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 45
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 47
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_UCP_CONST_BUFFER (R600_MAX_USER_CONST_BUFFERS)
71 #define R600_BUFFER_INFO_CONST_BUFFER (R600_MAX_USER_CONST_BUFFERS + 1)
72 #define R600_GS_RING_CONST_BUFFER (R600_MAX_USER_CONST_BUFFERS + 2)
73 /* Currently R600_MAX_CONST_BUFFERS just fits on the hw, which has a limit
74 * of 16 const buffers.
75 * UCP/SAMPLE_POSITIONS are never accessed by same shader stage so they can use the same id.
77 * In order to support d3d 11 mandated minimum of 15 user const buffers
78 * we'd have to squash all use cases into one driver buffer.
80 #define R600_SAMPLE_POSITIONS_CONST_BUFFER (R600_MAX_USER_CONST_BUFFERS)
82 #define R600_MAX_CONST_BUFFER_SIZE (4096 * sizeof(float[4]))
84 #ifdef PIPE_ARCH_BIG_ENDIAN
85 #define R600_BIG_ENDIAN 1
87 #define R600_BIG_ENDIAN 0
90 #define R600_DIRTY_ATOM_WORD_BITS (sizeof(unsigned long) * 8)
91 #define R600_DIRTY_ATOM_ARRAY_LEN DIV_ROUND_UP(R600_NUM_ATOMS, R600_DIRTY_ATOM_WORD_BITS)
95 union r600_shader_key
;
97 /* This is an atom containing GPU commands that never change.
98 * This is supposed to be copied directly into the CS. */
99 struct r600_command_buffer
{
106 struct r600_db_state
{
107 struct r600_atom atom
;
108 struct r600_surface
*rsurf
;
111 struct r600_db_misc_state
{
112 struct r600_atom atom
;
113 bool occlusion_query_enabled
;
114 bool flush_depthstencil_through_cb
;
115 bool flush_depthstencil_in_place
;
116 bool copy_depth
, copy_stencil
;
117 unsigned copy_sample
;
118 unsigned log_samples
;
119 unsigned db_shader_control
;
123 struct r600_cb_misc_state
{
124 struct r600_atom atom
;
125 unsigned cb_color_control
; /* this comes from blend state */
126 unsigned blend_colormask
; /* 8*4 bits for 8 RGBA colorbuffers */
128 unsigned nr_ps_color_outputs
;
133 struct r600_clip_misc_state
{
134 struct r600_atom atom
;
135 unsigned pa_cl_clip_cntl
; /* from rasterizer */
136 unsigned pa_cl_vs_out_cntl
; /* from vertex shader */
137 unsigned clip_plane_enable
; /* from rasterizer */
138 unsigned clip_dist_write
; /* from vertex shader */
139 boolean clip_disable
; /* from vertex shader */
142 struct r600_alphatest_state
{
143 struct r600_atom atom
;
144 unsigned sx_alpha_test_control
; /* this comes from dsa state */
145 unsigned sx_alpha_ref
; /* this comes from dsa state */
147 bool cb0_export_16bpc
; /* from set_framebuffer_state */
150 struct r600_vgt_state
{
151 struct r600_atom atom
;
152 uint32_t vgt_multi_prim_ib_reset_en
;
153 uint32_t vgt_multi_prim_ib_reset_indx
;
154 uint32_t vgt_indx_offset
;
155 bool last_draw_was_indirect
;
158 struct r600_blend_color
{
159 struct r600_atom atom
;
160 struct pipe_blend_color state
;
163 struct r600_clip_state
{
164 struct r600_atom atom
;
165 struct pipe_clip_state state
;
168 struct r600_cs_shader_state
{
169 struct r600_atom atom
;
170 unsigned kernel_index
;
172 struct r600_pipe_compute
*shader
;
175 struct r600_framebuffer
{
176 struct r600_atom atom
;
177 struct pipe_framebuffer_state state
;
178 unsigned compressed_cb_mask
;
182 bool is_msaa_resolve
;
185 struct r600_sample_mask
{
186 struct r600_atom atom
;
187 uint16_t sample_mask
; /* there are only 8 bits on EG, 16 bits on Cayman */
190 struct r600_config_state
{
191 struct r600_atom atom
;
192 unsigned sq_gpr_resource_mgmt_1
;
193 unsigned sq_gpr_resource_mgmt_2
;
196 struct r600_stencil_ref
203 struct r600_stencil_ref_state
{
