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_viewport_state
{
223 struct r600_atom atom
;
224 struct pipe_viewport_state state
[R600_MAX_VIEWPORTS
];
228 struct r600_shader_stages_state
{
229 struct r600_atom atom
;
230 unsigned geom_enable
;
233 struct r600_gs_rings_state
{
234 struct r600_atom atom
;
236 struct pipe_constant_buffer esgs_ring
;
237 struct pipe_constant_buffer gsvs_ring
;
240 /* This must start from 16. */
242 #define DBG_NO_CP_DMA (1 << 30)
244 #define DBG_NO_SB (1 << 21)
245 #define DBG_SB_CS (1 << 22)
246 #define DBG_SB_DRY_RUN (1 << 23)
247 #define DBG_SB_STAT (1 << 24)
248 #define DBG_SB_DUMP (1 << 25)
249 #define DBG_SB_NO_FALLBACK (1 << 26)
250 #define DBG_SB_DISASM (1 << 27)
251 #define DBG_SB_SAFEMATH (1 << 28)
254 struct r600_common_screen b
;
256 bool has_compressed_msaa_texturing
;
258 /*for compute global memory binding, we allocate stuff here, instead of
260 * XXX: Not sure if this is the best place for global_pool. Also,
261 * it's not thread safe, so it won't work with multiple contexts. */
262 struct compute_memory_pool
*global_pool
;
265 struct r600_pipe_sampler_view
{
266 struct pipe_sampler_view base
;
267 struct list_head list
;
268 struct r600_resource
*tex_resource
;
269 uint32_t tex_resource_words
[8];
270 bool skip_mip_address_reloc
;
271 bool is_stencil_sampler
;
274 struct r600_rasterizer_state
{
275 struct r600_command_buffer buffer
;
278 unsigned sprite_coord_enable
;
279 unsigned clip_plane_enable
;
280 unsigned pa_sc_line_stipple
;
281 unsigned pa_cl_clip_cntl
;
282 unsigned pa_su_sc_mode_cntl
;
287 bool multisample_enable
;
290 struct r600_poly_offset_state
{
291 struct r600_atom atom
;
292 enum pipe_format zs_format
;
297 struct r600_blend_state
{
298 struct r600_command_buffer buffer
;
299 struct r600_command_buffer buffer_no_blend
;
300 unsigned cb_target_mask
;
301 unsigned cb_color_control
;
302 unsigned cb_color_control_no_blend
;
307 struct r600_dsa_state
{
308 struct r600_command_buffer buffer
;
313 unsigned sx_alpha_test_control
;
316 struct r600_pipe_shader
;
318 struct r600_pipe_shader_selector
{
319 struct r600_pipe_shader
*current
;
321 struct tgsi_token
*tokens
;
322 struct pipe_stream_output_info so
;
323 struct tgsi_shader_info info
;
325 unsigned num_shaders
;
327 /* PIPE_SHADER_[VERTEX|FRAGMENT|...] */
330 /* geometry shader properties */
331 unsigned gs_output_prim
;
332 unsigned gs_max_out_vertices
;
333 unsigned gs_num_invocations
;
336 uint64_t lds_patch_outputs_written_mask
;
337 uint64_t lds_outputs_written_mask
;
338 unsigned nr_ps_max_color_exports
;
341 struct r600_pipe_sampler_state
{
342 uint32_t tex_sampler_words
[3];
343 union pipe_color_union border_color
;
344 bool border_color_use
;
345 bool seamless_cube_map
;
348 /* needed for blitter save */
349 #define NUM_TEX_UNITS 16
351 struct r600_seamless_cube_map
{
352 struct r600_atom atom
;
356 struct r600_samplerview_state
{
357 struct r600_atom atom
;
358 struct r600_pipe_sampler_view
*views
[NUM_TEX_UNITS
];
359 uint32_t enabled_mask
;
361 uint32_t compressed_depthtex_mask
