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"
32 #include "r600_llvm.h"
33 #include "r600_public.h"
35 #include "util/u_suballoc.h"
36 #include "util/list.h"
37 #include "util/u_transfer.h"
39 #include "tgsi/tgsi_scan.h"
41 #define R600_NUM_ATOMS 52
43 #define R600_MAX_VIEWPORTS 16
46 #define R600_CONTEXT_INV_VERTEX_CACHE (R600_CONTEXT_PRIVATE_FLAG << 0)
47 #define R600_CONTEXT_INV_TEX_CACHE (R600_CONTEXT_PRIVATE_FLAG << 1)
48 #define R600_CONTEXT_INV_CONST_CACHE (R600_CONTEXT_PRIVATE_FLAG << 2)
49 /* read-write caches */
50 #define R600_CONTEXT_FLUSH_AND_INV (R600_CONTEXT_PRIVATE_FLAG << 3)
51 #define R600_CONTEXT_FLUSH_AND_INV_CB_META (R600_CONTEXT_PRIVATE_FLAG << 4)
52 #define R600_CONTEXT_FLUSH_AND_INV_DB_META (R600_CONTEXT_PRIVATE_FLAG << 5)
53 #define R600_CONTEXT_FLUSH_AND_INV_DB (R600_CONTEXT_PRIVATE_FLAG << 6)
54 #define R600_CONTEXT_FLUSH_AND_INV_CB (R600_CONTEXT_PRIVATE_FLAG << 7)
55 /* engine synchronization */
56 #define R600_CONTEXT_PS_PARTIAL_FLUSH (R600_CONTEXT_PRIVATE_FLAG << 8)
57 #define R600_CONTEXT_WAIT_3D_IDLE (R600_CONTEXT_PRIVATE_FLAG << 9)
58 #define R600_CONTEXT_WAIT_CP_DMA_IDLE (R600_CONTEXT_PRIVATE_FLAG << 10)
60 /* the number of CS dwords for flushing and drawing */
61 #define R600_MAX_FLUSH_CS_DWORDS 16
62 #define R600_MAX_DRAW_CS_DWORDS 58
63 #define R600_TRACE_CS_DWORDS 7
65 #define R600_MAX_USER_CONST_BUFFERS 13
66 #define R600_MAX_DRIVER_CONST_BUFFERS 3
67 #define R600_MAX_CONST_BUFFERS (R600_MAX_USER_CONST_BUFFERS + R600_MAX_DRIVER_CONST_BUFFERS)
69 /* start driver buffers after user buffers */
70 #define R600_BUFFER_INFO_CONST_BUFFER (R600_MAX_USER_CONST_BUFFERS)
71 #define R600_UCP_SIZE (4*4*8)
72 #define R600_BUFFER_INFO_OFFSET (R600_UCP_SIZE)
74 #define R600_LDS_INFO_CONST_BUFFER (R600_MAX_USER_CONST_BUFFERS + 1)
76 * Note GS doesn't use a constant buffer binding, just a resource index,
77 * so it's fine to have it exist at index 16.
79 #define R600_GS_RING_CONST_BUFFER (R600_MAX_USER_CONST_BUFFERS + 2)
80 /* Currently R600_MAX_CONST_BUFFERS just fits on the hw, which has a limit
81 * of 16 const buffers.
82 * UCP/SAMPLE_POSITIONS are never accessed by same shader stage so they can use the same id.
84 * In order to support d3d 11 mandated minimum of 15 user const buffers
85 * we'd have to squash all use cases into one driver buffer.
87 #define R600_MAX_CONST_BUFFER_SIZE (4096 * sizeof(float[4]))
90 #define R600_HW_STAGE_PS 0
91 #define R600_HW_STAGE_VS 1
92 #define R600_HW_STAGE_GS 2
93 #define R600_HW_STAGE_ES 3
94 #define EG_HW_STAGE_LS 4
95 #define EG_HW_STAGE_HS 5
97 #define R600_NUM_HW_STAGES 4
98 #define EG_NUM_HW_STAGES 6
100 #ifdef PIPE_ARCH_BIG_ENDIAN
101 #define R600_BIG_ENDIAN 1
103 #define R600_BIG_ENDIAN 0
107 struct r600_bytecode
;
108 union r600_shader_key
;
110 /* This is an atom containing GPU commands that never change.
