i965: Convert scalar_* flags to a scalar_stage array.
[mesa.git] / src / mesa / drivers / dri / i965 / brw_gs.c
1 /*
2 * Copyright © 2013 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_vec4_gs.c
26 *
27 * State atom for client-programmable geometry shaders, and support code.
28 */
29
30 #include "brw_gs.h"
31 #include "brw_context.h"
32 #include "brw_vec4_gs_visitor.h"
33 #include "brw_state.h"
34 #include "brw_ff_gs.h"
35 #include "brw_nir.h"
36
37 static void
38 assign_gs_binding_table_offsets(const struct brw_device_info *devinfo,
39 const struct gl_shader_program *shader_prog,
40 const struct gl_program *prog,
41 struct brw_gs_prog_data *prog_data)
42 {
43 /* In gen6 we reserve the first BRW_MAX_SOL_BINDINGS entries for transform
44 * feedback surfaces.
45 */
46 uint32_t reserved = devinfo->gen == 6 ? BRW_MAX_SOL_BINDINGS : 0;
47
48 brw_assign_common_binding_table_offsets(MESA_SHADER_GEOMETRY, devinfo,
49 shader_prog, prog,
50 &prog_data->base.base,
51 reserved);
52 }
53
54 bool
55 brw_codegen_gs_prog(struct brw_context *brw,
56 struct gl_shader_program *prog,
57 struct brw_geometry_program *gp,
58 struct brw_gs_prog_key *key)
59 {
60 struct brw_compiler *compiler = brw->intelScreen->compiler;
61 struct gl_shader *shader = prog->_LinkedShaders[MESA_SHADER_GEOMETRY];
62 struct brw_stage_state *stage_state = &brw->gs.base;
63 struct brw_gs_prog_data prog_data;
64 memset(&prog_data, 0, sizeof(prog_data));
65
66 assign_gs_binding_table_offsets(brw->intelScreen->devinfo, prog,
67 &gp->program.Base, &prog_data);
68
69 /* Allocate the references to the uniforms that will end up in the
70 * prog_data associated with the compiled program, and which will be freed
71 * by the state cache.
72 *
73 * Note: param_count needs to be num_uniform_components * 4, since we add
74 * padding around uniform values below vec4 size, so the worst case is that
75 * every uniform is a float which gets padded to the size of a vec4.
76 */
77 struct gl_shader *gs = prog->_LinkedShaders[MESA_SHADER_GEOMETRY];
78 int param_count = gp->program.Base.nir->num_uniforms * 4;
79
80 prog_data.base.base.param =
81 rzalloc_array(NULL, const gl_constant_value *, param_count);
82 prog_data.base.base.pull_param =
83 rzalloc_array(NULL, const gl_constant_value *, param_count);
84 prog_data.base.base.image_param =
85 rzalloc_array(NULL, struct brw_image_param, gs->NumImages);
86 prog_data.base.base.nr_params = param_count;
87 prog_data.base.base.nr_image_params = gs->NumImages;
88
89 brw_nir_setup_glsl_uniforms(gp->program.Base.nir, prog, &gp->program.Base,
90 &prog_data.base.base,
91 compiler->scalar_stage[MESA_SHADER_GEOMETRY]);
92
93 GLbitfield64 outputs_written = gp->program.Base.OutputsWritten;
94
95 brw_compute_vue_map(brw->intelScreen->devinfo,
96 &prog_data.base.vue_map, outputs_written,
97 prog ? prog->SeparateShader : false);
98
99 if (unlikely(INTEL_DEBUG & DEBUG_GS))
100 brw_dump_ir("geometry", prog, gs, NULL);
101
102 int st_index = -1;
103 if (INTEL_DEBUG & DEBUG_SHADER_TIME)
104 st_index = brw_get_shader_time_index(brw, prog, NULL, ST_GS);
105
106 void *mem_ctx = ralloc_context(NULL);
107 unsigned program_size;
108 char *error_str;
109 const unsigned *program =
110 brw_compile_gs(brw->intelScreen->compiler, brw, mem_ctx, key,
111 &prog_data, shader->Program->nir, prog,
112 st_index, &program_size, &error_str);
113 if (program == NULL) {
