diff options
Diffstat (limited to 'src/armnn/backends/ClWorkloads/ClSoftmaxUint8Workload.cpp')
-rw-r--r-- | src/armnn/backends/ClWorkloads/ClSoftmaxUint8Workload.cpp | 39 |
1 files changed, 39 insertions, 0 deletions
diff --git a/src/armnn/backends/ClWorkloads/ClSoftmaxUint8Workload.cpp b/src/armnn/backends/ClWorkloads/ClSoftmaxUint8Workload.cpp new file mode 100644 index 0000000000..9e856fea94 --- /dev/null +++ b/src/armnn/backends/ClWorkloads/ClSoftmaxUint8Workload.cpp @@ -0,0 +1,39 @@ +// +// Copyright © 2017 Arm Ltd. All rights reserved. +// See LICENSE file in the project root for full license information. +// + +#include "ClSoftmaxUint8Workload.hpp" +#include "backends/ClTensorHandle.hpp" +#include "backends/CpuTensorHandle.hpp" + +namespace armnn +{ + +ClSoftmaxUint8Workload::ClSoftmaxUint8Workload(const SoftmaxQueueDescriptor& descriptor, const WorkloadInfo& info) + : Uint8Workload<SoftmaxQueueDescriptor>(descriptor, info) +{ + m_Data.ValidateInputsOutputs("ClSoftmaxUint8Workload", 1, 1); + + arm_compute::ICLTensor& input = static_cast<ClTensorHandle*>(m_Data.m_Inputs[0])->GetTensor(); + arm_compute::ICLTensor& output = static_cast<ClTensorHandle*>(m_Data.m_Outputs[0])->GetTensor(); + + const auto outputQuantization = output.info()->quantization_info(); + + if ((outputQuantization.scale != (1.0f / 256.0f)) || (outputQuantization.offset != 0)) + { + throw InvalidArgumentException( + "Invalid quantization for output. Only scale = 1.0f / 256.0f and offset = 0 supported"); + } + + m_SoftmaxLayer.configure(&input, &output, descriptor.m_Parameters.m_Beta); +} + +void ClSoftmaxUint8Workload::Execute() const +{ + ARMNN_SCOPED_PROFILING_EVENT(Compute::GpuAcc, "ClSoftmaxUint8Workload_Execute"); + + m_SoftmaxLayer.run(); +} + +} //namespace armnn |