summaryrefslogtreecommitdiff
path: root/libavcodec/x86/h264_deblock_10bit.asm
blob: d63ca02c7a2525fe2eb687ffe9ae123cb52723bb (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
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
;*****************************************************************************
;* MMX/SSE2/AVX-optimized 10-bit H.264 deblocking code
;*****************************************************************************
;* Copyright (C) 2005-2011 x264 project
;*
;* Authors: Oskar Arvidsson <oskar@irock.se>
;*          Loren Merritt <lorenm@u.washington.edu>
;*          Jason Garrett-Glaser <darkshikari@gmail.com>
;*
;* 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
;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
;******************************************************************************

%include "libavutil/x86/x86util.asm"

SECTION_RODATA

pw_pixel_max: times 8 dw ((1 << 10)-1)

SECTION .text

cextern pw_2
cextern pw_3
cextern pw_4

; out: %4 = |%1-%2|-%3
; clobbers: %5
%macro ABS_SUB 5
    psubusw %5, %2, %1
    psubusw %4, %1, %2
    por     %4, %5
    psubw   %4, %3
%endmacro

; out: %4 = |%1-%2|<%3
%macro DIFF_LT   5
    psubusw %4, %2, %1
    psubusw %5, %1, %2
    por     %5, %4 ; |%1-%2|
    pxor    %4, %4
    psubw   %5, %3 ; |%1-%2|-%3
    pcmpgtw %4, %5 ; 0 > |%1-%2|-%3
%endmacro

%macro LOAD_AB 4
    movd       %1, %3
    movd       %2, %4
    SPLATW     %1, %1
    SPLATW     %2, %2
%endmacro

; in:  %2=tc reg
; out: %1=splatted tc
%macro LOAD_TC 2
    movd        %1, [%2]
    punpcklbw   %1, %1
%if mmsize == 8
    pshufw      %1, %1, 0
%else
    pshuflw     %1, %1, 01010000b
    pshufd      %1, %1, 01010000b
%endif
    psraw       %1, 6
%endmacro

; in: %1=p1, %2=p0, %3=q0, %4=q1
;     %5=alpha, %6=beta, %7-%9=tmp
; out: %7=mask
%macro LOAD_MASK 9
    ABS_SUB     %2, %3, %5, %8, %7 ; |p0-q0| - alpha
    ABS_SUB     %1, %2, %6, %9, %7 ; |p1-p0| - beta
    pand        %8, %9
    ABS_SUB     %3, %4, %6, %9, %7 ; |q1-q0| - beta
    pxor        %7, %7
    pand        %8, %9
    pcmpgtw     %7, %8
%endmacro

; in: %1=p0, %2=q0, %3=p1, %4=q1, %5=mask, %6=tmp, %7=tmp
; out: %1=p0', m2=q0'
%macro DEBLOCK_P0_Q0 7
    psubw   %3, %4
    pxor    %7, %7
    paddw   %3, [pw_4]
    psubw   %7, %5
    psubw   %6, %2, %1
    psllw   %6, 2
    paddw   %3, %6
    psraw   %3, 3
    mova    %6, [pw_pixel_max]
    CLIPW   %3, %7, %5
    pxor    %7, %7
    paddw   %1, %3
    psubw   %2, %3
    CLIPW   %1, %7, %6
    CLIPW   %2, %7, %6
%endmacro

; in: %1=x2, %2=x1, %3=p0, %4=q0 %5=mask&tc, %6=tmp
%macro LUMA_Q1 6
    pavgw       %6, %3, %4      ; (p0+q0+1)>>1
    paddw       %1, %6
    pxor        %6, %6
    psraw       %1, 1
    psubw       %6, %5
    psubw       %1, %2
    CLIPW       %1, %6, %5
    paddw       %1, %2
%endmacro

%macro LUMA_DEBLOCK_ONE 3
    DIFF_LT     m5, %1, bm, m4, m6
    pxor        m6, m6
    mova        %3, m4
    pcmpgtw     m6, tcm
    pand        m4, tcm
    pandn       m6, m7
    pand        m4, m6
    LUMA_Q1 m5, %2, m1, m2, m4, m6
%endmacro

