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
|
/* -*- mode: C; c-file-style: "k&r"; tab-width 4; indent-tabs-mode: t; -*- */
/*
* Copyright (C) 2014 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>
*/
#include "pipe/p_state.h"
#include "util/u_memory.h"
#include "util/u_inlines.h"
#include "freedreno_query_hw.h"
#include "freedreno_context.h"
#include "freedreno_util.h"
struct fd_hw_sample_period {
struct fd_hw_sample *start, *end;
struct list_head list;
};
/* maps query_type to sample provider idx: */
static int pidx(unsigned query_type)
{
switch (query_type) {
case PIPE_QUERY_OCCLUSION_COUNTER:
return 0;
case PIPE_QUERY_OCCLUSION_PREDICATE:
return 1;
default:
return -1;
}
}
static struct fd_hw_sample *
get_sample(struct fd_context *ctx, struct fd_ringbuffer *ring,
unsigned query_type)
{
struct fd_hw_sample *samp = NULL;
int idx = pidx(query_type);
if (!ctx->sample_cache[idx]) {
ctx->sample_cache[idx] =
ctx->sample_providers[idx]->get_sample(ctx, ring);
}
fd_hw_sample_reference(ctx, &samp, ctx->sample_cache[idx]);
return samp;
}
static void
clear_sample_cache(struct fd_context *ctx)
{
int i;
for (i = 0; i < ARRAY_SIZE(ctx->sample_cache); i++)
fd_hw_sample_reference(ctx, &ctx->sample_cache[i], NULL);
}
static bool
is_active(struct fd_hw_query *hq, enum fd_render_stage stage)
{
return !!(hq->provider->active & stage);
}
static void
resume_query(struct fd_context *ctx, struct fd_hw_query *hq,
struct fd_ringbuffer *ring)
{
assert(!hq->period);
hq->period = util_slab_alloc(&ctx->sample_period_pool);
list_inithead(&hq->period->list);
hq->period->start = get_sample(ctx, ring, hq->base.type);
/* NOTE: util_slab_alloc() does not zero out the buffer: */
hq->period->end = NULL;
}
static void
pause_query(struct fd_context *ctx, struct fd_hw_query *hq,
struct fd_ringbuffer *ring)
{
assert(hq->period && !hq->period->end);
hq->period->end = get_sample(ctx, ring, hq->base.type);
list_addtail(&hq->period->list, &hq->current_periods);
hq->period = NULL;
}
static void
destroy_periods(struct fd_context *ctx, struct list_head *list)
{
struct fd_hw_sample_period *period, *s;
LIST_FOR_EACH_ENTRY_SAFE(period, s, list, list) {
fd_hw_sample_reference(ctx, &period->start, NULL);
fd_hw_sample_reference(ctx, &period->end, NULL);
list_del(&period->list);
util_slab_free(&ctx->sample_period_pool, period);
}
}
static void
fd_hw_destroy_query(struct fd_context *ctx, struct fd_query *q)
{
struct fd_hw_query *hq = fd_hw_query(q);
destroy_periods(ctx, &hq->periods);
destroy_periods(ctx, &hq->current_periods);
list_del(&hq->list);
free(hq);
}
static void
fd_hw_begin_query(struct fd_context *ctx, struct fd_query *q)
{
struct fd_hw_query *hq = fd_hw_query(q);
if (q->active)
return;
/* begin_query() should clear previous results: */
destroy_periods(ctx, &hq->periods);
if (is_active(hq, ctx->stage))
resume_query(ctx, hq, ctx->ring);
q->active = true;
/* add to active list: */
list_del(&hq->list);
list_addtail(&hq->list, &ctx->active_queries);
}
static void
fd_hw_end_query(struct fd_context *ctx, struct fd_query *q)
{
struct fd_hw_query *hq = fd_hw_query(q);
if (!q->active)
return;
if (is_active(hq, ctx->stage))
pause_query(ctx, hq, ctx->ring);
q->active = false;
/* move to current list: */
list_del(&hq->list);
list_addtail(&hq->list, &ctx->current_queries);
}
/* helper to get ptr to specified sample: */
static void * sampptr(struct fd_hw_sample *samp, uint32_t n, void *ptr)
{
return ((char *)ptr) + (samp->tile_stride * n) + samp->offset;
}
static boolean
fd_hw_get_query_result(struct fd_context *ctx, struct fd_query *q,
boolean wait, union pipe_query_result *result)
{
struct fd_hw_query *hq = fd_hw_query(q);
const struct fd_hw_sample_provider *p = hq->provider;
struct fd_hw_sample_period *period;
if (q->active)
return false;
/* if the app tries to read back the query result before the
* batch is submitted, that forces us to flush so that there