; /* which textures are depth */
362 uint32_t compressed_colortex_mask
;
363 boolean dirty_buffer_constants
;
366 struct r600_sampler_states
{
367 struct r600_atom atom
;
368 struct r600_pipe_sampler_state
*states
[NUM_TEX_UNITS
];
369 uint32_t enabled_mask
;
371 uint32_t has_bordercolor_mask
; /* which states contain the border color */
374 struct r600_textures_info
{
375 struct r600_samplerview_state views
;
376 struct r600_sampler_states states
;
377 bool is_array_sampler
[NUM_TEX_UNITS
];
380 struct r600_shader_driver_constants_info
{
381 /* currently 128 bytes for UCP/samplepos + sampler buffer constants */
385 bool texture_const_dirty
;
386 bool ps_sample_pos_dirty
;
389 struct r600_constbuf_state
391 struct r600_atom atom
;
392 struct pipe_constant_buffer cb
[PIPE_MAX_CONSTANT_BUFFERS
];
393 uint32_t enabled_mask
;
397 struct r600_vertexbuf_state
399 struct r600_atom atom
;
400 struct pipe_vertex_buffer vb
[PIPE_MAX_ATTRIBS
];
401 uint32_t enabled_mask
; /* non-NULL buffers */
405 /* CSO (constant state object, in other words, immutable state). */
406 struct r600_cso_state
408 struct r600_atom atom
;
409 void *cso
; /* e.g. r600_blend_state */
410 struct r600_command_buffer
*cb
;
413 struct r600_scissor_state
415 struct r600_atom atom
;
416 struct pipe_scissor_state scissor
[R600_MAX_VIEWPORTS
];
418 bool enable
; /* r6xx only */
421 struct r600_fetch_shader
{
422 struct r600_resource
*buffer
;
426 struct r600_shader_state
{
427 struct r600_atom atom
;
428 struct r600_pipe_shader
*shader
;
431 struct r600_context
{
432 struct r600_common_context b
;
433 struct r600_screen
*screen
;
434 struct blitter_context
*blitter
;
435 struct u_suballocator
*allocator_fetch_shader
;
438 boolean has_vertex_cache
;
439 unsigned default_gprs
[EG_NUM_HW_STAGES
];
440 unsigned current_gprs
[EG_NUM_HW_STAGES
];
441 unsigned r6xx_num_clause_temp_gprs
;
443 /* Miscellaneous state objects. */
444 void *custom_dsa_flush
;
445 void *custom_blend_resolve
;
446 void *custom_blend_decompress
;
447 void *custom_blend_fastclear
;
448 /* With rasterizer discard, there doesn't have to be a pixel shader.
449 * In that case, we bind this one: */
450 void *dummy_pixel_shader
;
451 /* These dummy CMASK and FMASK buffers are used to get around the R6xx hardware
452 * bug where valid CMASK and FMASK are required to be present to avoid
453 * a hardlock in certain operations but aren't actually used
454 * for anything useful. */
455 struct r600_resource
*dummy_fmask
;
456 struct r600_resource
*dummy_cmask
;
458 /* State binding slots are here. */
459 struct r600_atom
*atoms
[R600_NUM_ATOMS
];
460 /* Dirty atom bitmask for fast tests */
461 uint64_t dirty_atoms
;
462 /* States for CS initialization. */
463 struct r600_command_buffer start_cs_cmd
; /* invariant state mostly */
464 /** Compute specific registers initializations. The start_cs_cmd atom
465 * must be emitted before start_compute_cs_cmd. */
466 struct r600_command_buffer start_compute_cs_cmd
;
467 /* Register states. */
468 struct r600_alphatest_state alphatest_state
;
469 struct r600_cso_state blend_state