%macro LUMA_H_STORE 2
%if mmsize == 8
    movq        [r0-4], m0
    movq        [r0+r1-4], m1
    movq        [r0+r1*2-4], m2
    movq        [r0+%2-4], m3
%else
    movq        [r0-4], m0
    movhps      [r0+r1-4], m0
    movq        [r0+r1*2-4], m1
    movhps      [%1-4], m1
    movq        [%1+r1-4], m2
    movhps      [%1+r1*2-4], m2
    movq        [%1+%2-4], m3
    movhps      [%1+r1*4-4], m3
%endif
%endmacro

%macro DEBLOCK_LUMA 0
;-----------------------------------------------------------------------------
; void deblock_v_luma( uint16_t *pix, int stride, int alpha, int beta, int8_t *tc0 )
;-----------------------------------------------------------------------------
cglobal deblock_v_luma_10, 5,5,8*(mmsize/16)
    %assign pad 5*mmsize+12-(stack_offset&15)
    %define tcm [rsp]
    %define ms1 [rsp+mmsize]
    %define ms2 [rsp+mmsize*2]
    %define am  [rsp+mmsize*3]
    %define bm  [rsp+mmsize*4]
    SUB        rsp, pad
    shl        r2d, 2
    shl        r3d, 2
    LOAD_AB     m4, m5, r2d, r3d
    mov         r3, 32/mmsize
    mov         r2, r0
    sub         r0, r1
    mova        am, m4
    sub         r0, r1
    mova        bm, m5
    sub         r0, r1
.loop:
    mova        m0, [r0+r1]
    mova        m1, [r0+r1*2]
    mova        m2, [r2]
    mova        m3, [r2+r1]

    LOAD_MASK   m0, m1, m2, m3, am, bm, m7, m4, m6
    LOAD_TC     m6, r4
    mova       tcm, m6

    mova        m5, [r0]
    LUMA_DEBLOCK_ONE m1, m0, ms1
    mova   [r0+r1], m5

    mova        m5, [r2+r1*2]
    LUMA_DEBLOCK_ONE m2, m3, ms2
    mova   [r2+r1], m5

    pxor        m5, m5
    mova        m6, tcm
    pcmpgtw     m5, tcm
    psubw       m6, ms1
    pandn       m5, m7
    psubw       m6, ms2
    pand        m5, m6
    DEBLOCK_P0_Q0 m1, m2, m0, m3, m5, m7, m6
    mova [r0+r1*2], m1
    mova      [r2], m2

    add         r0, mmsize
    add         r2, mmsize
    add         r4, mmsize/8
    dec         r3
    jg .loop
    ADD         rsp, pad
    RET

cglobal deblock_h_luma_10, 5,6,8*(mmsize/16)
    %assign pad 7*mmsize+12-(stack_offset&15)
    %define tcm [rsp]
    %define ms1 [rsp+mmsize]
    %define ms2 [rsp+mmsize*2]
    %define p1m [rsp+mmsize*3]
    %define p2m [rsp+mmsize*4]
    %define am  [rsp+mmsize*5]
    %define bm  [rsp+mmsize*6]
    SUB        rsp, pad
    shl        r2d, 2
    shl        r3d, 2
    LOAD_AB     m4, m5, r2d, r3d
    mov         r3, r1
    mova        am, m4
    add         r3, r1
    mov         r5, 32/mmsize
    mova        bm, m5
    add         r3, r1
%if mmsize == 16
    mov         r2, r0
    add         r2, r3
%endif
.loop:
%if mmsize == 8
    movq        m2, [r0-8]     ; y q2 q1 q0
    movq        m7, [r0+0]
    movq        m5, [r0+r1-8]
    movq        m3, [r0+r1+0]
    movq        m0, [r0+r1*2-8]
    movq        m6, [r0+r1*2+0]
    movq        m1, [r0+r3-8]
    TRANSPOSE4x4W 2, 5, 0, 1, 4
    SWAP         2, 7
    movq        m7, [r0+r3]
    TRANSPOSE4x4W 2, 3, 6, 7, 4
%else
    movu        m5, [r0-8]     ; y q2 q1 q0 p0 p1 p2 x
    movu        m0, [r0+r1-8]
    movu        m2, [r0+r1*2-8]
    movu        m3, [r2-8]
    TRANSPOSE4x4W 5, 0, 2, 3, 6
    mova       tcm, m3

