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
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
|
/*
* Copyrigh 2016 Red Hat Inc.
* Based on anv:
* Copyright © 2015 Intel Corporation
*
* 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.
*/
#include "tu_private.h"
#include <assert.h>
#include <fcntl.h>
#include <stdbool.h>
#include <string.h>
#include <unistd.h>
#include "registers/adreno_pm4.xml.h"
#include "registers/adreno_common.xml.h"
#include "registers/a6xx.xml.h"
#include "nir/nir_builder.h"
#include "util/os_time.h"
#include "tu_cs.h"
#define NSEC_PER_SEC 1000000000ull
#define WAIT_TIMEOUT 5
/* It seems like sample counts need to be copied over to 16-byte aligned
* memory. */
struct PACKED slot_value {
uint64_t value;
uint64_t __padding;
};
struct PACKED occlusion_query_slot {
struct slot_value available; /* 0 when unavailable, 1 when available */
struct slot_value begin;
struct slot_value end;
struct slot_value result;
};
/* Returns the IOVA of a given uint64_t field in a given slot of a query
* pool. */
#define query_iova(type, pool, query, field) \
pool->bo.iova + pool->stride * query + offsetof(type, field) + \
offsetof(struct slot_value, value)
#define occlusion_query_iova(pool, query, field) \
query_iova(struct occlusion_query_slot, pool, query, field)
#define query_is_available(type, slot) \
((type*)slot)->available.value
#define occlusion_query_is_available(slot) \
query_is_available(struct occlusion_query_slot, slot)
/*
* Returns a pointer to a given slot in a query pool.
*/
static void* slot_address(struct tu_query_pool *pool, uint32_t query)
{
return (char*)pool->bo.map + query * pool->stride;
}
VkResult
tu_CreateQueryPool(VkDevice _device,
const VkQueryPoolCreateInfo *pCreateInfo,
const VkAllocationCallbacks *pAllocator,
VkQueryPool *pQueryPool)
{
TU_FROM_HANDLE(tu_device, device, _device);
assert(pCreateInfo->sType == VK_STRUCTURE_TYPE_QUERY_POOL_CREATE_INFO);
assert(pCreateInfo->queryCount > 0);
uint32_t slot_size;
switch (pCreateInfo->queryType) {
case VK_QUERY_TYPE_OCCLUSION:
slot_size = sizeof(struct occlusion_query_slot);
break;
case VK_QUERY_TYPE_PIPELINE_STATISTICS:
case VK_QUERY_TYPE_TIMESTAMP:
unreachable("Unimplemented query type");
default:
assert(!"Invalid query type");
}
struct tu_query_pool *pool =
vk_alloc2(&device->alloc, pAllocator, sizeof(*pool), 8,
VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
if (!pool)
return vk_error(device->instance, VK_ERROR_OUT_OF_HOST_MEMORY);
VkResult result = tu_bo_init_new(device, &pool->bo,
pCreateInfo->queryCount * slot_size);
if (result != VK_SUCCESS) {
vk_free2(&device->alloc, pAllocator, pool);
return result;
}
result = tu_bo_map(device, &pool->bo);
if (result != VK_SUCCESS) {
tu_bo_finish(device, &pool->bo);
vk_free2(&device->alloc, pAllocator, pool);
return result;
}
/* Initialize all query statuses to unavailable */
memset(pool->bo.map, 0, pool->bo.size);
pool->type = pCreateInfo->queryType;
pool->stride = slot_size;
pool->size = pCreateInfo->queryCount;
pool->pipeline_statistics = pCreateInfo->pipelineStatistics;
*pQueryPool = tu_query_pool_to_handle(pool);
return VK_SUCCESS;
}
void
tu_DestroyQueryPool(VkDevice _device,
VkQueryPool _pool,
const VkAllocationCallbacks *pAllocator)