204 struct r600_atom atom
;
205 struct r600_stencil_ref state
;
206 struct pipe_stencil_ref pipe_state
;
209 struct r600_viewport_state
{
210 struct r600_atom atom
;
211 struct pipe_viewport_state state
[R600_MAX_VIEWPORTS
];
215 struct r600_shader_stages_state
{
216 struct r600_atom atom
;
217 unsigned geom_enable
;
220 struct r600_gs_rings_state
{
221 struct r600_atom atom
;
223 struct pipe_constant_buffer esgs_ring
;
224 struct pipe_constant_buffer gsvs_ring
;
227 /* This must start from 16. */
229 #define DBG_LLVM (1 << 29)
230 #define DBG_NO_CP_DMA (1 << 30)
232 #define DBG_NO_SB (1 << 21)
233 #define DBG_SB_CS (1 << 22)
234 #define DBG_SB_DRY_RUN (1 << 23)
235 #define DBG_SB_STAT (1 << 24)
236 #define DBG_SB_DUMP (1 << 25)
237 #define DBG_SB_NO_FALLBACK (1 << 26)
238 #define DBG_SB_DISASM (1 << 27)
239 #define DBG_SB_SAFEMATH (1 << 28)
242 struct r600_common_screen b
;
244 bool has_compressed_msaa_texturing
;
246 /*for compute global memory binding, we allocate stuff here, instead of
248 * XXX: Not sure if this is the best place for global_pool. Also,
249 * it's not thread safe, so it won't work with multiple contexts. */
250 struct compute_memory_pool
*global_pool
;
253 struct r600_pipe_sampler_view
{
254 struct pipe_sampler_view base
;
255 struct list_head list
;
256 struct r600_resource
*tex_resource
;
257 uint32_t tex_resource_words
[8];
258 bool skip_mip_address_reloc
;
261 struct r600_rasterizer_state
{
262 struct r600_command_buffer buffer
;
265 unsigned sprite_coord_enable
;
266 unsigned clip_plane_enable
;
267 unsigned pa_sc_line_stipple
;
268 unsigned pa_cl_clip_cntl
;
269 unsigned pa_su_sc_mode_cntl
;
274 bool multisample_enable
;
277 struct r600_poly_offset_state
{
278 struct r600_atom atom
;
279 enum pipe_format zs_format
;
284 struct r600_blend_state
{
285 struct r600_command_buffer buffer
;
286 struct r600_command_buffer buffer_no_blend
;
287 unsigned cb_target_mask
;
288 unsigned cb_color_control
;
289 unsigned cb_color_control_no_blend
;
294 struct r600_dsa_state
{
295 struct r600_command_buffer buffer
;
300 unsigned sx_alpha_test_control
;
303 struct r600_pipe_shader
;
305 struct r600_pipe_shader_selector
{
306 struct r600_pipe_shader
*current
;
308 struct tgsi_token
*tokens
;
309 struct pipe_stream_output_info so
;
310 struct tgsi_shader_info info
;
312 unsigned num_shaders
;
314 /* PIPE_SHADER_[VERTEX|FRAGMENT|...] */
317 /* geometry shader properties */
318 unsigned gs_output_prim
;
319 unsigned gs_max_out_vertices
;
320 unsigned gs_num_invocations
;
322 unsigned nr_ps_max_color_exports
;
325 struct r600_pipe_sampler_state
{
326 uint32_t tex_sampler_words
[3];
327 union pipe_color_union border_color
;
328 bool border_color_use
;
329 bool seamless_cube_map
;
332 /* needed for blitter save */
333 #define NUM_TEX_UNITS 16
335 struct r600_seamless_cube_map
{
336 struct r600_atom atom
;
340 struct r600_samplerview_state
{
341 struct r600_atom atom
;
342 struct r600_pipe_sampler_view
*views
[NUM_TEX_UNITS
];
343 uint32_t enabled_mask
;
345 uint32_t compressed_depthtex_mask
; /* which textures are depth */
346 uint32_t compressed_colortex_mask
;
347 boolean dirty_buffer_constants
;
350 struct r600_sampler_states
{
351 struct r600_atom atom
;
352 struct r600_pipe_sampler_state
*states
[NUM_TEX_UNITS
];
353 uint32_t enabled_mask
;
355 uint32_t has_bordercolor_mask
; /* which states contain the border color */
358 struct r600_textures_info
{
359 struct r600_samplerview_state views
;
360 struct r600_sampler_states states