111 * This is supposed to be copied directly into the CS. */
112 struct r600_command_buffer
{
119 struct r600_db_state
{
120 struct r600_atom atom
;
121 struct r600_surface
*rsurf
;
124 struct r600_db_misc_state
{
125 struct r600_atom atom
;
126 bool occlusion_query_enabled
;
127 bool flush_depthstencil_through_cb
;
128 bool flush_depth_inplace
;
129 bool flush_stencil_inplace
;
130 bool copy_depth
, copy_stencil
;
131 unsigned copy_sample
;
132 unsigned log_samples
;
133 unsigned db_shader_control
;
135 uint8_t ps_conservative_z
;
138 struct r600_cb_misc_state
{
139 struct r600_atom atom
;
140 unsigned cb_color_control
; /* this comes from blend state */
141 unsigned blend_colormask
; /* 8*4 bits for 8 RGBA colorbuffers */
143 unsigned nr_ps_color_outputs
;
148 struct r600_clip_misc_state
{
149 struct r600_atom atom
;
150 unsigned pa_cl_clip_cntl
; /* from rasterizer */
151 unsigned pa_cl_vs_out_cntl
; /* from vertex shader */
152 unsigned clip_plane_enable
; /* from rasterizer */
153 unsigned clip_dist_write
; /* from vertex shader */
154 boolean clip_disable
; /* from vertex shader */
157 struct r600_alphatest_state
{
158 struct r600_atom atom
;
159 unsigned sx_alpha_test_control
; /* this comes from dsa state */
160 unsigned sx_alpha_ref
; /* this comes from dsa state */
162 bool cb0_export_16bpc
; /* from set_framebuffer_state */
165 struct r600_vgt_state
{
166 struct r600_atom atom
;
167 uint32_t vgt_multi_prim_ib_reset_en
;
168 uint32_t vgt_multi_prim_ib_reset_indx
;
169 uint32_t vgt_indx_offset
;
170 bool last_draw_was_indirect
;
173 struct r600_blend_color
{
174 struct r600_atom atom
;
175 struct pipe_blend_color state
;
178 struct r600_clip_state
{
179 struct r600_atom atom
;
180 struct pipe_clip_state state
;
183 struct r600_cs_shader_state
{
184 struct r600_atom atom
;
185 unsigned kernel_index
;
187 struct r600_pipe_compute
*shader
;
190 struct r600_framebuffer
{
191 struct r600_atom atom
;
192 struct pipe_framebuffer_state state
;
193 unsigned compressed_cb_mask
;
197 bool is_msaa_resolve
;
200 struct r600_sample_mask
{
201 struct r600_atom atom
;
202 uint16_t sample_mask
; /* there are only 8 bits on EG, 16 bits on Cayman */
205 struct r600_config_state
{
206 struct r600_atom atom
;
207 unsigned sq_gpr_resource_mgmt_1
;
208 unsigned sq_gpr_resource_mgmt_2
;
209 unsigned sq_gpr_resource_mgmt_3
;
210 bool dyn_gpr_enabled
;
213 struct r600_stencil_ref
220 struct r600_stencil_ref_state
{
221 struct r600_atom atom
;
222 struct r600_stencil_ref state
;
223 struct pipe_stencil_ref pipe_state
;
226 struct r600_viewport_state
{
227 struct r600_atom atom
;
228 struct pipe_viewport_state state
[R600_MAX_VIEWPORTS
];
232 struct r600_shader_stages_state
{
233 struct r600_atom atom
;
234 unsigned geom_enable
;
237 struct r600_gs_rings_state
{
238 struct r600_atom atom
;
240 struct pipe_constant_buffer esgs_ring
;
241 struct pipe_constant_buffer gsvs_ring
;
244 /* This must start from 16. */
246 #define DBG_LLVM (1 << 29)
247 #define DBG_NO_CP_DMA (1 << 30)
249 #define DBG_NO_SB (1 << 21)
250 #define DBG_SB_CS (1 << 22)
251 #define DBG_SB_DRY_RUN (1 << 23)
252 #define DBG_SB_STAT (1 << 24)
253 #define DBG_SB_DUMP (1 << 25)
254 #define DBG_SB_NO_FALLBACK (1 << 26)
255 #define DBG_SB_DISASM (1 << 27)
256 #define DBG_SB_SAFEMATH (1 << 28)
259 struct r600_common_screen b
;
261 bool has_compressed_msaa_texturing
;
263 /*for compute global memory binding, we allocate stuff here, instead of
265 * XXX: Not sure if this is the best place for global_pool. Also,
266 * it's not thread safe, so it won't work with multiple contexts. */
267 struct compute_memory_pool
*global_pool
;
270 struct r600_pipe_sampler_view
{
271 struct pipe_sampler_view base
;
272 struct list_head list
;
273 struct r600_resource
*tex_resource
;
274 uint32_t tex_resource_words
[8];
275 bool skip_mip_address_reloc
;
276 bool is_stencil_sampler
;
