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_slab.h"
31 #include "r600_llvm.h"
32 #include "r600_public.h"
33 #include "r600_shader.h"
34 #include "r600_resource.h"
35 #include "evergreen_compute.h"
37 #define R600_MAX_CONST_BUFFERS 2
38 #define R600_MAX_CONST_BUFFER_SIZE 4096
40 #ifdef PIPE_ARCH_BIG_ENDIAN
41 #define R600_BIG_ENDIAN 1
43 #define R600_BIG_ENDIAN 0
46 enum r600_atom_flags
{
47 /* When set, atoms are added at the beginning of the dirty list
48 * instead of the end. */
52 /* This encapsulates a state or an operation which can emitted into the GPU
53 * command stream. It's not limited to states only, it can be used for anything
54 * that wants to write commands into the CS (e.g. cache flushes). */
56 void (*emit
)(struct r600_context
*ctx
, struct r600_atom
*state
);
59 enum r600_atom_flags flags
;
62 struct list_head head
;
65 /* This is an atom containing GPU commands that never change.
66 * This is supposed to be copied directly into the CS. */
67 struct r600_command_buffer
{
68 struct r600_atom atom
;
74 struct r600_surface_sync_cmd
{
75 struct r600_atom atom
;
76 unsigned flush_flags
; /* CP_COHER_CNTL */
79 struct r600_db_misc_state
{
80 struct r600_atom atom
;
81 bool occlusion_query_enabled
;
82 bool flush_depthstencil_through_cb
;
83 bool copy_depth
, copy_stencil
;
86 struct r600_cb_misc_state
{
87 struct r600_atom atom
;
88 unsigned cb_color_control
; /* this comes from blend state */
89 unsigned blend_colormask
; /* 8*4 bits for 8 RGBA colorbuffers */
91 unsigned nr_ps_color_outputs
;
96 struct r600_alphatest_state
{
97 struct r600_atom atom
;
98 unsigned sx_alpha_test_control
; /* this comes from dsa state */
99 unsigned sx_alpha_ref
; /* this comes from dsa state */
101 bool cb0_export_16bpc
; /* from set_framebuffer_state */
104 struct r600_cs_shader_state
{
105 struct r600_atom atom
;
106 struct r600_pipe_compute
*shader
;
109 enum r600_pipe_state_id
{
110 R600_PIPE_STATE_BLEND
= 0,
111 R600_PIPE_STATE_BLEND_COLOR
,
112 R600_PIPE_STATE_CONFIG
,
113 R600_PIPE_STATE_SEAMLESS_CUBEMAP
,
114 R600_PIPE_STATE_CLIP
,
115 R600_PIPE_STATE_SCISSOR
,
116 R600_PIPE_STATE_VIEWPORT
,
117 R600_PIPE_STATE_RASTERIZER
,
119 R600_PIPE_STATE_FRAMEBUFFER
,
121 R600_PIPE_STATE_STENCIL_REF
,
122 R600_PIPE_STATE_PS_SHADER
,
123 R600_PIPE_STATE_VS_SHADER
,
124 R600_PIPE_STATE_CONSTANT
,
125 R600_PIPE_STATE_SAMPLER
,
126 R600_PIPE_STATE_RESOURCE
,
127 R600_PIPE_STATE_POLYGON_OFFSET
,
128 R600_PIPE_STATE_FETCH_SHADER
,
133 struct compute_memory_pool
;
134 void compute_memory_pool_delete(struct compute_memory_pool
* pool
);
135 struct compute_memory_pool
* compute_memory_pool_new(
136 struct r600_screen
*rscreen
);
138 struct r600_pipe_fences
{
139 struct r600_resource
*bo
;
142 /* linked list of preallocated blocks */
143 struct list_head blocks
;
144 /* linked list of freed fences */
145 struct list_head pool
;
150 struct pipe_screen screen
;
151 struct radeon_winsys
*ws
;
153 enum chip_class chip_class
;
154 struct radeon_info info
;
156 struct r600_tiling_info tiling_info
;
157 struct r600_pipe_fences fences
;
159 /*for compute global memory binding, we allocate stuff here, instead of
161 * XXX: Not sure if this is the best place for global_pool. Also,
162 * it's not thread safe, so it won't work with multiple contexts. */
