gfx_cd optimization wip, line glitch fixed
[picodrive.git] / Pico / cd / gfx_cd.c
1 // TODO...
2
3 // #include <string.h>
4 #include "../PicoInt.h"
5
6 #define rot_comp Pico_mcd->rot_comp
7
8 static const int Table_Rot_Time[] =
9 {
10         0x00054000, 0x00048000, 0x00040000, 0x00036000,          //; 008-032               ; briefing - sprite
11         0x0002E000, 0x00028000, 0x00024000, 0x00022000,          //; 036-064               ; arbre souvent
12         0x00021000, 0x00020000, 0x0001E000, 0x0001B800,          //; 068-096               ; map thunderstrike
13         0x00019800, 0x00017A00, 0x00015C00, 0x00013E00,          //; 100-128               ; logo défoncé
14
15         0x00012000, 0x00011800, 0x00011000, 0x00010800,          //; 132-160               ; briefing - map
16         0x00010000, 0x0000F800, 0x0000F000, 0x0000E800,          //; 164-192
17         0x0000E000, 0x0000D800, 0x0000D000, 0x0000C800,          //; 196-224
18         0x0000C000, 0x0000B800, 0x0000B000, 0x0000A800,          //; 228-256               ; batman visage
19
20         0x0000A000, 0x00009F00, 0x00009E00, 0x00009D00,          //; 260-288
21         0x00009C00, 0x00009B00, 0x00009A00, 0x00009900,          //; 292-320
22         0x00009800, 0x00009700, 0x00009600, 0x00009500,          //; 324-352
23         0x00009400, 0x00009300, 0x00009200, 0x00009100,          //; 356-384
24
25         0x00009000, 0x00008F00, 0x00008E00, 0x00008D00,          //; 388-416
26         0x00008C00, 0x00008B00, 0x00008A00, 0x00008900,          //; 420-448
27         0x00008800, 0x00008700, 0x00008600, 0x00008500,          //; 452-476
28         0x00008400, 0x00008300, 0x00008200, 0x00008100,          //; 480-512
29 };
30
31
32 static void gfx_cd_start(void)
33 {
34         int upd_len;
35
36         dprintf("gfx_cd_start()");
37
38         // rot_comp.XD_Mul = ((rot_comp.Reg_5C & 0x1f) + 1) * 4; // unused
39         rot_comp.Function = (rot_comp.Reg_58 & 7) | (Pico_mcd->s68k_regs[3] & 0x18);    // Jmp_Adr
40         // rot_comp.Buffer_Adr = (rot_comp.Reg_5E & 0xfff8) << 2; // unused?
41         rot_comp.YD = (rot_comp.Reg_60 >> 3) & 7;
42         rot_comp.Vector_Adr = (rot_comp.Reg_66 & 0xfffe) << 2;
43
44         upd_len = (rot_comp.Reg_62 >> 3) & 0x3f;
45         upd_len = Table_Rot_Time[upd_len];
46         rot_comp.Draw_Speed = rot_comp.Float_Part = upd_len;
47
48         rot_comp.Reg_58 |= 0x8000;      // Stamp_Size,  we start a new GFX operation
49
50         switch (rot_comp.Reg_58 & 6)    // Scr_16?
51         {
52                 case 0: // ?
