steniu01 | db00668 | 2017-08-09 16:26:22 +0100 | [diff] [blame] | 1 | /* |
Gian Marco Iodice | ff1fe3e | 2021-01-02 09:58:51 +0000 | [diff] [blame^] | 2 | * Copyright (c) 2016-2021 Arm Limited. |
steniu01 | db00668 | 2017-08-09 16:26:22 +0100 | [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 "helpers.h" |
| 25 | |
| 26 | #undef CONVERT_SAT |
| 27 | |
Gian Marco Iodice | 1246b63 | 2017-08-16 18:38:32 +0100 | [diff] [blame] | 28 | #if defined(DATA_TYPE) && defined(STRIDE_X) && defined(WEIGHTS_DEPTH) |
| 29 | |
steniu01 | db00668 | 2017-08-09 16:26:22 +0100 | [diff] [blame] | 30 | #if STRIDE_X == 1 |
| 31 | #define CONVOLUTION1x5(acc, src_row_ptr, weights_row_ptr) CONVOLUTION1x5_STRIDE1(acc, src_row_ptr, weights_row_ptr) |
| 32 | #elif STRIDE_X == 2 /* STRIDE_X == 1 */ |
| 33 | #define CONVOLUTION1x5(acc, src_row_ptr, weights_row_ptr) CONVOLUTION1x5_STRIDE2(acc, src_row_ptr, weights_row_ptr) |
| 34 | #else /* STRIDE_X not equals 1 or 2 */ |
| 35 | #error "STRIDE_X larger than 2 is not supported" |
| 36 | #endif /* STRIDE_X == 2 */ |
| 37 | |
| 38 | #define CONVOLUTION1x5_STRIDE1(acc, src_row_ptr, weights_row_ptr) \ |
| 39 | ({ \ |
| 40 | VEC_DATA_TYPE(DATA_TYPE, 4) \ |
| 41 | weights_values0 = vload4(0, weights_row_ptr); \ |
| 42 | DATA_TYPE weights_value1 = *(weights_row_ptr + 4); \ |
| 43 | VEC_DATA_TYPE(DATA_TYPE, 8) \ |
| 44 | src0 = vload8(0, src_row_ptr); \ |
| 45 | VEC_DATA_TYPE(DATA_TYPE, 4) \ |
| 46 | src1 = vload4(0, src_row_ptr + 8); \ |
| 47 | \ |
| 48 | acc += src0 * (VEC_DATA_TYPE(DATA_TYPE, 8))weights_values0.s0; \ |
| 49 | acc += (VEC_DATA_TYPE(DATA_TYPE, 8))(src0.s1234, src0.s567, src1.s0) * (VEC_DATA_TYPE(DATA_TYPE, 8))weights_values0.s1; \ |
| 50 | acc += (VEC_DATA_TYPE(DATA_TYPE, 8))(src0.s234, src0.s567, src1.s01) * (VEC_DATA_TYPE(DATA_TYPE, 8))weights_values0.s2; \ |
| 51 | acc += (VEC_DATA_TYPE(DATA_TYPE, 8))(src0.s345, src0.s67, src1.s012) * (VEC_DATA_TYPE(DATA_TYPE, 8))weights_values0.s3; \ |
| 52 | acc += (VEC_DATA_TYPE(DATA_TYPE, 8))(src0.s45, src0.s67, src1.s0123) * (VEC_DATA_TYPE(DATA_TYPE, 8))weights_value1; \ |
| 53 | }) |
| 54 | |
| 55 | #define CONVOLUTION1x5_STRIDE2(acc, src_row_ptr, weights_row_ptr) \ |
| 56 | ({ \ |
| 57 | VEC_DATA_TYPE(DATA_TYPE, 4) \ |
| 58 | weights_values0 = vload4(0, weights_row_ptr); \ |
| 59 | DATA_TYPE weights_value1 = *(weights_row_ptr + 4); \ |
| 60 | VEC_DATA_TYPE(DATA_TYPE, 16) \ |
| 61 | src0 = vload16(0, src_row_ptr); \ |
| 62 | VEC_DATA_TYPE(DATA_TYPE, 4) \ |
| 63 | src1 = vload4(0, src_row_ptr + 16); \ |
| 64 | acc += src0.even * (VEC_DATA_TYPE(DATA_TYPE, 8))weights_values0.s0; \ |
| 65 | acc += (VEC_DATA_TYPE(DATA_TYPE, 8))(src0.s1357, src0.s9BDF) * (VEC_DATA_TYPE(DATA_TYPE, 8))weights_values0.s1; \ |
| 66 | acc += (VEC_DATA_TYPE(DATA_TYPE, 8))(src0.s2468, src0.sACE, src1.s0) * (VEC_DATA_TYPE(DATA_TYPE, 8))weights_values0.s2; \ |
| 67 | \ |
| 68 | acc += (VEC_DATA_TYPE(DATA_TYPE, 8))(src0.s3579, src0.sBDF, src1.s1) * (VEC_DATA_TYPE(DATA_TYPE, 8))weights_values0.s3; \ |
