IVGCVSW-2325 Reduce duplication in LayerTests by reusing the ElementwiseTestHelper

* Refactored tests for element-wise operations to use ElementwiseTestHelper
* Moved the etasts for each operation in a separate file under backendsCommon/test/layerTests

Signed-off-by: Aron Virginas-Tar <Aron.Virginas-Tar@arm.com>
Change-Id: Icefb6b35df78b9619f69378229789505bf92670e

22 files changed, 3248 insertions, 2959 deletions

diff --git a/src/backends/backendsCommon/common.mk b/src/backends/backendsCommon/common.mk
index d7fe9b159f..9f097431e0 100644
--- a/src/backends/backendsCommon/common.mk
+++ b/src/backends/backendsCommon/common.mk
@@ -33,7 +33,15 @@ COMMON_TEST_SOURCES := \
     test/CommonTestUtils.cpp \
     test/JsonPrinterTestImpl.cpp \
     test/LayerTests.cpp \
-    test/TensorCopyUtils.cpp
+    test/TensorCopyUtils.cpp \
+    test/layerTests/AdditionTestImpl.cpp \
+    test/layerTests/DivisionTestImpl.cpp \
+    test/layerTests/EqualTestImpl.cpp \
+    test/layerTests/GreaterTestImpl.cpp \
+    test/layerTests/MaximumTestImpl.cpp \
+    test/layerTests/MinimumTestImpl.cpp \
+    test/layerTests/MultiplicationTestImpl.cpp \
+    test/layerTests/SubtractionTestImpl.cpp
 
 ifeq ($(ARMNN_COMPUTE_REF_ENABLED),1)
 COMMON_TEST_SOURCES += \
diff --git a/src/backends/backendsCommon/test/CMakeLists.txt b/src/backends/backendsCommon/test/CMakeLists.txt
index 73574732f3..7dca047514 100644
--- a/src/backends/backendsCommon/test/CMakeLists.txt
+++ b/src/backends/backendsCommon/test/CMakeLists.txt
@@ -57,6 +57,24 @@ list(APPEND armnnBackendsCommonUnitTests_sources
     TensorCopyUtils.hpp
     WorkloadFactoryHelper.hpp
     WorkloadTestUtils.hpp
+    layerTests/AdditionTestImpl.cpp
+    layerTests/AdditionTestImpl.hpp
+    layerTests/DivisionTestImpl.cpp
+    layerTests/DivisionTestImpl.hpp
+    layerTests/ElementwiseTestImpl.hpp
+    layerTests/EqualTestImpl.cpp
+    layerTests/EqualTestImpl.hpp
+    layerTests/GreaterTestImpl.cpp
+    layerTests/GreaterTestImpl.hpp
+    layerTests/LayerTestResult.hpp
+    layerTests/MaximumTestImpl.cpp
+    layerTests/MaximumTestImpl.hpp
+    layerTests/MinimumTestImpl.cpp
+    layerTests/MinimumTestImpl.hpp
+    layerTests/MultiplicationTestImpl.cpp
+    layerTests/MultiplicationTestImpl.hpp
+    layerTests/SubtractionTestImpl.cpp
+    layerTests/SubtractionTestImpl.hpp
 )
 
 if (ARMNNREF)
diff --git a/src/backends/backendsCommon/test/LayerTests.cpp b/src/backends/backendsCommon/test/LayerTests.cpp
index 5fd8f3e641..2d71e60ca4 100644
--- a/src/backends/backendsCommon/test/LayerTests.cpp
+++ b/src/backends/backendsCommon/test/LayerTests.cpp
@@ -2789,1927 +2789,6 @@ LayerTestResult<float,3> ConcatTest(
     return ret;
 }
 
