vp9intrapred_16bpp.asm 74.5 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
;******************************************************************************
;* VP9 Intra prediction SIMD optimizations
;*
;* Copyright (c) 2015 Ronald S. Bultje <rsbultje gmail com>
;* Copyright (c) 2015 Henrik Gramner <henrik gramner com>
;*
;* This file is part of FFmpeg.
;*
;* FFmpeg is free software; you can redistribute it and/or
;* modify it under the terms of the GNU Lesser General Public
;* License as published by the Free Software Foundation; either
;* version 2.1 of the License, or (at your option) any later version.
;*
;* FFmpeg is distributed in the hope that it will be useful,
;* but WITHOUT ANY WARRANTY; without even the implied warranty of
;* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
;* Lesser General Public License for more details.
;*
;* You should have received a copy of the GNU Lesser General Public
;* License along with FFmpeg; if not, write to the Free Software
;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
;******************************************************************************

%include "libavutil/x86/x86util.asm"

SECTION_RODATA 32

pd_2: times 8 dd 2
pd_4: times 8 dd 4
pd_8: times 8 dd 8

32 33 34 35
pb_2to15_14_15: db 2,3,4,5,6,7,8,9,10,11,12,13,14,15,14,15
pb_4_5_8to13_8x0: db 4,5,8,9,10,11,12,13,0,0,0,0,0,0,0,0
pb_0to7_67x4: db 0,1,2,3,4,5,6,7,6,7,6,7,6,7,6,7

36 37 38 39 40
cextern pw_1
cextern pw_1023
cextern pw_4095
cextern pd_16
cextern pd_32
41 42 43 44 45
cextern pd_65535;

; FIXME most top-only functions (ddl, vl, v, dc_top) can be modified to take
; only 3 registers on x86-32, which would make it one cycle faster, but that
; would make the code quite a bit uglier...
46 47 48

SECTION .text

49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84
%macro SCRATCH 3-4
%if ARCH_X86_64
    SWAP                %1, %2
%if %0 == 4
%define reg_%4 m%2
%endif
%else
    mova              [%3], m%1
%if %0 == 4
%define reg_%4 [%3]
%endif
%endif
%endmacro

%macro UNSCRATCH 3-4
%if ARCH_X86_64
    SWAP                %1, %2
%else
    mova               m%1, [%3]
%endif
%if %0 == 4
%undef reg_%4
%endif
%endmacro

%macro PRELOAD 2-3
%if ARCH_X86_64
    mova               m%1, [%2]
%if %0 == 3
%define reg_%3 m%1
%endif
%elif %0 == 3
%define reg_%3 [%2]
%endif
%endmacro

85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603
INIT_MMX mmx
cglobal vp9_ipred_v_4x4_16, 2, 4, 1, dst, stride, l, a
    movifnidn               aq, amp
    mova                    m0, [aq]
    DEFINE_ARGS dst, stride, stride3
    lea               stride3q, [strideq*3]
    mova      [dstq+strideq*0], m0
    mova      [dstq+strideq*1], m0
    mova      [dstq+strideq*2], m0
    mova      [dstq+stride3q ], m0
    RET

INIT_XMM sse
cglobal vp9_ipred_v_8x8_16, 2, 4, 1, dst, stride, l, a
    movifnidn               aq, amp
    mova                    m0, [aq]
    DEFINE_ARGS dst, stride, stride3
    lea               stride3q, [strideq*3]
    mova      [dstq+strideq*0], m0
    mova      [dstq+strideq*1], m0
    mova      [dstq+strideq*2], m0
    mova      [dstq+stride3q ], m0
    lea                   dstq, [dstq+strideq*4]
    mova      [dstq+strideq*0], m0
    mova      [dstq+strideq*1], m0
    mova      [dstq+strideq*2], m0
    mova      [dstq+stride3q ], m0
    RET

INIT_XMM sse
cglobal vp9_ipred_v_16x16_16, 2, 4, 2, dst, stride, l, a
    movifnidn               aq, amp
    mova                    m0, [aq]
    mova                    m1, [aq+mmsize]
    DEFINE_ARGS dst, stride, stride3, cnt
    lea               stride3q, [strideq*3]
    mov                   cntd, 4
.loop:
    mova   [dstq+strideq*0+ 0], m0
    mova   [dstq+strideq*0+16], m1
    mova   [dstq+strideq*1+ 0], m0
    mova   [dstq+strideq*1+16], m1
    mova   [dstq+strideq*2+ 0], m0
    mova   [dstq+strideq*2+16], m1
    mova   [dstq+stride3q + 0], m0
    mova   [dstq+stride3q +16], m1
    lea                   dstq, [dstq+strideq*4]
    dec               cntd
    jg .loop
    RET

INIT_XMM sse
cglobal vp9_ipred_v_32x32_16, 2, 4, 4, dst, stride, l, a
    movifnidn               aq, amp
    mova                    m0, [aq+mmsize*0]
    mova                    m1, [aq+mmsize*1]
    mova                    m2, [aq+mmsize*2]
    mova                    m3, [aq+mmsize*3]
    DEFINE_ARGS dst, stride, cnt
    mov                   cntd, 16
.loop:
    mova   [dstq+strideq*0+ 0], m0
    mova   [dstq+strideq*0+16], m1
    mova   [dstq+strideq*0+32], m2
    mova   [dstq+strideq*0+48], m3
    mova   [dstq+strideq*1+ 0], m0
    mova   [dstq+strideq*1+16], m1
    mova   [dstq+strideq*1+32], m2
    mova   [dstq+strideq*1+48], m3
    lea                   dstq, [dstq+strideq*2]
    dec               cntd
    jg .loop
    RET

INIT_MMX mmxext
cglobal vp9_ipred_h_4x4_16, 3, 3, 4, dst, stride, l, a
    mova                    m3, [lq]
    DEFINE_ARGS dst, stride, stride3
    lea               stride3q, [strideq*3]
    pshufw                  m0, m3, q3333
    pshufw                  m1, m3, q2222
    pshufw                  m2, m3, q1111
    pshufw                  m3, m3, q0000
    mova      [dstq+strideq*0], m0
    mova      [dstq+strideq*1], m1
    mova      [dstq+strideq*2], m2
    mova      [dstq+stride3q ], m3
    RET

INIT_XMM sse2
cglobal vp9_ipred_h_8x8_16, 3, 3, 4, dst, stride, l, a
    mova                    m2, [lq]
    DEFINE_ARGS dst, stride, stride3
    lea               stride3q, [strideq*3]
    punpckhwd               m3, m2, m2
    pshufd                  m0, m3, q3333
    pshufd                  m1, m3, q2222
    mova      [dstq+strideq*0], m0
    mova      [dstq+strideq*1], m1
    pshufd                  m0, m3, q1111
    pshufd                  m1, m3, q0000
    mova      [dstq+strideq*2], m0
    mova      [dstq+stride3q ], m1
    lea                   dstq, [dstq+strideq*4]
    punpcklwd               m2, m2
    pshufd                  m0, m2, q3333
    pshufd                  m1, m2, q2222
    mova      [dstq+strideq*0], m0
    mova      [dstq+strideq*1], m1
    pshufd                  m0, m2, q1111
    pshufd                  m1, m2, q0000
    mova      [dstq+strideq*2], m0
    mova      [dstq+stride3q ], m1
    RET

INIT_XMM sse2
cglobal vp9_ipred_h_16x16_16, 3, 5, 4, dst, stride, l, stride3, cnt
    mov                   cntd, 3
    lea               stride3q, [strideq*3]
.loop:
    movh                    m3, [lq+cntq*8]
    punpcklwd               m3, m3
    pshufd                  m0, m3, q3333
    pshufd                  m1, m3, q2222
    pshufd                  m2, m3, q1111
    pshufd                  m3, m3, q0000
    mova    [dstq+strideq*0+ 0], m0
    mova    [dstq+strideq*0+16], m0
    mova    [dstq+strideq*1+ 0], m1
    mova    [dstq+strideq*1+16], m1
    mova    [dstq+strideq*2+ 0], m2
    mova    [dstq+strideq*2+16], m2
    mova    [dstq+stride3q + 0], m3
    mova    [dstq+stride3q +16], m3
    lea                   dstq, [dstq+strideq*4]
    dec                   cntd
    jge .loop
    RET

INIT_XMM sse2
cglobal vp9_ipred_h_32x32_16, 3, 5, 4, dst, stride, l, stride3, cnt
    mov                   cntd, 7
    lea               stride3q, [strideq*3]
.loop:
    movh                    m3, [lq+cntq*8]
    punpcklwd               m3, m3
    pshufd                  m0, m3, q3333
    pshufd                  m1, m3, q2222
    pshufd                  m2, m3, q1111
    pshufd                  m3, m3, q0000
    mova   [dstq+strideq*0+ 0], m0
    mova   [dstq+strideq*0+16], m0
    mova   [dstq+strideq*0+32], m0
    mova   [dstq+strideq*0+48], m0
    mova   [dstq+strideq*1+ 0], m1
    mova   [dstq+strideq*1+16], m1
    mova   [dstq+strideq*1+32], m1
    mova   [dstq+strideq*1+48], m1
    mova   [dstq+strideq*2+ 0], m2
    mova   [dstq+strideq*2+16], m2
    mova   [dstq+strideq*2+32], m2
    mova   [dstq+strideq*2+48], m2
    mova   [dstq+stride3q + 0], m3
    mova   [dstq+stride3q +16], m3
    mova   [dstq+stride3q +32], m3
    mova   [dstq+stride3q +48], m3
    lea                   dstq, [dstq+strideq*4]
    dec                   cntd
    jge .loop
    RET

INIT_MMX mmxext
cglobal vp9_ipred_dc_4x4_16, 4, 4, 2, dst, stride, l, a
    mova                    m0, [lq]
    paddw                   m0, [aq]
    DEFINE_ARGS dst, stride, stride3
    lea               stride3q, [strideq*3]
    pmaddwd                 m0, [pw_1]
    pshufw                  m1, m0, q3232
    paddd                   m0, [pd_4]
    paddd                   m0, m1
    psrad                   m0, 3
    pshufw                  m0, m0, q0000
    mova      [dstq+strideq*0], m0
    mova      [dstq+strideq*1], m0
    mova      [dstq+strideq*2], m0
    mova      [dstq+stride3q ], m0
    RET

INIT_XMM sse2
cglobal vp9_ipred_dc_8x8_16, 4, 4, 2, dst, stride, l, a
    mova                    m0, [lq]
    paddw                   m0, [aq]
    DEFINE_ARGS dst, stride, stride3
    lea               stride3q, [strideq*3]
    pmaddwd                 m0, [pw_1]
    pshufd                  m1, m0, q3232
    paddd                   m0, m1
    pshufd                  m1, m0, q1111
    paddd                   m0, [pd_8]
    paddd                   m0, m1
    psrad                   m0, 4
    pshuflw                 m0, m0, q0000
    punpcklqdq              m0, m0
    mova      [dstq+strideq*0], m0
    mova      [dstq+strideq*1], m0
    mova      [dstq+strideq*2], m0
    mova      [dstq+stride3q ], m0
    lea                   dstq, [dstq+strideq*4]
    mova      [dstq+strideq*0], m0
    mova      [dstq+strideq*1], m0
    mova      [dstq+strideq*2], m0
    mova      [dstq+stride3q ], m0
    RET

