-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathfilm_grain.asm
2405 lines (2208 loc) · 71.2 KB
/
film_grain.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
; Copyright © 2019, VideoLAN and dav1d authors
; Copyright © 2019, Two Orioles, LLC
; All rights reserved.
;
; Redistribution and use in source and binary forms, with or without
; modification, are permitted provided that the following conditions are met:
;
; 1. Redistributions of source code must retain the above copyright notice, this
; list of conditions and the following disclaimer.
;
; 2. Redistributions in binary form must reproduce the above copyright notice,
; this list of conditions and the following disclaimer in the documentation
; and/or other materials provided with the distribution.
;
; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
; ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
; WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
; DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR
; ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
; (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
; ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
; SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
%include "config.asm"
%include "ext/x86/x86inc.asm"
%if ARCH_X86_64
SECTION_RODATA 32
pb_8x_27_17_8x_17_27: times 8 db 27, 17
times 8 db 17, 27
pw_1024: times 16 dw 1024
pb_mask: db 0, 0x80, 0x80, 0, 0x80, 0, 0, 0x80, 0x80, 0, 0, 0x80, 0, 0x80, 0x80, 0
rnd_next_upperbit_mask: dw 0x100B, 0x2016, 0x402C, 0x8058
byte_blend: db 0, 0, 0, 0xff, 0, 0, 0, 0
pw_seed_xor: times 2 dw 0xb524
times 2 dw 0x49d8
pd_m65536: dd ~0xffff
pb_23_22: times 2 db 23, 22
pb_1: times 4 db 1
hmul_bits: dw 32768, 16384, 8192, 4096
round: dw 2048, 1024, 512
mul_bits: dw 256, 128, 64, 32, 16
round_vals: dw 32, 64, 128, 256, 512
max: dw 255, 240, 235
min: dw 0, 16
pb_27_17_17_27: db 27, 17, 17, 27
pw_1: dw 1
%macro JMP_TABLE 1-*
%xdefine %1_table %%table
%xdefine %%base %1_table
%xdefine %%prefix mangle(private_prefix %+ _%1)
%%table:
%rep %0 - 1
dd %%prefix %+ .ar%2 - %%base
%rotate 1
%endrep
%endmacro
ALIGN 4
JMP_TABLE generate_grain_y_avx2, 0, 1, 2, 3
JMP_TABLE generate_grain_uv_420_avx2, 0, 1, 2, 3
JMP_TABLE generate_grain_uv_422_avx2, 0, 1, 2, 3
JMP_TABLE generate_grain_uv_444_avx2, 0, 1, 2, 3
struc FGData
.seed: resd 1
