Anthony Barbier | 6ff3b19 | 2017-09-04 18:44:23 +0100 | [diff] [blame] | 1 | /* |
| 2 | * Copyright (c) 2016, 2017 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 "arm_compute/core/NEON/kernels/NEMeanStdDevKernel.h" |
| 25 | |
| 26 | #include "arm_compute/core/Error.h" |
| 27 | #include "arm_compute/core/Helpers.h" |
| 28 | #include "arm_compute/core/IAccessWindow.h" |
| 29 | #include "arm_compute/core/TensorInfo.h" |
| 30 | #include "arm_compute/core/Types.h" |
| 31 | #include "arm_compute/core/Validate.h" |
| 32 | |
| 33 | #include <arm_neon.h> |
| 34 | #include <cmath> |
| 35 | #include <tuple> |
| 36 | #include <utility> |
| 37 | |
| 38 | using namespace arm_compute; |
| 39 | |
| 40 | namespace arm_compute |
| 41 | { |
| 42 | class Coordinates; |
| 43 | } // namespace arm_compute |
| 44 | |
| 45 | namespace |
| 46 | { |
| 47 | template <bool calc_sum_squared> |
| 48 | std::pair<uint64x1_t, uint64x1_t> accumulate(const Window &window, Iterator &iterator) |
| 49 | { |
| 50 | uint64x1_t sum = vdup_n_u64(0); |
| 51 | uint64x1_t sum_squared = vdup_n_u64(0); |
| 52 | |
| 53 | // Calculate sum |
| 54 | execute_window_loop(window, [&](const Coordinates & id) |
| 55 | { |
| 56 | const uint8x16_t in_data = vld1q_u8(iterator.ptr()); |
| 57 | |
| 58 | // Sum of the low and high elements of data |
| 59 | const uint16x8_t tmp0 = vaddl_u8(vget_low_u8(in_data), vget_high_u8(in_data)); |
| 60 | const uint32x4_t tmp1 = vaddl_u16(vget_low_u16(tmp0), vget_high_u16(tmp0)); |
| 61 | const uint32x2_t tmp2 = vadd_u32(vget_low_u32(tmp1), vget_high_u32(tmp1)); |
| 62 | |
| 63 | // Update sum |
| 64 | sum = vpadal_u32(sum, tmp2); |
| 65 | |
| 66 | if(calc_sum_squared) |
| 67 | { |
| 68 | const uint16x8_t square_data_low = vmull_u8(vget_low_u8(in_data), vget_low_u8(in_data)); |
| 69 | const uint16x8_t square_data_high = vmull_u8(vget_high_u8(in_data), vget_high_u8(in_data)); |
| 70 | |
| 71 | // Sum of the low and high elements of data |
| 72 | const uint32x4_t tmp0_low = vaddl_u16(vget_low_u16(square_data_low), vget_high_u16(square_data_low)); |
| 73 | const uint32x4_t tmp0_high = vaddl_u16(vget_low_u16(square_data_high), vget_high_u16(square_data_high)); |
| 74 | const uint32x4_t tmp1 = vaddq_u32(tmp0_low, tmp0_high); |
| 75 | const uint32x2_t tmp2 = vadd_u32(vget_low_u32(tmp1), vget_high_u32(tmp1)); |
| 76 | |
| 77 | // Update sum |
| 78 | sum_squared = vpadal_u32(sum_squared, tmp2); |
| 79 | } |
| 80 | }, |
| 81 | iterator); |
| 82 | |
| 83 | return std::make_pair(sum, sum_squared); |
| 84 | } |
| 85 | } // namespace |
| 86 | |
| 87 | NEMeanStdDevKernel::NEMeanStdDevKernel() |
Giorgio Arena | a261181 | 2017-07-21 10:08:48 +0100 | [diff] [blame] | 88 | : _input(nullptr), _mean(nullptr), _stddev(nullptr), _global_sum(nullptr), _global_sum_squared(nullptr), _mtx(), _border_size(0) |
Anthony Barbier | 6ff3b19 | 2017-09-04 18:44:23 +0100 | [diff] [blame] | 89 | { |
| 90 | } |
| 91 | |
Giorgio Arena | a261181 | 2017-07-21 10:08:48 +0100 | [diff] [blame] | 92 | BorderSize NEMeanStdDevKernel::border_size() const |
| 93 | { |
| 94 | return _border_size; |
| 95 | } |
| 96 | |
Anthony Barbier | 6ff3b19 | 2017-09-04 18:44:23 +0100 | [diff] [blame] | 97 | void NEMeanStdDevKernel::configure(const IImage *input, float *mean, uint64_t *global_sum, float *stddev, uint64_t *global_sum_squared) |
