Merge ../mesa 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 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
103 #define R600_MAP_BUFFER_ALIGNMENT 64
104
105 struct r600_common_context;
106
107 struct radeon_shader_reloc {
108 char *name;
109 uint64_t offset;
110 };
111
112 struct radeon_shader_binary {
113 /** Shader code */
114 unsigned char *code;
115 unsigned code_size;
116
117 /** Config/Context register state that accompanies this shader.
118 * This is a stream of dword pairs. First dword contains the
119 * register address, the second dword contains the value.*/
120 unsigned char *config;
121 unsigned config_size;
122
123 /** The number of bytes of config information for each global symbol.
124 */
125 unsigned config_size_per_symbol;
126
127 /** Constant data accessed by the shader. This will be uploaded
128 * into a constant buffer. */
129 unsigned char *rodata;
130 unsigned rodata_size;
131
132 /** List of symbol offsets for the shader */
133 uint64_t *global_symbol_offsets;
134 unsigned global_symbol_count;
135
136 struct radeon_shader_reloc *relocs;
137 unsigned reloc_count;
138
139 /** Disassembled shader in a string. */
140 char *disasm_string;
141 };
142
143 struct r600_resource {
144 struct u_resource b;
145
146 /* Winsys objects. */
147 struct pb_buffer *buf;
148 struct radeon_winsys_cs_handle *cs_buf;
149 uint64_t gpu_address;
150
151 /* Resource state. */
152 enum radeon_bo_domain domains;
153
154 /* The buffer range which is initialized (with a write transfer,
155 * streamout, DMA, or as a random access target). The rest of
156 * the buffer is considered invalid and can be mapped unsynchronized.
157 *
158 * This allows unsychronized mapping of a buffer range which hasn't
159 * been used yet. It's for applications which forget to use
160 * the unsynchronized map flag and expect the driver to figure it out.
161 */
162 struct util_range valid_buffer_range;
163
164 /* For buffers only. This indicates that a write operation has been
165 * performed by TC L2, but the cache hasn't been flushed.
166 * Any hw block which doesn't use or bypasses TC L2 should check this
167 * flag and flush the cache before using the buffer.
168 *
169 * For example, TC L2 must be flushed if a buffer which has been
170 * modified by a shader store instruction is about to be used as
171 * an index buffer. The reason is that VGT DMA index fetching doesn't
172 * use TC L2.
173 */
174 bool TC_L2_dirty;
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;
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 slice_tile_max;
198 unsigned base_address_reg;
199 };
200
201 struct r600_texture {
202 struct r600_resource resource;
203
204 unsigned size;
205 unsigned pitch_override;
206 bool is_depth;
207 unsigned dirty_level_mask; /* each bit says if that mipmap is compressed */
208 unsigned stencil_dirty_level_mask; /* each bit says if that mipmap is compressed */
209 struct r600_texture *flushed_depth_texture;
210 boolean is_flushing_texture;
211 struct radeon_surf surface;
212
213 /* Colorbuffer compression and fast clear. */
214 struct r600_fmask_info fmask;
215 struct r600_cmask_info cmask;
216 struct r600_resource *cmask_buffer;
217 unsigned cb_color_info; /* fast clear enable bit */
218 unsigned color_clear_value[2];
219
220 /* Depth buffer compression and fast clear. */
221 struct r600_resource *htile_buffer;
222 bool depth_cleared; /* if it was cleared at least once */
223 float depth_clear_value;
224
225 bool non_disp_tiling; /* R600-Cayman only */
226 };
227
228 struct r600_surface {
