gallium/radeon: degrade tiled textures mapped often to linear
[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 /* Pipeline & streamout query controls. */
54 #define R600_CONTEXT_START_PIPELINE_STATS (1u << 1)
55 #define R600_CONTEXT_STOP_PIPELINE_STATS (1u << 2)
56 #define R600_CONTEXT_PRIVATE_FLAG (1u << 3)
57
58 /* special primitive types */
59 #define R600_PRIM_RECTANGLE_LIST PIPE_PRIM_MAX
60
61 /* Debug flags. */
62 /* logging */
63 #define DBG_TEX (1 << 0)
64 /* gap - reuse */
65 #define DBG_COMPUTE (1 << 2)
66 #define DBG_VM (1 << 3)
67 /* gap - reuse */
68 /* shader logging */
69 #define DBG_FS (1 << 5)
70 #define DBG_VS (1 << 6)
71 #define DBG_GS (1 << 7)
72 #define DBG_PS (1 << 8)
73 #define DBG_CS (1 << 9)
74 #define DBG_TCS (1 << 10)
75 #define DBG_TES (1 << 11)
76 #define DBG_NO_IR (1 << 12)
77 #define DBG_NO_TGSI (1 << 13)
78 #define DBG_NO_ASM (1 << 14)
79 #define DBG_PREOPT_IR (1 << 15)
80 /* gaps */
81 #define DBG_TEST_DMA (1 << 20)
82 /* Bits 21-31 are reserved for the r600g driver. */
83 /* features */
84 #define DBG_NO_ASYNC_DMA (1llu << 32)
85 #define DBG_NO_HYPERZ (1llu << 33)
86 #define DBG_NO_DISCARD_RANGE (1llu << 34)
87 #define DBG_NO_2D_TILING (1llu << 35)
88 #define DBG_NO_TILING (1llu << 36)
89 #define DBG_SWITCH_ON_EOP (1llu << 37)
90 #define DBG_FORCE_DMA (1llu << 38)
91 #define DBG_PRECOMPILE (1llu << 39)
92 #define DBG_INFO (1llu << 40)
93 #define DBG_NO_WC (1llu << 41)
94 #define DBG_CHECK_VM (1llu << 42)
95 #define DBG_NO_DCC (1llu << 43)
96 #define DBG_NO_DCC_CLEAR (1llu << 44)
97 #define DBG_NO_RB_PLUS (1llu << 45)
98 #define DBG_SI_SCHED (1llu << 46)
99 #define DBG_MONOLITHIC_SHADERS (1llu << 47)
100 #define DBG_NO_CE (1llu << 48)
101
102 #define R600_MAP_BUFFER_ALIGNMENT 64
103 #define R600_MAX_VIEWPORTS 16
104
105 enum r600_coherency {
106 R600_COHERENCY_NONE, /* no cache flushes needed */
107 R600_COHERENCY_SHADER,
108 R600_COHERENCY_CB_META,
109 };
110
111 #ifdef PIPE_ARCH_BIG_ENDIAN
112 #define R600_BIG_ENDIAN 1
113 #else
114 #define R600_BIG_ENDIAN 0
115 #endif
116
117 struct r600_common_context;
118 struct r600_perfcounters;
119 struct tgsi_shader_info;
120
121 struct radeon_shader_reloc {
122 char name[32];
123 uint64_t offset;
124 };
125
126 struct radeon_shader_binary {
127 /** Shader code */
128 unsigned char *code;
129 unsigned code_size;
130
131 /** Config/Context register state that accompanies this shader.
132 * This is a stream of dword pairs. First dword contains the
133 * register address, the second dword contains the value.*/
134 unsigned char *config;
135 unsigned config_size;
136
137 /** The number of bytes of config information for each global symbol.
138 */
139 unsigned config_size_per_symbol;
140
141 /** Constant data accessed by the shader. This will be uploaded
142 * into a constant buffer. */
143 unsigned char *rodata;
144 unsigned rodata_size;
145
146 /** List of symbol offsets for the shader */
147 uint64_t *global_symbol_offsets;
148 unsigned global_symbol_count;
149
150 struct radeon_shader_reloc *relocs;
151 unsigned reloc_count;
152
153 /** Disassembled shader in a string. */
154 char *disasm_string;
155 };
156
157 void radeon_shader_binary_init(struct radeon_shader_binary *b);
158 void radeon_shader_binary_clean(struct radeon_shader_binary *b);
159
160 /* Only 32-bit buffer allocations are supported, gallium doesn't support more
161 * at the moment.