;
470 struct r600_blend_color blend_color
;
471 struct r600_cb_misc_state cb_misc_state
;
472 struct r600_clip_misc_state clip_misc_state
;
473 struct r600_clip_state clip_state
;
474 struct r600_db_misc_state db_misc_state
;
475 struct r600_db_state db_state
;
476 struct r600_cso_state dsa_state
;
477 struct r600_framebuffer framebuffer
;
478 struct r600_poly_offset_state poly_offset_state
;
479 struct r600_cso_state rasterizer_state
;
480 struct r600_sample_mask sample_mask
;
481 struct r600_scissor_state scissor
;
482 struct r600_seamless_cube_map seamless_cube_map
;
483 struct r600_config_state config_state
;
484 struct r600_stencil_ref_state stencil_ref
;
485 struct r600_vgt_state vgt_state
;
486 struct r600_viewport_state viewport
;
487 /* Shaders and shader resources. */
488 struct r600_cso_state vertex_fetch_shader
;
489 struct r600_shader_state hw_shader_stages
[EG_NUM_HW_STAGES
];
490 struct r600_cs_shader_state cs_shader_state
;
491 struct r600_shader_stages_state shader_stages
;
492 struct r600_gs_rings_state gs_rings
;
493 struct r600_constbuf_state constbuf_state
[PIPE_SHADER_TYPES
];
494 struct r600_textures_info samplers
[PIPE_SHADER_TYPES
];
496 struct r600_shader_driver_constants_info driver_consts
[PIPE_SHADER_TYPES
];
498 /** Vertex buffers for fetch shaders */
499 struct r600_vertexbuf_state vertex_buffer_state
;
500 /** Vertex buffers for compute shaders */
501 struct r600_vertexbuf_state cs_vertex_buffer_state
;
503 /* Additional context states. */
504 unsigned compute_cb_target_mask
;
505 struct r600_pipe_shader_selector
*ps_shader
;
506 struct r600_pipe_shader_selector
*vs_shader
;
507 struct r600_pipe_shader_selector
*gs_shader
;
509 struct r600_pipe_shader_selector
*tcs_shader
;
510 struct r600_pipe_shader_selector
*tes_shader
;
512 struct r600_pipe_shader_selector
*fixed_func_tcs_shader
;
514 struct r600_rasterizer_state
*rasterizer
;
516 bool force_blend_disable
;
517 boolean dual_src_blend
;
522 struct pipe_index_buffer index_buffer
;
524 /* Last draw state (-1 = unset). */
525 int last_primitive_type
; /* Last primitive type used in draw_vbo. */
526 int last_start_instance
;
529 struct r600_isa
*isa
;
530 float sample_positions
[4 * 16];
532 bool tess_state_dirty
;
533 struct r600_pipe_shader_selector
*last_ls
;
534 struct r600_pipe_shader_selector
*last_tcs
;
535 unsigned last_num_tcs_input_cp
;
539 static inline void r600_emit_command_buffer(struct radeon_winsys_cs
*cs
,
540 struct r600_command_buffer
*cb
)
542 assert(cs
->cdw
+ cb
->num_dw
<= cs
->max_dw
);
543 memcpy(cs
->buf
+ cs
->cdw
, cb
->buf
, 4 * cb
->num_dw
);
544 cs
->cdw
+= cb
->num_dw
;
547 static inline void r600_set_atom_dirty(struct r600_context
*rctx
,
548 struct r600_atom
*atom
,
553 assert(atom
->id
!= 0);
554 assert(atom
->id
< sizeof(mask
) * 8);
555 mask
= 1ull << atom
->id
;
557 rctx
->dirty_atoms
|= mask
;
559 rctx
->dirty_atoms
&= ~mask
;
562 static inline void r600_mark_atom_dirty(struct r600_context
*rctx
,
563 struct r600_atom
*atom
)
565 r600_set_atom_dirty(rctx
, atom
, true);
