2 * Copyright 2012 Advanced Micro Devices, Inc.
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 * on the rights to use, copy, modify, merge, publish, distribute, sub
8 * license, and/or sell copies of the Software, and to permit persons to whom
9 * the 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 NON-INFRINGEMENT. IN NO EVENT SHALL
18 * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
19 * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
20 * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
21 * USE OR OTHER DEALINGS IN THE SOFTWARE.
24 * Tom Stellard <thomas.stellard@amd.com>
25 * Michel Dänzer <michel.daenzer@amd.com>
26 * Christian König <christian.koenig@amd.com>
29 /* The compiler middle-end architecture: Explaining (non-)monolithic shaders
30 * -------------------------------------------------------------------------
32 * Typically, there is one-to-one correspondence between API and HW shaders,
33 * that is, for every API shader, there is exactly one shader binary in
36 * The problem with that is that we also have to emulate some API states
37 * (e.g. alpha-test, and many others) in shaders too. The two obvious ways
38 * to deal with it are:
39 * - each shader has multiple variants for each combination of emulated states,
40 * and the variants are compiled on demand, possibly relying on a shader
41 * cache for good performance
42 * - patch shaders at the binary level
44 * This driver uses something completely different. The emulated states are
45 * usually implemented at the beginning or end of shaders. Therefore, we can
46 * split the shader into 3 parts:
47 * - prolog part (shader code dependent on states)
48 * - main part (the API shader)
49 * - epilog part (shader code dependent on states)
51 * Each part is compiled as a separate shader and the final binaries are
52 * concatenated. This type of shader is called non-monolithic, because it
53 * consists of multiple independent binaries. Creating a new shader variant
54 * is therefore only a concatenation of shader parts (binaries) and doesn't
55 * involve any compilation. The main shader parts are the only parts that are
56 * compiled when applications create shader objects. The prolog and epilog
57 * parts are compiled on the first use and saved, so that their binaries can
58 * be reused by many other shaders.
60 * One of the roles of the prolog part is to compute vertex buffer addresses
61 * for vertex shaders. A few of the roles of the epilog part are color buffer
62 * format conversions in pixel shaders that we have to do manually, and write
63 * tessellation factors in tessellation control shaders. The prolog and epilog
64 * have many other important responsibilities in various shader stages.
65 * They don't just "emulate legacy stuff".
67 * Monolithic shaders are shaders where the parts are combined before LLVM
68 * compilation, and the whole thing is compiled and optimized as one unit with
69 * one binary on the output. The result is the same as the non-monolithic
70 * shader, but the final code can be better, because LLVM can optimize across
71 * all shader parts. Monolithic shaders aren't usually used except for these
74 * 1) Some rarely-used states require modification of the main shader part
75 * itself, and in such cases, only the monolithic shader variant is
76 * compiled, and that's always done on the first use.
78 * 2) When we do cross-stage optimizations for separate shader objects and
79 * e.g. eliminate unused shader varyings, the resulting optimized shader
80 * variants are always compiled as monolithic shaders, and always
81 * asynchronously (i.e. not stalling ongoing rendering). We call them
82 * "optimized monolithic" shaders. The important property here is that
83 * the non-monolithic unoptimized shader variant is always available for use
84 * when the asynchronous compilation of the optimized shader is not done
87 * Starting with GFX9 chips, some shader stages are merged, and the number of
88 * shader parts per shader increased. The complete new list of shader parts is:
89 * - 1st shader: prolog part
90 * - 1st shader: main part
91 * - 2nd shader: prolog part
92 * - 2nd shader: main part
93 * - 2nd shader: epilog part
96 /* How linking shader inputs and outputs between vertex, tessellation, and
97 * geometry shaders works.
99 * Inputs and outputs between shaders are stored in a buffer. This buffer
100 * lives in LDS (typical case for tessellation), but it can also live
101 * in memory (ESGS). Each input or output has a fixed location within a vertex.
102 * The highest used input or output determines the stride between vertices.
104 * Since GS and tessellation are only possible in the OpenGL core profile,
105 * only these semantics are valid for per-vertex data:
112 * CULLDIST0..1 (not implemented)
115 * For example, a shader only writing GENERIC0 has the output stride of 5.
117 * Only these semantics are valid for per-patch data:
125 * That's how independent shaders agree on input and output locations.
