1 /*
2 * Copyright © 2009-2011 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
20 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
21 * IN THE SOFTWARE.
22 */
23
24 #ifdef HAVE_CONFIG_H
25 #include "config.h"
26 #endif
27
28 #include <assert.h>
29 #include <stdint.h>
30 #include <stdlib.h>
31 #include <stdio.h>
32 #include <stdbool.h>
33 #include <stdarg.h>
34 #include <string.h>
35
36 #include "libdrm_macros.h"
37 #include "xf86drm.h"
38 #include "intel_chipset.h"
39 #include "intel_bufmgr.h"
40
41 /* The compiler throws ~90 warnings. Do not spam the build, until we fix them. */
42 #pragma GCC diagnostic ignored "-Wmissing-field-initializers"
43
44 /* Struct for tracking drm_intel_decode state. */
45 struct drm_intel_decode {
46 /** stdio file where the output should land. Defaults to stdout. */
47 FILE *out;
48
49 /** PCI device ID. */
50 uint32_t devid;
51
52 /**
53 * Shorthand device identifier: 3 is 915, 4 is 965, 5 is
54 * Ironlake, etc.
55 */
56 int gen;
57
58 /** GPU address of the start of the current packet. */
59 uint32_t hw_offset;
60 /** CPU virtual address of the start of the current packet. */
61 uint32_t *data;
62 /** DWORDs of remaining batchbuffer data starting from the packet. */
63 uint32_t count;
64
65 /** GPU address of the start of the batchbuffer data. */
66 uint32_t base_hw_offset;
67 /** CPU Virtual address of the start of the batchbuffer data. */
68 uint32_t *base_data;
69 /** Number of DWORDs of batchbuffer data. */
70 uint32_t base_count;
71
72 /** @{
73 * GPU head and tail pointers, which will be noted in the dump, or ~0.
74 */
75 uint32_t head, tail;
76 /** @} */
77
78 /**
79 * Whether to dump the dwords after MI_BATCHBUFFER_END.
80 *
81 * This sometimes provides clues in corrupted batchbuffers,
82 * and is used by the intel-gpu-tools.
83 */
84 bool dump_past_end;
85
86 bool overflowed;
87 };
88
89 static FILE *out;
90 static uint32_t saved_s2 = 0, saved_s4 = 0;
91 static char saved_s2_set = 0, saved_s4_set = 0;
92 static uint32_t head_offset = 0xffffffff; /* undefined */
93 static uint32_t tail_offset = 0xffffffff; /* undefined */
94
95 #ifndef ARRAY_SIZE
96 #define ARRAY_SIZE(A) (sizeof(A)/sizeof(A[0]))
97 #endif
98
99 #define BUFFER_FAIL(_count, _len, _name) do { \
100 fprintf(out, "Buffer size too small in %s (%d < %d)\n", \
101 (_name), (_count), (_len)); \
102 return _count; \
103 } while (0)
104
int_as_float(uint32_t intval)105 static float int_as_float(uint32_t intval)
106 {
107 union intfloat {
108 uint32_t i;
109 float f;
110 } uval;
111
112 uval.i = intval;
113 return uval.f;
114 }
115
116 static void DRM_PRINTFLIKE(3, 4)
instr_out(struct drm_intel_decode * ctx,unsigned int index,const char * fmt,...)117 instr_out(struct drm_intel_decode *ctx, unsigned int index,
118 const char *fmt, ...)
119 {
120 va_list va;
121 const char *parseinfo;
122 uint32_t offset = ctx->hw_offset + index * 4;
123
124 if (index > ctx->count) {
125 if (!ctx->overflowed) {
126 fprintf(out, "ERROR: Decode attempted to continue beyond end of batchbuffer\n");
127 ctx->overflowed = true;
128 }
129 return;
130 }
131
132 if (offset == head_offset)
133 parseinfo = "HEAD";
134 else if (offset == tail_offset)
135 parseinfo = "TAIL";
136 else
137 parseinfo = " ";
138
139 fprintf(out, "0x%08x: %s 0x%08x: %s", offset, parseinfo,
140 ctx->data[index], index == 0 ? "" : " ");
141 va_start(va, fmt);
142 vfprintf(out, fmt, va);
143 va_end(va);
144 }
145
146 static int
decode_MI_SET_CONTEXT(struct drm_intel_decode * ctx)147 decode_MI_SET_CONTEXT(struct drm_intel_decode *ctx)
148 {
149 uint32_t data = ctx->data[1];
150 if (ctx->gen > 7)
151 return 1;
152
153 instr_out(ctx, 0, "MI_SET_CONTEXT\n");
154 instr_out(ctx, 1, "gtt offset = 0x%x%s%s\n",
155 data & ~0xfff,
156 data & (1<<1)? ", Force Restore": "",
157 data & (1<<0)? ", Restore Inhibit": "");
158
159 return 2;
160 }
161
162 static int
decode_MI_WAIT_FOR_EVENT(struct drm_intel_decode * ctx)163 decode_MI_WAIT_FOR_EVENT(struct drm_intel_decode *ctx)
164 {
165 const char *cc_wait;
166 int cc_shift = 0;
167 uint32_t data = ctx->data[0];
168
169 if (ctx->gen <= 5)
170 cc_shift = 9;
171 else
172 cc_shift = 16;
173
174 switch ((data >> cc_shift) & 0x1f) {
175 case 1:
176 cc_wait = ", cc wait 1";
177 break;
178 case 2:
179 cc_wait = ", cc wait 2";
180 break;
181 case 3:
182 cc_wait = ", cc wait 3";
183 break;
184 case 4:
185 cc_wait = ", cc wait 4";
186 break;
187 case 5:
188 cc_wait = ", cc wait 4";
189 break;
190 default:
191 cc_wait = "";
192 break;
193 }
194
195 if (ctx->gen <= 5) {
196 instr_out(ctx, 0, "MI_WAIT_FOR_EVENT%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n",
197 data & (1<<18)? ", pipe B start vblank wait": "",
198 data & (1<<17)? ", pipe A start vblank wait": "",
199 data & (1<<16)? ", overlay flip pending wait": "",
200 data & (1<<14)? ", pipe B hblank wait": "",
201 data & (1<<13)? ", pipe A hblank wait": "",
202 cc_wait,
203 data & (1<<8)? ", plane C pending flip wait": "",
204 data & (1<<7)? ", pipe B vblank wait": "",
205 data & (1<<6)? ", plane B pending flip wait": "",
206 data & (1<<5)? ", pipe B scan line wait": "",
207 data & (1<<4)? ", fbc idle wait": "",
208 data & (1<<3)? ", pipe A vblank wait": "",
209 data & (1<<2)? ", plane A pending flip wait": "",
210 data & (1<<1)? ", plane A scan line wait": "");
211 } else {
212 instr_out(ctx, 0, "MI_WAIT_FOR_EVENT%s%s%s%s%s%s%s%s%s%s%s%s\n",
213 data & (1<<20)? ", sprite C pending flip wait": "", /* ivb */
214 cc_wait,
215 data & (1<<13)? ", pipe B hblank wait": "",
216 data & (1<<11)? ", pipe B vblank wait": "",
217 data & (1<<10)? ", sprite B pending flip wait": "",
218 data & (1<<9)? ", plane B pending flip wait": "",
219 data & (1<<8)? ", plane B scan line wait": "",
220 data & (1<<5)? ", pipe A hblank wait": "",
221 data & (1<<3)? ", pipe A vblank wait": "",
222 data & (1<<2)? ", sprite A pending flip wait": "",
223 data & (1<<1)? ", plane A pending flip wait": "",
224 data & (1<<0)? ", plane A scan line wait": "");
225 }
226
227 return 1;
228 }
229
230 static int
decode_mi(struct drm_intel_decode * ctx)231 decode_mi(struct drm_intel_decode *ctx)
232 {
233 unsigned int opcode, len = -1;
234 const char *post_sync_op = "";
235 uint32_t *data = ctx->data;
236
237 struct {
238 uint32_t opcode;
239 int len_mask;
240 unsigned int min_len;
241 unsigned int max_len;
242 const char *name;
243 int (*func)(struct drm_intel_decode *ctx);
244 } opcodes_mi[] = {
245 { 0x08, 0, 1, 1, "MI_ARB_ON_OFF" },
246 { 0x0a, 0, 1, 1, "MI_BATCH_BUFFER_END" },
247 { 0x30, 0x3f, 3, 3, "MI_BATCH_BUFFER" },
248 { 0x31, 0x3f, 2, 2, "MI_BATCH_BUFFER_START" },
249 { 0x14, 0x3f, 3, 3, "MI_DISPLAY_BUFFER_INFO" },
250 { 0x04, 0, 1, 1, "MI_FLUSH" },
251 { 0x22, 0x1f, 3, 3, "MI_LOAD_REGISTER_IMM" },
252 { 0x13, 0x3f, 2, 2, "MI_LOAD_SCAN_LINES_EXCL" },
253 { 0x12, 0x3f, 2, 2, "MI_LOAD_SCAN_LINES_INCL" },
254 { 0x00, 0, 1, 1, "MI_NOOP" },
255 { 0x11, 0x3f, 2, 2, "MI_OVERLAY_FLIP" },
256 { 0x07, 0, 1, 1, "MI_REPORT_HEAD" },
257 { 0x18, 0x3f, 2, 2, "MI_SET_CONTEXT", decode_MI_SET_CONTEXT },
258 { 0x20, 0x3f, 3, 4, "MI_STORE_DATA_IMM" },
259 { 0x21, 0x3f, 3, 4, "MI_STORE_DATA_INDEX" },
260 { 0x24, 0x3f, 3, 3, "MI_STORE_REGISTER_MEM" },
261 { 0x02, 0, 1, 1, "MI_USER_INTERRUPT" },
262 { 0x03, 0, 1, 1, "MI_WAIT_FOR_EVENT", decode_MI_WAIT_FOR_EVENT },
263 { 0x16, 0x7f, 3, 3, "MI_SEMAPHORE_MBOX" },
264 { 0x26, 0x1f, 3, 4, "MI_FLUSH_DW" },
265 { 0x28, 0x3f, 3, 3, "MI_REPORT_PERF_COUNT" },
266 { 0x29, 0xff, 3, 3, "MI_LOAD_REGISTER_MEM" },
267 { 0x0b, 0, 1, 1, "MI_SUSPEND_FLUSH"},
268 }, *opcode_mi = NULL;
269
270 /* check instruction length */
271 for (opcode = 0; opcode < sizeof(opcodes_mi) / sizeof(opcodes_mi[0]);
272 opcode++) {
273 if ((data[0] & 0x1f800000) >> 23 == opcodes_mi[opcode].opcode) {
274 len = 1;
275 if (opcodes_mi[opcode].max_len > 1) {
276 len =
277 (data[0] & opcodes_mi[opcode].len_mask) + 2;
278 if (len < opcodes_mi[opcode].min_len
279 || len > opcodes_mi[opcode].max_len) {
280 fprintf(out,
281 "Bad length (%d) in %s, [%d, %d]\n",
282 len, opcodes_mi[opcode].name,
283 opcodes_mi[opcode].min_len,
284 opcodes_mi[opcode].max_len);
285 }
286 }
287 opcode_mi = &opcodes_mi[opcode];
288 break;
289 }
290 }
291
292 if (opcode_mi && opcode_mi->func)
293 return opcode_mi->func(ctx);
294
295 switch ((data[0] & 0x1f800000) >> 23) {
296 case 0x0a:
297 instr_out(ctx, 0, "MI_BATCH_BUFFER_END\n");
298 return -1;
299 case 0x16:
300 instr_out(ctx, 0, "MI_SEMAPHORE_MBOX%s%s%s%s %u\n",
301 data[0] & (1 << 22) ? " global gtt," : "",
302 data[0] & (1 << 21) ? " update semaphore," : "",
303 data[0] & (1 << 20) ? " compare semaphore," : "",
304 data[0] & (1 << 18) ? " use compare reg" : "",
305 (data[0] & (0x3 << 16)) >> 16);
306 instr_out(ctx, 1, "value\n");
307 instr_out(ctx, 2, "address\n");
308 return len;
309 case 0x21:
310 instr_out(ctx, 0, "MI_STORE_DATA_INDEX%s\n",
311 data[0] & (1 << 21) ? " use per-process HWS," : "");
312 instr_out(ctx, 1, "index\n");
313 instr_out(ctx, 2, "dword\n");
314 if (len == 4)
315 instr_out(ctx, 3, "upper dword\n");
316 return len;
317 case 0x00:
318 if (data[0] & (1 << 22))
319 instr_out(ctx, 0,
320 "MI_NOOP write NOPID reg, val=0x%x\n",
321 data[0] & ((1 << 22) - 1));
322 else
323 instr_out(ctx, 0, "MI_NOOP\n");
324 return len;
325 case 0x26:
326 switch (data[0] & (0x3 << 14)) {
327 case (0 << 14):
328 post_sync_op = "no write";
329 break;
330 case (1 << 14):
331 post_sync_op = "write data";
332 break;
333 case (2 << 14):
334 post_sync_op = "reserved";
335 break;
336 case (3 << 14):
337 post_sync_op = "write TIMESTAMP";
338 break;
339 }
340 instr_out(ctx, 0,
341 "MI_FLUSH_DW%s%s%s%s post_sync_op='%s' %s%s\n",
342 data[0] & (1 << 22) ?
343 " enable protected mem (BCS-only)," : "",
344 data[0] & (1 << 21) ? " store in hws," : "",
345 data[0] & (1 << 18) ? " invalidate tlb," : "",
346 data[0] & (1 << 17) ? " flush gfdt," : "",
347 post_sync_op,
348 data[0] & (1 << 8) ? " enable notify interrupt," : "",
349 data[0] & (1 << 7) ?
350 " invalidate video state (BCS-only)," : "");
351 if (data[0] & (1 << 21))
352 instr_out(ctx, 1, "hws index\n");
353 else
354 instr_out(ctx, 1, "address\n");
355 instr_out(ctx, 2, "dword\n");
356 if (len == 4)
357 instr_out(ctx, 3, "upper dword\n");
358 return len;
359 }
360
361 for (opcode = 0; opcode < sizeof(opcodes_mi) / sizeof(opcodes_mi[0]);
362 opcode++) {
363 if ((data[0] & 0x1f800000) >> 23 == opcodes_mi[opcode].opcode) {
364 unsigned int i;
365
366 instr_out(ctx, 0, "%s\n",
367 opcodes_mi[opcode].name);
368 for (i = 1; i < len; i++) {
369 instr_out(ctx, i, "dword %d\n", i);
370 }
371
372 return len;
373 }
374 }
375
376 instr_out(ctx, 0, "MI UNKNOWN\n");
377 return 1;
378 }
379
380 static void
decode_2d_br00(struct drm_intel_decode * ctx,const char * cmd)381 decode_2d_br00(struct drm_intel_decode *ctx, const char *cmd)
382 {
383 instr_out(ctx, 0,
384 "%s (rgb %sabled, alpha %sabled, src tile %d, dst tile %d)\n",
385 cmd,
386 (ctx->data[0] & (1 << 20)) ? "en" : "dis",
387 (ctx->data[0] & (1 << 21)) ? "en" : "dis",
388 (ctx->data[0] >> 15) & 1,
389 (ctx->data[0] >> 11) & 1);
390 }
391
392 static void
decode_2d_br01(struct drm_intel_decode * ctx)393 decode_2d_br01(struct drm_intel_decode *ctx)
394 {
395 const char *format;
396 switch ((ctx->data[1] >> 24) & 0x3) {
397 case 0:
398 format = "8";
399 break;
400 case 1:
401 format = "565";
402 break;
403 case 2:
404 format = "1555";
405 break;
406 case 3:
407 format = "8888";
408 break;
409 }
410
411 instr_out(ctx, 1,
412 "format %s, pitch %d, rop 0x%02x, "
413 "clipping %sabled, %s%s \n",
414 format,
415 (short)(ctx->data[1] & 0xffff),
416 (ctx->data[1] >> 16) & 0xff,
417 ctx->data[1] & (1 << 30) ? "en" : "dis",
418 ctx->data[1] & (1 << 31) ? "solid pattern enabled, " : "",
419 ctx->data[1] & (1 << 31) ?
