summaryrefslogtreecommitdiff
path: root/libavcodec/x86/h264_chromamc.asm
blob: 99faf44651d98fb1e00ee00986a40ad5487220c7 (plain)
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
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
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
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
604
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
658
659
660
661
662
663
664
665
666
667
668
669
;******************************************************************************
;* MMX/SSSE3-optimized functions for H264 chroma MC
;* Copyright (c) 2005 Zoltan Hidvegi <hzoli -a- hzoli -d- com>,
;*               2005-2008 Loren Merritt
;*
;* This file is part of Libav.
;*
;* Libav 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.
;*
;* Libav 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 Libav; if not, write to the Free Software
;* 51, Inc., Foundation Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
;******************************************************************************

%include "x86inc.asm"
%include "x86util.asm"

SECTION_RODATA

rnd_rv40_2d_tbl: times 4 dw  0
                 times 4 dw 16
                 times 4 dw 32
                 times 4 dw 16
                 times 4 dw 32
                 times 4 dw 28
                 times 4 dw 32
                 times 4 dw 28
                 times 4 dw  0
                 times 4 dw 32
                 times 4 dw 16
                 times 4 dw 32
                 times 4 dw 32
                 times 4 dw 28
                 times 4 dw 32
                 times 4 dw 28
rnd_rv40_1d_tbl: times 4 dw  0
                 times 4 dw  2
                 times 4 dw  4
                 times 4 dw  2
                 times 4 dw  4
                 times 4 dw  3
                 times 4 dw  4
                 times 4 dw  3
                 times 4 dw  0
                 times 4 dw  4
                 times 4 dw  2
                 times 4 dw  4
                 times 4 dw  4
                 times 4 dw  3
                 times 4 dw  4
                 times 4 dw  3

cextern pw_3
cextern pw_4
cextern pw_8
cextern pw_28
cextern pw_32
cextern pw_64

SECTION .text

%macro mv0_pixels_mc8 0
    lea           r4, [r2*2 ]
.next4rows
    movq         mm0, [r1   ]
    movq         mm1, [r1+r2]
    CHROMAMC_AVG mm0, [r0   ]
    CHROMAMC_AVG mm1, [r0+r2]
    movq     [r0   ], mm0
    movq     [r0+r2], mm1
    add           r0, r4
    add           r1, r4
    movq         mm0, [r1   ]
    movq         mm1, [r1+r2]
    CHROMAMC_AVG mm0, [r0   ]
    CHROMAMC_AVG mm1, [r0+r2]
    add           r1, r4
    movq     [r0   ], mm0
    movq     [r0+r2], mm1
    add           r0, r4
    sub          r3d, 4
    jne .next4rows
%endmacro

%macro chroma_mc8_mmx_func 3
; put/avg_h264_chroma_mc8_mmx_*(uint8_t *dst /*align 8*/, uint8_t *src /*align 1*/,
;                              int stride, int h, int mx, int my)
cglobal %1_%2_chroma_mc8_%3, 6, 7, 0
%ifdef ARCH_X86_64
    movsxd        r2, r2d
%endif
    mov          r6d, r5d
    or           r6d, r4d
    jne .at_least_one_non_zero
    ; mx == 0 AND my == 0 - no filter needed
    mv0_pixels_mc8
    REP_RET

.at_least_one_non_zero
%ifidn %2, rv40
%ifdef PIC
%define rnd_1d_rv40 r11
%define rnd_2d_rv40 r11
%else ; no-PIC
%define rnd_1d_rv40 rnd_rv40_1d_tbl
%define rnd_2d_rv40 rnd_rv40_2d_tbl
%endif
%ifdef ARCH_X86_64
    mov          r10, r5
    and          r10, 6         ; &~1 for mx/my=[0,7]
    lea          r10, [r10*4+r4]
    sar         r10d, 1
%define rnd_bias r10
%define dest_reg r0
%else ; x86-32
    mov           r0, r5
    and           r0, 6         ; &~1 for mx/my=[0,7]
    lea           r0, [r0*4+r4]
    sar          r0d, 1
%define rnd_bias r0
%define dest_reg r5
%endif
%else ; vc1, h264
%define rnd_bias  0
%define dest_reg r0
%endif

    test         r5d, r5d
    mov           r6, 1
    je .my_is_zero
    test         r4d, r4d
    mov           r6, r2        ; dxy = x ? 1 : stride
    jne .both_non_zero
.my_is_zero
    ; mx == 0 XOR my == 0 - 1 dimensional filter only
    or           r4d, r5d       ; x + y

