radeonsi/gfx10: fix the legacy pipeline by storing as_ngg in the shader cache
[mesa.git] / src / gallium / drivers / radeonsi / si_state.h
1 /*
2 * Copyright 2012 Advanced Micro Devices, Inc.
3 * All Rights Reserved.
4 *
5 * Permission is hereby granted, free of charge, to any person obtaining a
6 * copy of this software and associated documentation files (the "Software"),
7 * to deal in the Software without restriction, including without limitation
8 * on the rights to use, copy, modify, merge, publish, distribute, sub
9 * license, and/or sell copies of the Software, and to permit persons to whom
10 * the Software is furnished to do so, subject to the following conditions:
11 *
12 * The above copyright notice and this permission notice (including the next
13 * paragraph) shall be included in all copies or substantial portions of the
14 * Software.
15 *
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
19 * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
20 * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
21 * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
22 * USE OR OTHER DEALINGS IN THE SOFTWARE.
23 */
24
25 #ifndef SI_STATE_H
26 #define SI_STATE_H
27
28 #include "si_pm4.h"
29
30 #include "pipebuffer/pb_slab.h"
31 #include "util/u_blitter.h"
32
33 #define SI_NUM_GRAPHICS_SHADERS (PIPE_SHADER_TESS_EVAL+1)
34 #define SI_NUM_SHADERS (PIPE_SHADER_COMPUTE+1)
35
36 #define SI_NUM_VERTEX_BUFFERS SI_MAX_ATTRIBS
37 #define SI_NUM_SAMPLERS 32 /* OpenGL textures units per shader */
38 #define SI_NUM_CONST_BUFFERS 16
39 #define SI_NUM_IMAGES 16
40 #define SI_NUM_SHADER_BUFFERS 16
41
42 struct si_screen;
43 struct si_shader;
44 struct si_shader_ctx_state;
45 struct si_shader_selector;
46 struct si_texture;
47 struct si_qbo_state;
48
49 struct si_state_blend {
50 struct si_pm4_state pm4;
51 uint32_t cb_target_mask;
52 /* Set 0xf or 0x0 (4 bits) per render target if the following is
53 * true. ANDed with spi_shader_col_format.
54 */
55 unsigned cb_target_enabled_4bit;
56 unsigned blend_enable_4bit;
57 unsigned need_src_alpha_4bit;
58 unsigned commutative_4bit;
59 unsigned dcc_msaa_corruption_4bit;
60 bool alpha_to_coverage:1;
61 bool alpha_to_one:1;
62 bool dual_src_blend:1;
63 bool logicop_enable:1;
64 };
65
66 struct si_state_rasterizer {
67 struct si_pm4_state pm4;
68 /* poly offset states for 16-bit, 24-bit, and 32-bit zbuffers */
69 struct si_pm4_state *pm4_poly_offset;
70 unsigned pa_sc_line_stipple;
71 unsigned pa_cl_clip_cntl;
72 float line_width;
73 float max_point_size;
74 unsigned sprite_coord_enable:8;
75 unsigned clip_plane_enable:8;
76 unsigned half_pixel_center:1;
77 unsigned flatshade:1;
78 unsigned flatshade_first:1;
79 unsigned two_side:1;
80 unsigned multisample_enable:1;
81 unsigned force_persample_interp:1;
82 unsigned line_stipple_enable:1;
83 unsigned poly_stipple_enable:1;
84 unsigned line_smooth:1;
85 unsigned poly_smooth:1;
86 unsigned uses_poly_offset:1;
87 unsigned clamp_fragment_color:1;
88 unsigned clamp_vertex_color:1;