INIT_XMM sse2
cglobal vp9_ipred_dc_16x16_16, 4, 4, 2, dst, stride, l, a
    mova                    m0, [lq]
    paddw                   m0, [lq+mmsize]
    paddw                   m0, [aq]
    paddw                   m0, [aq+mmsize]
    DEFINE_ARGS dst, stride, stride3, cnt
    lea               stride3q, [strideq*3]
    mov                   cntd, 4
    pmaddwd                 m0, [pw_1]
    pshufd                  m1, m0, q3232
    paddd                   m0, m1
    pshufd                  m1, m0, q1111
    paddd                   m0, [pd_16]
    paddd                   m0, m1
    psrad                   m0, 5
    pshuflw                 m0, m0, q0000
    punpcklqdq              m0, m0
.loop:
    mova   [dstq+strideq*0+ 0], m0
    mova   [dstq+strideq*0+16], m0
    mova   [dstq+strideq*1+ 0], m0
    mova   [dstq+strideq*1+16], m0
    mova   [dstq+strideq*2+ 0], m0
    mova   [dstq+strideq*2+16], m0
    mova   [dstq+stride3q + 0], m0
    mova   [dstq+stride3q +16], m0
    lea                   dstq, [dstq+strideq*4]
    dec                   cntd
    jg .loop
    RET

INIT_XMM sse2
cglobal vp9_ipred_dc_32x32_16, 4, 4, 2, dst, stride, l, a
    mova                    m0, [lq+mmsize*0]
    paddw                   m0, [lq+mmsize*1]
    paddw                   m0, [lq+mmsize*2]
    paddw                   m0, [lq+mmsize*3]
    paddw                   m0, [aq+mmsize*0]
    paddw                   m0, [aq+mmsize*1]
    paddw                   m0, [aq+mmsize*2]
    paddw                   m0, [aq+mmsize*3]
    DEFINE_ARGS dst, stride, stride3, cnt
    lea               stride3q, [strideq*3]
    mov                   cntd, 16
    pmaddwd                 m0, [pw_1]
    pshufd                  m1, m0, q3232
    paddd                   m0, m1
    pshufd                  m1, m0, q1111
    paddd                   m0, [pd_32]
    paddd                   m0, m1
    psrad                   m0, 6
    pshuflw                 m0, m0, q0000
    punpcklqdq              m0, m0
.loop:
    mova   [dstq+strideq*0+ 0], m0
    mova   [dstq+strideq*0+16], m0
    mova   [dstq+strideq*0+32], m0
    mova   [dstq+strideq*0+48], m0
    mova   [dstq+strideq*1+ 0], m0
    mova   [dstq+strideq*1+16], m0
    mova   [dstq+strideq*1+32], m0
    mova   [dstq+strideq*1+48], m0
    lea                   dstq, [dstq+strideq*2]
    dec                   cntd
    jg .loop
    RET

%macro DC_1D_FNS 2
INIT_MMX mmxext
cglobal vp9_ipred_dc_%1_4x4_16, 4, 4, 2, dst, stride, l, a
    mova                    m0, [%2]
    DEFINE_ARGS dst, stride, stride3
    lea               stride3q, [strideq*3]
    pmaddwd                 m0, [pw_1]
    pshufw                  m1, m0, q3232
    paddd                   m0, [pd_2]
    paddd                   m0, m1
    psrad                   m0, 2
    pshufw                  m0, m0, q0000
    mova      [dstq+strideq*0], m0
    mova      [dstq+strideq*1], m0
    mova      [dstq+strideq*2], m0
    mova      [dstq+stride3q ], m0
    RET

INIT_XMM sse2
cglobal vp9_ipred_dc_%1_8x8_16, 4, 4, 2, dst, stride, l, a
    mova                    m0, [%2]
    DEFINE_ARGS dst, stride, stride3
    lea               stride3q, [strideq*3]
    pmaddwd                 m0, [pw_1]
    pshufd                  m1, m0, q3232
    paddd                   m0, m1
    pshufd                  m1, m0, q1111
    paddd                   m0, [pd_4]
    paddd                   m0, m1
    psrad                   m0, 3
    pshuflw                 m0, m0, q0000
    punpcklqdq              m0, m0
    mova      [dstq+strideq*0], m0
    mova      [dstq+strideq*1], m0
    mova      [dstq+strideq*2], m0
    mova      [dstq+stride3q ], m0
    lea                   dstq, [dstq+strideq*4]
    mova      [dstq+strideq*0], m0
    mova      [dstq+strideq*1], m0
    mova      [dstq+strideq*2], m0
    mova      [dstq+stride3q ], m0
    RET

INIT_XMM sse2
cglobal vp9_ipred_dc_%1_16x16_16, 4, 4, 2, dst, stride, l, a
    mova                    m0, [%2]
    paddw                   m0, [%2+mmsize]
    DEFINE_ARGS dst, stride, stride3, cnt
    lea               stride3q, [strideq*3]
    mov                   cntd, 4
    pmaddwd                 m0, [pw_1]
    pshufd                  m1, m0, q3232
    paddd                   m0, m1
    pshufd                  m1, m0, q1111
    paddd                   m0, [pd_8]
    paddd                   m0, m1
    psrad                   m0, 4
    pshuflw                 m0, m0, q0000
    punpcklqdq              m0, m0
.loop:
    mova   [dstq+strideq*0+ 0], m0
    mova   [dstq+strideq*0+16], m0
    mova   [dstq+strideq*1+ 0], m0
    mova   [dstq+strideq*1+16], m0
    mova   [dstq+strideq*2+ 0], m0
    mova   [dstq+strideq*2+16], m0
    mova   [dstq+stride3q + 0], m0
    mova   [dstq+stride3q +16], m0
    lea                   dstq, [dstq+strideq*4]
    dec                   cntd
    jg .loop
    RET

INIT_XMM sse2
cglobal vp9_ipred_dc_%1_32x32_16, 4, 4, 2, dst, stride, l, a
    mova                    m0, [%2+mmsize*0]
    paddw                   m0, [%2+mmsize*1]
    paddw                   m0, [%2+mmsize*2]
    paddw                   m0, [%2+mmsize*3]
    DEFINE_ARGS dst, stride, cnt
    mov                   cntd, 16
    pmaddwd                 m0, [pw_1]
    pshufd                  m1, m0, q3232
    paddd                   m0, m1
    pshufd                  m1, m0, q1111
    paddd                   m0, [pd_16]
    paddd                   m0, m1
    psrad                   m0, 5
    pshuflw                 m0, m0, q0000
    punpcklqdq              m0, m0
.loop:
    mova   [dstq+strideq*0+ 0], m0
    mova   [dstq+strideq*0+16], m0
    mova   [dstq+strideq*0+32], m0
    mova   [dstq+strideq*0+48], m0
    mova   [dstq+strideq*1+ 0], m0
    mova   [dstq+strideq*1+16], m0
    mova   [dstq+strideq*1+32], m0
    mova   [dstq+strideq*1+48], m0
    lea                   dstq, [dstq+strideq*2]
    dec                   cntd
    jg .loop
    RET
%endmacro

DC_1D_FNS top,  aq
DC_1D_FNS left, lq

INIT_MMX mmxext
cglobal vp9_ipred_tm_4x4_10, 4, 4, 6, dst, stride, l, a
    mova                    m5, [pw_1023]
.body:
    mova                    m4, [aq]
    mova                    m3, [lq]
    movd                    m0, [aq-4]
    pshufw                  m0, m0, q1111
    psubw                   m4, m0
    DEFINE_ARGS dst, stride, stride3
    lea               stride3q, [strideq*3]
    pshufw                  m0, m3, q3333
    pshufw                  m1, m3, q2222
    pshufw                  m2, m3, q1111
    pshufw                  m3, m3, q0000
    paddw                   m0, m4
    paddw                   m1, m4
    paddw                   m2, m4
    paddw                   m3, m4
    pxor                    m4, m4
    pmaxsw                  m0, m4
    pmaxsw                  m1, m4
    pmaxsw                  m2, m4
    pmaxsw                  m3, m4
    pminsw                  m0, m5
    pminsw                  m1, m5
    pminsw                  m2, m5
    pminsw                  m3, m5
    mova      [dstq+strideq*0], m0
    mova      [dstq+strideq*1], m1
    mova      [dstq+strideq*2], m2
    mova      [dstq+stride3q ], m3
    RET

cglobal vp9_ipred_tm_4x4_12, 4, 4, 6, dst, stride, l, a
    mova                    m5, [pw_4095]
    jmp mangle(private_prefix %+ _ %+ vp9_ipred_tm_4x4_10 %+ SUFFIX).body

INIT_XMM sse2
cglobal vp9_ipred_tm_8x8_10, 4, 5, 7, dst, stride, l, a
    mova                    m4, [pw_1023]
.body:
    pxor                    m6, m6
    mova                    m5, [aq]
    movd                    m0, [aq-4]
    pshuflw                 m0, m0, q1111
    punpcklqdq              m0, m0
    psubw                   m5, m0
    DEFINE_ARGS dst, stride, l, stride3, cnt
    lea               stride3q, [strideq*3]
    mov                   cntd, 1
.loop:
    movh                    m3, [lq+cntq*8]
    punpcklwd               m3, m3
    pshufd                  m0, m3, q3333
    pshufd                  m1, m3, q2222
    pshufd                  m2, m3, q1111
    pshufd                  m3, m3, q0000
    paddw                   m0, m5
    paddw                   m1, m5
    paddw                   m2, m5
    paddw                   m3, m5
    pmaxsw                  m0, m6
    pmaxsw                  m1, m6
    pmaxsw                  m2, m6
    pmaxsw                  m3, m6
    pminsw                  m0, m4
    pminsw                  m1, m4
    pminsw                  m2, m4
    pminsw                  m3, m4
    mova      [dstq+strideq*0], m0
    mova      [dstq+strideq*1], m1
    mova      [dstq+strideq*2], m2
    mova      [dstq+stride3q ], m3
    lea                   dstq, [dstq+strideq*4]
    dec                   cntd
    jge .loop
    RET

cglobal vp9_ipred_tm_8x8_12, 4, 5, 7, dst, stride, l, a
    mova                    m4, [pw_4095]
    jmp mangle(private_prefix %+ _ %+ vp9_ipred_tm_8x8_10 %+ SUFFIX).body

INIT_XMM sse2
cglobal vp9_ipred_tm_16x16_10, 4, 4, 8, dst, stride, l, a
    mova                    m7, [pw_1023]
.body:
    pxor                    m6, m6
    mova                    m4, [aq]
    mova                    m5, [aq+mmsize]
    movd                    m0, [aq-4]
    pshuflw                 m0, m0, q1111
    punpcklqdq              m0, m0
    psubw                   m4, m0
    psubw                   m5, m0
    DEFINE_ARGS dst, stride, l, cnt
    mov                   cntd, 7
.loop:
    movd                    m3, [lq+cntq*4]
    punpcklwd               m3, m3
    pshufd                  m2, m3, q1111
    pshufd                  m3, m3, q0000
    paddw                   m0, m2, m4
    paddw                   m2, m5
    paddw                   m1, m3, m4
    paddw                   m3, m5
    pmaxsw                  m0, m6
    pmaxsw                  m2, m6
    pmaxsw                  m1, m6
    pmaxsw                  m3, m6
    pminsw                  m0, m7
    pminsw                  m2, m7
    pminsw                  m1, m7
    pminsw                  m3, m7
    mova   [dstq+strideq*0+ 0], m0
    mova   [dstq+strideq*0+16], m2
    mova   [dstq+strideq*1+ 0], m1
    mova   [dstq+strideq*1+16], m3
    lea                   dstq, [dstq+strideq*2]
    dec                   cntd
    jge .loop
    RET

cglobal vp9_ipred_tm_16x16_12, 4, 4, 8, dst, stride, l, a
    mova                    m7, [pw_4095]
    jmp mangle(private_prefix %+ _ %+ vp9_ipred_tm_16x16_10 %+ SUFFIX).body

