radeonsi: don't allocate DCC for the temporary MSAA resolve surface
[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 #define R600_RESOURCE_FLAG_DISABLE_DCC (PIPE_RESOURCE_FLAG_DRV_PRIV << 3)
52
53 #define R600_CONTEXT_STREAMOUT_FLUSH (1u << 0)
54 /* Pipeline & streamout query controls. */
55 #define R600_CONTEXT_START_PIPELINE_STATS (1u << 1)
56 #define R600_CONTEXT_STOP_PIPELINE_STATS (1u << 2)
57 #define R600_CONTEXT_PRIVATE_FLAG (1u << 3)
58
59 /* special primitive types */
60 #define R600_PRIM_RECTANGLE_LIST PIPE_PRIM_MAX
61
62 /* Debug flags. */
63 /* logging */
64 #define DBG_TEX (1 << 0)
65 /* gap - reuse */
66 #define DBG_COMPUTE (1 << 2)
67 #define DBG_VM (1 << 3)
68 /* gap - reuse */
69 /* shader logging */
70 #define DBG_FS (1 << 5)
71 #define DBG_VS (1 << 6)
72 #define DBG_GS (1 << 7)
73 #define DBG_PS (1 << 8)
74 #define DBG_CS (1 << 9)
75 #define DBG_TCS (1 << 10)
76 #define DBG_TES (1 << 11)
77 #define DBG_NO_IR (1 << 12)
78 #define DBG_NO_TGSI (1 << 13)
79 #define DBG_NO_ASM (1 << 14)
80 #define DBG_PREOPT_IR (1 << 15)
81 /* gaps */
82 #define DBG_TEST_DMA (1 << 20)
83 /* Bits 21-31 are reserved for the r600g driver. */
84 /* features */
85 #define DBG_NO_ASYNC_DMA (1llu << 32)
86 #define DBG_NO_HYPERZ (1llu << 33)
87 #define DBG_NO_DISCARD_RANGE (1llu << 34)
88 #define DBG_NO_2D_TILING (1llu << 35)
89 #define DBG_NO_TILING (1llu << 36)
90 #define DBG_SWITCH_ON_EOP (1llu << 37)
91 #define DBG_FORCE_DMA (1llu << 38)
92 #define DBG_PRECOMPILE (1llu << 39)
93 #define DBG_INFO (1llu << 40)
94 #define DBG_NO_WC (1llu << 41)
95 #define DBG_CHECK_VM (1llu << 42)
96 #define DBG_NO_DCC (1llu << 43)
97 #define DBG_NO_DCC_CLEAR (1llu << 44)
98 #define DBG_NO_RB_PLUS (1llu << 45)
99 #define DBG_SI_SCHED (1llu << 46)
100 #define DBG_MONOLITHIC_SHADERS (1llu << 47)
101 #define DBG_NO_CE (1llu << 48)
102
103 #define R600_MAP_BUFFER_ALIGNMENT 64
104 #define R600_MAX_VIEWPORTS 16
105
106 enum r600_coherency {
107 R600_COHERENCY_NONE, /* no cache flushes needed */
108 R600_COHERENCY_SHADER,
109 R600_COHERENCY_CB_META,
110 };
111
112 #ifdef PIPE_ARCH_BIG_ENDIAN
113 #define R600_BIG_ENDIAN 1
114 #else
115 #define R600_BIG_ENDIAN 0
116 #endif
117
118 struct r600_common_context;
119 struct r600_perfcounters;
120 struct tgsi_shader_info;
121
122 struct radeon_shader_reloc {
123 char name[32];
124 uint64_t offset;
125 };
126
127 struct radeon_shader_binary {
128 /** Shader code */
129 unsigned char *code;
130 unsigned code_size;
131
132 /** Config/Context register state that accompanies this shader.
133 * This is a stream of dword pairs. First dword contains the
134 * register address, the second dword contains the value.*/
135 unsigned char *config;
136 unsigned config_size;
137
138 /** The number of bytes of config information for each global symbol.
