gpu_neon: partially support range regs

3 files changed, 54 insertions, 19 deletions

diff --git a/plugins/gpu_neon/gpu.c b/plugins/gpu_neon/gpu.c
index bacb05b..c43fd31 100644
--- a/plugins/gpu_neon/gpu.c
+++ b/plugins/gpu_neon/gpu.c
@@ -14,6 +14,7 @@
 
 #define ARRAY_SIZE(x) (sizeof(x) / sizeof((x)[0]))
 #define unlikely(x) __builtin_expect((x), 0)
+#define noinline __attribute__((noinline))
 
 //#define log_io printf
 #define log_io(...)
@@ -34,15 +35,40 @@ long GPUshutdown(void)
   return vout_finish();
 }
 
+static noinline void update_width(void)
+{
+  int sw = gpu.screen.x2 - gpu.screen.x1;
+  if (sw <= 0 || sw >= 2560)
+    // full width
+    gpu.screen.w = gpu.screen.hres;
+  else
+    gpu.screen.w = sw * gpu.screen.hres / 2560;
+}
+
+static noinline void update_height(void)
+{
+  int sh = gpu.screen.y2 - gpu.screen.y1;
+  if (gpu.status.dheight)
+    sh *= 2;
+  if (sh <= 0)
+    sh = gpu.screen.vres;
+
+  gpu.screen.h = sh;
+}
+
 void GPUwriteStatus(uint32_t data)
 {
   static const short hres[8] = { 256, 368, 320, 384, 512, 512, 640, 640 };
   static const short vres[4] = { 240, 480, 256, 480 };
   uint32_t cmd = data >> 24;
 
-  switch (data >> 24) {
+  if (cmd < ARRAY_SIZE(gpu.regs))
+    gpu.regs[cmd] = data;
+
+  switch (cmd) {
     case 0x00:
       gpu.status.reg = 0x14802000;
+      gpu.status.blanking = 1;
       break;
     case 0x03:
       gpu.status.blanking = data & 1;
@@ -54,19 +80,24 @@ void GPUwriteStatus(uint32_t data)
       gpu.screen.x = data & 0x3ff;
       gpu.screen.y = (data >> 10) & 0x3ff;
       break;
+    case 0x06:
+      gpu.screen.x1 = data & 0xfff;
+      gpu.screen.x2 = (data >> 12) & 0xfff;
+      update_width();
+      break;
     case 0x07:
       gpu.screen.y1 = data & 0x3ff;
       gpu.screen.y2 = (data >> 10) & 0x3ff;
+      update_height();
       break;
     case 0x08:
       gpu.status.reg = (gpu.status.reg & ~0x7f0000) | ((data & 0x3F) << 17) | ((data & 0x40) << 10);
-      gpu.screen.w = hres[(gpu.status.reg >> 16) & 7];
-      gpu.screen.h = vres[(gpu.status.reg >> 19) & 3];
+      gpu.screen.hres = hres[(gpu.status.reg >> 16) & 7];
+      gpu.screen.vres = vres[(gpu.status.reg >> 19) & 3];
+      update_width();
+      update_height();
       break;
   }
-
-  if (cmd < ARRAY_SIZE(gpu.regs))
-    gpu.regs[cmd] = data;
 }
 
 const unsigned char cmd_lengths[256] =
diff --git a/plugins/gpu_neon/gpu.h b/plugins/gpu_neon/gpu.h
index ebf2640..8813a44 100644
--- a/plugins/gpu_neon/gpu.h
+++ b/plugins/gpu_neon/gpu.h
@@ -45,7 +45,9 @@ struct psx_gpu {
     };
   } status;
   struct {
+    int hres, vres;
     int x, y, w, h;
+    int x1, x2;
     int y1, y2;
   } screen;
   struct {
diff --git a/plugins/gpu_neon/vout_fb.c b/plugins/gpu_neon/vout_fb.c
index 7b77d78..fdcfd51 100644
--- a/plugins/gpu_neon/vout_fb.c
+++ b/plugins/gpu_neon/vout_fb.c
@@ -28,39 +28,40 @@ int vout_finish(void)
 static void blit(void)
 {
   static uint32_t old_status, old_h;
-  int x = gpu.screen.x & ~3; // alignment needed by blitter
+  int x = gpu.screen.x & ~1; // alignment needed by blitter
   int y = gpu.screen.y;
   int w = gpu.screen.w;
-  int h;
+  int h = gpu.screen.h;
+  int stride = gpu.screen.hres;
+  int doffs;
   uint16_t *srcs;
   uint8_t  *dest;
 
   srcs = &gpu.vram[y * 1024 + x];
 
-  h = gpu.screen.y2 - gpu.screen.y1;
-  if (gpu.status.dheight)
-    h *= 2;
-
-  if (h <= 0)
-    return;
-
   if ((gpu.status.reg ^ old_status) & ((7<<16)|(1<<21)) || h != old_h) // width|rgb24 change?
   {
     old_status = gpu.status.reg;
     old_h = h;
-    screen_buf = cbs->pl_fbdev_set_mode(w, h, gpu.status.rgb24 ? 24 : 16);
+    screen_buf = cbs->pl_fbdev_set_mode(stride, h, gpu.status.rgb24 ? 24 : 16);
   }
+
   dest = screen_buf;
 
+  // only do centering, at least for now
+  doffs = (stride - w) / 2 & ~1;
+
   if (gpu.status.rgb24)
   {
 #ifndef MAEMO
-    for (; h-- > 0; dest += w * 3, srcs += 1024)
+    dest += (doffs / 8) * 24;
+    for (; h-- > 0; dest += stride * 3, srcs += 1024)
     {
       bgr888_to_rgb888(dest, srcs, w * 3);
     }
 #else
-    for (; h-- > 0; dest += w * 2, srcs += 1024)
+    dest += doffs * 2;
+    for (; h-- > 0; dest += stride * 2, srcs += 1024)
     {
       bgr888_to_rgb565(dest, srcs, w * 3);
     }
@@ -68,7 +69,8 @@ static void blit(void)
   }
   else
   {
-    for (; h-- > 0; dest += w * 2, srcs += 1024)
+    dest += doffs * 2;
+    for (; h-- > 0; dest += stride * 2, srcs += 1024)
     {
       bgr555_to_rgb565(dest, srcs, w * 2);
     }