Explorar el Código

accurate_sprites performance improvement, PSP is untested

git-svn-id: file:///home/notaz/opt/svn/PicoDrive@478 be3aeb3a-fb24-0410-a615-afba39da0efa
notaz hace 16 años
padre
commit
e5fa981777
Se han modificado 6 ficheros con 291 adiciones y 229 borrados
  1. 167 183
      Pico/Draw.c
  2. 41 26
      Pico/Draw.s
  3. 37 0
      Pico/Draw_amips.s
  4. 2 2
      Pico/Pico.c
  5. 23 10
      platform/gp2x/emu.c
  6. 21 8
      platform/psp/emu.c

+ 167 - 183
Pico/Draw.c

@@ -6,6 +6,17 @@
 
 // For commercial use, separate licencing terms must be obtained.
 
+/*
+ * The renderer has 4 modes now:
+ * - normal
+ * - shadow/hilight (s/h)
+ * - "sonic mode" for midline palette changes
+ * - accurate sprites (AS)
+ *
+ * AS and s/h both use upper bits for both priority and shadow/hilight flags.
+ * "sonic mode" is autodetected, shadow/hilight is enabled by emulated game.
+ * AS is enabled by user and takes priority over "sonic mode".
+ */
 
 #include "PicoInt.h"
 
@@ -26,8 +37,6 @@ static int  HighCacheB[41+1];
 static int  HighCacheS[80+1];   // and sprites
 static int  HighPreSpr[80*2+1]; // slightly preprocessed sprites
 
-static unsigned char HighSprZ[320+8+8]; // Z-buffer for accurate sprites
-
 int rendstatus = 0;
 int Scanline = 0; // Scanline
 
@@ -36,6 +45,9 @@ static int skip_next_line=0;
 
 //unsigned short ppt[] = { 0x0f11, 0x0ff1, 0x01f1, 0x011f, 0x01ff, 0x0f1f, 0x0f0e, 0x0e7c };
 
+static void (*DrawAllSpritesLoPri)(int *hcache, int maxwidth, int prio, int sh) = NULL;
+static void (*DrawAllSpritesHiPri)(int *hcache, int maxwidth, int prio, int sh) = NULL;
+
 struct TileStrip
 {
   int nametab; // Position in VRAM of name table (for this tile line)
@@ -50,9 +62,9 @@ struct TileStrip
 #ifdef _ASM_DRAW_C
 void DrawWindow(int tstart, int tend, int prio, int sh);
 void BackFill(int reg7, int sh);
-void DrawSprite(int *sprite, int **hc, int sh);
+void DrawSprite(int *sprite, int **hc, int sh, int as);
 void DrawTilesFromCache(int *hc, int sh, int rlim);
-void DrawSpritesFromCache(int *hc, int sh);
+void DrawSpritesFromCache(int *hc, int maxwidth, int prio, int sh);
 void DrawLayer(int plane_sh, int *hcache, int cellskip, int maxcells);
 void FinalizeLineBGR444(int sh);
 void FinalizeLineRGB555(int sh);
@@ -122,7 +134,7 @@ static int TileFlip(int sx,int addr,int pal)
 #define sh_pix(x) \
   if(!t); \
   else if(t==0xe) pd[x]=(unsigned char)((pd[x]&0x3f)|0x80); /* hilight */ \
-  else if(t==0xf) pd[x]=(unsigned char)((pd[x]&0x3f)|0xc0); /* shadow  */ \
+  else if(t==0xf) pd[x]=(unsigned char)( pd[x]      |0xc0); /* shadow  */ \
   else pd[x]=(unsigned char)(pal|t)
 
 #ifndef _ASM_DRAW_C
@@ -170,115 +182,104 @@ static int TileFlipSH(int sx,int addr,int pal)
 }
 #endif
 
-static int TileNormZ(int sx,int addr,int pal,int zval)
+#define tilepixelAS(mask,index,shift) \
+  if (!(pd[index]&0xc0)) { t=pack&mask; if (t) pd[index]=(pal|(t>>shift)); }
+
+static int TileNormAS(int sx,int addr,int pal)
 {
-  unsigned int pack=0; unsigned int t=0;
   unsigned char *pd = HighCol+sx;
-  unsigned char *zb = HighSprZ+sx;
-  int collision = 0;
+  unsigned int pack=0; unsigned int t=0;
 
   pack=*(unsigned int *)(Pico.vram+addr); // Get 8 pixels
   if (pack)
   {
-    t=pack&0x0000f000; if(t) { if(zb[0]) collision=1; if(zval>zb[0]) { pd[0]=(unsigned char)(pal|(t>>12)); zb[0]=(char)zval; } }
-    t=pack&0x00000f00; if(t) { if(zb[1]) collision=1; if(zval>zb[1]) { pd[1]=(unsigned char)(pal|(t>> 8)); zb[1]=(char)zval; } }
-    t=pack&0x000000f0; if(t) { if(zb[2]) collision=1; if(zval>zb[2]) { pd[2]=(unsigned char)(pal|(t>> 4)); zb[2]=(char)zval; } }
-    t=pack&0x0000000f; if(t) { if(zb[3]) collision=1; if(zval>zb[3]) { pd[3]=(unsigned char)(pal|(t    )); zb[3]=(char)zval; } }
-    t=pack&0xf0000000; if(t) { if(zb[4]) collision=1; if(zval>zb[4]) { pd[4]=(unsigned char)(pal|(t>>28)); zb[4]=(char)zval; } }
-    t=pack&0x0f000000; if(t) { if(zb[5]) collision=1; if(zval>zb[5]) { pd[5]=(unsigned char)(pal|(t>>24)); zb[5]=(char)zval; } }
-    t=pack&0x00f00000; if(t) { if(zb[6]) collision=1; if(zval>zb[6]) { pd[6]=(unsigned char)(pal|(t>>20)); zb[6]=(char)zval; } }
-    t=pack&0x000f0000; if(t) { if(zb[7]) collision=1; if(zval>zb[7]) { pd[7]=(unsigned char)(pal|(t>>16)); zb[7]=(char)zval; } }
-    if (collision) Pico.video.status|=0x20;
+    tilepixelAS(0x0000f000, 0, 12);
+    tilepixelAS(0x00000f00, 1,  8);
+    tilepixelAS(0x000000f0, 2,  4);
+    tilepixelAS(0x0000000f, 3,  0);
+    tilepixelAS(0xf0000000, 4, 28);
+    tilepixelAS(0x0f000000, 5, 24);
+    tilepixelAS(0x00f00000, 6, 20);
+    tilepixelAS(0x000f0000, 7, 16);
     return 0;
   }
 
