r600g: implement MSAA depth-stencil decompression and resolve
[profile/ivi/mesa.git] / src / gallium / drivers / r600 / r600_pipe.h
1 /*
2  * Copyright 2010 Jerome Glisse <glisse@freedesktop.org>
3  *
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:
10  *
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
13  * Software.
14  *
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.
22  *
23  * Authors:
24  *      Jerome Glisse
25  */
26 #ifndef R600_PIPE_H
27 #define R600_PIPE_H
28
29 #include "util/u_slab.h"
30 #include "r600.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"
36
37 #define R600_MAX_CONST_BUFFERS 2
38 #define R600_MAX_CONST_BUFFER_SIZE 4096
39
40 #ifdef PIPE_ARCH_BIG_ENDIAN
41 #define R600_BIG_ENDIAN 1
42 #else
43 #define R600_BIG_ENDIAN 0
44 #endif
45
46 enum r600_atom_flags {
47         /* When set, atoms are added at the beginning of the dirty list
48          * instead of the end. */
49         EMIT_EARLY = (1 << 0)
50 };
51
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). */
55 struct r600_atom {
56         void (*emit)(struct r600_context *ctx, struct r600_atom *state);
57
58         unsigned                num_dw;
59         enum r600_atom_flags    flags;
60         bool                    dirty;
61
62         struct list_head        head;
63 };
64
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;
69         uint32_t *buf;
70         unsigned max_num_dw;
71         unsigned pkt_flags;
72 };
73
74 struct r600_surface_sync_cmd {
75         struct r600_atom atom;
76         unsigned flush_flags; /* CP_COHER_CNTL */
77 };
78
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;
84         unsigned copy_sample;
85 };
86
87 struct r600_cb_misc_state {
88         struct r600_atom atom;
89         unsigned cb_color_control; /* this comes from blend state */
90         unsigned blend_colormask; /* 8*4 bits for 8 RGBA colorbuffers */
91         unsigned nr_cbufs;
92         unsigned nr_ps_color_outputs;
93         bool multiwrite;
94         bool dual_src_blend;
95 };
96
97 struct r600_alphatest_state {
98         struct r600_atom atom;
99         unsigned sx_alpha_test_control; /* this comes from dsa state */
100         unsigned sx_alpha_ref; /* this comes from dsa state */
101         bool bypass;
102         bool cb0_export_16bpc; /* from set_framebuffer_state */
103 };
104
105 struct r600_cs_shader_state {
106         struct r600_atom atom;
107         struct r600_pipe_compute *shader;
108 };
109
110 struct r600_sample_mask {
111         struct r600_atom atom;
112         uint16_t sample_mask; /* there are only 8 bits on EG, 16 bits on Cayman */
113 };
114
115 enum r600_pipe_state_id {
116         R600_PIPE_STATE_BLEND = 0,
117         R600_PIPE_STATE_BLEND_COLOR,
118         R600_PIPE_STATE_CONFIG,
119         R600_PIPE_STATE_SEAMLESS_CUBEMAP,
120         R600_PIPE_STATE_CLIP,
121         R600_PIPE_STATE_SCISSOR,
122         R600_PIPE_STATE_VIEWPORT,
123         R600_PIPE_STATE_RASTERIZER,
124         R600_PIPE_STATE_VGT,
125         R600_PIPE_STATE_FRAMEBUFFER,
126         R600_PIPE_STATE_DSA,
127         R600_PIPE_STATE_STENCIL_REF,
128         R600_PIPE_STATE_PS_SHADER,
129         R600_PIPE_STATE_VS_SHADER,
130         R600_PIPE_STATE_CONSTANT,
131         R600_PIPE_STATE_SAMPLER,
132         R600_PIPE_STATE_RESOURCE,
133         R600_PIPE_STATE_POLYGON_OFFSET,
134         R600_PIPE_STATE_FETCH_SHADER,
135         R600_PIPE_STATE_SPI,
136         R600_PIPE_NSTATES
137 };
138
139 struct compute_memory_pool;
140 void compute_memory_pool_delete(struct compute_memory_pool* pool);
141 struct compute_memory_pool* compute_memory_pool_new(
142         struct r600_screen *rscreen);
143
144 struct r600_pipe_fences {
145         struct r600_resource            *bo;
146         unsigned                        *data;
147         unsigned                        next_index;
148         /* linked list of preallocated blocks */
149         struct list_head                blocks;
150         /* linked list of freed fences */
151         struct list_head                pool;
152         pipe_mutex                      mutex;
153 };
154
155 struct r600_screen {
156         struct pipe_screen              screen;
157         struct radeon_winsys            *ws;
158         unsigned                        family;
159         enum chip_class                 chip_class;
160         struct radeon_info              info;
161         bool                            has_streamout;
162         struct r600_tiling_info         tiling_info;
163         struct r600_pipe_fences         fences;
164
165         /*for compute global memory binding, we allocate stuff here, instead of
166          * buffers.