420 "mono pattern transparency enabled, " : "");
421
422 }
423
424 static int
decode_2d(struct drm_intel_decode * ctx)425 decode_2d(struct drm_intel_decode *ctx)
426 {
427 unsigned int opcode, len;
428 uint32_t *data = ctx->data;
429
430 struct {
431 uint32_t opcode;
432 unsigned int min_len;
433 unsigned int max_len;
434 const char *name;
435 } opcodes_2d[] = {
436 { 0x40, 5, 5, "COLOR_BLT" },
437 { 0x43, 6, 6, "SRC_COPY_BLT" },
438 { 0x01, 8, 8, "XY_SETUP_BLT" },
439 { 0x11, 9, 9, "XY_SETUP_MONO_PATTERN_SL_BLT" },
440 { 0x03, 3, 3, "XY_SETUP_CLIP_BLT" },
441 { 0x24, 2, 2, "XY_PIXEL_BLT" },
442 { 0x25, 3, 3, "XY_SCANLINES_BLT" },
443 { 0x26, 4, 4, "Y_TEXT_BLT" },
444 { 0x31, 5, 134, "XY_TEXT_IMMEDIATE_BLT" },
445 { 0x50, 6, 6, "XY_COLOR_BLT" },
446 { 0x51, 6, 6, "XY_PAT_BLT" },
447 { 0x76, 8, 8, "XY_PAT_CHROMA_BLT" },
448 { 0x72, 7, 135, "XY_PAT_BLT_IMMEDIATE" },
449 { 0x77, 9, 137, "XY_PAT_CHROMA_BLT_IMMEDIATE" },
450 { 0x52, 9, 9, "XY_MONO_PAT_BLT" },
451 { 0x59, 7, 7, "XY_MONO_PAT_FIXED_BLT" },
452 { 0x53, 8, 8, "XY_SRC_COPY_BLT" },
453 { 0x54, 8, 8, "XY_MONO_SRC_COPY_BLT" },
454 { 0x71, 9, 137, "XY_MONO_SRC_COPY_IMMEDIATE_BLT" },
455 { 0x55, 9, 9, "XY_FULL_BLT" },
456 { 0x55, 9, 137, "XY_FULL_IMMEDIATE_PATTERN_BLT" },
457 { 0x56, 9, 9, "XY_FULL_MONO_SRC_BLT" },
458 { 0x75, 10, 138, "XY_FULL_MONO_SRC_IMMEDIATE_PATTERN_BLT" },
459 { 0x57, 12, 12, "XY_FULL_MONO_PATTERN_BLT" },
460 { 0x58, 12, 12, "XY_FULL_MONO_PATTERN_MONO_SRC_BLT"},
461 };
462
463 switch ((data[0] & 0x1fc00000) >> 22) {
464 case 0x25:
465 instr_out(ctx, 0,
466 "XY_SCANLINES_BLT (pattern seed (%d, %d), dst tile %d)\n",
467 (data[0] >> 12) & 0x8,
468 (data[0] >> 8) & 0x8, (data[0] >> 11) & 1);
469
470 len = (data[0] & 0x000000ff) + 2;
471 if (len != 3)
472 fprintf(out, "Bad count in XY_SCANLINES_BLT\n");
473
474 instr_out(ctx, 1, "dest (%d,%d)\n",
475 data[1] & 0xffff, data[1] >> 16);
476 instr_out(ctx, 2, "dest (%d,%d)\n",
477 data[2] & 0xffff, data[2] >> 16);
478 return len;
479 case 0x01:
480 decode_2d_br00(ctx, "XY_SETUP_BLT");
481
482 len = (data[0] & 0x000000ff) + 2;
483 if (len != 8)
484 fprintf(out, "Bad count in XY_SETUP_BLT\n");
485
486 decode_2d_br01(ctx);
487 instr_out(ctx, 2, "cliprect (%d,%d)\n",
488 data[2] & 0xffff, data[2] >> 16);
489 instr_out(ctx, 3, "cliprect (%d,%d)\n",
490 data[3] & 0xffff, data[3] >> 16);
491 instr_out(ctx, 4, "setup dst offset 0x%08x\n",
492 data[4]);
493 instr_out(ctx, 5, "setup background color\n");
494 instr_out(ctx, 6, "setup foreground color\n");
495 instr_out(ctx, 7, "color pattern offset\n");
496 return len;
497 case 0x03:
498 decode_2d_br00(ctx, "XY_SETUP_CLIP_BLT");
499
500 len = (data[0] & 0x000000ff) + 2;
501 if (len != 3)
502 fprintf(out, "Bad count in XY_SETUP_CLIP_BLT\n");
503
504 instr_out(ctx, 1, "cliprect (%d,%d)\n",
505 data[1] & 0xffff, data[2] >> 16);
506 instr_out(ctx, 2, "cliprect (%d,%d)\n",
507 data[2] & 0xffff, data[3] >> 16);
508 return len;
509 case 0x11:
510 decode_2d_br00(ctx, "XY_SETUP_MONO_PATTERN_SL_BLT");
511
512 len = (data[0] & 0x000000ff) + 2;
513 if (len != 9)
514 fprintf(out,
515 "Bad count in XY_SETUP_MONO_PATTERN_SL_BLT\n");
516
517 decode_2d_br01(ctx);
518 instr_out(ctx, 2, "cliprect (%d,%d)\n",
519 data[2] & 0xffff, data[2] >> 16);
520 instr_out(ctx, 3, "cliprect (%d,%d)\n",
521 data[3] & 0xffff, data[3] >> 16);
522 instr_out(ctx, 4, "setup dst offset 0x%08x\n",
523 data[4]);
524 instr_out(ctx, 5, "setup background color\n");
525 instr_out(ctx, 6, "setup foreground color\n");
526 instr_out(ctx, 7, "mono pattern dw0\n");
527 instr_out(ctx, 8, "mono pattern dw1\n");
528 return len;
529 case 0x50:
530 decode_2d_br00(ctx, "XY_COLOR_BLT");
531
532 len = (data[0] & 0x000000ff) + 2;
533 if (len != 6)
534 fprintf(out, "Bad count in XY_COLOR_BLT\n");
535
536 decode_2d_br01(ctx);
537 instr_out(ctx, 2, "(%d,%d)\n",
538 data[2] & 0xffff, data[2] >> 16);
539 instr_out(ctx, 3, "(%d,%d)\n",
540 data[3] & 0xffff, data[3] >> 16);
541 instr_out(ctx, 4, "offset 0x%08x\n", data[4]);
542 instr_out(ctx, 5, "color\n");
543 return len;
544 case 0x53:
545 decode_2d_br00(ctx, "XY_SRC_COPY_BLT");
546
547 len = (data[0] & 0x000000ff) + 2;
548 if (len != 8)
549 fprintf(out, "Bad count in XY_SRC_COPY_BLT\n");
550
551 decode_2d_br01(ctx);
552 instr_out(ctx, 2, "dst (%d,%d)\n",
553 data[2] & 0xffff, data[2] >> 16);
554 instr_out(ctx, 3, "dst (%d,%d)\n",
555 data[3] & 0xffff, data[3] >> 16);
556 instr_out(ctx, 4, "dst offset 0x%08x\n", data[4]);
557 instr_out(ctx, 5, "src (%d,%d)\n",
558 data[5] & 0xffff, data[5] >> 16);
559 instr_out(ctx, 6, "src pitch %d\n",
560 (short)(data[6] & 0xffff));
561 instr_out(ctx, 7, "src offset 0x%08x\n", data[7]);
562 return len;
563 }
564
565 for (opcode = 0; opcode < sizeof(opcodes_2d) / sizeof(opcodes_2d[0]);
566 opcode++) {
567 if ((data[0] & 0x1fc00000) >> 22 == opcodes_2d[opcode].opcode) {
568 unsigned int i;
569
570 len = 1;
571 instr_out(ctx, 0, "%s\n",
572 opcodes_2d[opcode].name);
573 if (opcodes_2d[opcode].max_len > 1) {
574 len = (data[0] & 0x000000ff) + 2;
575 if (len < opcodes_2d[opcode].min_len ||
576 len > opcodes_2d[opcode].max_len) {
577 fprintf(out, "Bad count in %s\n",
578 opcodes_2d[opcode].name);
579 }
580 }
581
582 for (i = 1; i < len; i++) {
583 instr_out(ctx, i, "dword %d\n", i);
584 }
585
586 return len;
587 }
588 }
589
590 instr_out(ctx, 0, "2D UNKNOWN\n");
591 return 1;
592 }
593
594 static int
decode_3d_1c(struct drm_intel_decode * ctx)595 decode_3d_1c(struct drm_intel_decode *ctx)
596 {
597 uint32_t *data = ctx->data;
598 uint32_t opcode;
599
600 opcode = (data[0] & 0x00f80000) >> 19;
601
602 switch (opcode) {
603 case 0x11:
604 instr_out(ctx, 0,
605 "3DSTATE_DEPTH_SUBRECTANGLE_DISABLE\n");
606 return 1;
607 case 0x10:
608 instr_out(ctx, 0, "3DSTATE_SCISSOR_ENABLE %s\n",
609 data[0] & 1 ? "enabled" : "disabled");
610 return 1;
611 case 0x01:
612 instr_out(ctx, 0, "3DSTATE_MAP_COORD_SET_I830\n");
613 return 1;
614 case 0x0a:
615 instr_out(ctx, 0, "3DSTATE_MAP_CUBE_I830\n");
616 return 1;
617 case 0x05:
618 instr_out(ctx, 0, "3DSTATE_MAP_TEX_STREAM_I830\n");
619 return 1;
620 }
621
622 instr_out(ctx, 0, "3D UNKNOWN: 3d_1c opcode = 0x%x\n",
623 opcode);
624 return 1;
625 }
626
627 /** Sets the string dstname to describe the destination of the PS instruction */
628 static void
i915_get_instruction_dst(uint32_t * data,int i,char * dstname,int do_mask)629 i915_get_instruction_dst(uint32_t *data, int i, char *dstname, int do_mask)
630 {
631 uint32_t a0 = data[i];
632 int dst_nr = (a0 >> 14) & 0xf;
633 char dstmask[8];
634 const char *sat;
635
636 if (do_mask) {
637 if (((a0 >> 10) & 0xf) == 0xf) {
638 dstmask[0] = 0;
639 } else {
640 int dstmask_index = 0;
641
642 dstmask[dstmask_index++] = '.';
643 if (a0 & (1 << 10))
644 dstmask[dstmask_index++] = 'x';
645 if (a0 & (1 << 11))
646 dstmask[dstmask_index++] = 'y';
647 if (a0 & (1 << 12))
648 dstmask[dstmask_index++] = 'z';
649 if (a0 & (1 << 13))
650 dstmask[dstmask_index++] = 'w';
651 dstmask[dstmask_index++] = 0;
652 }
653
654 if (a0 & (1 << 22))
655 sat = ".sat";
656 else
657 sat = "";
658 } else {
659 dstmask[0] = 0;
660 sat = "";
661 }
662
663 switch ((a0 >> 19) & 0x7) {
664 case 0:
665 if (dst_nr > 15)
666 fprintf(out, "bad destination reg R%d\n", dst_nr);
667 sprintf(dstname, "R%d%s%s", dst_nr, dstmask, sat);
668 break;
669 case 4:
670 if (dst_nr > 0)
671 fprintf(out, "bad destination reg oC%d\n", dst_nr);
672 sprintf(dstname, "oC%s%s", dstmask, sat);
673 break;
674 case 5:
675 if (dst_nr > 0)
676 fprintf(out, "bad destination reg oD%d\n", dst_nr);
677 sprintf(dstname, "oD%s%s", dstmask, sat);
678 break;
679 case 6:
680 if (dst_nr > 3)
681 fprintf(out, "bad destination reg U%d\n", dst_nr);
682 sprintf(dstname, "U%d%s%s", dst_nr, dstmask, sat);
683 break;
684 default:
685 sprintf(dstname, "RESERVED");
686 break;
687 }
688 }
689
690 static const char *
i915_get_channel_swizzle(uint32_t select)691 i915_get_channel_swizzle(uint32_t select)
692 {
693 switch (select & 0x7) {
694 case 0:
695 return (select & 8) ? "-x" : "x";
696 case 1:
697 return (select & 8) ? "-y" : "y";
698 case 2:
699 return (select & 8) ? "-z" : "z";
700 case 3:
701 return (select & 8) ? "-w" : "w";
702 case 4:
703 return (select & 8) ? "-0" : "0";
704 case 5:
705 return (select & 8) ? "-1" : "1";
706 default:
707 return (select & 8) ? "-bad" : "bad";
708 }
709 }
710
711 static void
i915_get_instruction_src_name(uint32_t src_type,uint32_t src_nr,char * name)712 i915_get_instruction_src_name(uint32_t src_type, uint32_t src_nr, char *name)
713 {
714 switch (src_type) {
715 case 0:
716 sprintf(name, "R%d", src_nr);
717 if (src_nr > 15)
718 fprintf(out, "bad src reg %s\n", name);
719 break;
720 case 1:
721 if (src_nr < 8)
722 sprintf(name, "T%d", src_nr);
723 else if (src_nr == 8)
724 sprintf(name, "DIFFUSE");
725 else if (src_nr == 9)
726 sprintf(name, "SPECULAR");
727 else if (src_nr == 10)
728 sprintf(name, "FOG");
729 else {
730 fprintf(out, "bad src reg T%d\n", src_nr);
731 sprintf(name, "RESERVED");
732 }
733 break;
734 case 2:
735 sprintf(name, "C%d", src_nr);
736 if (src_nr > 31)
737 fprintf(out, "bad src reg %s\n", name);
738 break;
739 case 4:
740 sprintf(name, "oC");
741 if (src_nr > 0)
742 fprintf(out, "bad src reg oC%d\n", src_nr);
743 break;
744 case 5:
745 sprintf(name, "oD");
746 if (src_nr > 0)
747 fprintf(out, "bad src reg oD%d\n", src_nr);
748 break;
749 case 6:
750 sprintf(name, "U%d", src_nr);
751 if (src_nr > 3)
752 fprintf(out, "bad src reg %s\n", name);
753 break;
754 default:
755 fprintf(out, "bad src reg type %d\n", src_type);
756 sprintf(name, "RESERVED");
757 break;
758 }
759 }
760
i915_get_instruction_src0(uint32_t * data,int i,char * srcname)761 static void i915_get_instruction_src0(uint32_t *data, int i, char *srcname)
762 {
763 uint32_t a0 = data[i];
764 uint32_t a1 = data[i + 1];
765 int src_nr = (a0 >> 2) & 0x1f;
766 const char *swizzle_x = i915_get_channel_swizzle((a1 >> 28) & 0xf);
767 const char *swizzle_y = i915_get_channel_swizzle((a1 >> 24) & 0xf);
768 const char *swizzle_z = i915_get_channel_swizzle((a1 >> 20) & 0xf);
769 const char *swizzle_w = i915_get_channel_swizzle((a1 >> 16) & 0xf);
770 char swizzle[100];
771
772 i915_get_instruction_src_name((a0 >> 7) & 0x7, src_nr, srcname);
773 sprintf(swizzle, ".%s%s%s%s", swizzle_x, swizzle_y, swizzle_z,
774 swizzle_w);
775 if (strcmp(swizzle, ".xyzw") != 0)
776 strcat(srcname, swizzle);
777 }
778
i915_get_instruction_src1(uint32_t * data,int i,char * srcname)779 static void i915_get_instruction_src1(uint32_t *data, int i, char *srcname)
780 {
781 uint32_t a1 = data[i + 1];
782 uint32_t a2 = data[i + 2];
783 int src_nr = (a1 >> 8) & 0x1f;
784 const char *swizzle_x = i915_get_channel_swizzle((a1 >> 4) & 0xf);
785 const char *swizzle_y = i915_get_channel_swizzle((a1 >> 0) & 0xf);
786 const char *swizzle_z = i915_get_channel_swizzle((a2 >> 28) & 0xf);
787 const char *swizzle_w = i915_get_channel_swizzle((a2 >> 24) & 0xf);
788 char swizzle[100];
789
790 i915_get_instruction_src_name((a1 >> 13) & 0x7, src_nr, srcname);
791 sprintf(swizzle, ".%s%s%s%s", swizzle_x, swizzle_y, swizzle_z,
792 swizzle_w);
793 if (strcmp(swizzle, ".xyzw") != 0)
794 strcat(srcname, swizzle);
795 }
796
i915_get_instruction_src2(uint32_t * data,int i,char * srcname)797 static void i915_get_instruction_src2(uint32_t *data, int i, char *srcname)
798 {
799 uint32_t a2 = data[i + 2];
800 int src_nr = (a2 >> 16) & 0x1f;
801 const char *swizzle_x = i915_get_channel_swizzle((a2 >> 12) & 0xf);
802 const char *swizzle_y = i915_get_channel_swizzle((a2 >> 8) & 0xf);
803 const char *swizzle_z = i915_get_channel_swizzle((a2 >> 4) & 0xf);
804 const char *swizzle_w = i915_get_channel_swizzle((a2 >> 0) & 0xf);
805 char swizzle[100];
806
807 i915_get_instruction_src_name((a2 >> 21) & 0x7, src_nr, srcname);
808 sprintf(swizzle, ".%s%s%s%s", swizzle_x, swizzle_y, swizzle_z,
809 swizzle_w);
810 if (strcmp(swizzle, ".xyzw") != 0)
811 strcat(srcname, swizzle);
812 }
813
814 static void
i915_get_instruction_addr(uint32_t src_type,uint32_t src_nr,char * name)815 i915_get_instruction_addr(uint32_t src_type, uint32_t src_nr, char *name)
816 {
817 switch (src_type) {
818 case 0:
819 sprintf(name, "R%d", src_nr);
820 if (src_nr > 15)
821 fprintf(out, "bad src reg %s\n", name);
822 break;
823 case 1:
824 if (src_nr < 8)
825 sprintf(name, "T%d", src_nr);
826 else if (src_nr == 8)
827 sprintf(name, "DIFFUSE");
828 else if (src_nr == 9)
829 sprintf(name, "SPECULAR");
830 else if (src_nr == 10)
831 sprintf(name, "FOG");
832 else {
833 fprintf(out, "bad src reg T%d\n", src_nr);
834 sprintf(name, "RESERVED");
835 }
836 break;
837 case 4:
838 sprintf(name, "oC");
839 if (src_nr > 0)
840 fprintf(out, "bad src reg oC%d\n", src_nr);
841 break;
842 case 5:
843 sprintf(name, "oD");
844 if (src_nr > 0)
845 fprintf(out, "bad src reg oD%d\n", src_nr);
846 break;
847 default:
848 fprintf(out, "bad src reg type %d\n", src_type);
849 sprintf(name, "RESERVED");
850 break;
851 }
852 }
853
854 static void
i915_decode_alu1(struct drm_intel_decode * ctx,int i,char * instr_prefix,const char * op_name)855 i915_decode_alu1(struct drm_intel_decode *ctx,
856 int i, char *instr_prefix, const char *op_name)
857 {
858 char dst[100], src0[100];
859
860 i915_get_instruction_dst(ctx->data, i, dst, 1);
861 i915_get_instruction_src0(ctx->data, i, src0);
862
863 instr_out(ctx, i++, "%s: %s %s, %s\n", instr_prefix,
864 op_name, dst, src0);
865 instr_out(ctx, i++, "%s\n", instr_prefix);
866 instr_out(ctx, i++, "%s\n", instr_prefix);
867 }
868
869 static void
i915_decode_alu2(struct drm_intel_decode * ctx,int i,char * instr_prefix,const char * op_name)870 i915_decode_alu2(struct drm_intel_decode *ctx,
871 int i, char *instr_prefix, const char *op_name)
872 {
873 char dst[100], src0[100], src1[100];
874
875 i915_get_instruction_dst(ctx->data, i, dst, 1);
876 i915_get_instruction_src0(ctx->data, i, src0);
877 i915_get_instruction_src1(ctx->data, i, src1);
878
879 instr_out(ctx, i++, "%s: %s %s, %s, %s\n", instr_prefix,
880 op_name, dst, src0, src1);
881 instr_out(ctx, i++, "%s\n", instr_prefix);
882 instr_out(ctx, i++, "%s\n", instr_prefix);
883 }
884
885 static void
i915_decode_alu3(struct drm_intel_decode * ctx,int i,char * instr_prefix,const char * op_name)886 i915_decode_alu3(struct drm_intel_decode *ctx,
887 int i, char *instr_prefix, const char *op_name)
888 {
889 char dst[100], src0[100], src1[100], src2[100];
890
891 i915_get_instruction_dst(ctx->data, i, dst, 1);
892 i915_get_instruction_src0(ctx->data, i, src0);
893 i915_get_instruction_src1(ctx->data, i, src1);
894 i915_get_instruction_src2(ctx->data, i, src2);
895
896 instr_out(ctx, i++, "%s: %s %s, %s, %s, %s\n", instr_prefix,
897 op_name, dst, src0, src1, src2);
898 instr_out(ctx, i++, "%s\n", instr_prefix);
899 instr_out(ctx, i++, "%s\n", instr_prefix);
900 }
901
902 static void
i915_decode_tex(struct drm_intel_decode * ctx,int i,const char * instr_prefix,const char * tex_name)903 i915_decode_tex(struct drm_intel_decode *ctx, int i,
904 const char *instr_prefix, const char *tex_name)
905 {
906 uint32_t t0 = ctx->data[i];
907 uint32_t t1 = ctx->data[i + 1];
908 char dst_name[100];
909 char addr_name[100];
910 int sampler_nr;
911
912 i915_get_instruction_dst(ctx->data, i, dst_name, 0);
913 i915_get_instruction_addr((t1 >> 24) & 0x7,
