze->ulControl + 0xe0, sizeof(gpu.ex_regs));
       gpu.status.reg = freeze->ulStatus;
@@ -669,6 +700,14 @@ void GPUrearmedCallbacks(const struct rearmed_cbs *cbs)
   gpu.state.hcnt = cbs->gpu_hcnt;
   gpu.state.frame_count = cbs->gpu_frame_count;
   gpu.state.allow_interlace = cbs->gpu_neon.allow_interlace;
+  gpu.state.enhancement_enable = cbs->gpu_neon.enhancement_enable;
+
+  gpu.mmap = cbs->mmap;
+  gpu.munmap = cbs->munmap;
+
+  // delayed vram mmap
+  if (gpu.vram == NULL)
+    map_vram();
 
   if (cbs->pl_vout_set_raw_vram)
     cbs->pl_vout_set_raw_vram(gpu.vram);
diff --git a/plugins/gpulib/gpu.h b/plugins/gpulib/gpu.h
index 1cbe38c..d11f991 100644
--- a/plugins/gpulib/gpu.h
+++ b/plugins/gpulib/gpu.h
@@ -17,10 +17,9 @@ extern "C" {
 #define CMD_BUFFER_LEN          1024
 
 struct psx_gpu {
-  uint16_t vram[1024 * 512];
-  uint16_t guard[1024 * 512]; // overdraw guard
   uint32_t cmd_buffer[CMD_BUFFER_LEN];
   uint32_t regs[16];
+  uint16_t *vram;
   union {
     uint32_t reg;
     struct {
@@ -67,6 +66,8 @@ struct psx_gpu {
     uint32_t old_interlace:1;
     uint32_t allow_interlace:2;
     uint32_t blanked:1;
+    uint32_t enhancement_enable:1;
+    uint32_t enhancement_active:1;
     uint32_t *frame_count;
     uint32_t *hcnt; /* hsync count */
     struct {
@@ -87,6 +88,10 @@ struct psx_gpu {
     uint32_t last_flip_frame;
     uint32_t pending_fill[3];
   } frameskip;
+  uint16_t *(*get_enhancement_bufer)
+    (int *x, int *y, int *w, int *h, int *vram_h);
+  void *(*mmap)(unsigned int size);
+  void  (*munmap)(void *ptr, unsigned int size);
 };
 
 extern struct psx_gpu gpu;
@@ -98,11 +103,13 @@ int do_cmd_list(uint32_t *list, int count, int *last_cmd);
 struct rearmed_cbs;
 
 int  renderer_init(void);
+void renderer_finish(void);
 void renderer_sync_ecmds(uint32_t * ecmds);
 void renderer_update_caches(int x, int y, int w, int h);
 void renderer_flush_queues(void);
 void renderer_set_interlace(int enable, int is_odd);
 void renderer_set_config(const struct rearmed_cbs *config);
+void renderer_notify_res_change(void);
 
 int  vout_init(void);
 int  vout_finish(void);
diff --git a/plugins/gpulib/gpulib.mak b/plugins/gpulib/gpulib.mak
index 349a0c8..ad6a8ad 100644
--- a/plugins/gpulib/gpulib.mak
+++ b/plugins/gpulib/gpulib.mak
@@ -2,7 +2,7 @@
 # always adding gpulib to deps in case cspace is needed
 # users must include ../../config.mak
 
-LDFLAGS += -shared
+LDFLAGS += -shared -Wl,--no-undefined
 CFLAGS += $(PLUGIN_CFLAGS)
 ifeq "$(ARCH)" "arm"
  EXT =
@@ -17,10 +17,10 @@ endif
 GPULIB_A = ../gpulib/gpulib$(EXT).a
 
 ifdef BIN_STANDLALONE
-TARGETS += $(BIN_STANDLALONE)$(EXT)
+TARGETS += $(BIN_STANDLALONE)
 endif
 ifdef BIN_GPULIB
-TARGETS += $(BIN_GPULIB)$(EXT)
+TARGETS += $(BIN_GPULIB)
 endif
 CC_STANDLALONE = $(CC)
 CC_GPULIB = $(CC)
@@ -34,7 +34,7 @@ ifdef BIN_STANDLALONE
 ifneq ($(findstring .cpp,$(SRC_STANDALONE)),)
 CC_STANDLALONE = $(CXX)
 endif
-$(BIN_STANDLALONE)$(EXT): $(SRC) $(SRC_STANDALONE) $(GPULIB_A)
+$(BIN_STANDLALONE): $(SRC) $(SRC_STANDALONE) $(GPULIB_A)
 	$(CC_STANDLALONE) -o $@ $(CFLAGS) $(LDFLAGS) $^ $(LDLIBS) $(LDLIBS_STANDALONE)
 	ln -fs $(PLUGINDIR)/$@ ../
 endif
@@ -43,7 +43,7 @@ ifdef BIN_GPULIB
 ifneq ($(findstring .cpp,$(SRC_GPULIB)),)
 CC_GPULIB = $(CXX)
 endif
-$(BIN_GPULIB)$(EXT): $(SRC) $(SRC_GPULIB) $(GPULIB_A)
+$(BIN_GPULIB): $(SRC) $(SRC_GPULIB) $(GPULIB_A)
 	$(CC_GPULIB) -o $@ $(CFLAGS) $(LDFLAGS) $^ $(LDLIBS) $(LDLIBS_GPULIB)
 	ln -fs $(PLUGINDIR)/$@ ../
 endif
diff --git a/plugins/gpulib/vout_pl.c b/plugins/gpulib/vout_pl.c
index 0bd1ecf..49f53d6 100644
--- a/plugins/gpulib/vout_pl.c
+++ b/plugins/gpulib/vout_pl.c
@@ -15,7 +15,6 @@
 #include "../../frontend/plugin_lib.h"
 
