-
Notifications
You must be signed in to change notification settings - Fork 232
/
Copy pathmc-a.asm
2226 lines (2108 loc) · 48.2 KB
/
mc-a.asm
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
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
948
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
;*****************************************************************************
;* mc-a.asm: x86 motion compensation
;*****************************************************************************
;* Copyright (C) 2003-2020 x264 project
;*
;* Authors: Loren Merritt <[email protected]>
;* Fiona Glaser <[email protected]>
;* Laurent Aimar <[email protected]>
;* Dylan Yudaken <[email protected]>
;* Holger Lubitz <[email protected]>
;* Min Chen <chenm001.163.com>
;* Oskar Arvidsson <[email protected]>
;*
;* This program is free software; you can redistribute it and/or modify
;* it under the terms of the GNU General Public License as published by
;* the Free Software Foundation; either version 2 of the License, or
;* (at your option) any later version.
;*
;* This program is distributed in the hope that it will be useful,
;* but WITHOUT ANY WARRANTY; without even the implied warranty of
;* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
;* GNU General Public License for more details.
;*
;* You should have received a copy of the GNU General Public License
;* along with this program; if not, write to the Free Software
;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02111, USA.
;*
;* This program is also available under a commercial proprietary license.
;* For more information, contact us at [email protected].
;*****************************************************************************
%include "x86inc.asm"
%include "x86util.asm"
SECTION_RODATA 32
ch_shuf: times 2 db 0,2,2,4,4,6,6,8,1,3,3,5,5,7,7,9
ch_shuf_adj: times 8 db 0
times 8 db 2
times 8 db 4
times 8 db 6
sq_1: times 1 dq 1
SECTION .text
cextern pb_0
cextern pw_1
cextern pw_4
cextern pw_8
cextern pw_32
cextern pw_64
cextern pw_512
cextern pw_00ff
cextern pw_pixel_max
cextern sw_64
cextern pd_32
cextern deinterleave_shufd
;=============================================================================
; implicit weighted biprediction
;=============================================================================
; assumes log2_denom = 5, offset = 0, weight1 + weight2 = 64
%if WIN64
DECLARE_REG_TMP 0,1,2,3,4,5,4,5
%macro AVG_START 0-1 0
PROLOGUE 6,7,%1
%endmacro
%elif UNIX64
DECLARE_REG_TMP 0,1,2,3,4,5,7,8
%macro AVG_START 0-1 0
PROLOGUE 6,9,%1
%endmacro
%else
DECLARE_REG_TMP 1,2,3,4,5,6,1,2
%macro AVG_START 0-1 0
PROLOGUE 0,7,%1
mov t0, r0m
mov t1, r1m
mov t2, r2m
mov t3, r3m
mov t4, r4m
mov t5, r5m
