IVGCVSW-7400 POW IVGCVSW-7278 SQUARED_DIFFERENCE to CpuAcc and GpuAcc

* Add POW SQUARED_DIFFERENCE and Unit tests for CpuAcc and GpuAcc

Signed-off-by: John Mcloughlin <john.mcloughlin@arm.com>
Signed-off-by: Teresa Charlin <teresa.charlinreyes@arm.com>

Change-Id: Ifa78af2a2fda2074586d8e4d9a506b1b13fa5755

4 files changed, 130 insertions, 2 deletions

diff --git a/src/backends/neon/workloads/CMakeLists.txt b/src/backends/neon/workloads/CMakeLists.txt
index a3eb883079..f3f12842c1 100644
--- a/src/backends/neon/workloads/CMakeLists.txt
+++ b/src/backends/neon/workloads/CMakeLists.txt
@@ -1,5 +1,5 @@
 #
-# Copyright © 2017 Arm Ltd and Contributors. All rights reserved.
+# Copyright © 2017,2023 Arm Ltd and Contributors. All rights reserved.
 # SPDX-License-Identifier: MIT
 #
 
@@ -44,6 +44,8 @@ list(APPEND armnnNeonBackendWorkloads_sources
     NeonDequantizeWorkload.hpp
     NeonDetectionPostProcessWorkload.cpp
     NeonDetectionPostProcessWorkload.hpp
+    NeonElementwiseBinaryWorkload.cpp
+    NeonElementwiseBinaryWorkload.hpp
     NeonExpWorkload.cpp
     NeonExpWorkload.hpp
     NeonFillWorkload.cpp
diff --git a/src/backends/neon/workloads/NeonElementwiseBinaryWorkload.cpp b/src/backends/neon/workloads/NeonElementwiseBinaryWorkload.cpp
new file mode 100644
index 0000000000..839ec6c5ad
--- /dev/null
+++ b/src/backends/neon/workloads/NeonElementwiseBinaryWorkload.cpp
@@ -0,0 +1,92 @@
+//
+// Copyright © 2023 Arm Ltd and Contributors. All rights reserved.
+// SPDX-License-Identifier: MIT
+//
+
+#include "NeonElementwiseBinaryWorkload.hpp"
+#include "NeonWorkloadUtils.hpp"
+
+#include <aclCommon/ArmComputeTensorUtils.hpp>
+#include <aclCommon/ArmComputeUtils.hpp>
+
+#include <armnn/utility/PolymorphicDowncast.hpp>
+#include <armnn/backends/TensorHandle.hpp>
+
+#include <arm_compute/runtime/NEON/functions/NEElementwiseOperations.h>
+
+namespace armnn
+{
+
+arm_compute::Status NeonElementwiseBinaryWorkloadValidate(const TensorInfo& input0,
+                                                          const TensorInfo& input1,
+                                                          const TensorInfo& output,
+                                                          const ElementwiseBinaryDescriptor& descriptor,
+                                                          const ActivationDescriptor* activationDescriptor)
+{
+    const arm_compute::TensorInfo aclInput0 = armcomputetensorutils::BuildArmComputeTensorInfo(input0);
+    const arm_compute::TensorInfo aclInput1 = armcomputetensorutils::BuildArmComputeTensorInfo(input1);
+    const arm_compute::TensorInfo aclOutput = armcomputetensorutils::BuildArmComputeTensorInfo(output);
+
+    const arm_compute::ActivationLayerInfo activationInfo = ConvertActivationDescriptorToAclActivationLayerInfo(
+            activationDescriptor);
+
+    switch (descriptor.m_Operation)
+    {
+        case armnn::BinaryOperation::Power:
+            return arm_compute::NEElementwisePower::validate(&aclInput0,
+                                                             &aclInput1,
+                                                             &aclOutput,
+                                                             activationInfo);
+        case armnn::BinaryOperation::SqDiff:
+            return arm_compute::NEElementwiseSquaredDiff::validate(&aclInput0,
+                                                                   &aclInput1,
+                                                                   &aclOutput,
+                                                                   activationInfo);
+        default:
+            throw InvalidArgumentException("Unknown binary operator", CHECK_LOCATION());
+    }
+}
+
+
+NeonElementwiseBinaryWorkload::NeonElementwiseBinaryWorkload(const ElementwiseBinaryQueueDescriptor& descriptor,
+                                                             const WorkloadInfo& info)
+    : NeonBaseWorkload<ElementwiseBinaryQueueDescriptor>(descriptor, info)
