i965: only try print GLSL IR once when using INTEL_DEBUG to dump ir
[mesa.git] / src / mesa / drivers / dri / i965 / brw_tes.c
1 /*
2 * Copyright © 2014 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
21 * DEALINGS IN THE SOFTWARE.
22 */
23
24 /**
25 * \file brw_tes.c
26 *
27 * Tessellation evaluation shader state upload code.
28 */
29
30 #include "brw_context.h"
31 #include "brw_nir.h"
32 #include "brw_program.h"
33 #include "brw_shader.h"
34 #include "brw_state.h"
35 #include "program/prog_parameter.h"
36
37 static void
38 brw_tes_debug_recompile(struct brw_context *brw,
39 struct gl_shader_program *shader_prog,
40 const struct brw_tes_prog_key *key)
41 {
42 struct brw_cache_item *c = NULL;
43 const struct brw_tes_prog_key *old_key = NULL;
44 bool found = false;
45
46 perf_debug("Recompiling tessellation evaluation shader for program %d\n",
47 shader_prog->Name);
48
49 for (unsigned int i = 0; i < brw->cache.size; i++) {
50 for (c = brw->cache.items[i]; c; c = c->next) {
51 if (c->cache_id == BRW_CACHE_TES_PROG) {
52 old_key = c->key;
53
54 if (old_key->program_string_id == key->program_string_id)
55 break;
56 }
57 }
58 if (c)
59 break;
60 }
61
62 if (!c) {
63 perf_debug(" Didn't find previous compile in the shader cache for "
64 "debug\n");
65 return;
66 }
67
68 found |= brw_debug_recompile_sampler_key(brw, &old_key->tex, &key->tex);
69 found |= key_debug(brw, "inputs read", old_key->inputs_read,
70 key->inputs_read);
71 found |= key_debug(brw, "patch inputs read", old_key->patch_inputs_read,
72 key->patch_inputs_read);
73
74 if (!found) {
75 perf_debug(" Something else\n");
76 }
77 }
78
79 static bool
80 brw_codegen_tes_prog(struct brw_context *brw,
81 struct gl_shader_program *shader_prog,
82 struct brw_program *tep,
83 struct brw_tes_prog_key *key)
84 {
85 const struct brw_compiler *compiler = brw->screen->compiler;
86 const struct gen_device_info *devinfo = &brw->screen->devinfo;
87 struct brw_stage_state *stage_state = &brw->tes.base;
88 nir_shader *nir = tep->program.nir;
89 struct brw_tes_prog_data prog_data;
90 bool start_busy = false;
91 double start_time = 0;
92
93 memset(&prog_data, 0, sizeof(prog_data));
94
95 brw_assign_common_binding_table_offsets(MESA_SHADER_TESS_EVAL, devinfo,
96 shader_prog, &tep->program,
97 &prog_data.base.base, 0);
98
99 switch (tep->program.info.tes.spacing) {
100 case GL_EQUAL:
101 prog_data.partitioning = BRW_TESS_PARTITIONING_INTEGER;
102 break;
103 case GL_FRACTIONAL_ODD:
104 prog_data.partitioning = BRW_TESS_PARTITIONING_ODD_FRACTIONAL;
105 break;
106 case GL_FRACTIONAL_EVEN:
107 prog_data.partitioning = BRW_TESS_PARTITIONING_EVEN_FRACTIONAL;
108 break;
109 default:
110 unreachable("invalid domain shader spacing");
111 }
112
113 switch (tep->program.info.tes.primitive_mode) {
114 case GL_QUADS:
115 prog_data.domain = BRW_TESS_DOMAIN_QUAD;
116 break;
117 case GL_TRIANGLES:
118 prog_data.domain = BRW_TESS_DOMAIN_TRI;
119 break;
120 case GL_ISOLINES:
121 prog_data.domain = BRW_TESS_DOMAIN_ISOLINE;
122 break;
123 default:
124 unreachable("invalid domain shader primitive mode");
125 }
126
127 if (tep->program.info.tes.point_mode) {
128 prog_data.output_topology = BRW_TESS_OUTPUT_TOPOLOGY_POINT;
129 } else if (tep->program.info.tes.primitive_mode == GL_ISOLINES) {
130 prog_data.output_topology = BRW_TESS_OUTPUT_TOPOLOGY_LINE;
131 } else {
132 /* Hardware winding order is backwards from OpenGL */
133 switch (tep->program.info.tes.vertex_order) {
134 case GL_CCW:
135 prog_data.output_topology = BRW_TESS_OUTPUT_TOPOLOGY_TRI_CW;
136 break;
137 case GL_CW:
138 prog_data.output_topology = BRW_TESS_OUTPUT_TOPOLOGY_TRI_CCW;
139 break;
140 default:
141 unreachable("invalid domain shader vertex order");
142 }
143 }
144
145 /* Allocate the references to the uniforms that will end up in the
146 * prog_data associated with the compiled program, and which will be freed
147 * by the state cache.
