1 /*
2 * Copyright © 2007-2011 Intel Corporation
3 *
4 * Permission is hereby granted, free of charge, to any person obtaining a
5 * copy of this software and associated documentation files (the "Software"),
6 * to deal in the Software without restriction, including without limitation
7 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
8 * and/or sell copies of the Software, and to permit persons to whom the
9 * Software is furnished to do so, subject to the following conditions:
10 *
11 * The above copyright notice and this permission notice (including the next
12 * paragraph) shall be included in all copies or substantial portions of the
13 * Software.
14 *
15 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
18 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
20 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
21 * SOFTWARE.
22 *
23 * Authors:
24 * Eric Anholt <eric@anholt.net>
25 * Chris Wilson <chris@chris-wilson.co.uk>
26 *
27 */
28
29 #ifdef HAVE_CONFIG_H
30 #include "config.h"
31 #endif
32
33 #include <sys/mman.h>
34 #include <assert.h>
35
36 #include "sna.h"
37 #include "sna_reg.h"
38
39 #include "gen3_render.h"
40
41 #include "kgem_debug.h"
42
43 enum type {
44 T_FLOAT32,
45 T_FLOAT16,
46 };
47
48 static struct state {
49 struct vertex_buffer {
50 int handle;
51 void *base;
52 const char *ptr;
53 int pitch;
54
55 struct kgem_bo *current;
56 } vb;
57 struct vertex_elements {
58 int offset;
59 bool valid;
60 enum type type;
61 int size;
62 uint8_t swizzle[4];
63 } ve[33];
64 int num_ve;
65 } state;
66
int_as_float(int i)67 static float int_as_float(int i)
68 {
69 union {
70 float f;
71 int i;
72 } x;
73 x.i = i;
74 return x.f;
75 }
76
gen3_update_vertex_buffer_addr(struct kgem * kgem,uint32_t offset)77 static void gen3_update_vertex_buffer_addr(struct kgem *kgem,
78 uint32_t offset)
79 {
80 uint32_t handle;
81 struct kgem_bo *bo = NULL;
82 void *base, *ptr;
83 int i;
84
85 offset *= sizeof(uint32_t);
86
87 for (i = 0; i < kgem->nreloc; i++)
88 if (kgem->reloc[i].offset == offset)
89 break;
90 assert(i < kgem->nreloc);
91 handle = kgem->reloc[i].target_handle;
92
93 if (handle == 0) {
94 base = kgem->batch;
95 } else {
96 list_for_each_entry(bo, &kgem->next_request->buffers, request)
97 if (bo->handle == handle)
98 break;
99 assert(&bo->request != &kgem->next_request->buffers);
100 base = kgem_bo_map__debug(kgem, bo);
101 }
102 ptr = (char *)base + kgem->reloc[i].delta;
103
104 state.vb.current = bo;
105 state.vb.base = base;
106 state.vb.ptr = ptr;
107 }
108
gen3_update_vertex_buffer_pitch(struct kgem * kgem,uint32_t offset)109 static void gen3_update_vertex_buffer_pitch(struct kgem *kgem,
110 uint32_t offset)
111 {
112 state.vb.pitch = kgem->batch[offset] >> 16 & 0x3f;
113 state.vb.pitch *= sizeof(uint32_t);
114 }
115
gen3_update_vertex_elements(struct kgem * kgem,uint32_t data)116 static void gen3_update_vertex_elements(struct kgem *kgem, uint32_t data)
117 {
118 state.ve[1].valid = 1;
119
120 switch ((data >> 6) & 7) {
121 case 1:
122 state.ve[1].type = T_FLOAT32;
123 state.ve[1].size = 3;
124 state.ve[1].swizzle[0] = 1;
125 state.ve[1].swizzle[1] = 1;
126 state.ve[1].swizzle[2] = 1;
127 state.ve[1].swizzle[3] = 3;
128 break;
129 case 2:
130 state.ve[1].type = T_FLOAT32;
131 state.ve[1].size = 4;
132 state.ve[1].swizzle[0] = 1;
133 state.ve[1].swizzle[1] = 1;
134 state.ve[1].swizzle[2] = 1;
135 state.ve[1].swizzle[3] = 1;
136 break;
137 case 3:
138 state.ve[1].type = T_FLOAT32;
139 state.ve[1].size = 2;
140 state.ve[1].swizzle[0] = 1;
141 state.ve[1].swizzle[1] = 1;
142 state.ve[1].swizzle[2] = 2;
143 state.ve[1].swizzle[3] = 3;
144 break;
145 case 4:
146 state.ve[1].type = T_FLOAT32;
147 state.ve[1].size = 3;
148 state.ve[1].swizzle[0] = 1;
149 state.ve[1].swizzle[1] = 1;
150 state.ve[1].swizzle[2] = 3;
151 state.ve[1].swizzle[3] = 1;
152 break;
153 }
154
155 state.ve[2].valid = 0;
156 state.ve[3].valid = 0;
157 }
158
gen3_update_vertex_texcoords(struct kgem * kgem,uint32_t data)159 static void gen3_update_vertex_texcoords(struct kgem *kgem, uint32_t data)
160 {
161 int id;
162 for (id = 0; id < 8; id++) {
163 uint32_t fmt = (data >> (id*4)) & 0xf;
164 int width;
165
166 state.ve[id+4].valid = fmt != 0xf;
167
168 width = 0;
169 switch (fmt) {
170 case 0:
171 state.ve[id+4].type = T_FLOAT32;
172 width = state.ve[id+4].size = 2;
173 break;
174 case 1:
175 state.ve[id+4].type = T_FLOAT32;
176 width = state.ve[id+4].size = 3;
177 break;
178 case 2:
179 state.ve[id+4].type = T_FLOAT32;
180 width = state.ve[id+4].size = 4;
181 break;
182 case 3:
183 state.ve[id+4].type = T_FLOAT32;
184 width = state.ve[id+4].size = 1;
185 break;
186 case 4:
187 state.ve[id+4].type = T_FLOAT16;
188 width = state.ve[id+4].size = 2;
189 break;
190 case 5:
191 state.ve[id+4].type = T_FLOAT16;
192 width = state.ve[id+4].size = 4;
193 break;
194 }
195
196 state.ve[id+4].swizzle[0] = width > 0 ? 1 : 2;
197 state.ve[id+4].swizzle[1] = width > 1 ? 1 : 2;
198 state.ve[id+4].swizzle[2] = width > 2 ? 1 : 2;
199 state.ve[id+4].swizzle[3] = width > 3 ? 1 : 2;
200 }
201 }
202
gen3_update_vertex_elements_offsets(struct kgem * kgem)203 static void gen3_update_vertex_elements_offsets(struct kgem *kgem)
204 {
205 int i, offset;
206
207 for (i = offset = 0; i < ARRAY_SIZE(state.ve); i++) {
208 if (!state.ve[i].valid)
209 continue;
210
211 state.ve[i].offset = offset;
212 offset += 4 * state.ve[i].size;
213 state.num_ve = i;
214 }
215 }
216
vertices_float32_out(const struct vertex_elements * ve,const float * f,int max)217 static void vertices_float32_out(const struct vertex_elements *ve, const float *f, int max)
218 {
219 int c;
220
221 ErrorF("(");
222 for (c = 0; c < max; c++) {
223 switch (ve->swizzle[c]) {
224 case 0: ErrorF("#"); break;
225 case 1: ErrorF("%f", f[c]); break;
226 case 2: ErrorF("0.0"); break;
227 case 3: ErrorF("1.0"); break;
228 case 4: ErrorF("0x1"); break;
229 case 5: break;
230 default: ErrorF("?");
231 }
232 if (c < max-1)
233 ErrorF(", ");
234 }
235 ErrorF(")");
236 }
237
ve_out(const struct vertex_elements * ve,const void * ptr)238 static void ve_out(const struct vertex_elements *ve, const void *ptr)
239 {
240 switch (ve->type) {
241 case T_FLOAT32:
242 vertices_float32_out(ve, ptr, ve->size);
243 break;
244 case T_FLOAT16:
245 //vertices_float16_out(ve, ptr, ve->size);
246 break;
247 }