   return 1; // Tile blank
 }
 
-static int TileFlipZ(int sx,int addr,int pal,int zval)
+static int TileFlipAS(int sx,int addr,int pal)
 {
-  unsigned int pack=0; unsigned int t=0;
   unsigned char *pd = HighCol+sx;
-  unsigned char *zb = HighSprZ+sx;
-  int collision = 0;
+  unsigned int pack=0; unsigned int t=0;
 
   pack=*(unsigned int *)(Pico.vram+addr); // Get 8 pixels
   if (pack)
   {
-    t=pack&0x000f0000; if(t) { if(zb[0]) collision=1; if(zval>zb[0]) { pd[0]=(unsigned char)(pal|(t>>16)); zb[0]=(char)zval; } }
-    t=pack&0x00f00000; if(t) { if(zb[1]) collision=1; if(zval>zb[1]) { pd[1]=(unsigned char)(pal|(t>>20)); zb[1]=(char)zval; } }
-    t=pack&0x0f000000; if(t) { if(zb[2]) collision=1; if(zval>zb[2]) { pd[2]=(unsigned char)(pal|(t>>24)); zb[2]=(char)zval; } }
-    t=pack&0xf0000000; if(t) { if(zb[3]) collision=1; if(zval>zb[3]) { pd[3]=(unsigned char)(pal|(t>>28)); zb[3]=(char)zval; } }
-    t=pack&0x0000000f; if(t) { if(zb[4]) collision=1; if(zval>zb[4]) { pd[4]=(unsigned char)(pal|(t    )); zb[4]=(char)zval; } }
-    t=pack&0x000000f0; if(t) { if(zb[5]) collision=1; if(zval>zb[5]) { pd[5]=(unsigned char)(pal|(t>> 4)); zb[5]=(char)zval; } }
-    t=pack&0x00000f00; if(t) { if(zb[6]) collision=1; if(zval>zb[6]) { pd[6]=(unsigned char)(pal|(t>> 8)); zb[6]=(char)zval; } }
-    t=pack&0x0000f000; if(t) { if(zb[7]) collision=1; if(zval>zb[7]) { pd[7]=(unsigned char)(pal|(t>>12)); zb[7]=(char)zval; } }
-    if (collision) Pico.video.status|=0x20;
+    tilepixelAS(0x000f0000, 0, 16);
+    tilepixelAS(0x00f00000, 1, 20);
+    tilepixelAS(0x0f000000, 2, 24);
+    tilepixelAS(0xf0000000, 3, 28);
+    tilepixelAS(0x0000000f, 4,  0);
+    tilepixelAS(0x000000f0, 5,  4);
+    tilepixelAS(0x00000f00, 6,  8);
+    tilepixelAS(0x0000f000, 7, 12);
     return 0;
   }
   return 1; // Tile blank
 }
 
+// there is a problem with transparent hi pri tiles (on layer), it will clear high bits
+// and sprite tiles will be drawn needlessly. Hopefully that won't happen much..
+#define sh_pixAS(x) \
+  if(!t); \
+  else if(t==0xe) pd[x]=(unsigned char)((pd[x]&0x3f)|0x80); /* hilight */ \
+  else if(t==0xf) pd[x]=(unsigned char)( pd[x]      |0xc0); /* shadow  */ \
+  else if(!(pd[x]&0xc0)) pd[x]=(unsigned char)(pal|t)
 
-#define sh_pixZ(x) \
-  if(t) { \
-    if(zb[x]) collision=1; \
-    if(zval>zb[x]) { \
-      if     (t==0xe) { pd[x]=(unsigned char)((pd[x]&0x3f)|0x80); /* hilight */ } \
-      else if(t==0xf) { pd[x]=(unsigned char)((pd[x]&0x3f)|0xc0); /* shadow  */ } \
-      else            { zb[x]=(char)zval; pd[x]=(unsigned char)(pal|t); } \
-    } \
-  }
-
-static int TileNormZSH(int sx,int addr,int pal,int zval)
+static int TileNormSHAS(int sx,int addr,int pal)
 {
   unsigned int pack=0; unsigned int t=0;
   unsigned char *pd = HighCol+sx;
-  unsigned char *zb = HighSprZ+sx;
-  int collision = 0;
 
   pack=*(unsigned int *)(Pico.vram+addr); // Get 8 pixels
   if (pack)
   {
-    t=(pack&0x0000f000)>>12; sh_pixZ(0);
-    t=(pack&0x00000f00)>> 8; sh_pixZ(1);
-    t=(pack&0x000000f0)>> 4; sh_pixZ(2);
-    t=(pack&0x0000000f)    ; sh_pixZ(3);
-    t=(pack&0xf0000000)>>28; sh_pixZ(4);
-    t=(pack&0x0f000000)>>24; sh_pixZ(5);
-    t=(pack&0x00f00000)>>20; sh_pixZ(6);
-    t=(pack&0x000f0000)>>16; sh_pixZ(7);
-    if(collision) Pico.video.status|=0x20;
+    t=(pack&0x0000f000)>>12; sh_pixAS(0);
+    t=(pack&0x00000f00)>> 8; sh_pixAS(1);
+    t=(pack&0x000000f0)>> 4; sh_pixAS(2);
+    t=(pack&0x0000000f)    ; sh_pixAS(3);
+    t=(pack&0xf0000000)>>28; sh_pixAS(4);
+    t=(pack&0x0f000000)>>24; sh_pixAS(5);
+    t=(pack&0x00f00000)>>20; sh_pixAS(6);
+    t=(pack&0x000f0000)>>16; sh_pixAS(7);
     return 0;
   }
 
