6 // I owe a debt of gratitude to Curt Vendel and to John Mathieson--to Curt
7 // for supplying the Oberon ASIC nets and to John for making them available
8 // to Curt. ;-) Without that excellent documentation which shows *exactly*
9 // what's going on inside the TOM chip, we'd all still be guessing as to how
10 // the wily blitter and other pieces of the Jaguar puzzle actually work.
16 // Various conditional compilation goodies...
18 //#define USE_ORIGINAL_BLITTER
19 //#define USE_MIDSUMMER_BLITTER
20 #define USE_MIDSUMMER_BLITTER_MKII
22 // External global variables
24 extern int jaguar_active_memory_dumps;
26 // Local global variables
28 int start_logging = 0;
29 uint8 blitter_working = 0;
31 // Blitter register RAM (most of it is hidden from the user)
33 static uint8 blitter_ram[0x100];
37 bool specialLog = false;
38 extern int effect_start;
39 extern int blit_start_log;
40 void BlitterMidsummer(uint32 cmd);
41 void BlitterMidsummer2(void);
43 #define REG(A) (((uint32)blitter_ram[(A)] << 24) | ((uint32)blitter_ram[(A)+1] << 16) \
44 | ((uint32)blitter_ram[(A)+2] << 8) | (uint32)blitter_ram[(A)+3])
45 #define WREG(A,D) (blitter_ram[(A)] = ((D)>>24)&0xFF, blitter_ram[(A)+1] = ((D)>>16)&0xFF, \
46 blitter_ram[(A)+2] = ((D)>>8)&0xFF, blitter_ram[(A)+3] = (D)&0xFF)
48 // Blitter registers (offsets from F02200)
50 #define A1_BASE ((UINT32)0x00)
51 #define A1_FLAGS ((UINT32)0x04)
52 #define A1_CLIP ((UINT32)0x08) // Height and width values for clipping
53 #define A1_PIXEL ((UINT32)0x0C) // Integer part of the pixel (Y.i and X.i)
54 #define A1_STEP ((UINT32)0x10) // Integer part of the step
55 #define A1_FSTEP ((UINT32)0x14) // Fractional part of the step
56 #define A1_FPIXEL ((UINT32)0x18) // Fractional part of the pixel (Y.f and X.f)
57 #define A1_INC ((UINT32)0x1C) // Integer part of the increment
58 #define A1_FINC ((UINT32)0x20) // Fractional part of the increment
59 #define A2_BASE ((UINT32)0x24)
60 #define A2_FLAGS ((UINT32)0x28)
61 #define A2_MASK ((UINT32)0x2C) // Modulo values for x and y (M.y and M.x)
62 #define A2_PIXEL ((UINT32)0x30) // Integer part of the pixel (no fractional part for A2)
63 #define A2_STEP ((UINT32)0x34) // Integer part of the step (no fractional part for A2)
64 #define COMMAND ((UINT32)0x38)
65 #define PIXLINECOUNTER ((UINT32)0x3C) // Inner & outer loop values
66 #define SRCDATA ((UINT32)0x40)
67 #define DSTDATA ((UINT32)0x48)
68 #define DSTZ ((UINT32)0x50)
69 #define SRCZINT ((UINT32)0x58)
70 #define SRCZFRAC ((UINT32)0x60)
71 #define PATTERNDATA ((UINT32)0x68)
72 #define INTENSITYINC ((UINT32)0x70)
73 #define ZINC ((UINT32)0x74)
74 #define COLLISIONCTRL ((UINT32)0x78)
75 #define PHRASEINT0 ((UINT32)0x7C)
76 #define PHRASEINT1 ((UINT32)0x80)
77 #define PHRASEINT2 ((UINT32)0x84)
78 #define PHRASEINT3 ((UINT32)0x88)
79 #define PHRASEZ0 ((UINT32)0x8C)
80 #define PHRASEZ1 ((UINT32)0x90)
81 #define PHRASEZ2 ((UINT32)0x94)
82 #define PHRASEZ3 ((UINT32)0x98)
84 // Blitter command bits
86 #define SRCEN (cmd & 0x00000001)
87 #define SRCENZ (cmd & 0x00000002)
88 #define SRCENX (cmd & 0x00000004)
89 #define DSTEN (cmd & 0x00000008)
90 #define DSTENZ (cmd & 0x00000010)
91 #define DSTWRZ (cmd & 0x00000020)
92 #define CLIPA1 (cmd & 0x00000040)
94 #define UPDA1F (cmd & 0x00000100)
95 #define UPDA1 (cmd & 0x00000200)
96 #define UPDA2 (cmd & 0x00000400)
98 #define DSTA2 (cmd & 0x00000800)
100 #define Z_OP_INF (cmd & 0x00040000)
101 #define Z_OP_EQU (cmd & 0x00080000)
102 #define Z_OP_SUP (cmd & 0x00100000)
104 #define LFU_NAN (cmd & 0x00200000)
105 #define LFU_NA (cmd & 0x00400000)
106 #define LFU_AN (cmd & 0x00800000)
107 #define LFU_A (cmd & 0x01000000)
109 #define CMPDST (cmd & 0x02000000)
110 #define BCOMPEN (cmd & 0x04000000)
111 #define DCOMPEN (cmd & 0x08000000)
113 #define PATDSEL (cmd & 0x00010000)
114 #define ADDDSEL (cmd & 0x00020000)
115 #define TOPBEN (cmd & 0x00004000)
116 #define TOPNEN (cmd & 0x00008000)
117 #define BKGWREN (cmd & 0x10000000)
118 #define GOURD (cmd & 0x00001000)
119 #define GOURZ (cmd & 0x00002000)
120 #define SRCSHADE (cmd & 0x40000000)
128 #define XSIGNSUB_A1 (REG(A1_FLAGS)&0x080000)
129 #define XSIGNSUB_A2 (REG(A2_FLAGS)&0x080000)
131 #define YSIGNSUB_A1 (REG(A1_FLAGS)&0x100000)
132 #define YSIGNSUB_A2 (REG(A2_FLAGS)&0x100000)
134 #define YADD1_A1 (REG(A1_FLAGS)&0x040000)
135 #define YADD1_A2 (REG(A2_FLAGS)&0x040000)
137 /*******************************************************************************
138 ********************** STUFF CUT BELOW THIS LINE! ******************************
139 *******************************************************************************/
140 #ifdef USE_ORIGINAL_BLITTER // We're ditching this crap for now...
142 //Put 'em back, once we fix the problem!!! [KO]
144 #define PIXEL_SHIFT_1(a) (((~a##_x) >> 16) & 7)
145 #define PIXEL_OFFSET_1(a) (((((UINT32)a##_y >> 16) * a##_width / 8) + (((UINT32)a##_x >> 19) & ~7)) * (1 + a##_pitch) + (((UINT32)a##_x >> 19) & 7))
146 #define READ_PIXEL_1(a) ((JaguarReadByte(a##_addr+PIXEL_OFFSET_1(a), BLITTER) >> PIXEL_SHIFT_1(a)) & 0x01)
147 //#define READ_PIXEL_1(a) ((JaguarReadByte(a##_addr+PIXEL_OFFSET_1(a)) >> PIXEL_SHIFT_1(a)) & 0x01)
150 #define PIXEL_SHIFT_2(a) (((~a##_x) >> 15) & 6)
151 #define PIXEL_OFFSET_2(a) (((((UINT32)a##_y >> 16) * a##_width / 4) + (((UINT32)a##_x >> 18) & ~7)) * (1 + a##_pitch) + (((UINT32)a##_x >> 18) & 7))
152 #define READ_PIXEL_2(a) ((JaguarReadByte(a##_addr+PIXEL_OFFSET_2(a), BLITTER) >> PIXEL_SHIFT_2(a)) & 0x03)
153 //#define READ_PIXEL_2(a) ((JaguarReadByte(a##_addr+PIXEL_OFFSET_2(a)) >> PIXEL_SHIFT_2(a)) & 0x03)
156 #define PIXEL_SHIFT_4(a) (((~a##_x) >> 14) & 4)
157 #define PIXEL_OFFSET_4(a) (((((UINT32)a##_y >> 16) * (a##_width/2)) + (((UINT32)a##_x >> 17) & ~7)) * (1 + a##_pitch) + (((UINT32)a##_x >> 17) & 7))
158 #define READ_PIXEL_4(a) ((JaguarReadByte(a##_addr+PIXEL_OFFSET_4(a), BLITTER) >> PIXEL_SHIFT_4(a)) & 0x0f)
159 //#define READ_PIXEL_4(a) ((JaguarReadByte(a##_addr+PIXEL_OFFSET_4(a)) >> PIXEL_SHIFT_4(a)) & 0x0f)
162 #define PIXEL_OFFSET_8(a) (((((UINT32)a##_y >> 16) * a##_width) + (((UINT32)a##_x >> 16) & ~7)) * (1 + a##_pitch) + (((UINT32)a##_x >> 16) & 7))
163 #define READ_PIXEL_8(a) (JaguarReadByte(a##_addr+PIXEL_OFFSET_8(a), BLITTER))
164 //#define READ_PIXEL_8(a) (JaguarReadByte(a##_addr+PIXEL_OFFSET_8(a)))
167 #define PIXEL_OFFSET_16(a) (((((UINT32)a##_y >> 16) * a##_width) + (((UINT32)a##_x >> 16) & ~3)) * (1 + a##_pitch) + (((UINT32)a##_x >> 16) & 3))
168 #define READ_PIXEL_16(a) (JaguarReadWord(a##_addr+(PIXEL_OFFSET_16(a)<<1), BLITTER))
169 //#define READ_PIXEL_16(a) (JaguarReadWord(a##_addr+(PIXEL_OFFSET_16(a)<<1)))
172 #define PIXEL_OFFSET_32(a) (((((UINT32)a##_y >> 16) * a##_width) + (((UINT32)a##_x >> 16) & ~1)) * (1 + a##_pitch) + (((UINT32)a##_x >> 16) & 1))
173 #define READ_PIXEL_32(a) (JaguarReadLong(a##_addr+(PIXEL_OFFSET_32(a)<<2), BLITTER))
174 //#define READ_PIXEL_32(a) (JaguarReadLong(a##_addr+(PIXEL_OFFSET_32(a)<<2)))
177 #define READ_PIXEL(a,f) (\
178 (((f>>3)&0x07) == 0) ? (READ_PIXEL_1(a)) : \
179 (((f>>3)&0x07) == 1) ? (READ_PIXEL_2(a)) : \
180 (((f>>3)&0x07) == 2) ? (READ_PIXEL_4(a)) : \
181 (((f>>3)&0x07) == 3) ? (READ_PIXEL_8(a)) : \
182 (((f>>3)&0x07) == 4) ? (READ_PIXEL_16(a)) : \
183 (((f>>3)&0x07) == 5) ? (READ_PIXEL_32(a)) : 0)
185 // 16 bpp z data read
186 #define ZDATA_OFFSET_16(a) (PIXEL_OFFSET_16(a) + a##_zoffs * 4)
187 #define READ_ZDATA_16(a) (JaguarReadWord(a##_addr+(ZDATA_OFFSET_16(a)<<1), BLITTER))
188 //#define READ_ZDATA_16(a) (JaguarReadWord(a##_addr+(ZDATA_OFFSET_16(a)<<1)))
191 #define READ_ZDATA(a,f) (READ_ZDATA_16(a))
193 // 16 bpp z data write
194 #define WRITE_ZDATA_16(a,d) { JaguarWriteWord(a##_addr+(ZDATA_OFFSET_16(a)<<1), d, BLITTER); }
195 //#define WRITE_ZDATA_16(a,d) { JaguarWriteWord(a##_addr+(ZDATA_OFFSET_16(a)<<1), d); }
198 #define WRITE_ZDATA(a,f,d) WRITE_ZDATA_16(a,d);
201 #define READ_RDATA_1(r,a,p) ((p) ? ((REG(r+(((UINT32)a##_x >> 19) & 0x04))) >> (((UINT32)a##_x >> 16) & 0x1F)) & 0x0001 : (REG(r) & 0x0001))
204 #define READ_RDATA_2(r,a,p) ((p) ? ((REG(r+(((UINT32)a##_x >> 18) & 0x04))) >> (((UINT32)a##_x >> 15) & 0x3E)) & 0x0003 : (REG(r) & 0x0003))
207 #define READ_RDATA_4(r,a,p) ((p) ? ((REG(r+(((UINT32)a##_x >> 17) & 0x04))) >> (((UINT32)a##_x >> 14) & 0x28)) & 0x000F : (REG(r) & 0x000F))
210 #define READ_RDATA_8(r,a,p) ((p) ? ((REG(r+(((UINT32)a##_x >> 16) & 0x04))) >> (((UINT32)a##_x >> 13) & 0x18)) & 0x00FF : (REG(r) & 0x00FF))
212 // 16 bpp r data read
213 #define READ_RDATA_16(r,a,p) ((p) ? ((REG(r+(((UINT32)a##_x >> 15) & 0x04))) >> (((UINT32)a##_x >> 12) & 0x10)) & 0xFFFF : (REG(r) & 0xFFFF))
215 // 32 bpp r data read
216 #define READ_RDATA_32(r,a,p) ((p) ? REG(r+(((UINT32)a##_x >> 14) & 0x04)) : REG(r))
218 // register data read
219 #define READ_RDATA(r,a,f,p) (\
220 (((f>>3)&0x07) == 0) ? (READ_RDATA_1(r,a,p)) : \
221 (((f>>3)&0x07) == 1) ? (READ_RDATA_2(r,a,p)) : \
222 (((f>>3)&0x07) == 2) ? (READ_RDATA_4(r,a,p)) : \
223 (((f>>3)&0x07) == 3) ? (READ_RDATA_8(r,a,p)) : \
224 (((f>>3)&0x07) == 4) ? (READ_RDATA_16(r,a,p)) : \
225 (((f>>3)&0x07) == 5) ? (READ_RDATA_32(r,a,p)) : 0)
228 #define WRITE_PIXEL_1(a,d) { JaguarWriteByte(a##_addr+PIXEL_OFFSET_1(a), (JaguarReadByte(a##_addr+PIXEL_OFFSET_1(a), BLITTER)&(~(0x01 << PIXEL_SHIFT_1(a))))|(d<<PIXEL_SHIFT_1(a)), BLITTER); }
229 //#define WRITE_PIXEL_1(a,d) { JaguarWriteByte(a##_addr+PIXEL_OFFSET_1(a), (JaguarReadByte(a##_addr+PIXEL_OFFSET_1(a))&(~(0x01 << PIXEL_SHIFT_1(a))))|(d<<PIXEL_SHIFT_1(a))); }
232 #define WRITE_PIXEL_2(a,d) { JaguarWriteByte(a##_addr+PIXEL_OFFSET_2(a), (JaguarReadByte(a##_addr+PIXEL_OFFSET_2(a), BLITTER)&(~(0x03 << PIXEL_SHIFT_2(a))))|(d<<PIXEL_SHIFT_2(a)), BLITTER); }
233 //#define WRITE_PIXEL_2(a,d) { JaguarWriteByte(a##_addr+PIXEL_OFFSET_2(a), (JaguarReadByte(a##_addr+PIXEL_OFFSET_2(a))&(~(0x03 << PIXEL_SHIFT_2(a))))|(d<<PIXEL_SHIFT_2(a))); }
236 #define WRITE_PIXEL_4(a,d) { JaguarWriteByte(a##_addr+PIXEL_OFFSET_4(a), (JaguarReadByte(a##_addr+PIXEL_OFFSET_4(a), BLITTER)&(~(0x0f << PIXEL_SHIFT_4(a))))|(d<<PIXEL_SHIFT_4(a)), BLITTER); }
237 //#define WRITE_PIXEL_4(a,d) { JaguarWriteByte(a##_addr+PIXEL_OFFSET_4(a), (JaguarReadByte(a##_addr+PIXEL_OFFSET_4(a))&(~(0x0f << PIXEL_SHIFT_4(a))))|(d<<PIXEL_SHIFT_4(a))); }
240 #define WRITE_PIXEL_8(a,d) { JaguarWriteByte(a##_addr+PIXEL_OFFSET_8(a), d, BLITTER); }
241 //#define WRITE_PIXEL_8(a,d) { JaguarWriteByte(a##_addr+PIXEL_OFFSET_8(a), d); }
243 // 16 bpp pixel write
244 //#define WRITE_PIXEL_16(a,d) { JaguarWriteWord(a##_addr+(PIXEL_OFFSET_16(a)<<1),d); }
245 #define WRITE_PIXEL_16(a,d) { JaguarWriteWord(a##_addr+(PIXEL_OFFSET_16(a)<<1), d, BLITTER); if (specialLog) WriteLog("Pixel write address: %08X\n", a##_addr+(PIXEL_OFFSET_16(a)<<1)); }
246 //#define WRITE_PIXEL_16(a,d) { JaguarWriteWord(a##_addr+(PIXEL_OFFSET_16(a)<<1), d); if (specialLog) WriteLog("Pixel write address: %08X\n", a##_addr+(PIXEL_OFFSET_16(a)<<1)); }
248 // 32 bpp pixel write
249 #define WRITE_PIXEL_32(a,d) { JaguarWriteLong(a##_addr+(PIXEL_OFFSET_32(a)<<2), d, BLITTER); }
250 //#define WRITE_PIXEL_32(a,d) { JaguarWriteLong(a##_addr+(PIXEL_OFFSET_32(a)<<2), d); }
253 #define WRITE_PIXEL(a,f,d) {\
254 switch ((f>>3)&0x07) { \
255 case 0: WRITE_PIXEL_1(a,d); break; \
256 case 1: WRITE_PIXEL_2(a,d); break; \
257 case 2: WRITE_PIXEL_4(a,d); break; \
258 case 3: WRITE_PIXEL_8(a,d); break; \
259 case 4: WRITE_PIXEL_16(a,d); break; \
260 case 5: WRITE_PIXEL_32(a,d); break; \
263 // Width in Pixels of a Scanline
264 // This is a pretranslation of the value found in the A1 & A2 flags: It's really a floating point value
265 // of the form EEEEMM where MM is the mantissa with an implied "1." in front of it and the EEEE value is
266 // the exponent. Valid values for the exponent range from 0 to 11 (decimal). It's easiest to think of it
267 // as a floating point bit pattern being followed by a number of zeroes. So, e.g., 001101 translates to
268 // 1.01 (the "1." being implied) x (2 ^ 3) or 1010 -> 10 in base 10 (i.e., 1.01 with the decimal place
269 // being shifted to the right 3 places).
270 /*static uint32 blitter_scanline_width[48] =
272 0, 0, 0, 0, // Note: This would really translate to 1, 1, 1, 1
282 1024, 1280, 1536, 1792,
283 2048, 2560, 3072, 3584
286 //static uint8 * tom_ram_8;
287 //static uint8 * paletteRam;
297 static uint32 a1_addr;
298 static uint32 a2_addr;
299 static int32 a1_zoffs;
300 static int32 a2_zoffs;
301 static uint32 xadd_a1_control;
302 static uint32 xadd_a2_control;
303 static int32 a1_pitch;
304 static int32 a2_pitch;
305 static uint32 n_pixels;
306 static uint32 n_lines;
309 static int32 a1_width;
312 static int32 a2_width;
313 static int32 a2_mask_x;
314 static int32 a2_mask_y;
315 static int32 a1_xadd;
316 static int32 a1_yadd;
317 static int32 a2_xadd;
318 static int32 a2_yadd;
319 static uint8 a1_phrase_mode;
320 static uint8 a2_phrase_mode;
321 static int32 a1_step_x = 0;
322 static int32 a1_step_y = 0;
323 static int32 a2_step_x = 0;
324 static int32 a2_step_y = 0;
325 static uint32 outer_loop;
326 static uint32 inner_loop;
327 static uint32 a2_psize;
328 static uint32 a1_psize;
329 static uint32 gouraud_add;
330 //static uint32 gouraud_data;
331 //static uint16 gint[4];
332 //static uint16 gfrac[4];
333 //static uint8 gcolour[4];
336 static int gd_ia, gd_ca;
337 static int colour_index = 0;
339 static uint32 z_i[4];
341 static int32 a1_clip_x, a1_clip_y;
343 // In the spirit of "get it right first, *then* optimize" I've taken the liberty
344 // of removing all the unnecessary code caching. If it turns out to be a good way
345 // to optimize the blitter, then we may revisit it in the future...
348 // Generic blit handler
350 void blitter_generic(uint32 cmd)
353 Blit! (0018FA70 <- 008DDC40) count: 2 x 13, A1/2_FLAGS: 00014218/00013C18 [cmd: 1401060C]
354 CMD -> src: SRCENX dst: DSTEN misc: a1ctl: UPDA1 UPDA2 mode: ity: PATDSEL z-op: op: LFU_CLEAR ctrl: BCOMPEN BKGWREN
355 A1 step values: -2 (X), 1 (Y)
356 A2 step values: -1 (X), 1 (Y) [mask (unused): 00000000 - FFFFFFFF/FFFFFFFF]
357 A1 -> pitch: 1 phrases, depth: 8bpp, z-off: 0, width: 320 (21), addctl: XADDPIX YADD0 XSIGNADD YSIGNADD
358 A2 -> pitch: 1 phrases, depth: 8bpp, z-off: 0, width: 192 (1E), addctl: XADDPIX YADD0 XSIGNADD YSIGNADD
359 A1 x/y: 100/12, A2 x/y: 106/0 Pattern: 000000F300000000
362 // specialLog = true;
363 /*if (cmd == 0x1401060C && blit_start_log)
364 specialLog = true;//*/
366 //uint32 logGo = ((cmd == 0x01800E01 && REG(A1_BASE) == 0x898000) ? 1 : 0);
367 uint32 srcdata, srczdata, dstdata, dstzdata, writedata, inhibit;
368 uint32 bppSrc = (DSTA2 ? 1 << ((REG(A1_FLAGS) >> 3) & 0x07) : 1 << ((REG(A2_FLAGS) >> 3) & 0x07));
372 WriteLog("About to do n x m blit (BM width is ? pixels)...\n");
373 WriteLog("A1_STEP_X/Y = %08X/%08X, A2_STEP_X/Y = %08X/%08X\n", a1_step_x, a1_step_y, a2_step_x, a2_step_y);
387 WriteLog(" A1_X/Y = %08X/%08X, A2_X/Y = %08X/%08X\n", a1_x, a1_y, a2_x, a2_y);
389 uint32 a1_start = a1_x, a2_start = a2_x, bitPos = 0;
391 //Kludge for Hover Strike...
392 //I wonder if this kludge is in conjunction with the SRCENX down below...
393 // This isn't so much a kludge but the way things work in BCOMPEN mode...!
394 if (BCOMPEN && SRCENX)
396 if (n_pixels < bppSrc)
397 bitPos = bppSrc - n_pixels;
400 inner_loop = n_pixels;
405 WriteLog(" A1_X/Y = %08X/%08X, A2_X/Y = %08X/%08X\n", a1_x, a1_y, a2_x, a2_y);
407 srcdata = srczdata = dstdata = dstzdata = writedata = inhibit = 0;
409 if (!DSTA2) // Data movement: A1 <- A2
411 // load src data and Z
413 if (SRCEN || SRCENX) // Not sure if this is correct... (seems to be...!)
415 srcdata = READ_PIXEL(a2, REG(A2_FLAGS));
418 srczdata = READ_ZDATA(a2, REG(A2_FLAGS));
419 else if (cmd & 0x0001C020) // PATDSEL | TOPBEN | TOPNEN | DSTWRZ
420 srczdata = READ_RDATA(SRCZINT, a2, REG(A2_FLAGS), a2_phrase_mode);
422 else // Use SRCDATA register...
424 srcdata = READ_RDATA(SRCDATA, a2, REG(A2_FLAGS), a2_phrase_mode);
426 if (cmd & 0x0001C020) // PATDSEL | TOPBEN | TOPNEN | DSTWRZ
427 srczdata = READ_RDATA(SRCZINT, a2, REG(A2_FLAGS), a2_phrase_mode);
430 // load dst data and Z
433 dstdata = READ_PIXEL(a1, REG(A1_FLAGS));
436 dstzdata = READ_ZDATA(a1, REG(A1_FLAGS));
438 dstzdata = READ_RDATA(DSTZ, a1, REG(A1_FLAGS), a1_phrase_mode);
442 dstdata = READ_RDATA(DSTDATA, a1, REG(A1_FLAGS), a1_phrase_mode);
445 dstzdata = READ_RDATA(DSTZ, a1, REG(A1_FLAGS), a1_phrase_mode);
448 /*This wasn't working... // a1 clipping
449 if (cmd & 0x00000040)
451 if (a1_x < 0 || a1_y < 0 || (a1_x >> 16) >= (REG(A1_CLIP) & 0x7FFF)
452 || (a1_y >> 16) >= ((REG(A1_CLIP) >> 16) & 0x7FFF))
457 srczdata = z_i[colour_index] >> 16;
459 // apply z comparator
460 if (Z_OP_INF && srczdata < dstzdata) inhibit = 1;
461 if (Z_OP_EQU && srczdata == dstzdata) inhibit = 1;
462 if (Z_OP_SUP && srczdata > dstzdata) inhibit = 1;
464 // apply data comparator
465 // Note: DCOMPEN only works in 8/16 bpp modes! !!! FIX !!!
466 // Does BCOMPEN only work in 1 bpp mode???
467 // No, but it always does a 1 bit expansion no matter what the BPP of the channel is set to. !!! FIX !!!
468 // This is bit tricky... We need to fix the XADD value so that it acts like a 1BPP value while inside
470 if (DCOMPEN | BCOMPEN)
472 //Temp, for testing Hover Strike
473 //Doesn't seem to do it... Why?
474 //What needs to happen here is twofold. First, the address generator in the outer loop has
475 //to honor the BPP when calculating the start address (which it kinda does already). Second,
476 //it has to step bit by bit when using BCOMPEN. How to do this???
478 //small problem with this approach: it's not accurate... We need a proper address to begin with
479 //and *then* we can do the bit stepping from there the way it's *supposed* to be done... !!! FIX !!!
482 uint32 pixShift = (~bitPos) & (bppSrc - 1);
483 srcdata = (srcdata >> pixShift) & 0x01;
486 // if (bitPos % bppSrc == 0)
487 // a2_x += 0x00010000;
490 Interesting (Hover Strike--large letter):
492 Blit! (0018FA70 <- 008DDC40) count: 2 x 13, A1/2_FLAGS: 00014218/00013C18 [cmd: 1401060C]
493 CMD -> src: SRCENX dst: DSTEN misc: a1ctl: UPDA1 UPDA2 mode: ity: PATDSEL z-op: op: LFU_CLEAR ctrl: BCOMPEN BKGWREN
494 A1 step values: -2 (X), 1 (Y)
495 A2 step values: -1 (X), 1 (Y) [mask (unused): 00000000 - FFFFFFFF/FFFFFFFF]
496 A1 -> pitch: 1 phrases, depth: 8bpp, z-off: 0, width: 320 (21), addctl: XADDPIX YADD0 XSIGNADD YSIGNADD
497 A2 -> pitch: 1 phrases, depth: 8bpp, z-off: 0, width: 192 (1E), addctl: XADDPIX YADD0 XSIGNADD YSIGNADD
498 A1 x/y: 100/12, A2 x/y: 106/0 Pattern: 000000F300000000
500 Blit! (0018FA70 <- 008DDC40) count: 8 x 13, A1/2_FLAGS: 00014218/00013C18 [cmd: 1401060C]
501 CMD -> src: SRCENX dst: DSTEN misc: a1ctl: UPDA1 UPDA2 mode: ity: PATDSEL z-op: op: LFU_CLEAR ctrl: BCOMPEN BKGWREN
502 A1 step values: -8 (X), 1 (Y)
503 A2 step values: -1 (X), 1 (Y) [mask (unused): 00000000 - FFFFFFFF/FFFFFFFF]
504 A1 -> pitch: 1 phrases, depth: 8bpp, z-off: 0, width: 320 (21), addctl: XADDPIX YADD0 XSIGNADD YSIGNADD
505 A2 -> pitch: 1 phrases, depth: 8bpp, z-off: 0, width: 192 (1E), addctl: XADDPIX YADD0 XSIGNADD YSIGNADD
506 A1 x/y: 102/12, A2 x/y: 107/0 Pattern: 000000F300000000
508 Blit! (0018FA70 <- 008DDC40) count: 1 x 13, A1/2_FLAGS: 00014218/00013C18 [cmd: 1401060C]
509 CMD -> src: SRCENX dst: DSTEN misc: a1ctl: UPDA1 UPDA2 mode: ity: PATDSEL z-op: op: LFU_CLEAR ctrl: BCOMPEN BKGWREN
510 A1 step values: -1 (X), 1 (Y)
511 A2 step values: -1 (X), 1 (Y) [mask (unused): 00000000 - FFFFFFFF/FFFFFFFF]
512 A1 -> pitch: 1 phrases, depth: 8bpp, z-off: 0, width: 320 (21), addctl: XADDPIX YADD0 XSIGNADD YSIGNADD
513 A2 -> pitch: 1 phrases, depth: 8bpp, z-off: 0, width: 192 (1E), addctl: XADDPIX YADD0 XSIGNADD YSIGNADD
514 A1 x/y: 118/12, A2 x/y: 70/0 Pattern: 000000F300000000
516 Blit! (0018FA70 <- 008DDC40) count: 8 x 13, A1/2_FLAGS: 00014218/00013C18 [cmd: 1401060C]
517 CMD -> src: SRCENX dst: DSTEN misc: a1ctl: UPDA1 UPDA2 mode: ity: PATDSEL z-op: op: LFU_CLEAR ctrl: BCOMPEN BKGWREN
518 A1 step values: -8 (X), 1 (Y)
519 A2 step values: -1 (X), 1 (Y) [mask (unused): 00000000 - FFFFFFFF/FFFFFFFF]
520 A1 -> pitch: 1 phrases, depth: 8bpp, z-off: 0, width: 320 (21), addctl: XADDPIX YADD0 XSIGNADD YSIGNADD
521 A2 -> pitch: 1 phrases, depth: 8bpp, z-off: 0, width: 192 (1E), addctl: XADDPIX YADD0 XSIGNADD YSIGNADD
522 A1 x/y: 119/12, A2 x/y: 71/0 Pattern: 000000F300000000
524 Blit! (0018FA70 <- 008DDC40) count: 1 x 13, A1/2_FLAGS: 00014218/00013C18 [cmd: 1401060C]
525 CMD -> src: SRCENX dst: DSTEN misc: a1ctl: UPDA1 UPDA2 mode: ity: PATDSEL z-op: op: LFU_CLEAR ctrl: BCOMPEN BKGWREN
526 A1 step values: -1 (X), 1 (Y)
527 A2 step values: -1 (X), 1 (Y) [mask (unused): 00000000 - FFFFFFFF/FFFFFFFF]
528 A1 -> pitch: 1 phrases, depth: 8bpp, z-off: 0, width: 320 (21), addctl: XADDPIX YADD0 XSIGNADD YSIGNADD
529 A2 -> pitch: 1 phrases, depth: 8bpp, z-off: 0, width: 192 (1E), addctl: XADDPIX YADD0 XSIGNADD YSIGNADD
530 A1 x/y: 127/12, A2 x/y: 66/0 Pattern: 000000F300000000
532 Blit! (0018FA70 <- 008DDC40) count: 8 x 13, A1/2_FLAGS: 00014218/00013C18 [cmd: 1401060C]
533 CMD -> src: SRCENX dst: DSTEN misc: a1ctl: UPDA1 UPDA2 mode: ity: PATDSEL z-op: op: LFU_CLEAR ctrl: BCOMPEN BKGWREN
534 A1 step values: -8 (X), 1 (Y)
535 A2 step values: -1 (X), 1 (Y) [mask (unused): 00000000 - FFFFFFFF/FFFFFFFF]
536 A1 -> pitch: 1 phrases, depth: 8bpp, z-off: 0, width: 320 (21), addctl: XADDPIX YADD0 XSIGNADD YSIGNADD
537 A2 -> pitch: 1 phrases, depth: 8bpp, z-off: 0, width: 192 (1E), addctl: XADDPIX YADD0 XSIGNADD YSIGNADD
538 A1 x/y: 128/12, A2 x/y: 67/0 Pattern: 000000F300000000
544 //WriteLog("Blitter: BCOMPEN set on command %08X inhibit prev:%u, now:", cmd, inhibit);
545 // compare source pixel with pattern pixel
547 Blit! (000B8250 <- 0012C3A0) count: 16 x 1, A1/2_FLAGS: 00014420/00012000 [cmd: 05810001]
548 CMD -> src: SRCEN dst: misc: a1ctl: mode: ity: PATDSEL z-op: op: LFU_REPLACE ctrl: BCOMPEN
549 A1 -> pitch: 1 phrases, depth: 16bpp, z-off: 0, width: 384 (22), addctl: XADDPIX YADD0 XSIGNADD YSIGNADD
550 A2 -> pitch: 1 phrases, depth: 1bpp, z-off: 0, width: 16 (10), addctl: XADDPIX YADD0 XSIGNADD YSIGNADD
554 // AvP is still wrong, could be cuz it's doing A1 -> A2...
556 // Src is the 1bpp bitmap... DST is the PATTERN!!!
557 // This seems to solve at least ONE of the problems with MC3D...
558 // Why should this be inverted???
559 // Bcuz it is. This is supposed to be used only for a bit -> pixel expansion...
560 /* if (srcdata == READ_RDATA(PATTERNDATA, a2, REG(A2_FLAGS), a2_phrase_mode))
561 // if (srcdata != READ_RDATA(PATTERNDATA, a2, REG(A2_FLAGS), a2_phrase_mode))
563 /* uint32 A2bpp = 1 << ((REG(A2_FLAGS) >> 3) & 0x07);
564 if (A2bpp == 1 || A2bpp == 16 || A2bpp == 8)
565 inhibit = (srcdata == 0 ? 1: 0);
566 // inhibit = !srcdata;
568 WriteLog("Blitter: Bad BPP (%u) selected for BCOMPEN mode!\n", A2bpp);//*/
569 // What it boils down to is this:
576 // compare destination pixel with pattern pixel
577 if (dstdata == READ_RDATA(PATTERNDATA, a1, REG(A1_FLAGS), a1_phrase_mode))
578 // if (dstdata != READ_RDATA(PATTERNDATA, a1, REG(A1_FLAGS), a1_phrase_mode))
582 // This is DEFINITELY WRONG
583 // if (a1_phrase_mode || a2_phrase_mode)
584 // inhibit = !inhibit;
589 inhibit |= (((a1_x >> 16) < a1_clip_x && (a1_x >> 16) >= 0
590 && (a1_y >> 16) < a1_clip_y && (a1_y >> 16) >= 0) ? 0 : 1);
593 // compute the write data and store
596 // Houston, we have a problem...
597 // Look here, at PATDSEL and GOURD. If both are active (as they are on the BIOS intro), then there's
599 //Blit! (00100000 <- 000095D0) count: 3 x 1, A1/2_FLAGS: 00014220/00004020 [cmd: 00011008]
600 // CMD -> src: dst: DSTEN misc: a1ctl: mode: GOURD ity: PATDSEL z-op: op: LFU_CLEAR ctrl:
601 // A1 -> pitch: 1 phrases, depth: 16bpp, z-off: 0, width: 320 (21), addctl: XADDPIX YADD0 XSIGNADD YSIGNADD
602 // A2 -> pitch: 1 phrases, depth: 16bpp, z-off: 0, width: 256 (20), addctl: XADDPHR YADD0 XSIGNADD YSIGNADD
603 // A1 x/y: 90/171, A2 x/y: 808/0 Pattern: 776D770077007700
607 // use pattern data for write data
608 writedata = READ_RDATA(PATTERNDATA, a1, REG(A1_FLAGS), a1_phrase_mode);
612 /*if (blit_start_log)
613 WriteLog("BLIT: ADDDSEL srcdata: %08X\, dstdata: %08X, ", srcdata, dstdata);//*/
615 // intensity addition
616 //Ok, this is wrong... Or is it? Yes, it's wrong! !!! FIX !!!
617 /* writedata = (srcdata & 0xFF) + (dstdata & 0xFF);
618 if (!(TOPBEN) && writedata > 0xFF)
621 writedata |= (srcdata & 0xF00) + (dstdata & 0xF00);
622 if (!(TOPNEN) && writedata > 0xFFF)
623 // writedata = 0xFFF;
625 writedata |= (srcdata & 0xF000) + (dstdata & 0xF000);//*/
626 //notneeded--writedata &= 0xFFFF;
627 /*if (blit_start_log)
628 WriteLog("writedata: %08X\n", writedata);//*/
630 Hover Strike ADDDSEL blit:
632 Blit! (00098D90 <- 0081DDC0) count: 320 x 287, A1/2_FLAGS: 00004220/00004020 [cmd: 00020208]
633 CMD -> src: dst: DSTEN misc: a1ctl: UPDA1 mode: ity: ADDDSEL z-op: op: LFU_CLEAR ctrl:
634 A1 step values: -320 (X), 1 (Y)
635 A1 -> pitch: 1 phrases, depth: 16bpp, z-off: 0, width: 320 (21), addctl: XADDPHR YADD0 XSIGNADD YSIGNADD
636 A2 -> pitch: 1 phrases, depth: 16bpp, z-off: 0, width: 256 (20), addctl: XADDPHR YADD0 XSIGNADD YSIGNADD
637 A1 x/y: 0/0, A2 x/y: 3288/0 Pattern: 0000000000000000 SRCDATA: 00FD00FD00FD00FD
639 writedata = (srcdata & 0xFF) + (dstdata & 0xFF);
643 //This is correct now, but slow...
644 int16 s = (srcdata & 0xFF) | (srcdata & 0x80 ? 0xFF00 : 0x0000),
653 writedata = (uint32)sum;
656 //This doesn't seem right... Looks like it would muck up the low byte... !!! FIX !!!
657 writedata |= (srcdata & 0xF00) + (dstdata & 0xF00);
659 if (!TOPNEN && writedata > 0xFFF)
664 writedata |= (srcdata & 0xF000) + (dstdata & 0xF000);
668 if (LFU_NAN) writedata |= ~srcdata & ~dstdata;
669 if (LFU_NA) writedata |= ~srcdata & dstdata;
670 if (LFU_AN) writedata |= srcdata & ~dstdata;
671 if (LFU_A) writedata |= srcdata & dstdata;
674 //Although, this looks like it's OK... (even if it is shitty!)
675 //According to JTRM, this is part of the four things the blitter does with the write data (the other
676 //three being PATDSEL, ADDDSEL, and LFU (default). I'm not sure which gets precedence, this or PATDSEL
677 //(see above blit example)...
679 writedata = ((gd_c[colour_index]) << 8) | (gd_i[colour_index] >> 16);
683 int intensity = srcdata & 0xFF;
684 int ia = gd_ia >> 16;
686 ia = 0xFFFFFF00 | ia;
690 if (intensity > 0xFF)
692 writedata = (srcdata & 0xFF00) | intensity;
701 //Tried 2nd below for Hover Strike: No dice.