-LayerTestResult<float,4> AdditionTest(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    unsigned int batchSize = 2;
-    unsigned int channels  = 2;
-    unsigned int height    = 2;
-    unsigned int width     = 3;
-
-    armnn::TensorInfo inputTensorInfo1, inputTensorInfo2;
-    armnn::TensorInfo outputTensorInfo;
-
-    unsigned int shape[] = {batchSize, channels, height, width};
-
-    inputTensorInfo1 = armnn::TensorInfo(4, shape, armnn::DataType::Float32);
-    inputTensorInfo2 = armnn::TensorInfo(4, shape, armnn::DataType::Float32);
-    outputTensorInfo = armnn::TensorInfo(4, shape, armnn::DataType::Float32);
-
-
-    auto input1 = MakeTensor<float, 4>(inputTensorInfo1, std::vector<float>(
-        {
-            0.0f, 2.0f, 1.0f,
-            0.2f, 1.0f, 2.0f,
-
-            1.0f, 2.0f, 1.0f,
-            0.2f, 1.0f, 2.0f,
-
-            0.0f, 2.0f, 1.0f,
-            4.2f, 1.0f, 2.0f,
-
-            0.0f, 0.0f, 1.0f,
-            0.2f, 1.0f, 2.0f,
-        }));
-
-    auto input2 = MakeTensor<float, 4>(inputTensorInfo2, std::vector<float>(
-        {
-            1.0f, 2.0f, 1.0f,
-            0.0f, 1.0f, 2.0f,
-
-            1.0f, 2.0f, -2.0f,
-            0.2f, 1.0f, 2.0f,
-
-            0.0f, 2.0f, 1.0f,
-            4.2f, 0.0f, -3.0f,
-
-            0.0f, 0.0f, 1.0f,
-            0.7f, 1.0f, 5.0f,
-        }));
-
-    LayerTestResult<float,4> ret(outputTensorInfo);
-    ret.outputExpected = MakeTensor<float, 4>(outputTensorInfo, std::vector<float>(
-        {
-            1.0f, 4.0f, 2.0f,
-            0.2f, 2.0f, 4.0f,
-
-            2.0f, 4.0f, -1.0f,
-            0.4f, 2.0f, 4.0f,
-
-            0.0f, 4.0f, 2.0f,
-            8.4f, 1.0f, -1.0f,
-
-            0.0f, 0.0f, 2.0f,
-            0.9f, 2.0f, 7.0f,
-        }));
-
-    std::unique_ptr<armnn::ITensorHandle> inputHandle1 = workloadFactory.CreateTensorHandle(inputTensorInfo1);
-    std::unique_ptr<armnn::ITensorHandle> inputHandle2 = workloadFactory.CreateTensorHandle(inputTensorInfo2);
-    std::unique_ptr<armnn::ITensorHandle> outputHandle = workloadFactory.CreateTensorHandle(outputTensorInfo);
-
-    armnn::AdditionQueueDescriptor data;
-    armnn::WorkloadInfo info;
-    AddInputToWorkload(data, info, inputTensorInfo1, inputHandle1.get());
-    AddInputToWorkload(data, info, inputTensorInfo2, inputHandle2.get());
-    AddOutputToWorkload(data, info, outputTensorInfo, outputHandle.get());
-
-    std::unique_ptr<armnn::IWorkload> workload = workloadFactory.CreateAddition(data, info);
-
-    inputHandle1->Allocate();
-    inputHandle2->Allocate();
-    outputHandle->Allocate();
-
-    CopyDataToITensorHandle(inputHandle1.get(), &input1[0][0][0][0]);
-    CopyDataToITensorHandle(inputHandle2.get(), &input2[0][0][0][0]);
-
-    workload->PostAllocationConfigure();
-    workload->Execute();
-
-    CopyDataFromITensorHandle(&ret.output[0][0][0][0], outputHandle.get());
-
-    return ret;
-}
-
-LayerTestResult<float, 5> Addition5dTest(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    unsigned int depth     = 2;
-    unsigned int batchSize = 2;
-    unsigned int channels  = 2;
-    unsigned int height    = 2;
-    unsigned int width     = 3;
-
-    armnn::TensorInfo inputTensorInfo1, inputTensorInfo2;
-    armnn::TensorInfo outputTensorInfo;
-
-    unsigned int shape[] = {depth, batchSize, channels, height, width};
-
-    inputTensorInfo1 = armnn::TensorInfo(5, shape, armnn::DataType::Float32);
-    inputTensorInfo2 = armnn::TensorInfo(5, shape, armnn::DataType::Float32);
-    outputTensorInfo = armnn::TensorInfo(5, shape, armnn::DataType::Float32);
-
-
-    auto input1 = MakeTensor<float, 5>(inputTensorInfo1, std::vector<float>(
-        {
-            2.6f, 4.0f, 4.4f,  2.7f, 4.6f, 2.8f,
-            2.3f, 1.9f, 3.4f,  2.9f, 2.2f, 4.5f,
-
-            2.8f, 1.9f, 2.3f,  2.6f, 4.7f, 3.5f,
-            0.4f, 1.5f, 2.1f,  0.7f, 5.0f, 1.1f,
-
-
-            1.0f, 2.7f, 0.0f,  0.6f, 0.8f, 0.9f,
-            1.0f, 2.6f, 0.4f,  3.8f, 0.4f, 0.8f,
-
-            0.5f, 4.3f, 3.1f,  4.4f, 0.7f, 1.4f,
-            0.4f, 4.4f, 0.7f,  0.6f, 4.7f, 1.2f,
-
-        }));
-
-    auto input2 = MakeTensor<float, 5>(inputTensorInfo2, std::vector<float>(
-        {
-            4.4f, 3.0f, 1.0f,  0.0f, 3.9f, 3.1f,
-            1.7f, 2.9f, 1.3f,  0.4f, 0.4f, 4.3f,
-
-            4.5f, 0.2f, 2.2f,  4.1f, 3.9f, 3.0f,
-            0.1f, 2.5f, 4.1f,  4.6f, 1.5f, 0.0f,
-
-
-            0.5f, 4.9f, 2.5f,  1.5f, 3.4f, 4.5f,
-            2.0f, 3.0f, 4.9f,  1.6f, 2.4f, 3.4f,
-
-            3.6f, 1.8f, 1.3f,  2.6f, 2.1f, 4.8f,
-            2.0f, 4.3f, 4.0f,  0.2f, 0.6f, 4.4f,
-        }));
-
-    LayerTestResult<float, 5> ret(outputTensorInfo);
-    ret.outputExpected = MakeTensor<float, 5>(outputTensorInfo, std::vector<float>(
-        {
-            7.0f, 7.0f, 5.4f,  2.7f, 8.5f, 5.9f,
-            4.0f, 4.8f, 4.7f,  3.3f, 2.6f, 8.8f,
-
-            7.3f, 2.1f, 4.5f,  6.7f, 8.6f, 6.5f,
-            0.5f, 4.0f, 6.2f,  5.3f, 6.5f, 1.1f,
-
-
-            1.5f, 7.6f, 2.5f,  2.1f, 4.2f, 5.4f,
-            3.0f, 5.6f, 5.3f,  5.4f, 2.8f, 4.2f,
-
-            4.1f, 6.1f, 4.4f,  7.0f, 2.8f, 6.2f,
-            2.4f, 8.7f, 4.7f,  0.8f, 5.3f, 5.6f,
-        }));
-
-    std::unique_ptr<armnn::ITensorHandle> inputHandle1 = workloadFactory.CreateTensorHandle(inputTensorInfo1);
-    std::unique_ptr<armnn::ITensorHandle> inputHandle2 = workloadFactory.CreateTensorHandle(inputTensorInfo2);
-    std::unique_ptr<armnn::ITensorHandle> outputHandle = workloadFactory.CreateTensorHandle(outputTensorInfo);
-
-    armnn::AdditionQueueDescriptor data;
-    armnn::WorkloadInfo info;
-    AddInputToWorkload(data, info, inputTensorInfo1, inputHandle1.get());
-    AddInputToWorkload(data, info, inputTensorInfo2, inputHandle2.get());
-    AddOutputToWorkload(data, info, outputTensorInfo, outputHandle.get());
-
-    std::unique_ptr<armnn::IWorkload> workload = workloadFactory.CreateAddition(data, info);
-
-    inputHandle1->Allocate();
-    inputHandle2->Allocate();
-    outputHandle->Allocate();
-
-    CopyDataToITensorHandle(inputHandle1.get(), &input1[0][0][0][0][0]);
-    CopyDataToITensorHandle(inputHandle2.get(), &input2[0][0][0][0][0]);
-
-    workload->PostAllocationConfigure();
-    workload->Execute();
-
-    CopyDataFromITensorHandle(&ret.output[0][0][0][0][0], outputHandle.get());
-
-    return ret;
-}
-
-template<armnn::DataType ArmnnType, typename T = armnn::ResolveType<ArmnnType>>
-LayerTestResult<T, 4> AdditionBroadcastTestImpl(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
-    float qScale,
-    int32_t qOffset)
-{
-    armnn::TensorInfo inputTensorInfo1 = armnn::TensorInfo({1, 3, 2, 1}, ArmnnType);
-    armnn::TensorInfo inputTensorInfo2 = armnn::TensorInfo({1, 1, 2, 3}, ArmnnType);
-    armnn::TensorInfo outputTensorInfo = armnn::TensorInfo({1, 3, 2, 3}, ArmnnType);
-
-    if (armnn::IsQuantizedType<T>())
-    {
-        inputTensorInfo1.SetQuantizationScale(qScale);
-        inputTensorInfo1.SetQuantizationOffset(qOffset);
-        inputTensorInfo2.SetQuantizationScale(qScale);
-        inputTensorInfo2.SetQuantizationOffset(qOffset);
-        outputTensorInfo.SetQuantizationScale(qScale);
-        outputTensorInfo.SetQuantizationOffset(qOffset);
-    }
-
-    auto input1 = MakeTensor<T, 4>(inputTensorInfo1, QuantizedVector<T>(qScale, qOffset,
-        {
-            0.0f,
-            1.0f,
-
-            2.0f,
-            3.0f,
-
-            4.0f,
-            5.0f,
-        }));
-
-    auto input2 = MakeTensor<T, 4>(inputTensorInfo2, QuantizedVector<T>(qScale, qOffset,
-        {
-            0.5f, 1.5f, 2.5f,
-            3.5f, 4.5f, 5.5f,
-        }));
-
-    LayerTestResult<T,4> ret(outputTensorInfo);
-    ret.outputExpected = MakeTensor<T, 4>(outputTensorInfo, QuantizedVector<T>(qScale, qOffset,
-        {
-            0.5f, 1.5f, 2.5f,
-            4.5f, 5.5f, 6.5f,
-
-            2.5f, 3.5f, 4.5f,
-            6.5f, 7.5f, 8.5f,
-
-            4.5f, 5.5f, 6.5f,
-            8.5f, 9.5f, 10.5f,
-        }));
-
-    std::unique_ptr<armnn::ITensorHandle> inputHandle1 = workloadFactory.CreateTensorHandle(inputTensorInfo1);
-    std::unique_ptr<armnn::ITensorHandle> inputHandle2 = workloadFactory.CreateTensorHandle(inputTensorInfo2);
-    std::unique_ptr<armnn::ITensorHandle> outputHandle = workloadFactory.CreateTensorHandle(outputTensorInfo);
-
-    armnn::AdditionQueueDescriptor data;
-    armnn::WorkloadInfo info;
-    AddInputToWorkload(data, info, inputTensorInfo1, inputHandle1.get());
-    AddInputToWorkload(data, info, inputTensorInfo2, inputHandle2.get());
-    AddOutputToWorkload(data, info, outputTensorInfo, outputHandle.get());
-
-    std::unique_ptr<armnn::IWorkload> workload = workloadFactory.CreateAddition(data, info);
-
-    inputHandle1->Allocate();
-    inputHandle2->Allocate();
-    outputHandle->Allocate();
-
-    CopyDataToITensorHandle(inputHandle1.get(), &input1[0][0][0][0]);
-    CopyDataToITensorHandle(inputHandle2.get(), &input2[0][0][0][0]);
-
-    workload->PostAllocationConfigure();
-    workload->Execute();
-
-    CopyDataFromITensorHandle(&ret.output[0][0][0][0], outputHandle.get());
-
-    return ret;
-}
-
-template<armnn::DataType ArmnnType, typename T = armnn::ResolveType<ArmnnType>>
-LayerTestResult<T, 4> AdditionBroadcast1ElementTestImpl(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
-    float qScale,
-    int32_t qOffset)
-{
-    armnn::TensorInfo inputTensorInfo1 = armnn::TensorInfo({1, 3, 2, 3}, ArmnnType);
-    armnn::TensorInfo inputTensorInfo2 = armnn::TensorInfo({1, 1, 1, 1}, ArmnnType);
-    armnn::TensorInfo outputTensorInfo = armnn::TensorInfo({1, 3, 2, 3}, ArmnnType);
-
-    if (armnn::IsQuantizedType<T>())
-    {
-        inputTensorInfo1.SetQuantizationScale(qScale);
-        inputTensorInfo1.SetQuantizationOffset(qOffset);
-        inputTensorInfo2.SetQuantizationScale(qScale);
-        inputTensorInfo2.SetQuantizationOffset(qOffset);
-        outputTensorInfo.SetQuantizationScale(qScale);
-        outputTensorInfo.SetQuantizationOffset(qOffset);
-    }
-
-    auto input1 = MakeTensor<T, 4>(inputTensorInfo1, QuantizedVector<T>(qScale, qOffset,
-        {
-             0.0f,  1.0f,  2.0f,
-             3.0f,  4.0f,  5.0f,
-             6.0f,  7.0f,  8.0f,
-             9.0f, 10.0f, 11.0f,
-            12.0f, 13.0f, 14.0f,
-            15.0f, 16.0f, 17.0f,
-        }));
-
-    auto input2 = MakeTensor<T, 4>(inputTensorInfo2, QuantizedVector<T>(qScale, qOffset,
-        {
-            0.5f,
-        }));
-
-    LayerTestResult<T,4> ret(outputTensorInfo);
-    ret.outputExpected = MakeTensor<T, 4>(outputTensorInfo, QuantizedVector<T>(qScale, qOffset,
-        {
-             0.5f,  1.5f,  2.5f,
-             3.5f,  4.5f,  5.5f,
-             6.5f,  7.5f,  8.5f,
-             9.5f, 10.5f, 11.5f,
-            12.5f, 13.5f, 14.5f,
-            15.5f, 16.5f, 17.5f,
-        }));
-
-    std::unique_ptr<armnn::ITensorHandle> inputHandle1 = workloadFactory.CreateTensorHandle(inputTensorInfo1);
-    std::unique_ptr<armnn::ITensorHandle> inputHandle2 = workloadFactory.CreateTensorHandle(inputTensorInfo2);
-    std::unique_ptr<armnn::ITensorHandle> outputHandle = workloadFactory.CreateTensorHandle(outputTensorInfo);
-
-    armnn::AdditionQueueDescriptor data;
-    armnn::WorkloadInfo info;
-    AddInputToWorkload(data, info, inputTensorInfo1, inputHandle1.get());
-    AddInputToWorkload(data, info, inputTensorInfo2, inputHandle2.get());
-    AddOutputToWorkload(data, info, outputTensorInfo, outputHandle.get());
-
-    std::unique_ptr<armnn::IWorkload> workload = workloadFactory.CreateAddition(data, info);
-
-    inputHandle1->Allocate();
-    inputHandle2->Allocate();
-    outputHandle->Allocate();
-
-    CopyDataToITensorHandle(inputHandle1.get(), &input1[0][0][0][0]);
-    CopyDataToITensorHandle(inputHandle2.get(), &input2[0][0][0][0]);
-
-    workload->PostAllocationConfigure();
-    workload->Execute();
-
-    CopyDataFromITensorHandle(&ret.output[0][0][0][0], outputHandle.get());
-
-    return ret;
-}
-
-LayerTestResult<float, 4> AdditionBroadcastTest(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    return AdditionBroadcastTestImpl<armnn::DataType::Float32>(
-        workloadFactory, memoryManager, 0.0f, 0);
-}
-
-LayerTestResult<uint8_t, 4> AdditionBroadcastUint8Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    return AdditionBroadcastTestImpl<armnn::DataType::QuantisedAsymm8>(
-        workloadFactory, memoryManager, 2.f, 0);
-}
-
-LayerTestResult<int16_t, 4> AdditionBroadcastInt16Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    return AdditionBroadcastTestImpl<armnn::DataType::QuantisedSymm16>(
-        workloadFactory, memoryManager, 2.f, 0);
-}
-
-LayerTestResult<float, 4> AdditionBroadcast1ElementTest(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    return AdditionBroadcast1ElementTestImpl<armnn::DataType::Float32>(
-        workloadFactory, memoryManager, 0.0f, 0);
-}
-
-LayerTestResult<uint8_t, 4> AdditionBroadcast1ElementUint8Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    return AdditionBroadcast1ElementTestImpl<armnn::DataType::QuantisedAsymm8>(
-        workloadFactory, memoryManager, 0.1333333f, 128);
-}
-
-LayerTestResult<int16_t, 4> AdditionBroadcast1ElementInt16Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    return AdditionBroadcast1ElementTestImpl<armnn::DataType::QuantisedSymm16>(
-        workloadFactory, memoryManager, 0.1333333f, 0);
-}
-
-LayerTestResult<float,4> CompareAdditionTest(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
-    armnn::IWorkloadFactory& refWorkloadFactory)
-{
-    unsigned int batchSize = 4;
-    unsigned int channels  = 1;
-    unsigned int height    = 2;
-    unsigned int width     = 3;
-
-    armnn::TensorInfo inputTensorInfo1, inputTensorInfo2;
-    armnn::TensorInfo outputTensorInfo;
-
-    unsigned int shape[] = {batchSize, channels, height, width};
-
-    inputTensorInfo1 = armnn::TensorInfo(4, shape, armnn::DataType::Float32);
-    inputTensorInfo2 = armnn::TensorInfo(4, shape, armnn::DataType::Float32);
-    outputTensorInfo = armnn::TensorInfo(4, shape, armnn::DataType::Float32);
-
-    auto input1 = MakeRandomTensor<float, 4>(inputTensorInfo1, 1232);
-    auto input2 = MakeRandomTensor<float, 4>(inputTensorInfo2, 456);
-
-    LayerTestResult<float,4> ret(outputTensorInfo);
-
-    std::unique_ptr<armnn::ITensorHandle> inputHandle1 = workloadFactory.CreateTensorHandle(inputTensorInfo1);
-    std::unique_ptr<armnn::ITensorHandle> inputHandle2 = workloadFactory.CreateTensorHandle(inputTensorInfo2);
-    std::unique_ptr<armnn::ITensorHandle> outputHandle = workloadFactory.CreateTensorHandle(outputTensorInfo);
-
-    std::unique_ptr<armnn::ITensorHandle> inputHandle1Ref = refWorkloadFactory.CreateTensorHandle(inputTensorInfo1);
-    std::unique_ptr<armnn::ITensorHandle> inputHandle2Ref = refWorkloadFactory.CreateTensorHandle(inputTensorInfo2);
-    std::unique_ptr<armnn::ITensorHandle> outputHandleRef = refWorkloadFactory.CreateTensorHandle(outputTensorInfo);
-
-    armnn::AdditionQueueDescriptor data;
-    armnn::WorkloadInfo info;
-    AddInputToWorkload(data, info, inputTensorInfo1, inputHandle1.get());
-    AddInputToWorkload(data, info, inputTensorInfo2, inputHandle2.get());
-    AddOutputToWorkload(data, info, outputTensorInfo, outputHandle.get());
-
-    armnn::AdditionQueueDescriptor refData = data;
-    armnn::WorkloadInfo refInfo = info;
-    SetWorkloadInput(refData, refInfo, 0, inputTensorInfo1, inputHandle1Ref.get());
-    SetWorkloadInput(refData, refInfo, 1, inputTensorInfo2, inputHandle2Ref.get());
-    SetWorkloadOutput(refData, refInfo, 0, outputTensorInfo, outputHandleRef.get());
-
-    std::unique_ptr<armnn::IWorkload> workload = workloadFactory.CreateAddition(data, info);
-    std::unique_ptr<armnn::IWorkload> workloadRef = refWorkloadFactory.CreateAddition(refData, refInfo);
-
-    inputHandle1->Allocate();
-    inputHandle2->Allocate();
-    outputHandle->Allocate();
-    inputHandle1Ref->Allocate();
-    inputHandle2Ref->Allocate();
-    outputHandleRef->Allocate();
-
-    CopyDataToITensorHandle(inputHandle1.get(), &input1[0][0][0][0]);
-    CopyDataToITensorHandle(inputHandle2.get(), &input2[0][0][0][0]);
-    CopyDataToITensorHandle(inputHandle1Ref.get(), &input1[0][0][0][0]);
-    CopyDataToITensorHandle(inputHandle2Ref.get(), &input2[0][0][0][0]);
-
-    workload->PostAllocationConfigure();
-    workload->Execute();
-    workloadRef->PostAllocationConfigure();
-    workloadRef->Execute();
-
-    CopyDataFromITensorHandle(&ret.output[0][0][0][0], outputHandle.get());
-    CopyDataFromITensorHandle(&ret.outputExpected[0][0][0][0], outputHandleRef.get());
-
-    return ret;
-}
-
-namespace {
-template<armnn::DataType ArmnnType, typename T = armnn::ResolveType<ArmnnType>>
-LayerTestResult<T, 4> DivisionTestHelper(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
-    const unsigned int shape0[4],
-    const std::vector<T>& values0,
-    float scale0,
-    int32_t offset0,
-    const unsigned int shape1[4],
-    const std::vector<T> & values1,
-    float scale1,
-    int32_t offset1,
-    const unsigned int outShape[4],
-    const std::vector<T> & outValues,
-    float outScale,
-    int32_t outOffset)
-{
-    armnn::TensorInfo inputTensorInfo0(4, shape0, ArmnnType);
-    armnn::TensorInfo inputTensorInfo1(4, shape1, ArmnnType);
-    armnn::TensorInfo outputTensorInfo(4, outShape, ArmnnType);
-
-    inputTensorInfo0.SetQuantizationScale(scale0);
-    inputTensorInfo0.SetQuantizationOffset(offset0);
-
-    inputTensorInfo1.SetQuantizationScale(scale1);
-    inputTensorInfo1.SetQuantizationOffset(offset1);
-
-    outputTensorInfo.SetQuantizationScale(outScale);
-    outputTensorInfo.SetQuantizationOffset(outOffset);
-
-    auto input0 = MakeTensor<T, 4>(inputTensorInfo0, values0);
-    auto input1 = MakeTensor<T, 4>(inputTensorInfo1, values1);
-
-    LayerTestResult<T, 4> result(outputTensorInfo);
-    result.outputExpected = MakeTensor<T, 4>(outputTensorInfo, outValues);
-
-    std::unique_ptr<armnn::ITensorHandle> inputHandle0 = workloadFactory.CreateTensorHandle(inputTensorInfo0);
-    std::unique_ptr<armnn::ITensorHandle> inputHandle1 = workloadFactory.CreateTensorHandle(inputTensorInfo1);
-    std::unique_ptr<armnn::ITensorHandle> outputHandle = workloadFactory.CreateTensorHandle(outputTensorInfo);
-
-    armnn::DivisionQueueDescriptor data;
-    armnn::WorkloadInfo info;
-    AddInputToWorkload(data,  info, inputTensorInfo0, inputHandle0.get());
-    AddInputToWorkload(data,  info, inputTensorInfo1, inputHandle1.get());
-    AddOutputToWorkload(data, info, outputTensorInfo, outputHandle.get());
-
-    std::unique_ptr<armnn::IWorkload> workload = workloadFactory.CreateDivision(data, info);
-
-    inputHandle0->Allocate();
-    inputHandle1->Allocate();
-    outputHandle->Allocate();
-
-    CopyDataToITensorHandle(inputHandle0.get(), &input0[0][0][0][0]);
-    CopyDataToITensorHandle(inputHandle1.get(), &input1[0][0][0][0]);
-
-    workload->PostAllocationConfigure();
-    workload->Execute();
-
-    CopyDataFromITensorHandle(&result.output[0][0][0][0], outputHandle.get());
-
-    return result;
-}
-} // anonymous namespace
-
-LayerTestResult<float,4> DivisionByZeroTest(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    const unsigned int width = 2;
-    const unsigned int height = 2;
-    const unsigned int channelCount = 2;
-    const unsigned int batchSize = 2;
-
-    unsigned int shape[] = { batchSize, channelCount, height, width };
-
-    std::vector<float> input0({
-                                1.f,  1.f,  1.f,  1.f,  0.f, 0.f, 0.f, 0.f,
-                               -1.f, -1.f, -1.f, -1.f,  5.f, 5.f, 5.f, 5.f });
-
-    std::vector<float> input1({
-                               0.f, 0.f, -0.f, -0.f,  0.f, 0.f, -0.f, -0.f,
-                               0.f, 0.f, -0.f, -0.f,  5.f, 5.f,  5.f,  5.f });
-
-    std::vector<float> output({
-                               INFINITY, INFINITY, -INFINITY, -INFINITY,  NAN, NAN, -NAN, -NAN,
-                               -INFINITY, -INFINITY, INFINITY, INFINITY,  1, 1, 1, 1 });
-
-    return DivisionTestHelper<armnn::DataType::Float32>(workloadFactory,
-                                                        memoryManager,
-                                                        shape, input0, 1.0f, 0,
-                                                        shape, input1, 1.0f, 0,
-                                                        shape, output, 1.0f, 0);
-}
-
-LayerTestResult<float,4> DivisionTest(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    const unsigned int width = 2;
-    const unsigned int height = 2;
-    const unsigned int channelCount = 2;
-    const unsigned int batchSize = 2;
-
-    unsigned int shape[] = { batchSize, channelCount, height, width };
-
-    std::vector<float> input0({
-                                      2,  2,  2,  2,    3,  3,  3,  3,
-                                      4,  4,  4,  4,    5,  5,  5,  5 });
-
-    std::vector<float> input1({
-                                      1,  1,  1,  1,    2,  2,  2,  2,
-                                      4,  4,  4,  4,    4,  4,  4,  4 });
-
-    std::vector<float> output({
-                                      2,  2,  2,  2,    1.5,  1.5,  1.5,  1.5,
-                                      1, 1, 1, 1,  1.25, 1.25, 1.25, 1.25 });
-
-
-    return DivisionTestHelper<armnn::DataType::Float32>(workloadFactory,
-                                                        memoryManager,
-                                                        shape, input0, 1.0f, 0,
-                                                        shape, input1, 1.0f, 0,
-                                                        shape, output, 1.0f, 0);
-}
-
-LayerTestResult<float, 4> DivisionBroadcast1ElementTest(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    unsigned int shape0[] = { 1, 2, 2, 2 };
-    std::vector<float> input0({ 2, 4, 6, 8, 10, 12, 14, 16});
-
-    unsigned int shape1[] = { 1, 1, 1, 1 };
-    std::vector<float> input1({ 2 });
-
-    std::vector<float> output({ 1, 2, 3, 4, 5, 6, 7, 8});
-
-
-    return DivisionTestHelper<armnn::DataType::Float32>(workloadFactory,
-                                                        memoryManager,
-                                                        shape0, input0, 1.0f, 0,
-                                                        shape1, input1, 1.0f, 0,
-                                                        shape0, output, 1.0f, 0);
-}
-
-LayerTestResult<float, 4> DivisionBroadcast1DVectorTest(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    unsigned int shape0[] = { 1, 3, 3, 2 };
-    std::vector<float> input0({
-                                      1,   4,       3,  8,      5, 12,
-                                      7,   16,      9, 20,     11, 24,
-                                      13,  28,     15, 32,     17, 36});
-
-    unsigned int shape1[] = { 1, 1, 1, 2 };
-    std::vector<float> input1({ 1, 2 });
-
-    std::vector<float> output({
-                                      1,   2,      3,  4,      5,  6,
-                                      7,   8,      9, 10,     11, 12,
-                                      13, 14,     15, 16,     17, 18});
-
-    return DivisionTestHelper<armnn::DataType::Float32>(workloadFactory,
-                                                        memoryManager,
-                                                        shape0, input0, 1.0f, 0,
-                                                        shape1, input1, 1.0f, 0,
-                                                        shape0, output, 1.0f, 0);
-}
-
-LayerTestResult<uint8_t,4> DivisionUint8Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    const unsigned int width = 2;
-    const unsigned int height = 2;
-    const unsigned int channelCount = 2;
-    const unsigned int batchSize = 2;
-
-    unsigned int shape[] = { batchSize, channelCount, height, width };
-
-    std::vector<uint8_t> input0({2,  2,  2,  2,    3,  3,  3,  3,
-                                 4,  4,  4,  4,    5,  5,  5,  5 });
-
-    std::vector<uint8_t> input1({1,  1,  1,  1,    2,  2,  2,  2,
-                                 4,  4,  4,  4,    4,  4,  4,  4 });
-
-    std::vector<uint8_t> output({8,  8,  8,  8,    6,  6,  6,  6,
-                                 4,  4,  4,  4,    5,  5,  5,  5});
-
-
-    return DivisionTestHelper<armnn::DataType::QuantisedAsymm8>(workloadFactory,
-                                                                memoryManager,
-                                                                shape, input0, 1.0f,  0,
-                                                                shape, input1, 1.0f,  0,
-                                                                shape, output, 0.25f, 0);
-}
-
-LayerTestResult<uint8_t, 4> DivisionBroadcast1ElementUint8Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    unsigned int shape0[] = { 1, 2, 2, 2 };
-    std::vector<uint8_t> input0({ 2, 4, 6, 8, 10, 12, 14, 16});
-
-    unsigned int shape1[] = { 1, 1, 1, 1 };
-    std::vector<uint8_t> input1({ 2 });
-
-    std::vector<uint8_t> output({ 1, 2, 3, 4, 5, 6, 7, 8});
-
-    return DivisionTestHelper<armnn::DataType::QuantisedAsymm8>(workloadFactory,
-                                                                memoryManager,
-                                                                shape0, input0, 1.0f, 0,
-                                                                shape1, input1, 1.0f, 0,
-                                                                shape0, output, 1.0f, 0);
-}
-
-LayerTestResult<uint8_t, 4> DivisionBroadcast1DVectorUint8Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    unsigned int shape0[] = { 1, 3, 3, 2 };
-    std::vector<uint8_t> input0({1,   4,     3,  8,      5,  12,
-                                 7,   16,    9,  20,     11, 24,
-                                 13,  28,    15, 32,     17, 36});
-
-    unsigned int shape1[] = { 1, 1, 1, 2 };
-    std::vector<uint8_t> input1({ 1, 2 });
-
-    std::vector<uint8_t> output({1,   2,      3,  4,      5,  6,
-                                 7,   8,      9, 10,     11, 12,
-                                 13, 14,     15, 16,     17, 18});
-
-    return DivisionTestHelper<armnn::DataType::QuantisedAsymm8>(workloadFactory,
-                                                                memoryManager,
-                                                                shape0, input0, 1.0f, 0,
-                                                                shape1, input1, 1.0f, 0,
-                                                                shape0, output, 1.0f, 0);
-}
-
-LayerTestResult<int16_t,4> DivisionInt16Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    unsigned int shape[] = { 2, 2, 2, 2 };
-
-    std::vector<int16_t> input0({2,  2,  2,  2,    3,  3,  3,  3,
-                                 4,  4,  4,  4,    5,  5,  5,  5 });
-
-    std::vector<int16_t> input1({1,  1,  1,  1,    2,  2,  2,  2,
-                                 4,  4,  4,  4,    4,  4,  4,  4 });
-
-    std::vector<int16_t> output({8,  8,  8,  8,    6,  6,  6,  6,
-                                 4,  4,  4,  4,    5,  5,  5,  5});
-
-
-    return DivisionTestHelper<armnn::DataType::QuantisedSymm16>(workloadFactory,
-                                                                memoryManager,
-                                                                shape, input0, 1.0f,  0,
-                                                                shape, input1, 1.0f,  0,
-                                                                shape, output, 0.25f, 0);
-}
-
-LayerTestResult<int16_t, 4> DivisionBroadcast1ElementInt16Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    unsigned int shape0[] = { 1, 2, 2, 2 };
-    std::vector<int16_t> input0({ 2, 4, 6, 8, 10, 12, 14, 16});
-
-    unsigned int shape1[] = { 1, 1, 1, 1 };
-    std::vector<int16_t> input1({ 2 });
-
-    std::vector<int16_t> output({ 1, 2, 3, 4, 5, 6, 7, 8});
-
-    return DivisionTestHelper<armnn::DataType::QuantisedSymm16>(workloadFactory,
-                                                                memoryManager,
-                                                                shape0, input0, 1.0f, 0,
-                                                                shape1, input1, 1.0f, 0,
-                                                                shape0, output, 1.0f, 0);
-}
-
-LayerTestResult<int16_t, 4> DivisionBroadcast1DVectorInt16Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    unsigned int shape0[] = { 1, 3, 3, 2 };
-    std::vector<int16_t> input0({1,   4,     3,  8,      5,  12,
-                                 7,   16,    9,  20,     11, 24,
-                                 13,  28,    15, 32,     17, 36});
-
-    unsigned int shape1[] = { 1, 1, 1, 2 };
-    std::vector<int16_t> input1({ 1, 2 });
-
-    std::vector<int16_t> output({1,   2,      3,  4,      5,  6,
-                                 7,   8,      9, 10,     11, 12,
-                                 13, 14,     15, 16,     17, 18});
-
-    return DivisionTestHelper<armnn::DataType::QuantisedSymm16>(workloadFactory,
-                                                                memoryManager,
-                                                                shape0, input0, 1.0f, 0,
-                                                                shape1, input1, 1.0f, 0,
-                                                                shape0, output, 1.0f, 0);
-}
-
-template<typename DescriptorType>
-std::unique_ptr<armnn::IWorkload> CreateWorkload(
-    const armnn::IWorkloadFactory& workloadFactory,
-    const armnn::WorkloadInfo& info,
-    const DescriptorType& descriptor)
-{
-    return CreateWorkload(workloadFactory, info, descriptor);
-};
-
-template<>
-std::unique_ptr<armnn::IWorkload> CreateWorkload<armnn::MaximumQueueDescriptor>(
-    const armnn::IWorkloadFactory& workloadFactory,
-    const armnn::WorkloadInfo& info,
-    const armnn::MaximumQueueDescriptor& descriptor)
-{
-    return workloadFactory.CreateMaximum(descriptor, info);
-}
-
-template<>
-std::unique_ptr<armnn::IWorkload> CreateWorkload<armnn::MinimumQueueDescriptor>(
-    const armnn::IWorkloadFactory& workloadFactory,
-    const armnn::WorkloadInfo& info,
-    const armnn::MinimumQueueDescriptor& descriptor)
-{
-    return workloadFactory.CreateMinimum(descriptor, info);
-}
-
-template<>
-std::unique_ptr<armnn::IWorkload> CreateWorkload<armnn::EqualQueueDescriptor>(
-        const armnn::IWorkloadFactory& workloadFactory,
-        const armnn::WorkloadInfo& info,
-        const armnn::EqualQueueDescriptor& descriptor)
-{
-    return workloadFactory.CreateEqual(descriptor, info);
-}
-
-template<>
-std::unique_ptr<armnn::IWorkload> CreateWorkload<armnn::GreaterQueueDescriptor>(
-        const armnn::IWorkloadFactory& workloadFactory,
-        const armnn::WorkloadInfo& info,
-        const armnn::GreaterQueueDescriptor& descriptor)
-{
-    return workloadFactory.CreateGreater(descriptor, info);
-}
-
-namespace {
-
-template <typename Descriptor,
-          armnn::DataType ArmnnTypeInput,
-          armnn::DataType ArmnnTypeOutput,
-          typename TInput = armnn::ResolveType<ArmnnTypeInput>,
-          typename TOutput = armnn::ResolveType<ArmnnTypeOutput>>
-LayerTestResult<TOutput, 4> ElementwiseTestHelper(
-    armnn::IWorkloadFactory & workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr & memoryManager,
-    const unsigned int shape0[4], std::vector<TInput> values0,
-    const unsigned int shape1[4], std::vector<TInput> values1,
-    const unsigned int outShape[4], std::vector<TOutput> outValues,
-    float qScale = 0.0f, int qOffset = 0)
-{
-    const uint32_t dimensionCount = 4;
-    armnn::TensorInfo inputTensorInfo0{dimensionCount, shape0, ArmnnTypeInput};
-    armnn::TensorInfo inputTensorInfo1{dimensionCount, shape1, ArmnnTypeInput};
-    armnn::TensorInfo outputTensorInfo{dimensionCount, outShape, ArmnnTypeOutput};
-
-    auto input0 = MakeTensor<TInput, 4>(inputTensorInfo0, values0);
-    auto input1 = MakeTensor<TInput, 4>(inputTensorInfo1, values1);
-
-    if (armnn::IsQuantizedType<TInput>())
-    {
-        inputTensorInfo0.SetQuantizationScale(qScale);
-        inputTensorInfo0.SetQuantizationOffset(qOffset);
-
-        inputTensorInfo1.SetQuantizationScale(qScale);
-        inputTensorInfo1.SetQuantizationOffset(qOffset);
-
-        outputTensorInfo.SetQuantizationScale(qScale);
-        outputTensorInfo.SetQuantizationOffset(qOffset);
-    }
-
-    LayerTestResult<TOutput,4> ret(outputTensorInfo);
-
-    if(ArmnnTypeOutput == armnn::DataType::Boolean)
-    {
-        ret.compareBoolean = true;
-    }
-
-    std::unique_ptr<armnn::ITensorHandle> inputHandle0 = workloadFactory.CreateTensorHandle(inputTensorInfo0);
-    std::unique_ptr<armnn::ITensorHandle> inputHandle1 = workloadFactory.CreateTensorHandle(inputTensorInfo1);
-    std::unique_ptr<armnn::ITensorHandle> outputHandle = workloadFactory.CreateTensorHandle(outputTensorInfo);
-
-    Descriptor data;
-    armnn::WorkloadInfo info;
-    AddInputToWorkload(data, info, inputTensorInfo0, inputHandle0.get());
-    AddInputToWorkload(data, info, inputTensorInfo1, inputHandle1.get());
-    AddOutputToWorkload(data, info, outputTensorInfo, outputHandle.get());
-    auto workload = CreateWorkload<Descriptor>(workloadFactory, info, data);
-
-    inputHandle0->Allocate();
-    inputHandle1->Allocate();
-    outputHandle->Allocate();
-
-    CopyDataToITensorHandle(inputHandle0.get(), &input0[0][0][0][0]);
-    CopyDataToITensorHandle(inputHandle1.get(), &input1[0][0][0][0]);
-
-    workload->PostAllocationConfigure();
-    ExecuteWorkload(*workload, memoryManager);
-
-    CopyDataFromITensorHandle(&ret.output[0][0][0][0], outputHandle.get());
-
-    ret.outputExpected = MakeTensor<TOutput, 4>(outputTensorInfo, outValues);
-    return ret;
-}
-
-template <typename Descriptor, armnn::DataType ArmnnT, typename T = armnn::ResolveType<ArmnnT>>
-LayerTestResult<T, 4> ElementwiseTestHelper(
-    armnn::IWorkloadFactory & workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr & memoryManager,
-    const unsigned int shape0[4], std::vector<T> values0,
-    const unsigned int shape1[4], std::vector<T> values1,
-    const unsigned int outShape[4], std::vector<T> outValues,
-    float qScale = 0.0f, int qOffset = 0)
-{
-    return ElementwiseTestHelper<Descriptor, ArmnnT, ArmnnT>
-        (workloadFactory,
-         memoryManager,
-         shape0,
-         values0,
-         shape1,
-         values1,
-         outShape,
-         outValues,
-         qScale,
-         qOffset);
-}
-}
-
-LayerTestResult<uint8_t, 4> EqualSimpleTest(armnn::IWorkloadFactory& workloadFactory,
-                                            const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    const unsigned int width = 2;
-    const unsigned int height = 2;
-    const unsigned int channelCount = 2;
-    const unsigned int batchSize = 2;
-
-    unsigned int shape[] = { batchSize, channelCount, height, width };
-
-    std::vector<float> input0({ 1, 1, 1, 1,  5, 5, 5, 5,
-                                3, 3, 3, 3,  4, 4, 4, 4 });
-
-    std::vector<float> input1({ 1, 1, 1, 1,  3, 3, 3, 3,
-                                5, 5, 5, 5,  4, 4, 4, 4 });
-
-    std::vector<uint8_t> output({ 1, 1, 1, 1,  0, 0, 0, 0,
-                                  0, 0, 0, 0,  1, 1, 1, 1 });
-
-    return ElementwiseTestHelper<armnn::EqualQueueDescriptor, armnn::DataType::Float32, armnn::DataType::Boolean>(
-        workloadFactory,
-        memoryManager,
-        shape,
-        input0,
-        shape,
-        input1,
-        shape,
-        output);
-}
-
-LayerTestResult<uint8_t, 4> EqualBroadcast1ElementTest(
-        armnn::IWorkloadFactory& workloadFactory,
-        const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    unsigned int shape0[] = { 1, 2, 2, 2 };
-    std::vector<float> input0({ 1, 2, 3, 4, 5, 6, 7, 8});
-
-    unsigned int shape1[] = { 1, 1, 1, 1 };
-    std::vector<float> input1({ 1 });
-
-    std::vector<uint8_t> output({ 1, 0, 0, 0, 0, 0, 0, 0});
-
-    return ElementwiseTestHelper<armnn::EqualQueueDescriptor, armnn::DataType::Float32, armnn::DataType::Boolean>(
-        workloadFactory,
-        memoryManager,
-        shape0,
-        input0,
-        shape1,
-        input1,
-        shape0,
-        output);
-}
-
-LayerTestResult<uint8_t, 4> EqualBroadcast1DVectorTest(
-        armnn::IWorkloadFactory& workloadFactory,
-        const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    const unsigned int shape0[] = { 1, 2, 2, 3 };
-    const unsigned int shape1[] = { 1, 1, 1, 3 };
-
-    std::vector<float> input0({ 1, 2, 3, 4, 5, 6,
-                                7, 8, 9, 10, 11, 12 });
-
-    std::vector<float> input1({ 1, 2, 3});
-
-    std::vector<uint8_t> output({ 1, 1, 1, 0, 0, 0,
-                                  0, 0, 0, 0, 0, 0 });
-
-    return ElementwiseTestHelper<armnn::EqualQueueDescriptor, armnn::DataType::Float32, armnn::DataType::Boolean>(
-        workloadFactory,
-        memoryManager,
-        shape0,
-        input0,
-        shape1,
-        input1,
-        shape0,
-        output);
-}
-
-LayerTestResult<uint8_t, 4> EqualUint8Test(
-        armnn::IWorkloadFactory& workloadFactory,
-        const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    unsigned int shape[] = { 2, 2, 2, 2 };
-
-    // See dequantized values to the right.
-    std::vector<uint8_t> input0({ 1, 1, 1, 1, 6, 6, 6, 6,
-                                  3, 3, 3, 3, 7, 7, 7, 7 });
-
-    std::vector<uint8_t> input1({ 2, 2, 2, 2, 6, 6, 6, 6,
-                                  3, 3, 3, 3, 5, 5, 5, 5 });
-
-    std::vector<uint8_t> output({ 0, 0, 0, 0, 1, 1, 1, 1,
-                                  1, 1, 1, 1, 0, 0, 0, 0 });
-
-    return ElementwiseTestHelper<armnn::EqualQueueDescriptor,
-                                 armnn::DataType::QuantisedAsymm8,
-                                 armnn::DataType::Boolean>(
-        workloadFactory,
-        memoryManager,
-        shape,
-        input0,
-        shape,
-        input1,
-        shape,
-        output,
-        1.0f,
-        0);
-}
-
-LayerTestResult<uint8_t, 4> EqualBroadcast1ElementUint8Test(
-        armnn::IWorkloadFactory& workloadFactory,
-        const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    const unsigned int shape0[] = { 1, 2, 2, 3 };
-    const unsigned int shape1[] = { 1, 1, 1, 1 };
-
-    std::vector<uint8_t> input0({ 1, 2, 3, 4, 5, 6,
-                                  7, 8, 9, 10, 11, 12 });
-
-    std::vector<uint8_t> input1({ 1 });
-
-    std::vector<uint8_t> output({ 1, 0, 0, 0, 0, 0,
-                                  0, 0, 0, 0, 0, 0 });
-
-    return ElementwiseTestHelper<armnn::EqualQueueDescriptor,
-                                 armnn::DataType::QuantisedAsymm8,
-                                 armnn::DataType::Boolean>(
-        workloadFactory,
-        memoryManager,
-        shape0,
-        input0,
-        shape1,
-        input1,
-        shape0,
-        output,
-        1.0f,
-        0);
-}
-
-LayerTestResult<uint8_t, 4> EqualBroadcast1DVectorUint8Test(
-        armnn::IWorkloadFactory& workloadFactory,
-        const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    const unsigned int shape0[] = { 1, 2, 2, 3 };
-    const unsigned int shape1[] = { 1, 1, 1, 3 };
-
-    std::vector<uint8_t> input0({ 1, 2, 3, 4, 5, 6,
-                                  7, 8, 9, 10, 11, 12 });
-
-    std::vector<uint8_t> input1({ 1, 1, 3});
-
-    std::vector<uint8_t> output({ 1, 0, 1, 0, 0, 0,
-                                  0, 0, 0, 0, 0, 0 });
-
-    return ElementwiseTestHelper<armnn::EqualQueueDescriptor,
-                                 armnn::DataType::QuantisedAsymm8,
-                                 armnn::DataType::Boolean>(
-        workloadFactory,
-        memoryManager,
-        shape0,
-        input0,
-        shape1,
-        input1,
-        shape0,
-        output,
-        1.0f,
-        0);
-}
-
-LayerTestResult<uint8_t, 4> GreaterSimpleTest(armnn::IWorkloadFactory& workloadFactory,
-                                            const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    const unsigned int width = 2;
-    const unsigned int height = 2;
-    const unsigned int channelCount = 2;
-    const unsigned int batchSize = 2;
-
-    unsigned int shape[] = { batchSize, channelCount, height, width };
-
-    std::vector<float> input0({ 1, 1, 1, 1,  5, 5, 5, 5,
-                                3, 3, 3, 3,  4, 4, 4, 4 });
-
-    std::vector<float> input1({ 1, 1, 1, 1,  3, 3, 3, 3,
-                                5, 5, 5, 5,  4, 4, 4, 4 });
-
-    std::vector<uint8_t> output({ 0, 0, 0, 0,  1, 1, 1, 1,
-                                  0, 0, 0, 0,  0, 0, 0, 0 });
-
-    return ElementwiseTestHelper<armnn::GreaterQueueDescriptor, armnn::DataType::Float32, armnn::DataType::Boolean>(
-        workloadFactory,
-        memoryManager,
-        shape,
-        input0,
-        shape,
-        input1,
-        shape,
-        output);
-}
-
-LayerTestResult<uint8_t, 4> GreaterBroadcast1ElementTest(
-        armnn::IWorkloadFactory& workloadFactory,
-        const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    unsigned int shape0[] = { 1, 2, 2, 2 };
-    std::vector<float> input0({ 1, 2, 3, 4, 5, 6, 7, 8});
-
-    unsigned int shape1[] = { 1, 1, 1, 1 };
-    std::vector<float> input1({ 1 });
-
-    std::vector<uint8_t> output({ 0, 1, 1, 1, 1, 1, 1, 1});
-
-    return ElementwiseTestHelper<armnn::GreaterQueueDescriptor, armnn::DataType::Float32, armnn::DataType::Boolean>(
-        workloadFactory,
-        memoryManager,
-        shape0,
-        input0,
-        shape1,
-        input1,
-        shape0,
-        output);
-}
-
-LayerTestResult<uint8_t, 4> GreaterBroadcast1DVectorTest(
-        armnn::IWorkloadFactory& workloadFactory,
-        const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    const unsigned int shape0[] = { 1, 2, 2, 3 };
-    const unsigned int shape1[] = { 1, 1, 1, 3 };
-
-    std::vector<float> input0({ 1, 2.9f, 2.1f, 4, 5, 6,
-                                7, 8, 9, 10, 11, 12 });
-
-    std::vector<float> input1({ 1, 3, 2});
-
-    std::vector<uint8_t> output({ 0, 0, 1, 1, 1, 1,
-                                  1, 1, 1, 1, 1, 1 });
-
-    return ElementwiseTestHelper<armnn::GreaterQueueDescriptor, armnn::DataType::Float32, armnn::DataType::Boolean>(
-        workloadFactory,
-        memoryManager,
-        shape0,
-        input0,
-        shape1,
-        input1,
-        shape0,
-        output);
-}
-
-LayerTestResult<uint8_t, 4> GreaterUint8Test(
-        armnn::IWorkloadFactory& workloadFactory,
-        const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    unsigned int shape[] = { 2, 2, 2, 2 };
-
-    // See dequantized values to the right.
-    std::vector<uint8_t> input0({ 1, 1, 1, 1, 6, 6, 6, 6,
-                                  3, 3, 3, 3, 5, 5, 5, 5 });
-
-    std::vector<uint8_t> input1({ 2, 2, 2, 2, 6, 6, 6, 6,
-                                  2, 2, 2, 2, 5, 5, 5, 5 });
-
-    std::vector<uint8_t> output({ 0, 0, 0, 0, 0, 0, 0, 0,
-                                  1, 1, 1, 1, 0, 0, 0, 0 });
-
-    return ElementwiseTestHelper<armnn::GreaterQueueDescriptor,
-                                 armnn::DataType::QuantisedAsymm8,
-                                 armnn::DataType::Boolean>(
-        workloadFactory,
-        memoryManager,
-        shape,
-        input0,
-        shape,
-        input1,
-        shape,
-        output,
-        1.0f,
-        0);
-}
-
-LayerTestResult<uint8_t, 4> GreaterBroadcast1ElementUint8Test(
-        armnn::IWorkloadFactory& workloadFactory,
-        const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    const unsigned int shape0[] = { 1, 2, 2, 3 };
-    const unsigned int shape1[] = { 1, 1, 1, 1 };
-
-    std::vector<uint8_t> input0({ 1, 2, 3, 4, 5, 6,
-                                  7, 8, 9, 10, 11, 12 });
-
-    std::vector<uint8_t> input1({ 1 });
-
-    std::vector<uint8_t> output({ 0, 1, 1, 1, 1, 1,
-                                  1, 1, 1, 1, 1, 1 });
-
-    return ElementwiseTestHelper<armnn::GreaterQueueDescriptor,
-                                 armnn::DataType::QuantisedAsymm8,
-                                 armnn::DataType::Boolean>(
-        workloadFactory,
-        memoryManager,
-        shape0,
-        input0,
-        shape1,
-        input1,
-        shape0,
-        output,
-        1.0f,
-        0);
-}
-
-LayerTestResult<uint8_t, 4> GreaterBroadcast1DVectorUint8Test(
-        armnn::IWorkloadFactory& workloadFactory,
-        const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    const unsigned int shape0[] = { 1, 2, 2, 3 };
-    const unsigned int shape1[] = { 1, 1, 1, 3 };
-
-    std::vector<uint8_t> input0({ 1, 2, 3, 4, 5, 6,
-                                  7, 8, 9, 10, 11, 12 });
-
-    std::vector<uint8_t> input1({ 1, 1, 3});
-
-    std::vector<uint8_t> output({ 0, 1, 0, 1, 1, 1,
-                                  1, 1, 1, 1, 1, 1 });
-
-    return ElementwiseTestHelper<armnn::GreaterQueueDescriptor,
-                                 armnn::DataType::QuantisedAsymm8,
-                                 armnn::DataType::Boolean>(
-        workloadFactory,
-        memoryManager,
-        shape0,
-        input0,
-        shape1,
-        input1,
-        shape0,
-        output,
-        1.0f,
-        0);
-}
-
-LayerTestResult<float, 4> MaximumSimpleTest(armnn::IWorkloadFactory& workloadFactory,
-                                           const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    const unsigned int width = 2;
-    const unsigned int height = 2;
-    const unsigned int channelCount = 2;
-    const unsigned int batchSize = 2;
-
-    unsigned int shape[] = { batchSize, channelCount, height, width };
-
-    std::vector<float> input0({ 1, 1, 1, 1,  5, 5, 5, 5,
-                                3, 3, 3, 3,  4, 4, 4, 4 });
-
-    std::vector<float> input1({ 2, 2, 2, 2,  3, 3, 3, 3,
-                                4, 4, 4, 4,  5, 5, 5, 5 });
-
-    std::vector<float> output({ 2, 2, 2, 2,  5, 5, 5, 5,
-                                4, 4, 4, 4,  5, 5, 5, 5 });
-
-    return ElementwiseTestHelper<armnn::MaximumQueueDescriptor, armnn::DataType::Float32>(
-        workloadFactory,
-        memoryManager,
-        shape,
-        input0,
-        shape,
-        input1,
-        shape,
-        output);
-}
-
-LayerTestResult<float, 4> MaximumBroadcast1ElementTest(
-        armnn::IWorkloadFactory& workloadFactory,
-        const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    unsigned int shape0[] = { 1, 2, 2, 2 };
-    std::vector<float> input0({ 1, 2, 3, 4, 5, 6, 7, 8});
-
-    unsigned int shape1[] = { 1, 1, 1, 1 };
-    std::vector<float> input1({ 2 });
-
-    std::vector<float> output({ 2, 2, 3, 4, 5, 6, 7, 8});
-
-    return ElementwiseTestHelper<armnn::MaximumQueueDescriptor, armnn::DataType::Float32>(
-        workloadFactory,
-        memoryManager,
-        shape0,
-        input0,
-        shape1,
-        input1,
-        shape0,
-        output);
-}
-
-LayerTestResult<float, 4> MaximumBroadcast1DVectorTest(
-        armnn::IWorkloadFactory& workloadFactory,
-        const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    const unsigned int shape0[] = { 1, 2, 2, 3 };
-    const unsigned int shape1[] = { 1, 1, 1, 3 };
-
-    std::vector<float> input0({ 1, 2, 3, 4, 5, 6,
-                                  7, 8, 9, 10, 11, 12 });
-
-    std::vector<float> input1({ 1, 2, 3});
-
-    std::vector<float> output({ 1, 2, 3, 4, 5, 6,
-                                7, 8, 9, 10, 11, 12 });
-
-    return ElementwiseTestHelper<armnn::MaximumQueueDescriptor, armnn::DataType::Float32>(
-        workloadFactory,
-        memoryManager,
-        shape0,
-        input0,
-        shape1,
-        input1,
-        shape0,
-        output);
-}
-
-LayerTestResult<uint8_t, 4> MaximumUint8Test(
-        armnn::IWorkloadFactory& workloadFactory,
-        const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    unsigned int shape[] = { 2, 2, 2, 2 };
-
-    // See dequantized values to the right.
-    std::vector<uint8_t> input0({ 1, 1, 1, 1, 6, 6, 6, 6,
-                                  3, 3, 3, 3, 4, 4, 4, 4 });
-
-    std::vector<uint8_t> input1({ 2, 2, 2, 2, 3, 3, 3, 3,
-                                  4, 4, 4, 4, 5, 5, 5, 5 });
-
-    std::vector<uint8_t> output({ 2, 2, 2, 2, 6, 6, 6, 6,
-                                  4, 4, 4, 4, 5, 5, 5, 5 });
-
-    return ElementwiseTestHelper<armnn::MaximumQueueDescriptor, armnn::DataType::QuantisedAsymm8>(
-        workloadFactory,
-        memoryManager,
-        shape,
-        input0,
-        shape,
-        input1,
-        shape,
-        output,
-        1.0f,
-        0);
-}
-
-LayerTestResult<uint8_t, 4> MaximumBroadcast1ElementUint8Test(
-        armnn::IWorkloadFactory& workloadFactory,
-        const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    const unsigned int shape0[] = { 1, 2, 2, 3 };
-    const unsigned int shape1[] = { 1, 1, 1, 1 };
-
-    std::vector<uint8_t> input0({ 1, 2, 3, 4,  5, 6,
-                                  7, 8, 9, 10, 11, 12 });
-
-    std::vector<uint8_t> input1({2});
-
-    std::vector<uint8_t> output({ 2, 2, 3, 4, 5, 6,
-                                  7, 8, 9, 10, 11, 12 });
-
-    return ElementwiseTestHelper<armnn::MaximumQueueDescriptor, armnn::DataType::QuantisedAsymm8>(
-        workloadFactory,
-        memoryManager,
-        shape0,
-        input0,
-        shape1,
-        input1,
-        shape0,
-        output,
-        1.0f,
-        0);
-}
-
-LayerTestResult<uint8_t, 4> MaximumBroadcast1DVectorUint8Test(
-        armnn::IWorkloadFactory& workloadFactory,
-        const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    const unsigned int shape0[] = { 1, 2, 2, 3 };
-    const unsigned int shape1[] = { 1, 1, 1, 3 };
-
-    std::vector<uint8_t> input0({ 1, 2, 3, 4, 5, 6,
-                                  7, 8, 9, 10, 11, 12 });
-
-    std::vector<uint8_t> input1({ 1, 10, 3});
-
-    std::vector<uint8_t> output({ 1, 10, 3, 4, 10, 6,
-                                  7, 10, 9, 10, 11, 12 });
-
-    return ElementwiseTestHelper<armnn::MaximumQueueDescriptor, armnn::DataType::QuantisedAsymm8>(
-        workloadFactory,
-        memoryManager,
-        shape0,
-        input0,
-        shape1,
-        input1,
-        shape0,
-        output,
-        1.0f,
-        0);
-}
-
-LayerTestResult<int16_t, 4> MaximumInt16Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    unsigned int shape[] = { 2, 2, 2, 2 };
-
-    std::vector<int16_t> input0({ 1, 1, 1, 1, 6, 6, 6, 6,
-                                  3, 3, 3, 3, 4, 4, 4, 4 });
-
-    std::vector<int16_t> input1({ 2, 2, 2, 2, 3, 3, 3, 3,
-                                  4, 4, 4, 4, 5, 5, 5, 5 });
-
-    std::vector<int16_t> output({ 2, 2, 2, 2, 6, 6, 6, 6,
-                                  4, 4, 4, 4, 5, 5, 5, 5 });
-
-    return ElementwiseTestHelper<armnn::MaximumQueueDescriptor, armnn::DataType::QuantisedSymm16>(
-        workloadFactory,
-        memoryManager,
-        shape,
-        input0,
-        shape,
-        input1,
-        shape,
-        output,
-        1.0f,
-        0);
-}
-
-LayerTestResult<int16_t, 4> MaximumBroadcast1ElementInt16Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    const unsigned int shape0[] = { 1, 2, 2, 3 };
-    const unsigned int shape1[] = { 1, 1, 1, 1 };
-
-    std::vector<int16_t> input0({ 1, 2, 3, 4,  5, 6,
-                                  7, 8, 9, 10, 11, 12 });
-
-    std::vector<int16_t> input1({2});
-
-    std::vector<int16_t> output({ 2, 2, 3, 4, 5, 6,
-                                  7, 8, 9, 10, 11, 12 });
-
-    return ElementwiseTestHelper<armnn::MaximumQueueDescriptor, armnn::DataType::QuantisedSymm16>(
-        workloadFactory,
-        memoryManager,
-        shape0,
-        input0,
-        shape1,
-        input1,
-        shape0,
-        output,
-        1.0f,
-        0);
-}
-
-LayerTestResult<int16_t, 4> MaximumBroadcast1DVectorInt16Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    const unsigned int shape0[] = { 1, 2, 2, 3 };
-    const unsigned int shape1[] = { 1, 1, 1, 3 };
-
-    std::vector<int16_t> input0({ 1, 2, 3, 4, 5, 6,
-                                  7, 8, 9, 10, 11, 12 });
-
-    std::vector<int16_t> input1({ 1, 10, 3});
-
-    std::vector<int16_t> output({ 1, 10, 3, 4, 10, 6,
-                                  7, 10, 9, 10, 11, 12 });
-
-    return ElementwiseTestHelper<armnn::MaximumQueueDescriptor, armnn::DataType::QuantisedSymm16>(
-        workloadFactory,
-        memoryManager,
-        shape0,
-        input0,
-        shape1,
-        input1,
-        shape0,
-        output,
-        1.0f,
-        0);
-}
-
-LayerTestResult<float, 4> MinimumBroadcast1ElementTest1(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    unsigned int shape0[] = { 1, 2, 2, 2 };
-    std::vector<float> input0({ 1, 2, 3, 4, 5, 6, 7, 8});
-
-    unsigned int shape1[] = { 1, 1, 1, 1 };
-    std::vector<float> input1({ 2 });
-
-    std::vector<float> output({ 1, 2, 2, 2, 2, 2, 2, 2});
-
-    return ElementwiseTestHelper<armnn::MinimumQueueDescriptor, armnn::DataType::Float32>(
-        workloadFactory,
-        memoryManager,
-        shape0,
-        input0,
-        shape1,
-        input1,
-        shape0,
-        output);
-}
-
-
-LayerTestResult<float, 4> MinimumBroadcast1ElementTest2(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    unsigned int shape0[] = { 1, 2, 2, 2 };
-    std::vector<float> input0({ 1, 6, 3, 2, 8, 9, 1, 10});
-
-    unsigned int shape1[] = { 1, 1, 1, 1 };
-    std::vector<float> input1({ 5 });
-
-    std::vector<float> output({ 1, 5, 3, 2, 5, 5, 1, 5});
-
-    return ElementwiseTestHelper<armnn::MinimumQueueDescriptor, armnn::DataType::Float32>(
-        workloadFactory,
-        memoryManager,
-        shape0,
-        input0,
-        shape1,
-        input1,
-        shape0,
-        output);
-}
-
-LayerTestResult<uint8_t, 4> MinimumBroadcast1DVectorUint8Test(
-    armnn::IWorkloadFactory & workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr & memoryManager)
-{
-    const unsigned int shape0[] = { 1, 2, 2, 3 };
-    const unsigned int shape1[] = { 1, 1, 1, 3 };
-
-    std::vector<uint8_t> input0({ 1, 2, 3, 3, 2, 1,
-                                  7, 1, 2, 3, 4, 5 });
-
-    std::vector<uint8_t> input1({ 1, 2, 3});
-
-    std::vector<uint8_t> output({ 1, 2, 3, 1, 2, 1,
-                                  1, 1, 2, 1, 2, 3 });
-
-    return ElementwiseTestHelper<armnn::MinimumQueueDescriptor, armnn::DataType::QuantisedAsymm8>(
-        workloadFactory,
-        memoryManager,
-        shape0,
-        input0,
-        shape1,
-        input1,
-        shape0,
-        output,
-        1.0f,
-        0);
-}
-
-LayerTestResult<int16_t, 4> MinimumInt16Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    unsigned int shape[] = { 2, 2, 2, 2 };
-
-    std::vector<int16_t> input0({ 1, 1, 1, 1, 6, 6, 6, 6,
-                                  3, 3, 3, 3, 4, 4, 4, 4 });
-
-    std::vector<int16_t> input1({ 2, 2, 2, 2, 3, 3, 3, 3,
-                                  4, 4, 4, 4, 5, 5, 5, 5 });
-
-    std::vector<int16_t> output({ 1, 1, 1, 1, 3, 3, 3, 3,
-                                  3, 3, 3, 3, 4, 4, 4, 4 });
-
-    return ElementwiseTestHelper<armnn::MinimumQueueDescriptor, armnn::DataType::QuantisedSymm16>(
-        workloadFactory,
-        memoryManager,
-        shape,
-        input0,
-        shape,
-        input1,
-        shape,
-        output,
-        1.0f,
-        0);
-}
-
-LayerTestResult<int16_t, 4> MinimumBroadcast1ElementInt16Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    const unsigned int shape0[] = { 1, 2, 2, 3 };
-    const unsigned int shape1[] = { 1, 1, 1, 1 };
-
-    std::vector<int16_t> input0({ 1, 2, 3, 4,  5, 6,
-                                  7, 8, 9, 10, 11, 12 });
-
-    std::vector<int16_t> input1({2});
-
-    std::vector<int16_t> output({ 1, 2, 2, 2, 2, 2,
-                                  2, 2, 2, 2, 2, 2 });
-
-    return ElementwiseTestHelper<armnn::MinimumQueueDescriptor, armnn::DataType::QuantisedSymm16>(
-        workloadFactory,
-        memoryManager,
-        shape0,
-        input0,
-        shape1,
-        input1,
-        shape0,
-        output,
-        1.0f,
-        0);
-}
-
-LayerTestResult<int16_t, 4> MinimumBroadcast1DVectorInt16Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    const unsigned int shape0[] = { 1, 2, 2, 3 };
-    const unsigned int shape1[] = { 1, 1, 1, 3 };
-
-    std::vector<int16_t> input0({ 1, 2, 3, 4, 5, 6,
-                                  7, 8, 9, 10, 11, 12 });
-
-    std::vector<int16_t> input1({ 1, 10, 3});
-
-    std::vector<int16_t> output({ 1, 2, 3, 1, 5, 3,
-                                  1, 8, 3, 1, 10, 3 });
-
-    return ElementwiseTestHelper<armnn::MinimumQueueDescriptor, armnn::DataType::QuantisedSymm16>(
-        workloadFactory,
-        memoryManager,
-        shape0,
-        input0,
-        shape1,
-        input1,
-        shape0,
-        output,
-        1.0f,
-        0);
-}
-
-namespace {
-template<std::size_t NumDims>
-LayerTestResult<float,NumDims> MultiplicationTestHelper(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
-    const unsigned int shape0[NumDims],
-    const std::vector<float> & values0,
-    const unsigned int shape1[NumDims],
-    const std::vector<float> & values1,
-    const unsigned int outShape[NumDims],
-    const std::vector<float> & outValues)
-{
-    armnn::TensorInfo inputTensorInfo0{NumDims, shape0, armnn::DataType::Float32};
-    armnn::TensorInfo inputTensorInfo1{NumDims, shape1, armnn::DataType::Float32};
-    armnn::TensorInfo outputTensorInfo{NumDims, outShape, armnn::DataType::Float32};
-
-    auto input0 = MakeTensor<float, NumDims>(inputTensorInfo0, values0);
-    auto input1 = MakeTensor<float, NumDims>(inputTensorInfo1, values1);
-
-    LayerTestResult<float,NumDims> ret(outputTensorInfo);
-
-    std::unique_ptr<armnn::ITensorHandle> inputHandle0 = workloadFactory.CreateTensorHandle(inputTensorInfo0);
-    std::unique_ptr<armnn::ITensorHandle> inputHandle1 = workloadFactory.CreateTensorHandle(inputTensorInfo1);
-    std::unique_ptr<armnn::ITensorHandle> outputHandle = workloadFactory.CreateTensorHandle(outputTensorInfo);
-
-    armnn::MultiplicationQueueDescriptor data;
-    armnn::WorkloadInfo info;
-    AddInputToWorkload(data, info, inputTensorInfo0, inputHandle0.get());
-    AddInputToWorkload(data, info, inputTensorInfo1, inputHandle1.get());
-    AddOutputToWorkload(data, info, outputTensorInfo, outputHandle.get());
-
-    std::unique_ptr<armnn::IWorkload> workload = workloadFactory.CreateMultiplication(data, info);
-
-    inputHandle0->Allocate();
-    inputHandle1->Allocate();
-    outputHandle->Allocate();
-
-    CopyDataToITensorHandle(inputHandle0.get(), input0.origin());
-    CopyDataToITensorHandle(inputHandle1.get(), input1.origin());
-
-    workload->PostAllocationConfigure();
-    workload->Execute();
-
-    CopyDataFromITensorHandle(ret.output.origin(), outputHandle.get());
-
-    ret.outputExpected = MakeTensor<float, NumDims>(outputTensorInfo, outValues);
-    return ret;
-}
-} // anonymous namespace
-
-
-LayerTestResult<float,4> MultiplicationTest(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    const unsigned int width = 2;
-    const unsigned int height = 2;
-    const unsigned int channelCount = 2;
-    const unsigned int batchSize = 2;
-
-    unsigned int shape[] = { batchSize, channelCount, height, width };
-
-    std::vector<float> input0({
-        1,  1,  1,  1,    2,  2,  2,  2,
-        3,  3,  3,  3,    4,  4,  4,  4 });
-
-    std::vector<float> input1({
-        2,  2,  2,  2,    3,  3,  3,  3,
-        4,  4,  4,  4,    5,  5,  5,  5 });
-
-    std::vector<float> output({
-        2,  2,  2,  2,    6,  6,  6,  6,
-        12, 12, 12, 12,  20, 20, 20, 20 });
-
-    return MultiplicationTestHelper<4>(workloadFactory,
-                                       memoryManager,
-                                       shape,
-                                       input0,
-                                       shape,
-                                       input1,
-                                       shape,
-                                       output);
-}
-
-LayerTestResult<float,5> Multiplication5dTest(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    const unsigned int width = 3;
-    const unsigned int height = 2;
-    const unsigned int channelCount = 2;
-    const unsigned int batchSize = 2;
-    const unsigned int depth = 2;
-
-    unsigned int shape[] = { depth, batchSize, channelCount, height, width };
-
-    std::vector<float> input0({
-        1.80f, 0.20f, 2.30f,  1.30f, 2.10f, 1.00f,
-        2.60f, 0.60f, 2.10f,  2.30f, 2.30f, 2.00f,
-
-        2.50f, 1.00f, 2.90f,  3.10f, 1.50f, 2.40f,
-        2.80f, 1.10f, 1.00f,  3.20f, 1.00f, 2.30f,
-
-
-        0.30f, 2.20f, 1.00f,  0.20f, 1.60f, 1.40f,
-        0.80f, 3.20f, 0.10f,  0.10f, 3.10f, 2.10f,
-
-        1.50f, 2.40f, 1.40f,  0.70f, 2.40f, 1.40f,
-        1.60f, 1.20f, 1.90f,  0.80f, 0.00f, 0.10f,
-    });
-
-    std::vector<float> input1({
-        0.70f, 1.00f, 2.90f,  2.20f, 3.10f, 2.80f,
-        1.80f, 2.00f, 0.50f,  2.30f, 1.20f, 2.70f,
-
-        2.40f, 0.20f, 3.20f,  1.60f, 0.20f, 2.50f,
-        2.30f, 0.70f, 2.70f,  1.80f, 2.90f, 2.70f,
-
-
-        3.20f, 3.20f, 0.70f,  1.90f, 2.70f, 2.50f,
-        2.40f, 0.90f, 2.30f,  1.80f, 2.50f, 2.00f,
-
-        1.60f, 2.20f, 1.60f,  2.00f, 0.30f, 3.20f,
-        0.40f, 3.00f, 2.60f,  0.30f, 0.00f, 2.50f,
-    });
-
-    std::vector<float> output({
-        1.26f, 0.20f, 6.67f,  2.86f, 6.51f, 2.80f,
-        4.68f, 1.20f, 1.05f,  5.29f, 2.76f, 5.40f,
-
-        6.00f, 0.20f, 9.28f,  4.96f, 0.30f, 6.00f,
-        6.44f, 0.77f, 2.70f,  5.76f, 2.90f, 6.21f,
-
-
-        0.96f, 7.04f, 0.70f,  0.38f, 4.32f, 3.50f,
-        1.92f, 2.88f, 0.23f,  0.18f, 7.75f, 4.20f,
-
-        2.40f, 5.28f, 2.24f,  1.40f, 0.72f, 4.48f,
-        0.64f, 3.60f, 4.94f,  0.24f, 0.00f, 0.25f,
-    });
-
-    return MultiplicationTestHelper<5>(workloadFactory,
-                                       memoryManager,
-                                       shape,
-                                       input0,
-                                       shape,
-                                       input1,
-                                       shape,
-                                       output);
-}
-
-LayerTestResult<float, 4> MultiplicationBroadcast1ElementTest(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    unsigned int shape0[] = { 1, 2, 2, 2 };
-    std::vector<float> input0({ 1, 2, 3, 4, 5, 6, 7, 8});
-
-    unsigned int shape1[] = { 1, 1, 1, 1 };
-    std::vector<float> input1({ 2 });
-
-    std::vector<float> output({ 2, 4, 6, 8, 10, 12, 14, 16});
-
-    return MultiplicationTestHelper<4>(workloadFactory,
-                                       memoryManager,
-                                       shape0,
-                                       input0,
-                                       shape1,
-                                       input1,
-                                       shape0,
-                                       output);
-}
-
-LayerTestResult<float, 4> MultiplicationBroadcast1DVectorTest(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    unsigned int shape0[] = { 1, 3, 3, 2 };
-    std::vector<float> input0({
-        1,   2,      3,  4,      5,  6,
-        7,   8,      9, 10,     11, 12,
-        13, 14,     15, 16,     17, 18});
-
-    unsigned int shape1[] = { 1, 1, 1, 2 };
-    std::vector<float> input1({ 1, 2 });
-
-    std::vector<float> output({
-        1,   4,       3,  8,      5, 12,
-        7,   16,      9, 20,     11, 24,
-        13,  28,     15, 32,     17, 36});
-
-    return MultiplicationTestHelper<4>(workloadFactory,
-                                       memoryManager,
-                                       shape0,
-                                       input0,
-                                       shape1,
-                                       input1,
-                                       shape0,
-                                       output);
-}
-
-LayerTestResult<float,4> CompareMultiplicationTest(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
-    armnn::IWorkloadFactory& refWorkloadFactory)
-{
-    const unsigned int width = 16;
-    const unsigned int height = 32;
-    const unsigned int channelCount = 2;
-    const unsigned int batchSize = 5;
-
-    armnn::TensorInfo inputTensorInfo0;
-    armnn::TensorInfo inputTensorInfo1;
-    armnn::TensorInfo outputTensorInfo;
-
-    constexpr unsigned int shape[] = { batchSize, channelCount, height, width };
-
-    inputTensorInfo0 = armnn::TensorInfo(4, shape, armnn::DataType::Float32);
-    inputTensorInfo1 = armnn::TensorInfo(4, shape, armnn::DataType::Float32);
-    outputTensorInfo = armnn::TensorInfo(4, shape, armnn::DataType::Float32);
-
-    LayerTestResult<float,4> comparisonResult(outputTensorInfo);
-
-    auto input0 = MakeRandomTensor<float, 4>(inputTensorInfo0, 803506992);
-    auto input1 = MakeRandomTensor<float, 4>(inputTensorInfo1, 54902257);
-
-    std::unique_ptr<armnn::ITensorHandle> inputHandle0 = workloadFactory.CreateTensorHandle(inputTensorInfo0);
-    std::unique_ptr<armnn::ITensorHandle> inputHandle1 = workloadFactory.CreateTensorHandle(inputTensorInfo1);
-    std::unique_ptr<armnn::ITensorHandle> outputHandle = workloadFactory.CreateTensorHandle(outputTensorInfo);
-
-    std::unique_ptr<armnn::ITensorHandle> inputHandle0Ref = refWorkloadFactory.CreateTensorHandle(inputTensorInfo0);
-    std::unique_ptr<armnn::ITensorHandle> inputHandle1Ref = refWorkloadFactory.CreateTensorHandle(inputTensorInfo1);
-    std::unique_ptr<armnn::ITensorHandle> outputHandleRef = refWorkloadFactory.CreateTensorHandle(outputTensorInfo);
-
-    armnn::MultiplicationQueueDescriptor data;
-    armnn::WorkloadInfo info;
-    AddInputToWorkload(data, info, inputTensorInfo0, inputHandle0.get());
-    AddInputToWorkload(data, info, inputTensorInfo1, inputHandle1.get());
-    AddOutputToWorkload(data, info, outputTensorInfo, outputHandle.get());
-
-    armnn::MultiplicationQueueDescriptor refData = data;
-    armnn::WorkloadInfo refInfo = info;
-    SetWorkloadInput(refData, refInfo, 0, inputTensorInfo0, inputHandle0Ref.get());
-    SetWorkloadInput(refData, refInfo, 1, inputTensorInfo1, inputHandle1Ref.get());
-    SetWorkloadOutput(refData, refInfo, 0, outputTensorInfo, outputHandleRef.get());
-
-    std::unique_ptr<armnn::IWorkload> workload = workloadFactory.CreateMultiplication(data, info);
-    std::unique_ptr<armnn::IWorkload> workloadRef = refWorkloadFactory.CreateMultiplication(refData, refInfo);
-
-    inputHandle0->Allocate();
-    inputHandle1->Allocate();
-    outputHandle->Allocate();
-    inputHandle0Ref->Allocate();
-    inputHandle1Ref->Allocate();
-    outputHandleRef->Allocate();
-
-    CopyDataToITensorHandle(inputHandle0.get(), &input0[0][0][0][0]);
-    CopyDataToITensorHandle(inputHandle1.get(), &input1[0][0][0][0]);
-    CopyDataToITensorHandle(inputHandle0Ref.get(), &input0[0][0][0][0]);
-    CopyDataToITensorHandle(inputHandle1Ref.get(), &input1[0][0][0][0]);
-
-    workload->PostAllocationConfigure();
-    workload->Execute();
-    workloadRef->PostAllocationConfigure();
-    workloadRef->Execute();
-    CopyDataFromITensorHandle(&comparisonResult.output[0][0][0][0], outputHandle.get());
-    CopyDataFromITensorHandle(&comparisonResult.outputExpected[0][0][0][0], outputHandleRef.get());
-
-    return comparisonResult;
-}
-
 LayerTestResult<float,4> CompareBatchNormTest(
     armnn::IWorkloadFactory& workloadFactory,
     const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
@@ -8467,664 +6546,6 @@ LayerTestResult<uint16_t, 3> ConcatUint16Test(
     return ret;
 }
 