.num_y_points: resd 1
.y_points: resb 14 * 2
.chroma_scaling_from_luma: resd 1
.num_uv_points: resd 2
.uv_points: resb 2 * 10 * 2
.scaling_shift: resd 1
.ar_coeff_lag: resd 1
.ar_coeffs_y: resb 24
.ar_coeffs_uv: resb 2 * 28 ; includes padding
.ar_coeff_shift: resq 1
.grain_scale_shift: resd 1
.uv_mult: resd 2
.uv_luma_mult: resd 2
.uv_offset: resd 2
.overlap_flag: resd 1
.clip_to_restricted_range: resd 1
endstruc
cextern gaussian_sequence
SECTION .text
INIT_XMM avx2
cglobal generate_grain_y, 2, 9, 16, buf, fg_data
lea r4, [pb_mask]
%define base r4-pb_mask
movq xm1, [base+rnd_next_upperbit_mask]
movq xm4, [base+mul_bits]
movq xm7, [base+hmul_bits]
mov r2d, [fg_dataq+FGData.grain_scale_shift]
vpbroadcastw xm8, [base+round+r2*2]
mova xm5, [base+pb_mask]
vpbroadcastw xm0, [fg_dataq+FGData.seed]
vpbroadcastd xm9, [base+pd_m65536]
mov r2, -73*82
sub bufq, r2
lea r3, [gaussian_sequence]
.loop:
pand xm2, xm0, xm1
psrlw xm3, xm2, 10
por xm2, xm3 ; bits 0xf, 0x1e, 0x3c and 0x78 are set
pmullw xm2, xm4 ; bits 0x0f00 are set
pshufb xm2, xm5, xm2 ; set 15th bit for next 4 seeds
psllq xm6, xm2, 30
por xm2, xm6
psllq xm6, xm2, 15
por xm2, xm6 ; aggregate each bit into next seed's high bit
pmulhuw xm3, xm0, xm7
por xm2, xm3 ; 4 next output seeds
pshuflw xm0, xm2, q3333
psrlw xm2, 5
pmovzxwd xm3, xm2
mova xm6, xm9
vpgatherdd xm2, [r3+xm3*2], xm6
pandn xm2, xm9, xm2
packusdw xm2, xm2
pmulhrsw xm2, xm8
packsswb xm2, xm2
movd [bufq+r2], xm2
add r2, 4
jl .loop
; auto-regression code
movsxd r2, [fg_dataq+FGData.ar_coeff_lag]
movsxd r2, [base+generate_grain_y_avx2_table+r2*4]
lea r2, [r2+base+generate_grain_y_avx2_table]
jmp r2
.ar1:
DEFINE_ARGS buf, fg_data, cf3, shift, val3, min, max, x, val0
mov shiftd, [fg_dataq+FGData.ar_coeff_shift]
movsx cf3d, byte [fg_dataq+FGData.ar_coeffs_y+3]
movd xm4, [fg_dataq+FGData.ar_coeffs_y]
DEFINE_ARGS buf, h, cf3, shift, val3, min, max, x, val0
pinsrb xm4, [pb_1], 3
pmovsxbw xm4, xm4
pshufd xm5, xm4, q1111
pshufd xm4, xm4, q0000
vpbroadcastw xm3, [base+round_vals+shiftq*2-12] ; rnd
sub bufq, 82*73-(82*3+79)
mov hd, 70
mov mind, -128
mov maxd, 127
.y_loop_ar1:
mov xq, -76
movsx val3d, byte [bufq+xq-1]
.x_loop_ar1:
pmovsxbw xm0, [bufq+xq-82-1] ; top/left
pmovsxbw xm2, [bufq+xq-82+0] ; top
pmovsxbw xm1, [bufq+xq-82+1] ; top/right
punpcklwd xm0, xm2
punpcklwd xm1, xm3
pmaddwd xm0, xm4