%ifidn %2, rv40
%ifdef PIC
    lea          r11, [rnd_rv40_1d_tbl]
%endif
%ifndef ARCH_X86_64
    mov           r5, r0m
%endif
%endif

    movd          m5, r4d
    movq          m4, [pw_8]
    movq          m6, [rnd_1d_%2+rnd_bias*8] ; mm6 = rnd >> 3
    punpcklwd     m5, m5
    punpckldq     m5, m5        ; mm5 = B = x
    pxor          m7, m7
    psubw         m4, m5        ; mm4 = A = 8-x

.next1drow
    movq          m0, [r1   ]   ; mm0 = src[0..7]
    movq          m2, [r1+r6]   ; mm1 = src[1..8]

    movq          m1, m0
    movq          m3, m2
    punpcklbw     m0, m7
    punpckhbw     m1, m7
    punpcklbw     m2, m7
    punpckhbw     m3, m7
    pmullw        m0, m4        ; [mm0,mm1] = A * src[0..7]
    pmullw        m1, m4
    pmullw        m2, m5        ; [mm2,mm3] = B * src[1..8]
    pmullw        m3, m5

    paddw         m0, m6
    paddw         m1, m6
    paddw         m0, m2
    paddw         m1, m3
    psrlw         m0, 3
    psrlw         m1, 3
    packuswb      m0, m1
    CHROMAMC_AVG  m0, [dest_reg]
    movq  [dest_reg], m0        ; dst[0..7] = (A * src[0..7] + B * src[1..8] + (rnd >> 3)) >> 3

    add     dest_reg, r2
    add           r1, r2
    dec           r3d
    jne .next1drow
    REP_RET

.both_non_zero ; general case, bilinear
    movd          m4, r4d         ; x
    movd          m6, r5d         ; y
%ifidn %2, rv40
%ifdef PIC
    lea          r11, [rnd_rv40_2d_tbl]
%endif
%ifndef ARCH_X86_64
    mov           r5, r0m
%endif
%endif
    mov           r6, rsp         ; backup stack pointer
    and          rsp, ~(mmsize-1) ; align stack
    sub          rsp, 16          ; AA and DD

    punpcklwd     m4, m4
    punpcklwd     m6, m6
    punpckldq     m4, m4          ; mm4 = x words
    punpckldq     m6, m6          ; mm6 = y words
    movq          m5, m4
    pmullw        m4, m6          ; mm4 = x * y
    psllw         m5, 3
    psllw         m6, 3
    movq          m7, m5
    paddw         m7, m6
    movq     [rsp+8], m4          ; DD = x * y
    psubw         m5, m4          ; mm5 = B = 8x - xy
    psubw         m6, m4          ; mm6 = C = 8y - xy
    paddw         m4, [pw_64]
    psubw         m4, m7          ; mm4 = A = xy - (8x+8y) + 64
    pxor          m7, m7
    movq     [rsp  ], m4

    movq          m0, [r1  ]      ; mm0 = src[0..7]
    movq          m1, [r1+1]      ; mm1 = src[1..8]
.next2drow
    add           r1, r2

    movq          m2, m0
    movq          m3, m1
    punpckhbw     m0, m7
    punpcklbw     m1, m7
    punpcklbw     m2, m7
    punpckhbw     m3, m7
    pmullw        m0, [rsp]
    pmullw        m2, [rsp]
    pmullw        m1, m5
    pmullw        m3, m5
    paddw         m2, m1          ; mm2 = A * src[0..3] + B * src[1..4]
    paddw         m3, m0          ; mm3 = A * src[4..7] + B * src[5..8]

    movq          m0, [r1]
    movq          m1, m0
    punpcklbw     m0, m7
    punpckhbw     m1, m7
    pmullw        m0, m6
    pmullw        m1, m6
    paddw         m2, m0
    paddw         m3, m1          ; [mm2,mm3] += C * src[0..7]

    movq          m1, [r1+1]
    movq          m0, m1
    movq          m4, m1
    punpcklbw     m0, m7
    punpckhbw     m4, m7
    pmullw        m0, [rsp+8]
    pmullw        m4, [rsp+8]
    paddw         m2, m0
    paddw         m3, m4          ; [mm2,mm3] += D * src[1..8]
    movq          m0, [r1]

    paddw         m2, [rnd_2d_%2+rnd_bias*8]
    paddw         m3, [rnd_2d_%2+rnd_bias*8]
    psrlw         m2, 6
    psrlw         m3, 6
    packuswb      m2, m3
    CHROMAMC_AVG  m2, [dest_reg]
    movq  [dest_reg], m2          ; dst[0..7] = ([mm2,mm3] + rnd) >> 6

    add     dest_reg, r2
    dec          r3d
    jne .next2drow
    mov          rsp, r6          ; restore stack pointer
    RET
%endmacro

