summaryrefslogtreecommitdiff
path: root/src/gallium/drivers/svga/svga_shader.c
blob: 55f79224779e703aefb928209c657fcf1c13a04b (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
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
/**********************************************************
 * Copyright 2008-2012 VMware, Inc.  All rights reserved.
 *
 * 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 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 "util/u_bitmask.h"
#include "util/u_memory.h"
#include "svga_context.h"
#include "svga_cmd.h"
#include "svga_format.h"
#include "svga_shader.h"


/**
 * This bit isn't really used anywhere.  It only serves to help
 * generate a unique "signature" for the vertex shader output bitmask.
 * Shader input/output signatures are used to resolve shader linking
 * issues.
 */
#define FOG_GENERIC_BIT (((uint64_t) 1) << 63)


/**
 * Use the shader info to generate a bitmask indicating which generic
 * inputs are used by the shader.  A set bit indicates that GENERIC[i]
 * is used.
 */
uint64_t
svga_get_generic_inputs_mask(const struct tgsi_shader_info *info)
{
   unsigned i;
   uint64_t mask = 0x0;

   for (i = 0; i < info->num_inputs; i++) {
      if (info->input_semantic_name[i] == TGSI_SEMANTIC_GENERIC) {
         unsigned j = info->input_semantic_index[i];
         assert(j < sizeof(mask) * 8);
         mask |= ((uint64_t) 1) << j;
      }
   }

   return mask;
}


/**
 * Scan shader info to return a bitmask of written outputs.
 */
uint64_t
svga_get_generic_outputs_mask(const struct tgsi_shader_info *info)
{
   unsigned i;
   uint64_t mask = 0x0;

   for (i = 0; i < info->num_outputs; i++) {
      switch (info->output_semantic_name[i]) {
      case TGSI_SEMANTIC_GENERIC:
         {
            unsigned j = info->output_semantic_index[i];
            assert(j < sizeof(mask) * 8);
            mask |= ((uint64_t) 1) << j;
         }
         break;
      case TGSI_SEMANTIC_FOG:
         mask |= FOG_GENERIC_BIT;
         break;
      }
   }

   return mask;
}



/**
 * Given a mask of used generic variables (as returned by the above functions)
 * fill in a table which maps those indexes to small integers.
 * This table is used by the remap_generic_index() function in
 * svga_tgsi_decl_sm30.c
 * Example: if generics_mask = binary(1010) it means that GENERIC[1] and
 * GENERIC[3] are used.  The remap_table will contain:
 *   table[1] = 0;
 *   table[3] = 1;
 * The remaining table entries will be filled in with the next unused
 * generic index (in this example, 2).
 */
void
svga_remap_generics(uint64_t generics_mask,
                    int8_t remap_table[MAX_GENERIC_VARYING])
{
   /* Note texcoord[0] is reserved so start at 1 */
   unsigned count = 1, i;

   for (i = 0; i < MAX_GENERIC_VARYING; i++) {
      remap_table[i] = -1;
   }

   /* for each bit set in generic_mask */
   while (generics_mask) {
      unsigned index = ffsll(generics_mask) - 1;
      remap_table[index] = count++;
      generics_mask &= ~((uint64_t) 1 << index);
   }
}


/**
 * Use the generic remap table to map a TGSI generic varying variable
 * index to a small integer.  If the remapping table doesn't have a
 * valid value for the given index (the table entry is -1) it means
 * the fragment shader doesn't use that VS output.  Just allocate
 * the next free value in that case.  Alternately, we could cull
 * VS instructions that write to register, or replace the register
 * with a dummy temp register.
 * XXX TODO: we should do one of the later as it would save precious
 * texcoord registers.
 */
int
svga_remap_generic_index(int8_t remap_table[MAX_GENERIC_VARYING],
                         int generic_index)
{
   assert(generic_index < MAX_GENERIC_VARYING);

   if (generic_index >= MAX_GENERIC_VARYING) {
      /* just don't return a random/garbage value */
      generic_index = MAX_GENERIC_VARYING - 1;
   }

   if (remap_table[generic_index] == -1) {
      /* This is a VS output that has no matching PS input.  Find a
       * free index.
       */
      int i, max = 0;
      for (i = 0; i < MAX_GENERIC_VARYING; i++) {
         max = MAX2(max, remap_table[i]);
      }
      remap_table[generic_index] = max + 1;
   }

   return remap_table[generic_index];
}


/**
 * Initialize the shader-neutral fields of svga_compile_key from context
 * state.  This is basically the texture-related state.
 */
void
svga_init_shader_key_common(const struct svga_context *svga,
                            enum pipe_shader_type shader,
                            struct svga_compile_key *key)
{
   unsigned i, idx = 0;

   assert(shader < ARRAY_SIZE(svga->curr.num_sampler_views));