914 (t1 >> 17) & 0xf, addr_name);
915 sampler_nr = t0 & 0xf;
916
917 instr_out(ctx, i++, "%s: %s %s, S%d, %s\n", instr_prefix,
918 tex_name, dst_name, sampler_nr, addr_name);
919 instr_out(ctx, i++, "%s\n", instr_prefix);
920 instr_out(ctx, i++, "%s\n", instr_prefix);
921 }
922
923 static void
i915_decode_dcl(struct drm_intel_decode * ctx,int i,char * instr_prefix)924 i915_decode_dcl(struct drm_intel_decode *ctx, int i, char *instr_prefix)
925 {
926 uint32_t d0 = ctx->data[i];
927 const char *sampletype;
928 int dcl_nr = (d0 >> 14) & 0xf;
929 const char *dcl_x = d0 & (1 << 10) ? "x" : "";
930 const char *dcl_y = d0 & (1 << 11) ? "y" : "";
931 const char *dcl_z = d0 & (1 << 12) ? "z" : "";
932 const char *dcl_w = d0 & (1 << 13) ? "w" : "";
933 char dcl_mask[10];
934
935 switch ((d0 >> 19) & 0x3) {
936 case 1:
937 sprintf(dcl_mask, ".%s%s%s%s", dcl_x, dcl_y, dcl_z, dcl_w);
938 if (strcmp(dcl_mask, ".") == 0)
939 fprintf(out, "bad (empty) dcl mask\n");
940
941 if (dcl_nr > 10)
942 fprintf(out, "bad T%d dcl register number\n", dcl_nr);
943 if (dcl_nr < 8) {
944 if (strcmp(dcl_mask, ".x") != 0 &&
945 strcmp(dcl_mask, ".xy") != 0 &&
946 strcmp(dcl_mask, ".xz") != 0 &&
947 strcmp(dcl_mask, ".w") != 0 &&
948 strcmp(dcl_mask, ".xyzw") != 0) {
949 fprintf(out, "bad T%d.%s dcl mask\n", dcl_nr,
950 dcl_mask);
951 }
952 instr_out(ctx, i++, "%s: DCL T%d%s\n",
953 instr_prefix, dcl_nr, dcl_mask);
954 } else {
955 if (strcmp(dcl_mask, ".xz") == 0)
956 fprintf(out, "errataed bad dcl mask %s\n",
957 dcl_mask);
958 else if (strcmp(dcl_mask, ".xw") == 0)
959 fprintf(out, "errataed bad dcl mask %s\n",
960 dcl_mask);
961 else if (strcmp(dcl_mask, ".xzw") == 0)
962 fprintf(out, "errataed bad dcl mask %s\n",
963 dcl_mask);
964
965 if (dcl_nr == 8) {
966 instr_out(ctx, i++,
967 "%s: DCL DIFFUSE%s\n", instr_prefix,
968 dcl_mask);
969 } else if (dcl_nr == 9) {
970 instr_out(ctx, i++,
971 "%s: DCL SPECULAR%s\n", instr_prefix,
972 dcl_mask);
973 } else if (dcl_nr == 10) {
974 instr_out(ctx, i++,
975 "%s: DCL FOG%s\n", instr_prefix,
976 dcl_mask);
977 }
978 }
979 instr_out(ctx, i++, "%s\n", instr_prefix);
980 instr_out(ctx, i++, "%s\n", instr_prefix);
981 break;
982 case 3:
983 switch ((d0 >> 22) & 0x3) {
984 case 0:
985 sampletype = "2D";
986 break;
987 case 1:
988 sampletype = "CUBE";
989 break;
990 case 2:
991 sampletype = "3D";
992 break;
993 default:
994 sampletype = "RESERVED";
995 break;
996 }
997 if (dcl_nr > 15)
998 fprintf(out, "bad S%d dcl register number\n", dcl_nr);
999 instr_out(ctx, i++, "%s: DCL S%d %s\n",
1000 instr_prefix, dcl_nr, sampletype);
1001 instr_out(ctx, i++, "%s\n", instr_prefix);
1002 instr_out(ctx, i++, "%s\n", instr_prefix);
1003 break;
1004 default:
1005 instr_out(ctx, i++, "%s: DCL RESERVED%d\n",
1006 instr_prefix, dcl_nr);
1007 instr_out(ctx, i++, "%s\n", instr_prefix);
1008 instr_out(ctx, i++, "%s\n", instr_prefix);
1009 }
1010 }
1011
1012 static void
i915_decode_instruction(struct drm_intel_decode * ctx,int i,char * instr_prefix)1013 i915_decode_instruction(struct drm_intel_decode *ctx,
1014 int i, char *instr_prefix)
1015 {
1016 switch ((ctx->data[i] >> 24) & 0x1f) {
1017 case 0x0:
1018 instr_out(ctx, i++, "%s: NOP\n", instr_prefix);
1019 instr_out(ctx, i++, "%s\n", instr_prefix);
1020 instr_out(ctx, i++, "%s\n", instr_prefix);
1021 break;
1022 case 0x01:
1023 i915_decode_alu2(ctx, i, instr_prefix, "ADD");
1024 break;
1025 case 0x02:
1026 i915_decode_alu1(ctx, i, instr_prefix, "MOV");
1027 break;
1028 case 0x03:
1029 i915_decode_alu2(ctx, i, instr_prefix, "MUL");
1030 break;
1031 case 0x04:
1032 i915_decode_alu3(ctx, i, instr_prefix, "MAD");
1033 break;
1034 case 0x05:
1035 i915_decode_alu3(ctx, i, instr_prefix, "DP2ADD");
1036 break;
1037 case 0x06:
1038 i915_decode_alu2(ctx, i, instr_prefix, "DP3");
1039 break;
1040 case 0x07:
1041 i915_decode_alu2(ctx, i, instr_prefix, "DP4");
1042 break;
1043 case 0x08:
1044 i915_decode_alu1(ctx, i, instr_prefix, "FRC");
1045 break;
1046 case 0x09:
1047 i915_decode_alu1(ctx, i, instr_prefix, "RCP");
1048 break;
1049 case 0x0a:
1050 i915_decode_alu1(ctx, i, instr_prefix, "RSQ");
1051 break;
1052 case 0x0b:
1053 i915_decode_alu1(ctx, i, instr_prefix, "EXP");
1054 break;
1055 case 0x0c:
1056 i915_decode_alu1(ctx, i, instr_prefix, "LOG");
1057 break;
1058 case 0x0d:
1059 i915_decode_alu2(ctx, i, instr_prefix, "CMP");
1060 break;
1061 case 0x0e:
1062 i915_decode_alu2(ctx, i, instr_prefix, "MIN");
1063 break;
1064 case 0x0f:
1065 i915_decode_alu2(ctx, i, instr_prefix, "MAX");
1066 break;
1067 case 0x10:
1068 i915_decode_alu1(ctx, i, instr_prefix, "FLR");
1069 break;
1070 case 0x11:
1071 i915_decode_alu1(ctx, i, instr_prefix, "MOD");
1072 break;
1073 case 0x12:
1074 i915_decode_alu1(ctx, i, instr_prefix, "TRC");
1075 break;
1076 case 0x13:
1077 i915_decode_alu2(ctx, i, instr_prefix, "SGE");
1078 break;
1079 case 0x14:
1080 i915_decode_alu2(ctx, i, instr_prefix, "SLT");
1081 break;
1082 case 0x15:
1083 i915_decode_tex(ctx, i, instr_prefix, "TEXLD");
1084 break;
1085 case 0x16:
1086 i915_decode_tex(ctx, i, instr_prefix, "TEXLDP");
1087 break;
1088 case 0x17:
1089 i915_decode_tex(ctx, i, instr_prefix, "TEXLDB");
1090 break;
1091 case 0x19:
1092 i915_decode_dcl(ctx, i, instr_prefix);
1093 break;
1094 default:
1095 instr_out(ctx, i++, "%s: unknown\n", instr_prefix);
1096 instr_out(ctx, i++, "%s\n", instr_prefix);
1097 instr_out(ctx, i++, "%s\n", instr_prefix);
1098 break;
1099 }
1100 }
1101
1102 static const char *
decode_compare_func(uint32_t op)1103 decode_compare_func(uint32_t op)
1104 {
1105 switch (op & 0x7) {
1106 case 0:
1107 return "always";
1108 case 1:
1109 return "never";
1110 case 2:
1111 return "less";
1112 case 3:
1113 return "equal";
1114 case 4:
1115 return "lequal";
1116 case 5:
1117 return "greater";
1118 case 6:
1119 return "notequal";
1120 case 7:
1121 return "gequal";
1122 }
1123 return "";
1124 }
1125
1126 static const char *
decode_stencil_op(uint32_t op)1127 decode_stencil_op(uint32_t op)
1128 {
1129 switch (op & 0x7) {
1130 case 0:
1131 return "keep";
1132 case 1:
1133 return "zero";
1134 case 2:
1135 return "replace";
1136 case 3:
1137 return "incr_sat";
1138 case 4:
1139 return "decr_sat";
1140 case 5:
1141 return "greater";
1142 case 6:
1143 return "incr";
1144 case 7:
1145 return "decr";
1146 }
1147 return "";
1148 }
1149
1150 #if 0
1151 static const char *
1152 decode_logic_op(uint32_t op)
1153 {
1154 switch (op & 0xf) {
1155 case 0:
1156 return "clear";
1157 case 1:
1158 return "nor";
1159 case 2:
1160 return "and_inv";
1161 case 3:
1162 return "copy_inv";
1163 case 4:
1164 return "and_rvrse";
1165 case 5:
1166 return "inv";
1167 case 6:
1168 return "xor";
1169 case 7:
1170 return "nand";
1171 case 8:
1172 return "and";
1173 case 9:
1174 return "equiv";
1175 case 10:
1176 return "noop";
1177 case 11:
1178 return "or_inv";
1179 case 12:
1180 return "copy";
1181 case 13:
1182 return "or_rvrse";
1183 case 14:
1184 return "or";
1185 case 15:
1186 return "set";
1187 }
1188 return "";
1189 }
1190 #endif
1191
1192 static const char *
decode_blend_fact(uint32_t op)1193 decode_blend_fact(uint32_t op)
1194 {
1195 switch (op & 0xf) {
1196 case 1:
1197 return "zero";
1198 case 2:
1199 return "one";
1200 case 3:
1201 return "src_colr";
1202 case 4:
1203 return "inv_src_colr";
1204 case 5:
1205 return "src_alpha";
1206 case 6:
1207 return "inv_src_alpha";
1208 case 7:
1209 return "dst_alpha";
1210 case 8:
1211 return "inv_dst_alpha";
1212 case 9:
1213 return "dst_colr";
1214 case 10:
1215 return "inv_dst_colr";
1216 case 11:
1217 return "src_alpha_sat";
1218 case 12:
1219 return "cnst_colr";
1220 case 13:
1221 return "inv_cnst_colr";
1222 case 14:
1223 return "cnst_alpha";
1224 case 15:
1225 return "inv_const_alpha";
1226 }
1227 return "";
1228 }
1229
1230 static const char *
decode_tex_coord_mode(uint32_t mode)1231 decode_tex_coord_mode(uint32_t mode)
1232 {
1233 switch (mode & 0x7) {
1234 case 0:
1235 return "wrap";
1236 case 1:
1237 return "mirror";
1238 case 2:
1239 return "clamp_edge";
1240 case 3:
1241 return "cube";
1242 case 4:
1243 return "clamp_border";
1244 case 5:
1245 return "mirror_once";
1246 }
1247 return "";
1248 }
1249
1250 static const char *
decode_sample_filter(uint32_t mode)1251 decode_sample_filter(uint32_t mode)
1252 {
1253 switch (mode & 0x7) {
1254 case 0:
1255 return "nearest";
1256 case 1:
1257 return "linear";
1258 case 2:
1259 return "anisotropic";
1260 case 3:
1261 return "4x4_1";
1262 case 4:
1263 return "4x4_2";
1264 case 5:
1265 return "4x4_flat";
1266 case 6:
1267 return "6x5_mono";
1268 }
1269 return "";
1270 }
1271
1272 static int
decode_3d_1d(struct drm_intel_decode * ctx)1273 decode_3d_1d(struct drm_intel_decode *ctx)
1274 {
1275 unsigned int len, i, c, idx, word, map, sampler, instr;
1276 const char *format, *zformat, *type;
1277 uint32_t opcode;
1278 uint32_t *data = ctx->data;
1279 uint32_t devid = ctx->devid;
1280
1281 struct {
1282 uint32_t opcode;
1283 int i830_only;
1284 unsigned int min_len;
1285 unsigned int max_len;
1286 const char *name;
1287 } opcodes_3d_1d[] = {
1288 { 0x86, 0, 4, 4, "3DSTATE_CHROMA_KEY" },
1289 { 0x88, 0, 2, 2, "3DSTATE_CONSTANT_BLEND_COLOR" },
1290 { 0x99, 0, 2, 2, "3DSTATE_DEFAULT_DIFFUSE" },
1291 { 0x9a, 0, 2, 2, "3DSTATE_DEFAULT_SPECULAR" },
1292 { 0x98, 0, 2, 2, "3DSTATE_DEFAULT_Z" },
1293 { 0x97, 0, 2, 2, "3DSTATE_DEPTH_OFFSET_SCALE" },
1294 { 0x9d, 0, 65, 65, "3DSTATE_FILTER_COEFFICIENTS_4X4" },
1295 { 0x9e, 0, 4, 4, "3DSTATE_MONO_FILTER" },
1296 { 0x89, 0, 4, 4, "3DSTATE_FOG_MODE" },
1297 { 0x8f, 0, 2, 16, "3DSTATE_MAP_PALLETE_LOAD_32" },
1298 { 0x83, 0, 2, 2, "3DSTATE_SPAN_STIPPLE" },
1299 { 0x8c, 1, 2, 2, "3DSTATE_MAP_COORD_TRANSFORM_I830" },
1300 { 0x8b, 1, 2, 2, "3DSTATE_MAP_VERTEX_TRANSFORM_I830" },
1301 { 0x8d, 1, 3, 3, "3DSTATE_W_STATE_I830" },
1302 { 0x01, 1, 2, 2, "3DSTATE_COLOR_FACTOR_I830" },
1303 { 0x02, 1, 2, 2, "3DSTATE_MAP_COORD_SETBIND_I830"},
1304 }, *opcode_3d_1d;
1305
1306 opcode = (data[0] & 0x00ff0000) >> 16;
1307
1308 switch (opcode) {
1309 case 0x07:
1310 /* This instruction is unusual. A 0 length means just
1311 * 1 DWORD instead of 2. The 0 length is specified in
1312 * one place to be unsupported, but stated to be
1313 * required in another, and 0 length LOAD_INDIRECTs
1314 * appear to cause no harm at least.
1315 */
1316 instr_out(ctx, 0, "3DSTATE_LOAD_INDIRECT\n");
1317 len = (data[0] & 0x000000ff) + 1;
1318 i = 1;
1319 if (data[0] & (0x01 << 8)) {
1320 instr_out(ctx, i++, "SIS.0\n");
1321 instr_out(ctx, i++, "SIS.1\n");
1322 }
1323 if (data[0] & (0x02 << 8)) {
1324 instr_out(ctx, i++, "DIS.0\n");
1325 }
1326 if (data[0] & (0x04 << 8)) {
1327 instr_out(ctx, i++, "SSB.0\n");
1328 instr_out(ctx, i++, "SSB.1\n");
1329 }
1330 if (data[0] & (0x08 << 8)) {
1331 instr_out(ctx, i++, "MSB.0\n");
1332 instr_out(ctx, i++, "MSB.1\n");
1333 }
1334 if (data[0] & (0x10 << 8)) {
1335 instr_out(ctx, i++, "PSP.0\n");
1336 instr_out(ctx, i++, "PSP.1\n");
1337 }
1338 if (data[0] & (0x20 << 8)) {
1339 instr_out(ctx, i++, "PSC.0\n");
1340 instr_out(ctx, i++, "PSC.1\n");
1341 }
1342 if (len != i) {
1343 fprintf(out, "Bad count in 3DSTATE_LOAD_INDIRECT\n");
1344 return len;
1345 }
1346 return len;
1347 case 0x04:
1348 instr_out(ctx, 0,
1349 "3DSTATE_LOAD_STATE_IMMEDIATE_1\n");
1350 len = (data[0] & 0x0000000f) + 2;
1351 i = 1;
1352 for (word = 0; word <= 8; word++) {
1353 if (data[0] & (1 << (4 + word))) {
1354 /* save vertex state for decode */
1355 if (!IS_GEN2(devid)) {
1356 int tex_num;
1357
1358 if (word == 2) {
1359 saved_s2_set = 1;
1360 saved_s2 = data[i];
1361 }
1362 if (word == 4) {
1363 saved_s4_set = 1;
1364 saved_s4 = data[i];
1365 }
1366
1367 switch (word) {
1368 case 0:
1369 instr_out(ctx, i,
1370 "S0: vbo offset: 0x%08x%s\n",
1371 data[i] & (~1),
1372 data[i] & 1 ?
1373 ", auto cache invalidate disabled"
1374 : "");
1375 break;
1376 case 1:
1377 instr_out(ctx, i,
1378 "S1: vertex width: %i, vertex pitch: %i\n",
1379 (data[i] >> 24) &
1380 0x3f,
1381 (data[i] >> 16) &
1382 0x3f);
1383 break;
1384 case 2:
1385 instr_out(ctx, i,
1386 "S2: texcoord formats: ");
1387 for (tex_num = 0;
1388 tex_num < 8; tex_num++) {
1389 switch ((data[i] >>
1390 tex_num *
1391 4) & 0xf) {
1392 case 0:
1393 fprintf(out,
1394 "%i=2D ",
1395 tex_num);
1396 break;
1397 case 1:
1398 fprintf(out,
1399 "%i=3D ",
1400 tex_num);
1401 break;
1402 case 2:
1403 fprintf(out,
1404 "%i=4D ",
1405 tex_num);
1406 break;
1407 case 3:
1408 fprintf(out,
1409 "%i=1D ",
1410 tex_num);
1411 break;
1412 case 4:
1413 fprintf(out,
1414 "%i=2D_16 ",
1415 tex_num);
1416 break;
1417 case 5:
1418 fprintf(out,
1419 "%i=4D_16 ",
1420 tex_num);
1421 break;
1422 case 0xf:
1423 fprintf(out,
1424 "%i=NP ",
1425 tex_num);
1426 break;
1427 }
1428 }
1429 fprintf(out, "\n");
1430
1431 break;
1432 case 3:
1433 instr_out(ctx, i,
1434 "S3: not documented\n");
1435 break;
1436 case 4:
1437 {
1438 const char *cullmode = "";
1439 const char *vfmt_xyzw = "";
1440 switch ((data[i] >> 13)
1441 & 0x3) {
1442 case 0:
1443 cullmode =
1444 "both";
1445 break;
1446 case 1:
1447 cullmode =
1448 "none";
1449 break;
1450 case 2:
1451 cullmode = "cw";
1452 break;
1453 case 3:
1454 cullmode =
1455 "ccw";
1456 break;
1457 }
1458 switch (data[i] &
1459 (7 << 6 | 1 <<
1460 2)) {
1461 case 1 << 6:
1462 vfmt_xyzw =
1463 "XYZ,";
1464 break;
1465 case 2 << 6:
1466 vfmt_xyzw =
1467 "XYZW,";
1468 break;
1469 case 3 << 6:
1470 vfmt_xyzw =
1471 "XY,";
1472 break;
1473 case 4 << 6:
1474 vfmt_xyzw =
1475 "XYW,";
1476 break;
1477 case 1 << 6 | 1 << 2:
1478 vfmt_xyzw =
1479 "XYZF,";
1480 break;
1481 case 2 << 6 | 1 << 2:
1482 vfmt_xyzw =
1483 "XYZWF,";
1484 break;
1485 case 3 << 6 | 1 << 2:
1486 vfmt_xyzw =
1487 "XYF,";
1488 break;
1489 case 4 << 6 | 1 << 2:
1490 vfmt_xyzw =
1491 "XYWF,";
1492 break;
1493 }
1494 instr_out(ctx, i,
1495 "S4: point_width=%i, line_width=%.1f,"
1496 "%s%s%s%s%s cullmode=%s, vfmt=%s%s%s%s%s%s "
1497 "%s%s%s%s%s\n",
1498 (data[i] >>
1499 23) & 0x1ff,
1500 ((data[i] >>
1501 19) & 0xf) /
1502 2.0,
1503 data[i] & (0xf
1504 <<
1505 15)
1506 ?
1507 " flatshade="
1508 : "",
1509 data[i] & (1
1510 <<
1511 18)
1512 ? "Alpha," :
1513 "",
1514 data[i] & (1
1515 <<
1516 17)
1517 ? "Fog," : "",
1518 data[i] & (1
1519 <<
1520 16)
1521 ? "Specular,"
1522 : "",
1523 data[i] & (1
1524 <<
1525 15)
1526 ? "Color," :
1527 "", cullmode,
1528 data[i] & (1
1529 <<
1530 12)
1531 ?
1532 "PointWidth,"
1533 : "",
1534 data[i] & (1
1535 <<
1536 11)
1537 ? "SpecFog," :
1538 "",
1539 data[i] & (1
1540 <<
1541 10)
1542 ? "Color," :
1543 "",
1544 data[i] & (1
1545 <<
1546 9)
1547 ? "DepthOfs,"
1548 : "",
1549 vfmt_xyzw,
1550 data[i] & (1
1551 <<
1552 9)
1553 ? "FogParam,"
1554 : "",
1555 data[i] & (1
1556 <<
1557 5)
1558 ?
1559 "force default diffuse, "
1560 : "",
1561 data[i] & (1
1562 <<
1563 4)
1564 ?
1565 "force default specular, "
1566 : "",
1567 data[i] & (1
1568 <<
1569 3)
1570 ?
1571 "local depth ofs enable, "
1572 : "",
1573 data[i] & (1
1574 <<
1575 1)
1576 ?
1577 "point sprite enable, "
1578 : "",
1579 data[i] & (1
1580 <<
1581 0)
1582 ?
1583 "line AA enable, "
1584 : "");
1585 break;
1586 }
1587 case 5:
1588 {
1589 instr_out(ctx, i,
1590 "S5:%s%s%s%s%s"
1591 "%s%s%s%s stencil_ref=0x%x, stencil_test=%s, "
1592 "stencil_fail=%s, stencil_pass_z_fail=%s, "
1593 "stencil_pass_z_pass=%s, %s%s%s%s\n",
1594 data[i] & (0xf
1595 <<
1596 28)
1597 ?
