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 "util/u_blitter.h"
30 #include "util/u_slab.h"
32 #include "r600_llvm.h"
33 #include "r600_public.h"
34 #include "r600_shader.h"
35 #include "r600_resource.h"
36 #include "evergreen_compute.h"
38 #define R600_MAX_ATOM 17
40 #define R600_MAX_CONST_BUFFERS 2
41 #define R600_MAX_CONST_BUFFER_SIZE 4096
43 #ifdef PIPE_ARCH_BIG_ENDIAN
44 #define R600_BIG_ENDIAN 1
46 #define R600_BIG_ENDIAN 0
49 /* This encapsulates a state or an operation which can emitted into the GPU
50 * command stream. It's not limited to states only, it can be used for anything
51 * that wants to write commands into the CS (e.g. cache flushes). */
53 void (*emit
)(struct r600_context
*ctx
, struct r600_atom
*state
);
59 /* This is an atom containing GPU commands that never change.
60 * This is supposed to be copied directly into the CS. */
61 struct r600_command_buffer
{
62 struct r600_atom atom
;
68 struct r600_db_misc_state
{
69 struct r600_atom atom
;
70 bool occlusion_query_enabled
;
71 bool flush_depthstencil_through_cb
;
72 bool copy_depth
, copy_stencil
;
77 struct r600_cb_misc_state
{
78 struct r600_atom atom
;
79 unsigned cb_color_control
; /* this comes from blend state */
80 unsigned blend_colormask
; /* 8*4 bits for 8 RGBA colorbuffers */
82 unsigned nr_ps_color_outputs
;
87 struct r600_alphatest_state
{
88 struct r600_atom atom
;
89 unsigned sx_alpha_test_control
; /* this comes from dsa state */
90 unsigned sx_alpha_ref
; /* this comes from dsa state */
92 bool cb0_export_16bpc
; /* from set_framebuffer_state */
95 struct r600_cs_shader_state
{
96 struct r600_atom atom
;
97 struct r600_pipe_compute
*shader
;
100 struct r600_sample_mask
{
101 struct r600_atom atom
;
102 uint16_t sample_mask
; /* there are only 8 bits on EG, 16 bits on Cayman */
105 enum r600_pipe_state_id
{
106 R600_PIPE_STATE_BLEND
= 0,
107 R600_PIPE_STATE_BLEND_COLOR
,
108 R600_PIPE_STATE_CONFIG
,
109 R600_PIPE_STATE_SEAMLESS_CUBEMAP
,
110 R600_PIPE_STATE_CLIP
,
111 R600_PIPE_STATE_SCISSOR
,
112 R600_PIPE_STATE_VIEWPORT
,
113 R600_PIPE_STATE_RASTERIZER
,
115 R600_PIPE_STATE_FRAMEBUFFER
,
117 R600_PIPE_STATE_STENCIL_REF
,
118 R600_PIPE_STATE_PS_SHADER
,
119 R600_PIPE_STATE_VS_SHADER
,
120 R600_PIPE_STATE_CONSTANT
,
121 R600_PIPE_STATE_SAMPLER
,
122 R600_PIPE_STATE_RESOURCE
,
123 R600_PIPE_STATE_POLYGON_OFFSET
,
124 R600_PIPE_STATE_FETCH_SHADER
,
129 struct compute_memory_pool
;
130 void compute_memory_pool_delete(struct compute_memory_pool
* pool
);
131 struct compute_memory_pool
* compute_memory_pool_new(
132 struct r600_screen
*rscreen
);
134 struct r600_pipe_fences
{
135 struct r600_resource
*bo
;
138 /* linked list of preallocated blocks */
139 struct list_head blocks
;
140 /* linked list of freed fences */
141 struct list_head pool
;
146 struct pipe_screen screen
;
147 struct radeon_winsys
*ws
;
149 enum chip_class chip_class
;
150 struct radeon_info info
;
152 struct r600_tiling_info tiling_info
;
153 struct r600_pipe_fences fences
;
155 /*for compute global memory binding, we allocate stuff here, instead of
157 * XXX: Not sure if this is the best place for global_pool. Also,
158 * it's not thread safe, so it won't work with multiple contexts. */
