Merge remote-tracking branch 'public/master' into vulkan
[mesa.git] / src / gallium / drivers / radeon / r600_pipe_common.h
1 /*
2 * Copyright 2013 Advanced Micro Devices, Inc.
3 *
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:
10 *
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
13 * Software.
14 *
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 FROM,
20 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
21 * SOFTWARE.
22 *
23 * Authors: Marek Olšák <maraeo@gmail.com>
24 *
25 */
26
27 /**
28 * This file contains common screen and context structures and functions
29 * for r600g and radeonsi.
30 */
31
32 #ifndef R600_PIPE_COMMON_H
33 #define R600_PIPE_COMMON_H
34
35 #include <stdio.h>
36
37 #include "radeon/radeon_winsys.h"
38
39 #include "util/u_blitter.h"
40 #include "util/list.h"
41 #include "util/u_range.h"
42 #include "util/u_slab.h"
43 #include "util/u_suballoc.h"
44 #include "util/u_transfer.h"
45
46 #define ATI_VENDOR_ID 0x1002
47
48 #define R600_RESOURCE_FLAG_TRANSFER (PIPE_RESOURCE_FLAG_DRV_PRIV << 0)
49 #define R600_RESOURCE_FLAG_FLUSHED_DEPTH (PIPE_RESOURCE_FLAG_DRV_PRIV << 1)
50 #define R600_RESOURCE_FLAG_FORCE_TILING (PIPE_RESOURCE_FLAG_DRV_PRIV << 2)
51
52 #define R600_CONTEXT_STREAMOUT_FLUSH (1u << 0)
53 #define R600_CONTEXT_PRIVATE_FLAG (1u << 1)
54
55 /* special primitive types */
56 #define R600_PRIM_RECTANGLE_LIST PIPE_PRIM_MAX
57
58 /* Debug flags. */
59 /* logging */
60 #define DBG_TEX (1 << 0)
61 /* gap - reuse */
62 #define DBG_COMPUTE (1 << 2)
63 #define DBG_VM (1 << 3)
64 /* gap - reuse */
65 /* shader logging */
66 #define DBG_FS (1 << 5)
67 #define DBG_VS (1 << 6)
68 #define DBG_GS (1 << 7)
69 #define DBG_PS (1 << 8)
70 #define DBG_CS (1 << 9)
71 #define DBG_TCS (1 << 10)
72 #define DBG_TES (1 << 11)
73 #define DBG_NO_IR (1 << 12)
74 #define DBG_NO_TGSI (1 << 13)
75 #define DBG_NO_ASM (1 << 14)
76 #define DBG_PREOPT_IR (1 << 15)
77 /* Bits 21-31 are reserved for the r600g driver. */
78 /* features */
79 #define DBG_NO_ASYNC_DMA (1llu << 32)
80 #define DBG_NO_HYPERZ (1llu << 33)
81 #define DBG_NO_DISCARD_RANGE (1llu << 34)
82 #define DBG_NO_2D_TILING (1llu << 35)
83 #define DBG_NO_TILING (1llu << 36)
84 #define DBG_SWITCH_ON_EOP (1llu << 37)
85 #define DBG_FORCE_DMA (1llu << 38)
86 #define DBG_PRECOMPILE (1llu << 39)
87 #define DBG_INFO (1llu << 40)
88 #define DBG_NO_WC (1llu << 41)
89 #define DBG_CHECK_VM (1llu << 42)
90 #define DBG_NO_DCC (1llu << 43)
91 #define DBG_NO_DCC_CLEAR (1llu << 44)
92 #define DBG_NO_RB_PLUS (1llu << 45)
93 #define DBG_SI_SCHED (1llu << 46)
94 #define DBG_MONOLITHIC_SHADERS (1llu << 47)
95
96 #define R600_MAP_BUFFER_ALIGNMENT 64
97
98 struct r600_common_context;
99 struct r600_perfcounters;
100
101 struct radeon_shader_reloc {
102 char name[32];
103 uint64_t offset;
104 };
105
106 struct radeon_shader_binary {
107 /** Shader code */
108 unsigned char *code;
109 unsigned code_size;
110
111 /** Config/Context register state that accompanies this shader.
