2 * Copyright 2013 Advanced Micro Devices, Inc.
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:
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
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
23 * Authors: Marek Olšák <maraeo@gmail.com>
28 * This file contains common screen and context structures and functions
29 * for r600g and radeonsi.
32 #ifndef R600_PIPE_COMMON_H
33 #define R600_PIPE_COMMON_H
37 #include "radeon/radeon_winsys.h"
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"
46 #define ATI_VENDOR_ID 0x1002
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)
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)
59 /* special primitive types */
60 #define R600_PRIM_RECTANGLE_LIST PIPE_PRIM_MAX
64 #define DBG_TEX (1 << 0)
66 #define DBG_COMPUTE (1 << 2)
67 #define DBG_VM (1 << 3)
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)
82 #define DBG_TEST_DMA (1 << 20)
83 /* Bits 21-31 are reserved for the r600g driver. */
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 #define DBG_UNSAFE_MATH (1llu << 49)
103 #define DBG_NO_DCC_FB (1llu << 50)
105 #define R600_MAP_BUFFER_ALIGNMENT 64
106 #define R600_MAX_VIEWPORTS 16
108 enum r600_coherency
{
109 R600_COHERENCY_NONE
, /* no cache flushes needed */
110 R600_COHERENCY_SHADER
,
111 R600_COHERENCY_CB_META
,
114 #ifdef PIPE_ARCH_BIG_ENDIAN
115 #define R600_BIG_ENDIAN 1
117 #define R600_BIG_ENDIAN 0
120 struct r600_common_context
;
121 struct r600_perfcounters
;
122 struct tgsi_shader_info
;
124 struct radeon_shader_reloc
{
129 struct radeon_shader_binary
{
134 /** Config/Context register state that accompanies this shader.
135 * This is a stream of dword pairs. First dword contains the
136 * register address, the second dword contains the value.*/
137 unsigned char *config
;
138 unsigned config_size
;
140 /** The number of bytes of config information for each global symbol.
142 unsigned config_size_per_symbol
;
144 /** Constant data accessed by the shader. This will be uploaded
145 * into a constant buffer. */
146 unsigned char *rodata
;
147 unsigned rodata_size
;
149 /** List of symbol offsets for the shader */
150 uint64_t *global_symbol_offsets
;
151 unsigned global_symbol_count
;
153 struct radeon_shader_reloc
*relocs
;
154 unsigned reloc_count
;
156 /** Disassembled shader in a string. */
158 char *llvm_ir_string
;
161 void radeon_shader_binary_init(struct radeon_shader_binary
*b
);
162 void radeon_shader_binary_clean(struct radeon_shader_binary
*b
);
164 /* Only 32-bit buffer allocations are supported, gallium doesn't support more
167 struct r600_resource
{
170 /* Winsys objects. */
171 struct pb_buffer
*buf
;
172 uint64_t gpu_address
;
173 /* Memory usage if the buffer placement is optimal. */
177 /* Resource state. */
178 enum radeon_bo_domain domains
;
180 /* The buffer range which is initialized (with a write transfer,
181 * streamout, DMA, or as a random access target). The rest of
182 * the buffer is considered invalid and can be mapped unsynchronized.
184 * This allows unsychronized mapping of a buffer range which hasn't
185 * been used yet. It's for applications which forget to use
186 * the unsynchronized map flag and expect the driver to figure it out.
188 struct util_range valid_buffer_range
;
190 /* For buffers only. This indicates that a write operation has been
191 * performed by TC L2, but the cache hasn't been flushed.
192 * Any hw block which doesn't use or bypasses TC L2 should check this
193 * flag and flush the cache before using the buffer.
