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_scalar(const struct glsl_type *type); 36 int type_size_vec4(const struct glsl_type *type); 37 int type_size_dvec4(const struct glsl_type *type); 38 39 static inline int 40 type_size_scalar_bytes(const struct glsl_type *type) 41 { 42 return type_size_scalar(type) * 4; 43 } 44 45 static inline int 46 type_size_vec4_bytes(const struct glsl_type *type) 47 { 48 return type_size_vec4(type) * 16; 49 } 50 51 /* Flags set in the instr->pass_flags field by i965 analysis passes */ 52 enum { 53 BRW_NIR_NON_BOOLEAN = 0x0, 54 55 /* Indicates that the given instruction's destination is a boolean 56 * value but that it needs to be resolved before it can be used. 57 * On Gen <= 5, CMP instructions return a 32-bit value where the bottom 58 * bit represents the actual true/false value of the compare and the top 59 * 31 bits are undefined. In order to use this value, we have to do a 60 * "resolve" operation by replacing the value of the CMP with -(x & 1) 61 * to sign-extend the bottom bit to 0/~0. 62 */ 63 BRW_NIR_BOOLEAN_NEEDS_RESOLVE = 0x1, 64 65 /* Indicates that the given instruction's destination is a boolean 66 * value that has intentionally been left unresolved. Not all boolean 67 * values need to be resolved immediately. For instance, if we have 68 * 69 * CMP r1 r2 r3 70 * CMP r4 r5 r6 71 * AND r7 r1 r4 72 * 73 * We don't have to resolve the result of the two CMP instructions 74 * immediately because the AND still does an AND of the bottom bits. 75 * Instead, we can save ourselves instructions by delaying the resolve 76 * until after the AND. The result of the two CMP instructions is left 77 * as BRW_NIR_BOOLEAN_UNRESOLVED. 78 */ 79 BRW_NIR_BOOLEAN_UNRESOLVED = 0x2, 80 81 /* Indicates a that the given instruction's destination is a boolean 82 * value that does not need a resolve. For instance, if you AND two 83 * values that are BRW_NIR_BOOLEAN_NEEDS_RESOLVE then we know that both 84 * values will be 0/~0 before we get them and the result of the AND is 85 * also guaranteed to be 0/~0 and does not need a resolve. 86 */ 87 BRW_NIR_BOOLEAN_NO_RESOLVE = 0x3, 88 89 /* A mask to mask the boolean status values off of instr->pass_flags */ 90 BRW_NIR_BOOLEAN_MASK = 0x3, 91 }; 92 93 void brw_nir_analyze_boolean_resolves(nir_shader *nir); 94 95 nir_shader *brw_preprocess_nir(const struct brw_compiler *compiler, 96 nir_shader *nir); 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 unsigned dispatch_width); 104 void brw_nir_lower_vs_inputs(nir_shader *nir, 105 const uint8_t *vs_attrib_wa_flags); 106 void brw_nir_lower_vue_inputs(nir_shader *nir, 107 const struct brw_vue_map *vue_map); 108 void brw_nir_lower_tes_inputs(nir_shader *nir, const struct brw_vue_map *vue); 109 void brw_nir_lower_fs_inputs(nir_shader *nir, 110 const struct gen_device_info *devinfo, 111 const struct brw_wm_prog_key *key); 112 void brw_nir_lower_vue_outputs(nir_shader *nir, bool is_scalar); 113 void brw_nir_lower_tcs_outputs(nir_shader *nir, const struct brw_vue_map *vue, 114 GLenum tes_primitive_mode); 115 void brw_nir_lower_fs_outputs(nir_shader *nir); 116 void brw_nir_lower_cs_shared(nir_shader *nir); 117 118 nir_shader *brw_postprocess_nir(nir_shader *nir, 119 const struct brw_compiler *compiler, 120 bool is_scalar); 121 122 bool brw_nir_apply_attribute_workarounds(nir_shader *nir, 123 const uint8_t *attrib_wa_flags); 124 125 bool brw_nir_apply_trig_workarounds(nir_shader *nir); 126 127 void brw_nir_apply_tcs_quads_workaround(nir_shader *nir); 128 129 nir_shader *brw_nir_apply_sampler_key(nir_shader *nir, 130 const struct brw_compiler *compiler, 131 const struct brw_sampler_prog_key_data *key, 132 bool is_scalar); 133 134 enum brw_reg_type brw_type_for_nir_type(const struct gen_device_info *devinfo, 135 nir_alu_type type); 136 137 enum glsl_base_type brw_glsl_base_type_for_nir_type(nir_alu_type type); 138 139 void brw_nir_setup_glsl_uniforms(void *mem_ctx, nir_shader *shader, 140 const struct gl_program *prog, 141 struct brw_stage_prog_data *stage_prog_data, 142 bool is_scalar); 143 144 void brw_nir_setup_arb_uniforms(void *mem_ctx, nir_shader *shader, 145 struct gl_program *prog, 146 struct brw_stage_prog_data *stage_prog_data); 147 148 void brw_nir_lower_patch_vertices_in_to_uniform(nir_shader *nir); 149 150 void brw_nir_analyze_ubo_ranges(const struct brw_compiler *compiler, 151 nir_shader *nir, 152 struct brw_ubo_range out_ranges[4]); 153 154 bool brw_nir_opt_peephole_ffma(nir_shader *shader); 155 156 nir_shader *brw_nir_optimize(nir_shader *nir, 157 const struct brw_compiler *compiler, 158 bool is_scalar); 159 160 #define BRW_NIR_FRAG_OUTPUT_INDEX_SHIFT 0 161 #define BRW_NIR_FRAG_OUTPUT_INDEX_MASK INTEL_MASK(0, 0) 162 #define BRW_NIR_FRAG_OUTPUT_LOCATION_SHIFT 1 163 #define BRW_NIR_FRAG_OUTPUT_LOCATION_MASK INTEL_MASK(31, 1) 164 165 #ifdef __cplusplus 166 } 167 #endif 168 169 #endif /* BRW_NIR_H */ 170