source/common/x86/ipfilter8.asm

   1 ;*****************************************************************************
   2 ;* Copyright (C) 2013 x265 project
   3 ;*
   4 ;* Authors: Min Chen <chenm003@163.com>
   5 ;*          Nabajit Deka <nabajit@multicorewareinc.com>
   6 ;*          Praveen Kumar Tiwari <praveen@multicorewareinc.com>
   7 ;*
   8 ;* This program is free software; you can redistribute it and/or modify
   9 ;* it under the terms of the GNU General Public License as published by
  10 ;* the Free Software Foundation; either version 2 of the License, or
  11 ;* (at your option) any later version.
  12 ;*
  13 ;* This program is distributed in the hope that it will be useful,
  14 ;* but WITHOUT ANY WARRANTY; without even the implied warranty of
  15 ;* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  16 ;* GNU General Public License for more details.
  17 ;*
  18 ;* You should have received a copy of the GNU General Public License
  19 ;* along with this program; if not, write to the Free Software
  20 ;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02111, USA.
  21 ;*
  22 ;* This program is also available under a commercial proprietary license.
  23 ;* For more information, contact us at license @ x265.com.
  24 ;*****************************************************************************/
  25
  26 %include "x86inc.asm"
  27 %include "x86util.asm"
  28
  29 SECTION_RODATA 32
  30 tab_Tm:    db 0, 1, 2, 3, 1, 2, 3, 4, 2, 3, 4, 5, 3, 4, 5, 6
  31            db 4, 5, 6, 7, 5, 6, 7, 8, 6, 7, 8, 9, 7, 8, 9, 10
  32            db 8, 9,10,11, 9,10,11,12,10,11,12,13,11,12,13, 14
  33
  34 ALIGN 32
  35 const interp4_vpp_shuf, times 2 db 0, 4, 1, 5, 2, 6, 3, 7, 8, 12, 9, 13, 10, 14, 11, 15
  36
  37 ALIGN 32
  38 const interp4_vpp_shuf1, dd 0, 1, 1, 2, 2, 3, 3, 4
  39                          dd 2, 3, 3, 4, 4, 5, 5, 6
  40
  41 ALIGN 32
  42 tab_Lm:    db 0, 1, 2, 3, 4,  5,  6,  7,  1, 2, 3, 4,  5,  6,  7,  8
  43            db 2, 3, 4, 5, 6,  7,  8,  9,  3, 4, 5, 6,  7,  8,  9,  10
  44            db 4, 5, 6, 7, 8,  9,  10, 11, 5, 6, 7, 8,  9,  10, 11, 12
  45            db 6, 7, 8, 9, 10, 11, 12, 13, 7, 8, 9, 10, 11, 12, 13, 14
  46
  47 tab_Vm:    db 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1
  48            db 2, 3, 2, 3, 2, 3, 2, 3, 2, 3, 2, 3, 2, 3, 2, 3
  49
  50 tab_Cm:    db 0, 2, 1, 3, 0, 2, 1, 3, 0, 2, 1, 3, 0, 2, 1, 3
  51
  52 tab_c_526336:   times 4 dd 8192*64+2048
  53
  54 tab_ChromaCoeff: db  0, 64,  0,  0
  55                  db -2, 58, 10, -2
  56                  db -4, 54, 16, -2
  57                  db -6, 46, 28, -4
  58                  db -4, 36, 36, -4
  59                  db -4, 28, 46, -6
  60                  db -2, 16, 54, -4
  61                  db -2, 10, 58, -2
  62
  63 tab_ChromaCoeffV: times 4 dw 0, 64
  64                   times 4 dw 0, 0
  65
  66                   times 4 dw -2, 58
  67                   times 4 dw 10, -2
  68
  69                   times 4 dw -4, 54
  70                   times 4 dw 16, -2
  71
  72                   times 4 dw -6, 46
  73                   times 4 dw 28, -4
  74
  75                   times 4 dw -4, 36
  76                   times 4 dw 36, -4
  77
  78                   times 4 dw -4, 28
  79                   times 4 dw 46, -6
  80
  81                   times 4 dw -2, 16
  82                   times 4 dw 54, -4
  83
  84                   times 4 dw -2, 10
  85                   times 4 dw 58, -2
  86
  87 tab_LumaCoeff:   db   0, 0,  0,  64,  0,   0,  0,  0
  88                  db  -1, 4, -10, 58,  17, -5,  1,  0
  89                  db  -1, 4, -11, 40,  40, -11, 4, -1
  90                  db   0, 1, -5,  17,  58, -10, 4, -1
  91
  92 tab_LumaCoeffV: times 4 dw 0, 0
  93                 times 4 dw 0, 64
  94                 times 4 dw 0, 0
  95                 times 4 dw 0, 0
  96
  97                 times 4 dw -1, 4
  98                 times 4 dw -10, 58
  99                 times 4 dw 17, -5
 100                 times 4 dw 1, 0
 101
 102                 times 4 dw -1, 4
 103                 times 4 dw -11, 40
 104                 times 4 dw 40, -11
 105                 times 4 dw 4, -1
 106
 107                 times 4 dw 0, 1
 108                 times 4 dw -5, 17
 109                 times 4 dw 58, -10
 110                 times 4 dw 4, -1
 111
 112 tab_LumaCoeffVer: times 8 db 0, 0
 113                   times 8 db 0, 64
 114                   times 8 db 0, 0
 115                   times 8 db 0, 0
 116
 117                   times 8 db -1, 4
 118                   times 8 db -10, 58
 119                   times 8 db 17, -5
 120                   times 8 db 1, 0
 121
 122                   times 8 db -1, 4
 123                   times 8 db -11, 40
 124                   times 8 db 40, -11
 125                   times 8 db 4, -1
 126
 127                   times 8 db 0, 1
 128                   times 8 db -5, 17
 129                   times 8 db 58, -10
 130                   times 8 db 4, -1
 131
 132 ALIGN 32
 133 tab_LumaCoeffVer_32: times 16 db 0, 0
 134                      times 16 db 0, 64
 135                      times 16 db 0, 0
 136                      times 16 db 0, 0
 137
 138                      times 16 db -1, 4
 139                      times 16 db -10, 58
 140                      times 16 db 17, -5
 141                      times 16 db 1, 0
 142
 143                      times 16 db -1, 4
 144                      times 16 db -11, 40
 145                      times 16 db 40, -11
 146                      times 16 db 4, -1
 147
 148                      times 16 db 0, 1
 149                      times 16 db -5, 17
 150                      times 16 db 58, -10
 151                      times 16 db 4, -1
 152
 153 ALIGN 32
 154 tab_ChromaCoeffVer_32: times 16 db 0, 64
 155                        times 16 db 0, 0
 156
 157                        times 16 db -2, 58
 158                        times 16 db 10, -2
 159
 160                        times 16 db -4, 54
 161                        times 16 db 16, -2
 162
 163                        times 16 db -6, 46
 164                        times 16 db 28, -4
 165
 166                        times 16 db -4, 36
 167                        times 16 db 36, -4
 168
 169                        times 16 db -4, 28
 170                        times 16 db 46, -6
 171
 172                        times 16 db -2, 16
 173                        times 16 db 54, -4
 174
 175                        times 16 db -2, 10
 176                        times 16 db 58, -2
 177
 178 tab_c_64_n64:   times 8 db 64, -64
 179
 180 const interp4_shuf, times 2 db 0, 1, 8, 9, 4, 5, 12, 13, 2, 3, 10, 11, 6, 7, 14, 15
 181
 182 ALIGN 32
 183 interp4_horiz_shuf1:    db 0, 1, 2, 3, 1, 2, 3, 4, 2, 3, 4, 5, 3, 4, 5, 6
 184                         db 8, 9, 10, 11, 9, 10, 11, 12, 10, 11, 12, 13, 11, 12, 13, 14
 185
 186 SECTION .text
 187
 188 cextern pb_128
 189 cextern pw_1
 190 cextern pw_512
 191 cextern pw_2000
 192
 193 %macro FILTER_H4_w2_2 3
 194     movh        %2, [srcq - 1]
 195     pshufb      %2, %2, Tm0
 196     movh        %1, [srcq + srcstrideq - 1]
 197     pshufb      %1, %1, Tm0
 198     punpcklqdq  %2, %1
 199     pmaddubsw   %2, coef2
 200     phaddw      %2, %2
 201     pmulhrsw    %2, %3
 202     packuswb    %2, %2
 203     movd        r4, %2
 204     mov         [dstq], r4w
 205     shr         r4, 16
 206     mov         [dstq + dststrideq], r4w
 207 %endmacro
 208
 209 ;-----------------------------------------------------------------------------
 210 ; void interp_4tap_horiz_pp_2x4(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
 211 ;-----------------------------------------------------------------------------
 212 INIT_XMM sse4
 213 cglobal interp_4tap_horiz_pp_2x4, 4, 6, 5, src, srcstride, dst, dststride
 214 %define coef2       m4
 215 %define Tm0         m3
 216 %define t2          m2
 217 %define t1          m1
 218 %define t0          m0
 219
 220 mov         r4d,        r4m
 221
 222 %ifdef PIC
 223 lea         r5,          [tab_ChromaCoeff]
 224 movd        coef2,       [r5 + r4 * 4]
 225 %else
 226 movd        coef2,       [tab_ChromaCoeff + r4 * 4]
 227 %endif
 228
 229 pshufd      coef2,       coef2,      0
 230 mova        t2,          [pw_512]
 231 mova        Tm0,         [tab_Tm]
 232
 233 %rep 2
 234 FILTER_H4_w2_2   t0, t1, t2
 235 lea         srcq,       [srcq + srcstrideq * 2]
 236 lea         dstq,       [dstq + dststrideq * 2]
 237 %endrep
 238
 239 RET
 240
 241 ;-----------------------------------------------------------------------------
 242 ; void interp_4tap_horiz_pp_2x8(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
 243 ;-----------------------------------------------------------------------------
 244 INIT_XMM sse4
 245 cglobal interp_4tap_horiz_pp_2x8, 4, 6, 5, src, srcstride, dst, dststride
 246 %define coef2       m4
 247 %define Tm0         m3
 248 %define t2          m2
 249 %define t1          m1
 250 %define t0          m0
 251
 252 mov         r4d,        r4m
 253
 254 %ifdef PIC
 255 lea         r5,          [tab_ChromaCoeff]
 256 movd        coef2,       [r5 + r4 * 4]
 257 %else
 258 movd        coef2,       [tab_ChromaCoeff + r4 * 4]
 259 %endif
 260
 261 pshufd      coef2,       coef2,      0
 262 mova        t2,          [pw_512]
 263 mova        Tm0,         [tab_Tm]
 264
 265 %rep 4
 266 FILTER_H4_w2_2   t0, t1, t2
 267 lea         srcq,       [srcq + srcstrideq * 2]
 268 lea         dstq,       [dstq + dststrideq * 2]
 269 %endrep
 270
 271 RET
 272
 273 ;-----------------------------------------------------------------------------
 274 ; void interp_4tap_horiz_pp_2x16(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
 275 ;-----------------------------------------------------------------------------
 276 INIT_XMM sse4
 277 cglobal interp_4tap_horiz_pp_2x16, 4, 6, 5, src, srcstride, dst, dststride
 278 %define coef2       m4
 279 %define Tm0         m3
 280 %define t2          m2
 281 %define t1          m1
 282 %define t0          m0
 283
 284 mov         r4d,        r4m
 285
 286 %ifdef PIC
 287 lea         r5,          [tab_ChromaCoeff]
 288 movd        coef2,       [r5 + r4 * 4]
 289 %else
 290 movd        coef2,       [tab_ChromaCoeff + r4 * 4]
 291 %endif
 292
 293 pshufd      coef2,       coef2,      0
 294 mova        t2,          [pw_512]
 295 mova        Tm0,         [tab_Tm]
 296
 297 mov         r5d,        16/2
 298
 299 .loop:
 300 FILTER_H4_w2_2   t0, t1, t2
 301 lea         srcq,       [srcq + srcstrideq * 2]
 302 lea         dstq,       [dstq + dststrideq * 2]
 303 dec         r5d
 304 jnz         .loop
 305
 306 RET
 307
 308 %macro FILTER_H4_w4_2 3
 309     movh        %2, [srcq - 1]
 310     pshufb      %2, %2, Tm0
 311     pmaddubsw   %2, coef2
 312     movh        %1, [srcq + srcstrideq - 1]
 313     pshufb      %1, %1, Tm0
 314     pmaddubsw   %1, coef2
 315     phaddw      %2, %1
 316     pmulhrsw    %2, %3
 317     packuswb    %2, %2
 318     movd        [dstq], %2
 319     palignr     %2, %2, 4
 320     movd        [dstq + dststrideq], %2
 321 %endmacro
 322
 323 ;-----------------------------------------------------------------------------
 324 ; void interp_4tap_horiz_pp_4x2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
 325 ;-----------------------------------------------------------------------------
 326 INIT_XMM sse4
 327 cglobal interp_4tap_horiz_pp_4x2, 4, 6, 5, src, srcstride, dst, dststride
 328 %define coef2       m4
 329 %define Tm0         m3
 330 %define t2          m2
 331 %define t1          m1
 332 %define t0          m0
 333
 334 mov         r4d,        r4m
 335
 336 %ifdef PIC
 337 lea         r5,          [tab_ChromaCoeff]
 338 movd        coef2,       [r5 + r4 * 4]
 339 %else
 340 movd        coef2,       [tab_ChromaCoeff + r4 * 4]
 341 %endif
 342
 343 pshufd      coef2,       coef2,      0
 344 mova        t2,          [pw_512]
 345 mova        Tm0,         [tab_Tm]
 346
 347 FILTER_H4_w4_2   t0, t1, t2
 348
 349 RET
 350
 351 ;-----------------------------------------------------------------------------
 352 ; void interp_4tap_horiz_pp_4x4(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
 353 ;-----------------------------------------------------------------------------
 354 INIT_XMM sse4
 355 cglobal interp_4tap_horiz_pp_4x4, 4, 6, 5, src, srcstride, dst, dststride
 356 %define coef2       m4
 357 %define Tm0         m3
 358 %define t2          m2
 359 %define t1          m1
 360 %define t0          m0
 361
 362 mov         r4d,        r4m
 363
 364 %ifdef PIC
 365 lea         r5,          [tab_ChromaCoeff]
 366 movd        coef2,       [r5 + r4 * 4]
 367 %else
 368 movd        coef2,       [tab_ChromaCoeff + r4 * 4]
 369 %endif
 370
 371 pshufd      coef2,       coef2,      0
 372 mova        t2,          [pw_512]
 373 mova        Tm0,         [tab_Tm]
 374
 375 %rep 2
 376 FILTER_H4_w4_2   t0, t1, t2
 377 lea         srcq,       [srcq + srcstrideq * 2]
 378 lea         dstq,       [dstq + dststrideq * 2]
 379 %endrep
 380
 381 RET
 382
 383 ;-----------------------------------------------------------------------------
 384 ; void interp_4tap_horiz_pp_4x8(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
 385 ;-----------------------------------------------------------------------------
 386 INIT_XMM sse4
 387 cglobal interp_4tap_horiz_pp_4x8, 4, 6, 5, src, srcstride, dst, dststride
 388 %define coef2       m4
 389 %define Tm0         m3
 390 %define t2          m2
 391 %define t1          m1
 392 %define t0          m0
 393
 394 mov         r4d,        r4m
 395
 396 %ifdef PIC
 397 lea         r5,          [tab_ChromaCoeff]
 398 movd        coef2,       [r5 + r4 * 4]
 399 %else
 400 movd        coef2,       [tab_ChromaCoeff + r4 * 4]
 401 %endif
 402
 403 pshufd      coef2,       coef2,      0
 404 mova        t2,          [pw_512]
 405 mova        Tm0,         [tab_Tm]
 406
 407 %rep 4
 408 FILTER_H4_w4_2   t0, t1, t2
 409 lea         srcq,       [srcq + srcstrideq * 2]
 410 lea         dstq,       [dstq + dststrideq * 2]
 411 %endrep
 412
 413 RET
 414
 415 ;-----------------------------------------------------------------------------
 416 ; void interp_4tap_horiz_pp_4x16(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
 417 ;-----------------------------------------------------------------------------
 418 INIT_XMM sse4
 419 cglobal interp_4tap_horiz_pp_4x16, 4, 6, 5, src, srcstride, dst, dststride
 420 %define coef2       m4
 421 %define Tm0         m3
 422 %define t2          m2
 423 %define t1          m1
 424 %define t0          m0
 425
 426 mov         r4d,        r4m
 427
 428 %ifdef PIC
 429 lea         r5,          [tab_ChromaCoeff]
 430 movd        coef2,       [r5 + r4 * 4]
 431 %else
 432 movd        coef2,       [tab_ChromaCoeff + r4 * 4]
 433 %endif
 434
 435 pshufd      coef2,       coef2,      0
 436 mova        t2,          [pw_512]
 437 mova        Tm0,         [tab_Tm]
 438
 439 %rep 8
 440 FILTER_H4_w4_2   t0, t1, t2
 441 lea         srcq,       [srcq + srcstrideq * 2]
 442 lea         dstq,       [dstq + dststrideq * 2]
 443 %endrep
 444
 445 RET
 446
 447 ;-----------------------------------------------------------------------------
 448 ; void interp_4tap_horiz_pp_4x32(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
 449 ;-----------------------------------------------------------------------------
 450 INIT_XMM sse4
 451 cglobal interp_4tap_horiz_pp_4x32, 4, 6, 5, src, srcstride, dst, dststride
 452 %define coef2       m4
 453 %define Tm0         m3
 454 %define t2          m2
 455 %define t1          m1
 456 %define t0          m0
 457
 458 mov         r4d,        r4m
 459
 460 %ifdef PIC
 461 lea         r5,          [tab_ChromaCoeff]
 462 movd        coef2,       [r5 + r4 * 4]
 463 %else
 464 movd        coef2,       [tab_ChromaCoeff + r4 * 4]
 465 %endif
 466
 467 pshufd      coef2,       coef2,      0
 468 mova        t2,          [pw_512]
 469 mova        Tm0,         [tab_Tm]
 470
 471 mov         r5d,        32/2
 472
 473 .loop:
 474 FILTER_H4_w4_2   t0, t1, t2
 475 lea         srcq,       [srcq + srcstrideq * 2]
 476 lea         dstq,       [dstq + dststrideq * 2]
 477 dec         r5d
 478 jnz         .loop
 479
 480 RET
 481
 482 ALIGN 32
 483 const interp_4tap_8x8_horiz_shuf,   dd 0, 4, 1, 5, 2, 6, 3, 7
 484
 485
 486 %macro FILTER_H4_w6 3
 487     movu        %1, [srcq - 1]
 488     pshufb      %2, %1, Tm0
 489     pmaddubsw   %2, coef2
 490     pshufb      %1, %1, Tm1
 491     pmaddubsw   %1, coef2
 492     phaddw      %2, %1
 493     pmulhrsw    %2, %3
 494     packuswb    %2, %2
 495     movd        [dstq],      %2
 496     pextrw      [dstq + 4], %2, 2
 497 %endmacro
 498
 499 %macro FILTER_H4_w8 3
 500     movu        %1, [srcq - 1]
 501     pshufb      %2, %1, Tm0
 502     pmaddubsw   %2, coef2
 503     pshufb      %1, %1, Tm1
 504     pmaddubsw   %1, coef2
 505     phaddw      %2, %1
 506     pmulhrsw    %2, %3
 507     packuswb    %2, %2
 508     movh        [dstq],      %2
 509 %endmacro
 510
 511 %macro FILTER_H4_w12 3
 512     movu        %1, [srcq - 1]
 513     pshufb      %2, %1, Tm0
 514     pmaddubsw   %2, coef2
 515     pshufb      %1, %1, Tm1
 516     pmaddubsw   %1, coef2
 517     phaddw      %2, %1
 518     pmulhrsw    %2, %3
 519     movu        %1, [srcq - 1 + 8]
 520     pshufb      %1, %1, Tm0
 521     pmaddubsw   %1, coef2
 522     phaddw      %1, %1
 523     pmulhrsw    %1, %3
 524     packuswb    %2, %1
 525     movh        [dstq],      %2
 526     pextrd      [dstq + 8], %2, 2
 527 %endmacro
 528
 529 %macro FILTER_H4_w16 4
 530     movu        %1, [srcq - 1]
 531     pshufb      %2, %1, Tm0
 532     pmaddubsw   %2, coef2
 533     pshufb      %1, %1, Tm1
 534     pmaddubsw   %1, coef2
 535     phaddw      %2, %1
 536     movu        %1, [srcq - 1 + 8]
 537     pshufb      %4, %1, Tm0
 538     pmaddubsw   %4, coef2
 539     pshufb      %1, %1, Tm1
 540     pmaddubsw   %1, coef2
 541     phaddw      %4, %1
 542     pmulhrsw    %2, %3
 543     pmulhrsw    %4, %3
 544     packuswb    %2, %4
 545     movu        [dstq],      %2
 546 %endmacro
 547
 548 %macro FILTER_H4_w24 4
 549     movu        %1, [srcq - 1]
 550     pshufb      %2, %1, Tm0
 551     pmaddubsw   %2, coef2
 552     pshufb      %1, %1, Tm1
 553     pmaddubsw   %1, coef2
 554     phaddw      %2, %1
 555     movu        %1, [srcq - 1 + 8]
 556     pshufb      %4, %1, Tm0
 557     pmaddubsw   %4, coef2
 558     pshufb      %1, %1, Tm1
 559     pmaddubsw   %1, coef2
 560     phaddw      %4, %1
 561     pmulhrsw    %2, %3
 562     pmulhrsw    %4, %3
 563     packuswb    %2, %4
 564     movu        [dstq],          %2
 565     movu        %1, [srcq - 1 + 16]
 566     pshufb      %2, %1, Tm0
 567     pmaddubsw   %2, coef2
 568     pshufb      %1, %1, Tm1
 569     pmaddubsw   %1, coef2
 570     phaddw      %2, %1
 571     pmulhrsw    %2, %3
 572     packuswb    %2, %2
 573     movh        [dstq + 16],     %2
 574 %endmacro
 575
 576 %macro FILTER_H4_w32 4
 577     movu        %1, [srcq - 1]
 578     pshufb      %2, %1, Tm0
 579     pmaddubsw   %2, coef2
 580     pshufb      %1, %1, Tm1
 581     pmaddubsw   %1, coef2
 582     phaddw      %2, %1
 583     movu        %1, [srcq - 1 + 8]
 584     pshufb      %4, %1, Tm0
 585     pmaddubsw   %4, coef2
 586     pshufb      %1, %1, Tm1
 587     pmaddubsw   %1, coef2
 588     phaddw      %4, %1
 589     pmulhrsw    %2, %3
 590     pmulhrsw    %4, %3
 591     packuswb    %2, %4
 592     movu        [dstq],      %2
 593     movu        %1, [srcq - 1 + 16]
 594     pshufb      %2, %1, Tm0
 595     pmaddubsw   %2, coef2
 596     pshufb      %1, %1, Tm1
 597     pmaddubsw   %1, coef2
 598     phaddw      %2, %1
 599     movu        %1, [srcq - 1 + 24]
 600     pshufb      %4, %1, Tm0
 601     pmaddubsw   %4, coef2
 602     pshufb      %1, %1, Tm1
 603     pmaddubsw   %1, coef2
 604     phaddw      %4, %1
 605     pmulhrsw    %2, %3
 606     pmulhrsw    %4, %3
 607     packuswb    %2, %4
 608     movu        [dstq + 16],      %2
 609 %endmacro
 610
 611 %macro FILTER_H4_w16o 5
 612     movu        %1, [srcq + %5 - 1]
 613     pshufb      %2, %1, Tm0
 614     pmaddubsw   %2, coef2
 615     pshufb      %1, %1, Tm1
 616     pmaddubsw   %1, coef2
 617     phaddw      %2, %1
 618     movu        %1, [srcq + %5 - 1 + 8]
 619     pshufb      %4, %1, Tm0
 620     pmaddubsw   %4, coef2
 621     pshufb      %1, %1, Tm1
 622     pmaddubsw   %1, coef2
 623     phaddw      %4, %1
 624     pmulhrsw    %2, %3
 625     pmulhrsw    %4, %3
 626     packuswb    %2, %4
 627     movu        [dstq + %5],      %2
 628 %endmacro
 629
 630 %macro FILTER_H4_w48 4
 631     FILTER_H4_w16o %1, %2, %3, %4, 0
 632     FILTER_H4_w16o %1, %2, %3, %4, 16
 633     FILTER_H4_w16o %1, %2, %3, %4, 32
 634 %endmacro
 635
 636 %macro FILTER_H4_w64 4
 637     FILTER_H4_w16o %1, %2, %3, %4, 0
 638     FILTER_H4_w16o %1, %2, %3, %4, 16
 639     FILTER_H4_w16o %1, %2, %3, %4, 32
 640     FILTER_H4_w16o %1, %2, %3, %4, 48
 641 %endmacro
 642
 643 ;-----------------------------------------------------------------------------
 644 ; void interp_4tap_horiz_pp_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
 645 ;-----------------------------------------------------------------------------
 646 %macro IPFILTER_CHROMA 2
 647 INIT_XMM sse4
 648 cglobal interp_4tap_horiz_pp_%1x%2, 4, 6, 6, src, srcstride, dst, dststride
 649 %define coef2       m5
 650 %define Tm0         m4
 651 %define Tm1         m3
 652 %define t2          m2
 653 %define t1          m1
 654 %define t0          m0
 655
 656 mov         r4d,        r4m
 657
 658 %ifdef PIC
 659 lea         r5,          [tab_ChromaCoeff]
 660 movd        coef2,       [r5 + r4 * 4]
 661 %else
 662 movd        coef2,       [tab_ChromaCoeff + r4 * 4]
 663 %endif
 664
 665 mov           r5d,       %2
 666
 667 pshufd      coef2,       coef2,      0
 668 mova        t2,          [pw_512]
 669 mova        Tm0,         [tab_Tm]
 670 mova        Tm1,         [tab_Tm + 16]
 671
 672 .loop:
 673 FILTER_H4_w%1   t0, t1, t2
 674 add         srcq,        srcstrideq
 675 add         dstq,        dststrideq
 676
 677 dec         r5d
 678 jnz        .loop
 679
 680 RET
 681 %endmacro
 682
 683
 684 IPFILTER_CHROMA 6,   8
 685 IPFILTER_CHROMA 8,   2
 686 IPFILTER_CHROMA 8,   4
 687 IPFILTER_CHROMA 8,   6
 688 IPFILTER_CHROMA 8,   8
 689 IPFILTER_CHROMA 8,  16
 690 IPFILTER_CHROMA 8,  32
 691 IPFILTER_CHROMA 12, 16
 692
 693 IPFILTER_CHROMA 6,  16
 694 IPFILTER_CHROMA 8,  12
 695 IPFILTER_CHROMA 8,  64
 696 IPFILTER_CHROMA 12, 32
 697
 698 ;-----------------------------------------------------------------------------
 699 ; void interp_4tap_horiz_pp_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
 700 ;-----------------------------------------------------------------------------
 701 %macro IPFILTER_CHROMA_W 2
 702 INIT_XMM sse4
 703 cglobal interp_4tap_horiz_pp_%1x%2, 4, 6, 7, src, srcstride, dst, dststride
 704 %define coef2       m6
 705 %define Tm0         m5
 706 %define Tm1         m4
 707 %define t3          m3
 708 %define t2          m2
 709 %define t1          m1
 710 %define t0          m0
 711
 712 mov         r4d,         r4m
 713
 714 %ifdef PIC
 715 lea         r5,          [tab_ChromaCoeff]
 716 movd        coef2,       [r5 + r4 * 4]
 717 %else
 718 movd        coef2,       [tab_ChromaCoeff + r4 * 4]
 719 %endif
 720
 721 mov         r5d,          %2
 722
 723 pshufd      coef2,       coef2,      0
 724 mova        t2,          [pw_512]
 725 mova        Tm0,         [tab_Tm]
 726 mova        Tm1,         [tab_Tm + 16]
 727
 728 .loop:
 729 FILTER_H4_w%1   t0, t1, t2, t3
 730 add         srcq,        srcstrideq
 731 add         dstq,        dststrideq
 732
 733 dec         r5d
 734 jnz        .loop
 735
 736 RET
 737 %endmacro
 738
 739 IPFILTER_CHROMA_W 16,  4
 740 IPFILTER_CHROMA_W 16,  8
 741 IPFILTER_CHROMA_W 16, 12
 742 IPFILTER_CHROMA_W 16, 16
 743 IPFILTER_CHROMA_W 16, 32
 744 IPFILTER_CHROMA_W 32,  8
 745 IPFILTER_CHROMA_W 32, 16
 746 IPFILTER_CHROMA_W 32, 24
 747 IPFILTER_CHROMA_W 24, 32
 748 IPFILTER_CHROMA_W 32, 32
 749
 750 IPFILTER_CHROMA_W 16, 24
 751 IPFILTER_CHROMA_W 16, 64
 752 IPFILTER_CHROMA_W 32, 48
 753 IPFILTER_CHROMA_W 24, 64
 754 IPFILTER_CHROMA_W 32, 64
 755
 756 IPFILTER_CHROMA_W 64, 64
 757 IPFILTER_CHROMA_W 64, 32
 758 IPFILTER_CHROMA_W 64, 48
 759 IPFILTER_CHROMA_W 48, 64
 760 IPFILTER_CHROMA_W 64, 16
 761
 762
 763 %macro FILTER_H8_W8 7-8   ; t0, t1, t2, t3, coef, c512, src, dst
 764     movu        %1, %7
 765     pshufb      %2, %1, [tab_Lm +  0]
 766     pmaddubsw   %2, %5
 767     pshufb      %3, %1, [tab_Lm + 16]
 768     pmaddubsw   %3, %5
 769     phaddw      %2, %3
 770     pshufb      %4, %1, [tab_Lm + 32]
 771     pmaddubsw   %4, %5
 772     pshufb      %1, %1, [tab_Lm + 48]
 773     pmaddubsw   %1, %5
 774     phaddw      %4, %1
 775     phaddw      %2, %4
 776   %if %0 == 8
 777     pmulhrsw    %2, %6
 778     packuswb    %2, %2
 779     movh        %8, %2
 780   %endif
 781 %endmacro
 782
 783 %macro FILTER_H8_W4 2
 784     movu        %1, [r0 - 3 + r5]
 785     pshufb      %2, %1, [tab_Lm]
 786     pmaddubsw   %2, m3
 787     pshufb      m7, %1, [tab_Lm + 16]
 788     pmaddubsw   m7, m3
 789     phaddw      %2, m7
 790     phaddw      %2, %2
 791 %endmacro
 792
 793 ;----------------------------------------------------------------------------------------------------------------------------
 794 ; void interp_8tap_horiz_%3_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
 795 ;----------------------------------------------------------------------------------------------------------------------------
 796 %macro IPFILTER_LUMA 3
 797 INIT_XMM sse4
 798 cglobal interp_8tap_horiz_%3_%1x%2, 4,7,8
 799
 800     mov       r4d, r4m
 801
 802 %ifdef PIC
 803     lea       r6, [tab_LumaCoeff]
 804     movh      m3, [r6 + r4 * 8]
 805 %else
 806     movh      m3, [tab_LumaCoeff + r4 * 8]
 807 %endif
 808     punpcklqdq  m3, m3
 809
 810 %ifidn %3, pp
 811     mova      m2, [pw_512]
 812 %else
 813     mova      m2, [pw_2000]
 814 %endif
 815
 816     mov       r4d, %2
 817 %ifidn %3, ps
 818     add       r3, r3
 819     cmp       r5m, byte 0
 820     je        .loopH
 821     lea       r6, [r1 + 2 * r1]
 822     sub       r0, r6
 823     add       r4d, 7
 824 %endif
 825
 826 .loopH:
 827     xor       r5, r5
 828 %rep %1 / 8
 829   %ifidn %3, pp
 830     FILTER_H8_W8  m0, m1, m4, m5, m3, m2, [r0 - 3 + r5], [r2 + r5]
 831   %else
 832     FILTER_H8_W8  m0, m1, m4, m5, m3, UNUSED, [r0 - 3 + r5]
 833     psubw     m1, m2
 834     movu      [r2 + 2 * r5], m1
 835   %endif
 836     add       r5, 8
 837 %endrep
 838
 839 %rep (%1 % 8) / 4
 840     FILTER_H8_W4  m0, m1
 841   %ifidn %3, pp
 842     pmulhrsw  m1, m2
 843     packuswb  m1, m1
 844     movd      [r2 + r5], m1
 845   %else
 846     psubw     m1, m2
 847     movh      [r2 + 2 * r5], m1
 848   %endif
 849 %endrep
 850
 851     add       r0, r1
 852     add       r2, r3
 853
 854     dec       r4d
 855     jnz       .loopH
 856     RET
 857 %endmacro
 858
 859
 860 INIT_YMM avx2
 861 cglobal interp_8tap_horiz_pp_4x4, 4,6,6
 862     mov             r4d, r4m
 863
 864 %ifdef PIC
 865     lea             r5, [tab_LumaCoeff]
 866     vpbroadcastq    m0, [r5 + r4 * 8]
 867 %else
 868     vpbroadcastq    m0, [tab_LumaCoeff + r4 * 8]
 869 %endif
 870
 871     mova            m1, [tab_Lm]
 872     vpbroadcastd    m2, [pw_1]
 873
 874     ; register map
 875     ; m0 - interpolate coeff
 876     ; m1 - shuffle order table
 877     ; m2 - constant word 1
 878
 879     sub             r0, 3
 880     ; Row 0-1
 881     vbroadcasti128  m3, [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
 882     pshufb          m3, m1
 883     pmaddubsw       m3, m0
 884     pmaddwd         m3, m2
 885     vbroadcasti128  m4, [r0 + r1]                   ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
 886     pshufb          m4, m1
 887     pmaddubsw       m4, m0
 888     pmaddwd         m4, m2
 889     phaddd          m3, m4                          ; DWORD [R1D R1C R0D R0C R1B R1A R0B R0A]
 890
 891     ; Row 2-3
 892     lea             r0, [r0 + r1 * 2]
 893     vbroadcasti128  m4, [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
 894     pshufb          m4, m1
 895     pmaddubsw       m4, m0
 896     pmaddwd         m4, m2
 897     vbroadcasti128  m5, [r0 + r1]                   ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
 898     pshufb          m5, m1
 899     pmaddubsw       m5, m0
 900     pmaddwd         m5, m2
 901     phaddd          m4, m5                          ; DWORD [R3D R3C R2D R2C R3B R3A R2B R2A]
 902
 903     packssdw        m3, m4                          ; WORD [R3D R3C R2D R2C R1D R1C R0D R0C R3B R3A R2B R2A R1B R1A R0B R0A]
 904     pmulhrsw        m3, [pw_512]
 905     vextracti128    xm4, m3, 1
 906     packuswb        xm3, xm4                        ; BYTE [R3D R3C R2D R2C R1D R1C R0D R0C R3B R3A R2B R2A R1B R1A R0B R0A]
 907     pshufb          xm3, [interp4_shuf]             ; [row3 row1 row2 row0]
 908
 909     lea             r0, [r3 * 3]
 910     movd            [r2], xm3
 911     pextrd          [r2+r3], xm3, 2
 912     pextrd          [r2+r3*2], xm3, 1
 913     pextrd          [r2+r0], xm3, 3
 914     RET
 915
 916 INIT_YMM avx2
 917 cglobal interp_8tap_horiz_pp_8x4, 4, 6, 7
 918     mov             r4d, r4m
 919
 920 %ifdef PIC
 921     lea             r5, [tab_LumaCoeff]
 922     vpbroadcastq    m0, [r5 + r4 * 8]
 923 %else
 924     vpbroadcastq    m0, [tab_LumaCoeff + r4 * 8]
 925 %endif
 926
 927     mova            m1, [tab_Lm]
 928     mova            m2, [tab_Lm + 32]
 929
 930     ; register map
 931     ; m0     - interpolate coeff
 932     ; m1, m2 - shuffle order table
 933
 934     sub             r0, 3
 935     lea             r5, [r1 * 3]
 936     lea             r4, [r3 * 3]
 937
 938     ; Row 0
 939     vbroadcasti128  m3, [r0]                        ; [x E D C B A 9 8 7 6 5 4 3 2 1 0]
 940     pshufb          m4, m3, m2
 941     pshufb          m3, m1
 942     pmaddubsw       m3, m0
 943     pmaddubsw       m4, m0
 944     phaddw          m3, m4
 945     ; Row 1
 946     vbroadcasti128  m4, [r0 + r1]                   ; [x E D C B A 9 8 7 6 5 4 3 2 1 0]
 947     pshufb          m5, m4, m2
 948     pshufb          m4, m1
 949     pmaddubsw       m4, m0
 950     pmaddubsw       m5, m0
 951     phaddw          m4, m5
 952
 953     phaddw          m3, m4                          ; WORD [R1H R1G R1D R1C R0H R0G R0D R0C R1F R1E R1B R1A R0F R0E R0B R0A]
 954     pmulhrsw        m3, [pw_512]
 955
 956     ; Row 2
 957     vbroadcasti128  m4, [r0 + r1 * 2]               ; [x E D C B A 9 8 7 6 5 4 3 2 1 0]
 958     pshufb          m5, m4, m2
 959     pshufb          m4, m1
 960     pmaddubsw       m4, m0
 961     pmaddubsw       m5, m0
 962     phaddw          m4, m5
 963     ; Row 3
 964     vbroadcasti128  m5, [r0 + r5]                   ; [x E D C B A 9 8 7 6 5 4 3 2 1 0]
 965     pshufb          m6, m5, m2
 966     pshufb          m5, m1
 967     pmaddubsw       m5, m0
 968     pmaddubsw       m6, m0
 969     phaddw          m5, m6
 970
 971     phaddw          m4, m5                          ; WORD [R3H R3G R3D R3C R2H R2G R2D R2C R3F R3E R3B R3A R2F R2E R2B R2A]
 972     pmulhrsw        m4, [pw_512]
 973
 974     packuswb        m3, m4
 975     vextracti128    xm4, m3, 1
 976     punpcklwd       xm5, xm3, xm4
 977
 978     movq            [r2], xm5
 979     movhps          [r2 + r3], xm5
 980
 981     punpckhwd       xm5, xm3, xm4
 982     movq            [r2 + r3 * 2], xm5
 983     movhps          [r2 + r4], xm5
 984     RET
 985
 986 %macro IPFILTER_LUMA_AVX2_8xN 2
 987 INIT_YMM avx2
 988 cglobal interp_8tap_horiz_pp_%1x%2, 4, 7, 7
 989     mov             r4d, r4m
 990
 991 %ifdef PIC
 992     lea             r5, [tab_LumaCoeff]
 993     vpbroadcastq    m0, [r5 + r4 * 8]
 994 %else
 995     vpbroadcastq    m0, [tab_LumaCoeff + r4 * 8]
 996 %endif
 997
 998     mova            m1, [tab_Lm]
 999     mova            m2, [tab_Lm + 32]
1000
1001     ; register map
1002     ; m0     - interpolate coeff
1003     ; m1, m2 - shuffle order table
1004
1005     sub             r0, 3
1006     lea             r5, [r1 * 3]
1007     lea             r6, [r3 * 3]
1008     mov             r4d, %2 / 4
1009 .loop:
1010     ; Row 0
1011     vbroadcasti128  m3, [r0]                        ; [x E D C B A 9 8 7 6 5 4 3 2 1 0]
1012     pshufb          m4, m3, m2
1013     pshufb          m3, m1
1014     pmaddubsw       m3, m0
1015     pmaddubsw       m4, m0
1016     phaddw          m3, m4
1017     ; Row 1
1018     vbroadcasti128  m4, [r0 + r1]                   ; [x E D C B A 9 8 7 6 5 4 3 2 1 0]
1019     pshufb          m5, m4, m2
1020     pshufb          m4, m1
1021     pmaddubsw       m4, m0
1022     pmaddubsw       m5, m0
1023     phaddw          m4, m5
1024
1025     phaddw          m3, m4                          ; WORD [R1H R1G R1D R1C R0H R0G R0D R0C R1F R1E R1B R1A R0F R0E R0B R0A]
1026     pmulhrsw        m3, [pw_512]
1027
1028     ; Row 2
1029     vbroadcasti128  m4, [r0 + r1 * 2]               ; [x E D C B A 9 8 7 6 5 4 3 2 1 0]
1030     pshufb          m5, m4, m2
1031     pshufb          m4, m1
1032     pmaddubsw       m4, m0
1033     pmaddubsw       m5, m0
1034     phaddw          m4, m5
1035     ; Row 3
1036     vbroadcasti128  m5, [r0 + r5]                   ; [x E D C B A 9 8 7 6 5 4 3 2 1 0]
1037     pshufb          m6, m5, m2
1038     pshufb          m5, m1
1039     pmaddubsw       m5, m0
1040     pmaddubsw       m6, m0
1041     phaddw          m5, m6
1042
1043     phaddw          m4, m5                          ; WORD [R3H R3G R3D R3C R2H R2G R2D R2C R3F R3E R3B R3A R2F R2E R2B R2A]
1044     pmulhrsw        m4, [pw_512]
1045
1046     packuswb        m3, m4
1047     vextracti128    xm4, m3, 1
1048     punpcklwd       xm5, xm3, xm4
1049
1050     movq            [r2], xm5
1051     movhps          [r2 + r3], xm5
1052
1053     punpckhwd       xm5, xm3, xm4
1054     movq            [r2 + r3 * 2], xm5
1055     movhps          [r2 + r6], xm5
1056
1057     lea             r0, [r0 + r1 * 4]
1058     lea             r2, [r2 + r3 * 4]
1059     dec             r4d
1060     jnz             .loop
1061     RET
1062 %endmacro
1063
1064 IPFILTER_LUMA_AVX2_8xN 8, 8
1065 IPFILTER_LUMA_AVX2_8xN 8, 16
1066 IPFILTER_LUMA_AVX2_8xN 8, 32
1067
1068 %macro IPFILTER_LUMA_AVX2 2
1069 INIT_YMM avx2
1070 cglobal interp_8tap_horiz_pp_%1x%2, 4,6,8
1071     sub               r0,        3
1072     mov               r4d,       r4m
1073 %ifdef PIC
1074     lea               r5,        [tab_LumaCoeff]
1075     vpbroadcastd      m0,        [r5 + r4 * 8]
1076     vpbroadcastd      m1,        [r5 + r4 * 8 + 4]
1077 %else
1078     vpbroadcastd      m0,         [tab_LumaCoeff + r4 * 8]
1079     vpbroadcastd      m1,         [tab_LumaCoeff + r4 * 8 + 4]
1080 %endif
1081     movu              m3,         [tab_Tm + 16]
1082     vpbroadcastd      m7,         [pw_1]
1083
1084     ; register map
1085     ; m0 , m1 interpolate coeff
1086     ; m2 , m2  shuffle order table
1087     ; m7 - pw_1
1088
1089     mov               r4d,        %2/2
1090 .loop:
1091     ; Row 0
1092     vbroadcasti128    m4,         [r0]                        ; [x E D C B A 9 8 7 6 5 4 3 2 1 0]
1093     pshufb            m5,         m4,     m3
1094     pshufb            m4,         [tab_Tm]
1095     pmaddubsw         m4,         m0
1096     pmaddubsw         m5,         m1
1097     paddw             m4,         m5
1098     pmaddwd           m4,         m7
1099     vbroadcasti128    m5,         [r0 + 8]                    ; second 8 elements in Row0
1100     pshufb            m6,         m5,     m3
1101     pshufb            m5,         [tab_Tm]
1102     pmaddubsw         m5,         m0
1103     pmaddubsw         m6,         m1
1104     paddw             m5,         m6
1105     pmaddwd           m5,         m7
1106     packssdw          m4,         m5                          ; [17 16 15 14 07 06 05 04 13 12 11 10 03 02 01 00]
1107     pmulhrsw          m4,         [pw_512]
1108     vbroadcasti128    m2,         [r0 + r1]                        ; [x E D C B A 9 8 7 6 5 4 3 2 1 0]
1109     pshufb            m5,         m2,     m3
1110     pshufb            m2,         [tab_Tm]
1111     pmaddubsw         m2,         m0
1112     pmaddubsw         m5,         m1
1113     paddw             m2,         m5
1114     pmaddwd           m2,         m7
1115     vbroadcasti128    m5,         [r0 + r1 + 8]                    ; second 8 elements in Row0
1116     pshufb            m6,         m5,     m3
1117     pshufb            m5,         [tab_Tm]
1118     pmaddubsw         m5,         m0
1119     pmaddubsw         m6,         m1
1120     paddw             m5,         m6
1121     pmaddwd           m5,         m7
1122     packssdw          m2,         m5                          ; [17 16 15 14 07 06 05 04 13 12 11 10 03 02 01 00]
1123     pmulhrsw          m2,         [pw_512]
1124     packuswb          m4,         m2
1125     vpermq            m4,         m4,     11011000b
1126     vextracti128      xm5,        m4,     1
1127     pshufd            xm4,        xm4,    11011000b
1128     pshufd            xm5,        xm5,    11011000b
1129     movu              [r2],       xm4
1130     movu              [r2+r3],    xm5
1131     lea               r0,         [r0 + r1 * 2]
1132     lea               r2,         [r2 + r3 * 2]
1133     dec               r4d
1134     jnz              .loop
1135     RET
1136 %endmacro
1137
1138 %macro IPFILTER_LUMA_32x_avx2 2
1139 INIT_YMM avx2
1140 cglobal interp_8tap_horiz_pp_%1x%2, 4,6,8
1141     sub               r0,         3
1142     mov               r4d,        r4m
1143 %ifdef PIC
1144     lea               r5,         [tab_LumaCoeff]
1145     vpbroadcastd      m0,         [r5 + r4 * 8]
1146     vpbroadcastd      m1,         [r5 + r4 * 8 + 4]
1147 %else
1148     vpbroadcastd      m0,         [tab_LumaCoeff + r4 * 8]
1149     vpbroadcastd      m1,         [tab_LumaCoeff + r4 * 8 + 4]
1150 %endif
1151     movu              m3,         [tab_Tm + 16]
1152     vpbroadcastd      m7,         [pw_1]
1153
1154     ; register map
1155     ; m0 , m1 interpolate coeff
1156     ; m2 , m2  shuffle order table
1157     ; m7 - pw_1
1158
1159     mov               r4d,        %2
1160 .loop:
1161     ; Row 0
1162     vbroadcasti128    m4,         [r0]                        ; [x E D C B A 9 8 7 6 5 4 3 2 1 0]
1163     pshufb            m5,         m4,     m3
1164     pshufb            m4,         [tab_Tm]
1165     pmaddubsw         m4,         m0
1166     pmaddubsw         m5,         m1
1167     paddw             m4,         m5
1168     pmaddwd           m4,         m7
1169     vbroadcasti128    m5,         [r0 + 8]
1170     pshufb            m6,         m5,     m3
1171     pshufb            m5,         [tab_Tm]
1172     pmaddubsw         m5,         m0
1173     pmaddubsw         m6,         m1
1174     paddw             m5,         m6
1175     pmaddwd           m5,         m7
1176     packssdw          m4,         m5                          ; [17 16 15 14 07 06 05 04 13 12 11 10 03 02 01 00]
1177     pmulhrsw          m4,         [pw_512]
1178     vbroadcasti128    m2,         [r0 + 16]
1179     pshufb            m5,         m2,     m3
1180     pshufb            m2,         [tab_Tm]
1181     pmaddubsw         m2,         m0
1182     pmaddubsw         m5,         m1
1183     paddw             m2,         m5
1184     pmaddwd           m2,         m7
1185     vbroadcasti128    m5,         [r0 + 24]
1186     pshufb            m6,         m5,     m3
1187     pshufb            m5,         [tab_Tm]
1188     pmaddubsw         m5,         m0
1189     pmaddubsw         m6,         m1
1190     paddw             m5,         m6
1191     pmaddwd           m5,         m7
1192     packssdw          m2,         m5
1193     pmulhrsw          m2,         [pw_512]
1194     packuswb          m4,         m2
1195     vpermq            m4,         m4,     11011000b
1196     vextracti128      xm5,        m4,     1
1197     pshufd            xm4,        xm4,    11011000b
1198     pshufd            xm5,        xm5,    11011000b
1199     movu              [r2],       xm4
1200     movu              [r2 + 16],  xm5
1201     lea               r0,         [r0 + r1]
1202     lea               r2,         [r2 + r3]
1203     dec               r4d
1204     jnz               .loop
1205     RET
1206 %endmacro
1207
1208 %macro IPFILTER_LUMA_64x_avx2 2
1209 INIT_YMM avx2
1210 cglobal interp_8tap_horiz_pp_%1x%2, 4,6,8
1211     sub               r0,    3
1212     mov               r4d,   r4m
1213 %ifdef PIC
1214     lea               r5,        [tab_LumaCoeff]
1215     vpbroadcastd      m0,        [r5 + r4 * 8]
1216     vpbroadcastd      m1,        [r5 + r4 * 8 + 4]
1217 %else
1218     vpbroadcastd      m0,        [tab_LumaCoeff + r4 * 8]
1219     vpbroadcastd      m1,        [tab_LumaCoeff + r4 * 8 + 4]
1220 %endif