   return 1; // Tile blank
 }
 
-static int TileFlipZSH(int sx,int addr,int pal,int zval)
+static int TileFlipSHAS(int sx,int addr,int pal)
 {
   unsigned int pack=0; unsigned int t=0;
   unsigned char *pd = HighCol+sx;
-  unsigned char *zb = HighSprZ+sx;
-  int collision = 0;
 
   pack=*(unsigned int *)(Pico.vram+addr); // Get 8 pixels
   if (pack)
   {
-    t=(pack&0x000f0000)>>16; sh_pixZ(0);
-    t=(pack&0x00f00000)>>20; sh_pixZ(1);
-    t=(pack&0x0f000000)>>24; sh_pixZ(2);
-    t=(pack&0xf0000000)>>28; sh_pixZ(3);
-    t=(pack&0x0000000f)    ; sh_pixZ(4);
-    t=(pack&0x000000f0)>> 4; sh_pixZ(5);
-    t=(pack&0x00000f00)>> 8; sh_pixZ(6);
-    t=(pack&0x0000f000)>>12; sh_pixZ(7);
-    if(collision) Pico.video.status|=0x20;
+    t=(pack&0x000f0000)>>16; sh_pixAS(0);
+    t=(pack&0x00f00000)>>20; sh_pixAS(1);
+    t=(pack&0x0f000000)>>24; sh_pixAS(2);
+    t=(pack&0xf0000000)>>28; sh_pixAS(3);
+    t=(pack&0x0000000f)    ; sh_pixAS(4);
+    t=(pack&0x000000f0)>> 4; sh_pixAS(5);
+    t=(pack&0x00000f00)>> 8; sh_pixAS(6);
+    t=(pack&0x0000f000)>>12; sh_pixAS(7);
     return 0;
   }
   return 1; // Tile blank
 }
 
+
 // --------------------------------------------
 
 #ifndef _ASM_DRAW_C
@@ -725,7 +726,7 @@ last_cut_tile:
 // Index + 0  :    hhhhvvvv ab--hhvv yyyyyyyy yyyyyyyy // a: offscreen h, b: offs. v, h: horiz. size
 // Index + 4  :    xxxxxxxx xxxxxxxx pccvhnnn nnnnnnnn // x: x coord + 8
 
-static void DrawSprite(int *sprite, int **hc, int sh)
+static void DrawSprite(int *sprite, int **hc, int sh, int as)
 {
   int width=0,height=0;
   int row=0,code=0;
@@ -753,74 +754,21 @@ static void DrawSprite(int *sprite, int **hc, int sh)
 
   tile<<=4; tile+=(row&7)<<1; // Tile address
 
-  if(code&0x8000) { // high priority - cache it
+  if (code&0x8000) { // high priority - cache it
     *(*hc)++ = (tile<<16)|((code&0x0800)<<5)|((sx<<6)&0x0000ffc0)|((code>>9)&0x30)|((sprite[0]>>16)&0xf);
-  } else {
-    delta<<=4; // Delta of address
-    pal=((code>>9)&0x30)|(sh<<6);
-
-    if(sh && (code&0x6000) == 0x6000) {
-      if(code&0x0800) fTileFunc=TileFlipSH;
-      else            fTileFunc=TileNormSH;
-    } else {
-      if(code&0x0800) fTileFunc=TileFlip;
-      else            fTileFunc=TileNorm;
-    }
-
-    for (; width; width--,sx+=8,tile+=delta)
-    {
-      if(sx<=0)   continue;
-      if(sx>=328) break; // Offscreen
-
-      tile&=0x7fff; // Clip tile address
-      fTileFunc(sx,tile,pal);
-    }
+    // we need all for accurate sprites, cached will be used to recover ones overwritten by high layer
+    if (!as) return;
   }
-}
-#endif
-
-
-// Index + 0  :    hhhhvvvv s---hhvv yyyyyyyy yyyyyyyy // s: skip flag, h: horiz. size
-// Index + 4  :    xxxxxxxx xxxxxxxx pccvhnnn nnnnnnnn // x: x coord + 8
-
-static void DrawSpriteZ(int pack, int pack2, int shpri, int sprio)
-{
-  int width=0,height=0;
-  int row=0;
-  int pal;
-  int tile=0,delta=0;
-  int sx, sy;
-  int (*fTileFunc)(int sx,int addr,int pal,int zval);
-
-  // parse the sprite data
-  sx=pack2>>16; // X
-  sy=(pack <<16)>>16; // Y
-  width=pack>>28;
-  height=(pack>>24)&7; // Width and height in tiles
 
-  row=Scanline-sy; // Row of the sprite we are on
-
-  if (pack2&0x1000) row=(height<<3)-1-row; // Flip Y
-
-  tile=pack2&0x7ff; // Tile number
-  tile+=row>>3; // Tile number increases going down
-  delta=height; // Delta to increase tile by going right
-  if (pack2&0x0800) { tile+=delta*(width-1); delta=-delta; } // Flip X
-
-  tile<<=4; tile+=(row&7)<<1; // Tile address
   delta<<=4; // Delta of address
-  pal=((pack2>>9)&0x30);
-  if((shpri&1)&&!(shpri&2)) pal|=0x40;
-
-  shpri&=1;
-  if((pack2&0x6000) != 0x6000) shpri = 0;
-  shpri |= (pack2&0x0800)>>10;
-  switch(shpri) {
-    default:
-    case 0: fTileFunc=TileNormZ;   break;
-    case 1: fTileFunc=TileNormZSH; break;
-    case 2: fTileFunc=TileFlipZ;   break;
-    case 3: fTileFunc=TileFlipZSH; break;
+  pal=((code>>9)&0x30)|((sh|as)<<6);
+
+  if (sh && (code&0x6000) == 0x6000) {
+    if(code&0x0800) fTileFunc=TileFlipSH;
+    else            fTileFunc=TileNormSH;
+  } else {
+    if(code&0x0800) fTileFunc=TileFlip;
+    else            fTileFunc=TileNorm;
   }
 
