hevc_mc.asm 43 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
; /*
; * Provide SSE luma and chroma mc functions for HEVC decoding
; * Copyright (c) 2013 Pierre-Edouard LEPERE
; *
; * 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
24 25 26 27 28 29 30 31
pw_8:                   times 8 dw (1 << 9)
pw_10:                  times 8 dw (1 << 11)
pw_12:                  times 8 dw (1 << 13)
pw_bi_8:                times 8 dw (1 << 8)
pw_bi_10:               times 8 dw (1 << 10)
pw_bi_12:               times 8 dw (1 << 12)
max_pixels_10:          times 8  dw ((1 << 10)-1)
max_pixels_12:          times 8  dw ((1 << 12)-1)
32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56
zero:                   times 4  dd 0
one_per_32:             times 4  dd 1

SECTION .text
%macro EPEL_TABLE 4
hevc_epel_filters_%4_%1 times %2 d%3 -2, 58
                        times %2 d%3 10, -2
                        times %2 d%3 -4, 54
                        times %2 d%3 16, -2
                        times %2 d%3 -6, 46
                        times %2 d%3 28, -4
                        times %2 d%3 -4, 36
                        times %2 d%3 36, -4
                        times %2 d%3 -4, 28
                        times %2 d%3 46, -6
                        times %2 d%3 -2, 16
                        times %2 d%3 54, -4
                        times %2 d%3 -2, 10
                        times %2 d%3 58, -2
%endmacro



EPEL_TABLE  8, 8, b, sse4
EPEL_TABLE 10, 4, w, sse4
57
EPEL_TABLE 12, 4, w, sse4
58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75

%macro QPEL_TABLE 4
hevc_qpel_filters_%4_%1 times %2 d%3  -1,  4
                        times %2 d%3 -10, 58
                        times %2 d%3  17, -5
                        times %2 d%3   1,  0
                        times %2 d%3  -1,  4
                        times %2 d%3 -11, 40
                        times %2 d%3  40,-11
                        times %2 d%3   4, -1
                        times %2 d%3   0,  1
                        times %2 d%3  -5, 17
                        times %2 d%3  58,-10
                        times %2 d%3   4, -1
%endmacro

QPEL_TABLE  8, 8, b, sse4
QPEL_TABLE 10, 4, w, sse4
76
QPEL_TABLE 12, 4, w, sse4
77

78 79
%define MAX_PB_SIZE  64

80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121
%define hevc_qpel_filters_sse4_14 hevc_qpel_filters_sse4_10

%if ARCH_X86_64

%macro SIMPLE_BILOAD 4   ;width, tab, r1, r2
%if %1 <= 4
    movq              %3, [%2]                                              ; load data from source2
%elif %1 <= 8
    movdqa            %3, [%2]                                              ; load data from source2
%elif %1 <= 12
    movdqa            %3, [%2]                                              ; load data from source2
    movq              %4, [%2+16]                                           ; load data from source2
%else
    movdqa            %3, [%2]                                              ; load data from source2
    movdqa            %4, [%2+16]                                           ; load data from source2
%endif
%endmacro

%macro SIMPLE_LOAD 4    ;width, bitd, tab, r1
%if %1 == 2 || (%2 == 8 && %1 <= 4)
    movd              %4, [%3]                                               ; load data from source
%elif %1 == 4 || (%2 == 8 && %1 <= 8)
    movq              %4, [%3]                                               ; load data from source
%else
    movdqu            %4, [%3]                                               ; load data from source
%endif
%endmacro

%macro SIMPLE_8LOAD 5    ;width, bitd, tab, r1, r2
%if %1 == 2 || (%2 == 8 && %1 <= 4)
    movq              %4, [%3]                                              ; load data from source2
%elif %1 == 4 || (%2 == 8 && %1 <= 8)
    movdqa            %4, [%3]                                              ; load data from source2
%elif %1 <= 12
    movdqa            %4, [%3]                                              ; load data from source2
    movq              %5, [%3+16]                                           ; load data from source2
%else
    movdqa            %4, [%3]                                              ; load data from source2
    movdqa            %5, [%3+16]                                           ; load data from source2
%endif
%endmacro

122
%macro EPEL_FILTER 2-4                            ; bit depth, filter index
123 124 125 126 127 128 129
%ifdef PIC
    lea         rfilterq, [hevc_epel_filters_sse4_%1]
%else
    %define rfilterq hevc_epel_filters_sse4_%1
%endif
    sub              %2q, 1
    shl              %2q, 5                      ; multiply by 32
130
%if %0 == 2
131 132
    movdqa           m14, [rfilterq + %2q]        ; get 2 first values of filters
    movdqa           m15, [rfilterq + %2q+16]     ; get 2 last values of filters
133 134 135 136
%else
    movdqa           %3, [rfilterq + %2q]        ; get 2 first values of filters
    movdqa           %4, [rfilterq + %2q+16]     ; get 2 last values of filters
%endif
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
%endmacro

%macro EPEL_HV_FILTER 1
%ifdef PIC
    lea         rfilterq, [hevc_epel_filters_sse4_%1]
%else
    %define rfilterq hevc_epel_filters_sse4_%1
%endif
    sub              mxq, 1
    sub              myq, 1
    shl              mxq, 5                      ; multiply by 32
    shl              myq, 5                      ; multiply by 32
    movdqa           m14, [rfilterq + mxq]        ; get 2 first values of filters
    movdqa           m15, [rfilterq + mxq+16]     ; get 2 last values of filters
    lea           r3srcq, [srcstrideq*3]

%ifdef PIC
    lea         rfilterq, [hevc_epel_filters_sse4_10]
%else
    %define rfilterq hevc_epel_filters_sse4_10
%endif
    movdqa           m12, [rfilterq + myq]        ; get 2 first values of filters
    movdqa           m13, [rfilterq + myq+16]     ; get 2 last values of filters
%endmacro

%macro QPEL_FILTER 2
%ifdef PIC
    lea         rfilterq, [hevc_qpel_filters_sse4_%1]
%else
    %define rfilterq hevc_qpel_filters_sse4_%1
%endif
    lea              %2q, [%2q*8-8]
    movdqa           m12, [rfilterq + %2q*8]       ; get 4 first values of filters
    movdqa           m13, [rfilterq + %2q*8 + 16]  ; get 4 first values of filters
    movdqa           m14, [rfilterq + %2q*8 + 32]  ; get 4 first values of filters
    movdqa           m15, [rfilterq + %2q*8 + 48]  ; get 4 first values of filters
%endmacro

%macro EPEL_LOAD 4
%ifdef PIC
    lea rfilterq, [%2]
%else
    %define rfilterq %2
%endif
181 182 183 184 185 186 187 188 189
%if (%1 == 8 && %4 <= 4)
%define %%load movd
%elif (%1 == 8 && %4 <= 8) || (%1 > 8 && %4 <= 4)
%define %%load movq
%else
%define %%load movdqu
%endif

    %%load            m0, [rfilterq ]
190
%ifnum %3
191 192 193
    %%load            m1, [rfilterq+  %3]
    %%load            m2, [rfilterq+2*%3]
    %%load            m3, [rfilterq+3*%3]
194
%else
195 196 197
    %%load            m1, [rfilterq+  %3q]
    %%load            m2, [rfilterq+2*%3q]
    %%load            m3, [rfilterq+r3srcq]
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
%endif