163 struct compute_memory_pool
*global_pool
;
166 struct r600_pipe_sampler_view
{
167 struct pipe_sampler_view base
;
168 struct r600_resource
*tex_resource
;
169 uint32_t tex_resource_words
[8];
172 struct r600_pipe_rasterizer
{
173 struct r600_pipe_state rstate
;
176 unsigned sprite_coord_enable
;
177 unsigned clip_plane_enable
;
178 unsigned pa_sc_line_stipple
;
179 unsigned pa_cl_clip_cntl
;
185 struct r600_pipe_blend
{
186 struct r600_pipe_state rstate
;
187 unsigned cb_target_mask
;
188 unsigned cb_color_control
;
192 struct r600_pipe_dsa
{
193 struct r600_pipe_state rstate
;
197 unsigned sx_alpha_test_control
;
200 struct r600_vertex_element
203 struct pipe_vertex_element elements
[PIPE_MAX_ATTRIBS
];
204 struct r600_resource
*fetch_shader
;
206 struct r600_pipe_state rstate
;
209 struct r600_pipe_shader
;
211 struct r600_pipe_shader_selector
{
212 struct r600_pipe_shader
*current
;
214 struct tgsi_token
*tokens
;
215 struct pipe_stream_output_info so
;
217 unsigned num_shaders
;
219 /* PIPE_SHADER_[VERTEX|FRAGMENT|...] */
222 unsigned nr_ps_max_color_exports
;
225 struct r600_pipe_shader
{
226 struct r600_pipe_shader_selector
*selector
;
227 struct r600_pipe_shader
*next_variant
;
228 struct r600_shader shader
;
229 struct r600_pipe_state rstate
;
230 struct r600_resource
*bo
;
231 struct r600_resource
*bo_fetch
;
232 struct r600_vertex_element vertex_elements
;
233 unsigned sprite_coord_enable
;
235 unsigned pa_cl_vs_out_cntl
;
236 unsigned nr_ps_color_outputs
;
238 unsigned db_shader_control
;
239 unsigned ps_depth_export
;
242 struct r600_pipe_sampler_state
{
243 struct r600_pipe_state rstate
;
244 boolean seamless_cube_map
;
247 /* needed for blitter save */
248 #define NUM_TEX_UNITS 16
250 struct r600_samplerview_state
252 struct r600_atom atom
;
253 struct r600_pipe_sampler_view
*views
[NUM_TEX_UNITS
];
254 uint32_t enabled_mask
;
256 uint32_t depth_texture_mask
; /* which textures are depth */
259 struct r600_textures_info
{
260 struct r600_samplerview_state views
;
262 struct r600_pipe_sampler_state
*samplers
[NUM_TEX_UNITS
];
265 bool is_array_sampler
[NUM_TEX_UNITS
];
269 struct pipe_reference reference
;
270 unsigned index
; /* in the shared bo */
271 struct r600_resource
*sleep_bo
;
272 struct list_head head
;
275 #define FENCE_BLOCK_SIZE 16
277 struct r600_fence_block
{
278 struct r600_fence fences
[FENCE_BLOCK_SIZE
];
279 struct list_head head
;
282 #define R600_CONSTANT_ARRAY_SIZE 256
283 #define R600_RESOURCE_ARRAY_SIZE 160
285 struct r600_stencil_ref
292 struct r600_constbuf_state
294 struct r600_atom atom
;
295 struct pipe_constant_buffer cb
[PIPE_MAX_CONSTANT_BUFFERS
];
296 uint32_t enabled_mask
;
300 struct r600_vertexbuf_state
302 struct r600_atom atom
;
303 struct pipe_vertex_buffer vb
[PIPE_MAX_ATTRIBS
];
304 uint32_t enabled_mask
; /* non-NULL buffers */
308 struct r600_context
{
309 struct pipe_context context
;
310 struct blitter_context
*blitter
;
311 enum radeon_family family
;
312 enum chip_class chip_class
;
313 boolean has_vertex_cache
;
314 unsigned r6xx_num_clause_temp_gprs
;
315 void *custom_dsa_flush
;
316 struct r600_screen
*screen
;
317 struct radeon_winsys
*ws
;
318 struct r600_pipe_state
*states
[R600_PIPE_NSTATES
];
319 struct r600_vertex_element
*vertex_elements
;
320 struct pipe_framebuffer_state framebuffer