702 if (/*a1_phrase_mode || */BKGWREN || !inhibit)
703 // if (/*a1_phrase_mode || BKGWREN ||*/ !inhibit)
705 /*if (((REG(A1_FLAGS) >> 3) & 0x07) == 5)
707 uint32 offset = a1_addr+(PIXEL_OFFSET_32(a1)<<2);
708 // (((((UINT32)a##_y >> 16) * a##_width) + (((UINT32)a##_x >> 16) & ~1)) * (1 + a##_pitch) + (((UINT32)a##_x >> 16) & 1))
709 if ((offset >= 0x1FF020 && offset <= 0x1FF03F) || (offset >= 0x1FF820 && offset <= 0x1FF83F))
710 WriteLog("32bpp pixel write: A1 Phrase mode --> ");
712 // write to the destination
713 WRITE_PIXEL(a1, REG(A1_FLAGS), writedata);
715 WRITE_ZDATA(a1, REG(A1_FLAGS), srczdata);
718 else // if (DSTA2) // Data movement: A1 -> A2
720 // load src data and Z
723 srcdata = READ_PIXEL(a1, REG(A1_FLAGS));
725 srczdata = READ_ZDATA(a1, REG(A1_FLAGS));
726 else if (cmd & 0x0001C020) // PATDSEL | TOPBEN | TOPNEN | DSTWRZ
727 srczdata = READ_RDATA(SRCZINT, a1, REG(A1_FLAGS), a1_phrase_mode);
731 srcdata = READ_RDATA(SRCDATA, a1, REG(A1_FLAGS), a1_phrase_mode);
732 if (cmd & 0x001C020) // PATDSEL | TOPBEN | TOPNEN | DSTWRZ
733 srczdata = READ_RDATA(SRCZINT, a1, REG(A1_FLAGS), a1_phrase_mode);
736 // load dst data and Z
739 dstdata = READ_PIXEL(a2, REG(A2_FLAGS));
741 dstzdata = READ_ZDATA(a2, REG(A2_FLAGS));
743 dstzdata = READ_RDATA(DSTZ, a2, REG(A2_FLAGS), a2_phrase_mode);
747 dstdata = READ_RDATA(DSTDATA, a2, REG(A2_FLAGS), a2_phrase_mode);
749 dstzdata = READ_RDATA(DSTZ, a2, REG(A2_FLAGS), a2_phrase_mode);
753 srczdata = z_i[colour_index] >> 16;
755 // apply z comparator
756 if (Z_OP_INF && srczdata < dstzdata) inhibit = 1;
757 if (Z_OP_EQU && srczdata == dstzdata) inhibit = 1;
758 if (Z_OP_SUP && srczdata > dstzdata) inhibit = 1;
760 // apply data comparator
761 //NOTE: The bit comparator (BCOMPEN) is NOT the same at the data comparator!
762 if (DCOMPEN | BCOMPEN)
766 // compare source pixel with pattern pixel
767 // AvP: Numbers are correct, but sprites are not!
768 //This doesn't seem to be a problem... But could still be wrong...
769 /* if (srcdata == READ_RDATA(PATTERNDATA, a1, REG(A1_FLAGS), a1_phrase_mode))
770 // if (srcdata != READ_RDATA(PATTERNDATA, a1, REG(A1_FLAGS), a1_phrase_mode))
772 // This is probably not 100% correct... It works in the 1bpp case
773 // (in A1 <- A2 mode, that is...)
774 // AvP: This is causing blocks to be written instead of bit patterns...
776 // NOTE: We really should separate out the BCOMPEN & DCOMPEN stuff!
777 /* uint32 A1bpp = 1 << ((REG(A1_FLAGS) >> 3) & 0x07);
778 if (A1bpp == 1 || A1bpp == 16 || A1bpp == 8)
779 inhibit = (srcdata == 0 ? 1: 0);
781 WriteLog("Blitter: Bad BPP (%u) selected for BCOMPEN mode!\n", A1bpp);//*/
782 // What it boils down to is this:
788 // compare destination pixel with pattern pixel
789 if (dstdata == READ_RDATA(PATTERNDATA, a2, REG(A2_FLAGS), a2_phrase_mode))
790 // if (dstdata != READ_RDATA(PATTERNDATA, a2, REG(A2_FLAGS), a2_phrase_mode))
794 // This is DEFINITELY WRONG
795 // if (a1_phrase_mode || a2_phrase_mode)
796 // inhibit = !inhibit;
801 inhibit |= (((a1_x >> 16) < a1_clip_x && (a1_x >> 16) >= 0
802 && (a1_y >> 16) < a1_clip_y && (a1_y >> 16) >= 0) ? 0 : 1);
805 // compute the write data and store
810 // use pattern data for write data
811 writedata = READ_RDATA(PATTERNDATA, a2, REG(A2_FLAGS), a2_phrase_mode);
815 // intensity addition
816 writedata = (srcdata & 0xFF) + (dstdata & 0xFF);
817 if (!(TOPBEN) && writedata > 0xFF)
819 writedata |= (srcdata & 0xF00) + (dstdata & 0xF00);
820 if (!(TOPNEN) && writedata > 0xFFF)
822 writedata |= (srcdata & 0xF000) + (dstdata & 0xF000);
827 writedata |= ~srcdata & ~dstdata;
829 writedata |= ~srcdata & dstdata;
831 writedata |= srcdata & ~dstdata;
833 writedata |= srcdata & dstdata;
837 writedata = ((gd_c[colour_index]) << 8) | (gd_i[colour_index] >> 16);
841 int intensity = srcdata & 0xFF;
842 int ia = gd_ia >> 16;
844 ia = 0xFFFFFF00 | ia;
848 if (intensity > 0xFF)
850 writedata = (srcdata & 0xFF00) | intensity;
859 if (/*a2_phrase_mode || */BKGWREN || !inhibit)
863 uint32 offset = a2_addr+(PIXEL_OFFSET_16(a2)<<1);
864 // (((((UINT32)a##_y >> 16) * a##_width) + (((UINT32)a##_x >> 16) & ~1)) * (1 + a##_pitch) + (((UINT32)a##_x >> 16) & 1))
865 WriteLog("[%08X:%04X] ", offset, writedata);
867 // write to the destination
868 WRITE_PIXEL(a2, REG(A2_FLAGS), writedata);
871 WRITE_ZDATA(a2, REG(A2_FLAGS), srczdata);
875 // Update x and y (inner loop)
876 //Now it does! But crappy, crappy, crappy! !!! FIX !!! [DONE]
877 //This is less than ideal, but it works...
880 a1_x += a1_xadd, a1_y += a1_yadd;
881 a2_x = (a2_x + a2_xadd) & a2_mask_x, a2_y = (a2_y + a2_yadd) & a2_mask_y;
885 a1_y += a1_yadd, a2_y = (a2_y + a2_yadd) & a2_mask_y;
889 if (bitPos % bppSrc == 0)
890 a2_x = (a2_x + a2_xadd) & a2_mask_x;
894 a2_x = (a2_x + a2_xadd) & a2_mask_x;
895 if (bitPos % bppSrc == 0)
901 z_i[colour_index] += zadd;
903 if (GOURD || SRCSHADE)
905 gd_i[colour_index] += gd_ia;
906 //Hmm, this doesn't seem to do anything...
907 //But it is correct according to the JTRM...!
908 if ((int32)gd_i[colour_index] < 0)
909 gd_i[colour_index] = 0;
910 if (gd_i[colour_index] > 0x00FFFFFF)
911 gd_i[colour_index] = 0x00FFFFFF;//*/
913 gd_c[colour_index] += gd_ca;
914 if ((int32)gd_c[colour_index] < 0)
915 gd_c[colour_index] = 0;
916 if (gd_c[colour_index] > 0x000000FF)
917 gd_c[colour_index] = 0x000000FF;//*/
920 if (GOURD || SRCSHADE || GOURZ)
923 //This screws things up WORSE (for the BIOS opening screen)
924 // if (a1_phrase_mode || a2_phrase_mode)
925 colour_index = (colour_index + 1) & 0x03;
930 Here's the problem... The phrase mode code!
931 Blit! (00100000 -> 00148000) count: 327 x 267, A1/2_FLAGS: 00004420/00004420 [cmd: 41802E01]
932 CMD -> src: SRCEN dst: misc: a1ctl: UPDA1 UPDA2 mode: DSTA2 GOURZ ity: z-op: op: LFU_REPLACE ctrl: SRCSHADE
933 A1 step values: -327 (X), 1 (Y)
934 A2 step values: -327 (X), 1 (Y) [mask (unused): 00000000 - FFFFFFFF/FFFFFFFF]
935 A1 -> pitch: 1 phrases, depth: 16bpp, z-off: 0, width: 384 (22), addctl: XADDPHR YADD0 XSIGNADD YSIGNADD
936 A2 -> pitch: 1 phrases, depth: 16bpp, z-off: 0, width: 384 (22), addctl: XADDPHR YADD0 XSIGNADD YSIGNADD
937 A1 x/y: 28/58, A2 x/y: 28/58 Pattern: 00EA7BEA77EA77EA SRCDATA: 7BFF7BFF7BFF7BFF
939 Below fixes it, but then borks:
942 Blit! (00110000 <- 0010B2A8) count: 12 x 12, A1/2_FLAGS: 000042E2/00000020 [cmd: 09800609]
943 CMD -> src: SRCEN dst: DSTEN misc: a1ctl: UPDA1 UPDA2 mode: ity: z-op: op: LFU_REPLACE ctrl: DCOMPEN
944 A1 step values: -15 (X), 1 (Y)
945 A2 step values: -4 (X), 0 (Y) [mask (unused): 00000000 - FFFFFFFF/FFFFFFFF]
946 A1 -> pitch: 4 phrases, depth: 16bpp, z-off: 3, width: 320 (21), addctl: XADDPHR YADD0 XSIGNADD YSIGNADD
947 A2 -> pitch: 1 phrases, depth: 16bpp, z-off: 0, width: 1 (00), addctl: XADDPHR YADD0 XSIGNADD YSIGNADD
948 A1 x/y: 173/144, A2 x/y: 4052/0
950 Lesse, with pre-add we'd have:
953 00001111222233334444555566667777
956 |rolls back to here. Hmm.
959 //NOTE: The way to fix the CD BIOS is to uncomment below and comment the stuff after
960 // the phrase mode mucking around. But it fucks up everything else...
961 //#define SCREWY_CD_DEPENDENT
962 #ifdef SCREWY_CD_DEPENDENT
966 a2_y += a2_step_y;//*/
969 //New: Phrase mode taken into account! :-p
970 /* if (a1_phrase_mode) // v1
972 // Bump the pointer to the next phrase boundary
973 // Even though it works, this is crappy... Clean it up!
974 uint32 size = 64 / a1_psize;
976 // Crappy kludge... ('aligning' source to destination)
977 if (a2_phrase_mode && DSTA2)
979 uint32 extra = (a2_start >> 16) % size;
983 uint32 newx = (a1_x >> 16) / size;
984 uint32 newxrem = (a1_x >> 16) % size;
986 a1_x |= (((newx + (newxrem == 0 ? 0 : 1)) * size) & 0xFFFF) << 16;
988 if (a1_phrase_mode) // v2
990 // Bump the pointer to the next phrase boundary
991 // Even though it works, this is crappy... Clean it up!
992 uint32 size = 64 / a1_psize;
994 // Crappy kludge... ('aligning' source to destination)
995 if (a2_phrase_mode && DSTA2)
997 uint32 extra = (a2_start >> 16) % size;
1001 uint32 pixelSize = (size - 1) << 16;
1002 a1_x = (a1_x + pixelSize) & ~pixelSize;
1005 /* if (a2_phrase_mode) // v1
1007 // Bump the pointer to the next phrase boundary
1008 // Even though it works, this is crappy... Clean it up!
1009 uint32 size = 64 / a2_psize;
1011 // Crappy kludge... ('aligning' source to destination)
1012 // Prolly should do this for A1 channel as well... [DONE]
1013 if (a1_phrase_mode && !DSTA2)
1015 uint32 extra = (a1_start >> 16) % size;
1016 a2_x += extra << 16;
1019 uint32 newx = (a2_x >> 16) / size;
1020 uint32 newxrem = (a2_x >> 16) % size;
1022 a2_x |= (((newx + (newxrem == 0 ? 0 : 1)) * size) & 0xFFFF) << 16;
1024 if (a2_phrase_mode) // v1
1026 // Bump the pointer to the next phrase boundary
1027 // Even though it works, this is crappy... Clean it up!
1028 uint32 size = 64 / a2_psize;
1030 // Crappy kludge... ('aligning' source to destination)
1031 // Prolly should do this for A1 channel as well... [DONE]
1032 if (a1_phrase_mode && !DSTA2)
1034 uint32 extra = (a1_start >> 16) % size;
1035 a2_x += extra << 16;
1038 uint32 pixelSize = (size - 1) << 16;
1039 a2_x = (a2_x + pixelSize) & ~pixelSize;
1042 //Not entirely: This still mucks things up... !!! FIX !!!
1043 //Should this go before or after the phrase mode mucking around?
1044 #ifndef SCREWY_CD_DEPENDENT
1048 a2_y += a2_step_y;//*/
1052 // write values back to registers
1053 WREG(A1_PIXEL, (a1_y & 0xFFFF0000) | ((a1_x >> 16) & 0xFFFF));
1054 WREG(A1_FPIXEL, (a1_y << 16) | (a1_x & 0xFFFF));
1055 WREG(A2_PIXEL, (a2_y & 0xFFFF0000) | ((a2_x >> 16) & 0xFFFF));
1059 void blitter_blit(uint32 cmd)
1061 //Apparently this is doing *something*, just not sure exactly what...
1062 /*if (cmd == 0x41802E01)
1064 WriteLog("BLIT: Found our blit. Was: %08X ", cmd);
1066 WriteLog("Is: %08X\n", cmd);
1069 uint32 pitchValue[4] = { 0, 1, 3, 2 };
1072 dst = (cmd >> 3) & 0x07;
1073 misc = (cmd >> 6) & 0x03;
1074 a1ctl = (cmd >> 8) & 0x7;
1075 mode = (cmd >> 11) & 0x07;
1076 ity = (cmd >> 14) & 0x0F;
1077 zop = (cmd >> 18) & 0x07;
1078 op = (cmd >> 21) & 0x0F;
1079 ctrl = (cmd >> 25) & 0x3F;
1081 // Addresses in A1/2_BASE are *phrase* aligned, i.e., bottom three bits are ignored!
1082 // NOTE: This fixes Rayman's bad collision detection AND keeps T2K working!
1083 a1_addr = REG(A1_BASE) & 0xFFFFFFF8;
1084 a2_addr = REG(A2_BASE) & 0xFFFFFFF8;
1086 a1_zoffs = (REG(A1_FLAGS) >> 6) & 7;
1087 a2_zoffs = (REG(A2_FLAGS) >> 6) & 7;
1089 xadd_a1_control = (REG(A1_FLAGS) >> 16) & 0x03;
1090 xadd_a2_control = (REG(A2_FLAGS) >> 16) & 0x03;
1092 a1_pitch = pitchValue[(REG(A1_FLAGS) & 0x03)];
1093 a2_pitch = pitchValue[(REG(A2_FLAGS) & 0x03)];
1095 n_pixels = REG(PIXLINECOUNTER) & 0xFFFF;
1096 n_lines = (REG(PIXLINECOUNTER) >> 16) & 0xFFFF;
1098 a1_x = (REG(A1_PIXEL) << 16) | (REG(A1_FPIXEL) & 0xFFFF);
1099 a1_y = (REG(A1_PIXEL) & 0xFFFF0000) | (REG(A1_FPIXEL) >> 16);
1100 //According to the JTRM, X is restricted to 15 bits and Y is restricted to 12.
1101 //But it seems to fuck up T2K! !!! FIX !!!
1102 //Could it be sign extended??? Doesn't seem to be so according to JTRM
1103 // a1_x &= 0x7FFFFFFF, a1_y &= 0x0FFFFFFF;
1104 //Actually, it says that the X is 16 bits. But it still seems to mess with the Y when restricted to 12...
1105 // a1_y &= 0x0FFFFFFF;
1107 // a1_width = blitter_scanline_width[((REG(A1_FLAGS) & 0x00007E00) >> 9)];
1108 // According to JTRM, this must give a *whole number* of phrases in the current
1109 // pixel size (this means the lookup above is WRONG)... !!! FIX !!!
1110 UINT32 m = (REG(A1_FLAGS) >> 9) & 0x03, e = (REG(A1_FLAGS) >> 11) & 0x0F;
1111 a1_width = ((0x04 | m) << e) >> 2;//*/
1113 a2_x = (REG(A2_PIXEL) & 0x0000FFFF) << 16;
1114 a2_y = (REG(A2_PIXEL) & 0xFFFF0000);
1115 //According to the JTRM, X is restricted to 15 bits and Y is restricted to 12.
1116 //But it seems to fuck up T2K! !!! FIX !!!
1117 // a2_x &= 0x7FFFFFFF, a2_y &= 0x0FFFFFFF;
1118 //Actually, it says that the X is 16 bits. But it still seems to mess with the Y when restricted to 12...
1119 // a2_y &= 0x0FFFFFFF;
1121 // a2_width = blitter_scanline_width[((REG(A2_FLAGS) & 0x00007E00) >> 9)];
1122 // According to JTRM, this must give a *whole number* of phrases in the current
1123 // pixel size (this means the lookup above is WRONG)... !!! FIX !!!
1124 m = (REG(A2_FLAGS) >> 9) & 0x03, e = (REG(A2_FLAGS) >> 11) & 0x0F;
1125 a2_width = ((0x04 | m) << e) >> 2;//*/
1126 a2_mask_x = ((REG(A2_MASK) & 0x0000FFFF) << 16) | 0xFFFF;
1127 a2_mask_y = (REG(A2_MASK) & 0xFFFF0000) | 0xFFFF;
1129 // Check for "use mask" flag
1130 if (!(REG(A2_FLAGS) & 0x8000))
1132 a2_mask_x = 0xFFFFFFFF; // must be 16.16
1133 a2_mask_y = 0xFFFFFFFF; // must be 16.16
1138 // According to the official documentation, a hardware bug ties A2's yadd bit to A1's...
1139 a2_yadd = a1_yadd = (YADD1_A1 ? 1 << 16 : 0);
1144 // determine a1_xadd
1145 switch (xadd_a1_control)
1148 // This is a documented Jaguar bug relating to phrase mode and truncation... Look into it!
1149 // add phrase offset to X and truncate
1154 // add pixelsize (1) to X
1158 // add zero (for those nice vertical lines)
1162 // add the contents of the increment register
1163 a1_xadd = (REG(A1_INC) << 16) | (REG(A1_FINC) & 0x0000FFFF);
1164 a1_yadd = (REG(A1_INC) & 0xFFFF0000) | (REG(A1_FINC) >> 16);
1169 //Blit! (0011D000 -> 000B9600) count: 228 x 1, A1/2_FLAGS: 00073820/00064220 [cmd: 41802801]
1170 // A1 -> pitch: 1 phrases, depth: 16bpp, z-off: 0, width: 128 (1C), addctl: XADDINC YADD1 XSIGNADD YSIGNADD
1171 // A2 -> pitch: 1 phrases, depth: 16bpp, z-off: 0, width: 320 (21), addctl: XADD0 YADD1 XSIGNADD YSIGNADD
1172 //if (YADD1_A1 && YADD1_A2 && xadd_a2_control == XADD0 && xadd_a1_control == XADDINC)// &&
1173 // UINT32 a1f = REG(A1_FLAGS), a2f = REG(A2_FLAGS);
1174 //Ok, so this ISN'T it... Prolly the XADDPHR code above that's doing it...
1175 //if (REG(A1_FLAGS) == 0x00073820 && REG(A2_FLAGS) == 0x00064220 && cmd == 0x41802801)
1176 // A1 x/y: 14368/7, A2 x/y: 150/36
1177 //This is it... The problem...
1178 //if ((a1_x >> 16) == 14368) // 14368 = $3820
1179 // return; //Lesse what we got...
1189 // determine a2_xadd
1190 switch (xadd_a2_control)
1193 // add phrase offset to X and truncate
1198 // add pixelsize (1) to X
1202 // add zero (for those nice vertical lines)
1205 //This really isn't a valid bit combo for A2... Shouldn't this cause the blitter to just say no?
1207 WriteLog("BLIT: Asked to use invalid bit combo (XADDINC) for A2...\n");
1208 // add the contents of the increment register
1209 // since there is no register for a2 we just add 1
1210 //Let's do nothing, since it's not listed as a valid bit combo...
1211 // a2_xadd = 1 << 16;
1218 // Modify outer loop steps based on blitter command
1226 a1_step_x = (REG(A1_FSTEP) & 0xFFFF),
1227 a1_step_y = (REG(A1_FSTEP) >> 16);
1230 a1_step_x |= ((REG(A1_STEP) & 0x0000FFFF) << 16),
1231 a1_step_y |= ((REG(A1_STEP) & 0xFFFF0000));
1234 a2_step_x = (REG(A2_STEP) & 0x0000FFFF) << 16,
1235 a2_step_y = (REG(A2_STEP) & 0xFFFF0000);
1237 outer_loop = n_lines;
1242 a1_clip_x = REG(A1_CLIP) & 0x7FFF,
1243 a1_clip_y = (REG(A1_CLIP) >> 16) & 0x7FFF;
1245 // This phrase sizing is incorrect as well... !!! FIX !!! [NOTHING TO FIX]
1246 // Err, this is pixel size... (and it's OK)
1247 a2_psize = 1 << ((REG(A2_FLAGS) >> 3) & 0x07);
1248 a1_psize = 1 << ((REG(A1_FLAGS) >> 3) & 0x07);
1255 for(int v=0; v<4; v++)
1256 z_i[v] = REG(PHRASEZ0 + v*4);
1260 if (GOURD || GOURZ || SRCSHADE)
1262 gd_c[0] = blitter_ram[PATTERNDATA + 6];
1263 gd_i[0] = ((uint32)blitter_ram[PATTERNDATA + 7] << 16)
1264 | ((uint32)blitter_ram[SRCDATA + 6] << 8) | blitter_ram[SRCDATA + 7];
1266 gd_c[1] = blitter_ram[PATTERNDATA + 4];
1267 gd_i[1] = ((uint32)blitter_ram[PATTERNDATA + 5] << 16)
1268 | ((uint32)blitter_ram[SRCDATA + 4] << 8) | blitter_ram[SRCDATA + 5];
1270 gd_c[2] = blitter_ram[PATTERNDATA + 2];
1271 gd_i[2] = ((uint32)blitter_ram[PATTERNDATA + 3] << 16)
1272 | ((uint32)blitter_ram[SRCDATA + 2] << 8) | blitter_ram[SRCDATA + 3];
1274 gd_c[3] = blitter_ram[PATTERNDATA + 0];
1275 gd_i[3] = ((uint32)blitter_ram[PATTERNDATA + 1] << 16)
1276 | ((uint32)blitter_ram[SRCDATA + 0] << 8) | blitter_ram[SRCDATA + 1];
1278 gouraud_add = REG(INTENSITYINC);
1280 gd_ia = gouraud_add & 0x00FFFFFF;
1281 if (gd_ia & 0x00800000)
1282 gd_ia = 0xFF000000 | gd_ia;
1284 gd_ca = (gouraud_add >> 24) & 0xFF;
1285 if (gd_ca & 0x00000080)
1286 gd_ca = 0xFFFFFF00 | gd_ca;
1289 // Bit comparitor fixing...
1292 // Determine the data flow direction...
1294 a2_step_x /= (1 << ((REG(A2_FLAGS) >> 3) & 0x07));
1298 /* if (BCOMPEN)//Kludge for Hover Strike... !!! FIX !!!
1300 // Determine the data flow direction...
1308 WriteLog("Blit!\n");
1309 WriteLog(" cmd = 0x%.8x\n",cmd);
1310 WriteLog(" a1_base = %08X\n", a1_addr);
1311 WriteLog(" a1_pitch = %d\n", a1_pitch);
1312 WriteLog(" a1_psize = %d\n", a1_psize);
1313 WriteLog(" a1_width = %d\n", a1_width);
1314 WriteLog(" a1_xadd = %f (phrase=%d)\n", (float)a1_xadd / 65536.0, a1_phrase_mode);
1315 WriteLog(" a1_yadd = %f\n", (float)a1_yadd / 65536.0);
1316 WriteLog(" a1_xstep = %f\n", (float)a1_step_x / 65536.0);
1317 WriteLog(" a1_ystep = %f\n", (float)a1_step_y / 65536.0);
1318 WriteLog(" a1_x = %f\n", (float)a1_x / 65536.0);
1319 WriteLog(" a1_y = %f\n", (float)a1_y / 65536.0);
1320 WriteLog(" a1_zoffs = %i\n",a1_zoffs);
1322 WriteLog(" a2_base = %08X\n", a2_addr);
1323 WriteLog(" a2_pitch = %d\n", a2_pitch);
1324 WriteLog(" a2_psize = %d\n", a2_psize);
1325 WriteLog(" a2_width = %d\n", a2_width);
1326 WriteLog(" a2_xadd = %f (phrase=%d)\n", (float)a2_xadd / 65536.0, a2_phrase_mode);
1327 WriteLog(" a2_yadd = %f\n", (float)a2_yadd / 65536.0);
1328 WriteLog(" a2_xstep = %f\n", (float)a2_step_x / 65536.0);
1329 WriteLog(" a2_ystep = %f\n", (float)a2_step_y / 65536.0);
1330 WriteLog(" a2_x = %f\n", (float)a2_x / 65536.0);
1331 WriteLog(" a2_y = %f\n", (float)a2_y / 65536.0);
1332 WriteLog(" a2_mask_x= 0x%.4x\n",a2_mask_x);
1333 WriteLog(" a2_mask_y= 0x%.4x\n",a2_mask_y);
1334 WriteLog(" a2_zoffs = %i\n",a2_zoffs);
1336 WriteLog(" count = %d x %d\n", n_pixels, n_lines);
1338 WriteLog(" command = %08X\n", cmd);
1339 WriteLog(" dsten = %i\n",DSTEN);
1340 WriteLog(" srcen = %i\n",SRCEN);
1341 WriteLog(" patdsel = %i\n",PATDSEL);
1342 WriteLog(" color = 0x%.8x\n",REG(PATTERNDATA));
1343 WriteLog(" dcompen = %i\n",DCOMPEN);
1344 WriteLog(" bcompen = %i\n",BCOMPEN);
1345 WriteLog(" cmpdst = %i\n",CMPDST);
1346 WriteLog(" GOURZ = %i\n",GOURZ);
1347 WriteLog(" GOURD = %i\n",GOURD);
1348 WriteLog(" SRCSHADE= %i\n",SRCSHADE);
1352 //NOTE: Pitch is ignored!
1354 //This *might* be the altimeter blits (they are)...
1355 //On captured screen, x-pos for black (inner) is 259, for pink is 257
1356 //Black is short by 3, pink is short by 1...
1358 Blit! (00110000 <- 000BF010) count: 9 x 31, A1/2_FLAGS: 000042E2/00010020 [cmd: 00010200]
1359 CMD -> src: dst: misc: a1ctl: UPDA1 mode: ity: PATDSEL z-op: op: LFU_CLEAR ctrl:
1360 A1 -> pitch: 4 phrases, depth: 16bpp, z-off: 3, width: 320 (21), addctl: XADDPHR YADD0 XSIGNADD YSIGNADD
1361 A2 -> pitch: 1 phrases, depth: 16bpp, z-off: 0, width: 1 (00), addctl: XADDPIX YADD0 XSIGNADD YSIGNADD
1362 A1 x/y: 262/124, A2 x/y: 128/0
1363 Blit! (00110000 <- 000BF010) count: 5 x 38, A1/2_FLAGS: 000042E2/00010020 [cmd: 00010200]
1364 CMD -> src: dst: misc: a1ctl: UPDA1 mode: ity: PATDSEL z-op: op: LFU_CLEAR ctrl:
1365 A1 -> pitch: 4 phrases, depth: 16bpp, z-off: 3, width: 320 (21), addctl: XADDPHR YADD0 XSIGNADD YSIGNADD
1366 A2 -> pitch: 1 phrases, depth: 16bpp, z-off: 0, width: 1 (00), addctl: XADDPIX YADD0 XSIGNADD YSIGNADD
1367 A1 x/y: 264/117, A2 x/y: 407/0
1369 Blit! (00110000 <- 000BF010) count: 9 x 23, A1/2_FLAGS: 000042E2/00010020 [cmd: 00010200]
1370 CMD -> src: dst: misc: a1ctl: UPDA1 mode: ity: PATDSEL z-op: op: LFU_CLEAR ctrl:
1371 A1 step values: -10 (X), 1 (Y)
1372 A1 -> pitch: 4(2) phrases, depth: 16bpp, z-off: 3, width: 320 (21), addctl: XADDPHR YADD0 XSIGNADD YSIGNADD
1373 A2 -> pitch: 1(0) phrases, depth: 16bpp, z-off: 0, width: 1 (00), addctl: XADDPIX YADD0 XSIGNADD YSIGNADD
1374 A1 x/y: 262/132, A2 x/y: 129/0
1375 Blit! (00110000 <- 000BF010) count: 5 x 27, A1/2_FLAGS: 000042E2/00010020 [cmd: 00010200]
1376 CMD -> src: dst: misc: a1ctl: UPDA1 mode: ity: PATDSEL z-op: op: LFU_CLEAR ctrl:
1377 A1 step values: -8 (X), 1 (Y)
1378 A1 -> pitch: 4(2) phrases, depth: 16bpp, z-off: 3, width: 320 (21), addctl: XADDPHR YADD0 XSIGNADD YSIGNADD
1379 A2 -> pitch: 1(0) phrases, depth: 16bpp, z-off: 0, width: 1 (00), addctl: XADDPIX YADD0 XSIGNADD YSIGNADD
1380 A1 x/y: 264/128, A2 x/y: 336/0
1382 264v vCursor ends up here...
1386 262v vCursor ends up here...
1390 Fixed! Now for more:
1392 ; This looks like the ship icon in the upper left corner...
1394 Blit! (00110000 <- 0010B2A8) count: 11 x 12, A1/2_FLAGS: 000042E2/00000020 [cmd: 09800609]
1395 CMD -> src: SRCEN dst: DSTEN misc: a1ctl: UPDA1 UPDA2 mode: ity: z-op: op: LFU_REPLACE ctrl: DCOMPEN
1396 A1 step values: -12 (X), 1 (Y)
1397 A2 step values: 0 (X), 0 (Y) [mask (unused): 00000000 - FFFFFFFF/FFFFFFFF]
1398 A1 -> pitch: 4 phrases, depth: 16bpp, z-off: 3, width: 320 (21), addctl: XADDPHR YADD0 XSIGNADD YSIGNADD
1399 A2 -> pitch: 1 phrases, depth: 16bpp, z-off: 0, width: 1 (00), addctl: XADDPHR YADD0 XSIGNADD YSIGNADD
1400 A1 x/y: 20/24, A2 x/y: 5780/0
1404 More (not sure this is a blitter problem as much as it's a GPU problem):
1405 All but the "M" are trashed...
1406 This does *NOT* look like a blitter problem, as it's rendering properly...
1407 Actually, if you look at the A1 step values, there IS a discrepancy!
1411 Blit! (00110000 <- 0010B2A8) count: 12 x 12, A1/2_FLAGS: 000042E2/00000020 [cmd: 09800609]
1412 CMD -> src: SRCEN dst: DSTEN misc: a1ctl: UPDA1 UPDA2 mode: ity: z-op: op: LFU_REPLACE ctrl: DCOMPEN
1413 A1 step values: -14 (X), 1 (Y)
1414 A2 step values: -4 (X), 0 (Y) [mask (unused): 00000000 - FFFFFFFF/FFFFFFFF]
1415 A1 -> pitch: 4 phrases, depth: 16bpp, z-off: 3, width: 320 (21), addctl: XADDPHR YADD0 XSIGNADD YSIGNADD
1416 A2 -> pitch: 1 phrases, depth: 16bpp, z-off: 0, width: 1 (00), addctl: XADDPHR YADD0 XSIGNADD YSIGNADD
1417 A1 x/y: 134/144, A2 x/y: 2516/0
1422 Blit! (00110000 <- 0010B2A8) count: 12 x 12, A1/2_FLAGS: 000042E2/00000020 [cmd: 09800609]
1423 CMD -> src: SRCEN dst: DSTEN misc: a1ctl: UPDA1 UPDA2 mode: ity: z-op: op: LFU_REPLACE ctrl: DCOMPEN
1424 A1 step values: -13 (X), 1 (Y)
1425 A2 step values: -4 (X), 0 (Y) [mask (unused): 00000000 - FFFFFFFF/FFFFFFFF]
1426 A1 -> pitch: 4 phrases, depth: 16bpp, z-off: 3, width: 320 (21), addctl: XADDPHR YADD0 XSIGNADD YSIGNADD
1427 A2 -> pitch: 1 phrases, depth: 16bpp, z-off: 0, width: 1 (00), addctl: XADDPHR YADD0 XSIGNADD YSIGNADD
1428 A1 x/y: 147/144, A2 x/y: 2660/0
1432 Blit! (00110000 <- 0010B2A8) count: 12 x 12, A1/2_FLAGS: 000042E2/00000020 [cmd: 09800609]
1433 CMD -> src: SRCEN dst: DSTEN misc: a1ctl: UPDA1 UPDA2 mode: ity: z-op: op: LFU_REPLACE ctrl: DCOMPEN
1434 A1 step values: -12 (X), 1 (Y)
1435 A2 step values: 0 (X), 0 (Y) [mask (unused): 00000000 - FFFFFFFF/FFFFFFFF]
1436 A1 -> pitch: 4 phrases, depth: 16bpp, z-off: 3, width: 320 (21), addctl: XADDPHR YADD0 XSIGNADD YSIGNADD
1437 A2 -> pitch: 1 phrases, depth: 16bpp, z-off: 0, width: 1 (00), addctl: XADDPHR YADD0 XSIGNADD YSIGNADD
1438 A1 x/y: 160/144, A2 x/y: 3764/0
1442 Blit! (00110000 <- 0010B2A8) count: 12 x 12, A1/2_FLAGS: 000042E2/00000020 [cmd: 09800609]
1443 CMD -> src: SRCEN dst: DSTEN misc: a1ctl: UPDA1 UPDA2 mode: ity: z-op: op: LFU_REPLACE ctrl: DCOMPEN
1444 A1 step values: -15 (X), 1 (Y)
1445 A2 step values: -4 (X), 0 (Y) [mask (unused): 00000000 - FFFFFFFF/FFFFFFFF]
1446 A1 -> pitch: 4 phrases, depth: 16bpp, z-off: 3, width: 320 (21), addctl: XADDPHR YADD0 XSIGNADD YSIGNADD
1447 A2 -> pitch: 1 phrases, depth: 16bpp, z-off: 0, width: 1 (00), addctl: XADDPHR YADD0 XSIGNADD YSIGNADD
1448 A1 x/y: 173/144, A2 x/y: 4052/0
1451 //extern int op_start_log;
1454 char * ctrlStr[4] = { "XADDPHR\0", "XADDPIX\0", "XADD0\0", "XADDINC\0" };
1455 char * bppStr[8] = { "1bpp\0", "2bpp\0", "4bpp\0", "8bpp\0", "16bpp\0", "32bpp\0", "???\0", "!!!\0" };
1456 char * opStr[16] = { "LFU_CLEAR", "LFU_NSAND", "LFU_NSAD", "LFU_NOTS", "LFU_SAND", "LFU_NOTD", "LFU_N_SXORD", "LFU_NSORND",
1457 "LFU_SAD", "LFU_XOR", "LFU_D", "LFU_NSORD", "LFU_REPLACE", "LFU_SORND", "LFU_SORD", "LFU_ONE" };
1458 uint32 /*src = cmd & 0x07, dst = (cmd >> 3) & 0x07, misc = (cmd >> 6) & 0x03,
1459 a1ctl = (cmd >> 8) & 0x07,*/ mode = (cmd >> 11) & 0x07/*, ity = (cmd >> 14) & 0x0F,
1460 zop = (cmd >> 18) & 0x07, op = (cmd >> 21) & 0x0F, ctrl = (cmd >> 25) & 0x3F*/;
1461 UINT32 a1f = REG(A1_FLAGS), a2f = REG(A2_FLAGS);
1462 uint32 p1 = a1f & 0x07, p2 = a2f & 0x07,
1463 d1 = (a1f >> 3) & 0x07, d2 = (a2f >> 3) & 0x07,
1464 zo1 = (a1f >> 6) & 0x07, zo2 = (a2f >> 6) & 0x07,
1465 w1 = (a1f >> 9) & 0x3F, w2 = (a2f >> 9) & 0x3F,
1466 ac1 = (a1f >> 16) & 0x1F, ac2 = (a2f >> 16) & 0x1F;
1467 UINT32 iw1 = ((0x04 | (w1 & 0x03)) << ((w1 & 0x3C) >> 2)) >> 2;
1468 UINT32 iw2 = ((0x04 | (w2 & 0x03)) << ((w2 & 0x3C) >> 2)) >> 2;
1469 WriteLog("Blit! (%08X %s %08X) count: %d x %d, A1/2_FLAGS: %08X/%08X [cmd: %08X]\n", a1_addr, (mode&0x01 ? "->" : "<-"), a2_addr, n_pixels, n_lines, a1f, a2f, cmd);
1470 // WriteLog(" CMD -> src: %d, dst: %d, misc: %d, a1ctl: %d, mode: %d, ity: %1X, z-op: %d, op: %1X, ctrl: %02X\n", src, dst, misc, a1ctl, mode, ity, zop, op, ctrl);
1472 WriteLog(" CMD -> src: %s%s%s ", (cmd & 0x0001 ? "SRCEN " : ""), (cmd & 0x0002 ? "SRCENZ " : ""), (cmd & 0x0004 ? "SRCENX" : ""));
1473 WriteLog("dst: %s%s%s ", (cmd & 0x0008 ? "DSTEN " : ""), (cmd & 0x0010 ? "DSTENZ " : ""), (cmd & 0x0020 ? "DSTWRZ" : ""));
1474 WriteLog("misc: %s%s ", (cmd & 0x0040 ? "CLIP_A1 " : ""), (cmd & 0x0080 ? "???" : ""));
1475 WriteLog("a1ctl: %s%s%s ", (cmd & 0x0100 ? "UPDA1F " : ""), (cmd & 0x0200 ? "UPDA1 " : ""), (cmd & 0x0400 ? "UPDA2" : ""));
1476 WriteLog("mode: %s%s%s ", (cmd & 0x0800 ? "DSTA2 " : ""), (cmd & 0x1000 ? "GOURD " : ""), (cmd & 0x2000 ? "GOURZ" : ""));
1477 WriteLog("ity: %s%s%s%s ", (cmd & 0x4000 ? "TOPBEN " : ""), (cmd & 0x8000 ? "TOPNEN " : ""), (cmd & 0x00010000 ? "PATDSEL" : ""), (cmd & 0x00020000 ? "ADDDSEL" : ""));
1478 WriteLog("z-op: %s%s%s ", (cmd & 0x00040000 ? "ZMODELT " : ""), (cmd & 0x00080000 ? "ZMODEEQ " : ""), (cmd & 0x00100000 ? "ZMODEGT" : ""));
1479 WriteLog("op: %s ", opStr[(cmd >> 21) & 0x0F]);
1480 WriteLog("ctrl: %s%s%s%s%s%s\n", (cmd & 0x02000000 ? "CMPDST " : ""), (cmd & 0x04000000 ? "BCOMPEN " : ""), (cmd & 0x08000000 ? "DCOMPEN " : ""), (cmd & 0x10000000 ? "BKGWREN " : ""), (cmd & 0x20000000 ? "BUSHI " : ""), (cmd & 0x40000000 ? "SRCSHADE" : ""));
1483 WriteLog(" A1 step values: %d (X), %d (Y)\n", a1_step_x >> 16, a1_step_y >> 16);
1486 WriteLog(" A2 step values: %d (X), %d (Y) [mask (%sused): %08X - %08X/%08X]\n", a2_step_x >> 16, a2_step_y >> 16, (a2f & 0x8000 ? "" : "un"), REG(A2_MASK), a2_mask_x, a2_mask_y);
1488 WriteLog(" A1 -> pitch: %d phrases, depth: %s, z-off: %d, width: %d (%02X), addctl: %s %s %s %s\n", 1 << p1, bppStr[d1], zo1, iw1, w1, ctrlStr[ac1&0x03], (ac1&0x04 ? "YADD1" : "YADD0"), (ac1&0x08 ? "XSIGNSUB" : "XSIGNADD"), (ac1&0x10 ? "YSIGNSUB" : "YSIGNADD"));
1489 WriteLog(" A2 -> pitch: %d phrases, depth: %s, z-off: %d, width: %d (%02X), addctl: %s %s %s %s\n", 1 << p2, bppStr[d2], zo2, iw2, w2, ctrlStr[ac2&0x03], (ac2&0x04 ? "YADD1" : "YADD0"), (ac2&0x08 ? "XSIGNSUB" : "XSIGNADD"), (ac2&0x10 ? "YSIGNSUB" : "YSIGNADD"));
1490 WriteLog(" A1 x/y: %d/%d, A2 x/y: %d/%d Pattern: %08X%08X SRCDATA: %08X%08X\n", a1_x >> 16, a1_y >> 16, a2_x >> 16, a2_y >> 16, REG(PATTERNDATA), REG(PATTERNDATA + 4), REG(SRCDATA), REG(SRCDATA + 4));
1491 // blit_start_log = 0;
1492 // op_start_log = 1;
1495 blitter_working = 1;
1496 //#ifndef USE_GENERIC_BLITTER
1497 // if (!blitter_execute_cached_code(blitter_in_cache(cmd)))
1499 blitter_generic(cmd);
1501 /*if (blit_start_log)
1503 if (a1_addr == 0xF03000 && a2_addr == 0x004D58)
1505 WriteLog("\nBytes at 004D58:\n");
1506 for(int i=0x004D58; i<0x004D58+(10*127*4); i++)
1507 WriteLog("%02X ", JaguarReadByte(i));
1508 WriteLog("\nBytes at F03000:\n");
1509 for(int i=0xF03000; i<0xF03000+(6*127*4); i++)
1510 WriteLog("%02X ", JaguarReadByte(i));
1515 blitter_working = 0;
1517 #endif // of the #if 0 near the top...