248 }
249
indirect_vertex_out(struct kgem * kgem,uint32_t v)250 static void indirect_vertex_out(struct kgem *kgem, uint32_t v)
251 {
252 const struct vertex_buffer *vb = &state.vb;
253 int i = 1;
254
255 do {
256 const struct vertex_elements *ve = &state.ve[i];
257 const void *ptr = vb->ptr + v * vb->pitch + ve->offset;
258
259 if (!ve->valid)
260 continue;
261
262 ve_out(ve, ptr);
263
264 while (++i <= state.num_ve && !state.ve[i].valid)
265 ;
266
267 if (i <= state.num_ve)
268 ErrorF(", ");
269 } while (i <= state.num_ve);
270 }
271
inline_vertex_out(struct kgem * kgem,void * base)272 static int inline_vertex_out(struct kgem *kgem, void *base)
273 {
274 const struct vertex_buffer *vb = &state.vb;
275 int i = 1;
276
277 do {
278 const struct vertex_elements *ve = &state.ve[i];
279 const void *ptr = (char *)base + ve->offset;
280
281 if (!ve->valid)
282 continue;
283
284 ve_out(ve, ptr);
285
286 while (++i <= state.num_ve && !state.ve[i].valid)
287 ;
288
289 if (i <= state.num_ve)
290 ErrorF(", ");
291 } while (i <= state.num_ve);
292
293 return vb->pitch;
294 }
295
296 static int
gen3_decode_3d_1c(struct kgem * kgem,uint32_t offset)297 gen3_decode_3d_1c(struct kgem *kgem, uint32_t offset)
298 {
299 uint32_t *data = kgem->batch + offset;
300 uint32_t opcode;
301
302 opcode = (data[0] & 0x00f80000) >> 19;
303
304 switch (opcode) {
305 case 0x11:
306 kgem_debug_print(data, offset, 0, "3DSTATE_DEPTH_SUBRECTANGLE_DISABLE\n");
307 return 1;
308 case 0x10:
309 kgem_debug_print(data, offset, 0, "3DSTATE_SCISSOR_ENABLE %s\n",
310 data[0]&1?"enabled":"disabled");
311 return 1;
312 case 0x01:
313 kgem_debug_print(data, offset, 0, "3DSTATE_MAP_COORD_SET_I830\n");
314 return 1;
315 case 0x0a:
316 kgem_debug_print(data, offset, 0, "3DSTATE_MAP_CUBE_I830\n");
317 return 1;
318 case 0x05:
319 kgem_debug_print(data, offset, 0, "3DSTATE_MAP_TEX_STREAM_I830\n");
320 return 1;
321 }
322
323 kgem_debug_print(data, offset, 0, "3D UNKNOWN: 3d_1c opcode = 0x%x\n",
324 opcode);
325 assert(0);
326 return 1;
327 }
328
329 /** Sets the string dstname to describe the destination of the PS instruction */
330 static void
gen3_get_instruction_dst(uint32_t * data,int i,char * dstname,int do_mask)331 gen3_get_instruction_dst(uint32_t *data, int i, char *dstname, int do_mask)
332 {
333 uint32_t a0 = data[i];
334 int dst_nr = (a0 >> 14) & 0xf;
335 char dstmask[8];
336 const char *sat;
337
338 if (do_mask) {
339 if (((a0 >> 10) & 0xf) == 0xf) {
340 dstmask[0] = 0;
341 } else {
342 int dstmask_index = 0;
343
344 dstmask[dstmask_index++] = '.';
345 if (a0 & (1 << 10))
346 dstmask[dstmask_index++] = 'x';
347 if (a0 & (1 << 11))
348 dstmask[dstmask_index++] = 'y';
349 if (a0 & (1 << 12))
350 dstmask[dstmask_index++] = 'z';
351 if (a0 & (1 << 13))
352 dstmask[dstmask_index++] = 'w';
353 dstmask[dstmask_index++] = 0;
354 }
355
356 if (a0 & (1 << 22))
357 sat = ".sat";
358 else
359 sat = "";
360 } else {
361 dstmask[0] = 0;
362 sat = "";
363 }
364
365 switch ((a0 >> 19) & 0x7) {
366 case 0:
367 assert(dst_nr <= 15);
368 sprintf(dstname, "R%d%s%s", dst_nr, dstmask, sat);
369 break;
370 case 4:
371 assert(dst_nr == 0);
372 sprintf(dstname, "oC%s%s", dstmask, sat);
373 break;
374 case 5:
375 assert(dst_nr == 0);
376 sprintf(dstname, "oD%s%s", dstmask, sat);
377 break;
378 case 6:
379 assert(dst_nr <= 3);
380 sprintf(dstname, "U%d%s%s", dst_nr, dstmask, sat);
381 break;
382 default:
383 sprintf(dstname, "RESERVED");
384 break;
385 }
386 }
387
388 static const char *
gen3_get_channel_swizzle(uint32_t select)389 gen3_get_channel_swizzle(uint32_t select)
390 {
391 switch (select & 0x7) {
392 case 0:
393 return (select & 8) ? "-x" : "x";
394 case 1:
395 return (select & 8) ? "-y" : "y";
396 case 2:
397 return (select & 8) ? "-z" : "z";
398 case 3:
399 return (select & 8) ? "-w" : "w";
400 case 4:
401 return (select & 8) ? "-0" : "0";
402 case 5:
403 return (select & 8) ? "-1" : "1";
404 default:
405 return (select & 8) ? "-bad" : "bad";
406 }
407 }
408
409 static void
gen3_get_instruction_src_name(uint32_t src_type,uint32_t src_nr,char * name)410 gen3_get_instruction_src_name(uint32_t src_type, uint32_t src_nr, char *name)
411 {
412 switch (src_type) {
413 case 0:
414 sprintf(name, "R%d", src_nr);
415 assert(src_nr <= 15);
416 break;
417 case 1:
418 if (src_nr < 8)
419 sprintf(name, "T%d", src_nr);
420 else if (src_nr == 8)
421 sprintf(name, "DIFFUSE");
422 else if (src_nr == 9)
423 sprintf(name, "SPECULAR");
424 else if (src_nr == 10)
425 sprintf(name, "FOG");
426 else {
427 assert(0);
428 sprintf(name, "RESERVED");
429 }
430 break;
431 case 2:
432 sprintf(name, "C%d", src_nr);
433 assert(src_nr <= 31);
434 break;
435 case 4:
436 sprintf(name, "oC");
437 assert(src_nr == 0);
438 break;
439 case 5:
440 sprintf(name, "oD");
441 assert(src_nr == 0);
442 break;
443 case 6:
444 sprintf(name, "U%d", src_nr);
445 assert(src_nr <= 3);
446 break;
447 default:
448 sprintf(name, "RESERVED");
449 assert(0);
450 break;
451 }
452 }
453
454 static void
gen3_get_instruction_src0(uint32_t * data,int i,char * srcname)455 gen3_get_instruction_src0(uint32_t *data, int i, char *srcname)
456 {
457 uint32_t a0 = data[i];
458 uint32_t a1 = data[i + 1];
459 int src_nr = (a0 >> 2) & 0x1f;
460 const char *swizzle_x = gen3_get_channel_swizzle((a1 >> 28) & 0xf);
461 const char *swizzle_y = gen3_get_channel_swizzle((a1 >> 24) & 0xf);
462 const char *swizzle_z = gen3_get_channel_swizzle((a1 >> 20) & 0xf);
463 const char *swizzle_w = gen3_get_channel_swizzle((a1 >> 16) & 0xf);
464 char swizzle[100];
465
466 gen3_get_instruction_src_name((a0 >> 7) & 0x7, src_nr, srcname);
467 sprintf(swizzle, ".%s%s%s%s", swizzle_x, swizzle_y, swizzle_z, swizzle_w);
468 if (strcmp(swizzle, ".xyzw") != 0)
469 strcat(srcname, swizzle);
470 }
471
472 static void
gen3_get_instruction_src1(uint32_t * data,int i,char * srcname)473 gen3_get_instruction_src1(uint32_t *data, int i, char *srcname)
474 {
475 uint32_t a1 = data[i + 1];
476 uint32_t a2 = data[i + 2];
477 int src_nr = (a1 >> 8) & 0x1f;
478 const char *swizzle_x = gen3_get_channel_swizzle((a1 >> 4) & 0xf);
479 const char *swizzle_y = gen3_get_channel_swizzle((a1 >> 0) & 0xf);
480 const char *swizzle_z = gen3_get_channel_swizzle((a2 >> 28) & 0xf);
481 const char *swizzle_w = gen3_get_channel_swizzle((a2 >> 24) & 0xf);
482 char swizzle[100];
483
