2 * Copyright 2010 Jerome Glisse <glisse@freedesktop.org>
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 * on the rights to use, copy, modify, merge, publish, distribute, sub
8 * license, and/or sell copies of the Software, and to permit persons to whom
9 * the 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 NON-INFRINGEMENT. IN NO EVENT SHALL
18 * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
19 * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
20 * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
21 * USE OR OTHER DEALINGS IN THE SOFTWARE.
29 #include "radeon/r600_pipe_common.h"
30 #include "radeon/r600_cs.h"
31 #include "r600_public.h"
33 #include "util/u_suballoc.h"
34 #include "util/list.h"
35 #include "util/u_transfer.h"
36 #include "util/u_memory.h"
38 #include "tgsi/tgsi_scan.h"
40 #define R600_NUM_ATOMS 52
43 #define R600_CONTEXT_INV_VERTEX_CACHE (R600_CONTEXT_PRIVATE_FLAG << 0)
44 #define R600_CONTEXT_INV_TEX_CACHE (R600_CONTEXT_PRIVATE_FLAG << 1)
45 #define R600_CONTEXT_INV_CONST_CACHE (R600_CONTEXT_PRIVATE_FLAG << 2)
46 /* read-write caches */
47 #define R600_CONTEXT_FLUSH_AND_INV (R600_CONTEXT_PRIVATE_FLAG << 3)
48 #define R600_CONTEXT_FLUSH_AND_INV_CB_META (R600_CONTEXT_PRIVATE_FLAG << 4)
49 #define R600_CONTEXT_FLUSH_AND_INV_DB_META (R600_CONTEXT_PRIVATE_FLAG << 5)
50 #define R600_CONTEXT_FLUSH_AND_INV_DB (R600_CONTEXT_PRIVATE_FLAG << 6)
51 #define R600_CONTEXT_FLUSH_AND_INV_CB (R600_CONTEXT_PRIVATE_FLAG << 7)
52 /* engine synchronization */
53 #define R600_CONTEXT_PS_PARTIAL_FLUSH (R600_CONTEXT_PRIVATE_FLAG << 8)
54 #define R600_CONTEXT_WAIT_3D_IDLE (R600_CONTEXT_PRIVATE_FLAG << 9)
55 #define R600_CONTEXT_WAIT_CP_DMA_IDLE (R600_CONTEXT_PRIVATE_FLAG << 10)
57 /* the number of CS dwords for flushing and drawing */
58 #define R600_MAX_FLUSH_CS_DWORDS 18
59 #define R600_MAX_DRAW_CS_DWORDS 58
60 #define R600_MAX_PFP_SYNC_ME_DWORDS 16
62 #define R600_MAX_USER_CONST_BUFFERS 13
63 #define R600_MAX_DRIVER_CONST_BUFFERS 3
64 #define R600_MAX_CONST_BUFFERS (R600_MAX_USER_CONST_BUFFERS + R600_MAX_DRIVER_CONST_BUFFERS)
66 /* start driver buffers after user buffers */
67 #define R600_BUFFER_INFO_CONST_BUFFER (R600_MAX_USER_CONST_BUFFERS)
68 #define R600_UCP_SIZE (4*4*8)
69 #define R600_BUFFER_INFO_OFFSET (R600_UCP_SIZE)
71 #define R600_LDS_INFO_CONST_BUFFER (R600_MAX_USER_CONST_BUFFERS + 1)
73 * Note GS doesn't use a constant buffer binding, just a resource index,
74 * so it's fine to have it exist at index 16.
76 #define R600_GS_RING_CONST_BUFFER (R600_MAX_USER_CONST_BUFFERS + 2)
77 /* Currently R600_MAX_CONST_BUFFERS just fits on the hw, which has a limit
78 * of 16 const buffers.
79 * UCP/SAMPLE_POSITIONS are never accessed by same shader stage so they can use the same id.
81 * In order to support d3d 11 mandated minimum of 15 user const buffers
82 * we'd have to squash all use cases into one driver buffer.
84 #define R600_MAX_CONST_BUFFER_SIZE (4096 * sizeof(float[4]))
87 #define R600_HW_STAGE_PS 0
88 #define R600_HW_STAGE_VS 1
89 #define R600_HW_STAGE_GS 2
90 #define R600_HW_STAGE_ES 3
91 #define EG_HW_STAGE_LS 4
92 #define EG_HW_STAGE_HS 5
94 #define R600_NUM_HW_STAGES 4
95 #define EG_NUM_HW_STAGES 6
99 union r600_shader_key
;
101 /* This is an atom containing GPU commands that never change.
