radeonsi: invalidate caches at the beginning of the prim discard compute IB
[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 bool alpha_to_coverage:1;
60 bool alpha_to_one:1;
61 bool dual_src_blend:1;
62 bool logicop_enable:1;
63 };
64
65 struct si_state_rasterizer {
66 struct si_pm4_state pm4;
67 /* poly offset states for 16-bit, 24-bit, and 32-bit zbuffers */
68 struct si_pm4_state *pm4_poly_offset;
69 unsigned pa_sc_line_stipple;
70 unsigned pa_cl_clip_cntl;
71 float line_width;
72 float max_point_size;
73 unsigned sprite_coord_enable:8;
74 unsigned clip_plane_enable:8;
75 unsigned half_pixel_center:1;
76 unsigned flatshade:1;
77 unsigned two_side:1;
78 unsigned multisample_enable:1;
79 unsigned force_persample_interp:1;
80 unsigned line_stipple_enable:1;
81 unsigned poly_stipple_enable:1;
82 unsigned line_smooth:1;
83 unsigned poly_smooth:1;
84 unsigned uses_poly_offset:1;
85 unsigned clamp_fragment_color:1;
86 unsigned clamp_vertex_color:1;
87 unsigned rasterizer_discard:1;
88 unsigned scissor_enable:1;
89 unsigned clip_halfz:1;
90 unsigned cull_front:1;
91 unsigned cull_back:1;
92 unsigned depth_clamp_any:1;
93 unsigned provoking_vertex_first:1;
94 };
95
96 struct si_dsa_stencil_ref_part {
97 uint8_t valuemask[2];
98 uint8_t writemask[2];
99 };
100
101 struct si_dsa_order_invariance {
102 /** Whether the final result in Z/S buffers is guaranteed to be
103 * invariant under changes to the order in which fragments arrive. */
104 bool zs:1;
105
106 /** Whether the set of fragments that pass the combined Z/S test is
107 * guaranteed to be invariant under changes to the order in which
108 * fragments arrive. */
109 bool pass_set:1;
110
111 /** Whether the last fragment that passes the combined Z/S test at each
112 * sample is guaranteed to be invariant under changes to the order in
113 * which fragments arrive. */
114 bool pass_last:1;
115 };
116
117 struct si_state_dsa {
118 struct si_pm4_state pm4;
119 struct si_dsa_stencil_ref_part stencil_ref;
120
121 /* 0 = without stencil buffer, 1 = when both Z and S buffers are present */
122 struct si_dsa_order_invariance order_invariance[2];
123
124 ubyte alpha_func:3;
125 bool depth_enabled:1;
126 bool depth_write_enabled:1;
127 bool stencil_enabled:1;
128 bool stencil_write_enabled:1;
129 bool db_can_write:1;
130
131 };
132
133 struct si_stencil_ref {
134 struct pipe_stencil_ref state;
135 struct si_dsa_stencil_ref_part dsa_part;
136 };
137
138 struct si_vertex_elements
139 {
140 struct si_resource *instance_divisor_factor_buffer;
141 uint32_t rsrc_word3[SI_MAX_ATTRIBS];
142 uint16_t src_offset[SI_MAX_ATTRIBS];
143 uint8_t fix_fetch[SI_MAX_ATTRIBS];
144 uint8_t format_size[SI_MAX_ATTRIBS];
145 uint8_t vertex_buffer_index[SI_MAX_ATTRIBS];
146
147 /* Bitmask of elements that always need a fixup to be applied. */
148 uint16_t fix_fetch_always;
149
150 /* Bitmask of elements whose fetch should always be opencoded. */
151 uint16_t fix_fetch_opencode;
152
153 /* Bitmask of elements which need to be opencoded if the vertex buffer
154 * is unaligned. */
155 uint16_t fix_fetch_unaligned;
156
157 /* For elements in fix_fetch_unaligned: whether the effective
158 * element load size as seen by the hardware is a dword (as opposed
159 * to a short).
