1 /*
2 * Copyright © 2014-2017 Broadcom
3 * Copyright (C) 2012 Rob Clark <robclark@freedesktop.org>
4 *
5 * Permission is hereby granted, free of charge, to any person obtaining a
6 * copy of this software and associated documentation files (the "Software"),
7 * to deal in the Software without restriction, including without limitation
8 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
9 * and/or sell copies of the Software, and to permit persons to whom the
10 * Software is furnished to do so, subject to the following conditions:
11 *
12 * The above copyright notice and this permission notice (including the next
13 * paragraph) shall be included in all copies or substantial portions of the
14 * Software.
15 *
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
21 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
22 * IN THE SOFTWARE.
23 */
24
25 #ifndef V3D_CONTEXT_H
26 #define V3D_CONTEXT_H
27
28 #ifdef V3D_VERSION
29 #include "broadcom/common/v3d_macros.h"
30 #endif
31
32 #include <stdio.h>
33
34 #include "pipe/p_context.h"
35 #include "pipe/p_state.h"
36 #include "util/bitset.h"
37 #include "util/slab.h"
38 #include "xf86drm.h"
39 #include "drm-uapi/v3d_drm.h"
40 #include "v3d_screen.h"
41 #include "broadcom/common/v3d_limits.h"
42
43 #include "broadcom/simulator/v3d_simulator.h"
44
45 struct v3d_job;
46 struct v3d_bo;
47 void v3d_job_add_bo(struct v3d_job *job, struct v3d_bo *bo);
48
49 #include "v3d_bufmgr.h"
50 #include "v3d_resource.h"
51 #include "v3d_cl.h"
52
53 #ifdef USE_V3D_SIMULATOR
54 #define using_v3d_simulator true
55 #else
56 #define using_v3d_simulator false
57 #endif
58
59 #define V3D_DIRTY_BLEND (1ull << 0)
60 #define V3D_DIRTY_RASTERIZER (1ull << 1)
61 #define V3D_DIRTY_ZSA (1ull << 2)
62 #define V3D_DIRTY_COMPTEX (1ull << 3)
63 #define V3D_DIRTY_VERTTEX (1ull << 4)
64 #define V3D_DIRTY_GEOMTEX (1ull << 5)
65 #define V3D_DIRTY_FRAGTEX (1ull << 6)
66
67 #define V3D_DIRTY_SHADER_IMAGE (1ull << 9)
68 #define V3D_DIRTY_BLEND_COLOR (1ull << 10)
69 #define V3D_DIRTY_STENCIL_REF (1ull << 11)
70 #define V3D_DIRTY_SAMPLE_STATE (1ull << 12)
71 #define V3D_DIRTY_FRAMEBUFFER (1ull << 13)
72 #define V3D_DIRTY_STIPPLE (1ull << 14)
73 #define V3D_DIRTY_VIEWPORT (1ull << 15)
74 #define V3D_DIRTY_CONSTBUF (1ull << 16)
75 #define V3D_DIRTY_VTXSTATE (1ull << 17)
76 #define V3D_DIRTY_VTXBUF (1ull << 18)
77 #define V3D_DIRTY_SCISSOR (1ull << 19)
78 #define V3D_DIRTY_FLAT_SHADE_FLAGS (1ull << 20)
79 #define V3D_DIRTY_PRIM_MODE (1ull << 21)
80 #define V3D_DIRTY_CLIP (1ull << 22)
81 #define V3D_DIRTY_UNCOMPILED_CS (1ull << 23)
82 #define V3D_DIRTY_UNCOMPILED_VS (1ull << 24)
83 #define V3D_DIRTY_UNCOMPILED_GS (1ull << 25)