102 * This is supposed to be copied directly into the CS. */
103 struct r600_command_buffer
{
110 struct r600_db_state
{
111 struct r600_atom atom
;
112 struct r600_surface
*rsurf
;
115 struct r600_db_misc_state
{
116 struct r600_atom atom
;
117 bool occlusion_queries_disabled
;
118 bool flush_depthstencil_through_cb
;
119 bool flush_depth_inplace
;
120 bool flush_stencil_inplace
;
121 bool copy_depth
, copy_stencil
;
122 unsigned copy_sample
;
123 unsigned log_samples
;
124 unsigned db_shader_control
;
126 uint8_t ps_conservative_z
;
129 struct r600_cb_misc_state
{
130 struct r600_atom atom
;
131 unsigned cb_color_control
; /* this comes from blend state */
132 unsigned blend_colormask
; /* 8*4 bits for 8 RGBA colorbuffers */
134 unsigned nr_ps_color_outputs
;
139 struct r600_clip_misc_state
{
140 struct r600_atom atom
;
141 unsigned pa_cl_clip_cntl
; /* from rasterizer */
142 unsigned pa_cl_vs_out_cntl
; /* from vertex shader */
143 unsigned clip_plane_enable
; /* from rasterizer */
144 unsigned clip_dist_write
; /* from vertex shader */
145 boolean clip_disable
; /* from vertex shader */
146 boolean vs_out_viewport
; /* from vertex shader */
149 struct r600_alphatest_state
{
150 struct r600_atom atom
;
151 unsigned sx_alpha_test_control
; /* this comes from dsa state */
152 unsigned sx_alpha_ref
; /* this comes from dsa state */
154 bool cb0_export_16bpc
; /* from set_framebuffer_state */
157 struct r600_vgt_state
{
158 struct r600_atom atom
;
159 uint32_t vgt_multi_prim_ib_reset_en
;
160 uint32_t vgt_multi_prim_ib_reset_indx
;
161 uint32_t vgt_indx_offset
;
162 bool last_draw_was_indirect
;
165 struct r600_blend_color
{
166 struct r600_atom atom
;
167 struct pipe_blend_color state
;
170 struct r600_clip_state
{
171 struct r600_atom atom
;
172 struct pipe_clip_state state
;
175 struct r600_cs_shader_state
{
176 struct r600_atom atom
;
177 unsigned kernel_index
;
179 struct r600_pipe_compute
*shader
;
182 struct r600_framebuffer
{
183 struct r600_atom atom
;
184 struct pipe_framebuffer_state state
;
185 unsigned compressed_cb_mask
;
189 bool is_msaa_resolve
;
192 struct r600_sample_mask
{
193 struct r600_atom atom
;
194 uint16_t sample_mask
; /* there are only 8 bits on EG, 16 bits on Cayman */
197 struct r600_config_state
{
198 struct r600_atom atom
;
199 unsigned sq_gpr_resource_mgmt_1
;
200 unsigned sq_gpr_resource_mgmt_2
;
201 unsigned sq_gpr_resource_mgmt_3
;
202 bool dyn_gpr_enabled
;
205 struct r600_stencil_ref
212 struct r600_stencil_ref_state
{
213 struct r600_atom atom
;
214 struct r600_stencil_ref state
;
215 struct pipe_stencil_ref pipe_state
;
218 struct r600_shader_stages_state
{
219 struct r600_atom atom
;
220 unsigned geom_enable
;
223 struct r600_gs_rings_state
{
224 struct r600_atom atom
;
226 struct pipe_constant_buffer esgs_ring
;
227 struct pipe_constant_buffer gsvs_ring
;
230 /* This must start from 16. */
232 #define DBG_NO_CP_DMA (1 << 30)
234 #define DBG_NO_SB (1 << 21)
235 #define DBG_SB_CS (1 << 22)
236 #define DBG_SB_DRY_RUN (1 << 23)
237 #define DBG_SB_STAT (1 << 24)
238 #define DBG_SB_DUMP (1 << 25)
239 #define DBG_SB_NO_FALLBACK (1 << 26)
240 #define DBG_SB_DISASM (1 << 27)
241 #define DBG_SB_SAFEMATH (1 << 28)
244 struct r600_common_screen b
;
246 bool has_compressed_msaa_texturing
;
248 /*for compute global memory binding, we allocate stuff here, instead of
250 * XXX: Not sure if this is the best place for global_pool. Also,
251 * it's not thread safe, so it won't work with multiple contexts. */
252 struct compute_memory_pool
*global_pool
;
255 struct r600_pipe_sampler_view
{
256 struct pipe_sampler_view base
;
257 struct list_head list
;
258 struct r600_resource
*tex_resource
;
259 uint32_t tex_resource_words
[8];
260 bool skip_mip_address_reloc
;
261 bool is_stencil_sampler