167          * XXX: Not sure if this is the best place for global_pool.  Also,
168          * it's not thread safe, so it won't work with multiple contexts. */
169         struct compute_memory_pool *global_pool;
170 };
171
172 struct r600_pipe_sampler_view {
173         struct pipe_sampler_view        base;
174         struct r600_resource            *tex_resource;
175         uint32_t                        tex_resource_words[8];
176 };
177
178 struct r600_pipe_rasterizer {
179         struct r600_pipe_state          rstate;
180         boolean                         flatshade;
181         boolean                         two_side;
182         unsigned                        sprite_coord_enable;
183         unsigned                        clip_plane_enable;
184         unsigned                        pa_sc_line_stipple;
185         unsigned                        pa_cl_clip_cntl;
186         float                           offset_units;
187         float                           offset_scale;
188         bool                            scissor_enable;
189         bool                            multisample_enable;
190 };
191
192 struct r600_pipe_blend {
193         struct r600_pipe_state          rstate;
194         unsigned                        cb_target_mask;
195         unsigned                        cb_color_control;
196         bool                            dual_src_blend;
197         bool                            alpha_to_one;
198 };
199
200 struct r600_pipe_dsa {
201         struct r600_pipe_state          rstate;
202         unsigned                        alpha_ref;
203         ubyte                           valuemask[2];
204         ubyte                           writemask[2];
205         unsigned                        sx_alpha_test_control;
206 };
207
208 struct r600_vertex_element
209 {
210         unsigned                        count;
211         struct pipe_vertex_element      elements[PIPE_MAX_ATTRIBS];
212         struct r600_resource            *fetch_shader;
213         unsigned                        fs_size;
214         struct r600_pipe_state          rstate;
215 };
216
217 struct r600_pipe_shader;
218
219 struct r600_pipe_shader_selector {
220         struct r600_pipe_shader *current;
221
222         struct tgsi_token       *tokens;
223         struct pipe_stream_output_info  so;
224
225         unsigned        num_shaders;
226
227         /* PIPE_SHADER_[VERTEX|FRAGMENT|...] */
228         unsigned        type;
229
230         unsigned        nr_ps_max_color_exports;
231 };
232
233 struct r600_pipe_shader {
234         struct r600_pipe_shader_selector *selector;
235         struct r600_pipe_shader *next_variant;
236         struct r600_shader              shader;
237         struct r600_pipe_state          rstate;
238         struct r600_resource            *bo;
239         struct r600_resource            *bo_fetch;
240         struct r600_vertex_element      vertex_elements;
241         unsigned        sprite_coord_enable;
242         unsigned        flatshade;
243         unsigned        pa_cl_vs_out_cntl;
244         unsigned        nr_ps_color_outputs;
245         unsigned        key;
246         unsigned                db_shader_control;
247         unsigned                ps_depth_export;
248 };
249
250 struct r600_pipe_sampler_state {
251         uint32_t                        tex_sampler_words[3];
252         uint32_t                        border_color[4];
253         bool                            border_color_use;
254         bool                            seamless_cube_map;
255 };
256
257 /* needed for blitter save */
258 #define NUM_TEX_UNITS 16
259
260 struct r600_seamless_cube_map {
261         struct r600_atom                atom;
262         bool                            enabled;
263 };
264
265 struct r600_samplerview_state {
266         struct r600_atom                atom;
267         struct r600_pipe_sampler_view   *views[NUM_TEX_UNITS];
268         uint32_t                        enabled_mask;
269         uint32_t                        dirty_mask;
270         uint32_t                        depth_texture_mask; /* which textures are depth */
271 };
272
273 struct r600_textures_info {
274         struct r600_samplerview_state   views;
275         struct r600_atom                atom_sampler;
276         struct r600_pipe_sampler_state  *samplers[NUM_TEX_UNITS];
277         unsigned                        n_samplers;
278         bool                            is_array_sampler[NUM_TEX_UNITS];
279 };
280
281 struct r600_fence {
282         struct pipe_reference           reference;
283         unsigned                        index; /* in the shared bo */
284         struct r600_resource            *sleep_bo;
285         struct list_head                head;
286 };
287
288 #define FENCE_BLOCK_SIZE 16
289
290 struct r600_fence_block {
291         struct r600_fence               fences[FENCE_BLOCK_SIZE];
292         struct list_head                head;
293 };
294
295 #define R600_CONSTANT_ARRAY_SIZE 256
296 #define R600_RESOURCE_ARRAY_SIZE 160
