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"
37 #include "tgsi/tgsi_scan.h"
39 #define R600_NUM_ATOMS 52
42 #define R600_CONTEXT_INV_VERTEX_CACHE (R600_CONTEXT_PRIVATE_FLAG << 0)
43 #define R600_CONTEXT_INV_TEX_CACHE (R600_CONTEXT_PRIVATE_FLAG << 1)
44 #define R600_CONTEXT_INV_CONST_CACHE (R600_CONTEXT_PRIVATE_FLAG << 2)
45 /* read-write caches */
46 #define R600_CONTEXT_FLUSH_AND_INV (R600_CONTEXT_PRIVATE_FLAG << 3)
47 #define R600_CONTEXT_FLUSH_AND_INV_CB_META (R600_CONTEXT_PRIVATE_FLAG << 4)
48 #define R600_CONTEXT_FLUSH_AND_INV_DB_META (R600_CONTEXT_PRIVATE_FLAG << 5)
49 #define R600_CONTEXT_FLUSH_AND_INV_DB (R600_CONTEXT_PRIVATE_FLAG << 6)
50 #define R600_CONTEXT_FLUSH_AND_INV_CB (R600_CONTEXT_PRIVATE_FLAG << 7)
51 /* engine synchronization */
52 #define R600_CONTEXT_PS_PARTIAL_FLUSH (R600_CONTEXT_PRIVATE_FLAG << 8)
53 #define R600_CONTEXT_WAIT_3D_IDLE (R600_CONTEXT_PRIVATE_FLAG << 9)
54 #define R600_CONTEXT_WAIT_CP_DMA_IDLE (R600_CONTEXT_PRIVATE_FLAG << 10)
56 /* the number of CS dwords for flushing and drawing */
57 #define R600_MAX_FLUSH_CS_DWORDS 18
58 #define R600_MAX_DRAW_CS_DWORDS 58
60 #define R600_MAX_USER_CONST_BUFFERS 13
61 #define R600_MAX_DRIVER_CONST_BUFFERS 3
62 #define R600_MAX_CONST_BUFFERS (R600_MAX_USER_CONST_BUFFERS + R600_MAX_DRIVER_CONST_BUFFERS)
64 /* start driver buffers after user buffers */
65 #define R600_BUFFER_INFO_CONST_BUFFER (R600_MAX_USER_CONST_BUFFERS)
66 #define R600_UCP_SIZE (4*4*8)
67 #define R600_BUFFER_INFO_OFFSET (R600_UCP_SIZE)
69 #define R600_LDS_INFO_CONST_BUFFER (R600_MAX_USER_CONST_BUFFERS + 1)
71 * Note GS doesn't use a constant buffer binding, just a resource index,
72 * so it's fine to have it exist at index 16.
74 #define R600_GS_RING_CONST_BUFFER (R600_MAX_USER_CONST_BUFFERS + 2)
75 /* Currently R600_MAX_CONST_BUFFERS just fits on the hw, which has a limit
76 * of 16 const buffers.
77 * UCP/SAMPLE_POSITIONS are never accessed by same shader stage so they can use the same id.
79 * In order to support d3d 11 mandated minimum of 15 user const buffers
80 * we'd have to squash all use cases into one driver buffer.
82 #define R600_MAX_CONST_BUFFER_SIZE (4096 * sizeof(float[4]))
85 #define R600_HW_STAGE_PS 0
86 #define R600_HW_STAGE_VS 1
87 #define R600_HW_STAGE_GS 2
88 #define R600_HW_STAGE_ES 3
89 #define EG_HW_STAGE_LS 4
90 #define EG_HW_STAGE_HS 5
92 #define R600_NUM_HW_STAGES 4
93 #define EG_NUM_HW_STAGES 6
95 #ifdef PIPE_ARCH_BIG_ENDIAN
96 #define R600_BIG_ENDIAN 1
98 #define R600_BIG_ENDIAN 0
102 struct r600_bytecode
;
103 union r600_shader_key
;
105 /* This is an atom containing GPU commands that never change.
