| OLD | NEW |
|---|
| (Empty) |
| 1 ; | |
| 2 ; Copyright (c) 2014 The WebM project authors. All Rights Reserved. | |
| 3 ; | |
| 4 ; Use of this source code is governed by a BSD-style license | |
| 5 ; that can be found in the LICENSE file in the root of the source | |
| 6 ; tree. An additional intellectual property rights grant can be found | |
| 7 ; in the file PATENTS. All contributing project authors may | |
| 8 ; be found in the AUTHORS file in the root of the source tree. | |
| 9 ; | |
| 10 | |
| 11 EXPORT |vp9_v_predictor_4x4_neon| | |
| 12 EXPORT |vp9_v_predictor_8x8_neon| | |
| 13 EXPORT |vp9_v_predictor_16x16_neon| | |
| 14 EXPORT |vp9_v_predictor_32x32_neon| | |
| 15 EXPORT |vp9_h_predictor_4x4_neon| | |
| 16 EXPORT |vp9_h_predictor_8x8_neon| | |
| 17 EXPORT |vp9_h_predictor_16x16_neon| | |
| 18 EXPORT |vp9_h_predictor_32x32_neon| | |
| 19 EXPORT |vp9_tm_predictor_4x4_neon| | |
| 20 EXPORT |vp9_tm_predictor_8x8_neon| | |
| 21 EXPORT |vp9_tm_predictor_16x16_neon| | |
| 22 EXPORT |vp9_tm_predictor_32x32_neon| | |
| 23 ARM | |
| 24 REQUIRE8 | |
| 25 PRESERVE8 | |
| 26 | |
| 27 AREA ||.text||, CODE, READONLY, ALIGN=2 | |
| 28 | |
| 29 ;void vp9_v_predictor_4x4_neon(uint8_t *dst, ptrdiff_t y_stride, | |
| 30 ; const uint8_t *above, | |
| 31 ; const uint8_t *left) | |
| 32 ; r0 uint8_t *dst | |
| 33 ; r1 ptrdiff_t y_stride | |
| 34 ; r2 const uint8_t *above | |
| 35 ; r3 const uint8_t *left | |
| 36 | |
| 37 |vp9_v_predictor_4x4_neon| PROC | |
| 38 vld1.32 {d0[0]}, [r2] | |
| 39 vst1.32 {d0[0]}, [r0], r1 | |
| 40 vst1.32 {d0[0]}, [r0], r1 | |
| 41 vst1.32 {d0[0]}, [r0], r1 | |
| 42 vst1.32 {d0[0]}, [r0], r1 | |
| 43 bx lr | |
| 44 ENDP ; |vp9_v_predictor_4x4_neon| | |
| 45 | |
| 46 ;void vp9_v_predictor_8x8_neon(uint8_t *dst, ptrdiff_t y_stride, | |
| 47 ; const uint8_t *above, | |
| 48 ; const uint8_t *left) | |
| 49 ; r0 uint8_t *dst | |
| 50 ; r1 ptrdiff_t y_stride | |
| 51 ; r2 const uint8_t *above | |
| 52 ; r3 const uint8_t *left | |
| 53 | |
| 54 |vp9_v_predictor_8x8_neon| PROC | |
| 55 vld1.8 {d0}, [r2] | |
| 56 vst1.8 {d0}, [r0], r1 | |
| 57 vst1.8 {d0}, [r0], r1 | |
| 58 vst1.8 {d0}, [r0], r1 | |
| 59 vst1.8 {d0}, [r0], r1 | |
| 60 vst1.8 {d0}, [r0], r1 | |
| 61 vst1.8 {d0}, [r0], r1 | |
| 62 vst1.8 {d0}, [r0], r1 | |
| 63 vst1.8 {d0}, [r0], r1 | |
| 64 bx lr | |
| 65 ENDP ; |vp9_v_predictor_8x8_neon| | |
| 66 | |
| 67 ;void vp9_v_predictor_16x16_neon(uint8_t *dst, ptrdiff_t y_stride, | |
| 68 ; const uint8_t *above, | |