;
321 unsigned compute_cb_target_mask
;
322 unsigned db_shader_control
;
323 unsigned pa_sc_line_stipple
;
324 unsigned pa_cl_clip_cntl
;
325 /* for saving when using blitter */
326 struct pipe_stencil_ref stencil_ref
;
327 struct pipe_viewport_state viewport
;
328 struct pipe_clip_state clip
;
329 struct r600_pipe_shader_selector
*ps_shader
;
330 struct r600_pipe_shader_selector
*vs_shader
;
331 struct r600_pipe_rasterizer
*rasterizer
;
332 struct r600_pipe_state vgt
;
333 struct r600_pipe_state spi
;
334 struct pipe_query
*current_render_cond
;
335 unsigned current_render_cond_mode
;
336 struct pipe_query
*saved_render_cond
;
337 unsigned saved_render_cond_mode
;
338 /* shader information */
341 unsigned sprite_coord_enable
;
343 boolean export_16bpc
;
346 struct u_upload_mgr
*uploader
;
347 struct util_slab_mempool pool_transfers
;
349 unsigned default_ps_gprs
, default_vs_gprs
;
351 /* States based on r600_atom. */
352 struct list_head dirty_states
;
353 struct r600_command_buffer start_cs_cmd
; /* invariant state mostly */
354 /** Compute specific registers initializations. The start_cs_cmd atom
355 * must be emitted before start_compute_cs_cmd. */
356 struct r600_command_buffer start_compute_cs_cmd
;
357 struct r600_surface_sync_cmd surface_sync_cmd
;
358 struct r600_atom r6xx_flush_and_inv_cmd
;
359 struct r600_alphatest_state alphatest_state
;
360 struct r600_cb_misc_state cb_misc_state
;
361 struct r600_db_misc_state db_misc_state
;
362 /** Vertex buffers for fetch shaders */
363 struct r600_vertexbuf_state vertex_buffer_state
;
364 /** Vertex buffers for compute shaders */
365 struct r600_vertexbuf_state cs_vertex_buffer_state
;
366 struct r600_constbuf_state vs_constbuf_state
;
367 struct r600_constbuf_state ps_constbuf_state
;
368 struct r600_textures_info vs_samplers
;
369 struct r600_textures_info ps_samplers
;
370 struct r600_cs_shader_state cs_shader_state
;
372 struct radeon_winsys_cs
*cs
;
374 struct r600_range
*range
;
376 struct r600_block
**blocks
;
377 struct list_head dirty
;
378 struct list_head enable_list
;
379 unsigned pm4_dirty_cdwords
;
380 unsigned ctx_pm4_ndwords
;
382 /* The list of active queries. Only one query of each type can be active. */
383 int num_occlusion_queries
;
385 /* Manage queries in two separate groups:
386 * The timer ones and the others (streamout, occlusion).
388 * We do this because we should only suspend non-timer queries for u_blitter,
389 * and later if the non-timer queries are suspended, the context flush should
390 * only suspend and resume the timer queries. */
391 struct list_head active_timer_queries
;
392 unsigned num_cs_dw_timer_queries_suspend
;
393 struct list_head active_nontimer_queries
;
394 unsigned num_cs_dw_nontimer_queries_suspend
;
396 unsigned num_cs_dw_streamout_end
;
398 unsigned backend_mask
;
399 unsigned max_db
; /* for OQ */
401 boolean predicate_drawing
;
403 unsigned num_so_targets
;
404 struct r600_so_target
*so_targets
[PIPE_MAX_SO_BUFFERS
];
405 boolean streamout_start
;
406 unsigned streamout_append_bitmask
;
408 /* There is no scissor enable bit on r6xx, so we must use a workaround.
409 * These track the current scissor state. */
411 struct pipe_scissor_state scissor_state
;
413 /* With rasterizer discard, there doesn't have to be a pixel shader.
