deblock-a.S 28.7 KB
Newer Older
1 2 3
/*****************************************************************************
 * deblock.S: aarch64 deblocking
 *****************************************************************************
Henrik Gramner's avatar
Henrik Gramner committed
4
 * Copyright (C) 2009-2018 x264 project
5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33
 *
 * Authors: Mans Rullgard <mans@mansr.com>
 *          Janne Grunau <janne-x264@jannau.net>
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02111, USA.
 *
 * This program is also available under a commercial proprietary license.
 * For more information, contact us at licensing@x264.com.
 *****************************************************************************/

#include "asm.S"

.macro h264_loop_filter_start
    cmp             w2,  #0
    ldr             w6,  [x4]
    ccmp            w3,  #0, #0, ne
    mov             v24.s[0], w6
34
    and             w8,  w6,  w6,  lsl #16
35
    b.eq            1f
36
    ands            w8,  w8,  w8,  lsl #8
37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110
    b.ge            2f
1:
    ret
2:
.endm

.macro h264_loop_filter_luma
    dup             v22.16b, w2                     // alpha
    uxtl            v24.8h,  v24.8b
    uabd            v21.16b, v16.16b, v0.16b        // abs(p0 - q0)
    uxtl            v24.4s,  v24.4h
    uabd            v28.16b, v18.16b, v16.16b       // abs(p1 - p0)
    sli             v24.8h,  v24.8h,  #8
    uabd            v30.16b, v2.16b,  v0.16b        // abs(q1 - q0)
    sli             v24.4s,  v24.4s,  #16
    cmhi            v21.16b, v22.16b, v21.16b       // < alpha
    dup             v22.16b, w3                     // beta
    cmlt            v23.16b, v24.16b, #0
    cmhi            v28.16b, v22.16b, v28.16b       // < beta
    cmhi            v30.16b, v22.16b, v30.16b       // < beta
    bic             v21.16b, v21.16b, v23.16b
    uabd            v17.16b, v20.16b, v16.16b       // abs(p2 - p0)
    and             v21.16b, v21.16b, v28.16b
    uabd            v19.16b,  v4.16b,  v0.16b       // abs(q2 - q0)
    cmhi            v17.16b, v22.16b, v17.16b       // < beta
    and             v21.16b, v21.16b, v30.16b
    cmhi            v19.16b, v22.16b, v19.16b       // < beta
    and             v17.16b, v17.16b, v21.16b
    and             v19.16b, v19.16b, v21.16b
    and             v24.16b, v24.16b, v21.16b
    urhadd          v28.16b, v16.16b,  v0.16b
    sub             v21.16b, v24.16b, v17.16b
    uqadd           v23.16b, v18.16b, v24.16b
    uhadd           v20.16b, v20.16b, v28.16b
    sub             v21.16b, v21.16b, v19.16b
    uhadd           v28.16b,  v4.16b, v28.16b
    umin            v23.16b, v23.16b, v20.16b
    uqsub           v22.16b, v18.16b, v24.16b
    uqadd           v4.16b,   v2.16b, v24.16b
    umax            v23.16b, v23.16b, v22.16b
    uqsub           v22.16b,  v2.16b, v24.16b
    umin            v28.16b,  v4.16b, v28.16b
    uxtl            v4.8h,    v0.8b
    umax            v28.16b, v28.16b, v22.16b
    uxtl2           v20.8h,   v0.16b
    usubw           v4.8h,    v4.8h,  v16.8b
    usubw2          v20.8h,  v20.8h,  v16.16b
    shl             v4.8h,    v4.8h,  #2
    shl             v20.8h,  v20.8h,  #2
    uaddw           v4.8h,    v4.8h,  v18.8b
    uaddw2          v20.8h,  v20.8h,  v18.16b
    usubw           v4.8h,    v4.8h,   v2.8b
    usubw2          v20.8h,  v20.8h,   v2.16b
    rshrn           v4.8b,    v4.8h,  #3
    rshrn2          v4.16b,  v20.8h,  #3
    bsl             v17.16b, v23.16b, v18.16b
    bsl             v19.16b, v28.16b,  v2.16b
    neg             v23.16b, v21.16b
    uxtl            v28.8h,  v16.8b
    smin            v4.16b,   v4.16b, v21.16b
    uxtl2           v21.8h,  v16.16b
    smax            v4.16b,   v4.16b, v23.16b
    uxtl            v22.8h,   v0.8b
    uxtl2           v24.8h,   v0.16b
    saddw           v28.8h,  v28.8h,  v4.8b
    saddw2          v21.8h,  v21.8h,  v4.16b
    ssubw           v22.8h,  v22.8h,  v4.8b
    ssubw2          v24.8h,  v24.8h,  v4.16b
    sqxtun          v16.8b,  v28.8h
    sqxtun2         v16.16b, v21.8h
    sqxtun          v0.8b,   v22.8h
    sqxtun2         v0.16b,  v24.8h
.endm