;
264 struct r600_rasterizer_state
{
265 struct r600_command_buffer buffer
;
268 unsigned sprite_coord_enable
;
269 unsigned clip_plane_enable
;
270 unsigned pa_sc_line_stipple
;
271 unsigned pa_cl_clip_cntl
;
272 unsigned pa_su_sc_mode_cntl
;
276 bool offset_units_unscaled
;
278 bool multisample_enable
;
281 struct r600_poly_offset_state
{
282 struct r600_atom atom
;
283 enum pipe_format zs_format
;
286 bool offset_units_unscaled
;
289 struct r600_blend_state
{
290 struct r600_command_buffer buffer
;
291 struct r600_command_buffer buffer_no_blend
;
292 unsigned cb_target_mask
;
293 unsigned cb_color_control
;
294 unsigned cb_color_control_no_blend
;
299 struct r600_dsa_state
{
300 struct r600_command_buffer buffer
;
305 unsigned sx_alpha_test_control
;
308 struct r600_pipe_shader
;
310 struct r600_pipe_shader_selector
{
311 struct r600_pipe_shader
*current
;
313 struct tgsi_token
*tokens
;
314 struct pipe_stream_output_info so
;
315 struct tgsi_shader_info info
;
317 unsigned num_shaders
;
319 /* PIPE_SHADER_[VERTEX|FRAGMENT|...] */
322 /* geometry shader properties */
323 unsigned gs_output_prim
;
324 unsigned gs_max_out_vertices
;
325 unsigned gs_num_invocations
;
328 uint64_t lds_patch_outputs_written_mask
;
329 uint64_t lds_outputs_written_mask
;
330 unsigned nr_ps_max_color_exports
;
333 struct r600_pipe_sampler_state
{
334 uint32_t tex_sampler_words
[3];
335 union pipe_color_union border_color
;
336 bool border_color_use
;
337 bool seamless_cube_map
;
340 /* needed for blitter save */
341 #define NUM_TEX_UNITS 16
343 struct r600_seamless_cube_map
{
344 struct r600_atom atom
;
348 struct r600_samplerview_state
{
349 struct r600_atom atom
;
350 struct r600_pipe_sampler_view
*views
[NUM_TEX_UNITS
];
351 uint32_t enabled_mask
;
353 uint32_t compressed_depthtex_mask
; /* which textures are depth */
354 uint32_t compressed_colortex_mask
;
355 boolean dirty_buffer_constants
;
358 struct r600_sampler_states
{
359 struct r600_atom atom
;
360 struct r600_pipe_sampler_state
*states
[NUM_TEX_UNITS
];
361 uint32_t enabled_mask
;
363 uint32_t has_bordercolor_mask
; /* which states contain the border color */
366 struct r600_textures_info
{
367 struct r600_samplerview_state views
;
368 struct r600_sampler_states states
;
369 bool is_array_sampler
[NUM_TEX_UNITS
];
372 struct r600_shader_driver_constants_info
{
373 /* currently 128 bytes for UCP/samplepos + sampler buffer constants */
377 bool texture_const_dirty
;
378 bool ps_sample_pos_dirty
;
381 struct r600_constbuf_state
383 struct r600_atom atom
;
384 struct pipe_constant_buffer cb
[PIPE_MAX_CONSTANT_BUFFERS
];
385 uint32_t enabled_mask
;
389 struct r600_vertexbuf_state
391 struct r600_atom atom
;
392 struct pipe_vertex_buffer vb
[PIPE_MAX_ATTRIBS
];
393 uint32_t enabled_mask
; /* non-NULL buffers */
397 /* CSO (constant state object, in other words, immutable state). */
398 struct r600_cso_state
400 struct r600_atom atom
;
401 void *cso
; /* e.g. r600_blend_state */
402 struct r600_command_buffer
*cb
;
405 struct r600_fetch_shader
{
406 struct r600_resource
*buffer
;
410 struct r600_shader_state
{
411 struct r600_atom atom
;
412 struct r600_pipe_shader
*shader
;
415 struct r600_context
{
416 struct r600_common_context b
;
417 struct r600_screen
*screen
;
418 struct blitter_context
*blitter
;
419 struct u_suballocator
*allocator_fetch_shader
;
422 boolean has_vertex_cache
;
423 unsigned default_gprs
[EG_NUM_HW_STAGES
];
424 unsigned current_gprs
[EG_NUM_HW_STAGES
];
425 unsigned r6xx_num_clause_temp_gprs
;
427 /* Miscellaneous state objects. */
428 void *custom_dsa_flush
;
429 void *custom_blend_resolve
;
430 void *custom_blend_decompress
;
431 void *custom_blend_fastclear
;
432 /* With rasterizer discard, there doesn't have to be a pixel shader.
