src/backends/neon/workloads/NeonSubtractionFloatWorkload.cpp

   1 //
   2 // Copyright © 2017 Arm Ltd. All rights reserved.
   3 // SPDX-License-Identifier: MIT
   4 //
   5
   6 #include "NeonSubtractionFloatWorkload.hpp"
   7 #include <aclCommon/ArmComputeTensorUtils.hpp>
   8 #include <backendsCommon/CpuTensorHandle.hpp>
   9
  10 namespace armnn
  11 {
  12
  13 arm_compute::Status NeonSubtractionWorkloadValidate(const TensorInfo& input0,
  14                                                     const TensorInfo& input1,
  15                                                     const TensorInfo& output)
  16 {
  17     const arm_compute::TensorInfo aclInput0 = armcomputetensorutils::BuildArmComputeTensorInfo(input0);
  18     const arm_compute::TensorInfo aclInput1 = armcomputetensorutils::BuildArmComputeTensorInfo(input1);
  19     const arm_compute::TensorInfo aclOutput = armcomputetensorutils::BuildArmComputeTensorInfo(output);
  20
  21     return arm_compute::NEArithmeticSubtraction::validate(&aclInput0,
  22                                                           &aclInput1,
  23                                                           &aclOutput,
  24                                                           arm_compute::ConvertPolicy::SATURATE);
  25 }
  26
  27 NeonSubtractionFloatWorkload::NeonSubtractionFloatWorkload(const SubtractionQueueDescriptor& descriptor,
  28                                                            const WorkloadInfo& info)
  29     : FloatWorkload<SubtractionQueueDescriptor>(descriptor, info)
  30 {
  31     m_Data.ValidateInputsOutputs("NeonSubtractionFloatWorkload", 2, 1);
  32
  33     arm_compute::ITensor& input1 = boost::polymorphic_downcast<INeonTensorHandle*>(m_Data.m_Inputs[0])->GetTensor();
  34     arm_compute::ITensor& input2 = boost::polymorphic_downcast<INeonTensorHandle*>(m_Data.m_Inputs[1])->GetTensor();
  35     arm_compute::ITensor& output = boost::polymorphic_downcast<INeonTensorHandle*>(m_Data.m_Outputs[0])->GetTensor();
  36
  37     m_SubLayer.configure(&input1, &input2, &output, arm_compute::ConvertPolicy::SATURATE);
  38 }
  39
  40 void NeonSubtractionFloatWorkload::Execute() const
  41 {
  42     ARMNN_SCOPED_PROFILING_EVENT_NEON("NeonSubtractionFloatWorkload_Execute");
  43     m_SubLayer.run();
  44 }
  45
  46 } //namespace armnn