1 files changed, 46 insertions, 1 deletions
diff --git a/src/backends/backendsCommon/LayerSupportBase.cpp b/src/backends/backendsCommon/LayerSupportBase.cpp
index 220590e197..89a0772602 100644
--- a/src/backends/backendsCommon/LayerSupportBase.cpp
+++ b/src/backends/backendsCommon/LayerSupportBase.cpp
@@ -4,13 +4,13 @@
 //
 
 #include <armnn/Deprecated.hpp>
-#include <armnn/Descriptors.hpp>
 #include <armnn/Exceptions.hpp>
 #include <armnn/Types.hpp>
 
 #include <backendsCommon/LayerSupportBase.hpp>
 
 #include <armnn/utility/IgnoreUnused.hpp>
+#include <armnn/utility/PolymorphicDowncast.hpp>
 
 namespace
 {
@@ -37,6 +37,51 @@ bool DefaultLayerSupport(const char* func,
 namespace armnn
 {
 
+bool LayerSupportBase::IsLayerSupported(const LayerType& type,
+                                        const std::vector<TensorInfo>& infos,
+                                        const BaseDescriptor& descriptor,
+                                        const Optional<LstmInputParamsInfo>&,
+                                        const Optional<QuantizedLstmInputParamsInfo>&,
+                                        Optional<std::string&> reasonIfUnsupported) const
+{
+    switch(type)
+    {
+        case LayerType::MemCopy:
+            return IsMemCopySupported(infos[0], infos[1], reasonIfUnsupported);
+        case LayerType::MemImport:
+            return IsMemImportSupported(infos[0], infos[1], reasonIfUnsupported);
+        case LayerType::StandIn:
+        {
+            auto desc = *(PolymorphicDowncast<const StandInDescriptor*>(&descriptor));
+
+            if (infos.size() != (desc.m_NumInputs + desc.m_NumOutputs))
+            {
+                throw InvalidArgumentException("Number of StandIn layer TensorInfos does not equal "
+                                               "the combined number of input and output slots assigned "
+                                               "to the StandIn descriptor");
+            }
+
+            std::vector<const TensorInfo*> inputInfos;
+            for (uint32_t i = 0; i < desc.m_NumInputs; i++)
+            {
+                inputInfos.push_back(&infos[i]);
+            }
+            std::vector<const TensorInfo*> outputInfos;
+            for (uint32_t i = desc.m_NumInputs; i < infos.size(); i++)
+            {
+                outputInfos.push_back(&infos[i]);
+            }
+
+            return IsStandInSupported(inputInfos,
+                                      outputInfos,
+                                      desc,
+                                      reasonIfUnsupported);
+        }
+        default:
+            return DefaultLayerSupport(__func__, __FILE__, __LINE__, reasonIfUnsupported);
+    }
+}
+
 bool LayerSupportBase::IsActivationSupported(const TensorInfo&, // input
                                              const TensorInfo&, //output
                                              const ActivationDescriptor&, // descriptor