%if %1 == 8
%if %4 > 8
    SBUTTERFLY        bw, 0, 1, 10
    SBUTTERFLY        bw, 2, 3, 10
%else
    punpcklbw         m0, m1
    punpcklbw         m2, m3
%endif
%else
%if %4 > 4
    SBUTTERFLY        wd, 0, 1, 10
    SBUTTERFLY        wd, 2, 3, 10
%else
    punpcklwd         m0, m1
    punpcklwd         m2, m3
%endif
%endif
%endmacro


%macro QPEL_H_LOAD 4
%assign %%stride (%1+7)/8
%if %1 == 8
%if %3 <= 4
%define %%load movd
%elif %3 == 8
%define %%load movq
%else
%define %%load movdqu
%endif
%else
%if %3 == 2
%define %%load movd
%elif %3 == 4
%define %%load movq
%else
%define %%load movdqu
%endif
%endif
    %%load            m0, [%2-3*%%stride]        ;load data from source
    %%load            m1, [%2-2*%%stride]
    %%load            m2, [%2-%%stride  ]
    %%load            m3, [%2           ]
    %%load            m4, [%2+%%stride  ]
    %%load            m5, [%2+2*%%stride]
    %%load            m6, [%2+3*%%stride]
    %%load            m7, [%2+4*%%stride]

%if %1 == 8
%if %3 > 8
    SBUTTERFLY        wd, 0, 1, %4
    SBUTTERFLY        wd, 2, 3, %4
    SBUTTERFLY        wd, 4, 5, %4
    SBUTTERFLY        wd, 6, 7, %4
%else
    punpcklwd         m0, m1
    punpcklwd         m2, m3
    punpcklwd         m4, m5
    punpcklwd         m6, m7
%endif
%else
%if %3 > 4
    SBUTTERFLY        dq, 0, 1, %4
    SBUTTERFLY        dq, 2, 3, %4
    SBUTTERFLY        dq, 4, 5, %4
    SBUTTERFLY        dq, 6, 7, %4
%else
    punpckldq         m0, m1
    punpckldq         m2, m3
    punpckldq         m4, m5
    punpckldq         m6, m7
%endif
%endif
%endmacro

275 276 277 278 279 280
%macro QPEL_V_LOAD 5
    lea              %5q, [%2]
    sub              %5q, r3srcq
    movdqu            m0, [%5q            ]      ;load x- 3*srcstride
    movdqu            m1, [%5q+   %3q     ]      ;load x- 2*srcstride
    movdqu            m2, [%5q+ 2*%3q     ]      ;load x-srcstride
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
    movdqu            m3, [%2       ]      ;load x
    movdqu            m4, [%2+   %3q]      ;load x+stride
    movdqu            m5, [%2+ 2*%3q]      ;load x+2*stride
    movdqu            m6, [%2+r3srcq]      ;load x+3*stride
    movdqu            m7, [%2+ 4*%3q]      ;load x+4*stride
%if %1 == 8
%if %4 > 8
    SBUTTERFLY        bw, 0, 1, 8
    SBUTTERFLY        bw, 2, 3, 8
    SBUTTERFLY        bw, 4, 5, 8
    SBUTTERFLY        bw, 6, 7, 8
%else
    punpcklbw         m0, m1
    punpcklbw         m2, m3
    punpcklbw         m4, m5
    punpcklbw         m6, m7
%endif
%else
%if %4 > 4
    SBUTTERFLY        wd, 0, 1, 8
    SBUTTERFLY        wd, 2, 3, 8
    SBUTTERFLY        wd, 4, 5, 8
    SBUTTERFLY        wd, 6, 7, 8
%else
    punpcklwd         m0, m1
    punpcklwd         m2, m3
    punpcklwd         m4, m5
    punpcklwd         m6, m7
%endif
%endif
%endmacro

313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335
%macro PEL_12STORE2 3
    movd           [%1], %2
%endmacro
%macro PEL_12STORE4 3
    movq           [%1], %2
%endmacro
%macro PEL_12STORE6 3
    movq           [%1], %2
    psrldq            %2, 8
    movd         [%1+8], %2
%endmacro
%macro PEL_12STORE8 3
    movdqa         [%1], %2
%endmacro
%macro PEL_12STORE12 3
    movdqa         [%1], %2
    movq        [%1+16], %3
%endmacro
%macro PEL_12STORE16 3
    PEL_12STORE8      %1, %2, %3
    movdqa       [%1+16], %3
%endmacro

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
%macro PEL_10STORE2 3
    movd           [%1], %2
%endmacro
%macro PEL_10STORE4 3
    movq           [%1], %2
%endmacro
%macro PEL_10STORE6 3
    movq           [%1], %2
    psrldq            %2, 8
    movd         [%1+8], %2
%endmacro
%macro PEL_10STORE8 3
    movdqa         [%1], %2
%endmacro
%macro PEL_10STORE12 3
    movdqa         [%1], %2
    movq        [%1+16], %3
%endmacro
%macro PEL_10STORE16 3
    PEL_10STORE8      %1, %2, %3
    movdqa       [%1+16], %3
%endmacro

%macro PEL_8STORE2 3
    pextrw          [%1], %2, 0
%endmacro
%macro PEL_8STORE4 3
    movd            [%1], %2
%endmacro
%macro PEL_8STORE6 3
    movd            [%1], %2
    pextrw        [%1+4], %2, 2
%endmacro
%macro PEL_8STORE8 3
    movq           [%1], %2
%endmacro
%macro PEL_8STORE12 3
    movq            [%1], %2
    psrldq            %2, 8
    movd          [%1+8], %2
%endmacro
%macro PEL_8STORE16 3
    movdqa          [%1], %2
%endmacro

381 382 383
%macro LOOP_END 3
    add              %1q, 2*MAX_PB_SIZE          ; dst += dststride
    add              %2q, %3q                    ; src += srcstride
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
    dec          heightd                         ; cmp height
    jnz               .loop                      ; height loop
%endmacro


%macro MC_PIXEL_COMPUTE 2 ;width, bitdepth
%if %2 == 8
%if %1 > 8
    punpckhbw         m1, m0, m2
    psllw             m1, 14-%2
%endif
    punpcklbw         m0, m2
%endif
    psllw             m0, 14-%2
%endmacro


%macro EPEL_COMPUTE 4 ; bitdepth, width, filter1, filter2
%if %1 == 8
    pmaddubsw         m0, %3   ;x1*c1+x2*c2
    pmaddubsw         m2, %4   ;x3*c3+x4*c4
    paddw             m0, m2
%if %2 > 8
    pmaddubsw         m1, %3
    pmaddubsw         m3, %4
    paddw             m1, m3
%endif
%else
    pmaddwd           m0, %3
    pmaddwd           m2, %4
    paddd             m0, m2
%if %2 > 4
    pmaddwd           m1, %3
    pmaddwd           m3, %4
    paddd             m1, m3
%endif
420
%if %1 != 8
421 422
    psrad             m0, %1-8
    psrad             m1, %1-8
423
%endif
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
    packssdw          m0, m1
%endif
%endmacro

%macro QPEL_HV_COMPUTE 4     ; width, bitdepth, filter idx
%ifdef PIC
    lea         rfilterq, [hevc_qpel_filters_sse4_%2]
%else
    %define rfilterq hevc_qpel_filters_sse4_%2
%endif

