diff options
author | Francis Murtagh <francis.murtagh@arm.com> | 2018-08-29 12:42:10 +0100 |
---|---|---|
committer | Matthew Bentham <matthew.bentham@arm.com> | 2018-09-17 17:21:23 +0100 |
commit | e7a86a4a3363993fb41b1ea62f23b3643b8b0c78 (patch) | |
tree | 6d054cae92a13412129525e4f9ea441e7d8c6b73 /src/armnn/backends/ClLayerSupport.cpp | |
parent | a68241066c3e797dab70f515d2c55aaa74abf564 (diff) | |
download | armnn-e7a86a4a3363993fb41b1ea62f23b3643b8b0c78.tar.gz |
IVGCVSW-1200 Division layer
*IVGCVSW-1772 Create QueueDescriptors
*IVGCVSW-1773 Add a CL implementation of the DivisionWorkload
*IVGCVSW-1774 Add Neon implementation of the DivisionWorkload
*IVGCVSW-1775 Add a Ref implementation of the DivisionWorkload
*IVGCVSW-1776 Add a Division Layer
* Added simple division unit tests with broadcasting
Change-Id: I05751fb7f868789f6c06f91e8d25e52b4f12ab5e
Diffstat (limited to 'src/armnn/backends/ClLayerSupport.cpp')
-rw-r--r-- | src/armnn/backends/ClLayerSupport.cpp | 14 |
1 files changed, 13 insertions, 1 deletions
diff --git a/src/armnn/backends/ClLayerSupport.cpp b/src/armnn/backends/ClLayerSupport.cpp index b00a218a72..77e74f57c6 100644 --- a/src/armnn/backends/ClLayerSupport.cpp +++ b/src/armnn/backends/ClLayerSupport.cpp @@ -17,11 +17,11 @@ #include "ClWorkloads/ClAdditionFloat32Workload.hpp" #include "ClWorkloads/ClActivationFloat32Workload.hpp" #include "ClWorkloads/ClBatchNormalizationFloat32Workload.hpp" - #include "ClWorkloads/ClConvertFp16ToFp32Workload.hpp" #include "ClWorkloads/ClConvertFp32ToFp16Workload.hpp" #include "ClWorkloads/ClConvolution2dBaseWorkload.hpp" #include "ClWorkloads/ClDepthwiseConvolutionBaseWorkload.hpp" +#include "ClWorkloads/ClDivisionFloatWorkload.hpp" #include "ClWorkloads/ClL2NormalizationFloat32Workload.hpp" #include "ClWorkloads/ClMultiplicationFloat32Workload.hpp" #include "ClWorkloads/ClFullyConnectedFloat32Workload.hpp" @@ -238,6 +238,18 @@ bool IsDepthwiseConvolutionSupportedCl(const TensorInfo& input, biases); } +bool IsDivisionSupportedCl(const TensorInfo& input0, + const TensorInfo& input1, + const TensorInfo& output, + std::string* reasonIfUnsupported) +{ + FORWARD_WORKLOAD_VALIDATE_FUNC(ClDivisionWorkloadValidate, + reasonIfUnsupported, + input0, + input1, + output); +} + bool IsFullyConnectedSupportedCl(const TensorInfo& input, const TensorInfo& output, const TensorInfo& weights, |