INIT_XMM sse2
604
cglobal vp9_ipred_tm_32x32_10, 4, 4, 10, 32 * -ARCH_X86_32, dst, stride, l, a
605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657
    mova                    m0, [pw_1023]
.body:
    pxor                    m1, m1
%if ARCH_X86_64
    SWAP                     0, 8
    SWAP                     1, 9
%define reg_min m9
%define reg_max m8
%else
    mova              [rsp+ 0], m0
    mova              [rsp+16], m1
%define reg_min [rsp+16]
%define reg_max [rsp+ 0]
%endif

    mova                    m4, [aq+mmsize*0]
    mova                    m5, [aq+mmsize*1]
    mova                    m6, [aq+mmsize*2]
    mova                    m7, [aq+mmsize*3]
    movd                    m0, [aq-4]
    pshuflw                 m0, m0, q1111
    punpcklqdq              m0, m0
    psubw                   m4, m0
    psubw                   m5, m0
    psubw                   m6, m0
    psubw                   m7, m0
    DEFINE_ARGS dst, stride, l, cnt
    mov                   cntd, 31
.loop:
    pinsrw                  m3, [lq+cntq*2], 0
    punpcklwd               m3, m3
    pshufd                  m3, m3, q0000
    paddw                   m0, m3, m4
    paddw                   m1, m3, m5
    paddw                   m2, m3, m6
    paddw                   m3, m7
    pmaxsw                  m0, reg_min
    pmaxsw                  m1, reg_min
    pmaxsw                  m2, reg_min
    pmaxsw                  m3, reg_min
    pminsw                  m0, reg_max
    pminsw                  m1, reg_max
    pminsw                  m2, reg_max
    pminsw                  m3, reg_max
    mova   [dstq+strideq*0+ 0], m0
    mova   [dstq+strideq*0+16], m1
    mova   [dstq+strideq*0+32], m2
    mova   [dstq+strideq*0+48], m3
    add                   dstq, strideq
    dec                   cntd
    jge .loop
    RET

658
cglobal vp9_ipred_tm_32x32_12, 4, 4, 10, 32 * -ARCH_X86_32, dst, stride, l, a
659 660
    mova                    m0, [pw_4095]
    jmp mangle(private_prefix %+ _ %+ vp9_ipred_tm_32x32_10 %+ SUFFIX).body
661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947

; Directional intra predicion functions
;
; in the functions below, 'abcdefgh' refers to above data (sometimes simply
; abbreviated as a[N-M]). 'stuvwxyz' refers to left data (sometimes simply
; abbreviated as l[N-M]). * is top-left data. ABCDEFG or A[N-M] is filtered
; above data, STUVWXYZ or L[N-M] is filtered left data, and # is filtered
; top-left data.

; left=(left+2*center+right+2)>>2
%macro LOWPASS 3 ; left [dst], center, right
    paddw                  m%1, m%3
    psraw                  m%1, 1
    pavgw                  m%1, m%2
%endmacro

; abcdefgh (src) -> bcdefghh (dst)
; dst/src can be the same register
%macro SHIFT_RIGHT 2-3 [pb_2to15_14_15] ; dst, src, [ssse3_shift_reg]
%if cpuflag(ssse3)
    pshufb                  %1, %2, %3              ; abcdefgh -> bcdefghh
%else
    psrldq                  %1, %2, 2               ; abcdefgh -> bcdefgh.
    pshufhw                 %1, %1, q2210           ; bcdefgh. -> bcdefghh
%endif
%endmacro

; abcdefgh (src) -> bcdefghh (dst1) and cdefghhh (dst2)
%macro SHIFT_RIGHTx2 3-4 [pb_2to15_14_15] ; dst1, dst2, src, [ssse3_shift_reg]
%if cpuflag(ssse3)
    pshufb                  %1, %3, %4              ; abcdefgh -> bcdefghh
    pshufb                  %2, %1, %4              ; bcdefghh -> cdefghhh
%else
    psrldq                  %1, %3, 2               ; abcdefgh -> bcdefgh.
    psrldq                  %2, %3, 4               ; abcdefgh -> cdefgh..
    pshufhw                 %1, %1, q2210           ; bcdefgh. -> bcdefghh
    pshufhw                 %2, %2, q1110           ; cdefgh.. -> cdefghhh
%endif
%endmacro

%macro DL_FUNCS 0
cglobal vp9_ipred_dl_4x4_16, 2, 4, 3, dst, stride, l, a
    movifnidn               aq, amp
    movu                    m1, [aq]                ; abcdefgh
    pshufhw                 m0, m1, q3310           ; abcdefhh
    SHIFT_RIGHT             m1, m1                  ; bcdefghh
    psrldq                  m2, m1, 2               ; cdefghh.
    LOWPASS                  0,  1,  2              ; BCDEFGh.
    pshufd                  m1, m0, q3321           ; DEFGh...
    movh      [dstq+strideq*0], m0
    movh      [dstq+strideq*2], m1
    add                   dstq, strideq
    psrldq                  m0, 2                   ; CDEFGh..
    psrldq                  m1, 2                   ; EFGh....
    movh      [dstq+strideq*0], m0
    movh      [dstq+strideq*2], m1
    RET

cglobal vp9_ipred_dl_8x8_16, 2, 4, 5, dst, stride, l, a
    movifnidn               aq, amp
    mova                    m0, [aq]                ; abcdefgh
%if cpuflag(ssse3)
    mova                    m4, [pb_2to15_14_15]
%endif
    SHIFT_RIGHTx2           m1, m2, m0, m4          ; bcdefghh/cdefghhh
    LOWPASS                  0,  1,  2              ; BCDEFGHh
    shufps                  m1, m0, m2, q3332       ; FGHhhhhh
    shufps                  m3, m0, m1, q2121       ; DEFGHhhh
    DEFINE_ARGS dst, stride, stride5
    lea               stride5q, [strideq*5]

    mova      [dstq+strideq*0], m0
    mova      [dstq+strideq*4], m1
    SHIFT_RIGHT             m0, m0, m4              ; CDEFGHhh
    pshuflw                 m1, m1, q3321           ; GHhhhhhh
    pshufd                  m2, m0, q3321           ; EFGHhhhh
    mova      [dstq+strideq*1], m0
    mova      [dstq+stride5q ], m1
    lea                   dstq, [dstq+strideq*2]
    pshuflw                 m1, m1, q3321           ; Hhhhhhhh
    mova      [dstq+strideq*0], m3
    mova      [dstq+strideq*4], m1
    pshuflw                 m1, m1, q3321           ; hhhhhhhh
    mova      [dstq+strideq*1], m2
    mova      [dstq+stride5q ], m1
    RET

cglobal vp9_ipred_dl_16x16_16, 2, 4, 5, dst, stride, l, a
    movifnidn               aq, amp
    mova                    m0, [aq]                ; abcdefgh
    mova                    m3, [aq+mmsize]         ; ijklmnop
    PALIGNR                 m1, m3, m0, 2, m4       ; bcdefghi
    PALIGNR                 m2, m3, m0, 4, m4       ; cdefghij
    LOWPASS                  0,  1,  2              ; BCDEFGHI
%if cpuflag(ssse3)
    mova                    m4, [pb_2to15_14_15]
%endif
    SHIFT_RIGHTx2           m2, m1, m3, m4          ; jklmnopp/klmnoppp
    LOWPASS                  1,  2,  3              ; JKLMNOPp
    pshufd                  m2, m2, q3333           ; pppppppp
    DEFINE_ARGS dst, stride, cnt
    mov                   cntd, 8

.loop:
    mova   [dstq+strideq*0+ 0], m0
    mova   [dstq+strideq*0+16], m1
    mova   [dstq+strideq*8+ 0], m1
    mova   [dstq+strideq*8+16], m2
    add                   dstq, strideq
%if cpuflag(avx)
    vpalignr                m0, m1, m0, 2
%else
    PALIGNR                 m3, m1, m0, 2, m4
    mova                    m0, m3
%endif
    SHIFT_RIGHT             m1, m1, m4
    dec                   cntd
    jg .loop
    RET

cglobal vp9_ipred_dl_32x32_16, 2, 5, 7, dst, stride, l, a
    movifnidn               aq, amp
    mova                    m0, [aq+mmsize*0]       ; abcdefgh
    mova                    m1, [aq+mmsize*1]       ; ijklmnop
    mova                    m2, [aq+mmsize*2]       ; qrstuvwx
    mova                    m3, [aq+mmsize*3]       ; yz012345
    PALIGNR                 m4, m1, m0, 2, m6
    PALIGNR                 m5, m1, m0, 4, m6
    LOWPASS                  0,  4,  5              ; BCDEFGHI
    PALIGNR                 m4, m2, m1, 2, m6
    PALIGNR                 m5, m2, m1, 4, m6
    LOWPASS                  1,  4,  5              ; JKLMNOPQ
    PALIGNR                 m4, m3, m2, 2, m6
    PALIGNR                 m5, m3, m2, 4, m6
    LOWPASS                  2,  4,  5              ; RSTUVWXY
%if cpuflag(ssse3)
    mova                    m6, [pb_2to15_14_15]
%endif
    SHIFT_RIGHTx2           m4, m5, m3, m6
    LOWPASS                  3,  4,  5              ; Z0123455
    pshufd                  m4, m4, q3333           ; 55555555
    DEFINE_ARGS dst, stride, stride8, stride24, cnt
    mov                   cntd, 8
    lea               stride8q, [strideq*8]
    lea              stride24q, [stride8q*3]

.loop:
    mova  [dstq+stride8q*0+ 0], m0
    mova  [dstq+stride8q*0+16], m1
    mova  [dstq+stride8q*0+32], m2
    mova  [dstq+stride8q*0+48], m3
    mova  [dstq+stride8q*1+ 0], m1
    mova  [dstq+stride8q*1+16], m2
    mova  [dstq+stride8q*1+32], m3
    mova  [dstq+stride8q*1+48], m4
    mova  [dstq+stride8q*2+ 0], m2
    mova  [dstq+stride8q*2+16], m3
    mova  [dstq+stride8q*2+32], m4
    mova  [dstq+stride8q*2+48], m4
    mova  [dstq+stride24q + 0], m3
    mova  [dstq+stride24q +16], m4
    mova  [dstq+stride24q +32], m4
    mova  [dstq+stride24q +48], m4
    add                   dstq, strideq
%if cpuflag(avx)
    vpalignr                m0, m1, m0, 2
    vpalignr                m1, m2, m1, 2
    vpalignr                m2, m3, m2, 2
%else
    PALIGNR                 m5, m1, m0, 2, m6
    mova                    m0, m5
    PALIGNR                 m5, m2, m1, 2, m6
    mova                    m1, m5
    PALIGNR                 m5, m3, m2, 2, m6
    mova                    m2, m5
%endif
    SHIFT_RIGHT             m3, m3, m6
    dec                   cntd
    jg .loop
    RET
%endmacro