106 * This is supposed to be copied directly into the CS. */
107 struct r600_command_buffer
{
114 struct r600_db_state
{
115 struct r600_atom atom
;
116 struct r600_surface
*rsurf
;
119 struct r600_db_misc_state
{
120 struct r600_atom atom
;
121 bool occlusion_queries_disabled
;
122 bool flush_depthstencil_through_cb
;
123 bool flush_depth_inplace
;
124 bool flush_stencil_inplace
;
125 bool copy_depth
, copy_stencil
;
126 unsigned copy_sample
;
127 unsigned log_samples
;
128 unsigned db_shader_control
;
130 uint8_t ps_conservative_z
;
133 struct r600_cb_misc_state
{
134 struct r600_atom atom
;
135 unsigned cb_color_control
; /* this comes from blend state */
136 unsigned blend_colormask
; /* 8*4 bits for 8 RGBA colorbuffers */
138 unsigned nr_ps_color_outputs
;
143 struct r600_clip_misc_state
{
144 struct r600_atom atom
;
145 unsigned pa_cl_clip_cntl
; /* from rasterizer */
146 unsigned pa_cl_vs_out_cntl
; /* from vertex shader */
147 unsigned clip_plane_enable
; /* from rasterizer */
148 unsigned clip_dist_write
; /* from vertex shader */
149 boolean clip_disable
; /* from vertex shader */
150 boolean vs_out_viewport
; /* from vertex shader */
153 struct r600_alphatest_state
{
154 struct r600_atom atom
;
155 unsigned sx_alpha_test_control
; /* this comes from dsa state */
156 unsigned sx_alpha_ref
; /* this comes from dsa state */
158 bool cb0_export_16bpc
; /* from set_framebuffer_state */
161 struct r600_vgt_state
{
162 struct r600_atom atom
;
163 uint32_t vgt_multi_prim_ib_reset_en
;
164 uint32_t vgt_multi_prim_ib_reset_indx
;
165 uint32_t vgt_indx_offset
;
166 bool last_draw_was_indirect
;
169 struct r600_blend_color
{
170 struct r600_atom atom
;
171 struct pipe_blend_color state
;
174 struct r600_clip_state
{
175 struct r600_atom atom
;
176 struct pipe_clip_state state
;
179 struct r600_cs_shader_state
{
180 struct r600_atom atom
;
181 unsigned kernel_index
;
183 struct r600_pipe_compute
*shader
;
186 struct r600_framebuffer
{
187 struct r600_atom atom
;
188 struct pipe_framebuffer_state state
;
189 unsigned compressed_cb_mask
;
193 bool is_msaa_resolve
;
196 struct r600_sample_mask
{
197 struct r600_atom atom
;
198 uint16_t sample_mask
; /* there are only 8 bits on EG, 16 bits on Cayman */
201 struct r600_config_state
{
202 struct r600_atom atom
;
203 unsigned sq_gpr_resource_mgmt_1
;
204 unsigned sq_gpr_resource_mgmt_2
;
205 unsigned sq_gpr_resource_mgmt_3
;
206 bool dyn_gpr_enabled
;
209 struct r600_stencil_ref
216 struct r600_stencil_ref_state
{
217 struct r600_atom atom
;
218 struct r600_stencil_ref state
;
219 struct pipe_stencil_ref pipe_state
;
222 struct r600_shader_stages_state
{
223 struct r600_atom atom
;
224 unsigned geom_enable
;
227 struct r600_gs_rings_state
{
228 struct r600_atom atom
;
230 struct pipe_constant_buffer esgs_ring
;
231 struct pipe_constant_buffer gsvs_ring
;
234 /* This must start from 16. */
236 #define DBG_NO_CP_DMA (1 << 30)
238 #define DBG_NO_SB (1 << 21)
239 #define DBG_SB_CS (1 << 22)
240 #define DBG_SB_DRY_RUN (1 << 23)
241 #define DBG_SB_STAT (1 << 24)
242 #define DBG_SB_DUMP (1 << 25)
243 #define DBG_SB_NO_FALLBACK (1 << 26)
244 #define DBG_SB_DISASM (1 << 27)
245 #define DBG_SB_SAFEMATH (1 << 28)
248 struct r600_common_screen b
;
250 bool has_compressed_msaa_texturing
;
252 /*for compute global memory binding, we allocate stuff here, instead of
254 * XXX: Not sure if this is the best place for global_pool. Also,
255 * it's not thread safe, so it won't work with multiple contexts. */
256 struct compute_memory_pool
*global_pool
;
259 struct r600_pipe_sampler_view
{
260 struct pipe_sampler_view base
;
261 struct list_head list
;
262 struct r600_resource
*tex_resource
;
263 uint32_t tex_resource_words
[8];
264 bool skip_mip_address_reloc