484 gen3_get_instruction_src_name((a1 >> 13) & 0x7, src_nr, srcname);
485 sprintf(swizzle, ".%s%s%s%s", swizzle_x, swizzle_y, swizzle_z, swizzle_w);
486 if (strcmp(swizzle, ".xyzw") != 0)
487 strcat(srcname, swizzle);
488 }
489
490 static void
gen3_get_instruction_src2(uint32_t * data,int i,char * srcname)491 gen3_get_instruction_src2(uint32_t *data, int i, char *srcname)
492 {
493 uint32_t a2 = data[i + 2];
494 int src_nr = (a2 >> 16) & 0x1f;
495 const char *swizzle_x = gen3_get_channel_swizzle((a2 >> 12) & 0xf);
496 const char *swizzle_y = gen3_get_channel_swizzle((a2 >> 8) & 0xf);
497 const char *swizzle_z = gen3_get_channel_swizzle((a2 >> 4) & 0xf);
498 const char *swizzle_w = gen3_get_channel_swizzle((a2 >> 0) & 0xf);
499 char swizzle[100];
500
501 gen3_get_instruction_src_name((a2 >> 21) & 0x7, src_nr, srcname);
502 sprintf(swizzle, ".%s%s%s%s", swizzle_x, swizzle_y, swizzle_z, swizzle_w);
503 if (strcmp(swizzle, ".xyzw") != 0)
504 strcat(srcname, swizzle);
505 }
506
507 static void
gen3_get_instruction_addr(uint32_t src_type,uint32_t src_nr,char * name)508 gen3_get_instruction_addr(uint32_t src_type, uint32_t src_nr, char *name)
509 {
510 switch (src_type) {
511 case 0:
512 sprintf(name, "R%d", src_nr);
513 assert(src_nr <= 15);
514 break;
515 case 1:
516 if (src_nr < 8)
517 sprintf(name, "T%d", src_nr);
518 else if (src_nr == 8)
519 sprintf(name, "DIFFUSE");
520 else if (src_nr == 9)
521 sprintf(name, "SPECULAR");
522 else if (src_nr == 10)
523 sprintf(name, "FOG");
524 else {
525 assert(0);
526 sprintf(name, "RESERVED");
527 }
528 break;
529 case 4:
530 sprintf(name, "oC");
531 assert(src_nr == 0);
532 break;
533 case 5:
534 sprintf(name, "oD");
535 assert(src_nr == 0);
536 break;
537 default:
538 assert(0);
539 sprintf(name, "RESERVED");
540 break;
541 }
542 }
543
544 static void
gen3_decode_alu1(uint32_t * data,uint32_t offset,int i,char * instr_prefix,const char * op_name)545 gen3_decode_alu1(uint32_t *data, uint32_t offset,
546 int i, char *instr_prefix, const char *op_name)
547 {
548 char dst[100], src0[100];
549
550 gen3_get_instruction_dst(data, i, dst, 1);
551 gen3_get_instruction_src0(data, i, src0);
552
553 kgem_debug_print(data, offset, i++, "%s: %s %s, %s\n", instr_prefix,
554 op_name, dst, src0);
555 kgem_debug_print(data, offset, i++, "%s\n", instr_prefix);
556 kgem_debug_print(data, offset, i++, "%s\n", instr_prefix);
557 }
558
559 static void
gen3_decode_alu2(uint32_t * data,uint32_t offset,int i,char * instr_prefix,const char * op_name)560 gen3_decode_alu2(uint32_t *data, uint32_t offset,
561 int i, char *instr_prefix, const char *op_name)
562 {
563 char dst[100], src0[100], src1[100];
564
565 gen3_get_instruction_dst(data, i, dst, 1);
566 gen3_get_instruction_src0(data, i, src0);
567 gen3_get_instruction_src1(data, i, src1);
568
569 kgem_debug_print(data, offset, i++, "%s: %s %s, %s, %s\n", instr_prefix,
570 op_name, dst, src0, src1);
571 kgem_debug_print(data, offset, i++, "%s\n", instr_prefix);
572 kgem_debug_print(data, offset, i++, "%s\n", instr_prefix);
573 }
574
575 static void
gen3_decode_alu3(uint32_t * data,uint32_t offset,int i,char * instr_prefix,const char * op_name)576 gen3_decode_alu3(uint32_t *data, uint32_t offset,
577 int i, char *instr_prefix, const char *op_name)
578 {
579 char dst[100], src0[100], src1[100], src2[100];
580
581 gen3_get_instruction_dst(data, i, dst, 1);
582 gen3_get_instruction_src0(data, i, src0);
583 gen3_get_instruction_src1(data, i, src1);
584 gen3_get_instruction_src2(data, i, src2);
585
586 kgem_debug_print(data, offset, i++, "%s: %s %s, %s, %s, %s\n", instr_prefix,
587 op_name, dst, src0, src1, src2);
588 kgem_debug_print(data, offset, i++, "%s\n", instr_prefix);
589 kgem_debug_print(data, offset, i++, "%s\n", instr_prefix);
590 }
591
592 static void
gen3_decode_tex(uint32_t * data,uint32_t offset,int i,char * instr_prefix,const char * tex_name)593 gen3_decode_tex(uint32_t *data, uint32_t offset, int i, char *instr_prefix,
594 const char *tex_name)
595 {
596 uint32_t t0 = data[i];
597 uint32_t t1 = data[i + 1];
598 char dst_name[100];
599 char addr_name[100];
600 int sampler_nr;
601
602 gen3_get_instruction_dst(data, i, dst_name, 0);
603 gen3_get_instruction_addr((t1 >> 24) & 0x7,
604 (t1 >> 17) & 0xf,
605 addr_name);
606 sampler_nr = t0 & 0xf;
607
608 kgem_debug_print(data, offset, i++, "%s: %s %s, S%d, %s\n", instr_prefix,
609 tex_name, dst_name, sampler_nr, addr_name);
610 kgem_debug_print(data, offset, i++, "%s\n", instr_prefix);
611 kgem_debug_print(data, offset, i++, "%s\n", instr_prefix);
612 }
613
614 static void
gen3_decode_dcl(uint32_t * data,uint32_t offset,int i,char * instr_prefix)615 gen3_decode_dcl(uint32_t *data, uint32_t offset, int i, char *instr_prefix)
616 {
617 uint32_t d0 = data[i];
618 const char *sampletype;
619 int dcl_nr = (d0 >> 14) & 0xf;
620 const char *dcl_x = d0 & (1 << 10) ? "x" : "";
621 const char *dcl_y = d0 & (1 << 11) ? "y" : "";
622 const char *dcl_z = d0 & (1 << 12) ? "z" : "";
623 const char *dcl_w = d0 & (1 << 13) ? "w" : "";
624 char dcl_mask[10];
625
626 switch ((d0 >> 19) & 0x3) {
627 case 1:
628 sprintf(dcl_mask, ".%s%s%s%s", dcl_x, dcl_y, dcl_z, dcl_w);
629 assert (strcmp(dcl_mask, "."));
630
631 assert(dcl_nr <= 10);
632 if (dcl_nr < 8) {
633 if (strcmp(dcl_mask, ".x") != 0 &&
634 strcmp(dcl_mask, ".xy") != 0 &&
635 strcmp(dcl_mask, ".xz") != 0 &&
636 strcmp(dcl_mask, ".w") != 0 &&
637 strcmp(dcl_mask, ".xyzw") != 0) {
638 assert(0);
639 }
640 kgem_debug_print(data, offset, i++, "%s: DCL T%d%s\n", instr_prefix,
641 dcl_nr, dcl_mask);
642 } else {
643 if (strcmp(dcl_mask, ".xz") == 0)
644 assert(0);
645 else if (strcmp(dcl_mask, ".xw") == 0)
646 assert(0);
647 else if (strcmp(dcl_mask, ".xzw") == 0)
648 assert(0);
649
650 if (dcl_nr == 8) {
651 kgem_debug_print(data, offset, i++, "%s: DCL DIFFUSE%s\n", instr_prefix,
652 dcl_mask);
653 } else if (dcl_nr == 9) {
654 kgem_debug_print(data, offset, i++, "%s: DCL SPECULAR%s\n", instr_prefix,
655 dcl_mask);
656 } else if (dcl_nr == 10) {
657 kgem_debug_print(data, offset, i++, "%s: DCL FOG%s\n", instr_prefix,
658 dcl_mask);
659 }
660 }
661 kgem_debug_print(data, offset, i++, "%s\n", instr_prefix);
662 kgem_debug_print(data, offset, i++, "%s\n", instr_prefix);
663 break;
664 case 3:
665 switch ((d0 >> 22) & 0x3) {
666 case 0:
667 sampletype = "2D";
