4 // Original source by Cal2
5 // GCC/SDL port by Niels Wagenaar (Linux/WIN32) and Caz (BeOS)
6 // Extensive cleanups/fixes/rewrites by James L. Hammons
15 //#define OP_DEBUG_BMP
17 #define BLEND_Y(dst, src) op_blend_y[(((uint16)dst<<8)) | ((uint16)(src))]
18 #define BLEND_CR(dst, src) op_blend_cr[(((uint16)dst)<<8) | ((uint16)(src))]
20 #define OBJECT_TYPE_BITMAP 0 // 000
21 #define OBJECT_TYPE_SCALE 1 // 001
22 #define OBJECT_TYPE_GPU 2 // 010
23 #define OBJECT_TYPE_BRANCH 3 // 011
24 #define OBJECT_TYPE_STOP 4 // 100
26 #define CONDITION_EQUAL 0
27 #define CONDITION_LESS_THAN 1
28 #define CONDITION_GREATER_THAN 2
29 #define CONDITION_OP_FLAG_SET 3
30 #define CONDITION_SECOND_HALF_LINE 4
32 #define OPFLAG_RELEASE 8 // Bus release bit
33 #define OPFLAG_TRANS 4 // Transparency bit
34 #define OPFLAG_RMW 2 // Read-Modify-Write bit
35 #define OPFLAG_REFLECT 1 // Horizontal mirror bit
37 // Private function prototypes
39 void OPProcessFixedBitmap(uint64 p0, uint64 p1, bool render);
40 void OPProcessScaledBitmap(uint64 p0, uint64 p1, uint64 p2, bool render);
41 void DumpScaledObject(uint64 p0, uint64 p1, uint64 p2);
42 void DumpFixedObject(uint64 p0, uint64 p1);
43 uint64 op_load_phrase(uint32 offset);
45 // Local global variables
47 static uint8 * op_blend_y;
48 static uint8 * op_blend_cr;
49 // There may be a problem with this "RAM" overlapping (and thus being independent of)
50 // some of the regular TOM RAM...
51 static uint8 objectp_ram[0x40]; // This is based at $F00000
52 uint8 objectp_running;
53 //bool objectp_stop_reading_list;
55 static uint8 op_bitmap_bit_depth[8] = { 1, 2, 4, 8, 16, 24, 32, 0 };
56 //static uint32 op_bitmap_bit_size[8] =
57 // { (uint32)(0.125*65536), (uint32)(0.25*65536), (uint32)(0.5*65536), (uint32)(1*65536),
58 // (uint32)(2*65536), (uint32)(1*65536), (uint32)(1*65536), (uint32)(1*65536) };
59 static uint32 op_pointer;
61 int32 phraseWidthToPixels[8] = { 64, 32, 16, 8, 4, 2, 0, 0 };
65 // Object Processor initialization
69 // Blend tables (64K each)
70 memory_malloc_secure((void **)&op_blend_y, 0x10000, "Jaguar Object processor Y blend lookup table");
71 memory_malloc_secure((void **)&op_blend_cr, 0x10000, "Jaguar Object processor CR blend lookup table");
73 // Here we calculate the saturating blend of a signed 4-bit value and an
74 // existing Cyan/Red value as well as a signed 8-bit value and an existing intensity...
75 // Note: CRY is 4 bits Cyan, 4 bits Red, 16 bits intensitY
76 for(int i=0; i<256*256; i++)
78 int y = (i >> 8) & 0xFF;
79 int dy = (INT8)i; // Sign extend the Y index
80 int c1 = (i >> 8) & 0x0F;
81 int dc1 = (INT8)(i << 4) >> 4; // Sign extend the R index
82 int c2 = (i >> 12) & 0x0F;
83 int dc2 = (INT8)(i & 0xF0) >> 4; // Sign extend the C index
103 op_blend_cr[i] = (c2 << 4) | c1;
110 // Object Processor reset
114 memset(objectp_ram, 0x00, 0x40);
121 { "(BITMAP)", "(SCALED BITMAP)", "(GPU INT)", "(BRANCH)", "(STOP)", "???", "???", "???" };
123 { "\"==\"", "\"<\"", "\">\"", "(opflag set)", "(second half line)", "?", "?", "?" };
125 uint32 olp = op_get_list_pointer();
126 WriteLog("OP: OLP = %08X\n", olp);
127 WriteLog("OP: Phrase dump\n ----------\n");
128 for(uint32 i=0; i<0x100; i+=8)
130 uint32 hi = JaguarReadLong(olp + i, OP), lo = JaguarReadLong(olp + i + 4, OP);
131 WriteLog("\t%08X: %08X %08X %s", olp + i, hi, lo, opType[lo & 0x07]);
132 if ((lo & 0x07) == 3)
134 uint16 ypos = (lo >> 3) & 0x7FF;
135 uint8 cc = (lo >> 14) & 0x03;
136 uint32 link = ((hi << 11) | (lo >> 21)) & 0x3FFFF8;
137 WriteLog(" YPOS=%u, CC=%s, link=%08X", ypos, ccType[cc], link);
140 if ((lo & 0x07) == 0)
141 DumpFixedObject(op_load_phrase(olp+i), op_load_phrase(olp+i+8));
142 if ((lo & 0x07) == 1)
143 DumpScaledObject(op_load_phrase(olp+i), op_load_phrase(olp+i+8), op_load_phrase(olp+i+16));
149 // Object Processor memory access
150 // Memory range: F00010 - F00027
152 // F00010-F00017 R xxxxxxxx xxxxxxxx OB - current object code from the graphics processor
153 // F00020-F00023 W xxxxxxxx xxxxxxxx OLP - start of the object list
154 // F00026 W -------- -------x OBF - object processor flag
157 uint8 OPReadByte(uint32 offset, uint32 who/*=UNKNOWN*/)
160 return objectp_ram[offset];
163 uint16 OPReadWord(uint32 offset, uint32 who/*=UNKNOWN*/)
166 return GET16(objectp_ram, offset);
169 void OPWriteByte(uint32 offset, uint8 data, uint32 who/*=UNKNOWN*/)
172 objectp_ram[offset] = data;
175 void OPWriteWord(uint32 offset, uint16 data, uint32 who/*=UNKNOWN*/)
178 SET16(objectp_ram, offset, data);
180 /*if (offset == 0x20)
181 WriteLog("OP: Setting lo list pointer: %04X\n", data);
183 WriteLog("OP: Setting hi list pointer: %04X\n", data);//*/
186 uint32 op_get_list_pointer(void)
188 // Note: This register is LO / HI WORD, hence the funky look of this...
189 // return (objectp_ram[0x22] << 24) | (objectp_ram[0x23] << 16) | (objectp_ram[0x20] << 8) | objectp_ram[0x21];
190 return GET16(objectp_ram, 0x20) | (GET16(objectp_ram, 0x22) << 16);
193 // This is WRONG, since the OBF is only 16 bits wide!!! [FIXED]
195 uint32 op_get_status_register(void)
197 // return (objectp_ram[0x26] << 24) | (objectp_ram[0x27] << 16) | (objectp_ram[0x28] << 8) | objectp_ram[0x29];
198 // return GET32(objectp_ram, 0x26);
199 return GET16(objectp_ram, 0x26);
202 // This is WRONG, since the OBF is only 16 bits wide!!! [FIXED]
204 void op_set_status_register(uint32 data)
206 /* objectp_ram[0x26] = (data & 0xFF000000) >> 24;
207 objectp_ram[0x27] = (data & 0x00FF0000) >> 16;
208 objectp_ram[0x28] = (data & 0x0000FF00) >> 8;
209 objectp_ram[0x29] |= (data & 0xFE);*/
210 objectp_ram[0x26] = (data & 0x0000FF00) >> 8;
211 objectp_ram[0x27] |= (data & 0xFE);
214 void op_set_current_object(uint64 object)
216 //Not sure this is right... Wouldn't it just be stored 64 bit BE?
217 // Stored as least significant 32 bits first, ms32 last in big endian
218 /* objectp_ram[0x13] = object & 0xFF; object >>= 8;
219 objectp_ram[0x12] = object & 0xFF; object >>= 8;
220 objectp_ram[0x11] = object & 0xFF; object >>= 8;
221 objectp_ram[0x10] = object & 0xFF; object >>= 8;
223 objectp_ram[0x17] = object & 0xFF; object >>= 8;
224 objectp_ram[0x16] = object & 0xFF; object >>= 8;
225 objectp_ram[0x15] = object & 0xFF; object >>= 8;
226 objectp_ram[0x14] = object & 0xFF;*/
227 // Let's try regular good old big endian...
228 objectp_ram[0x17] = object & 0xFF; object >>= 8;
229 objectp_ram[0x16] = object & 0xFF; object >>= 8;
230 objectp_ram[0x15] = object & 0xFF; object >>= 8;
231 objectp_ram[0x14] = object & 0xFF; object >>= 8;
233 objectp_ram[0x13] = object & 0xFF; object >>= 8;
234 objectp_ram[0x12] = object & 0xFF; object >>= 8;
235 objectp_ram[0x11] = object & 0xFF; object >>= 8;
236 objectp_ram[0x10] = object & 0xFF;
239 uint64 op_load_phrase(uint32 offset)
241 offset &= ~0x07; // 8 byte alignment
242 return ((uint64)JaguarReadLong(offset, OP) << 32) | (uint64)JaguarReadLong(offset+4, OP);
245 void OPStorePhrase(uint32 offset, uint64 p)
247 offset &= ~0x07; // 8 byte alignment
248 JaguarWriteLong(offset, p >> 32, OP);
249 JaguarWriteLong(offset + 4, p & 0xFFFFFFFF, OP);
253 // Debugging routines
255 void DumpScaledObject(uint64 p0, uint64 p1, uint64 p2)
257 WriteLog(" (SCALED BITMAP)");
258 WriteLog(" %08X --> phrase %08X %08X\n", op_pointer, (uint32)(p1>>32), (uint32)(p1&0xFFFFFFFF));
259 WriteLog(" %08X --> phrase %08X %08X ", op_pointer+8, (uint32)(p2>>32), (uint32)(p2&0xFFFFFFFF));
260 uint8 bitdepth = (p1 >> 12) & 0x07;
261 int16 ypos = ((p0 >> 3) & 0x3FF); // ??? What if not interlaced (/2)?