568 static inline void r600_emit_atom(struct r600_context
*rctx
, struct r600_atom
*atom
)
570 atom
->emit(&rctx
->b
, atom
);
571 r600_set_atom_dirty(rctx
, atom
, false);
574 static inline void r600_set_cso_state(struct r600_context
*rctx
,
575 struct r600_cso_state
*state
, void *cso
)
578 r600_set_atom_dirty(rctx
, &state
->atom
, cso
!= NULL
);
581 static inline void r600_set_cso_state_with_cb(struct r600_context
*rctx
,
582 struct r600_cso_state
*state
, void *cso
,
583 struct r600_command_buffer
*cb
)
586 state
->atom
.num_dw
= cb
? cb
->num_dw
: 0;
587 r600_set_cso_state(rctx
, state
, cso
);
590 /* compute_memory_pool.c */
591 struct compute_memory_pool
;
592 void compute_memory_pool_delete(struct compute_memory_pool
* pool
);
593 struct compute_memory_pool
* compute_memory_pool_new(
594 struct r600_screen
*rscreen
);
596 /* evergreen_state.c */
597 struct pipe_sampler_view
*
598 evergreen_create_sampler_view_custom(struct pipe_context
*ctx
,
599 struct pipe_resource
*texture
,
600 const struct pipe_sampler_view
*state
,
601 unsigned width0
, unsigned height0
,
602 unsigned force_level
);
603 void evergreen_init_common_regs(struct r600_context
*ctx
,
604 struct r600_command_buffer
*cb
,
605 enum chip_class ctx_chip_class
,
606 enum radeon_family ctx_family
,
608 void cayman_init_common_regs(struct r600_command_buffer
*cb
,
609 enum chip_class ctx_chip_class
,
610 enum radeon_family ctx_family
,
613 void evergreen_init_state_functions(struct r600_context
*rctx
);
614 void evergreen_init_atom_start_cs(struct r600_context
*rctx
);
615 void evergreen_update_ps_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
616 void evergreen_update_es_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
617 void evergreen_update_gs_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
618 void evergreen_update_vs_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
619 void evergreen_update_ls_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
620 void evergreen_update_hs_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
621 void *evergreen_create_db_flush_dsa(struct r600_context
*rctx
);
622 void *evergreen_create_resolve_blend(struct r600_context
*rctx
);
623 void *evergreen_create_decompress_blend(struct r600_context
*rctx
);
624 void *evergreen_create_fastclear_blend(struct r600_context
*rctx
);
625 boolean
evergreen_is_format_supported(struct pipe_screen
*screen
,
626 enum pipe_format format
,
627 enum pipe_texture_target target
,
628 unsigned sample_count
,
630 void evergreen_init_color_surface(struct r600_context
*rctx
,
631 struct r600_surface
*surf
);
632 void evergreen_init_color_surface_rat(struct r600_context
*rctx
,
633 struct r600_surface
*surf
);
634 void evergreen_update_db_shader_control(struct r600_context
* rctx
);
635 bool evergreen_adjust_gprs(struct r600_context
*rctx
);
637 void r600_init_blit_functions(struct r600_context
*rctx
);
638 void r600_decompress_depth_textures(struct r600_context
*rctx
,
639 struct r600_samplerview_state
*textures
);
640 void r600_decompress_color_textures(struct r600_context