112 * This is a stream of dword pairs. First dword contains the
113 * register address, the second dword contains the value.*/
114 unsigned char *config;
115 unsigned config_size;
116
117 /** The number of bytes of config information for each global symbol.
118 */
119 unsigned config_size_per_symbol;
120
121 /** Constant data accessed by the shader. This will be uploaded
122 * into a constant buffer. */
123 unsigned char *rodata;
124 unsigned rodata_size;
125
126 /** List of symbol offsets for the shader */
127 uint64_t *global_symbol_offsets;
128 unsigned global_symbol_count;
129
130 struct radeon_shader_reloc *relocs;
131 unsigned reloc_count;
132
133 /** Disassembled shader in a string. */
134 char *disasm_string;
135 };
136
137 void radeon_shader_binary_init(struct radeon_shader_binary *b);
138 void radeon_shader_binary_clean(struct radeon_shader_binary *b);
139
140 struct r600_resource {
141 struct u_resource b;
142
143 /* Winsys objects. */
144 struct pb_buffer *buf;
145 uint64_t gpu_address;
146
147 /* Resource state. */
148 enum radeon_bo_domain domains;
149
150 /* The buffer range which is initialized (with a write transfer,
151 * streamout, DMA, or as a random access target). The rest of
152 * the buffer is considered invalid and can be mapped unsynchronized.
153 *
154 * This allows unsychronized mapping of a buffer range which hasn't
155 * been used yet. It's for applications which forget to use
156 * the unsynchronized map flag and expect the driver to figure it out.
157 */
158 struct util_range valid_buffer_range;
159
160 /* For buffers only. This indicates that a write operation has been
161 * performed by TC L2, but the cache hasn't been flushed.
162 * Any hw block which doesn't use or bypasses TC L2 should check this
163 * flag and flush the cache before using the buffer.
164 *
165 * For example, TC L2 must be flushed if a buffer which has been
166 * modified by a shader store instruction is about to be used as
167 * an index buffer. The reason is that VGT DMA index fetching doesn't
168 * use TC L2.
169 */
170 bool TC_L2_dirty;
171
172 /* Whether the resource has been exported via resource_get_handle. */
173 bool is_shared;
174 unsigned external_usage; /* PIPE_HANDLE_USAGE_* */
175 };
176
177 struct r600_transfer {
178 struct pipe_transfer transfer;
179 struct r600_resource *staging;
180 unsigned offset;
181 };
182
183 struct r600_fmask_info {
184 unsigned offset;
185 unsigned size;
186 unsigned alignment;
187 unsigned pitch_in_pixels;
188 unsigned bank_height;
189 unsigned slice_tile_max;
190 unsigned tile_mode_index;
191 };
192
193 struct r600_cmask_info {
194 unsigned offset;
195 unsigned size;
196 unsigned alignment;
197 unsigned pitch;
198 unsigned height;
199 unsigned xalign;
200 unsigned yalign;
201 unsigned slice_tile_max;
202 unsigned base_address_reg;
203 };
204
205 struct r600_htile_info {
206 unsigned pitch;
207 unsigned height;
208 unsigned xalign;
209 unsigned yalign;
210 };
211
212 struct r600_texture {
213 struct r600_resource resource;
214
215 unsigned size;
216 bool is_depth;
217 unsigned dirty_level_mask; /* each bit says if that mipmap is compressed */
218 unsigned stencil_dirty_level_mask; /* each bit says if that mipmap is compressed */
219 struct r600_texture *flushed_depth_texture;
220 boolean is_flushing_texture;
221 struct radeon_surf surface;
222
223 /* Colorbuffer compression and fast clear. */
