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;
85 struct r600_cb_misc_state {
86 struct r600_atom atom;
87 unsigned cb_color_control; /* this comes from blend state */
88 unsigned blend_colormask; /* 8*4 bits for 8 RGBA colorbuffers */
90 unsigned nr_ps_color_outputs;
95 enum r600_pipe_state_id {
96 R600_PIPE_STATE_BLEND = 0,
97 R600_PIPE_STATE_BLEND_COLOR,
98 R600_PIPE_STATE_CONFIG,
99 R600_PIPE_STATE_SEAMLESS_CUBEMAP,
100 R600_PIPE_STATE_CLIP,
101 R600_PIPE_STATE_SCISSOR,
102 R600_PIPE_STATE_VIEWPORT,
103 R600_PIPE_STATE_RASTERIZER,
105 R600_PIPE_STATE_FRAMEBUFFER,
107 R600_PIPE_STATE_STENCIL_REF,
108 R600_PIPE_STATE_PS_SHADER,
109 R600_PIPE_STATE_VS_SHADER,
110 R600_PIPE_STATE_CONSTANT,
111 R600_PIPE_STATE_SAMPLER,
112 R600_PIPE_STATE_RESOURCE,
113 R600_PIPE_STATE_POLYGON_OFFSET,
114 R600_PIPE_STATE_FETCH_SHADER,
119 struct compute_memory_pool;
120 void compute_memory_pool_delete(struct compute_memory_pool* pool);
121 struct compute_memory_pool* compute_memory_pool_new(
122 struct r600_screen *rscreen);
124 struct r600_pipe_fences {
125 struct r600_resource *bo;
128 /* linked list of preallocated blocks */
129 struct list_head blocks;
130 /* linked list of freed fences */
131 struct list_head pool;
136 struct pipe_screen screen;
137 struct radeon_winsys *ws;
139 enum chip_class chip_class;
140 struct radeon_info info;
142 struct r600_tiling_info tiling_info;
143 struct r600_pipe_fences fences;
145 bool use_surface_alloc;
147 /*for compute global memory binding, we allocate stuff here, instead of
149 * XXX: Not sure if this is the best place for global_pool. Also,
150 * it's not thread safe, so it won't work with multiple contexts. */
151 struct compute_memory_pool *global_pool;
154 struct r600_pipe_sampler_view {
155 struct pipe_sampler_view base;
156 struct r600_resource *tex_resource;
157 uint32_t tex_resource_words[8];
160 struct r600_pipe_rasterizer {
161 struct r600_pipe_state rstate;
164 unsigned sprite_coord_enable;
165 unsigned clip_plane_enable;
166 unsigned pa_sc_line_stipple;
167 unsigned pa_cl_clip_cntl;
173 struct r600_pipe_blend {
174 struct r600_pipe_state rstate;
175 unsigned cb_target_mask;
176 unsigned cb_color_control;
180 struct r600_pipe_dsa {
181 struct r600_pipe_state rstate;
185 unsigned sx_alpha_test_control;
188 struct r600_vertex_element
191 struct pipe_vertex_element elements[PIPE_MAX_ATTRIBS];
192 struct r600_resource *fetch_shader;
194 struct r600_pipe_state rstate;
197 struct r600_pipe_shader;
199 struct r600_pipe_shader_selector {
200 struct r600_pipe_shader *current;
202 struct tgsi_token *tokens;
203 struct pipe_stream_output_info so;
205 unsigned num_shaders;
207 /* PIPE_SHADER_[VERTEX|FRAGMENT|...] */
210 unsigned nr_ps_max_color_exports;
213 struct r600_pipe_shader {
214 struct r600_pipe_shader_selector *selector;
215 struct r600_pipe_shader *next_variant;
216 struct r600_shader shader;
217 struct r600_pipe_state rstate;
218 struct r600_resource *bo;
219 struct r600_resource *bo_fetch;
220 struct r600_vertex_element vertex_elements;
221 unsigned sprite_coord_enable;
223 unsigned pa_cl_vs_out_cntl;
224 unsigned nr_ps_color_outputs;
226 unsigned db_shader_control;
227 unsigned ps_depth_export;
230 struct r600_pipe_sampler_state {
231 struct r600_pipe_state rstate;
232 boolean seamless_cube_map;
235 /* needed for blitter save */
236 #define NUM_TEX_UNITS 16
238 struct r600_samplerview_state
240 struct r600_atom atom;
241 struct r600_pipe_sampler_view *views[NUM_TEX_UNITS];
242 uint32_t enabled_mask;
244 uint32_t depth_texture_mask; /* which textures are depth */
247 struct r600_textures_info {
248 struct r600_samplerview_state views;
