gallium/radeon: set num_banks in the winsys
[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_CONTEXT_STREAMOUT_FLUSH (1u << 0)
51 #define R600_CONTEXT_PRIVATE_FLAG (1u << 1)
52
53 /* special primitive types */
54 #define R600_PRIM_RECTANGLE_LIST PIPE_PRIM_MAX
55
56 /* Debug flags. */
57 /* logging */
58 #define DBG_TEX (1 << 0)
59 /* gap - reuse */
60 #define DBG_COMPUTE (1 << 2)
61 #define DBG_VM (1 << 3)
62 #define DBG_TRACE_CS (1 << 4)
63 /* shader logging */
64 #define DBG_FS (1 << 5)
65 #define DBG_VS (1 << 6)
66 #define DBG_GS (1 << 7)
67 #define DBG_PS (1 << 8)
68 #define DBG_CS (1 << 9)
69 #define DBG_TCS (1 << 10)
70 #define DBG_TES (1 << 11)
71 #define DBG_NO_IR (1 << 12)
72 #define DBG_NO_TGSI (1 << 13)
73 #define DBG_NO_ASM (1 << 14)
74 #define DBG_PREOPT_IR (1 << 15)
75 /* Bits 21-31 are reserved for the r600g driver. */
76 /* features */
77 #define DBG_NO_ASYNC_DMA (1llu << 32)
78 #define DBG_NO_HYPERZ (1llu << 33)
79 #define DBG_NO_DISCARD_RANGE (1llu << 34)
80 #define DBG_NO_2D_TILING (1llu << 35)
81 #define DBG_NO_TILING (1llu << 36)
82 #define DBG_SWITCH_ON_EOP (1llu << 37)
83 #define DBG_FORCE_DMA (1llu << 38)
84 #define DBG_PRECOMPILE (1llu << 39)
85 #define DBG_INFO (1llu << 40)
86 #define DBG_NO_WC (1llu << 41)
87 #define DBG_CHECK_VM (1llu << 42)
88 #define DBG_NO_DCC (1llu << 43)
89 #define DBG_NO_DCC_CLEAR (1llu << 44)
90 #define DBG_NO_RB_PLUS (1llu << 45)
91 #define DBG_SI_SCHED (1llu << 46)
92
93 #define R600_MAP_BUFFER_ALIGNMENT 64
94
95 struct r600_common_context;
96 struct r600_perfcounters;
97
98 struct radeon_shader_reloc {
99 char *name;
100 uint64_t offset;
101 };
102
103 struct radeon_shader_binary {
104 /** Shader code */
105 unsigned char *code;
106 unsigned code_size;
107
108 /** Config/Context register state that accompanies this shader.
109 * This is a stream of dword pairs. First dword contains the
110 * register address, the second dword contains the value.*/
111 unsigned char *config;
112 unsigned config_size;
113
114 /** The number of bytes of config information for each global symbol.
115 */
116 unsigned config_size_per_symbol;
117
118 /** Constant data accessed by the shader. This will be uploaded
119 * into a constant buffer. */
120 unsigned char *rodata;
121 unsigned rodata_size;
122
123 /** List of symbol offsets for the shader */
124 uint64_t *global_symbol_offsets;
125 unsigned global_symbol_count;
126
127 struct radeon_shader_reloc *relocs;
128 unsigned reloc_count;
129
130 /** Disassembled shader in a string. */
131 char *disasm_string;
132 };
133
134 void radeon_shader_binary_init(struct radeon_shader_binary *b);
135 void radeon_shader_binary_clean(struct radeon_shader_binary *b);
136
137 struct r600_resource {
138 struct u_resource b;
139
140 /* Winsys objects. */
141 struct pb_buffer *buf;
142 uint64_t gpu_address;
143
144 /* Resource state. */
145 enum radeon_bo_domain domains;
146
147 /* The buffer range which is initialized (with a write transfer,
148 * streamout, DMA, or as a random access target). The rest of
149 * the buffer is considered invalid and can be mapped unsynchronized.
150 *
151 * This allows unsychronized mapping of a buffer range which hasn't
152 * been used yet. It's for applications which forget to use
153 * the unsynchronized map flag and expect the driver to figure it out.
154 */
155 struct util_range valid_buffer_range;
156
157 /* For buffers only. This indicates that a write operation has been
158 * performed by TC L2, but the cache hasn't been flushed.