162 */
163 struct r600_resource {
164 struct u_resource b;
165
166 /* Winsys objects. */
167 struct pb_buffer *buf;
168 uint64_t gpu_address;
169
170 /* Resource state. */
171 enum radeon_bo_domain domains;
172
173 /* The buffer range which is initialized (with a write transfer,
174 * streamout, DMA, or as a random access target). The rest of
175 * the buffer is considered invalid and can be mapped unsynchronized.
176 *
177 * This allows unsychronized mapping of a buffer range which hasn't
178 * been used yet. It's for applications which forget to use
179 * the unsynchronized map flag and expect the driver to figure it out.
180 */
181 struct util_range valid_buffer_range;
182
183 /* For buffers only. This indicates that a write operation has been
184 * performed by TC L2, but the cache hasn't been flushed.
185 * Any hw block which doesn't use or bypasses TC L2 should check this
186 * flag and flush the cache before using the buffer.
187 *
188 * For example, TC L2 must be flushed if a buffer which has been
189 * modified by a shader store instruction is about to be used as
190 * an index buffer. The reason is that VGT DMA index fetching doesn't
191 * use TC L2.
192 */
193 bool TC_L2_dirty;
194
195 /* Whether the resource has been exported via resource_get_handle. */
196 bool is_shared;
197 unsigned external_usage; /* PIPE_HANDLE_USAGE_* */
198 };
199
200 struct r600_transfer {
201 struct pipe_transfer transfer;
202 struct r600_resource *staging;
203 unsigned offset;
204 };
205
206 struct r600_fmask_info {
207 uint64_t offset;
208 uint64_t size;
209 unsigned alignment;
210 unsigned pitch_in_pixels;
211 unsigned bank_height;
212 unsigned slice_tile_max;
213 unsigned tile_mode_index;
214 };
215
216 struct r600_cmask_info {
217 uint64_t offset;
218 uint64_t size;
219 unsigned alignment;
220 unsigned pitch;
221 unsigned height;
222 unsigned xalign;
223 unsigned yalign;
224 unsigned slice_tile_max;
225 unsigned base_address_reg;
226 };
227
228 struct r600_htile_info {
229 unsigned pitch;
230 unsigned height;
231 unsigned xalign;
232 unsigned yalign;
233 };
234
235 struct r600_texture {
236 struct r600_resource resource;
237
238 uint64_t size;
239 unsigned num_level0_transfers;
240 bool is_depth;
241 unsigned dirty_level_mask; /* each bit says if that mipmap is compressed */
242 unsigned stencil_dirty_level_mask; /* each bit says if that mipmap is compressed */
243 struct r600_texture *flushed_depth_texture;
244 boolean is_flushing_texture;
245 struct radeon_surf surface;
246
247 /* Colorbuffer compression and fast clear. */
248 struct r600_fmask_info fmask;
249 struct r600_cmask_info cmask;
250 struct r600_resource *cmask_buffer;
251 uint64_t dcc_offset; /* 0 = disabled */
252 unsigned cb_color_info; /* fast clear enable bit */
253 unsigned color_clear_value[2];
254
255 /* Depth buffer compression and fast clear. */
256 struct r600_htile_info htile;
257 struct r600_resource *htile_buffer;
258 bool depth_cleared; /* if it was cleared at least once */
259 float depth_clear_value;
260 bool stencil_cleared; /* if it was cleared at least once */
261 uint8_t stencil_clear_value;
262
263 bool non_disp_tiling; /* R600-Cayman only */
264
265 /* Counter that should be non-zero if the texture is bound to a
266 * framebuffer. Implemented in radeonsi only.
