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 //WAS: int16 ypos = ((p0 >> 3) & 0x3FF); // ??? What if not interlaced (/2)?
262 int16 ypos = ((p0 >> 3) & 0x7FF); // ??? What if not interlaced (/2)?
263 int32 xpos = p1 & 0xFFF;
264 xpos = (xpos & 0x800 ? xpos | 0xFFFFF000 : xpos);
265 uint32 iwidth = ((p1 >> 28) & 0x3FF);
266 uint32 dwidth = ((p1 >> 18) & 0x3FF); // Unsigned!
267 uint16 height = ((p0 >> 14) & 0x3FF);
268 uint32 link = ((p0 >> 24) & 0x7FFFF) << 3;
269 uint32 ptr = ((p0 >> 43) & 0x1FFFFF) << 3;
270 uint32 firstPix = (p1 >> 49) & 0x3F;
271 uint8 flags = (p1 >> 45) & 0x0F;
272 uint8 idx = (p1 >> 38) & 0x7F;
273 uint32 pitch = (p1 >> 15) & 0x07;
274 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",
275 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);
276 uint32 hscale = p2 & 0xFF;
277 uint32 vscale = (p2 >> 8) & 0xFF;
278 uint32 remainder = (p2 >> 16) & 0xFF;
279 WriteLog(" [hsc: %02X, vsc: %02X, rem: %02X]\n", hscale, vscale, remainder);
282 void DumpFixedObject(uint64 p0, uint64 p1)
284 WriteLog(" (BITMAP)");
285 WriteLog(" %08X --> phrase %08X %08X\n", op_pointer, (uint32)(p1>>32), (uint32)(p1&0xFFFFFFFF));
286 uint8 bitdepth = (p1 >> 12) & 0x07;
287 //WAS: int16 ypos = ((p0 >> 3) & 0x3FF); // ??? What if not interlaced (/2)?
288 int16 ypos = ((p0 >> 3) & 0x7FF); // ??? What if not interlaced (/2)?
289 int32 xpos = p1 & 0xFFF;
290 xpos = (xpos & 0x800 ? xpos | 0xFFFFF000 : xpos);
291 uint32 iwidth = ((p1 >> 28) & 0x3FF);
292 uint32 dwidth = ((p1 >> 18) & 0x3FF); // Unsigned!
293 uint16 height = ((p0 >> 14) & 0x3FF);
294 uint32 link = ((p0 >> 24) & 0x7FFFF) << 3;
295 uint32 ptr = ((p0 >> 43) & 0x1FFFFF) << 3;
296 uint32 firstPix = (p1 >> 49) & 0x3F;
297 uint8 flags = (p1 >> 45) & 0x0F;
298 uint8 idx = (p1 >> 38) & 0x7F;
299 uint32 pitch = (p1 >> 15) & 0x07;
300 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",
301 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);
305 // Object Processor main routine
307 //Need to fix this so that when an GPU object IRQ happens, we can pick up OP processing
308 //where we left off. !!! FIX !!!
309 void OPProcessList(int scanline, bool render)
311 extern int op_start_log;
312 // char * condition_to_str[8] =
313 // { "==", "<", ">", "(opflag set)", "(second half line)", "?", "?", "?" };
315 op_pointer = op_get_list_pointer();
317 // objectp_stop_reading_list = false;
319 // *** BEGIN OP PROCESSOR TESTING ONLY ***
320 extern bool interactiveMode;
322 extern int objectPtr;
324 int bitmapCounter = 0;
325 // *** END OP PROCESSOR TESTING ONLY ***
327 // if (op_pointer) WriteLog(" new op list at 0x%.8x scanline %i\n",op_pointer,scanline);
330 // *** BEGIN OP PROCESSOR TESTING ONLY ***
331 if (interactiveMode && bitmapCounter == objectPtr)
335 // *** END OP PROCESSOR TESTING ONLY ***
336 // if (objectp_stop_reading_list)
339 uint64 p0 = op_load_phrase(op_pointer);
341 if (scanline == tom_get_vdb() && op_start_log)
342 //if (scanline == 215 && op_start_log)
344 WriteLog("%08X --> phrase %08X %08X", op_pointer - 8, (int)(p0>>32), (int)(p0&0xFFFFFFFF));
345 if ((p0 & 0x07) == OBJECT_TYPE_BITMAP)
347 WriteLog(" (BITMAP) ");
348 uint64 p1 = op_load_phrase(op_pointer);
349 WriteLog("\n%08X --> phrase %08X %08X ", op_pointer, (int)(p1>>32), (int)(p1&0xFFFFFFFF));
350 uint8 bitdepth = (p1 >> 12) & 0x07;
351 //WAS: int16 ypos = ((p0 >> 3) & 0x3FF); // ??? What if not interlaced (/2)?
352 int16 ypos = ((p0 >> 3) & 0x7FF); // ??? What if not interlaced (/2)?
353 int32 xpos = p1 & 0xFFF;
354 xpos = (xpos & 0x800 ? xpos | 0xFFFFF000 : xpos);
355 uint32 iwidth = ((p1 >> 28) & 0x3FF);
356 uint32 dwidth = ((p1 >> 18) & 0x3FF); // Unsigned!
357 uint16 height = ((p0 >> 14) & 0x3FF);
358 uint32 link = ((p0 >> 24) & 0x7FFFF) << 3;
359 uint32 ptr = ((p0 >> 43) & 0x1FFFFF) << 3;
360 uint32 firstPix = (p1 >> 49) & 0x3F;
361 uint8 flags = (p1 >> 45) & 0x0F;
362 uint8 idx = (p1 >> 38) & 0x7F;
363 uint32 pitch = (p1 >> 15) & 0x07;
364 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",
365 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);
367 if ((p0 & 0x07) == OBJECT_TYPE_SCALE)
369 WriteLog(" (SCALED BITMAP)");
370 uint64 p1 = op_load_phrase(op_pointer), p2 = op_load_phrase(op_pointer+8);
371 WriteLog("\n%08X --> phrase %08X %08X ", op_pointer, (int)(p1>>32), (int)(p1&0xFFFFFFFF));
372 WriteLog("\n%08X --> phrase %08X %08X ", op_pointer+8, (int)(p2>>32), (int)(p2&0xFFFFFFFF));
373 uint8 bitdepth = (p1 >> 12) & 0x07;
374 //WAS: int16 ypos = ((p0 >> 3) & 0x3FF); // ??? What if not interlaced (/2)?
375 int16 ypos = ((p0 >> 3) & 0x7FF); // ??? What if not interlaced (/2)?
376 int32 xpos = p1 & 0xFFF;
377 xpos = (xpos & 0x800 ? xpos | 0xFFFFF000 : xpos);
378 uint32 iwidth = ((p1 >> 28) & 0x3FF);
379 uint32 dwidth = ((p1 >> 18) & 0x3FF); // Unsigned!
380 uint16 height = ((p0 >> 14) & 0x3FF);
381 uint32 link = ((p0 >> 24) & 0x7FFFF) << 3;
382 uint32 ptr = ((p0 >> 43) & 0x1FFFFF) << 3;
383 uint32 firstPix = (p1 >> 49) & 0x3F;
384 uint8 flags = (p1 >> 45) & 0x0F;
385 uint8 idx = (p1 >> 38) & 0x7F;
386 uint32 pitch = (p1 >> 15) & 0x07;
387 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",
388 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);
389 uint32 hscale = p2 & 0xFF;
390 uint32 vscale = (p2 >> 8) & 0xFF;
391 uint32 remainder = (p2 >> 16) & 0xFF;
392 WriteLog(" [hsc: %02X, vsc: %02X, rem: %02X]\n", hscale, vscale, remainder);
394 if ((p0 & 0x07) == OBJECT_TYPE_GPU)
395 WriteLog(" (GPU)\n");
396 if ((p0 & 0x07) == OBJECT_TYPE_BRANCH)
398 WriteLog(" (BRANCH)\n");
399 uint8 * jaguar_mainRam = GetRamPtr();
400 WriteLog("[RAM] --> ");
401 for(int k=0; k<8; k++)
402 WriteLog("%02X ", jaguar_mainRam[op_pointer-8 + k]);
405 if ((p0 & 0x07) == OBJECT_TYPE_STOP)
406 WriteLog(" --> List end\n");
409 // WriteLog("%08X type %i\n", op_pointer, (uint8)p0 & 0x07);
410 switch ((uint8)p0 & 0x07)
412 case OBJECT_TYPE_BITMAP:
414 //WAS: uint16 ypos = (p0 >> 3) & 0x3FF;
415 uint16 ypos = (p0 >> 3) & 0x7FF;
416 // This is only theory implied by Rayman...!