111
function deblock_v_luma_neon, export=1
112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133
    h264_loop_filter_start

    ld1             {v0.16b},  [x0], x1
    ld1             {v2.16b},  [x0], x1
    ld1             {v4.16b},  [x0], x1
    sub             x0,  x0,  x1, lsl #2
    sub             x0,  x0,  x1, lsl #1
    ld1             {v20.16b},  [x0], x1
    ld1             {v18.16b},  [x0], x1
    ld1             {v16.16b},  [x0], x1

    h264_loop_filter_luma

    sub             x0,  x0,  x1, lsl #1
    st1             {v17.16b}, [x0], x1
    st1             {v16.16b}, [x0], x1
    st1             {v0.16b},  [x0], x1
    st1             {v19.16b}, [x0]

    ret
endfunc

134
function deblock_h_luma_neon, export=1
135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182
    h264_loop_filter_start

    sub             x0,  x0,  #4
    ld1             {v6.8b},  [x0], x1
    ld1             {v20.8b}, [x0], x1
    ld1             {v18.8b}, [x0], x1
    ld1             {v16.8b}, [x0], x1
    ld1             {v0.8b},  [x0], x1
    ld1             {v2.8b},  [x0], x1
    ld1             {v4.8b},  [x0], x1
    ld1             {v26.8b}, [x0], x1
    ld1             {v6.d}[1],  [x0], x1
    ld1             {v20.d}[1], [x0], x1
    ld1             {v18.d}[1], [x0], x1
    ld1             {v16.d}[1], [x0], x1
    ld1             {v0.d}[1],  [x0], x1
    ld1             {v2.d}[1],  [x0], x1
    ld1             {v4.d}[1],  [x0], x1
    ld1             {v26.d}[1], [x0], x1

    transpose_8x16.b v6, v20, v18, v16, v0, v2, v4, v26, v21, v23

    h264_loop_filter_luma

    transpose_4x16.b v17, v16, v0, v19, v21, v23, v25, v27

    sub             x0,  x0,  x1, lsl #4
    add             x0,  x0,  #2
    st1             {v17.s}[0],  [x0], x1
    st1             {v16.s}[0], [x0], x1
    st1             {v0.s}[0],  [x0], x1
    st1             {v19.s}[0], [x0], x1
    st1             {v17.s}[1],  [x0], x1
    st1             {v16.s}[1], [x0], x1
    st1             {v0.s}[1],  [x0], x1
    st1             {v19.s}[1], [x0], x1
    st1             {v17.s}[2],  [x0], x1
    st1             {v16.s}[2], [x0], x1
    st1             {v0.s}[2],  [x0], x1
    st1             {v19.s}[2], [x0], x1
    st1             {v17.s}[3],  [x0], x1
    st1             {v16.s}[3], [x0], x1
    st1             {v0.s}[3],  [x0], x1
    st1             {v19.s}[3], [x0], x1

    ret
endfunc

183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304
.macro h264_loop_filter_start_intra
    orr             w4,  w2,  w3
    cmp             w4,  #0
    b.ne            1f
    ret
1:
    dup             v30.16b, w2                // alpha
    dup             v31.16b, w3                // beta
.endm

.macro h264_loop_filter_luma_intra
    uabd            v16.16b, v7.16b,  v0.16b        // abs(p0 - q0)
    uabd            v17.16b, v6.16b,  v7.16b        // abs(p1 - p0)
    uabd            v18.16b, v1.16b,  v0.16b        // abs(q1 - q0)
    cmhi            v19.16b, v30.16b, v16.16b       // < alpha
    cmhi            v17.16b, v31.16b, v17.16b       // < beta
    cmhi            v18.16b, v31.16b, v18.16b       // < beta

    movi            v29.16b, #2
    ushr            v30.16b, v30.16b, #2            // alpha >> 2
    add             v30.16b, v30.16b, v29.16b       // (alpha >> 2) + 2
    cmhi            v16.16b, v30.16b, v16.16b       // < (alpha >> 2) + 2

    and             v19.16b, v19.16b, v17.16b
    and             v19.16b, v19.16b, v18.16b
    shrn            v20.8b,  v19.8h,  #4
    mov             x4, v20.d[0]
    cbz             x4, 9f