+{
+    m_Data.ValidateInputsOutputs("NeonElementwiseBinaryWorkload", 2, 1);
+
+    arm_compute::ITensor& input1 = PolymorphicDowncast<IAclTensorHandle*>(m_Data.m_Inputs[0])->GetTensor();
+    arm_compute::ITensor& input2 = PolymorphicDowncast<IAclTensorHandle*>(m_Data.m_Inputs[1])->GetTensor();
+    arm_compute::ITensor& output = PolymorphicDowncast<IAclTensorHandle*>(m_Data.m_Outputs[0])->GetTensor();
+
+    const arm_compute::ActivationLayerInfo activationInfo = ConvertAdditionalInfoToAclActivationLayerInfo(descriptor);
+
+    ARMNN_SCOPED_PROFILING_EVENT(Compute::Undefined, "NeonElementwiseBinaryWorkload_configure");
+
+    switch (descriptor.m_Parameters.m_Operation)
+    {
+        case armnn::BinaryOperation::Power:
+        {
+            auto powerLayer = std::make_unique<arm_compute::NEElementwisePower>();
+            powerLayer->configure(&input1, &input2, &output, activationInfo);
+            m_ElementwiseBinaryLayer.reset(powerLayer.release());
+            break;
+        }
+        case armnn::BinaryOperation::SqDiff:
+        {
+            auto SqDiffLayer = std::make_unique<arm_compute::NEElementwiseSquaredDiff>();
+            SqDiffLayer->configure(&input1, &input2, &output, activationInfo);
+            m_ElementwiseBinaryLayer.reset(SqDiffLayer.release());
+            break;
+        }
+        default:
+            throw InvalidArgumentException("Unknown binary operator", CHECK_LOCATION());
+    }
+}
+
+void NeonElementwiseBinaryWorkload::Execute() const
+{
+    ARMNN_SCOPED_PROFILING_EVENT_NEON_GUID("NeonElementwiseBinaryWorkload_Execute", this->GetGuid());
+    m_ElementwiseBinaryLayer->run();
+}
+
+} //namespace armnn
\ No newline at end of file
diff --git a/src/backends/neon/workloads/NeonElementwiseBinaryWorkload.hpp b/src/backends/neon/workloads/NeonElementwiseBinaryWorkload.hpp
new file mode 100644
index 0000000000..9cdc7ae11b
--- /dev/null
+++ b/src/backends/neon/workloads/NeonElementwiseBinaryWorkload.hpp
@@ -0,0 +1,33 @@
+//
+// Copyright © 2023 Arm Ltd and Contributors. All rights reserved.
+// SPDX-License-Identifier: MIT
+//
+
+#pragma once
+
+#include "NeonBaseWorkload.hpp"
+
+#include <arm_compute/core/Error.h>
+#include <arm_compute/core/Types.h>
+#include <arm_compute/runtime/IFunction.h>
+
+namespace armnn
+{
+
+arm_compute::Status NeonElementwiseBinaryWorkloadValidate(const TensorInfo& input0,
+                                                          const TensorInfo& input1,
+                                                          const TensorInfo& output,
+                                                          const ElementwiseBinaryDescriptor& descriptor,
+                                                          const ActivationDescriptor* activationDescriptor = nullptr);
+
+class NeonElementwiseBinaryWorkload : public NeonBaseWorkload<ElementwiseBinaryQueueDescriptor>
+{
+public:
+    NeonElementwiseBinaryWorkload(const ElementwiseBinaryQueueDescriptor& descriptor, const WorkloadInfo& info);
+    virtual void Execute() const override;
+
+private:
+    std::unique_ptr<arm_compute::IFunction> m_ElementwiseBinaryLayer;
+};
+
+} //namespace armnn
\ No newline at end of file
diff --git a/src/backends/neon/workloads/NeonWorkloads.hpp b/src/backends/neon/workloads/NeonWorkloads.hpp
index 01fd2f7dba..024748690c 100644
--- a/src/backends/neon/workloads/NeonWorkloads.hpp
+++ b/src/backends/neon/workloads/NeonWorkloads.hpp
@@ -1,5 +1,5 @@
 //
-// Copyright © 2017 Arm Ltd and Contributors. All rights reserved.
+// Copyright © 2017, 2023 Arm Ltd and Contributors. All rights reserved.
 // SPDX-License-Identifier: MIT
 //
 
@@ -25,6 +25,7 @@
 #include "NeonDequantizeWorkload.hpp"
 #include "NeonDetectionPostProcessWorkload.hpp"
 #include "NeonDivisionWorkload.hpp"
+#include "NeonElementwiseBinaryWorkload.hpp"
 #include "NeonExpWorkload.hpp"
 #include "NeonFillWorkload.hpp"
 #include "NeonFloorFloatWorkload.hpp"