d945e53cbc30b018a4e5202844a39ddcef7362bc
[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 R600_RESOURCE_FLAG_TRANSFER (PIPE_RESOURCE_FLAG_DRV_PRIV << 0)
47 #define R600_RESOURCE_FLAG_FLUSHED_DEPTH (PIPE_RESOURCE_FLAG_DRV_PRIV << 1)
48 #define R600_RESOURCE_FLAG_FORCE_TILING (PIPE_RESOURCE_FLAG_DRV_PRIV << 2)
49
50 #define R600_QUERY_DRAW_CALLS (PIPE_QUERY_DRIVER_SPECIFIC + 0)
51 #define R600_QUERY_REQUESTED_VRAM (PIPE_QUERY_DRIVER_SPECIFIC + 1)
52 #define R600_QUERY_REQUESTED_GTT (PIPE_QUERY_DRIVER_SPECIFIC + 2)
53 #define R600_QUERY_BUFFER_WAIT_TIME (PIPE_QUERY_DRIVER_SPECIFIC + 3)
54 #define R600_QUERY_NUM_CS_FLUSHES (PIPE_QUERY_DRIVER_SPECIFIC + 4)
55 #define R600_QUERY_NUM_BYTES_MOVED (PIPE_QUERY_DRIVER_SPECIFIC + 5)
56 #define R600_QUERY_VRAM_USAGE (PIPE_QUERY_DRIVER_SPECIFIC + 6)
57 #define R600_QUERY_GTT_USAGE (PIPE_QUERY_DRIVER_SPECIFIC + 7)
58 #define R600_QUERY_GPU_TEMPERATURE (PIPE_QUERY_DRIVER_SPECIFIC + 8)
59 #define R600_QUERY_CURRENT_GPU_SCLK (PIPE_QUERY_DRIVER_SPECIFIC + 9)
60 #define R600_QUERY_CURRENT_GPU_MCLK (PIPE_QUERY_DRIVER_SPECIFIC + 10)
61 #define R600_QUERY_GPU_LOAD (PIPE_QUERY_DRIVER_SPECIFIC + 11)
62 #define R600_QUERY_NUM_COMPILATIONS (PIPE_QUERY_DRIVER_SPECIFIC + 12)
63 #define R600_QUERY_NUM_SHADERS_CREATED (PIPE_QUERY_DRIVER_SPECIFIC + 13)
64
65 #define R600_CONTEXT_STREAMOUT_FLUSH (1u << 0)
66 #define R600_CONTEXT_PRIVATE_FLAG (1u << 1)
67
68 /* special primitive types */
69 #define R600_PRIM_RECTANGLE_LIST PIPE_PRIM_MAX
70
71 /* Debug flags. */
72 /* logging */
73 #define DBG_TEX (1 << 0)
74 #define DBG_TEXMIP (1 << 1)
75 #define DBG_COMPUTE (1 << 2)
76 #define DBG_VM (1 << 3)
77 #define DBG_TRACE_CS (1 << 4)
78 /* shader logging */
79 #define DBG_FS (1 << 5)
80 #define DBG_VS (1 << 6)
81 #define DBG_GS (1 << 7)
82 #define DBG_PS (1 << 8)
83 #define DBG_CS (1 << 9)
84 #define DBG_TCS (1 << 10)
85 #define DBG_TES (1 << 11)
86 #define DBG_NO_IR (1 << 12)
87 #define DBG_NO_TGSI (1 << 13)
88 #define DBG_NO_ASM (1 << 14)
89 /* Bits 21-31 are reserved for the r600g driver. */
90 /* features */
91 #define DBG_NO_ASYNC_DMA (1llu << 32)
92 #define DBG_NO_HYPERZ (1llu << 33)
93 #define DBG_NO_DISCARD_RANGE (1llu << 34)
94 #define DBG_NO_2D_TILING (1llu << 35)
95 #define DBG_NO_TILING (1llu << 36)
96 #define DBG_SWITCH_ON_EOP (1llu << 37)
97 #define DBG_FORCE_DMA (1llu << 38)
98 #define DBG_PRECOMPILE (1llu << 39)
99 #define DBG_INFO (1llu << 40)
100 #define DBG_NO_WC (1llu << 41)
101 #define DBG_CHECK_VM (1llu << 42)
102 #define DBG_NO_DCC (1llu << 43)
103 #define DBG_NO_DCC_CLEAR (1llu << 44)
104
105 #define R600_MAP_BUFFER_ALIGNMENT 64
106
107 struct r600_common_context;
108
109 struct radeon_shader_reloc {
110 char *name;
111 uint64_t offset;
112 };
113
114 struct radeon_shader_binary {
115 /** Shader code */
116 unsigned char *code;
117 unsigned code_size;
118
119 /** Config/Context register state that accompanies this shader.