| 69 ; const uint8_t *left) | |
| 70 ; r0 uint8_t *dst | |
| 71 ; r1 ptrdiff_t y_stride | |
| 72 ; r2 const uint8_t *above | |
| 73 ; r3 const uint8_t *left | |
| 74 | |
| 75 |vp9_v_predictor_16x16_neon| PROC | |
| 76 vld1.8 {q0}, [r2] | |
| 77 vst1.8 {q0}, [r0], r1 | |
| 78 vst1.8 {q0}, [r0], r1 | |
| 79 vst1.8 {q0}, [r0], r1 | |
| 80 vst1.8 {q0}, [r0], r1 | |
| 81 vst1.8 {q0}, [r0], r1 | |
| 82 vst1.8 {q0}, [r0], r1 | |
| 83 vst1.8 {q0}, [r0], r1 | |
| 84 vst1.8 {q0}, [r0], r1 | |
| 85 vst1.8 {q0}, [r0], r1 | |
| 86 vst1.8 {q0}, [r0], r1 | |
| 87 vst1.8 {q0}, [r0], r1 | |
| 88 vst1.8 {q0}, [r0], r1 | |
| 89 vst1.8 {q0}, [r0], r1 | |
| 90 vst1.8 {q0}, [r0], r1 | |
| 91 vst1.8 {q0}, [r0], r1 | |
| 92 vst1.8 {q0}, [r0], r1 | |
| 93 bx lr | |
| 94 ENDP ; |vp9_v_predictor_16x16_neon| | |
| 95 | |
| 96 ;void vp9_v_predictor_32x32_neon(uint8_t *dst, ptrdiff_t y_stride, | |
| 97 ; const uint8_t *above, | |
| 98 ; const uint8_t *left) | |
| 99 ; r0 uint8_t *dst | |
| 100 ; r1 ptrdiff_t y_stride | |
| 101 ; r2 const uint8_t *above | |
| 102 ; r3 const uint8_t *left | |
| 103 | |
| 104 |vp9_v_predictor_32x32_neon| PROC | |
| 105 vld1.8 {q0, q1}, [r2] | |
| 106 mov r2, #2 | |
| 107 loop_v | |
| 108 vst1.8 {q0, q1}, [r0], r1 | |
| 109 vst1.8 {q0, q1}, [r0], r1 | |
| 110 vst1.8 {q0, q1}, [r0], r1 | |
| 111 vst1.8 {q0, q1}, [r0], r1 | |
| 112 vst1.8 {q0, q1}, [r0], r1 | |
| 113 vst1.8 {q0, q1}, [r0], r1 | |
| 114 vst1.8 {q0, q1}, [r0], r1 | |
| 115 vst1.8 {q0, q1}, [r0], r1 | |
| 116 vst1.8 {q0, q1}, [r0], r1 | |
| 117 vst1.8 {q0, q1}, [r0], r1 | |
| 118 vst1.8 {q0, q1}, [r0], r1 | |
| 119 vst1.8 {q0, q1}, [r0], r1 | |
| 120 vst1.8 {q0, q1}, [r0], r1 | |
| 121 vst1.8 {q0, q1}, [r0], r1 | |
| 122 vst1.8 {q0, q1}, [r0], r1 | |
| 123 vst1.8 {q0, q1}, [r0], r1 | |
| 124 subs r2, r2, #1 | |
| 125 bgt loop_v | |
| 126 bx lr | |
| 127 ENDP ; |vp9_v_predictor_32x32_neon| | |
| 128 | |
| 129 ;void vp9_h_predictor_4x4_neon(uint8_t *dst, ptrdiff_t y_stride, | |
| 130 ; const uint8_t *above, | |
| 131 ; const uint8_t *left) | |
| 132 ; r0 uint8_t *dst | |
| 133 ; r1 ptrdiff_t y_stride | |
| 134 ; r2 const uint8_t *above | |
| 135 ; r3 const uint8_t *left | |
| 136 | |
| 137 |vp9_h_predictor_4x4_neon| PROC | |
| 138 vld1.32 {d1[0]}, [r3] | |
| 139 vdup.8 d0, d1[0] | |
| 140 vst1.32 {d0[0]}, [r0], r1 | |
| 141 vdup.8 d0, d1[1] | |
| 142 vst1.32 {d0[0]}, [r0], r1 | |
| 143 vdup.8 d0, d1[2] | |
| 144 vst1.32 {d0[0]}, [r0], r1 | |