159 struct compute_memory_pool
*global_pool
;
162 struct r600_pipe_sampler_view
{
163 struct pipe_sampler_view base
;
164 struct r600_resource
*tex_resource
;
165 uint32_t tex_resource_words
[8];
168 struct r600_pipe_rasterizer
{
169 struct r600_pipe_state rstate
;
172 unsigned sprite_coord_enable
;
173 unsigned clip_plane_enable
;
174 unsigned pa_sc_line_stipple
;
175 unsigned pa_cl_clip_cntl
;
179 bool multisample_enable
;
182 struct r600_pipe_blend
{
183 struct r600_pipe_state rstate
;
184 unsigned cb_target_mask
;
185 unsigned cb_color_control
;
190 struct r600_pipe_dsa
{
191 struct r600_pipe_state rstate
;
195 unsigned sx_alpha_test_control
;
198 struct r600_vertex_element
201 struct pipe_vertex_element elements
[PIPE_MAX_ATTRIBS
];
202 struct r600_resource
*fetch_shader
;
204 struct r600_pipe_state rstate
;
207 struct r600_pipe_shader
;
209 struct r600_pipe_shader_selector
{
210 struct r600_pipe_shader
*current
;
212 struct tgsi_token
*tokens
;
213 struct pipe_stream_output_info so
;
215 unsigned num_shaders
;
217 /* PIPE_SHADER_[VERTEX|FRAGMENT|...] */
220 unsigned nr_ps_max_color_exports
;
223 struct r600_pipe_shader
{
224 struct r600_pipe_shader_selector
*selector
;
225 struct r600_pipe_shader
*next_variant
;
226 struct r600_shader shader
;
227 struct r600_pipe_state rstate
;
228 struct r600_resource
*bo
;
229 struct r600_resource
*bo_fetch
;
230 struct r600_vertex_element vertex_elements
;
231 unsigned sprite_coord_enable
;
233 unsigned pa_cl_vs_out_cntl
;
234 unsigned nr_ps_color_outputs
;
236 unsigned db_shader_control
;
237 unsigned ps_depth_export
;
240 struct r600_pipe_sampler_state
{
241 uint32_t tex_sampler_words
[3];
242 uint32_t border_color
[4];
243 bool border_color_use
;
244 bool seamless_cube_map
;
247 /* needed for blitter save */
248 #define NUM_TEX_UNITS 16
250 struct r600_seamless_cube_map
{
251 struct r600_atom atom
;
255 struct r600_samplerview_state
{
256 struct r600_atom atom
;
257 struct r600_pipe_sampler_view
*views
[NUM_TEX_UNITS
];
258 uint32_t enabled_mask
;
260 uint32_t compressed_depthtex_mask
; /* which textures are depth */
261 uint32_t compressed_colortex_mask
;
264 struct r600_textures_info
{
265 struct r600_samplerview_state views
;
266 struct r600_atom atom_sampler
;
267 struct r600_pipe_sampler_state
*samplers
[NUM_TEX_UNITS
];
269 bool is_array_sampler
[NUM_TEX_UNITS
];
273 struct pipe_reference reference
;
274 unsigned index
; /* in the shared bo */
275 struct r600_resource
*sleep_bo
;
276 struct list_head head
;
279 #define FENCE_BLOCK_SIZE 16
281 struct r600_fence_block
{
282 struct r600_fence fences
[FENCE_BLOCK_SIZE
];
283 struct list_head head
;
286 #define R600_CONSTANT_ARRAY_SIZE 256
287 #define R600_RESOURCE_ARRAY_SIZE 160
289 struct r600_stencil_ref
296 struct r600_constbuf_state
298 struct r600_atom atom
;
299 struct pipe_constant_buffer cb
[PIPE_MAX_CONSTANT_BUFFERS
];
300 uint32_t enabled_mask
;
304 struct r600_vertexbuf_state
306 struct r600_atom atom
;
307 struct pipe_vertex_buffer vb
[PIPE_MAX_ATTRIBS
];
308 uint32_t enabled_mask
; /* non-NULL buffers */
312 struct r600_context
{
313 struct pipe_context context
;
314 struct blitter_context