120 * This is a stream of dword pairs. First dword contains the
121 * register address, the second dword contains the value.*/
122 unsigned char *config;
123 unsigned config_size;
124
125 /** The number of bytes of config information for each global symbol.
126 */
127 unsigned config_size_per_symbol;
128
129 /** Constant data accessed by the shader. This will be uploaded
130 * into a constant buffer. */
131 unsigned char *rodata;
132 unsigned rodata_size;
133
134 /** List of symbol offsets for the shader */
135 uint64_t *global_symbol_offsets;
136 unsigned global_symbol_count;
137
138 struct radeon_shader_reloc *relocs;
139 unsigned reloc_count;
140
141 /** Disassembled shader in a string. */
142 char *disasm_string;
143 };
144
145 struct r600_resource {
146 struct u_resource b;
147
148 /* Winsys objects. */
149 struct pb_buffer *buf;
150 struct radeon_winsys_cs_handle *cs_buf;
151 uint64_t gpu_address;
152
153 /* Resource state. */
154 enum radeon_bo_domain domains;
155
156 /* The buffer range which is initialized (with a write transfer,
157 * streamout, DMA, or as a random access target). The rest of
158 * the buffer is considered invalid and can be mapped unsynchronized.
159 *
160 * This allows unsychronized mapping of a buffer range which hasn't
161 * been used yet. It's for applications which forget to use
162 * the unsynchronized map flag and expect the driver to figure it out.
163 */
164 struct util_range valid_buffer_range;
165
166 /* For buffers only. This indicates that a write operation has been
167 * performed by TC L2, but the cache hasn't been flushed.
168 * Any hw block which doesn't use or bypasses TC L2 should check this
169 * flag and flush the cache before using the buffer.
170 *
171 * For example, TC L2 must be flushed if a buffer which has been
172 * modified by a shader store instruction is about to be used as
173 * an index buffer. The reason is that VGT DMA index fetching doesn't
174 * use TC L2.
175 */
176 bool TC_L2_dirty;
177 };
178
179 struct r600_transfer {
180 struct pipe_transfer transfer;
181 struct r600_resource *staging;
182 unsigned offset;
183 };
184
185 struct r600_fmask_info {
186 unsigned offset;
187 unsigned size;
188 unsigned alignment;
189 unsigned pitch;
190 unsigned bank_height;
191 unsigned slice_tile_max;
192 unsigned tile_mode_index;
193 };
194
195 struct r600_cmask_info {
196 unsigned offset;
197 unsigned size;
198 unsigned alignment;
199 unsigned slice_tile_max;
200 unsigned base_address_reg;
201 };
202
203 struct r600_texture {
204 struct r600_resource resource;
205
206 unsigned size;
207 unsigned pitch_override;
208 bool is_depth;
209 unsigned dirty_level_mask; /* each bit says if that mipmap is compressed */
210 unsigned stencil_dirty_level_mask; /* each bit says if that mipmap is compressed */
211 struct r600_texture *flushed_depth_texture;
212 boolean is_flushing_texture;
213 struct radeon_surf surface;
214
215 /* Colorbuffer compression and fast clear. */
216 struct r600_fmask_info fmask;
217 struct r600_cmask_info cmask;
218 struct r600_resource *cmask_buffer;
219 struct r600_resource *dcc_buffer;
220 unsigned cb_color_info; /* fast clear enable bit */
221 unsigned color_clear_value[2];
222
223 /* Depth buffer compression and fast clear. */
224 struct r600_resource *htile_buffer;
225 bool depth_cleared; /* if it was cleared at least once */
226 float depth_clear_value;
227
228 bool non_disp_tiling; /* R600-Cayman only */
229 };
230
231 struct r600_surface {
232 struct pipe_surface base;
233
234 bool color_initialized;
235 bool depth_initialized;
236
237 /* Misc. color flags. */
238 bool alphatest_bypass;
239 bool export_16bpc;
240
241 /* Color registers. */
242 unsigned cb_color_info;
243 unsigned cb_color_base;
244 unsigned cb_color_view;
245 unsigned cb_color_size; /* R600 only */