414 * In that case, we bind this one: */
415 void *dummy_pixel_shader
;
417 boolean dual_src_blend
;
420 struct pipe_index_buffer index_buffer
;
423 static INLINE
void r600_emit_atom(struct r600_context
*rctx
, struct r600_atom
*atom
)
425 atom
->emit(rctx
, atom
);
427 if (atom
->head
.next
&& atom
->head
.prev
)
428 LIST_DELINIT(&atom
->head
);
431 static INLINE
void r600_atom_dirty(struct r600_context
*rctx
, struct r600_atom
*state
)
434 if (state
->flags
& EMIT_EARLY
) {
435 LIST_ADD(&state
->head
, &rctx
->dirty_states
);
437 LIST_ADDTAIL(&state
->head
, &rctx
->dirty_states
);
443 /* evergreen_state.c */
444 void evergreen_init_state_functions(struct r600_context
*rctx
);
445 void evergreen_init_atom_start_cs(struct r600_context
*rctx
);
446 void evergreen_pipe_shader_ps(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
447 void evergreen_pipe_shader_vs(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
448 void evergreen_fetch_shader(struct pipe_context
*ctx
, struct r600_vertex_element
*ve
);
449 void *evergreen_create_db_flush_dsa(struct r600_context
*rctx
);
450 void evergreen_polygon_offset_update(struct r600_context
*rctx
);
451 boolean
evergreen_is_format_supported(struct pipe_screen
*screen
,
452 enum pipe_format format
,
453 enum pipe_texture_target target
,
454 unsigned sample_count
,
456 void evergreen_cb(struct r600_context
*rctx
, struct r600_pipe_state
*rstate
,
457 const struct pipe_framebuffer_state
*state
, int cb
);
460 void evergreen_update_dual_export_state(struct r600_context
* rctx
);
463 void r600_copy_buffer(struct pipe_context
*ctx
, struct
464 pipe_resource
*dst
, unsigned dstx
,
465 struct pipe_resource
*src
, const struct pipe_box
*src_box
);
466 void r600_init_blit_functions(struct r600_context
*rctx
);
467 void r600_blit_uncompress_depth(struct pipe_context
*ctx
,
468 struct r600_resource_texture
*texture
,
469 struct r600_resource_texture
*staging
,
470 unsigned first_level
, unsigned last_level
,
471 unsigned first_layer
, unsigned last_layer
);
472 void r600_flush_depth_textures(struct r600_context
*rctx
,
473 struct r600_samplerview_state
*textures
);
475 bool r600_init_resource(struct r600_screen
*rscreen
,
476 struct r600_resource
*res
,
477 unsigned size
, unsigned alignment
,
478 unsigned bind
, unsigned usage
);
479 struct pipe_resource
*r600_buffer_create(struct pipe_screen
*screen
,
480 const struct pipe_resource
*templ
);
483 void r600_flush(struct pipe_context
*ctx
, struct pipe_fence_handle
**fence
,
487 void r600_init_query_functions(struct r600_context
*rctx
);
488 void r600_suspend_nontimer_queries(struct r600_context
*ctx
);
489 void r600_resume_nontimer_queries(struct r600_context
*ctx
);
490 void r600_suspend_timer_queries(struct r600_context
*ctx
);
491 void r600_resume_timer_queries(struct r600_context
*ctx
);
493 /* r600_resource.c */
494 void r600_init_context_resource_functions(struct r600_context
*r600
);
497 int r600_pipe_shader_create(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
499 int r600_compute_shader_create(struct pipe_context
* ctx
,
500 LLVMModuleRef mod
, struct r600_bytecode
* bytecode
);