668 break;
669 case 1:
670 sampletype = "CUBE";
671 break;
672 case 2:
673 sampletype = "3D";
674 break;
675 default:
676 sampletype = "RESERVED";
677 break;
678 }
679 assert(dcl_nr <= 15);
680 kgem_debug_print(data, offset, i++, "%s: DCL S%d %s\n", instr_prefix,
681 dcl_nr, sampletype);
682 kgem_debug_print(data, offset, i++, "%s\n", instr_prefix);
683 kgem_debug_print(data, offset, i++, "%s\n", instr_prefix);
684 break;
685 default:
686 kgem_debug_print(data, offset, i++, "%s: DCL RESERVED%d\n", instr_prefix, dcl_nr);
687 kgem_debug_print(data, offset, i++, "%s\n", instr_prefix);
688 kgem_debug_print(data, offset, i++, "%s\n", instr_prefix);
689 }
690 }
691
692 static void
gen3_decode_instruction(uint32_t * data,uint32_t offset,int i,char * instr_prefix)693 gen3_decode_instruction(uint32_t *data, uint32_t offset,
694 int i, char *instr_prefix)
695 {
696 switch ((data[i] >> 24) & 0x1f) {
697 case 0x0:
698 kgem_debug_print(data, offset, i++, "%s: NOP\n", instr_prefix);
699 kgem_debug_print(data, offset, i++, "%s\n", instr_prefix);
700 kgem_debug_print(data, offset, i++, "%s\n", instr_prefix);
701 break;
702 case 0x01:
703 gen3_decode_alu2(data, offset, i, instr_prefix, "ADD");
704 break;
705 case 0x02:
706 gen3_decode_alu1(data, offset, i, instr_prefix, "MOV");
707 break;
708 case 0x03:
709 gen3_decode_alu2(data, offset, i, instr_prefix, "MUL");
710 break;
711 case 0x04:
712 gen3_decode_alu3(data, offset, i, instr_prefix, "MAD");
713 break;
714 case 0x05:
715 gen3_decode_alu3(data, offset, i, instr_prefix, "DP2ADD");
716 break;
717 case 0x06:
718 gen3_decode_alu2(data, offset, i, instr_prefix, "DP3");
719 break;
720 case 0x07:
721 gen3_decode_alu2(data, offset, i, instr_prefix, "DP4");
722 break;
723 case 0x08:
724 gen3_decode_alu1(data, offset, i, instr_prefix, "FRC");
725 break;
726 case 0x09:
727 gen3_decode_alu1(data, offset, i, instr_prefix, "RCP");
728 break;
729 case 0x0a:
730 gen3_decode_alu1(data, offset, i, instr_prefix, "RSQ");
731 break;
732 case 0x0b:
733 gen3_decode_alu1(data, offset, i, instr_prefix, "EXP");
734 break;
735 case 0x0c:
736 gen3_decode_alu1(data, offset, i, instr_prefix, "LOG");
737 break;
738 case 0x0d:
739 gen3_decode_alu2(data, offset, i, instr_prefix, "CMP");
740 break;
741 case 0x0e:
742 gen3_decode_alu2(data, offset, i, instr_prefix, "MIN");
743 break;
744 case 0x0f:
745 gen3_decode_alu2(data, offset, i, instr_prefix, "MAX");
746 break;
747 case 0x10:
748 gen3_decode_alu1(data, offset, i, instr_prefix, "FLR");
749 break;
750 case 0x11:
751 gen3_decode_alu1(data, offset, i, instr_prefix, "MOD");
752 break;
753 case 0x12:
754 gen3_decode_alu1(data, offset, i, instr_prefix, "TRC");
755 break;
756 case 0x13:
757 gen3_decode_alu2(data, offset, i, instr_prefix, "SGE");
758 break;
759 case 0x14:
760 gen3_decode_alu2(data, offset, i, instr_prefix, "SLT");
761 break;
762 case 0x15:
763 gen3_decode_tex(data, offset, i, instr_prefix, "TEXLD");
764 break;
765 case 0x16:
766 gen3_decode_tex(data, offset, i, instr_prefix, "TEXLDP");
767 break;
768 case 0x17:
769 gen3_decode_tex(data, offset, i, instr_prefix, "TEXLDB");
770 break;
771 case 0x19:
772 gen3_decode_dcl(data, offset, i, instr_prefix);
773 break;
774 default:
775 kgem_debug_print(data, offset, i++, "%s: unknown\n", instr_prefix);
776 kgem_debug_print(data, offset, i++, "%s\n", instr_prefix);
777 kgem_debug_print(data, offset, i++, "%s\n", instr_prefix);
778 break;
779 }
780 }
781
782 static const char *
gen3_decode_compare_func(uint32_t op)783 gen3_decode_compare_func(uint32_t op)
784 {
785 switch (op&0x7) {
786 case 0: return "always";
787 case 1: return "never";
788 case 2: return "less";
789 case 3: return "equal";
790 case 4: return "lequal";
791 case 5: return "greater";
792 case 6: return "notequal";
793 case 7: return "gequal";
794 }
795 return "";
796 }
797
798 static const char *
gen3_decode_stencil_op(uint32_t op)799 gen3_decode_stencil_op(uint32_t op)
800 {
801 switch (op&0x7) {
802 case 0: return "keep";
803 case 1: return "zero";
804 case 2: return "replace";
805 case 3: return "incr_sat";
806 case 4: return "decr_sat";
807 case 5: return "greater";
808 case 6: return "incr";
809 case 7: return "decr";
810 }
811 return "";
812 }
813
814 #if 0
815 /* part of MODES_4 */
816 static const char *
817 gen3_decode_logic_op(uint32_t op)
818 {
819 switch (op&0xf) {
820 case 0: return "clear";
821 case 1: return "nor";
822 case 2: return "and_inv";
823 case 3: return "copy_inv";
824 case 4: return "and_rvrse";
825 case 5: return "inv";
826 case 6: return "xor";
827 case 7: return "nand";
828 case 8: return "and";
829 case 9: return "equiv";
830 case 10: return "noop";
831 case 11: return "or_inv";
832 case 12: return "copy";
833 case 13: return "or_rvrse";
834 case 14: return "or";
835 case 15: return "set";
836 }
837 return "";
838 }
839 #endif
840
841 static const char *
gen3_decode_blend_fact(uint32_t op)842 gen3_decode_blend_fact(uint32_t op)
843 {
844 switch (op&0xf) {
845 case 1: return "zero";
846 case 2: return "one";
847 case 3: return "src_colr";
848 case 4: return "inv_src_colr";
849 case 5: return "src_alpha";
850 case 6: return "inv_src_alpha";
851 case 7: return "dst_alpha";
852 case 8: return "inv_dst_alpha";
853 case 9: return "dst_colr";
854 case 10: return "inv_dst_colr";
855 case 11: return "src_alpha_sat";
856 case 12: return "cnst_colr";
857 case 13: return "inv_cnst_colr";
858 case 14: return "cnst_alpha";
859 case 15: return "inv_const_alpha";
860 }
861 return "";
862 }
863
864 static const char *
decode_tex_coord_mode(uint32_t mode)865 decode_tex_coord_mode(uint32_t mode)
866 {
867 switch (mode&0x7) {
868 case 0: return "wrap";
869 case 1: return "mirror";
870 case 2: return "clamp_edge";
871 case 3: return "cube";
872 case 4: return "clamp_border";
873 case 5: return "mirror_once";
874 }
875 return "";
876 }
877
878 static const char *
gen3_decode_sample_filter(uint32_t mode)879 gen3_decode_sample_filter(uint32_t mode)
880 {
881 switch (mode&0x7) {
882 case 0: return "nearest";
883 case 1: return "linear";
884 case 2: return "anisotropic";
885 case 3: return "4x4_1";
886 case 4: return "4x4_2";
887 case 5: return "4x4_flat";
888 case 6: return "6x5_mono";
889 }
890 return "";
891 }
892
893 static int
gen3_decode_load_state_immediate_1(struct kgem * kgem,uint32_t offset)894 gen3_decode_load_state_immediate_1(struct kgem *kgem, uint32_t offset)