297
298 struct r600_stencil_ref
299 {
300         ubyte ref_value[2];
301         ubyte valuemask[2];
302         ubyte writemask[2];
303 };
304
305 struct r600_constbuf_state
306 {
307         struct r600_atom                atom;
308         struct pipe_constant_buffer     cb[PIPE_MAX_CONSTANT_BUFFERS];
309         uint32_t                        enabled_mask;
310         uint32_t                        dirty_mask;
311 };
312
313 struct r600_vertexbuf_state
314 {
315         struct r600_atom                atom;
316         struct pipe_vertex_buffer       vb[PIPE_MAX_ATTRIBS];
317         uint32_t                        enabled_mask; /* non-NULL buffers */
318         uint32_t                        dirty_mask;
319 };
320
321 struct r600_context {
322         struct pipe_context             context;
323         struct blitter_context          *blitter;
324         enum radeon_family              family;
325         enum chip_class                 chip_class;
326         boolean                         has_vertex_cache;
327         unsigned                        r6xx_num_clause_temp_gprs;
328         void                            *custom_dsa_flush;
329         struct r600_screen              *screen;
330         struct radeon_winsys            *ws;
331         struct r600_pipe_state          *states[R600_PIPE_NSTATES];
332         struct r600_vertex_element      *vertex_elements;
333         struct pipe_framebuffer_state   framebuffer;
334         unsigned                        compute_cb_target_mask;
335         unsigned                        db_shader_control;
336         unsigned                        pa_sc_line_stipple;
337         unsigned                        pa_cl_clip_cntl;
338         /* for saving when using blitter */
339         struct pipe_stencil_ref         stencil_ref;
340         struct pipe_viewport_state      viewport;
341         struct pipe_clip_state          clip;
342         struct r600_pipe_shader_selector        *ps_shader;
343         struct r600_pipe_shader_selector        *vs_shader;
344         struct r600_pipe_rasterizer     *rasterizer;
345         struct r600_pipe_state          vgt;
346         struct r600_pipe_state          spi;
347         struct pipe_query               *current_render_cond;
348         unsigned                        current_render_cond_mode;
349         struct pipe_query               *saved_render_cond;
350         unsigned                        saved_render_cond_mode;
351         /* shader information */
352         boolean                         two_side;
353         boolean                         spi_dirty;
354         unsigned                        sprite_coord_enable;
355         boolean                         flatshade;
356         boolean                         export_16bpc;
357         unsigned                        nr_cbufs;
358         bool                            alpha_to_one;
359         bool                            multisample_enable;
360         bool                            cb0_is_integer;
361
362         struct u_upload_mgr             *uploader;
363         struct util_slab_mempool        pool_transfers;
364
365         unsigned default_ps_gprs, default_vs_gprs;
366
367         /* States based on r600_atom. */
368         struct list_head                dirty_states;
369         struct r600_command_buffer      start_cs_cmd; /* invariant state mostly */
370         /** Compute specific registers initializations.  The start_cs_cmd atom
371          *  must be emitted before start_compute_cs_cmd. */
372         struct r600_command_buffer      start_compute_cs_cmd;
373         struct r600_surface_sync_cmd    surface_sync_cmd;
374         struct r600_atom                r6xx_flush_and_inv_cmd;
375         struct r600_alphatest_state     alphatest_state;
376         struct r600_cb_misc_state       cb_misc_state;
377         struct r600_db_misc_state       db_misc_state;
378         /** Vertex buffers for fetch shaders */
379         struct r600_vertexbuf_state     vertex_buffer_state;
380         /** Vertex buffers for compute shaders */
381         struct r600_vertexbuf_state     cs_vertex_buffer_state;
382         struct r600_constbuf_state      vs_constbuf_state;
383         struct r600_constbuf_state      ps_constbuf_state;
384         struct r600_textures_info       vs_samplers;
385         struct r600_textures_info       ps_samplers;
386         struct r600_seamless_cube_map   seamless_cube_map;
387         struct r600_cs_shader_state     cs_shader_state;
388         struct r600_sample_mask         sample_mask;
389
390         struct radeon_winsys_cs *cs;
391
392         struct r600_range       *range;
393         unsigned                nblocks;
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;
399
400         /* The list of active queries. Only one query of each type can be active. */
401         int                     num_occlusion_queries;
402
403         /* Manage queries in two separate groups:
404          * The timer ones and the others (streamout, occlusion).
