aboutsummaryrefslogtreecommitdiffstats
path: root/src/mesa/drivers/dri/i965/brw_disk_cache.c
blob: 1b4cb437cbc227fdfddb72bd1b00487041d0a269 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
/*
 * Copyright © 2014 Intel Corporation
 *
 * Permission is hereby granted, free of charge, to any person obtaining a
 * copy of this software and associated documentation files (the "Software"),
 * to deal in the Software without restriction, including without limitation
 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
 * and/or sell copies of the Software, and to permit persons to whom the
 * Software is furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice (including the next
 * paragraph) shall be included in all copies or substantial portions of the
 * Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
 * IN THE SOFTWARE.
 */

#include "compiler/glsl/ir_uniform.h"
#include "compiler/glsl/shader_cache.h"
#include "main/mtypes.h"
#include "util/blob.h"
#include "util/build_id.h"
#include "util/debug.h"
#include "util/disk_cache.h"
#include "util/macros.h"
#include "util/mesa-sha1.h"

#include "compiler/brw_eu.h"
#include "dev/gen_debug.h"

#include "brw_context.h"
#include "brw_program.h"
#include "brw_cs.h"
#include "brw_gs.h"
#include "brw_state.h"
#include "brw_vs.h"
#include "brw_wm.h"

static bool
debug_enabled_for_stage(gl_shader_stage stage)
{
   static const uint64_t stage_debug_flags[] = {
      DEBUG_VS, DEBUG_TCS, DEBUG_TES, DEBUG_GS, DEBUG_WM, DEBUG_CS,
   };
   assert((int)stage >= 0 && stage < ARRAY_SIZE(stage_debug_flags));
   return (INTEL_DEBUG & stage_debug_flags[stage]) != 0;
}

static void
gen_shader_sha1(struct gl_program *prog, gl_shader_stage stage,
                void *key, unsigned char *out_sha1)
{
   char sha1_buf[41];
   unsigned char sha1[20];
   char manifest[256];
   int offset = 0;

   _mesa_sha1_format(sha1_buf, prog->sh.data->sha1);
   offset += snprintf(manifest, sizeof(manifest), "program: %s\n", sha1_buf);

   _mesa_sha1_compute(key, brw_prog_key_size(stage), sha1);
   _mesa_sha1_format(sha1_buf, sha1);
   offset += snprintf(manifest + offset, sizeof(manifest) - offset,
                      "%s_key: %s\n", _mesa_shader_stage_to_abbrev(stage),
                      sha1_buf);

   _mesa_sha1_compute(manifest, strlen(manifest), out_sha1);
}

static bool
read_blob_program_data(struct blob_reader *binary, struct gl_program *prog,
                       gl_shader_stage stage, const uint8_t **program,
                       struct brw_stage_prog_data *prog_data)
{
   return
      brw_read_blob_program_data(binary, prog, stage, program, prog_data) &&
      (binary->current == binary->end);
}

static bool
read_and_upload(struct brw_context *brw, struct disk_cache *cache,
                struct gl_program *prog, gl_shader_stage stage)
{
   unsigned char binary_sha1[20];

   union brw_any_prog_key prog_key;

   switch (stage) {
   case MESA_SHADER_VERTEX:
      brw_vs_populate_key(brw, &prog_key.vs);
      break;
   case MESA_SHADER_TESS_CTRL:
      brw_tcs_populate_key(brw, &prog_key.tcs);
      break;
   case MESA_SHADER_TESS_EVAL:
      brw_tes_populate_key(brw, &prog_key.tes);
      break;
   case MESA_SHADER_GEOMETRY:
      brw_gs_populate_key(brw, &prog_key.gs);
      break;
   case MESA_SHADER_FRAGMENT:
      brw_wm_populate_key(brw, &prog_key.wm);
      break;
   case MESA_SHADER_COMPUTE:
      brw_cs_populate_key(brw, &prog_key.cs);
      break;
   default:
      unreachable("Unsupported stage!");
   }

   /* We don't care what instance of the program it is for the disk cache hash
    * lookup, so set the id to 0 for the sha1 hashing. program_string_id will
    * be set below.
    */
   prog_key.base.program_string_id = 0;

   gen_shader_sha1(prog, stage, &prog_key, binary_sha1);

   size_t buffer_size;
   uint8_t *buffer = disk_cache_get(cache, binary_sha1, &buffer_size);
   if (buffer == NULL) {
      if (brw->ctx._Shader->Flags & GLSL_CACHE_INFO) {
         char sha1_buf[41];
         _mesa_sha1_format(sha1_buf, binary_sha1);
         fprintf(stderr, "No cached %s binary found for: %s\n",
                 _mesa_shader_stage_to_abbrev(stage), sha1_buf);
      }
      return false;
   }

   if (brw->ctx._Shader->Flags & GLSL_CACHE_INFO) {
      char sha1_buf[41];
      _mesa_sha1_format(sha1_buf, binary_sha1);
      fprintf(stderr, "attempting to populate bo cache with binary: %s\n",
              sha1_buf);
   }