| 145 vdup.8 d0, d1[3] | |
| 146 vst1.32 {d0[0]}, [r0], r1 | |
| 147 bx lr | |
| 148 ENDP ; |vp9_h_predictor_4x4_neon| | |
| 149 | |
| 150 ;void vp9_h_predictor_8x8_neon(uint8_t *dst, ptrdiff_t y_stride, | |
| 151 ; const uint8_t *above, | |
| 152 ; const uint8_t *left) | |
| 153 ; r0 uint8_t *dst | |
| 154 ; r1 ptrdiff_t y_stride | |
| 155 ; r2 const uint8_t *above | |
| 156 ; r3 const uint8_t *left | |
| 157 | |
| 158 |vp9_h_predictor_8x8_neon| PROC | |
| 159 vld1.64 {d1}, [r3] | |
| 160 vdup.8 d0, d1[0] | |
| 161 vst1.64 {d0}, [r0], r1 | |
| 162 vdup.8 d0, d1[1] | |
| 163 vst1.64 {d0}, [r0], r1 | |
| 164 vdup.8 d0, d1[2] | |
| 165 vst1.64 {d0}, [r0], r1 | |
| 166 vdup.8 d0, d1[3] | |
| 167 vst1.64 {d0}, [r0], r1 | |
| 168 vdup.8 d0, d1[4] | |
| 169 vst1.64 {d0}, [r0], r1 | |
| 170 vdup.8 d0, d1[5] | |
| 171 vst1.64 {d0}, [r0], r1 | |
| 172 vdup.8 d0, d1[6] | |
| 173 vst1.64 {d0}, [r0], r1 | |
| 174 vdup.8 d0, d1[7] | |
| 175 vst1.64 {d0}, [r0], r1 | |
| 176 bx lr | |
| 177 ENDP ; |vp9_h_predictor_8x8_neon| | |
| 178 | |
| 179 ;void vp9_h_predictor_16x16_neon(uint8_t *dst, ptrdiff_t y_stride, | |
| 180 ; const uint8_t *above, | |
| 181 ; const uint8_t *left) | |
| 182 ; r0 uint8_t *dst | |
| 183 ; r1 ptrdiff_t y_stride | |
| 184 ; r2 const uint8_t *above | |
| 185 ; r3 const uint8_t *left | |
| 186 | |
| 187 |vp9_h_predictor_16x16_neon| PROC | |
| 188 vld1.8 {q1}, [r3] | |
| 189 vdup.8 q0, d2[0] | |
| 190 vst1.8 {q0}, [r0], r1 | |
| 191 vdup.8 q0, d2[1] | |
| 192 vst1.8 {q0}, [r0], r1 | |
| 193 vdup.8 q0, d2[2] | |
| 194 vst1.8 {q0}, [r0], r1 | |
| 195 vdup.8 q0, d2[3] | |
| 196 vst1.8 {q0}, [r0], r1 | |
| 197 vdup.8 q0, d2[4] | |
| 198 vst1.8 {q0}, [r0], r1 | |
| 199 vdup.8 q0, d2[5] | |
| 200 vst1.8 {q0}, [r0], r1 | |
| 201 vdup.8 q0, d2[6] | |
| 202 vst1.8 {q0}, [r0], r1 | |
| 203 vdup.8 q0, d2[7] | |
| 204 vst1.8 {q0}, [r0], r1 | |
| 205 vdup.8 q0, d3[0] | |
| 206 vst1.8 {q0}, [r0], r1 | |
| 207 vdup.8 q0, d3[1] | |
| 208 vst1.8 {q0}, [r0], r1 | |
| 209 vdup.8 q0, d3[2] | |
| 210 vst1.8 {q0}, [r0], r1 | |
| 211 vdup.8 q0, d3[3] | |
| 212 vst1.8 {q0}, [r0], r1 | |
| 213 vdup.8 q0, d3[4] | |
| 214 vst1.8 {q0}, [r0], r1 | |
| 215 vdup.8 q0, d3[5] | |
| 216 vst1.8 {q0}, [r0], r1 | |
| 217 vdup.8 q0, d3[6] | |
| 218 vst1.8 {q0}, [r0], r1 | |
| 219 vdup.8 q0, d3[7] | |
| 220 vst1.8 {q0}, [r0], r1 | |
| 221 bx lr | |
| 222 ENDP ; |vp9_h_predictor_16x16_neon| | |
| 223 | |
| 224 ;void vp9_h_predictor_32x32_neon(uint8_t *dst, ptrdiff_t y_stride, | |