1518 /*******************************************************************************
1519 ********************** STUFF CUT ABOVE THIS LINE! ******************************
1520 *******************************************************************************/
1522 void blitter_init(void)
1527 void blitter_reset(void)
1529 memset(blitter_ram, 0x00, 0xA0);
1532 void blitter_done(void)
1534 WriteLog("BLIT: Done.\n");
1537 uint8 BlitterReadByte(uint32 offset, uint32 who/*=UNKNOWN*/)
1542 //This isn't cycle accurate--how to fix? !!! FIX !!!
1543 //Probably have to do some multi-threaded implementation or at least a reentrant safe implementation...
1544 if (offset == (0x38 + 3))
1545 return 0x01; // always idle
1547 // CHECK HERE ONCE THIS FIX HAS BEEN TESTED: [ ]
1549 if (offset >= 0x04 && offset <= 0x07)
1550 //This is it. I wonder if it just ignores the lower three bits?
1551 //No, this is a documented Jaguar I bug. It also bites the read at $F02230 as well...
1552 return blitter_ram[offset + 0x08]; // A1_PIXEL ($F0220C) read at $F02204
1554 if (offset >= 0x2C && offset <= 0x2F)
1555 return blitter_ram[offset + 0x04]; // A2_PIXEL ($F02230) read at $F0222C
1557 return blitter_ram[offset];
1561 uint16 BlitterReadWord(uint32 offset, uint32 who/*=UNKNOWN*/)
1563 return ((uint16)BlitterReadByte(offset, who) << 8) | (uint16)BlitterReadByte(offset+1, who);
1567 uint32 BlitterReadLong(uint32 offset, uint32 who/*=UNKNOWN*/)
1569 return (BlitterReadWord(offset, who) << 16) | BlitterReadWord(offset+2, who);
1572 void BlitterWriteByte(uint32 offset, uint8 data, uint32 who/*=UNKNOWN*/)
1574 /*if (offset & 0xFF == 0x7B)
1575 WriteLog("--> Wrote to B_STOP: value -> %02X\n", data);*/
1577 /*if ((offset >= PATTERNDATA) && (offset < PATTERNDATA + 8))
1579 printf("--> %s wrote %02X to byte %u of PATTERNDATA...\n", whoName[who], data, offset - PATTERNDATA);
1583 // This handles writes to INTENSITY0-3 by also writing them to their proper places in
1584 // PATTERNDATA & SOURCEDATA (should do the same for the Z registers! !!! FIX !!! [DONE])
1585 if ((offset >= 0x7C) && (offset <= 0x9B))
1589 // INTENSITY registers 0-3
1591 case 0x7D: blitter_ram[PATTERNDATA + 7] = data; break;
1592 case 0x7E: blitter_ram[SRCDATA + 6] = data; break;
1593 case 0x7F: blitter_ram[SRCDATA + 7] = data; break;
1596 case 0x81: blitter_ram[PATTERNDATA + 5] = data; break;
1597 case 0x82: blitter_ram[SRCDATA + 4] = data; break;
1598 case 0x83: blitter_ram[SRCDATA + 5] = data; break;
1601 case 0x85: blitter_ram[PATTERNDATA + 3] = data; break;
1602 case 0x86: blitter_ram[SRCDATA + 2] = data; break;
1603 case 0x87: blitter_ram[SRCDATA + 3] = data; break;
1606 case 0x89: blitter_ram[PATTERNDATA + 1] = data; break;
1607 case 0x8A: blitter_ram[SRCDATA + 0] = data; break;
1608 case 0x8B: blitter_ram[SRCDATA + 1] = data; break;
1612 case 0x8C: blitter_ram[SRCZINT + 6] = data; break;
1613 case 0x8D: blitter_ram[SRCZINT + 7] = data; break;
1614 case 0x8E: blitter_ram[SRCZFRAC + 6] = data; break;
1615 case 0x8F: blitter_ram[SRCZFRAC + 7] = data; break;
1617 case 0x90: blitter_ram[SRCZINT + 4] = data; break;
1618 case 0x91: blitter_ram[SRCZINT + 5] = data; break;
1619 case 0x92: blitter_ram[SRCZFRAC + 4] = data; break;
1620 case 0x93: blitter_ram[SRCZFRAC + 5] = data; break;
1622 case 0x94: blitter_ram[SRCZINT + 2] = data; break;
1623 case 0x95: blitter_ram[SRCZINT + 3] = data; break;
1624 case 0x96: blitter_ram[SRCZFRAC + 2] = data; break;
1625 case 0x97: blitter_ram[SRCZFRAC + 3] = data; break;
1627 case 0x98: blitter_ram[SRCZINT + 0] = data; break;
1628 case 0x99: blitter_ram[SRCZINT + 1] = data; break;
1629 case 0x9A: blitter_ram[SRCZFRAC + 0] = data; break;
1630 case 0x9B: blitter_ram[SRCZFRAC + 1] = data; break;
1634 // It looks weird, but this is how the 64 bit registers are actually handled...!
1636 else if ((offset >= SRCDATA + 0) && (offset <= SRCDATA + 3)
1637 || (offset >= DSTDATA + 0) && (offset <= DSTDATA + 3)
1638 || (offset >= DSTZ + 0) && (offset <= DSTZ + 3)
1639 || (offset >= SRCZINT + 0) && (offset <= SRCZINT + 3)
1640 || (offset >= SRCZFRAC + 0) && (offset <= SRCZFRAC + 3)
1641 || (offset >= PATTERNDATA + 0) && (offset <= PATTERNDATA + 3))
1643 blitter_ram[offset + 4] = data;
1645 else if ((offset >= SRCDATA + 4) && (offset <= SRCDATA + 7)
1646 || (offset >= DSTDATA + 4) && (offset <= DSTDATA + 7)
1647 || (offset >= DSTZ + 4) && (offset <= DSTZ + 7)
1648 || (offset >= SRCZINT + 4) && (offset <= SRCZINT + 7)
1649 || (offset >= SRCZFRAC + 4) && (offset <= SRCZFRAC + 7)
1650 || (offset >= PATTERNDATA + 4) && (offset <= PATTERNDATA + 7))
1652 blitter_ram[offset - 4] = data;
1655 blitter_ram[offset] = data;
1658 void BlitterWriteWord(uint32 offset, uint16 data, uint32 who/*=UNKNOWN*/)
1660 /*if (((offset & 0xFF) >= PATTERNDATA) && ((offset & 0xFF) < PATTERNDATA + 8))
1662 printf("----> %s wrote %04X to byte %u of PATTERNDATA...\n", whoName[who], data, offset - (0xF02200 + PATTERNDATA));
1666 /* if (offset & 0xFF == A1_PIXEL && data == 14368)
1668 WriteLog("\n1\nA1_PIXEL written by %s (%u)...\n\n\n", whoName[who], data);
1669 extern bool doGPUDis;
1672 if ((offset & 0xFF) == (A1_PIXEL + 2) && data == 14368)
1674 WriteLog("\n2\nA1_PIXEL written by %s (%u)...\n\n\n", whoName[who], data);
1675 extern bool doGPUDis;
1680 BlitterWriteByte(offset + 0, data >> 8, who);
1681 BlitterWriteByte(offset + 1, data & 0xFF, who);
1683 if ((offset & 0xFF) == 0x3A)
1684 // I.e., the second write of 32-bit value--not convinced this is the best way to do this!
1685 // But then again, according to the Jaguar docs, this is correct...!
1686 /*extern int blit_start_log;
1687 extern bool doGPUDis;
1690 WriteLog("BLIT: Blitter started by %s...\n", whoName[who]);
1693 #ifdef USE_ORIGINAL_BLITTER
1694 blitter_blit(GET32(blitter_ram, 0x38));
1696 #ifdef USE_MIDSUMMER_BLITTER
1697 BlitterMidsummer(GET32(blitter_ram, 0x38));
1699 #ifdef USE_MIDSUMMER_BLITTER_MKII
1700 BlitterMidsummer2();
1705 void BlitterWriteLong(uint32 offset, uint32 data, uint32 who/*=UNKNOWN*/)
1707 /*if (((offset & 0xFF) >= PATTERNDATA) && ((offset & 0xFF) < PATTERNDATA + 8))
1709 printf("------> %s wrote %08X to byte %u of PATTERNDATA...\n", whoName[who], data, offset - (0xF02200 + PATTERNDATA));
1713 /* if ((offset & 0xFF) == A1_PIXEL && (data & 0xFFFF) == 14368)
1715 WriteLog("\n3\nA1_PIXEL written by %s (%u)...\n\n\n", whoName[who], data);
1716 extern bool doGPUDis;
1721 BlitterWriteWord(offset + 0, data >> 16, who);
1722 BlitterWriteWord(offset + 2, data & 0xFFFF, who);
1727 char * opStr[16] = { "LFU_CLEAR", "LFU_NSAND", "LFU_NSAD", "LFU_NOTS", "LFU_SAND", "LFU_NOTD", "LFU_N_SXORD", "LFU_NSORND",
1728 "LFU_SAD", "LFU_XOR", "LFU_D", "LFU_NSORD", "LFU_REPLACE", "LFU_SORND", "LFU_SORD", "LFU_ONE" };
1729 uint32 cmd = GET32(blitter_ram, 0x38);
1730 UINT32 m = (REG(A1_FLAGS) >> 9) & 0x03, e = (REG(A1_FLAGS) >> 11) & 0x0F;
1731 UINT32 a1_width = ((0x04 | m) << e) >> 2;
1732 m = (REG(A2_FLAGS) >> 9) & 0x03, e = (REG(A2_FLAGS) >> 11) & 0x0F;
1733 UINT32 a2_width = ((0x04 | m) << e) >> 2;
1735 WriteLog("Blit!\n");
1736 WriteLog(" COMMAND = %08X\n", cmd);
1737 WriteLog(" a1_base = %08X\n", REG(A1_BASE));
1738 WriteLog(" a1_flags = %08X (%c %c %c %c%c . %c%c%c%c%c%c %c%c%c %c%c%c . %c%c)\n", REG(A1_FLAGS),
1739 (REG(A1_FLAGS) & 0x100000 ? '1' : '0'),
1740 (REG(A1_FLAGS) & 0x080000 ? '1' : '0'),
1741 (REG(A1_FLAGS) & 0x040000 ? '1' : '0'),
1742 (REG(A1_FLAGS) & 0x020000 ? '1' : '0'),
1743 (REG(A1_FLAGS) & 0x010000 ? '1' : '0'),
1744 (REG(A1_FLAGS) & 0x004000 ? '1' : '0'),
1745 (REG(A1_FLAGS) & 0x002000 ? '1' : '0'),
1746 (REG(A1_FLAGS) & 0x001000 ? '1' : '0'),
1747 (REG(A1_FLAGS) & 0x000800 ? '1' : '0'),
1748 (REG(A1_FLAGS) & 0x000400 ? '1' : '0'),
1749 (REG(A1_FLAGS) & 0x000200 ? '1' : '0'),
1750 (REG(A1_FLAGS) & 0x000100 ? '1' : '0'),
1751 (REG(A1_FLAGS) & 0x000080 ? '1' : '0'),
1752 (REG(A1_FLAGS) & 0x000040 ? '1' : '0'),
1753 (REG(A1_FLAGS) & 0x000020 ? '1' : '0'),
1754 (REG(A1_FLAGS) & 0x000010 ? '1' : '0'),
1755 (REG(A1_FLAGS) & 0x000008 ? '1' : '0'),
1756 (REG(A1_FLAGS) & 0x000002 ? '1' : '0'),
1757 (REG(A1_FLAGS) & 0x000001 ? '1' : '0'));
1758 WriteLog(" pitch=%u, pixSz=%u, zOff=%u, width=%u, xCtrl=%u\n",
1759 REG(A1_FLAGS) & 0x00003, (REG(A1_FLAGS) & 0x00038) >> 3,
1760 (REG(A1_FLAGS) & 0x001C0) >> 6, a1_width, (REG(A1_FLAGS) & 0x30000) >> 16);
1761 WriteLog(" a1_clip = %u, %u (%08X)\n", GET16(blitter_ram, A1_CLIP + 2), GET16(blitter_ram, A1_CLIP + 0), GET32(blitter_ram, A1_CLIP));
1762 WriteLog(" a1_pixel = %d, %d (%08X)\n", (int16)GET16(blitter_ram, A1_PIXEL + 2), (int16)GET16(blitter_ram, A1_PIXEL + 0), GET32(blitter_ram, A1_PIXEL));
1763 WriteLog(" a1_step = %d, %d (%08X)\n", (int16)GET16(blitter_ram, A1_STEP + 2), (int16)GET16(blitter_ram, A1_STEP + 0), GET32(blitter_ram, A1_STEP));
1764 WriteLog(" a1_fstep = %u, %u (%08X)\n", GET16(blitter_ram, A1_FSTEP + 2), GET16(blitter_ram, A1_FSTEP + 0), GET32(blitter_ram, A1_FSTEP));
1765 WriteLog(" a1_fpixel= %u, %u (%08X)\n", GET16(blitter_ram, A1_FPIXEL + 2), GET16(blitter_ram, A1_FPIXEL + 0), GET32(blitter_ram, A1_FPIXEL));
1766 WriteLog(" a1_inc = %d, %d (%08X)\n", (int16)GET16(blitter_ram, A1_INC + 2), (int16)GET16(blitter_ram, A1_INC + 0), GET32(blitter_ram, A1_INC));
1767 WriteLog(" a1_finc = %u, %u (%08X)\n", GET16(blitter_ram, A1_FINC + 2), GET16(blitter_ram, A1_FINC + 0), GET32(blitter_ram, A1_FINC));
1769 WriteLog(" a2_base = %08X\n", REG(A2_BASE));
1770 WriteLog(" a2_flags = %08X (%c %c %c %c%c %c %c%c%c%c%c%c %c%c%c %c%c%c . %c%c)\n", REG(A2_FLAGS),
1771 (REG(A2_FLAGS) & 0x100000 ? '1' : '0'),
1772 (REG(A2_FLAGS) & 0x080000 ? '1' : '0'),
1773 (REG(A2_FLAGS) & 0x040000 ? '1' : '0'),
1774 (REG(A2_FLAGS) & 0x020000 ? '1' : '0'),
1775 (REG(A2_FLAGS) & 0x010000 ? '1' : '0'),
1776 (REG(A2_FLAGS) & 0x008000 ? '1' : '0'),
1777 (REG(A2_FLAGS) & 0x004000 ? '1' : '0'),
1778 (REG(A2_FLAGS) & 0x002000 ? '1' : '0'),
1779 (REG(A2_FLAGS) & 0x001000 ? '1' : '0'),
1780 (REG(A2_FLAGS) & 0x000800 ? '1' : '0'),
1781 (REG(A2_FLAGS) & 0x000400 ? '1' : '0'),
1782 (REG(A2_FLAGS) & 0x000200 ? '1' : '0'),
1783 (REG(A2_FLAGS) & 0x000100 ? '1' : '0'),
1784 (REG(A2_FLAGS) & 0x000080 ? '1' : '0'),
1785 (REG(A2_FLAGS) & 0x000040 ? '1' : '0'),
1786 (REG(A2_FLAGS) & 0x000020 ? '1' : '0'),
1787 (REG(A2_FLAGS) & 0x000010 ? '1' : '0'),
1788 (REG(A2_FLAGS) & 0x000008 ? '1' : '0'),
1789 (REG(A2_FLAGS) & 0x000002 ? '1' : '0'),
1790 (REG(A2_FLAGS) & 0x000001 ? '1' : '0'));
1791 WriteLog(" pitch=%u, pixSz=%u, zOff=%u, width=%u, xCtrl=%u\n",
1792 REG(A2_FLAGS) & 0x00003, (REG(A2_FLAGS) & 0x00038) >> 3,
1793 (REG(A2_FLAGS) & 0x001C0) >> 6, a2_width, (REG(A2_FLAGS) & 0x30000) >> 16);
1794 WriteLog(" a2_mask = %u, %u (%08X)\n", GET16(blitter_ram, A2_MASK + 2), GET16(blitter_ram, A2_MASK + 0), GET32(blitter_ram, A2_MASK));
1795 WriteLog(" a2_pixel = %d, %d (%08X)\n", (int16)GET16(blitter_ram, A2_PIXEL + 2), (int16)GET16(blitter_ram, A2_PIXEL + 0), GET32(blitter_ram, A2_PIXEL));
1796 WriteLog(" a2_step = %d, %d (%08X)\n", (int16)GET16(blitter_ram, A2_STEP + 2), (int16)GET16(blitter_ram, A2_STEP + 0), GET32(blitter_ram, A2_STEP));
1798 WriteLog(" count = %d x %d\n", GET16(blitter_ram, PIXLINECOUNTER + 2), GET16(blitter_ram, PIXLINECOUNTER));
1800 WriteLog(" SRCEN = %s\n", (SRCEN ? "1" : "0"));
1801 WriteLog(" SRCENZ = %s\n", (SRCENZ ? "1" : "0"));
1802 WriteLog(" SRCENX = %s\n", (SRCENX ? "1" : "0"));
1803 WriteLog(" DSTEN = %s\n", (DSTEN ? "1" : "0"));
1804 WriteLog(" DSTENZ = %s\n", (DSTENZ ? "1" : "0"));
1805 WriteLog(" DSTWRZ = %s\n", (DSTWRZ ? "1" : "0"));
1806 WriteLog(" CLIPA1 = %s\n", (CLIPA1 ? "1" : "0"));
1807 WriteLog(" UPDA1F = %s\n", (UPDA1F ? "1" : "0"));
1808 WriteLog(" UPDA1 = %s\n", (UPDA1 ? "1" : "0"));
1809 WriteLog(" UPDA2 = %s\n", (UPDA2 ? "1" : "0"));
1810 WriteLog(" DSTA2 = %s\n", (DSTA2 ? "1" : "0"));
1811 WriteLog(" ZOP = %s %s %s\n", (Z_OP_INF ? "<" : ""), (Z_OP_EQU ? "=" : ""), (Z_OP_SUP ? ">" : ""));
1812 WriteLog("--LFUFUNC = %s\n", opStr[(cmd >> 21) & 0x0F]);
1813 WriteLog("| PATDSEL = %s (PD=%08X%08X)\n", (PATDSEL ? "1" : "0"), REG(PATTERNDATA), REG(PATTERNDATA + 4));
1814 WriteLog("--ADDDSEL = %s\n", (ADDDSEL ? "1" : "0"));
1815 WriteLog(" CMPDST = %s\n", (CMPDST ? "1" : "0"));
1816 WriteLog(" BCOMPEN = %s\n", (BCOMPEN ? "1" : "0"));
1817 WriteLog(" DCOMPEN = %s\n", (DCOMPEN ? "1" : "0"));
1818 WriteLog(" TOPBEN = %s\n", (TOPBEN ? "1" : "0"));
1819 WriteLog(" TOPNEN = %s\n", (TOPNEN ? "1" : "0"));
1820 WriteLog(" BKGWREN = %s\n", (BKGWREN ? "1" : "0"));
1821 WriteLog(" GOURD = %s (II=%08X, SD=%08X%08X)\n", (GOURD ? "1" : "0"), REG(INTENSITYINC), REG(SRCDATA), REG(SRCDATA + 4));
1822 WriteLog(" GOURZ = %s (ZI=%08X, ZD=%08X%08X, SZ1=%08X%08X, SZ2=%08X%08X)\n", (GOURZ ? "1" : "0"), REG(ZINC), REG(DSTZ), REG(DSTZ + 4),
1823 REG(SRCZINT), REG(SRCZINT + 4), REG(SRCZFRAC), REG(SRCZFRAC + 4));
1824 WriteLog(" SRCSHADE = %s\n", (SRCSHADE ? "1" : "0"));
1828 #ifdef USE_MIDSUMMER_BLITTER
1830 // Here's an attempt to write a blitter that conforms to the Midsummer specs--since
1831 // it's supposedly backwards compatible, it should work well...
1833 //#define LOG_BLITTER_MEMORY_ACCESSES
1835 #define DATINIT (false)
1836 #define TXTEXT (false)
1837 #define POLYGON (false)
1839 void BlitterMidsummer(uint32 cmd)
1841 uint32 outer_loop, inner_loop, a1_addr, a2_addr;
1842 int32 a1_x, a1_y, a2_x, a2_y, a1_width, a2_width;
1843 uint8 a1_phrase_mode, a2_phrase_mode;
1845 a1_addr = REG(A1_BASE) & 0xFFFFFFF8;
1846 a2_addr = REG(A2_BASE) & 0xFFFFFFF8;
1847 a1_x = (REG(A1_PIXEL) << 16) | (REG(A1_FPIXEL) & 0xFFFF);
1848 a1_y = (REG(A1_PIXEL) & 0xFFFF0000) | (REG(A1_FPIXEL) >> 16);
1849 UINT32 m = (REG(A1_FLAGS) >> 9) & 0x03, e = (REG(A1_FLAGS) >> 11) & 0x0F;
1850 a1_width = ((0x04 | m) << e) >> 2;//*/
1851 a2_x = (REG(A2_PIXEL) & 0x0000FFFF) << 16;
1852 a2_y = (REG(A2_PIXEL) & 0xFFFF0000);
1853 m = (REG(A2_FLAGS) >> 9) & 0x03, e = (REG(A2_FLAGS) >> 11) & 0x0F;
1854 a2_width = ((0x04 | m) << e) >> 2;//*/
1856 a1_phrase_mode = a2_phrase_mode = 0;
1858 if ((blitter_ram[A1_FLAGS + 1] & 0x03) == 0)
1861 if ((blitter_ram[A2_FLAGS + 1] & 0x03) == 0)
1864 #define INNER0 (inner_loop == 0)
1865 #define OUTER0 (outer_loop == 0)
1867 // $01800005 has SRCENX, may have to investigate further...
1868 // $00011008 has GOURD & DSTEN.
1869 // $41802F41 has SRCSHADE, CLIPA1
1870 /*bool logBlit = false;
1871 if (cmd != 0x00010200 && cmd != 0x01800001 && cmd != 0x01800005
1872 && cmd != 0x00011008 && cmd !=0x41802F41)
1878 uint64 srcData = GET64(blitter_ram, SRCDATA), srcXtraData,
1879 dstData = GET64(blitter_ram, DSTDATA), writeData;
1880 uint32 srcAddr, dstAddr;
1881 uint8 bitCount, a1PixelSize, a2PixelSize;
1883 // JTRM says phrase mode only works for 8BPP or higher, so let's try this...
1884 uint32 phraseOffset[8] = { 8, 8, 8, 8, 4, 2, 0, 0 };
1885 uint8 pixelShift[8] = { 3, 2, 1, 0, 1, 2, 0, 0 };
1887 a1PixelSize = (blitter_ram[A1_FLAGS + 3] >> 3) & 0x07;
1888 a2PixelSize = (blitter_ram[A2_FLAGS + 3] >> 3) & 0x07;
1890 outer_loop = GET16(blitter_ram, PIXLINECOUNTER + 0);
1892 if (outer_loop == 0)
1893 outer_loop = 0x10000;
1895 // We just list the states here and jump from state to state in order to
1896 // keep things somewhat clear. Optimization/cleanups later.
1898 //idle: // Blitter is idle, and will not perform any bus activity
1900 idle Blitter is off the bus, and no activity takes place.
1901 if GO if DATINIT goto init_if
1910 inner Inner loop is active, read and write cycles are performed
1912 inner: // Run inner loop state machine (asserts step from its idle state)
1913 inner_loop = GET16(blitter_ram, PIXLINECOUNTER + 2);
1915 if (inner_loop == 0)
1916 inner_loop = 0x10000;
1919 ------------------------------
1920 idle: Inactive, blitter is idle or passing round outer loop
1921 idle Another state in the outer loop is active. No bus transfers are performed.
1923 if SRCENX goto sreadx
1924 else if TXTEXT goto txtread
1925 else if SRCEN goto sread
1926 else if DSTEN goto dread
1927 else if DSTENZ goto dzread
1944 sreadx Extra source data read at the start of an inner loop pass.
1946 if SRCENZ goto szreadx
1947 else if TXTEXT goto txtread
1948 else if SRCEN goto sread
1949 else if DSTEN goto dread
1950 else if DSTENZ goto dzread
1953 sreadx: // Extra source data read
1968 szreadx Extra source Z read as the start of an inner loop pass.
1970 if TXTEXT goto txtread
1973 szreadx: // Extra source Z read
1980 txtread Read texture data from external memory. This state is only used for external texture.
1981 TEXTEXT is the condition TEXTMODE=1.
1984 else if DSTEN goto dread
1985 else if DSTENZ goto dzread
1988 txtread: // Read external texture data
1999 sread Source data read.
2001 if SRCENZ goto szread
2002 else if DSTEN goto dread
2003 else if DSTENZ goto dzread
2006 sread: // Source data read
2007 //The JTRM doesn't really specify the internal structure of the source data read, but I would
2008 //imagine that if it's in phrase mode that it starts by reading the phrase that the window is
2009 //pointing at. Likewise, the pixel (if in BPP 1, 2 & 4, chopped) otherwise. It probably still
2010 //transfers an entire phrase even in pixel mode.
2011 //Odd thought: Does it expand, e.g., 1 BPP pixels into 32 BPP internally? Hmm...
2014 a1_addr = REG(A1_BASE) & 0xFFFFFFF8;
2015 a2_addr = REG(A2_BASE) & 0xFFFFFFF8;
2016 a1_zoffs = (REG(A1_FLAGS) >> 6) & 7;
2017 a2_zoffs = (REG(A2_FLAGS) >> 6) & 7;
2018 xadd_a1_control = (REG(A1_FLAGS) >> 16) & 0x03;
2019 xadd_a2_control = (REG(A2_FLAGS) >> 16) & 0x03;
2020 a1_pitch = pitchValue[(REG(A1_FLAGS) & 0x03)];
2021 a2_pitch = pitchValue[(REG(A2_FLAGS) & 0x03)];
2022 n_pixels = REG(PIXLINECOUNTER) & 0xFFFF;
2023 n_lines = (REG(PIXLINECOUNTER) >> 16) & 0xFFFF;
2024 a1_x = (REG(A1_PIXEL) << 16) | (REG(A1_FPIXEL) & 0xFFFF);
2025 a1_y = (REG(A1_PIXEL) & 0xFFFF0000) | (REG(A1_FPIXEL) >> 16);
2026 a2_psize = 1 << ((REG(A2_FLAGS) >> 3) & 0x07);
2027 a1_psize = 1 << ((REG(A1_FLAGS) >> 3) & 0x07);
2030 a1_width = ((0x04 | m) << e) >> 2;
2031 a2_width = ((0x04 | m) << e) >> 2;
2033 // write values back to registers
2034 WREG(A1_PIXEL, (a1_y & 0xFFFF0000) | ((a1_x >> 16) & 0xFFFF));
2035 WREG(A1_FPIXEL, (a1_y << 16) | (a1_x & 0xFFFF));
2036 WREG(A2_PIXEL, (a2_y & 0xFFFF0000) | ((a2_x >> 16) & 0xFFFF));
2038 // Calculate the address to be read...
2040 //Need to fix phrase mode calcs here, since they should *step* by eight, not mulitply.
2041 //Also, need to fix various differing BPP modes here, since offset won't be correct except
2042 //for 8BPP. !!! FIX !!!
2043 srcAddr = (DSTA2 ? a1_addr : a2_addr);
2045 /* if ((DSTA2 ? a1_phrase_mode : a2_phrase_mode) == 1)
2047 srcAddr += (((DSTA2 ? a1_x : a2_x) >> 16)
2048 + (((DSTA2 ? a1_y : a2_y) >> 16) * (DSTA2 ? a1_width : a2_width)));
2052 // uint32 pixAddr = ((DSTA2 ? a1_x : a2_x) >> 16)
2053 // + (((DSTA2 ? a1_y : a2_y) >> 16) * (DSTA2 ? a1_width : a2_width));
2054 int32 pixAddr = (int16)((DSTA2 ? a1_x : a2_x) >> 16)
2055 + ((int16)((DSTA2 ? a1_y : a2_y) >> 16) * (DSTA2 ? a1_width : a2_width));
2057 if ((DSTA2 ? a1PixelSize : a2PixelSize) < 3)
2058 pixAddr >>= pixelShift[(DSTA2 ? a1PixelSize : a2PixelSize)];
2059 else if ((DSTA2 ? a1PixelSize : a2PixelSize) > 3)
2060 pixAddr <<= pixelShift[(DSTA2 ? a1PixelSize : a2PixelSize)];
2067 if ((DSTA2 ? a1_phrase_mode : a2_phrase_mode) == 1)
2069 srcData = ((uint64)JaguarReadLong(srcAddr, BLITTER) << 32)
2070 | (uint64)JaguarReadLong(srcAddr + 4, BLITTER);
2074 //1,2,&4BPP are wrong here... !!! FIX !!!
2075 if ((DSTA2 ? a1PixelSize : a2PixelSize) == 0) // 1 BPP
2076 srcData = JaguarReadByte(srcAddr, BLITTER);
2077 if ((DSTA2 ? a1PixelSize : a2PixelSize) == 1) // 2 BPP
2078 srcData = JaguarReadByte(srcAddr, BLITTER);
2079 if ((DSTA2 ? a1PixelSize : a2PixelSize) == 2) // 4 BPP
2080 srcData = JaguarReadByte(srcAddr, BLITTER);
2081 if ((DSTA2 ? a1PixelSize : a2PixelSize) == 3) // 8 BPP
2082 srcData = JaguarReadByte(srcAddr, BLITTER);
2083 if ((DSTA2 ? a1PixelSize : a2PixelSize) == 4) // 16 BPP
2084 srcData = JaguarReadWord(srcAddr, BLITTER);
2085 if ((DSTA2 ? a1PixelSize : a2PixelSize) == 5) // 32 BPP
2086 srcData = JaguarReadLong(srcAddr, BLITTER);
2089 #ifdef LOG_BLITTER_MEMORY_ACCESSES
2091 WriteLog("BLITTER: srcAddr=%08X, srcData=%08X %08X\n", srcAddr, (uint32)(srcData >> 32), (uint32)(srcData & 0xFFFFFFFF));
2103 szread: // Source Z read
2105 szread Source Z read.
2108 else if DSTENZ goto dzread
2118 dread: // Destination data read
2120 dread Destination data read.
2122 if DSTENZ goto dzread
2125 // Calculate the destination address to be read...
2127 //Need to fix phrase mode calcs here, since they should *step* by eight, not mulitply.
2128 //Also, need to fix various differing BPP modes here, since offset won't be correct except
2129 //for 8BPP. !!! FIX !!!
2130 dstAddr = (DSTA2 ? a2_addr : a1_addr);
2133 // uint32 pixAddr = ((DSTA2 ? a2_x : a1_x) >> 16)
2134 // + (((DSTA2 ? a2_y : a1_y) >> 16) * (DSTA2 ? a2_width : a1_width));
2135 int32 pixAddr = (int16)((DSTA2 ? a2_x : a1_x) >> 16)
2136 + ((int16)((DSTA2 ? a2_y : a1_y) >> 16) * (DSTA2 ? a2_width : a1_width));
2138 if ((DSTA2 ? a2PixelSize : a1PixelSize) < 3)
2139 pixAddr >>= pixelShift[(DSTA2 ? a2PixelSize : a1PixelSize)];
2140 else if ((DSTA2 ? a2PixelSize : a1PixelSize) > 3)
2141 pixAddr <<= pixelShift[(DSTA2 ? a2PixelSize : a1PixelSize)];
2148 if ((DSTA2 ? a2_phrase_mode : a1_phrase_mode) == 1)
2150 dstData = ((uint64)JaguarReadLong(srcAddr, BLITTER) << 32)
2151 | (uint64)JaguarReadLong(srcAddr + 4, BLITTER);
2155 //1,2,&4BPP are wrong here... !!! FIX !!!
2156 if ((DSTA2 ? a2PixelSize : a1PixelSize) == 0) // 1 BPP
2157 dstData = JaguarReadByte(dstAddr, BLITTER);
2158 if ((DSTA2 ? a2PixelSize : a1PixelSize) == 1) // 2 BPP
2159 dstData = JaguarReadByte(dstAddr, BLITTER);
2160 if ((DSTA2 ? a2PixelSize : a1PixelSize) == 2) // 4 BPP
2161 dstData = JaguarReadByte(dstAddr, BLITTER);
2162 if ((DSTA2 ? a2PixelSize : a1PixelSize) == 3) // 8 BPP
2163 dstData = JaguarReadByte(dstAddr, BLITTER);
2164 if ((DSTA2 ? a2PixelSize : a1PixelSize) == 4) // 16 BPP
2165 dstData = JaguarReadWord(dstAddr, BLITTER);
2166 if ((DSTA2 ? a2PixelSize : a1PixelSize) == 5) // 32 BPP
2167 dstData = JaguarReadLong(dstAddr, BLITTER);
2170 #ifdef LOG_BLITTER_MEMORY_ACCESSES
2172 WriteLog("BLITTER (dread): dstAddr=%08X, dstData=%08X %08X\n", dstAddr, (uint32)(dstData >> 32), (uint32)(dstData & 0xFFFFFFFF));
2180 dzread: // Destination Z read
2182 dzread Destination Z read.
2187 dwrite: // Destination data write
2189 dwrite Destination write. Every pass round the inner loop must go through this state..
2191 if DSTWRZ goto dzwrite
2192 else if INNER0 goto idle
2193 else if TXTEXT goto txtread
2194 else if SRCEN goto sread
2195 else if DSTEN goto dread
2196 else if DSTENZ goto dzread
2205 a1_xadd = 1.000000 (phrase=0)
2214 a2_xadd = 1.000000 (phrase=1)
2218 a2_mask_x= 0xFFFFFFFF
2219 a2_mask_y= 0xFFFFFFFF
2229 --LFUFUNC = LFU_CLEAR
2230 | PATDSEL = 1 (PD=77C7 7700 7700 7700)
2232 GOURD = 1 (II=00FC 1A00, SD=FF00 0000 0000 0000)
2235 //Still need to do CLIPA1 and SRCSHADE and GOURD and GOURZ...
2237 // Check clipping...
2241 uint16 x = a1_x >> 16, y = a1_y >> 16;
2243 if (x >= GET16(blitter_ram, A1_CLIP + 2) || y >= GET16(blitter_ram, A1_CLIP))
2247 // Figure out what gets written...
2251 writeData = GET64(blitter_ram, PATTERNDATA);
2252 //GOURD works properly only in 16BPP mode...
2253 //SRCDATA holds the intensity fractions...
2254 //Does GOURD get calc'ed here or somewhere else???
2255 //Temporary testing kludge...
2257 // writeData >>= 48;
2258 // writeData = 0xFF88;
2259 //OK, it's not writing an entire strip of pixels... Why?
2260 //bad incrementing, that's why!
2264 // Apparently this only works with 16-bit pixels. Not sure if it works in phrase mode either.
2265 //Also, take TOPBEN & TOPNEN into account here as well...
2266 writeData = srcData + dstData;
2268 else // LFUFUNC is the default...
2273 writeData |= ~srcData & ~dstData;
2275 writeData |= ~srcData & dstData;
2277 writeData |= srcData & ~dstData;
2279 writeData |= srcData & dstData;
2282 // Calculate the address to be written...
2284 dstAddr = (DSTA2 ? a2_addr : a1_addr);
2286 /* if ((DSTA2 ? a2_phrase_mode : a1_phrase_mode) == 1)
2288 //both of these calculate the wrong address because they don't take into account
2290 dstAddr += ((DSTA2 ? a2_x : a1_x) >> 16)
2291 + (((DSTA2 ? a2_y : a1_y) >> 16) * (DSTA2 ? a2_width : a1_width));
2295 /* dstAddr += ((DSTA2 ? a2_x : a1_x) >> 16)
2296 + (((DSTA2 ? a2_y : a1_y) >> 16) * (DSTA2 ? a2_width : a1_width));*/
2297 // uint32 pixAddr = ((DSTA2 ? a2_x : a1_x) >> 16)
2298 // + (((DSTA2 ? a2_y : a1_y) >> 16) * (DSTA2 ? a2_width : a1_width));
2299 int32 pixAddr = (int16)((DSTA2 ? a2_x : a1_x) >> 16)
2300 + ((int16)((DSTA2 ? a2_y : a1_y) >> 16) * (DSTA2 ? a2_width : a1_width));
2302 if ((DSTA2 ? a2PixelSize : a1PixelSize) < 3)
2303 pixAddr >>= pixelShift[(DSTA2 ? a2PixelSize : a1PixelSize)];
2304 else if ((DSTA2 ? a2PixelSize : a1PixelSize) > 3)
2305 pixAddr <<= pixelShift[(DSTA2 ? a2PixelSize : a1PixelSize)];
2312 if ((DSTA2 ? a2_phrase_mode : a1_phrase_mode) == 1)
2314 JaguarWriteLong(dstAddr, writeData >> 32, BLITTER);
2315 JaguarWriteLong(dstAddr + 4, writeData & 0xFFFFFFFF, BLITTER);
2319 //1,2,&4BPP are wrong here... !!! FIX !!!