-namespace
-{
-template <typename T>
-LayerTestResult<T, 4> AdditionQuantizeTestHelper(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
-    const unsigned int shape0[4],
-    const std::vector<T>& values0,
-    float scale0,
-    int32_t offset0,
-    const unsigned int shape1[4],
-    const std::vector<T> & values1,
-    float scale1,
-    int32_t offset1,
-    const unsigned int outShape[4],
-    const std::vector<T> & outValues,
-    float outScale,
-    int32_t outOffset)
-{
-    auto dataType = (std::is_same<T, uint8_t>::value ?
-                     armnn::DataType::QuantisedAsymm8 :
-                     armnn::DataType::QuantisedSymm16);
-
-    armnn::TensorInfo inputTensorInfo0(4, shape0, dataType);
-    armnn::TensorInfo inputTensorInfo1(4, shape1, dataType);
-    armnn::TensorInfo outputTensorInfo(4, outShape, dataType);
-
-    inputTensorInfo0.SetQuantizationScale(scale0);
-    inputTensorInfo0.SetQuantizationOffset(offset0);
-
-    inputTensorInfo1.SetQuantizationScale(scale1);
-    inputTensorInfo1.SetQuantizationOffset(offset1);
-
-    outputTensorInfo.SetQuantizationScale(outScale);
-    outputTensorInfo.SetQuantizationOffset(outOffset);
-
-    auto input0 = MakeTensor<T, 4>(inputTensorInfo0, values0);
-    auto input1 = MakeTensor<T, 4>(inputTensorInfo1, values1);
-
-    LayerTestResult<T, 4> result(outputTensorInfo);
-    result.outputExpected = MakeTensor<T, 4>(outputTensorInfo, outValues);
-
-    std::unique_ptr<armnn::ITensorHandle> inputHandle0 = workloadFactory.CreateTensorHandle(inputTensorInfo0);
-    std::unique_ptr<armnn::ITensorHandle> inputHandle1 = workloadFactory.CreateTensorHandle(inputTensorInfo1);
-    std::unique_ptr<armnn::ITensorHandle> outputHandle = workloadFactory.CreateTensorHandle(outputTensorInfo);
-
-    armnn::AdditionQueueDescriptor data;
-    armnn::WorkloadInfo info;
-    AddInputToWorkload(data,  info, inputTensorInfo0, inputHandle0.get());
-    AddInputToWorkload(data,  info, inputTensorInfo1, inputHandle1.get());
-    AddOutputToWorkload(data, info, outputTensorInfo, outputHandle.get());
-
-    std::unique_ptr<armnn::IWorkload> workload = workloadFactory.CreateAddition(data, info);
-
-    inputHandle0->Allocate();
-    inputHandle1->Allocate();
-    outputHandle->Allocate();
-
-    CopyDataToITensorHandle(inputHandle0.get(), &input0[0][0][0][0]);
-    CopyDataToITensorHandle(inputHandle1.get(), &input1[0][0][0][0]);
-
-    workload->PostAllocationConfigure();
-    workload->Execute();
-
-    CopyDataFromITensorHandle(&result.output[0][0][0][0], outputHandle.get());
-
-    return result;
-}
-} // anonymous namespace
-
-LayerTestResult<uint8_t, 4> AdditionUint8Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    const unsigned int shape0[] = { 1, 2, 2, 3 };
-    const unsigned int shape1[] = { 1, 2, 2, 3 };
-
-    std::vector<uint8_t> input0(
-    {
-        63,  35,  77,  70,  56, 112, //  420, 224,  518,  469,  371, 763
-        203,  28, 252, 168, 245,  91  // 1400, 175, 1743, 1155, 1694, 616
-    });
-
-    std::vector<uint8_t> input1(
-    {
-        21,   7, 175, 231, 175, 210, // 126,   28, 1204, 1596, 1204, 1449
-        126, 161,  63,  21, 105, 126  // 861, 1106,  420,  126,  714,  861
-    });
-
-    std::vector<uint8_t> output(
-    {
-        81,  39, 249, 255, 228, 255, //  546,  252, 1722, 2065(clamped), 1575, 2212(clamped)
-        255, 186, 255, 186, 255, 214, // 2261(clamped), 1281, 2163(clamped), 1281, 2408(clamped), 1477
-    });
-
-    return AdditionQuantizeTestHelper(workloadFactory,
-                                      memoryManager,
-                                      shape0, input0, 7.0f, 3,
-                                      shape1, input1, 7.0f, 3,
-                                      shape0, output, 7.0f, 3);
-}
-
-LayerTestResult<int16_t, 4> AdditionInt16Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    const unsigned int shape0[] = { 1, 2, 2, 3 };
-    const unsigned int shape1[] = { 1, 2, 2, 3 };
-
-    std::vector<int16_t> input0(
-        {
-            63,  35,  77,  70,  56, 112, //  441, 245,  539,  490,  392, 184
-            203,  28, 252, 168, 245,  91  // 1421, 196, 1764, 1176, 1715, 637
-        });
-
-    std::vector<int16_t> input1(
-        {
-            21,   7, 175, 231, 175, 210, // 126,   28, 1204, 1596, 1204, 1449
-            126, 161,  63,  21, 105, 126  // 861, 1106,  420,  126,  714,  861
-        });
-
-    std::vector<int16_t> output(
-        {
-            84,  42, 252, 301, 231, 322, //  588,  294, 1764, 2107(clamped), 1617, 2254(clamped)
-            329, 189, 315, 189, 350, 217, // 2303(clamped), 1323, 2205(clamped), 1323, 2450(clamped), 1519
-        });
-
-    return AdditionQuantizeTestHelper(workloadFactory,
-                                      memoryManager,
-                                      shape0, input0, 7.0f, 0,
-                                      shape1, input1, 7.0f, 0,
-                                      shape0, output, 7.0f, 0);
-}
-
-namespace
-{
-template<armnn::DataType ArmnnType, typename T = armnn::ResolveType<ArmnnType>>
-LayerTestResult<T, 4> MultiplicationQuantizeTestHelper(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
-    const unsigned int shape0[4],
-    const std::vector<T> & values0,
-    float scale0,
-    int32_t offset0,
-    const unsigned int shape1[4],
-    const std::vector<T> & values1,
-    float scale1,
-    int32_t offset1,
-    const unsigned int outShape[4],
-    const std::vector<T> & outValues,
-    float outScale,
-    int32_t outOffset)
-{
-    armnn::TensorInfo inputTensorInfo0(4, shape0, ArmnnType);
-    armnn::TensorInfo inputTensorInfo1(4, shape1, ArmnnType);
-    armnn::TensorInfo outputTensorInfo(4, outShape, ArmnnType);
-
-    inputTensorInfo0.SetQuantizationScale(scale0);
-    inputTensorInfo0.SetQuantizationOffset(offset0);
-
-    inputTensorInfo1.SetQuantizationScale(scale1);
-    inputTensorInfo1.SetQuantizationOffset(offset1);
-
-    outputTensorInfo.SetQuantizationScale(outScale);
-    outputTensorInfo.SetQuantizationOffset(outOffset);
-
-    auto input0 = MakeTensor<T, 4>(inputTensorInfo0, values0);
-    auto input1 = MakeTensor<T, 4>(inputTensorInfo1, values1);
-
-    LayerTestResult<T, 4> result(outputTensorInfo);
-    result.outputExpected = MakeTensor<T, 4>(outputTensorInfo, outValues);
-
-    std::unique_ptr<armnn::ITensorHandle> inputHandle0 = workloadFactory.CreateTensorHandle(inputTensorInfo0);
-    std::unique_ptr<armnn::ITensorHandle> inputHandle1 = workloadFactory.CreateTensorHandle(inputTensorInfo1);
-    std::unique_ptr<armnn::ITensorHandle> outputHandle = workloadFactory.CreateTensorHandle(outputTensorInfo);
-
-    armnn::MultiplicationQueueDescriptor data;
-    armnn::WorkloadInfo info;
-    AddInputToWorkload(data,  info, inputTensorInfo0, inputHandle0.get());
-    AddInputToWorkload(data,  info, inputTensorInfo1, inputHandle1.get());
-    AddOutputToWorkload(data, info, outputTensorInfo, outputHandle.get());
-
-    std::unique_ptr<armnn::IWorkload> workload = workloadFactory.CreateMultiplication(data, info);
-
-    inputHandle0->Allocate();
-    inputHandle1->Allocate();
-    outputHandle->Allocate();
-
-    CopyDataToITensorHandle(inputHandle0.get(), &input0[0][0][0][0]);
-    CopyDataToITensorHandle(inputHandle1.get(), &input1[0][0][0][0]);
-
-    workload->PostAllocationConfigure();
-    workload->Execute();
-
-    CopyDataFromITensorHandle(&result.output[0][0][0][0], outputHandle.get());
-
-    return result;
-}
-} // anonymous namespace
-
-LayerTestResult<uint8_t, 4> MultiplicationUint8Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    unsigned int batchSize = 1;
-    unsigned int channels = 2;
-    unsigned int height = 2;
-    unsigned int width = 3;
-    const unsigned int shape[] = { batchSize, channels, height, width };
-
-    // See dequantized values to the right.
-    std::vector<uint8_t> input0({
-         62,  37,   3, 172,  13, 111, // 244, 144,   8, 684,  48, 440,
-        188,  20,  73,  31,  23,  31  // 748,  76, 288, 120,  88, 120
-    });
-
-    // See dequantized values to the right.
-    std::vector<uint8_t> input1({
-        126, 240, 252, 183, 121, 247, // 384, 726, 762, 555, 369, 747,
-         48, 115, 151,  79,  78,  97  // 150, 351, 459, 243, 240, 297
-    });
-
-    // See dequantized values to the right.
-    std::vector<uint8_t> output(
-    {
-         64,  72,   0, 255,   8, 236, //  93696, 104544, 6096(clamped), 379620(clamped), 17712, 328680,
-         77,  15,  92,  16,  10,  21, // 112200,  26676,        132192,           29160, 21120,  35640
-    });
-
-    // Scale/offset chosen to have output values out of range.
-    return MultiplicationQuantizeTestHelper<armnn::DataType::QuantisedAsymm8>(workloadFactory,
-                                                                              memoryManager,
-                                                                              shape,
-                                                                              input0,
-                                                                              4.0f,
-                                                                              1,
-                                                                              shape,
-                                                                              input1,
-                                                                              3.0f,
-                                                                              -2,
-                                                                              shape,
-                                                                              output,
-                                                                              1366.255f,
-                                                                              -5);
-}
-
-LayerTestResult<uint8_t, 4> MultiplicationBroadcast1ElementUint8Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    const unsigned int shape0[] = { 1, 2, 2, 3 };
-    const unsigned int shape1[] = { 1, 1, 1, 1 };
-
-    std::vector<uint8_t> input0({
-        1, 2, 3,    4,  5,  6,
-        7, 8, 9,   10, 11, 12
-    });
-
-    std::vector<uint8_t> input1({2});
-
-    std::vector<uint8_t> output({
-        2,  4,   6,     8, 10, 12,
-        14, 16, 18,    20, 22, 24
-    });
-
-    return MultiplicationQuantizeTestHelper<armnn::DataType::QuantisedAsymm8>(workloadFactory,
-                                                                              memoryManager,
-                                                                              shape0,
-                                                                              input0,
-                                                                              1.0f,
-                                                                              0,
-                                                                              shape1,
-                                                                              input1,
-                                                                              1.0f,
-                                                                              0,
-                                                                              shape0,
-                                                                              output,
-                                                                              1.0f,
-                                                                              0);
-}
-
-LayerTestResult<uint8_t, 4> MultiplicationBroadcast1DVectorUint8Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    const unsigned int shape0[] = { 1, 2, 2, 3 };
-    const unsigned int shape1[] = { 1, 1, 1, 3 };
-
-    std::vector<uint8_t> input0({
-        1, 2, 3,    4,  5,  6,
-        7, 8, 9,   10, 11, 12
-    });
-
-    std::vector<uint8_t> input1({1, 2, 3});
-
-    std::vector<uint8_t> output({
-        1,  4,   9,     4, 10, 18,
-        7, 16,  27,    10, 22, 36
-    });
-
-    return MultiplicationQuantizeTestHelper<armnn::DataType::QuantisedAsymm8>(workloadFactory,
-                                                                              memoryManager,
-                                                                              shape0,
-                                                                              input0,
-                                                                              1.0f,
-                                                                              0,
-                                                                              shape1,
-                                                                              input1,
-                                                                              1.0f,
-                                                                              0,
-                                                                              shape0,
-                                                                              output,
-                                                                              1.0f,
-                                                                              0);
-}
-
-LayerTestResult<int16_t, 4> MultiplicationInt16Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    const unsigned int shape[] = { 1, 2, 2, 3 };
-
-    std::vector<int16_t> input0(
-    {
-        6,   7,  8,  9, 10, 11,
-        12, 13, 14, 15, 16, 17
-    });
-
-    std::vector<int16_t> input1(
-    {
-        1, 2, 3,  4,  5,  6,
-        7, 8, 9, 10, 11, 12
-    });
-
-    std::vector<int16_t> output(
-    {
-        6,   14,  24,  36,  50,  66,
-        84, 104, 126, 150, 176, 204
-    });
-
-    return MultiplicationQuantizeTestHelper<armnn::DataType::QuantisedSymm16>(workloadFactory,
-                                                                              memoryManager,
-                                                                              shape,
-                                                                              input0,
-                                                                              1.0f,
-                                                                              0,
-                                                                              shape,
-                                                                              input1,
-                                                                              1.0f,
-                                                                              0,
-                                                                              shape,
-                                                                              output,
-                                                                              1.0f,
-                                                                              0);
-}
-
-LayerTestResult<int16_t, 4> MultiplicationBroadcast1ElementInt16Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    const unsigned int shape0[] = { 1, 2, 2, 3 };
-    const unsigned int shape1[] = { 1, 1, 1, 1 };
-
-    std::vector<int16_t> input0(
-    {
-        1, 2, 3,  4,  5,  6,
-        7, 8, 9, 10, 11, 12
-    });
-
-    std::vector<int16_t> input1({2});
-
-    std::vector<int16_t> output(
-    {
-        2,   4,  6,  8, 10, 12,
-        14, 16, 18, 20, 22, 24
-    });
-
-    return MultiplicationQuantizeTestHelper<armnn::DataType::QuantisedSymm16>(workloadFactory,
-                                                                              memoryManager,
-                                                                              shape0,
-                                                                              input0,
-                                                                              1.0f,
-                                                                              0,
-                                                                              shape1,
-                                                                              input1,
-                                                                              1.0f,
-                                                                              0,
-                                                                              shape0,
-                                                                              output,
-                                                                              1.0f,
-                                                                              0);
-}
-
-LayerTestResult<int16_t, 4> MultiplicationBroadcast1DVectorInt16Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    const unsigned int shape0[] = { 1, 2, 2, 3 };
-    const unsigned int shape1[] = { 1, 1, 1, 3 };
-
-    std::vector<int16_t> input0(
-    {
-        1, 2, 3,  4,  5,  6,
-        7, 8, 9, 10, 11, 12
-    });
-
-    std::vector<int16_t> input1({1, 2, 3});
-
-    std::vector<int16_t> output(
-    {
-        1,  4,  9,  4, 10, 18,
-        7, 16, 27, 10, 22, 36
-    });
-
-    return MultiplicationQuantizeTestHelper<armnn::DataType::QuantisedSymm16>(workloadFactory,
-                                                                              memoryManager,
-                                                                              shape0,
-                                                                              input0,
-                                                                              1.0f,
-                                                                              0,
-                                                                              shape1,
-                                                                              input1,
-                                                                              1.0f,
-                                                                              0,
-                                                                              shape0,
-                                                                              output,
-                                                                              1.0f,
-                                                                              0);
-}
-
-namespace
-{
-template<armnn::DataType ArmnnType, typename T = armnn::ResolveType<ArmnnType>>
-LayerTestResult<T, 4> SubtractionTestHelper(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
-    const unsigned int shape0[4],
-    const std::vector<T>& values0,
-    float scale0,
-    int32_t offset0,
-    const unsigned int shape1[4],
-    const std::vector<T> & values1,
-    float scale1,
-    int32_t offset1,
-    const unsigned int outShape[4],
-    const std::vector<T> & outValues,
-    float outScale,
-    int32_t outOffset)
-{
-    armnn::TensorInfo inputTensorInfo0(4, shape0, ArmnnType);
-    armnn::TensorInfo inputTensorInfo1(4, shape1, ArmnnType);
-    armnn::TensorInfo outputTensorInfo(4, outShape, ArmnnType);
-
-    inputTensorInfo0.SetQuantizationScale(scale0);
-    inputTensorInfo0.SetQuantizationOffset(offset0);
-
-    inputTensorInfo1.SetQuantizationScale(scale1);
-    inputTensorInfo1.SetQuantizationOffset(offset1);
-
-    outputTensorInfo.SetQuantizationScale(outScale);
-    outputTensorInfo.SetQuantizationOffset(outOffset);
-
-    auto input0 = MakeTensor<T, 4>(inputTensorInfo0, values0);
-    auto input1 = MakeTensor<T, 4>(inputTensorInfo1, values1);
-
-    LayerTestResult<T, 4> result(outputTensorInfo);
-    result.outputExpected = MakeTensor<T, 4>(outputTensorInfo, outValues);
-
-    std::unique_ptr<armnn::ITensorHandle> inputHandle0 = workloadFactory.CreateTensorHandle(inputTensorInfo0);
-    std::unique_ptr<armnn::ITensorHandle> inputHandle1 = workloadFactory.CreateTensorHandle(inputTensorInfo1);
-    std::unique_ptr<armnn::ITensorHandle> outputHandle = workloadFactory.CreateTensorHandle(outputTensorInfo);
-
-    armnn::SubtractionQueueDescriptor data;
-    armnn::WorkloadInfo info;
-    AddInputToWorkload(data,  info, inputTensorInfo0, inputHandle0.get());
-    AddInputToWorkload(data,  info, inputTensorInfo1, inputHandle1.get());
-    AddOutputToWorkload(data, info, outputTensorInfo, outputHandle.get());
-
-    std::unique_ptr<armnn::IWorkload> workload = workloadFactory.CreateSubtraction(data, info);
-
-    inputHandle0->Allocate();
-    inputHandle1->Allocate();
-    outputHandle->Allocate();
-
-    CopyDataToITensorHandle(inputHandle0.get(), &input0[0][0][0][0]);
-    CopyDataToITensorHandle(inputHandle1.get(), &input1[0][0][0][0]);
-
-    workload->PostAllocationConfigure();
-    workload->Execute();
-
-    CopyDataFromITensorHandle(&result.output[0][0][0][0], outputHandle.get());
-
-    return result;
-}
-} // anonymous namespace
-
-LayerTestResult<uint8_t, 4> SubtractionUint8Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    const unsigned int shape0[] = { 1, 1, 2, 2 };
-    const unsigned int shape1[] = { 1, 1, 2, 2 };
-
-    std::vector<uint8_t> input0({ 10, 12, 14, 16 });
-    std::vector<uint8_t> input1({ 1, 2, 1, 2 });
-    std::vector<uint8_t> output({ 3, 3, 5, 5 });
-
-    return SubtractionTestHelper<armnn::DataType::QuantisedAsymm8>(workloadFactory,
-                                                                   memoryManager,
-                                                                   shape0, input0, 0.5f, 2,
-                                                                   shape1, input1, 1.0f, 0,
-                                                                   shape0, output, 1.0f, 0);
-}
-
-LayerTestResult<uint8_t, 4> SubtractionBroadcast1ElementUint8Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    const unsigned int shape0[] = { 1, 1, 2, 2 };
-    const unsigned int shape1[] = { 1, 1, 1, 1 };
-
-    std::vector<uint8_t> input0({ 10, 12, 14, 16 });
-    std::vector<uint8_t> input1({ 2 });
-    std::vector<uint8_t> output({ 5, 6, 7, 8 });
-
-    return SubtractionTestHelper<armnn::DataType::QuantisedAsymm8>(workloadFactory,
-                                                                   memoryManager,
-                                                                   shape0, input0, 0.5f, 2,
-                                                                   shape1, input1, 1.0f, 0,
-                                                                   shape0, output, 1.0f, 3);
-}
-
-LayerTestResult<uint8_t, 4> SubtractionBroadcastUint8Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    const unsigned int shape0[] = { 1, 1, 2, 2 };
-    const unsigned int shape1[] = { 1, 1, 2, 1 };
-
-    std::vector<uint8_t> input0({ 10, 12, 14, 16 });
-    std::vector<uint8_t> input1({ 2, 1 });
-    std::vector<uint8_t> output({ 8, 11, 12, 15 });
-
-    return SubtractionTestHelper<armnn::DataType::QuantisedAsymm8>(workloadFactory,
-                                                                   memoryManager,
-                                                                   shape0, input0, 1.0f, 0,
-                                                                   shape1, input1, 1.0f, 0,
-                                                                   shape0, output, 1.0f, 0);
-}
-
-LayerTestResult<float, 4> SubtractionTest(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    const unsigned int shape0[] = { 1, 1, 2, 2 };
-    const unsigned int shape1[] = { 1, 1, 2, 2 };
-
-    std::vector<float> input0({ 1,  2, 3, 4 });
-    std::vector<float> input1({ 1, -1, 0, 2 });
-    std::vector<float> output({ 0,  3, 3, 2 });
-
-    return SubtractionTestHelper<armnn::DataType::Float32>(workloadFactory,
-                                                           memoryManager,
-                                                           shape0, input0, 1.0f, 0,
-                                                           shape1, input1, 1.0f, 0,
-                                                           shape0, output, 1.0f, 0);
-}
-
-LayerTestResult<float, 4> SubtractionBroadcast1ElementTest(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    const unsigned int shape0[] = { 1, 1, 2, 2 };
-    const unsigned int shape1[] = { 1, 1, 1, 1 };
-
-    std::vector<float> input0({ 1,  2, 3, 4 });
-    std::vector<float> input1({ 10 });
-    std::vector<float> output({ -9,  -8, -7, -6 });
-
-    return SubtractionTestHelper<armnn::DataType::Float32>(workloadFactory,
-                                                           memoryManager,
-                                                           shape0, input0, 1.0f, 0,
-                                                           shape1, input1, 1.0f, 0,
-                                                           shape0, output, 1.0f, 0);
-}
-
-LayerTestResult<float, 4> SubtractionBroadcastTest(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    const unsigned int shape0[] = { 1, 1, 2, 2 };
-    const unsigned int shape1[] = { 1, 1, 1, 2 };
-
-    std::vector<float> input0({ 1,  2, 3, 4 });
-    std::vector<float> input1({ 10, -5 });
-    std::vector<float> output({ -9,  7, -7, 9 });
-
-    return SubtractionTestHelper<armnn::DataType::Float32>(workloadFactory,
-                                                           memoryManager,
-                                                           shape0, input0, 1.0f, 0,
-                                                           shape1, input1, 1.0f, 0,
-                                                           shape0, output, 1.0f, 0);
-}
-
-LayerTestResult<int16_t, 4> SubtractionInt16Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    const unsigned int shape0[] = { 1, 1, 2, 2 };
-    const unsigned int shape1[] = { 1, 1, 2, 2 };
-
-    std::vector<int16_t> input0({ 10, 12, 14, 16 });
-    std::vector<int16_t> input1({ 1, 2, 1, 2 });
-    std::vector<int16_t> output({ 3, 3, 5, 5 });
-
-    return SubtractionTestHelper<armnn::DataType::QuantisedSymm16>(workloadFactory,
-                                                                   memoryManager,
-                                                                   shape0, input0, 0.5f, 0,
-                                                                   shape1, input1, 1.0f, 0,
-                                                                   shape0, output, 1.0f, 0);
-}
-
-LayerTestResult<int16_t, 4> SubtractionBroadcast1ElementInt16Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    const unsigned int shape0[] = { 1, 1, 2, 2 };
-    const unsigned int shape1[] = { 1, 1, 1, 1 };
-
-    std::vector<int16_t> input0({ 10, 12, 14, 16 });
-    std::vector<int16_t> input1({ 2 });
-    std::vector<int16_t> output({ 3, 4, 5, 6 });
-
-    return SubtractionTestHelper<armnn::DataType::QuantisedSymm16>(workloadFactory,
-                                                                   memoryManager,
-                                                                   shape0, input0, 0.5f, 0,
-                                                                   shape1, input1, 1.0f, 0,
-                                                                   shape0, output, 1.0f, 0);
-}
-
-LayerTestResult<int16_t, 4> SubtractionBroadcastInt16Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    const unsigned int shape0[] = { 1, 1, 2, 2 };
-    const unsigned int shape1[] = { 1, 1, 2, 1 };
-
-    std::vector<int16_t> input0({ 10, 12, 14, 16 });
-    std::vector<int16_t> input1({ 2, 1 });
-    std::vector<int16_t> output({ 8, 11, 12, 15 });
-
-    return SubtractionTestHelper<armnn::DataType::QuantisedSymm16>(workloadFactory,
-                                                                   memoryManager,
-                                                                   shape0, input0, 1.0f, 0,
-                                                                   shape1, input1, 1.0f, 0,
-                                                                   shape0, output, 1.0f, 0);
-}
-
 LayerTestResult<float, 4> BatchNormTest(
     armnn::IWorkloadFactory& workloadFactory,
     const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
@@ -10021,111 +7442,6 @@ LayerTestResult<float, 2> FullyConnectedLargeTest(
     return FullyConnectedLargeTestCommon<armnn::DataType::Float32>(workloadFactory, memoryManager, transposeWeights);
 }
 
