COMPMID-791: Generic Depthwise Convolution Layer NEON QASYMM8

Change-Id: I33cf54e68f6c097ac58b6f16c3f9a720978f09cd
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/117289
Tested-by: Jenkins <bsgcomp@arm.com>
Reviewed-by: Michalis Spyrou <michalis.spyrou@arm.com>
Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>

1 files changed, 17 insertions, 3 deletions

diff --git a/arm_compute/core/NEON/kernels/NEDepthwiseVectorToTensorKernel.h b/arm_compute/core/NEON/kernels/NEDepthwiseVectorToTensorKernel.h
index 19000905b0..458cbd7812 100644
--- a/arm_compute/core/NEON/kernels/NEDepthwiseVectorToTensorKernel.h
+++ b/arm_compute/core/NEON/kernels/NEDepthwiseVectorToTensorKernel.h
@@ -56,7 +56,7 @@ public:
     NEDepthwiseVectorToTensorKernel &operator=(NEDepthwiseVectorToTensorKernel &&) = default;
     /** Set the input and output of the kernel.
      *
-     * @param[in]  input  The input vector to convert. Data type supported: F32.
+     * @param[in]  input  The input vector to convert. Data type supported: QASYMM8/S32/F32.
      * @param[out] output The output tensor. 3 lower dimensions represent a single input [width, height, IFM]. Data type supported: same as @p input.
      * @param[in]  conv_w The converted tensor's width.
      * @param[in]  conv_h The converted tensor's height.
@@ -67,8 +67,22 @@ public:
     void run(const Window &window, const ThreadInfo &info) override;
 
 private:
-    const ITensor *_input;
-    ITensor       *_output;
+    /** Template function to run the vector to tensor reshape used for the depthwise convolution layer case
+     *
+     * @param[in] window Region on which to execute the kernel. (Must be a valid region of the window returned by window()).
+     */
+    template <typename T>
+    void vector_to_tensor(const Window &window);
+    /** Common signature for all the specialised depthwise vector to tensor functions
+     *
+     * @param[in] window Region on which to execute the kernel.
+     */
+    using DepthwiseVectorToTensorFunctionPtr = void (NEDepthwiseVectorToTensorKernel::*)(const Window &window);
+
+private:
+    DepthwiseVectorToTensorFunctionPtr _func;
+    const ITensor                     *_input;
+    ITensor                           *_output;
     std::pair<size_t, size_t> _conv_dims;
 };
 } // arm_compute