250 struct r600_pipe_sampler_state *samplers[NUM_TEX_UNITS];
253 bool is_array_sampler[NUM_TEX_UNITS];
257 struct pipe_reference reference;
258 unsigned index; /* in the shared bo */
259 struct r600_resource *sleep_bo;
260 struct list_head head;
263 #define FENCE_BLOCK_SIZE 16
265 struct r600_fence_block {
266 struct r600_fence fences[FENCE_BLOCK_SIZE];
267 struct list_head head;
270 #define R600_CONSTANT_ARRAY_SIZE 256
271 #define R600_RESOURCE_ARRAY_SIZE 160
273 struct r600_stencil_ref
280 struct r600_constbuf_state
282 struct r600_atom atom;
283 struct pipe_constant_buffer cb[PIPE_MAX_CONSTANT_BUFFERS];
284 uint32_t enabled_mask;
288 struct r600_vertexbuf_state
290 struct r600_atom atom;
291 struct pipe_vertex_buffer vb[PIPE_MAX_ATTRIBS];
292 uint32_t enabled_mask; /* non-NULL buffers */
296 struct r600_context {
297 struct pipe_context context;
298 struct blitter_context *blitter;
299 enum radeon_family family;
300 enum chip_class chip_class;
301 boolean has_vertex_cache;
302 unsigned r6xx_num_clause_temp_gprs;
303 void *custom_dsa_flush;
304 struct r600_screen *screen;
305 struct radeon_winsys *ws;
306 struct r600_pipe_state *states[R600_PIPE_NSTATES];
307 struct r600_vertex_element *vertex_elements;
308 struct pipe_framebuffer_state framebuffer;
309 unsigned compute_cb_target_mask;
310 unsigned sx_alpha_test_control;
311 unsigned db_shader_control;
312 unsigned pa_sc_line_stipple;
313 unsigned pa_cl_clip_cntl;
314 /* for saving when using blitter */
315 struct pipe_stencil_ref stencil_ref;
316 struct pipe_viewport_state viewport;
317 struct pipe_clip_state clip;
318 struct r600_pipe_shader_selector *ps_shader;
319 struct r600_pipe_shader_selector *vs_shader;
320 struct r600_pipe_compute *cs_shader;
321 struct r600_pipe_rasterizer *rasterizer;
322 struct r600_pipe_state vgt;
323 struct r600_pipe_state spi;
324 struct pipe_query *current_render_cond;
325 unsigned current_render_cond_mode;
326 struct pipe_query *saved_render_cond;
327 unsigned saved_render_cond_mode;
328 /* shader information */
331 unsigned sprite_coord_enable;
333 boolean export_16bpc;
335 boolean alpha_ref_dirty;
338 struct u_upload_mgr *uploader;
339 struct util_slab_mempool pool_transfers;
341 unsigned default_ps_gprs, default_vs_gprs;
343 /* States based on r600_atom. */
344 struct list_head dirty_states;
345 struct r600_command_buffer start_cs_cmd; /* invariant state mostly */
346 /** Compute specific registers initializations. The start_cs_cmd atom
347 * must be emitted before start_compute_cs_cmd. */
348 struct r600_command_buffer start_compute_cs_cmd;
349 struct r600_surface_sync_cmd surface_sync_cmd;
350 struct r600_atom r6xx_flush_and_inv_cmd;
351 struct r600_cb_misc_state cb_misc_state;
352 struct r600_db_misc_state db_misc_state;
353 /** Vertex buffers for fetch shaders */
354 struct r600_vertexbuf_state vertex_buffer_state;
355 /** Vertex buffers for compute shaders */
356 struct r600_vertexbuf_state cs_vertex_buffer_state;
357 struct r600_constbuf_state vs_constbuf_state;
358 struct r600_constbuf_state ps_constbuf_state;
359 struct r600_textures_info vs_samplers;
360 struct r600_textures_info ps_samplers;
362 struct radeon_winsys_cs *cs;
364 struct r600_range *range;
366 struct r600_block **blocks;
367 struct list_head dirty;
368 struct list_head resource_dirty;
369 struct list_head enable_list;
370 unsigned pm4_dirty_cdwords;
371 unsigned ctx_pm4_ndwords;
373 /* The list of active queries. Only one query of each type can be active. */
374 int num_occlusion_queries;
376 /* Manage queries in two separate groups:
377 * The timer ones and the others (streamout, occlusion).