    ushll           v20.8h,  v6.8b,   #1
    ushll           v22.8h,  v1.8b,   #1
    ushll2          v21.8h,  v6.16b,  #1
    ushll2          v23.8h,  v1.16b,  #1
    uaddw           v20.8h,  v20.8h,  v7.8b
    uaddw           v22.8h,  v22.8h,  v0.8b
    uaddw2          v21.8h,  v21.8h,  v7.16b
    uaddw2          v23.8h,  v23.8h,  v0.16b
    uaddw           v20.8h,  v20.8h,  v1.8b
    uaddw           v22.8h,  v22.8h,  v6.8b
    uaddw2          v21.8h,  v21.8h,  v1.16b
    uaddw2          v23.8h,  v23.8h,  v6.16b

    rshrn           v24.8b,  v20.8h,  #2 // p0'_1
    rshrn           v25.8b,  v22.8h,  #2 // q0'_1
    rshrn2          v24.16b, v21.8h,  #2 // p0'_1
    rshrn2          v25.16b, v23.8h,  #2 // q0'_1

    uabd            v17.16b, v5.16b,  v7.16b        // abs(p2 - p0)
    uabd            v18.16b, v2.16b,  v0.16b        // abs(q2 - q0)
    cmhi            v17.16b, v31.16b, v17.16b       // < beta
    cmhi            v18.16b, v31.16b, v18.16b       // < beta

    and             v17.16b, v16.16b, v17.16b  // if_2 && if_3
    and             v18.16b, v16.16b, v18.16b  // if_2 && if_4

    not             v30.16b, v17.16b
    not             v31.16b, v18.16b

    and             v30.16b, v30.16b, v19.16b  // if_1 && !(if_2 && if_3)
    and             v31.16b, v31.16b, v19.16b  // if_1 && !(if_2 && if_4)

    and             v17.16b, v19.16b, v17.16b  // if_1 && if_2 && if_3
    and             v18.16b, v19.16b, v18.16b  // if_1 && if_2 && if_4

    //calc            p, v7, v6, v5, v4, v17, v7, v6, v5, v4
    uaddl           v26.8h,  v5.8b,   v7.8b
    uaddl2          v27.8h,  v5.16b,  v7.16b
    uaddw           v26.8h,  v26.8h,  v0.8b
    uaddw2          v27.8h,  v27.8h,  v0.16b
    add             v20.8h,  v20.8h,  v26.8h
    add             v21.8h,  v21.8h,  v27.8h
    uaddw           v20.8h,  v20.8h,  v0.8b
    uaddw2          v21.8h,  v21.8h,  v0.16b
    rshrn           v20.8b,  v20.8h,  #3 // p0'_2
    rshrn2          v20.16b, v21.8h,  #3 // p0'_2
    uaddw           v26.8h,  v26.8h,  v6.8b
    uaddw2          v27.8h,  v27.8h,  v6.16b
    rshrn           v21.8b,  v26.8h,  #2 // p1'_2
    rshrn2          v21.16b, v27.8h,  #2 // p1'_2
    uaddl           v28.8h,  v4.8b,   v5.8b
    uaddl2          v29.8h,  v4.16b,  v5.16b
    shl             v28.8h,  v28.8h,  #1
    shl             v29.8h,  v29.8h,  #1
    add             v28.8h,  v28.8h,  v26.8h
    add             v29.8h,  v29.8h,  v27.8h
    rshrn           v19.8b,  v28.8h,  #3 // p2'_2
    rshrn2          v19.16b, v29.8h,  #3 // p2'_2

    //calc            q, v0, v1, v2, v3, v18, v0, v1, v2, v3
    uaddl           v26.8h,  v2.8b,   v0.8b
    uaddl2          v27.8h,  v2.16b,  v0.16b
    uaddw           v26.8h,  v26.8h,  v7.8b
    uaddw2          v27.8h,  v27.8h,  v7.16b
    add             v22.8h,  v22.8h,  v26.8h
    add             v23.8h,  v23.8h,  v27.8h
    uaddw           v22.8h,  v22.8h,  v7.8b
    uaddw2          v23.8h,  v23.8h,  v7.16b
    rshrn           v22.8b,  v22.8h,  #3 // q0'_2
    rshrn2          v22.16b, v23.8h,  #3 // q0'_2
    uaddw           v26.8h,  v26.8h,  v1.8b
    uaddw2          v27.8h,  v27.8h,  v1.16b
    rshrn           v23.8b,  v26.8h,  #2 // q1'_2
    rshrn2          v23.16b, v27.8h,  #2 // q1'_2
    uaddl           v28.8h,  v2.8b,   v3.8b
    uaddl2          v29.8h,  v2.16b,  v3.16b
    shl             v28.8h,  v28.8h,  #1
    shl             v29.8h,  v29.8h,  #1
    add             v28.8h,  v28.8h,  v26.8h
    add             v29.8h,  v29.8h,  v27.8h
    rshrn           v26.8b,  v28.8h,  #3 // q2'_2
    rshrn2          v26.16b, v29.8h,  #3 // q2'_2

    bit             v7.16b,  v24.16b, v30.16b  // p0'_1
    bit             v0.16b,  v25.16b, v31.16b  // q0'_1
    bit             v7.16b, v20.16b,  v17.16b  // p0'_2
    bit             v6.16b, v21.16b,  v17.16b  // p1'_2
    bit             v5.16b, v19.16b,  v17.16b  // p2'_2
    bit             v0.16b, v22.16b,  v18.16b  // q0'_2
    bit             v1.16b, v23.16b,  v18.16b  // q1'_2
    bit             v2.16b, v26.16b,  v18.16b  // q2'_2
.endm