405          *
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;
413
414         unsigned                num_cs_dw_streamout_end;
415
416         unsigned                backend_mask;
417         unsigned                max_db; /* for OQ */
418         unsigned                flags;
419         boolean                 predicate_drawing;
420
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;
425
426         /* There is no scissor enable bit on r6xx, so we must use a workaround.
427          * These track the current scissor state. */
428         bool                    scissor_enable;
429         struct pipe_scissor_state scissor_state;
430
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;
434
435         boolean                 dual_src_blend;
436
437         /* Index buffer. */
438         struct pipe_index_buffer index_buffer;
439 };
440
441 static INLINE void r600_emit_atom(struct r600_context *rctx, struct r600_atom *atom)
442 {
443         atom->emit(rctx, atom);
444         atom->dirty = false;
445         if (atom->head.next && atom->head.prev)
446                 LIST_DELINIT(&atom->head);
447 }
448
449 static INLINE void r600_atom_dirty(struct r600_context *rctx, struct r600_atom *state)
450 {
451         if (!state->dirty) {
452                 if (state->flags & EMIT_EARLY) {
453                         LIST_ADD(&state->head, &rctx->dirty_states);
454                 } else {
455                         LIST_ADDTAIL(&state->head, &rctx->dirty_states);
456                 }
457                 state->dirty = true;
458         }
459 }
460
461 /* evergreen_state.c */
462 void evergreen_init_state_functions(struct r600_context *rctx);
463 void evergreen_init_atom_start_cs(struct r600_context *rctx);
464 void evergreen_pipe_shader_ps(struct pipe_context *ctx, struct r600_pipe_shader *shader);
465 void evergreen_pipe_shader_vs(struct pipe_context *ctx, struct r600_pipe_shader *shader);
466 void evergreen_fetch_shader(struct pipe_context *ctx, struct r600_vertex_element *ve);
467 void *evergreen_create_db_flush_dsa(struct r600_context *rctx);
468 void evergreen_polygon_offset_update(struct r600_context *rctx);
469 boolean evergreen_is_format_supported(struct pipe_screen *screen,
470                                       enum pipe_format format,
471                                       enum pipe_texture_target target,
472                                       unsigned sample_count,
473                                       unsigned usage);
474 void evergreen_init_color_surface(struct r600_context *rctx,
475                                   struct r600_surface *surf);
476 void evergreen_update_dual_export_state(struct r600_context * rctx);
477
478 /* r600_blit.c */
479 void r600_copy_buffer(struct pipe_context *ctx, struct
480                       pipe_resource *dst, unsigned dstx,
481                       struct pipe_resource *src, const struct pipe_box *src_box);
482 void r600_init_blit_functions(struct r600_context *rctx);
483 void r600_blit_uncompress_depth(struct pipe_context *ctx,
484                 struct r600_resource_texture *texture,
485                 struct r600_resource_texture *staging,
486                 unsigned first_level, unsigned last_level,
487                 unsigned first_layer, unsigned last_layer,
488                 unsigned first_sample, unsigned last_sample);
489 void r600_flush_depth_textures(struct r600_context *rctx,
490                                struct r600_samplerview_state *textures);
491
492 /* r600_buffer.c */
493 bool r600_init_resource(struct r600_screen *rscreen,
494                         struct r600_resource *res,
495                         unsigned size, unsigned alignment,
496                         unsigned bind, unsigned usage);
497 struct pipe_resource *r600_buffer_create(struct pipe_screen *screen,
498                                          const struct pipe_resource *templ);
499
500 /* r600_pipe.c */
501 void r600_flush(struct pipe_context *ctx, struct pipe_fence_handle **fence,
502                 unsigned flags);
503
504 /* r600_query.c */
505 void r600_init_query_functions(struct r600_context *rctx);
506 void r600_suspend_nontimer_queries(struct r600_context *ctx);
507 void r600_resume_nontimer_queries(struct r600_context *ctx);