139 */
140 unsigned config_size_per_symbol;
141
142 /** Constant data accessed by the shader. This will be uploaded
143 * into a constant buffer. */
144 unsigned char *rodata;
145 unsigned rodata_size;
146
147 /** List of symbol offsets for the shader */
148 uint64_t *global_symbol_offsets;
149 unsigned global_symbol_count;
150
151 struct radeon_shader_reloc *relocs;
152 unsigned reloc_count;
153
154 /** Disassembled shader in a string. */
155 char *disasm_string;
156 };
157
158 void radeon_shader_binary_init(struct radeon_shader_binary *b);
159 void radeon_shader_binary_clean(struct radeon_shader_binary *b);
160
161 /* Only 32-bit buffer allocations are supported, gallium doesn't support more
162 * at the moment.
163 */
164 struct r600_resource {
165 struct u_resource b;
166
167 /* Winsys objects. */
168 struct pb_buffer *buf;
169 uint64_t gpu_address;
170
171 /* Resource state. */
172 enum radeon_bo_domain domains;
173
174 /* The buffer range which is initialized (with a write transfer,
175 * streamout, DMA, or as a random access target). The rest of
176 * the buffer is considered invalid and can be mapped unsynchronized.
177 *
178 * This allows unsychronized mapping of a buffer range which hasn't
179 * been used yet. It's for applications which forget to use
180 * the unsynchronized map flag and expect the driver to figure it out.
181 */
182 struct util_range valid_buffer_range;
183
184 /* For buffers only. This indicates that a write operation has been
185 * performed by TC L2, but the cache hasn't been flushed.
186 * Any hw block which doesn't use or bypasses TC L2 should check this
187 * flag and flush the cache before using the buffer.
188 *
189 * For example, TC L2 must be flushed if a buffer which has been
190 * modified by a shader store instruction is about to be used as
191 * an index buffer. The reason is that VGT DMA index fetching doesn't
192 * use TC L2.
193 */
194 bool TC_L2_dirty;
195
196 /* Whether the resource has been exported via resource_get_handle. */
197 bool is_shared;
198 unsigned external_usage; /* PIPE_HANDLE_USAGE_* */
199 };
200
201 struct r600_transfer {
202 struct pipe_transfer transfer;
203 struct r600_resource *staging;
204 unsigned offset;
205 };
206
207 struct r600_fmask_info {
208 uint64_t offset;
209 uint64_t size;
210 unsigned alignment;
211 unsigned pitch_in_pixels;
212 unsigned bank_height;
213 unsigned slice_tile_max;
214 unsigned tile_mode_index;
215 };
216
217 struct r600_cmask_info {
218 uint64_t offset;
219 uint64_t size;
220 unsigned alignment;
221 unsigned pitch;
222 unsigned height;
223 unsigned xalign;
224 unsigned yalign;
225 unsigned slice_tile_max;
226 unsigned base_address_reg;
227 };
228
229 struct r600_htile_info {
230 unsigned pitch;
231 unsigned height;
232 unsigned xalign;
233 unsigned yalign;
234 };
235
236 struct r600_texture {
237 struct r600_resource resource;
238
239 uint64_t size;
240 unsigned num_level0_transfers;
241 bool is_depth;
242 unsigned dirty_level_mask; /* each bit says if that mipmap is compressed */
243 unsigned stencil_dirty_level_mask; /* each bit says if that mipmap is compressed */
244 struct r600_texture *flushed_depth_texture;
245 boolean is_flushing_texture;
246 struct radeon_surf surface;
247
248 /* Colorbuffer compression and fast clear. */
249 struct r600_fmask_info fmask;
250 struct r600_cmask_info cmask;
251 struct r600_resource *cmask_buffer;
252 uint64_t dcc_offset; /* 0 = disabled */
253 unsigned cb_color_info; /* fast clear enable bit */
254 unsigned color_clear_value[2];
255
256 /* Depth buffer compression and fast clear. */
257 struct r600_htile_info htile;
258 struct r600_resource *htile_buffer;
259 bool depth_cleared; /* if it was cleared at least once */
260 float depth_clear_value;
261 bool stencil_cleared; /* if it was cleared at least once */
262 uint8_t stencil_clear_value;
263
264 bool non_disp_tiling; /* R600-Cayman only */
265
266 /* Counter that should be non-zero if the texture is bound to a
267 * framebuffer. Implemented in radeonsi only.