;
265 bool is_stencil_sampler
;
268 struct r600_rasterizer_state
{
269 struct r600_command_buffer buffer
;
272 unsigned sprite_coord_enable
;
273 unsigned clip_plane_enable
;
274 unsigned pa_sc_line_stipple
;
275 unsigned pa_cl_clip_cntl
;
276 unsigned pa_su_sc_mode_cntl
;
281 bool multisample_enable
;
284 struct r600_poly_offset_state
{
285 struct r600_atom atom
;
286 enum pipe_format zs_format
;
291 struct r600_blend_state
{
292 struct r600_command_buffer buffer
;
293 struct r600_command_buffer buffer_no_blend
;
294 unsigned cb_target_mask
;
295 unsigned cb_color_control
;
296 unsigned cb_color_control_no_blend
;
301 struct r600_dsa_state
{
302 struct r600_command_buffer buffer
;
307 unsigned sx_alpha_test_control
;
310 struct r600_pipe_shader
;
312 struct r600_pipe_shader_selector
{
313 struct r600_pipe_shader
*current
;
315 struct tgsi_token
*tokens
;
316 struct pipe_stream_output_info so
;
317 struct tgsi_shader_info info
;
319 unsigned num_shaders
;
321 /* PIPE_SHADER_[VERTEX|FRAGMENT|...] */
324 /* geometry shader properties */
325 unsigned gs_output_prim
;
326 unsigned gs_max_out_vertices
;
327 unsigned gs_num_invocations
;
330 uint64_t lds_patch_outputs_written_mask
;
331 uint64_t lds_outputs_written_mask
;
332 unsigned nr_ps_max_color_exports
;
335 struct r600_pipe_sampler_state
{
336 uint32_t tex_sampler_words
[3];
337 union pipe_color_union border_color
;
338 bool border_color_use
;
339 bool seamless_cube_map
;
342 /* needed for blitter save */
343 #define NUM_TEX_UNITS 16
345 struct r600_seamless_cube_map
{
346 struct r600_atom atom
;
350 struct r600_samplerview_state
{
351 struct r600_atom atom
;
352 struct r600_pipe_sampler_view
*views
[NUM_TEX_UNITS
];
353 uint32_t enabled_mask
;
355 uint32_t compressed_depthtex_mask
; /* which textures are depth */
356 uint32_t compressed_colortex_mask
;
357 boolean dirty_buffer_constants
;
360 struct r600_sampler_states
{
361 struct r600_atom atom
;
362 struct r600_pipe_sampler_state
*states
[NUM_TEX_UNITS
];
363 uint32_t enabled_mask
;
365 uint32_t has_bordercolor_mask
; /* which states contain the border color */
368 struct r600_textures_info
{
369 struct r600_samplerview_state views
;
370 struct r600_sampler_states states
;
371 bool is_array_sampler
[NUM_TEX_UNITS
];
374 struct r600_shader_driver_constants_info
{
375 /* currently 128 bytes for UCP/samplepos + sampler buffer constants */
379 bool texture_const_dirty
;
380 bool ps_sample_pos_dirty
;
383 struct r600_constbuf_state
385 struct r600_atom atom
;
386 struct pipe_constant_buffer cb
[PIPE_MAX_CONSTANT_BUFFERS
];
387 uint32_t enabled_mask
;
391 struct r600_vertexbuf_state
393 struct r600_atom atom
;
394 struct pipe_vertex_buffer vb
[PIPE_MAX_ATTRIBS
];
395 uint32_t enabled_mask
; /* non-NULL buffers */
399 /* CSO (constant state object, in other words, immutable state). */
400 struct r600_cso_state
402 struct r600_atom atom
;
403 void *cso
; /* e.g. r600_blend_state */
404 struct r600_command_buffer
*cb
;
407 struct r600_fetch_shader
{
408 struct r600_resource
*buffer
;
412 struct r600_shader_state
{
413 struct r600_atom atom
;
414 struct r600_pipe_shader
*shader
;
417 struct r600_context
{
418 struct r600_common_context b
;
419 struct r600_screen
*screen
;
420 struct blitter_context
*blitter
;
421 struct u_suballocator
*allocator_fetch_shader
;
424 boolean has_vertex_cache
;
425 unsigned default_gprs
[EG_NUM_HW_STAGES
];
426 unsigned current_gprs
[EG_NUM_HW_STAGES
];
427 unsigned r6xx_num_clause_temp_gprs
;
429 /* Miscellaneous state objects. */
430 void *custom_dsa_flush
;
431 void *custom_blend_resolve
;
432 void *custom_blend_decompress
;
433 void *custom_blend_fastclear
;
434 /* With rasterizer discard, there doesn't have to be a pixel shader.