-LayerTestResult<float, 4> AdditionAfterMaxPoolTest(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
-{
-    // Create Initial Tensor
-    // 1, 2, 3
-    // 4, 5, 6
-    // 7, 8, 9
-
-    armnn::TensorInfo poolingInputTensorInfo({ 1, 1, 3, 3}, armnn::DataType::Float32);
-    armnn::TensorInfo poolingOutputTensorInfo({ 1, 1, 2, 2}, armnn::DataType::Float32);
-
-    boost::multi_array<float, 4> poolingInput = MakeTensor<float,4>(poolingInputTensorInfo,
-                                                            {1, 2, 3,
-                                                             4, 5, 6,
-                                                             7, 8, 9
-                                                            });
-
-    std::unique_ptr<armnn::ITensorHandle> poolingInputHandle =
-            workloadFactory.CreateTensorHandle(poolingInputTensorInfo);
-    std::unique_ptr<armnn::ITensorHandle> poolingOutputHandle =
-            workloadFactory.CreateTensorHandle(poolingOutputTensorInfo);
-
-    // Apply MaxPool poolSize = 1x1, stride=2x2
-    // Result =
-    // 1, 3
-    // 7, 9
-    armnn::Pooling2dDescriptor descriptor;
-    descriptor.m_PoolHeight = 1;
-    descriptor.m_PoolWidth = 1;
-    descriptor.m_StrideX = 2;
-    descriptor.m_StrideY = 2;
-    descriptor.m_PoolType = armnn::PoolingAlgorithm::Max;
-
-    armnn::Pooling2dQueueDescriptor queueDescriptor;
-    queueDescriptor.m_Parameters = descriptor;
-    armnn::WorkloadInfo workloadInfo;
-    AddInputToWorkload(queueDescriptor, workloadInfo, poolingInputTensorInfo, poolingInputHandle.get());
-    AddOutputToWorkload(queueDescriptor, workloadInfo, poolingOutputTensorInfo, poolingOutputHandle.get());
-
-    // Create the MaxPool
-    std::unique_ptr<armnn::IWorkload> workload = workloadFactory.CreatePooling2d(queueDescriptor, workloadInfo);
-
-    //LayerTestResult<float, 4> result(poolingOutputTensorInfo);
-    auto shape( GetTensorShapeAsArray<4>(poolingOutputTensorInfo));
-    boost::multi_array<float, 4> resultMaxPool;
-    resultMaxPool.resize(shape);
-
-
-    // Create addition with another tensor the same size
-    // This would be the result to apply a Conv2d with kernel ones(2) and stride 1x1
-    // with the initial tensor.
-    // 12, 16
-    // 24, 28
-
-    armnn::TensorInfo addInputTensorInfo({ 1,1,2,2}, armnn::DataType::Float32);
-    armnn::TensorInfo addOutputTensorInfo({ 1,1,2,2}, armnn::DataType::Float32);
-
-    boost::multi_array<float, 4> addInput = MakeTensor<float,4>(addInputTensorInfo,
-                                                                    {12, 16,
-                                                                     24, 28,
-                                                                    });
-
-    // Expected output tensor after MaxPool and Addition.
-    LayerTestResult<float,4> addRet(addOutputTensorInfo);
-    addRet.outputExpected = MakeTensor<float, 4>(addOutputTensorInfo, std::vector<float>(
-            {
-                    13, 19,
-                    31, 37
-            }));
-
-    std::unique_ptr<armnn::ITensorHandle> addInputHandle = workloadFactory.CreateTensorHandle(addInputTensorInfo);
-    std::unique_ptr<armnn::ITensorHandle> addOutputHandle = workloadFactory.CreateTensorHandle(addOutputTensorInfo);
-
-    armnn::AdditionQueueDescriptor data;
-    armnn::WorkloadInfo info;
-
-    // Add the output of the MaxPool and the new tensor
-    AddInputToWorkload(data, info, poolingOutputTensorInfo, poolingOutputHandle.get());
-    AddInputToWorkload(data, info, addInputTensorInfo, addInputHandle.get());
-    AddOutputToWorkload(data, info, addOutputTensorInfo, addOutputHandle.get());
-
-    std::unique_ptr<armnn::IWorkload> addWorkload = workloadFactory.CreateAddition(data, info);
-
-    poolingInputHandle->Allocate();
-    poolingOutputHandle->Allocate();
-    addInputHandle->Allocate();
-    addOutputHandle->Allocate();
-
-    CopyDataToITensorHandle(poolingInputHandle.get(), &poolingInput[0][0][0][0]);
-    CopyDataFromITensorHandle(&resultMaxPool[0][0][0][0], poolingOutputHandle.get());
-
-    CopyDataToITensorHandle(poolingOutputHandle.get(), &resultMaxPool[0][0][0][0]);
-    CopyDataToITensorHandle(addInputHandle.get(), &addInput[0][0][0][0]);
-
-    workload->PostAllocationConfigure();
-    workload->Execute();
-    addWorkload->PostAllocationConfigure();
-    addWorkload->Execute();
-
-    CopyDataFromITensorHandle(&addRet.output[0][0][0][0], addOutputHandle.get());
-
-    return addRet;
-}
-
 LayerTestResult<float, 4> SpaceToBatchNdSimpleFloat32Test(
     armnn::IWorkloadFactory& workloadFactory,
     const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
diff --git a/src/backends/backendsCommon/test/LayerTests.hpp b/src/backends/backendsCommon/test/LayerTests.hpp
index 235c5dcbd2..bb21202731 100644
--- a/src/backends/backendsCommon/test/LayerTests.hpp
+++ b/src/backends/backendsCommon/test/LayerTests.hpp
@@ -7,6 +7,17 @@
 #include <armnn/ArmNN.hpp>
 #include <armnn/Tensor.hpp>
 
+#include <backendsCommon/test/layerTests/LayerTestResult.hpp>
+
+#include <backendsCommon/test/layerTests/AdditionTestImpl.hpp>
+#include <backendsCommon/test/layerTests/DivisionTestImpl.hpp>
+#include <backendsCommon/test/layerTests/EqualTestImpl.hpp>
+#include <backendsCommon/test/layerTests/GreaterTestImpl.hpp>
+#include <backendsCommon/test/layerTests/MaximumTestImpl.hpp>
+#include <backendsCommon/test/layerTests/MinimumTestImpl.hpp>
+#include <backendsCommon/test/layerTests/MultiplicationTestImpl.hpp>
+#include <backendsCommon/test/layerTests/SubtractionTestImpl.hpp>
+
 #include <Half.hpp>
 #include "TensorCopyUtils.hpp"
 #include "WorkloadTestUtils.hpp"
@@ -32,38 +43,6 @@ namespace armnn
 class IWorkloadFactory;
 }
 
-template <std::size_t n>
-boost::array<unsigned int, n> GetTensorShapeAsArray(const armnn::TensorInfo& tensorInfo)
-{
-    BOOST_ASSERT_MSG(n == tensorInfo.GetNumDimensions(),
-        "Attempting to construct a shape array of mismatching size");
-
-    boost::array<unsigned int, n> shape;
-    for (unsigned int i = 0; i < n; i++)
-    {
-        shape[i] = tensorInfo.GetShape()[i];
-    }
-    return shape;
-}
-
-template <typename T, std::size_t n>
-struct LayerTestResult
-{
-    LayerTestResult(const armnn::TensorInfo& outputInfo)
-    {
-        auto shape( GetTensorShapeAsArray<n>(outputInfo) );
-        output.resize(shape);
-        outputExpected.resize(shape);
-        supported = true;
-        compareBoolean = false;
-    }
-
-    boost::multi_array<T, n> output;
-    boost::multi_array<T, n> outputExpected;
-    bool supported;
-    bool compareBoolean;
-};
-
 LayerTestResult<float, 4> SimpleConvolution2d3x5Test(
     armnn::IWorkloadFactory& workloadFactory,
     const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
@@ -727,39 +706,6 @@ LayerTestResult<float, 3> ConcatTest(
     armnn::IWorkloadFactory& workloadFactory,
     const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
 
-LayerTestResult<float, 4> AdditionTest(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<float, 5> Addition5dTest(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<float, 4> AdditionBroadcast1ElementTest(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<float, 4> AdditionBroadcastTest(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<float, 4> CompareAdditionTest(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
-    armnn::IWorkloadFactory& refWorkloadFactory);
-
-LayerTestResult<float, 4> SubtractionTest(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<float, 4> SubtractionBroadcast1ElementTest(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<float, 4> SubtractionBroadcastTest(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
 LayerTestResult<float, 4> CompareActivationTest(
     armnn::IWorkloadFactory& workloadFactory,
     const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
@@ -767,43 +713,6 @@ LayerTestResult<float, 4> CompareActivationTest(
     armnn::ActivationFunction f,
     unsigned int batchSize);
 
-LayerTestResult<float, 4> DivisionTest(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<float, 4> DivisionByZeroTest(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<float, 4> DivisionBroadcast1ElementTest(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<float, 4> DivisionBroadcast1DVectorTest(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<float, 4> MultiplicationTest(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<float, 5> Multiplication5dTest(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<float, 4> MultiplicationBroadcast1ElementTest(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<float, 4> MultiplicationBroadcast1DVectorTest(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<float, 4> CompareMultiplicationTest(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
-    armnn::IWorkloadFactory& refWorkloadFactory);
-
 LayerTestResult<float, 4> BatchNormTest(
     armnn::IWorkloadFactory& workloadFactory,
     const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
@@ -1177,54 +1086,6 @@ LayerTestResult<uint8_t, 3> ConcatUint8DifferentQParamsTest(
     armnn::IWorkloadFactory& workloadFactory,
     const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
 
-LayerTestResult<uint8_t, 4> AdditionUint8Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<uint8_t, 4> AdditionBroadcast1ElementUint8Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<uint8_t, 4> AdditionBroadcastUint8Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<int16_t, 4> AdditionInt16Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<int16_t, 4> AdditionBroadcastInt16Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<int16_t, 4> AdditionBroadcast1ElementInt16Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<uint8_t, 4> SubtractionUint8Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<uint8_t, 4> SubtractionBroadcast1ElementUint8Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<uint8_t, 4> SubtractionBroadcastUint8Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<int16_t, 4> SubtractionInt16Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<int16_t, 4> SubtractionBroadcast1ElementInt16Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<int16_t, 4> SubtractionBroadcastInt16Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
 LayerTestResult<uint8_t, 4> CompareActivationUint8Test(
     armnn::IWorkloadFactory& workloadFactory,
     const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
@@ -1243,54 +1104,6 @@ LayerTestResult<uint8_t, 2> CompareSoftmaxUint8Test(
     armnn::IWorkloadFactory& refWorkloadFactory,
     float beta);
 