| 225 ; const uint8_t *above, | |
| 226 ; const uint8_t *left) | |
| 227 ; r0 uint8_t *dst | |
| 228 ; r1 ptrdiff_t y_stride | |
| 229 ; r2 const uint8_t *above | |
| 230 ; r3 const uint8_t *left | |
| 231 | |
| 232 |vp9_h_predictor_32x32_neon| PROC | |
| 233 sub r1, r1, #16 | |
| 234 mov r2, #2 | |
| 235 loop_h | |
| 236 vld1.8 {q1}, [r3]! | |
| 237 vdup.8 q0, d2[0] | |
| 238 vst1.8 {q0}, [r0]! | |
| 239 vst1.8 {q0}, [r0], r1 | |
| 240 vdup.8 q0, d2[1] | |
| 241 vst1.8 {q0}, [r0]! | |
| 242 vst1.8 {q0}, [r0], r1 | |
| 243 vdup.8 q0, d2[2] | |
| 244 vst1.8 {q0}, [r0]! | |
| 245 vst1.8 {q0}, [r0], r1 | |
| 246 vdup.8 q0, d2[3] | |
| 247 vst1.8 {q0}, [r0]! | |
| 248 vst1.8 {q0}, [r0], r1 | |
| 249 vdup.8 q0, d2[4] | |
| 250 vst1.8 {q0}, [r0]! | |
| 251 vst1.8 {q0}, [r0], r1 | |
| 252 vdup.8 q0, d2[5] | |
| 253 vst1.8 {q0}, [r0]! | |
| 254 vst1.8 {q0}, [r0], r1 | |
| 255 vdup.8 q0, d2[6] | |
| 256 vst1.8 {q0}, [r0]! | |
| 257 vst1.8 {q0}, [r0], r1 | |
| 258 vdup.8 q0, d2[7] | |
| 259 vst1.8 {q0}, [r0]! | |
| 260 vst1.8 {q0}, [r0], r1 | |
| 261 vdup.8 q0, d3[0] | |
| 262 vst1.8 {q0}, [r0]! | |
| 263 vst1.8 {q0}, [r0], r1 | |
| 264 vdup.8 q0, d3[1] | |
| 265 vst1.8 {q0}, [r0]! | |
| 266 vst1.8 {q0}, [r0], r1 | |
| 267 vdup.8 q0, d3[2] | |
| 268 vst1.8 {q0}, [r0]! | |
| 269 vst1.8 {q0}, [r0], r1 | |
| 270 vdup.8 q0, d3[3] | |
| 271 vst1.8 {q0}, [r0]! | |
| 272 vst1.8 {q0}, [r0], r1 | |
| 273 vdup.8 q0, d3[4] | |
| 274 vst1.8 {q0}, [r0]! | |
| 275 vst1.8 {q0}, [r0], r1 | |
| 276 vdup.8 q0, d3[5] | |
| 277 vst1.8 {q0}, [r0]! | |
| 278 vst1.8 {q0}, [r0], r1 | |
| 279 vdup.8 q0, d3[6] | |
| 280 vst1.8 {q0}, [r0]! | |
| 281 vst1.8 {q0}, [r0], r1 | |
| 282 vdup.8 q0, d3[7] | |
| 283 vst1.8 {q0}, [r0]! | |
| 284 vst1.8 {q0}, [r0], r1 | |
| 285 subs r2, r2, #1 | |
| 286 bgt loop_h | |
| 287 bx lr | |
| 288 ENDP ; |vp9_h_predictor_32x32_neon| | |
| 289 | |
| 290 ;void vp9_tm_predictor_4x4_neon (uint8_t *dst, ptrdiff_t y_stride, | |
| 291 ; const uint8_t *above, | |
| 292 ; const uint8_t *left) | |
| 293 ; r0 uint8_t *dst | |
| 294 ; r1 ptrdiff_t y_stride | |
| 295 ; r2 const uint8_t *above | |
| 296 ; r3 const uint8_t *left | |
| 297 | |
| 298 |vp9_tm_predictor_4x4_neon| PROC | |
| 299 ; Load ytop_left = above[-1]; | |
| 300 sub r12, r2, #1 | |
| 301 ldrb r12, [r12] | |
| 302 vdup.u8 d0, r12 | |
| 303 | |
| 304 ; Load above 4 pixels | |
| 305 vld1.32 {d2[0]}, [r2] | |
| 306 | |
| 307 ; Compute above - ytop_left | |