84 #define V3D_DIRTY_UNCOMPILED_FS (1ull << 26)
85
86 #define V3D_DIRTY_COMPILED_CS (1ull << 29)
87 #define V3D_DIRTY_COMPILED_VS (1ull << 30)
88 #define V3D_DIRTY_COMPILED_GS_BIN (1ULL << 31)
89 #define V3D_DIRTY_COMPILED_GS (1ULL << 32)
90 #define V3D_DIRTY_COMPILED_FS (1ull << 33)
91
92 #define V3D_DIRTY_FS_INPUTS (1ull << 38)
93 #define V3D_DIRTY_GS_INPUTS (1ull << 39)
94 #define V3D_DIRTY_STREAMOUT (1ull << 40)
95 #define V3D_DIRTY_OQ (1ull << 41)
96 #define V3D_DIRTY_CENTROID_FLAGS (1ull << 42)
97 #define V3D_DIRTY_NOPERSPECTIVE_FLAGS (1ull << 43)
98 #define V3D_DIRTY_SSBO (1ull << 44)
99
100 #define V3D_MAX_FS_INPUTS 64
101
102 #define MAX_JOB_SCISSORS 16
103
104 enum v3d_sampler_state_variant {
105 V3D_SAMPLER_STATE_BORDER_0,
106 V3D_SAMPLER_STATE_F16,
107 V3D_SAMPLER_STATE_F16_UNORM,
108 V3D_SAMPLER_STATE_F16_SNORM,
109 V3D_SAMPLER_STATE_F16_BGRA,
110 V3D_SAMPLER_STATE_F16_BGRA_UNORM,
111 V3D_SAMPLER_STATE_F16_BGRA_SNORM,
112 V3D_SAMPLER_STATE_F16_A,
113 V3D_SAMPLER_STATE_F16_A_SNORM,
114 V3D_SAMPLER_STATE_F16_A_UNORM,
115 V3D_SAMPLER_STATE_F16_LA,
116 V3D_SAMPLER_STATE_F16_LA_UNORM,
117 V3D_SAMPLER_STATE_F16_LA_SNORM,
118 V3D_SAMPLER_STATE_32,
119 V3D_SAMPLER_STATE_32_UNORM,
120 V3D_SAMPLER_STATE_32_SNORM,
121 V3D_SAMPLER_STATE_32_A,
122 V3D_SAMPLER_STATE_32_A_UNORM,
123 V3D_SAMPLER_STATE_32_A_SNORM,
124 V3D_SAMPLER_STATE_1010102U,
125 V3D_SAMPLER_STATE_16U,
126 V3D_SAMPLER_STATE_16I,
127 V3D_SAMPLER_STATE_8I,
128 V3D_SAMPLER_STATE_8U,
129
130 V3D_SAMPLER_STATE_VARIANT_COUNT,
131 };
132
133 enum v3d_flush_cond {
134 /* Flush job unless we are flushing for transform feedback, where we
135 * handle flushing in the driver via the 'Wait for TF' packet.
136 */
137 V3D_FLUSH_DEFAULT,
138 /* Always flush the job, even for cases where we would normally not
139 * do it, such as transform feedback.
140 */
141 V3D_FLUSH_ALWAYS,
142 /* Flush job if it is not the current FBO job. This is intended to
143 * skip automatic flushes of the current job for resources that we
144 * expect to be externally synchronized by the application using
145 * glMemoryBarrier(), such as SSBOs and shader images.
146 */
147 V3D_FLUSH_NOT_CURRENT_JOB,
148 };
149
150 struct v3d_sampler_view {
151 struct pipe_sampler_view base;
152 uint32_t p0;
153 uint32_t p1;
154 /* Precomputed swizzles to pass in to the shader key. */
155 uint8_t swizzle[4];
156
157 uint8_t texture_shader_state[32];
158 /* V3D 4.x: Texture state struct. */
159 struct v3d_bo *bo;
160
161 enum v3d_sampler_state_variant sampler_variant;
162
163 /* Actual texture to be read by this sampler view. May be different
164 * from base.texture in the case of having a shadow tiled copy of a
165 * raster texture.