INIT_XMM sse2
DL_FUNCS
INIT_XMM ssse3
DL_FUNCS
INIT_XMM avx
DL_FUNCS

%macro DR_FUNCS 1 ; stack_mem_for_32x32_32bit_function
cglobal vp9_ipred_dr_4x4_16, 4, 4, 3, dst, stride, l, a
    movh                    m0, [lq]                ; wxyz....
    movhps                  m0, [aq-2]              ; wxyz*abc
    movd                    m1, [aq+6]              ; d.......
    PALIGNR                 m1, m0, 2, m2           ; xyz*abcd
    psrldq                  m2, m1, 2               ; yz*abcd.
    LOWPASS                  0, 1, 2                ; XYZ#ABC.
    DEFINE_ARGS dst, stride, stride3
    lea               stride3q, [strideq*3]

    movh      [dstq+stride3q ], m0
    psrldq                  m0, 2                   ; YZ#ABC..
    movh      [dstq+strideq*2], m0
    psrldq                  m0, 2                   ; Z#ABC...
    movh      [dstq+strideq*1], m0
    psrldq                  m0, 2                   ; #ABC....
    movh      [dstq+strideq*0], m0
    RET

cglobal vp9_ipred_dr_8x8_16, 4, 4, 5, dst, stride, l, a
    mova                    m0, [lq]                ; stuvwxyz
    movu                    m1, [aq-2]              ; *abcdefg
    mova                    m2, [aq]                ; abcdefgh
    psrldq                  m3, m2, 2               ; bcdefgh.
    LOWPASS                  3,  2, 1               ; ABCDEFG.
    PALIGNR                 m1, m0, 2, m4           ; tuvwxyz*
    PALIGNR                 m2, m1, 2, m4           ; uvwxyz*a
    LOWPASS                  2,  1, 0               ; TUVWXYZ#
    DEFINE_ARGS dst, stride, dst4, stride3
    lea               stride3q, [strideq*3]
    lea                  dst4q, [dstq+strideq*4]

    movhps [dstq +stride3q +0], m2
    movh   [dstq+ stride3q +8], m3
    mova   [dst4q+stride3q +0], m2
    PALIGNR                 m1, m3, m2, 2, m0
    psrldq                  m3, 2
    movhps [dstq +strideq*2+0], m1
    movh   [dstq+ strideq*2+8], m3
    mova   [dst4q+strideq*2+0], m1
    PALIGNR                 m2, m3, m1, 2, m0
    psrldq                  m3, 2
    movhps [dstq +strideq*1+0], m2
    movh   [dstq+ strideq*1+8], m3
    mova   [dst4q+strideq*1+0], m2
    PALIGNR                 m1, m3, m2, 2, m0
    psrldq                  m3, 2
    movhps [dstq +strideq*0+0], m1
    movh   [dstq+ strideq*0+8], m3
    mova   [dst4q+strideq*0+0], m1
    RET

cglobal vp9_ipred_dr_16x16_16, 4, 4, 7, dst, stride, l, a
    mova                    m0, [lq]                ; klmnopqr
    mova                    m1, [lq+mmsize]         ; stuvwxyz
    movu                    m2, [aq-2]              ; *abcdefg
    movu                    m3, [aq+mmsize-2]       ; hijklmno
    mova                    m4, [aq]                ; abcdefgh
    mova                    m5, [aq+mmsize]         ; ijklmnop
    psrldq                  m6, m5, 2               ; jklmnop.
    LOWPASS                  6,  5, 3               ; IJKLMNO.
    PALIGNR                 m5, m4, 2, m3           ; bcdefghi
    LOWPASS                  5,  4, 2               ; ABCDEFGH
    PALIGNR                 m2, m1, 2, m3           ; tuvwxyz*
    PALIGNR                 m4, m2, 2, m3           ; uvwxyz*a
    LOWPASS                  4,  2, 1               ; TUVWXYZ#
    PALIGNR                 m1, m0, 2, m3           ; lmnopqrs
    PALIGNR                 m2, m1, 2, m3           ; mnopqrst
    LOWPASS                  2, 1, 0                ; LMNOPQRS
    DEFINE_ARGS dst, stride, dst8, cnt
    lea                  dst8q, [dstq+strideq*8]
    mov                   cntd, 8

.loop:
    sub                  dst8q, strideq
    mova  [dst8q+strideq*0+ 0], m4
    mova  [dst8q+strideq*0+16], m5
    mova  [dst8q+strideq*8+ 0], m2
    mova  [dst8q+strideq*8+16], m4
%if cpuflag(avx)
    vpalignr                m2, m4, m2, 2
    vpalignr                m4, m5, m4, 2
    vpalignr                m5, m6, m5, 2
%else
    PALIGNR                 m0, m4, m2, 2, m1
    mova                    m2, m0
    PALIGNR                 m0, m5, m4, 2, m1
    mova                    m4, m0
    PALIGNR                 m0, m6, m5, 2, m1
    mova                    m5, m0
%endif
    psrldq                  m6, 2
    dec                   cntd
    jg .loop
    RET

cglobal vp9_ipred_dr_32x32_16, 4, 5, 10 + notcpuflag(ssse3), \
948
                               %1 * ARCH_X86_32 * -mmsize, dst, stride, l, a
949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637
    mova                    m0, [aq+mmsize*3]       ; a[24-31]
    movu                    m1, [aq+mmsize*3-2]     ; a[23-30]
    psrldq                  m2, m0, 2               ; a[25-31].
    LOWPASS                  2,  0, 1               ; A[24-30].
    mova                    m1, [aq+mmsize*2]       ; a[16-23]
    movu                    m3, [aq+mmsize*2-2]     ; a[15-22]
    PALIGNR                 m0, m1, 2, m4           ; a[17-24]
    LOWPASS                  0,  1, 3               ; A[16-23]
    mova                    m3, [aq+mmsize*1]       ; a[8-15]
    movu                    m4, [aq+mmsize*1-2]     ; a[7-14]
    PALIGNR                 m1, m3, 2, m5           ; a[9-16]
    LOWPASS                  1,  3, 4               ; A[8-15]
    mova                    m4, [aq+mmsize*0]       ; a[0-7]
    movu                    m5, [aq+mmsize*0-2]     ; *a[0-6]
    PALIGNR                 m3, m4, 2, m6           ; a[1-8]
    LOWPASS                  3,  4, 5               ; A[0-7]
    SCRATCH                  1,  8, rsp+0*mmsize
    SCRATCH                  3,  9, rsp+1*mmsize
%if notcpuflag(ssse3)
    SCRATCH                  0, 10, rsp+2*mmsize
%endif
    mova                    m6, [lq+mmsize*3]       ; l[24-31]
    PALIGNR                 m5, m6, 2, m0           ; l[25-31]*
    PALIGNR                 m4, m5, 2, m0           ; l[26-31]*a
    LOWPASS                  4,  5, 6               ; L[25-31]#
    mova                    m7, [lq+mmsize*2]       ; l[16-23]
    PALIGNR                 m6, m7, 2, m0           ; l[17-24]
    PALIGNR                 m5, m6, 2, m0           ; l[18-25]
    LOWPASS                  5,  6, 7               ; L[17-24]
    mova                    m1, [lq+mmsize*1]       ; l[8-15]
    PALIGNR                 m7, m1, 2, m0           ; l[9-16]
    PALIGNR                 m6, m7, 2, m0           ; l[10-17]
    LOWPASS                  6,  7, 1               ; L[9-16]
    mova                    m3, [lq+mmsize*0]       ; l[0-7]
    PALIGNR                 m1, m3, 2, m0           ; l[1-8]
    PALIGNR                 m7, m1, 2, m0           ; l[2-9]
    LOWPASS                  7,  1, 3               ; L[1-8]
%if cpuflag(ssse3)
%if cpuflag(avx)
    UNSCRATCH                1,  8, rsp+0*mmsize
%endif
    UNSCRATCH                3,  9, rsp+1*mmsize
%else
    UNSCRATCH                0, 10, rsp+2*mmsize
%endif
    DEFINE_ARGS dst8, stride, stride8, stride24, cnt
    lea               stride8q, [strideq*8]
    lea              stride24q, [stride8q*3]
    lea                  dst8q, [dst8q+strideq*8]
    mov                   cntd, 8

.loop:
    sub                  dst8q, strideq
%if notcpuflag(avx)
    UNSCRATCH                1,  8, rsp+0*mmsize
%if notcpuflag(ssse3)
    UNSCRATCH                3,  9, rsp+1*mmsize
%endif
%endif
    mova [dst8q+stride8q*0+ 0], m4
    mova [dst8q+stride8q*0+16], m3
    mova [dst8q+stride8q*0+32], m1
    mova [dst8q+stride8q*0+48], m0
    mova [dst8q+stride8q*1+ 0], m5
    mova [dst8q+stride8q*1+16], m4
    mova [dst8q+stride8q*1+32], m3
    mova [dst8q+stride8q*1+48], m1
    mova [dst8q+stride8q*2+ 0], m6
    mova [dst8q+stride8q*2+16], m5
    mova [dst8q+stride8q*2+32], m4
    mova [dst8q+stride8q*2+48], m3
    mova [dst8q+stride24q + 0], m7
    mova [dst8q+stride24q +16], m6
    mova [dst8q+stride24q +32], m5
    mova [dst8q+stride24q +48], m4
%if cpuflag(avx)
    vpalignr                m7, m6, m7, 2
    vpalignr                m6, m5, m6, 2
    vpalignr                m5, m4, m5, 2
    vpalignr                m4, m3, m4, 2
    vpalignr                m3, m1, m3, 2
    vpalignr                m1, m0, m1, 2
    vpalignr                m0, m2, m0, 2
%else
    SCRATCH                  2,  8, rsp+0*mmsize
%if notcpuflag(ssse3)
    SCRATCH                  0,  9, rsp+1*mmsize
%endif
    PALIGNR                 m2, m6, m7, 2, m0
    mova                    m7, m2
    PALIGNR                 m2, m5, m6, 2, m0
    mova                    m6, m2
    PALIGNR                 m2, m4, m5, 2, m0
    mova                    m5, m2
    PALIGNR                 m2, m3, m4, 2, m0
    mova                    m4, m2
    PALIGNR                 m2, m1, m3, 2, m0
    mova                    m3, m2
%if notcpuflag(ssse3)
    UNSCRATCH                0,  9, rsp+1*mmsize
    SCRATCH                  3,  9, rsp+1*mmsize
%endif
    PALIGNR                 m2, m0, m1, 2, m3
    mova                    m1, m2
    UNSCRATCH                2,  8, rsp+0*mmsize
    SCRATCH                  1,  8, rsp+0*mmsize
    PALIGNR                 m1, m2, m0, 2, m3
    mova                    m0, m1
%endif
    psrldq                  m2, 2
    dec                   cntd
    jg .loop
    RET
%endmacro