53                         rot_comp.Stamp_Map_Adr = (rot_comp.Reg_5A & 0xff80) << 2;
54                         break;
55                 case 2: // .Dot_32
56                         rot_comp.Stamp_Map_Adr = (rot_comp.Reg_5A & 0xffe0) << 2;
57                         break;
58                 case 4: // .Scr_16
59                         rot_comp.Stamp_Map_Adr = 0x20000;
60                         break;
61                 case 6: // .Scr_16_Dot_32
62                         rot_comp.Stamp_Map_Adr = (rot_comp.Reg_5A & 0xe000) << 2;
63                         break;
64         }
65
66         gfx_cd_update();
67 }
68
69
70 static void gfx_completed(void)
71 {
72         rot_comp.Reg_58 &= 0x7fff;      // Stamp_Size
73         rot_comp.Reg_64  = 0;
74         if (Pico_mcd->s68k_regs[0x33] & (1<<1))
75         {
76                 dprintf("gfx_cd irq 1");
77                 SekInterruptS68k(1);
78         }
79 }
80
81
82 static void gfx_do(void)
83 {
84         unsigned int eax, ebx, ecx, edx, esi, edi, pixel;
85         unsigned int XD, Buffer_Adr, H_Dot;
86         unsigned int func = rot_comp.Function;
87         unsigned short *stamp_base;
88         int DYXS;
89
90         XD = rot_comp.Reg_60 & 7;
91         Buffer_Adr = ((rot_comp.Reg_5E & 0xfff8) + rot_comp.YD) << 2;
92         stamp_base = (unsigned short *) (Pico_mcd->word_ram2M + rot_comp.Stamp_Map_Adr);
93         H_Dot = rot_comp.Reg_62 & 0x1ff;
94         ecx = *(unsigned int *)(Pico_mcd->word_ram2M + rot_comp.Vector_Adr);
95         edx = ecx >> 16;
96         ecx = (ecx & 0xffff) << 8;
97         edx <<= 8;
98         DYXS = *(int *)(Pico_mcd->word_ram2M + rot_comp.Vector_Adr + 4);
99         rot_comp.Vector_Adr += 8;
100
101         // MAKE_IMAGE_LINE
102         while (H_Dot)
103         {
104                 if (func & 2)           // mode 32x32 dot
105                 {
106                         if (func & 4)   // 16x16 screen
107                         {
108                                 eax = (ecx >> (11+5)) & 0x007f;
109                                 ebx = (edx >> (11-2)) & 0x3f80;
110                         }
111                         else            // 1x1 screen
112                         {
113                                 eax = (ecx >> (11+5)) & 0x07;
114                                 ebx = (edx >> (11+2)) & 0x38;
115                         }
116                 }
117                 else                    // mode 16x16 dot
118                 {
119                         if (func & 4)   // 16x16 screen
120                         {
121                                 eax = (ecx >> (11+4)) & 0x00ff;
122                                 ebx = (edx >> (11-4)) & 0xff00;
123                         }
124                         else            // 1x1 screen
125                         {
126                                 eax = (ecx >> (11+4)) & 0x0f;
127                                 ebx = (edx >> (11+0)) & 0xf0;
128                         }
129                 }
130                 ebx += eax;
131
132                 // MAKE_IMAGE_PIXEL
133                 if (!(func & 1))        // NOT TILED
134                 {
135                         int mask = (func & 4) ? 0x00800000 : 0x00f80000;
136                         if ((ecx | edx) & mask)
137                         {
138                                 if (func & 0x18) goto Next_Pixel;
139                                 pixel = 0;
140                                 goto Pixel_Out;
141                         }
142                 }
143
144                 // esi = rot_comp.Stamp_Map_Adr;
145                 edi = stamp_base[ebx] | (stamp_base[ebx+1] << 16);
146                 esi = edi;
147                 edi >>= (11+1);
148                 esi = (esi & 0x7ff) << 7;
149                 if (!esi) { pixel = 0; goto Pixel_Out; }
150                 edi &= (0x1c>>1);
151                 eax = ecx;
152                 ebx = edx;
153                 if (func & 2) edi |= 1; // 32 dots?
