2 * Copyright © 2010 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
24 #include "brw_context.h"
29 #include "brw_vec4_tes.h"
30 #include "glsl/glsl_parser_extras.h"
31 #include "main/shaderobj.h"
32 #include "main/uniforms.h"
33 #include "util/debug.h"
36 shader_debug_log_mesa(void *data
, const char *fmt
, ...)
38 struct brw_context
*brw
= (struct brw_context
*)data
;
43 _mesa_gl_vdebug(&brw
->ctx
, &msg_id
,
44 MESA_DEBUG_SOURCE_SHADER_COMPILER
,
45 MESA_DEBUG_TYPE_OTHER
,
46 MESA_DEBUG_SEVERITY_NOTIFICATION
, fmt
, args
);
51 shader_perf_log_mesa(void *data
, const char *fmt
, ...)
53 struct brw_context
*brw
= (struct brw_context
*)data
;
58 if (unlikely(INTEL_DEBUG
& DEBUG_PERF
)) {
60 va_copy(args_copy
, args
);
61 vfprintf(stderr
, fmt
, args_copy
);
65 if (brw
->perf_debug
) {
67 _mesa_gl_vdebug(&brw
->ctx
, &msg_id
,
68 MESA_DEBUG_SOURCE_SHADER_COMPILER
,
69 MESA_DEBUG_TYPE_PERFORMANCE
,
70 MESA_DEBUG_SEVERITY_MEDIUM
, fmt
, args
);
76 brw_compiler_create(void *mem_ctx
, const struct brw_device_info
*devinfo
)
78 struct brw_compiler
*compiler
= rzalloc(mem_ctx
, struct brw_compiler
);
80 compiler
->devinfo
= devinfo
;
81 compiler
->shader_debug_log
= shader_debug_log_mesa
;
82 compiler
->shader_perf_log
= shader_perf_log_mesa
;
84 brw_fs_alloc_reg_sets(compiler
);
85 brw_vec4_alloc_reg_set(compiler
);
87 compiler
->scalar_stage
[MESA_SHADER_VERTEX
] =
88 devinfo
->gen
>= 8 && !(INTEL_DEBUG
& DEBUG_VEC4VS
);
89 compiler
->scalar_stage
[MESA_SHADER_TESS_CTRL
] = false;
90 compiler
->scalar_stage
[MESA_SHADER_TESS_EVAL
] =
91 devinfo
->gen
>= 8 && env_var_as_boolean("INTEL_SCALAR_TES", true);
92 compiler
->scalar_stage
[MESA_SHADER_GEOMETRY
] =
93 devinfo
->gen
>= 8 && env_var_as_boolean("INTEL_SCALAR_GS", false);
94 compiler
->scalar_stage
[MESA_SHADER_FRAGMENT
] = true;
95 compiler
->scalar_stage
[MESA_SHADER_COMPUTE
] = true;
97 nir_shader_compiler_options
*nir_options
=
98 rzalloc(compiler
, nir_shader_compiler_options
);
99 nir_options
->native_integers
= true;
100 nir_options
->lower_fdiv
= true;
101 /* In order to help allow for better CSE at the NIR level we tell NIR
102 * to split all ffma instructions during opt_algebraic and we then
103 * re-combine them as a later step.
105 nir_options
->lower_ffma
= true;
106 nir_options
->lower_sub
= true;
107 nir_options
->lower_fdiv
= true;
108 nir_options
->lower_scmp
= true;
109 nir_options
->lower_fmod
= true;
110 nir_options
->lower_bitfield_insert
= true;
111 nir_options
->lower_uadd_carry
= true;
112 nir_options
->lower_usub_borrow
= true;
114 /* In the vec4 backend, our dpN instruction replicates its result to all
115 * the components of a vec4. We would like NIR to give us replicated fdot
116 * instructions because it can optimize better for us.
118 * For the FS backend, it should be lowered away by the scalarizing pass so
119 * we should never see fdot anyway.
121 nir_options
->fdot_replicates
= true;
123 /* We want the GLSL compiler to emit code that uses condition codes */
124 for (int i
= 0; i
< MESA_SHADER_STAGES
; i
++) {
125 compiler
->glsl_compiler_options
[i
].MaxUnrollIterations
= 32;
126 compiler
->glsl_compiler_options
[i
].MaxIfDepth
=
127 devinfo
->gen
< 6 ? 16 : UINT_MAX
;
129 compiler
->glsl_compiler_options
[i
].EmitCondCodes
= true;
130 compiler
->glsl_compiler_options
[i
].EmitNoNoise
= true;
131 compiler
->glsl_compiler_options
[i
].EmitNoMainReturn
= true;
132 compiler
->glsl_compiler_options
[i
].EmitNoIndirectInput
= true;
133 compiler
->glsl_compiler_options
[i
].EmitNoIndirectUniform
= false;
134 compiler
->glsl_compiler_options
[i
].LowerClipDistance
= true;
136 bool is_scalar
= compiler
->scalar_stage
[i
];
138 compiler
->glsl_compiler_options
[i
].EmitNoIndirectOutput
= is_scalar
;
139 compiler
->glsl_compiler_options
[i
].EmitNoIndirectTemp
= is_scalar
;
140 compiler
->glsl_compiler_options
[i
].OptimizeForAOS
= !is_scalar
;
142 /* !ARB_gpu_shader5 */
143 if (devinfo
->gen
< 7)
144 compiler
->glsl_compiler_options
[i
].EmitNoIndirectSampler
= true;
146 compiler
->glsl_compiler_options
[i
].NirOptions
= nir_options
;
148 compiler
->glsl_compiler_options
[i
].LowerBufferInterfaceBlocks
= true;
151 compiler
->glsl_compiler_options
[MESA_SHADER_TESS_CTRL
].EmitNoIndirectInput
= false;
152 compiler
->glsl_compiler_options
[MESA_SHADER_TESS_EVAL
].EmitNoIndirectInput
= false;
154 if (compiler
->scalar_stage
[MESA_SHADER_GEOMETRY
])
155 compiler
->glsl_compiler_options
[MESA_SHADER_GEOMETRY
].EmitNoIndirectInput
= false;
157 compiler
->glsl_compiler_options
[MESA_SHADER_COMPUTE
]
158 .LowerShaderSharedVariables
= true;
163 extern "C" struct gl_shader
*
164 brw_new_shader(struct gl_context
*ctx
, GLuint name
, GLuint type
)
166 struct brw_shader
*shader
;
168 shader
= rzalloc(NULL
, struct brw_shader
);
170 shader
->base
.Type
= type
;
171 shader
->base
.Stage
= _mesa_shader_enum_to_shader_stage(type
);
172 shader
->base
.Name
= name
;
173 _mesa_init_shader(ctx
, &shader
->base
);
176 return &shader
->base
;
180 brw_mark_surface_used(struct brw_stage_prog_data
*prog_data
,
183 assert(surf_index
< BRW_MAX_SURFACES
);
185 prog_data
->binding_table
.size_bytes
=
186 MAX2(prog_data
->binding_table
.size_bytes
, (surf_index
+ 1) * 4);
190 brw_type_for_base_type(const struct glsl_type
*type
)
192 switch (type
->base_type
) {
193 case GLSL_TYPE_FLOAT
:
194 return BRW_REGISTER_TYPE_F
;
197 case GLSL_TYPE_SUBROUTINE
:
198 return BRW_REGISTER_TYPE_D
;
200 return BRW_REGISTER_TYPE_UD
;
201 case GLSL_TYPE_ARRAY
:
202 return brw_type_for_base_type(type
->fields
.array
);
203 case GLSL_TYPE_STRUCT
:
204 case GLSL_TYPE_SAMPLER
:
205 case GLSL_TYPE_ATOMIC_UINT
:
206 /* These should be overridden with the type of the member when
207 * dereferenced into. BRW_REGISTER_TYPE_UD seems like a likely
208 * way to trip up if we don't.