1598 " write_disable="
1599 : "",
1600 data[i] & (1
1601 <<
1602 31)
1603 ? "Alpha," :
1604 "",
1605 data[i] & (1
1606 <<
1607 30)
1608 ? "Red," : "",
1609 data[i] & (1
1610 <<
1611 29)
1612 ? "Green," :
1613 "",
1614 data[i] & (1
1615 <<
1616 28)
1617 ? "Blue," :
1618 "",
1619 data[i] & (1
1620 <<
1621 27)
1622 ?
1623 " force default point size,"
1624 : "",
1625 data[i] & (1
1626 <<
1627 26)
1628 ?
1629 " last pixel enable,"
1630 : "",
1631 data[i] & (1
1632 <<
1633 25)
1634 ?
1635 " global depth ofs enable,"
1636 : "",
1637 data[i] & (1
1638 <<
1639 24)
1640 ?
1641 " fog enable,"
1642 : "",
1643 (data[i] >>
1644 16) & 0xff,
1645 decode_compare_func
1646 (data[i] >>
1647 13),
1648 decode_stencil_op
1649 (data[i] >>
1650 10),
1651 decode_stencil_op
1652 (data[i] >>
1653 7),
1654 decode_stencil_op
1655 (data[i] >>
1656 4),
1657 data[i] & (1
1658 <<
1659 3)
1660 ?
1661 "stencil write enable, "
1662 : "",
1663 data[i] & (1
1664 <<
1665 2)
1666 ?
1667 "stencil test enable, "
1668 : "",
1669 data[i] & (1
1670 <<
1671 1)
1672 ?
1673 "color dither enable, "
1674 : "",
1675 data[i] & (1
1676 <<
1677 0)
1678 ?
1679 "logicop enable, "
1680 : "");
1681 }
1682 break;
1683 case 6:
1684 instr_out(ctx, i,
1685 "S6: %salpha_test=%s, alpha_ref=0x%x, "
1686 "depth_test=%s, %ssrc_blnd_fct=%s, dst_blnd_fct=%s, "
1687 "%s%stristrip_provoking_vertex=%i\n",
1688 data[i] & (1 << 31) ?
1689 "alpha test enable, "
1690 : "",
1691 decode_compare_func
1692 (data[i] >> 28),
1693 data[i] & (0xff <<
1694 20),
1695 decode_compare_func
1696 (data[i] >> 16),
1697 data[i] & (1 << 15) ?
1698 "cbuf blend enable, "
1699 : "",
1700 decode_blend_fact(data
1701 [i]
1702 >>
1703 8),
1704 decode_blend_fact(data
1705 [i]
1706 >>
1707 4),
1708 data[i] & (1 << 3) ?
1709 "depth write enable, "
1710 : "",
1711 data[i] & (1 << 2) ?
1712 "cbuf write enable, "
1713 : "",
1714 data[i] & (0x3));
1715 break;
1716 case 7:
1717 instr_out(ctx, i,
1718 "S7: depth offset constant: 0x%08x\n",
1719 data[i]);
1720 break;
1721 }
1722 } else {
1723 instr_out(ctx, i,
1724 "S%d: 0x%08x\n", word, data[i]);
1725 }
1726 i++;
1727 }
1728 }
1729 if (len != i) {
1730 fprintf(out,
1731 "Bad count in 3DSTATE_LOAD_STATE_IMMEDIATE_1\n");
1732 }
1733 return len;
1734 case 0x03:
1735 instr_out(ctx, 0,
1736 "3DSTATE_LOAD_STATE_IMMEDIATE_2\n");
1737 len = (data[0] & 0x0000000f) + 2;
1738 i = 1;
1739 for (word = 6; word <= 14; word++) {
1740 if (data[0] & (1 << word)) {
1741 if (word == 6)
1742 instr_out(ctx, i++,
1743 "TBCF\n");
1744 else if (word >= 7 && word <= 10) {
1745 instr_out(ctx, i++,
1746 "TB%dC\n", word - 7);
1747 instr_out(ctx, i++,
1748 "TB%dA\n", word - 7);
1749 } else if (word >= 11 && word <= 14) {
1750 instr_out(ctx, i,
1751 "TM%dS0: offset=0x%08x, %s\n",
1752 word - 11,
1753 data[i] & 0xfffffffe,
1754 data[i] & 1 ? "use fence" :
1755 "");
1756 i++;
1757 instr_out(ctx, i,
1758 "TM%dS1: height=%i, width=%i, %s\n",
1759 word - 11, data[i] >> 21,
1760 (data[i] >> 10) & 0x3ff,
1761 data[i] & 2 ? (data[i] & 1 ?
1762 "y-tiled" :
1763 "x-tiled") :
1764 "");
1765 i++;
1766 instr_out(ctx, i,
1767 "TM%dS2: pitch=%i, \n",
1768 word - 11,
1769 ((data[i] >> 21) + 1) * 4);
1770 i++;
1771 instr_out(ctx, i++,
1772 "TM%dS3\n", word - 11);
1773 instr_out(ctx, i++,
1774 "TM%dS4: dflt color\n",
1775 word - 11);
1776 }
1777 }
1778 }
1779 if (len != i) {
1780 fprintf(out,
1781 "Bad count in 3DSTATE_LOAD_STATE_IMMEDIATE_2\n");
1782 }
1783 return len;
1784 case 0x00:
1785 instr_out(ctx, 0, "3DSTATE_MAP_STATE\n");
1786 len = (data[0] & 0x0000003f) + 2;
1787 instr_out(ctx, 1, "mask\n");
1788
1789 i = 2;
1790 for (map = 0; map <= 15; map++) {
1791 if (data[1] & (1 << map)) {
1792 int width, height, pitch, dword;
1793 const char *tiling;
1794
1795 dword = data[i];
1796 instr_out(ctx, i++,
1797 "map %d MS2 %s%s%s\n", map,
1798 dword & (1 << 31) ?
1799 "untrusted surface, " : "",
1800 dword & (1 << 1) ?
1801 "vertical line stride enable, " : "",
1802 dword & (1 << 0) ?
1803 "vertical ofs enable, " : "");
1804
1805 dword = data[i];
1806 width = ((dword >> 10) & ((1 << 11) - 1)) + 1;
1807 height = ((dword >> 21) & ((1 << 11) - 1)) + 1;
1808
1809 tiling = "none";
1810 if (dword & (1 << 2))
1811 tiling = "fenced";
1812 else if (dword & (1 << 1))
1813 tiling = dword & (1 << 0) ? "Y" : "X";
1814 type = " BAD";
1815 format = "BAD";
1816 switch ((dword >> 7) & 0x7) {
1817 case 1:
1818 type = "8b";
1819 switch ((dword >> 3) & 0xf) {
1820 case 0:
1821 format = "I";
1822 break;
1823 case 1:
1824 format = "L";
1825 break;
1826 case 4:
1827 format = "A";
1828 break;
1829 case 5:
1830 format = " mono";
1831 break;
1832 }
1833 break;
1834 case 2:
1835 type = "16b";
1836 switch ((dword >> 3) & 0xf) {
1837 case 0:
1838 format = " rgb565";
1839 break;
1840 case 1:
1841 format = " argb1555";
1842 break;
1843 case 2:
1844 format = " argb4444";
1845 break;
1846 case 5:
1847 format = " ay88";
1848 break;
1849 case 6:
1850 format = " bump655";
1851 break;
1852 case 7:
1853 format = "I";
1854 break;
1855 case 8:
1856 format = "L";
1857 break;
1858 case 9:
1859 format = "A";
1860 break;
1861 }
1862 break;
1863 case 3:
1864 type = "32b";
1865 switch ((dword >> 3) & 0xf) {
1866 case 0:
1867 format = " argb8888";
1868 break;
1869 case 1:
1870 format = " abgr8888";
1871 break;
1872 case 2:
1873 format = " xrgb8888";
1874 break;
1875 case 3:
1876 format = " xbgr8888";
1877 break;
1878 case 4:
1879 format = " qwvu8888";
1880 break;
1881 case 5:
1882 format = " axvu8888";
1883 break;
1884 case 6:
1885 format = " lxvu8888";
1886 break;
1887 case 7:
1888 format = " xlvu8888";
1889 break;
1890 case 8:
1891 format = " argb2101010";
1892 break;
1893 case 9:
1894 format = " abgr2101010";
1895 break;
1896 case 10:
1897 format = " awvu2101010";
1898 break;
1899 case 11:
1900 format = " gr1616";
1901 break;
1902 case 12:
1903 format = " vu1616";
1904 break;
1905 case 13:
1906 format = " xI824";
1907 break;
1908 case 14:
1909 format = " xA824";
1910 break;
1911 case 15:
1912 format = " xL824";
1913 break;
1914 }
1915 break;
1916 case 5:
1917 type = "422";
1918 switch ((dword >> 3) & 0xf) {
1919 case 0:
1920 format = " yuv_swapy";
1921 break;
1922 case 1:
1923 format = " yuv";
1924 break;
1925 case 2:
1926 format = " yuv_swapuv";
1927 break;
1928 case 3:
1929 format = " yuv_swapuvy";
1930 break;
1931 }
1932 break;
1933 case 6:
1934 type = "compressed";
1935 switch ((dword >> 3) & 0x7) {
1936 case 0:
1937 format = " dxt1";
1938 break;
1939 case 1:
1940 format = " dxt2_3";
1941 break;
1942 case 2:
1943 format = " dxt4_5";
1944 break;
1945 case 3:
1946 format = " fxt1";
1947 break;
1948 case 4:
1949 format = " dxt1_rb";
1950 break;
1951 }
1952 break;
1953 case 7:
1954 type = "4b indexed";
1955 switch ((dword >> 3) & 0xf) {
1956 case 7:
1957 format = " argb8888";
1958 break;
1959 }
1960 break;
1961 }
1962 dword = data[i];
1963 instr_out(ctx, i++,
1964 "map %d MS3 [width=%d, height=%d, format=%s%s, tiling=%s%s]\n",
1965 map, width, height, type, format,
1966 tiling,
1967 dword & (1 << 9) ? " palette select" :
1968 "");
1969
1970 dword = data[i];
1971 pitch =
1972 4 * (((dword >> 21) & ((1 << 11) - 1)) + 1);
1973 instr_out(ctx, i++,
1974 "map %d MS4 [pitch=%d, max_lod=%i, vol_depth=%i, cube_face_ena=%x, %s]\n",
1975 map, pitch, (dword >> 9) & 0x3f,
1976 dword & 0xff, (dword >> 15) & 0x3f,
1977 dword & (1 << 8) ? "miplayout legacy"
1978 : "miplayout right");
1979 }
1980 }
1981 if (len != i) {
1982 fprintf(out, "Bad count in 3DSTATE_MAP_STATE\n");
1983 return len;
1984 }
1985 return len;
1986 case 0x06:
1987 instr_out(ctx, 0,
1988 "3DSTATE_PIXEL_SHADER_CONSTANTS\n");
1989 len = (data[0] & 0x000000ff) + 2;
1990
1991 i = 2;
1992 for (c = 0; c <= 31; c++) {
1993 if (data[1] & (1 << c)) {
1994 instr_out(ctx, i, "C%d.X = %f\n", c,
1995 int_as_float(data[i]));
1996 i++;
1997 instr_out(ctx, i, "C%d.Y = %f\n",
1998 c, int_as_float(data[i]));
1999 i++;
2000 instr_out(ctx, i, "C%d.Z = %f\n",
2001 c, int_as_float(data[i]));
2002 i++;
2003 instr_out(ctx, i, "C%d.W = %f\n",
2004 c, int_as_float(data[i]));
2005 i++;
2006 }
2007 }
2008 if (len != i) {
2009 fprintf(out,
2010 "Bad count in 3DSTATE_PIXEL_SHADER_CONSTANTS\n");
2011 }
2012 return len;
2013 case 0x05:
2014 instr_out(ctx, 0, "3DSTATE_PIXEL_SHADER_PROGRAM\n");
2015 len = (data[0] & 0x000000ff) + 2;
2016 if ((len - 1) % 3 != 0 || len > 370) {
2017 fprintf(out,
2018 "Bad count in 3DSTATE_PIXEL_SHADER_PROGRAM\n");
2019 }
2020 i = 1;
2021 for (instr = 0; instr < (len - 1) / 3; instr++) {
2022 char instr_prefix[10];
2023
2024 sprintf(instr_prefix, "PS%03d", instr);
2025 i915_decode_instruction(ctx, i,
2026 instr_prefix);
2027 i += 3;
2028 }
2029 return len;
2030 case 0x01:
2031 if (IS_GEN2(devid))
2032 break;
2033 instr_out(ctx, 0, "3DSTATE_SAMPLER_STATE\n");
2034 instr_out(ctx, 1, "mask\n");
2035 len = (data[0] & 0x0000003f) + 2;
2036 i = 2;
2037 for (sampler = 0; sampler <= 15; sampler++) {
2038 if (data[1] & (1 << sampler)) {
2039 uint32_t dword;
2040 const char *mip_filter = "";
2041
2042 dword = data[i];
2043 switch ((dword >> 20) & 0x3) {
2044 case 0:
2045 mip_filter = "none";
2046 break;
2047 case 1:
2048 mip_filter = "nearest";
2049 break;
2050 case 3:
2051 mip_filter = "linear";
2052 break;
2053 }
2054 instr_out(ctx, i++,
2055 "sampler %d SS2:%s%s%s "
2056 "base_mip_level=%i, mip_filter=%s, mag_filter=%s, min_filter=%s "
2057 "lod_bias=%.2f,%s max_aniso=%i, shadow_func=%s\n",
2058 sampler,
2059 dword & (1 << 31) ? " reverse gamma,"
2060 : "",
2061 dword & (1 << 30) ? " packed2planar,"
2062 : "",
2063 dword & (1 << 29) ?
2064 " colorspace conversion," : "",
2065 (dword >> 22) & 0x1f, mip_filter,
2066 decode_sample_filter(dword >> 17),
2067 decode_sample_filter(dword >> 14),
2068 ((dword >> 5) & 0x1ff) / (0x10 * 1.0),
2069 dword & (1 << 4) ? " shadow," : "",
2070 dword & (1 << 3) ? 4 : 2,
2071 decode_compare_func(dword));
2072 dword = data[i];
2073 instr_out(ctx, i++,
2074 "sampler %d SS3: min_lod=%.2f,%s "
2075 "tcmode_x=%s, tcmode_y=%s, tcmode_z=%s,%s texmap_idx=%i,%s\n",
2076 sampler,
2077 ((dword >> 24) & 0xff) / (0x10 * 1.0),
2078 dword & (1 << 17) ?
2079 " kill pixel enable," : "",
2080 decode_tex_coord_mode(dword >> 12),
2081 decode_tex_coord_mode(dword >> 9),
2082 decode_tex_coord_mode(dword >> 6),
2083 dword & (1 << 5) ?
