Giorgio Arena | 93a690e | 2017-08-01 16:09:33 +0100 | [diff] [blame] | 1 | /* |
| 2 | * Copyright (c) 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 | |
| 25 | #include "helpers.h" |
| 26 | |
| 27 | #if CONV_STRIDE_X == 1 |
| 28 | #define convolution1x3 convolution1x3_stride_1 |
| 29 | #elif CONV_STRIDE_X == 2 |
| 30 | #define convolution1x3 convolution1x3_stride_2 |
| 31 | #elif CONV_STRIDE_X == 3 |
| 32 | #define convolution1x3 convolution1x3_stride_3 |
| 33 | #else /* CONV_STRIDE_X */ |
| 34 | #error "Stride not supported" |
| 35 | #endif /* CONV_STRIDE_X */ |
| 36 | |
| 37 | /** Compute a 1D horizontal convolution of size 3 and stride 1 for floating point type. |
| 38 | * |
| 39 | * @param[in] left_pixel Pointer to the left pixel. |
| 40 | * @param[in] left_coeff Weight of the left pixel |
| 41 | * @param[in] middle_coeff Weight of the middle pixel |
| 42 | * @param[in] right_coeff Weight of the right pixel |
| 43 | * |
| 44 | * @return a float2 containing 2 convoluted values. |
| 45 | */ |
| 46 | inline float2 convolution1x3_stride_1(__global const uchar *left_pixel, |
| 47 | const float left_coeff, |
| 48 | const float middle_coeff, |
| 49 | const float right_coeff) |
| 50 | { |
| 51 | float4 temp = vload4(0, (__global float *)left_pixel); |
| 52 | |
| 53 | float2 left = CONVERT(temp.s01, float2); |
| 54 | float2 middle = CONVERT(temp.s12, float2); |
| 55 | float2 right = CONVERT(temp.s23, float2); |
| 56 | |
| 57 | return left * (float2)left_coeff + middle * (float2)middle_coeff + right * (float2)right_coeff; |
| 58 | } |
| 59 | |
| 60 | /** Compute a 1D horizontal convolution of size 3 and stride 2 for floating point type. |
| 61 | * |
| 62 | * @param[in] left_pixel Pointer to the left pixel. |
| 63 | * @param[in] left_coeff Weight of the left pixel |
| 64 | * @param[in] middle_coeff Weight of the middle pixel |
| 65 | * @param[in] right_coeff Weight of the right pixel |
| 66 | * |
| 67 | * @return a float2 containing 2 convoluted values. |
| 68 | */ |
| 69 | inline float2 convolution1x3_stride_2(__global const uchar *left_pixel, |
| 70 | const float left_coeff, |
| 71 | const float middle_coeff, |
| 72 | const float right_coeff) |
| 73 | { |
| 74 | float4 temp0 = vload4(0, (__global float *)left_pixel); |
| 75 | float temp1 = *((__global float *)(left_pixel + 4 * sizeof(float))); |
| 76 | |
| 77 | float2 left = CONVERT(temp0.s02, float2); |
| 78 | float2 middle = CONVERT(temp0.s13, float2); |
| 79 | float2 right = CONVERT((float2)(temp0.s2, temp1), float2); |
| 80 | |
| 81 | return left * (float2)left_coeff + middle * (float2)middle_coeff + right * (float2)right_coeff; |
| 82 | } |
| 83 | |
| 84 | /** Compute a 1D horizontal convolution of size 3 and stride 3 for floating point type. |
| 85 | * |
| 86 | * @param[in] left_pixel Pointer to the left pixel. |
| 87 | * @param[in] left_coeff Weight of the left pixel |
| 88 | * @param[in] middle_coeff Weight of the middle pixel |
| 89 | * @param[in] right_coeff Weight of the right pixel |
| 90 | * |
| 91 | * @return a float2 containing 2 convoluted values. |
| 92 | */ |
| 93 | inline float2 convolution1x3_stride_3(__global const uchar *left_pixel, |
| 94 | const float left_coeff, |
| 95 | const float middle_coeff, |
| 96 | const float right_coeff) |
| 97 | { |
| 98 | float4 temp0 = vload4(0, (__global float *)left_pixel); |
| 99 | float2 temp1 = vload2(0, (__global float *)(left_pixel + 4 * sizeof(float))); |
| 100 | |
| 101 | float2 left = CONVERT(temp0.s03, float2); |
| 102 | float2 middle = CONVERT((float2)(temp0.s1, temp1.s0), float2); |
| 103 | float2 right = CONVERT((float2)(temp0.s2, temp1.s1), float2); |
| 104 | |
| 105 | return left * (float2)left_coeff + middle * (float2)middle_coeff + right * (float2)right_coeff; |
| 106 | } |
| 107 | |
| 108 | /** Apply a 3x3 convolution matrix to a single channel F32 input image and return the result. |
| 109 | * |
| 110 | * Convolution matrix layout: |
| 111 | * |
| 112 | * [ mat0, mat1, mat2 ]\n |
| 113 | * [ mat3, mat4, mat5 ]\n |
| 114 | * [ mat6, mat7, mat8 ]\n |
| 115 | * |
| 116 | * @param[in] src A pointer to source Image structure |
| 117 | * @param[in] mat0 Coefficient from the convolution matrix |
| 118 | * @param[in] mat1 Coefficient from the convolution matrix |
| 119 | * @param[in] mat2 Coefficient from the convolution matrix |
| 120 | * @param[in] mat3 Coefficient from the convolution matrix |
| 121 | * @param[in] mat4 Coefficient from the convolution matrix |
| 122 | * @param[in] mat5 Coefficient from the convolution matrix |