-LayerTestResult<uint8_t, 4> MultiplicationUint8Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<uint8_t, 4> MultiplicationBroadcast1ElementUint8Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<uint8_t, 4> MultiplicationBroadcast1DVectorUint8Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<int16_t, 4> MultiplicationInt16Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<int16_t, 4> MultiplicationBroadcast1ElementInt16Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<int16_t, 4> MultiplicationBroadcast1DVectorInt16Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<uint8_t, 4> DivisionUint8Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<uint8_t, 4> DivisionBroadcast1ElementUint8Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<uint8_t, 4> DivisionBroadcast1DVectorUint8Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<int16_t, 4> DivisionInt16Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<int16_t, 4> DivisionBroadcast1ElementInt16Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<int16_t, 4> DivisionBroadcast1DVectorInt16Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
 LayerTestResult<uint8_t, 4> SimpleConvolution2d3x5Uint8Test(
     armnn::IWorkloadFactory& workloadFactory,
     const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
@@ -1417,54 +1230,6 @@ LayerTestResult<uint8_t, 3> Concatenation3dDim2DiffInputDimsUint8Test(
     const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
     bool useSubtensor);
 
-LayerTestResult<uint8_t, 4> EqualSimpleTest(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<uint8_t, 4> EqualBroadcast1ElementTest(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<uint8_t, 4> EqualBroadcast1DVectorTest(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<uint8_t, 4> EqualUint8Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<uint8_t, 4> EqualBroadcast1ElementUint8Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<uint8_t, 4> EqualBroadcast1DVectorUint8Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<uint8_t, 4> GreaterSimpleTest(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<uint8_t, 4> GreaterBroadcast1ElementTest(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<uint8_t, 4> GreaterBroadcast1DVectorTest(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<uint8_t, 4> GreaterUint8Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<uint8_t, 4> GreaterBroadcast1ElementUint8Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<uint8_t, 4> GreaterBroadcast1DVectorUint8Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
 LayerTestResult<float, 2> FullyConnectedLargeTest(
     armnn::IWorkloadFactory& workloadFactory,
     const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
@@ -1647,34 +1412,6 @@ LayerTestResult<T, 3> MeanVts3Test(
     armnn::IWorkloadFactory& workloadFactory,
     const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
 
