Port CLGEMMConvolutionLayer

Details:
port CLWeightsReshapeKernel to ClWeightsReshapeKernel
port CLGEMMConvolutionLayer to ClGemmConvolution

Resolves: COMPMID-4515

Change-Id: I7d5b4ec72db2742f6eb9f3ffc88f717c35b4f2a3
Signed-off-by: Manuel Bottini <manuel.bottini@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5983
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>

2 files changed, 0 insertions, 288 deletions

diff --git a/src/core/CL/kernels/CLWeightsReshapeKernel.cpp b/src/core/CL/kernels/CLWeightsReshapeKernel.cpp
deleted file mode 100644
index 45e3505d0f..0000000000
--- a/src/core/CL/kernels/CLWeightsReshapeKernel.cpp
+++ /dev/null
@@ -1,167 +0,0 @@
-/*
- * Copyright (c) 2017-2021 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#include "src/core/CL/kernels/CLWeightsReshapeKernel.h"
-#include "arm_compute/core/CL/ICLTensor.h"
-#include "arm_compute/core/Error.h"
-#include "arm_compute/core/utils/misc/ShapeCalculator.h"
-#include "src/core/helpers/AutoConfiguration.h"
-#include "src/core/helpers/WindowHelpers.h"
-#include "support/StringSupport.h"
-
-namespace arm_compute
-{
-using namespace arm_compute::misc::shape_calculator;
-
-namespace
-{
-Status validate_arguments(const ITensorInfo *input, const ITensorInfo *biases, const ITensorInfo *output, unsigned int num_groups)
-{
-    ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, output);
-    ARM_COMPUTE_RETURN_ERROR_ON(input->data_type() == DataType::UNKNOWN);
-    ARM_COMPUTE_RETURN_ERROR_ON(num_groups == 0);
-    ARM_COMPUTE_RETURN_ERROR_ON(input->data_layout() == DataLayout::NHWC && num_groups > 1);
-    ARM_COMPUTE_RETURN_ERROR_ON(input->num_dimensions() > 4 && num_groups > 1);
-    ARM_COMPUTE_RETURN_ERROR_ON((input->dimension(3) % num_groups) != 0);
-
-    if(biases != nullptr)
-    {
-        ARM_COMPUTE_RETURN_ERROR_ON(!is_data_type_float(input->data_type()));
-        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, biases);
-        ARM_COMPUTE_RETURN_ERROR_ON((input->num_dimensions() == 4) && (biases->num_dimensions() != 1));
-        ARM_COMPUTE_RETURN_ERROR_ON((input->num_dimensions() == 5) && (biases->num_dimensions() != 2));
-        ARM_COMPUTE_RETURN_ERROR_ON((input->num_dimensions() == 4) && (biases->dimension(0) != input->tensor_shape()[3]));
-        ARM_COMPUTE_RETURN_ERROR_ON((input->num_dimensions() == 5) && (biases->dimension(0) != input->tensor_shape()[3] || biases->dimension(1) != input->tensor_shape()[4]));
-    }
-
-    // Checks performed when output is configured
-    if(output->total_size() != 0)
-    {
-        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DIMENSIONS(output->tensor_shape(), compute_weights_reshaped_shape(*input, biases != nullptr, num_groups));
-        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);
-        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_QUANTIZATION_INFO(input, output);
-    }
-
-    return Status{};
-}
-} // namespace
-
-CLWeightsReshapeKernel::CLWeightsReshapeKernel()
-    : _input(nullptr), _biases(nullptr), _output(nullptr)
-{
-    _type = CLKernelType::ELEMENTWISE;
-}
-
-void CLWeightsReshapeKernel::configure(const ICLTensor *input, const ICLTensor *biases, ICLTensor *output, unsigned int num_groups)
-{
-    configure(CLKernelLibrary::get().get_compile_context(), input, biases, output, num_groups);
-}
-
-void CLWeightsReshapeKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, const ICLTensor *biases, ICLTensor *output, unsigned int num_groups)
-{
-    ARM_COMPUTE_ERROR_ON_NULLPTR(input, output);
-
-    // Output tensor auto inizialitation if not yet initialized
-    auto_init_if_empty(*output->info(), input->info()->clone()->set_tensor_shape(compute_weights_reshaped_shape(*input->info(), (biases != nullptr), num_groups)));
-
-    // Perform validation step
-    ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input->info(),
-                                                  (biases != nullptr) ? biases->info() : nullptr,
-                                                  output->info(), num_groups));
-
-    auto padding_info = get_padding_info({ input, biases, output });
-
-    const DataType data_type = input->info()->data_type();
-
-    _biases = biases;
-    _output = output;
-    _input  = input;
-
-    // Create build options
-    CLBuildOptions build_opts;
-    build_opts.add_option("-DDATA_TYPE=" + get_cl_unsigned_type_from_element_size(data_size_from_type(data_type)));
-    build_opts.add_option("-DNUM_GROUPS=" + support::cpp11::to_string(num_groups));
-    build_opts.add_option_if(biases != nullptr, "-DHAS_BIAS");
-
-    // Create kernel