229 struct pipe_surface base;
230
231 bool color_initialized;
232 bool depth_initialized;
233
234 /* Misc. color flags. */
235 bool alphatest_bypass;
236 bool export_16bpc;
237
238 /* Color registers. */
239 unsigned cb_color_info;
240 unsigned cb_color_base;
241 unsigned cb_color_view;
242 unsigned cb_color_size; /* R600 only */
243 unsigned cb_color_dim; /* EG only */
244 unsigned cb_color_pitch; /* EG and later */
245 unsigned cb_color_slice; /* EG and later */
246 unsigned cb_color_attrib; /* EG and later */
247 unsigned cb_dcc_control; /* VI and later */
248 unsigned cb_color_fmask; /* CB_COLORn_FMASK (EG and later) or CB_COLORn_FRAG (r600) */
249 unsigned cb_color_fmask_slice; /* EG and later */
250 unsigned cb_color_cmask; /* CB_COLORn_TILE (r600 only) */
251 unsigned cb_color_mask; /* R600 only */
252 struct r600_resource *cb_buffer_fmask; /* Used for FMASK relocations. R600 only */
253 struct r600_resource *cb_buffer_cmask; /* Used for CMASK relocations. R600 only */
254
255 /* DB registers. */
256 unsigned db_depth_info; /* R600 only, then SI and later */
257 unsigned db_z_info; /* EG and later */
258 unsigned db_depth_base; /* DB_Z_READ/WRITE_BASE (EG and later) or DB_DEPTH_BASE (r600) */
259 unsigned db_depth_view;
260 unsigned db_depth_size;
261 unsigned db_depth_slice; /* EG and later */
262 unsigned db_stencil_base; /* EG and later */
263 unsigned db_stencil_info; /* EG and later */
264 unsigned db_prefetch_limit; /* R600 only */
265 unsigned db_htile_surface;
266 unsigned db_htile_data_base;
267 unsigned db_preload_control; /* EG and later */
268 unsigned pa_su_poly_offset_db_fmt_cntl;
269 };
270
271 struct r600_tiling_info {
272 unsigned num_channels;
273 unsigned num_banks;
274 unsigned group_bytes;
275 };
276
277 struct r600_common_screen {
278 struct pipe_screen b;
279 struct radeon_winsys *ws;
280 enum radeon_family family;
281 enum chip_class chip_class;
282 struct radeon_info info;
283 struct r600_tiling_info tiling_info;
284 uint64_t debug_flags;
285 bool has_cp_dma;
286 bool has_streamout;
287
288 /* Auxiliary context. Mainly used to initialize resources.
289 * It must be locked prior to using and flushed before unlocking. */
290 struct pipe_context *aux_context;
291 pipe_mutex aux_context_lock;
292
293 struct r600_resource *trace_bo;
294 uint32_t *trace_ptr;
295 unsigned cs_count;
296
297 /* This must be in the screen, because UE4 uses one context for
298 * compilation and another one for rendering.
299 */
300 unsigned num_compilations;
301 /* Along with ST_DEBUG=precompile, this should show if applications
302 * are loading shaders on demand. This is a monotonic counter.
303 */
304 unsigned num_shaders_created;
305
306 /* GPU load thread. */
307 pipe_mutex gpu_load_mutex;
308 pipe_thread gpu_load_thread;
309 unsigned gpu_load_counter_busy;
310 unsigned gpu_load_counter_idle;
311 volatile unsigned gpu_load_stop_thread; /* bool */
312
313 char renderer_string[64];
314 };
315
316 /* This encapsulates a state or an operation which can emitted into the GPU
317 * command stream. */
318 struct r600_atom {
319 void (*emit)(struct r600_common_context *ctx, struct r600_atom *state);
320 unsigned num_dw;
321 unsigned short id;
322 };
323
324 struct r600_so_target {
325 struct pipe_stream_output_target b;
326
327 /* The buffer where BUFFER_FILLED_SIZE is stored. */
328 struct r600_resource *buf_filled_size;
329 unsigned buf_filled_size_offset;
330 bool buf_filled_size_valid;
331
332 unsigned stride_in_dw;
333 };
334
335 struct r600_streamout {