159 * Any hw block which doesn't use or bypasses TC L2 should check this
160 * flag and flush the cache before using the buffer.
161 *
162 * For example, TC L2 must be flushed if a buffer which has been
163 * modified by a shader store instruction is about to be used as
164 * an index buffer. The reason is that VGT DMA index fetching doesn't
165 * use TC L2.
166 */
167 bool TC_L2_dirty;
168 };
169
170 struct r600_transfer {
171 struct pipe_transfer transfer;
172 struct r600_resource *staging;
173 unsigned offset;
174 };
175
176 struct r600_fmask_info {
177 unsigned offset;
178 unsigned size;
179 unsigned alignment;
180 unsigned pitch_in_pixels;
181 unsigned bank_height;
182 unsigned slice_tile_max;
183 unsigned tile_mode_index;
184 };
185
186 struct r600_cmask_info {
187 unsigned offset;
188 unsigned size;
189 unsigned alignment;
190 unsigned pitch;
191 unsigned height;
192 unsigned xalign;
193 unsigned yalign;
194 unsigned slice_tile_max;
195 unsigned base_address_reg;
196 };
197
198 struct r600_htile_info {
199 unsigned pitch;
200 unsigned height;
201 unsigned xalign;
202 unsigned yalign;
203 };
204
205 struct r600_texture {
206 struct r600_resource resource;
207
208 unsigned size;
209 bool is_depth;
210 unsigned dirty_level_mask; /* each bit says if that mipmap is compressed */
211 unsigned stencil_dirty_level_mask; /* each bit says if that mipmap is compressed */
212 struct r600_texture *flushed_depth_texture;
213 boolean is_flushing_texture;
214 struct radeon_surf surface;
215
216 /* Colorbuffer compression and fast clear. */
217 struct r600_fmask_info fmask;
218 struct r600_cmask_info cmask;
219 struct r600_resource *cmask_buffer;
220 struct r600_resource *dcc_buffer;
221 unsigned cb_color_info; /* fast clear enable bit */
222 unsigned color_clear_value[2];
223
224 /* Depth buffer compression and fast clear. */
225 struct r600_htile_info htile;
226 struct r600_resource *htile_buffer;
227 bool depth_cleared; /* if it was cleared at least once */
228 float depth_clear_value;
229 bool stencil_cleared; /* if it was cleared at least once */
230 uint8_t stencil_clear_value;
231
232 bool non_disp_tiling; /* R600-Cayman only */
233 };
234
235 struct r600_surface {
236 struct pipe_surface base;
237
238 bool color_initialized;
239 bool depth_initialized;
240
241 /* Misc. color flags. */
242 bool alphatest_bypass;
243 bool export_16bpc;
244 bool color_is_int8;
245
246 /* Color registers. */
247 unsigned cb_color_info;
248 unsigned cb_color_base;
249 unsigned cb_color_view;
250 unsigned cb_color_size; /* R600 only */
251 unsigned cb_color_dim; /* EG only */
252 unsigned cb_color_pitch; /* EG and later */
253 unsigned cb_color_slice; /* EG and later */
254 unsigned cb_dcc_base; /* VI and later */
255 unsigned cb_color_attrib; /* EG and later */
256 unsigned cb_dcc_control; /* VI and later */
257 unsigned cb_color_fmask; /* CB_COLORn_FMASK (EG and later) or CB_COLORn_FRAG (r600) */
258 unsigned cb_color_fmask_slice; /* EG and later */
259 unsigned cb_color_cmask; /* CB_COLORn_TILE (r600 only) */
260 unsigned cb_color_mask; /* R600 only */
261 unsigned spi_shader_col_format; /* SI+, no blending, no alpha-to-coverage. */
262 unsigned spi_shader_col_format_alpha; /* SI+, alpha-to-coverage */
263 unsigned spi_shader_col_format_blend; /* SI+, blending without alpha. */
264 unsigned spi_shader_col_format_blend_alpha; /* SI+, blending with alpha. */
265 struct r600_resource *cb_buffer_fmask; /* Used for FMASK relocations. R600 only */
266 struct r600_resource *cb_buffer_cmask; /* Used for CMASK relocations. R600 only */
267
268 /* DB registers. */
269 unsigned db_depth_info; /* R600 only, then SI and later */
270 unsigned db_z_info; /* EG and later */
271 unsigned db_depth_base; /* DB_Z_READ/WRITE_BASE (EG and later) or DB_DEPTH_BASE (r600) */
272 unsigned db_depth_view;
273 unsigned db_depth_size;
274 unsigned db_depth_slice; /* EG and later */
275 unsigned db_stencil_base; /* EG and later */
276 unsigned db_stencil_info; /* EG and later */
277 unsigned db_prefetch_limit; /* R600 only */
278 unsigned db_htile_surface;
279 unsigned db_htile_data_base;
280 unsigned db_preload_control; /* EG and later */
281 unsigned pa_su_poly_offset_db_fmt_cntl;
282 };
283
284 struct r600_tiling_info {
285 unsigned group_bytes;
286 };
287
288 struct r600_common_screen {
289 struct pipe_screen b;
290 struct radeon_winsys *ws;
291 enum radeon_family family;
292 enum chip_class chip_class;
293 struct radeon_info info;
294 struct r600_tiling_info tiling_info;
295 uint64_t debug_flags;
296 bool has_cp_dma;
297 bool has_streamout;
298
299 /* Auxiliary context. Mainly used to initialize resources.