224 struct r600_fmask_info fmask;
225 struct r600_cmask_info cmask;
226 struct r600_resource *cmask_buffer;
227 unsigned dcc_offset; /* 0 = disabled */
228 unsigned cb_color_info; /* fast clear enable bit */
229 unsigned color_clear_value[2];
230
231 /* Depth buffer compression and fast clear. */
232 struct r600_htile_info htile;
233 struct r600_resource *htile_buffer;
234 bool depth_cleared; /* if it was cleared at least once */
235 float depth_clear_value;
236 bool stencil_cleared; /* if it was cleared at least once */
237 uint8_t stencil_clear_value;
238
239 bool non_disp_tiling; /* R600-Cayman only */
240 };
241
242 struct r600_surface {
243 struct pipe_surface base;
244
245 bool color_initialized;
246 bool depth_initialized;
247
248 /* Misc. color flags. */
249 bool alphatest_bypass;
250 bool export_16bpc;
251 bool color_is_int8;
252
253 /* Color registers. */
254 unsigned cb_color_info;
255 unsigned cb_color_base;
256 unsigned cb_color_view;
257 unsigned cb_color_size; /* R600 only */
258 unsigned cb_color_dim; /* EG only */
259 unsigned cb_color_pitch; /* EG and later */
260 unsigned cb_color_slice; /* EG and later */
261 unsigned cb_dcc_base; /* VI and later */
262 unsigned cb_color_attrib; /* EG and later */
263 unsigned cb_dcc_control; /* VI and later */
264 unsigned cb_color_fmask; /* CB_COLORn_FMASK (EG and later) or CB_COLORn_FRAG (r600) */
265 unsigned cb_color_fmask_slice; /* EG and later */
266 unsigned cb_color_cmask; /* CB_COLORn_TILE (r600 only) */
267 unsigned cb_color_mask; /* R600 only */
268 unsigned spi_shader_col_format; /* SI+, no blending, no alpha-to-coverage. */
269 unsigned spi_shader_col_format_alpha; /* SI+, alpha-to-coverage */
270 unsigned spi_shader_col_format_blend; /* SI+, blending without alpha. */
271 unsigned spi_shader_col_format_blend_alpha; /* SI+, blending with alpha. */
272 struct r600_resource *cb_buffer_fmask; /* Used for FMASK relocations. R600 only */
273 struct r600_resource *cb_buffer_cmask; /* Used for CMASK relocations. R600 only */
274
275 /* DB registers. */
276 unsigned db_depth_info; /* R600 only, then SI and later */
277 unsigned db_z_info; /* EG and later */
278 unsigned db_depth_base; /* DB_Z_READ/WRITE_BASE (EG and later) or DB_DEPTH_BASE (r600) */
279 unsigned db_depth_view;
280 unsigned db_depth_size;
281 unsigned db_depth_slice; /* EG and later */
282 unsigned db_stencil_base; /* EG and later */
283 unsigned db_stencil_info; /* EG and later */
284 unsigned db_prefetch_limit; /* R600 only */
285 unsigned db_htile_surface;
286 unsigned db_htile_data_base;
287 unsigned db_preload_control; /* EG and later */
288 unsigned pa_su_poly_offset_db_fmt_cntl;
289 };
290
291 struct r600_common_screen {
292 struct pipe_screen b;
293 struct radeon_winsys *ws;
294 enum radeon_family family;
295 enum chip_class chip_class;
296 struct radeon_info info;
297 uint64_t debug_flags;
298 bool has_cp_dma;
299 bool has_streamout;
300
301 /* Auxiliary context. Mainly used to initialize resources.
302 * It must be locked prior to using and flushed before unlocking. */
303 struct pipe_context *aux_context;
304 pipe_mutex aux_context_lock;
305
306 /* This must be in the screen, because UE4 uses one context for
307 * compilation and another one for rendering.
308 */
309 unsigned num_compilations;
310 /* Along with ST_DEBUG=precompile, this should show if applications
311 * are loading shaders on demand. This is a monotonic counter.