305
function deblock_v_luma_intra_neon, export=1
306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330
    h264_loop_filter_start_intra

    ld1             {v0.16b},  [x0], x1 // q0
    ld1             {v1.16b},  [x0], x1 // q1
    ld1             {v2.16b},  [x0], x1 // q2
    ld1             {v3.16b},  [x0], x1 // q3
    sub             x0,  x0,  x1, lsl #3
    ld1             {v4.16b},  [x0], x1 // p3
    ld1             {v5.16b},  [x0], x1 // p2
    ld1             {v6.16b},  [x0], x1 // p1
    ld1             {v7.16b},  [x0]     // p0

    h264_loop_filter_luma_intra

    sub             x0,  x0,  x1, lsl #1
    st1             {v5.16b}, [x0], x1  // p2
    st1             {v6.16b}, [x0], x1  // p1
    st1             {v7.16b}, [x0], x1  // p0
    st1             {v0.16b}, [x0], x1  // q0
    st1             {v1.16b}, [x0], x1  // q1
    st1             {v2.16b}, [x0]      // q2
9:
    ret
endfunc

331
function deblock_h_luma_intra_neon, export=1
332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378
    h264_loop_filter_start_intra

    sub             x0,  x0,  #4
    ld1             {v4.8b},  [x0], x1
    ld1             {v5.8b},  [x0], x1
    ld1             {v6.8b},  [x0], x1
    ld1             {v7.8b},  [x0], x1
    ld1             {v0.8b},  [x0], x1
    ld1             {v1.8b},  [x0], x1
    ld1             {v2.8b},  [x0], x1
    ld1             {v3.8b},  [x0], x1
    ld1             {v4.d}[1],  [x0], x1
    ld1             {v5.d}[1],  [x0], x1
    ld1             {v6.d}[1],  [x0], x1
    ld1             {v7.d}[1],  [x0], x1
    ld1             {v0.d}[1],  [x0], x1
    ld1             {v1.d}[1],  [x0], x1
    ld1             {v2.d}[1],  [x0], x1
    ld1             {v3.d}[1],  [x0], x1

    transpose_8x16.b v4, v5, v6, v7, v0, v1, v2, v3, v21, v23

    h264_loop_filter_luma_intra

    transpose_8x16.b v4, v5, v6, v7, v0, v1, v2, v3, v21, v23

    sub             x0,  x0,  x1, lsl #4
    st1             {v4.8b},  [x0], x1
    st1             {v5.8b},  [x0], x1
    st1             {v6.8b},  [x0], x1
    st1             {v7.8b},  [x0], x1
    st1             {v0.8b},  [x0], x1
    st1             {v1.8b},  [x0], x1
    st1             {v2.8b},  [x0], x1
    st1             {v3.8b},  [x0], x1
    st1             {v4.d}[1],  [x0], x1
    st1             {v5.d}[1],  [x0], x1
    st1             {v6.d}[1],  [x0], x1
    st1             {v7.d}[1],  [x0], x1
    st1             {v0.d}[1],  [x0], x1
    st1             {v1.d}[1],  [x0], x1
    st1             {v2.d}[1],  [x0], x1
    st1             {v3.d}[1],  [x0], x1
9:
    ret
endfunc