508 void r600_suspend_timer_queries(struct r600_context *ctx);
509 void r600_resume_timer_queries(struct r600_context *ctx);
510
511 /* r600_resource.c */
512 void r600_init_context_resource_functions(struct r600_context *r600);
513
514 /* r600_shader.c */
515 int r600_pipe_shader_create(struct pipe_context *ctx, struct r600_pipe_shader *shader);
516 #ifdef HAVE_OPENCL
517 int r600_compute_shader_create(struct pipe_context * ctx,
518         LLVMModuleRef mod,  struct r600_bytecode * bytecode);
519 #endif
520 void r600_pipe_shader_destroy(struct pipe_context *ctx, struct r600_pipe_shader *shader);
521
522 /* r600_state.c */
523 void r600_set_scissor_state(struct r600_context *rctx,
524                             const struct pipe_scissor_state *state);
525 void r600_init_state_functions(struct r600_context *rctx);
526 void r600_init_atom_start_cs(struct r600_context *rctx);
527 void r600_pipe_shader_ps(struct pipe_context *ctx, struct r600_pipe_shader *shader);
528 void r600_pipe_shader_vs(struct pipe_context *ctx, struct r600_pipe_shader *shader);
529 void r600_fetch_shader(struct pipe_context *ctx, struct r600_vertex_element *ve);
530 void *r600_create_db_flush_dsa(struct r600_context *rctx);
531 void r600_polygon_offset_update(struct r600_context *rctx);
532 void r600_adjust_gprs(struct r600_context *rctx);
533 boolean r600_is_format_supported(struct pipe_screen *screen,
534                                  enum pipe_format format,
535                                  enum pipe_texture_target target,
536                                  unsigned sample_count,
537                                  unsigned usage);
538 void r600_update_dual_export_state(struct r600_context * rctx);
539
540 /* r600_texture.c */
541 void r600_init_screen_texture_functions(struct pipe_screen *screen);
542 void r600_init_surface_functions(struct r600_context *r600);
543 uint32_t r600_translate_texformat(struct pipe_screen *screen, enum pipe_format format,
544                                   const unsigned char *swizzle_view,
545                                   uint32_t *word4_p, uint32_t *yuv_format_p);
546 unsigned r600_texture_get_offset(struct r600_resource_texture *rtex,
547                                         unsigned level, unsigned layer);
548
549 /* r600_translate.c */
550 void r600_translate_index_buffer(struct r600_context *r600,
551                                  struct pipe_index_buffer *ib,
552                                  unsigned count);
553
554 /* r600_state_common.c */
555 void r600_init_atom(struct r600_atom *atom,
556                     void (*emit)(struct r600_context *ctx, struct r600_atom *state),
557                     unsigned num_dw, enum r600_atom_flags flags);
558 void r600_init_common_atoms(struct r600_context *rctx);
559 unsigned r600_get_cb_flush_flags(struct r600_context *rctx);
560 void r600_texture_barrier(struct pipe_context *ctx);
561 void r600_set_index_buffer(struct pipe_context *ctx,
562                            const struct pipe_index_buffer *ib);
563 void r600_vertex_buffers_dirty(struct r600_context *rctx);
564 void r600_set_vertex_buffers(struct pipe_context *ctx, unsigned count,
565                              const struct pipe_vertex_buffer *input);
566 void r600_sampler_views_dirty(struct r600_context *rctx,
567                               struct r600_samplerview_state *state);
568 void r600_set_sampler_views(struct r600_context *rctx,
569                             struct r600_textures_info *dst,
570                             unsigned count,
571                             struct pipe_sampler_view **views);
572 void r600_bind_vs_samplers(struct pipe_context *ctx, unsigned count, void **states);
573 void r600_bind_ps_samplers(struct pipe_context *ctx, unsigned count, void **states);
574 void *r600_create_vertex_elements(struct pipe_context *ctx,
575                                   unsigned count,
576                                   const struct pipe_vertex_element *elements);
577 void r600_delete_vertex_element(struct pipe_context *ctx, void *state);
578 void r600_bind_blend_state(struct pipe_context *ctx, void *state);
579 void r600_set_blend_color(struct pipe_context *ctx,
580                           const struct pipe_blend_color *state);