312 */
313 unsigned num_shaders_created;
314
315 /* GPU load thread. */
316 pipe_mutex gpu_load_mutex;
317 pipe_thread gpu_load_thread;
318 unsigned gpu_load_counter_busy;
319 unsigned gpu_load_counter_idle;
320 volatile unsigned gpu_load_stop_thread; /* bool */
321
322 char renderer_string[64];
323
324 /* Performance counters. */
325 struct r600_perfcounters *perfcounters;
326
327 /* If pipe_screen wants to re-emit the framebuffer state of all
328 * contexts, it should atomically increment this. Each context will
329 * compare this with its own last known value of the counter before
330 * drawing and re-emit the framebuffer state accordingly.
331 */
332 unsigned dirty_fb_counter;
333
334 /* Atomically increment this counter when an existing texture's
335 * metadata is enabled or disabled in a way that requires changing
336 * contexts' compressed texture binding masks.
337 */
338 unsigned compressed_colortex_counter;
339
340 void (*query_opaque_metadata)(struct r600_common_screen *rscreen,
341 struct r600_texture *rtex,
342 struct radeon_bo_metadata *md);
343 };
344
345 /* This encapsulates a state or an operation which can emitted into the GPU
346 * command stream. */
347 struct r600_atom {
348 void (*emit)(struct r600_common_context *ctx, struct r600_atom *state);
349 unsigned num_dw;
350 unsigned short id;
351 };
352
353 struct r600_so_target {
354 struct pipe_stream_output_target b;
355
356 /* The buffer where BUFFER_FILLED_SIZE is stored. */
357 struct r600_resource *buf_filled_size;
358 unsigned buf_filled_size_offset;
359 bool buf_filled_size_valid;
360
361 unsigned stride_in_dw;
362 };
363
364 struct r600_streamout {
365 struct r600_atom begin_atom;
366 bool begin_emitted;
367 unsigned num_dw_for_end;
368
369 unsigned enabled_mask;
370 unsigned num_targets;
371 struct r600_so_target *targets[PIPE_MAX_SO_BUFFERS];
372
373 unsigned append_bitmask;
374 bool suspended;
375
376 /* External state which comes from the vertex shader,
377 * it must be set explicitly when binding a shader. */
378 unsigned *stride_in_dw;
379 unsigned enabled_stream_buffers_mask; /* stream0 buffers0-3 in 4 LSB */
380
381 /* The state of VGT_STRMOUT_BUFFER_(CONFIG|EN). */
382 unsigned hw_enabled_mask;
383
384 /* The state of VGT_STRMOUT_(CONFIG|EN). */
385 struct r600_atom enable_atom;
386 bool streamout_enabled;
387 bool prims_gen_query_enabled;
388 int num_prims_gen_queries;
389 };
390
391 struct r600_ring {
392 struct radeon_winsys_cs *cs;
393 void (*flush)(void *ctx, unsigned flags,
394 struct pipe_fence_handle **fence);
395 };
396
397 struct r600_common_context {
398 struct pipe_context b; /* base class */
399
400 struct r600_common_screen *screen;
401 struct radeon_winsys *ws;
402 struct radeon_winsys_ctx *ctx;
403 enum radeon_family family;
404 enum chip_class chip_class;
405 struct r600_ring gfx;
406 struct r600_ring dma;
407 struct pipe_fence_handle *last_sdma_fence;
408 unsigned initial_gfx_cs_size;
409 unsigned gpu_reset_counter;
410 unsigned last_dirty_fb_counter;
411 unsigned last_compressed_colortex_counter;
412
413 struct u_upload_mgr *uploader;
414 struct u_suballocator *allocator_so_filled_size;
415 struct util_slab_mempool pool_transfers;
416
417 /* Current unaccounted memory usage. */
418 uint64_t vram;
419 uint64_t gtt;
420
421 /* States. */
422 struct r600_streamout streamout;
423
424 /* Additional context states. */
425 unsigned flags; /* flush flags */
426
427 /* Queries. */
428 /* The list of active queries. Only one query of each type can be active. */
429 int num_occlusion_queries;
430 /* Keep track of non-timer queries, because they should be suspended
431 * during context flushing.
