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 DEALINGS
25 #include "brw_shader.h"
26 #include "glsl/glsl_parser_extras.h"
27 #include "glsl/nir/glsl_to_nir.h"
28 #include "program/prog_to_nir.h"
31 nir_optimize(nir_shader *nir, bool is_scalar)
36 nir_lower_vars_to_ssa(nir);
37 nir_validate_shader(nir);
40 nir_lower_alu_to_scalar(nir);
41 nir_validate_shader(nir);
44 progress |= nir_copy_prop(nir);
45 nir_validate_shader(nir);
48 nir_lower_phis_to_scalar(nir);
49 nir_validate_shader(nir);
52 progress |= nir_copy_prop(nir);
53 nir_validate_shader(nir);
54 progress |= nir_opt_dce(nir);
55 nir_validate_shader(nir);
56 progress |= nir_opt_cse(nir);
57 nir_validate_shader(nir);
58 progress |= nir_opt_peephole_select(nir);
59 nir_validate_shader(nir);
60 progress |= nir_opt_algebraic(nir);
61 nir_validate_shader(nir);
62 progress |= nir_opt_constant_folding(nir);
63 nir_validate_shader(nir);
64 progress |= nir_opt_remove_phis(nir);
65 nir_validate_shader(nir);
66 progress |= nir_opt_undef(nir);
67 nir_validate_shader(nir);
72 brw_create_nir(struct brw_context *brw,
73 const struct gl_shader_program *shader_prog,
74 const struct gl_program *prog,
75 gl_shader_stage stage,
78 struct gl_context *ctx = &brw->ctx;
79 const nir_shader_compiler_options *options =
80 ctx->Const.ShaderCompilerOptions[stage].NirOptions;
81 struct gl_shader *shader = shader_prog ? shader_prog->_LinkedShaders[stage] : NULL;
84 /* First, lower the GLSL IR or Mesa IR to NIR */
86 nir = glsl_to_nir(shader, options);
88 nir = prog_to_nir(prog, options);
89 nir_convert_to_ssa(nir); /* turn registers into SSA */
91 nir_validate_shader(nir);
93 brw_process_nir(nir, brw->intelScreen->devinfo, shader_prog, stage, is_scalar);
95 static GLuint msg_id = 0;
96 _mesa_gl_debug(&brw->ctx, &msg_id,
97 MESA_DEBUG_SOURCE_SHADER_COMPILER,
98 MESA_DEBUG_TYPE_OTHER,
99 MESA_DEBUG_SEVERITY_NOTIFICATION,
101 _mesa_shader_stage_to_abbrev(stage));
107 brw_process_nir(nir_shader *nir,
108 const struct brw_device_info *devinfo,
109 const struct gl_shader_program *shader_prog,
110 gl_shader_stage stage, bool is_scalar)
112 bool debug_enabled = INTEL_DEBUG & intel_debug_flag_for_shader_stage(stage);
114 nir_lower_global_vars_to_local(nir);
115 nir_validate_shader(nir);
117 nir_lower_tex_projector(nir);
118 nir_validate_shader(nir);
120 nir_normalize_cubemap_coords(nir);
121 nir_validate_shader(nir);
123 nir_split_var_copies(nir);
124 nir_validate_shader(nir);
126 nir_optimize(nir, is_scalar);
128 /* Lower a bunch of stuff */
129 nir_lower_var_copies(nir);
130 nir_validate_shader(nir);
132 /* Get rid of split copies */
133 nir_optimize(nir, is_scalar);
136 nir_assign_var_locations(&nir->uniforms,
139 nir_assign_var_locations(&nir->inputs, &nir->num_inputs, type_size_scalar);
140 nir_assign_var_locations(&nir->outputs, &nir->num_outputs, type_size_scalar);
141 nir_lower_io(nir, type_size_scalar);
143 nir_assign_var_locations(&nir->uniforms,
147 nir_assign_var_locations(&nir->inputs, &nir->num_inputs, type_size_vec4);
149 foreach_list_typed(nir_variable, var, node, &nir->outputs)
150 var->data.driver_location = var->data.location;
152 nir_lower_io(nir, type_size_vec4);
155 nir_validate_shader(nir);
157 nir_remove_dead_variables(nir);
158 nir_validate_shader(nir);
161 nir_lower_samplers(nir, shader_prog);
163 nir_lower_samplers_for_vk(nir);
165 nir_validate_shader(nir);
167 nir_lower_system_values(nir);
168 nir_validate_shader(nir);
170 nir_lower_atomics(nir);
171 nir_validate_shader(nir);
173 nir_optimize(nir, is_scalar);
175 if (devinfo->gen >= 6) {
176 /* Try and fuse multiply-adds */
177 nir_opt_peephole_ffma(nir);
178 nir_validate_shader(nir);
181 nir_opt_algebraic_late(nir);
182 nir_validate_shader(nir);
184 nir_lower_locals_to_regs(nir);
185 nir_validate_shader(nir);
187 nir_lower_to_source_mods(nir);
188 nir_validate_shader(nir);
190 nir_validate_shader(nir);
192 nir_validate_shader(nir);
194 if (unlikely(debug_enabled)) {
195 /* Re-index SSA defs so we print more sensible numbers. */
196 nir_foreach_overload(nir, overload) {
198 nir_index_ssa_defs(overload->impl);
201 fprintf(stderr, "NIR (SSA form) for %s shader:\n",
202 _mesa_shader_stage_to_string(stage));
203 nir_print_shader(nir, stderr);
206 nir_convert_from_ssa(nir, is_scalar);
207 nir_validate_shader(nir);
210 nir_lower_vec_to_movs(nir);
211 nir_validate_shader(nir);
214 /* This is the last pass we run before we start emitting stuff. It
215 * determines when we need to insert boolean resolves on Gen <= 5. We
216 * run it last because it stashes data in instr->pass_flags and we don't
217 * want that to be squashed by other NIR passes.
219 if (devinfo->gen <= 5)
220 brw_nir_analyze_boolean_resolves(nir);
224 if (unlikely(debug_enabled)) {
225 fprintf(stderr, "NIR (final form) for %s shader:\n",
226 _mesa_shader_stage_to_string(stage));
227 nir_print_shader(nir, stderr);
232 brw_type_for_nir_type(nir_alu_type type)
235 case nir_type_unsigned:
236 return BRW_REGISTER_TYPE_UD;
239 return BRW_REGISTER_TYPE_D;
241 return BRW_REGISTER_TYPE_F;
243 unreachable("unknown type");
246 return BRW_REGISTER_TYPE_F;
249 /* Returns the glsl_base_type corresponding to a nir_alu_type.
250 * This is used by both brw_vec4_nir and brw_fs_nir.
253 brw_glsl_base_type_for_nir_type(nir_alu_type type)
257 return GLSL_TYPE_FLOAT;
260 return GLSL_TYPE_INT;
262 case nir_type_unsigned:
263 return GLSL_TYPE_UINT;
266 unreachable("bad type");