268 */
269 uint32_t framebuffers_bound;
270 };
271
272 struct r600_surface {
273 struct pipe_surface base;
274 const struct radeon_surf_level *level_info;
275
276 bool color_initialized;
277 bool depth_initialized;
278
279 /* Misc. color flags. */
280 bool alphatest_bypass;
281 bool export_16bpc;
282 bool color_is_int8;
283
284 /* Color registers. */
285 unsigned cb_color_info;
286 unsigned cb_color_base;
287 unsigned cb_color_view;
288 unsigned cb_color_size; /* R600 only */
289 unsigned cb_color_dim; /* EG only */
290 unsigned cb_color_pitch; /* EG and later */
291 unsigned cb_color_slice; /* EG 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 void (*apply_opaque_metadata)(struct r600_common_screen *rscreen,
384 struct r600_texture *rtex,
385 struct radeon_bo_metadata *md);
386 };
387
388 /* This encapsulates a state or an operation which can emitted into the GPU
389 * command stream. */
390 struct r600_atom {
391 void (*emit)(struct r600_common_context *ctx, struct r600_atom *state);
392 unsigned num_dw;
393 unsigned short id;
394 };
395
396 struct r600_so_target {
397 struct pipe_stream_output_target b;
398
399 /* The buffer where BUFFER_FILLED_SIZE is stored. */
400 struct r600_resource *buf_filled_size;
401 unsigned buf_filled_size_offset;
402 bool buf_filled_size_valid;
403
404 unsigned stride_in_dw;
405 };
406
407 struct r600_streamout {
408 struct r600_atom begin_atom;
409 bool begin_emitted;
410 unsigned num_dw_for_end;
411
412 unsigned enabled_mask;
413 unsigned num_targets;
414 struct r600_so_target *targets[PIPE_MAX_SO_BUFFERS];
415
416 unsigned append_bitmask;
417 bool suspended;
418
419 /* External state which comes from the vertex shader,
420 * it must be set explicitly when binding a shader. */
421 unsigned *stride_in_dw;
422 unsigned enabled_stream_buffers_mask; /* stream0 buffers0-3 in 4 LSB */
423
424 /* The state of VGT_STRMOUT_BUFFER_(CONFIG|EN). */
425 unsigned hw_enabled_mask;
426
427 /* The state of VGT_STRMOUT_(CONFIG|EN). */
428 struct r600_atom enable_atom;
429 bool streamout_enabled;
430 bool prims_gen_query_enabled;
431 int num_prims_gen_queries;
432 };
433
434 struct r600_signed_scissor {
435 int minx;
436 int miny;
437 int maxx;
438 int maxy;
439 };
440
441 struct r600_scissors {
442 struct r600_atom atom;
443 unsigned dirty_mask;
444 struct pipe_scissor_state states[R600_MAX_VIEWPORTS];
445 };
446
447 struct r600_viewports {
448 struct r600_atom atom;
449 unsigned dirty_mask;
450 struct pipe_viewport_state states[R600_MAX_VIEWPORTS];
451 struct r600_signed_scissor as_scissor[R600_MAX_VIEWPORTS];
452 };
453
454 struct r600_ring {
455 struct radeon_winsys_cs *cs;
456 void (*flush)(void *ctx, unsigned flags,
457 struct pipe_fence_handle **fence);
458 };
459
460 struct r600_common_context {
461 struct pipe_context b; /* base class */
462
463 struct r600_common_screen *screen;
464 struct radeon_winsys *ws;
465 struct radeon_winsys_ctx *ctx;
466 enum radeon_family family;
467 enum chip_class chip_class;
468 struct r600_ring gfx;
469 struct r600_ring dma;
470 struct pipe_fence_handle *last_sdma_fence;
471 unsigned initial_gfx_cs_size;
472 unsigned gpu_reset_counter;