%endmacro
%endif
%macro AVG_END 0-1 2 ; rows
lea t2, [t2+t3*2*SIZEOF_PIXEL]
lea t4, [t4+t5*2*SIZEOF_PIXEL]
lea t0, [t0+t1*2*SIZEOF_PIXEL]
sub eax, %1
jg .height_loop
RET
%endmacro
%if HIGH_BIT_DEPTH
%macro BIWEIGHT_MMX 2
movh m0, %1
movh m1, %2
punpcklwd m0, m1
pmaddwd m0, m3
paddd m0, m4
psrad m0, 6
%endmacro
%macro BIWEIGHT_START_MMX 0
movzx t6d, word r6m
mov t7d, 64
sub t7d, t6d
shl t7d, 16
add t6d, t7d
movd m3, t6d
SPLATD m3, m3
mova m4, [pd_32]
pxor m5, m5
%endmacro
%else ;!HIGH_BIT_DEPTH
%macro BIWEIGHT_MMX 2
movh m0, %1
movh m1, %2
punpcklbw m0, m5
punpcklbw m1, m5
pmullw m0, m2
pmullw m1, m3
paddw m0, m1
paddw m0, m4
psraw m0, 6
%endmacro
%macro BIWEIGHT_START_MMX 0
movd m2, r6m
SPLATW m2, m2 ; weight_dst
mova m3, [pw_64]
psubw m3, m2 ; weight_src
mova m4, [pw_32] ; rounding
pxor m5, m5
%endmacro
%endif ;HIGH_BIT_DEPTH
%macro BIWEIGHT_SSSE3 2
movh m0, %1
movh m1, %2
punpcklbw m0, m1
pmaddubsw m0, m3
pmulhrsw m0, m4
%endmacro
%macro BIWEIGHT_START_SSSE3 0
movzx t6d, byte r6m ; FIXME x86_64
%if mmsize > 16
vbroadcasti128 m4, [pw_512]
%else
mova m4, [pw_512]
%endif
lea t7d, [t6+(64<<8)]
shl t6d, 8
sub t7d, t6d
%if cpuflag(avx512)
vpbroadcastw m3, t7d
%else
movd xm3, t7d
%if cpuflag(avx2)
vpbroadcastw m3, xm3
%else
SPLATW m3, m3 ; weight_dst,src
%endif
%endif
%endmacro
%if HIGH_BIT_DEPTH
%macro BIWEIGHT_ROW 4
BIWEIGHT [%2], [%3]
%if %4==mmsize/4
packssdw m0, m0
CLIPW m0, m5, m7
movh [%1], m0
%else
SWAP 0, 6
BIWEIGHT [%2+mmsize/2], [%3+mmsize/2]
packssdw m6, m0
CLIPW m6, m5, m7
mova [%1], m6
%endif
%endmacro
%else ;!HIGH_BIT_DEPTH
%macro BIWEIGHT_ROW 4
BIWEIGHT [%2], [%3]
%if %4==mmsize/2
packuswb m0, m0
movh [%1], m0
%else
SWAP 0, 6
BIWEIGHT [%2+mmsize/2], [%3+mmsize/2]
packuswb m6, m0
mova [%1], m6
%endif
%endmacro
%endif ;HIGH_BIT_DEPTH
;-----------------------------------------------------------------------------
; int pixel_avg_weight_w16( pixel *dst, intptr_t, pixel *src1, intptr_t, pixel *src2, intptr_t, int i_weight )
;-----------------------------------------------------------------------------
%macro AVG_WEIGHT 1-2 0
cglobal pixel_avg_weight_w%1
BIWEIGHT_START
AVG_START %2
%if HIGH_BIT_DEPTH
mova m7, [pw_pixel_max]
%endif
.height_loop:
%if mmsize==16 && %1==mmsize/(2*SIZEOF_PIXEL)
BIWEIGHT [t2], [t4]
SWAP 0, 6
BIWEIGHT [t2+SIZEOF_PIXEL*t3], [t4+SIZEOF_PIXEL*t5]
%if HIGH_BIT_DEPTH
packssdw m6, m0
CLIPW m6, m5, m7
%else ;!HIGH_BIT_DEPTH
packuswb m6, m0
%endif ;HIGH_BIT_DEPTH
movlps [t0], m6
movhps [t0+SIZEOF_PIXEL*t1], m6
%else
%assign x 0
%rep (%1*SIZEOF_PIXEL+mmsize-1)/mmsize
BIWEIGHT_ROW t0+x, t2+x, t4+x, %1
BIWEIGHT_ROW t0+x+SIZEOF_PIXEL*t1, t2+x+SIZEOF_PIXEL*t3, t4+x+SIZEOF_PIXEL*t5, %1
%assign x x+mmsize
%endrep
%endif
AVG_END
%endmacro