581 void r600_bind_dsa_state(struct pipe_context *ctx, void *state);
582 void r600_set_max_scissor(struct r600_context *rctx);
583 void r600_bind_rs_state(struct pipe_context *ctx, void *state);
584 void r600_delete_rs_state(struct pipe_context *ctx, void *state);
585 void r600_sampler_view_destroy(struct pipe_context *ctx,
586                                struct pipe_sampler_view *state);
587 void r600_delete_sampler(struct pipe_context *ctx, void *state);
588 void r600_delete_state(struct pipe_context *ctx, void *state);
589 void r600_bind_vertex_elements(struct pipe_context *ctx, void *state);
590 void *r600_create_shader_state_ps(struct pipe_context *ctx,
591                    const struct pipe_shader_state *state);
592 void *r600_create_shader_state_vs(struct pipe_context *ctx,
593                    const struct pipe_shader_state *state);
594 void r600_bind_ps_shader(struct pipe_context *ctx, void *state);
595 void r600_bind_vs_shader(struct pipe_context *ctx, void *state);
596 void r600_delete_ps_shader(struct pipe_context *ctx, void *state);
597 void r600_delete_vs_shader(struct pipe_context *ctx, void *state);
598 void r600_constant_buffers_dirty(struct r600_context *rctx, struct r600_constbuf_state *state);
599 void r600_set_constant_buffer(struct pipe_context *ctx, uint shader, uint index,
600                               struct pipe_constant_buffer *cb);
601 struct pipe_stream_output_target *
602 r600_create_so_target(struct pipe_context *ctx,
603                       struct pipe_resource *buffer,
604                       unsigned buffer_offset,
605                       unsigned buffer_size);
606 void r600_so_target_destroy(struct pipe_context *ctx,
607                             struct pipe_stream_output_target *target);
608 void r600_set_so_targets(struct pipe_context *ctx,
609                          unsigned num_targets,
610                          struct pipe_stream_output_target **targets,
611                          unsigned append_bitmask);
612 void r600_set_sample_mask(struct pipe_context *pipe, unsigned sample_mask);
613 void r600_set_pipe_stencil_ref(struct pipe_context *ctx,
614                                const struct pipe_stencil_ref *state);
615 void r600_draw_vbo(struct pipe_context *ctx, const struct pipe_draw_info *info);
616 uint32_t r600_translate_stencil_op(int s_op);
617 uint32_t r600_translate_fill(uint32_t func);
618 unsigned r600_tex_wrap(unsigned wrap);
619 unsigned r600_tex_filter(unsigned filter);
620 unsigned r600_tex_mipfilter(unsigned filter);
621 unsigned r600_tex_compare(unsigned compare);
622
623 /*
624  * Helpers for building command buffers
625  */
626
627 #define PKT3_SET_CONFIG_REG     0x68
628 #define PKT3_SET_CONTEXT_REG    0x69
629 #define PKT3_SET_CTL_CONST      0x6F
630 #define PKT3_SET_LOOP_CONST                    0x6C
631
632 #define R600_CONFIG_REG_OFFSET  0x08000
633 #define R600_CONTEXT_REG_OFFSET 0x28000
634 #define R600_CTL_CONST_OFFSET   0x3CFF0
635 #define R600_LOOP_CONST_OFFSET                 0X0003E200
636 #define EG_LOOP_CONST_OFFSET               0x0003A200
637
638 #define PKT_TYPE_S(x)                   (((x) & 0x3) << 30)
639 #define PKT_COUNT_S(x)                  (((x) & 0x3FFF) << 16)
640 #define PKT3_IT_OPCODE_S(x)             (((x) & 0xFF) << 8)
641 #define PKT3_PREDICATE(x)               (((x) >> 0) & 0x1)
642 #define PKT3(op, count, predicate) (PKT_TYPE_S(3) | PKT_COUNT_S(count) | PKT3_IT_OPCODE_S(op) | PKT3_PREDICATE(predicate))
643
644 #define RADEON_CP_PACKET3_COMPUTE_MODE 0x00000002
645
646 /*Evergreen Compute packet3*/
647 #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)
648
649 static INLINE void r600_store_value(struct r600_command_buffer *cb, unsigned value)
650 {
651         cb->buf[cb->atom.num_dw++] = value;
652 }
653
654 static INLINE void r600_store_config_reg_seq(struct r600_command_buffer *cb, unsigned reg, unsigned num)
655 {
656         assert(reg < R600_CONTEXT_REG_OFFSET);
657         assert(cb->atom.num_dw+2+num <= cb->max_num_dw);
658         cb->buf[cb->atom.num_dw++] = PKT3(PKT3_SET_CONFIG_REG, num, 0);
659         cb->buf[cb->atom.num_dw++] = (reg - R600_CONFIG_REG_OFFSET) >> 2;
660 }
661
662 /**
663  * Needs cb->pkt_flags set to  RADEON_CP_PACKET3_COMPUTE_MODE for compute
664  * shaders.