126 * The si_shader_io_get_unique_index function assigns the locations.
128 * For tessellation, other required information for calculating the input and
129 * output addresses like the vertex stride, the patch stride, and the offsets
130 * where per-vertex and per-patch data start, is passed to the shader via
131 * user data SGPRs. The offsets and strides are calculated at draw time and
132 * aren't available at compile time.
138 #include <llvm-c/Core.h> /* LLVMModuleRef */
139 #include <llvm-c/TargetMachine.h>
140 #include "tgsi/tgsi_scan.h"
141 #include "util/u_queue.h"
142 #include "si_state.h"
144 struct ac_shader_binary
;
146 #define SI_MAX_VS_OUTPUTS 40
148 /* SGPR user data indices */
150 /* GFX9 merged shaders have RW_BUFFERS among the first 8 system SGPRs,
151 * and these two are used for other purposes.
153 SI_SGPR_RW_BUFFERS
, /* rings (& stream-out, VS only) */
154 SI_SGPR_RW_BUFFERS_HI
,
155 SI_SGPR_CONST_BUFFERS
,
156 SI_SGPR_CONST_BUFFERS_HI
,
157 SI_SGPR_SAMPLERS
, /* images & sampler states interleaved */
161 SI_SGPR_SHADER_BUFFERS
,
162 SI_SGPR_SHADER_BUFFERS_HI
,
163 SI_NUM_RESOURCE_SGPRS
,
165 /* all VS variants */
166 SI_SGPR_VERTEX_BUFFERS
= SI_NUM_RESOURCE_SGPRS
,
167 SI_SGPR_VERTEX_BUFFERS_HI
,
169 SI_SGPR_START_INSTANCE
,
171 SI_SGPR_VS_STATE_BITS
,
175 SI_SGPR_TES_OFFCHIP_LAYOUT
= SI_NUM_RESOURCE_SGPRS
,
176 SI_TES_NUM_USER_SGPR
,
178 /* GFX6-8: TCS only */
179 GFX6_SGPR_TCS_OFFCHIP_LAYOUT
= SI_NUM_RESOURCE_SGPRS
,
180 GFX6_SGPR_TCS_OUT_OFFSETS
,
181 GFX6_SGPR_TCS_OUT_LAYOUT
,
182 GFX6_SGPR_TCS_IN_LAYOUT
,
183 GFX6_TCS_NUM_USER_SGPR
,
185 /* GFX9: Merged LS-HS (VS-TCS) only. */
186 GFX9_SGPR_TCS_OFFCHIP_LAYOUT
= SI_VS_NUM_USER_SGPR
,
187 GFX9_SGPR_TCS_OUT_OFFSETS
,
188 GFX9_SGPR_TCS_OUT_LAYOUT
,
189 GFX9_SGPR_unused_to_align_the_next_pointer
,
190 GFX9_SGPR_TCS_CONST_BUFFERS
,
191 GFX9_SGPR_TCS_CONST_BUFFERS_HI
,
192 GFX9_SGPR_TCS_SAMPLERS
, /* images & sampler states interleaved */
193 GFX9_SGPR_TCS_SAMPLERS_HI
,
194 GFX9_SGPR_TCS_IMAGES
,
195 GFX9_SGPR_TCS_IMAGES_HI
,
196 GFX9_SGPR_TCS_SHADER_BUFFERS
,
197 GFX9_SGPR_TCS_SHADER_BUFFERS_HI
,
198 GFX9_TCS_NUM_USER_SGPR
,
200 /* GFX9: Merged ES-GS (VS-GS or TES-GS). */
201 GFX9_SGPR_GS_CONST_BUFFERS
= SI_VS_NUM_USER_SGPR
,
202 GFX9_SGPR_GS_CONST_BUFFERS_HI
,
203 GFX9_SGPR_GS_SAMPLERS
,
204 GFX9_SGPR_GS_SAMPLERS_HI
,
206 GFX9_SGPR_GS_IMAGES_HI
,
207 GFX9_SGPR_GS_SHADER_BUFFERS
,
208 GFX9_SGPR_GS_SHADER_BUFFERS_HI
,
209 GFX9_GS_NUM_USER_SGPR
,
212 GFX6_GS_NUM_USER_SGPR
= SI_NUM_RESOURCE_SGPRS
,
213 SI_GSCOPY_NUM_USER_SGPR
= SI_SGPR_RW_BUFFERS_HI
+ 1,
216 SI_SGPR_ALPHA_REF
= SI_NUM_RESOURCE_SGPRS
,
220 SI_SGPR_GRID_SIZE
= SI_NUM_RESOURCE_SGPRS
,
221 SI_SGPR_BLOCK_SIZE
= SI_SGPR_GRID_SIZE
+ 3,