417 // It seems that if the YPOS is zero, then bump the YPOS value so that it coincides with
418 // the VDB value. With interlacing, this would be slightly more tricky.
419 // There's probably another bit somewhere that enables this mode--but so far, doesn't seem
420 // to affect any other game in a negative way (that I've seen).
421 // Either that, or it's an undocumented bug...
423 //No, the reason this was needed is that the OP code before was wrong. Any value
424 //less than VDB will get written to the top line of the display!
426 // ypos = TOMReadWord(0xF00046, OP) / 2; // Get the VDB value
427 uint32 height = (p0 & 0xFFC000) >> 14;
428 uint32 oldOPP = op_pointer - 8;
429 // *** BEGIN OP PROCESSOR TESTING ONLY ***
430 if (inhibit && op_start_log)
431 WriteLog("!!! ^^^ This object is INHIBITED! ^^^ !!!\n");
433 if (!inhibit) // For OP testing only!
434 // *** END OP PROCESSOR TESTING ONLY ***
435 if (scanline >= ypos && height > 0)
437 uint64 p1 = op_load_phrase(op_pointer);
439 //WriteLog("OP: Writing scanline %d with ypos == %d...\n", scanline, ypos);
440 //WriteLog("--> Writing %u BPP bitmap...\n", op_bitmap_bit_depth[(p1 >> 12) & 0x07]);
441 // OPProcessFixedBitmap(scanline, p0, p1, render);
442 OPProcessFixedBitmap(p0, p1, render);
446 //???Does this really happen??? Doesn't seem to work if you do this...!
447 //Probably not. Must be a bug in the documentation...!
448 // uint32 link = (p0 & 0x7FFFF000000) >> 21;
449 // SET16(objectp_ram, 0x20, link & 0xFFFF); // OLP
450 // SET16(objectp_ram, 0x22, link >> 16);
451 /* uint32 height = (p0 & 0xFFC000) >> 14;
454 // NOTE: Would subtract 2 if in interlaced mode...!
455 // uint64 height = ((p0 & 0xFFC000) - 0x4000) & 0xFFC000;
459 uint64 data = (p0 & 0xFFFFF80000000000LL) >> 40;
460 uint64 dwidth = (p1 & 0xFFC0000) >> 15;
463 p0 &= ~0xFFFFF80000FFC000LL; // Mask out old data...
464 p0 |= (uint64)height << 14;
466 OPStorePhrase(oldOPP, p0);
468 op_pointer = (p0 & 0x000007FFFF000000LL) >> 21;
471 case OBJECT_TYPE_SCALE:
473 //WAS: uint16 ypos = (p0 >> 3) & 0x3FF;
474 uint16 ypos = (p0 >> 3) & 0x7FF;
475 uint32 height = (p0 & 0xFFC000) >> 14;
476 uint32 oldOPP = op_pointer - 8;
477 // *** BEGIN OP PROCESSOR TESTING ONLY ***
478 if (inhibit && op_start_log)
480 WriteLog("!!! ^^^ This object is INHIBITED! ^^^ !!! (scanline=%u, ypos=%u, height=%u)\n", scanline, ypos, height);
481 DumpScaledObject(p0, op_load_phrase(op_pointer), op_load_phrase(op_pointer+8));
484 if (!inhibit) // For OP testing only!
485 // *** END OP PROCESSOR TESTING ONLY ***
486 if (scanline >= ypos && height > 0)
488 uint64 p1 = op_load_phrase(op_pointer);
490 uint64 p2 = op_load_phrase(op_pointer);
492 //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));
493 OPProcessScaledBitmap(p0, p1, p2, render);
497 uint8 remainder = p2 >> 16, vscale = p2 >> 8;
498 //Actually, we should skip this object if it has a vscale of zero.
499 //Or do we? Not sure... Atari Karts has a few lines that look like:
501 //000E8268 --> phrase 00010000 7000B00D
502 // [7 (0) x 1 @ (13, 0) (8 bpp), l: 000E82A0, p: 000E0FC0 fp: 00, fl:RELEASE, idx:00, pt:01]
503 // [hsc: 9A, vsc: 00, rem: 00]
504 // Could it be the vscale is overridden if the DWIDTH is zero? Hmm...
507 vscale = 0x20; // OP bug??? Nope, it isn't...! Or is it?
509 /*extern int start_logging;
511 WriteLog("--> Returned from scaled bitmap processing (rem=%02X, vscale=%02X)...\n", remainder, vscale);*/
513 //--> Returned from scaled bitmap processing (rem=20, vscale=80)...
514 //There are other problems here, it looks like...
516 OP: Scaled bitmap 4x? 4bpp at 38,? hscale=7C fpix=0 data=00075E28 pitch 1 hflipped=no dwidth=? (linked to 00071118) Transluency=no
517 --> Returned from scaled bitmap processing (rem=50, vscale=7C)...
518 OP: Scaled bitmap 4x? 4bpp at 38,? hscale=7C fpix=0 data=00075E28 pitch 1 hflipped=no dwidth=? (linked to 00071118) Transluency=no
519 --> Returned from scaled bitmap processing (rem=30, vscale=7C)...
520 OP: Scaled bitmap 4x? 4bpp at 38,? hscale=7C fpix=0 data=00075E28 pitch 1 hflipped=no dwidth=? (linked to 00071118) Transluency=no
521 --> Returned from scaled bitmap processing (rem=10, vscale=7C)...
522 OP: Scaled bitmap 4x? 4bpp at 36,? hscale=7E fpix=0 data=000756A8 pitch 1 hflipped=no dwidth=? (linked to 00073058) Transluency=no
523 --> Returned from scaled bitmap processing (rem=00, vscale=7E)...
524 OP: Scaled bitmap 4x? 4bpp at 34,? hscale=80 fpix=0 data=000756C8 pitch 1 hflipped=no dwidth=? (linked to 00073078) Transluency=no
525 --> Returned from scaled bitmap processing (rem=00, vscale=80)...
526 OP: Scaled bitmap 4x? 4bpp at 36,? hscale=7E fpix=0 data=000756C8 pitch 1 hflipped=no dwidth=? (linked to 00073058) Transluency=no
527 --> Returned from scaled bitmap processing (rem=5E, vscale=7E)...
528 OP: Scaled bitmap 4x? 4bpp at 34,? hscale=80 fpix=0 data=000756E8 pitch 1 hflipped=no dwidth=? (linked to 00073078) Transluency=no
529 --> Returned from scaled bitmap processing (rem=60, vscale=80)...
530 OP: Scaled bitmap 4x? 4bpp at 36,? hscale=7E fpix=0 data=000756C8 pitch 1 hflipped=no dwidth=? (linked to 00073058) Transluency=no
531 --> Returned from scaled bitmap processing (rem=3E, vscale=7E)...
532 OP: Scaled bitmap 4x? 4bpp at 34,? hscale=80 fpix=0 data=000756E8 pitch 1 hflipped=no dwidth=? (linked to 00073078) Transluency=no
533 --> Returned from scaled bitmap processing (rem=40, vscale=80)...
534 OP: Scaled bitmap 4x? 4bpp at 36,? hscale=7E fpix=0 data=000756C8 pitch 1 hflipped=no dwidth=? (linked to 00073058) Transluency=no
535 --> Returned from scaled bitmap processing (rem=1E, vscale=7E)...
536 OP: Scaled bitmap 4x? 4bpp at 34,? hscale=80 fpix=0 data=000756E8 pitch 1 hflipped=no dwidth=? (linked to 00073078) Transluency=no
537 --> Returned from scaled bitmap processing (rem=20, vscale=80)...