379 * We do this because we should only suspend non-timer queries for u_blitter,
380 * and later if the non-timer queries are suspended, the context flush should
381 * only suspend and resume the timer queries. */
382 struct list_head active_timer_queries;
383 unsigned num_cs_dw_timer_queries_suspend;
384 struct list_head active_nontimer_queries;
385 unsigned num_cs_dw_nontimer_queries_suspend;
387 unsigned num_cs_dw_streamout_end;
389 unsigned backend_mask;
390 unsigned max_db; /* for OQ */
392 boolean predicate_drawing;
393 struct r600_range ps_resources;
394 struct r600_range vs_resources;
395 int num_ps_resources, num_vs_resources;
397 unsigned num_so_targets;
398 struct r600_so_target *so_targets[PIPE_MAX_SO_BUFFERS];
399 boolean streamout_start;
400 unsigned streamout_append_bitmask;
402 /* There is no scissor enable bit on r6xx, so we must use a workaround.
403 * These track the current scissor state. */
405 struct pipe_scissor_state scissor_state;
407 /* With rasterizer discard, there doesn't have to be a pixel shader.
408 * In that case, we bind this one: */
409 void *dummy_pixel_shader;
411 boolean dual_src_blend;
414 struct pipe_index_buffer index_buffer;
417 static INLINE void r600_emit_atom(struct r600_context *rctx, struct r600_atom *atom)
419 atom->emit(rctx, atom);
421 if (atom->head.next && atom->head.prev)
422 LIST_DELINIT(&atom->head);
425 static INLINE void r600_atom_dirty(struct r600_context *rctx, struct r600_atom *state)
428 if (state->flags & EMIT_EARLY) {
429 LIST_ADD(&state->head, &rctx->dirty_states);
431 LIST_ADDTAIL(&state->head, &rctx->dirty_states);
437 /* evergreen_state.c */
438 void evergreen_init_state_functions(struct r600_context *rctx);
439 void evergreen_init_atom_start_cs(struct r600_context *rctx);
440 void evergreen_pipe_shader_ps(struct pipe_context *ctx, struct r600_pipe_shader *shader);
441 void evergreen_pipe_shader_vs(struct pipe_context *ctx, struct r600_pipe_shader *shader);
442 void evergreen_fetch_shader(struct pipe_context *ctx, struct r600_vertex_element *ve);
443 void *evergreen_create_db_flush_dsa(struct r600_context *rctx);
444 void evergreen_polygon_offset_update(struct r600_context *rctx);
445 boolean evergreen_is_format_supported(struct pipe_screen *screen,
446 enum pipe_format format,
447 enum pipe_texture_target target,
448 unsigned sample_count,
450 void evergreen_cb(struct r600_context *rctx, struct r600_pipe_state *rstate,
451 const struct pipe_framebuffer_state *state, int cb);
454 void evergreen_update_dual_export_state(struct r600_context * rctx);
457 void r600_init_blit_functions(struct r600_context *rctx);
458 void r600_blit_uncompress_depth(struct pipe_context *ctx,
459 struct r600_resource_texture *texture,
460 struct r600_resource_texture *staging,
461 unsigned first_level, unsigned last_level,
462 unsigned first_layer, unsigned last_layer);
463 void r600_flush_depth_textures(struct r600_context *rctx,
464 struct r600_samplerview_state *textures);
466 bool r600_init_resource(struct r600_screen *rscreen,
467 struct r600_resource *res,
468 unsigned size, unsigned alignment,
469 unsigned bind, unsigned usage);
470 struct pipe_resource *r600_buffer_create(struct pipe_screen *screen,
471 const struct pipe_resource *templ);
