summaryrefslogtreecommitdiffstats
path: root/src/mesa/drivers/dri/i965/brw_tcs.c
blob: 3f8798a82aeba2d72df8aeac015c58642c5301b5 (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
/*
 * 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.
 */

/**
 * \file brw_tcs.c
 *
 * Tessellation control shader state upload code.
 */

#include "brw_context.h"
#include "brw_nir.h"
#include "brw_program.h"
#include "brw_shader.h"
#include "brw_state.h"
#include "program/prog_parameter.h"
#include "nir_builder.h"

static nir_shader *
create_passthrough_tcs(void *mem_ctx, const struct brw_compiler *compiler,
                       const nir_shader_compiler_options *options,
                       const struct brw_tcs_prog_key *key)
{
   nir_builder b;
   nir_builder_init_simple_shader(&b, mem_ctx, MESA_SHADER_TESS_CTRL,
                                  options);
   nir_shader *nir = b.shader;
   nir_variable *var;
   nir_intrinsic_instr *load;
   nir_intrinsic_instr *store;
   nir_ssa_def *zero = nir_imm_int(&b, 0);
   nir_ssa_def *invoc_id =
      nir_load_system_value(&b, nir_intrinsic_load_invocation_id, 0);

   nir->info->inputs_read = key->outputs_written;
   nir->info->outputs_written = key->outputs_written;
   nir->info->tcs.vertices_out = key->input_vertices;
   nir->info->name = ralloc_strdup(nir, "passthrough");
   nir->num_uniforms = 8 * sizeof(uint32_t);

   var = nir_variable_create(nir, nir_var_uniform, glsl_vec4_type(), "hdr_0");
   var->data.location = 0;
   var = nir_variable_create(nir, nir_var_uniform, glsl_vec4_type(), "hdr_1");
   var->data.location = 1;

   /* Write the patch URB header. */
   for (int i = 0; i <= 1; i++) {
      load = nir_intrinsic_instr_create(nir, nir_intrinsic_load_uniform);
      load->num_components = 4;
      load->src[0] = nir_src_for_ssa(zero);
      nir_ssa_dest_init(&load->instr, &load->dest, 4, 32, NULL);
      nir_intrinsic_set_base(load, i * 4 * sizeof(uint32_t));
      nir_builder_instr_insert(&b, &load->instr);

      store = nir_intrinsic_instr_create(nir, nir_intrinsic_store_output);
      store->num_components = 4;
      store->src[0] = nir_src_for_ssa(&load->dest.ssa);
      store->src[1] = nir_src_for_ssa(zero);
      nir_intrinsic_set_base(store, VARYING_SLOT_TESS_LEVEL_INNER - i);
      nir_intrinsic_set_write_mask(store, WRITEMASK_XYZW);
      nir_builder_instr_insert(&b, &store->instr);
   }

   /* Copy inputs to outputs. */
   uint64_t varyings = key->outputs_written;

   while (varyings != 0) {
      const int varying = ffsll(varyings) - 1;

      load = nir_intrinsic_instr_create(nir,
                                        nir_intrinsic_load_per_vertex_input);
      load->num_components = 4;
      load->src[0] = nir_src_for_ssa(invoc_id);
      load->src[1] = nir_src_for_ssa(zero);
      nir_ssa_dest_init(&load->instr, &load->dest, 4, 32, NULL);
      nir_intrinsic_set_base(load, varying);
      nir_builder_instr_insert(&b, &load->instr);

      store = nir_intrinsic_instr_create(nir,
                                         nir_intrinsic_store_per_vertex_output);
      store->num_components = 4;
      store->src[0] = nir_src_for_ssa(&load->dest.ssa);
      store->src[1] = nir_src_for_ssa(invoc_id);
      store->src[2] = nir_src_for_ssa(zero);
      nir_intrinsic_set_base(store, varying);
      nir_intrinsic_set_write_mask(store, WRITEMASK_XYZW);
      nir_builder_instr_insert(&b, &store->instr);

      varyings &= ~BITFIELD64_BIT(varying);
   }

   nir_validate_shader(nir);