539 //Here's another problem:
540 // [hsc: 20, vsc: 20, rem: 00]
541 // Since we're not checking for $E0 (but that's what we get from the above), we end
542 // up repeating this scanline unnecessarily... !!! FIX !!! [DONE, but... still not quite
543 // right. Either that, or the Accolade team that wrote Bubsy screwed up royal.]
544 //Also note: $E0 = 7.0 which IS a legal vscale value...
546 // if (remainder & 0x80) // I.e., it's negative
547 // if ((remainder & 0x80) || remainder == 0) // I.e., it's <= 0
548 // if ((remainder - 1) >= 0xE0) // I.e., it's <= 0
549 // if ((remainder >= 0xE1) || remainder == 0)// I.e., it's <= 0
550 // if ((remainder >= 0xE1 && remainder <= 0xFF) || remainder == 0)// I.e., it's <= 0
551 if (remainder <= 0x20) // I.e., it's <= 0
553 uint64 data = (p0 & 0xFFFFF80000000000LL) >> 40;
554 uint64 dwidth = (p1 & 0xFFC0000) >> 15;
556 // while (remainder & 0x80)
557 // while ((remainder & 0x80) || remainder == 0)
558 // while ((remainder - 1) >= 0xE0)
559 // while ((remainder >= 0xE1) || remainder == 0)
560 // while ((remainder >= 0xE1 && remainder <= 0xFF) || remainder == 0)
561 while (remainder <= 0x20)
571 p0 &= ~0xFFFFF80000FFC000LL; // Mask out old data...
572 p0 |= (uint64)height << 14;
574 OPStorePhrase(oldOPP, p0);
577 remainder -= 0x20; // 1.0f in [3.5] fixed point format
580 WriteLog("--> Finished writebacks...\n");*/
582 //WriteLog(" [%08X%08X -> ", (uint32)(p2>>32), (uint32)(p2&0xFFFFFFFF));
583 p2 &= ~0x0000000000FF0000LL;
584 p2 |= (uint64)remainder << 16;
585 //WriteLog("%08X%08X]\n", (uint32)(p2>>32), (uint32)(p2&0xFFFFFFFF));
586 OPStorePhrase(oldOPP+16, p2);
587 //remainder = (uint8)(p2 >> 16), vscale = (uint8)(p2 >> 8);
588 //WriteLog(" [after]: rem=%02X, vscale=%02X\n", remainder, vscale);
590 op_pointer = (p0 & 0x000007FFFF000000LL) >> 21;
593 case OBJECT_TYPE_GPU:
595 //WriteLog("OP: Asserting GPU IRQ #3...\n");
596 op_set_current_object(p0);
597 GPUSetIRQLine(3, ASSERT_LINE);
598 //Also, OP processing is suspended from this point until OBF (F00026) is written to...
601 //OPSuspendedByGPU = true;
602 //Dunno if the OP keeps processing from where it was interrupted, or if it just continues
603 //on the next scanline...
604 // --> It continues from where it was interrupted! !!! FIX !!!
607 case OBJECT_TYPE_BRANCH:
609 uint16 ypos = (p0 >> 3) & 0x7FF;
610 uint8 cc = (p0 >> 14) & 0x03;
611 uint32 link = (p0 >> 21) & 0x3FFFF8;
613 // if ((ypos!=507)&&(ypos!=25))
614 // WriteLog("\t%i%s%i link=0x%.8x\n",scanline,condition_to_str[cc],ypos>>1,link);
617 case CONDITION_EQUAL:
618 if (TOMReadWord(0xF00006, OP) == ypos || ypos == 0x7FF)
621 case CONDITION_LESS_THAN:
622 if (TOMReadWord(0xF00006, OP) < ypos)
625 case CONDITION_GREATER_THAN:
626 if (TOMReadWord(0xF00006, OP) > ypos)
629 case CONDITION_OP_FLAG_SET:
630 if (op_get_status_register() & 0x01)
633 case CONDITION_SECOND_HALF_LINE:
634 // This basically means branch if bit 10 of HC is set
635 WriteLog("OP: Unexpected CONDITION_SECOND_HALF_LINE in BRANCH object\nOP: shuting down\n");
640 WriteLog("OP: Unimplemented branch condition %i\n", cc);
644 case OBJECT_TYPE_STOP:
648 //WriteLog("OP: --> STOP\n");
649 // op_set_status_register(((p0>>3) & 0xFFFFFFFF));
650 //This seems more likely...
651 op_set_current_object(p0);
655 tom_set_pending_object_int();
656 if (tom_irq_enabled(IRQ_OPFLAG))// && jaguar_interrupt_handler_is_valid(64))
657 m68k_set_irq(7); // Cause an NMI to occur...
664 WriteLog("op: unknown object type %i\n", ((uint8)p0 & 0x07));
671 // Store fixed size bitmap in line buffer
673 void OPProcessFixedBitmap(uint64 p0, uint64 p1, bool render)
675 // Need to make sure that when writing that it stays within the line buffer...
676 // LBUF ($F01800 - $F01D9E) 360 x 32-bit RAM
677 uint8 depth = (p1 >> 12) & 0x07; // Color depth of image
678 int32 xpos = ((int16)((p1 << 4) & 0xFFFF)) >> 4;// Image xpos in LBUF
679 uint32 iwidth = (p1 >> 28) & 0x3FF; // Image width in *phrases*
680 uint32 data = (p0 >> 40) & 0xFFFFF8; // Pixel data address
681 //#ifdef OP_DEBUG_BMP
682 uint32 firstPix = (p1 >> 49) & 0x3F;
683 // "The LSB is significant only for scaled objects..." -JTRM
684 // "In 1 BPP mode, all five bits are significant. In 2 BPP mode, the top four are significant..."
687 // We can ignore the RELEASE (high order) bit for now--probably forever...!
688 // uint8 flags = (p1 >> 45) & 0x0F; // REFLECT, RMW, TRANS, RELEASE
689 //Optimize: break these out to their own BOOL values
690 uint8 flags = (p1 >> 45) & 0x07; // REFLECT (0), RMW (1), TRANS (2)
691 bool flagREFLECT = (flags & OPFLAG_REFLECT ? true : false),
692 flagRMW = (flags & OPFLAG_RMW ? true : false),
693 flagTRANS = (flags & OPFLAG_TRANS ? true : false);
694 // "For images with 1 to 4 bits/pixel the top 7 to 4 bits of the index
695 // provide the most significant bits of the palette address."
696 uint8 index = (p1 >> 37) & 0xFE; // CLUT index offset (upper pix, 1-4 bpp)
697 uint32 pitch = (p1 >> 15) & 0x07; // Phrase pitch
698 pitch <<= 3; // Optimization: Multiply pitch by 8
700 // int16 scanlineWidth = tom_getVideoModeWidth();
701 uint8 * tom_ram_8 = tom_get_ram_pointer();
702 uint8 * paletteRAM = &tom_ram_8[0x400];
703 // This is OK as long as it's used correctly: For 16-bit RAM to RAM direct copies--NOT
704 // for use when using endian-corrected data (i.e., any of the *_word_read functions!)
705 uint16 * paletteRAM16 = (uint16 *)paletteRAM;
707 // WriteLog("bitmap %ix? %ibpp at %i,? firstpix=? data=0x%.8x pitch %i hflipped=%s dwidth=? (linked to ?) RMW=%s Tranparent=%s\n",
708 // iwidth, op_bitmap_bit_depth[bitdepth], xpos, ptr, pitch, (flags&OPFLAG_REFLECT ? "yes" : "no"), (flags&OPFLAG_RMW ? "yes" : "no"), (flags&OPFLAG_TRANS ? "yes" : "no"));
710 // Is it OK to have a 0 for the data width??? (i.e., undocumented?)
711 // Seems to be... Seems that dwidth *can* be zero (i.e., reuse same line) as well.
712 // Pitch == 0 is OK too...
713 // if (!render || op_pointer == 0 || ptr == 0 || pitch == 0)
714 //I'm not convinced that we need to concern ourselves with data & op_pointer here either!