*blitter
;
315 enum radeon_family family
;
316 enum chip_class chip_class
;
317 boolean has_vertex_cache
;
318 unsigned r6xx_num_clause_temp_gprs
;
319 void *custom_dsa_flush
;
320 void *custom_blend_resolve
;
321 void *custom_blend_decompress
;
323 struct r600_screen
*screen
;
324 struct radeon_winsys
*ws
;
325 struct r600_pipe_state
*states
[R600_PIPE_NSTATES
];
326 struct r600_vertex_element
*vertex_elements
;
327 struct pipe_framebuffer_state framebuffer
;
328 unsigned compressed_cb_mask
;
329 unsigned compute_cb_target_mask
;
330 unsigned db_shader_control
;
331 unsigned pa_sc_line_stipple
;
332 unsigned pa_cl_clip_cntl
;
333 /* for saving when using blitter */
334 struct pipe_stencil_ref stencil_ref
;
335 struct pipe_viewport_state viewport
;
336 struct pipe_clip_state clip
;
337 struct r600_pipe_shader_selector
*ps_shader
;
338 struct r600_pipe_shader_selector
*vs_shader
;
339 struct r600_pipe_rasterizer
*rasterizer
;
340 struct r600_pipe_state vgt
;
341 struct r600_pipe_state spi
;
342 struct pipe_query
*current_render_cond
;
343 unsigned current_render_cond_mode
;
344 struct pipe_query
*saved_render_cond
;
345 unsigned saved_render_cond_mode
;
346 /* shader information */
349 unsigned sprite_coord_enable
;
351 boolean export_16bpc
;
354 bool multisample_enable
;
357 struct u_upload_mgr
*uploader
;
358 struct util_slab_mempool pool_transfers
;
360 unsigned default_ps_gprs
, default_vs_gprs
;
362 /* States based on r600_atom. */
363 struct r600_command_buffer start_cs_cmd
; /* invariant state mostly */
364 struct r600_atom
*atoms
[R600_MAX_ATOM
];
365 /** Compute specific registers initializations. The start_cs_cmd atom
366 * must be emitted before start_compute_cs_cmd. */
367 struct r600_command_buffer start_compute_cs_cmd
;
368 struct r600_alphatest_state alphatest_state
;
369 struct r600_cb_misc_state cb_misc_state
;
370 struct r600_db_misc_state db_misc_state
;
371 /** Vertex buffers for fetch shaders */
372 struct r600_vertexbuf_state vertex_buffer_state
;
373 /** Vertex buffers for compute shaders */
374 struct r600_vertexbuf_state cs_vertex_buffer_state
;
375 struct r600_constbuf_state constbuf_state
[PIPE_SHADER_TYPES
];
376 struct r600_textures_info vs_samplers
;
377 struct r600_textures_info ps_samplers
;
378 struct r600_seamless_cube_map seamless_cube_map
;
379 struct r600_cs_shader_state cs_shader_state
;
380 struct r600_sample_mask sample_mask
;
382 /* current external blend state (from state tracker) */
383 struct r600_pipe_blend
*blend
;
384 /* state with disabled blending - used internally with blend_override */
385 struct r600_pipe_blend
*no_blend
;
387 /* 1 - override current blend state with no_blend, 0 - use external state */
388 unsigned blend_override
;
390 struct radeon_winsys_cs
*cs
;
392 struct r600_range
*range
;
394 struct r600_block
**blocks
;
395 struct list_head dirty
;
396 struct list_head enable_list
;
397 unsigned pm4_dirty_cdwords
;
398 unsigned ctx_pm4_ndwords
;
400 /* The list of active queries. Only one query of each type can be active. */
401 int num_occlusion_queries
;
403 /* Manage queries in two separate groups:
404 * The timer ones and the others (streamout, occlusion).