   nir = brw_preprocess_nir(compiler, nir);

   return nir;
}

static void
brw_tcs_debug_recompile(struct brw_context *brw, struct gl_program *prog,
                       const struct brw_tcs_prog_key *key)
{
   struct brw_cache_item *c = NULL;
   const struct brw_tcs_prog_key *old_key = NULL;
   bool found = false;

   perf_debug("Recompiling tessellation control shader for program %d\n",
              prog->Id);

   for (unsigned int i = 0; i < brw->cache.size; i++) {
      for (c = brw->cache.items[i]; c; c = c->next) {
         if (c->cache_id == BRW_CACHE_TCS_PROG) {
            old_key = c->key;

            if (old_key->program_string_id == key->program_string_id)
               break;
         }
      }
      if (c)
         break;
   }

   if (!c) {
      perf_debug("  Didn't find previous compile in the shader cache for "
                 "debug\n");
      return;
   }

   found |= key_debug(brw, "input vertices", old_key->input_vertices,
                      key->input_vertices);
   found |= key_debug(brw, "outputs written", old_key->outputs_written,
                      key->outputs_written);
   found |= key_debug(brw, "patch outputs written", old_key->patch_outputs_written,
                      key->patch_outputs_written);
   found |= key_debug(brw, "TES primitive mode", old_key->tes_primitive_mode,
                      key->tes_primitive_mode);
   found |= key_debug(brw, "quads and equal_spacing workaround",
                      old_key->quads_workaround, key->quads_workaround);
   found |= brw_debug_recompile_sampler_key(brw, &old_key->tex, &key->tex);

   if (!found) {
      perf_debug("  Something else\n");
   }
}

static bool
brw_codegen_tcs_prog(struct brw_context *brw,
                     struct gl_shader_program *shader_prog,
                     struct brw_program *tcp,
                     struct brw_program *tep,
                     struct brw_tcs_prog_key *key)
{
   struct gl_context *ctx = &brw->ctx;
   const struct brw_compiler *compiler = brw->screen->compiler;
   const struct gen_device_info *devinfo = compiler->devinfo;
   struct brw_stage_state *stage_state = &brw->tcs.base;
   nir_shader *nir;
   struct brw_tcs_prog_data prog_data;
   bool start_busy = false;
   double start_time = 0;

   void *mem_ctx = ralloc_context(NULL);
   if (tcp) {
      nir = tcp->program.nir;
   } else {
      /* Create a dummy nir_shader.  We won't actually use NIR code to
       * generate assembly (it's easier to generate assembly directly),
       * but the whole compiler assumes one of these exists.
       */
      const nir_shader_compiler_options *options =
         ctx->Const.ShaderCompilerOptions[MESA_SHADER_TESS_CTRL].NirOptions;
      nir = create_passthrough_tcs(mem_ctx, compiler, options, key);
   }

   memset(&prog_data, 0, sizeof(prog_data));

   /* Allocate the references to the uniforms that will end up in the
    * prog_data associated with the compiled program, and which will be freed
    * by the state cache.
    *
    * Note: param_count needs to be num_uniform_components * 4, since we add
    * padding around uniform values below vec4 size, so the worst case is that
    * every uniform is a float which gets padded to the size of a vec4.
    */
   int param_count = nir->num_uniforms / 4;

   prog_data.base.base.param =
      rzalloc_array(NULL, const gl_constant_value *, param_count);
   prog_data.base.base.pull_param =
      rzalloc_array(NULL, const gl_constant_value *, param_count);
   prog_data.base.base.nr_params = param_count;

   if (tcp) {
      brw_assign_common_binding_table_offsets(MESA_SHADER_TESS_CTRL, devinfo,
                                              shader_prog, &tcp->program,
                                              &prog_data.base.base, 0);

      prog_data.base.base.image_param =
         rzalloc_array(NULL, struct brw_image_param,
                       tcp->program.info.num_images);
      prog_data.base.base.nr_image_params = tcp->program.info.num_images;