2320 if ((DSTA2 ? a2PixelSize : a1PixelSize) == 0) // 1 BPP
2321 JaguarWriteByte(dstAddr, writeData, BLITTER);
2322 if ((DSTA2 ? a2PixelSize : a1PixelSize) == 1) // 2 BPP
2323 JaguarWriteByte(dstAddr, writeData, BLITTER);
2324 if ((DSTA2 ? a2PixelSize : a1PixelSize) == 2) // 4 BPP
2325 JaguarWriteByte(dstAddr, writeData, BLITTER);
2326 if ((DSTA2 ? a2PixelSize : a1PixelSize) == 3) // 8 BPP
2327 JaguarWriteByte(dstAddr, writeData, BLITTER);
2328 if ((DSTA2 ? a2PixelSize : a1PixelSize) == 4) // 16 BPP
2329 JaguarWriteWord(dstAddr, writeData, BLITTER);
2330 if ((DSTA2 ? a2PixelSize : a1PixelSize) == 5) // 32 BPP
2331 JaguarWriteLong(dstAddr, writeData, BLITTER);
2334 #ifdef LOG_BLITTER_MEMORY_ACCESSES
2336 WriteLog("BLITTER: dstAddr=%08X, writeData=%08X %08X\n", dstAddr, (uint32)(writeData >> 32), (uint32)(writeData & 0xFFFFFFFF));
2339 inhibitWrite://Should this go here? or on the other side of the X/Y incrementing?
2340 //Seems OK here... for now.
2342 // Do funky X/Y incrementation here as well... !!! FIX !!!
2344 // Handle A1 channel stepping
2346 if ((blitter_ram[A1_FLAGS + 1] & 0x03) == 0)
2347 a1_x += phraseOffset[a1PixelSize] << 16;
2348 else if ((blitter_ram[A1_FLAGS + 1] & 0x03) == 1)
2349 a1_x += (blitter_ram[A1_FLAGS + 1] & 0x08 ? -1 << 16 : 1 << 16);
2350 /* else if ((blitter_ram[A1_FLAGS + 1] & 0x03) == 2)
2352 else if ((blitter_ram[A1_FLAGS + 1] & 0x03) == 3)
2354 //Always add the FINC here??? That was the problem with the BIOS screen... So perhaps.
2355 a1_x += GET16(blitter_ram, A1_FINC + 2);
2356 a1_y += GET16(blitter_ram, A1_FINC + 0);
2358 a1_x += GET16(blitter_ram, A1_INC + 2) << 16;
2359 a1_y += GET16(blitter_ram, A1_INC + 0) << 16;
2362 if ((blitter_ram[A1_FLAGS + 1] & 0x04) && (blitter_ram[A1_FLAGS + 1] & 0x03 != 3))
2363 a1_y += (blitter_ram[A1_FLAGS + 1] & 0x10 ? -1 << 16 : 1 << 16);
2365 // Handle A2 channel stepping
2367 if ((blitter_ram[A2_FLAGS + 1] & 0x03) == 0)
2368 a2_x += phraseOffset[a2PixelSize] << 16;
2369 else if ((blitter_ram[A2_FLAGS + 1] & 0x03) == 1)
2370 a2_x += (blitter_ram[A2_FLAGS + 1] & 0x08 ? -1 << 16 : 1 << 16);
2371 /* else if ((blitter_ram[A2_FLAGS + 1] & 0x03) == 2)
2374 if (blitter_ram[A2_FLAGS + 1] & 0x04)
2375 a2_y += (blitter_ram[A2_FLAGS + 1] & 0x10 ? -1 << 16 : 1 << 16);
2377 //Need to fix this so that it subtracts (saturating, of course) the correct number of pixels
2378 //in phrase mode... !!! FIX !!! [DONE]
2379 //Need to fix this so that it counts down the correct item. Does it count the
2380 //source or the destination phrase mode???
2381 //It shouldn't matter, because we *should* end up processing the same amount
2382 //the same number of pixels... Not sure though.
2383 if ((DSTA2 ? a2_phrase_mode : a1_phrase_mode) == 1)
2385 if (inner_loop < phraseOffset[DSTA2 ? a2PixelSize : a1PixelSize])
2388 inner_loop -= phraseOffset[DSTA2 ? a2PixelSize : a1PixelSize];
2409 dzwrite: // Destination Z write
2411 dzwrite Destination Z write.
2414 else if TXTEXT goto txtread
2415 else if SRCEN goto sread
2416 else if DSTEN goto dread
2417 else if DSTENZ goto dzread
2434 ------------------------------
2435 if INDONE if OUTER0 goto idle
2436 else if UPDA1F goto a1fupdate
2437 else if UPDA1 goto a1update
2438 else if GOURZ.POLYGON goto zfupdate
2439 else if UPDA2 goto a2update
2440 else if DATINIT goto init_if
2453 //kill this, for now...
2454 // else if (GOURZ.POLYGON)
2463 a1fupdate: // Update A1 pointer fractions and more (see below)
2465 a1fupdate A1 step fraction is added to A1 pointer fraction
2466 POLYGON true: A1 step delta X and Y fraction parts are added to the A1
2467 step X and Y fraction parts (the value prior to this add is used for
2468 the step to pointer add).
2469 POLYGON true: inner count step fraction is added to the inner count
2471 POLYGON.GOURD true: the I fraction step is added to the computed
2472 intensity fraction parts +
2473 POLYGON.GOURD true: the I fraction step delta is added to the I
2478 #define A1_PIXEL ((UINT32)0x0C) // Integer part of the pixel (Y.i and X.i)
2479 #define A1_STEP ((UINT32)0x10) // Integer part of the step
2480 #define A1_FSTEP ((UINT32)0x14) // Fractional part of the step
2481 #define A1_FPIXEL ((UINT32)0x18) // Fractional part of the pixel (Y.f and X.f)
2484 // This is all kinda murky. All we have are the Midsummer docs to give us any guidance,
2485 // and it's incomplete or filled with errors (like above). Aarrrgggghhhhh!
2487 //This isn't right. Is it? I don't think the fractional parts are signed...
2488 // a1_x += (int32)((int16)GET16(blitter_ram, A1_FSTEP + 2));
2489 // a1_y += (int32)((int16)GET16(blitter_ram, A1_FSTEP + 0));
2490 a1_x += GET16(blitter_ram, A1_FSTEP + 2);
2491 a1_y += GET16(blitter_ram, A1_FSTEP + 0);
2495 a1update: // Update A1 pointer integers
2497 a1update A1 step is added to A1 pointer, with carry from the fractional add
2498 POLYGON true: A1 step delta X and Y integer parts are added to the A1
2499 step X and Y integer parts, with carry from the corresponding
2500 fractional part add (again, the value prior to this add is used for
2501 the step to pointer add).
2502 POLYGON true: inner count step is added to the inner count, with carry
2503 POLYGON.GOURD true: the I step is added to the computed intensities,
2505 POLYGON.GOURD true: the I step delta is added to the I step, with
2506 carry the texture X and Y step delta values are added to the X and Y
2508 if GOURZ.POLYGON goto zfupdate
2509 else if UPDA2 goto a2update
2510 else if DATINIT goto init_if
2513 a1_x += (int32)(GET16(blitter_ram, A1_STEP + 2) << 16);
2514 a1_y += (int32)(GET16(blitter_ram, A1_STEP + 0) << 16);
2517 //kill this, for now...
2518 // if (GOURZ.POLYGON)
2528 zfupdate: // Update computed Z step fractions
2530 zfupdate the Z fraction step is added to the computed Z fraction parts +
2531 the Z fraction step delta is added to the Z fraction step
2536 zupdate: // Update computed Z step integers
2538 zupdate the Z step is added to the computed Zs, with carry +
2539 the Z step delta is added to the Z step, with carry
2540 if UPDA2 goto a2update
2541 else if DATINIT goto init_if
2551 a2update: // Update A2 pointer
2553 a2update A2 step is added to the A2 pointer
2554 if DATINIT goto init_if
2557 a2_x += (int32)(GET16(blitter_ram, A2_STEP + 2) << 16);
2558 a2_y += (int32)(GET16(blitter_ram, A2_STEP + 0) << 16);
2566 init_if: // Initialise intensity fractions and texture X
2568 init_if Initialise the fractional part of the computed intensity fields, from
2569 the increment and step registers. The texture X integer and fractional
2570 parts can also be initialised.
2575 init_ii: // Initialise intensity integers and texture Y
2577 init_ii Initialise the integer part of the computed intensity, and texture Y
2578 integer and fractional parts
2579 if GOURZ goto init_zf
2587 init_zf: // Initialise Z fractions
2589 init_zf Initialise the fractional part of the computed Z fields.
2594 init_zi: // Initialise Z integers
2596 init_zi Initialise the integer part of the computed Z fields.
2603 The outer loop state machine fires off the inner loop, and controls the updating
2604 process between passes through the inner loop.
2606 + -- these functions are irrelevant if the DATINIT function is enabled, which it
2609 All these states will complete in one clock cycle, with the exception of the idle
2610 state, which means the blitter is quiescent; and the inner state, which takes as
2611 long as is required to complete one strip of pixels. It is therefore possible for
2612 the blitter to spend a maximum of nine clock cycles of inactivity between passes
2613 through the inner loop.
2623 // Here's attempt #2--taken from the Oberon chip specs!
2626 #ifdef USE_MIDSUMMER_BLITTER_MKII
2628 void ADDRGEN(uint32 &, uint32 &, bool, bool,
2629 uint16, uint16, uint32, uint8, uint8, uint8, uint8,
2630 uint16, uint16, uint32, uint8, uint8, uint8, uint8);
2631 void ADDARRAY(uint16 * addq, uint8 daddasel, uint8 daddbsel, uint8 daddmode,
2632 uint64 dstd, uint32 iinc, uint8 initcin[], uint64 initinc, uint16 initpix,
2633 uint32 istep, uint64 patd, uint64 srcd, uint64 srcz1, uint64 srcz2,
2634 uint32 zinc, uint32 zstep);
2635 void ADD16SAT(uint16 &r, uint8 &co, uint16 a, uint16 b, uint8 cin, bool sat, bool eightbit, bool hicinh);
2636 void ADDAMUX(int16 &adda_x, int16 &adda_y, uint8 addasel, int16 a1_step_x, int16 a1_step_y,
2637 int16 a1_stepf_x, int16 a1_stepf_y, int16 a2_step_x, int16 a2_step_y,
2638 int16 a1_inc_x, int16 a1_inc_y, int16 a1_incf_x, int16 a1_incf_y, uint8 adda_xconst,
2639 bool adda_yconst, bool addareg, bool suba_x, bool suba_y);
2640 void ADDBMUX(int16 &addb_x, int16 &addb_y, uint8 addbsel, int16 a1_x, int16 a1_y,
2641 int16 a2_x, int16 a2_y, int16 a1_frac_x, int16 a1_frac_y);
2642 void DATAMUX(int16 &data_x, int16 &data_y, uint32 gpu_din, int16 addq_x, int16 addq_y, bool addqsel);
2643 void ADDRADD(int16 &addq_x, int16 &addq_y, bool a1fracldi,
2644 uint16 adda_x, uint16 adda_y, uint16 addb_x, uint16 addb_y, uint8 modx, bool suba_x, bool suba_y);
2645 void DATA(uint64 &wdata, uint8 &dcomp, uint8 &zcomp, bool &nowrite,
2646 bool big_pix, bool cmpdst, uint8 daddasel, uint8 daddbsel, uint8 daddmode, bool daddq_sel, uint8 data_sel,
2647 uint8 dbinh, uint8 dend, uint8 dstart, uint64 dstd, uint32 iinc, uint8 lfu_func, uint64 &patd, bool patdadd,
2648 bool phrase_mode, uint64 srcd, bool srcdread, bool srczread, bool srcz2add, uint8 zmode,
2649 bool bcompen, bool bkgwren, bool dcompen, uint8 icount, uint8 pixsize,
2650 uint64 &srcz, uint64 dstz, uint32 zinc);
2651 void COMP_CTRL(uint8 &dbinh, bool &nowrite,
2652 bool bcompen, bool big_pix, bool bkgwren, uint8 dcomp, bool dcompen, uint8 icount,
2653 uint8 pixsize, bool phrase_mode, uint8 srcd, uint8 zcomp);
2654 #define VERBOSE_BLITTER_LOGGING
2655 bool logBlit = false;
2657 void BlitterMidsummer2(void)
2659 // Here's what the specs say the state machine does. Note that this can probably be
2660 // greatly simplified (also, it's different from what John has in his Oberon docs):
2661 //Will remove stuff that isn't in Jaguar I once fully described (stuff like texture won't
2662 //be described here at all)...
2664 uint32 cmd = GET32(blitter_ram, COMMAND);
2668 cmd != 0x00010200 && // PATDSEL
2670 && cmd != 0x01800005
2671 //Boot ROM ATARI letters:
2672 && cmd != 0x00011008 // DSTEN GOURD PATDSEL
2673 //Boot ROM spinning cube:
2674 && cmd != 0x41802F41 // SRCEN CLIP_A1 UPDA1 UPDA1F UPDA2 DSTA2 GOURZ ZMODE=0 LFUFUNC=C SRCSHADE
2676 && cmd != 0x01800E01 // SRCEN UPDA1 UPDA2 DSTA2 LFUFUNC=C
2677 //T2K TEMPEST letters:
2678 && cmd != 0x09800741 // SRCEN CLIP_A1 UPDA1 UPDA1F UPDA2 LFUFUNC=C DCOMPEN
2679 //Static letters on Cybermorph intro screen:
2680 && cmd != 0x09800609 // SRCEN DSTEN UPDA1 UPDA2 LFUFUNC=C DCOMPEN
2681 //Static pic on title screen:
2682 && cmd != 0x01800601 // SRCEN UPDA1 UPDA2 LFUFUNC=C
2683 //Turning letters on Cybermorph intro screen:
2684 && cmd != 0x09800F41 // SRCEN CLIP_A1 UPDA1 UPDA1F UPDA2 DSTA2 LFUFUNC=C DCOMPEN
2685 && cmd != 0x00113078 // DSTEN DSTENZ DSTWRZ CLIP_A1 GOURD GOURZ PATDSEL ZMODE=4
2686 && cmd != 0x09900F39 // SRCEN DSTEN DSTENZ DSTWRZ UPDA1 UPDA1F UPDA2 DSTA2 ZMODE=4 LFUFUNC=C DCOMPEN
2687 && cmd != 0x09800209 // SRCEN DSTEN UPDA1 LFUFUNC=C DCOMPEN
2688 && cmd != 0x00011200 // UPDA1 GOURD PATDSEL
2689 //Start of Hover Strike (clearing screen):
2690 && cmd != 0x00010000 // PATDSEL
2691 //Hover Strike text:
2692 && cmd != 0x1401060C // SRCENX DSTEN UPDA1 UPDA2 PATDSEL BCOMPEN BKGWREN
2693 //Hover Strike 3D stuff
2694 // && cmd != 0x01902839 // SRCEN DSTEN DSTENZ DSTWRZ DSTA2 GOURZ ZMODE=4 LFUFUNC=C
2695 //Hover Strike darkening on intro to play (briefing) screen
2696 && cmd != 0x00020208 // DSTEN UPDA1 ADDDSEL
2697 //Trevor McFur stuff:
2698 && cmd != 0x05810601 // SRCEN UPDA1 UPDA2 PATDSEL BCOMPEN
2699 && cmd != 0x01800201 // SRCEN UPDA1 LFUFUNC=C
2701 && cmd != 0x00011000 // GOURD PATDSEL
2702 && cmd != 0x00011040 // CLIP_A1 GOURD PATDSEL
2706 if (blit_start_log == 0) // Wait for the signal...
2707 logBlit = false;//*/
2709 Some T2K unique blits:
2710 logBlit = F, cmd = 00010200 *
2711 logBlit = F, cmd = 00011000
2712 logBlit = F, cmd = 00011040
2713 logBlit = F, cmd = 01800005 *
2714 logBlit = F, cmd = 09800741 *
2716 Hover Strike mission selection screen:
2717 Blit! (CMD = 01902839) // SRCEN DSTEN DSTENZ DSTWRZ DSTA2 GOURZ ZMODE=4 LFUFUNC=C
2719 Checkered Flag blits in the screw up zone:
2720 Blit! (CMD = 01800001) // SRCEN LFUFUNC=C
2721 Blit! (CMD = 01800000) // LFUFUNC=C
2722 Blit! (CMD = 00010000) // PATDSEL
2724 Wolfenstein 3D in the fuckup zone:
2725 Blit! (CMD = 01800000) // LFUFUNC=C
2728 //printf("logBlit = %s, cmd = %08X\n", (logBlit ? "T" : "F"), cmd);
2733 Blit! (CMD = 00011040)
2734 Flags: CLIP_A1 GOURD PATDSEL
2736 a1_base = 00100000, a2_base = 0081F6A8
2737 a1_x = 00A7, a1_y = 0014, a1_frac_x = 0000, a1_frac_y = 0000, a2_x = 0001, a2_y = 0000
2738 a1_step_x = FE80, a1_step_y = 0001, a1_stepf_x = 0000, a1_stepf_y = 0000, a2_step_x = FFF8, a2_step_y = 0001
2739 a1_inc_x = 0001, a1_inc_y = 0000, a1_incf_x = 0000, a1_incf_y = 0000
2740 a1_win_x = 0180, a1_win_y = 0118, a2_mask_x = 0000, a2_mask_y = 0000
2741 a2_mask=F a1add=+phr/+0 a2add=+phr/+0
2742 a1_pixsize = 4, a2_pixsize = 4
2746 if (cmd == 0x00011040
2747 && (GET16(blitter_ram, A1_PIXEL + 2) == 0x00A7) && (GET16(blitter_ram, A1_PIXEL + 0) == 0x0014)
2748 && (GET16(blitter_ram, A2_PIXEL + 2) == 0x0001) && (GET16(blitter_ram, A2_PIXEL + 0) == 0x0000)
2749 && (GET16(blitter_ram, PIXLINECOUNTER + 2) == 18))
2752 // Line states passed in via the command register
2754 bool srcen = (SRCEN), srcenx = (SRCENX), srcenz = (SRCENZ),
2755 dsten = (DSTEN), dstenz = (DSTENZ), dstwrz = (DSTWRZ), clip_a1 = (CLIPA1),
2756 upda1 = (UPDA1), upda1f = (UPDA1F), upda2 = (UPDA2), dsta2 = (DSTA2),
2757 gourd = (GOURD), gourz = (GOURZ), topben = (TOPBEN), topnen = (TOPNEN),
2758 patdsel = (PATDSEL), adddsel = (ADDDSEL), cmpdst = (CMPDST), bcompen = (BCOMPEN),
2759 dcompen = (DCOMPEN), bkgwren = (BKGWREN), srcshade = (SRCSHADE);
2761 uint8 zmode = (cmd & 0x01C0000) >> 18, lfufunc = (cmd & 0x1E00000) >> 21;
2763 //Where to find various lines:
2765 // gourd -> dcontrol, inner, outer, state
2766 // gourz -> dcontrol, inner, outer, state
2767 // cmpdst -> blit, data, datacomp, state
2768 // bcompen -> acontrol, inner, mcontrol, state
2769 // dcompen -> inner, state
2770 // bkgwren -> inner, state
2771 // srcshade -> dcontrol, inner, state
2772 // adddsel -> dcontrol
2773 //NOTE: ADDDSEL takes precedence over PATDSEL, PATDSEL over LFU_FUNC
2774 #ifdef VERBOSE_BLITTER_LOGGING
2777 char zfs[512], lfus[512];
2778 zfs[0] = lfus[0] = 0;
2779 if (dstwrz || dstenz || gourz)
2780 sprintf(zfs, " ZMODE=%X", zmode);
2781 if (!(patdsel || adddsel))
2782 sprintf(lfus, " LFUFUNC=%X", lfufunc);
2783 printf("\nBlit! (CMD = %08X)\nFlags:%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n", cmd,
2784 (srcen ? " SRCEN" : ""), (srcenx ? " SRCENX" : ""), (srcenz ? " SRCENZ" : ""),
2785 (dsten ? " DSTEN" : ""), (dstenz ? " DSTENZ" : ""), (dstwrz ? " DSTWRZ" : ""),
2786 (clip_a1 ? " CLIP_A1" : ""), (upda1 ? " UPDA1" : ""), (upda1f ? " UPDA1F" : ""),
2787 (upda2 ? " UPDA2" : ""), (dsta2 ? " DSTA2" : ""), (gourd ? " GOURD" : ""),
2788 (gourz ? " GOURZ" : ""), (topben ? " TOPBEN" : ""), (topnen ? " TOPNEN" : ""),
2789 (patdsel ? " PATDSEL" : ""), (adddsel ? " ADDDSEL" : ""), zfs, lfus, (cmpdst ? " CMPDST" : ""),
2790 (bcompen ? " BCOMPEN" : ""), (dcompen ? " DCOMPEN" : ""), (bkgwren ? " BKGWREN" : ""),
2791 (srcshade ? " SRCSHADE" : ""));
2792 printf(" count = %d x %d\n", GET16(blitter_ram, PIXLINECOUNTER + 2), GET16(blitter_ram, PIXLINECOUNTER));
2797 // Lines that don't exist in Jaguar I (and will never be asserted)
2799 bool polygon = false, datinit = false, a1_stepld = false, a2_stepld = false, ext_int = false;
2800 bool istepadd = false, istepfadd = false, finneradd = false, inneradd = false;
2801 bool zstepfadd = false, zstepadd = false;
2803 // Various state lines (initial state--basically the reset state of the FDSYNCs)
2805 bool go = true, idle = true, inner = false, a1fupdate = false, a1update = false,
2806 zfupdate = false, zupdate = false, a2update = false, init_if = false, init_ii = false,
2807 init_zf = false, init_zi = false;
2809 bool outer0 = false, indone = false;
2811 bool idlei, inneri, a1fupdatei, a1updatei, zfupdatei, zupdatei, a2updatei, init_ifi, init_iii,
2814 bool notgzandp = !(gourz && polygon);
2816 // Various registers set up by user
2818 uint16 ocount = GET16(blitter_ram, PIXLINECOUNTER);
2819 uint8 a1_pitch = blitter_ram[A1_FLAGS + 3] & 0x03;
2820 uint8 a2_pitch = blitter_ram[A2_FLAGS + 3] & 0x03;
2821 uint8 a1_pixsize = (blitter_ram[A1_FLAGS + 3] & 0x38) >> 3;
2822 uint8 a2_pixsize = (blitter_ram[A2_FLAGS + 3] & 0x38) >> 3;
2823 uint8 a1_zoffset = (GET16(blitter_ram, A1_FLAGS + 2) >> 6) & 0x07;
2824 uint8 a2_zoffset = (GET16(blitter_ram, A2_FLAGS + 2) >> 6) & 0x07;
2825 uint8 a1_width = (blitter_ram[A1_FLAGS + 2] >> 1) & 0x3F;
2826 uint8 a2_width = (blitter_ram[A2_FLAGS + 2] >> 1) & 0x3F;
2827 bool a2_mask = blitter_ram[A2_FLAGS + 2] & 0x80;
2828 uint8 a1addx = blitter_ram[A1_FLAGS + 1] & 0x03, a2addx = blitter_ram[A2_FLAGS + 1] & 0x03;
2829 bool a1addy = blitter_ram[A1_FLAGS + 1] & 0x04, a2addy = blitter_ram[A2_FLAGS + 1] & 0x04;
2830 bool a1xsign = blitter_ram[A1_FLAGS + 1] & 0x08, a2xsign = blitter_ram[A2_FLAGS + 1] & 0x08;
2831 bool a1ysign = blitter_ram[A1_FLAGS + 1] & 0x10, a2ysign = blitter_ram[A2_FLAGS + 1] & 0x10;
2832 uint32 a1_base = GET32(blitter_ram, A1_BASE) & 0xFFFFFFF8; // Phrase aligned by ignoring bottom 3 bits
2833 uint32 a2_base = GET32(blitter_ram, A2_BASE) & 0xFFFFFFF8;
2835 uint16 a1_win_x = GET16(blitter_ram, A1_CLIP + 2) & 0x7FFF;
2836 uint16 a1_win_y = GET16(blitter_ram, A1_CLIP + 0) & 0x7FFF;
2837 int16 a1_x = (int16)GET16(blitter_ram, A1_PIXEL + 2);
2838 int16 a1_y = (int16)GET16(blitter_ram, A1_PIXEL + 0);
2839 int16 a1_step_x = (int16)GET16(blitter_ram, A1_STEP + 2);
2840 int16 a1_step_y = (int16)GET16(blitter_ram, A1_STEP + 0);
2841 uint16 a1_stepf_x = GET16(blitter_ram, A1_FSTEP + 2);
2842 uint16 a1_stepf_y = GET16(blitter_ram, A1_FSTEP + 0);
2843 uint16 a1_frac_x = GET16(blitter_ram, A1_FPIXEL + 2);
2844 uint16 a1_frac_y = GET16(blitter_ram, A1_FPIXEL + 0);
2845 int16 a1_inc_x = (int16)GET16(blitter_ram, A1_INC + 2);
2846 int16 a1_inc_y = (int16)GET16(blitter_ram, A1_INC + 0);
2847 uint16 a1_incf_x = GET16(blitter_ram, A1_FINC + 2);
2848 uint16 a1_incf_y = GET16(blitter_ram, A1_FINC + 0);
2850 int16 a2_x = (int16)GET16(blitter_ram, A2_PIXEL + 2);
2851 int16 a2_y = (int16)GET16(blitter_ram, A2_PIXEL + 0);
2852 uint16 a2_mask_x = GET16(blitter_ram, A2_MASK + 2);
2853 uint16 a2_mask_y = GET16(blitter_ram, A2_MASK + 0);
2854 int16 a2_step_x = (int16)GET16(blitter_ram, A2_STEP + 2);
2855 int16 a2_step_y = (int16)GET16(blitter_ram, A2_STEP + 0);
2857 uint64 srcd1 = GET64(blitter_ram, SRCDATA);
2859 uint64 dstd = GET64(blitter_ram, DSTDATA);
2860 uint64 patd = GET64(blitter_ram, PATTERNDATA);
2861 uint32 iinc = GET32(blitter_ram, INTENSITYINC);
2862 uint64 srcz1 = GET64(blitter_ram, SRCZINT);
2863 uint64 srcz2 = GET64(blitter_ram, SRCZFRAC);
2864 uint64 dstz = GET64(blitter_ram, DSTZ);
2865 uint32 zinc = GET32(blitter_ram, ZINC);
2866 uint32 collision = GET32(blitter_ram, COLLISIONCTRL);// 0=RESUME, 1=ABORT, 2=STOPEN
2868 uint8 pixsize = (dsta2 ? a2_pixsize : a1_pixsize); // From ACONTROL
2870 //Testing Trevor McFur--I *think* it's the circle on the lower RHS of the screen...
2872 if (cmd == 0x05810601 && (GET16(blitter_ram, PIXLINECOUNTER + 2) == 96)
2873 && (GET16(blitter_ram, PIXLINECOUNTER + 0) == 72))
2876 //if (cmd == 0x1401060C) patd = 0xFFFFFFFFFFFFFFFFLL;
2877 //if (cmd == 0x1401060C) patd = 0x00000000000000FFLL;
2878 //If it's still not working (bcompen-patd) then see who's writing what to patd and where...
2879 //Still not OK. Check to see who's writing what to where in patd!
2880 //It looks like M68K is writing to the top half of patd... Hmm...
2882 ----> M68K wrote 0000 to byte 15737344 of PATTERNDATA...
2883 --> M68K wrote 00 to byte 0 of PATTERNDATA...
2884 --> M68K wrote 00 to byte 1 of PATTERNDATA...
2885 ----> M68K wrote 00FF to byte 15737346 of PATTERNDATA...
2886 --> M68K wrote 00 to byte 2 of PATTERNDATA...
2887 --> M68K wrote FF to byte 3 of PATTERNDATA...
2888 logBlit = F, cmd = 1401060C
2890 Wren0 := ND6 (wren\[0], gpua\[5], gpua\[6..8], bliten, gpu_memw);
2891 Wren1 := ND6 (wren\[1], gpua[5], gpua\[6..8], bliten, gpu_memw);
2892 Wren2 := ND6 (wren\[2], gpua\[5], gpua[6], gpua\[7..8], bliten, gpu_memw);
2893 Wren3 := ND6 (wren\[3], gpua[5], gpua[6], gpua\[7..8], bliten, gpu_memw);
2896 Dec0 := D38GH (a1baseld, a1flagld, a1winld, a1ptrld, a1stepld, a1stepfld, a1fracld, a1incld, gpua[2..4], wren\[0]);
2898 Dec1 := D38GH (a1incfld, a2baseld, a2flagld, a2maskld, a2ptrldg, a2stepld, cmdldt, countldt, gpua[2..4], wren\[1]);
2900 Dec2 := D38GH (srcd1ldg[0..1], dstdldg[0..1], dstzldg[0..1], srcz1ldg[0..1], gpua[2..4], wren\[2]);
2902 Dec3 := D38GH (srcz2ld[0..1], patdld[0..1], iincld, zincld, stopld, intld[0], gpua[2..4], wren\[3]);
2904 wren[3] is asserted when gpu address bus = 0 011x xx00
2905 patdld[0] -> 0 0110 1000 -> $F02268 (lo 32 bits)
2906 patdld[1] -> 0 0110 1100 -> $F0226C (hi 32 bits)
2908 So... It's reversed! The data organization of the patd register is [low 32][high 32]! !!! FIX !!! [DONE]
2909 And fix all the other 64 bit registers [DONE]
2911 /*if (cmd == 0x1401060C)
2913 printf("logBlit = %s, cmd = %08X\n", (logBlit ? "T" : "F"), cmd);
2917 if ((cmd == 0x00010200) && (GET16(blitter_ram, PIXLINECOUNTER + 2) == 9))
2920 ; Pink altimeter bar
2922 Blit! (00110000 <- 000BF010) count: 9 x 23, A1/2_FLAGS: 000042E2/00010020 [cmd: 00010200]
2923 CMD -> src: dst: misc: a1ctl: UPDA1 mode: ity: PATDSEL z-op: op: LFU_CLEAR ctrl:
2924 A1 step values: -10 (X), 1 (Y)
2925 A1 -> pitch: 4 phrases, depth: 16bpp, z-off: 3, width: 320 (21), addctl: XADDPHR YADD0 XSIGNADD YSIGNADD
2926 A2 -> pitch: 1 phrases, depth: 16bpp, z-off: 0, width: 1 (00), addctl: XADDPIX YADD0 XSIGNADD YSIGNADD
2927 A1 x/y: 262/132, A2 x/y: 129/0
2928 ;x-coord is 257 in pic, so add 5
2929 ;20 for ship, 33 for #... Let's see if we can find 'em!
2931 ; Black altimeter bar
2933 Blit! (00110000 <- 000BF010) count: 5 x 29, A1/2_FLAGS: 000042E2/00010020 [cmd: 00010200]
2934 CMD -> src: dst: misc: a1ctl: UPDA1 mode: ity: PATDSEL z-op: op: LFU_CLEAR ctrl:
2935 A1 step values: -8 (X), 1 (Y)
2936 A1 -> pitch: 4 phrases, depth: 16bpp, z-off: 3, width: 320 (21), addctl: XADDPHR YADD0 XSIGNADD YSIGNADD
2937 A2 -> pitch: 1 phrases, depth: 16bpp, z-off: 0, width: 1 (00), addctl: XADDPIX YADD0 XSIGNADD YSIGNADD
2938 A1 x/y: 264/126, A2 x/y: 336/0
2940 Here's the pink bar--note that it's phrase mode without dread, so how does this work???
2941 Not sure, but I *think* that somehow it MUXes the data at the write site in on the left or right side
2942 of the write data when masked in phrase mode. I'll have to do some tracing to see if this is the mechanism
2945 Blit! (CMD = 00010200)
2946 Flags: UPDA1 PATDSEL
2948 a1_base = 00110010, a2_base = 000BD7E0
2949 a1_x = 0106, a1_y = 0090, a1_frac_x = 0000, a1_frac_y = 8000, a2_x = 025A, a2_y = 0000
2950 a1_step_x = FFF6, a1_step_y = 0001, a1_stepf_x = 5E00, a1_stepf_y = D100, a2_step_x = FFF7, a2_step_y = 0001
2951 a1_inc_x = 0001, a1_inc_y = FFFF, a1_incf_x = 0000, a1_incf_y = E000
2952 a1_win_x = 0000, a1_win_y = 0000, a2_mask_x = 0000, a2_mask_y = 0000
2953 a2_mask=F a1add=+phr/+0 a2add=+1/+0
2954 a1_pixsize = 4, a2_pixsize = 4
2955 srcd=BAC673AC2C92E578 dstd=0000000000000000 patd=74C074C074C074C0 iinc=0002E398
2956 srcz1=7E127E12000088DA srcz2=DBE06DF000000000 dstz=0000000000000000 zinc=FFFE4840, coll=0
2958 [in=T a1f=F a1=F zf=F z=F a2=F iif=F iii=F izf=F izi=F]
2959 Entering INNER state...
2960 Entering DWRITE state...
2961 Dest write address/pix address: 0016A830/0 [dstart=20 dend=40 pwidth=8 srcshift=0][daas=0 dabs=0 dam=7 ds=0 daq=F] [7400000074C074C0] (icount=0007, inc=2)
2962 Entering A1_ADD state [a1_x=0106, a1_y=0090, addasel=0, addbsel=0, modx=2, addareg=F, adda_xconst=2, adda_yconst=0]...
2963 Entering DWRITE state...
2964 Dest write address/pix address: 0016A850/0 [dstart=0 dend=40 pwidth=8 srcshift=0][daas=0 dabs=0 dam=7 ds=0 daq=F] [74C074C074C074C0] (icount=0003, inc=4)
2965 Entering A1_ADD state [a1_x=0108, a1_y=0090, addasel=0, addbsel=0, modx=2, addareg=F, adda_xconst=2, adda_yconst=0]...
2966 Entering DWRITE state...
2967 Dest write address/pix address: 0016A870/0 [dstart=0 dend=30 pwidth=8 srcshift=0][daas=0 dabs=0 dam=7 ds=0 daq=F] [74C074C074C00000] (icount=FFFF, inc=4)
2968 Entering A1_ADD state [a1_x=010C, a1_y=0090, addasel=0, addbsel=0, modx=2, addareg=F, adda_xconst=2, adda_yconst=0]...
2969 Entering IDLE_INNER state...
2970 Leaving INNER state... (ocount=000A)
2971 [in=F a1f=F a1=T zf=F z=F a2=F iif=F iii=F izf=F izi=F]
2972 Entering A1UPDATE state... (272/144 -> 262/145)
2973 [in=T a1f=F a1=F zf=F z=F a2=F iif=F iii=F izf=F izi=F]
2974 Entering INNER state...
2979 a2addy = a1addy; // A2 channel Y add bit is tied to A1's
2981 //if (logBlit && (ocount > 20)) logBlit = false;
2982 #ifdef VERBOSE_BLITTER_LOGGING
2985 printf(" a1_base = %08X, a2_base = %08X\n", a1_base, a2_base);
2986 printf(" a1_x = %04X, a1_y = %04X, a1_frac_x = %04X, a1_frac_y = %04X, a2_x = %04X, a2_y = %04X\n", (uint16)a1_x, (uint16)a1_y, a1_frac_x, a1_frac_y, (uint16)a2_x, (uint16)a2_y);
2987 printf(" a1_step_x = %04X, a1_step_y = %04X, a1_stepf_x = %04X, a1_stepf_y = %04X, a2_step_x = %04X, a2_step_y = %04X\n", (uint16)a1_step_x, (uint16)a1_step_y, a1_stepf_x, a1_stepf_y, (uint16)a2_step_x, (uint16)a2_step_y);
2988 printf(" a1_inc_x = %04X, a1_inc_y = %04X, a1_incf_x = %04X, a1_incf_y = %04X\n", (uint16)a1_inc_x, (uint16)a1_inc_y, a1_incf_x, a1_incf_y);
2989 printf(" a1_win_x = %04X, a1_win_y = %04X, a2_mask_x = %04X, a2_mask_y = %04X\n", a1_win_x, a1_win_y, a2_mask_x, a2_mask_y);
2990 char x_add_str[4][4] = { "phr", "1", "0", "inc" };
2991 printf(" a2_mask=%s a1add=%s%s/%s%s a2add=%s%s/%s%s\n", (a2_mask ? "T" : "F"), (a1xsign ? "-" : "+"), x_add_str[a1addx],
2992 (a1ysign ? "-" : "+"), (a1addy ? "1" : "0"), (a2xsign ? "-" : "+"), x_add_str[a2addx],
2993 (a2ysign ? "-" : "+"), (a2addy ? "1" : "0"));
2994 printf(" a1_pixsize = %u, a2_pixsize = %u\n", a1_pixsize, a2_pixsize);
2995 printf(" srcd=%08X%08X dstd=%08X%08X patd=%08X%08X iinc=%08X\n",
2996 (uint32)(srcd1 >> 32), (uint32)(srcd1 & 0xFFFFFFFF),
2997 (uint32)(dstd >> 32), (uint32)(dstd & 0xFFFFFFFF),
2998 (uint32)(patd >> 32), (uint32)(patd & 0xFFFFFFFF), iinc);
2999 printf(" srcz1=%08X%08X srcz2=%08X%08X dstz=%08X%08X zinc=%08X, coll=%X\n",
3000 (uint32)(srcz1 >> 32), (uint32)(srcz1 & 0xFFFFFFFF),
3001 (uint32)(srcz2 >> 32), (uint32)(srcz2 & 0xFFFFFFFF),
3002 (uint32)(dstz >> 32), (uint32)(dstz & 0xFFFFFFFF), zinc, collision);
3006 // Various state lines set up by user
3008 bool phrase_mode = ((!dsta2 && a1addx == 0) || (dsta2 && a2addx == 0) ? true : false); // From ACONTROL
3009 #ifdef VERBOSE_BLITTER_LOGGING
3012 printf(" Phrase mode is %s\n", (phrase_mode ? "ON" : "off"));
3018 // Stopgap vars to simulate various lines
3020 uint16 a1FracCInX = 0, a1FracCInY = 0;
3026 if ((idle && !go) || (inner && outer0 && indone))
3028 #ifdef VERBOSE_BLITTER_LOGGING
3031 printf(" Entering IDLE state...\n");
3037 //Instead of a return, let's try breaking out of the loop...
