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
|
//
// Copyright © 2021 Arm Ltd and Contributors. All rights reserved.
// SPDX-License-Identifier: MIT
//
#include "NeonReduceWorkload.hpp"
#include <aclCommon/ArmComputeUtils.hpp>
#include <aclCommon/ArmComputeTensorUtils.hpp>
#include <neon/NeonTensorHandle.hpp>
#include "NeonWorkloadUtils.hpp"
namespace armnn
{
using namespace armcomputetensorutils;
arm_compute::Status NeonReduceWorkloadValidate(const TensorInfo& input,
const TensorInfo& output,
const ReduceDescriptor& desc)
{
if ( desc.m_vAxis.size()==1 || desc.m_vAxis.empty())
{
const arm_compute::TensorInfo aclInputInfo = armcomputetensorutils::BuildArmComputeTensorInfo(input);
const arm_compute::TensorInfo aclOutputInfo = armcomputetensorutils::BuildArmComputeTensorInfo(output);
arm_compute::Coordinates coords = BuildArmComputeReductionCoordinates(aclInputInfo.num_dimensions(),
input.GetNumDimensions(),
desc.m_vAxis);
return arm_compute::NEReductionOperation::validate(&aclInputInfo,
&aclOutputInfo,
static_cast<unsigned int>(coords[0]),
ConvertReductionOperationToAcl(desc),
desc.m_KeepDims);
}
else
{
// Validate layer if there are multiple axes.
arm_compute::Status status;
IS_MULTI_AXES_REDUCE_SUPPORTED(NeonReduceWorkloadValidate, input, desc, status);
return status;
}
}
NeonReduceWorkload::NeonReduceWorkload(const ReduceQueueDescriptor& descriptor, const WorkloadInfo& info)
: BaseWorkload<ReduceQueueDescriptor>(descriptor, info)
{
m_Data.ValidateInputsOutputs("NeonReduceWorkload", 1, 1);
arm_compute::ITensor& input = static_cast<IAclTensorHandle*>(m_Data.m_Inputs[0])->GetTensor();
arm_compute::ITensor& output = static_cast<IAclTensorHandle*>(m_Data.m_Outputs[0])->GetTensor();
arm_compute::Coordinates coords = BuildArmComputeReductionCoordinates(input.info()->num_dimensions(),
info.m_InputTensorInfos[0].GetNumDimensions(),
m_Data.m_Parameters.m_vAxis);
m_Layer.configure(&input,
&output,
static_cast<unsigned int>(coords[0]),
ConvertReductionOperationToAcl(m_Data.m_Parameters),
m_Data.m_Parameters.m_KeepDims);
}
void NeonReduceWorkload::Execute() const
{
ARMNN_SCOPED_PROFILING_EVENT_NEON("NeonReduceWorkload_Execute");
m_Layer.run();
}
} //namespace armnn
|