pmaddwd xm1, xm5
paddd xm0, xm1
.x_loop_ar1_inner:
movd val0d, xm0
psrldq xm0, 4
imul val3d, cf3d
add val3d, val0d
%if WIN64
sarx val3d, val3d, shiftd
%else
sar val3d, shiftb
%endif
movsx val0d, byte [bufq+xq]
add val3d, val0d
cmp val3d, maxd
cmovns val3d, maxd
cmp val3d, mind
cmovs val3d, mind
mov byte [bufq+xq], val3b
; keep val3d in-place as left for next x iteration
inc xq
jz .x_loop_ar1_end
test xq, 3
jnz .x_loop_ar1_inner
jmp .x_loop_ar1
.x_loop_ar1_end:
add bufq, 82
dec hd
jg .y_loop_ar1
.ar0:
RET
.ar2:
DEFINE_ARGS buf, fg_data, shift
mov shiftd, [fg_dataq+FGData.ar_coeff_shift]
vpbroadcastw xm14, [base+round_vals-12+shiftq*2]
movq xm15, [base+byte_blend+1]
pmovsxbw xm8, [fg_dataq+FGData.ar_coeffs_y+0] ; cf0-7
movd xm9, [fg_dataq+FGData.ar_coeffs_y+8] ; cf8-11
pmovsxbw xm9, xm9
DEFINE_ARGS buf, fg_data, h, x
pshufd xm12, xm9, q0000
pshufd xm13, xm9, q1111
pshufd xm11, xm8, q3333
pshufd xm10, xm8, q2222
pshufd xm9, xm8, q1111
pshufd xm8, xm8, q0000
pmovzxwd xm14, xm14
sub bufq, 82*73-(82*3+79)
mov hd, 70
.y_loop_ar2:
mov xq, -76
.x_loop_ar2:
pmovsxbw xm0, [bufq+xq-82*2-2] ; y=-2,x=[-2,+5]
pmovsxbw xm1, [bufq+xq-82*1-2] ; y=-1,x=[-2,+5]
psrldq xm2, xm0, 2 ; y=-2,x=[-1,+5]
psrldq xm3, xm1, 2 ; y=-1,x=[-1,+5]
psrldq xm4, xm1, 4 ; y=-1,x=[+0,+5]
punpcklwd xm2, xm0, xm2
punpcklwd xm3, xm4
pmaddwd xm2, xm8
pmaddwd xm3, xm11
paddd xm2, xm3
psrldq xm4, xm0, 4 ; y=-2,x=[+0,+5]
psrldq xm5, xm0, 6 ; y=-2,x=[+1,+5]
psrldq xm6, xm0, 8 ; y=-2,x=[+2,+5]
punpcklwd xm4, xm5
punpcklwd xm6, xm1
psrldq xm7, xm1, 6 ; y=-1,x=[+1,+5]
psrldq xm1, xm1, 8 ; y=-1,x=[+2,+5]
punpcklwd xm7, xm1
pmaddwd xm4, xm9
pmaddwd xm6, xm10
pmaddwd xm7, xm12
paddd xm4, xm6
paddd xm2, xm7
paddd xm2, xm4
paddd xm2, xm14
movq xm0, [bufq+xq-2] ; y=0,x=[-2,+5]
.x_loop_ar2_inner:
pmovsxbw xm1, xm0
pmaddwd xm3, xm1, xm13
paddd xm3, xm2
psrldq xm1, 4 ; y=0,x=0
psrldq xm2, 4 ; shift top to next pixel
psrad xm3, [fg_dataq+FGData.ar_coeff_shift]
; don't packssdw since we only care about one value
paddw xm3, xm1
packsswb xm3, xm3
pextrb [bufq+xq], xm3, 0
pslldq xm3, 2
pand xm3, xm15
pandn xm0, xm15, xm0
por xm0, xm3
psrldq xm0, 1
inc xq
jz .x_loop_ar2_end
test xq, 3
jnz .x_loop_ar2_inner
jmp .x_loop_ar2
.x_loop_ar2_end:
add bufq, 82
dec hd
jg .y_loop_ar2
RET
.ar3:
DEFINE_ARGS buf, fg_data, shift
%if WIN64
SUB rsp, 16*12