    movu        m4, [r2+r1-8]
    movu        m1, [r2+r1*2-8]
    movu        m3, [r2+r3-8]
    movu        m7, [r2+r1*4-8]
    TRANSPOSE4x4W 4, 1, 3, 7, 6

    mova        m6, tcm
    punpcklqdq  m6, m7
    punpckhqdq  m5, m4
    SBUTTERFLY qdq, 0, 1, 7
    SBUTTERFLY qdq, 2, 3, 7
%endif

    mova       p2m, m6
    LOAD_MASK   m0, m1, m2, m3, am, bm, m7, m4, m6
    LOAD_TC     m6, r4
    mova       tcm, m6

    LUMA_DEBLOCK_ONE m1, m0, ms1
    mova       p1m, m5

    mova        m5, p2m
    LUMA_DEBLOCK_ONE m2, m3, ms2
    mova       p2m, m5

    pxor        m5, m5
    mova        m6, tcm
    pcmpgtw     m5, tcm
    psubw       m6, ms1
    pandn       m5, m7
    psubw       m6, ms2
    pand        m5, m6
    DEBLOCK_P0_Q0 m1, m2, m0, m3, m5, m7, m6
    mova        m0, p1m
    mova        m3, p2m
    TRANSPOSE4x4W 0, 1, 2, 3, 4
    LUMA_H_STORE r2, r3

    add         r4, mmsize/8
    lea         r0, [r0+r1*(mmsize/2)]
    lea         r2, [r2+r1*(mmsize/2)]
    dec         r5
    jg .loop
    ADD        rsp, pad
    RET
%endmacro

%if ARCH_X86_64
; in:  m0=p1, m1=p0, m2=q0, m3=q1, m8=p2, m9=q2
;      m12=alpha, m13=beta
; out: m0=p1', m3=q1', m1=p0', m2=q0'
; clobbers: m4, m5, m6, m7, m10, m11, m14
%macro DEBLOCK_LUMA_INTER_SSE2 0
    LOAD_MASK   m0, m1, m2, m3, m12, m13, m7, m4, m6
    LOAD_TC     m6, r4
    DIFF_LT     m8, m1, m13, m10, m4
    DIFF_LT     m9, m2, m13, m11, m4
    pand        m6, m7

    mova       m14, m6
    pxor        m4, m4
    pcmpgtw     m6, m4
    pand        m6, m14

    mova        m5, m10
    pand        m5, m6
    LUMA_Q1 m8, m0, m1, m2, m5, m4

    mova        m5, m11
    pand        m5, m6
    LUMA_Q1 m9, m3, m1, m2, m5, m4

    pxor        m4, m4
    psubw       m6, m10
    pcmpgtw     m4, m14
    pandn       m4, m7
    psubw       m6, m11
    pand        m4, m6
    DEBLOCK_P0_Q0 m1, m2, m0, m3, m4, m5, m6

    SWAP         0, 8
    SWAP         3, 9
%endmacro

%macro DEBLOCK_LUMA_64 0
cglobal deblock_v_luma_10, 5,5,15
    %define p2 m8
    %define p1 m0
    %define p0 m1
    %define q0 m2
    %define q1 m3
    %define q2 m9
    %define mask0 m7
    %define mask1 m10
    %define mask2 m11
    shl        r2d, 2
    shl        r3d, 2
    LOAD_AB    m12, m13, r2d, r3d
    mov         r2, r0
    sub         r0, r1
    sub         r0, r1
    sub         r0, r1
    mov         r3, 2
.loop:
    mova        p2, [r0]
    mova        p1, [r0+r1]
    mova        p0, [r0+r1*2]
    mova        q0, [r2]
    mova        q1, [r2+r1]
    mova        q2, [r2+r1*2]
    DEBLOCK_LUMA_INTER_SSE2
    mova   [r0+r1], p1
    mova [r0+r1*2], p0
    mova      [r2], q0
    mova   [r2+r1], q1
    add         r0, mmsize
    add         r2, mmsize
    add         r4, 2
    dec         r3
    jg .loop
    REP_RET