154                 switch (edi)
155                 {
156                         case 0x00:      // No_Flip_0, 16x16 dots
157                                 ebx = (ebx >> 9) & 0x3c;
158                                 ebx += esi;
159                                 edi = (eax & 0x3800) ^ 0x1000;          // bswap
160                                 eax = ((eax >> 8) & 0x40) + ebx;
161                                 break;
162                         case 0x01:      // No_Flip_0, 32x32 dots
163                                 ebx = (ebx >> 9) & 0x7c;
164                                 ebx += esi;
165                                 edi = (eax & 0x3800) ^ 0x1000;          // bswap
166                                 eax = ((eax >> 7) & 0x180) + ebx;
167                                 break;
168                         case 0x02:      // No_Flip_90, 16x16 dots
169                                 eax = (eax >> 9) & 0x3c;
170                                 eax += esi;
171                                 edi = (ebx & 0x3800) ^ 0x2800;          // bswap
172                                 eax += ((ebx >> 8) & 0x40) ^ 0x40;
173                                 break;
174                         case 0x03:      // No_Flip_90, 32x32 dots
175                                 eax = (eax >> 9) & 0x7c;
176                                 eax += esi;
177                                 edi = (ebx & 0x3800) ^ 0x2800;          // bswap
178                                 eax += ((ebx >> 7) & 0x180) ^ 0x180;
179                                 break;
180                         case 0x04:      // No_Flip_180, 16x16 dots
181                                 ebx = ((ebx >> 9) & 0x3c) ^ 0x3c;
182                                 ebx += esi;
183                                 edi = (eax & 0x3800) ^ 0x2800;          // bswap and flip
184                                 eax = (((eax >> 8) & 0x40) ^ 0x40) + ebx;
185                                 break;
186                         case 0x05:      // No_Flip_180, 32x32 dots
187                                 ebx = ((ebx >> 9) & 0x7c) ^ 0x7c;
188                                 ebx += esi;
189                                 edi = (eax & 0x3800) ^ 0x2800;          // bswap and flip
190                                 eax = (((eax >> 7) & 0x180) ^ 0x180) + ebx;
191                                 break;
192                         case 0x06:      // No_Flip_270, 16x16 dots
193                                 eax = ((eax >> 9) & 0x3c) ^ 0x3c;
194                                 eax += esi;
195                                 edi = (ebx & 0x3800) ^ 0x1000;          // bswap
196                                 eax += (ebx >> 8) & 0x40;
197                                 break;
198                         case 0x07:      // No_Flip_270, 32x32 dots
199                                 eax = ((eax >> 9) & 0x7c) ^ 0x7c;
200                                 eax += esi;
201                                 edi = (ebx & 0x3800) ^ 0x1000;          // bswap
202                                 eax += (ebx >> 7) & 0x180;
203                                 break;
204                         case 0x08:      // Flip_0, 16x16 dots
205                                 ebx = (ebx >> 9) & 0x3c;
206                                 ebx += esi;
207                                 edi = (eax & 0x3800) ^ 0x2800;          // bswap, flip
208                                 eax = (((eax >> 8) & 0x40) ^ 0x40) + ebx;
209                                 break;
210                         case 0x09:      // Flip_0, 32x32 dots
211                                 ebx = (ebx >> 9) & 0x7c;
212                                 ebx += esi;
213                                 edi = (eax & 0x3800) ^ 0x2800;          // bswap, flip
214                                 eax = (((eax >> 7) & 0x180) ^ 0x180) + ebx;
215                                 break;
216                         case 0x0a:      // Flip_90, 16x16 dots
217                                 eax = ((eax >> 9) & 0x3c) ^ 0x3c;
218                                 eax += esi;
219                                 edi = (ebx & 0x3800) ^ 0x2800;          // bswap, flip
220                                 eax += ((ebx >> 8) & 0x40) ^ 0x40;
221                                 break;
222                         case 0x0b:      // Flip_90, 32x32 dots