      brw_nir_setup_glsl_uniforms(nir, shader_prog, &tcp->program,
                                  &prog_data.base.base,
                                  compiler->scalar_stage[MESA_SHADER_TESS_CTRL]);
   } else {
      /* Upload the Patch URB Header as the first two uniforms.
       * Do the annoying scrambling so the shader doesn't have to.
       */
      const float **param = (const float **) prog_data.base.base.param;
      static float zero = 0.0f;
      for (int i = 0; i < 8; i++)
         param[i] = &zero;

      if (key->tes_primitive_mode == GL_QUADS) {
         for (int i = 0; i < 4; i++)
            param[7 - i] = &ctx->TessCtrlProgram.patch_default_outer_level[i];

         param[3] = &ctx->TessCtrlProgram.patch_default_inner_level[0];
         param[2] = &ctx->TessCtrlProgram.patch_default_inner_level[1];
      } else if (key->tes_primitive_mode == GL_TRIANGLES) {
         for (int i = 0; i < 3; i++)
            param[7 - i] = &ctx->TessCtrlProgram.patch_default_outer_level[i];

         param[4] = &ctx->TessCtrlProgram.patch_default_inner_level[0];
      } else {
         assert(key->tes_primitive_mode == GL_ISOLINES);
         param[7] = &ctx->TessCtrlProgram.patch_default_outer_level[1];
         param[6] = &ctx->TessCtrlProgram.patch_default_outer_level[0];
      }
   }

   int st_index = -1;
   if (unlikely((INTEL_DEBUG & DEBUG_SHADER_TIME) && tep))
      st_index = brw_get_shader_time_index(brw, &tep->program, ST_TCS, true);

   if (unlikely(brw->perf_debug)) {
      start_busy = brw->batch.last_bo && drm_intel_bo_busy(brw->batch.last_bo);
      start_time = get_time();
   }

   unsigned program_size;
   char *error_str;
   const unsigned *program =
      brw_compile_tcs(compiler, brw, mem_ctx, key, &prog_data, nir, st_index,
                      &program_size, &error_str);
   if (program == NULL) {
      if (tep) {
         tep->program.sh.data->LinkStatus = false;
         ralloc_strcat(&tep->program.sh.data->InfoLog, error_str);
      }

      _mesa_problem(NULL, "Failed to compile tessellation control shader: "
                    "%s\n", error_str);

      ralloc_free(mem_ctx);
      return false;
   }

   if (unlikely(brw->perf_debug)) {
      if (tcp) {
         if (tcp->compiled_once) {
            brw_tcs_debug_recompile(brw, &tcp->program, key);
         }
         tcp->compiled_once = true;
      }

      if (start_busy && !drm_intel_bo_busy(brw->batch.last_bo)) {
         perf_debug("TCS compile took %.03f ms and stalled the GPU\n",
                    (get_time() - start_time) * 1000);
      }
   }

   /* Scratch space is used for register spilling */
   brw_alloc_stage_scratch(brw, stage_state,
                           prog_data.base.base.total_scratch,
                           devinfo->max_tcs_threads);

   brw_upload_cache(&brw->cache, BRW_CACHE_TCS_PROG,
                    key, sizeof(*key),
                    program, program_size,
                    &prog_data, sizeof(prog_data),
                    &stage_state->prog_offset, &brw->tcs.base.prog_data);
   ralloc_free(mem_ctx);

   return true;
}

void
brw_tcs_populate_key(struct brw_context *brw,
                     struct brw_tcs_prog_key *key)
{
   struct brw_program *tcp = (struct brw_program *) brw->tess_ctrl_program;
   struct brw_program *tep = (struct brw_program *) brw->tess_eval_program;
   struct gl_program *tes_prog = &tep->program;

   uint64_t per_vertex_slots = tes_prog->info.inputs_read;
   uint32_t per_patch_slots = tes_prog->info.patch_inputs_read;

   memset(key, 0, sizeof(*key));

   if (tcp) {
      struct gl_program *prog = &tcp->program;
      per_vertex_slots |= prog->info.outputs_written;
      per_patch_slots |= prog->info.patch_outputs_written;
   }

   if (brw->gen < 8 || !tcp)
      key->input_vertices = brw->ctx.TessCtrlProgram.patch_vertices;
   key->outputs_written = per_vertex_slots;
   key->patch_outputs_written = per_patch_slots;

