xref: /aosp_15_r20/external/armnn/src/backends/neon/workloads/NeonMeanWorkload.cpp (revision 89c4ff92f2867872bb9e2354d150bf0c8c502810)
1 //
2 // Copyright © 2017 Arm Ltd and Contributors. All rights reserved.
3 // SPDX-License-Identifier: MIT
4 //
5 
6 #include "NeonMeanWorkload.hpp"
7 
8 #include <aclCommon/ArmComputeTensorUtils.hpp>
9 
10 #include <neon/NeonTensorHandle.hpp>
11 
12 #include "NeonWorkloadUtils.hpp"
13 
14 namespace armnn
15 {
16 using namespace armcomputetensorutils;
17 
NeonMeanWorkloadValidate(const TensorInfo & input,const TensorInfo & output,const MeanDescriptor & descriptor)18 arm_compute::Status NeonMeanWorkloadValidate(const TensorInfo& input,
19                                              const TensorInfo& output,
20                                              const MeanDescriptor& descriptor)
21 {
22     const arm_compute::TensorInfo aclInputInfo  = armcomputetensorutils::BuildArmComputeTensorInfo(input);
23     const arm_compute::TensorInfo aclOutputInfo = armcomputetensorutils::BuildArmComputeTensorInfo(output);
24 
25     arm_compute::Coordinates coords = BuildArmComputeReductionCoordinates(aclInputInfo.num_dimensions(),
26                                                                           input.GetNumDimensions(),
27                                                                           descriptor.m_Axis);
28 
29     return arm_compute::NEReduceMean::validate(&aclInputInfo, coords, descriptor.m_KeepDims, &aclOutputInfo);
30 }
31 
NeonMeanWorkload(const MeanQueueDescriptor & descriptor,const WorkloadInfo & info)32 NeonMeanWorkload::NeonMeanWorkload(const MeanQueueDescriptor& descriptor, const WorkloadInfo& info)
33     : NeonBaseWorkload<MeanQueueDescriptor>(descriptor, info)
34 {
35     // Report Profiling Details
36     ARMNN_REPORT_PROFILING_WORKLOAD_DESC("NeonMeanWorkload_Construct",
37                                          descriptor.m_Parameters,
38                                          info,
39                                          this->GetGuid());
40 
41     m_Data.ValidateInputsOutputs("NeonMeanWorkload", 1, 1);
42 
43     arm_compute::ITensor& input  = static_cast<IAclTensorHandle*>(m_Data.m_Inputs[0])->GetTensor();
44     arm_compute::ITensor& output = static_cast<IAclTensorHandle*>(m_Data.m_Outputs[0])->GetTensor();
45 
46     arm_compute::Coordinates coords = BuildArmComputeReductionCoordinates(input.info()->num_dimensions(),
47                                                                           info.m_InputTensorInfos[0].GetNumDimensions(),
48                                                                           m_Data.m_Parameters.m_Axis);
49 
50     m_Layer.configure(&input, coords, m_Data.m_Parameters.m_KeepDims, &output);
51 }
52 
Execute() const53 void NeonMeanWorkload::Execute() const
54 {
55     ARMNN_SCOPED_PROFILING_EVENT_NEON_GUID("NeonMeanWorkload_Execute", this->GetGuid());
56     m_Layer.run();
57 }
58 
59 } //namespace armnn
60