   for (; width; width--,sx+=8,tile+=delta)
@@ -829,9 +777,10 @@ static void DrawSpriteZ(int pack, int pack2, int shpri, int sprio)
     if(sx>=328) break; // Offscreen
 
     tile&=0x7fff; // Clip tile address
-    fTileFunc(sx,tile,pal,sprio);
+    fTileFunc(sx,tile,pal);
   }
 }
+#endif
 
 static void DrawSpriteInterlace(unsigned int *sprite)
 {
@@ -877,7 +826,7 @@ static void DrawSpriteInterlace(unsigned int *sprite)
 }
 
 
-static void DrawAllSpritesInterlace(int pri, int maxwidth)
+static void DrawAllSpritesInterlace(int *hcache, int maxwidth, int pri, int sh)
 {
   struct PicoVideo *pvid=&Pico.video;
   int i,u,table,link=0,sline=Scanline<<1;
@@ -925,7 +874,7 @@ static void DrawAllSpritesInterlace(int pri, int maxwidth)
 
 
 #ifndef _ASM_DRAW_C
-static void DrawSpritesFromCache(int *hc, int sh)
+static void DrawSpritesFromCache(int *hc, int maxwidth, int prio, int sh)
 {
   int code, tile, sx, delta, width;
   int pal;
@@ -963,6 +912,43 @@ static void DrawSpritesFromCache(int *hc, int sh)
 }
 #endif
 
+static void DrawSpritesFromCacheAS(int *hc, int maxwidth, int prio, int sh)
+{
+  int code, tile, sx, delta, width;
+  int pal;
+  int (*fTileFunc)(int sx,int addr,int pal);
+
+  // *(*hc)++ = (tile<<16)|((code&0x0800)<<5)|((sx<<6)&0x0000ffc0)|((code>>9)&0x30)|((sprite[0]>>24)&0xf);
+
+  while((code=*hc++)) {
+    pal=(code&0x30);
+    delta=code&0xf;
+    width=delta>>2; delta&=3;
+    width++; delta++; // Width and height in tiles
+    if (code&0x10000) delta=-delta; // Flip X
+    delta<<=4;
+    tile=((unsigned int)code>>17)<<1;
+    sx=(code<<16)>>22; // sx can be negative (start offscreen), so sign extend
+
+    if(sh && pal == 0x30) { //
+      if(code&0x10000) fTileFunc=TileFlipSHAS;
+      else             fTileFunc=TileNormSHAS;
+    } else {
+      if(code&0x10000) fTileFunc=TileFlipAS;
+      else             fTileFunc=TileNormAS;
+    }
+
+    for (; width; width--,sx+=8,tile+=delta)
+    {
+      if(sx<=0)   continue;
+      if(sx>=328) break; // Offscreen
+
+      tile&=0x7fff; // Clip tile address
+      fTileFunc(sx,tile,pal);
+    }
+  }
+}
+
 
 // Index + 0  :    ----hhvv -lllllll -------y yyyyyyyy
 // Index + 4  :    -------x xxxxxxxx pccvhnnn nnnnnnnn
@@ -1066,23 +1052,12 @@ static void DrawAllSprites(int *hcache, int maxwidth, int prio, int sh)
   int max_line_sprites = 20; // 20 sprites, 40 tiles
   int *ps, pack, rs = rendstatus, scan = Scanline;
 
-  if(rs&8) {
-    DrawAllSpritesInterlace(prio, maxwidth);
-    return;
-  }
   if (rs & (PDRAW_SPRITES_MOVED|PDRAW_DIRTY_SPRITES)) {
     //dprintf("PrepareSprites(%i) [%i]", (rs>>4)&1, scan);
     PrepareSprites(rs & PDRAW_DIRTY_SPRITES);
     rendstatus = rs & ~(PDRAW_SPRITES_MOVED|PDRAW_DIRTY_SPRITES);
   }
-  if (!(SpriteBlocks & (1<<(scan>>3)))) return;
-
-  if (((rs&PDRAW_ACC_SPRITES)||sh) && prio==0)
-    memset(HighSprZ, 0, 328);
-  if (!(rs&PDRAW_ACC_SPRITES)&&prio) {
-    if(hcache[0]) DrawSpritesFromCache(hcache, sh);
-    return;
-  }
+  if (!(SpriteBlocks & (1<<(scan>>3)))) { *hcache = 0; return; }
 
   if (PicoOpt & POPT_DIS_SPRITE_LIM)
     max_line_sprites = 80;
@@ -1125,27 +1100,18 @@ static void DrawAllSprites(int *hcache, int maxwidth, int prio, int sh)
 
     if (pack & 0x00800000) continue;
 
-    // accurate sprites
-    //dprintf("P:%i",((sx>>15)&1));
-    if (rs & PDRAW_ACC_SPRITES) {
-      // might need to skip this sprite
-      if ((pack2&0x8000) ^ (prio<<15)) continue;
-      DrawSpriteZ(pack, pack2, sh|(prio<<1), n^0xff);
-      continue;
-    }
-
     // sprite is good, save it's pointer
     sprites[i++]=ps;
   }
 
+  n = (rs & PDRAW_ACC_SPRITES) ? 1 : 0;
+
   // Go through sprites backwards:
-  if (!(rs & PDRAW_ACC_SPRITES)) {
-    for (i--; i>=0; i--)
-      DrawSprite(sprites[i],&hcache,sh);
+  for (i--; i>=0; i--)
+    DrawSprite(sprites[i],&hcache,sh,n);
 
-    // terminate cache list
-    *hcache = 0;
-  }
+  // terminate cache list
+  *hcache = 0;
 }
 