210 return BRW_REGISTER_TYPE_UD
;
211 case GLSL_TYPE_IMAGE
:
212 return BRW_REGISTER_TYPE_UD
;
214 case GLSL_TYPE_ERROR
:
215 case GLSL_TYPE_INTERFACE
:
216 case GLSL_TYPE_DOUBLE
:
217 unreachable("not reached");
220 return BRW_REGISTER_TYPE_F
;
223 enum brw_conditional_mod
224 brw_conditional_for_comparison(unsigned int op
)
228 return BRW_CONDITIONAL_L
;
229 case ir_binop_greater
:
230 return BRW_CONDITIONAL_G
;
231 case ir_binop_lequal
:
232 return BRW_CONDITIONAL_LE
;
233 case ir_binop_gequal
:
234 return BRW_CONDITIONAL_GE
;
236 case ir_binop_all_equal
: /* same as equal for scalars */
237 return BRW_CONDITIONAL_Z
;
238 case ir_binop_nequal
:
239 case ir_binop_any_nequal
: /* same as nequal for scalars */
240 return BRW_CONDITIONAL_NZ
;
242 unreachable("not reached: bad operation for comparison");
247 brw_math_function(enum opcode op
)
250 case SHADER_OPCODE_RCP
:
251 return BRW_MATH_FUNCTION_INV
;
252 case SHADER_OPCODE_RSQ
:
253 return BRW_MATH_FUNCTION_RSQ
;
254 case SHADER_OPCODE_SQRT
:
255 return BRW_MATH_FUNCTION_SQRT
;
256 case SHADER_OPCODE_EXP2
:
257 return BRW_MATH_FUNCTION_EXP
;
258 case SHADER_OPCODE_LOG2
:
259 return BRW_MATH_FUNCTION_LOG
;
260 case SHADER_OPCODE_POW
:
261 return BRW_MATH_FUNCTION_POW
;
262 case SHADER_OPCODE_SIN
:
263 return BRW_MATH_FUNCTION_SIN
;
264 case SHADER_OPCODE_COS
:
265 return BRW_MATH_FUNCTION_COS
;
266 case SHADER_OPCODE_INT_QUOTIENT
:
267 return BRW_MATH_FUNCTION_INT_DIV_QUOTIENT
;
268 case SHADER_OPCODE_INT_REMAINDER
:
269 return BRW_MATH_FUNCTION_INT_DIV_REMAINDER
;
271 unreachable("not reached: unknown math function");
276 brw_texture_offset(int *offsets
, unsigned num_components
)
278 if (!offsets
) return 0; /* nonconstant offset; caller will handle it. */
280 /* Combine all three offsets into a single unsigned dword:
282 * bits 11:8 - U Offset (X component)
283 * bits 7:4 - V Offset (Y component)
284 * bits 3:0 - R Offset (Z component)
286 unsigned offset_bits
= 0;
287 for (unsigned i
= 0; i
< num_components
; i
++) {
288 const unsigned shift
= 4 * (2 - i
);
289 offset_bits
|= (offsets
[i
] << shift
) & (0xF << shift
);
295 brw_instruction_name(enum opcode op
)
298 case BRW_OPCODE_ILLEGAL
... BRW_OPCODE_NOP
:
299 assert(opcode_descs
[op
].name
);
300 return opcode_descs
[op
].name
;
301 case FS_OPCODE_FB_WRITE
:
303 case FS_OPCODE_FB_WRITE_LOGICAL
:
304 return "fb_write_logical";
305 case FS_OPCODE_PACK_STENCIL_REF
:
306 return "pack_stencil_ref";
307 case FS_OPCODE_BLORP_FB_WRITE
:
308 return "blorp_fb_write";
309 case FS_OPCODE_REP_FB_WRITE
:
310 return "rep_fb_write";
312 case SHADER_OPCODE_RCP
:
314 case SHADER_OPCODE_RSQ
:
316 case SHADER_OPCODE_SQRT
:
318 case SHADER_OPCODE_EXP2
:
320 case SHADER_OPCODE_LOG2
:
322 case SHADER_OPCODE_POW
:
324 case SHADER_OPCODE_INT_QUOTIENT
:
326 case SHADER_OPCODE_INT_REMAINDER
:
328 case SHADER_OPCODE_SIN
:
330 case SHADER_OPCODE_COS
:
333 case SHADER_OPCODE_TEX
:
335 case SHADER_OPCODE_TEX_LOGICAL
:
336 return "tex_logical";
337 case SHADER_OPCODE_TXD
:
339 case SHADER_OPCODE_TXD_LOGICAL
:
340 return "txd_logical";
341 case SHADER_OPCODE_TXF
:
343 case SHADER_OPCODE_TXF_LOGICAL
:
344 return "txf_logical";
345 case SHADER_OPCODE_TXL
:
347 case SHADER_OPCODE_TXL_LOGICAL
:
348 return "txl_logical";
349 case SHADER_OPCODE_TXS
:
351 case SHADER_OPCODE_TXS_LOGICAL
:
352 return "txs_logical";
355 case FS_OPCODE_TXB_LOGICAL
:
356 return "txb_logical";
357 case SHADER_OPCODE_TXF_CMS
:
359 case SHADER_OPCODE_TXF_CMS_LOGICAL
:
360 return "txf_cms_logical";
361 case SHADER_OPCODE_TXF_CMS_W
:
363 case SHADER_OPCODE_TXF_CMS_W_LOGICAL
:
364 return "txf_cms_w_logical";
365 case SHADER_OPCODE_TXF_UMS
:
367 case SHADER_OPCODE_TXF_UMS_LOGICAL
:
368 return "txf_ums_logical";
369 case SHADER_OPCODE_TXF_MCS
:
371 case SHADER_OPCODE_TXF_MCS_LOGICAL
:
372 return "txf_mcs_logical";
373 case SHADER_OPCODE_LOD
:
375 case SHADER_OPCODE_LOD_LOGICAL
:
376 return "lod_logical";
377 case SHADER_OPCODE_TG4
:
379 case SHADER_OPCODE_TG4_LOGICAL
:
380 return "tg4_logical";
381 case SHADER_OPCODE_TG4_OFFSET
:
383 case SHADER_OPCODE_TG4_OFFSET_LOGICAL
:
384 return "tg4_offset_logical";
385 case SHADER_OPCODE_SAMPLEINFO
:
388 case SHADER_OPCODE_SHADER_TIME_ADD