474 void r600_flush(struct pipe_context *ctx, struct pipe_fence_handle **fence,
478 void r600_init_query_functions(struct r600_context *rctx);
479 void r600_suspend_nontimer_queries(struct r600_context *ctx);
480 void r600_resume_nontimer_queries(struct r600_context *ctx);
481 void r600_suspend_timer_queries(struct r600_context *ctx);
482 void r600_resume_timer_queries(struct r600_context *ctx);
484 /* r600_resource.c */
485 void r600_init_context_resource_functions(struct r600_context *r600);
488 int r600_pipe_shader_create(struct pipe_context *ctx, struct r600_pipe_shader *shader);
490 int r600_compute_shader_create(struct pipe_context * ctx,
491 LLVMModuleRef mod, struct r600_bytecode * bytecode);
493 void r600_pipe_shader_destroy(struct pipe_context *ctx, struct r600_pipe_shader *shader);
496 void r600_set_scissor_state(struct r600_context *rctx,
497 const struct pipe_scissor_state *state);
498 void r600_update_sampler_states(struct r600_context *rctx);
499 void r600_init_state_functions(struct r600_context *rctx);
500 void r600_init_atom_start_cs(struct r600_context *rctx);
501 void r600_pipe_shader_ps(struct pipe_context *ctx, struct r600_pipe_shader *shader);
502 void r600_pipe_shader_vs(struct pipe_context *ctx, struct r600_pipe_shader *shader);
503 void r600_fetch_shader(struct pipe_context *ctx, struct r600_vertex_element *ve);
504 void *r600_create_db_flush_dsa(struct r600_context *rctx);
505 void r600_polygon_offset_update(struct r600_context *rctx);
506 void r600_adjust_gprs(struct r600_context *rctx);
507 boolean r600_is_format_supported(struct pipe_screen *screen,
508 enum pipe_format format,
509 enum pipe_texture_target target,
510 unsigned sample_count,
512 void r600_update_dual_export_state(struct r600_context * rctx);
515 void r600_init_screen_texture_functions(struct pipe_screen *screen);
516 void r600_init_surface_functions(struct r600_context *r600);
517 uint32_t r600_translate_texformat(struct pipe_screen *screen, enum pipe_format format,
518 const unsigned char *swizzle_view,
519 uint32_t *word4_p, uint32_t *yuv_format_p);
520 unsigned r600_texture_get_offset(struct r600_resource_texture *rtex,
521 unsigned level, unsigned layer);
523 /* r600_translate.c */
524 void r600_translate_index_buffer(struct r600_context *r600,
525 struct pipe_index_buffer *ib,
528 /* r600_state_common.c */
529 void r600_init_atom(struct r600_atom *atom,
530 void (*emit)(struct r600_context *ctx, struct r600_atom *state),
531 unsigned num_dw, enum r600_atom_flags flags);
532 void r600_init_common_atoms(struct r600_context *rctx);
533 unsigned r600_get_cb_flush_flags(struct r600_context *rctx);
534 void r600_texture_barrier(struct pipe_context *ctx);
535 void r600_set_index_buffer(struct pipe_context *ctx,
536 const struct pipe_index_buffer *ib);
537 void r600_vertex_buffers_dirty(struct r600_context *rctx);
538 void r600_set_vertex_buffers(struct pipe_context *ctx, unsigned count,
539 const struct pipe_vertex_buffer *input);
540 void r600_sampler_views_dirty(struct r600_context *rctx,
541 struct r600_samplerview_state *state);
542 void r600_set_sampler_views(struct r600_context *rctx,
543 struct r600_textures_info *dst,
545 struct pipe_sampler_view **views);
546 void *r600_create_vertex_elements(struct pipe_context *ctx,
