2 * Copyright © 2014 Intel Corporation
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:
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
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.
27 * Tessellation evaluation shader state upload code.
30 #include "brw_context.h"
32 #include "brw_program.h"
33 #include "brw_shader.h"
34 #include "brw_state.h"
35 #include "program/prog_parameter.h"
38 brw_tes_debug_recompile(struct brw_context
*brw
, struct gl_program
*prog
,
39 const struct brw_tes_prog_key
*key
)
41 struct brw_cache_item
*c
= NULL
;
42 const struct brw_tes_prog_key
*old_key
= NULL
;
45 perf_debug("Recompiling tessellation evaluation shader for program %d\n",
48 for (unsigned int i
= 0; i
< brw
->cache
.size
; i
++) {
49 for (c
= brw
->cache
.items
[i
]; c
; c
= c
->next
) {
50 if (c
->cache_id
== BRW_CACHE_TES_PROG
) {
53 if (old_key
->program_string_id
== key
->program_string_id
)
62 perf_debug(" Didn't find previous compile in the shader cache for "
67 found
|= brw_debug_recompile_sampler_key(brw
, &old_key
->tex
, &key
->tex
);
68 found
|= key_debug(brw
, "inputs read", old_key
->inputs_read
,
70 found
|= key_debug(brw
, "patch inputs read", old_key
->patch_inputs_read
,
71 key
->patch_inputs_read
);
74 perf_debug(" Something else\n");
79 brw_codegen_tes_prog(struct brw_context
*brw
,
80 struct brw_program
*tep
,
81 struct brw_tes_prog_key
*key
)
83 const struct brw_compiler
*compiler
= brw
->screen
->compiler
;
84 const struct gen_device_info
*devinfo
= &brw
->screen
->devinfo
;
85 struct brw_stage_state
*stage_state
= &brw
->tes
.base
;
86 nir_shader
*nir
= tep
->program
.nir
;
87 struct brw_tes_prog_data prog_data
;
88 bool start_busy
= false;
89 double start_time
= 0;
91 memset(&prog_data
, 0, sizeof(prog_data
));
93 brw_assign_common_binding_table_offsets(devinfo
, &tep
->program
,
94 &prog_data
.base
.base
, 0);
96 switch (tep
->program
.info
.tes
.spacing
) {
98 prog_data
.partitioning
= BRW_TESS_PARTITIONING_INTEGER
;
100 case GL_FRACTIONAL_ODD
:
101 prog_data
.partitioning
= BRW_TESS_PARTITIONING_ODD_FRACTIONAL
;
103 case GL_FRACTIONAL_EVEN
:
104 prog_data
.partitioning
= BRW_TESS_PARTITIONING_EVEN_FRACTIONAL
;
107 unreachable("invalid domain shader spacing");
110 switch (tep
->program
.info
.tes
.primitive_mode
) {
112 prog_data
.domain
= BRW_TESS_DOMAIN_QUAD
;
115 prog_data
.domain
= BRW_TESS_DOMAIN_TRI
;
118 prog_data
.domain
= BRW_TESS_DOMAIN_ISOLINE
;
121 unreachable("invalid domain shader primitive mode");
124 if (tep
->program
.info
.tes
.point_mode
) {
125 prog_data
.output_topology
= BRW_TESS_OUTPUT_TOPOLOGY_POINT
;
126 } else if (tep
->program
.info
.tes
.primitive_mode
== GL_ISOLINES
) {
127 prog_data
.output_topology
= BRW_TESS_OUTPUT_TOPOLOGY_LINE
;
129 /* Hardware winding order is backwards from OpenGL */
130 switch (tep
->program
.info
.tes
.vertex_order
) {
132 prog_data
.output_topology
= BRW_TESS_OUTPUT_TOPOLOGY_TRI_CW
;
135 prog_data
.output_topology
= BRW_TESS_OUTPUT_TOPOLOGY_TRI_CCW
;
138 unreachable("invalid domain shader vertex order");
142 /* Allocate the references to the uniforms that will end up in the
143 * prog_data associated with the compiled program, and which will be freed
144 * by the state cache.
146 * Note: param_count needs to be num_uniform_components * 4, since we add
147 * padding around uniform values below vec4 size, so the worst case is that
148 * every uniform is a float which gets padded to the size of a vec4.