279 struct r600_rasterizer_state
{
280 struct r600_command_buffer buffer
;
283 unsigned sprite_coord_enable
;
284 unsigned clip_plane_enable
;
285 unsigned pa_sc_line_stipple
;
286 unsigned pa_cl_clip_cntl
;
287 unsigned pa_su_sc_mode_cntl
;
292 bool multisample_enable
;
295 struct r600_poly_offset_state
{
296 struct r600_atom atom
;
297 enum pipe_format zs_format
;
302 struct r600_blend_state
{
303 struct r600_command_buffer buffer
;
304 struct r600_command_buffer buffer_no_blend
;
305 unsigned cb_target_mask
;
306 unsigned cb_color_control
;
307 unsigned cb_color_control_no_blend
;
312 struct r600_dsa_state
{
313 struct r600_command_buffer buffer
;
318 unsigned sx_alpha_test_control
;
321 struct r600_pipe_shader
;
323 struct r600_pipe_shader_selector
{
324 struct r600_pipe_shader
*current
;
326 struct tgsi_token
*tokens
;
327 struct pipe_stream_output_info so
;
328 struct tgsi_shader_info info
;
330 unsigned num_shaders
;
332 /* PIPE_SHADER_[VERTEX|FRAGMENT|...] */
335 /* geometry shader properties */
336 unsigned gs_output_prim
;
337 unsigned gs_max_out_vertices
;
338 unsigned gs_num_invocations
;
341 uint64_t lds_patch_outputs_written_mask
;
342 uint64_t lds_outputs_written_mask
;
343 unsigned nr_ps_max_color_exports
;
346 struct r600_pipe_sampler_state
{
347 uint32_t tex_sampler_words
[3];
348 union pipe_color_union border_color
;
349 bool border_color_use
;
350 bool seamless_cube_map
;
353 /* needed for blitter save */
354 #define NUM_TEX_UNITS 16
356 struct r600_seamless_cube_map
{
357 struct r600_atom atom
;
361 struct r600_samplerview_state
{
362 struct r600_atom atom
;
363 struct r600_pipe_sampler_view
*views
[NUM_TEX_UNITS
];
364 uint32_t enabled_mask
;
366 uint32_t compressed_depthtex_mask
; /* which textures are depth */
367 uint32_t compressed_colortex_mask
;
368 boolean dirty_buffer_constants
;
371 struct r600_sampler_states
{
372 struct r600_atom atom
;
373 struct r600_pipe_sampler_state
*states
[NUM_TEX_UNITS
];
374 uint32_t enabled_mask
;
376 uint32_t has_bordercolor_mask
; /* which states contain the border color */
379 struct r600_textures_info
{
380 struct r600_samplerview_state views
;
381 struct r600_sampler_states states
;
382 bool is_array_sampler
[NUM_TEX_UNITS
];
385 struct r600_shader_driver_constants_info
{
386 /* currently 128 bytes for UCP/samplepos + sampler buffer constants */
390 bool texture_const_dirty
;
391 bool ps_sample_pos_dirty
;
394 struct r600_constbuf_state
396 struct r600_atom atom
;
397 struct pipe_constant_buffer cb
[PIPE_MAX_CONSTANT_BUFFERS
];
398 uint32_t enabled_mask
;
402 struct r600_vertexbuf_state
404 struct r600_atom atom
;
405 struct pipe_vertex_buffer vb
[PIPE_MAX_ATTRIBS
];
406 uint32_t enabled_mask
; /* non-NULL buffers */
410 /* CSO (constant state object, in other words, immutable state). */
411 struct r600_cso_state
413 struct r600_atom atom
;
414 void *cso
; /* e.g. r600_blend_state */
415 struct r600_command_buffer
*cb
;
418 struct r600_scissor_state
420 struct r600_atom atom
;
421 struct pipe_scissor_state scissor
[R600_MAX_VIEWPORTS
];
423 bool enable
; /* r6xx only */
426 struct r600_fetch_shader
{
427 struct r600_resource
*buffer
;
431 struct r600_shader_state
{
432 struct r600_atom atom
;
433 struct r600_pipe_shader
*shader
;
436 struct r600_context
{
437 struct r600_common_context b
;
438 struct r600_screen
*screen
;
439 struct blitter_context
*blitter
;
440 struct u_suballocator
*allocator_fetch_shader
;
443 boolean has_vertex_cache
;
444 boolean keep_tiling_flags
;
445 unsigned default_gprs
[EG_NUM_HW_STAGES
];
446 unsigned current_gprs
[EG_NUM_HW_STAGES
];
447 unsigned r6xx_num_clause_temp_gprs
;
449 /* Miscellaneous state objects. */
450 void *custom_dsa_flush
;
451 void *custom_blend_resolve
;
452 void *custom_blend_decompress
;
453 void *custom_blend_fastclear
;
454 /* With rasterizer discard, there doesn't have to be a pixel shader.