195 * For example, TC L2 must be flushed if a buffer which has been
196 * modified by a shader store instruction is about to be used as
197 * an index buffer. The reason is that VGT DMA index fetching doesn't
202 /* Whether the resource has been exported via resource_get_handle. */
204 unsigned external_usage
; /* PIPE_HANDLE_USAGE_* */
207 struct r600_transfer
{
208 struct pipe_transfer transfer
;
209 struct r600_resource
*staging
;
213 struct r600_fmask_info
{
217 unsigned pitch_in_pixels
;
218 unsigned bank_height
;
219 unsigned slice_tile_max
;
220 unsigned tile_mode_index
;
223 struct r600_cmask_info
{
231 unsigned slice_tile_max
;
232 unsigned base_address_reg
;
235 struct r600_htile_info
{
242 struct r600_texture
{
243 struct r600_resource resource
;
246 unsigned num_level0_transfers
;
251 unsigned dirty_level_mask
; /* each bit says if that mipmap is compressed */
252 unsigned stencil_dirty_level_mask
; /* each bit says if that mipmap is compressed */
253 struct r600_texture
*flushed_depth_texture
;
254 struct radeon_surf surface
;
256 /* Colorbuffer compression and fast clear. */
257 struct r600_fmask_info fmask
;
258 struct r600_cmask_info cmask
;
259 struct r600_resource
*cmask_buffer
;
260 uint64_t dcc_offset
; /* 0 = disabled */
261 unsigned cb_color_info
; /* fast clear enable bit */
262 unsigned color_clear_value
[2];
263 unsigned last_msaa_resolve_target_micro_mode
;
265 /* Depth buffer compression and fast clear. */
266 struct r600_htile_info htile
;
267 struct r600_resource
*htile_buffer
;
268 bool depth_cleared
; /* if it was cleared at least once */
269 float depth_clear_value
;
270 bool stencil_cleared
; /* if it was cleared at least once */
271 uint8_t stencil_clear_value
;
273 bool non_disp_tiling
; /* R600-Cayman only */
275 /* Whether the texture is a displayable back buffer and needs DCC
276 * decompression, which is expensive. Therefore, it's enabled only
277 * if statistics suggest that it will pay off and it's allocated
278 * separately. It can't be bound as a sampler by apps. Limited to
279 * target == 2D and last_level == 0. If enabled, dcc_offset contains
280 * the absolute GPUVM address, not the relative one.
282 struct r600_resource
*dcc_separate_buffer
;
283 /* When DCC is temporarily disabled, the separate buffer is here. */
284 struct r600_resource
*last_dcc_separate_buffer
;
285 /* We need to track DCC dirtiness, because st/dri usually calls
286 * flush_resource twice per frame (not a bug) and we don't wanna
287 * decompress DCC twice. Also, the dirty tracking must be done even
288 * if DCC isn't used, because it's required by the DCC usage analysis
289 * for a possible future enablement.
291 bool separate_dcc_dirty
;
292 /* Statistics gathering for the DCC enablement heuristic. */
293 bool dcc_gather_statistics
;
294 /* Estimate of how much this color buffer is written to in units of
295 * full-screen draws: ps_invocations / (width * height)
296 * Shader kills, late Z, and blending with trivial discards make it
297 * inaccurate (we need to count CB updates, not PS invocations).
299 unsigned ps_draw_ratio
;
300 /* The number of clears since the last DCC usage analysis. */
301 unsigned num_slow_clears
;
303 /* Counter that should be non-zero if the texture is bound to a
304 * framebuffer. Implemented in radeonsi only.