379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423
.macro h264_loop_filter_chroma
    dup             v22.16b, w2              // alpha
    uxtl            v24.8h,  v24.8b
    uabd            v26.16b, v16.16b, v0.16b   // abs(p0 - q0)
    uxtl            v4.8h,   v0.8b
    uxtl2           v5.8h,   v0.16b
    uabd            v28.16b, v18.16b, v16.16b  // abs(p1 - p0)
    usubw           v4.8h,   v4.8h,   v16.8b
    usubw2          v5.8h,   v5.8h,   v16.16b
    sli             v24.8h,  v24.8h,  #8
    shl             v4.8h,   v4.8h,   #2
    shl             v5.8h,   v5.8h,   #2
    uabd            v30.16b, v2.16b,  v0.16b   // abs(q1 - q0)
    uxtl            v24.4s,  v24.4h
    uaddw           v4.8h,   v4.8h,   v18.8b
    uaddw2          v5.8h,   v5.8h,   v18.16b
    cmhi            v26.16b, v22.16b, v26.16b  // < alpha
    usubw           v4.8h,   v4.8h,   v2.8b
    usubw2          v5.8h,   v5.8h,   v2.16b
    sli             v24.4s,  v24.4s,  #16
    dup             v22.16b, w3              // beta
    rshrn           v4.8b,   v4.8h,   #3
    rshrn2          v4.16b,  v5.8h,   #3
    cmhi            v28.16b, v22.16b, v28.16b  // < beta
    cmhi            v30.16b, v22.16b, v30.16b  // < beta
    smin            v4.16b,  v4.16b,  v24.16b
    neg             v25.16b, v24.16b
    and             v26.16b, v26.16b, v28.16b
    smax            v4.16b,  v4.16b,  v25.16b
    and             v26.16b, v26.16b, v30.16b
    uxtl            v22.8h,  v0.8b
    uxtl2           v23.8h,  v0.16b
    and             v4.16b,  v4.16b,  v26.16b
    uxtl            v28.8h,  v16.8b
    uxtl2           v29.8h,  v16.16b
    saddw           v28.8h,  v28.8h,  v4.8b
    saddw2          v29.8h,  v29.8h,  v4.16b
    ssubw           v22.8h,  v22.8h,  v4.8b
    ssubw2          v23.8h,  v23.8h,  v4.16b
    sqxtun          v16.8b,  v28.8h
    sqxtun          v0.8b,   v22.8h
    sqxtun2         v16.16b, v29.8h
    sqxtun2         v0.16b,  v23.8h
.endm

424
function deblock_v_chroma_neon, export=1
425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441
    h264_loop_filter_start

    sub             x0,  x0,  x1, lsl #1
    ld1             {v18.16b}, [x0], x1
    ld1             {v16.16b}, [x0], x1
    ld1             {v0.16b},  [x0], x1
    ld1             {v2.16b},  [x0]

    h264_loop_filter_chroma

    sub             x0,  x0,  x1, lsl #1
    st1             {v16.16b}, [x0], x1
    st1             {v0.16b},  [x0], x1

    ret
endfunc

442
function deblock_h_chroma_neon, export=1
443 444 445
    h264_loop_filter_start

    sub             x0,  x0,  #4
446
deblock_h_chroma:
447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474
    ld1             {v18.d}[0], [x0], x1
    ld1             {v16.d}[0], [x0], x1
    ld1             {v0.d}[0],  [x0], x1
    ld1             {v2.d}[0],  [x0], x1
    ld1             {v18.d}[1], [x0], x1
    ld1             {v16.d}[1], [x0], x1
    ld1             {v0.d}[1],  [x0], x1
    ld1             {v2.d}[1],  [x0], x1

    transpose4x8.h  v18, v16, v0, v2, v28, v29, v30, v31

    h264_loop_filter_chroma

    transpose4x8.h  v18, v16, v0, v2, v28, v29, v30, v31

    sub             x0,  x0,  x1, lsl #3
    st1             {v18.d}[0], [x0], x1
    st1             {v16.d}[0], [x0], x1
    st1             {v0.d}[0],  [x0], x1
    st1             {v2.d}[0],  [x0], x1
    st1             {v18.d}[1], [x0], x1
    st1             {v16.d}[1], [x0], x1
    st1             {v0.d}[1],  [x0], x1
    st1             {v2.d}[1],  [x0], x1

    ret
endfunc

475
function deblock_h_chroma_422_neon, export=1
476
    add             x5,  x0,  x1
477
    sub             x0,  x0,  #4
478
    add             x1,  x1,  x1
479
    h264_loop_filter_start
480
    mov             x7,  x30
481
    bl              deblock_h_chroma
482 483 484 485 486 487
    mov             x30, x7
    sub             x0,  x5,  #4
    mov             v24.s[0], w6
    b               deblock_h_chroma
endfunc

488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518
.macro h264_loop_filter_chroma8
    dup             v22.8b,  w2                 // alpha
    uxtl            v24.8h,  v24.8b
    uabd            v26.8b,  v16.8b,  v17.8b    // abs(p0 - q0)
    uxtl            v4.8h,   v17.8b
    uabd            v28.8b,  v18.8b,  v16.8b    // abs(p1 - p0)
    usubw           v4.8h,   v4.8h,   v16.8b
    sli             v24.8h,  v24.8h,  #8
    shl             v4.8h,   v4.8h,   #2
    uabd            v30.8b,  v19.8b,  v17.8b    // abs(q1 - q0)
    uaddw           v4.8h,   v4.8h,   v18.8b
    cmhi            v26.8b,  v22.8b,  v26.8b    // < alpha
    usubw           v4.8h,   v4.8h,   v19.8b
    dup             v22.8b,  w3                 // beta
    rshrn           v4.8b,   v4.8h,   #3
    cmhi            v28.8b,  v22.8b,  v28.8b    // < beta
    cmhi            v30.8b,  v22.8b,  v30.8b    // < beta
    smin            v4.8b,   v4.8b,   v24.8b
    neg             v25.8b,  v24.8b
    and             v26.8b,  v26.8b,  v28.8b
    smax            v4.8b,   v4.8b,   v25.8b
    and             v26.8b,  v26.8b,  v30.8b
    uxtl            v22.8h,  v17.8b
    and             v4.8b,   v4.8b,   v26.8b
    uxtl            v28.8h,  v16.8b
    saddw           v28.8h,  v28.8h,  v4.8b
    ssubw           v22.8h,  v22.8h,  v4.8b
    sqxtun          v16.8b,  v28.8h
    sqxtun          v17.8b,  v22.8h
.endm