455 * In that case, we bind this one: */
456 void *dummy_pixel_shader
;
457 /* These dummy CMASK and FMASK buffers are used to get around the R6xx hardware
458 * bug where valid CMASK and FMASK are required to be present to avoid
459 * a hardlock in certain operations but aren't actually used
460 * for anything useful. */
461 struct r600_resource
*dummy_fmask
;
462 struct r600_resource
*dummy_cmask
;
464 /* State binding slots are here. */
465 struct r600_atom
*atoms
[R600_NUM_ATOMS
];
466 /* Dirty atom bitmask for fast tests */
467 uint64_t dirty_atoms
;
468 /* States for CS initialization. */
469 struct r600_command_buffer start_cs_cmd
; /* invariant state mostly */
470 /** Compute specific registers initializations. The start_cs_cmd atom
471 * must be emitted before start_compute_cs_cmd. */
472 struct r600_command_buffer start_compute_cs_cmd
;
473 /* Register states. */
474 struct r600_alphatest_state alphatest_state
;
475 struct r600_cso_state blend_state
;
476 struct r600_blend_color blend_color
;
477 struct r600_cb_misc_state cb_misc_state
;
478 struct r600_clip_misc_state clip_misc_state
;
479 struct r600_clip_state clip_state
;
480 struct r600_db_misc_state db_misc_state
;
481 struct r600_db_state db_state
;
482 struct r600_cso_state dsa_state
;
483 struct r600_framebuffer framebuffer
;
484 struct r600_poly_offset_state poly_offset_state
;
485 struct r600_cso_state rasterizer_state
;
486 struct r600_sample_mask sample_mask
;
487 struct r600_scissor_state scissor
;
488 struct r600_seamless_cube_map seamless_cube_map
;
489 struct r600_config_state config_state
;
490 struct r600_stencil_ref_state stencil_ref
;
491 struct r600_vgt_state vgt_state
;
492 struct r600_viewport_state viewport
;
493 /* Shaders and shader resources. */
494 struct r600_cso_state vertex_fetch_shader
;
495 struct r600_shader_state hw_shader_stages
[EG_NUM_HW_STAGES
];
496 struct r600_cs_shader_state cs_shader_state
;
497 struct r600_shader_stages_state shader_stages
;
498 struct r600_gs_rings_state gs_rings
;
499 struct r600_constbuf_state constbuf_state
[PIPE_SHADER_TYPES
];
500 struct r600_textures_info samplers
[PIPE_SHADER_TYPES
];
502 struct r600_shader_driver_constants_info driver_consts
[PIPE_SHADER_TYPES
];
504 /** Vertex buffers for fetch shaders */
505 struct r600_vertexbuf_state vertex_buffer_state
;
506 /** Vertex buffers for compute shaders */
507 struct r600_vertexbuf_state cs_vertex_buffer_state
;
509 /* Additional context states. */
510 unsigned compute_cb_target_mask
;
511 struct r600_pipe_shader_selector
*ps_shader
;
512 struct r600_pipe_shader_selector
*vs_shader
;
513 struct r600_pipe_shader_selector
*gs_shader
;
515 struct r600_pipe_shader_selector
*tcs_shader
;
516 struct r600_pipe_shader_selector
*tes_shader
;
518 struct r600_pipe_shader_selector
*fixed_func_tcs_shader
;
520 struct r600_rasterizer_state
*rasterizer
;
522 bool force_blend_disable
;
523 boolean dual_src_blend
;
528 struct pipe_index_buffer index_buffer
;
530 /* Last draw state (-1 = unset). */
531 int last_primitive_type
; /* Last primitive type used in draw_vbo. */
532 int last_start_instance
;
535 struct r600_isa
*isa
;
536 float sample_positions
[4 * 16];
538 bool tess_state_dirty
;
539 struct r600_pipe_shader_selector
*last_ls
;
540 struct r600_pipe_shader_selector
*last_tcs
;
541 unsigned last_num_tcs_input_cp
;
545 static inline void r600_emit_command_buffer(struct radeon_winsys_cs
*cs
,
546 struct r600_command_buffer
*cb
)
548 assert(cs
->cdw
+ cb
->num_dw
<= cs
->max_dw
);
549 memcpy(cs
->buf
+ cs
->cdw
, cb
->buf
, 4 * cb
->num_dw
);
550 cs
->cdw
+= cb
->num_dw
;
553 static inline void r600_set_atom_dirty(struct r600_context
*rctx
,
554 struct r600_atom
*atom
,
559 assert(atom
->id
!= 0);
560 assert(atom
->id
< sizeof(mask
) * 8);
561 mask
= 1ull << atom
->id
;
563 rctx
->dirty_atoms
|= mask
;
565 rctx
->dirty_atoms
&= ~mask
;