267 */
268 uint32_t framebuffers_bound;
269 };
270
271 struct r600_surface {
272 struct pipe_surface base;
273 const struct radeon_surf_level *level_info;
274
275 bool color_initialized;
276 bool depth_initialized;
277
278 /* Misc. color flags. */
279 bool alphatest_bypass;
280 bool export_16bpc;
281 bool color_is_int8;
282
283 /* Color registers. */
284 unsigned cb_color_info;
285 unsigned cb_color_base;
286 unsigned cb_color_view;
287 unsigned cb_color_size; /* R600 only */
288 unsigned cb_color_dim; /* EG only */
289 unsigned cb_color_pitch; /* EG and later */
290 unsigned cb_color_slice; /* EG and later */
291 unsigned cb_dcc_base; /* VI and later */
292 unsigned cb_color_attrib; /* EG and later */
293 unsigned cb_dcc_control; /* VI and later */
294 unsigned cb_color_fmask; /* CB_COLORn_FMASK (EG and later) or CB_COLORn_FRAG (r600) */
295 unsigned cb_color_fmask_slice; /* EG and later */
296 unsigned cb_color_cmask; /* CB_COLORn_TILE (r600 only) */
297 unsigned cb_color_mask; /* R600 only */
298 unsigned spi_shader_col_format; /* SI+, no blending, no alpha-to-coverage. */
299 unsigned spi_shader_col_format_alpha; /* SI+, alpha-to-coverage */
300 unsigned spi_shader_col_format_blend; /* SI+, blending without alpha. */
301 unsigned spi_shader_col_format_blend_alpha; /* SI+, blending with alpha. */
302 struct r600_resource *cb_buffer_fmask; /* Used for FMASK relocations. R600 only */
303 struct r600_resource *cb_buffer_cmask; /* Used for CMASK relocations. R600 only */
304
305 /* DB registers. */
306 unsigned db_depth_info; /* R600 only, then SI and later */
307 unsigned db_z_info; /* EG and later */
308 unsigned db_depth_base; /* DB_Z_READ/WRITE_BASE (EG and later) or DB_DEPTH_BASE (r600) */
309 unsigned db_depth_view;
310 unsigned db_depth_size;
311 unsigned db_depth_slice; /* EG and later */
312 unsigned db_stencil_base; /* EG and later */
313 unsigned db_stencil_info; /* EG and later */
314 unsigned db_prefetch_limit; /* R600 only */
315 unsigned db_htile_surface;
316 unsigned db_htile_data_base;
317 unsigned db_preload_control; /* EG and later */
318 unsigned pa_su_poly_offset_db_fmt_cntl;
319 };
320
321 struct r600_common_screen {
322 struct pipe_screen b;
323 struct radeon_winsys *ws;
324 enum radeon_family family;
325 enum chip_class chip_class;
326 struct radeon_info info;
327 uint64_t debug_flags;
328 bool has_cp_dma;
329 bool has_streamout;
330
331 /* Texture filter settings. */
332 int force_aniso; /* -1 = disabled */
333
334 /* Auxiliary context. Mainly used to initialize resources.
335 * It must be locked prior to using and flushed before unlocking. */
336 struct pipe_context *aux_context;
337 pipe_mutex aux_context_lock;
338
339 /* This must be in the screen, because UE4 uses one context for
340 * compilation and another one for rendering.
341 */
342 unsigned num_compilations;
343 /* Along with ST_DEBUG=precompile, this should show if applications
344 * are loading shaders on demand. This is a monotonic counter.
345 */
346 unsigned num_shaders_created;
347
348 /* GPU load thread. */
349 pipe_mutex gpu_load_mutex;
350 pipe_thread gpu_load_thread;
351 unsigned gpu_load_counter_busy;
352 unsigned gpu_load_counter_idle;
353 volatile unsigned gpu_load_stop_thread; /* bool */
354
355 char renderer_string[64];
356
357 /* Performance counters. */
358 struct r600_perfcounters *perfcounters;
359
360 /* If pipe_screen wants to re-emit the framebuffer state of all
361 * contexts, it should atomically increment this. Each context will
362 * compare this with its own last known value of the counter before
363 * drawing and re-emit the framebuffer state accordingly.
364 */
365 unsigned dirty_fb_counter;
366
367 /* Atomically increment this counter when an existing texture's
368 * metadata is enabled or disabled in a way that requires changing
369 * contexts' compressed texture binding masks.
370 */
371 unsigned compressed_colortex_counter;
372
373 /* Atomically increment this counter when an existing texture's
374 * backing buffer or tile mode parameters have changed that requires
375 * recomputation of shader descriptors.