406 * We do this because we should only suspend non-timer queries for u_blitter,
407 * and later if the non-timer queries are suspended, the context flush should
408 * only suspend and resume the timer queries. */
409 struct list_head active_timer_queries
;
410 unsigned num_cs_dw_timer_queries_suspend
;
411 struct list_head active_nontimer_queries
;
412 unsigned num_cs_dw_nontimer_queries_suspend
;
414 unsigned num_cs_dw_streamout_end
;
416 unsigned backend_mask
;
417 unsigned max_db
; /* for OQ */
419 boolean predicate_drawing
;
421 unsigned num_so_targets
;
422 struct r600_so_target
*so_targets
[PIPE_MAX_SO_BUFFERS
];
423 boolean streamout_start
;
424 unsigned streamout_append_bitmask
;
426 /* There is no scissor enable bit on r6xx, so we must use a workaround.
427 * These track the current scissor state. */
429 struct pipe_scissor_state scissor_state
;
431 /* With rasterizer discard, there doesn't have to be a pixel shader.
432 * In that case, we bind this one: */
433 void *dummy_pixel_shader
;
435 boolean dual_src_blend
;
438 struct pipe_index_buffer index_buffer
;
440 /* Dummy CMASK and FMASK buffers used to get around the R6xx hardware
441 * bug where valid CMASK and FMASK are required to be present to avoid
442 * a hardlock in certain operations but aren't actually used
443 * for anything useful. */
444 struct r600_resource
*dummy_fmask
;
445 struct r600_resource
*dummy_cmask
;
448 static INLINE
void r600_emit_atom(struct r600_context
*rctx
, struct r600_atom
*atom
)
450 atom
->emit(rctx
, atom
);
454 static INLINE
void r600_atom_dirty(struct r600_context
*rctx
, struct r600_atom
*state
)
459 /* evergreen_state.c */
460 void evergreen_init_common_regs(struct r600_command_buffer
*cb
,
461 enum chip_class ctx_chip_class
,
462 enum radeon_family ctx_family
,
465 void evergreen_init_state_functions(struct r600_context
*rctx
);
466 void evergreen_init_atom_start_cs(struct r600_context
*rctx
);
467 void evergreen_pipe_shader_ps(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
468 void evergreen_pipe_shader_vs(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
469 void evergreen_fetch_shader(struct pipe_context
*ctx
, struct r600_vertex_element
*ve
);
470 void *evergreen_create_db_flush_dsa(struct r600_context
*rctx
);
471 void *evergreen_create_resolve_blend(struct r600_context
*rctx
);
472 void *evergreen_create_decompress_blend(struct r600_context
*rctx
);
473 void evergreen_polygon_offset_update(struct r600_context
*rctx
);
474 boolean
evergreen_is_format_supported(struct pipe_screen
*screen
,
475 enum pipe_format format
,
476 enum pipe_texture_target target
,
477 unsigned sample_count
,
479 void evergreen_init_color_surface(struct r600_context
*rctx
,
480 struct r600_surface
*surf
);
481 void evergreen_update_dual_export_state(struct r600_context
* rctx
);
484 void r600_copy_buffer(struct pipe_context
*ctx
, struct
485 pipe_resource
*dst
, unsigned dstx
,
486 struct pipe_resource
*src
, const struct pipe_box