519
function deblock_h_chroma_mbaff_neon, export=1
520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541
    h264_loop_filter_start

    sub             x4,  x0,  #4
    sub             x0,  x0,  #2

    ld1             {v18.8b}, [x4], x1
    ld1             {v16.8b}, [x4], x1
    ld1             {v17.8b},  [x4], x1
    ld1             {v19.8b},  [x4]

    transpose4x4.h  v18, v16, v17, v19, v28, v29, v30, v31

    h264_loop_filter_chroma8

    st2             {v16.h,v17.h}[0], [x0], x1
    st2             {v16.h,v17.h}[1], [x0], x1
    st2             {v16.h,v17.h}[2], [x0], x1
    st2             {v16.h,v17.h}[3], [x0]

    ret
endfunc

542
.macro h264_loop_filter_chroma_intra width=16
543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577
    uabd            v26.16b, v16.16b, v17.16b  // abs(p0 - q0)
    uabd            v27.16b, v18.16b, v16.16b  // abs(p1 - p0)
    uabd            v28.16b, v19.16b, v17.16b  // abs(q1 - q0)
    cmhi            v26.16b, v30.16b, v26.16b  // < alpha
    cmhi            v27.16b, v31.16b, v27.16b  // < beta
    cmhi            v28.16b, v31.16b, v28.16b  // < beta
    and             v26.16b, v26.16b, v27.16b
    and             v26.16b, v26.16b, v28.16b

    ushll           v4.8h,   v18.8b,  #1
    ushll           v6.8h,   v19.8b,  #1
.ifc \width, 16
    ushll2          v5.8h,   v18.16b, #1
    ushll2          v7.8h,   v19.16b, #1
    uaddl2          v21.8h,  v16.16b, v19.16b
    uaddl2          v23.8h,  v17.16b, v18.16b
.endif
    uaddl           v20.8h,  v16.8b,  v19.8b
    uaddl           v22.8h,  v17.8b,  v18.8b
    add             v20.8h,  v20.8h,  v4.8h     // mlal?
    add             v22.8h,  v22.8h,  v6.8h
.ifc \width, 16
    add             v21.8h,  v21.8h,  v5.8h
    add             v23.8h,  v23.8h,  v7.8h
.endif
    uqrshrn         v24.8b,  v20.8h,  #2
    uqrshrn         v25.8b,  v22.8h,  #2
.ifc \width, 16
    uqrshrn2        v24.16b, v21.8h,  #2
    uqrshrn2        v25.16b, v23.8h,  #2
.endif
    bit             v16.16b, v24.16b, v26.16b
    bit             v17.16b, v25.16b, v26.16b
.endm

578
function deblock_v_chroma_intra_neon, export=1
579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595
    h264_loop_filter_start_intra

    sub             x0,  x0,  x1, lsl #1
    ld1             {v18.16b}, [x0], x1
    ld1             {v16.16b}, [x0], x1
    ld1             {v17.16b}, [x0], x1
    ld1             {v19.16b}, [x0]

    h264_loop_filter_chroma_intra

    sub             x0,  x0,  x1, lsl #1
    st1             {v16.16b}, [x0], x1
    st1             {v17.16b}, [x0], x1

    ret
endfunc

596
function deblock_h_chroma_intra_mbaff_neon, export=1
597 598 599 600 601 602 603 604 605 606 607
    h264_loop_filter_start_intra

    sub             x4,  x0,  #4
    sub             x0,  x0,  #2
    ld1             {v18.8b}, [x4], x1
    ld1             {v16.8b}, [x4], x1
    ld1             {v17.8b}, [x4], x1
    ld1             {v19.8b}, [x4], x1

    transpose4x4.h  v18, v16, v17, v19, v26, v27, v28, v29

608
    h264_loop_filter_chroma_intra width=8
609 610 611 612 613 614 615 616 617