%define BIWEIGHT BIWEIGHT_MMX
%define BIWEIGHT_START BIWEIGHT_START_MMX
INIT_MMX mmx2
AVG_WEIGHT 4
AVG_WEIGHT 8
AVG_WEIGHT 16
%if HIGH_BIT_DEPTH
INIT_XMM sse2
AVG_WEIGHT 4, 8
AVG_WEIGHT 8, 8
AVG_WEIGHT 16, 8
%else ;!HIGH_BIT_DEPTH
INIT_XMM sse2
AVG_WEIGHT 8, 7
AVG_WEIGHT 16, 7
%define BIWEIGHT BIWEIGHT_SSSE3
%define BIWEIGHT_START BIWEIGHT_START_SSSE3
INIT_MMX ssse3
AVG_WEIGHT 4
INIT_XMM ssse3
AVG_WEIGHT 8, 7
AVG_WEIGHT 16, 7
INIT_YMM avx2
cglobal pixel_avg_weight_w16
BIWEIGHT_START
AVG_START 5
.height_loop:
movu xm0, [t2]
movu xm1, [t4]
vinserti128 m0, m0, [t2+t3], 1
vinserti128 m1, m1, [t4+t5], 1
SBUTTERFLY bw, 0, 1, 2
pmaddubsw m0, m3
pmaddubsw m1, m3
pmulhrsw m0, m4
pmulhrsw m1, m4
packuswb m0, m1
mova [t0], xm0
vextracti128 [t0+t1], m0, 1
AVG_END
INIT_YMM avx512
cglobal pixel_avg_weight_w8
BIWEIGHT_START
kxnorb k1, k1, k1
kaddb k1, k1, k1
AVG_START 5
.height_loop:
movq xm0, [t2]
movq xm2, [t4]
movq xm1, [t2+t3]
movq xm5, [t4+t5]
lea t2, [t2+t3*2]
lea t4, [t4+t5*2]
vpbroadcastq m0 {k1}, [t2]
vpbroadcastq m2 {k1}, [t4]
vpbroadcastq m1 {k1}, [t2+t3]
vpbroadcastq m5 {k1}, [t4+t5]
punpcklbw m0, m2
punpcklbw m1, m5
pmaddubsw m0, m3
pmaddubsw m1, m3
pmulhrsw m0, m4
pmulhrsw m1, m4
packuswb m0, m1
vextracti128 xmm1, m0, 1
movq [t0], xm0
movhps [t0+t1], xm0
lea t0, [t0+t1*2]
movq [t0], xmm1
movhps [t0+t1], xmm1
AVG_END 4
INIT_ZMM avx512
cglobal pixel_avg_weight_w16
BIWEIGHT_START
AVG_START 5
.height_loop:
movu xm0, [t2]
movu xm1, [t4]
vinserti128 ym0, [t2+t3], 1
vinserti128 ym1, [t4+t5], 1
lea t2, [t2+t3*2]
lea t4, [t4+t5*2]
vinserti32x4 m0, [t2], 2
vinserti32x4 m1, [t4], 2
vinserti32x4 m0, [t2+t3], 3
vinserti32x4 m1, [t4+t5], 3
SBUTTERFLY bw, 0, 1, 2
pmaddubsw m0, m3
pmaddubsw m1, m3
pmulhrsw m0, m4
pmulhrsw m1, m4
packuswb m0, m1
mova [t0], xm0
vextracti128 [t0+t1], ym0, 1
lea t0, [t0+t1*2]
vextracti32x4 [t0], m0, 2
vextracti32x4 [t0+t1], m0, 3
AVG_END 4
%endif ;HIGH_BIT_DEPTH
;=============================================================================
; P frame explicit weighted prediction
;=============================================================================
%if HIGH_BIT_DEPTH
; width
%macro WEIGHT_START 1
mova m0, [r4+ 0] ; 1<<denom
mova m3, [r4+16]
movd m2, [r4+32] ; denom
mova m4, [pw_pixel_max]
paddw m2, [sq_1] ; denom+1
%endmacro
; src1, src2
%macro WEIGHT 2
movh m5, [%1]
movh m6, [%2]
punpcklwd m5, m0
punpcklwd m6, m0
pmaddwd m5, m3
pmaddwd m6, m3
psrad m5, m2
psrad m6, m2
packssdw m5, m6
%endmacro
; src, dst, width
%macro WEIGHT_TWO_ROW 4
%assign x 0
%rep (%3+mmsize/2-1)/(mmsize/2)
%if %3-x/2 <= 4 && mmsize == 16
WEIGHT %1+x, %1+r3+x
CLIPW m5, [pb_0], m4
movh [%2+x], m5
movhps [%2+r1+x], m5
%else
WEIGHT %1+x, %1+x+mmsize/2
SWAP 5, 7