3044 // INNER LOOP ACTIVE
3046 Entering DWRITE state... (icount=0000, inc=4)
3047 Entering IDLE_INNER state...
3048 Leaving INNER state... (ocount=00EF)
3049 [in=T a1f=F a1=T zf=F z=F a2=F iif=F iii=F izf=F izi=F]
3050 Entering INNER state...
3052 [in=F a1f=F a1=F zf=F z=F a2=F iif=F iii=F izf=F izi=F]
3055 if ((idle && go && !datinit)
3056 || (inner && !indone)
3057 || (inner && indone && !outer0 && !upda1f && !upda1 && notgzandp && !upda2 && !datinit)
3058 || (a1update && !upda2 && notgzandp && !datinit)
3059 || (zupdate && !upda2 && !datinit)
3060 || (a2update && !datinit)
3061 || (init_ii && !gourz)
3069 // A1 FRACTION UPDATE
3071 if (inner && indone && !outer0 && upda1f)
3078 // A1 POINTER UPDATE
3081 || (inner && indone && !outer0 && !upda1f && upda1))
3088 // Z FRACTION UPDATE
3090 if ((a1update && gourz && polygon)
3091 || (inner && indone && !outer0 && !upda1f && !upda1 && gourz && polygon))
3107 // A2 POINTER UPDATE
3109 if ((a1update && upda2 && notgzandp)
3110 || (zupdate && upda2)
3111 || (inner && indone && !outer0 && !upda1f && notgzandp && !upda1 && upda2))
3118 // INITIALIZE INTENSITY FRACTION
3120 if ((zupdate && !upda2 && datinit)
3121 || (a1update && !upda2 && datinit && notgzandp)
3122 || (inner && indone && !outer0 && !upda1f && !upda1 && notgzandp && !upda2 && datinit)
3123 || (a2update && datinit)
3124 || (idle && go && datinit))
3131 // INITIALIZE INTENSITY INTEGER
3140 // INITIALIZE Z FRACTION
3142 if (init_ii && gourz)
3149 // INITIALIZE Z INTEGER
3158 // Here we move the fooi into their foo counterparts in order to simulate the moving
3159 // of data into the various FDSYNCs... Each time we loop we simulate one clock cycle...
3163 a1fupdate = a1fupdatei;
3164 a1update = a1updatei;
3165 zfupdate = zfupdatei; // *
3166 zupdate = zupdatei; // *
3167 a2update = a2updatei;
3168 init_if = init_ifi; // *
3169 init_ii = init_iii; // *
3170 init_zf = init_zfi; // *
3171 init_zi = init_zii; // *
3172 // * denotes states that will never assert for Jaguar I
3173 #ifdef VERBOSE_BLITTER_LOGGING
3176 printf(" [in=%c a1f=%c a1=%c zf=%c z=%c a2=%c iif=%c iii=%c izf=%c izi=%c]\n",
3177 (inner ? 'T' : 'F'), (a1fupdate ? 'T' : 'F'), (a1update ? 'T' : 'F'), (zfupdate ? 'T' : 'F'),
3178 (zupdate ? 'T' : 'F'), (a2update ? 'T' : 'F'), (init_if ? 'T' : 'F'), (init_ii ? 'T' : 'F'),
3179 (init_zf ? 'T' : 'F'), (init_zi ? 'T' : 'F'));
3184 // Now, depending on how we want to handle things, we could either put the implementation
3185 // of the various pieces up above, or handle them down below here.
3187 // Let's try postprocessing for now...
3192 #ifdef VERBOSE_BLITTER_LOGGING
3195 printf(" Entering INNER state...\n");
3199 uint16 icount = GET16(blitter_ram, PIXLINECOUNTER + 2);
3200 bool idle_inner = true, step = true, sreadx = false, szreadx = false, sread = false,
3201 szread = false, dread = false, dzread = false, dwrite = false, dzwrite = false;
3202 bool inner0 = false;
3203 bool idle_inneri, sreadxi, szreadxi, sreadi, szreadi, dreadi, dzreadi, dwritei, dzwritei;
3205 // State lines that will never assert in Jaguar I
3207 bool textext = false, txtread = false;
3211 bool sshftld = true; // D flipflop (D -> Q): instart -> sshftld
3212 //NOTE: sshftld probably is only asserted at the beginning of the inner loop. !!! FIX !!!
3214 Blit! (CMD = 01800005)
3215 Flags: SRCEN SRCENX LFUFUNC=C
3217 a1_base = 00037290, a2_base = 000095D0
3218 a1_x = 0000, a1_y = 0000, a2_x = 0002, a2_y = 0000
3219 a1_pixsize = 4, a2_pixsize = 4
3220 srcd=0000000000000000, dstd=0000000000000000, patd=0000000000000000
3222 [in=T a1f=F a1=F zf=F z=F a2=F iif=F iii=F izf=F izi=F]
3223 Entering INNER state...
3224 Entering SREADX state... [dstart=0 dend=20 pwidth=8 srcshift=20]
3225 Source extra read address/pix address: 000095D4/0 [0000001C00540038]
3226 Entering A2_ADD state [a2_x=0002, a2_y=0000, addasel=0, addbsel=1, modx=2, addareg=F, adda_xconst=2, adda_yconst=0]...
3227 Entering SREAD state... [dstart=0 dend=20 pwidth=8 srcshift=0]
3228 Source read address/pix address: 000095D8/0 [0054003800009814]
3229 Entering A2_ADD state [a2_x=0004, a2_y=0000, addasel=0, addbsel=1, modx=2, addareg=F, adda_xconst=2, adda_yconst=0]...
3230 Entering DWRITE state...
3231 Dest write address/pix address: 00037290/0 [dstart=0 dend=20 pwidth=8 srcshift=0] (icount=026E, inc=4)
3232 Entering A1_ADD state [a1_x=0000, a1_y=0000, addasel=0, addbsel=0, modx=2, addareg=F, adda_xconst=2, adda_yconst=0]...
3233 Entering SREAD state... [dstart=0 dend=20 pwidth=8 srcshift=0]
3234 Source read address/pix address: 000095E0/0 [00009968000377C7]
3235 Entering A2_ADD state [a2_x=0008, a2_y=0000, addasel=0, addbsel=1, modx=2, addareg=F, adda_xconst=2, adda_yconst=0]...
3236 Entering DWRITE state...
3237 Dest write address/pix address: 00037298/0 [dstart=0 dend=20 pwidth=8 srcshift=0] (icount=026A, inc=4)
3238 Entering A1_ADD state [a1_x=0004, a1_y=0000, addasel=0, addbsel=0, modx=2, addareg=F, adda_xconst=2, adda_yconst=0]...
3241 // while (!idle_inner)
3246 if ((idle_inner && !step)
3247 || (dzwrite && step && inner0)
3248 || (dwrite && step && !dstwrz && inner0))
3250 #ifdef VERBOSE_BLITTER_LOGGING
3253 printf(" Entering IDLE_INNER state...\n");
3261 idle_inneri = false;
3263 // EXTRA SOURCE DATA READ
3265 if ((idle_inner && step && srcenx)
3266 || (sreadx && !step))
3273 // EXTRA SOURCE ZED READ
3275 if ((sreadx && step && srcenz)
3276 || (szreadx && !step))
3283 // TEXTURE DATA READ (not implemented because not in Jaguar I)
3287 if ((szreadx && step && !textext)
3288 || (sreadx && step && !srcenz && srcen)
3289 || (idle_inner && step && !srcenx && !textext && srcen)
3290 || (dzwrite && step && !inner0 && !textext && srcen)
3291 || (dwrite && step && !dstwrz && !inner0 && !textext && srcen)
3292 || (txtread && step && srcen)
3293 || (sread && !step))
3302 if ((sread && step && srcenz)
3303 || (szread && !step))
3310 // DESTINATION DATA READ
3312 if ((szread && step && dsten)
3313 || (sread && step && !srcenz && dsten)
3314 || (sreadx && step && !srcenz && !textext && !srcen && dsten)
3315 || (idle_inner && step && !srcenx && !textext && !srcen && dsten)
3316 || (dzwrite && step && !inner0 && !textext && !srcen && dsten)
3317 || (dwrite && step && !dstwrz && !inner0 && !textext && !srcen && dsten)
3318 || (txtread && step && !srcen && dsten)
3319 || (dread && !step))
3326 // DESTINATION ZED READ
3328 if ((dread && step && dstenz)
3329 || (szread && step && !dsten && dstenz)
3330 || (sread && step && !srcenz && !dsten && dstenz)
3331 || (sreadx && step && !srcenz && !textext && !srcen && !dsten && dstenz)
3332 || (idle_inner && step && !srcenx && !textext && !srcen && !dsten && dstenz)
3333 || (dzwrite && step && !inner0 && !textext && !srcen && !dsten && dstenz)
3334 || (dwrite && step && !dstwrz && !inner0 && !textext && !srcen && !dsten && dstenz)
3335 || (txtread && step && !srcen && !dsten && dstenz)
3336 || (dzread && !step))
3343 // DESTINATION DATA WRITE
3345 if ((dzread && step)
3346 || (dread && step && !dstenz)
3347 || (szread && step && !dsten && !dstenz)
3348 || (sread && step && !srcenz && !dsten && !dstenz)
3349 || (txtread && step && !srcen && !dsten && !dstenz)
3350 || (sreadx && step && !srcenz && !textext && !srcen && !dsten && !dstenz)
3351 || (idle_inner && step && !srcenx && !textext && !srcen && !dsten && !dstenz)
3352 || (dzwrite && step && !inner0 && !textext && !srcen && !dsten && !dstenz)
3353 || (dwrite && step && !dstwrz && !inner0 && !textext && !srcen && !dsten && !dstenz)
3354 || (dwrite && !step))
3361 // DESTINATION ZED WRITE
3363 if ((dzwrite && !step)
3364 || (dwrite && step && dstwrz))
3371 //Kludge: A QnD way to make sure that sshftld is asserted only for the first
3372 // cycle of the inner loop...
3373 sshftld = idle_inner;
3375 // Here we move the fooi into their foo counterparts in order to simulate the moving
3376 // of data into the various FDSYNCs... Each time we loop we simulate one clock cycle...
3378 idle_inner = idle_inneri;
3388 // Here's a few more decodes--not sure if they're supposed to go here or not...
3390 bool srca_addi = (sreadxi && !srcenz) || (sreadi && !srcenz) || szreadxi || szreadi;
3392 bool dsta_addi = (dwritei && !dstwrz) || dzwritei;
3394 bool gensrc = sreadxi || szreadxi || sreadi || szreadi;
3395 bool gendst = dreadi || dzreadi || dwritei || dzwritei;
3396 bool gena2i = (gensrc && !dsta2) || (gendst && dsta2);
3398 bool zaddr = szreadx || szread || dzread || dzwrite;
3400 // Some stuff from MCONTROL.NET--not sure if this is the correct use of this decode or not...
3401 /*Fontread\ := OND1 (fontread\, sread[1], sreadx[1], bcompen);
3402 Fontread := INV1 (fontread, fontread\);
3403 Justt := NAN3 (justt, fontread\, phrase_mode, tactive\);
3404 Justify := TS (justify, justt, busen);*/
3405 bool fontread = (sread || sreadx) && bcompen;
3406 bool justify = !(!fontread && phrase_mode /*&& tactive*/);
3408 /* Generate inner loop update enables */
3410 A1_addi := MX2 (a1_addi, dsta_addi, srca_addi, dsta2);
3411 A2_addi := MX2 (a2_addi, srca_addi, dsta_addi, dsta2);
3412 A1_add := FD1 (a1_add, a1_add\, a1_addi, clk);
3413 A2_add := FD1 (a2_add, a2_add\, a2_addi, clk);
3414 A2_addb := BUF1 (a2_addb, a2_add);
3416 bool a1_add = (dsta2 ? srca_addi : dsta_addi);
3417 bool a2_add = (dsta2 ? dsta_addi : srca_addi);
3419 /* Address adder input A register selection
3420 000 A1 step integer part
3421 001 A1 step fraction part
3422 010 A1 increment integer part
3423 011 A1 increment fraction part
3427 bit 1 = /a2update . (a1_add . a1addx[0..1])
3428 bit 0 = /a2update . ( a1fupdate
3429 + a1_add . atick[0] . a1addx[0..1])
3430 The /a2update term on bits 0 and 1 is redundant.
3431 Now look-ahead based
3433 uint8 addasel = (a1fupdate || (a1_add && a1addx == 3) ? 0x01 : 0x00);
3434 addasel |= (a1_add && a1addx == 3 ? 0x02 : 0x00);
3435 addasel |= (a2update ? 0x04 : 0x00);
3436 /* Address adder input A X constant selection
3437 adda_xconst[0..2] generate a power of 2 in the range 1-64 or all
3438 zeroes when they are all 1
3439 Remember - these are pixels, so to add one phrase the pixel size
3440 has to be taken into account to get the appropriate value.
3442 if a1addx[0..1] are 00 set 6 - pixel size
3443 if a1addx[0..1] are 01 set the value 000
3444 if a1addx[0..1] are 10 set the value 111
3446 JLH: Also, 11 will likewise set the value to 111
3448 uint8 a1_xconst = 6 - a1_pixsize, a2_xconst = 6 - a2_pixsize;
3452 else if (a1addx & 0x02)
3457 else if (a2addx & 0x02)
3460 uint8 adda_xconst = (a2_add ? a2_xconst : a1_xconst);
3461 /* Address adder input A Y constant selection
3462 22 June 94 - This was erroneous, because only the a1addy bit was reflected here.
3463 Therefore, the selection has to be controlled by a bug fix bit.
3464 JLH: Bug fix bit in Jaguar II--not in Jaguar I!
3466 bool adda_yconst = a1addy;
3467 /* Address adder input A register versus constant selection
3468 given by a1_add . a1addx[0..1]
3471 + a2_add . a2addx[0..1]
3474 bool addareg = ((a1_add && a1addx == 3) || a1update || a1fupdate
3475 || (a2_add && a2addx == 3) || a2update ? true : false);
3476 /* The adders can be put into subtract mode in add pixel size
3477 mode when the corresponding flags are set */
3478 bool suba_x = ((a1_add && a1xsign && a1addx == 1) || (a2_add && a2xsign && a2addx == 1) ? true : false);
3479 bool suba_y = ((a1_add && a1addy && a1ysign) || (a2_add && a2addy && a2ysign) ? true : false);
3480 /* Address adder input B selection
3487 + (a1_add . atick[0] . a1addx[0..1])
3488 + a1fupdate . a1_stepld
3489 + a1update . a1_stepld
3490 + a2update . a2_stepld
3491 Bit 0 = a2update + a2_add
3492 + a1fupdate . a1_stepld
3493 + a1update . a1_stepld
3494 + a2update . a2_stepld
3496 uint8 addbsel = (a2update || a2_add || (a1fupdate && a1_stepld)
3497 || (a1update && a1_stepld) || (a2update && a2_stepld) ? 0x01 : 0x00);
3498 addbsel |= (a1fupdate || (a1_add && a1addx == 3) || (a1fupdate && a1_stepld)
3499 || (a1update && a1_stepld) || (a2update && a2_stepld) ? 0x02 : 0x00);
3501 /* The modulo bits are used to align X onto a phrase boundary when
3502 it is being updated by one phrase
3509 Masking is enabled for a1 when a1addx[0..1] is 00, and the value
3510 is 6 - the pixel size (again!)
3512 uint8 maska1 = (a1_add && a1addx == 0 ? 6 - a1_pixsize : 0);
3513 uint8 maska2 = (a2_add && a2addx == 0 ? 6 - a2_pixsize : 0);
3514 uint8 modx = (a2_add ? maska2 : maska1);
3515 /* Generate load strobes for the increment updates */
3517 /*A1pldt := NAN2 (a1pldt, atick[1], a1_add);
3518 A1ptrldi := NAN2 (a1ptrldi, a1update\, a1pldt);
3520 A1fldt := NAN4 (a1fldt, atick[0], a1_add, a1addx[0..1]);
3521 A1fracldi := NAN2 (a1fracldi, a1fupdate\, a1fldt);
3523 A2pldt := NAN2 (a2pldt, atick[1], a2_add);
3524 A2ptrldi := NAN2 (a2ptrldi, a2update\, a2pldt);*/
3525 bool a1fracldi = a1fupdate || (a1_add && a1addx == 3);
3527 // Some more from DCONTROL...
3528 // atick[] just MAY be important here! We're assuming it's true and dropping the term...
3529 // That will probably screw up some of the lower terms that seem to rely on the timing of it...
3530 #warning srcdreadd is not properly initialized!
3531 bool srcdreadd = false; // Set in INNER.NET
3532 //Shadeadd\ := NAN2H (shadeadd\, dwrite, srcshade);
3533 //Shadeadd := INV2 (shadeadd, shadeadd\);
3534 bool shadeadd = dwrite && srcshade;
3535 /* Data adder control, input A selection
3536 000 Destination data
3537 001 Initialiser pixel value
3538 100 Source data - computed intensity fraction
3539 101 Pattern data - computed intensity
3540 110 Source zed 1 - computed zed
3541 111 Source zed 2 - computed zed fraction
3543 Bit 0 = dwrite . gourd . atick[1]
3544 + dzwrite . gourz . atick[0]
3547 + init_if + init_ii + init_zf + init_zi
3548 Bit 1 = dzwrite . gourz . (atick[0] + atick[1])
3551 Bit 2 = (gourd + gourz) . /(init_if + init_ii + init_zf + init_zi)
3554 uint8 daddasel = ((dwrite && gourd) || (dzwrite && gourz) || istepadd || zstepfadd
3555 || init_if || init_ii || init_zf || init_zi ? 0x01 : 0x00);
3556 daddasel |= ((dzwrite && gourz) || zstepadd || zstepfadd ? 0x02 : 0x00);
3557 daddasel |= (((gourd || gourz) && !(init_if || init_ii || init_zf || init_zi))
3558 || (dwrite && srcshade) ? 0x04 : 0x00);
3559 /* Data adder control, input B selection
3561 0001 Data initialiser increment
3562 0100 Bottom 16 bits of I increment repeated four times
3563 0101 Top 16 bits of I increment repeated four times
3564 0110 Bottom 16 bits of Z increment repeated four times
3565 0111 Top 16 bits of Z increment repeated four times
3566 1100 Bottom 16 bits of I step repeated four times
3567 1101 Top 16 bits of I step repeated four times
3568 1110 Bottom 16 bits of Z step repeated four times
3569 1111 Top 16 bits of Z step repeated four times
3571 Bit 0 = dwrite . gourd . atick[1]
3572 + dzwrite . gourz . atick[1]
3576 + init_if + init_ii + init_zf + init_zi
3577 Bit 1 = dzwrite . gourz . (atick[0] + atick[1])
3580 Bit 2 = dwrite . gourd . (atick[0] + atick[1])
3581 + dzwrite . gourz . (atick[0] + atick[1])
3583 + istepadd + istepfadd + zstepadd + zstepfadd
3584 Bit 3 = istepadd + istepfadd + zstepadd + zstepfadd
3586 uint8 daddbsel = ((dwrite && gourd) || (dzwrite && gourz) || (dwrite && srcshade)
3587 || istepadd || zstepadd || init_if || init_ii || init_zf || init_zi ? 0x01 : 0x00);
3588 daddbsel |= ((dzwrite && gourz) || zstepadd || zstepfadd ? 0x02 : 0x00);
3589 daddbsel |= ((dwrite && gourd) || (dzwrite && gourz) || (dwrite && srcshade)
3590 || istepadd || istepfadd || zstepadd || zstepfadd ? 0x04 : 0x00);
3591 daddbsel |= (istepadd && istepfadd && zstepadd && zstepfadd ? 0x08 : 0x00);
3592 /* Data adder mode control
3593 000 16-bit normal add
3594 001 16-bit saturating add with carry
3595 010 8-bit saturating add with carry, carry into top byte is
3597 011 8-bit saturating add with carry, carry into top byte and
3598 between top nybbles is inhibited (CRY)
3599 100 16-bit normal add with carry
3600 101 16-bit saturating add
3601 110 8-bit saturating add, carry into top byte is inhibited
3602 111 8-bit saturating add, carry into top byte and between top
3603 nybbles is inhibited
3605 The first five are used for Gouraud calculations, the latter three
3606 for adding source and destination data
3608 Bit 0 = dzwrite . gourz . atick[1]
3609 + dwrite . gourd . atick[1] . /topnen . /topben . /ext_int
3610 + dwrite . gourd . atick[1] . topnen . topben . /ext_int
3612 + istepadd . /topnen . /topben . /ext_int
3613 + istepadd . topnen . topben . /ext_int
3614 + /gourd . /gourz . /topnen . /topben
3615 + /gourd . /gourz . topnen . topben
3616 + shadeadd . /topnen . /topben
3617 + shadeadd . topnen . topben
3618 + init_ii . /topnen . /topben . /ext_int
3619 + init_ii . topnen . topben . /ext_int
3622 Bit 1 = dwrite . gourd . atick[1] . /topben . /ext_int
3623 + istepadd . /topben . /ext_int
3624 + /gourd . /gourz . /topben
3625 + shadeadd . /topben
3626 + init_ii . /topben . /ext_int
3628 Bit 2 = /gourd . /gourz
3630 + dwrite . gourd . atick[1] . ext_int
3631 + istepadd . ext_int
3634 uint8 daddmode = ((dzwrite && gourz) || (dwrite && gourd && !topnen && !topben && !ext_int)
3635 || (dwrite && gourd && topnen && topben && !ext_int) || zstepadd
3636 || (istepadd && !topnen && !topben && !ext_int)
3637 || (istepadd && topnen && topben && !ext_int) || (!gourd && !gourz && !topnen && !topben)
3638 || (!gourd && !gourz && topnen && topben) || (shadeadd && !topnen && !topben)
3639 || (shadeadd && topnen && topben) || (init_ii && !topnen && !topben && !ext_int)
3640 || (init_ii && topnen && topben && !ext_int) || init_zi ? 0x01 : 0x00);
3641 daddmode |= ((dwrite && gourd && !topben && !ext_int) || (istepadd && !topben && !ext_int)
3642 || (!gourd && !gourz && !topben) || (shadeadd && !topben)
3643 || (init_ii && !topben && !ext_int) ? 0x02 : 0x00);
3644 daddmode |= ((!gourd && !gourz) || shadeadd || (dwrite && gourd && ext_int)
3645 || (istepadd && ext_int) || (init_ii && ext_int) ? 0x04 : 0x00);
3646 /* Data add load controls
3647 Pattern fraction (dest data) is loaded on
3648 dwrite . gourd . atick[0]
3649 + istepfadd . /datinit
3651 Pattern data is loaded on
3652 dwrite . gourd . atick[1]
3653 + istepadd . /datinit . /datinit
3655 Source z1 is loaded on
3656 dzwrite . gourz . atick[1]
3657 + zstepadd . /datinit . /datinit
3659 Source z2 is loaded on
3660 dzwrite . gourz . atick[0]
3663 Texture map shaded data is loaded on
3664 srcdreadd . srcshade
3666 bool patfadd = (dwrite && gourd) || (istepfadd && !datinit) || init_if;
3667 bool patdadd = (dwrite && gourd) || (istepadd && !datinit) || init_ii;
3668 bool srcz1add = (dzwrite && gourz) || (zstepadd && !datinit) || init_zi;
3669 bool srcz2add = (dzwrite && gourz) || zstepfadd || init_zf;
3670 bool srcshadd = srcdreadd && srcshade;
3671 bool daddq_sel = patfadd || patdadd || srcz1add || srcz2add || srcshadd;
3672 /* Select write data
3673 This has to be controlled from stage 1 of the pipe-line, delayed
3674 by one tick, as the write occurs in the cycle after the ack.
3681 Bit 0 = /patdsel . /adddsel
3686 uint8 data_sel = ((!patdsel && !adddsel) || dzwrite ? 0x01 : 0x00)
3687 | (adddsel || dzwrite ? 0x02 : 0x00);
3689 uint32 address, pixAddr;
3690 ADDRGEN(address, pixAddr, gena2i, zaddr,
3691 a1_x, a1_y, a1_base, a1_pitch, a1_pixsize, a1_width, a1_zoffset,
3692 a2_x, a2_y, a2_base, a2_pitch, a2_pixsize, a2_width, a2_zoffset);
3694 //Here's my guess as to how the addresses get truncated to phrase boundaries in phrase mode...
3696 address &= 0xFFFFF8;
3698 /* Generate source alignment shift
3699 -------------------------------
3700 The source alignment shift for data move is the difference between
3701 the source and destination X pointers, multiplied by the pixel
3702 size. Only the low six bits of the pointers are of interest, as
3703 pixel sizes are always a power of 2 and window rows are always
3706 When not in phrase mode, the top 3 bits of the shift value are
3709 Source shifting is also used to extract bits for bit-to-byte
3710 expansion in phrase mode. This involves only the bottom three
3711 bits of the shift value, and is based on the offset within the
3712 phrase of the destination X pointer, in pixels.
3714 Source shifting is disabled when srcen is not set.
3716 uint8 dstxp = (dsta2 ? a2_x : a1_x) & 0x3F;
3717 uint8 srcxp = (dsta2 ? a1_x : a2_x) & 0x3F;
3718 uint8 shftv = ((dstxp - srcxp) << pixsize) & 0x3F;
3719 /* The phrase mode alignment count is given by the phrase offset
3720 of the first pixel, for bit to byte expansion */
3724 pobb = dstxp & 0x07;
3726 pobb = dstxp & 0x03;
3728 pobb = dstxp & 0x01;
3730 bool pobbsel = phrase_mode && bcompen;
3731 uint8 loshd = (pobbsel ? pobb : shftv) & 0x07;
3732 uint8 shfti = (srcen || pobbsel ? (sshftld ? loshd : srcshift & 0x07) : 0);
3733 /* Enable for high bits is srcen . phrase_mode */
3734 shfti |= (srcen && phrase_mode ? (sshftld ? shftv & 0x38 : srcshift & 0x38) : 0);
3738 Note that there's a problem here--even though it's NOT in phrase mode, it's still calculating
3739 a source shift... !!! FIX !!!
3740 Actually, the problem is the code that utilizes the source shift even when it's not needed... I think.
3742 Blit! (CMD = 01800609)
3743 Flags: SRCEN DSTEN UPDA1 UPDA2 LFUFUNC=C
3745 a1_base = 001F8300, a2_base = 00812F80
3746 a1_x = 0007, a1_y = 0000, a1_frac_x = 0000, a1_frac_y = 0000, a2_x = 0000, a2_y = 0000
3747 a1_step_x = FFF6, a1_step_y = 0001, a1_stepf_x = 0000, a1_stepf_y = 0000, a2_step_x = FFF6, a2_step_y = 0001
3748 a1_inc_x = 0000, a1_inc_y = 0000, a1_incf_x = 0000, a1_incf_y = 0000
3749 a1_win_x = 0000, a1_win_y = 0000, a2_mask_x = 0000, a2_mask_y = 0000
3750 a2_mask=F a1add=+1/+0 a2add=+1/+0
3751 a1_pixsize = 2, a2_pixsize = 2
3752 srcd=0000000000000000 dstd=0000000000000000 patd=0000000000000000 iinc=00000000
3753 srcz1=0000000000000000 srcz2=0000000000000000 dstz=0000000000000000 zinc=00000000, coll=0
3755 [in=T a1f=F a1=F zf=F z=F a2=F iif=F iii=F izf=F izi=F]
3756 Entering INNER state...
3757 Entering SREAD state... Source read address/pix address: 00812F80/0 [0000000000000000]
3758 Entering A2_ADD state [a2_x=0000, a2_y=0000, addasel=0, addbsel=1, modx=0, addareg=F, adda_xconst=0, adda_yconst=0]...
3759 Entering DREAD state... Dest read address/pix address: 001F8303/4 [0000000000000000]
3760 Entering DWRITE state... Dest write address/pix address: 001F8303/4 srcz=0000000000000000]
3762 [dcomp=FF zcomp=00 dbinh=00]
3764 [srcz=0000000000000000 dstz=0000000000000000 zwdata=0000000000000000 mask=000F]
3765 [0000000000000000] (icount=0009, inc=1)
3766 [dstart=4 dend=8 pwidth=4 srcshift=4][daas=0 dabs=0 dam=7 ds=1 daq=F]
3767 Entering A1_ADD state [a1_x=0007, a1_y=0000, addasel=0, addbsel=0, modx=0, addareg=F, adda_xconst=0, adda_yconst=0]...
3772 #ifdef VERBOSE_BLITTER_LOGGING
3775 printf(" Entering SREADX state...");
3776 //printf(" [dstart=%X dend=%X pwidth=%X srcshift=%X]\n", dstart, dend, pwidth, srcshift);
3780 //uint32 srcAddr, pixAddr;
3781 //ADDRGEN(srcAddr, pixAddr, gena2i, zaddr,
3782 // a1_x, a1_y, a1_base, a1_pitch, a1_pixsize, a1_width, a1_zoffset,
3783 // a2_x, a2_y, a2_base, a2_pitch, a2_pixsize, a2_width, a2_zoffset);
3785 srcd1 = ((uint64)JaguarReadLong(address + 0, BLITTER) << 32)
3786 | (uint64)JaguarReadLong(address + 4, BLITTER);
3787 //Kludge to take pixel size into account...
3788 //Hmm. If we're not in phrase mode, this is most likely NOT going to be used...
3789 //Actually, it would be--because of BCOMPEN expansion, for example...
3798 else if (pixsize == 4)
3804 #ifdef VERBOSE_BLITTER_LOGGING
3807 printf(" Source extra read address/pix address: %08X/%1X [%08X%08X]\n", address, pixAddr,
3808 (uint32)(srcd1 >> 32), (uint32)(srcd1 & 0xFFFFFFFF));
3816 #ifdef VERBOSE_BLITTER_LOGGING
3819 printf(" Entering SZREADX state...");
3824 srcz1 = ((uint64)JaguarReadLong(address, BLITTER) << 32) | (uint64)JaguarReadLong(address + 4, BLITTER);
3825 #ifdef VERBOSE_BLITTER_LOGGING
3828 printf(" Src Z extra read address/pix address: %08X/%1X [%08X%08X]\n", address, pixAddr,
3829 (uint32)(dstz >> 32), (uint32)(dstz & 0xFFFFFFFF));
3837 #ifdef VERBOSE_BLITTER_LOGGING
3840 printf(" Entering SREAD state...");
3841 //printf(" [dstart=%X dend=%X pwidth=%X srcshift=%X]\n", dstart, dend, pwidth, srcshift);
3845 //uint32 srcAddr, pixAddr;
3846 //ADDRGEN(srcAddr, pixAddr, gena2i, zaddr,
3847 // a1_x, a1_y, a1_base, a1_pitch, a1_pixsize, a1_width, a1_zoffset,
3848 // a2_x, a2_y, a2_base, a2_pitch, a2_pixsize, a2_width, a2_zoffset);
3850 srcd1 = ((uint64)JaguarReadLong(address, BLITTER) << 32) | (uint64)JaguarReadLong(address + 4, BLITTER);
3851 //Kludge to take pixel size into account...
3860 else if (pixsize == 4)
3866 #ifdef VERBOSE_BLITTER_LOGGING
3869 printf(" Source read address/pix address: %08X/%1X [%08X%08X]\n", address, pixAddr,
3870 (uint32)(srcd1 >> 32), (uint32)(srcd1 & 0xFFFFFFFF));
3878 #ifdef VERBOSE_BLITTER_LOGGING
3881 printf(" Entering SZREAD state...");
3886 srcz1 = ((uint64)JaguarReadLong(address, BLITTER) << 32) | (uint64)JaguarReadLong(address + 4, BLITTER);
3887 //Kludge to take pixel size into account... I believe that it only has to take 16BPP mode into account. Not sure tho.
3888 if (!phrase_mode && pixsize == 4)
3891 #ifdef VERBOSE_BLITTER_LOGGING
3894 printf(" Src Z read address/pix address: %08X/%1X [%08X%08X]\n", address, pixAddr,
3895 (uint32)(dstz >> 32), (uint32)(dstz & 0xFFFFFFFF));
3903 #ifdef VERBOSE_BLITTER_LOGGING
3906 printf(" Entering DREAD state...");
3910 //uint32 dstAddr, pixAddr;
3911 //ADDRGEN(dstAddr, pixAddr, gena2i, zaddr,
3912 // a1_x, a1_y, a1_base, a1_pitch, a1_pixsize, a1_width, a1_zoffset,
3913 // a2_x, a2_y, a2_base, a2_pitch, a2_pixsize, a2_width, a2_zoffset);
3914 dstd = ((uint64)JaguarReadLong(address, BLITTER) << 32) | (uint64)JaguarReadLong(address + 4, BLITTER);
3915 //Kludge to take pixel size into account...
3920 else if (pixsize == 4)
3925 #ifdef VERBOSE_BLITTER_LOGGING
3928 printf(" Dest read address/pix address: %08X/%1X [%08X%08X]\n", address, pixAddr,
3929 (uint32)(dstd >> 32), (uint32)(dstd & 0xFFFFFFFF));
3937 // Is Z always 64 bit read? Or sometimes 16 bit (dependent on phrase_mode)?
3938 #ifdef VERBOSE_BLITTER_LOGGING
3941 printf(" Entering DZREAD state...");
3945 dstz = ((uint64)JaguarReadLong(address, BLITTER) << 32) | (uint64)JaguarReadLong(address + 4, BLITTER);
3946 //Kludge to take pixel size into account... I believe that it only has to take 16BPP mode into account. Not sure tho.
3947 if (!phrase_mode && pixsize == 4)
3950 #ifdef VERBOSE_BLITTER_LOGGING
3953 printf(" Dest Z read address/pix address: %08X/%1X [%08X%08X]\n", address, pixAddr,
3954 (uint32)(dstz >> 32), (uint32)(dstz & 0xFFFFFFFF));
3960 // These vars should probably go further up in the code... !!! FIX !!!
3961 // We can't preassign these unless they're static...
3962 //uint64 srcz = 0; // These are assigned to shut up stupid compiler warnings--dwrite is ALWAYS asserted
3963 //bool winhibit = false;
3966 //NOTE: SRCSHADE requires GOURZ to be set to work properly--another Jaguar I bug
3969 #ifdef VERBOSE_BLITTER_LOGGING
3972 printf(" Entering DWRITE state...");
3976 //Counter is done on the dwrite state...! (We'll do it first, since it affects dstart/dend calculations.)
3977 //Here's the voodoo for figuring the correct amount of pixels in phrase mode (or not):
3978 int8 inct = -((dsta2 ? a2_x : a1_x) & 0x07); // From INNER_CNT
3980 inc = (!phrase_mode || (phrase_mode && (inct & 0x01)) ? 0x01 : 0x00);
3981 inc |= (phrase_mode && (((pixsize == 3 || pixsize == 4) && (inct & 0x02)) || pixsize == 5 && !(inct & 0x01)) ? 0x02 : 0x00);
3982 inc |= (phrase_mode && ((pixsize == 3 && (inct & 0x04)) || (pixsize == 4 && !(inct & 0x03))) ? 0x04 : 0x00);
3983 inc |= (phrase_mode && pixsize == 3 && !(inct & 0x07) ? 0x08 : 0x00);
3985 uint16 oldicount = icount; // Save icount to detect underflow...
3988 if (icount == 0 || ((icount & 0x8000) && !(oldicount & 0x8000)))
3990 // X/Y stepping is also done here, I think...No. It's done when a1_add or a2_add is asserted...
3992 //*********************************************************************************
3993 //Start & end write mask computations...
3994 //*********************************************************************************
3999 dstart = (dstxp & 0x07) << 3;
4001 dstart = (dstxp & 0x03) << 4;
4003 dstart = (dstxp & 0x01) << 5;
4005 dstart = (phrase_mode ? dstart : pixAddr & 0x07);
4007 //This is the other Jaguar I bug... Normally, should ALWAYS select a1_x here.
4008 uint16 dstxwr = (dsta2 ? a2_x : a1_x) & 0x7FFE;
4009 uint16 pseq = dstxwr ^ (a1_win_x & 0x7FFE);
4010 pseq = (pixsize == 5 ? pseq : pseq & 0x7FFC);
4011 pseq = ((pixsize & 0x06) == 4 ? pseq : pseq & 0x7FF8);
4012 bool penden = clip_a1 && (pseq == 0);
4013 uint8 window_mask = 0;
4016 window_mask = (a1_win_x & 0x07) << 3;
4018 window_mask = (a1_win_x & 0x03) << 4;
4020 window_mask = (a1_win_x & 0x01) << 5;
4022 window_mask = (penden ? window_mask : 0);
4025 Entering SREADX state... [dstart=0 dend=20 pwidth=8 srcshift=20]
4026 Source extra read address/pix address: 000095D0/0 [000004E40000001C]
4027 Entering A2_ADD state [a2_x=0002, a2_y=0000, addasel=0, addbsel=1, modx=2, addareg=F, adda_xconst=2, adda_yconst=0]...
4028 Entering SREAD state... [dstart=0 dend=20 pwidth=8 srcshift=20]
4029 Source read address/pix address: 000095D8/0 [0054003800009814]
4030 Entering A2_ADD state [a2_x=0004, a2_y=0000, addasel=0, addbsel=1, modx=2, addareg=F, adda_xconst=2, adda_yconst=0]...
4031 Entering DWRITE state...
4032 Dest write address/pix address: 00037290/0 [dstart=0 dend=20 pwidth=8 srcshift=20][daas=0 dabs=0 dam=7 ds=1 daq=F] [0000001C00000000] (icount=026E, inc=4)
4033 Entering A1_ADD state [a1_x=0000, a1_y=0000, addasel=0, addbsel=0, modx=2, addareg=F, adda_xconst=2, adda_yconst=0]...
4035 (icount=026E, inc=4)
4036 icount & 0x03 = 0x02
4039 window_mask = 0x1000
4041 Therefore, it chooses the inner_mask over the window_mask every time! Argh!
4042 This is because we did this wrong:
4043 Innerm[3-5] := AN2 (inner_mask[3-5], imb[3-5], inner0);
4044 NOTE! This doesn't fix the problem because inner0 is asserted too late to help here. !!! FIX !!! [Should be DONE]
4047 /* The mask to be used if within one phrase of the end of the inner
4049 uint8 inner_mask = 0;
4052 inner_mask = (icount & 0x07) << 3;
4054 inner_mask = (icount & 0x03) << 4;
4056 inner_mask = (icount & 0x01) << 5;
4059 /* The actual mask used should be the lesser of the window masks and
4060 the inner mask, where is all cases 000 means 1000. */
4061 window_mask = (window_mask == 0 ? 0x40 : window_mask);
4062 inner_mask = (inner_mask == 0 ? 0x40 : inner_mask);
4063 uint8 emask = (window_mask > inner_mask ? inner_mask : window_mask);
4064 /* The mask to be used for the pixel size, to which must be added
4066 uint8 pma = pixAddr + (1 << pixsize);
4067 /* Select the mask */
4068 uint8 dend = (phrase_mode ? emask : pma);
4070 /* The cycle width in phrase mode is normally one phrase. However,
4071 at the start and end it may be narrower. The start and end masks
4072 are used to generate this. The width is given by:
4074 8 - start mask - (8 - end mask)
4075 = end mask - start mask
4077 This is only used for writes in phrase mode.
4078 Start and end from the address level of the pipeline are used.
4080 uint8 pwidth = (((dend | dstart) & 0x07) == 0 ? 0x08 : (dend - dstart) & 0x07);
4082 //uint32 dstAddr, pixAddr;
4083 //ADDRGEN(dstAddr, pixAddr, gena2i, zaddr,
4084 // a1_x, a1_y, a1_base, a1_pitch, a1_pixsize, a1_width, a1_zoffset,
4085 // a2_x, a2_y, a2_base, a2_pitch, a2_pixsize, a2_width, a2_zoffset);
4086 #ifdef VERBOSE_BLITTER_LOGGING
4089 printf(" Dest write address/pix address: %08X/%1X", address, pixAddr);
4094 //More testing... This is almost certainly wrong, but how else does this work???