435 * In that case, we bind this one: */
436 void *dummy_pixel_shader
;
437 /* These dummy CMASK and FMASK buffers are used to get around the R6xx hardware
438 * bug where valid CMASK and FMASK are required to be present to avoid
439 * a hardlock in certain operations but aren't actually used
440 * for anything useful. */
441 struct r600_resource
*dummy_fmask
;
442 struct r600_resource
*dummy_cmask
;
444 /* State binding slots are here. */
445 struct r600_atom
*atoms
[R600_NUM_ATOMS
];
446 /* Dirty atom bitmask for fast tests */
447 uint64_t dirty_atoms
;
448 /* States for CS initialization. */
449 struct r600_command_buffer start_cs_cmd
; /* invariant state mostly */
450 /** Compute specific registers initializations. The start_cs_cmd atom
451 * must be emitted before start_compute_cs_cmd. */
452 struct r600_command_buffer start_compute_cs_cmd
;
453 /* Register states. */
454 struct r600_alphatest_state alphatest_state
;
455 struct r600_cso_state blend_state
;
456 struct r600_blend_color blend_color
;
457 struct r600_cb_misc_state cb_misc_state
;
458 struct r600_clip_misc_state clip_misc_state
;
459 struct r600_clip_state clip_state
;
460 struct r600_db_misc_state db_misc_state
;
461 struct r600_db_state db_state
;
462 struct r600_cso_state dsa_state
;
463 struct r600_framebuffer framebuffer
;
464 struct r600_poly_offset_state poly_offset_state
;
465 struct r600_cso_state rasterizer_state
;
466 struct r600_sample_mask sample_mask
;
467 struct r600_seamless_cube_map seamless_cube_map
;
468 struct r600_config_state config_state
;
469 struct r600_stencil_ref_state stencil_ref
;
470 struct r600_vgt_state vgt_state
;
471 /* Shaders and shader resources. */
472 struct r600_cso_state vertex_fetch_shader
;
473 struct r600_shader_state hw_shader_stages
[EG_NUM_HW_STAGES
];
474 struct r600_cs_shader_state cs_shader_state
;
475 struct r600_shader_stages_state shader_stages
;
476 struct r600_gs_rings_state gs_rings
;
477 struct r600_constbuf_state constbuf_state
[PIPE_SHADER_TYPES
];
478 struct r600_textures_info samplers
[PIPE_SHADER_TYPES
];
480 struct r600_shader_driver_constants_info driver_consts
[PIPE_SHADER_TYPES
];
482 /** Vertex buffers for fetch shaders */
483 struct r600_vertexbuf_state vertex_buffer_state
;
484 /** Vertex buffers for compute shaders */
485 struct r600_vertexbuf_state cs_vertex_buffer_state
;
487 /* Additional context states. */
488 unsigned compute_cb_target_mask
;
489 struct r600_pipe_shader_selector
*ps_shader
;
490 struct r600_pipe_shader_selector
*vs_shader
;
491 struct r600_pipe_shader_selector
*gs_shader
;
493 struct r600_pipe_shader_selector
*tcs_shader
;
494 struct r600_pipe_shader_selector
*tes_shader
;
496 struct r600_pipe_shader_selector
*fixed_func_tcs_shader
;
498 struct r600_rasterizer_state
*rasterizer
;
500 bool force_blend_disable
;
501 boolean dual_src_blend
;
506 struct pipe_index_buffer index_buffer
;
508 /* Last draw state (-1 = unset). */
509 int last_primitive_type
; /* Last primitive type used in draw_vbo. */
510 int last_start_instance
;
513 struct r600_isa
*isa
;
514 float sample_positions
[4 * 16];
516 bool tess_state_dirty
;
517 struct r600_pipe_shader_selector
*last_ls
;
518 struct r600_pipe_shader_selector
*last_tcs
;
519 unsigned last_num_tcs_input_cp