WEIGHT %1+r3+x, %1+r3+x+mmsize/2
CLIPW m5, [pb_0], m4
CLIPW m7, [pb_0], m4
mova [%2+x], m7
mova [%2+r1+x], m5
%endif
%assign x x+mmsize
%endrep
%endmacro
%else ; !HIGH_BIT_DEPTH
%macro WEIGHT_START 1
%if cpuflag(avx2)
vbroadcasti128 m3, [r4]
vbroadcasti128 m4, [r4+16]
%else
mova m3, [r4]
mova m4, [r4+16]
%if notcpuflag(ssse3)
movd m5, [r4+32]
%endif
%endif
pxor m2, m2
%endmacro
; src1, src2, dst1, dst2, fast
%macro WEIGHT_ROWx2 5
movh m0, [%1 ]
movh m1, [%1+mmsize/2]
movh m6, [%2 ]
movh m7, [%2+mmsize/2]
punpcklbw m0, m2
punpcklbw m1, m2
punpcklbw m6, m2
punpcklbw m7, m2
%if cpuflag(ssse3)
%if %5==0
psllw m0, 7
psllw m1, 7
psllw m6, 7
psllw m7, 7
%endif
pmulhrsw m0, m3
pmulhrsw m1, m3
pmulhrsw m6, m3
pmulhrsw m7, m3
paddw m0, m4
paddw m1, m4
paddw m6, m4
paddw m7, m4
%else
pmullw m0, m3
pmullw m1, m3
pmullw m6, m3
pmullw m7, m3
paddsw m0, m4 ;1<<(denom-1)+(offset<<denom)
paddsw m1, m4
paddsw m6, m4
paddsw m7, m4
psraw m0, m5
psraw m1, m5
psraw m6, m5
psraw m7, m5
%endif
packuswb m0, m1
packuswb m6, m7
mova [%3], m0
mova [%4], m6
%endmacro
; src1, src2, dst1, dst2, width, fast
%macro WEIGHT_COL 6
%if cpuflag(avx2)
%if %5==16
movu xm0, [%1]
vinserti128 m0, m0, [%2], 1
punpckhbw m1, m0, m2
punpcklbw m0, m0, m2
%if %6==0
psllw m0, 7
psllw m1, 7
%endif
pmulhrsw m0, m3
pmulhrsw m1, m3
paddw m0, m4
paddw m1, m4
packuswb m0, m1
mova [%3], xm0
vextracti128 [%4], m0, 1
%else
movq xm0, [%1]
vinserti128 m0, m0, [%2], 1
punpcklbw m0, m2
%if %6==0
psllw m0, 7
%endif
pmulhrsw m0, m3
paddw m0, m4
packuswb m0, m0
vextracti128 xm1, m0, 1
%if %5 == 8
movq [%3], xm0
movq [%4], xm1
%else
movd [%3], xm0
movd [%4], xm1
%endif
%endif
%else
movh m0, [%1]
movh m1, [%2]
punpcklbw m0, m2
punpcklbw m1, m2
%if cpuflag(ssse3)
%if %6==0
psllw m0, 7
psllw m1, 7
%endif
pmulhrsw m0, m3
pmulhrsw m1, m3
paddw m0, m4
paddw m1, m4
%else
pmullw m0, m3
pmullw m1, m3
paddsw m0, m4 ;1<<(denom-1)+(offset<<denom)
paddsw m1, m4
psraw m0, m5
psraw m1, m5
%endif
%if %5 == 8
packuswb m0, m1
movh [%3], m0
movhps [%4], m0
%else
packuswb m0, m0
packuswb m1, m1
movd [%3], m0 ; width 2 can write garbage for the last 2 bytes
movd [%4], m1
%endif
%endif
%endmacro
; src, dst, width
%macro WEIGHT_TWO_ROW 4
%assign x 0
%rep %3
%if (%3-x) >= mmsize
WEIGHT_ROWx2 %1+x, %1+r3+x, %2+x, %2+r1+x, %4
%assign x (x+mmsize)
%else
%assign w %3-x
%if w == 20
%assign w 16
%endif
WEIGHT_COL %1+x, %1+r3+x, %2+x, %2+r1+x, w, %4
%assign x (x+w)
%endif
%if x >= %3
%exitrep
%endif
%endrep
%endmacro
%endif ; HIGH_BIT_DEPTH
;-----------------------------------------------------------------------------
;void mc_weight_wX( pixel *dst, intptr_t i_dst_stride, pixel *src, intptr_t i_src_stride, weight_t *weight, int h )