306 uint32_t framebuffers_bound
;
309 struct r600_surface
{
310 struct pipe_surface base
;
311 const struct radeon_surf_level
*level_info
;
313 bool color_initialized
;
314 bool depth_initialized
;
316 /* Misc. color flags. */
317 bool alphatest_bypass
;
321 /* Color registers. */
322 unsigned cb_color_info
;
323 unsigned cb_color_base
;
324 unsigned cb_color_view
;
325 unsigned cb_color_size
; /* R600 only */
326 unsigned cb_color_dim
; /* EG only */
327 unsigned cb_color_pitch
; /* EG and later */
328 unsigned cb_color_slice
; /* EG and later */
329 unsigned cb_color_attrib
; /* EG and later */
330 unsigned cb_dcc_control
; /* VI and later */
331 unsigned cb_color_fmask
; /* CB_COLORn_FMASK (EG and later) or CB_COLORn_FRAG (r600) */
332 unsigned cb_color_fmask_slice
; /* EG and later */
333 unsigned cb_color_cmask
; /* CB_COLORn_TILE (r600 only) */
334 unsigned cb_color_mask
; /* R600 only */
335 unsigned spi_shader_col_format
; /* SI+, no blending, no alpha-to-coverage. */
336 unsigned spi_shader_col_format_alpha
; /* SI+, alpha-to-coverage */
337 unsigned spi_shader_col_format_blend
; /* SI+, blending without alpha. */
338 unsigned spi_shader_col_format_blend_alpha
; /* SI+, blending with alpha. */
339 struct r600_resource
*cb_buffer_fmask
; /* Used for FMASK relocations. R600 only */
340 struct r600_resource
*cb_buffer_cmask
; /* Used for CMASK relocations. R600 only */
343 unsigned db_depth_info
; /* R600 only, then SI and later */
344 unsigned db_z_info
; /* EG and later */
345 unsigned db_depth_base
; /* DB_Z_READ/WRITE_BASE (EG and later) or DB_DEPTH_BASE (r600) */
346 unsigned db_depth_view
;
347 unsigned db_depth_size
;
348 unsigned db_depth_slice
; /* EG and later */
349 unsigned db_stencil_base
; /* EG and later */
350 unsigned db_stencil_info
; /* EG and later */
351 unsigned db_prefetch_limit
; /* R600 only */
352 unsigned db_htile_surface
;
353 unsigned db_htile_data_base
;
354 unsigned db_preload_control
; /* EG and later */
357 struct r600_common_screen
{
358 struct pipe_screen b
;
359 struct radeon_winsys
*ws
;
360 enum radeon_family family
;
361 enum chip_class chip_class
;
362 struct radeon_info info
;
363 uint64_t debug_flags
;
367 /* Texture filter settings. */
368 int force_aniso
; /* -1 = disabled */
370 /* Auxiliary context. Mainly used to initialize resources.
371 * It must be locked prior to using and flushed before unlocking. */
372 struct pipe_context
*aux_context
;
373 pipe_mutex aux_context_lock
;
375 /* This must be in the screen, because UE4 uses one context for
376 * compilation and another one for rendering.
378 unsigned num_compilations
;
379 /* Along with ST_DEBUG=precompile, this should show if applications
380 * are loading shaders on demand. This is a monotonic counter.
382 unsigned num_shaders_created
;
384 /* GPU load thread. */
385 pipe_mutex gpu_load_mutex
;
386 pipe_thread gpu_load_thread
;
387 unsigned gpu_load_counter_busy
;
388 unsigned gpu_load_counter_idle
;
389 volatile unsigned gpu_load_stop_thread
; /* bool */
391 char renderer_string
[64];
393 /* Performance counters. */
394 struct r600_perfcounters
*perfcounters
;
396 /* If pipe_screen wants to re-emit the framebuffer state of all
397 * contexts, it should atomically increment this. Each context will
398 * compare this with its own last known value of the counter before
399 * drawing and re-emit the framebuffer state accordingly.
401 unsigned dirty_fb_counter
;
403 /* Atomically increment this counter when an existing texture's
404 * metadata is enabled or disabled in a way that requires changing
405 * contexts' compressed texture binding masks.
407 unsigned compressed_colortex_counter
;
409 /* Atomically increment this counter when an existing texture's
410 * backing buffer or tile mode parameters have changed that requires
411 * recomputation of shader descriptors.