160 */
161 uint16_t hw_load_is_dword;
162
163 /* Bitmask of vertex buffers requiring alignment check */
164 uint16_t vb_alignment_check_mask;
165
166 uint8_t count;
167 bool uses_instance_divisors;
168
169 uint16_t first_vb_use_mask;
170 /* Vertex buffer descriptor list size aligned for optimal prefetch. */
171 uint16_t desc_list_byte_size;
172 uint16_t instance_divisor_is_one; /* bitmask of inputs */
173 uint16_t instance_divisor_is_fetched; /* bitmask of inputs */
174 };
175
176 union si_state {
177 struct {
178 struct si_state_blend *blend;
179 struct si_state_rasterizer *rasterizer;
180 struct si_state_dsa *dsa;
181 struct si_pm4_state *poly_offset;
182 struct si_pm4_state *ls;
183 struct si_pm4_state *hs;
184 struct si_pm4_state *es;
185 struct si_pm4_state *gs;
186 struct si_pm4_state *vgt_shader_config;
187 struct si_pm4_state *vs;
188 struct si_pm4_state *ps;
189 } named;
190 struct si_pm4_state *array[0];
191 };
192
193 #define SI_STATE_IDX(name) \
194 (offsetof(union si_state, named.name) / sizeof(struct si_pm4_state *))
195 #define SI_STATE_BIT(name) (1 << SI_STATE_IDX(name))
196 #define SI_NUM_STATES (sizeof(union si_state) / sizeof(struct si_pm4_state *))
197
198 static inline unsigned si_states_that_always_roll_context(void)
199 {
200 return (SI_STATE_BIT(blend) |
201 SI_STATE_BIT(rasterizer) |
202 SI_STATE_BIT(dsa) |
203 SI_STATE_BIT(poly_offset) |
204 SI_STATE_BIT(vgt_shader_config));
205 }
206
207 union si_state_atoms {
208 struct {
209 /* The order matters. */
210 struct si_atom render_cond;
211 struct si_atom streamout_begin;
212 struct si_atom streamout_enable; /* must be after streamout_begin */
213 struct si_atom framebuffer;
214 struct si_atom msaa_sample_locs;
215 struct si_atom db_render_state;
216 struct si_atom dpbb_state;
217 struct si_atom msaa_config;
218 struct si_atom sample_mask;
219 struct si_atom cb_render_state;
220 struct si_atom blend_color;
221 struct si_atom clip_regs;
222 struct si_atom clip_state;
223 struct si_atom shader_pointers;
224 struct si_atom guardband;
225 struct si_atom scissors;
226 struct si_atom viewports;
227 struct si_atom stencil_ref;
228 struct si_atom spi_map;
229 struct si_atom scratch_state;
230 struct si_atom window_rectangles;
231 } s;
232 struct si_atom array[0];
233 };
234
235 #define SI_ATOM_BIT(name) (1 << (offsetof(union si_state_atoms, s.name) / \
236 sizeof(struct si_atom)))
237 #define SI_NUM_ATOMS (sizeof(union si_state_atoms)/sizeof(struct si_atom*))
238
239 static inline unsigned si_atoms_that_always_roll_context(void)
240 {
241 return (SI_ATOM_BIT(streamout_begin) |
242 SI_ATOM_BIT(streamout_enable) |
243 SI_ATOM_BIT(framebuffer) |
244 SI_ATOM_BIT(msaa_sample_locs) |
245 SI_ATOM_BIT(sample_mask) |