%if %2 == 8
    pmaddubsw         m0, [rfilterq + %3q*8   ]   ;x1*c1+x2*c2
    pmaddubsw         m2, [rfilterq + %3q*8+16]   ;x3*c3+x4*c4
    pmaddubsw         m4, [rfilterq + %3q*8+32]   ;x5*c5+x6*c6
    pmaddubsw         m6, [rfilterq + %3q*8+48]   ;x7*c7+x8*c8
    paddw             m0, m2
    paddw             m4, m6
    paddw             m0, m4
%else
    pmaddwd           m0, [rfilterq + %3q*8   ]
    pmaddwd           m2, [rfilterq + %3q*8+16]
    pmaddwd           m4, [rfilterq + %3q*8+32]
    pmaddwd           m6, [rfilterq + %3q*8+48]
    paddd             m0, m2
    paddd             m4, m6
    paddd             m0, m4
451
%if %2 != 8
452
    psrad             m0, %2-8
453
%endif
454 455 456 457 458 459 460 461
%if %1 > 4
    pmaddwd           m1, [rfilterq + %3q*8   ]
    pmaddwd           m3, [rfilterq + %3q*8+16]
    pmaddwd           m5, [rfilterq + %3q*8+32]
    pmaddwd           m7, [rfilterq + %3q*8+48]
    paddd             m1, m3
    paddd             m5, m7
    paddd             m1, m5
462
%if %2 != 8
463
    psrad             m1, %2-8
464
%endif
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
%endif
    p%4               m0, m1
%endif
%endmacro

%macro QPEL_COMPUTE 2     ; width, bitdepth
%if %2 == 8
    pmaddubsw         m0, m12   ;x1*c1+x2*c2
    pmaddubsw         m2, m13   ;x3*c3+x4*c4
    pmaddubsw         m4, m14   ;x5*c5+x6*c6
    pmaddubsw         m6, m15   ;x7*c7+x8*c8
    paddw             m0, m2
    paddw             m4, m6
    paddw             m0, m4
%if %1 > 8
    pmaddubsw         m1, m12
    pmaddubsw         m3, m13
    pmaddubsw         m5, m14
    pmaddubsw         m7, m15
    paddw             m1, m3
    paddw             m5, m7
    paddw             m1, m5
%endif
%else
    pmaddwd           m0, m12
    pmaddwd           m2, m13
    pmaddwd           m4, m14
    pmaddwd           m6, m15
    paddd             m0, m2
    paddd             m4, m6
    paddd             m0, m4
496
%if %2 != 8
497
    psrad             m0, %2-8
498
%endif
499 500 501 502 503 504 505 506
%if %1 > 4
    pmaddwd           m1, m12
    pmaddwd           m3, m13
    pmaddwd           m5, m14
    pmaddwd           m7, m15
    paddd             m1, m3
    paddd             m5, m7
    paddd             m1, m5
507
%if %2 != 8
508 509 510
    psrad             m1, %2-8
%endif
%endif
511
%endif
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
%endmacro

%macro BI_COMPUTE 7     ; width, bitd, src1l, src1h, scr2l, scr2h, pw
    paddsw            %3, %5
%if %1 > 8
    paddsw            %4, %6
%endif
    UNI_COMPUTE       %1, %2, %3, %4, %7
%endmacro

%macro UNI_COMPUTE 5
    pmulhrsw          %3, %5
%if %1 > 8 || (%2 > 8 && %1 > 4)
    pmulhrsw          %4, %5
%endif
%if %2 == 8
    packuswb          %3, %4
%else
    pminsw            %3, [max_pixels_%2]
    pmaxsw            %3, [zero]
%if %1 > 8
    pminsw            %4, [max_pixels_%2]
    pmaxsw            %4, [zero]
%endif
%endif
%endmacro

INIT_XMM sse4                                    ; adds ff_ and _sse4 to function name
; ******************************
; void put_hevc_mc_pixels(int16_t *dst, ptrdiff_t dststride,
;                         uint8_t *_src, ptrdiff_t _srcstride,
;                         int height, int mx, int my)
; ******************************

%macro HEVC_PUT_HEVC_PEL_PIXELS 2
547
cglobal hevc_put_hevc_pel_pixels%1_%2, 4, 4, 3, dst, src, srcstride,height
548 549 550 551 552
    pxor               m2, m2
.loop
    SIMPLE_LOAD       %1, %2, srcq, m0
    MC_PIXEL_COMPUTE  %1, %2
    PEL_10STORE%1     dstq, m0, m1
553
    LOOP_END         dst, src, srcstride
554 555
    RET

556
cglobal hevc_put_hevc_uni_pel_pixels%1_%2, 5, 5, 2, dst, dststride, src, srcstride,height
557 558 559
.loop
    SIMPLE_LOAD       %1, %2, srcq, m0
    PEL_%2STORE%1   dstq, m0, m1
560 561
    add             dstq, dststrideq             ; dst += dststride
    add             srcq, srcstrideq             ; src += srcstride
562 563 564 565
    dec          heightd                         ; cmp height
    jnz               .loop                      ; height loop
    RET

566
cglobal hevc_put_hevc_bi_pel_pixels%1_%2, 6, 6, 6, dst, dststride, src, srcstride, src2, height
567 568 569 570 571 572 573 574
    pxor              m2, m2
    movdqa            m5, [pw_bi_%2]
.loop
    SIMPLE_LOAD       %1, %2, srcq, m0
    SIMPLE_BILOAD     %1, src2q, m3, m4
    MC_PIXEL_COMPUTE  %1, %2
    BI_COMPUTE        %1, %2, m0, m1, m3, m4, m5
    PEL_%2STORE%1   dstq, m0, m1
575 576
    add             dstq, dststrideq             ; dst += dststride
    add             srcq, srcstrideq             ; src += srcstride
577
    add            src2q, 2*MAX_PB_SIZE          ; src += srcstride
578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593
    dec          heightd                         ; cmp height
    jnz               .loop                      ; height loop
    RET

%endmacro


; ******************************
; void put_hevc_epel_hX(int16_t *dst, ptrdiff_t dststride,
;                       uint8_t *_src, ptrdiff_t _srcstride,
;                       int width, int height, int mx, int my,
;                       int16_t* mcbuffer)
; ******************************


%macro HEVC_PUT_HEVC_EPEL 2
594
cglobal hevc_put_hevc_epel_h%1_%2, 5, 6, 6, dst, src, srcstride, height, mx, rfilter
595
%assign %%stride ((%2 + 7)/8)
596
    EPEL_FILTER       %2, mx, m4, m5
597 598
.loop
    EPEL_LOAD         %2, srcq-%%stride, %%stride, %1
599
    EPEL_COMPUTE      %2, %1, m4, m5
600
    PEL_10STORE%1      dstq, m0, m1
601
    LOOP_END         dst, src, srcstride
602 603
    RET

604
cglobal hevc_put_hevc_uni_epel_h%1_%2, 6, 7, 7, dst, dststride, src, srcstride, height, mx, rfilter
605
%assign %%stride ((%2 + 7)/8)
606 607
    movdqa            m6, [pw_%2]
    EPEL_FILTER       %2, mx, m4, m5
608 609
.loop
    EPEL_LOAD         %2, srcq-%%stride, %%stride, %1
610 611
    EPEL_COMPUTE      %2, %1, m4, m5
    UNI_COMPUTE       %1, %2, m0, m1, m6
612
    PEL_%2STORE%1   dstq, m0, m1
613 614
    add             dstq, dststrideq             ; dst += dststride
    add             srcq, srcstrideq             ; src += srcstride
615 616 617 618
    dec          heightd                         ; cmp height
    jnz               .loop                      ; height loop
    RET