   /* In case the number of samplers and sampler_views doesn't match,
    * loop over the lower of the two counts.
    */
   key->num_textures = MIN2(svga->curr.num_sampler_views[shader],
                            svga->curr.num_samplers[shader]);

   for (i = 0; i < key->num_textures; i++) {
      struct pipe_sampler_view *view = svga->curr.sampler_views[shader][i];
      const struct svga_sampler_state *sampler = svga->curr.sampler[shader][i];
      if (view && sampler) {
         assert(view->texture);
         assert(view->texture->target < (1 << 4)); /* texture_target:4 */

         /* 1D/2D array textures with one slice are treated as non-arrays
          * by the SVGA3D device.  Convert the texture type here so that
          * we emit the right TEX/SAMPLE instruction in the shader.
          */
         if (view->texture->target == PIPE_TEXTURE_1D_ARRAY ||
             view->texture->target == PIPE_TEXTURE_2D_ARRAY) {
            if (view->texture->array_size == 1) {
               key->tex[i].is_array = 0;
            }
            else {
               assert(view->texture->array_size > 1);
               key->tex[i].is_array = 1;
            }
         }

         if (!sampler->normalized_coords) {
            assert(idx < (1 << 5));  /* width_height_idx:5 bitfield */
            key->tex[i].width_height_idx = idx++;
            key->tex[i].unnormalized = TRUE;
            ++key->num_unnormalized_coords;
         }

         key->tex[i].swizzle_r = view->swizzle_r;
         key->tex[i].swizzle_g = view->swizzle_g;
         key->tex[i].swizzle_b = view->swizzle_b;
         key->tex[i].swizzle_a = view->swizzle_a;
      }
   }
}


/** Search for a compiled shader variant with the same compile key */
struct svga_shader_variant *
svga_search_shader_key(const struct svga_shader *shader,
                       const struct svga_compile_key *key)
{
   struct svga_shader_variant *variant = shader->variants;

   assert(key);

   for ( ; variant; variant = variant->next) {
      if (svga_compile_keys_equal(key, &variant->key))
         return variant;
   }
   return NULL;
}

/** Search for a shader with the same token key */
struct svga_shader *
svga_search_shader_token_key(struct svga_shader *pshader,
                             const struct svga_token_key *key)
{
   struct svga_shader *shader = pshader;

   assert(key);

   for ( ; shader; shader = shader->next) {
      if (memcmp(key, &shader->token_key, sizeof(struct svga_token_key)) == 0)
         return shader;
   }
   return NULL;
}

/**
 * Helper function to define a gb shader for non-vgpu10 device
 */
static enum pipe_error
define_gb_shader_vgpu9(struct svga_context *svga,
                       SVGA3dShaderType type,
                       struct svga_shader_variant *variant,
                       unsigned codeLen)
{
   struct svga_winsys_screen *sws = svga_screen(svga->pipe.screen)->sws;
   enum pipe_error ret;

   /**
    * Create gb memory for the shader and upload the shader code.
    * Kernel module will allocate an id for the shader and issue
    * the DefineGBShader command.
    */
   variant->gb_shader = sws->shader_create(sws, type,
                                           variant->tokens, codeLen);

   if (!variant->gb_shader)
      return PIPE_ERROR_OUT_OF_MEMORY;

   ret = SVGA3D_BindGBShader(svga->swc, variant->gb_shader);

   return ret;
}

/**
 * Helper function to define a gb shader for vgpu10 device
 */
static enum pipe_error
define_gb_shader_vgpu10(struct svga_context *svga,
                        SVGA3dShaderType type,
                        struct svga_shader_variant *variant,
                        unsigned codeLen)
{
   struct svga_winsys_context *swc = svga->swc;
   enum pipe_error ret;

