IVGCVSW-6260 ConstTensorsAsInput: Fully Connected Cl and Neon support.

  * IVGCVSW-6940 ConstTensorsAsInput: DepthwiseConvolution2d - Complete Neon and Cl Bug Fix
  * Bug fix to enable Cl and Neon Backend Compatibility ConstantTensorsAsInputs
  * Updated Cl and Neon FullyConnected workloads to handle constant
    weights and bias as inputs rather than reading from member variables.
  * Prevent non const weights and biases passing CL and NEON validate
    for Depthwise Convolution.

Signed-off-by: Cathal Corbett <cathal.corbett@arm.com>
Change-Id: I0f505ff5998a183152f843d0f6cc74327ba920e7

1 files changed, 4 insertions, 0 deletions

diff --git a/src/backends/aclCommon/ArmComputeTensorUtils.cpp b/src/backends/aclCommon/ArmComputeTensorUtils.cpp
index 1960332ccf..01c071d43b 100644
--- a/src/backends/aclCommon/ArmComputeTensorUtils.cpp
+++ b/src/backends/aclCommon/ArmComputeTensorUtils.cpp
@@ -147,6 +147,10 @@ arm_compute::TensorShape BuildArmComputeTensorShape(const armnn::TensorShape& te
 
 // Utility function used to build a TensorInfo object, that can be used to initialise
 // ARM Compute Tensor and CLTensor allocators.
+// Note: this utility ignores the value of armnn::TensorInfo.IsConstant(). ACL tensors
+// default to constant but Arm NN ones default to non constant. In the cases where
+// we expect ACL to treat a tensor as constant that value must be set after this
+// utility has been called.
 arm_compute::TensorInfo BuildArmComputeTensorInfo(const armnn::TensorInfo& tensorInfo)
 {
     bool multiScales = tensorInfo.HasMultipleQuantizationScales();