89 unsigned rasterizer_discard:1;
90 unsigned scissor_enable:1;
91 unsigned clip_halfz:1;
92 unsigned cull_front:1;
93 unsigned cull_back:1;
94 unsigned depth_clamp_any:1;
95 unsigned provoking_vertex_first:1;
96 };
97
98 struct si_dsa_stencil_ref_part {
99 uint8_t valuemask[2];
100 uint8_t writemask[2];
101 };
102
103 struct si_dsa_order_invariance {
104 /** Whether the final result in Z/S buffers is guaranteed to be
105 * invariant under changes to the order in which fragments arrive. */
106 bool zs:1;
107
108 /** Whether the set of fragments that pass the combined Z/S test is
109 * guaranteed to be invariant under changes to the order in which
110 * fragments arrive. */
111 bool pass_set:1;
112
113 /** Whether the last fragment that passes the combined Z/S test at each
114 * sample is guaranteed to be invariant under changes to the order in
115 * which fragments arrive. */
116 bool pass_last:1;
117 };
118
119 struct si_state_dsa {
120 struct si_pm4_state pm4;
121 struct si_dsa_stencil_ref_part stencil_ref;
122
123 /* 0 = without stencil buffer, 1 = when both Z and S buffers are present */
124 struct si_dsa_order_invariance order_invariance[2];
125
126 ubyte alpha_func:3;
127 bool depth_enabled:1;
128 bool depth_write_enabled:1;
129 bool stencil_enabled:1;
130 bool stencil_write_enabled:1;
131 bool db_can_write:1;
132
133 };
134
135 struct si_stencil_ref {
136 struct pipe_stencil_ref state;
137 struct si_dsa_stencil_ref_part dsa_part;
138 };
139
140 struct si_vertex_elements
141 {
142 struct si_resource *instance_divisor_factor_buffer;
143 uint32_t rsrc_word3[SI_MAX_ATTRIBS];
144 uint16_t src_offset[SI_MAX_ATTRIBS];
145 uint8_t fix_fetch[SI_MAX_ATTRIBS];
146 uint8_t format_size[SI_MAX_ATTRIBS];
147 uint8_t vertex_buffer_index[SI_MAX_ATTRIBS];
148
149 /* Bitmask of elements that always need a fixup to be applied. */
150 uint16_t fix_fetch_always;
151
152 /* Bitmask of elements whose fetch should always be opencoded. */
153 uint16_t fix_fetch_opencode;
154
155 /* Bitmask of elements which need to be opencoded if the vertex buffer
156 * is unaligned. */
157 uint16_t fix_fetch_unaligned;
158
159 /* For elements in fix_fetch_unaligned: whether the effective
160 * element load size as seen by the hardware is a dword (as opposed
161 * to a short).
162 */
163 uint16_t hw_load_is_dword;
164
165 /* Bitmask of vertex buffers requiring alignment check */
166 uint16_t vb_alignment_check_mask;
167
168 uint8_t count;
169 bool uses_instance_divisors;
170
171 uint16_t first_vb_use_mask;
172 /* Vertex buffer descriptor list size aligned for optimal prefetch. */
173 uint16_t desc_list_byte_size;
174 uint16_t instance_divisor_is_one; /* bitmask of inputs */
175 uint16_t instance_divisor_is_fetched; /* bitmask of inputs */
176 };
177
178 union si_state {
179 struct {
180 struct si_state_blend *blend;
181 struct si_state_rasterizer *rasterizer;
182 struct si_state_dsa *dsa;
183 struct si_pm4_state *poly_offset;
184 struct si_pm4_state *ls;
185 struct si_pm4_state *hs;