413 unsigned dirty_tex_descriptor_counter
;
415 void (*query_opaque_metadata
)(struct r600_common_screen
*rscreen
,
416 struct r600_texture
*rtex
,
417 struct radeon_bo_metadata
*md
);
419 void (*apply_opaque_metadata
)(struct r600_common_screen
*rscreen
,
420 struct r600_texture
*rtex
,
421 struct radeon_bo_metadata
*md
);
424 /* This encapsulates a state or an operation which can emitted into the GPU
427 void (*emit
)(struct r600_common_context
*ctx
, struct r600_atom
*state
);
432 struct r600_so_target
{
433 struct pipe_stream_output_target b
;
435 /* The buffer where BUFFER_FILLED_SIZE is stored. */
436 struct r600_resource
*buf_filled_size
;
437 unsigned buf_filled_size_offset
;
438 bool buf_filled_size_valid
;
440 unsigned stride_in_dw
;
443 struct r600_streamout
{
444 struct r600_atom begin_atom
;
446 unsigned num_dw_for_end
;
448 unsigned enabled_mask
;
449 unsigned num_targets
;
450 struct r600_so_target
*targets
[PIPE_MAX_SO_BUFFERS
];
452 unsigned append_bitmask
;
455 /* External state which comes from the vertex shader,
456 * it must be set explicitly when binding a shader. */
457 unsigned *stride_in_dw
;
458 unsigned enabled_stream_buffers_mask
; /* stream0 buffers0-3 in 4 LSB */
460 /* The state of VGT_STRMOUT_BUFFER_(CONFIG|EN). */
461 unsigned hw_enabled_mask
;
463 /* The state of VGT_STRMOUT_(CONFIG|EN). */
464 struct r600_atom enable_atom
;
465 bool streamout_enabled
;
466 bool prims_gen_query_enabled
;
467 int num_prims_gen_queries
;
470 struct r600_signed_scissor
{
477 struct r600_scissors
{
478 struct r600_atom atom
;
480 struct pipe_scissor_state states
[R600_MAX_VIEWPORTS
];
483 struct r600_viewports
{
484 struct r600_atom atom
;
486 struct pipe_viewport_state states
[R600_MAX_VIEWPORTS
];
487 struct r600_signed_scissor as_scissor
[R600_MAX_VIEWPORTS
];
491 struct radeon_winsys_cs
*cs
;
492 void (*flush
)(void *ctx
, unsigned flags
,
493 struct pipe_fence_handle
**fence
);
496 /* Saved CS data for debugging features. */
497 struct radeon_saved_cs
{
501 struct radeon_bo_list_item
*bo_list
;
505 struct r600_common_context
{
506 struct pipe_context b
; /* base class */
508 struct r600_common_screen
*screen
;
509 struct radeon_winsys
*ws
;
510 struct radeon_winsys_ctx
*ctx
;
511 enum radeon_family family
;
512 enum chip_class chip_class
;
513 struct r600_ring gfx
;
514 struct r600_ring dma
;
515 struct pipe_fence_handle
*last_gfx_fence
;
516 struct pipe_fence_handle
*last_sdma_fence
;
517 unsigned initial_gfx_cs_size
;
518 unsigned gpu_reset_counter
;
519 unsigned last_dirty_fb_counter
;
520 unsigned last_compressed_colortex_counter
;
521 unsigned last_dirty_tex_descriptor_counter
;
523 struct u_upload_mgr
*uploader
;
524 struct u_suballocator
*allocator_zeroed_memory
;
525 struct util_slab_mempool pool_transfers
;
527 /* Current unaccounted memory usage. */
532 struct r600_streamout streamout
;
533 struct r600_scissors scissors
;
534 struct r600_viewports viewports
;
535 bool scissor_enabled
;
536 bool vs_writes_viewport_index
;
537 bool vs_disables_clipping_viewport
;
539 /* Additional context states. */
540 unsigned flags
; /* flush flags */
543 /* Maintain the list of active queries for pausing between IBs. */
544 int num_occlusion_queries
;
545 int num_perfect_occlusion_queries
;
546 struct list_head active_queries
;
547 unsigned num_cs_dw_queries_suspend
;
548 /* Additional hardware info. */
549 unsigned backend_mask
;
550 unsigned max_db
; /* for OQ */
552 unsigned num_draw_calls
;
553 unsigned num_spill_draw_calls
;
554 unsigned num_compute_calls
;
555 unsigned num_spill_compute_calls
;
556 unsigned num_dma_calls
;
557 uint64_t num_alloc_tex_transfer_bytes
;
558 unsigned last_tex_ps_draw_ratio
; /* for query */
560 /* Render condition. */
561 struct r600_atom render_cond_atom
;
562 struct pipe_query
*render_cond
;
563 unsigned render_cond_mode
;
564 bool render_cond_invert
;
565 bool render_cond_force_off
; /* for u_blitter */
567 /* MSAA sample locations.