 
@@ -1176,7 +1142,7 @@ static void FinalizeLineBGR444(int sh)
   unsigned short *pd=DrawLineDest;
   unsigned char  *ps=HighCol+8;
   unsigned short *pal=Pico.cram;
-  int len, i, t;
+  int len, i, t, mask=0xff;
 
   if (Pico.video.reg[12]&1) {
     len = 320;
@@ -1201,8 +1167,11 @@ static void FinalizeLineBGR444(int sh)
     }
   }
 
+  if (!sh && (rendstatus & PDRAW_ACC_SPRITES))
+    mask=0x3f; // accurate sprites
+
   for(i = 0; i < len; i++)
-    pd[i] = pal[ps[i]];
+    pd[i] = pal[ps[i] & mask];
 }
 
 
@@ -1247,15 +1216,22 @@ static void FinalizeLineRGB555(int sh)
     len = 256;
   }
 
+  {
 #ifndef PSP
-  for (i = 0; i < len; i++)
-    pd[i] = pal[ps[i]];
+    int mask=0xff;
+    if (!sh && (rendstatus & PDRAW_ACC_SPRITES))
+      mask=0x3f; // accurate sprites, upper bits are priority stuff
+
+    for (i = 0; i < len; i++)
+      pd[i] = pal[ps[i] & mask];
 #else
-  {
     extern void amips_clut(unsigned short *dst, unsigned char *src, unsigned short *pal, int count);
-    amips_clut(pd, ps, pal, len);
-  }
+    extern void amips_clut_6bit(unsigned short *dst, unsigned char *src, unsigned short *pal, int count);
+    if (!sh && (rendstatus & PDRAW_ACC_SPRITES))
+         amips_clut_6bit(pd, ps, pal, len);
+    else amips_clut(pd, ps, pal, len);
 #endif
+  }
 }
 #endif
 
@@ -1265,7 +1241,7 @@ static void FinalizeLine8bit(int sh)
   int len, rs = rendstatus;
   static int dirty_count;
 
-  if (!sh && Pico.m.dirtyPal == 1 && Scanline < 222)
+  if (!sh && !(rs & PDRAW_ACC_SPRITES) && Pico.m.dirtyPal == 1 && Scanline < 222)
   {
     // a hack for mid-frame palette changes
     if (!(rs & PDRAW_SONIC_MODE))
@@ -1320,7 +1296,7 @@ static void handle_early_blank(int scanline, int sh)
 
 // --------------------------------------------
 
-static int DrawDisplay(int sh)
+static int DrawDisplay(int sh, int as)
 {
   struct PicoVideo *pvid=&Pico.video;
   int win=0,edge=0,hvwind=0;
@@ -1354,16 +1330,16 @@ static int DrawDisplay(int sh)
     }
   }
 
-  DrawLayer(1|(sh<<1), HighCacheB, 0, maxcells);
+  DrawLayer(1|((sh|as)<<1), HighCacheB, 0, maxcells);
   if (hvwind == 1)
-    DrawWindow(0, maxcells>>1, 0, sh);
+    DrawWindow(0, maxcells>>1, 0, sh|as);
   else if (hvwind == 2) {
     // ahh, we have vertical window
-    DrawLayer(0|(sh<<1), HighCacheA, (win&0x80) ?    0 : edge<<1, (win&0x80) ?     edge<<1 : maxcells);
-    DrawWindow(                      (win&0x80) ? edge :       0, (win&0x80) ? maxcells>>1 : edge, 0, sh);
+    DrawLayer(0|((sh|as)<<1), HighCacheA, (win&0x80) ?    0 : edge<<1, (win&0x80) ?     edge<<1 : maxcells);
+    DrawWindow(                           (win&0x80) ? edge :       0, (win&0x80) ? maxcells>>1 : edge, 0, sh|as);
   } else
-    DrawLayer(0|(sh<<1), HighCacheA, 0, maxcells);
-  DrawAllSprites(HighCacheS, maxw, 0, sh);
+    DrawLayer(0|((sh|as)<<1), HighCacheA, 0, maxcells);
+  DrawAllSpritesLoPri(HighCacheS, maxw, 0, sh);
 
   if (HighCacheB[0]) DrawTilesFromCache(HighCacheB, sh, 328);
   if (hvwind == 1)
@@ -1373,7 +1349,7 @@ static int DrawDisplay(int sh)
     DrawWindow((win&0x80) ? edge : 0, (win&0x80) ? maxcells>>1 : edge, 1, sh);
   } else
     if (HighCacheA[0]) DrawTilesFromCache(HighCacheA, sh, 328);
-  DrawAllSprites(HighCacheS, maxw, 1, sh);
+  DrawAllSpritesHiPri(HighCacheS, maxw, 1, sh);
 
 #if 0
   {
@@ -1392,10 +1368,17 @@ PICO_INTERNAL void PicoFrameStart(void)
 {
   // prepare to do this frame
   rendstatus = (PicoOpt&0x80)>>5;    // accurate sprites, clear everything else
-  if (rendstatus)
-       Pico.video.status &= ~0x0020;
-  else Pico.video.status |=  0x0020; // sprite collision
-  if ((Pico.video.reg[12]&6) == 6) rendstatus |= PDRAW_INTERLACE; // interlace mode
+  if ((Pico.video.reg[12]&6) == 6) {
+    rendstatus |= PDRAW_INTERLACE; // interlace mode
+    DrawAllSpritesLoPri = DrawAllSpritesInterlace;
+    DrawAllSpritesHiPri = DrawAllSpritesInterlace;
+  }
+  else
+  {
+    DrawAllSpritesLoPri = DrawAllSprites;
+    DrawAllSpritesHiPri = rendstatus ? DrawSpritesFromCacheAS : DrawSpritesFromCache;
+  }
+
   if (Pico.m.dirtyPal) Pico.m.dirtyPal = 2; // reset dirty if needed
 
   PrepareSprites(1);
@@ -1404,11 +1387,12 @@ PICO_INTERNAL void PicoFrameStart(void)
 