INIT_XMM sse2
DR_FUNCS 3
INIT_XMM ssse3
DR_FUNCS 2
INIT_XMM avx
DR_FUNCS 2

%macro VL_FUNCS 1 ; stack_mem_for_32x32_32bit_function
cglobal vp9_ipred_vl_4x4_16, 2, 4, 3, dst, stride, l, a
    movifnidn               aq, amp
    movu                    m0, [aq]                ; abcdefgh
    psrldq                  m1, m0, 2               ; bcdefgh.
    psrldq                  m2, m0, 4               ; cdefgh..
    LOWPASS                  2,  1, 0               ; BCDEFGH.
    pavgw                   m1, m0                  ; ABCDEFG.
    DEFINE_ARGS dst, stride, stride3
    lea               stride3q, [strideq*3]

    movh      [dstq+strideq*0], m1
    movh      [dstq+strideq*1], m2
    psrldq                  m1, 2
    psrldq                  m2, 2
    movh      [dstq+strideq*2], m1
    movh      [dstq+stride3q ], m2
    RET

cglobal vp9_ipred_vl_8x8_16, 2, 4, 4, dst, stride, l, a
    movifnidn               aq, amp
    mova                    m0, [aq]                ; abcdefgh
%if cpuflag(ssse3)
    mova                    m3, [pb_2to15_14_15]
%endif
    SHIFT_RIGHTx2           m1, m2, m0, m3          ; bcdefghh/cdefghhh
    LOWPASS                  2,  1, 0               ; BCDEFGHh
    pavgw                   m1, m0                  ; ABCDEFGh
    DEFINE_ARGS dst, stride, stride3
    lea               stride3q, [strideq*3]

    mova      [dstq+strideq*0], m1
    mova      [dstq+strideq*1], m2
    SHIFT_RIGHT             m1, m1, m3
    SHIFT_RIGHT             m2, m2, m3
    mova      [dstq+strideq*2], m1
    mova      [dstq+stride3q ], m2
    lea                   dstq, [dstq+strideq*4]
    SHIFT_RIGHT             m1, m1, m3
    SHIFT_RIGHT             m2, m2, m3
    mova      [dstq+strideq*0], m1
    mova      [dstq+strideq*1], m2
    SHIFT_RIGHT             m1, m1, m3
    SHIFT_RIGHT             m2, m2, m3
    mova      [dstq+strideq*2], m1
    mova      [dstq+stride3q ], m2
    RET

cglobal vp9_ipred_vl_16x16_16, 2, 4, 6, dst, stride, l, a
    movifnidn               aq, amp
    mova                    m0, [aq]
    mova                    m1, [aq+mmsize]
    PALIGNR                 m2, m1, m0, 2, m3
    PALIGNR                 m3, m1, m0, 4, m4
    LOWPASS                  3,  2,  0
    pavgw                   m2, m0
%if cpuflag(ssse3)
    mova                    m4, [pb_2to15_14_15]
%endif
    SHIFT_RIGHTx2           m5, m0, m1, m4
    LOWPASS                  0,  5,  1
    pavgw                   m1, m5
    DEFINE_ARGS dst, stride, cnt
    mov                   cntd, 8

.loop:
    mova   [dstq+strideq*0+ 0], m2
    mova   [dstq+strideq*0+16], m1
    mova   [dstq+strideq*1+ 0], m3
    mova   [dstq+strideq*1+16], m0
    lea                   dstq, [dstq+strideq*2]
%if cpuflag(avx)
    vpalignr                m2, m1, m2, 2
    vpalignr                m3, m0, m3, 2
%else
    PALIGNR                 m5, m1, m2, 2, m4
    mova                    m2, m5
    PALIGNR                 m5, m0, m3, 2, m4
    mova                    m3, m5
%endif
    SHIFT_RIGHT             m1, m1, m4
    SHIFT_RIGHT             m0, m0, m4
    dec                   cntd
    jg .loop
    RET

cglobal vp9_ipred_vl_32x32_16, 2, 5, 11, %1 * mmsize * ARCH_X86_32, dst, stride, l, a
    movifnidn               aq, amp
    mova                    m0, [aq+mmsize*0]
    mova                    m1, [aq+mmsize*1]
    mova                    m2, [aq+mmsize*2]
    PALIGNR                 m6, m1, m0, 2, m5
    PALIGNR                 m7, m1, m0, 4, m5
    LOWPASS                  7,  6,  0
    pavgw                   m6, m0
    SCRATCH                  6,  8, rsp+0*mmsize
    PALIGNR                 m4, m2, m1, 2, m0
    PALIGNR                 m5, m2, m1, 4, m0
    LOWPASS                  5,  4,  1
    pavgw                   m4, m1
    mova                    m0, [aq+mmsize*3]
    PALIGNR                 m1, m0, m2, 2, m6
    PALIGNR                 m3, m0, m2, 4, m6
    LOWPASS                  3,  1,  2
    pavgw                   m2, m1
%if cpuflag(ssse3)
    PRELOAD                 10, pb_2to15_14_15, shuf
%endif
    SHIFT_RIGHTx2           m6, m1, m0, reg_shuf
    LOWPASS                  1,  6,  0
    pavgw                   m0, m6
%if ARCH_X86_64
    pshufd                  m9, m6, q3333
%endif
%if cpuflag(avx)
    UNSCRATCH                6,  8, rsp+0*mmsize
%endif
    DEFINE_ARGS dst, stride, cnt, stride16, stride17
    mov              stride16q, strideq
    mov                   cntd, 8
    shl              stride16q, 4
    lea              stride17q, [stride16q+strideq]

    ; FIXME m8 is unused for avx, so we could save one register here for win64
.loop:
%if notcpuflag(avx)
    UNSCRATCH                6,  8, rsp+0*mmsize
%endif
    mova   [dstq+strideq*0+ 0], m6
    mova   [dstq+strideq*0+16], m4
    mova   [dstq+strideq*0+32], m2
    mova   [dstq+strideq*0+48], m0
    mova   [dstq+strideq*1+ 0], m7
    mova   [dstq+strideq*1+16], m5
    mova   [dstq+strideq*1+32], m3
    mova   [dstq+strideq*1+48], m1
    mova   [dstq+stride16q+ 0], m4
    mova   [dstq+stride16q+16], m2
    mova   [dstq+stride16q+32], m0
%if ARCH_X86_64
    mova   [dstq+stride16q+48], m9
%endif
    mova   [dstq+stride17q+ 0], m5
    mova   [dstq+stride17q+16], m3
    mova   [dstq+stride17q+32], m1
%if ARCH_X86_64
    mova   [dstq+stride17q+48], m9
%endif
    lea                   dstq, [dstq+strideq*2]
%if cpuflag(avx)
    vpalignr                m6, m4, m6, 2
    vpalignr                m4, m2, m4, 2
    vpalignr                m2, m0, m2, 2
    vpalignr                m7, m5, m7, 2
    vpalignr                m5, m3, m5, 2
    vpalignr                m3, m1, m3, 2
%else
    SCRATCH                  3,  8, rsp+0*mmsize
%if notcpuflag(ssse3)
    SCRATCH                  1, 10, rsp+1*mmsize
%endif
    PALIGNR                 m3, m4, m6, 2, m1
    mova                    m6, m3
    PALIGNR                 m3, m2, m4, 2, m1
    mova                    m4, m3
    PALIGNR                 m3, m0, m2, 2, m1
    mova                    m2, m3
    PALIGNR                 m3, m5, m7, 2, m1
    mova                    m7, m3
    UNSCRATCH                3,  8, rsp+0*mmsize
    SCRATCH                  6,  8, rsp+0*mmsize
%if notcpuflag(ssse3)
    UNSCRATCH                1, 10, rsp+1*mmsize
    SCRATCH                  7, 10, rsp+1*mmsize
%endif
    PALIGNR                 m6, m3, m5, 2, m7
    mova                    m5, m6
    PALIGNR                 m6, m1, m3, 2, m7
    mova                    m3, m6
%if notcpuflag(ssse3)
    UNSCRATCH                7, 10, rsp+1*mmsize
%endif
%endif
    SHIFT_RIGHT             m1, m1, reg_shuf
    SHIFT_RIGHT             m0, m0, reg_shuf
    dec                   cntd
    jg .loop

%if ARCH_X86_32
    DEFINE_ARGS dst, stride, stride3
    lea               stride3q, [strideq*3]
%assign %%n 0
%rep 4
    mova   [dstq+strideq*0+48], m0
    mova   [dstq+strideq*1+48], m0
    mova   [dstq+strideq*2+48], m0
    mova   [dstq+stride3q +48], m0
%if %%n < 3
    lea                   dstq, [dstq+strideq*4]
%endif
%assign %%n (%%n+1)
%endrep
%endif
    RET
%endmacro

INIT_XMM sse2
VL_FUNCS 2
INIT_XMM ssse3
VL_FUNCS 1
INIT_XMM avx
VL_FUNCS 1

%macro VR_FUNCS 0
cglobal vp9_ipred_vr_4x4_16, 4, 4, 3, dst, stride, l, a
    movu                    m0, [aq-2]
    movhps                  m1, [lq]
    PALIGNR                 m0, m1, 10, m2          ; xyz*abcd
    pslldq                  m1, m0, 2               ; .xyz*abc
    pslldq                  m2, m0, 4               ; ..xyz*ab
    LOWPASS                  2,  1, 0               ; ..YZ#ABC
    pavgw                   m1, m0                  ; ....#ABC
    DEFINE_ARGS dst, stride, stride3
    lea               stride3q, [strideq*3]

    movhps    [dstq+strideq*0], m1
    movhps    [dstq+strideq*1], m2
    shufps                  m0, m2, m1, q3210
%if cpuflag(ssse3)
    pshufb                  m2, [pb_4_5_8to13_8x0]
%else
    pshuflw                 m2, m2, q2222
    psrldq                  m2, 6
%endif
    psrldq                  m0, 6
    movh      [dstq+strideq*2], m0
    movh      [dstq+stride3q ], m2
    RET

cglobal vp9_ipred_vr_8x8_16, 4, 4, 5, dst, stride, l, a
    movu                    m1, [aq-2]              ; *abcdefg
    movu                    m2, [lq]                ; stuvwxyz
    mova                    m0, [aq]                ; abcdefgh
    PALIGNR                 m3, m1, m2, 14, m4      ; z*abcdef
    LOWPASS                  3,  1,  0
    pavgw                   m0, m1
    PALIGNR                 m1, m2,  2, m4          ; tuvwxyz*
    pslldq                  m4, m2,  2              ; .stuvwxy
    LOWPASS                  4,  2,  1
    DEFINE_ARGS dst, stride, stride3
    lea               stride3q, [strideq*3]

    mova      [dstq+strideq*0], m0
    mova      [dstq+strideq*1], m3
    PALIGNR                 m0, m4, 14, m1
    pslldq                  m4, 2
    PALIGNR                 m3, m4, 14, m1
    pslldq                  m4, 2
    mova      [dstq+strideq*2], m0
    mova      [dstq+stride3q ], m3
    lea                   dstq, [dstq+strideq*4]
    PALIGNR                 m0, m4, 14, m1
    pslldq                  m4, 2
    PALIGNR                 m3, m4, 14, m1
    pslldq                  m4, 2
    mova      [dstq+strideq*0], m0
    mova      [dstq+strideq*1], m3
    PALIGNR                 m0, m4, 14, m1
    pslldq                  m4, 2
    PALIGNR                 m3, m4, 14, m4
    mova      [dstq+strideq*2], m0
    mova      [dstq+stride3q ], m3
    RET

cglobal vp9_ipred_vr_16x16_16, 4, 4, 8, dst, stride, l, a
    movu                    m1, [aq-2]              ; *abcdefg
    movu                    m2, [aq+mmsize-2]       ; hijklmno
    mova                    m3, [aq]                ; abcdefgh
    mova                    m4, [aq+mmsize]         ; ijklmnop
    mova                    m5, [lq+mmsize]         ; stuvwxyz
    PALIGNR                 m0, m1, m5, 14, m6      ; z*abcdef
    movu                    m6, [aq+mmsize-4]       ; ghijklmn
    LOWPASS                  6,  2,  4
    pavgw                   m2, m4
    LOWPASS                  0,  1,  3
    pavgw                   m3, m1
    PALIGNR                 m1, m5,  2, m7          ; tuvwxyz*
    movu                    m7, [lq+mmsize-2]       ; rstuvwxy
    LOWPASS                  1,  5,  7
    movu                    m5, [lq+2]              ; lmnopqrs
    pslldq                  m4, m5,  2              ; .lmnopqr
    pslldq                  m7, m5,  4              ; ..lmnopq
    LOWPASS                  5,  4,  7
    psrld                   m4, m1, 16
    psrld                   m7, m5, 16
    pand                    m1, [pd_65535]
    pand                    m5, [pd_65535]
    packssdw                m7, m4
    packssdw                m5, m1
    DEFINE_ARGS dst, stride, cnt
    mov                   cntd, 8

