2 * Copyright © 2014 Broadcom
3 * Copyright (C) 2012 Rob Clark <robclark@freedesktop.org>
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:
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
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
30 #include "pipe/p_context.h"
31 #include "pipe/p_state.h"
32 #include "util/slab.h"
37 #include "vc4_bufmgr.h"
38 #include "vc4_resource.h"
42 #ifndef DRM_VC4_PARAM_SUPPORTS_ETC1
43 #define DRM_VC4_PARAM_SUPPORTS_ETC1 4
45 #ifndef DRM_VC4_PARAM_SUPPORTS_THREADED_FS
46 #define DRM_VC4_PARAM_SUPPORTS_THREADED_FS 5
49 #ifdef USE_VC4_SIMULATOR
50 #define using_vc4_simulator true
52 #define using_vc4_simulator false
55 #define VC4_DIRTY_BLEND (1 << 0)
56 #define VC4_DIRTY_RASTERIZER (1 << 1)
57 #define VC4_DIRTY_ZSA (1 << 2)
58 #define VC4_DIRTY_FRAGTEX (1 << 3)
59 #define VC4_DIRTY_VERTTEX (1 << 4)
61 #define VC4_DIRTY_BLEND_COLOR (1 << 7)
62 #define VC4_DIRTY_STENCIL_REF (1 << 8)
63 #define VC4_DIRTY_SAMPLE_MASK (1 << 9)
64 #define VC4_DIRTY_FRAMEBUFFER (1 << 10)
65 #define VC4_DIRTY_STIPPLE (1 << 11)
66 #define VC4_DIRTY_VIEWPORT (1 << 12)
67 #define VC4_DIRTY_CONSTBUF (1 << 13)
68 #define VC4_DIRTY_VTXSTATE (1 << 14)
69 #define VC4_DIRTY_VTXBUF (1 << 15)
71 #define VC4_DIRTY_SCISSOR (1 << 17)
72 #define VC4_DIRTY_FLAT_SHADE_FLAGS (1 << 18)
73 #define VC4_DIRTY_PRIM_MODE (1 << 19)
74 #define VC4_DIRTY_CLIP (1 << 20)
75 #define VC4_DIRTY_UNCOMPILED_VS (1 << 21)
76 #define VC4_DIRTY_UNCOMPILED_FS (1 << 22)
77 #define VC4_DIRTY_COMPILED_CS (1 << 23)
78 #define VC4_DIRTY_COMPILED_VS (1 << 24)
79 #define VC4_DIRTY_COMPILED_FS (1 << 25)
80 #define VC4_DIRTY_FS_INPUTS (1 << 26)
82 struct vc4_sampler_view
{
83 struct pipe_sampler_view base
;
86 bool force_first_level
;
88 * Resource containing the actual texture that will be sampled.
90 * We may need to rebase the .base.texture resource to work around the
91 * lack of GL_TEXTURE_BASE_LEVEL, or to upload the texture as tiled.
93 struct pipe_resource
*texture
;
96 struct vc4_sampler_state
{
97 struct pipe_sampler_state base
;
101 struct vc4_texture_stateobj
{
102 struct pipe_sampler_view
*textures
[PIPE_MAX_SAMPLERS
];
103 unsigned num_textures
;
104 struct pipe_sampler_state
*samplers
[PIPE_MAX_SAMPLERS
];
105 unsigned num_samplers
;
108 struct vc4_shader_uniform_info
{
109 enum quniform_contents
*contents
;
112 uint32_t num_texture_samples
;
115 struct vc4_uncompiled_shader
{
116 /** A name for this program, so you can track it in shader-db output. */
118 /** How many variants of this program were compiled, for shader-db. */
119 uint32_t compiled_variant_count
;
120 struct pipe_shader_state base
;
123 struct vc4_ubo_range
{
125 * offset in bytes from the start of the ubo where this range is
128 * Only set once used is set.
133 * offset in bytes from the start of the gallium uniforms where the
138 /** size in bytes of this ubo range */
142 struct vc4_fs_inputs
{
144 * Array of the meanings of the VPM inputs this shader needs.
