1 /*
2 * Copyright © 2015 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 DEALINGS
21 * IN THE SOFTWARE.
22 */
23
24 #ifndef BRW_NIR_H
25 #define BRW_NIR_H
26
27 #include "brw_reg.h"
28 #include "compiler/nir/nir.h"
29 #include "brw_compiler.h"
30
31 #ifdef __cplusplus
32 extern "C" {
33 #endif
34
35 int type_size_vec4(const struct glsl_type *type, bool bindless);
36 int type_size_dvec4(const struct glsl_type *type, bool bindless);
37
38 static inline int
type_size_scalar_bytes(const struct glsl_type * type,bool bindless)39 type_size_scalar_bytes(const struct glsl_type *type, bool bindless)
40 {
41 return glsl_count_dword_slots(type, bindless) * 4;
42 }
43
44 static inline int
type_size_vec4_bytes(const struct glsl_type * type,bool bindless)45 type_size_vec4_bytes(const struct glsl_type *type, bool bindless)
46 {
47 return type_size_vec4(type, bindless) * 16;
48 }
49
50 /* Flags set in the instr->pass_flags field by i965 analysis passes */
51 enum {
52 BRW_NIR_NON_BOOLEAN = 0x0,
53
54 /* Indicates that the given instruction's destination is a boolean
55 * value but that it needs to be resolved before it can be used.
56 * On Gen <= 5, CMP instructions return a 32-bit value where the bottom
57 * bit represents the actual true/false value of the compare and the top
58 * 31 bits are undefined. In order to use this value, we have to do a
59 * "resolve" operation by replacing the value of the CMP with -(x & 1)
60 * to sign-extend the bottom bit to 0/~0.
61 */
62 BRW_NIR_BOOLEAN_NEEDS_RESOLVE = 0x1,
63
64 /* Indicates that the given instruction's destination is a boolean
65 * value that has intentionally been left unresolved. Not all boolean
66 * values need to be resolved immediately. For instance, if we have
67 *
68 * CMP r1 r2 r3
69 * CMP r4 r5 r6
70 * AND r7 r1 r4
71 *
72 * We don't have to resolve the result of the two CMP instructions
73 * immediately because the AND still does an AND of the bottom bits.
74 * Instead, we can save ourselves instructions by delaying the resolve
75 * until after the AND. The result of the two CMP instructions is left
76 * as BRW_NIR_BOOLEAN_UNRESOLVED.
77 */
78 BRW_NIR_BOOLEAN_UNRESOLVED = 0x2,
79
80 /* Indicates a that the given instruction's destination is a boolean
81 * value that does not need a resolve. For instance, if you AND two
82 * values that are BRW_NIR_BOOLEAN_NEEDS_RESOLVE then we know that both
83 * values will be 0/~0 before we get them and the result of the AND is
84 * also guaranteed to be 0/~0 and does not need a resolve.
85 */
86 BRW_NIR_BOOLEAN_NO_RESOLVE = 0x3,
87
88 /* A mask to mask the boolean status values off of instr->pass_flags */
89 BRW_NIR_BOOLEAN_MASK = 0x3,
90 };
91
92 void brw_nir_analyze_boolean_resolves(nir_shader *nir);
93
94 void brw_preprocess_nir(const struct brw_compiler *compiler,
95 nir_shader *nir,
96 const nir_shader *softfp64);
97
98 void
99 brw_nir_link_shaders(const struct brw_compiler *compiler,
100 nir_shader *producer, nir_shader *consumer);
101
102 bool brw_nir_lower_cs_intrinsics(nir_shader *nir);
103 bool brw_nir_lower_alpha_to_coverage(nir_shader *shader);
104 void brw_nir_lower_legacy_clipping(nir_shader *nir,
105 int nr_userclip_plane_consts,
106 struct brw_stage_prog_data *prog_data);
107 void brw_nir_lower_vs_inputs(nir_shader *nir,
108 const uint8_t *vs_attrib_wa_flags);
109 void brw_nir_lower_vue_inputs(nir_shader *nir,
110 const struct brw_vue_map *vue_map);
111 void brw_nir_lower_tes_inputs(nir_shader *nir, const struct brw_vue_map *vue);