246 SI_ATOM_BIT(blend_color) |
247 SI_ATOM_BIT(clip_state) |
248 SI_ATOM_BIT(scissors) |
249 SI_ATOM_BIT(viewports) |
250 SI_ATOM_BIT(stencil_ref) |
251 SI_ATOM_BIT(scratch_state) |
252 SI_ATOM_BIT(window_rectangles));
253 }
254
255 struct si_shader_data {
256 uint32_t sh_base[SI_NUM_SHADERS];
257 };
258
259 /* The list of registers whose emitted values are remembered by si_context. */
260 enum si_tracked_reg {
261 SI_TRACKED_DB_RENDER_CONTROL, /* 2 consecutive registers */
262 SI_TRACKED_DB_COUNT_CONTROL,
263
264 SI_TRACKED_DB_RENDER_OVERRIDE2,
265 SI_TRACKED_DB_SHADER_CONTROL,
266
267 SI_TRACKED_CB_TARGET_MASK,
268 SI_TRACKED_CB_DCC_CONTROL,
269
270 SI_TRACKED_SX_PS_DOWNCONVERT, /* 3 consecutive registers */
271 SI_TRACKED_SX_BLEND_OPT_EPSILON,
272 SI_TRACKED_SX_BLEND_OPT_CONTROL,
273
274 SI_TRACKED_PA_SC_LINE_CNTL, /* 2 consecutive registers */
275 SI_TRACKED_PA_SC_AA_CONFIG,
276
277 SI_TRACKED_DB_EQAA,
278 SI_TRACKED_PA_SC_MODE_CNTL_1,
279
280 SI_TRACKED_PA_SU_PRIM_FILTER_CNTL,
281 SI_TRACKED_PA_SU_SMALL_PRIM_FILTER_CNTL,
282
283 SI_TRACKED_PA_CL_VS_OUT_CNTL,
284 SI_TRACKED_PA_CL_CLIP_CNTL,
285
286 SI_TRACKED_PA_SC_BINNER_CNTL_0,
287 SI_TRACKED_DB_DFSM_CONTROL,
288
289 SI_TRACKED_PA_CL_GB_VERT_CLIP_ADJ, /* 4 consecutive registers */
290 SI_TRACKED_PA_CL_GB_VERT_DISC_ADJ,
291 SI_TRACKED_PA_CL_GB_HORZ_CLIP_ADJ,
292 SI_TRACKED_PA_CL_GB_HORZ_DISC_ADJ,
293
294 SI_TRACKED_PA_SU_HARDWARE_SCREEN_OFFSET,
295 SI_TRACKED_PA_SU_VTX_CNTL,
296
297 SI_TRACKED_PA_SC_CLIPRECT_RULE,
298
299 SI_TRACKED_VGT_ESGS_RING_ITEMSIZE,
300
301 SI_TRACKED_VGT_GSVS_RING_OFFSET_1, /* 4 consecutive registers */
302 SI_TRACKED_VGT_GSVS_RING_OFFSET_2,
303 SI_TRACKED_VGT_GSVS_RING_OFFSET_3,
304 SI_TRACKED_VGT_GS_OUT_PRIM_TYPE,
305
306 SI_TRACKED_VGT_GSVS_RING_ITEMSIZE,
307 SI_TRACKED_VGT_GS_MAX_VERT_OUT,
308
309 SI_TRACKED_VGT_GS_VERT_ITEMSIZE, /* 4 consecutive registers */
310 SI_TRACKED_VGT_GS_VERT_ITEMSIZE_1,
311 SI_TRACKED_VGT_GS_VERT_ITEMSIZE_2,
312 SI_TRACKED_VGT_GS_VERT_ITEMSIZE_3,
313
314 SI_TRACKED_VGT_GS_INSTANCE_CNT,
315 SI_TRACKED_VGT_GS_ONCHIP_CNTL,
316 SI_TRACKED_VGT_GS_MAX_PRIMS_PER_SUBGROUP,
317 SI_TRACKED_VGT_GS_MODE,
318 SI_TRACKED_VGT_PRIMITIVEID_EN,
319 SI_TRACKED_VGT_REUSE_OFF,
320 SI_TRACKED_SPI_VS_OUT_CONFIG,
321 SI_TRACKED_SPI_SHADER_POS_FORMAT,
322 SI_TRACKED_PA_CL_VTE_CNTL,
323
324 SI_TRACKED_SPI_PS_INPUT_ENA, /* 2 consecutive registers */
325 SI_TRACKED_SPI_PS_INPUT_ADDR,
326
327 SI_TRACKED_SPI_BARYC_CNTL,
328 SI_TRACKED_SPI_PS_IN_CONTROL,
329
330 SI_TRACKED_SPI_SHADER_Z_FORMAT, /* 2 consecutive registers */
331 SI_TRACKED_SPI_SHADER_COL_FORMAT,
332
333 SI_TRACKED_CB_SHADER_MASK,
334 SI_TRACKED_VGT_TF_PARAM,
335 SI_TRACKED_VGT_VERTEX_REUSE_BLOCK_CNTL,
336
337 SI_NUM_TRACKED_REGS,
338 };
339