1221     movu              m3,        [tab_Tm + 16]
1222     vpbroadcastd      m7,        [pw_1]
1223
1224     ; register map
1225     ; m0 , m1 interpolate coeff
1226     ; m2 , m2  shuffle order table
1227     ; m7 - pw_1
1228
1229     mov               r4d,   %2
1230 .loop:
1231     ; Row 0
1232     vbroadcasti128    m4,        [r0]                        ; [x E D C B A 9 8 7 6 5 4 3 2 1 0]
1233     pshufb            m5,        m4,    m3
1234     pshufb            m4,        [tab_Tm]
1235     pmaddubsw         m4,        m0
1236     pmaddubsw         m5,        m1
1237     paddw             m4,        m5
1238     pmaddwd           m4,        m7
1239     vbroadcasti128    m5,        [r0 + 8]
1240     pshufb            m6,        m5,    m3
1241     pshufb            m5,        [tab_Tm]
1242     pmaddubsw         m5,        m0
1243     pmaddubsw         m6,        m1
1244     paddw             m5,        m6
1245     pmaddwd           m5,        m7
1246     packssdw          m4,        m5                          ; [17 16 15 14 07 06 05 04 13 12 11 10 03 02 01 00]
1247     pmulhrsw          m4,        [pw_512]
1248     vbroadcasti128    m2,        [r0 + 16]
1249     pshufb            m5,        m2,    m3
1250     pshufb            m2,        [tab_Tm]
1251     pmaddubsw         m2,        m0
1252     pmaddubsw         m5,        m1
1253     paddw             m2,        m5
1254     pmaddwd           m2,        m7
1255     vbroadcasti128    m5,        [r0 + 24]
1256     pshufb            m6,        m5,    m3
1257     pshufb            m5,        [tab_Tm]
1258     pmaddubsw         m5,        m0
1259     pmaddubsw         m6,        m1
1260     paddw             m5,        m6
1261     pmaddwd           m5,        m7
1262     packssdw          m2,        m5
1263     pmulhrsw          m2,        [pw_512]
1264     packuswb          m4,        m2
1265     vpermq            m4,        m4,    11011000b
1266     vextracti128      xm5,       m4,    1
1267     pshufd            xm4,       xm4,   11011000b
1268     pshufd            xm5,       xm5,   11011000b
1269     movu              [r2],      xm4
1270     movu              [r2 + 16], xm5
1271
1272     vbroadcasti128    m4,        [r0 + 32]
1273     pshufb            m5,        m4,    m3
1274     pshufb            m4,        [tab_Tm]
1275     pmaddubsw         m4,        m0
1276     pmaddubsw         m5,        m1
1277     paddw             m4,        m5
1278     pmaddwd           m4,        m7
1279     vbroadcasti128    m5,        [r0 + 40]
1280     pshufb            m6,        m5,    m3
1281     pshufb            m5,        [tab_Tm]
1282     pmaddubsw         m5,        m0
1283     pmaddubsw         m6,        m1
1284     paddw             m5,        m6
1285     pmaddwd           m5,        m7
1286     packssdw          m4,        m5
1287     pmulhrsw          m4,        [pw_512]
1288     vbroadcasti128    m2,        [r0 + 48]
1289     pshufb            m5,        m2,    m3
1290     pshufb            m2,        [tab_Tm]
1291     pmaddubsw         m2,        m0
1292     pmaddubsw         m5,        m1
1293     paddw             m2,        m5
1294     pmaddwd           m2,        m7
1295     vbroadcasti128    m5,        [r0 + 56]
1296     pshufb            m6,        m5,    m3
1297     pshufb            m5,        [tab_Tm]
1298     pmaddubsw         m5,        m0
1299     pmaddubsw         m6,        m1
1300     paddw             m5,        m6
1301     pmaddwd           m5,        m7
1302     packssdw          m2,        m5
1303     pmulhrsw          m2,        [pw_512]
1304     packuswb          m4,        m2
1305     vpermq            m4,        m4,    11011000b
1306     vextracti128      xm5,       m4,    1
1307     pshufd            xm4,       xm4,   11011000b
1308     pshufd            xm5,       xm5,   11011000b
1309     movu              [r2 +32],  xm4
1310     movu              [r2 + 48], xm5
1311
1312     lea               r0,        [r0 + r1]
1313     lea               r2,        [r2 + r3]
1314     dec               r4d
1315     jnz               .loop
1316     RET
1317 %endmacro
1318
1319 INIT_YMM avx2
1320 cglobal interp_8tap_horiz_pp_48x64, 4,6,8
1321     sub               r0,         3
1322     mov               r4d,        r4m
1323 %ifdef PIC
1324     lea               r5,         [tab_LumaCoeff]
1325     vpbroadcastd      m0,         [r5 + r4 * 8]
1326     vpbroadcastd      m1,         [r5 + r4 * 8 + 4]
1327 %else
1328     vpbroadcastd      m0,         [tab_LumaCoeff + r4 * 8]
1329     vpbroadcastd      m1,         [tab_LumaCoeff + r4 * 8 + 4]
1330 %endif
1331     movu              m3,         [tab_Tm + 16]
1332     vpbroadcastd      m7,         [pw_1]
1333
1334     ; register map
1335     ; m0 , m1 interpolate coeff
1336     ; m2 , m2  shuffle order table
1337     ; m7 - pw_1
1338
1339     mov               r4d,        64
1340 .loop:
1341     ; Row 0
1342     vbroadcasti128    m4,         [r0]                        ; [x E D C B A 9 8 7 6 5 4 3 2 1 0]
1343     pshufb            m5,         m4,     m3
1344     pshufb            m4,         [tab_Tm]
1345     pmaddubsw         m4,         m0
1346     pmaddubsw         m5,         m1
1347     paddw             m4,         m5
1348     pmaddwd           m4,         m7
1349     vbroadcasti128    m5,         [r0 + 8]
1350     pshufb            m6,         m5,     m3
1351     pshufb            m5,         [tab_Tm]
1352     pmaddubsw         m5,         m0
1353     pmaddubsw         m6,         m1
1354     paddw             m5,         m6
1355     pmaddwd           m5,         m7
1356     packssdw          m4,         m5                          ; [17 16 15 14 07 06 05 04 13 12 11 10 03 02 01 00]
1357     pmulhrsw          m4,         [pw_512]
1358
1359     vbroadcasti128    m2,         [r0 + 16]
1360     pshufb            m5,         m2,     m3
1361     pshufb            m2,         [tab_Tm]
1362     pmaddubsw         m2,         m0
1363     pmaddubsw         m5,         m1
1364     paddw             m2,         m5
1365     pmaddwd           m2,         m7
1366     vbroadcasti128    m5,         [r0 + 24]
1367     pshufb            m6,         m5,     m3
1368     pshufb            m5,         [tab_Tm]
1369     pmaddubsw         m5,         m0
1370     pmaddubsw         m6,         m1
1371     paddw             m5,         m6
1372     pmaddwd           m5,         m7
1373     packssdw          m2,         m5
1374     pmulhrsw          m2,         [pw_512]
1375     packuswb          m4,         m2
1376     vpermq            m4,         m4,     11011000b
1377     vextracti128      xm5,        m4,     1
1378     pshufd            xm4,        xm4,    11011000b
1379     pshufd            xm5,        xm5,    11011000b
1380     movu              [r2],       xm4
1381     movu              [r2 + 16],  xm5
1382
1383     vbroadcasti128    m4,         [r0 + 32]
1384     pshufb            m5,         m4,     m3
1385     pshufb            m4,         [tab_Tm]
1386     pmaddubsw         m4,         m0
1387     pmaddubsw         m5,         m1
1388     paddw             m4,         m5
1389     pmaddwd           m4,         m7
1390     vbroadcasti128    m5,         [r0 + 40]
1391     pshufb            m6,         m5,     m3
1392     pshufb            m5,         [tab_Tm]
1393     pmaddubsw         m5,         m0
1394     pmaddubsw         m6,         m1
1395     paddw             m5,         m6
1396     pmaddwd           m5,         m7
1397     packssdw          m4,         m5
1398     pmulhrsw          m4,         [pw_512]
1399     packuswb          m4,         m4
1400     vpermq            m4,         m4,     11011000b
1401     pshufd            xm4,        xm4,    11011000b
1402     movu              [r2 + 32],  xm4
1403
1404     lea               r0,         [r0 + r1]
1405     lea               r2,         [r2 + r3]
1406     dec               r4d
1407     jnz               .loop
1408     RET
1409
1410 INIT_YMM avx2
1411 cglobal interp_4tap_horiz_pp_4x4, 4,6,6
1412     mov             r4d, r4m
1413
1414 %ifdef PIC
1415     lea               r5,           [tab_ChromaCoeff]
1416     vpbroadcastd      m0,           [r5 + r4 * 4]
1417 %else
1418     vpbroadcastd      m0,           [tab_ChromaCoeff + r4 * 4]
1419 %endif
1420
1421     vpbroadcastd      m2,           [pw_1]
1422     vbroadcasti128    m1,           [tab_Tm]
1423
1424     ; register map
1425     ; m0 - interpolate coeff
1426     ; m1 - shuffle order table
1427     ; m2 - constant word 1
1428
1429     dec                r0
1430
1431     ; Row 0-1
1432     vbroadcasti128    m3,           [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
1433     vinserti128       m3,           m3,      [r0 + r1],     1
1434     pshufb            m3,           m1
1435     pmaddubsw         m3,           m0
1436     pmaddwd           m3,           m2
1437
1438     ; Row 2-3
1439     lea               r0,           [r0 + r1 * 2]
1440     vbroadcasti128    m4,           [r0]                      ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
1441     vinserti128       m4,           m4,      [r0 + r1],     1
1442     pshufb            m4,           m1
1443     pmaddubsw         m4,           m0
1444     pmaddwd           m4,           m2
1445
1446     packssdw          m3,           m4
1447     pmulhrsw          m3,           [pw_512]
1448     vextracti128      xm4,          m3,     1
1449     packuswb          xm3,          xm4
1450
1451     lea               r0,           [r3 * 3]
1452     movd              [r2],         xm3
1453     pextrd            [r2+r3],      xm3,     2
1454     pextrd            [r2+r3*2],    xm3,     1
1455     pextrd            [r2+r0],      xm3,     3
1456     RET
1457
1458 INIT_YMM avx2
1459 cglobal interp_4tap_horiz_pp_32x32, 4,6,7
1460     mov             r4d, r4m
1461
1462 %ifdef PIC
1463     lea               r5,           [tab_ChromaCoeff]
1464     vpbroadcastd      m0,           [r5 + r4 * 4]
1465 %else
1466     vpbroadcastd      m0,           [tab_ChromaCoeff + r4 * 4]
1467 %endif
1468
1469     mova              m1,           [interp4_horiz_shuf1]
1470     vpbroadcastd      m2,           [pw_1]
1471     mova              m6,           [pw_512]
1472     ; register map
1473     ; m0 - interpolate coeff
1474     ; m1 - shuffle order table
1475     ; m2 - constant word 1
1476
1477     dec               r0
1478     mov               r4d,          32
1479
1480 .loop:
1481     ; Row 0
1482     vbroadcasti128    m3,           [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
1483     pshufb            m3,           m1
1484     pmaddubsw         m3,           m0
1485     pmaddwd           m3,           m2
1486     vbroadcasti128    m4,           [r0 + 4]
1487     pshufb            m4,           m1
1488     pmaddubsw         m4,           m0
1489     pmaddwd           m4,           m2
1490     packssdw          m3,           m4
1491     pmulhrsw          m3,           m6
1492
1493     vbroadcasti128    m4,           [r0 + 16]
1494     pshufb            m4,           m1
1495     pmaddubsw         m4,           m0
1496     pmaddwd           m4,           m2
1497     vbroadcasti128    m5,           [r0 + 20]
1498     pshufb            m5,           m1
1499     pmaddubsw         m5,           m0
1500     pmaddwd           m5,           m2
1501     packssdw          m4,           m5
1502     pmulhrsw          m4,           m6
1503
1504     packuswb          m3,           m4
1505     vpermq            m3,           m3,      11011000b
1506
1507     movu              [r2],         m3
1508     lea               r2,           [r2 + r3]
1509     lea               r0,           [r0 + r1]
1510     dec               r4d
1511     jnz               .loop
1512     RET
1513
1514
1515 INIT_YMM avx2
1516 cglobal interp_4tap_horiz_pp_16x16, 4, 6, 7
1517     mov               r4d,          r4m
1518
1519 %ifdef PIC
1520     lea               r5,           [tab_ChromaCoeff]
1521     vpbroadcastd      m0,           [r5 + r4 * 4]
1522 %else
1523     vpbroadcastd      m0,           [tab_ChromaCoeff + r4 * 4]
1524 %endif
1525
1526     mova              m6,           [pw_512]
1527     mova              m1,           [interp4_horiz_shuf1]
1528     vpbroadcastd      m2,           [pw_1]
1529
1530     ; register map
1531     ; m0 - interpolate coeff
1532     ; m1 - shuffle order table
1533     ; m2 - constant word 1
1534
1535     dec               r0
1536     mov               r4d,          8
1537
1538 .loop:
1539     ; Row 0
1540     vbroadcasti128    m3,           [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
1541     pshufb            m3,           m1
1542     pmaddubsw         m3,           m0
1543     pmaddwd           m3,           m2
1544     vbroadcasti128    m4,           [r0 + 4]                    ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
1545     pshufb            m4,           m1
1546     pmaddubsw         m4,           m0
1547     pmaddwd           m4,           m2
1548     packssdw          m3,           m4
1549     pmulhrsw          m3,           m6
1550
1551     ; Row 1
1552     vbroadcasti128    m4,           [r0 + r1]                   ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
1553     pshufb            m4,           m1
1554     pmaddubsw         m4,           m0
1555     pmaddwd           m4,           m2
1556     vbroadcasti128    m5,           [r0 + r1 + 4]               ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
1557     pshufb            m5,           m1
1558     pmaddubsw         m5,           m0
1559     pmaddwd           m5,           m2
1560     packssdw          m4,           m5
1561     pmulhrsw          m4,           m6
1562
1563     packuswb          m3,           m4
1564     vpermq            m3,           m3,      11011000b
1565
1566     vextracti128      xm4,          m3,       1
1567     movu              [r2],         xm3
1568     movu              [r2 + r3],    xm4
1569     lea               r2,           [r2 + r3 * 2]
1570     lea               r0,           [r0 + r1 * 2]
1571     dec               r4d
1572     jnz               .loop
1573     RET
1574 ;--------------------------------------------------------------------------------------------------------------
1575 ; void interp_8tap_horiz_pp_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
1576 ;--------------------------------------------------------------------------------------------------------------
1577     IPFILTER_LUMA 4, 4, pp
1578     IPFILTER_LUMA 4, 8, pp
1579     IPFILTER_LUMA 12, 16, pp
1580     IPFILTER_LUMA 4, 16, pp
1581
1582 INIT_YMM avx2
1583 cglobal interp_4tap_horiz_pp_8x8, 4,6,6
1584     mov               r4d,    r4m
1585
1586 %ifdef PIC
1587     lea               r5,           [tab_ChromaCoeff]
1588     vpbroadcastd      m0,           [r5 + r4 * 4]
1589 %else
1590     vpbroadcastd      m0,           [tab_ChromaCoeff + r4 * 4]
1591 %endif
1592
1593     movu              m1,           [tab_Tm]
1594     vpbroadcastd      m2,           [pw_1]
1595
1596     ; register map
1597     ; m0 - interpolate coeff
1598     ; m1 - shuffle order table
1599     ; m2 - constant word 1
1600
1601     sub               r0,           1
1602     mov               r4d,          2
1603
1604 .loop:
1605     ; Row 0
1606     vbroadcasti128    m3,           [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
1607     pshufb            m3,           m1
1608     pmaddubsw         m3,           m0
1609     pmaddwd           m3,           m2
1610
1611     ; Row 1
1612     vbroadcasti128    m4,           [r0 + r1]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
1613     pshufb            m4,           m1
1614     pmaddubsw         m4,           m0
1615     pmaddwd           m4,           m2
1616     packssdw          m3,           m4
1617     pmulhrsw          m3,           [pw_512]
1618     lea               r0,           [r0 + r1 * 2]
1619
1620     ; Row 2
1621     vbroadcasti128    m4,           [r0 ]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
1622     pshufb            m4,           m1
1623     pmaddubsw         m4,           m0
1624     pmaddwd           m4,           m2
1625
1626     ; Row 3
1627     vbroadcasti128    m5,           [r0 + r1]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
1628     pshufb            m5,           m1
1629     pmaddubsw         m5,           m0
1630     pmaddwd           m5,           m2
1631     packssdw          m4,           m5
1632     pmulhrsw          m4,           [pw_512]
1633
1634     packuswb          m3,           m4
1635     mova              m5,           [interp_4tap_8x8_horiz_shuf]
1636     vpermd            m3,           m5,     m3
1637     vextracti128      xm4,          m3,     1
1638     movq              [r2],         xm3
1639     movhps            [r2 + r3],    xm3
1640     lea               r2,           [r2 + r3 * 2]
1641     movq              [r2],         xm4
1642     movhps            [r2 + r3],    xm4
1643     lea               r2,           [r2 + r3 * 2]
1644     lea               r0,           [r0 + r1*2]
1645     dec               r4d
1646     jnz               .loop
1647     RET
1648
1649     IPFILTER_LUMA_AVX2 16, 4
1650     IPFILTER_LUMA_AVX2 16, 8
1651     IPFILTER_LUMA_AVX2 16, 12
1652     IPFILTER_LUMA_AVX2 16, 16
1653     IPFILTER_LUMA_AVX2 16, 32
1654     IPFILTER_LUMA_AVX2 16, 64
1655
1656     IPFILTER_LUMA_32x_avx2 32 , 8
1657     IPFILTER_LUMA_32x_avx2 32 , 16
1658     IPFILTER_LUMA_32x_avx2 32 , 24
1659     IPFILTER_LUMA_32x_avx2 32 , 32
1660     IPFILTER_LUMA_32x_avx2 32 , 64
1661
1662     IPFILTER_LUMA_64x_avx2 64 , 64
1663     IPFILTER_LUMA_64x_avx2 64 , 48
1664     IPFILTER_LUMA_64x_avx2 64 , 32
1665     IPFILTER_LUMA_64x_avx2 64 , 16
1666
1667 ;--------------------------------------------------------------------------------------------------------------
1668 ; void interp_8tap_horiz_pp_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
1669 ;--------------------------------------------------------------------------------------------------------------
1670 %macro IPFILTER_LUMA_PP_W8 2
1671 INIT_XMM sse4
1672 cglobal interp_8tap_horiz_pp_%1x%2, 4,6,7
1673     mov         r4d, r4m
1674
1675 %ifdef PIC
1676     lea         r5, [tab_LumaCoeff]
1677     movh        m3, [r5 + r4 * 8]
1678 %else
1679     movh        m3, [tab_LumaCoeff + r4 * 8]
1680 %endif
1681     pshufd      m0, m3, 0                       ; m0 = coeff-L
1682     pshufd      m1, m3, 0x55                    ; m1 = coeff-H
1683     lea         r5, [tab_Tm]                    ; r5 = shuffle
1684     mova        m2, [pw_512]                    ; m2 = 512
1685
1686     mov         r4d, %2
1687 .loopH:
1688 %assign x 0
1689 %rep %1 / 8
1690     movu        m3, [r0 - 3 + x]                ; m3 = [F E D C B A 9 8 7 6 5 4 3 2 1 0]
1691     pshufb      m4, m3, [r5 + 0*16]             ; m4 = [6 5 4 3 5 4 3 2 4 3 2 1 3 2 1 0]
1692     pshufb      m5, m3, [r5 + 1*16]             ; m5 = [A 9 8 7 9 8 7 6 8 7 6 5 7 6 5 4]
1693     pshufb          m3, [r5 + 2*16]             ; m3 = [E D C B D C B A C B A 9 B A 9 8]
1694     pmaddubsw   m4, m0
1695     pmaddubsw   m6, m5, m1
1696     pmaddubsw   m5, m0
1697     pmaddubsw   m3, m1
1698     paddw       m4, m6
1699     paddw       m5, m3
1700     phaddw      m4, m5
1701     pmulhrsw    m4, m2
1702     packuswb    m4, m4
1703     movh        [r2 + x], m4
1704 %assign x x+8
1705 %endrep
1706
1707     add       r0, r1
1708     add       r2, r3
1709
1710     dec       r4d
1711     jnz      .loopH
1712     RET
1713 %endmacro
1714
1715 IPFILTER_LUMA_PP_W8      8,  4
1716 IPFILTER_LUMA_PP_W8      8,  8
1717 IPFILTER_LUMA_PP_W8      8, 16
1718 IPFILTER_LUMA_PP_W8      8, 32
1719 IPFILTER_LUMA_PP_W8     16,  4
1720 IPFILTER_LUMA_PP_W8     16,  8
1721 IPFILTER_LUMA_PP_W8     16, 12
1722 IPFILTER_LUMA_PP_W8     16, 16
1723 IPFILTER_LUMA_PP_W8     16, 32
1724 IPFILTER_LUMA_PP_W8     16, 64
1725 IPFILTER_LUMA_PP_W8     24, 32
1726 IPFILTER_LUMA_PP_W8     32,  8
1727 IPFILTER_LUMA_PP_W8     32, 16
1728 IPFILTER_LUMA_PP_W8     32, 24
1729 IPFILTER_LUMA_PP_W8     32, 32
1730 IPFILTER_LUMA_PP_W8     32, 64
1731 IPFILTER_LUMA_PP_W8     48, 64
1732 IPFILTER_LUMA_PP_W8     64, 16
1733 IPFILTER_LUMA_PP_W8     64, 32
1734 IPFILTER_LUMA_PP_W8     64, 48
1735 IPFILTER_LUMA_PP_W8     64, 64
1736
1737 ;----------------------------------------------------------------------------------------------------------------------------
1738 ; void interp_8tap_horiz_ps_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
1739 ;----------------------------------------------------------------------------------------------------------------------------
1740     IPFILTER_LUMA 4, 4, ps
1741     IPFILTER_LUMA 8, 8, ps
1742     IPFILTER_LUMA 8, 4, ps
1743     IPFILTER_LUMA 4, 8, ps
1744     IPFILTER_LUMA 16, 16, ps
1745     IPFILTER_LUMA 16, 8, ps
1746     IPFILTER_LUMA 8, 16, ps
1747     IPFILTER_LUMA 16, 12, ps
1748     IPFILTER_LUMA 12, 16, ps
1749     IPFILTER_LUMA 16, 4, ps
1750     IPFILTER_LUMA 4, 16, ps
1751     IPFILTER_LUMA 32, 32, ps
1752     IPFILTER_LUMA 32, 16, ps
1753     IPFILTER_LUMA 16, 32, ps
1754     IPFILTER_LUMA 32, 24, ps
1755     IPFILTER_LUMA 24, 32, ps
1756     IPFILTER_LUMA 32, 8, ps
1757     IPFILTER_LUMA 8, 32, ps
1758     IPFILTER_LUMA 64, 64, ps
1759     IPFILTER_LUMA 64, 32, ps
1760     IPFILTER_LUMA 32, 64, ps
1761     IPFILTER_LUMA 64, 48, ps
1762     IPFILTER_LUMA 48, 64, ps
1763     IPFILTER_LUMA 64, 16, ps
1764     IPFILTER_LUMA 16, 64, ps
1765
1766 ;-----------------------------------------------------------------------------
1767 ; Interpolate HV
1768 ;-----------------------------------------------------------------------------
1769 %macro FILTER_HV8_START 7 ; (t0, t1, t2, t3, t4, off_src, off_coeff) -> (t3, t5), (t4, t1), [2]
1770     mova        %5, [r0 +  (%6 + 0) * 16]
1771     mova        %1, [r0 +  (%6 + 1) * 16]
1772     mova        %2, [r0 +  (%6 + 2) * 16]
1773     punpcklwd   %3, %5, %1
1774     punpckhwd   %5, %1
1775     pmaddwd     %3, [r5 + (%7) * 16]   ; R3 = L[0+1] -- Row 0
1776     pmaddwd     %5, [r5 + (%7) * 16]   ; R0 = H[0+1]
1777     punpcklwd   %4, %1, %2
1778     punpckhwd   %1, %2
1779     pmaddwd     %4, [r5 + (%7) * 16]   ; R4 = L[1+2] -- Row 1
1780     pmaddwd     %1, [r5 + (%7) * 16]   ; R1 = H[1+2]
1781 %endmacro ; FILTER_HV8_START
1782
1783 %macro FILTER_HV8_MID 10 ; (Row3, prevRow, sum0L, sum1L, sum0H, sum1H, t6, t7, off_src, off_coeff) -> [6]
1784     mova        %8, [r0 +  (%9 + 0) * 16]
1785     mova        %1, [r0 +  (%9 + 1) * 16]
1786     punpcklwd   %7, %2, %8
1787     punpckhwd   %2, %8
1788     pmaddwd     %7, [r5 + %10 * 16]
1789     pmaddwd     %2, [r5 + %10 * 16]
1790     paddd       %3, %7              ; R3 = L[0+1+2+3] -- Row 0
1791     paddd       %5, %2              ; R0 = H[0+1+2+3]
1792     punpcklwd   %7, %8, %1
1793     punpckhwd   %8, %1
1794     pmaddwd     %7, [r5 + %10 * 16]
1795     pmaddwd     %8, [r5 + %10 * 16]
1796     paddd       %4, %7              ; R4 = L[1+2+3+4] -- Row 1
1797     paddd       %6, %8              ; R1 = H[1+2+3+4]
1798 %endmacro ; FILTER_HV8_MID
1799
1800 ; Round and Saturate
1801 %macro FILTER_HV8_END 4 ; output in [1, 3]
1802     paddd       %1, [tab_c_526336]
1803     paddd       %2, [tab_c_526336]
1804     paddd       %3, [tab_c_526336]
1805     paddd       %4, [tab_c_526336]
1806     psrad       %1, 12
1807     psrad       %2, 12
1808     psrad       %3, 12
1809     psrad       %4, 12
1810     packssdw    %1, %2
1811     packssdw    %3, %4
1812
1813     ; TODO: is merge better? I think this way is short dependency link
1814     packuswb    %1, %3
1815 %endmacro ; FILTER_HV8_END
1816
1817 ;-----------------------------------------------------------------------------
1818 ; void interp_8tap_hv_pp_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int idxX, int idxY)
1819 ;-----------------------------------------------------------------------------
1820 INIT_XMM ssse3
1821 cglobal interp_8tap_hv_pp_8x8, 4, 7, 8, 0-15*16
1822 %define coef        m7
1823 %define stk_buf     rsp
1824
1825     mov         r4d,        r4m
1826     mov         r5d,        r5m
1827
1828 %ifdef PIC
1829     lea         r6,         [tab_LumaCoeff]
1830     movh        coef,       [r6 + r4 * 8]
1831 %else
1832     movh        coef,       [tab_LumaCoeff + r4 * 8]
1833 %endif
1834     punpcklqdq  coef,       coef
1835
1836     ; move to row -3
1837     lea         r6,         [r1 + r1 * 2]
1838     sub         r0,         r6
1839
1840     xor         r6,         r6
1841     mov         r4,         rsp
1842
1843 .loopH:
1844     FILTER_H8_W8 m0, m1, m2, m3, coef, [pw_512], [r0 - 3]
1845     psubw       m1,         [pw_2000]
1846     mova        [r4],       m1
1847
1848     add         r0,         r1
1849     add         r4,         16
1850     inc         r6
1851     cmp         r6,         8+7
1852     jnz         .loopH
1853
1854     ; ready to phase V
1855     ; Here all of mN is free
1856
1857     ; load coeff table
1858     shl         r5,         6
1859     lea         r6,         [tab_LumaCoeffV]
1860     lea         r5,         [r5 + r6]
1861
1862     ; load intermedia buffer
1863     mov         r0,         stk_buf
1864
1865     ; register mapping
1866     ; r0 - src
1867     ; r5 - coeff
1868     ; r6 - loop_i
1869
1870     ; let's go
1871     xor         r6,         r6
1872
1873     ; TODO: this loop have more than 70 instructions, I think it is more than Intel loop decode cache
1874 .loopV:
1875
1876     FILTER_HV8_START    m1, m2, m3, m4, m0,             0, 0
1877     FILTER_HV8_MID      m6, m2, m3, m4, m0, m1, m7, m5, 3, 1
1878     FILTER_HV8_MID      m5, m6, m3, m4, m0, m1, m7, m2, 5, 2
1879     FILTER_HV8_MID      m6, m5, m3, m4, m0, m1, m7, m2, 7, 3
1880     FILTER_HV8_END      m3, m0, m4, m1
1881
1882     movh        [r2],       m3
1883     movhps      [r2 + r3],  m3
1884
1885     lea         r0,         [r0 + 16 * 2]
1886     lea         r2,         [r2 + r3 * 2]
1887
1888     inc         r6
1889     cmp         r6,         8/2
1890     jnz         .loopV
1891
1892     RET
1893
1894 ;-----------------------------------------------------------------------------
1895 ;void interp_4tap_vert_pp_2x4(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
1896 ;-----------------------------------------------------------------------------
1897 INIT_XMM sse4
1898 cglobal interp_4tap_vert_pp_2x4, 4, 6, 8
1899
1900 mov         r4d,       r4m
1901 sub         r0,        r1
1902
1903 %ifdef PIC
1904 lea         r5,        [tab_ChromaCoeff]
1905 movd        m0,        [r5 + r4 * 4]
1906 %else
1907 movd        m0,        [tab_ChromaCoeff + r4 * 4]
1908 %endif
1909 lea         r4,        [r1 * 3]
1910 lea         r5,        [r0 + 4 * r1]
1911 pshufb      m0,        [tab_Cm]
1912 mova        m1,        [pw_512]
1913
1914 movd        m2,        [r0]
1915 movd        m3,        [r0 + r1]
1916 movd        m4,        [r0 + 2 * r1]
1917 movd        m5,        [r0 + r4]
1918
1919 punpcklbw   m2,        m3
1920 punpcklbw   m6,        m4,        m5
1921 punpcklbw   m2,        m6
1922
1923 pmaddubsw   m2,        m0
1924
1925 movd        m6,        [r5]
1926
1927 punpcklbw   m3,        m4
1928 punpcklbw   m7,        m5,        m6
1929 punpcklbw   m3,        m7
1930
1931 pmaddubsw   m3,        m0
1932
1933 phaddw      m2,        m3
1934
1935 pmulhrsw    m2,        m1
1936
1937 movd        m7,        [r5 + r1]
1938
1939 punpcklbw   m4,        m5
1940 punpcklbw   m3,        m6,        m7
1941 punpcklbw   m4,        m3
1942
1943 pmaddubsw   m4,        m0
1944
1945 movd        m3,        [r5 + 2 * r1]
1946
1947 punpcklbw   m5,        m6
1948 punpcklbw   m7,        m3
1949 punpcklbw   m5,        m7
1950
1951 pmaddubsw   m5,        m0
1952
1953 phaddw      m4,        m5
1954
1955 pmulhrsw    m4,        m1
1956 packuswb    m2,        m4
1957
1958 pextrw      [r2],      m2, 0
1959 pextrw      [r2 + r3], m2, 2
1960 lea         r2,        [r2 + 2 * r3]
1961 pextrw      [r2],      m2, 4
1962 pextrw      [r2 + r3], m2, 6
1963
1964 RET
1965
1966 ;-----------------------------------------------------------------------------
1967 ; void interp_4tap_vert_pp_2x8(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
1968 ;-----------------------------------------------------------------------------
1969 %macro FILTER_V4_W2_H4 2
1970 INIT_XMM sse4
1971 cglobal interp_4tap_vert_pp_2x%2, 4, 6, 8
1972
1973 mov         r4d,       r4m
1974 sub         r0,        r1
1975
1976 %ifdef PIC
1977 lea         r5,        [tab_ChromaCoeff]
1978 movd        m0,        [r5 + r4 * 4]
1979 %else
1980 movd        m0,        [tab_ChromaCoeff + r4 * 4]
1981 %endif
1982
1983 pshufb      m0,        [tab_Cm]
1984
1985 mova        m1,        [pw_512]
1986
1987 mov         r4d,       %2
1988 lea         r5,        [3 * r1]
1989
1990 .loop:
1991 movd        m2,        [r0]
1992 movd        m3,        [r0 + r1]
1993 movd        m4,        [r0 + 2 * r1]
1994 movd        m5,        [r0 + r5]
1995
1996 punpcklbw   m2,        m3
1997 punpcklbw   m6,        m4,        m5
1998 punpcklbw   m2,        m6
1999
2000 pmaddubsw   m2,        m0
2001
2002 lea         r0,        [r0 + 4 * r1]
2003 movd        m6,        [r0]
2004
2005 punpcklbw   m3,        m4
2006 punpcklbw   m7,        m5,        m6
2007 punpcklbw   m3,        m7
2008
2009 pmaddubsw   m3,        m0
2010
2011 phaddw      m2,        m3
2012
2013 pmulhrsw    m2,        m1
2014
2015 movd        m7,        [r0 + r1]
2016
2017 punpcklbw   m4,        m5
2018 punpcklbw   m3,        m6,        m7
2019 punpcklbw   m4,        m3
2020
2021 pmaddubsw   m4,        m0
2022
2023 movd        m3,        [r0 + 2 * r1]
2024
2025 punpcklbw   m5,        m6
2026 punpcklbw   m7,        m3
2027 punpcklbw   m5,        m7
2028
2029 pmaddubsw   m5,        m0
2030
2031 phaddw      m4,        m5
2032
2033 pmulhrsw    m4,        m1
2034 packuswb    m2,        m4
2035
2036 pextrw      [r2],      m2, 0
2037 pextrw      [r2 + r3], m2, 2
2038 lea         r2,        [r2 + 2 * r3]
2039 pextrw      [r2],      m2, 4
2040 pextrw      [r2 + r3], m2, 6
2041
2042 lea         r2,        [r2 + 2 * r3]
2043
2044 sub         r4,        4
2045 jnz        .loop
2046 RET
2047 %endmacro
2048
2049 FILTER_V4_W2_H4 2, 8
2050
2051 FILTER_V4_W2_H4 2, 16
2052
2053 ;-----------------------------------------------------------------------------
2054 ; void interp_4tap_vert_pp_4x2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
2055 ;-----------------------------------------------------------------------------
2056 INIT_XMM sse4
2057 cglobal interp_4tap_vert_pp_4x2, 4, 6, 6
2058
2059 mov         r4d,       r4m
2060 sub         r0,        r1
2061
2062 %ifdef PIC
2063 lea         r5,        [tab_ChromaCoeff]
2064 movd        m0,        [r5 + r4 * 4]
2065 %else
2066 movd        m0,        [tab_ChromaCoeff + r4 * 4]
2067 %endif
2068
2069 pshufb      m0,        [tab_Cm]
2070 lea         r5,        [r0 + 2 * r1]
2071
2072 movd        m2,        [r0]
2073 movd        m3,        [r0 + r1]
2074 movd        m4,        [r5]
2075 movd        m5,        [r5 + r1]
2076
2077 punpcklbw   m2,        m3
2078 punpcklbw   m1,        m4,        m5
2079 punpcklbw   m2,        m1
2080
2081 pmaddubsw   m2,        m0
2082
2083 movd        m1,        [r0 + 4 * r1]
2084
2085 punpcklbw   m3,        m4
2086 punpcklbw   m5,        m1
2087 punpcklbw   m3,        m5
2088
2089 pmaddubsw   m3,        m0
2090
2091 phaddw      m2,        m3
2092
2093 pmulhrsw    m2,        [pw_512]
2094 packuswb    m2,        m2
2095 movd        [r2],      m2
2096 pextrd      [r2 + r3], m2,  1
2097
2098 RET
2099
2100 ;-----------------------------------------------------------------------------
2101 ; void interp_4tap_vert_pp_4x4(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
2102 ;-----------------------------------------------------------------------------
2103 INIT_XMM sse4
2104 cglobal interp_4tap_vert_pp_4x4, 4, 6, 8
2105
2106 mov         r4d,       r4m
2107 sub         r0,        r1
2108
2109 %ifdef PIC
2110 lea         r5,        [tab_ChromaCoeff]
2111 movd        m0,        [r5 + r4 * 4]
2112 %else
2113 movd        m0,        [tab_ChromaCoeff + r4 * 4]
2114 %endif
2115
2116 pshufb      m0,        [tab_Cm]
2117 mova        m1,        [pw_512]
2118 lea         r5,        [r0 + 4 * r1]
2119 lea         r4,        [r1 * 3]
2120
2121 movd        m2,        [r0]
2122 movd        m3,        [r0 + r1]
2123 movd        m4,        [r0 + 2 * r1]
2124 movd        m5,        [r0 + r4]
2125
2126 punpcklbw   m2,        m3
2127 punpcklbw   m6,        m4,        m5
2128 punpcklbw   m2,        m6
2129
2130 pmaddubsw   m2,        m0
2131
2132 movd        m6,        [r5]
2133
2134 punpcklbw   m3,        m4
2135 punpcklbw   m7,        m5,        m6
2136 punpcklbw   m3,        m7
2137
2138 pmaddubsw   m3,        m0
2139
2140 phaddw      m2,        m3
2141
2142 pmulhrsw    m2,        m1
2143
2144 movd        m7,        [r5 + r1]
2145
2146 punpcklbw   m4,        m5
2147 punpcklbw   m3,        m6,        m7
2148 punpcklbw   m4,        m3
2149
2150 pmaddubsw   m4,        m0
2151
2152 movd        m3,        [r5 + 2 * r1]
2153
2154 punpcklbw   m5,        m6
2155 punpcklbw   m7,        m3
2156 punpcklbw   m5,        m7
2157
2158 pmaddubsw   m5,        m0
2159
2160 phaddw      m4,        m5
2161
2162 pmulhrsw    m4,        m1
2163
2164 packuswb    m2,        m4
2165 movd        [r2],      m2
2166 pextrd      [r2 + r3], m2, 1
2167 lea         r2,        [r2 + 2 * r3]
2168 pextrd      [r2],      m2, 2
2169 pextrd      [r2 + r3], m2, 3
2170
2171 RET
2172
2173 INIT_YMM avx2
2174 cglobal interp_4tap_vert_pp_4x4, 4, 6, 3
2175     mov             r4d, r4m
2176     shl             r4d, 6
2177     sub             r0, r1
2178
2179 %ifdef PIC
2180     lea             r5, [tab_ChromaCoeffVer_32]
2181     add             r5, r4
2182 %else
2183     lea             r5, [tab_ChromaCoeffVer_32 + r4]
2184 %endif
2185
2186     lea             r4, [r1 * 3]
2187
2188     movd            xm1, [r0]
2189     pinsrd          xm1, [r0 + r1], 1
2190     pinsrd          xm1, [r0 + r1 * 2], 2
2191     pinsrd          xm1, [r0 + r4], 3                       ; m1 = row[3 2 1 0]
2192     lea             r0, [r0 + r1 * 4]
2193     movd            xm2, [r0]
2194     pinsrd          xm2, [r0 + r1], 1
2195     pinsrd          xm2, [r0 + r1 * 2], 2                   ; m2 = row[x 6 5 4]
2196     vinserti128     m1, m1, xm2, 1                          ; m1 = row[x 6 5 4 3 2 1 0]
2197     mova            m2, [interp4_vpp_shuf1]
2198     vpermd          m0, m2, m1                              ; m0 = row[4 3 3 2 2 1 1 0]
2199     mova            m2, [interp4_vpp_shuf1 + mmsize]
2200     vpermd          m1, m2, m1                              ; m1 = row[6 5 5 4 4 3 3 2]
2201
2202     mova            m2, [interp4_vpp_shuf]
2203     pshufb          m0, m0, m2
2204     pshufb          m1, m1, m2
2205     pmaddubsw       m0, [r5]
2206     pmaddubsw       m1, [r5 + mmsize]
2207     paddw           m0, m1                                  ; m0 = WORD ROW[3 2 1 0]
2208     pmulhrsw        m0, [pw_512]
2209     vextracti128    xm1, m0, 1
2210     packuswb        xm0, xm1
2211     lea             r5, [r3 * 3]
2212     movd            [r2], xm0
2213     pextrd          [r2 + r3], xm0, 1
2214     pextrd          [r2 + r3 * 2], xm0, 2
2215     pextrd          [r2 + r5], xm0, 3
2216     RET
2217
2218 ;-----------------------------------------------------------------------------
2219 ; void interp_4tap_vert_pp_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
2220 ;-----------------------------------------------------------------------------
2221 %macro FILTER_V4_W4_H4 2
2222 INIT_XMM sse4
2223 cglobal interp_4tap_vert_pp_%1x%2, 4, 6, 8
2224
2225 mov         r4d,       r4m
2226 sub         r0,        r1
2227
2228 %ifdef PIC
2229 lea         r5,        [tab_ChromaCoeff]
2230 movd        m0,        [r5 + r4 * 4]
2231 %else
2232 movd        m0,        [tab_ChromaCoeff + r4 * 4]
2233 %endif
2234
2235 pshufb      m0,        [tab_Cm]
2236
2237 mova        m1,        [pw_512]
2238
2239 mov         r4d,       %2
2240
2241 lea         r5,        [3 * r1]
2242
2243 .loop:
2244 movd        m2,        [r0]
2245 movd        m3,        [r0 + r1]
2246 movd        m4,        [r0 + 2 * r1]
2247 movd        m5,        [r0 + r5]
2248
2249 punpcklbw   m2,        m3
2250 punpcklbw   m6,        m4,        m5
2251 punpcklbw   m2,        m6
2252
2253 pmaddubsw   m2,        m0
2254
2255 lea         r0,        [r0 + 4 * r1]
2256 movd        m6,        [r0]
2257
2258 punpcklbw   m3,        m4
2259 punpcklbw   m7,        m5,        m6
2260 punpcklbw   m3,        m7
2261
2262 pmaddubsw   m3,        m0
2263
2264 phaddw      m2,        m3
2265
2266 pmulhrsw    m2,        m1
2267
2268 movd        m7,        [r0 + r1]
2269
2270 punpcklbw   m4,        m5
2271 punpcklbw   m3,        m6,        m7
2272 punpcklbw   m4,        m3
2273
2274 pmaddubsw   m4,        m0
2275
2276 movd        m3,        [r0 + 2 * r1]
2277
2278 punpcklbw   m5,        m6
2279 punpcklbw   m7,        m3
2280 punpcklbw   m5,        m7
2281
2282 pmaddubsw   m5,        m0
2283
2284 phaddw      m4,        m5
2285
2286 pmulhrsw    m4,        m1
2287 packuswb    m2,        m4
2288 movd        [r2],      m2
2289 pextrd      [r2 + r3], m2,  1
2290 lea         r2,        [r2 + 2 * r3]
2291 pextrd      [r2],      m2, 2
2292 pextrd      [r2 + r3], m2, 3
2293
2294 lea         r2,        [r2 + 2 * r3]
2295
2296 sub         r4,        4
2297 jnz        .loop
2298 RET
2299 %endmacro
2300
2301 FILTER_V4_W4_H4 4,  8
2302 FILTER_V4_W4_H4 4, 16
2303
2304 FILTER_V4_W4_H4 4, 32
2305
2306 %macro FILTER_V4_W8_H2 0
2307 punpcklbw   m1,        m2
2308 punpcklbw   m7,        m3,        m0
2309
2310 pmaddubsw   m1,        m6
2311 pmaddubsw   m7,        m5
2312
2313 paddw       m1,        m7
2314
2315 pmulhrsw    m1,        m4
2316 packuswb    m1,        m1
2317 %endmacro
2318
2319 %macro FILTER_V4_W8_H3 0
2320 punpcklbw   m2,        m3
2321 punpcklbw   m7,        m0,        m1
2322
2323 pmaddubsw   m2,        m6
2324 pmaddubsw   m7,        m5
2325
2326 paddw       m2,        m7
2327
2328 pmulhrsw    m2,        m4
2329 packuswb    m2,        m2
2330 %endmacro
2331
2332 %macro FILTER_V4_W8_H4 0
2333 punpcklbw   m3,        m0
2334 punpcklbw   m7,        m1,        m2
2335
2336 pmaddubsw   m3,        m6
2337 pmaddubsw   m7,        m5
2338
2339 paddw       m3,        m7
2340
2341 pmulhrsw    m3,        m4
2342 packuswb    m3,        m3
2343 %endmacro
2344
2345 %macro FILTER_V4_W8_H5 0
2346 punpcklbw   m0,        m1
2347 punpcklbw   m7,        m2,        m3
2348
2349 pmaddubsw   m0,        m6
2350 pmaddubsw   m7,        m5
2351
2352 paddw       m0,        m7
2353
2354 pmulhrsw    m0,        m4
2355 packuswb    m0,        m0
2356 %endmacro
2357
2358 %macro FILTER_V4_W8_8x2 2
2359 FILTER_V4_W8 %1, %2
2360 movq        m0,        [r0 + 4 * r1]
2361
2362 FILTER_V4_W8_H2
2363
2364 movh        [r2 + r3], m1
2365 %endmacro
2366
2367 %macro FILTER_V4_W8_8x4 2
2368 FILTER_V4_W8_8x2 %1, %2
2369 ;8x3
2370 lea         r6,        [r0 + 4 * r1]
2371 movq        m1,        [r6 + r1]
2372
2373 FILTER_V4_W8_H3
2374
2375 movh        [r2 + 2 * r3], m2
2376
2377 ;8x4
2378 movq        m2,        [r6 + 2 * r1]
2379
2380 FILTER_V4_W8_H4
2381
2382 lea         r5,        [r2 + 2 * r3]
2383 movh        [r5 + r3], m3
2384 %endmacro
2385
2386 %macro FILTER_V4_W8_8x6 2
2387 FILTER_V4_W8_8x4 %1, %2
2388 ;8x5
2389 lea         r6,        [r6 + 2 * r1]
2390 movq        m3,        [r6 + r1]
2391
2392 FILTER_V4_W8_H5
2393
2394 movh        [r2 + 4 * r3], m0
2395
2396 ;8x6
2397 movq        m0,        [r0 + 8 * r1]
2398
2399 FILTER_V4_W8_H2
2400
2401 lea         r5,        [r2 + 4 * r3]
2402 movh        [r5 + r3], m1
2403 %endmacro
2404
2405 ;-----------------------------------------------------------------------------
2406 ; void interp_4tap_vert_pp_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
2407 ;-----------------------------------------------------------------------------
2408 %macro FILTER_V4_W8 2
2409 INIT_XMM sse4
2410 cglobal interp_4tap_vert_pp_%1x%2, 4, 7, 8
2411
2412 mov         r4d,       r4m
2413
2414 sub         r0,        r1
2415 movq        m0,        [r0]
2416 movq        m1,        [r0 + r1]
2417 movq        m2,        [r0 + 2 * r1]
2418 lea         r5,        [r0 + 2 * r1]
2419 movq        m3,        [r5 + r1]
2420
2421 punpcklbw   m0,        m1
2422 punpcklbw   m4,        m2,          m3
2423
2424 %ifdef PIC
2425 lea         r6,        [tab_ChromaCoeff]
2426 movd        m5,        [r6 + r4 * 4]
2427 %else
2428 movd        m5,        [tab_ChromaCoeff + r4 * 4]
2429 %endif
2430
2431 pshufb      m6,        m5,       [tab_Vm]
2432 pmaddubsw   m0,        m6
2433
2434 pshufb      m5,        [tab_Vm + 16]
2435 pmaddubsw   m4,        m5
2436
2437 paddw       m0,        m4
2438
2439 mova        m4,        [pw_512]
2440
2441 pmulhrsw    m0,        m4
2442 packuswb    m0,        m0
2443 movh        [r2],      m0
2444 %endmacro
2445
2446 ;-----------------------------------------------------------------------------