895 {
896 const uint32_t *data = kgem->batch + offset;
897 int len, i, word;
898
899 kgem_debug_print(data, offset, 0, "3DSTATE_LOAD_STATE_IMMEDIATE_1\n");
900 len = (data[0] & 0x0000000f) + 2;
901 i = 1;
902 for (word = 0; word <= 8; word++) {
903 if (data[0] & (1 << (4 + word))) {
904 switch (word) {
905 case 0:
906 kgem_debug_print(data, offset, i, "S0: vbo offset: 0x%08x%s\n",
907 data[i]&(~1),data[i]&1?", auto cache invalidate disabled":"");
908 gen3_update_vertex_buffer_addr(kgem, offset + i);
909 break;
910 case 1:
911 kgem_debug_print(data, offset, i, "S1: vertex width: %i, vertex pitch: %i\n",
912 (data[i]>>24)&0x3f,(data[i]>>16)&0x3f);
913 gen3_update_vertex_buffer_pitch(kgem, offset + i);
914 break;
915 case 2:
916 {
917 char buf[200];
918 int len = 0;
919 int tex_num;
920 for (tex_num = 0; tex_num < 8; tex_num++) {
921 switch((data[i]>>tex_num*4)&0xf) {
922 case 0: len += sprintf(buf + len, "%i=2D ", tex_num); break;
923 case 1: len += sprintf(buf + len, "%i=3D ", tex_num); break;
924 case 2: len += sprintf(buf + len, "%i=4D ", tex_num); break;
925 case 3: len += sprintf(buf + len, "%i=1D ", tex_num); break;
926 case 4: len += sprintf(buf + len, "%i=2D_16 ", tex_num); break;
927 case 5: len += sprintf(buf + len, "%i=4D_16 ", tex_num); break;
928 case 0xf: len += sprintf(buf + len, "%i=NP ", tex_num); break;
929 }
930 }
931 kgem_debug_print(data, offset, i, "S2: texcoord formats: %s\n", buf);
932 gen3_update_vertex_texcoords(kgem, data[i]);
933 }
934
935 break;
936 case 3:
937 kgem_debug_print(data, offset, i, "S3: not documented\n");
938 break;
939 case 4:
940 {
941 const char *cullmode = "";
942 const char *vfmt_xyzw = "";
943 switch((data[i]>>13)&0x3) {
944 case 0: cullmode = "both"; break;
945 case 1: cullmode = "none"; break;
946 case 2: cullmode = "cw"; break;
947 case 3: cullmode = "ccw"; break;
948 }
949 switch(data[i] & (7<<6 | 1<<2)) {
950 case 1<<6: vfmt_xyzw = "XYZ,"; break;
951 case 2<<6: vfmt_xyzw = "XYZW,"; break;
952 case 3<<6: vfmt_xyzw = "XY,"; break;
953 case 4<<6: vfmt_xyzw = "XYW,"; break;
954 case 1<<6 | 1<<2: vfmt_xyzw = "XYZF,"; break;
955 case 2<<6 | 1<<2: vfmt_xyzw = "XYZWF,"; break;
956 case 3<<6 | 1<<2: vfmt_xyzw = "XYF,"; break;
957 case 4<<6 | 1<<2: vfmt_xyzw = "XYWF,"; break;
958 }
959 kgem_debug_print(data, offset, i, "S4: point_width=%i, line_width=%.1f,"
960 "%s%s%s%s%s cullmode=%s, vfmt=%s%s%s%s%s%s%s%s "
961 "%s%s%s\n",
962 (data[i]>>23)&0x1ff,
963 ((data[i]>>19)&0xf) / 2.0,
964 data[i]&(0xf<<15)?" flatshade=":"",
965 data[i]&(1<<18)?"Alpha,":"",
966 data[i]&(1<<17)?"Fog,":"",
967 data[i]&(1<<16)?"Specular,":"",
968 data[i]&(1<<15)?"Color,":"",
969 cullmode,
970 data[i]&(1<<12)?"PointWidth,":"",
971 data[i]&(1<<11)?"SpecFog,":"",
972 data[i]&(1<<10)?"Color,":"",
973 data[i]&(1<<9)?"DepthOfs,":"",
974 vfmt_xyzw,
975 data[i]&(1<<9)?"FogParam,":"",
976 data[i]&(1<<5)?"force default diffuse, ":"",
977 data[i]&(1<<4)?"force default specular, ":"",
978 data[i]&(1<<3)?"local depth ofs enable, ":"",
979 data[i]&(1<<1)?"point sprite enable, ":"",
980 data[i]&(1<<0)?"line AA enable, ":"");
981 gen3_update_vertex_elements(kgem, data[i]);
982 break;
983 }
984 case 5:
985 {
986 kgem_debug_print(data, offset, i, "S5:%s%s%s%s%s"
987 "%s%s%s%s stencil_ref=0x%x, stencil_test=%s, "
988 "stencil_fail=%s, stencil_pass_z_fail=%s, "
989 "stencil_pass_z_pass=%s, %s%s%s%s\n",
990 data[i]&(0xf<<28)?" write_disable=":"",
991 data[i]&(1<<31)?"Alpha,":"",
992 data[i]&(1<<30)?"Red,":"",
993 data[i]&(1<<29)?"Green,":"",
994 data[i]&(1<<28)?"Blue,":"",
995 data[i]&(1<<27)?" force default point size,":"",
996 data[i]&(1<<26)?" last pixel enable,":"",
997 data[i]&(1<<25)?" global depth ofs enable,":"",
998 data[i]&(1<<24)?" fog enable,":"",
999 (data[i]>>16)&0xff,
1000 gen3_decode_compare_func(data[i]>>13),
1001 gen3_decode_stencil_op(data[i]>>10),
1002 gen3_decode_stencil_op(data[i]>>7),
1003 gen3_decode_stencil_op(data[i]>>4),
1004 data[i]&(1<<3)?"stencil write enable, ":"",
1005 data[i]&(1<<2)?"stencil test enable, ":"",
1006 data[i]&(1<<1)?"color dither enable, ":"",
1007 data[i]&(1<<0)?"logicop enable, ":"");
1008 }
1009 break;
1010 case 6:
1011 kgem_debug_print(data, offset, i, "S6: %salpha_test=%s, alpha_ref=0x%x, "
1012 "depth_test=%s, %ssrc_blnd_fct=%s, dst_blnd_fct=%s, "
1013 "%s%stristrip_provoking_vertex=%i\n",
1014 data[i]&(1<<31)?"alpha test enable, ":"",
1015 gen3_decode_compare_func(data[i]>>28),
1016 data[i]&(0xff<<20),
1017 gen3_decode_compare_func(data[i]>>16),
1018 data[i]&(1<<15)?"cbuf blend enable, ":"",
1019 gen3_decode_blend_fact(data[i]>>8),
1020 gen3_decode_blend_fact(data[i]>>4),
1021 data[i]&(1<<3)?"depth write enable, ":"",
1022 data[i]&(1<<2)?"cbuf write enable, ":"",
1023 data[i]&(0x3));
1024 break;
1025 case 7:
1026 kgem_debug_print(data, offset, i, "S7: depth offset constant: 0x%08x\n", data[i]);
1027 break;
1028 }
1029 i++;
1030 }
1031 }
1032
1033 assert(len == i);
1034 return len;
1035 }
1036
1037 static int
gen3_decode_3d_1d(struct kgem * kgem,uint32_t offset)1038 gen3_decode_3d_1d(struct kgem *kgem, uint32_t offset)
1039 {
1040 uint32_t *data = kgem->batch + offset;
1041 unsigned int len, i, c, idx, word, map, sampler, instr;
1042 const char *format, *zformat, *type;
1043 uint32_t opcode;
1044 static const struct {
1045 uint32_t opcode;
1046 int min_len;
1047 int max_len;
1048 const char *name;
1049 } opcodes_3d_1d[] = {
1050 { 0x86, 4, 4, "3DSTATE_CHROMA_KEY" },
1051 { 0x88, 2, 2, "3DSTATE_CONSTANT_BLEND_COLOR" },
1052 { 0x99, 2, 2, "3DSTATE_DEFAULT_DIFFUSE" },
1053 { 0x9a, 2, 2, "3DSTATE_DEFAULT_SPECULAR" },
1054 { 0x98, 2, 2, "3DSTATE_DEFAULT_Z" },
1055 { 0x97, 2, 2, "3DSTATE_DEPTH_OFFSET_SCALE" },
1056 { 0x9d, 65, 65, "3DSTATE_FILTER_COEFFICIENTS_4X4" },
1057 { 0x9e, 4, 4, "3DSTATE_MONO_FILTER" },
1058 { 0x89, 4, 4, "3DSTATE_FOG_MODE" },
1059 { 0x8f, 2, 16, "3DSTATE_MAP_PALLETE_LOAD_32" },
1060 { 0x83, 2, 2, "3DSTATE_SPAN_STIPPLE" },
1061 }, *opcode_3d_1d;
1062
1063 opcode = (data[0] & 0x00ff0000) >> 16;
1064
1065 switch (opcode) {
1066 case 0x07:
1067 /* This instruction is unusual. A 0 length means just 1 DWORD instead of
1068 * 2. The 0 length is specified in one place to be unsupported, but
1069 * stated to be required in another, and 0 length LOAD_INDIRECTs appear
1070 * to cause no harm at least.