473 unsigned last_dirty_fb_counter;
474 unsigned last_compressed_colortex_counter;
475 unsigned last_dirty_tex_descriptor_counter;
476
477 struct u_upload_mgr *uploader;
478 struct u_suballocator *allocator_zeroed_memory;
479 struct util_slab_mempool pool_transfers;
480
481 /* Current unaccounted memory usage. */
482 uint64_t vram;
483 uint64_t gtt;
484
485 /* States. */
486 struct r600_streamout streamout;
487 struct r600_scissors scissors;
488 struct r600_viewports viewports;
489 bool scissor_enabled;
490 bool vs_writes_viewport_index;
491 bool vs_disables_clipping_viewport;
492
493 /* Additional context states. */
494 unsigned flags; /* flush flags */
495
496 /* Queries. */
497 /* Maintain the list of active queries for pausing between IBs. */
498 int num_occlusion_queries;
499 int num_perfect_occlusion_queries;
500 struct list_head active_queries;
501 unsigned num_cs_dw_queries_suspend;
502 /* Additional hardware info. */
503 unsigned backend_mask;
504 unsigned max_db; /* for OQ */
505 /* Misc stats. */
506 unsigned num_draw_calls;
507 unsigned num_dma_calls;
508 uint64_t num_alloc_tex_transfer_bytes;
509
510 /* Render condition. */
511 struct r600_atom render_cond_atom;
512 struct pipe_query *render_cond;
513 unsigned render_cond_mode;
514 boolean render_cond_invert;
515 bool render_cond_force_off; /* for u_blitter */
516
517 /* MSAA sample locations.
518 * The first index is the sample index.
519 * The second index is the coordinate: X, Y. */
520 float sample_locations_1x[1][2];
521 float sample_locations_2x[2][2];
522 float sample_locations_4x[4][2];
523 float sample_locations_8x[8][2];
524 float sample_locations_16x[16][2];
525
526 /* The list of all texture buffer objects in this context.
527 * This list is walked when a buffer is invalidated/reallocated and
528 * the GPU addresses are updated. */
529 struct list_head texture_buffers;
530
531 struct pipe_debug_callback debug;
532
533 /* Copy one resource to another using async DMA. */
534 void (*dma_copy)(struct pipe_context *ctx,
535 struct pipe_resource *dst,
536 unsigned dst_level,
537 unsigned dst_x, unsigned dst_y, unsigned dst_z,
538 struct pipe_resource *src,
539 unsigned src_level,
540 const struct pipe_box *src_box);
541
542 void (*clear_buffer)(struct pipe_context *ctx, struct pipe_resource *dst,
543 uint64_t offset, uint64_t size, unsigned value,
544 enum r600_coherency coher);
545
546 void (*blit_decompress_depth)(struct pipe_context *ctx,
547 struct r600_texture *texture,
548 struct r600_texture *staging,
549 unsigned first_level, unsigned last_level,
550 unsigned first_layer, unsigned last_layer,
551 unsigned first_sample, unsigned last_sample);
552
553 void (*decompress_dcc)(struct pipe_context *ctx,
554 struct r600_texture *rtex);
555
556 /* Reallocate the buffer and update all resource bindings where
557 * the buffer is bound, including all resource descriptors. */
558 void (*invalidate_buffer)(struct pipe_context *ctx, struct pipe_resource *buf);
559
560 /* Enable or disable occlusion queries. */
561 void (*set_occlusion_query_state)(struct pipe_context *ctx, bool enable);
562
563 /* This ensures there is enough space in the command stream. */