2447 ; void interp_4tap_vert_pp_8x2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
2448 ;-----------------------------------------------------------------------------
2449 FILTER_V4_W8_8x2 8, 2
2450
2451 RET
2452
2453 ;-----------------------------------------------------------------------------
2454 ; void interp_4tap_vert_pp_8x4(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
2455 ;-----------------------------------------------------------------------------
2456 FILTER_V4_W8_8x4 8, 4
2457
2458 RET
2459
2460 ;-----------------------------------------------------------------------------
2461 ; void interp_4tap_vert_pp_8x6(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
2462 ;-----------------------------------------------------------------------------
2463 FILTER_V4_W8_8x6 8, 6
2464
2465 RET
2466
2467 ;-------------------------------------------------------------------------------------------------------------
2468 ; void interp_4tap_vert_ps_4x2(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
2469 ;-------------------------------------------------------------------------------------------------------------
2470 INIT_XMM sse4
2471 cglobal interp_4tap_vert_ps_4x2, 4, 6, 6
2472
2473 mov         r4d, r4m
2474 sub         r0, r1
2475 add         r3d, r3d
2476
2477 %ifdef PIC
2478 lea         r5, [tab_ChromaCoeff]
2479 movd        m0, [r5 + r4 * 4]
2480 %else
2481 movd        m0, [tab_ChromaCoeff + r4 * 4]
2482 %endif
2483
2484 pshufb      m0, [tab_Cm]
2485
2486 movd        m2, [r0]
2487 movd        m3, [r0 + r1]
2488 lea         r5, [r0 + 2 * r1]
2489 movd        m4, [r5]
2490 movd        m5, [r5 + r1]
2491
2492 punpcklbw   m2, m3
2493 punpcklbw   m1, m4, m5
2494 punpcklbw   m2, m1
2495
2496 pmaddubsw   m2, m0
2497
2498 movd        m1, [r0 + 4 * r1]
2499
2500 punpcklbw   m3, m4
2501 punpcklbw   m5, m1
2502 punpcklbw   m3, m5
2503
2504 pmaddubsw   m3, m0
2505
2506 phaddw      m2, m3
2507
2508 psubw       m2, [pw_2000]
2509 movh        [r2], m2
2510 movhps      [r2 + r3], m2
2511
2512 RET
2513
2514 ;-------------------------------------------------------------------------------------------------------------
2515 ; void interp_4tap_vert_ps_4x4(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
2516 ;-------------------------------------------------------------------------------------------------------------
2517 INIT_XMM sse4
2518 cglobal interp_4tap_vert_ps_4x4, 4, 6, 7
2519
2520     mov        r4d, r4m
2521     sub        r0, r1
2522     add        r3d, r3d
2523
2524 %ifdef PIC
2525     lea        r5, [tab_ChromaCoeff]
2526     movd       m0, [r5 + r4 * 4]
2527 %else
2528     movd       m0, [tab_ChromaCoeff + r4 * 4]
2529 %endif
2530
2531     pshufb     m0, [tab_Cm]
2532
2533     lea        r4, [r1 * 3]
2534     lea        r5, [r0 + 4 * r1]
2535
2536     movd       m2, [r0]
2537     movd       m3, [r0 + r1]
2538     movd       m4, [r0 + 2 * r1]
2539     movd       m5, [r0 + r4]
2540
2541     punpcklbw  m2, m3
2542     punpcklbw  m6, m4, m5
2543     punpcklbw  m2, m6
2544
2545     pmaddubsw  m2, m0
2546
2547     movd       m6, [r5]
2548
2549     punpcklbw  m3, m4
2550     punpcklbw  m1, m5, m6
2551     punpcklbw  m3, m1
2552
2553     pmaddubsw  m3, m0
2554
2555     phaddw     m2, m3
2556
2557     mova       m1, [pw_2000]
2558
2559     psubw      m2, m1
2560     movh       [r2], m2
2561     movhps     [r2 + r3], m2
2562
2563     movd       m2, [r5 + r1]
2564
2565     punpcklbw  m4, m5
2566     punpcklbw  m3, m6, m2
2567     punpcklbw  m4, m3
2568
2569     pmaddubsw  m4, m0
2570
2571     movd       m3, [r5 + 2 * r1]
2572
2573     punpcklbw  m5, m6
2574     punpcklbw  m2, m3
2575     punpcklbw  m5, m2
2576
2577     pmaddubsw  m5, m0
2578
2579     phaddw     m4, m5
2580
2581     psubw      m4, m1
2582     lea        r2, [r2 + 2 * r3]
2583     movh       [r2], m4
2584     movhps     [r2 + r3], m4
2585
2586     RET
2587
2588 ;---------------------------------------------------------------------------------------------------------------
2589 ; void interp_4tap_vert_ps_%1x%2(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
2590 ;---------------------------------------------------------------------------------------------------------------
2591 %macro FILTER_V_PS_W4_H4 2
2592 INIT_XMM sse4
2593 cglobal interp_4tap_vert_ps_%1x%2, 4, 6, 8
2594
2595     mov        r4d, r4m
2596     sub        r0, r1
2597     add        r3d, r3d
2598
2599 %ifdef PIC
2600     lea        r5, [tab_ChromaCoeff]
2601     movd       m0, [r5 + r4 * 4]
2602 %else
2603     movd       m0, [tab_ChromaCoeff + r4 * 4]
2604 %endif
2605
2606     pshufb     m0, [tab_Cm]
2607
2608     mova       m1, [pw_2000]
2609
2610     mov        r4d, %2/4
2611     lea        r5, [3 * r1]
2612
2613 .loop:
2614     movd       m2, [r0]
2615     movd       m3, [r0 + r1]
2616     movd       m4, [r0 + 2 * r1]
2617     movd       m5, [r0 + r5]
2618
2619     punpcklbw  m2, m3
2620     punpcklbw  m6, m4, m5
2621     punpcklbw  m2, m6
2622
2623     pmaddubsw  m2, m0
2624
2625     lea        r0, [r0 + 4 * r1]
2626     movd       m6, [r0]
2627
2628     punpcklbw  m3, m4
2629     punpcklbw  m7, m5, m6
2630     punpcklbw  m3, m7
2631
2632     pmaddubsw  m3, m0
2633
2634     phaddw     m2, m3
2635
2636     psubw      m2, m1
2637     movh       [r2], m2
2638     movhps     [r2 + r3], m2
2639
2640     movd       m2, [r0 + r1]
2641
2642     punpcklbw  m4, m5
2643     punpcklbw  m3, m6, m2
2644     punpcklbw  m4, m3
2645
2646     pmaddubsw  m4, m0
2647
2648     movd       m3, [r0 + 2 * r1]
2649
2650     punpcklbw  m5, m6
2651     punpcklbw  m2, m3
2652     punpcklbw  m5, m2
2653
2654     pmaddubsw  m5, m0
2655
2656     phaddw     m4, m5
2657
2658     psubw      m4, m1
2659     lea        r2, [r2 + 2 * r3]
2660     movh       [r2], m4
2661     movhps     [r2 + r3], m4
2662
2663     lea        r2, [r2 + 2 * r3]
2664
2665     dec        r4d
2666     jnz        .loop
2667     RET
2668 %endmacro
2669
2670 FILTER_V_PS_W4_H4 4, 8
2671 FILTER_V_PS_W4_H4 4, 16
2672
2673 FILTER_V_PS_W4_H4 4, 32
2674
2675 ;--------------------------------------------------------------------------------------------------------------
2676 ; void interp_4tap_vert_ps_8x%2(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
2677 ;--------------------------------------------------------------------------------------------------------------
2678 %macro FILTER_V_PS_W8_H8_H16_H2 2
2679 INIT_XMM sse4
2680 cglobal interp_4tap_vert_ps_%1x%2, 4, 6, 7
2681
2682     mov        r4d, r4m
2683     sub        r0, r1
2684     add        r3d, r3d
2685
2686 %ifdef PIC
2687     lea        r5, [tab_ChromaCoeff]
2688     movd       m5, [r5 + r4 * 4]
2689 %else
2690     movd       m5, [tab_ChromaCoeff + r4 * 4]
2691 %endif
2692
2693     pshufb     m6, m5, [tab_Vm]
2694     pshufb     m5, [tab_Vm + 16]
2695     mova       m4, [pw_2000]
2696
2697     mov        r4d, %2/2
2698     lea        r5, [3 * r1]
2699
2700 .loopH:
2701     movq       m0, [r0]
2702     movq       m1, [r0 + r1]
2703     movq       m2, [r0 + 2 * r1]
2704     movq       m3, [r0 + r5]
2705
2706     punpcklbw  m0, m1
2707     punpcklbw  m1, m2
2708     punpcklbw  m2, m3
2709
2710     pmaddubsw  m0, m6
2711     pmaddubsw  m2, m5
2712
2713     paddw      m0, m2
2714
2715     psubw      m0, m4
2716     movu       [r2], m0
2717
2718     movq       m0, [r0 + 4 * r1]
2719
2720     punpcklbw  m3, m0
2721
2722     pmaddubsw  m1, m6
2723     pmaddubsw  m3, m5
2724
2725     paddw      m1, m3
2726     psubw      m1, m4
2727
2728     movu       [r2 + r3], m1
2729
2730     lea        r0, [r0 + 2 * r1]
2731     lea        r2, [r2 + 2 * r3]
2732
2733     dec        r4d
2734     jnz       .loopH
2735
2736     RET
2737 %endmacro
2738
2739 FILTER_V_PS_W8_H8_H16_H2 8, 2
2740 FILTER_V_PS_W8_H8_H16_H2 8, 4
2741 FILTER_V_PS_W8_H8_H16_H2 8, 6
2742
2743 FILTER_V_PS_W8_H8_H16_H2 8, 12
2744 FILTER_V_PS_W8_H8_H16_H2 8, 64
2745
2746 ;--------------------------------------------------------------------------------------------------------------
2747 ; void interp_4tap_vert_ps_8x%2(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
2748 ;--------------------------------------------------------------------------------------------------------------
2749 %macro FILTER_V_PS_W8_H8_H16_H32 2
2750 INIT_XMM sse4
2751 cglobal interp_4tap_vert_ps_%1x%2, 4, 6, 8
2752
2753     mov        r4d, r4m
2754     sub        r0, r1
2755     add        r3d, r3d
2756
2757 %ifdef PIC
2758     lea        r5, [tab_ChromaCoeff]
2759     movd       m5, [r5 + r4 * 4]
2760 %else
2761     movd       m5, [tab_ChromaCoeff + r4 * 4]
2762 %endif
2763
2764     pshufb     m6, m5, [tab_Vm]
2765     pshufb     m5, [tab_Vm + 16]
2766     mova       m4, [pw_2000]
2767
2768     mov        r4d, %2/4
2769     lea        r5, [3 * r1]
2770
2771 .loop:
2772     movq       m0, [r0]
2773     movq       m1, [r0 + r1]
2774     movq       m2, [r0 + 2 * r1]
2775     movq       m3, [r0 + r5]
2776
2777     punpcklbw  m0, m1
2778     punpcklbw  m1, m2
2779     punpcklbw  m2, m3
2780
2781     pmaddubsw  m0, m6
2782     pmaddubsw  m7, m2, m5
2783
2784     paddw      m0, m7
2785
2786     psubw       m0, m4
2787     movu       [r2], m0
2788
2789     lea        r0, [r0 + 4 * r1]
2790     movq       m0, [r0]
2791
2792     punpcklbw  m3, m0
2793
2794     pmaddubsw  m1, m6
2795     pmaddubsw  m7, m3, m5
2796
2797     paddw      m1, m7
2798
2799     psubw      m1, m4
2800     movu       [r2 + r3], m1
2801
2802     movq       m1, [r0 + r1]
2803
2804     punpcklbw  m0, m1
2805
2806     pmaddubsw  m2, m6
2807     pmaddubsw  m0, m5
2808
2809     paddw      m2, m0
2810
2811     psubw      m2, m4
2812     lea        r2, [r2 + 2 * r3]
2813     movu       [r2], m2
2814
2815     movq       m2, [r0 + 2 * r1]
2816
2817     punpcklbw  m1, m2
2818
2819     pmaddubsw  m3, m6
2820     pmaddubsw  m1, m5
2821
2822     paddw      m3, m1
2823     psubw      m3, m4
2824
2825     movu       [r2 + r3], m3
2826
2827     lea        r2, [r2 + 2 * r3]
2828
2829     dec        r4d
2830     jnz        .loop
2831     RET
2832 %endmacro
2833
2834 FILTER_V_PS_W8_H8_H16_H32 8,  8
2835 FILTER_V_PS_W8_H8_H16_H32 8, 16
2836 FILTER_V_PS_W8_H8_H16_H32 8, 32
2837
2838 ;------------------------------------------------------------------------------------------------------------
2839 ;void interp_4tap_vert_ps_6x8(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
2840 ;------------------------------------------------------------------------------------------------------------
2841 %macro FILTER_V_PS_W6 2
2842 INIT_XMM sse4
2843 cglobal interp_4tap_vert_ps_6x%2, 4, 6, 8
2844
2845     mov        r4d, r4m
2846     sub        r0, r1
2847     add        r3d, r3d
2848
2849 %ifdef PIC
2850     lea        r5, [tab_ChromaCoeff]
2851     movd       m5, [r5 + r4 * 4]
2852 %else
2853     movd       m5, [tab_ChromaCoeff + r4 * 4]
2854 %endif
2855
2856     pshufb     m6, m5, [tab_Vm]
2857     pshufb     m5, [tab_Vm + 16]
2858     mova       m4, [pw_2000]
2859     lea        r5, [3 * r1]
2860     mov        r4d, %2/4
2861
2862 .loop:
2863     movq       m0, [r0]
2864     movq       m1, [r0 + r1]
2865     movq       m2, [r0 + 2 * r1]
2866     movq       m3, [r0 + r5]
2867
2868     punpcklbw  m0, m1
2869     punpcklbw  m1, m2
2870     punpcklbw  m2, m3
2871
2872     pmaddubsw  m0, m6
2873     pmaddubsw  m7, m2, m5
2874
2875     paddw      m0, m7
2876     psubw      m0, m4
2877
2878     movh       [r2], m0
2879     pshufd     m0, m0, 2
2880     movd       [r2 + 8], m0
2881
2882     lea        r0, [r0 + 4 * r1]
2883     movq       m0, [r0]
2884     punpcklbw  m3, m0
2885
2886     pmaddubsw  m1, m6
2887     pmaddubsw  m7, m3, m5
2888
2889     paddw      m1, m7
2890     psubw      m1, m4
2891
2892     movh       [r2 + r3], m1
2893     pshufd     m1, m1, 2
2894     movd       [r2 + r3 + 8], m1
2895
2896     movq       m1, [r0 + r1]
2897     punpcklbw  m0, m1
2898
2899     pmaddubsw  m2, m6
2900     pmaddubsw  m0, m5
2901
2902     paddw      m2, m0
2903     psubw      m2, m4
2904
2905     lea        r2,[r2 + 2 * r3]
2906     movh       [r2], m2
2907     pshufd     m2, m2, 2
2908     movd       [r2 + 8], m2
2909
2910     movq       m2,[r0 + 2 * r1]
2911     punpcklbw  m1, m2
2912
2913     pmaddubsw  m3, m6
2914     pmaddubsw  m1, m5
2915
2916     paddw      m3, m1
2917     psubw      m3, m4
2918
2919     movh       [r2 + r3], m3
2920     pshufd     m3, m3, 2
2921     movd       [r2 + r3 + 8], m3
2922
2923     lea        r2, [r2 + 2 * r3]
2924
2925     dec        r4d
2926     jnz        .loop
2927     RET
2928 %endmacro
2929
2930 FILTER_V_PS_W6 6, 8
2931 FILTER_V_PS_W6 6, 16
2932
2933 ;---------------------------------------------------------------------------------------------------------------
2934 ; void interp_4tap_vert_ps_12x16(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
2935 ;---------------------------------------------------------------------------------------------------------------
2936 %macro FILTER_V_PS_W12 2
2937 INIT_XMM sse4
2938 cglobal interp_4tap_vert_ps_12x%2, 4, 6, 8
2939
2940     mov        r4d, r4m
2941     sub        r0, r1
2942     add        r3d, r3d
2943
2944 %ifdef PIC
2945     lea        r5, [tab_ChromaCoeff]
2946     movd       m0, [r5 + r4 * 4]
2947 %else
2948     movd       m0, [tab_ChromaCoeff + r4 * 4]
2949 %endif
2950
2951     pshufb     m1, m0, [tab_Vm]
2952     pshufb     m0, [tab_Vm + 16]
2953
2954     mov        r4d, %2/2
2955
2956 .loop:
2957     movu       m2, [r0]
2958     movu       m3, [r0 + r1]
2959
2960     punpcklbw  m4, m2, m3
2961     punpckhbw  m2, m3
2962
2963     pmaddubsw  m4, m1
2964     pmaddubsw  m2, m1
2965
2966     lea        r0, [r0 + 2 * r1]
2967     movu       m5, [r0]
2968     movu       m7, [r0 + r1]
2969
2970     punpcklbw  m6, m5, m7
2971     pmaddubsw  m6, m0
2972     paddw      m4, m6
2973
2974     punpckhbw  m6, m5, m7
2975     pmaddubsw  m6, m0
2976     paddw      m2, m6
2977
2978     mova       m6, [pw_2000]
2979
2980     psubw      m4, m6
2981     psubw      m2, m6
2982
2983     movu       [r2], m4
2984     movh       [r2 + 16], m2
2985
2986     punpcklbw  m4, m3, m5
2987     punpckhbw  m3, m5
2988
2989     pmaddubsw  m4, m1
2990     pmaddubsw  m3, m1
2991
2992     movu       m2, [r0 + 2 * r1]
2993
2994     punpcklbw  m5, m7, m2
2995     punpckhbw  m7, m2
2996
2997     pmaddubsw  m5, m0
2998     pmaddubsw  m7, m0
2999
3000     paddw      m4, m5
3001     paddw      m3, m7
3002
3003     psubw      m4, m6
3004     psubw      m3, m6
3005
3006     movu       [r2 + r3], m4
3007     movh       [r2 + r3 + 16], m3
3008
3009     lea        r2, [r2 + 2 * r3]
3010
3011     dec        r4d
3012     jnz        .loop
3013     RET
3014 %endmacro
3015
3016 FILTER_V_PS_W12 12, 16
3017 FILTER_V_PS_W12 12, 32
3018
3019 ;---------------------------------------------------------------------------------------------------------------
3020 ; void interp_4tap_vert_ps_16x%2(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
3021 ;---------------------------------------------------------------------------------------------------------------
3022 %macro FILTER_V_PS_W16 2
3023 INIT_XMM sse4
3024 cglobal interp_4tap_vert_ps_%1x%2, 4, 6, 8
3025
3026     mov        r4d, r4m
3027     sub        r0, r1
3028     add        r3d, r3d
3029
3030 %ifdef PIC
3031     lea        r5, [tab_ChromaCoeff]
3032     movd       m0, [r5 + r4 * 4]
3033 %else
3034     movd       m0, [tab_ChromaCoeff + r4 * 4]
3035 %endif
3036
3037     pshufb     m1, m0, [tab_Vm]
3038     pshufb     m0, [tab_Vm + 16]
3039     mov        r4d, %2/2
3040
3041 .loop:
3042     movu       m2, [r0]
3043     movu       m3, [r0 + r1]
3044
3045     punpcklbw  m4, m2, m3
3046     punpckhbw  m2, m3
3047
3048     pmaddubsw  m4, m1
3049     pmaddubsw  m2, m1
3050
3051     lea        r0, [r0 + 2 * r1]
3052     movu       m5, [r0]
3053     movu       m7, [r0 + r1]
3054
3055     punpcklbw  m6, m5, m7
3056     pmaddubsw  m6, m0
3057     paddw      m4, m6
3058
3059     punpckhbw  m6, m5, m7
3060     pmaddubsw  m6, m0
3061     paddw      m2, m6
3062
3063     mova       m6, [pw_2000]
3064
3065     psubw      m4, m6
3066     psubw      m2, m6
3067
3068     movu       [r2], m4
3069     movu       [r2 + 16], m2
3070
3071     punpcklbw  m4, m3, m5
3072     punpckhbw  m3, m5
3073
3074     pmaddubsw  m4, m1
3075     pmaddubsw  m3, m1
3076
3077     movu       m5, [r0 + 2 * r1]
3078
3079     punpcklbw  m2, m7, m5
3080     punpckhbw  m7, m5
3081
3082     pmaddubsw  m2, m0
3083     pmaddubsw  m7, m0
3084
3085     paddw      m4, m2
3086     paddw      m3, m7
3087
3088     psubw      m4, m6
3089     psubw      m3, m6
3090
3091     movu       [r2 + r3], m4
3092     movu       [r2 + r3 + 16], m3
3093
3094     lea        r2, [r2 + 2 * r3]
3095
3096     dec        r4d
3097     jnz        .loop
3098     RET
3099 %endmacro
3100
3101 FILTER_V_PS_W16 16,  4
3102 FILTER_V_PS_W16 16,  8
3103 FILTER_V_PS_W16 16, 12
3104 FILTER_V_PS_W16 16, 16
3105 FILTER_V_PS_W16 16, 32
3106
3107 FILTER_V_PS_W16 16, 24
3108 FILTER_V_PS_W16 16, 64
3109
3110 ;--------------------------------------------------------------------------------------------------------------
3111 ;void interp_4tap_vert_ps_24x32(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
3112 ;--------------------------------------------------------------------------------------------------------------
3113 %macro FILTER_V4_PS_W24 2
3114 INIT_XMM sse4
3115 cglobal interp_4tap_vert_ps_24x%2, 4, 6, 8
3116
3117     mov        r4d, r4m
3118     sub        r0, r1
3119     add        r3d, r3d
3120
3121 %ifdef PIC
3122     lea        r5, [tab_ChromaCoeff]
3123     movd       m0, [r5 + r4 * 4]
3124 %else
3125     movd       m0, [tab_ChromaCoeff + r4 * 4]
3126 %endif
3127
3128     pshufb     m1, m0, [tab_Vm]
3129     pshufb     m0, [tab_Vm + 16]
3130
3131     mov        r4d, %2/2
3132
3133 .loop:
3134     movu       m2, [r0]
3135     movu       m3, [r0 + r1]
3136
3137     punpcklbw  m4, m2, m3
3138     punpckhbw  m2, m3
3139
3140     pmaddubsw  m4, m1
3141     pmaddubsw  m2, m1
3142
3143     lea        r5, [r0 + 2 * r1]
3144
3145     movu       m5, [r5]
3146     movu       m7, [r5 + r1]
3147
3148     punpcklbw  m6, m5, m7
3149     pmaddubsw  m6, m0
3150     paddw      m4, m6
3151
3152     punpckhbw  m6, m5, m7
3153     pmaddubsw  m6, m0
3154     paddw      m2, m6
3155
3156     mova       m6, [pw_2000]
3157
3158     psubw      m4, m6
3159     psubw      m2, m6
3160
3161     movu       [r2], m4
3162     movu       [r2 + 16], m2
3163
3164     punpcklbw  m4, m3, m5
3165     punpckhbw  m3, m5
3166
3167     pmaddubsw  m4, m1
3168     pmaddubsw  m3, m1
3169
3170     movu       m2, [r5 + 2 * r1]
3171
3172     punpcklbw  m5, m7, m2
3173     punpckhbw  m7, m2
3174
3175     pmaddubsw  m5, m0
3176     pmaddubsw  m7, m0
3177
3178     paddw      m4, m5
3179     paddw      m3, m7
3180
3181     psubw      m4, m6
3182     psubw      m3, m6
3183
3184     movu       [r2 + r3], m4
3185     movu       [r2 + r3 + 16], m3
3186
3187     movq       m2, [r0 + 16]
3188     movq       m3, [r0 + r1 + 16]
3189     movq       m4, [r5 + 16]
3190     movq       m5, [r5 + r1 + 16]
3191
3192     punpcklbw  m2, m3
3193     punpcklbw  m7, m4, m5
3194
3195     pmaddubsw  m2, m1
3196     pmaddubsw  m7, m0
3197
3198     paddw      m2, m7
3199     psubw      m2, m6
3200
3201     movu       [r2 + 32], m2
3202
3203     movq       m2, [r5 + 2 * r1 + 16]
3204
3205     punpcklbw  m3, m4
3206     punpcklbw  m5, m2
3207
3208     pmaddubsw  m3, m1
3209     pmaddubsw  m5, m0
3210
3211     paddw      m3, m5
3212     psubw      m3,  m6
3213
3214     movu       [r2 + r3 + 32], m3
3215
3216     mov        r0, r5
3217     lea        r2, [r2 + 2 * r3]
3218
3219     dec        r4d
3220     jnz        .loop
3221     RET
3222 %endmacro
3223
3224 FILTER_V4_PS_W24 24, 32
3225
3226 FILTER_V4_PS_W24 24, 64
3227
3228 ;---------------------------------------------------------------------------------------------------------------
3229 ; void interp_4tap_vert_ps_32x%2(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
3230 ;---------------------------------------------------------------------------------------------------------------
3231 %macro FILTER_V_PS_W32 2
3232 INIT_XMM sse4
3233 cglobal interp_4tap_vert_ps_%1x%2, 4, 6, 8
3234
3235     mov        r4d, r4m
3236     sub        r0, r1
3237     add        r3d, r3d
3238
3239 %ifdef PIC
3240     lea        r5, [tab_ChromaCoeff]
3241     movd       m0, [r5 + r4 * 4]
3242 %else
3243     movd       m0, [tab_ChromaCoeff + r4 * 4]
3244 %endif
3245
3246     pshufb     m1, m0, [tab_Vm]
3247     pshufb     m0, [tab_Vm + 16]
3248
3249     mova       m7, [pw_2000]
3250
3251     mov        r4d, %2
3252
3253 .loop:
3254     movu       m2, [r0]
3255     movu       m3, [r0 + r1]
3256
3257     punpcklbw  m4, m2, m3
3258     punpckhbw  m2, m3
3259
3260     pmaddubsw  m4, m1
3261     pmaddubsw  m2, m1
3262
3263     lea        r5, [r0 + 2 * r1]
3264     movu       m3, [r5]
3265     movu       m5, [r5 + r1]
3266
3267     punpcklbw  m6, m3, m5
3268     punpckhbw  m3, m5
3269
3270     pmaddubsw  m6, m0
3271     pmaddubsw  m3, m0
3272
3273     paddw      m4, m6
3274     paddw      m2, m3
3275
3276     psubw      m4, m7
3277     psubw      m2, m7
3278
3279     movu       [r2], m4
3280     movu       [r2 + 16], m2
3281
3282     movu       m2, [r0 + 16]
3283     movu       m3, [r0 + r1 + 16]
3284
3285     punpcklbw  m4, m2, m3
3286     punpckhbw  m2, m3
3287
3288     pmaddubsw  m4, m1
3289     pmaddubsw  m2, m1
3290
3291     movu       m3, [r5 + 16]
3292     movu       m5, [r5 + r1 + 16]
3293
3294     punpcklbw  m6, m3, m5
3295     punpckhbw  m3, m5
3296
3297     pmaddubsw  m6, m0
3298     pmaddubsw  m3, m0
3299
3300     paddw      m4, m6
3301     paddw      m2, m3
3302
3303     psubw      m4, m7
3304     psubw      m2, m7
3305
3306     movu       [r2 + 32], m4
3307     movu       [r2 + 48], m2
3308
3309     lea        r0, [r0 + r1]
3310     lea        r2, [r2 + r3]
3311
3312     dec        r4d
3313     jnz        .loop
3314     RET
3315 %endmacro
3316
3317 FILTER_V_PS_W32 32,  8
3318 FILTER_V_PS_W32 32, 16
3319 FILTER_V_PS_W32 32, 24
3320 FILTER_V_PS_W32 32, 32
3321
3322 FILTER_V_PS_W32 32, 48
3323 FILTER_V_PS_W32 32, 64
3324
3325 ;-----------------------------------------------------------------------------
3326 ; void interp_4tap_vert_pp_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
3327 ;-----------------------------------------------------------------------------
3328 %macro FILTER_V4_W8_H8_H16_H32 2
3329 INIT_XMM sse4
3330 cglobal interp_4tap_vert_pp_%1x%2, 4, 6, 8
3331
3332 mov         r4d,       r4m
3333 sub         r0,        r1
3334
3335 %ifdef PIC
3336 lea         r5,        [tab_ChromaCoeff]
3337 movd        m5,        [r5 + r4 * 4]
3338 %else
3339 movd        m5,        [tab_ChromaCoeff + r4 * 4]
3340 %endif
3341
3342 pshufb      m6,        m5,       [tab_Vm]
3343 pshufb      m5,        [tab_Vm + 16]
3344 mova        m4,        [pw_512]
3345 lea         r5,        [r1 * 3]
3346
3347 mov         r4d,       %2
3348
3349 .loop:
3350 movq        m0,        [r0]
3351 movq        m1,        [r0 + r1]
3352 movq        m2,        [r0 + 2 * r1]
3353 movq        m3,        [r0 + r5]
3354
3355 punpcklbw   m0,        m1
3356 punpcklbw   m1,        m2
3357 punpcklbw   m2,        m3
3358
3359 pmaddubsw   m0,        m6
3360 pmaddubsw   m7,        m2, m5
3361
3362 paddw       m0,        m7
3363
3364 pmulhrsw    m0,        m4
3365 packuswb    m0,        m0
3366 movh        [r2],      m0
3367
3368 lea         r0,        [r0 + 4 * r1]
3369 movq        m0,        [r0]
3370
3371 punpcklbw   m3,        m0
3372
3373 pmaddubsw   m1,        m6
3374 pmaddubsw   m7,        m3, m5
3375
3376 paddw       m1,        m7
3377
3378 pmulhrsw    m1,        m4
3379 packuswb    m1,        m1
3380 movh        [r2 + r3], m1
3381
3382 movq        m1,        [r0 + r1]
3383
3384 punpcklbw   m0,        m1
3385
3386 pmaddubsw   m2,        m6
3387 pmaddubsw   m0,        m5
3388
3389 paddw       m2,        m0
3390
3391 pmulhrsw    m2,        m4
3392
3393 movq        m7,        [r0 + 2 * r1]
3394 punpcklbw   m1,        m7
3395
3396 pmaddubsw   m3,        m6
3397 pmaddubsw   m1,        m5
3398
3399 paddw       m3,        m1
3400
3401 pmulhrsw    m3,        m4
3402 packuswb    m2,        m3
3403
3404 lea         r2,        [r2 + 2 * r3]
3405 movh        [r2],      m2
3406 movhps      [r2 + r3], m2
3407
3408 lea         r2,        [r2 + 2 * r3]
3409
3410 sub         r4,         4
3411 jnz        .loop
3412 RET
3413 %endmacro
3414
3415 FILTER_V4_W8_H8_H16_H32 8,  8
3416 FILTER_V4_W8_H8_H16_H32 8, 16
3417 FILTER_V4_W8_H8_H16_H32 8, 32
3418
3419 FILTER_V4_W8_H8_H16_H32 8, 12
3420 FILTER_V4_W8_H8_H16_H32 8, 64
3421
3422 %macro PROCESS_CHROMA_AVX2_W8_8R 0
3423     movq            xm1, [r0]                       ; m1 = row 0
3424     movq            xm2, [r0 + r1]                  ; m2 = row 1
3425     punpcklbw       xm1, xm2                        ; m1 = [17 07 16 06 15 05 14 04 13 03 12 02 11 01 10 00]
3426     movq            xm3, [r0 + r1 * 2]              ; m3 = row 2
3427     punpcklbw       xm2, xm3                        ; m2 = [27 17 26 16 25 15 24 14 23 13 22 12 21 11 20 10]
3428     vinserti128     m5, m1, xm2, 1                  ; m5 = [27 17 26 16 25 15 24 14 23 13 22 12 21 11 20 10] - [17 07 16 06 15 05 14 04 13 03 12 02 11 01 10 00]
3429     pmaddubsw       m5, [r5]
3430     movq            xm4, [r0 + r4]                  ; m4 = row 3
3431     punpcklbw       xm3, xm4                        ; m3 = [37 27 36 26 35 25 34 24 33 23 32 22 31 21 30 20]
3432     lea             r0, [r0 + r1 * 4]
3433     movq            xm1, [r0]                       ; m1 = row 4
3434     punpcklbw       xm4, xm1                        ; m4 = [47 37 46 36 45 35 44 34 43 33 42 32 41 31 40 30]
3435     vinserti128     m2, m3, xm4, 1                  ; m2 = [47 37 46 36 45 35 44 34 43 33 42 32 41 31 40 30] - [37 27 36 26 35 25 34 24 33 23 32 22 31 21 30 20]
3436     pmaddubsw       m0, m2, [r5 + 1 * mmsize]
3437     paddw           m5, m0
3438     pmaddubsw       m2, [r5]
3439     movq            xm3, [r0 + r1]                  ; m3 = row 5
3440     punpcklbw       xm1, xm3                        ; m1 = [57 47 56 46 55 45 54 44 53 43 52 42 51 41 50 40]
3441     movq            xm4, [r0 + r1 * 2]              ; m4 = row 6
3442     punpcklbw       xm3, xm4                        ; m3 = [67 57 66 56 65 55 64 54 63 53 62 52 61 51 60 50]
3443     vinserti128     m1, m1, xm3, 1                  ; m1 = [67 57 66 56 65 55 64 54 63 53 62 52 61 51 60 50] - [57 47 56 46 55 45 54 44 53 43 52 42 51 41 50 40]
3444     pmaddubsw       m0, m1, [r5 + 1 * mmsize]
3445     paddw           m2, m0
3446     pmaddubsw       m1, [r5]
3447     movq            xm3, [r0 + r4]                  ; m3 = row 7
3448     punpcklbw       xm4, xm3                        ; m4 = [77 67 76 66 75 65 74 64 73 63 72 62 71 61 70 60]
3449     lea             r0, [r0 + r1 * 4]
3450     movq            xm0, [r0]                       ; m0 = row 8
3451     punpcklbw       xm3, xm0                        ; m3 = [87 77 86 76 85 75 84 74 83 73 82 72 81 71 80 70]
3452     vinserti128     m4, m4, xm3, 1                  ; m4 = [87 77 86 76 85 75 84 74 83 73 82 72 81 71 80 70] - [77 67 76 66 75 65 74 64 73 63 72 62 71 61 70 60]
3453     pmaddubsw       m3, m4, [r5 + 1 * mmsize]
3454     paddw           m1, m3
3455     pmaddubsw       m4, [r5]
3456     movq            xm3, [r0 + r1]                  ; m3 = row 9
3457     punpcklbw       xm0, xm3                        ; m0 = [97 87 96 86 95 85 94 84 93 83 92 82 91 81 90 80]
3458     movq            xm6, [r0 + r1 * 2]              ; m6 = row 10
3459     punpcklbw       xm3, xm6                        ; m3 = [A7 97 A6 96 A5 95 A4 94 A3 93 A2 92 A1 91 A0 90]
3460     vinserti128     m0, m0, xm3, 1                  ; m0 = [A7 97 A6 96 A5 95 A4 94 A3 93 A2 92 A1 91 A0 90] - [97 87 96 86 95 85 94 84 93 83 92 82 91 81 90 80]
3461     pmaddubsw       m0, [r5 + 1 * mmsize]
3462     paddw           m4, m0
3463 %endmacro
3464
3465 INIT_YMM avx2
3466 cglobal interp_4tap_vert_pp_8x8, 4, 6, 7
3467     mov             r4d, r4m
3468     shl             r4d, 6
3469
3470 %ifdef PIC
3471     lea             r5, [tab_ChromaCoeffVer_32]
3472     add             r5, r4
3473 %else
3474     lea             r5, [tab_ChromaCoeffVer_32 + r4]
3475 %endif
3476
3477     lea             r4, [r1 * 3]
3478     sub             r0, r1
3479     PROCESS_CHROMA_AVX2_W8_8R
3480     lea             r4, [r3 * 3]
3481     mova            m3, [pw_512]
3482     pmulhrsw        m5, m3                          ; m5 = word: row 0, row 1
3483     pmulhrsw        m2, m3                          ; m2 = word: row 2, row 3
3484     pmulhrsw        m1, m3                          ; m1 = word: row 4, row 5
3485     pmulhrsw        m4, m3                          ; m4 = word: row 6, row 7
3486     packuswb        m5, m2
3487     packuswb        m1, m4
3488     vextracti128    xm2, m5, 1
3489     vextracti128    xm4, m1, 1
3490     movq            [r2], xm5
3491     movq            [r2 + r3], xm2
3492     movhps          [r2 + r3 * 2], xm5
3493     movhps          [r2 + r4], xm2
3494     lea             r2, [r2 + r3 * 4]
3495     movq            [r2], xm1
3496     movq            [r2 + r3], xm4
3497     movhps          [r2 + r3 * 2], xm1
3498     movhps          [r2 + r4], xm4
3499     RET
3500
3501 ;-----------------------------------------------------------------------------
3502 ;void interp_4tap_vert_pp_6x8(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
3503 ;-----------------------------------------------------------------------------
3504 %macro FILTER_V4_W6_H4 2
3505 INIT_XMM sse4
3506 cglobal interp_4tap_vert_pp_6x%2, 4, 6, 8
3507
3508 mov         r4d,       r4m
3509 sub         r0,        r1
3510
3511 %ifdef PIC
3512 lea         r5,        [tab_ChromaCoeff]
3513 movd        m5,        [r5 + r4 * 4]
3514 %else
3515 movd        m5,        [tab_ChromaCoeff + r4 * 4]
3516 %endif
3517
3518 pshufb      m6,        m5,       [tab_Vm]
3519 pshufb      m5,        [tab_Vm + 16]
3520 mova        m4,        [pw_512]
3521
3522 mov         r4d,       %2
3523 lea         r5,        [3 * r1]
3524
3525 .loop:
3526 movq        m0,        [r0]
3527 movq        m1,        [r0 + r1]
3528 movq        m2,        [r0 + 2 * r1]
3529 movq        m3,        [r0 + r5]
3530
3531 punpcklbw   m0,        m1
3532 punpcklbw   m1,        m2
3533 punpcklbw   m2,        m3
3534
3535 pmaddubsw   m0,        m6
3536 pmaddubsw   m7,        m2, m5
3537
3538 paddw       m0,        m7
3539
3540 pmulhrsw    m0,        m4
3541 packuswb    m0,        m0
3542 movd        [r2],      m0
3543 pextrw      [r2 + 4],  m0,    2
3544
3545 lea         r0,        [r0 + 4 * r1]
3546
3547 movq        m0,        [r0]
3548 punpcklbw   m3,        m0
3549
3550 pmaddubsw   m1,        m6
3551 pmaddubsw   m7,        m3, m5
3552
3553 paddw       m1,        m7
3554
3555 pmulhrsw    m1,        m4
3556 packuswb    m1,        m1
3557 movd        [r2 + r3],      m1
3558 pextrw      [r2 + r3 + 4],  m1,    2
3559
3560 movq        m1,        [r0 + r1]
3561 punpcklbw   m7,        m0,        m1
3562
3563 pmaddubsw   m2,        m6
3564 pmaddubsw   m7,        m5
3565
3566 paddw       m2,        m7
3567
3568 pmulhrsw    m2,        m4
3569 packuswb    m2,        m2
3570 lea         r2,        [r2 + 2 * r3]
3571 movd        [r2],      m2
3572 pextrw      [r2 + 4],  m2,    2
3573
3574 movq        m2,        [r0 + 2 * r1]
3575 punpcklbw   m1,        m2
3576
3577 pmaddubsw   m3,        m6
3578 pmaddubsw   m1,        m5
3579
3580 paddw       m3,        m1
3581
3582 pmulhrsw    m3,        m4
3583 packuswb    m3,        m3
3584
3585 movd        [r2 + r3],        m3
3586 pextrw      [r2 + r3 + 4],    m3,    2
3587
3588 lea         r2,        [r2 + 2 * r3]
3589
3590 sub         r4,         4
3591 jnz        .loop
3592 RET
3593 %endmacro
3594
3595 FILTER_V4_W6_H4 6, 8
3596
3597 FILTER_V4_W6_H4 6, 16
3598
3599 ;-----------------------------------------------------------------------------
3600 ; void interp_4tap_vert_pp_12x16(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
3601 ;-----------------------------------------------------------------------------
3602 %macro FILTER_V4_W12_H2 2
3603 INIT_XMM sse4
3604 cglobal interp_4tap_vert_pp_12x%2, 4, 6, 8
3605
3606 mov         r4d,       r4m
3607 sub         r0,        r1
3608
3609 %ifdef PIC
3610 lea         r5,        [tab_ChromaCoeff]
3611 movd        m0,        [r5 + r4 * 4]
3612 %else
3613 movd        m0,        [tab_ChromaCoeff + r4 * 4]
3614 %endif
3615
3616 pshufb      m1,        m0,       [tab_Vm]
3617 pshufb      m0,        [tab_Vm + 16]
3618
3619 mov         r4d,       %2
3620
3621 .loop:
3622 movu        m2,        [r0]
3623 movu        m3,        [r0 + r1]
3624
3625 punpcklbw   m4,        m2,        m3
3626 punpckhbw   m2,        m3
3627
3628 pmaddubsw   m4,        m1
3629 pmaddubsw   m2,        m1
3630
3631 lea         r0,        [r0 + 2 * r1]
3632 movu        m5,        [r0]
3633 movu        m7,        [r0 + r1]
3634
3635 punpcklbw   m6,        m5,        m7
3636 pmaddubsw   m6,        m0
3637 paddw       m4,        m6
3638
3639 punpckhbw   m6,        m5,        m7
3640 pmaddubsw   m6,        m0
3641 paddw       m2,        m6
3642
3643 mova        m6,        [pw_512]
3644
3645 pmulhrsw    m4,        m6
3646 pmulhrsw    m2,        m6
3647
3648 packuswb    m4,        m2
3649
3650 movh         [r2],     m4
3651 pextrd       [r2 + 8], m4,  2
3652
3653 punpcklbw   m4,        m3,        m5
3654 punpckhbw   m3,        m5
3655
3656 pmaddubsw   m4,        m1
3657 pmaddubsw   m3,        m1
3658
3659 movu        m5,        [r0 + 2 * r1]
3660
3661 punpcklbw   m2,        m7,        m5
3662 punpckhbw   m7,        m5
3663
3664 pmaddubsw   m2,        m0
3665 pmaddubsw   m7,        m0
3666
3667 paddw       m4,        m2
3668 paddw       m3,        m7
3669
3670 pmulhrsw    m4,        m6
3671 pmulhrsw    m3,        m6
3672
3673 packuswb    m4,        m3
3674
3675 movh        [r2 + r3],      m4
3676 pextrd      [r2 + r3 + 8],  m4,  2
3677
3678 lea         r2,        [r2 + 2 * r3]
3679
3680 sub         r4,        2
3681 jnz        .loop
3682 RET
3683 %endmacro
3684
3685 FILTER_V4_W12_H2 12, 16
3686
3687 FILTER_V4_W12_H2 12, 32
3688
3689 ;-----------------------------------------------------------------------------
3690 ; void interp_4tap_vert_pp_16x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
3691 ;-----------------------------------------------------------------------------
3692 %macro FILTER_V4_W16_H2 2
3693 INIT_XMM sse4
3694 cglobal interp_4tap_vert_pp_16x%2, 4, 6, 8
3695
3696 mov         r4d,       r4m
3697 sub         r0,        r1
3698
3699 %ifdef PIC
3700 lea         r5,        [tab_ChromaCoeff]
3701 movd        m0,        [r5 + r4 * 4]
3702 %else
3703 movd        m0,        [tab_ChromaCoeff + r4 * 4]
3704 %endif
3705
3706 pshufb      m1,        m0,       [tab_Vm]
3707 pshufb      m0,        [tab_Vm + 16]
3708
3709 mov         r4d,       %2/2
3710
3711 .loop:
3712 movu        m2,        [r0]
3713 movu        m3,        [r0 + r1]
3714
3715 punpcklbw   m4,        m2,        m3
3716 punpckhbw   m2,        m3
3717
3718 pmaddubsw   m4,        m1
3719 pmaddubsw   m2,        m1
3720
3721 lea         r0,        [r0 + 2 * r1]
3722 movu        m5,        [r0]
3723 movu        m6,        [r0 + r1]
3724
3725 punpckhbw   m7,        m5,        m6
3726 pmaddubsw   m7,        m0
3727 paddw       m2,        m7
3728
3729 punpcklbw   m7,        m5,        m6
3730 pmaddubsw   m7,        m0
3731 paddw       m4,        m7
3732
3733 mova        m7,        [pw_512]
3734
3735 pmulhrsw    m4,        m7
3736 pmulhrsw    m2,        m7
3737
3738 packuswb    m4,        m2
3739
3740 movu        [r2],      m4
3741
3742 punpcklbw   m4,        m3,        m5
3743 punpckhbw   m3,        m5
3744
3745 pmaddubsw   m4,        m1
3746 pmaddubsw   m3,        m1
3747
3748 movu        m5,        [r0 + 2 * r1]
3749
3750 punpcklbw   m2,        m6,        m5
3751 punpckhbw   m6,        m5
3752
3753 pmaddubsw   m2,        m0
3754 pmaddubsw   m6,        m0
3755
3756 paddw       m4,        m2
3757 paddw       m3,        m6
3758
3759 pmulhrsw    m4,        m7
3760 pmulhrsw    m3,        m7
3761
3762 packuswb    m4,        m3
3763
3764 movu        [r2 + r3],      m4
3765
3766 lea         r2,        [r2 + 2 * r3]
3767
3768 dec         r4d
3769 jnz        .loop
3770 RET
3771 %endmacro
3772
3773 FILTER_V4_W16_H2 16,  4
3774 FILTER_V4_W16_H2 16,  8
3775 FILTER_V4_W16_H2 16, 12
3776 FILTER_V4_W16_H2 16, 16
3777 FILTER_V4_W16_H2 16, 32
3778
3779 FILTER_V4_W16_H2 16, 24
3780 FILTER_V4_W16_H2 16, 64
3781
3782 INIT_YMM avx2
3783 %if ARCH_X86_64 == 1
3784 cglobal interp_4tap_vert_pp_16x16, 4, 6, 15
3785     mov             r4d, r4m
3786     shl             r4d, 6
3787
3788 %ifdef PIC
3789     lea             r5, [tab_ChromaCoeffVer_32]
3790     add             r5, r4
3791 %else
3792     lea             r5, [tab_ChromaCoeffVer_32 + r4]
3793 %endif
3794
3795     mova            m12, [r5]
3796     mova            m13, [r5 + mmsize]
3797     lea             r4, [r1 * 3]
3798     sub             r0, r1
3799     lea             r5, [r3 * 3]
3800     mova            m14, [pw_512]
3801
3802     movu            xm0, [r0]                       ; m0 = row 0
3803     movu            xm1, [r0 + r1]                  ; m1 = row 1
3804     punpckhbw       xm2, xm0, xm1
3805     punpcklbw       xm0, xm1
3806     vinserti128     m0, m0, xm2, 1
3807     pmaddubsw       m0, m12
3808     movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
3809     punpckhbw       xm3, xm1, xm2
3810     punpcklbw       xm1, xm2
3811     vinserti128     m1, m1, xm3, 1
3812     pmaddubsw       m1, m12
3813     movu            xm3, [r0 + r4]                  ; m3 = row 3
3814     punpckhbw       xm4, xm2, xm3
3815     punpcklbw       xm2, xm3
3816     vinserti128     m2, m2, xm4, 1
3817     pmaddubsw       m4, m2, m13
3818     paddw           m0, m4
3819     pmaddubsw       m2, m12
3820     lea             r0, [r0 + r1 * 4]
3821     movu            xm4, [r0]                       ; m4 = row 4
3822     punpckhbw       xm5, xm3, xm4
3823     punpcklbw       xm3, xm4
3824     vinserti128     m3, m3, xm5, 1
3825     pmaddubsw       m5, m3, m13
3826     paddw           m1, m5
3827     pmaddubsw       m3, m12
3828     movu            xm5, [r0 + r1]                  ; m5 = row 5
3829     punpckhbw       xm6, xm4, xm5
3830     punpcklbw       xm4, xm5
3831     vinserti128     m4, m4, xm6, 1
3832     pmaddubsw       m6, m4, m13
3833     paddw           m2, m6
3834     pmaddubsw       m4, m12
3835     movu            xm6, [r0 + r1 * 2]              ; m6 = row 6
3836     punpckhbw       xm7, xm5, xm6
3837     punpcklbw       xm5, xm6
3838     vinserti128     m5, m5, xm7, 1
3839     pmaddubsw       m7, m5, m13
3840     paddw           m3, m7
3841     pmaddubsw       m5, m12
3842     movu            xm7, [r0 + r4]                  ; m7 = row 7
3843     punpckhbw       xm8, xm6, xm7
3844     punpcklbw       xm6, xm7
3845     vinserti128     m6, m6, xm8, 1
3846     pmaddubsw       m8, m6, m13
3847     paddw           m4, m8
3848     pmaddubsw       m6, m12
3849     lea             r0, [r0 + r1 * 4]
3850     movu            xm8, [r0]                       ; m8 = row 8
3851     punpckhbw       xm9, xm7, xm8
3852     punpcklbw       xm7, xm8
3853     vinserti128     m7, m7, xm9, 1
3854     pmaddubsw       m9, m7, m13
3855     paddw           m5, m9
3856     pmaddubsw       m7, m12
3857     movu            xm9, [r0 + r1]                  ; m9 = row 9
3858     punpckhbw       xm10, xm8, xm9
3859     punpcklbw       xm8, xm9
3860     vinserti128     m8, m8, xm10, 1
3861     pmaddubsw       m10, m8, m13
3862     paddw           m6, m10
3863     pmaddubsw       m8, m12
3864     movu            xm10, [r0 + r1 * 2]             ; m10 = row 10
3865     punpckhbw       xm11, xm9, xm10
3866     punpcklbw       xm9, xm10
3867     vinserti128     m9, m9, xm11, 1
3868     pmaddubsw       m11, m9, m13
3869     paddw           m7, m11
3870     pmaddubsw       m9, m12
3871
3872     pmulhrsw        m0, m14                         ; m0 = word: row 0
3873     pmulhrsw        m1, m14                         ; m1 = word: row 1
3874     pmulhrsw        m2, m14                         ; m2 = word: row 2
3875     pmulhrsw        m3, m14                         ; m3 = word: row 3
3876     pmulhrsw        m4, m14                         ; m4 = word: row 4
3877     pmulhrsw        m5, m14                         ; m5 = word: row 5
3878     pmulhrsw        m6, m14                         ; m6 = word: row 6
3879     pmulhrsw        m7, m14                         ; m7 = word: row 7
3880     packuswb        m0, m1
3881     packuswb        m2, m3
3882     packuswb        m4, m5
3883     packuswb        m6, m7
3884     vpermq          m0, m0, 11011000b
3885     vpermq          m2, m2, 11011000b
3886     vpermq          m4, m4, 11011000b
3887     vpermq          m6, m6, 11011000b