186 struct si_pm4_state *es;
187 struct si_pm4_state *gs;
188 struct si_pm4_state *vgt_shader_config;
189 struct si_pm4_state *vs;
190 struct si_pm4_state *ps;
191 } named;
192 struct si_pm4_state *array[0];
193 };
194
195 #define SI_STATE_IDX(name) \
196 (offsetof(union si_state, named.name) / sizeof(struct si_pm4_state *))
197 #define SI_STATE_BIT(name) (1 << SI_STATE_IDX(name))
198 #define SI_NUM_STATES (sizeof(union si_state) / sizeof(struct si_pm4_state *))
199
200 static inline unsigned si_states_that_always_roll_context(void)
201 {
202 return (SI_STATE_BIT(blend) |
203 SI_STATE_BIT(rasterizer) |
204 SI_STATE_BIT(dsa) |
205 SI_STATE_BIT(poly_offset) |
206 SI_STATE_BIT(vgt_shader_config));
207 }
208
209 union si_state_atoms {
210 struct {
211 /* The order matters. */
212 struct si_atom render_cond;
213 struct si_atom streamout_begin;
214 struct si_atom streamout_enable; /* must be after streamout_begin */
215 struct si_atom framebuffer;
216 struct si_atom msaa_sample_locs;
217 struct si_atom db_render_state;
218 struct si_atom dpbb_state;
219 struct si_atom msaa_config;
220 struct si_atom sample_mask;
221 struct si_atom cb_render_state;
222 struct si_atom blend_color;
223 struct si_atom clip_regs;
224 struct si_atom clip_state;
225 struct si_atom shader_pointers;
226 struct si_atom guardband;
227 struct si_atom scissors;
228 struct si_atom viewports;
229 struct si_atom stencil_ref;
230 struct si_atom spi_map;
231 struct si_atom scratch_state;
232 struct si_atom window_rectangles;
233 struct si_atom shader_query;
234 } s;
235 struct si_atom array[0];
236 };
237
238 #define SI_ATOM_BIT(name) (1 << (offsetof(union si_state_atoms, s.name) / \
239 sizeof(struct si_atom)))
240 #define SI_NUM_ATOMS (sizeof(union si_state_atoms)/sizeof(struct si_atom*))
241
242 static inline unsigned si_atoms_that_always_roll_context(void)
243 {
244 return (SI_ATOM_BIT(streamout_begin) |
245 SI_ATOM_BIT(streamout_enable) |
246 SI_ATOM_BIT(framebuffer) |
247 SI_ATOM_BIT(msaa_sample_locs) |
248 SI_ATOM_BIT(sample_mask) |
249 SI_ATOM_BIT(blend_color) |
250 SI_ATOM_BIT(clip_state) |
251 SI_ATOM_BIT(scissors) |
252 SI_ATOM_BIT(viewports) |
253 SI_ATOM_BIT(stencil_ref) |
254 SI_ATOM_BIT(scratch_state) |
255 SI_ATOM_BIT(window_rectangles));
256 }
257
258 struct si_shader_data {
259 uint32_t sh_base[SI_NUM_SHADERS];
260 };
261
262 /* The list of registers whose emitted values are remembered by si_context. */
263 enum si_tracked_reg {
264 SI_TRACKED_DB_RENDER_CONTROL, /* 2 consecutive registers */
265 SI_TRACKED_DB_COUNT_CONTROL,
266
267 SI_TRACKED_DB_RENDER_OVERRIDE2,
268 SI_TRACKED_DB_SHADER_CONTROL,
269
270 SI_TRACKED_CB_TARGET_MASK,
271 SI_TRACKED_CB_DCC_CONTROL,
272
273 SI_TRACKED_SX_PS_DOWNCONVERT, /* 3 consecutive registers */
274 SI_TRACKED_SX_BLEND_OPT_EPSILON,
275 SI_TRACKED_SX_BLEND_OPT_CONTROL,
276
277 SI_TRACKED_PA_SC_LINE_CNTL, /* 2 consecutive registers */
278 SI_TRACKED_PA_SC_AA_CONFIG,
279
280 SI_TRACKED_DB_EQAA,
281 SI_TRACKED_PA_SC_MODE_CNTL_1,