340 struct si_tracked_regs {
341 uint64_t reg_saved;
342 uint32_t reg_value[SI_NUM_TRACKED_REGS];
343 uint32_t spi_ps_input_cntl[32];
344 };
345
346 /* Private read-write buffer slots. */
347 enum {
348 SI_ES_RING_ESGS,
349 SI_GS_RING_ESGS,
350
351 SI_RING_GSVS,
352
353 SI_VS_STREAMOUT_BUF0,
354 SI_VS_STREAMOUT_BUF1,
355 SI_VS_STREAMOUT_BUF2,
356 SI_VS_STREAMOUT_BUF3,
357
358 SI_HS_CONST_DEFAULT_TESS_LEVELS,
359 SI_VS_CONST_INSTANCE_DIVISORS,
360 SI_VS_CONST_CLIP_PLANES,
361 SI_PS_CONST_POLY_STIPPLE,
362 SI_PS_CONST_SAMPLE_POSITIONS,
363
364 /* Image descriptor of color buffer 0 for KHR_blend_equation_advanced. */
365 SI_PS_IMAGE_COLORBUF0,
366 SI_PS_IMAGE_COLORBUF0_HI,
367 SI_PS_IMAGE_COLORBUF0_FMASK,
368 SI_PS_IMAGE_COLORBUF0_FMASK_HI,
369
370 SI_NUM_RW_BUFFERS,
371 };
372
373 /* Indices into sctx->descriptors, laid out so that gfx and compute pipelines
374 * are contiguous:
375 *
376 * 0 - rw buffers
377 * 1 - vertex const and shader buffers
378 * 2 - vertex samplers and images
379 * 3 - fragment const and shader buffer
380 * ...
381 * 11 - compute const and shader buffers
382 * 12 - compute samplers and images
383 */
384 enum {
385 SI_SHADER_DESCS_CONST_AND_SHADER_BUFFERS,
386 SI_SHADER_DESCS_SAMPLERS_AND_IMAGES,
387 SI_NUM_SHADER_DESCS,
388 };
389
390 #define SI_DESCS_RW_BUFFERS 0
391 #define SI_DESCS_FIRST_SHADER 1
392 #define SI_DESCS_FIRST_COMPUTE (SI_DESCS_FIRST_SHADER + \
393 PIPE_SHADER_COMPUTE * SI_NUM_SHADER_DESCS)
394 #define SI_NUM_DESCS (SI_DESCS_FIRST_SHADER + \
395 SI_NUM_SHADERS * SI_NUM_SHADER_DESCS)
396
397 #define SI_DESCS_SHADER_MASK(name) \
398 u_bit_consecutive(SI_DESCS_FIRST_SHADER + \
399 PIPE_SHADER_##name * SI_NUM_SHADER_DESCS, \
400 SI_NUM_SHADER_DESCS)
401
402 static inline unsigned
403 si_const_and_shader_buffer_descriptors_idx(unsigned shader)
404 {
405 return SI_DESCS_FIRST_SHADER + shader * SI_NUM_SHADER_DESCS +
406 SI_SHADER_DESCS_CONST_AND_SHADER_BUFFERS;
407 }
408
409 static inline unsigned
410 si_sampler_and_image_descriptors_idx(unsigned shader)
411 {
412 return SI_DESCS_FIRST_SHADER + shader * SI_NUM_SHADER_DESCS +
413 SI_SHADER_DESCS_SAMPLERS_AND_IMAGES;
414 }
415
416 /* This represents descriptors in memory, such as buffer resources,
417 * image resources, and sampler states.
418 */
419 struct si_descriptors {
420 /* The list of descriptors in malloc'd memory. */
421 uint32_t *list;
422 /* The list in mapped GPU memory. */
423 uint32_t *gpu_list;
424
425 /* The buffer where the descriptors have been uploaded. */
426 struct si_resource *buffer;
427 uint64_t gpu_address;
428
429 /* The maximum number of descriptors. */
430 uint32_t num_elements;
431
432 /* Slots that are used by currently-bound shaders.
433 * It determines which slots are uploaded.