   struct blob_reader binary;
   blob_reader_init(&binary, buffer, buffer_size);

   const uint8_t *program;
   struct brw_stage_prog_data *prog_data =
      ralloc_size(NULL, sizeof(union brw_any_prog_data));
   if (!read_blob_program_data(&binary, prog, stage, &program, prog_data)) {
      /* Something very bad has gone wrong discard the item from the cache and
       * rebuild from source.
       */
      if (brw->ctx._Shader->Flags & GLSL_CACHE_INFO) {
         fprintf(stderr, "Error reading program from cache (invalid i965 "
                 "cache item)\n");
      }

      disk_cache_remove(cache, binary_sha1);
      ralloc_free(prog_data);
      free(buffer);
      return false;
   }

   enum brw_cache_id cache_id;
   struct brw_stage_state *stage_state;

   switch (stage) {
   case MESA_SHADER_VERTEX:
      cache_id = BRW_CACHE_VS_PROG;
      stage_state = &brw->vs.base;
      break;
   case MESA_SHADER_TESS_CTRL:
      cache_id = BRW_CACHE_TCS_PROG;
      stage_state = &brw->tcs.base;
      break;
   case MESA_SHADER_TESS_EVAL:
      cache_id = BRW_CACHE_TES_PROG;
      stage_state = &brw->tes.base;
      break;
   case MESA_SHADER_GEOMETRY:
      cache_id = BRW_CACHE_GS_PROG;
      stage_state = &brw->gs.base;
      break;
   case MESA_SHADER_FRAGMENT:
      cache_id = BRW_CACHE_FS_PROG;
      stage_state = &brw->wm.base;
      break;
   case MESA_SHADER_COMPUTE:
      cache_id = BRW_CACHE_CS_PROG;
      stage_state = &brw->cs.base;
      break;
   default:
      unreachable("Unsupported stage!");
   }

   prog_key.base.program_string_id = brw_program(prog)->id;

   brw_alloc_stage_scratch(brw, stage_state, prog_data->total_scratch);

   if (unlikely(debug_enabled_for_stage(stage))) {
      fprintf(stderr, "NIR for %s program %d loaded from disk shader cache:\n",
              _mesa_shader_stage_to_abbrev(stage), brw_program(prog)->id);
      brw_program_deserialize_driver_blob(&brw->ctx, prog, stage);
      nir_shader *nir = prog->nir;
      nir_print_shader(nir, stderr);
      fprintf(stderr, "Native code for %s %s shader %s from disk cache:\n",
              nir->info.label ? nir->info.label : "unnamed",
              _mesa_shader_stage_to_string(nir->info.stage), nir->info.name);
      brw_disassemble(&brw->screen->devinfo, program, 0,
                      prog_data->program_size, stderr);
   }

   brw_upload_cache(&brw->cache, cache_id, &prog_key, brw_prog_key_size(stage),
                    program, prog_data->program_size, prog_data,
                    brw_prog_data_size(stage), &stage_state->prog_offset,
                    &stage_state->prog_data);

   prog->program_written_to_cache = true;

   ralloc_free(prog_data);
   free(buffer);

   return true;
}

bool
brw_disk_cache_upload_program(struct brw_context *brw, gl_shader_stage stage)
{
   struct disk_cache *cache = brw->ctx.Cache;
   if (cache == NULL)
      return false;

   struct gl_program *prog = brw->ctx._Shader->CurrentProgram[stage];
   if (prog == NULL)
      return false;

   if (prog->sh.data->spirv)
      return false;

   if (brw->ctx._Shader->Flags & GLSL_CACHE_FALLBACK)
      goto fail;

   if (!read_and_upload(brw, cache, prog, stage))
      goto fail;

   if (brw->ctx._Shader->Flags & GLSL_CACHE_INFO) {
      fprintf(stderr, "read gen program from cache\n");
   }

   return true;

fail:
   prog->program_written_to_cache = false;
   if (brw->ctx._Shader->Flags & GLSL_CACHE_INFO) {
      fprintf(stderr, "falling back to nir %s.\n",
              _mesa_shader_stage_to_abbrev(prog->info.stage));
   }

   brw_program_deserialize_driver_blob(&brw->ctx, prog, stage);

   return false;
}

static void
write_program_data(struct brw_context *brw, struct gl_program *prog,
                   void *key, struct brw_stage_prog_data *prog_data,
                   uint32_t prog_offset, struct disk_cache *cache,
                   gl_shader_stage stage)
{
   struct blob binary;
   blob_init(&binary);

   const void *program_map = brw->cache.map + prog_offset;
   /* TODO: Improve perf for non-LLC. It would be best to save it at program
    * generation time when the program is in normal memory accessible with
    * cache to the CPU. Another easier change would be to use
    * _mesa_streaming_load_memcpy to read from the program mapped memory. */
   brw_write_blob_program_data(&binary, stage, program_map, prog_data);

   unsigned char sha1[20];
   char buf[41];
   gen_shader_sha1(prog, stage, key, sha1);
   _mesa_sha1_format(buf, sha1);
   if (brw->ctx._Shader->Flags & GLSL_CACHE_INFO) {
      fprintf(stderr, "putting binary in cache: %s\n", buf);
   }