619
cglobal hevc_put_hevc_bi_epel_h%1_%2, 7, 8, 7, dst, dststride, src, srcstride, src2, height, mx, rfilter
620 621
    movdqa            m6, [pw_bi_%2]
    EPEL_FILTER       %2, mx, m4, m5
622 623
.loop
    EPEL_LOAD         %2, srcq-%%stride, %%stride, %1
624
    EPEL_COMPUTE      %2, %1, m4, m5
625
    SIMPLE_BILOAD     %1, src2q, m2, m3
626
    BI_COMPUTE        %1, %2, m0, m1, m2, m3, m6
627
    PEL_%2STORE%1   dstq, m0, m1
628 629
    add             dstq, dststrideq             ; dst += dststride
    add             srcq, srcstrideq             ; src += srcstride
630
    add            src2q, 2*MAX_PB_SIZE          ; src += srcstride
631 632 633 634 635 636 637 638 639 640 641
    dec          heightd                         ; cmp height
    jnz               .loop                      ; height loop
    RET

; ******************************
; void put_hevc_epel_v(int16_t *dst, ptrdiff_t dststride,
;                      uint8_t *_src, ptrdiff_t _srcstride,
;                      int width, int height, int mx, int my,
;                      int16_t* mcbuffer)
; ******************************

642
cglobal hevc_put_hevc_epel_v%1_%2, 6, 7, 6, dst, src, srcstride, height, r3src, my, rfilter
643 644
    lea           r3srcq, [srcstrideq*3]
    sub             srcq, srcstrideq
645
    EPEL_FILTER       %2, my, m4, m5
646 647
.loop
    EPEL_LOAD         %2, srcq, srcstride, %1
648
    EPEL_COMPUTE      %2, %1, m4, m5
649
    PEL_10STORE%1     dstq, m0, m1
650
    LOOP_END          dst, src, srcstride
651 652
    RET

653
cglobal hevc_put_hevc_uni_epel_v%1_%2, 7, 8, 7, dst, dststride, src, srcstride, height, r3src, my, rfilter
654
    lea           r3srcq, [srcstrideq*3]
655
    movdqa            m6, [pw_%2]
656
    sub             srcq, srcstrideq
657
    EPEL_FILTER       %2, my, m4, m5
658 659
.loop
    EPEL_LOAD         %2, srcq, srcstride, %1
660 661
    EPEL_COMPUTE      %2, %1, m4, m5
    UNI_COMPUTE       %1, %2, m0, m1, m6
662
    PEL_%2STORE%1   dstq, m0, m1
663 664
    add             dstq, dststrideq             ; dst += dststride
    add             srcq, srcstrideq             ; src += srcstride
665 666 667 668 669
    dec          heightd                         ; cmp height
    jnz               .loop                      ; height loop
    RET


670
cglobal hevc_put_hevc_bi_epel_v%1_%2, 8, 9, 7, dst, dststride, src, srcstride, src2, height, r3src, my, rfilter
671
    lea           r3srcq, [srcstrideq*3]
672
    movdqa            m6, [pw_bi_%2]
673
    sub             srcq, srcstrideq
674
    EPEL_FILTER       %2, my, m4, m5
675 676
.loop
    EPEL_LOAD         %2, srcq, srcstride, %1
677
    EPEL_COMPUTE      %2, %1, m4, m5
678
    SIMPLE_BILOAD     %1, src2q, m2, m3
679
    BI_COMPUTE        %1, %2, m0, m1, m2, m3, m6
680
    PEL_%2STORE%1   dstq, m0, m1
681 682
    add             dstq, dststrideq             ; dst += dststride
    add             srcq, srcstrideq             ; src += srcstride
683
    add            src2q, 2*MAX_PB_SIZE          ; src += srcstride
684 685 686 687 688 689 690 691 692 693 694 695 696
    dec          heightd                         ; cmp height
    jnz               .loop                      ; height loop
    RET
%endmacro


; ******************************
; void put_hevc_epel_hv(int16_t *dst, ptrdiff_t dststride,
;                       uint8_t *_src, ptrdiff_t _srcstride,
;                       int width, int height, int mx, int my)
; ******************************

%macro HEVC_PUT_HEVC_EPEL_HV 2
697
cglobal hevc_put_hevc_epel_hv%1_%2, 6, 8, 12 , dst, src, srcstride, height, mx, my, r3src, rfilter
698 699 700 701 702 703
%assign %%stride ((%2 + 7)/8)
    sub             srcq, srcstrideq
    EPEL_HV_FILTER    %2
    EPEL_LOAD         %2, srcq-%%stride, %%stride, %1
    EPEL_COMPUTE      %2, %1, m14, m15
    SWAP              m4, m0
704
    add             srcq, srcstrideq
705 706 707
    EPEL_LOAD         %2, srcq-%%stride, %%stride, %1
    EPEL_COMPUTE      %2, %1, m14, m15
    SWAP              m5, m0
708
    add             srcq, srcstrideq
709 710 711
    EPEL_LOAD         %2, srcq-%%stride, %%stride, %1
    EPEL_COMPUTE      %2, %1, m14, m15
    SWAP              m6, m0
712
    add             srcq, srcstrideq
713 714 715 716 717 718 719 720 721 722 723 724 725 726 727
.loop
    EPEL_LOAD         %2, srcq-%%stride, %%stride, %1
    EPEL_COMPUTE      %2, %1, m14, m15
    SWAP              m7, m0
    punpcklwd         m0, m4, m5
    punpcklwd         m2, m6, m7
%if %1 > 4
    punpckhwd         m1, m4, m5
    punpckhwd         m3, m6, m7
%endif
    EPEL_COMPUTE      14, %1, m12, m13
    PEL_10STORE%1     dstq, m0, m1
    movdqa            m4, m5
    movdqa            m5, m6
    movdqa            m6, m7
728
    LOOP_END         dst, src, srcstride
729 730 731 732 733 734 735 736 737
    RET

cglobal hevc_put_hevc_uni_epel_hv%1_%2, 7, 9, 12 , dst, dststride, src, srcstride, height, mx, my, r3src, rfilter
%assign %%stride ((%2 + 7)/8)
    sub             srcq, srcstrideq
    EPEL_HV_FILTER    %2
    EPEL_LOAD         %2, srcq-%%stride, %%stride, %1
    EPEL_COMPUTE      %2, %1, m14, m15
    SWAP              m4, m0
738
    add             srcq, srcstrideq
739 740 741
    EPEL_LOAD         %2, srcq-%%stride, %%stride, %1
    EPEL_COMPUTE      %2, %1, m14, m15
    SWAP              m5, m0
742
    add             srcq, srcstrideq
743 744 745
    EPEL_LOAD         %2, srcq-%%stride, %%stride, %1
    EPEL_COMPUTE      %2, %1, m14, m15
    SWAP              m6, m0
746
    add             srcq, srcstrideq
747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762
.loop
    EPEL_LOAD         %2, srcq-%%stride, %%stride, %1
    EPEL_COMPUTE      %2, %1, m14, m15
    SWAP              m7, m0
    punpcklwd         m0, m4, m5
    punpcklwd         m2, m6, m7
%if %1 > 4
    punpckhwd         m1, m4, m5
    punpckhwd         m3, m6, m7
%endif
    EPEL_COMPUTE      14, %1, m12, m13
    UNI_COMPUTE       %1, %2, m0, m1, [pw_%2]
    PEL_%2STORE%1   dstq, m0, m1
    movdqa            m4, m5
    movdqa            m5, m6
    movdqa            m6, m7
763 764
    add             dstq, dststrideq             ; dst += dststride
    add             srcq, srcstrideq             ; src += srcstride
765 766 767 768 769
    dec          heightd                         ; cmp height
    jnz               .loop                      ; height loop
    RET