665  */
666 static INLINE void r600_store_context_reg_seq(struct r600_command_buffer *cb, unsigned reg, unsigned num)
667 {
668         assert(reg >= R600_CONTEXT_REG_OFFSET && reg < R600_CTL_CONST_OFFSET);
669         assert(cb->atom.num_dw+2+num <= cb->max_num_dw);
670         cb->buf[cb->atom.num_dw++] = PKT3(PKT3_SET_CONTEXT_REG, num, 0) | cb->pkt_flags;
671         cb->buf[cb->atom.num_dw++] = (reg - R600_CONTEXT_REG_OFFSET) >> 2;
672 }
673
674 /**
675  * Needs cb->pkt_flags set to  RADEON_CP_PACKET3_COMPUTE_MODE for compute
676  * shaders.
677  */
678 static INLINE void r600_store_ctl_const_seq(struct r600_command_buffer *cb, unsigned reg, unsigned num)
679 {
680         assert(reg >= R600_CTL_CONST_OFFSET);
681         assert(cb->atom.num_dw+2+num <= cb->max_num_dw);
682         cb->buf[cb->atom.num_dw++] = PKT3(PKT3_SET_CTL_CONST, num, 0) | cb->pkt_flags;
683         cb->buf[cb->atom.num_dw++] = (reg - R600_CTL_CONST_OFFSET) >> 2;
684 }
685
686 static INLINE void r600_store_loop_const_seq(struct r600_command_buffer *cb, unsigned reg, unsigned num)
687 {
688         assert(reg >= R600_LOOP_CONST_OFFSET);
689         assert(cb->atom.num_dw+2+num <= cb->max_num_dw);
690         cb->buf[cb->atom.num_dw++] = PKT3(PKT3_SET_LOOP_CONST, num, 0);
691         cb->buf[cb->atom.num_dw++] = (reg - R600_LOOP_CONST_OFFSET) >> 2;
692 }
693
694 /**
695  * Needs cb->pkt_flags set to  RADEON_CP_PACKET3_COMPUTE_MODE for compute
696  * shaders.
697  */
698 static INLINE void eg_store_loop_const_seq(struct r600_command_buffer *cb, unsigned reg, unsigned num)
699 {
700         assert(reg >= EG_LOOP_CONST_OFFSET);
701         assert(cb->atom.num_dw+2+num <= cb->max_num_dw);
702         cb->buf[cb->atom.num_dw++] = PKT3(PKT3_SET_LOOP_CONST, num, 0) | cb->pkt_flags;
703         cb->buf[cb->atom.num_dw++] = (reg - EG_LOOP_CONST_OFFSET) >> 2;
704 }
705
706 static INLINE void r600_store_config_reg(struct r600_command_buffer *cb, unsigned reg, unsigned value)
707 {
708         r600_store_config_reg_seq(cb, reg, 1);
709         r600_store_value(cb, value);
710 }
711
712 static INLINE void r600_store_context_reg(struct r600_command_buffer *cb, unsigned reg, unsigned value)
713 {
714         r600_store_context_reg_seq(cb, reg, 1);
715         r600_store_value(cb, value);
716 }
717
718 static INLINE void r600_store_ctl_const(struct r600_command_buffer *cb, unsigned reg, unsigned value)
719 {
720         r600_store_ctl_const_seq(cb, reg, 1);
721         r600_store_value(cb, value);
722 }
723
724 static INLINE void r600_store_loop_const(struct r600_command_buffer *cb, unsigned reg, unsigned value)
725 {
726         r600_store_loop_const_seq(cb, reg, 1);
727         r600_store_value(cb, value);
728 }
729
730 static INLINE void eg_store_loop_const(struct r600_command_buffer *cb, unsigned reg, unsigned value)
731 {
732         eg_store_loop_const_seq(cb, reg, 1);
733         r600_store_value(cb, value);
734 }
735
736 void r600_init_command_buffer(struct r600_command_buffer *cb, unsigned num_dw, enum r600_atom_flags flags);
737 void r600_release_command_buffer(struct r600_command_buffer *cb);
738
739 /*
740  * Helpers for emitting state into a command stream directly.