cglobal deblock_h_luma_10, 5,7,15
    shl        r2d, 2
    shl        r3d, 2
    LOAD_AB    m12, m13, r2d, r3d
    mov         r2, r1
    add         r2, r1
    add         r2, r1
    mov         r5, r0
    add         r5, r2
    mov         r6, 2
.loop:
    movu        m8, [r0-8]     ; y q2 q1 q0 p0 p1 p2 x
    movu        m0, [r0+r1-8]
    movu        m2, [r0+r1*2-8]
    movu        m9, [r5-8]
    movu        m5, [r5+r1-8]
    movu        m1, [r5+r1*2-8]
    movu        m3, [r5+r2-8]
    movu        m7, [r5+r1*4-8]

    TRANSPOSE4x4W 8, 0, 2, 9, 10
    TRANSPOSE4x4W 5, 1, 3, 7, 10

    punpckhqdq  m8, m5
    SBUTTERFLY qdq, 0, 1, 10
    SBUTTERFLY qdq, 2, 3, 10
    punpcklqdq  m9, m7

    DEBLOCK_LUMA_INTER_SSE2

    TRANSPOSE4x4W 0, 1, 2, 3, 4
    LUMA_H_STORE r5, r2
    add         r4, 2
    lea         r0, [r0+r1*8]
    lea         r5, [r5+r1*8]
    dec         r6
    jg .loop
    REP_RET
%endmacro

INIT_XMM sse2
DEBLOCK_LUMA_64
%if HAVE_AVX_EXTERNAL
INIT_XMM avx
DEBLOCK_LUMA_64
%endif
%endif

%macro SWAPMOVA 2
%ifid %1
    SWAP %1, %2
%else
    mova %1, %2
%endif
%endmacro

; in: t0-t2: tmp registers
;     %1=p0 %2=p1 %3=p2 %4=p3 %5=q0 %6=q1 %7=mask0
;     %8=mask1p %9=2 %10=p0' %11=p1' %12=p2'
%macro LUMA_INTRA_P012 12 ; p0..p3 in memory
%if ARCH_X86_64
    paddw     t0, %3, %2
    mova      t2, %4
    paddw     t2, %3
%else
    mova      t0, %3
    mova      t2, %4
    paddw     t0, %2
    paddw     t2, %3
%endif
    paddw     t0, %1
    paddw     t2, t2
    paddw     t0, %5
    paddw     t2, %9
    paddw     t0, %9    ; (p2 + p1 + p0 + q0 + 2)
    paddw     t2, t0    ; (2*p3 + 3*p2 + p1 + p0 + q0 + 4)

    psrlw     t2, 3
    psrlw     t1, t0, 2
    psubw     t2, %3
    psubw     t1, %2
    pand      t2, %8
    pand      t1, %8
    paddw     t2, %3
    paddw     t1, %2
    SWAPMOVA %11, t1

    psubw     t1, t0, %3
    paddw     t0, t0
    psubw     t1, %5
    psubw     t0, %3
    paddw     t1, %6
    paddw     t1, %2
    paddw     t0, %6
    psrlw     t1, 2     ; (2*p1 + p0 + q1 + 2)/4
    psrlw     t0, 3     ; (p2 + 2*p1 + 2*p0 + 2*q0 + q1 + 4)>>3

    pxor      t0, t1
    pxor      t1, %1
    pand      t0, %8
    pand      t1, %7
    pxor      t0, t1
    pxor      t0, %1
    SWAPMOVA %10, t0
    SWAPMOVA %12, t2
%endmacro

%macro LUMA_INTRA_INIT 1
    %xdefine pad %1*mmsize+((gprsize*3) % mmsize)-(stack_offset&15)
    %define t0 m4
    %define t1 m5
    %define t2 m6
    %define t3 m7
    %assign i 4
%rep %1
    CAT_XDEFINE t, i, [rsp+mmsize*(i-4)]
    %assign i i+1
%endrep
    SUB    rsp, pad
%endmacro