%macro chroma_mc4_mmx_func 3
cglobal %1_%2_chroma_mc4_%3, 6, 6, 0
%ifdef ARCH_X86_64
    movsxd        r2, r2d
%endif
    pxor          m7, m7
    movd          m2, r4d         ; x
    movd          m3, r5d         ; y
    movq          m4, [pw_8]
    movq          m5, [pw_8]
    punpcklwd     m2, m2
    punpcklwd     m3, m3
    punpcklwd     m2, m2
    punpcklwd     m3, m3
    psubw         m4, m2
    psubw         m5, m3

%ifidn %2, rv40
%ifdef PIC
   lea           r11, [rnd_rv40_2d_tbl]
%define rnd_2d_rv40 r11
%else
%define rnd_2d_rv40 rnd_rv40_2d_tbl
%endif
    and           r5, 6         ; &~1 for mx/my=[0,7]
    lea           r5, [r5*4+r4]
    sar          r5d, 1
%define rnd_bias r5
%else ; vc1, h264
%define rnd_bias 0
%endif

    movd          m0, [r1  ]
    movd          m6, [r1+1]
    add           r1, r2
    punpcklbw     m0, m7
    punpcklbw     m6, m7
    pmullw        m0, m4
    pmullw        m6, m2
    paddw         m6, m0

.next2rows
    movd          m0, [r1  ]
    movd          m1, [r1+1]
    add           r1, r2
    punpcklbw     m0, m7
    punpcklbw     m1, m7
    pmullw        m0, m4
    pmullw        m1, m2
    paddw         m1, m0
    movq          m0, m1

    pmullw        m6, m5
    pmullw        m1, m3
    paddw         m6, [rnd_2d_%2+rnd_bias*8]
    paddw         m1, m6
    psrlw         m1, 6
    packuswb      m1, m1
    CHROMAMC_AVG4 m1, m6, [r0]
    movd        [r0], m1
    add           r0, r2

    movd          m6, [r1  ]
    movd          m1, [r1+1]
    add           r1, r2
    punpcklbw     m6, m7
    punpcklbw     m1, m7
    pmullw        m6, m4
    pmullw        m1, m2
    paddw         m1, m6
    movq          m6, m1
    pmullw        m0, m5
    pmullw        m1, m3
    paddw         m0, [rnd_2d_%2+rnd_bias*8]
    paddw         m1, m0
    psrlw         m1, 6
    packuswb      m1, m1
    CHROMAMC_AVG4 m1, m0, [r0]
    movd        [r0], m1
    add           r0, r2
    sub          r3d, 2
    jnz .next2rows
    REP_RET
%endmacro

%macro chroma_mc2_mmx_func 3
cglobal %1_%2_chroma_mc2_%3, 6, 7, 0
%ifdef ARCH_X86_64
    movsxd        r2, r2d
%endif

    mov          r6d, r4d
    shl          r4d, 16
    sub          r4d, r6d
    add          r4d, 8
    imul         r5d, r4d         ; x*y<<16 | y*(8-x)
    shl          r4d, 3
    sub          r4d, r5d         ; x*(8-y)<<16 | (8-x)*(8-y)

    movd          m5, r4d
    movd          m6, r5d
    punpckldq     m5, m5          ; mm5 = {A,B,A,B}
    punpckldq     m6, m6          ; mm6 = {C,D,C,D}
    pxor          m7, m7
    movd          m2, [r1]
    punpcklbw     m2, m7
    pshufw        m2, m2, 0x94    ; mm0 = src[0,1,1,2]

.nextrow
    add           r1, r2
    movq          m1, m2
    pmaddwd       m1, m5          ; mm1 = A * src[0,1] + B * src[1,2]
    movd          m0, [r1]
    punpcklbw     m0, m7
    pshufw        m0, m0, 0x94    ; mm0 = src[0,1,1,2]
    movq          m2, m0
    pmaddwd       m0, m6
    paddw         m1, [rnd_2d_%2]
    paddw         m1, m0          ; mm1 += C * src[0,1] + D * src[1,2]
    psrlw         m1, 6
    packssdw      m1, m7
    packuswb      m1, m7
    CHROMAMC_AVG4 m1, m3, [r0]
    movd         r5d, m1
    mov         [r0], r5w
    add           r0, r2
    sub          r3d, 1
    jnz .nextrow
    REP_RET
%endmacro