3888     vextracti128    xm1, m0, 1
3889     vextracti128    xm3, m2, 1
3890     vextracti128    xm5, m4, 1
3891     vextracti128    xm7, m6, 1
3892     movu            [r2], xm0
3893     movu            [r2 + r3], xm1
3894     movu            [r2 + r3 * 2], xm2
3895     movu            [r2 + r5], xm3
3896     lea             r2, [r2 + r3 * 4]
3897     movu            [r2], xm4
3898     movu            [r2 + r3], xm5
3899     movu            [r2 + r3 * 2], xm6
3900     movu            [r2 + r5], xm7
3901     lea             r2, [r2 + r3 * 4]
3902
3903     movu            xm11, [r0 + r4]                 ; m11 = row 11
3904     punpckhbw       xm6, xm10, xm11
3905     punpcklbw       xm10, xm11
3906     vinserti128     m10, m10, xm6, 1
3907     pmaddubsw       m6, m10, m13
3908     paddw           m8, m6
3909     pmaddubsw       m10, m12
3910     lea             r0, [r0 + r1 * 4]
3911     movu            xm6, [r0]                       ; m6 = row 12
3912     punpckhbw       xm7, xm11, xm6
3913     punpcklbw       xm11, xm6
3914     vinserti128     m11, m11, xm7, 1
3915     pmaddubsw       m7, m11, m13
3916     paddw           m9, m7
3917     pmaddubsw       m11, m12
3918
3919     movu            xm7, [r0 + r1]                  ; m7 = row 13
3920     punpckhbw       xm0, xm6, xm7
3921     punpcklbw       xm6, xm7
3922     vinserti128     m6, m6, xm0, 1
3923     pmaddubsw       m0, m6, m13
3924     paddw           m10, m0
3925     pmaddubsw       m6, m12
3926     movu            xm0, [r0 + r1 * 2]              ; m0 = row 14
3927     punpckhbw       xm1, xm7, xm0
3928     punpcklbw       xm7, xm0
3929     vinserti128     m7, m7, xm1, 1
3930     pmaddubsw       m1, m7, m13
3931     paddw           m11, m1
3932     pmaddubsw       m7, m12
3933     movu            xm1, [r0 + r4]                  ; m1 = row 15
3934     punpckhbw       xm2, xm0, xm1
3935     punpcklbw       xm0, xm1
3936     vinserti128     m0, m0, xm2, 1
3937     pmaddubsw       m2, m0, m13
3938     paddw           m6, m2
3939     pmaddubsw       m0, m12
3940     lea             r0, [r0 + r1 * 4]
3941     movu            xm2, [r0]                       ; m2 = row 16
3942     punpckhbw       xm3, xm1, xm2
3943     punpcklbw       xm1, xm2
3944     vinserti128     m1, m1, xm3, 1
3945     pmaddubsw       m3, m1, m13
3946     paddw           m7, m3
3947     pmaddubsw       m1, m12
3948     movu            xm3, [r0 + r1]                  ; m3 = row 17
3949     punpckhbw       xm4, xm2, xm3
3950     punpcklbw       xm2, xm3
3951     vinserti128     m2, m2, xm4, 1
3952     pmaddubsw       m2, m13
3953     paddw           m0, m2
3954     movu            xm4, [r0 + r1 * 2]              ; m4 = row 18
3955     punpckhbw       xm5, xm3, xm4
3956     punpcklbw       xm3, xm4
3957     vinserti128     m3, m3, xm5, 1
3958     pmaddubsw       m3, m13
3959     paddw           m1, m3
3960
3961     pmulhrsw        m8, m14                         ; m8 = word: row 8
3962     pmulhrsw        m9, m14                         ; m9 = word: row 9
3963     pmulhrsw        m10, m14                        ; m10 = word: row 10
3964     pmulhrsw        m11, m14                        ; m11 = word: row 11
3965     pmulhrsw        m6, m14                         ; m6 = word: row 12
3966     pmulhrsw        m7, m14                         ; m7 = word: row 13
3967     pmulhrsw        m0, m14                         ; m0 = word: row 14
3968     pmulhrsw        m1, m14                         ; m1 = word: row 15
3969     packuswb        m8, m9
3970     packuswb        m10, m11
3971     packuswb        m6, m7
3972     packuswb        m0, m1
3973     vpermq          m8, m8, 11011000b
3974     vpermq          m10, m10, 11011000b
3975     vpermq          m6, m6, 11011000b
3976     vpermq          m0, m0, 11011000b
3977     vextracti128    xm9, m8, 1
3978     vextracti128    xm11, m10, 1
3979     vextracti128    xm7, m6, 1
3980     vextracti128    xm1, m0, 1
3981     movu            [r2], xm8
3982     movu            [r2 + r3], xm9
3983     movu            [r2 + r3 * 2], xm10
3984     movu            [r2 + r5], xm11
3985     lea             r2, [r2 + r3 * 4]
3986     movu            [r2], xm6
3987     movu            [r2 + r3], xm7
3988     movu            [r2 + r3 * 2], xm0
3989     movu            [r2 + r5], xm1
3990     RET
3991 %endif
3992
3993 ;-----------------------------------------------------------------------------
3994 ;void interp_4tap_vert_pp_24x32(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
3995 ;-----------------------------------------------------------------------------
3996 %macro FILTER_V4_W24 2
3997 INIT_XMM sse4
3998 cglobal interp_4tap_vert_pp_24x%2, 4, 6, 8
3999
4000 mov         r4d,       r4m
4001 sub         r0,        r1
4002
4003 %ifdef PIC
4004 lea         r5,        [tab_ChromaCoeff]
4005 movd        m0,        [r5 + r4 * 4]
4006 %else
4007 movd        m0,        [tab_ChromaCoeff + r4 * 4]
4008 %endif
4009
4010 pshufb      m1,        m0,       [tab_Vm]
4011 pshufb      m0,        [tab_Vm + 16]
4012
4013 mov         r4d,       %2
4014
4015 .loop:
4016 movu        m2,        [r0]
4017 movu        m3,        [r0 + r1]
4018
4019 punpcklbw   m4,        m2,        m3
4020 punpckhbw   m2,        m3
4021
4022 pmaddubsw   m4,        m1
4023 pmaddubsw   m2,        m1
4024
4025 lea         r5,        [r0 + 2 * r1]
4026 movu        m5,        [r5]
4027 movu        m7,        [r5 + r1]
4028
4029 punpcklbw   m6,        m5,        m7
4030 pmaddubsw   m6,        m0
4031 paddw       m4,        m6
4032
4033 punpckhbw   m6,        m5,        m7
4034 pmaddubsw   m6,        m0
4035 paddw       m2,        m6
4036
4037 mova        m6,        [pw_512]
4038
4039 pmulhrsw    m4,        m6
4040 pmulhrsw    m2,        m6
4041
4042 packuswb    m4,        m2
4043
4044 movu        [r2],      m4
4045
4046 punpcklbw   m4,        m3,        m5
4047 punpckhbw   m3,        m5
4048
4049 pmaddubsw   m4,        m1
4050 pmaddubsw   m3,        m1
4051
4052 movu        m2,        [r5 + 2 * r1]
4053
4054 punpcklbw   m5,        m7,        m2
4055 punpckhbw   m7,        m2
4056
4057 pmaddubsw   m5,        m0
4058 pmaddubsw   m7,        m0
4059
4060 paddw       m4,        m5
4061 paddw       m3,        m7
4062
4063 pmulhrsw    m4,        m6
4064 pmulhrsw    m3,        m6
4065
4066 packuswb    m4,        m3
4067
4068 movu        [r2 + r3],      m4
4069
4070 movq        m2,        [r0 + 16]
4071 movq        m3,        [r0 + r1 + 16]
4072 movq        m4,        [r5 + 16]
4073 movq        m5,        [r5 + r1 + 16]
4074
4075 punpcklbw   m2,        m3
4076 punpcklbw   m4,        m5
4077
4078 pmaddubsw   m2,        m1
4079 pmaddubsw   m4,        m0
4080
4081 paddw       m2,        m4
4082
4083 pmulhrsw    m2,        m6
4084
4085 movq        m3,        [r0 + r1 + 16]
4086 movq        m4,        [r5 + 16]
4087 movq        m5,        [r5 + r1 + 16]
4088 movq        m7,        [r5 + 2 * r1 + 16]
4089
4090 punpcklbw   m3,        m4
4091 punpcklbw   m5,        m7
4092
4093 pmaddubsw   m3,        m1
4094 pmaddubsw   m5,        m0
4095
4096 paddw       m3,        m5
4097
4098 pmulhrsw    m3,        m6
4099 packuswb    m2,        m3
4100
4101 movh        [r2 + 16], m2
4102 movhps      [r2 + r3 + 16], m2
4103
4104 mov         r0,        r5
4105 lea         r2,        [r2 + 2 * r3]
4106
4107 sub         r4,        2
4108 jnz        .loop
4109 RET
4110 %endmacro
4111
4112 FILTER_V4_W24 24, 32
4113
4114 FILTER_V4_W24 24, 64
4115
4116 ;-----------------------------------------------------------------------------
4117 ; void interp_4tap_vert_pp_32x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
4118 ;-----------------------------------------------------------------------------
4119 %macro FILTER_V4_W32 2
4120 INIT_XMM sse4
4121 cglobal interp_4tap_vert_pp_%1x%2, 4, 6, 8
4122
4123 mov         r4d,       r4m
4124 sub         r0,        r1
4125
4126 %ifdef PIC
4127 lea         r5,        [tab_ChromaCoeff]
4128 movd        m0,        [r5 + r4 * 4]
4129 %else
4130 movd        m0,        [tab_ChromaCoeff + r4 * 4]
4131 %endif
4132
4133 pshufb      m1,        m0,       [tab_Vm]
4134 pshufb      m0,        [tab_Vm + 16]
4135
4136 mova        m7,        [pw_512]
4137
4138 mov         r4d,       %2
4139
4140 .loop:
4141 movu        m2,        [r0]
4142 movu        m3,        [r0 + r1]
4143
4144 punpcklbw   m4,        m2,        m3
4145 punpckhbw   m2,        m3
4146
4147 pmaddubsw   m4,        m1
4148 pmaddubsw   m2,        m1
4149
4150 lea         r5,        [r0 + 2 * r1]
4151 movu        m3,        [r5]
4152 movu        m5,        [r5 + r1]
4153
4154 punpcklbw   m6,        m3,        m5
4155 punpckhbw   m3,        m5
4156
4157 pmaddubsw   m6,        m0
4158 pmaddubsw   m3,        m0
4159
4160 paddw       m4,        m6
4161 paddw       m2,        m3
4162
4163 pmulhrsw    m4,        m7
4164 pmulhrsw    m2,        m7
4165
4166 packuswb    m4,        m2
4167
4168 movu        [r2],      m4
4169
4170 movu        m2,        [r0 + 16]
4171 movu        m3,        [r0 + r1 + 16]
4172
4173 punpcklbw   m4,        m2,        m3
4174 punpckhbw   m2,        m3
4175
4176 pmaddubsw   m4,        m1
4177 pmaddubsw   m2,        m1
4178
4179 movu        m3,        [r5 + 16]
4180 movu        m5,        [r5 + r1 + 16]
4181
4182 punpcklbw   m6,        m3,        m5
4183 punpckhbw   m3,        m5
4184
4185 pmaddubsw   m6,        m0
4186 pmaddubsw   m3,        m0
4187
4188 paddw       m4,        m6
4189 paddw       m2,        m3
4190
4191 pmulhrsw    m4,        m7
4192 pmulhrsw    m2,        m7
4193
4194 packuswb    m4,        m2
4195
4196 movu        [r2 + 16], m4
4197
4198 lea         r0,        [r0 + r1]
4199 lea         r2,        [r2 + r3]
4200
4201 dec         r4
4202 jnz        .loop
4203 RET
4204 %endmacro
4205
4206 FILTER_V4_W32 32,  8
4207 FILTER_V4_W32 32, 16
4208 FILTER_V4_W32 32, 24
4209 FILTER_V4_W32 32, 32
4210
4211 FILTER_V4_W32 32, 48
4212 FILTER_V4_W32 32, 64
4213
4214 INIT_YMM avx2
4215 %if ARCH_X86_64 == 1
4216 cglobal interp_4tap_vert_pp_32x32, 4, 7, 13
4217     mov             r4d, r4m
4218     shl             r4d, 6
4219
4220 %ifdef PIC
4221     lea             r5, [tab_ChromaCoeffVer_32]
4222     add             r5, r4
4223 %else
4224     lea             r5, [tab_ChromaCoeffVer_32 + r4]
4225 %endif
4226
4227     mova            m10, [r5]
4228     mova            m11, [r5 + mmsize]
4229     lea             r4, [r1 * 3]
4230     sub             r0, r1
4231     lea             r5, [r3 * 3]
4232     mova            m12, [pw_512]
4233     mov             r6d, 8
4234 .loopW:
4235     movu            m0, [r0]                        ; m0 = row 0
4236     movu            m1, [r0 + r1]                   ; m1 = row 1
4237     punpcklbw       m2, m0, m1
4238     punpckhbw       m3, m0, m1
4239     pmaddubsw       m2, m10
4240     pmaddubsw       m3, m10
4241     movu            m0, [r0 + r1 * 2]               ; m0 = row 2
4242     punpcklbw       m4, m1, m0
4243     punpckhbw       m5, m1, m0
4244     pmaddubsw       m4, m10
4245     pmaddubsw       m5, m10
4246     movu            m1, [r0 + r4]                   ; m1 = row 3
4247     punpcklbw       m6, m0, m1
4248     punpckhbw       m7, m0, m1
4249     pmaddubsw       m8, m6, m11
4250     pmaddubsw       m9, m7, m11
4251     pmaddubsw       m6, m10
4252     pmaddubsw       m7, m10
4253     paddw           m2, m8
4254     paddw           m3, m9
4255     pmulhrsw        m2, m12
4256     pmulhrsw        m3, m12
4257     packuswb        m2, m3
4258     movu            [r2], m2
4259
4260     lea             r0, [r0 + r1 * 4]
4261     movu            m0, [r0]                        ; m0 = row 4
4262     punpcklbw       m2, m1, m0
4263     punpckhbw       m3, m1, m0
4264     pmaddubsw       m8, m2, m11
4265     pmaddubsw       m9, m3, m11
4266     pmaddubsw       m2, m10
4267     pmaddubsw       m3, m10
4268     paddw           m4, m8
4269     paddw           m5, m9
4270     pmulhrsw        m4, m12
4271     pmulhrsw        m5, m12
4272     packuswb        m4, m5
4273     movu            [r2 + r3], m4
4274
4275     movu            m1, [r0 + r1]                   ; m1 = row 5
4276     punpcklbw       m4, m0, m1
4277     punpckhbw       m5, m0, m1
4278     pmaddubsw       m4, m11
4279     pmaddubsw       m5, m11
4280     paddw           m6, m4
4281     paddw           m7, m5
4282     pmulhrsw        m6, m12
4283     pmulhrsw        m7, m12
4284     packuswb        m6, m7
4285     movu            [r2 + r3 * 2], m6
4286
4287     movu            m0, [r0 + r1 * 2]               ; m0 = row 6
4288     punpcklbw       m6, m1, m0
4289     punpckhbw       m7, m1, m0
4290     pmaddubsw       m6, m11
4291     pmaddubsw       m7, m11
4292     paddw           m2, m6
4293     paddw           m3, m7
4294     pmulhrsw        m2, m12
4295     pmulhrsw        m3, m12
4296     packuswb        m2, m3
4297     movu            [r2 + r5], m2
4298
4299     lea             r2, [r2 + r3 * 4]
4300     dec             r6d
4301     jnz             .loopW
4302     RET
4303 %endif
4304
4305 ;-----------------------------------------------------------------------------
4306 ; void interp_4tap_vert_pp_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
4307 ;-----------------------------------------------------------------------------
4308 %macro FILTER_V4_W16n_H2 2
4309 INIT_XMM sse4
4310 cglobal interp_4tap_vert_pp_%1x%2, 4, 7, 8
4311
4312 mov         r4d,       r4m
4313 sub         r0,        r1
4314
4315 %ifdef PIC
4316 lea         r5,        [tab_ChromaCoeff]
4317 movd        m0,        [r5 + r4 * 4]
4318 %else
4319 movd        m0,        [tab_ChromaCoeff + r4 * 4]
4320 %endif
4321
4322 pshufb      m1,        m0,       [tab_Vm]
4323 pshufb      m0,        [tab_Vm + 16]
4324
4325 mov         r4d,       %2/2
4326
4327 .loop:
4328
4329 mov         r6d,       %1/16
4330
4331 .loopW:
4332
4333 movu        m2,        [r0]
4334 movu        m3,        [r0 + r1]
4335
4336 punpcklbw   m4,        m2,        m3
4337 punpckhbw   m2,        m3
4338
4339 pmaddubsw   m4,        m1
4340 pmaddubsw   m2,        m1
4341
4342 lea         r5,        [r0 + 2 * r1]
4343 movu        m5,        [r5]
4344 movu        m6,        [r5 + r1]
4345
4346 punpckhbw   m7,        m5,        m6
4347 pmaddubsw   m7,        m0
4348 paddw       m2,        m7
4349
4350 punpcklbw   m7,        m5,        m6
4351 pmaddubsw   m7,        m0
4352 paddw       m4,        m7
4353
4354 mova        m7,        [pw_512]
4355
4356 pmulhrsw    m4,        m7
4357 pmulhrsw    m2,        m7
4358
4359 packuswb    m4,        m2
4360
4361 movu        [r2],      m4
4362
4363 punpcklbw   m4,        m3,        m5
4364 punpckhbw   m3,        m5
4365
4366 pmaddubsw   m4,        m1
4367 pmaddubsw   m3,        m1
4368
4369 movu        m5,        [r5 + 2 * r1]
4370
4371 punpcklbw   m2,        m6,        m5
4372 punpckhbw   m6,        m5
4373
4374 pmaddubsw   m2,        m0
4375 pmaddubsw   m6,        m0
4376
4377 paddw       m4,        m2
4378 paddw       m3,        m6
4379
4380 pmulhrsw    m4,        m7
4381 pmulhrsw    m3,        m7
4382
4383 packuswb    m4,        m3
4384
4385 movu        [r2 + r3],      m4
4386
4387 add         r0,        16
4388 add         r2,        16
4389 dec         r6d
4390 jnz         .loopW
4391
4392 lea         r0,        [r0 + r1 * 2 - %1]
4393 lea         r2,        [r2 + r3 * 2 - %1]
4394
4395 dec         r4d
4396 jnz        .loop
4397 RET
4398 %endmacro
4399
4400 FILTER_V4_W16n_H2 64, 64
4401 FILTER_V4_W16n_H2 64, 32
4402 FILTER_V4_W16n_H2 64, 48
4403 FILTER_V4_W16n_H2 48, 64
4404 FILTER_V4_W16n_H2 64, 16
4405
4406
4407 ;-----------------------------------------------------------------------------
4408 ; void filterConvertPelToShort(pixel *src, intptr_t srcStride, int16_t *dst, int width, int height)
4409 ;-----------------------------------------------------------------------------
4410 INIT_XMM ssse3
4411 cglobal luma_p2s, 3, 7, 6
4412
4413     ; load width and height
4414     mov         r3d, r3m
4415     mov         r4d, r4m
4416
4417     ; load constant
4418     mova        m4, [pb_128]
4419     mova        m5, [tab_c_64_n64]
4420
4421 .loopH:
4422
4423     xor         r5d, r5d
4424 .loopW:
4425     lea         r6, [r0 + r5]
4426
4427     movh        m0, [r6]
4428     punpcklbw   m0, m4
4429     pmaddubsw   m0, m5
4430
4431     movh        m1, [r6 + r1]
4432     punpcklbw   m1, m4
4433     pmaddubsw   m1, m5
4434
4435     movh        m2, [r6 + r1 * 2]
4436     punpcklbw   m2, m4
4437     pmaddubsw   m2, m5
4438
4439     lea         r6, [r6 + r1 * 2]
4440     movh        m3, [r6 + r1]
4441     punpcklbw   m3, m4
4442     pmaddubsw   m3, m5
4443
4444     add         r5, 8
4445     cmp         r5, r3
4446     jg          .width4
4447     movu        [r2 + r5 * 2 + FENC_STRIDE * 0 - 16], m0
4448     movu        [r2 + r5 * 2 + FENC_STRIDE * 2 - 16], m1
4449     movu        [r2 + r5 * 2 + FENC_STRIDE * 4 - 16], m2
4450     movu        [r2 + r5 * 2 + FENC_STRIDE * 6 - 16], m3
4451     je          .nextH
4452     jmp         .loopW
4453
4454 .width4:
4455     movh        [r2 + r5 * 2 + FENC_STRIDE * 0 - 16], m0
4456     movh        [r2 + r5 * 2 + FENC_STRIDE * 2 - 16], m1
4457     movh        [r2 + r5 * 2 + FENC_STRIDE * 4 - 16], m2
4458     movh        [r2 + r5 * 2 + FENC_STRIDE * 6 - 16], m3
4459
4460 .nextH:
4461     lea         r0, [r0 + r1 * 4]
4462     add         r2, FENC_STRIDE * 8
4463
4464     sub         r4d, 4
4465     jnz         .loopH
4466
4467     RET
4468
4469 %macro PROCESS_LUMA_W4_4R 0
4470     movd        m0, [r0]
4471     movd        m1, [r0 + r1]
4472     punpcklbw   m2, m0, m1                     ; m2=[0 1]
4473
4474     lea         r0, [r0 + 2 * r1]
4475     movd        m0, [r0]
4476     punpcklbw   m1, m0                         ; m1=[1 2]
4477     punpcklqdq  m2, m1                         ; m2=[0 1 1 2]
4478     pmaddubsw   m4, m2, [r6 + 0 * 16]          ; m4=[0+1 1+2]
4479
4480     movd        m1, [r0 + r1]
4481     punpcklbw   m5, m0, m1                     ; m2=[2 3]
4482     lea         r0, [r0 + 2 * r1]
4483     movd        m0, [r0]
4484     punpcklbw   m1, m0                         ; m1=[3 4]
4485     punpcklqdq  m5, m1                         ; m5=[2 3 3 4]
4486     pmaddubsw   m2, m5, [r6 + 1 * 16]          ; m2=[2+3 3+4]
4487     paddw       m4, m2                         ; m4=[0+1+2+3 1+2+3+4]                   Row1-2
4488     pmaddubsw   m5, [r6 + 0 * 16]              ; m5=[2+3 3+4]                           Row3-4
4489
4490     movd        m1, [r0 + r1]
4491     punpcklbw   m2, m0, m1                     ; m2=[4 5]
4492     lea         r0, [r0 + 2 * r1]
4493     movd        m0, [r0]
4494     punpcklbw   m1, m0                         ; m1=[5 6]
4495     punpcklqdq  m2, m1                         ; m2=[4 5 5 6]
4496     pmaddubsw   m1, m2, [r6 + 2 * 16]          ; m1=[4+5 5+6]
4497     paddw       m4, m1                         ; m4=[0+1+2+3+4+5 1+2+3+4+5+6]           Row1-2
4498     pmaddubsw   m2, [r6 + 1 * 16]              ; m2=[4+5 5+6]
4499     paddw       m5, m2                         ; m5=[2+3+4+5 3+4+5+6]                   Row3-4
4500
4501     movd        m1, [r0 + r1]
4502     punpcklbw   m2, m0, m1                     ; m2=[6 7]
4503     lea         r0, [r0 + 2 * r1]
4504     movd        m0, [r0]
4505     punpcklbw   m1, m0                         ; m1=[7 8]
4506     punpcklqdq  m2, m1                         ; m2=[6 7 7 8]
4507     pmaddubsw   m1, m2, [r6 + 3 * 16]          ; m1=[6+7 7+8]
4508     paddw       m4, m1                         ; m4=[0+1+2+3+4+5+6+7 1+2+3+4+5+6+7+8]   Row1-2 end
4509     pmaddubsw   m2, [r6 + 2 * 16]              ; m2=[6+7 7+8]
4510     paddw       m5, m2                         ; m5=[2+3+4+5+6+7 3+4+5+6+7+8]           Row3-4
4511
4512     movd        m1, [r0 + r1]
4513     punpcklbw   m2, m0, m1                     ; m2=[8 9]
4514     movd        m0, [r0 + 2 * r1]
4515     punpcklbw   m1, m0                         ; m1=[9 10]
4516     punpcklqdq  m2, m1                         ; m2=[8 9 9 10]
4517     pmaddubsw   m2, [r6 + 3 * 16]              ; m2=[8+9 9+10]
4518     paddw       m5, m2                         ; m5=[2+3+4+5+6+7+8+9 3+4+5+6+7+8+9+10]  Row3-4 end
4519 %endmacro
4520
4521 %macro PROCESS_LUMA_W8_4R 0
4522     movq       m0, [r0]
4523     movq       m1, [r0 + r1]
4524     punpcklbw  m0, m1
4525     pmaddubsw  m7, m0, [r6 + 0 *16]            ;m7=[0+1]               Row1
4526
4527     lea        r0, [r0 + 2 * r1]
4528     movq       m0, [r0]
4529     punpcklbw  m1, m0
4530     pmaddubsw  m6, m1, [r6 + 0 *16]            ;m6=[1+2]               Row2
4531
4532     movq       m1, [r0 + r1]
4533     punpcklbw  m0, m1
4534     pmaddubsw  m5, m0, [r6 + 0 *16]            ;m5=[2+3]               Row3
4535     pmaddubsw  m0, [r6 + 1 * 16]
4536     paddw      m7, m0                          ;m7=[0+1+2+3]           Row1
4537
4538     lea        r0, [r0 + 2 * r1]
4539     movq       m0, [r0]
4540     punpcklbw  m1, m0
4541     pmaddubsw  m4, m1, [r6 + 0 *16]            ;m4=[3+4]               Row4
4542     pmaddubsw  m1, [r6 + 1 * 16]
4543     paddw      m6, m1                          ;m6 = [1+2+3+4]         Row2
4544
4545     movq       m1, [r0 + r1]
4546     punpcklbw  m0, m1
4547     pmaddubsw  m2, m0, [r6 + 1 * 16]
4548     pmaddubsw  m0, [r6 + 2 * 16]
4549     paddw      m7, m0                          ;m7=[0+1+2+3+4+5]       Row1
4550     paddw      m5, m2                          ;m5=[2+3+4+5]           Row3
4551
4552     lea        r0, [r0 + 2 * r1]
4553     movq       m0, [r0]
4554     punpcklbw  m1, m0
4555     pmaddubsw  m2, m1, [r6 + 1 * 16]
4556     pmaddubsw  m1, [r6 + 2 * 16]
4557     paddw      m6, m1                          ;m6=[1+2+3+4+5+6]       Row2
4558     paddw      m4, m2                          ;m4=[3+4+5+6]           Row4
4559
4560     movq       m1, [r0 + r1]
4561     punpcklbw  m0, m1
4562     pmaddubsw  m2, m0, [r6 + 2 * 16]
4563     pmaddubsw  m0, [r6 + 3 * 16]
4564     paddw      m7, m0                          ;m7=[0+1+2+3+4+5+6+7]   Row1 end
4565     paddw      m5, m2                          ;m5=[2+3+4+5+6+7]       Row3
4566
4567     lea        r0, [r0 + 2 * r1]
4568     movq       m0, [r0]
4569     punpcklbw  m1, m0
4570     pmaddubsw  m2, m1, [r6 + 2 * 16]
4571     pmaddubsw  m1, [r6 + 3 * 16]
4572     paddw      m6, m1                          ;m6=[1+2+3+4+5+6+7+8]   Row2 end
4573     paddw      m4, m2                          ;m4=[3+4+5+6+7+8]       Row4
4574
4575     movq       m1, [r0 + r1]
4576     punpcklbw  m0, m1
4577     pmaddubsw  m0, [r6 + 3 * 16]
4578     paddw      m5, m0                          ;m5=[2+3+4+5+6+7+8+9]   Row3 end
4579
4580     movq       m0, [r0 + 2 * r1]
4581     punpcklbw  m1, m0
4582     pmaddubsw  m1, [r6 + 3 * 16]
4583     paddw      m4, m1                          ;m4=[3+4+5+6+7+8+9+10]  Row4 end
4584 %endmacro
4585
4586 ;-------------------------------------------------------------------------------------------------------------
4587 ; void interp_8tap_vert_%3_4x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
4588 ;-------------------------------------------------------------------------------------------------------------
4589 %macro FILTER_VER_LUMA_4xN 3
4590 INIT_XMM sse4
4591 cglobal interp_8tap_vert_%3_%1x%2, 5, 7, 6
4592     lea       r5, [3 * r1]
4593     sub       r0, r5
4594     shl       r4d, 6
4595 %ifidn %3,ps
4596     add       r3d, r3d
4597 %endif
4598
4599 %ifdef PIC
4600     lea       r5, [tab_LumaCoeffVer]
4601     lea       r6, [r5 + r4]
4602 %else
4603     lea       r6, [tab_LumaCoeffVer + r4]
4604 %endif
4605
4606 %ifidn %3,pp
4607     mova      m3, [pw_512]
4608 %else
4609     mova      m3, [pw_2000]
4610 %endif
4611
4612     mov       r4d, %2/4
4613     lea       r5, [4 * r1]
4614
4615 .loopH:
4616     PROCESS_LUMA_W4_4R
4617
4618 %ifidn %3,pp
4619     pmulhrsw  m4, m3
4620     pmulhrsw  m5, m3
4621
4622     packuswb  m4, m5
4623
4624     movd      [r2], m4
4625     pextrd    [r2 + r3], m4, 1
4626     lea       r2, [r2 + 2 * r3]
4627     pextrd    [r2], m4, 2
4628     pextrd    [r2 + r3], m4, 3
4629 %else
4630     psubw     m4, m3
4631     psubw     m5, m3
4632
4633     movlps    [r2], m4
4634     movhps    [r2 + r3], m4
4635     lea       r2, [r2 + 2 * r3]
4636     movlps    [r2], m5
4637     movhps    [r2 + r3], m5
4638 %endif
4639
4640     sub       r0, r5
4641     lea       r2, [r2 + 2 * r3]
4642
4643     dec       r4d
4644     jnz       .loopH
4645
4646     RET
4647 %endmacro
4648
4649
4650 INIT_YMM avx2
4651 cglobal interp_8tap_vert_pp_4x4, 4,6,8
4652     mov             r4d, r4m
4653     lea             r5, [r1 * 3]
4654     sub             r0, r5
4655
4656     ; TODO: VPGATHERDD
4657     movd            xm1, [r0]                       ; m1 = row0
4658     movd            xm2, [r0 + r1]                  ; m2 = row1
4659     punpcklbw       xm1, xm2                        ; m1 = [13 03 12 02 11 01 10 00]
4660
4661     movd            xm3, [r0 + r1 * 2]              ; m3 = row2
4662     punpcklbw       xm2, xm3                        ; m2 = [23 13 22 12 21 11 20 10]
4663     movd            xm4, [r0 + r5]
4664     punpcklbw       xm3, xm4                        ; m3 = [33 23 32 22 31 21 30 20]
4665     punpcklwd       xm1, xm3                        ; m1 = [33 23 13 03 32 22 12 02 31 21 11 01 30 20 10 00]
4666
4667     lea             r0, [r0 + r1 * 4]
4668     movd            xm5, [r0]                       ; m5 = row4
4669     punpcklbw       xm4, xm5                        ; m4 = [43 33 42 32 41 31 40 30]
4670     punpcklwd       xm2, xm4                        ; m2 = [43 33 21 13 42 32 22 12 41 31 21 11 40 30 20 10]
4671     vinserti128     m1, m1, xm2, 1                  ; m1 = [43 33 21 13 42 32 22 12 41 31 21 11 40 30 20 10] - [33 23 13 03 32 22 12 02 31 21 11 01 30 20 10 00]
4672     movd            xm2, [r0 + r1]                  ; m2 = row5
4673     punpcklbw       xm5, xm2                        ; m5 = [53 43 52 42 51 41 50 40]
4674     punpcklwd       xm3, xm5                        ; m3 = [53 43 44 23 52 42 32 22 51 41 31 21 50 40 30 20]
4675     movd            xm6, [r0 + r1 * 2]              ; m6 = row6
4676     punpcklbw       xm2, xm6                        ; m2 = [63 53 62 52 61 51 60 50]
4677     punpcklwd       xm4, xm2                        ; m4 = [63 53 43 33 62 52 42 32 61 51 41 31 60 50 40 30]
4678     vinserti128     m3, m3, xm4, 1                  ; m3 = [63 53 43 33 62 52 42 32 61 51 41 31 60 50 40 30] - [53 43 44 23 52 42 32 22 51 41 31 21 50 40 30 20]
4679     movd            xm4, [r0 + r5]                  ; m4 = row7
4680     punpcklbw       xm6, xm4                        ; m6 = [73 63 72 62 71 61 70 60]
4681     punpcklwd       xm5, xm6                        ; m5 = [73 63 53 43 72 62 52 42 71 61 51 41 70 60 50 40]
4682
4683     lea             r0, [r0 + r1 * 4]
4684     movd            xm7, [r0]                       ; m7 = row8
4685     punpcklbw       xm4, xm7                        ; m4 = [83 73 82 72 81 71 80 70]
4686     punpcklwd       xm2, xm4                        ; m2 = [83 73 63 53 82 72 62 52 81 71 61 51 80 70 60 50]
4687     vinserti128     m5, m5, xm2, 1                  ; m5 = [83 73 63 53 82 72 62 52 81 71 61 51 80 70 60 50] - [73 63 53 43 72 62 52 42 71 61 51 41 70 60 50 40]
4688     movd            xm2, [r0 + r1]                  ; m2 = row9
4689     punpcklbw       xm7, xm2                        ; m7 = [93 83 92 82 91 81 90 80]
4690     punpcklwd       xm6, xm7                        ; m6 = [93 83 73 63 92 82 72 62 91 81 71 61 90 80 70 60]
4691     movd            xm7, [r0 + r1 * 2]              ; m7 = rowA
4692     punpcklbw       xm2, xm7                        ; m2 = [A3 93 A2 92 A1 91 A0 90]
4693     punpcklwd       xm4, xm2                        ; m4 = [A3 93 83 73 A2 92 82 72 A1 91 81 71 A0 90 80 70]
4694     vinserti128     m6, m6, xm4, 1                  ; m6 = [A3 93 83 73 A2 92 82 72 A1 91 81 71 A0 90 80 70] - [93 83 73 63 92 82 72 62 91 81 71 61 90 80 70 60]
4695
4696     ; load filter coeff
4697 %ifdef PIC
4698     lea             r5, [tab_LumaCoeff]
4699     vpbroadcastd    m0, [r5 + r4 * 8 + 0]
4700     vpbroadcastd    m2, [r5 + r4 * 8 + 4]
4701 %else
4702     vpbroadcastd    m0, [tab_LumaCoeff + r4 * 8 + 0]
4703     vpbroadcastd    m2, [tab_LumaCoeff + r4 * 8 + 4]
4704 %endif
4705
4706     pmaddubsw       m1, m0
4707     pmaddubsw       m3, m0
4708     pmaddubsw       m5, m2
4709     pmaddubsw       m6, m2
4710     vbroadcasti128  m0, [pw_1]
4711     pmaddwd         m1, m0
4712     pmaddwd         m3, m0
4713     pmaddwd         m5, m0
4714     pmaddwd         m6, m0
4715     paddd           m1, m5                          ; m1 = DQWORD ROW[1 0]
4716     paddd           m3, m6                          ; m3 = DQWORD ROW[3 2]
4717     packssdw        m1, m3                          ; m1 =  QWORD ROW[3 1 2 0]
4718
4719     ; TODO: does it overflow?
4720     pmulhrsw        m1, [pw_512]
4721     vextracti128    xm2, m1, 1
4722     packuswb        xm1, xm2                        ; m1 =  DWORD ROW[3 1 2 0]
4723     movd            [r2], xm1
4724     pextrd          [r2 + r3], xm1, 2
4725     pextrd          [r2 + r3 * 2], xm1, 1
4726     lea             r4, [r3 * 3]
4727     pextrd          [r2 + r4], xm1, 3
4728     RET
4729
4730 INIT_YMM avx2
4731 cglobal interp_8tap_vert_ps_4x4, 4, 6, 5
4732     mov             r4d, r4m
4733     shl             r4d, 7
4734
4735 %ifdef PIC
4736     lea             r5, [tab_LumaCoeffVer_32]
4737     add             r5, r4
4738 %else
4739     lea             r5, [tab_LumaCoeffVer_32 + r4]
4740 %endif
4741
4742     lea             r4, [r1 * 3]
4743     sub             r0, r4
4744
4745     add             r3d, r3d
4746
4747     movd            xm1, [r0]
4748     pinsrd          xm1, [r0 + r1], 1
4749     pinsrd          xm1, [r0 + r1 * 2], 2
4750     pinsrd          xm1, [r0 + r4], 3                       ; m1 = row[3 2 1 0]
4751     lea             r0, [r0 + r1 * 4]
4752     movd            xm2, [r0]
4753     pinsrd          xm2, [r0 + r1], 1
4754     pinsrd          xm2, [r0 + r1 * 2], 2
4755     pinsrd          xm2, [r0 + r4], 3                       ; m2 = row[7 6 5 4]
4756     vinserti128     m1, m1, xm2, 1                          ; m1 = row[7 6 5 4 3 2 1 0]
4757     lea             r0, [r0 + r1 * 4]
4758     movd            xm3, [r0]
4759     pinsrd          xm3, [r0 + r1], 1
4760     pinsrd          xm3, [r0 + r1 * 2], 2                   ; m3 = row[x 10 9 8]
4761     vinserti128     m2, m2, xm3, 1                          ; m2 = row[x 10 9 8 7 6 5 4]
4762     mova            m3, [interp4_vpp_shuf1]
4763     vpermd          m0, m3, m1                              ; m0 = row[4 3 3 2 2 1 1 0]
4764     vpermd          m4, m3, m2                              ; m4 = row[8 7 7 6 6 5 5 4]
4765     mova            m3, [interp4_vpp_shuf1 + mmsize]
4766     vpermd          m1, m3, m1                              ; m1 = row[6 5 5 4 4 3 3 2]
4767     vpermd          m2, m3, m2                              ; m2 = row[10 9 9 8 8 7 7 6]
4768
4769     mova            m3, [interp4_vpp_shuf]
4770     pshufb          m0, m0, m3
4771     pshufb          m1, m1, m3
4772     pshufb          m4, m4, m3
4773     pshufb          m2, m2, m3
4774     pmaddubsw       m0, [r5]
4775     pmaddubsw       m1, [r5 + mmsize]
4776     pmaddubsw       m4, [r5 + 2 * mmsize]
4777     pmaddubsw       m2, [r5 + 3 * mmsize]
4778     paddw           m0, m1
4779     paddw           m0, m4
4780     paddw           m0, m2                                  ; m0 = WORD ROW[3 2 1 0]
4781
4782     vbroadcasti128  m3, [pw_2000]
4783     psubw           m0, m3
4784     vextracti128    xm2, m0, 1
4785     lea             r5, [r3 * 3]
4786     movq            [r2], xm0
4787     movhps          [r2 + r3], xm0
4788     movq            [r2 + r3 * 2], xm2
4789     movhps          [r2 + r5], xm2
4790     RET
4791
4792 ;-------------------------------------------------------------------------------------------------------------
4793 ; void interp_8tap_vert_pp_4x4(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
4794 ;-------------------------------------------------------------------------------------------------------------
4795 FILTER_VER_LUMA_4xN 4, 4, pp
4796
4797 ;-------------------------------------------------------------------------------------------------------------
4798 ; void interp_8tap_vert_pp_4x8(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
4799 ;-------------------------------------------------------------------------------------------------------------
4800 FILTER_VER_LUMA_4xN 4, 8, pp
4801
4802 ;-------------------------------------------------------------------------------------------------------------
4803 ; void interp_8tap_vert_pp_4x16(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
4804 ;-------------------------------------------------------------------------------------------------------------
4805 FILTER_VER_LUMA_4xN 4, 16, pp
4806
4807 ;-------------------------------------------------------------------------------------------------------------
4808 ; void interp_8tap_vert_ps_4x4(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
4809 ;-------------------------------------------------------------------------------------------------------------
4810 FILTER_VER_LUMA_4xN 4, 4, ps
4811
4812 ;-------------------------------------------------------------------------------------------------------------
4813 ; void interp_8tap_vert_ps_4x8(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
4814 ;-------------------------------------------------------------------------------------------------------------
4815 FILTER_VER_LUMA_4xN 4, 8, ps
4816
4817 ;-------------------------------------------------------------------------------------------------------------
4818 ; void interp_8tap_vert_ps_4x16(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
4819 ;-------------------------------------------------------------------------------------------------------------
4820 FILTER_VER_LUMA_4xN 4, 16, ps
4821
4822 %macro PROCESS_LUMA_AVX2_W8_8R 0
4823     movq            xm1, [r0]                       ; m1 = row 0
4824     movq            xm2, [r0 + r1]                  ; m2 = row 1
4825     punpcklbw       xm1, xm2                        ; m1 = [17 07 16 06 15 05 14 04 13 03 12 02 11 01 10 00]
4826     movq            xm3, [r0 + r1 * 2]              ; m3 = row 2
4827     punpcklbw       xm2, xm3                        ; m2 = [27 17 26 16 25 15 24 14 23 13 22 12 21 11 20 10]
4828     vinserti128     m5, m1, xm2, 1                  ; m5 = [27 17 26 16 25 15 24 14 23 13 22 12 21 11 20 10] - [17 07 16 06 15 05 14 04 13 03 12 02 11 01 10 00]
4829     pmaddubsw       m5, [r5]
4830     movq            xm4, [r0 + r4]                  ; m4 = row 3
4831     punpcklbw       xm3, xm4                        ; m3 = [37 27 36 26 35 25 34 24 33 23 32 22 31 21 30 20]
4832     lea             r0, [r0 + r1 * 4]
4833     movq            xm1, [r0]                       ; m1 = row 4
4834     punpcklbw       xm4, xm1                        ; m4 = [47 37 46 36 45 35 44 34 43 33 42 32 41 31 40 30]
4835     vinserti128     m2, m3, xm4, 1                  ; m2 = [47 37 46 36 45 35 44 34 43 33 42 32 41 31 40 30] - [37 27 36 26 35 25 34 24 33 23 32 22 31 21 30 20]
4836     pmaddubsw       m0, m2, [r5 + 1 * mmsize]
4837     paddw           m5, m0
4838     pmaddubsw       m2, [r5]
4839     movq            xm3, [r0 + r1]                  ; m3 = row 5
4840     punpcklbw       xm1, xm3                        ; m1 = [57 47 56 46 55 45 54 44 53 43 52 42 51 41 50 40]
4841     movq            xm4, [r0 + r1 * 2]              ; m4 = row 6
4842     punpcklbw       xm3, xm4                        ; m3 = [67 57 66 56 65 55 64 54 63 53 62 52 61 51 60 50]
4843     vinserti128     m1, m1, xm3, 1                  ; m1 = [67 57 66 56 65 55 64 54 63 53 62 52 61 51 60 50] - [57 47 56 46 55 45 54 44 53 43 52 42 51 41 50 40]
4844     pmaddubsw       m3, m1, [r5 + 2 * mmsize]
4845     paddw           m5, m3
4846     pmaddubsw       m0, m1, [r5 + 1 * mmsize]
4847     paddw           m2, m0
4848     pmaddubsw       m1, [r5]
4849     movq            xm3, [r0 + r4]                  ; m3 = row 7
4850     punpcklbw       xm4, xm3                        ; m4 = [77 67 76 66 75 65 74 64 73 63 72 62 71 61 70 60]
4851     lea             r0, [r0 + r1 * 4]
4852     movq            xm0, [r0]                       ; m0 = row 8
4853     punpcklbw       xm3, xm0                        ; m3 = [87 77 86 76 85 75 84 74 83 73 82 72 81 71 80 70]
4854     vinserti128     m4, m4, xm3, 1                  ; m4 = [87 77 86 76 85 75 84 74 83 73 82 72 81 71 80 70] - [77 67 76 66 75 65 74 64 73 63 72 62 71 61 70 60]
4855     pmaddubsw       m3, m4, [r5 + 3 * mmsize]
4856     paddw           m5, m3
4857     pmaddubsw       m3, m4, [r5 + 2 * mmsize]
4858     paddw           m2, m3
4859     pmaddubsw       m3, m4, [r5 + 1 * mmsize]
4860     paddw           m1, m3
4861     pmaddubsw       m4, [r5]
4862     movq            xm3, [r0 + r1]                  ; m3 = row 9
4863     punpcklbw       xm0, xm3                        ; m0 = [97 87 96 86 95 85 94 84 93 83 92 82 91 81 90 80]
4864     movq            xm6, [r0 + r1 * 2]              ; m6 = row 10
4865     punpcklbw       xm3, xm6                        ; m3 = [A7 97 A6 96 A5 95 A4 94 A3 93 A2 92 A1 91 A0 90]
4866     vinserti128     m0, m0, xm3, 1                  ; m0 = [A7 97 A6 96 A5 95 A4 94 A3 93 A2 92 A1 91 A0 90] - [97 87 96 86 95 85 94 84 93 83 92 82 91 81 90 80]
4867     pmaddubsw       m3, m0, [r5 + 3 * mmsize]
4868     paddw           m2, m3
4869     pmaddubsw       m3, m0, [r5 + 2 * mmsize]
4870     paddw           m1, m3
4871     pmaddubsw       m0, [r5 + 1 * mmsize]
4872     paddw           m4, m0
4873
4874     movq            xm3, [r0 + r4]                  ; m3 = row 11
4875     punpcklbw       xm6, xm3                        ; m6 = [B7 A7 B6 A6 B5 A5 B4 A4 B3 A3 B2 A2 B1 A1 B0 A0]
4876     lea             r0, [r0 + r1 * 4]
4877     movq            xm0, [r0]                       ; m0 = row 12
4878     punpcklbw       xm3, xm0                        ; m3 = [C7 B7 C6 B6 C5 B5 C4 B4 C3 B3 C2 B2 C1 B1 C0 B0]
4879     vinserti128     m6, m6, xm3, 1                  ; m6 = [C7 B7 C6 B6 C5 B5 C4 B4 C3 B3 C2 B2 C1 B1 C0 B0] - [B7 A7 B6 A6 B5 A5 B4 A4 B3 A3 B2 A2 B1 A1 B0 A0]
4880     pmaddubsw       m3, m6, [r5 + 3 * mmsize]
4881     paddw           m1, m3
4882     pmaddubsw       m6, [r5 + 2 * mmsize]
4883     paddw           m4, m6
4884     movq            xm3, [r0 + r1]                  ; m3 = row 13
4885     punpcklbw       xm0, xm3                        ; m0 = [D7 C7 D6 C6 D5 C5 D4 C4 D3 C3 D2 C2 D1 C1 D0 C0]
4886     movq            xm6, [r0 + r1 * 2]              ; m6 = row 14
4887     punpcklbw       xm3, xm6                        ; m3 = [E7 D7 E6 D6 E5 D5 E4 D4 E3 D3 E2 D2 E1 D1 E0 D0]
4888     vinserti128     m0, m0, xm3, 1                  ; m0 = [E7 D7 E6 D6 E5 D5 E4 D4 E3 D3 E2 D2 E1 D1 E0 D0] - [D7 C7 D6 C6 D5 C5 D4 C4 D3 C3 D2 C2 D1 C1 D0 C0]
4889     pmaddubsw       m0, [r5 + 3 * mmsize]
4890     paddw           m4, m0
4891 %endmacro
4892
4893 %macro PROCESS_LUMA_AVX2_W8_4R 0
4894     movq            xm1, [r0]                       ; m1 = row 0
4895     movq            xm2, [r0 + r1]                  ; m2 = row 1
4896     punpcklbw       xm1, xm2                        ; m1 = [17 07 16 06 15 05 14 04 13 03 12 02 11 01 10 00]
4897     movq            xm3, [r0 + r1 * 2]              ; m3 = row 2
4898     punpcklbw       xm2, xm3                        ; m2 = [27 17 26 16 25 15 24 14 23 13 22 12 21 11 20 10]
4899     vinserti128     m5, m1, xm2, 1                  ; m5 = [27 17 26 16 25 15 24 14 23 13 22 12 21 11 20 10] - [17 07 16 06 15 05 14 04 13 03 12 02 11 01 10 00]
4900     pmaddubsw       m5, [r5]
4901     movq            xm4, [r0 + r4]                  ; m4 = row 3
4902     punpcklbw       xm3, xm4                        ; m3 = [37 27 36 26 35 25 34 24 33 23 32 22 31 21 30 20]
4903     lea             r0, [r0 + r1 * 4]
4904     movq            xm1, [r0]                       ; m1 = row 4
4905     punpcklbw       xm4, xm1                        ; m4 = [47 37 46 36 45 35 44 34 43 33 42 32 41 31 40 30]
4906     vinserti128     m2, m3, xm4, 1                  ; m2 = [47 37 46 36 45 35 44 34 43 33 42 32 41 31 40 30] - [37 27 36 26 35 25 34 24 33 23 32 22 31 21 30 20]
4907     pmaddubsw       m0, m2, [r5 + 1 * mmsize]
4908     paddw           m5, m0
4909     pmaddubsw       m2, [r5]
4910     movq            xm3, [r0 + r1]                  ; m3 = row 5
4911     punpcklbw       xm1, xm3                        ; m1 = [57 47 56 46 55 45 54 44 53 43 52 42 51 41 50 40]
4912     movq            xm4, [r0 + r1 * 2]              ; m4 = row 6
4913     punpcklbw       xm3, xm4                        ; m3 = [67 57 66 56 65 55 64 54 63 53 62 52 61 51 60 50]
4914     vinserti128     m1, m1, xm3, 1                  ; m1 = [67 57 66 56 65 55 64 54 63 53 62 52 61 51 60 50] - [57 47 56 46 55 45 54 44 53 43 52 42 51 41 50 40]