715 if (!render || iwidth == 0)
718 //#define OP_DEBUG_BMP
719 //#ifdef OP_DEBUG_BMP
720 // 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",
721 // 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"));
724 // int32 leftMargin = xpos, rightMargin = (xpos + (phraseWidthToPixels[depth] * iwidth)) - 1;
725 int32 startPos = xpos, endPos = xpos +
726 (!flagREFLECT ? (phraseWidthToPixels[depth] * iwidth) - 1
727 : -((phraseWidthToPixels[depth] * iwidth) + 1));
728 uint32 clippedWidth = 0, phraseClippedWidth = 0, dataClippedWidth = 0;//, phrasePixel = 0;
729 bool in24BPPMode = (((GET16(tom_ram_8, 0x0028) >> 1) & 0x03) == 1 ? true : false); // VMODE
730 // Not sure if this is Jaguar Two only location or what...
731 // From the docs, it is... If we want to limit here we should think of something else.
732 // int32 limit = GET16(tom_ram_8, 0x0008); // LIMIT
734 int32 lbufWidth = (!in24BPPMode ? limit - 1 : (limit / 2) - 1); // Zero based limit...
736 // If the image is completely to the left or right of the line buffer, then bail.
737 //If in REFLECT mode, then these values are swapped! !!! FIX !!! [DONE]
738 //There are four possibilities:
739 // 1. image sits on left edge and no REFLECT; starts out of bounds but ends in bounds.
740 // 2. image sits on left edge and REFLECT; starts in bounds but ends out of bounds.
741 // 3. image sits on right edge and REFLECT; starts out of bounds but ends in bounds.
742 // 4. image sits on right edge and no REFLECT; starts in bounds but ends out of bounds.
743 //Numbers 2 & 4 can be caught by checking the LBUF clip while in the inner loop,
744 // numbers 1 & 3 are of concern.
745 // This *indirectly* handles only cases 2 & 4! And is WRONG is REFLECT is set...!
746 // if (rightMargin < 0 || leftMargin > lbufWidth)
748 // It might be easier to swap these (if REFLECTed) and just use XPOS down below...
749 // That way, you could simply set XPOS to leftMargin if !REFLECT and to rightMargin otherwise.
750 // Still have to be careful with the DATA and IWIDTH values though...
752 // if ((!flagREFLECT && (rightMargin < 0 || leftMargin > lbufWidth))
753 // || (flagREFLECT && (leftMargin < 0 || rightMargin > lbufWidth)))
755 if ((!flagREFLECT && (endPos < 0 || startPos > lbufWidth))
756 || (flagREFLECT && (startPos < 0 || endPos > lbufWidth)))
759 // Otherwise, find the clip limits and clip the phrase as well...
760 // NOTE: I'm fudging here by letting the actual blit overstep the bounds of the
761 // line buffer, but it shouldn't matter since there are two unused line
762 // buffers below and nothing above and I'll at most write 8 bytes outside
763 // the line buffer... I could use a fractional clip begin/end value, but
764 // this makes the blit a *lot* more hairy. I might fix this in the future
765 // if it becomes necessary. (JLH)
766 // Probably wouldn't be *that* hairy. Just use a delta that tells the inner loop
767 // which pixel in the phrase is being written, and quit when either end of phrases
768 // is reached or line buffer extents are surpassed.
770 //This stuff is probably wrong as well... !!! FIX !!!
771 //The strange thing is that it seems to work, but that's no guarantee that it's bulletproof!
772 //Yup. Seems that JagMania doesn't work correctly with this...
773 //Dunno if this is the problem, but Atari Karts is showing *some* of the road now...
778 clippedWidth = 0 - leftMargin,
779 phraseClippedWidth = clippedWidth / phraseWidthToPixels[depth],
780 leftMargin = 0 - (clippedWidth % phraseWidthToPixels[depth]);
783 if (rightMargin > lbufWidth)
784 clippedWidth = rightMargin - lbufWidth,
785 phraseClippedWidth = clippedWidth / phraseWidthToPixels[depth];//,
786 // rightMargin = lbufWidth + (clippedWidth % phraseWidthToPixels[depth]);
787 // rightMargin = lbufWidth;
790 WriteLog("OP: We're about to encounter a divide by zero error!\n");
791 // NOTE: We're just using endPos to figure out how much, if any, to clip by.
792 // ALSO: There may be another case where we start out of bounds and end out of bounds...!
794 if (startPos < 0) // Case #1: Begin out, end in, L to R
795 clippedWidth = 0 - startPos,
796 dataClippedWidth = phraseClippedWidth = clippedWidth / phraseWidthToPixels[depth],
797 startPos = 0 - (clippedWidth % phraseWidthToPixels[depth]);
799 if (endPos < 0) // Case #2: Begin in, end out, R to L
800 clippedWidth = 0 - endPos,
801 phraseClippedWidth = clippedWidth / phraseWidthToPixels[depth];
803 if (endPos > lbufWidth) // Case #3: Begin in, end out, L to R
804 clippedWidth = endPos - lbufWidth,
805 phraseClippedWidth = clippedWidth / phraseWidthToPixels[depth];
807 if (startPos > lbufWidth) // Case #4: Begin out, end in, R to L
808 clippedWidth = startPos - lbufWidth,
809 dataClippedWidth = phraseClippedWidth = clippedWidth / phraseWidthToPixels[depth],
810 startPos = lbufWidth + (clippedWidth % phraseWidthToPixels[depth]);
812 // If the image is sitting on the line buffer left or right edge, we need to compensate
813 // by decreasing the image phrase width accordingly.
814 iwidth -= phraseClippedWidth;
816 // Also, if we're clipping the phrase we need to make sure we're in the correct part of
818 // data += phraseClippedWidth * (pitch << 3);
819 data += dataClippedWidth * pitch;
821 // NOTE: When the bitmap is in REFLECT mode, the XPOS marks the *right* side of the
822 // bitmap! This makes clipping & etc. MUCH, much easier...!
823 // uint32 lbufAddress = 0x1800 + (!in24BPPMode ? leftMargin * 2 : leftMargin * 4);
824 //Why does this work right when multiplying startPos by 2 (instead of 4) for 24 BPP mode?
825 //Is this a bug in the OP?
826 uint32 lbufAddress = 0x1800 + (!in24BPPMode ? startPos * 2 : startPos * 2);
827 uint8 * currentLineBuffer = &tom_ram_8[lbufAddress];
831 // Hmm. We check above for 24 BPP mode, but don't do anything about it below...
832 // If we *were* in 24 BPP mode, how would you convert CRY to RGB24? Seems to me
833 // that if you're in CRY mode then you wouldn't be able to use 24 BPP bitmaps
835 // This seems to be the case (at least according to the Midsummer docs)...!
837 if (depth == 0) // 1 BPP
839 // The LSB of flags is OPFLAG_REFLECT, so sign extend it and or 2 into it.
840 int32 lbufDelta = ((int8)((flags << 7) & 0xFF) >> 5) | 0x02;
842 // Fetch 1st phrase...
843 uint64 pixels = ((uint64)JaguarReadLong(data, OP) << 32) | JaguarReadLong(data + 4, OP);
844 //Note that firstPix should only be honored *if* we start with the 1st phrase of the bitmap
845 //i.e., we didn't clip on the margin... !!! FIX !!!
846 pixels <<= firstPix; // Skip first N pixels (N=firstPix)...
847 int i = firstPix; // Start counter at right spot...
853 uint8 bit = pixels >> 63;
854 if (flagTRANS && bit == 0)
859 //Optimize: Set palleteRAM16 to beginning of palette RAM + index*2 and use only [bit] as index...
860 //Won't optimize RMW case though...
861 // This is the *only* correct use of endian-dependent code
862 // (i.e., mem-to-mem direct copying)!
863 *(uint16 *)currentLineBuffer = paletteRAM16[index | bit];
866 BLEND_CR(*currentLineBuffer, paletteRAM[(index | bit) << 1]),
867 *(currentLineBuffer + 1) =
868 BLEND_Y(*(currentLineBuffer + 1), paletteRAM[((index | bit) << 1) + 1]);
871 currentLineBuffer += lbufDelta;
875 // Fetch next phrase...