%assign stack_size_padded (stack_size_padded+16*12)
%assign stack_size (stack_size+16*12)
%else
ALLOC_STACK 16*12
%endif
mov shiftd, [fg_dataq+FGData.ar_coeff_shift]
vpbroadcastw xm14, [base+round_vals-12+shiftq*2]
movq xm15, [base+byte_blend]
pmovsxbw xm0, [fg_dataq+FGData.ar_coeffs_y+ 0] ; cf0-7
pmovsxbw xm1, [fg_dataq+FGData.ar_coeffs_y+ 8] ; cf8-15
pmovsxbw xm2, [fg_dataq+FGData.ar_coeffs_y+16] ; cf16-23
pshufd xm9, xm0, q1111
pshufd xm10, xm0, q2222
pshufd xm11, xm0, q3333
pshufd xm0, xm0, q0000
pshufd xm6, xm1, q1111
pshufd xm7, xm1, q2222
pshufd xm8, xm1, q3333
pshufd xm1, xm1, q0000
pshufd xm3, xm2, q1111
psrldq xm13, xm2, 10
pinsrw xm2, [pw_1], 5
pshufd xm4, xm2, q2222
pshufd xm2, xm2, q0000
pinsrw xm13, [base+round_vals+shiftq*2-10], 3
mova [rsp+ 0*16], xm0
mova [rsp+ 1*16], xm9
mova [rsp+ 2*16], xm10
mova [rsp+ 3*16], xm11
mova [rsp+ 4*16], xm1
mova [rsp+ 5*16], xm6
mova [rsp+ 6*16], xm7
mova [rsp+ 7*16], xm8
mova [rsp+ 8*16], xm2
mova [rsp+ 9*16], xm3
mova [rsp+10*16], xm4
DEFINE_ARGS buf, fg_data, h, x
sub bufq, 82*73-(82*3+79)
mov hd, 70
.y_loop_ar3:
mov xq, -76
.x_loop_ar3:
movu xm0, [bufq+xq-82*3-3] ; y=-3,x=[-3,+12]
movu xm1, [bufq+xq-82*2-3] ; y=-2,x=[-3,+12]
movu xm2, [bufq+xq-82*1-3] ; y=-1,x=[-3,+12]
pxor xm3, xm3
pcmpgtb xm6, xm3, xm2
pcmpgtb xm5, xm3, xm1
pcmpgtb xm4, xm3, xm0
punpckhbw xm3, xm0, xm4
punpcklbw xm0, xm4
punpckhbw xm4, xm1, xm5
punpcklbw xm1, xm5
punpckhbw xm5, xm2, xm6
punpcklbw xm2, xm6
psrldq xm6, xm0, 2
psrldq xm7, xm0, 4
psrldq xm8, xm0, 6
psrldq xm9, xm0, 8
palignr xm10, xm3, xm0, 10
palignr xm11, xm3, xm0, 12
punpcklwd xm0, xm6
punpcklwd xm7, xm8
punpcklwd xm9, xm10
punpcklwd xm11, xm1
pmaddwd xm0, [rsp+ 0*16]
pmaddwd xm7, [rsp+ 1*16]
pmaddwd xm9, [rsp+ 2*16]
pmaddwd xm11, [rsp+ 3*16]
paddd xm0, xm7
paddd xm9, xm11
paddd xm0, xm9
psrldq xm6, xm1, 2
psrldq xm7, xm1, 4
psrldq xm8, xm1, 6
psrldq xm9, xm1, 8
palignr xm10, xm4, xm1, 10
palignr xm11, xm4, xm1, 12
psrldq xm12, xm2, 2
punpcklwd xm6, xm7
punpcklwd xm8, xm9
punpcklwd xm10, xm11
punpcklwd xm12, xm2, xm12
pmaddwd xm6, [rsp+ 4*16]
pmaddwd xm8, [rsp+ 5*16]
pmaddwd xm10, [rsp+ 6*16]
pmaddwd xm12, [rsp+ 7*16]
paddd xm6, xm8
paddd xm10, xm12
paddd xm6, xm10
paddd xm0, xm6
psrldq xm6, xm2, 4
psrldq xm7, xm2, 6
psrldq xm8, xm2, 8
palignr xm9, xm5, xm2, 10
palignr xm5, xm5, xm2, 12
punpcklwd xm6, xm7
punpcklwd xm8, xm9