568 * The first index is the sample index.
569 * The second index is the coordinate: X, Y. */
570 float sample_locations_1x
[1][2];
571 float sample_locations_2x
[2][2];
572 float sample_locations_4x
[4][2];
573 float sample_locations_8x
[8][2];
574 float sample_locations_16x
[16][2];
576 /* Statistics gathering for the DCC enablement heuristic. It can't be
577 * in r600_texture because r600_texture can be shared by multiple
578 * contexts. This is for back buffers only. We shouldn't get too many
581 * X11 DRI3 rotates among a finite set of back buffers. They should
582 * all fit in this array. If they don't, separate DCC might never be
583 * enabled by DCC stat gathering.
586 struct r600_texture
*tex
;
587 /* Query queue: 0 = usually active, 1 = waiting, 2 = readback. */
588 struct pipe_query
*ps_stats
[3];
589 /* If all slots are used and another slot is needed,
590 * the least recently used slot is evicted based on this. */
591 int64_t last_use_timestamp
;
595 /* The list of all texture buffer objects in this context.
596 * This list is walked when a buffer is invalidated/reallocated and
597 * the GPU addresses are updated. */
598 struct list_head texture_buffers
;
600 struct pipe_debug_callback debug
;
602 /* Copy one resource to another using async DMA. */
603 void (*dma_copy
)(struct pipe_context
*ctx
,
604 struct pipe_resource
*dst
,
606 unsigned dst_x
, unsigned dst_y
, unsigned dst_z
,
607 struct pipe_resource
*src
,
609 const struct pipe_box
*src_box
);
611 void (*clear_buffer
)(struct pipe_context
*ctx
, struct pipe_resource
*dst
,
612 uint64_t offset
, uint64_t size
, unsigned value
,
613 enum r600_coherency coher
);
615 void (*blit_decompress_depth
)(struct pipe_context
*ctx
,
616 struct r600_texture
*texture
,
617 struct r600_texture
*staging
,
618 unsigned first_level
, unsigned last_level
,
619 unsigned first_layer
, unsigned last_layer
,
620 unsigned first_sample
, unsigned last_sample
);
622 void (*decompress_dcc
)(struct pipe_context
*ctx
,
623 struct r600_texture
*rtex
);
625 /* Reallocate the buffer and update all resource bindings where
626 * the buffer is bound, including all resource descriptors. */
627 void (*invalidate_buffer
)(struct pipe_context
*ctx
, struct pipe_resource
*buf
);
629 /* Enable or disable occlusion queries. */
630 void (*set_occlusion_query_state
)(struct pipe_context
*ctx
, bool enable
);
632 /* This ensures there is enough space in the command stream. */
633 void (*need_gfx_cs_space
)(struct pipe_context
*ctx
, unsigned num_dw
,
634 bool include_draw_vbo
);
636 void (*set_atom_dirty
)(struct r600_common_context
*ctx
,
637 struct r600_atom
*atom
, bool dirty
);
639 void (*check_vm_faults
)(struct r600_common_context
*ctx
,
640 struct radeon_saved_cs
*saved
,
641 enum ring_type ring
);
645 bool r600_rings_is_buffer_referenced(struct r600_common_context
*ctx
,
646 struct pb_buffer
*buf
,
647 enum radeon_bo_usage usage
);
648 void *r600_buffer_map_sync_with_rings(struct r600_common_context
*ctx
,
649 struct r600_resource
*resource
,
651 void r600_buffer_subdata(struct pipe_context
