Gian Marco Iodice | d2fab73 | 2018-03-02 11:18:12 +0000 | [diff] [blame] | 1 | /* |
Michele Di Giorgio | d9eaf61 | 2020-07-08 11:12:57 +0100 | [diff] [blame] | 2 | * Copyright (c) 2018-2020 Arm Limited. |
Gian Marco Iodice | d2fab73 | 2018-03-02 11:18:12 +0000 | [diff] [blame] | 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/runtime/CL/functions/CLWinogradConvolutionLayer.h" |
| 25 | |
| 26 | #include "arm_compute/core/CL/ICLTensor.h" |
| 27 | #include "arm_compute/core/Utils.h" |
| 28 | #include "arm_compute/core/Validate.h" |
| 29 | #include "arm_compute/core/utils/misc/ShapeCalculator.h" |
| 30 | #include "arm_compute/runtime/CL/CLScheduler.h" |
| 31 | |
| 32 | using namespace arm_compute; |
| 33 | |
Gian Marco Iodice | 2213d4b | 2018-04-27 10:39:06 +0100 | [diff] [blame] | 34 | namespace |
| 35 | { |
Gian Marco Iodice | d30714a | 2018-08-15 16:53:27 +0100 | [diff] [blame] | 36 | Size2D winograd_output_tile(const Size2D &input_dims, const Size2D &kernel_dims, DataLayout data_layout) |
Gian Marco Iodice | 2213d4b | 2018-04-27 10:39:06 +0100 | [diff] [blame] | 37 | { |
| 38 | Size2D output_tile = Size2D{}; |
| 39 | |
Gian Marco Iodice | f1c2bf0 | 2018-06-13 14:05:54 +0100 | [diff] [blame] | 40 | const unsigned int kernel_max_dim = std::max(kernel_dims.width, kernel_dims.height); |
| 41 | |
| 42 | // Check if the input spatial dimensions are smaller than 4 |
Gian Marco Iodice | d30714a | 2018-08-15 16:53:27 +0100 | [diff] [blame] | 43 | const bool is_input_lt4_nchw = (input_dims.width <= 4 && input_dims.height <= 4) && (data_layout == DataLayout::NCHW); |
Gian Marco Iodice | f1c2bf0 | 2018-06-13 14:05:54 +0100 | [diff] [blame] | 44 | |
| 45 | if(kernel_max_dim == 3U) |
Gian Marco Iodice | 2213d4b | 2018-04-27 10:39:06 +0100 | [diff] [blame] | 46 | { |
Gian Marco Iodice | f1c2bf0 | 2018-06-13 14:05:54 +0100 | [diff] [blame] | 47 | if(kernel_dims == Size2D(3U, 3U)) |
| 48 | { |
Gian Marco Iodice | d30714a | 2018-08-15 16:53:27 +0100 | [diff] [blame] | 49 | output_tile = is_input_lt4_nchw ? Size2D(2U, 2U) : Size2D(4U, 4U); |
Gian Marco Iodice | f1c2bf0 | 2018-06-13 14:05:54 +0100 | [diff] [blame] | 50 | } |
| 51 | else if(kernel_dims == Size2D(3U, 1U)) |
| 52 | { |
Gian Marco Iodice | d30714a | 2018-08-15 16:53:27 +0100 | [diff] [blame] | 53 | output_tile = is_input_lt4_nchw ? Size2D(2U, 1U) : Size2D(4U, 1U); |
Gian Marco Iodice | f1c2bf0 | 2018-06-13 14:05:54 +0100 | [diff] [blame] | 54 | } |
| 55 | else |
| 56 | { |
Gian Marco Iodice | d30714a | 2018-08-15 16:53:27 +0100 | [diff] [blame] | 57 | output_tile = is_input_lt4_nchw ? Size2D(1U, 2U) : Size2D(1U, 4U); |
Gian Marco Iodice | f1c2bf0 | 2018-06-13 14:05:54 +0100 | [diff] [blame] | 58 | } |
Gian Marco Iodice | 2213d4b | 2018-04-27 10:39:06 +0100 | [diff] [blame] | 59 | } |
Gian Marco Iodice | f1c2bf0 | 2018-06-13 14:05:54 +0100 | [diff] [blame] | 60 | else if(kernel_max_dim == 5U) |
Gian Marco Iodice | 2213d4b | 2018-04-27 10:39:06 +0100 | [diff] [blame] | 61 | { |
Gian Marco Iodice | 876be2a | 2018-07-03 12:22:09 +0100 | [diff] [blame] | 62 | output_tile = Size2D(kernel_dims.width == 1 ? 1U : 4U, |