:
389 return "shader_time_add";
391 case SHADER_OPCODE_UNTYPED_ATOMIC
:
392 return "untyped_atomic";
393 case SHADER_OPCODE_UNTYPED_ATOMIC_LOGICAL
:
394 return "untyped_atomic_logical";
395 case SHADER_OPCODE_UNTYPED_SURFACE_READ
:
396 return "untyped_surface_read";
397 case SHADER_OPCODE_UNTYPED_SURFACE_READ_LOGICAL
:
398 return "untyped_surface_read_logical";
399 case SHADER_OPCODE_UNTYPED_SURFACE_WRITE
:
400 return "untyped_surface_write";
401 case SHADER_OPCODE_UNTYPED_SURFACE_WRITE_LOGICAL
:
402 return "untyped_surface_write_logical";
403 case SHADER_OPCODE_TYPED_ATOMIC
:
404 return "typed_atomic";
405 case SHADER_OPCODE_TYPED_ATOMIC_LOGICAL
:
406 return "typed_atomic_logical";
407 case SHADER_OPCODE_TYPED_SURFACE_READ
:
408 return "typed_surface_read";
409 case SHADER_OPCODE_TYPED_SURFACE_READ_LOGICAL
:
410 return "typed_surface_read_logical";
411 case SHADER_OPCODE_TYPED_SURFACE_WRITE
:
412 return "typed_surface_write";
413 case SHADER_OPCODE_TYPED_SURFACE_WRITE_LOGICAL
:
414 return "typed_surface_write_logical";
415 case SHADER_OPCODE_MEMORY_FENCE
:
416 return "memory_fence";
418 case SHADER_OPCODE_LOAD_PAYLOAD
:
419 return "load_payload";
421 case SHADER_OPCODE_GEN4_SCRATCH_READ
:
422 return "gen4_scratch_read";
423 case SHADER_OPCODE_GEN4_SCRATCH_WRITE
:
424 return "gen4_scratch_write";
425 case SHADER_OPCODE_GEN7_SCRATCH_READ
:
426 return "gen7_scratch_read";
427 case SHADER_OPCODE_URB_WRITE_SIMD8
:
428 return "gen8_urb_write_simd8";
429 case SHADER_OPCODE_URB_WRITE_SIMD8_PER_SLOT
:
430 return "gen8_urb_write_simd8_per_slot";
431 case SHADER_OPCODE_URB_WRITE_SIMD8_MASKED
:
432 return "gen8_urb_write_simd8_masked";
433 case SHADER_OPCODE_URB_WRITE_SIMD8_MASKED_PER_SLOT
:
434 return "gen8_urb_write_simd8_masked_per_slot";
435 case SHADER_OPCODE_URB_READ_SIMD8
:
436 return "urb_read_simd8";
437 case SHADER_OPCODE_URB_READ_SIMD8_PER_SLOT
:
438 return "urb_read_simd8_per_slot";
440 case SHADER_OPCODE_FIND_LIVE_CHANNEL
:
441 return "find_live_channel";
442 case SHADER_OPCODE_BROADCAST
:
445 case VEC4_OPCODE_MOV_BYTES
:
447 case VEC4_OPCODE_PACK_BYTES
:
449 case VEC4_OPCODE_UNPACK_UNIFORM
:
450 return "unpack_uniform";
452 case FS_OPCODE_DDX_COARSE
:
454 case FS_OPCODE_DDX_FINE
:
456 case FS_OPCODE_DDY_COARSE
:
458 case FS_OPCODE_DDY_FINE
:
461 case FS_OPCODE_CINTERP
:
463 case FS_OPCODE_LINTERP
:
466 case FS_OPCODE_PIXEL_X
:
468 case FS_OPCODE_PIXEL_Y
:
471 case FS_OPCODE_GET_BUFFER_SIZE
:
472 return "fs_get_buffer_size";
474 case FS_OPCODE_UNIFORM_PULL_CONSTANT_LOAD
:
475 return "uniform_pull_const";
476 case FS_OPCODE_UNIFORM_PULL_CONSTANT_LOAD_GEN7
:
477 return "uniform_pull_const_gen7";
478 case FS_OPCODE_VARYING_PULL_CONSTANT_LOAD
:
479 return "varying_pull_const";
480 case FS_OPCODE_VARYING_PULL_CONSTANT_LOAD_GEN7
:
481 return "varying_pull_const_gen7";
483 case FS_OPCODE_MOV_DISPATCH_TO_FLAGS
:
484 return "mov_dispatch_to_flags";
485 case FS_OPCODE_DISCARD_JUMP
:
486 return "discard_jump";
488 case FS_OPCODE_SET_SAMPLE_ID
:
489 return "set_sample_id";
490 case FS_OPCODE_SET_SIMD4X2_OFFSET
:
491 return "set_simd4x2_offset";
493 case FS_OPCODE_PACK_HALF_2x16_SPLIT
:
494 return "pack_half_2x16_split";
495 case FS_OPCODE_UNPACK_HALF_2x16_SPLIT_X
:
496 return "unpack_half_2x16_split_x";
497 case FS_OPCODE_UNPACK_HALF_2x16_SPLIT_Y
:
498 return "unpack_half_2x16_split_y";
500 case FS_OPCODE_PLACEHOLDER_HALT
:
501 return "placeholder_halt";
503 case FS_OPCODE_INTERPOLATE_AT_CENTROID
:
504 return "interp_centroid";
505 case FS_OPCODE_INTERPOLATE_AT_SAMPLE
:
506 return "interp_sample";
507 case FS_OPCODE_INTERPOLATE_AT_SHARED_OFFSET
:
508 return "interp_shared_offset";
509 case FS_OPCODE_INTERPOLATE_AT_PER_SLOT_OFFSET
:
510 return "interp_per_slot_offset";
512 case VS_OPCODE_URB_WRITE
:
513 return "vs_urb_write";
514 case VS_OPCODE_PULL_CONSTANT_LOAD
:
515 return "pull_constant_load";
516 case VS_OPCODE_PULL_CONSTANT_LOAD_GEN7
:
517 return "pull_constant_load_gen7";
519 case VS_OPCODE_SET_SIMD4X2_HEADER_GEN9
:
520 return "set_simd4x2_header_gen9";
522 case VS_OPCODE_GET_BUFFER_SIZE
:
523 return "vs_get_buffer_size";
525 case VS_OPCODE_UNPACK_FLAGS_SIMD4X2
:
526 return "unpack_flags_simd4x2";