 PICO_INTERNAL int PicoLine(int scan)
 {
-  int sh;
+  int sh, as = 0;
   if (skip_next_line>0) { skip_next_line--; return 0; } // skip_next_line rendering lines
 
   Scanline=scan;
   sh=(Pico.video.reg[0xC]&8)>>3; // shadow/hilight?
+  if (rendstatus & PDRAW_ACC_SPRITES) as|=1; // accurate sprites
 
   if (rendstatus & PDRAW_EARLY_BLANK) {
     if (scan > 0) handle_early_blank(scan, sh);
@@ -1419,9 +1403,9 @@ PICO_INTERNAL int PicoLine(int scan)
     skip_next_line = PicoScanBegin(scan);
 
   // Draw screen:
-  BackFill(Pico.video.reg[7], sh);
+  BackFill(Pico.video.reg[7], sh|as);
   if (Pico.video.reg[1]&0x40)
-    DrawDisplay(sh);
+    DrawDisplay(sh, as);
 
   if (FinalizeLine != NULL)
     FinalizeLine(sh);

+ 41 - 26
Pico/Draw.s

@@ -896,7 +896,7 @@ DrawTilesFromCache:
 @ @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
 
 
-.global DrawSpritesFromCache @ int *hc, int sh
+.global DrawSpritesFromCache @ int *hc, int maxwidth, int prio, int sh
 
 DrawSpritesFromCache:
     stmfd   sp!, {r4-r11,lr}
@@ -911,7 +911,7 @@ DrawSpritesFromCache:
     mov     r12,#0xf
 .endif
     ldr     lr, =(Pico+0x10000) @ lr=Pico.vram
-    mov     r6, r1, lsl #31
+    mov     r6, r3, lsl #31
     orr     r6, r6, #1<<30
 
     mov     r10, r0
@@ -1032,11 +1032,12 @@ DrawSpritesFromCache:
 @ + 0  :    hhhhvvvv ab--hhvv yyyyyyyy yyyyyyyy // a: offscreen h, b: offs. v, h: horiz. size
 @ + 4  :    xxxxxxxx xxxxxxxx pccvhnnn nnnnnnnn // x: x coord + 8
 
-.global DrawSprite @ unsigned int *sprite, int **hc, int sh
+.global DrawSprite @ unsigned int *sprite, int **hc, int sh, int acc_sprites
 
 DrawSprite:
     stmfd   sp!, {r4-r9,r11,lr}
 
+    orr     r8, r3, r2, lsl #4
     ldr     r3, [r0]        @ sprite[0]
     ldr     r7, =Scanline
     mov     r6, r3, lsr #28
@@ -1050,10 +1051,10 @@ DrawSprite:
     ldr     r9, [r0, #4]
     sub     r7, r7, r4, asr #16 @ r7=row=Scanline-sy
 
-    tst     r2, r2
     mov     r2, r9, asr #16 @ r2=sx
-    bic     r9, r9, #0xfe000000
-    orrne   r9, r9, #1<<31  @ r9=code|(sh<<31)
+    mov     r9, r9, lsl #16
+    mov     r9, r9, lsr #16
+    orr     r9, r9, r8, lsl #27 @ r9=code|sh[31]|as[27]
 
     tst     r9, #0x1000
     movne   r4, r5, lsl #3
@@ -1075,6 +1076,7 @@ DrawSprite:
     tst     r9, #0x8000
     bne     .dspr_cache       @ if(code&0x8000) // high priority - cache it
 
+.dspr_continue:
     @ cache some stuff to avoid mem access
 .if OVERRIDE_HIGHCOL
     ldr     r11,=HighCol
@@ -1089,11 +1091,10 @@ DrawSprite:
     mov     r5, r5, lsl #4     @ delta<<=4; // Delta of address
     and     r4, r9, #0x6000
     orr     r9, r9, r4, lsl #16
-    orr     r9, r9, #0x10000000 @ r9=scc1 ???? ... <code> (s=shadow/hilight, cc=pal)
+    orrs    r9, r9, #0x10000000 @ r9=scc1 a??? ... <code> (s=shadow/hilight, cc=pal, a=acc_spr)
 
-    tst     r9, #1<<31
     mov     r3, r4, lsr #9     @ r3=pal=((code>>9)&0x30);
-    orrne   r3, r3, #0x40      @ shadow by default
+    orrmi   r3, r3, #0x40      @ shadow by default
 
     add     r6, r6, #1         @ inc now
     adds    r0, r2, #0         @ mov sx to r0 and set ZV flags
@@ -1184,19 +1185,21 @@ DrawSprite:
     mov     r4, r8, lsl #16     @ tile
     tst     r9, #0x0800
     orrne   r4, r4, #0x10000    @ code&0x0800
-    mov     r2, r2, lsl #22
-    orr     r4, r4, r2, lsr #16 @ (sx<<6)&0x0000ffc0
-    and     r2, r9, #0x6000
-    orr     r4, r4, r2, lsr #9  @ (code>>9)&0x30
+    mov     r0, r2, lsl #22
+    orr     r4, r4, r0, lsr #16 @ (sx<<6)&0x0000ffc0
+    and     r0, r9, #0x6000
+    orr     r4, r4, r0, lsr #9  @ (code>>9)&0x30
     mov     r3, r3, lsl #12
-    ldr     r2, [r1]
+    ldr     r0, [r1]
     orr     r4, r4, r3, lsr #28 @ (sprite[0]>>24)&0xf
 
-    str     r4, [r2], #4
-    str     r2, [r1]
+    str     r4, [r0], #4
+    str     r0, [r1]
 
-    ldmfd   sp!, {r4-r9,r11,lr}
-    bx      lr
+    tst     r9, #(1<<27)
+    ldmeqfd sp!, {r4-r9,r11,lr}
+    bne     .dspr_continue      @ draw anyway if accurate sprites enabled
+    bxeq    lr
 