| 308 vsubl.u8 q3, d2, d0 | |
| 309 | |
| 310 ; Load left row by row and compute left + (above - ytop_left) | |
| 311 ; 1st row and 2nd row | |
| 312 ldrb r12, [r3], #1 | |
| 313 ldrb r2, [r3], #1 | |
| 314 vdup.u16 q1, r12 | |
| 315 vdup.u16 q2, r2 | |
| 316 vadd.s16 q1, q1, q3 | |
| 317 vadd.s16 q2, q2, q3 | |
| 318 vqmovun.s16 d0, q1 | |
| 319 vqmovun.s16 d1, q2 | |
| 320 vst1.32 {d0[0]}, [r0], r1 | |
| 321 vst1.32 {d1[0]}, [r0], r1 | |
| 322 | |
| 323 ; 3rd row and 4th row | |
| 324 ldrb r12, [r3], #1 | |
| 325 ldrb r2, [r3], #1 | |
| 326 vdup.u16 q1, r12 | |
| 327 vdup.u16 q2, r2 | |
| 328 vadd.s16 q1, q1, q3 | |
| 329 vadd.s16 q2, q2, q3 | |
| 330 vqmovun.s16 d0, q1 | |
| 331 vqmovun.s16 d1, q2 | |
| 332 vst1.32 {d0[0]}, [r0], r1 | |
| 333 vst1.32 {d1[0]}, [r0], r1 | |
| 334 bx lr | |
| 335 ENDP ; |vp9_tm_predictor_4x4_neon| | |
| 336 | |
| 337 ;void vp9_tm_predictor_8x8_neon (uint8_t *dst, ptrdiff_t y_stride, | |
| 338 ; const uint8_t *above, | |
| 339 ; const uint8_t *left) | |
| 340 ; r0 uint8_t *dst | |
| 341 ; r1 ptrdiff_t y_stride | |
| 342 ; r2 const uint8_t *above | |
| 343 ; r3 const uint8_t *left | |
| 344 | |
| 345 |vp9_tm_predictor_8x8_neon| PROC | |
| 346 ; Load ytop_left = above[-1]; | |
| 347 sub r12, r2, #1 | |
| 348 ldrb r12, [r12] | |
| 349 vdup.u8 d0, r12 | |
| 350 | |
| 351 ; preload 8 left | |
| 352 vld1.8 {d30}, [r3] | |
| 353 | |
| 354 ; Load above 8 pixels | |
| 355 vld1.64 {d2}, [r2] | |
| 356 | |
| 357 vmovl.u8 q10, d30 | |
| 358 | |
| 359 ; Compute above - ytop_left | |
| 360 vsubl.u8 q3, d2, d0 | |
| 361 | |
| 362 ; Load left row by row and compute left + (above - ytop_left) | |
| 363 ; 1st row and 2nd row | |
| 364 vdup.16 q0, d20[0] | |
| 365 vdup.16 q1, d20[1] | |
| 366 vadd.s16 q0, q3, q0 | |
| 367 vadd.s16 q1, q3, q1 | |
| 368 | |
| 369 ; 3rd row and 4th row | |
| 370 vdup.16 q8, d20[2] | |
| 371 vdup.16 q9, d20[3] | |
| 372 vadd.s16 q8, q3, q8 | |
| 373 vadd.s16 q9, q3, q9 | |
| 374 | |
| 375 vqmovun.s16 d0, q0 | |
| 376 vqmovun.s16 d1, q1 | |
| 377 vqmovun.s16 d2, q8 | |
| 378 vqmovun.s16 d3, q9 | |
| 379 | |
| 380 vst1.64 {d0}, [r0], r1 | |
| 381 vst1.64 {d1}, [r0], r1 | |
| 382 vst1.64 {d2}, [r0], r1 | |
| 383 vst1.64 {d3}, [r0], r1 | |
| 384 | |
| 385 ; 5th row and 6th row | |
| 386 vdup.16 q0, d21[0] | |
| 387 vdup.16 q1, d21[1] | |
| 388 vadd.s16 q0, q3, q0 | |
| 389 vadd.s16 q1, q3, q1 | |
| 390 | |
| 391 ; 7th row and 8th row | |
| 392 vdup.16 q8, d21[2] | |
| 393 vdup.16 q9, d21[3] | |