| 63 | kernel_dims.height == 1 ? 1U : 4U); |
Gian Marco Iodice | 2213d4b | 2018-04-27 10:39:06 +0100 | [diff] [blame] | 64 | } |
giuros01 | 3bfacb2 | 2019-04-01 12:07:02 +0100 | [diff] [blame] | 65 | else if(kernel_max_dim == 7U) |
| 66 | { |
Gian Marco Iodice | 4fbcac6 | 2019-04-04 10:14:58 +0100 | [diff] [blame] | 67 | output_tile = Size2D(kernel_dims.width == 1 ? 1U : 2U, |
| 68 | kernel_dims.height == 1 ? 1U : 2U); |
giuros01 | 3bfacb2 | 2019-04-01 12:07:02 +0100 | [diff] [blame] | 69 | } |
Gian Marco Iodice | 2213d4b | 2018-04-27 10:39:06 +0100 | [diff] [blame] | 70 | |
| 71 | return output_tile; |
| 72 | } |
| 73 | |
| 74 | bool check_support_fast_math(const Size2D &output_tile, const Size2D &kernel_size) |
| 75 | { |
| 76 | // Check if we want to configure a Winograd configuration which requires fast math |
| 77 | using WinogradConfiguration = std::pair<std::pair<int, int>, std::pair<int, int>>; |
| 78 | |
| 79 | std::vector<WinogradConfiguration> fast_math_winograd = |
| 80 | { |
giuros01 | 3bfacb2 | 2019-04-01 12:07:02 +0100 | [diff] [blame] | 81 | WinogradConfiguration(std::pair<int, int>(4, 4), std::pair<int, int>(5, 5)), |
| 82 | WinogradConfiguration(std::pair<int, int>(2, 2), std::pair<int, int>(7, 7)) |
Gian Marco Iodice | 2213d4b | 2018-04-27 10:39:06 +0100 | [diff] [blame] | 83 | }; |
| 84 | |
| 85 | auto p = std::make_pair(std::pair<int, int>(output_tile.width, output_tile.height), |
| 86 | std::pair<int, int>(kernel_size.width, kernel_size.height)); |
| 87 | |
| 88 | return std::find(fast_math_winograd.begin(), fast_math_winograd.end(), p) != fast_math_winograd.end(); |
| 89 | } |
| 90 | } // namespace |
| 91 | |
Gian Marco Iodice | d2fab73 | 2018-03-02 11:18:12 +0000 | [diff] [blame] | 92 | CLWinogradConvolutionLayer::CLWinogradConvolutionLayer(std::shared_ptr<IMemoryManager> memory_manager) |
Manuel Bottini | 0d0028c | 2018-10-02 16:41:52 +0100 | [diff] [blame] | 93 | : _memory_group(memory_manager), _batched_mm(memory_manager), _input_transform(), _filter_transform(), _output_transform(), _input0(), _input1(), _batched_mm_output(), _original_weights(nullptr), |
| 94 | _is_prepared(false) |
Gian Marco Iodice | d2fab73 | 2018-03-02 11:18:12 +0000 | [diff] [blame] | 95 | { |
| 96 | } |
| 97 | |
Gian Marco Iodice | 2213d4b | 2018-04-27 10:39:06 +0100 | [diff] [blame] | 98 | void CLWinogradConvolutionLayer::configure(ICLTensor *input, const ICLTensor *weights, const ICLTensor *biases, ICLTensor *output, const PadStrideInfo &conv_info, const ActivationLayerInfo &act_info, |
| 99 | bool enable_fast_math) |
Gian Marco Iodice | d2fab73 | 2018-03-02 11:18:12 +0000 | [diff] [blame] | 100 | { |
Manuel Bottini | 2b84be5 | 2020-04-08 10:15:51 +0100 | [diff] [blame] | 101 | configure(CLKernelLibrary::get().get_compile_context(), input, weights, biases, output, conv_info, act_info, enable_fast_math); |
| 102 | } |
| 103 | |
| 104 | void CLWinogradConvolutionLayer::configure(const CLCompileContext &compile_context, ICLTensor *input, const ICLTensor *weights, const ICLTensor *biases, ICLTensor *output, |
