1# Copyright 2021 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- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x8c4__aarch32_neondot_cortex_a55 7 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 8 k-block: 8 9 assembly: true 10- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x8c4__aarch32_neondot_ld64 11 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 12 k-block: 8 13 assembly: true 14- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x8__aarch32_neon_mlal_lane_prfm_ld64 15 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 16 k-block: 8 17 assembly: true 18- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x8__aarch32_neon_mlal_lane_ld64 19 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 20 k-block: 8 21 assembly: true 22- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x8__aarch64_neon_mlal_lane_prfm_ld64 23 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 24 k-block: 8 25 assembly: true 26- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x8__aarch64_neon_mlal_lane_ld64 27 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 28 k-block: 8 29 assembly: true 30- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x8c2__neon_mull_ld1r 31 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 32 k-block: 8 33- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x8c2__neon_mull_ld1r 34 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 35 k-block: 8 36- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x8c2__neon_mull_ld1r 37 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 38 k-block: 8 39- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x8c2__neon_mull_ld1r 40 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 41 k-block: 8 42- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x16c2__neon_mull_ld1r 43 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 44 k-block: 8 45- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x16c2__neon_mull_ld1r 46 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 47 k-block: 8 48- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x16c2__neon_mull_ld1r 49 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 50 k-block: 8 51- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x16c2__neon_mull_ld1r 52 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 53 k-block: 8 54- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x8c2__neon_mlal_ld1r 55 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 56 k-block: 16 57- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x8c2__neon_mlal_ld1r 58 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 59 k-block: 16 60- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x8c2__neon_mlal_ld1r 61 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 62 k-block: 16 63- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x8c2__neon_mlal_ld1r 64 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 65 k-block: 16 66- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x16c2__neon_mlal_ld1r 67 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 68 k-block: 16 69- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x16c2__neon_mlal_ld1r 70 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 71 k-block: 16 72- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x16c2__neon_mlal_ld1r 73 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 74 k-block: 16 75- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x16c2__neon_mlal_ld1r 76 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 77 k-block: 16 78- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x8c2__neon_mull_ld2r 79 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 80 k-block: 8 81- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x8c2__neon_mull_ld2r 82 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 83 k-block: 8 84- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x8c2__neon_mull_ld2r 85 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 86 k-block: 8 87- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x8c2__neon_mull_ld2r 88 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 89 k-block: 8 90- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x16c2__neon_mull_ld2r 91 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 92 k-block: 8 93- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x16c2__neon_mull_ld2r 94 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 95 k-block: 8 96- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x16c2__neon_mull_ld2r 97 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 98 k-block: 8 99- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x16c2__neon_mull_ld2r 100 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 101 k-block: 8 102- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x8c2__neon_mlal_ld2r 103 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 104 k-block: 16 105- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x8c2__neon_mlal_ld2r 106 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 107 k-block: 16 108- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x8c2__neon_mlal_ld2r 109 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 110 k-block: 16 111- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x8c2__neon_mlal_ld2r 112 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 113 k-block: 16 114- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x16c2__neon_mlal_ld2r 115 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 116 k-block: 16 117- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x16c2__neon_mlal_ld2r 118 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 119 k-block: 16 120- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x16c2__neon_mlal_ld2r 121 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 122 k-block: 16 123- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x16c2__neon_mlal_ld2r 124 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 125 k-block: 16 126- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x8c2__neon_mull_ld4r 127 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 128 k-block: 8 129- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x8c2__neon_mull_ld4r 130 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 131 k-block: 8 132- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x8c2__neon_mull_ld4r 133 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 134 k-block: 8 135- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x8c2__neon_mull_ld4r 136 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 137 k-block: 8 138- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x16c2__neon_mull_ld4r 139 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 140 k-block: 8 141- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x16c2__neon_mull_ld4r 142 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 143 k-block: 8 144- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x16c2__neon_mull_ld4r 145 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 146 k-block: 8 147- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x16c2__neon_mull_ld4r 148 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 149 k-block: 8 150- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x8c2__neon_mlal_ld4r 151 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 152 k-block: 16 153- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x8c2__neon_mlal_ld4r 154 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 155 k-block: 16 156- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x8c2__neon_mlal_ld4r 157 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 158 k-block: 16 159- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x8c2__neon_mlal_ld4r 160 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 161 k-block: 16 162- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x16c2__neon_mlal_ld4r 163 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 164 k-block: 16 165- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x16c2__neon_mlal_ld4r 166 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 167 k-block: 16 168- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x16c2__neon_mlal_ld4r 169 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 170 k-block: 16 171- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x16c2__neon_mlal_ld4r 172 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 173 k-block: 16 174- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x8c4s2__neon_mull 175 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 176 k-block: 8 177- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x8c4s2__neon_mull 178 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 179 k-block: 8 180- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x8c4s2__neon_mull 181 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 182 k-block: 8 183- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x8c4s2__neon_mull 184 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 185 k-block: 8 186- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x16c4s2__neon_mull 187 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 188 k-block: 8 189- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x16c4s2__neon_mull 190 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 191 k-block: 8 192- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x16c4s2__neon_mull 193 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 194 k-block: 8 195- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x16c4s2__neon_mull 196 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 197 k-block: 8 198- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x8c4s2__neon_mlal 199 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 200 k-block: 16 201- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x8c4s2__neon_mlal 202 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 203 k-block: 16 204- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x8c4s2__neon_mlal 205 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 206 k-block: 16 207- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x8c4s2__neon_mlal 208 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 209 k-block: 16 210- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x16c4s2__neon_mlal 211 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 212 k-block: 16 213- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x16c4s2__neon_mlal 214 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 215 k-block: 16 216- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x16c4s2__neon_mlal 217 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 218 k-block: 16 219- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x16c4s2__neon_mlal 220 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 221 k-block: 16 222- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x8c2s4__neon_mull 223 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 224 k-block: 8 225- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x8c2s4__neon_mull 226 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 227 k-block: 8 228- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x8c2s4__neon_mull 229 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 230 k-block: 8 231- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x8c2s4__neon_mull 232 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 233 k-block: 8 234- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x16c2s4__neon_mull 235 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 236 k-block: 8 237- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x16c2s4__neon_mull 238 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 239 k-block: 8 240- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x16c2s4__neon_mull 241 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 242 k-block: 8 243- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x16c2s4__neon_mull 244 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 245 k-block: 8 246- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x8c2s4__neon_mlal 247 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 248 k-block: 16 249- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x8c2s4__neon_mlal 250 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 251 k-block: 16 252- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x8c2s4__neon_mlal 253 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 254 k-block: 16 255- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x8c2s4__neon_mlal 256 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 257 k-block: 16 258- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x16c2s4__neon_mlal 259 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 260 k-block: 16 261- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x16c2s4__neon_mlal 262 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 263 k-block: 16 264- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x16c2s4__neon_mlal 265 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 266 k-block: 16 267- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x16c2s4__neon_mlal 268 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 269 k-block: 16 270- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x8c4__neon_mull_dup 271 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 272 k-block: 8 273- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x8c4__neon_mull_dup 274 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 275 k-block: 8 276- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x8c4__neon_mull_dup 277 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 278 k-block: 8 279- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x8c4__neon_mull_dup 280 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 281 k-block: 8 282- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x16c4__neon_mull_dup 283 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 284 k-block: 8 285- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x16c4__neon_mull_dup 286 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 287 k-block: 8 288- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x16c4__neon_mull_dup 289 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 290 k-block: 8 291- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x16c4__neon_mull_dup 292 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 293 k-block: 8 294- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x8c4__neon_mlal_dup 295 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 296 k-block: 16 297- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x8c4__neon_mlal_dup 298 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 299 k-block: 16 300- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x8c4__neon_mlal_dup 301 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 302 k-block: 16 303- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x8c4__neon_mlal_dup 304 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 305 k-block: 16 306- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x16c4__neon_mlal_dup 307 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 308 k-block: 16 309- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x16c4__neon_mlal_dup 310 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 311 k-block: 16 312- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x16c4__neon_mlal_dup 313 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 314 k-block: 16 315- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x16c4__neon_mlal_dup 316 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 317 k-block: 16 318- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x8c4__neon_mull_ld1r 319 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 320 k-block: 8 321- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x8c4__neon_mull_ld1r 322 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 323 k-block: 8 324- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x8c4__neon_mull_ld1r 325 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 326 k-block: 8 327- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x8c4__neon_mull_ld1r 328 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 329 k-block: 8 330- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x16c4__neon_mull_ld1r 331 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 332 k-block: 8 333- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x16c4__neon_mull_ld1r 334 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 335 k-block: 8 336- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x16c4__neon_mull_ld1r 337 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 338 k-block: 8 339- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x16c4__neon_mull_ld1r 340 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 341 k-block: 8 342- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x8c4__neon_mlal_ld1r 343 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 344 k-block: 16 345- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x8c4__neon_mlal_ld1r 346 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 347 k-block: 16 348- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x8c4__neon_mlal_ld1r 349 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 350 k-block: 16 351- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x8c4__neon_mlal_ld1r 352 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 353 k-block: 16 354- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x16c4__neon_mlal_ld1r 355 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 356 k-block: 16 357- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x16c4__neon_mlal_ld1r 358 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 359 k-block: 16 360- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x16c4__neon_mlal_ld1r 361 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 362 k-block: 16 363- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x16c4__neon_mlal_ld1r 364 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 365 k-block: 16 366- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x8c4__neon_mull_ld2r 367 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 368 k-block: 8 369- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x8c4__neon_mull_ld2r 370 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 371 k-block: 8 372- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x8c4__neon_mull_ld2r 373 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 374 k-block: 8 375- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x8c4__neon_mull_ld2r 376 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 377 k-block: 8 378- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x16c4__neon_mull_ld2r 379 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 380 k-block: 8 381- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x16c4__neon_mull_ld2r 382 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 383 k-block: 8 384- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x16c4__neon_mull_ld2r 385 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 386 k-block: 8 387- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x16c4__neon_mull_ld2r 388 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 389 k-block: 8 390- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x8c4__neon_mlal_ld2r 391 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 392 k-block: 16 393- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x8c4__neon_mlal_ld2r 394 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 395 k-block: 16 396- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x8c4__neon_mlal_ld2r 397 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 398 k-block: 16 399- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x8c4__neon_mlal_ld2r 400 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 401 k-block: 16 402- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x16c4__neon_mlal_ld2r 403 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 404 k-block: 16 405- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x16c4__neon_mlal_ld2r 406 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 407 k-block: 16 408- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x16c4__neon_mlal_ld2r 409 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 410 k-block: 16 411- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x16c4__neon_mlal_ld2r 412 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 413 k-block: 16 414- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x8c2__neon_mull_dup 415 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 416 k-block: 8 417- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x8c2__neon_mull_dup 418 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 419 k-block: 8 420- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x8c2__neon_mull_dup 421 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 422 k-block: 8 423- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x8c2__neon_mull_dup 424 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 425 k-block: 8 426- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x16c2__neon_mull_dup 