%define rnd_1d_h264 pw_4
%define rnd_2d_h264 pw_32
%define rnd_1d_vc1  pw_3
%define rnd_2d_vc1  pw_28

%macro NOTHING 2-3
%endmacro
%macro DIRECT_AVG 2
    PAVG          %1, %2
%endmacro
%macro COPY_AVG 3
    movd          %2, %3
    PAVG          %1, %2
%endmacro

INIT_MMX
%define CHROMAMC_AVG  NOTHING
%define CHROMAMC_AVG4 NOTHING
chroma_mc8_mmx_func put, h264, mmx_rnd
chroma_mc8_mmx_func put, vc1,  mmx_nornd
chroma_mc8_mmx_func put, rv40, mmx
chroma_mc4_mmx_func put, h264, mmx
chroma_mc4_mmx_func put, rv40, mmx
chroma_mc2_mmx_func put, h264, mmx2

%define CHROMAMC_AVG  DIRECT_AVG
%define CHROMAMC_AVG4 COPY_AVG
%define PAVG          pavgb
chroma_mc8_mmx_func avg, h264, mmx2_rnd
chroma_mc8_mmx_func avg, vc1,  mmx2_nornd
chroma_mc8_mmx_func avg, rv40, mmx2
chroma_mc4_mmx_func avg, h264, mmx2
chroma_mc4_mmx_func avg, rv40, mmx2
chroma_mc2_mmx_func avg, h264, mmx2

%define PAVG          pavgusb
chroma_mc8_mmx_func avg, h264, 3dnow_rnd
chroma_mc8_mmx_func avg, vc1,  3dnow_nornd
chroma_mc8_mmx_func avg, rv40, 3dnow
chroma_mc4_mmx_func avg, h264, 3dnow
chroma_mc4_mmx_func avg, rv40, 3dnow

%macro chroma_mc8_ssse3_func 3
cglobal %1_%2_chroma_mc8_%3, 6, 7, 8
%ifdef ARCH_X86_64
    movsxd        r2, r2d
%endif
    mov          r6d, r5d
    or           r6d, r4d
    jne .at_least_one_non_zero
    ; mx == 0 AND my == 0 - no filter needed
    mv0_pixels_mc8
    REP_RET

.at_least_one_non_zero
    test         r5d, r5d
    je .my_is_zero
    test         r4d, r4d
    je .mx_is_zero

    ; general case, bilinear
    mov          r6d, r4d
    shl          r4d, 8
    sub           r4, r6
    add           r4, 8           ; x*288+8 = x<<8 | (8-x)
    mov           r6, 8
    sub          r6d, r5d
    imul          r6, r4          ; (8-y)*(x*255+8) = (8-y)*x<<8 | (8-y)*(8-x)
    imul         r4d, r5d         ;    y *(x*255+8) =    y *x<<8 |    y *(8-x)

    movd          m7, r6d
    movd          m6, r4d
    movdqa        m5, [rnd_2d_%2]
    pshuflw       m7, m7, 0
    pshuflw       m6, m6, 0
    movlhps       m7, m7
    movlhps       m6, m6

    movq          m0, [r1     ]
    movq          m1, [r1   +1]
    punpcklbw     m0, m1
    add           r1, r2
.next2rows
    movq          m1, [r1     ]
    movq          m2, [r1   +1]
    movq          m3, [r1+r2  ]
    movq          m4, [r1+r2+1]
    lea           r1, [r1+r2*2]
    punpcklbw     m1, m2
    punpcklbw     m3, m4
    movdqa        m2, m1
    movdqa        m4, m3
    pmaddubsw     m0, m7
    pmaddubsw     m1, m6
    pmaddubsw     m2, m7
    pmaddubsw     m3, m6
    paddw         m0, m5
    paddw         m2, m5
    paddw         m1, m0
    paddw         m3, m2
    movdqa        m0, m4
    psrlw         m1, 6
    psrlw         m3, 6
%ifidn %1, avg
    movq          m2, [r0   ]
    movhps        m2, [r0+r2]
%endif
    packuswb      m1, m3
    CHROMAMC_AVG  m1, m2
    movq     [r0   ], m1
    movhps   [r0+r2], m1
    sub          r3d, 2
    lea           r0, [r0+r2*2]
    jg .next2rows
    REP_RET

.my_is_zero
    mov          r5d, r4d
    shl          r4d, 8
    add           r4, 8
    sub           r4, r5          ; 255*x+8 = x<<8 | (8-x)
    movd          m7, r4d
    movdqa        m6, [rnd_1d_%2]
    pshuflw       m7, m7, 0
    movlhps       m7, m7