262 int32 xpos = p1 & 0xFFF;
263 xpos = (xpos & 0x800 ? xpos | 0xFFFFF000 : xpos);
264 uint32 iwidth = ((p1 >> 28) & 0x3FF);
265 uint32 dwidth = ((p1 >> 18) & 0x3FF); // Unsigned!
266 uint16 height = ((p0 >> 14) & 0x3FF);
267 uint32 link = ((p0 >> 24) & 0x7FFFF) << 3;
268 uint32 ptr = ((p0 >> 43) & 0x1FFFFF) << 3;
269 uint32 firstPix = (p1 >> 49) & 0x3F;
270 uint8 flags = (p1 >> 45) & 0x0F;
271 uint8 idx = (p1 >> 38) & 0x7F;
272 uint32 pitch = (p1 >> 15) & 0x07;
273 WriteLog("\n [%u (%u) x %u @ (%i, %u) (%u bpp), l: %08X, p: %08X fp: %02X, fl:%s%s%s%s, idx:%02X, pt:%02X]\n",
274 iwidth, dwidth, height, xpos, ypos, op_bitmap_bit_depth[bitdepth], link, ptr, firstPix, (flags&OPFLAG_REFLECT ? "REFLECT " : ""), (flags&OPFLAG_RMW ? "RMW " : ""), (flags&OPFLAG_TRANS ? "TRANS " : ""), (flags&OPFLAG_RELEASE ? "RELEASE" : ""), idx, pitch);
275 uint32 hscale = p2 & 0xFF;
276 uint32 vscale = (p2 >> 8) & 0xFF;
277 uint32 remainder = (p2 >> 16) & 0xFF;
278 WriteLog(" [hsc: %02X, vsc: %02X, rem: %02X]\n", hscale, vscale, remainder);
281 void DumpFixedObject(uint64 p0, uint64 p1)
283 WriteLog(" (BITMAP)");
284 WriteLog(" %08X --> phrase %08X %08X\n", op_pointer, (uint32)(p1>>32), (uint32)(p1&0xFFFFFFFF));
285 uint8 bitdepth = (p1 >> 12) & 0x07;
286 int16 ypos = ((p0 >> 3) & 0x3FF); // ??? What if not interlaced (/2)?
287 int32 xpos = p1 & 0xFFF;
288 xpos = (xpos & 0x800 ? xpos | 0xFFFFF000 : xpos);
289 uint32 iwidth = ((p1 >> 28) & 0x3FF);
290 uint32 dwidth = ((p1 >> 18) & 0x3FF); // Unsigned!
291 uint16 height = ((p0 >> 14) & 0x3FF);
292 uint32 link = ((p0 >> 24) & 0x7FFFF) << 3;
293 uint32 ptr = ((p0 >> 43) & 0x1FFFFF) << 3;
294 uint32 firstPix = (p1 >> 49) & 0x3F;
295 uint8 flags = (p1 >> 45) & 0x0F;
296 uint8 idx = (p1 >> 38) & 0x7F;
297 uint32 pitch = (p1 >> 15) & 0x07;
298 WriteLog(" [%u (%u) x %u @ (%i, %u) (%u bpp), l: %08X, p: %08X fp: %02X, fl:%s%s%s%s, idx:%02X, pt:%02X]\n",
299 iwidth, dwidth, height, xpos, ypos, op_bitmap_bit_depth[bitdepth], link, ptr, firstPix, (flags&OPFLAG_REFLECT ? "REFLECT " : ""), (flags&OPFLAG_RMW ? "RMW " : ""), (flags&OPFLAG_TRANS ? "TRANS " : ""), (flags&OPFLAG_RELEASE ? "RELEASE" : ""), idx, pitch);
303 // Object Processor main routine
305 //Need to fix this so that when an GPU object IRQ happens, we can pick up OP processing
306 //where we left off. !!! FIX !!!
307 void OPProcessList(int scanline, bool render)
309 extern int op_start_log;
310 // char * condition_to_str[8] =
311 // { "==", "<", ">", "(opflag set)", "(second half line)", "?", "?", "?" };
313 op_pointer = op_get_list_pointer();
315 // objectp_stop_reading_list = false;
317 // *** BEGIN OP PROCESSOR TESTING ONLY ***
318 extern bool interactiveMode;
320 extern int objectPtr;
322 int bitmapCounter = 0;
323 // *** END OP PROCESSOR TESTING ONLY ***
325 // if (op_pointer) WriteLog(" new op list at 0x%.8x scanline %i\n",op_pointer,scanline);
328 // *** BEGIN OP PROCESSOR TESTING ONLY ***
329 if (interactiveMode && bitmapCounter == objectPtr)
333 // *** END OP PROCESSOR TESTING ONLY ***
334 // if (objectp_stop_reading_list)
337 uint64 p0 = op_load_phrase(op_pointer);
339 if (scanline == tom_get_vdb() && op_start_log)
340 //if (scanline == 215 && op_start_log)
342 WriteLog("%08X --> phrase %08X %08X", op_pointer - 8, (int)(p0>>32), (int)(p0&0xFFFFFFFF));
343 if ((p0 & 0x07) == OBJECT_TYPE_BITMAP)
345 WriteLog(" (BITMAP) ");
346 uint64 p1 = op_load_phrase(op_pointer);
347 WriteLog("\n%08X --> phrase %08X %08X ", op_pointer, (int)(p1>>32), (int)(p1&0xFFFFFFFF));
348 uint8 bitdepth = (p1 >> 12) & 0x07;
349 int16 ypos = ((p0 >> 3) & 0x3FF); // ??? What if not interlaced (/2)?
350 int32 xpos = p1 & 0xFFF;
351 xpos = (xpos & 0x800 ? xpos | 0xFFFFF000 : xpos);
352 uint32 iwidth = ((p1 >> 28) & 0x3FF);
353 uint32 dwidth = ((p1 >> 18) & 0x3FF); // Unsigned!
354 uint16 height = ((p0 >> 14) & 0x3FF);
355 uint32 link = ((p0 >> 24) & 0x7FFFF) << 3;
356 uint32 ptr = ((p0 >> 43) & 0x1FFFFF) << 3;
357 uint32 firstPix = (p1 >> 49) & 0x3F;
358 uint8 flags = (p1 >> 45) & 0x0F;
359 uint8 idx = (p1 >> 38) & 0x7F;
360 uint32 pitch = (p1 >> 15) & 0x07;
361 WriteLog("\n [%u (%u) x %u @ (%i, %u) (%u bpp), l: %08X, p: %08X fp: %02X, fl:%s%s%s%s, idx:%02X, pt:%02X]\n",
362 iwidth, dwidth, height, xpos, ypos, op_bitmap_bit_depth[bitdepth], link, ptr, firstPix, (flags&OPFLAG_REFLECT ? "REFLECT " : ""), (flags&OPFLAG_RMW ? "RMW " : ""), (flags&OPFLAG_TRANS ? "TRANS " : ""), (flags&OPFLAG_RELEASE ? "RELEASE" : ""), idx, pitch);
364 if ((p0 & 0x07) == OBJECT_TYPE_SCALE)
366 WriteLog(" (SCALED BITMAP)");
367 uint64 p1 = op_load_phrase(op_pointer), p2 = op_load_phrase(op_pointer+8);
368 WriteLog("\n%08X --> phrase %08X %08X ", op_pointer, (int)(p1>>32), (int)(p1&0xFFFFFFFF));
369 WriteLog("\n%08X --> phrase %08X %08X ", op_pointer+8, (int)(p2>>32), (int)(p2&0xFFFFFFFF));
370 uint8 bitdepth = (p1 >> 12) & 0x07;
371 int16 ypos = ((p0 >> 3) & 0x3FF); // ??? What if not interlaced (/2)?
372 int32 xpos = p1 & 0xFFF;
373 xpos = (xpos & 0x800 ? xpos | 0xFFFFF000 : xpos);
374 uint32 iwidth = ((p1 >> 28) & 0x3FF);
375 uint32 dwidth = ((p1 >> 18) & 0x3FF); // Unsigned!
376 uint16 height = ((p0 >> 14) & 0x3FF);
377 uint32 link = ((p0 >> 24) & 0x7FFFF) << 3;
378 uint32 ptr = ((p0 >> 43) & 0x1FFFFF) << 3;
379 uint32 firstPix = (p1 >> 49) & 0x3F;
380 uint8 flags = (p1 >> 45) & 0x0F;
381 uint8 idx = (p1 >> 38) & 0x7F;
382 uint32 pitch = (p1 >> 15) & 0x07;
383 WriteLog("\n [%u (%u) x %u @ (%i, %u) (%u bpp), l: %08X, p: %08X fp: %02X, fl:%s%s%s%s, idx:%02X, pt:%02X]\n",
384 iwidth, dwidth, height, xpos, ypos, op_bitmap_bit_depth[bitdepth], link, ptr, firstPix, (flags&OPFLAG_REFLECT ? "REFLECT " : ""), (flags&OPFLAG_RMW ? "RMW " : ""), (flags&OPFLAG_TRANS ? "TRANS " : ""), (flags&OPFLAG_RELEASE ? "RELEASE" : ""), idx, pitch);
385 uint32 hscale = p2 & 0xFF;
386 uint32 vscale = (p2 >> 8) & 0xFF;
387 uint32 remainder = (p2 >> 16) & 0xFF;
388 WriteLog(" [hsc: %02X, vsc: %02X, rem: %02X]\n", hscale, vscale, remainder);
390 if ((p0 & 0x07) == OBJECT_TYPE_GPU)
391 WriteLog(" (GPU)\n");
392 if ((p0 & 0x07) == OBJECT_TYPE_BRANCH)
394 WriteLog(" (BRANCH)\n");
395 uint8 * jaguar_mainRam = GetRamPtr();
396 WriteLog("[RAM] --> ");
397 for(int k=0; k<8; k++)
398 WriteLog("%02X ", jaguar_mainRam[op_pointer-8 + k]);
401 if ((p0 & 0x07) == OBJECT_TYPE_STOP)
402 WriteLog(" --> List end\n");
405 // WriteLog("%08X type %i\n", op_pointer, (uint8)p0 & 0x07);
406 switch ((uint8)p0 & 0x07)
408 case OBJECT_TYPE_BITMAP:
410 uint16 ypos = (p0 >> 3) & 0x3FF;
411 // This is only theory implied by Rayman...!