; in: %1-%3=tmp, %4=p2, %5=q2
%macro LUMA_INTRA_INTER 5
    LOAD_AB t0, t1, r2d, r3d
    mova    %1, t0
    LOAD_MASK m0, m1, m2, m3, %1, t1, t0, t2, t3
%if ARCH_X86_64
    mova    %2, t0        ; mask0
    psrlw   t3, %1, 2
%else
    mova    t3, %1
    mova    %2, t0        ; mask0
    psrlw   t3, 2
%endif
    paddw   t3, [pw_2]    ; alpha/4+2
    DIFF_LT m1, m2, t3, t2, t0 ; t2 = |p0-q0| < alpha/4+2
    pand    t2, %2
    mova    t3, %5        ; q2
    mova    %1, t2        ; mask1
    DIFF_LT t3, m2, t1, t2, t0 ; t2 = |q2-q0| < beta
    pand    t2, %1
    mova    t3, %4        ; p2
    mova    %3, t2        ; mask1q
    DIFF_LT t3, m1, t1, t2, t0 ; t2 = |p2-p0| < beta
    pand    t2, %1
    mova    %1, t2        ; mask1p
%endmacro

%macro LUMA_H_INTRA_LOAD 0
%if mmsize == 8
    movu    t0, [r0-8]
    movu    t1, [r0+r1-8]
    movu    m0, [r0+r1*2-8]
    movu    m1, [r0+r4-8]
    TRANSPOSE4x4W 4, 5, 0, 1, 2
    mova    t4, t0        ; p3
    mova    t5, t1        ; p2

    movu    m2, [r0]
    movu    m3, [r0+r1]
    movu    t0, [r0+r1*2]
    movu    t1, [r0+r4]
    TRANSPOSE4x4W 2, 3, 4, 5, 6
    mova    t6, t0        ; q2
    mova    t7, t1        ; q3
%else
    movu    t0, [r0-8]
    movu    t1, [r0+r1-8]
    movu    m0, [r0+r1*2-8]
    movu    m1, [r0+r5-8]
    movu    m2, [r4-8]
    movu    m3, [r4+r1-8]
    movu    t2, [r4+r1*2-8]
    movu    t3, [r4+r5-8]
    TRANSPOSE8x8W 4, 5, 0, 1, 2, 3, 6, 7, t4, t5
    mova    t4, t0        ; p3
    mova    t5, t1        ; p2
    mova    t6, t2        ; q2
    mova    t7, t3        ; q3
%endif
%endmacro

; in: %1=q3 %2=q2' %3=q1' %4=q0' %5=p0' %6=p1' %7=p2' %8=p3 %9=tmp
%macro LUMA_H_INTRA_STORE 9
%if mmsize == 8
    TRANSPOSE4x4W %1, %2, %3, %4, %9
    movq       [r0-8], m%1
    movq       [r0+r1-8], m%2
    movq       [r0+r1*2-8], m%3
    movq       [r0+r4-8], m%4
    movq       m%1, %8
    TRANSPOSE4x4W %5, %6, %7, %1, %9
    movq       [r0], m%5
    movq       [r0+r1], m%6
    movq       [r0+r1*2], m%7
    movq       [r0+r4], m%1
%else
    TRANSPOSE2x4x4W %1, %2, %3, %4, %9
    movq       [r0-8], m%1
    movq       [r0+r1-8], m%2
    movq       [r0+r1*2-8], m%3
    movq       [r0+r5-8], m%4
    movhps     [r4-8], m%1
    movhps     [r4+r1-8], m%2
    movhps     [r4+r1*2-8], m%3
    movhps     [r4+r5-8], m%4
%ifnum %8
    SWAP       %1, %8
%else
    mova       m%1, %8
%endif
    TRANSPOSE2x4x4W %5, %6, %7, %1, %9
    movq       [r0], m%5
    movq       [r0+r1], m%6
    movq       [r0+r1*2], m%7
    movq       [r0+r5], m%1
    movhps     [r4], m%5
    movhps     [r4+r1], m%6
    movhps     [r4+r1*2], m%7
    movhps     [r4+r5], m%1
%endif
%endmacro