{
TU_FROM_HANDLE(tu_device, device, _device);
TU_FROM_HANDLE(tu_query_pool, pool, _pool);
if (!pool)
return;
tu_bo_finish(device, &pool->bo);
vk_free2(&device->alloc, pAllocator, pool);
}
/* Wait on the the availability status of a query up until a timeout. */
static VkResult
wait_for_available(struct tu_device *device, struct tu_query_pool *pool,
uint32_t query)
{
/* TODO: Use the MSM_IOVA_WAIT ioctl to wait on the available bit in a
* scheduler friendly way instead of busy polling once the patch has landed
* upstream. */
struct occlusion_query_slot *slot = slot_address(pool, query);
uint64_t abs_timeout = os_time_get_absolute_timeout(
WAIT_TIMEOUT * NSEC_PER_SEC);
while(os_time_get_nano() < abs_timeout) {
if (occlusion_query_is_available(slot))
return VK_SUCCESS;
}
return vk_error(device->instance, VK_TIMEOUT);
}
/* Writes a query value to a buffer from the CPU. */
static void
write_query_value_cpu(char* base,
uint32_t offset,
uint64_t value,
VkQueryResultFlags flags)
{
if (flags & VK_QUERY_RESULT_64_BIT) {
*(uint64_t*)(base + (offset * sizeof(uint64_t))) = value;
} else {
*(uint32_t*)(base + (offset * sizeof(uint32_t))) = value;
}
}
static VkResult
get_occlusion_query_pool_results(struct tu_device *device,
struct tu_query_pool *pool,
uint32_t firstQuery,
uint32_t queryCount,
size_t dataSize,
void *pData,
VkDeviceSize stride,
VkQueryResultFlags flags)
{
assert(dataSize >= stride * queryCount);
char *result_base = pData;
VkResult result = VK_SUCCESS;
for (uint32_t i = 0; i < queryCount; i++) {
uint32_t query = firstQuery + i;
struct occlusion_query_slot *slot = slot_address(pool, query);
bool available = occlusion_query_is_available(slot);
if ((flags & VK_QUERY_RESULT_WAIT_BIT) && !available) {
VkResult wait_result = wait_for_available(device, pool, query);
if (wait_result != VK_SUCCESS)
return wait_result;
available = true;
} else if (!(flags & VK_QUERY_RESULT_PARTIAL_BIT) && !available) {
/* From the Vulkan 1.1.130 spec:
*
* If VK_QUERY_RESULT_WAIT_BIT and VK_QUERY_RESULT_PARTIAL_BIT are
* both not set then no result values are written to pData for
* queries that are in the unavailable state at the time of the
* call, and vkGetQueryPoolResults returns VK_NOT_READY. However,
* availability state is still written to pData for those queries
* if VK_QUERY_RESULT_WITH_AVAILABILITY_BIT is set.
*/
result = VK_NOT_READY;
if (!(flags & VK_QUERY_RESULT_WITH_AVAILABILITY_BIT)) {
result_base += stride;
continue;
}
}
if (available)
write_query_value_cpu(result_base, 0, slot->result.value, flags);
else if (flags & VK_QUERY_RESULT_PARTIAL_BIT)
/* From the Vulkan 1.1.130 spec:
*
* If VK_QUERY_RESULT_PARTIAL_BIT is set, VK_QUERY_RESULT_WAIT_BIT
* is not set, and the query’s status is unavailable, an
* intermediate result value between zero and the final result
* value is written to pData for that query.
*
* Just return 0 here for simplicity since it's a valid result.
*/
write_query_value_cpu(result_base, 0, 0, flags);
if (flags & VK_QUERY_RESULT_WITH_AVAILABILITY_BIT)
/* From the Vulkan 1.1.130 spec:
*
* If VK_QUERY_RESULT_WITH_AVAILABILITY_BIT is set, the final
* integer value written for each query is non-zero if the query’s
* status was available or zero if the status was unavailable.