*rctx
,
641 struct r600_samplerview_state
*textures
);
642 void r600_resource_copy_region(struct pipe_context
*ctx
,
643 struct pipe_resource
*dst
,
645 unsigned dstx
, unsigned dsty
, unsigned dstz
,
646 struct pipe_resource
*src
,
648 const struct pipe_box
*src_box
);
651 int r600_pipe_shader_create(struct pipe_context
*ctx
,
652 struct r600_pipe_shader
*shader
,
653 union r600_shader_key key
);
655 void r600_pipe_shader_destroy(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
658 struct pipe_sampler_view
*
659 r600_create_sampler_view_custom(struct pipe_context
*ctx
,
660 struct pipe_resource
*texture
,
661 const struct pipe_sampler_view
*state
,
662 unsigned width_first_level
, unsigned height_first_level
);
663 void r600_init_state_functions(struct r600_context
*rctx
);
664 void r600_init_atom_start_cs(struct r600_context
*rctx
);
665 void r600_update_ps_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
666 void r600_update_es_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
667 void r600_update_gs_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
668 void r600_update_vs_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
669 void *r600_create_db_flush_dsa(struct r600_context
*rctx
);
670 void *r600_create_resolve_blend(struct r600_context
*rctx
);
671 void *r700_create_resolve_blend(struct r600_context
*rctx
);
672 void *r600_create_decompress_blend(struct r600_context
*rctx
);
673 bool r600_adjust_gprs(struct r600_context
*rctx
);
674 boolean
r600_is_format_supported(struct pipe_screen
*screen
,
675 enum pipe_format format
,
676 enum pipe_texture_target target
,
677 unsigned sample_count
,
679 void r600_update_db_shader_control(struct r600_context
* rctx
);
681 /* r600_hw_context.c */
682 void r600_context_gfx_flush(void *context
, unsigned flags
,
683 struct pipe_fence_handle
**fence
);
684 void r600_begin_new_cs(struct r600_context
*ctx
);
685 void r600_flush_emit(struct r600_context
*ctx
);
686 void r600_need_cs_space(struct r600_context
*ctx
, unsigned num_dw
, boolean count_draw_in
);
687 void r600_cp_dma_copy_buffer(struct r600_context
*rctx
,
688 struct pipe_resource
*dst
, uint64_t dst_offset
,
689 struct pipe_resource
*src
, uint64_t src_offset
,
691 void evergreen_cp_dma_clear_buffer(struct r600_context
*rctx
,
692 struct pipe_resource
*dst
, uint64_t offset
,
693 unsigned size
, uint32_t clear_value
);
694 void r600_dma_copy_buffer(struct r600_context
*rctx
,
695 struct pipe_resource
*dst
,
696 struct pipe_resource
*src
,
702 * evergreen_hw_context.c
704 void evergreen_dma_copy_buffer(struct r600_context
*rctx
,
705 struct pipe_resource
*dst
,
706 struct pipe_resource
*src
,
710 void evergreen_setup_tess_constants(struct r600_context
*rctx
,
711 const struct pipe_draw_info
*info
,
712 unsigned *num_patches
);
713 uint32_t evergreen_get_ls_hs_config(struct r600_context
*rctx
,
714 const struct pipe_draw_info
*info
,
715 unsigned num_patches
);
716 void evergreen_set_ls_hs_config(struct r600_context
*rctx
,
717 struct radeon_winsys_cs
*cs
,
718 uint32_t ls_hs_config
);