| 69 | acc += (VEC_DATA_TYPE(DATA_TYPE, 8))(src0.s468a, src0.sCE, src1.s02) * (VEC_DATA_TYPE(DATA_TYPE, 8))weights_value1; \ |
| 70 | }) |
| 71 | |
| 72 | /** This kernel performs a direct convolution to convolve the low three dimensions. |
| 73 | * |
| 74 | * @note The data type must be passed at compile time using -DDATA_TYPE: e.g. -DDATA_TYPE=float |
| 75 | * @note The third dimensions of the weights tensors must be passed at compile time using -DWEIGHTS_DEPTH |
Gian Marco Iodice | 1246b63 | 2017-08-16 18:38:32 +0100 | [diff] [blame] | 76 | * @note If biases are used then -DHAS_BIAS has to be passed at compile time |
steniu01 | db00668 | 2017-08-09 16:26:22 +0100 | [diff] [blame] | 77 | * |
| 78 | * @param[in] src_ptr Pointer to the source tensor. Supported data types: F16/F32 |
| 79 | * @param[in] src_stride_x Stride of the source tensor in X dimension (in bytes) |
| 80 | * @param[in] src_step_x src_stride_x * number of elements along X processed per workitem(in bytes) |
| 81 | * @param[in] src_stride_y Stride of the source tensor in Y dimension (in bytes) |
| 82 | * @param[in] src_step_y src_stride_y * number of elements along Y processed per workitem(in bytes) |
| 83 | * @param[in] src_stride_z Stride of the source tensor in Z dimension (in bytes) |
| 84 | * @param[in] src_step_z src_stride_z * number of elements along Z processed per workitem(in bytes) |
| 85 | * @param[in] src_offset_first_element_in_bytes The offset of the first element in the source tensor |
| 86 | * @param[out] dst_ptr Pointer to the destination tensor. Supported data types: same as @p src_ptr |
| 87 | * @param[in] dst_stride_x Stride of the destination tensor in X dimension (in bytes) |
| 88 | * @param[in] dst_step_x dst_stride_x * number of elements along X processed per workitem(in bytes) |
| 89 | * @param[in] dst_stride_y Stride of the destination tensor in Y dimension (in bytes) |
| 90 | * @param[in] dst_step_y dst_stride_y * number of elements along Z processed per workitem(in bytes) |
| 91 | * @param[in] dst_stride_z Stride of the destination tensor in Z dimension (in bytes) |
| 92 | * @param[in] dst_step_z dst_stride_z * number of elements along Z processed per workitem(in bytes) |
| 93 | * @param[in] dst_offset_first_element_in_bytes The offset of the first element in the destination tensor |
Joel Liang | f1f3ebd | 2017-11-10 09:59:19 +0800 | [diff] [blame] | 94 | * @param[in] weights_ptr Pointer to the weights tensor. Supported data types: same as @p src_ptr |
steniu01 | db00668 | 2017-08-09 16:26:22 +0100 | [diff] [blame] | 95 | * @param[in] weights_stride_x Stride of the weights tensor in X dimension (in bytes) |
| 96 | * @param[in] weights_step_x weights_stride_x * number of elements along X processed per workitem(in bytes) |
| 97 | * @param[in] weights_stride_y Stride of the weights tensor in Y dimension (in bytes) |
