keidav01 | d74dc91 | 2018-12-10 18:16:07 +0000 | [diff] [blame] | 1 | // |
| 2 | // Copyright © 2017 Arm Ltd. All rights reserved. |
| 3 | // SPDX-License-Identifier: MIT |
| 4 | // |
| 5 | |
| 6 | #include "ClStridedSliceWorkload.hpp" |
| 7 | |
| 8 | #include "ClWorkloadUtils.hpp" |
| 9 | |
| 10 | #include <aclCommon/ArmComputeUtils.hpp> |
| 11 | #include <aclCommon/ArmComputeTensorUtils.hpp> |
| 12 | |
| 13 | #include <backendsCommon/CpuTensorHandle.hpp> |
| 14 | |
| 15 | #include <cl/ClLayerSupport.hpp> |
| 16 | #include <cl/ClTensorHandle.hpp> |
| 17 | #include <cl/ClLayerSupport.hpp> |
| 18 | |
| 19 | namespace armnn |
| 20 | { |
| 21 | |
| 22 | using namespace armcomputetensorutils; |
| 23 | |
| 24 | arm_compute::Status ClStridedSliceWorkloadValidate(const TensorInfo& input, |
| 25 | const TensorInfo& output, |
| 26 | const StridedSliceDescriptor& descriptor) |
| 27 | { |
| 28 | const arm_compute::TensorInfo aclInputInfo = BuildArmComputeTensorInfo(input, descriptor.m_DataLayout); |
| 29 | const arm_compute::TensorInfo aclOutputInfo = BuildArmComputeTensorInfo(output, descriptor.m_DataLayout); |
| 30 | |
| 31 | arm_compute::Coordinates starts; |
| 32 | arm_compute::Coordinates ends; |
| 33 | arm_compute::Coordinates strides; |
| 34 | |
| 35 | std::tie(starts, ends, strides) = SetClStridedSliceData(descriptor.m_Begin, descriptor.m_End, descriptor.m_Stride); |
| 36 | |
| 37 | int32_t begin_mask = descriptor.m_BeginMask; |
| 38 | int32_t end_mask = descriptor.m_EndMask; |
| 39 | int32_t shrink_axis_mask = descriptor.m_ShrinkAxisMask; |
| 40 | |
| 41 | return arm_compute::CLStridedSlice::validate(&aclInputInfo, |
| 42 | &aclOutputInfo, |
| 43 | starts, |
| 44 | ends, |
| 45 | strides, |
| 46 | begin_mask, |
| 47 | end_mask, |
| 48 | shrink_axis_mask); |
| 49 | } |
| 50 | |
| 51 | ClStridedSliceWorkload::ClStridedSliceWorkload(const StridedSliceQueueDescriptor& descriptor, |
| 52 | const WorkloadInfo& info) |
| 53 | : BaseWorkload<StridedSliceQueueDescriptor>(descriptor, info) |
| 54 | { |
| 55 | m_Data.ValidateInputsOutputs("ClStridedSliceWorkload", 1, 1); |
| 56 | |
| 57 | arm_compute::ICLTensor& input = static_cast<IClTensorHandle*>(m_Data.m_Inputs[0])->GetTensor(); |
| 58 | arm_compute::ICLTensor& output = static_cast<IClTensorHandle*>(m_Data.m_Outputs[0])->GetTensor(); |
| 59 | |
| 60 | arm_compute::Coordinates starts; |
| 61 | arm_compute::Coordinates ends; |
| 62 | arm_compute::Coordinates strides; |
| 63 | |
| 64 | std::tie(starts, ends, strides) = SetClStridedSliceData(m_Data.m_Parameters.m_Begin, |
| 65 | m_Data.m_Parameters.m_End, |
| 66 | m_Data.m_Parameters.m_Stride); |
| 67 | |
| 68 | int32_t begin_mask = m_Data.m_Parameters.m_BeginMask; |
| 69 | int32_t end_mask = m_Data.m_Parameters.m_EndMask; |
| 70 | int32_t shrink_axis_mask = m_Data.m_Parameters.m_ShrinkAxisMask; |
| 71 | |
| 72 | arm_compute::DataLayout aclDataLayout = ConvertDataLayout(m_Data.m_Parameters.m_DataLayout); |
| 73 | input.info()->set_data_layout(aclDataLayout); |
| 74 | output.info()->set_data_layout(aclDataLayout); |
| 75 | |
| 76 | m_StridedSliceLayer.configure(&input, |
| 77 | &output, |
| 78 | starts, |
| 79 | ends, |
| 80 | strides, |
| 81 | begin_mask, |
| 82 | end_mask, |
| 83 | shrink_axis_mask); |
| 84 | } |
| 85 | |
| 86 | void ClStridedSliceWorkload::Execute() const |
| 87 | { |
| 88 | ARMNN_SCOPED_PROFILING_EVENT_CL("ClStridedSliceWorkload_Execute"); |
| 89 | RunClFunction(m_StridedSliceLayer, CHECK_LOCATION()); |
| 90 | } |
| 91 | |
| 92 | } //namespace armnn |