XNNPACK Team | b455b12 | 2019-09-27 18:10:33 -0700 | [diff] [blame] | 1 | // Copyright 2019 Google LLC |
| 2 | // |
| 3 | // This source code is licensed under the BSD-style license found in the |
| 4 | // LICENSE file in the root directory of this source tree. |
| 5 | |
| 6 | #include <stddef.h> |
Marat Dukhan | f8c8046 | 2019-10-02 16:27:03 -0700 | [diff] [blame] | 7 | #include <stdint.h> |
XNNPACK Team | b455b12 | 2019-09-27 18:10:33 -0700 | [diff] [blame] | 8 | #include <string.h> |
| 9 | |
| 10 | #include <xnnpack/im2col.h> |
| 11 | |
| 12 | |
| 13 | void xnn_im2col_conv2d( |
| 14 | size_t output_height, |
| 15 | size_t output_width, |
| 16 | size_t kernel_height, |
| 17 | size_t kernel_width, |
| 18 | size_t subsampling_height, |
| 19 | size_t subsampling_width, |
| 20 | size_t dilation_height, |
| 21 | size_t dilation_width, |
| 22 | size_t input_width, |
| 23 | size_t input_padding_top, |
| 24 | size_t input_padding_left, |
| 25 | size_t group_input_channels_in_bytes, |
| 26 | size_t input_pixel_stride_in_bytes, |
| 27 | const void* input, |
| 28 | void* output) |
| 29 | { |
| 30 | for (size_t output_y = 0; output_y < output_height; output_y++) { |
| 31 | for (size_t output_x = 0; output_x < output_width; output_x++) { |
| 32 | for (size_t kernel_y = 0; kernel_y < kernel_height; kernel_y++) { |
| 33 | const size_t input_y = output_y * subsampling_height + kernel_y * dilation_height - input_padding_top; |
| 34 | if (input_y < output_height) { |
| 35 | for (size_t kernel_x = 0; kernel_x < kernel_width; kernel_x++) { |
| 36 | const size_t input_x = output_x * subsampling_width + kernel_x * dilation_width - input_padding_left; |
| 37 | if (input_x < output_width) { |
Marat Dukhan | e0df831 | 2019-10-22 18:16:56 -0700 | [diff] [blame^] | 38 | memcpy(output, |
XNNPACK Team | b455b12 | 2019-09-27 18:10:33 -0700 | [diff] [blame] | 39 | (const void*) ((uintptr_t) input + (input_y * input_width + input_x) * input_pixel_stride_in_bytes), |
| 40 | group_input_channels_in_bytes); |
| 41 | } else { |
| 42 | memset(output, 0, group_input_channels_in_bytes); |
| 43 | } |
| 44 | output = (void*) ((uintptr_t) output + group_input_channels_in_bytes); |
| 45 | } |
| 46 | } else { |
| 47 | memset(output, 0, kernel_width * group_input_channels_in_bytes); |
| 48 | output = (void*) ((uintptr_t) output + kernel_width * group_input_channels_in_bytes); |
| 49 | } |
| 50 | } |
| 51 | } |
| 52 | } |
| 53 | } |