1071 */
1072 kgem_debug_print(data, offset, 0, "3DSTATE_LOAD_INDIRECT\n");
1073 len = (data[0] & 0x000000ff) + 1;
1074 i = 1;
1075 if (data[0] & (0x01 << 8)) {
1076 kgem_debug_print(data, offset, i++, "SIS.0\n");
1077 kgem_debug_print(data, offset, i++, "SIS.1\n");
1078 }
1079 if (data[0] & (0x02 << 8)) {
1080 kgem_debug_print(data, offset, i++, "DIS.0\n");
1081 }
1082 if (data[0] & (0x04 << 8)) {
1083 kgem_debug_print(data, offset, i++, "SSB.0\n");
1084 kgem_debug_print(data, offset, i++, "SSB.1\n");
1085 }
1086 if (data[0] & (0x08 << 8)) {
1087 kgem_debug_print(data, offset, i++, "MSB.0\n");
1088 kgem_debug_print(data, offset, i++, "MSB.1\n");
1089 }
1090 if (data[0] & (0x10 << 8)) {
1091 kgem_debug_print(data, offset, i++, "PSP.0\n");
1092 kgem_debug_print(data, offset, i++, "PSP.1\n");
1093 }
1094 if (data[0] & (0x20 << 8)) {
1095 kgem_debug_print(data, offset, i++, "PSC.0\n");
1096 kgem_debug_print(data, offset, i++, "PSC.1\n");
1097 }
1098 assert(len == i);
1099 return len;
1100 case 0x04:
1101 return gen3_decode_load_state_immediate_1(kgem, offset);
1102 case 0x03:
1103 kgem_debug_print(data, offset, 0, "3DSTATE_LOAD_STATE_IMMEDIATE_2\n");
1104 len = (data[0] & 0x0000000f) + 2;
1105 i = 1;
1106 for (word = 6; word <= 14; word++) {
1107 if (data[0] & (1 << word)) {
1108 if (word == 6)
1109 kgem_debug_print(data, offset, i++, "TBCF\n");
1110 else if (word >= 7 && word <= 10) {
1111 kgem_debug_print(data, offset, i++, "TB%dC\n", word - 7);
1112 kgem_debug_print(data, offset, i++, "TB%dA\n", word - 7);
1113 } else if (word >= 11 && word <= 14) {
1114 kgem_debug_print(data, offset, i, "TM%dS0: offset=0x%08x, %s\n",
1115 word - 11,
1116 data[i]&0xfffffffe,
1117 data[i]&1?"use fence":"");
1118 i++;
1119 kgem_debug_print(data, offset, i, "TM%dS1: height=%i, width=%i, %s\n",
1120 word - 11,
1121 data[i]>>21, (data[i]>>10)&0x3ff,
1122 data[i]&2?(data[i]&1?"y-tiled":"x-tiled"):"");
1123 i++;
1124 kgem_debug_print(data, offset, i, "TM%dS2: pitch=%i, \n",
1125 word - 11,
1126 ((data[i]>>21) + 1)*4);
1127 i++;
1128 kgem_debug_print(data, offset, i++, "TM%dS3\n", word - 11);
1129 kgem_debug_print(data, offset, i++, "TM%dS4: dflt color\n", word - 11);
1130 }
1131 }
1132 }
1133 assert(len == i);
1134 return len;
1135 case 0x00:
1136 kgem_debug_print(data, offset, 0, "3DSTATE_MAP_STATE\n");
1137 len = (data[0] & 0x0000003f) + 2;
1138 kgem_debug_print(data, offset, 1, "mask\n");
1139
1140 i = 2;
1141 for (map = 0; map <= 15; map++) {
1142 if (data[1] & (1 << map)) {
1143 int width, height, pitch, dword;
1144 struct drm_i915_gem_relocation_entry *reloc;
1145 const char *tiling;
1146
1147 reloc = kgem_debug_get_reloc_entry(kgem, &data[i] - kgem->batch);
1148 assert(reloc->target_handle);
1149
1150 dword = data[i];
1151 kgem_debug_print(data, offset, i++, "map %d MS2 %s%s%s, handle=%d\n", map,
1152 dword&(1<<31)?"untrusted surface, ":"",
1153 dword&(1<<1)?"vertical line stride enable, ":"",
1154 dword&(1<<0)?"vertical ofs enable, ":"",
1155 reloc->target_handle);
1156
1157 dword = data[i];
1158 width = ((dword >> 10) & ((1 << 11) - 1))+1;
1159 height = ((dword >> 21) & ((1 << 11) - 1))+1;
1160
1161 tiling = "none";
1162 if (dword & (1 << 2))
1163 tiling = "fenced";
1164 else if (dword & (1 << 1))
1165 tiling = dword & (1 << 0) ? "Y" : "X";
1166 type = " BAD";
1167 format = " (invalid)";
1168 switch ((dword>>7) & 0x7) {
1169 case 1:
1170 type = "8";
1171 switch ((dword>>3) & 0xf) {
1172 case 0: format = "I"; break;
1173 case 1: format = "L"; break;
1174 case 4: format = "A"; break;
1175 case 5: format = " mono"; break;
1176 }
1177 break;
1178 case 2:
1179 type = "16";
1180 switch ((dword>>3) & 0xf) {
1181 case 0: format = " rgb565"; break;
1182 case 1: format = " argb1555"; break;
1183 case 2: format = " argb4444"; break;
1184 case 3: format = " ay88"; break;
1185 case 5: format = " 88dvdu"; break;
1186 case 6: format = " bump655"; break;
1187 case 7: format = "I"; break;
1188 case 8: format = "L"; break;
1189 case 9: format = "A"; break;
1190 }
1191 break;
1192 case 3:
1193 type = "32";
1194 switch ((dword>>3) & 0xf) {
1195 case 0: format = " argb8888"; break;
1196 case 1: format = " abgr8888"; break;
1197 case 2: format = " xrgb8888"; break;
1198 case 3: format = " xbgr8888"; break;
1199 case 4: format = " qwvu8888"; break;
1200 case 5: format = " axvu8888"; break;
1201 case 6: format = " lxvu8888"; break;
1202 case 7: format = " xlvu8888"; break;
1203 case 8: format = " argb2101010"; break;
1204 case 9: format = " abgr2101010"; break;
1205 case 10: format = " awvu2101010"; break;
1206 case 11: format = " gr1616"; break;
1207 case 12: format = " vu1616"; break;
1208 case 13: format = " xI824"; break;
1209 case 14: format = " xA824"; break;
1210 case 15: format = " xL824"; break;
1211 }
1212 break;
1213 case 5:
1214 type = "422";
1215 switch ((dword>>3) & 0xf) {
1216 case 0: format = " yuv_swapy"; break;
1217 case 1: format = " yuv"; break;
1218 case 2: format = " yuv_swapuv"; break;
1219 case 3: format = " yuv_swapuvy"; break;
1220 }
1221 break;
1222 case 6:
1223 type = "compressed";
1224 switch ((dword>>3) & 0x7) {
1225 case 0: format = " dxt1"; break;
1226 case 1: format = " dxt2_3"; break;
1227 case 2: format = " dxt4_5"; break;
1228 case 3: format = " fxt1"; break;
1229 case 4: format = " dxt1_rb"; break;
1230 }
1231 break;
1232 case 7:
1233 type = "4b indexed";
1234 switch ((dword>>3) & 0xf) {
1235 case 7: format = " argb8888"; break;
1236 }
1237 break;
1238 default:
1239 format = "BAD";
1240 break;
1241 }
1242 dword = data[i];
1243 kgem_debug_print(data, offset, i++, "map %d MS3 [width=%d, height=%d, format=%s%s, tiling=%s%s]\n",
1244 map, width, height, type, format, tiling,
1245 dword&(1<<9)?" palette select":"");
1246
1247 dword = data[i];