.loop:
    mova   [dstq+strideq*0+ 0], m3
    mova   [dstq+strideq*0+16], m2
    mova   [dstq+strideq*1+ 0], m0
    mova   [dstq+strideq*1+16], m6
    lea                   dstq, [dstq+strideq*2]
    PALIGNR                 m2, m3, 14, m4
    PALIGNR                 m3, m7, 14, m4
    pslldq                  m7, 2
    PALIGNR                 m6, m0, 14, m4
    PALIGNR                 m0, m5, 14, m4
    pslldq                  m5, 2
    dec                   cntd
    jg .loop
    RET

cglobal vp9_ipred_vr_32x32_16, 4, 5, 14, 6 * mmsize * ARCH_X86_32, dst, stride, l, a
    movu                    m0, [aq+mmsize*0-2]     ; *a[0-6]
    movu                    m1, [aq+mmsize*1-2]     ; a[7-14]
    movu                    m2, [aq+mmsize*2-2]     ; a[15-22]
    movu                    m3, [aq+mmsize*3-2]     ; a[23-30]
    mova                    m4, [aq+mmsize*3+0]     ; a[24-31]
    movu                    m5, [aq+mmsize*3-4]     ; a[22-29]
    LOWPASS                  5,  3,  4              ; A[23-30]
    SCRATCH                  5,  8, rsp+0*mmsize
    pavgw                   m3, m4
    mova                    m4, [aq+mmsize*2+0]     ; a[16-23]
    movu                    m6, [aq+mmsize*2-4]     ; a[14-21]
    LOWPASS                  6,  2,  4              ; A[15-22]
    SCRATCH                  6,  9, rsp+1*mmsize
    pavgw                   m2, m4
    mova                    m4, [aq+mmsize*1+0]     ; a[8-15]
    movu                    m7, [aq+mmsize*1-4]     ; a[6-13]
    LOWPASS                  7,  1,  4              ; A[7-14]
    SCRATCH                  7, 10, rsp+2*mmsize
    pavgw                   m1, m4
    mova                    m4, [aq+mmsize*0+0]     ; a[0-7]
    mova                    m5, [lq+mmsize*3+0]     ; l[24-31]
    PALIGNR                 m6, m0, m5, 14, m7      ; l[31]*a[0-5]
    LOWPASS                  6,  0,  4              ; #A[0-6]
    SCRATCH                  6, 11, rsp+3*mmsize
    pavgw                   m4, m0
    PALIGNR                 m0, m5,  2, m7          ; l[25-31]*
    movu                    m7, [lq+mmsize*3-2]     ; l[23-30]
    LOWPASS                  0,  5,  7              ; L[24-31]
    movu                    m5, [lq+mmsize*2-2]     ; l[15-22]
    mova                    m7, [lq+mmsize*2+0]     ; l[16-23]
    movu                    m6, [lq+mmsize*2+2]     ; l[17-24]
    LOWPASS                  5,  7,  6              ; L[16-23]
    psrld                   m7, m0, 16
    psrld                   m6, m5, 16
    pand                    m0, [pd_65535]
    pand                    m5, [pd_65535]
    packssdw                m6, m7
    packssdw                m5, m0
    SCRATCH                  5, 12, rsp+4*mmsize
    SCRATCH                  6, 13, rsp+5*mmsize
    movu                    m6, [lq+mmsize*1-2]     ; l[7-14]
    mova                    m0, [lq+mmsize*1+0]     ; l[8-15]
    movu                    m5, [lq+mmsize*1+2]     ; l[9-16]
    LOWPASS                  6,  0,  5              ; L[8-15]
    movu                    m0, [lq+mmsize*0+2]     ; l[1-8]
    pslldq                  m5, m0,  2              ; .l[1-7]
    pslldq                  m7, m0,  4              ; ..l[1-6]
    LOWPASS                  0,  5,  7
    psrld                   m5, m6, 16
    psrld                   m7, m0, 16
    pand                    m6, [pd_65535]
    pand                    m0, [pd_65535]
    packssdw                m7, m5
    packssdw                m0, m6
    UNSCRATCH                6, 13, rsp+5*mmsize
    DEFINE_ARGS dst, stride, stride16, cnt, stride17
    mov              stride16q, strideq
    mov                   cntd, 8
    shl              stride16q, 4
%if ARCH_X86_64
    lea              stride17q, [stride16q+strideq]
%endif

.loop:
    mova   [dstq+strideq*0+ 0], m4
    mova   [dstq+strideq*0+16], m1
    mova   [dstq+strideq*0+32], m2
    mova   [dstq+strideq*0+48], m3
%if ARCH_X86_64
    mova   [dstq+strideq*1+ 0], m11
    mova   [dstq+strideq*1+16], m10
    mova   [dstq+strideq*1+32], m9
    mova   [dstq+strideq*1+48], m8
%endif
    mova   [dstq+stride16q+ 0], m6
    mova   [dstq+stride16q+16], m4
    mova   [dstq+stride16q+32], m1
    mova   [dstq+stride16q+48], m2
%if ARCH_X86_64
    mova   [dstq+stride17q+ 0], m12
    mova   [dstq+stride17q+16], m11
    mova   [dstq+stride17q+32], m10
    mova   [dstq+stride17q+48], m9
%endif
    lea                   dstq, [dstq+strideq*2]
    PALIGNR                 m3, m2,  14, m5
    PALIGNR                 m2, m1,  14, m5
    PALIGNR                 m1, m4,  14, m5
    PALIGNR                 m4, m6,  14, m5
    PALIGNR                 m6, m7,  14, m5
    pslldq                  m7, 2
%if ARCH_X86_64
    PALIGNR                 m8, m9,  14, m5
    PALIGNR                 m9, m10, 14, m5
    PALIGNR                m10, m11, 14, m5
    PALIGNR                m11, m12, 14, m5
    PALIGNR                m12, m0,  14, m5
    pslldq                  m0, 2
%endif
    dec                   cntd
    jg .loop

%if ARCH_X86_32
    UNSCRATCH                5, 12, rsp+4*mmsize
    UNSCRATCH                4, 11, rsp+3*mmsize
    UNSCRATCH                3, 10, rsp+2*mmsize
    UNSCRATCH                2,  9, rsp+1*mmsize
    UNSCRATCH                1,  8, rsp+0*mmsize
    mov                   dstq, dstm
    mov                   cntd, 8
    add                   dstq, strideq
.loop2:
    mova   [dstq+strideq*0+ 0], m4
    mova   [dstq+strideq*0+16], m3
    mova   [dstq+strideq*0+32], m2
    mova   [dstq+strideq*0+48], m1
    mova   [dstq+stride16q+ 0], m5
    mova   [dstq+stride16q+16], m4
    mova   [dstq+stride16q+32], m3
    mova   [dstq+stride16q+48], m2
    lea                   dstq, [dstq+strideq*2]
    PALIGNR                 m1, m2,  14, m6
    PALIGNR                 m2, m3,  14, m6
    PALIGNR                 m3, m4,  14, m6
    PALIGNR                 m4, m5,  14, m6
    PALIGNR                 m5, m0,  14, m6
    pslldq                  m0, 2
    dec                   cntd
    jg .loop2
%endif
    RET
%endmacro

INIT_XMM sse2
VR_FUNCS
INIT_XMM ssse3
VR_FUNCS
INIT_XMM avx
VR_FUNCS

%macro HU_FUNCS 1 ; stack_mem_for_32x32_32bit_function
cglobal vp9_ipred_hu_4x4_16, 3, 3, 3, dst, stride, l, a
    movh                    m0, [lq]                ; abcd
%if cpuflag(ssse3)
    pshufb                  m0, [pb_0to7_67x4]      ; abcddddd
%else
    punpcklqdq              m0, m0
    pshufhw                 m0, m0, q3333           ; abcddddd
%endif
    psrldq                  m1, m0,  2              ; bcddddd.
    psrldq                  m2, m0,  4              ; cddddd..
    LOWPASS                  2,  1,  0              ; BCDddd..
    pavgw                   m1, m0                  ; abcddddd
    SBUTTERFLY          wd,  1,  2,  0              ; aBbCcDdd, dddddddd
    PALIGNR                 m2, m1,  4, m0          ; bCcDdddd
    DEFINE_ARGS dst, stride, stride3
    lea               stride3q, [strideq*3]

    movh      [dstq+strideq*0], m1                  ; aBbC
    movh      [dstq+strideq*1], m2                  ; bCcD
    movhps    [dstq+strideq*2], m1                  ; cDdd
    movhps    [dstq+stride3q ], m2                  ; dddd
    RET

cglobal vp9_ipred_hu_8x8_16, 3, 3, 4, dst, stride, l, a
    mova                    m0, [lq]
%if cpuflag(ssse3)
    mova                    m3, [pb_2to15_14_15]
%endif
    SHIFT_RIGHTx2           m1, m2, m0, m3
    LOWPASS                  2,  1,  0
    pavgw                   m1, m0
    SBUTTERFLY          wd,  1,  2,  0
    shufps                  m0, m1, m2, q1032
    pshufd                  m3, m2, q3332
    DEFINE_ARGS dst, stride, stride3
    lea               stride3q, [strideq*3]

    mova     [dstq+strideq *0], m1
    mova     [dstq+strideq *2], m0
    mova     [dstq+strideq *4], m2
    mova     [dstq+stride3q*2], m3
    add                   dstq, strideq
%if cpuflag(avx)
    vpalignr                m1, m2, m1, 4
%else
    PALIGNR                 m0, m2, m1, 4, m3
    mova                    m1, m0
%endif
    pshufd                  m2, m2, q3321
    shufps                  m0, m1, m2, q1032
    pshufd                  m3, m2, q3332
    mova     [dstq+strideq *0], m1
    mova     [dstq+strideq *2], m0
    mova     [dstq+strideq *4], m2
    mova     [dstq+stride3q*2], m3
    RET

cglobal vp9_ipred_hu_16x16_16, 3, 4, 6 + notcpuflag(ssse3), dst, stride, l, a
    mova                    m0, [lq]
    mova                    m3, [lq+mmsize]
    movu                    m1, [lq+2]
    movu                    m2, [lq+4]
    LOWPASS                  2,  1,  0
    pavgw                   m1, m0
    SBUTTERFLY           wd, 1,  2,  0
%if cpuflag(ssse3)
    mova                    m5, [pb_2to15_14_15]
%endif
    SHIFT_RIGHTx2           m0, m4, m3, m5
    LOWPASS                  4,  0,  3
    pavgw                   m3, m0
    SBUTTERFLY           wd, 3,  4,  5
    pshufd                  m0, m0, q3333
    DEFINE_ARGS dst, stride, stride3, cnt
    lea               stride3q, [strideq*3]
    mov                   cntd, 4