   /**
    * Shaders in VGPU10 enabled device reside in the device COTable.
    * SVGA driver will allocate an integer ID for the shader and
    * issue DXDefineShader and DXBindShader commands.
    */
   variant->id = util_bitmask_add(svga->shader_id_bm);
   if (variant->id == UTIL_BITMASK_INVALID_INDEX) {
      return PIPE_ERROR_OUT_OF_MEMORY;
   }

   /* Create gb memory for the shader and upload the shader code */
   variant->gb_shader = swc->shader_create(swc,
                                           variant->id, type,
                                           variant->tokens, codeLen);

   if (!variant->gb_shader) {
      /* Free the shader ID */
      assert(variant->id != UTIL_BITMASK_INVALID_INDEX);
      goto fail_no_allocation;
   }

   /**
    * Since we don't want to do any flush within state emission to avoid
    * partial state in a command buffer, it's important to make sure that
    * there is enough room to send both the DXDefineShader & DXBindShader
    * commands in the same command buffer. So let's send both
    * commands in one command reservation. If it fails, we'll undo
    * the shader creation and return an error.
    */
   ret = SVGA3D_vgpu10_DefineAndBindShader(swc, variant->gb_shader,
                                           variant->id, type, codeLen);

   if (ret != PIPE_OK)
      goto fail;

   return PIPE_OK;

fail:
   swc->shader_destroy(swc, variant->gb_shader);
   variant->gb_shader = NULL;

fail_no_allocation:
   util_bitmask_clear(svga->shader_id_bm, variant->id);
   variant->id = UTIL_BITMASK_INVALID_INDEX;

   return PIPE_ERROR_OUT_OF_MEMORY;
}

/**
 * Issue the SVGA3D commands to define a new shader.
 * \param variant  contains the shader tokens, etc.  The result->id field will
 *                 be set here.
 */
enum pipe_error
svga_define_shader(struct svga_context *svga,
                   SVGA3dShaderType type,
                   struct svga_shader_variant *variant)
{
   unsigned codeLen = variant->nr_tokens * sizeof(variant->tokens[0]);
   enum pipe_error ret;

   SVGA_STATS_TIME_PUSH(svga_sws(svga), SVGA_STATS_TIME_DEFINESHADER);

   variant->id = UTIL_BITMASK_INVALID_INDEX;

   if (svga_have_gb_objects(svga)) {
      if (svga_have_vgpu10(svga))
         ret = define_gb_shader_vgpu10(svga, type, variant, codeLen);
      else
         ret = define_gb_shader_vgpu9(svga, type, variant, codeLen);
   }
   else {
      /* Allocate an integer ID for the shader */
      variant->id = util_bitmask_add(svga->shader_id_bm);
      if (variant->id == UTIL_BITMASK_INVALID_INDEX) {
         ret = PIPE_ERROR_OUT_OF_MEMORY;
         goto done;
      }

      /* Issue SVGA3D device command to define the shader */
      ret = SVGA3D_DefineShader(svga->swc,
                                variant->id,
                                type,
                                variant->tokens,
                                codeLen);
      if (ret != PIPE_OK) {
         /* free the ID */
         assert(variant->id != UTIL_BITMASK_INVALID_INDEX);
         util_bitmask_clear(svga->shader_id_bm, variant->id);
         variant->id = UTIL_BITMASK_INVALID_INDEX;
      }
   }

done:
   SVGA_STATS_TIME_POP(svga_sws(svga));
   return ret;
}


/**
 * Issue the SVGA3D commands to set/bind a shader.
 * \param result  the shader to bind.
 */
enum pipe_error
svga_set_shader(struct svga_context *svga,
                SVGA3dShaderType type,
                struct svga_shader_variant *variant)
{
   enum pipe_error ret;
   unsigned id = variant ? variant->id : SVGA3D_INVALID_ID;

   assert(type == SVGA3D_SHADERTYPE_VS ||
          type == SVGA3D_SHADERTYPE_GS ||
          type == SVGA3D_SHADERTYPE_PS);

