1/* 2 * Copyright (c) 2019-2022 Arm Limited. 3 * 4 * SPDX-License-Identifier: MIT 5 * 6 * Permission is hereby granted, free of charge, to any person obtaining a copy 7 * of this software and associated documentation files (the "Software"), to 8 * deal in the Software without restriction, including without limitation the 9 * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or 10 * sell copies of the Software, and to permit persons to whom the Software is 11 * furnished to do so, subject to the following conditions: 12 * 13 * The above copyright notice and this permission notice shall be included in all 14 * copies or substantial portions of the Software. 15 * 16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE 19 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER 20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, 21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE 22 * SOFTWARE. 23 */ 24#include "helpers.h" 25 26#if defined(VEC_SIZE) && defined(DATA_TYPE) && defined(EPSILON) && defined(WIDTH) 27/** This function normalizes the input 2D tensor across the first dimension with respect to mean and standard deviation of the same dimension. 28 * 29 * @attention Vector size should be given as a preprocessor argument using -DVEC_SIZE=size. e.g. -DVEC_SIZE=16 30 * @attention Data type should be passed using the -DDATA_TYPE compile flag, e.g. -DDATA_TYPE=float 31 * @attention Width of the input tensor should be passed using the -DWIDTH compile flag, e.g. -DWIDTH=16 32 * @attention Normalization epsilon parameter should be given as a preprocessor argument with -DEPSILON=value. e.g. -DEPSILON=0.001f 33 * 34 * @param[in] input_ptr Pointer to the first source tensor. Supported data types: F16/F32 35 * @param[in] input_stride_x Stride of the first source tensor in X dimension (in bytes) 36 * @param[in] input_step_x input_stride_x * number of elements along X processed per workitem(in bytes) 37 * @param[in] input_stride_y Stride of the first source tensor in Y dimension (in bytes) 38 * @param[in] input_step_y input_stride_y * number of elements along Y processed per workitem(in bytes) 39 * @param[in] input_offset_first_element_in_bytes The offset of the first element in the first source tensor 40 * @param[out] output_ptr (Optional) Pointer to the destination tensor. Supported data types: same as @p input_ptr 41 * @param[in] output_stride_x (Optional) Stride of the destination tensor in X dimension (in bytes) 42 * @param[in] output_step_x (Optional) output_stride_x * number of elements along X processed per workitem(in bytes) 43 * @param[in] output_stride_y (Optional) Stride of the destination tensor in Y dimension (in bytes) 44 * @param[in] output_step_y (Optional) output_stride_y * number of elements along Y processed per workitem(in bytes) 45 * @param[in] output_offset_first_element_in_bytes (Optional) The offset of the first element in the destination tensor 46 */ 47__kernel void mean_stddev_normalization( 48 IMAGE_DECLARATION(input) 49#ifndef IN_PLACE 50 , 51 IMAGE_DECLARATION(output) 52#endif /* IN_PLACE */ 53) 54{ 55 // Get pixels pointer 56 Image in = CONVERT_TO_IMAGE_STRUCT(input); 57#ifdef IN_PLACE 58 Image out = in; 59#else /* IN_PLACE */ 60 Image out = CONVERT_TO_IMAGE_STRUCT(output); 61#endif /* IN_PLACE */ 62 63 VEC_DATA_TYPE(DATA_TYPE, VEC_SIZE) 64 sum = 0.f; 65#ifdef MEANSTDNORM_HALF 66 VEC_DATA_TYPE(float, VEC_SIZE) 67#else /* MEANSTDNORM_HALF */ 68 VEC_DATA_TYPE(DATA_TYPE, VEC_SIZE) 69#endif /* MEANSTDNORM_HALF */ 70 sum_sq = 0.f; 71 // Calculate partial sum 72 int i = 0; 73 for(; i <= (WIDTH - VEC_SIZE); i += VEC_SIZE) 74 { 75 // Load data 76 VEC_DATA_TYPE(DATA_TYPE, VEC_SIZE) 77 data = VLOAD(VEC_SIZE)(0, (__global DATA_TYPE *)offset(&in, i, 0)); 78 79 sum += data; 80#ifdef MEANSTDNORM_HALF 81 VEC_DATA_TYPE(float, VEC_SIZE) 82 dsq = CONVERT(data * data, VEC_DATA_TYPE(float, VEC_SIZE)); 83 sum_sq += dsq; 84#else /* MEANSTDNORM_HALF */ 85 sum_sq += data * data; 86#endif /* MEANSTDNORM_HALF */ 87 } 88 // Perform reduction 89 sum = SUM_REDUCE(sum, VEC_SIZE); 90 sum_sq = SUM_REDUCE(sum_sq, VEC_SIZE); 91 92#if VEC_SIZE > 1 93#define sum sum.s0 94#define sum_sq sum_sq.s0 95#endif // VEC_SIZE > 1 96 97 // Left-overs loop 98 for(; i < WIDTH; ++i) 99 { 100 DATA_TYPE data = *((__global DATA_TYPE *)offset(&in, i, 0)); 101 102 sum += data; 103 sum_sq += data * data; 104 } 105 106 DATA_TYPE mean = sum / WIDTH; 107 DATA_TYPE var = (sum_sq / WIDTH) - (mean * mean); 108 DATA_TYPE stddev_inv = 1.f / sqrt(var + EPSILON); 109 110 i = 0; 111 for(; i <= (WIDTH - VEC_SIZE); i += VEC_SIZE) 112 { 113 VEC_DATA_TYPE(DATA_TYPE, VEC_SIZE) 114 data = VLOAD(VEC_SIZE)(0, (__global DATA_TYPE *)offset(&in, i, 0)); 115 116 VEC_DATA_TYPE(DATA_TYPE, VEC_SIZE) 117 res = (data - mean) * stddev_inv; 118 VSTORE(VEC_SIZE) 119 (res, 0, (__global DATA_TYPE *)offset(&out, i, 0)); 120 } 121 for(; i < WIDTH; ++i) 122 { 123 DATA_TYPE data = *((__global DATA_TYPE *)offset(&in, i, 0)); 124 125 *((__global DATA_TYPE *)offset(&out, i, 0)) = (data - mean) * stddev_inv; 126 } 127} 128#endif /* defined(VEC_SIZE) && defined(DATA_TYPE) && defined(EPSILON) && defined(WIDTH) */ 129