412 // It seems that if the YPOS is zero, then bump the YPOS value so that it coincides with
413 // the VDB value. With interlacing, this would be slightly more tricky.
414 // There's probably another bit somewhere that enables this mode--but so far, doesn't seem
415 // to affect any other game in a negative way (that I've seen).
416 // Either that, or it's an undocumented bug...
418 //No, the reason this was needed is that the OP code before was wrong. Any value
419 //less than VDB will get written to the top line of the display!
421 // ypos = TOMReadWord(0xF00046, OP) / 2; // Get the VDB value
422 uint32 height = (p0 & 0xFFC000) >> 14;
423 uint32 oldOPP = op_pointer - 8;
424 // *** BEGIN OP PROCESSOR TESTING ONLY ***
425 if (inhibit && op_start_log)
426 WriteLog("!!! ^^^ This object is INHIBITED! ^^^ !!!\n");
428 if (!inhibit) // For OP testing only!
429 // *** END OP PROCESSOR TESTING ONLY ***
430 if (scanline >= ypos && height > 0)
432 uint64 p1 = op_load_phrase(op_pointer);
434 //WriteLog("OP: Writing scanline %d with ypos == %d...\n", scanline, ypos);
435 //WriteLog("--> Writing %u BPP bitmap...\n", op_bitmap_bit_depth[(p1 >> 12) & 0x07]);
436 // OPProcessFixedBitmap(scanline, p0, p1, render);
437 OPProcessFixedBitmap(p0, p1, render);
441 //???Does this really happen??? Doesn't seem to work if you do this...!
442 //Probably not. Must be a bug in the documentation...!
443 // uint32 link = (p0 & 0x7FFFF000000) >> 21;
444 // SET16(objectp_ram, 0x20, link & 0xFFFF); // OLP
445 // SET16(objectp_ram, 0x22, link >> 16);
446 /* uint32 height = (p0 & 0xFFC000) >> 14;
449 // NOTE: Would subtract 2 if in interlaced mode...!
450 // uint64 height = ((p0 & 0xFFC000) - 0x4000) & 0xFFC000;
454 uint64 data = (p0 & 0xFFFFF80000000000LL) >> 40;
455 uint64 dwidth = (p1 & 0xFFC0000) >> 15;
458 p0 &= ~0xFFFFF80000FFC000LL; // Mask out old data...
459 p0 |= (uint64)height << 14;
461 OPStorePhrase(oldOPP, p0);
463 op_pointer = (p0 & 0x000007FFFF000000LL) >> 21;
466 case OBJECT_TYPE_SCALE:
468 uint16 ypos = (p0 >> 3) & 0x3FF;
469 uint32 height = (p0 & 0xFFC000) >> 14;
470 uint32 oldOPP = op_pointer - 8;
471 // *** BEGIN OP PROCESSOR TESTING ONLY ***
472 if (inhibit && op_start_log)
474 WriteLog("!!! ^^^ This object is INHIBITED! ^^^ !!! (scanline=%u, ypos=%u, height=%u)\n", scanline, ypos, height);
475 DumpScaledObject(p0, op_load_phrase(op_pointer), op_load_phrase(op_pointer+8));
478 if (!inhibit) // For OP testing only!
479 // *** END OP PROCESSOR TESTING ONLY ***
480 if (scanline >= ypos && height > 0)
482 uint64 p1 = op_load_phrase(op_pointer);
484 uint64 p2 = op_load_phrase(op_pointer);
486 //WriteLog("OP: %08X (%d) %08X%08X %08X%08X %08X%08X\n", oldOPP, scanline, (uint32)(p0>>32), (uint32)(p0&0xFFFFFFFF), (uint32)(p1>>32), (uint32)(p1&0xFFFFFFFF), (uint32)(p2>>32), (uint32)(p2&0xFFFFFFFF));
487 OPProcessScaledBitmap(p0, p1, p2, render);
491 uint8 remainder = p2 >> 16, vscale = p2 >> 8;
492 //Actually, we should skip this object if it has a vscale of zero.
493 //Or do we? Not sure... Atari Karts has a few lines that look like:
495 //000E8268 --> phrase 00010000 7000B00D
496 // [7 (0) x 1 @ (13, 0) (8 bpp), l: 000E82A0, p: 000E0FC0 fp: 00, fl:RELEASE, idx:00, pt:01]
497 // [hsc: 9A, vsc: 00, rem: 00]
498 // Could it be the vscale is overridden if the DWIDTH is zero? Hmm...
501 vscale = 0x20; // OP bug??? Nope, it isn't...! Or is it?
503 /*extern int start_logging;
505 WriteLog("--> Returned from scaled bitmap processing (rem=%02X, vscale=%02X)...\n", remainder, vscale);*/
507 //--> Returned from scaled bitmap processing (rem=20, vscale=80)...
508 //There are other problems here, it looks like...
510 OP: Scaled bitmap 4x? 4bpp at 38,? hscale=7C fpix=0 data=00075E28 pitch 1 hflipped=no dwidth=? (linked to 00071118) Transluency=no
511 --> Returned from scaled bitmap processing (rem=50, vscale=7C)...
512 OP: Scaled bitmap 4x? 4bpp at 38,? hscale=7C fpix=0 data=00075E28 pitch 1 hflipped=no dwidth=? (linked to 00071118) Transluency=no
513 --> Returned from scaled bitmap processing (rem=30, vscale=7C)...
514 OP: Scaled bitmap 4x? 4bpp at 38,? hscale=7C fpix=0 data=00075E28 pitch 1 hflipped=no dwidth=? (linked to 00071118) Transluency=no
515 --> Returned from scaled bitmap processing (rem=10, vscale=7C)...
516 OP: Scaled bitmap 4x? 4bpp at 36,? hscale=7E fpix=0 data=000756A8 pitch 1 hflipped=no dwidth=? (linked to 00073058) Transluency=no
517 --> Returned from scaled bitmap processing (rem=00, vscale=7E)...
518 OP: Scaled bitmap 4x? 4bpp at 34,? hscale=80 fpix=0 data=000756C8 pitch 1 hflipped=no dwidth=? (linked to 00073078) Transluency=no
519 --> Returned from scaled bitmap processing (rem=00, vscale=80)...
520 OP: Scaled bitmap 4x? 4bpp at 36,? hscale=7E fpix=0 data=000756C8 pitch 1 hflipped=no dwidth=? (linked to 00073058) Transluency=no
521 --> Returned from scaled bitmap processing (rem=5E, vscale=7E)...
522 OP: Scaled bitmap 4x? 4bpp at 34,? hscale=80 fpix=0 data=000756E8 pitch 1 hflipped=no dwidth=? (linked to 00073078) Transluency=no
523 --> Returned from scaled bitmap processing (rem=60, vscale=80)...
524 OP: Scaled bitmap 4x? 4bpp at 36,? hscale=7E fpix=0 data=000756C8 pitch 1 hflipped=no dwidth=? (linked to 00073058) Transluency=no
525 --> Returned from scaled bitmap processing (rem=3E, vscale=7E)...
526 OP: Scaled bitmap 4x? 4bpp at 34,? hscale=80 fpix=0 data=000756E8 pitch 1 hflipped=no dwidth=? (linked to 00073078) Transluency=no
527 --> Returned from scaled bitmap processing (rem=40, vscale=80)...
528 OP: Scaled bitmap 4x? 4bpp at 36,? hscale=7E fpix=0 data=000756C8 pitch 1 hflipped=no dwidth=? (linked to 00073058) Transluency=no
529 --> Returned from scaled bitmap processing (rem=1E, vscale=7E)...
530 OP: Scaled bitmap 4x? 4bpp at 34,? hscale=80 fpix=0 data=000756E8 pitch 1 hflipped=no dwidth=? (linked to 00073078) Transluency=no
531 --> Returned from scaled bitmap processing (rem=20, vscale=80)...
533 //Here's another problem:
534 // [hsc: 20, vsc: 20, rem: 00]
535 // Since we're not checking for $E0 (but that's what we get from the above), we end
536 // up repeating this scanline unnecessarily... !!! FIX !!! [DONE, but... still not quite
537 // right. Either that, or the Accolade team that wrote Bubsy screwed up royal.]
538 //Also note: $E0 = 7.0 which IS a legal vscale value...
