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
|
/*
* Copyright (C) 2018-2025 Intel Corporation
*
* SPDX-License-Identifier: MIT
*
*/
#pragma once
#include "shared/source/command_stream/command_stream_receiver.h"
#include "shared/source/memory_manager/allocation_properties.h"
#include "shared/source/memory_manager/internal_allocation_storage.h"
#include "shared/source/memory_manager/memory_manager.h"
#include "opencl/source/command_queue/command_queue_hw.h"
#include "opencl/source/mem_obj/buffer.h"
#include "opencl/source/memory_manager/mem_obj_surface.h"
namespace NEO {
template <typename GfxFamily>
cl_int CommandQueueHw<GfxFamily>::enqueueFillBuffer(
Buffer *buffer,
const void *pattern,
size_t patternSize,
size_t offset,
size_t size,
cl_uint numEventsInWaitList,
const cl_event *eventWaitList,
cl_event *event) {
auto memoryManager = getDevice().getMemoryManager();
DEBUG_BREAK_IF(nullptr == memoryManager);
auto commandStreamReceiverOwnership = getGpgpuCommandStreamReceiver().obtainUniqueOwnership();
auto storageWithAllocations = getGpgpuCommandStreamReceiver().getInternalAllocationStorage();
auto allocationType = AllocationType::fillPattern;
auto patternAllocation = storageWithAllocations->obtainReusableAllocation(patternSize, allocationType).release();
commandStreamReceiverOwnership.unlock();
if (!patternAllocation) {
patternAllocation = memoryManager->allocateGraphicsMemoryWithProperties({getDevice().getRootDeviceIndex(), alignUp(patternSize, MemoryConstants::cacheLineSize), AllocationType::fillPattern, getDevice().getDeviceBitfield()});
}
if (patternSize == 1) {
auto patternValue = *reinterpret_cast<const uint8_t *>(pattern);
int patternInt = static_cast<uint32_t>((patternValue << 24) | (patternValue << 16) | (patternValue << 8) | patternValue);
memcpy_s(patternAllocation->getUnderlyingBuffer(), sizeof(uint32_t), &patternInt, sizeof(uint32_t));
} else if (patternSize == 2) {
auto patternValue = *reinterpret_cast<const uint16_t *>(pattern);
int patternInt = static_cast<uint32_t>((patternValue << 16) | patternValue);
memcpy_s(patternAllocation->getUnderlyingBuffer(), sizeof(uint32_t), &patternInt, sizeof(uint32_t));
} else {
memcpy_s(patternAllocation->getUnderlyingBuffer(), patternSize, pattern, patternSize);
}
const bool isStateless = forceStateless(buffer->getSize());
const bool useHeapless = this->getHeaplessModeEnabled();
auto builtInType = EBuiltInOps::adjustBuiltinType<EBuiltInOps::fillBuffer>(isStateless, useHeapless);
auto &builder = BuiltInDispatchBuilderOp::getBuiltinDispatchInfoBuilder(builtInType,
this->getClDevice());
BuiltInOwnershipWrapper builtInLock(builder, this->context);
BuiltinOpParams dc;
auto multiGraphicsAllocation = MultiGraphicsAllocation(getDevice().getRootDeviceIndex());
multiGraphicsAllocation.addAllocation(patternAllocation);
MemObj patternMemObj(this->context, 0, {}, 0, 0, alignUp(patternSize, 4), patternAllocation->getUnderlyingBuffer(),
patternAllocation->getUnderlyingBuffer(), std::move(multiGraphicsAllocation), false, false, true);
dc.srcMemObj = &patternMemObj;
dc.dstMemObj = buffer;
dc.dstOffset = {offset, 0, 0};
dc.size = {size, 0, 0};
MultiDispatchInfo dispatchInfo(dc);
builder.buildDispatchInfos(dispatchInfo);
MemObjSurface s1(buffer);
GeneralSurface s2(patternAllocation);
Surface *surfaces[] = {&s1, &s2};
const auto enqueueResult = enqueueHandler<CL_COMMAND_FILL_BUFFER>(
surfaces,
false,
dispatchInfo,
numEventsInWaitList,
eventWaitList,
event);
auto storageForAllocation = getGpgpuCommandStreamReceiver().getInternalAllocationStorage();
storageForAllocation->storeAllocationWithTaskCount(std::unique_ptr<GraphicsAllocation>(patternAllocation), REUSABLE_ALLOCATION, taskCount);
return enqueueResult;
}
} // namespace NEO
|