770
cglobal hevc_put_hevc_bi_epel_hv%1_%2, 8, 10, 16, dst, dststride, src, srcstride, src2, height, mx, my, r3src, rfilter
771 772 773 774 775 776
%assign %%stride ((%2 + 7)/8)
    sub             srcq, srcstrideq
    EPEL_HV_FILTER    %2
    EPEL_LOAD         %2, srcq-%%stride, %%stride, %1
    EPEL_COMPUTE      %2, %1, m14, m15
    SWAP              m4, m0
777
    add             srcq, srcstrideq
778 779 780
    EPEL_LOAD         %2, srcq-%%stride, %%stride, %1
    EPEL_COMPUTE      %2, %1, m14, m15
    SWAP              m5, m0
781
    add             srcq, srcstrideq
782 783 784
    EPEL_LOAD         %2, srcq-%%stride, %%stride, %1
    EPEL_COMPUTE      %2, %1, m14, m15
    SWAP              m6, m0
785
    add             srcq, srcstrideq
786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802
.loop
    EPEL_LOAD         %2, srcq-%%stride, %%stride, %1
    EPEL_COMPUTE      %2, %1, m14, m15
    SWAP              m7, m0
    punpcklwd         m0, m4, m5
    punpcklwd         m2, m6, m7
%if %1 > 4
    punpckhwd         m1, m4, m5
    punpckhwd         m3, m6, m7
%endif
    EPEL_COMPUTE      14, %1, m12, m13
    SIMPLE_BILOAD     %1, src2q, m8, m9
    BI_COMPUTE        %1, %2, m0, m1, m8, m9, [pw_bi_%2]
    PEL_%2STORE%1   dstq, m0, m1
    movdqa            m4, m5
    movdqa            m5, m6
    movdqa            m6, m7
803 804
    add             dstq, dststrideq             ; dst += dststride
    add             srcq, srcstrideq             ; src += srcstride
805
    add            src2q, 2*MAX_PB_SIZE          ; src += srcstride
806 807 808 809 810 811 812 813 814 815 816 817
    dec          heightd                         ; cmp height
    jnz               .loop                      ; height loop
    RET
%endmacro

; ******************************
; void put_hevc_qpel_hX_X_X(int16_t *dst, ptrdiff_t dststride,
;                       uint8_t *_src, ptrdiff_t _srcstride,
;                       int width, int height, int mx, int my)
; ******************************

%macro HEVC_PUT_HEVC_QPEL 2
818
cglobal hevc_put_hevc_qpel_h%1_%2, 5, 6, 15, dst, src, srcstride, height, mx, rfilter
819 820 821 822 823 824 825 826
    QPEL_FILTER       %2, mx
.loop
    QPEL_H_LOAD       %2, srcq, %1, 10
    QPEL_COMPUTE      %1, %2
%if %2 > 8
    packssdw          m0, m1
%endif
    PEL_10STORE%1     dstq, m0, m1
827
    LOOP_END          dst, src, srcstride
828 829 830 831 832 833 834 835 836 837 838 839 840
    RET

cglobal hevc_put_hevc_uni_qpel_h%1_%2, 6, 7, 15 , dst, dststride, src, srcstride, height, mx, rfilter
    movdqa            m9, [pw_%2]
    QPEL_FILTER       %2, mx
.loop
    QPEL_H_LOAD       %2, srcq, %1, 10
    QPEL_COMPUTE      %1, %2
%if %2 > 8
    packssdw          m0, m1
%endif
    UNI_COMPUTE       %1, %2, m0, m1, m9
    PEL_%2STORE%1   dstq, m0, m1
841 842
    add             dstq, dststrideq             ; dst += dststride
    add             srcq, srcstrideq             ; src += srcstride
843 844 845 846
    dec          heightd                         ; cmp height
    jnz               .loop                      ; height loop
    RET

847
cglobal hevc_put_hevc_bi_qpel_h%1_%2, 7, 8, 16 , dst, dststride, src, srcstride, src2, height, mx, rfilter
848 849 850 851 852 853 854 855 856 857 858
    movdqa            m9, [pw_bi_%2]
    QPEL_FILTER       %2, mx
.loop
    QPEL_H_LOAD       %2, srcq, %1, 10
    QPEL_COMPUTE      %1, %2
%if %2 > 8
    packssdw          m0, m1
%endif
    SIMPLE_BILOAD     %1, src2q, m10, m11
    BI_COMPUTE        %1, %2, m0, m1, m10, m11, m9
    PEL_%2STORE%1   dstq, m0, m1
859 860
    add             dstq, dststrideq             ; dst += dststride
    add             srcq, srcstrideq             ; src += srcstride
861
    add            src2q, 2*MAX_PB_SIZE          ; src += srcstride
862 863 864 865 866 867 868 869 870 871 872
    dec          heightd                         ; cmp height
    jnz               .loop                      ; height loop
    RET


; ******************************
; void put_hevc_qpel_vX_X_X(int16_t *dst, ptrdiff_t dststride,
;                       uint8_t *_src, ptrdiff_t _srcstride,
;                       int width, int height, int mx, int my)
; ******************************

873
cglobal hevc_put_hevc_qpel_v%1_%2, 6, 8, 15, dst, src, srcstride, height, r3src, my, rfilter
874 875 876
    lea           r3srcq, [srcstrideq*3]
    QPEL_FILTER       %2, my
.loop
877
    QPEL_V_LOAD       %2, srcq, srcstride, %1, r7
878 879 880 881 882
    QPEL_COMPUTE      %1, %2
%if %2 > 8
    packssdw          m0, m1
%endif
    PEL_10STORE%1     dstq, m0, m1
883
    LOOP_END         dst, src, srcstride
884 885
    RET

886
cglobal hevc_put_hevc_uni_qpel_v%1_%2, 7, 9, 15, dst, dststride, src, srcstride, height, r3src, my, rfilter
887 888 889 890
    movdqa            m9, [pw_%2]
    lea           r3srcq, [srcstrideq*3]
    QPEL_FILTER       %2, my
.loop
891
    QPEL_V_LOAD       %2, srcq, srcstride, %1, r8
892 893
    QPEL_COMPUTE      %1, %2
%if %2 > 8
894
    packssdw          m0, m1
895 896 897
%endif
    UNI_COMPUTE       %1, %2, m0, m1, m9
    PEL_%2STORE%1   dstq, m0, m1
898 899
    add             dstq, dststrideq             ; dst += dststride
    add             srcq, srcstrideq             ; src += srcstride
900 901 902 903
    dec          heightd                         ; cmp height
    jnz               .loop                      ; height loop
    RET