564 void (*need_gfx_cs_space)(struct pipe_context *ctx, unsigned num_dw,
565 bool include_draw_vbo);
566
567 void (*set_atom_dirty)(struct r600_common_context *ctx,
568 struct r600_atom *atom, bool dirty);
569 };
570
571 /* r600_buffer.c */
572 boolean r600_rings_is_buffer_referenced(struct r600_common_context *ctx,
573 struct pb_buffer *buf,
574 enum radeon_bo_usage usage);
575 void *r600_buffer_map_sync_with_rings(struct r600_common_context *ctx,
576 struct r600_resource *resource,
577 unsigned usage);
578 bool r600_init_resource(struct r600_common_screen *rscreen,
579 struct r600_resource *res,
580 uint64_t size, unsigned alignment);
581 struct pipe_resource *r600_buffer_create(struct pipe_screen *screen,
582 const struct pipe_resource *templ,
583 unsigned alignment);
584 struct pipe_resource * r600_aligned_buffer_create(struct pipe_screen *screen,
585 unsigned bind,
586 unsigned usage,
587 unsigned size,
588 unsigned alignment);
589 struct pipe_resource *
590 r600_buffer_from_user_memory(struct pipe_screen *screen,
591 const struct pipe_resource *templ,
592 void *user_memory);
593 void
594 r600_invalidate_resource(struct pipe_context *ctx,
595 struct pipe_resource *resource);
596
597 /* r600_common_pipe.c */
598 void r600_draw_rectangle(struct blitter_context *blitter,
599 int x1, int y1, int x2, int y2, float depth,
600 enum blitter_attrib_type type,
601 const union pipe_color_union *attrib);
602 bool r600_common_screen_init(struct r600_common_screen *rscreen,
603 struct radeon_winsys *ws);
604 void r600_destroy_common_screen(struct r600_common_screen *rscreen);
605 void r600_preflush_suspend_features(struct r600_common_context *ctx);
606 void r600_postflush_resume_features(struct r600_common_context *ctx);
607 bool r600_common_context_init(struct r600_common_context *rctx,
608 struct r600_common_screen *rscreen);
609 void r600_common_context_cleanup(struct r600_common_context *rctx);
610 void r600_context_add_resource_size(struct pipe_context *ctx, struct pipe_resource *r);
611 bool r600_can_dump_shader(struct r600_common_screen *rscreen,
612 unsigned processor);
613 void r600_screen_clear_buffer(struct r600_common_screen *rscreen, struct pipe_resource *dst,
614 uint64_t offset, uint64_t size, unsigned value,
615 enum r600_coherency coher);
616 struct pipe_resource *r600_resource_create_common(struct pipe_screen *screen,
617 const struct pipe_resource *templ);
618 const char *r600_get_llvm_processor_name(enum radeon_family family);
619 void r600_need_dma_space(struct r600_common_context *ctx, unsigned num_dw,
620 struct r600_resource *dst, struct r600_resource *src);
621 void r600_dma_emit_wait_idle(struct r600_common_context *rctx);
622
623 /* r600_gpu_load.c */
624 void r600_gpu_load_kill_thread(struct r600_common_screen *rscreen);
625 uint64_t r600_gpu_load_begin(struct r600_common_screen *rscreen);
626 unsigned r600_gpu_load_end(struct r600_common_screen *rscreen, uint64_t begin);
627
628 /* r600_perfcounters.c */
629 void r600_perfcounters_destroy(struct r600_common_screen *rscreen);
630
631 /* r600_query.c */
632 void r600_init_screen_query_functions(struct r600_common_screen *rscreen);