548 const struct pipe_vertex_element *elements);
549 void r600_delete_vertex_element(struct pipe_context *ctx, void *state);
550 void r600_bind_blend_state(struct pipe_context *ctx, void *state);
551 void r600_set_blend_color(struct pipe_context *ctx,
552 const struct pipe_blend_color *state);
553 void r600_bind_dsa_state(struct pipe_context *ctx, void *state);
554 void r600_set_max_scissor(struct r600_context *rctx);
555 void r600_bind_rs_state(struct pipe_context *ctx, void *state);
556 void r600_delete_rs_state(struct pipe_context *ctx, void *state);
557 void r600_sampler_view_destroy(struct pipe_context *ctx,
558 struct pipe_sampler_view *state);
559 void r600_delete_state(struct pipe_context *ctx, void *state);
560 void r600_bind_vertex_elements(struct pipe_context *ctx, void *state);
561 void *r600_create_shader_state_ps(struct pipe_context *ctx,
562 const struct pipe_shader_state *state);
563 void *r600_create_shader_state_vs(struct pipe_context *ctx,
564 const struct pipe_shader_state *state);
565 void r600_bind_ps_shader(struct pipe_context *ctx, void *state);
566 void r600_bind_vs_shader(struct pipe_context *ctx, void *state);
567 void r600_delete_ps_shader(struct pipe_context *ctx, void *state);
568 void r600_delete_vs_shader(struct pipe_context *ctx, void *state);
569 void r600_constant_buffers_dirty(struct r600_context *rctx, struct r600_constbuf_state *state);
570 void r600_set_constant_buffer(struct pipe_context *ctx, uint shader, uint index,
571 struct pipe_constant_buffer *cb);
572 struct pipe_stream_output_target *
573 r600_create_so_target(struct pipe_context *ctx,
574 struct pipe_resource *buffer,
575 unsigned buffer_offset,
576 unsigned buffer_size);
577 void r600_so_target_destroy(struct pipe_context *ctx,
578 struct pipe_stream_output_target *target);
579 void r600_set_so_targets(struct pipe_context *ctx,
580 unsigned num_targets,
581 struct pipe_stream_output_target **targets,
582 unsigned append_bitmask);
583 void r600_set_pipe_stencil_ref(struct pipe_context *ctx,
584 const struct pipe_stencil_ref *state);
585 void r600_draw_vbo(struct pipe_context *ctx, const struct pipe_draw_info *info);
586 uint32_t r600_translate_stencil_op(int s_op);
587 uint32_t r600_translate_fill(uint32_t func);
588 unsigned r600_tex_wrap(unsigned wrap);
589 unsigned r600_tex_filter(unsigned filter);
590 unsigned r600_tex_mipfilter(unsigned filter);
591 unsigned r600_tex_compare(unsigned compare);
594 * Helpers for building command buffers
597 #define PKT3_SET_CONFIG_REG 0x68
598 #define PKT3_SET_CONTEXT_REG 0x69
599 #define PKT3_SET_CTL_CONST 0x6F
600 #define PKT3_SET_LOOP_CONST 0x6C
602 #define R600_CONFIG_REG_OFFSET 0x08000
603 #define R600_CONTEXT_REG_OFFSET 0x28000
604 #define R600_CTL_CONST_OFFSET 0x3CFF0
605 #define R600_LOOP_CONST_OFFSET 0X0003E200
606 #define EG_LOOP_CONST_OFFSET 0x0003A200
608 #define PKT_TYPE_S(x) (((x) & 0x3) << 30)
609 #define PKT_COUNT_S(x) (((x) & 0x3FFF) << 16)
610 #define PKT3_IT_OPCODE_S(x) (((x) & 0xFF) << 8)
611 #define PKT3_PREDICATE(x) (((x) >> 0) & 0x1)
612 #define PKT3(op, count, predicate) (PKT_TYPE_S(3) | PKT_COUNT_S(count) | PKT3_IT_OPCODE_S(op) | PKT3_PREDICATE(predicate))
614 static INLINE void r600_store_value(struct r600_command_buffer *cb, unsigned value)