2084 " normalized coords," : "",
2085 (dword >> 1) & 0xf,
2086 dword & (1 << 0) ? " deinterlacer," :
2087 "");
2088 dword = data[i];
2089 instr_out(ctx, i++,
2090 "sampler %d SS4: border color\n",
2091 sampler);
2092 }
2093 }
2094 if (len != i) {
2095 fprintf(out, "Bad count in 3DSTATE_SAMPLER_STATE\n");
2096 }
2097 return len;
2098 case 0x85:
2099 len = (data[0] & 0x0000000f) + 2;
2100
2101 if (len != 2)
2102 fprintf(out,
2103 "Bad count in 3DSTATE_DEST_BUFFER_VARIABLES\n");
2104
2105 instr_out(ctx, 0,
2106 "3DSTATE_DEST_BUFFER_VARIABLES\n");
2107
2108 switch ((data[1] >> 8) & 0xf) {
2109 case 0x0:
2110 format = "g8";
2111 break;
2112 case 0x1:
2113 format = "x1r5g5b5";
2114 break;
2115 case 0x2:
2116 format = "r5g6b5";
2117 break;
2118 case 0x3:
2119 format = "a8r8g8b8";
2120 break;
2121 case 0x4:
2122 format = "ycrcb_swapy";
2123 break;
2124 case 0x5:
2125 format = "ycrcb_normal";
2126 break;
2127 case 0x6:
2128 format = "ycrcb_swapuv";
2129 break;
2130 case 0x7:
2131 format = "ycrcb_swapuvy";
2132 break;
2133 case 0x8:
2134 format = "a4r4g4b4";
2135 break;
2136 case 0x9:
2137 format = "a1r5g5b5";
2138 break;
2139 case 0xa:
2140 format = "a2r10g10b10";
2141 break;
2142 default:
2143 format = "BAD";
2144 break;
2145 }
2146 switch ((data[1] >> 2) & 0x3) {
2147 case 0x0:
2148 zformat = "u16";
2149 break;
2150 case 0x1:
2151 zformat = "f16";
2152 break;
2153 case 0x2:
2154 zformat = "u24x8";
2155 break;
2156 default:
2157 zformat = "BAD";
2158 break;
2159 }
2160 instr_out(ctx, 1,
2161 "%s format, %s depth format, early Z %sabled\n",
2162 format, zformat,
2163 (data[1] & (1 << 31)) ? "en" : "dis");
2164 return len;
2165
2166 case 0x8e:
2167 {
2168 const char *name, *tiling;
2169
2170 len = (data[0] & 0x0000000f) + 2;
2171 if (len != 3)
2172 fprintf(out,
2173 "Bad count in 3DSTATE_BUFFER_INFO\n");
2174
2175 switch ((data[1] >> 24) & 0x7) {
2176 case 0x3:
2177 name = "color";
2178 break;
2179 case 0x7:
2180 name = "depth";
2181 break;
2182 default:
2183 name = "unknown";
2184 break;
2185 }
2186
2187 tiling = "none";
2188 if (data[1] & (1 << 23))
2189 tiling = "fenced";
2190 else if (data[1] & (1 << 22))
2191 tiling = data[1] & (1 << 21) ? "Y" : "X";
2192
2193 instr_out(ctx, 0, "3DSTATE_BUFFER_INFO\n");
2194 instr_out(ctx, 1,
2195 "%s, tiling = %s, pitch=%d\n", name, tiling,
2196 data[1] & 0xffff);
2197
2198 instr_out(ctx, 2, "address\n");
2199 return len;
2200 }
2201 case 0x81:
2202 len = (data[0] & 0x0000000f) + 2;
2203
2204 if (len != 3)
2205 fprintf(out,
2206 "Bad count in 3DSTATE_SCISSOR_RECTANGLE\n");
2207
2208 instr_out(ctx, 0, "3DSTATE_SCISSOR_RECTANGLE\n");
2209 instr_out(ctx, 1, "(%d,%d)\n",
2210 data[1] & 0xffff, data[1] >> 16);
2211 instr_out(ctx, 2, "(%d,%d)\n",
2212 data[2] & 0xffff, data[2] >> 16);
2213
2214 return len;
2215 case 0x80:
2216 len = (data[0] & 0x0000000f) + 2;
2217
2218 if (len != 5)
2219 fprintf(out,
2220 "Bad count in 3DSTATE_DRAWING_RECTANGLE\n");
2221
2222 instr_out(ctx, 0, "3DSTATE_DRAWING_RECTANGLE\n");
2223 instr_out(ctx, 1, "%s\n",
2224 data[1] & (1 << 30) ? "depth ofs disabled " : "");
2225 instr_out(ctx, 2, "(%d,%d)\n",
2226 data[2] & 0xffff, data[2] >> 16);
2227 instr_out(ctx, 3, "(%d,%d)\n",
2228 data[3] & 0xffff, data[3] >> 16);
2229 instr_out(ctx, 4, "(%d,%d)\n",
2230 data[4] & 0xffff, data[4] >> 16);
2231
2232 return len;
2233 case 0x9c:
2234 len = (data[0] & 0x0000000f) + 2;
2235
2236 if (len != 7)
2237 fprintf(out, "Bad count in 3DSTATE_CLEAR_PARAMETERS\n");
2238
2239 instr_out(ctx, 0, "3DSTATE_CLEAR_PARAMETERS\n");
2240 instr_out(ctx, 1, "prim_type=%s, clear=%s%s%s\n",
2241 data[1] & (1 << 16) ? "CLEAR_RECT" : "ZONE_INIT",
2242 data[1] & (1 << 2) ? "color," : "",
2243 data[1] & (1 << 1) ? "depth," : "",
2244 data[1] & (1 << 0) ? "stencil," : "");
2245 instr_out(ctx, 2, "clear color\n");
2246 instr_out(ctx, 3, "clear depth/stencil\n");
2247 instr_out(ctx, 4, "color value (rgba8888)\n");
2248 instr_out(ctx, 5, "depth value %f\n",
2249 int_as_float(data[5]));
2250 instr_out(ctx, 6, "clear stencil\n");
2251 return len;
2252 }
2253
2254 for (idx = 0; idx < ARRAY_SIZE(opcodes_3d_1d); idx++) {
2255 opcode_3d_1d = &opcodes_3d_1d[idx];
2256 if (opcode_3d_1d->i830_only && !IS_GEN2(devid))
2257 continue;
2258
2259 if (((data[0] & 0x00ff0000) >> 16) == opcode_3d_1d->opcode) {
2260 len = 1;
2261
2262 instr_out(ctx, 0, "%s\n",
2263 opcode_3d_1d->name);
2264 if (opcode_3d_1d->max_len > 1) {
2265 len = (data[0] & 0x0000ffff) + 2;
2266 if (len < opcode_3d_1d->min_len ||
2267 len > opcode_3d_1d->max_len) {
2268 fprintf(out, "Bad count in %s\n",
2269 opcode_3d_1d->name);
2270 }
2271 }
2272
2273 for (i = 1; i < len; i++) {
2274 instr_out(ctx, i, "dword %d\n", i);
2275 }
2276
2277 return len;
2278 }
2279 }
2280
2281 instr_out(ctx, 0, "3D UNKNOWN: 3d_1d opcode = 0x%x\n",
2282 opcode);
2283 return 1;
2284 }
2285
2286 static int
decode_3d_primitive(struct drm_intel_decode * ctx)2287 decode_3d_primitive(struct drm_intel_decode *ctx)
2288 {
2289 uint32_t *data = ctx->data;
2290 uint32_t count = ctx->count;
2291 char immediate = (data[0] & (1 << 23)) == 0;
2292 unsigned int len, i, j, ret;
2293 const char *primtype;
2294 int original_s2 = saved_s2;
2295 int original_s4 = saved_s4;
2296
2297 switch ((data[0] >> 18) & 0xf) {
2298 case 0x0:
2299 primtype = "TRILIST";
2300 break;
2301 case 0x1:
2302 primtype = "TRISTRIP";
2303 break;
2304 case 0x2:
2305 primtype = "TRISTRIP_REVERSE";
2306 break;
2307 case 0x3:
2308 primtype = "TRIFAN";
2309 break;
2310 case 0x4:
2311 primtype = "POLYGON";
2312 break;
2313 case 0x5:
2314 primtype = "LINELIST";
2315 break;
2316 case 0x6:
2317 primtype = "LINESTRIP";
2318 break;
2319 case 0x7:
2320 primtype = "RECTLIST";
2321 break;
2322 case 0x8:
2323 primtype = "POINTLIST";
2324 break;
2325 case 0x9:
2326 primtype = "DIB";
2327 break;
2328 case 0xa:
2329 primtype = "CLEAR_RECT";
2330 saved_s4 = 3 << 6;
2331 saved_s2 = ~0;
2332 break;
2333 default:
2334 primtype = "unknown";
2335 break;
2336 }
2337
2338 /* XXX: 3DPRIM_DIB not supported */
2339 if (immediate) {
2340 len = (data[0] & 0x0003ffff) + 2;
2341 instr_out(ctx, 0, "3DPRIMITIVE inline %s\n",
2342 primtype);
2343 if (count < len)
2344 BUFFER_FAIL(count, len, "3DPRIMITIVE inline");
2345 if (!saved_s2_set || !saved_s4_set) {
2346 fprintf(out, "unknown vertex format\n");
2347 for (i = 1; i < len; i++) {
2348 instr_out(ctx, i,
2349 " vertex data (%f float)\n",
2350 int_as_float(data[i]));
2351 }
2352 } else {
2353 unsigned int vertex = 0;
2354 for (i = 1; i < len;) {
2355 unsigned int tc;
2356
2357 #define VERTEX_OUT(fmt, ...) do { \
2358 if (i < len) \
2359 instr_out(ctx, i, " V%d."fmt"\n", vertex, __VA_ARGS__); \
2360 else \
2361 fprintf(out, " missing data in V%d\n", vertex); \
2362 i++; \
2363 } while (0)
2364
2365 VERTEX_OUT("X = %f", int_as_float(data[i]));
2366 VERTEX_OUT("Y = %f", int_as_float(data[i]));
2367 switch (saved_s4 >> 6 & 0x7) {
2368 case 0x1:
2369 VERTEX_OUT("Z = %f",
2370 int_as_float(data[i]));
2371 break;
2372 case 0x2:
2373 VERTEX_OUT("Z = %f",
2374 int_as_float(data[i]));
2375 VERTEX_OUT("W = %f",
2376 int_as_float(data[i]));
2377 break;
2378 case 0x3:
2379 break;
2380 case 0x4:
2381 VERTEX_OUT("W = %f",
2382 int_as_float(data[i]));
2383 break;
2384 default:
2385 fprintf(out, "bad S4 position mask\n");
2386 }
2387
2388 if (saved_s4 & (1 << 10)) {
2389 VERTEX_OUT
2390 ("color = (A=0x%02x, R=0x%02x, G=0x%02x, "
2391 "B=0x%02x)", data[i] >> 24,
2392 (data[i] >> 16) & 0xff,
2393 (data[i] >> 8) & 0xff,
2394 data[i] & 0xff);
2395 }
2396 if (saved_s4 & (1 << 11)) {
2397 VERTEX_OUT
2398 ("spec = (A=0x%02x, R=0x%02x, G=0x%02x, "
2399 "B=0x%02x)", data[i] >> 24,
2400 (data[i] >> 16) & 0xff,
2401 (data[i] >> 8) & 0xff,
2402 data[i] & 0xff);
2403 }
2404 if (saved_s4 & (1 << 12))
2405 VERTEX_OUT("width = 0x%08x)", data[i]);
2406
2407 for (tc = 0; tc <= 7; tc++) {
2408 switch ((saved_s2 >> (tc * 4)) & 0xf) {
2409 case 0x0:
2410 VERTEX_OUT("T%d.X = %f", tc,
2411 int_as_float(data
2412 [i]));
2413 VERTEX_OUT("T%d.Y = %f", tc,
2414 int_as_float(data
2415 [i]));
2416 break;
2417 case 0x1:
2418 VERTEX_OUT("T%d.X = %f", tc,
2419 int_as_float(data
2420 [i]));
2421 VERTEX_OUT("T%d.Y = %f", tc,
2422 int_as_float(data
2423 [i]));
2424 VERTEX_OUT("T%d.Z = %f", tc,
2425 int_as_float(data
2426 [i]));
2427 break;
2428 case 0x2:
2429 VERTEX_OUT("T%d.X = %f", tc,
2430 int_as_float(data
2431 [i]));
2432 VERTEX_OUT("T%d.Y = %f", tc,
2433 int_as_float(data
2434 [i]));
2435 VERTEX_OUT("T%d.Z = %f", tc,
2436 int_as_float(data
2437 [i]));
2438 VERTEX_OUT("T%d.W = %f", tc,
2439 int_as_float(data
2440 [i]));
2441 break;
2442 case 0x3:
2443 VERTEX_OUT("T%d.X = %f", tc,
2444 int_as_float(data
2445 [i]));
2446 break;
2447 case 0x4:
2448 VERTEX_OUT
2449 ("T%d.XY = 0x%08x half-float",
2450 tc, data[i]);
2451 break;
2452 case 0x5:
2453 VERTEX_OUT
2454 ("T%d.XY = 0x%08x half-float",
2455 tc, data[i]);
2456 VERTEX_OUT
2457 ("T%d.ZW = 0x%08x half-float",
2458 tc, data[i]);
2459 break;
2460 case 0xf:
2461 break;
2462 default:
2463 fprintf(out,
2464 "bad S2.T%d format\n",
2465 tc);
2466 }
2467 }
2468 vertex++;
2469 }
2470 }
2471
2472 ret = len;
2473 } else {
2474 /* indirect vertices */
2475 len = data[0] & 0x0000ffff; /* index count */
2476 if (data[0] & (1 << 17)) {
2477 /* random vertex access */
2478 if (count < (len + 1) / 2 + 1) {
2479 BUFFER_FAIL(count, (len + 1) / 2 + 1,
2480 "3DPRIMITIVE random indirect");
2481 }
2482 instr_out(ctx, 0,
2483 "3DPRIMITIVE random indirect %s (%d)\n",
2484 primtype, len);
2485 if (len == 0) {
2486 /* vertex indices continue until 0xffff is
2487 * found
2488 */
2489 for (i = 1; i < count; i++) {
2490 if ((data[i] & 0xffff) == 0xffff) {
2491 instr_out(ctx, i,
2492 " indices: (terminator)\n");
2493 ret = i;
2494 goto out;
2495 } else if ((data[i] >> 16) == 0xffff) {
2496 instr_out(ctx, i,
2497 " indices: 0x%04x, (terminator)\n",
2498 data[i] & 0xffff);
2499 ret = i;
2500 goto out;
2501 } else {
2502 instr_out(ctx, i,
2503 " indices: 0x%04x, 0x%04x\n",
2504 data[i] & 0xffff,
2505 data[i] >> 16);
2506 }
2507 }
2508 fprintf(out,
2509 "3DPRIMITIVE: no terminator found in index buffer\n");
2510 ret = count;
2511 goto out;
2512 } else {
2513 /* fixed size vertex index buffer */
2514 for (j = 1, i = 0; i < len; i += 2, j++) {
2515 if (i * 2 == len - 1) {
2516 instr_out(ctx, j,
2517 " indices: 0x%04x\n",
2518 data[j] & 0xffff);
2519 } else {
2520 instr_out(ctx, j,
2521 " indices: 0x%04x, 0x%04x\n",
2522 data[j] & 0xffff,
2523 data[j] >> 16);
2524 }
2525 }
2526 }
2527 ret = (len + 1) / 2 + 1;
2528 goto out;
2529 } else {
2530 /* sequential vertex access */
2531 instr_out(ctx, 0,
2532 "3DPRIMITIVE sequential indirect %s, %d starting from "
2533 "%d\n", primtype, len, data[1] & 0xffff);
2534 instr_out(ctx, 1, " start\n");
2535 ret = 2;
2536 goto out;
2537 }
2538 }
2539
2540 out:
2541 saved_s2 = original_s2;
2542 saved_s4 = original_s4;
2543 return ret;
2544 }
2545
2546 static int
decode_3d(struct drm_intel_decode * ctx)2547 decode_3d(struct drm_intel_decode *ctx)
2548 {
2549 uint32_t opcode;
2550 unsigned int idx;
2551 uint32_t *data = ctx->data;
2552
2553 struct {
2554 uint32_t opcode;
2555 unsigned int min_len;
2556 unsigned int max_len;
2557 const char *name;
2558 } opcodes_3d[] = {
2559 { 0x06, 1, 1, "3DSTATE_ANTI_ALIASING" },
2560 { 0x08, 1, 1, "3DSTATE_BACKFACE_STENCIL_OPS" },
2561 { 0x09, 1, 1, "3DSTATE_BACKFACE_STENCIL_MASKS" },
2562 { 0x16, 1, 1, "3DSTATE_COORD_SET_BINDINGS" },
2563 { 0x15, 1, 1, "3DSTATE_FOG_COLOR" },
2564 { 0x0b, 1, 1, "3DSTATE_INDEPENDENT_ALPHA_BLEND" },
2565 { 0x0d, 1, 1, "3DSTATE_MODES_4" },
2566 { 0x0c, 1, 1, "3DSTATE_MODES_5" },
2567 { 0x07, 1, 1, "3DSTATE_RASTERIZATION_RULES"},
2568 }, *opcode_3d;
2569
2570 opcode = (data[0] & 0x1f000000) >> 24;
2571
2572 switch (opcode) {
2573 case 0x1f:
2574 return decode_3d_primitive(ctx);
2575 case 0x1d:
2576 return decode_3d_1d(ctx);
2577 case 0x1c:
2578 return decode_3d_1c(ctx);
2579 }
2580
2581 for (idx = 0; idx < ARRAY_SIZE(opcodes_3d); idx++) {
2582 opcode_3d = &opcodes_3d[idx];
2583 if (opcode == opcode_3d->opcode) {
2584 unsigned int len = 1, i;
2585
2586 instr_out(ctx, 0, "%s\n", opcode_3d->name);
2587 if (opcode_3d->max_len > 1) {
2588 len = (data[0] & 0xff) + 2;
2589 if (len < opcode_3d->min_len ||
2590 len > opcode_3d->max_len) {
2591 fprintf(out, "Bad count in %s\n",
2592 opcode_3d->name);
2593 }
2594 }
2595
2596 for (i = 1; i < len; i++) {
2597 instr_out(ctx, i, "dword %d\n", i);
2598 }
2599 return len;
2600 }
2601 }
2602
2603 instr_out(ctx, 0, "3D UNKNOWN: 3d opcode = 0x%x\n", opcode);
2604 return 1;
2605 }
2606
get_965_surfacetype(unsigned int surfacetype)2607 static const char *get_965_surfacetype(unsigned int surfacetype)
2608 {
2609 switch (surfacetype) {
2610 case 0:
2611 return "1D";
2612 case 1:
2613 return "2D";
2614 case 2:
2615 return "3D";
2616 case 3:
2617 return "CUBE";
2618 case 4:
2619 return "BUFFER";
2620 case 7:
2621 return "NULL";
2622 default:
2623 return "unknown";
2624 }
2625 }
2626
get_965_depthformat(unsigned int depthformat)2627 static const char *get_965_depthformat(unsigned int depthformat)
2628 {
2629 switch (depthformat) {
2630 case 0:
2631 return "s8_z24float";
2632 case 1:
2633 return "z32float";
2634 case 2:
2635 return "z24s8";
2636 case 5:
2637 return "z16";
2638 default:
2639 return "unknown";
2640 }
2641 }
2642
get_965_element_component(uint32_t data,int component)2643 static const char *get_965_element_component(uint32_t data, int component)
