From 27400b90a9cb3fe028c5b724b58ce0e82d89b5e8 Mon Sep 17 00:00:00 2001 From: Michele Di Giorgio Date: Thu, 1 Nov 2018 13:44:05 +0000 Subject: 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 --- arm_compute/runtime/CL/functions/CLWidthConcatenateLayer.h | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-) (limited to 'arm_compute/runtime/CL/functions/CLWidthConcatenateLayer.h') 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 @@ -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 _concat_kernels_vector; + CLWidthConcatenate2TensorsKernel _concat_x2_kernel; + CLWidthConcatenate4TensorsKernel _concat_x4_kernel; unsigned int _num_inputs; }; } // namespace arm_compute -- cgit v1.2.1