| 98 | * @param[in] weights_step_y weights_stride_y * number of elements along y processed per workitem(in bytes) |
| 99 | * @param[in] weights_stride_z Stride of the weights tensor in Z dimension (in bytes) |
| 100 | * @param[in] weights_step_z weights_stride_z * number of elements along Z processed per workitem(in bytes) |
| 101 | * @param[in] weights_offset_first_element_in_bytes The offset of the first element in the weights tensor |
| 102 | * @param[in] biases_ptr Pointer to the biases tensor. Same as @p src_ptr |
| 103 | * @param[in] biases_stride_x Stride of the biases tensor in X dimension (in bytes) |
| 104 | * @param[in] biases_step_x biases_stride_x * number of elements along X processed per workitem(in bytes) |
| 105 | * @param[in] biases_offset_first_element_in_bytes The offset of the first element in the biases tensor |
| 106 | * @param[in] weights_stride_w Stride of the weights tensor in the 4th dimension |
| 107 | */ |
steniu01 | db00668 | 2017-08-09 16:26:22 +0100 | [diff] [blame] | 108 | __kernel void direct_convolution5x5( |
| 109 | TENSOR3D_DECLARATION(src), |
| 110 | TENSOR3D_DECLARATION(dst), |
| 111 | TENSOR3D_DECLARATION(weights), |
| 112 | #ifdef HAS_BIAS |
| 113 | VECTOR_DECLARATION(biases), |
| 114 | #endif /* defined(HAS_BIAS) */ |
| 115 | unsigned int weights_stride_w) |
| 116 | { |
| 117 | Image src = CONVERT_TO_IMAGE_STRUCT(src); |
| 118 | Tensor3D weights = CONVERT_TO_TENSOR3D_STRUCT_NO_STEP(weights); |
| 119 | Tensor3D dst = CONVERT_TO_TENSOR3D_STRUCT(dst); |
| 120 | |
| 121 | VEC_DATA_TYPE(DATA_TYPE, 8) |
Pablo Tello | 3d31946 | 2018-06-21 15:13:17 +0100 | [diff] [blame] | 122 | values0 = 0; |
steniu01 | db00668 | 2017-08-09 16:26:22 +0100 | [diff] [blame] | 123 | |
| 124 | __global uchar *weights_addr = (__global uchar *)tensor3D_offset(&weights, 0, 0, 0); |
| 125 | __global uchar *src_addr = (__global uchar *)offset(&src, 0, 0); |
| 126 | |
| 127 | const int kernel_index = get_global_id(2); |
| 128 | weights_addr += kernel_index * weights_stride_w; |
| 129 | |
Gian Marco Iodice | 744b5ed | 2017-10-06 15:44:27 +0100 | [diff] [blame] | 130 | for(volatile int d = 0; d < WEIGHTS_DEPTH; ++d) |
steniu01 | db00668 | 2017-08-09 16:26:22 +0100 | [diff] [blame] | 131 | { |
Pablo Tello | 3d31946 | 2018-06-21 15:13:17 +0100 | [diff] [blame] | 132 | CONVOLUTION1x5(values0, (__global DATA_TYPE *)src_addr, (__global DATA_TYPE *)weights_addr); |
| 133 | CONVOLUTION1x5(values0, (__global DATA_TYPE *)(src_addr + 1 * src_stride_y), (__global DATA_TYPE *)(weights_addr + 1 * weights_stride_y)); |
| 134 | CONVOLUTION1x5(values0, (__global DATA_TYPE *)(src_addr + 2 * src_stride_y), (__global DATA_TYPE *)(weights_addr + 2 * weights_stride_y)); |
| 135 | CONVOLUTION1x5(values0, (__global DATA_TYPE *)(src_addr + 3 * src_stride_y), (__global DATA_TYPE *)(weights_addr + 3 * weights_stride_y)); |
| 136 | CONVOLUTION1x5(values0, (__global DATA_TYPE *)(src_addr + 4 * src_stride_y), (__global DATA_TYPE *)(weights_addr + 4 * weights_stride_y)); |