;
523 static inline void r600_emit_command_buffer(struct radeon_winsys_cs
*cs
,
524 struct r600_command_buffer
*cb
)
526 assert(cs
->cdw
+ cb
->num_dw
<= cs
->max_dw
);
527 memcpy(cs
->buf
+ cs
->cdw
, cb
->buf
, 4 * cb
->num_dw
);
528 cs
->cdw
+= cb
->num_dw
;
531 static inline void r600_set_atom_dirty(struct r600_context
*rctx
,
532 struct r600_atom
*atom
,
537 assert(atom
->id
!= 0);
538 assert(atom
->id
< sizeof(mask
) * 8);
539 mask
= 1ull << atom
->id
;
541 rctx
->dirty_atoms
|= mask
;
543 rctx
->dirty_atoms
&= ~mask
;
546 static inline void r600_mark_atom_dirty(struct r600_context
*rctx
,
547 struct r600_atom
*atom
)
549 r600_set_atom_dirty(rctx
, atom
, true);
552 static inline void r600_emit_atom(struct r600_context
*rctx
, struct r600_atom
*atom
)
554 atom
->emit(&rctx
->b
, atom
);
555 r600_set_atom_dirty(rctx
, atom
, false);
558 static inline void r600_set_cso_state(struct r600_context
*rctx
,
559 struct r600_cso_state
*state
, void *cso
)
562 r600_set_atom_dirty(rctx
, &state
->atom
, cso
!= NULL
);
565 static inline void r600_set_cso_state_with_cb(struct r600_context
*rctx
,
566 struct r600_cso_state
*state
, void *cso
,
567 struct r600_command_buffer
*cb
)
570 state
->atom
.num_dw
= cb
? cb
->num_dw
: 0;
571 r600_set_cso_state(rctx
, state
, cso
);
574 /* compute_memory_pool.c */
575 struct compute_memory_pool
;
576 void compute_memory_pool_delete(struct compute_memory_pool
* pool
);
577 struct compute_memory_pool
* compute_memory_pool_new(
578 struct r600_screen
*rscreen
);
580 /* evergreen_state.c */
581 struct pipe_sampler_view
*
582 evergreen_create_sampler_view_custom(struct pipe_context
*ctx
,
583 struct pipe_resource
*texture
,
584 const struct pipe_sampler_view
*state
,
585 unsigned width0
, unsigned height0
,
586 unsigned force_level
);
587 void evergreen_init_common_regs(struct r600_context
*ctx
,
588 struct r600_command_buffer
*cb
,
589 enum chip_class ctx_chip_class
,
590 enum radeon_family ctx_family
,
592 void cayman_init_common_regs(struct r600_command_buffer
*cb
,
593 enum chip_class ctx_chip_class
,
594 enum radeon_family ctx_family
,
597 void evergreen_init_state_functions(struct r600_context
*rctx
);
598 void evergreen_init_atom_start_cs(struct r600_context
*rctx
);
599 void evergreen_update_ps_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
600 void evergreen_update_es_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
601 void evergreen_update_gs_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
602 void evergreen_update_vs_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
603 void evergreen_update_ls_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
604 void evergreen_update_hs_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
605 void *evergreen_create_db_flush_dsa(struct r600_context
*rctx
);
606 void *evergreen_create_resolve_blend(struct r600_context
*rctx
);
607 void *evergreen_create_decompress_blend(struct r600_context
*rctx
);
608 void *evergreen_create_fastclear_blend(struct r600_context
*rctx
);
609 boolean
evergreen_is_format_supported(struct pipe_screen
*screen
,
610 enum pipe_format format
,
611 enum pipe_texture_target target
,
612 unsigned sample_count
,
614 void evergreen_init_color_surface(struct r600_context