246 unsigned cb_color_dim; /* EG only */
247 unsigned cb_color_pitch; /* EG and later */
248 unsigned cb_color_slice; /* EG and later */
249 unsigned cb_dcc_base; /* VI and later */
250 unsigned cb_color_attrib; /* EG and later */
251 unsigned cb_dcc_control; /* VI and later */
252 unsigned cb_color_fmask; /* CB_COLORn_FMASK (EG and later) or CB_COLORn_FRAG (r600) */
253 unsigned cb_color_fmask_slice; /* EG and later */
254 unsigned cb_color_cmask; /* CB_COLORn_TILE (r600 only) */
255 unsigned cb_color_mask; /* R600 only */
256 struct r600_resource *cb_buffer_fmask; /* Used for FMASK relocations. R600 only */
257 struct r600_resource *cb_buffer_cmask; /* Used for CMASK relocations. R600 only */
258
259 /* DB registers. */
260 unsigned db_depth_info; /* R600 only, then SI and later */
261 unsigned db_z_info; /* EG and later */
262 unsigned db_depth_base; /* DB_Z_READ/WRITE_BASE (EG and later) or DB_DEPTH_BASE (r600) */
263 unsigned db_depth_view;
264 unsigned db_depth_size;
265 unsigned db_depth_slice; /* EG and later */
266 unsigned db_stencil_base; /* EG and later */
267 unsigned db_stencil_info; /* EG and later */
268 unsigned db_prefetch_limit; /* R600 only */
269 unsigned db_htile_surface;
270 unsigned db_htile_data_base;
271 unsigned db_preload_control; /* EG and later */
272 unsigned pa_su_poly_offset_db_fmt_cntl;
273 };
274
275 struct r600_tiling_info {
276 unsigned num_channels;
277 unsigned num_banks;
278 unsigned group_bytes;
279 };
280
281 struct r600_common_screen {
282 struct pipe_screen b;
283 struct radeon_winsys *ws;
284 enum radeon_family family;
285 enum chip_class chip_class;
286 struct radeon_info info;
287 struct r600_tiling_info tiling_info;
288 uint64_t debug_flags;
289 bool has_cp_dma;
290 bool has_streamout;
291
292 /* Auxiliary context. Mainly used to initialize resources.
293 * It must be locked prior to using and flushed before unlocking. */
294 struct pipe_context *aux_context;
295 pipe_mutex aux_context_lock;
296
297 struct r600_resource *trace_bo;
298 uint32_t *trace_ptr;
299 unsigned cs_count;
300
301 /* This must be in the screen, because UE4 uses one context for
302 * compilation and another one for rendering.
303 */
304 unsigned num_compilations;
305 /* Along with ST_DEBUG=precompile, this should show if applications
306 * are loading shaders on demand. This is a monotonic counter.
307 */
308 unsigned num_shaders_created;
309
310 /* GPU load thread. */
311 pipe_mutex gpu_load_mutex;
312 pipe_thread gpu_load_thread;
313 unsigned gpu_load_counter_busy;
314 unsigned gpu_load_counter_idle;
315 volatile unsigned gpu_load_stop_thread; /* bool */
316
317 char renderer_string[64];
318 };
319
320 /* This encapsulates a state or an operation which can emitted into the GPU
321 * command stream. */
322 struct r600_atom {
323 void (*emit)(struct r600_common_context *ctx, struct r600_atom *state);
324 unsigned num_dw;
325 unsigned short id;
326 };
327
328 struct r600_so_target {
329 struct pipe_stream_output_target b;
330
331 /* The buffer where BUFFER_FILLED_SIZE is stored. */
332 struct r600_resource *buf_filled_size;
333 unsigned buf_filled_size_offset;
334 bool buf_filled_size_valid;
335
336 unsigned stride_in_dw;
337 };
338
339 struct r600_streamout {
340 struct r600_atom begin_atom;
341 bool begin_emitted;
342 unsigned num_dw_for_end;
343
344 unsigned enabled_mask;
345 unsigned num_targets;
346 struct r600_so_target *targets[PIPE_MAX_SO_BUFFERS];
347
348 unsigned append_bitmask;
349 bool suspended;
350
351 /* External state which comes from the vertex shader,
352 * it must be set explicitly when binding a shader. */
353 unsigned *stride_in_dw;
354 unsigned enabled_stream_buffers_mask; /* stream0 buffers0-3 in 4 LSB */
355
356 /* The state of VGT_STRMOUT_BUFFER_(CONFIG|EN). */