punpcklwd xm5, xm14
pmaddwd xm6, [rsp+ 8*16]
pmaddwd xm8, [rsp+ 9*16]
pmaddwd xm5, [rsp+10*16]
paddd xm0, xm6
paddd xm8, xm5
paddd xm0, xm8
movq xm1, [bufq+xq-3] ; y=0,x=[-3,+4]
.x_loop_ar3_inner:
pmovsxbw xm2, xm1
pmaddwd xm2, xm13
pshufd xm3, xm2, q1111
paddd xm2, xm3 ; left+cur
paddd xm2, xm0 ; add top
psrldq xm0, 4
psrad xm2, [fg_dataq+FGData.ar_coeff_shift]
; don't packssdw since we only care about one value
packsswb xm2, xm2
pextrb [bufq+xq], xm2, 0
pslldq xm2, 3
pand xm2, xm15
pandn xm1, xm15, xm1
por xm1, xm2
psrldq xm1, 1
inc xq
jz .x_loop_ar3_end
test xq, 3
jnz .x_loop_ar3_inner
jmp .x_loop_ar3
.x_loop_ar3_end:
add bufq, 82
dec hd
jg .y_loop_ar3
RET
%macro generate_grain_uv_fn 3 ; ss_name, ss_x, ss_y
INIT_XMM avx2
cglobal generate_grain_uv_%1, 4, 10, 16, buf, bufy, fg_data, uv
lea r4, [pb_mask]
%define base r4-pb_mask
movq xm1, [base+rnd_next_upperbit_mask]
movq xm4, [base+mul_bits]
movq xm7, [base+hmul_bits]
mov r5d, [fg_dataq+FGData.grain_scale_shift]
vpbroadcastw xm8, [base+round+r5*2]
mova xm5, [base+pb_mask]
vpbroadcastw xm0, [fg_dataq+FGData.seed]
vpbroadcastw xm9, [base+pw_seed_xor+uvq*4]
pxor xm0, xm9
vpbroadcastd xm9, [base+pd_m65536]
lea r6, [gaussian_sequence]
%if %2
mov r7d, 73-35*%3
add bufq, 44
.loop_y:
mov r5, -44
.loop_x:
%else
mov r5, -73*82
sub bufq, r5
.loop:
%endif
pand xm2, xm0, xm1
psrlw xm3, xm2, 10
por xm2, xm3 ; bits 0xf, 0x1e, 0x3c and 0x78 are set
pmullw xm2, xm4 ; bits 0x0f00 are set
pshufb xm2, xm5, xm2 ; set 15th bit for next 4 seeds
psllq xm6, xm2, 30
por xm2, xm6
psllq xm6, xm2, 15
por xm2, xm6 ; aggregate each bit into next seed's high bit
pmulhuw xm3, xm0, xm7
por xm2, xm3 ; 4 next output seeds
pshuflw xm0, xm2, q3333
psrlw xm2, 5
pmovzxwd xm3, xm2
mova xm6, xm9
vpgatherdd xm2, [r6+xm3*2], xm6
pandn xm2, xm9, xm2
packusdw xm2, xm2
pmulhrsw xm2, xm8
packsswb xm2, xm2
movd [bufq+r5], xm2
add r5, 4
%if %2
jl .loop_x
add bufq, 82
dec r7d
jg .loop_y
%else
jl .loop
%endif
; auto-regression code
movsxd r5, [fg_dataq+FGData.ar_coeff_lag]
movsxd r5, [base+generate_grain_uv_%1_avx2_table+r5*4]
lea r5, [r5+base+generate_grain_uv_%1_avx2_table]
jmp r5
.ar0:
INIT_YMM avx2
DEFINE_ARGS buf, bufy, fg_data, uv, unused, shift
imul uvd, 28
mov shiftd, [fg_dataq+FGData.ar_coeff_shift]
movd xm4, [fg_dataq+FGData.ar_coeffs_uv+uvq]
movd xm3, [base+hmul_bits+shiftq*2]
DEFINE_ARGS buf, bufy, h