222 SI_CS_NUM_USER_SGPR
= SI_SGPR_BLOCK_SIZE
+ 3
225 /* LLVM function parameter indices */
227 SI_NUM_RESOURCE_PARAMS
= 5,
229 /* PS only parameters */
230 SI_PARAM_ALPHA_REF
= SI_NUM_RESOURCE_PARAMS
,
232 SI_PARAM_PERSP_SAMPLE
,
233 SI_PARAM_PERSP_CENTER
,
234 SI_PARAM_PERSP_CENTROID
,
235 SI_PARAM_PERSP_PULL_MODEL
,
236 SI_PARAM_LINEAR_SAMPLE
,
237 SI_PARAM_LINEAR_CENTER
,
238 SI_PARAM_LINEAR_CENTROID
,
239 SI_PARAM_LINE_STIPPLE_TEX
,
240 SI_PARAM_POS_X_FLOAT
,
241 SI_PARAM_POS_Y_FLOAT
,
242 SI_PARAM_POS_Z_FLOAT
,
243 SI_PARAM_POS_W_FLOAT
,
246 SI_PARAM_SAMPLE_COVERAGE
,
247 SI_PARAM_POS_FIXED_PT
,
249 /* CS only parameters */
250 SI_PARAM_GRID_SIZE
= SI_NUM_RESOURCE_PARAMS
,
255 SI_NUM_PARAMS
= SI_PARAM_POS_FIXED_PT
+ 9, /* +8 for COLOR[0..1] */
258 /* Fields of driver-defined VS state SGPR. */
259 /* Clamp vertex color output (only used in VS as VS). */
260 #define S_VS_STATE_CLAMP_VERTEX_COLOR(x) (((unsigned)(x) & 0x1) << 0)
261 #define C_VS_STATE_CLAMP_VERTEX_COLOR 0xFFFFFFFE
262 #define S_VS_STATE_INDEXED(x) (((unsigned)(x) & 0x1) << 1)
263 #define C_VS_STATE_INDEXED 0xFFFFFFFD
264 #define S_VS_STATE_LS_OUT_PATCH_SIZE(x) (((unsigned)(x) & 0x1FFF) << 8)
265 #define C_VS_STATE_LS_OUT_PATCH_SIZE 0xFFE000FF
266 #define S_VS_STATE_LS_OUT_VERTEX_SIZE(x) (((unsigned)(x) & 0xFF) << 24)
267 #define C_VS_STATE_LS_OUT_VERTEX_SIZE 0x00FFFFFF
269 /* SI-specific system values. */
271 TGSI_SEMANTIC_DEFAULT_TESSOUTER_SI
= TGSI_SEMANTIC_COUNT
,
272 TGSI_SEMANTIC_DEFAULT_TESSINNER_SI
,
275 /* For VS shader key fix_fetch. */
277 SI_FIX_FETCH_NONE
= 0,
278 SI_FIX_FETCH_A2_SNORM
,
279 SI_FIX_FETCH_A2_SSCALED
,
280 SI_FIX_FETCH_A2_SINT
,
281 SI_FIX_FETCH_RGBA_32_UNORM
,
282 SI_FIX_FETCH_RGBX_32_UNORM
,
283 SI_FIX_FETCH_RGBA_32_SNORM
,
284 SI_FIX_FETCH_RGBX_32_SNORM
,
285 SI_FIX_FETCH_RGBA_32_USCALED
,
286 SI_FIX_FETCH_RGBA_32_SSCALED
,
287 SI_FIX_FETCH_RGBA_32_FIXED
,
288 SI_FIX_FETCH_RGBX_32_FIXED
,
289 SI_FIX_FETCH_RG_64_FLOAT
,
290 SI_FIX_FETCH_RGB_64_FLOAT
,
291 SI_FIX_FETCH_RGBA_64_FLOAT
,
292 SI_FIX_FETCH_RGB_8
, /* A = 1.0 */
293 SI_FIX_FETCH_RGB_8_INT
, /* A = 1 */
295 SI_FIX_FETCH_RGB_16_INT
,
300 /* State of the context creating the shader object. */
301 struct si_compiler_ctx_state
{
302 /* Should only be used by si_init_shader_selector_async and
303 * si_build_shader_variant if thread_index == -1 (non-threaded). */
304 LLVMTargetMachineRef tm
;
306 /* Used if thread_index == -1 or if debug.async is true. */
307 struct pipe_debug_callback debug
;
309 /* Used for creating the log string for gallium/ddebug. */
310 bool is_debug_context
;
313 /* A shader selector is a gallium CSO and contains shader variants and
314 * binaries for one TGSI program. This can be shared by multiple contexts.