434 */
435 uint32_t first_active_slot;
436 uint32_t num_active_slots;
437
438 /* The SH register offset relative to USER_DATA*_0 where the pointer
439 * to the descriptor array will be stored. */
440 short shader_userdata_offset;
441 /* The size of one descriptor. */
442 ubyte element_dw_size;
443 /* If there is only one slot enabled, bind it directly instead of
444 * uploading descriptors. -1 if disabled. */
445 signed char slot_index_to_bind_directly;
446 };
447
448 struct si_buffer_resources {
449 struct pipe_resource **buffers; /* this has num_buffers elements */
450
451 enum radeon_bo_priority priority:6;
452 enum radeon_bo_priority priority_constbuf:6;
453
454 /* The i-th bit is set if that element is enabled (non-NULL resource). */
455 unsigned enabled_mask;
456 unsigned writable_mask;
457 };
458
459 #define si_pm4_state_changed(sctx, member) \
460 ((sctx)->queued.named.member != (sctx)->emitted.named.member)
461
462 #define si_pm4_state_enabled_and_changed(sctx, member) \
463 ((sctx)->queued.named.member && si_pm4_state_changed(sctx, member))
464
465 #define si_pm4_bind_state(sctx, member, value) \
466 do { \
467 (sctx)->queued.named.member = (value); \
468 (sctx)->dirty_states |= SI_STATE_BIT(member); \
469 } while(0)
470
471 #define si_pm4_delete_state(sctx, member, value) \
472 do { \
473 if ((sctx)->queued.named.member == (value)) { \
474 (sctx)->queued.named.member = NULL; \
475 } \
476 si_pm4_free_state(sctx, (struct si_pm4_state *)(value), \
477 SI_STATE_IDX(member)); \
478 } while(0)
479
480 /* si_descriptors.c */
481 void si_set_mutable_tex_desc_fields(struct si_screen *sscreen,
482 struct si_texture *tex,
483 const struct legacy_surf_level *base_level_info,
484 unsigned base_level, unsigned first_level,
485 unsigned block_width, bool is_stencil,
486 uint32_t *state);
487 void si_update_ps_colorbuf0_slot(struct si_context *sctx);
488 void si_get_pipe_constant_buffer(struct si_context *sctx, uint shader,
489 uint slot, struct pipe_constant_buffer *cbuf);
490 void si_get_shader_buffers(struct si_context *sctx,
491 enum pipe_shader_type shader,
492 uint start_slot, uint count,
493 struct pipe_shader_buffer *sbuf);
494 void si_set_ring_buffer(struct si_context *sctx, uint slot,
495 struct pipe_resource *buffer,
496 unsigned stride, unsigned num_records,
497 bool add_tid, bool swizzle,
498 unsigned element_size, unsigned index_stride, uint64_t offset);
499 void si_init_all_descriptors(struct si_context *sctx);
500 bool si_upload_vertex_buffer_descriptors(struct si_context *sctx);
501 bool si_upload_graphics_shader_descriptors(struct si_context *sctx);
502 bool si_upload_compute_shader_descriptors(struct si_context *sctx);
503 void si_release_all_descriptors(struct si_context *sctx);
504 void si_gfx_resources_add_all_to_bo_list(struct si_context *sctx);
505 void si_compute_resources_add_all_to_bo_list(struct si_context *sctx);
506 void si_all_descriptors_begin_new_cs(struct si_context *sctx);
507 void si_upload_const_buffer(struct si_context *sctx, struct si_resource **buf,
508 const uint8_t *ptr, unsigned size, uint32_t *const_offset);
509 void si_update_all_texture_descriptors(struct si_context *sctx);
510 void si_shader_change_notify(struct si_context *sctx);
511 void si_update_needs_color_decompress_masks(struct si_context *sctx);
512 void si_emit_graphics_shader_pointers(struct si_context *sctx);
513 void si_emit_compute_shader_pointers(struct si_context *sctx);
514 void si_set_rw_buffer(struct si_context *sctx,
515 uint slot, const struct pipe_constant_buffer *input);
516 void si_set_rw_shader_buffer(struct si_context *sctx, uint slot,
517 const struct pipe_shader_buffer *sbuffer);
518 void si_set_active_descriptors(struct si_context *sctx, unsigned desc_idx,
519 uint64_t new_active_mask);
520 void si_set_active_descriptors_for_shader(struct si_context *sctx,
521 struct si_shader_selector *sel);