540 // if (remainder & 0x80) // I.e., it's negative
541 // if ((remainder & 0x80) || remainder == 0) // I.e., it's <= 0
542 // if ((remainder - 1) >= 0xE0) // I.e., it's <= 0
543 // if ((remainder >= 0xE1) || remainder == 0)// I.e., it's <= 0
544 // if ((remainder >= 0xE1 && remainder <= 0xFF) || remainder == 0)// I.e., it's <= 0
545 if (remainder <= 0x20) // I.e., it's <= 0
547 uint64 data = (p0 & 0xFFFFF80000000000LL) >> 40;
548 uint64 dwidth = (p1 & 0xFFC0000) >> 15;
550 // while (remainder & 0x80)
551 // while ((remainder & 0x80) || remainder == 0)
552 // while ((remainder - 1) >= 0xE0)
553 // while ((remainder >= 0xE1) || remainder == 0)
554 // while ((remainder >= 0xE1 && remainder <= 0xFF) || remainder == 0)
555 while (remainder <= 0x20)
565 p0 &= ~0xFFFFF80000FFC000LL; // Mask out old data...
566 p0 |= (uint64)height << 14;
568 OPStorePhrase(oldOPP, p0);
571 remainder -= 0x20; // 1.0f in [3.5] fixed point format
574 WriteLog("--> Finished writebacks...\n");*/
576 //WriteLog(" [%08X%08X -> ", (uint32)(p2>>32), (uint32)(p2&0xFFFFFFFF));
577 p2 &= ~0x0000000000FF0000LL;
578 p2 |= (uint64)remainder << 16;
579 //WriteLog("%08X%08X]\n", (uint32)(p2>>32), (uint32)(p2&0xFFFFFFFF));
580 OPStorePhrase(oldOPP+16, p2);
581 //remainder = (uint8)(p2 >> 16), vscale = (uint8)(p2 >> 8);
582 //WriteLog(" [after]: rem=%02X, vscale=%02X\n", remainder, vscale);
584 op_pointer = (p0 & 0x000007FFFF000000LL) >> 21;
587 case OBJECT_TYPE_GPU:
589 //WriteLog("OP: Asserting GPU IRQ #3...\n");
590 op_set_current_object(p0);
591 GPUSetIRQLine(3, ASSERT_LINE);
592 //Also, OP processing is suspended from this point until OBF (F00026) is written to...
595 //OPSuspendedByGPU = true;
596 //Dunno if the OP keeps processing from where it was interrupted, or if it just continues
597 //on the next scanline...
598 // --> It continues from where it was interrupted! !!! FIX !!!
601 case OBJECT_TYPE_BRANCH:
603 uint16 ypos = (p0 >> 3) & 0x7FF;
604 uint8 cc = (p0 >> 14) & 0x03;
605 uint32 link = (p0 >> 21) & 0x3FFFF8;
607 // if ((ypos!=507)&&(ypos!=25))
608 // WriteLog("\t%i%s%i link=0x%.8x\n",scanline,condition_to_str[cc],ypos>>1,link);
611 case CONDITION_EQUAL:
612 if (TOMReadWord(0xF00006, OP) == ypos || ypos == 0x7FF)
615 case CONDITION_LESS_THAN:
616 if (TOMReadWord(0xF00006, OP) < ypos)
619 case CONDITION_GREATER_THAN:
620 if (TOMReadWord(0xF00006, OP) > ypos)
623 case CONDITION_OP_FLAG_SET:
624 if (op_get_status_register() & 0x01)
627 case CONDITION_SECOND_HALF_LINE:
628 // This basically means branch if bit 10 of HC is set
629 WriteLog("OP: Unexpected CONDITION_SECOND_HALF_LINE in BRANCH object\nOP: shuting down\n");
634 WriteLog("OP: Unimplemented branch condition %i\n", cc);
638 case OBJECT_TYPE_STOP:
642 //WriteLog("OP: --> STOP\n");
643 // op_set_status_register(((p0>>3) & 0xFFFFFFFF));
644 //This seems more likely...
645 op_set_current_object(p0);
649 tom_set_pending_object_int();
650 if (tom_irq_enabled(IRQ_OPFLAG))// && jaguar_interrupt_handler_is_valid(64))
651 m68k_set_irq(7); // Cause an NMI to occur...
658 WriteLog("op: unknown object type %i\n", ((uint8)p0 & 0x07));
665 // Store fixed size bitmap in line buffer
667 void OPProcessFixedBitmap(uint64 p0, uint64 p1, bool render)
669 // Need to make sure that when writing that it stays within the line buffer...
670 // LBUF ($F01800 - $F01D9E) 360 x 32-bit RAM
671 uint8 depth = (p1 >> 12) & 0x07; // Color depth of image
672 int32 xpos = ((int16)((p1 << 4) & 0xFFFF)) >> 4;// Image xpos in LBUF
673 uint32 iwidth = (p1 >> 28) & 0x3FF; // Image width in *phrases*
674 uint32 data = (p0 >> 40) & 0xFFFFF8; // Pixel data address
675 //#ifdef OP_DEBUG_BMP
676 uint32 firstPix = (p1 >> 49) & 0x3F;
677 // "The LSB is significant only for scaled objects..." -JTRM
678 // "In 1 BPP mode, all five bits are significant. In 2 BPP mode, the top four are significant..."
681 // We can ignore the RELEASE (high order) bit for now--probably forever...!
682 // uint8 flags = (p1 >> 45) & 0x0F; // REFLECT, RMW, TRANS, RELEASE
683 //Optimize: break these out to their own BOOL values
684 uint8 flags = (p1 >> 45) & 0x07; // REFLECT (0), RMW (1), TRANS (2)
685 bool flagREFLECT = (flags & OPFLAG_REFLECT ? true : false),
686 flagRMW = (flags & OPFLAG_RMW ? true : false),
687 flagTRANS = (flags & OPFLAG_TRANS ? true : false);
688 // "For images with 1 to 4 bits/pixel the top 7 to 4 bits of the index
689 // provide the most significant bits of the palette address."
690 uint8 index = (p1 >> 37) & 0xFE; // CLUT index offset (upper pix, 1-4 bpp)
691 uint32 pitch = (p1 >> 15) & 0x07; // Phrase pitch
692 pitch <<= 3; // Optimization: Multiply pitch by 8
694 // int16 scanlineWidth = tom_getVideoModeWidth();
695 uint8 * tom_ram_8 = tom_get_ram_pointer();
696 uint8 * paletteRAM = &tom_ram_8[0x400];
697 // This is OK as long as it's used correctly: For 16-bit RAM to RAM direct copies--NOT
698 // for use when using endian-corrected data (i.e., any of the *_word_read functions!)
699 uint16 * paletteRAM16 = (uint16 *)paletteRAM;
701 // WriteLog("bitmap %ix? %ibpp at %i,? firstpix=? data=0x%.8x pitch %i hflipped=%s dwidth=? (linked to ?) RMW=%s Tranparent=%s\n",
702 // iwidth, op_bitmap_bit_depth[bitdepth], xpos, ptr, pitch, (flags&OPFLAG_REFLECT ? "yes" : "no"), (flags&OPFLAG_RMW ? "yes" : "no"), (flags&OPFLAG_TRANS ? "yes" : "no"));
704 // Is it OK to have a 0 for the data width??? (i.e., undocumented?)
705 // Seems to be... Seems that dwidth *can* be zero (i.e., reuse same line) as well.
706 // Pitch == 0 is OK too...
707 // if (!render || op_pointer == 0 || ptr == 0 || pitch == 0)
708 //I'm not convinced that we need to concern ourselves with data & op_pointer here either!
709 if (!render || iwidth == 0)
712 //#define OP_DEBUG_BMP
713 //#ifdef OP_DEBUG_BMP
714 // WriteLog("bitmap %ix%i %ibpp at %i,%i firstpix=%i data=0x%.8x pitch %i hflipped=%s dwidth=%i (linked to 0x%.8x) Transluency=%s\n",
715 // iwidth, height, op_bitmap_bit_depth[bitdepth], xpos, ypos, firstPix, ptr, pitch, (flags&OPFLAG_REFLECT ? "yes" : "no"), dwidth, op_pointer, (flags&OPFLAG_RMW ? "yes" : "no"));
718 // int32 leftMargin = xpos, rightMargin = (xpos + (phraseWidthToPixels[depth] * iwidth)) - 1;
719 int32 startPos = xpos, endPos = xpos +
720 (!flagREFLECT ? (phraseWidthToPixels[depth] * iwidth) - 1
721 : -((phraseWidthToPixels[depth] * iwidth) + 1));
722 uint32 clippedWidth = 0, phraseClippedWidth = 0, dataClippedWidth = 0;//, phrasePixel = 0;
723 bool in24BPPMode = (((GET16(tom_ram_8, 0x0028) >> 1) & 0x03) == 1 ? true : false); // VMODE
724 // Not sure if this is Jaguar Two only location or what...
725 // From the docs, it is... If we want to limit here we should think of something else.
726 // int32 limit = GET16(tom_ram_8, 0x0008); // LIMIT
728 int32 lbufWidth = (!in24BPPMode ? limit - 1 : (limit / 2) - 1); // Zero based limit...
730 // If the image is completely to the left or right of the line buffer, then bail.
731 //If in REFLECT mode, then these values are swapped! !!! FIX !!! [DONE]
732 //There are four possibilities:
733 // 1. image sits on left edge and no REFLECT; starts out of bounds but ends in bounds.
734 // 2. image sits on left edge and REFLECT; starts in bounds but ends out of bounds.
735 // 3. image sits on right edge and REFLECT; starts out of bounds but ends in bounds.
736 // 4. image sits on right edge and no REFLECT; starts in bounds but ends out of bounds.
737 //Numbers 2 & 4 can be caught by checking the LBUF clip while in the inner loop,
738 // numbers 1 & 3 are of concern.
739 // This *indirectly* handles only cases 2 & 4! And is WRONG is REFLECT is set...!
740 // if (rightMargin < 0 || leftMargin > lbufWidth)
742 // It might be easier to swap these (if REFLECTed) and just use XPOS down below...