steniu01 | db00668 | 2017-08-09 16:26:22 +0100 | [diff] [blame] | 137 | |
| 138 | src_addr += src_stride_z; |
| 139 | weights_addr += weights_stride_z; |
| 140 | } |
| 141 | |
| 142 | #ifdef HAS_BIAS |
| 143 | Vector biases = CONVERT_TO_VECTOR_STRUCT_NO_STEP(biases); |
| 144 | |
Pablo Tello | 3d31946 | 2018-06-21 15:13:17 +0100 | [diff] [blame] | 145 | values0 += (VEC_DATA_TYPE(DATA_TYPE, 8)) * ((__global DATA_TYPE *)(vector_offset(&biases, kernel_index))); |
steniu01 | db00668 | 2017-08-09 16:26:22 +0100 | [diff] [blame] | 146 | #endif /* defined(HAS_BIAS) */ |
| 147 | |
Pablo Tello | 3d31946 | 2018-06-21 15:13:17 +0100 | [diff] [blame] | 148 | vstore8(values0, 0, (__global DATA_TYPE *)dst.ptr); |
steniu01 | db00668 | 2017-08-09 16:26:22 +0100 | [diff] [blame] | 149 | } |
| 150 | #endif // defined(DATA_TYPE) && defined(STRIDE_X) && defined(WEIGHTS_DEPTH) |
Gian Marco Iodice | 1246b63 | 2017-08-16 18:38:32 +0100 | [diff] [blame] | 151 | |
| 152 | #if defined(WEIGHTS_DEPTH) |
| 153 | |
| 154 | #define CONVOLUTION1x5_BIFROST(acc, src0, weights_row00, weights_row01) \ |
| 155 | ({ \ |
| 156 | acc.s0 = mad(src0.s0, weights_row00.s0, acc.s0); \ |
| 157 | acc.s1 = mad(src0.s1, weights_row00.s0, acc.s1); \ |
| 158 | acc.s2 = mad(src0.s2, weights_row00.s0, acc.s2); \ |
| 159 | acc.s3 = mad(src0.s3, weights_row00.s0, acc.s3); \ |
| 160 | acc.s0 = mad(src0.s1, weights_row00.s1, acc.s0); \ |
| 161 | acc.s1 = mad(src0.s2, weights_row00.s1, acc.s1); \ |
| 162 | acc.s2 = mad(src0.s3, weights_row00.s1, acc.s2); \ |
| 163 | acc.s3 = mad(src0.s4, weights_row00.s1, acc.s3); \ |
| 164 | acc.s0 = mad(src0.s2, weights_row00.s2, acc.s0); \ |
| 165 | acc.s1 = mad(src0.s3, weights_row00.s2, acc.s1); \ |
| 166 | acc.s2 = mad(src0.s4, weights_row00.s2, acc.s2); \ |
| 167 | acc.s3 = mad(src0.s5, weights_row00.s2, acc.s3); \ |
| 168 | acc.s0 = mad(src0.s3, weights_row00.s3, acc.s0); \ |
| 169 | acc.s1 = mad(src0.s4, weights_row00.s3, acc.s1); \ |
| 170 | acc.s2 = mad(src0.s5, weights_row00.s3, acc.s2); \ |
| 171 | acc.s3 = mad(src0.s6, weights_row00.s3, acc.s3); \ |
| 172 | acc.s0 = mad(src0.s4, weights_row01, acc.s0); \ |
| 173 | acc.s1 = mad(src0.s5, weights_row01, acc.s1); \ |
| 174 | acc.s2 = mad(src0.s6, weights_row01, acc.s2); \ |
| 175 | acc.s3 = mad(src0.s7, weights_row01, acc.s3); \ |
| 176 | }) |
| 177 | |
| 178 | /** An optimized direct convolution 5x5 OpenCL kernel for Bifrost architectures when the data type is F32 |
| 179 | * |
| 180 | * @note This OpenCL kernel works only with stride_x and stride_y equal to 1 |
| 181 | * @note The third dimensions of the weights tensors must be passed at compile time using -DWEIGHTS_DEPTH |
| 182 | * @note If biases are used then -DHAS_BIAS has to be passed at compile time |
| 183 | * |
| 184 | * @param[in] src_ptr Pointer to the source tensor. Supported data types: F32 |
| 185 | * @param[in] src_stride_x Stride of the source tensor in X dimension (in bytes) |