568 static inline void r600_mark_atom_dirty(struct r600_context
*rctx
,
569 struct r600_atom
*atom
)
571 r600_set_atom_dirty(rctx
, atom
, true);
574 void r600_trace_emit(struct r600_context
*rctx
);
576 static inline void r600_emit_atom(struct r600_context
*rctx
, struct r600_atom
*atom
)
578 atom
->emit(&rctx
->b
, atom
);
579 r600_set_atom_dirty(rctx
, atom
, false);
580 if (rctx
->screen
->b
.trace_bo
) {
581 r600_trace_emit(rctx
);
585 static inline void r600_set_cso_state(struct r600_context
*rctx
,
586 struct r600_cso_state
*state
, void *cso
)
589 r600_set_atom_dirty(rctx
, &state
->atom
, cso
!= NULL
);
592 static inline void r600_set_cso_state_with_cb(struct r600_context
*rctx
,
593 struct r600_cso_state
*state
, void *cso
,
594 struct r600_command_buffer
*cb
)
597 state
->atom
.num_dw
= cb
? cb
->num_dw
: 0;
598 r600_set_cso_state(rctx
, state
, cso
);
601 /* compute_memory_pool.c */
602 struct compute_memory_pool
;
603 void compute_memory_pool_delete(struct compute_memory_pool
* pool
);
604 struct compute_memory_pool
* compute_memory_pool_new(
605 struct r600_screen
*rscreen
);
607 /* evergreen_state.c */
608 struct pipe_sampler_view
*
609 evergreen_create_sampler_view_custom(struct pipe_context
*ctx
,
610 struct pipe_resource
*texture
,
611 const struct pipe_sampler_view
*state
,
612 unsigned width0
, unsigned height0
,
613 unsigned force_level
);
614 void evergreen_init_common_regs(struct r600_context
*ctx
,
615 struct r600_command_buffer
*cb
,
616 enum chip_class ctx_chip_class
,
617 enum radeon_family ctx_family
,
619 void cayman_init_common_regs(struct r600_command_buffer
*cb
,
620 enum chip_class ctx_chip_class
,
621 enum radeon_family ctx_family
,
624 void evergreen_init_state_functions(struct r600_context
*rctx
);
625 void evergreen_init_atom_start_cs(struct r600_context
*rctx
);
626 void evergreen_update_ps_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
627 void evergreen_update_es_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
628 void evergreen_update_gs_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
629 void evergreen_update_vs_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
630 void evergreen_update_ls_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
631 void evergreen_update_hs_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
632 void *evergreen_create_db_flush_dsa(struct r600_context
*rctx
);
633 void *evergreen_create_resolve_blend(struct r600_context
*rctx
);
634 void *evergreen_create_decompress_blend(struct r600_context
*rctx
);
635 void *evergreen_create_fastclear_blend(struct r600_context
*rctx
);
636 boolean
evergreen_is_format_supported(struct pipe_screen
*screen
,
637 enum pipe_format format
,
638 enum pipe_texture_target target
,
639 unsigned sample_count
,
641 void evergreen_init_color_surface(struct r600_context
*rctx
,
642 struct r600_surface
*surf
);
643 void evergreen_init_color_surface_rat(struct r600_context
*rctx
,
644 struct r600_surface
*surf
);
645 void evergreen_update_db_shader_control(struct r600_context
* rctx
);
646 bool evergreen_adjust_gprs(struct r600_context
*rctx
);
648 void r600_init_blit_functions(struct r600_context
*rctx
);
649 void r600_decompress_depth_textures(struct r600_context
*rctx
,
650 struct r600_samplerview_state
*textures
);
651 void r600_decompress_color_textures(struct r600_context
*rctx
,
652 struct r600_samplerview_state
*textures
);
653 void r600_resource_copy_region(struct pipe_context
*ctx
,
654 struct pipe_resource
*dst
,
656 unsigned dstx
, unsigned dsty
, unsigned dstz
,
657 struct pipe_resource
*src
,
659 const struct pipe_box
*src_box
);
662 int r600_pipe_shader_create(struct pipe_context
*ctx
,
663 struct r600_pipe_shader
*shader
,
664 union r600_shader_key key
);
666 void r600_pipe_shader_destroy(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
669 struct pipe_sampler_view
*
670 r600_create_sampler_view_custom(struct pipe_context