336 struct r600_atom begin_atom;
337 bool begin_emitted;
338 unsigned num_dw_for_end;
339
340 unsigned enabled_mask;
341 unsigned num_targets;
342 struct r600_so_target *targets[PIPE_MAX_SO_BUFFERS];
343
344 unsigned append_bitmask;
345 bool suspended;
346
347 /* External state which comes from the vertex shader,
348 * it must be set explicitly when binding a shader. */
349 unsigned *stride_in_dw;
350 unsigned enabled_stream_buffers_mask; /* stream0 buffers0-3 in 4 LSB */
351
352 /* The state of VGT_STRMOUT_BUFFER_(CONFIG|EN). */
353 unsigned hw_enabled_mask;
354
355 /* The state of VGT_STRMOUT_(CONFIG|EN). */
356 struct r600_atom enable_atom;
357 bool streamout_enabled;
358 bool prims_gen_query_enabled;
359 int num_prims_gen_queries;
360 };
361
362 struct r600_ring {
363 struct radeon_winsys_cs *cs;
364 bool flushing;
365 void (*flush)(void *ctx, unsigned flags,
366 struct pipe_fence_handle **fence);
367 };
368
369 struct r600_rings {
370 struct r600_ring gfx;
371 struct r600_ring dma;
372 };
373
374 struct r600_common_context {
375 struct pipe_context b; /* base class */
376
377 struct r600_common_screen *screen;
378 struct radeon_winsys *ws;
379 struct radeon_winsys_ctx *ctx;
380 enum radeon_family family;
381 enum chip_class chip_class;
382 struct r600_rings rings;
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 pipe_query *current_render_cond;
421 unsigned current_render_cond_mode;
422 boolean current_render_cond_cond;
423 boolean predicate_drawing;
424 /* For context flushing. */
425 struct pipe_query *saved_render_cond;
426 boolean saved_render_cond_cond;
427 unsigned saved_render_cond_mode;
428
429 /* MSAA sample locations.
430 * The first index is the sample index.
431 * The second index is the coordinate: X, Y. */
432 float sample_locations_1x[1][2];
433 float sample_locations_2x[2][2];
434 float sample_locations_4x[4][2];
435 float sample_locations_8x[8][2];
436 float sample_locations_16x[16][2];
437
438 /* The list of all texture buffer objects in this context.
439 * This list is walked when a buffer is invalidated/reallocated and
440 * the GPU addresses are updated. */
441 struct list_head texture_buffers;
442
443 /* Copy one resource to another using async DMA. */
444 void (*dma_copy)(struct pipe_context *ctx,
445 struct pipe_resource *dst,
446 unsigned dst_level,
447 unsigned dst_x, unsigned dst_y, unsigned dst_z,
448 struct pipe_resource *src,
449 unsigned src_level,
450 const struct pipe_box *src_box);
451
452 void (*clear_buffer)(struct pipe_context *ctx, struct pipe_resource *dst,
453 unsigned offset, unsigned size, unsigned value,
454 bool is_framebuffer);
455
456 void (*blit_decompress_depth)(struct pipe_context *ctx,
457 struct r600_texture *texture,
458 struct r600_texture *staging,
459 unsigned first_level, unsigned last_level,
460 unsigned first_layer, unsigned last_layer,
461 unsigned first_sample, unsigned last_sample);
462
463 /* Reallocate the buffer and update all resource bindings where
464 * the buffer is bound, including all resource descriptors. */
465 void (*invalidate_buffer)(struct pipe_context *ctx, struct pipe_resource *buf);
466
467 /* Enable or disable occlusion queries. */
468 void (*set_occlusion_query_state)(struct pipe_context *ctx, bool enable);
469
470 /* This ensures there is enough space in the command stream. */
471 void (*need_gfx_cs_space)(struct pipe_context *ctx, unsigned num_dw,
472 bool include_draw_vbo);
473
474 void (*set_atom_dirty)(struct r600_common_context *ctx,
475 struct r600_atom *atom, bool dirty);
476 };
477