| 123 | * @param[in] mat6 Coefficient from the convolution matrix |
| 124 | * @param[in] mat0 Coefficient from the convolution matrix |
| 125 | * @param[in] mat7 Coefficient from the convolution matrix |
| 126 | * @param[in] mat8 Coefficient from the convolution matrix |
| 127 | * |
| 128 | * @return a float2 containing 2 convoluted values. |
| 129 | */ |
| 130 | inline float2 convolution3x3( |
| 131 | Image *src, |
| 132 | const float mat0, const float mat1, const float mat2, |
| 133 | const float mat3, const float mat4, const float mat5, |
| 134 | const float mat6, const float mat7, const float mat8) |
| 135 | { |
| 136 | float2 pixels; |
| 137 | |
| 138 | pixels = convolution1x3(offset(src, 0, 0), mat0, mat1, mat2); |
| 139 | pixels += convolution1x3(offset(src, 0, 1), mat3, mat4, mat5); |
| 140 | pixels += convolution1x3(offset(src, 0, 2), mat6, mat7, mat8); |
| 141 | |
| 142 | return pixels; |
| 143 | } |
| 144 | |
| 145 | /** This function computes the horizontal integral of the image. |
| 146 | * |
| 147 | * @param[in] src_ptr Pointer to the source image. Supported data types: U8 |
| 148 | * @param[in] src_stride_x Stride of the source image in X dimension (in bytes) |
| 149 | * @param[in] src_step_x src_stride_x * number of elements along X processed per workitem(in bytes) |
| 150 | * @param[in] src_stride_y Stride of the source image in Y dimension (in bytes) |
| 151 | * @param[in] src_step_y src_stride_y * number of elements along Y processed per workitem(in bytes) |
| 152 | * @param[in] src_offset_first_element_in_bytes The offset of the first element in the source image |
| 153 | * @param[in] src_stride_z Stride of the source tensor in Z dimension (in bytes) |
| 154 | * @param[in] src_step_z src_stride_z * number of elements along Y processed per workitem(in bytes) |
| 155 | * @param[in] dst_ptr Pointer to the destination tensor. Supported data types: F16/F32 |
| 156 | * @param[in] dst_stride_x Stride of the destination tensor in X dimension (in bytes) |
| 157 | * @param[in] dst_step_x dst_stride_x * number of elements along X processed per workitem(in bytes) |
| 158 | * @param[in] dst_stride_y Stride of the destination tensor in Y dimension (in bytes) |
| 159 | * @param[in] dst_step_y dst_stride_y * number of elements along Y processed per workitem(in bytes) |
| 160 | * @param[in] dst_stride_z Stride of the destination tensor in Z dimension (in bytes) |
| 161 | * @param[in] dst_step_z dst_stride_z * number of elements along Y processed per workitem(in bytes) |
| 162 | * @param[in] dst_offset_first_element_in_bytes The offset of the first element in the destination tensor |
| 163 | * @param[in] weights_ptr Pointer to the weights tensor. Supported data types: F16/F32 |
| 164 | * @param[in] weights_stride_x Stride of the weights tensor in X dimension (in bytes) |
| 165 | * @param[in] weights_step_x weights_stride_x * number of elements along X processed per workitem(in bytes) |
| 166 | * @param[in] weights_stride_y Stride of the weights tensor in Y dimension (in bytes) |
| 167 | * @param[in] weights_step_y weights_stride_y * number of elements along Y processed per workitem(in bytes) |
| 168 | * @param[in] weights_stride_z Stride of the weights tensor in Z dimension (in bytes) |
| 169 | * @param[in] weights_step_z weights_stride_z * number of elements along Y processed per workitem(in bytes) |
| 170 | * @param[in] weights_offset_first_element_in_bytes The offset of the first element in the weights tensor |
| 171 | */ |
| 172 | |
| 173 | __kernel void depthwise_convolution_3x3(TENSOR3D_DECLARATION(src), TENSOR3D_DECLARATION(dst), TENSOR3D_DECLARATION(weights)) |
| 174 | { |
| 175 | Image src = CONVERT_TENSOR3D_TO_IMAGE_STRUCT(src); |
| 176 | Image dst = CONVERT_TENSOR3D_TO_IMAGE_STRUCT(dst); |
| 177 | Tensor3D weights = CONVERT_TO_TENSOR3D_STRUCT(weights); |
| 178 | |
| 179 | uchar3 offset = (uchar3)(0, 1, 2) * (uchar3)weights_stride_y; |
| 180 | float3 weights_values0 = vload3(0, (__global float *)(weights.ptr + offset.s0)); |
| 181 | float3 weights_values1 = vload3(0, (__global float *)(weights.ptr + offset.s1)); |
| 182 | float3 weights_values2 = vload3(0, (__global float *)(weights.ptr + offset.s2)); |
| 183 | |
| 184 | float2 pixels = convolution3x3(&src, weights_values0.s0, weights_values0.s1, weights_values0.s2, |
| 185 | weights_values1.s0, weights_values1.s1, weights_values1.s2, |
| 186 | weights_values2.s0, weights_values2.s1, weights_values2.s2); |
| 187 | |
| 188 | vstore2(pixels, 0, (__global float *)dst.ptr); |
| 189 | } |