*ctx
,
671 struct pipe_resource
*texture
,
672 const struct pipe_sampler_view
*state
,
673 unsigned width_first_level
, unsigned height_first_level
);
674 void r600_init_state_functions(struct r600_context
*rctx
);
675 void r600_init_atom_start_cs(struct r600_context
*rctx
);
676 void r600_update_ps_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
677 void r600_update_es_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
678 void r600_update_gs_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
679 void r600_update_vs_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
680 void *r600_create_db_flush_dsa(struct r600_context
*rctx
);
681 void *r600_create_resolve_blend(struct r600_context
*rctx
);
682 void *r700_create_resolve_blend(struct r600_context
*rctx
);
683 void *r600_create_decompress_blend(struct r600_context
*rctx
);
684 bool r600_adjust_gprs(struct r600_context
*rctx
);
685 boolean
r600_is_format_supported(struct pipe_screen
*screen
,
686 enum pipe_format format
,
687 enum pipe_texture_target target
,
688 unsigned sample_count
,
690 void r600_update_db_shader_control(struct r600_context
* rctx
);
692 /* r600_hw_context.c */
693 void r600_context_gfx_flush(void *context
, unsigned flags
,
694 struct pipe_fence_handle
**fence
);
695 void r600_begin_new_cs(struct r600_context
*ctx
);
696 void r600_flush_emit(struct r600_context
*ctx
);
697 void r600_need_cs_space(struct r600_context
*ctx
, unsigned num_dw
, boolean count_draw_in
);
698 void r600_cp_dma_copy_buffer(struct r600_context
*rctx
,
699 struct pipe_resource
*dst
, uint64_t dst_offset
,
700 struct pipe_resource
*src
, uint64_t src_offset
,
702 void evergreen_cp_dma_clear_buffer(struct r600_context
*rctx
,
703 struct pipe_resource
*dst
, uint64_t offset
,
704 unsigned size
, uint32_t clear_value
);
705 void r600_dma_copy_buffer(struct r600_context
*rctx
,
706 struct pipe_resource
*dst
,
707 struct pipe_resource
*src
,
713 * evergreen_hw_context.c
715 void evergreen_dma_copy_buffer(struct r600_context
*rctx
,
716 struct pipe_resource
*dst
,
717 struct pipe_resource
*src
,
721 void evergreen_setup_tess_constants(struct r600_context
*rctx
,
722 const struct pipe_draw_info
*info
,
723 unsigned *num_patches
);
724 uint32_t evergreen_get_ls_hs_config(struct r600_context
*rctx
,
725 const struct pipe_draw_info
*info
,
726 unsigned num_patches
);
727 void evergreen_set_ls_hs_config(struct r600_context
*rctx
,
728 struct radeon_winsys_cs
*cs
,
729 uint32_t ls_hs_config
);
730 void evergreen_set_lds_alloc(struct r600_context
*rctx
,
731 struct radeon_winsys_cs
*cs
,
734 /* r600_state_common.c */
735 void r600_init_common_state_functions(struct r600_context
*rctx
);
736 void r600_emit_cso_state(struct r600_context
*rctx
, struct r600_atom
*atom
);
737 void r600_emit_alphatest_state(struct r600_context
*rctx
, struct r600_atom
*atom
);
738 void r600_emit_blend_color(struct r600_context
*rctx
, struct r600_atom
*atom
);
739 void r600_emit_vgt_state(struct r600_context
*rctx
, struct r600_atom
*atom
);
740 void r600_emit_clip_misc_state(struct r600_context
*rctx
, struct r600_atom
*atom
);
741 void r600_emit_stencil_ref(struct r600_context
*rctx
, struct r600_atom
*atom
);
742 void r600_emit_viewport_state(struct r600_context
*rctx
, struct r600_atom
*atom
);
743 void r600_emit_shader(struct r600_context
*rctx
, struct r600_atom
*a
);
744 void r600_add_atom(struct r600_context
*rctx
, struct r600_atom
*atom
, unsigned id
);
745 void r600_init_atom(struct r600_context
*rctx
, struct r600_atom
*atom
, unsigned id
,
746 void (*emit
)(struct r600_context
*ctx
, struct r600_atom
*state
),
748 void r600_vertex_buffers_dirty(struct r600_context
*rctx
);
749 void r600_sampler_views_dirty(struct r600_context
*rctx
,
750 struct r600_samplerview_state
*state
);
751 void r600_sampler_states_dirty(struct r600_context
*rctx
,
752 struct r600_sampler_states