376 */
377 unsigned dirty_tex_descriptor_counter;
378
379 void (*query_opaque_metadata)(struct r600_common_screen *rscreen,
380 struct r600_texture *rtex,
381 struct radeon_bo_metadata *md);
382 };
383
384 /* This encapsulates a state or an operation which can emitted into the GPU
385 * command stream. */
386 struct r600_atom {
387 void (*emit)(struct r600_common_context *ctx, struct r600_atom *state);
388 unsigned num_dw;
389 unsigned short id;
390 };
391
392 struct r600_so_target {
393 struct pipe_stream_output_target b;
394
395 /* The buffer where BUFFER_FILLED_SIZE is stored. */
396 struct r600_resource *buf_filled_size;
397 unsigned buf_filled_size_offset;
398 bool buf_filled_size_valid;
399
400 unsigned stride_in_dw;
401 };
402
403 struct r600_streamout {
404 struct r600_atom begin_atom;
405 bool begin_emitted;
406 unsigned num_dw_for_end;
407
408 unsigned enabled_mask;
409 unsigned num_targets;
410 struct r600_so_target *targets[PIPE_MAX_SO_BUFFERS];
411
412 unsigned append_bitmask;
413 bool suspended;
414
415 /* External state which comes from the vertex shader,
416 * it must be set explicitly when binding a shader. */
417 unsigned *stride_in_dw;
418 unsigned enabled_stream_buffers_mask; /* stream0 buffers0-3 in 4 LSB */
419
420 /* The state of VGT_STRMOUT_BUFFER_(CONFIG|EN). */
421 unsigned hw_enabled_mask;
422
423 /* The state of VGT_STRMOUT_(CONFIG|EN). */
424 struct r600_atom enable_atom;
425 bool streamout_enabled;
426 bool prims_gen_query_enabled;
427 int num_prims_gen_queries;
428 };
429
430 struct r600_signed_scissor {
431 int minx;
432 int miny;
433 int maxx;
434 int maxy;
435 };
436
437 struct r600_scissors {
438 struct r600_atom atom;
439 unsigned dirty_mask;
440 struct pipe_scissor_state states[R600_MAX_VIEWPORTS];
441 };
442
443 struct r600_viewports {
444 struct r600_atom atom;
445 unsigned dirty_mask;
446 struct pipe_viewport_state states[R600_MAX_VIEWPORTS];
447 struct r600_signed_scissor as_scissor[R600_MAX_VIEWPORTS];
448 };
449
450 struct r600_ring {
451 struct radeon_winsys_cs *cs;
452 void (*flush)(void *ctx, unsigned flags,
453 struct pipe_fence_handle **fence);
454 };
455
456 struct r600_common_context {
457 struct pipe_context b; /* base class */
458
459 struct r600_common_screen *screen;
460 struct radeon_winsys *ws;
461 struct radeon_winsys_ctx *ctx;
462 enum radeon_family family;
463 enum chip_class chip_class;
464 struct r600_ring gfx;
465 struct r600_ring dma;
466 struct pipe_fence_handle *last_sdma_fence;
467 unsigned initial_gfx_cs_size;
468 unsigned gpu_reset_counter;
469 unsigned last_dirty_fb_counter;
470 unsigned last_compressed_colortex_counter;
471 unsigned last_dirty_tex_descriptor_counter;
472
473 struct u_upload_mgr *uploader;
474 struct u_suballocator *allocator_so_filled_size;
475 struct util_slab_mempool pool_transfers;
476
477 /* Current unaccounted memory usage. */
478 uint64_t vram;
479 uint64_t gtt;
480
481 /* States. */
482 struct r600_streamout streamout;
483 struct r600_scissors scissors;
484 struct r600_viewports viewports;
485 bool scissor_enabled;
486 bool vs_writes_viewport_index;
487 bool vs_disables_clipping_viewport;
488
489 /* Additional context states. */
490 unsigned flags; /* flush flags */
491
492 /* Queries. */
493 /* Maintain the list of active queries for pausing between IBs. */
494 int num_occlusion_queries;
495 int num_perfect_occlusion_queries;
496 struct list_head active_queries;
497 unsigned num_cs_dw_queries_suspend;
498 /* Additional hardware info. */
499 unsigned backend_mask;
500 unsigned max_db; /* for OQ */
501 /* Misc stats. */
502 unsigned num_draw_calls;
503 unsigned num_dma_calls;
504
505 /* Render condition. */
506 struct r600_atom render_cond_atom;
507 struct pipe_query *render_cond;
508 unsigned render_cond_mode;
509 boolean render_cond_invert;
510 bool render_cond_force_off; /* for u_blitter */
511
512 /* MSAA sample locations.
513 * The first index is the sample index.
514 * The second index is the coordinate: X, Y. */
515 float sample_locations_1x[1][2];
516 float sample_locations_2x[2][2];
517 float sample_locations_4x[4][2];
518 float sample_locations_8x[8][2];
519 float sample_locations_16x[16][2];
520
521 /* The list of all texture buffer objects in this context.