166 */
167 struct pipe_resource *texture;
168 };
169
170 struct v3d_sampler_state {
171 struct pipe_sampler_state base;
172 uint32_t p0;
173 uint32_t p1;
174
175 /* V3D 3.x: Packed texture state. */
176 uint8_t texture_shader_state[32];
177 /* V3D 4.x: Sampler state struct. */
178 struct pipe_resource *sampler_state;
179 uint32_t sampler_state_offset[V3D_SAMPLER_STATE_VARIANT_COUNT];
180
181 bool border_color_variants;
182 };
183
184 struct v3d_texture_stateobj {
185 struct pipe_sampler_view *textures[V3D_MAX_TEXTURE_SAMPLERS];
186 unsigned num_textures;
187 struct pipe_sampler_state *samplers[V3D_MAX_TEXTURE_SAMPLERS];
188 unsigned num_samplers;
189 struct v3d_cl_reloc texture_state[V3D_MAX_TEXTURE_SAMPLERS];
190 };
191
192 struct v3d_shader_uniform_info {
193 enum quniform_contents *contents;
194 uint32_t *data;
195 uint32_t count;
196 };
197
198 struct v3d_uncompiled_shader {
199 /** A name for this program, so you can track it in shader-db output. */
200 uint32_t program_id;
201 /** How many variants of this program were compiled, for shader-db. */
202 uint32_t compiled_variant_count;
203 struct pipe_shader_state base;
204 uint32_t num_tf_outputs;
205 struct v3d_varying_slot *tf_outputs;
206 uint16_t tf_specs[16];
207 uint16_t tf_specs_psiz[16];
208 uint32_t num_tf_specs;
209 };
210
211 struct v3d_compiled_shader {
212 struct pipe_resource *resource;
213 uint32_t offset;
214
215 union {
216 struct v3d_prog_data *base;
217 struct v3d_vs_prog_data *vs;
218 struct v3d_gs_prog_data *gs;
219 struct v3d_fs_prog_data *fs;
220 struct v3d_compute_prog_data *compute;
221 } prog_data;
222
223 /**
224 * V3D_DIRTY_* flags that, when set in v3d->dirty, mean that the
225 * uniforms have to be rewritten (and therefore the shader state
226 * reemitted).
227 */
228 uint64_t uniform_dirty_bits;
229 };
230
231 struct v3d_program_stateobj {
232 struct v3d_uncompiled_shader *bind_vs, *bind_gs, *bind_fs, *bind_compute;
233 struct v3d_compiled_shader *cs, *vs, *gs_bin, *gs, *fs, *compute;
234
235 struct hash_table *cache[MESA_SHADER_STAGES];
236
237 struct v3d_bo *spill_bo;
238 int spill_size_per_thread;
239 };
240
241 struct v3d_constbuf_stateobj {
242 struct pipe_constant_buffer cb[PIPE_MAX_CONSTANT_BUFFERS];
243 uint32_t enabled_mask;
244 uint32_t dirty_mask;
245 };
246
247 struct v3d_vertexbuf_stateobj {
248 struct pipe_vertex_buffer vb[PIPE_MAX_ATTRIBS];
249 unsigned count;
250 uint32_t enabled_mask;
251 uint32_t dirty_mask;
252 };
253
254 struct v3d_vertex_stateobj {
255 struct pipe_vertex_element pipe[V3D_MAX_VS_INPUTS / 4];
256 unsigned num_elements;
257
258 uint8_t attrs[16 * (V3D_MAX_VS_INPUTS / 4)];
259 struct pipe_resource *defaults;
260 uint32_t defaults_offset;
261 };
262
263 struct v3d_stream_output_target {
264 struct pipe_stream_output_target base;
265 /* Number of transform feedback vertices written to this target */
266 uint32_t recorded_vertex_count;
267 };
268
269 struct v3d_streamout_stateobj {
270 struct pipe_stream_output_target *targets[PIPE_MAX_SO_BUFFERS];
271 /* Number of vertices we've written into the buffer so far. */
272 uint32_t offsets[PIPE_MAX_SO_BUFFERS];
273 unsigned num_targets;
274 };
275
276 struct v3d_ssbo_stateobj {
277 struct pipe_shader_buffer sb[PIPE_MAX_SHADER_BUFFERS];
278 uint32_t enabled_mask;
279 };
280
281 /* Hash table key for v3d->jobs */
282 struct v3d_job_key {
283 struct pipe_surface *cbufs[4];
284 struct pipe_surface *zsbuf;
285 struct pipe_surface *bbuf;
286 };
287
288 enum v3d_ez_state {
289 V3D_EZ_UNDECIDED = 0,
290 V3D_EZ_GT_GE,
291 V3D_EZ_LT_LE,
292 V3D_EZ_DISABLED,
293 };
294
295 struct v3d_image_view {
296 struct pipe_image_view base;
297 /* V3D 4.x texture shader state struct */
298 struct pipe_resource *tex_state;
299 uint32_t tex_state_offset;
300 };
301
302 struct v3d_shaderimg_stateobj {
303 struct v3d_image_view si[PIPE_MAX_SHADER_IMAGES];
304 uint32_t enabled_mask;
305 };
306
307 struct v3d_perfmon_state {
308 /* The kernel perfmon id */
309 uint32_t kperfmon_id;
310 /* True if at least one job was submitted with this perfmon. */
311 bool job_submitted;
312 /* Fence to be signaled when the last job submitted with this perfmon
313 * is executed by the GPU.