   /* We need to specialize our code generation for tessellation levels
    * based on the domain the DS is expecting to tessellate.
    */
   key->tes_primitive_mode = tep->program.info.tes.primitive_mode;
   key->quads_workaround = brw->gen < 9 &&
                           tep->program.info.tes.primitive_mode == GL_QUADS &&
                           tep->program.info.tes.spacing == GL_EQUAL;

   if (tcp) {
      key->program_string_id = tcp->id;

      /* _NEW_TEXTURE */
      brw_populate_sampler_prog_key_data(&brw->ctx, &tcp->program, &key->tex);
   }
}

void
brw_upload_tcs_prog(struct brw_context *brw)
{
   struct gl_shader_program **current = brw->ctx._Shader->CurrentProgram;
   struct brw_stage_state *stage_state = &brw->tcs.base;
   struct brw_tcs_prog_key key;
   /* BRW_NEW_TESS_PROGRAMS */
   struct brw_program *tcp = (struct brw_program *) brw->tess_ctrl_program;
   MAYBE_UNUSED struct brw_program *tep =
      (struct brw_program *) brw->tess_eval_program;
   assert(tep);

   if (!brw_state_dirty(brw,
                        _NEW_TEXTURE,
                        BRW_NEW_PATCH_PRIMITIVE |
                        BRW_NEW_TESS_PROGRAMS))
      return;

   brw_tcs_populate_key(brw, &key);

   if (!brw_search_cache(&brw->cache, BRW_CACHE_TCS_PROG,
                         &key, sizeof(key),
                         &stage_state->prog_offset,
                         &brw->tcs.base.prog_data)) {
      bool success = brw_codegen_tcs_prog(brw, current[MESA_SHADER_TESS_CTRL],
                                          tcp, tep, &key);
      assert(success);
      (void)success;
   }
}


bool
brw_tcs_precompile(struct gl_context *ctx,
                   struct gl_shader_program *shader_prog,
                   struct gl_program *prog)
{
   struct brw_context *brw = brw_context(ctx);
   struct brw_tcs_prog_key key;
   uint32_t old_prog_offset = brw->tcs.base.prog_offset;
   struct brw_stage_prog_data *old_prog_data = brw->tcs.base.prog_data;
   bool success;

   struct brw_program *btcp = brw_program(prog);
   const struct gl_linked_shader *tes =
      shader_prog->_LinkedShaders[MESA_SHADER_TESS_EVAL];

   memset(&key, 0, sizeof(key));

   key.program_string_id = btcp->id;
   brw_setup_tex_for_precompile(brw, &key.tex, prog);

   /* Guess that the input and output patches have the same dimensionality. */
   if (brw->gen < 8) {
      key.input_vertices = shader_prog->
         _LinkedShaders[MESA_SHADER_TESS_CTRL]->info.TessCtrl.VerticesOut;
   }

   struct brw_program *btep;
   if (tes) {
      btep = brw_program(tes->Program);
      key.tes_primitive_mode = tes->info.TessEval.PrimitiveMode;
      key.quads_workaround = brw->gen < 9 &&
                             tes->info.TessEval.PrimitiveMode == GL_QUADS &&
                             tes->info.TessEval.Spacing == GL_EQUAL;
   } else {
      btep = NULL;
      key.tes_primitive_mode = GL_TRIANGLES;
   }

   key.outputs_written = prog->nir->info->outputs_written;
   key.patch_outputs_written = prog->nir->info->patch_outputs_written;

   success = brw_codegen_tcs_prog(brw, shader_prog, btcp, btep, &key);

   brw->tcs.base.prog_offset = old_prog_offset;
   brw->tcs.base.prog_data = old_prog_data;

   return success;
}