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