877 pixels = ((uint64)JaguarReadLong(data, OP) << 32) | JaguarReadLong(data + 4, OP);
880 else if (depth == 1) // 2 BPP
883 WriteLog("OP: Fixed bitmap @ 2 BPP requesting FIRSTPIX! (fp=%u)\n", firstPix);
884 index &= 0xFC; // Top six bits form CLUT index
885 // The LSB is OPFLAG_REFLECT, so sign extend it and or 2 into it.
886 int32 lbufDelta = ((int8)((flags << 7) & 0xFF) >> 5) | 0x02;
891 uint64 pixels = ((uint64)JaguarReadLong(data, OP) << 32) | JaguarReadLong(data + 4, OP);
894 for(int i=0; i<32; i++)
896 uint8 bits = pixels >> 62;
897 // Seems to me that both of these are in the same endian, so we could cast it as
898 // uint16 * and do straight across copies (what about 24 bpp? Treat it differently...)
899 // This only works for the palettized modes (1 - 8 BPP), since we actually have to
900 // copy data from memory in 16 BPP mode (or does it? Isn't this the same as the CLUT case?)
901 // No, it isn't because we read the memory in an endian safe way--this *won't* work...
902 if (flagTRANS && bits == 0)
907 *(uint16 *)currentLineBuffer = paletteRAM16[index | bits];
910 BLEND_CR(*currentLineBuffer, paletteRAM[(index | bits) << 1]),
911 *(currentLineBuffer + 1) =
912 BLEND_Y(*(currentLineBuffer + 1), paletteRAM[((index | bits) << 1) + 1]);
915 currentLineBuffer += lbufDelta;
920 else if (depth == 2) // 4 BPP
923 WriteLog("OP: Fixed bitmap @ 4 BPP requesting FIRSTPIX! (fp=%u)\n", firstPix);
924 index &= 0xF0; // Top four bits form CLUT index
925 // The LSB is OPFLAG_REFLECT, so sign extend it and or 2 into it.
926 int32 lbufDelta = ((int8)((flags << 7) & 0xFF) >> 5) | 0x02;
931 uint64 pixels = ((uint64)JaguarReadLong(data, OP) << 32) | JaguarReadLong(data + 4, OP);
934 for(int i=0; i<16; i++)
936 uint8 bits = pixels >> 60;
937 // Seems to me that both of these are in the same endian, so we could cast it as
938 // uint16 * and do straight across copies (what about 24 bpp? Treat it differently...)
939 // This only works for the palettized modes (1 - 8 BPP), since we actually have to
940 // copy data from memory in 16 BPP mode (or does it? Isn't this the same as the CLUT case?)
941 // No, it isn't because we read the memory in an endian safe way--this *won't* work...
942 if (flagTRANS && bits == 0)
947 *(uint16 *)currentLineBuffer = paletteRAM16[index | bits];
950 BLEND_CR(*currentLineBuffer, paletteRAM[(index | bits) << 1]),
951 *(currentLineBuffer + 1) =
952 BLEND_Y(*(currentLineBuffer + 1), paletteRAM[((index | bits) << 1) + 1]);
955 currentLineBuffer += lbufDelta;
960 else if (depth == 3) // 8 BPP
962 // The LSB is OPFLAG_REFLECT, so sign extend it and or 2 into it.
963 int32 lbufDelta = ((int8)((flags << 7) & 0xFF) >> 5) | 0x02;
965 // Fetch 1st phrase...
966 uint64 pixels = ((uint64)JaguarReadLong(data, OP) << 32) | JaguarReadLong(data + 4, OP);
967 //Note that firstPix should only be honored *if* we start with the 1st phrase of the bitmap
968 //i.e., we didn't clip on the margin... !!! FIX !!!
969 firstPix &= 0x30; // Only top two bits are valid for 8 BPP
970 pixels <<= firstPix; // Skip first N pixels (N=firstPix)...
971 int i = firstPix >> 3; // Start counter at right spot...
977 uint8 bits = pixels >> 56;
978 // Seems to me that both of these are in the same endian, so we could cast it as
979 // uint16 * and do straight across copies (what about 24 bpp? Treat it differently...)
980 // This only works for the palettized modes (1 - 8 BPP), since we actually have to
981 // copy data from memory in 16 BPP mode (or does it? Isn't this the same as the CLUT case?)
982 // No, it isn't because we read the memory in an endian safe way--this *won't* work...
983 if (flagTRANS && bits == 0)
988 *(uint16 *)currentLineBuffer = paletteRAM16[bits];
991 BLEND_CR(*currentLineBuffer, paletteRAM[bits << 1]),
992 *(currentLineBuffer + 1) =
993 BLEND_Y(*(currentLineBuffer + 1), paletteRAM[(bits << 1) + 1]);
996 currentLineBuffer += lbufDelta;
1000 // Fetch next phrase...
1002 pixels = ((uint64)JaguarReadLong(data, OP) << 32) | JaguarReadLong(data + 4, OP);
1005 else if (depth == 4) // 16 BPP
1008 WriteLog("OP: Fixed bitmap @ 16 BPP requesting FIRSTPIX! (fp=%u)\n", firstPix);
1009 // The LSB is OPFLAG_REFLECT, so sign extend it and or 2 into it.
1010 int32 lbufDelta = ((int8)((flags << 7) & 0xFF) >> 5) | 0x02;
1015 uint64 pixels = ((uint64)JaguarReadLong(data, OP) << 32) | JaguarReadLong(data + 4, OP);
1018 for(int i=0; i<4; i++)
1020 uint8 bitsHi = pixels >> 56, bitsLo = pixels >> 48;
1021 // Seems to me that both of these are in the same endian, so we could cast it as
1022 // uint16 * and do straight across copies (what about 24 bpp? Treat it differently...)
1023 // This only works for the palettized modes (1 - 8 BPP), since we actually have to
1024 // copy data from memory in 16 BPP mode (or does it? Isn't this the same as the CLUT case?)
1025 // No, it isn't because we read the memory in an endian safe way--it *won't* work...
1026 if (flagTRANS && (bitsLo | bitsHi) == 0)
1031 *currentLineBuffer = bitsHi,
1032 *(currentLineBuffer + 1) = bitsLo;
1034 *currentLineBuffer =
1035 BLEND_CR(*currentLineBuffer, bitsHi),
1036 *(currentLineBuffer + 1) =
1037 BLEND_Y(*(currentLineBuffer + 1), bitsLo);
1040 currentLineBuffer += lbufDelta;
1045 else if (depth == 5) // 24 BPP
1047 //Looks like Iron Soldier is the only game that uses 24BPP mode...
1048 //There *might* be others...
1049 //WriteLog("OP: Writing 24 BPP bitmap!\n");
1051 WriteLog("OP: Fixed bitmap @ 24 BPP requesting FIRSTPIX! (fp=%u)\n", firstPix);
1052 // Not sure, but I think RMW only works with 16 BPP and below, and only in CRY mode...
1053 // The LSB of flags is OPFLAG_REFLECT, so sign extend it and OR 4 into it.
1054 int32 lbufDelta = ((int8)((flags << 7) & 0xFF) >> 4) | 0x04;
1059 uint64 pixels = ((uint64)JaguarReadLong(data, OP) << 32) | JaguarReadLong(data + 4, OP);
1062 for(int i=0; i<2; i++)
1064 // We don't use a 32-bit var here because of endian issues...!
1065 uint8 bits3 = pixels >> 56, bits2 = pixels >> 48,
1066 bits1 = pixels >> 40, bits0 = pixels >> 32;
1068 if (flagTRANS && (bits3 | bits2 | bits1 | bits0) == 0)
1071 *currentLineBuffer = bits3,
1072 *(currentLineBuffer + 1) = bits2,
1073 *(currentLineBuffer + 2) = bits1,
1074 *(currentLineBuffer + 3) = bits0;
1076 currentLineBuffer += lbufDelta;
1084 // Store scaled bitmap in line buffer
1086 void OPProcessScaledBitmap(uint64 p0, uint64 p1, uint64 p2, bool render)
1088 // Need to make sure that when writing that it stays within the line buffer...
