radeonsi: rename r600_texture -> si_texture, rxxx -> xxx or sxxx
[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_selector;
45 struct si_texture;
46 struct si_qbo_state;
47
48 /* State atoms are callbacks which write a sequence of packets into a GPU
49 * command buffer (AKA indirect buffer, AKA IB, AKA command stream, AKA CS).
50 */
51 struct si_atom {
52 void (*emit)(struct si_context *ctx);
53 };
54
55 struct si_state_blend {
56 struct si_pm4_state pm4;
57 uint32_t cb_target_mask;
58 /* Set 0xf or 0x0 (4 bits) per render target if the following is
59 * true. ANDed with spi_shader_col_format.
60 */
61 unsigned cb_target_enabled_4bit;
62 unsigned blend_enable_4bit;
63 unsigned need_src_alpha_4bit;
64 unsigned commutative_4bit;
65 bool alpha_to_coverage:1;
66 bool alpha_to_one:1;
67 bool dual_src_blend:1;
68 bool logicop_enable:1;
69 };
70
71 struct si_state_rasterizer {
72 struct si_pm4_state pm4;
73 /* poly offset states for 16-bit, 24-bit, and 32-bit zbuffers */
74 struct si_pm4_state *pm4_poly_offset;
75 unsigned pa_sc_line_stipple;
76 unsigned pa_cl_clip_cntl;
77 float line_width;
78 float max_point_size;
79 unsigned sprite_coord_enable:8;
80 unsigned clip_plane_enable:8;
81 unsigned flatshade:1;
82 unsigned two_side:1;
83 unsigned multisample_enable:1;
84 unsigned force_persample_interp:1;
85 unsigned line_stipple_enable:1;
86 unsigned poly_stipple_enable:1;
87 unsigned line_smooth:1;
88 unsigned poly_smooth:1;
89 unsigned uses_poly_offset:1;
90 unsigned clamp_fragment_color:1;
91 unsigned clamp_vertex_color:1;
92 unsigned rasterizer_discard:1;
93 unsigned scissor_enable:1;
94 unsigned clip_halfz:1;
95 };
96
97 struct si_dsa_stencil_ref_part {
98 uint8_t valuemask[2];
99 uint8_t writemask[2];
100 };
101
102 struct si_dsa_order_invariance {
103 /** Whether the final result in Z/S buffers is guaranteed to be
104 * invariant under changes to the order in which fragments arrive. */
105 bool zs:1;
106
107 /** Whether the set of fragments that pass the combined Z/S test is
108 * guaranteed to be invariant under changes to the order in which
109 * fragments arrive. */
110 bool pass_set:1;
111
112 /** Whether the last fragment that passes the combined Z/S test at each
113 * sample is guaranteed to be invariant under changes to the order in
114 * which fragments arrive. */
115 bool pass_last:1;
116 };
117
118 struct si_state_dsa {
119 struct si_pm4_state pm4;
120 struct si_dsa_stencil_ref_part stencil_ref;
121
122 /* 0 = without stencil buffer, 1 = when both Z and S buffers are present */
123 struct si_dsa_order_invariance order_invariance[2];
124
125 ubyte alpha_func:3;
126 bool depth_enabled:1;
127 bool depth_write_enabled:1;
128 bool stencil_enabled:1;
129 bool stencil_write_enabled:1;
130 bool db_can_write:1;
131
132 };
133
134 struct si_stencil_ref {
135 struct pipe_stencil_ref state;
136 struct si_dsa_stencil_ref_part dsa_part;
137 };
138
139 struct si_vertex_elements
140 {
141 uint32_t instance_divisors[SI_MAX_ATTRIBS];
142 uint32_t rsrc_word3[SI_MAX_ATTRIBS];
143 uint16_t src_offset[SI_MAX_ATTRIBS];
144 uint8_t fix_fetch[SI_MAX_ATTRIBS];
145 uint8_t format_size[SI_MAX_ATTRIBS];