pmovsxbw xm4, xm4
%if %2
vpbroadcastd m7, [pb_1]
vpbroadcastw m6, [hmul_bits+2+%3*2]
%endif
vpbroadcastw m4, xm4
vpbroadcastw m3, xm3
pxor m12, m12
%if %2
sub bufq, 82*(73-35*%3)+82-(82*3+41)
%else
sub bufq, 82*70-3
%endif
add bufyq, 3+82*3
mov hd, 70-35*%3
.y_loop_ar0:
%if %2
; first 32 pixels
movu xm8, [bufyq]
%if %3
movu xm9, [bufyq+82]
%endif
movu xm10, [bufyq+16]
%if %3
movu xm11, [bufyq+82+16]
%endif
vinserti128 m8, [bufyq+32], 1
%if %3
vinserti128 m9, [bufyq+82+32], 1
%endif
vinserti128 m10, [bufyq+48], 1
%if %3
vinserti128 m11, [bufyq+82+48], 1
%endif
pmaddubsw m8, m7, m8
%if %3
pmaddubsw m9, m7, m9
%endif
pmaddubsw m10, m7, m10
%if %3
pmaddubsw m11, m7, m11
paddw m8, m9
paddw m10, m11
%endif
pmulhrsw m8, m6
pmulhrsw m10, m6
%else
xor r3d, r3d
; first 32x2 pixels
.x_loop_ar0:
movu m8, [bufyq+r3]
pcmpgtb m9, m12, m8
punpckhbw m10, m8, m9
punpcklbw m8, m9
%endif
pmullw m8, m4
pmullw m10, m4
pmulhrsw m8, m3
pmulhrsw m10, m3
%if %2
movu m0, [bufq]
%else
movu m0, [bufq+r3]
%endif
pcmpgtb m1, m12, m0
punpckhbw m9, m0, m1
punpcklbw m0, m1
paddw m0, m8
paddw m9, m10
packsswb m0, m9
%if %2
movu [bufq], m0
%else
movu [bufq+r3], m0
add r3d, 32
cmp r3d, 64
jl .x_loop_ar0
%endif
; last 6/12 pixels
movu xm8, [bufyq+32*2]
%if %2
%if %3
movu xm9, [bufyq+32*2+82]
%endif
pmaddubsw xm8, xm7, xm8
%if %3
pmaddubsw xm9, xm7, xm9
paddw xm8, xm9
%endif
pmulhrsw xm8, xm6
pmullw xm8, xm4
pmulhrsw xm8, xm3
movq xm0, [bufq+32]
pcmpgtb xm9, xm12, xm0
punpcklbw xm9, xm0, xm9
paddw xm8, xm9
packsswb xm8, xm8
vpblendw xm0, xm8, xm0, 1000b
movq [bufq+32], xm0
%else
pcmpgtb xm9, xm12, xm8
punpckhbw xm10, xm8, xm9
punpcklbw xm8, xm9
pmullw xm10, xm4
pmullw xm8, xm4
pmulhrsw xm10, xm3
pmulhrsw xm8, xm3
movu xm0, [bufq+64]
pcmpgtb xm9, xm12, xm0
punpcklbw xm1, xm0, xm9
punpckhbw xm9, xm0, xm9
paddw xm1, xm8
paddw xm9, xm10
packsswb xm1, xm9
vpblendw xm0, xm1, xm0, 11000000b
movu [bufq+64], xm0
%endif
add bufq, 82
add bufyq, 82<<%3
dec hd
jg .y_loop_ar0
RET
.ar1:
INIT_XMM avx2
DEFINE_ARGS buf, bufy, fg_data, uv, val3, cf3, min, max, x, shift
imul uvd, 28
mov shiftd, [fg_dataq+FGData.ar_coeff_shift]
movsx cf3d, byte [fg_dataq+FGData.ar_coeffs_uv+uvq+3]
movd xm4, [fg_dataq+FGData.ar_coeffs_uv+uvq]
pinsrb xm4, [fg_dataq+FGData.ar_coeffs_uv+uvq+4], 3
DEFINE_ARGS buf, bufy, h, val0, val3, cf3, min, max, x, shift
pmovsxbw xm4, xm4