1089 // LBUF ($F01800 - $F01D9E) 360 x 32-bit RAM
1090 uint8 depth = (p1 >> 12) & 0x07; // Color depth of image
1091 int32 xpos = ((int16)((p1 << 4) & 0xFFFF)) >> 4;// Image xpos in LBUF
1092 uint32 iwidth = (p1 >> 28) & 0x3FF; // Image width in *phrases*
1093 uint32 data = (p0 >> 40) & 0xFFFFF8; // Pixel data address
1094 //#ifdef OP_DEBUG_BMP
1095 // Prolly should use this... Though not sure exactly how.
1096 //Use the upper bits as an offset into the phrase depending on the BPP. That's how!
1097 uint32 firstPix = (p1 >> 49) & 0x3F;
1098 //This is WEIRD! I'm sure I saw Atari Karts request 8 BPP FIRSTPIX! What happened???
1100 WriteLog("OP: FIRSTPIX != 0! (Scaled BM)\n");
1102 // We can ignore the RELEASE (high order) bit for now--probably forever...!
1103 // uint8 flags = (p1 >> 45) & 0x0F; // REFLECT, RMW, TRANS, RELEASE
1104 //Optimize: break these out to their own BOOL values [DONE]
1105 uint8 flags = (p1 >> 45) & 0x07; // REFLECT (0), RMW (1), TRANS (2)
1106 bool flagREFLECT = (flags & OPFLAG_REFLECT ? true : false),
1107 flagRMW = (flags & OPFLAG_RMW ? true : false),
1108 flagTRANS = (flags & OPFLAG_TRANS ? true : false);
1109 uint8 index = (p1 >> 37) & 0xFE; // CLUT index offset (upper pix, 1-4 bpp)
1110 uint32 pitch = (p1 >> 15) & 0x07; // Phrase pitch
1112 uint8 * tom_ram_8 = tom_get_ram_pointer();
1113 uint8 * paletteRAM = &tom_ram_8[0x400];
1114 // This is OK as long as it's used correctly: For 16-bit RAM to RAM direct copies--NOT
1115 // for use when using endian-corrected data (i.e., any of the *ReadWord functions!)
1116 uint16 * paletteRAM16 = (uint16 *)paletteRAM;
1118 uint8 hscale = p2 & 0xFF;
1119 // uint8 horizontalRemainder = hscale; // Not sure if it starts full, but seems reasonable [It's not!]
1120 uint8 horizontalRemainder = 0; // Let's try zero! Seems to work! Yay!
1121 int32 scaledWidthInPixels = (iwidth * phraseWidthToPixels[depth] * hscale) >> 5;
1122 uint32 scaledPhrasePixels = (phraseWidthToPixels[depth] * hscale) >> 5;
1124 // WriteLog("bitmap %ix? %ibpp at %i,? firstpix=? data=0x%.8x pitch %i hflipped=%s dwidth=? (linked to ?) RMW=%s Tranparent=%s\n",
1125 // iwidth, op_bitmap_bit_depth[bitdepth], xpos, ptr, pitch, (flags&OPFLAG_REFLECT ? "yes" : "no"), (flags&OPFLAG_RMW ? "yes" : "no"), (flags&OPFLAG_TRANS ? "yes" : "no"));
1127 // Looks like an hscale of zero means don't draw!
1128 if (!render || iwidth == 0 || hscale == 0)
1131 /*extern int start_logging;
1133 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",
1134 iwidth, op_bitmap_bit_depth[depth], xpos, hscale, firstPix, data, pitch, (flagREFLECT ? "yes" : "no"), op_pointer, (flagRMW ? "yes" : "no"));*/
1135 //#define OP_DEBUG_BMP
1136 //#ifdef OP_DEBUG_BMP
1137 // 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",
1138 // 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"));
1141 int32 startPos = xpos, endPos = xpos +
1142 (!flagREFLECT ? scaledWidthInPixels - 1 : -(scaledWidthInPixels + 1));
1143 uint32 clippedWidth = 0, phraseClippedWidth = 0, dataClippedWidth = 0;
1144 bool in24BPPMode = (((GET16(tom_ram_8, 0x0028) >> 1) & 0x03) == 1 ? true : false); // VMODE
1145 // Not sure if this is Jaguar Two only location or what...
1146 // From the docs, it is... If we want to limit here we should think of something else.
1147 // int32 limit = GET16(tom_ram_8, 0x0008); // LIMIT
1149 int32 lbufWidth = (!in24BPPMode ? limit - 1 : (limit / 2) - 1); // Zero based limit...
1151 // If the image is completely to the left or right of the line buffer, then bail.
1152 //If in REFLECT mode, then these values are swapped! !!! FIX !!! [DONE]
1153 //There are four possibilities:
1154 // 1. image sits on left edge and no REFLECT; starts out of bounds but ends in bounds.
1155 // 2. image sits on left edge and REFLECT; starts in bounds but ends out of bounds.
1156 // 3. image sits on right edge and REFLECT; starts out of bounds but ends in bounds.
1157 // 4. image sits on right edge and no REFLECT; starts in bounds but ends out of bounds.
1158 //Numbers 2 & 4 can be caught by checking the LBUF clip while in the inner loop,
1159 // numbers 1 & 3 are of concern.
1160 // This *indirectly* handles only cases 2 & 4! And is WRONG if REFLECT is set...!
1161 // if (rightMargin < 0 || leftMargin > lbufWidth)
1163 // It might be easier to swap these (if REFLECTed) and just use XPOS down below...
1164 // That way, you could simply set XPOS to leftMargin if !REFLECT and to rightMargin otherwise.
1165 // Still have to be careful with the DATA and IWIDTH values though...
1167 if ((!flagREFLECT && (endPos < 0 || startPos > lbufWidth))
1168 || (flagREFLECT && (startPos < 0 || endPos > lbufWidth)))
1171 // Otherwise, find the clip limits and clip the phrase as well...
1172 // NOTE: I'm fudging here by letting the actual blit overstep the bounds of the
1173 // line buffer, but it shouldn't matter since there are two unused line
1174 // buffers below and nothing above and I'll at most write 40 bytes outside
1175 // the line buffer... I could use a fractional clip begin/end value, but
1176 // this makes the blit a *lot* more hairy. I might fix this in the future
1177 // if it becomes necessary. (JLH)
1178 // Probably wouldn't be *that* hairy. Just use a delta that tells the inner loop
1179 // which pixel in the phrase is being written, and quit when either end of phrases
1180 // is reached or line buffer extents are surpassed.
1182 //This stuff is probably wrong as well... !!! FIX !!!
1183 //The strange thing is that it seems to work, but that's no guarantee that it's bulletproof!
1184 //Yup. Seems that JagMania doesn't work correctly with this...
1185 //Dunno if this is the problem, but Atari Karts is showing *some* of the road now...
1186 //Actually, it is! Or, it was. It doesn't seem to be clipping here, so the problem lies
1187 //elsewhere! Hmm. Putting the scaling code into the 1/2/8 BPP cases seems to draw the ground
1188 // a bit more accurately... Strange!
1189 //It's probably a case of the REFLECT flag being set and the background being written
1190 //from the right side of the screen...
1191 //But no, it isn't... At least if the diagnostics are telling the truth!
1193 // NOTE: We're just using endPos to figure out how much, if any, to clip by.
1194 // ALSO: There may be another case where we start out of bounds and end out of bounds...!
1197 //There's a problem here with scaledPhrasePixels in that it can be forced to zero when
1198 //the scaling factor is small. So fix it already! !!! FIX !!!
1199 /*if (scaledPhrasePixels == 0)
1201 WriteLog("OP: [Scaled] We're about to encounter a divide by zero error!\n");
1202 DumpScaledObject(p0, p1, p2);
1204 //NOTE: I'm almost 100% sure that this is wrong... And it is! :-p
1206 //Try a simple example...
1207 // Let's say we have a 8 BPP scanline with an hscale of $80 (4). Our xpos is -10,
1208 // non-flipped. Pixels in the bitmap are XYZXYZXYZXYZXYZ.