522 * This list is walked when a buffer is invalidated/reallocated and
523 * the GPU addresses are updated. */
524 struct list_head texture_buffers;
525
526 struct pipe_debug_callback debug;
527
528 /* Copy one resource to another using async DMA. */
529 void (*dma_copy)(struct pipe_context *ctx,
530 struct pipe_resource *dst,
531 unsigned dst_level,
532 unsigned dst_x, unsigned dst_y, unsigned dst_z,
533 struct pipe_resource *src,
534 unsigned src_level,
535 const struct pipe_box *src_box);
536
537 void (*clear_buffer)(struct pipe_context *ctx, struct pipe_resource *dst,
538 uint64_t offset, uint64_t size, unsigned value,
539 enum r600_coherency coher);
540
541 void (*blit_decompress_depth)(struct pipe_context *ctx,
542 struct r600_texture *texture,
543 struct r600_texture *staging,
544 unsigned first_level, unsigned last_level,
545 unsigned first_layer, unsigned last_layer,
546 unsigned first_sample, unsigned last_sample);
547
548 void (*decompress_dcc)(struct pipe_context *ctx,
549 struct r600_texture *rtex);
550
551 /* Reallocate the buffer and update all resource bindings where
552 * the buffer is bound, including all resource descriptors. */
553 void (*invalidate_buffer)(struct pipe_context *ctx, struct pipe_resource *buf);
554
555 /* Enable or disable occlusion queries. */
556 void (*set_occlusion_query_state)(struct pipe_context *ctx, bool enable);
557
558 /* This ensures there is enough space in the command stream. */
559 void (*need_gfx_cs_space)(struct pipe_context *ctx, unsigned num_dw,
560 bool include_draw_vbo);
561
562 void (*set_atom_dirty)(struct r600_common_context *ctx,
563 struct r600_atom *atom, bool dirty);
564 };
565
566 /* r600_buffer.c */
567 boolean r600_rings_is_buffer_referenced(struct r600_common_context *ctx,
568 struct pb_buffer *buf,
569 enum radeon_bo_usage usage);
570 void *r600_buffer_map_sync_with_rings(struct r600_common_context *ctx,
571 struct r600_resource *resource,
572 unsigned usage);
573 bool r600_init_resource(struct r600_common_screen *rscreen,
574 struct r600_resource *res,
575 uint64_t size, unsigned alignment);
576 struct pipe_resource *r600_buffer_create(struct pipe_screen *screen,
577 const struct pipe_resource *templ,
578 unsigned alignment);
579 struct pipe_resource * r600_aligned_buffer_create(struct pipe_screen *screen,
580 unsigned bind,
581 unsigned usage,
582 unsigned size,
583 unsigned alignment);
584 struct pipe_resource *
585 r600_buffer_from_user_memory(struct pipe_screen *screen,
586 const struct pipe_resource *templ,
587 void *user_memory);
588 void
589 r600_invalidate_resource(struct pipe_context *ctx,
590 struct pipe_resource *resource);
591
592 /* r600_common_pipe.c */
593 void r600_draw_rectangle(struct blitter_context *blitter,
594 int x1, int y1, int x2, int y2, float depth,
595 enum blitter_attrib_type type,
596 const union pipe_color_union *attrib);
597 bool r600_common_screen_init(struct r600_common_screen *rscreen,
598 struct radeon_winsys *ws);
599 void r600_destroy_common_screen(struct r600_common_screen *rscreen);
600 void r600_preflush_suspend_features(struct r600_common_context *ctx);
601 void r600_postflush_resume_features(struct r600_common_context *ctx);
602 bool r600_common_context_init(struct r600_common_context *rctx,
603 struct r600_common_screen *rscreen);
604 void r600_common_context_cleanup(struct r600_common_context *rctx);
605 void r600_context_add_resource_size(struct pipe_context *ctx, struct pipe_resource *r);
606 bool r600_can_dump_shader(struct r600_common_screen *rscreen,
607 unsigned processor);
608 void r600_screen_clear_buffer(struct r600_common_screen *rscreen, struct pipe_resource *dst,
