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