300 * It must be locked prior to using and flushed before unlocking. */
301 struct pipe_context *aux_context;
302 pipe_mutex aux_context_lock;
303
304 struct r600_resource *trace_bo;
305 uint32_t *trace_ptr;
306 unsigned cs_count;
307
308 /* This must be in the screen, because UE4 uses one context for
309 * compilation and another one for rendering.
310 */
311 unsigned num_compilations;
312 /* Along with ST_DEBUG=precompile, this should show if applications
313 * are loading shaders on demand. This is a monotonic counter.
314 */
315 unsigned num_shaders_created;
316
317 /* GPU load thread. */
318 pipe_mutex gpu_load_mutex;
319 pipe_thread gpu_load_thread;
320 unsigned gpu_load_counter_busy;
321 unsigned gpu_load_counter_idle;
322 volatile unsigned gpu_load_stop_thread; /* bool */
323
324 char renderer_string[64];
325
326 /* Performance counters. */
327 struct r600_perfcounters *perfcounters;
328 };
329
330 /* This encapsulates a state or an operation which can emitted into the GPU
331 * command stream. */
332 struct r600_atom {
333 void (*emit)(struct r600_common_context *ctx, struct r600_atom *state);
334 unsigned num_dw;
335 unsigned short id;
336 };
337
338 struct r600_so_target {
339 struct pipe_stream_output_target b;
340
341 /* The buffer where BUFFER_FILLED_SIZE is stored. */
342 struct r600_resource *buf_filled_size;
343 unsigned buf_filled_size_offset;
344 bool buf_filled_size_valid;
345
346 unsigned stride_in_dw;
347 };
348
349 struct r600_streamout {
350 struct r600_atom begin_atom;
351 bool begin_emitted;
352 unsigned num_dw_for_end;
353
354 unsigned enabled_mask;
355 unsigned num_targets;
356 struct r600_so_target *targets[PIPE_MAX_SO_BUFFERS];
357
358 unsigned append_bitmask;
359 bool suspended;
360
361 /* External state which comes from the vertex shader,
362 * it must be set explicitly when binding a shader. */
363 unsigned *stride_in_dw;
364 unsigned enabled_stream_buffers_mask; /* stream0 buffers0-3 in 4 LSB */
365
366 /* The state of VGT_STRMOUT_BUFFER_(CONFIG|EN). */
367 unsigned hw_enabled_mask;
368
369 /* The state of VGT_STRMOUT_(CONFIG|EN). */
370 struct r600_atom enable_atom;
371 bool streamout_enabled;
372 bool prims_gen_query_enabled;
373 int num_prims_gen_queries;
374 };
375
376 struct r600_ring {
377 struct radeon_winsys_cs *cs;
378 void (*flush)(void *ctx, unsigned flags,
379 struct pipe_fence_handle **fence);
380 };
381
382 struct r600_common_context {
383 struct pipe_context b; /* base class */
384
385 struct r600_common_screen *screen;
386 struct radeon_winsys *ws;
387 struct radeon_winsys_ctx *ctx;
388 enum radeon_family family;
389 enum chip_class chip_class;
390 struct r600_ring gfx;
391 struct r600_ring dma;
392 struct pipe_fence_handle *last_sdma_fence;
393 unsigned initial_gfx_cs_size;
394 unsigned gpu_reset_counter;
395
396 struct u_upload_mgr *uploader;
397 struct u_suballocator *allocator_so_filled_size;
398 struct util_slab_mempool pool_transfers;
399
400 /* Current unaccounted memory usage. */
401 uint64_t vram;
402 uint64_t gtt;
403
404 /* States. */
405 struct r600_streamout streamout;
406
407 /* Additional context states. */
408 unsigned flags; /* flush flags */
409
410 /* Queries. */
411 /* The list of active queries. Only one query of each type can be active. */
412 int num_occlusion_queries;
413 /* Keep track of non-timer queries, because they should be suspended
414 * during context flushing.