743 // That way, you could simply set XPOS to leftMargin if !REFLECT and to rightMargin otherwise.
744 // Still have to be careful with the DATA and IWIDTH values though...
746 // if ((!flagREFLECT && (rightMargin < 0 || leftMargin > lbufWidth))
747 // || (flagREFLECT && (leftMargin < 0 || rightMargin > lbufWidth)))
749 if ((!flagREFLECT && (endPos < 0 || startPos > lbufWidth))
750 || (flagREFLECT && (startPos < 0 || endPos > lbufWidth)))
753 // Otherwise, find the clip limits and clip the phrase as well...
754 // NOTE: I'm fudging here by letting the actual blit overstep the bounds of the
755 // line buffer, but it shouldn't matter since there are two unused line
756 // buffers below and nothing above and I'll at most write 8 bytes outside
757 // the line buffer... I could use a fractional clip begin/end value, but
758 // this makes the blit a *lot* more hairy. I might fix this in the future
759 // if it becomes necessary. (JLH)
760 // Probably wouldn't be *that* hairy. Just use a delta that tells the inner loop
761 // which pixel in the phrase is being written, and quit when either end of phrases
762 // is reached or line buffer extents are surpassed.
764 //This stuff is probably wrong as well... !!! FIX !!!
765 //The strange thing is that it seems to work, but that's no guarantee that it's bulletproof!
766 //Yup. Seems that JagMania doesn't work correctly with this...
767 //Dunno if this is the problem, but Atari Karts is showing *some* of the road now...
772 clippedWidth = 0 - leftMargin,
773 phraseClippedWidth = clippedWidth / phraseWidthToPixels[depth],
774 leftMargin = 0 - (clippedWidth % phraseWidthToPixels[depth]);
777 if (rightMargin > lbufWidth)
778 clippedWidth = rightMargin - lbufWidth,
779 phraseClippedWidth = clippedWidth / phraseWidthToPixels[depth];//,
780 // rightMargin = lbufWidth + (clippedWidth % phraseWidthToPixels[depth]);
781 // rightMargin = lbufWidth;
784 WriteLog("OP: We're about to encounter a divide by zero error!\n");
785 // NOTE: We're just using endPos to figure out how much, if any, to clip by.
786 // ALSO: There may be another case where we start out of bounds and end out of bounds...!
788 if (startPos < 0) // Case #1: Begin out, end in, L to R
789 clippedWidth = 0 - startPos,
790 dataClippedWidth = phraseClippedWidth = clippedWidth / phraseWidthToPixels[depth],
791 startPos = 0 - (clippedWidth % phraseWidthToPixels[depth]);
793 if (endPos < 0) // Case #2: Begin in, end out, R to L
794 clippedWidth = 0 - endPos,
795 phraseClippedWidth = clippedWidth / phraseWidthToPixels[depth];
797 if (endPos > lbufWidth) // Case #3: Begin in, end out, L to R
798 clippedWidth = endPos - lbufWidth,
799 phraseClippedWidth = clippedWidth / phraseWidthToPixels[depth];
801 if (startPos > lbufWidth) // Case #4: Begin out, end in, R to L
802 clippedWidth = startPos - lbufWidth,
803 dataClippedWidth = phraseClippedWidth = clippedWidth / phraseWidthToPixels[depth],
804 startPos = lbufWidth + (clippedWidth % phraseWidthToPixels[depth]);
806 // If the image is sitting on the line buffer left or right edge, we need to compensate
807 // by decreasing the image phrase width accordingly.
808 iwidth -= phraseClippedWidth;
810 // Also, if we're clipping the phrase we need to make sure we're in the correct part of
812 // data += phraseClippedWidth * (pitch << 3);
813 data += dataClippedWidth * pitch;
815 // NOTE: When the bitmap is in REFLECT mode, the XPOS marks the *right* side of the
816 // bitmap! This makes clipping & etc. MUCH, much easier...!
817 // uint32 lbufAddress = 0x1800 + (!in24BPPMode ? leftMargin * 2 : leftMargin * 4);
818 //Why does this work right when multiplying startPos by 2 (instead of 4) for 24 BPP mode?
819 //Is this a bug in the OP?
820 uint32 lbufAddress = 0x1800 + (!in24BPPMode ? startPos * 2 : startPos * 2);
821 uint8 * currentLineBuffer = &tom_ram_8[lbufAddress];
825 // Hmm. We check above for 24 BPP mode, but don't do anything about it below...
826 // If we *were* in 24 BPP mode, how would you convert CRY to RGB24? Seems to me
827 // that if you're in CRY mode then you wouldn't be able to use 24 BPP bitmaps
829 // This seems to be the case (at least according to the Midsummer docs)...!
831 if (depth == 0) // 1 BPP
833 // The LSB of flags is OPFLAG_REFLECT, so sign extend it and or 2 into it.
834 int32 lbufDelta = ((int8)((flags << 7) & 0xFF) >> 5) | 0x02;
836 // Fetch 1st phrase...
837 uint64 pixels = ((uint64)JaguarReadLong(data, OP) << 32) | JaguarReadLong(data + 4, OP);
838 //Note that firstPix should only be honored *if* we start with the 1st phrase of the bitmap
839 //i.e., we didn't clip on the margin... !!! FIX !!!
840 pixels <<= firstPix; // Skip first N pixels (N=firstPix)...
841 int i = firstPix; // Start counter at right spot...
847 uint8 bit = pixels >> 63;
848 if (flagTRANS && bit == 0)
853 //Optimize: Set palleteRAM16 to beginning of palette RAM + index*2 and use only [bit] as index...
854 //Won't optimize RMW case though...
855 // This is the *only* correct use of endian-dependent code
856 // (i.e., mem-to-mem direct copying)!
857 *(uint16 *)currentLineBuffer = paletteRAM16[index | bit];
860 BLEND_CR(*currentLineBuffer, paletteRAM[(index | bit) << 1]),
861 *(currentLineBuffer + 1) =
862 BLEND_Y(*(currentLineBuffer + 1), paletteRAM[((index | bit) << 1) + 1]);
865 currentLineBuffer += lbufDelta;
869 // Fetch next phrase...
871 pixels = ((uint64)JaguarReadLong(data, OP) << 32) | JaguarReadLong(data + 4, OP);
874 else if (depth == 1) // 2 BPP
877 WriteLog("OP: Fixed bitmap @ 2 BPP requesting FIRSTPIX! (fp=%u)\n", firstPix);
878 index &= 0xFC; // Top six bits form CLUT index
879 // The LSB is OPFLAG_REFLECT, so sign extend it and or 2 into it.
880 int32 lbufDelta = ((int8)((flags << 7) & 0xFF) >> 5) | 0x02;
885 uint64 pixels = ((uint64)JaguarReadLong(data, OP) << 32) | JaguarReadLong(data + 4, OP);
888 for(int i=0; i<32; i++)
890 uint8 bits = pixels >> 62;
891 // Seems to me that both of these are in the same endian, so we could cast it as
892 // uint16 * and do straight across copies (what about 24 bpp? Treat it differently...)
893 // This only works for the palettized modes (1 - 8 BPP), since we actually have to
894 // copy data from memory in 16 BPP mode (or does it? Isn't this the same as the CLUT case?)
895 // No, it isn't because we read the memory in an endian safe way--this *won't* work...
896 if (flagTRANS && bits == 0)
901 *(uint16 *)currentLineBuffer = paletteRAM16[index | bits];
904 BLEND_CR(*currentLineBuffer, paletteRAM[(index | bits) << 1]),
905 *(currentLineBuffer + 1) =
906 BLEND_Y(*(currentLineBuffer + 1), paletteRAM[((index | bits) << 1) + 1]);
909 currentLineBuffer += lbufDelta;
914 else if (depth == 2) // 4 BPP
917 WriteLog("OP: Fixed bitmap @ 4 BPP requesting FIRSTPIX! (fp=%u)\n", firstPix);
918 index &= 0xF0; // Top four bits form CLUT index
919 // The LSB is OPFLAG_REFLECT, so sign extend it and or 2 into it.
920 int32 lbufDelta = ((int8)((flags << 7) & 0xFF) >> 5) | 0x02;
925 uint64 pixels = ((uint64)JaguarReadLong(data, OP) << 32) | JaguarReadLong(data + 4, OP);
928 for(int i=0; i<16; i++)
930 uint8 bits = pixels >> 60;
931 // Seems to me that both of these are in the same endian, so we could cast it as
932 // uint16 * and do straight across copies (what about 24 bpp? Treat it differently...)
933 // This only works for the palettized modes (1 - 8 BPP), since we actually have to
934 // copy data from memory in 16 BPP mode (or does it? Isn't this the same as the CLUT case?)
935 // No, it isn't because we read the memory in an endian safe way--this *won't* work...
936 if (flagTRANS && bits == 0)
941 *(uint16 *)currentLineBuffer = paletteRAM16[index | bits];
944 BLEND_CR(*currentLineBuffer, paletteRAM[(index | bits) << 1]),
945 *(currentLineBuffer + 1) =
946 BLEND_Y(*(currentLineBuffer + 1), paletteRAM[((index | bits) << 1) + 1]);
949 currentLineBuffer += lbufDelta;
954 else if (depth == 3) // 8 BPP
956 // The LSB is OPFLAG_REFLECT, so sign extend it and or 2 into it.
957 int32 lbufDelta = ((int8)((flags << 7) & 0xFF) >> 5) | 0x02;
959 // Fetch 1st phrase...
960 uint64 pixels = ((uint64)JaguarReadLong(data, OP) << 32) | JaguarReadLong(data + 4, OP);
961 //Note that firstPix should only be honored *if* we start with the 1st phrase of the bitmap
962 //i.e., we didn't clip on the margin... !!! FIX !!!