316 struct si_shader_selector
{
317 struct pipe_reference reference
;
318 struct si_screen
*screen
;
319 struct util_queue_fence ready
;
320 struct si_compiler_ctx_state compiler_ctx_state
;
323 struct si_shader
*first_variant
; /* immutable after the first variant */
324 struct si_shader
*last_variant
; /* mutable */
326 /* The compiled TGSI shader expecting a prolog and/or epilog (not
327 * uploaded to a buffer).
329 struct si_shader
*main_shader_part
;
330 struct si_shader
*main_shader_part_ls
; /* as_ls is set in the key */
331 struct si_shader
*main_shader_part_es
; /* as_es is set in the key */
333 struct si_shader
*gs_copy_shader
;
335 struct tgsi_token
*tokens
;
336 struct pipe_stream_output_info so
;
337 struct tgsi_shader_info info
;
339 /* PIPE_SHADER_[VERTEX|FRAGMENT|...] */
341 bool vs_needs_prolog
;
344 unsigned esgs_itemsize
;
345 unsigned gs_input_verts_per_prim
;
346 unsigned gs_output_prim
;
347 unsigned gs_max_out_vertices
;
348 unsigned gs_num_invocations
;
349 unsigned max_gs_stream
; /* count - 1 */
350 unsigned gsvs_vertex_size
;
351 unsigned max_gsvs_emit_size
;
354 unsigned color_attr_index
[2];
355 unsigned db_shader_control
;
356 /* Set 0xf or 0x0 (4 bits) per each written output.
357 * ANDed with spi_shader_col_format.
359 unsigned colors_written_4bit
;
364 uint64_t outputs_written
; /* "get_unique_index" bits */
365 uint32_t patch_outputs_written
; /* "get_unique_index" bits */
366 uint32_t outputs_written2
; /* "get_unique_index2" bits */
368 uint64_t inputs_read
; /* "get_unique_index" bits */
369 uint32_t inputs_read2
; /* "get_unique_index2" bits */
372 /* Valid shader configurations:
374 * API shaders VS | TCS | TES | GS |pass| PS
375 * are compiled as: | | | |thru|
377 * Only VS & PS: VS | | | | | PS
378 * GFX6 - with GS: ES | | | GS | VS | PS
379 * - with tess: LS | HS | VS | | | PS
380 * - with both: LS | HS | ES | GS | VS | PS
381 * GFX9 - with GS: -> | | | GS | VS | PS
382 * - with tess: -> | HS | VS | | | PS
383 * - with both: -> | HS | -> | GS | VS | PS
385 * -> = merged with the next stage
388 /* Common VS bits between the shader key and the prolog key. */
389 struct si_vs_prolog_bits
{
390 unsigned instance_divisors
[SI_MAX_ATTRIBS
];
393 /* Common VS bits between the shader key and the epilog key. */
394 struct si_vs_epilog_bits
{
395 unsigned export_prim_id
:1; /* when PS needs it and GS is disabled */
398 /* Common TCS bits between the shader key and the epilog key. */
399 struct si_tcs_epilog_bits
{
400 unsigned prim_mode
:3;
401 unsigned tes_reads_tess_factors
:1;
404 struct si_gs_prolog_bits
{
405 unsigned tri_strip_adj_fix
:1;
408 /* Common PS bits between the shader key and the prolog key. */
409 struct si_ps_prolog_bits
{
410 unsigned color_two_side
:1;
411 unsigned flatshade_colors
:1;
412 unsigned poly_stipple
:1;
413 unsigned force_persp_sample_interp
:1;
414 unsigned force_linear_sample_interp
:1;
415 unsigned force_persp_center_interp
:1;
416 unsigned force_linear_center_interp
:1;
417 unsigned bc_optimize_for_persp
:1;
418 unsigned bc_optimize_for_linear
:1;