282
283 SI_TRACKED_PA_SU_PRIM_FILTER_CNTL,
284 SI_TRACKED_PA_SU_SMALL_PRIM_FILTER_CNTL,
285
286 SI_TRACKED_PA_CL_VS_OUT_CNTL,
287 SI_TRACKED_PA_CL_CLIP_CNTL,
288
289 SI_TRACKED_PA_SC_BINNER_CNTL_0,
290 SI_TRACKED_DB_DFSM_CONTROL,
291
292 SI_TRACKED_PA_CL_GB_VERT_CLIP_ADJ, /* 4 consecutive registers */
293 SI_TRACKED_PA_CL_GB_VERT_DISC_ADJ,
294 SI_TRACKED_PA_CL_GB_HORZ_CLIP_ADJ,
295 SI_TRACKED_PA_CL_GB_HORZ_DISC_ADJ,
296
297 SI_TRACKED_PA_SU_HARDWARE_SCREEN_OFFSET,
298 SI_TRACKED_PA_SU_VTX_CNTL,
299
300 SI_TRACKED_PA_SC_CLIPRECT_RULE,
301
302 SI_TRACKED_VGT_ESGS_RING_ITEMSIZE,
303
304 SI_TRACKED_VGT_GSVS_RING_OFFSET_1, /* 3 consecutive registers */
305 SI_TRACKED_VGT_GSVS_RING_OFFSET_2,
306 SI_TRACKED_VGT_GSVS_RING_OFFSET_3,
307
308 SI_TRACKED_VGT_GSVS_RING_ITEMSIZE,
309 SI_TRACKED_VGT_GS_MAX_VERT_OUT,
310
311 SI_TRACKED_VGT_GS_VERT_ITEMSIZE, /* 4 consecutive registers */
312 SI_TRACKED_VGT_GS_VERT_ITEMSIZE_1,
313 SI_TRACKED_VGT_GS_VERT_ITEMSIZE_2,
314 SI_TRACKED_VGT_GS_VERT_ITEMSIZE_3,
315
316 SI_TRACKED_VGT_GS_INSTANCE_CNT,
317 SI_TRACKED_VGT_GS_ONCHIP_CNTL,
318 SI_TRACKED_VGT_GS_MAX_PRIMS_PER_SUBGROUP,
319 SI_TRACKED_VGT_GS_MODE,
320 SI_TRACKED_VGT_PRIMITIVEID_EN,
321 SI_TRACKED_VGT_REUSE_OFF,
322 SI_TRACKED_SPI_VS_OUT_CONFIG,
323 SI_TRACKED_PA_CL_VTE_CNTL,
324 SI_TRACKED_PA_CL_NGG_CNTL,
325 SI_TRACKED_GE_MAX_OUTPUT_PER_SUBGROUP,
326 SI_TRACKED_GE_NGG_SUBGRP_CNTL,
327
328 SI_TRACKED_SPI_SHADER_IDX_FORMAT, /* 2 consecutive registers */
329 SI_TRACKED_SPI_SHADER_POS_FORMAT,
330
331 SI_TRACKED_SPI_PS_INPUT_ENA, /* 2 consecutive registers */
332 SI_TRACKED_SPI_PS_INPUT_ADDR,
333
334 SI_TRACKED_SPI_BARYC_CNTL,
335 SI_TRACKED_SPI_PS_IN_CONTROL,
336
337 SI_TRACKED_SPI_SHADER_Z_FORMAT, /* 2 consecutive registers */
338 SI_TRACKED_SPI_SHADER_COL_FORMAT,
339
340 SI_TRACKED_CB_SHADER_MASK,
341 SI_TRACKED_VGT_TF_PARAM,
342 SI_TRACKED_VGT_VERTEX_REUSE_BLOCK_CNTL,
343
344 SI_NUM_TRACKED_REGS,
345 };
346
347 struct si_tracked_regs {
348 uint64_t reg_saved;
349 uint32_t reg_value[SI_NUM_TRACKED_REGS];
350 uint32_t spi_ps_input_cntl[32];
351 };
352
353 /* Private read-write buffer slots. */
354 enum {
355 SI_ES_RING_ESGS,
356 SI_GS_RING_ESGS,
357
358 SI_RING_GSVS,
359
360 SI_VS_STREAMOUT_BUF0,
361 SI_VS_STREAMOUT_BUF1,
362 SI_VS_STREAMOUT_BUF2,
363 SI_VS_STREAMOUT_BUF3,
364
365 SI_HS_CONST_DEFAULT_TESS_LEVELS,
366 SI_VS_CONST_INSTANCE_DIVISORS,
367 SI_VS_CONST_CLIP_PLANES,
368 SI_PS_CONST_POLY_STIPPLE,
369 SI_PS_CONST_SAMPLE_POSITIONS,
370
371 /* Image descriptor of color buffer 0 for KHR_blend_equation_advanced. */
372 SI_PS_IMAGE_COLORBUF0,
373 SI_PS_IMAGE_COLORBUF0_HI,
374 SI_PS_IMAGE_COLORBUF0_FMASK,
375 SI_PS_IMAGE_COLORBUF0_FMASK_HI,
376
377 GFX10_GS_QUERY_BUF,
378
379 SI_NUM_RW_BUFFERS,
380 };
381
382 /* Indices into sctx->descriptors, laid out so that gfx and compute pipelines
383 * are contiguous:
384 *
385 * 0 - rw buffers
386 * 1 - vertex const and shader buffers
387 * 2 - vertex samplers and images
388 * 3 - fragment const and shader buffer
389 * ...