357 unsigned hw_enabled_mask;
358
359 /* The state of VGT_STRMOUT_(CONFIG|EN). */
360 struct r600_atom enable_atom;
361 bool streamout_enabled;
362 bool prims_gen_query_enabled;
363 int num_prims_gen_queries;
364 };
365
366 struct r600_ring {
367 struct radeon_winsys_cs *cs;
368 void (*flush)(void *ctx, unsigned flags,
369 struct pipe_fence_handle **fence);
370 };
371
372 struct r600_common_context {
373 struct pipe_context b; /* base class */
374
375 struct r600_common_screen *screen;
376 struct radeon_winsys *ws;
377 struct radeon_winsys_ctx *ctx;
378 enum radeon_family family;
379 enum chip_class chip_class;
380 struct r600_ring gfx;
381 struct r600_ring dma;
382 struct pipe_fence_handle *last_sdma_fence;
383 unsigned initial_gfx_cs_size;
384 unsigned gpu_reset_counter;
385
386 struct u_upload_mgr *uploader;
387 struct u_suballocator *allocator_so_filled_size;
388 struct util_slab_mempool pool_transfers;
389
390 /* Current unaccounted memory usage. */
391 uint64_t vram;
392 uint64_t gtt;
393
394 /* States. */
395 struct r600_streamout streamout;
396
397 /* Additional context states. */
398 unsigned flags; /* flush flags */
399
400 /* Queries. */
401 /* The list of active queries. Only one query of each type can be active. */
402 int num_occlusion_queries;
403 /* Keep track of non-timer queries, because they should be suspended
404 * during context flushing.
405 * The timer queries (TIME_ELAPSED) shouldn't be suspended for blits,
406 * but they should be suspended between IBs. */
407 struct list_head active_nontimer_queries;
408 struct list_head active_timer_queries;
409 unsigned num_cs_dw_nontimer_queries_suspend;
410 unsigned num_cs_dw_timer_queries_suspend;
411 /* If queries have been suspended. */
412 bool queries_suspended_for_flush;
413 /* Additional hardware info. */
414 unsigned backend_mask;
415 unsigned max_db; /* for OQ */
416 /* Misc stats. */
417 unsigned num_draw_calls;
418
419 /* Render condition. */
420 struct r600_atom render_cond_atom;
421 struct pipe_query *render_cond;
422 unsigned render_cond_mode;
423 boolean render_cond_invert;
424 bool render_cond_force_off; /* for u_blitter */
425
426 /* MSAA sample locations.
427 * The first index is the sample index.
428 * The second index is the coordinate: X, Y. */
429 float sample_locations_1x[1][2];
430 float sample_locations_2x[2][2];
431 float sample_locations_4x[4][2];
432 float sample_locations_8x[8][2];
433 float sample_locations_16x[16][2];
434
435 /* The list of all texture buffer objects in this context.
436 * This list is walked when a buffer is invalidated/reallocated and
437 * the GPU addresses are updated. */
438 struct list_head texture_buffers;
439
440 /* Copy one resource to another using async DMA. */
441 void (*dma_copy)(struct pipe_context *ctx,
442 struct pipe_resource *dst,
443 unsigned dst_level,
444 unsigned dst_x, unsigned dst_y, unsigned dst_z,
445 struct pipe_resource *src,
446 unsigned src_level,
447 const struct pipe_box *src_box);
448
449 void (*clear_buffer)(struct pipe_context *ctx, struct pipe_resource *dst,
450 unsigned offset, unsigned size, unsigned value,
451 bool is_framebuffer);
452
453 void (*blit_decompress_depth)(struct pipe_context *ctx,
454 struct r600_texture *texture,
455 struct r600_texture *staging,
456 unsigned first_level, unsigned last_level,
457 unsigned first_layer, unsigned last_layer,
458 unsigned first_sample, unsigned last_sample);
459
460 /* Reallocate the buffer and update all resource bindings where
461 * the buffer is bound, including all resource descriptors. */
462 void (*invalidate_buffer)(struct pipe_context *ctx, struct pipe_resource *buf);
463
464 /* Enable or disable occlusion queries. */
465 void (*set_occlusion_query_state)(struct pipe_context *ctx, bool enable);
466
467 /* This ensures there is enough space in the command stream. */