502 void r600_pipe_shader_destroy(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
505 void r600_set_scissor_state(struct r600_context
*rctx
,
506 const struct pipe_scissor_state
*state
);
507 void r600_update_sampler_states(struct r600_context
*rctx
);
508 void r600_init_state_functions(struct r600_context
*rctx
);
509 void r600_init_atom_start_cs(struct r600_context
*rctx
);
510 void r600_pipe_shader_ps(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
511 void r600_pipe_shader_vs(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
);
512 void r600_fetch_shader(struct pipe_context
*ctx
, struct r600_vertex_element
*ve
);
513 void *r600_create_db_flush_dsa(struct r600_context
*rctx
);
514 void r600_polygon_offset_update(struct r600_context
*rctx
);
515 void r600_adjust_gprs(struct r600_context
*rctx
);
516 boolean
r600_is_format_supported(struct pipe_screen
*screen
,
517 enum pipe_format format
,
518 enum pipe_texture_target target
,
519 unsigned sample_count
,
521 void r600_update_dual_export_state(struct r600_context
* rctx
);
524 void r600_init_screen_texture_functions(struct pipe_screen
*screen
);
525 void r600_init_surface_functions(struct r600_context
*r600
);
526 uint32_t r600_translate_texformat(struct pipe_screen
*screen
, enum pipe_format format
,
527 const unsigned char *swizzle_view
,
528 uint32_t *word4_p
, uint32_t *yuv_format_p
);
529 unsigned r600_texture_get_offset(struct r600_resource_texture
*rtex
,
530 unsigned level
, unsigned layer
);
532 /* r600_translate.c */
533 void r600_translate_index_buffer(struct r600_context
*r600
,
534 struct pipe_index_buffer
*ib
,
537 /* r600_state_common.c */
538 void r600_init_atom(struct r600_atom
*atom
,
539 void (*emit
)(struct r600_context
*ctx
, struct r600_atom
*state
),
540 unsigned num_dw
, enum r600_atom_flags flags
);
541 void r600_init_common_atoms(struct r600_context
*rctx
);
542 unsigned r600_get_cb_flush_flags(struct r600_context
*rctx
);
543 void r600_texture_barrier(struct pipe_context
*ctx
);
544 void r600_set_index_buffer(struct pipe_context
*ctx
,
545 const struct pipe_index_buffer
*ib
);
546 void r600_vertex_buffers_dirty(struct r600_context
*rctx
);
547 void r600_set_vertex_buffers(struct pipe_context
*ctx
, unsigned count
,
548 const struct pipe_vertex_buffer
*input
);
549 void r600_sampler_views_dirty(struct r600_context
*rctx
,
550 struct r600_samplerview_state
*state
);
551 void r600_set_sampler_views(struct r600_context
*rctx
,
552 struct r600_textures_info
*dst
,
554 struct pipe_sampler_view
**views
);
555 void *r600_create_vertex_elements(struct pipe_context
*ctx
,
557 const struct pipe_vertex_element
*elements
);
558 void r600_delete_vertex_element(struct pipe_context
*ctx
, void *state
);
559 void r600_bind_blend_state(struct pipe_context
*ctx
, void *state
);
560 void r600_set_blend_color(struct pipe_context
*ctx
,
561 const struct pipe_blend_color
*state
);
562 void r600_bind_dsa_state(struct pipe_context
*ctx
, void *state
);
563 void r600_set_max_scissor(struct r600_context
*rctx
);
564 void r600_bind_rs_state(struct pipe_context
*ctx
, void *state
);
565 void r600_delete_rs_state(struct pipe_context