;
361 bool is_array_sampler
[NUM_TEX_UNITS
];
363 /* cube array txq workaround */
364 uint32_t *txq_constants
;
365 /* buffer related workarounds */
366 uint32_t *buffer_constants
;
369 struct r600_constbuf_state
371 struct r600_atom atom
;
372 struct pipe_constant_buffer cb
[PIPE_MAX_CONSTANT_BUFFERS
];
373 uint32_t enabled_mask
;
377 struct r600_vertexbuf_state
379 struct r600_atom atom
;
380 struct pipe_vertex_buffer vb
[PIPE_MAX_ATTRIBS
];
381 uint32_t enabled_mask
; /* non-NULL buffers */
385 /* CSO (constant state object, in other words, immutable state). */
386 struct r600_cso_state
388 struct r600_atom atom
;
389 void *cso
; /* e.g. r600_blend_state */
390 struct r600_command_buffer
*cb
;
393 struct r600_scissor_state
395 struct r600_atom atom
;
396 struct pipe_scissor_state scissor
[R600_MAX_VIEWPORTS
];
398 bool enable
; /* r6xx only */
401 struct r600_fetch_shader
{
402 struct r600_resource
*buffer
;
406 struct r600_shader_state
{
407 struct r600_atom atom
;
408 struct r600_pipe_shader
*shader
;
411 struct r600_context
{
412 struct r600_common_context b
;
413 struct r600_screen
*screen
;
414 struct blitter_context
*blitter
;
415 struct u_suballocator
*allocator_fetch_shader
;
418 boolean has_vertex_cache
;
419 boolean keep_tiling_flags
;
420 unsigned default_ps_gprs
, default_vs_gprs
;
421 unsigned r6xx_num_clause_temp_gprs
;
423 /* Miscellaneous state objects. */
424 void *custom_dsa_flush
;
425 void *custom_blend_resolve
;
426 void *custom_blend_decompress
;
427 void *custom_blend_fastclear
;
428 /* With rasterizer discard, there doesn't have to be a pixel shader.
429 * In that case, we bind this one: */
430 void *dummy_pixel_shader
;
431 /* These dummy CMASK and FMASK buffers are used to get around the R6xx hardware
432 * bug where valid CMASK and FMASK are required to be present to avoid
433 * a hardlock in certain operations but aren't actually used
434 * for anything useful. */
435 struct r600_resource
*dummy_fmask
;
436 struct r600_resource
*dummy_cmask
;
438 /* State binding slots are here. */
439 struct r600_atom
*atoms
[R600_NUM_ATOMS
];
440 /* Dirty atom bitmask for fast tests */
441 unsigned long dirty_atoms
[R600_DIRTY_ATOM_ARRAY_LEN
];
442 /* States for CS initialization. */
443 struct r600_command_buffer start_cs_cmd
; /* invariant state mostly */
444 /** Compute specific registers initializations. The start_cs_cmd atom
445 * must be emitted before start_compute_cs_cmd. */
446 struct r600_command_buffer start_compute_cs_cmd
;
447 /* Register states. */
448 struct r600_alphatest_state alphatest_state
;
449 struct r600_cso_state blend_state
;
450 struct r600_blend_color blend_color
;
451 struct r600_cb_misc_state cb_misc_state
;
452 struct r600_clip_misc_state clip_misc_state
;
453 struct r600_clip_state clip_state
;
454 struct r600_db_misc_state db_misc_state
;
455 struct r600_db_state db_state
;
456 struct r600_cso_state dsa_state
;
457 struct r600_framebuffer framebuffer
;
458 struct r600_poly_offset_state poly_offset_state
;
459 struct r600_cso_state rasterizer_state
;
460 struct r600_sample_mask sample_mask
;
461 struct r600_scissor_state scissor
;
462 struct r600_seamless_cube_map seamless_cube_map
;
463 struct r600_config_state config_state
;
464 struct r600_stencil_ref_state stencil_ref
;
465 struct r600_vgt_state vgt_state
;
466 struct r600_viewport_state viewport
;
467 /* Shaders and shader resources. */
468 struct r600_cso_state vertex_fetch_shader
;
469 struct r600_shader_state vertex_shader