314 */
315 struct v3d_fence *last_job_fence;
316 uint8_t counters[DRM_V3D_MAX_PERF_COUNTERS];
317 uint64_t values[DRM_V3D_MAX_PERF_COUNTERS];
318 };
319
320 /**
321 * A complete bin/render job.
322 *
323 * This is all of the state necessary to submit a bin/render to the kernel.
324 * We want to be able to have multiple in progress at a time, so that we don't
325 * need to flush an existing CL just to switch to rendering to a new render
326 * target (which would mean reading back from the old render target when
327 * starting to render to it again).
328 */
329 struct v3d_job {
330 struct v3d_context *v3d;
331 struct v3d_cl bcl;
332 struct v3d_cl rcl;
333 struct v3d_cl indirect;
334 struct v3d_bo *tile_alloc;
335 struct v3d_bo *tile_state;
336
337 struct drm_v3d_submit_cl submit;
338
339 /**
340 * Set of all BOs referenced by the job. This will be used for making
341 * the list of BOs that the kernel will need to have paged in to
342 * execute our job.
343 */
344 struct set *bos;
345
346 /** Sum of the sizes of the BOs referenced by the job. */
347 uint32_t referenced_size;
348
349 struct set *write_prscs;
350 struct set *tf_write_prscs;
351
352 /* Size of the submit.bo_handles array. */
353 uint32_t bo_handles_size;
354
355 /** @{
356 * Surfaces to submit rendering for.
357 * For blit operations, bbuf is the source surface, and cbufs[0] is
358 * the destination surface.
359 */
360 uint32_t nr_cbufs;
361 struct pipe_surface *cbufs[4];
362 struct pipe_surface *zsbuf;
363 struct pipe_surface *bbuf;
364 /** @} */
365 /** @{
366 * Bounding box of the scissor across all queued drawing.
367 *
368 * Note that the max values are exclusive.
369 */
370 uint32_t draw_min_x;
371 uint32_t draw_min_y;
372 uint32_t draw_max_x;
373 uint32_t draw_max_y;
374
375 /** @} */
376 /** @{
377 * List of scissor rects used for all queued drawing. All scissor
378 * rects will be contained in the draw_{min/max}_{x/y} bounding box.
379 *
380 * This is used as an optimization when all drawing is scissored to
381 * limit tile flushing only to tiles that intersect a scissor rect.
382 * If scissor is used together with non-scissored drawing, then
383 * the optimization is disabled.
384 */
385 struct {
386 bool disabled;
387 uint32_t count;
388 struct {
389 uint32_t min_x, min_y;
390 uint32_t max_x, max_y;
391 } rects[MAX_JOB_SCISSORS];
392 } scissor;
393
394 /** @} */
395 /** @{
396 * Width/height of the color framebuffer being rendered to,
397 * for V3D_TILE_RENDERING_MODE_CONFIG.