.next2xrows
    movq          m0, [r1     ]
    movq          m1, [r1   +1]
    movq          m2, [r1+r2  ]
    movq          m3, [r1+r2+1]
    punpcklbw     m0, m1
    punpcklbw     m2, m3
    pmaddubsw     m0, m7
    pmaddubsw     m2, m7
%ifidn %1, avg
    movq          m4, [r0   ]
    movhps        m4, [r0+r2]
%endif
    paddw         m0, m6
    paddw         m2, m6
    psrlw         m0, 3
    psrlw         m2, 3
    packuswb      m0, m2
    CHROMAMC_AVG  m0, m4
    movq     [r0   ], m0
    movhps   [r0+r2], m0
    sub          r3d, 2
    lea           r0, [r0+r2*2]
    lea           r1, [r1+r2*2]
    jg .next2xrows
    REP_RET

.mx_is_zero
    mov          r4d, r5d
    shl          r5d, 8
    add           r5, 8
    sub           r5, r4          ; 255*y+8 = y<<8 | (8-y)
    movd          m7, r5d
    movdqa        m6, [rnd_1d_%2]
    pshuflw       m7, m7, 0
    movlhps       m7, m7

.next2yrows
    movq          m0, [r1     ]
    movq          m1, [r1+r2  ]
    movdqa        m2, m1
    movq          m3, [r1+r2*2]
    punpcklbw     m0, m1
    punpcklbw     m2, m3
    pmaddubsw     m0, m7
    pmaddubsw     m2, m7
%ifidn %1, avg
    movq          m4, [r0   ]
    movhps        m4, [r0+r2]
%endif
    paddw         m0, m6
    paddw         m2, m6
    psrlw         m0, 3
    psrlw         m2, 3
    packuswb      m0, m2
    CHROMAMC_AVG  m0, m4
    movq     [r0   ], m0
    movhps   [r0+r2], m0
    sub          r3d, 2
    lea           r0, [r0+r2*2]
    lea           r1, [r1+r2*2]
    jg .next2yrows
    REP_RET
%endmacro

%macro chroma_mc4_ssse3_func 3
cglobal %1_%2_chroma_mc4_%3, 6, 7, 0
%ifdef ARCH_X86_64
    movsxd        r2, r2d
%endif
    mov           r6, r4
    shl          r4d, 8
    sub          r4d, r6d
    add          r4d, 8           ; x*288+8
    mov           r6, 8
    sub          r6d, r5d
    imul         r6d, r4d         ; (8-y)*(x*255+8) = (8-y)*x<<8 | (8-y)*(8-x)
    imul         r4d, r5d         ;    y *(x*255+8) =    y *x<<8 |    y *(8-x)

    movd          m7, r6d
    movd          m6, r4d
    movq          m5, [pw_32]
    pshufw        m7, m7, 0
    pshufw        m6, m6, 0

    movd          m0, [r1     ]
    punpcklbw     m0, [r1   +1]
    add           r1, r2
.next2rows
    movd          m1, [r1     ]
    movd          m3, [r1+r2  ]
    punpcklbw     m1, [r1   +1]
    punpcklbw     m3, [r1+r2+1]
    lea           r1, [r1+r2*2]
    movq          m2, m1
    movq          m4, m3
    pmaddubsw     m0, m7
    pmaddubsw     m1, m6
    pmaddubsw     m2, m7
    pmaddubsw     m3, m6
    paddw         m0, m5
    paddw         m2, m5
    paddw         m1, m0
    paddw         m3, m2
    movq          m0, m4
    psrlw         m1, 6
    psrlw         m3, 6
    packuswb      m1, m1
    packuswb      m3, m3
    CHROMAMC_AVG  m1, [r0  ]
    CHROMAMC_AVG  m3, [r0+r2]
    movd     [r0   ], m1
    movd     [r0+r2], m3
    sub          r3d, 2
    lea           r0, [r0+r2*2]
    jg .next2rows
    REP_RET
%endmacro

%define CHROMAMC_AVG NOTHING
INIT_XMM
chroma_mc8_ssse3_func put, h264, ssse3_rnd
chroma_mc8_ssse3_func put, vc1,  ssse3_nornd
INIT_MMX
chroma_mc4_ssse3_func put, h264, ssse3

%define CHROMAMC_AVG DIRECT_AVG
%define PAVG         pavgb
INIT_XMM
chroma_mc8_ssse3_func avg, h264, ssse3_rnd
chroma_mc8_ssse3_func avg, vc1,  ssse3_nornd
INIT_MMX
chroma_mc4_ssse3_func avg, h264, ssse3