*rctx
,
615 struct r600_surface
*surf
);
616 void evergreen_init_color_surface_rat(struct r600_context
*rctx
,
617 struct r600_surface
*surf
);
618 void evergreen_update_db_shader_control(struct r600_context
* rctx
);
619 bool evergreen_adjust_gprs(struct r600_context
*rctx
);
621 void r600_init_blit_functions(struct r600_context
*rctx
);
622 void r600_decompress_depth_textures(struct r600_context
*rctx
,
623 struct r600_samplerview_state
*textures
);
624 void r600_decompress_color_textures(struct r600_context
*rctx
,
625 struct r600_samplerview_state
*textures
);
626 void r600_resource_copy_region(struct pipe_context
*ctx
,
627 struct pipe_resource
*dst
,
629 unsigned dstx
, unsigned dsty
, unsigned dstz
,
630 struct pipe_resource
*src
,
632 const struct pipe_box
*src_box
);
635 int r600_pipe_shader_create(struct pipe_context
*ctx
,
636 struct r600_pipe_shader
*shader
,
637 union r600_shader_key key
);
639 void r600_pipe_shader_destroy(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
642 struct pipe_sampler_view
*
643 r600_create_sampler_view_custom(struct pipe_context
*ctx
,
644 struct pipe_resource
*texture
,
645 const struct pipe_sampler_view
*state
,
646 unsigned width_first_level
, unsigned height_first_level
);
647 void r600_init_state_functions(struct r600_context
*rctx
);
648 void r600_init_atom_start_cs(struct r600_context
*rctx
);
649 void r600_update_ps_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
650 void r600_update_es_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
651 void r600_update_gs_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
652 void r600_update_vs_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
653 void *r600_create_db_flush_dsa(struct r600_context
*rctx
);
654 void *r600_create_resolve_blend(struct r600_context
*rctx
);
655 void *r700_create_resolve_blend(struct r600_context
*rctx
);
656 void *r600_create_decompress_blend(struct r600_context
*rctx
);
657 bool r600_adjust_gprs(struct r600_context
*rctx
);
658 boolean
r600_is_format_supported(struct pipe_screen
*screen
,
659 enum pipe_format format
,
660 enum pipe_texture_target target
,
661 unsigned sample_count
,
663 void r600_update_db_shader_control(struct r600_context
* rctx
);
665 /* r600_hw_context.c */
666 void r600_context_gfx_flush(void *context
, unsigned flags
,
667 struct pipe_fence_handle
**fence
);
668 void r600_begin_new_cs(struct r600_context
*ctx
);
669 void r600_flush_emit(struct r600_context
*ctx
);
670 void r600_need_cs_space(struct r600_context
*ctx
, unsigned num_dw
, boolean count_draw_in
);
671 void r600_cp_dma_copy_buffer(struct r600_context
*rctx
,
672 struct pipe_resource
*dst
, uint64_t dst_offset
,
673 struct pipe_resource
*src
, uint64_t src_offset
,
675 void evergreen_cp_dma_clear_buffer(struct r600_context
*rctx
,
676 struct pipe_resource
*dst
, uint64_t offset
,
677 unsigned size
, uint32_t clear_value
);
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 uint32_t r600_translate_colorformat(enum chip_class chip
, enum pipe_format format
);
744 uint32_t r600_colorformat_endian_swap(uint32_t colorformat
);
747 struct pipe_video_codec
*r600_uvd_create_decoder(struct pipe_context
*context
,
748 const struct pipe_video_codec
*decoder
);
750 struct pipe_video_buffer