421 /* Common PS bits between the shader key and the epilog key. */
422 struct si_ps_epilog_bits
{
423 unsigned spi_shader_col_format
;
424 unsigned color_is_int8
:8;
425 unsigned color_is_int10
:8;
426 unsigned last_cbuf
:3;
427 unsigned alpha_func
:3;
428 unsigned alpha_to_one
:1;
429 unsigned poly_line_smoothing
:1;
430 unsigned clamp_color
:1;
433 union si_shader_part_key
{
435 struct si_vs_prolog_bits states
;
436 unsigned num_input_sgprs
:6;
437 /* For merged stages such as LS-HS, HS input VGPRs are first. */
438 unsigned num_merged_next_stage_vgprs
:3;
439 unsigned last_input
:4;
441 /* Prologs for monolithic shaders shouldn't set EXEC. */
442 unsigned is_monolithic
:1;
445 struct si_vs_epilog_bits states
;
446 unsigned prim_id_param_offset
:5;
449 struct si_tcs_epilog_bits states
;
452 struct si_gs_prolog_bits states
;
453 /* Prologs of monolithic shaders shouldn't set EXEC. */
454 unsigned is_monolithic
:1;
457 struct si_ps_prolog_bits states
;
458 unsigned num_input_sgprs
:6;
459 unsigned num_input_vgprs
:5;
460 /* Color interpolation and two-side color selection. */
461 unsigned colors_read
:8; /* color input components read */
462 unsigned num_interp_inputs
:5; /* BCOLOR is at this location */
463 unsigned face_vgpr_index
:5;
465 char color_attr_index
[2];
466 char color_interp_vgpr_index
[2]; /* -1 == constant */
469 struct si_ps_epilog_bits states
;
470 unsigned colors_written
:8;
472 unsigned writes_stencil
:1;
473 unsigned writes_samplemask
:1;
477 struct si_shader_key
{
478 /* Prolog and epilog flags. */
481 struct si_vs_prolog_bits prolog
;
482 struct si_vs_epilog_bits epilog
;
485 struct si_vs_prolog_bits ls_prolog
; /* for merged LS-HS */
486 struct si_shader_selector
*ls
; /* for merged LS-HS */
487 struct si_tcs_epilog_bits epilog
;
488 } tcs
; /* tessellation control shader */
490 struct si_vs_epilog_bits epilog
; /* same as VS */
491 } tes
; /* tessellation evaluation shader */
493 struct si_vs_prolog_bits vs_prolog
; /* for merged ES-GS */
494 struct si_shader_selector
*es
; /* for merged ES-GS */
495 struct si_gs_prolog_bits prolog
;
498 struct si_ps_prolog_bits prolog
;
499 struct si_ps_epilog_bits epilog
;
503 /* These two are initially set according to the NEXT_SHADER property,
504 * or guessed if the property doesn't seem correct.
506 unsigned as_es
:1; /* export shader, which precedes GS */
507 unsigned as_ls
:1; /* local shader, which precedes TCS */
509 /* Flags for monolithic compilation only. */
511 /* One byte for every input: SI_FIX_FETCH_* enums. */
512 uint8_t vs_fix_fetch
[SI_MAX_ATTRIBS
];
513 uint64_t ff_tcs_inputs_to_copy
; /* for fixed-func TCS */
516 /* Optimization flags for asynchronous compilation only. */
519 uint64_t kill_outputs
; /* "get_unique_index" bits */
520 uint32_t kill_outputs2
; /* "get_unique_index2" bits */
521 unsigned clip_disable
:1;
522 } hw_vs
; /* HW VS (it can be VS, TES, GS) */
524 /* For shaders where monolithic variants have better code.
526 * This is a flag that has no effect on code generation,
527 * but forces monolithic shaders to be used as soon as
528 * possible, because it's in the "opt" group.