390 * 11 - compute const and shader buffers
391 * 12 - compute samplers and images
392 */
393 enum {
394 SI_SHADER_DESCS_CONST_AND_SHADER_BUFFERS,
395 SI_SHADER_DESCS_SAMPLERS_AND_IMAGES,
396 SI_NUM_SHADER_DESCS,
397 };
398
399 #define SI_DESCS_RW_BUFFERS 0
400 #define SI_DESCS_FIRST_SHADER 1
401 #define SI_DESCS_FIRST_COMPUTE (SI_DESCS_FIRST_SHADER + \
402 PIPE_SHADER_COMPUTE * SI_NUM_SHADER_DESCS)
403 #define SI_NUM_DESCS (SI_DESCS_FIRST_SHADER + \
404 SI_NUM_SHADERS * SI_NUM_SHADER_DESCS)
405
406 #define SI_DESCS_SHADER_MASK(name) \
407 u_bit_consecutive(SI_DESCS_FIRST_SHADER + \
408 PIPE_SHADER_##name * SI_NUM_SHADER_DESCS, \
409 SI_NUM_SHADER_DESCS)
410
411 static inline unsigned
412 si_const_and_shader_buffer_descriptors_idx(unsigned shader)
413 {
414 return SI_DESCS_FIRST_SHADER + shader * SI_NUM_SHADER_DESCS +
415 SI_SHADER_DESCS_CONST_AND_SHADER_BUFFERS;
416 }
417
418 static inline unsigned
419 si_sampler_and_image_descriptors_idx(unsigned shader)
420 {
421 return SI_DESCS_FIRST_SHADER + shader * SI_NUM_SHADER_DESCS +
422 SI_SHADER_DESCS_SAMPLERS_AND_IMAGES;
423 }
424
425 /* This represents descriptors in memory, such as buffer resources,
426 * image resources, and sampler states.
427 */
428 struct si_descriptors {
429 /* The list of descriptors in malloc'd memory. */
430 uint32_t *list;
431 /* The list in mapped GPU memory. */
432 uint32_t *gpu_list;
433
434 /* The buffer where the descriptors have been uploaded. */
435 struct si_resource *buffer;
436 uint64_t gpu_address;
437
438 /* The maximum number of descriptors. */
439 uint32_t num_elements;
440
441 /* Slots that are used by currently-bound shaders.
442 * It determines which slots are uploaded.
443 */
444 uint32_t first_active_slot;
445 uint32_t num_active_slots;
446
447 /* The SH register offset relative to USER_DATA*_0 where the pointer
448 * to the descriptor array will be stored. */
449 short shader_userdata_offset;
450 /* The size of one descriptor. */
451 ubyte element_dw_size;
452 /* If there is only one slot enabled, bind it directly instead of
453 * uploading descriptors. -1 if disabled. */
454 signed char slot_index_to_bind_directly;
455 };
456
457 struct si_buffer_resources {
458 struct pipe_resource **buffers; /* this has num_buffers elements */
459 unsigned *offsets; /* this has num_buffers elements */
460
461 enum radeon_bo_priority priority:6;
462 enum radeon_bo_priority priority_constbuf:6;
463
464 /* The i-th bit is set if that element is enabled (non-NULL resource). */
465 unsigned enabled_mask;
466 unsigned writable_mask;
467 };
468
469 #define si_pm4_state_changed(sctx, member) \
470 ((sctx)->queued.named.member != (sctx)->emitted.named.member)
471
472 #define si_pm4_state_enabled_and_changed(sctx, member) \
473 ((sctx)->queued.named.member && si_pm4_state_changed(sctx, member))
474
475 #define si_pm4_bind_state(sctx, member, value) \
476 do { \
477 (sctx)->queued.named.member = (value); \
478 (sctx)->dirty_states |= SI_STATE_BIT(member); \
479 } while(0)
480
481 #define si_pm4_delete_state(sctx, member, value) \
482 do { \
483 if ((sctx)->queued.named.member == (value)) { \
484 (sctx)->queued.named.member = NULL; \
485 } \
486 si_pm4_free_state(sctx, (struct si_pm4_state *)(value), \
487 SI_STATE_IDX(member)); \
488 } while(0)
489
490 /* si_descriptors.c */
491 void si_set_mutable_tex_desc_fields(struct si_screen *sscreen,
492 struct si_texture *tex,
493 const struct legacy_surf_level *base_level_info,
494 unsigned base_level, unsigned first_level,
495 unsigned block_width, bool is_stencil,
496 uint32_t *state);