*r600_video_buffer_create(struct pipe_context
*pipe
,
751 const struct pipe_video_buffer
*tmpl
);
754 * Helpers for building command buffers
757 #define PKT3_SET_CONFIG_REG 0x68
758 #define PKT3_SET_CONTEXT_REG 0x69
759 #define PKT3_SET_CTL_CONST 0x6F
760 #define PKT3_SET_LOOP_CONST 0x6C
762 #define R600_CONFIG_REG_OFFSET 0x08000
763 #define R600_CONTEXT_REG_OFFSET 0x28000
764 #define R600_CTL_CONST_OFFSET 0x3CFF0
765 #define R600_LOOP_CONST_OFFSET 0X0003E200
766 #define EG_LOOP_CONST_OFFSET 0x0003A200
768 #define PKT_TYPE_S(x) (((x) & 0x3) << 30)
769 #define PKT_COUNT_S(x) (((x) & 0x3FFF) << 16)
770 #define PKT3_IT_OPCODE_S(x) (((x) & 0xFF) << 8)
771 #define PKT3_PREDICATE(x) (((x) >> 0) & 0x1)
772 #define PKT3(op, count, predicate) (PKT_TYPE_S(3) | PKT_COUNT_S(count) | PKT3_IT_OPCODE_S(op) | PKT3_PREDICATE(predicate))
774 #define RADEON_CP_PACKET3_COMPUTE_MODE 0x00000002
776 /*Evergreen Compute packet3*/
777 #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)
779 static inline void r600_store_value(struct r600_command_buffer
*cb
, unsigned value
)
781 cb
->buf
[cb
->num_dw
++] = value
;
784 static inline void r600_store_array(struct r600_command_buffer
*cb
, unsigned num
, unsigned *ptr
)
786 assert(cb
->num_dw
+num
<= cb
->max_num_dw
);
787 memcpy(&cb
->buf
[cb
->num_dw
], ptr
, num
* sizeof(ptr
[0]));
791 static inline void r600_store_config_reg_seq(struct r600_command_buffer
*cb
, unsigned reg
, unsigned num
)
793 assert(reg
< R600_CONTEXT_REG_OFFSET
);
794 assert(cb
->num_dw
+2+num
<= cb
->max_num_dw
);
795 cb
->buf
[cb
->num_dw
++] = PKT3(PKT3_SET_CONFIG_REG
, num
, 0);
796 cb
->buf
[cb
->num_dw
++] = (reg
- R600_CONFIG_REG_OFFSET
) >> 2;
800 * Needs cb->pkt_flags set to RADEON_CP_PACKET3_COMPUTE_MODE for compute
803 static inline void r600_store_context_reg_seq(struct r600_command_buffer
*cb
, unsigned reg
, unsigned num
)
805 assert(reg
>= R600_CONTEXT_REG_OFFSET
&& reg
< R600_CTL_CONST_OFFSET
);
806 assert(cb
->num_dw
+2+num
<= cb
->max_num_dw
);
807 cb
->buf
[cb
->num_dw
++] = PKT3(PKT3_SET_CONTEXT_REG
, num
, 0) | cb
->pkt_flags
;
808 cb
->buf
[cb
->num_dw
++] = (reg
- R600_CONTEXT_REG_OFFSET
) >> 2;
812 * Needs cb->pkt_flags set to RADEON_CP_PACKET3_COMPUTE_MODE for compute
815 static inline void r600_store_ctl_const_seq(struct r600_command_buffer
*cb
, unsigned reg
, unsigned num
)
817 assert(reg
>= R600_CTL_CONST_OFFSET
);
818 assert(cb
->num_dw
+2+num
<= cb
->max_num_dw
);
819 cb
->buf
[cb
->num_dw
++] = PKT3(PKT3_SET_CTL_CONST
, num
, 0) | cb
->pkt_flags
;
820 cb
->buf
[cb
->num_dw
++] = (reg
- R600_CTL_CONST_OFFSET
) >> 2;
823 static inline void r600_store_loop_const_seq(struct r600_command_buffer
*cb
, unsigned reg
, unsigned num
)
825 assert(reg
>= R600_LOOP_CONST_OFFSET
);
826 assert(cb
->num_dw
+2+num
<= cb
->max_num_dw
);
827 cb
->buf
[cb
->num_dw
++] = PKT3(PKT3_SET_LOOP_CONST
, num
, 0);
828 cb
->buf
[cb
->num_dw
++] = (reg
- R600_LOOP_CONST_OFFSET
) >> 2;
832 * Needs cb->pkt_flags set to RADEON_CP_PACKET3_COMPUTE_MODE for compute
835 static inline void eg_store_loop_const_seq(struct r600_command_buffer
*cb
, unsigned reg
, unsigned num
)
837 assert(reg
>= EG_LOOP_CONST_OFFSET
);
838 assert(cb
->num_dw
+2+num
<= cb
->max_num_dw
);
839 cb
->buf
[cb
->num_dw