963 firstPix &= 0x30; // Only top two bits are valid for 8 BPP
964 pixels <<= firstPix; // Skip first N pixels (N=firstPix)...
965 int i = firstPix >> 3; // Start counter at right spot...
971 uint8 bits = pixels >> 56;
972 // Seems to me that both of these are in the same endian, so we could cast it as
973 // uint16 * and do straight across copies (what about 24 bpp? Treat it differently...)
974 // This only works for the palettized modes (1 - 8 BPP), since we actually have to
975 // copy data from memory in 16 BPP mode (or does it? Isn't this the same as the CLUT case?)
976 // No, it isn't because we read the memory in an endian safe way--this *won't* work...
977 if (flagTRANS && bits == 0)
982 *(uint16 *)currentLineBuffer = paletteRAM16[bits];
985 BLEND_CR(*currentLineBuffer, paletteRAM[bits << 1]),
986 *(currentLineBuffer + 1) =
987 BLEND_Y(*(currentLineBuffer + 1), paletteRAM[(bits << 1) + 1]);
990 currentLineBuffer += lbufDelta;
994 // Fetch next phrase...
996 pixels = ((uint64)JaguarReadLong(data, OP) << 32) | JaguarReadLong(data + 4, OP);
999 else if (depth == 4) // 16 BPP
1002 WriteLog("OP: Fixed bitmap @ 16 BPP requesting FIRSTPIX! (fp=%u)\n", firstPix);
1003 // The LSB is OPFLAG_REFLECT, so sign extend it and or 2 into it.
1004 int32 lbufDelta = ((int8)((flags << 7) & 0xFF) >> 5) | 0x02;
1009 uint64 pixels = ((uint64)JaguarReadLong(data, OP) << 32) | JaguarReadLong(data + 4, OP);
1012 for(int i=0; i<4; i++)
1014 uint8 bitsHi = pixels >> 56, bitsLo = pixels >> 48;
1015 // Seems to me that both of these are in the same endian, so we could cast it as
1016 // uint16 * and do straight across copies (what about 24 bpp? Treat it differently...)
1017 // This only works for the palettized modes (1 - 8 BPP), since we actually have to
1018 // copy data from memory in 16 BPP mode (or does it? Isn't this the same as the CLUT case?)
1019 // No, it isn't because we read the memory in an endian safe way--it *won't* work...
1020 if (flagTRANS && (bitsLo | bitsHi) == 0)
1025 *currentLineBuffer = bitsHi,
1026 *(currentLineBuffer + 1) = bitsLo;
1028 *currentLineBuffer =
1029 BLEND_CR(*currentLineBuffer, bitsHi),
1030 *(currentLineBuffer + 1) =
1031 BLEND_Y(*(currentLineBuffer + 1), bitsLo);
1034 currentLineBuffer += lbufDelta;
1039 else if (depth == 5) // 24 BPP
1041 //Looks like Iron Soldier is the only game that uses 24BPP mode...
1042 //There *might* be others...
1043 //WriteLog("OP: Writing 24 BPP bitmap!\n");
1045 WriteLog("OP: Fixed bitmap @ 24 BPP requesting FIRSTPIX! (fp=%u)\n", firstPix);
1046 // Not sure, but I think RMW only works with 16 BPP and below, and only in CRY mode...
1047 // The LSB of flags is OPFLAG_REFLECT, so sign extend it and OR 4 into it.
1048 int32 lbufDelta = ((int8)((flags << 7) & 0xFF) >> 4) | 0x04;
1053 uint64 pixels = ((uint64)JaguarReadLong(data, OP) << 32) | JaguarReadLong(data + 4, OP);
1056 for(int i=0; i<2; i++)
1058 // We don't use a 32-bit var here because of endian issues...!
1059 uint8 bits3 = pixels >> 56, bits2 = pixels >> 48,
1060 bits1 = pixels >> 40, bits0 = pixels >> 32;
1062 if (flagTRANS && (bits3 | bits2 | bits1 | bits0) == 0)
1065 *currentLineBuffer = bits3,
1066 *(currentLineBuffer + 1) = bits2,
1067 *(currentLineBuffer + 2) = bits1,
1068 *(currentLineBuffer + 3) = bits0;
1070 currentLineBuffer += lbufDelta;
1078 // Store scaled bitmap in line buffer
1080 void OPProcessScaledBitmap(uint64 p0, uint64 p1, uint64 p2, bool render)
1082 // Need to make sure that when writing that it stays within the line buffer...
1083 // LBUF ($F01800 - $F01D9E) 360 x 32-bit RAM
1084 uint8 depth = (p1 >> 12) & 0x07; // Color depth of image
1085 int32 xpos = ((int16)((p1 << 4) & 0xFFFF)) >> 4;// Image xpos in LBUF
1086 uint32 iwidth = (p1 >> 28) & 0x3FF; // Image width in *phrases*
1087 uint32 data = (p0 >> 40) & 0xFFFFF8; // Pixel data address
1088 //#ifdef OP_DEBUG_BMP
1089 // Prolly should use this... Though not sure exactly how.
1090 //Use the upper bits as an offset into the phrase depending on the BPP. That's how!
1091 uint32 firstPix = (p1 >> 49) & 0x3F;
1092 //This is WEIRD! I'm sure I saw Atari Karts request 8 BPP FIRSTPIX! What happened???
1094 WriteLog("OP: FIRSTPIX != 0! (Scaled BM)\n");
1096 // We can ignore the RELEASE (high order) bit for now--probably forever...!
1097 // uint8 flags = (p1 >> 45) & 0x0F; // REFLECT, RMW, TRANS, RELEASE
1098 //Optimize: break these out to their own BOOL values [DONE]
1099 uint8 flags = (p1 >> 45) & 0x07; // REFLECT (0), RMW (1), TRANS (2)
1100 bool flagREFLECT = (flags & OPFLAG_REFLECT ? true : false),
1101 flagRMW = (flags & OPFLAG_RMW ? true : false),
1102 flagTRANS = (flags & OPFLAG_TRANS ? true : false);
1103 uint8 index = (p1 >> 37) & 0xFE; // CLUT index offset (upper pix, 1-4 bpp)
1104 uint32 pitch = (p1 >> 15) & 0x07; // Phrase pitch
1106 uint8 * tom_ram_8 = tom_get_ram_pointer();
1107 uint8 * paletteRAM = &tom_ram_8[0x400];
1108 // This is OK as long as it's used correctly: For 16-bit RAM to RAM direct copies--NOT
1109 // for use when using endian-corrected data (i.e., any of the *ReadWord functions!)
1110 uint16 * paletteRAM16 = (uint16 *)paletteRAM;
1112 uint8 hscale = p2 & 0xFF;
1113 // uint8 horizontalRemainder = hscale; // Not sure if it starts full, but seems reasonable [It's not!]
1114 uint8 horizontalRemainder = 0; // Let's try zero! Seems to work! Yay!
1115 int32 scaledWidthInPixels = (iwidth * phraseWidthToPixels[depth] * hscale) >> 5;
1116 uint32 scaledPhrasePixels = (phraseWidthToPixels[depth] * hscale) >> 5;
1118 // WriteLog("bitmap %ix? %ibpp at %i,? firstpix=? data=0x%.8x pitch %i hflipped=%s dwidth=? (linked to ?) RMW=%s Tranparent=%s\n",
1119 // iwidth, op_bitmap_bit_depth[bitdepth], xpos, ptr, pitch, (flags&OPFLAG_REFLECT ? "yes" : "no"), (flags&OPFLAG_RMW ? "yes" : "no"), (flags&OPFLAG_TRANS ? "yes" : "no"));
1121 // Looks like an hscale of zero means don't draw!
1122 if (!render || iwidth == 0 || hscale == 0)
1125 /*extern int start_logging;
1127 WriteLog("OP: Scaled bitmap %ix? %ibpp at %i,? hscale=%02X fpix=%i data=%08X pitch %i hflipped=%s dwidth=? (linked to %08X) Transluency=%s\n",
1128 iwidth, op_bitmap_bit_depth[depth], xpos, hscale, firstPix, data, pitch, (flagREFLECT ? "yes" : "no"), op_pointer, (flagRMW ? "yes" : "no"));*/
1129 //#define OP_DEBUG_BMP
1130 //#ifdef OP_DEBUG_BMP
1131 // WriteLog("OP: Scaled bitmap %ix%i %ibpp at %i,%i firstpix=%i data=0x%.8x pitch %i hflipped=%s dwidth=%i (linked to 0x%.8x) Transluency=%s\n",
1132 // iwidth, height, op_bitmap_bit_depth[bitdepth], xpos, ypos, firstPix, ptr, pitch, (flags&OPFLAG_REFLECT ? "yes" : "no"), dwidth, op_pointer, (flags&OPFLAG_RMW ? "yes" : "no"));
1135 int32 startPos = xpos, endPos = xpos +
1136 (!flagREFLECT ? scaledWidthInPixels - 1 : -(scaledWidthInPixels + 1));
1137 uint32 clippedWidth = 0, phraseClippedWidth = 0, dataClippedWidth = 0;
1138 bool in24BPPMode = (((GET16(tom_ram_8, 0x0028) >> 1) & 0x03) == 1 ? true : false); // VMODE
1139 // Not sure if this is Jaguar Two only location or what...
1140 // From the docs, it is... If we want to limit here we should think of something else.
1141 // int32 limit = GET16(tom_ram_8, 0x0008); // LIMIT
1143 int32 lbufWidth = (!in24BPPMode ? limit - 1 : (limit / 2) - 1); // Zero based limit...
1145 // If the image is completely to the left or right of the line buffer, then bail.
1146 //If in REFLECT mode, then these values are swapped! !!! FIX !!! [DONE]
1147 //There are four possibilities:
1148 // 1. image sits on left edge and no REFLECT; starts out of bounds but ends in bounds.