;-----------------------------------------------------------------------------
%macro WEIGHTER 1
cglobal mc_weight_w%1, 6,6,8
FIX_STRIDES r1, r3
WEIGHT_START %1
%if cpuflag(ssse3) && HIGH_BIT_DEPTH == 0
; we can merge the shift step into the scale factor
; if (m3<<7) doesn't overflow an int16_t
cmp byte [r4+1], 0
jz .fast
%endif
.loop:
WEIGHT_TWO_ROW r2, r0, %1, 0
lea r0, [r0+r1*2]
lea r2, [r2+r3*2]
sub r5d, 2
jg .loop
RET
%if cpuflag(ssse3) && HIGH_BIT_DEPTH == 0
.fast:
psllw m3, 7
.fastloop:
WEIGHT_TWO_ROW r2, r0, %1, 1
lea r0, [r0+r1*2]
lea r2, [r2+r3*2]
sub r5d, 2
jg .fastloop
RET
%endif
%endmacro
INIT_MMX mmx2
WEIGHTER 4
WEIGHTER 8
WEIGHTER 12
WEIGHTER 16
WEIGHTER 20
INIT_XMM sse2
WEIGHTER 8
WEIGHTER 16
WEIGHTER 20
%if HIGH_BIT_DEPTH
WEIGHTER 12
%else
INIT_MMX ssse3
WEIGHTER 4
INIT_XMM ssse3
WEIGHTER 8
WEIGHTER 16
WEIGHTER 20
INIT_YMM avx2
WEIGHTER 8
WEIGHTER 16
WEIGHTER 20
%endif
%macro OFFSET_OP 7
mov%6 m0, [%1]
mov%6 m1, [%2]
%if HIGH_BIT_DEPTH
p%5usw m0, m2
p%5usw m1, m2
%ifidn %5,add
pminsw m0, m3
pminsw m1, m3
%endif
%else
p%5usb m0, m2
p%5usb m1, m2
%endif
mov%7 [%3], m0
mov%7 [%4], m1
%endmacro
%macro OFFSET_TWO_ROW 4
%assign x 0
%rep %3
%if (%3*SIZEOF_PIXEL-x) >= mmsize
OFFSET_OP (%1+x), (%1+x+r3), (%2+x), (%2+x+r1), %4, u, a
%assign x (x+mmsize)
%else
%if HIGH_BIT_DEPTH
OFFSET_OP (%1+x), (%1+x+r3), (%2+x), (%2+x+r1), %4, h, h
%else
OFFSET_OP (%1+x), (%1+x+r3), (%2+x), (%2+x+r1), %4, d, d
%endif
%exitrep
%endif
%if x >= %3*SIZEOF_PIXEL
%exitrep
%endif
%endrep
%endmacro
;-----------------------------------------------------------------------------
;void mc_offset_wX( pixel *src, intptr_t i_src_stride, pixel *dst, intptr_t i_dst_stride, weight_t *w, int h )
;-----------------------------------------------------------------------------
%macro OFFSET 2
cglobal mc_offset%2_w%1, 6,6
FIX_STRIDES r1, r3
mova m2, [r4]
%if HIGH_BIT_DEPTH
%ifidn %2,add
mova m3, [pw_pixel_max]
%endif
%endif
.loop:
OFFSET_TWO_ROW r2, r0, %1, %2
lea r0, [r0+r1*2]
lea r2, [r2+r3*2]
sub r5d, 2
jg .loop
RET
%endmacro
%macro OFFSETPN 1
OFFSET %1, add
OFFSET %1, sub
%endmacro
INIT_MMX mmx2
OFFSETPN 4
OFFSETPN 8
OFFSETPN 12
OFFSETPN 16
OFFSETPN 20
INIT_XMM sse2
OFFSETPN 12
OFFSETPN 16
OFFSETPN 20
%if HIGH_BIT_DEPTH
INIT_XMM sse2
OFFSETPN 8
%endif
;=============================================================================
; pixel avg
;=============================================================================
;-----------------------------------------------------------------------------
; void pixel_avg_4x4( pixel *dst, intptr_t dst_stride, pixel *src1, intptr_t src1_stride,
; pixel *src2, intptr_t src2_stride, int weight );
;-----------------------------------------------------------------------------
