IVGCVSW-2043 - Merger using ACL for innermost concat axis

     * Add ClMergerWorkload and NeonMergerWorkload to call ACL for innermost concat axis
     * Modify layer support to call ClMergerWorkloadValidate and NeonMergerWorkloadValidate when concat axis is inner most
     * Add m_ConcatAxis to MergerDescriptor
     * Modify MergerQueueDescriptor::Validate to check sub tensor only when using subtensor

!android-nn-driver:166

Change-Id: I56676b43964c8d6d726387b41b3cc34a512c0f0a

4 files changed, 15 insertions, 1 deletions

diff --git a/src/backends/backendsCommon/ILayerSupport.cpp b/src/backends/backendsCommon/ILayerSupport.cpp
index dc106e344e..55dd447369 100644
--- a/src/backends/backendsCommon/ILayerSupport.cpp
+++ b/src/backends/backendsCommon/ILayerSupport.cpp
@@ -192,6 +192,7 @@ bool ILayerSupport::IsMeanSupported(const TensorInfo& input,
 }
 
 bool ILayerSupport::IsMergerSupported(const std::vector<const TensorInfo*> inputs,
+                                      const TensorInfo& output,
                                       const OriginsDescriptor& descriptor,
                                       Optional<std::string&> reasonIfUnsupported) const
 {
diff --git a/src/backends/backendsCommon/WorkloadData.cpp b/src/backends/backendsCommon/WorkloadData.cpp
index e1146543ff..18ab4a8709 100644
--- a/src/backends/backendsCommon/WorkloadData.cpp
+++ b/src/backends/backendsCommon/WorkloadData.cpp
@@ -390,6 +390,16 @@ void MergerQueueDescriptor::Validate(const WorkloadInfo& workloadInfo) const
         throw InvalidArgumentException("MergerQueueDescriptor: At least one TensorInfo output needs to be provided.");
     }
 
+    if(m_Parameters.GetConcatAxis() > workloadInfo.m_InputTensorInfos[0].GetShape().GetNumDimensions())
+    {
+        throw InvalidArgumentException("Invalid Concatenation Axis provided");
+    }
+
+    if (workloadInfo.m_InputTensorInfos[0].GetShape().GetNumDimensions() - m_Parameters.GetConcatAxis() == 1)
+    {
+        return;
+    }
+
     if (workloadInfo.m_InputTensorInfos.size() != m_ViewOrigins.size())
     {
         throw InvalidArgumentException(
diff --git a/src/backends/backendsCommon/WorkloadFactory.cpp b/src/backends/backendsCommon/WorkloadFactory.cpp
index dc38f1a721..2e2824e521 100644
--- a/src/backends/backendsCommon/WorkloadFactory.cpp
+++ b/src/backends/backendsCommon/WorkloadFactory.cpp
@@ -471,7 +471,9 @@ bool IWorkloadFactory::IsLayerSupported(const BackendId& backendId,
             auto endPtr = boost::make_transform_iterator(inputs.end(), getTensorInfoPtr);
             std::vector<const TensorInfo*> inputPtrs(beginPtr, endPtr);
 
-            result = layerSupportObject->IsMergerSupported(inputPtrs, cLayer->GetParameters(), reason);
+            const TensorInfo& output = layer.GetOutputSlot(0).GetTensorInfo();
+
+            result = layerSupportObject->IsMergerSupported(inputPtrs, output, cLayer->GetParameters(), reason);
             break;
         }
         case LayerType::Multiplication:
diff --git a/src/backends/backendsCommon/test/LayerTests.cpp b/src/backends/backendsCommon/test/LayerTests.cpp
index bd8b38da01..f5689e74bf 100755
--- a/src/backends/backendsCommon/test/LayerTests.cpp
+++ b/src/backends/backendsCommon/test/LayerTests.cpp
@@ -2235,6 +2235,7 @@ void Concatenate(
     }
 
     armnn::OriginsDescriptor viewsDescriptor = CreateMergerDescriptorForConcatenation(inputTensorInfos, concatDim);
+    queueDescriptor.m_Parameters = viewsDescriptor;
 
     queueDescriptor.m_ViewOrigins.reserve(viewsDescriptor.GetNumViews());
     for (unsigned int i = 0; i < viewsDescriptor.GetNumViews(); ++i)