2644 {
2645 uint32_t component_control = (data >> (16 + (3 - component) * 4)) & 0x7;
2646
2647 switch (component_control) {
2648 case 0:
2649 return "nostore";
2650 case 1:
2651 switch (component) {
2652 case 0:
2653 return "X";
2654 case 1:
2655 return "Y";
2656 case 2:
2657 return "Z";
2658 case 3:
2659 return "W";
2660 default:
2661 return "fail";
2662 }
2663 case 2:
2664 return "0.0";
2665 case 3:
2666 return "1.0";
2667 case 4:
2668 return "0x1";
2669 case 5:
2670 return "VID";
2671 default:
2672 return "fail";
2673 }
2674 }
2675
get_965_prim_type(uint32_t primtype)2676 static const char *get_965_prim_type(uint32_t primtype)
2677 {
2678 switch (primtype) {
2679 case 0x01:
2680 return "point list";
2681 case 0x02:
2682 return "line list";
2683 case 0x03:
2684 return "line strip";
2685 case 0x04:
2686 return "tri list";
2687 case 0x05:
2688 return "tri strip";
2689 case 0x06:
2690 return "tri fan";
2691 case 0x07:
2692 return "quad list";
2693 case 0x08:
2694 return "quad strip";
2695 case 0x09:
2696 return "line list adj";
2697 case 0x0a:
2698 return "line strip adj";
2699 case 0x0b:
2700 return "tri list adj";
2701 case 0x0c:
2702 return "tri strip adj";
2703 case 0x0d:
2704 return "tri strip reverse";
2705 case 0x0e:
2706 return "polygon";
2707 case 0x0f:
2708 return "rect list";
2709 case 0x10:
2710 return "line loop";
2711 case 0x11:
2712 return "point list bf";
2713 case 0x12:
2714 return "line strip cont";
2715 case 0x13:
2716 return "line strip bf";
2717 case 0x14:
2718 return "line strip cont bf";
2719 case 0x15:
2720 return "tri fan no stipple";
2721 default:
2722 return "fail";
2723 }
2724 }
2725
2726 static int
i965_decode_urb_fence(struct drm_intel_decode * ctx,int len)2727 i965_decode_urb_fence(struct drm_intel_decode *ctx, int len)
2728 {
2729 uint32_t vs_fence, clip_fence, gs_fence, sf_fence, vfe_fence, cs_fence;
2730 uint32_t *data = ctx->data;
2731
2732 if (len != 3)
2733 fprintf(out, "Bad count in URB_FENCE\n");
2734
2735 vs_fence = data[1] & 0x3ff;
2736 gs_fence = (data[1] >> 10) & 0x3ff;
2737 clip_fence = (data[1] >> 20) & 0x3ff;
2738 sf_fence = data[2] & 0x3ff;
2739 vfe_fence = (data[2] >> 10) & 0x3ff;
2740 cs_fence = (data[2] >> 20) & 0x7ff;
2741
2742 instr_out(ctx, 0, "URB_FENCE: %s%s%s%s%s%s\n",
2743 (data[0] >> 13) & 1 ? "cs " : "",
2744 (data[0] >> 12) & 1 ? "vfe " : "",
2745 (data[0] >> 11) & 1 ? "sf " : "",
2746 (data[0] >> 10) & 1 ? "clip " : "",
2747 (data[0] >> 9) & 1 ? "gs " : "",
2748 (data[0] >> 8) & 1 ? "vs " : "");
2749 instr_out(ctx, 1,
2750 "vs fence: %d, clip_fence: %d, gs_fence: %d\n",
2751 vs_fence, clip_fence, gs_fence);
2752 instr_out(ctx, 2,
2753 "sf fence: %d, vfe_fence: %d, cs_fence: %d\n",
2754 sf_fence, vfe_fence, cs_fence);
2755 if (gs_fence < vs_fence)
2756 fprintf(out, "gs fence < vs fence!\n");
2757 if (clip_fence < gs_fence)
2758 fprintf(out, "clip fence < gs fence!\n");
2759 if (sf_fence < clip_fence)
2760 fprintf(out, "sf fence < clip fence!\n");
2761 if (cs_fence < sf_fence)
2762 fprintf(out, "cs fence < sf fence!\n");
2763
2764 return len;
2765 }
2766
2767 static void
state_base_out(struct drm_intel_decode * ctx,unsigned int index,const char * name)2768 state_base_out(struct drm_intel_decode *ctx, unsigned int index,
2769 const char *name)
2770 {
2771 if (ctx->data[index] & 1) {
2772 instr_out(ctx, index,
2773 "%s state base address 0x%08x\n", name,
2774 ctx->data[index] & ~1);
2775 } else {
2776 instr_out(ctx, index, "%s state base not updated\n",
2777 name);
2778 }
2779 }
2780
2781 static void
state_max_out(struct drm_intel_decode * ctx,unsigned int index,const char * name)2782 state_max_out(struct drm_intel_decode *ctx, unsigned int index,
2783 const char *name)
2784 {
2785 if (ctx->data[index] & 1) {
2786 if (ctx->data[index] == 1) {
2787 instr_out(ctx, index,
2788 "%s state upper bound disabled\n", name);
2789 } else {
2790 instr_out(ctx, index,
2791 "%s state upper bound 0x%08x\n", name,
2792 ctx->data[index] & ~1);
2793 }
2794 } else {
2795 instr_out(ctx, index,
2796 "%s state upper bound not updated\n", name);
2797 }
2798 }
2799
2800 static int
gen7_3DSTATE_VIEWPORT_STATE_POINTERS_CC(struct drm_intel_decode * ctx)2801 gen7_3DSTATE_VIEWPORT_STATE_POINTERS_CC(struct drm_intel_decode *ctx)
2802 {
2803 instr_out(ctx, 0, "3DSTATE_VIEWPORT_STATE_POINTERS_CC\n");
2804 instr_out(ctx, 1, "pointer to CC viewport\n");
2805
2806 return 2;
2807 }
2808
2809 static int
gen7_3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP(struct drm_intel_decode * ctx)2810 gen7_3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP(struct drm_intel_decode *ctx)
2811 {
2812 instr_out(ctx, 0, "3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP\n");
2813 instr_out(ctx, 1, "pointer to SF_CLIP viewport\n");
2814
2815 return 2;
2816 }
2817
2818 static int
gen7_3DSTATE_BLEND_STATE_POINTERS(struct drm_intel_decode * ctx)2819 gen7_3DSTATE_BLEND_STATE_POINTERS(struct drm_intel_decode *ctx)
2820 {
2821 instr_out(ctx, 0, "3DSTATE_BLEND_STATE_POINTERS\n");
2822 instr_out(ctx, 1, "pointer to BLEND_STATE at 0x%08x (%s)\n",
2823 ctx->data[1] & ~1,
2824 (ctx->data[1] & 1) ? "changed" : "unchanged");
2825
2826 return 2;
2827 }
2828
2829 static int
gen7_3DSTATE_DEPTH_STENCIL_STATE_POINTERS(struct drm_intel_decode * ctx)2830 gen7_3DSTATE_DEPTH_STENCIL_STATE_POINTERS(struct drm_intel_decode *ctx)
2831 {
2832 instr_out(ctx, 0, "3DSTATE_DEPTH_STENCIL_STATE_POINTERS\n");
2833 instr_out(ctx, 1,
2834 "pointer to DEPTH_STENCIL_STATE at 0x%08x (%s)\n",
2835 ctx->data[1] & ~1,
2836 (ctx->data[1] & 1) ? "changed" : "unchanged");
2837
2838 return 2;
2839 }
2840
2841 static int
gen7_3DSTATE_HIER_DEPTH_BUFFER(struct drm_intel_decode * ctx)2842 gen7_3DSTATE_HIER_DEPTH_BUFFER(struct drm_intel_decode *ctx)
2843 {
2844 instr_out(ctx, 0, "3DSTATE_HIER_DEPTH_BUFFER\n");
2845 instr_out(ctx, 1, "pitch %db\n",
2846 (ctx->data[1] & 0x1ffff) + 1);
2847 instr_out(ctx, 2, "pointer to HiZ buffer\n");
2848
2849 return 3;
2850 }
2851
2852 static int
gen6_3DSTATE_CC_STATE_POINTERS(struct drm_intel_decode * ctx)2853 gen6_3DSTATE_CC_STATE_POINTERS(struct drm_intel_decode *ctx)
2854 {
2855 instr_out(ctx, 0, "3DSTATE_CC_STATE_POINTERS\n");
2856 instr_out(ctx, 1, "blend change %d\n", ctx->data[1] & 1);
2857 instr_out(ctx, 2, "depth stencil change %d\n",
2858 ctx->data[2] & 1);
2859 instr_out(ctx, 3, "cc change %d\n", ctx->data[3] & 1);
2860
2861 return 4;
2862 }
2863
2864 static int
gen7_3DSTATE_CC_STATE_POINTERS(struct drm_intel_decode * ctx)2865 gen7_3DSTATE_CC_STATE_POINTERS(struct drm_intel_decode *ctx)
2866 {
2867 instr_out(ctx, 0, "3DSTATE_CC_STATE_POINTERS\n");
2868 instr_out(ctx, 1, "pointer to COLOR_CALC_STATE at 0x%08x "
2869 "(%s)\n",
2870 ctx->data[1] & ~1,
2871 (ctx->data[1] & 1) ? "changed" : "unchanged");
2872
2873 return 2;
2874 }
2875
2876 static int
gen7_3DSTATE_URB_unit(struct drm_intel_decode * ctx,const char * unit)2877 gen7_3DSTATE_URB_unit(struct drm_intel_decode *ctx, const char *unit)
2878 {
2879 int start_kb = ((ctx->data[1] >> 25) & 0x3f) * 8;
2880 /* the field is # of 512-bit rows - 1, we print bytes */
2881 int entry_size = (((ctx->data[1] >> 16) & 0x1ff) + 1);
2882 int nr_entries = ctx->data[1] & 0xffff;
2883
2884 instr_out(ctx, 0, "3DSTATE_URB_%s\n", unit);
2885 instr_out(ctx, 1,
2886 "%dKB start, size=%d 64B rows, nr_entries=%d, total size %dB\n",
2887 start_kb, entry_size, nr_entries, nr_entries * 64 * entry_size);
2888
2889 return 2;
2890 }
2891
2892 static int
gen7_3DSTATE_URB_VS(struct drm_intel_decode * ctx)2893 gen7_3DSTATE_URB_VS(struct drm_intel_decode *ctx)
2894 {
2895 return gen7_3DSTATE_URB_unit(ctx, "VS");
2896 }
2897
2898 static int
gen7_3DSTATE_URB_HS(struct drm_intel_decode * ctx)2899 gen7_3DSTATE_URB_HS(struct drm_intel_decode *ctx)
2900 {
2901 return gen7_3DSTATE_URB_unit(ctx, "HS");
2902 }
2903
2904 static int
gen7_3DSTATE_URB_DS(struct drm_intel_decode * ctx)2905 gen7_3DSTATE_URB_DS(struct drm_intel_decode *ctx)
2906 {
2907 return gen7_3DSTATE_URB_unit(ctx, "DS");
2908 }
2909
2910 static int
gen7_3DSTATE_URB_GS(struct drm_intel_decode * ctx)2911 gen7_3DSTATE_URB_GS(struct drm_intel_decode *ctx)
2912 {
2913 return gen7_3DSTATE_URB_unit(ctx, "GS");
2914 }
2915
2916 static int
gen7_3DSTATE_CONSTANT(struct drm_intel_decode * ctx,const char * unit)2917 gen7_3DSTATE_CONSTANT(struct drm_intel_decode *ctx, const char *unit)
2918 {
2919 int rlen[4];
2920
2921 rlen[0] = (ctx->data[1] >> 0) & 0xffff;
2922 rlen[1] = (ctx->data[1] >> 16) & 0xffff;
2923 rlen[2] = (ctx->data[2] >> 0) & 0xffff;
2924 rlen[3] = (ctx->data[2] >> 16) & 0xffff;
2925
2926 instr_out(ctx, 0, "3DSTATE_CONSTANT_%s\n", unit);
2927 instr_out(ctx, 1, "len 0 = %d, len 1 = %d\n", rlen[0], rlen[1]);
2928 instr_out(ctx, 2, "len 2 = %d, len 3 = %d\n", rlen[2], rlen[3]);
2929 instr_out(ctx, 3, "pointer to constbuf 0\n");
2930 instr_out(ctx, 4, "pointer to constbuf 1\n");
2931 instr_out(ctx, 5, "pointer to constbuf 2\n");
2932 instr_out(ctx, 6, "pointer to constbuf 3\n");
2933
2934 return 7;
2935 }
2936
2937 static int
gen7_3DSTATE_CONSTANT_VS(struct drm_intel_decode * ctx)2938 gen7_3DSTATE_CONSTANT_VS(struct drm_intel_decode *ctx)
2939 {
2940 return gen7_3DSTATE_CONSTANT(ctx, "VS");
2941 }
2942
2943 static int
gen7_3DSTATE_CONSTANT_GS(struct drm_intel_decode * ctx)2944 gen7_3DSTATE_CONSTANT_GS(struct drm_intel_decode *ctx)
2945 {
2946 return gen7_3DSTATE_CONSTANT(ctx, "GS");
2947 }
2948
2949 static int
gen7_3DSTATE_CONSTANT_PS(struct drm_intel_decode * ctx)2950 gen7_3DSTATE_CONSTANT_PS(struct drm_intel_decode *ctx)
2951 {
2952 return gen7_3DSTATE_CONSTANT(ctx, "PS");
2953 }
2954
2955 static int
gen7_3DSTATE_CONSTANT_DS(struct drm_intel_decode * ctx)2956 gen7_3DSTATE_CONSTANT_DS(struct drm_intel_decode *ctx)
2957 {
2958 return gen7_3DSTATE_CONSTANT(ctx, "DS");
2959 }
2960
2961 static int
gen7_3DSTATE_CONSTANT_HS(struct drm_intel_decode * ctx)2962 gen7_3DSTATE_CONSTANT_HS(struct drm_intel_decode *ctx)
2963 {
2964 return gen7_3DSTATE_CONSTANT(ctx, "HS");
2965 }
2966
2967
2968 static int
gen6_3DSTATE_WM(struct drm_intel_decode * ctx)2969 gen6_3DSTATE_WM(struct drm_intel_decode *ctx)
2970 {
2971 instr_out(ctx, 0, "3DSTATE_WM\n");
2972 instr_out(ctx, 1, "kernel start pointer 0\n");
2973 instr_out(ctx, 2,
2974 "SPF=%d, VME=%d, Sampler Count %d, "
2975 "Binding table count %d\n",
2976 (ctx->data[2] >> 31) & 1,
2977 (ctx->data[2] >> 30) & 1,
2978 (ctx->data[2] >> 27) & 7,
2979 (ctx->data[2] >> 18) & 0xff);
2980 instr_out(ctx, 3, "scratch offset\n");
2981 instr_out(ctx, 4,
2982 "Depth Clear %d, Depth Resolve %d, HiZ Resolve %d, "
2983 "Dispatch GRF start[0] %d, start[1] %d, start[2] %d\n",
2984 (ctx->data[4] & (1 << 30)) != 0,
2985 (ctx->data[4] & (1 << 28)) != 0,
2986 (ctx->data[4] & (1 << 27)) != 0,
2987 (ctx->data[4] >> 16) & 0x7f,
2988 (ctx->data[4] >> 8) & 0x7f,
2989 (ctx->data[4] & 0x7f));
2990 instr_out(ctx, 5,
2991 "MaxThreads %d, PS KillPixel %d, PS computed Z %d, "
2992 "PS use sourceZ %d, Thread Dispatch %d, PS use sourceW %d, "
2993 "Dispatch32 %d, Dispatch16 %d, Dispatch8 %d\n",
2994 ((ctx->data[5] >> 25) & 0x7f) + 1,
2995 (ctx->data[5] & (1 << 22)) != 0,
2996 (ctx->data[5] & (1 << 21)) != 0,
2997 (ctx->data[5] & (1 << 20)) != 0,
2998 (ctx->data[5] & (1 << 19)) != 0,
2999 (ctx->data[5] & (1 << 8)) != 0,
3000 (ctx->data[5] & (1 << 2)) != 0,
3001 (ctx->data[5] & (1 << 1)) != 0,
3002 (ctx->data[5] & (1 << 0)) != 0);
3003 instr_out(ctx, 6,
3004 "Num SF output %d, Pos XY offset %d, ZW interp mode %d , "
3005 "Barycentric interp mode 0x%x, Point raster rule %d, "
3006 "Multisample mode %d, "
3007 "Multisample Dispatch mode %d\n",
3008 (ctx->data[6] >> 20) & 0x3f,
3009 (ctx->data[6] >> 18) & 3,
3010 (ctx->data[6] >> 16) & 3,
3011 (ctx->data[6] >> 10) & 0x3f,
3012 (ctx->data[6] & (1 << 9)) != 0,
3013 (ctx->data[6] >> 1) & 3,
3014 (ctx->data[6] & 1));
3015 instr_out(ctx, 7, "kernel start pointer 1\n");
3016 instr_out(ctx, 8, "kernel start pointer 2\n");
3017
3018 return 9;
3019 }
3020
3021 static int
gen7_3DSTATE_WM(struct drm_intel_decode * ctx)3022 gen7_3DSTATE_WM(struct drm_intel_decode *ctx)
3023 {
3024 const char *computed_depth = "";
3025 const char *early_depth = "";
3026 const char *zw_interp = "";
3027
3028 switch ((ctx->data[1] >> 23) & 0x3) {
3029 case 0:
3030 computed_depth = "";
3031 break;
3032 case 1:
3033 computed_depth = "computed depth";
3034 break;
3035 case 2:
3036 computed_depth = "computed depth >=";
3037 break;
3038 case 3:
3039 computed_depth = "computed depth <=";
3040 break;
3041 }
3042
3043 switch ((ctx->data[1] >> 21) & 0x3) {
3044 case 0:
3045 early_depth = "";
3046 break;
3047 case 1:
3048 early_depth = ", EDSC_PSEXEC";
3049 break;
3050 case 2:
3051 early_depth = ", EDSC_PREPS";
3052 break;
3053 case 3:
3054 early_depth = ", BAD EDSC";
3055 break;
3056 }
3057
3058 switch ((ctx->data[1] >> 17) & 0x3) {
3059 case 0:
3060 early_depth = "";
3061 break;
3062 case 1:
3063 early_depth = ", BAD ZW interp";
3064 break;
3065 case 2:
3066 early_depth = ", ZW centroid";
3067 break;
3068 case 3:
3069 early_depth = ", ZW sample";
3070 break;
3071 }
3072
3073 instr_out(ctx, 0, "3DSTATE_WM\n");
3074 instr_out(ctx, 1, "(%s%s%s%s%s%s)%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n",
3075 (ctx->data[1] & (1 << 11)) ? "PP " : "",
3076 (ctx->data[1] & (1 << 12)) ? "PC " : "",
3077 (ctx->data[1] & (1 << 13)) ? "PS " : "",
3078 (ctx->data[1] & (1 << 14)) ? "NPP " : "",
3079 (ctx->data[1] & (1 << 15)) ? "NPC " : "",