;
470 struct r600_shader_state pixel_shader
;
471 struct r600_shader_state geometry_shader
;
472 struct r600_shader_state export_shader
;
473 struct r600_cs_shader_state cs_shader_state
;
474 struct r600_shader_stages_state shader_stages
;
475 struct r600_gs_rings_state gs_rings
;
476 struct r600_constbuf_state constbuf_state
[PIPE_SHADER_TYPES
];
477 struct r600_textures_info samplers
[PIPE_SHADER_TYPES
];
478 /** Vertex buffers for fetch shaders */
479 struct r600_vertexbuf_state vertex_buffer_state
;
480 /** Vertex buffers for compute shaders */
481 struct r600_vertexbuf_state cs_vertex_buffer_state
;
483 /* Additional context states. */
484 unsigned compute_cb_target_mask
;
485 struct r600_pipe_shader_selector
*ps_shader
;
486 struct r600_pipe_shader_selector
*vs_shader
;
487 struct r600_pipe_shader_selector
*gs_shader
;
488 struct r600_rasterizer_state
*rasterizer
;
490 bool force_blend_disable
;
491 boolean dual_src_blend
;
496 struct pipe_index_buffer index_buffer
;
498 /* Last draw state (-1 = unset). */
499 int last_primitive_type
; /* Last primitive type used in draw_vbo. */
500 int last_start_instance
;
503 struct r600_isa
*isa
;
506 static inline void r600_emit_command_buffer(struct radeon_winsys_cs
*cs
,
507 struct r600_command_buffer
*cb
)
509 assert(cs
->cdw
+ cb
->num_dw
<= cs
->max_dw
);
510 memcpy(cs
->buf
+ cs
->cdw
, cb
->buf
, 4 * cb
->num_dw
);
511 cs
->cdw
+= cb
->num_dw
;
514 static inline void r600_set_atom_dirty(struct r600_context
*rctx
,
515 struct r600_atom
*atom
,
523 assert(atom
->id
!= 0);
524 w
= atom
->id
/ R600_DIRTY_ATOM_WORD_BITS
;
525 mask
= 1ul << (atom
->id
% R600_DIRTY_ATOM_WORD_BITS
);
527 rctx
->dirty_atoms
[w
] |= mask
;
529 rctx
->dirty_atoms
[w
] &= ~mask
;
532 static inline void r600_mark_atom_dirty(struct r600_context
*rctx
,
533 struct r600_atom
*atom
)
535 r600_set_atom_dirty(rctx
, atom
, true);
538 static inline unsigned int r600_next_dirty_atom(struct r600_context
*rctx
,
541 #if !defined(DEBUG) && defined(HAVE___BUILTIN_CTZ)
542 unsigned int w
= id
/ R600_DIRTY_ATOM_WORD_BITS
;
543 unsigned int bit
= id
% R600_DIRTY_ATOM_WORD_BITS
;
544 unsigned long bits
, mask
= (1ul << bit
) - 1;
546 for (; w
< R600_DIRTY_ATOM_ARRAY_LEN
; w
++, mask
= 0ul) {
547 bits
= rctx
->dirty_atoms
[w
] & ~mask
;
550 return w
* R600_DIRTY_ATOM_WORD_BITS
+ __builtin_ctzl(bits
);
553 return R600_NUM_ATOMS
;
555 for (; id
< R600_NUM_ATOMS
; id
++) {
556 bool dirty
= !!(rctx
->dirty_atoms
[id
/ R600_DIRTY_ATOM_WORD_BITS
] &
557 (1ul << (id
% R600_DIRTY_ATOM_WORD_BITS
)));
558 assert(dirty
== (rctx
->atoms
[id
] && rctx
->atoms
[id
]->dirty
));
567 void r600_trace_emit(struct r600_context
*rctx
);
569 static inline void r600_emit_atom(struct r600_context
*rctx
, struct r600_atom
*atom
)
571 atom
->emit(&rctx
->b
, atom
);
572 r600_set_atom_dirty(rctx
, atom
, false);
573 if (rctx
->screen
->b
.trace_bo
) {
574 r600_trace_emit(rctx
);
578 static inline void r600_set_cso_state(struct r600_context
*rctx
,
579 struct r600_cso_state
*state
, void *cso
)
582 r600_set_atom_dirty(rctx
, &state
->atom
, cso
!= NULL
);
585 static inline void r600_set_cso_state_with_cb(struct r600_context
*rctx
,
586 struct r600_cso_state
*state
, void *cso
,
587 struct r600_command_buffer
*cb
)
590 state
->atom
.num_dw
= cb
? cb
->num_dw
: 0;
591 r600_set_cso_state(rctx
, state
, cso
);
594 /* compute_memory_pool.c */
595 struct compute_memory_pool
;
596 void compute_memory_pool_delete(struct compute_memory_pool
* pool
);
