summaryrefslogtreecommitdiff
path: root/src/gallium/drivers/freedreno/freedreno_context.h
blob: 85f17c81a9d549a533ffa471f5e7beb593717eb8 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
/*
 * Copyright (C) 2012 Rob Clark <robclark@freedesktop.org>
 *
 * Permission is hereby granted, free of charge, to any person obtaining a
 * copy of this software and associated documentation files (the "Software"),
 * to deal in the Software without restriction, including without limitation
 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
 * and/or sell copies of the Software, and to permit persons to whom the
 * Software is furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice (including the next
 * paragraph) shall be included in all copies or substantial portions of the
 * Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
 * SOFTWARE.
 *
 * Authors:
 *    Rob Clark <robclark@freedesktop.org>
 */

#ifndef FREEDRENO_CONTEXT_H_
#define FREEDRENO_CONTEXT_H_

#include "pipe/p_context.h"
#include "indices/u_primconvert.h"
#include "util/u_blitter.h"
#include "util/list.h"
#include "util/slab.h"
#include "util/u_string.h"

#include "freedreno_batch.h"
#include "freedreno_screen.h"
#include "freedreno_gmem.h"
#include "freedreno_util.h"

#define BORDER_COLOR_UPLOAD_SIZE (2 * PIPE_MAX_SAMPLERS * BORDERCOLOR_SIZE)

struct fd_vertex_stateobj;

struct fd_texture_stateobj {
	struct pipe_sampler_view *textures[PIPE_MAX_SAMPLERS];
	unsigned num_textures;
	unsigned valid_textures;
	struct pipe_sampler_state *samplers[PIPE_MAX_SAMPLERS];
	unsigned num_samplers;
	unsigned valid_samplers;
	/* number of samples per sampler, 2 bits per sampler: */
	uint32_t samples;
};

struct fd_program_stateobj {
	void *vp, *fp;
};

struct fd_constbuf_stateobj {
	struct pipe_constant_buffer cb[PIPE_MAX_CONSTANT_BUFFERS];
	uint32_t enabled_mask;
};

struct fd_shaderbuf_stateobj {
	struct pipe_shader_buffer sb[PIPE_MAX_SHADER_BUFFERS];
	uint32_t enabled_mask;
};

struct fd_shaderimg_stateobj {
	struct pipe_image_view si[PIPE_MAX_SHADER_IMAGES];
	uint32_t enabled_mask;
};

struct fd_vertexbuf_stateobj {
	struct pipe_vertex_buffer vb[PIPE_MAX_ATTRIBS];
	unsigned count;
	uint32_t enabled_mask;
};

struct fd_vertex_stateobj {
	struct pipe_vertex_element pipe[PIPE_MAX_ATTRIBS];
	unsigned num_elements;
};

struct fd_streamout_stateobj {
	struct pipe_stream_output_target *targets[PIPE_MAX_SO_BUFFERS];
	unsigned num_targets;
	/* Track offset from vtxcnt for streamout data.  This counter
	 * is just incremented by # of vertices on each draw until
	 * reset or new streamout buffer bound.
	 *
	 * When we eventually have GS, the CPU won't actually know the
	 * number of vertices per draw, so I think we'll have to do
	 * something more clever.
	 */
	unsigned offsets[PIPE_MAX_SO_BUFFERS];
};

#define MAX_GLOBAL_BUFFERS 16
struct fd_global_bindings_stateobj {
	struct pipe_resource *buf[MAX_GLOBAL_BUFFERS];
	uint32_t enabled_mask;
};

/* group together the vertex and vertexbuf state.. for ease of passing
 * around, and because various internal operations (gmem<->mem, etc)
 * need their own vertex state:
 */
struct fd_vertex_state {
	struct fd_vertex_stateobj *vtx;
	struct fd_vertexbuf_stateobj vertexbuf;
};

/* global 3d pipeline dirty state: */
enum fd_dirty_3d_state {
	FD_DIRTY_BLEND       = BIT(0),
	FD_DIRTY_RASTERIZER  = BIT(1),
	FD_DIRTY_ZSA         = BIT(2),
	FD_DIRTY_BLEND_COLOR = BIT(3),
	FD_DIRTY_STENCIL_REF = BIT(4),
	FD_DIRTY_SAMPLE_MASK = BIT(5),
	FD_DIRTY_FRAMEBUFFER = BIT(6),
	FD_DIRTY_STIPPLE     = BIT(7),
	FD_DIRTY_VIEWPORT    = BIT(8),
	FD_DIRTY_VTXSTATE    = BIT(9),
	FD_DIRTY_VTXBUF      = BIT(10),