415 * The timer queries (TIME_ELAPSED) shouldn't be suspended for blits,
416 * but they should be suspended between IBs. */
417 struct list_head active_nontimer_queries;
418 struct list_head active_timer_queries;
419 unsigned num_cs_dw_nontimer_queries_suspend;
420 bool nontimer_queries_suspended_by_flush;
421 unsigned num_cs_dw_timer_queries_suspend;
422 /* Additional hardware info. */
423 unsigned backend_mask;
424 unsigned max_db; /* for OQ */
425 /* Misc stats. */
426 unsigned num_draw_calls;
427
428 /* Render condition. */
429 struct r600_atom render_cond_atom;
430 struct pipe_query *render_cond;
431 unsigned render_cond_mode;
432 boolean render_cond_invert;
433 bool render_cond_force_off; /* for u_blitter */
434
435 /* MSAA sample locations.
436 * The first index is the sample index.
437 * The second index is the coordinate: X, Y. */
438 float sample_locations_1x[1][2];
439 float sample_locations_2x[2][2];
440 float sample_locations_4x[4][2];
441 float sample_locations_8x[8][2];
442 float sample_locations_16x[16][2];
443
444 /* The list of all texture buffer objects in this context.
445 * This list is walked when a buffer is invalidated/reallocated and
446 * the GPU addresses are updated. */
447 struct list_head texture_buffers;
448
449 struct pipe_debug_callback debug;
450
451 /* Copy one resource to another using async DMA. */
452 void (*dma_copy)(struct pipe_context *ctx,
453 struct pipe_resource *dst,
454 unsigned dst_level,
455 unsigned dst_x, unsigned dst_y, unsigned dst_z,
456 struct pipe_resource *src,
457 unsigned src_level,
458 const struct pipe_box *src_box);
459
460 void (*clear_buffer)(struct pipe_context *ctx, struct pipe_resource *dst,
461 unsigned offset, unsigned size, unsigned value,
462 bool is_framebuffer);
463
464 void (*blit_decompress_depth)(struct pipe_context *ctx,
465 struct r600_texture *texture,
466 struct r600_texture *staging,
467 unsigned first_level, unsigned last_level,
468 unsigned first_layer, unsigned last_layer,
469 unsigned first_sample, unsigned last_sample);
470
471 /* Reallocate the buffer and update all resource bindings where
472 * the buffer is bound, including all resource descriptors. */
473 void (*invalidate_buffer)(struct pipe_context *ctx, struct pipe_resource *buf);
474
475 /* Enable or disable occlusion queries. */
476 void (*set_occlusion_query_state)(struct pipe_context *ctx, bool enable);
477
478 /* This ensures there is enough space in the command stream. */
479 void (*need_gfx_cs_space)(struct pipe_context *ctx, unsigned num_dw,
480 bool include_draw_vbo);
481
482 void (*set_atom_dirty)(struct r600_common_context *ctx,
483 struct r600_atom *atom, bool dirty);
484 };
485
486 /* r600_buffer.c */
487 boolean r600_rings_is_buffer_referenced(struct r600_common_context *ctx,
488 struct pb_buffer *buf,
489 enum radeon_bo_usage usage);
490 void *r600_buffer_map_sync_with_rings(struct r600_common_context *ctx,
491 struct r600_resource *resource,
492 unsigned usage);
493 bool r600_init_resource(struct r600_common_screen *rscreen,
494 struct r600_resource *res,
495 unsigned size, unsigned alignment,
496 bool use_reusable_pool);
497 struct pipe_resource *r600_buffer_create(struct pipe_screen *screen,
498 const struct pipe_resource *templ,
499 unsigned alignment);
500 struct pipe_resource * r600_aligned_buffer_create(struct pipe_screen *screen,
501 unsigned bind,
502 unsigned usage,
503 unsigned size,