433 * In that case, we bind this one: */
434 void *dummy_pixel_shader
;
435 /* These dummy CMASK and FMASK buffers are used to get around the R6xx hardware
436 * bug where valid CMASK and FMASK are required to be present to avoid
437 * a hardlock in certain operations but aren't actually used
438 * for anything useful. */
439 struct r600_resource
*dummy_fmask
;
440 struct r600_resource
*dummy_cmask
;
442 /* State binding slots are here. */
443 struct r600_atom
*atoms
[R600_NUM_ATOMS
];
444 /* Dirty atom bitmask for fast tests */
445 uint64_t dirty_atoms
;
446 /* States for CS initialization. */
447 struct r600_command_buffer start_cs_cmd
; /* invariant state mostly */
448 /** Compute specific registers initializations. The start_cs_cmd atom
449 * must be emitted before start_compute_cs_cmd. */
450 struct r600_command_buffer start_compute_cs_cmd
;
451 /* Register states. */
452 struct r600_alphatest_state alphatest_state
;
453 struct r600_cso_state blend_state
;
454 struct r600_blend_color blend_color
;
455 struct r600_cb_misc_state cb_misc_state
;
456 struct r600_clip_misc_state clip_misc_state
;
457 struct r600_clip_state clip_state
;
458 struct r600_db_misc_state db_misc_state
;
459 struct r600_db_state db_state
;
460 struct r600_cso_state dsa_state
;
461 struct r600_framebuffer framebuffer
;
462 struct r600_poly_offset_state poly_offset_state
;
463 struct r600_cso_state rasterizer_state
;
464 struct r600_sample_mask sample_mask
;
465 struct r600_seamless_cube_map seamless_cube_map
;
466 struct r600_config_state config_state
;
467 struct r600_stencil_ref_state stencil_ref
;
468 struct r600_vgt_state vgt_state
;
469 /* Shaders and shader resources. */
470 struct r600_cso_state vertex_fetch_shader
;
471 struct r600_shader_state hw_shader_stages
[EG_NUM_HW_STAGES
];
472 struct r600_cs_shader_state cs_shader_state
;
473 struct r600_shader_stages_state shader_stages
;
474 struct r600_gs_rings_state gs_rings
;
475 struct r600_constbuf_state constbuf_state
[PIPE_SHADER_TYPES
];
476 struct r600_textures_info samplers
[PIPE_SHADER_TYPES
];
478 struct r600_shader_driver_constants_info driver_consts
[PIPE_SHADER_TYPES
];
480 /** Vertex buffers for fetch shaders */
481 struct r600_vertexbuf_state vertex_buffer_state
;
482 /** Vertex buffers for compute shaders */
483 struct r600_vertexbuf_state cs_vertex_buffer_state
;
485 /* Additional context states. */
486 unsigned compute_cb_target_mask
;
487 struct r600_pipe_shader_selector
*ps_shader
;
488 struct r600_pipe_shader_selector
*vs_shader
;
489 struct r600_pipe_shader_selector
*gs_shader
;
491 struct r600_pipe_shader_selector
*tcs_shader
;
492 struct r600_pipe_shader_selector
*tes_shader
;
494 struct r600_pipe_shader_selector
*fixed_func_tcs_shader
;
496 struct r600_rasterizer_state
*rasterizer
;
498 bool force_blend_disable
;
499 boolean dual_src_blend
;
504 struct pipe_index_buffer index_buffer
;
506 /* Last draw state (-1 = unset). */
507 int last_primitive_type
; /* Last primitive type used in draw_vbo. */
508 int last_start_instance
;
511 struct r600_isa
*isa
;
512 float sample_positions
[4 * 16];
514 bool tess_state_dirty
;
515 struct r600_pipe_shader_selector
*last_ls
;
516 struct r600_pipe_shader_selector
*last_tcs
;
517 unsigned last_num_tcs_input_cp
;
521 static inline void r600_emit_command_buffer(struct radeon_winsys_cs
*cs
,
522 struct r600_command_buffer
*cb
)
524 assert(cs
->current
.cdw
+ cb
->num_dw
<= cs
->current
.max_dw
);
525 memcpy(cs
->current
.buf
+ cs
->current
.cdw
, cb
->buf
, 4 * cb
->num_dw
);
526 cs
->current
.cdw
+= cb
->num_dw
;
529 static inline void r600_set_atom_dirty(struct r600_context
*rctx
,
530 struct r600_atom
*atom
,
535 assert(atom
->id
!= 0);
536 assert(atom
->id
< sizeof(mask
) * 8);
537 mask
= 1ull << atom
->id
;
539 rctx
->dirty_atoms
|= mask
;
541 rctx
->dirty_atoms
&= ~mask
;
544 static inline void r600_mark_atom_dirty(struct r600_context
*rctx
,
545 struct r600_atom
*atom
)
547 r600_set_atom_dirty(rctx
, atom
, true);