4915     pmaddubsw       m3, m1, [r5 + 2 * mmsize]
4916     paddw           m5, m3
4917     pmaddubsw       m0, m1, [r5 + 1 * mmsize]
4918     paddw           m2, m0
4919     movq            xm3, [r0 + r4]                  ; m3 = row 7
4920     punpcklbw       xm4, xm3                        ; m4 = [77 67 76 66 75 65 74 64 73 63 72 62 71 61 70 60]
4921     lea             r0, [r0 + r1 * 4]
4922     movq            xm0, [r0]                       ; m0 = row 8
4923     punpcklbw       xm3, xm0                        ; m3 = [87 77 86 76 85 75 84 74 83 73 82 72 81 71 80 70]
4924     vinserti128     m4, m4, xm3, 1                  ; m4 = [87 77 86 76 85 75 84 74 83 73 82 72 81 71 80 70] - [77 67 76 66 75 65 74 64 73 63 72 62 71 61 70 60]
4925     pmaddubsw       m3, m4, [r5 + 3 * mmsize]
4926     paddw           m5, m3
4927     pmaddubsw       m3, m4, [r5 + 2 * mmsize]
4928     paddw           m2, m3
4929     movq            xm3, [r0 + r1]                  ; m3 = row 9
4930     punpcklbw       xm0, xm3                        ; m0 = [97 87 96 86 95 85 94 84 93 83 92 82 91 81 90 80]
4931     movq            xm6, [r0 + r1 * 2]              ; m6 = row 10
4932     punpcklbw       xm3, xm6                        ; m3 = [A7 97 A6 96 A5 95 A4 94 A3 93 A2 92 A1 91 A0 90]
4933     vinserti128     m0, m0, xm3, 1                  ; m0 = [A7 97 A6 96 A5 95 A4 94 A3 93 A2 92 A1 91 A0 90] - [97 87 96 86 95 85 94 84 93 83 92 82 91 81 90 80]
4934     pmaddubsw       m3, m0, [r5 + 3 * mmsize]
4935     paddw           m2, m3
4936 %endmacro
4937
4938 ;-------------------------------------------------------------------------------------------------------------
4939 ; void interp_8tap_vert_%3_8x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
4940 ;-------------------------------------------------------------------------------------------------------------
4941 %macro FILTER_VER_LUMA_8xN 3
4942 INIT_XMM sse4
4943 cglobal interp_8tap_vert_%3_%1x%2, 5, 7, 8
4944     lea       r5, [3 * r1]
4945     sub       r0, r5
4946     shl       r4d, 6
4947
4948 %ifidn %3,ps
4949     add       r3d, r3d
4950 %endif
4951
4952 %ifdef PIC
4953     lea       r5, [tab_LumaCoeffVer]
4954     lea       r6, [r5 + r4]
4955 %else
4956     lea       r6, [tab_LumaCoeffVer + r4]
4957 %endif
4958
4959  %ifidn %3,pp
4960     mova      m3, [pw_512]
4961 %else
4962     mova      m3, [pw_2000]
4963 %endif
4964
4965     mov       r4d, %2/4
4966     lea       r5, [4 * r1]
4967
4968 .loopH:
4969     PROCESS_LUMA_W8_4R
4970
4971 %ifidn %3,pp
4972     pmulhrsw  m7, m3
4973     pmulhrsw  m6, m3
4974     pmulhrsw  m5, m3
4975     pmulhrsw  m4, m3
4976
4977     packuswb  m7, m6
4978     packuswb  m5, m4
4979
4980     movlps    [r2], m7
4981     movhps    [r2 + r3], m7
4982     lea       r2, [r2 + 2 * r3]
4983     movlps    [r2], m5
4984     movhps    [r2 + r3], m5
4985 %else
4986     psubw     m7, m3
4987     psubw     m6, m3
4988     psubw     m5, m3
4989     psubw     m4, m3
4990
4991     movu      [r2], m7
4992     movu      [r2 + r3], m6
4993     lea       r2, [r2 + 2 * r3]
4994     movu      [r2], m5
4995     movu      [r2 + r3], m4
4996 %endif
4997
4998     sub       r0, r5
4999     lea       r2, [r2 + 2 * r3]
5000
5001     dec       r4d
5002     jnz       .loopH
5003
5004     RET
5005 %endmacro
5006
5007 %macro FILTER_VER_LUMA_AVX2_8xN 2
5008 INIT_YMM avx2
5009 cglobal interp_8tap_vert_pp_%1x%2, 4, 7, 8, 0-gprsize
5010     mov             r4d, r4m
5011     shl             r4d, 7
5012
5013 %ifdef PIC
5014     lea             r5, [tab_LumaCoeffVer_32]
5015     add             r5, r4
5016 %else
5017     lea             r5, [tab_LumaCoeffVer_32 + r4]
5018 %endif
5019     lea             r4, [r1 * 3]
5020     sub             r0, r4
5021     lea             r6, [r1 * 4]
5022     mov             word [rsp], %2 / 8
5023     mova            m7, [pw_512]
5024
5025 .loop:
5026     PROCESS_LUMA_AVX2_W8_8R
5027     pmulhrsw        m5, m7                          ; m5 = word: row 0, row 1
5028     pmulhrsw        m2, m7                          ; m2 = word: row 2, row 3
5029     pmulhrsw        m1, m7                          ; m1 = word: row 4, row 5
5030     pmulhrsw        m4, m7                          ; m4 = word: row 6, row 7
5031     packuswb        m5, m2
5032     packuswb        m1, m4
5033     vextracti128    xm2, m5, 1
5034     vextracti128    xm4, m1, 1
5035     movq            [r2], xm5
5036     movq            [r2 + r3], xm2
5037     lea             r2, [r2 + r3 * 2]
5038     movhps          [r2], xm5
5039     movhps          [r2 + r3], xm2
5040     lea             r2, [r2 + r3 * 2]
5041     movq            [r2], xm1
5042     movq            [r2 + r3], xm4
5043     lea             r2, [r2 + r3 * 2]
5044     movhps          [r2], xm1
5045     movhps          [r2 + r3], xm4
5046     lea             r2, [r2 + r3 * 2]
5047     sub             r0, r6
5048     dec             word [rsp]
5049     jnz             .loop
5050     RET
5051 %endmacro
5052
5053 INIT_YMM avx2
5054 cglobal interp_8tap_vert_pp_8x8, 4, 6, 7
5055     mov             r4d, r4m
5056     shl             r4d, 7
5057
5058 %ifdef PIC
5059     lea             r5, [tab_LumaCoeffVer_32]
5060     add             r5, r4
5061 %else
5062     lea             r5, [tab_LumaCoeffVer_32 + r4]
5063 %endif
5064
5065     lea             r4, [r1 * 3]
5066     sub             r0, r4
5067     PROCESS_LUMA_AVX2_W8_8R
5068     lea             r4, [r3 * 3]
5069     mova            m3, [pw_512]
5070     pmulhrsw        m5, m3                          ; m5 = word: row 0, row 1
5071     pmulhrsw        m2, m3                          ; m2 = word: row 2, row 3
5072     pmulhrsw        m1, m3                          ; m1 = word: row 4, row 5
5073     pmulhrsw        m4, m3                          ; m4 = word: row 6, row 7
5074     packuswb        m5, m2
5075     packuswb        m1, m4
5076     vextracti128    xm2, m5, 1
5077     vextracti128    xm4, m1, 1
5078     movq            [r2], xm5
5079     movq            [r2 + r3], xm2
5080     movhps          [r2 + r3 * 2], xm5
5081     movhps          [r2 + r4], xm2
5082     lea             r2, [r2 + r3 * 4]
5083     movq            [r2], xm1
5084     movq            [r2 + r3], xm4
5085     movhps          [r2 + r3 * 2], xm1
5086     movhps          [r2 + r4], xm4
5087     RET
5088
5089 INIT_YMM avx2
5090 cglobal interp_8tap_vert_pp_8x4, 4, 6, 7
5091     mov             r4d, r4m
5092     shl             r4d, 7
5093
5094 %ifdef PIC
5095     lea             r5, [tab_LumaCoeffVer_32]
5096     add             r5, r4
5097 %else
5098     lea             r5, [tab_LumaCoeffVer_32 + r4]
5099 %endif
5100
5101     lea             r4, [r1 * 3]
5102     sub             r0, r4
5103     PROCESS_LUMA_AVX2_W8_4R
5104     lea             r4, [r3 * 3]
5105     mova            m3, [pw_512]
5106     pmulhrsw        m5, m3                          ; m5 = word: row 0, row 1
5107     pmulhrsw        m2, m3                          ; m2 = word: row 2, row 3
5108     packuswb        m5, m2
5109     vextracti128    xm2, m5, 1
5110     movq            [r2], xm5
5111     movq            [r2 + r3], xm2
5112     movhps          [r2 + r3 * 2], xm5
5113     movhps          [r2 + r4], xm2
5114     RET
5115
5116 ;-------------------------------------------------------------------------------------------------------------
5117 ; void interp_8tap_vert_pp_8x4(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
5118 ;-------------------------------------------------------------------------------------------------------------
5119 FILTER_VER_LUMA_8xN 8, 4, pp
5120
5121 ;-------------------------------------------------------------------------------------------------------------
5122 ; void interp_8tap_vert_pp_8x8(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
5123 ;-------------------------------------------------------------------------------------------------------------
5124 FILTER_VER_LUMA_8xN 8, 8, pp
5125
5126 ;-------------------------------------------------------------------------------------------------------------
5127 ; void interp_8tap_vert_pp_8x16(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
5128 ;-------------------------------------------------------------------------------------------------------------
5129 FILTER_VER_LUMA_8xN 8, 16, pp
5130 FILTER_VER_LUMA_AVX2_8xN 8, 16
5131
5132 ;-------------------------------------------------------------------------------------------------------------
5133 ; void interp_8tap_vert_pp_8x32(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
5134 ;-------------------------------------------------------------------------------------------------------------
5135 FILTER_VER_LUMA_8xN 8, 32, pp
5136 FILTER_VER_LUMA_AVX2_8xN 8, 32
5137
5138 ;-------------------------------------------------------------------------------------------------------------
5139 ; void interp_8tap_vert_ps_8x4(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
5140 ;-------------------------------------------------------------------------------------------------------------
5141 FILTER_VER_LUMA_8xN 8, 4, ps
5142
5143 ;-------------------------------------------------------------------------------------------------------------
5144 ; void interp_8tap_vert_ps_8x8(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
5145 ;-------------------------------------------------------------------------------------------------------------
5146 FILTER_VER_LUMA_8xN 8, 8, ps
5147
5148 ;-------------------------------------------------------------------------------------------------------------
5149 ; void interp_8tap_vert_ps_8x16(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
5150 ;-------------------------------------------------------------------------------------------------------------
5151 FILTER_VER_LUMA_8xN 8, 16, ps
5152
5153 ;-------------------------------------------------------------------------------------------------------------
5154 ; void interp_8tap_vert_ps_8x32(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
5155 ;-------------------------------------------------------------------------------------------------------------
5156 FILTER_VER_LUMA_8xN 8, 32, ps
5157
5158 ;-------------------------------------------------------------------------------------------------------------
5159 ; void interp_8tap_vert_%3_12x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
5160 ;-------------------------------------------------------------------------------------------------------------
5161 %macro FILTER_VER_LUMA_12xN 3
5162 INIT_XMM sse4
5163 cglobal interp_8tap_vert_%3_%1x%2, 5, 7, 8
5164     lea       r5, [3 * r1]
5165     sub       r0, r5
5166     shl       r4d, 6
5167 %ifidn %3,ps
5168     add       r3d, r3d
5169 %endif
5170
5171 %ifdef PIC
5172     lea       r5, [tab_LumaCoeffVer]
5173     lea       r6, [r5 + r4]
5174 %else
5175     lea       r6, [tab_LumaCoeffVer + r4]
5176 %endif
5177
5178  %ifidn %3,pp
5179     mova      m3, [pw_512]
5180 %else
5181     mova      m3, [pw_2000]
5182 %endif
5183
5184     mov       r4d, %2/4
5185
5186 .loopH:
5187     PROCESS_LUMA_W8_4R
5188
5189 %ifidn %3,pp
5190     pmulhrsw  m7, m3
5191     pmulhrsw  m6, m3
5192     pmulhrsw  m5, m3
5193     pmulhrsw  m4, m3
5194
5195     packuswb  m7, m6
5196     packuswb  m5, m4
5197
5198     movlps    [r2], m7
5199     movhps    [r2 + r3], m7
5200     lea       r5, [r2 + 2 * r3]
5201     movlps    [r5], m5
5202     movhps    [r5 + r3], m5
5203 %else
5204     psubw     m7, m3
5205     psubw     m6, m3
5206     psubw     m5, m3
5207     psubw     m4, m3
5208
5209     movu      [r2], m7
5210     movu      [r2 + r3], m6
5211     lea       r5, [r2 + 2 * r3]
5212     movu      [r5], m5
5213     movu      [r5 + r3], m4
5214 %endif
5215
5216     lea       r5, [8 * r1 - 8]
5217     sub       r0, r5
5218 %ifidn %3,pp
5219     add       r2, 8
5220 %else
5221     add       r2, 16
5222 %endif
5223
5224     PROCESS_LUMA_W4_4R
5225
5226 %ifidn %3,pp
5227     pmulhrsw  m4, m3
5228     pmulhrsw  m5, m3
5229
5230     packuswb  m4, m5
5231
5232     movd      [r2], m4
5233     pextrd    [r2 + r3], m4, 1
5234     lea       r5, [r2 + 2 * r3]
5235     pextrd    [r5], m4, 2
5236     pextrd    [r5 + r3], m4, 3
5237 %else
5238     psubw     m4, m3
5239     psubw     m5, m3
5240
5241     movlps    [r2], m4
5242     movhps    [r2 + r3], m4
5243     lea       r5, [r2 + 2 * r3]
5244     movlps    [r5], m5
5245     movhps    [r5 + r3], m5
5246 %endif
5247
5248     lea       r5, [4 * r1 + 8]
5249     sub       r0, r5
5250 %ifidn %3,pp
5251     lea       r2, [r2 + 4 * r3 - 8]
5252 %else
5253     lea       r2, [r2 + 4 * r3 - 16]
5254 %endif
5255
5256     dec       r4d
5257     jnz       .loopH
5258
5259     RET
5260 %endmacro
5261
5262 ;-------------------------------------------------------------------------------------------------------------
5263 ; void interp_8tap_vert_pp_12x16(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
5264 ;-------------------------------------------------------------------------------------------------------------
5265 FILTER_VER_LUMA_12xN 12, 16, pp
5266
5267 ;-------------------------------------------------------------------------------------------------------------
5268 ; void interp_8tap_vert_ps_12x16(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
5269 ;-------------------------------------------------------------------------------------------------------------
5270 FILTER_VER_LUMA_12xN 12, 16, ps
5271
5272 INIT_YMM avx2
5273 %if ARCH_X86_64 == 1
5274 cglobal interp_8tap_vert_pp_12x16, 4, 7, 15
5275     mov             r4d, r4m
5276     shl             r4d, 7
5277
5278 %ifdef PIC
5279     lea             r5, [tab_LumaCoeffVer_32]
5280     add             r5, r4
5281 %else
5282     lea             r5, [tab_LumaCoeffVer_32 + r4]
5283 %endif
5284
5285     lea             r4, [r1 * 3]
5286     sub             r0, r4
5287     lea             r6, [r3 * 3]
5288     mova            m14, [pw_512]
5289
5290     movu            xm0, [r0]                       ; m0 = row 0
5291     movu            xm1, [r0 + r1]                  ; m1 = row 1
5292     punpckhbw       xm2, xm0, xm1
5293     punpcklbw       xm0, xm1
5294     vinserti128     m0, m0, xm2, 1
5295     pmaddubsw       m0, [r5]
5296     movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
5297     punpckhbw       xm3, xm1, xm2
5298     punpcklbw       xm1, xm2
5299     vinserti128     m1, m1, xm3, 1
5300     pmaddubsw       m1, [r5]
5301     movu            xm3, [r0 + r4]                  ; m3 = row 3
5302     punpckhbw       xm4, xm2, xm3
5303     punpcklbw       xm2, xm3
5304     vinserti128     m2, m2, xm4, 1
5305     pmaddubsw       m4, m2, [r5 + 1 * mmsize]
5306     paddw           m0, m4
5307     pmaddubsw       m2, [r5]
5308     lea             r0, [r0 + r1 * 4]
5309     movu            xm4, [r0]                       ; m4 = row 4
5310     punpckhbw       xm5, xm3, xm4
5311     punpcklbw       xm3, xm4
5312     vinserti128     m3, m3, xm5, 1
5313     pmaddubsw       m5, m3, [r5 + 1 * mmsize]
5314     paddw           m1, m5
5315     pmaddubsw       m3, [r5]
5316     movu            xm5, [r0 + r1]                  ; m5 = row 5
5317     punpckhbw       xm6, xm4, xm5
5318     punpcklbw       xm4, xm5
5319     vinserti128     m4, m4, xm6, 1
5320     pmaddubsw       m6, m4, [r5 + 2 * mmsize]
5321     paddw           m0, m6
5322     pmaddubsw       m6, m4, [r5 + 1 * mmsize]
5323     paddw           m2, m6
5324     pmaddubsw       m4, [r5]
5325     movu            xm6, [r0 + r1 * 2]              ; m6 = row 6
5326     punpckhbw       xm7, xm5, xm6
5327     punpcklbw       xm5, xm6
5328     vinserti128     m5, m5, xm7, 1
5329     pmaddubsw       m7, m5, [r5 + 2 * mmsize]
5330     paddw           m1, m7
5331     pmaddubsw       m7, m5, [r5 + 1 * mmsize]
5332     paddw           m3, m7
5333     pmaddubsw       m5, [r5]
5334     movu            xm7, [r0 + r4]                  ; m7 = row 7
5335     punpckhbw       xm8, xm6, xm7
5336     punpcklbw       xm6, xm7
5337     vinserti128     m6, m6, xm8, 1
5338     pmaddubsw       m8, m6, [r5 + 3 * mmsize]
5339     paddw           m0, m8
5340     pmaddubsw       m8, m6, [r5 + 2 * mmsize]
5341     paddw           m2, m8
5342     pmaddubsw       m8, m6, [r5 + 1 * mmsize]
5343     paddw           m4, m8
5344     pmaddubsw       m6, [r5]
5345     lea             r0, [r0 + r1 * 4]
5346     movu            xm8, [r0]                       ; m8 = row 8
5347     punpckhbw       xm9, xm7, xm8
5348     punpcklbw       xm7, xm8
5349     vinserti128     m7, m7, xm9, 1
5350     pmaddubsw       m9, m7, [r5 + 3 * mmsize]
5351     paddw           m1, m9
5352     pmaddubsw       m9, m7, [r5 + 2 * mmsize]
5353     paddw           m3, m9
5354     pmaddubsw       m9, m7, [r5 + 1 * mmsize]
5355     paddw           m5, m9
5356     pmaddubsw       m7, [r5]
5357     movu            xm9, [r0 + r1]                  ; m9 = row 9
5358     punpckhbw       xm10, xm8, xm9
5359     punpcklbw       xm8, xm9
5360     vinserti128     m8, m8, xm10, 1
5361     pmaddubsw       m10, m8, [r5 + 3 * mmsize]
5362     paddw           m2, m10
5363     pmaddubsw       m10, m8, [r5 + 2 * mmsize]
5364     paddw           m4, m10
5365     pmaddubsw       m10, m8, [r5 + 1 * mmsize]
5366     paddw           m6, m10
5367     pmaddubsw       m8, [r5]
5368     movu            xm10, [r0 + r1 * 2]             ; m10 = row 10
5369     punpckhbw       xm11, xm9, xm10
5370     punpcklbw       xm9, xm10
5371     vinserti128     m9, m9, xm11, 1
5372     pmaddubsw       m11, m9, [r5 + 3 * mmsize]
5373     paddw           m3, m11
5374     pmaddubsw       m11, m9, [r5 + 2 * mmsize]
5375     paddw           m5, m11
5376     pmaddubsw       m11, m9, [r5 + 1 * mmsize]
5377     paddw           m7, m11
5378     pmaddubsw       m9, [r5]
5379     movu            xm11, [r0 + r4]                 ; m11 = row 11
5380     punpckhbw       xm12, xm10, xm11
5381     punpcklbw       xm10, xm11
5382     vinserti128     m10, m10, xm12, 1
5383     pmaddubsw       m12, m10, [r5 + 3 * mmsize]
5384     paddw           m4, m12
5385     pmaddubsw       m12, m10, [r5 + 2 * mmsize]
5386     paddw           m6, m12
5387     pmaddubsw       m12, m10, [r5 + 1 * mmsize]
5388     paddw           m8, m12
5389     pmaddubsw       m10, [r5]
5390     lea             r0, [r0 + r1 * 4]
5391     movu            xm12, [r0]                      ; m12 = row 12
5392     punpckhbw       xm13, xm11, xm12
5393     punpcklbw       xm11, xm12
5394     vinserti128     m11, m11, xm13, 1
5395     pmaddubsw       m13, m11, [r5 + 3 * mmsize]
5396     paddw           m5, m13
5397     pmaddubsw       m13, m11, [r5 + 2 * mmsize]
5398     paddw           m7, m13
5399     pmaddubsw       m13, m11, [r5 + 1 * mmsize]
5400     paddw           m9, m13
5401     pmaddubsw       m11, [r5]
5402
5403     pmulhrsw        m0, m14                         ; m0 = word: row 0
5404     pmulhrsw        m1, m14                         ; m1 = word: row 1
5405     pmulhrsw        m2, m14                         ; m2 = word: row 2
5406     pmulhrsw        m3, m14                         ; m3 = word: row 3
5407     pmulhrsw        m4, m14                         ; m4 = word: row 4
5408     pmulhrsw        m5, m14                         ; m5 = word: row 5
5409     packuswb        m0, m1
5410     packuswb        m2, m3
5411     packuswb        m4, m5
5412     vpermq          m0, m0, 11011000b
5413     vpermq          m2, m2, 11011000b
5414     vpermq          m4, m4, 11011000b
5415     vextracti128    xm1, m0, 1
5416     vextracti128    xm3, m2, 1
5417     vextracti128    xm5, m4, 1
5418     movq            [r2], xm0
5419     pextrd          [r2 + 8], xm0, 2
5420     movq            [r2 + r3], xm1
5421     pextrd          [r2 + r3 + 8], xm1, 2
5422     movq            [r2 + r3 * 2], xm2
5423     pextrd          [r2 + r3 * 2 + 8], xm2, 2
5424     movq            [r2 + r6], xm3
5425     pextrd          [r2 + r6 + 8], xm3, 2
5426     lea             r2, [r2 + r3 * 4]
5427     movq            [r2], xm4
5428     pextrd          [r2 + 8], xm4, 2
5429     movq            [r2 + r3], xm5
5430     pextrd          [r2 + r3 + 8], xm5, 2
5431
5432     movu            xm13, [r0 + r1]                 ; m13 = row 13
5433     punpckhbw       xm0, xm12, xm13
5434     punpcklbw       xm12, xm13
5435     vinserti128     m12, m12, xm0, 1
5436     pmaddubsw       m0, m12, [r5 + 3 * mmsize]
5437     paddw           m6, m0
5438     pmaddubsw       m0, m12, [r5 + 2 * mmsize]
5439     paddw           m8, m0
5440     pmaddubsw       m0, m12, [r5 + 1 * mmsize]
5441     paddw           m10, m0
5442     pmaddubsw       m12, [r5]
5443     movu            xm0, [r0 + r1 * 2]              ; m0 = row 14
5444     punpckhbw       xm1, xm13, xm0
5445     punpcklbw       xm13, xm0
5446     vinserti128     m13, m13, xm1, 1
5447     pmaddubsw       m1, m13, [r5 + 3 * mmsize]
5448     paddw           m7, m1
5449     pmaddubsw       m1, m13, [r5 + 2 * mmsize]
5450     paddw           m9, m1
5451     pmaddubsw       m1, m13, [r5 + 1 * mmsize]
5452     paddw           m11, m1
5453     pmaddubsw       m13, [r5]
5454
5455     pmulhrsw        m6, m14                         ; m6 = word: row 6
5456     pmulhrsw        m7, m14                         ; m7 = word: row 7
5457     packuswb        m6, m7
5458     vpermq          m6, m6, 11011000b
5459     vextracti128    xm7, m6, 1
5460     movq            [r2 + r3 * 2], xm6
5461     pextrd          [r2 + r3 * 2 + 8], xm6, 2
5462     movq            [r2 + r6], xm7
5463     pextrd          [r2 + r6 + 8], xm7, 2
5464     lea             r2, [r2 + r3 * 4]
5465
5466     movu            xm1, [r0 + r4]                  ; m1 = row 15
5467     punpckhbw       xm2, xm0, xm1
5468     punpcklbw       xm0, xm1
5469     vinserti128     m0, m0, xm2, 1
5470     pmaddubsw       m2, m0, [r5 + 3 * mmsize]
5471     paddw           m8, m2
5472     pmaddubsw       m2, m0, [r5 + 2 * mmsize]
5473     paddw           m10, m2
5474     pmaddubsw       m2, m0, [r5 + 1 * mmsize]
5475     paddw           m12, m2
5476     pmaddubsw       m0, [r5]
5477     lea             r0, [r0 + r1 * 4]
5478     movu            xm2, [r0]                       ; m2 = row 16
5479     punpckhbw       xm3, xm1, xm2
5480     punpcklbw       xm1, xm2
5481     vinserti128     m1, m1, xm3, 1
5482     pmaddubsw       m3, m1, [r5 + 3 * mmsize]
5483     paddw           m9, m3
5484     pmaddubsw       m3, m1, [r5 + 2 * mmsize]
5485     paddw           m11, m3
5486     pmaddubsw       m3, m1, [r5 + 1 * mmsize]
5487     paddw           m13, m3
5488     pmaddubsw       m1, [r5]
5489     movu            xm3, [r0 + r1]                  ; m3 = row 17
5490     punpckhbw       xm4, xm2, xm3
5491     punpcklbw       xm2, xm3
5492     vinserti128     m2, m2, xm4, 1
5493     pmaddubsw       m4, m2, [r5 + 3 * mmsize]
5494     paddw           m10, m4
5495     pmaddubsw       m4, m2, [r5 + 2 * mmsize]
5496     paddw           m12, m4
5497     pmaddubsw       m2, [r5 + 1 * mmsize]
5498     paddw           m0, m2
5499     movu            xm4, [r0 + r1 * 2]              ; m4 = row 18
5500     punpckhbw       xm5, xm3, xm4
5501     punpcklbw       xm3, xm4
5502     vinserti128     m3, m3, xm5, 1
5503     pmaddubsw       m5, m3, [r5 + 3 * mmsize]
5504     paddw           m11, m5
5505     pmaddubsw       m5, m3, [r5 + 2 * mmsize]
5506     paddw           m13, m5
5507     pmaddubsw       m3, [r5 + 1 * mmsize]
5508     paddw           m1, m3
5509     movu            xm5, [r0 + r4]                  ; m5 = row 19
5510     punpckhbw       xm6, xm4, xm5
5511     punpcklbw       xm4, xm5
5512     vinserti128     m4, m4, xm6, 1
5513     pmaddubsw       m6, m4, [r5 + 3 * mmsize]
5514     paddw           m12, m6
5515     pmaddubsw       m4, [r5 + 2 * mmsize]
5516     paddw           m0, m4
5517     lea             r0, [r0 + r1 * 4]
5518     movu            xm6, [r0]                       ; m6 = row 20
5519     punpckhbw       xm7, xm5, xm6
5520     punpcklbw       xm5, xm6
5521     vinserti128     m5, m5, xm7, 1
5522     pmaddubsw       m7, m5, [r5 + 3 * mmsize]
5523     paddw           m13, m7
5524     pmaddubsw       m5, [r5 + 2 * mmsize]
5525     paddw           m1, m5
5526     movu            xm7, [r0 + r1]                  ; m7 = row 21
5527     punpckhbw       xm2, xm6, xm7
5528     punpcklbw       xm6, xm7
5529     vinserti128     m6, m6, xm2, 1
5530     pmaddubsw       m6, [r5 + 3 * mmsize]
5531     paddw           m0, m6
5532     movu            xm2, [r0 + r1 * 2]              ; m2 = row 22
5533     punpckhbw       xm3, xm7, xm2
5534     punpcklbw       xm7, xm2
5535     vinserti128     m7, m7, xm3, 1
5536     pmaddubsw       m7, [r5 + 3 * mmsize]
5537     paddw           m1, m7
5538
5539     pmulhrsw        m8, m14                         ; m8 = word: row 8
5540     pmulhrsw        m9, m14                         ; m9 = word: row 9
5541     pmulhrsw        m10, m14                        ; m10 = word: row 10
5542     pmulhrsw        m11, m14                        ; m11 = word: row 11
5543     pmulhrsw        m12, m14                        ; m12 = word: row 12
5544     pmulhrsw        m13, m14                        ; m13 = word: row 13
5545     pmulhrsw        m0, m14                         ; m0 = word: row 14
5546     pmulhrsw        m1, m14                         ; m1 = word: row 15
5547     packuswb        m8, m9
5548     packuswb        m10, m11
5549     packuswb        m12, m13
5550     packuswb        m0, m1
5551     vpermq          m8, m8, 11011000b
5552     vpermq          m10, m10, 11011000b
5553     vpermq          m12, m12, 11011000b
5554     vpermq          m0, m0, 11011000b
5555     vextracti128    xm9, m8, 1
5556     vextracti128    xm11, m10, 1
5557     vextracti128    xm13, m12, 1
5558     vextracti128    xm1, m0, 1
5559     movq            [r2], xm8
5560     pextrd          [r2 + 8], xm8, 2
5561     movq            [r2 + r3], xm9
5562     pextrd          [r2 + r3 + 8], xm9, 2
5563     movq            [r2 + r3 * 2], xm10
5564     pextrd          [r2 + r3 * 2 + 8], xm10, 2
5565     movq            [r2 + r6], xm11
5566     pextrd          [r2 + r6 + 8], xm11, 2
5567     lea             r2, [r2 + r3 * 4]
5568     movq            [r2], xm12
5569     pextrd          [r2 + 8], xm12, 2
5570     movq            [r2 + r3], xm13
5571     pextrd          [r2 + r3 + 8], xm13, 2
5572     movq            [r2 + r3 * 2], xm0
5573     pextrd          [r2 + r3 * 2 + 8], xm0, 2
5574     movq            [r2 + r6], xm1
5575     pextrd          [r2 + r6 + 8], xm1, 2
5576     RET
5577 %endif
5578
5579 INIT_YMM avx2
5580 %if ARCH_X86_64 == 1
5581 cglobal interp_8tap_vert_pp_16x16, 4, 7, 15
5582     mov             r4d, r4m
5583     shl             r4d, 7
5584
5585 %ifdef PIC
5586     lea             r5, [tab_LumaCoeffVer_32]
5587     add             r5, r4
5588 %else
5589     lea             r5, [tab_LumaCoeffVer_32 + r4]
5590 %endif
5591
5592     lea             r4, [r1 * 3]
5593     sub             r0, r4
5594     lea             r6, [r3 * 3]
5595     mova            m14, [pw_512]
5596
5597     movu            xm0, [r0]                       ; m0 = row 0
5598     movu            xm1, [r0 + r1]                  ; m1 = row 1
5599     punpckhbw       xm2, xm0, xm1
5600     punpcklbw       xm0, xm1
5601     vinserti128     m0, m0, xm2, 1
5602     pmaddubsw       m0, [r5]
5603     movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
5604     punpckhbw       xm3, xm1, xm2
5605     punpcklbw       xm1, xm2
5606     vinserti128     m1, m1, xm3, 1
5607     pmaddubsw       m1, [r5]
5608     movu            xm3, [r0 + r4]                  ; m3 = row 3
5609     punpckhbw       xm4, xm2, xm3
5610     punpcklbw       xm2, xm3
5611     vinserti128     m2, m2, xm4, 1
5612     pmaddubsw       m4, m2, [r5 + 1 * mmsize]
5613     paddw           m0, m4
5614     pmaddubsw       m2, [r5]
5615     lea             r0, [r0 + r1 * 4]
5616     movu            xm4, [r0]                       ; m4 = row 4
5617     punpckhbw       xm5, xm3, xm4
5618     punpcklbw       xm3, xm4
5619     vinserti128     m3, m3, xm5, 1
5620     pmaddubsw       m5, m3, [r5 + 1 * mmsize]
5621     paddw           m1, m5
5622     pmaddubsw       m3, [r5]
5623     movu            xm5, [r0 + r1]                  ; m5 = row 5
5624     punpckhbw       xm6, xm4, xm5
5625     punpcklbw       xm4, xm5
5626     vinserti128     m4, m4, xm6, 1
5627     pmaddubsw       m6, m4, [r5 + 2 * mmsize]
5628     paddw           m0, m6
5629     pmaddubsw       m6, m4, [r5 + 1 * mmsize]
5630     paddw           m2, m6
5631     pmaddubsw       m4, [r5]
5632     movu            xm6, [r0 + r1 * 2]              ; m6 = row 6
5633     punpckhbw       xm7, xm5, xm6
5634     punpcklbw       xm5, xm6
5635     vinserti128     m5, m5, xm7, 1
5636     pmaddubsw       m7, m5, [r5 + 2 * mmsize]
5637     paddw           m1, m7
5638     pmaddubsw       m7, m5, [r5 + 1 * mmsize]
5639     paddw           m3, m7
5640     pmaddubsw       m5, [r5]
5641     movu            xm7, [r0 + r4]                  ; m7 = row 7
5642     punpckhbw       xm8, xm6, xm7
5643     punpcklbw       xm6, xm7
5644     vinserti128     m6, m6, xm8, 1
5645     pmaddubsw       m8, m6, [r5 + 3 * mmsize]
5646     paddw           m0, m8
5647     pmaddubsw       m8, m6, [r5 + 2 * mmsize]
5648     paddw           m2, m8
5649     pmaddubsw       m8, m6, [r5 + 1 * mmsize]
5650     paddw           m4, m8
5651     pmaddubsw       m6, [r5]
5652     lea             r0, [r0 + r1 * 4]
5653     movu            xm8, [r0]                       ; m8 = row 8
5654     punpckhbw       xm9, xm7, xm8
5655     punpcklbw       xm7, xm8
5656     vinserti128     m7, m7, xm9, 1
5657     pmaddubsw       m9, m7, [r5 + 3 * mmsize]
5658     paddw           m1, m9
5659     pmaddubsw       m9, m7, [r5 + 2 * mmsize]
5660     paddw           m3, m9
5661     pmaddubsw       m9, m7, [r5 + 1 * mmsize]
5662     paddw           m5, m9
5663     pmaddubsw       m7, [r5]
5664     movu            xm9, [r0 + r1]                  ; m9 = row 9
5665     punpckhbw       xm10, xm8, xm9
5666     punpcklbw       xm8, xm9
5667     vinserti128     m8, m8, xm10, 1
5668     pmaddubsw       m10, m8, [r5 + 3 * mmsize]
5669     paddw           m2, m10
5670     pmaddubsw       m10, m8, [r5 + 2 * mmsize]
5671     paddw           m4, m10
5672     pmaddubsw       m10, m8, [r5 + 1 * mmsize]
5673     paddw           m6, m10
5674     pmaddubsw       m8, [r5]
5675     movu            xm10, [r0 + r1 * 2]             ; m10 = row 10
5676     punpckhbw       xm11, xm9, xm10
5677     punpcklbw       xm9, xm10
5678     vinserti128     m9, m9, xm11, 1
5679     pmaddubsw       m11, m9, [r5 + 3 * mmsize]
5680     paddw           m3, m11
5681     pmaddubsw       m11, m9, [r5 + 2 * mmsize]
5682     paddw           m5, m11
5683     pmaddubsw       m11, m9, [r5 + 1 * mmsize]
5684     paddw           m7, m11
5685     pmaddubsw       m9, [r5]
5686     movu            xm11, [r0 + r4]                 ; m11 = row 11
5687     punpckhbw       xm12, xm10, xm11
5688     punpcklbw       xm10, xm11
5689     vinserti128     m10, m10, xm12, 1
5690     pmaddubsw       m12, m10, [r5 + 3 * mmsize]
5691     paddw           m4, m12
5692     pmaddubsw       m12, m10, [r5 + 2 * mmsize]
5693     paddw           m6, m12
5694     pmaddubsw       m12, m10, [r5 + 1 * mmsize]
5695     paddw           m8, m12
5696     pmaddubsw       m10, [r5]
5697     lea             r0, [r0 + r1 * 4]
5698     movu            xm12, [r0]                      ; m12 = row 12
5699     punpckhbw       xm13, xm11, xm12
5700     punpcklbw       xm11, xm12
5701     vinserti128     m11, m11, xm13, 1
5702     pmaddubsw       m13, m11, [r5 + 3 * mmsize]
5703     paddw           m5, m13
5704     pmaddubsw       m13, m11, [r5 + 2 * mmsize]
5705     paddw           m7, m13
5706     pmaddubsw       m13, m11, [r5 + 1 * mmsize]
5707     paddw           m9, m13
5708     pmaddubsw       m11, [r5]
5709
5710     pmulhrsw        m0, m14                         ; m0 = word: row 0
5711     pmulhrsw        m1, m14                         ; m1 = word: row 1
5712     pmulhrsw        m2, m14                         ; m2 = word: row 2
5713     pmulhrsw        m3, m14                         ; m3 = word: row 3
5714     pmulhrsw        m4, m14                         ; m4 = word: row 4
5715     pmulhrsw        m5, m14                         ; m5 = word: row 5
5716     packuswb        m0, m1
5717     packuswb        m2, m3
5718     packuswb        m4, m5
5719     vpermq          m0, m0, 11011000b
5720     vpermq          m2, m2, 11011000b
5721     vpermq          m4, m4, 11011000b
5722     vextracti128    xm1, m0, 1
5723     vextracti128    xm3, m2, 1
5724     vextracti128    xm5, m4, 1
5725     movu            [r2], xm0
5726     movu            [r2 + r3], xm1
5727     movu            [r2 + r3 * 2], xm2
5728     movu            [r2 + r6], xm3
5729     lea             r2, [r2 + r3 * 4]
5730     movu            [r2], xm4
5731     movu            [r2 + r3], xm5
5732
5733     movu            xm13, [r0 + r1]                 ; m13 = row 13
5734     punpckhbw       xm0, xm12, xm13
5735     punpcklbw       xm12, xm13
5736     vinserti128     m12, m12, xm0, 1
5737     pmaddubsw       m0, m12, [r5 + 3 * mmsize]
5738     paddw           m6, m0
5739     pmaddubsw       m0, m12, [r5 + 2 * mmsize]
5740     paddw           m8, m0
5741     pmaddubsw       m0, m12, [r5 + 1 * mmsize]
5742     paddw           m10, m0
5743     pmaddubsw       m12, [r5]
5744     movu            xm0, [r0 + r1 * 2]              ; m0 = row 14
5745     punpckhbw       xm1, xm13, xm0
5746     punpcklbw       xm13, xm0
5747     vinserti128     m13, m13, xm1, 1
5748     pmaddubsw       m1, m13, [r5 + 3 * mmsize]
5749     paddw           m7, m1
5750     pmaddubsw       m1, m13, [r5 + 2 * mmsize]
5751     paddw           m9, m1
5752     pmaddubsw       m1, m13, [r5 + 1 * mmsize]
5753     paddw           m11, m1
5754     pmaddubsw       m13, [r5]
5755
5756     pmulhrsw        m6, m14                         ; m6 = word: row 6
5757     pmulhrsw        m7, m14                         ; m7 = word: row 7
5758     packuswb        m6, m7
5759     vpermq          m6, m6, 11011000b
5760     vextracti128    xm7, m6, 1
5761     movu            [r2 + r3 * 2], xm6
5762     movu            [r2 + r6], xm7
5763     lea             r2, [r2 + r3 * 4]
5764
5765     movu            xm1, [r0 + r4]                  ; m1 = row 15
5766     punpckhbw       xm2, xm0, xm1
5767     punpcklbw       xm0, xm1
5768     vinserti128     m0, m0, xm2, 1
5769     pmaddubsw       m2, m0, [r5 + 3 * mmsize]
5770     paddw           m8, m2
5771     pmaddubsw       m2, m0, [r5 + 2 * mmsize]
5772     paddw           m10, m2
5773     pmaddubsw       m2, m0, [r5 + 1 * mmsize]
5774     paddw           m12, m2
5775     pmaddubsw       m0, [r5]
5776     lea             r0, [r0 + r1 * 4]
5777     movu            xm2, [r0]                       ; m2 = row 16
5778     punpckhbw       xm3, xm1, xm2
5779     punpcklbw       xm1, xm2
5780     vinserti128     m1, m1, xm3, 1
5781     pmaddubsw       m3, m1, [r5 + 3 * mmsize]
5782     paddw           m9, m3
5783     pmaddubsw       m3, m1, [r5 + 2 * mmsize]
5784     paddw           m11, m3
5785     pmaddubsw       m3, m1, [r5 + 1 * mmsize]
5786     paddw           m13, m3
5787     pmaddubsw       m1, [r5]
5788     movu            xm3, [r0 + r1]                  ; m3 = row 17
5789     punpckhbw       xm4, xm2, xm3
5790     punpcklbw       xm2, xm3
5791     vinserti128     m2, m2, xm4, 1
5792     pmaddubsw       m4, m2, [r5 + 3 * mmsize]
5793     paddw           m10, m4
5794     pmaddubsw       m4, m2, [r5 + 2 * mmsize]
5795     paddw           m12, m4
5796     pmaddubsw       m2, [r5 + 1 * mmsize]
5797     paddw           m0, m2
5798     movu            xm4, [r0 + r1 * 2]              ; m4 = row 18
5799     punpckhbw       xm5, xm3, xm4
5800     punpcklbw       xm3, xm4
5801     vinserti128     m3, m3, xm5, 1
5802     pmaddubsw       m5, m3, [r5 + 3 * mmsize]
5803     paddw           m11, m5
5804     pmaddubsw       m5, m3, [r5 + 2 * mmsize]
5805     paddw           m13, m5
5806     pmaddubsw       m3, [r5 + 1 * mmsize]
5807     paddw           m1, m3
5808     movu            xm5, [r0 + r4]                  ; m5 = row 19
5809     punpckhbw       xm6, xm4, xm5
5810     punpcklbw       xm4, xm5
5811     vinserti128     m4, m4, xm6, 1
5812     pmaddubsw       m6, m4, [r5 + 3 * mmsize]
5813     paddw           m12, m6
5814     pmaddubsw       m4, [r5 + 2 * mmsize]
5815     paddw           m0, m4
5816     lea             r0, [r0 + r1 * 4]
5817     movu            xm6, [r0]                       ; m6 = row 20
5818     punpckhbw       xm7, xm5, xm6
5819     punpcklbw       xm5, xm6
5820     vinserti128     m5, m5, xm7, 1
5821     pmaddubsw       m7, m5, [r5 + 3 * mmsize]
5822     paddw           m13, m7
5823     pmaddubsw       m5, [r5 + 2 * mmsize]
5824     paddw           m1, m5
5825     movu            xm7, [r0 + r1]                  ; m7 = row 21
5826     punpckhbw       xm2, xm6, xm7
5827     punpcklbw       xm6, xm7
5828     vinserti128     m6, m6, xm2, 1
5829     pmaddubsw       m6, [r5 + 3 * mmsize]
5830     paddw           m0, m6
5831     movu            xm2, [r0 + r1 * 2]              ; m2 = row 22
5832     punpckhbw       xm3, xm7, xm2
5833     punpcklbw       xm7, xm2
5834     vinserti128     m7, m7, xm3, 1
5835     pmaddubsw       m7, [r5 + 3 * mmsize]
5836     paddw           m1, m7
5837
5838     pmulhrsw        m8, m14                         ; m8 = word: row 8
5839     pmulhrsw        m9, m14                         ; m9 = word: row 9
5840     pmulhrsw        m10, m14                        ; m10 = word: row 10
5841     pmulhrsw        m11, m14                        ; m11 = word: row 11
5842     pmulhrsw        m12, m14                        ; m12 = word: row 12
5843     pmulhrsw        m13, m14                        ; m13 = word: row 13
5844     pmulhrsw        m0, m14                         ; m0 = word: row 14
5845     pmulhrsw        m1, m14                         ; m1 = word: row 15
5846     packuswb        m8, m9
5847     packuswb        m10, m11
5848     packuswb        m12, m13
5849     packuswb        m0, m1
5850     vpermq          m8, m8, 11011000b
5851     vpermq          m10, m10, 11011000b
5852     vpermq          m12, m12, 11011000b
5853     vpermq          m0, m0, 11011000b
5854     vextracti128    xm9, m8, 1
5855     vextracti128    xm11, m10, 1
5856     vextracti128    xm13, m12, 1
5857     vextracti128    xm1, m0, 1
5858     movu            [r2], xm8
5859     movu            [r2 + r3], xm9
5860     movu            [r2 + r3 * 2], xm10
5861     movu            [r2 + r6], xm11
5862     lea             r2, [r2 + r3 * 4]
5863     movu            [r2], xm12
5864     movu            [r2 + r3], xm13
5865     movu            [r2 + r3 * 2], xm0
5866     movu            [r2 + r6], xm1
5867     RET
5868 %endif
5869
5870 INIT_YMM avx2
5871 %if ARCH_X86_64 == 1
5872 cglobal interp_8tap_vert_pp_16x12, 4, 7, 15
5873     mov             r4d, r4m
5874     shl             r4d, 7
5875
5876 %ifdef PIC
5877     lea             r5, [tab_LumaCoeffVer_32]
5878     add             r5, r4
5879 %else
5880     lea             r5, [tab_LumaCoeffVer_32 + r4]
5881 %endif
5882
5883     lea             r4, [r1 * 3]
5884     sub             r0, r4
5885     lea             r6, [r3 * 3]
5886     mova            m14, [pw_512]
5887
5888     movu            xm0, [r0]                       ; m0 = row 0
5889     movu            xm1, [r0 + r1]                  ; m1 = row 1
5890     punpckhbw       xm2, xm0, xm1
5891     punpcklbw       xm0, xm1
5892     vinserti128     m0, m0, xm2, 1
5893     pmaddubsw       m0, [r5]
5894     movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
5895     punpckhbw       xm3, xm1, xm2
5896     punpcklbw       xm1, xm2
5897     vinserti128     m1, m1, xm3, 1
5898     pmaddubsw       m1, [r5]
5899     movu            xm3, [r0 + r4]                  ; m3 = row 3
5900     punpckhbw       xm4, xm2, xm3
5901     punpcklbw       xm2, xm3
5902     vinserti128     m2, m2, xm4, 1
5903     pmaddubsw       m4, m2, [r5 + 1 * mmsize]
5904     paddw           m0, m4
5905     pmaddubsw       m2, [r5]
5906     lea             r0, [r0 + r1 * 4]
5907     movu            xm4, [r0]                       ; m4 = row 4
5908     punpckhbw       xm5, xm3, xm4
5909     punpcklbw       xm3, xm4
5910     vinserti128     m3, m3, xm5, 1
5911     pmaddubsw       m5, m3, [r5 + 1 * mmsize]
5912     paddw           m1, m5
5913     pmaddubsw       m3, [r5]
5914     movu            xm5, [r0 + r1]                  ; m5 = row 5
5915     punpckhbw       xm6, xm4, xm5
5916     punpcklbw       xm4, xm5
5917     vinserti128     m4, m4, xm6, 1
5918     pmaddubsw       m6, m4, [r5 + 2 * mmsize]
5919     paddw           m0, m6
5920     pmaddubsw       m6, m4, [r5 + 1 * mmsize]
5921     paddw           m2, m6
5922     pmaddubsw       m4, [r5]
5923     movu            xm6, [r0 + r1 * 2]              ; m6 = row 6
5924     punpckhbw       xm7, xm5, xm6
5925     punpcklbw       xm5, xm6