398 */
399 uint32_t draw_width;
400 uint32_t draw_height;
401 uint32_t num_layers;
402
403 /** @} */
404 /** @{ Tile information, depending on MSAA and float color buffer. */
405 uint32_t draw_tiles_x; /** @< Number of tiles wide for framebuffer. */
406 uint32_t draw_tiles_y; /** @< Number of tiles high for framebuffer. */
407
408 uint32_t tile_width; /** @< Width of a tile. */
409 uint32_t tile_height; /** @< Height of a tile. */
410 /** maximum internal_bpp of all color render targets. */
411 uint32_t internal_bpp;
412
413 /** Whether the current rendering is in a 4X MSAA tile buffer. */
414 bool msaa;
415 /** @} */
416
417 /* Bitmask of PIPE_CLEAR_* of buffers that were cleared before the
418 * first rendering.
419 */
420 uint32_t clear;
421 /* Bitmask of PIPE_CLEAR_* of buffers that have been read by a draw
422 * call without having been cleared first.
423 */
424 uint32_t load;
425 /* Bitmask of PIPE_CLEAR_* of buffers that have been rendered to
426 * (either clears or draws) and should be stored.
427 */
428 uint32_t store;
429 uint32_t clear_color[4][4];
430 float clear_z;
431 uint8_t clear_s;
432
433 /**
434 * Set if some drawing (triangles, blits, or just a glClear()) has
435 * been done to the FBO, meaning that we need to
436 * DRM_IOCTL_V3D_SUBMIT_CL.
437 */
438 bool needs_flush;
439
440 /* Set if any shader has dirtied cachelines in the TMU that need to be
441 * flushed before job end.
442 */
443 bool tmu_dirty_rcl;
444
445 /**
446 * Set if a packet enabling TF has been emitted in the job (V3D 4.x).
447 */
448 bool tf_enabled;
449
450 bool needs_primitives_generated;
451
452 /**
453 * Current EZ state for drawing. Updated at the start of draw after
454 * we've decided on the shader being rendered.
455 */
456 enum v3d_ez_state ez_state;
457 /**
458 * The first EZ state that was used for drawing with a decided EZ
459 * direction (so either UNDECIDED, GT, or LT).
460 */
461 enum v3d_ez_state first_ez_state;
462
463 /**
464 * Number of draw calls (not counting full buffer clears) queued in
465 * the current job.
466 */
467 uint32_t draw_calls_queued;
468
469 /**
470 * Number of draw calls (not counting full buffer clears) queued in
471 * the current job during active transform feedback.
472 */
473 uint32_t tf_draw_calls_queued;
474
475 struct v3d_job_key key;
476 };
477
478 struct v3d_context {
479 struct pipe_context base;
480
481 int fd;
482 struct v3d_screen *screen;
483
484 /** The 3D rendering job for the currently bound FBO. */
485 struct v3d_job *job;
486
487 /* Map from struct v3d_job_key to the job for that FBO.
488 */
489 struct hash_table *jobs;
490
491 /**
492 * Map from v3d_resource to a job writing to that resource.
493 *
494 * Primarily for flushing jobs rendering to textures that are now
495 * being read from.
496 */
497 struct hash_table *write_jobs;
498
499 struct slab_child_pool transfer_pool;
500 struct blitter_context *blitter;
501
502 /** bitfield of V3D_DIRTY_* */
503 uint64_t dirty;
504
505 uint32_t next_uncompiled_program_id;
506 uint64_t next_compiled_program_id;
507
508 struct v3d_compiler_state *compiler_state;
509
510 uint8_t prim_mode;
511
512 /** Maximum index buffer valid for the current shader_rec. */
513 uint32_t max_index;
514
515 /** Sync object that our RCL or TFU job will update as its out_sync. */
516 uint32_t out_sync;
517
518 /* Stream uploader used by gallium internals. This could also be used
519 * by driver internals, but we tend to use the v3d_cl.h interfaces
520 * instead.
521 */
522 struct u_upload_mgr *uploader;
523 /* State uploader used inside the driver. This is for packing bits of
524 * long-term state inside buffers, since the kernel interfaces
525 * allocate a page at a time.