550 static inline void r600_emit_atom(struct r600_context
*rctx
, struct r600_atom
*atom
)
552 atom
->emit(&rctx
->b
, atom
);
553 r600_set_atom_dirty(rctx
, atom
, false);
556 static inline void r600_set_cso_state(struct r600_context
*rctx
,
557 struct r600_cso_state
*state
, void *cso
)
560 r600_set_atom_dirty(rctx
, &state
->atom
, cso
!= NULL
);
563 static inline void r600_set_cso_state_with_cb(struct r600_context
*rctx
,
564 struct r600_cso_state
*state
, void *cso
,
565 struct r600_command_buffer
*cb
)
568 state
->atom
.num_dw
= cb
? cb
->num_dw
: 0;
569 r600_set_cso_state(rctx
, state
, cso
);
572 /* compute_memory_pool.c */
573 struct compute_memory_pool
;
574 void compute_memory_pool_delete(struct compute_memory_pool
* pool
);
575 struct compute_memory_pool
* compute_memory_pool_new(
576 struct r600_screen
*rscreen
);
578 /* evergreen_state.c */
579 struct pipe_sampler_view
*
580 evergreen_create_sampler_view_custom(struct pipe_context
*ctx
,
581 struct pipe_resource
*texture
,
582 const struct pipe_sampler_view
*state
,
583 unsigned width0
, unsigned height0
,
584 unsigned force_level
);
585 void evergreen_init_common_regs(struct r600_context
*ctx
,
586 struct r600_command_buffer
*cb
,
587 enum chip_class ctx_chip_class
,
588 enum radeon_family ctx_family
,
590 void cayman_init_common_regs(struct r600_command_buffer
*cb
,
591 enum chip_class ctx_chip_class
,
592 enum radeon_family ctx_family
,
595 void evergreen_init_state_functions(struct r600_context
*rctx
);
596 void evergreen_init_atom_start_cs(struct r600_context
*rctx
);
597 void evergreen_update_ps_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
598 void evergreen_update_es_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
599 void evergreen_update_gs_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
600 void evergreen_update_vs_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
601 void evergreen_update_ls_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
602 void evergreen_update_hs_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
603 void *evergreen_create_db_flush_dsa(struct r600_context
*rctx
);
604 void *evergreen_create_resolve_blend(struct r600_context
*rctx
);
605 void *evergreen_create_decompress_blend(struct r600_context
*rctx
);
606 void *evergreen_create_fastclear_blend(struct r600_context
*rctx
);
607 boolean
evergreen_is_format_supported(struct pipe_screen
*screen
,
608 enum pipe_format format
,
609 enum pipe_texture_target target
,
610 unsigned sample_count
,
612 void evergreen_init_color_surface(struct r600_context
*rctx
,
613 struct r600_surface
*surf
);
614 void evergreen_init_color_surface_rat(struct r600_context
*rctx
,
615 struct r600_surface
*surf
);
616 void evergreen_update_db_shader_control(struct r600_context
* rctx
);
617 bool evergreen_adjust_gprs(struct r600_context
*rctx
);
619 void r600_init_blit_functions(struct r600_context
*rctx
);
620 void r600_decompress_depth_textures(struct r600_context
*rctx
,
621 struct r600_samplerview_state
*textures
);
622 void r600_decompress_color_textures(struct r600_context
*rctx
,
623 struct r600_samplerview_state
*textures
);
624 void r600_resource_copy_region(struct pipe_context
*ctx
,
625 struct pipe_resource
*dst
,
627 unsigned dstx
, unsigned dsty
, unsigned dstz
,
628 struct pipe_resource
*src
,
630 const struct pipe_box
*src_box
);
633 int r600_pipe_shader_create(struct pipe_context
*ctx
,
634 struct r600_pipe_shader
*shader
,
635 union r600_shader_key key
);
637 void r600_pipe_shader_destroy(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
640 struct pipe_sampler_view
*
641 r600_create_sampler_view_custom(struct pipe_context
*ctx
,
642 struct pipe_resource
*texture
,
643 const struct pipe_sampler_view
*state
,
644 unsigned width_first_level
, unsigned height_first_level
);
645 void r600_init_state_functions(struct r600_context
*rctx
);
646 void r600_init_atom_start_cs(struct r600_context
*rctx
);
647 void r600_update_ps_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