5926     vinserti128     m5, m5, xm7, 1
5927     pmaddubsw       m7, m5, [r5 + 2 * mmsize]
5928     paddw           m1, m7
5929     pmaddubsw       m7, m5, [r5 + 1 * mmsize]
5930     paddw           m3, m7
5931     pmaddubsw       m5, [r5]
5932     movu            xm7, [r0 + r4]                  ; m7 = row 7
5933     punpckhbw       xm8, xm6, xm7
5934     punpcklbw       xm6, xm7
5935     vinserti128     m6, m6, xm8, 1
5936     pmaddubsw       m8, m6, [r5 + 3 * mmsize]
5937     paddw           m0, m8
5938     pmaddubsw       m8, m6, [r5 + 2 * mmsize]
5939     paddw           m2, m8
5940     pmaddubsw       m8, m6, [r5 + 1 * mmsize]
5941     paddw           m4, m8
5942     pmaddubsw       m6, [r5]
5943     lea             r0, [r0 + r1 * 4]
5944     movu            xm8, [r0]                       ; m8 = row 8
5945     punpckhbw       xm9, xm7, xm8
5946     punpcklbw       xm7, xm8
5947     vinserti128     m7, m7, xm9, 1
5948     pmaddubsw       m9, m7, [r5 + 3 * mmsize]
5949     paddw           m1, m9
5950     pmaddubsw       m9, m7, [r5 + 2 * mmsize]
5951     paddw           m3, m9
5952     pmaddubsw       m9, m7, [r5 + 1 * mmsize]
5953     paddw           m5, m9
5954     pmaddubsw       m7, [r5]
5955     movu            xm9, [r0 + r1]                  ; m9 = row 9
5956     punpckhbw       xm10, xm8, xm9
5957     punpcklbw       xm8, xm9
5958     vinserti128     m8, m8, xm10, 1
5959     pmaddubsw       m10, m8, [r5 + 3 * mmsize]
5960     paddw           m2, m10
5961     pmaddubsw       m10, m8, [r5 + 2 * mmsize]
5962     paddw           m4, m10
5963     pmaddubsw       m10, m8, [r5 + 1 * mmsize]
5964     paddw           m6, m10
5965     pmaddubsw       m8, [r5]
5966     movu            xm10, [r0 + r1 * 2]             ; m10 = row 10
5967     punpckhbw       xm11, xm9, xm10
5968     punpcklbw       xm9, xm10
5969     vinserti128     m9, m9, xm11, 1
5970     pmaddubsw       m11, m9, [r5 + 3 * mmsize]
5971     paddw           m3, m11
5972     pmaddubsw       m11, m9, [r5 + 2 * mmsize]
5973     paddw           m5, m11
5974     pmaddubsw       m11, m9, [r5 + 1 * mmsize]
5975     paddw           m7, m11
5976     pmaddubsw       m9, [r5]
5977     movu            xm11, [r0 + r4]                 ; m11 = row 11
5978     punpckhbw       xm12, xm10, xm11
5979     punpcklbw       xm10, xm11
5980     vinserti128     m10, m10, xm12, 1
5981     pmaddubsw       m12, m10, [r5 + 3 * mmsize]
5982     paddw           m4, m12
5983     pmaddubsw       m12, m10, [r5 + 2 * mmsize]
5984     paddw           m6, m12
5985     pmaddubsw       m12, m10, [r5 + 1 * mmsize]
5986     paddw           m8, m12
5987     pmaddubsw       m10, [r5]
5988     lea             r0, [r0 + r1 * 4]
5989     movu            xm12, [r0]                      ; m12 = row 12
5990     punpckhbw       xm13, xm11, xm12
5991     punpcklbw       xm11, xm12
5992     vinserti128     m11, m11, xm13, 1
5993     pmaddubsw       m13, m11, [r5 + 3 * mmsize]
5994     paddw           m5, m13
5995     pmaddubsw       m13, m11, [r5 + 2 * mmsize]
5996     paddw           m7, m13
5997     pmaddubsw       m13, m11, [r5 + 1 * mmsize]
5998     paddw           m9, m13
5999     pmaddubsw       m11, [r5]
6000
6001     pmulhrsw        m0, m14                         ; m0 = word: row 0
6002     pmulhrsw        m1, m14                         ; m1 = word: row 1
6003     pmulhrsw        m2, m14                         ; m2 = word: row 2
6004     pmulhrsw        m3, m14                         ; m3 = word: row 3
6005     pmulhrsw        m4, m14                         ; m4 = word: row 4
6006     pmulhrsw        m5, m14                         ; m5 = word: row 5
6007     packuswb        m0, m1
6008     packuswb        m2, m3
6009     packuswb        m4, m5
6010     vpermq          m0, m0, 11011000b
6011     vpermq          m2, m2, 11011000b
6012     vpermq          m4, m4, 11011000b
6013     vextracti128    xm1, m0, 1
6014     vextracti128    xm3, m2, 1
6015     vextracti128    xm5, m4, 1
6016     movu            [r2], xm0
6017     movu            [r2 + r3], xm1
6018     movu            [r2 + r3 * 2], xm2
6019     movu            [r2 + r6], xm3
6020     lea             r2, [r2 + r3 * 4]
6021     movu            [r2], xm4
6022     movu            [r2 + r3], xm5
6023
6024     movu            xm13, [r0 + r1]                 ; m13 = row 13
6025     punpckhbw       xm0, xm12, xm13
6026     punpcklbw       xm12, xm13
6027     vinserti128     m12, m12, xm0, 1
6028     pmaddubsw       m0, m12, [r5 + 3 * mmsize]
6029     paddw           m6, m0
6030     pmaddubsw       m0, m12, [r5 + 2 * mmsize]
6031     paddw           m8, m0
6032     pmaddubsw       m0, m12, [r5 + 1 * mmsize]
6033     paddw           m10, m0
6034     movu            xm0, [r0 + r1 * 2]              ; m0 = row 14
6035     punpckhbw       xm1, xm13, xm0
6036     punpcklbw       xm13, xm0
6037     vinserti128     m13, m13, xm1, 1
6038     pmaddubsw       m1, m13, [r5 + 3 * mmsize]
6039     paddw           m7, m1
6040     pmaddubsw       m1, m13, [r5 + 2 * mmsize]
6041     paddw           m9, m1
6042     pmaddubsw       m1, m13, [r5 + 1 * mmsize]
6043     paddw           m11, m1
6044
6045     pmulhrsw        m6, m14                         ; m6 = word: row 6
6046     pmulhrsw        m7, m14                         ; m7 = word: row 7
6047     packuswb        m6, m7
6048     vpermq          m6, m6, 11011000b
6049     vextracti128    xm7, m6, 1
6050     movu            [r2 + r3 * 2], xm6
6051     movu            [r2 + r6], xm7
6052     lea             r2, [r2 + r3 * 4]
6053
6054     movu            xm1, [r0 + r4]                  ; m1 = row 15
6055     punpckhbw       xm2, xm0, xm1
6056     punpcklbw       xm0, xm1
6057     vinserti128     m0, m0, xm2, 1
6058     pmaddubsw       m2, m0, [r5 + 3 * mmsize]
6059     paddw           m8, m2
6060     pmaddubsw       m2, m0, [r5 + 2 * mmsize]
6061     paddw           m10, m2
6062     lea             r0, [r0 + r1 * 4]
6063     movu            xm2, [r0]                       ; m2 = row 16
6064     punpckhbw       xm3, xm1, xm2
6065     punpcklbw       xm1, xm2
6066     vinserti128     m1, m1, xm3, 1
6067     pmaddubsw       m3, m1, [r5 + 3 * mmsize]
6068     paddw           m9, m3
6069     pmaddubsw       m3, m1, [r5 + 2 * mmsize]
6070     paddw           m11, m3
6071     movu            xm3, [r0 + r1]                  ; m3 = row 17
6072     punpckhbw       xm4, xm2, xm3
6073     punpcklbw       xm2, xm3
6074     vinserti128     m2, m2, xm4, 1
6075     pmaddubsw       m4, m2, [r5 + 3 * mmsize]
6076     paddw           m10, m4
6077     movu            xm4, [r0 + r1 * 2]              ; m4 = row 18
6078     punpckhbw       xm5, xm3, xm4
6079     punpcklbw       xm3, xm4
6080     vinserti128     m3, m3, xm5, 1
6081     pmaddubsw       m5, m3, [r5 + 3 * mmsize]
6082     paddw           m11, m5
6083
6084     pmulhrsw        m8, m14                         ; m8 = word: row 8
6085     pmulhrsw        m9, m14                         ; m9 = word: row 9
6086     pmulhrsw        m10, m14                        ; m10 = word: row 10
6087     pmulhrsw        m11, m14                        ; m11 = word: row 11
6088     packuswb        m8, m9
6089     packuswb        m10, m11
6090     vpermq          m8, m8, 11011000b
6091     vpermq          m10, m10, 11011000b
6092     vextracti128    xm9, m8, 1
6093     vextracti128    xm11, m10, 1
6094     movu            [r2], xm8
6095     movu            [r2 + r3], xm9
6096     movu            [r2 + r3 * 2], xm10
6097     movu            [r2 + r6], xm11
6098     RET
6099 %endif
6100
6101 INIT_YMM avx2
6102 %if ARCH_X86_64 == 1
6103 cglobal interp_8tap_vert_pp_16x8, 4, 7, 15
6104     mov             r4d, r4m
6105     shl             r4d, 7
6106
6107 %ifdef PIC
6108     lea             r5, [tab_LumaCoeffVer_32]
6109     add             r5, r4
6110 %else
6111     lea             r5, [tab_LumaCoeffVer_32 + r4]
6112 %endif
6113
6114     lea             r4, [r1 * 3]
6115     sub             r0, r4
6116     lea             r6, [r3 * 3]
6117     mova            m14, [pw_512]
6118
6119     movu            xm0, [r0]                       ; m0 = row 0
6120     movu            xm1, [r0 + r1]                  ; m1 = row 1
6121     punpckhbw       xm2, xm0, xm1
6122     punpcklbw       xm0, xm1
6123     vinserti128     m0, m0, xm2, 1
6124     pmaddubsw       m0, [r5]
6125     movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
6126     punpckhbw       xm3, xm1, xm2
6127     punpcklbw       xm1, xm2
6128     vinserti128     m1, m1, xm3, 1
6129     pmaddubsw       m1, [r5]
6130     movu            xm3, [r0 + r4]                  ; m3 = row 3
6131     punpckhbw       xm4, xm2, xm3
6132     punpcklbw       xm2, xm3
6133     vinserti128     m2, m2, xm4, 1
6134     pmaddubsw       m4, m2, [r5 + 1 * mmsize]
6135     paddw           m0, m4
6136     pmaddubsw       m2, [r5]
6137     lea             r0, [r0 + r1 * 4]
6138     movu            xm4, [r0]                       ; m4 = row 4
6139     punpckhbw       xm5, xm3, xm4
6140     punpcklbw       xm3, xm4
6141     vinserti128     m3, m3, xm5, 1
6142     pmaddubsw       m5, m3, [r5 + 1 * mmsize]
6143     paddw           m1, m5
6144     pmaddubsw       m3, [r5]
6145     movu            xm5, [r0 + r1]                  ; m5 = row 5
6146     punpckhbw       xm6, xm4, xm5
6147     punpcklbw       xm4, xm5
6148     vinserti128     m4, m4, xm6, 1
6149     pmaddubsw       m6, m4, [r5 + 2 * mmsize]
6150     paddw           m0, m6
6151     pmaddubsw       m6, m4, [r5 + 1 * mmsize]
6152     paddw           m2, m6
6153     pmaddubsw       m4, [r5]
6154     movu            xm6, [r0 + r1 * 2]              ; m6 = row 6
6155     punpckhbw       xm7, xm5, xm6
6156     punpcklbw       xm5, xm6
6157     vinserti128     m5, m5, xm7, 1
6158     pmaddubsw       m7, m5, [r5 + 2 * mmsize]
6159     paddw           m1, m7
6160     pmaddubsw       m7, m5, [r5 + 1 * mmsize]
6161     paddw           m3, m7
6162     pmaddubsw       m5, [r5]
6163     movu            xm7, [r0 + r4]                  ; m7 = row 7
6164     punpckhbw       xm8, xm6, xm7
6165     punpcklbw       xm6, xm7
6166     vinserti128     m6, m6, xm8, 1
6167     pmaddubsw       m8, m6, [r5 + 3 * mmsize]
6168     paddw           m0, m8
6169     pmaddubsw       m8, m6, [r5 + 2 * mmsize]
6170     paddw           m2, m8
6171     pmaddubsw       m8, m6, [r5 + 1 * mmsize]
6172     paddw           m4, m8
6173     pmaddubsw       m6, [r5]
6174     lea             r0, [r0 + r1 * 4]
6175     movu            xm8, [r0]                       ; m8 = row 8
6176     punpckhbw       xm9, xm7, xm8
6177     punpcklbw       xm7, xm8
6178     vinserti128     m7, m7, xm9, 1
6179     pmaddubsw       m9, m7, [r5 + 3 * mmsize]
6180     paddw           m1, m9
6181     pmaddubsw       m9, m7, [r5 + 2 * mmsize]
6182     paddw           m3, m9
6183     pmaddubsw       m9, m7, [r5 + 1 * mmsize]
6184     paddw           m5, m9
6185     pmaddubsw       m7, [r5]
6186     movu            xm9, [r0 + r1]                  ; m9 = row 9
6187     punpckhbw       xm10, xm8, xm9
6188     punpcklbw       xm8, xm9
6189     vinserti128     m8, m8, xm10, 1
6190     pmaddubsw       m10, m8, [r5 + 3 * mmsize]
6191     paddw           m2, m10
6192     pmaddubsw       m10, m8, [r5 + 2 * mmsize]
6193     paddw           m4, m10
6194     pmaddubsw       m10, m8, [r5 + 1 * mmsize]
6195     paddw           m6, m10
6196     movu            xm10, [r0 + r1 * 2]             ; m10 = row 10
6197     punpckhbw       xm11, xm9, xm10
6198     punpcklbw       xm9, xm10
6199     vinserti128     m9, m9, xm11, 1
6200     pmaddubsw       m11, m9, [r5 + 3 * mmsize]
6201     paddw           m3, m11
6202     pmaddubsw       m11, m9, [r5 + 2 * mmsize]
6203     paddw           m5, m11
6204     pmaddubsw       m11, m9, [r5 + 1 * mmsize]
6205     paddw           m7, m11
6206     movu            xm11, [r0 + r4]                 ; m11 = row 11
6207     punpckhbw       xm12, xm10, xm11
6208     punpcklbw       xm10, xm11
6209     vinserti128     m10, m10, xm12, 1
6210     pmaddubsw       m12, m10, [r5 + 3 * mmsize]
6211     paddw           m4, m12
6212     pmaddubsw       m12, m10, [r5 + 2 * mmsize]
6213     paddw           m6, m12
6214     lea             r0, [r0 + r1 * 4]
6215     movu            xm12, [r0]                      ; m12 = row 12
6216     punpckhbw       xm13, xm11, xm12
6217     punpcklbw       xm11, xm12
6218     vinserti128     m11, m11, xm13, 1
6219     pmaddubsw       m13, m11, [r5 + 3 * mmsize]
6220     paddw           m5, m13
6221     pmaddubsw       m13, m11, [r5 + 2 * mmsize]
6222     paddw           m7, m13
6223
6224     pmulhrsw        m0, m14                         ; m0 = word: row 0
6225     pmulhrsw        m1, m14                         ; m1 = word: row 1
6226     pmulhrsw        m2, m14                         ; m2 = word: row 2
6227     pmulhrsw        m3, m14                         ; m3 = word: row 3
6228     pmulhrsw        m4, m14                         ; m4 = word: row 4
6229     pmulhrsw        m5, m14                         ; m5 = word: row 5
6230     packuswb        m0, m1
6231     packuswb        m2, m3
6232     packuswb        m4, m5
6233     vpermq          m0, m0, 11011000b
6234     vpermq          m2, m2, 11011000b
6235     vpermq          m4, m4, 11011000b
6236     vextracti128    xm1, m0, 1
6237     vextracti128    xm3, m2, 1
6238     vextracti128    xm5, m4, 1
6239     movu            [r2], xm0
6240     movu            [r2 + r3], xm1
6241     movu            [r2 + r3 * 2], xm2
6242     movu            [r2 + r6], xm3
6243     lea             r2, [r2 + r3 * 4]
6244     movu            [r2], xm4
6245     movu            [r2 + r3], xm5
6246
6247     movu            xm13, [r0 + r1]                 ; m13 = row 13
6248     punpckhbw       xm0, xm12, xm13
6249     punpcklbw       xm12, xm13
6250     vinserti128     m12, m12, xm0, 1
6251     pmaddubsw       m0, m12, [r5 + 3 * mmsize]
6252     paddw           m6, m0
6253     movu            xm0, [r0 + r1 * 2]              ; m0 = row 14
6254     punpckhbw       xm1, xm13, xm0
6255     punpcklbw       xm13, xm0
6256     vinserti128     m13, m13, xm1, 1
6257     pmaddubsw       m1, m13, [r5 + 3 * mmsize]
6258     paddw           m7, m1
6259
6260     pmulhrsw        m6, m14                         ; m6 = word: row 6
6261     pmulhrsw        m7, m14                         ; m7 = word: row 7
6262     packuswb        m6, m7
6263     vpermq          m6, m6, 11011000b
6264     vextracti128    xm7, m6, 1
6265     movu            [r2 + r3 * 2], xm6
6266     movu            [r2 + r6], xm7
6267     RET
6268 %endif
6269
6270 INIT_YMM avx2
6271 %if ARCH_X86_64 == 1
6272 cglobal interp_8tap_vert_pp_16x4, 4, 7, 13
6273     mov             r4d, r4m
6274     shl             r4d, 7
6275
6276 %ifdef PIC
6277     lea             r5, [tab_LumaCoeffVer_32]
6278     add             r5, r4
6279 %else
6280     lea             r5, [tab_LumaCoeffVer_32 + r4]
6281 %endif
6282
6283     lea             r4, [r1 * 3]
6284     sub             r0, r4
6285     lea             r6, [r3 * 3]
6286     mova            m12, [pw_512]
6287
6288     movu            xm0, [r0]                       ; m0 = row 0
6289     movu            xm1, [r0 + r1]                  ; m1 = row 1
6290     punpckhbw       xm2, xm0, xm1
6291     punpcklbw       xm0, xm1
6292     vinserti128     m0, m0, xm2, 1
6293     pmaddubsw       m0, [r5]
6294     movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
6295     punpckhbw       xm3, xm1, xm2
6296     punpcklbw       xm1, xm2
6297     vinserti128     m1, m1, xm3, 1
6298     pmaddubsw       m1, [r5]
6299     movu            xm3, [r0 + r4]                  ; m3 = row 3
6300     punpckhbw       xm4, xm2, xm3
6301     punpcklbw       xm2, xm3
6302     vinserti128     m2, m2, xm4, 1
6303     pmaddubsw       m4, m2, [r5 + 1 * mmsize]
6304     paddw           m0, m4
6305     pmaddubsw       m2, [r5]
6306     lea             r0, [r0 + r1 * 4]
6307     movu            xm4, [r0]                       ; m4 = row 4
6308     punpckhbw       xm5, xm3, xm4
6309     punpcklbw       xm3, xm4
6310     vinserti128     m3, m3, xm5, 1
6311     pmaddubsw       m5, m3, [r5 + 1 * mmsize]
6312     paddw           m1, m5
6313     pmaddubsw       m3, [r5]
6314     movu            xm5, [r0 + r1]                  ; m5 = row 5
6315     punpckhbw       xm6, xm4, xm5
6316     punpcklbw       xm4, xm5
6317     vinserti128     m4, m4, xm6, 1
6318     pmaddubsw       m6, m4, [r5 + 2 * mmsize]
6319     paddw           m0, m6
6320     pmaddubsw       m6, m4, [r5 + 1 * mmsize]
6321     paddw           m2, m6
6322     movu            xm6, [r0 + r1 * 2]              ; m6 = row 6
6323     punpckhbw       xm7, xm5, xm6
6324     punpcklbw       xm5, xm6
6325     vinserti128     m5, m5, xm7, 1
6326     pmaddubsw       m7, m5, [r5 + 2 * mmsize]
6327     paddw           m1, m7
6328     pmaddubsw       m7, m5, [r5 + 1 * mmsize]
6329     paddw           m3, m7
6330     movu            xm7, [r0 + r4]                  ; m7 = row 7
6331     punpckhbw       xm8, xm6, xm7
6332     punpcklbw       xm6, xm7
6333     vinserti128     m6, m6, xm8, 1
6334     pmaddubsw       m8, m6, [r5 + 3 * mmsize]
6335     paddw           m0, m8
6336     pmaddubsw       m8, m6, [r5 + 2 * mmsize]
6337     paddw           m2, m8
6338     lea             r0, [r0 + r1 * 4]
6339     movu            xm8, [r0]                       ; m8 = row 8
6340     punpckhbw       xm9, xm7, xm8
6341     punpcklbw       xm7, xm8
6342     vinserti128     m7, m7, xm9, 1
6343     pmaddubsw       m9, m7, [r5 + 3 * mmsize]
6344     paddw           m1, m9
6345     pmaddubsw       m9, m7, [r5 + 2 * mmsize]
6346     paddw           m3, m9
6347     movu            xm9, [r0 + r1]                  ; m9 = row 9
6348     punpckhbw       xm10, xm8, xm9
6349     punpcklbw       xm8, xm9
6350     vinserti128     m8, m8, xm10, 1
6351     pmaddubsw       m10, m8, [r5 + 3 * mmsize]
6352     paddw           m2, m10
6353     movu            xm10, [r0 + r1 * 2]             ; m10 = row 10
6354     punpckhbw       xm11, xm9, xm10
6355     punpcklbw       xm9, xm10
6356     vinserti128     m9, m9, xm11, 1
6357     pmaddubsw       m11, m9, [r5 + 3 * mmsize]
6358     paddw           m3, m11
6359
6360     pmulhrsw        m0, m12                         ; m0 = word: row 0
6361     pmulhrsw        m1, m12                         ; m1 = word: row 1
6362     pmulhrsw        m2, m12                         ; m2 = word: row 2
6363     pmulhrsw        m3, m12                         ; m3 = word: row 3
6364     packuswb        m0, m1
6365     packuswb        m2, m3
6366     vpermq          m0, m0, 11011000b
6367     vpermq          m2, m2, 11011000b
6368     vextracti128    xm1, m0, 1
6369     vextracti128    xm3, m2, 1
6370     movu            [r2], xm0
6371     movu            [r2 + r3], xm1
6372     movu            [r2 + r3 * 2], xm2
6373     movu            [r2 + r6], xm3
6374     RET
6375 %endif
6376
6377 %macro FILTER_VER_LUMA_AVX2_16xN 2
6378 INIT_YMM avx2
6379 %if ARCH_X86_64 == 1
6380 cglobal interp_8tap_vert_pp_%1x%2, 4, 9, 15
6381     mov             r4d, r4m
6382     shl             r4d, 7
6383
6384 %ifdef PIC
6385     lea             r5, [tab_LumaCoeffVer_32]
6386     add             r5, r4
6387 %else
6388     lea             r5, [tab_LumaCoeffVer_32 + r4]
6389 %endif
6390
6391     lea             r4, [r1 * 3]
6392     sub             r0, r4
6393     lea             r6, [r3 * 3]
6394     lea             r7, [r1 * 4]
6395     mova            m14, [pw_512]
6396     mov             r8d, %2 / 16
6397
6398 .loop:
6399     movu            xm0, [r0]                       ; m0 = row 0
6400     movu            xm1, [r0 + r1]                  ; m1 = row 1
6401     punpckhbw       xm2, xm0, xm1
6402     punpcklbw       xm0, xm1
6403     vinserti128     m0, m0, xm2, 1
6404     pmaddubsw       m0, [r5]
6405     movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
6406     punpckhbw       xm3, xm1, xm2
6407     punpcklbw       xm1, xm2
6408     vinserti128     m1, m1, xm3, 1
6409     pmaddubsw       m1, [r5]
6410     movu            xm3, [r0 + r4]                  ; m3 = row 3
6411     punpckhbw       xm4, xm2, xm3
6412     punpcklbw       xm2, xm3
6413     vinserti128     m2, m2, xm4, 1
6414     pmaddubsw       m4, m2, [r5 + 1 * mmsize]
6415     paddw           m0, m4
6416     pmaddubsw       m2, [r5]
6417     lea             r0, [r0 + r1 * 4]
6418     movu            xm4, [r0]                       ; m4 = row 4
6419     punpckhbw       xm5, xm3, xm4
6420     punpcklbw       xm3, xm4
6421     vinserti128     m3, m3, xm5, 1
6422     pmaddubsw       m5, m3, [r5 + 1 * mmsize]
6423     paddw           m1, m5
6424     pmaddubsw       m3, [r5]
6425     movu            xm5, [r0 + r1]                  ; m5 = row 5
6426     punpckhbw       xm6, xm4, xm5
6427     punpcklbw       xm4, xm5
6428     vinserti128     m4, m4, xm6, 1
6429     pmaddubsw       m6, m4, [r5 + 2 * mmsize]
6430     paddw           m0, m6
6431     pmaddubsw       m6, m4, [r5 + 1 * mmsize]
6432     paddw           m2, m6
6433     pmaddubsw       m4, [r5]
6434     movu            xm6, [r0 + r1 * 2]              ; m6 = row 6
6435     punpckhbw       xm7, xm5, xm6
6436     punpcklbw       xm5, xm6
6437     vinserti128     m5, m5, xm7, 1
6438     pmaddubsw       m7, m5, [r5 + 2 * mmsize]
6439     paddw           m1, m7
6440     pmaddubsw       m7, m5, [r5 + 1 * mmsize]
6441     paddw           m3, m7
6442     pmaddubsw       m5, [r5]
6443     movu            xm7, [r0 + r4]                  ; m7 = row 7
6444     punpckhbw       xm8, xm6, xm7
6445     punpcklbw       xm6, xm7
6446     vinserti128     m6, m6, xm8, 1
6447     pmaddubsw       m8, m6, [r5 + 3 * mmsize]
6448     paddw           m0, m8
6449     pmaddubsw       m8, m6, [r5 + 2 * mmsize]
6450     paddw           m2, m8
6451     pmaddubsw       m8, m6, [r5 + 1 * mmsize]
6452     paddw           m4, m8
6453     pmaddubsw       m6, [r5]
6454     lea             r0, [r0 + r1 * 4]
6455     movu            xm8, [r0]                       ; m8 = row 8
6456     punpckhbw       xm9, xm7, xm8
6457     punpcklbw       xm7, xm8
6458     vinserti128     m7, m7, xm9, 1
6459     pmaddubsw       m9, m7, [r5 + 3 * mmsize]
6460     paddw           m1, m9
6461     pmaddubsw       m9, m7, [r5 + 2 * mmsize]
6462     paddw           m3, m9
6463     pmaddubsw       m9, m7, [r5 + 1 * mmsize]
6464     paddw           m5, m9
6465     pmaddubsw       m7, [r5]
6466     movu            xm9, [r0 + r1]                  ; m9 = row 9
6467     punpckhbw       xm10, xm8, xm9
6468     punpcklbw       xm8, xm9
6469     vinserti128     m8, m8, xm10, 1
6470     pmaddubsw       m10, m8, [r5 + 3 * mmsize]
6471     paddw           m2, m10
6472     pmaddubsw       m10, m8, [r5 + 2 * mmsize]
6473     paddw           m4, m10
6474     pmaddubsw       m10, m8, [r5 + 1 * mmsize]
6475     paddw           m6, m10
6476     pmaddubsw       m8, [r5]
6477     movu            xm10, [r0 + r1 * 2]             ; m10 = row 10
6478     punpckhbw       xm11, xm9, xm10
6479     punpcklbw       xm9, xm10
6480     vinserti128     m9, m9, xm11, 1
6481     pmaddubsw       m11, m9, [r5 + 3 * mmsize]
6482     paddw           m3, m11
6483     pmaddubsw       m11, m9, [r5 + 2 * mmsize]
6484     paddw           m5, m11
6485     pmaddubsw       m11, m9, [r5 + 1 * mmsize]
6486     paddw           m7, m11
6487     pmaddubsw       m9, [r5]
6488     movu            xm11, [r0 + r4]                 ; m11 = row 11
6489     punpckhbw       xm12, xm10, xm11
6490     punpcklbw       xm10, xm11
6491     vinserti128     m10, m10, xm12, 1
6492     pmaddubsw       m12, m10, [r5 + 3 * mmsize]
6493     paddw           m4, m12
6494     pmaddubsw       m12, m10, [r5 + 2 * mmsize]
6495     paddw           m6, m12
6496     pmaddubsw       m12, m10, [r5 + 1 * mmsize]
6497     paddw           m8, m12
6498     pmaddubsw       m10, [r5]
6499     lea             r0, [r0 + r1 * 4]
6500     movu            xm12, [r0]                      ; m12 = row 12
6501     punpckhbw       xm13, xm11, xm12
6502     punpcklbw       xm11, xm12
6503     vinserti128     m11, m11, xm13, 1
6504     pmaddubsw       m13, m11, [r5 + 3 * mmsize]
6505     paddw           m5, m13
6506     pmaddubsw       m13, m11, [r5 + 2 * mmsize]
6507     paddw           m7, m13
6508     pmaddubsw       m13, m11, [r5 + 1 * mmsize]
6509     paddw           m9, m13
6510     pmaddubsw       m11, [r5]
6511
6512     pmulhrsw        m0, m14                         ; m0 = word: row 0
6513     pmulhrsw        m1, m14                         ; m1 = word: row 1
6514     pmulhrsw        m2, m14                         ; m2 = word: row 2
6515     pmulhrsw        m3, m14                         ; m3 = word: row 3
6516     pmulhrsw        m4, m14                         ; m4 = word: row 4
6517     pmulhrsw        m5, m14                         ; m5 = word: row 5
6518     packuswb        m0, m1
6519     packuswb        m2, m3
6520     packuswb        m4, m5
6521     vpermq          m0, m0, 11011000b
6522     vpermq          m2, m2, 11011000b
6523     vpermq          m4, m4, 11011000b
6524     vextracti128    xm1, m0, 1
6525     vextracti128    xm3, m2, 1
6526     vextracti128    xm5, m4, 1
6527     movu            [r2], xm0
6528     movu            [r2 + r3], xm1
6529     movu            [r2 + r3 * 2], xm2
6530     movu            [r2 + r6], xm3
6531     lea             r2, [r2 + r3 * 4]
6532     movu            [r2], xm4
6533     movu            [r2 + r3], xm5
6534
6535     movu            xm13, [r0 + r1]                 ; m13 = row 13
6536     punpckhbw       xm0, xm12, xm13
6537     punpcklbw       xm12, xm13
6538     vinserti128     m12, m12, xm0, 1
6539     pmaddubsw       m0, m12, [r5 + 3 * mmsize]
6540     paddw           m6, m0
6541     pmaddubsw       m0, m12, [r5 + 2 * mmsize]
6542     paddw           m8, m0
6543     pmaddubsw       m0, m12, [r5 + 1 * mmsize]
6544     paddw           m10, m0
6545     pmaddubsw       m12, [r5]
6546     movu            xm0, [r0 + r1 * 2]              ; m0 = row 14
6547     punpckhbw       xm1, xm13, xm0
6548     punpcklbw       xm13, xm0
6549     vinserti128     m13, m13, xm1, 1
6550     pmaddubsw       m1, m13, [r5 + 3 * mmsize]
6551     paddw           m7, m1
6552     pmaddubsw       m1, m13, [r5 + 2 * mmsize]
6553     paddw           m9, m1
6554     pmaddubsw       m1, m13, [r5 + 1 * mmsize]
6555     paddw           m11, m1
6556     pmaddubsw       m13, [r5]
6557
6558     pmulhrsw        m6, m14                         ; m6 = word: row 6
6559     pmulhrsw        m7, m14                         ; m7 = word: row 7
6560     packuswb        m6, m7
6561     vpermq          m6, m6, 11011000b
6562     vextracti128    xm7, m6, 1
6563     movu            [r2 + r3 * 2], xm6
6564     movu            [r2 + r6], xm7
6565     lea             r2, [r2 + r3 * 4]
6566
6567     movu            xm1, [r0 + r4]                  ; m1 = row 15
6568     punpckhbw       xm2, xm0, xm1
6569     punpcklbw       xm0, xm1
6570     vinserti128     m0, m0, xm2, 1
6571     pmaddubsw       m2, m0, [r5 + 3 * mmsize]
6572     paddw           m8, m2
6573     pmaddubsw       m2, m0, [r5 + 2 * mmsize]
6574     paddw           m10, m2
6575     pmaddubsw       m2, m0, [r5 + 1 * mmsize]
6576     paddw           m12, m2
6577     pmaddubsw       m0, [r5]
6578     lea             r0, [r0 + r1 * 4]
6579     movu            xm2, [r0]                       ; m2 = row 16
6580     punpckhbw       xm3, xm1, xm2
6581     punpcklbw       xm1, xm2
6582     vinserti128     m1, m1, xm3, 1
6583     pmaddubsw       m3, m1, [r5 + 3 * mmsize]
6584     paddw           m9, m3
6585     pmaddubsw       m3, m1, [r5 + 2 * mmsize]
6586     paddw           m11, m3
6587     pmaddubsw       m3, m1, [r5 + 1 * mmsize]
6588     paddw           m13, m3
6589     pmaddubsw       m1, [r5]
6590     movu            xm3, [r0 + r1]                  ; m3 = row 17
6591     punpckhbw       xm4, xm2, xm3
6592     punpcklbw       xm2, xm3
6593     vinserti128     m2, m2, xm4, 1
6594     pmaddubsw       m4, m2, [r5 + 3 * mmsize]
6595     paddw           m10, m4
6596     pmaddubsw       m4, m2, [r5 + 2 * mmsize]
6597     paddw           m12, m4
6598     pmaddubsw       m2, [r5 + 1 * mmsize]
6599     paddw           m0, m2
6600     movu            xm4, [r0 + r1 * 2]              ; m4 = row 18
6601     punpckhbw       xm5, xm3, xm4
6602     punpcklbw       xm3, xm4
6603     vinserti128     m3, m3, xm5, 1
6604     pmaddubsw       m5, m3, [r5 + 3 * mmsize]
6605     paddw           m11, m5
6606     pmaddubsw       m5, m3, [r5 + 2 * mmsize]
6607     paddw           m13, m5
6608     pmaddubsw       m3, [r5 + 1 * mmsize]
6609     paddw           m1, m3
6610     movu            xm5, [r0 + r4]                  ; m5 = row 19
6611     punpckhbw       xm6, xm4, xm5
6612     punpcklbw       xm4, xm5
6613     vinserti128     m4, m4, xm6, 1
6614     pmaddubsw       m6, m4, [r5 + 3 * mmsize]
6615     paddw           m12, m6
6616     pmaddubsw       m4, [r5 + 2 * mmsize]
6617     paddw           m0, m4
6618     lea             r0, [r0 + r1 * 4]
6619     movu            xm6, [r0]                       ; m6 = row 20
6620     punpckhbw       xm7, xm5, xm6
6621     punpcklbw       xm5, xm6
6622     vinserti128     m5, m5, xm7, 1
6623     pmaddubsw       m7, m5, [r5 + 3 * mmsize]
6624     paddw           m13, m7
6625     pmaddubsw       m5, [r5 + 2 * mmsize]
6626     paddw           m1, m5
6627     movu            xm7, [r0 + r1]                  ; m7 = row 21
6628     punpckhbw       xm2, xm6, xm7
6629     punpcklbw       xm6, xm7
6630     vinserti128     m6, m6, xm2, 1
6631     pmaddubsw       m6, [r5 + 3 * mmsize]
6632     paddw           m0, m6
6633     movu            xm2, [r0 + r1 * 2]              ; m2 = row 22
6634     punpckhbw       xm3, xm7, xm2
6635     punpcklbw       xm7, xm2
6636     vinserti128     m7, m7, xm3, 1
6637     pmaddubsw       m7, [r5 + 3 * mmsize]
6638     paddw           m1, m7
6639
6640     pmulhrsw        m8, m14                         ; m8 = word: row 8
6641     pmulhrsw        m9, m14                         ; m9 = word: row 9
6642     pmulhrsw        m10, m14                        ; m10 = word: row 10
6643     pmulhrsw        m11, m14                        ; m11 = word: row 11
6644     pmulhrsw        m12, m14                        ; m12 = word: row 12
6645     pmulhrsw        m13, m14                        ; m13 = word: row 13
6646     pmulhrsw        m0, m14                         ; m0 = word: row 14
6647     pmulhrsw        m1, m14                         ; m1 = word: row 15
6648     packuswb        m8, m9
6649     packuswb        m10, m11
6650     packuswb        m12, m13
6651     packuswb        m0, m1
6652     vpermq          m8, m8, 11011000b
6653     vpermq          m10, m10, 11011000b
6654     vpermq          m12, m12, 11011000b
6655     vpermq          m0, m0, 11011000b
6656     vextracti128    xm9, m8, 1
6657     vextracti128    xm11, m10, 1
6658     vextracti128    xm13, m12, 1
6659     vextracti128    xm1, m0, 1
6660     movu            [r2], xm8
6661     movu            [r2 + r3], xm9
6662     movu            [r2 + r3 * 2], xm10
6663     movu            [r2 + r6], xm11
6664     lea             r2, [r2 + r3 * 4]
6665     movu            [r2], xm12
6666     movu            [r2 + r3], xm13
6667     movu            [r2 + r3 * 2], xm0
6668     movu            [r2 + r6], xm1
6669     lea             r2, [r2 + r3 * 4]
6670     sub             r0, r7
6671     dec             r8d
6672     jnz             .loop
6673     RET
6674 %endif
6675 %endmacro
6676
6677 FILTER_VER_LUMA_AVX2_16xN 16, 32
6678 FILTER_VER_LUMA_AVX2_16xN 16, 64
6679
6680 %macro PROCESS_LUMA_AVX2_W16_16R 0
6681     movu            xm0, [r0]                       ; m0 = row 0
6682     movu            xm1, [r0 + r1]                  ; m1 = row 1
6683     punpckhbw       xm2, xm0, xm1
6684     punpcklbw       xm0, xm1
6685     vinserti128     m0, m0, xm2, 1
6686     pmaddubsw       m0, [r5]
6687     movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
6688     punpckhbw       xm3, xm1, xm2
6689     punpcklbw       xm1, xm2
6690     vinserti128     m1, m1, xm3, 1
6691     pmaddubsw       m1, [r5]
6692     movu            xm3, [r0 + r4]                  ; m3 = row 3
6693     punpckhbw       xm4, xm2, xm3
6694     punpcklbw       xm2, xm3
6695     vinserti128     m2, m2, xm4, 1
6696     pmaddubsw       m4, m2, [r5 + 1 * mmsize]
6697     paddw           m0, m4
6698     pmaddubsw       m2, [r5]
6699     lea             r7, [r0 + r1 * 4]
6700     movu            xm4, [r7]                       ; m4 = row 4
6701     punpckhbw       xm5, xm3, xm4
6702     punpcklbw       xm3, xm4
6703     vinserti128     m3, m3, xm5, 1
6704     pmaddubsw       m5, m3, [r5 + 1 * mmsize]
6705     paddw           m1, m5
6706     pmaddubsw       m3, [r5]
6707     movu            xm5, [r7 + r1]                  ; m5 = row 5
6708     punpckhbw       xm6, xm4, xm5
6709     punpcklbw       xm4, xm5
6710     vinserti128     m4, m4, xm6, 1
6711     pmaddubsw       m6, m4, [r5 + 2 * mmsize]
6712     paddw           m0, m6
6713     pmaddubsw       m6, m4, [r5 + 1 * mmsize]
6714     paddw           m2, m6
6715     pmaddubsw       m4, [r5]
6716     movu            xm6, [r7 + r1 * 2]              ; m6 = row 6
6717     punpckhbw       xm7, xm5, xm6
6718     punpcklbw       xm5, xm6
6719     vinserti128     m5, m5, xm7, 1
6720     pmaddubsw       m7, m5, [r5 + 2 * mmsize]
6721     paddw           m1, m7
6722     pmaddubsw       m7, m5, [r5 + 1 * mmsize]
6723     paddw           m3, m7
6724     pmaddubsw       m5, [r5]
6725     movu            xm7, [r7 + r4]                  ; m7 = row 7
6726     punpckhbw       xm8, xm6, xm7
6727     punpcklbw       xm6, xm7
6728     vinserti128     m6, m6, xm8, 1
6729     pmaddubsw       m8, m6, [r5 + 3 * mmsize]
6730     paddw           m0, m8
6731     pmaddubsw       m8, m6, [r5 + 2 * mmsize]
6732     paddw           m2, m8
6733     pmaddubsw       m8, m6, [r5 + 1 * mmsize]
6734     paddw           m4, m8
6735     pmaddubsw       m6, [r5]
6736     lea             r7, [r7 + r1 * 4]
6737     movu            xm8, [r7]                       ; m8 = row 8
6738     punpckhbw       xm9, xm7, xm8
6739     punpcklbw       xm7, xm8
6740     vinserti128     m7, m7, xm9, 1
6741     pmaddubsw       m9, m7, [r5 + 3 * mmsize]
6742     paddw           m1, m9
6743     pmaddubsw       m9, m7, [r5 + 2 * mmsize]
6744     paddw           m3, m9
6745     pmaddubsw       m9, m7, [r5 + 1 * mmsize]
6746     paddw           m5, m9
6747     pmaddubsw       m7, [r5]
6748     movu            xm9, [r7 + r1]                  ; m9 = row 9
6749     punpckhbw       xm10, xm8, xm9
6750     punpcklbw       xm8, xm9
6751     vinserti128     m8, m8, xm10, 1
6752     pmaddubsw       m10, m8, [r5 + 3 * mmsize]
6753     paddw           m2, m10
6754     pmaddubsw       m10, m8, [r5 + 2 * mmsize]
6755     paddw           m4, m10
6756     pmaddubsw       m10, m8, [r5 + 1 * mmsize]
6757     paddw           m6, m10
6758     pmaddubsw       m8, [r5]
6759     movu            xm10, [r7 + r1 * 2]             ; m10 = row 10
6760     punpckhbw       xm11, xm9, xm10
6761     punpcklbw       xm9, xm10
6762     vinserti128     m9, m9, xm11, 1
6763     pmaddubsw       m11, m9, [r5 + 3 * mmsize]
6764     paddw           m3, m11
6765     pmaddubsw       m11, m9, [r5 + 2 * mmsize]
6766     paddw           m5, m11
6767     pmaddubsw       m11, m9, [r5 + 1 * mmsize]
6768     paddw           m7, m11
6769     pmaddubsw       m9, [r5]
6770     movu            xm11, [r7 + r4]                 ; m11 = row 11
6771     punpckhbw       xm12, xm10, xm11
6772     punpcklbw       xm10, xm11
6773     vinserti128     m10, m10, xm12, 1
6774     pmaddubsw       m12, m10, [r5 + 3 * mmsize]
6775     paddw           m4, m12
6776     pmaddubsw       m12, m10, [r5 + 2 * mmsize]
6777     paddw           m6, m12
6778     pmaddubsw       m12, m10, [r5 + 1 * mmsize]
6779     paddw           m8, m12
6780     pmaddubsw       m10, [r5]
6781     lea             r7, [r7 + r1 * 4]
6782     movu            xm12, [r7]                      ; m12 = row 12
6783     punpckhbw       xm13, xm11, xm12
6784     punpcklbw       xm11, xm12
6785     vinserti128     m11, m11, xm13, 1
6786     pmaddubsw       m13, m11, [r5 + 3 * mmsize]
6787     paddw           m5, m13
6788     pmaddubsw       m13, m11, [r5 + 2 * mmsize]
6789     paddw           m7, m13
6790     pmaddubsw       m13, m11, [r5 + 1 * mmsize]
6791     paddw           m9, m13
6792     pmaddubsw       m11, [r5]
6793
6794     pmulhrsw        m0, m14                         ; m0 = word: row 0
6795     pmulhrsw        m1, m14                         ; m1 = word: row 1
6796     pmulhrsw        m2, m14                         ; m2 = word: row 2
6797     pmulhrsw        m3, m14                         ; m3 = word: row 3
6798     pmulhrsw        m4, m14                         ; m4 = word: row 4
6799     pmulhrsw        m5, m14                         ; m5 = word: row 5
6800     packuswb        m0, m1
6801     packuswb        m2, m3
6802     packuswb        m4, m5
6803     vpermq          m0, m0, 11011000b
6804     vpermq          m2, m2, 11011000b
6805     vpermq          m4, m4, 11011000b
6806     vextracti128    xm1, m0, 1
6807     vextracti128    xm3, m2, 1
6808     vextracti128    xm5, m4, 1
6809     movu            [r2], xm0
6810     movu            [r2 + r3], xm1
6811     movu            [r2 + r3 * 2], xm2
6812     movu            [r2 + r6], xm3
6813     lea             r8, [r2 + r3 * 4]
6814     movu            [r8], xm4
6815     movu            [r8 + r3], xm5
6816
6817     movu            xm13, [r7 + r1]                 ; m13 = row 13
6818     punpckhbw       xm0, xm12, xm13
6819     punpcklbw       xm12, xm13
6820     vinserti128     m12, m12, xm0, 1
6821     pmaddubsw       m0, m12, [r5 + 3 * mmsize]
6822     paddw           m6, m0
6823     pmaddubsw       m0, m12, [r5 + 2 * mmsize]
6824     paddw           m8, m0
6825     pmaddubsw       m0, m12, [r5 + 1 * mmsize]
6826     paddw           m10, m0
6827     pmaddubsw       m12, [r5]
6828     movu            xm0, [r7 + r1 * 2]              ; m0 = row 14
6829     punpckhbw       xm1, xm13, xm0
6830     punpcklbw       xm13, xm0
6831     vinserti128     m13, m13, xm1, 1
6832     pmaddubsw       m1, m13, [r5 + 3 * mmsize]
6833     paddw           m7, m1
6834     pmaddubsw       m1, m13, [r5 + 2 * mmsize]
6835     paddw           m9, m1
6836     pmaddubsw       m1, m13, [r5 + 1 * mmsize]
6837     paddw           m11, m1
6838     pmaddubsw       m13, [r5]
6839
6840     pmulhrsw        m6, m14                         ; m6 = word: row 6
6841     pmulhrsw        m7, m14                         ; m7 = word: row 7
6842     packuswb        m6, m7
6843     vpermq          m6, m6, 11011000b
6844     vextracti128    xm7, m6, 1
6845     movu            [r8 + r3 * 2], xm6
6846     movu            [r8 + r6], xm7
6847     lea             r8, [r8 + r3 * 4]
6848
6849     movu            xm1, [r7 + r4]                  ; m1 = row 15
6850     punpckhbw       xm2, xm0, xm1
6851     punpcklbw       xm0, xm1
6852     vinserti128     m0, m0, xm2, 1
6853     pmaddubsw       m2, m0, [r5 + 3 * mmsize]
6854     paddw           m8, m2
6855     pmaddubsw       m2, m0, [r5 + 2 * mmsize]
6856     paddw           m10, m2
6857     pmaddubsw       m2, m0, [r5 + 1 * mmsize]
6858     paddw           m12, m2
6859     pmaddubsw       m0, [r5]
6860     lea             r7, [r7 + r1 * 4]
6861     movu            xm2, [r7]                       ; m2 = row 16
6862     punpckhbw       xm3, xm1, xm2
6863     punpcklbw       xm1, xm2
6864     vinserti128     m1, m1, xm3, 1
6865     pmaddubsw       m3, m1, [r5 + 3 * mmsize]
6866     paddw           m9, m3
6867     pmaddubsw       m3, m1, [r5 + 2 * mmsize]
6868     paddw           m11, m3
6869     pmaddubsw       m3, m1, [r5 + 1 * mmsize]
6870     paddw           m13, m3
6871     pmaddubsw       m1, [r5]
6872     movu            xm3, [r7 + r1]                  ; m3 = row 17
6873     punpckhbw       xm4, xm2, xm3
6874     punpcklbw       xm2, xm3
6875     vinserti128     m2, m2, xm4, 1
6876     pmaddubsw       m4, m2, [r5 + 3 * mmsize]
6877     paddw           m10, m4
6878     pmaddubsw       m4, m2, [r5 + 2 * mmsize]
6879     paddw           m12, m4
6880     pmaddubsw       m2, [r5 + 1 * mmsize]
6881     paddw           m0, m2
6882     movu            xm4, [r7 + r1 * 2]              ; m4 = row 18
6883     punpckhbw       xm5, xm3, xm4
6884     punpcklbw       xm3, xm4
6885     vinserti128     m3, m3, xm5, 1
6886     pmaddubsw       m5, m3, [r5 + 3 * mmsize]
6887     paddw           m11, m5
6888     pmaddubsw       m5, m3, [r5 + 2 * mmsize]
6889     paddw           m13, m5
6890     pmaddubsw       m3, [r5 + 1 * mmsize]
6891     paddw           m1, m3
6892     movu            xm5, [r7 + r4]                  ; m5 = row 19
6893     punpckhbw       xm6, xm4, xm5
6894     punpcklbw       xm4, xm5
6895     vinserti128     m4, m4, xm6, 1
6896     pmaddubsw       m6, m4, [r5 + 3 * mmsize]
6897     paddw           m12, m6
6898     pmaddubsw       m4, [r5 + 2 * mmsize]