597 struct compute_memory_pool
* compute_memory_pool_new(
598 struct r600_screen
*rscreen
);
600 /* evergreen_state.c */
601 struct pipe_sampler_view
*
602 evergreen_create_sampler_view_custom(struct pipe_context
*ctx
,
603 struct pipe_resource
*texture
,
604 const struct pipe_sampler_view
*state
,
605 unsigned width0
, unsigned height0
,
606 unsigned force_level
);
607 void evergreen_init_common_regs(struct r600_command_buffer
*cb
,
608 enum chip_class ctx_chip_class
,
609 enum radeon_family ctx_family
,
611 void cayman_init_common_regs(struct r600_command_buffer
*cb
,
612 enum chip_class ctx_chip_class
,
613 enum radeon_family ctx_family
,
616 void evergreen_init_state_functions(struct r600_context
*rctx
);
617 void evergreen_init_atom_start_cs(struct r600_context
*rctx
);
618 void evergreen_update_ps_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
619 void evergreen_update_es_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
620 void evergreen_update_gs_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
621 void evergreen_update_vs_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
622 void *evergreen_create_db_flush_dsa(struct r600_context
*rctx
);
623 void *evergreen_create_resolve_blend(struct r600_context
*rctx
);
624 void *evergreen_create_decompress_blend(struct r600_context
*rctx
);
625 void *evergreen_create_fastclear_blend(struct r600_context
*rctx
);
626 boolean
evergreen_is_format_supported(struct pipe_screen
*screen
,
627 enum pipe_format format
,
628 enum pipe_texture_target target
,
629 unsigned sample_count
,
631 void evergreen_init_color_surface(struct r600_context
*rctx
,
632 struct r600_surface
*surf
);
633 void evergreen_init_color_surface_rat(struct r600_context
*rctx
,
634 struct r600_surface
*surf
);
635 void evergreen_update_db_shader_control(struct r600_context
* rctx
);
638 void r600_init_blit_functions(struct r600_context
*rctx
);
639 void r600_decompress_depth_textures(struct r600_context
*rctx
,
640 struct r600_samplerview_state
*textures
);
641 void r600_decompress_color_textures(struct r600_context
*rctx
,
642 struct r600_samplerview_state
*textures
);
643 void r600_resource_copy_region(struct pipe_context
*ctx
,
644 struct pipe_resource
*dst
,
646 unsigned dstx
, unsigned dsty
, unsigned dstz
,
647 struct pipe_resource
*src
,
649 const struct pipe_box
*src_box
);
652 int r600_pipe_shader_create(struct pipe_context
*ctx
,
653 struct r600_pipe_shader
*shader
,
654 union r600_shader_key key
);
656 void r600_pipe_shader_destroy(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
659 struct pipe_sampler_view
*
660 r600_create_sampler_view_custom(struct pipe_context
*ctx
,
661 struct pipe_resource
*texture
,
662 const struct pipe_sampler_view
*state
,
663 unsigned width_first_level
, unsigned height_first_level
);
664 void r600_init_state_functions(struct r600_context
*rctx
);
665 void r600_init_atom_start_cs(struct r600_context
*rctx
);
666 void r600_update_ps_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
667 void r600_update_es_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
668 void r600_update_gs_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
669 void r600_update_vs_state(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
670 void *r600_create_db_flush_dsa(struct r600_context
*rctx
);
671 void *r600_create_resolve_blend(struct r600_context
*rctx
);
672 void *r700_create_resolve_blend(struct r600_context
*rctx
);
673 void *r600_create_decompress_blend(struct r600_context