    st2             {v16.h,v17.h}[0], [x0], x1
    st2             {v16.h,v17.h}[1], [x0], x1
    st2             {v16.h,v17.h}[2], [x0], x1
    st2             {v16.h,v17.h}[3], [x0], x1

    ret
endfunc

618
function deblock_h_chroma_intra_neon, export=1
619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647
    h264_loop_filter_start_intra

    sub             x4,  x0,  #4
    sub             x0,  x0,  #2
    ld1             {v18.d}[0], [x4], x1
    ld1             {v16.d}[0], [x4], x1
    ld1             {v17.d}[0], [x4], x1
    ld1             {v19.d}[0], [x4], x1
    ld1             {v18.d}[1], [x4], x1
    ld1             {v16.d}[1], [x4], x1
    ld1             {v17.d}[1], [x4], x1
    ld1             {v19.d}[1], [x4], x1

    transpose4x8.h  v18, v16, v17, v19, v26, v27, v28, v29

    h264_loop_filter_chroma_intra

    st2             {v16.h,v17.h}[0], [x0], x1
    st2             {v16.h,v17.h}[1], [x0], x1
    st2             {v16.h,v17.h}[2], [x0], x1
    st2             {v16.h,v17.h}[3], [x0], x1
    st2             {v16.h,v17.h}[4], [x0], x1
    st2             {v16.h,v17.h}[5], [x0], x1
    st2             {v16.h,v17.h}[6], [x0], x1
    st2             {v16.h,v17.h}[7], [x0], x1

    ret
endfunc

648
function deblock_h_chroma_422_intra_neon, export=1
649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698
    h264_loop_filter_start_intra

    sub             x4,  x0,  #4
    sub             x0,  x0,  #2
    ld1             {v18.d}[0], [x4], x1
    ld1             {v16.d}[0], [x4], x1
    ld1             {v17.d}[0], [x4], x1
    ld1             {v19.d}[0], [x4], x1
    ld1             {v18.d}[1], [x4], x1
    ld1             {v16.d}[1], [x4], x1
    ld1             {v17.d}[1], [x4], x1
    ld1             {v19.d}[1], [x4], x1

    transpose4x8.h  v18, v16, v17, v19, v26, v27, v28, v29

    h264_loop_filter_chroma_intra

    st2             {v16.h,v17.h}[0], [x0], x1
    st2             {v16.h,v17.h}[1], [x0], x1
    st2             {v16.h,v17.h}[2], [x0], x1
    st2             {v16.h,v17.h}[3], [x0], x1
    st2             {v16.h,v17.h}[4], [x0], x1
    st2             {v16.h,v17.h}[5], [x0], x1
    st2             {v16.h,v17.h}[6], [x0], x1
    st2             {v16.h,v17.h}[7], [x0], x1

    ld1             {v18.d}[0], [x4], x1
    ld1             {v16.d}[0], [x4], x1
    ld1             {v17.d}[0], [x4], x1
    ld1             {v19.d}[0], [x4], x1
    ld1             {v18.d}[1], [x4], x1
    ld1             {v16.d}[1], [x4], x1
    ld1             {v17.d}[1], [x4], x1
    ld1             {v19.d}[1], [x4], x1

    transpose4x8.h  v18, v16, v17, v19, v26, v27, v28, v29

    h264_loop_filter_chroma_intra

    st2             {v16.h,v17.h}[0], [x0], x1
    st2             {v16.h,v17.h}[1], [x0], x1
    st2             {v16.h,v17.h}[2], [x0], x1
    st2             {v16.h,v17.h}[3], [x0], x1
    st2             {v16.h,v17.h}[4], [x0], x1
    st2             {v16.h,v17.h}[5], [x0], x1
    st2             {v16.h,v17.h}[6], [x0], x1
    st2             {v16.h,v17.h}[7], [x0], x1

    ret
endfunc
699

700 701 702 703 704 705
// void deblock_strength( uint8_t nnz[X264_SCAN8_SIZE],
//                        int8_t ref[2][X264_SCAN8_LUMA_SIZE],
//                        int16_t mv[2][X264_SCAN8_LUMA_SIZE][2],
//                        uint8_t bs[2][8][4], int mvy_limit,
//                        int bframe )
function deblock_strength_neon, export=1
706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813
    movi        v4.16b, #0
    lsl         w4,  w4,  #8
    add         x3,  x3,  #32
    sub         w4,  w4,  #(1<<8)-3
    movi        v5.16b, #0
    dup         v6.8h,  w4
    mov         x6,  #-32

bframe:
    // load bytes ref
    add         x2,  x2,  #16
    ld1        {v31.d}[1], [x1], #8
    ld1        {v1.16b}, [x1], #16
    movi        v0.16b,  #0
    ld1        {v2.16b}, [x1], #16
    ext         v3.16b,  v0.16b,  v1.16b,  #15
    ext         v0.16b,  v0.16b,  v2.16b,  #15
    unzip       v21.4s,  v22.4s,  v1.4s,   v2.4s
    unzip       v23.4s,  v20.4s,  v3.4s,   v0.4s
    ext         v21.16b, v31.16b, v22.16b, #12