4095 //Seems to kinda work... But still, this doesn't seem to make any sense!
4096 if (phrase_mode && !dsten)
4097 dstd = ((uint64)JaguarReadLong(address, BLITTER) << 32) | (uint64)JaguarReadLong(address + 4, BLITTER);
4099 //Testing only... for now...
4100 //This is wrong because the write data is a combination of srcd and dstd--either run
4101 //thru the LFU or in PATDSEL or ADDDSEL mode. [DONE now, thru DATA module]
4102 // Precedence is ADDDSEL > PATDSEL > LFU.
4103 //Also, doesn't take into account the start & end masks, or the phrase width...
4106 // srcd2 = xxxx xxxx 0123 4567, srcd = 8901 2345 xxxx xxxx, srcshift = $20 (32)
4107 uint64 srcd = (srcd2 << (64 - srcshift)) | (srcd1 >> srcshift);
4108 //bleh, ugly ugly ugly
4112 //Z DATA() stuff done here... And it has to be done before any Z shifting...
4113 //Note that we need to have phrase mode start/end support here... (Not since we moved it from dzwrite...!)
4115 Here are a couple of Cybermorph blits with Z:
4116 $00113078 // DSTEN DSTENZ DSTWRZ CLIP_A1 GOURD GOURZ PATDSEL ZMODE=4
4117 $09900F39 // SRCEN DSTEN DSTENZ DSTWRZ UPDA1 UPDA1F UPDA2 DSTA2 ZMODE=4 LFUFUNC=C DCOMPEN
4119 We're having the same phrase mode overwrite problem we had with the pixels... !!! FIX !!!
4120 Odd. It's equating 0 with 0... Even though ZMODE is $04 (less than)!
4125 void ADDARRAY(uint16 * addq, uint8 daddasel, uint8 daddbsel, uint8 daddmode,
4126 uint64 dstd, uint32 iinc, uint8 initcin[], uint64 initinc, uint16 initpix,
4127 uint32 istep, uint64 patd, uint64 srcd, uint64 srcz1, uint64 srcz2,
4128 uint32 zinc, uint32 zstep)
4131 uint8 initcin[4] = { 0, 0, 0, 0 };
4132 ADDARRAY(addq, 7/*daddasel*/, 6/*daddbsel*/, 0/*daddmode*/, 0, 0, initcin, 0, 0, 0, 0, 0, srcz1, srcz2, zinc, 0);
4133 srcz2 = ((uint64)addq[3] << 48) | ((uint64)addq[2] << 32) | ((uint64)addq[1] << 16) | (uint64)addq[0];
4134 ADDARRAY(addq, 6/*daddasel*/, 7/*daddbsel*/, 1/*daddmode*/, 0, 0, initcin, 0, 0, 0, 0, 0, srcz1, srcz2, zinc, 0);
4135 srcz1 = ((uint64)addq[3] << 48) | ((uint64)addq[2] << 32) | ((uint64)addq[1] << 16) | (uint64)addq[0];
4137 #ifdef VERBOSE_BLITTER_LOGGING
4140 printf("\n[srcz1=%08X%08X, srcz2=%08X%08X, zinc=%08X",
4141 (uint32)(srcz1 >> 32), (uint32)(srcz1 & 0xFFFFFFFF),
4142 (uint32)(srcz2 >> 32), (uint32)(srcz2 & 0xFFFFFFFF), zinc);
4148 uint8 zSrcShift = srcshift & 0x30;
4149 srcz = (srcz2 << (64 - zSrcShift)) | (srcz1 >> zSrcShift);
4150 //bleh, ugly ugly ugly
4154 #ifdef VERBOSE_BLITTER_LOGGING
4157 printf(" srcz=%08X%08X]\n", (uint32)(srcz >> 32), (uint32)(srcz & 0xFFFFFFFF));
4162 //When in SRCSHADE mode, it adds the IINC to the read source (from LFU???)
4163 //According to following line, it gets LFU mode. But does it feed the source into the LFU
4165 //Dest write address/pix address: 0014E83E/0 [dstart=0 dend=10 pwidth=8 srcshift=0][daas=4 dabs=5 dam=7 ds=1 daq=F] [0000000000006505] (icount=003F, inc=1)
4169 //NOTE: This is basically doubling the work done by DATA--since this is what
4170 // ADDARRAY is loaded with when srschshade is enabled... !!! FIX !!!
4171 // Also note that it doesn't work properly unless GOURZ is set--there's the clue!
4173 uint8 initcin[4] = { 0, 0, 0, 0 };
4174 ADDARRAY(addq, 4/*daddasel*/, 5/*daddbsel*/, 7/*daddmode*/, dstd, iinc, initcin, 0, 0, 0, patd, srcd, 0, 0, 0, 0);
4175 srcd = ((uint64)addq[3] << 48) | ((uint64)addq[2] << 32) | ((uint64)addq[1] << 16) | (uint64)addq[0];
4177 //Seems to work... Not 100% sure tho.
4180 //Temporary kludge, to see if the fractional pattern does anything...
4182 //But it seems to mess up in Cybermorph... the shading should be smooth but it isn't...
4183 //Seems the carry out is lost again... !!! FIX !!! [DONE--see below]
4187 uint8 initcin[4] = { 0, 0, 0, 0 };
4188 ADDARRAY(addq, 4/*daddasel*/, 4/*daddbsel*/, 0/*daddmode*/, dstd, iinc, initcin, 0, 0, 0, patd, srcd, 0, 0, 0, 0);
4189 srcd1 = ((uint64)addq[3] << 48) | ((uint64)addq[2] << 32) | ((uint64)addq[1] << 16) | (uint64)addq[0];
4192 //Note that we still don't take atick[0] & [1] into account here, so this will skip half of the data needed... !!! FIX !!!
4193 //Not yet enumerated: dbinh, srcdread, srczread
4194 //Also, should do srcshift on the z value in phrase mode... !!! FIX !!! [DONE]
4195 //As well as add a srcz variable we can set external to this state... !!! FIX !!! [DONE]
4199 DATA(wdata, dcomp, zcomp, winhibit,
4200 true, cmpdst, daddasel, daddbsel, daddmode, daddq_sel, data_sel, 0/*dbinh*/,
4201 dend, dstart, dstd, iinc, lfufunc, patd, patdadd,
4202 phrase_mode, srcd, false/*srcdread*/, false/*srczread*/, srcz2add, zmode,
4203 bcompen, bkgwren, dcompen, icount & 0x07, pixsize,
4206 Seems that the phrase mode writes with DCOMPEN and DSTEN are corrupting inside of DATA: !!! FIX !!!
4207 It's fairly random as well. 7CFE -> 7DFE, 7FCA -> 78CA, 7FA4 -> 78A4, 7F88 -> 8F88
4208 It could be related to an uninitialized variable, like the zmode bug...
4210 It was a bug in the dech38el data--it returned $FF for ungated instead of $00...
4212 Blit! (CMD = 09800609)
4213 Flags: SRCEN DSTEN UPDA1 UPDA2 LFUFUNC=C DCOMPEN
4215 a1_base = 00110000, a2_base = 0010B2A8
4216 a1_x = 004B, a1_y = 00D8, a1_frac_x = 0000, a1_frac_y = 0000, a2_x = 0704, a2_y = 0000
4217 a1_step_x = FFF3, a1_step_y = 0001, a1_stepf_x = 0000, a1_stepf_y = 0000, a2_step_x = FFFC, a2_step_y = 0000
4218 a1_inc_x = 0000, a1_inc_y = 0000, a1_incf_x = 0000, a1_incf_y = 0000
4219 a1_win_x = 0000, a1_win_y = 0000, a2_mask_x = 0000, a2_mask_y = 0000
4220 a2_mask=F a1add=+phr/+0 a2add=+phr/+0
4221 a1_pixsize = 4, a2_pixsize = 4
4222 srcd=0000000000000000 dstd=0000000000000000 patd=0000000000000000 iinc=00000000
4223 srcz1=0000000000000000 srcz2=0000000000000000 dstz=0000000000000000 zinc=00000000, coll=0
4225 [in=T a1f=F a1=F zf=F z=F a2=F iif=F iii=F izf=F izi=F]
4226 Entering INNER state...
4227 Entering SREAD state... Source read address/pix address: 0010C0B0/0 [0000000078047804]
4228 Entering A2_ADD state [a2_x=0704, a2_y=0000, addasel=0, addbsel=1, modx=2, addareg=F, adda_xconst=2, adda_yconst=0]...
4229 Entering DREAD state...
4230 Dest read address/pix address: 00197240/0 [0000000000000028]
4231 Entering DWRITE state...
4232 Dest write address/pix address: 00197240/0 [dstart=30 dend=40 pwidth=8 srcshift=30][daas=0 dabs=0 dam=7 ds=1 daq=F] [0000000000000028] (icount=0009, inc=1)
4233 Entering A1_ADD state [a1_x=004B, a1_y=00D8, addasel=0, addbsel=0, modx=2, addareg=F, adda_xconst=2, adda_yconst=0]...
4234 Entering SREAD state... Source read address/pix address: 0010C0B8/0 [7804780478047804]
4235 Entering A2_ADD state [a2_x=0708, a2_y=0000, addasel=0, addbsel=1, modx=2, addareg=F, adda_xconst=2, adda_yconst=0]...
4236 Entering DREAD state...
4237 Dest read address/pix address: 00197260/0 [0028000000200008]
4238 Entering DWRITE state...
4239 Dest write address/pix address: 00197260/0 [dstart=0 dend=40 pwidth=8 srcshift=30][daas=0 dabs=0 dam=7 ds=1 daq=F] [0028780478047804] (icount=0005, inc=4)
4240 Entering A1_ADD state [a1_x=004C, a1_y=00D8, addasel=0, addbsel=0, modx=2, addareg=F, adda_xconst=2, adda_yconst=0]...
4241 Entering SREAD state... Source read address/pix address: 0010C0C0/0 [0000000000000000]
4242 Entering A2_ADD state [a2_x=070C, a2_y=0000, addasel=0, addbsel=1, modx=2, addareg=F, adda_xconst=2, adda_yconst=0]...
4243 Entering DREAD state...
4244 Dest read address/pix address: 00197280/0 [0008001800180018]
4245 Entering DWRITE state...
4246 Dest write address/pix address: 00197280/0 [dstart=0 dend=40 pwidth=8 srcshift=30][daas=0 dabs=0 dam=7 ds=1 daq=F] [7804780478040018] (icount=0001, inc=4)
4247 Entering A1_ADD state [a1_x=0050, a1_y=00D8, addasel=0, addbsel=0, modx=2, addareg=F, adda_xconst=2, adda_yconst=0]...
4248 Entering SREAD state... Source read address/pix address: 0010C0C8/0 [000078047BFE7BFE]
4249 Entering A2_ADD state [a2_x=0710, a2_y=0000, addasel=0, addbsel=1, modx=2, addareg=F, adda_xconst=2, adda_yconst=0]...
4250 Entering DREAD state...
4251 Dest read address/pix address: 001972A0/0 [0008002000000000]
4252 Entering DWRITE state...
4253 Dest write address/pix address: 001972A0/0 [dstart=0 dend=10 pwidth=8 srcshift=30][daas=0 dabs=0 dam=7 ds=1 daq=F] [0008002000000000] (icount=FFFD, inc=4)
4254 Entering A1_ADD state [a1_x=0054, a1_y=00D8, addasel=0, addbsel=0, modx=2, addareg=F, adda_xconst=2, adda_yconst=0]...
4255 Entering IDLE_INNER state...
4258 //Why isn't this taken care of in DATA? Because, DATA is modifying its local copy instead of the one used here.
4259 //!!! FIX !!! [DONE]
4268 a1_outside // A1 pointer is outside window bounds
4277 // The address is outside if negative, or if greater than or equal
4278 // to the window size
4280 A1_xcomp := MAG_15 (a1xgr, a1xeq, a1xlt, a1_x{0..14}, a1_win_x{0..14});
4281 A1_ycomp := MAG_15 (a1ygr, a1yeq, a1ylt, a1_y{0..14}, a1_win_y{0..14});
4282 A1_outside := OR6 (a1_outside, a1_x{15}, a1xgr, a1xeq, a1_y{15}, a1ygr, a1yeq);
4284 //NOTE: There seems to be an off-by-one bug here in the clip_a1 section... !!! FIX !!!
4285 // Actually, seems to be related to phrase mode writes...
4286 // Or is it? Could be related to non-15-bit compares as above?
4287 if (clip_a1 && ((a1_x & 0x8000) || (a1_y & 0x8000) || (a1_x >= a1_win_x) || (a1_y >= a1_win_y)))
4294 JaguarWriteLong(address + 0, wdata >> 32, BLITTER);
4295 JaguarWriteLong(address + 4, wdata & 0xFFFFFFFF, BLITTER);
4300 JaguarWriteLong(address, wdata & 0xFFFFFFFF, BLITTER);
4301 else if (pixsize == 4)
4302 JaguarWriteWord(address, wdata & 0x0000FFFF, BLITTER);
4304 JaguarWriteByte(address, wdata & 0x000000FF, BLITTER);
4308 #ifdef VERBOSE_BLITTER_LOGGING
4311 printf(" [%08X%08X]", (uint32)(wdata >> 32), (uint32)(wdata & 0xFFFFFFFF));
4312 printf(" (icount=%04X, inc=%u)\n", icount, (uint16)inc);
4313 printf(" [dstart=%X dend=%X pwidth=%X srcshift=%X]", dstart, dend, pwidth, srcshift);
4314 printf("[daas=%X dabs=%X dam=%X ds=%X daq=%s]\n", daddasel, daddbsel, daddmode, data_sel, (daddq_sel ? "T" : "F"));
4322 // OK, here's the big insight: When NOT in GOURZ mode, srcz1 & 2 function EXACTLY the same way that
4323 // srcd1 & 2 work--there's an implicit shift from srcz1 to srcz2 whenever srcz1 is read.
4324 // OTHERWISE, srcz1 is the integer for the computed Z and srcz2 is the fractional part.
4325 // Writes to srcz1 & 2 follow the same pattern as the other 64-bit registers--low 32 at the low address,
4326 // high 32 at the high address (little endian!).
4327 // NOTE: GOURZ is still not properly supported. Check patd/patf handling...
4328 // Phrase mode start/end masks are not properly supported either...
4329 #ifdef VERBOSE_BLITTER_LOGGING
4332 printf(" Entering DZWRITE state...");
4333 printf(" Dest Z write address/pix address: %08X/%1X [%08X%08X]\n", address, pixAddr,
4334 (uint32)(srcz >> 32), (uint32)(srcz & 0xFFFFFFFF));
4338 //This is not correct... !!! FIX !!!
4339 //Should be OK now... We'll see...
4340 //Nope. Having the same starstep write problems in phrase mode as we had with pixels... !!! FIX !!!
4341 //This is not causing the problem in Hover Strike... :-/
4342 //The problem was with the SREADX not shifting. Still problems with Z comparisons & other text in pregame screen...
4347 JaguarWriteLong(address + 0, srcz >> 32, BLITTER);
4348 JaguarWriteLong(address + 4, srcz & 0xFFFFFFFF, BLITTER);
4353 JaguarWriteWord(address, srcz & 0x0000FFFF, BLITTER);
4356 #ifdef VERBOSE_BLITTER_LOGGING
4359 // printf(" [%08X%08X]\n", (uint32)(srcz >> 32), (uint32)(srcz & 0xFFFFFFFF));
4361 //printf(" [dstart=%X dend=%X pwidth=%X srcshift=%X]", dstart, dend, pwidth, srcshift);
4362 printf(" [dstart=? dend=? pwidth=? srcshift=%X]", srcshift);
4363 printf("[daas=%X dabs=%X dam=%X ds=%X daq=%s]\n", daddasel, daddbsel, daddmode, data_sel, (daddq_sel ? "T" : "F"));
4370 This is because the address generator was using only 15 bits of the X when it should have
4373 There's a slight problem here: The X pointer isn't wrapping like it should when it hits
4374 the edge of the window... Notice how the X isn't reset at the edge of the window:
4376 Blit! (CMD = 00010000)
4379 a1_base = 000E8008, a2_base = 0001FA68
4380 a1_x = 0000, a1_y = 0000, a1_frac_x = 0000, a1_frac_y = 0000, a2_x = 0000, a2_y = 0000
4381 a1_step_x = 0000, a1_step_y = 0000, a1_stepf_x = 0000, a1_stepf_y = 0000, a2_step_x = 0000, a2_step_y = 0000
4382 a1_inc_x = 0000, a1_inc_y = 0000, a1_incf_x = 0000, a1_incf_y = 0000
4383 a1_win_x = 0000, a1_win_y = 0000, a2_mask_x = 0000, a2_mask_y = 0000
4384 a2_mask=F a1add=+phr/+0 a2add=+phr/+0
4385 a1_pixsize = 5, a2_pixsize = 5
4386 srcd=7717771777177717 dstd=0000000000000000 patd=7730773077307730 iinc=00000000
4387 srcz1=0000000000000000 srcz2=0000000000000000 dstz=0000000000000000 zinc=00000000, coll=0
4389 [in=T a1f=F a1=F zf=F z=F a2=F iif=F iii=F izf=F izi=F]
4390 Entering INNER state...
4391 Entering DWRITE state... Dest write address/pix address: 000E8008/0 [7730773077307730] (icount=009E, inc=2)
4392 srcz=0000000000000000][dcomp=AA zcomp=00 dbinh=00]
4393 [srcz=0000000000000000 dstz=0000000000000000 zwdata=0000000000000000 mask=7FFF]
4394 [dstart=0 dend=40 pwidth=8 srcshift=0][daas=0 dabs=0 dam=7 ds=0 daq=F]
4395 Entering A1_ADD state [a1_x=0000, a1_y=0000, addasel=0, addbsel=0, modx=1, addareg=F, adda_xconst=1, adda_yconst=0]...
4396 Entering DWRITE state... Dest write address/pix address: 000E8018/0 [7730773077307730] (icount=009C, inc=2)
4397 srcz=0000000000000000][dcomp=AA zcomp=00 dbinh=00]
4398 [srcz=0000000000000000 dstz=0000000000000000 zwdata=0000000000000000 mask=7FFF]
4399 [dstart=0 dend=40 pwidth=8 srcshift=0][daas=0 dabs=0 dam=7 ds=0 daq=F]
4400 Entering A1_ADD state [a1_x=0002, a1_y=0000, addasel=0, addbsel=0, modx=1, addareg=F, adda_xconst=1, adda_yconst=0]...
4404 Entering A1_ADD state [a1_x=009C, a1_y=0000, addasel=0, addbsel=0, modx=1, addareg=F, adda_xconst=1, adda_yconst=0]...
4405 Entering DWRITE state... Dest write address/pix address: 000E84F8/0 [7730773077307730] (icount=0000, inc=2)
4406 srcz=0000000000000000][dcomp=AA zcomp=00 dbinh=00]
4407 [srcz=0000000000000000 dstz=0000000000000000 zwdata=0000000000000000 mask=7FFF]
4408 [dstart=0 dend=40 pwidth=8 srcshift=0][daas=0 dabs=0 dam=7 ds=0 daq=F]
4409 Entering A1_ADD state [a1_x=009E, a1_y=0000, addasel=0, addbsel=0, modx=1, addareg=F, adda_xconst=1, adda_yconst=0]...
4410 Entering IDLE_INNER state...
4412 Leaving INNER state... (ocount=0104)
4413 [in=T a1f=F a1=F zf=F z=F a2=F iif=F iii=F izf=F izi=F]
4415 Entering INNER state...
4416 Entering DWRITE state... Dest write address/pix address: 000E8508/0 [7730773077307730] (icount=009E, inc=2)
4417 srcz=0000000000000000][dcomp=AA zcomp=00 dbinh=00]
4418 [srcz=0000000000000000 dstz=0000000000000000 zwdata=0000000000000000 mask=7FFF]
4419 [dstart=0 dend=40 pwidth=8 srcshift=0][daas=0 dabs=0 dam=7 ds=0 daq=F]
4420 Entering A1_ADD state [a1_x=00A0, a1_y=0000, addasel=0, addbsel=0, modx=1, addareg=F, adda_xconst=1, adda_yconst=0]...
4421 Entering DWRITE state... Dest write address/pix address: 000E8518/0 [7730773077307730] (icount=009C, inc=2)
4422 srcz=0000000000000000][dcomp=AA zcomp=00 dbinh=00]
4423 [srcz=0000000000000000 dstz=0000000000000000 zwdata=0000000000000000 mask=7FFF]
4424 [dstart=0 dend=40 pwidth=8 srcshift=0][daas=0 dabs=0 dam=7 ds=0 daq=F]
4425 Entering A1_ADD state [a1_x=00A2, a1_y=0000, addasel=0, addbsel=0, modx=1, addareg=F, adda_xconst=1, adda_yconst=0]...
4431 #ifdef VERBOSE_BLITTER_LOGGING
4434 //printf(" Entering A1_ADD state [addasel=%X, addbsel=%X, modx=%X, addareg=%s, adda_xconst=%u, adda_yconst=%s]...\n", addasel, addbsel, modx, (addareg ? "T" : "F"), adda_xconst, (adda_yconst ? "1" : "0"));
4435 printf(" Entering A1_ADD state [a1_x=%04X, a1_y=%04X, addasel=%X, addbsel=%X, modx=%X, addareg=%s, adda_xconst=%u, adda_yconst=%s]...\n", a1_x, a1_y, addasel, addbsel, modx, (addareg ? "T" : "F"), adda_xconst, (adda_yconst ? "1" : "0"));
4439 int16 adda_x, adda_y, addb_x, addb_y, data_x, data_y, addq_x, addq_y;
4440 ADDAMUX(adda_x, adda_y, addasel, a1_step_x, a1_step_y, a1_stepf_x, a1_stepf_y, a2_step_x, a2_step_y,
4441 a1_inc_x, a1_inc_y, a1_incf_x, a1_incf_y, adda_xconst, adda_yconst, addareg, suba_x, suba_y);
4442 ADDBMUX(addb_x, addb_y, addbsel, a1_x, a1_y, a2_x, a2_y, a1_frac_x, a1_frac_y);
4443 ADDRADD(addq_x, addq_y, a1fracldi, adda_x, adda_y, addb_x, addb_y, modx, suba_x, suba_y);
4445 #if 0//def VERBOSE_BLITTER_LOGGING
4448 printf(" [adda_x=%d, adda_y=%d, addb_x=%d, addb_y=%d, addq_x=%d, addq_y=%d]\n", adda_x, adda_y, addb_x, addb_y, addq_x, addq_y);
4452 //Now, write to what???
4453 //a2ptrld comes from a2ptrldi...
4454 //I believe it's addbsel that determines the writeback...
4455 // This is where atick[0] & [1] come in, in determining which part (fractional, integer)
4456 // gets written to...
4459 //Kludge, to get A1 channel increment working...
4462 a1_frac_x = addq_x, a1_frac_y = addq_y;
4464 addasel = 2, addbsel = 0, a1fracldi = false;
4465 ADDAMUX(adda_x, adda_y, addasel, a1_step_x, a1_step_y, a1_stepf_x, a1_stepf_y, a2_step_x, a2_step_y,
4466 a1_inc_x, a1_inc_y, a1_incf_x, a1_incf_y, adda_xconst, adda_yconst, addareg, suba_x, suba_y);
4467 ADDBMUX(addb_x, addb_y, addbsel, a1_x, a1_y, a2_x, a2_y, a1_frac_x, a1_frac_y);
4468 ADDRADD(addq_x, addq_y, a1fracldi, adda_x, adda_y, addb_x, addb_y, modx, suba_x, suba_y);
4470 a1_x = addq_x, a1_y = addq_y;
4473 a1_x = addq_x, a1_y = addq_y;
4478 #ifdef VERBOSE_BLITTER_LOGGING
4481 //printf(" Entering A2_ADD state [addasel=%X, addbsel=%X, modx=%X, addareg=%s, adda_xconst=%u, adda_yconst=%s]...\n", addasel, addbsel, modx, (addareg ? "T" : "F"), adda_xconst, (adda_yconst ? "1" : "0"));
4482 printf(" Entering A2_ADD state [a2_x=%04X, a2_y=%04X, addasel=%X, addbsel=%X, modx=%X, addareg=%s, adda_xconst=%u, adda_yconst=%s]...\n", a2_x, a2_y, addasel, addbsel, modx, (addareg ? "T" : "F"), adda_xconst, (adda_yconst ? "1" : "0"));
4486 //void ADDAMUX(int16 &adda_x, int16 &adda_y, uint8 addasel, int16 a1_step_x, int16 a1_step_y,
4487 // int16 a1_stepf_x, int16 a1_stepf_y, int16 a2_step_x, int16 a2_step_y,
4488 // int16 a1_inc_x, int16 a1_inc_y, int16 a1_incf_x, int16 a1_incf_y, uint8 adda_xconst,
4489 // bool adda_yconst, bool addareg, bool suba_x, bool suba_y)
4490 //void ADDBMUX(int16 &addb_x, int16 &addb_y, uint8 addbsel, int16 a1_x, int16 a1_y,
4491 // int16 a2_x, int16 a2_y, int16 a1_frac_x, int16 a1_frac_y)
4492 //void ADDRADD(int16 &addq_x, int16 &addq_y, bool a1fracldi,
4493 // int16 adda_x, int16 adda_y, int16 addb_x, int16 addb_y, uint8 modx, bool suba_x, bool suba_y)
4494 //void DATAMUX(int16 &data_x, int16 &data_y, uint32 gpu_din, int16 addq_x, int16 addq_y, bool addqsel)
4495 int16 adda_x, adda_y, addb_x, addb_y, data_x, data_y, addq_x, addq_y;
4496 ADDAMUX(adda_x, adda_y, addasel, a1_step_x, a1_step_y, a1_stepf_x, a1_stepf_y, a2_step_x, a2_step_y,
4497 a1_inc_x, a1_inc_y, a1_incf_x, a1_incf_y, adda_xconst, adda_yconst, addareg, suba_x, suba_y);
4498 ADDBMUX(addb_x, addb_y, addbsel, a1_x, a1_y, a2_x, a2_y, a1_frac_x, a1_frac_y);
4499 ADDRADD(addq_x, addq_y, a1fracldi, adda_x, adda_y, addb_x, addb_y, modx, suba_x, suba_y);
4501 #if 0//def VERBOSE_BLITTER_LOGGING
4504 printf(" [adda_x=%d, adda_y=%d, addb_x=%d, addb_y=%d, addq_x=%d, addq_y=%d]\n", adda_x, adda_y, addb_x, addb_y, addq_x, addq_y);
4508 //Now, write to what???
4509 //a2ptrld comes from a2ptrldi...
4510 //I believe it's addbsel that determines the writeback...
4516 Flags: SRCEN CLIP_A1 UPDA1 UPDA1F UPDA2 DSTA2 GOURZ ZMODE=0 LFUFUNC=C SRCSHADE
4518 a1_base = 0015B000, a2_base = 0014B000
4519 a1_x = 0000, a1_y = 0000, a1_frac_x = 8000, a1_frac_y = 8000, a2_x = 001F, a2_y = 0038
4520 a1_step_x = FFFFFFC0, a1_step_y = 0001, a1_stepf_x = 0000, a1_stepf_y = 2AAA, a2_step_x = FFFFFFC0, a2_step_y = 0001
4521 a1_inc_x = 0001, a1_inc_y = 0000, a1_incf_x = 0000, a1_incf_y = 0000
4522 a1_win_x = 0040, a1_win_y = 0040, a2_mask_x = 0000, a2_mask_y = 0000
4523 a2_mask=F a1add=+inc/+0 a2add=+1/+0
4524 a1_pixsize = 4, a2_pixsize = 4
4525 srcd=FF00FF00FF00FF00 dstd=0000000000000000 patd=0000000000000000 iinc=00000000
4526 srcz1=0000000000000000 srcz2=0000000000000000 dstz=0000000000000000 zinc=00000000, col=0
4528 [in=T a1f=F a1=F zf=F z=F a2=F iif=F iii=F izf=F izi=F]
4529 Entering INNER state...
4530 Entering SREAD state... Source read address/pix address: 0015B000/0 [6505650565056505]
4531 Entering A1_ADD state [a1_x=0000, a1_y=0000, addasel=3, addbsel=2, modx=0, addareg=T, adda_xconst=7, adda_yconst=0]...
4532 Entering DWRITE state...
4533 Dest write address/pix address: 0014E83E/0 [dstart=0 dend=10 pwidth=8 srcshift=0][daas=4 dabs=5 dam=7 ds=1 daq=F] [0000000000006505] (icount=003F, inc=1)
4534 Entering A2_ADD state [a2_x=001F, a2_y=0038, addasel=0, addbsel=1, modx=0, addareg=F, adda_xconst=0, adda_yconst=0]...
4535 Entering SREAD state... Source read address/pix address: 0015B000/0 [6505650565056505]
4536 Entering A1_ADD state [a1_x=FFFF8000, a1_y=FFFF8000, addasel=3, addbsel=2, modx=0, addareg=T, adda_xconst=7, adda_yconst=0]...
4537 Entering DWRITE state...
4538 Dest write address/pix address: 0014E942/0 [dstart=0 dend=10 pwidth=8 srcshift=0][daas=4 dabs=5 dam=7 ds=1 daq=F] [0000000000006505] (icount=003E, inc=1)
4539 Entering A2_ADD state [a2_x=0021, a2_y=0039, addasel=0, addbsel=1, modx=0, addareg=F, adda_xconst=0, adda_yconst=0]...
4540 Entering SREAD state... Source read address/pix address: 0015B000/0 [6505650565056505]
4541 Entering A1_ADD state [a1_x=FFFF8000, a1_y=FFFF8000, addasel=3, addbsel=2, modx=0, addareg=T, adda_xconst=7, adda_yconst=0]...
4542 Entering DWRITE state...
4543 Dest write address/pix address: 0014EA46/0 [dstart=0 dend=10 pwidth=8 srcshift=0][daas=4 dabs=5 dam=7 ds=1 daq=F] [0000000000006505] (icount=003D, inc=1)
4544 Entering A2_ADD state [a2_x=0023, a2_y=003A, addasel=0, addbsel=1, modx=0, addareg=F, adda_xconst=0, adda_yconst=0]...
4545 Entering SREAD state... Source read address/pix address: 0015B000/0 [6505650565056505]
4546 Entering A1_ADD state [a1_x=FFFF8000, a1_y=FFFF8000, addasel=3, addbsel=2, modx=0, addareg=T, adda_xconst=7, adda_yconst=0]...
4547 Entering DWRITE state...
4548 Dest write address/pix address: 0014EB4A/0 [dstart=0 dend=10 pwidth=8 srcshift=0][daas=4 dabs=5 dam=7 ds=1 daq=F] [0000000000006505] (icount=003C, inc=1)
4549 Entering A2_ADD state [a2_x=0025, a2_y=003B, addasel=0, addbsel=1, modx=0, addareg=F, adda_xconst=0, adda_yconst=0]...
4551 Entering SREAD state... Source read address/pix address: 0015B000/0 [6505650565056505]
4552 Entering A1_ADD state [a1_x=FFFF8000, a1_y=FFFF8000, addasel=3, addbsel=2, modx=0, addareg=T, adda_xconst=7, adda_yconst=0]...
4553 Entering DWRITE state...
4554 Dest write address/pix address: 0015283A/0 [dstart=0 dend=10 pwidth=8 srcshift=0][daas=4 dabs=5 dam=7 ds=1 daq=F] [0000000000006505] (icount=0000, inc=1)
4555 Entering A2_ADD state [a2_x=009D, a2_y=0077, addasel=0, addbsel=1, modx=0, addareg=F, adda_xconst=0, adda_yconst=0]...
4556 Entering IDLE_INNER state...
4557 Leaving INNER state... (ocount=0036)
4558 [in=F a1f=T a1=F zf=F z=F a2=F iif=F iii=F izf=F izi=F]
4559 Entering A1FUPDATE state...
4560 [in=F a1f=F a1=T zf=F z=F a2=F iif=F iii=F izf=F izi=F]
4561 Entering A1UPDATE state... (-32768/-32768 -> 32704/-32767)
4562 [in=F a1f=F a1=F zf=F z=F a2=T iif=F iii=F izf=F izi=F]
4563 Entering A2UPDATE state... (159/120 -> 95/121)
4564 [in=T a1f=F a1=F zf=F z=F a2=F iif=F iii=F izf=F izi=F]
4565 Entering INNER state...
4568 #ifdef VERBOSE_BLITTER_LOGGING
4571 printf(" Leaving INNER state...");
4576 // The outer counter is updated here as well on the clock cycle...
4578 /* the inner loop is started whenever another state is about to
4579 cause the inner state to go active */
4580 //Instart := ND7 (instart, innert[0], innert[2..7]);
4582 //Actually, it's done only when inner gets asserted without the 2nd line of conditions
4583 //(inner AND !indone)
4585 //Since we don't get here until the inner loop is finished (indone = true) we can get
4586 //away with doing it here...!
4591 #ifdef VERBOSE_BLITTER_LOGGING
4594 printf(" (ocount=%04X)\n", ocount);
4602 #ifdef VERBOSE_BLITTER_LOGGING
4605 printf(" Entering A1FUPDATE state...\n");
4609 uint32 a1_frac_xt = (uint32)a1_frac_x + (uint32)a1_stepf_x;
4610 uint32 a1_frac_yt = (uint32)a1_frac_y + (uint32)a1_stepf_y;
4611 a1FracCInX = a1_frac_xt >> 16;
4612 a1FracCInY = a1_frac_yt >> 16;
4613 a1_frac_x = (uint16)(a1_frac_xt & 0xFFFF);
4614 a1_frac_y = (uint16)(a1_frac_yt & 0xFFFF);
4619 #ifdef VERBOSE_BLITTER_LOGGING
4622 printf(" Entering A1UPDATE state... (%d/%d -> ", a1_x, a1_y);
4626 a1_x += a1_step_x + a1FracCInX;
4627 a1_y += a1_step_y + a1FracCInY;
4628 #ifdef VERBOSE_BLITTER_LOGGING
4631 printf("%d/%d)\n", a1_x, a1_y);
4639 #ifdef VERBOSE_BLITTER_LOGGING
4642 printf(" Entering A2UPDATE state... (%d/%d -> ", a2_x, a2_y);
4648 #ifdef VERBOSE_BLITTER_LOGGING
4651 printf("%d/%d)\n", a2_x, a2_y);
4658 // We never get here! !!! FIX !!!
4660 #ifdef VERBOSE_BLITTER_LOGGING
4663 printf("Done!\na1_x=%04X a1_y=%04X a1_frac_x=%04X a1_frac_y=%04X a2_x=%04X a2_y%04X\n",
4664 GET16(blitter_ram, A1_PIXEL + 2),
4665 GET16(blitter_ram, A1_PIXEL + 0),
4666 GET16(blitter_ram, A1_FPIXEL + 2),
4667 GET16(blitter_ram, A1_FPIXEL + 0),
4668 GET16(blitter_ram, A2_PIXEL + 2),
4669 GET16(blitter_ram, A2_PIXEL + 0));
4674 // Write values back to registers (in real blitter, these are continuously updated)
4675 SET16(blitter_ram, A1_PIXEL + 2, a1_x);
4676 SET16(blitter_ram, A1_PIXEL + 0, a1_y);
4677 SET16(blitter_ram, A1_FPIXEL + 2, a1_frac_x);
4678 SET16(blitter_ram, A1_FPIXEL + 0, a1_frac_y);
4679 SET16(blitter_ram, A2_PIXEL + 2, a2_x);
4680 SET16(blitter_ram, A2_PIXEL + 0, a2_y);
4682 #ifdef VERBOSE_BLITTER_LOGGING
4685 printf("Writeback!\na1_x=%04X a1_y=%04X a1_frac_x=%04X a1_frac_y=%04X a2_x=%04X a2_y%04X\n",
4686 GET16(blitter_ram, A1_PIXEL + 2),
4687 GET16(blitter_ram, A1_PIXEL + 0),
4688 GET16(blitter_ram, A1_FPIXEL + 2),
4689 GET16(blitter_ram, A1_FPIXEL + 0),
4690 GET16(blitter_ram, A2_PIXEL + 2),
4691 GET16(blitter_ram, A2_PIXEL + 0));
4698 int16 a1_x = (int16)GET16(blitter_ram, A1_PIXEL + 2);
4699 int16 a1_y = (int16)GET16(blitter_ram, A1_PIXEL + 0);
4700 uint16 a1_frac_x = GET16(blitter_ram, A1_FPIXEL + 2);
4701 uint16 a1_frac_y = GET16(blitter_ram, A1_FPIXEL + 0);
4702 int16 a2_x = (int16)GET16(blitter_ram, A2_PIXEL + 2);
4703 int16 a2_y = (int16)GET16(blitter_ram, A2_PIXEL + 0);
4705 Seems that the ending a1_x should be written between blits, but it doesn't seem to be...
4707 Blit! (CMD = 01800000)
4710 a1_base = 00050000, a2_base = 00070000
4711 a1_x = 0000, a1_y = 0000, a1_frac_x = 49CD, a1_frac_y = 0000, a2_x = 0033, a2_y = 0001
4712 a1_step_x = 0000, a1_step_y = 0000, a1_stepf_x = 939A, a1_stepf_y = 0000, a2_step_x = 0000, a2_step_y = 0000
4713 a1_inc_x = 0000, a1_inc_y = 0000, a1_incf_x = 0000, a1_incf_y = 0000
4714 a1_win_x = 0100, a1_win_y = 0020, a2_mask_x = 0000, a2_mask_y = 0000
4715 a2_mask=F a1add=+phr/+0 a2add=+phr/+0
4716 a1_pixsize = 4, a2_pixsize = 3
4717 srcd=DEDEDEDEDEDEDEDE dstd=0000000000000000 patd=0000000000000000 iinc=00000000
4718 srcz1=0000000000000000 srcz2=0000000000000000 dstz=0000000000000000 zinc=00000000, coll=0
4721 Blit! (CMD = 01800000)
4724 a1_base = 00050000, a2_base = 00070000
4725 a1_x = 0000, a1_y = 0000, a1_frac_x = 49CD, a1_frac_y = 0000, a2_x = 0033, a2_y = 0001
4726 a1_step_x = 0000, a1_step_y = 0000, a1_stepf_x = 939A, a1_stepf_y = 0000, a2_step_x = 0000, a2_step_y = 0000
4727 a1_inc_x = 0000, a1_inc_y = 0000, a1_incf_x = 0000, a1_incf_y = 0000
4728 a1_win_x = 0100, a1_win_y = 0020, a2_mask_x = 0000, a2_mask_y = 0000
4729 a2_mask=F a1add=+phr/+0 a2add=+phr/+0
4730 a1_pixsize = 4, a2_pixsize = 3
4731 srcd=D6D6D6D6D6D6D6D6 dstd=0000000000000000 patd=0000000000000000 iinc=00000000
4732 srcz1=0000000000000000 srcz2=0000000000000000 dstz=0000000000000000 zinc=00000000, coll=0
4738 // Various pieces of the blitter puzzle are teased out here...