*rctx
);
674 bool r600_adjust_gprs(struct r600_context
*rctx
);
675 boolean
r600_is_format_supported(struct pipe_screen
*screen
,
676 enum pipe_format format
,
677 enum pipe_texture_target target
,
678 unsigned sample_count
,
680 void r600_update_db_shader_control(struct r600_context
* rctx
);
682 /* r600_hw_context.c */
683 void r600_context_gfx_flush(void *context
, unsigned flags
,
684 struct pipe_fence_handle
**fence
);
685 void r600_begin_new_cs(struct r600_context
*ctx
);
686 void r600_flush_emit(struct r600_context
*ctx
);
687 void r600_need_cs_space(struct r600_context
*ctx
, unsigned num_dw
, boolean count_draw_in
);
688 void r600_cp_dma_copy_buffer(struct r600_context
*rctx
,
689 struct pipe_resource
*dst
, uint64_t dst_offset
,
690 struct pipe_resource
*src
, uint64_t src_offset
,
692 void evergreen_cp_dma_clear_buffer(struct r600_context
*rctx
,
693 struct pipe_resource
*dst
, uint64_t offset
,
694 unsigned size
, uint32_t clear_value
);
695 void r600_dma_copy_buffer(struct r600_context
*rctx
,
696 struct pipe_resource
*dst
,
697 struct pipe_resource
*src
,
703 * evergreen_hw_context.c
705 void evergreen_dma_copy_buffer(struct r600_context
*rctx
,
706 struct pipe_resource
*dst
,
707 struct pipe_resource
*src
,
712 /* r600_state_common.c */
713 void r600_init_common_state_functions(struct r600_context
*rctx
);
714 void r600_emit_cso_state(struct r600_context
*rctx
, struct r600_atom
*atom
);
715 void r600_emit_alphatest_state(struct r600_context
*rctx
, struct r600_atom
*atom
);
716 void r600_emit_blend_color(struct r600_context
*rctx
, struct r600_atom
*atom
);
717 void r600_emit_vgt_state(struct r600_context
*rctx
, struct r600_atom
*atom
);
718 void r600_emit_clip_misc_state(struct r600_context
*rctx
, struct r600_atom
*atom
);
719 void r600_emit_stencil_ref(struct r600_context
*rctx
, struct r600_atom
*atom
);
720 void r600_emit_viewport_state(struct r600_context
*rctx
, struct r600_atom
*atom
);
721 void r600_emit_shader(struct r600_context
*rctx
, struct r600_atom
*a
);
722 void r600_add_atom(struct r600_context
*rctx
, struct r600_atom
*atom
, unsigned id
);
723 void r600_init_atom(struct r600_context
*rctx
, struct r600_atom
*atom
, unsigned id
,
724 void (*emit
)(struct r600_context
*ctx
, struct r600_atom
*state
),
726 void r600_vertex_buffers_dirty(struct r600_context
*rctx
);
727 void r600_sampler_views_dirty(struct r600_context
*rctx
,
728 struct r600_samplerview_state
*state
);
729 void r600_sampler_states_dirty(struct r600_context
*rctx
,
730 struct r600_sampler_states
*state
);
731 void r600_constant_buffers_dirty(struct r600_context
*rctx
, struct r600_constbuf_state
*state
);
732 void r600_set_sample_locations_constant_buffer(struct r600_context
*rctx
);
733 uint32_t r600_translate_stencil_op(int s_op
);
734 uint32_t r600_translate_fill(uint32_t func
);
735 unsigned r600_tex_wrap(unsigned wrap
);
736 unsigned r600_tex_filter(unsigned filter
);
737 unsigned r600_tex_mipfilter(unsigned filter
);
738 unsigned r600_tex_compare(unsigned compare
);
739 bool sampler_state_needs_border_color(const struct pipe_sampler_state
*state
);
740 struct pipe_surface
*r600_create_surface_custom(struct pipe_context
*pipe
,
741 struct pipe_resource
*texture
,
742 const struct pipe_surface
*templ
,
743 unsigned width
, unsigned height
);
744 unsigned r600_get_swizzle_combined(const unsigned char *swizzle_format
,
745 const unsigned char *swizzle_view
,
747 uint32_t r600_translate_texformat(struct pipe_screen
*screen