3080 (ctx->data[1] & (1 << 16)) ? "NPS " : "",
3081 (ctx->data[1] & (1 << 30)) ? ", depth clear" : "",
3082 (ctx->data[1] & (1 << 29)) ? "" : ", disabled",
3083 (ctx->data[1] & (1 << 28)) ? ", depth resolve" : "",
3084 (ctx->data[1] & (1 << 27)) ? ", hiz resolve" : "",
3085 (ctx->data[1] & (1 << 25)) ? ", kill" : "",
3086 computed_depth,
3087 early_depth,
3088 zw_interp,
3089 (ctx->data[1] & (1 << 20)) ? ", source depth" : "",
3090 (ctx->data[1] & (1 << 19)) ? ", source W" : "",
3091 (ctx->data[1] & (1 << 10)) ? ", coverage" : "",
3092 (ctx->data[1] & (1 << 4)) ? ", poly stipple" : "",
3093 (ctx->data[1] & (1 << 3)) ? ", line stipple" : "",
3094 (ctx->data[1] & (1 << 2)) ? ", point UL" : ", point UR"
3095 );
3096 instr_out(ctx, 2, "MS\n");
3097
3098 return 3;
3099 }
3100
3101 static int
gen4_3DPRIMITIVE(struct drm_intel_decode * ctx)3102 gen4_3DPRIMITIVE(struct drm_intel_decode *ctx)
3103 {
3104 instr_out(ctx, 0,
3105 "3DPRIMITIVE: %s %s\n",
3106 get_965_prim_type((ctx->data[0] >> 10) & 0x1f),
3107 (ctx->data[0] & (1 << 15)) ? "random" : "sequential");
3108 instr_out(ctx, 1, "vertex count\n");
3109 instr_out(ctx, 2, "start vertex\n");
3110 instr_out(ctx, 3, "instance count\n");
3111 instr_out(ctx, 4, "start instance\n");
3112 instr_out(ctx, 5, "index bias\n");
3113
3114 return 6;
3115 }
3116
3117 static int
gen7_3DPRIMITIVE(struct drm_intel_decode * ctx)3118 gen7_3DPRIMITIVE(struct drm_intel_decode *ctx)
3119 {
3120 bool indirect = !!(ctx->data[0] & (1 << 10));
3121
3122 instr_out(ctx, 0,
3123 "3DPRIMITIVE: %s%s\n",
3124 indirect ? " indirect" : "",
3125 (ctx->data[0] & (1 << 8)) ? " predicated" : "");
3126 instr_out(ctx, 1, "%s %s\n",
3127 get_965_prim_type(ctx->data[1] & 0x3f),
3128 (ctx->data[1] & (1 << 8)) ? "random" : "sequential");
3129 instr_out(ctx, 2, indirect ? "ignored" : "vertex count\n");
3130 instr_out(ctx, 3, indirect ? "ignored" : "start vertex\n");
3131 instr_out(ctx, 4, indirect ? "ignored" : "instance count\n");
3132 instr_out(ctx, 5, indirect ? "ignored" : "start instance\n");
3133 instr_out(ctx, 6, indirect ? "ignored" : "index bias\n");
3134
3135 return 7;
3136 }
3137
3138 static int
decode_3d_965(struct drm_intel_decode * ctx)3139 decode_3d_965(struct drm_intel_decode *ctx)
3140 {
3141 uint32_t opcode;
3142 unsigned int len;
3143 unsigned int i, j, sba_len;
3144 const char *desc1 = NULL;
3145 uint32_t *data = ctx->data;
3146 uint32_t devid = ctx->devid;
3147
3148 struct {
3149 uint32_t opcode;
3150 uint32_t len_mask;
3151 int unsigned min_len;
3152 int unsigned max_len;
3153 const char *name;
3154 int gen;
3155 int (*func)(struct drm_intel_decode *ctx);
3156 } opcodes_3d[] = {
3157 { 0x6000, 0x00ff, 3, 3, "URB_FENCE" },
3158 { 0x6001, 0xffff, 2, 2, "CS_URB_STATE" },
3159 { 0x6002, 0x00ff, 2, 2, "CONSTANT_BUFFER" },
3160 { 0x6101, 0xffff, 6, 10, "STATE_BASE_ADDRESS" },
3161 { 0x6102, 0xffff, 2, 2, "STATE_SIP" },
3162 { 0x6104, 0xffff, 1, 1, "3DSTATE_PIPELINE_SELECT" },
3163 { 0x680b, 0xffff, 1, 1, "3DSTATE_VF_STATISTICS" },
3164 { 0x6904, 0xffff, 1, 1, "3DSTATE_PIPELINE_SELECT" },
3165 { 0x7800, 0xffff, 7, 7, "3DSTATE_PIPELINED_POINTERS" },
3166 { 0x7801, 0x00ff, 4, 6, "3DSTATE_BINDING_TABLE_POINTERS" },
3167 { 0x7802, 0x00ff, 4, 4, "3DSTATE_SAMPLER_STATE_POINTERS" },
3168 { 0x7805, 0x00ff, 7, 7, "3DSTATE_DEPTH_BUFFER", 7 },
3169 { 0x7805, 0x00ff, 3, 3, "3DSTATE_URB" },
3170 { 0x7804, 0x00ff, 3, 3, "3DSTATE_CLEAR_PARAMS" },
3171 { 0x7806, 0x00ff, 3, 3, "3DSTATE_STENCIL_BUFFER" },
3172 { 0x790f, 0x00ff, 3, 3, "3DSTATE_HIER_DEPTH_BUFFER", 6 },
3173 { 0x7807, 0x00ff, 3, 3, "3DSTATE_HIER_DEPTH_BUFFER", 7, gen7_3DSTATE_HIER_DEPTH_BUFFER },
3174 { 0x7808, 0x00ff, 5, 257, "3DSTATE_VERTEX_BUFFERS" },
3175 { 0x7809, 0x00ff, 3, 256, "3DSTATE_VERTEX_ELEMENTS" },
3176 { 0x780a, 0x00ff, 3, 3, "3DSTATE_INDEX_BUFFER" },
3177 { 0x780b, 0xffff, 1, 1, "3DSTATE_VF_STATISTICS" },
3178 { 0x780d, 0x00ff, 4, 4, "3DSTATE_VIEWPORT_STATE_POINTERS" },
3179 { 0x780e, 0xffff, 4, 4, NULL, 6, gen6_3DSTATE_CC_STATE_POINTERS },
3180 { 0x780e, 0x00ff, 2, 2, NULL, 7, gen7_3DSTATE_CC_STATE_POINTERS },
3181 { 0x780f, 0x00ff, 2, 2, "3DSTATE_SCISSOR_POINTERS" },
3182 { 0x7810, 0x00ff, 6, 6, "3DSTATE_VS" },
3183 { 0x7811, 0x00ff, 7, 7, "3DSTATE_GS" },
3184 { 0x7812, 0x00ff, 4, 4, "3DSTATE_CLIP" },
3185 { 0x7813, 0x00ff, 20, 20, "3DSTATE_SF", 6 },
3186 { 0x7813, 0x00ff, 7, 7, "3DSTATE_SF", 7 },
3187 { 0x7814, 0x00ff, 3, 3, "3DSTATE_WM", 7, gen7_3DSTATE_WM },
3188 { 0x7814, 0x00ff, 9, 9, "3DSTATE_WM", 6, gen6_3DSTATE_WM },
3189 { 0x7815, 0x00ff, 5, 5, "3DSTATE_CONSTANT_VS_STATE", 6 },
3190 { 0x7815, 0x00ff, 7, 7, "3DSTATE_CONSTANT_VS", 7, gen7_3DSTATE_CONSTANT_VS },
3191 { 0x7816, 0x00ff, 5, 5, "3DSTATE_CONSTANT_GS_STATE", 6 },
3192 { 0x7816, 0x00ff, 7, 7, "3DSTATE_CONSTANT_GS", 7, gen7_3DSTATE_CONSTANT_GS },
3193 { 0x7817, 0x00ff, 5, 5, "3DSTATE_CONSTANT_PS_STATE", 6 },
3194 { 0x7817, 0x00ff, 7, 7, "3DSTATE_CONSTANT_PS", 7, gen7_3DSTATE_CONSTANT_PS },
3195 { 0x7818, 0xffff, 2, 2, "3DSTATE_SAMPLE_MASK" },
3196 { 0x7819, 0x00ff, 7, 7, "3DSTATE_CONSTANT_HS", 7, gen7_3DSTATE_CONSTANT_HS },
3197 { 0x781a, 0x00ff, 7, 7, "3DSTATE_CONSTANT_DS", 7, gen7_3DSTATE_CONSTANT_DS },
3198 { 0x781b, 0x00ff, 7, 7, "3DSTATE_HS" },
3199 { 0x781c, 0x00ff, 4, 4, "3DSTATE_TE" },
3200 { 0x781d, 0x00ff, 6, 6, "3DSTATE_DS" },
3201 { 0x781e, 0x00ff, 3, 3, "3DSTATE_STREAMOUT" },
3202 { 0x781f, 0x00ff, 14, 14, "3DSTATE_SBE" },
3203 { 0x7820, 0x00ff, 8, 8, "3DSTATE_PS" },
3204 { 0x7821, 0x00ff, 2, 2, NULL, 7, gen7_3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP },
3205 { 0x7823, 0x00ff, 2, 2, NULL, 7, gen7_3DSTATE_VIEWPORT_STATE_POINTERS_CC },
3206 { 0x7824, 0x00ff, 2, 2, NULL, 7, gen7_3DSTATE_BLEND_STATE_POINTERS },
3207 { 0x7825, 0x00ff, 2, 2, NULL, 7, gen7_3DSTATE_DEPTH_STENCIL_STATE_POINTERS },
3208 { 0x7826, 0x00ff, 2, 2, "3DSTATE_BINDING_TABLE_POINTERS_VS" },
3209 { 0x7827, 0x00ff, 2, 2, "3DSTATE_BINDING_TABLE_POINTERS_HS" },
3210 { 0x7828, 0x00ff, 2, 2, "3DSTATE_BINDING_TABLE_POINTERS_DS" },
3211 { 0x7829, 0x00ff, 2, 2, "3DSTATE_BINDING_TABLE_POINTERS_GS" },
3212 { 0x782a, 0x00ff, 2, 2, "3DSTATE_BINDING_TABLE_POINTERS_PS" },
3213 { 0x782b, 0x00ff, 2, 2, "3DSTATE_SAMPLER_STATE_POINTERS_VS" },
3214 { 0x782c, 0x00ff, 2, 2, "3DSTATE_SAMPLER_STATE_POINTERS_HS" },
3215 { 0x782d, 0x00ff, 2, 2, "3DSTATE_SAMPLER_STATE_POINTERS_DS" },
3216 { 0x782e, 0x00ff, 2, 2, "3DSTATE_SAMPLER_STATE_POINTERS_GS" },
3217 { 0x782f, 0x00ff, 2, 2, "3DSTATE_SAMPLER_STATE_POINTERS_PS" },
3218 { 0x7830, 0x00ff, 2, 2, NULL, 7, gen7_3DSTATE_URB_VS },
3219 { 0x7831, 0x00ff, 2, 2, NULL, 7, gen7_3DSTATE_URB_HS },
3220 { 0x7832, 0x00ff, 2, 2, NULL, 7, gen7_3DSTATE_URB_DS },
3221 { 0x7833, 0x00ff, 2, 2, NULL, 7, gen7_3DSTATE_URB_GS },
3222 { 0x7900, 0xffff, 4, 4, "3DSTATE_DRAWING_RECTANGLE" },
3223 { 0x7901, 0xffff, 5, 5, "3DSTATE_CONSTANT_COLOR" },
3224 { 0x7905, 0xffff, 5, 7, "3DSTATE_DEPTH_BUFFER" },
3225 { 0x7906, 0xffff, 2, 2, "3DSTATE_POLY_STIPPLE_OFFSET" },
3226 { 0x7907, 0xffff, 33, 33, "3DSTATE_POLY_STIPPLE_PATTERN" },
3227 { 0x7908, 0xffff, 3, 3, "3DSTATE_LINE_STIPPLE" },
3228 { 0x7909, 0xffff, 2, 2, "3DSTATE_GLOBAL_DEPTH_OFFSET_CLAMP" },
3229 { 0x7909, 0xffff, 2, 2, "3DSTATE_CLEAR_PARAMS" },
3230 { 0x790a, 0xffff, 3, 3, "3DSTATE_AA_LINE_PARAMETERS" },
3231 { 0x790b, 0xffff, 4, 4, "3DSTATE_GS_SVB_INDEX" },
3232 { 0x790d, 0xffff, 3, 3, "3DSTATE_MULTISAMPLE", 6 },
3233 { 0x790d, 0xffff, 4, 4, "3DSTATE_MULTISAMPLE", 7 },
3234 { 0x7910, 0x00ff, 2, 2, "3DSTATE_CLEAR_PARAMS" },
3235 { 0x7912, 0x00ff, 2, 2, "3DSTATE_PUSH_CONSTANT_ALLOC_VS" },
3236 { 0x7913, 0x00ff, 2, 2, "3DSTATE_PUSH_CONSTANT_ALLOC_HS" },
3237 { 0x7914, 0x00ff, 2, 2, "3DSTATE_PUSH_CONSTANT_ALLOC_DS" },
3238 { 0x7915, 0x00ff, 2, 2, "3DSTATE_PUSH_CONSTANT_ALLOC_GS" },
3239 { 0x7916, 0x00ff, 2, 2, "3DSTATE_PUSH_CONSTANT_ALLOC_PS" },
3240 { 0x7917, 0x00ff, 2, 2+128*2, "3DSTATE_SO_DECL_LIST" },
3241 { 0x7918, 0x00ff, 4, 4, "3DSTATE_SO_BUFFER" },
3242 { 0x7a00, 0x00ff, 4, 6, "PIPE_CONTROL" },
3243 { 0x7b00, 0x00ff, 7, 7, NULL, 7, gen7_3DPRIMITIVE },
3244 { 0x7b00, 0x00ff, 6, 6, NULL, 0, gen4_3DPRIMITIVE },
3245 }, *opcode_3d = NULL;
3246
3247 opcode = (data[0] & 0xffff0000) >> 16;
3248
3249 for (i = 0; i < ARRAY_SIZE(opcodes_3d); i++) {
3250 if (opcode != opcodes_3d[i].opcode)
3251 continue;
3252
3253 /* If it's marked as not our gen, skip. */
3254 if (opcodes_3d[i].gen && opcodes_3d[i].gen != ctx->gen)
3255 continue;
3256
3257 opcode_3d = &opcodes_3d[i];
3258 break;
3259 }
3260
3261 if (opcode_3d) {
3262 if (opcode_3d->max_len == 1)
3263 len = 1;
3264 else
3265 len = (data[0] & opcode_3d->len_mask) + 2;
3266
3267 if (len < opcode_3d->min_len ||
3268 len > opcode_3d->max_len) {
3269 fprintf(out, "Bad length %d in %s, expected %d-%d\n",
3270 len, opcode_3d->name,
3271 opcode_3d->min_len, opcode_3d->max_len);
3272 }
3273 } else {
3274 len = (data[0] & 0x0000ffff) + 2;
3275 }
3276
3277 switch (opcode) {
3278 case 0x6000:
3279 return i965_decode_urb_fence(ctx, len);
3280 case 0x6001:
3281 instr_out(ctx, 0, "CS_URB_STATE\n");
3282 instr_out(ctx, 1,
3283 "entry_size: %d [%d bytes], n_entries: %d\n",
3284 (data[1] >> 4) & 0x1f,
3285 (((data[1] >> 4) & 0x1f) + 1) * 64, data[1] & 0x7);
3286 return len;
3287 case 0x6002:
3288 instr_out(ctx, 0, "CONSTANT_BUFFER: %s\n",
3289 (data[0] >> 8) & 1 ? "valid" : "invalid");
3290 instr_out(ctx, 1,
3291 "offset: 0x%08x, length: %d bytes\n", data[1] & ~0x3f,
3292 ((data[1] & 0x3f) + 1) * 64);
3293 return len;
3294 case 0x6101:
3295 i = 0;
3296 instr_out(ctx, 0, "STATE_BASE_ADDRESS\n");
3297 i++;
3298
3299 if (IS_GEN6(devid) || IS_GEN7(devid))
3300 sba_len = 10;
3301 else if (IS_GEN5(devid))
3302 sba_len = 8;
3303 else
3304 sba_len = 6;
3305 if (len != sba_len)
3306 fprintf(out, "Bad count in STATE_BASE_ADDRESS\n");
3307
3308 state_base_out(ctx, i++, "general");
3309 state_base_out(ctx, i++, "surface");
3310 if (IS_GEN6(devid) || IS_GEN7(devid))
3311 state_base_out(ctx, i++, "dynamic");
3312 state_base_out(ctx, i++, "indirect");
3313 if (IS_GEN5(devid) || IS_GEN6(devid) || IS_GEN7(devid))
3314 state_base_out(ctx, i++, "instruction");
3315
3316 state_max_out(ctx, i++, "general");
3317 if (IS_GEN6(devid) || IS_GEN7(devid))
3318 state_max_out(ctx, i++, "dynamic");
3319 state_max_out(ctx, i++, "indirect");
3320 if (IS_GEN5(devid) || IS_GEN6(devid) || IS_GEN7(devid))
3321 state_max_out(ctx, i++, "instruction");
3322
3323 return len;
3324 case 0x7800:
3325 instr_out(ctx, 0, "3DSTATE_PIPELINED_POINTERS\n");
3326 instr_out(ctx, 1, "VS state\n");
3327 instr_out(ctx, 2, "GS state\n");
3328 instr_out(ctx, 3, "Clip state\n");
3329 instr_out(ctx, 4, "SF state\n");
3330 instr_out(ctx, 5, "WM state\n");
3331 instr_out(ctx, 6, "CC state\n");
3332 return len;
3333 case 0x7801:
3334 if (len != 6 && len != 4)
3335 fprintf(out,
3336 "Bad count in 3DSTATE_BINDING_TABLE_POINTERS\n");
3337 if (len == 6) {
3338 instr_out(ctx, 0,
3339 "3DSTATE_BINDING_TABLE_POINTERS\n");
3340 instr_out(ctx, 1, "VS binding table\n");
3341 instr_out(ctx, 2, "GS binding table\n");
3342 instr_out(ctx, 3, "Clip binding table\n");
3343 instr_out(ctx, 4, "SF binding table\n");
3344 instr_out(ctx, 5, "WM binding table\n");
3345 } else {
3346 instr_out(ctx, 0,
3347 "3DSTATE_BINDING_TABLE_POINTERS: VS mod %d, "
3348 "GS mod %d, PS mod %d\n",
3349 (data[0] & (1 << 8)) != 0,
3350 (data[0] & (1 << 9)) != 0,
3351 (data[0] & (1 << 12)) != 0);
3352 instr_out(ctx, 1, "VS binding table\n");
3353 instr_out(ctx, 2, "GS binding table\n");
3354 instr_out(ctx, 3, "WM binding table\n");
3355 }
3356
3357 return len;
3358 case 0x7802:
3359 instr_out(ctx, 0,
3360 "3DSTATE_SAMPLER_STATE_POINTERS: VS mod %d, "
3361 "GS mod %d, PS mod %d\n", (data[0] & (1 << 8)) != 0,
3362 (data[0] & (1 << 9)) != 0,
3363 (data[0] & (1 << 12)) != 0);
3364 instr_out(ctx, 1, "VS sampler state\n");
3365 instr_out(ctx, 2, "GS sampler state\n");
3366 instr_out(ctx, 3, "WM sampler state\n");
3367 return len;
3368 case 0x7805:
3369 /* Actually 3DSTATE_DEPTH_BUFFER on gen7. */
3370 if (ctx->gen == 7)
3371 break;
3372
3373 instr_out(ctx, 0, "3DSTATE_URB\n");
3374 instr_out(ctx, 1,
3375 "VS entries %d, alloc size %d (1024bit row)\n",
3376 data[1] & 0xffff, ((data[1] >> 16) & 0x07f) + 1);
3377 instr_out(ctx, 2,
3378 "GS entries %d, alloc size %d (1024bit row)\n",
3379 (data[2] >> 8) & 0x3ff, (data[2] & 7) + 1);
3380 return len;
3381
3382 case 0x7808:
3383 if ((len - 1) % 4 != 0)
3384 fprintf(out, "Bad count in 3DSTATE_VERTEX_BUFFERS\n");
3385 instr_out(ctx, 0, "3DSTATE_VERTEX_BUFFERS\n");
3386
3387 for (i = 1; i < len;) {
3388 int idx, access;
3389 if (IS_GEN6(devid)) {
3390 idx = 26;
3391 access = 20;
3392 } else {
3393 idx = 27;
3394 access = 26;
3395 }
3396 instr_out(ctx, i,
3397 "buffer %d: %s, pitch %db\n", data[i] >> idx,
3398 data[i] & (1 << access) ? "random" :
3399 "sequential", data[i] & 0x07ff);
3400 i++;
3401 instr_out(ctx, i++, "buffer address\n");
3402 instr_out(ctx, i++, "max index\n");
3403 instr_out(ctx, i++, "mbz\n");
3404 }
3405 return len;
3406
3407 case 0x7809:
3408 if ((len + 1) % 2 != 0)
3409 fprintf(out, "Bad count in 3DSTATE_VERTEX_ELEMENTS\n");
3410 instr_out(ctx, 0, "3DSTATE_VERTEX_ELEMENTS\n");
3411
3412 for (i = 1; i < len;) {
3413 instr_out(ctx, i,
3414 "buffer %d: %svalid, type 0x%04x, "
3415 "src offset 0x%04x bytes\n",
3416 data[i] >> ((IS_GEN6(devid) || IS_GEN7(devid)) ? 26 : 27),
3417 data[i] & (1 << ((IS_GEN6(devid) || IS_GEN7(devid)) ? 25 : 26)) ?