 @ @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
 
@@ -1452,23 +1455,29 @@ FinalizeLineBGR444:
     bne     .fl_loopcpBGR444_hi
 
     sub     r3, r4, #0x40*3*2
+    mov     r6, #1
 
 
 .fl_noshBGR444:
+    ldr     r12,=rendstatus
+    eors    r6, r6, #1          @ sh is 0
+    ldr     r12,[r12]
+    mov     lr, #0xff
+    tstne   r12,#(1<<2)         @ and PDRAW_ACC_SPRITES
+
 .if OVERRIDE_HIGHCOL
     ldr     r1, =HighCol
-    mov     lr, #0xff
+    movne   lr, #0x3f
     ldr     r1, [r1]
     mov     lr, lr, lsl #1
     add     r1, r1, #8
 .else
     ldr     r1, =(HighCol+8)
-    mov     lr, #0xff
+    movne   lr, #0x3f
     mov     lr, lr, lsl #1
 .endif
 
 .fl_loopBGR444:
-
     ldr     r12, [r1], #4
     subs    r2, r2, #1
 
@@ -1478,11 +1487,10 @@ FinalizeLineBGR444:
     ldrh    r5, [r3, r5]
     and     r6, lr, r12, lsr #15
     ldrh    r6, [r3, r6]
+    and     r12,lr, r12, lsr #23
+    ldrh    r12,[r3, r12]              @ 1c.i.
     orr     r4, r4, r5, lsl #16
-
-    and     r5, lr, r12, lsr #23
-    ldrh    r5, [r3, r5]              @ 2c.i.
-    orr     r5, r6, r5, lsl #16
+    orr     r5, r6, r12,lsl #16
 
     stmia   r0!, {r4,r5}
     bne     .fl_loopBGR444
@@ -1617,8 +1625,16 @@ FinalizeLineRGB555:
     bne     .fl_loopcpRGB555_hi
 
     sub     r3, r3, #0x40*2
+    mov     r6, #1
 
 .fl_noshRGB555:
+    ldr     r12,=rendstatus
+    eors    r6, r6, #1          @ sh is 0
+    ldr     r12,[r12]
+    mov     lr, #0xff
+    tstne   r12,#(1<<2)         @ and PDRAW_ACC_SPRITES
+    movne   lr, #0x3f
+
 .if OVERRIDE_HIGHCOL
     ldr     r1, =HighCol
     ldr     r0, =DrawLineDest
@@ -1632,7 +1648,6 @@ FinalizeLineRGB555:
 .endif
 
     ldrb    r12, [r8, #12]
-    mov     lr, #0xff
     mov     lr, lr, lsl #1
 
     tst     r12, #1

+ 37 - 0
Pico/Draw_amips.s

@@ -45,6 +45,43 @@ amips_clut_loop:
     nop
 
 
+.global amips_clut_6bit
+
+amips_clut_6bit:
+    srl     $a3, 2
+    li      $t4, 0
+    li      $t5, 0
+    li      $t6, 0
+    li      $t7, 0
+amips_clut_loop6:
+    lbu     $t0, 0($a1)           # tried lw here, no improvement noticed
+    lbu     $t1, 1($a1)
+    lbu     $t2, 2($a1)
+    lbu     $t3, 3($a1)
+    ins     $t4, $t0, 1, 6
+    ins     $t5, $t1, 1, 6
+    ins     $t6, $t2, 1, 6
+    ins     $t7, $t3, 1, 6
+    addu    $t0, $t4, $a2
+    addu    $t1, $t5, $a2
+    addu    $t2, $t6, $a2
+    addu    $t3, $t7, $a2
+    lhu     $t0, 0($t0)
+    lhu     $t1, 0($t1)
+    lhu     $t2, 0($t2)
+    lhu     $t3, 0($t3)
+    ins     $t0, $t1, 16, 16      # ins rt, rs, pos, size - Insert size bits starting
+    ins     $t2, $t3, 16, 16      #  from the LSB of rs into rt starting at position pos
+    sw      $t0, 0($a0)
+    sw      $t2, 4($a0)
+    addiu   $a0, 8
+    addiu   $a3, -1
+    bnez    $a3, amips_clut_loop6
+    addiu   $a1, 4
+    jr      $ra
+    nop
+
+
 # $a0 - pd, $a1 - tile word, $a2 - pal
 # ext rt, rs, pos, size  // Extract size bits from position pos in rs and store in rt 
 

+ 2 - 2
Pico/Pico.c

@@ -157,7 +157,7 @@ int PicoReset(void)
   Pico.m.dirtyPal = 1;
 
   PicoDetectRegion();
-  Pico.video.status = 0x3408 | Pico.m.pal; // 'always set' bits | vblank | pal
+  Pico.video.status = 0x3428 | Pico.m.pal; // 'always set' bits | vblank | collision | pal
 
   PsndReset(); // pal must be known here
 
@@ -320,7 +320,7 @@ PICO_INTERNAL void PicoSyncZ80(int m68k_cycles_done)
   z80_cycle_aim = cycles_68k_to_z80(m68k_cycles_done);
   cnt = z80_cycle_aim - z80_cycle_cnt;
 
-  elprintf(EL_ANOMALY, "z80 sync %i (%i|%i -> %i|%i)", cnt, z80_cycle_cnt, z80_cycle_cnt / 228,
+  elprintf(EL_BUSREQ, "z80 sync %i (%i|%i -> %i|%i)", cnt, z80_cycle_cnt, z80_cycle_cnt / 228,
     z80_cycle_aim, z80_cycle_aim / 228);
 
   if (cnt > 0)