| 186 | * @param[in] src_step_x src_stride_x * number of elements along X processed per workitem(in bytes) |
| 187 | * @param[in] src_stride_y Stride of the source tensor in Y dimension (in bytes) |
| 188 | * @param[in] src_step_y src_stride_y * number of elements along Y processed per workitem(in bytes) |
| 189 | * @param[in] src_stride_z Stride of the source tensor in Z dimension (in bytes) |
| 190 | * @param[in] src_step_z src_stride_z * number of elements along Z processed per workitem(in bytes) |
| 191 | * @param[in] src_offset_first_element_in_bytes The offset of the first element in the source tensor |
| 192 | * @param[out] dst_ptr Pointer to the destination tensor. Supported data types: same as @p src_ptr |
| 193 | * @param[in] dst_stride_x Stride of the destination tensor in X dimension (in bytes) |
| 194 | * @param[in] dst_step_x dst_stride_x * number of elements along X processed per workitem(in bytes) |
| 195 | * @param[in] dst_stride_y Stride of the destination tensor in Y dimension (in bytes) |
| 196 | * @param[in] dst_step_y dst_stride_y * number of elements along Z processed per workitem(in bytes) |
| 197 | * @param[in] dst_stride_z Stride of the destination tensor in Z dimension (in bytes) |
| 198 | * @param[in] dst_step_z dst_stride_z * number of elements along Z processed per workitem(in bytes) |
| 199 | * @param[in] dst_offset_first_element_in_bytes The offset of the first element in the destination tensor |
Joel Liang | f1f3ebd | 2017-11-10 09:59:19 +0800 | [diff] [blame] | 200 | * @param[in] weights_ptr Pointer to the weights tensor. Supported data types: same as @p src_ptr |
Gian Marco Iodice | 1246b63 | 2017-08-16 18:38:32 +0100 | [diff] [blame] | 201 | * @param[in] weights_stride_x Stride of the weights tensor in X dimension (in bytes) |
| 202 | * @param[in] weights_step_x weights_stride_x * number of elements along X processed per workitem(in bytes) |
| 203 | * @param[in] weights_stride_y Stride of the weights tensor in Y dimension (in bytes) |
| 204 | * @param[in] weights_step_y weights_stride_y * number of elements along y processed per workitem(in bytes) |
| 205 | * @param[in] weights_stride_z Stride of the weights tensor in Z dimension (in bytes) |
| 206 | * @param[in] weights_step_z weights_stride_z * number of elements along Z processed per workitem(in bytes) |
| 207 | * @param[in] weights_offset_first_element_in_bytes The offset of the first element in the weights tensor |
| 208 | * @param[in] biases_ptr Pointer to the biases tensor. Same as @p src_ptr |
| 209 | * @param[in] biases_stride_x Stride of the biases tensor in X dimension (in bytes) |
| 210 | * @param[in] biases_step_x biases_stride_x * number of elements along X processed per workitem(in bytes) |
| 211 | * @param[in] biases_offset_first_element_in_bytes The offset of the first element in the biases tensor |
| 212 | * @param[in] weights_stride_w Stride of the weights tensor in the 4th dimension |
| 213 | */ |