6899     paddw           m0, m4
6900     lea             r7, [r7 + r1 * 4]
6901     movu            xm6, [r7]                       ; m6 = row 20
6902     punpckhbw       xm7, xm5, xm6
6903     punpcklbw       xm5, xm6
6904     vinserti128     m5, m5, xm7, 1
6905     pmaddubsw       m7, m5, [r5 + 3 * mmsize]
6906     paddw           m13, m7
6907     pmaddubsw       m5, [r5 + 2 * mmsize]
6908     paddw           m1, m5
6909     movu            xm7, [r7 + r1]                  ; m7 = row 21
6910     punpckhbw       xm2, xm6, xm7
6911     punpcklbw       xm6, xm7
6912     vinserti128     m6, m6, xm2, 1
6913     pmaddubsw       m6, [r5 + 3 * mmsize]
6914     paddw           m0, m6
6915     movu            xm2, [r7 + r1 * 2]              ; m2 = row 22
6916     punpckhbw       xm3, xm7, xm2
6917     punpcklbw       xm7, xm2
6918     vinserti128     m7, m7, xm3, 1
6919     pmaddubsw       m7, [r5 + 3 * mmsize]
6920     paddw           m1, m7
6921
6922     pmulhrsw        m8, m14                         ; m8 = word: row 8
6923     pmulhrsw        m9, m14                         ; m9 = word: row 9
6924     pmulhrsw        m10, m14                        ; m10 = word: row 10
6925     pmulhrsw        m11, m14                        ; m11 = word: row 11
6926     pmulhrsw        m12, m14                        ; m12 = word: row 12
6927     pmulhrsw        m13, m14                        ; m13 = word: row 13
6928     pmulhrsw        m0, m14                         ; m0 = word: row 14
6929     pmulhrsw        m1, m14                         ; m1 = word: row 15
6930     packuswb        m8, m9
6931     packuswb        m10, m11
6932     packuswb        m12, m13
6933     packuswb        m0, m1
6934     vpermq          m8, m8, 11011000b
6935     vpermq          m10, m10, 11011000b
6936     vpermq          m12, m12, 11011000b
6937     vpermq          m0, m0, 11011000b
6938     vextracti128    xm9, m8, 1
6939     vextracti128    xm11, m10, 1
6940     vextracti128    xm13, m12, 1
6941     vextracti128    xm1, m0, 1
6942     movu            [r8], xm8
6943     movu            [r8 + r3], xm9
6944     movu            [r8 + r3 * 2], xm10
6945     movu            [r8 + r6], xm11
6946     lea             r8, [r8 + r3 * 4]
6947     movu            [r8], xm12
6948     movu            [r8 + r3], xm13
6949     movu            [r8 + r3 * 2], xm0
6950     movu            [r8 + r6], xm1
6951 %endmacro
6952
6953 %macro PROCESS_LUMA_AVX2_W16_8R 0
6954     movu            xm0, [r0]                       ; m0 = row 0
6955     movu            xm1, [r0 + r1]                  ; m1 = row 1
6956     punpckhbw       xm2, xm0, xm1
6957     punpcklbw       xm0, xm1
6958     vinserti128     m0, m0, xm2, 1
6959     pmaddubsw       m0, [r5]
6960     movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
6961     punpckhbw       xm3, xm1, xm2
6962     punpcklbw       xm1, xm2
6963     vinserti128     m1, m1, xm3, 1
6964     pmaddubsw       m1, [r5]
6965     movu            xm3, [r0 + r4]                  ; m3 = row 3
6966     punpckhbw       xm4, xm2, xm3
6967     punpcklbw       xm2, xm3
6968     vinserti128     m2, m2, xm4, 1
6969     pmaddubsw       m4, m2, [r5 + 1 * mmsize]
6970     paddw           m0, m4
6971     pmaddubsw       m2, [r5]
6972     lea             r7, [r0 + r1 * 4]
6973     movu            xm4, [r7]                       ; m4 = row 4
6974     punpckhbw       xm5, xm3, xm4
6975     punpcklbw       xm3, xm4
6976     vinserti128     m3, m3, xm5, 1
6977     pmaddubsw       m5, m3, [r5 + 1 * mmsize]
6978     paddw           m1, m5
6979     pmaddubsw       m3, [r5]
6980     movu            xm5, [r7 + r1]                  ; m5 = row 5
6981     punpckhbw       xm6, xm4, xm5
6982     punpcklbw       xm4, xm5
6983     vinserti128     m4, m4, xm6, 1
6984     pmaddubsw       m6, m4, [r5 + 2 * mmsize]
6985     paddw           m0, m6
6986     pmaddubsw       m6, m4, [r5 + 1 * mmsize]
6987     paddw           m2, m6
6988     pmaddubsw       m4, [r5]
6989     movu            xm6, [r7 + r1 * 2]              ; m6 = row 6
6990     punpckhbw       xm7, xm5, xm6
6991     punpcklbw       xm5, xm6
6992     vinserti128     m5, m5, xm7, 1
6993     pmaddubsw       m7, m5, [r5 + 2 * mmsize]
6994     paddw           m1, m7
6995     pmaddubsw       m7, m5, [r5 + 1 * mmsize]
6996     paddw           m3, m7
6997     pmaddubsw       m5, [r5]
6998     movu            xm7, [r7 + r4]                  ; m7 = row 7
6999     punpckhbw       xm8, xm6, xm7
7000     punpcklbw       xm6, xm7
7001     vinserti128     m6, m6, xm8, 1
7002     pmaddubsw       m8, m6, [r5 + 3 * mmsize]
7003     paddw           m0, m8
7004     pmaddubsw       m8, m6, [r5 + 2 * mmsize]
7005     paddw           m2, m8
7006     pmaddubsw       m8, m6, [r5 + 1 * mmsize]
7007     paddw           m4, m8
7008     pmaddubsw       m6, [r5]
7009     lea             r7, [r7 + r1 * 4]
7010     movu            xm8, [r7]                       ; m8 = row 8
7011     punpckhbw       xm9, xm7, xm8
7012     punpcklbw       xm7, xm8
7013     vinserti128     m7, m7, xm9, 1
7014     pmaddubsw       m9, m7, [r5 + 3 * mmsize]
7015     paddw           m1, m9
7016     pmaddubsw       m9, m7, [r5 + 2 * mmsize]
7017     paddw           m3, m9
7018     pmaddubsw       m9, m7, [r5 + 1 * mmsize]
7019     paddw           m5, m9
7020     pmaddubsw       m7, [r5]
7021     movu            xm9, [r7 + r1]                  ; m9 = row 9
7022     punpckhbw       xm10, xm8, xm9
7023     punpcklbw       xm8, xm9
7024     vinserti128     m8, m8, xm10, 1
7025     pmaddubsw       m10, m8, [r5 + 3 * mmsize]
7026     paddw           m2, m10
7027     pmaddubsw       m10, m8, [r5 + 2 * mmsize]
7028     paddw           m4, m10
7029     pmaddubsw       m10, m8, [r5 + 1 * mmsize]
7030     paddw           m6, m10
7031     movu            xm10, [r7 + r1 * 2]             ; m10 = row 10
7032     punpckhbw       xm11, xm9, xm10
7033     punpcklbw       xm9, xm10
7034     vinserti128     m9, m9, xm11, 1
7035     pmaddubsw       m11, m9, [r5 + 3 * mmsize]
7036     paddw           m3, m11
7037     pmaddubsw       m11, m9, [r5 + 2 * mmsize]
7038     paddw           m5, m11
7039     pmaddubsw       m11, m9, [r5 + 1 * mmsize]
7040     paddw           m7, m11
7041     movu            xm11, [r7 + r4]                 ; m11 = row 11
7042     punpckhbw       xm12, xm10, xm11
7043     punpcklbw       xm10, xm11
7044     vinserti128     m10, m10, xm12, 1
7045     pmaddubsw       m12, m10, [r5 + 3 * mmsize]
7046     paddw           m4, m12
7047     pmaddubsw       m12, m10, [r5 + 2 * mmsize]
7048     paddw           m6, m12
7049     lea             r7, [r7 + r1 * 4]
7050     movu            xm12, [r7]                      ; m12 = row 12
7051     punpckhbw       xm13, xm11, xm12
7052     punpcklbw       xm11, xm12
7053     vinserti128     m11, m11, xm13, 1
7054     pmaddubsw       m13, m11, [r5 + 3 * mmsize]
7055     paddw           m5, m13
7056     pmaddubsw       m13, m11, [r5 + 2 * mmsize]
7057     paddw           m7, m13
7058
7059     pmulhrsw        m0, m14                         ; m0 = word: row 0
7060     pmulhrsw        m1, m14                         ; m1 = word: row 1
7061     pmulhrsw        m2, m14                         ; m2 = word: row 2
7062     pmulhrsw        m3, m14                         ; m3 = word: row 3
7063     pmulhrsw        m4, m14                         ; m4 = word: row 4
7064     pmulhrsw        m5, m14                         ; m5 = word: row 5
7065     packuswb        m0, m1
7066     packuswb        m2, m3
7067     packuswb        m4, m5
7068     vpermq          m0, m0, 11011000b
7069     vpermq          m2, m2, 11011000b
7070     vpermq          m4, m4, 11011000b
7071     vextracti128    xm1, m0, 1
7072     vextracti128    xm3, m2, 1
7073     vextracti128    xm5, m4, 1
7074     movu            [r2], xm0
7075     movu            [r2 + r3], xm1
7076     movu            [r2 + r3 * 2], xm2
7077     movu            [r2 + r6], xm3
7078     lea             r8, [r2 + r3 * 4]
7079     movu            [r8], xm4
7080     movu            [r8 + r3], xm5
7081
7082     movu            xm13, [r7 + r1]                 ; m13 = row 13
7083     punpckhbw       xm0, xm12, xm13
7084     punpcklbw       xm12, xm13
7085     vinserti128     m12, m12, xm0, 1
7086     pmaddubsw       m0, m12, [r5 + 3 * mmsize]
7087     paddw           m6, m0
7088     movu            xm0, [r7 + r1 * 2]              ; m0 = row 14
7089     punpckhbw       xm1, xm13, xm0
7090     punpcklbw       xm13, xm0
7091     vinserti128     m13, m13, xm1, 1
7092     pmaddubsw       m1, m13, [r5 + 3 * mmsize]
7093     paddw           m7, m1
7094
7095     pmulhrsw        m6, m14                         ; m6 = word: row 6
7096     pmulhrsw        m7, m14                         ; m7 = word: row 7
7097     packuswb        m6, m7
7098     vpermq          m6, m6, 11011000b
7099     vextracti128    xm7, m6, 1
7100     movu            [r8 + r3 * 2], xm6
7101     movu            [r8 + r6], xm7
7102 %endmacro
7103
7104 INIT_YMM avx2
7105 %if ARCH_X86_64 == 1
7106 cglobal interp_8tap_vert_pp_24x32, 4, 11, 15
7107     mov             r4d, r4m
7108     shl             r4d, 7
7109
7110 %ifdef PIC
7111     lea             r5, [tab_LumaCoeffVer_32]
7112     add             r5, r4
7113 %else
7114     lea             r5, [tab_LumaCoeffVer_32 + r4]
7115 %endif
7116
7117     lea             r4, [r1 * 3]
7118     sub             r0, r4
7119     lea             r6, [r3 * 3]
7120     lea             r10, [r1 * 4]
7121     mova            m14, [pw_512]
7122     mov             r9d, 2
7123 .loopH:
7124     PROCESS_LUMA_AVX2_W16_16R
7125     add             r2, 16
7126     add             r0, 16
7127
7128     movq            xm1, [r0]                       ; m1 = row 0
7129     movq            xm2, [r0 + r1]                  ; m2 = row 1
7130     punpcklbw       xm1, xm2
7131     movq            xm3, [r0 + r1 * 2]              ; m3 = row 2
7132     punpcklbw       xm2, xm3
7133     vinserti128     m5, m1, xm2, 1
7134     pmaddubsw       m5, [r5]
7135     movq            xm4, [r0 + r4]                  ; m4 = row 3
7136     punpcklbw       xm3, xm4
7137     lea             r7, [r0 + r1 * 4]
7138     movq            xm1, [r7]                       ; m1 = row 4
7139     punpcklbw       xm4, xm1
7140     vinserti128     m2, m3, xm4, 1
7141     pmaddubsw       m0, m2, [r5 + 1 * mmsize]
7142     paddw           m5, m0
7143     pmaddubsw       m2, [r5]
7144     movq            xm3, [r7 + r1]                  ; m3 = row 5
7145     punpcklbw       xm1, xm3
7146     movq            xm4, [r7 + r1 * 2]              ; m4 = row 6
7147     punpcklbw       xm3, xm4
7148     vinserti128     m1, m1, xm3, 1
7149     pmaddubsw       m3, m1, [r5 + 2 * mmsize]
7150     paddw           m5, m3
7151     pmaddubsw       m0, m1, [r5 + 1 * mmsize]
7152     paddw           m2, m0
7153     pmaddubsw       m1, [r5]
7154     movq            xm3, [r7 + r4]                  ; m3 = row 7
7155     punpcklbw       xm4, xm3
7156     lea             r7, [r7 + r1 * 4]
7157     movq            xm0, [r7]                       ; m0 = row 8
7158     punpcklbw       xm3, xm0
7159     vinserti128     m4, m4, xm3, 1
7160     pmaddubsw       m3, m4, [r5 + 3 * mmsize]
7161     paddw           m5, m3
7162     pmaddubsw       m3, m4, [r5 + 2 * mmsize]
7163     paddw           m2, m3
7164     pmaddubsw       m3, m4, [r5 + 1 * mmsize]
7165     paddw           m1, m3
7166     pmaddubsw       m4, [r5]
7167     movq            xm3, [r7 + r1]                  ; m3 = row 9
7168     punpcklbw       xm0, xm3
7169     movq            xm6, [r7 + r1 * 2]              ; m6 = row 10
7170     punpcklbw       xm3, xm6
7171     vinserti128     m0, m0, xm3, 1
7172     pmaddubsw       m3, m0, [r5 + 3 * mmsize]
7173     paddw           m2, m3
7174     pmaddubsw       m3, m0, [r5 + 2 * mmsize]
7175     paddw           m1, m3
7176     pmaddubsw       m3, m0, [r5 + 1 * mmsize]
7177     paddw           m4, m3
7178     pmaddubsw       m0, [r5]
7179
7180     movq            xm3, [r7 + r4]                  ; m3 = row 11
7181     punpcklbw       xm6, xm3
7182     lea             r7, [r7 + r1 * 4]
7183     movq            xm7, [r7]                       ; m7 = row 12
7184     punpcklbw       xm3, xm7
7185     vinserti128     m6, m6, xm3, 1
7186     pmaddubsw       m3, m6, [r5 + 3 * mmsize]
7187     paddw           m1, m3
7188     pmaddubsw       m3, m6, [r5 + 2 * mmsize]
7189     paddw           m4, m3
7190     pmaddubsw       m3, m6, [r5 + 1 * mmsize]
7191     paddw           m0, m3
7192     pmaddubsw       m6, [r5]
7193     movq            xm3, [r7 + r1]                  ; m3 = row 13
7194     punpcklbw       xm7, xm3
7195     movq            xm8, [r7 + r1 * 2]              ; m8 = row 14
7196     punpcklbw       xm3, xm8
7197     vinserti128     m7, m7, xm3, 1
7198     pmaddubsw       m3, m7, [r5 + 3 * mmsize]
7199     paddw           m4, m3
7200     pmaddubsw       m3, m7, [r5 + 2 * mmsize]
7201     paddw           m0, m3
7202     pmaddubsw       m3, m7, [r5 + 1 * mmsize]
7203     paddw           m6, m3
7204     pmaddubsw       m7, [r5]
7205     movq            xm3, [r7 + r4]                  ; m3 = row 15
7206     punpcklbw       xm8, xm3
7207     lea             r7, [r7 + r1 * 4]
7208     movq            xm9, [r7]                       ; m9 = row 16
7209     punpcklbw       xm3, xm9
7210     vinserti128     m8, m8, xm3, 1
7211     pmaddubsw       m3, m8, [r5 + 3 * mmsize]
7212     paddw           m0, m3
7213     pmaddubsw       m3, m8, [r5 + 2 * mmsize]
7214     paddw           m6, m3
7215     pmaddubsw       m3, m8, [r5 + 1 * mmsize]
7216     paddw           m7, m3
7217     pmaddubsw       m8, [r5]
7218     movq            xm3, [r7 + r1]                  ; m3 = row 17
7219     punpcklbw       xm9, xm3
7220     movq            xm10, [r7 + r1 * 2]             ; m10 = row 18
7221     punpcklbw       xm3, xm10
7222     vinserti128     m9, m9, xm3, 1
7223     pmaddubsw       m3, m9, [r5 + 3 * mmsize]
7224     paddw           m6, m3
7225     pmaddubsw       m3, m9, [r5 + 2 * mmsize]
7226     paddw           m7, m3
7227     pmaddubsw       m3, m9, [r5 + 1 * mmsize]
7228     paddw           m8, m3
7229     movq            xm3, [r7 + r4]                  ; m3 = row 19
7230     punpcklbw       xm10, xm3
7231     lea             r7, [r7 + r1 * 4]
7232     movq            xm9, [r7]                       ; m9 = row 20
7233     punpcklbw       xm3, xm9
7234     vinserti128     m10, m10, xm3, 1
7235     pmaddubsw       m3, m10, [r5 + 3 * mmsize]
7236     paddw           m7, m3
7237     pmaddubsw       m3, m10, [r5 + 2 * mmsize]
7238     paddw           m8, m3
7239     movq            xm3, [r7 + r1]                  ; m3 = row 21
7240     punpcklbw       xm9, xm3
7241     movq            xm10, [r7 + r1 * 2]             ; m10 = row 22
7242     punpcklbw       xm3, xm10
7243     vinserti128     m9, m9, xm3, 1
7244     pmaddubsw       m3, m9, [r5 + 3 * mmsize]
7245     paddw           m8, m3
7246
7247     pmulhrsw        m5, m14                         ; m5 = word: row 0, row 1
7248     pmulhrsw        m2, m14                         ; m2 = word: row 2, row 3
7249     pmulhrsw        m1, m14                         ; m1 = word: row 4, row 5
7250     pmulhrsw        m4, m14                         ; m4 = word: row 6, row 7
7251     pmulhrsw        m0, m14                         ; m0 = word: row 8, row 9
7252     pmulhrsw        m6, m14                         ; m6 = word: row 10, row 11
7253     pmulhrsw        m7, m14                         ; m7 = word: row 12, row 13
7254     pmulhrsw        m8, m14                         ; m8 = word: row 14, row 15
7255     packuswb        m5, m2
7256     packuswb        m1, m4
7257     packuswb        m0, m6
7258     packuswb        m7, m8
7259     vextracti128    xm2, m5, 1
7260     vextracti128    xm4, m1, 1
7261     vextracti128    xm6, m0, 1
7262     vextracti128    xm8, m7, 1
7263     movq            [r2], xm5
7264     movq            [r2 + r3], xm2
7265     movhps          [r2 + r3 * 2], xm5
7266     movhps          [r2 + r6], xm2
7267     lea             r8, [r2 + r3 * 4]
7268     movq            [r8], xm1
7269     movq            [r8 + r3], xm4
7270     movhps          [r8 + r3 * 2], xm1
7271     movhps          [r8 + r6], xm4
7272     lea             r8, [r8 + r3 * 4]
7273     movq            [r8], xm0
7274     movq            [r8 + r3], xm6
7275     movhps          [r8 + r3 * 2], xm0
7276     movhps          [r8 + r6], xm6
7277     lea             r8, [r8 + r3 * 4]
7278     movq            [r8], xm7
7279     movq            [r8 + r3], xm8
7280     movhps          [r8 + r3 * 2], xm7
7281     movhps          [r8 + r6], xm8
7282
7283     sub             r7, r10
7284     lea             r0, [r7 - 16]
7285     lea             r2, [r8 + r3 * 4 - 16]
7286     dec             r9d
7287     jnz             .loopH
7288     RET
7289 %endif
7290
7291 %macro FILTER_VER_LUMA_AVX2_32xN 2
7292 INIT_YMM avx2
7293 %if ARCH_X86_64 == 1
7294 cglobal interp_8tap_vert_pp_%1x%2, 4, 12, 15
7295     mov             r4d, r4m
7296     shl             r4d, 7
7297
7298 %ifdef PIC
7299     lea             r5, [tab_LumaCoeffVer_32]
7300     add             r5, r4
7301 %else
7302     lea             r5, [tab_LumaCoeffVer_32 + r4]
7303 %endif
7304
7305     lea             r4, [r1 * 3]
7306     sub             r0, r4
7307     lea             r6, [r3 * 3]
7308     lea             r11, [r1 * 4]
7309     mova            m14, [pw_512]
7310     mov             r9d, %2 / 16
7311 .loopH:
7312     mov             r10d, %1 / 16
7313 .loopW:
7314     PROCESS_LUMA_AVX2_W16_16R
7315     add             r2, 16
7316     add             r0, 16
7317     dec             r10d
7318     jnz             .loopW
7319     sub             r7, r11
7320     lea             r0, [r7 - 16]
7321     lea             r2, [r8 + r3 * 4 - 16]
7322     dec             r9d
7323     jnz             .loopH
7324     RET
7325 %endif
7326 %endmacro
7327
7328 FILTER_VER_LUMA_AVX2_32xN 32, 32
7329 FILTER_VER_LUMA_AVX2_32xN 32, 64
7330
7331 INIT_YMM avx2
7332 %if ARCH_X86_64 == 1
7333 cglobal interp_8tap_vert_pp_32x16, 4, 10, 15
7334     mov             r4d, r4m
7335     shl             r4d, 7
7336
7337 %ifdef PIC
7338     lea             r5, [tab_LumaCoeffVer_32]
7339     add             r5, r4
7340 %else
7341     lea             r5, [tab_LumaCoeffVer_32 + r4]
7342 %endif
7343
7344     lea             r4, [r1 * 3]
7345     sub             r0, r4
7346     lea             r6, [r3 * 3]
7347     mova            m14, [pw_512]
7348     mov             r9d, 2
7349 .loopW:
7350     PROCESS_LUMA_AVX2_W16_16R
7351     add             r2, 16
7352     add             r0, 16
7353     dec             r9d
7354     jnz             .loopW
7355     RET
7356 %endif
7357
7358 INIT_YMM avx2
7359 %if ARCH_X86_64 == 1
7360 cglobal interp_8tap_vert_pp_32x24, 4, 10, 15
7361     mov             r4d, r4m
7362     shl             r4d, 7
7363
7364 %ifdef PIC
7365     lea             r5, [tab_LumaCoeffVer_32]
7366     add             r5, r4
7367 %else
7368     lea             r5, [tab_LumaCoeffVer_32 + r4]
7369 %endif
7370
7371     lea             r4, [r1 * 3]
7372     sub             r0, r4
7373     lea             r6, [r3 * 3]
7374     mova            m14, [pw_512]
7375     mov             r9d, 2
7376 .loopW:
7377     PROCESS_LUMA_AVX2_W16_16R
7378     add             r2, 16
7379     add             r0, 16
7380     dec             r9d
7381     jnz             .loopW
7382     lea             r9, [r1 * 4]
7383     sub             r7, r9
7384     lea             r0, [r7 - 16]
7385     lea             r2, [r8 + r3 * 4 - 16]
7386     mov             r9d, 2
7387 .loop:
7388     PROCESS_LUMA_AVX2_W16_8R
7389     add             r2, 16
7390     add             r0, 16
7391     dec             r9d
7392     jnz             .loop
7393     RET
7394 %endif
7395
7396 INIT_YMM avx2
7397 %if ARCH_X86_64 == 1
7398 cglobal interp_8tap_vert_pp_32x8, 4, 10, 15
7399     mov             r4d, r4m
7400     shl             r4d, 7
7401
7402 %ifdef PIC
7403     lea             r5, [tab_LumaCoeffVer_32]
7404     add             r5, r4
7405 %else
7406     lea             r5, [tab_LumaCoeffVer_32 + r4]
7407 %endif
7408
7409     lea             r4, [r1 * 3]
7410     sub             r0, r4
7411     lea             r6, [r3 * 3]
7412     mova            m14, [pw_512]
7413     mov             r9d, 2
7414 .loopW:
7415     PROCESS_LUMA_AVX2_W16_8R
7416     add             r2, 16
7417     add             r0, 16
7418     dec             r9d
7419     jnz             .loopW
7420     RET
7421 %endif
7422
7423 INIT_YMM avx2
7424 %if ARCH_X86_64 == 1
7425 cglobal interp_8tap_vert_pp_48x64, 4, 12, 15
7426     mov             r4d, r4m
7427     shl             r4d, 7
7428
7429 %ifdef PIC
7430     lea             r5, [tab_LumaCoeffVer_32]
7431     add             r5, r4
7432 %else
7433     lea             r5, [tab_LumaCoeffVer_32 + r4]
7434 %endif
7435
7436     lea             r4, [r1 * 3]
7437     sub             r0, r4
7438     lea             r6, [r3 * 3]
7439     lea             r11, [r1 * 4]
7440     mova            m14, [pw_512]
7441     mov             r9d, 4
7442 .loopH:
7443     mov             r10d, 3
7444 .loopW:
7445     PROCESS_LUMA_AVX2_W16_16R
7446     add             r2, 16
7447     add             r0, 16
7448     dec             r10d
7449     jnz             .loopW
7450     sub             r7, r11
7451     lea             r0, [r7 - 32]
7452     lea             r2, [r8 + r3 * 4 - 32]
7453     dec             r9d
7454     jnz             .loopH
7455     RET
7456 %endif
7457
7458 %macro FILTER_VER_LUMA_AVX2_64xN 2
7459 INIT_YMM avx2
7460 %if ARCH_X86_64 == 1
7461 cglobal interp_8tap_vert_pp_%1x%2, 4, 12, 15
7462     mov             r4d, r4m
7463     shl             r4d, 7
7464
7465 %ifdef PIC
7466     lea             r5, [tab_LumaCoeffVer_32]
7467     add             r5, r4
7468 %else
7469     lea             r5, [tab_LumaCoeffVer_32 + r4]
7470 %endif
7471
7472     lea             r4, [r1 * 3]
7473     sub             r0, r4
7474     lea             r6, [r3 * 3]
7475     lea             r11, [r1 * 4]
7476     mova            m14, [pw_512]
7477     mov             r9d, %2 / 16
7478 .loopH:
7479     mov             r10d, %1 / 16
7480 .loopW:
7481     PROCESS_LUMA_AVX2_W16_16R
7482     add             r2, 16
7483     add             r0, 16
7484     dec             r10d
7485     jnz             .loopW
7486     sub             r7, r11
7487     lea             r0, [r7 - 48]
7488     lea             r2, [r8 + r3 * 4 - 48]
7489     dec             r9d
7490     jnz             .loopH
7491     RET
7492 %endif
7493 %endmacro
7494
7495 FILTER_VER_LUMA_AVX2_64xN 64, 32
7496 FILTER_VER_LUMA_AVX2_64xN 64, 48
7497 FILTER_VER_LUMA_AVX2_64xN 64, 64
7498
7499 INIT_YMM avx2
7500 %if ARCH_X86_64 == 1
7501 cglobal interp_8tap_vert_pp_64x16, 4, 10, 15
7502     mov             r4d, r4m
7503     shl             r4d, 7
7504
7505 %ifdef PIC
7506     lea             r5, [tab_LumaCoeffVer_32]
7507     add             r5, r4
7508 %else
7509     lea             r5, [tab_LumaCoeffVer_32 + r4]
7510 %endif
7511
7512     lea             r4, [r1 * 3]
7513     sub             r0, r4
7514     lea             r6, [r3 * 3]
7515     mova            m14, [pw_512]
7516     mov             r9d, 4
7517 .loopW:
7518     PROCESS_LUMA_AVX2_W16_16R
7519     add             r2, 16
7520     add             r0, 16
7521     dec             r9d
7522     jnz             .loopW
7523     RET
7524 %endif
7525
7526 ;-------------------------------------------------------------------------------------------------------------
7527 ; void interp_8tap_vert_%3_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
7528 ;-------------------------------------------------------------------------------------------------------------
7529 %macro FILTER_VER_LUMA 3
7530 INIT_XMM sse4
7531 cglobal interp_8tap_vert_%3_%1x%2, 5, 7, 8 ,0-gprsize
7532     lea       r5, [3 * r1]
7533     sub       r0, r5
7534     shl       r4d, 6
7535 %ifidn %3,ps
7536     add       r3d, r3d
7537 %endif
7538
7539 %ifdef PIC
7540     lea       r5, [tab_LumaCoeffVer]
7541     lea       r6, [r5 + r4]
7542 %else
7543     lea       r6, [tab_LumaCoeffVer + r4]
7544 %endif
7545
7546 %ifidn %3,pp
7547     mova      m3, [pw_512]
7548 %else
7549     mova      m3, [pw_2000]
7550 %endif
7551     mov       dword [rsp], %2/4
7552
7553 .loopH:
7554     mov       r4d, (%1/8)
7555 .loopW:
7556     PROCESS_LUMA_W8_4R
7557 %ifidn %3,pp
7558     pmulhrsw  m7, m3
7559     pmulhrsw  m6, m3
7560     pmulhrsw  m5, m3
7561     pmulhrsw  m4, m3
7562
7563     packuswb  m7, m6
7564     packuswb  m5, m4
7565
7566     movlps    [r2], m7
7567     movhps    [r2 + r3], m7
7568     lea       r5, [r2 + 2 * r3]
7569     movlps    [r5], m5
7570     movhps    [r5 + r3], m5
7571 %else
7572     psubw     m7, m3
7573     psubw     m6, m3
7574     psubw     m5, m3
7575     psubw     m4, m3
7576
7577     movu      [r2], m7
7578     movu      [r2 + r3], m6
7579     lea       r5, [r2 + 2 * r3]
7580     movu      [r5], m5
7581     movu      [r5 + r3], m4
7582 %endif
7583
7584     lea       r5, [8 * r1 - 8]
7585     sub       r0, r5
7586 %ifidn %3,pp
7587     add       r2, 8
7588 %else
7589     add       r2, 16
7590 %endif
7591     dec       r4d
7592     jnz       .loopW
7593
7594     lea       r0, [r0 + 4 * r1 - %1]
7595 %ifidn %3,pp
7596     lea       r2, [r2 + 4 * r3 - %1]
7597 %else
7598     lea       r2, [r2 + 4 * r3 - 2 * %1]
7599 %endif
7600
7601     dec       dword [rsp]
7602     jnz       .loopH
7603
7604     RET
7605 %endmacro
7606
7607 FILTER_VER_LUMA 16, 4, pp
7608 FILTER_VER_LUMA 16, 8, pp
7609 FILTER_VER_LUMA 16, 12, pp
7610 FILTER_VER_LUMA 16, 16, pp
7611 FILTER_VER_LUMA 16, 32, pp
7612 FILTER_VER_LUMA 16, 64, pp
7613 FILTER_VER_LUMA 24, 32, pp
7614 FILTER_VER_LUMA 32, 8, pp
7615 FILTER_VER_LUMA 32, 16, pp
7616 FILTER_VER_LUMA 32, 24, pp
7617 FILTER_VER_LUMA 32, 32, pp
7618 FILTER_VER_LUMA 32, 64, pp
7619 FILTER_VER_LUMA 48, 64, pp
7620 FILTER_VER_LUMA 64, 16, pp
7621 FILTER_VER_LUMA 64, 32, pp
7622 FILTER_VER_LUMA 64, 48, pp
7623 FILTER_VER_LUMA 64, 64, pp
7624
7625 FILTER_VER_LUMA 16, 4, ps
7626 FILTER_VER_LUMA 16, 8, ps
7627 FILTER_VER_LUMA 16, 12, ps
7628 FILTER_VER_LUMA 16, 16, ps
7629 FILTER_VER_LUMA 16, 32, ps
7630 FILTER_VER_LUMA 16, 64, ps
7631 FILTER_VER_LUMA 24, 32, ps
7632 FILTER_VER_LUMA 32, 8, ps
7633 FILTER_VER_LUMA 32, 16, ps
7634 FILTER_VER_LUMA 32, 24, ps
7635 FILTER_VER_LUMA 32, 32, ps
7636 FILTER_VER_LUMA 32, 64, ps
7637 FILTER_VER_LUMA 48, 64, ps
7638 FILTER_VER_LUMA 64, 16, ps
7639 FILTER_VER_LUMA 64, 32, ps
7640 FILTER_VER_LUMA 64, 48, ps
7641 FILTER_VER_LUMA 64, 64, ps
7642
7643 %macro PROCESS_LUMA_SP_W4_4R 0
7644     movq       m0, [r0]
7645     movq       m1, [r0 + r1]
7646     punpcklwd  m0, m1                          ;m0=[0 1]
7647     pmaddwd    m0, [r6 + 0 *16]                ;m0=[0+1]  Row1
7648
7649     lea        r0, [r0 + 2 * r1]
7650     movq       m4, [r0]
7651     punpcklwd  m1, m4                          ;m1=[1 2]
7652     pmaddwd    m1, [r6 + 0 *16]                ;m1=[1+2]  Row2
7653
7654     movq       m5, [r0 + r1]
7655     punpcklwd  m4, m5                          ;m4=[2 3]
7656     pmaddwd    m2, m4, [r6 + 0 *16]            ;m2=[2+3]  Row3
7657     pmaddwd    m4, [r6 + 1 * 16]
7658     paddd      m0, m4                          ;m0=[0+1+2+3]  Row1
7659
7660     lea        r0, [r0 + 2 * r1]
7661     movq       m4, [r0]
7662     punpcklwd  m5, m4                          ;m5=[3 4]
7663     pmaddwd    m3, m5, [r6 + 0 *16]            ;m3=[3+4]  Row4
7664     pmaddwd    m5, [r6 + 1 * 16]
7665     paddd      m1, m5                          ;m1 = [1+2+3+4]  Row2
7666
7667     movq       m5, [r0 + r1]
7668     punpcklwd  m4, m5                          ;m4=[4 5]
7669     pmaddwd    m6, m4, [r6 + 1 * 16]
7670     paddd      m2, m6                          ;m2=[2+3+4+5]  Row3
7671     pmaddwd    m4, [r6 + 2 * 16]
7672     paddd      m0, m4                          ;m0=[0+1+2+3+4+5]  Row1
7673
7674     lea        r0, [r0 + 2 * r1]
7675     movq       m4, [r0]
7676     punpcklwd  m5, m4                          ;m5=[5 6]
7677     pmaddwd    m6, m5, [r6 + 1 * 16]
7678     paddd      m3, m6                          ;m3=[3+4+5+6]  Row4
7679     pmaddwd    m5, [r6 + 2 * 16]
7680     paddd      m1, m5                          ;m1=[1+2+3+4+5+6]  Row2
7681
7682     movq       m5, [r0 + r1]
7683     punpcklwd  m4, m5                          ;m4=[6 7]
7684     pmaddwd    m6, m4, [r6 + 2 * 16]
7685     paddd      m2, m6                          ;m2=[2+3+4+5+6+7]  Row3
7686     pmaddwd    m4, [r6 + 3 * 16]
7687     paddd      m0, m4                          ;m0=[0+1+2+3+4+5+6+7]  Row1 end
7688
7689     lea        r0, [r0 + 2 * r1]
7690     movq       m4, [r0]
7691     punpcklwd  m5, m4                          ;m5=[7 8]
7692     pmaddwd    m6, m5, [r6 + 2 * 16]
7693     paddd      m3, m6                          ;m3=[3+4+5+6+7+8]  Row4
7694     pmaddwd    m5, [r6 + 3 * 16]
7695     paddd      m1, m5                          ;m1=[1+2+3+4+5+6+7+8]  Row2 end
7696
7697     movq       m5, [r0 + r1]
7698     punpcklwd  m4, m5                          ;m4=[8 9]
7699     pmaddwd    m4, [r6 + 3 * 16]
7700     paddd      m2, m4                          ;m2=[2+3+4+5+6+7+8+9]  Row3 end
7701
7702     movq       m4, [r0 + 2 * r1]
7703     punpcklwd  m5, m4                          ;m5=[9 10]
7704     pmaddwd    m5, [r6 + 3 * 16]
7705     paddd      m3, m5                          ;m3=[3+4+5+6+7+8+9+10]  Row4 end
7706 %endmacro
7707
7708 ;--------------------------------------------------------------------------------------------------------------
7709 ; void interp_8tap_vert_sp_%1x%2(int16_t *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
7710 ;--------------------------------------------------------------------------------------------------------------
7711 %macro FILTER_VER_LUMA_SP 2
7712 INIT_XMM sse4
7713 cglobal interp_8tap_vert_sp_%1x%2, 5, 7, 8 ,0-gprsize
7714
7715     add       r1d, r1d
7716     lea       r5, [r1 + 2 * r1]
7717     sub       r0, r5
7718     shl       r4d, 6
7719
7720 %ifdef PIC
7721     lea       r5, [tab_LumaCoeffV]
7722     lea       r6, [r5 + r4]
7723 %else
7724     lea       r6, [tab_LumaCoeffV + r4]
7725 %endif
7726
7727     mova      m7, [tab_c_526336]
7728
7729     mov       dword [rsp], %2/4
7730 .loopH:
7731     mov       r4d, (%1/4)
7732 .loopW:
7733     PROCESS_LUMA_SP_W4_4R
7734
7735     paddd     m0, m7
7736     paddd     m1, m7
7737     paddd     m2, m7
7738     paddd     m3, m7
7739
7740     psrad     m0, 12
7741     psrad     m1, 12
7742     psrad     m2, 12
7743     psrad     m3, 12
7744
7745     packssdw  m0, m1
7746     packssdw  m2, m3
7747
7748     packuswb  m0, m2
7749
7750     movd      [r2], m0
7751     pextrd    [r2 + r3], m0, 1
7752     lea       r5, [r2 + 2 * r3]
7753     pextrd    [r5], m0, 2
7754     pextrd    [r5 + r3], m0, 3
7755
7756     lea       r5, [8 * r1 - 2 * 4]
7757     sub       r0, r5
7758     add       r2, 4
7759
7760     dec       r4d
7761     jnz       .loopW
7762
7763     lea       r0, [r0 + 4 * r1 - 2 * %1]
7764     lea       r2, [r2 + 4 * r3 - %1]
7765
7766     dec       dword [rsp]
7767     jnz       .loopH
7768
7769     RET
7770 %endmacro
7771
7772 ;--------------------------------------------------------------------------------------------------------------
7773 ; void interp_8tap_vert_sp_%1x%2(int16_t *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
7774 ;--------------------------------------------------------------------------------------------------------------
7775     FILTER_VER_LUMA_SP 4, 4
7776     FILTER_VER_LUMA_SP 8, 8
7777     FILTER_VER_LUMA_SP 8, 4
7778     FILTER_VER_LUMA_SP 4, 8
7779     FILTER_VER_LUMA_SP 16, 16
7780     FILTER_VER_LUMA_SP 16, 8
7781     FILTER_VER_LUMA_SP 8, 16
7782     FILTER_VER_LUMA_SP 16, 12
7783     FILTER_VER_LUMA_SP 12, 16
7784     FILTER_VER_LUMA_SP 16, 4
7785     FILTER_VER_LUMA_SP 4, 16
7786     FILTER_VER_LUMA_SP 32, 32
7787     FILTER_VER_LUMA_SP 32, 16
7788     FILTER_VER_LUMA_SP 16, 32
7789     FILTER_VER_LUMA_SP 32, 24
7790     FILTER_VER_LUMA_SP 24, 32
7791     FILTER_VER_LUMA_SP 32, 8
7792     FILTER_VER_LUMA_SP 8, 32
7793     FILTER_VER_LUMA_SP 64, 64
7794     FILTER_VER_LUMA_SP 64, 32
7795     FILTER_VER_LUMA_SP 32, 64
7796     FILTER_VER_LUMA_SP 64, 48
7797     FILTER_VER_LUMA_SP 48, 64
7798     FILTER_VER_LUMA_SP 64, 16
7799     FILTER_VER_LUMA_SP 16, 64
7800
7801 ; TODO: combin of U and V is more performance, but need more register
7802 ; TODO: use two path for height alignment to 4 and otherwise may improvement 10% performance, but code is more complex, so I disable it
7803 INIT_XMM ssse3
7804 cglobal chroma_p2s, 3, 7, 4
7805
7806     ; load width and height
7807     mov         r3d, r3m
7808     mov         r4d, r4m
7809
7810     ; load constant
7811     mova        m2, [pb_128]
7812     mova        m3, [tab_c_64_n64]
7813
7814 .loopH:
7815
7816     xor         r5d, r5d
7817 .loopW:
7818     lea         r6, [r0 + r5]
7819
7820     movh        m0, [r6]
7821     punpcklbw   m0, m2
7822     pmaddubsw   m0, m3
7823
7824     movh        m1, [r6 + r1]
7825     punpcklbw   m1, m2
7826     pmaddubsw   m1, m3
7827
7828     add         r5d, 8
7829     cmp         r5d, r3d
7830     lea         r6, [r2 + r5 * 2]
7831     jg          .width4
7832     movu        [r6 + FENC_STRIDE / 2 * 0 - 16], m0
7833     movu        [r6 + FENC_STRIDE / 2 * 2 - 16], m1
7834     je          .nextH
7835     jmp         .loopW
7836
7837 .width4:
7838     test        r3d, 4
7839     jz          .width2
7840     test        r3d, 2
7841     movh        [r6 + FENC_STRIDE / 2 * 0 - 16], m0
7842     movh        [r6 + FENC_STRIDE / 2 * 2 - 16], m1
7843     lea         r6, [r6 + 8]
7844     pshufd      m0, m0, 2
7845     pshufd      m1, m1, 2
7846     jz          .nextH
7847
7848 .width2:
7849     movd        [r6 + FENC_STRIDE / 2 * 0 - 16], m0
7850     movd        [r6 + FENC_STRIDE / 2 * 2 - 16], m1
7851
7852 .nextH:
7853     lea         r0, [r0 + r1 * 2]
7854     add         r2, FENC_STRIDE / 2 * 4
7855
7856     sub         r4d, 2
7857     jnz         .loopH
7858
7859     RET
7860
7861 %macro PROCESS_CHROMA_SP_W4_4R 0
7862     movq       m0, [r0]
7863     movq       m1, [r0 + r1]
7864     punpcklwd  m0, m1                          ;m0=[0 1]
7865     pmaddwd    m0, [r6 + 0 *16]                ;m0=[0+1]         Row1
7866
7867     lea        r0, [r0 + 2 * r1]
7868     movq       m4, [r0]
7869     punpcklwd  m1, m4                          ;m1=[1 2]
7870     pmaddwd    m1, [r6 + 0 *16]                ;m1=[1+2]         Row2
7871
7872     movq       m5, [r0 + r1]
7873     punpcklwd  m4, m5                          ;m4=[2 3]
7874     pmaddwd    m2, m4, [r6 + 0 *16]            ;m2=[2+3]         Row3
7875     pmaddwd    m4, [r6 + 1 * 16]
7876     paddd      m0, m4                          ;m0=[0+1+2+3]     Row1 done
7877
7878     lea        r0, [r0 + 2 * r1]
7879     movq       m4, [r0]
7880     punpcklwd  m5, m4                          ;m5=[3 4]
7881     pmaddwd    m3, m5, [r6 + 0 *16]            ;m3=[3+4]         Row4
7882     pmaddwd    m5, [r6 + 1 * 16]
7883     paddd      m1, m5                          ;m1 = [1+2+3+4]   Row2
7884
7885     movq       m5, [r0 + r1]
7886     punpcklwd  m4, m5                          ;m4=[4 5]
7887     pmaddwd    m4, [r6 + 1 * 16]
7888     paddd      m2, m4                          ;m2=[2+3+4+5]     Row3
7889
7890     movq       m4, [r0 + 2 * r1]
7891     punpcklwd  m5, m4                          ;m5=[5 6]
7892     pmaddwd    m5, [r6 + 1 * 16]
7893     paddd      m3, m5                          ;m3=[3+4+5+6]     Row4
7894 %endmacro
7895
7896 ;--------------------------------------------------------------------------------------------------------------
7897 ; void interp_4tap_vert_sp_%1x%2(int16_t *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
7898 ;--------------------------------------------------------------------------------------------------------------
7899 %macro FILTER_VER_CHROMA_SP 2
7900 INIT_XMM sse4
7901 cglobal interp_4tap_vert_sp_%1x%2, 5, 7, 7 ,0-gprsize
7902
7903     add       r1d, r1d
7904     sub       r0, r1
7905     shl       r4d, 5
7906
7907 %ifdef PIC
7908     lea       r5, [tab_ChromaCoeffV]
7909     lea       r6, [r5 + r4]
7910 %else
7911     lea       r6, [tab_ChromaCoeffV + r4]
7912 %endif
7913
7914     mova      m6, [tab_c_526336]
7915
7916     mov       dword [rsp], %2/4
7917
7918 .loopH:
7919     mov       r4d, (%1/4)
7920 .loopW:
7921     PROCESS_CHROMA_SP_W4_4R
7922
7923     paddd     m0, m6
7924     paddd     m1, m6
7925     paddd     m2, m6
7926     paddd     m3, m6
7927
7928     psrad     m0, 12
7929     psrad     m1, 12
7930     psrad     m2, 12
7931     psrad     m3, 12
7932
7933     packssdw  m0, m1
7934     packssdw  m2, m3
7935
7936     packuswb  m0, m2
7937
7938     movd      [r2], m0
7939     pextrd    [r2 + r3], m0, 1
7940     lea       r5, [r2 + 2 * r3]
7941     pextrd    [r5], m0, 2
7942     pextrd    [r5 + r3], m0, 3
7943
7944     lea       r5, [4 * r1 - 2 * 4]
7945     sub       r0, r5
7946     add       r2, 4
7947
7948     dec       r4d
7949     jnz       .loopW
7950
7951     lea       r0, [r0 + 4 * r1 - 2 * %1]
7952     lea       r2, [r2 + 4 * r3 - %1]
7953
7954     dec       dword [rsp]
7955     jnz       .loopH
7956
7957     RET
7958 %endmacro
7959
7960     FILTER_VER_CHROMA_SP 4, 4
7961     FILTER_VER_CHROMA_SP 4, 8
7962     FILTER_VER_CHROMA_SP 16, 16
7963     FILTER_VER_CHROMA_SP 16, 8
7964     FILTER_VER_CHROMA_SP 16, 12
7965     FILTER_VER_CHROMA_SP 12, 16
7966     FILTER_VER_CHROMA_SP 16, 4
7967     FILTER_VER_CHROMA_SP 4, 16
7968     FILTER_VER_CHROMA_SP 32, 32
7969     FILTER_VER_CHROMA_SP 32, 16
7970     FILTER_VER_CHROMA_SP 16, 32
7971     FILTER_VER_CHROMA_SP 32, 24
7972     FILTER_VER_CHROMA_SP 24, 32
7973     FILTER_VER_CHROMA_SP 32, 8
7974
7975     FILTER_VER_CHROMA_SP 16, 24
7976     FILTER_VER_CHROMA_SP 16, 64
7977     FILTER_VER_CHROMA_SP 12, 32
7978     FILTER_VER_CHROMA_SP 4, 32
7979     FILTER_VER_CHROMA_SP 32, 64
7980     FILTER_VER_CHROMA_SP 32, 48
7981     FILTER_VER_CHROMA_SP 24, 64
7982
7983     FILTER_VER_CHROMA_SP 64, 64
7984     FILTER_VER_CHROMA_SP 64, 32
7985     FILTER_VER_CHROMA_SP 64, 48
7986     FILTER_VER_CHROMA_SP 48, 64
7987     FILTER_VER_CHROMA_SP 64, 16
7988
7989
7990 %macro PROCESS_CHROMA_SP_W2_4R 1
7991     movd       m0, [r0]
7992     movd       m1, [r0 + r1]
7993     punpcklwd  m0, m1                          ;m0=[0 1]
7994
7995     lea        r0, [r0 + 2 * r1]
7996     movd       m2, [r0]
7997     punpcklwd  m1, m2                          ;m1=[1 2]
7998     punpcklqdq m0, m1                          ;m0=[0 1 1 2]
7999     pmaddwd    m0, [%1 + 0 *16]                ;m0=[0+1 1+2] Row 1-2
8000
8001     movd       m1, [r0 + r1]
8002     punpcklwd  m2, m1                          ;m2=[2 3]
8003
8004     lea        r0, [r0 + 2 * r1]
8005     movd       m3, [r0]
8006     punpcklwd  m1, m3                          ;m2=[3 4]
8007     punpcklqdq m2, m1                          ;m2=[2 3 3 4]
8008
8009     pmaddwd    m4, m2, [%1 + 1 * 16]           ;m4=[2+3 3+4] Row 1-2
8010     pmaddwd    m2, [%1 + 0 * 16]               ;m2=[2+3 3+4] Row 3-4
8011     paddd      m0, m4                          ;m0=[0+1+2+3 1+2+3+4] Row 1-2
8012
8013     movd       m1, [r0 + r1]
8014     punpcklwd  m3, m1                          ;m3=[4 5]
8015
8016     movd       m4, [r0 + 2 * r1]
8017     punpcklwd  m1, m4                          ;m1=[5 6]
8018     punpcklqdq m3, m1                          ;m2=[4 5 5 6]
8019     pmaddwd    m3, [%1 + 1 * 16]               ;m3=[4+5 5+6] Row 3-4
8020     paddd      m2, m3                          ;m2=[2+3+4+5 3+4+5+6] Row 3-4
8021 %endmacro
8022
8023 ;-------------------------------------------------------------------------------------------------------------------
8024 ; void interp_4tap_vertical_sp_%1x%2(int16_t *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
8025 ;-------------------------------------------------------------------------------------------------------------------
8026 %macro FILTER_VER_CHROMA_SP_W2_4R 2
8027 INIT_XMM sse4
8028 cglobal interp_4tap_vert_sp_%1x%2, 5, 6, 6
8029
8030     add       r1d, r1d
8031     sub       r0, r1
8032     shl       r4d, 5
8033
8034 %ifdef PIC
8035     lea       r5, [tab_ChromaCoeffV]
8036     lea       r5, [r5 + r4]
8037 %else
8038     lea       r5, [tab_ChromaCoeffV + r4]
8039 %endif
8040
8041     mova      m5, [tab_c_526336]
8042
8043     mov       r4d, (%2/4)
8044
8045 .loopH:
8046     PROCESS_CHROMA_SP_W2_4R r5
8047
8048     paddd     m0, m5
8049     paddd     m2, m5
8050
8051     psrad     m0, 12
8052     psrad     m2, 12
8053
8054     packssdw  m0, m2
8055     packuswb  m0, m0
8056
8057     pextrw    [r2], m0, 0
8058     pextrw    [r2 + r3], m0, 1
8059     lea       r2, [r2 + 2 * r3]
8060     pextrw    [r2], m0, 2
8061     pextrw    [r2 + r3], m0, 3
8062
8063     lea       r2, [r2 + 2 * r3]
8064
