IVGCVSW-4375 Add support for Transpose

 * Added TransposeLayer
 * Added CL, Neon and Ref Workloads
 * Added Transpose utilities
 * Added Serializer and Deserializer support
 * Added Quantizer support

Signed-off-by: Mike Kelly <mike.kelly@arm.com>
Change-Id: I04c755ba7cb5b1edf72b3c9f3c0314878032e3c7

4 files changed, 90 insertions, 0 deletions

diff --git a/src/backends/neon/workloads/CMakeLists.txt b/src/backends/neon/workloads/CMakeLists.txt
index 02ffedcf09..a932f8b852 100644
--- a/src/backends/neon/workloads/CMakeLists.txt
+++ b/src/backends/neon/workloads/CMakeLists.txt
@@ -98,6 +98,8 @@ list(APPEND armnnNeonBackendWorkloads_sources
     NeonSubtractionWorkload.hpp
     NeonTransposeConvolution2dWorkload.cpp
     NeonTransposeConvolution2dWorkload.hpp
+    NeonTransposeWorkload.cpp
+    NeonTransposeWorkload.hpp
     NeonWorkloads.hpp
     NeonWorkloadUtils.hpp
 )
diff --git a/src/backends/neon/workloads/NeonTransposeWorkload.cpp b/src/backends/neon/workloads/NeonTransposeWorkload.cpp
new file mode 100644
index 0000000000..c11f2df2d2
--- /dev/null
+++ b/src/backends/neon/workloads/NeonTransposeWorkload.cpp
@@ -0,0 +1,48 @@
+//
+// Copyright © 2020 Arm Ltd. All rights reserved.
+// SPDX-License-Identifier: MIT
+//
+
+#include "NeonTransposeWorkload.hpp"
+#include <neon/NeonTensorHandle.hpp>
+#include <aclCommon/ArmComputeTensorUtils.hpp>
+
+#include <arm_compute/core/Error.h>
+
+namespace armnn
+{
+
+arm_compute::Status NeonTransposeWorkloadValidate(const TensorInfo& input,
+                                                  const TensorInfo& output,
+                                                  const TransposeDescriptor& descriptor)
+{
+    const arm_compute::TensorInfo aclInputInfo = armcomputetensorutils::BuildArmComputeTensorInfo(input);
+    const arm_compute::TensorInfo aclOutputInfo = armcomputetensorutils::BuildArmComputeTensorInfo(output);
+    const armnn::PermutationVector& mappings = descriptor.m_DimMappings;
+
+    return arm_compute::NEPermute::validate(&aclInputInfo, &aclOutputInfo,
+                                            armcomputetensorutils::BuildArmComputeTransposeVector(mappings));
+}
+
+NeonTransposeWorkload::NeonTransposeWorkload(const TransposeQueueDescriptor& descriptor,
+                                             const WorkloadInfo& info)
+        : BaseWorkload<TransposeQueueDescriptor>(descriptor, info)
+{
+    m_Data.ValidateInputsOutputs(GetName(), 1, 1);
+
+    const arm_compute::ITensor& input = static_cast<IAclTensorHandle*>(m_Data.m_Inputs[0])->GetTensor();
+    arm_compute::ITensor& output = static_cast<IAclTensorHandle*>(m_Data.m_Outputs[0])->GetTensor();
+    const armnn::PermutationVector& mappings = m_Data.m_Parameters.m_DimMappings;
+
+    // Run the layer.
+    m_PermuteFunction.configure(&input, &output,
+                                armcomputetensorutils::BuildArmComputeTransposeVector(mappings));
+}
+
+void NeonTransposeWorkload::Execute() const
+{
+    ARMNN_SCOPED_PROFILING_EVENT_NEON(GetName() + "_Execute");
+    m_PermuteFunction.run();
+}
+
+} // namespace armnn
diff --git a/src/backends/neon/workloads/NeonTransposeWorkload.hpp b/src/backends/neon/workloads/NeonTransposeWorkload.hpp
new file mode 100644
index 0000000000..aab7b705e3
--- /dev/null
+++ b/src/backends/neon/workloads/NeonTransposeWorkload.hpp
@@ -0,0 +1,39 @@
+//
+// Copyright © 2020 Arm Ltd. All rights reserved.
+// SPDX-License-Identifier: MIT
+//
+
+#pragma once
+
+#include <backendsCommon/Workload.hpp>
+#include <backendsCommon/WorkloadData.hpp>
+#include <neon/workloads/NeonWorkloadUtils.hpp>
+
+#include <armnn/TypesUtils.hpp>
+#include <arm_compute/runtime/NEON/functions/NEPermute.h>
+
+#include <string>
+
+namespace armnn
+{
+arm_compute::Status NeonTransposeWorkloadValidate(const TensorInfo& input, const TensorInfo& output,
+                                                  const TransposeDescriptor& descriptor);
+
+class NeonTransposeWorkload : public BaseWorkload<TransposeQueueDescriptor>
+{
+public:
+    static const std::string& GetName()
+    {
+        static const std::string name = std::string("NeonTransposeWorkload");
+        return name;
+    }
+
+    NeonTransposeWorkload(const TransposeQueueDescriptor& descriptor, const WorkloadInfo& info);
+    void Execute() const override;
+
+private:
+    using BaseWorkload<TransposeQueueDescriptor>::m_Data;
+    mutable arm_compute::NEPermute m_PermuteFunction;
+};
+
+} // namespace armnn
diff --git a/src/backends/neon/workloads/NeonWorkloads.hpp b/src/backends/neon/workloads/NeonWorkloads.hpp
index b08483c443..52cd76f14b 100644
--- a/src/backends/neon/workloads/NeonWorkloads.hpp
+++ b/src/backends/neon/workloads/NeonWorkloads.hpp
@@ -50,3 +50,4 @@
 #include "NeonStridedSliceWorkload.hpp"
 #include "NeonSubtractionWorkload.hpp"
 #include "NeonTransposeConvolution2dWorkload.hpp"
+#include "NeonTransposeWorkload.hpp"