| 98 | { |
| 99 | ARM_COMPUTE_ERROR_ON_TENSOR_NOT_2D(input); |
| 100 | ARM_COMPUTE_ERROR_ON(nullptr == mean); |
| 101 | ARM_COMPUTE_ERROR_ON(nullptr == global_sum); |
| 102 | ARM_COMPUTE_ERROR_ON(stddev && nullptr == global_sum_squared); |
| 103 | ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8); |
| 104 | |
| 105 | _input = input; |
| 106 | _mean = mean; |
| 107 | _stddev = stddev; |
| 108 | _global_sum = global_sum; |
| 109 | _global_sum_squared = global_sum_squared; |
| 110 | |
| 111 | constexpr unsigned int num_elems_processed_per_iteration = 16; |
| 112 | |
Giorgio Arena | fc2817d | 2017-06-27 17:26:37 +0100 | [diff] [blame] | 113 | _border_size = BorderSize(ceil_to_multiple(input->info()->dimension(0), num_elems_processed_per_iteration) - input->info()->dimension(0)); |
Giorgio Arena | a261181 | 2017-07-21 10:08:48 +0100 | [diff] [blame] | 114 | |
Anthony Barbier | 6ff3b19 | 2017-09-04 18:44:23 +0100 | [diff] [blame] | 115 | // Configure kernel window |
| 116 | Window win = calculate_max_window(*input->info(), Steps(num_elems_processed_per_iteration)); |
| 117 | |
| 118 | update_window_and_padding(win, AccessWindowHorizontal(input->info(), 0, num_elems_processed_per_iteration)); |
| 119 | |
| 120 | INEKernel::configure(win); |
| 121 | } |
| 122 | |
Moritz Pflanzer | c186b57 | 2017-09-07 09:48:04 +0100 | [diff] [blame] | 123 | void NEMeanStdDevKernel::run(const Window &window, const ThreadInfo &info) |
Anthony Barbier | 6ff3b19 | 2017-09-04 18:44:23 +0100 | [diff] [blame] | 124 | { |
Moritz Pflanzer | c186b57 | 2017-09-07 09:48:04 +0100 | [diff] [blame] | 125 | ARM_COMPUTE_UNUSED(info); |
Anthony Barbier | 6ff3b19 | 2017-09-04 18:44:23 +0100 | [diff] [blame] | 126 | ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this); |
| 127 | ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(INEKernel::window(), window); |
| 128 | Iterator input(_input, window); |
| 129 | |
| 130 | uint64x1_t local_sum = vdup_n_u64(0); |
| 131 | uint64x1_t local_sum_squared = vdup_n_u64(0); |
| 132 | |
| 133 | if(_stddev != nullptr) |
| 134 | { |
| 135 | std::tie(local_sum, local_sum_squared) = accumulate<true>(window, input); |
| 136 | } |
| 137 | else |
| 138 | { |
| 139 | std::tie(local_sum, local_sum_squared) = accumulate<false>(window, input); |
| 140 | } |
| 141 | |
| 142 | const float num_pixels = _input->info()->dimension(0) * _input->info()->dimension(1); |
| 143 | |
| 144 | // Merge sum and calculate mean and stddev |
Michalis Spyrou | 07781ac | 2017-08-31 15:11:41 +0100 | [diff] [blame] | 145 | std::unique_lock<arm_compute::Mutex> lock(_mtx); |
Anthony Barbier | 6ff3b19 | 2017-09-04 18:44:23 +0100 | [diff] [blame] | 146 | |
| 147 | *_global_sum += vget_lane_u64(local_sum, 0); |
| 148 | |
| 149 | const float mean = *_global_sum / num_pixels; |
| 150 | *_mean = mean; |
| 151 | |
| 152 | if(_stddev != nullptr) |
| 153 | { |
| 154 | const uint64_t tmp_sum_squared = vget_lane_u64(local_sum_squared, 0); |
| 155 | *_global_sum_squared += tmp_sum_squared; |
| 156 | *_stddev = std::sqrt((*_global_sum_squared / num_pixels) - (mean * mean)); |
| 157 | } |
| 158 | |
| 159 | lock.unlock(); |
| 160 | } |