++] = PKT3(PKT3_SET_LOOP_CONST
, num
, 0) | cb
->pkt_flags
;
840 cb
->buf
[cb
->num_dw
++] = (reg
- EG_LOOP_CONST_OFFSET
) >> 2;
843 static inline void r600_store_config_reg(struct r600_command_buffer
*cb
, unsigned reg
, unsigned value
)
845 r600_store_config_reg_seq(cb
, reg
, 1);
846 r600_store_value(cb
, value
);
849 static inline void r600_store_context_reg(struct r600_command_buffer
*cb
, unsigned reg
, unsigned value
)
851 r600_store_context_reg_seq(cb
, reg
, 1);
852 r600_store_value(cb
, value
);
855 static inline void r600_store_ctl_const(struct r600_command_buffer
*cb
, unsigned reg
, unsigned value
)
857 r600_store_ctl_const_seq(cb
, reg
, 1);
858 r600_store_value(cb
, value
);
861 static inline void r600_store_loop_const(struct r600_command_buffer
*cb
, unsigned reg
, unsigned value
)
863 r600_store_loop_const_seq(cb
, reg
, 1);
864 r600_store_value(cb
, value
);
867 static inline void eg_store_loop_const(struct r600_command_buffer
*cb
, unsigned reg
, unsigned value
)
869 eg_store_loop_const_seq(cb
, reg
, 1);
870 r600_store_value(cb
, value
);
873 void r600_init_command_buffer(struct r600_command_buffer
*cb
, unsigned num_dw
);
874 void r600_release_command_buffer(struct r600_command_buffer
*cb
);
876 static inline void radeon_compute_set_context_reg_seq(struct radeon_winsys_cs
*cs
, unsigned reg
, unsigned num
)
878 radeon_set_context_reg_seq(cs
, reg
, num
);
879 /* Set the compute bit on the packet header */
880 cs
->buf
[cs
->cdw
- 2] |= RADEON_CP_PACKET3_COMPUTE_MODE
;
883 static inline void radeon_set_ctl_const_seq(struct radeon_winsys_cs
*cs
, unsigned reg
, unsigned num
)
885 assert(reg
>= R600_CTL_CONST_OFFSET
);
886 assert(cs
->cdw
+2+num
<= cs
->max_dw
);
887 cs
->buf
[cs
->cdw
++] = PKT3(PKT3_SET_CTL_CONST
, num
, 0);
888 cs
->buf
[cs
->cdw
++] = (reg
- R600_CTL_CONST_OFFSET
) >> 2;
891 static inline void radeon_compute_set_context_reg(struct radeon_winsys_cs
*cs
, unsigned reg
, unsigned value
)
893 radeon_compute_set_context_reg_seq(cs
, reg
, 1);
894 radeon_emit(cs
, value
);
897 static inline void radeon_set_context_reg_flag(struct radeon_winsys_cs
*cs
, unsigned reg
, unsigned value
, unsigned flag
)
899 if (flag
& RADEON_CP_PACKET3_COMPUTE_MODE
) {
900 radeon_compute_set_context_reg(cs
, reg
, value
);
902 radeon_set_context_reg(cs
, reg
, value
);
906 static inline void radeon_set_ctl_const(struct radeon_winsys_cs
*cs
, unsigned reg
, unsigned value
)
908 radeon_set_ctl_const_seq(cs
, reg
, 1);
909 radeon_emit(cs
, value
);
915 static inline uint32_t S_FIXED(float value
, uint32_t frac_bits
)
917 return value
* (1 << frac_bits
);
920 /* 12.4 fixed-point */
921 static inline unsigned r600_pack_float_12p4(float x
)
924 x
>= 4096 ? 0xffff : x
* 16;
927 /* Return if the depth format can be read without the DB->CB copy on r6xx-r7xx. */
928 static inline bool r600_can_read_depth(struct r600_texture
*rtex
)
930 return rtex
->resource
.b
.b
.nr_samples
<= 1 &&
931 (rtex
->resource
.b
.b
.format
== PIPE_FORMAT_Z16_UNORM
||
932 rtex
->resource
.b
.b
.format
== PIPE_FORMAT_Z32_FLOAT
);
935 #define V_028A6C_OUTPRIM_TYPE_POINTLIST 0
936 #define V_028A6C_OUTPRIM_TYPE_LINESTRIP 1
937 #define V_028A6C_OUTPRIM_TYPE_TRISTRIP 2
939 unsigned r600_conv_prim_to_gs_out(unsigned mode
);