From 7075fe2c5ee6f7cfe7cfd9454d905235e70b9ac4 Mon Sep 17 00:00:00 2001 From: Adnan AlSinan Date: Mon, 5 Jul 2021 13:12:52 +0100 Subject: Reorganize the kernels into nhwc, nchw and common folders The Following kernels have been split into nchw/nhwc kernels files: - batchnormalization_layer - batch_to_space - channel_shuffle - depth_to_space - dequantization_layer - im2col - normalization_layer - normalize_planar_yuv_layer - normalize_planar_yuv_layer_quantized - pooling_layer - pooling_layer_quantized - remap - reorg_layer - scale - scale_quantized - space_to_batch - space_to_depth - upsample_layer - winograd_filter_transform - winograd_input_transform - winograd_output_transform The following kernels have been moved to nchw folder: - direct_convolution1x1 - direct_convolution3x3 - direct_convolution5x5 - direct_convolution_quantized - prior_box_layer The following kernels have been moved to nhwc folder: - direct_convolution - dwc_native_fp_nhwc - dwc_native_quantized_nhwc The following kernels have been removed: - sobel_filter While the rest kerenls have been moved to the common folder. Partially resolves COMPMID-4453 Signed-off-by: Adnan AlSinan Change-Id: Ic327ac935687ec351c610c65a3c6357f364a5a58 Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5919 Tested-by: Arm Jenkins Reviewed-by: Georgios Pinitas Comments-Addressed: Arm Jenkins --- src/core/CL/cl_kernels/elementwise_unary.cl | 93 ----------------------------- 1 file changed, 93 deletions(-) delete mode 100644 src/core/CL/cl_kernels/elementwise_unary.cl (limited to 'src/core/CL/cl_kernels/elementwise_unary.cl') diff --git a/src/core/CL/cl_kernels/elementwise_unary.cl b/src/core/CL/cl_kernels/elementwise_unary.cl deleted file mode 100644 index d2d9d97d33..0000000000 --- a/src/core/CL/cl_kernels/elementwise_unary.cl +++ /dev/null @@ -1,93 +0,0 @@ -/* - * Copyright (c) 2018-2021 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 "helpers.h" -#include "warp_helpers.h" - -#if defined(DATA_TYPE) && defined(OPERATION) - -// Calculate exponential -#define exp_op(input) exp(input) -// Calculate reverse square root -#define rsqrt_op(input) rsqrt(input) -// Calculate negative -#define neg_op(input) (-input) -// Calculate sine -#define sin_op(input) sin(input) -// Calculate abs for floating point values -#define fabs_op(input) fabs(input) -// Calculate natural_log -#define natural_log_op(input) log(input) -// Calculate round (Cannot use round function as it rounds halfway cases away from zero). -#if defined(VEC_SIZE) -#define VEC_TYPE VEC_DATA_TYPE(DATA_TYPE, VEC_SIZE) -#define round_op(input) CONVERT(CONVERT_SAT_ROUND(input, VEC_DATA_TYPE(int, VEC_SIZE), rte), VEC_TYPE) -#define logical_not_op(input) CONVERT(CONVERT(!input, VEC_TYPE) & ((VEC_TYPE)0x1), VEC_TYPE) -#else // defined(VEC_SIZE) -#define round_op(input) CONVERT(CONVERT_SAT_ROUND(input, int, rte), DATA_TYPE) -#define logical_not_op(input) ((!input) & 0x1) -#endif // defined(VEC_SIZE) - -/** Applies element wise unary operator in a tensor. - * - * @param[in] in_ptr Pointer to the source image. Supported data types: F16/32. - * @param[in] in_stride_x Stride of the source tensor in X dimension (in bytes) - * @param[in] in_step_x in_stride_x * number of elements along X processed per workitem(in bytes) - * @param[in] in_stride_y Stride of the source tensor in Y dimension (in bytes) - * @param[in] in_step_y in_stride_y * number of elements along Y processed per workitem(in bytes) - * @param[in] in_stride_z Stride of the source tensor in Z dimension (in bytes) - * @param[in] in_step_z in_stride_z * number of elements along Z processed per workitem(in bytes) - * @param[in] in_offset_first_element_in_bytes Offset of the first element in the source image - * @param[out] out_ptr Pointer to the destination image. Supported data types: F16/32. - * @param[in] out_stride_x Stride of the destination image in X dimension (in bytes) - * @param[in] out_step_x out_stride_x * number of elements along X processed per workitem(in bytes) - * @param[in] out_step_y Stride of the destination tensor in Y dimension (in bytes) - * @param[in] out_step_y out_stride_y * number of elements along Y processed per workitem(in bytes) - * @param[in] out_stride_z Stride of the destination tensor in Z dimension (in bytes) - * @param[in] out_step_z out_stride_z * number of elements along Z processed per workitem(in bytes) - * @param[in] out_offset_first_element_in_bytes Offset of the first element in the destination image - */ -__kernel void elementwise_unary( - TENSOR3D_DECLARATION(in), - TENSOR3D_DECLARATION(out)) -{ - Tensor3D in = CONVERT_TO_TENSOR3D_STRUCT(in); - Tensor3D out = CONVERT_TO_TENSOR3D_STRUCT(out); - -#if defined(VEC_SIZE) && defined(LAST_ACCESSED_X) - // Check if access on width gets out of bounds - // If it does shift access vector to access elements within bounds - const int xi = (int)(get_global_id(0) * VEC_SIZE); - in.ptr -= max(xi - (int)LAST_ACCESSED_X, 0) * in_stride_x; - out.ptr -= max(xi - (int)LAST_ACCESSED_X, 0) * out_stride_x; - - VEC_DATA_TYPE(DATA_TYPE, VEC_SIZE) - data = VLOAD(VEC_SIZE)(0, (__global DATA_TYPE *)in.ptr); - - VSTORE(VEC_SIZE) - (OPERATION(data), 0, (__global DATA_TYPE *)out.ptr); -#else // !defined(VEC_SIZE) || !defined(LAST_ACCESSED_X) - *((__global DATA_TYPE *)(out.ptr)) = (DATA_TYPE)(OPERATION(*((__global DATA_TYPE *)in.ptr))); -#endif // defined(VEC_SIZE) && defined(LAST_ACCESSED_X) -} -#endif // defined(DATA_TYPE) && defined(OPERATION) -- cgit v1.2.1