904
cglobal hevc_put_hevc_bi_qpel_v%1_%2, 8, 10, 16, dst, dststride, src, srcstride, src2, height, r3src, my, rfilter
905 906 907 908 909
    movdqa            m9, [pw_bi_%2]
    lea           r3srcq, [srcstrideq*3]
    QPEL_FILTER       %2, my
.loop
    SIMPLE_BILOAD     %1, src2q, m10, m11
910
    QPEL_V_LOAD       %2, srcq, srcstride, %1, r9
911 912 913 914 915 916
    QPEL_COMPUTE      %1, %2
%if %2 > 8
    packssdw          m0, m1
%endif
    BI_COMPUTE        %1, %2, m0, m1, m10, m11, m9
    PEL_%2STORE%1   dstq, m0, m1
917 918
    add             dstq, dststrideq             ; dst += dststride
    add             srcq, srcstrideq             ; src += srcstride
919
    add            src2q, 2*MAX_PB_SIZE          ; src += srcstride
920 921 922 923 924 925 926 927 928 929 930 931
    dec          heightd                         ; cmp height
    jnz               .loop                      ; height loop
    RET
%endmacro


; ******************************
; void put_hevc_qpel_hvX_X(int16_t *dst, ptrdiff_t dststride,
;                       uint8_t *_src, ptrdiff_t _srcstride,
;                       int height, int mx, int my)
; ******************************
%macro HEVC_PUT_HEVC_QPEL_HV 2
932
cglobal hevc_put_hevc_qpel_hv%1_%2, 6, 8, 12, dst, src, srcstride, height, mx, my, r3src, rfilter
933 934 935 936 937 938 939
    lea              mxq, [mxq*8-8]
    lea              myq, [myq*8-8]
    lea           r3srcq, [srcstrideq*3]
    sub             srcq, r3srcq
    QPEL_H_LOAD       %2, srcq, %1, 15
    QPEL_HV_COMPUTE   %1, %2, mx, ackssdw
    SWAP              m8, m0
940
    add             srcq, srcstrideq
941 942 943
    QPEL_H_LOAD       %2, srcq, %1, 15
    QPEL_HV_COMPUTE   %1, %2, mx, ackssdw
    SWAP              m9, m0
944
    add             srcq, srcstrideq
945 946 947
    QPEL_H_LOAD       %2, srcq, %1, 15
    QPEL_HV_COMPUTE   %1, %2, mx, ackssdw
    SWAP             m10, m0
948
    add             srcq, srcstrideq
949 950 951
    QPEL_H_LOAD       %2, srcq, %1, 15
    QPEL_HV_COMPUTE   %1, %2, mx, ackssdw
    SWAP             m11, m0
952
    add             srcq, srcstrideq
953 954 955
    QPEL_H_LOAD       %2, srcq, %1, 15
    QPEL_HV_COMPUTE   %1, %2, mx, ackssdw
    SWAP             m12, m0
956
    add             srcq, srcstrideq
957 958 959
    QPEL_H_LOAD       %2, srcq, %1, 15
    QPEL_HV_COMPUTE   %1, %2, mx, ackssdw
    SWAP             m13, m0
960
    add             srcq, srcstrideq
961 962 963
    QPEL_H_LOAD       %2, srcq, %1, 15
    QPEL_HV_COMPUTE   %1, %2, mx, ackssdw
    SWAP             m14, m0
964
    add             srcq, srcstrideq
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
.loop
    QPEL_H_LOAD       %2, srcq, %1, 15
    QPEL_HV_COMPUTE   %1, %2, mx, ackssdw
    SWAP             m15, m0
    punpcklwd         m0, m8, m9
    punpcklwd         m2, m10, m11
    punpcklwd         m4, m12, m13
    punpcklwd         m6, m14, m15
%if %1 > 4
    punpckhwd         m1, m8, m9
    punpckhwd         m3, m10, m11
    punpckhwd         m5, m12, m13
    punpckhwd         m7, m14, m15
%endif
    QPEL_HV_COMPUTE   %1, 14, my, ackssdw
    PEL_10STORE%1     dstq, m0, m1
%if %1 <= 4
    movq              m8, m9
    movq              m9, m10
    movq             m10, m11
    movq             m11, m12
    movq             m12, m13
    movq             m13, m14
    movq             m14, m15
%else
    movdqa            m8, m9
    movdqa            m9, m10
    movdqa           m10, m11
    movdqa           m11, m12
    movdqa           m12, m13
    movdqa           m13, m14
    movdqa           m14, m15
%endif
998
    LOOP_END         dst, src, srcstride
999 1000 1001 1002 1003 1004 1005 1006 1007 1008
    RET

cglobal hevc_put_hevc_uni_qpel_hv%1_%2, 7, 9, 12 , dst, dststride, src, srcstride, height, mx, my, r3src, rfilter
    lea              mxq, [mxq*8-8]
    lea              myq, [myq*8-8]
    lea           r3srcq, [srcstrideq*3]
    sub             srcq, r3srcq
    QPEL_H_LOAD       %2, srcq, %1, 15
    QPEL_HV_COMPUTE   %1, %2, mx, ackssdw
    SWAP              m8, m0
1009
    add             srcq, srcstrideq
1010 1011 1012
    QPEL_H_LOAD       %2, srcq, %1, 15
    QPEL_HV_COMPUTE   %1, %2, mx, ackssdw
    SWAP              m9, m0
1013
    add             srcq, srcstrideq
1014 1015 1016
    QPEL_H_LOAD       %2, srcq, %1, 15
    QPEL_HV_COMPUTE   %1, %2, mx, ackssdw
    SWAP             m10, m0
1017
    add             srcq, srcstrideq
1018 1019 1020
    QPEL_H_LOAD       %2, srcq, %1, 15
    QPEL_HV_COMPUTE   %1, %2, mx, ackssdw
    SWAP             m11, m0
1021
    add             srcq, srcstrideq
1022 1023 1024
    QPEL_H_LOAD       %2, srcq, %1, 15
    QPEL_HV_COMPUTE   %1, %2, mx, ackssdw
    SWAP             m12, m0
1025
    add             srcq, srcstrideq
1026 1027 1028
    QPEL_H_LOAD       %2, srcq, %1, 15
    QPEL_HV_COMPUTE   %1, %2, mx, ackssdw
    SWAP             m13, m0
1029
    add             srcq, srcstrideq
1030 1031 1032
    QPEL_H_LOAD       %2, srcq, %1, 15
    QPEL_HV_COMPUTE   %1, %2, mx, ackssdw
    SWAP             m14, m0
1033
    add             srcq, srcstrideq
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
.loop
    QPEL_H_LOAD       %2, srcq, %1, 15
    QPEL_HV_COMPUTE   %1, %2, mx, ackssdw
    SWAP             m15, m0
    punpcklwd         m0, m8, m9
    punpcklwd         m2, m10, m11
    punpcklwd         m4, m12, m13
    punpcklwd         m6, m14, m15
%if %1 > 4
    punpckhwd         m1, m8, m9
    punpckhwd         m3, m10, m11
    punpckhwd         m5, m12, m13
    punpckhwd         m7, m14, m15
%endif
    QPEL_HV_COMPUTE   %1, 14, my, ackusdw
    UNI_COMPUTE       %1, %2, m0, m1, [pw_%2]
    PEL_%2STORE%1   dstq, m0, m1