146 uint8_t vertex_buffer_index[SI_MAX_ATTRIBS];
147
148 uint8_t count;
149 bool uses_instance_divisors;
150
151 uint16_t first_vb_use_mask;
152 /* Vertex buffer descriptor list size aligned for optimal prefetch. */
153 uint16_t desc_list_byte_size;
154 uint16_t instance_divisor_is_one; /* bitmask of inputs */
155 uint16_t instance_divisor_is_fetched; /* bitmask of inputs */
156 };
157
158 union si_state {
159 struct {
160 struct si_state_blend *blend;
161 struct si_state_rasterizer *rasterizer;
162 struct si_state_dsa *dsa;
163 struct si_pm4_state *poly_offset;
164 struct si_pm4_state *ls;
165 struct si_pm4_state *hs;
166 struct si_pm4_state *es;
167 struct si_pm4_state *gs;
168 struct si_pm4_state *vgt_shader_config;
169 struct si_pm4_state *vs;
170 struct si_pm4_state *ps;
171 } named;
172 struct si_pm4_state *array[0];
173 };
174
175 #define SI_STATE_IDX(name) \
176 (offsetof(union si_state, named.name) / sizeof(struct si_pm4_state *))
177 #define SI_STATE_BIT(name) (1 << SI_STATE_IDX(name))
178 #define SI_NUM_STATES (sizeof(union si_state) / sizeof(struct si_pm4_state *))
179
180 static inline unsigned si_states_that_roll_context(void)
181 {
182 return (SI_STATE_BIT(blend) |
183 SI_STATE_BIT(rasterizer) |
184 SI_STATE_BIT(dsa) |
185 SI_STATE_BIT(poly_offset) |
186 SI_STATE_BIT(es) |
187 SI_STATE_BIT(gs) |
188 SI_STATE_BIT(vgt_shader_config) |
189 SI_STATE_BIT(vs) |
190 SI_STATE_BIT(ps));
191 }
192
193 union si_state_atoms {
194 struct {
195 /* The order matters. */
196 struct si_atom render_cond;
197 struct si_atom streamout_begin;
198 struct si_atom streamout_enable; /* must be after streamout_begin */
199 struct si_atom framebuffer;
200 struct si_atom msaa_sample_locs;
201 struct si_atom db_render_state;
202 struct si_atom dpbb_state;
203 struct si_atom msaa_config;
204 struct si_atom sample_mask;
205 struct si_atom cb_render_state;
206 struct si_atom blend_color;
207 struct si_atom clip_regs;
208 struct si_atom clip_state;
209 struct si_atom shader_pointers;
210 struct si_atom guardband;
211 struct si_atom scissors;
212 struct si_atom viewports;
213 struct si_atom stencil_ref;
214 struct si_atom spi_map;
215 struct si_atom scratch_state;
216 } s;
217 struct si_atom array[0];
218 };
219
220 #define SI_ATOM_BIT(name) (1 << (offsetof(union si_state_atoms, s.name) / \
221 sizeof(struct si_atom)))
222 #define SI_NUM_ATOMS (sizeof(union si_state_atoms)/sizeof(struct si_atom*))
223
224 static inline unsigned si_atoms_that_roll_context(void)
225 {
226 return (SI_ATOM_BIT(streamout_begin) |
227 SI_ATOM_BIT(streamout_enable) |
228 SI_ATOM_BIT(framebuffer) |
229 SI_ATOM_BIT(msaa_sample_locs) |
230 SI_ATOM_BIT(db_render_state) |
231 SI_ATOM_BIT(dpbb_state) |
232 SI_ATOM_BIT(msaa_config) |
233 SI_ATOM_BIT(sample_mask) |
234 SI_ATOM_BIT(cb_render_state) |
235 SI_ATOM_BIT(blend_color) |
236 SI_ATOM_BIT(clip_regs) |
237 SI_ATOM_BIT(clip_state) |
238 SI_ATOM_BIT(guardband) |
239 SI_ATOM_BIT(scissors) |
240 SI_ATOM_BIT(viewports) |
241 SI_ATOM_BIT(stencil_ref) |
242 SI_ATOM_BIT(spi_map) |
243 SI_ATOM_BIT(scratch_state));
244 }
245
246 struct si_shader_data {
247 uint32_t sh_base[SI_NUM_SHADERS];
248 };
249