528 case GS_OPCODE_URB_WRITE
:
529 return "gs_urb_write";
530 case GS_OPCODE_URB_WRITE_ALLOCATE
:
531 return "gs_urb_write_allocate";
532 case GS_OPCODE_THREAD_END
:
533 return "gs_thread_end";
534 case GS_OPCODE_SET_WRITE_OFFSET
:
535 return "set_write_offset";
536 case GS_OPCODE_SET_VERTEX_COUNT
:
537 return "set_vertex_count";
538 case GS_OPCODE_SET_DWORD_2
:
539 return "set_dword_2";
540 case GS_OPCODE_PREPARE_CHANNEL_MASKS
:
541 return "prepare_channel_masks";
542 case GS_OPCODE_SET_CHANNEL_MASKS
:
543 return "set_channel_masks";
544 case GS_OPCODE_GET_INSTANCE_ID
:
545 return "get_instance_id";
546 case GS_OPCODE_FF_SYNC
:
548 case GS_OPCODE_SET_PRIMITIVE_ID
:
549 return "set_primitive_id";
550 case GS_OPCODE_SVB_WRITE
:
551 return "gs_svb_write";
552 case GS_OPCODE_SVB_SET_DST_INDEX
:
553 return "gs_svb_set_dst_index";
554 case GS_OPCODE_FF_SYNC_SET_PRIMITIVES
:
555 return "gs_ff_sync_set_primitives";
556 case CS_OPCODE_CS_TERMINATE
:
557 return "cs_terminate";
558 case SHADER_OPCODE_BARRIER
:
560 case SHADER_OPCODE_MULH
:
562 case SHADER_OPCODE_MOV_INDIRECT
:
563 return "mov_indirect";
565 case VEC4_OPCODE_URB_READ
:
567 case TCS_OPCODE_GET_INSTANCE_ID
:
568 return "tcs_get_instance_id";
569 case TCS_OPCODE_URB_WRITE
:
570 return "tcs_urb_write";
571 case TCS_OPCODE_SET_INPUT_URB_OFFSETS
:
572 return "tcs_set_input_urb_offsets";
573 case TCS_OPCODE_SET_OUTPUT_URB_OFFSETS
:
574 return "tcs_set_output_urb_offsets";
575 case TCS_OPCODE_GET_PRIMITIVE_ID
:
576 return "tcs_get_primitive_id";
577 case TCS_OPCODE_CREATE_BARRIER_HEADER
:
578 return "tcs_create_barrier_header";
579 case TCS_OPCODE_SRC0_010_IS_ZERO
:
580 return "tcs_src0<0,1,0>_is_zero";
581 case TCS_OPCODE_RELEASE_INPUT
:
582 return "tcs_release_input";
583 case TCS_OPCODE_THREAD_END
:
584 return "tcs_thread_end";
585 case TES_OPCODE_CREATE_INPUT_READ_HEADER
:
586 return "tes_create_input_read_header";
587 case TES_OPCODE_ADD_INDIRECT_URB_OFFSET
:
588 return "tes_add_indirect_urb_offset";
589 case TES_OPCODE_GET_PRIMITIVE_ID
:
590 return "tes_get_primitive_id";
593 unreachable("not reached");
597 brw_saturate_immediate(enum brw_reg_type type
, struct brw_reg
*reg
)
603 } imm
= { reg
->ud
}, sat_imm
= { 0 };
606 case BRW_REGISTER_TYPE_UD
:
607 case BRW_REGISTER_TYPE_D
:
608 case BRW_REGISTER_TYPE_UW
:
609 case BRW_REGISTER_TYPE_W
:
610 case BRW_REGISTER_TYPE_UQ
:
611 case BRW_REGISTER_TYPE_Q
:
614 case BRW_REGISTER_TYPE_F
:
615 sat_imm
.f
= CLAMP(imm
.f
, 0.0f
, 1.0f
);
617 case BRW_REGISTER_TYPE_UB
:
618 case BRW_REGISTER_TYPE_B
:
619 unreachable("no UB/B immediates");
620 case BRW_REGISTER_TYPE_V
:
621 case BRW_REGISTER_TYPE_UV
:
622 case BRW_REGISTER_TYPE_VF
:
623 unreachable("unimplemented: saturate vector immediate");
624 case BRW_REGISTER_TYPE_DF
:
625 case BRW_REGISTER_TYPE_HF
:
626 unreachable("unimplemented: saturate DF/HF immediate");
629 if (imm
.ud
!= sat_imm
.ud
) {
630 reg
->ud
= sat_imm
.ud
;
637 brw_negate_immediate(enum brw_reg_type type
, struct brw_reg
*reg
)
640 case BRW_REGISTER_TYPE_D
:
641 case BRW_REGISTER_TYPE_UD
:
644 case BRW_REGISTER_TYPE_W
:
645 case BRW_REGISTER_TYPE_UW
:
646 reg
->d
= -(int16_t)reg
->ud
;
648 case BRW_REGISTER_TYPE_F
:
651 case BRW_REGISTER_TYPE_VF
:
652 reg
->ud
^= 0x80808080;
654 case BRW_REGISTER_TYPE_UB
:
655 case BRW_REGISTER_TYPE_B
:
656 unreachable("no UB/B immediates");
657 case BRW_REGISTER_TYPE_UV
:
658 case BRW_REGISTER_TYPE_V
:
659 assert(!"unimplemented: negate UV/V immediate");
660 case BRW_REGISTER_TYPE_UQ
:
661 case BRW_REGISTER_TYPE_Q
:
662 assert(!"unimplemented: negate UQ/Q immediate");
663 case BRW_REGISTER_TYPE_DF
:
664 case BRW_REGISTER_TYPE_HF
:
665 assert(!"unimplemented: negate DF/HF immediate");
672 brw_abs_immediate(enum brw_reg_type type
, struct brw_reg
*reg
)
675 case BRW_REGISTER_TYPE_D
:
676 reg
->d
= abs(reg
->d
);
678 case BRW_REGISTER_TYPE_W
:
679 reg
->d
= abs((int16_t)reg
->ud
);
681 case BRW_REGISTER_TYPE_F
:
682 reg
->f
= fabsf(reg
->f
);
684 case BRW_REGISTER_TYPE_VF
:
685 reg
->ud
&= ~0x80808080;
687 case BRW_REGISTER_TYPE_UB
:
688 case BRW_REGISTER_TYPE_B
:
689 unreachable("no UB/B immediates");
690 case BRW_REGISTER_TYPE_UQ
:
691 case BRW_REGISTER_TYPE_UD