-    _kernel = create_kernel(compile_context, "reshape_to_columns", build_opts.options());
-
-    // Configure window
-    Window win = calculate_max_window(*input->info(), Steps());
-    ICLKernel::configure_internal(win);
-
-    ARM_COMPUTE_ERROR_ON(has_padding_changed(padding_info));
-}
-
-Status CLWeightsReshapeKernel::validate(const ITensorInfo *input, const ITensorInfo *biases, const ITensorInfo *output, unsigned int num_groups)
-{
-    ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input, biases, output, num_groups));
-    return Status{};
-}
-
-void CLWeightsReshapeKernel::run(const Window &window, cl::CommandQueue &queue)
-{
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_MISMATCHING_WINDOWS(ICLKernel::window(), window);
-
-    Window out_window;
-    out_window.use_tensor_dimensions(_output->info()->tensor_shape());
-
-    Window in_slice  = window.first_slice_window_3D();
-    Window out_slice = out_window.first_slice_window_2D();
-
-    Window biases_window;
-    Window biases_slice;
-
-    unsigned int idx = num_arguments_per_3D_tensor() + num_arguments_per_2D_tensor();
-    idx += (_biases != nullptr) ? num_arguments_per_1D_tensor() : 0;
-    _kernel.setArg<cl_uint>(idx++, _input->info()->dimension(0));
-    _kernel.setArg<cl_uint>(idx++, _input->info()->dimension(1));
-    _kernel.setArg<cl_uint>(idx++, _input->info()->dimension(2));
-    _kernel.setArg<cl_uint>(idx++, _input->info()->dimension(3));
-    _kernel.setArg<cl_uint>(idx++, _output->info()->strides_in_bytes().z());
-
-    if(_biases != nullptr)
-    {
-        biases_window.use_tensor_dimensions(_biases->info()->tensor_shape());
-        biases_slice = biases_window.first_slice_window_1D();
-    }
-
-    do
-    {
-        // Set arguments
-        unsigned idx = 0;
-        add_3D_tensor_argument(idx, _input, in_slice);
-        add_2D_tensor_argument(idx, _output, out_slice);
-        if(_biases != nullptr)
-        {
-            add_1D_tensor_argument(idx, _biases, biases_slice);
-            ARM_COMPUTE_UNUSED(biases_window.slide_window_slice_1D(biases_slice));
-        }
-
-        // Run kernel
-        enqueue(queue, *this, in_slice, lws_hint());
-    }
-    while(window.slide_window_slice_4D(in_slice) && out_window.slide_window_slice_2D(out_slice));
-}
-} // namespace arm_compute
diff --git a/src/core/CL/kernels/CLWeightsReshapeKernel.h b/src/core/CL/kernels/CLWeightsReshapeKernel.h
deleted file mode 100644
index 9ac60a7a1a..0000000000
--- a/src/core/CL/kernels/CLWeightsReshapeKernel.h
+++ /dev/null
@@ -1,121 +0,0 @@
-/*
- * Copyright (c) 2017-2021 Arm Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#ifndef ARM_COMPUTE_CLWEIGHTSRESHAPEKERNEL_H
-#define ARM_COMPUTE_CLWEIGHTSRESHAPEKERNEL_H
-
-#include "src/core/CL/ICLKernel.h"
-
-namespace arm_compute
-{
-/** OpenCL kernel to perform reshaping on the weights used by convolution and locally connected layer
- *
- * Rearranges each 3-dimensional kernel to a single row leading to a matrix with linearized kernels.
- * In combination with the @ref opencl::kernels::ClIm2ColKernel can transform a convolution to a matrix multiplication.
- *
- * For example assuming a 3D weight kernel of 3x3 dimensions and depth of 2 we have:
- * @f[
- * \left( \begin{array}{ccc}
- * a000 & a001 & a002 \\
- * a010 & a011 & a012 \\
- * a020 & a021 & a022 \\
- * \end{array} \right)
- * \left( \begin{array}{ccc}
- * a100 & a101 & a102 \\
- * a110 & a111 & a112 \\
- * a120 & a121 & a122 \\
- * \end{array} \right)
- * \rightarrow
- * \left( \begin{array}{ccccccccc}
- * a000 & a001 & a002 & a010 & a011 & a012 & a020 & a021 & a022 & a100 & a101 & a102 & a110 & a111 & a112 & a120 & a121 & a122 \\
- * \end{array} \right)
- * @f]
- */
-class CLWeightsReshapeKernel : public ICLKernel
-{
-public:
-    /** Constructor.*/
-    CLWeightsReshapeKernel();
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLWeightsReshapeKernel(const CLWeightsReshapeKernel &) = delete;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    CLWeightsReshapeKernel &operator=(const CLWeightsReshapeKernel &) = delete;
-    /** Allow instances of this class to be moved */
-    CLWeightsReshapeKernel(CLWeightsReshapeKernel &&) = default;
-    /** Allow instances of this class to be moved */
-    CLWeightsReshapeKernel &operator=(CLWeightsReshapeKernel &&) = default;
-    /** Default destructor */
-    ~CLWeightsReshapeKernel() = default;
-    /** Set the input and output of the kernel.
-     *
-     * @param[in]  input      The input tensor to convert. Weights are 4D tensor with dimensions [kernel_x, kernel_y, IFM, OFM] if shared,