*/
write_query_value_cpu(result_base, 1, available, flags);
result_base += stride;
}
return result;
}
VkResult
tu_GetQueryPoolResults(VkDevice _device,
VkQueryPool queryPool,
uint32_t firstQuery,
uint32_t queryCount,
size_t dataSize,
void *pData,
VkDeviceSize stride,
VkQueryResultFlags flags)
{
TU_FROM_HANDLE(tu_device, device, _device);
TU_FROM_HANDLE(tu_query_pool, pool, queryPool);
assert(firstQuery + queryCount <= pool->size);
switch (pool->type) {
case VK_QUERY_TYPE_OCCLUSION: {
return get_occlusion_query_pool_results(device, pool, firstQuery,
queryCount, dataSize, pData, stride, flags);
}
case VK_QUERY_TYPE_PIPELINE_STATISTICS:
case VK_QUERY_TYPE_TIMESTAMP:
unreachable("Unimplemented query type");
default:
assert(!"Invalid query type");
}
return VK_SUCCESS;
}
/* Copies a query value from one buffer to another from the GPU. */
static void
copy_query_value_gpu(struct tu_cmd_buffer *cmdbuf,
struct tu_cs *cs,
uint64_t src_iova,
uint64_t base_write_iova,
uint32_t offset,
VkQueryResultFlags flags) {
uint32_t element_size = flags & VK_QUERY_RESULT_64_BIT ?
sizeof(uint64_t) : sizeof(uint32_t);
uint64_t write_iova = base_write_iova + (offset * element_size);
tu_cs_emit_pkt7(cs, CP_MEM_TO_MEM, 5);
uint32_t mem_to_mem_flags = flags & VK_QUERY_RESULT_64_BIT ?
CP_MEM_TO_MEM_0_DOUBLE : 0;
tu_cs_emit(cs, mem_to_mem_flags);
tu_cs_emit_qw(cs, write_iova);
tu_cs_emit_qw(cs, src_iova);
}
static void
emit_copy_occlusion_query_pool_results(struct tu_cmd_buffer *cmdbuf,
struct tu_cs *cs,
struct tu_query_pool *pool,
uint32_t firstQuery,
uint32_t queryCount,
struct tu_buffer *buffer,
VkDeviceSize dstOffset,
VkDeviceSize stride,
VkQueryResultFlags flags)
{
/* From the Vulkan 1.1.130 spec:
*
* vkCmdCopyQueryPoolResults is guaranteed to see the effect of previous
* uses of vkCmdResetQueryPool in the same queue, without any additional
* synchronization.
*
* To ensure that previous writes to the available bit are coherent, first
* wait for all writes to complete.
*/
tu_cs_emit_pkt7(cs, CP_WAIT_MEM_WRITES, 0);
for (uint32_t i = 0; i < queryCount; i++) {
uint32_t query = firstQuery + i;
uint64_t available_iova = occlusion_query_iova(pool, query, available);
uint64_t result_iova = occlusion_query_iova(pool, query, result);
uint64_t buffer_iova = tu_buffer_iova(buffer) + dstOffset + i * stride;
/* Wait for the available bit to be set if executed with the
* VK_QUERY_RESULT_WAIT_BIT flag. */
if (flags & VK_QUERY_RESULT_WAIT_BIT) {
tu_cs_emit_pkt7(cs, CP_WAIT_REG_MEM, 6);
tu_cs_emit(cs, CP_WAIT_REG_MEM_0_FUNCTION(WRITE_EQ) |
CP_WAIT_REG_MEM_0_POLL_MEMORY);
tu_cs_emit_qw(cs, available_iova);
tu_cs_emit(cs, CP_WAIT_REG_MEM_3_REF(0x1));
tu_cs_emit(cs, CP_WAIT_REG_MEM_4_MASK(~0));
tu_cs_emit(cs, CP_WAIT_REG_MEM_5_DELAY_LOOP_CYCLES(16));
}
if (flags & VK_QUERY_RESULT_PARTIAL_BIT) {
/* Unconditionally copying the bo->result into the buffer here is
* valid because we only set bo->result on vkCmdEndQuery. Thus, even
* if the query is unavailable, this will copy the correct partial
* value of 0.