:
692 case BRW_REGISTER_TYPE_UW
:
693 case BRW_REGISTER_TYPE_UV
:
694 /* Presumably the absolute value modifier on an unsigned source is a
695 * nop, but it would be nice to confirm.
697 assert(!"unimplemented: abs unsigned immediate");
698 case BRW_REGISTER_TYPE_V
:
699 assert(!"unimplemented: abs V immediate");
700 case BRW_REGISTER_TYPE_Q
:
701 assert(!"unimplemented: abs Q immediate");
702 case BRW_REGISTER_TYPE_DF
:
703 case BRW_REGISTER_TYPE_HF
:
704 assert(!"unimplemented: abs DF/HF immediate");
710 backend_shader::backend_shader(const struct brw_compiler
*compiler
,
713 const nir_shader
*shader
,
714 struct brw_stage_prog_data
*stage_prog_data
)
715 : compiler(compiler
),
717 devinfo(compiler
->devinfo
),
719 stage_prog_data(stage_prog_data
),
724 debug_enabled
= INTEL_DEBUG
& intel_debug_flag_for_shader_stage(stage
);
725 stage_name
= _mesa_shader_stage_to_string(stage
);
726 stage_abbrev
= _mesa_shader_stage_to_abbrev(stage
);
730 backend_reg::equals(const backend_reg
&r
) const
732 return memcmp((brw_reg
*)this, (brw_reg
*)&r
, sizeof(brw_reg
)) == 0 &&
733 reg_offset
== r
.reg_offset
;
737 backend_reg::is_zero() const
746 backend_reg::is_one() const
751 return type
== BRW_REGISTER_TYPE_F
757 backend_reg::is_negative_one() const
763 case BRW_REGISTER_TYPE_F
:
765 case BRW_REGISTER_TYPE_D
:
773 backend_reg::is_null() const
775 return file
== ARF
&& nr
== BRW_ARF_NULL
;
780 backend_reg::is_accumulator() const
782 return file
== ARF
&& nr
== BRW_ARF_ACCUMULATOR
;
786 backend_reg::in_range(const backend_reg
&r
, unsigned n
) const
788 return (file
== r
.file
&&
790 reg_offset
>= r
.reg_offset
&&
791 reg_offset
< r
.reg_offset
+ n
);
795 backend_instruction::is_commutative() const
803 case SHADER_OPCODE_MULH
:
806 /* MIN and MAX are commutative. */
807 if (conditional_mod
== BRW_CONDITIONAL_GE
||
808 conditional_mod
== BRW_CONDITIONAL_L
) {
818 backend_instruction::is_3src() const
820 return ::is_3src(opcode
);
824 backend_instruction::is_tex() const
826 return (opcode
== SHADER_OPCODE_TEX
||
827 opcode
== FS_OPCODE_TXB
||
828 opcode
== SHADER_OPCODE_TXD
||
829 opcode
== SHADER_OPCODE_TXF
||
830 opcode
== SHADER_OPCODE_TXF_CMS
||
831 opcode
== SHADER_OPCODE_TXF_CMS_W
||
832 opcode
== SHADER_OPCODE_TXF_UMS
||
833 opcode
== SHADER_OPCODE_TXF_MCS
||
834 opcode
== SHADER_OPCODE_TXL
||
835 opcode
== SHADER_OPCODE_TXS
||
836 opcode
== SHADER_OPCODE_LOD
||
837 opcode
== SHADER_OPCODE_TG4
||
838 opcode
== SHADER_OPCODE_TG4_OFFSET
);
842 backend_instruction::is_math() const
844 return (opcode
== SHADER_OPCODE_RCP
||
845 opcode
== SHADER_OPCODE_RSQ
||
846 opcode
== SHADER_OPCODE_SQRT
||
847 opcode
== SHADER_OPCODE_EXP2
||
848 opcode
== SHADER_OPCODE_LOG2
||
849 opcode
== SHADER_OPCODE_SIN
||
850 opcode
== SHADER_OPCODE_COS
||
851 opcode
== SHADER_OPCODE_INT_QUOTIENT
||
852 opcode
== SHADER_OPCODE_INT_REMAINDER
||
853 opcode
== SHADER_OPCODE_POW
);
857 backend_instruction::is_control_flow() const
861 case BRW_OPCODE_WHILE
:
863 case BRW_OPCODE_ELSE
:
864 case BRW_OPCODE_ENDIF
:
865 case BRW_OPCODE_BREAK
:
866 case BRW_OPCODE_CONTINUE
:
874 backend_instruction::can_do_source_mods() const
877 case BRW_OPCODE_ADDC
:
879 case BRW_OPCODE_BFI1
:
880 case BRW_OPCODE_BFI2
:
881 case BRW_OPCODE_BFREV
:
882 case BRW_OPCODE_CBIT
:
885 case BRW_OPCODE_SUBB
:
893 backend_instruction::can_do_saturate() const
903 case BRW_OPCODE_F16TO32
:
904 case BRW_OPCODE_F32TO16
:
905 case BRW_OPCODE_LINE
:
909 case BRW_OPCODE_MATH
:
912 case SHADER_OPCODE_MULH
:
914 case BRW_OPCODE_RNDD
:
915 case BRW_OPCODE_RNDE
:
916 case BRW_OPCODE_RNDU
:
917 case BRW_OPCODE_RNDZ
:
921 case FS_OPCODE_LINTERP
:
922 case SHADER_OPCODE_COS
:
923 case SHADER_OPCODE_EXP2
:
924 case SHADER_OPCODE_LOG2
:
925 case SHADER_OPCODE_POW
:
926 case SHADER_OPCODE_RCP
:
927 case SHADER_OPCODE_RSQ
:
928 case SHADER_OPCODE_SIN
:
929 case SHADER_OPCODE_SQRT
:
937 backend_instruction::can_do_cmod() const
941 case BRW_OPCODE_ADDC
:
946 case BRW_OPCODE_CMPN
:
951 case BRW_OPCODE_F16TO32
:
952 case BRW_OPCODE_F32TO16
:
954 case BRW_OPCODE_LINE
:
958 case BRW_OPCODE_MACH
:
965 case BRW_OPCODE_RNDD
:
966 case BRW_OPCODE_RNDE
:
967 case BRW_OPCODE_RNDU
:
