diff options
| author | Mike Kelly <mike.kelly@arm.com> | 2023-07-07 15:43:06 +0100 |
|---|---|---|
| committer | Mike Kelly <mike.kelly@arm.com> | 2023-07-14 00:00:53 +0100 |
| commit | 4cc341cf8b5a6e6bb0543504cbbfde6fa11a2cdb (patch) | |
| tree | 7cac128e9ec6f2fd27f1afdb55f44b870f39e0b3 /src/armnn/layers/ConcatLayer.cpp | |
| parent | 6963b33221c23af4a8eff19ff4a5773230b0befd (diff) | |
| download | armnn-4cc341cf8b5a6e6bb0543504cbbfde6fa11a2cdb.tar.gz | |
IVGCVSW-7830 Add backend optimizations to remove Reshapes where possible
* Added optimization to remove reshapes for Neon and Ref Backends
by using overridden TensorInfos
* Added ability to delete Subgraphs during Optimization
* Fixed naming error in NeonEndToEndTests and CLEndToEndTests
* Added LayerNameAndTypeCheck for testing.
* Fixed error where layers were not marked as altered when removed in
CLBackend
Signed-off-by: Mike Kelly <mike.kelly@arm.com>
Change-Id: I1ac25cd4ec9821470d961831ae2c8d24882276cc
Diffstat (limited to 'src/armnn/layers/ConcatLayer.cpp')
| -rw-r--r-- | src/armnn/layers/ConcatLayer.cpp | 3 |
1 files changed, 2 insertions, 1 deletions
diff --git a/src/armnn/layers/ConcatLayer.cpp b/src/armnn/layers/ConcatLayer.cpp index 7a1b689b2c..4629bf245e 100644 --- a/src/armnn/layers/ConcatLayer.cpp +++ b/src/armnn/layers/ConcatLayer.cpp @@ -120,7 +120,7 @@ void ConcatLayer::CreateTensors(const TensorHandleFactoryRegistry& registry, // 3) the input does not come from a Constant layer or input layer // 4) the input is only read by this concat layer // 5) if concat along x or y (2 innermost dimensions) and the previous layers do not require padding - // 6) the input does not have an Overridden TensorInfo + // 6) neither the inputs nor the output have an Overridden TensorInfo if (slot && parentInfo.IsTypeSpaceMatch(info) && //(1) factoryId == slot->GetTensorHandleFactoryId() && //(2) @@ -128,6 +128,7 @@ void ConcatLayer::CreateTensors(const TensorHandleFactoryRegistry& registry, slot->GetOwningLayer().GetType() != LayerType::Input && //(3) slot->GetNumConnections() == 1 && canUseSubTensorOnXorY && //(5) + !GetOutputSlot(0).GetConnection(0)->IsTensorInfoOverridden() && //(6) !currentLayer->GetInputSlot(i).IsTensorInfoOverridden()) //(6) { ARMNN_NO_DEPRECATE_WARN_BEGIN |