From 4fcda0101ec3d110c1d6d7bee5c83416b645528a Mon Sep 17 00:00:00 2001 From: telsoa01 Date: Fri, 9 Mar 2018 14:13:49 +0000 Subject: Release 18.02 Change-Id: Id3c11dc5ee94ef664374a988fcc6901e9a232fa6 --- .../NeonMultiplicationFloat32Workload.cpp | 41 ++++++++++++++++++++++ 1 file changed, 41 insertions(+) create mode 100644 src/armnn/backends/NeonWorkloads/NeonMultiplicationFloat32Workload.cpp (limited to 'src/armnn/backends/NeonWorkloads/NeonMultiplicationFloat32Workload.cpp') diff --git a/src/armnn/backends/NeonWorkloads/NeonMultiplicationFloat32Workload.cpp b/src/armnn/backends/NeonWorkloads/NeonMultiplicationFloat32Workload.cpp new file mode 100644 index 0000000000..58ce7b74ba --- /dev/null +++ b/src/armnn/backends/NeonWorkloads/NeonMultiplicationFloat32Workload.cpp @@ -0,0 +1,41 @@ +// +// Copyright © 2017 Arm Ltd. All rights reserved. +// See LICENSE file in the project root for full license information. +// + +#include "NeonMultiplicationFloat32Workload.hpp" + + +namespace armnn +{ + +NeonMultiplicationFloat32Workload::NeonMultiplicationFloat32Workload(const MultiplicationQueueDescriptor& descriptor, + const WorkloadInfo& info) + : Float32Workload(descriptor, info) +{ + m_Data.ValidateInputsOutputs("NeonMultiplicationFloat32Workload", 2, 1); + + arm_compute::ITensor& input1 = boost::polymorphic_downcast(m_Data.m_Inputs[0])->GetTensor(); + arm_compute::ITensor& input2 = boost::polymorphic_downcast(m_Data.m_Inputs[1])->GetTensor(); + arm_compute::ITensor& output = boost::polymorphic_downcast(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. + m_PixelWiseMultiplication.configure(&input1, + &input2, + &output, + 1.0f, + arm_compute::ConvertPolicy::SATURATE, + arm_compute::RoundingPolicy::TO_ZERO); +} + +void NeonMultiplicationFloat32Workload::Execute() const +{ + ARMNN_SCOPED_PROFILING_EVENT(Compute::CpuAcc, "NeonMultiplicationFloat32Workload_Execute"); + m_PixelWiseMultiplication.run(); +} + +} //namespace armnn + + -- cgit v1.2.1