250 /* The list of registers whose emitted values are remembered by si_context. */
251 enum si_tracked_reg {
252 SI_TRACKED_DB_RENDER_CONTROL, /* 2 consecutive registers */
253 SI_TRACKED_DB_COUNT_CONTROL,
254
255 SI_TRACKED_DB_RENDER_OVERRIDE2,
256 SI_TRACKED_DB_SHADER_CONTROL,
257
258 SI_TRACKED_CB_TARGET_MASK,
259 SI_TRACKED_CB_DCC_CONTROL,
260
261 SI_TRACKED_SX_PS_DOWNCONVERT, /* 3 consecutive registers */
262 SI_TRACKED_SX_BLEND_OPT_EPSILON,
263 SI_TRACKED_SX_BLEND_OPT_CONTROL,
264
265 SI_TRACKED_PA_SC_LINE_CNTL, /* 2 consecutive registers */
266 SI_TRACKED_PA_SC_AA_CONFIG,
267
268 SI_TRACKED_DB_EQAA,
269 SI_TRACKED_PA_SC_MODE_CNTL_1,
270
271 SI_TRACKED_PA_SU_SMALL_PRIM_FILTER_CNTL,
272
273 SI_TRACKED_PA_CL_VS_OUT_CNTL,
274 SI_TRACKED_PA_CL_CLIP_CNTL,
275
276 SI_TRACKED_PA_SC_BINNER_CNTL_0,
277 SI_TRACKED_DB_DFSM_CONTROL,
278
279 SI_NUM_TRACKED_REGS,
280 };
281
282 struct si_tracked_regs {
283 uint32_t reg_saved;
284 uint32_t reg_value[SI_NUM_TRACKED_REGS];
285 };
286
287 /* Private read-write buffer slots. */
288 enum {
289 SI_ES_RING_ESGS,
290 SI_GS_RING_ESGS,
291
292 SI_RING_GSVS,
293
294 SI_VS_STREAMOUT_BUF0,
295 SI_VS_STREAMOUT_BUF1,
296 SI_VS_STREAMOUT_BUF2,
297 SI_VS_STREAMOUT_BUF3,
298
299 SI_HS_CONST_DEFAULT_TESS_LEVELS,
300 SI_VS_CONST_INSTANCE_DIVISORS,
301 SI_VS_CONST_CLIP_PLANES,
302 SI_PS_CONST_POLY_STIPPLE,
303 SI_PS_CONST_SAMPLE_POSITIONS,
304
305 /* Image descriptor of color buffer 0 for KHR_blend_equation_advanced. */
306 SI_PS_IMAGE_COLORBUF0,
307 SI_PS_IMAGE_COLORBUF0_HI,
308 SI_PS_IMAGE_COLORBUF0_FMASK,
309 SI_PS_IMAGE_COLORBUF0_FMASK_HI,
310
311 SI_NUM_RW_BUFFERS,
312 };
313
314 /* Indices into sctx->descriptors, laid out so that gfx and compute pipelines
315 * are contiguous:
316 *
317 * 0 - rw buffers
318 * 1 - vertex const and shader buffers
319 * 2 - vertex samplers and images
320 * 3 - fragment const and shader buffer
321 * ...
322 * 11 - compute const and shader buffers
323 * 12 - compute samplers and images
324 */
325 enum {
326 SI_SHADER_DESCS_CONST_AND_SHADER_BUFFERS,
327 SI_SHADER_DESCS_SAMPLERS_AND_IMAGES,
328 SI_NUM_SHADER_DESCS,
329 };
330
331 #define SI_DESCS_RW_BUFFERS 0
332 #define SI_DESCS_FIRST_SHADER 1
333 #define SI_DESCS_FIRST_COMPUTE (SI_DESCS_FIRST_SHADER + \
334 PIPE_SHADER_COMPUTE * SI_NUM_SHADER_DESCS)
335 #define SI_NUM_DESCS (SI_DESCS_FIRST_SHADER + \
336 SI_NUM_SHADERS * SI_NUM_SHADER_DESCS)
337
338 #define SI_DESCS_SHADER_MASK(name) \
339 u_bit_consecutive(SI_DESCS_FIRST_SHADER + \
340 PIPE_SHADER_##name * SI_NUM_SHADER_DESCS, \
341 SI_NUM_SHADER_DESCS)
342
343 /* This represents descriptors in memory, such as buffer resources,
344 * image resources, and sampler states.
345 */
346 struct si_descriptors {
347 /* The list of descriptors in malloc'd memory. */
348 uint32_t *list;
349 /* The list in mapped GPU memory. */
350 uint32_t *gpu_list;
351
352 /* The buffer where the descriptors have been uploaded. */
353 struct r600_resource *buffer;
354 uint64_t gpu_address;
355
356 /* The maximum number of descriptors. */
357 uint32_t num_elements;
358
359 /* Slots that are used by currently-bound shaders.
360 * It determines which slots are uploaded.