648 void r600_update_es_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
649 void r600_update_gs_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
650 void r600_update_vs_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
651 void *r600_create_db_flush_dsa(struct r600_context
*rctx
);
652 void *r600_create_resolve_blend(struct r600_context
*rctx
);
653 void *r700_create_resolve_blend(struct r600_context
*rctx
);
654 void *r600_create_decompress_blend(struct r600_context
*rctx
);
655 bool r600_adjust_gprs(struct r600_context
*rctx
);
656 boolean
r600_is_format_supported(struct pipe_screen
*screen
,
657 enum pipe_format format
,
658 enum pipe_texture_target target
,
659 unsigned sample_count
,
661 void r600_update_db_shader_control(struct r600_context
* rctx
);
663 /* r600_hw_context.c */
664 void r600_context_gfx_flush(void *context
, unsigned flags
,
665 struct pipe_fence_handle
**fence
);
666 void r600_begin_new_cs(struct r600_context
*ctx
);
667 void r600_flush_emit(struct r600_context
*ctx
);
668 void r600_need_cs_space(struct r600_context
*ctx
, unsigned num_dw
, boolean count_draw_in
);
669 void r600_emit_pfp_sync_me(struct r600_context
*rctx
);
670 void r600_cp_dma_copy_buffer(struct r600_context
*rctx
,
671 struct pipe_resource
*dst
, uint64_t dst_offset
,
672 struct pipe_resource
*src
, uint64_t src_offset
,
674 void evergreen_cp_dma_clear_buffer(struct r600_context
*rctx
,
675 struct pipe_resource
*dst
, uint64_t offset
,
676 unsigned size
, uint32_t clear_value
,
677 enum r600_coherency coher
);
678 void r600_dma_copy_buffer(struct r600_context
*rctx
,
679 struct pipe_resource
*dst
,
680 struct pipe_resource
*src
,
686 * evergreen_hw_context.c
688 void evergreen_dma_copy_buffer(struct r600_context
*rctx
,
689 struct pipe_resource
*dst
,
690 struct pipe_resource
*src
,
694 void evergreen_setup_tess_constants(struct r600_context
*rctx
,
695 const struct pipe_draw_info
*info
,
696 unsigned *num_patches
);
697 uint32_t evergreen_get_ls_hs_config(struct r600_context
*rctx
,
698 const struct pipe_draw_info
*info
,
699 unsigned num_patches
);
700 void evergreen_set_ls_hs_config(struct r600_context
*rctx
,
701 struct radeon_winsys_cs
*cs
,
702 uint32_t ls_hs_config
);
703 void evergreen_set_lds_alloc(struct r600_context
*rctx
,
704 struct radeon_winsys_cs
*cs
,
707 /* r600_state_common.c */
708 void r600_init_common_state_functions(struct r600_context
*rctx
);
709 void r600_emit_cso_state(struct r600_context
*rctx
, struct r600_atom
*atom
);
710 void r600_emit_alphatest_state(struct r600_context
*rctx
, struct r600_atom
*atom
);
711 void r600_emit_blend_color(struct r600_context
*rctx
, struct r600_atom
*atom
);
712 void r600_emit_vgt_state(struct r600_context
*rctx
, struct r600_atom
*atom
);
713 void r600_emit_clip_misc_state(struct r600_context
*rctx
, struct r600_atom
*atom
);
714 void r600_emit_stencil_ref(struct r600_context
*rctx
, struct r600_atom
*atom
);
715 void r600_emit_shader(struct r600_context
*rctx
, struct r600_atom
*a
);
716 void r600_add_atom(struct r600_context
*rctx
, struct r600_atom
*atom
, unsigned id
);
717 void r600_init_atom(struct r600_context
*rctx
, struct r600_atom
*atom
, unsigned id
,
718 void (*emit
)(struct r600_context
*ctx
, struct r600_atom
*state
),
720 void r600_vertex_buffers_dirty(struct r600_context
*rctx
);
721 void r600_sampler_views_dirty(struct r600_context
*rctx
,
722 struct r600_samplerview_state
*state
);
723 void r600_sampler_states_dirty(struct r600_context
*rctx
,
724 struct r600_sampler_states
*state
);
725 void r600_constant_buffers_dirty(struct r600_context
*rctx
, struct r600_constbuf_state
*state
);
726 void r600_set_sample_locations_constant_buffer(struct r600_context
*rctx
);
727 uint32_t r600_translate_stencil_op(int s_op
);
728 uint32_t r600_translate_fill(uint32_t func
);
729 unsigned r600_tex_wrap(unsigned wrap
);
730 unsigned r600_tex_mipfilter(unsigned filter
);
731 unsigned r600_tex_compare(unsigned compare
);