   disk_cache_put(cache, sha1, binary.data, binary.size, NULL);

   prog->program_written_to_cache = true;
   blob_finish(&binary);
}

void
brw_disk_cache_write_render_programs(struct brw_context *brw)
{
   struct disk_cache *cache = brw->ctx.Cache;
   if (cache == NULL)
      return;

   struct gl_program *prog;
   gl_shader_stage stage;
   for (stage = MESA_SHADER_VERTEX; stage <= MESA_SHADER_FRAGMENT; stage++) {
      prog = brw->ctx._Shader->CurrentProgram[stage];
      if (prog && prog->sh.data->spirv)
         return;
   }

   prog = brw->ctx._Shader->CurrentProgram[MESA_SHADER_VERTEX];
   if (prog && !prog->program_written_to_cache) {
      struct brw_vs_prog_key vs_key;
      brw_vs_populate_key(brw, &vs_key);
      vs_key.base.program_string_id = 0;

      write_program_data(brw, prog, &vs_key, brw->vs.base.prog_data,
                         brw->vs.base.prog_offset, cache,
                         MESA_SHADER_VERTEX);
   }

   prog = brw->ctx._Shader->CurrentProgram[MESA_SHADER_TESS_CTRL];
   if (prog && !prog->program_written_to_cache) {
      struct brw_tcs_prog_key tcs_key;
      brw_tcs_populate_key(brw, &tcs_key);
      tcs_key.base.program_string_id = 0;

      write_program_data(brw, prog, &tcs_key, brw->tcs.base.prog_data,
                         brw->tcs.base.prog_offset, cache,
                         MESA_SHADER_TESS_CTRL);
   }

   prog = brw->ctx._Shader->CurrentProgram[MESA_SHADER_TESS_EVAL];
   if (prog && !prog->program_written_to_cache) {
      struct brw_tes_prog_key tes_key;
      brw_tes_populate_key(brw, &tes_key);
      tes_key.base.program_string_id = 0;

      write_program_data(brw, prog, &tes_key, brw->tes.base.prog_data,
                         brw->tes.base.prog_offset, cache,
                         MESA_SHADER_TESS_EVAL);
   }

   prog = brw->ctx._Shader->CurrentProgram[MESA_SHADER_GEOMETRY];
   if (prog && !prog->program_written_to_cache) {
      struct brw_gs_prog_key gs_key;
      brw_gs_populate_key(brw, &gs_key);
      gs_key.base.program_string_id = 0;

      write_program_data(brw, prog, &gs_key, brw->gs.base.prog_data,
                         brw->gs.base.prog_offset, cache,
                         MESA_SHADER_GEOMETRY);
   }

   prog = brw->ctx._Shader->CurrentProgram[MESA_SHADER_FRAGMENT];
   if (prog && !prog->program_written_to_cache) {
      struct brw_wm_prog_key wm_key;
      brw_wm_populate_key(brw, &wm_key);
      wm_key.base.program_string_id = 0;

      write_program_data(brw, prog, &wm_key, brw->wm.base.prog_data,
                         brw->wm.base.prog_offset, cache,
                         MESA_SHADER_FRAGMENT);
   }
}

void
brw_disk_cache_write_compute_program(struct brw_context *brw)
{
   struct disk_cache *cache = brw->ctx.Cache;
   if (cache == NULL)
      return;

   struct gl_program *prog =
      brw->ctx._Shader->CurrentProgram[MESA_SHADER_COMPUTE];

   if (prog && prog->sh.data->spirv)
      return;

   if (prog && !prog->program_written_to_cache) {
      struct brw_cs_prog_key cs_key;
      brw_cs_populate_key(brw, &cs_key);
      cs_key.base.program_string_id = 0;

      write_program_data(brw, prog, &cs_key, brw->cs.base.prog_data,
                         brw->cs.base.prog_offset, cache,
                         MESA_SHADER_COMPUTE);
   }
}

void
brw_disk_cache_init(struct intel_screen *screen)
{
#ifdef ENABLE_SHADER_CACHE
   if (INTEL_DEBUG & DEBUG_DISK_CACHE_DISABLE_MASK)
      return;

   /* array length: print length + null char + 1 extra to verify it is unused */
   char renderer[11];
   ASSERTED int len = snprintf(renderer, sizeof(renderer), "i965_%04x",
                                   screen->deviceID);
   assert(len == sizeof(renderer) - 2);

   const struct build_id_note *note =
      build_id_find_nhdr_for_addr(brw_disk_cache_init);
   assert(note && build_id_length(note) == 20 /* sha1 */);

   const uint8_t *id_sha1 = build_id_data(note);
   assert(id_sha1);

   char timestamp[41];
   _mesa_sha1_format(timestamp, id_sha1);

   const uint64_t driver_flags =
      brw_get_compiler_config_value(screen->compiler);
   screen->disk_cache = disk_cache_create(renderer, timestamp, driver_flags);
#endif
}