, enum pipe_format format
,
748 const unsigned char *swizzle_view
,
749 uint32_t *word4_p
, uint32_t *yuv_format_p
);
750 uint32_t r600_translate_colorformat(enum chip_class chip
, enum pipe_format format
);
751 uint32_t r600_colorformat_endian_swap(uint32_t colorformat
);
754 struct pipe_video_codec
*r600_uvd_create_decoder(struct pipe_context
*context
,
755 const struct pipe_video_codec
*decoder
);
757 struct pipe_video_buffer
*r600_video_buffer_create(struct pipe_context
*pipe
,
758 const struct pipe_video_buffer
*tmpl
);
761 * Helpers for building command buffers
764 #define PKT3_SET_CONFIG_REG 0x68
765 #define PKT3_SET_CONTEXT_REG 0x69
766 #define PKT3_SET_CTL_CONST 0x6F
767 #define PKT3_SET_LOOP_CONST 0x6C
769 #define R600_CONFIG_REG_OFFSET 0x08000
770 #define R600_CONTEXT_REG_OFFSET 0x28000
771 #define R600_CTL_CONST_OFFSET 0x3CFF0
772 #define R600_LOOP_CONST_OFFSET 0X0003E200
773 #define EG_LOOP_CONST_OFFSET 0x0003A200
775 #define PKT_TYPE_S(x) (((x) & 0x3) << 30)
776 #define PKT_COUNT_S(x) (((x) & 0x3FFF) << 16)
777 #define PKT3_IT_OPCODE_S(x) (((x) & 0xFF) << 8)
778 #define PKT3_PREDICATE(x) (((x) >> 0) & 0x1)
779 #define PKT3(op, count, predicate) (PKT_TYPE_S(3) | PKT_COUNT_S(count) | PKT3_IT_OPCODE_S(op) | PKT3_PREDICATE(predicate))
781 #define RADEON_CP_PACKET3_COMPUTE_MODE 0x00000002
783 /*Evergreen Compute packet3*/
784 #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)
786 static inline void r600_store_value(struct r600_command_buffer
*cb
, unsigned value
)
788 cb
->buf
[cb
->num_dw
++] = value
;
791 static inline void r600_store_array(struct r600_command_buffer
*cb
, unsigned num
, unsigned *ptr
)
793 assert(cb
->num_dw
+num
<= cb
->max_num_dw
);
794 memcpy(&cb
->buf
[cb
->num_dw
], ptr
, num
* sizeof(ptr
[0]));
798 static inline void r600_store_config_reg_seq(struct r600_command_buffer
*cb
, unsigned reg
, unsigned num
)
800 assert(reg
< R600_CONTEXT_REG_OFFSET
);
801 assert(cb
->num_dw
+2+num
<= cb
->max_num_dw
);
802 cb
->buf
[cb
->num_dw
++] = PKT3(PKT3_SET_CONFIG_REG
, num
, 0);
803 cb
->buf
[cb
->num_dw
++] = (reg
- R600_CONFIG_REG_OFFSET
) >> 2;
807 * Needs cb->pkt_flags set to RADEON_CP_PACKET3_COMPUTE_MODE for compute
810 static inline void r600_store_context_reg_seq(struct r600_command_buffer
*cb
, unsigned reg
, unsigned num
)
812 assert(reg
>= R600_CONTEXT_REG_OFFSET
&& reg
< R600_CTL_CONST_OFFSET
);
813 assert(cb
->num_dw
+2+num
<= cb
->max_num_dw
);
814 cb
->buf
[cb
->num_dw
++] = PKT3(PKT3_SET_CONTEXT_REG
, num
, 0) | cb
->pkt_flags
;
815 cb
->buf
[cb
->num_dw
++] = (reg
- R600_CONTEXT_REG_OFFSET
) >> 2;
819 * Needs cb->pkt_flags set to RADEON_CP_PACKET3_COMPUTE_MODE for compute
822 static inline void r600_store_ctl_const_seq(struct r600_command_buffer
*cb
, unsigned reg
, unsigned num
)
824 assert(reg
>= R600_CTL_CONST_OFFSET
);
825 assert(cb
->num_dw
+2+num
<= cb
->max_num_dw
);
826 cb
->buf
[cb
->num_dw
++] = PKT3(PKT3_SET_CTL_CONST
, num
, 0) | cb
->pkt_flags
;
827 cb
->buf
[cb
->num_dw
++] = (reg
- R600_CTL_CONST_OFFSET
) >> 2;
830 static inline void r600_store_loop_const_seq(struct r600_command_buffer
*cb
, unsigned reg
, unsigned num
)
832 assert(reg
>= R600_LOOP_CONST_OFFSET
);
833 assert(cb
->num_dw
+2+num
<= cb
->max_num_dw
);
834 cb
->buf
[cb
->num_dw
++] = PKT3(PKT3_SET_LOOP_CONST
, num
, 0);
835 cb
->buf
[cb
->num_dw
++] = (reg
- R600_LOOP_CONST_OFFSET
) >> 2;
839 * Needs cb->pkt_flags set to RADEON_CP_PACKET3_COMPUTE_MODE for compute
842 static inline void eg_store_loop_const_seq(struct r600_command_buffer