223                                 eax = ((eax >> 9) & 0x7c) ^ 0x7c;
224                                 eax += esi;
225                                 edi = (ebx & 0x3800) ^ 0x2800;          // bswap, flip
226                                 eax += ((ebx >> 7) & 0x180) ^ 0x180;
227                                 break;
228                         case 0x0c:      // Flip_180, 16x16 dots
229                                 ebx = ((ebx >> 9) & 0x3c) ^ 0x3c;
230                                 ebx += esi;
231                                 edi = (eax & 0x3800) ^ 0x1000;          // bswap
232                                 eax = ((eax >> 8) & 0x40) + ebx;
233                                 break;
234                         case 0x0d:      // Flip_180, 32x32 dots
235                                 ebx = ((ebx >> 9) & 0x7c) ^ 0x7c;
236                                 ebx += esi;
237                                 edi = (eax & 0x3800) ^ 0x1000;          // bswap
238                                 eax = ((eax >> 7) & 0x180) + ebx;
239                                 break;
240                         case 0x0e:      // Flip_270, 16x16 dots
241                                 eax = (eax >> 9) & 0x3c;
242                                 eax += esi;
243                                 edi = (ebx & 0x3800) ^ 0x1000;          // bswap, flip
244                                 eax += (ebx >> 8) & 0x40;
245                                 break;
246                         case 0x0f:      // Flip_270, 32x32 dots
247                                 eax = (eax >> 9) & 0x7c;
248                                 eax += esi;
249                                 edi = (ebx & 0x3800) ^ 0x1000;          // bswap, flip
250                                 eax += (ebx >> 7) & 0x180;
251                                 break;
252                 }
253
254                 pixel = *(Pico_mcd->word_ram2M + (edi >> 12) + eax);
255                 if (!(edi & 0x800)) pixel >>= 4;
256                 else pixel &= 0x0f;
257
258 Pixel_Out:
259                 if (!pixel && (func & 0x18)) goto Next_Pixel;
260                 esi = Buffer_Adr + ((XD>>1)^1);                         // pixel addr
261                 eax = *(Pico_mcd->word_ram2M + esi);                    // old pixel
262                 if (XD & 1)
263                 {
264                         if ((eax & 0x0f) && (func & 0x18) == 0x08) goto Next_Pixel; // underwrite
265                         *(Pico_mcd->word_ram2M + esi) = pixel | (eax & 0xf0);
266                 }
267                 else
268                 {
269                         if ((eax & 0xf0) && (func & 0x18) == 0x08) goto Next_Pixel; // underwrite
270                         *(Pico_mcd->word_ram2M + esi) = (pixel << 4) | (eax & 0xf);
271                 }
272
273
274 Next_Pixel:
275                 ecx += (DYXS << 16) >> 16;      // rot_comp.DXS;
276                 edx +=  DYXS >> 16;             // rot_comp.DYS;
277                 XD++;
278                 if (XD >= 8)
279                 {
280                         Buffer_Adr += ((rot_comp.Reg_5C & 0x1f) + 1) << 5;
281                         XD = 0;
282                 }
283                 H_Dot--;
284         }
285         // end while
286
287
288 //nothing_to_draw:
289         rot_comp.YD++;
290         // rot_comp.V_Dot--; // will be done by caller
291 }
292
293
294 void gfx_cd_update(void)
295 {
296         unsigned char *V_Dot = (unsigned char *) &rot_comp.Reg_64;
297         int jobs;
298
299         dprintf("gfx_cd_update, Reg_64 = %04x", rot_comp.Reg_64);
300
301         if (!*V_Dot)
302         {
303                 gfx_completed();
304                 return;
305         }
306
307         jobs = rot_comp.Float_Part >> 16;
308
309         if (!jobs)
310         {
311                 rot_comp.Float_Part += rot_comp.Draw_Speed;
312                 return;
313         }
314
315         rot_comp.Float_Part &= 0xffff;
316         rot_comp.Float_Part += rot_comp.Draw_Speed;
317
318         while (jobs--)
319         {
320                 if (PicoOpt & 0x1000)
321                         gfx_do();       // jmp [Jmp_Adr]:
322
323                 (*V_Dot)--;             // dec byte [V_Dot]
324
325                 if (!*V_Dot)
326                 {
327                         // GFX_Completed:
328                         gfx_completed();
