diff options
| author | Mike Kelly <mike.kelly@arm.com> | 2020-11-12 10:58:48 +0000 |
|---|---|---|
| committer | Jim Flynn <jim.flynn@arm.com> | 2020-11-13 14:25:30 +0000 |
| commit | 07810fc2fcdd34db74222d90cc73ef12a88e7b78 (patch) | |
| tree | 8becef8453674822d079815b06ae37310b97d2cf /src/backends/neon/workloads/NeonDivisionWorkload.cpp | |
| parent | 8502adeafbbb1db0acefa62560d93453e38dcadb (diff) | |
| download | armnn-07810fc2fcdd34db74222d90cc73ef12a88e7b78.tar.gz | |
IVGCVSW-5328-5329 Fuse Activation
* Added Fused Activation Optimization to both CL and Neon backends.
* Added Fused Activation support to all the CL and Neon workloads
that support it.
* Changed ProfilingTest network to be a Convolution layer
followed by an Abs layer rather than an Activation layer.
* Added IBackendInternal::OptimizeSubgraphView function that can accept a
ModelOptions.
* Network will now call OptimizeSubgraphView passing in the ModelOptions.
Signed-off-by: Keith Davis <keith.davis@arm.com>
Signed-off-by: Mike Kelly <mike.kelly@arm.com>
Signed-off-by: Teresa Charlin <teresa.charlinreyes@arm.com>
Change-Id: Ib536ac3cbafc7d9b35c139ad9a65b7735262cd9d
Diffstat (limited to 'src/backends/neon/workloads/NeonDivisionWorkload.cpp')
| -rw-r--r-- | src/backends/neon/workloads/NeonDivisionWorkload.cpp | 20 |
1 files changed, 15 insertions, 5 deletions
diff --git a/src/backends/neon/workloads/NeonDivisionWorkload.cpp b/src/backends/neon/workloads/NeonDivisionWorkload.cpp index fc353f136d..1a26d9510a 100644 --- a/src/backends/neon/workloads/NeonDivisionWorkload.cpp +++ b/src/backends/neon/workloads/NeonDivisionWorkload.cpp @@ -6,23 +6,31 @@ #include "NeonDivisionWorkload.hpp" #include <aclCommon/ArmComputeTensorUtils.hpp> +#include <aclCommon/ArmComputeUtils.hpp> + #include <armnn/utility/PolymorphicDowncast.hpp> + #include <backendsCommon/CpuTensorHandle.hpp> namespace armnn { arm_compute::Status NeonDivisionWorkloadValidate(const TensorInfo& input0, - const TensorInfo& input1, - const TensorInfo& output) + const TensorInfo& input1, + const TensorInfo& output, + const ActivationDescriptor* activationDescriptor) { const arm_compute::TensorInfo aclInput0 = armcomputetensorutils::BuildArmComputeTensorInfo(input0); const arm_compute::TensorInfo aclInput1 = armcomputetensorutils::BuildArmComputeTensorInfo(input1); const arm_compute::TensorInfo aclOutput = armcomputetensorutils::BuildArmComputeTensorInfo(output); + const arm_compute::ActivationLayerInfo activationInfo = ConvertActivationDescriptorToAclActivationLayerInfo( + activationDescriptor); + return arm_compute::NEElementwiseDivision::validate(&aclInput0, - &aclInput1, - &aclOutput); + &aclInput1, + &aclOutput, + activationInfo); } NeonDivisionWorkload::NeonDivisionWorkload(const DivisionQueueDescriptor& descriptor, @@ -35,7 +43,9 @@ NeonDivisionWorkload::NeonDivisionWorkload(const DivisionQueueDescriptor& descri arm_compute::ITensor& input1 = PolymorphicDowncast<IAclTensorHandle*>(m_Data.m_Inputs[1])->GetTensor(); arm_compute::ITensor& output = PolymorphicDowncast<IAclTensorHandle*>(m_Data.m_Outputs[0])->GetTensor(); - m_DivLayer.configure(&input0, &input1, &output); + const arm_compute::ActivationLayerInfo activationInfo = ConvertAdditionalInfoToAclActivationLayerInfo(descriptor); + + m_DivLayer.configure(&input0, &input1, &output, activationInfo); } void NeonDivisionWorkload::Execute() const |