497 void si_update_ps_colorbuf0_slot(struct si_context *sctx);
498 void si_get_pipe_constant_buffer(struct si_context *sctx, uint shader,
499 uint slot, struct pipe_constant_buffer *cbuf);
500 void si_get_shader_buffers(struct si_context *sctx,
501 enum pipe_shader_type shader,
502 uint start_slot, uint count,
503 struct pipe_shader_buffer *sbuf);
504 void si_set_ring_buffer(struct si_context *sctx, uint slot,
505 struct pipe_resource *buffer,
506 unsigned stride, unsigned num_records,
507 bool add_tid, bool swizzle,
508 unsigned element_size, unsigned index_stride, uint64_t offset);
509 void si_init_all_descriptors(struct si_context *sctx);
510 bool si_upload_vertex_buffer_descriptors(struct si_context *sctx);
511 bool si_upload_graphics_shader_descriptors(struct si_context *sctx);
512 bool si_upload_compute_shader_descriptors(struct si_context *sctx);
513 void si_release_all_descriptors(struct si_context *sctx);
514 void si_gfx_resources_add_all_to_bo_list(struct si_context *sctx);
515 void si_compute_resources_add_all_to_bo_list(struct si_context *sctx);
516 void si_all_descriptors_begin_new_cs(struct si_context *sctx);
517 void si_upload_const_buffer(struct si_context *sctx, struct si_resource **buf,
518 const uint8_t *ptr, unsigned size, uint32_t *const_offset);
519 void si_update_all_texture_descriptors(struct si_context *sctx);
520 void si_shader_change_notify(struct si_context *sctx);
521 void si_update_needs_color_decompress_masks(struct si_context *sctx);
522 void si_emit_graphics_shader_pointers(struct si_context *sctx);
523 void si_emit_compute_shader_pointers(struct si_context *sctx);
524 void si_set_rw_buffer(struct si_context *sctx,
525 uint slot, const struct pipe_constant_buffer *input);
526 void si_set_rw_shader_buffer(struct si_context *sctx, uint slot,
527 const struct pipe_shader_buffer *sbuffer);
528 void si_set_active_descriptors(struct si_context *sctx, unsigned desc_idx,
529 uint64_t new_active_mask);
530 void si_set_active_descriptors_for_shader(struct si_context *sctx,
531 struct si_shader_selector *sel);
532 bool si_bindless_descriptor_can_reclaim_slab(void *priv,
533 struct pb_slab_entry *entry);
534 struct pb_slab *si_bindless_descriptor_slab_alloc(void *priv, unsigned heap,
535 unsigned entry_size,
536 unsigned group_index);
537 void si_bindless_descriptor_slab_free(void *priv, struct pb_slab *pslab);
538 void si_rebind_buffer(struct si_context *sctx, struct pipe_resource *buf);
539 /* si_state.c */
540 void si_init_state_compute_functions(struct si_context *sctx);
541 void si_init_state_functions(struct si_context *sctx);
542 void si_init_screen_state_functions(struct si_screen *sscreen);
543 void
544 si_make_buffer_descriptor(struct si_screen *screen, struct si_resource *buf,
545 enum pipe_format format,
546 unsigned offset, unsigned size,
547 uint32_t *state);
548 struct pipe_sampler_view *
549 si_create_sampler_view_custom(struct pipe_context *ctx,
550 struct pipe_resource *texture,
551 const struct pipe_sampler_view *state,
552 unsigned width0, unsigned height0,
553 unsigned force_level);
554 void si_update_fb_dirtiness_after_rendering(struct si_context *sctx);
555 void si_update_ps_iter_samples(struct si_context *sctx);
556 void si_save_qbo_state(struct si_context *sctx, struct si_qbo_state *st);
557 void si_restore_qbo_state(struct si_context *sctx, struct si_qbo_state *st);
558 void si_set_occlusion_query_state(struct si_context *sctx,
559 bool old_perfect_enable);
560
561 struct si_fast_udiv_info32 {
562 unsigned multiplier; /* the "magic number" multiplier */