609 uint64_t offset, uint64_t size, unsigned value,
610 enum r600_coherency coher);
611 struct pipe_resource *r600_resource_create_common(struct pipe_screen *screen,
612 const struct pipe_resource *templ);
613 const char *r600_get_llvm_processor_name(enum radeon_family family);
614 void r600_need_dma_space(struct r600_common_context *ctx, unsigned num_dw,
615 struct r600_resource *dst, struct r600_resource *src);
616 void r600_dma_emit_wait_idle(struct r600_common_context *rctx);
617
618 /* r600_gpu_load.c */
619 void r600_gpu_load_kill_thread(struct r600_common_screen *rscreen);
620 uint64_t r600_gpu_load_begin(struct r600_common_screen *rscreen);
621 unsigned r600_gpu_load_end(struct r600_common_screen *rscreen, uint64_t begin);
622
623 /* r600_perfcounters.c */
624 void r600_perfcounters_destroy(struct r600_common_screen *rscreen);
625
626 /* r600_query.c */
627 void r600_init_screen_query_functions(struct r600_common_screen *rscreen);
628 void r600_query_init(struct r600_common_context *rctx);
629 void r600_suspend_queries(struct r600_common_context *ctx);
630 void r600_resume_queries(struct r600_common_context *ctx);
631 void r600_query_init_backend_mask(struct r600_common_context *ctx);
632
633 /* r600_streamout.c */
634 void r600_streamout_buffers_dirty(struct r600_common_context *rctx);
635 void r600_set_streamout_targets(struct pipe_context *ctx,
636 unsigned num_targets,
637 struct pipe_stream_output_target **targets,
638 const unsigned *offset);
639 void r600_emit_streamout_end(struct r600_common_context *rctx);
640 void r600_update_prims_generated_query_state(struct r600_common_context *rctx,
641 unsigned type, int diff);
642 void r600_streamout_init(struct r600_common_context *rctx);
643
644 /* r600_test_dma.c */
645 void r600_test_dma(struct r600_common_screen *rscreen);
646
647 /* r600_texture.c */
648 bool r600_prepare_for_dma_blit(struct r600_common_context *rctx,
649 struct r600_texture *rdst,
650 unsigned dst_level, unsigned dstx,
651 unsigned dsty, unsigned dstz,
652 struct r600_texture *rsrc,
653 unsigned src_level,
654 const struct pipe_box *src_box);
655 void r600_texture_get_fmask_info(struct r600_common_screen *rscreen,
656 struct r600_texture *rtex,
657 unsigned nr_samples,
658 struct r600_fmask_info *out);
659 void r600_texture_get_cmask_info(struct r600_common_screen *rscreen,
660 struct r600_texture *rtex,
661 struct r600_cmask_info *out);
662 bool r600_init_flushed_depth_texture(struct pipe_context *ctx,
663 struct pipe_resource *texture,
664 struct r600_texture **staging);
665 void r600_print_texture_info(struct r600_texture *rtex, FILE *f);
666 struct pipe_resource *r600_texture_create(struct pipe_screen *screen,
667 const struct pipe_resource *templ);
668 struct pipe_surface *r600_create_surface_custom(struct pipe_context *pipe,
669 struct pipe_resource *texture,
670 const struct pipe_surface *templ,
671 unsigned width, unsigned height);
672 unsigned r600_translate_colorswap(enum pipe_format format, bool do_endian_swap);
673 void evergreen_do_fast_color_clear(struct r600_common_context *rctx,
674 struct pipe_framebuffer_state *fb,
675 struct r600_atom *fb_state,
676 unsigned *buffers, unsigned *dirty_cbufs,
677 const union pipe_color_union *color);
678 void r600_texture_disable_dcc(struct r600_common_screen *rscreen,
679 struct r600_texture *rtex);
680 void r600_init_screen_texture_functions(struct r600_common_screen *rscreen);
681 void r600_init_context_texture_functions(struct r600_common_context *rctx);
682
683 /* r600_viewport.c */
684 void evergreen_apply_scissor_bug_workaround(struct r600_common_context *rctx,