732 bool sampler_state_needs_border_color(const struct pipe_sampler_state
*state
);
733 struct pipe_surface
*r600_create_surface_custom(struct pipe_context
*pipe
,
734 struct pipe_resource
*texture
,
735 const struct pipe_surface
*templ
,
736 unsigned width
, unsigned height
);
737 unsigned r600_get_swizzle_combined(const unsigned char *swizzle_format
,
738 const unsigned char *swizzle_view
,
740 uint32_t r600_translate_texformat(struct pipe_screen
*screen
, enum pipe_format format
,
741 const unsigned char *swizzle_view
,
742 uint32_t *word4_p
, uint32_t *yuv_format_p
,
743 bool do_endian_swap
);
744 uint32_t r600_translate_colorformat(enum chip_class chip
, enum pipe_format format
,
745 bool do_endian_swap
);
746 uint32_t r600_colorformat_endian_swap(uint32_t colorformat
, bool do_endian_swap
);
749 struct pipe_video_codec
*r600_uvd_create_decoder(struct pipe_context
*context
,
750 const struct pipe_video_codec
*decoder
);
752 struct pipe_video_buffer
*r600_video_buffer_create(struct pipe_context
*pipe
,
753 const struct pipe_video_buffer
*tmpl
);
756 * Helpers for building command buffers
759 #define PKT3_SET_CONFIG_REG 0x68
760 #define PKT3_SET_CONTEXT_REG 0x69
761 #define PKT3_SET_CTL_CONST 0x6F
762 #define PKT3_SET_LOOP_CONST 0x6C
764 #define R600_CONFIG_REG_OFFSET 0x08000
765 #define R600_CONTEXT_REG_OFFSET 0x28000
766 #define R600_CTL_CONST_OFFSET 0x3CFF0
767 #define R600_LOOP_CONST_OFFSET 0X0003E200
768 #define EG_LOOP_CONST_OFFSET 0x0003A200
770 #define PKT_TYPE_S(x) (((unsigned)(x) & 0x3) << 30)
771 #define PKT_COUNT_S(x) (((unsigned)(x) & 0x3FFF) << 16)
772 #define PKT3_IT_OPCODE_S(x) (((unsigned)(x) & 0xFF) << 8)
773 #define PKT3_PREDICATE(x) (((x) >> 0) & 0x1)
774 #define PKT3(op, count, predicate) (PKT_TYPE_S(3) | PKT_COUNT_S(count) | PKT3_IT_OPCODE_S(op) | PKT3_PREDICATE(predicate))
776 #define RADEON_CP_PACKET3_COMPUTE_MODE 0x00000002
778 /*Evergreen Compute packet3*/
779 #define PKT3C(op, count, predicate) (PKT_TYPE_S(3) | PKT3_IT_OPCODE_S(op) | PKT_COUNT_S(count) | PKT3_PREDICATE(predicate) | RADEON_CP_PACKET3_COMPUTE_MODE)
781 static inline void r600_store_value(struct r600_command_buffer
*cb
, unsigned value
)
783 cb
->buf
[cb
->num_dw
++] = value
;
786 static inline void r600_store_array(struct r600_command_buffer
*cb
, unsigned num
, unsigned *ptr
)
788 assert(cb
->num_dw
+num
<= cb
->max_num_dw
);
789 memcpy(&cb
->buf
[cb
->num_dw
], ptr
, num
* sizeof(ptr
[0]));
793 static inline void r600_store_config_reg_seq(struct r600_command_buffer
*cb
, unsigned reg
, unsigned num
)
795 assert(reg
< R600_CONTEXT_REG_OFFSET
);
796 assert(cb
->num_dw
+2+num
<= cb
->max_num_dw
);
797 cb
->buf
[cb
->num_dw
++] = PKT3(PKT3_SET_CONFIG_REG
, num
, 0);
798 cb
->buf
[cb
->num_dw
++] = (reg
- R600_CONFIG_REG_OFFSET
) >> 2;
802 * Needs cb->pkt_flags set to RADEON_CP_PACKET3_COMPUTE_MODE for compute
805 static inline void r600_store_context_reg_seq(struct r600_command_buffer
*cb
, unsigned reg
, unsigned num
)
807 assert(reg
>= R600_CONTEXT_REG_OFFSET
&& reg
< R600_CTL_CONST_OFFSET
);
808 assert(cb
->num_dw
+2+num
<= cb
->max_num_dw
);
809 cb
->buf
[cb
->num_dw
++] = PKT3(PKT3_SET_CONTEXT_REG
, num
, 0) | cb
->pkt_flags
;
810 cb
->buf
[cb
->num_dw
++] = (reg
- R600_CONTEXT_REG_OFFSET
) >> 2;
814 * Needs cb->pkt_flags set to RADEON_CP_PACKET3_COMPUTE_MODE for compute
817 static inline void r600_store_ctl_const_seq(struct r600_command_buffer
*cb
, unsigned reg
, unsigned num
)
819 assert(reg
>= R600_CTL_CONST_OFFSET
);
820 assert(cb
->num_dw
+2+num
<= cb
->max_num_dw
);
821 cb
->buf
[cb
->num_dw
++] = PKT3(PKT3_SET_CTL_CONST
, num
, 0) | cb
->pkt_flags
;
822 cb
->buf
[cb
->num_dw
++] = (reg
- R600_CTL_CONST_OFFSET
) >> 2;
825 static inline void r600_store_loop_const_seq(struct r600_command_buffer
*cb
, unsigned reg
, unsigned num
)
827 assert(reg
>= R600_LOOP_CONST_OFFSET
);
828 assert(cb
->num_dw
+2+num
<= cb
->max_num_dw
);
829 cb
->buf
[cb
->num_dw
++] = PKT3(PKT3_SET_LOOP_CONST
, num
, 0);
830 cb
->buf
[cb
->num_dw
++] = (reg
- R600_LOOP_CONST_OFFSET