522 bool si_bindless_descriptor_can_reclaim_slab(void *priv,
523 struct pb_slab_entry *entry);
524 struct pb_slab *si_bindless_descriptor_slab_alloc(void *priv, unsigned heap,
525 unsigned entry_size,
526 unsigned group_index);
527 void si_bindless_descriptor_slab_free(void *priv, struct pb_slab *pslab);
528 void si_rebind_buffer(struct si_context *sctx, struct pipe_resource *buf,
529 uint64_t old_va);
530 /* si_state.c */
531 void si_init_state_compute_functions(struct si_context *sctx);
532 void si_init_state_functions(struct si_context *sctx);
533 void si_init_screen_state_functions(struct si_screen *sscreen);
534 void
535 si_make_buffer_descriptor(struct si_screen *screen, struct si_resource *buf,
536 enum pipe_format format,
537 unsigned offset, unsigned size,
538 uint32_t *state);
539 void
540 si_make_texture_descriptor(struct si_screen *screen,
541 struct si_texture *tex,
542 bool sampler,
543 enum pipe_texture_target target,
544 enum pipe_format pipe_format,
545 const unsigned char state_swizzle[4],
546 unsigned first_level, unsigned last_level,
547 unsigned first_layer, unsigned last_layer,
548 unsigned width, unsigned height, unsigned depth,
549 uint32_t *state,
550 uint32_t *fmask_state);
551 struct pipe_sampler_view *
552 si_create_sampler_view_custom(struct pipe_context *ctx,
553 struct pipe_resource *texture,
554 const struct pipe_sampler_view *state,
555 unsigned width0, unsigned height0,
556 unsigned force_level);
557 void si_update_fb_dirtiness_after_rendering(struct si_context *sctx);
558 void si_update_ps_iter_samples(struct si_context *sctx);
559 void si_save_qbo_state(struct si_context *sctx, struct si_qbo_state *st);
560 void si_set_occlusion_query_state(struct si_context *sctx,
561 bool old_perfect_enable);
562
563 struct si_fast_udiv_info32 {
564 unsigned multiplier; /* the "magic number" multiplier */
565 unsigned pre_shift; /* shift for the dividend before multiplying */
566 unsigned post_shift; /* shift for the dividend after multiplying */
567 int increment; /* 0 or 1; if set then increment the numerator, using one of
568 the two strategies */
569 };
570
571 struct si_fast_udiv_info32
572 si_compute_fast_udiv_info32(uint32_t D, unsigned num_bits);
573
574 /* si_state_binning.c */
575 void si_emit_dpbb_state(struct si_context *sctx);
576
577 /* si_state_shaders.c */
578 void *si_get_ir_binary(struct si_shader_selector *sel);
579 bool si_shader_cache_load_shader(struct si_screen *sscreen, void *ir_binary,
580 struct si_shader *shader);
581 bool si_shader_cache_insert_shader(struct si_screen *sscreen, void *ir_binary,
582 struct si_shader *shader,
583 bool insert_into_disk_cache);
584 bool si_update_shaders(struct si_context *sctx);
585 void si_init_shader_functions(struct si_context *sctx);
586 bool si_init_shader_cache(struct si_screen *sscreen);
587 void si_destroy_shader_cache(struct si_screen *sscreen);
588 void si_schedule_initial_compile(struct si_context *sctx, unsigned processor,
589 struct util_queue_fence *ready_fence,
590 struct si_compiler_ctx_state *compiler_ctx_state,
591 void *job, util_queue_execute_func execute);
592 void si_get_active_slot_masks(const struct tgsi_shader_info *info,
593 uint32_t *const_and_shader_buffers,
594 uint64_t *samplers_and_images);
595 int si_shader_select_with_key(struct si_screen *sscreen,
596 struct si_shader_ctx_state *state,
597 struct si_compiler_ctx_state *compiler_state,
598 struct si_shader_key *key,
599 int thread_index,
600 bool optimized_or_none);
601 void si_shader_selector_key_vs(struct si_context *sctx,
602 struct si_shader_selector *vs,
603 struct si_shader_key *key,
604 struct si_vs_prolog_bits *prolog_key);
605
606 /* si_state_draw.c */
607 void si_emit_surface_sync(struct si_context *sctx, struct radeon_cmdbuf *cs,
608 unsigned cp_coher_cntl);
609 void si_prim_discard_signal_next_compute_ib_start(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