1209 // Scaled up, they would be XXXXYYYYZZZZXXXXYYYYZZZZXXXXYYYYZZZZ...
1211 // Normally, we would expect this in the line buffer:
1212 // ZZXXXXYYYYZZZZXXXXYYYYZZZZ...
1214 // But instead we're getting:
1215 // XXXXYYYYZZZZXXXXYYYYZZZZ...
1217 // or are we??? It would seem so, simply by virtue of the fact that we're NOT starting
1218 // on negative boundary--or are we? Hmm...
1219 // cw = 10, dcw = pcw = 10 / ([8 * 4 = 32] 32) = 0, sp = -10
1221 // Let's try a real world example:
1223 //OP: Scaled bitmap (70, 8 BPP, spp=28) sp (-400) < 0... [new sp=-8, cw=400, dcw=pcw=14]
1224 //OP: Scaled bitmap (6F, 8 BPP, spp=27) sp (-395) < 0... [new sp=-17, cw=395, dcw=pcw=14]
1226 // Really, spp is 27.75 in the second case...
1227 // So... If we do 395 / 27.75, we get 14. Ok so far... If we scale that against the
1228 // start position (14 * 27.75), we get -6.5... NOT -17!
1230 //Now it seems we're working OK, at least for the first case...
1231 uint32 scaledPhrasePixelsUS = phraseWidthToPixels[depth] * hscale;
1233 if (startPos < 0) // Case #1: Begin out, end in, L to R
1235 extern int start_logging;
1237 WriteLog("OP: Scaled bitmap (%02X, %u BPP, spp=%u) start pos (%i) < 0...", hscale, op_bitmap_bit_depth[depth], scaledPhrasePixels, startPos);
1238 // clippedWidth = 0 - startPos,
1239 clippedWidth = (0 - startPos) << 5,
1240 // dataClippedWidth = phraseClippedWidth = clippedWidth / scaledPhrasePixels,
1241 dataClippedWidth = phraseClippedWidth = (clippedWidth / scaledPhrasePixelsUS) >> 5,
1242 // startPos = 0 - (clippedWidth % scaledPhrasePixels);
1243 startPos += (dataClippedWidth * scaledPhrasePixelsUS) >> 5;
1245 WriteLog(" [new sp=%i, cw=%i, dcw=pcw=%i]\n", startPos, clippedWidth, dataClippedWidth);
1248 if (endPos < 0) // Case #2: Begin in, end out, R to L
1249 clippedWidth = 0 - endPos,
1250 phraseClippedWidth = clippedWidth / scaledPhrasePixels;
1252 if (endPos > lbufWidth) // Case #3: Begin in, end out, L to R
1253 clippedWidth = endPos - lbufWidth,
1254 phraseClippedWidth = clippedWidth / scaledPhrasePixels;
1256 if (startPos > lbufWidth) // Case #4: Begin out, end in, R to L
1257 clippedWidth = startPos - lbufWidth,
1258 dataClippedWidth = phraseClippedWidth = clippedWidth / scaledPhrasePixels,
1259 startPos = lbufWidth + (clippedWidth % scaledPhrasePixels);
1261 extern int op_start_log;
1262 if (op_start_log && clippedWidth != 0)
1263 WriteLog("OP: Clipped line. SP=%i, EP=%i, clip=%u, iwidth=%u, hscale=%02X\n", startPos, endPos, clippedWidth, iwidth, hscale);
1264 if (op_start_log && startPos == 13)
1266 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);
1267 DumpScaledObject(p0, p1, p2);
1269 // If the image is sitting on the line buffer left or right edge, we need to compensate
1270 // by decreasing the image phrase width accordingly.
1271 iwidth -= phraseClippedWidth;
1273 // Also, if we're clipping the phrase we need to make sure we're in the correct part of
1275 // data += phraseClippedWidth * (pitch << 3);
1276 data += dataClippedWidth * (pitch << 3);
1278 // NOTE: When the bitmap is in REFLECT mode, the XPOS marks the *right* side of the
1279 // bitmap! This makes clipping & etc. MUCH, much easier...!
1280 // uint32 lbufAddress = 0x1800 + (!in24BPPMode ? leftMargin * 2 : leftMargin * 4);
1281 // uint32 lbufAddress = 0x1800 + (!in24BPPMode ? startPos * 2 : startPos * 4);
1282 uint32 lbufAddress = 0x1800 + startPos * 2;
1283 uint8 * currentLineBuffer = &tom_ram_8[lbufAddress];
1284 uint8 * lineBufferLowerLimit = &tom_ram_8[0x1800],
1285 * lineBufferUpperLimit = &tom_ram_8[0x1800 + 719];
1289 // Hmm. We check above for 24 BPP mode, but don't do anything about it below...
1290 // If we *were* in 24 BPP mode, how would you convert CRY to RGB24? Seems to me
1291 // that if you're in CRY mode then you wouldn't be able to use 24 BPP bitmaps
1293 // This seems to be the case (at least according to the Midsummer docs)...!
1295 if (depth == 0) // 1 BPP
1298 WriteLog("OP: Scaled bitmap @ 1 BPP requesting FIRSTPIX!\n");
1299 // The LSB of flags is OPFLAG_REFLECT, so sign extend it and or 2 into it.
1300 int32 lbufDelta = ((int8)((flags << 7) & 0xFF) >> 5) | 0x02;
1303 uint64 pixels = ((uint64)JaguarReadLong(data, OP) << 32) | JaguarReadLong(data + 4, OP);
1305 while ((int32)iwidth > 0)
1307 uint8 bits = pixels >> 63;
1309 if (flagTRANS && bits == 0)
1314 // This is the *only* correct use of endian-dependent code
1315 // (i.e., mem-to-mem direct copying)!
1316 *(uint16 *)currentLineBuffer = paletteRAM16[index | bits];
1318 *currentLineBuffer =
1319 BLEND_CR(*currentLineBuffer, paletteRAM[(index | bits) << 1]),
1320 *(currentLineBuffer + 1) =
1321 BLEND_Y(*(currentLineBuffer + 1), paletteRAM[((index | bits) << 1) + 1]);
1324 currentLineBuffer += lbufDelta;
1326 horizontalRemainder -= 0x20; // Subtract 1.0f in [3.5] fixed point format
1327 while (horizontalRemainder & 0x80)
1329 horizontalRemainder += hscale;
1336 int phrasesToSkip = pixCount / 64, pixelShift = pixCount % 64;
1338 data += (pitch << 3) * phrasesToSkip;
1339 pixels = ((uint64)JaguarReadLong(data, OP) << 32) | JaguarReadLong(data + 4, OP);
1340 pixels <<= 1 * pixelShift;
1341 iwidth -= phrasesToSkip;
1342 pixCount = pixelShift;
1346 else if (depth == 1) // 2 BPP
1349 WriteLog("OP: Scaled bitmap @ 2 BPP requesting FIRSTPIX!\n");
1350 index &= 0xFC; // Top six bits form CLUT index
1351 // The LSB is OPFLAG_REFLECT, so sign extend it and or 2 into it.
1352 int32 lbufDelta = ((int8)((flags << 7) & 0xFF) >> 5) | 0x02;
1355 uint64 pixels = ((uint64)JaguarReadLong(data, OP) << 32) | JaguarReadLong(data + 4, OP);
1357 while ((int32)iwidth > 0)
1359 uint8 bits = pixels >> 62;
1361 if (flagTRANS && bits == 0)
1366 // This is the *only* correct use of endian-dependent code
1367 // (i.e., mem-to-mem direct copying)!