-LayerTestResult<float, 4> MinimumBroadcast1ElementTest1(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<float, 4> MinimumBroadcast1ElementTest2(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<uint8_t, 4> MinimumBroadcast1DVectorUint8Test(
-    armnn::IWorkloadFactory & workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr & memoryManager);
-
-LayerTestResult<int16_t , 4> MinimumInt16Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<int16_t, 4> MinimumBroadcast1ElementInt16Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<int16_t, 4> MinimumBroadcast1DVectorInt16Test(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
-LayerTestResult<float, 4> AdditionAfterMaxPoolTest(
-    armnn::IWorkloadFactory& workloadFactory,
-    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
-
 LayerTestResult<float, 4> SpaceToBatchNdSimpleFloat32Test(
     armnn::IWorkloadFactory& workloadFactory,
     const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
diff --git a/src/backends/backendsCommon/test/layerTests/AdditionTestImpl.cpp b/src/backends/backendsCommon/test/layerTests/AdditionTestImpl.cpp
new file mode 100644
index 0000000000..c6d3982f92
--- /dev/null
+++ b/src/backends/backendsCommon/test/layerTests/AdditionTestImpl.cpp
@@ -0,0 +1,617 @@
+//
+// Copyright © 2017 Arm Ltd. All rights reserved.
+// SPDX-License-Identifier: MIT
+//
+
+#include "AdditionTestImpl.hpp"
+
+#include "ElementwiseTestImpl.hpp"
+
+template<>
+std::unique_ptr<armnn::IWorkload> CreateWorkload<armnn::AdditionQueueDescriptor>(
+    const armnn::IWorkloadFactory& workloadFactory,
+    const armnn::WorkloadInfo& info,
+    const armnn::AdditionQueueDescriptor& descriptor)
+{
+    return workloadFactory.CreateAddition(descriptor, info);
+}
+
+LayerTestResult<float,4> AdditionTest(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    unsigned int batchSize = 2u;
+    unsigned int channels  = 2u;
+    unsigned int height    = 2u;
+    unsigned int width     = 3u;
+
+    unsigned int shape[] = { batchSize, channels, height, width };
+
+    std::vector<float> input1 =
+    {
+        0.0f, 2.0f, 1.0f,
+        0.2f, 1.0f, 2.0f,
+
+        1.0f, 2.0f, 1.0f,
+        0.2f, 1.0f, 2.0f,
+
+        0.0f, 2.0f, 1.0f,
+        4.2f, 1.0f, 2.0f,
+
+        0.0f, 0.0f, 1.0f,
+        0.2f, 1.0f, 2.0f,
+    };
+
+    std::vector<float> input2 =
+    {
+        1.0f, 2.0f,  1.0f,
+        0.0f, 1.0f,  2.0f,
+
+        1.0f, 2.0f, -2.0f,
+        0.2f, 1.0f,  2.0f,
+
+        0.0f, 2.0f,  1.0f,
+        4.2f, 0.0f, -3.0f,
+
+        0.0f, 0.0f,  1.0f,
+        0.7f, 1.0f,  5.0f,
+    };
+
+
+    std::vector<float> output
+    {
+        1.0f, 4.0f,  2.0f,
+        0.2f, 2.0f,  4.0f,
+
+        2.0f, 4.0f, -1.0f,
+        0.4f, 2.0f,  4.0f,
+
+        0.0f, 4.0f,  2.0f,
+        8.4f, 1.0f, -1.0f,
+
+        0.0f, 0.0f,  2.0f,
+        0.9f, 2.0f,  7.0f,
+    };
+
+    return ElementwiseTestHelper<4, armnn::AdditionQueueDescriptor, armnn::DataType::Float32>(
+        workloadFactory,
+        memoryManager,
+        shape,
+        input1,
+        shape,
+        input2,
+        shape,
+        output);
+}
+
+LayerTestResult<float, 5> Addition5dTest(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    unsigned int depth     = 2u;
+    unsigned int batchSize = 2u;
+    unsigned int channels  = 2u;
+    unsigned int height    = 2u;
+    unsigned int width     = 3u;
+
+    unsigned int shape[] = { depth, batchSize, channels, height, width };
+
+    std::vector<float> input1 =
+    {
+        2.6f, 4.0f, 4.4f,  2.7f, 4.6f, 2.8f,
+        2.3f, 1.9f, 3.4f,  2.9f, 2.2f, 4.5f,
+
+        2.8f, 1.9f, 2.3f,  2.6f, 4.7f, 3.5f,
+        0.4f, 1.5f, 2.1f,  0.7f, 5.0f, 1.1f,
+
+
+        1.0f, 2.7f, 0.0f,  0.6f, 0.8f, 0.9f,
+        1.0f, 2.6f, 0.4f,  3.8f, 0.4f, 0.8f,
+
+        0.5f, 4.3f, 3.1f,  4.4f, 0.7f, 1.4f,
+        0.4f, 4.4f, 0.7f,  0.6f, 4.7f, 1.2f,
+
+    };
+
+    std::vector<float> input2 =
+    {
+        4.4f, 3.0f, 1.0f,  0.0f, 3.9f, 3.1f,
+        1.7f, 2.9f, 1.3f,  0.4f, 0.4f, 4.3f,
+
+        4.5f, 0.2f, 2.2f,  4.1f, 3.9f, 3.0f,
+        0.1f, 2.5f, 4.1f,  4.6f, 1.5f, 0.0f,
+
+
+        0.5f, 4.9f, 2.5f,  1.5f, 3.4f, 4.5f,
+        2.0f, 3.0f, 4.9f,  1.6f, 2.4f, 3.4f,
+
+        3.6f, 1.8f, 1.3f,  2.6f, 2.1f, 4.8f,
+        2.0f, 4.3f, 4.0f,  0.2f, 0.6f, 4.4f,
+    };
+
+    std::vector<float> output =
+    {
+        7.0f, 7.0f, 5.4f,  2.7f, 8.5f, 5.9f,
+        4.0f, 4.8f, 4.7f,  3.3f, 2.6f, 8.8f,
+
+        7.3f, 2.1f, 4.5f,  6.7f, 8.6f, 6.5f,
+        0.5f, 4.0f, 6.2f,  5.3f, 6.5f, 1.1f,
+
+
+        1.5f, 7.6f, 2.5f,  2.1f, 4.2f, 5.4f,
+        3.0f, 5.6f, 5.3f,  5.4f, 2.8f, 4.2f,
+
+        4.1f, 6.1f, 4.4f,  7.0f, 2.8f, 6.2f,
+        2.4f, 8.7f, 4.7f,  0.8f, 5.3f, 5.6f,
+    };
+
+    return ElementwiseTestHelper<5, armnn::AdditionQueueDescriptor, armnn::DataType::Float32>(
+        workloadFactory,
+        memoryManager,
+        shape,
+        input1,
+        shape,
+        input2,
+        shape,
+        output);
+}
+
+template<armnn::DataType ArmnnType, typename T = armnn::ResolveType<ArmnnType>>
+LayerTestResult<T, 4> AdditionBroadcastTestImpl(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
+    float qScale,
+    int32_t qOffset)
+{
+    armnn::TensorInfo inputTensorInfo1 = armnn::TensorInfo({1, 3, 2, 1}, ArmnnType);
+    armnn::TensorInfo inputTensorInfo2 = armnn::TensorInfo({1, 1, 2, 3}, ArmnnType);
+    armnn::TensorInfo outputTensorInfo = armnn::TensorInfo({1, 3, 2, 3}, ArmnnType);
+
+    if (armnn::IsQuantizedType<T>())
+    {
+        inputTensorInfo1.SetQuantizationScale(qScale);
+        inputTensorInfo1.SetQuantizationOffset(qOffset);
+        inputTensorInfo2.SetQuantizationScale(qScale);
+        inputTensorInfo2.SetQuantizationOffset(qOffset);
+        outputTensorInfo.SetQuantizationScale(qScale);
+        outputTensorInfo.SetQuantizationOffset(qOffset);
+    }
+
+    auto input1 = MakeTensor<T, 4>(inputTensorInfo1, QuantizedVector<T>(qScale, qOffset,
+    {
+        0.0f,
+        1.0f,
+
+        2.0f,
+        3.0f,
+
+        4.0f,
+        5.0f,
+    }));
+
+    auto input2 = MakeTensor<T, 4>(inputTensorInfo2, QuantizedVector<T>(qScale, qOffset,
+    {
+        0.5f, 1.5f, 2.5f,
+        3.5f, 4.5f, 5.5f,
+    }));
+
+    LayerTestResult<T,4> ret(outputTensorInfo);
+    ret.outputExpected = MakeTensor<T, 4>(outputTensorInfo, QuantizedVector<T>(qScale, qOffset,
+    {
+        0.5f, 1.5f, 2.5f,
+        4.5f, 5.5f, 6.5f,
+
+        2.5f, 3.5f, 4.5f,
+        6.5f, 7.5f, 8.5f,
+
+        4.5f, 5.5f, 6.5f,
+        8.5f, 9.5f, 10.5f,
+    }));
+
+    std::unique_ptr<armnn::ITensorHandle> inputHandle1 = workloadFactory.CreateTensorHandle(inputTensorInfo1);
+    std::unique_ptr<armnn::ITensorHandle> inputHandle2 = workloadFactory.CreateTensorHandle(inputTensorInfo2);
+    std::unique_ptr<armnn::ITensorHandle> outputHandle = workloadFactory.CreateTensorHandle(outputTensorInfo);
+
+    armnn::AdditionQueueDescriptor data;
+    armnn::WorkloadInfo info;
+    AddInputToWorkload(data, info, inputTensorInfo1, inputHandle1.get());
+    AddInputToWorkload(data, info, inputTensorInfo2, inputHandle2.get());
+    AddOutputToWorkload(data, info, outputTensorInfo, outputHandle.get());
+
+    std::unique_ptr<armnn::IWorkload> workload = workloadFactory.CreateAddition(data, info);
+
+    inputHandle1->Allocate();
+    inputHandle2->Allocate();
+    outputHandle->Allocate();
+
+    CopyDataToITensorHandle(inputHandle1.get(), &input1[0][0][0][0]);
+    CopyDataToITensorHandle(inputHandle2.get(), &input2[0][0][0][0]);
+
+    workload->PostAllocationConfigure();
+    workload->Execute();
+
+    CopyDataFromITensorHandle(&ret.output[0][0][0][0], outputHandle.get());
+
+    return ret;
+}
+
+template<armnn::DataType ArmnnType, typename T = armnn::ResolveType<ArmnnType>>
+LayerTestResult<T, 4> AdditionBroadcast1ElementTestImpl(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
+    float qScale,
+    int32_t qOffset)
+{
+    armnn::TensorInfo inputTensorInfo1 = armnn::TensorInfo({1, 3, 2, 3}, ArmnnType);
+    armnn::TensorInfo inputTensorInfo2 = armnn::TensorInfo({1, 1, 1, 1}, ArmnnType);
+    armnn::TensorInfo outputTensorInfo = armnn::TensorInfo({1, 3, 2, 3}, ArmnnType);
+
+    if (armnn::IsQuantizedType<T>())
+    {
+        inputTensorInfo1.SetQuantizationScale(qScale);
+        inputTensorInfo1.SetQuantizationOffset(qOffset);
+        inputTensorInfo2.SetQuantizationScale(qScale);
+        inputTensorInfo2.SetQuantizationOffset(qOffset);
+        outputTensorInfo.SetQuantizationScale(qScale);
+        outputTensorInfo.SetQuantizationOffset(qOffset);
+    }
+
+    auto input1 = MakeTensor<T, 4>(inputTensorInfo1, QuantizedVector<T>(qScale, qOffset,
+    {
+            0.0f,  1.0f,  2.0f,
+            3.0f,  4.0f,  5.0f,
+            6.0f,  7.0f,  8.0f,
+            9.0f, 10.0f, 11.0f,
+        12.0f, 13.0f, 14.0f,
+        15.0f, 16.0f, 17.0f,
+    }));
+
+    auto input2 = MakeTensor<T, 4>(inputTensorInfo2, QuantizedVector<T>(qScale, qOffset,
+    {
+        0.5f,
+    }));
+
+    LayerTestResult<T,4> ret(outputTensorInfo);
+    ret.outputExpected = MakeTensor<T, 4>(outputTensorInfo, QuantizedVector<T>(qScale, qOffset,
+    {
+            0.5f,  1.5f,  2.5f,
+            3.5f,  4.5f,  5.5f,
+            6.5f,  7.5f,  8.5f,
+            9.5f, 10.5f, 11.5f,
+        12.5f, 13.5f, 14.5f,
+        15.5f, 16.5f, 17.5f,
+    }));
+
+    std::unique_ptr<armnn::ITensorHandle> inputHandle1 = workloadFactory.CreateTensorHandle(inputTensorInfo1);
+    std::unique_ptr<armnn::ITensorHandle> inputHandle2 = workloadFactory.CreateTensorHandle(inputTensorInfo2);
+    std::unique_ptr<armnn::ITensorHandle> outputHandle = workloadFactory.CreateTensorHandle(outputTensorInfo);
+
+    armnn::AdditionQueueDescriptor data;
+    armnn::WorkloadInfo info;
+    AddInputToWorkload(data, info, inputTensorInfo1, inputHandle1.get());
+    AddInputToWorkload(data, info, inputTensorInfo2, inputHandle2.get());
+    AddOutputToWorkload(data, info, outputTensorInfo, outputHandle.get());
+
+    std::unique_ptr<armnn::IWorkload> workload = workloadFactory.CreateAddition(data, info);
+
+    inputHandle1->Allocate();
+    inputHandle2->Allocate();
+    outputHandle->Allocate();
+
+    CopyDataToITensorHandle(inputHandle1.get(), &input1[0][0][0][0]);
+    CopyDataToITensorHandle(inputHandle2.get(), &input2[0][0][0][0]);
+
+    workload->PostAllocationConfigure();
+    workload->Execute();
+
+    CopyDataFromITensorHandle(&ret.output[0][0][0][0], outputHandle.get());
+
+    return ret;
+}
+
+LayerTestResult<float, 4> AdditionBroadcastTest(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    return AdditionBroadcastTestImpl<armnn::DataType::Float32>(
+        workloadFactory, memoryManager, 0.0f, 0);
+}
+
+LayerTestResult<uint8_t, 4> AdditionBroadcastUint8Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    return AdditionBroadcastTestImpl<armnn::DataType::QuantisedAsymm8>(
+        workloadFactory, memoryManager, 2.f, 0);
+}
+
+LayerTestResult<int16_t, 4> AdditionBroadcastInt16Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    return AdditionBroadcastTestImpl<armnn::DataType::QuantisedSymm16>(
+        workloadFactory, memoryManager, 2.f, 0);
+}
+
+LayerTestResult<float, 4> AdditionBroadcast1ElementTest(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    return AdditionBroadcast1ElementTestImpl<armnn::DataType::Float32>(
+        workloadFactory, memoryManager, 0.0f, 0);
+}
+
+LayerTestResult<uint8_t, 4> AdditionBroadcast1ElementUint8Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    return AdditionBroadcast1ElementTestImpl<armnn::DataType::QuantisedAsymm8>(
+        workloadFactory, memoryManager, 0.1333333f, 128);
+}
+
+LayerTestResult<int16_t, 4> AdditionBroadcast1ElementInt16Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    return AdditionBroadcast1ElementTestImpl<armnn::DataType::QuantisedSymm16>(
+        workloadFactory, memoryManager, 0.1333333f, 0);
+}
+
+LayerTestResult<uint8_t, 4> AdditionUint8Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    const unsigned int shape0[] = { 1, 2, 2, 3 };
+    const unsigned int shape1[] = { 1, 2, 2, 3 };
+
+    std::vector<uint8_t> input0(
+    {
+        63,  35,  77,  70,  56, 112, //  420, 224,  518,  469,  371, 763
+        203,  28, 252, 168, 245,  91  // 1400, 175, 1743, 1155, 1694, 616
+    });
+
+    std::vector<uint8_t> input1(
+    {
+        21,   7, 175, 231, 175, 210, // 126,   28, 1204, 1596, 1204, 1449
+        126, 161,  63,  21, 105, 126  // 861, 1106,  420,  126,  714,  861
+    });
+
+    std::vector<uint8_t> output(
+    {
+        81,  39, 249, 255, 228, 255, //  546,  252, 1722, 2065(clamped), 1575, 2212(clamped)
+        255, 186, 255, 186, 255, 214, // 2261(clamped), 1281, 2163(clamped), 1281, 2408(clamped), 1477
+    });
+
+    return ElementwiseTestHelper<4, armnn::AdditionQueueDescriptor, armnn::DataType::QuantisedAsymm8>(
+        workloadFactory,
+        memoryManager,
+        shape0,
+        input0,
+        7.0f,
+        3,
+        shape1,
+        input1,
+        7.0f,
+        3,
+        shape0,
+        output,
+        7.0f,
+        3);
+}
+
+LayerTestResult<int16_t, 4> AdditionInt16Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    const unsigned int shape0[] = { 1, 2, 2, 3 };
+    const unsigned int shape1[] = { 1, 2, 2, 3 };
+
+    std::vector<int16_t> input0 =
+    {
+        63,  35,  77,  70,  56, 112, //  441, 245,  539,  490,  392, 184
+        203,  28, 252, 168, 245,  91  // 1421, 196, 1764, 1176, 1715, 637
+    };
+
+    std::vector<int16_t> input1 =
+    {
+        21,   7, 175, 231, 175, 210, // 126,   28, 1204, 1596, 1204, 1449
+        126, 161,  63,  21, 105, 126  // 861, 1106,  420,  126,  714,  861
+    };
+
+    std::vector<int16_t> output =
+    {
+        84,  42, 252, 301, 231, 322, //  588,  294, 1764, 2107(clamped), 1617, 2254(clamped)
+        329, 189, 315, 189, 350, 217, // 2303(clamped), 1323, 2205(clamped), 1323, 2450(clamped), 1519
+    };
+
+    return ElementwiseTestHelper<4, armnn::AdditionQueueDescriptor, armnn::DataType::QuantisedSymm16>(
+        workloadFactory,
+        memoryManager,
+        shape0,
+        input0,
+        7.0f,
+        0,
+        shape1,
+        input1,
+        7.0f,
+        0,
+        shape0,
+        output,
+        7.0f,
+        0);
+}
+
+LayerTestResult<float, 4> AdditionAfterMaxPoolTest(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    // Create Initial Tensor
+    // 1, 2, 3
+    // 4, 5, 6
+    // 7, 8, 9
+
+    armnn::TensorInfo poolingInputTensorInfo({ 1, 1, 3, 3}, armnn::DataType::Float32);
+    armnn::TensorInfo poolingOutputTensorInfo({ 1, 1, 2, 2}, armnn::DataType::Float32);
+
+    boost::multi_array<float, 4> poolingInput = MakeTensor<float,4>(poolingInputTensorInfo,
+                                                            {1, 2, 3,
+                                                             4, 5, 6,
+                                                             7, 8, 9
+                                                            });
+
+    std::unique_ptr<armnn::ITensorHandle> poolingInputHandle =
+            workloadFactory.CreateTensorHandle(poolingInputTensorInfo);
+    std::unique_ptr<armnn::ITensorHandle> poolingOutputHandle =
+            workloadFactory.CreateTensorHandle(poolingOutputTensorInfo);
+
+    // Apply MaxPool poolSize = 1x1, stride=2x2
+    // Result =
+    // 1, 3
+    // 7, 9
+    armnn::Pooling2dDescriptor descriptor;
+    descriptor.m_PoolHeight = 1;
+    descriptor.m_PoolWidth = 1;
+    descriptor.m_StrideX = 2;
+    descriptor.m_StrideY = 2;
+    descriptor.m_PoolType = armnn::PoolingAlgorithm::Max;
+
+    armnn::Pooling2dQueueDescriptor queueDescriptor;
+    queueDescriptor.m_Parameters = descriptor;
+    armnn::WorkloadInfo workloadInfo;
+    AddInputToWorkload(queueDescriptor, workloadInfo, poolingInputTensorInfo, poolingInputHandle.get());
+    AddOutputToWorkload(queueDescriptor, workloadInfo, poolingOutputTensorInfo, poolingOutputHandle.get());
+
+    // Create the MaxPool
+    std::unique_ptr<armnn::IWorkload> workload = workloadFactory.CreatePooling2d(queueDescriptor, workloadInfo);
+
+    //LayerTestResult<float, 4> result(poolingOutputTensorInfo);
+    auto shape( GetTensorShapeAsArray<4>(poolingOutputTensorInfo));
+    boost::multi_array<float, 4> resultMaxPool;
+    resultMaxPool.resize(shape);
+
+
+    // Create addition with another tensor the same size
+    // This would be the result to apply a Conv2d with kernel ones(2) and stride 1x1
+    // with the initial tensor.
+    // 12, 16
+    // 24, 28
+
+    armnn::TensorInfo addInputTensorInfo({ 1,1,2,2}, armnn::DataType::Float32);
+    armnn::TensorInfo addOutputTensorInfo({ 1,1,2,2}, armnn::DataType::Float32);
+
+    boost::multi_array<float, 4> addInput = MakeTensor<float,4>(addInputTensorInfo,
+                                                                    {12, 16,
+                                                                     24, 28,
+                                                                    });
+
+    // Expected output tensor after MaxPool and Addition.
+    LayerTestResult<float,4> addRet(addOutputTensorInfo);
+    addRet.outputExpected = MakeTensor<float, 4>(addOutputTensorInfo, std::vector<float>(
+    {
+        13, 19,
+        31, 37
+    }));
+
+    std::unique_ptr<armnn::ITensorHandle> addInputHandle = workloadFactory.CreateTensorHandle(addInputTensorInfo);
+    std::unique_ptr<armnn::ITensorHandle> addOutputHandle = workloadFactory.CreateTensorHandle(addOutputTensorInfo);
+
+    armnn::AdditionQueueDescriptor data;
+    armnn::WorkloadInfo info;
+
+    // Add the output of the MaxPool and the new tensor
+    AddInputToWorkload(data, info, poolingOutputTensorInfo, poolingOutputHandle.get());
+    AddInputToWorkload(data, info, addInputTensorInfo, addInputHandle.get());
+    AddOutputToWorkload(data, info, addOutputTensorInfo, addOutputHandle.get());
+
+    std::unique_ptr<armnn::IWorkload> addWorkload = workloadFactory.CreateAddition(data, info);
+
+    poolingInputHandle->Allocate();
+    poolingOutputHandle->Allocate();
+    addInputHandle->Allocate();
+    addOutputHandle->Allocate();
+
+    CopyDataToITensorHandle(poolingInputHandle.get(), &poolingInput[0][0][0][0]);
+    CopyDataFromITensorHandle(&resultMaxPool[0][0][0][0], poolingOutputHandle.get());
+
+    CopyDataToITensorHandle(poolingOutputHandle.get(), &resultMaxPool[0][0][0][0]);
+    CopyDataToITensorHandle(addInputHandle.get(), &addInput[0][0][0][0]);
+
+    workload->PostAllocationConfigure();
+    workload->Execute();
+    addWorkload->PostAllocationConfigure();
+    addWorkload->Execute();
+
+    CopyDataFromITensorHandle(&addRet.output[0][0][0][0], addOutputHandle.get());
+
+    return addRet;
+}
+
+LayerTestResult<float,4> CompareAdditionTest(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
+    armnn::IWorkloadFactory& refWorkloadFactory)
+{
+    unsigned int batchSize = 4;
+    unsigned int channels  = 1;
+    unsigned int height    = 2;
+    unsigned int width     = 3;
+
+    armnn::TensorInfo inputTensorInfo1, inputTensorInfo2;
+    armnn::TensorInfo outputTensorInfo;
+
+    unsigned int shape[] = {batchSize, channels, height, width};
+
+    inputTensorInfo1 = armnn::TensorInfo(4, shape, armnn::DataType::Float32);
+    inputTensorInfo2 = armnn::TensorInfo(4, shape, armnn::DataType::Float32);
+    outputTensorInfo = armnn::TensorInfo(4, shape, armnn::DataType::Float32);
+
+    auto input1 = MakeRandomTensor<float, 4>(inputTensorInfo1, 1232);
+    auto input2 = MakeRandomTensor<float, 4>(inputTensorInfo2, 456);
+
+    LayerTestResult<float,4> ret(outputTensorInfo);
+
+    std::unique_ptr<armnn::ITensorHandle> inputHandle1 = workloadFactory.CreateTensorHandle(inputTensorInfo1);
+    std::unique_ptr<armnn::ITensorHandle> inputHandle2 = workloadFactory.CreateTensorHandle(inputTensorInfo2);
+    std::unique_ptr<armnn::ITensorHandle> outputHandle = workloadFactory.CreateTensorHandle(outputTensorInfo);
+
+    std::unique_ptr<armnn::ITensorHandle> inputHandle1Ref = refWorkloadFactory.CreateTensorHandle(inputTensorInfo1);
+    std::unique_ptr<armnn::ITensorHandle> inputHandle2Ref = refWorkloadFactory.CreateTensorHandle(inputTensorInfo2);
+    std::unique_ptr<armnn::ITensorHandle> outputHandleRef = refWorkloadFactory.CreateTensorHandle(outputTensorInfo);
+
+    armnn::AdditionQueueDescriptor data;
+    armnn::WorkloadInfo info;
+    AddInputToWorkload(data, info, inputTensorInfo1, inputHandle1.get());
+    AddInputToWorkload(data, info, inputTensorInfo2, inputHandle2.get());
+    AddOutputToWorkload(data, info, outputTensorInfo, outputHandle.get());
+
+    armnn::AdditionQueueDescriptor refData = data;
+    armnn::WorkloadInfo refInfo = info;
+    SetWorkloadInput(refData, refInfo, 0, inputTensorInfo1, inputHandle1Ref.get());
+    SetWorkloadInput(refData, refInfo, 1, inputTensorInfo2, inputHandle2Ref.get());
+    SetWorkloadOutput(refData, refInfo, 0, outputTensorInfo, outputHandleRef.get());
+
+    std::unique_ptr<armnn::IWorkload> workload = workloadFactory.CreateAddition(data, info);
+    std::unique_ptr<armnn::IWorkload> workloadRef = refWorkloadFactory.CreateAddition(refData, refInfo);
+
+    inputHandle1->Allocate();
+    inputHandle2->Allocate();
+    outputHandle->Allocate();
+    inputHandle1Ref->Allocate();
+    inputHandle2Ref->Allocate();
+    outputHandleRef->Allocate();
+
+    CopyDataToITensorHandle(inputHandle1.get(), &input1[0][0][0][0]);
+    CopyDataToITensorHandle(inputHandle2.get(), &input2[0][0][0][0]);
+    CopyDataToITensorHandle(inputHandle1Ref.get(), &input1[0][0][0][0]);
+    CopyDataToITensorHandle(inputHandle2Ref.get(), &input2[0][0][0][0]);
+
+    workload->PostAllocationConfigure();
+    workload->Execute();
+    workloadRef->PostAllocationConfigure();
+    workloadRef->Execute();
+
+    CopyDataFromITensorHandle(&ret.output[0][0][0][0], outputHandle.get());
+    CopyDataFromITensorHandle(&ret.outputExpected[0][0][0][0], outputHandleRef.get());
+
+    return ret;
+}
diff --git a/src/backends/backendsCommon/test/layerTests/AdditionTestImpl.hpp b/src/backends/backendsCommon/test/layerTests/AdditionTestImpl.hpp
new file mode 100644
index 0000000000..60ef975563
--- /dev/null
+++ b/src/backends/backendsCommon/test/layerTests/AdditionTestImpl.hpp
@@ -0,0 +1,60 @@
+//
+// Copyright © 2017 Arm Ltd. All rights reserved.
+// SPDX-License-Identifier: MIT
+//
+
+#pragma once
+
+#include "LayerTestResult.hpp"
+
+#include <backendsCommon/IBackendInternal.hpp>
+#include <backendsCommon/WorkloadFactory.hpp>
+
+LayerTestResult<float, 4> AdditionTest(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<float, 5> Addition5dTest(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<float, 4> AdditionBroadcast1ElementTest(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<float, 4> AdditionBroadcastTest(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<uint8_t, 4> AdditionUint8Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<uint8_t, 4> AdditionBroadcast1ElementUint8Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<uint8_t, 4> AdditionBroadcastUint8Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<int16_t, 4> AdditionInt16Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<int16_t, 4> AdditionBroadcastInt16Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<int16_t, 4> AdditionBroadcast1ElementInt16Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<float, 4> AdditionAfterMaxPoolTest(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<float, 4> CompareAdditionTest(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
+    armnn::IWorkloadFactory& refWorkloadFactory);
diff --git a/src/backends/backendsCommon/test/layerTests/DivisionTestImpl.cpp b/src/backends/backendsCommon/test/layerTests/DivisionTestImpl.cpp
new file mode 100644
index 0000000000..0316ea185b
--- /dev/null
+++ b/src/backends/backendsCommon/test/layerTests/DivisionTestImpl.cpp
@@ -0,0 +1,350 @@
+//
+// Copyright © 2017 Arm Ltd. All rights reserved.
+// SPDX-License-Identifier: MIT
+//
+
+#include "DivisionTestImpl.hpp"
+
+#include "ElementwiseTestImpl.hpp"
+
+template<>
+std::unique_ptr<armnn::IWorkload> CreateWorkload<armnn::DivisionQueueDescriptor>(
+    const armnn::IWorkloadFactory& workloadFactory,
+    const armnn::WorkloadInfo& info,
+    const armnn::DivisionQueueDescriptor& descriptor)
+{
+    return workloadFactory.CreateDivision(descriptor, info);
+}
+
+LayerTestResult<float, 4> DivisionByZeroTest(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    const unsigned int width        = 2u;
+    const unsigned int height       = 2u;
+    const unsigned int channelCount = 2u;
+    const unsigned int batchSize    = 2u;
+
+    unsigned int shape[] = { batchSize, channelCount, height, width };
+
+    std::vector<float> input0 =
+    {
+         1.f,  1.f,  1.f,  1.f,  0.f, 0.f, 0.f, 0.f,
+        -1.f, -1.f, -1.f, -1.f,  5.f, 5.f, 5.f, 5.f
+    };
+
+    std::vector<float> input1 =
+    {
+        0.f, 0.f, -0.f, -0.f,  0.f, 0.f, -0.f, -0.f,
+        0.f, 0.f, -0.f, -0.f,  5.f, 5.f,  5.f,  5.f
+    };
+
+    std::vector<float> output =
+    {
+         INFINITY,  INFINITY, -INFINITY, -INFINITY,  NAN,  NAN, -NAN, -NAN,
+        -INFINITY, -INFINITY,  INFINITY,  INFINITY,    1,    1,    1,    1
+    };
+
+    return ElementwiseTestHelper<4, armnn::DivisionQueueDescriptor, armnn::DataType::Float32>(
+        workloadFactory,
+        memoryManager,
+        shape,
+        input0,
+        shape,
+        input1,
+        shape,
+        output);
+}
+
+LayerTestResult<float, 4> DivisionTest(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    const unsigned int width        = 2u;
+    const unsigned int height       = 2u;
+    const unsigned int channelCount = 2u;
+    const unsigned int batchSize    = 2u;
+
+    unsigned int shape[] = { batchSize, channelCount, height, width };
+
+    std::vector<float> input0 =
+    {
+        2.f, 2.f, 2.f, 2.f, 3.f, 3.f, 3.f, 3.f,
+        4.f, 4.f, 4.f, 4.f, 5.f, 5.f, 5.f, 5.f
+    };
+
+    std::vector<float> input1 =
+    {
+        1.f, 1.f, 1.f, 1.f, 2.f, 2.f, 2.f, 2.f,
+        4.f, 4.f, 4.f, 4.f, 4.f, 4.f, 4.f, 4.f
+    };
+
+    std::vector<float> output =
+    {
+        2.f, 2.f, 2.f, 2.f, 1.50f, 1.50f, 1.50f, 1.50f,
+        1.f, 1.f, 1.f, 1.f, 1.25f, 1.25f, 1.25f, 1.25f
+    };
+
+    return ElementwiseTestHelper<4, armnn::DivisionQueueDescriptor, armnn::DataType::Float32>(
+        workloadFactory,
+        memoryManager,
+        shape,
+        input0,
+        shape,
+        input1,
+        shape,
+        output);
+}
+
+LayerTestResult<float, 4> DivisionBroadcast1ElementTest(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    unsigned int shape0[] = { 1, 2, 2, 2 };
+    unsigned int shape1[] = { 1, 1, 1, 1 };
+
+    std::vector<float> input0({ 2, 4, 6, 8, 10, 12, 14, 16});
+
+    std::vector<float> input1({ 2 });
+
+    std::vector<float> output({ 1, 2, 3, 4, 5, 6, 7, 8});
+
+    return ElementwiseTestHelper<4, armnn::DivisionQueueDescriptor, armnn::DataType::Float32>(
+        workloadFactory,
+        memoryManager,
+        shape0,
+        input0,
+        shape1,
+        input1,
+        shape0,
+        output);
+}
+
+LayerTestResult<float, 4> DivisionBroadcast1DVectorTest(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    unsigned int shape0[] = { 1, 3, 3, 2 };
+    unsigned int shape1[] = { 1, 1, 1, 2 };
+
+    std::vector<float> input0 =
+    {
+         1.f,  4.f,  3.f,  8.f,  5.f, 12.f,
+         7.f, 16.f,  9.f, 20.f, 11.f, 24.f,
+        13.f, 28.f, 15.f, 32.f, 17.f, 36.f
+    };
+
+    std::vector<float> input1 = { 1.f, 2.f };
+
+    std::vector<float> output =
+    {
+         1.f,  2.f,  3.f,  4.f,  5.f,  6.f,
+         7.f,  8.f,  9.f, 10.f, 11.f, 12.f,
+        13.f, 14.f, 15.f, 16.f, 17.f, 18.f
+    };
+
+    return ElementwiseTestHelper<4, armnn::DivisionQueueDescriptor, armnn::DataType::Float32>(
+        workloadFactory,
+        memoryManager,
+        shape0,
+        input0,
+        shape1,
+        input1,
+        shape0,
+        output);
+}
+
+LayerTestResult<uint8_t, 4> DivisionUint8Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    const unsigned int width        = 2u;
+    const unsigned int height       = 2u;
+    const unsigned int channelCount = 2u;
+    const unsigned int batchSize    = 2u;
+
+    unsigned int shape[] = { batchSize, channelCount, height, width };
+
+    std::vector<uint8_t> input0 =
+    {
+        2, 2, 2, 2,  3, 3, 3, 3,
+        4, 4, 4, 4,  5, 5, 5, 5
+    };
+
+    std::vector<uint8_t> input1 =
+    {
+        1, 1, 1, 1,  2, 2, 2, 2,
+        4, 4, 4, 4,  4, 4, 4, 4
+    };
+
+    std::vector<uint8_t> output =
+    {
+        8, 8, 8, 8,  6, 6, 6, 6,
+        4, 4, 4, 4,  5, 5, 5, 5
+    };
+
+    return ElementwiseTestHelper<4, armnn::DivisionQueueDescriptor, armnn::DataType::QuantisedAsymm8>(
+        workloadFactory,
+        memoryManager,
+        shape,
+        input0,
+        shape,
+        input1,
+        shape,
+        output,
+        0.25f,
+        0);
+}
+
+LayerTestResult<uint8_t, 4> DivisionBroadcast1ElementUint8Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    unsigned int shape0[] = { 1, 2, 2, 2 };
+    unsigned int shape1[] = { 1, 1, 1, 1 };
+
+    std::vector<uint8_t> input0 = { 2, 4, 6, 8, 10, 12, 14, 16};
+
+    std::vector<uint8_t> input1 = { 2 };
+
+    std::vector<uint8_t> output = { 1, 2, 3, 4, 5, 6, 7, 8};
+
+    return ElementwiseTestHelper<4, armnn::DivisionQueueDescriptor, armnn::DataType::QuantisedAsymm8>(
+        workloadFactory,
+        memoryManager,
+        shape0,
+        input0,
+        shape1,
+        input1,
+        shape0,
+        output);
+}
+
+LayerTestResult<uint8_t, 4> DivisionBroadcast1DVectorUint8Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    unsigned int shape0[] = { 1, 3, 3, 2 };
+    unsigned int shape1[] = { 1, 1, 1, 2 };
+
+    std::vector<uint8_t> input0 =
+    {
+        1,  4,    3,  8,    5, 12,
+        7, 16,    9, 20,   11, 24,
+       13, 28,   15, 32,   17, 36
+    };
+
+    std::vector<uint8_t> input1 = { 1, 2 };
+
+    std::vector<uint8_t> output =
+    {
+        1,  2,    3,  4,    5,  6,
+        7,  8,    9, 10,   11, 12,
+       13, 14,   15, 16,   17, 18
+    };
+
+    return ElementwiseTestHelper<4, armnn::DivisionQueueDescriptor, armnn::DataType::QuantisedAsymm8>(
+        workloadFactory,
+        memoryManager,
+        shape0,
+        input0,
+        shape1,
+        input1,
+        shape0,
+        output);
+}
+
+LayerTestResult<int16_t,4> DivisionInt16Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    unsigned int shape[] = { 2, 2, 2, 2 };
+
+    std::vector<int16_t> input0 =
+    {
+        2, 2, 2, 2,  3, 3, 3, 3,
+        4, 4, 4, 4,  5, 5, 5, 5
+    };
+
+    std::vector<int16_t> input1 =
+    {
+        1, 1, 1, 1,  2, 2, 2, 2,
+        4, 4, 4, 4,  4, 4, 4, 4
+    };
+
+    std::vector<int16_t> output =
+    {
+        8, 8, 8, 8,  6, 6, 6, 6,
+        4, 4, 4, 4,  5, 5, 5, 5
+    };
+
+    return ElementwiseTestHelper<4, armnn::DivisionQueueDescriptor, armnn::DataType::QuantisedSymm16>(
+        workloadFactory,
+        memoryManager,
+        shape,
+        input0,
+        shape,
+        input1,
+        shape,
+        output,
+        0.25f,
+        0);
+}
+
+LayerTestResult<int16_t, 4> DivisionBroadcast1ElementInt16Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    unsigned int shape0[] = { 1, 2, 2, 2 };
+    unsigned int shape1[] = { 1, 1, 1, 1 };
+
+    std::vector<int16_t> input0 = { 2, 4, 6, 8, 10, 12, 14, 16};
+
+    std::vector<int16_t> input1 = { 2 };
+
+    std::vector<int16_t> output = { 1, 2, 3, 4, 5, 6, 7, 8};
+
+    return ElementwiseTestHelper<4, armnn::DivisionQueueDescriptor, armnn::DataType::QuantisedSymm16>(
+        workloadFactory,
+        memoryManager,
+        shape0,
+        input0,
+        shape1,
+        input1,
+        shape0,
+        output);
+}
+
+LayerTestResult<int16_t, 4> DivisionBroadcast1DVectorInt16Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    unsigned int shape0[] = { 1, 3, 3, 2 };
+    unsigned int shape1[] = { 1, 1, 1, 2 };
+
+    std::vector<int16_t> input0 =
+    {
+         1,  4,    3,  8,    5, 12,
+         7, 16,    9, 20,   11, 24,
+        13, 28,   15, 32,   17, 36
+    };
+
+    std::vector<int16_t> input1 = { 1, 2 };
+
+    std::vector<int16_t> output =
+    {
+         1,  2,    3,  4,    5,  6,
+         7,  8,    9, 10,   11, 12,
+        13, 14,   15, 16,   17, 18
+    };
+
+    return ElementwiseTestHelper<4, armnn::DivisionQueueDescriptor, armnn::DataType::QuantisedSymm16>(
+        workloadFactory,
+        memoryManager,
+        shape0,
+        input0,
+        shape1,
+        input1,
+        shape0,
+        output);
+}
diff --git a/src/backends/backendsCommon/test/layerTests/DivisionTestImpl.hpp b/src/backends/backendsCommon/test/layerTests/DivisionTestImpl.hpp
new file mode 100644
index 0000000000..e06b494b7d