719 void evergreen_set_lds_alloc(struct r600_context
*rctx
,
720 struct radeon_winsys_cs
*cs
,
723 /* r600_state_common.c */
724 void r600_init_common_state_functions(struct r600_context
*rctx
);
725 void r600_emit_cso_state(struct r600_context
*rctx
, struct r600_atom
*atom
);
726 void r600_emit_alphatest_state(struct r600_context
*rctx
, struct r600_atom
*atom
);
727 void r600_emit_blend_color(struct r600_context
*rctx
, struct r600_atom
*atom
);
728 void r600_emit_vgt_state(struct r600_context
*rctx
, struct r600_atom
*atom
);
729 void r600_emit_clip_misc_state(struct r600_context
*rctx
, struct r600_atom
*atom
);
730 void r600_emit_stencil_ref(struct r600_context
*rctx
, struct r600_atom
*atom
);
731 void r600_emit_viewport_state(struct r600_context
*rctx
, struct r600_atom
*atom
);
732 void r600_emit_shader(struct r600_context
*rctx
, struct r600_atom
*a
);
733 void r600_add_atom(struct r600_context
*rctx
, struct r600_atom
*atom
, unsigned id
);
734 void r600_init_atom(struct r600_context
*rctx
, struct r600_atom
*atom
, unsigned id
,
735 void (*emit
)(struct r600_context
*ctx
, struct r600_atom
*state
),
737 void r600_vertex_buffers_dirty(struct r600_context
*rctx
);
738 void r600_sampler_views_dirty(struct r600_context
*rctx
,
739 struct r600_samplerview_state
*state
);
740 void r600_sampler_states_dirty(struct r600_context
*rctx
,
741 struct r600_sampler_states
*state
);
742 void r600_constant_buffers_dirty(struct r600_context
*rctx
, struct r600_constbuf_state
*state
);
743 void r600_set_sample_locations_constant_buffer(struct r600_context
*rctx
);
744 uint32_t r600_translate_stencil_op(int s_op
);
745 uint32_t r600_translate_fill(uint32_t func
);
746 unsigned r600_tex_wrap(unsigned wrap
);
747 unsigned r600_tex_mipfilter(unsigned filter
);
748 unsigned r600_tex_compare(unsigned compare
);
749 bool sampler_state_needs_border_color(const struct pipe_sampler_state
*state
);
750 struct pipe_surface
*r600_create_surface_custom(struct pipe_context
*pipe
,
751 struct pipe_resource
*texture
,
752 const struct pipe_surface
*templ
,
753 unsigned width
, unsigned height
);
754 unsigned r600_get_swizzle_combined(const unsigned char *swizzle_format
,
755 const unsigned char *swizzle_view
,
757 uint32_t r600_translate_texformat(struct pipe_screen
*screen
, enum pipe_format format
,
758 const unsigned char *swizzle_view
,
759 uint32_t *word4_p
, uint32_t *yuv_format_p
);
760 uint32_t r600_translate_colorformat(enum chip_class chip
, enum pipe_format format
);
761 uint32_t r600_colorformat_endian_swap(uint32_t colorformat
);
764 struct pipe_video_codec
*r600_uvd_create_decoder(struct pipe_context
*context
,
765 const struct pipe_video_codec
*decoder
);
767 struct pipe_video_buffer
*r600_video_buffer_create(struct pipe_context
*pipe
,
768 const struct pipe_video_buffer
*tmpl
);
771 * Helpers for building command buffers
774 #define PKT3_SET_CONFIG_REG 0x68
775 #define PKT3_SET_CONTEXT_REG 0x69
776 #define PKT3_SET_CTL_CONST 0x6F
777 #define PKT3_SET_LOOP_CONST 0x6C
779 #define R600_CONFIG_REG_OFFSET 0x08000
780 #define R600_CONTEXT_REG_OFFSET 0x28000