| 214 | __kernel void direct_convolution5x5_f32_bifrost( |
| 215 | TENSOR3D_DECLARATION(src), |
| 216 | TENSOR3D_DECLARATION(dst), |
| 217 | TENSOR3D_DECLARATION(weights), |
| 218 | #ifdef HAS_BIAS |
| 219 | VECTOR_DECLARATION(biases), |
| 220 | #endif /* defined(HAS_BIAS) */ |
| 221 | unsigned int weights_stride_w) |
| 222 | { |
| 223 | // Get the kernel index |
| 224 | const int kernel_index = get_global_id(2); |
| 225 | |
| 226 | Image src = CONVERT_TO_IMAGE_STRUCT(src); |
| 227 | Tensor3D dst = CONVERT_TO_TENSOR3D_STRUCT(dst); |
| 228 | |
Pablo Tello | 3d31946 | 2018-06-21 15:13:17 +0100 | [diff] [blame] | 229 | float4 values0 = 0.0f; |
| 230 | float4 values1 = 0.0f; |
Gian Marco Iodice | 1246b63 | 2017-08-16 18:38:32 +0100 | [diff] [blame] | 231 | |
| 232 | __global uchar *weights_addr = (__global uchar *)(weights_ptr + weights_offset_first_element_in_bytes + kernel_index * weights_stride_w); |
| 233 | __global uchar *src_addr = (__global uchar *)offset(&src, 0, 0); |
| 234 | |
| 235 | // Note: Since each work-item computes 4x2 elements, we need to load 6 rows from the input tensor |
| 236 | |
| 237 | for(ushort d = 0; d < (ushort)WEIGHTS_DEPTH; ++d) |
| 238 | { |
| 239 | // Load the weights from row0 and row1 |
| 240 | float4 weights_row00 = vload4(0, (__global float *)(weights_addr + 0 * weights_stride_y)); |
| 241 | float weights_row01 = *((__global float *)(weights_addr + 0 * weights_stride_y) + 4); |
| 242 | float4 weights_row10 = vload4(0, (__global float *)(weights_addr + 1 * weights_stride_y)); |
| 243 | float weights_row11 = *((__global float *)(weights_addr + 1 * weights_stride_y) + 4); |
| 244 | float8 src0; |
| 245 | |
| 246 | // Load values from row0 of input tensor |
| 247 | src0 = vload8(0, (__global float *)(src_addr + 0 * src_stride_y)); |
| 248 | |
| 249 | // Accumulate |
Pablo Tello | 3d31946 | 2018-06-21 15:13:17 +0100 | [diff] [blame] | 250 | CONVOLUTION1x5_BIFROST(values0, src0, weights_row00, weights_row01); |
Gian Marco Iodice | 1246b63 | 2017-08-16 18:38:32 +0100 | [diff] [blame] | 251 | |
| 252 | // Load values from row1 of input tensor |
| 253 | src0 = vload8(0, (__global float *)(src_addr + 1 * src_stride_y)); |
| 254 | |
| 255 | // Accumulate |
Pablo Tello | 3d31946 | 2018-06-21 15:13:17 +0100 | [diff] [blame] | 256 | CONVOLUTION1x5_BIFROST(values0, src0, weights_row10, weights_row11); |
| 257 | CONVOLUTION1x5_BIFROST(values1, src0, weights_row00, weights_row01); |
Gian Marco Iodice | 1246b63 | 2017-08-16 18:38:32 +0100 | [diff] [blame] | 258 | |
| 259 | // Load values from row2 of input tensor |
| 260 | src0 = vload8(0, (__global float *)(src_addr + 2 * src_stride_y)); |
| 261 | |
| 262 | // Load weights from row2 |
| 263 | weights_row00 = vload4(0, (__global float *)(weights_addr + 2 * weights_stride_y)); |
| 264 | weights_row01 = *((__global float *)(weights_addr + 2 * weights_stride_y) + 4); |
| 265 | |
| 266 | // Accumulate |