114 ralloc_free(mem_ctx);
115 return false;
116 }
117
118 /* Scratch space is used for register spilling */
119 if (prog_data.base.base.total_scratch) {
120 brw_get_scratch_bo(brw, &stage_state->scratch_bo,
121 prog_data.base.base.total_scratch *
122 brw->max_gs_threads);
123 }
124
125 brw_upload_cache(&brw->cache, BRW_CACHE_GS_PROG,
126 key, sizeof(*key),
127 program, program_size,
128 &prog_data, sizeof(prog_data),
129 &stage_state->prog_offset, &brw->gs.prog_data);
130 ralloc_free(mem_ctx);
131
132 return true;
133 }
134
135 static bool
136 brw_gs_state_dirty(struct brw_context *brw)
137 {
138 return brw_state_dirty(brw,
139 _NEW_TEXTURE,
140 BRW_NEW_GEOMETRY_PROGRAM |
141 BRW_NEW_TRANSFORM_FEEDBACK);
142 }
143
144 static void
145 brw_gs_populate_key(struct brw_context *brw,
146 struct brw_gs_prog_key *key)
147 {
148 struct gl_context *ctx = &brw->ctx;
149 struct brw_stage_state *stage_state = &brw->gs.base;
150 struct brw_geometry_program *gp =
151 (struct brw_geometry_program *) brw->geometry_program;
152 struct gl_program *prog = &gp->program.Base;
153
154 memset(key, 0, sizeof(*key));
155
156 key->program_string_id = gp->id;
157
158 /* _NEW_TEXTURE */
159 brw_populate_sampler_prog_key_data(ctx, prog, stage_state->sampler_count,
160 &key->tex);
161 }
162
163 void
164 brw_upload_gs_prog(struct brw_context *brw)
165 {
166 struct gl_context *ctx = &brw->ctx;
167 struct gl_shader_program **current = ctx->_Shader->CurrentProgram;
168 struct brw_stage_state *stage_state = &brw->gs.base;
169 struct brw_gs_prog_key key;
170 /* BRW_NEW_GEOMETRY_PROGRAM */
171 struct brw_geometry_program *gp =
172 (struct brw_geometry_program *) brw->geometry_program;
173
174 if (!brw_gs_state_dirty(brw))
175 return;
176
177 if (gp == NULL) {
178 /* No geometry shader. Vertex data just passes straight through. */
179 if (brw->gen == 6 &&
180 (brw->ctx.NewDriverState & BRW_NEW_TRANSFORM_FEEDBACK)) {
181 gen6_brw_upload_ff_gs_prog(brw);
182 return;
183 }
184
185 /* Other state atoms had better not try to access prog_data, since
186 * there's no GS program.
187 */
188 brw->gs.prog_data = NULL;
189 brw->gs.base.prog_data = NULL;
190
191 return;
192 }
193
194 brw_gs_populate_key(brw, &key);
195
196 if (!brw_search_cache(&brw->cache, BRW_CACHE_GS_PROG,
197 &key, sizeof(key),
198 &stage_state->prog_offset, &brw->gs.prog_data)) {
199 bool success = brw_codegen_gs_prog(brw, current[MESA_SHADER_GEOMETRY],
200 gp, &key);
201 assert(success);
202 (void)success;
203 }
204 brw->gs.base.prog_data = &brw->gs.prog_data->base.base;
205 }
206
207 bool
208 brw_gs_precompile(struct gl_context *ctx,
209 struct gl_shader_program *shader_prog,
210 struct gl_program *prog)
211 {
212 struct brw_context *brw = brw_context(ctx);
213 struct brw_gs_prog_key key;
214 uint32_t old_prog_offset = brw->gs.base.prog_offset;
215 struct brw_gs_prog_data *old_prog_data = brw->gs.prog_data;
216 bool success;
217
218 struct gl_geometry_program *gp = (struct gl_geometry_program *) prog;
219 struct brw_geometry_program *bgp = brw_geometry_program(gp);
220
221 memset(&key, 0, sizeof(key));
222
223 brw_setup_tex_for_precompile(brw, &key.tex, prog);
224 key.program_string_id = bgp->id;
225
226 success = brw_codegen_gs_prog(brw, shader_prog, bgp, &key);
227
228 brw->gs.base.prog_offset = old_prog_offset;
229 brw->gs.prog_data = old_prog_data;
230
231 return success;
232 }