%if %1 <= 4
    movq              m8, m9
    movq              m9, m10
    movq             m10, m11
    movq             m11, m12
    movq             m12, m13
    movq             m13, m14
    movq             m14, m15
%else
    movdqa            m8, m9
    movdqa            m9, m10
    movdqa           m10, m11
    movdqa           m11, m12
    movdqa           m12, m13
    movdqa           m13, m14
    movdqa           m14, m15
%endif
1069 1070
    add             dstq, dststrideq             ; dst += dststride
    add             srcq, srcstrideq             ; src += srcstride
1071 1072 1073 1074
    dec          heightd                         ; cmp height
    jnz               .loop                      ; height loop
    RET

1075
cglobal hevc_put_hevc_bi_qpel_hv%1_%2, 8, 10, 16, dst, dststride, src, srcstride, src2, height, mx, my, r3src, rfilter
1076 1077 1078 1079 1080 1081 1082
    lea              mxq, [mxq*8-8]
    lea              myq, [myq*8-8]
    lea           r3srcq, [srcstrideq*3]
    sub             srcq, r3srcq
    QPEL_H_LOAD       %2, srcq, %1, 15
    QPEL_HV_COMPUTE   %1, %2, mx, ackssdw
    SWAP              m8, m0
1083
    add             srcq, srcstrideq
1084 1085 1086
    QPEL_H_LOAD       %2, srcq, %1, 15
    QPEL_HV_COMPUTE   %1, %2, mx, ackssdw
    SWAP              m9, m0
1087
    add             srcq, srcstrideq
1088 1089 1090
    QPEL_H_LOAD       %2, srcq, %1, 15
    QPEL_HV_COMPUTE   %1, %2, mx, ackssdw
    SWAP             m10, m0
1091
    add             srcq, srcstrideq
1092 1093 1094
    QPEL_H_LOAD       %2, srcq, %1, 15
    QPEL_HV_COMPUTE   %1, %2, mx, ackssdw
    SWAP             m11, m0
1095
    add             srcq, srcstrideq
1096 1097 1098
    QPEL_H_LOAD       %2, srcq, %1, 15
    QPEL_HV_COMPUTE   %1, %2, mx, ackssdw
    SWAP             m12, m0
1099
    add             srcq, srcstrideq
1100 1101 1102
    QPEL_H_LOAD       %2, srcq, %1, 15
    QPEL_HV_COMPUTE   %1, %2, mx, ackssdw
    SWAP             m13, m0
1103
    add             srcq, srcstrideq
1104 1105 1106
    QPEL_H_LOAD       %2, srcq, %1, 15
    QPEL_HV_COMPUTE   %1, %2, mx, ackssdw
    SWAP             m14, m0
1107
    add             srcq, srcstrideq
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
.loop
    QPEL_H_LOAD       %2, srcq, %1, 15
    QPEL_HV_COMPUTE   %1, %2, mx, ackssdw
    SWAP             m15, m0
    punpcklwd         m0, m8, m9
    punpcklwd         m2, m10, m11
    punpcklwd         m4, m12, m13
    punpcklwd         m6, m14, m15
%if %1 > 4
    punpckhwd         m1, m8, m9
    punpckhwd         m3, m10, m11
    punpckhwd         m5, m12, m13
    punpckhwd         m7, m14, m15
%endif
    QPEL_HV_COMPUTE   %1, 14, my, ackssdw
    SIMPLE_BILOAD     %1, src2q, m8, m9 ;m9 not used in this case
    BI_COMPUTE        %1, %2, m0, m1, m8, m9, [pw_bi_%2]
    PEL_%2STORE%1   dstq, m0, m1

%if %1 <= 4
    movq              m8, m9
    movq              m9, m10
    movq             m10, m11
    movq             m11, m12
    movq             m12, m13
    movq             m13, m14
    movq             m14, m15
%else
    movdqa            m8, m9
    movdqa            m9, m10
    movdqa           m10, m11
    movdqa           m11, m12
    movdqa           m12, m13
    movdqa           m13, m14
    movdqa           m14, m15
%endif
1144 1145
    add             dstq, dststrideq             ; dst += dststride
    add             srcq, srcstrideq             ; src += srcstride
1146
    add            src2q, 2*MAX_PB_SIZE          ; src += srcstride
1147 1148 1149 1150 1151 1152
    dec          heightd                         ; cmp height
    jnz               .loop                      ; height loop
    RET
%endmacro

%macro WEIGHTING_FUNCS 2
1153 1154 1155 1156 1157 1158 1159 1160 1161
%if WIN64 || ARCH_X86_32
cglobal hevc_put_hevc_uni_w%1_%2, 4, 5, 7, dst, dststride, src, srcstride, height, denom, wx, ox
    mov             r4d, denomm
%define SHIFT  r4d
%else
cglobal hevc_put_hevc_uni_w%1_%2, 6, 6, 7, dst, dststride, src, srcstride, height, denom, wx, ox
%define SHIFT  denomd
%endif
    lea           SHIFT, [SHIFT+14-%2]          ; shift = 14 - bitd + denom
1162 1163 1164
%if %1 <= 4
    pxor             m1, m1
%endif
1165 1166
    movd             m2, wxm        ; WX
    movd             m4, SHIFT      ; shift
1167 1168 1169
%if %1 <= 4
    punpcklwd        m2, m1
%else
1170
    punpcklwd        m2, m2
1171
%endif
1172
    dec           SHIFT
1173
    movdqu           m5, [one_per_32]
1174 1175 1176
    movd             m6, SHIFT
    pshufd           m2, m2, 0
    mov           SHIFT, oxm
1177
    pslld            m5, m6
1178 1179 1180 1181 1182 1183 1184 1185
%if %2 != 8
    shl           SHIFT, %2-8       ; ox << (bitd - 8)
%endif
    movd             m3, SHIFT      ; OX
    pshufd           m3, m3, 0
%if WIN64 || ARCH_X86_32
    mov           SHIFT, heightm
%endif
1186 1187
.loop
   SIMPLE_LOAD        %1, 10, srcq, m0
1188 1189 1190 1191 1192 1193 1194
%if %1 <= 4
    punpcklwd         m0, m1
    pmaddwd           m0, m2
    paddd             m0, m5
    psrad             m0, m4
    paddd             m0, m3
%else
1195 1196 1197 1198 1199 1200 1201 1202 1203 1204
    pmulhw            m6, m0, m2
    pmullw            m0, m2
    punpckhwd         m1, m0, m6
    punpcklwd         m0, m6
    paddd             m0, m5
    paddd             m1, m5
    psrad             m0, m4
    psrad             m1, m4
    paddd             m0, m3
    paddd             m1, m3
1205
%endif
1206 1207 1208 1209 1210 1211 1212
    packusdw          m0, m1
%if %2 == 8
    packuswb          m0, m0
%else
    pminsw            m0, [max_pixels_%2]
%endif
    PEL_%2STORE%1   dstq, m0, m1
1213
    add             dstq, dststrideq             ; dst += dststride
1214
    add             srcq, 2*MAX_PB_SIZE          ; src += srcstride
1215 1216 1217 1218
    dec          heightd                         ; cmp height
    jnz               .loop                      ; height loop
    RET