%if ARCH_X86_64
;-----------------------------------------------------------------------------
; void deblock_v_luma_intra( uint16_t *pix, int stride, int alpha, int beta )
;-----------------------------------------------------------------------------
%macro DEBLOCK_LUMA_INTRA_64 0
cglobal deblock_v_luma_intra_10, 4,7,16
    %define t0 m1
    %define t1 m2
    %define t2 m4
    %define p2 m8
    %define p1 m9
    %define p0 m10
    %define q0 m11
    %define q1 m12
    %define q2 m13
    %define aa m5
    %define bb m14
    lea     r4, [r1*4]
    lea     r5, [r1*3] ; 3*stride
    neg     r4
    add     r4, r0     ; pix-4*stride
    mov     r6, 2
    mova    m0, [pw_2]
    shl    r2d, 2
    shl    r3d, 2
    LOAD_AB aa, bb, r2d, r3d
.loop:
    mova    p2, [r4+r1]
    mova    p1, [r4+2*r1]
    mova    p0, [r4+r5]
    mova    q0, [r0]
    mova    q1, [r0+r1]
    mova    q2, [r0+2*r1]

    LOAD_MASK p1, p0, q0, q1, aa, bb, m3, t0, t1
    mova    t2, aa
    psrlw   t2, 2
    paddw   t2, m0 ; alpha/4+2
    DIFF_LT p0, q0, t2, m6, t0 ; m6 = |p0-q0| < alpha/4+2
    DIFF_LT p2, p0, bb, t1, t0 ; m7 = |p2-p0| < beta
    DIFF_LT q2, q0, bb, m7, t0 ; t1 = |q2-q0| < beta
    pand    m6, m3
    pand    m7, m6
    pand    m6, t1
    LUMA_INTRA_P012 p0, p1, p2, [r4], q0, q1, m3, m6, m0, [r4+r5], [r4+2*r1], [r4+r1]
    LUMA_INTRA_P012 q0, q1, q2, [r0+r5], p0, p1, m3, m7, m0, [r0], [r0+r1], [r0+2*r1]
    add     r0, mmsize
    add     r4, mmsize
    dec     r6
    jg .loop
    REP_RET

;-----------------------------------------------------------------------------
; void deblock_h_luma_intra( uint16_t *pix, int stride, int alpha, int beta )
;-----------------------------------------------------------------------------
cglobal deblock_h_luma_intra_10, 4,7,16
    %define t0 m15
    %define t1 m14
    %define t2 m2
    %define q3 m5
    %define q2 m8
    %define q1 m9
    %define q0 m10
    %define p0 m11
    %define p1 m12
    %define p2 m13
    %define p3 m4
    %define spill [rsp]
    %assign pad 24-(stack_offset&15)
    SUB     rsp, pad
    lea     r4, [r1*4]
    lea     r5, [r1*3] ; 3*stride
    add     r4, r0     ; pix+4*stride
    mov     r6, 2
    mova    m0, [pw_2]
    shl    r2d, 2
    shl    r3d, 2
.loop:
    movu    q3, [r0-8]
    movu    q2, [r0+r1-8]
    movu    q1, [r0+r1*2-8]
    movu    q0, [r0+r5-8]
    movu    p0, [r4-8]
    movu    p1, [r4+r1-8]
    movu    p2, [r4+r1*2-8]
    movu    p3, [r4+r5-8]
    TRANSPOSE8x8W 5, 8, 9, 10, 11, 12, 13, 4, 1