361 */
362 uint32_t first_active_slot;
363 uint32_t num_active_slots;
364
365 /* The SH register offset relative to USER_DATA*_0 where the pointer
366 * to the descriptor array will be stored. */
367 short shader_userdata_offset;
368 /* The size of one descriptor. */
369 ubyte element_dw_size;
370 /* If there is only one slot enabled, bind it directly instead of
371 * uploading descriptors. -1 if disabled. */
372 signed char slot_index_to_bind_directly;
373 };
374
375 struct si_buffer_resources {
376 struct pipe_resource **buffers; /* this has num_buffers elements */
377
378 enum radeon_bo_usage shader_usage:4; /* READ, WRITE, or READWRITE */
379 enum radeon_bo_usage shader_usage_constbuf:4;
380 enum radeon_bo_priority priority:6;
381 enum radeon_bo_priority priority_constbuf:6;
382
383 /* The i-th bit is set if that element is enabled (non-NULL resource). */
384 unsigned enabled_mask;
385 };
386
387 #define si_pm4_state_changed(sctx, member) \
388 ((sctx)->queued.named.member != (sctx)->emitted.named.member)
389
390 #define si_pm4_state_enabled_and_changed(sctx, member) \
391 ((sctx)->queued.named.member && si_pm4_state_changed(sctx, member))
392
393 #define si_pm4_bind_state(sctx, member, value) \
394 do { \
395 (sctx)->queued.named.member = (value); \
396 (sctx)->dirty_states |= SI_STATE_BIT(member); \
397 } while(0)
398
399 #define si_pm4_delete_state(sctx, member, value) \
400 do { \
401 if ((sctx)->queued.named.member == (value)) { \
402 (sctx)->queued.named.member = NULL; \
403 } \
404 si_pm4_free_state(sctx, (struct si_pm4_state *)(value), \
405 SI_STATE_IDX(member)); \
406 } while(0)
407
408 /* si_descriptors.c */
409 void si_set_mutable_tex_desc_fields(struct si_screen *sscreen,
410 struct si_texture *tex,
411 const struct legacy_surf_level *base_level_info,
412 unsigned base_level, unsigned first_level,
413 unsigned block_width, bool is_stencil,
414 uint32_t *state);
415 void si_update_ps_colorbuf0_slot(struct si_context *sctx);
416 void si_get_pipe_constant_buffer(struct si_context *sctx, uint shader,
417 uint slot, struct pipe_constant_buffer *cbuf);
418 void si_get_shader_buffers(struct si_context *sctx,
419 enum pipe_shader_type shader,
420 uint start_slot, uint count,
421 struct pipe_shader_buffer *sbuf);
422 void si_set_ring_buffer(struct si_context *sctx, uint slot,
423 struct pipe_resource *buffer,
424 unsigned stride, unsigned num_records,
425 bool add_tid, bool swizzle,
426 unsigned element_size, unsigned index_stride, uint64_t offset);
427 void si_init_all_descriptors(struct si_context *sctx);
428 bool si_upload_vertex_buffer_descriptors(struct si_context *sctx);
429 bool si_upload_graphics_shader_descriptors(struct si_context *sctx);
430 bool si_upload_compute_shader_descriptors(struct si_context *sctx);
431 void si_release_all_descriptors(struct si_context *sctx);
432 void si_all_descriptors_begin_new_cs(struct si_context *sctx);
433 void si_all_resident_buffers_begin_new_cs(struct si_context *sctx);
434 void si_upload_const_buffer(struct si_context *sctx, struct r600_resource **rbuffer,
435 const uint8_t *ptr, unsigned size, uint32_t *const_offset);
436 void si_update_all_texture_descriptors(struct si_context *sctx);
437 void si_shader_change_notify(struct si_context *sctx);
438 void si_update_needs_color_decompress_masks(struct si_context *sctx);
439 void si_emit_graphics_shader_pointers(struct si_context *sctx);
440 void si_emit_compute_shader_pointers(struct si_context *sctx);
441 void si_set_rw_buffer(struct si_context *sctx,
442 uint slot, const struct pipe_constant_buffer *input);
443 void si_set_active_descriptors(struct si_context *sctx, unsigned desc_idx,
444 uint64_t new_active_mask);
445 void si_set_active_descriptors_for_shader(struct si_context *sctx,
446 struct si_shader_selector *sel);
447 bool si_bindless_descriptor_can_reclaim_slab(void *priv,
448 struct pb_slab_entry *entry);
449 struct pb_slab *si_bindless_descriptor_slab_alloc(void *priv, unsigned heap,