*src_box
);
487 void r600_init_blit_functions(struct r600_context
*rctx
);
488 void r600_blit_decompress_depth(struct pipe_context
*ctx
,
489 struct r600_texture
*texture
,
490 struct r600_texture
*staging
,
491 unsigned first_level
, unsigned last_level
,
492 unsigned first_layer
, unsigned last_layer
,
493 unsigned first_sample
, unsigned last_sample
);
494 void r600_decompress_depth_textures(struct r600_context
*rctx
,
495 struct r600_samplerview_state
*textures
);
496 void r600_decompress_color_textures(struct r600_context
*rctx
,
497 struct r600_samplerview_state
*textures
);
500 bool r600_init_resource(struct r600_screen
*rscreen
,
501 struct r600_resource
*res
,
502 unsigned size
, unsigned alignment
,
503 unsigned bind
, unsigned usage
);
504 struct pipe_resource
*r600_buffer_create(struct pipe_screen
*screen
,
505 const struct pipe_resource
*templ
,
509 void r600_flush(struct pipe_context
*ctx
, struct pipe_fence_handle
**fence
,
513 void r600_init_query_functions(struct r600_context
*rctx
);
514 void r600_suspend_nontimer_queries(struct r600_context
*ctx
);
515 void r600_resume_nontimer_queries(struct r600_context
*ctx
);
516 void r600_suspend_timer_queries(struct r600_context
*ctx
);
517 void r600_resume_timer_queries(struct r600_context
*ctx
);
519 /* r600_resource.c */
520 void r600_init_context_resource_functions(struct r600_context
*r600
);
523 int r600_pipe_shader_create(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
525 int r600_compute_shader_create(struct pipe_context
* ctx
,
526 LLVMModuleRef mod
, struct r600_bytecode
* bytecode
);
528 void r600_pipe_shader_destroy(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
531 void r600_set_scissor_state(struct r600_context
*rctx
,
532 const struct pipe_scissor_state
*state
);
533 void r600_init_state_functions(struct r600_context
*rctx
);
534 void r600_init_atom_start_cs(struct r600_context
*rctx
);
535 void r600_pipe_shader_ps(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
536 void r600_pipe_shader_vs(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
537 void r600_fetch_shader(struct pipe_context
*ctx
, struct r600_vertex_element
*ve
);
538 void *r600_create_db_flush_dsa(struct r600_context
*rctx
);
539 void *r600_create_resolve_blend(struct r600_context
*rctx
);
540 void *r700_create_resolve_blend(struct r600_context
*rctx
);
541 void *r600_create_decompress_blend(struct r600_context
*rctx
);
542 void r600_polygon_offset_update(struct r600_context
*rctx
);
543 void r600_adjust_gprs(struct r600_context
*rctx
);
544 boolean
r600_is_format_supported(struct pipe_screen
*screen
,
545 enum pipe_format format
,
546 enum pipe_texture_target target
,
547 unsigned sample_count
,
549 void r600_update_dual_export_state(struct r600_context
* rctx
);
552 void r600_init_screen_texture_functions(struct pipe_screen
*screen
);
553 void r600_init_surface_functions(struct r600_context
*r600
);
554 uint32_t r600_translate_texformat(struct pipe_screen