.loop:
    mova  [dstq+strideq *0+ 0], m1
    mova  [dstq+strideq *0+16], m2
    mova  [dstq+strideq *4+ 0], m2
    mova  [dstq+strideq *4+16], m3
    mova  [dstq+strideq *8+ 0], m3
    mova  [dstq+strideq *8+16], m4
    mova  [dstq+stride3q*4+ 0], m4
    mova  [dstq+stride3q*4+16], m0
    add                   dstq, strideq
%if cpuflag(avx)
    vpalignr                m1, m2, m1, 4
    vpalignr                m2, m3, m2, 4
    vpalignr                m3, m4, m3, 4
    vpalignr                m4, m0, m4, 4
%else
    PALIGNR                 m5, m2, m1, 4, m6
    mova                    m1, m5
    PALIGNR                 m5, m3, m2, 4, m6
    mova                    m2, m5
    PALIGNR                 m5, m4, m3, 4, m6
    mova                    m3, m5
    PALIGNR                 m5, m0, m4, 4, m6
    mova                    m4, m5
%endif
    dec                   cntd
    jg .loop
    RET

cglobal vp9_ipred_hu_32x32_16, 3, 7, 10 + notcpuflag(ssse3), \
1638
                               %1 * -mmsize * ARCH_X86_32, dst, stride, l, a
1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894
    mova                    m2, [lq+mmsize*0+0]
    movu                    m1, [lq+mmsize*0+2]
    movu                    m0, [lq+mmsize*0+4]
    LOWPASS                  0,  1,  2
    pavgw                   m1, m2
    SBUTTERFLY           wd, 1,  0,  2
    SCRATCH                  1,  8, rsp+0*mmsize
    mova                    m4, [lq+mmsize*1+0]
    movu                    m3, [lq+mmsize*1+2]
    movu                    m2, [lq+mmsize*1+4]
    LOWPASS                  2,  3,  4
    pavgw                   m3, m4
    SBUTTERFLY           wd, 3,  2,  4
    mova                    m6, [lq+mmsize*2+0]
    movu                    m5, [lq+mmsize*2+2]
    movu                    m4, [lq+mmsize*2+4]
    LOWPASS                  4,  5,  6
    pavgw                   m5, m6
    SBUTTERFLY           wd, 5,  4,  6
    mova                    m7, [lq+mmsize*3+0]
    SCRATCH                  0,  9, rsp+1*mmsize
%if cpuflag(ssse3)
    mova                    m0, [pb_2to15_14_15]
%endif
    SHIFT_RIGHTx2           m1, m6, m7, m0
    LOWPASS                  6,  1,  7
    pavgw                   m7, m1
    SBUTTERFLY           wd, 7,  6,  0
    pshufd                  m1, m1, q3333
    UNSCRATCH                0,  9, rsp+1*mmsize
    DEFINE_ARGS dst, stride, cnt, stride3, stride4, stride20, stride28
    lea               stride3q, [strideq*3]
    lea               stride4q, [strideq*4]
    lea              stride28q, [stride4q*8]
    lea              stride20q, [stride4q*5]
    sub              stride28q, stride4q
    mov                   cntd, 4

.loop:
%if ARCH_X86_64
    SWAP                     1,  8
%else
    mova        [rsp+1*mmsize], m1
    mova                    m1, [rsp+0*mmsize]
%endif
    mova  [dstq+strideq *0+ 0], m1
    mova  [dstq+strideq *0+16], m0
    mova  [dstq+strideq *0+32], m3
    mova  [dstq+strideq *0+48], m2
    mova  [dstq+stride4q*1+ 0], m0
    mova  [dstq+stride4q*1+16], m3
    mova  [dstq+stride4q*1+32], m2
    mova  [dstq+stride4q*1+48], m5
    mova  [dstq+stride4q*2+ 0], m3
    mova  [dstq+stride4q*2+16], m2
    mova  [dstq+stride4q*2+32], m5
    mova  [dstq+stride4q*2+48], m4
%if cpuflag(avx)
    vpalignr                m1, m0, m1, 4
    vpalignr                m0, m3, m0, 4
    vpalignr                m3, m2, m3, 4
%else
    SCRATCH                  6,  9, rsp+2*mmsize
%if notcpuflag(ssse3)
    SCRATCH                  7, 10, rsp+3*mmsize
%endif
    PALIGNR                 m6, m0, m1, 4, m7
    mova                    m1, m6
    PALIGNR                 m6, m3, m0, 4, m7
    mova                    m0, m6
    PALIGNR                 m6, m2, m3, 4, m7
    mova                    m3, m6
    UNSCRATCH                6,  9, rsp+2*mmsize
    SCRATCH                  0,  9, rsp+2*mmsize
%if notcpuflag(ssse3)
    UNSCRATCH                7, 10, rsp+3*mmsize
    SCRATCH                  3, 10, rsp+3*mmsize
%endif
%endif
%if ARCH_X86_64
    SWAP                     1,  8
%else
    mova        [rsp+0*mmsize], m1
    mova                    m1, [rsp+1*mmsize]
%endif
    mova  [dstq+stride3q*4+ 0], m2
    mova  [dstq+stride3q*4+16], m5
    mova  [dstq+stride3q*4+32], m4
    mova  [dstq+stride3q*4+48], m7
    mova  [dstq+stride4q*4+ 0], m5
    mova  [dstq+stride4q*4+16], m4
    mova  [dstq+stride4q*4+32], m7
    mova  [dstq+stride4q*4+48], m6
    mova  [dstq+stride20q + 0], m4
    mova  [dstq+stride20q +16], m7
    mova  [dstq+stride20q +32], m6
    mova  [dstq+stride20q +48], m1
    mova  [dstq+stride3q*8+ 0], m7
    mova  [dstq+stride3q*8+16], m6
    mova  [dstq+stride3q*8+32], m1
    mova  [dstq+stride3q*8+48], m1
    mova  [dstq+stride28q + 0], m6
    mova  [dstq+stride28q +16], m1
    mova  [dstq+stride28q +32], m1
    mova  [dstq+stride28q +48], m1
%if cpuflag(avx)
    vpalignr                m2, m5, m2, 4
    vpalignr                m5, m4, m5, 4
    vpalignr                m4, m7, m4, 4
    vpalignr                m7, m6, m7, 4
    vpalignr                m6, m1, m6, 4
%else
    PALIGNR                 m0, m5, m2, 4, m3
    mova                    m2, m0
    PALIGNR                 m0, m4, m5, 4, m3
    mova                    m5, m0
    PALIGNR                 m0, m7, m4, 4, m3
    mova                    m4, m0
    PALIGNR                 m0, m6, m7, 4, m3
    mova                    m7, m0
    PALIGNR                 m0, m1, m6, 4, m3
    mova                    m6, m0
    UNSCRATCH                0,  9, rsp+2*mmsize
%if notcpuflag(ssse3)
    UNSCRATCH                3, 10, rsp+3*mmsize
%endif
%endif
    add                   dstq, strideq
    dec                   cntd
    jg .loop
    RET
%endmacro

INIT_XMM sse2
HU_FUNCS 4
INIT_XMM ssse3
HU_FUNCS 3
INIT_XMM avx
HU_FUNCS 2

%macro HD_FUNCS 0
cglobal vp9_ipred_hd_4x4_16, 4, 4, 4, dst, stride, l, a
    movh                    m0, [lq]
    movhps                  m0, [aq-2]
    psrldq                  m1, m0, 2
    psrldq                  m2, m0, 4
    LOWPASS                  2,  1,  0
    pavgw                   m1, m0
    punpcklwd               m1, m2
    DEFINE_ARGS dst, stride, stride3
    lea               stride3q, [strideq*3]

    movh      [dstq+stride3q ], m1
    movhps    [dstq+strideq*1], m1
    movhlps                 m2, m2
    PALIGNR                 m2, m1, 4, m0
    movh      [dstq+strideq*2], m2
    movhps    [dstq+strideq*0], m2
    RET

cglobal vp9_ipred_hd_8x8_16, 4, 4, 5, dst, stride, l, a
    mova                    m0, [lq]
    movu                    m1, [aq-2]
    PALIGNR                 m2, m1, m0, 2, m3
    PALIGNR                 m3, m1, m0, 4, m4
    LOWPASS                  3,  2,  0
    pavgw                   m2, m0
    SBUTTERFLY           wd, 2,  3,  0
    psrldq                  m0, m1,  2
    psrldq                  m4, m1,  4
    LOWPASS                  1,  0,  4
    DEFINE_ARGS dst8, mstride, cnt
    lea                  dst8q, [dst8q+mstrideq*8]
    neg               mstrideq
    mov                   cntd, 4

.loop:
    add                  dst8q, mstrideq
    mova    [dst8q+mstrideq*0], m2
    mova    [dst8q+mstrideq*4], m3
%if cpuflag(avx)
    vpalignr                m2, m3, m2, 4
    vpalignr                m3, m1, m3, 4
%else
    PALIGNR                 m0, m3, m2, 4, m4
    mova                    m2, m0
    PALIGNR                 m0, m1, m3, 4, m4
    mova                    m3, m0
%endif
    psrldq                  m1, 4
    dec                   cntd
    jg .loop
    RET

cglobal vp9_ipred_hd_16x16_16, 4, 4, 8, dst, stride, l, a
    mova                    m2, [lq]
    movu                    m1, [lq+2]
    movu                    m0, [lq+4]
    LOWPASS                  0,  1,  2
    pavgw                   m1, m2
    mova                    m4, [lq+mmsize]
    movu                    m5, [aq-2]
    PALIGNR                 m3, m5, m4, 2, m6
    PALIGNR                 m2, m5, m4, 4, m6
    LOWPASS                  2,  3,  4
    pavgw                   m3, m4
    SBUTTERFLY           wd, 1,  0,  4
    SBUTTERFLY           wd, 3,  2,  4
    mova                    m6, [aq]
    movu                    m4, [aq+2]
    LOWPASS                  4,  6,  5
    movu                    m5, [aq+mmsize-2]
    psrldq                  m6, m5,  2
    psrldq                  m7, m5,  4
    LOWPASS                  5,  6,  7
    DEFINE_ARGS dst, mstride, mstride3, cnt
    lea                   dstq, [dstq+mstrideq*8]
    lea                   dstq, [dstq+mstrideq*8]
    neg               mstrideq
    lea              mstride3q, [mstrideq*3]
    mov                   cntd, 4