) >> 2;
834 * Needs cb->pkt_flags set to RADEON_CP_PACKET3_COMPUTE_MODE for compute
837 static inline void eg_store_loop_const_seq(struct r600_command_buffer
*cb
, unsigned reg
, unsigned num
)
839 assert(reg
>= EG_LOOP_CONST_OFFSET
);
840 assert(cb
->num_dw
+2+num
<= cb
->max_num_dw
);
841 cb
->buf
[cb
->num_dw
++] = PKT3(PKT3_SET_LOOP_CONST
, num
, 0) | cb
->pkt_flags
;
842 cb
->buf
[cb
->num_dw
++] = (reg
- EG_LOOP_CONST_OFFSET
) >> 2;
845 static inline void r600_store_config_reg(struct r600_command_buffer
*cb
, unsigned reg
, unsigned value
)
847 r600_store_config_reg_seq(cb
, reg
, 1);
848 r600_store_value(cb
, value
);
851 static inline void r600_store_context_reg(struct r600_command_buffer
*cb
, unsigned reg
, unsigned value
)
853 r600_store_context_reg_seq(cb
, reg
, 1);
854 r600_store_value(cb
, value
);
857 static inline void r600_store_ctl_const(struct r600_command_buffer
*cb
, unsigned reg
, unsigned value
)
859 r600_store_ctl_const_seq(cb
, reg
, 1);
860 r600_store_value(cb
, value
);
863 static inline void r600_store_loop_const(struct r600_command_buffer
*cb
, unsigned reg
, unsigned value
)
865 r600_store_loop_const_seq(cb
, reg
, 1);
866 r600_store_value(cb
, value
);
869 static inline void eg_store_loop_const(struct r600_command_buffer
*cb
, unsigned reg
, unsigned value
)
871 eg_store_loop_const_seq(cb
, reg
, 1);
872 r600_store_value(cb
, value
);
875 void r600_init_command_buffer(struct r600_command_buffer
*cb
, unsigned num_dw
);
876 void r600_release_command_buffer(struct r600_command_buffer
*cb
);
878 static inline void radeon_compute_set_context_reg_seq(struct radeon_winsys_cs
*cs
, unsigned reg
, unsigned num
)
880 radeon_set_context_reg_seq(cs
, reg
, num
);
881 /* Set the compute bit on the packet header */
882 cs
->current
.buf
[cs
->current
.cdw
- 2] |= RADEON_CP_PACKET3_COMPUTE_MODE
;
885 static inline void radeon_set_ctl_const_seq(struct radeon_winsys_cs
*cs
, unsigned reg
, unsigned num
)
887 assert(reg
>= R600_CTL_CONST_OFFSET
);
888 assert(cs
->current
.cdw
+ 2 + num
<= cs
->current
.max_dw
);
889 radeon_emit(cs
, PKT3(PKT3_SET_CTL_CONST
, num
, 0));
890 radeon_emit(cs
, (reg
- R600_CTL_CONST_OFFSET
) >> 2);
893 static inline void radeon_compute_set_context_reg(struct radeon_winsys_cs
*cs
, unsigned reg
, unsigned value
)
895 radeon_compute_set_context_reg_seq(cs
, reg
, 1);
896 radeon_emit(cs
, value
);
899 static inline void radeon_set_context_reg_flag(struct radeon_winsys_cs
*cs
, unsigned reg
, unsigned value
, unsigned flag
)
901 if (flag
& RADEON_CP_PACKET3_COMPUTE_MODE
) {
902 radeon_compute_set_context_reg(cs
, reg
, value
);
904 radeon_set_context_reg(cs
, reg
, value
);
908 static inline void radeon_set_ctl_const(struct radeon_winsys_cs
*cs
, unsigned reg
, unsigned value
)
910 radeon_set_ctl_const_seq(cs
, reg
, 1);
911 radeon_emit(cs
, value
);
917 static inline uint32_t S_FIXED(float value
, uint32_t frac_bits
)
919 return value
* (1 << frac_bits
);
922 /* 12.4 fixed-point */
923 static inline unsigned r600_pack_float_12p4(float x
)
926 x
>= 4096 ? 0xffff : x
* 16;
929 /* Return if the depth format can be read without the DB->CB copy on r6xx-r7xx. */
930 static inline bool r600_can_read_depth(struct r600_texture
*rtex
)
932 return rtex
->resource
.b
.b
.nr_samples
<= 1 &&
933 (rtex
->resource
.b
.b
.format
== PIPE_FORMAT_Z16_UNORM
||
934 rtex
->resource
.b
.b
.format
== PIPE_FORMAT_Z32_FLOAT
);
937 static inline unsigned r600_get_flush_flags(enum r600_coherency coher
)
941 case R600_COHERENCY_NONE
:
943 case R600_COHERENCY_SHADER
:
944 return R600_CONTEXT_INV_CONST_CACHE
|
945 R600_CONTEXT_INV_VERTEX_CACHE
|
946 R600_CONTEXT_INV_TEX_CACHE
|
947 R600_CONTEXT_STREAMOUT_FLUSH
;
948 case R600_COHERENCY_CB_META
:
949 return R600_CONTEXT_FLUSH_AND_INV_CB
|
950 R600_CONTEXT_FLUSH_AND_INV_CB_META
;
954 #define V_028A6C_OUTPRIM_TYPE_POINTLIST 0
955 #define V_028A6C_OUTPRIM_TYPE_LINESTRIP 1
956 #define V_028A6C_OUTPRIM_TYPE_TRISTRIP 2
958 unsigned r600_conv_prim_to_gs_out(unsigned mode
);