*screen
, enum pipe_format format
,
555 const unsigned char *swizzle_view
,
556 uint32_t *word4_p
, uint32_t *yuv_format_p
);
557 unsigned r600_texture_get_offset(struct r600_texture
*rtex
,
558 unsigned level
, unsigned layer
);
560 /* r600_translate.c */
561 void r600_translate_index_buffer(struct r600_context
*r600
,
562 struct pipe_index_buffer
*ib
,
565 /* r600_state_common.c */
566 void r600_init_common_state_functions(struct r600_context
*rctx
);
567 void r600_emit_alphatest_state(struct r600_context
*rctx
, struct r600_atom
*atom
);
568 void r600_init_atom(struct r600_context
*rctx
, struct r600_atom
*atom
, unsigned id
,
569 void (*emit
)(struct r600_context
*ctx
, struct r600_atom
*state
),
571 void r600_vertex_buffers_dirty(struct r600_context
*rctx
);
572 void r600_sampler_views_dirty(struct r600_context
*rctx
,
573 struct r600_samplerview_state
*state
);
574 void r600_set_max_scissor(struct r600_context
*rctx
);
575 void r600_constant_buffers_dirty(struct r600_context
*rctx
, struct r600_constbuf_state
*state
);
576 void r600_draw_rectangle(struct blitter_context
*blitter
,
577 unsigned x1
, unsigned y1
, unsigned x2
, unsigned y2
, float depth
,
578 enum blitter_attrib_type type
, const union pipe_color_union
*attrib
);
579 uint32_t r600_translate_stencil_op(int s_op
);
580 uint32_t r600_translate_fill(uint32_t func
);
581 unsigned r600_tex_wrap(unsigned wrap
);
582 unsigned r600_tex_filter(unsigned filter
);
583 unsigned r600_tex_mipfilter(unsigned filter
);
584 unsigned r600_tex_compare(unsigned compare
);
587 * Helpers for building command buffers
590 #define PKT3_SET_CONFIG_REG 0x68
591 #define PKT3_SET_CONTEXT_REG 0x69
592 #define PKT3_SET_CTL_CONST 0x6F
593 #define PKT3_SET_LOOP_CONST 0x6C
595 #define R600_CONFIG_REG_OFFSET 0x08000
596 #define R600_CONTEXT_REG_OFFSET 0x28000
597 #define R600_CTL_CONST_OFFSET 0x3CFF0
598 #define R600_LOOP_CONST_OFFSET 0X0003E200
599 #define EG_LOOP_CONST_OFFSET 0x0003A200
601 #define PKT_TYPE_S(x) (((x) & 0x3) << 30)
602 #define PKT_COUNT_S(x) (((x) & 0x3FFF) << 16)
603 #define PKT3_IT_OPCODE_S(x) (((x) & 0xFF) << 8)
604 #define PKT3_PREDICATE(x) (((x) >> 0) & 0x1)
605 #define PKT3(op, count, predicate) (PKT_TYPE_S(3) | PKT_COUNT_S(count) | PKT3_IT_OPCODE_S(op) | PKT3_PREDICATE(predicate))
607 #define RADEON_CP_PACKET3_COMPUTE_MODE 0x00000002
609 /*Evergreen Compute packet3*/
610 #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)
612 static INLINE
void r600_store_value(struct r600_command_buffer
*cb
, unsigned value
)
614 cb
->buf
[cb
->atom
.num_dw
++] = value
;
617 static INLINE
void r600_store_config_reg_seq(struct r600_command_buffer
*cb
, unsigned reg
, unsigned num
)
619 assert(reg
< R600_CONTEXT_REG_OFFSET
);
620 assert(cb
->atom
.num_dw
+2+num
<= cb
->max_num_dw
);
621 cb
->buf
[cb
->atom
.num_dw
++] = PKT3(PKT3_SET_CONFIG_REG
, num
, 0);
622 cb
->buf
[cb
->atom
.num_dw
++] = (reg
- R600_CONFIG_REG_OFFSET
) >> 2;