.loop:
    add                  dstq, mstrideq
    mova [dstq+mstride3q*4+ 0], m2
    mova [dstq+mstride3q*4+16], m4
    mova [dstq+mstrideq *8+ 0], m3
    mova [dstq+mstrideq *8+16], m2
    mova [dstq+mstrideq *4+ 0], m0
    mova [dstq+mstrideq *4+16], m3
    mova [dstq+mstrideq *0+ 0], m1
    mova [dstq+mstrideq *0+16], m0
%if cpuflag(avx)
    vpalignr                m1, m0, m1, 4
    vpalignr                m0, m3, m0, 4
    vpalignr                m3, m2, m3, 4
    vpalignr                m2, m4, m2, 4
    vpalignr                m4, m5, m4, 4
%else
    PALIGNR                 m6, m0, m1, 4, m7
    mova                    m1, m6
    PALIGNR                 m6, m3, m0, 4, m7
    mova                    m0, m6
    PALIGNR                 m6, m2, m3, 4, m7
    mova                    m3, m6
    PALIGNR                 m6, m4, m2, 4, m7
    mova                    m2, m6
    PALIGNR                 m6, m5, m4, 4, m7
    mova                    m4, m6
%endif
    psrldq                  m5, 4
    dec                   cntd
    jg .loop
    RET

cglobal vp9_ipred_hd_32x32_16, 4, 4 + 3 * ARCH_X86_64, 14, \
1895
                               10 * -mmsize * ARCH_X86_32, dst, stride, l, a
1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135
    mova                    m2, [lq+mmsize*0+0]
    movu                    m1, [lq+mmsize*0+2]
    movu                    m0, [lq+mmsize*0+4]
    LOWPASS                  0,  1,  2
    pavgw                   m1, m2
    SBUTTERFLY           wd, 1,  0,  2
    mova                    m4, [lq+mmsize*1+0]
    movu                    m3, [lq+mmsize*1+2]
    movu                    m2, [lq+mmsize*1+4]
    LOWPASS                  2,  3,  4
    pavgw                   m3, m4
    SBUTTERFLY           wd, 3,  2,  4
    SCRATCH                  0,  8, rsp+0*mmsize
    SCRATCH                  1,  9, rsp+1*mmsize
    SCRATCH                  2, 10, rsp+2*mmsize
    SCRATCH                  3, 11, rsp+3*mmsize
    mova                    m6, [lq+mmsize*2+0]
    movu                    m5, [lq+mmsize*2+2]
    movu                    m4, [lq+mmsize*2+4]
    LOWPASS                  4,  5,  6
    pavgw                   m5, m6
    SBUTTERFLY           wd, 5,  4,  6
    mova                    m0, [lq+mmsize*3+0]
    movu                    m1, [aq+mmsize*0-2]
    PALIGNR                 m7, m1, m0, 2, m2
    PALIGNR                 m6, m1, m0, 4, m2
    LOWPASS                  6,  7,  0
    pavgw                   m7, m0
    SBUTTERFLY           wd, 7,  6,  0
    mova                    m2, [aq+mmsize*0+0]
    movu                    m0, [aq+mmsize*0+2]
    LOWPASS                  0,  2,  1
    movu                    m1, [aq+mmsize*1-2]
    mova                    m2, [aq+mmsize*1+0]
    movu                    m3, [aq+mmsize*1+2]
    LOWPASS                  1,  2,  3
    SCRATCH                  6, 12, rsp+6*mmsize
    SCRATCH                  7, 13, rsp+7*mmsize
    movu                    m2, [aq+mmsize*2-2]
    mova                    m3, [aq+mmsize*2+0]
    movu                    m6, [aq+mmsize*2+2]
    LOWPASS                  2,  3,  6
    movu                    m3, [aq+mmsize*3-2]
    psrldq                  m6, m3,  2
    psrldq                  m7, m3,  4
    LOWPASS                  3,  6,  7
    UNSCRATCH                6, 12, rsp+6*mmsize
    UNSCRATCH                7, 13, rsp+7*mmsize
%if ARCH_X86_32
    mova        [rsp+4*mmsize], m4
    mova        [rsp+5*mmsize], m5
    ; we already backed up m6/m7 earlier on x86-32 in SCRATCH, so we don't need
    ; to do it again here
%endif
    DEFINE_ARGS dst, stride, cnt, stride3, stride4, stride20, stride28
    mov                   cntd, 4
    lea               stride3q, [strideq*3]
%if ARCH_X86_64
    lea               stride4q, [strideq*4]
    lea              stride28q, [stride4q*8]
    lea              stride20q, [stride4q*5]
    sub              stride28q, stride4q
%endif
    add                   dstq, stride3q

    ; x86-32 doesn't have enough registers, so on that platform, we split
    ; the loop in 2... Otherwise you spend most of the loop (un)scratching
.loop:
%if ARCH_X86_64
    mova  [dstq+stride28q + 0], m9
    mova  [dstq+stride28q +16], m8
    mova  [dstq+stride28q +32], m11
    mova  [dstq+stride28q +48], m10
    mova  [dstq+stride3q*8+ 0], m8
    mova  [dstq+stride3q*8+16], m11
    mova  [dstq+stride3q*8+32], m10
    mova  [dstq+stride3q*8+48], m5
    mova  [dstq+stride20q + 0], m11
    mova  [dstq+stride20q +16], m10
    mova  [dstq+stride20q +32], m5
    mova  [dstq+stride20q +48], m4
    mova  [dstq+stride4q*4+ 0], m10
    mova  [dstq+stride4q*4+16], m5
    mova  [dstq+stride4q*4+32], m4
    mova  [dstq+stride4q*4+48], m7
%endif
    mova  [dstq+stride3q*4+ 0], m5
    mova  [dstq+stride3q*4+16], m4
    mova  [dstq+stride3q*4+32], m7
    mova  [dstq+stride3q*4+48], m6
    mova  [dstq+strideq* 8+ 0], m4
    mova  [dstq+strideq* 8+16], m7
    mova  [dstq+strideq* 8+32], m6
    mova  [dstq+strideq* 8+48], m0
    mova  [dstq+strideq* 4+ 0], m7
    mova  [dstq+strideq* 4+16], m6
    mova  [dstq+strideq* 4+32], m0
    mova  [dstq+strideq* 4+48], m1
    mova  [dstq+strideq* 0+ 0], m6
    mova  [dstq+strideq* 0+16], m0
    mova  [dstq+strideq* 0+32], m1
    mova  [dstq+strideq* 0+48], m2
    sub                   dstq, strideq
%if cpuflag(avx)
%if ARCH_X86_64
    vpalignr                m9, m8,  m9,  4
    vpalignr                m8, m11, m8,  4
    vpalignr               m11, m10, m11, 4
    vpalignr               m10, m5,  m10, 4
%endif
    vpalignr                m5, m4,  m5,  4
    vpalignr                m4, m7,  m4,  4
    vpalignr                m7, m6,  m7,  4
    vpalignr                m6, m0,  m6,  4
    vpalignr                m0, m1,  m0,  4
    vpalignr                m1, m2,  m1,  4
    vpalignr                m2, m3,  m2,  4
%else
%if ARCH_X86_64
    PALIGNR                m12, m8,  m9,  4, m13
    mova                    m9, m12
    PALIGNR                m12, m11, m8,  4, m13
    mova                    m8, m12
    PALIGNR                m12, m10, m11, 4, m13
    mova                   m11, m12
    PALIGNR                m12, m5,  m10, 4, m13
    mova                   m10, m12
%endif
    SCRATCH                  3, 12, rsp+8*mmsize, sh
%if notcpuflag(ssse3)
    SCRATCH                  2, 13, rsp+9*mmsize
%endif
    PALIGNR                 m3, m4,  m5,  4, m2
    mova                    m5, m3
    PALIGNR                 m3, m7,  m4,  4, m2
    mova                    m4, m3
    PALIGNR                 m3, m6,  m7,  4, m2
    mova                    m7, m3
    PALIGNR                 m3, m0,  m6,  4, m2
    mova                    m6, m3
    PALIGNR                 m3, m1,  m0,  4, m2
    mova                    m0, m3
%if notcpuflag(ssse3)
    UNSCRATCH                2, 13, rsp+9*mmsize
    SCRATCH                  0, 13, rsp+9*mmsize
%endif
    PALIGNR                 m3, m2,  m1,  4, m0
    mova                    m1, m3
    PALIGNR                 m3, reg_sh,  m2,  4, m0
    mova                    m2, m3
%if notcpuflag(ssse3)
    UNSCRATCH                0, 13, rsp+9*mmsize
%endif
    UNSCRATCH                3, 12, rsp+8*mmsize, sh
%endif
    psrldq                  m3, 4
    dec                   cntd
    jg .loop

%if ARCH_X86_32
    UNSCRATCH                0,  8, rsp+0*mmsize
    UNSCRATCH                1,  9, rsp+1*mmsize
    UNSCRATCH                2, 10, rsp+2*mmsize
    UNSCRATCH                3, 11, rsp+3*mmsize
    mova                    m4, [rsp+4*mmsize]
    mova                    m5, [rsp+5*mmsize]
    mova                    m6, [rsp+6*mmsize]
    mova                    m7, [rsp+7*mmsize]
    DEFINE_ARGS dst, stride, stride5, stride3
    lea               stride5q, [strideq*5]
    lea                   dstq, [dstq+stride5q*4]
    DEFINE_ARGS dst, stride, cnt, stride3
    mov                   cntd, 4
.loop_2:
    mova  [dstq+stride3q*4+ 0], m1
    mova  [dstq+stride3q*4+16], m0
    mova  [dstq+stride3q*4+32], m3
    mova  [dstq+stride3q*4+48], m2
    mova  [dstq+strideq* 8+ 0], m0
    mova  [dstq+strideq* 8+16], m3
    mova  [dstq+strideq* 8+32], m2
    mova  [dstq+strideq* 8+48], m5
    mova  [dstq+strideq* 4+ 0], m3
    mova  [dstq+strideq* 4+16], m2
    mova  [dstq+strideq* 4+32], m5
    mova  [dstq+strideq* 4+48], m4
    mova  [dstq+strideq* 0+ 0], m2
    mova  [dstq+strideq* 0+16], m5
    mova  [dstq+strideq* 0+32], m4
    mova  [dstq+strideq* 0+48], m7
    sub                   dstq, strideq
%if cpuflag(avx)
    vpalignr                m1, m0,  m1,  4
    vpalignr                m0, m3,  m0,  4
    vpalignr                m3, m2,  m3,  4
    vpalignr                m2, m5,  m2,  4
    vpalignr                m5, m4,  m5,  4
    vpalignr                m4, m7,  m4,  4
    vpalignr                m7, m6,  m7,  4
%else
    SCRATCH                  6, 12, rsp+8*mmsize, sh
%if notcpuflag(ssse3)
    SCRATCH                  7, 13, rsp+9*mmsize
%endif
    PALIGNR                 m6, m0,  m1,  4, m7
    mova                    m1, m6
    PALIGNR                 m6, m3,  m0,  4, m7
    mova                    m0, m6
    PALIGNR                 m6, m2,  m3,  4, m7
    mova                    m3, m6
    PALIGNR                 m6, m5,  m2,  4, m7
    mova                    m2, m6
    PALIGNR                 m6, m4,  m5,  4, m7
    mova                    m5, m6
%if notcpuflag(ssse3)
    UNSCRATCH                7, 13, rsp+9*mmsize
    SCRATCH                  5, 13, rsp+9*mmsize
%endif
    PALIGNR                 m6, m7,  m4,  4, m5
    mova                    m4, m6
    PALIGNR                 m6, reg_sh,  m7,  4, m5
    mova                    m7, m6
%if notcpuflag(ssse3)
    UNSCRATCH                5, 13, rsp+9*mmsize
%endif
    UNSCRATCH                6, 12, rsp+8*mmsize, sh
%endif
    psrldq                  m6, 4
    dec                   cntd
    jg .loop_2
%endif
    RET
%endmacro

INIT_XMM sse2
HD_FUNCS
INIT_XMM ssse3
HD_FUNCS
INIT_XMM avx
HD_FUNCS