Sang-Hoon Park | 68dd25f | 2020-10-19 16:00:11 +0100 | [diff] [blame] | 105 | const PadStrideInfo &conv_info, |
Manuel Bottini | 2b84be5 | 2020-04-08 10:15:51 +0100 | [diff] [blame] | 106 | const ActivationLayerInfo &act_info, bool enable_fast_math) |
| 107 | { |
Gian Marco Iodice | 247f52c | 2018-03-22 11:24:56 +0000 | [diff] [blame] | 108 | // Get indices for the width and height |
Gian Marco Iodice | d2fab73 | 2018-03-02 11:18:12 +0000 | [diff] [blame] | 109 | const size_t idx_width = get_data_layout_dimension_index(input->info()->data_layout(), DataLayoutDimension::WIDTH); |
| 110 | const size_t idx_height = get_data_layout_dimension_index(input->info()->data_layout(), DataLayoutDimension::HEIGHT); |
| 111 | |
Gian Marco Iodice | 2213d4b | 2018-04-27 10:39:06 +0100 | [diff] [blame] | 112 | // Input shape, kernel size and output tile |
| 113 | const Size2D input_dims = Size2D(input->info()->tensor_shape()[idx_width], input->info()->tensor_shape()[idx_height]); |
| 114 | const Size2D kernel_size = Size2D(weights->info()->tensor_shape()[idx_width], weights->info()->tensor_shape()[idx_height]); |
Gian Marco Iodice | d30714a | 2018-08-15 16:53:27 +0100 | [diff] [blame] | 115 | const Size2D output_tile = winograd_output_tile(input_dims, kernel_size, input->info()->data_layout()); |
Gian Marco Iodice | 247f52c | 2018-03-22 11:24:56 +0000 | [diff] [blame] | 116 | |
Gian Marco Iodice | 2213d4b | 2018-04-27 10:39:06 +0100 | [diff] [blame] | 117 | // Check if the Winograd configuration requires fast math |
| 118 | if(!enable_fast_math) |
| 119 | { |
Vidhya Sudhan Loganathan | a25d16c | 2018-11-16 11:33:12 +0000 | [diff] [blame] | 120 | ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::F32); //disable winograd for fp16 if fast math is false. |
Gian Marco Iodice | 2213d4b | 2018-04-27 10:39:06 +0100 | [diff] [blame] | 121 | ARM_COMPUTE_ERROR_ON_MSG(check_support_fast_math(output_tile, kernel_size), "This Winograd configuration requires enable_fast_math=true"); |
| 122 | } |
Gian Marco Iodice | e52a300 | 2018-04-11 15:59:10 +0100 | [diff] [blame] | 123 | const WinogradInfo winograd_info = WinogradInfo(output_tile, |
Gian Marco Iodice | 2213d4b | 2018-04-27 10:39:06 +0100 | [diff] [blame] | 124 | kernel_size, |
| 125 | input_dims, |
Gian Marco Iodice | 247f52c | 2018-03-22 11:24:56 +0000 | [diff] [blame] | 126 | conv_info, |
| 127 | input->info()->data_layout()); |
Gian Marco Iodice | d2fab73 | 2018-03-02 11:18:12 +0000 | [diff] [blame] | 128 | |
Georgios Pinitas | e043767 | 2018-05-02 14:07:55 +0100 | [diff] [blame] | 129 | _is_prepared = false; |
| 130 | _original_weights = weights; |
| 131 | |
Gian Marco Iodice | d2fab73 | 2018-03-02 11:18:12 +0000 | [diff] [blame] | 132 | // Manage intermediate tensors |
| 133 | _memory_group.manage(&_input0); |
| 134 | _memory_group.manage(&_batched_mm_output); |
| 135 | |
| 136 | // Do not manage _input1 as it contains the weights |
| 137 | |
| 138 | // Configure input transform |