*ctx
,
652 struct pipe_resource
*buffer
,
653 unsigned usage
, unsigned offset
,
654 unsigned size
, const void *data
);
655 bool r600_init_resource(struct r600_common_screen
*rscreen
,
656 struct r600_resource
*res
,
657 uint64_t size
, unsigned alignment
);
658 struct pipe_resource
*r600_buffer_create(struct pipe_screen
*screen
,
659 const struct pipe_resource
*templ
,
661 struct pipe_resource
* r600_aligned_buffer_create(struct pipe_screen
*screen
,
666 struct pipe_resource
*
667 r600_buffer_from_user_memory(struct pipe_screen
*screen
,
668 const struct pipe_resource
*templ
,
671 r600_invalidate_resource(struct pipe_context
*ctx
,
672 struct pipe_resource
*resource
);
674 /* r600_common_pipe.c */
675 void r600_draw_rectangle(struct blitter_context
*blitter
,
676 int x1
, int y1
, int x2
, int y2
, float depth
,
677 enum blitter_attrib_type type
,
678 const union pipe_color_union
*attrib
);
679 bool r600_common_screen_init(struct r600_common_screen
*rscreen
,
680 struct radeon_winsys
*ws
);
681 void r600_destroy_common_screen(struct r600_common_screen
*rscreen
);
682 void r600_preflush_suspend_features(struct r600_common_context
*ctx
);
683 void r600_postflush_resume_features(struct r600_common_context
*ctx
);
684 bool r600_common_context_init(struct r600_common_context
*rctx
,
685 struct r600_common_screen
*rscreen
,
686 unsigned context_flags
);
687 void r600_common_context_cleanup(struct r600_common_context
*rctx
);
688 void r600_context_add_resource_size(struct pipe_context
*ctx
, struct pipe_resource
*r
);
689 bool r600_can_dump_shader(struct r600_common_screen
*rscreen
,
691 void r600_screen_clear_buffer(struct r600_common_screen
*rscreen
, struct pipe_resource
*dst
,
692 uint64_t offset
, uint64_t size
, unsigned value
,
693 enum r600_coherency coher
);
694 struct pipe_resource
*r600_resource_create_common(struct pipe_screen
*screen
,
695 const struct pipe_resource
*templ
);
696 const char *r600_get_llvm_processor_name(enum radeon_family family
);
697 void r600_need_dma_space(struct r600_common_context
*ctx
, unsigned num_dw
,
698 struct r600_resource
*dst
, struct r600_resource
*src
);
699 void r600_dma_emit_wait_idle(struct r600_common_context
*rctx
);
700 void radeon_save_cs(struct radeon_winsys
*ws
, struct radeon_winsys_cs
*cs
,
701 struct radeon_saved_cs
*saved
);
702 void radeon_clear_saved_cs(struct radeon_saved_cs
*saved
);
704 /* r600_gpu_load.c */
705 void r600_gpu_load_kill_thread(struct r600_common_screen
*rscreen
);
706 uint64_t r600_gpu_load_begin(struct r600_common_screen
*rscreen
);
707 unsigned r600_gpu_load_end(struct r600_common_screen
*rscreen
, uint64_t begin
);
709 /* r600_perfcounters.c */
710 void r600_perfcounters_destroy(struct r600_common_screen
*rscreen
);
713 void r600_init_screen_query_functions(struct r600_common_screen
*rscreen
);
714 void r600_query_init(struct r600_common_context
*rctx
);
715 void r600_suspend_queries(struct r600_common_context
*ctx
);
716 void r600_resume_queries(struct r600_common_context
*ctx
);
717 void r600_query_init_backend_mask(struct r600_common_context