*/
copy_query_value_gpu(cmdbuf, cs, result_iova, buffer_iova,
0 /* offset */, flags);
} else {
/* Conditionally copy bo->result into the buffer based on whether the
* query is available.
*
* NOTE: For the conditional packets to be executed, CP_COND_EXEC
* tests that ADDR0 != 0 and ADDR1 < REF. The packet here simply tests
* that 0 < available < 2, aka available == 1.
*/
tu_cs_reserve(cs, 7 + 6);
tu_cs_emit_pkt7(cs, CP_COND_EXEC, 6);
tu_cs_emit_qw(cs, available_iova);
tu_cs_emit_qw(cs, available_iova);
tu_cs_emit(cs, CP_COND_EXEC_4_REF(0x2));
tu_cs_emit(cs, 6); /* Cond execute the next 6 DWORDS */
/* Start of conditional execution */
copy_query_value_gpu(cmdbuf, cs, result_iova, buffer_iova,
0 /* offset */, flags);
/* End of conditional execution */
}
if (flags & VK_QUERY_RESULT_WITH_AVAILABILITY_BIT) {
copy_query_value_gpu(cmdbuf, cs, available_iova, buffer_iova,
1 /* offset */, flags);
}
}
tu_bo_list_add(&cmdbuf->bo_list, buffer->bo, MSM_SUBMIT_BO_WRITE);
}
void
tu_CmdCopyQueryPoolResults(VkCommandBuffer commandBuffer,
VkQueryPool queryPool,
uint32_t firstQuery,
uint32_t queryCount,
VkBuffer dstBuffer,
VkDeviceSize dstOffset,
VkDeviceSize stride,
VkQueryResultFlags flags)
{
TU_FROM_HANDLE(tu_cmd_buffer, cmdbuf, commandBuffer);
TU_FROM_HANDLE(tu_query_pool, pool, queryPool);
TU_FROM_HANDLE(tu_buffer, buffer, dstBuffer);
struct tu_cs *cs = &cmdbuf->cs;
assert(firstQuery + queryCount <= pool->size);
switch (pool->type) {
case VK_QUERY_TYPE_OCCLUSION: {
return emit_copy_occlusion_query_pool_results(cmdbuf, cs, pool,
firstQuery, queryCount, buffer, dstOffset, stride, flags);
}
case VK_QUERY_TYPE_PIPELINE_STATISTICS:
case VK_QUERY_TYPE_TIMESTAMP:
unreachable("Unimplemented query type");
default:
assert(!"Invalid query type");
}
}
static void
emit_reset_occlusion_query_pool(struct tu_cmd_buffer *cmdbuf,
struct tu_query_pool *pool,
uint32_t firstQuery,
uint32_t queryCount)
{
struct tu_cs *cs = &cmdbuf->cs;
for (uint32_t i = 0; i < queryCount; i++) {
uint32_t query = firstQuery + i;
uint64_t available_iova = occlusion_query_iova(pool, query, available);
uint64_t result_iova = occlusion_query_iova(pool, query, result);
tu_cs_emit_pkt7(cs, CP_MEM_WRITE, 4);
tu_cs_emit_qw(cs, available_iova);
tu_cs_emit_qw(cs, 0x0);
tu_cs_emit_pkt7(cs, CP_MEM_WRITE, 4);
tu_cs_emit_qw(cs, result_iova);
tu_cs_emit_qw(cs, 0x0);
}
}
void
tu_CmdResetQueryPool(VkCommandBuffer commandBuffer,
VkQueryPool queryPool,
uint32_t firstQuery,
uint32_t queryCount)
{
TU_FROM_HANDLE(tu_cmd_buffer, cmdbuf, commandBuffer);
TU_FROM_HANDLE(tu_query_pool, pool, queryPool);
switch (pool->type) {
case VK_QUERY_TYPE_OCCLUSION:
emit_reset_occlusion_query_pool(cmdbuf, pool, firstQuery, queryCount);
break;
case VK_QUERY_TYPE_PIPELINE_STATISTICS:
case VK_QUERY_TYPE_TIMESTAMP:
unreachable("Unimplemented query type");
default:
assert(!"Invalid query type");
}
tu_bo_list_add(&cmdbuf->bo_list, &pool->bo, MSM_SUBMIT_BO_WRITE);
}
static void
emit_begin_occlusion_query(struct tu_cmd_buffer *cmdbuf,
struct tu_query_pool *pool,
uint32_t query)
{
/* From the Vulkan 1.1.130 spec:
*
* A query must begin and end inside the same subpass of a render pass
* instance, or must both begin and end outside of a render pass
* instance.