468 void (*need_gfx_cs_space)(struct pipe_context *ctx, unsigned num_dw,
469 bool include_draw_vbo);
470
471 void (*set_atom_dirty)(struct r600_common_context *ctx,
472 struct r600_atom *atom, bool dirty);
473 };
474
475 /* r600_buffer.c */
476 boolean r600_rings_is_buffer_referenced(struct r600_common_context *ctx,
477 struct radeon_winsys_cs_handle *buf,
478 enum radeon_bo_usage usage);
479 void *r600_buffer_map_sync_with_rings(struct r600_common_context *ctx,
480 struct r600_resource *resource,
481 unsigned usage);
482 bool r600_init_resource(struct r600_common_screen *rscreen,
483 struct r600_resource *res,
484 unsigned size, unsigned alignment,
485 bool use_reusable_pool);
486 struct pipe_resource *r600_buffer_create(struct pipe_screen *screen,
487 const struct pipe_resource *templ,
488 unsigned alignment);
489 struct pipe_resource * r600_aligned_buffer_create(struct pipe_screen *screen,
490 unsigned bind,
491 unsigned usage,
492 unsigned size,
493 unsigned alignment);
494 struct pipe_resource *
495 r600_buffer_from_user_memory(struct pipe_screen *screen,
496 const struct pipe_resource *templ,
497 void *user_memory);
498
499 /* r600_common_pipe.c */
500 void r600_draw_rectangle(struct blitter_context *blitter,
501 int x1, int y1, int x2, int y2, float depth,
502 enum blitter_attrib_type type,
503 const union pipe_color_union *attrib);
504 bool r600_common_screen_init(struct r600_common_screen *rscreen,
505 struct radeon_winsys *ws);
506 void r600_destroy_common_screen(struct r600_common_screen *rscreen);
507 void r600_preflush_suspend_features(struct r600_common_context *ctx);
508 void r600_postflush_resume_features(struct r600_common_context *ctx);
509 bool r600_common_context_init(struct r600_common_context *rctx,
510 struct r600_common_screen *rscreen);
511 void r600_common_context_cleanup(struct r600_common_context *rctx);
512 void r600_context_add_resource_size(struct pipe_context *ctx, struct pipe_resource *r);
513 bool r600_can_dump_shader(struct r600_common_screen *rscreen,
514 const struct tgsi_token *tokens);
515 void r600_screen_clear_buffer(struct r600_common_screen *rscreen, struct pipe_resource *dst,
516 unsigned offset, unsigned size, unsigned value,
517 bool is_framebuffer);
518 struct pipe_resource *r600_resource_create_common(struct pipe_screen *screen,
519 const struct pipe_resource *templ);
520 const char *r600_get_llvm_processor_name(enum radeon_family family);
521 void r600_need_dma_space(struct r600_common_context *ctx, unsigned num_dw);
522
523 /* r600_gpu_load.c */
524 void r600_gpu_load_kill_thread(struct r600_common_screen *rscreen);
525 uint64_t r600_gpu_load_begin(struct r600_common_screen *rscreen);
526 unsigned r600_gpu_load_end(struct r600_common_screen *rscreen, uint64_t begin);
527
528 /* r600_query.c */
529 void r600_init_screen_query_functions(struct r600_common_screen *rscreen);
530 void r600_query_init(struct r600_common_context *rctx);
531 void r600_suspend_nontimer_queries(struct r600_common_context *ctx);
532 void r600_resume_nontimer_queries(struct r600_common_context *ctx);
533 void r600_suspend_timer_queries(struct r600_common_context *ctx);
534 void r600_resume_timer_queries(struct r600_common_context *ctx);
535 void r600_query_init_backend_mask(struct r600_common_context *ctx);
536
537 /* r600_streamout.c */
538 void r600_streamout_buffers_dirty(struct r600_common_context *rctx);
539 void r600_set_streamout_targets(struct pipe_context *ctx,
540 unsigned num_targets,
541 struct pipe_stream_output_target **targets,
542 const unsigned *offset);
543 void r600_emit_streamout_end(struct r600_common_context *rctx);
544 void r600_update_prims_generated_query_state(struct r600_common_context *rctx,
545 unsigned type, int diff);