Manuel Bottini | 2b84be5 | 2020-04-08 10:15:51 +0100 | [diff] [blame] | 139 | _input_transform.configure(compile_context, input, &_input0, winograd_info); |
Gian Marco Iodice | d2fab73 | 2018-03-02 11:18:12 +0000 | [diff] [blame] | 140 | |
| 141 | // Configure filter transform |
Manuel Bottini | 2b84be5 | 2020-04-08 10:15:51 +0100 | [diff] [blame] | 142 | _filter_transform.configure(compile_context, weights, &_input1, winograd_info); |
Gian Marco Iodice | d2fab73 | 2018-03-02 11:18:12 +0000 | [diff] [blame] | 143 | |
| 144 | // Configure batched matrix multiply |
Manuel Bottini | 2b84be5 | 2020-04-08 10:15:51 +0100 | [diff] [blame] | 145 | _batched_mm.configure(compile_context, &_input0, &_input1, nullptr, &_batched_mm_output, 1.0f, 0.0f, GEMMInfo(false, false, true /* Reshape weights only for the first run*/, 0, false, false, |
| 146 | GEMMLowpOutputStageInfo(), |
| 147 | (input->info()->data_type() == DataType::F16))); |
Gian Marco Iodice | d2fab73 | 2018-03-02 11:18:12 +0000 | [diff] [blame] | 148 | |
| 149 | // Configure output transform |
Manuel Bottini | 2b84be5 | 2020-04-08 10:15:51 +0100 | [diff] [blame] | 150 | _output_transform.configure(compile_context, &_batched_mm_output, biases, output, winograd_info, act_info); |
Isabella Gottardi | 3f217ec | 2018-02-12 14:59:19 +0000 | [diff] [blame] | 151 | |
Gian Marco Iodice | d2fab73 | 2018-03-02 11:18:12 +0000 | [diff] [blame] | 152 | // Allocate temporary tensors |
| 153 | _input0.allocator()->allocate(); |
Gian Marco Iodice | d2fab73 | 2018-03-02 11:18:12 +0000 | [diff] [blame] | 154 | _batched_mm_output.allocator()->allocate(); |
| 155 | } |
| 156 | |
Isabella Gottardi | 3f217ec | 2018-02-12 14:59:19 +0000 | [diff] [blame] | 157 | Status CLWinogradConvolutionLayer::validate(const ITensorInfo *input, const ITensorInfo *weights, const ITensorInfo *biases, const ITensorInfo *output, const PadStrideInfo &conv_info, |
Gian Marco Iodice | 2213d4b | 2018-04-27 10:39:06 +0100 | [diff] [blame] | 158 | const ActivationLayerInfo &act_info, bool enable_fast_math) |
Gian Marco Iodice | d2fab73 | 2018-03-02 11:18:12 +0000 | [diff] [blame] | 159 | { |
Gian Marco Iodice | d2fab73 | 2018-03-02 11:18:12 +0000 | [diff] [blame] | 160 | // Get indeces for the width and height |
| 161 | const size_t idx_width = get_data_layout_dimension_index(input->data_layout(), DataLayoutDimension::WIDTH); |
| 162 | const size_t idx_height = get_data_layout_dimension_index(input->data_layout(), DataLayoutDimension::HEIGHT); |
| 163 | |
Gian Marco Iodice | 2213d4b | 2018-04-27 10:39:06 +0100 | [diff] [blame] | 164 | // Input shape, kernel size and output tile |
| 165 | const Size2D input_dims = Size2D(input->tensor_shape()[idx_width], input->tensor_shape()[idx_height]); |
| 166 | const Size2D kernel_size = Size2D(weights->tensor_shape()[idx_width], weights->tensor_shape()[idx_height]); |
Gian Marco Iodice | d30714a | 2018-08-15 16:53:27 +0100 | [diff] [blame] | 167 | const Size2D output_tile = winograd_output_tile(input_dims, kernel_size, input->data_layout()); |