+ 23 - 10
platform/gp2x/emu.c

@@ -291,26 +291,39 @@ static void blit(const char *fps, const char *notice)
 		// 8bit accurate renderer
 		if (Pico.m.dirtyPal)
 		{
+			int pallen = 0x40;
 			Pico.m.dirtyPal = 0;
-			if(Pico.video.reg[0xC]&8) { // shadow/hilight mode
+			if (Pico.video.reg[0xC]&8) // shadow/hilight mode
+			{
 				vidConvCpyRGB32(localPal, Pico.cram, 0x40);
 				vidConvCpyRGB32sh(localPal+0x40, Pico.cram, 0x40);
 				vidConvCpyRGB32hi(localPal+0x80, Pico.cram, 0x40);
 				memcpy32(localPal+0xc0, localPal+0x40, 0x40);
-				localPal[0xc0] = 0x0000c000;
-				localPal[0xd0] = 0x00c00000;
-				localPal[0xe0] = 0x00000000; // reserved pixels for OSD
-				localPal[0xf0] = 0x00ffffff;
-				gp2x_video_setpalette(localPal, 0x100);
-			} else if (rendstatus & PDRAW_SONIC_MODE) { // mid-frame palette changes
+				pallen = 0x100;
+			}
+			else if (rendstatus & PDRAW_ACC_SPRITES) {
+				vidConvCpyRGB32(localPal, Pico.cram, 0x40);
+				memcpy32(localPal+0x40, localPal, 0x40);
+				memcpy32(localPal+0x80, localPal, 0x40);
+				memcpy32(localPal+0xc0, localPal, 0x40);
+				pallen = 0x100;
+			}
+			else if (rendstatus & PDRAW_SONIC_MODE) { // mid-frame palette changes
 				vidConvCpyRGB32(localPal, Pico.cram, 0x40);
 				vidConvCpyRGB32(localPal+0x40, HighPal, 0x40);
 				vidConvCpyRGB32(localPal+0x80, HighPal+0x40, 0x40);
-				gp2x_video_setpalette(localPal, 0xc0);
-			} else {
+				pallen = 0xc0;
+			}
+			else {
 				vidConvCpyRGB32(localPal, Pico.cram, 0x40);
-				gp2x_video_setpalette(localPal, 0x40);
 			}
+			if (pallen > 0xc0) {
+				localPal[0xc0] = 0x0000c000;
+				localPal[0xd0] = 0x00c00000;
+				localPal[0xe0] = 0x00000000; // reserved pixels for OSD
+				localPal[0xf0] = 0x00ffffff;
+			}
+			gp2x_video_setpalette(localPal, pallen);
 		}
 	}
 

+ 21 - 8
platform/psp/emu.c

@@ -162,6 +162,9 @@ void emu_setDefaultConfig(void)
 
 
 extern void amips_clut(unsigned short *dst, unsigned char *src, unsigned short *pal, int count);
+extern void amips_clut_6bit(unsigned short *dst, unsigned char *src, unsigned short *pal, int count);
+
+extern void (*amips_clut_f)(unsigned short *dst, unsigned char *src, unsigned short *pal, int count) = NULL;
 
 struct Vertex
 {
@@ -235,7 +238,7 @@ static void set_scaling_params(void)
 	*/
 }
 
-static void do_pal_update(int allow_sh)
+static void do_pal_update(int allow_sh, int allow_as)
 {
 	unsigned int *dpal=(void *)localPal;
 	int i;
@@ -244,6 +247,9 @@ static void do_pal_update(int allow_sh)
 	//	dpal[i] = ((spal[i]&0x000f000f)<< 1)|((spal[i]&0x00f000f0)<<3)|((spal[i]&0x0f000f00)<<4);
 	do_pal_convert(localPal, Pico.cram, currentConfig.gamma, currentConfig.gamma2);
 
+	Pico.m.dirtyPal = 0;
+	need_pal_upload = 1;
+
 	if (allow_sh && (Pico.video.reg[0xC]&8)) // shadow/hilight?
 	{
 		// shadowed pixels
@@ -260,8 +266,12 @@ static void do_pal_update(int allow_sh)
 		}
 		localPal[0xe0] = 0;
 	}
-	Pico.m.dirtyPal = 0;
-	need_pal_upload = 1;
+	else if (allow_as && (rendstatus & PDRAW_ACC_SPRITES))
+	{
+		memcpy32(localPal+0x40, localPal, 0x40);
+		memcpy32(localPal+0x80, localPal, 0x40);
+		memcpy32(localPal+0xc0, localPal, 0x40);
+	}
 }
 
 static void do_slowmode_lines(int line_to)
@@ -272,7 +282,7 @@ static void do_slowmode_lines(int line_to)
 	if (!(Pico.video.reg[1]&8)) { line = 8; dst += 512*8; src += 512*8; }
 
 	for (; line < line_to; line++, dst+=512, src+=512)
-		amips_clut(dst, src, localPal, line_len);
+		amips_clut_f(dst, src, localPal, line_len);
 }
 
 static void EmuScanPrepare(void)
@@ -282,7 +292,10 @@ static void EmuScanPrepare(void)
 
 	dynamic_palette = 0;
 	if (Pico.m.dirtyPal)
-		do_pal_update(1);
+		do_pal_update(1, 1);
+	if ((rendstatus & PDRAW_ACC_SPRITES) && !(Pico.video.reg[0xC]&8))
+	     amips_clut_f = amips_clut_6bit;
+	else amips_clut_f = amips_clut;
 }
 
 static int EmuScanSlowBegin(unsigned int num)
@@ -304,13 +317,13 @@ static int EmuScanSlowEnd(unsigned int num)
 			do_slowmode_lines(num);
 			dynamic_palette = 1;
 		}
-		do_pal_update(1);
+		do_pal_update(1, 0);
 	}
 
 	if (dynamic_palette) {
 		int line_len = (Pico.video.reg[12]&1) ? 320 : 256;
 		void *dst = (char *)VRAM_STUFF + 512*240 + 512*2*num;
-		amips_clut(dst, HighCol + 8, localPal, line_len);
+		amips_clut_f(dst, HighCol + 8, localPal, line_len);
 	}
 
 	return 0;
@@ -344,7 +357,7 @@ static void blitscreen_clut(void)
 		}
 
 		if ((PicoOpt&0x10) && Pico.m.dirtyPal)
-			do_pal_update(0);
+			do_pal_update(0, 0);
 
 		sceKernelDcacheWritebackAll();