diff options
author | Conor Kennedy <conor.kennedy@arm.com> | 2019-03-08 08:24:41 +0000 |
---|---|---|
committer | Kevin May <kevin.may@arm.com> | 2019-03-11 10:38:20 +0000 |
commit | b99480bc522fd98a8e587478b78906e8a57936df (patch) | |
tree | fc6b5ce8e28eda60fa791cec097f65572b641408 /src/backends/neon/workloads/NeonMultiplicationWorkload.cpp | |
parent | a6d45c77688978518ca2113ff68ca7d98aaca36a (diff) | |
download | armnn-b99480bc522fd98a8e587478b78906e8a57936df.tar.gz |
IVGCVSW-2792 Update Neon Mul/Sub workload for latest CL master
* Update Neon Subtraction/Multiplication to use BaseWorkload
* Add Uint8 tests for Subtraction/Multiplication
Change-Id: Ied2aa048924817a4db0025b9bd9e54151667bd74
Signed-off-by: Conor Kennedy <conor.kennedy@arm.com>
Diffstat (limited to 'src/backends/neon/workloads/NeonMultiplicationWorkload.cpp')
-rw-r--r-- | src/backends/neon/workloads/NeonMultiplicationWorkload.cpp | 63 |
1 files changed, 63 insertions, 0 deletions
diff --git a/src/backends/neon/workloads/NeonMultiplicationWorkload.cpp b/src/backends/neon/workloads/NeonMultiplicationWorkload.cpp new file mode 100644 index 0000000000..6398b65946 --- /dev/null +++ b/src/backends/neon/workloads/NeonMultiplicationWorkload.cpp @@ -0,0 +1,63 @@ +// +// Copyright © 2017 Arm Ltd. All rights reserved. +// SPDX-License-Identifier: MIT +// + +#include "NeonMultiplicationWorkload.hpp" + +#include "NeonWorkloadUtils.hpp" + +#include <arm_compute/runtime/NEON/functions/NEPixelWiseMultiplication.h> + +namespace armnn +{ + +arm_compute::Status NeonMultiplicationWorkloadValidate(const TensorInfo& input0, + const TensorInfo& input1, + const TensorInfo& output) +{ + const arm_compute::TensorInfo aclInput1 = armcomputetensorutils::BuildArmComputeTensorInfo(input0); + const arm_compute::TensorInfo aclInput2 = armcomputetensorutils::BuildArmComputeTensorInfo(input1); + const arm_compute::TensorInfo aclOutput = armcomputetensorutils::BuildArmComputeTensorInfo(output); + + // At the time of writing, configure() will fail if a rounding policy other than TO_ZERO is supplied to it, + // when providing a scale of 1.0 for F32 tensors, even though the provided rounding policy appears to be + // ignored for F32 tensors. + return arm_compute::NEPixelWiseMultiplication::validate(&aclInput1, + &aclInput2, + &aclOutput, + 1.0f, + arm_compute::ConvertPolicy::SATURATE, + arm_compute::RoundingPolicy::TO_ZERO); +} + +NeonMultiplicationWorkload::NeonMultiplicationWorkload(const MultiplicationQueueDescriptor& descriptor, + const WorkloadInfo& info) + : BaseWorkload<MultiplicationQueueDescriptor>(descriptor, info) +{ + m_Data.ValidateInputsOutputs("NeonMultiplicationWorkload", 2, 1); + + arm_compute::ITensor& input1 = boost::polymorphic_downcast<INeonTensorHandle*>(m_Data.m_Inputs[0])->GetTensor(); + arm_compute::ITensor& input2 = boost::polymorphic_downcast<INeonTensorHandle*>(m_Data.m_Inputs[1])->GetTensor(); + arm_compute::ITensor& output = boost::polymorphic_downcast<INeonTensorHandle*>(m_Data.m_Outputs[0])->GetTensor(); + + // At the time of writing, configure() will fail if a rounding policy other than TO_ZERO is supplied to it, + // when providing a scale of 1.0 for F32 tensors, even though the provided rounding policy appears to be + // ignored for F32 tensors. + auto layer = std::make_unique<arm_compute::NEPixelWiseMultiplication>(); + layer->configure(&input1, + &input2, + &output, + 1.0f, + arm_compute::ConvertPolicy::SATURATE, + arm_compute::RoundingPolicy::TO_ZERO); + m_PixelWiseMultiplication.reset(layer.release()); +} + +void NeonMultiplicationWorkload::Execute() const +{ + ARMNN_SCOPED_PROFILING_EVENT_NEON("NeonMultiplicationWorkload_Execute"); + m_PixelWiseMultiplication->run(); +} + +} //namespace armnn |