432 * The timer queries (TIME_ELAPSED) shouldn't be suspended for blits,
433 * but they should be suspended between IBs. */
434 struct list_head active_nontimer_queries;
435 struct list_head active_timer_queries;
436 unsigned num_cs_dw_nontimer_queries_suspend;
437 bool nontimer_queries_suspended_by_flush;
438 unsigned num_cs_dw_timer_queries_suspend;
439 /* Additional hardware info. */
440 unsigned backend_mask;
441 unsigned max_db; /* for OQ */
442 /* Misc stats. */
443 unsigned num_draw_calls;
444
445 /* Render condition. */
446 struct r600_atom render_cond_atom;
447 struct pipe_query *render_cond;
448 unsigned render_cond_mode;
449 boolean render_cond_invert;
450 bool render_cond_force_off; /* for u_blitter */
451
452 /* MSAA sample locations.
453 * The first index is the sample index.
454 * The second index is the coordinate: X, Y. */
455 float sample_locations_1x[1][2];
456 float sample_locations_2x[2][2];
457 float sample_locations_4x[4][2];
458 float sample_locations_8x[8][2];
459 float sample_locations_16x[16][2];
460
461 /* The list of all texture buffer objects in this context.
462 * This list is walked when a buffer is invalidated/reallocated and
463 * the GPU addresses are updated. */
464 struct list_head texture_buffers;
465
466 struct pipe_debug_callback debug;
467
468 /* Copy one resource to another using async DMA. */
469 void (*dma_copy)(struct pipe_context *ctx,
470 struct pipe_resource *dst,
471 unsigned dst_level,
472 unsigned dst_x, unsigned dst_y, unsigned dst_z,
473 struct pipe_resource *src,
474 unsigned src_level,
475 const struct pipe_box *src_box);
476
477 void (*clear_buffer)(struct pipe_context *ctx, struct pipe_resource *dst,
478 unsigned offset, unsigned size, unsigned value,
479 bool is_framebuffer);
480
481 void (*blit_decompress_depth)(struct pipe_context *ctx,
482 struct r600_texture *texture,
483 struct r600_texture *staging,
484 unsigned first_level, unsigned last_level,
485 unsigned first_layer, unsigned last_layer,
486 unsigned first_sample, unsigned last_sample);
487
488 void (*decompress_dcc)(struct pipe_context *ctx,
489 struct r600_texture *rtex);
490
491 /* Reallocate the buffer and update all resource bindings where
492 * the buffer is bound, including all resource descriptors. */
493 void (*invalidate_buffer)(struct pipe_context *ctx, struct pipe_resource *buf);
494
495 /* Enable or disable occlusion queries. */
496 void (*set_occlusion_query_state)(struct pipe_context *ctx, bool enable);
497
498 /* This ensures there is enough space in the command stream. */
499 void (*need_gfx_cs_space)(struct pipe_context *ctx, unsigned num_dw,
500 bool include_draw_vbo);
501
502 void (*set_atom_dirty)(struct r600_common_context *ctx,
503 struct r600_atom *atom, bool dirty);
504 };
505
506 /* r600_buffer.c */
507 boolean r600_rings_is_buffer_referenced(struct r600_common_context *ctx,
508 struct pb_buffer *buf,
509 enum radeon_bo_usage usage);
510 void *r600_buffer_map_sync_with_rings(struct r600_common_context *ctx,
511 struct r600_resource *resource,
512 unsigned usage);
513 bool r600_init_resource(struct r600_common_screen *rscreen,
514 struct r600_resource *res,
515 unsigned size, unsigned alignment,
516 bool use_reusable_pool);
517 struct pipe_resource *r600_buffer_create(struct pipe_screen *screen,
518 const struct pipe_resource *templ,
519 unsigned alignment);
520 struct pipe_resource * r600_aligned_buffer_create(struct pipe_screen *screen,
521 unsigned bind,
522 unsigned usage,