633 void r600_query_init(struct r600_common_context *rctx);
634 void r600_suspend_queries(struct r600_common_context *ctx);
635 void r600_resume_queries(struct r600_common_context *ctx);
636 void r600_query_init_backend_mask(struct r600_common_context *ctx);
637
638 /* r600_streamout.c */
639 void r600_streamout_buffers_dirty(struct r600_common_context *rctx);
640 void r600_set_streamout_targets(struct pipe_context *ctx,
641 unsigned num_targets,
642 struct pipe_stream_output_target **targets,
643 const unsigned *offset);
644 void r600_emit_streamout_end(struct r600_common_context *rctx);
645 void r600_update_prims_generated_query_state(struct r600_common_context *rctx,
646 unsigned type, int diff);
647 void r600_streamout_init(struct r600_common_context *rctx);
648
649 /* r600_test_dma.c */
650 void r600_test_dma(struct r600_common_screen *rscreen);
651
652 /* r600_texture.c */
653 bool r600_prepare_for_dma_blit(struct r600_common_context *rctx,
654 struct r600_texture *rdst,
655 unsigned dst_level, unsigned dstx,
656 unsigned dsty, unsigned dstz,
657 struct r600_texture *rsrc,
658 unsigned src_level,
659 const struct pipe_box *src_box);
660 void r600_texture_get_fmask_info(struct r600_common_screen *rscreen,
661 struct r600_texture *rtex,
662 unsigned nr_samples,
663 struct r600_fmask_info *out);
664 void r600_texture_get_cmask_info(struct r600_common_screen *rscreen,
665 struct r600_texture *rtex,
666 struct r600_cmask_info *out);
667 bool r600_init_flushed_depth_texture(struct pipe_context *ctx,
668 struct pipe_resource *texture,
669 struct r600_texture **staging);
670 void r600_print_texture_info(struct r600_texture *rtex, FILE *f);
671 struct pipe_resource *r600_texture_create(struct pipe_screen *screen,
672 const struct pipe_resource *templ);
673 struct pipe_surface *r600_create_surface_custom(struct pipe_context *pipe,
674 struct pipe_resource *texture,
675 const struct pipe_surface *templ,
676 unsigned width, unsigned height);
677 unsigned r600_translate_colorswap(enum pipe_format format, bool do_endian_swap);
678 void evergreen_do_fast_color_clear(struct r600_common_context *rctx,
679 struct pipe_framebuffer_state *fb,
680 struct r600_atom *fb_state,
681 unsigned *buffers, unsigned *dirty_cbufs,
682 const union pipe_color_union *color);
683 bool r600_texture_disable_dcc(struct r600_common_screen *rscreen,
684 struct r600_texture *rtex);
685 void r600_init_screen_texture_functions(struct r600_common_screen *rscreen);
686 void r600_init_context_texture_functions(struct r600_common_context *rctx);
687
688 /* r600_viewport.c */
689 void evergreen_apply_scissor_bug_workaround(struct r600_common_context *rctx,
690 struct pipe_scissor_state *scissor);
691 void r600_set_scissor_enable(struct r600_common_context *rctx, bool enable);
692 void r600_update_vs_writes_viewport_index(struct r600_common_context *rctx,
693 struct tgsi_shader_info *info);
694 void r600_init_viewport_functions(struct r600_common_context *rctx);
695
696 /* cayman_msaa.c */
697 extern const uint32_t eg_sample_locs_2x[4];
698 extern const unsigned eg_max_dist_2x;