146 * It doesn't include those that aren't part of the VPM, like
147 * point/line coordinates.
149 struct vc4_varying_slot
*input_slots
;
153 struct vc4_compiled_shader
{
157 struct vc4_shader_uniform_info uniforms
;
159 struct vc4_ubo_range
*ubo_ranges
;
160 uint32_t num_ubo_ranges
;
163 * VC4_DIRTY_* flags that, when set in vc4->dirty, mean that the
164 * uniforms have to be rewritten (and therefore the shader state
167 uint32_t uniform_dirty_bits
;
169 /** bitmask of which inputs are color inputs, for flat shade handling. */
170 uint32_t color_inputs
;
172 bool disable_early_z
;
174 /* Set if the compile failed, likely due to register allocation
175 * failure. In this case, we have no shader to run and should not try
184 /* Byte offsets for the start of the vertex attributes 0-7, and the
185 * total size as "attribute" 8.
187 uint8_t vattr_offsets
[9];
190 const struct vc4_fs_inputs
*fs_inputs
;
193 struct vc4_program_stateobj
{
194 struct vc4_uncompiled_shader
*bind_vs
, *bind_fs
;
195 struct vc4_compiled_shader
*cs
, *vs
, *fs
;
198 struct vc4_constbuf_stateobj
{
199 struct pipe_constant_buffer cb
[PIPE_MAX_CONSTANT_BUFFERS
];
200 uint32_t enabled_mask
;
204 struct vc4_vertexbuf_stateobj
{
205 struct pipe_vertex_buffer vb
[PIPE_MAX_ATTRIBS
];
207 uint32_t enabled_mask
;
211 struct vc4_vertex_stateobj
{
212 struct pipe_vertex_element pipe
[PIPE_MAX_ATTRIBS
];
213 unsigned num_elements
;
216 /* Hash table key for vc4->jobs */
218 struct pipe_surface
*cbuf
;
219 struct pipe_surface
*zsbuf
;
223 * A complete bin/render job.
225 * This is all of the state necessary to submit a bin/render to the kernel.
226 * We want to be able to have multiple in progress at a time, so that we don't
227 * need to flush an existing CL just to switch to rendering to a new render
228 * target (which would mean reading back from the old render target when
229 * starting to render to it again).
233 struct vc4_cl shader_rec
;
234 struct vc4_cl uniforms
;
235 struct vc4_cl bo_handles
;
236 struct vc4_cl bo_pointers
;
237 uint32_t shader_rec_count
;
239 * Amount of memory used by the BOs in bo_pointers.
241 * Used for checking when we should flush the job early so we don't
246 /** @{ Surfaces to submit rendering for. */
247 struct pipe_surface
*color_read
;
248 struct pipe_surface
*color_write
;
249 struct pipe_surface
*zs_read
;
250 struct pipe_surface
*zs_write
;
251 struct pipe_surface
*msaa_color_write
;
252 struct pipe_surface
*msaa_zs_write
;
255 * Bounding box of the scissor across all queued drawing.
257 * Note that the max values are exclusive.
265 * Width/height of the color framebuffer being rendered to,
266 * for VC4_TILE_RENDERING_MODE_CONFIG.
269 uint32_t draw_height
;
271 /** @{ Tile information, depending on MSAA and float color buffer. */
272 uint32_t draw_tiles_x
; /** @< Number of tiles wide for framebuffer. */
273 uint32_t draw_tiles_y
; /** @< Number of tiles high for framebuffer. */
275 uint32_t tile_width
; /** @< Width of a tile. */
276 uint32_t tile_height
; /** @< Height of a tile. */
277 /** Whether the current rendering is in a 4X MSAA tile buffer. */
281 /* Bitmask of PIPE_CLEAR_* of buffers that were cleared before the
285 /* Bitmask of PIPE_CLEAR_* of buffers that have been rendered to
286 * (either clears or draws).