*ctx
, void *state
);
566 void r600_sampler_view_destroy(struct pipe_context
*ctx
,
567 struct pipe_sampler_view
*state
);
568 void r600_delete_state(struct pipe_context
*ctx
, void *state
);
569 void r600_bind_vertex_elements(struct pipe_context
*ctx
, void *state
);
570 void *r600_create_shader_state_ps(struct pipe_context
*ctx
,
571 const struct pipe_shader_state
*state
);
572 void *r600_create_shader_state_vs(struct pipe_context
*ctx
,
573 const struct pipe_shader_state
*state
);
574 void r600_bind_ps_shader(struct pipe_context
*ctx
, void *state
);
575 void r600_bind_vs_shader(struct pipe_context
*ctx
, void *state
);
576 void r600_delete_ps_shader(struct pipe_context
*ctx
, void *state
);
577 void r600_delete_vs_shader(struct pipe_context
*ctx
, void *state
);
578 void r600_constant_buffers_dirty(struct r600_context
*rctx
, struct r600_constbuf_state
*state
);
579 void r600_set_constant_buffer(struct pipe_context
*ctx
, uint shader
, uint index
,
580 struct pipe_constant_buffer
*cb
);
581 struct pipe_stream_output_target
*
582 r600_create_so_target(struct pipe_context
*ctx
,
583 struct pipe_resource
*buffer
,
584 unsigned buffer_offset
,
585 unsigned buffer_size
);
586 void r600_so_target_destroy(struct pipe_context
*ctx
,
587 struct pipe_stream_output_target
*target
);
588 void r600_set_so_targets(struct pipe_context
*ctx
,
589 unsigned num_targets
,
590 struct pipe_stream_output_target
**targets
,
591 unsigned append_bitmask
);
592 void r600_set_pipe_stencil_ref(struct pipe_context
*ctx
,
593 const struct pipe_stencil_ref
*state
);
594 void r600_draw_vbo(struct pipe_context
*ctx
, const struct pipe_draw_info
*info
);
595 uint32_t r600_translate_stencil_op(int s_op
);
596 uint32_t r600_translate_fill(uint32_t func
);
597 unsigned r600_tex_wrap(unsigned wrap
);
598 unsigned r600_tex_filter(unsigned filter
);
599 unsigned r600_tex_mipfilter(unsigned filter
);
600 unsigned r600_tex_compare(unsigned compare
);
603 * Helpers for building command buffers
606 #define PKT3_SET_CONFIG_REG 0x68
607 #define PKT3_SET_CONTEXT_REG 0x69
608 #define PKT3_SET_CTL_CONST 0x6F
609 #define PKT3_SET_LOOP_CONST 0x6C
611 #define R600_CONFIG_REG_OFFSET 0x08000
612 #define R600_CONTEXT_REG_OFFSET 0x28000
613 #define R600_CTL_CONST_OFFSET 0x3CFF0
614 #define R600_LOOP_CONST_OFFSET 0X0003E200
615 #define EG_LOOP_CONST_OFFSET 0x0003A200
617 #define PKT_TYPE_S(x) (((x) & 0x3) << 30)
618 #define PKT_COUNT_S(x) (((x) & 0x3FFF) << 16)
619 #define PKT3_IT_OPCODE_S(x) (((x) & 0xFF) << 8)
620 #define PKT3_PREDICATE(x) (((x) >> 0) & 0x1)
621 #define PKT3(op, count, predicate) (PKT_TYPE_S(3) | PKT_COUNT_S(count) | PKT3_IT_OPCODE_S(op) | PKT3_PREDICATE(predicate))
623 #define RADEON_CP_PACKET3_COMPUTE_MODE 0x00000002
625 /*Evergreen Compute packet3*/
626 #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)
628 static INLINE
void r600_store_value(struct r600_command_buffer
*cb
, unsigned value
)
630 cb
->buf
[cb
->atom
.num_dw
++] = value
;
633 static INLINE
void r600_store_config_reg_seq(struct r600_command_buffer
*cb
, unsigned reg
, unsigned num
)
635 assert(reg
< R600_CONTEXT_REG_OFFSET
);
636 assert(cb
->atom
.num_dw