530 unsigned prefer_mono
:1;
534 struct si_shader_config
{
537 unsigned spilled_sgprs
;
538 unsigned spilled_vgprs
;
539 unsigned private_mem_vgprs
;
541 unsigned spi_ps_input_ena
;
542 unsigned spi_ps_input_addr
;
544 unsigned scratch_bytes_per_wave
;
549 /* GCN-specific shader info. */
550 struct si_shader_info
{
551 ubyte vs_output_param_offset
[SI_MAX_VS_OUTPUTS
];
552 ubyte num_input_sgprs
;
553 ubyte num_input_vgprs
;
554 char face_vgpr_index
;
555 bool uses_instanceid
;
556 ubyte nr_pos_exports
;
557 ubyte nr_param_exports
;
561 struct si_compiler_ctx_state compiler_ctx_state
;
563 struct si_shader_selector
*selector
;
564 struct si_shader_selector
*previous_stage_sel
; /* for refcounting */
565 struct si_shader
*next_variant
;
567 struct si_shader_part
*prolog
;
568 struct si_shader
*previous_stage
; /* for GFX9 */
569 struct si_shader_part
*prolog2
;
570 struct si_shader_part
*epilog
;
572 struct si_pm4_state
*pm4
;
573 struct r600_resource
*bo
;
574 struct r600_resource
*scratch_bo
;
575 struct si_shader_key key
;
576 struct util_queue_fence optimized_ready
;
577 bool compilation_failed
;
580 bool is_binary_shared
;
581 bool is_gs_copy_shader
;
583 /* The following data is all that's needed for binary shaders. */
584 struct ac_shader_binary binary
;
585 struct si_shader_config config
;
586 struct si_shader_info info
;
588 /* Shader key + LLVM IR + disassembly + statistics.
589 * Generated for debug contexts only.
592 size_t shader_log_size
;
595 struct si_shader_part
{
596 struct si_shader_part
*next
;
597 union si_shader_part_key key
;
598 struct ac_shader_binary binary
;
599 struct si_shader_config config
;
604 si_generate_gs_copy_shader(struct si_screen
*sscreen
,
605 LLVMTargetMachineRef tm
,
606 struct si_shader_selector
*gs_selector
,
607 struct pipe_debug_callback
*debug
);
608 int si_compile_tgsi_shader(struct si_screen
*sscreen
,
609 LLVMTargetMachineRef tm
,
610 struct si_shader
*shader
,
612 struct pipe_debug_callback
*debug
);
613 int si_shader_create(struct si_screen
*sscreen
, LLVMTargetMachineRef tm
,
614 struct si_shader
*shader
,
615 struct pipe_debug_callback
*debug
);
616 int si_compile_llvm(struct si_screen
*sscreen
,
617 struct ac_shader_binary
*binary
,
618 struct si_shader_config
*conf
,
619 LLVMTargetMachineRef tm
,
621 struct pipe_debug_callback
*debug
,
624 void si_shader_destroy(struct si_shader
*shader
);
625 unsigned si_shader_io_get_unique_index(unsigned semantic_name
, unsigned index
);
626 unsigned si_shader_io_get_unique_index2(unsigned name
, unsigned index
);
627 int si_shader_binary_upload(struct si_screen
*sscreen
, struct si_shader
*shader
);
628 void si_shader_dump(struct si_screen
*sscreen
, struct si_shader
*shader
,
629 struct pipe_debug_callback
*debug
, unsigned processor
,
630 FILE *f
, bool check_debug_option
);
631 void si_multiwave_lds_size_workaround(struct si_screen
*sscreen
,
633 void si_shader_apply_scratch_relocs(struct si_context
*sctx
,
634 struct si_shader
*shader
,
635 struct si_shader_config
*config
,
636 uint64_t scratch_va
);
637 void si_shader_binary_read_config(struct ac_shader_binary
*binary
,
638 struct si_shader_config
*conf
,
639 unsigned symbol_offset
);
640 unsigned si_get_spi_shader_z_format(bool writes_z
, bool writes_stencil
,
641 bool writes_samplemask
);
642 const char *si_get_shader_name(struct si_shader
*shader
, unsigned processor
);
644 /* Inline helpers. */
646 /* Return the pointer to the main shader part's pointer. */
647 static inline struct si_shader
**
648 si_get_main_shader_part(struct si_shader_selector
*sel
,
649 struct si_shader_key
*key
)
652 return &sel
->main_shader_part_ls
;
654 return &sel
->main_shader_part_es
;
655 return &sel
->main_shader_part
;