478 /* r600_buffer.c */
479 boolean r600_rings_is_buffer_referenced(struct r600_common_context *ctx,
480 struct radeon_winsys_cs_handle *buf,
481 enum radeon_bo_usage usage);
482 void *r600_buffer_map_sync_with_rings(struct r600_common_context *ctx,
483 struct r600_resource *resource,
484 unsigned usage);
485 bool r600_init_resource(struct r600_common_screen *rscreen,
486 struct r600_resource *res,
487 unsigned size, unsigned alignment,
488 bool use_reusable_pool);
489 struct pipe_resource *r600_buffer_create(struct pipe_screen *screen,
490 const struct pipe_resource *templ,
491 unsigned alignment);
492 struct pipe_resource *
493 r600_buffer_from_user_memory(struct pipe_screen *screen,
494 const struct pipe_resource *templ,
495 void *user_memory);
496
497 /* r600_common_pipe.c */
498 void r600_draw_rectangle(struct blitter_context *blitter,
499 int x1, int y1, int x2, int y2, float depth,
500 enum blitter_attrib_type type,
501 const union pipe_color_union *attrib);
502 bool r600_common_screen_init(struct r600_common_screen *rscreen,
503 struct radeon_winsys *ws);
504 void r600_destroy_common_screen(struct r600_common_screen *rscreen);
505 void r600_preflush_suspend_features(struct r600_common_context *ctx);
506 void r600_postflush_resume_features(struct r600_common_context *ctx);
507 bool r600_common_context_init(struct r600_common_context *rctx,
508 struct r600_common_screen *rscreen);
509 void r600_common_context_cleanup(struct r600_common_context *rctx);
510 void r600_context_add_resource_size(struct pipe_context *ctx, struct pipe_resource *r);
511 bool r600_can_dump_shader(struct r600_common_screen *rscreen,
512 const struct tgsi_token *tokens);
513 void r600_screen_clear_buffer(struct r600_common_screen *rscreen, struct pipe_resource *dst,
514 unsigned offset, unsigned size, unsigned value,
515 bool is_framebuffer);
516 struct pipe_resource *r600_resource_create_common(struct pipe_screen *screen,
517 const struct pipe_resource *templ);
518 const char *r600_get_llvm_processor_name(enum radeon_family family);
519 void r600_need_dma_space(struct r600_common_context *ctx, unsigned num_dw);
520
521 /* r600_gpu_load.c */
522 void r600_gpu_load_kill_thread(struct r600_common_screen *rscreen);
523 uint64_t r600_gpu_load_begin(struct r600_common_screen *rscreen);
524 unsigned r600_gpu_load_end(struct r600_common_screen *rscreen, uint64_t begin);
525
526 /* r600_query.c */
527 void r600_query_init(struct r600_common_context *rctx);
528 void r600_suspend_nontimer_queries(struct r600_common_context *ctx);
529 void r600_resume_nontimer_queries(struct r600_common_context *ctx);
530 void r600_suspend_timer_queries(struct r600_common_context *ctx);
531 void r600_resume_timer_queries(struct r600_common_context *ctx);
532 void r600_query_init_backend_mask(struct r600_common_context *ctx);
533
534 /* r600_streamout.c */
535 void r600_streamout_buffers_dirty(struct r600_common_context *rctx);
536 void r600_set_streamout_targets(struct pipe_context *ctx,
537 unsigned num_targets,
538 struct pipe_stream_output_target **targets,
539 const unsigned *offset);
540 void r600_emit_streamout_end(struct r600_common_context *rctx);
541 void r600_update_prims_generated_query_state(struct r600_common_context *rctx,
542 unsigned type, int diff);
543 void r600_streamout_init(struct r600_common_context *rctx);
544
545 /* r600_texture.c */
546 void r600_texture_get_fmask_info(struct r600_common_screen *rscreen,
547 struct r600_texture *rtex,
548 unsigned nr_samples,
549 struct r600_fmask_info *out);