    LOAD_AB m1, m2, r2d, r3d
    LOAD_MASK q1, q0, p0, p1, m1, m2, m3, t0, t1
    psrlw   m1, 2
    paddw   m1, m0 ; alpha/4+2
    DIFF_LT p0, q0, m1, m6, t0 ; m6 = |p0-q0| < alpha/4+2
    DIFF_LT q2, q0, m2, t1, t0 ; t1 = |q2-q0| < beta
    DIFF_LT p0, p2, m2, m7, t0 ; m7 = |p2-p0| < beta
    pand    m6, m3
    pand    m7, m6
    pand    m6, t1

    mova spill, q3
    LUMA_INTRA_P012 q0, q1, q2, q3, p0, p1, m3, m6, m0, m5, m1, q2
    LUMA_INTRA_P012 p0, p1, p2, p3, q0, q1, m3, m7, m0, p0, m6, p2
    mova    m7, spill

    LUMA_H_INTRA_STORE 7, 8, 1, 5, 11, 6, 13, 4, 14

    lea     r0, [r0+r1*8]
    lea     r4, [r4+r1*8]
    dec     r6
    jg .loop
    ADD    rsp, pad
    RET
%endmacro

INIT_XMM sse2
DEBLOCK_LUMA_INTRA_64
%if HAVE_AVX_EXTERNAL
INIT_XMM avx
DEBLOCK_LUMA_INTRA_64
%endif

%endif

%macro DEBLOCK_LUMA_INTRA 0
;-----------------------------------------------------------------------------
; void deblock_v_luma_intra( uint16_t *pix, int stride, int alpha, int beta )
;-----------------------------------------------------------------------------
cglobal deblock_v_luma_intra_10, 4,7,8*(mmsize/16)
    LUMA_INTRA_INIT 3
    lea     r4, [r1*4]
    lea     r5, [r1*3]
    neg     r4
    add     r4, r0
    mov     r6, 32/mmsize
    shl    r2d, 2
    shl    r3d, 2
.loop:
    mova    m0, [r4+r1*2] ; p1
    mova    m1, [r4+r5]   ; p0
    mova    m2, [r0]      ; q0
    mova    m3, [r0+r1]   ; q1
    LUMA_INTRA_INTER t4, t5, t6, [r4+r1], [r0+r1*2]
    LUMA_INTRA_P012 m1, m0, t3, [r4], m2, m3, t5, t4, [pw_2], [r4+r5], [r4+2*r1], [r4+r1]
    mova    t3, [r0+r1*2] ; q2
    LUMA_INTRA_P012 m2, m3, t3, [r0+r5], m1, m0, t5, t6, [pw_2], [r0], [r0+r1], [r0+2*r1]
    add     r0, mmsize
    add     r4, mmsize
    dec     r6
    jg .loop
    ADD    rsp, pad
    RET

;-----------------------------------------------------------------------------
; void deblock_h_luma_intra( uint16_t *pix, int stride, int alpha, int beta )
;-----------------------------------------------------------------------------
cglobal deblock_h_luma_intra_10, 4,7,8*(mmsize/16)
    LUMA_INTRA_INIT 8
%if mmsize == 8
    lea     r4, [r1*3]
    mov     r5, 32/mmsize
%else
    lea     r4, [r1*4]
    lea     r5, [r1*3] ; 3*stride
    add     r4, r0     ; pix+4*stride
    mov     r6, 32/mmsize
%endif
    shl    r2d, 2
    shl    r3d, 2
.loop:
    LUMA_H_INTRA_LOAD
    LUMA_INTRA_INTER t8, t9, t10, t5, t6

    LUMA_INTRA_P012 m1, m0, t3, t4, m2, m3, t9, t8, [pw_2], t8, t5, t11
    mova    t3, t6     ; q2
    LUMA_INTRA_P012 m2, m3, t3, t7, m1, m0, t9, t10, [pw_2], m4, t6, m5

    mova    m2, t4
    mova    m0, t11
    mova    m1, t5
    mova    m3, t8
    mova    m6, t6

    LUMA_H_INTRA_STORE 2, 0, 1, 3, 4, 6, 5, t7, 7

    lea     r0, [r0+r1*(mmsize/2)]
%if mmsize == 8
    dec     r5
%else
    lea     r4, [r4+r1*(mmsize/2)]
    dec     r6
%endif
    jg .loop
    ADD    rsp, pad
    RET
%endmacro