1248 pitch = 4*(((dword >> 21) & ((1 << 11) - 1))+1);
1249 kgem_debug_print(data, offset, i++, "map %d MS4 [pitch=%d, max_lod=%i, vol_depth=%i, cube_face_ena=%x, %s]\n",
1250 map, pitch,
1251 (dword>>9)&0x3f, dword&0xff, (dword>>15)&0x3f,
1252 dword&(1<<8)?"miplayout legacy":"miplayout right");
1253 }
1254 }
1255 assert(len == i);
1256 return len;
1257 case 0x06:
1258 kgem_debug_print(data, offset, 0, "3DSTATE_PIXEL_SHADER_CONSTANTS\n");
1259 len = (data[0] & 0x000000ff) + 2;
1260
1261 i = 2;
1262 for (c = 0; c <= 31; c++) {
1263 if (data[1] & (1 << c)) {
1264 kgem_debug_print(data, offset, i, "C%d.X = %f\n",
1265 c, int_as_float(data[i]));
1266 i++;
1267 kgem_debug_print(data, offset, i, "C%d.Y = %f\n",
1268 c, int_as_float(data[i]));
1269 i++;
1270 kgem_debug_print(data, offset, i, "C%d.Z = %f\n",
1271 c, int_as_float(data[i]));
1272 i++;
1273 kgem_debug_print(data, offset, i, "C%d.W = %f\n",
1274 c, int_as_float(data[i]));
1275 i++;
1276 }
1277 }
1278 assert(len == i);
1279 return len;
1280 case 0x05:
1281 kgem_debug_print(data, offset, 0, "3DSTATE_PIXEL_SHADER_PROGRAM\n");
1282 len = (data[0] & 0x000000ff) + 2;
1283 assert(((len-1) % 3) == 0);
1284 assert(len <= 370);
1285 i = 1;
1286 for (instr = 0; instr < (len - 1) / 3; instr++) {
1287 char instr_prefix[10];
1288
1289 sprintf(instr_prefix, "PS%03d", instr);
1290 gen3_decode_instruction(data, offset, i, instr_prefix);
1291 i += 3;
1292 }
1293 return len;
1294 case 0x01:
1295 kgem_debug_print(data, offset, 0, "3DSTATE_SAMPLER_STATE\n");
1296 kgem_debug_print(data, offset, 1, "mask\n");
1297 len = (data[0] & 0x0000003f) + 2;
1298 i = 2;
1299 for (sampler = 0; sampler <= 15; sampler++) {
1300 if (data[1] & (1 << sampler)) {
1301 uint32_t dword;
1302 const char *mip_filter = "";
1303 dword = data[i];
1304 switch ((dword>>20)&0x3) {
1305 case 0: mip_filter = "none"; break;
1306 case 1: mip_filter = "nearest"; break;
1307 case 3: mip_filter = "linear"; break;
1308 }
1309 kgem_debug_print(data, offset, i++, "sampler %d SS2:%s%s%s "
1310 "base_mip_level=%i, mip_filter=%s, mag_filter=%s, min_filter=%s "
1311 "lod_bias=%.2f,%s max_aniso=%i, shadow_func=%s\n", sampler,
1312 dword&(1<<31)?" reverse gamma,":"",
1313 dword&(1<<30)?" packed2planar,":"",
1314 dword&(1<<29)?" colorspace conversion,":"",
1315 (dword>>22)&0x1f,
1316 mip_filter,
1317 gen3_decode_sample_filter(dword>>17),
1318 gen3_decode_sample_filter(dword>>14),
1319 ((dword>>5)&0x1ff)/(0x10*1.0),
1320 dword&(1<<4)?" shadow,":"",
1321 dword&(1<<3)?4:2,
1322 gen3_decode_compare_func(dword));
1323 dword = data[i];
1324 kgem_debug_print(data, offset, i++, "sampler %d SS3: min_lod=%.2f,%s "
1325 "tcmode_x=%s, tcmode_y=%s, tcmode_z=%s,%s texmap_idx=%i,%s\n",
1326 sampler, ((dword>>24)&0xff)/(0x10*1.0),
1327 dword&(1<<17)?" kill pixel enable,":"",
1328 decode_tex_coord_mode(dword>>12),
1329 decode_tex_coord_mode(dword>>9),
1330 decode_tex_coord_mode(dword>>6),
1331 dword&(1<<5)?" normalized coords,":"",
1332 (dword>>1)&0xf,
1333 dword&(1<<0)?" deinterlacer,":"");
1334 kgem_debug_print(data, offset, i++, "sampler %d SS4: border color\n",
1335 sampler);
1336 }
1337 }
1338 assert(len == i);
1339 return len;
1340 case 0x85:
1341 len = (data[0] & 0x0000000f) + 2;
1342 assert(len == 2);
1343
1344 kgem_debug_print(data, offset, 0,
1345 "3DSTATE_DEST_BUFFER_VARIABLES\n");
1346
1347 switch ((data[1] >> 8) & 0xf) {
1348 case 0x0: format = "g8"; break;
1349 case 0x1: format = "x1r5g5b5"; break;
1350 case 0x2: format = "r5g6b5"; break;
1351 case 0x3: format = "a8r8g8b8"; break;
1352 case 0x4: format = "ycrcb_swapy"; break;
1353 case 0x5: format = "ycrcb_normal"; break;
1354 case 0x6: format = "ycrcb_swapuv"; break;
1355 case 0x7: format = "ycrcb_swapuvy"; break;
1356 case 0x8: format = "a4r4g4b4"; break;
1357 case 0x9: format = "a1r5g5b5"; break;
1358 case 0xa: format = "a2r10g10b10"; break;
1359 default: format = "BAD"; break;
1360 }
1361 switch ((data[1] >> 2) & 0x3) {
1362 case 0x0: zformat = "u16"; break;
1363 case 0x1: zformat = "f16"; break;
1364 case 0x2: zformat = "u24x8"; break;
1365 default: zformat = "BAD"; break;
1366 }
1367 kgem_debug_print(data, offset, 1, "%s format, %s depth format, early Z %sabled\n",
1368 format, zformat,
1369 (data[1] & (1 << 31)) ? "en" : "dis");
1370 return len;
1371
1372 case 0x8e:
1373 {
1374 const char *name, *tiling;
1375
1376 len = (data[0] & 0x0000000f) + 2;
1377 assert(len == 3);
1378
1379 switch((data[1] >> 24) & 0x7) {
1380 case 0x3: name = "color"; break;
1381 case 0x7: name = "depth"; break;
1382 default: name = "unknown"; break;
1383 }
1384
1385 tiling = "none";
1386 if (data[1] & (1 << 23))
1387 tiling = "fenced";
1388 else if (data[1] & (1 << 22))
1389 tiling = data[1] & (1 << 21) ? "Y" : "X";
1390
1391 kgem_debug_print(data, offset, 0, "3DSTATE_BUFFER_INFO\n");
1392 kgem_debug_print(data, offset, 1, "%s, tiling = %s, pitch=%d\n", name, tiling, data[1]&0xffff);
1393
1394 kgem_debug_print(data, offset, 2, "address\n");
1395 return len;
1396 }
1397 case 0x81:
1398 len = (data[0] & 0x0000000f) + 2;
1399 assert(len == 3);
1400
1401 kgem_debug_print(data, offset, 0,
1402 "3DSTATE_SCISSOR_RECTANGLE\n");
1403 kgem_debug_print(data, offset, 1, "(%d,%d)\n",
1404 data[1] & 0xffff, data[1] >> 16);
1405 kgem_debug_print(data, offset, 2, "(%d,%d)\n",
1406 data[2] & 0xffff, data[2] >> 16);
1407
1408 return len;
1409 case 0x80:
1410 len = (data[0] & 0x0000000f) + 2;
1411 assert(len == 5);
1412
1413 kgem_debug_print(data, offset, 0,
1414 "3DSTATE_DRAWING_RECTANGLE\n");
1415 kgem_debug_print(data, offset, 1, "%s\n",
1416 data[1]&(1<<30)?"depth ofs disabled ":"");
1417 kgem_debug_print(data, offset, 2, "(%d,%d)\n",
1418 data[2] & 0xffff, data[2] >> 16);
1419 kgem_debug_print(data, offset, 3, "(%d,%d)\n",