550 void r600_texture_get_cmask_info(struct r600_common_screen *rscreen,
551 struct r600_texture *rtex,
552 struct r600_cmask_info *out);
553 bool r600_init_flushed_depth_texture(struct pipe_context *ctx,
554 struct pipe_resource *texture,
555 struct r600_texture **staging);
556 struct pipe_resource *r600_texture_create(struct pipe_screen *screen,
557 const struct pipe_resource *templ);
558 struct pipe_surface *r600_create_surface_custom(struct pipe_context *pipe,
559 struct pipe_resource *texture,
560 const struct pipe_surface *templ,
561 unsigned width, unsigned height);
562 unsigned r600_translate_colorswap(enum pipe_format format);
563 void evergreen_do_fast_color_clear(struct r600_common_context *rctx,
564 struct pipe_framebuffer_state *fb,
565 struct r600_atom *fb_state,
566 unsigned *buffers, unsigned *dirty_cbufs,
567 const union pipe_color_union *color);
568 void r600_init_screen_texture_functions(struct r600_common_screen *rscreen);
569 void r600_init_context_texture_functions(struct r600_common_context *rctx);
570
571 /* cayman_msaa.c */
572 extern const uint32_t eg_sample_locs_2x[4];
573 extern const unsigned eg_max_dist_2x;
574 extern const uint32_t eg_sample_locs_4x[4];
575 extern const unsigned eg_max_dist_4x;
576 void cayman_get_sample_position(struct pipe_context *ctx, unsigned sample_count,
577 unsigned sample_index, float *out_value);
578 void cayman_init_msaa(struct pipe_context *ctx);
579 void cayman_emit_msaa_sample_locs(struct radeon_winsys_cs *cs, int nr_samples);
580 void cayman_emit_msaa_config(struct radeon_winsys_cs *cs, int nr_samples,
581 int ps_iter_samples, int overrast_samples);
582
583
584 /* Inline helpers. */
585
586 static inline struct r600_resource *r600_resource(struct pipe_resource *r)
587 {
588 return (struct r600_resource*)r;
589 }
590
591 static inline void
592 r600_resource_reference(struct r600_resource **ptr, struct r600_resource *res)
593 {
594 pipe_resource_reference((struct pipe_resource **)ptr,
595 (struct pipe_resource *)res);
596 }
597
598 static inline unsigned r600_tex_aniso_filter(unsigned filter)
599 {
600 if (filter <= 1) return 0;
601 if (filter <= 2) return 1;
602 if (filter <= 4) return 2;
603 if (filter <= 8) return 3;
604 /* else */ return 4;
605 }
606
607 static inline unsigned r600_wavefront_size(enum radeon_family family)
608 {
609 switch (family) {
610 case CHIP_RV610:
611 case CHIP_RS780:
612 case CHIP_RV620:
613 case CHIP_RS880:
614 return 16;
615 case CHIP_RV630:
616 case CHIP_RV635:
617 case CHIP_RV730:
618 case CHIP_RV710:
619 case CHIP_PALM:
620 case CHIP_CEDAR:
621 return 32;
622 default:
623 return 64;
624 }
625 }
626
627 static inline enum radeon_bo_priority
628 r600_get_sampler_view_priority(struct r600_resource *res)
629 {
630 if (res->b.b.target == PIPE_BUFFER)
631 return RADEON_PRIO_SAMPLER_BUFFER;
632
633 if (res->b.b.nr_samples > 1)
634 return RADEON_PRIO_SAMPLER_TEXTURE_MSAA;
635
636 return RADEON_PRIO_SAMPLER_TEXTURE;
637 }
638
639 #define COMPUTE_DBG(rscreen, fmt, args...) \
640 do { \
641 if ((rscreen->b.debug_flags & DBG_COMPUTE)) fprintf(stderr, fmt, ##args); \
642 } while (0);
643
644 #define R600_ERR(fmt, args...) \
645 fprintf(stderr, "EE %s:%d %s - "fmt, __FILE__, __LINE__, __func__, ##args)
646
647 /* For MSAA sample positions. */
648 #define FILL_SREG(s0x, s0y, s1x, s1y, s2x, s2y, s3x, s3y) \
649 (((s0x) & 0xf) | (((s0y) & 0xf) << 4) | \
650 (((s1x) & 0xf) << 8) | (((s1y) & 0xf) << 12) | \
651 (((s2x) & 0xf) << 16) | (((s2y) & 0xf) << 20) | \
652 (((s3x) & 0xf) << 24) | (((s3y) & 0xf) << 28))
653
654 #endif