523 unsigned size,
524 unsigned alignment);
525 struct pipe_resource *
526 r600_buffer_from_user_memory(struct pipe_screen *screen,
527 const struct pipe_resource *templ,
528 void *user_memory);
529 void
530 r600_invalidate_resource(struct pipe_context *ctx,
531 struct pipe_resource *resource);
532
533 /* r600_common_pipe.c */
534 void r600_draw_rectangle(struct blitter_context *blitter,
535 int x1, int y1, int x2, int y2, float depth,
536 enum blitter_attrib_type type,
537 const union pipe_color_union *attrib);
538 bool r600_common_screen_init(struct r600_common_screen *rscreen,
539 struct radeon_winsys *ws);
540 void r600_destroy_common_screen(struct r600_common_screen *rscreen);
541 void r600_preflush_suspend_features(struct r600_common_context *ctx);
542 void r600_postflush_resume_features(struct r600_common_context *ctx);
543 bool r600_common_context_init(struct r600_common_context *rctx,
544 struct r600_common_screen *rscreen);
545 void r600_common_context_cleanup(struct r600_common_context *rctx);
546 void r600_context_add_resource_size(struct pipe_context *ctx, struct pipe_resource *r);
547 bool r600_can_dump_shader(struct r600_common_screen *rscreen,
548 unsigned processor);
549 void r600_screen_clear_buffer(struct r600_common_screen *rscreen, struct pipe_resource *dst,
550 unsigned offset, unsigned size, unsigned value,
551 bool is_framebuffer);
552 struct pipe_resource *r600_resource_create_common(struct pipe_screen *screen,
553 const struct pipe_resource *templ);
554 const char *r600_get_llvm_processor_name(enum radeon_family family);
555 void r600_need_dma_space(struct r600_common_context *ctx, unsigned num_dw);
556
557 /* r600_gpu_load.c */
558 void r600_gpu_load_kill_thread(struct r600_common_screen *rscreen);
559 uint64_t r600_gpu_load_begin(struct r600_common_screen *rscreen);
560 unsigned r600_gpu_load_end(struct r600_common_screen *rscreen, uint64_t begin);
561
562 /* r600_perfcounters.c */
563 void r600_perfcounters_destroy(struct r600_common_screen *rscreen);
564
565 /* r600_query.c */
566 void r600_init_screen_query_functions(struct r600_common_screen *rscreen);
567 void r600_query_init(struct r600_common_context *rctx);
568 void r600_suspend_nontimer_queries(struct r600_common_context *ctx);
569 void r600_resume_nontimer_queries(struct r600_common_context *ctx);
570 void r600_suspend_timer_queries(struct r600_common_context *ctx);
571 void r600_resume_timer_queries(struct r600_common_context *ctx);
572 void r600_query_init_backend_mask(struct r600_common_context *ctx);
573
574 /* r600_streamout.c */
575 void r600_streamout_buffers_dirty(struct r600_common_context *rctx);
576 void r600_set_streamout_targets(struct pipe_context *ctx,
577 unsigned num_targets,
578 struct pipe_stream_output_target **targets,
579 const unsigned *offset);
580 void r600_emit_streamout_end(struct r600_common_context *rctx);
581 void r600_update_prims_generated_query_state(struct r600_common_context *rctx,
582 unsigned type, int diff);
583 void r600_streamout_init(struct r600_common_context *rctx);
584
585 /* r600_texture.c */
586 void r600_texture_get_fmask_info(struct r600_common_screen *rscreen,
587 struct r600_texture *rtex,
588 unsigned nr_samples,
589 struct r600_fmask_info *out);
590 void r600_texture_get_cmask_info(struct r600_common_screen *rscreen,
591 struct r600_texture *rtex,
592 struct r600_cmask_info *out);
593 bool r600_init_flushed_depth_texture(struct pipe_context *ctx,
594 struct pipe_resource *texture,