Gian Marco Iodice | 247f52c | 2018-03-22 11:24:56 +0000 | [diff] [blame] | 168 | |
Pablo Tello | 7786f25 | 2019-08-05 12:22:37 +0100 | [diff] [blame] | 169 | ARM_COMPUTE_RETURN_ERROR_ON_MSG(((conv_info.pad_left() > (kernel_size.x() / 2u)) || (conv_info.pad_right() > (kernel_size.x() / 2u))), "Winograd only supports padding up to half kernel size"); |
| 170 | ARM_COMPUTE_RETURN_ERROR_ON_MSG(((conv_info.pad_top() > (kernel_size.y() / 2u)) || (conv_info.pad_bottom() > (kernel_size.y() / 2u))), "Winograd only supports padding up to half kernel size"); |
| 171 | |
Gian Marco Iodice | 2213d4b | 2018-04-27 10:39:06 +0100 | [diff] [blame] | 172 | // Check if the Winograd configuration requires fast math |
| 173 | if(!enable_fast_math) |
| 174 | { |
Vidhya Sudhan Loganathan | a25d16c | 2018-11-16 11:33:12 +0000 | [diff] [blame] | 175 | ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::F32); //disable winograd for fp16 if fast math is false. |
Gian Marco Iodice | 2213d4b | 2018-04-27 10:39:06 +0100 | [diff] [blame] | 176 | ARM_COMPUTE_RETURN_ERROR_ON_MSG(check_support_fast_math(output_tile, kernel_size), "This Winograd configuration requires enable_fast_math=true"); |
| 177 | } |
Gian Marco Iodice | e52a300 | 2018-04-11 15:59:10 +0100 | [diff] [blame] | 178 | |
| 179 | const WinogradInfo winograd_info = WinogradInfo(output_tile, |
Gian Marco Iodice | 2213d4b | 2018-04-27 10:39:06 +0100 | [diff] [blame] | 180 | kernel_size, |
| 181 | input_dims, |
Gian Marco Iodice | 247f52c | 2018-03-22 11:24:56 +0000 | [diff] [blame] | 182 | conv_info, |
| 183 | input->data_layout()); |
Gian Marco Iodice | d2fab73 | 2018-03-02 11:18:12 +0000 | [diff] [blame] | 184 | |
| 185 | // Validate input transform |
Gian Marco Iodice | 247f52c | 2018-03-22 11:24:56 +0000 | [diff] [blame] | 186 | const TensorShape input0_shape = misc::shape_calculator::compute_winograd_input_transform_shape(*input, winograd_info); |
Gian Marco Iodice | d2fab73 | 2018-03-02 11:18:12 +0000 | [diff] [blame] | 187 | const TensorInfo input0 = input->clone()->set_tensor_shape(input0_shape); |
Gian Marco Iodice | 247f52c | 2018-03-22 11:24:56 +0000 | [diff] [blame] | 188 | ARM_COMPUTE_RETURN_ON_ERROR(CLWinogradInputTransform::validate(input, &input0, winograd_info)); |
Gian Marco Iodice | d2fab73 | 2018-03-02 11:18:12 +0000 | [diff] [blame] | 189 | |
| 190 | // Validate filter transform |
Gian Marco Iodice | 247f52c | 2018-03-22 11:24:56 +0000 | [diff] [blame] | 191 | const TensorShape input1_shape = misc::shape_calculator::compute_winograd_filter_transform_shape(*weights, winograd_info); |
Gian Marco Iodice | d2fab73 | 2018-03-02 11:18:12 +0000 | [diff] [blame] | 192 | const TensorInfo input1 = weights->clone()->set_tensor_shape(input1_shape); |
Gian Marco Iodice | 247f52c | 2018-03-22 11:24:56 +0000 | [diff] [blame] | 193 | ARM_COMPUTE_RETURN_ON_ERROR(CLWinogradFilterTransformKernel::validate(weights, &input1, winograd_info)); |
Gian Marco Iodice | d2fab73 | 2018-03-02 11:18:12 +0000 | [diff] [blame] | 194 | |