4744 INT24/ address // byte address
4745 pixa[0..2] // bit part of address, un-pipe-lined
4761 apipe // load address pipe-line latch
4762 clk // co-processor clock
4763 gena2 // generate A2 as opposed to A1
4764 zaddr // generate Z address
4768 void ADDRGEN(uint32 &address, uint32 &pixa, bool gena2, bool zaddr,
4769 uint16 a1_x, uint16 a1_y, uint32 a1_base, uint8 a1_pitch, uint8 a1_pixsize, uint8 a1_width, uint8 a1_zoffset,
4770 uint16 a2_x, uint16 a2_y, uint32 a2_base, uint8 a2_pitch, uint8 a2_pixsize, uint8 a2_width, uint8 a2_zoffset)
4772 // uint16 x = (gena2 ? a2_x : a1_x) & 0x7FFF;
4773 uint16 x = (gena2 ? a2_x : a1_x) & 0xFFFF; // Actually uses all 16 bits to generate address...!
4774 uint16 y = (gena2 ? a2_y : a1_y) & 0x0FFF;
4775 uint8 width = (gena2 ? a2_width : a1_width);
4776 uint8 pixsize = (gena2 ? a2_pixsize : a1_pixsize);
4777 uint8 pitch = (gena2 ? a2_pitch : a1_pitch);
4778 uint32 base = (gena2 ? a2_base : a1_base) >> 3;//Only upper 21 bits are passed around the bus? Seems like it...
4779 uint8 zoffset = (gena2 ? a2_zoffset : a1_zoffset);
4781 uint32 ytm = ((uint32)y << 2) + (width & 0x02 ? (uint32)y << 1 : 0) + (width & 0x01 ? (uint32)y : 0);
4783 uint32 ya = (ytm << (width >> 2)) >> 2;
4787 /*uint32*/ pixa = pa << pixsize;
4789 uint8 pt = ((pitch & 0x01) && !(pitch & 0x02) ? 0x01 : 0x00)
4790 | (!(pitch & 0x01) && (pitch & 0x02) ? 0x02 : 0x00);
4791 // uint32 phradr = pixa << pt;
4792 uint32 phradr = (pixa >> 6) << pt;
4793 uint32 shup = (pitch == 0x03 ? (pixa >> 6) : 0);
4795 uint8 za = (zaddr ? zoffset : 0) & 0x03;
4796 // uint32 addr = za + (phradr & 0x07) + (shup << 1) + base;
4797 uint32 addr = za + phradr + (shup << 1) + base;
4798 /*uint32*/ address = ((pixa & 0x38) >> 3) | ((addr & 0x1FFFFF) << 3);
4799 #if 0//def VERBOSE_BLITTER_LOGGING
4802 printf(" [gena2=%s, x=%04X, y=%04X, w=%1X, pxsz=%1X, ptch=%1X, b=%08X, zoff=%1X]\n", (gena2 ? "T" : "F"), x, y, width, pixsize, pitch, base, zoffset);
4803 printf(" [ytm=%X, ya=%X, pa=%X, pixa=%X, pt=%X, phradr=%X, shup=%X, za=%X, addr=%X, address=%X]\n", ytm, ya, pa, pixa, pt, phradr, shup, za, addr, address);
4809 Entering INNER state...
4810 [gena2=T, x=0002, y=0000, w=20, pxsz=4, ptch=0, b=000012BA, zoff=0]
4811 [ytm=0, ya=0, pa=2, pixa=20, pt=0, phradr=0, shup=0, za=0, addr=12BA, address=95D4]
4812 Entering SREADX state... [dstart=0 dend=20 pwidth=8 srcshift=20]
4813 Source extra read address/pix address: 000095D4/0 [0000001C00540038]
4814 Entering A2_ADD state [a2_x=0002, a2_y=0000, addasel=0, addbsel=1, modx=2, addareg=F, adda_xconst=2, adda_yconst=0]...
4815 [gena2=T, x=0004, y=0000, w=20, pxsz=4, ptch=0, b=000012BA, zoff=0]
4816 [ytm=0, ya=0, pa=4, pixa=40, pt=0, phradr=1, shup=0, za=0, addr=12BB, address=95D8]
4817 Entering SREAD state... [dstart=0 dend=20 pwidth=8 srcshift=0]
4818 Source read address/pix address: 000095D8/0 [0054003800009814]
4819 Entering A2_ADD state [a2_x=0004, a2_y=0000, addasel=0, addbsel=1, modx=2, addareg=F, adda_xconst=2, adda_yconst=0]...
4820 [gena2=F, x=0000, y=0000, w=20, pxsz=4, ptch=0, b=00006E52, zoff=0]
4821 [ytm=0, ya=0, pa=0, pixa=0, pt=0, phradr=0, shup=0, za=0, addr=6E52, address=37290]
4822 Entering DWRITE state...
4823 Dest write address/pix address: 00037290/0 [dstart=0 dend=20 pwidth=8 srcshift=0] (icount=026E, inc=4)
4824 Entering A1_ADD state [a1_x=0000, a1_y=0000, addasel=0, addbsel=0, modx=2, addareg=F, adda_xconst=2, adda_yconst=0]...
4825 [gena2=T, x=0008, y=0000, w=20, pxsz=4, ptch=0, b=000012BA, zoff=0]
4826 [ytm=0, ya=0, pa=8, pixa=80, pt=0, phradr=2, shup=0, za=0, addr=12BC, address=95E0]
4830 Entering SREAD state...
4831 [gena2=T, x=0004, y=0000, w=20, pxsz=4, ptch=0, b=000010AC, zoff=0]
4832 [ytm=0, ya=0, pa=4, pixa=0, pt=0, phradr=40, shup=0, za=0, addr=10AC, address=8560]
4833 Source read address/pix address: 00008560/0 [8C27981B327E00F0]
4835 2nd pass (still wrong):
4836 Entering SREAD state...
4837 [gena2=T, x=0004, y=0000, w=20, pxsz=4, ptch=0, b=000010AC, zoff=0]
4838 [ytm=0, ya=0, pa=4, pixa=0, pt=0, phradr=40, shup=0, za=0, addr=10EC, address=8760]
4839 Source read address/pix address: 00008760/0 [00E06DC04581880C]
4842 Entering SREAD state...
4843 [gena2=T, x=0004, y=0000, w=20, pxsz=4, ptch=0, b=000010AC, zoff=0]
4844 [ytm=0, ya=0, pa=4, pixa=0, pt=0, phradr=1, shup=0, za=0, addr=10AD, address=8568]
4845 Source read address/pix address: 00008568/0 [6267981A327C00F0]
4847 OK, now we're back into incorrect (or is it?):
4848 Entering SREADX state... [dstart=0 dend=20 pwidth=8 srcshift=20]
4849 Source extra read address/pix address: 000095D4/0 [0000 001C 0054 0038]
4850 Entering A2_ADD state [a2_x=0002, a2_y=0000, addasel=0, addbsel=1, modx=2, addareg=F, adda_xconst=2, adda_yconst=0]...
4851 Entering SREAD state... [dstart=0 dend=20 pwidth=8 srcshift=0]
4852 Source read address/pix address: 000095D8/0 [0054 0038 0000 9814]
4853 Entering A2_ADD state [a2_x=0004, a2_y=0000, addasel=0, addbsel=1, modx=2, addareg=F, adda_xconst=2, adda_yconst=0]...
4854 I think this may be correct...!
4859 // source and destination address update conditions
4861 Sraat0 := AN2 (sraat[0], sreadxi, srcenz\);
4862 Sraat1 := AN2 (sraat[1], sreadi, srcenz\);
4863 Srca_addi := OR4 (srca_addi, szreadxi, szreadi, sraat[0..1]);
4864 Srca_add := FD1Q (srca_add, srca_addi, clk);
4866 Dstaat := AN2 (dstaat, dwritei, dstwrz\);
4867 Dsta_addi := OR2 (dsta_addi, dzwritei, dstaat);
4868 // Dsta_add := FD1Q (dsta_add, dsta_addi, clk);
4870 // source and destination address generate conditions
4872 Gensrc := OR4 (gensrc, sreadxi, szreadxi, sreadi, szreadi);
4873 Gendst := OR4 (gendst, dreadi, dzreadi, dwritei, dzwritei);
4874 Dsta2\ := INV1 (dsta2\, dsta2);
4875 Gena2t0 := NAN2 (gena2t[0], gensrc, dsta2\);
4876 Gena2t1 := NAN2 (gena2t[1], gendst, dsta2);
4877 Gena2i := NAN2 (gena2i, gena2t[0..1]);
4878 Gena2 := FD1QU (gena2, gena2i, clk);
4880 Zaddr := OR4 (zaddr, szreadx, szread, dzread, dzwrite);
4885 // Basically, the above translates to:
4886 bool srca_addi = (sreadxi && !srcenz) || (sreadi && !srcenz) || szreadxi || szreadi;
4888 bool dsta_addi = (dwritei && !dstwrz) || dzwritei;
4890 bool gensrc = sreadxi || szreadxi || sreadi || szreadi;
4891 bool gendst = dreadi || szreadi || dwritei || dzwritei;
4892 bool gena2i = (gensrc && !dsta2) || (gendst && dsta2);
4894 bool zaddr = szreadx || szread || dzread || dzwrite;
4898 // source data reads
4900 Srcdpset\ := NAN2 (srcdpset\, readreq, sread);
4901 Srcdpt1 := NAN2 (srcdpt[1], srcdpend, srcdack\);
4902 Srcdpt2 := NAN2 (srcdpt[2], srcdpset\, srcdpt[1]);
4903 Srcdpend := FD2Q (srcdpend, srcdpt[2], clk, reset\);
4905 Srcdxpset\ := NAN2 (srcdxpset\, readreq, sreadx);
4906 Srcdxpt1 := NAN2 (srcdxpt[1], srcdxpend, srcdxack\);
4907 Srcdxpt2 := NAN2 (srcdxpt[2], srcdxpset\, srcdxpt[1]);
4908 Srcdxpend := FD2Q (srcdxpend, srcdxpt[2], clk, reset\);
4910 Sdpend := OR2 (sdpend, srcdxpend, srcdpend);
4911 Srcdreadt := AN2 (srcdreadt, sdpend, read_ack);
4913 //2/9/92 - enhancement?
4914 //Load srcdread on the next tick as well to modify it in srcshade
4916 Srcdreadd := FD1Q (srcdreadd, srcdreadt, clk);
4917 Srcdread := AOR1 (srcdread, srcshade, srcdreadd, srcdreadt);
4921 Srczpset\ := NAN2 (srczpset\, readreq, szread);
4922 Srczpt1 := NAN2 (srczpt[1], srczpend, srczack\);
4923 Srczpt2 := NAN2 (srczpt[2], srczpset\, srczpt[1]);
4924 Srczpend := FD2Q (srczpend, srczpt[2], clk, reset\);
4926 Srczxpset\ := NAN2 (srczxpset\, readreq, szreadx);
4927 Srczxpt1 := NAN2 (srczxpt[1], srczxpend, srczxack\);
4928 Srczxpt2 := NAN2 (srczxpt[2], srczxpset\, srczxpt[1]);
4929 Srczxpend := FD2Q (srczxpend, srczxpt[2], clk, reset\);
4931 Szpend := OR2 (szpend, srczpend, srczxpend);
4932 Srczread := AN2 (srczread, szpend, read_ack);
4934 // destination data reads
4936 Dstdpset\ := NAN2 (dstdpset\, readreq, dread);
4937 Dstdpt0 := NAN2 (dstdpt[0], dstdpend, dstdack\);
4938 Dstdpt1 := NAN2 (dstdpt[1], dstdpset\, dstdpt[0]);
4939 Dstdpend := FD2Q (dstdpend, dstdpt[1], clk, reset\);
4940 Dstdread := AN2 (dstdread, dstdpend, read_ack);
4942 // destination zed reads
4944 Dstzpset\ := NAN2 (dstzpset\, readreq, dzread);
4945 Dstzpt0 := NAN2 (dstzpt[0], dstzpend, dstzack\);
4946 Dstzpt1 := NAN2 (dstzpt[1], dstzpset\, dstzpt[0]);
4947 Dstzpend := FD2Q (dstzpend, dstzpt[1], clk, reset\);
4948 Dstzread := AN2 (dstzread, dstzpend, read_ack);
4953 // Basically, the above translates to:
4954 bool srcdpend = (readreq && sread) || (srcdpend && !srcdack);
4955 bool srcdxpend = (readreq && sreadx) || (srcdxpend && !srcdxack);
4956 bool sdpend = srcxpend || srcdpend;
4957 bool srcdread = ((sdpend && read_ack) && srcshade) || (sdpend && read_ack);//the latter term is lookahead
4961 ////////////////////////////////////////////////////////////////////////////////////////////
4962 ////////////////////////////////////////////////////////////////////////////////////////////
4963 // Here's an important bit: The source data adder logic. Need to track down the inputs!!! //
4964 ////////////////////////////////////////////////////////////////////////////////////////////
4965 ////////////////////////////////////////////////////////////////////////////////////////////
4972 daddasel[0..2] // data adder input A selection
4977 initcin[0..3] // carry into the adders from the initializers
4978 initinc[0..63] // the initialisation increment
4979 initpix[0..15] // Data initialiser pixel value
4991 void ADDARRAY(uint16 * addq, uint8 daddasel, uint8 daddbsel, uint8 daddmode,
4992 uint64 dstd, uint32 iinc, uint8 initcin[], uint64 initinc, uint16 initpix,
4993 uint32 istep, uint64 patd, uint64 srcd, uint64 srcz1, uint64 srcz2,
4994 uint32 zinc, uint32 zstep)
4996 uint32 initpix2 = ((uint32)initpix << 16) | initpix;
4997 uint32 addalo[8], addahi[8];
4998 addalo[0] = dstd & 0xFFFFFFFF;
4999 addalo[1] = initpix2;
5002 addalo[4] = srcd & 0xFFFFFFFF;
5003 addalo[5] = patd & 0xFFFFFFFF;
5004 addalo[6] = srcz1 & 0xFFFFFFFF;
5005 addalo[7] = srcz2 & 0xFFFFFFFF;
5006 addahi[0] = dstd >> 32;
5007 addahi[1] = initpix2;
5010 addahi[4] = srcd >> 32;
5011 addahi[5] = patd >> 32;
5012 addahi[6] = srcz1 >> 32;
5013 addahi[7] = srcz2 >> 32;
5015 adda[0] = addalo[daddasel] & 0xFFFF;
5016 adda[1] = addalo[daddasel] >> 16;
5017 adda[2] = addahi[daddasel] & 0xFFFF;
5018 adda[3] = addahi[daddasel] >> 16;
5021 wordmux[0] = iinc & 0xFFFF;
5022 wordmux[1] = iinc >> 16;
5023 wordmux[2] = zinc & 0xFFFF;
5024 wordmux[3] = zinc >> 16;;
5025 wordmux[4] = istep & 0xFFFF;
5026 wordmux[5] = istep >> 16;;
5027 wordmux[6] = zstep & 0xFFFF;
5028 wordmux[7] = zstep >> 16;;
5029 uint16 word = wordmux[((daddbsel & 0x08) >> 1) | (daddbsel & 0x03)];
5031 bool dbsel2 = daddbsel & 0x04;
5032 bool iincsel = (daddbsel & 0x01) && !(daddbsel & 0x04);
5034 if (!dbsel2 && !iincsel)
5035 addb[0] = srcd & 0xFFFF,
5036 addb[1] = (srcd >> 16) & 0xFFFF,
5037 addb[2] = (srcd >> 32) & 0xFFFF,
5038 addb[3] = (srcd >> 48) & 0xFFFF;
5039 else if (dbsel2 && !iincsel)
5040 addb[0] = addb[1] = addb[2] = addb[3] = word;
5041 else if (!dbsel2 && iincsel)
5042 addb[0] = initinc & 0xFFFF,
5043 addb[1] = (initinc >> 16) & 0xFFFF,
5044 addb[2] = (initinc >> 32) & 0xFFFF,
5045 addb[3] = (initinc >> 48) & 0xFFFF;
5047 addb[0] = addb[1] = addb[2] = addb[3] = 0;
5049 uint8 cinsel = (daddmode >= 1 && daddmode <= 4 ? 1 : 0);
5051 static uint8 co[4];//These are preserved between calls...
5054 for(int i=0; i<4; i++)
5055 cin[i] = initcin[i] | (co[i] & cinsel);
5057 bool eightbit = daddmode & 0x02;
5058 bool sat = daddmode & 0x03;
5059 bool hicinh = ((daddmode & 0x03) == 0x03);
5061 //Note that the carry out is saved between calls to this function...
5062 for(int i=0; i<4; i++)
5063 ADD16SAT(addq[i], co[i], adda[i], addb[i], cin[i], sat, eightbit, hicinh);
5079 void ADD16SAT(uint16 &r, uint8 &co, uint16 a, uint16 b, uint8 cin, bool sat, bool eightbit, bool hicinh)
5083 printf("--> [sat=%s 8b=%s hicinh=%s] %04X + %04X (+ %u) = ", (sat ? "T" : "F"), (eightbit ? "T" : "F"), (hicinh ? "T" : "F"), a, b, cin);
5087 uint32 qt = (a & 0xFF) + (b & 0xFF) + cin;
5088 carry[0] = (qt & 0x0100 ? 1 : 0);
5089 uint16 q = qt & 0x00FF;
5090 carry[1] = (carry[0] && !eightbit ? carry[0] : 0);
5091 qt = (a & 0x0F00) + (b & 0x0F00) + (carry[1] << 8);
5092 carry[2] = (qt & 0x1000 ? 1 : 0);
5094 carry[3] = (carry[2] && !hicinh ? carry[2] : 0);
5095 qt = (a & 0xF000) + (b & 0xF000) + (carry[3] << 12);
5096 co = (qt & 0x10000 ? 1 : 0);
5099 uint8 btop = (eightbit ? (b & 0x0080) >> 7 : (b & 0x8000) >> 15);
5100 uint8 ctop = (eightbit ? carry[0] : co);
5102 bool saturate = sat && (btop ^ ctop);
5103 bool hisaturate = saturate && !eightbit;
5106 printf("bt=%u ct=%u s=%u hs=%u] ", btop, ctop, saturate, hisaturate);
5110 r = (saturate ? (ctop ? 0x00FF : 0x0000) : q & 0x00FF);
5111 r |= (hisaturate ? (ctop ? 0xFF00 : 0x0000) : q & 0xFF00);
5114 printf("%04X (co=%u)\n", r, co);
5119 /** ADDAMUX - Address adder input A selection *******************
5121 This module generates the data loaded into the address adder input A. This is
5122 the update value, and can be one of four registers : A1 step, A2 step, A1
5123 increment and A1 fraction. It can complement these values to perform
5124 subtraction, and it can generate constants to increment / decrement the window
5127 addasel[0..2] select the register to add
5129 000 A1 step integer part
5130 001 A1 step fraction part
5131 010 A1 increment integer part
5132 011 A1 increment fraction part
5135 adda_xconst[0..2] generate a power of 2 in the range 1-64 or all zeroes when
5138 addareg selects register value to be added as opposed to constant
5141 suba_x, suba_y complement the X and Y values
5167 void ADDAMUX(int16 &adda_x, int16 &adda_y, uint8 addasel, int16 a1_step_x, int16 a1_step_y,
5168 int16 a1_stepf_x, int16 a1_stepf_y, int16 a2_step_x, int16 a2_step_y,
5169 int16 a1_inc_x, int16 a1_inc_y, int16 a1_incf_x, int16 a1_incf_y, uint8 adda_xconst,
5170 bool adda_yconst, bool addareg, bool suba_x, bool suba_y)
5173 /*INT16/ addac_x, addac_y, addar_x, addar_y, addart_x, addart_y,
5174 INT16/ addas_x, addas_y, suba_x16, suba_y16
5178 Zero := TIE0 (zero);*/
5180 /* Multiplex the register terms */
5182 /*Addaselb[0-2] := BUF8 (addaselb[0-2], addasel[0-2]);
5183 Addart_x := MX4 (addart_x, a1_step_x, a1_stepf_x, a1_inc_x, a1_incf_x, addaselb[0..1]);
5184 Addar_x := MX2 (addar_x, addart_x, a2_step_x, addaselb[2]);
5185 Addart_y := MX4 (addart_y, a1_step_y, a1_stepf_y, a1_inc_y, a1_incf_y, addaselb[0..1]);
5186 Addar_y := MX2 (addar_y, addart_y, a2_step_y, addaselb[2]);*/
5188 ////////////////////////////////////// C++ CODE //////////////////////////////////////
5189 int16 xterm[4], yterm[4];
5190 xterm[0] = a1_step_x, xterm[1] = a1_stepf_x, xterm[2] = a1_inc_x, xterm[3] = a1_incf_x;
5191 yterm[0] = a1_step_y, yterm[1] = a1_stepf_y, yterm[2] = a1_inc_y, yterm[3] = a1_incf_y;
5192 int16 addar_x = (addasel & 0x04 ? a2_step_x : xterm[addasel & 0x03]);
5193 int16 addar_y = (addasel & 0x04 ? a2_step_y : yterm[addasel & 0x03]);
5194 //////////////////////////////////////////////////////////////////////////////////////
5196 /* Generate a constant value - this is a power of 2 in the range
5197 0-64, or zero. The control bits are adda_xconst[0..2], when they
5198 are all 1 the result is 0.
5199 Constants for Y can only be 0 or 1 */
5201 /*Addac_xlo := D38H (addac_x[0..6], unused[0], adda_xconst[0..2]);
5202 Unused[0] := DUMMY (unused[0]);
5204 Addac_x := JOIN (addac_x, addac_x[0..6], zero, zero, zero, zero, zero, zero, zero, zero, zero);
5205 Addac_y := JOIN (addac_y, adda_yconst, zero, zero, zero, zero, zero, zero, zero, zero, zero, zero,
5206 zero, zero, zero, zero, zero);*/
5207 ////////////////////////////////////// C++ CODE //////////////////////////////////////
5208 int16 addac_x = (adda_xconst == 0x07 ? 0 : 1 << adda_xconst);
5209 int16 addac_y = (adda_yconst ? 0x01 : 0);
5210 //////////////////////////////////////////////////////////////////////////////////////
5212 /* Select between constant value and register value */
5214 /*Addas_x := MX2 (addas_x, addac_x, addar_x, addareg);
5215 Addas_y := MX2 (addas_y, addac_y, addar_y, addareg);*/
5216 ////////////////////////////////////// C++ CODE //////////////////////////////////////
5217 int16 addas_x = (addareg ? addar_x : addac_x);
5218 int16 addas_y = (addareg ? addar_y : addac_y);
5219 //////////////////////////////////////////////////////////////////////////////////////
5221 /* Complement these values (complement flag gives adder carry in)*/
5223 /*Suba_x16 := JOIN (suba_x16, suba_x, suba_x, suba_x, suba_x, suba_x, suba_x, suba_x, suba_x, suba_x,
5224 suba_x, suba_x, suba_x, suba_x, suba_x, suba_x, suba_x);
5225 Suba_y16 := JOIN (suba_y16, suba_y, suba_y, suba_y, suba_y, suba_y, suba_y, suba_y, suba_y, suba_y,
5226 suba_y, suba_y, suba_y, suba_y, suba_y, suba_y, suba_y);
5227 Adda_x := EO (adda_x, suba_x16, addas_x);
5228 Adda_y := EO (adda_y, suba_y16, addas_y);*/
5229 ////////////////////////////////////// C++ CODE //////////////////////////////////////
5230 adda_x = addas_x ^ (suba_x ? 0xFFFF : 0x0000);
5231 adda_y = addas_y ^ (suba_y ? 0xFFFF : 0x0000);
5232 //////////////////////////////////////////////////////////////////////////////////////
5237 /** ADDBMUX - Address adder input B selection *******************
5239 This module selects the register to be updated by the address
5240 adder. This can be one of three registers, the A1 and A2
5241 pointers, or the A1 fractional part. It can also be zero, so that the step
5242 registers load directly into the pointers.
5257 INT16/ zero16 :LOCAL;
5259 void ADDBMUX(int16 &addb_x, int16 &addb_y, uint8 addbsel, int16 a1_x, int16 a1_y,
5260 int16 a2_x, int16 a2_y, int16 a1_frac_x, int16 a1_frac_y)
5263 /*Zero := TIE0 (zero);
5264 Zero16 := JOIN (zero16, zero, zero, zero, zero, zero, zero, zero,
5265 zero, zero, zero, zero, zero, zero, zero, zero, zero);
5266 Addbselb[0-1] := BUF8 (addbselb[0-1], addbsel[0-1]);
5267 Addb_x := MX4 (addb_x, a1_x, a2_x, a1_frac_x, zero16, addbselb[0..1]);
5268 Addb_y := MX4 (addb_y, a1_y, a2_y, a1_frac_y, zero16, addbselb[0..1]);*/
5269 ////////////////////////////////////// C++ CODE //////////////////////////////////////
5270 int16 xterm[4], yterm[4];
5271 xterm[0] = a1_x, xterm[1] = a2_x, xterm[2] = a1_frac_x, xterm[3] = 0;
5272 yterm[0] = a1_y, yterm[1] = a2_y, yterm[2] = a1_frac_y, yterm[3] = 0;
5273 addb_x = xterm[addbsel & 0x03];
5274 addb_y = yterm[addbsel & 0x03];
5275 //////////////////////////////////////////////////////////////////////////////////////
5280 /** DATAMUX - Address local data bus selection ******************
5282 Select between the adder output and the input data bus
5295 INT16/ gpu_lo, gpu_hi
5298 void DATAMUX(int16 &data_x, int16 &data_y, uint32 gpu_din, int16 addq_x, int16 addq_y, bool addqsel)
5300 /*Gpu_lo := JOIN (gpu_lo, gpu_din{0..15});
5301 Gpu_hi := JOIN (gpu_hi, gpu_din{16..31});
5303 Addqselb := BUF8 (addqselb, addqsel);
5304 Data_x := MX2 (data_x, gpu_lo, addq_x, addqselb);
5305 Data_y := MX2 (data_y, gpu_hi, addq_y, addqselb);*/
5306 ////////////////////////////////////// C++ CODE //////////////////////////////////////
5307 data_x = (addqsel ? addq_x : (int16)(gpu_din & 0xFFFF));
5308 data_y = (addqsel ? addq_y : (int16)(gpu_din >> 16));
5309 //////////////////////////////////////////////////////////////////////////////////////
5314 /******************************************************************
5318 Blitter Address Adder
5319 ---------------------
5320 The blitter address adder is a pair of sixteen bit adders, one
5321 each for X and Y. The multiplexing of the input terms is
5322 performed elsewhere, but this adder can also perform modulo
5323 arithmetic to align X-addresses onto phrase boundaries.
5325 modx[0..2] take values
5332 ******************************************************************/
5334 /*IMPORT duplo, tosh;
5340 a1fracldi // propagate address adder carry
5345 clk[0] // co-processor clock
5353 Zero := TIE0 (zero);*/
5354 void ADDRADD(int16 &addq_x, int16 &addq_y, bool a1fracldi,
5355 uint16 adda_x, uint16 adda_y, uint16 addb_x, uint16 addb_y, uint8 modx, bool suba_x, bool suba_y)
5358 /* Perform the addition */
5360 /*Adder_x := ADD16 (addqt_x[0..15], co_x, adda_x{0..15}, addb_x{0..15}, ci_x);
5361 Adder_y := ADD16 (addq_y[0..15], co_y, adda_y{0..15}, addb_y{0..15}, ci_y);*/
5363 /* latch carry and propagate if required */
5365 /*Cxt0 := AN2 (cxt[0], co_x, a1fracldi);
5366 Cxt1 := FD1Q (cxt[1], cxt[0], clk[0]);
5367 Ci_x := EO (ci_x, cxt[1], suba_x);
5369 yt0 := AN2 (cyt[0], co_y, a1fracldi);
5370 Cyt1 := FD1Q (cyt[1], cyt[0], clk[0]);
5371 Ci_y := EO (ci_y, cyt[1], suba_y);*/
5373 ////////////////////////////////////// C++ CODE //////////////////////////////////////
5374 //I'm sure the following will generate a bunch of warnings, but will have to do for now.
5375 static uint16 co_x = 0, co_y = 0; // Carry out has to propogate between function calls...
5376 uint16 ci_x = co_x ^ (suba_x ? 1 : 0);
5377 uint16 ci_y = co_y ^ (suba_y ? 1 : 0);
5378 uint32 addqt_x = adda_x + addb_x + ci_x;
5379 uint32 addqt_y = adda_y + addb_y + ci_y;
5380 co_x = ((addqt_x & 0x10000) && a1fracldi ? 1 : 0);
5381 co_y = ((addqt_y & 0x10000) && a1fracldi ? 1 : 0);
5382 //////////////////////////////////////////////////////////////////////////////////////
5384 /* Mask low bits of X to 0 if required */
5386 /*Masksel := D38H (unused[0], masksel[0..4], maskbit[5], unused[1], modx[0..2]);
5388 Maskbit[0-4] := OR2 (maskbit[0-4], masksel[0-4], maskbit[1-5]);
5390 Mask[0-5] := MX2 (addq_x[0-5], addqt_x[0-5], zero, maskbit[0-5]);
5392 Addq_x := JOIN (addq_x, addq_x[0..5], addqt_x[6..15]);
5393 Addq_y := JOIN (addq_y, addq_y[0..15]);*/
5395 ////////////////////////////////////// C++ CODE //////////////////////////////////////
5396 int16 mask[8] = { 0xFFFF, 0xFFFE, 0xFFFC, 0xFFF8, 0xFFF0, 0xFFE0, 0xFFC0, 0x0000 };
5397 addq_x = addqt_x & mask[modx];
5398 addq_y = addqt_y & 0xFFFF;
5399 //////////////////////////////////////////////////////////////////////////////////////
5401 //Unused[0-1] := DUMMY (unused[0-1]);
5408 wdata[0..63] // co-processor write data bus
5410 dcomp[0..7] // data byte equal flags
5411 srcd[0..7] // bits to use for bit to byte expansion
5412 zcomp[0..3] // output from Z comparators
5414 a1_x[0..1] // low two bits of A1 X pointer
5415 big_pix // pixel organisation is big-endian
5416 blitter_active // blitter is active
5417 clk // co-processor clock
5418 cmpdst // compare dest rather than source
5419 colorld // load the pattern color fields
5420 daddasel[0..2] // data adder input A selection
5421 daddbsel[0..3] // data adder input B selection
5422 daddmode[0..2] // data adder mode
5423 daddq_sel // select adder output vs. GPU data
5424 data[0..63] // co-processor read data bus
5425 data_ena // enable write data
5426 data_sel[0..1] // select data to write
5427 dbinh\[0..7] // byte oriented changed data inhibits
5428 dend[0..5] // end of changed write data zone
5429 dpipe[0..1] // load computed data pipe-line latch
5430 dstart[0..5] // start of changed write data zone
5431 dstdld[0..1] // dest data load (two halves)
5432 dstzld[0..1] // dest zed load (two halves)
5433 ext_int // enable extended precision intensity calculations
5434 INT32/ gpu_din // GPU data bus
5435 iincld // I increment load
5436 iincldx // alternate I increment load
5437 init_if // initialise I fraction phase
5438 init_ii // initialise I integer phase
5439 init_zf // initialise Z fraction phase
5440 intld[0..3] // computed intensities load
5441 istepadd // intensity step integer add
5442 istepfadd // intensity step fraction add
5443 istepld // I step load
5444 istepdld // I step delta load
5445 lfu_func[0..3] // LFU function code
5446 patdadd // pattern data gouraud add
5447 patdld[0..1] // pattern data load (two halves)
5448 pdsel[0..1] // select pattern data type
5449 phrase_mode // phrase write mode
5450 reload // transfer contents of double buffers
5451 reset\ // system reset
5452 srcd1ld[0..1] // source register 1 load (two halves)
5453 srcdread // source data read load enable
5454 srczread // source zed read load enable
5455 srcshift[0..5] // source alignment shift
5456 srcz1ld[0..1] // source zed 1 load (two halves)
5457 srcz2add // zed fraction gouraud add
5458 srcz2ld[0..1] // source zed 2 load (two halves)
5459 textrgb // texture mapping in RGB mode
5460 txtd[0..63] // data from the texture unit
5461 zedld[0..3] // computed zeds load
5462 zincld // Z increment load
5463 zmode[0..2] // Z comparator mode
5464 zpipe[0..1] // load computed zed pipe-line latch
5465 zstepadd // zed step integer add
5466 zstepfadd // zed step fraction add
5467 zstepld // Z step load
5468 zstepdld // Z step delta load
5472 void DATA(uint64 &wdata, uint8 &dcomp, uint8 &zcomp, bool &nowrite,
5473 bool big_pix, bool cmpdst, uint8 daddasel, uint8 daddbsel, uint8 daddmode, bool daddq_sel, uint8 data_sel,
5474 uint8 dbinh, uint8 dend, uint8 dstart, uint64 dstd, uint32 iinc, uint8 lfu_func, uint64 &patd, bool patdadd,
5475 bool phrase_mode, uint64 srcd, bool srcdread, bool srczread, bool srcz2add, uint8 zmode,
5476 bool bcompen, bool bkgwren, bool dcompen, uint8 icount, uint8 pixsize,
5477 uint64 &srcz, uint64 dstz, uint32 zinc)
5480 Stuff we absolutely *need* to have passed in/out:
5482 patdadd, dstd, srcd, patd, daddasel, daddbsel, daddmode, iinc, srcz1, srcz2, big_pix, phrase_mode, cmpdst
5484 changed patd (wdata I guess...) (Nope. We pass it back directly now...)
5487 // Source data registers
5489 /*Data_src := DATA_SRC (srcdlo, srcdhi, srcz[0..1], srczo[0..1], srczp[0..1], srcz1[0..1], srcz2[0..1], big_pix,
5490 clk, gpu_din, intld[0..3], local_data0, local_data1, srcd1ld[0..1], srcdread, srczread, srcshift[0..5],
5491 srcz1ld[0..1], srcz2add, srcz2ld[0..1], zedld[0..3], zpipe[0..1]);
5492 Srcd[0-7] := JOIN (srcd[0-7], srcdlo{0-7});
5493 Srcd[8-31] := JOIN (srcd[8-31], srcdlo{8-31});
5494 Srcd[32-63] := JOIN (srcd[32-63], srcdhi{0-31});*/
5496 // Destination data registers
5498 /*Data_dst := DATA_DST (dstd[0..63], dstz[0..1], clk, dstdld[0..1], dstzld[0..1], load_data[0..1]);
5499 Dstdlo := JOIN (dstdlo, dstd[0..31]);
5500 Dstdhi := JOIN (dstdhi, dstd[32..63]);*/
5502 // Pattern and Color data registers
5504 // Looks like this is simply another register file for the pattern data registers. No adding or anything funky
5505 // going on. Note that patd & patdv will output the same info.
5506 // Patdldl/h (patdld[0..1]) can select the local_data bus to overwrite the current pattern data...
5507 // Actually, it can be either patdld OR patdadd...!
5508 /*Data_pat := DATA_PAT (colord[0..15], int0dp[8..10], int1dp[8..10], int2dp[8..10], int3dp[8..10], mixsel[0..2],
5509 patd[0..63], patdv[0..1], clk, colorld, dpipe[0], ext_int, gpu_din, intld[0..3], local_data0, local_data1,
5510 patdadd, patdld[0..1], reload, reset\);
5511 Patdlo := JOIN (patdlo, patd[0..31]);
5512 Patdhi := JOIN (patdhi, patd[32..63]);*/
5514 // Multiplying data Mixer (NOT IN JAGUAR I)
5516 /*Datamix := DATAMIX (patdo[0..1], clk, colord[0..15], dpipe[1], dstd[0..63], int0dp[8..10], int1dp[8..10],
5517 int2dp[8..10], int3dp[8..10], mixsel[0..2], patd[0..63], pdsel[0..1], srcd[0..63], textrgb, txtd[0..63]);*/
5519 // Logic function unit
5521 /*Lfu := LFU (lfu[0..1], srcdlo, srcdhi, dstdlo, dstdhi, lfu_func[0..3]);*/
5522 ////////////////////////////////////// C++ CODE //////////////////////////////////////
5523 uint64 funcmask[2] = { 0, 0xFFFFFFFFFFFFFFFFLL };
5524 uint64 func0 = funcmask[lfu_func & 0x01];
5525 uint64 func1 = funcmask[(lfu_func >> 1) & 0x01];
5526 uint64 func2 = funcmask[(lfu_func >> 2) & 0x01];
5527 uint64 func3 = funcmask[(lfu_func >> 3) & 0x01];
5528 uint64 lfu = (~srcd & ~dstd & func0) | (~srcd & dstd & func1) | (srcd & ~dstd & func2) | (srcd & dstd & func3);
5529 //////////////////////////////////////////////////////////////////////////////////////
5531 // Increment and Step Registers
5533 // Does it do anything without the step add lines? Check it!
5534 // No. This is pretty much just a register file without the Jaguar II lines...
5535 /*Inc_step := INC_STEP (iinc, istep[0..31], zinc, zstep[0..31], clk, ext_int, gpu_din, iincld, iincldx, istepadd,
5536 istepfadd, istepld, istepdld, reload, reset\, zincld, zstepadd, zstepfadd, zstepld, zstepdld);
5537 Istep := JOIN (istep, istep[0..31]);
5538 Zstep := JOIN (zstep, zstep[0..31]);*/
5540 // Pixel data comparator
5542 /*Datacomp := DATACOMP (dcomp[0..7], cmpdst, dstdlo, dstdhi, patdlo, patdhi, srcdlo, srcdhi);*/
5543 ////////////////////////////////////// C++ CODE //////////////////////////////////////
5545 uint64 cmpd = patd ^ (cmpdst ? dstd : srcd);
5547 if ((cmpd & 0x00000000000000FFLL) == 0)
5549 if ((cmpd & 0x000000000000FF00LL) == 0)
5551 if ((cmpd & 0x0000000000FF0000LL) == 0)
5553 if ((cmpd & 0x00000000FF000000LL) == 0)
5555 if ((cmpd & 0x000000FF00000000LL) == 0)
5557 if ((cmpd & 0x0000FF0000000000LL) == 0)
5559 if ((cmpd & 0x00FF000000000000LL) == 0)
5561 if ((cmpd & 0xFF00000000000000LL) == 0)
5563 //////////////////////////////////////////////////////////////////////////////////////
5565 // Zed comparator for Z-buffer operations
5567 /*Zedcomp := ZEDCOMP (zcomp[0..3], srczp[0..1], dstz[0..1], zmode[0..2]);*/
5568 ////////////////////////////////////// C++ CODE //////////////////////////////////////
5569 //srczp is srcz pipelined, also it goes through a source shift as well...