	FD_DIRTY_SCISSOR     = BIT(12),
	FD_DIRTY_STREAMOUT   = BIT(13),
	FD_DIRTY_UCP         = BIT(14),
	FD_DIRTY_BLEND_DUAL  = BIT(15),

	/* These are a bit redundent with fd_dirty_shader_state, and possibly
	 * should be removed.  (But OTOH kinda convenient in some places)
	 */
	FD_DIRTY_PROG        = BIT(16),
	FD_DIRTY_CONST       = BIT(17),
	FD_DIRTY_TEX         = BIT(18),

	/* only used by a2xx.. possibly can be removed.. */
	FD_DIRTY_TEXSTATE    = BIT(19),
};

/* per shader-stage dirty state: */
enum fd_dirty_shader_state {
	FD_DIRTY_SHADER_PROG  = BIT(0),
	FD_DIRTY_SHADER_CONST = BIT(1),
	FD_DIRTY_SHADER_TEX   = BIT(2),
	FD_DIRTY_SHADER_SSBO  = BIT(3),
	FD_DIRTY_SHADER_IMAGE = BIT(4),
};

struct fd_context {
	struct pipe_context base;

	struct fd_device *dev;
	struct fd_screen *screen;
	struct fd_pipe *pipe;

	struct util_queue flush_queue;

	struct blitter_context *blitter;
	void *clear_rs_state;
	struct primconvert_context *primconvert;

	/* slab for pipe_transfer allocations: */
	struct slab_child_pool transfer_pool;

	/**
	 * query related state:
	 */
	/*@{*/
	/* slabs for fd_hw_sample and fd_hw_sample_period allocations: */
	struct slab_mempool sample_pool;
	struct slab_mempool sample_period_pool;

	/* sample-providers for hw queries: */
	const struct fd_hw_sample_provider *hw_sample_providers[MAX_HW_SAMPLE_PROVIDERS];

	/* list of active queries: */
	struct list_head hw_active_queries;

	/* sample-providers for accumulating hw queries: */
	const struct fd_acc_sample_provider *acc_sample_providers[MAX_HW_SAMPLE_PROVIDERS];

	/* list of active accumulating queries: */
	struct list_head acc_active_queries;
	/*@}*/

	/* table with PIPE_PRIM_MAX entries mapping PIPE_PRIM_x to
	 * DI_PT_x value to use for draw initiator.  There are some
	 * slight differences between generation:
	 */
	const uint8_t *primtypes;
	uint32_t primtype_mask;

	/* shaders used by clear, and gmem->mem blits: */
	struct fd_program_stateobj solid_prog; // TODO move to screen?

	/* shaders used by mem->gmem blits: */
	struct fd_program_stateobj blit_prog[MAX_RENDER_TARGETS]; // TODO move to screen?
	struct fd_program_stateobj blit_z, blit_zs;

	/* Stats/counters:
	 */
	struct {
		uint64_t prims_emitted;
		uint64_t prims_generated;
		uint64_t draw_calls;
		uint64_t batch_total, batch_sysmem, batch_gmem, batch_nondraw, batch_restore;
		uint64_t staging_uploads, shadow_uploads;
		uint64_t vs_regs, fs_regs;
	} stats;

	/* Current batch.. the rule here is that you can deref ctx->batch
	 * in codepaths from pipe_context entrypoints.  But not in code-
	 * paths from fd_batch_flush() (basically, the stuff that gets
	 * called from GMEM code), since in those code-paths the batch
	 * you care about is not necessarily the same as ctx->batch.
	 */
	struct fd_batch *batch;

	/* NULL if there has been rendering since last flush.  Otherwise
	 * keeps a reference to the last fence so we can re-use it rather
	 * than having to flush no-op batch.
	 */
	struct pipe_fence_handle *last_fence;

	/* Are we in process of shadowing a resource? Used to detect recursion
	 * in transfer_map, and skip unneeded synchronization.
	 */
	bool in_shadow : 1;

	/* Ie. in blit situation where we no longer care about previous framebuffer
	 * contents.  Main point is to eliminate blits from fd_try_shadow_resource().
	 * For example, in case of texture upload + gen-mipmaps.
	 */
	bool in_blit : 1;

	struct pipe_scissor_state scissor;