Isabella Gottardi | 3f217ec | 2018-02-12 14:59:19 +0000 | [diff] [blame] | 195 | // Validate batched matrix multiply |
Gian Marco Iodice | d2fab73 | 2018-03-02 11:18:12 +0000 | [diff] [blame] | 196 | TensorShape batched_mm_output_shape = input0.tensor_shape(); |
| 197 | batched_mm_output_shape[0] = input1.tensor_shape()[0]; |
| 198 | const TensorInfo batched_mm_output = input0.clone()->set_tensor_shape(batched_mm_output_shape); |
Vidhya Sudhan Loganathan | a25d16c | 2018-11-16 11:33:12 +0000 | [diff] [blame] | 199 | ARM_COMPUTE_RETURN_ON_ERROR(CLGEMM::validate(&input0, &input1, nullptr, &batched_mm_output, 1.0f, 0.0f, GEMMInfo(false, false, true /* Reshape weights only for the first run*/, 0, false, false, |
| 200 | GEMMLowpOutputStageInfo(), (input->data_type() == DataType::F16)))); |
Gian Marco Iodice | d2fab73 | 2018-03-02 11:18:12 +0000 | [diff] [blame] | 201 | |
Gian Marco Iodice | 247f52c | 2018-03-22 11:24:56 +0000 | [diff] [blame] | 202 | // Configure output transform |
Gian Marco Iodice | 607339f | 2019-08-01 15:53:23 +0100 | [diff] [blame] | 203 | ARM_COMPUTE_RETURN_ON_ERROR(CLWinogradOutputTransformKernel::validate(&batched_mm_output, biases, output, winograd_info, act_info)); |
Isabella Gottardi | 3f217ec | 2018-02-12 14:59:19 +0000 | [diff] [blame] | 204 | |
Gian Marco Iodice | d2fab73 | 2018-03-02 11:18:12 +0000 | [diff] [blame] | 205 | return Status{}; |
| 206 | } |
| 207 | |
| 208 | void CLWinogradConvolutionLayer::run() |
| 209 | { |
Georgios Pinitas | e043767 | 2018-05-02 14:07:55 +0100 | [diff] [blame] | 210 | prepare(); |
Gian Marco Iodice | d2fab73 | 2018-03-02 11:18:12 +0000 | [diff] [blame] | 211 | |
Georgios Pinitas | da953f2 | 2019-04-02 17:27:03 +0100 | [diff] [blame] | 212 | MemoryGroupResourceScope scope_mg(_memory_group); |
Gian Marco Iodice | d2fab73 | 2018-03-02 11:18:12 +0000 | [diff] [blame] | 213 | |
| 214 | // Run input transform |
| 215 | _input_transform.run(); |
| 216 | |
| 217 | // Run batched matrix multiplication |
| 218 | _batched_mm.run(); |
| 219 | |
| 220 | // Run output transform |
| 221 | CLScheduler::get().enqueue(_output_transform); |
Georgios Pinitas | e043767 | 2018-05-02 14:07:55 +0100 | [diff] [blame] | 222 | } |
Georgios Pinitas | 82b5148 | 2018-04-24 15:14:12 +0100 | [diff] [blame] | 223 | |
Georgios Pinitas | e043767 | 2018-05-02 14:07:55 +0100 | [diff] [blame] | 224 | void CLWinogradConvolutionLayer::prepare() |
| 225 | { |
| 226 | if(!_is_prepared) |
| 227 | { |
| 228 | // Run filter transform and mark original weights as unused |
| 229 | _input1.allocator()->allocate(); |
| 230 | CLScheduler::get().enqueue(_filter_transform, false); |
| 231 | _original_weights->mark_as_unused(); |
| 232 | |
| 233 | // Prepare GEMM and release reshaped weights if marked unused by CLGEMM |
| 234 | _batched_mm.prepare(); |
| 235 | if(!_input1.is_used()) |
| 236 | { |
| 237 | _input1.allocator()->free(); |
| 238 | } |
| 239 | |
| 240 | CLScheduler::get().queue().finish(); |
| 241 | _is_prepared = true; |
| 242 | } |
Gian Marco Iodice | d2fab73 | 2018-03-02 11:18:12 +0000 | [diff] [blame] | 243 | } |