*state
);
753 void r600_constant_buffers_dirty(struct r600_context
*rctx
, struct r600_constbuf_state
*state
);
754 void r600_set_sample_locations_constant_buffer(struct r600_context
*rctx
);
755 uint32_t r600_translate_stencil_op(int s_op
);
756 uint32_t r600_translate_fill(uint32_t func
);
757 unsigned r600_tex_wrap(unsigned wrap
);
758 unsigned r600_tex_filter(unsigned filter
);
759 unsigned r600_tex_mipfilter(unsigned filter
);
760 unsigned r600_tex_compare(unsigned compare
);
761 bool sampler_state_needs_border_color(const struct pipe_sampler_state
*state
);
762 struct pipe_surface
*r600_create_surface_custom(struct pipe_context
*pipe
,
763 struct pipe_resource
*texture
,
764 const struct pipe_surface
*templ
,
765 unsigned width
, unsigned height
);
766 unsigned r600_get_swizzle_combined(const unsigned char *swizzle_format
,
767 const unsigned char *swizzle_view
,
769 uint32_t r600_translate_texformat(struct pipe_screen
*screen
, enum pipe_format format
,
770 const unsigned char *swizzle_view
,
771 uint32_t *word4_p
, uint32_t *yuv_format_p
);
772 uint32_t r600_translate_colorformat(enum chip_class chip
, enum pipe_format format
);
773 uint32_t r600_colorformat_endian_swap(uint32_t colorformat
);
776 struct pipe_video_codec
*r600_uvd_create_decoder(struct pipe_context
*context
,
777 const struct pipe_video_codec
*decoder
);
779 struct pipe_video_buffer
*r600_video_buffer_create(struct pipe_context
*pipe
,
780 const struct pipe_video_buffer
*tmpl
);
783 * Helpers for building command buffers
786 #define PKT3_SET_CONFIG_REG 0x68
787 #define PKT3_SET_CONTEXT_REG 0x69
788 #define PKT3_SET_CTL_CONST 0x6F
789 #define PKT3_SET_LOOP_CONST 0x6C
791 #define R600_CONFIG_REG_OFFSET 0x08000
792 #define R600_CONTEXT_REG_OFFSET 0x28000
793 #define R600_CTL_CONST_OFFSET 0x3CFF0
794 #define R600_LOOP_CONST_OFFSET 0X0003E200
795 #define EG_LOOP_CONST_OFFSET 0x0003A200
797 #define PKT_TYPE_S(x) (((x) & 0x3) << 30)
798 #define PKT_COUNT_S(x) (((x) & 0x3FFF) << 16)
799 #define PKT3_IT_OPCODE_S(x) (((x) & 0xFF) << 8)
800 #define PKT3_PREDICATE(x) (((x) >> 0) & 0x1)
801 #define PKT3(op, count, predicate) (PKT_TYPE_S(3) | PKT_COUNT_S(count) | PKT3_IT_OPCODE_S(op) | PKT3_PREDICATE(predicate))
803 #define RADEON_CP_PACKET3_COMPUTE_MODE 0x00000002
805 /*Evergreen Compute packet3*/
806 #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)
808 static inline void r600_store_value(struct r600_command_buffer
*cb
, unsigned value
)
810 cb
->buf
[cb
->num_dw
++] = value
;
813 static inline void r600_store_array(struct r600_command_buffer
*cb
, unsigned num
, unsigned *ptr
)
815 assert(cb
->num_dw
+num
<= cb
->max_num_dw
);
816 memcpy(&cb
->buf
[cb
->num_dw
], ptr
, num
* sizeof(ptr
[0]));
820 static inline void r600_store_config_reg_seq(struct r600_command_buffer
*cb
, unsigned reg
, unsigned num
)
822 assert(reg
< R600_CONTEXT_REG_OFFSET
);
823 assert(cb
->num_dw
+2+num
<= cb
->max_num_dw
);
824 cb
->buf
[cb
->num_dw
++] = PKT3(PKT3_SET_CONFIG_REG
, num
, 0);
825 cb
->buf
[cb
->num_dw
++] = (reg
- R600_CONFIG_REG_OFFSET
) >> 2;
829 * Needs cb->pkt_flags set to RADEON_CP_PACKET3_COMPUTE_MODE for compute
832 static inline void r600_store_context_reg_seq(struct r600_command_buffer
*cb
, unsigned reg
, unsigned num
)
834 assert(reg
>= R600_CONTEXT_REG_OFFSET
&& 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_CONTEXT_REG
, num
, 0) | cb
->pkt_flags
;
837 cb
->buf
[cb
->num_dw
++] = (reg
- R600_CONTEXT_REG_OFFSET
) >> 2;
841 * Needs cb->pkt_flags set to RADEON_CP_PACKET3_COMPUTE_MODE for compute
844 static inline void r600_store_ctl_const_seq(struct r600_command_buffer
*cb
, unsigned reg
, unsigned num
)
846 assert(reg
>= R600_CTL_CONST_OFFSET
);
847 assert(cb
->num_dw
+2+num
<= cb
->max_num_dw
);
848 cb
->buf
[cb
->num_dw
++] = PKT3(PKT3_SET_CTL_CONST
, num
, 0) | cb
->pkt_flags
;
849 cb
->buf
[cb
->num_dw
++] = (reg
- R600_CTL_CONST_OFFSET