595 struct r600_texture **staging);
596 void r600_print_texture_info(struct r600_texture *rtex, FILE *f);
597 struct pipe_resource *r600_texture_create(struct pipe_screen *screen,
598 const struct pipe_resource *templ);
599 struct pipe_surface *r600_create_surface_custom(struct pipe_context *pipe,
600 struct pipe_resource *texture,
601 const struct pipe_surface *templ,
602 unsigned width, unsigned height);
603 unsigned r600_translate_colorswap(enum pipe_format format);
604 void evergreen_do_fast_color_clear(struct r600_common_context *rctx,
605 struct pipe_framebuffer_state *fb,
606 struct r600_atom *fb_state,
607 unsigned *buffers, unsigned *dirty_cbufs,
608 const union pipe_color_union *color);
609 void r600_texture_disable_dcc(struct r600_common_screen *rscreen,
610 struct r600_texture *rtex);
611 void r600_init_screen_texture_functions(struct r600_common_screen *rscreen);
612 void r600_init_context_texture_functions(struct r600_common_context *rctx);
613
614 /* cayman_msaa.c */
615 extern const uint32_t eg_sample_locs_2x[4];
616 extern const unsigned eg_max_dist_2x;
617 extern const uint32_t eg_sample_locs_4x[4];
618 extern const unsigned eg_max_dist_4x;
619 void cayman_get_sample_position(struct pipe_context *ctx, unsigned sample_count,
620 unsigned sample_index, float *out_value);
621 void cayman_init_msaa(struct pipe_context *ctx);
622 void cayman_emit_msaa_sample_locs(struct radeon_winsys_cs *cs, int nr_samples);
623 void cayman_emit_msaa_config(struct radeon_winsys_cs *cs, int nr_samples,
624 int ps_iter_samples, int overrast_samples);
625
626
627 /* Inline helpers. */
628
629 static inline struct r600_resource *r600_resource(struct pipe_resource *r)
630 {
631 return (struct r600_resource*)r;
632 }
633
634 static inline void
635 r600_resource_reference(struct r600_resource **ptr, struct r600_resource *res)
636 {
637 pipe_resource_reference((struct pipe_resource **)ptr,
638 (struct pipe_resource *)res);
639 }
640
641 static inline unsigned r600_tex_aniso_filter(unsigned filter)
642 {
643 if (filter <= 1) return 0;
644 if (filter <= 2) return 1;
645 if (filter <= 4) return 2;
646 if (filter <= 8) return 3;
647 /* else */ return 4;
648 }
649
650 static inline unsigned r600_wavefront_size(enum radeon_family family)
651 {
652 switch (family) {
653 case CHIP_RV610:
654 case CHIP_RS780:
655 case CHIP_RV620:
656 case CHIP_RS880:
657 return 16;
658 case CHIP_RV630:
659 case CHIP_RV635:
660 case CHIP_RV730:
661 case CHIP_RV710:
662 case CHIP_PALM:
663 case CHIP_CEDAR:
664 return 32;
665 default:
666 return 64;
667 }
668 }
669
670 static inline enum radeon_bo_priority
671 r600_get_sampler_view_priority(struct r600_resource *res)
672 {
673 if (res->b.b.target == PIPE_BUFFER)
674 return RADEON_PRIO_SAMPLER_BUFFER;
675
676 if (res->b.b.nr_samples > 1)
677 return RADEON_PRIO_SAMPLER_TEXTURE_MSAA;
678
679 return RADEON_PRIO_SAMPLER_TEXTURE;
680 }
681
682 #define COMPUTE_DBG(rscreen, fmt, args...) \
683 do { \
684 if ((rscreen->b.debug_flags & DBG_COMPUTE)) fprintf(stderr, fmt, ##args); \
685 } while (0);
686
687 #define R600_ERR(fmt, args...) \
688 fprintf(stderr, "EE %s:%d %s - " fmt, __FILE__, __LINE__, __func__, ##args)
689
690 /* For MSAA sample positions. */
691 #define FILL_SREG(s0x, s0y, s1x, s1y, s2x, s2y, s3x, s3y) \
692 (((s0x) & 0xf) | (((s0y) & 0xf) << 4) | \
693 (((s1x) & 0xf) << 8) | (((s1y) & 0xf) << 12) | \
694 (((s2x) & 0xf) << 16) | (((s2y) & 0xf) << 20) | \
695 (((s3x) & 0xf) << 24) | (((s3y) & 0xf) << 28))
696
697 #endif