781 #define R600_CTL_CONST_OFFSET 0x3CFF0
782 #define R600_LOOP_CONST_OFFSET 0X0003E200
783 #define EG_LOOP_CONST_OFFSET 0x0003A200
785 #define PKT_TYPE_S(x) (((x) & 0x3) << 30)
786 #define PKT_COUNT_S(x) (((x) & 0x3FFF) << 16)
787 #define PKT3_IT_OPCODE_S(x) (((x) & 0xFF) << 8)
788 #define PKT3_PREDICATE(x) (((x) >> 0) & 0x1)
789 #define PKT3(op, count, predicate) (PKT_TYPE_S(3) | PKT_COUNT_S(count) | PKT3_IT_OPCODE_S(op) | PKT3_PREDICATE(predicate))
791 #define RADEON_CP_PACKET3_COMPUTE_MODE 0x00000002
793 /*Evergreen Compute packet3*/
794 #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)
796 static inline void r600_store_value(struct r600_command_buffer
*cb
, unsigned value
)
798 cb
->buf
[cb
->num_dw
++] = value
;
801 static inline void r600_store_array(struct r600_command_buffer
*cb
, unsigned num
, unsigned *ptr
)
803 assert(cb
->num_dw
+num
<= cb
->max_num_dw
);
804 memcpy(&cb
->buf
[cb
->num_dw
], ptr
, num
* sizeof(ptr
[0]));
808 static inline void r600_store_config_reg_seq(struct r600_command_buffer
*cb
, unsigned reg
, unsigned num
)
810 assert(reg
< R600_CONTEXT_REG_OFFSET
);
811 assert(cb
->num_dw
+2+num
<= cb
->max_num_dw
);
812 cb
->buf
[cb
->num_dw
++] = PKT3(PKT3_SET_CONFIG_REG
, num
, 0);
813 cb
->buf
[cb
->num_dw
++] = (reg
- R600_CONFIG_REG_OFFSET
) >> 2;
817 * Needs cb->pkt_flags set to RADEON_CP_PACKET3_COMPUTE_MODE for compute
820 static inline void r600_store_context_reg_seq(struct r600_command_buffer
*cb
, unsigned reg
, unsigned num
)
822 assert(reg
>= R600_CONTEXT_REG_OFFSET
&& reg
< R600_CTL_CONST_OFFSET
);
823 assert(cb
->num_dw
+2+num
<= cb
->max_num_dw
);
824 cb
->buf
[cb
->num_dw
++] = PKT3(PKT3_SET_CONTEXT_REG
, num
, 0) | cb
->pkt_flags
;
825 cb
->buf
[cb
->num_dw
++] = (reg
- R600_CONTEXT_REG_OFFSET
) >> 2;
829 * Needs cb->pkt_flags set to RADEON_CP_PACKET3_COMPUTE_MODE for compute
832 static inline void r600_store_ctl_const_seq(struct r600_command_buffer
*cb
, unsigned reg
, unsigned num
)
834 assert(reg
>= R600_CTL_CONST_OFFSET
);
835 assert(cb
->num_dw
+2+num
<= cb
->max_num_dw
);
836 cb
->buf
[cb
->num_dw
++] = PKT3(PKT3_SET_CTL_CONST
, num
, 0) | cb
->pkt_flags
;
837 cb
->buf
[cb
->num_dw
++] = (reg
- R600_CTL_CONST_OFFSET
) >> 2;
840 static inline void r600_store_loop_const_seq(struct r600_command_buffer
*cb
, unsigned reg
, unsigned num
)
842 assert(reg
>= R600_LOOP_CONST_OFFSET
);
843 assert(cb
->num_dw
+2+num
<= cb
->max_num_dw
);
844 cb
->buf
[cb
->num_dw
++] = PKT3(PKT3_SET_LOOP_CONST
, num
, 0);
845 cb
->buf
[cb
->num_dw
++] = (reg
- R600_LOOP_CONST_OFFSET
) >> 2;
849 * Needs cb->pkt_flags set to RADEON_CP_PACKET3_COMPUTE_MODE for compute
852 static inline void eg_store_loop_const_seq(struct r600_command_buffer
*cb
, unsigned reg
, unsigned num
)
854 assert(reg
>= EG_LOOP_CONST_OFFSET
);
855 assert(cb
->num_dw
+2+num
<= cb
->max_num_dw
);
856 cb
->buf
[cb
->num_dw
++] = PKT3(PKT3_SET_LOOP_CONST
, num
, 0) | cb
->pkt_flags
;
857 cb
->buf
[cb
->num_dw
++] = (reg
- EG_LOOP_CONST_OFFSET
) >> 2;
860 static inline void r600_store_config_reg(struct r600_command_buffer