1149 // 2. image sits on left edge and REFLECT; starts in bounds but ends out of bounds.
1150 // 3. image sits on right edge and REFLECT; starts out of bounds but ends in bounds.
1151 // 4. image sits on right edge and no REFLECT; starts in bounds but ends out of bounds.
1152 //Numbers 2 & 4 can be caught by checking the LBUF clip while in the inner loop,
1153 // numbers 1 & 3 are of concern.
1154 // This *indirectly* handles only cases 2 & 4! And is WRONG if REFLECT is set...!
1155 // if (rightMargin < 0 || leftMargin > lbufWidth)
1157 // It might be easier to swap these (if REFLECTed) and just use XPOS down below...
1158 // That way, you could simply set XPOS to leftMargin if !REFLECT and to rightMargin otherwise.
1159 // Still have to be careful with the DATA and IWIDTH values though...
1161 if ((!flagREFLECT && (endPos < 0 || startPos > lbufWidth))
1162 || (flagREFLECT && (startPos < 0 || endPos > lbufWidth)))
1165 // Otherwise, find the clip limits and clip the phrase as well...
1166 // NOTE: I'm fudging here by letting the actual blit overstep the bounds of the
1167 // line buffer, but it shouldn't matter since there are two unused line
1168 // buffers below and nothing above and I'll at most write 40 bytes outside
1169 // the line buffer... I could use a fractional clip begin/end value, but
1170 // this makes the blit a *lot* more hairy. I might fix this in the future
1171 // if it becomes necessary. (JLH)
1172 // Probably wouldn't be *that* hairy. Just use a delta that tells the inner loop
1173 // which pixel in the phrase is being written, and quit when either end of phrases
1174 // is reached or line buffer extents are surpassed.
1176 //This stuff is probably wrong as well... !!! FIX !!!
1177 //The strange thing is that it seems to work, but that's no guarantee that it's bulletproof!
1178 //Yup. Seems that JagMania doesn't work correctly with this...
1179 //Dunno if this is the problem, but Atari Karts is showing *some* of the road now...
1180 //Actually, it is! Or, it was. It doesn't seem to be clipping here, so the problem lies
1181 //elsewhere! Hmm. Putting the scaling code into the 1/2/8 BPP cases seems to draw the ground
1182 // a bit more accurately... Strange!
1183 //It's probably a case of the REFLECT flag being set and the background being written
1184 //from the right side of the screen...
1185 //But no, it isn't... At least if the diagnostics are telling the truth!
1187 // NOTE: We're just using endPos to figure out how much, if any, to clip by.
1188 // ALSO: There may be another case where we start out of bounds and end out of bounds...!
1191 //There's a problem here with scaledPhrasePixels in that it can be forced to zero when
1192 //the scaling factor is small. So fix it already! !!! FIX !!!
1193 /*if (scaledPhrasePixels == 0)
1195 WriteLog("OP: [Scaled] We're about to encounter a divide by zero error!\n");
1196 DumpScaledObject(p0, p1, p2);
1198 //NOTE: I'm almost 100% sure that this is wrong... And it is! :-p
1200 //Try a simple example...
1201 // Let's say we have a 8 BPP scanline with an hscale of $80 (4). Our xpos is -10,
1202 // non-flipped. Pixels in the bitmap are XYZXYZXYZXYZXYZ.
1203 // Scaled up, they would be XXXXYYYYZZZZXXXXYYYYZZZZXXXXYYYYZZZZ...
1205 // Normally, we would expect this in the line buffer:
1206 // ZZXXXXYYYYZZZZXXXXYYYYZZZZ...
1208 // But instead we're getting:
1209 // XXXXYYYYZZZZXXXXYYYYZZZZ...
1211 // or are we??? It would seem so, simply by virtue of the fact that we're NOT starting
1212 // on negative boundary--or are we? Hmm...
1213 // cw = 10, dcw = pcw = 10 / ([8 * 4 = 32] 32) = 0, sp = -10
1215 // Let's try a real world example:
1217 //OP: Scaled bitmap (70, 8 BPP, spp=28) sp (-400) < 0... [new sp=-8, cw=400, dcw=pcw=14]
1218 //OP: Scaled bitmap (6F, 8 BPP, spp=27) sp (-395) < 0... [new sp=-17, cw=395, dcw=pcw=14]
1220 // Really, spp is 27.75 in the second case...
1221 // So... If we do 395 / 27.75, we get 14. Ok so far... If we scale that against the
1222 // start position (14 * 27.75), we get -6.5... NOT -17!
1224 //Now it seems we're working OK, at least for the first case...
1225 uint32 scaledPhrasePixelsUS = phraseWidthToPixels[depth] * hscale;
1227 if (startPos < 0) // Case #1: Begin out, end in, L to R
1229 extern int start_logging;
1231 WriteLog("OP: Scaled bitmap (%02X, %u BPP, spp=%u) start pos (%i) < 0...", hscale, op_bitmap_bit_depth[depth], scaledPhrasePixels, startPos);
1232 // clippedWidth = 0 - startPos,
1233 clippedWidth = (0 - startPos) << 5,
1234 // dataClippedWidth = phraseClippedWidth = clippedWidth / scaledPhrasePixels,
1235 dataClippedWidth = phraseClippedWidth = (clippedWidth / scaledPhrasePixelsUS) >> 5,
1236 // startPos = 0 - (clippedWidth % scaledPhrasePixels);
1237 startPos += (dataClippedWidth * scaledPhrasePixelsUS) >> 5;
1239 WriteLog(" [new sp=%i, cw=%i, dcw=pcw=%i]\n", startPos, clippedWidth, dataClippedWidth);
1242 if (endPos < 0) // Case #2: Begin in, end out, R to L
1243 clippedWidth = 0 - endPos,
1244 phraseClippedWidth = clippedWidth / scaledPhrasePixels;
1246 if (endPos > lbufWidth) // Case #3: Begin in, end out, L to R
1247 clippedWidth = endPos - lbufWidth,
1248 phraseClippedWidth = clippedWidth / scaledPhrasePixels;
1250 if (startPos > lbufWidth) // Case #4: Begin out, end in, R to L
1251 clippedWidth = startPos - lbufWidth,
1252 dataClippedWidth = phraseClippedWidth = clippedWidth / scaledPhrasePixels,
1253 startPos = lbufWidth + (clippedWidth % scaledPhrasePixels);
1255 extern int op_start_log;
1256 if (op_start_log && clippedWidth != 0)
1257 WriteLog("OP: Clipped line. SP=%i, EP=%i, clip=%u, iwidth=%u, hscale=%02X\n", startPos, endPos, clippedWidth, iwidth, hscale);
1258 if (op_start_log && startPos == 13)
1260 WriteLog("OP: Scaled line. SP=%i, EP=%i, clip=%u, iwidth=%u, hscale=%02X, depth=%u, firstPix=%u\n", startPos, endPos, clippedWidth, iwidth, hscale, depth, firstPix);
1261 DumpScaledObject(p0, p1, p2);
1263 // If the image is sitting on the line buffer left or right edge, we need to compensate
1264 // by decreasing the image phrase width accordingly.
1265 iwidth -= phraseClippedWidth;
1267 // Also, if we're clipping the phrase we need to make sure we're in the correct part of
1269 // data += phraseClippedWidth * (pitch << 3);
1270 data += dataClippedWidth * (pitch << 3);
1272 // NOTE: When the bitmap is in REFLECT mode, the XPOS marks the *right* side of the
1273 // bitmap! This makes clipping & etc. MUCH, much easier...!
1274 // uint32 lbufAddress = 0x1800 + (!in24BPPMode ? leftMargin * 2 : leftMargin * 4);
1275 // uint32 lbufAddress = 0x1800 + (!in24BPPMode ? startPos * 2 : startPos * 4);
1276 uint32 lbufAddress = 0x1800 + startPos * 2;
1277 uint8 * currentLineBuffer = &tom_ram_8[lbufAddress];
1278 uint8 * lineBufferLowerLimit = &tom_ram_8[0x1800],
1279 * lineBufferUpperLimit = &tom_ram_8[0x1800 + 719];
1283 // Hmm. We check above for 24 BPP mode, but don't do anything about it below...
1284 // If we *were* in 24 BPP mode, how would you convert CRY to RGB24? Seems to me
1285 // that if you're in CRY mode then you wouldn't be able to use 24 BPP bitmaps
1287 // This seems to be the case (at least according to the Midsummer docs)...!
1289 if (depth == 0) // 1 BPP
1292 WriteLog("OP: Scaled bitmap @ 1 BPP requesting FIRSTPIX!\n");
1293 // The LSB of flags is OPFLAG_REFLECT, so sign extend it and or 2 into it.
1294 int32 lbufDelta = ((int8)((flags << 7) & 0xFF) >> 5) | 0x02;
1297 uint64 pixels = ((uint64)JaguarReadLong(data, OP) << 32) | JaguarReadLong(data + 4, OP);
1299 while ((int32)iwidth > 0)
1301 uint8 bits = pixels >> 63;
1303 if (flagTRANS && bits == 0)
1308 // This is the *only* correct use of endian-dependent code
1309 // (i.e., mem-to-mem direct copying)!