5570 /*The shift is basically like so (each piece is 16 bits long):
5573 srcz1lolo srcz1lohi srcz1hilo srcz1hihi srcrz2lolo srcz2lohi srcz2hilo
5575 with srcshift bits 4 & 5 selecting the start position
5577 //So... basically what we have here is:
5580 if ((((srcz & 0x000000000000FFFFLL) < (dstz & 0x000000000000FFFFLL)) && (zmode & 0x01))
5581 || (((srcz & 0x000000000000FFFFLL) == (dstz & 0x000000000000FFFFLL)) && (zmode & 0x02))
5582 || (((srcz & 0x000000000000FFFFLL) > (dstz & 0x000000000000FFFFLL)) && (zmode & 0x04)))
5585 if ((((srcz & 0x00000000FFFF0000LL) < (dstz & 0x00000000FFFF0000LL)) && (zmode & 0x01))
5586 || (((srcz & 0x00000000FFFF0000LL) == (dstz & 0x00000000FFFF0000LL)) && (zmode & 0x02))
5587 || (((srcz & 0x00000000FFFF0000LL) > (dstz & 0x00000000FFFF0000LL)) && (zmode & 0x04)))
5590 if ((((srcz & 0x0000FFFF00000000LL) < (dstz & 0x0000FFFF00000000LL)) && (zmode & 0x01))
5591 || (((srcz & 0x0000FFFF00000000LL) == (dstz & 0x0000FFFF00000000LL)) && (zmode & 0x02))
5592 || (((srcz & 0x0000FFFF00000000LL) > (dstz & 0x0000FFFF00000000LL)) && (zmode & 0x04)))
5595 if ((((srcz & 0xFFFF000000000000LL) < (dstz & 0xFFFF000000000000LL)) && (zmode & 0x01))
5596 || (((srcz & 0xFFFF000000000000LL) == (dstz & 0xFFFF000000000000LL)) && (zmode & 0x02))
5597 || (((srcz & 0xFFFF000000000000LL) > (dstz & 0xFFFF000000000000LL)) && (zmode & 0x04)))
5600 //TEMP, TO TEST IF ZCOMP IS THE CULPRIT...
5601 //Nope, this is NOT the problem...
5603 // We'll do the comparison/bit/byte inhibits here, since that's they way it happens
5604 // in the real thing (dcomp goes out to COMP_CTRL and back into DATA through dbinh)...
5608 COMP_CTRL(dbinht, nowrite,
5609 bcompen, true/*big_pix*/, bkgwren, dcomp, dcompen, icount, pixsize, phrase_mode, srcd & 0xFF, zcomp);
5615 #ifdef VERBOSE_BLITTER_LOGGING
5618 printf("\n[dcomp=%02X zcomp=%02X dbinh=%02X]\n", dcomp, zcomp, dbinh);
5623 //////////////////////////////////////////////////////////////////////////////////////
5626 // The data initializer - allows all four initial values to be computed from one (NOT IN JAGUAR I)
5628 /*Datinit := DATINIT (initcin[0..3], initinc[0..63], initpix[0..15], a1_x[0..1], big_pix, clk, iinc, init_if, init_ii,
5629 init_zf, istep[0..31], zinc, zstep[0..31]);*/
5631 // Adder array for Z and intensity increments
5633 /*Addarray := ADDARRAY (addq[0..3], clk, daddasel[0..2], daddbsel[0..3], daddmode[0..2], dstdlo, dstdhi, iinc,
5634 initcin[0..3], initinc[0..63], initpix[0..15], istep, patdv[0..1], srcdlo, srcdhi, srcz1[0..1],
5635 srcz2[0..1], reset\, zinc, zstep);*/
5636 /*void ADDARRAY(uint16 * addq, uint8 daddasel, uint8 daddbsel, uint8 daddmode,
5637 uint64 dstd, uint32 iinc, uint8 initcin[], uint64 initinc, uint16 initpix,
5638 uint32 istep, uint64 patd, uint64 srcd, uint64 srcz1, uint64 srcz2,
5639 uint32 zinc, uint32 zstep)*/
5640 ////////////////////////////////////// C++ CODE //////////////////////////////////////
5642 uint8 initcin[4] = { 0, 0, 0, 0 };
5643 ADDARRAY(addq, daddasel, daddbsel, daddmode, dstd, iinc, initcin, 0, 0, 0, patd, srcd, 0, 0, 0, 0);
5645 //This is normally done asynchronously above (thru local_data) when in patdadd mode...
5646 //And now it's passed back to the caller to be persistent between calls...!
5647 //But it's causing some serious fuck-ups in T2K now... !!! FIX !!! [DONE--???]
5648 //Weird! It doesn't anymore...!
5650 patd = ((uint64)addq[3] << 48) | ((uint64)addq[2] << 32) | ((uint64)addq[1] << 16) | (uint64)addq[0];
5651 //////////////////////////////////////////////////////////////////////////////////////
5653 // Local data bus multiplexer
5655 /*Local_mux := LOCAL_MUX (local_data[0..1], load_data[0..1],
5656 addq[0..3], gpu_din, data[0..63], blitter_active, daddq_sel);
5657 Local_data0 := JOIN (local_data0, local_data[0]);
5658 Local_data1 := JOIN (local_data1, local_data[1]);*/
5659 ////////////////////////////////////// C++ CODE //////////////////////////////////////
5660 //////////////////////////////////////////////////////////////////////////////////////
5662 // Data output multiplexer and tri-state drive
5664 /*Data_mux := DATA_MUX (wdata[0..63], addq[0..3], big_pix, dstdlo, dstdhi, dstz[0..1], data_sel[0..1], data_ena,
5665 dstart[0..5], dend[0..5], dbinh\[0..7], lfu[0..1], patdo[0..1], phrase_mode, srczo[0..1]);*/
5666 ////////////////////////////////////// C++ CODE //////////////////////////////////////
5667 // NOTE: patdo comes from DATAMIX and can be considered the same as patd for Jaguar I
5669 //////////////////////////////////////////////////////////////////////////////////////
5673 wdata[0..63] // co-processor rwrite data bus
5676 big_pix // Pixel organisation is big-endian
5681 data_sel[0..1] // source of write data
5682 data_ena // enable write data onto read/write bus
5683 dstart[0..5] // start of changed write data
5684 dend[0..5] // end of changed write data
5685 dbinh\[0..7] // byte oriented changed data inhibits
5688 phrase_mode // phrase write mode
5693 /*INT32/ addql[0..1], ddatlo, ddathi zero32
5697 Phrase_mode\ := INV1 (phrase_mode\, phrase_mode);
5698 Zero := TIE0 (zero);
5699 Zero32 := JOIN (zero32, zero, zero, zero, zero, zero, zero, zero, zero, zero, zero, zero, zero, zero, zero, zero, zero, zero, zero, zero, zero, zero, zero, zero, zero, zero, zero, zero, zero, zero, zero, zero, zero);*/
5701 /* Generate a changed data mask */
5703 /*Edis := OR6 (edis\, dend[0..5]);
5704 Ecoarse := DECL38E (e_coarse\[0..7], dend[3..5], edis\);
5705 E_coarse[0] := INV1 (e_coarse[0], e_coarse\[0]);
5706 Efine := DECL38E (unused[0], e_fine\[1..7], dend[0..2], e_coarse[0]);*/
5707 ////////////////////////////////////// C++ CODE //////////////////////////////////////
5708 uint8 decl38e[2][8] = { { 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF },
5709 { 0xFE, 0xFD, 0xFB, 0xF7, 0xEF, 0xDF, 0xBF, 0x7F } };
5710 uint8 dech38[8] = { 0x01, 0x02, 0x04, 0x08, 0x10, 0x20, 0x40, 0x80 };
5711 uint8 dech38el[2][8] = { { 0x01, 0x02, 0x04, 0x08, 0x10, 0x20, 0x40, 0x80 },
5712 { 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00 } };
5714 int en = (dend & 0x3F ? 1 : 0);
5715 uint8 e_coarse = decl38e[en][(dend & 0x38) >> 3]; // Actually, this is e_coarse inverted...
5716 uint8 e_fine = decl38e[(e_coarse & 0x01) ^ 0x01][dend & 0x07];
5718 //////////////////////////////////////////////////////////////////////////////////////
5720 /*Scoarse := DECH38 (s_coarse[0..7], dstart[3..5]);
5721 Sfen\ := INV1 (sfen\, s_coarse[0]);
5722 Sfine := DECH38EL (s_fine[0..7], dstart[0..2], sfen\);*/
5723 ////////////////////////////////////// C++ CODE //////////////////////////////////////
5724 uint8 s_coarse = dech38[(dstart & 0x38) >> 3];
5725 uint8 s_fine = dech38el[(s_coarse & 0x01) ^ 0x01][dstart & 0x07];
5726 //////////////////////////////////////////////////////////////////////////////////////
5728 /*Maskt[0] := BUF1 (maskt[0], s_fine[0]);
5729 Maskt[1-7] := OAN1P (maskt[1-7], maskt[0-6], s_fine[1-7], e_fine\[1-7]);*/
5730 ////////////////////////////////////// C++ CODE //////////////////////////////////////
5731 uint16 maskt = s_fine & 0x0001;
5732 maskt |= (((maskt & 0x0001) || (s_fine & 0x02)) && (e_fine & 0x02) ? 0x0002 : 0x0000);
5733 maskt |= (((maskt & 0x0002) || (s_fine & 0x04)) && (e_fine & 0x04) ? 0x0004 : 0x0000);
5734 maskt |= (((maskt & 0x0004) || (s_fine & 0x08)) && (e_fine & 0x08) ? 0x0008 : 0x0000);
5735 maskt |= (((maskt & 0x0008) || (s_fine & 0x10)) && (e_fine & 0x10) ? 0x0010 : 0x0000);
5736 maskt |= (((maskt & 0x0010) || (s_fine & 0x20)) && (e_fine & 0x20) ? 0x0020 : 0x0000);
5737 maskt |= (((maskt & 0x0020) || (s_fine & 0x40)) && (e_fine & 0x40) ? 0x0040 : 0x0000);
5738 maskt |= (((maskt & 0x0040) || (s_fine & 0x80)) && (e_fine & 0x80) ? 0x0080 : 0x0000);
5739 //////////////////////////////////////////////////////////////////////////////////////
5741 /* Produce a look-ahead on the ripple carry:
5742 masktla = s_coarse[0] . /e_coarse[0] */
5743 /*Masktla := AN2 (masktla, s_coarse[0], e_coarse\[0]);
5744 Maskt[8] := OAN1P (maskt[8], masktla, s_coarse[1], e_coarse\[1]);
5745 Maskt[9-14] := OAN1P (maskt[9-14], maskt[8-13], s_coarse[2-7], e_coarse\[2-7]);*/
5746 ////////////////////////////////////// C++ CODE //////////////////////////////////////
5747 maskt |= (((s_coarse & e_coarse & 0x01) || (s_coarse & 0x02)) && (e_coarse & 0x02) ? 0x0100 : 0x0000);
5748 maskt |= (((maskt & 0x0100) || (s_coarse & 0x04)) && (e_coarse & 0x04) ? 0x0200 : 0x0000);
5749 maskt |= (((maskt & 0x0200) || (s_coarse & 0x08)) && (e_coarse & 0x08) ? 0x0400 : 0x0000);
5750 maskt |= (((maskt & 0x0400) || (s_coarse & 0x10)) && (e_coarse & 0x10) ? 0x0800 : 0x0000);
5751 maskt |= (((maskt & 0x0800) || (s_coarse & 0x20)) && (e_coarse & 0x20) ? 0x1000 : 0x0000);
5752 maskt |= (((maskt & 0x1000) || (s_coarse & 0x40)) && (e_coarse & 0x40) ? 0x2000 : 0x0000);
5753 maskt |= (((maskt & 0x2000) || (s_coarse & 0x80)) && (e_coarse & 0x80) ? 0x4000 : 0x0000);
5754 //////////////////////////////////////////////////////////////////////////////////////
5756 /* The bit terms are mirrored for big-endian pixels outside phrase
5757 mode. The byte terms are mirrored for big-endian pixels in phrase
5760 /*Mirror_bit := AN2M (mir_bit, phrase_mode\, big_pix);
5761 Mirror_byte := AN2H (mir_byte, phrase_mode, big_pix);
5763 Masktb[14] := BUF1 (masktb[14], maskt[14]);
5764 Masku[0] := MX4 (masku[0], maskt[0], maskt[7], maskt[14], zero, mir_bit, mir_byte);
5765 Masku[1] := MX4 (masku[1], maskt[1], maskt[6], maskt[14], zero, mir_bit, mir_byte);
5766 Masku[2] := MX4 (masku[2], maskt[2], maskt[5], maskt[14], zero, mir_bit, mir_byte);
5767 Masku[3] := MX4 (masku[3], maskt[3], maskt[4], masktb[14], zero, mir_bit, mir_byte);
5768 Masku[4] := MX4 (masku[4], maskt[4], maskt[3], masktb[14], zero, mir_bit, mir_byte);
5769 Masku[5] := MX4 (masku[5], maskt[5], maskt[2], masktb[14], zero, mir_bit, mir_byte);
5770 Masku[6] := MX4 (masku[6], maskt[6], maskt[1], masktb[14], zero, mir_bit, mir_byte);
5771 Masku[7] := MX4 (masku[7], maskt[7], maskt[0], masktb[14], zero, mir_bit, mir_byte);
5772 Masku[8] := MX2 (masku[8], maskt[8], maskt[13], mir_byte);
5773 Masku[9] := MX2 (masku[9], maskt[9], maskt[12], mir_byte);
5774 Masku[10] := MX2 (masku[10], maskt[10], maskt[11], mir_byte);
5775 Masku[11] := MX2 (masku[11], maskt[11], maskt[10], mir_byte);
5776 Masku[12] := MX2 (masku[12], maskt[12], maskt[9], mir_byte);
5777 Masku[13] := MX2 (masku[13], maskt[13], maskt[8], mir_byte);
5778 Masku[14] := MX2 (masku[14], maskt[14], maskt[0], mir_byte);*/
5779 ////////////////////////////////////// C++ CODE //////////////////////////////////////
5780 bool mir_bit = true/*big_pix*/ && !phrase_mode;
5781 bool mir_byte = true/*big_pix*/ && phrase_mode;
5782 uint16 masku = maskt;
5787 masku |= (maskt >> 7) & 0x0001;
5788 masku |= (maskt >> 5) & 0x0002;
5789 masku |= (maskt >> 3) & 0x0004;
5790 masku |= (maskt >> 1) & 0x0008;
5791 masku |= (maskt << 1) & 0x0010;
5792 masku |= (maskt << 3) & 0x0020;
5793 masku |= (maskt << 5) & 0x0040;
5794 masku |= (maskt << 7) & 0x0080;
5800 masku |= (maskt >> 14) & 0x0001;
5801 masku |= (maskt >> 13) & 0x0002;
5802 masku |= (maskt >> 12) & 0x0004;
5803 masku |= (maskt >> 11) & 0x0008;
5804 masku |= (maskt >> 10) & 0x0010;
5805 masku |= (maskt >> 9) & 0x0020;
5806 masku |= (maskt >> 8) & 0x0040;
5807 masku |= (maskt >> 7) & 0x0080;
5809 masku |= (maskt >> 5) & 0x0100;
5810 masku |= (maskt >> 3) & 0x0200;
5811 masku |= (maskt >> 1) & 0x0400;
5812 masku |= (maskt << 1) & 0x0800;
5813 masku |= (maskt << 3) & 0x1000;
5814 masku |= (maskt << 5) & 0x2000;
5815 masku |= (maskt << 7) & 0x4000;
5817 //////////////////////////////////////////////////////////////////////////////////////
5819 /* The maskt terms define the area for changed data, but the byte
5820 inhibit terms can override these */
5822 /*Mask[0-7] := AN2 (mask[0-7], masku[0-7], dbinh\[0]);
5823 Mask[8-14] := AN2H (mask[8-14], masku[8-14], dbinh\[1-7]);*/
5824 ////////////////////////////////////// C++ CODE //////////////////////////////////////
5825 uint16 mask = masku & (!(dbinh & 0x01) ? 0xFFFF : 0xFF00);
5826 mask &= ~(((uint16)dbinh & 0x00FE) << 7);
5827 //////////////////////////////////////////////////////////////////////////////////////
5829 /*Addql[0] := JOIN (addql[0], addq[0..1]);
5830 Addql[1] := JOIN (addql[1], addq[2..3]);
5832 Dsel0b[0-1] := BUF8 (dsel0b[0-1], data_sel[0]);
5833 Dsel1b[0-1] := BUF8 (dsel1b[0-1], data_sel[1]);
5834 Ddatlo := MX4 (ddatlo, patd[0], lfu[0], addql[0], zero32, dsel0b[0], dsel1b[0]);
5835 Ddathi := MX4 (ddathi, patd[1], lfu[1], addql[1], zero32, dsel0b[1], dsel1b[1]);*/
5836 ////////////////////////////////////// C++ CODE //////////////////////////////////////
5840 dmux[2] = ((uint64)addq[3] << 48) | ((uint64)addq[2] << 32) | ((uint64)addq[1] << 16) | (uint64)addq[0];
5842 uint64 ddat = dmux[data_sel];
5843 //////////////////////////////////////////////////////////////////////////////////////
5845 /*Zed_sel := AN2 (zed_sel, data_sel[0..1]);
5846 Zed_selb[0-1] := BUF8 (zed_selb[0-1], zed_sel);
5848 Dat[0-7] := MX4 (dat[0-7], dstdlo{0-7}, ddatlo{0-7}, dstzlo{0-7}, srczlo{0-7}, mask[0-7], zed_selb[0]);
5849 Dat[8-15] := MX4 (dat[8-15], dstdlo{8-15}, ddatlo{8-15}, dstzlo{8-15}, srczlo{8-15}, mask[8], zed_selb[0]);
5850 Dat[16-23] := MX4 (dat[16-23], dstdlo{16-23}, ddatlo{16-23}, dstzlo{16-23}, srczlo{16-23}, mask[9], zed_selb[0]);
5851 Dat[24-31] := MX4 (dat[24-31], dstdlo{24-31}, ddatlo{24-31}, dstzlo{24-31}, srczlo{24-31}, mask[10], zed_selb[0]);
5852 Dat[32-39] := MX4 (dat[32-39], dstdhi{0-7}, ddathi{0-7}, dstzhi{0-7}, srczhi{0-7}, mask[11], zed_selb[1]);
5853 Dat[40-47] := MX4 (dat[40-47], dstdhi{8-15}, ddathi{8-15}, dstzhi{8-15}, srczhi{8-15}, mask[12], zed_selb[1]);
5854 Dat[48-55] := MX4 (dat[48-55], dstdhi{16-23}, ddathi{16-23}, dstzhi{16-23}, srczhi{16-23}, mask[13], zed_selb[1]);
5855 Dat[56-63] := MX4 (dat[56-63], dstdhi{24-31}, ddathi{24-31}, dstzhi{24-31}, srczhi{24-31}, mask[14], zed_selb[1]);*/
5856 ////////////////////////////////////// C++ CODE //////////////////////////////////////
5857 wdata = ((ddat & mask) | (dstd & ~mask)) & 0x00000000000000FFLL;
5858 wdata |= (mask & 0x0100 ? ddat : dstd) & 0x000000000000FF00LL;
5859 wdata |= (mask & 0x0200 ? ddat : dstd) & 0x0000000000FF0000LL;
5860 wdata |= (mask & 0x0400 ? ddat : dstd) & 0x00000000FF000000LL;
5861 wdata |= (mask & 0x0800 ? ddat : dstd) & 0x000000FF00000000LL;
5862 wdata |= (mask & 0x1000 ? ddat : dstd) & 0x0000FF0000000000LL;
5863 wdata |= (mask & 0x2000 ? ddat : dstd) & 0x00FF000000000000LL;
5864 wdata |= (mask & 0x4000 ? ddat : dstd) & 0xFF00000000000000LL;
5867 printf("\n[ddat=%08X%08X dstd=%08X%08X wdata=%08X%08X mask=%04X]\n",
5868 (uint32)(ddat >> 32), (uint32)(ddat & 0xFFFFFFFF),
5869 (uint32)(dstd >> 32), (uint32)(dstd & 0xFFFFFFFF),
5870 (uint32)(wdata >> 32), (uint32)(wdata & 0xFFFFFFFF), mask);
5873 //This is a crappy way of handling this, but it should work for now...
5875 zwdata = ((srcz & mask) | (dstz & ~mask)) & 0x00000000000000FFLL;
5876 zwdata |= (mask & 0x0100 ? srcz : dstz) & 0x000000000000FF00LL;
5877 zwdata |= (mask & 0x0200 ? srcz : dstz) & 0x0000000000FF0000LL;
5878 zwdata |= (mask & 0x0400 ? srcz : dstz) & 0x00000000FF000000LL;
5879 zwdata |= (mask & 0x0800 ? srcz : dstz) & 0x000000FF00000000LL;
5880 zwdata |= (mask & 0x1000 ? srcz : dstz) & 0x0000FF0000000000LL;
5881 zwdata |= (mask & 0x2000 ? srcz : dstz) & 0x00FF000000000000LL;
5882 zwdata |= (mask & 0x4000 ? srcz : dstz) & 0xFF00000000000000LL;
5885 printf("\n[srcz=%08X%08X dstz=%08X%08X zwdata=%08X%08X mask=%04X]\n",
5886 (uint32)(srcz >> 32), (uint32)(srcz & 0xFFFFFFFF),
5887 (uint32)(dstz >> 32), (uint32)(dstz & 0xFFFFFFFF),
5888 (uint32)(zwdata >> 32), (uint32)(zwdata & 0xFFFFFFFF), mask);
5892 //////////////////////////////////////////////////////////////////////////////////////
5894 /*Data_enab[0-1] := BUF8 (data_enab[0-1], data_ena);
5895 Datadrv[0-31] := TS (wdata[0-31], dat[0-31], data_enab[0]);
5896 Datadrv[32-63] := TS (wdata[32-63], dat[32-63], data_enab[1]);
5898 Unused[0] := DUMMY (unused[0]);
5903 /** COMP_CTRL - Comparator output control logic *****************
5905 This block is responsible for taking the comparator outputs and
5906 using them as appropriate to inhibit writes. Two methods are
5907 supported for inhibiting write data:
5909 - suppression of the inner loop controlled write operation
5910 - a set of eight byte inhibit lines to write back dest data
5912 The first technique is used in pixel oriented modes, the second in
5913 phrase mode, but the phrase mode form is only applicable to eight
5914 and sixteen bit pixel modes.
5916 Writes can be suppressed by data being equal, by the Z comparator
5917 conditions being met, or by the bit to pixel expansion scheme.
5919 Pipe-lining issues: the data derived comparator outputs are stable
5920 until the next data read, well after the affected write from this
5921 operation. However, the inner counter bits can count immediately
5922 before the ack for the last write. Therefore, it is necessary to
5923 delay bcompbit select terms by one inner loop pipe-line stage,
5924 when generating the select for the data control - the output is
5925 delayed one further tick to give it write data timing (2/34).
5927 There is also a problem with computed data - the new values are
5928 calculated before the write associated with the old value has been
5929 performed. The is taken care of within the zed comparator by
5930 pipe-lining the comparator inputs where appropriate.
5933 //#define LOG_COMP_CTRL
5935 dbinh\[0..7] // destination byte inhibit lines
5936 nowrite // suppress inner loop write operation
5938 bcompen // bit selector inhibit enable
5939 big_pix // pixels are big-endian
5940 bkgwren // enable dest data write in pix inhibit
5941 clk // co-processor clock
5942 dcomp[0..7] // output of data byte comparators
5943 dcompen // data comparator inhibit enable
5944 icount[0..2] // low bits of inner count
5945 pixsize[0..2] // destination pixel size
5946 phrase_mode // phrase write mode
5947 srcd[0..7] // bits to use for bit to byte expansion
5948 step_inner // inner loop advance
5949 zcomp[0..3] // output of word zed comparators
5951 void COMP_CTRL(uint8 &dbinh, bool &nowrite,
5952 bool bcompen, bool big_pix, bool bkgwren, uint8 dcomp, bool dcompen, uint8 icount,
5953 uint8 pixsize, bool phrase_mode, uint8 srcd, uint8 zcomp)
5957 /*Bkgwren\ := INV1 (bkgwren\, bkgwren);
5958 Phrase_mode\ := INV1 (phrase_mode\, phrase_mode);
5959 Pixsize\[0-2] := INV2 (pixsize\[0-2], pixsize[0-2]);*/
5961 /* The bit comparator bits are derived from the source data, which
5962 will have been suitably aligned for phrase mode. The contents of
5963 the inner counter are used to select which bit to use.
5965 When not in phrase mode the inner count value is used to select
5966 one bit. It is assumed that the count has already occurred, so,
5967 7 selects bit 0, etc. In big-endian pixel mode, this turns round,
5968 so that a count of 7 selects bit 7.
5970 In phrase mode, the eight bits are used directly, and this mode is
5971 only applicable to 8-bit pixel mode (2/34) */
5973 /*Bcompselt[0-2] := EO (bcompselt[0-2], icount[0-2], big_pix);
5974 Bcompbit := MX8 (bcompbit, srcd[7], srcd[6], srcd[5],
5975 srcd[4], srcd[3], srcd[2], srcd[1], srcd[0], bcompselt[0..2]);
5976 Bcompbit\ := INV1 (bcompbit\, bcompbit);*/
5977 ////////////////////////////////////// C++ CODE //////////////////////////////////////
5978 #ifdef LOG_COMP_CTRL
5981 printf("\n [bcompen=%s dcompen=%s phrase_mode=%s bkgwren=%s dcomp=%02X zcomp=%02X]", (bcompen ? "T" : "F"), (dcompen ? "T" : "F"), (phrase_mode ? "T" : "F"), (bkgwren ? "T" : "F"), dcomp, zcomp);
5986 uint8 bcompselt = (big_pix ? ~icount : icount) & 0x07;
5987 uint8 bitmask[8] = { 0x80, 0x40, 0x20, 0x10, 0x08, 0x04, 0x02, 0x01 };
5988 bool bcompbit = srcd & bitmask[bcompselt];
5989 //////////////////////////////////////////////////////////////////////////////////////
5991 /* pipe-line the count */
5992 /*Bcompsel[0-2] := FDSYNC (bcompsel[0-2], bcompselt[0-2], step_inner, clk);
5993 Bcompbt := MX8 (bcompbitpt, srcd[7], srcd[6], srcd[5],
5994 srcd[4], srcd[3], srcd[2], srcd[1], srcd[0], bcompsel[0..2]);
5995 Bcompbitp := FD1Q (bcompbitp, bcompbitpt, clk);
5996 Bcompbitp\ := INV1 (bcompbitp\, bcompbitp);*/
5998 /* For pixel mode, generate the write inhibit signal for all modes
5999 on bit inhibit, for 8 and 16 bit modes on comparator inhibit, and
6000 for 16 bit mode on Z inhibit
6002 Nowrite = bcompen . /bcompbit . /phrase_mode
6003 + dcompen . dcomp[0] . /phrase_mode . pixsize = 011
6004 + dcompen . dcomp[0..1] . /phrase_mode . pixsize = 100
6005 + zcomp[0] . /phrase_mode . pixsize = 100
6008 /*Nowt0 := NAN3 (nowt[0], bcompen, bcompbit\, phrase_mode\);
6009 Nowt1 := ND6 (nowt[1], dcompen, dcomp[0], phrase_mode\, pixsize\[2], pixsize[0..1]);
6010 Nowt2 := ND7 (nowt[2], dcompen, dcomp[0..1], phrase_mode\, pixsize[2], pixsize\[0..1]);
6011 Nowt3 := NAN5 (nowt[3], zcomp[0], phrase_mode\, pixsize[2], pixsize\[0..1]);
6012 Nowt4 := NAN4 (nowt[4], nowt[0..3]);
6013 Nowrite := AN2 (nowrite, nowt[4], bkgwren\);*/
6014 ////////////////////////////////////// C++ CODE //////////////////////////////////////
6015 nowrite = ((bcompen && !bcompbit && !phrase_mode)
6016 || (dcompen && (dcomp & 0x01) && !phrase_mode && (pixsize == 3))
6017 || (dcompen && ((dcomp & 0x03) == 0x03) && !phrase_mode && (pixsize == 4))
6018 || ((zcomp & 0x01) && !phrase_mode && (pixsize == 4)))
6020 //////////////////////////////////////////////////////////////////////////////////////
6022 /*Winht := NAN3 (winht, bcompen, bcompbitp\, phrase_mode\);
6023 Winhibit := NAN4 (winhibit, winht, nowt[1..3]);*/
6024 ////////////////////////////////////// C++ CODE //////////////////////////////////////
6025 //This is the same as above, but with bcompbit delayed one tick and called 'winhibit'
6026 //Small difference: Besides the pipeline effect, it's also not using !bkgwren...
6027 // bool winhibit = (bcompen && !
6028 bool winhibit = (bcompen && !bcompbit && !phrase_mode)
6029 || (dcompen && (dcomp & 0x01) && !phrase_mode && (pixsize == 3))
6030 || (dcompen && ((dcomp & 0x03) == 0x03) && !phrase_mode && (pixsize == 4))
6031 || ((zcomp & 0x01) && !phrase_mode && (pixsize == 4));
6032 #ifdef LOG_COMP_CTRL
6035 printf("[nw=%s wi=%s]", (nowrite ? "T" : "F"), (winhibit ? "T" : "F"));
6039 //////////////////////////////////////////////////////////////////////////////////////
6041 /* For phrase mode, generate the byte inhibit signals for eight bit
6042 mode 011, or sixteen bit mode 100
6043 dbinh\[0] = pixsize[2] . zcomp[0]
6044 + pixsize[2] . dcomp[0] . dcomp[1] . dcompen
6045 + /pixsize[2] . dcomp[0] . dcompen
6046 + /srcd[0] . bcompen
6048 Inhibits 0-3 are also used when not in phrase mode to write back
6052 /*Srcd\[0-7] := INV1 (srcd\[0-7], srcd[0-7]);
6054 Di0t0 := NAN2H (di0t[0], pixsize[2], zcomp[0]);
6055 Di0t1 := NAN4H (di0t[1], pixsize[2], dcomp[0..1], dcompen);
6056 Di0t2 := NAN2 (di0t[2], srcd\[0], bcompen);
6057 Di0t3 := NAN3 (di0t[3], pixsize\[2], dcomp[0], dcompen);
6058 Di0t4 := NAN4 (di0t[4], di0t[0..3]);
6059 Dbinh[0] := ANR1P (dbinh\[0], di0t[4], phrase_mode, winhibit);*/
6060 ////////////////////////////////////// C++ CODE //////////////////////////////////////
6062 bool di0t0_1 = ((pixsize & 0x04) && (zcomp & 0x01))
6063 || ((pixsize & 0x04) && (dcomp & 0x01) && (dcomp & 0x02) && dcompen);
6064 bool di0t4 = di0t0_1
6065 || (!(srcd & 0x01) && bcompen)
6066 || (!(pixsize & 0x04) && (dcomp & 0x01) && dcompen);
6067 dbinh |= (!((di0t4 && phrase_mode) || winhibit) ? 0x01 : 0x00);
6068 #ifdef LOG_COMP_CTRL
6071 printf("[di0t0_1=%s di0t4=%s]", (di0t0_1 ? "T" : "F"), (di0t4 ? "T" : "F"));
6075 //////////////////////////////////////////////////////////////////////////////////////
6077 /*Di1t0 := NAN3 (di1t[0], pixsize\[2], dcomp[1], dcompen);
6078 Di1t1 := NAN2 (di1t[1], srcd\[1], bcompen);
6079 Di1t2 := NAN4 (di1t[2], di0t[0..1], di1t[0..1]);
6080 Dbinh[1] := ANR1 (dbinh\[1], di1t[2], phrase_mode, winhibit);*/
6081 ////////////////////////////////////// C++ CODE //////////////////////////////////////
6082 bool di1t2 = di0t0_1
6083 || (!(srcd & 0x02) && bcompen)
6084 || (!(pixsize & 0x04) && (dcomp & 0x02) && dcompen);
6085 dbinh |= (!((di1t2 && phrase_mode) || winhibit) ? 0x02 : 0x00);
6086 #ifdef LOG_COMP_CTRL
6089 printf("[di1t2=%s]", (di1t2 ? "T" : "F"));
6093 //////////////////////////////////////////////////////////////////////////////////////
6095 /*Di2t0 := NAN2H (di2t[0], pixsize[2], zcomp[1]);
6096 Di2t1 := NAN4H (di2t[1], pixsize[2], dcomp[2..3], dcompen);
6097 Di2t2 := NAN2 (di2t[2], srcd\[2], bcompen);
6098 Di2t3 := NAN3 (di2t[3], pixsize\[2], dcomp[2], dcompen);
6099 Di2t4 := NAN4 (di2t[4], di2t[0..3]);
6100 Dbinh[2] := ANR1 (dbinh\[2], di2t[4], phrase_mode, winhibit);*/
6101 ////////////////////////////////////// C++ CODE //////////////////////////////////////
6102 //[bcompen=F dcompen=T phrase_mode=T bkgwren=F][nw=F wi=F]
6103 //[di0t0_1=F di0t4=F][di1t2=F][di2t0_1=T di2t4=T][di3t2=T][di4t0_1=F di2t4=F][di5t2=F][di6t0_1=F di6t4=F][di7t2=F]
6104 //[dcomp=$00 dbinh=$0C][7804780400007804] (icount=0005, inc=4)
6105 bool di2t0_1 = ((pixsize & 0x04) && (zcomp & 0x02))
6106 || ((pixsize & 0x04) && (dcomp & 0x04) && (dcomp & 0x08) && dcompen);
6107 bool di2t4 = di2t0_1
6108 || (!(srcd & 0x04) && bcompen)
6109 || (!(pixsize & 0x04) && (dcomp & 0x04) && dcompen);
6110 dbinh |= (!((di2t4 && phrase_mode) || winhibit) ? 0x04 : 0x00);
6111 #ifdef LOG_COMP_CTRL
6114 printf("[di2t0_1=%s di2t4=%s]", (di2t0_1 ? "T" : "F"), (di2t4 ? "T" : "F"));
6118 //////////////////////////////////////////////////////////////////////////////////////
6120 /*Di3t0 := NAN3 (di3t[0], pixsize\[2], dcomp[3], dcompen);
6121 Di3t1 := NAN2 (di3t[1], srcd\[3], bcompen);
6122 Di3t2 := NAN4 (di3t[2], di2t[0..1], di3t[0..1]);
6123 Dbinh[3] := ANR1 (dbinh\[3], di3t[2], phrase_mode, winhibit);*/
6124 ////////////////////////////////////// C++ CODE //////////////////////////////////////
6125 bool di3t2 = di2t0_1
6126 || (!(srcd & 0x08) && bcompen)
6127 || (!(pixsize & 0x04) && (dcomp & 0x08) && dcompen);
6128 dbinh |= (!((di3t2 && phrase_mode) || winhibit) ? 0x08 : 0x00);
6129 #ifdef LOG_COMP_CTRL
6132 printf("[di3t2=%s]", (di3t2 ? "T" : "F"));
6136 //////////////////////////////////////////////////////////////////////////////////////
6138 /*Di4t0 := NAN2H (di4t[0], pixsize[2], zcomp[2]);
6139 Di4t1 := NAN4H (di4t[1], pixsize[2], dcomp[4..5], dcompen);
6140 Di4t2 := NAN2 (di4t[2], srcd\[4], bcompen);
6141 Di4t3 := NAN3 (di4t[3], pixsize\[2], dcomp[4], dcompen);
6142 Di4t4 := NAN4 (di4t[4], di4t[0..3]);
6143 Dbinh[4] := NAN2 (dbinh\[4], di4t[4], phrase_mode);*/
6144 ////////////////////////////////////// C++ CODE //////////////////////////////////////
6145 bool di4t0_1 = ((pixsize & 0x04) && (zcomp & 0x04))
6146 || ((pixsize & 0x04) && (dcomp & 0x10) && (dcomp & 0x20) && dcompen);
6147 bool di4t4 = di4t0_1
6148 || (!(srcd & 0x10) && bcompen)
6149 || (!(pixsize & 0x04) && (dcomp & 0x10) && dcompen);
6150 dbinh |= (!(di4t4 && phrase_mode) ? 0x10 : 0x00);
6151 #ifdef LOG_COMP_CTRL
6154 printf("[di4t0_1=%s di2t4=%s]", (di4t0_1 ? "T" : "F"), (di4t4 ? "T" : "F"));
6158 //////////////////////////////////////////////////////////////////////////////////////
6160 /*Di5t0 := NAN3 (di5t[0], pixsize\[2], dcomp[5], dcompen);
6161 Di5t1 := NAN2 (di5t[1], srcd\[5], bcompen);
6162 Di5t2 := NAN4 (di5t[2], di4t[0..1], di5t[0..1]);
6163 Dbinh[5] := NAN2 (dbinh\[5], di5t[2], phrase_mode);*/
6164 ////////////////////////////////////// C++ CODE //////////////////////////////////////
6165 bool di5t2 = di4t0_1
6166 || (!(srcd & 0x20) && bcompen)
6167 || (!(pixsize & 0x04) && (dcomp & 0x20) && dcompen);
6168 dbinh |= (!(di5t2 && phrase_mode) ? 0x20 : 0x00);
6169 #ifdef LOG_COMP_CTRL
6172 printf("[di5t2=%s]", (di5t2 ? "T" : "F"));
6176 //////////////////////////////////////////////////////////////////////////////////////
6178 /*Di6t0 := NAN2H (di6t[0], pixsize[2], zcomp[3]);
6179 Di6t1 := NAN4H (di6t[1], pixsize[2], dcomp[6..7], dcompen);
6180 Di6t2 := NAN2 (di6t[2], srcd\[6], bcompen);
6181 Di6t3 := NAN3 (di6t[3], pixsize\[2], dcomp[6], dcompen);
6182 Di6t4 := NAN4 (di6t[4], di6t[0..3]);
6183 Dbinh[6] := NAN2 (dbinh\[6], di6t[4], phrase_mode);*/
6184 ////////////////////////////////////// C++ CODE //////////////////////////////////////
6185 bool di6t0_1 = ((pixsize & 0x04) && (zcomp & 0x08))
6186 || ((pixsize & 0x04) && (dcomp & 0x40) && (dcomp & 0x80) && dcompen);
6187 bool di6t4 = di6t0_1
6188 || (!(srcd & 0x40) && bcompen)
6189 || (!(pixsize & 0x04) && (dcomp & 0x40) && dcompen);
6190 dbinh |= (!(di6t4 && phrase_mode) ? 0x40 : 0x00);
6191 #ifdef LOG_COMP_CTRL
6194 printf("[di6t0_1=%s di6t4=%s]", (di6t0_1 ? "T" : "F"), (di6t4 ? "T" : "F"));
6198 //////////////////////////////////////////////////////////////////////////////////////
6200 /*Di7t0 := NAN3 (di7t[0], pixsize\[2], dcomp[7], dcompen);
6201 Di7t1 := NAN2 (di7t[1], srcd\[7], bcompen);
6202 Di7t2 := NAN4 (di7t[2], di6t[0..1], di7t[0..1]);
6203 Dbinh[7] := NAN2 (dbinh\[7], di7t[2], phrase_mode);*/
6204 ////////////////////////////////////// C++ CODE //////////////////////////////////////
6205 bool di7t2 = di6t0_1
6206 || (!(srcd & 0x80) && bcompen)
6207 || (!(pixsize & 0x04) && (dcomp & 0x80) && dcompen);
6208 dbinh |= (!(di7t2 && phrase_mode) ? 0x80 : 0x00);
6209 #ifdef LOG_COMP_CTRL
6212 printf("[di7t2=%s]", (di7t2 ? "T" : "F"));
6216 //////////////////////////////////////////////////////////////////////////////////////
6221 #ifdef LOG_COMP_CTRL
6224 printf("[dcomp=$%02X dbinh=$%02X]\n ", dcomp, dbinh);
6231 ////////////////////////////////////// C++ CODE //////////////////////////////////////
6232 //////////////////////////////////////////////////////////////////////////////////////