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
|
//
// Copyright © 2017 Arm Ltd and Contributors. All rights reserved.
// SPDX-License-Identifier: MIT
//
#include <ResolveType.hpp>
#include "WorkloadUtils.hpp"
#include <armnn/backends/MemCopyWorkload.hpp>
#include <armnn/backends/TensorHandle.hpp>
#include <armnn/utility/PolymorphicDowncast.hpp>
#include <cstring>
namespace armnn
{
namespace
{
template <typename SrcTensorHandleType, typename DstTensorHandleType>
void GatherTensorHandlePairs(const MemCopyQueueDescriptor& descriptor,
std::vector<std::pair<SrcTensorHandleType*, DstTensorHandleType*>>& tensorHandlePairs)
{
const unsigned int numInputs = static_cast<unsigned int>(descriptor.m_Inputs.size());
tensorHandlePairs.reserve(numInputs);
for (unsigned int i = 0; i < numInputs; ++i)
{
SrcTensorHandleType* const srcTensorHandle = PolymorphicDowncast<SrcTensorHandleType*>(
descriptor.m_Inputs[i]);
DstTensorHandleType* const dstTensorHandle = PolymorphicDowncast<DstTensorHandleType*>(
descriptor.m_Outputs[i]);
tensorHandlePairs.emplace_back(srcTensorHandle, dstTensorHandle);
}
}
} //namespace
CopyMemGenericWorkload::CopyMemGenericWorkload(const MemCopyQueueDescriptor& descriptor,
const WorkloadInfo& info)
: BaseWorkload<MemCopyQueueDescriptor>(descriptor, info)
{
GatherTensorHandlePairs(descriptor, m_TensorHandlePairs);
}
void CopyMemGenericWorkload::Execute() const
{
ARMNN_SCOPED_PROFILING_EVENT(Compute::Undefined, "CopyMemGeneric_Execute");
auto copyFunc = [](void* dst, const void* src, size_t size)
{
memcpy(dst, src, size);
};
for (const auto& pair : m_TensorHandlePairs)
{
CopyTensorContentsGeneric(pair.first, pair.second, copyFunc);
}
}
void CopyMemGenericWorkload::ExecuteAsync(WorkingMemDescriptor& descriptor)
{
ARMNN_SCOPED_PROFILING_EVENT(Compute::Undefined, "CopyMemGeneric_Execute_WorkingMemDescriptor");
std::vector<TensorHandlePair> tensorHandlePairs;
GatherTensorHandlePairs(descriptor, tensorHandlePairs);
auto copyFunc = [](void* dst, const void* src, size_t size)
{
memcpy(dst, src, size);
};
for (const auto& pair : tensorHandlePairs)
{
CopyTensorContentsGeneric(pair.first, pair.second, copyFunc);
}
}
} //namespace armnn
|