112 void brw_nir_lower_fs_inputs(nir_shader *nir,
113 const struct gen_device_info *devinfo,
114 const struct brw_wm_prog_key *key);
115 void brw_nir_lower_vue_outputs(nir_shader *nir);
116 void brw_nir_lower_tcs_outputs(nir_shader *nir, const struct brw_vue_map *vue,
117 GLenum tes_primitive_mode);
118 void brw_nir_lower_fs_outputs(nir_shader *nir);
119
120 bool brw_nir_lower_conversions(nir_shader *nir);
121
122 bool brw_nir_lower_scoped_barriers(nir_shader *nir);
123
124 bool brw_nir_lower_image_load_store(nir_shader *nir,
125 const struct gen_device_info *devinfo,
126 bool *uses_atomic_load_store);
127 void brw_nir_rewrite_image_intrinsic(nir_intrinsic_instr *intrin,
128 nir_ssa_def *index);
129 void brw_nir_rewrite_bindless_image_intrinsic(nir_intrinsic_instr *intrin,
130 nir_ssa_def *handle);
131
132 bool brw_nir_lower_mem_access_bit_sizes(nir_shader *shader,
133 const struct gen_device_info *devinfo);
134
135 void brw_postprocess_nir(nir_shader *nir,
136 const struct brw_compiler *compiler,
137 bool is_scalar);
138
139 bool brw_nir_clamp_image_1d_2d_array_sizes(nir_shader *shader);
140
141 bool brw_nir_apply_attribute_workarounds(nir_shader *nir,
142 const uint8_t *attrib_wa_flags);
143
144 bool brw_nir_apply_trig_workarounds(nir_shader *nir);
145
146 void brw_nir_apply_tcs_quads_workaround(nir_shader *nir);
147
148 void brw_nir_apply_key(nir_shader *nir,
149 const struct brw_compiler *compiler,
150 const struct brw_base_prog_key *key,
151 unsigned max_subgroup_size,
152 bool is_scalar);
153
154 enum brw_conditional_mod brw_cmod_for_nir_comparison(nir_op op);
155 uint32_t brw_aop_for_nir_intrinsic(const nir_intrinsic_instr *atomic);
156 enum brw_reg_type brw_type_for_nir_type(const struct gen_device_info *devinfo,
157 nir_alu_type type);
158
159 enum glsl_base_type brw_glsl_base_type_for_nir_type(nir_alu_type type);
160
161 void brw_nir_setup_glsl_uniforms(void *mem_ctx, nir_shader *shader,
162 const struct gl_program *prog,
163 struct brw_stage_prog_data *stage_prog_data,
164 bool is_scalar);
165
166 void brw_nir_setup_arb_uniforms(void *mem_ctx, nir_shader *shader,
167 struct gl_program *prog,
168 struct brw_stage_prog_data *stage_prog_data);
169
170 void brw_nir_lower_gl_images(nir_shader *shader,
171 const struct gl_program *prog);
172
173 void brw_nir_analyze_ubo_ranges(const struct brw_compiler *compiler,
174 nir_shader *nir,
175 const struct brw_vs_prog_key *vs_key,
176 struct brw_ubo_range out_ranges[4]);
177
178 bool brw_nir_opt_peephole_ffma(nir_shader *shader);
179
180 void brw_nir_optimize(nir_shader *nir,
181 const struct brw_compiler *compiler,
182 bool is_scalar,
183 bool allow_copies);
184
185 nir_shader *brw_nir_create_passthrough_tcs(void *mem_ctx,
186 const struct brw_compiler *compiler,
187 const nir_shader_compiler_options *options,
188 const struct brw_tcs_prog_key *key);
189
190 #define BRW_NIR_FRAG_OUTPUT_INDEX_SHIFT 0
191 #define BRW_NIR_FRAG_OUTPUT_INDEX_MASK INTEL_MASK(0, 0)
192 #define BRW_NIR_FRAG_OUTPUT_LOCATION_SHIFT 1
193 #define BRW_NIR_FRAG_OUTPUT_LOCATION_MASK INTEL_MASK(31, 1)
194
195 bool brw_nir_move_interpolation_to_top(nir_shader *nir);
196 bool brw_nir_demote_sample_qualifiers(nir_shader *nir);
197 void brw_nir_populate_wm_prog_data(const nir_shader *shader,
198 const struct gen_device_info *devinfo,
199 const struct brw_wm_prog_key *key,
200 struct brw_wm_prog_data *prog_data);
201
202 #ifdef __cplusplus
203 }
204 #endif
205
206 #endif /* BRW_NIR_H */
207