504 unsigned alignment);
505 struct pipe_resource *
506 r600_buffer_from_user_memory(struct pipe_screen *screen,
507 const struct pipe_resource *templ,
508 void *user_memory);
509 void
510 r600_invalidate_resource(struct pipe_context *ctx,
511 struct pipe_resource *resource);
512
513 /* r600_common_pipe.c */
514 void r600_draw_rectangle(struct blitter_context *blitter,
515 int x1, int y1, int x2, int y2, float depth,
516 enum blitter_attrib_type type,
517 const union pipe_color_union *attrib);
518 bool r600_common_screen_init(struct r600_common_screen *rscreen,
519 struct radeon_winsys *ws);
520 void r600_destroy_common_screen(struct r600_common_screen *rscreen);
521 void r600_preflush_suspend_features(struct r600_common_context *ctx);
522 void r600_postflush_resume_features(struct r600_common_context *ctx);
523 bool r600_common_context_init(struct r600_common_context *rctx,
524 struct r600_common_screen *rscreen);
525 void r600_common_context_cleanup(struct r600_common_context *rctx);
526 void r600_context_add_resource_size(struct pipe_context *ctx, struct pipe_resource *r);
527 bool r600_can_dump_shader(struct r600_common_screen *rscreen,
528 unsigned processor);
529 void r600_screen_clear_buffer(struct r600_common_screen *rscreen, struct pipe_resource *dst,
530 unsigned offset, unsigned size, unsigned value,
531 bool is_framebuffer);
532 struct pipe_resource *r600_resource_create_common(struct pipe_screen *screen,
533 const struct pipe_resource *templ);
534 const char *r600_get_llvm_processor_name(enum radeon_family family);
535 void r600_need_dma_space(struct r600_common_context *ctx, unsigned num_dw);
536
537 /* r600_gpu_load.c */
538 void r600_gpu_load_kill_thread(struct r600_common_screen *rscreen);
539 uint64_t r600_gpu_load_begin(struct r600_common_screen *rscreen);
540 unsigned r600_gpu_load_end(struct r600_common_screen *rscreen, uint64_t begin);
541
542 /* r600_perfcounters.c */
543 void r600_perfcounters_destroy(struct r600_common_screen *rscreen);
544
545 /* r600_query.c */
546 void r600_init_screen_query_functions(struct r600_common_screen *rscreen);
547 void r600_query_init(struct r600_common_context *rctx);
548 void r600_suspend_nontimer_queries(struct r600_common_context *ctx);
549 void r600_resume_nontimer_queries(struct r600_common_context *ctx);
550 void r600_suspend_timer_queries(struct r600_common_context *ctx);
551 void r600_resume_timer_queries(struct r600_common_context *ctx);
552 void r600_query_init_backend_mask(struct r600_common_context *ctx);
553
554 /* r600_streamout.c */
555 void r600_streamout_buffers_dirty(struct r600_common_context *rctx);
556 void r600_set_streamout_targets(struct pipe_context *ctx,
557 unsigned num_targets,
558 struct pipe_stream_output_target **targets,
559 const unsigned *offset);
560 void r600_emit_streamout_end(struct r600_common_context *rctx);
561 void r600_update_prims_generated_query_state(struct r600_common_context *rctx,
562 unsigned type, int diff);
563 void r600_streamout_init(struct r600_common_context *rctx);
564
565 /* r600_texture.c */
566 void r600_texture_get_fmask_info(struct r600_common_screen *rscreen,
567 struct r600_texture *rtex,
568 unsigned nr_samples,
569 struct r600_fmask_info *out);
570 void r600_texture_get_cmask_info(struct r600_common_screen *rscreen,
571 struct r600_texture *rtex,
572 struct r600_cmask_info *out);
573 bool r600_init_flushed_depth_texture(struct pipe_context *ctx,