+2+num
<= cb
->max_num_dw
);
637 cb
->buf
[cb
->atom
.num_dw
++] = PKT3(PKT3_SET_CONFIG_REG
, num
, 0);
638 cb
->buf
[cb
->atom
.num_dw
++] = (reg
- R600_CONFIG_REG_OFFSET
) >> 2;
642 * Needs cb->pkt_flags set to RADEON_CP_PACKET3_COMPUTE_MODE for compute
645 static INLINE
void r600_store_context_reg_seq(struct r600_command_buffer
*cb
, unsigned reg
, unsigned num
)
647 assert(reg
>= R600_CONTEXT_REG_OFFSET
&& reg
< R600_CTL_CONST_OFFSET
);
648 assert(cb
->atom
.num_dw
+2+num
<= cb
->max_num_dw
);
649 cb
->buf
[cb
->atom
.num_dw
++] = PKT3(PKT3_SET_CONTEXT_REG
, num
, 0) | cb
->pkt_flags
;
650 cb
->buf
[cb
->atom
.num_dw
++] = (reg
- R600_CONTEXT_REG_OFFSET
) >> 2;
654 * Needs cb->pkt_flags set to RADEON_CP_PACKET3_COMPUTE_MODE for compute
657 static INLINE
void r600_store_ctl_const_seq(struct r600_command_buffer
*cb
, unsigned reg
, unsigned num
)
659 assert(reg
>= R600_CTL_CONST_OFFSET
);
660 assert(cb
->atom
.num_dw
+2+num
<= cb
->max_num_dw
);
661 cb
->buf
[cb
->atom
.num_dw
++] = PKT3(PKT3_SET_CTL_CONST
, num
, 0) | cb
->pkt_flags
;
662 cb
->buf
[cb
->atom
.num_dw
++] = (reg
- R600_CTL_CONST_OFFSET
) >> 2;
665 static INLINE
void r600_store_loop_const_seq(struct r600_command_buffer
*cb
, unsigned reg
, unsigned num
)
667 assert(reg
>= R600_LOOP_CONST_OFFSET
);
668 assert(cb
->atom
.num_dw
+2+num
<= cb
->max_num_dw
);
669 cb
->buf
[cb
->atom
.num_dw
++] = PKT3(PKT3_SET_LOOP_CONST
, num
, 0);
670 cb
->buf
[cb
->atom
.num_dw
++] = (reg
- R600_LOOP_CONST_OFFSET
) >> 2;
674 * Needs cb->pkt_flags set to RADEON_CP_PACKET3_COMPUTE_MODE for compute
677 static INLINE
void eg_store_loop_const_seq(struct r600_command_buffer
*cb
, unsigned reg
, unsigned num
)
679 assert(reg
>= EG_LOOP_CONST_OFFSET
);
680 assert(cb
->atom
.num_dw
+2+num
<= cb
->max_num_dw
);
681 cb
->buf
[cb
->atom
.num_dw
++] = PKT3(PKT3_SET_LOOP_CONST
, num
, 0) | cb
->pkt_flags
;
682 cb
->buf
[cb
->atom
.num_dw
++] = (reg
- EG_LOOP_CONST_OFFSET
) >> 2;
685 static INLINE
void r600_store_config_reg(struct r600_command_buffer
*cb
, unsigned reg
, unsigned value
)
687 r600_store_config_reg_seq(cb
, reg
, 1);
688 r600_store_value(cb
, value
);
691 static INLINE
void r600_store_context_reg(struct r600_command_buffer
*cb
, unsigned reg
, unsigned value
)
693 r600_store_context_reg_seq(cb
, reg
, 1);
694 r600_store_value(cb
, value
);
697 static INLINE
void r600_store_ctl_const(struct r600_command_buffer
*cb
, unsigned reg
, unsigned value
)
699 r600_store_ctl_const_seq(cb
, reg
, 1);
700 r600_store_value(cb
, value
);
703 static INLINE
void r600_store_loop_const(struct r600_command_buffer
*cb
, unsigned reg
, unsigned value
)
705 r600_store_loop_const_seq(cb
, reg
, 1);
706 r600_store_value(cb
, value
);
709 static INLINE
void eg_store_loop_const(struct r600_command_buffer
*cb
, unsigned reg
, unsigned value
)
711 eg_store_loop_const_seq(cb
, reg
, 1);
712 r600_store_value(cb
, value
);
715 void r600_init_command_buffer(struct r600_command_buffer
*cb
, unsigned num_dw
, enum r600_atom_flags flags
);
716 void r600_release_command_buffer(struct r600_command_buffer
*cb
);
719 * Helpers for emitting state into a command stream directly.
