diff options
author | Nattapat Chaimanowong <nattapat.chaimanowong@arm.com> | 2018-12-20 14:14:06 +0000 |
---|---|---|
committer | Matteo Martincigh <matteo.martincigh@arm.com> | 2018-12-20 16:11:34 +0000 |
commit | 4e6597a26059bfa31ea24e190e2afe0558cea4c8 (patch) | |
tree | 45f38dc5fb24653ceebad21be278e1f2922a6802 /src/backends/neon/workloads/NeonMaximumWorkload.cpp | |
parent | a959ee5f7753c073d62b0bf16ae86d18ce6fd8cc (diff) | |
download | armnn-4e6597a26059bfa31ea24e190e2afe0558cea4c8.tar.gz |
IVGCVSW-2212 Add Neon support for Maximum operator
Change-Id: Iddae3486641d2d195827ae4e0a9dfa3f7ccd65e3
Diffstat (limited to 'src/backends/neon/workloads/NeonMaximumWorkload.cpp')
-rw-r--r-- | src/backends/neon/workloads/NeonMaximumWorkload.cpp | 45 |
1 files changed, 45 insertions, 0 deletions
diff --git a/src/backends/neon/workloads/NeonMaximumWorkload.cpp b/src/backends/neon/workloads/NeonMaximumWorkload.cpp new file mode 100644 index 0000000000..9994af98d1 --- /dev/null +++ b/src/backends/neon/workloads/NeonMaximumWorkload.cpp @@ -0,0 +1,45 @@ +// +// Copyright © 2017 Arm Ltd. All rights reserved. +// SPDX-License-Identifier: MIT +// + +#include "NeonMaximumWorkload.hpp" +#include <aclCommon/ArmComputeTensorUtils.hpp> +#include <backendsCommon/CpuTensorHandle.hpp> + +namespace armnn +{ + +arm_compute::Status NeonMaximumWorkloadValidate(const TensorInfo& input0, + const TensorInfo& input1, + const TensorInfo& output) +{ + const arm_compute::TensorInfo aclInput0 = armcomputetensorutils::BuildArmComputeTensorInfo(input0); + const arm_compute::TensorInfo aclInput1 = armcomputetensorutils::BuildArmComputeTensorInfo(input1); + const arm_compute::TensorInfo aclOutput = armcomputetensorutils::BuildArmComputeTensorInfo(output); + + return arm_compute::NEElementwiseMax::validate(&aclInput0, + &aclInput1, + &aclOutput); +} + +NeonMaximumWorkload::NeonMaximumWorkload(const MaximumQueueDescriptor& descriptor, + const WorkloadInfo& info) + : BaseWorkload<MaximumQueueDescriptor>(descriptor, info) +{ + m_Data.ValidateInputsOutputs("NeonMaximumWorkload", 2, 1); + + arm_compute::ITensor& input0 = boost::polymorphic_downcast<INeonTensorHandle*>(m_Data.m_Inputs[0])->GetTensor(); + arm_compute::ITensor& input1 = boost::polymorphic_downcast<INeonTensorHandle*>(m_Data.m_Inputs[1])->GetTensor(); + arm_compute::ITensor& output = boost::polymorphic_downcast<INeonTensorHandle*>(m_Data.m_Outputs[0])->GetTensor(); + + m_MaxLayer.configure(&input0, &input1, &output); +} + +void NeonMaximumWorkload::Execute() const +{ + ARMNN_SCOPED_PROFILING_EVENT_NEON("NeonMaximumWorkload_Execute"); + m_MaxLayer.run(); +} + +} //namespace armnn |