450 unsigned entry_size,
451 unsigned group_index);
452 void si_bindless_descriptor_slab_free(void *priv, struct pb_slab *pslab);
453 void si_rebind_buffer(struct si_context *sctx, struct pipe_resource *buf,
454 uint64_t old_va);
455 /* si_state.c */
456 void si_init_state_functions(struct si_context *sctx);
457 void si_init_screen_state_functions(struct si_screen *sscreen);
458 void
459 si_make_buffer_descriptor(struct si_screen *screen, struct r600_resource *buf,
460 enum pipe_format format,
461 unsigned offset, unsigned size,
462 uint32_t *state);
463 void
464 si_make_texture_descriptor(struct si_screen *screen,
465 struct si_texture *tex,
466 bool sampler,
467 enum pipe_texture_target target,
468 enum pipe_format pipe_format,
469 const unsigned char state_swizzle[4],
470 unsigned first_level, unsigned last_level,
471 unsigned first_layer, unsigned last_layer,
472 unsigned width, unsigned height, unsigned depth,
473 uint32_t *state,
474 uint32_t *fmask_state);
475 struct pipe_sampler_view *
476 si_create_sampler_view_custom(struct pipe_context *ctx,
477 struct pipe_resource *texture,
478 const struct pipe_sampler_view *state,
479 unsigned width0, unsigned height0,
480 unsigned force_level);
481 void si_update_fb_dirtiness_after_rendering(struct si_context *sctx);
482 void si_update_ps_iter_samples(struct si_context *sctx);
483 void si_save_qbo_state(struct si_context *sctx, struct si_qbo_state *st);
484 void si_set_occlusion_query_state(struct si_context *sctx,
485 bool old_perfect_enable);
486
487 /* si_state_binning.c */
488 void si_emit_dpbb_state(struct si_context *sctx);
489
490 /* si_state_shaders.c */
491 bool si_update_shaders(struct si_context *sctx);
492 void si_init_shader_functions(struct si_context *sctx);
493 bool si_init_shader_cache(struct si_screen *sscreen);
494 void si_destroy_shader_cache(struct si_screen *sscreen);
495 void si_get_active_slot_masks(const struct tgsi_shader_info *info,
496 uint32_t *const_and_shader_buffers,
497 uint64_t *samplers_and_images);
498 void *si_get_blit_vs(struct si_context *sctx, enum blitter_attrib_type type,
499 unsigned num_layers);
500
501 /* si_state_draw.c */
502 void si_init_ia_multi_vgt_param_table(struct si_context *sctx);
503 void si_emit_cache_flush(struct si_context *sctx);
504 void si_draw_vbo(struct pipe_context *ctx, const struct pipe_draw_info *dinfo);
505 void si_draw_rectangle(struct blitter_context *blitter,
506 void *vertex_elements_cso,
507 blitter_get_vs_func get_vs,
508 int x1, int y1, int x2, int y2,
509 float depth, unsigned num_instances,
510 enum blitter_attrib_type type,
511 const union blitter_attrib *attrib);
512 void si_trace_emit(struct si_context *sctx);
513
514 /* si_state_msaa.c */
515 void si_init_msaa_functions(struct si_context *sctx);
516 void si_emit_sample_locations(struct radeon_cmdbuf *cs, int nr_samples);
517
518 /* si_state_streamout.c */
519 void si_streamout_buffers_dirty(struct si_context *sctx);
520 void si_emit_streamout_end(struct si_context *sctx);
521 void si_update_prims_generated_query_state(struct si_context *sctx,
522 unsigned type, int diff);
523 void si_init_streamout_functions(struct si_context *sctx);
524
525
526 static inline unsigned si_get_constbuf_slot(unsigned slot)
527 {
528 /* Constant buffers are in slots [16..31], ascending */
529 return SI_NUM_SHADER_BUFFERS + slot;
530 }
531
532 static inline unsigned si_get_shaderbuf_slot(unsigned slot)
533 {
534 /* shader buffers are in slots [15..0], descending */
535 return SI_NUM_SHADER_BUFFERS - 1 - slot;
536 }
537
538 static inline unsigned si_get_sampler_slot(unsigned slot)
539 {
540 /* samplers are in slots [8..39], ascending */
541 return SI_NUM_IMAGES / 2 + slot;
542 }
543
544 static inline unsigned si_get_image_slot(unsigned slot)
545 {
546 /* images are in slots [15..0] (sampler slots [7..0]), descending */
547 return SI_NUM_IMAGES - 1 - slot;
548 }
549
550 #endif