574 struct pipe_resource *texture,
575 struct r600_texture **staging);
576 void r600_print_texture_info(struct r600_texture *rtex, FILE *f);
577 struct pipe_resource *r600_texture_create(struct pipe_screen *screen,
578 const struct pipe_resource *templ);
579 struct pipe_surface *r600_create_surface_custom(struct pipe_context *pipe,
580 struct pipe_resource *texture,
581 const struct pipe_surface *templ,
582 unsigned width, unsigned height);
583 unsigned r600_translate_colorswap(enum pipe_format format);
584 void evergreen_do_fast_color_clear(struct r600_common_context *rctx,
585 struct pipe_framebuffer_state *fb,
586 struct r600_atom *fb_state,
587 unsigned *buffers, unsigned *dirty_cbufs,
588 const union pipe_color_union *color);
589 void r600_init_screen_texture_functions(struct r600_common_screen *rscreen);
590 void r600_init_context_texture_functions(struct r600_common_context *rctx);
591
592 /* cayman_msaa.c */
593 extern const uint32_t eg_sample_locs_2x[4];
594 extern const unsigned eg_max_dist_2x;
595 extern const uint32_t eg_sample_locs_4x[4];
596 extern const unsigned eg_max_dist_4x;
597 void cayman_get_sample_position(struct pipe_context *ctx, unsigned sample_count,
598 unsigned sample_index, float *out_value);
599 void cayman_init_msaa(struct pipe_context *ctx);
600 void cayman_emit_msaa_sample_locs(struct radeon_winsys_cs *cs, int nr_samples);
601 void cayman_emit_msaa_config(struct radeon_winsys_cs *cs, int nr_samples,
602 int ps_iter_samples, int overrast_samples);
603
604
605 /* Inline helpers. */
606
607 static inline struct r600_resource *r600_resource(struct pipe_resource *r)
608 {
609 return (struct r600_resource*)r;
610 }
611
612 static inline void
613 r600_resource_reference(struct r600_resource **ptr, struct r600_resource *res)
614 {
615 pipe_resource_reference((struct pipe_resource **)ptr,
616 (struct pipe_resource *)res);
617 }
618
619 static inline unsigned r600_tex_aniso_filter(unsigned filter)
620 {
621 if (filter <= 1) return 0;
622 if (filter <= 2) return 1;
623 if (filter <= 4) return 2;
624 if (filter <= 8) return 3;
625 /* else */ return 4;
626 }
627
628 static inline unsigned r600_wavefront_size(enum radeon_family family)
629 {
630 switch (family) {
631 case CHIP_RV610:
632 case CHIP_RS780:
633 case CHIP_RV620:
634 case CHIP_RS880:
635 return 16;
636 case CHIP_RV630:
637 case CHIP_RV635:
638 case CHIP_RV730:
639 case CHIP_RV710:
640 case CHIP_PALM:
641 case CHIP_CEDAR:
642 return 32;
643 default:
644 return 64;
645 }
646 }
647
648 static inline enum radeon_bo_priority
649 r600_get_sampler_view_priority(struct r600_resource *res)
650 {
651 if (res->b.b.target == PIPE_BUFFER)
652 return RADEON_PRIO_SAMPLER_BUFFER;
653
654 if (res->b.b.nr_samples > 1)
655 return RADEON_PRIO_SAMPLER_TEXTURE_MSAA;
656
657 return RADEON_PRIO_SAMPLER_TEXTURE;
658 }
659
660 #define COMPUTE_DBG(rscreen, fmt, args...) \
661 do { \
662 if ((rscreen->b.debug_flags & DBG_COMPUTE)) fprintf(stderr, fmt, ##args); \
663 } while (0);
664
665 #define R600_ERR(fmt, args...) \
666 fprintf(stderr, "EE %s:%d %s - "fmt, __FILE__, __LINE__, __func__, ##args)
667
668 /* For MSAA sample positions. */
669 #define FILL_SREG(s0x, s0y, s1x, s1y, s2x, s2y, s3x, s3y) \
670 (((s0x) & 0xf) | (((s0y) & 0xf) << 4) | \
671 (((s1x) & 0xf) << 8) | (((s1y) & 0xf) << 12) | \
672 (((s2x) & 0xf) << 16) | (((s2y) & 0xf) << 20) | \
673 (((s3x) & 0xf) << 24) | (((s3y) & 0xf) << 28))
674
675 #endif