150 int param_count
= nir
->num_uniforms
/ 4;
152 prog_data
.base
.base
.param
=
153 rzalloc_array(NULL
, const gl_constant_value
*, param_count
);
154 prog_data
.base
.base
.pull_param
=
155 rzalloc_array(NULL
, const gl_constant_value
*, param_count
);
156 prog_data
.base
.base
.image_param
=
157 rzalloc_array(NULL
, struct brw_image_param
,
158 tep
->program
.info
.num_images
);
159 prog_data
.base
.base
.nr_params
= param_count
;
160 prog_data
.base
.base
.nr_image_params
= tep
->program
.info
.num_images
;
162 brw_nir_setup_glsl_uniforms(nir
, &tep
->program
, &prog_data
.base
.base
,
163 compiler
->scalar_stage
[MESA_SHADER_TESS_EVAL
]);
166 if (unlikely(INTEL_DEBUG
& DEBUG_SHADER_TIME
))
167 st_index
= brw_get_shader_time_index(brw
, &tep
->program
, ST_TES
, true);
169 if (unlikely(brw
->perf_debug
)) {
170 start_busy
= brw
->batch
.last_bo
&& drm_intel_bo_busy(brw
->batch
.last_bo
);
171 start_time
= get_time();
174 void *mem_ctx
= ralloc_context(NULL
);
175 unsigned program_size
;
177 const unsigned *program
=
178 brw_compile_tes(compiler
, brw
, mem_ctx
, key
, &prog_data
, nir
,
179 &tep
->program
, st_index
, &program_size
, &error_str
);
180 if (program
== NULL
) {
181 tep
->program
.sh
.data
->LinkStatus
= false;
182 ralloc_strcat(&tep
->program
.sh
.data
->InfoLog
, error_str
);
184 _mesa_problem(NULL
, "Failed to compile tessellation evaluation shader: "
187 ralloc_free(mem_ctx
);
191 if (unlikely(brw
->perf_debug
)) {
192 if (tep
->compiled_once
) {
193 brw_tes_debug_recompile(brw
, &tep
->program
, key
);
195 if (start_busy
&& !drm_intel_bo_busy(brw
->batch
.last_bo
)) {
196 perf_debug("TES compile took %.03f ms and stalled the GPU\n",
197 (get_time() - start_time
) * 1000);
199 tep
->compiled_once
= true;
202 /* Scratch space is used for register spilling */
203 brw_alloc_stage_scratch(brw
, stage_state
,
204 prog_data
.base
.base
.total_scratch
,
205 devinfo
->max_tes_threads
);
207 brw_upload_cache(&brw
->cache
, BRW_CACHE_TES_PROG
,
209 program
, program_size
,
210 &prog_data
, sizeof(prog_data
),
211 &stage_state
->prog_offset
, &brw
->tes
.base
.prog_data
);
212 ralloc_free(mem_ctx
);
218 brw_tes_populate_key(struct brw_context
*brw
,
219 struct brw_tes_prog_key
*key
)
221 struct brw_program
*tcp
= (struct brw_program
*) brw
->tess_ctrl_program
;
222 struct brw_program
*tep
= (struct brw_program
*) brw
->tess_eval_program
;
223 struct gl_program
*prog
= &tep
->program
;
225 uint64_t per_vertex_slots
= prog
->info
.inputs_read
;
226 uint32_t per_patch_slots
= prog
->info
.patch_inputs_read
;
228 memset(key
, 0, sizeof(*key
));
230 key
->program_string_id
= tep
->id
;
232 /* The TCS may have additional outputs which aren't read by the
233 * TES (possibly for cross-thread communication). These need to
234 * be stored in the Patch URB Entry as well.
237 struct gl_program
*tcp_prog
= &tcp
->program
;
238 per_vertex_slots
|= tcp_prog
->info
.outputs_written
&
239 ~(VARYING_BIT_TESS_LEVEL_INNER
| VARYING_BIT_TESS_LEVEL_OUTER
);
240 per_patch_slots
|= tcp_prog
->info
.patch_outputs_written
;
243 key
->inputs_read
= per_vertex_slots
;
244 key
->patch_inputs_read
= per_patch_slots
;
247 brw_populate_sampler_prog_key_data(&brw
->ctx
, prog
, &key
->tex
);
251 brw_upload_tes_prog(struct brw_context
*brw
)
253 struct brw_stage_state
*stage_state
= &brw
->tes
.base
;
254 struct brw_tes_prog_key key
;
255 /* BRW_NEW_TESS_PROGRAMS */
256 struct brw_program
*tep
= (struct brw_program
*) brw
->tess_eval_program
;
258 if (!brw_state_dirty(brw
,
260 BRW_NEW_TESS_PROGRAMS
))
263 brw_tes_populate_key(brw
, &key
);
265 if (!brw_search_cache(&brw
->cache
, BRW_CACHE_TES_PROG
,
267 &stage_state
->prog_offset
,
268 &brw
->tes
.base
.prog_data
)) {
269 bool success
= brw_codegen_tes_prog(brw
, tep
, &key
);
277 brw_tes_precompile(struct gl_context
*ctx
,
278 struct gl_shader_program
*shader_prog
,
279 struct gl_program
*prog
)
281 struct brw_context
*brw
= brw_context(ctx
);
282 struct brw_tes_prog_key key
;
283 uint32_t old_prog_offset
= brw
->tes
.base
.prog_offset
;
284 struct brw_stage_prog_data
*old_prog_data
= brw
->tes
.base
.prog_data
;
287 struct brw_program
*btep
= brw_program(prog
);
289 memset(&key
, 0, sizeof(key
));
291 key
.program_string_id
= btep
->id
;
292 key
.inputs_read
= prog
->nir
->info
->inputs_read
;
293 key
.patch_inputs_read
= prog
->nir
->info
->patch_inputs_read
;
295 if (shader_prog
->_LinkedShaders
[MESA_SHADER_TESS_CTRL
]) {
296 struct gl_program
*tcp
=
297 shader_prog
->_LinkedShaders
[MESA_SHADER_TESS_CTRL
]->Program
;
298 key
.inputs_read
|= tcp
->nir
->info
->outputs_written
&
299 ~(VARYING_BIT_TESS_LEVEL_INNER
| VARYING_BIT_TESS_LEVEL_OUTER
);
300 key
.patch_inputs_read
|= tcp
->nir
->info
->patch_outputs_written
;
303 brw_setup_tex_for_precompile(brw
, &key
.tex
, prog
);
305 success
= brw_codegen_tes_prog(brw
, btep
, &key
);
307 brw
->tes
.base
.prog_offset
= old_prog_offset
;
308 brw
->tes
.base
.prog_data
= old_prog_data
;