968 case BRW_OPCODE_RNDZ
:
969 case BRW_OPCODE_SAD2
:
970 case BRW_OPCODE_SADA2
:
973 case BRW_OPCODE_SUBB
:
975 case FS_OPCODE_CINTERP
:
976 case FS_OPCODE_LINTERP
:
984 backend_instruction::reads_accumulator_implicitly() const
988 case BRW_OPCODE_MACH
:
989 case BRW_OPCODE_SADA2
:
997 backend_instruction::writes_accumulator_implicitly(const struct brw_device_info
*devinfo
) const
999 return writes_accumulator
||
1000 (devinfo
->gen
< 6 &&
1001 ((opcode
>= BRW_OPCODE_ADD
&& opcode
< BRW_OPCODE_NOP
) ||
1002 (opcode
>= FS_OPCODE_DDX_COARSE
&& opcode
<= FS_OPCODE_LINTERP
&&
1003 opcode
!= FS_OPCODE_CINTERP
)));
1007 backend_instruction::has_side_effects() const
1010 case SHADER_OPCODE_UNTYPED_ATOMIC
:
1011 case SHADER_OPCODE_UNTYPED_ATOMIC_LOGICAL
:
1012 case SHADER_OPCODE_GEN4_SCRATCH_WRITE
:
1013 case SHADER_OPCODE_UNTYPED_SURFACE_WRITE
:
1014 case SHADER_OPCODE_UNTYPED_SURFACE_WRITE_LOGICAL
:
1015 case SHADER_OPCODE_TYPED_ATOMIC
:
1016 case SHADER_OPCODE_TYPED_ATOMIC_LOGICAL
:
1017 case SHADER_OPCODE_TYPED_SURFACE_WRITE
:
1018 case SHADER_OPCODE_TYPED_SURFACE_WRITE_LOGICAL
:
1019 case SHADER_OPCODE_MEMORY_FENCE
:
1020 case SHADER_OPCODE_URB_WRITE_SIMD8
:
1021 case SHADER_OPCODE_URB_WRITE_SIMD8_PER_SLOT
:
1022 case SHADER_OPCODE_URB_WRITE_SIMD8_MASKED
:
1023 case SHADER_OPCODE_URB_WRITE_SIMD8_MASKED_PER_SLOT
:
1024 case FS_OPCODE_FB_WRITE
:
1025 case SHADER_OPCODE_BARRIER
:
1026 case TCS_OPCODE_RELEASE_INPUT
:
1034 backend_instruction::is_volatile() const
1037 case SHADER_OPCODE_UNTYPED_SURFACE_READ
:
1038 case SHADER_OPCODE_UNTYPED_SURFACE_READ_LOGICAL
:
1039 case SHADER_OPCODE_TYPED_SURFACE_READ
:
1040 case SHADER_OPCODE_TYPED_SURFACE_READ_LOGICAL
:
1049 inst_is_in_block(const bblock_t
*block
, const backend_instruction
*inst
)
1052 foreach_inst_in_block (backend_instruction
, i
, block
) {
1062 adjust_later_block_ips(bblock_t
*start_block
, int ip_adjustment
)
1064 for (bblock_t
*block_iter
= start_block
->next();
1065 !block_iter
->link
.is_tail_sentinel();
1066 block_iter
= block_iter
->next()) {
1067 block_iter
->start_ip
+= ip_adjustment
;
1068 block_iter
->end_ip
+= ip_adjustment
;
1073 backend_instruction::insert_after(bblock_t
*block
, backend_instruction
*inst
)
1075 if (!this->is_head_sentinel())
1076 assert(inst_is_in_block(block
, this) || !"Instruction not in block");
1080 adjust_later_block_ips(block
, 1);
1082 exec_node::insert_after(inst
);
1086 backend_instruction::insert_before(bblock_t
*block
, backend_instruction
*inst
)
1088 if (!this->is_tail_sentinel())
1089 assert(inst_is_in_block(block
, this) || !"Instruction not in block");
1093 adjust_later_block_ips(block
, 1);
1095 exec_node::insert_before(inst
);
1099 backend_instruction::insert_before(bblock_t
*block
, exec_list
*list
)
1101 assert(inst_is_in_block(block
, this) || !"Instruction not in block");
1103 unsigned num_inst
= list
->length();
1105 block
->end_ip
+= num_inst
;
1107 adjust_later_block_ips(block
, num_inst
);
1109 exec_node::insert_before(list
);
1113 backend_instruction::remove(bblock_t
*block
)
1115 assert(inst_is_in_block(block
, this) || !"Instruction not in block");
1117 adjust_later_block_ips(block
, -1);
1119 if (block
->start_ip
== block
->end_ip
) {
1120 block
->cfg
->remove_block(block
);
1125 exec_node::remove();
1129 backend_shader::dump_instructions()
1131 dump_instructions(NULL
);
1135 backend_shader::dump_instructions(const char *name
)
1137 FILE *file
= stderr
;
1138 if (name
&& geteuid() != 0) {
1139 file
= fopen(name
, "w");
1146 foreach_block_and_inst(block
, backend_instruction
, inst
, cfg
) {
1147 if (!unlikely(INTEL_DEBUG
& DEBUG_OPTIMIZER
))
1148 fprintf(file
, "%4d: ", ip
++);
1149 dump_instruction(inst
, file
);
1153 foreach_in_list(backend_instruction
, inst
, &instructions
) {
1154 if (!unlikely(INTEL_DEBUG
& DEBUG_OPTIMIZER
))
1155 fprintf(file
, "%4d: ", ip
++);
1156 dump_instruction(inst
, file
);
1160 if (file
!= stderr
) {
1166 backend_shader::calculate_cfg()
1170 cfg
= new(mem_ctx
) cfg_t(&this->instructions
);
1174 backend_shader::invalidate_cfg()
1176 ralloc_free(this->cfg
);
1181 * Sets up the starting offsets for the groups of binding table entries
1182 * commong to all pipeline stages.