 static const struct rearmed_cbs *cbs;
-static void *screen_buf;
 
 int vout_init(void)
 {
@@ -27,90 +26,77 @@ int vout_finish(void)
   return 0;
 }
 
-static void check_mode_change(void)
+static void check_mode_change(int force)
 {
   static uint32_t old_status;
   static int old_h;
+  int w = gpu.screen.hres;
+  int h = gpu.screen.h;
+  int w_out = w;
+  int h_out = h;
+
+  gpu.state.enhancement_active =
+    gpu.get_enhancement_bufer != NULL && gpu.state.enhancement_enable
+    && w <= 512 && h <= 256 && !gpu.status.rgb24;
+
+  if (gpu.state.enhancement_active) {
+    w_out *= 2;
+    h_out *= 2;
+  }
 
   // width|rgb24 change?
-  if ((gpu.status.reg ^ old_status) & ((7<<16)|(1<<21)) || gpu.screen.h != old_h)
+  if (force || (gpu.status.reg ^ old_status) & ((7<<16)|(1<<21)) || h != old_h)
   {
     old_status = gpu.status.reg;
-    old_h = gpu.screen.h;
-    screen_buf = cbs->pl_vout_set_mode(gpu.screen.hres, gpu.screen.h,
+    old_h = h;
+
+    cbs->pl_vout_set_mode(w_out, h_out, w, h,
       (gpu.status.rgb24 && !cbs->only_16bpp) ? 24 : 16);
   }
 }
 
-static void blit(void)
+void vout_update(void)
 {
   int x = gpu.screen.x & ~1; // alignment needed by blitter
   int y = gpu.screen.y;
   int w = gpu.screen.w;
   int h = gpu.screen.h;
   uint16_t *vram = gpu.vram;
-  int stride = gpu.screen.hres;
-  int fb_offs, doffs;
-  uint8_t *dest;
+  int vram_h = 512;
 
-  dest = (uint8_t *)screen_buf;
-  if (dest == NULL)
+  if (w == 0 || h == 0)
     return;
 
-  fb_offs = y * 1024 + x;
+  check_mode_change(0);
+  if (gpu.state.enhancement_active)
+    vram = gpu.get_enhancement_bufer(&x, &y, &w, &h, &vram_h);
 
-  // only do centering, at least for now
-  doffs = (stride - w) / 2 & ~1;
-
-  if (gpu.status.rgb24)
-  {
-    if (cbs->only_16bpp) {
-      dest += doffs * 2;
-      for (; h-- > 0; dest += stride * 2, fb_offs += 1024)
-      {
-        fb_offs &= 1024*512-1;
-        bgr888_to_rgb565(dest, vram + fb_offs, w * 3);
-      }
-    }
-    else {
-      dest += (doffs / 8) * 24;
-      for (; h-- > 0; dest += stride * 3, fb_offs += 1024)
-      {
-        fb_offs &= 1024*512-1;
-        bgr888_to_rgb888(dest, vram + fb_offs, w * 3);
-      }
-    }
-  }
-  else
-  {
-    dest += doffs * 2;
-    for (; h-- > 0; dest += stride * 2, fb_offs += 1024)
-    {
-      fb_offs &= 1024*512-1;
-      bgr555_to_rgb565(dest, vram + fb_offs, w * 2);
+  if (y + h > vram_h) {
+    if (y + h - vram_h > h / 2) {
+      // wrap
+      h -= vram_h - y;
+      y = 0;
     }
+    else
+      // clip
+      h = vram_h - y;
   }
 
-  screen_buf = cbs->pl_vout_flip();
-}
+  vram += y * 1024 + x;
 
-void vout_update(void)
-{
-  check_mode_change();
-  if (cbs->pl_vout_raw_flip)
-    cbs->pl_vout_raw_flip(gpu.screen.x, gpu.screen.y);
-  else
-    blit();
+  cbs->pl_vout_flip(vram, 1024, gpu.status.rgb24, w, h);
 }
 
 void vout_blank(void)
 {
-  check_mode_change();
-  if (cbs->pl_vout_raw_flip == NULL) {
-    int bytespp = gpu.status.rgb24 ? 3 : 2;
-    memset(screen_buf, 0, gpu.screen.hres * gpu.screen.h * bytespp);
-    screen_buf = cbs->pl_vout_flip();
+  int w = gpu.screen.hres;
+  int h = gpu.screen.h;
+  if (gpu.state.enhancement_active) {
+    w *= 2;
+    h *= 2;
   }
+  check_mode_change(0);
+  cbs->pl_vout_flip(NULL, 1024, gpu.status.rgb24, w, h);
 }
 
 long GPUopen(void **unused)
@@ -119,7 +105,7 @@ long GPUopen(void **unused)
   gpu.frameskip.frame_ready = 1;
 
   cbs->pl_vout_open();
-  screen_buf = cbs->pl_vout_flip();
+  check_mode_change(1);
   return 0;
 }