*ctx
);
719 /* r600_streamout.c */
720 void r600_streamout_buffers_dirty(struct r600_common_context
*rctx
);
721 void r600_set_streamout_targets(struct pipe_context
*ctx
,
722 unsigned num_targets
,
723 struct pipe_stream_output_target
**targets
,
724 const unsigned *offset
);
725 void r600_emit_streamout_end(struct r600_common_context
*rctx
);
726 void r600_update_prims_generated_query_state(struct r600_common_context
*rctx
,
727 unsigned type
, int diff
);
728 void r600_streamout_init(struct r600_common_context
*rctx
);
730 /* r600_test_dma.c */
731 void r600_test_dma(struct r600_common_screen
*rscreen
);
734 bool r600_prepare_for_dma_blit(struct r600_common_context
*rctx
,
735 struct r600_texture
*rdst
,
736 unsigned dst_level
, unsigned dstx
,
737 unsigned dsty
, unsigned dstz
,
738 struct r600_texture
*rsrc
,
740 const struct pipe_box
*src_box
);
741 void r600_texture_get_fmask_info(struct r600_common_screen
*rscreen
,
742 struct r600_texture
*rtex
,
744 struct r600_fmask_info
*out
);
745 void r600_texture_get_cmask_info(struct r600_common_screen
*rscreen
,
746 struct r600_texture
*rtex
,
747 struct r600_cmask_info
*out
);
748 bool r600_init_flushed_depth_texture(struct pipe_context
*ctx
,
749 struct pipe_resource
*texture
,
750 struct r600_texture
**staging
);
751 void r600_print_texture_info(struct r600_texture
*rtex
, FILE *f
);
752 struct pipe_resource
*r600_texture_create(struct pipe_screen
*screen
,
753 const struct pipe_resource
*templ
);
754 struct pipe_surface
*r600_create_surface_custom(struct pipe_context
*pipe
,
755 struct pipe_resource
*texture
,
756 const struct pipe_surface
*templ
,
757 unsigned width
, unsigned height
);
758 unsigned r600_translate_colorswap(enum pipe_format format
, bool do_endian_swap
);
759 void vi_separate_dcc_start_query(struct pipe_context
*ctx
,
760 struct r600_texture
*tex
);
761 void vi_separate_dcc_stop_query(struct pipe_context
*ctx
,
762 struct r600_texture
*tex
);
763 void vi_separate_dcc_process_and_reset_stats(struct pipe_context
*ctx
,
764 struct r600_texture
*tex
);
765 void vi_dcc_clear_level(struct r600_common_context
*rctx
,
766 struct r600_texture
*rtex
,
767 unsigned level
, unsigned clear_value
);
768 void evergreen_do_fast_color_clear(struct r600_common_context
*rctx
,
769 struct pipe_framebuffer_state
*fb
,
770 struct r600_atom
*fb_state
,
771 unsigned *buffers
, unsigned *dirty_cbufs
,
772 const union pipe_color_union
*color
);
773 bool r600_texture_disable_dcc(struct r600_common_screen
*rscreen
,
774 struct r600_texture
*rtex
);
775 void r600_init_screen_texture_functions(struct r600_common_screen
*rscreen
);
776 void r600_init_context_texture_functions(struct r600_common_context
*rctx
);
778 /* r600_viewport.c */
779 void evergreen_apply_scissor_bug_workaround(struct r600_common_context
*rctx
,
780 struct pipe_scissor_state
*scissor
);
781 void r600_set_scissor_enable(struct r600_common_context
*rctx
, bool enable
);
782 void r600_update_vs_writes_viewport_index(struct r600_common_context
*rctx
,
783 struct tgsi_shader_info
*info
);