8065     dec       r4d
8066     jnz       .loopH
8067
8068     RET
8069 %endmacro
8070
8071 FILTER_VER_CHROMA_SP_W2_4R 2, 4
8072 FILTER_VER_CHROMA_SP_W2_4R 2, 8
8073
8074 FILTER_VER_CHROMA_SP_W2_4R 2, 16
8075
8076 ;--------------------------------------------------------------------------------------------------------------
8077 ; void interp_4tap_vert_sp_4x2(int16_t *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
8078 ;--------------------------------------------------------------------------------------------------------------
8079 INIT_XMM sse4
8080 cglobal interp_4tap_vert_sp_4x2, 5, 6, 5
8081
8082     add        r1d, r1d
8083     sub        r0, r1
8084     shl        r4d, 5
8085
8086 %ifdef PIC
8087     lea        r5, [tab_ChromaCoeffV]
8088     lea        r5, [r5 + r4]
8089 %else
8090     lea        r5, [tab_ChromaCoeffV + r4]
8091 %endif
8092
8093     mova       m4, [tab_c_526336]
8094
8095     movq       m0, [r0]
8096     movq       m1, [r0 + r1]
8097     punpcklwd  m0, m1                          ;m0=[0 1]
8098     pmaddwd    m0, [r5 + 0 *16]                ;m0=[0+1]  Row1
8099
8100     lea        r0, [r0 + 2 * r1]
8101     movq       m2, [r0]
8102     punpcklwd  m1, m2                          ;m1=[1 2]
8103     pmaddwd    m1, [r5 + 0 *16]                ;m1=[1+2]  Row2
8104
8105     movq       m3, [r0 + r1]
8106     punpcklwd  m2, m3                          ;m4=[2 3]
8107     pmaddwd    m2, [r5 + 1 * 16]
8108     paddd      m0, m2                          ;m0=[0+1+2+3]  Row1 done
8109     paddd      m0, m4
8110     psrad      m0, 12
8111
8112     movq       m2, [r0 + 2 * r1]
8113     punpcklwd  m3, m2                          ;m5=[3 4]
8114     pmaddwd    m3, [r5 + 1 * 16]
8115     paddd      m1, m3                          ;m1 = [1+2+3+4]  Row2 done
8116     paddd      m1, m4
8117     psrad      m1, 12
8118
8119     packssdw   m0, m1
8120     packuswb   m0, m0
8121
8122     movd       [r2], m0
8123     pextrd     [r2 + r3], m0, 1
8124
8125     RET
8126
8127 ;-------------------------------------------------------------------------------------------------------------------
8128 ; void interp_4tap_vertical_sp_6x%2(int16_t *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
8129 ;-------------------------------------------------------------------------------------------------------------------
8130 %macro FILTER_VER_CHROMA_SP_W6_H4 2
8131 INIT_XMM sse4
8132 cglobal interp_4tap_vert_sp_6x%2, 5, 7, 7
8133
8134     add       r1d, r1d
8135     sub       r0, r1
8136     shl       r4d, 5
8137
8138 %ifdef PIC
8139     lea       r5, [tab_ChromaCoeffV]
8140     lea       r6, [r5 + r4]
8141 %else
8142     lea       r6, [tab_ChromaCoeffV + r4]
8143 %endif
8144
8145     mova      m6, [tab_c_526336]
8146
8147     mov       r4d, %2/4
8148
8149 .loopH:
8150     PROCESS_CHROMA_SP_W4_4R
8151
8152     paddd     m0, m6
8153     paddd     m1, m6
8154     paddd     m2, m6
8155     paddd     m3, m6
8156
8157     psrad     m0, 12
8158     psrad     m1, 12
8159     psrad     m2, 12
8160     psrad     m3, 12
8161
8162     packssdw  m0, m1
8163     packssdw  m2, m3
8164
8165     packuswb  m0, m2
8166
8167     movd      [r2], m0
8168     pextrd    [r2 + r3], m0, 1
8169     lea       r5, [r2 + 2 * r3]
8170     pextrd    [r5], m0, 2
8171     pextrd    [r5 + r3], m0, 3
8172
8173     lea       r5, [4 * r1 - 2 * 4]
8174     sub       r0, r5
8175     add       r2, 4
8176
8177     PROCESS_CHROMA_SP_W2_4R r6
8178
8179     paddd     m0, m6
8180     paddd     m2, m6
8181
8182     psrad     m0, 12
8183     psrad     m2, 12
8184
8185     packssdw  m0, m2
8186     packuswb  m0, m0
8187
8188     pextrw    [r2], m0, 0
8189     pextrw    [r2 + r3], m0, 1
8190     lea       r2, [r2 + 2 * r3]
8191     pextrw    [r2], m0, 2
8192     pextrw    [r2 + r3], m0, 3
8193
8194     sub       r0, 2 * 4
8195     lea       r2, [r2 + 2 * r3 - 4]
8196
8197     dec       r4d
8198     jnz       .loopH
8199
8200     RET
8201 %endmacro
8202
8203 FILTER_VER_CHROMA_SP_W6_H4 6, 8
8204
8205 FILTER_VER_CHROMA_SP_W6_H4 6, 16
8206
8207 %macro PROCESS_CHROMA_SP_W8_2R 0
8208     movu       m1, [r0]
8209     movu       m3, [r0 + r1]
8210     punpcklwd  m0, m1, m3
8211     pmaddwd    m0, [r5 + 0 * 16]                ;m0 = [0l+1l]  Row1l
8212     punpckhwd  m1, m3
8213     pmaddwd    m1, [r5 + 0 * 16]                ;m1 = [0h+1h]  Row1h
8214
8215     movu       m4, [r0 + 2 * r1]
8216     punpcklwd  m2, m3, m4
8217     pmaddwd    m2, [r5 + 0 * 16]                ;m2 = [1l+2l]  Row2l
8218     punpckhwd  m3, m4
8219     pmaddwd    m3, [r5 + 0 * 16]                ;m3 = [1h+2h]  Row2h
8220
8221     lea        r0, [r0 + 2 * r1]
8222     movu       m5, [r0 + r1]
8223     punpcklwd  m6, m4, m5
8224     pmaddwd    m6, [r5 + 1 * 16]                ;m6 = [2l+3l]  Row1l
8225     paddd      m0, m6                           ;m0 = [0l+1l+2l+3l]  Row1l sum
8226     punpckhwd  m4, m5
8227     pmaddwd    m4, [r5 + 1 * 16]                ;m6 = [2h+3h]  Row1h
8228     paddd      m1, m4                           ;m1 = [0h+1h+2h+3h]  Row1h sum
8229
8230     movu       m4, [r0 + 2 * r1]
8231     punpcklwd  m6, m5, m4
8232     pmaddwd    m6, [r5 + 1 * 16]                ;m6 = [3l+4l]  Row2l
8233     paddd      m2, m6                           ;m2 = [1l+2l+3l+4l]  Row2l sum
8234     punpckhwd  m5, m4
8235     pmaddwd    m5, [r5 + 1 * 16]                ;m1 = [3h+4h]  Row2h
8236     paddd      m3, m5                           ;m3 = [1h+2h+3h+4h]  Row2h sum
8237 %endmacro
8238
8239 ;--------------------------------------------------------------------------------------------------------------
8240 ; void interp_4tap_vert_sp_8x%2(int16_t *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
8241 ;--------------------------------------------------------------------------------------------------------------
8242 %macro FILTER_VER_CHROMA_SP_W8_H2 2
8243 INIT_XMM sse2
8244 cglobal interp_4tap_vert_sp_%1x%2, 5, 6, 8
8245
8246     add       r1d, r1d
8247     sub       r0, r1
8248     shl       r4d, 5
8249
8250 %ifdef PIC
8251     lea       r5, [tab_ChromaCoeffV]
8252     lea       r5, [r5 + r4]
8253 %else
8254     lea       r5, [tab_ChromaCoeffV + r4]
8255 %endif
8256
8257     mova      m7, [tab_c_526336]
8258
8259     mov       r4d, %2/2
8260 .loopH:
8261     PROCESS_CHROMA_SP_W8_2R
8262
8263     paddd     m0, m7
8264     paddd     m1, m7
8265     paddd     m2, m7
8266     paddd     m3, m7
8267
8268     psrad     m0, 12
8269     psrad     m1, 12
8270     psrad     m2, 12
8271     psrad     m3, 12
8272
8273     packssdw  m0, m1
8274     packssdw  m2, m3
8275
8276     packuswb  m0, m2
8277
8278     movlps    [r2], m0
8279     movhps    [r2 + r3], m0
8280
8281     lea       r2, [r2 + 2 * r3]
8282
8283     dec r4d
8284     jnz .loopH
8285
8286     RET
8287 %endmacro
8288
8289 FILTER_VER_CHROMA_SP_W8_H2 8, 2
8290 FILTER_VER_CHROMA_SP_W8_H2 8, 4
8291 FILTER_VER_CHROMA_SP_W8_H2 8, 6
8292 FILTER_VER_CHROMA_SP_W8_H2 8, 8
8293 FILTER_VER_CHROMA_SP_W8_H2 8, 16
8294 FILTER_VER_CHROMA_SP_W8_H2 8, 32
8295
8296 FILTER_VER_CHROMA_SP_W8_H2 8, 12
8297 FILTER_VER_CHROMA_SP_W8_H2 8, 64
8298
8299
8300 ;-----------------------------------------------------------------------------------------------------------------------------
8301 ; void interp_4tap_horiz_ps_2x%2(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
8302 ;-----------------------------------------------------------------------------------------------------------------------------
8303 %macro FILTER_HORIZ_CHROMA_2xN 2
8304 INIT_XMM sse4
8305 cglobal interp_4tap_horiz_ps_%1x%2, 4, 7, 4, src, srcstride, dst, dststride
8306 %define coef2  m3
8307 %define Tm0    m2
8308 %define t1     m1
8309 %define t0     m0
8310
8311     dec        srcq
8312     mov        r4d, r4m
8313     add        dststrided, dststrided
8314
8315 %ifdef PIC
8316     lea        r6, [tab_ChromaCoeff]
8317     movd       coef2, [r6 + r4 * 4]
8318 %else
8319     movd       coef2, [tab_ChromaCoeff + r4 * 4]
8320 %endif
8321
8322     pshufd     coef2, coef2, 0
8323     mova       t1, [pw_2000]
8324     mova       Tm0, [tab_Tm]
8325
8326     mov        r4d, %2
8327     cmp        r5m, byte 0
8328     je         .loopH
8329     sub        srcq, srcstrideq
8330     add        r4d, 3
8331
8332 .loopH:
8333     movh       t0, [srcq]
8334     pshufb     t0, t0, Tm0
8335     pmaddubsw  t0, coef2
8336     phaddw     t0, t0
8337     psubw      t0, t1
8338     movd       [dstq], t0
8339
8340     lea        srcq, [srcq + srcstrideq]
8341     lea        dstq, [dstq + dststrideq]
8342
8343     dec        r4d
8344     jnz        .loopH
8345
8346     RET
8347 %endmacro
8348
8349 FILTER_HORIZ_CHROMA_2xN 2, 4
8350 FILTER_HORIZ_CHROMA_2xN 2, 8
8351
8352 FILTER_HORIZ_CHROMA_2xN 2, 16
8353
8354 ;-----------------------------------------------------------------------------------------------------------------------------
8355 ; void interp_4tap_horiz_ps_4x%2(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
8356 ;-----------------------------------------------------------------------------------------------------------------------------
8357 %macro FILTER_HORIZ_CHROMA_4xN 2
8358 INIT_XMM sse4
8359 cglobal interp_4tap_horiz_ps_%1x%2, 4, 7, 4, src, srcstride, dst, dststride
8360 %define coef2  m3
8361 %define Tm0    m2
8362 %define t1     m1
8363 %define t0     m0
8364
8365     dec        srcq
8366     mov        r4d, r4m
8367     add        dststrided, dststrided
8368
8369 %ifdef PIC
8370     lea        r6, [tab_ChromaCoeff]
8371     movd       coef2, [r6 + r4 * 4]
8372 %else
8373     movd       coef2, [tab_ChromaCoeff + r4 * 4]
8374 %endif
8375
8376     pshufd     coef2, coef2, 0
8377     mova       t1, [pw_2000]
8378     mova       Tm0, [tab_Tm]
8379
8380     mov        r4d, %2
8381     cmp        r5m, byte 0
8382     je         .loopH
8383     sub        srcq, srcstrideq
8384     add        r4d, 3
8385
8386 .loopH:
8387     movh       t0, [srcq]
8388     pshufb     t0, t0, Tm0
8389     pmaddubsw  t0, coef2
8390     phaddw     t0, t0
8391     psubw      t0, t1
8392     movlps     [dstq], t0
8393
8394     lea        srcq, [srcq + srcstrideq]
8395     lea        dstq, [dstq + dststrideq]
8396
8397     dec        r4d
8398     jnz        .loopH
8399     RET
8400 %endmacro
8401
8402 FILTER_HORIZ_CHROMA_4xN 4, 2
8403 FILTER_HORIZ_CHROMA_4xN 4, 4
8404 FILTER_HORIZ_CHROMA_4xN 4, 8
8405 FILTER_HORIZ_CHROMA_4xN 4, 16
8406
8407 FILTER_HORIZ_CHROMA_4xN 4, 32
8408
8409 %macro PROCESS_CHROMA_W6 3
8410     movu       %1, [srcq]
8411     pshufb     %2, %1, Tm0
8412     pmaddubsw  %2, coef2
8413     pshufb     %1, %1, Tm1
8414     pmaddubsw  %1, coef2
8415     phaddw     %2, %1
8416     psubw      %2, %3
8417     movh       [dstq], %2
8418     pshufd     %2, %2, 2
8419     movd       [dstq + 8], %2
8420 %endmacro
8421
8422 %macro PROCESS_CHROMA_W12 3
8423     movu       %1, [srcq]
8424     pshufb     %2, %1, Tm0
8425     pmaddubsw  %2, coef2
8426     pshufb     %1, %1, Tm1
8427     pmaddubsw  %1, coef2
8428     phaddw     %2, %1
8429     psubw      %2, %3
8430     movu       [dstq], %2
8431     movu       %1, [srcq + 8]
8432     pshufb     %1, %1, Tm0
8433     pmaddubsw  %1, coef2
8434     phaddw     %1, %1
8435     psubw      %1, %3
8436     movh       [dstq + 16], %1
8437 %endmacro
8438
8439 ;-----------------------------------------------------------------------------------------------------------------------------
8440 ; void interp_4tap_horiz_ps_6x%2(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
8441 ;-----------------------------------------------------------------------------------------------------------------------------
8442 %macro FILTER_HORIZ_CHROMA 2
8443 INIT_XMM sse4
8444 cglobal interp_4tap_horiz_ps_%1x%2, 4, 7, 6, src, srcstride, dst, dststride
8445 %define coef2    m5
8446 %define Tm0      m4
8447 %define Tm1      m3
8448 %define t2       m2
8449 %define t1       m1
8450 %define t0       m0
8451
8452     dec     srcq
8453     mov     r4d, r4m
8454     add     dststrided, dststrided
8455
8456 %ifdef PIC
8457     lea     r6, [tab_ChromaCoeff]
8458     movd    coef2, [r6 + r4 * 4]
8459 %else
8460     movd    coef2, [tab_ChromaCoeff + r4 * 4]
8461 %endif
8462
8463     pshufd  coef2, coef2, 0
8464     mova    t2, [pw_2000]
8465     mova    Tm0, [tab_Tm]
8466     mova    Tm1, [tab_Tm + 16]
8467
8468     mov     r4d, %2
8469     cmp     r5m, byte 0
8470     je      .loopH
8471     sub     srcq, srcstrideq
8472     add     r4d, 3
8473
8474 .loopH:
8475     PROCESS_CHROMA_W%1  t0, t1, t2
8476     add     srcq, srcstrideq
8477     add     dstq, dststrideq
8478
8479     dec     r4d
8480     jnz     .loopH
8481
8482     RET
8483 %endmacro
8484
8485 FILTER_HORIZ_CHROMA 6, 8
8486 FILTER_HORIZ_CHROMA 12, 16
8487
8488 FILTER_HORIZ_CHROMA 6, 16
8489 FILTER_HORIZ_CHROMA 12, 32
8490
8491 %macro PROCESS_CHROMA_W8 3
8492     movu        %1, [srcq]
8493     pshufb      %2, %1, Tm0
8494     pmaddubsw   %2, coef2
8495     pshufb      %1, %1, Tm1
8496     pmaddubsw   %1, coef2
8497     phaddw      %2, %1
8498     psubw       %2, %3
8499     movu        [dstq], %2
8500 %endmacro
8501
8502 ;-----------------------------------------------------------------------------------------------------------------------------
8503 ; void interp_4tap_horiz_ps_8x%2(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
8504 ;-----------------------------------------------------------------------------------------------------------------------------
8505 %macro FILTER_HORIZ_CHROMA_8xN 2
8506 INIT_XMM sse4
8507 cglobal interp_4tap_horiz_ps_%1x%2, 4, 7, 6, src, srcstride, dst, dststride
8508 %define coef2    m5
8509 %define Tm0      m4
8510 %define Tm1      m3
8511 %define t2       m2
8512 %define t1       m1
8513 %define t0       m0
8514
8515     dec     srcq
8516     mov     r4d, r4m
8517     add     dststrided, dststrided
8518
8519 %ifdef PIC
8520     lea     r6, [tab_ChromaCoeff]
8521     movd    coef2, [r6 + r4 * 4]
8522 %else
8523     movd    coef2, [tab_ChromaCoeff + r4 * 4]
8524 %endif
8525
8526     pshufd  coef2, coef2, 0
8527     mova    t2, [pw_2000]
8528     mova    Tm0, [tab_Tm]
8529     mova    Tm1, [tab_Tm + 16]
8530
8531     mov     r4d, %2
8532     cmp     r5m, byte 0
8533     je      .loopH
8534     sub     srcq, srcstrideq
8535     add     r4d, 3
8536
8537 .loopH:
8538     PROCESS_CHROMA_W8  t0, t1, t2
8539     add     srcq, srcstrideq
8540     add     dstq, dststrideq
8541
8542     dec     r4d
8543     jnz     .loopH
8544
8545     RET
8546 %endmacro
8547
8548 FILTER_HORIZ_CHROMA_8xN 8, 2
8549 FILTER_HORIZ_CHROMA_8xN 8, 4
8550 FILTER_HORIZ_CHROMA_8xN 8, 6
8551 FILTER_HORIZ_CHROMA_8xN 8, 8
8552 FILTER_HORIZ_CHROMA_8xN 8, 16
8553 FILTER_HORIZ_CHROMA_8xN 8, 32
8554
8555 FILTER_HORIZ_CHROMA_8xN 8, 12
8556 FILTER_HORIZ_CHROMA_8xN 8, 64
8557
8558 %macro PROCESS_CHROMA_W16 4
8559     movu        %1, [srcq]
8560     pshufb      %2, %1, Tm0
8561     pmaddubsw   %2, coef2
8562     pshufb      %1, %1, Tm1
8563     pmaddubsw   %1, coef2
8564     phaddw      %2, %1
8565     movu        %1, [srcq + 8]
8566     pshufb      %4, %1, Tm0
8567     pmaddubsw   %4, coef2
8568     pshufb      %1, %1, Tm1
8569     pmaddubsw   %1, coef2
8570     phaddw      %4, %1
8571     psubw       %2, %3
8572     psubw       %4, %3
8573     movu        [dstq], %2
8574     movu        [dstq + 16], %4
8575 %endmacro
8576
8577 %macro PROCESS_CHROMA_W24 4
8578     movu        %1, [srcq]
8579     pshufb      %2, %1, Tm0
8580     pmaddubsw   %2, coef2
8581     pshufb      %1, %1, Tm1
8582     pmaddubsw   %1, coef2
8583     phaddw      %2, %1
8584     movu        %1, [srcq + 8]
8585     pshufb      %4, %1, Tm0
8586     pmaddubsw   %4, coef2
8587     pshufb      %1, %1, Tm1
8588     pmaddubsw   %1, coef2
8589     phaddw      %4, %1
8590     psubw       %2, %3
8591     psubw       %4, %3
8592     movu        [dstq], %2
8593     movu        [dstq + 16], %4
8594     movu        %1, [srcq + 16]
8595     pshufb      %2, %1, Tm0
8596     pmaddubsw   %2, coef2
8597     pshufb      %1, %1, Tm1
8598     pmaddubsw   %1, coef2
8599     phaddw      %2, %1
8600     psubw       %2, %3
8601     movu        [dstq + 32], %2
8602 %endmacro
8603
8604 %macro PROCESS_CHROMA_W32 4
8605     movu        %1, [srcq]
8606     pshufb      %2, %1, Tm0
8607     pmaddubsw   %2, coef2
8608     pshufb      %1, %1, Tm1
8609     pmaddubsw   %1, coef2
8610     phaddw      %2, %1
8611     movu        %1, [srcq + 8]
8612     pshufb      %4, %1, Tm0
8613     pmaddubsw   %4, coef2
8614     pshufb      %1, %1, Tm1
8615     pmaddubsw   %1, coef2
8616     phaddw      %4, %1
8617     psubw       %2, %3
8618     psubw       %4, %3
8619     movu        [dstq], %2
8620     movu        [dstq + 16], %4
8621     movu        %1, [srcq + 16]
8622     pshufb      %2, %1, Tm0
8623     pmaddubsw   %2, coef2
8624     pshufb      %1, %1, Tm1
8625     pmaddubsw   %1, coef2
8626     phaddw      %2, %1
8627     movu        %1, [srcq + 24]
8628     pshufb      %4, %1, Tm0
8629     pmaddubsw   %4, coef2
8630     pshufb      %1, %1, Tm1
8631     pmaddubsw   %1, coef2
8632     phaddw      %4, %1
8633     psubw       %2, %3
8634     psubw       %4, %3
8635     movu        [dstq + 32], %2
8636     movu        [dstq + 48], %4
8637 %endmacro
8638
8639 %macro PROCESS_CHROMA_W16o 5
8640     movu        %1, [srcq + %5]
8641     pshufb      %2, %1, Tm0
8642     pmaddubsw   %2, coef2
8643     pshufb      %1, %1, Tm1
8644     pmaddubsw   %1, coef2
8645     phaddw      %2, %1
8646     movu        %1, [srcq + %5 + 8]
8647     pshufb      %4, %1, Tm0
8648     pmaddubsw   %4, coef2
8649     pshufb      %1, %1, Tm1
8650     pmaddubsw   %1, coef2
8651     phaddw      %4, %1
8652     psubw       %2, %3
8653     psubw       %4, %3
8654     movu        [dstq + %5 * 2], %2
8655     movu        [dstq + %5 * 2 + 16], %4
8656 %endmacro
8657
8658 %macro PROCESS_CHROMA_W48 4
8659     PROCESS_CHROMA_W16o %1, %2, %3, %4, 0
8660     PROCESS_CHROMA_W16o %1, %2, %3, %4, 16
8661     PROCESS_CHROMA_W16o %1, %2, %3, %4, 32
8662 %endmacro
8663
8664 %macro PROCESS_CHROMA_W64 4
8665     PROCESS_CHROMA_W16o %1, %2, %3, %4, 0
8666     PROCESS_CHROMA_W16o %1, %2, %3, %4, 16
8667     PROCESS_CHROMA_W16o %1, %2, %3, %4, 32
8668     PROCESS_CHROMA_W16o %1, %2, %3, %4, 48
8669 %endmacro
8670
8671 ;------------------------------------------------------------------------------------------------------------------------------
8672 ; void interp_4tap_horiz_ps_%1x%2(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
8673 ;------------------------------------------------------------------------------------------------------------------------------
8674 %macro FILTER_HORIZ_CHROMA_WxN 2
8675 INIT_XMM sse4
8676 cglobal interp_4tap_horiz_ps_%1x%2, 4, 7, 7, src, srcstride, dst, dststride
8677 %define coef2    m6
8678 %define Tm0      m5
8679 %define Tm1      m4
8680 %define t3       m3
8681 %define t2       m2
8682 %define t1       m1
8683 %define t0       m0
8684
8685     dec     srcq
8686     mov     r4d, r4m
8687     add     dststrided, dststrided
8688
8689 %ifdef PIC
8690     lea     r6, [tab_ChromaCoeff]
8691     movd    coef2, [r6 + r4 * 4]
8692 %else
8693     movd    coef2, [tab_ChromaCoeff + r4 * 4]
8694 %endif
8695
8696     pshufd  coef2, coef2, 0
8697     mova    t2, [pw_2000]
8698     mova    Tm0, [tab_Tm]
8699     mova    Tm1, [tab_Tm + 16]
8700
8701     mov     r4d, %2
8702     cmp     r5m, byte 0
8703     je      .loopH
8704     sub     srcq, srcstrideq
8705     add     r4d, 3
8706
8707 .loopH:
8708     PROCESS_CHROMA_W%1   t0, t1, t2, t3
8709     add     srcq, srcstrideq
8710     add     dstq, dststrideq
8711
8712     dec     r4d
8713     jnz     .loopH
8714
8715     RET
8716 %endmacro
8717
8718 FILTER_HORIZ_CHROMA_WxN 16, 4
8719 FILTER_HORIZ_CHROMA_WxN 16, 8
8720 FILTER_HORIZ_CHROMA_WxN 16, 12
8721 FILTER_HORIZ_CHROMA_WxN 16, 16
8722 FILTER_HORIZ_CHROMA_WxN 16, 32
8723 FILTER_HORIZ_CHROMA_WxN 24, 32
8724 FILTER_HORIZ_CHROMA_WxN 32,  8
8725 FILTER_HORIZ_CHROMA_WxN 32, 16
8726 FILTER_HORIZ_CHROMA_WxN 32, 24
8727 FILTER_HORIZ_CHROMA_WxN 32, 32
8728
8729 FILTER_HORIZ_CHROMA_WxN 16, 24
8730 FILTER_HORIZ_CHROMA_WxN 16, 64
8731 FILTER_HORIZ_CHROMA_WxN 24, 64
8732 FILTER_HORIZ_CHROMA_WxN 32, 48
8733 FILTER_HORIZ_CHROMA_WxN 32, 64
8734
8735 FILTER_HORIZ_CHROMA_WxN 64, 64
8736 FILTER_HORIZ_CHROMA_WxN 64, 32
8737 FILTER_HORIZ_CHROMA_WxN 64, 48
8738 FILTER_HORIZ_CHROMA_WxN 48, 64
8739 FILTER_HORIZ_CHROMA_WxN 64, 16
8740
8741
8742 ;---------------------------------------------------------------------------------------------------------------
8743 ; void interp_4tap_vert_ps_%1x%2(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
8744 ;---------------------------------------------------------------------------------------------------------------
8745 %macro FILTER_V_PS_W16n 2
8746 INIT_XMM sse4
8747 cglobal interp_4tap_vert_ps_%1x%2, 4, 7, 8
8748
8749     mov        r4d, r4m
8750     sub        r0, r1
8751     add        r3d, r3d
8752
8753 %ifdef PIC
8754     lea        r5, [tab_ChromaCoeff]
8755     movd       m0, [r5 + r4 * 4]
8756 %else
8757     movd       m0, [tab_ChromaCoeff + r4 * 4]
8758 %endif
8759
8760     pshufb     m1, m0, [tab_Vm]
8761     pshufb     m0, [tab_Vm + 16]
8762     mov        r4d, %2/2
8763
8764 .loop:
8765
8766     mov         r6d,       %1/16
8767
8768 .loopW:
8769
8770     movu       m2, [r0]
8771     movu       m3, [r0 + r1]
8772
8773     punpcklbw  m4, m2, m3
8774     punpckhbw  m2, m3
8775
8776     pmaddubsw  m4, m1
8777     pmaddubsw  m2, m1
8778
8779     lea        r5, [r0 + 2 * r1]
8780     movu       m5, [r5]
8781     movu       m7, [r5 + r1]
8782
8783     punpcklbw  m6, m5, m7
8784     pmaddubsw  m6, m0
8785     paddw      m4, m6
8786
8787     punpckhbw  m6, m5, m7
8788     pmaddubsw  m6, m0
8789     paddw      m2, m6
8790
8791     mova       m6, [pw_2000]
8792
8793     psubw      m4, m6
8794     psubw      m2, m6
8795
8796     movu       [r2], m4
8797     movu       [r2 + 16], m2
8798
8799     punpcklbw  m4, m3, m5
8800     punpckhbw  m3, m5
8801
8802     pmaddubsw  m4, m1
8803     pmaddubsw  m3, m1
8804
8805     movu       m5, [r5 + 2 * r1]
8806
8807     punpcklbw  m2, m7, m5
8808     punpckhbw  m7, m5
8809
8810     pmaddubsw  m2, m0
8811     pmaddubsw  m7, m0
8812
8813     paddw      m4, m2
8814     paddw      m3, m7
8815
8816     psubw      m4, m6
8817     psubw      m3, m6
8818
8819     movu       [r2 + r3], m4
8820     movu       [r2 + r3 + 16], m3
8821
8822     add         r0,        16
8823     add         r2,        32
8824     dec         r6d
8825     jnz         .loopW
8826
8827     lea         r0,        [r0 + r1 * 2 - %1]
8828     lea         r2,        [r2 + r3 * 2 - %1 * 2]
8829
8830     dec        r4d
8831     jnz        .loop
8832     RET
8833 %endmacro
8834
8835 FILTER_V_PS_W16n 64, 64
8836 FILTER_V_PS_W16n 64, 32
8837 FILTER_V_PS_W16n 64, 48
8838 FILTER_V_PS_W16n 48, 64
8839 FILTER_V_PS_W16n 64, 16
8840
8841
8842 ;------------------------------------------------------------------------------------------------------------
8843 ;void interp_4tap_vert_ps_2x4(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
8844 ;------------------------------------------------------------------------------------------------------------
8845 INIT_XMM sse4
8846 cglobal interp_4tap_vert_ps_2x4, 4, 6, 7
8847
8848     mov         r4d, r4m
8849     sub         r0, r1
8850     add         r3d, r3d
8851
8852 %ifdef PIC
8853     lea         r5, [tab_ChromaCoeff]
8854     movd        m0, [r5 + r4 * 4]
8855 %else
8856     movd        m0, [tab_ChromaCoeff + r4 * 4]
8857 %endif
8858
8859     pshufb      m0, [tab_Cm]
8860
8861     lea         r5, [3 * r1]
8862
8863     movd        m2, [r0]
8864     movd        m3, [r0 + r1]
8865     movd        m4, [r0 + 2 * r1]
8866     movd        m5, [r0 + r5]
8867
8868     punpcklbw   m2, m3
8869     punpcklbw   m6, m4, m5
8870     punpcklbw   m2, m6
8871
8872     pmaddubsw   m2, m0
8873
8874     lea         r0, [r0 + 4 * r1]
8875     movd        m6, [r0]
8876
8877     punpcklbw   m3, m4
8878     punpcklbw   m1, m5, m6
8879     punpcklbw   m3, m1
8880
8881     pmaddubsw   m3, m0
8882     phaddw      m2, m3
8883
8884     mova        m1, [pw_2000]
8885
8886     psubw       m2, m1
8887
8888     movd        [r2], m2
8889     pextrd      [r2 + r3], m2, 2
8890
8891     movd        m2, [r0 + r1]
8892
8893     punpcklbw   m4, m5
8894     punpcklbw   m3, m6, m2
8895     punpcklbw   m4, m3
8896
8897     pmaddubsw   m4, m0
8898
8899     movd        m3, [r0 + 2 * r1]
8900
8901     punpcklbw   m5, m6
8902     punpcklbw   m2, m3
8903     punpcklbw   m5, m2
8904
8905     pmaddubsw   m5, m0
8906     phaddw      m4, m5
8907     psubw       m4, m1
8908
8909     lea         r2, [r2 + 2 * r3]
8910     movd        [r2], m4
8911     pextrd      [r2 + r3], m4, 2
8912
8913     RET
8914
8915 ;-------------------------------------------------------------------------------------------------------------
8916 ; void interp_4tap_vert_ps_2x8(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
8917 ;-------------------------------------------------------------------------------------------------------------
8918 %macro FILTER_V_PS_W2 2
8919 INIT_XMM sse4
8920 cglobal interp_4tap_vert_ps_2x%2, 4, 6, 8
8921
8922     mov        r4d, r4m
8923     sub        r0, r1
8924     add        r3d, r3d
8925
8926 %ifdef PIC
8927     lea        r5, [tab_ChromaCoeff]
8928     movd       m0, [r5 + r4 * 4]
8929 %else
8930     movd       m0, [tab_ChromaCoeff + r4 * 4]
8931 %endif
8932
8933     pshufb     m0, [tab_Cm]
8934
8935     mova       m1, [pw_2000]
8936     lea        r5, [3 * r1]
8937     mov        r4d, %2/4
8938 .loop:
8939     movd       m2, [r0]
8940     movd       m3, [r0 + r1]
8941     movd       m4, [r0 + 2 * r1]
8942     movd       m5, [r0 + r5]
8943
8944     punpcklbw  m2, m3
8945     punpcklbw  m6, m4, m5
8946     punpcklbw  m2, m6
8947
8948     pmaddubsw  m2, m0
8949
8950     lea        r0, [r0 + 4 * r1]
8951     movd       m6, [r0]
8952
8953     punpcklbw  m3, m4
8954     punpcklbw  m7, m5, m6
8955     punpcklbw  m3, m7
8956
8957     pmaddubsw  m3, m0
8958
8959     phaddw     m2, m3
8960     psubw      m2, m1
8961
8962
8963     movd       [r2], m2
8964     pshufd     m2, m2, 2
8965     movd       [r2 + r3], m2
8966
8967     movd       m2, [r0 + r1]
8968
8969     punpcklbw  m4, m5
8970     punpcklbw  m3, m6, m2
8971     punpcklbw  m4, m3
8972
8973     pmaddubsw  m4, m0
8974
8975     movd       m3, [r0 + 2 * r1]
8976
8977     punpcklbw  m5, m6
8978     punpcklbw  m2, m3
8979     punpcklbw  m5, m2
8980
8981     pmaddubsw  m5, m0
8982
8983     phaddw     m4, m5
8984
8985     psubw      m4, m1
8986
8987     lea        r2, [r2 + 2 * r3]
8988     movd       [r2], m4
8989     pshufd     m4 , m4 ,2
8990     movd       [r2 + r3], m4
8991
8992     lea        r2, [r2 + 2 * r3]
8993
8994     dec        r4d
8995     jnz        .loop
8996
8997 RET
8998 %endmacro
8999
9000 FILTER_V_PS_W2 2, 8
9001
9002 FILTER_V_PS_W2 2, 16
9003
9004 ;-----------------------------------------------------------------------------------------------------------------
9005 ; void interp_4tap_vert_ss_%1x%2(int16_t *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
9006 ;-----------------------------------------------------------------------------------------------------------------
9007 %macro FILTER_VER_CHROMA_SS 2
9008 INIT_XMM sse2
9009 cglobal interp_4tap_vert_ss_%1x%2, 5, 7, 6 ,0-gprsize
9010
9011     add       r1d, r1d
9012     add       r3d, r3d
9013     sub       r0, r1
9014     shl       r4d, 5
9015
9016 %ifdef PIC
9017     lea       r5, [tab_ChromaCoeffV]
9018     lea       r6, [r5 + r4]
9019 %else
9020     lea       r6, [tab_ChromaCoeffV + r4]
9021 %endif
9022
9023     mov       dword [rsp], %2/4
9024
9025 .loopH:
9026     mov       r4d, (%1/4)
9027 .loopW:
9028     PROCESS_CHROMA_SP_W4_4R
9029
9030     psrad     m0, 6
9031     psrad     m1, 6
9032     psrad     m2, 6
9033     psrad     m3, 6
9034
9035     packssdw  m0, m1
9036     packssdw  m2, m3
9037
9038     movlps    [r2], m0
9039     movhps    [r2 + r3], m0
9040     lea       r5, [r2 + 2 * r3]
9041     movlps    [r5], m2
9042     movhps    [r5 + r3], m2
9043
9044     lea       r5, [4 * r1 - 2 * 4]
9045     sub       r0, r5
9046     add       r2, 2 * 4
9047
9048     dec       r4d
9049     jnz       .loopW
9050
9051     lea       r0, [r0 + 4 * r1 - 2 * %1]
9052     lea       r2, [r2 + 4 * r3 - 2 * %1]
9053
9054     dec       dword [rsp]
9055     jnz       .loopH
9056
9057     RET
9058 %endmacro
9059
9060     FILTER_VER_CHROMA_SS 4, 4
9061     FILTER_VER_CHROMA_SS 4, 8
9062     FILTER_VER_CHROMA_SS 16, 16
9063     FILTER_VER_CHROMA_SS 16, 8
9064     FILTER_VER_CHROMA_SS 16, 12
9065     FILTER_VER_CHROMA_SS 12, 16
9066     FILTER_VER_CHROMA_SS 16, 4
9067     FILTER_VER_CHROMA_SS 4, 16
9068     FILTER_VER_CHROMA_SS 32, 32
9069     FILTER_VER_CHROMA_SS 32, 16
9070     FILTER_VER_CHROMA_SS 16, 32
9071     FILTER_VER_CHROMA_SS 32, 24
9072     FILTER_VER_CHROMA_SS 24, 32
9073     FILTER_VER_CHROMA_SS 32, 8
9074
9075     FILTER_VER_CHROMA_SS 16, 24
9076     FILTER_VER_CHROMA_SS 12, 32
9077     FILTER_VER_CHROMA_SS 4, 32
9078     FILTER_VER_CHROMA_SS 32, 64
9079     FILTER_VER_CHROMA_SS 16, 64
9080     FILTER_VER_CHROMA_SS 32, 48
9081     FILTER_VER_CHROMA_SS 24, 64
9082
9083     FILTER_VER_CHROMA_SS 64, 64
9084     FILTER_VER_CHROMA_SS 64, 32
9085     FILTER_VER_CHROMA_SS 64, 48
9086     FILTER_VER_CHROMA_SS 48, 64
9087     FILTER_VER_CHROMA_SS 64, 16
9088
9089
9090 ;---------------------------------------------------------------------------------------------------------------------
9091 ; void interp_4tap_vertical_ss_%1x%2(int16_t *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
9092 ;---------------------------------------------------------------------------------------------------------------------
9093 %macro FILTER_VER_CHROMA_SS_W2_4R 2
9094 INIT_XMM sse4
9095 cglobal interp_4tap_vert_ss_%1x%2, 5, 6, 5
9096
9097     add       r1d, r1d
9098     add       r3d, r3d
9099     sub       r0, r1
9100     shl       r4d, 5
9101
9102 %ifdef PIC
9103     lea       r5, [tab_ChromaCoeffV]
9104     lea       r5, [r5 + r4]
9105 %else
9106     lea       r5, [tab_ChromaCoeffV + r4]
9107 %endif
9108
9109     mov       r4d, (%2/4)
9110
9111 .loopH:
9112     PROCESS_CHROMA_SP_W2_4R r5
9113
9114     psrad     m0, 6
9115     psrad     m2, 6
9116
9117     packssdw  m0, m2
9118
9119     movd      [r2], m0
9120     pextrd    [r2 + r3], m0, 1
9121     lea       r2, [r2 + 2 * r3]
9122     pextrd    [r2], m0, 2
9123     pextrd    [r2 + r3], m0, 3
9124
9125     lea       r2, [r2 + 2 * r3]
9126
9127     dec       r4d
9128     jnz       .loopH
9129
9130     RET
9131 %endmacro
9132
9133 FILTER_VER_CHROMA_SS_W2_4R 2, 4
9134 FILTER_VER_CHROMA_SS_W2_4R 2, 8
9135
9136 FILTER_VER_CHROMA_SS_W2_4R 2, 16
9137
9138 ;---------------------------------------------------------------------------------------------------------------
9139 ; void interp_4tap_vert_ss_4x2(int16_t *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
9140 ;---------------------------------------------------------------------------------------------------------------
9141 INIT_XMM sse2
9142 cglobal interp_4tap_vert_ss_4x2, 5, 6, 4
9143
9144     add        r1d, r1d
9145     add        r3d, r3d
9146     sub        r0, r1
9147     shl        r4d, 5
9148
9149 %ifdef PIC
9150     lea        r5, [tab_ChromaCoeffV]
9151     lea        r5, [r5 + r4]
9152 %else
9153     lea        r5, [tab_ChromaCoeffV + r4]
9154 %endif
9155
9156     movq       m0, [r0]
9157     movq       m1, [r0 + r1]
9158     punpcklwd  m0, m1                          ;m0=[0 1]
9159     pmaddwd    m0, [r5 + 0 *16]                ;m0=[0+1]  Row1
9160
9161     lea        r0, [r0 + 2 * r1]
9162     movq       m2, [r0]
9163     punpcklwd  m1, m2                          ;m1=[1 2]
9164     pmaddwd    m1, [r5 + 0 *16]                ;m1=[1+2]  Row2
9165
9166     movq       m3, [r0 + r1]
9167     punpcklwd  m2, m3                          ;m4=[2 3]
9168     pmaddwd    m2, [r5 + 1 * 16]
9169     paddd      m0, m2                          ;m0=[0+1+2+3]  Row1 done
9170     psrad      m0, 6
9171
9172     movq       m2, [r0 + 2 * r1]
9173     punpcklwd  m3, m2                          ;m5=[3 4]
9174     pmaddwd    m3, [r5 + 1 * 16]
9175     paddd      m1, m3                          ;m1=[1+2+3+4]  Row2 done
9176     psrad      m1, 6
9177
9178     packssdw   m0, m1
9179
9180     movlps     [r2], m0
9181     movhps     [r2 + r3], m0
9182
9183     RET
9184
9185 ;-------------------------------------------------------------------------------------------------------------------
9186 ; void interp_4tap_vertical_ss_6x8(int16_t *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
9187 ;-------------------------------------------------------------------------------------------------------------------
9188 %macro FILTER_VER_CHROMA_SS_W6_H4 2
9189 INIT_XMM sse4
9190 cglobal interp_4tap_vert_ss_6x%2, 5, 7, 6
9191
9192     add       r1d, r1d
9193     add       r3d, r3d
9194     sub       r0, r1
9195     shl       r4d, 5
9196
9197 %ifdef PIC
9198     lea       r5, [tab_ChromaCoeffV]
9199     lea       r6, [r5 + r4]
9200 %else
9201     lea       r6, [tab_ChromaCoeffV + r4]
9202 %endif
9203
9204     mov       r4d, %2/4
9205
9206 .loopH:
9207     PROCESS_CHROMA_SP_W4_4R
9208
9209     psrad     m0, 6
9210     psrad     m1, 6
9211     psrad     m2, 6
9212     psrad     m3, 6
9213
9214     packssdw  m0, m1
9215     packssdw  m2, m3
9216
9217     movlps    [r2], m0
9218     movhps    [r2 + r3], m0
9219     lea       r5, [r2 + 2 * r3]
9220     movlps    [r5], m2
9221     movhps    [r5 + r3], m2
9222
9223     lea       r5, [4 * r1 - 2 * 4]
9224     sub       r0, r5
9225     add       r2, 2 * 4
9226
9227     PROCESS_CHROMA_SP_W2_4R r6
9228
9229     psrad     m0, 6
9230     psrad     m2, 6
9231
9232     packssdw  m0, m2
9233
9234     movd      [r2], m0
9235     pextrd    [r2 + r3], m0, 1
9236     lea       r2, [r2 + 2 * r3]
9237     pextrd    [r2], m0, 2
9238     pextrd    [r2 + r3], m0, 3
9239
9240     sub       r0, 2 * 4
9241     lea       r2, [r2 + 2 * r3 - 2 * 4]
9242
9243     dec       r4d
9244     jnz       .loopH
9245
9246     RET
9247 %endmacro
9248
9249 FILTER_VER_CHROMA_SS_W6_H4 6, 8
9250
9251 FILTER_VER_CHROMA_SS_W6_H4 6, 16
9252
9253
9254 ;----------------------------------------------------------------------------------------------------------------
9255 ; void interp_4tap_vert_ss_8x%2(int16_t *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
9256 ;----------------------------------------------------------------------------------------------------------------
9257 %macro FILTER_VER_CHROMA_SS_W8_H2 2
9258 INIT_XMM sse2
9259 cglobal interp_4tap_vert_ss_%1x%2, 5, 6, 7
9260
9261     add       r1d, r1d
9262     add       r3d, r3d
9263     sub       r0, r1
9264     shl       r4d, 5
9265
9266 %ifdef PIC
9267     lea       r5, [tab_ChromaCoeffV]
9268     lea       r5, [r5 + r4]
9269 %else
9270     lea       r5, [tab_ChromaCoeffV + r4]
9271 %endif
9272
9273     mov       r4d, %2/2
9274 .loopH:
9275     PROCESS_CHROMA_SP_W8_2R
9276
9277     psrad     m0, 6
9278     psrad     m1, 6
9279     psrad     m2, 6
9280     psrad     m3, 6
9281
9282     packssdw  m0, m1
9283     packssdw  m2, m3
9284
9285     movu      [r2], m0
9286     movu      [r2 + r3], m2
9287
9288     lea       r2, [r2 + 2 * r3]
9289
9290     dec       r4d
9291     jnz       .loopH
9292
9293     RET
9294 %endmacro
9295
9296 FILTER_VER_CHROMA_SS_W8_H2 8, 2
9297 FILTER_VER_CHROMA_SS_W8_H2 8, 4
9298 FILTER_VER_CHROMA_SS_W8_H2 8, 6
9299 FILTER_VER_CHROMA_SS_W8_H2 8, 8
9300 FILTER_VER_CHROMA_SS_W8_H2 8, 16
9301 FILTER_VER_CHROMA_SS_W8_H2 8, 32
9302
9303 FILTER_VER_CHROMA_SS_W8_H2 8, 12
9304 FILTER_VER_CHROMA_SS_W8_H2 8, 64
9305
9306 ;-----------------------------------------------------------------------------------------------------------------
9307 ; void interp_8tap_vert_ss_%1x%2(int16_t *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
9308 ;-----------------------------------------------------------------------------------------------------------------
9309 %macro FILTER_VER_LUMA_SS 2
9310 INIT_XMM sse2
9311 cglobal interp_8tap_vert_ss_%1x%2, 5, 7, 7 ,0-gprsize
9312
9313     add        r1d, r1d
9314     add        r3d, r3d
9315     lea        r5, [3 * r1]
9316     sub        r0, r5
9317     shl        r4d, 6
9318
9319 %ifdef PIC
9320     lea        r5, [tab_LumaCoeffV]
9321     lea        r6, [r5 + r4]
9322 %else
9323     lea        r6, [tab_LumaCoeffV + r4]
9324 %endif
9325
9326     mov        dword [rsp], %2/4
9327 .loopH:
9328     mov        r4d, (%1/4)
9329 .loopW:
9330     movq       m0, [r0]
9331     movq       m1, [r0 + r1]
9332     punpcklwd  m0, m1                          ;m0=[0 1]
9333     pmaddwd    m0, [r6 + 0 *16]                ;m0=[0+1]  Row1
9334
9335     lea        r0, [r0 + 2 * r1]
9336     movq       m4, [r0]
9337     punpcklwd  m1, m4                          ;m1=[1 2]
9338     pmaddwd    m1, [r6 + 0 *16]                ;m1=[1+2]  Row2
9339
9340     movq       m5, [r0 + r1]
9341     punpcklwd  m4, m5                          ;m4=[2 3]
9342     pmaddwd    m2, m4, [r6 + 0 *16]            ;m2=[2+3]  Row3
9343     pmaddwd    m4, [r6 + 1 * 16]
9344     paddd      m0, m4                          ;m0=[0+1+2+3]  Row1
9345
9346     lea        r0, [r0 + 2 * r1]
9347     movq       m4, [r0]
9348     punpcklwd  m5, m4                          ;m5=[3 4]
9349     pmaddwd    m3, m5, [r6 + 0 *16]            ;m3=[3+4]  Row4
9350     pmaddwd    m5, [r6 + 1 * 16]
9351     paddd      m1, m5                          ;m1 = [1+2+3+4]  Row2
9352
9353     movq       m5, [r0 + r1]
9354     punpcklwd  m4, m5                          ;m4=[4 5]
9355     pmaddwd    m6, m4, [r6 + 1 * 16]
9356     paddd      m2, m6                          ;m2=[2+3+4+5]  Row3
9357     pmaddwd    m4, [r6 + 2 * 16]
9358     paddd      m0, m4                          ;m0=[0+1+2+3+4+5]  Row1
9359
9360     lea        r0, [r0 + 2 * r1]
9361     movq       m4, [r0]
9362     punpcklwd  m5, m4                          ;m5=[5 6]
9363     pmaddwd    m6, m5, [r6 + 1 * 16]
9364     paddd      m3, m6                          ;m3=[3+4+5+6]  Row4
9365     pmaddwd    m5, [r6 + 2 * 16]
9366     paddd      m1, m5                          ;m1=[1+2+3+4+5+6]  Row2
9367
9368     movq       m5, [r0 + r1]
9369     punpcklwd  m4, m5                          ;m4=[6 7]
9370     pmaddwd    m6, m4, [r6 + 2 * 16]
9371     paddd      m2, m6                          ;m2=[2+3+4+5+6+7]  Row3
9372     pmaddwd    m4, [r6 + 3 * 16]
9373     paddd      m0, m4                          ;m0=[0+1+2+3+4+5+6+7]  Row1 end
9374     psrad      m0, 6
9375
9376     lea        r0, [r0 + 2 * r1]
9377     movq       m4, [r0]
9378     punpcklwd  m5, m4                          ;m5=[7 8]
9379     pmaddwd    m6, m5, [r6 + 2 * 16]
9380     paddd      m3, m6                          ;m3=[3+4+5+6+7+8]  Row4
9381     pmaddwd    m5, [r6 + 3 * 16]
9382     paddd      m1, m5                          ;m1=[1+2+3+4+5+6+7+8]  Row2 end
9383     psrad      m1, 6
9384
9385     packssdw   m0, m1
9386
9387     movlps     [r2], m0
9388     movhps     [r2 + r3], m0
9389
9390     movq       m5, [r0 + r1]
9391     punpcklwd  m4, m5                          ;m4=[8 9]
9392     pmaddwd    m4, [r6 + 3 * 16]
9393     paddd      m2, m4                          ;m2=[2+3+4+5+6+7+8+9]  Row3 end
9394     psrad      m2, 6
9395
9396     movq       m4, [r0 + 2 * r1]
9397     punpcklwd  m5, m4                          ;m5=[9 10]
9398     pmaddwd    m5, [r6 + 3 * 16]
9399     paddd      m3, m5                          ;m3=[3+4+5+6+7+8+9+10]  Row4 end
9400     psrad      m3, 6
9401
9402     packssdw   m2, m3
9403
9404     movlps     [r2 + 2 * r3], m2
9405     lea        r5, [3 * r3]
9406     movhps     [r2 + r5], m2
9407
9408     lea        r5, [8 * r1 - 2 * 4]
9409     sub        r0, r5
9410     add        r2, 2 * 4
9411
9412     dec        r4d
9413     jnz        .loopW
9414
9415     lea        r0, [r0 + 4 * r1 - 2 * %1]
9416     lea        r2, [r2 + 4 * r3 - 2 * %1]
9417
9418     dec        dword [rsp]
9419     jnz        .loopH
9420
9421     RET
9422 %endmacro
9423
9424     FILTER_VER_LUMA_SS 4, 4
9425     FILTER_VER_LUMA_SS 8, 8
9426     FILTER_VER_LUMA_SS 8, 4
9427     FILTER_VER_LUMA_SS 4, 8
9428     FILTER_VER_LUMA_SS 16, 16
9429     FILTER_VER_LUMA_SS 16, 8
9430     FILTER_VER_LUMA_SS 8, 16
9431     FILTER_VER_LUMA_SS 16, 12
9432     FILTER_VER_LUMA_SS 12, 16
9433     FILTER_VER_LUMA_SS 16, 4
9434     FILTER_VER_LUMA_SS 4, 16
9435     FILTER_VER_LUMA_SS 32, 32
9436     FILTER_VER_LUMA_SS 32, 16
9437     FILTER_VER_LUMA_SS 16, 32
9438     FILTER_VER_LUMA_SS 32, 24
9439     FILTER_VER_LUMA_SS 24, 32
9440     FILTER_VER_LUMA_SS 32, 8
9441     FILTER_VER_LUMA_SS 8, 32
9442     FILTER_VER_LUMA_SS 64, 64
9443     FILTER_VER_LUMA_SS 64, 32
9444     FILTER_VER_LUMA_SS 32, 64
9445     FILTER_VER_LUMA_SS 64, 48
9446     FILTER_VER_LUMA_SS 48, 64
9447     FILTER_VER_LUMA_SS 64, 16
9448     FILTER_VER_LUMA_SS 16, 64