1420 data[3] & 0xffff, data[3] >> 16);
1421 kgem_debug_print(data, offset, 4, "(%d,%d)\n",
1422 (int16_t)(data[4] & 0xffff),
1423 (int16_t)(data[4] >> 16));
1424
1425 return len;
1426 case 0x9c:
1427 len = (data[0] & 0x0000000f) + 2;
1428 assert(len == 7);
1429
1430 kgem_debug_print(data, offset, 0,
1431 "3DSTATE_CLEAR_PARAMETERS\n");
1432 kgem_debug_print(data, offset, 1, "prim_type=%s, clear=%s%s%s\n",
1433 data[1]&(1<<16)?"CLEAR_RECT":"ZONE_INIT",
1434 data[1]&(1<<2)?"color,":"",
1435 data[1]&(1<<1)?"depth,":"",
1436 data[1]&(1<<0)?"stencil,":"");
1437 kgem_debug_print(data, offset, 2, "clear color\n");
1438 kgem_debug_print(data, offset, 3, "clear depth/stencil\n");
1439 kgem_debug_print(data, offset, 4, "color value (rgba8888)\n");
1440 kgem_debug_print(data, offset, 5, "depth value %f\n",
1441 int_as_float(data[5]));
1442 kgem_debug_print(data, offset, 6, "clear stencil\n");
1443 return len;
1444 }
1445
1446 for (idx = 0; idx < ARRAY_SIZE(opcodes_3d_1d); idx++) {
1447 opcode_3d_1d = &opcodes_3d_1d[idx];
1448 if (((data[0] & 0x00ff0000) >> 16) == opcode_3d_1d->opcode) {
1449 len = (data[0] & 0xf) + 2;
1450 kgem_debug_print(data, offset, 0, "%s\n", opcode_3d_1d->name);
1451 for (i = 1; i < len; i++)
1452 kgem_debug_print(data, offset, i, "dword %d\n", i);
1453
1454 return len;
1455 }
1456 }
1457
1458 kgem_debug_print(data, offset, 0, "3D UNKNOWN: 3d_1d opcode = 0x%x\n", opcode);
1459 assert(0);
1460 return 1;
1461 }
1462
1463 #define VERTEX_OUT(fmt, ...) do { \
1464 kgem_debug_print(data, offset, i, " V%d."fmt"\n", vertex, __VA_ARGS__); \
1465 i++; \
1466 } while (0)
1467
1468 static int
gen3_decode_3d_primitive(struct kgem * kgem,uint32_t offset)1469 gen3_decode_3d_primitive(struct kgem *kgem, uint32_t offset)
1470 {
1471 uint32_t *data = kgem->batch + offset;
1472 char immediate = (data[0] & (1 << 23)) == 0;
1473 unsigned int len, i, ret;
1474 const char *primtype;
1475 unsigned int vertex = 0;
1476
1477 switch ((data[0] >> 18) & 0xf) {
1478 case 0x0: primtype = "TRILIST"; break;
1479 case 0x1: primtype = "TRISTRIP"; break;
1480 case 0x2: primtype = "TRISTRIP_REVERSE"; break;
1481 case 0x3: primtype = "TRIFAN"; break;
1482 case 0x4: primtype = "POLYGON"; break;
1483 case 0x5: primtype = "LINELIST"; break;
1484 case 0x6: primtype = "LINESTRIP"; break;
1485 case 0x7: primtype = "RECTLIST"; break;
1486 case 0x8: primtype = "POINTLIST"; break;
1487 case 0x9: primtype = "DIB"; break;
1488 case 0xa: primtype = "CLEAR_RECT"; assert(0); break;
1489 default: primtype = "unknown"; break;
1490 }
1491
1492 gen3_update_vertex_elements_offsets(kgem);
1493
1494 /* XXX: 3DPRIM_DIB not supported */
1495 if (immediate) {
1496 len = (data[0] & 0x0003ffff) + 2;
1497 kgem_debug_print(data, offset, 0, "3DPRIMITIVE inline %s\n", primtype);
1498 for (i = 1; i < len; ) {
1499 ErrorF(" [%d]: ", vertex);
1500 i += inline_vertex_out(kgem, data + i) / sizeof(uint32_t);
1501 ErrorF("\n");
1502 vertex++;
1503 }
1504
1505 ret = len;
1506 } else {
1507 /* indirect vertices */
1508 len = data[0] & 0x0000ffff; /* index count */
1509 if (data[0] & (1 << 17)) {
1510 /* random vertex access */
1511 kgem_debug_print(data, offset, 0,
1512 "3DPRIMITIVE random indirect %s (%d)\n", primtype, len);
1513 assert(0);
1514 if (len == 0) {
1515 /* vertex indices continue until 0xffff is found */
1516 } else {
1517 /* fixed size vertex index buffer */
1518 }
1519 ret = (len + 1) / 2 + 1;
1520 goto out;
1521 } else {
1522 /* sequential vertex access */
1523 vertex = data[1] & 0xffff;
1524 kgem_debug_print(data, offset, 0,
1525 "3DPRIMITIVE sequential indirect %s, %d starting from "
1526 "%d\n", primtype, len, vertex);
1527 kgem_debug_print(data, offset, 1, " start\n");
1528 for (i = 0; i < len; i++) {
1529 ErrorF(" [%d]: ", vertex);
1530 indirect_vertex_out(kgem, vertex++);
1531 ErrorF("\n");
1532 }
1533 ret = 2;
1534 goto out;
1535 }
1536 }
1537
1538 out:
1539 return ret;
1540 }
1541
kgem_gen3_decode_3d(struct kgem * kgem,uint32_t offset)1542 int kgem_gen3_decode_3d(struct kgem *kgem, uint32_t offset)
1543 {
1544 static const struct {
1545 uint32_t opcode;
1546 int min_len;
1547 int max_len;
1548 const char *name;
1549 } opcodes[] = {
1550 { 0x06, 1, 1, "3DSTATE_ANTI_ALIASING" },
1551 { 0x08, 1, 1, "3DSTATE_BACKFACE_STENCIL_OPS" },
1552 { 0x09, 1, 1, "3DSTATE_BACKFACE_STENCIL_MASKS" },
1553 { 0x16, 1, 1, "3DSTATE_COORD_SET_BINDINGS" },
1554 { 0x15, 1, 1, "3DSTATE_FOG_COLOR" },
1555 { 0x0b, 1, 1, "3DSTATE_INDEPENDENT_ALPHA_BLEND" },
1556 { 0x0d, 1, 1, "3DSTATE_MODES_4" },
1557 { 0x0c, 1, 1, "3DSTATE_MODES_5" },
1558 { 0x07, 1, 1, "3DSTATE_RASTERIZATION_RULES" },
1559 };
1560 uint32_t *data = kgem->batch + offset;
1561 uint32_t opcode;
1562 unsigned int idx;
1563
1564 opcode = (data[0] & 0x1f000000) >> 24;
1565
1566 switch (opcode) {
1567 case 0x1f:
1568 return gen3_decode_3d_primitive(kgem, offset);
1569 case 0x1d:
1570 return gen3_decode_3d_1d(kgem, offset);
1571 case 0x1c:
1572 return gen3_decode_3d_1c(kgem, offset);
1573 }
1574
1575 for (idx = 0; idx < ARRAY_SIZE(opcodes); idx++) {
1576 if (opcode == opcodes[idx].opcode) {
1577 unsigned int len = 1, i;
1578
1579 kgem_debug_print(data, offset, 0, "%s\n", opcodes[idx].name);
1580 if (opcodes[idx].max_len > 1) {
1581 len = (data[0] & 0xff) + 2;
1582 assert(len >= opcodes[idx].min_len ||
1583 len <= opcodes[idx].max_len);
1584 }
1585
1586 for (i = 1; i < len; i++)
1587 kgem_debug_print(data, offset, i, "dword %d\n", i);
1588 return len;
1589 }
1590 }
1591
1592 kgem_debug_print(data, offset, 0, "3D UNKNOWN: 3d opcode = 0x%x\n", opcode);
1593 return 1;
1594 }
1595
1596
kgem_gen3_finish_state(struct kgem * kgem)1597 void kgem_gen3_finish_state(struct kgem *kgem)
1598 {
1599 memset(&state, 0, sizeof(state));
1600 }
1601