722 static INLINE
unsigned r600_context_bo_reloc(struct r600_context
*ctx
, struct r600_resource
*rbo
,
723 enum radeon_bo_usage usage
)
726 return ctx
->ws
->cs_add_reloc(ctx
->cs
, rbo
->cs_buf
, usage
, rbo
->domains
) * 4;
729 static INLINE
void r600_write_value(struct radeon_winsys_cs
*cs
, unsigned value
)
731 cs
->buf
[cs
->cdw
++] = value
;
734 static INLINE
void r600_write_array(struct radeon_winsys_cs
*cs
, unsigned num
, unsigned *ptr
)
736 assert(cs
->cdw
+num
<= RADEON_MAX_CMDBUF_DWORDS
);
737 memcpy(&cs
->buf
[cs
->cdw
], ptr
, num
* sizeof(ptr
[0]));
741 static INLINE
void r600_write_config_reg_seq(struct radeon_winsys_cs
*cs
, unsigned reg
, unsigned num
)
743 assert(reg
< R600_CONTEXT_REG_OFFSET
);
744 assert(cs
->cdw
+2+num
<= RADEON_MAX_CMDBUF_DWORDS
);
745 cs
->buf
[cs
->cdw
++] = PKT3(PKT3_SET_CONFIG_REG
, num
, 0);
746 cs
->buf
[cs
->cdw
++] = (reg
- R600_CONFIG_REG_OFFSET
) >> 2;
749 static INLINE
void r600_write_context_reg_seq(struct radeon_winsys_cs
*cs
, unsigned reg
, unsigned num
)
751 assert(reg
>= R600_CONTEXT_REG_OFFSET
&& reg
< R600_CTL_CONST_OFFSET
);
752 assert(cs
->cdw
+2+num
<= RADEON_MAX_CMDBUF_DWORDS
);
753 cs
->buf
[cs
->cdw
++] = PKT3(PKT3_SET_CONTEXT_REG
, num
, 0);
754 cs
->buf
[cs
->cdw
++] = (reg
- R600_CONTEXT_REG_OFFSET
) >> 2;
757 static INLINE
void r600_write_compute_context_reg_seq(struct radeon_winsys_cs
*cs
, unsigned reg
, unsigned num
)
759 r600_write_context_reg_seq(cs
, reg
, num
);
760 /* Set the compute bit on the packet header */
761 cs
->buf
[cs
->cdw
- 2] |= RADEON_CP_PACKET3_COMPUTE_MODE
;
764 static INLINE
void r600_write_ctl_const_seq(struct radeon_winsys_cs
*cs
, unsigned reg
, unsigned num
)
766 assert(reg
>= R600_CTL_CONST_OFFSET
);
767 assert(cs
->cdw
+2+num
<= RADEON_MAX_CMDBUF_DWORDS
);
768 cs
->buf
[cs
->cdw
++] = PKT3(PKT3_SET_CTL_CONST
, num
, 0);
769 cs
->buf
[cs
->cdw
++] = (reg
- R600_CTL_CONST_OFFSET
) >> 2;
772 static INLINE
void r600_write_config_reg(struct radeon_winsys_cs
*cs
, unsigned reg
, unsigned value
)
774 r600_write_config_reg_seq(cs
, reg
, 1);
775 r600_write_value(cs
, value
);
778 static INLINE
void r600_write_context_reg(struct radeon_winsys_cs
*cs
, unsigned reg
, unsigned value
)
780 r600_write_context_reg_seq(cs
, reg
, 1);
781 r600_write_value(cs
, value
);
784 static INLINE
void r600_write_compute_context_reg(struct radeon_winsys_cs
*cs
, unsigned reg
, unsigned value
)
786 r600_write_compute_context_reg_seq(cs
, reg
, 1);
787 r600_write_value(cs
, value
);
790 static INLINE
void r600_write_ctl_const(struct radeon_winsys_cs
*cs
, unsigned reg
, unsigned value
)
792 r600_write_ctl_const_seq(cs
, reg
, 1);
793 r600_write_value(cs
, value
);
799 static INLINE
uint32_t S_FIXED(float value
, uint32_t frac_bits
)
801 return value
* (1 << frac_bits
);
803 #define ALIGN_DIVUP(x, y) (((x) + (y) - 1) / (y))
805 static inline unsigned r600_tex_aniso_filter(unsigned filter
)
807 if (filter
<= 1) return 0;
808 if (filter
<= 2) return 1;
809 if (filter
<= 4) return 2;
810 if (filter
<= 8) return 3;
814 /* 12.4 fixed-point */
815 static INLINE
unsigned r600_pack_float_12p4(float x
)
818 x
>= 4096 ? 0xffff : x
* 16;
821 static INLINE
uint64_t r600_resource_va(struct pipe_screen
*screen
, struct pipe_resource
*resource
)
823 struct r600_screen
*rscreen
= (struct r600_screen
*)screen
;
824 struct r600_resource
*rresource
= (struct r600_resource
*)resource
;
826 return rscreen
->ws
->buffer_get_virtual_address(rresource
->cs_buf
);