Optimize the calling of IsLayerSupported().

!armnn:8742

  * Done as part of 22.11/23.02 innovation days.
  * IsLayerSupported() is called in model prepare (delegate, android-nn-driver and shim/support_library)
    and again in ArmNN once model optimization is performed.
  * From calling IsLayerSupported() the first time, we should know that the layers are supported
    and what backend they are supported on.
  * Solution is to set the BackendId of the IConnectableLayer when IsLayerSupported() is called the first time,
  * In the Optimize() function we then check if the backend is set. If so, we do not call IsLayerSupported() again.
  * In the case a layer that is supported gets optimized, then the BackendId of that layer get set to "Unknown"
    for the new optimized layer and IsLayerSupported() will get called on the newly optimized layer.

Signed-off-by: Cathal Corbett <cathal.corbett@arm.com>
Change-Id: Ie5c6c9cd10d81f90b1ee78dd6e3442f353b6c109

1 files changed, 4 insertions, 1 deletions

diff --git a/ConversionUtils.cpp b/ConversionUtils.cpp
index f441e7dc..c691c554 100644
--- a/ConversionUtils.cpp
+++ b/ConversionUtils.cpp
@@ -1,5 +1,5 @@
 //
-// Copyright © 2017 Arm Ltd and Contributors. All rights reserved.
+// Copyright © 2017,2022 Arm Ltd and Contributors. All rights reserved.
 // SPDX-License-Identifier: MIT
 //
 
@@ -191,10 +191,12 @@ armnn::IConnectableLayer* ProcessActivation(const armnn::TensorInfo& tensorInfo,
         }
 
         bool isSupported = false;
+        armnn::BackendId setBackend;
         FORWARD_LAYER_SUPPORT_FUNC(__func__,
                                    IsActivationSupported,
                                    data.m_Backends,
                                    isSupported,
+                                   setBackend,
                                    prevLayer->GetOutputSlot(0).GetTensorInfo(),
                                    tensorInfo,
                                    activationDesc);
@@ -204,6 +206,7 @@ armnn::IConnectableLayer* ProcessActivation(const armnn::TensorInfo& tensorInfo,
         }
 
         activationLayer = data.m_Network->AddActivationLayer(activationDesc);
+        activationLayer->SetBackendId(setBackend);
 
         prevLayer->GetOutputSlot(0).Connect(activationLayer->GetInputSlot(0));
         activationLayer->GetOutputSlot(0).SetTensorInfo(tensorInfo);