626 * Needs cb->pkt_flags set to RADEON_CP_PACKET3_COMPUTE_MODE for compute
629 static INLINE
void r600_store_context_reg_seq(struct r600_command_buffer
*cb
, unsigned reg
, unsigned num
)
631 assert(reg
>= R600_CONTEXT_REG_OFFSET
&& reg
< R600_CTL_CONST_OFFSET
);
632 assert(cb
->atom
.num_dw
+2+num
<= cb
->max_num_dw
);
633 cb
->buf
[cb
->atom
.num_dw
++] = PKT3(PKT3_SET_CONTEXT_REG
, num
, 0) | cb
->pkt_flags
;
634 cb
->buf
[cb
->atom
.num_dw
++] = (reg
- R600_CONTEXT_REG_OFFSET
) >> 2;
638 * Needs cb->pkt_flags set to RADEON_CP_PACKET3_COMPUTE_MODE for compute
641 static INLINE
void r600_store_ctl_const_seq(struct r600_command_buffer
*cb
, unsigned reg
, unsigned num
)
643 assert(reg
>= R600_CTL_CONST_OFFSET
);
644 assert(cb
->atom
.num_dw
+2+num
<= cb
->max_num_dw
);
645 cb
->buf
[cb
->atom
.num_dw
++] = PKT3(PKT3_SET_CTL_CONST
, num
, 0) | cb
->pkt_flags
;
646 cb
->buf
[cb
->atom
.num_dw
++] = (reg
- R600_CTL_CONST_OFFSET
) >> 2;
649 static INLINE
void r600_store_loop_const_seq(struct r600_command_buffer
*cb
, unsigned reg
, unsigned num
)
651 assert(reg
>= R600_LOOP_CONST_OFFSET
);
652 assert(cb
->atom
.num_dw
+2+num
<= cb
->max_num_dw
);
653 cb
->buf
[cb
->atom
.num_dw
++] = PKT3(PKT3_SET_LOOP_CONST
, num
, 0);
654 cb
->buf
[cb
->atom
.num_dw
++] = (reg
- R600_LOOP_CONST_OFFSET
) >> 2;
658 * Needs cb->pkt_flags set to RADEON_CP_PACKET3_COMPUTE_MODE for compute
661 static INLINE
void eg_store_loop_const_seq(struct r600_command_buffer
*cb
, unsigned reg
, unsigned num
)
663 assert(reg
>= EG_LOOP_CONST_OFFSET
);
664 assert(cb
->atom
.num_dw
+2+num
<= cb
->max_num_dw
);
665 cb
->buf
[cb
->atom
.num_dw
++] = PKT3(PKT3_SET_LOOP_CONST
, num
, 0) | cb
->pkt_flags
;
666 cb
->buf
[cb
->atom
.num_dw
++] = (reg
- EG_LOOP_CONST_OFFSET
) >> 2;
669 static INLINE
void r600_store_config_reg(struct r600_command_buffer
*cb
, unsigned reg
, unsigned value
)
671 r600_store_config_reg_seq(cb
, reg
, 1);
672 r600_store_value(cb
, value
);
675 static INLINE
void r600_store_context_reg(struct r600_command_buffer
*cb
, unsigned reg
, unsigned value
)
677 r600_store_context_reg_seq(cb
, reg
, 1);
678 r600_store_value(cb
, value
);
681 static INLINE
void r600_store_ctl_const(struct r600_command_buffer
*cb
, unsigned reg
, unsigned value
)
683 r600_store_ctl_const_seq(cb
, reg
, 1);
684 r600_store_value(cb
, value
);
687 static INLINE
void r600_store_loop_const(struct r600_command_buffer
*cb
, unsigned reg
, unsigned value
)
689 r600_store_loop_const_seq(cb
, reg
, 1);
690 r600_store_value(cb
, value
);
693 static INLINE
void eg_store_loop_const(struct r600_command_buffer
*cb
, unsigned reg
, unsigned value
)
695 eg_store_loop_const_seq(cb
, reg
, 1);
696 r600_store_value(cb
, value
);
699 void r600_init_command_buffer(struct r600_context
*rctx
, struct r600_command_buffer
*cb
, unsigned id
, unsigned num_dw
);
700 void r600_release_command_buffer(struct r600_command_buffer
*cb
);
703 * Helpers for emitting state into a command stream directly.