%if ARCH_X86_64 == 0
INIT_MMX mmxext
DEBLOCK_LUMA
DEBLOCK_LUMA_INTRA
INIT_XMM sse2
DEBLOCK_LUMA
DEBLOCK_LUMA_INTRA
%if HAVE_AVX_EXTERNAL
INIT_XMM avx
DEBLOCK_LUMA
DEBLOCK_LUMA_INTRA
%endif
%endif

; in: %1=p0, %2=q0, %3=p1, %4=q1, %5=mask, %6=tmp, %7=tmp
; out: %1=p0', %2=q0'
%macro CHROMA_DEBLOCK_P0_Q0_INTRA 7
    mova    %6, [pw_2]
    paddw   %6, %3
    paddw   %6, %4
    paddw   %7, %6, %2
    paddw   %6, %1
    paddw   %6, %3
    paddw   %7, %4
    psraw   %6, 2
    psraw   %7, 2
    psubw   %6, %1
    psubw   %7, %2
    pand    %6, %5
    pand    %7, %5
    paddw   %1, %6
    paddw   %2, %7
%endmacro

%macro CHROMA_V_LOAD 1
    mova        m0, [r0]    ; p1
    mova        m1, [r0+r1] ; p0
    mova        m2, [%1]    ; q0
    mova        m3, [%1+r1] ; q1
%endmacro

%macro CHROMA_V_STORE 0
    mova [r0+1*r1], m1
    mova [r0+2*r1], m2
%endmacro

%macro CHROMA_V_LOAD_TC 2
    movd        %1, [%2]
    punpcklbw   %1, %1
    punpcklwd   %1, %1
    psraw       %1, 6
%endmacro

%macro DEBLOCK_CHROMA 0
;-----------------------------------------------------------------------------
; void deblock_v_chroma( uint16_t *pix, int stride, int alpha, int beta, int8_t *tc0 )
;-----------------------------------------------------------------------------
cglobal deblock_v_chroma_10, 5,7-(mmsize/16),8*(mmsize/16)
    mov         r5, r0
    sub         r0, r1
    sub         r0, r1
    shl        r2d, 2
    shl        r3d, 2
%if mmsize < 16
    mov         r6, 16/mmsize
.loop:
%endif
    CHROMA_V_LOAD r5
    LOAD_AB     m4, m5, r2d, r3d
    LOAD_MASK   m0, m1, m2, m3, m4, m5, m7, m6, m4
    pxor        m4, m4
    CHROMA_V_LOAD_TC m6, r4
    psubw       m6, [pw_3]
    pmaxsw      m6, m4
    pand        m7, m6
    DEBLOCK_P0_Q0 m1, m2, m0, m3, m7, m5, m6
    CHROMA_V_STORE
%if mmsize < 16
    add         r0, mmsize
    add         r5, mmsize
    add         r4, mmsize/4
    dec         r6
    jg .loop
    REP_RET
%else
    RET
%endif

;-----------------------------------------------------------------------------
; void deblock_v_chroma_intra( uint16_t *pix, int stride, int alpha, int beta )
;-----------------------------------------------------------------------------
cglobal deblock_v_chroma_intra_10, 4,6-(mmsize/16),8*(mmsize/16)
    mov         r4, r0
    sub         r0, r1
    sub         r0, r1
    shl        r2d, 2
    shl        r3d, 2
%if mmsize < 16
    mov         r5, 16/mmsize
.loop:
%endif
    CHROMA_V_LOAD r4
    LOAD_AB     m4, m5, r2d, r3d
    LOAD_MASK   m0, m1, m2, m3, m4, m5, m7, m6, m4
    CHROMA_DEBLOCK_P0_Q0_INTRA m1, m2, m0, m3, m7, m5, m6
    CHROMA_V_STORE
%if mmsize < 16
    add         r0, mmsize
    add         r4, mmsize
    dec         r5
    jg .loop
    REP_RET
%else
    RET
%endif
%endmacro

%if ARCH_X86_64 == 0
INIT_MMX mmxext
DEBLOCK_CHROMA
%endif
INIT_XMM sse2
DEBLOCK_CHROMA
%if HAVE_AVX_EXTERNAL
INIT_XMM avx
DEBLOCK_CHROMA
%endif