699 extern const uint32_t eg_sample_locs_4x[4];
700 extern const unsigned eg_max_dist_4x;
701 void cayman_get_sample_position(struct pipe_context *ctx, unsigned sample_count,
702 unsigned sample_index, float *out_value);
703 void cayman_init_msaa(struct pipe_context *ctx);
704 void cayman_emit_msaa_sample_locs(struct radeon_winsys_cs *cs, int nr_samples);
705 void cayman_emit_msaa_config(struct radeon_winsys_cs *cs, int nr_samples,
706 int ps_iter_samples, int overrast_samples);
707
708
709 /* Inline helpers. */
710
711 static inline struct r600_resource *r600_resource(struct pipe_resource *r)
712 {
713 return (struct r600_resource*)r;
714 }
715
716 static inline void
717 r600_resource_reference(struct r600_resource **ptr, struct r600_resource *res)
718 {
719 pipe_resource_reference((struct pipe_resource **)ptr,
720 (struct pipe_resource *)res);
721 }
722
723 static inline bool r600_get_strmout_en(struct r600_common_context *rctx)
724 {
725 return rctx->streamout.streamout_enabled ||
726 rctx->streamout.prims_gen_query_enabled;
727 }
728
729 #define SQ_TEX_XY_FILTER_POINT 0x00
730 #define SQ_TEX_XY_FILTER_BILINEAR 0x01
731 #define SQ_TEX_XY_FILTER_ANISO_POINT 0x02
732 #define SQ_TEX_XY_FILTER_ANISO_BILINEAR 0x03
733
734 static inline unsigned eg_tex_filter(unsigned filter, unsigned max_aniso)
735 {
736 if (filter == PIPE_TEX_FILTER_LINEAR)
737 return max_aniso > 1 ? SQ_TEX_XY_FILTER_ANISO_BILINEAR
738 : SQ_TEX_XY_FILTER_BILINEAR;
739 else
740 return max_aniso > 1 ? SQ_TEX_XY_FILTER_ANISO_POINT
741 : SQ_TEX_XY_FILTER_POINT;
742 }
743
744 static inline unsigned r600_tex_aniso_filter(unsigned filter)
745 {
746 if (filter < 2)
747 return 0;
748 if (filter < 4)
749 return 1;
750 if (filter < 8)
751 return 2;
752 if (filter < 16)
753 return 3;
754 return 4;
755 }
756
757 static inline unsigned r600_wavefront_size(enum radeon_family family)
758 {
759 switch (family) {
760 case CHIP_RV610:
761 case CHIP_RS780:
762 case CHIP_RV620:
763 case CHIP_RS880:
764 return 16;
765 case CHIP_RV630:
766 case CHIP_RV635:
767 case CHIP_RV730:
768 case CHIP_RV710:
769 case CHIP_PALM:
770 case CHIP_CEDAR:
771 return 32;
772 default:
773 return 64;
774 }
775 }
776
777 static inline enum radeon_bo_priority
778 r600_get_sampler_view_priority(struct r600_resource *res)
779 {
780 if (res->b.b.target == PIPE_BUFFER)
781 return RADEON_PRIO_SAMPLER_BUFFER;
782
783 if (res->b.b.nr_samples > 1)
784 return RADEON_PRIO_SAMPLER_TEXTURE_MSAA;
785
786 return RADEON_PRIO_SAMPLER_TEXTURE;
787 }
788
789 #define COMPUTE_DBG(rscreen, fmt, args...) \
790 do { \
791 if ((rscreen->b.debug_flags & DBG_COMPUTE)) fprintf(stderr, fmt, ##args); \
792 } while (0);
793
794 #define R600_ERR(fmt, args...) \
795 fprintf(stderr, "EE %s:%d %s - " fmt, __FILE__, __LINE__, __func__, ##args)
796
797 /* For MSAA sample positions. */
798 #define FILL_SREG(s0x, s0y, s1x, s1y, s2x, s2y, s3x, s3y) \
799 (((s0x) & 0xf) | (((unsigned)(s0y) & 0xf) << 4) | \
800 (((unsigned)(s1x) & 0xf) << 8) | (((unsigned)(s1y) & 0xf) << 12) | \
801 (((unsigned)(s2x) & 0xf) << 16) | (((unsigned)(s2y) & 0xf) << 20) | \
802 (((unsigned)(s3x) & 0xf) << 24) | (((unsigned)(s3y) & 0xf) << 28))
803
804 #endif