Pablo Tello | 3d31946 | 2018-06-21 15:13:17 +0100 | [diff] [blame] | 267 | CONVOLUTION1x5_BIFROST(values0, src0, weights_row00, weights_row01); |
| 268 | CONVOLUTION1x5_BIFROST(values1, src0, weights_row10, weights_row11); |
Gian Marco Iodice | 1246b63 | 2017-08-16 18:38:32 +0100 | [diff] [blame] | 269 | |
| 270 | // Load values from row3 of input tensor |
| 271 | src0 = vload8(0, (__global float *)(src_addr + 3 * src_stride_y)); |
| 272 | |
| 273 | // Load weights from row3 |
| 274 | weights_row10 = vload4(0, (__global float *)(weights_addr + 3 * weights_stride_y)); |
| 275 | weights_row11 = *((__global float *)(weights_addr + 3 * weights_stride_y) + 4); |
| 276 | |
| 277 | // Accumulate |
Pablo Tello | 3d31946 | 2018-06-21 15:13:17 +0100 | [diff] [blame] | 278 | CONVOLUTION1x5_BIFROST(values0, src0, weights_row10, weights_row11); |
| 279 | CONVOLUTION1x5_BIFROST(values1, src0, weights_row00, weights_row01); |
Gian Marco Iodice | 1246b63 | 2017-08-16 18:38:32 +0100 | [diff] [blame] | 280 | |
| 281 | // Load values from row4 of input tensor |
| 282 | src0 = vload8(0, (__global float *)(src_addr + 4 * src_stride_y)); |
| 283 | |
| 284 | // Load weights from row4 |
| 285 | weights_row00 = vload4(0, (__global float *)(weights_addr + 4 * weights_stride_y)); |
| 286 | weights_row01 = *((__global float *)(weights_addr + 4 * weights_stride_y) + 4); |
| 287 | |
Pablo Tello | 3d31946 | 2018-06-21 15:13:17 +0100 | [diff] [blame] | 288 | CONVOLUTION1x5_BIFROST(values0, src0, weights_row00, weights_row01); |
| 289 | CONVOLUTION1x5_BIFROST(values1, src0, weights_row10, weights_row11); |
Gian Marco Iodice | 1246b63 | 2017-08-16 18:38:32 +0100 | [diff] [blame] | 290 | |
| 291 | // Load values from row5 of input tensor |
| 292 | src0 = vload8(0, (__global float *)(src_addr + 5 * src_stride_y)); |
| 293 | |
| 294 | // Accumulate |
Pablo Tello | 3d31946 | 2018-06-21 15:13:17 +0100 | [diff] [blame] | 295 | CONVOLUTION1x5_BIFROST(values1, src0, weights_row00, weights_row01); |
Gian Marco Iodice | 1246b63 | 2017-08-16 18:38:32 +0100 | [diff] [blame] | 296 | |
| 297 | src_addr += src_stride_z; |
| 298 | weights_addr += weights_stride_z; |
| 299 | } |
| 300 | |
| 301 | #ifdef HAS_BIAS |
| 302 | Vector biases = CONVERT_TO_VECTOR_STRUCT_NO_STEP(biases); |
| 303 | |
| 304 | float4 bias = (float4) * ((__global float *)(vector_offset(&biases, kernel_index))); |
| 305 | |
Pablo Tello | 3d31946 | 2018-06-21 15:13:17 +0100 | [diff] [blame] | 306 | values0 += bias; |
| 307 | values1 += bias; |
Gian Marco Iodice | 1246b63 | 2017-08-16 18:38:32 +0100 | [diff] [blame] | 308 | #endif /* defined(HAS_BIAS) */ |
| 309 | |
Pablo Tello | 3d31946 | 2018-06-21 15:13:17 +0100 | [diff] [blame] | 310 | vstore4(values0, 0, (__global float *)(dst.ptr + 0 * dst_stride_y)); |
| 311 | vstore4(values1, 0, (__global float *)(dst.ptr + 1 * dst_stride_y)); |
Gian Marco Iodice | 1246b63 | 2017-08-16 18:38:32 +0100 | [diff] [blame] | 312 | } |
| 313 | #endif // defined(WEIGHTS_DEPTH) |