* are actually results to wait for:
*/
if (!LIST_IS_EMPTY(&hq->list)) {
/* if app didn't actually trigger any cmdstream, then
* we have nothing to do:
*/
if (!ctx->needs_flush)
return true;
DBG("reading query result forces flush!");
fd_context_render(&ctx->base);
}
util_query_clear_result(result, q->type);
if (LIST_IS_EMPTY(&hq->periods))
return true;
assert(LIST_IS_EMPTY(&hq->list));
assert(LIST_IS_EMPTY(&hq->current_periods));
assert(!hq->period);
/* if !wait, then check the last sample (the one most likely to
* not be ready yet) and bail if it is not ready:
*/
if (!wait) {
int ret;
period = LIST_ENTRY(struct fd_hw_sample_period,
hq->periods.prev, list);
ret = fd_bo_cpu_prep(period->end->bo, ctx->screen->pipe,
DRM_FREEDRENO_PREP_READ | DRM_FREEDRENO_PREP_NOSYNC);
if (ret)
return false;
fd_bo_cpu_fini(period->end->bo);
}
/* sum the result across all sample periods: */
LIST_FOR_EACH_ENTRY(period, &hq->periods, list) {
struct fd_hw_sample *start = period->start;
struct fd_hw_sample *end = period->end;
unsigned i;
/* start and end samples should be from same batch: */
assert(start->bo == end->bo);
assert(start->num_tiles == end->num_tiles);
for (i = 0; i < start->num_tiles; i++) {
void *ptr;
fd_bo_cpu_prep(start->bo, ctx->screen->pipe,
DRM_FREEDRENO_PREP_READ);
ptr = fd_bo_map(start->bo);
p->accumulate_result(ctx, sampptr(period->start, i, ptr),
sampptr(period->end, i, ptr), result);
fd_bo_cpu_fini(start->bo);
}
}
return true;
}
static const struct fd_query_funcs hw_query_funcs = {
.destroy_query = fd_hw_destroy_query,
.begin_query = fd_hw_begin_query,
.end_query = fd_hw_end_query,
.get_query_result = fd_hw_get_query_result,
};
struct fd_query *
fd_hw_create_query(struct fd_context *ctx, unsigned query_type)
{
struct fd_hw_query *hq;
struct fd_query *q;
int idx = pidx(query_type);
if ((idx < 0) || !ctx->sample_providers[idx])
return NULL;
hq = CALLOC_STRUCT(fd_hw_query);
if (!hq)
return NULL;
hq->provider = ctx->sample_providers[idx];
list_inithead(&hq->periods);
list_inithead(&hq->current_periods);
list_inithead(&hq->list);
q = &hq->base;
q->funcs = &hw_query_funcs;
q->type = query_type;
return q;
}
struct fd_hw_sample *
fd_hw_sample_init(struct fd_context *ctx, uint32_t size)
{
struct fd_hw_sample *samp = util_slab_alloc(&ctx->sample_pool);
pipe_reference_init(&samp->reference, 1);
samp->size = size;
samp->offset = ctx->next_sample_offset;
/* NOTE: util_slab_alloc() does not zero out the buffer: */
samp->bo = NULL;
samp->num_tiles = 0;
samp->tile_stride = 0;
ctx->next_sample_offset += size;
return samp;
}
void
__fd_hw_sample_destroy(struct fd_context *ctx, struct fd_hw_sample *samp)
{
if (samp->bo)
fd_bo_del(samp->bo);
util_slab_free(&ctx->sample_pool, samp);
}
static void
prepare_sample(struct fd_hw_sample *samp, struct fd_bo *bo,
uint32_t num_tiles, uint32_t tile_stride)
{
if (samp->bo) {
assert(samp->bo == bo);
assert(samp->num_tiles == num_tiles);
assert(samp->tile_stride == tile_stride);
return;
}
samp->bo = bo;
samp->num_tiles = num_tiles;
samp->tile_stride = tile_stride;
}
static void
prepare_query(struct fd_hw_query *hq, struct fd_bo *bo,
uint32_t num_tiles, uint32_t tile_stride)
{
struct fd_hw_sample_period *period, *s;
/* prepare all the samples in the query: */
LIST_FOR_EACH_ENTRY_SAFE(period, s, &hq->current_periods, list) {
prepare_sample(period->start, bo, num_tiles, tile_stride);
prepare_sample(period->end, bo, num_tiles, tile_stride);
/* move from current_periods list to periods list: */
list_del(&period->list);
list_addtail(&period->list, &hq->periods);
}
}
static void
prepare_queries(struct fd_context *ctx, struct fd_bo *bo,
uint32_t num_tiles, uint32_t tile_stride,
struct list_head *list, bool remove)
{
struct fd_hw_query *hq, *s;
LIST_FOR_EACH_ENTRY_SAFE(hq, s, list, list) {
prepare_query(hq, bo, num_tiles, tile_stride);
if (remove)
list_delinit(&hq->list);
}
}
/* called from gmem code once total storage requirements are known (ie.