546 void r600_streamout_init(struct r600_common_context *rctx);
547
548 /* r600_texture.c */
549 void r600_texture_get_fmask_info(struct r600_common_screen *rscreen,
550 struct r600_texture *rtex,
551 unsigned nr_samples,
552 struct r600_fmask_info *out);
553 void r600_texture_get_cmask_info(struct r600_common_screen *rscreen,
554 struct r600_texture *rtex,
555 struct r600_cmask_info *out);
556 bool r600_init_flushed_depth_texture(struct pipe_context *ctx,
557 struct pipe_resource *texture,
558 struct r600_texture **staging);
559 struct pipe_resource *r600_texture_create(struct pipe_screen *screen,
560 const struct pipe_resource *templ);
561 struct pipe_surface *r600_create_surface_custom(struct pipe_context *pipe,
562 struct pipe_resource *texture,
563 const struct pipe_surface *templ,
564 unsigned width, unsigned height);
565 unsigned r600_translate_colorswap(enum pipe_format format);
566 void evergreen_do_fast_color_clear(struct r600_common_context *rctx,
567 struct pipe_framebuffer_state *fb,
568 struct r600_atom *fb_state,
569 unsigned *buffers, unsigned *dirty_cbufs,
570 const union pipe_color_union *color);
571 void r600_init_screen_texture_functions(struct r600_common_screen *rscreen);
572 void r600_init_context_texture_functions(struct r600_common_context *rctx);
573
574 /* cayman_msaa.c */
575 extern const uint32_t eg_sample_locs_2x[4];
576 extern const unsigned eg_max_dist_2x;
577 extern const uint32_t eg_sample_locs_4x[4];
578 extern const unsigned eg_max_dist_4x;
579 void cayman_get_sample_position(struct pipe_context *ctx, unsigned sample_count,
580 unsigned sample_index, float *out_value);
581 void cayman_init_msaa(struct pipe_context *ctx);
582 void cayman_emit_msaa_sample_locs(struct radeon_winsys_cs *cs, int nr_samples);
583 void cayman_emit_msaa_config(struct radeon_winsys_cs *cs, int nr_samples,
584 int ps_iter_samples, int overrast_samples);
585
586
587 /* Inline helpers. */
588
589 static inline struct r600_resource *r600_resource(struct pipe_resource *r)
590 {
591 return (struct r600_resource*)r;
592 }
593
594 static inline void
595 r600_resource_reference(struct r600_resource **ptr, struct r600_resource *res)
596 {
597 pipe_resource_reference((struct pipe_resource **)ptr,
598 (struct pipe_resource *)res);
599 }
600
601 static inline unsigned r600_tex_aniso_filter(unsigned filter)
602 {
603 if (filter <= 1) return 0;
604 if (filter <= 2) return 1;
605 if (filter <= 4) return 2;
606 if (filter <= 8) return 3;
607 /* else */ return 4;
608 }
609
610 static inline unsigned r600_wavefront_size(enum radeon_family family)
611 {
612 switch (family) {
613 case CHIP_RV610:
614 case CHIP_RS780:
615 case CHIP_RV620:
616 case CHIP_RS880:
617 return 16;
618 case CHIP_RV630:
619 case CHIP_RV635:
620 case CHIP_RV730:
621 case CHIP_RV710:
622 case CHIP_PALM:
623 case CHIP_CEDAR:
624 return 32;
625 default:
626 return 64;
627 }
628 }
629
630 static inline enum radeon_bo_priority
631 r600_get_sampler_view_priority(struct r600_resource *res)
632 {
633 if (res->b.b.target == PIPE_BUFFER)
634 return RADEON_PRIO_SAMPLER_BUFFER;
635
636 if (res->b.b.nr_samples > 1)
637 return RADEON_PRIO_SAMPLER_TEXTURE_MSAA;
638
639 return RADEON_PRIO_SAMPLER_TEXTURE;
640 }
641
642 #define COMPUTE_DBG(rscreen, fmt, args...) \
643 do { \
644 if ((rscreen->b.debug_flags & DBG_COMPUTE)) fprintf(stderr, fmt, ##args); \
645 } while (0);
646
647 #define R600_ERR(fmt, args...) \
648 fprintf(stderr, "EE %s:%d %s - "fmt, __FILE__, __LINE__, __func__, ##args)
649
650 /* For MSAA sample positions. */
651 #define FILL_SREG(s0x, s0y, s1x, s1y, s2x, s2y, s3x, s3y) \
652 (((s0x) & 0xf) | (((s0y) & 0xf) << 4) | \
653 (((s1x) & 0xf) << 8) | (((s1y) & 0xf) << 12) | \
654 (((s2x) & 0xf) << 16) | (((s2y) & 0xf) << 20) | \
655 (((s3x) & 0xf) << 24) | (((s3y) & 0xf) << 28))
656
657 #endif