148 *
149 * Note: param_count needs to be num_uniform_components * 4, since we add
150 * padding around uniform values below vec4 size, so the worst case is that
151 * every uniform is a float which gets padded to the size of a vec4.
152 */
153 struct gl_linked_shader *tes =
154 shader_prog->_LinkedShaders[MESA_SHADER_TESS_EVAL];
155 int param_count = nir->num_uniforms / 4;
156
157 prog_data.base.base.param =
158 rzalloc_array(NULL, const gl_constant_value *, param_count);
159 prog_data.base.base.pull_param =
160 rzalloc_array(NULL, const gl_constant_value *, param_count);
161 prog_data.base.base.image_param =
162 rzalloc_array(NULL, struct brw_image_param, tes->NumImages);
163 prog_data.base.base.nr_params = param_count;
164 prog_data.base.base.nr_image_params = tes->NumImages;
165
166 prog_data.base.cull_distance_mask =
167 ((1 << tep->program.CullDistanceArraySize) - 1) <<
168 tep->program.ClipDistanceArraySize;
169
170 brw_nir_setup_glsl_uniforms(nir, shader_prog, &tep->program,
171 &prog_data.base.base,
172 compiler->scalar_stage[MESA_SHADER_TESS_EVAL]);
173
174 int st_index = -1;
175 if (unlikely(INTEL_DEBUG & DEBUG_SHADER_TIME))
176 st_index = brw_get_shader_time_index(brw, shader_prog, NULL, ST_TES);
177
178 if (unlikely(brw->perf_debug)) {
179 start_busy = brw->batch.last_bo && drm_intel_bo_busy(brw->batch.last_bo);
180 start_time = get_time();
181 }
182
183 void *mem_ctx = ralloc_context(NULL);
184 unsigned program_size;
185 char *error_str;
186 const unsigned *program =
187 brw_compile_tes(compiler, brw, mem_ctx, key, &prog_data, nir,
188 shader_prog, st_index, &program_size, &error_str);
189 if (program == NULL) {
190 if (shader_prog) {
191 shader_prog->LinkStatus = false;
192 ralloc_strcat(&shader_prog->InfoLog, error_str);
193 }
194
195 _mesa_problem(NULL, "Failed to compile tessellation evaluation shader: "
196 "%s\n", error_str);
197
198 ralloc_free(mem_ctx);
199 return false;
200 }
201
202 if (unlikely(brw->perf_debug)) {
203 struct brw_shader *btes = (struct brw_shader *) tes;
204 if (btes->compiled_once) {
205 brw_tes_debug_recompile(brw, shader_prog, key);
206 }
207 if (start_busy && !drm_intel_bo_busy(brw->batch.last_bo)) {
208 perf_debug("TES compile took %.03f ms and stalled the GPU\n",
209 (get_time() - start_time) * 1000);
210 }
211 btes->compiled_once = true;
212 }
213
214 /* Scratch space is used for register spilling */
215 brw_alloc_stage_scratch(brw, stage_state,
216 prog_data.base.base.total_scratch,
217 devinfo->max_tes_threads);
218
219 brw_upload_cache(&brw->cache, BRW_CACHE_TES_PROG,
220 key, sizeof(*key),
221 program, program_size,
222 &prog_data, sizeof(prog_data),
223 &stage_state->prog_offset, &brw->tes.base.prog_data);
224 ralloc_free(mem_ctx);
225
226 return true;
227 }
228
229 void
230 brw_tes_populate_key(struct brw_context *brw,
231 struct brw_tes_prog_key *key)
232 {
233 struct brw_program *tcp = (struct brw_program *) brw->tess_ctrl_program;
234 struct brw_program *tep = (struct brw_program *) brw->tess_eval_program;
235 struct gl_program *prog = &tep->program;
236
237 uint64_t per_vertex_slots = prog->info.inputs_read;
238 uint32_t per_patch_slots = prog->info.patch_inputs_read;
239
240 memset(key, 0, sizeof(*key));
241
242 key->program_string_id = tep->id;
243
244 /* The TCS may have additional outputs which aren't read by the
245 * TES (possibly for cross-thread communication). These need to
246 * be stored in the Patch URB Entry as well.