1368 *(uint16 *)currentLineBuffer = paletteRAM16[index | bits];
1370 *currentLineBuffer =
1371 BLEND_CR(*currentLineBuffer, paletteRAM[(index | bits) << 1]),
1372 *(currentLineBuffer + 1) =
1373 BLEND_Y(*(currentLineBuffer + 1), paletteRAM[((index | bits) << 1) + 1]);
1376 currentLineBuffer += lbufDelta;
1378 horizontalRemainder -= 0x20; // Subtract 1.0f in [3.5] fixed point format
1379 while (horizontalRemainder & 0x80)
1381 horizontalRemainder += hscale;
1388 int phrasesToSkip = pixCount / 32, pixelShift = pixCount % 32;
1390 data += (pitch << 3) * phrasesToSkip;
1391 pixels = ((uint64)JaguarReadLong(data, OP) << 32) | JaguarReadLong(data + 4, OP);
1392 pixels <<= 2 * pixelShift;
1393 iwidth -= phrasesToSkip;
1394 pixCount = pixelShift;
1398 else if (depth == 2) // 4 BPP
1401 WriteLog("OP: Scaled bitmap @ 4 BPP requesting FIRSTPIX!\n");
1402 index &= 0xF0; // Top four bits form CLUT index
1403 // The LSB is OPFLAG_REFLECT, so sign extend it and or 2 into it.
1404 int32 lbufDelta = ((int8)((flags << 7) & 0xFF) >> 5) | 0x02;
1407 uint64 pixels = ((uint64)JaguarReadLong(data, OP) << 32) | JaguarReadLong(data + 4, OP);
1409 while ((int32)iwidth > 0)
1411 uint8 bits = pixels >> 60;
1413 if (flagTRANS && bits == 0)
1418 // This is the *only* correct use of endian-dependent code
1419 // (i.e., mem-to-mem direct copying)!
1420 *(uint16 *)currentLineBuffer = paletteRAM16[index | bits];
1422 *currentLineBuffer =
1423 BLEND_CR(*currentLineBuffer, paletteRAM[(index | bits) << 1]),
1424 *(currentLineBuffer + 1) =
1425 BLEND_Y(*(currentLineBuffer + 1), paletteRAM[((index | bits) << 1) + 1]);
1428 currentLineBuffer += lbufDelta;
1430 horizontalRemainder -= 0x20; // Subtract 1.0f in [3.5] fixed point format
1431 while (horizontalRemainder & 0x80)
1433 horizontalRemainder += hscale;
1440 int phrasesToSkip = pixCount / 16, pixelShift = pixCount % 16;
1442 data += (pitch << 3) * phrasesToSkip;
1443 pixels = ((uint64)JaguarReadLong(data, OP) << 32) | JaguarReadLong(data + 4, OP);
1444 pixels <<= 4 * pixelShift;
1445 iwidth -= phrasesToSkip;
1446 pixCount = pixelShift;
1450 else if (depth == 3) // 8 BPP
1453 WriteLog("OP: Scaled bitmap @ 8 BPP requesting FIRSTPIX! (fp=%u)\n", firstPix);
1454 // The LSB is OPFLAG_REFLECT, so sign extend it and or 2 into it.
1455 int32 lbufDelta = ((int8)((flags << 7) & 0xFF) >> 5) | 0x02;
1458 uint64 pixels = ((uint64)JaguarReadLong(data, OP) << 32) | JaguarReadLong(data + 4, OP);
1460 while ((int32)iwidth > 0)
1462 uint8 bits = pixels >> 56;
1464 if (flagTRANS && bits == 0)
1469 // This is the *only* correct use of endian-dependent code
1470 // (i.e., mem-to-mem direct copying)!
1471 *(uint16 *)currentLineBuffer = paletteRAM16[bits];
1473 if (currentLineBuffer >= lineBufferLowerLimit && currentLineBuffer <= lineBufferUpperLimit)
1474 *(uint16 *)currentLineBuffer = paletteRAM16[bits];
1477 *currentLineBuffer =
1478 BLEND_CR(*currentLineBuffer, paletteRAM[bits << 1]),
1479 *(currentLineBuffer + 1) =
1480 BLEND_Y(*(currentLineBuffer + 1), paletteRAM[(bits << 1) + 1]);
1483 currentLineBuffer += lbufDelta;
1485 horizontalRemainder -= 0x20; // Subtract 1.0f in [3.5] fixed point format
1486 while (horizontalRemainder & 0x80)
1488 horizontalRemainder += hscale;
1495 int phrasesToSkip = pixCount / 8, pixelShift = pixCount % 8;
1497 data += (pitch << 3) * phrasesToSkip;
1498 pixels = ((uint64)JaguarReadLong(data, OP) << 32) | JaguarReadLong(data + 4, OP);
1499 pixels <<= 8 * pixelShift;
1500 iwidth -= phrasesToSkip;
1501 pixCount = pixelShift;
1505 else if (depth == 4) // 16 BPP
1508 WriteLog("OP: Scaled bitmap @ 16 BPP requesting FIRSTPIX!\n");
1509 // The LSB is OPFLAG_REFLECT, so sign extend it and OR 2 into it.
1510 int32 lbufDelta = ((int8)((flags << 7) & 0xFF) >> 5) | 0x02;
1513 uint64 pixels = ((uint64)JaguarReadLong(data, OP) << 32) | JaguarReadLong(data + 4, OP);
1515 while ((int32)iwidth > 0)
1517 uint8 bitsHi = pixels >> 56, bitsLo = pixels >> 48;
1519 if (flagTRANS && (bitsLo | bitsHi) == 0)
1524 *currentLineBuffer = bitsHi,
1525 *(currentLineBuffer + 1) = bitsLo;
1527 *currentLineBuffer =
1528 BLEND_CR(*currentLineBuffer, bitsHi),
1529 *(currentLineBuffer + 1) =
1530 BLEND_Y(*(currentLineBuffer + 1), bitsLo);
1533 currentLineBuffer += lbufDelta;
1535 horizontalRemainder -= 0x20; // Subtract 1.0f in [3.5] fixed point format
1536 while (horizontalRemainder & 0x80)
1538 horizontalRemainder += hscale;
1545 int phrasesToSkip = pixCount / 4, pixelShift = pixCount % 4;
1547 data += (pitch << 3) * phrasesToSkip;
1548 pixels = ((uint64)JaguarReadLong(data, OP) << 32) | JaguarReadLong(data + 4, OP);
1549 pixels <<= 16 * pixelShift;
1551 iwidth -= phrasesToSkip;
1553 pixCount = pixelShift;
1557 else if (depth == 5) // 24 BPP
1559 //I'm not sure that you can scale a 24 BPP bitmap properly--the JTRM seem to indicate as much.
1560 WriteLog("OP: Writing 24 BPP scaled bitmap!\n");
1562 WriteLog("OP: Scaled bitmap @ 24 BPP requesting FIRSTPIX!\n");
1563 // Not sure, but I think RMW only works with 16 BPP and below, and only in CRY mode...
1564 // The LSB is OPFLAG_REFLECT, so sign extend it and or 4 into it.
1565 int32 lbufDelta = ((int8)((flags << 7) & 0xFF) >> 4) | 0x04;
1570 uint64 pixels = ((uint64)JaguarReadLong(data, OP) << 32) | JaguarReadLong(data + 4, OP);
1571 data += pitch << 3; // Multiply pitch * 8 (optimize: precompute this value)
1573 for(int i=0; i<2; i++)
1575 uint8 bits3 = pixels >> 56, bits2 = pixels >> 48,
1576 bits1 = pixels >> 40, bits0 = pixels >> 32;
1578 if (flagTRANS && (bits3 | bits2 | bits1 | bits0) == 0)
1581 *currentLineBuffer = bits3,
1582 *(currentLineBuffer + 1) = bits2,
1583 *(currentLineBuffer + 2) = bits1,
1584 *(currentLineBuffer + 3) = bits0;
1586 currentLineBuffer += lbufDelta;
1591 /*if (depth == 3 && startPos == 13)
1594 WriteLog("OP: Writing in the margins...\n");
1595 for(int i=0; i<100*2; i+=2)
1596 // for(int i=0; i<14*2; i+=2)
1597 tom_ram_8[0x1800 + i] = 0xFF,
1598 tom_ram_8[0x1800 + i + 1] = 0xFF;
1600 // uint32 lbufAddress = 0x1800 + (!in24BPPMode ? startPos * 2 : startPos * 4);
1601 // uint8 * currentLineBuffer = &tom_ram_8[lbufAddress];