%macro AVGH 2
cglobal pixel_avg_%1x%2
mov eax, %2
cmp dword r6m, 32
jne pixel_avg_weight_w%1 %+ SUFFIX
%if cpuflag(avx2) && %1 == 16 ; all AVX2 machines can do fast 16-byte unaligned loads
jmp pixel_avg_w%1_avx2
%else
%if mmsize == 16 && %1 == 16
test dword r4m, 15
jz pixel_avg_w%1_sse2
%endif
jmp pixel_avg_w%1_mmx2
%endif
%endmacro
;-----------------------------------------------------------------------------
; void pixel_avg_w4( pixel *dst, intptr_t dst_stride, pixel *src1, intptr_t src1_stride,
; pixel *src2, intptr_t src2_stride, int height, int weight );
;-----------------------------------------------------------------------------
%macro AVG_FUNC 3
cglobal pixel_avg_w%1
AVG_START
.height_loop:
%assign x 0
%rep (%1*SIZEOF_PIXEL+mmsize-1)/mmsize
%2 m0, [t2+x]
%2 m1, [t2+x+SIZEOF_PIXEL*t3]
%if HIGH_BIT_DEPTH
pavgw m0, [t4+x]
pavgw m1, [t4+x+SIZEOF_PIXEL*t5]
%else ;!HIGH_BIT_DEPTH
pavgb m0, [t4+x]
pavgb m1, [t4+x+SIZEOF_PIXEL*t5]
%endif
%3 [t0+x], m0
%3 [t0+x+SIZEOF_PIXEL*t1], m1
%assign x x+mmsize
%endrep
AVG_END
%endmacro
%if HIGH_BIT_DEPTH
INIT_MMX mmx2
AVG_FUNC 4, movq, movq
AVGH 4, 16
AVGH 4, 8
AVGH 4, 4
AVGH 4, 2
AVG_FUNC 8, movq, movq
AVGH 8, 16
AVGH 8, 8
AVGH 8, 4
AVG_FUNC 16, movq, movq
AVGH 16, 16
AVGH 16, 8
INIT_XMM sse2
AVG_FUNC 4, movq, movq
AVGH 4, 16
AVGH 4, 8
AVGH 4, 4
AVGH 4, 2
AVG_FUNC 8, movdqu, movdqa
AVGH 8, 16
AVGH 8, 8
AVGH 8, 4
AVG_FUNC 16, movdqu, movdqa
AVGH 16, 16
AVGH 16, 8
%else ;!HIGH_BIT_DEPTH
INIT_MMX mmx2
AVG_FUNC 4, movd, movd
AVGH 4, 16
AVGH 4, 8
AVGH 4, 4
AVGH 4, 2
AVG_FUNC 8, movq, movq
AVGH 8, 16
AVGH 8, 8
AVGH 8, 4
AVG_FUNC 16, movq, movq
AVGH 16, 16
AVGH 16, 8
INIT_XMM sse2
AVG_FUNC 16, movdqu, movdqa
AVGH 16, 16
AVGH 16, 8
AVGH 8, 16
AVGH 8, 8
AVGH 8, 4
INIT_XMM ssse3
AVGH 16, 16
AVGH 16, 8
AVGH 8, 16
AVGH 8, 8
AVGH 8, 4
INIT_MMX ssse3
AVGH 4, 16
AVGH 4, 8
AVGH 4, 4
AVGH 4, 2
INIT_XMM avx2
AVG_FUNC 16, movdqu, movdqa
AVGH 16, 16
AVGH 16, 8
INIT_XMM avx512
AVGH 16, 16
AVGH 16, 8
AVGH 8, 16
AVGH 8, 8
AVGH 8, 4
%endif ;HIGH_BIT_DEPTH
;=============================================================================
; pixel avg2
;=============================================================================
%if HIGH_BIT_DEPTH
;-----------------------------------------------------------------------------
; void pixel_avg2_wN( uint16_t *dst, intptr_t dst_stride,
; uint16_t *src1, intptr_t src_stride,
; uint16_t *src2, int height );
;-----------------------------------------------------------------------------
%macro AVG2_W_ONE 1
cglobal pixel_avg2_w%1, 6,7,4
sub r4, r2
lea r6, [r4+r3*2]
.height_loop:
movu m0, [r2]
movu m1, [r2+r3*2]
%if cpuflag(avx) || mmsize == 8
pavgw m0, [r2+r4]
pavgw m1, [r2+r6]
%else