    eor         v0.16b,  v20.16b, v22.16b
    eor         v1.16b,  v21.16b, v22.16b
    orr         v4.16b,  v4.16b,  v0.16b
    orr         v5.16b,  v5.16b,  v1.16b

    ld1        {v21.8h}, [x2], #16      // mv + 0x10
    ld1        {v19.8h}, [x2], #16      // mv + 0x20
    ld1        {v22.8h}, [x2], #16      // mv + 0x30
    ld1        {v18.8h}, [x2], #16      // mv + 0x40
    ld1        {v23.8h}, [x2], #16      // mv + 0x50
    ext         v19.16b, v19.16b, v22.16b, #12
    ext         v18.16b, v18.16b, v23.16b, #12
    sabd        v0.8h,   v22.8h,  v19.8h
    ld1        {v19.8h}, [x2], #16      // mv + 0x60
    sabd        v1.8h,   v23.8h,  v18.8h
    ld1        {v24.8h}, [x2], #16      // mv + 0x70
    uqxtn       v0.8b,   v0.8h
    ld1        {v18.8h}, [x2], #16      // mv + 0x80
    ld1        {v25.8h}, [x2], #16      // mv + 0x90
    uqxtn2      v0.16b,  v1.8h
    ext         v19.16b, v19.16b, v24.16b, #12
    ext         v18.16b, v18.16b, v25.16b, #12
    sabd        v1.8h,   v24.8h,  v19.8h
    sabd        v2.8h,   v25.8h,  v18.8h
    uqxtn       v1.8b,   v1.8h
    uqxtn2      v1.16b,  v2.8h

    uqsub       v0.16b,  v0.16b,  v6.16b
    uqsub       v1.16b,  v1.16b,  v6.16b
    uqxtn       v0.8b,   v0.8h
    uqxtn2      v0.16b,  v1.8h

    sabd        v1.8h,   v22.8h,  v23.8h
    orr         v4.16b,  v4.16b,  v0.16b

    sabd        v0.8h,   v21.8h,  v22.8h
    sabd        v2.8h,   v23.8h,  v24.8h
    sabd        v3.8h,   v24.8h,  v25.8h
    uqxtn       v0.8b,   v0.8h
    uqxtn2      v0.16b,  v1.8h
    uqxtn       v1.8b,   v2.8h
    uqxtn2      v1.16b,  v3.8h

    uqsub       v0.16b,  v0.16b,  v6.16b
    uqsub       v1.16b,  v1.16b,  v6.16b
    uqxtn       v0.8b,   v0.8h
    uqxtn2      v0.16b,  v1.8h
    subs        w5,  w5,  #1
    orr         v5.16b,  v5.16b,  v0.16b
    b.eq        bframe

    movi        v6.16b, #1
    // load bytes nnz
    ld1        {v31.d}[1], [x0], #8
    ld1        {v1.16b}, [x0], #16
    movi        v0.16b,  #0
    ld1        {v2.16b}, [x0], #16
    ext         v3.16b,  v0.16b,  v1.16b,  #15
    ext         v0.16b,  v0.16b,  v2.16b,  #15
    unzip       v21.4s,  v22.4s,  v1.4s,   v2.4s
    unzip       v23.4s,  v20.4s,  v3.4s,   v0.4s
    ext         v21.16b, v31.16b, v22.16b, #12

    movrel      x7,  transpose_table
    ld1        {v7.16b}, [x7]
    orr         v0.16b,  v20.16b, v22.16b
    orr         v1.16b,  v21.16b, v22.16b
    umin        v0.16b,  v0.16b,  v6.16b
    umin        v1.16b,  v1.16b,  v6.16b
    umin        v4.16b,  v4.16b,  v6.16b        // mv ? 1 : 0
    umin        v5.16b,  v5.16b,  v6.16b
    add         v0.16b,  v0.16b,  v0.16b        // nnz ? 2 : 0
    add         v1.16b,  v1.16b,  v1.16b
    umax        v4.16b,  v4.16b,  v0.16b
    umax        v5.16b,  v5.16b,  v1.16b
    tbl         v6.16b, {v4.16b}, v7.16b
    st1        {v5.16b}, [x3], x6       // bs[1]
    st1        {v6.16b}, [x3]           // bs[0]
    ret
endfunc

const transpose_table
    .byte 0, 4,  8, 12
    .byte 1, 5,  9, 13
    .byte 2, 6, 10, 14
    .byte 3, 7, 11, 15
endconst