289 uint32_t clear_color
[2];
290 uint32_t clear_depth
; /**< 24-bit unorm depth */
291 uint8_t clear_stencil
;
294 * Set if some drawing (triangles, blits, or just a glClear()) has
295 * been done to the FBO, meaning that we need to
296 * DRM_IOCTL_VC4_SUBMIT_CL.
301 * Number of draw calls (not counting full buffer clears) queued in
304 uint32_t draw_calls_queued
;
306 /** Any flags to be passed in drm_vc4_submit_cl.flags. */
309 struct vc4_job_key key
;
313 struct pipe_context base
;
316 struct vc4_screen
*screen
;
318 /** The 3D rendering job for the currently bound FBO. */
321 /* Map from struct vc4_job_key to the job for that FBO.
323 struct hash_table
*jobs
;
326 * Map from vc4_resource to a job writing to that resource.
328 * Primarily for flushing jobs rendering to textures that are now
331 struct hash_table
*write_jobs
;
333 struct slab_child_pool transfer_pool
;
334 struct blitter_context
*blitter
;
336 /** bitfield of VC4_DIRTY_* */
339 struct primconvert_context
*primconvert
;
341 struct hash_table
*fs_cache
, *vs_cache
;
342 struct set
*fs_inputs_set
;
343 uint32_t next_uncompiled_program_id
;
344 uint64_t next_compiled_program_id
;
346 struct ra_regs
*regs
;
347 unsigned int reg_class_any
[2];
348 unsigned int reg_class_a_or_b
[2];
349 unsigned int reg_class_a_or_b_or_acc
[2];
350 unsigned int reg_class_r0_r3
;
351 unsigned int reg_class_r4_or_a
[2];
352 unsigned int reg_class_a
[2];
356 /** Maximum index buffer valid for the current shader_rec. */
358 /** Last index bias baked into the current shader_rec. */
359 uint32_t last_index_bias
;
361 /** Seqno of the last CL flush's job. */
362 uint64_t last_emit_seqno
;
364 struct u_upload_mgr
*uploader
;
366 /** @{ Current pipeline state objects */
367 struct pipe_scissor_state scissor
;
368 struct pipe_blend_state
*blend
;
369 struct vc4_rasterizer_state
*rasterizer
;
370 struct vc4_depth_stencil_alpha_state
*zsa
;
372 struct vc4_texture_stateobj verttex
, fragtex
;
374 struct vc4_program_stateobj prog
;
376 struct vc4_vertex_stateobj
*vtx
;
379 struct pipe_blend_color f
;
382 struct pipe_stencil_ref stencil_ref
;
383 unsigned sample_mask
;
384 struct pipe_framebuffer_state framebuffer
;
385 struct pipe_poly_stipple stipple
;
386 struct pipe_clip_state clip
;
387 struct pipe_viewport_state viewport
;
388 struct vc4_constbuf_stateobj constbuf
[PIPE_SHADER_TYPES
];
389 struct vc4_vertexbuf_stateobj vertexbuf
;
393 struct vc4_rasterizer_state
{
394 struct pipe_rasterizer_state base
;
396 /* VC4_CONFIGURATION_BITS */
397 uint8_t config_bits
[V3D21_CONFIGURATION_BITS_length
];
400 uint8_t depth_offset
[V3D21_DEPTH_OFFSET_length
];
401 uint8_t point_size
[V3D21_POINT_SIZE_length
];
402 uint8_t line_width
[V3D21_LINE_WIDTH_length
];
405 /** Raster order flags to be passed in struct drm_vc4_submit_cl.flags. */
406 uint32_t tile_raster_order_flags
;
409 struct vc4_depth_stencil_alpha_state
{
410 struct pipe_depth_stencil_alpha_state base
;
412 /* VC4_CONFIGURATION_BITS */
413 uint8_t config_bits
[V3D21_CONFIGURATION_BITS_length
];
415 /** Uniforms for stencil state.
417 * Index 0 is either the front config, or the front-and-back config.
418 * Index 1 is the back config if doing separate back stencil.
419 * Index 2 is the writemask config if it's not a common mask value.