1310 *(uint16 *)currentLineBuffer = paletteRAM16[index | bits];
1312 *currentLineBuffer =
1313 BLEND_CR(*currentLineBuffer, paletteRAM[(index | bits) << 1]),
1314 *(currentLineBuffer + 1) =
1315 BLEND_Y(*(currentLineBuffer + 1), paletteRAM[((index | bits) << 1) + 1]);
1318 currentLineBuffer += lbufDelta;
1320 horizontalRemainder -= 0x20; // Subtract 1.0f in [3.5] fixed point format
1321 while (horizontalRemainder & 0x80)
1323 horizontalRemainder += hscale;
1330 int phrasesToSkip = pixCount / 64, pixelShift = pixCount % 64;
1332 data += (pitch << 3) * phrasesToSkip;
1333 pixels = ((uint64)JaguarReadLong(data, OP) << 32) | JaguarReadLong(data + 4, OP);
1334 pixels <<= 1 * pixelShift;
1335 iwidth -= phrasesToSkip;
1336 pixCount = pixelShift;
1340 else if (depth == 1) // 2 BPP
1343 WriteLog("OP: Scaled bitmap @ 2 BPP requesting FIRSTPIX!\n");
1344 index &= 0xFC; // Top six bits form CLUT index
1345 // The LSB is OPFLAG_REFLECT, so sign extend it and or 2 into it.
1346 int32 lbufDelta = ((int8)((flags << 7) & 0xFF) >> 5) | 0x02;
1349 uint64 pixels = ((uint64)JaguarReadLong(data, OP) << 32) | JaguarReadLong(data + 4, OP);
1351 while ((int32)iwidth > 0)
1353 uint8 bits = pixels >> 62;
1355 if (flagTRANS && bits == 0)
1360 // This is the *only* correct use of endian-dependent code
1361 // (i.e., mem-to-mem direct copying)!
1362 *(uint16 *)currentLineBuffer = paletteRAM16[index | bits];
1364 *currentLineBuffer =
1365 BLEND_CR(*currentLineBuffer, paletteRAM[(index | bits) << 1]),
1366 *(currentLineBuffer + 1) =
1367 BLEND_Y(*(currentLineBuffer + 1), paletteRAM[((index | bits) << 1) + 1]);
1370 currentLineBuffer += lbufDelta;
1372 horizontalRemainder -= 0x20; // Subtract 1.0f in [3.5] fixed point format
1373 while (horizontalRemainder & 0x80)
1375 horizontalRemainder += hscale;
1382 int phrasesToSkip = pixCount / 32, pixelShift = pixCount % 32;
1384 data += (pitch << 3) * phrasesToSkip;
1385 pixels = ((uint64)JaguarReadLong(data, OP) << 32) | JaguarReadLong(data + 4, OP);
1386 pixels <<= 2 * pixelShift;
1387 iwidth -= phrasesToSkip;
1388 pixCount = pixelShift;
1392 else if (depth == 2) // 4 BPP
1395 WriteLog("OP: Scaled bitmap @ 4 BPP requesting FIRSTPIX!\n");
1396 index &= 0xF0; // Top four bits form CLUT index
1397 // The LSB is OPFLAG_REFLECT, so sign extend it and or 2 into it.
1398 int32 lbufDelta = ((int8)((flags << 7) & 0xFF) >> 5) | 0x02;
1401 uint64 pixels = ((uint64)JaguarReadLong(data, OP) << 32) | JaguarReadLong(data + 4, OP);
1403 while ((int32)iwidth > 0)
1405 uint8 bits = pixels >> 60;
1407 if (flagTRANS && bits == 0)
1412 // This is the *only* correct use of endian-dependent code
1413 // (i.e., mem-to-mem direct copying)!
1414 *(uint16 *)currentLineBuffer = paletteRAM16[index | bits];
1416 *currentLineBuffer =
1417 BLEND_CR(*currentLineBuffer, paletteRAM[(index | bits) << 1]),
1418 *(currentLineBuffer + 1) =
1419 BLEND_Y(*(currentLineBuffer + 1), paletteRAM[((index | bits) << 1) + 1]);
1422 currentLineBuffer += lbufDelta;
1424 horizontalRemainder -= 0x20; // Subtract 1.0f in [3.5] fixed point format
1425 while (horizontalRemainder & 0x80)
1427 horizontalRemainder += hscale;
1434 int phrasesToSkip = pixCount / 16, pixelShift = pixCount % 16;
1436 data += (pitch << 3) * phrasesToSkip;
1437 pixels = ((uint64)JaguarReadLong(data, OP) << 32) | JaguarReadLong(data + 4, OP);
1438 pixels <<= 4 * pixelShift;
1439 iwidth -= phrasesToSkip;
1440 pixCount = pixelShift;
1444 else if (depth == 3) // 8 BPP
1447 WriteLog("OP: Scaled bitmap @ 8 BPP requesting FIRSTPIX! (fp=%u)\n", firstPix);
1448 // The LSB is OPFLAG_REFLECT, so sign extend it and or 2 into it.
1449 int32 lbufDelta = ((int8)((flags << 7) & 0xFF) >> 5) | 0x02;
1452 uint64 pixels = ((uint64)JaguarReadLong(data, OP) << 32) | JaguarReadLong(data + 4, OP);
1454 while ((int32)iwidth > 0)
1456 uint8 bits = pixels >> 56;
1458 if (flagTRANS && bits == 0)
1463 // This is the *only* correct use of endian-dependent code
1464 // (i.e., mem-to-mem direct copying)!
1465 *(uint16 *)currentLineBuffer = paletteRAM16[bits];
1467 if (currentLineBuffer >= lineBufferLowerLimit && currentLineBuffer <= lineBufferUpperLimit)
1468 *(uint16 *)currentLineBuffer = paletteRAM16[bits];
1471 *currentLineBuffer =
1472 BLEND_CR(*currentLineBuffer, paletteRAM[bits << 1]),
1473 *(currentLineBuffer + 1) =
1474 BLEND_Y(*(currentLineBuffer + 1), paletteRAM[(bits << 1) + 1]);
1477 currentLineBuffer += lbufDelta;
1479 horizontalRemainder -= 0x20; // Subtract 1.0f in [3.5] fixed point format
1480 while (horizontalRemainder & 0x80)
1482 horizontalRemainder += hscale;
1489 int phrasesToSkip = pixCount / 8, pixelShift = pixCount % 8;
1491 data += (pitch << 3) * phrasesToSkip;
1492 pixels = ((uint64)JaguarReadLong(data, OP) << 32) | JaguarReadLong(data + 4, OP);
1493 pixels <<= 8 * pixelShift;
1494 iwidth -= phrasesToSkip;
1495 pixCount = pixelShift;
1499 else if (depth == 4) // 16 BPP
1502 WriteLog("OP: Scaled bitmap @ 16 BPP requesting FIRSTPIX!\n");
1503 // The LSB is OPFLAG_REFLECT, so sign extend it and OR 2 into it.
1504 int32 lbufDelta = ((int8)((flags << 7) & 0xFF) >> 5) | 0x02;
1507 uint64 pixels = ((uint64)JaguarReadLong(data, OP) << 32) | JaguarReadLong(data + 4, OP);
1509 while ((int32)iwidth > 0)
1511 uint8 bitsHi = pixels >> 56, bitsLo = pixels >> 48;
1513 if (flagTRANS && (bitsLo | bitsHi) == 0)
1518 *currentLineBuffer = bitsHi,
1519 *(currentLineBuffer + 1) = bitsLo;
1521 *currentLineBuffer =
1522 BLEND_CR(*currentLineBuffer, bitsHi),
1523 *(currentLineBuffer + 1) =
1524 BLEND_Y(*(currentLineBuffer + 1), bitsLo);
1527 currentLineBuffer += lbufDelta;
1529 horizontalRemainder -= 0x20; // Subtract 1.0f in [3.5] fixed point format
1530 while (horizontalRemainder & 0x80)
1532 horizontalRemainder += hscale;
1539 int phrasesToSkip = pixCount / 4, pixelShift = pixCount % 4;
1541 data += (pitch << 3) * phrasesToSkip;
1542 pixels = ((uint64)JaguarReadLong(data, OP) << 32) | JaguarReadLong(data + 4, OP);
1543 pixels <<= 16 * pixelShift;
1545 iwidth -= phrasesToSkip;
1547 pixCount = pixelShift;
1551 else if (depth == 5) // 24 BPP
1553 //I'm not sure that you can scale a 24 BPP bitmap properly--the JTRM seem to indicate as much.
1554 WriteLog("OP: Writing 24 BPP scaled bitmap!\n");
1556 WriteLog("OP: Scaled bitmap @ 24 BPP requesting FIRSTPIX!\n");
1557 // Not sure, but I think RMW only works with 16 BPP and below, and only in CRY mode...
1558 // The LSB is OPFLAG_REFLECT, so sign extend it and or 4 into it.
1559 int32 lbufDelta = ((int8)((flags << 7) & 0xFF) >> 4) | 0x04;
1564 uint64 pixels = ((uint64)JaguarReadLong(data, OP) << 32) | JaguarReadLong(data + 4, OP);
1565 data += pitch << 3; // Multiply pitch * 8 (optimize: precompute this value)
1567 for(int i=0; i<2; i++)
1569 uint8 bits3 = pixels >> 56, bits2 = pixels >> 48,
1570 bits1 = pixels >> 40, bits0 = pixels >> 32;
1572 if (flagTRANS && (bits3 | bits2 | bits1 | bits0) == 0)
1575 *currentLineBuffer = bits3,
1576 *(currentLineBuffer + 1) = bits2,
1577 *(currentLineBuffer + 2) = bits1,
1578 *(currentLineBuffer + 3) = bits0;
1580 currentLineBuffer += lbufDelta;
1585 /*if (depth == 3 && startPos == 13)
1588 WriteLog("OP: Writing in the margins...\n");
1589 for(int i=0; i<100*2; i+=2)
1590 // for(int i=0; i<14*2; i+=2)
1591 tom_ram_8[0x1800 + i] = 0xFF,
1592 tom_ram_8[0x1800 + i + 1] = 0xFF;
1594 // uint32 lbufAddress = 0x1800 + (!in24BPPMode ? startPos * 2 : startPos * 4);
1595 // uint8 * currentLineBuffer = &tom_ram_8[lbufAddress];