706 static INLINE
unsigned r600_context_bo_reloc(struct r600_context
*ctx
, struct r600_resource
*rbo
,
707 enum radeon_bo_usage usage
)
710 return ctx
->ws
->cs_add_reloc(ctx
->cs
, rbo
->cs_buf
, usage
, rbo
->domains
) * 4;
713 static INLINE
void r600_write_value(struct radeon_winsys_cs
*cs
, unsigned value
)
715 cs
->buf
[cs
->cdw
++] = value
;
718 static INLINE
void r600_write_array(struct radeon_winsys_cs
*cs
, unsigned num
, unsigned *ptr
)
720 assert(cs
->cdw
+num
<= RADEON_MAX_CMDBUF_DWORDS
);
721 memcpy(&cs
->buf
[cs
->cdw
], ptr
, num
* sizeof(ptr
[0]));
725 static INLINE
void r600_write_config_reg_seq(struct radeon_winsys_cs
*cs
, unsigned reg
, unsigned num
)
727 assert(reg
< R600_CONTEXT_REG_OFFSET
);
728 assert(cs
->cdw
+2+num
<= RADEON_MAX_CMDBUF_DWORDS
);
729 cs
->buf
[cs
->cdw
++] = PKT3(PKT3_SET_CONFIG_REG
, num
, 0);
730 cs
->buf
[cs
->cdw
++] = (reg
- R600_CONFIG_REG_OFFSET
) >> 2;
733 static INLINE
void r600_write_context_reg_seq(struct radeon_winsys_cs
*cs
, unsigned reg
, unsigned num
)
735 assert(reg
>= R600_CONTEXT_REG_OFFSET
&& reg
< R600_CTL_CONST_OFFSET
);
736 assert(cs
->cdw
+2+num
<= RADEON_MAX_CMDBUF_DWORDS
);
737 cs
->buf
[cs
->cdw
++] = PKT3(PKT3_SET_CONTEXT_REG
, num
, 0);
738 cs
->buf
[cs
->cdw
++] = (reg
- R600_CONTEXT_REG_OFFSET
) >> 2;
741 static INLINE
void r600_write_compute_context_reg_seq(struct radeon_winsys_cs
*cs
, unsigned reg
, unsigned num
)
743 r600_write_context_reg_seq(cs
, reg
, num
);
744 /* Set the compute bit on the packet header */
745 cs
->buf
[cs
->cdw
- 2] |= RADEON_CP_PACKET3_COMPUTE_MODE
;
748 static INLINE
void r600_write_ctl_const_seq(struct radeon_winsys_cs
*cs
, unsigned reg
, unsigned num
)
750 assert(reg
>= R600_CTL_CONST_OFFSET
);
751 assert(cs
->cdw
+2+num
<= RADEON_MAX_CMDBUF_DWORDS
);
752 cs
->buf
[cs
->cdw
++] = PKT3(PKT3_SET_CTL_CONST
, num
, 0);
753 cs
->buf
[cs
->cdw
++] = (reg
- R600_CTL_CONST_OFFSET
) >> 2;
756 static INLINE
void r600_write_config_reg(struct radeon_winsys_cs
*cs
, unsigned reg
, unsigned value
)
758 r600_write_config_reg_seq(cs
, reg
, 1);
759 r600_write_value(cs
, value
);
762 static INLINE
void r600_write_context_reg(struct radeon_winsys_cs
*cs
, unsigned reg
, unsigned value
)
764 r600_write_context_reg_seq(cs
, reg
, 1);
765 r600_write_value(cs
, value
);
768 static INLINE
void r600_write_compute_context_reg(struct radeon_winsys_cs
*cs
, unsigned reg
, unsigned value
)
770 r600_write_compute_context_reg_seq(cs
, reg
, 1);
771 r600_write_value(cs
, value
);
774 static INLINE
void r600_write_ctl_const(struct radeon_winsys_cs
*cs
, unsigned reg
, unsigned value
)
776 r600_write_ctl_const_seq(cs
, reg
, 1);
777 r600_write_value(cs
, value
);
783 static INLINE
uint32_t S_FIXED(float value
, uint32_t frac_bits
)
785 return value
* (1 << frac_bits
);
787 #define ALIGN_DIVUP(x, y) (((x) + (y) - 1) / (y))
789 static inline unsigned r600_tex_aniso_filter(unsigned filter
)
791 if (filter
<= 1) return 0;
792 if (filter
<= 2) return 1;
793 if (filter
<= 4) return 2;
794 if (filter
<= 8) return 3;
798 /* 12.4 fixed-point */
799 static INLINE
unsigned r600_pack_float_12p4(float x
)
802 x
>= 4096 ? 0xffff : x
* 16;
805 static INLINE
uint64_t r600_resource_va(struct pipe_screen
*screen
, struct pipe_resource
*resource
)
807 struct r600_screen
*rscreen
= (struct r600_screen
*)screen
;
808 struct r600_resource
*rresource
= (struct r600_resource
*)resource
;
810 return rscreen
->ws
->buffer_get_virtual_address(rresource
->cs_buf
);