427 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 428 k-block: 8 429- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x16c2__neon_mull_dup 430 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 431 k-block: 8 432- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x16c2__neon_mull_dup 433 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 434 k-block: 8 435- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x16c2__neon_mull_dup 436 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 437 k-block: 8 438- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x8c2__neon_mlal_dup 439 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 440 k-block: 16 441- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x8c2__neon_mlal_dup 442 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 443 k-block: 16 444- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x8c2__neon_mlal_dup 445 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 446 k-block: 16 447- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x8c2__neon_mlal_dup 448 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 449 k-block: 16 450- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x16c2__neon_mlal_dup 451 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 452 k-block: 16 453- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x16c2__neon_mlal_dup 454 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 455 k-block: 16 456- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x16c2__neon_mlal_dup 457 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 458 k-block: 16 459- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x16c2__neon_mlal_dup 460 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 461 k-block: 16 462- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x8c8__aarch64_neon_mlal 463 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 464 k-block: 16 465 assembly: true 466- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x8c8__aarch64_neon_mlal_prfm 467 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 468 k-block: 16 469 assembly: true 470- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x8c8__aarch64_neon_mlal_cortex_a53 471 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 472 k-block: 16 473 assembly: true 474- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x8c8__aarch64_neon_mlal_prfm_cortex_a53 475 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 476 k-block: 16 477 assembly: true 478- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x8c16__aarch64_neon_mlal 479 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 480 k-block: 16 481 assembly: true 482- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x8c8__aarch64_neon_mlal 483 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 484 k-block: 16 485 assembly: true 486- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x8c8__aarch64_neon_mlal_prfm 487 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 488 k-block: 16 489 assembly: true 490- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x8c8__aarch64_neon_mlal_cortex_a53 491 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 492 k-block: 16 493 assembly: true 494- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x8c8__aarch64_neon_mlal_prfm_cortex_a53 495 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 496 k-block: 16 497 assembly: true 498- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x16__aarch64_neon_mlal_lane_prfm_cortex_a53 499 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 500 k-block: 8 501 assembly: true 502- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x16__aarch64_neon_mlal_lane_cortex_a53 503 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 504 k-block: 8 505 assembly: true 506- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x16__aarch64_neon_mlal_lane_prfm_ld64 507 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 508 k-block: 8 509 assembly: true 510- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x16__aarch64_neon_mlal_lane_ld64 511 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 512 k-block: 8 513 assembly: true 514- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x16c4__aarch64_neondot_cortex_a55 515 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 516 k-block: 16 517 assembly: true 518- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x16c4__aarch64_neondot_ld64 519 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 520 k-block: 8 521 assembly: true 522- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x16c4__aarch64_neondot_ld128 523 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 524 k-block: 16 525 assembly: true 526- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x8c8__neon_mull 527 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 528 k-block: 8 529- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x8c8__neon_mull 530 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 531 k-block: 8 532- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x8c8__neon_mull 533 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 534 k-block: 8 535- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x8c8__neon_mull 536 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 537 k-block: 8 538- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x16c8__neon_mull 539 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 540 k-block: 8 541- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x16c8__neon_mull 542 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 543 k-block: 8 544- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x16c8__neon_mull 545 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 546 k-block: 8 547- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x16c8__neon_mull 548 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 549 k-block: 8 550- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x8c8__neon_mlal 551 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 552 k-block: 16 553- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x8c8__neon_mlal 554 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 555 k-block: 16 556- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x8c8__neon_mlal 557 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 558 k-block: 16 559- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x8c8__neon_mlal 560 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 561 k-block: 16 562- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x16c8__neon_mlal 563 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 564 k-block: 16 565- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x16c8__neon_mlal 566 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 567 k-block: 16 568- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x16c8__neon_mlal 569 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 570 k-block: 16 571- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x16c8__neon_mlal 572 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 573 