--- /dev/null
+++ b/src/backends/backendsCommon/test/layerTests/DivisionTestImpl.hpp
@@ -0,0 +1,51 @@
+//
+// Copyright © 2017 Arm Ltd. All rights reserved.
+// SPDX-License-Identifier: MIT
+//
+
+#pragma once
+
+#include "LayerTestResult.hpp"
+
+#include <backendsCommon/IBackendInternal.hpp>
+#include <backendsCommon/WorkloadFactory.hpp>
+
+LayerTestResult<float, 4> DivisionByZeroTest(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<float, 4> DivisionTest(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<float, 4> DivisionBroadcast1ElementTest(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<float, 4> DivisionBroadcast1DVectorTest(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<uint8_t, 4> DivisionUint8Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<uint8_t, 4> DivisionBroadcast1ElementUint8Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<uint8_t, 4> DivisionBroadcast1DVectorUint8Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<int16_t,4> DivisionInt16Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<int16_t, 4> DivisionBroadcast1ElementInt16Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<int16_t, 4> DivisionBroadcast1DVectorInt16Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
diff --git a/src/backends/backendsCommon/test/layerTests/ElementwiseTestImpl.hpp b/src/backends/backendsCommon/test/layerTests/ElementwiseTestImpl.hpp
new file mode 100644
index 0000000000..1bb1348d61
--- /dev/null
+++ b/src/backends/backendsCommon/test/layerTests/ElementwiseTestImpl.hpp
@@ -0,0 +1,206 @@
+//
+// Copyright © 2017 Arm Ltd. All rights reserved.
+// SPDX-License-Identifier: MIT
+//
+
+#pragma once
+
+#include "LayerTestResult.hpp"
+
+#include <armnn/ArmNN.hpp>
+
+#include <ResolveType.hpp>
+
+#include <backendsCommon/IBackendInternal.hpp>
+#include <backendsCommon/Workload.hpp>
+#include <backendsCommon/WorkloadData.hpp>
+#include <backendsCommon/WorkloadFactory.hpp>
+
+#include <backendsCommon/test/TensorCopyUtils.hpp>
+#include <backendsCommon/test/WorkloadTestUtils.hpp>
+
+#include <test/TensorHelpers.hpp>
+
+#include <memory>
+
+template<typename DescriptorType>
+std::unique_ptr<armnn::IWorkload> CreateWorkload(
+    const armnn::IWorkloadFactory& workloadFactory,
+    const armnn::WorkloadInfo& info,
+    const DescriptorType& descriptor)
+{
+    return CreateWorkload(workloadFactory, info, descriptor);
+}
+
+template <std::size_t NumDims,
+          typename Descriptor,
+          armnn::DataType ArmnnTypeInput,
+          armnn::DataType ArmnnTypeOutput,
+          typename TInput  = armnn::ResolveType<ArmnnTypeInput>,
+          typename TOutput = armnn::ResolveType<ArmnnTypeOutput>>
+LayerTestResult<TOutput, NumDims> ElementwiseTestHelper(
+    armnn::IWorkloadFactory & workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr & memoryManager,
+    const unsigned int shape0[NumDims],
+    std::vector<TInput> values0,
+    float quantScale0,
+    int quantOffset0,
+    const unsigned int shape1[NumDims],
+    std::vector<TInput> values1,
+    float quantScale1,
+    int quantOffset1,
+    const unsigned int outShape[NumDims],
+    std::vector<TOutput> outValues,
+    float outQuantScale,
+    int outQuantOffset)
+{
+    armnn::TensorInfo inputTensorInfo0{NumDims, shape0, ArmnnTypeInput};
+    armnn::TensorInfo inputTensorInfo1{NumDims, shape1, ArmnnTypeInput};
+    armnn::TensorInfo outputTensorInfo{NumDims, outShape, ArmnnTypeOutput};
+
+    auto input0 = MakeTensor<TInput, NumDims>(inputTensorInfo0, values0);
+    auto input1 = MakeTensor<TInput, NumDims>(inputTensorInfo1, values1);
+
+    inputTensorInfo0.SetQuantizationScale(quantScale0);
+    inputTensorInfo0.SetQuantizationOffset(quantOffset0);
+
+    inputTensorInfo1.SetQuantizationScale(quantScale1);
+    inputTensorInfo1.SetQuantizationOffset(quantOffset1);
+
+    outputTensorInfo.SetQuantizationScale(outQuantScale);
+    outputTensorInfo.SetQuantizationOffset(outQuantOffset);
+
+    LayerTestResult<TOutput, NumDims> ret(outputTensorInfo);
+
+    if(ArmnnTypeOutput == armnn::DataType::Boolean)
+    {
+        ret.compareBoolean = true;
+    }
+
+    std::unique_ptr<armnn::ITensorHandle> inputHandle0 = workloadFactory.CreateTensorHandle(inputTensorInfo0);
+    std::unique_ptr<armnn::ITensorHandle> inputHandle1 = workloadFactory.CreateTensorHandle(inputTensorInfo1);
+    std::unique_ptr<armnn::ITensorHandle> outputHandle = workloadFactory.CreateTensorHandle(outputTensorInfo);
+
+    Descriptor data;
+    armnn::WorkloadInfo info;
+    AddInputToWorkload(data, info, inputTensorInfo0, inputHandle0.get());
+    AddInputToWorkload(data, info, inputTensorInfo1, inputHandle1.get());
+    AddOutputToWorkload(data, info, outputTensorInfo, outputHandle.get());
+    auto workload = CreateWorkload<Descriptor>(workloadFactory, info, data);
+
+    inputHandle0->Allocate();
+    inputHandle1->Allocate();
+    outputHandle->Allocate();
+
+    CopyDataToITensorHandle(inputHandle0.get(), input0.origin());
+    CopyDataToITensorHandle(inputHandle1.get(), input1.origin());
+
+    workload->PostAllocationConfigure();
+    ExecuteWorkload(*workload, memoryManager);
+
+    CopyDataFromITensorHandle(ret.output.origin(), outputHandle.get());
+
+    ret.outputExpected = MakeTensor<TOutput, NumDims>(outputTensorInfo, outValues);
+    return ret;
+}
+
+template <std::size_t NumDims,
+          typename Descriptor,
+          armnn::DataType ArmnnType,
+          typename T = armnn::ResolveType<ArmnnType>>
+LayerTestResult<T, NumDims> ElementwiseTestHelper(
+    armnn::IWorkloadFactory & workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr & memoryManager,
+    const unsigned int shape0[NumDims],
+    std::vector<T> values0,
+    float quantScale0,
+    int quantOffset0,
+    const unsigned int shape1[NumDims],
+    std::vector<T> values1,
+    float quantScale1,
+    int quantOffset1,
+    const unsigned int outShape[NumDims],
+    std::vector<T> outValues,
+    float outQuantScale,
+    int outQuantOffset)
+{
+    return ElementwiseTestHelper<NumDims, Descriptor, ArmnnType, ArmnnType>(
+        workloadFactory,
+        memoryManager,
+        shape0,
+        values0,
+        quantScale0,
+        quantOffset0,
+        shape1,
+        values1,
+        quantScale1,
+        quantOffset1,
+        outShape,
+        outValues,
+        outQuantScale,
+        outQuantOffset);
+}
+
+template <std::size_t NumDims,
+          typename Descriptor,
+          armnn::DataType ArmnnTypeInput,
+          armnn::DataType ArmnnTypeOutput,
+          typename TInput  = armnn::ResolveType<ArmnnTypeInput>,
+          typename TOutput = armnn::ResolveType<ArmnnTypeOutput>>
+LayerTestResult<TOutput, NumDims> ElementwiseTestHelper(
+    armnn::IWorkloadFactory & workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr & memoryManager,
+    const unsigned int shape0[NumDims],
+    std::vector<TInput> values0,
+    const unsigned int shape1[NumDims],
+    std::vector<TInput> values1,
+    const unsigned int outShape[NumDims],
+    std::vector<TOutput> outValues,
+    float quantScale = 1.0f,
+    int quantOffset = 0)
+{
+    return ElementwiseTestHelper<NumDims, Descriptor, ArmnnTypeInput, ArmnnTypeOutput>(
+        workloadFactory,
+        memoryManager,
+        shape0,
+        values0,
+        quantScale,
+        quantOffset,
+        shape1,
+        values1,
+        quantScale,
+        quantOffset,
+        outShape,
+        outValues,
+        quantScale,
+        quantOffset);
+}
+
+template <std::size_t NumDims,
+          typename Descriptor,
+          armnn::DataType ArmnnType,
+          typename T = armnn::ResolveType<ArmnnType>>
+LayerTestResult<T, NumDims> ElementwiseTestHelper(
+    armnn::IWorkloadFactory & workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr & memoryManager,
+    const unsigned int shape0[NumDims],
+    std::vector<T> values0,
+    const unsigned int shape1[NumDims],
+    std::vector<T> values1,
+    const unsigned int outShape[NumDims],
+    std::vector<T> outValues,
+    float quantScale = 1.0f,
+    int quantOffset = 0)
+{
+    return ElementwiseTestHelper<NumDims, Descriptor, ArmnnType, ArmnnType>(
+        workloadFactory,
+        memoryManager,
+        shape0,
+        values0,
+        shape1,
+        values1,
+        outShape,
+        outValues,
+        quantScale,
+        quantOffset);
+}
diff --git a/src/backends/backendsCommon/test/layerTests/EqualTestImpl.cpp b/src/backends/backendsCommon/test/layerTests/EqualTestImpl.cpp
new file mode 100644
index 0000000000..fa72136255
--- /dev/null
+++ b/src/backends/backendsCommon/test/layerTests/EqualTestImpl.cpp
@@ -0,0 +1,187 @@
+//
+// Copyright © 2017 Arm Ltd. All rights reserved.
+// SPDX-License-Identifier: MIT
+//
+
+#include "EqualTestImpl.hpp"
+
+#include "ElementwiseTestImpl.hpp"
+
+template<>
+std::unique_ptr<armnn::IWorkload> CreateWorkload<armnn::EqualQueueDescriptor>(
+    const armnn::IWorkloadFactory& workloadFactory,
+    const armnn::WorkloadInfo& info,
+    const armnn::EqualQueueDescriptor& descriptor)
+{
+    return workloadFactory.CreateEqual(descriptor, info);
+}
+
+LayerTestResult<uint8_t, 4> EqualSimpleTest(armnn::IWorkloadFactory& workloadFactory,
+                                            const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    const unsigned int width        = 2u;
+    const unsigned int height       = 2u;
+    const unsigned int channelCount = 2u;
+    const unsigned int batchSize    = 2u;
+
+    unsigned int shape[] = { batchSize, channelCount, height, width };
+
+    std::vector<float> input0 =
+    {
+        1.f, 1.f, 1.f, 1.f,  5.f, 5.f, 5.f, 5.f,
+        3.f, 3.f, 3.f, 3.f,  4.f, 4.f, 4.f, 4.f
+    };
+
+    std::vector<float> input1({ 1, 1, 1, 1,  3, 3, 3, 3,
+                                5, 5, 5, 5,  4, 4, 4, 4 });
+
+    std::vector<uint8_t> output({ 1, 1, 1, 1,  0, 0, 0, 0,
+                                  0, 0, 0, 0,  1, 1, 1, 1 });
+
+    return ElementwiseTestHelper<4, armnn::EqualQueueDescriptor, armnn::DataType::Float32, armnn::DataType::Boolean>(
+        workloadFactory,
+        memoryManager,
+        shape,
+        input0,
+        shape,
+        input1,
+        shape,
+        output);
+}
+
+LayerTestResult<uint8_t, 4> EqualBroadcast1ElementTest(
+        armnn::IWorkloadFactory& workloadFactory,
+        const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    unsigned int shape0[] = { 1, 2, 2, 2 };
+    std::vector<float> input0({ 1, 2, 3, 4, 5, 6, 7, 8});
+
+    unsigned int shape1[] = { 1, 1, 1, 1 };
+    std::vector<float> input1({ 1 });
+
+    std::vector<uint8_t> output({ 1, 0, 0, 0, 0, 0, 0, 0});
+
+    return ElementwiseTestHelper<4, armnn::EqualQueueDescriptor, armnn::DataType::Float32, armnn::DataType::Boolean>(
+        workloadFactory,
+        memoryManager,
+        shape0,
+        input0,
+        shape1,
+        input1,
+        shape0,
+        output);
+}
+
+LayerTestResult<uint8_t, 4> EqualBroadcast1DVectorTest(
+        armnn::IWorkloadFactory& workloadFactory,
+        const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    const unsigned int shape0[] = { 1, 2, 2, 3 };
+    const unsigned int shape1[] = { 1, 1, 1, 3 };
+
+    std::vector<float> input0({ 1, 2, 3, 4, 5, 6,
+                                7, 8, 9, 10, 11, 12 });
+
+    std::vector<float> input1({ 1, 2, 3});
+
+    std::vector<uint8_t> output({ 1, 1, 1, 0, 0, 0,
+                                  0, 0, 0, 0, 0, 0 });
+
+    return ElementwiseTestHelper<4, armnn::EqualQueueDescriptor, armnn::DataType::Float32, armnn::DataType::Boolean>(
+        workloadFactory,
+        memoryManager,
+        shape0,
+        input0,
+        shape1,
+        input1,
+        shape0,
+        output);
+}
+
+LayerTestResult<uint8_t, 4> EqualUint8Test(
+        armnn::IWorkloadFactory& workloadFactory,
+        const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    unsigned int shape[] = { 2, 2, 2, 2 };
+
+    // See dequantized values to the right.
+    std::vector<uint8_t> input0({ 1, 1, 1, 1, 6, 6, 6, 6,
+                                  3, 3, 3, 3, 7, 7, 7, 7 });
+
+    std::vector<uint8_t> input1({ 2, 2, 2, 2, 6, 6, 6, 6,
+                                  3, 3, 3, 3, 5, 5, 5, 5 });
+
+    std::vector<uint8_t> output({ 0, 0, 0, 0, 1, 1, 1, 1,
+                                  1, 1, 1, 1, 0, 0, 0, 0 });
+
+    return ElementwiseTestHelper<4,
+                                 armnn::EqualQueueDescriptor,
+                                 armnn::DataType::QuantisedAsymm8,
+                                 armnn::DataType::Boolean>(
+        workloadFactory,
+        memoryManager,
+        shape,
+        input0,
+        shape,
+        input1,
+        shape,
+        output);
+}
+
+LayerTestResult<uint8_t, 4> EqualBroadcast1ElementUint8Test(
+        armnn::IWorkloadFactory& workloadFactory,
+        const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    const unsigned int shape0[] = { 1, 2, 2, 3 };
+    const unsigned int shape1[] = { 1, 1, 1, 1 };
+
+    std::vector<uint8_t> input0({ 1, 2, 3, 4, 5, 6,
+                                  7, 8, 9, 10, 11, 12 });
+
+    std::vector<uint8_t> input1({ 1 });
+
+    std::vector<uint8_t> output({ 1, 0, 0, 0, 0, 0,
+                                  0, 0, 0, 0, 0, 0 });
+
+    return ElementwiseTestHelper<4,
+                                 armnn::EqualQueueDescriptor,
+                                 armnn::DataType::QuantisedAsymm8,
+                                 armnn::DataType::Boolean>(
+        workloadFactory,
+        memoryManager,
+        shape0,
+        input0,
+        shape1,
+        input1,
+        shape0,
+        output);
+}
+
+LayerTestResult<uint8_t, 4> EqualBroadcast1DVectorUint8Test(
+        armnn::IWorkloadFactory& workloadFactory,
+        const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    const unsigned int shape0[] = { 1, 2, 2, 3 };
+    const unsigned int shape1[] = { 1, 1, 1, 3 };
+
+    std::vector<uint8_t> input0({ 1, 2, 3, 4, 5, 6,
+                                  7, 8, 9, 10, 11, 12 });
+
+    std::vector<uint8_t> input1({ 1, 1, 3});
+
+    std::vector<uint8_t> output({ 1, 0, 1, 0, 0, 0,
+                                  0, 0, 0, 0, 0, 0 });
+
+    return ElementwiseTestHelper<4,
+                                 armnn::EqualQueueDescriptor,
+                                 armnn::DataType::QuantisedAsymm8,
+                                 armnn::DataType::Boolean>(
+        workloadFactory,
+        memoryManager,
+        shape0,
+        input0,
+        shape1,
+        input1,
+        shape0,
+        output);
+}
diff --git a/src/backends/backendsCommon/test/layerTests/EqualTestImpl.hpp b/src/backends/backendsCommon/test/layerTests/EqualTestImpl.hpp
new file mode 100644
index 0000000000..e9560b38bd
--- /dev/null
+++ b/src/backends/backendsCommon/test/layerTests/EqualTestImpl.hpp
@@ -0,0 +1,34 @@
+//
+// Copyright © 2017 Arm Ltd. All rights reserved.
+// SPDX-License-Identifier: MIT
+//
+
+#pragma once
+
+#include "LayerTestResult.hpp"
+
+#include <backendsCommon/IBackendInternal.hpp>
+#include <backendsCommon/WorkloadFactory.hpp>
+
+LayerTestResult<uint8_t, 4> EqualSimpleTest(armnn::IWorkloadFactory& workloadFactory,
+                                            const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<uint8_t, 4> EqualBroadcast1ElementTest(
+        armnn::IWorkloadFactory& workloadFactory,
+        const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<uint8_t, 4> EqualBroadcast1DVectorTest(
+        armnn::IWorkloadFactory& workloadFactory,
+        const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<uint8_t, 4> EqualUint8Test(
+        armnn::IWorkloadFactory& workloadFactory,
+        const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<uint8_t, 4> EqualBroadcast1ElementUint8Test(
+        armnn::IWorkloadFactory& workloadFactory,
+        const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<uint8_t, 4> EqualBroadcast1DVectorUint8Test(
+        armnn::IWorkloadFactory& workloadFactory,
+        const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
diff --git a/src/backends/backendsCommon/test/layerTests/GreaterTestImpl.cpp b/src/backends/backendsCommon/test/layerTests/GreaterTestImpl.cpp
new file mode 100644
index 0000000000..b5bf560e3c
--- /dev/null
+++ b/src/backends/backendsCommon/test/layerTests/GreaterTestImpl.cpp
@@ -0,0 +1,229 @@
+//
+// Copyright © 2017 Arm Ltd. All rights reserved.
+// SPDX-License-Identifier: MIT
+//
+
+#include "GreaterTestImpl.hpp"
+
+#include "ElementwiseTestImpl.hpp"
+
+template<>
+std::unique_ptr<armnn::IWorkload> CreateWorkload<armnn::GreaterQueueDescriptor>(
+    const armnn::IWorkloadFactory& workloadFactory,
+    const armnn::WorkloadInfo& info,
+    const armnn::GreaterQueueDescriptor& descriptor)
+{
+    return workloadFactory.CreateGreater(descriptor, info);
+}
+
+LayerTestResult<uint8_t, 4> GreaterSimpleTest(armnn::IWorkloadFactory& workloadFactory,
+                                              const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    const unsigned int width        = 2u;
+    const unsigned int height       = 2u;
+    const unsigned int channelCount = 2u;
+    const unsigned int batchSize    = 2u;
+
+    unsigned int shape[] = { batchSize, channelCount, height, width };
+
+    std::vector<float> input0 =
+    {
+        1.f, 1.f, 1.f, 1.f,  5.f, 5.f, 5.f, 5.f,
+        3.f, 3.f, 3.f, 3.f,  4.f, 4.f, 4.f, 4.f
+    };
+
+    std::vector<float> input1 =
+    {
+        1.f, 1.f, 1.f, 1.f,  3.f, 3.f, 3.f, 3.f,
+        5.f, 5.f, 5.f, 5.f,  4.f, 4.f, 4.f, 4.f
+    };
+
+    std::vector<uint8_t> output =
+    {
+        0, 0, 0, 0,  1, 1, 1, 1,
+        0, 0, 0, 0,  0, 0, 0, 0
+    };
+
+    return ElementwiseTestHelper<4,
+                                 armnn::GreaterQueueDescriptor,
+                                 armnn::DataType::Float32,
+                                 armnn::DataType::Boolean>(
+        workloadFactory,
+        memoryManager,
+        shape,
+        input0,
+        shape,
+        input1,
+        shape,
+        output);
+}
+
+LayerTestResult<uint8_t, 4> GreaterBroadcast1ElementTest(
+        armnn::IWorkloadFactory& workloadFactory,
+        const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    unsigned int shape0[] = { 1, 2, 2, 2 };
+    unsigned int shape1[] = { 1, 1, 1, 1 };
+
+    std::vector<float> input0 = { 1, 2, 3, 4, 5, 6, 7, 8};
+    std::vector<float> input1 = { 1 };
+
+    std::vector<uint8_t> output = { 0, 1, 1, 1, 1, 1, 1, 1};
+
+    return ElementwiseTestHelper<4,
+                                 armnn::GreaterQueueDescriptor,
+                                 armnn::DataType::Float32,
+                                 armnn::DataType::Boolean>(
+        workloadFactory,
+        memoryManager,
+        shape0,
+        input0,
+        shape1,
+        input1,
+        shape0,
+        output);
+}
+
+LayerTestResult<uint8_t, 4> GreaterBroadcast1DVectorTest(
+        armnn::IWorkloadFactory& workloadFactory,
+        const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    const unsigned int shape0[] = { 1, 2, 2, 3 };
+    const unsigned int shape1[] = { 1, 1, 1, 3 };
+
+    std::vector<float> input0 =
+    {
+        1.0f, 2.9f, 2.1f,  4.0f,  5.0f,  6.0f,
+        7.0f, 8.0f, 9.0f, 10.0f, 11.0f, 12.0f
+    };
+
+    std::vector<float> input1 = { 1.f, 3.f, 2.f };
+
+    std::vector<uint8_t> output =
+    {
+        0, 0, 1, 1, 1, 1,
+        1, 1, 1, 1, 1, 1
+    };
+
+    return ElementwiseTestHelper<4,
+                                 armnn::GreaterQueueDescriptor,
+                                 armnn::DataType::Float32,
+                                 armnn::DataType::Boolean>(
+        workloadFactory,
+        memoryManager,
+        shape0,
+        input0,
+        shape1,
+        input1,
+        shape0,
+        output);
+}
+
+LayerTestResult<uint8_t, 4> GreaterUint8Test(
+        armnn::IWorkloadFactory& workloadFactory,
+        const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    unsigned int shape[] = { 2, 2, 2, 2 };
+
+    // See dequantized values to the right.
+    std::vector<uint8_t> input0 =
+    {
+        1, 1, 1, 1, 6, 6, 6, 6,
+        3, 3, 3, 3, 5, 5, 5, 5
+    };
+
+    std::vector<uint8_t> input1 =
+    {
+        2, 2, 2, 2, 6, 6, 6, 6,
+        2, 2, 2, 2, 5, 5, 5, 5
+    };
+
+    std::vector<uint8_t> output =
+    {
+        0, 0, 0, 0, 0, 0, 0, 0,
+        1, 1, 1, 1, 0, 0, 0, 0
+    };
+
+    return ElementwiseTestHelper<4,
+                                 armnn::GreaterQueueDescriptor,
+                                 armnn::DataType::QuantisedAsymm8,
+                                 armnn::DataType::Boolean>(
+        workloadFactory,
+        memoryManager,
+        shape,
+        input0,
+        shape,
+        input1,
+        shape,
+        output);
+}
+
+LayerTestResult<uint8_t, 4> GreaterBroadcast1ElementUint8Test(
+        armnn::IWorkloadFactory& workloadFactory,
+        const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    const unsigned int shape0[] = { 1, 2, 2, 3 };
+    const unsigned int shape1[] = { 1, 1, 1, 1 };
+
+    std::vector<uint8_t> input0 =
+    {
+        1, 2, 3,  4,  5,  6,
+        7, 8, 9, 10, 11, 12
+    };
+
+    std::vector<uint8_t> input1 = { 1 };
+
+    std::vector<uint8_t> output =
+    {
+        0, 1, 1, 1, 1, 1,
+        1, 1, 1, 1, 1, 1
+    };
+
+    return ElementwiseTestHelper<4,
+                                 armnn::GreaterQueueDescriptor,
+                                 armnn::DataType::QuantisedAsymm8,
+                                 armnn::DataType::Boolean>(
+        workloadFactory,
+        memoryManager,
+        shape0,
+        input0,
+        shape1,
+        input1,
+        shape0,
+        output);
+}
+
+LayerTestResult<uint8_t, 4> GreaterBroadcast1DVectorUint8Test(
+        armnn::IWorkloadFactory& workloadFactory,
+        const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    const unsigned int shape0[] = { 1, 2, 2, 3 };
+    const unsigned int shape1[] = { 1, 1, 1, 3 };
+
+    std::vector<uint8_t> input0 =
+    {
+        1, 2, 3,  4,  5,  6,
+        7, 8, 9, 10, 11, 12
+    };
+
+    std::vector<uint8_t> input1 = { 1, 1, 3 };
+
+    std::vector<uint8_t> output =
+    {
+        0, 1, 0, 1, 1, 1,
+        1, 1, 1, 1, 1, 1
+    };
+
+    return ElementwiseTestHelper<4,
+                                 armnn::GreaterQueueDescriptor,
+                                 armnn::DataType::QuantisedAsymm8,
+                                 armnn::DataType::Boolean>(
+        workloadFactory,
+        memoryManager,
+        shape0,
+        input0,
+        shape1,
+        input1,
+        shape0,
+        output);
+}
diff --git a/src/backends/backendsCommon/test/layerTests/GreaterTestImpl.hpp b/src/backends/backendsCommon/test/layerTests/GreaterTestImpl.hpp
new file mode 100644
index 0000000000..39f3a39451
--- /dev/null
+++ b/src/backends/backendsCommon/test/layerTests/GreaterTestImpl.hpp
@@ -0,0 +1,34 @@
+//
+// Copyright © 2017 Arm Ltd. All rights reserved.
+// SPDX-License-Identifier: MIT
+//
+
+#pragma once
+
+#include "LayerTestResult.hpp"
+
+#include <backendsCommon/IBackendInternal.hpp>
+#include <backendsCommon/WorkloadFactory.hpp>
+
+LayerTestResult<uint8_t, 4> GreaterSimpleTest(armnn::IWorkloadFactory& workloadFactory,
+                                              const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<uint8_t, 4> GreaterBroadcast1ElementTest(
+        armnn::IWorkloadFactory& workloadFactory,
+        const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<uint8_t, 4> GreaterBroadcast1DVectorTest(
+        armnn::IWorkloadFactory& workloadFactory,
+        const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<uint8_t, 4> GreaterUint8Test(
+        armnn::IWorkloadFactory& workloadFactory,
+        const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<uint8_t, 4> GreaterBroadcast1ElementUint8Test(
+        armnn::IWorkloadFactory& workloadFactory,
+        const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<uint8_t, 4> GreaterBroadcast1DVectorUint8Test(
+        armnn::IWorkloadFactory& workloadFactory,
+        const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
diff --git a/src/backends/backendsCommon/test/layerTests/LayerTestResult.hpp b/src/backends/backendsCommon/test/layerTests/LayerTestResult.hpp
new file mode 100644
index 0000000000..c277d2d5e1
--- /dev/null
+++ b/src/backends/backendsCommon/test/layerTests/LayerTestResult.hpp
@@ -0,0 +1,44 @@
+//
+// Copyright © 2017 Arm Ltd. All rights reserved.
+// SPDX-License-Identifier: MIT
+//
+
+#pragma once
+
+#include <armnn/Tensor.hpp>
+
+#include <boost/multi_array.hpp>
+
+#include <cstddef>
+
+template <std::size_t n>
+boost::array<unsigned int, n> GetTensorShapeAsArray(const armnn::TensorInfo& tensorInfo)
+{
+    BOOST_ASSERT_MSG(n == tensorInfo.GetNumDimensions(),
+        "Attempting to construct a shape array of mismatching size");
+
+    boost::array<unsigned int, n> shape;
+    for (unsigned int i = 0; i < n; i++)
+    {
+        shape[i] = tensorInfo.GetShape()[i];
+    }
+    return shape;
+}
+
+template <typename T, std::size_t n>
+struct LayerTestResult
+{
+    LayerTestResult(const armnn::TensorInfo& outputInfo)
+    {
+        auto shape( GetTensorShapeAsArray<n>(outputInfo) );
+        output.resize(shape);
+        outputExpected.resize(shape);
+        supported = true;
+        compareBoolean = false;
+    }
+
+    boost::multi_array<T, n> output;
+    boost::multi_array<T, n> outputExpected;
+    bool supported;
+    bool compareBoolean;
+};
diff --git a/src/backends/backendsCommon/test/layerTests/MaximumTestImpl.cpp b/src/backends/backendsCommon/test/layerTests/MaximumTestImpl.cpp
new file mode 100644
index 0000000000..d0e624d655
--- /dev/null
+++ b/src/backends/backendsCommon/test/layerTests/MaximumTestImpl.cpp
@@ -0,0 +1,301 @@
+//
+// Copyright © 2017 Arm Ltd. All rights reserved.
+// SPDX-License-Identifier: MIT
+//
+
+#include "MaximumTestImpl.hpp"
+
+#include "ElementwiseTestImpl.hpp"
+
+template<>
+std::unique_ptr<armnn::IWorkload> CreateWorkload<armnn::MaximumQueueDescriptor>(
+    const armnn::IWorkloadFactory& workloadFactory,
+    const armnn::WorkloadInfo& info,
+    const armnn::MaximumQueueDescriptor& descriptor)
+{
+    return workloadFactory.CreateMaximum(descriptor, info);
+}
+
+LayerTestResult<float, 4> MaximumSimpleTest(armnn::IWorkloadFactory& workloadFactory,
+                                           const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    const unsigned int width        = 2u;
+    const unsigned int height       = 2u;
+    const unsigned int channelCount = 2u;
+    const unsigned int batchSize    = 2u;
+
+    unsigned int shape[] = { batchSize, channelCount, height, width };
+
+    std::vector<float> input0 =
+    {
+        1.f, 1.f, 1.f, 1.f,  5.f, 5.f, 5.f, 5.f,
+        3.f, 3.f, 3.f, 3.f,  4.f, 4.f, 4.f, 4.f
+    };
+
+    std::vector<float> input1 =
+    {
+        2.f, 2.f, 2.f, 2.f,  3.f, 3.f, 3.f, 3.f,
+        4.f, 4.f, 4.f, 4.f,  5.f, 5.f, 5.f, 5.f
+    };
+
+    std::vector<float> output =
+    {
+        2.f, 2.f, 2.f, 2.f,  5.f, 5.f, 5.f, 5.f,
+        4.f, 4.f, 4.f, 4.f,  5.f, 5.f, 5.f, 5.f
+    };
+
+    return ElementwiseTestHelper<4, armnn::MaximumQueueDescriptor, armnn::DataType::Float32>(
+        workloadFactory,
+        memoryManager,
+        shape,
+        input0,
+        shape,
+        input1,
+        shape,
+        output);
+}
+
+LayerTestResult<float, 4> MaximumBroadcast1ElementTest(
+        armnn::IWorkloadFactory& workloadFactory,
+        const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    unsigned int shape0[] = { 1, 2, 2, 2 };
+    unsigned int shape1[] = { 1, 1, 1, 1 };
+
+    std::vector<float> input0 = { 1.f, 2.f, 3.f, 4.f, 5.f, 6.f, 7.f, 8.f };
+
+    std::vector<float> input1 = { 2.f };
+
+    std::vector<float> output = { 2.f, 2.f, 3.f, 4.f, 5.f, 6.f, 7.f, 8.f };
+
+    return ElementwiseTestHelper<4, armnn::MaximumQueueDescriptor, armnn::DataType::Float32>(
+        workloadFactory,
+        memoryManager,
+        shape0,
+        input0,
+        shape1,
+        input1,
+        shape0,
+        output);
+}
+
+LayerTestResult<float, 4> MaximumBroadcast1DVectorTest(
+        armnn::IWorkloadFactory& workloadFactory,
+        const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    const unsigned int shape0[] = { 1, 2, 2, 3 };
+    const unsigned int shape1[] = { 1, 1, 1, 3 };
+
+    std::vector<float> input0 =
+    {
+        1.f, 2.f, 3.f,  4.f,  5.f,  6.f,
+        7.f, 8.f, 9.f, 10.f, 11.f, 12.f
+    };
+
+    std::vector<float> input1 = { 1.f, 2.f, 3.f };
+
+    std::vector<float> output =
+    {
+        1.f, 2.f, 3.f,  4.f,  5.f,  6.f,
+        7.f, 8.f, 9.f, 10.f, 11.f, 12.f
+    };
+
+    return ElementwiseTestHelper<4, armnn::MaximumQueueDescriptor, armnn::DataType::Float32>(
+        workloadFactory,
+        memoryManager,
+        shape0,
+        input0,
+        shape1,
+        input1,
+        shape0,
+        output);
+}
+
+LayerTestResult<uint8_t, 4> MaximumUint8Test(
+        armnn::IWorkloadFactory& workloadFactory,
+        const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    unsigned int shape[] = { 2, 2, 2, 2 };
+
+    // See dequantized values to the right.
+    std::vector<uint8_t> input0 =
+    {
+        1, 1, 1, 1, 6, 6, 6, 6,
+        3, 3, 3, 3, 4, 4, 4, 4
+    };
+
+    std::vector<uint8_t> input1 =
+    {
+        2, 2, 2, 2, 3, 3, 3, 3,
+        4, 4, 4, 4, 5, 5, 5, 5
+    };
+
+    std::vector<uint8_t> output =
+    {
+        2, 2, 2, 2, 6, 6, 6, 6,
+        4, 4, 4, 4, 5, 5, 5, 5
+    };
+
+    return ElementwiseTestHelper<4, armnn::MaximumQueueDescriptor, armnn::DataType::QuantisedAsymm8>(
+        workloadFactory,
+        memoryManager,
+        shape,
+        input0,
+        shape,
+        input1,
+        shape,
+        output);
+}
+
+LayerTestResult<uint8_t, 4> MaximumBroadcast1ElementUint8Test(
+        armnn::IWorkloadFactory& workloadFactory,
+        const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    const unsigned int shape0[] = { 1, 2, 2, 3 };
+    const unsigned int shape1[] = { 1, 1, 1, 1 };
+
+    std::vector<uint8_t> input0 =
+    {
+        1, 2, 3,  4,  5,  6,
+        7, 8, 9, 10, 11, 12
+    };
+
+    std::vector<uint8_t> input1 = { 2 };
+
+    std::vector<uint8_t> output =
+    {
+        2, 2, 3,  4,  5,  6,
+        7, 8, 9, 10, 11, 12
+    };
+
+    return ElementwiseTestHelper<4, armnn::MaximumQueueDescriptor, armnn::DataType::QuantisedAsymm8>(
+        workloadFactory,
+        memoryManager,
+        shape0,
+        input0,
+        shape1,
+        input1,
+        shape0,
+        output);
+}
+
+LayerTestResult<uint8_t, 4> MaximumBroadcast1DVectorUint8Test(
+        armnn::IWorkloadFactory& workloadFactory,
+        const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    const unsigned int shape0[] = { 1, 2, 2, 3 };
+    const unsigned int shape1[] = { 1, 1, 1, 3 };
+
+    std::vector<uint8_t> input0 =
+    {
+        1, 2, 3,  4,  5,  6,
+        7, 8, 9, 10, 11, 12
+    };
+
+    std::vector<uint8_t> input1 = { 1, 10, 3 };
+
+    std::vector<uint8_t> output = {
+        1, 10, 3,  4, 10,  6,
+        7, 10, 9, 10, 11, 12
+    };
+
+    return ElementwiseTestHelper<4, armnn::MaximumQueueDescriptor, armnn::DataType::QuantisedAsymm8>(
+        workloadFactory,
+        memoryManager,
+        shape0,
+        input0,
+        shape1,
+        input1,
+        shape0,
+        output);
+}
+
+LayerTestResult<int16_t, 4> MaximumInt16Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    unsigned int shape[] = { 2, 2, 2, 2 };
+
+    std::vector<int16_t> input0({ 1, 1, 1, 1, 6, 6, 6, 6,
+                                  3, 3, 3, 3, 4, 4, 4, 4 });
+
+    std::vector<int16_t> input1({ 2, 2, 2, 2, 3, 3, 3, 3,
+                                  4, 4, 4, 4, 5, 5, 5, 5 });
+
+    std::vector<int16_t> output({ 2, 2, 2, 2, 6, 6, 6, 6,
+                                  4, 4, 4, 4, 5, 5, 5, 5 });
+
+    return ElementwiseTestHelper<4, armnn::MaximumQueueDescriptor, armnn::DataType::QuantisedSymm16>(
+        workloadFactory,
+        memoryManager,
+        shape,
+        input0,
+        shape,
+        input1,
+        shape,
+        output);
+}
+
+LayerTestResult<int16_t, 4> MaximumBroadcast1ElementInt16Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    const unsigned int shape0[] = { 1, 2, 2, 3 };
+    const unsigned int shape1[] = { 1, 1, 1, 1 };
+
+    std::vector<int16_t> input0 =
+    {
+        1, 2, 3,  4,  5,  6,
+        7, 8, 9, 10, 11, 12
+    };
+
+    std::vector<int16_t> input1 = { 2 };
+
+    std::vector<int16_t> output =
+    {
+        2, 2, 3,  4,  5,  6,
+        7, 8, 9, 10, 11, 12
+    };
+
+    return ElementwiseTestHelper<4, armnn::MaximumQueueDescriptor, armnn::DataType::QuantisedSymm16>(
+        workloadFactory,
+        memoryManager,
+        shape0,