3418 "" : "in", (data[i] >> 16) & 0x1ff,
3419 data[i] & 0x07ff);
3420 i++;
3421 instr_out(ctx, i, "(%s, %s, %s, %s), "
3422 "dst offset 0x%02x bytes\n",
3423 get_965_element_component(data[i], 0),
3424 get_965_element_component(data[i], 1),
3425 get_965_element_component(data[i], 2),
3426 get_965_element_component(data[i], 3),
3427 (data[i] & 0xff) * 4);
3428 i++;
3429 }
3430 return len;
3431
3432 case 0x780d:
3433 instr_out(ctx, 0,
3434 "3DSTATE_VIEWPORT_STATE_POINTERS\n");
3435 instr_out(ctx, 1, "clip\n");
3436 instr_out(ctx, 2, "sf\n");
3437 instr_out(ctx, 3, "cc\n");
3438 return len;
3439
3440 case 0x780a:
3441 instr_out(ctx, 0, "3DSTATE_INDEX_BUFFER\n");
3442 instr_out(ctx, 1, "beginning buffer address\n");
3443 instr_out(ctx, 2, "ending buffer address\n");
3444 return len;
3445
3446 case 0x780f:
3447 instr_out(ctx, 0, "3DSTATE_SCISSOR_POINTERS\n");
3448 instr_out(ctx, 1, "scissor rect offset\n");
3449 return len;
3450
3451 case 0x7810:
3452 instr_out(ctx, 0, "3DSTATE_VS\n");
3453 instr_out(ctx, 1, "kernel pointer\n");
3454 instr_out(ctx, 2,
3455 "SPF=%d, VME=%d, Sampler Count %d, "
3456 "Binding table count %d\n", (data[2] >> 31) & 1,
3457 (data[2] >> 30) & 1, (data[2] >> 27) & 7,
3458 (data[2] >> 18) & 0xff);
3459 instr_out(ctx, 3, "scratch offset\n");
3460 instr_out(ctx, 4,
3461 "Dispatch GRF start %d, VUE read length %d, "
3462 "VUE read offset %d\n", (data[4] >> 20) & 0x1f,
3463 (data[4] >> 11) & 0x3f, (data[4] >> 4) & 0x3f);
3464 instr_out(ctx, 5,
3465 "Max Threads %d, Vertex Cache %sable, "
3466 "VS func %sable\n", ((data[5] >> 25) & 0x7f) + 1,
3467 (data[5] & (1 << 1)) != 0 ? "dis" : "en",
3468 (data[5] & 1) != 0 ? "en" : "dis");
3469 return len;
3470
3471 case 0x7811:
3472 instr_out(ctx, 0, "3DSTATE_GS\n");
3473 instr_out(ctx, 1, "kernel pointer\n");
3474 instr_out(ctx, 2,
3475 "SPF=%d, VME=%d, Sampler Count %d, "
3476 "Binding table count %d\n", (data[2] >> 31) & 1,
3477 (data[2] >> 30) & 1, (data[2] >> 27) & 7,
3478 (data[2] >> 18) & 0xff);
3479 instr_out(ctx, 3, "scratch offset\n");
3480 instr_out(ctx, 4,
3481 "Dispatch GRF start %d, VUE read length %d, "
3482 "VUE read offset %d\n", (data[4] & 0xf),
3483 (data[4] >> 11) & 0x3f, (data[4] >> 4) & 0x3f);
3484 instr_out(ctx, 5,
3485 "Max Threads %d, Rendering %sable\n",
3486 ((data[5] >> 25) & 0x7f) + 1,
3487 (data[5] & (1 << 8)) != 0 ? "en" : "dis");
3488 instr_out(ctx, 6,
3489 "Reorder %sable, Discard Adjaceny %sable, "
3490 "GS %sable\n",
3491 (data[6] & (1 << 30)) != 0 ? "en" : "dis",
3492 (data[6] & (1 << 29)) != 0 ? "en" : "dis",
3493 (data[6] & (1 << 15)) != 0 ? "en" : "dis");
3494 return len;
3495
3496 case 0x7812:
3497 instr_out(ctx, 0, "3DSTATE_CLIP\n");
3498 instr_out(ctx, 1,
3499 "UserClip distance cull test mask 0x%x\n",
3500 data[1] & 0xff);
3501 instr_out(ctx, 2,
3502 "Clip %sable, API mode %s, Viewport XY test %sable, "
3503 "Viewport Z test %sable, Guardband test %sable, Clip mode %d, "
3504 "Perspective Divide %sable, Non-Perspective Barycentric %sable, "
3505 "Tri Provoking %d, Line Provoking %d, Trifan Provoking %d\n",
3506 (data[2] & (1 << 31)) != 0 ? "en" : "dis",
3507 (data[2] & (1 << 30)) != 0 ? "D3D" : "OGL",
3508 (data[2] & (1 << 28)) != 0 ? "en" : "dis",
3509 (data[2] & (1 << 27)) != 0 ? "en" : "dis",
3510 (data[2] & (1 << 26)) != 0 ? "en" : "dis",
3511 (data[2] >> 13) & 7,
3512 (data[2] & (1 << 9)) != 0 ? "dis" : "en",
3513 (data[2] & (1 << 8)) != 0 ? "en" : "dis",
3514 (data[2] >> 4) & 3, (data[2] >> 2) & 3,
3515 (data[2] & 3));
3516 instr_out(ctx, 3,
3517 "Min PointWidth %d, Max PointWidth %d, "
3518 "Force Zero RTAIndex %sable, Max VPIndex %d\n",
3519 (data[3] >> 17) & 0x7ff, (data[3] >> 6) & 0x7ff,
3520 (data[3] & (1 << 5)) != 0 ? "en" : "dis",
3521 (data[3] & 0xf));
3522 return len;
3523
3524 case 0x7813:
3525 if (ctx->gen == 7)
3526 break;
3527
3528 instr_out(ctx, 0, "3DSTATE_SF\n");
3529 instr_out(ctx, 1,
3530 "Attrib Out %d, Attrib Swizzle %sable, VUE read length %d, "
3531 "VUE read offset %d\n", (data[1] >> 22) & 0x3f,
3532 (data[1] & (1 << 21)) != 0 ? "en" : "dis",
3533 (data[1] >> 11) & 0x1f, (data[1] >> 4) & 0x3f);
3534 instr_out(ctx, 2,
3535 "Legacy Global DepthBias %sable, FrontFace fill %d, BF fill %d, "
3536 "VP transform %sable, FrontWinding_%s\n",
3537 (data[2] & (1 << 11)) != 0 ? "en" : "dis",
3538 (data[2] >> 5) & 3, (data[2] >> 3) & 3,
3539 (data[2] & (1 << 1)) != 0 ? "en" : "dis",
3540 (data[2] & 1) != 0 ? "CCW" : "CW");
3541 instr_out(ctx, 3,
3542 "AA %sable, CullMode %d, Scissor %sable, Multisample m ode %d\n",
3543 (data[3] & (1 << 31)) != 0 ? "en" : "dis",
3544 (data[3] >> 29) & 3,
3545 (data[3] & (1 << 11)) != 0 ? "en" : "dis",
3546 (data[3] >> 8) & 3);
3547 instr_out(ctx, 4,
3548 "Last Pixel %sable, SubPixel Precision %d, Use PixelWidth %d\n",
3549 (data[4] & (1 << 31)) != 0 ? "en" : "dis",
3550 (data[4] & (1 << 12)) != 0 ? 4 : 8,
3551 (data[4] & (1 << 11)) != 0);
3552 instr_out(ctx, 5,
3553 "Global Depth Offset Constant %f\n",
3554 *(float *)(&data[5]));
3555 instr_out(ctx, 6, "Global Depth Offset Scale %f\n",
3556 *(float *)(&data[6]));
3557 instr_out(ctx, 7, "Global Depth Offset Clamp %f\n",
3558 *(float *)(&data[7]));
3559
3560 for (i = 0, j = 0; i < 8; i++, j += 2)
3561 instr_out(ctx, i + 8,
3562 "Attrib %d (Override %s%s%s%s, Const Source %d, Swizzle Select %d, "
3563 "Source %d); Attrib %d (Override %s%s%s%s, Const Source %d, Swizzle Select %d, Source %d)\n",
3564 j + 1,
3565 (data[8 + i] & (1 << 31)) != 0 ? "W" : "",
3566 (data[8 + i] & (1 << 30)) != 0 ? "Z" : "",
3567 (data[8 + i] & (1 << 29)) != 0 ? "Y" : "",
3568 (data[8 + i] & (1 << 28)) != 0 ? "X" : "",
3569 (data[8 + i] >> 25) & 3,
3570 (data[8 + i] >> 22) & 3,
3571 (data[8 + i] >> 16) & 0x1f, j,
3572 (data[8 + i] & (1 << 15)) != 0 ? "W" : "",
3573 (data[8 + i] & (1 << 14)) != 0 ? "Z" : "",
3574 (data[8 + i] & (1 << 13)) != 0 ? "Y" : "",
3575 (data[8 + i] & (1 << 12)) != 0 ? "X" : "",
3576 (data[8 + i] >> 9) & 3,
3577 (data[8 + i] >> 6) & 3, (data[8 + i] & 0x1f));
3578 instr_out(ctx, 16,
3579 "Point Sprite TexCoord Enable\n");
3580 instr_out(ctx, 17, "Const Interp Enable\n");
3581 instr_out(ctx, 18,
3582 "Attrib 7-0 WrapShortest Enable\n");
3583 instr_out(ctx, 19,
3584 "Attrib 15-8 WrapShortest Enable\n");
3585
3586 return len;
3587
3588 case 0x7900:
3589 instr_out(ctx, 0, "3DSTATE_DRAWING_RECTANGLE\n");
3590 instr_out(ctx, 1, "top left: %d,%d\n",
3591 data[1] & 0xffff, (data[1] >> 16) & 0xffff);
3592 instr_out(ctx, 2, "bottom right: %d,%d\n",
3593 data[2] & 0xffff, (data[2] >> 16) & 0xffff);
3594 instr_out(ctx, 3, "origin: %d,%d\n",
3595 (int)data[3] & 0xffff, ((int)data[3] >> 16) & 0xffff);
3596
3597 return len;
3598
3599 case 0x7905:
3600 instr_out(ctx, 0, "3DSTATE_DEPTH_BUFFER\n");
3601 if (IS_GEN5(devid) || IS_GEN6(devid))
3602 instr_out(ctx, 1,
3603 "%s, %s, pitch = %d bytes, %stiled, HiZ %d, Seperate Stencil %d\n",
3604 get_965_surfacetype(data[1] >> 29),
3605 get_965_depthformat((data[1] >> 18) & 0x7),
3606 (data[1] & 0x0001ffff) + 1,
3607 data[1] & (1 << 27) ? "" : "not ",
3608 (data[1] & (1 << 22)) != 0,
3609 (data[1] & (1 << 21)) != 0);
3610 else
3611 instr_out(ctx, 1,
3612 "%s, %s, pitch = %d bytes, %stiled\n",
3613 get_965_surfacetype(data[1] >> 29),
3614 get_965_depthformat((data[1] >> 18) & 0x7),
3615 (data[1] & 0x0001ffff) + 1,
3616 data[1] & (1 << 27) ? "" : "not ");
3617 instr_out(ctx, 2, "depth offset\n");
3618 instr_out(ctx, 3, "%dx%d\n",
3619 ((data[3] & 0x0007ffc0) >> 6) + 1,
3620 ((data[3] & 0xfff80000) >> 19) + 1);
3621 instr_out(ctx, 4, "volume depth\n");
3622 if (len >= 6)
3623 instr_out(ctx, 5, "\n");
3624 if (len >= 7) {
3625 if (IS_GEN6(devid))
3626 instr_out(ctx, 6, "\n");
3627 else
3628 instr_out(ctx, 6,
3629 "render target view extent\n");
3630 }
3631
3632 return len;
3633
3634 case 0x7a00:
3635 if (IS_GEN6(devid) || IS_GEN7(devid)) {
3636 if (len != 4 && len != 5)
3637 fprintf(out, "Bad count in PIPE_CONTROL\n");
3638
3639 switch ((data[1] >> 14) & 0x3) {
3640 case 0:
3641 desc1 = "no write";
3642 break;
3643 case 1:
3644 desc1 = "qword write";
3645 break;
3646 case 2:
3647 desc1 = "PS_DEPTH_COUNT write";
3648 break;
3649 case 3:
3650 desc1 = "TIMESTAMP write";
3651 break;
3652 }
3653 instr_out(ctx, 0, "PIPE_CONTROL\n");
3654 instr_out(ctx, 1,
3655 "%s, %s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n",
3656 desc1,
3657 data[1] & (1 << 20) ? "cs stall, " : "",
3658 data[1] & (1 << 19) ?
3659 "global snapshot count reset, " : "",
3660 data[1] & (1 << 18) ? "tlb invalidate, " : "",
3661 data[1] & (1 << 17) ? "gfdt flush, " : "",
3662 data[1] & (1 << 17) ? "media state clear, " :
3663 "",
3664 data[1] & (1 << 13) ? "depth stall, " : "",
3665 data[1] & (1 << 12) ?
3666 "render target cache flush, " : "",
3667 data[1] & (1 << 11) ?
3668 "instruction cache invalidate, " : "",
3669 data[1] & (1 << 10) ?
3670 "texture cache invalidate, " : "",
3671 data[1] & (1 << 9) ?
3672 "indirect state invalidate, " : "",
3673 data[1] & (1 << 8) ? "notify irq, " : "",
3674 data[1] & (1 << 7) ? "PIPE_CONTROL flush, " :
3675 "",
3676 data[1] & (1 << 6) ? "protect mem app_id, " :
3677 "", data[1] & (1 << 5) ? "DC flush, " : "",
3678 data[1] & (1 << 4) ? "vf fetch invalidate, " :
3679 "",
3680 data[1] & (1 << 3) ?
3681 "constant cache invalidate, " : "",
3682 data[1] & (1 << 2) ?
3683 "state cache invalidate, " : "",
3684 data[1] & (1 << 1) ? "stall at scoreboard, " :
3685 "",
3686 data[1] & (1 << 0) ? "depth cache flush, " :
3687 "");
3688 if (len == 5) {
3689 instr_out(ctx, 2,
3690 "destination address\n");
3691 instr_out(ctx, 3,
3692 "immediate dword low\n");
3693 instr_out(ctx, 4,
3694 "immediate dword high\n");
3695 } else {
3696 for (i = 2; i < len; i++) {
3697 instr_out(ctx, i, "\n");
3698 }
3699 }
3700 return len;
3701 } else {
3702 if (len != 4)
3703 fprintf(out, "Bad count in PIPE_CONTROL\n");
3704
3705 switch ((data[0] >> 14) & 0x3) {
3706 case 0:
3707 desc1 = "no write";
3708 break;
3709 case 1:
3710 desc1 = "qword write";
3711 break;
3712 case 2:
3713 desc1 = "PS_DEPTH_COUNT write";
3714 break;
3715 case 3:
3716 desc1 = "TIMESTAMP write";
3717 break;
3718 }
3719 instr_out(ctx, 0,
3720 "PIPE_CONTROL: %s, %sdepth stall, %sRC write flush, "
3721 "%sinst flush\n",
3722 desc1,
3723 data[0] & (1 << 13) ? "" : "no ",
3724 data[0] & (1 << 12) ? "" : "no ",
3725 data[0] & (1 << 11) ? "" : "no ");
3726 instr_out(ctx, 1, "destination address\n");
3727 instr_out(ctx, 2, "immediate dword low\n");
3728 instr_out(ctx, 3, "immediate dword high\n");
3729 return len;
3730 }
3731 }
3732
3733 if (opcode_3d) {
3734 if (opcode_3d->func) {
3735 return opcode_3d->func(ctx);
3736 } else {
3737 instr_out(ctx, 0, "%s\n", opcode_3d->name);
3738
3739 for (i = 1; i < len; i++) {
3740 instr_out(ctx, i, "dword %d\n", i);
3741 }
3742 return len;
3743 }
3744 }
3745
3746 instr_out(ctx, 0, "3D UNKNOWN: 3d_965 opcode = 0x%x\n",
3747 opcode);
3748 return 1;
3749 }
3750
3751 static int
decode_3d_i830(struct drm_intel_decode * ctx)3752 decode_3d_i830(struct drm_intel_decode *ctx)
3753 {
3754 unsigned int idx;
3755 uint32_t opcode;
3756 uint32_t *data = ctx->data;
3757
3758 struct {
3759 uint32_t opcode;
3760 unsigned int min_len;
3761 unsigned int max_len;
3762 const char *name;
3763 } opcodes_3d[] = {
3764 { 0x02, 1, 1, "3DSTATE_MODES_3" },
3765 { 0x03, 1, 1, "3DSTATE_ENABLES_1" },
3766 { 0x04, 1, 1, "3DSTATE_ENABLES_2" },
3767 { 0x05, 1, 1, "3DSTATE_VFT0" },
3768 { 0x06, 1, 1, "3DSTATE_AA" },
3769 { 0x07, 1, 1, "3DSTATE_RASTERIZATION_RULES" },
3770 { 0x08, 1, 1, "3DSTATE_MODES_1" },
3771 { 0x09, 1, 1, "3DSTATE_STENCIL_TEST" },
3772 { 0x0a, 1, 1, "3DSTATE_VFT1" },
3773 { 0x0b, 1, 1, "3DSTATE_INDPT_ALPHA_BLEND" },
3774 { 0x0c, 1, 1, "3DSTATE_MODES_5" },
3775 { 0x0d, 1, 1, "3DSTATE_MAP_BLEND_OP" },
3776 { 0x0e, 1, 1, "3DSTATE_MAP_BLEND_ARG" },
3777 { 0x0f, 1, 1, "3DSTATE_MODES_2" },
3778 { 0x15, 1, 1, "3DSTATE_FOG_COLOR" },
3779 { 0x16, 1, 1, "3DSTATE_MODES_4"},
3780 }, *opcode_3d;
3781
3782 opcode = (data[0] & 0x1f000000) >> 24;
3783
3784 switch (opcode) {
3785 case 0x1f:
3786 return decode_3d_primitive(ctx);
3787 case 0x1d:
3788 return decode_3d_1d(ctx);
3789 case 0x1c:
3790 return decode_3d_1c(ctx);
3791 }
3792
3793 for (idx = 0; idx < ARRAY_SIZE(opcodes_3d); idx++) {
3794 opcode_3d = &opcodes_3d[idx];
3795 if ((data[0] & 0x1f000000) >> 24 == opcode_3d->opcode) {
3796 unsigned int len = 1, i;
3797
3798 instr_out(ctx, 0, "%s\n", opcode_3d->name);
3799 if (opcode_3d->max_len > 1) {
3800 len = (data[0] & 0xff) + 2;
3801 if (len < opcode_3d->min_len ||
3802 len > opcode_3d->max_len) {
3803 fprintf(out, "Bad count in %s\n",
3804 opcode_3d->name);
3805 }
3806 }
3807
3808 for (i = 1; i < len; i++) {
3809 instr_out(ctx, i, "dword %d\n", i);
3810 }
3811 return len;
3812 }
3813 }
3814
3815 instr_out(ctx, 0, "3D UNKNOWN: 3d_i830 opcode = 0x%x\n",
3816 opcode);
3817 return 1;
3818 }
3819
3820 struct drm_intel_decode *
drm_intel_decode_context_alloc(uint32_t devid)3821 drm_intel_decode_context_alloc(uint32_t devid)
3822 {
3823 struct drm_intel_decode *ctx;
3824
3825 ctx = calloc(1, sizeof(struct drm_intel_decode));
3826 if (!ctx)
3827 return NULL;
3828
3829 ctx->devid = devid;
3830 ctx->out = stdout;
3831
3832 if (IS_GEN9(devid))
3833 ctx->gen = 9;
3834 else if (IS_GEN8(devid))
3835 ctx->gen = 8;
3836 else if (IS_GEN7(devid))
3837 ctx->gen = 7;
3838 else if (IS_GEN6(devid))
3839 ctx->gen = 6;
3840 else if (IS_GEN5(devid))
3841 ctx->gen = 5;
3842 else if (IS_GEN4(devid))
3843 ctx->gen = 4;
3844 else if (IS_9XX(devid))
3845 ctx->gen = 3;
3846 else {
3847 assert(IS_GEN2(devid));
3848 ctx->gen = 2;
3849 }
3850
3851 return ctx;
3852 }
3853
3854 void
drm_intel_decode_context_free(struct drm_intel_decode * ctx)3855 drm_intel_decode_context_free(struct drm_intel_decode *ctx)
3856 {
3857 free(ctx);
3858 }
3859
3860 void
drm_intel_decode_set_dump_past_end(struct drm_intel_decode * ctx,int dump_past_end)3861 drm_intel_decode_set_dump_past_end(struct drm_intel_decode *ctx,
3862 int dump_past_end)
3863 {
3864 ctx->dump_past_end = !!dump_past_end;
3865 }
3866
3867 void
drm_intel_decode_set_batch_pointer(struct drm_intel_decode * ctx,void * data,uint32_t hw_offset,int count)3868 drm_intel_decode_set_batch_pointer(struct drm_intel_decode *ctx,
3869 void *data, uint32_t hw_offset, int count)
3870 {
3871 ctx->base_data = data;
3872 ctx->base_hw_offset = hw_offset;
3873 ctx->base_count = count;
3874 }
3875
3876 void
drm_intel_decode_set_head_tail(struct drm_intel_decode * ctx,uint32_t head,uint32_t tail)3877 drm_intel_decode_set_head_tail(struct drm_intel_decode *ctx,
3878 uint32_t head, uint32_t tail)
3879 {
3880 ctx->head = head;
3881 ctx->tail = tail;
3882 }
3883
3884 void
drm_intel_decode_set_output_file(struct drm_intel_decode * ctx,FILE * output)3885 drm_intel_decode_set_output_file(struct drm_intel_decode *ctx,
3886 FILE *output)
3887 {
3888 ctx->out = output;
3889 }
3890
3891 /**
3892 * Decodes an i830-i915 batch buffer, writing the output to stdout.
3893 *
3894 * \param data batch buffer contents
3895 * \param count number of DWORDs to decode in the batch buffer
3896 * \param hw_offset hardware address for the buffer
3897 */
3898 void
drm_intel_decode(struct drm_intel_decode * ctx)3899 drm_intel_decode(struct drm_intel_decode *ctx)
3900 {
3901 int ret;
3902 unsigned int index = 0;
3903 uint32_t devid;
3904 int size = ctx->base_count * 4;
3905 void *temp;
3906
3907 if (!ctx)
3908 return;
3909
3910 /* Put a scratch page full of obviously undefined data after
3911 * the batchbuffer. This lets us avoid a bunch of length
3912 * checking in statically sized packets.
3913 */
3914 temp = malloc(size + 4096);
3915 memcpy(temp, ctx->base_data, size);
3916 memset((char *)temp + size, 0xd0, 4096);
3917 ctx->data = temp;
3918
3919 ctx->hw_offset = ctx->base_hw_offset;
3920 ctx->count = ctx->base_count;
3921
3922 devid = ctx->devid;
3923 head_offset = ctx->head;
3924 tail_offset = ctx->tail;
3925 out = ctx->out;
3926
3927 saved_s2_set = 0;
3928 saved_s4_set = 1;
3929
3930 while (ctx->count > 0) {
3931 index = 0;
3932
3933 switch ((ctx->data[index] & 0xe0000000) >> 29) {
3934 case 0x0:
3935 ret = decode_mi(ctx);
3936
3937 /* If MI_BATCHBUFFER_END happened, then dump
3938 * the rest of the output in case we some day
3939 * want it in debugging, but don't decode it
3940 * since it'll just confuse in the common
3941 * case.
3942 */
3943 if (ret == -1) {
3944 if (ctx->dump_past_end) {
3945 index++;
3946 } else {
3947 for (index = index + 1; index < ctx->count;
3948 index++) {
3949 instr_out(ctx, index, "\n");
3950 }
3951 }
3952 } else
3953 index += ret;
3954 break;
3955 case 0x2:
3956 index += decode_2d(ctx);
3957 break;
3958 case 0x3:
3959 if (IS_9XX(devid) && !IS_GEN3(devid)) {
3960 index +=
3961 decode_3d_965(ctx);
3962 } else if (IS_GEN3(devid)) {
3963 index += decode_3d(ctx);
3964 } else {
3965 index +=
3966 decode_3d_i830(ctx);
3967 }
3968 break;
3969 default:
3970 instr_out(ctx, index, "UNKNOWN\n");
3971 index++;
3972 break;
3973 }
3974 fflush(out);
3975
3976 if (ctx->count < index)
3977 break;
3978
3979 ctx->count -= index;
3980 ctx->data += index;
3981 ctx->hw_offset += 4 * index;
3982 }
3983
3984 free(temp);
3985 }
3986