*cb
, unsigned reg
, unsigned num
)
844 assert(reg
>= EG_LOOP_CONST_OFFSET
);
845 assert(cb
->num_dw
+2+num
<= cb
->max_num_dw
);
846 cb
->buf
[cb
->num_dw
++] = PKT3(PKT3_SET_LOOP_CONST
, num
, 0) | cb
->pkt_flags
;
847 cb
->buf
[cb
->num_dw
++] = (reg
- EG_LOOP_CONST_OFFSET
) >> 2;
850 static inline void r600_store_config_reg(struct r600_command_buffer
*cb
, unsigned reg
, unsigned value
)
852 r600_store_config_reg_seq(cb
, reg
, 1);
853 r600_store_value(cb
, value
);
856 static inline void r600_store_context_reg(struct r600_command_buffer
*cb
, unsigned reg
, unsigned value
)
858 r600_store_context_reg_seq(cb
, reg
, 1);
859 r600_store_value(cb
, value
);
862 static inline void r600_store_ctl_const(struct r600_command_buffer
*cb
, unsigned reg
, unsigned value
)
864 r600_store_ctl_const_seq(cb
, reg
, 1);
865 r600_store_value(cb
, value
);
868 static inline void r600_store_loop_const(struct r600_command_buffer
*cb
, unsigned reg
, unsigned value
)
870 r600_store_loop_const_seq(cb
, reg
, 1);
871 r600_store_value(cb
, value
);
874 static inline void eg_store_loop_const(struct r600_command_buffer
*cb
, unsigned reg
, unsigned value
)
876 eg_store_loop_const_seq(cb
, reg
, 1);
877 r600_store_value(cb
, value
);
880 void r600_init_command_buffer(struct r600_command_buffer
*cb
, unsigned num_dw
);
881 void r600_release_command_buffer(struct r600_command_buffer
*cb
);
883 static inline void radeon_compute_set_context_reg_seq(struct radeon_winsys_cs
*cs
, unsigned reg
, unsigned num
)
885 radeon_set_context_reg_seq(cs
, reg
, num
);
886 /* Set the compute bit on the packet header */
887 cs
->buf
[cs
->cdw
- 2] |= RADEON_CP_PACKET3_COMPUTE_MODE
;
890 static inline void radeon_set_ctl_const_seq(struct radeon_winsys_cs
*cs
, unsigned reg
, unsigned num
)
892 assert(reg
>= R600_CTL_CONST_OFFSET
);
893 assert(cs
->cdw
+2+num
<= cs
->max_dw
);
894 cs
->buf
[cs
->cdw
++] = PKT3(PKT3_SET_CTL_CONST
, num
, 0);
895 cs
->buf
[cs
->cdw
++] = (reg
- R600_CTL_CONST_OFFSET
) >> 2;
898 static inline void radeon_compute_set_context_reg(struct radeon_winsys_cs
*cs
, unsigned reg
, unsigned value
)
900 radeon_compute_set_context_reg_seq(cs
, reg
, 1);
901 radeon_emit(cs
, value
);
904 static inline void radeon_set_context_reg_flag(struct radeon_winsys_cs
*cs
, unsigned reg
, unsigned value
, unsigned flag
)
906 if (flag
& RADEON_CP_PACKET3_COMPUTE_MODE
) {
907 radeon_compute_set_context_reg(cs
, reg
, value
);
909 radeon_set_context_reg(cs
, reg
, value
);
913 static inline void radeon_set_ctl_const(struct radeon_winsys_cs
*cs
, unsigned reg
, unsigned value
)
915 radeon_set_ctl_const_seq(cs
, reg
, 1);
916 radeon_emit(cs
, value
);
922 static inline uint32_t S_FIXED(float value
, uint32_t frac_bits
)
924 return value
* (1 << frac_bits
);
926 #define ALIGN_DIVUP(x, y) (((x) + (y) - 1) / (y))
928 /* 12.4 fixed-point */
929 static inline unsigned r600_pack_float_12p4(float x
)
932 x
>= 4096 ? 0xffff : x
* 16;
935 /* Return if the depth format can be read without the DB->CB copy on r6xx-r7xx. */
936 static inline bool r600_can_read_depth(struct r600_texture
*rtex
)
938 return rtex
->resource
.b
.b
.nr_samples
<= 1 &&
939 (rtex
->resource
.b
.b
.format
== PIPE_FORMAT_Z16_UNORM
||
940 rtex
->resource
.b
.b
.format
== PIPE_FORMAT_Z32_FLOAT
);
943 #define V_028A6C_OUTPRIM_TYPE_POINTLIST 0
944 #define V_028A6C_OUTPRIM_TYPE_LINESTRIP 1
945 #define V_028A6C_OUTPRIM_TYPE_TRISTRIP 2
947 unsigned r600_conv_prim_to_gs_out(unsigned mode
);