+        input0,
+        shape1,
+        input1,
+        shape0,
+        output);
+}
+
+LayerTestResult<int16_t, 4> MaximumBroadcast1DVectorInt16Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    const unsigned int shape0[] = { 1, 2, 2, 3 };
+    const unsigned int shape1[] = { 1, 1, 1, 3 };
+
+    std::vector<int16_t> input0 =
+    {
+        1, 2, 3,  4,  5,  6,
+        7, 8, 9, 10, 11, 12
+    };
+
+    std::vector<int16_t> input1 = { 1, 10, 3 };
+
+    std::vector<int16_t> output =
+    {
+        1, 10, 3,  4, 10,  6,
+        7, 10, 9, 10, 11, 12
+    };
+
+    return ElementwiseTestHelper<4, armnn::MaximumQueueDescriptor, armnn::DataType::QuantisedSymm16>(
+        workloadFactory,
+        memoryManager,
+        shape0,
+        input0,
+        shape1,
+        input1,
+        shape0,
+        output);
+}
diff --git a/src/backends/backendsCommon/test/layerTests/MaximumTestImpl.hpp b/src/backends/backendsCommon/test/layerTests/MaximumTestImpl.hpp
new file mode 100644
index 0000000000..b672431c15
--- /dev/null
+++ b/src/backends/backendsCommon/test/layerTests/MaximumTestImpl.hpp
@@ -0,0 +1,46 @@
+//
+// Copyright © 2017 Arm Ltd. All rights reserved.
+// SPDX-License-Identifier: MIT
+//
+
+#pragma once
+
+#include "LayerTestResult.hpp"
+
+#include <backendsCommon/IBackendInternal.hpp>
+#include <backendsCommon/WorkloadFactory.hpp>
+
+LayerTestResult<float, 4> MaximumSimpleTest(armnn::IWorkloadFactory& workloadFactory,
+                                           const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<float, 4> MaximumBroadcast1ElementTest(
+        armnn::IWorkloadFactory& workloadFactory,
+        const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<float, 4> MaximumBroadcast1DVectorTest(
+        armnn::IWorkloadFactory& workloadFactory,
+        const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<uint8_t, 4> MaximumUint8Test(
+        armnn::IWorkloadFactory& workloadFactory,
+        const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<uint8_t, 4> MaximumBroadcast1ElementUint8Test(
+        armnn::IWorkloadFactory& workloadFactory,
+        const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<uint8_t, 4> MaximumBroadcast1DVectorUint8Test(
+        armnn::IWorkloadFactory& workloadFactory,
+        const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<int16_t, 4> MaximumInt16Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<int16_t, 4> MaximumBroadcast1ElementInt16Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<int16_t, 4> MaximumBroadcast1DVectorInt16Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
diff --git a/src/backends/backendsCommon/test/layerTests/MinimumTestImpl.cpp b/src/backends/backendsCommon/test/layerTests/MinimumTestImpl.cpp
new file mode 100644
index 0000000000..eabad8f852
--- /dev/null
+++ b/src/backends/backendsCommon/test/layerTests/MinimumTestImpl.cpp
@@ -0,0 +1,196 @@
+//
+// Copyright © 2017 Arm Ltd. All rights reserved.
+// SPDX-License-Identifier: MIT
+//
+
+#include "MinimumTestImpl.hpp"
+
+#include "ElementwiseTestImpl.hpp"
+
+template<>
+std::unique_ptr<armnn::IWorkload> CreateWorkload<armnn::MinimumQueueDescriptor>(
+    const armnn::IWorkloadFactory& workloadFactory,
+    const armnn::WorkloadInfo& info,
+    const armnn::MinimumQueueDescriptor& descriptor)
+{
+    return workloadFactory.CreateMinimum(descriptor, info);
+}
+
+LayerTestResult<float, 4> MinimumBroadcast1ElementTest1(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    unsigned int shape0[] = { 1, 2, 2, 2 };
+    unsigned int shape1[] = { 1, 1, 1, 1 };
+
+    std::vector<float> input0 = { 1.f, 2.f, 3.f, 4.f, 5.f, 6.f, 7.f, 8.f };
+
+    std::vector<float> input1 = { 2.f };
+
+    std::vector<float> output = { 1.f, 2.f, 2.f, 2.f, 2.f, 2.f, 2.f, 2.f };
+
+    return ElementwiseTestHelper<4, armnn::MinimumQueueDescriptor, armnn::DataType::Float32>(
+        workloadFactory,
+        memoryManager,
+        shape0,
+        input0,
+        shape1,
+        input1,
+        shape0,
+        output);
+}
+
+LayerTestResult<float, 4> MinimumBroadcast1ElementTest2(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    unsigned int shape0[] = { 1, 2, 2, 2 };
+    unsigned int shape1[] = { 1, 1, 1, 1 };
+
+    std::vector<float> input0 = { 1.f, 6.f, 3.f, 2.f, 8.f, 9.f, 1.f, 10.f };
+
+    std::vector<float> input1 = { 5.f };
+
+    std::vector<float> output = { 1.f, 5.f, 3.f, 2.f, 5.f, 5.f, 1.f, 5.f };
+
+    return ElementwiseTestHelper<4, armnn::MinimumQueueDescriptor, armnn::DataType::Float32>(
+        workloadFactory,
+        memoryManager,
+        shape0,
+        input0,
+        shape1,
+        input1,
+        shape0,
+        output);
+}
+
+LayerTestResult<uint8_t, 4> MinimumBroadcast1DVectorUint8Test(
+    armnn::IWorkloadFactory & workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr & memoryManager)
+{
+    const unsigned int shape0[] = { 1, 2, 2, 3 };
+    const unsigned int shape1[] = { 1, 1, 1, 3 };
+
+    std::vector<uint8_t> input0 =
+    {
+        1, 2, 3, 3, 2, 1,
+        7, 1, 2, 3, 4, 5
+    };
+
+    std::vector<uint8_t> input1 = { 1, 2, 3 };
+
+    std::vector<uint8_t> output =
+    {
+        1, 2, 3, 1, 2, 1,
+        1, 1, 2, 1, 2, 3
+    };
+
+    return ElementwiseTestHelper<4, armnn::MinimumQueueDescriptor, armnn::DataType::QuantisedAsymm8>(
+        workloadFactory,
+        memoryManager,
+        shape0,
+        input0,
+        shape1,
+        input1,
+        shape0,
+        output);
+}
+
+LayerTestResult<int16_t, 4> MinimumInt16Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    unsigned int shape[] = { 2, 2, 2, 2 };
+
+    std::vector<int16_t> input0 =
+    {
+        1, 1, 1, 1, 6, 6, 6, 6,
+        3, 3, 3, 3, 4, 4, 4, 4
+    };
+
+    std::vector<int16_t> input1 =
+    {
+        2, 2, 2, 2, 3, 3, 3, 3,
+        4, 4, 4, 4, 5, 5, 5, 5
+    };
+
+    std::vector<int16_t> output
+    {
+        1, 1, 1, 1, 3, 3, 3, 3,
+        3, 3, 3, 3, 4, 4, 4, 4
+    };
+
+    return ElementwiseTestHelper<4, armnn::MinimumQueueDescriptor, armnn::DataType::QuantisedSymm16>(
+        workloadFactory,
+        memoryManager,
+        shape,
+        input0,
+        shape,
+        input1,
+        shape,
+        output);
+}
+
+LayerTestResult<int16_t, 4> MinimumBroadcast1ElementInt16Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    const unsigned int shape0[] = { 1, 2, 2, 3 };
+    const unsigned int shape1[] = { 1, 1, 1, 1 };
+
+    std::vector<int16_t> input0 =
+    {
+        1, 2, 3,  4,  5,  6,
+        7, 8, 9, 10, 11, 12
+    };
+
+    std::vector<int16_t> input1 = { 2 };
+
+    std::vector<int16_t> output =
+    {
+        1, 2, 2, 2, 2, 2,
+        2, 2, 2, 2, 2, 2
+    };
+
+    return ElementwiseTestHelper<4, armnn::MinimumQueueDescriptor, armnn::DataType::QuantisedSymm16>(
+        workloadFactory,
+        memoryManager,
+        shape0,
+        input0,
+        shape1,
+        input1,
+        shape0,
+        output);
+}
+
+LayerTestResult<int16_t, 4> MinimumBroadcast1DVectorInt16Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    const unsigned int shape0[] = { 1, 2, 2, 3 };
+    const unsigned int shape1[] = { 1, 1, 1, 3 };
+
+    std::vector<int16_t> input0 =
+    {
+        1, 2, 3,  4,  5,  6,
+        7, 8, 9, 10, 11, 12
+    };
+
+    std::vector<int16_t> input1 = { 1, 10, 3 };
+
+    std::vector<int16_t> output =
+    {
+        1, 2, 3, 1,  5, 3,
+        1, 8, 3, 1, 10, 3
+    };
+
+    return ElementwiseTestHelper<4, armnn::MinimumQueueDescriptor, armnn::DataType::QuantisedSymm16>(
+        workloadFactory,
+        memoryManager,
+        shape0,
+        input0,
+        shape1,
+        input1,
+        shape0,
+        output);
+}
diff --git a/src/backends/backendsCommon/test/layerTests/MinimumTestImpl.hpp b/src/backends/backendsCommon/test/layerTests/MinimumTestImpl.hpp
new file mode 100644
index 0000000000..bb84bc0fe2
--- /dev/null
+++ b/src/backends/backendsCommon/test/layerTests/MinimumTestImpl.hpp
@@ -0,0 +1,35 @@
+//
+// Copyright © 2017 Arm Ltd. All rights reserved.
+// SPDX-License-Identifier: MIT
+//
+
+#pragma once
+
+#include "LayerTestResult.hpp"
+
+#include <backendsCommon/IBackendInternal.hpp>
+#include <backendsCommon/WorkloadFactory.hpp>
+
+LayerTestResult<float, 4> MinimumBroadcast1ElementTest1(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<float, 4> MinimumBroadcast1ElementTest2(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<uint8_t, 4> MinimumBroadcast1DVectorUint8Test(
+    armnn::IWorkloadFactory & workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr & memoryManager);
+
+LayerTestResult<int16_t , 4> MinimumInt16Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<int16_t, 4> MinimumBroadcast1ElementInt16Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<int16_t, 4> MinimumBroadcast1DVectorInt16Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
diff --git a/src/backends/backendsCommon/test/layerTests/MultiplicationTestImpl.cpp b/src/backends/backendsCommon/test/layerTests/MultiplicationTestImpl.cpp
new file mode 100644
index 0000000000..99b1b18c50
--- /dev/null
+++ b/src/backends/backendsCommon/test/layerTests/MultiplicationTestImpl.cpp
@@ -0,0 +1,467 @@
+//
+// Copyright © 2017 Arm Ltd. All rights reserved.
+// SPDX-License-Identifier: MIT
+//
+
+#include "MultiplicationTestImpl.hpp"
+
+#include "ElementwiseTestImpl.hpp"
+
+template<>
+std::unique_ptr<armnn::IWorkload> CreateWorkload<armnn::MultiplicationQueueDescriptor>(
+    const armnn::IWorkloadFactory& workloadFactory,
+    const armnn::WorkloadInfo& info,
+    const armnn::MultiplicationQueueDescriptor& descriptor)
+{
+    return workloadFactory.CreateMultiplication(descriptor, info);
+}
+
+LayerTestResult<float, 4> MultiplicationTest(armnn::IWorkloadFactory& workloadFactory,
+                                             const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    const unsigned int width        = 2u;
+    const unsigned int height       = 2u;
+    const unsigned int channelCount = 2u;
+    const unsigned int batchSize    = 2u;
+
+    unsigned int shape[] = { batchSize, channelCount, height, width };
+
+    std::vector<float> input0 =
+    {
+        1, 1, 1, 1,  2, 2, 2, 2,
+        3, 3, 3, 3,  4, 4, 4, 4
+    };
+
+    std::vector<float> input1 =
+    {
+        2, 2, 2, 2,  3, 3, 3, 3,
+        4, 4, 4, 4,  5, 5, 5, 5
+    };
+
+    std::vector<float> output =
+    {
+         2,  2,  2,  2,   6,  6,  6,  6,
+        12, 12, 12, 12,  20, 20, 20, 20
+    };
+
+    return ElementwiseTestHelper<4, armnn::MultiplicationQueueDescriptor, armnn::DataType::Float32>(
+        workloadFactory,
+        memoryManager,
+        shape,
+        input0,
+        shape,
+        input1,
+        shape,
+        output);
+}
+
+LayerTestResult<float, 5> Multiplication5dTest(armnn::IWorkloadFactory& workloadFactory,
+                                               const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    const unsigned int width        = 3u;
+    const unsigned int height       = 2u;
+    const unsigned int channelCount = 2u;
+    const unsigned int batchSize    = 2u;;
+    const unsigned int depth        = 2u;
+
+    unsigned int shape[] = { depth, batchSize, channelCount, height, width };
+
+    std::vector<float> input0 =
+    {
+        1.80f, 0.20f, 2.30f,  1.30f, 2.10f, 1.00f,
+        2.60f, 0.60f, 2.10f,  2.30f, 2.30f, 2.00f,
+
+        2.50f, 1.00f, 2.90f,  3.10f, 1.50f, 2.40f,
+        2.80f, 1.10f, 1.00f,  3.20f, 1.00f, 2.30f,
+
+
+        0.30f, 2.20f, 1.00f,  0.20f, 1.60f, 1.40f,
+        0.80f, 3.20f, 0.10f,  0.10f, 3.10f, 2.10f,
+
+        1.50f, 2.40f, 1.40f,  0.70f, 2.40f, 1.40f,
+        1.60f, 1.20f, 1.90f,  0.80f, 0.00f, 0.10f,
+    };
+
+    std::vector<float> input1 =
+    {
+        0.70f, 1.00f, 2.90f,  2.20f, 3.10f, 2.80f,
+        1.80f, 2.00f, 0.50f,  2.30f, 1.20f, 2.70f,
+
+        2.40f, 0.20f, 3.20f,  1.60f, 0.20f, 2.50f,
+        2.30f, 0.70f, 2.70f,  1.80f, 2.90f, 2.70f,
+
+
+        3.20f, 3.20f, 0.70f,  1.90f, 2.70f, 2.50f,
+        2.40f, 0.90f, 2.30f,  1.80f, 2.50f, 2.00f,
+
+        1.60f, 2.20f, 1.60f,  2.00f, 0.30f, 3.20f,
+        0.40f, 3.00f, 2.60f,  0.30f, 0.00f, 2.50f,
+    };
+
+    std::vector<float> output =
+    {
+        1.26f, 0.20f, 6.67f,  2.86f, 6.51f, 2.80f,
+        4.68f, 1.20f, 1.05f,  5.29f, 2.76f, 5.40f,
+
+        6.00f, 0.20f, 9.28f,  4.96f, 0.30f, 6.00f,
+        6.44f, 0.77f, 2.70f,  5.76f, 2.90f, 6.21f,
+
+
+        0.96f, 7.04f, 0.70f,  0.38f, 4.32f, 3.50f,
+        1.92f, 2.88f, 0.23f,  0.18f, 7.75f, 4.20f,
+
+        2.40f, 5.28f, 2.24f,  1.40f, 0.72f, 4.48f,
+        0.64f, 3.60f, 4.94f,  0.24f, 0.00f, 0.25f,
+    };
+
+    return ElementwiseTestHelper<5, armnn::MultiplicationQueueDescriptor, armnn::DataType::Float32>(
+        workloadFactory,
+        memoryManager,
+        shape,
+        input0,
+        shape,
+        input1,
+        shape,
+        output);
+}
+
+LayerTestResult<float, 4> MultiplicationBroadcast1ElementTest(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    unsigned int shape0[] = { 1, 2, 2, 2 };
+    unsigned int shape1[] = { 1, 1, 1, 1 };
+
+    std::vector<float> input0 = { 1, 2, 3, 4, 5, 6, 7, 8};
+
+    std::vector<float> input1 = { 2 };
+
+    std::vector<float> output = { 2, 4, 6, 8, 10, 12, 14, 16};
+
+    return ElementwiseTestHelper<4, armnn::MultiplicationQueueDescriptor, armnn::DataType::Float32>(
+        workloadFactory,
+        memoryManager,
+        shape0,
+        input0,
+        shape1,
+        input1,
+        shape0,
+        output);
+}
+
+LayerTestResult<float, 4> MultiplicationBroadcast1DVectorTest(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    unsigned int shape0[] = { 1, 3, 3, 2 };
+    unsigned int shape1[] = { 1, 1, 1, 2 };
+
+    std::vector<float> input0 =
+    {
+        1,   2,    3,  4,    5,  6,
+        7,   8,    9, 10,   11, 12,
+        13, 14,   15, 16,   17, 18
+    };
+
+    std::vector<float> input1 = { 1, 2 };
+
+    std::vector<float> output =
+    {
+         1,  4,    3,  8,    5, 12,
+         7, 16,    9, 20,   11, 24,
+        13, 28,   15, 32,   17, 36
+    };
+
+    return ElementwiseTestHelper<4, armnn::MultiplicationQueueDescriptor, armnn::DataType::Float32>(
+        workloadFactory,
+        memoryManager,
+        shape0,
+        input0,
+        shape1,
+        input1,
+        shape0,
+        output);
+}
+
+LayerTestResult<uint8_t, 4> MultiplicationUint8Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    constexpr unsigned int batchSize = 1u;
+    constexpr unsigned int channels  = 2u;
+    constexpr unsigned int height    = 2u;
+    constexpr unsigned int width     = 3u;
+
+    const unsigned int shape[] = { batchSize, channels, height, width };
+
+    // See dequantized values to the right
+    std::vector<uint8_t> input0 =
+    {
+         62,  37,   3, 172,  13, 111, // 244, 144,   8, 684,  48, 440,
+        188,  20,  73,  31,  23,  31  // 748,  76, 288, 120,  88, 120
+    };
+
+    // See dequantized values to the right
+    std::vector<uint8_t> input1 =
+    {
+        126, 240, 252, 183, 121, 247, // 384, 726, 762, 555, 369, 747,
+         48, 115, 151,  79,  78,  97  // 150, 351, 459, 243, 240, 297
+    };
+
+    // See dequantized values to the right
+    std::vector<uint8_t> output =
+    {
+         64,  72,   0, 255,   8, 236, //  93696, 104544, 6096(clamped), 379620(clamped), 17712, 328680,
+         77,  15,  92,  16,  10,  21, // 112200,  26676,        132192,           29160, 21120,  35640
+    };
+
+    // Scale/offset chosen to have output values out of range
+    return ElementwiseTestHelper<4, armnn::MultiplicationQueueDescriptor, armnn::DataType::QuantisedAsymm8>(
+        workloadFactory,
+        memoryManager,
+        shape,
+        input0,
+        4.0f,
+        1,
+        shape,
+        input1,
+        3.0f,
+        -2,
+        shape,
+        output,
+        1366.255f,
+        -5);
+}
+
+LayerTestResult<uint8_t, 4> MultiplicationBroadcast1ElementUint8Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    const unsigned int shape0[] = { 1, 2, 2, 3 };
+    const unsigned int shape1[] = { 1, 1, 1, 1 };
+
+    std::vector<uint8_t> input0 =
+    {
+        1, 2, 3,    4,  5,  6,
+        7, 8, 9,   10, 11, 12
+    };
+
+    std::vector<uint8_t> input1 = { 2 };
+
+    std::vector<uint8_t> output =
+    {
+        2,  4,   6,    8, 10, 12,
+        14, 16, 18,   20, 22, 24
+    };
+
+    return ElementwiseTestHelper<4, armnn::MultiplicationQueueDescriptor, armnn::DataType::QuantisedAsymm8>(
+        workloadFactory,
+        memoryManager,
+        shape0,
+        input0,
+        shape1,
+        input1,
+        shape0,
+        output);
+}
+
+LayerTestResult<uint8_t, 4> MultiplicationBroadcast1DVectorUint8Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    const unsigned int shape0[] = { 1, 2, 2, 3 };
+    const unsigned int shape1[] = { 1, 1, 1, 3 };
+
+    std::vector<uint8_t> input0 =
+    {
+        1, 2, 3,    4,  5,  6,
+        7, 8, 9,   10, 11, 12
+    };
+
+    std::vector<uint8_t> input1 = { 1, 2, 3 };
+
+    std::vector<uint8_t> output =
+    {
+        1,  4,   9,     4, 10, 18,
+        7, 16,  27,    10, 22, 36
+    };
+
+    return ElementwiseTestHelper<4, armnn::MultiplicationQueueDescriptor, armnn::DataType::QuantisedAsymm8>(
+        workloadFactory,
+        memoryManager,
+        shape0,
+        input0,
+        shape1,
+        input1,
+        shape0,
+        output);
+}
+
+LayerTestResult<int16_t, 4> MultiplicationInt16Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    const unsigned int shape[] = { 1, 2, 2, 3 };
+
+    std::vector<int16_t> input0 =
+    {
+        6,   7,  8,  9, 10, 11,
+        12, 13, 14, 15, 16, 17
+    };
+
+    std::vector<int16_t> input1 =
+    {
+        1, 2, 3,  4,  5,  6,
+        7, 8, 9, 10, 11, 12
+    };
+
+    std::vector<int16_t> output =
+    {
+        6,   14,  24,  36,  50,  66,
+        84, 104, 126, 150, 176, 204
+    };
+
+    return ElementwiseTestHelper<4, armnn::MultiplicationQueueDescriptor, armnn::DataType::QuantisedSymm16>(
+        workloadFactory,
+        memoryManager,
+        shape,
+        input0,
+        shape,
+        input1,
+        shape,
+        output);
+}
+
+LayerTestResult<int16_t, 4> MultiplicationBroadcast1ElementInt16Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    const unsigned int shape0[] = { 1, 2, 2, 3 };
+    const unsigned int shape1[] = { 1, 1, 1, 1 };
+
+    std::vector<int16_t> input0 =
+    {
+        1, 2, 3,  4,  5,  6,
+        7, 8, 9, 10, 11, 12
+    };
+
+    std::vector<int16_t> input1 = { 2 };
+
+    std::vector<int16_t> output =
+    {
+        2,   4,  6,  8, 10, 12,
+        14, 16, 18, 20, 22, 24
+    };
+
+    return ElementwiseTestHelper<4, armnn::MultiplicationQueueDescriptor, armnn::DataType::QuantisedSymm16>(
+        workloadFactory,
+        memoryManager,
+        shape0,
+        input0,
+        shape1,
+        input1,
+        shape0,
+        output);
+}
+
+LayerTestResult<int16_t, 4> MultiplicationBroadcast1DVectorInt16Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    const unsigned int shape0[] = { 1, 2, 2, 3 };
+    const unsigned int shape1[] = { 1, 1, 1, 3 };
+
+    std::vector<int16_t> input0 =
+    {
+        1, 2, 3,  4,  5,  6,
+        7, 8, 9, 10, 11, 12
+    };
+
+    std::vector<int16_t> input1 = { 1, 2, 3 };
+
+    std::vector<int16_t> output =
+    {
+        1,  4,  9,  4, 10, 18,
+        7, 16, 27, 10, 22, 36
+    };
+
+    return ElementwiseTestHelper<4, armnn::MultiplicationQueueDescriptor, armnn::DataType::QuantisedSymm16>(
+        workloadFactory,
+        memoryManager,
+        shape0,
+        input0,
+        shape1,
+        input1,
+        shape0,
+        output);
+}
+
+LayerTestResult<float,4> CompareMultiplicationTest(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
+    armnn::IWorkloadFactory& refWorkloadFactory)
+{
+    const unsigned int width = 16;
+    const unsigned int height = 32;
+    const unsigned int channelCount = 2;
+    const unsigned int batchSize = 5;
+
+    armnn::TensorInfo inputTensorInfo0;
+    armnn::TensorInfo inputTensorInfo1;
+    armnn::TensorInfo outputTensorInfo;
+
+    constexpr unsigned int shape[] = { batchSize, channelCount, height, width };
+
+    inputTensorInfo0 = armnn::TensorInfo(4, shape, armnn::DataType::Float32);
+    inputTensorInfo1 = armnn::TensorInfo(4, shape, armnn::DataType::Float32);
+    outputTensorInfo = armnn::TensorInfo(4, shape, armnn::DataType::Float32);
+
+    LayerTestResult<float,4> comparisonResult(outputTensorInfo);
+
+    auto input0 = MakeRandomTensor<float, 4>(inputTensorInfo0, 803506992);
+    auto input1 = MakeRandomTensor<float, 4>(inputTensorInfo1, 54902257);
+
+    std::unique_ptr<armnn::ITensorHandle> inputHandle0 = workloadFactory.CreateTensorHandle(inputTensorInfo0);
+    std::unique_ptr<armnn::ITensorHandle> inputHandle1 = workloadFactory.CreateTensorHandle(inputTensorInfo1);
+    std::unique_ptr<armnn::ITensorHandle> outputHandle = workloadFactory.CreateTensorHandle(outputTensorInfo);
+
+    std::unique_ptr<armnn::ITensorHandle> inputHandle0Ref = refWorkloadFactory.CreateTensorHandle(inputTensorInfo0);
+    std::unique_ptr<armnn::ITensorHandle> inputHandle1Ref = refWorkloadFactory.CreateTensorHandle(inputTensorInfo1);
+    std::unique_ptr<armnn::ITensorHandle> outputHandleRef = refWorkloadFactory.CreateTensorHandle(outputTensorInfo);
+
+    armnn::MultiplicationQueueDescriptor data;
+    armnn::WorkloadInfo info;
+    AddInputToWorkload(data, info, inputTensorInfo0, inputHandle0.get());
+    AddInputToWorkload(data, info, inputTensorInfo1, inputHandle1.get());
+    AddOutputToWorkload(data, info, outputTensorInfo, outputHandle.get());
+
+    armnn::MultiplicationQueueDescriptor refData = data;
+    armnn::WorkloadInfo refInfo = info;
+    SetWorkloadInput(refData, refInfo, 0, inputTensorInfo0, inputHandle0Ref.get());
+    SetWorkloadInput(refData, refInfo, 1, inputTensorInfo1, inputHandle1Ref.get());
+    SetWorkloadOutput(refData, refInfo, 0, outputTensorInfo, outputHandleRef.get());
+
+    std::unique_ptr<armnn::IWorkload> workload    = workloadFactory.CreateMultiplication(data, info);
+    std::unique_ptr<armnn::IWorkload> workloadRef = refWorkloadFactory.CreateMultiplication(refData, refInfo);
+
+    inputHandle0->Allocate();
+    inputHandle1->Allocate();
+    outputHandle->Allocate();
+    inputHandle0Ref->Allocate();
+    inputHandle1Ref->Allocate();
+    outputHandleRef->Allocate();
+
+    CopyDataToITensorHandle(inputHandle0.get(), input0.origin());
+    CopyDataToITensorHandle(inputHandle1.get(), input1.origin());
+    CopyDataToITensorHandle(inputHandle0Ref.get(), input0.origin());
+    CopyDataToITensorHandle(inputHandle1Ref.get(), input1.origin());
+
+    workload->PostAllocationConfigure();
+    workload->Execute();
+    workloadRef->PostAllocationConfigure();
+    workloadRef->Execute();
+    CopyDataFromITensorHandle(comparisonResult.output.origin(), outputHandle.get());
+    CopyDataFromITensorHandle(comparisonResult.outputExpected.origin(), outputHandleRef.get());
+
+    return comparisonResult;
+}
diff --git a/src/backends/backendsCommon/test/layerTests/MultiplicationTestImpl.hpp b/src/backends/backendsCommon/test/layerTests/MultiplicationTestImpl.hpp
new file mode 100644
index 0000000000..c56675150b
--- /dev/null
+++ b/src/backends/backendsCommon/test/layerTests/MultiplicationTestImpl.hpp
@@ -0,0 +1,56 @@
+//
+// Copyright © 2017 Arm Ltd. All rights reserved.
+// SPDX-License-Identifier: MIT
+//
+
+#pragma once
+
+#include "LayerTestResult.hpp"
+
+#include <backendsCommon/IBackendInternal.hpp>
+#include <backendsCommon/WorkloadFactory.hpp>
+
+LayerTestResult<float, 4> MultiplicationTest(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<float, 5> Multiplication5dTest(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<float, 4> MultiplicationBroadcast1ElementTest(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<float, 4> MultiplicationBroadcast1DVectorTest(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<uint8_t, 4> MultiplicationUint8Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<uint8_t, 4> MultiplicationBroadcast1ElementUint8Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<uint8_t, 4> MultiplicationBroadcast1DVectorUint8Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<int16_t, 4> MultiplicationInt16Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<int16_t, 4> MultiplicationBroadcast1ElementInt16Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<int16_t, 4> MultiplicationBroadcast1DVectorInt16Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<float, 4> CompareMultiplicationTest(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager,
+    armnn::IWorkloadFactory& refWorkloadFactory);
diff --git a/src/backends/backendsCommon/test/layerTests/SubtractionTestImpl.cpp b/src/backends/backendsCommon/test/layerTests/SubtractionTestImpl.cpp
new file mode 100644
index 0000000000..61225d40e5
--- /dev/null
+++ b/src/backends/backendsCommon/test/layerTests/SubtractionTestImpl.cpp
@@ -0,0 +1,250 @@
+//
+// Copyright © 2017 Arm Ltd. All rights reserved.
+// SPDX-License-Identifier: MIT
+//
+
+#include "SubtractionTestImpl.hpp"
+
+#include "ElementwiseTestImpl.hpp"
+
+template<>
+std::unique_ptr<armnn::IWorkload> CreateWorkload<armnn::SubtractionQueueDescriptor>(
+    const armnn::IWorkloadFactory& workloadFactory,
+    const armnn::WorkloadInfo& info,
+    const armnn::SubtractionQueueDescriptor& descriptor)
+{
+    return workloadFactory.CreateSubtraction(descriptor, info);
+}
+
+LayerTestResult<uint8_t, 4> SubtractionUint8Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    const unsigned int shape0[] = { 1, 1, 2, 2 };
+    const unsigned int shape1[] = { 1, 1, 2, 2 };
+
+    std::vector<uint8_t> input0 = { 10, 12, 14, 16 };
+    std::vector<uint8_t> input1 = {  1,  2,  1,  2 };
+    std::vector<uint8_t> output = {  3,  3,  5,  5 };
+
+    return ElementwiseTestHelper<4, armnn::SubtractionQueueDescriptor, armnn::DataType::QuantisedAsymm8>(
+        workloadFactory,
+        memoryManager,
+        shape0,
+        input0,
+        0.5f,
+        2,
+        shape1,
+        input1,
+        1.0f,
+        0,
+        shape0,
+        output,
+        1.0f,
+        0);
+}
+
+LayerTestResult<uint8_t, 4> SubtractionBroadcast1ElementUint8Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    const unsigned int shape0[] = { 1, 1, 2, 2 };
+    const unsigned int shape1[] = { 1, 1, 1, 1 };
+
+    std::vector<uint8_t> input0 = { 10, 12, 14, 16 };
+
+    std::vector<uint8_t> input1 = { 2 };
+
+    std::vector<uint8_t> output = { 5, 6, 7, 8 };
+
+    return ElementwiseTestHelper<4, armnn::SubtractionQueueDescriptor, armnn::DataType::QuantisedAsymm8>(
+        workloadFactory,
+        memoryManager,
+        shape0,
+        input0,
+        0.5f,
+        2,
+        shape1,
+        input1,
+        1.0f,
+        0,
+        shape0,
+        output,
+        1.0f,
+        3);
+}
+
+LayerTestResult<uint8_t, 4> SubtractionBroadcastUint8Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    const unsigned int shape0[] = { 1, 1, 2, 2 };
+    const unsigned int shape1[] = { 1, 1, 2, 1 };
+
+    std::vector<uint8_t> input0 = { 10, 12, 14, 16 };
+
+    std::vector<uint8_t> input1 = { 2, 1 };
+
+    std::vector<uint8_t> output = { 8, 11, 12, 15 };
+
+    return ElementwiseTestHelper<4, armnn::SubtractionQueueDescriptor, armnn::DataType::QuantisedAsymm8>(
+        workloadFactory,
+        memoryManager,
+        shape0,
+        input0,
+        shape1,
+        input1,
+        shape0,
+        output);
+}
+
+LayerTestResult<float, 4> SubtractionTest(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    const unsigned int shape0[] = { 1, 1, 2, 2 };
+    const unsigned int shape1[] = { 1, 1, 2, 2 };
+
+    std::vector<float> input0 = { 1,  2, 3, 4 };
+    std::vector<float> input1 = { 1, -1, 0, 2 };
+    std::vector<float> output = { 0,  3, 3, 2 };
+
+    return ElementwiseTestHelper<4, armnn::SubtractionQueueDescriptor, armnn::DataType::Float32>(
+        workloadFactory,
+        memoryManager,
+        shape0,
+        input0,
+        shape1,
+        input1,
+        shape0,
+        output);
+}
+
+LayerTestResult<float, 4> SubtractionBroadcast1ElementTest(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    const unsigned int shape0[] = { 1, 1, 2, 2 };
+    const unsigned int shape1[] = { 1, 1, 1, 1 };
+
+    std::vector<float> input0 = { 1,  2, 3, 4 };
+
+    std::vector<float> input1 = { 10 };
+
+    std::vector<float> output = { -9,  -8, -7, -6 };
+
+    return ElementwiseTestHelper<4, armnn::SubtractionQueueDescriptor, armnn::DataType::Float32>(
+        workloadFactory,
+        memoryManager,
+        shape0,
+        input0,
+        shape1,
+        input1,
+        shape0,
+        output);
+}
+
+LayerTestResult<float, 4> SubtractionBroadcastTest(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    const unsigned int shape0[] = { 1, 1, 2, 2 };
+    const unsigned int shape1[] = { 1, 1, 1, 2 };
+
+    std::vector<float> input0 = { 1,  2, 3, 4 };
+
+    std::vector<float> input1 = { 10, -5 };
+
+    std::vector<float> output = { -9,  7, -7, 9 };
+
+    return ElementwiseTestHelper<4, armnn::SubtractionQueueDescriptor, armnn::DataType::Float32>(
+        workloadFactory,
+        memoryManager,
+        shape0,
+        input0,
+        shape1,
+        input1,
+        shape0,
+        output);
+}
+
+LayerTestResult<int16_t, 4> SubtractionInt16Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    const unsigned int shape[] = { 1, 1, 2, 2 };
+
+    std::vector<int16_t> input0 = { 10, 12, 14, 16 };
+    std::vector<int16_t> input1 = {  1,  2,  1,  2 };
+    std::vector<int16_t> output = {  3,  3,  5,  5 };
+
+    return ElementwiseTestHelper<4, armnn::SubtractionQueueDescriptor, armnn::DataType::QuantisedSymm16>(
+        workloadFactory,
+        memoryManager,
+        shape,
+        input0,
+        0.5f,
+        0,
+        shape,
+        input1,
+        1.0f,
+        0,
+        shape,
+        output,
+        1.0f,
+        0);
+}
+
+LayerTestResult<int16_t, 4> SubtractionBroadcast1ElementInt16Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    const unsigned int shape0[] = { 1, 1, 2, 2 };
+    const unsigned int shape1[] = { 1, 1, 1, 1 };
+
+    std::vector<int16_t> input0 = { 10, 12, 14, 16 };
+
+    std::vector<int16_t> input1 = { 2 };
+
+    std::vector<int16_t> output = { 3, 4, 5, 6 };
+
+    return ElementwiseTestHelper<4, armnn::SubtractionQueueDescriptor, armnn::DataType::QuantisedSymm16>(
+        workloadFactory,
+        memoryManager,
+        shape0,
+        input0,
+        0.5f,
+        0,
+        shape1,
+        input1,
+        1.0f,
+        0,
+        shape0,
+        output,
+        1.0f,
+        0);
+}
+
+LayerTestResult<int16_t, 4> SubtractionBroadcastInt16Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager)
+{
+    const unsigned int shape0[] = { 1, 1, 2, 2 };
+    const unsigned int shape1[] = { 1, 1, 2, 1 };
+
+    std::vector<int16_t> input0 = { 10, 12, 14, 16 };
+
+    std::vector<int16_t> input1 = { 2, 1 };
+
+    std::vector<int16_t> output = { 8, 11, 12, 15 };
+
+    return ElementwiseTestHelper<4, armnn::SubtractionQueueDescriptor, armnn::DataType::QuantisedSymm16>(
+        workloadFactory,
+        memoryManager,
+        shape0,
+        input0,
+        shape1,
+        input1,
+        shape0,
+        output);
+}
diff --git a/src/backends/backendsCommon/test/layerTests/SubtractionTestImpl.hpp b/src/backends/backendsCommon/test/layerTests/SubtractionTestImpl.hpp
new file mode 100644
index 0000000000..ca1742b77b
--- /dev/null
+++ b/src/backends/backendsCommon/test/layerTests/SubtractionTestImpl.hpp
@@ -0,0 +1,47 @@
+//
+// Copyright © 2017 Arm Ltd. All rights reserved.
+// SPDX-License-Identifier: MIT
+//
+
+#pragma once
+
+#include "LayerTestResult.hpp"
+
+#include <backendsCommon/IBackendInternal.hpp>
+#include <backendsCommon/WorkloadFactory.hpp>
+
+LayerTestResult<float, 4> SubtractionTest(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<float, 4> SubtractionBroadcast1ElementTest(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<float, 4> SubtractionBroadcastTest(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<uint8_t, 4> SubtractionUint8Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<uint8_t, 4> SubtractionBroadcast1ElementUint8Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<uint8_t, 4> SubtractionBroadcastUint8Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<int16_t, 4> SubtractionInt16Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<int16_t, 4> SubtractionBroadcast1ElementInt16Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);
+
+LayerTestResult<int16_t, 4> SubtractionBroadcastInt16Test(
+    armnn::IWorkloadFactory& workloadFactory,
+    const armnn::IBackendInternal::IMemoryManagerSharedPtr& memoryManager);