247 */
248 if (tcp) {
249 struct gl_program *tcp_prog = &tcp->program;
250 per_vertex_slots |= tcp_prog->info.outputs_written;
251 per_patch_slots |= tcp_prog->info.patch_outputs_written;
252 }
253
254 /* Ignore gl_TessLevelInner/Outer - we treat them as system values,
255 * not inputs, and they're always present in the URB entry regardless
256 * of whether or not we read them.
257 */
258 key->inputs_read = per_vertex_slots &
259 ~(VARYING_BIT_TESS_LEVEL_INNER | VARYING_BIT_TESS_LEVEL_OUTER);
260 key->patch_inputs_read = per_patch_slots;
261
262 /* _NEW_TEXTURE */
263 brw_populate_sampler_prog_key_data(&brw->ctx, prog, &key->tex);
264 }
265
266 void
267 brw_upload_tes_prog(struct brw_context *brw)
268 {
269 struct gl_shader_program **current = brw->ctx._Shader->CurrentProgram;
270 struct brw_stage_state *stage_state = &brw->tes.base;
271 struct brw_tes_prog_key key;
272 /* BRW_NEW_TESS_PROGRAMS */
273 struct brw_program *tep = (struct brw_program *) brw->tess_eval_program;
274
275 if (!brw_state_dirty(brw,
276 _NEW_TEXTURE,
277 BRW_NEW_TESS_PROGRAMS))
278 return;
279
280 brw_tes_populate_key(brw, &key);
281
282 if (!brw_search_cache(&brw->cache, BRW_CACHE_TES_PROG,
283 &key, sizeof(key),
284 &stage_state->prog_offset,
285 &brw->tes.base.prog_data)) {
286 bool success = brw_codegen_tes_prog(brw, current[MESA_SHADER_TESS_EVAL],
287 tep, &key);
288 assert(success);
289 (void)success;
290 }
291 }
292
293
294 bool
295 brw_tes_precompile(struct gl_context *ctx,
296 struct gl_shader_program *shader_prog,
297 struct gl_program *prog)
298 {
299 struct brw_context *brw = brw_context(ctx);
300 struct brw_tes_prog_key key;
301 uint32_t old_prog_offset = brw->tes.base.prog_offset;
302 struct brw_stage_prog_data *old_prog_data = brw->tes.base.prog_data;
303 bool success;
304
305 struct brw_program *btep = brw_program(prog);
306
307 memset(&key, 0, sizeof(key));
308
309 key.program_string_id = btep->id;
310 key.inputs_read = prog->nir->info->inputs_read;
311 key.patch_inputs_read = prog->nir->info->patch_inputs_read;
312
313 if (shader_prog->_LinkedShaders[MESA_SHADER_TESS_CTRL]) {
314 struct gl_program *tcp =
315 shader_prog->_LinkedShaders[MESA_SHADER_TESS_CTRL]->Program;
316 key.inputs_read |= tcp->nir->info->outputs_written;
317 key.patch_inputs_read |= tcp->nir->info->patch_outputs_written;
318 }
319
320 /* Ignore gl_TessLevelInner/Outer - they're system values. */
321 key.inputs_read &= ~(VARYING_BIT_TESS_LEVEL_INNER |
322 VARYING_BIT_TESS_LEVEL_OUTER);
323
324 brw_setup_tex_for_precompile(brw, &key.tex, prog);
325
326 success = brw_codegen_tes_prog(brw, shader_prog, btep, &key);
327
328 brw->tes.base.prog_offset = old_prog_offset;
329 brw->tes.base.prog_data = old_prog_data;
330
331 return success;
332 }