616 cb->buf[cb->atom.num_dw++] = value;
619 static INLINE void r600_store_config_reg_seq(struct r600_command_buffer *cb, unsigned reg, unsigned num)
621 assert(reg < R600_CONTEXT_REG_OFFSET);
622 assert(cb->atom.num_dw+2+num <= cb->max_num_dw);
623 cb->buf[cb->atom.num_dw++] = PKT3(PKT3_SET_CONFIG_REG, num, 0);
624 cb->buf[cb->atom.num_dw++] = (reg - R600_CONFIG_REG_OFFSET) >> 2;
628 * Needs cb->pkt_flags set to RADEON_CP_PACKET3_COMPUTE_MODE for compute
631 static INLINE void r600_store_context_reg_seq(struct r600_command_buffer *cb, unsigned reg, unsigned num)
633 assert(reg >= R600_CONTEXT_REG_OFFSET && reg < R600_CTL_CONST_OFFSET);
634 assert(cb->atom.num_dw+2+num <= cb->max_num_dw);
635 cb->buf[cb->atom.num_dw++] = PKT3(PKT3_SET_CONTEXT_REG, num, 0) | cb->pkt_flags;
636 cb->buf[cb->atom.num_dw++] = (reg - R600_CONTEXT_REG_OFFSET) >> 2;
640 * Needs cb->pkt_flags set to RADEON_CP_PACKET3_COMPUTE_MODE for compute
643 static INLINE void r600_store_ctl_const_seq(struct r600_command_buffer *cb, unsigned reg, unsigned num)
645 assert(reg >= R600_CTL_CONST_OFFSET);
646 assert(cb->atom.num_dw+2+num <= cb->max_num_dw);
647 cb->buf[cb->atom.num_dw++] = PKT3(PKT3_SET_CTL_CONST, num, 0) | cb->pkt_flags;
648 cb->buf[cb->atom.num_dw++] = (reg - R600_CTL_CONST_OFFSET) >> 2;
651 static INLINE void r600_store_loop_const_seq(struct r600_command_buffer *cb, unsigned reg, unsigned num)
653 assert(reg >= R600_LOOP_CONST_OFFSET);
654 assert(cb->atom.num_dw+2+num <= cb->max_num_dw);
655 cb->buf[cb->atom.num_dw++] = PKT3(PKT3_SET_LOOP_CONST, num, 0);
656 cb->buf[cb->atom.num_dw++] = (reg - R600_LOOP_CONST_OFFSET) >> 2;
660 * Needs cb->pkt_flags set to RADEON_CP_PACKET3_COMPUTE_MODE for compute
663 static INLINE void eg_store_loop_const_seq(struct r600_command_buffer *cb, unsigned reg, unsigned num)
665 assert(reg >= EG_LOOP_CONST_OFFSET);
666 assert(cb->atom.num_dw+2+num <= cb->max_num_dw);
667 cb->buf[cb->atom.num_dw++] = PKT3(PKT3_SET_LOOP_CONST, num, 0) | cb->pkt_flags;
668 cb->buf[cb->atom.num_dw++] = (reg - EG_LOOP_CONST_OFFSET) >> 2;
671 static INLINE void r600_store_config_reg(struct r600_command_buffer *cb, unsigned reg, unsigned value)
673 r600_store_config_reg_seq(cb, reg, 1);
674 r600_store_value(cb, value);
677 static INLINE void r600_store_context_reg(struct r600_command_buffer *cb, unsigned reg, unsigned value)
679 r600_store_context_reg_seq(cb, reg, 1);
680 r600_store_value(cb, value);
683 static INLINE void r600_store_ctl_const(struct r600_command_buffer *cb, unsigned reg, unsigned value)
685 r600_store_ctl_const_seq(cb, reg, 1);
686 r600_store_value(cb, value);
689 static INLINE void r600_store_loop_const(struct r600_command_buffer *cb, unsigned reg, unsigned value)
691 r600_store_loop_const_seq(cb, reg, 1);
692 r600_store_value(cb, value);
695 static INLINE void eg_store_loop_const(struct r600_command_buffer *cb, unsigned reg, unsigned value)
697 eg_store_loop_const_seq(cb, reg, 1);
698 r600_store_value(cb, value);
701 void r600_init_command_buffer(struct r600_command_buffer *cb, unsigned num_dw, enum r600_atom_flags flags);
702 void r600_release_command_buffer(struct r600_command_buffer *cb);
705 * Helpers for emitting state into a command stream directly.