	/* we don't have a disable/enable bit for scissor, so instead we keep
	 * a disabled-scissor state which matches the entire bound framebuffer
	 * and use that when scissor is not enabled.
	 */
	struct pipe_scissor_state disabled_scissor;

	/* Current gmem/tiling configuration.. gets updated on render_tiles()
	 * if out of date with current maximal-scissor/cpp:
	 *
	 * (NOTE: this is kind of related to the batch, but moving it there
	 * means we'd always have to recalc tiles ever batch)
	 */
	struct fd_gmem_stateobj gmem;
	struct fd_vsc_pipe      vsc_pipe[32];
	struct fd_tile          tile[512];

	/* which state objects need to be re-emit'd: */
	enum fd_dirty_3d_state dirty;

	/* per shader-stage dirty status: */
	enum fd_dirty_shader_state dirty_shader[PIPE_SHADER_TYPES];

	void *compute;
	struct pipe_blend_state *blend;
	struct pipe_rasterizer_state *rasterizer;
	struct pipe_depth_stencil_alpha_state *zsa;

	struct fd_texture_stateobj tex[PIPE_SHADER_TYPES];

	struct fd_program_stateobj prog;

	struct fd_vertex_state vtx;

	struct pipe_blend_color blend_color;
	struct pipe_stencil_ref stencil_ref;
	unsigned sample_mask;
	/* local context fb state, for when ctx->batch is null: */
	struct pipe_framebuffer_state framebuffer;
	struct pipe_poly_stipple stipple;
	struct pipe_viewport_state viewport;
	struct pipe_scissor_state viewport_scissor;
	struct fd_constbuf_stateobj constbuf[PIPE_SHADER_TYPES];
	struct fd_shaderbuf_stateobj shaderbuf[PIPE_SHADER_TYPES];
	struct fd_shaderimg_stateobj shaderimg[PIPE_SHADER_TYPES];
	struct fd_streamout_stateobj streamout;
	struct fd_global_bindings_stateobj global_bindings;
	struct pipe_clip_state ucp;

	struct pipe_query *cond_query;
	bool cond_cond; /* inverted rendering condition */
	uint cond_mode;

	struct pipe_debug_callback debug;

	/* GMEM/tile handling fxns: */
	void (*emit_tile_init)(struct fd_batch *batch);
	void (*emit_tile_prep)(struct fd_batch *batch, struct fd_tile *tile);
	void (*emit_tile_mem2gmem)(struct fd_batch *batch, struct fd_tile *tile);
	void (*emit_tile_renderprep)(struct fd_batch *batch, struct fd_tile *tile);
	void (*emit_tile_gmem2mem)(struct fd_batch *batch, struct fd_tile *tile);
	void (*emit_tile_fini)(struct fd_batch *batch);   /* optional */

	/* optional, for GMEM bypass: */
	void (*emit_sysmem_prep)(struct fd_batch *batch);
	void (*emit_sysmem_fini)(struct fd_batch *batch);

	/* draw: */
	bool (*draw_vbo)(struct fd_context *ctx, const struct pipe_draw_info *info,
			unsigned index_offset);
	bool (*clear)(struct fd_context *ctx, unsigned buffers,
			const union pipe_color_union *color, double depth, unsigned stencil);

	/* compute: */
	void (*launch_grid)(struct fd_context *ctx, const struct pipe_grid_info *info);

	/* constant emit:  (note currently not used/needed for a2xx) */
	void (*emit_const)(struct fd_ringbuffer *ring, gl_shader_stage type,
			uint32_t regid, uint32_t offset, uint32_t sizedwords,
			const uint32_t *dwords, struct pipe_resource *prsc);
	/* emit bo addresses as constant: */
	void (*emit_const_bo)(struct fd_ringbuffer *ring, gl_shader_stage type, boolean write,
			uint32_t regid, uint32_t num, struct pipe_resource **prscs, uint32_t *offsets);

	/* indirect-branch emit: */
	void (*emit_ib)(struct fd_ringbuffer *ring, struct fd_ringbuffer *target);

	/* query: */
	struct fd_query * (*create_query)(struct fd_context *ctx, unsigned query_type);
	void (*query_prepare)(struct fd_batch *batch, uint32_t num_tiles);
	void (*query_prepare_tile)(struct fd_batch *batch, uint32_t n,
			struct fd_ringbuffer *ring);
	void (*query_set_stage)(struct fd_batch *batch, enum fd_render_stage stage);