movu m2, [r2+r4]
movu m3, [r2+r6]
pavgw m0, m2
pavgw m1, m3
%endif
mova [r0], m0
mova [r0+r1*2], m1
lea r2, [r2+r3*4]
lea r0, [r0+r1*4]
sub r5d, 2
jg .height_loop
RET
%endmacro
%macro AVG2_W_TWO 3
cglobal pixel_avg2_w%1, 6,7,8
sub r4, r2
lea r6, [r4+r3*2]
.height_loop:
movu m0, [r2]
%2 m1, [r2+mmsize]
movu m2, [r2+r3*2]
%2 m3, [r2+r3*2+mmsize]
%if mmsize == 8
pavgw m0, [r2+r4]
pavgw m1, [r2+r4+mmsize]
pavgw m2, [r2+r6]
pavgw m3, [r2+r6+mmsize]
%else
movu m4, [r2+r4]
%2 m5, [r2+r4+mmsize]
movu m6, [r2+r6]
%2 m7, [r2+r6+mmsize]
pavgw m0, m4
pavgw m1, m5
pavgw m2, m6
pavgw m3, m7
%endif
mova [r0], m0
%3 [r0+mmsize], m1
mova [r0+r1*2], m2
%3 [r0+r1*2+mmsize], m3
lea r2, [r2+r3*4]
lea r0, [r0+r1*4]
sub r5d, 2
jg .height_loop
RET
%endmacro
INIT_MMX mmx2
AVG2_W_ONE 4
AVG2_W_TWO 8, movu, mova
INIT_XMM sse2
AVG2_W_ONE 8
AVG2_W_TWO 10, movd, movd
AVG2_W_TWO 16, movu, mova
INIT_YMM avx2
AVG2_W_ONE 16
INIT_MMX
cglobal pixel_avg2_w10_mmx2, 6,7
sub r4, r2
lea r6, [r4+r3*2]
.height_loop:
movu m0, [r2+ 0]
movu m1, [r2+ 8]
movh m2, [r2+16]
movu m3, [r2+r3*2+ 0]
movu m4, [r2+r3*2+ 8]
movh m5, [r2+r3*2+16]
pavgw m0, [r2+r4+ 0]
pavgw m1, [r2+r4+ 8]
pavgw m2, [r2+r4+16]
pavgw m3, [r2+r6+ 0]
pavgw m4, [r2+r6+ 8]
pavgw m5, [r2+r6+16]
mova [r0+ 0], m0
mova [r0+ 8], m1
movh [r0+16], m2
mova [r0+r1*2+ 0], m3
mova [r0+r1*2+ 8], m4
movh [r0+r1*2+16], m5
lea r2, [r2+r3*2*2]
lea r0, [r0+r1*2*2]
sub r5d, 2
jg .height_loop
RET
cglobal pixel_avg2_w16_mmx2, 6,7
sub r4, r2
lea r6, [r4+r3*2]
.height_loop:
movu m0, [r2+ 0]
movu m1, [r2+ 8]
movu m2, [r2+16]
movu m3, [r2+24]
movu m4, [r2+r3*2+ 0]
movu m5, [r2+r3*2+ 8]
movu m6, [r2+r3*2+16]
movu m7, [r2+r3*2+24]
pavgw m0, [r2+r4+ 0]
pavgw m1, [r2+r4+ 8]
pavgw m2, [r2+r4+16]
pavgw m3, [r2+r4+24]
pavgw m4, [r2+r6+ 0]
pavgw m5, [r2+r6+ 8]
pavgw m6, [r2+r6+16]
pavgw m7, [r2+r6+24]
mova [r0+ 0], m0
mova [r0+ 8], m1
mova [r0+16], m2
mova [r0+24], m3
mova [r0+r1*2+ 0], m4
mova [r0+r1*2+ 8], m5
mova [r0+r1*2+16], m6
mova [r0+r1*2+24], m7
lea r2, [r2+r3*2*2]
lea r0, [r0+r1*2*2]
sub r5d, 2
jg .height_loop
RET
cglobal pixel_avg2_w18_mmx2, 6,7
sub r4, r2
.height_loop:
movu m0, [r2+ 0]
movu m1, [r2+ 8]
movu m2, [r2+16]
movu m3, [r2+24]
movh m4, [r2+32]
pavgw m0, [r2+r4+ 0]
pavgw m1, [r2+r4+ 8]
pavgw m2, [r2+r4+16]
pavgw m3, [r2+r4+24]
pavgw m4, [r2+r4+32]
mova [r0+ 0], m0
mova [r0+ 8], m1
mova [r0+16], m2
mova [r0+24], m3
movh [r0+32], m4
lea r2, [r2+r3*2]
lea r0, [r0+r1*2]
dec r5d
jg .height_loop
RET
%macro PIXEL_AVG_W18 0
cglobal pixel_avg2_w18, 6,7
sub r4, r2
.height_loop:
movu m0, [r2+ 0]
movd xm2, [r2+32]
%if mmsize == 32
pavgw m0, [r2+r4+ 0]
movd xm1, [r2+r4+32]
pavgw xm2, xm1
%else
movu m1, [r2+16]
movu m3, [r2+r4+ 0]
movu m4, [r2+r4+16]
movd m5, [r2+r4+32]