526 */
527 struct u_upload_mgr *state_uploader;
528
529 struct pipe_shader_state *sand8_blit_vs;
530 struct pipe_shader_state *sand8_blit_fs_luma;
531 struct pipe_shader_state *sand8_blit_fs_chroma;
532
533 /** @{ Current pipeline state objects */
534 struct pipe_scissor_state scissor;
535 struct v3d_blend_state *blend;
536 struct v3d_rasterizer_state *rasterizer;
537 struct v3d_depth_stencil_alpha_state *zsa;
538
539 struct v3d_program_stateobj prog;
540 uint32_t compute_num_workgroups[3];
541 struct v3d_bo *compute_shared_memory;
542
543 struct v3d_vertex_stateobj *vtx;
544
545 struct {
546 struct pipe_blend_color f;
547 uint16_t hf[4];
548 } blend_color;
549 struct pipe_stencil_ref stencil_ref;
550 unsigned sample_mask;
551 struct pipe_framebuffer_state framebuffer;
552
553 /* Per render target, whether we should swap the R and B fields in the
554 * shader's color output and in blending. If render targets disagree
555 * on the R/B swap and use the constant color, then we would need to
556 * fall back to in-shader blending.
557 */
558 uint8_t swap_color_rb;
559
560 /* Per render target, whether we should treat the dst alpha values as
561 * one in blending.
562 *
563 * For RGBX formats, the tile buffer's alpha channel will be
564 * undefined.
565 */
566 uint8_t blend_dst_alpha_one;
567
568 bool active_queries;
569
570 /**
571 * If a compute job writes a resource read by a non-compute stage we
572 * should sync on the last compute job.
573 */
574 bool sync_on_last_compute_job;
575
576 uint32_t tf_prims_generated;
577 uint32_t prims_generated;
578
579 uint32_t n_primitives_generated_queries_in_flight;
580
581 struct pipe_poly_stipple stipple;
582 struct pipe_clip_state clip;
583 struct pipe_viewport_state viewport;
584 struct v3d_ssbo_stateobj ssbo[PIPE_SHADER_TYPES];
585 struct v3d_shaderimg_stateobj shaderimg[PIPE_SHADER_TYPES];
586 struct v3d_constbuf_stateobj constbuf[PIPE_SHADER_TYPES];
587 struct v3d_texture_stateobj tex[PIPE_SHADER_TYPES];
588 struct v3d_vertexbuf_stateobj vertexbuf;
589 struct v3d_streamout_stateobj streamout;
590 struct v3d_bo *current_oq;
591 struct pipe_resource *prim_counts;
592 uint32_t prim_counts_offset;
593 struct pipe_debug_callback debug;
594 struct v3d_perfmon_state *active_perfmon;
595 struct v3d_perfmon_state *last_perfmon;
596 /** @} */
597 };
598
599 struct v3d_rasterizer_state {
600 struct pipe_rasterizer_state base;
601
602 float point_size;
603
604 uint8_t depth_offset[9];
605 uint8_t depth_offset_z16[9];
606 };
607
608 struct v3d_depth_stencil_alpha_state {
609 struct pipe_depth_stencil_alpha_state base;
610
611 enum v3d_ez_state ez_state;
612
613 uint8_t stencil_front[6];
614 uint8_t stencil_back[6];
615 };
616
617 struct v3d_blend_state {
618 struct pipe_blend_state base;
619
620 /* Per-RT mask of whether blending is enabled. */
621 uint8_t blend_enables;
622 };
623
624 #define perf_debug(...) do { \
625 if (unlikely(V3D_DEBUG & V3D_DEBUG_PERF)) \
626 fprintf(stderr, __VA_ARGS__); \
627 if (unlikely(v3d->debug.debug_message)) \
628 pipe_debug_message(&v3d->debug, PERF_INFO, __VA_ARGS__); \
629 } while (0)
630
631 static inline struct v3d_context *
v3d_context(struct pipe_context * pcontext)632 v3d_context(struct pipe_context *pcontext)