* number of samples times number of tiles)
*/
void
fd_hw_query_prepare(struct fd_context *ctx, uint32_t num_tiles)
{
uint32_t tile_stride = ctx->next_sample_offset;
struct fd_bo *bo;
if (ctx->query_bo)
fd_bo_del(ctx->query_bo);
if (tile_stride > 0) {
bo = fd_bo_new(ctx->dev, tile_stride * num_tiles,
DRM_FREEDRENO_GEM_CACHE_WCOMBINE |
DRM_FREEDRENO_GEM_TYPE_KMEM);
} else {
bo = NULL;
}
ctx->query_bo = bo;
ctx->query_tile_stride = tile_stride;
prepare_queries(ctx, bo, num_tiles, tile_stride,
&ctx->active_queries, false);
prepare_queries(ctx, bo, num_tiles, tile_stride,
&ctx->current_queries, true);
/* reset things for next batch: */
ctx->next_sample_offset = 0;
}
void
fd_hw_query_prepare_tile(struct fd_context *ctx, uint32_t n,
struct fd_ringbuffer *ring)
{
uint32_t tile_stride = ctx->query_tile_stride;
uint32_t offset = tile_stride * n;
/* bail if no queries: */
if (tile_stride == 0)
return;
fd_wfi(ctx, ring);
OUT_PKT0 (ring, HW_QUERY_BASE_REG, 1);
OUT_RELOCW(ring, ctx->query_bo, offset, 0, 0);
}
void
fd_hw_query_set_stage(struct fd_context *ctx, struct fd_ringbuffer *ring,
enum fd_render_stage stage)
{
/* 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 ((ctx->stage == FD_STAGE_BLIT) &&
(stage != FD_STAGE_NULL))
return;
if (stage != ctx->stage) {
struct fd_hw_query *hq;
LIST_FOR_EACH_ENTRY(hq, &ctx->active_queries, list) {
bool was_active = is_active(hq, ctx->stage);
bool now_active = is_active(hq, stage);
if (now_active && !was_active)
resume_query(ctx, hq, ring);
else if (was_active && !now_active)
pause_query(ctx, hq, ring);
}
}
clear_sample_cache(ctx);
ctx->stage = stage;
}
void
fd_hw_query_register_provider(struct pipe_context *pctx,
const struct fd_hw_sample_provider *provider)
{
struct fd_context *ctx = fd_context(pctx);
int idx = pidx(provider->query_type);
assert((0 <= idx) && (idx < MAX_HW_SAMPLE_PROVIDERS));
assert(!ctx->sample_providers[idx]);
ctx->sample_providers[idx] = provider;
}
void
fd_hw_query_init(struct pipe_context *pctx)
{
struct fd_context *ctx = fd_context(pctx);
util_slab_create(&ctx->sample_pool, sizeof(struct fd_hw_sample),
16, UTIL_SLAB_SINGLETHREADED);
util_slab_create(&ctx->sample_period_pool, sizeof(struct fd_hw_sample_period),
16, UTIL_SLAB_SINGLETHREADED);
list_inithead(&ctx->active_queries);
list_inithead(&ctx->current_queries);
}
void
fd_hw_query_fini(struct pipe_context *pctx)
{
struct fd_context *ctx = fd_context(pctx);
util_slab_destroy(&ctx->sample_pool);
util_slab_destroy(&ctx->sample_period_pool);
}
|