741  */
742
743 static INLINE unsigned r600_context_bo_reloc(struct r600_context *ctx, struct r600_resource *rbo,
744                                              enum radeon_bo_usage usage)
745 {
746         assert(usage);
747         return ctx->ws->cs_add_reloc(ctx->cs, rbo->cs_buf, usage, rbo->domains) * 4;
748 }
749
750 static INLINE void r600_write_value(struct radeon_winsys_cs *cs, unsigned value)
751 {
752         cs->buf[cs->cdw++] = value;
753 }
754
755 static INLINE void r600_write_array(struct radeon_winsys_cs *cs, unsigned num, unsigned *ptr)
756 {
757         assert(cs->cdw+num <= RADEON_MAX_CMDBUF_DWORDS);
758         memcpy(&cs->buf[cs->cdw], ptr, num * sizeof(ptr[0]));
759         cs->cdw += num;
760 }
761
762 static INLINE void r600_write_config_reg_seq(struct radeon_winsys_cs *cs, unsigned reg, unsigned num)
763 {
764         assert(reg < R600_CONTEXT_REG_OFFSET);
765         assert(cs->cdw+2+num <= RADEON_MAX_CMDBUF_DWORDS);
766         cs->buf[cs->cdw++] = PKT3(PKT3_SET_CONFIG_REG, num, 0);
767         cs->buf[cs->cdw++] = (reg - R600_CONFIG_REG_OFFSET) >> 2;
768 }
769
770 static INLINE void r600_write_context_reg_seq(struct radeon_winsys_cs *cs, unsigned reg, unsigned num)
771 {
772         assert(reg >= R600_CONTEXT_REG_OFFSET && reg < R600_CTL_CONST_OFFSET);
773         assert(cs->cdw+2+num <= RADEON_MAX_CMDBUF_DWORDS);
774         cs->buf[cs->cdw++] = PKT3(PKT3_SET_CONTEXT_REG, num, 0);
775         cs->buf[cs->cdw++] = (reg - R600_CONTEXT_REG_OFFSET) >> 2;
776 }
777
778 static INLINE void r600_write_compute_context_reg_seq(struct radeon_winsys_cs *cs, unsigned reg, unsigned num)
779 {
780         r600_write_context_reg_seq(cs, reg, num);
781         /* Set the compute bit on the packet header */
782         cs->buf[cs->cdw - 2] |= RADEON_CP_PACKET3_COMPUTE_MODE;
783 }
784
785 static INLINE void r600_write_ctl_const_seq(struct radeon_winsys_cs *cs, unsigned reg, unsigned num)
786 {
787         assert(reg >= R600_CTL_CONST_OFFSET);
788         assert(cs->cdw+2+num <= RADEON_MAX_CMDBUF_DWORDS);
789         cs->buf[cs->cdw++] = PKT3(PKT3_SET_CTL_CONST, num, 0);
790         cs->buf[cs->cdw++] = (reg - R600_CTL_CONST_OFFSET) >> 2;
791 }
792
793 static INLINE void r600_write_config_reg(struct radeon_winsys_cs *cs, unsigned reg, unsigned value)
794 {
795         r600_write_config_reg_seq(cs, reg, 1);
796         r600_write_value(cs, value);
797 }
798
799 static INLINE void r600_write_context_reg(struct radeon_winsys_cs *cs, unsigned reg, unsigned value)
800 {
801         r600_write_context_reg_seq(cs, reg, 1);
802         r600_write_value(cs, value);
803 }
804
805 static INLINE void r600_write_compute_context_reg(struct radeon_winsys_cs *cs, unsigned reg, unsigned value)
806 {
807         r600_write_compute_context_reg_seq(cs, reg, 1);
808         r600_write_value(cs, value);
809 }
810
811 static INLINE void r600_write_ctl_const(struct radeon_winsys_cs *cs, unsigned reg, unsigned value)
812 {
813         r600_write_ctl_const_seq(cs, reg, 1);
814         r600_write_value(cs, value);
815 }
816
817 /*
818  * common helpers
819  */
820 static INLINE uint32_t S_FIXED(float value, uint32_t frac_bits)
821 {
822         return value * (1 << frac_bits);
823 }
824 #define ALIGN_DIVUP(x, y) (((x) + (y) - 1) / (y))
825
826 static inline unsigned r600_tex_aniso_filter(unsigned filter)
827 {
828         if (filter <= 1)   return 0;
829         if (filter <= 2)   return 1;
830         if (filter <= 4)   return 2;
831         if (filter <= 8)   return 3;
832          /* else */        return 4;
833 }
834
835 /* 12.4 fixed-point */
836 static INLINE unsigned r600_pack_float_12p4(float x)
837 {
838         return x <= 0    ? 0 :
839                x >= 4096 ? 0xffff : x * 16;
840 }
841
842 static INLINE uint64_t r600_resource_va(struct pipe_screen *screen, struct pipe_resource *resource)
843 {
844         struct r600_screen *rscreen = (struct r600_screen*)screen;
845         struct r600_resource *rresource = (struct r600_resource*)resource;
846
847         return rscreen->ws->buffer_get_virtual_address(rresource->cs_buf);
848 }
849
850 #endif