633 {
634 return (struct v3d_context *)pcontext;
635 }
636
637 static inline struct v3d_sampler_view *
v3d_sampler_view(struct pipe_sampler_view * psview)638 v3d_sampler_view(struct pipe_sampler_view *psview)
639 {
640 return (struct v3d_sampler_view *)psview;
641 }
642
643 static inline struct v3d_sampler_state *
v3d_sampler_state(struct pipe_sampler_state * psampler)644 v3d_sampler_state(struct pipe_sampler_state *psampler)
645 {
646 return (struct v3d_sampler_state *)psampler;
647 }
648
649 static inline struct v3d_stream_output_target *
v3d_stream_output_target(struct pipe_stream_output_target * ptarget)650 v3d_stream_output_target(struct pipe_stream_output_target *ptarget)
651 {
652 return (struct v3d_stream_output_target *)ptarget;
653 }
654
655 static inline uint32_t
v3d_stream_output_target_get_vertex_count(struct pipe_stream_output_target * ptarget)656 v3d_stream_output_target_get_vertex_count(struct pipe_stream_output_target *ptarget)
657 {
658 return v3d_stream_output_target(ptarget)->recorded_vertex_count;
659 }
660
661 int v3d_get_driver_query_group_info(struct pipe_screen *pscreen,
662 unsigned index,
663 struct pipe_driver_query_group_info *info);
664 int v3d_get_driver_query_info(struct pipe_screen *pscreen, unsigned index,
665 struct pipe_driver_query_info *info);
666
667 struct pipe_context *v3d_context_create(struct pipe_screen *pscreen,
668 void *priv, unsigned flags);
669 void v3d_program_init(struct pipe_context *pctx);
670 void v3d_program_fini(struct pipe_context *pctx);
671 void v3d_query_init(struct pipe_context *pctx);
672
673 static inline int
v3d_ioctl(int fd,unsigned long request,void * arg)674 v3d_ioctl(int fd, unsigned long request, void *arg)
675 {
676 if (using_v3d_simulator)
677 return v3d_simulator_ioctl(fd, request, arg);
678 else
679 return drmIoctl(fd, request, arg);
680 }
681
682 static inline bool
v3d_transform_feedback_enabled(struct v3d_context * v3d)683 v3d_transform_feedback_enabled(struct v3d_context *v3d)
684 {
685 return (v3d->prog.bind_vs->num_tf_specs != 0 ||
686 (v3d->prog.bind_gs && v3d->prog.bind_gs->num_tf_specs != 0)) &&
687 v3d->active_queries;
688 }
689
690 void v3d_set_shader_uniform_dirty_flags(struct v3d_compiled_shader *shader);
691 struct v3d_cl_reloc v3d_write_uniforms(struct v3d_context *v3d,
692 struct v3d_job *job,
693 struct v3d_compiled_shader *shader,
694 enum pipe_shader_type stage);
695
696 void v3d_flush(struct pipe_context *pctx);
697 void v3d_job_init(struct v3d_context *v3d);
698 struct v3d_job *v3d_job_create(struct v3d_context *v3d);
699 void v3d_job_free(struct v3d_context *v3d, struct v3d_job *job);
700 struct v3d_job *v3d_get_job(struct v3d_context *v3d,
701 uint32_t nr_cbufs,
702 struct pipe_surface **cbufs,
703 struct pipe_surface *zsbuf,
704 struct pipe_surface *bbuf);
705 struct v3d_job *v3d_get_job_for_fbo(struct v3d_context *v3d);
706 void v3d_job_add_bo(struct v3d_job *job, struct v3d_bo *bo);
707 void v3d_job_add_write_resource(struct v3d_job *job, struct pipe_resource *prsc);
708 void v3d_job_add_tf_write_resource(struct v3d_job *job, struct pipe_resource *prsc);
709 void v3d_job_submit(struct v3d_context *v3d, struct v3d_job *job);
710 void v3d_flush_jobs_using_bo(struct v3d_context *v3d, struct v3d_bo *bo);