1219
cglobal hevc_put_hevc_bi_w%1_%2, 5, 7, 10, dst, dststride, src, srcstride, src2, height, denom, wx0, wx1, ox0, ox1
1220
    mov              r6d, denomm
1221 1222 1223
%if %1 <= 4
    pxor              m1, m1
%endif
1224 1225 1226 1227
    movd              m2, wx0m         ; WX0
    lea              r6d, [r6d+14-%2]  ; shift = 14 - bitd + denom
    movd              m3, wx1m         ; WX1
    movd              m0, r6d          ; shift
1228 1229 1230 1231
%if %1 <= 4
    punpcklwd         m2, m1
    punpcklwd         m3, m1
%else
1232 1233
    punpcklwd         m2, m2
    punpcklwd         m3, m3
1234 1235
%endif
    inc              r6d
1236
    movd              m5, r6d          ; shift+1
1237
    pshufd            m2, m2, 0
1238
    mov              r6d, ox0m
1239
    pshufd            m3, m3, 0
1240 1241 1242 1243 1244 1245
    add              r6d, ox1m
%if %2 != 8
    shl              r6d, %2-8         ; ox << (bitd - 8)
%endif
    inc              r6d
    movd              m4, r6d          ; offset
1246
    pshufd            m4, m4, 0
1247 1248
    mov              r6d, heightm
    pslld             m4, m0
1249 1250 1251

.loop
   SIMPLE_LOAD        %1, 10, srcq,  m0
1252
   SIMPLE_LOAD        %1, 10, src2q, m8
1253 1254 1255 1256 1257 1258 1259 1260 1261
%if %1 <= 4
    punpcklwd         m0, m1
    punpcklwd         m8, m1
    pmaddwd           m0, m3
    pmaddwd           m8, m2
    paddd             m0, m4
    paddd             m0, m8
    psrad             m0, m5
%else
1262 1263
    pmulhw            m6, m0, m3
    pmullw            m0, m3
1264 1265
    pmulhw            m7, m8, m2
    pmullw            m8, m2
1266 1267
    punpckhwd         m1, m0, m6
    punpcklwd         m0, m6
1268 1269 1270 1271
    punpckhwd         m9, m8, m7
    punpcklwd         m8, m7
    paddd             m0, m8
    paddd             m1, m9
1272 1273 1274 1275
    paddd             m0, m4
    paddd             m1, m4
    psrad             m0, m5
    psrad             m1, m5
1276
%endif
1277 1278 1279 1280 1281 1282 1283
    packusdw          m0, m1
%if %2 == 8
    packuswb          m0, m0
%else
    pminsw            m0, [max_pixels_%2]
%endif
    PEL_%2STORE%1   dstq, m0, m1
1284
    add             dstq, dststrideq             ; dst += dststride
1285
    add             srcq, 2*MAX_PB_SIZE          ; src += srcstride
1286
    add            src2q, 2*MAX_PB_SIZE          ; src2 += srcstride
1287
    dec              r6d                         ; cmp height
1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301
    jnz               .loop                      ; height loop
    RET
%endmacro

WEIGHTING_FUNCS 2, 8
WEIGHTING_FUNCS 4, 8
WEIGHTING_FUNCS 6, 8
WEIGHTING_FUNCS 8, 8

WEIGHTING_FUNCS 2, 10
WEIGHTING_FUNCS 4, 10
WEIGHTING_FUNCS 6, 10
WEIGHTING_FUNCS 8, 10

1302 1303 1304 1305 1306
WEIGHTING_FUNCS 2, 12
WEIGHTING_FUNCS 4, 12
WEIGHTING_FUNCS 6, 12
WEIGHTING_FUNCS 8, 12

1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318
HEVC_PUT_HEVC_PEL_PIXELS  2, 8
HEVC_PUT_HEVC_PEL_PIXELS  4, 8
HEVC_PUT_HEVC_PEL_PIXELS  6, 8
HEVC_PUT_HEVC_PEL_PIXELS  8, 8
HEVC_PUT_HEVC_PEL_PIXELS 12, 8
HEVC_PUT_HEVC_PEL_PIXELS 16, 8

HEVC_PUT_HEVC_PEL_PIXELS 2, 10
HEVC_PUT_HEVC_PEL_PIXELS 4, 10
HEVC_PUT_HEVC_PEL_PIXELS 6, 10
HEVC_PUT_HEVC_PEL_PIXELS 8, 10

1319 1320 1321 1322
HEVC_PUT_HEVC_PEL_PIXELS 2, 12
HEVC_PUT_HEVC_PEL_PIXELS 4, 12
HEVC_PUT_HEVC_PEL_PIXELS 6, 12
HEVC_PUT_HEVC_PEL_PIXELS 8, 12
1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336

HEVC_PUT_HEVC_EPEL 2,  8
HEVC_PUT_HEVC_EPEL 4,  8
HEVC_PUT_HEVC_EPEL 6,  8
HEVC_PUT_HEVC_EPEL 8,  8
HEVC_PUT_HEVC_EPEL 12, 8
HEVC_PUT_HEVC_EPEL 16, 8


HEVC_PUT_HEVC_EPEL 2, 10
HEVC_PUT_HEVC_EPEL 4, 10
HEVC_PUT_HEVC_EPEL 6, 10
HEVC_PUT_HEVC_EPEL 8, 10

1337 1338 1339 1340
HEVC_PUT_HEVC_EPEL 2, 12
HEVC_PUT_HEVC_EPEL 4, 12
HEVC_PUT_HEVC_EPEL 6, 12
HEVC_PUT_HEVC_EPEL 8, 12
1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351

HEVC_PUT_HEVC_EPEL_HV 2,  8
HEVC_PUT_HEVC_EPEL_HV 4,  8
HEVC_PUT_HEVC_EPEL_HV 6,  8
HEVC_PUT_HEVC_EPEL_HV 8,  8

HEVC_PUT_HEVC_EPEL_HV 2, 10
HEVC_PUT_HEVC_EPEL_HV 4, 10
HEVC_PUT_HEVC_EPEL_HV 6, 10
HEVC_PUT_HEVC_EPEL_HV 8, 10

1352 1353 1354 1355
HEVC_PUT_HEVC_EPEL_HV 2, 12
HEVC_PUT_HEVC_EPEL_HV 4, 12
HEVC_PUT_HEVC_EPEL_HV 6, 12
HEVC_PUT_HEVC_EPEL_HV 8, 12
1356 1357 1358 1359 1360 1361 1362 1363 1364

HEVC_PUT_HEVC_QPEL 4,  8
HEVC_PUT_HEVC_QPEL 8,  8
HEVC_PUT_HEVC_QPEL 12, 8
HEVC_PUT_HEVC_QPEL 16, 8

HEVC_PUT_HEVC_QPEL 4, 10
HEVC_PUT_HEVC_QPEL 8, 10

1365 1366 1367
HEVC_PUT_HEVC_QPEL 4, 12
HEVC_PUT_HEVC_QPEL 8, 12

1368 1369 1370 1371 1372 1373 1374 1375 1376 1377
HEVC_PUT_HEVC_QPEL_HV 2, 8
HEVC_PUT_HEVC_QPEL_HV 4, 8
HEVC_PUT_HEVC_QPEL_HV 6, 8
HEVC_PUT_HEVC_QPEL_HV 8, 8

HEVC_PUT_HEVC_QPEL_HV 2, 10
HEVC_PUT_HEVC_QPEL_HV 4, 10
HEVC_PUT_HEVC_QPEL_HV 6, 10
HEVC_PUT_HEVC_QPEL_HV 8, 10

1378 1379 1380 1381 1382
HEVC_PUT_HEVC_QPEL_HV 2, 12
HEVC_PUT_HEVC_QPEL_HV 4, 12
HEVC_PUT_HEVC_QPEL_HV 6, 12
HEVC_PUT_HEVC_QPEL_HV 8, 12

1383
%endif ; ARCH_X86_64