| 394 vadd.s16 q8, q3, q8 | |
| 395 vadd.s16 q9, q3, q9 | |
| 396 | |
| 397 vqmovun.s16 d0, q0 | |
| 398 vqmovun.s16 d1, q1 | |
| 399 vqmovun.s16 d2, q8 | |
| 400 vqmovun.s16 d3, q9 | |
| 401 | |
| 402 vst1.64 {d0}, [r0], r1 | |
| 403 vst1.64 {d1}, [r0], r1 | |
| 404 vst1.64 {d2}, [r0], r1 | |
| 405 vst1.64 {d3}, [r0], r1 | |
| 406 | |
| 407 bx lr | |
| 408 ENDP ; |vp9_tm_predictor_8x8_neon| | |
| 409 | |
| 410 ;void vp9_tm_predictor_16x16_neon (uint8_t *dst, ptrdiff_t y_stride, | |
| 411 ; const uint8_t *above, | |
| 412 ; const uint8_t *left) | |
| 413 ; r0 uint8_t *dst | |
| 414 ; r1 ptrdiff_t y_stride | |
| 415 ; r2 const uint8_t *above | |
| 416 ; r3 const uint8_t *left | |
| 417 | |
| 418 |vp9_tm_predictor_16x16_neon| PROC | |
| 419 ; Load ytop_left = above[-1]; | |
| 420 sub r12, r2, #1 | |
| 421 ldrb r12, [r12] | |
| 422 vdup.u8 q0, r12 | |
| 423 | |
| 424 ; Load above 8 pixels | |
| 425 vld1.8 {q1}, [r2] | |
| 426 | |
| 427 ; preload 8 left into r12 | |
| 428 vld1.8 {d18}, [r3]! | |
| 429 | |
| 430 ; Compute above - ytop_left | |
| 431 vsubl.u8 q2, d2, d0 | |
| 432 vsubl.u8 q3, d3, d1 | |
| 433 | |
| 434 vmovl.u8 q10, d18 | |
| 435 | |
| 436 ; Load left row by row and compute left + (above - ytop_left) | |
| 437 ; Process 8 rows in each single loop and loop 2 times to process 16 rows. | |
| 438 mov r2, #2 | |
| 439 | |
| 440 loop_16x16_neon | |
| 441 ; Process two rows. | |
| 442 vdup.16 q0, d20[0] | |
| 443 vdup.16 q8, d20[1] | |
| 444 vadd.s16 q1, q0, q2 | |
| 445 vadd.s16 q0, q0, q3 | |
| 446 vadd.s16 q11, q8, q2 | |
| 447 vadd.s16 q8, q8, q3 | |
| 448 vqmovun.s16 d2, q1 | |
| 449 vqmovun.s16 d3, q0 | |
| 450 vqmovun.s16 d22, q11 | |
| 451 vqmovun.s16 d23, q8 | |
| 452 vdup.16 q0, d20[2] ; proload next 2 rows data | |
| 453 vdup.16 q8, d20[3] | |
| 454 vst1.64 {d2,d3}, [r0], r1 | |
| 455 vst1.64 {d22,d23}, [r0], r1 | |
| 456 | |
| 457 ; Process two rows. | |
| 458 vadd.s16 q1, q0, q2 | |
| 459 vadd.s16 q0, q0, q3 | |
| 460 vadd.s16 q11, q8, q2 | |
| 461 vadd.s16 q8, q8, q3 | |
| 462 vqmovun.s16 d2, q1 | |
| 463 vqmovun.s16 d3, q0 | |
| 464 vqmovun.s16 d22, q11 | |
| 465 vqmovun.s16 d23, q8 | |
| 466 vdup.16 q0, d21[0] ; proload next 2 rows data | |
| 467 vdup.16 q8, d21[1] | |
| 468 vst1.64 {d2,d3}, [r0], r1 | |
| 469 vst1.64 {d22,d23}, [r0], r1 | |
| 470 | |
| 471 vadd.s16 q1, q0, q2 | |
| 472 vadd.s16 q0, q0, q3 | |
| 473 vadd.s16 q11, q8, q2 | |
| 474 vadd.s16 q8, q8, q3 | |
| 475 vqmovun.s16 d2, q1 | |