1184 * Unused groups are initialized to 0xd0d0d0d0 to make it obvious that they're
1185 * unused but also make sure that addition of small offsets to them will
1186 * trigger some of our asserts that surface indices are < BRW_MAX_SURFACES.
1189 brw_assign_common_binding_table_offsets(gl_shader_stage stage
,
1190 const struct brw_device_info
*devinfo
,
1191 const struct gl_shader_program
*shader_prog
,
1192 const struct gl_program
*prog
,
1193 struct brw_stage_prog_data
*stage_prog_data
,
1194 uint32_t next_binding_table_offset
)
1196 const struct gl_shader
*shader
= NULL
;
1197 int num_textures
= _mesa_fls(prog
->SamplersUsed
);
1200 shader
= shader_prog
->_LinkedShaders
[stage
];
1202 stage_prog_data
->binding_table
.texture_start
= next_binding_table_offset
;
1203 next_binding_table_offset
+= num_textures
;
1206 assert(shader
->NumUniformBlocks
<= BRW_MAX_UBO
);
1207 stage_prog_data
->binding_table
.ubo_start
= next_binding_table_offset
;
1208 next_binding_table_offset
+= shader
->NumUniformBlocks
;
1210 assert(shader
->NumShaderStorageBlocks
<= BRW_MAX_SSBO
);
1211 stage_prog_data
->binding_table
.ssbo_start
= next_binding_table_offset
;
1212 next_binding_table_offset
+= shader
->NumShaderStorageBlocks
;
1214 stage_prog_data
->binding_table
.ubo_start
= 0xd0d0d0d0;
1215 stage_prog_data
->binding_table
.ssbo_start
= 0xd0d0d0d0;
1218 if (INTEL_DEBUG
& DEBUG_SHADER_TIME
) {
1219 stage_prog_data
->binding_table
.shader_time_start
= next_binding_table_offset
;
1220 next_binding_table_offset
++;
1222 stage_prog_data
->binding_table
.shader_time_start
= 0xd0d0d0d0;
1225 if (prog
->UsesGather
) {
1226 if (devinfo
->gen
>= 8) {
1227 stage_prog_data
->binding_table
.gather_texture_start
=
1228 stage_prog_data
->binding_table
.texture_start
;
1230 stage_prog_data
->binding_table
.gather_texture_start
= next_binding_table_offset
;
1231 next_binding_table_offset
+= num_textures
;
1234 stage_prog_data
->binding_table
.gather_texture_start
= 0xd0d0d0d0;
1237 if (shader
&& shader
->NumAtomicBuffers
) {
1238 stage_prog_data
->binding_table
.abo_start
= next_binding_table_offset
;
1239 next_binding_table_offset
+= shader
->NumAtomicBuffers
;
1241 stage_prog_data
->binding_table
.abo_start
= 0xd0d0d0d0;
1244 if (shader
&& shader
->NumImages
) {
1245 stage_prog_data
->binding_table
.image_start
= next_binding_table_offset
;
1246 next_binding_table_offset
+= shader
->NumImages
;
1248 stage_prog_data
->binding_table
.image_start
= 0xd0d0d0d0;
1251 /* This may or may not be used depending on how the compile goes. */
1252 stage_prog_data
->binding_table
.pull_constants_start
= next_binding_table_offset
;
1253 next_binding_table_offset
++;
1255 assert(next_binding_table_offset
<= BRW_MAX_SURFACES
);
1257 /* prog_data->base.binding_table.size will be set by brw_mark_surface_used. */
1261 setup_vec4_uniform_value(const gl_constant_value
**params
,
1262 const gl_constant_value
*values
,
1265 static const gl_constant_value zero
= { 0 };
1267 for (unsigned i
= 0; i
< n
; ++i
)
1268 params
[i
] = &values
[i
];
1270 for (unsigned i
= n
; i
< 4; ++i
)
1275 brw_setup_image_uniform_values(gl_shader_stage stage
,
1276 struct brw_stage_prog_data
*stage_prog_data
,
1277 unsigned param_start_index
,
1278 const gl_uniform_storage
*storage
)
1280 const gl_constant_value
**param
=
1281 &stage_prog_data
->param
[param_start_index
];
1283 for (unsigned i
= 0; i
< MAX2(storage
->array_elements
, 1); i
++) {
1284 const unsigned image_idx
= storage
->opaque
[stage
].index
+ i
;
1285 const brw_image_param
*image_param
=
1286 &stage_prog_data
->image_param
[image_idx
];
1288 /* Upload the brw_image_param structure. The order is expected to match
1289 * the BRW_IMAGE_PARAM_*_OFFSET defines.