) >> 2;
852 static inline void r600_store_loop_const_seq(struct r600_command_buffer
*cb
, unsigned reg
, unsigned num
)
854 assert(reg
>= R600_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);
857 cb
->buf
[cb
->num_dw
++] = (reg
- R600_LOOP_CONST_OFFSET
) >> 2;
861 * Needs cb->pkt_flags set to RADEON_CP_PACKET3_COMPUTE_MODE for compute
864 static inline void eg_store_loop_const_seq(struct r600_command_buffer
*cb
, unsigned reg
, unsigned num
)
866 assert(reg
>= EG_LOOP_CONST_OFFSET
);
867 assert(cb
->num_dw
+2+num
<= cb
->max_num_dw
);
868 cb
->buf
[cb
->num_dw
++] = PKT3(PKT3_SET_LOOP_CONST
, num
, 0) | cb
->pkt_flags
;
869 cb
->buf
[cb
->num_dw
++] = (reg
- EG_LOOP_CONST_OFFSET
) >> 2;
872 static inline void r600_store_config_reg(struct r600_command_buffer
*cb
, unsigned reg
, unsigned value
)
874 r600_store_config_reg_seq(cb
, reg
, 1);
875 r600_store_value(cb
, value
);
878 static inline void r600_store_context_reg(struct r600_command_buffer
*cb
, unsigned reg
, unsigned value
)
880 r600_store_context_reg_seq(cb
, reg
, 1);
881 r600_store_value(cb
, value
);
884 static inline void r600_store_ctl_const(struct r600_command_buffer
*cb
, unsigned reg
, unsigned value
)
886 r600_store_ctl_const_seq(cb
, reg
, 1);
887 r600_store_value(cb
, value
);
890 static inline void r600_store_loop_const(struct r600_command_buffer
*cb
, unsigned reg
, unsigned value
)
892 r600_store_loop_const_seq(cb
, reg
, 1);
893 r600_store_value(cb
, value
);
896 static inline void eg_store_loop_const(struct r600_command_buffer
*cb
, unsigned reg
, unsigned value
)
898 eg_store_loop_const_seq(cb
, reg
, 1);
899 r600_store_value(cb
, value
);
902 void r600_init_command_buffer(struct r600_command_buffer
*cb
, unsigned num_dw
);
903 void r600_release_command_buffer(struct r600_command_buffer
*cb
);
905 static inline void radeon_compute_set_context_reg_seq(struct radeon_winsys_cs
*cs
, unsigned reg
, unsigned num
)
907 radeon_set_context_reg_seq(cs
, reg
, num
);
908 /* Set the compute bit on the packet header */
909 cs
->buf
[cs
->cdw
- 2] |= RADEON_CP_PACKET3_COMPUTE_MODE
;
912 static inline void radeon_set_ctl_const_seq(struct radeon_winsys_cs
*cs
, unsigned reg
, unsigned num
)
914 assert(reg
>= R600_CTL_CONST_OFFSET
);
915 assert(cs
->cdw
+2+num
<= cs
->max_dw
);
916 cs
->buf
[cs
->cdw
++] = PKT3(PKT3_SET_CTL_CONST
, num
, 0);
917 cs
->buf
[cs
->cdw
++] = (reg
- R600_CTL_CONST_OFFSET
) >> 2;
920 static inline void radeon_compute_set_context_reg(struct radeon_winsys_cs
*cs
, unsigned reg
, unsigned value
)
922 radeon_compute_set_context_reg_seq(cs
, reg
, 1);
923 radeon_emit(cs
, value
);
926 static inline void radeon_set_context_reg_flag(struct radeon_winsys_cs
*cs
, unsigned reg
, unsigned value
, unsigned flag
)
928 if (flag
& RADEON_CP_PACKET3_COMPUTE_MODE
) {
929 radeon_compute_set_context_reg(cs
, reg
, value
);
931 radeon_set_context_reg(cs
, reg
, value
);
935 static inline void radeon_set_ctl_const(struct radeon_winsys_cs
*cs
, unsigned reg
, unsigned value
)
937 radeon_set_ctl_const_seq(cs
, reg
, 1);
938 radeon_emit(cs
, value
);
944 static inline uint32_t S_FIXED(float value
, uint32_t frac_bits
)
946 return value
* (1 << frac_bits
);
948 #define ALIGN_DIVUP(x, y) (((x) + (y) - 1) / (y))
950 /* 12.4 fixed-point */
951 static inline unsigned r600_pack_float_12p4(float x
)
954 x
>= 4096 ? 0xffff : x
* 16;
957 /* Return if the depth format can be read without the DB->CB copy on r6xx-r7xx. */
958 static inline bool r600_can_read_depth(struct r600_texture
*rtex
)
960 return rtex
->resource
.b
.b
.nr_samples
<= 1 &&
961 (rtex
->resource
.b
.b
.format
== PIPE_FORMAT_Z16_UNORM
||
962 rtex
->resource
.b
.b
.format
== PIPE_FORMAT_Z32_FLOAT
);
965 #define V_028A6C_OUTPRIM_TYPE_POINTLIST 0
966 #define V_028A6C_OUTPRIM_TYPE_LINESTRIP 1
967 #define V_028A6C_OUTPRIM_TYPE_TRISTRIP 2
969 unsigned r600_conv_prim_to_gs_out(unsigned mode
);