421 uint32_t stencil_uniforms
[3];
424 #define perf_debug(...) do { \
425 if (unlikely(vc4_debug & VC4_DEBUG_PERF)) \
426 fprintf(stderr, __VA_ARGS__); \
429 static inline struct vc4_context
*
430 vc4_context(struct pipe_context
*pcontext
)
432 return (struct vc4_context
*)pcontext
;
435 static inline struct vc4_sampler_view
*
436 vc4_sampler_view(struct pipe_sampler_view
*psview
)
438 return (struct vc4_sampler_view
*)psview
;
441 static inline struct vc4_sampler_state
*
442 vc4_sampler_state(struct pipe_sampler_state
*psampler
)
444 return (struct vc4_sampler_state
*)psampler
;
447 struct pipe_context
*vc4_context_create(struct pipe_screen
*pscreen
,
448 void *priv
, unsigned flags
);
449 void vc4_draw_init(struct pipe_context
*pctx
);
450 void vc4_state_init(struct pipe_context
*pctx
);
451 void vc4_program_init(struct pipe_context
*pctx
);
452 void vc4_program_fini(struct pipe_context
*pctx
);
453 void vc4_query_init(struct pipe_context
*pctx
);
454 void vc4_simulator_init(struct vc4_screen
*screen
);
455 void vc4_simulator_destroy(struct vc4_screen
*screen
);
456 int vc4_simulator_flush(struct vc4_context
*vc4
,
457 struct drm_vc4_submit_cl
*args
,
458 struct vc4_job
*job
);
459 int vc4_simulator_ioctl(int fd
, unsigned long request
, void *arg
);
460 void vc4_simulator_open_from_handle(int fd
, uint32_t winsys_stride
,
461 int handle
, uint32_t size
);
464 vc4_ioctl(int fd
, unsigned long request
, void *arg
)
466 if (using_vc4_simulator
)
467 return vc4_simulator_ioctl(fd
, request
, arg
);
469 return drmIoctl(fd
, request
, arg
);
472 void vc4_set_shader_uniform_dirty_flags(struct vc4_compiled_shader
*shader
);
473 void vc4_write_uniforms(struct vc4_context
*vc4
,
474 struct vc4_compiled_shader
*shader
,
475 struct vc4_constbuf_stateobj
*cb
,
476 struct vc4_texture_stateobj
*texstate
);
478 void vc4_flush(struct pipe_context
*pctx
);
479 void vc4_job_init(struct vc4_context
*vc4
);
480 struct vc4_job
*vc4_get_job(struct vc4_context
*vc4
,
481 struct pipe_surface
*cbuf
,
482 struct pipe_surface
*zsbuf
);
483 struct vc4_job
*vc4_get_job_for_fbo(struct vc4_context
*vc4
);
485 void vc4_job_submit(struct vc4_context
*vc4
, struct vc4_job
*job
);
486 void vc4_flush_jobs_writing_resource(struct vc4_context
*vc4
,
487 struct pipe_resource
*prsc
);
488 void vc4_flush_jobs_reading_resource(struct vc4_context
*vc4
,
489 struct pipe_resource
*prsc
);
490 void vc4_emit_state(struct pipe_context
*pctx
);
491 void vc4_generate_code(struct vc4_context
*vc4
, struct vc4_compile
*c
);
492 struct qpu_reg
*vc4_register_allocate(struct vc4_context
*vc4
, struct vc4_compile
*c
);
493 bool vc4_update_compiled_shaders(struct vc4_context
*vc4
, uint8_t prim_mode
);
495 bool vc4_rt_format_supported(enum pipe_format f
);
496 bool vc4_rt_format_is_565(enum pipe_format f
);
497 bool vc4_tex_format_supported(enum pipe_format f
);
498 uint8_t vc4_get_tex_format(enum pipe_format f
);
499 const uint8_t *vc4_get_format_swizzle(enum pipe_format f
);
500 void vc4_init_query_functions(struct vc4_context
*vc4
);
501 void vc4_blit(struct pipe_context
*pctx
, const struct pipe_blit_info
*blit_info
);
502 void vc4_blitter_save(struct vc4_context
*vc4
);
503 #endif /* VC4_CONTEXT_H */