| 476 vqmovun.s16 d3, q0 | |
| 477 vqmovun.s16 d22, q11 | |
| 478 vqmovun.s16 d23, q8 | |
| 479 vdup.16 q0, d21[2] ; proload next 2 rows data | |
| 480 vdup.16 q8, d21[3] | |
| 481 vst1.64 {d2,d3}, [r0], r1 | |
| 482 vst1.64 {d22,d23}, [r0], r1 | |
| 483 | |
| 484 | |
| 485 vadd.s16 q1, q0, q2 | |
| 486 vadd.s16 q0, q0, q3 | |
| 487 vadd.s16 q11, q8, q2 | |
| 488 vadd.s16 q8, q8, q3 | |
| 489 vqmovun.s16 d2, q1 | |
| 490 vqmovun.s16 d3, q0 | |
| 491 vqmovun.s16 d22, q11 | |
| 492 vqmovun.s16 d23, q8 | |
| 493 vld1.8 {d18}, [r3]! ; preload 8 left into r12 | |
| 494 vmovl.u8 q10, d18 | |
| 495 vst1.64 {d2,d3}, [r0], r1 | |
| 496 vst1.64 {d22,d23}, [r0], r1 | |
| 497 | |
| 498 subs r2, r2, #1 | |
| 499 bgt loop_16x16_neon | |
| 500 | |
| 501 bx lr | |
| 502 ENDP ; |vp9_tm_predictor_16x16_neon| | |
| 503 | |
| 504 ;void vp9_tm_predictor_32x32_neon (uint8_t *dst, ptrdiff_t y_stride, | |
| 505 ; const uint8_t *above, | |
| 506 ; const uint8_t *left) | |
| 507 ; r0 uint8_t *dst | |
| 508 ; r1 ptrdiff_t y_stride | |
| 509 ; r2 const uint8_t *above | |
| 510 ; r3 const uint8_t *left | |
| 511 | |
| 512 |vp9_tm_predictor_32x32_neon| PROC | |
| 513 ; Load ytop_left = above[-1]; | |
| 514 sub r12, r2, #1 | |
| 515 ldrb r12, [r12] | |
| 516 vdup.u8 q0, r12 | |
| 517 | |
| 518 ; Load above 32 pixels | |
| 519 vld1.8 {q1}, [r2]! | |
| 520 vld1.8 {q2}, [r2] | |
| 521 | |
| 522 ; preload 8 left pixels | |
| 523 vld1.8 {d26}, [r3]! | |
| 524 | |
| 525 ; Compute above - ytop_left | |
| 526 vsubl.u8 q8, d2, d0 | |
| 527 vsubl.u8 q9, d3, d1 | |
| 528 vsubl.u8 q10, d4, d0 | |
| 529 vsubl.u8 q11, d5, d1 | |
| 530 | |
| 531 vmovl.u8 q3, d26 | |
| 532 | |
| 533 ; Load left row by row and compute left + (above - ytop_left) | |
| 534 ; Process 8 rows in each single loop and loop 4 times to process 32 rows. | |
| 535 mov r2, #4 | |
| 536 | |
| 537 loop_32x32_neon | |
| 538 ; Process two rows. | |
| 539 vdup.16 q0, d6[0] | |
| 540 vdup.16 q2, d6[1] | |
| 541 vadd.s16 q12, q0, q8 | |
| 542 vadd.s16 q13, q0, q9 | |
| 543 vadd.s16 q14, q0, q10 | |
| 544 vadd.s16 q15, q0, q11 | |
| 545 vqmovun.s16 d0, q12 | |
| 546 vqmovun.s16 d1, q13 | |
| 547 vadd.s16 q12, q2, q8 | |
| 548 vadd.s16 q13, q2, q9 | |
| 549 vqmovun.s16 d2, q14 | |
| 550 vqmovun.s16 d3, q15 | |
| 551 vadd.s16 q14, q2, q10 | |
| 552 vadd.s16 q15, q2, q11 | |
| 553 vst1.64 {d0-d3}, [r0], r1 | |
| 554 vqmovun.s16 d24, q12 | |
| 555 vqmovun.s16 d25, q13 | |
| 556 vqmovun.s16 d26, q14 | |