*
* Unlike on an immediate-mode renderer, Turnip renders all tiles on
* vkCmdEndRenderPass, not individually on each vkCmdDraw*. As such, if a
* query begins/ends inside the same subpass of a render pass, we need to
* record the packets on the secondary draw command stream. cmdbuf->draw_cs
* is then run on every tile during render, so we just need to accumulate
* sample counts in slot->result to compute the query result.
*/
struct tu_cs *cs = cmdbuf->state.pass ? &cmdbuf->draw_cs : &cmdbuf->cs;
uint64_t begin_iova = occlusion_query_iova(pool, query, begin);
tu_cs_emit_regs(cs,
A6XX_RB_SAMPLE_COUNT_CONTROL(.copy = true));
tu_cs_emit_regs(cs,
A6XX_RB_SAMPLE_COUNT_ADDR_LO(begin_iova));
tu_cs_emit_pkt7(cs, CP_EVENT_WRITE, 1);
tu_cs_emit(cs, ZPASS_DONE);
}
void
tu_CmdBeginQuery(VkCommandBuffer commandBuffer,
VkQueryPool queryPool,
uint32_t query,
VkQueryControlFlags flags)
{
TU_FROM_HANDLE(tu_cmd_buffer, cmdbuf, commandBuffer);
TU_FROM_HANDLE(tu_query_pool, pool, queryPool);
assert(query < pool->size);
switch (pool->type) {
case VK_QUERY_TYPE_OCCLUSION:
/* In freedreno, there is no implementation difference between
* GL_SAMPLES_PASSED and GL_ANY_SAMPLES_PASSED, so we can similarly
* ignore the VK_QUERY_CONTROL_PRECISE_BIT flag here.
*/
emit_begin_occlusion_query(cmdbuf, pool, query);
break;
case VK_QUERY_TYPE_PIPELINE_STATISTICS:
case VK_QUERY_TYPE_TIMESTAMP:
unreachable("Unimplemented query type");
default:
assert(!"Invalid query type");
}
tu_bo_list_add(&cmdbuf->bo_list, &pool->bo, MSM_SUBMIT_BO_WRITE);
}
static void
emit_end_occlusion_query(struct tu_cmd_buffer *cmdbuf,
struct tu_query_pool *pool,
uint32_t query)
{
/* Ending an occlusion query happens in a few steps:
* 1) Set the slot->end to UINT64_MAX.
* 2) Set up the SAMPLE_COUNT registers and trigger a CP_EVENT_WRITE to
* write the current sample count value into slot->end.
* 3) Since (2) is asynchronous, wait until slot->end is not equal to
* UINT64_MAX before continuing via CP_WAIT_REG_MEM.
* 4) Accumulate the results of the query (slot->end - slot->begin) into
* slot->result.
* 5) If vkCmdEndQuery is *not* called from within the scope of a render
* pass, set the slot's available bit since the query is now done.
* 6) If vkCmdEndQuery *is* called from within the scope of a render
* pass, we cannot mark as available yet since the commands in
* draw_cs are not run until vkCmdEndRenderPass.