1291 setup_vec4_uniform_value(param
+ BRW_IMAGE_PARAM_SURFACE_IDX_OFFSET
,
1292 (const gl_constant_value
*)&image_param
->surface_idx
, 1);
1293 setup_vec4_uniform_value(param
+ BRW_IMAGE_PARAM_OFFSET_OFFSET
,
1294 (const gl_constant_value
*)image_param
->offset
, 2);
1295 setup_vec4_uniform_value(param
+ BRW_IMAGE_PARAM_SIZE_OFFSET
,
1296 (const gl_constant_value
*)image_param
->size
, 3);
1297 setup_vec4_uniform_value(param
+ BRW_IMAGE_PARAM_STRIDE_OFFSET
,
1298 (const gl_constant_value
*)image_param
->stride
, 4);
1299 setup_vec4_uniform_value(param
+ BRW_IMAGE_PARAM_TILING_OFFSET
,
1300 (const gl_constant_value
*)image_param
->tiling
, 3);
1301 setup_vec4_uniform_value(param
+ BRW_IMAGE_PARAM_SWIZZLING_OFFSET
,
1302 (const gl_constant_value
*)image_param
->swizzling
, 2);
1303 param
+= BRW_IMAGE_PARAM_SIZE
;
1305 brw_mark_surface_used(
1307 stage_prog_data
->binding_table
.image_start
+ image_idx
);
1312 * Decide which set of clip planes should be used when clipping via
1313 * gl_Position or gl_ClipVertex.
1315 gl_clip_plane
*brw_select_clip_planes(struct gl_context
*ctx
)
1317 if (ctx
->_Shader
->CurrentProgram
[MESA_SHADER_VERTEX
]) {
1318 /* There is currently a GLSL vertex shader, so clip according to GLSL
1319 * rules, which means compare gl_ClipVertex (or gl_Position, if
1320 * gl_ClipVertex wasn't assigned) against the eye-coordinate clip planes
1321 * that were stored in EyeUserPlane at the time the clip planes were
1324 return ctx
->Transform
.EyeUserPlane
;
1326 /* Either we are using fixed function or an ARB vertex program. In
1327 * either case the clip planes are going to be compared against
1328 * gl_Position (which is in clip coordinates) so we have to clip using
1329 * _ClipUserPlane, which was transformed into clip coordinates by Mesa
1332 return ctx
->Transform
._ClipUserPlane
;
1336 extern "C" const unsigned *
1337 brw_compile_tes(const struct brw_compiler
*compiler
,
1340 const struct brw_tes_prog_key
*key
,
1341 struct brw_tes_prog_data
*prog_data
,
1342 const nir_shader
*src_shader
,
1343 struct gl_shader_program
*shader_prog
,
1344 int shader_time_index
,
1345 unsigned *final_assembly_size
,
1348 const struct brw_device_info
*devinfo
= compiler
->devinfo
;
1349 struct gl_shader
*shader
=
1350 shader_prog
->_LinkedShaders
[MESA_SHADER_TESS_EVAL
];
1351 const bool is_scalar
= compiler
->scalar_stage
[MESA_SHADER_TESS_EVAL
];
1353 nir_shader
*nir
= nir_shader_clone(mem_ctx
, src_shader
);
1354 nir
= brw_nir_apply_sampler_key(nir
, devinfo
, &key
->tex
, is_scalar
);
1355 nir
->info
.inputs_read
= key
->inputs_read
;
1356 nir
->info
.patch_inputs_read
= key
->patch_inputs_read
;
1357 nir
= brw_nir_lower_io(nir
, compiler
->devinfo
, is_scalar
);
1358 nir
= brw_postprocess_nir(nir
, compiler
->devinfo
, is_scalar
);
1360 brw_compute_vue_map(devinfo
, &prog_data
->base
.vue_map
,
1361 nir
->info
.outputs_written
,
1362 nir
->info
.separate_shader
);
1364 unsigned output_size_bytes
= prog_data
->base
.vue_map
.num_slots
* 4 * 4;
1366 assert(output_size_bytes
>= 1);
1367 if (output_size_bytes
> GEN7_MAX_DS_URB_ENTRY_SIZE_BYTES
) {
1369 *error_str
= ralloc_strdup(mem_ctx
, "DS outputs exceed maximum size");
1373 /* URB entry sizes are stored as a multiple of 64 bytes. */
1374 prog_data
->base
.urb_entry_size
= ALIGN(output_size_bytes
, 64) / 64;
1376 struct brw_vue_map input_vue_map
;
1377 brw_compute_tess_vue_map(&input_vue_map
,
1378 nir
->info
.inputs_read
& ~VARYING_BIT_PRIMITIVE_ID
,
1379 nir
->info
.patch_inputs_read
);
1381 bool need_patch_header
= nir
->info
.system_values_read
&
1382 (BITFIELD64_BIT(SYSTEM_VALUE_TESS_LEVEL_OUTER
) |
1383 BITFIELD64_BIT(SYSTEM_VALUE_TESS_LEVEL_INNER
));
1385 /* The TES will pull most inputs using URB read messages.
1387 * However, we push the patch header for TessLevel factors when required,
1388 * as it's a tiny amount of extra data.
1390 prog_data
->base
.urb_read_length
= need_patch_header
? 1 : 0;
1392 if (unlikely(INTEL_DEBUG
& DEBUG_TES
)) {
1393 fprintf(stderr
, "TES Input ");
1394 brw_print_vue_map(stderr
, &input_vue_map
);
1395 fprintf(stderr
, "TES Output ");
1396 brw_print_vue_map(stderr
, &prog_data
->base
.vue_map
);
1400 fs_visitor
v(compiler
, log_data
, mem_ctx
, (void *) key
,
1401 &prog_data
->base
.base
, shader
->Program
, nir
, 8,
1402 shader_time_index
, &input_vue_map
);
1405 *error_str
= ralloc_strdup(mem_ctx
, v
.fail_msg
);
1409 prog_data
->base
.dispatch_mode
= DISPATCH_MODE_SIMD8
;
1411 fs_generator
g(compiler
, log_data
, mem_ctx
, (void *) key
,
1412 &prog_data
->base
.base
, v
.promoted_constants
, false,
1414 if (unlikely(INTEL_DEBUG
& DEBUG_TES
)) {
1415 g
.enable_debug(ralloc_asprintf(mem_ctx
,
1416 "%s tessellation evaluation shader %s",
1417 nir
->info
.label
? nir
->info
.label
1422 g
.generate_code(v
.cfg
, 8);
1424 return g
.get_assembly(final_assembly_size
);
1426 brw::vec4_tes_visitor
v(compiler
, log_data
, key
, prog_data
,
1427 nir
, mem_ctx
, shader_time_index
);
1430 *error_str
= ralloc_strdup(mem_ctx
, v
.fail_msg
);
1434 if (unlikely(INTEL_DEBUG
& DEBUG_TES
))
1435 v
.dump_instructions();
1437 return brw_vec4_generate_assembly(compiler
, log_data
, mem_ctx
, nir
,
1438 &prog_data
->base
, v
.cfg
,
1439 final_assembly_size
);