   if (svga_have_gb_objects(svga)) {
      struct svga_winsys_gb_shader *gbshader =
         variant ? variant->gb_shader : NULL;

      if (svga_have_vgpu10(svga))
         ret = SVGA3D_vgpu10_SetShader(svga->swc, type, gbshader, id);
      else
         ret = SVGA3D_SetGBShader(svga->swc, type, gbshader);
   }
   else {
      ret = SVGA3D_SetShader(svga->swc, type, id);
   }

   return ret;
}


struct svga_shader_variant *
svga_new_shader_variant(struct svga_context *svga)
{
   svga->hud.num_shaders++;
   return CALLOC_STRUCT(svga_shader_variant);
}


enum pipe_error
svga_destroy_shader_variant(struct svga_context *svga,
                            SVGA3dShaderType type,
                            struct svga_shader_variant *variant)
{
   enum pipe_error ret = PIPE_OK;

   if (svga_have_gb_objects(svga) && variant->gb_shader) {
      if (svga_have_vgpu10(svga)) {
         struct svga_winsys_context *swc = svga->swc;
         swc->shader_destroy(swc, variant->gb_shader);
         ret = SVGA3D_vgpu10_DestroyShader(svga->swc, variant->id);
         if (ret != PIPE_OK) {
            /* flush and try again */
            svga_context_flush(svga, NULL);
            ret = SVGA3D_vgpu10_DestroyShader(svga->swc, variant->id);
         }
         util_bitmask_clear(svga->shader_id_bm, variant->id);
      }
      else {
         struct svga_winsys_screen *sws = svga_screen(svga->pipe.screen)->sws;
         sws->shader_destroy(sws, variant->gb_shader);
      }
      variant->gb_shader = NULL;
   }
   else {
      if (variant->id != UTIL_BITMASK_INVALID_INDEX) {
         ret = SVGA3D_DestroyShader(svga->swc, variant->id, type);
         if (ret != PIPE_OK) {
            /* flush and try again */
            svga_context_flush(svga, NULL);
            ret = SVGA3D_DestroyShader(svga->swc, variant->id, type);
            assert(ret == PIPE_OK);
         }
         util_bitmask_clear(svga->shader_id_bm, variant->id);
      }
   }

   FREE((unsigned *)variant->tokens);
   FREE(variant);

   svga->hud.num_shaders--;

   return ret;
}

/*
 * Rebind shaders.
 * Called at the beginning of every new command buffer to ensure that
 * shaders are properly paged-in. Instead of sending the SetShader
 * command, this function sends a private allocation command to
 * page in a shader. This avoids emitting redundant state to the device
 * just to page in a resource.
 */
enum pipe_error
svga_rebind_shaders(struct svga_context *svga)
{
   struct svga_winsys_context *swc = svga->swc;
   struct svga_hw_draw_state *hw = &svga->state.hw_draw;
   enum pipe_error ret;

   assert(svga_have_vgpu10(svga));

   /**
    * If the underlying winsys layer does not need resource rebinding,
    * just clear the rebind flags and return.
    */
   if (swc->resource_rebind == NULL) {
      svga->rebind.flags.vs = 0;
      svga->rebind.flags.gs = 0;
      svga->rebind.flags.fs = 0;

      return PIPE_OK;
   }

   if (svga->rebind.flags.vs && hw->vs && hw->vs->gb_shader) {
      ret = swc->resource_rebind(swc, NULL, hw->vs->gb_shader, SVGA_RELOC_READ);
      if (ret != PIPE_OK)
         return ret;
   }
   svga->rebind.flags.vs = 0;

   if (svga->rebind.flags.gs && hw->gs && hw->gs->gb_shader) {
      ret = swc->resource_rebind(swc, NULL, hw->gs->gb_shader, SVGA_RELOC_READ);
      if (ret != PIPE_OK)
         return ret;
   }
   svga->rebind.flags.gs = 0;

   if (svga->rebind.flags.fs && hw->fs && hw->fs->gb_shader) {
      ret = swc->resource_rebind(swc, NULL, hw->fs->gb_shader, SVGA_RELOC_READ);
      if (ret != PIPE_OK)
         return ret;
   }
   svga->rebind.flags.fs = 0;

   return PIPE_OK;
}