	/* blitter: */
	bool (*blit)(struct fd_context *ctx, const struct pipe_blit_info *info);

	/* simple gpu "memcpy": */
	void (*mem_to_mem)(struct fd_ringbuffer *ring, struct pipe_resource *dst,
			unsigned dst_off, struct pipe_resource *src, unsigned src_off,
			unsigned sizedwords);

	/*
	 * Common pre-cooked VBO state (used for a3xx and later):
	 */

	/* for clear/gmem->mem vertices, and mem->gmem */
	struct pipe_resource *solid_vbuf;

	/* for mem->gmem tex coords: */
	struct pipe_resource *blit_texcoord_vbuf;

	/* vertex state for solid_vbuf:
	 *    - solid_vbuf / 12 / R32G32B32_FLOAT
	 */
	struct fd_vertex_state solid_vbuf_state;

	/* vertex state for blit_prog:
	 *    - blit_texcoord_vbuf / 8 / R32G32_FLOAT
	 *    - solid_vbuf / 12 / R32G32B32_FLOAT
	 */
	struct fd_vertex_state blit_vbuf_state;
};

static inline struct fd_context *
fd_context(struct pipe_context *pctx)
{
	return (struct fd_context *)pctx;
}

static inline void
fd_context_assert_locked(struct fd_context *ctx)
{
	pipe_mutex_assert_locked(ctx->screen->lock);
}

static inline void
fd_context_lock(struct fd_context *ctx)
{
	mtx_lock(&ctx->screen->lock);
}

static inline void
fd_context_unlock(struct fd_context *ctx)
{
	mtx_unlock(&ctx->screen->lock);
}

/* mark all state dirty: */
static inline void
fd_context_all_dirty(struct fd_context *ctx)
{
	ctx->dirty = ~0;
	for (unsigned i = 0; i < PIPE_SHADER_TYPES; i++)
		ctx->dirty_shader[i] = ~0;
}

static inline void
fd_context_all_clean(struct fd_context *ctx)
{
	ctx->dirty = 0;
	for (unsigned i = 0; i < PIPE_SHADER_TYPES; i++) {
		/* don't mark compute state as clean, since it is not emitted
		 * during normal draw call.  The places that call _all_dirty(),
		 * it is safe to mark compute state dirty as well, but the
		 * inverse is not true.
		 */
		if (i == PIPE_SHADER_COMPUTE)
			continue;
		ctx->dirty_shader[i] = 0;
	}
}

static inline struct pipe_scissor_state *
fd_context_get_scissor(struct fd_context *ctx)
{
	if (ctx->rasterizer && ctx->rasterizer->scissor)
		return &ctx->scissor;
	return &ctx->disabled_scissor;
}

static inline bool
fd_supported_prim(struct fd_context *ctx, unsigned prim)
{
	return (1 << prim) & ctx->primtype_mask;
}

static inline struct fd_batch *
fd_context_batch(struct fd_context *ctx)
{
	if (unlikely(!ctx->batch)) {
		struct fd_batch *batch =
			fd_batch_from_fb(&ctx->screen->batch_cache, ctx, &ctx->framebuffer);
		util_copy_framebuffer_state(&batch->framebuffer, &ctx->framebuffer);
		ctx->batch = batch;
		fd_context_all_dirty(ctx);
	}
	return ctx->batch;
}

static inline void
fd_batch_set_stage(struct fd_batch *batch, enum fd_render_stage stage)
{
	struct fd_context *ctx = batch->ctx;

	/* special case: internal blits (like mipmap level generation)
	 * go through normal draw path (via util_blitter_blit()).. but
	 * we need to ignore the FD_STAGE_DRAW which will be set, so we
	 * don't enable queries which should be paused during internal
	 * blits:
	 */
	if ((batch->stage == FD_STAGE_BLIT) &&
			(stage != FD_STAGE_NULL))
		return;

	if (ctx->query_set_stage)
		ctx->query_set_stage(batch, stage);

	batch->stage = stage;
}

void fd_context_setup_common_vbos(struct fd_context *ctx);
void fd_context_cleanup_common_vbos(struct fd_context *ctx);

struct pipe_context * fd_context_init(struct fd_context *ctx,
		struct pipe_screen *pscreen, const uint8_t *primtypes,
		void *priv, unsigned flags);

void fd_context_destroy(struct pipe_context *pctx);

#endif /* FREEDRENO_CONTEXT_H_ */