-     *                        and 5D tensor with dimensions [kernel_x, kernel_y, IFM, OFM,  num_patches] if unshared. Data types supported: All
-     * @param[in]  biases     The shared biases tensor to append.  Bias is 1D tensor with dimensions [OFM] if shared and 2D tensor with
-     *                        dimensions [OFM, num_patches] if unshared. Data types supported: F16/F32, for quantized types this must be nullptr.
-     *                        @warning Appending biases to weights reshaped matrix is not supported for quantized asymmetric types.
-     * @param[out] output     The output tensor. Should be a 2D Tensor if there are no groups and the weights are not shared; a 3D Tensor otherwise.
-     *                        Data types supported: Same as @p input
-     * @param[in]  num_groups (Optional) Number of groups when performing a grouped convolution. num_groups != 1 is only supported for NCHW data layout
-     *                        Number of groups greater than one are only supported for NCHW data layout, and the number of weights must be a multiple of it.
-     */
-    void configure(const ICLTensor *input, const ICLTensor *biases, ICLTensor *output, unsigned int num_groups = 1);
-    /** Set the input and output of the kernel.
-     *
-     * @param[in]  compile_context The compile context to be used.
-     * @param[in]  input           The input tensor to convert. Weights are 4D tensor with dimensions [kernel_x, kernel_y, IFM, OFM] if shared,
-     *                             and 5D tensor with dimensions [kernel_x, kernel_y, IFM, OFM,  num_patches] if unshared. Data types supported: All
-     * @param[in]  biases          The shared biases tensor to append.  Bias is 1D tensor with dimensions [OFM] if shared and 2D tensor with
-     *                             dimensions [OFM, num_patches] if unshared. Data types supported: F16/F32, for quantized types this must be nullptr.
-     *                             @warning Appending biases to weights reshaped matrix is not supported for quantized asymmetric types.
-     * @param[out] output          The output tensor. Should be a 2D Tensor if there are no groups and the weights are not shared; a 3D Tensor otherwise.
-     *                             Data types supported: Same as @p input
-     * @param[in]  num_groups      (Optional) Number of groups when performing a grouped convolution. num_groups != 1 is only supported for NCHW data layout
-     *                             Number of groups greater than one are only supported for NCHW data layout, and the number of weights must be a multiple of it.
-     */
-    void configure(const CLCompileContext &compile_context, const ICLTensor *input, const ICLTensor *biases, ICLTensor *output, unsigned int num_groups = 1);
-    /** Static function to check if given info will lead to a valid configuration of @ref CLWeightsReshapeKernel
-     *
-     * @param[in] input      The input tensor to convert. Weights are 4D tensor with dimensions [kernel_x, kernel_y, IFM, OFM] if shared,
-     *                       and 5D tensor with dimensions [kernel_x, kernel_y, IFM, OFM,  num_patches] if unshared. Data types supported: All
-     * @param[in] biases     The shared biases tensor to append.  Bias is 1D tensor with dimensions [OFM] if shared and 2D tensor with
-     *                       dimensions [OFM, num_patches] if unshared. Data types supported: F16/F32, for quantized types this must be nullptr.
-     *                       @warning Appending biases to weights reshaped matrix is not supported for quantized asymmetric types.
-     * @param[in] output     The output tensor. Should be a 2D Tensor if there are no groups and the weights are not shared; a 3D Tensor otherwise.
-     *                       Data types supported: Same as @p input
-     * @param[in] num_groups (Optional) Number of groups when performing a grouped convolution. num_groups != 1 is only supported for NCHW data layout
-     *                       Number of groups greater than one are only supported for NCHW data layout, and the number of weights must be a multiple of it.
-     *
-     * @return a status
-     */
-    static Status validate(const ITensorInfo *input, const ITensorInfo *biases, const ITensorInfo *output, unsigned int num_groups = 1);
-
-    // Inherited methods overridden:
-    void run(const Window &window, cl::CommandQueue &queue) override;
-
-private:
-    const ICLTensor *_input;
-    const ICLTensor *_biases;
-    ICLTensor       *_output;
-};
-} // namespace arm_compute
-#endif /*ARM_COMPUTE_CLWEIGHTSRESHAPEKERNEL_H */
\ No newline at end of file