784 void r600_init_viewport_functions(struct r600_common_context
*rctx
);
787 extern const uint32_t eg_sample_locs_2x
[4];
788 extern const unsigned eg_max_dist_2x
;
789 extern const uint32_t eg_sample_locs_4x
[4];
790 extern const unsigned eg_max_dist_4x
;
791 void cayman_get_sample_position(struct pipe_context
*ctx
, unsigned sample_count
,
792 unsigned sample_index
, float *out_value
);
793 void cayman_init_msaa(struct pipe_context
*ctx
);
794 void cayman_emit_msaa_sample_locs(struct radeon_winsys_cs
*cs
, int nr_samples
);
795 void cayman_emit_msaa_config(struct radeon_winsys_cs
*cs
, int nr_samples
,
796 int ps_iter_samples
, int overrast_samples
,
797 unsigned sc_mode_cntl_1
);
800 /* Inline helpers. */
802 static inline struct r600_resource
*r600_resource(struct pipe_resource
*r
)
804 return (struct r600_resource
*)r
;
808 r600_resource_reference(struct r600_resource
**ptr
, struct r600_resource
*res
)
810 pipe_resource_reference((struct pipe_resource
**)ptr
,
811 (struct pipe_resource
*)res
);
815 r600_texture_reference(struct r600_texture
**ptr
, struct r600_texture
*res
)
817 pipe_resource_reference((struct pipe_resource
**)ptr
, &res
->resource
.b
.b
);
820 static inline bool r600_get_strmout_en(struct r600_common_context
*rctx
)
822 return rctx
->streamout
.streamout_enabled
||
823 rctx
->streamout
.prims_gen_query_enabled
;
826 #define SQ_TEX_XY_FILTER_POINT 0x00
827 #define SQ_TEX_XY_FILTER_BILINEAR 0x01
828 #define SQ_TEX_XY_FILTER_ANISO_POINT 0x02
829 #define SQ_TEX_XY_FILTER_ANISO_BILINEAR 0x03
831 static inline unsigned eg_tex_filter(unsigned filter
, unsigned max_aniso
)
833 if (filter
== PIPE_TEX_FILTER_LINEAR
)
834 return max_aniso
> 1 ? SQ_TEX_XY_FILTER_ANISO_BILINEAR
835 : SQ_TEX_XY_FILTER_BILINEAR
;
837 return max_aniso
> 1 ? SQ_TEX_XY_FILTER_ANISO_POINT
838 : SQ_TEX_XY_FILTER_POINT
;
841 static inline unsigned r600_tex_aniso_filter(unsigned filter
)
854 static inline unsigned r600_wavefront_size(enum radeon_family family
)
874 static inline enum radeon_bo_priority
875 r600_get_sampler_view_priority(struct r600_resource
*res
)
877 if (res
->b
.b
.target
== PIPE_BUFFER
)
878 return RADEON_PRIO_SAMPLER_BUFFER
;
880 if (res
->b
.b
.nr_samples
> 1)
881 return RADEON_PRIO_SAMPLER_TEXTURE_MSAA
;
883 return RADEON_PRIO_SAMPLER_TEXTURE
;
887 r600_can_sample_zs(struct r600_texture
*tex
, bool stencil_sampler
)
889 return (stencil_sampler
&& tex
->can_sample_s
) ||
890 (!stencil_sampler
&& tex
->can_sample_z
);
893 #define COMPUTE_DBG(rscreen, fmt, args...) \
895 if ((rscreen->b.debug_flags & DBG_COMPUTE)) fprintf(stderr, fmt, ##args); \
898 #define R600_ERR(fmt, args...) \
899 fprintf(stderr, "EE %s:%d %s - " fmt, __FILE__, __LINE__, __func__, ##args)
901 /* For MSAA sample positions. */
902 #define FILL_SREG(s0x, s0y, s1x, s1y, s2x, s2y, s3x, s3y) \
903 (((s0x) & 0xf) | (((unsigned)(s0y) & 0xf) << 4) | \
904 (((unsigned)(s1x) & 0xf) << 8) | (((unsigned)(s1y) & 0xf) << 12) | \
905 (((unsigned)(s2x) & 0xf) << 16) | (((unsigned)(s2y) & 0xf) << 20) | \
906 (((unsigned)(s3x) & 0xf) << 24) | (((unsigned)(s3y) & 0xf) << 28))