*cb
, unsigned reg
, unsigned value
)
862 r600_store_config_reg_seq(cb
, reg
, 1);
863 r600_store_value(cb
, value
);
866 static inline void r600_store_context_reg(struct r600_command_buffer
*cb
, unsigned reg
, unsigned value
)
868 r600_store_context_reg_seq(cb
, reg
, 1);
869 r600_store_value(cb
, value
);
872 static inline void r600_store_ctl_const(struct r600_command_buffer
*cb
, unsigned reg
, unsigned value
)
874 r600_store_ctl_const_seq(cb
, reg
, 1);
875 r600_store_value(cb
, value
);
878 static inline void r600_store_loop_const(struct r600_command_buffer
*cb
, unsigned reg
, unsigned value
)
880 r600_store_loop_const_seq(cb
, reg
, 1);
881 r600_store_value(cb
, value
);
884 static inline void eg_store_loop_const(struct r600_command_buffer
*cb
, unsigned reg
, unsigned value
)
886 eg_store_loop_const_seq(cb
, reg
, 1);
887 r600_store_value(cb
, value
);
890 void r600_init_command_buffer(struct r600_command_buffer
*cb
, unsigned num_dw
);
891 void r600_release_command_buffer(struct r600_command_buffer
*cb
);
893 static inline void radeon_compute_set_context_reg_seq(struct radeon_winsys_cs
*cs
, unsigned reg
, unsigned num
)
895 radeon_set_context_reg_seq(cs
, reg
, num
);
896 /* Set the compute bit on the packet header */
897 cs
->buf
[cs
->cdw
- 2] |= RADEON_CP_PACKET3_COMPUTE_MODE
;
900 static inline void radeon_set_ctl_const_seq(struct radeon_winsys_cs
*cs
, unsigned reg
, unsigned num
)
902 assert(reg
>= R600_CTL_CONST_OFFSET
);
903 assert(cs
->cdw
+2+num
<= cs
->max_dw
);
904 cs
->buf
[cs
->cdw
++] = PKT3(PKT3_SET_CTL_CONST
, num
, 0);
905 cs
->buf
[cs
->cdw
++] = (reg
- R600_CTL_CONST_OFFSET
) >> 2;
908 static inline void radeon_compute_set_context_reg(struct radeon_winsys_cs
*cs
, unsigned reg
, unsigned value
)
910 radeon_compute_set_context_reg_seq(cs
, reg
, 1);
911 radeon_emit(cs
, value
);
914 static inline void radeon_set_context_reg_flag(struct radeon_winsys_cs
*cs
, unsigned reg
, unsigned value
, unsigned flag
)
916 if (flag
& RADEON_CP_PACKET3_COMPUTE_MODE
) {
917 radeon_compute_set_context_reg(cs
, reg
, value
);
919 radeon_set_context_reg(cs
, reg
, value
);
923 static inline void radeon_set_ctl_const(struct radeon_winsys_cs
*cs
, unsigned reg
, unsigned value
)
925 radeon_set_ctl_const_seq(cs
, reg
, 1);
926 radeon_emit(cs
, value
);
932 static inline uint32_t S_FIXED(float value
, uint32_t frac_bits
)
934 return value
* (1 << frac_bits
);
937 /* 12.4 fixed-point */
938 static inline unsigned r600_pack_float_12p4(float x
)
941 x
>= 4096 ? 0xffff : x
* 16;
944 /* Return if the depth format can be read without the DB->CB copy on r6xx-r7xx. */
945 static inline bool r600_can_read_depth(struct r600_texture
*rtex
)
947 return rtex
->resource
.b
.b
.nr_samples
<= 1 &&
948 (rtex
->resource
.b
.b
.format
== PIPE_FORMAT_Z16_UNORM
||
949 rtex
->resource
.b
.b
.format
== PIPE_FORMAT_Z32_FLOAT
);
952 #define V_028A6C_OUTPRIM_TYPE_POINTLIST 0
953 #define V_028A6C_OUTPRIM_TYPE_LINESTRIP 1
954 #define V_028A6C_OUTPRIM_TYPE_TRISTRIP 2
956 unsigned r600_conv_prim_to_gs_out(unsigned mode
);