685 struct pipe_scissor_state *scissor);
686 void r600_set_scissor_enable(struct r600_common_context *rctx, bool enable);
687 void r600_update_vs_writes_viewport_index(struct r600_common_context *rctx,
688 struct tgsi_shader_info *info);
689 void r600_init_viewport_functions(struct r600_common_context *rctx);
690
691 /* cayman_msaa.c */
692 extern const uint32_t eg_sample_locs_2x[4];
693 extern const unsigned eg_max_dist_2x;
694 extern const uint32_t eg_sample_locs_4x[4];
695 extern const unsigned eg_max_dist_4x;
696 void cayman_get_sample_position(struct pipe_context *ctx, unsigned sample_count,
697 unsigned sample_index, float *out_value);
698 void cayman_init_msaa(struct pipe_context *ctx);
699 void cayman_emit_msaa_sample_locs(struct radeon_winsys_cs *cs, int nr_samples);
700 void cayman_emit_msaa_config(struct radeon_winsys_cs *cs, int nr_samples,
701 int ps_iter_samples, int overrast_samples);
702
703
704 /* Inline helpers. */
705
706 static inline struct r600_resource *r600_resource(struct pipe_resource *r)
707 {
708 return (struct r600_resource*)r;
709 }
710
711 static inline void
712 r600_resource_reference(struct r600_resource **ptr, struct r600_resource *res)
713 {
714 pipe_resource_reference((struct pipe_resource **)ptr,
715 (struct pipe_resource *)res);
716 }
717
718 static inline bool r600_get_strmout_en(struct r600_common_context *rctx)
719 {
720 return rctx->streamout.streamout_enabled ||
721 rctx->streamout.prims_gen_query_enabled;
722 }
723
724 #define SQ_TEX_XY_FILTER_POINT 0x00
725 #define SQ_TEX_XY_FILTER_BILINEAR 0x01
726 #define SQ_TEX_XY_FILTER_ANISO_POINT 0x02
727 #define SQ_TEX_XY_FILTER_ANISO_BILINEAR 0x03
728
729 static inline unsigned eg_tex_filter(unsigned filter, unsigned max_aniso)
730 {
731 if (filter == PIPE_TEX_FILTER_LINEAR)
732 return max_aniso > 1 ? SQ_TEX_XY_FILTER_ANISO_BILINEAR
733 : SQ_TEX_XY_FILTER_BILINEAR;
734 else
735 return max_aniso > 1 ? SQ_TEX_XY_FILTER_ANISO_POINT
736 : SQ_TEX_XY_FILTER_POINT;
737 }
738
739 static inline unsigned r600_tex_aniso_filter(unsigned filter)
740 {
741 if (filter < 2)
742 return 0;
743 if (filter < 4)
744 return 1;
745 if (filter < 8)
746 return 2;
747 if (filter < 16)
748 return 3;
749 return 4;
750 }
751
752 static inline unsigned r600_wavefront_size(enum radeon_family family)
753 {
754 switch (family) {
755 case CHIP_RV610:
756 case CHIP_RS780:
757 case CHIP_RV620:
758 case CHIP_RS880:
759 return 16;
760 case CHIP_RV630:
761 case CHIP_RV635:
762 case CHIP_RV730:
763 case CHIP_RV710:
764 case CHIP_PALM:
765 case CHIP_CEDAR:
766 return 32;
767 default:
768 return 64;
769 }
770 }
771
772 static inline enum radeon_bo_priority
773 r600_get_sampler_view_priority(struct r600_resource *res)
774 {
775 if (res->b.b.target == PIPE_BUFFER)
776 return RADEON_PRIO_SAMPLER_BUFFER;
777
778 if (res->b.b.nr_samples > 1)
779 return RADEON_PRIO_SAMPLER_TEXTURE_MSAA;
780
781 return RADEON_PRIO_SAMPLER_TEXTURE;
782 }
783
784 #define COMPUTE_DBG(rscreen, fmt, args...) \
785 do { \
786 if ((rscreen->b.debug_flags & DBG_COMPUTE)) fprintf(stderr, fmt, ##args); \
787 } while (0);
788
789 #define R600_ERR(fmt, args...) \
790 fprintf(stderr, "EE %s:%d %s - " fmt, __FILE__, __LINE__, __func__, ##args)
791
792 /* For MSAA sample positions. */
793 #define FILL_SREG(s0x, s0y, s1x, s1y, s2x, s2y, s3x, s3y) \
794 (((s0x) & 0xf) | (((unsigned)(s0y) & 0xf) << 4) | \
795 (((unsigned)(s1x) & 0xf) << 8) | (((unsigned)(s1y) & 0xf) << 12) | \
796 (((unsigned)(s2x) & 0xf) << 16) | (((unsigned)(s2y) & 0xf) << 20) | \
797 (((unsigned)(s3x) & 0xf) << 24) | (((unsigned)(s3y) & 0xf) << 28))
798
799 #endif