k-block: 16 574- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x8c16__neon_mlal 575 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 576 k-block: 16 577- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x8c16__neon_mlal 578 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 579 k-block: 16 580- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x8c16__neon_mlal 581 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 582 k-block: 16 583- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x8c16__neon_mlal 584 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 585 k-block: 16 586- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x16c16__neon_mlal 587 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 588 k-block: 16 589- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x16c16__neon_mlal 590 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 591 k-block: 16 592- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x16c16__neon_mlal 593 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 594 k-block: 16 595- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x16c16__neon_mlal 596 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 597 k-block: 16 598- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x8c4__neondot 599 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 600 k-block: 8 601- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x8c4__neondot 602 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 603 k-block: 8 604- name: xnn_qs8_igemm_minmax_rndnu_ukernel_6x8c4__neondot 605 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 606 k-block: 8 607- name: xnn_qs8_igemm_minmax_rndnu_ukernel_8x8c4__neondot 608 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 609 k-block: 8 610- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x16c4__neondot 611 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 612 k-block: 8 613- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x16c4__neondot 614 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 615 k-block: 8 616- name: xnn_qs8_igemm_minmax_rndnu_ukernel_6x16c4__neondot 617 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 618 k-block: 8 619- name: xnn_qs8_igemm_minmax_rndnu_ukernel_8x16c4__neondot 620 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 621 k-block: 8 622- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x8__neon_mlal_lane 623 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 624 k-block: 8 625- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x8__neon_mlal_lane 626 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 627 k-block: 8 628- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x8__neon_mlal_lane 629 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 630 k-block: 8 631- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x8__neon_mlal_lane 632 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 633 k-block: 8 634- name: xnn_qs8_igemm_minmax_rndnu_ukernel_6x8__neon_mlal_lane 635 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 636 k-block: 8 637- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x16__neon_mlal_lane 638 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 639 k-block: 8 640- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x16__neon_mlal_lane 641 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 642 k-block: 8 643- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x16__neon_mlal_lane 644 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 645 k-block: 8 646- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x16__neon_mlal_lane 647 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 648 k-block: 8 649- name: xnn_qs8_igemm_minmax_rndnu_ukernel_6x16__neon_mlal_lane 650 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 651 k-block: 8 652- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x8__neon_mlal_lane_prfm 653 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 654 k-block: 8 655- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x8__neon_mlal_lane_prfm 656 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 657 k-block: 8 658- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x8__neon_mlal_lane_prfm 659 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 660 k-block: 8 661- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x8__neon_mlal_lane_prfm 662 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 663 k-block: 8 664- name: xnn_qs8_igemm_minmax_rndnu_ukernel_6x8__neon_mlal_lane_prfm 665 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 666 k-block: 8 667- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x16__neon_mlal_lane_prfm 668 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 669 k-block: 8 670- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x16__neon_mlal_lane_prfm 671 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 672 k-block: 8 673- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x16__neon_mlal_lane_prfm 674 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 675 k-block: 8 676- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x16__neon_mlal_lane_prfm 677 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 678 k-block: 8 679- name: xnn_qs8_igemm_minmax_rndnu_ukernel_6x16__neon_mlal_lane_prfm 680 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 681 k-block: 8 682- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x8__neon_mull_addw_dup 683 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 684 k-block: 8 685- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x8__neon_mull_addw_dup 686 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 687 k-block: 8 688- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x8__neon_mull_addw_dup 689 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 690 k-block: 8 691- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x8__neon_mull_addw_dup 692 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 693 k-block: 8 694- name: xnn_qs8_igemm_minmax_rndnu_ukernel_1x16__neon_mull_addw_dup 695 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 696 k-block: 8 697- name: xnn_qs8_igemm_minmax_rndnu_ukernel_2x16__neon_mull_addw_dup 698 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 699 k-block: 8 700- name: xnn_qs8_igemm_minmax_rndnu_ukernel_3x16__neon_mull_addw_dup 701 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 702 k-block: 8 703- name: xnn_qs8_igemm_minmax_rndnu_ukernel_4x16__neon_mull_addw_dup 704 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 705 k-block: 8 706- name: xnn_generate_qs8_igemm_rndnu_ukernel_4x8c4__aarch32_neondot_ld64 707 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 708 k-block: 8 709 assembly: true 710- name: xnn_generate_qs8_igemm_rndnu_ukernel_4x8__aarch32_neon_mlal_lane_prfm_ld64 711 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 712 k-block: 8 713 assembly: true 714- name: xnn_generate_qs8_igemm_rndnu_ukernel_4x8__aarch32_neon_mlal_lane_ld64 715 init: xnn_init_qs8_conv_minmax_rndnu_neon_params 716 k-block: 8 717 assembly: true 718