563 unsigned pre_shift; /* shift for the dividend before multiplying */
564 unsigned post_shift; /* shift for the dividend after multiplying */
565 int increment; /* 0 or 1; if set then increment the numerator, using one of
566 the two strategies */
567 };
568
569 struct si_fast_udiv_info32
570 si_compute_fast_udiv_info32(uint32_t D, unsigned num_bits);
571
572 /* si_state_binning.c */
573 void si_emit_dpbb_state(struct si_context *sctx);
574
575 /* si_state_shaders.c */
576 void *si_get_ir_binary(struct si_shader_selector *sel, bool as_ngg);
577 bool si_shader_cache_load_shader(struct si_screen *sscreen, void *ir_binary,
578 struct si_shader *shader);
579 bool si_shader_cache_insert_shader(struct si_screen *sscreen, void *ir_binary,
580 struct si_shader *shader,
581 bool insert_into_disk_cache);
582 bool si_update_shaders(struct si_context *sctx);
583 void si_init_shader_functions(struct si_context *sctx);
584 bool si_init_shader_cache(struct si_screen *sscreen);
585 void si_destroy_shader_cache(struct si_screen *sscreen);
586 void si_schedule_initial_compile(struct si_context *sctx, unsigned processor,
587 struct util_queue_fence *ready_fence,
588 struct si_compiler_ctx_state *compiler_ctx_state,
589 void *job, util_queue_execute_func execute);
590 void si_get_active_slot_masks(const struct tgsi_shader_info *info,
591 uint32_t *const_and_shader_buffers,
592 uint64_t *samplers_and_images);
593 int si_shader_select_with_key(struct si_screen *sscreen,
594 struct si_shader_ctx_state *state,
595 struct si_compiler_ctx_state *compiler_state,
596 struct si_shader_key *key,
597 int thread_index,
598 bool optimized_or_none);
599 void si_shader_selector_key_vs(struct si_context *sctx,
600 struct si_shader_selector *vs,
601 struct si_shader_key *key,
602 struct si_vs_prolog_bits *prolog_key);
603 unsigned si_get_input_prim(const struct si_shader_selector *gs);
604
605 /* si_state_draw.c */
606 void si_emit_surface_sync(struct si_context *sctx, struct radeon_cmdbuf *cs,
607 unsigned cp_coher_cntl);
608 void si_prim_discard_signal_next_compute_ib_start(struct si_context *sctx);
609 void gfx10_emit_cache_flush(struct si_context *sctx);
610 void si_emit_cache_flush(struct si_context *sctx);
611 void si_trace_emit(struct si_context *sctx);
612 void si_init_draw_functions(struct si_context *sctx);
613
614 /* si_state_msaa.c */
615 void si_init_msaa_functions(struct si_context *sctx);
616 void si_emit_sample_locations(struct radeon_cmdbuf *cs, int nr_samples);
617
618 /* si_state_streamout.c */
619 void si_streamout_buffers_dirty(struct si_context *sctx);
620 void si_emit_streamout_end(struct si_context *sctx);
621 void si_update_prims_generated_query_state(struct si_context *sctx,
622 unsigned type, int diff);
623 void si_init_streamout_functions(struct si_context *sctx);
624
625
626 static inline unsigned si_get_constbuf_slot(unsigned slot)
627 {
628 /* Constant buffers are in slots [16..31], ascending */
629 return SI_NUM_SHADER_BUFFERS + slot;
630 }
631
632 static inline unsigned si_get_shaderbuf_slot(unsigned slot)
633 {
634 /* shader buffers are in slots [15..0], descending */
635 return SI_NUM_SHADER_BUFFERS - 1 - slot;
636 }
637
638 static inline unsigned si_get_sampler_slot(unsigned slot)
639 {
640 /* samplers are in slots [8..39], ascending */
641 return SI_NUM_IMAGES / 2 + slot;
642 }
643
644 static inline unsigned si_get_image_slot(unsigned slot)
645 {
646 /* images are in slots [15..0] (sampler slots [7..0]), descending */
647 return SI_NUM_IMAGES - 1 - slot;
648 }
649
650 #endif