Nattapat Chaimanowong | 233b3d6 | 2018-10-12 12:02:18 +0100 | [diff] [blame] | 1 | // |
| 2 | // Copyright © 2017 Arm Ltd. All rights reserved. |
| 3 | // SPDX-License-Identifier: MIT |
| 4 | // |
| 5 | |
| 6 | #include "NeonConstantWorkload.hpp" |
| 7 | |
| 8 | #include <arm_compute/core/Types.h> |
Aron Virginas-Tar | c9cc804 | 2018-11-01 16:15:57 +0000 | [diff] [blame] | 9 | #include <Half.hpp> |
| 10 | #include <aclCommon/ArmComputeTensorUtils.hpp> |
| 11 | #include <neon/NeonTensorHandle.hpp> |
| 12 | #include <backendsCommon/CpuTensorHandle.hpp> |
| 13 | #include <backendsCommon/Workload.hpp> |
Nattapat Chaimanowong | 233b3d6 | 2018-10-12 12:02:18 +0100 | [diff] [blame] | 14 | |
| 15 | #include <boost/cast.hpp> |
| 16 | |
| 17 | namespace armnn |
| 18 | { |
| 19 | |
| 20 | NeonConstantWorkload::NeonConstantWorkload(const ConstantQueueDescriptor& descriptor, |
| 21 | const WorkloadInfo& info) |
| 22 | : BaseWorkload<ConstantQueueDescriptor>(descriptor, info) |
| 23 | , m_RanOnce(false) |
| 24 | { |
| 25 | } |
| 26 | |
| 27 | void NeonConstantWorkload::Execute() const |
| 28 | { |
| 29 | ARMNN_SCOPED_PROFILING_EVENT_NEON("NeonConstantWorkload_Execute"); |
| 30 | |
| 31 | using namespace armcomputetensorutils; |
| 32 | |
| 33 | // The intermediate tensor held by the corresponding layer output handler can be initialised with the |
| 34 | // given data on the first inference, then reused for subsequent inferences. |
| 35 | // The initialisation cannot happen at workload construction time since the ACL kernel for the next layer |
| 36 | // may not have been configured at the time. |
| 37 | if (!m_RanOnce) |
| 38 | { |
| 39 | const ConstantQueueDescriptor& data = this->m_Data; |
| 40 | |
| 41 | BOOST_ASSERT(data.m_LayerOutput != nullptr); |
| 42 | arm_compute::ITensor& output = |
| 43 | boost::polymorphic_downcast<NeonTensorHandle*>(data.m_Outputs[0])->GetTensor(); |
| 44 | arm_compute::DataType computeDataType = |
| 45 | boost::polymorphic_downcast<NeonTensorHandle*>(data.m_Outputs[0])->GetDataType(); |
| 46 | |
| 47 | switch (computeDataType) |
| 48 | { |
| 49 | case arm_compute::DataType::F16: |
| 50 | { |
| 51 | CopyArmComputeITensorData(data.m_LayerOutput->GetConstTensor<Half>(), output); |
| 52 | break; |
| 53 | } |
| 54 | case arm_compute::DataType::F32: |
| 55 | { |
| 56 | CopyArmComputeITensorData(data.m_LayerOutput->GetConstTensor<float>(), output); |
| 57 | break; |
| 58 | } |
| 59 | case arm_compute::DataType::QASYMM8: |
| 60 | { |
| 61 | CopyArmComputeITensorData(data.m_LayerOutput->GetConstTensor<uint8_t>(), output); |
| 62 | break; |
| 63 | } |
| 64 | default: |
| 65 | { |
| 66 | BOOST_ASSERT_MSG(false, "Unknown data type"); |
| 67 | break; |
| 68 | } |
| 69 | } |
| 70 | |
| 71 | m_RanOnce = true; |
| 72 | } |
| 73 | } |
| 74 | |
| 75 | } //namespace armnn |