329                         return;
330                 }
331         }
332 }
333
334
335 unsigned int gfx_cd_read(unsigned int a)
336 {
337         unsigned int d = 0;
338
339         switch (a) {
340                 case 0x58: d = rot_comp.Reg_58; break;
341                 case 0x5A: d = rot_comp.Reg_5A; break;
342                 case 0x5C: d = rot_comp.Reg_5C; break;
343                 case 0x5E: d = rot_comp.Reg_5E; break;
344                 case 0x60: d = rot_comp.Reg_60; break;
345                 case 0x62: d = rot_comp.Reg_62; break;
346                 case 0x64: d = rot_comp.Reg_64; break;
347                 case 0x66: break;
348                 default: dprintf("gfx_cd_read FIXME: unexpected address: %02x", a); break;
349         }
350
351         dprintf("gfx_cd_read(%02x) = %04x", a, d);
352
353         return 0;
354 }
355
356 void gfx_cd_write16(unsigned int a, unsigned int d)
357 {
358         dprintf("gfx_cd_write16(%x, %04x)", a, d);
359
360         switch (a) {
361                 case 0x58: // .Reg_Stamp_Size
362                         rot_comp.Reg_58 = d & 7;
363                         return;
364
365                 case 0x5A: // .Reg_Stamp_Adr
366                         rot_comp.Reg_5A = d & 0xffe0;
367                         return;
368
369                 case 0x5C: // .Reg_IM_VCell_Size
370                         rot_comp.Reg_5C = d & 0x1f;
371                         return;
372
373                 case 0x5E: // .Reg_IM_Adr
374                         rot_comp.Reg_5E = d & 0xFFF8;
375                         return;
376
377                 case 0x60: // .Reg_IM_Offset
378                         rot_comp.Reg_60 = d & 0x3f;
379                         return;
380
381                 case 0x62: // .Reg_IM_HDot_Size
382                         rot_comp.Reg_62 = d & 0x1ff;
383                         return;
384
385                 case 0x64: // .Reg_IM_VDot_Size
386                         rot_comp.Reg_64 = d & 0xff;     // V_Dot, must be 32bit?
387                         return;
388
389                 case 0x66: // .Reg_Vector_Adr
390                         rot_comp.Reg_66 = d & 0xfffe;
391                         if (Pico_mcd->s68k_regs[3]&4) return; // can't do tanformations in 1M mode
392                         gfx_cd_start();
393                         return;
394
395                 default: dprintf("gfx_cd_write16 FIXME: unexpected address: %02x", a); return;
396         }
397 }
398
399
400 void gfx_cd_reset(void)
401 {
402         memset(&rot_comp.Reg_58, 0, sizeof(rot_comp));
403 }
404
405
406 // --------------------------------
407
408 #include "cell_map.c"
409
410 typedef unsigned short u16;
411
412 // check: Heart of the alien, jaguar xj 220
413 void DmaSlowCell(unsigned int source, unsigned int a, int len, unsigned char inc)
414 {
415   unsigned char *base;
416   unsigned int asrc, a2;
417   u16 *r;
418
419   base = Pico_mcd->word_ram1M[Pico_mcd->s68k_regs[3]&1];
420
421   switch (Pico.video.type)
422   {
423     case 1: // vram
424       r = Pico.vram;
425       for(; len; len--)
426       {
427         asrc = cell_map(source >> 2) << 2;
428         asrc |= source & 2;
429         // if(a&1) d=(d<<8)|(d>>8); // ??
430         r[a>>1] = *(u16 *)(base + asrc);
431         source += 2;
432         // AutoIncrement
433         a=(u16)(a+inc);
434       }
435       rendstatus|=0x10;
436       break;
437
438     case 3: // cram
439       Pico.m.dirtyPal = 1;
440       r = Pico.cram;
441       for(a2=a&0x7f; len; len--)
442       {
443         asrc = cell_map(source >> 2) << 2;
444         asrc |= source & 2;
445         r[a2>>1] = *(u16 *)(base + asrc);
446         source += 2;
447         // AutoIncrement
448         a2+=inc;
449         // good dest?
450         if(a2 >= 0x80) break;
451       }
452       a=(a&0xff00)|a2;
453       break;
454
455     case 5: // vsram[a&0x003f]=d;
456       r = Pico.vsram;
457       for(a2=a&0x7f; len; len--)
458       {
459         asrc = cell_map(source >> 2) << 2;
460         asrc |= source & 2;
461         r[a2>>1] = *(u16 *)(base + asrc);
462         source += 2;
463         // AutoIncrement
464         a2+=inc;
465         // good dest?
466         if(a2 >= 0x80) break;
467       }
468       a=(a&0xff00)|a2;
469       break;
470   }
471   // remember addr
472   Pico.video.addr=(u16)a;
473 }
474