diff options
author | Michele Di Giorgio <michele.digiorgio@arm.com> | 2018-11-01 13:44:05 +0000 |
---|---|---|
committer | Georgios Pinitas <georgios.pinitas@arm.com> | 2018-11-13 14:49:34 +0000 |
commit | 27400b90a9cb3fe028c5b724b58ce0e82d89b5e8 (patch) | |
tree | 4b7dd9d4b29653ada018172dae826fe3e6ef5e08 /arm_compute/runtime/CL | |
parent | bb081cac4f386eb6db6e9927fce27c7027dd7be5 (diff) | |
download | ComputeLibrary-27400b90a9cb3fe028c5b724b58ce0e82d89b5e8.tar.gz |
COMPMID-1707: Create 3 special CLWidthConcatenate kernel to concatenate 2/4 and 8 tensors (Part 1)
Creating special cases for concatening 2 and 4 tensors.
Change-Id: I6a739a494ae45011acb65369e353f9ef96970b90
Diffstat (limited to 'arm_compute/runtime/CL')
-rw-r--r-- | arm_compute/runtime/CL/functions/CLWidthConcatenateLayer.h | 8 |
1 files changed, 7 insertions, 1 deletions
diff --git a/arm_compute/runtime/CL/functions/CLWidthConcatenateLayer.h b/arm_compute/runtime/CL/functions/CLWidthConcatenateLayer.h index 44462b02b2..55b65dadc4 100644 --- a/arm_compute/runtime/CL/functions/CLWidthConcatenateLayer.h +++ b/arm_compute/runtime/CL/functions/CLWidthConcatenateLayer.h @@ -29,6 +29,8 @@ #include "arm_compute/core/Window.h" #include "arm_compute/runtime/IFunction.h" +#include "arm_compute/core/CL/kernels/CLWidthConcatenate2TensorsKernel.h" +#include "arm_compute/core/CL/kernels/CLWidthConcatenate4TensorsKernel.h" #include "arm_compute/core/CL/kernels/CLWidthConcatenateLayerKernel.h" #include <memory> @@ -40,7 +42,9 @@ class ICLTensor; /** Basic function to execute concatenate tensors along x axis. This function calls the following kernel: * - * -# @ref CLDepthConcatenateLayerKernel + * -# @ref CLWidthConcatenateLayerKernel + * -# @ref CLWidthConcatenate2TensorsKernel (if there are exactly 2 input tensors) + * -# @ref CLWidthConcatenate4TensorsKernel (if there are exactly 4 input tensors) * */ class CLWidthConcatenateLayer : public IFunction @@ -74,6 +78,8 @@ public: private: std::unique_ptr<CLWidthConcatenateLayerKernel[]> _concat_kernels_vector; + CLWidthConcatenate2TensorsKernel _concat_x2_kernel; + CLWidthConcatenate4TensorsKernel _concat_x4_kernel; unsigned int _num_inputs; }; } // namespace arm_compute |