| 557 vqmovun.s16 d27, q15 | |
| 558 vdup.16 q1, d6[2] | |
| 559 vdup.16 q2, d6[3] | |
| 560 vst1.64 {d24-d27}, [r0], r1 | |
| 561 | |
| 562 ; Process two rows. | |
| 563 vadd.s16 q12, q1, q8 | |
| 564 vadd.s16 q13, q1, q9 | |
| 565 vadd.s16 q14, q1, q10 | |
| 566 vadd.s16 q15, q1, q11 | |
| 567 vqmovun.s16 d0, q12 | |
| 568 vqmovun.s16 d1, q13 | |
| 569 vadd.s16 q12, q2, q8 | |
| 570 vadd.s16 q13, q2, q9 | |
| 571 vqmovun.s16 d2, q14 | |
| 572 vqmovun.s16 d3, q15 | |
| 573 vadd.s16 q14, q2, q10 | |
| 574 vadd.s16 q15, q2, q11 | |
| 575 vst1.64 {d0-d3}, [r0], r1 | |
| 576 vqmovun.s16 d24, q12 | |
| 577 vqmovun.s16 d25, q13 | |
| 578 vqmovun.s16 d26, q14 | |
| 579 vqmovun.s16 d27, q15 | |
| 580 vdup.16 q0, d7[0] | |
| 581 vdup.16 q2, d7[1] | |
| 582 vst1.64 {d24-d27}, [r0], r1 | |
| 583 | |
| 584 ; Process two rows. | |
| 585 vadd.s16 q12, q0, q8 | |
| 586 vadd.s16 q13, q0, q9 | |
| 587 vadd.s16 q14, q0, q10 | |
| 588 vadd.s16 q15, q0, q11 | |
| 589 vqmovun.s16 d0, q12 | |
| 590 vqmovun.s16 d1, q13 | |
| 591 vadd.s16 q12, q2, q8 | |
| 592 vadd.s16 q13, q2, q9 | |
| 593 vqmovun.s16 d2, q14 | |
| 594 vqmovun.s16 d3, q15 | |
| 595 vadd.s16 q14, q2, q10 | |
| 596 vadd.s16 q15, q2, q11 | |
| 597 vst1.64 {d0-d3}, [r0], r1 | |
| 598 vqmovun.s16 d24, q12 | |
| 599 vqmovun.s16 d25, q13 | |
| 600 vqmovun.s16 d26, q14 | |
| 601 vqmovun.s16 d27, q15 | |
| 602 vdup.16 q0, d7[2] | |
| 603 vdup.16 q2, d7[3] | |
| 604 vst1.64 {d24-d27}, [r0], r1 | |
| 605 | |
| 606 ; Process two rows. | |
| 607 vadd.s16 q12, q0, q8 | |
| 608 vadd.s16 q13, q0, q9 | |
| 609 vadd.s16 q14, q0, q10 | |
| 610 vadd.s16 q15, q0, q11 | |
| 611 vqmovun.s16 d0, q12 | |
| 612 vqmovun.s16 d1, q13 | |
| 613 vadd.s16 q12, q2, q8 | |
| 614 vadd.s16 q13, q2, q9 | |
| 615 vqmovun.s16 d2, q14 | |
| 616 vqmovun.s16 d3, q15 | |
| 617 vadd.s16 q14, q2, q10 | |
| 618 vadd.s16 q15, q2, q11 | |
| 619 vst1.64 {d0-d3}, [r0], r1 | |
| 620 vqmovun.s16 d24, q12 | |
| 621 vqmovun.s16 d25, q13 | |
| 622 vld1.8 {d0}, [r3]! ; preload 8 left pixels | |
| 623 vqmovun.s16 d26, q14 | |
| 624 vqmovun.s16 d27, q15 | |
| 625 vmovl.u8 q3, d0 | |
| 626 vst1.64 {d24-d27}, [r0], r1 | |
| 627 | |
| 628 subs r2, r2, #1 | |
| 629 bgt loop_32x32_neon | |
| 630 | |
| 631 bx lr | |
| 632 ENDP ; |vp9_tm_predictor_32x32_neon| | |
| 633 | |
| 634 END | |
| OLD | NEW |
|---|
Chromium Code Reviews
Issue 812033011:
libvpx: Pull from upstream (Closed)
Base URL: svn://svn.chromium.org/chrome/trunk/deps/third_party/libvpx/