711 void v3d_flush_jobs_writing_resource(struct v3d_context *v3d,
712 struct pipe_resource *prsc,
713 enum v3d_flush_cond flush_cond,
714 bool is_compute_pipeline);
715 void v3d_flush_jobs_reading_resource(struct v3d_context *v3d,
716 struct pipe_resource *prsc,
717 enum v3d_flush_cond flush_cond,
718 bool is_compute_pipeline);
719 void v3d_update_compiled_shaders(struct v3d_context *v3d, uint8_t prim_mode);
720 void v3d_update_compiled_cs(struct v3d_context *v3d);
721
722 bool v3d_rt_format_supported(const struct v3d_device_info *devinfo,
723 enum pipe_format f);
724 bool v3d_tex_format_supported(const struct v3d_device_info *devinfo,
725 enum pipe_format f);
726 uint8_t v3d_get_rt_format(const struct v3d_device_info *devinfo, enum pipe_format f);
727 uint8_t v3d_get_tex_format(const struct v3d_device_info *devinfo, enum pipe_format f);
728 uint8_t v3d_get_tex_return_size(const struct v3d_device_info *devinfo,
729 enum pipe_format f,
730 enum pipe_tex_compare compare);
731 uint8_t v3d_get_tex_return_channels(const struct v3d_device_info *devinfo,
732 enum pipe_format f);
733 const uint8_t *v3d_get_format_swizzle(const struct v3d_device_info *devinfo,
734 enum pipe_format f);
735 void v3d_get_internal_type_bpp_for_output_format(const struct v3d_device_info *devinfo,
736 uint32_t format,
737 uint32_t *type,
738 uint32_t *bpp);
739 bool v3d_tfu_supports_tex_format(const struct v3d_device_info *devinfo,
740 uint32_t tex_format,
741 bool for_mipmap);
742 bool v3d_format_supports_tlb_msaa_resolve(const struct v3d_device_info *devinfo,
743 enum pipe_format f);
744
745 void v3d_init_query_functions(struct v3d_context *v3d);
746 void v3d_blit(struct pipe_context *pctx, const struct pipe_blit_info *blit_info);
747 void v3d_blitter_save(struct v3d_context *v3d);
748 bool v3d_generate_mipmap(struct pipe_context *pctx,
749 struct pipe_resource *prsc,
750 enum pipe_format format,
751 unsigned int base_level,
752 unsigned int last_level,
753 unsigned int first_layer,
754 unsigned int last_layer);
755
756 void
757 v3d_fence_unreference(struct v3d_fence **fence);
758
759 struct v3d_fence *v3d_fence_create(struct v3d_context *v3d);
760
761 bool v3d_fence_wait(struct v3d_screen *screen,
762 struct v3d_fence *fence,
763 uint64_t timeout_ns);
764
765 void v3d_update_primitive_counters(struct v3d_context *v3d);
766
767 bool v3d_line_smoothing_enabled(struct v3d_context *v3d);
768
769 float v3d_get_real_line_width(struct v3d_context *v3d);
770
771 void v3d_ensure_prim_counts_allocated(struct v3d_context *ctx);
772
773 void v3d_flag_dirty_sampler_state(struct v3d_context *v3d,
774 enum pipe_shader_type shader);
775
776 void v3d_create_texture_shader_state_bo(struct v3d_context *v3d,
777 struct v3d_sampler_view *so);
778
779 void v3d_get_tile_buffer_size(bool is_msaa,
780 uint32_t nr_cbufs,
781 struct pipe_surface **cbufs,
782 struct pipe_surface *bbuf,
783 uint32_t *tile_width,
784 uint32_t *tile_height,
785 uint32_t *max_bpp);
786
787 #ifdef v3dX
788 # include "v3dx_context.h"
789 #else
790 # define v3dX(x) v3d33_##x
791 # include "v3dx_context.h"
792 # undef v3dX
793
794 # define v3dX(x) v3d41_##x
795 # include "v3dx_context.h"
796 # undef v3dX
797 #endif
798
799 #endif /* V3D_CONTEXT_H */
800