*/
const struct tu_render_pass *pass = cmdbuf->state.pass;
struct tu_cs *cs = pass ? &cmdbuf->draw_cs : &cmdbuf->cs;
uint64_t available_iova = occlusion_query_iova(pool, query, available);
uint64_t begin_iova = occlusion_query_iova(pool, query, begin);
uint64_t end_iova = occlusion_query_iova(pool, query, end);
uint64_t result_iova = occlusion_query_iova(pool, query, result);
tu_cs_emit_pkt7(cs, CP_MEM_WRITE, 4);
tu_cs_emit_qw(cs, end_iova);
tu_cs_emit_qw(cs, 0xffffffffffffffffull);
tu_cs_emit_pkt7(cs, CP_WAIT_MEM_WRITES, 0);
tu_cs_emit_regs(cs,
A6XX_RB_SAMPLE_COUNT_CONTROL(.copy = true));
tu_cs_emit_regs(cs,
A6XX_RB_SAMPLE_COUNT_ADDR_LO(end_iova));
tu_cs_emit_pkt7(cs, CP_EVENT_WRITE, 1);
tu_cs_emit(cs, ZPASS_DONE);
tu_cs_emit_pkt7(cs, CP_WAIT_REG_MEM, 6);
tu_cs_emit(cs, CP_WAIT_REG_MEM_0_FUNCTION(WRITE_NE) |
CP_WAIT_REG_MEM_0_POLL_MEMORY);
tu_cs_emit_qw(cs, end_iova);
tu_cs_emit(cs, CP_WAIT_REG_MEM_3_REF(0xffffffff));
tu_cs_emit(cs, CP_WAIT_REG_MEM_4_MASK(~0));
tu_cs_emit(cs, CP_WAIT_REG_MEM_5_DELAY_LOOP_CYCLES(16));
/* result (dst) = result (srcA) + end (srcB) - begin (srcC) */
tu_cs_emit_pkt7(cs, CP_MEM_TO_MEM, 9);
tu_cs_emit(cs, CP_MEM_TO_MEM_0_DOUBLE | CP_MEM_TO_MEM_0_NEG_C);
tu_cs_emit_qw(cs, result_iova);
tu_cs_emit_qw(cs, result_iova);
tu_cs_emit_qw(cs, end_iova);
tu_cs_emit_qw(cs, begin_iova);
tu_cs_emit_pkt7(cs, CP_WAIT_MEM_WRITES, 0);
if (pass)
/* Technically, queries should be tracked per-subpass, but here we track
* at the render pass level to simply the code a bit. This is safe
* because the only commands that use the available bit are
* vkCmdCopyQueryPoolResults and vkCmdResetQueryPool, both of which
* cannot be invoked from inside a render pass scope.
*/
cs = &cmdbuf->draw_epilogue_cs;
tu_cs_emit_pkt7(cs, CP_MEM_WRITE, 4);
tu_cs_emit_qw(cs, available_iova);
tu_cs_emit_qw(cs, 0x1);
}
void
tu_CmdEndQuery(VkCommandBuffer commandBuffer,
VkQueryPool queryPool,
uint32_t query)
{
TU_FROM_HANDLE(tu_cmd_buffer, cmdbuf, commandBuffer);
TU_FROM_HANDLE(tu_query_pool, pool, queryPool);
assert(query < pool->size);
switch (pool->type) {
case VK_QUERY_TYPE_OCCLUSION:
emit_end_occlusion_query(cmdbuf, pool, query);
break;
case VK_QUERY_TYPE_PIPELINE_STATISTICS:
case VK_QUERY_TYPE_TIMESTAMP:
unreachable("Unimplemented query type");
default:
assert(!"Invalid query type");
}
tu_bo_list_add(&cmdbuf->bo_list, &pool->bo, MSM_SUBMIT_BO_WRITE);
}
void
tu_CmdWriteTimestamp(VkCommandBuffer commandBuffer,
VkPipelineStageFlagBits pipelineStage,
VkQueryPool queryPool,
uint32_t query)
{
}
|