708 static INLINE unsigned r600_context_bo_reloc(struct r600_context *ctx, struct r600_resource *rbo,
709 enum radeon_bo_usage usage)
712 return ctx->ws->cs_add_reloc(ctx->cs, rbo->cs_buf, usage, rbo->domains) * 4;
715 static INLINE void r600_write_value(struct radeon_winsys_cs *cs, unsigned value)
717 cs->buf[cs->cdw++] = value;
720 static INLINE void r600_write_array(struct radeon_winsys_cs *cs, unsigned num, unsigned *ptr)
722 assert(cs->cdw+num <= RADEON_MAX_CMDBUF_DWORDS);
723 memcpy(&cs->buf[cs->cdw], ptr, num * sizeof(ptr[0]));
727 static INLINE void r600_write_config_reg_seq(struct radeon_winsys_cs *cs, unsigned reg, unsigned num)
729 assert(reg < R600_CONTEXT_REG_OFFSET);
730 assert(cs->cdw+2+num <= RADEON_MAX_CMDBUF_DWORDS);
731 cs->buf[cs->cdw++] = PKT3(PKT3_SET_CONFIG_REG, num, 0);
732 cs->buf[cs->cdw++] = (reg - R600_CONFIG_REG_OFFSET) >> 2;
735 static INLINE void r600_write_context_reg_seq(struct radeon_winsys_cs *cs, unsigned reg, unsigned num)
737 assert(reg >= R600_CONTEXT_REG_OFFSET && reg < R600_CTL_CONST_OFFSET);
738 assert(cs->cdw+2+num <= RADEON_MAX_CMDBUF_DWORDS);
739 cs->buf[cs->cdw++] = PKT3(PKT3_SET_CONTEXT_REG, num, 0);
740 cs->buf[cs->cdw++] = (reg - R600_CONTEXT_REG_OFFSET) >> 2;
743 static INLINE void r600_write_ctl_const_seq(struct radeon_winsys_cs *cs, unsigned reg, unsigned num)
745 assert(reg >= R600_CTL_CONST_OFFSET);
746 assert(cs->cdw+2+num <= RADEON_MAX_CMDBUF_DWORDS);
747 cs->buf[cs->cdw++] = PKT3(PKT3_SET_CTL_CONST, num, 0);
748 cs->buf[cs->cdw++] = (reg - R600_CTL_CONST_OFFSET) >> 2;
751 static INLINE void r600_write_config_reg(struct radeon_winsys_cs *cs, unsigned reg, unsigned value)
753 r600_write_config_reg_seq(cs, reg, 1);
754 r600_write_value(cs, value);
757 static INLINE void r600_write_context_reg(struct radeon_winsys_cs *cs, unsigned reg, unsigned value)
759 r600_write_context_reg_seq(cs, reg, 1);
760 r600_write_value(cs, value);
763 static INLINE void r600_write_ctl_const(struct radeon_winsys_cs *cs, unsigned reg, unsigned value)
765 r600_write_ctl_const_seq(cs, reg, 1);
766 r600_write_value(cs, value);
772 static INLINE uint32_t S_FIXED(float value, uint32_t frac_bits)
774 return value * (1 << frac_bits);
776 #define ALIGN_DIVUP(x, y) (((x) + (y) - 1) / (y))
778 static inline unsigned r600_tex_aniso_filter(unsigned filter)
780 if (filter <= 1) return 0;
781 if (filter <= 2) return 1;
782 if (filter <= 4) return 2;
783 if (filter <= 8) return 3;
787 /* 12.4 fixed-point */
788 static INLINE unsigned r600_pack_float_12p4(float x)
791 x >= 4096 ? 0xffff : x * 16;
794 static INLINE uint64_t r600_resource_va(struct pipe_screen *screen, struct pipe_resource *resource)
796 struct r600_screen *rscreen = (struct r600_screen*)screen;
797 struct r600_resource *rresource = (struct r600_resource*)resource;
799 return rscreen->ws->buffer_get_virtual_address(rresource->cs_buf);