pshufd xm5, xm4, q1111
pshufd xm4, xm4, q0000
pmovsxwd xm3, [base+round_vals+shiftq*2-12] ; rnd
%if %2
vpbroadcastd xm7, [pb_1]
vpbroadcastw xm6, [hmul_bits+2+%3*2]
%endif
vpbroadcastd xm3, xm3
%if %2
sub bufq, 82*(73-35*%3)+44-(82*3+41)
%else
sub bufq, 82*70-(82-3)
%endif
add bufyq, 79+82*3
mov hd, 70-35*%3
mov mind, -128
mov maxd, 127
.y_loop_ar1:
mov xq, -(76>>%2)
movsx val3d, byte [bufq+xq-1]
.x_loop_ar1:
pmovsxbw xm0, [bufq+xq-82-1] ; top/left
%if %2
movq xm8, [bufyq+xq*2]
%if %3
movq xm9, [bufyq+xq*2+82]
%endif
%endif
psrldq xm2, xm0, 2 ; top
psrldq xm1, xm0, 4 ; top/right
%if %2
pmaddubsw xm8, xm7, xm8
%if %3
pmaddubsw xm9, xm7, xm9
paddw xm8, xm9
%endif
pmulhrsw xm8, xm6
%else
pmovsxbw xm8, [bufyq+xq]
%endif
punpcklwd xm0, xm2
punpcklwd xm1, xm8
pmaddwd xm0, xm4
pmaddwd xm1, xm5
paddd xm0, xm1
paddd xm0, xm3
.x_loop_ar1_inner:
movd val0d, xm0
psrldq xm0, 4
imul val3d, cf3d
add val3d, val0d
sarx val3d, val3d, shiftd
movsx val0d, byte [bufq+xq]
add val3d, val0d
cmp val3d, maxd
cmovns val3d, maxd
cmp val3d, mind
cmovs val3d, mind
mov byte [bufq+xq], val3b
; keep val3d in-place as left for next x iteration
inc xq
jz .x_loop_ar1_end
test xq, 3
jnz .x_loop_ar1_inner
jmp .x_loop_ar1
.x_loop_ar1_end:
add bufq, 82
add bufyq, 82<<%3
dec hd
jg .y_loop_ar1
RET
.ar2:
DEFINE_ARGS buf, bufy, fg_data, uv, unused, shift
mov shiftd, [fg_dataq+FGData.ar_coeff_shift]
imul uvd, 28
vpbroadcastw xm15, [base+round_vals-12+shiftq*2]
pmovsxbw xm8, [fg_dataq+FGData.ar_coeffs_uv+uvq+0] ; cf0-7
pmovsxbw xm9, [fg_dataq+FGData.ar_coeffs_uv+uvq+8] ; cf8-12
pinsrw xm9, [base+pw_1], 5
%if %2
vpbroadcastw xm7, [base+hmul_bits+2+%3*2]
vpbroadcastd xm6, [base+pb_1]
%endif
DEFINE_ARGS buf, bufy, fg_data, h, unused, x
pshufd xm12, xm9, q0000
pshufd xm13, xm9, q1111
pshufd xm14, xm9, q2222
pshufd xm11, xm8, q3333
pshufd xm10, xm8, q2222
pshufd xm9, xm8, q1111
pshufd xm8, xm8, q0000
%if %2
sub bufq, 82*(73-35*%3)+44-(82*3+41)
%else
sub bufq, 82*70-(82-3)
%endif
add bufyq, 79+82*3
mov hd, 70-35*%3
.y_loop_ar2:
mov xq, -(76>>%2)
.x_loop_ar2:
pmovsxbw xm0, [bufq+xq-82*2-2] ; y=-2,x=[-2,+5]
pmovsxbw xm1, [bufq+xq-82*1-2] ; y=-1,x=[-2,+5]
psrldq xm2, xm0, 2 ; y=-2,x=[-1,+5]
psrldq xm3, xm1, 2 ; y=-1,x=[-1,+5]
psrldq xm4, xm1, 4 ; y=-1,x=[+0,+5]
punpcklwd xm2, xm0, xm2
punpcklwd xm3, xm4
pmaddwd xm2, xm8
pmaddwd xm3, xm11
paddd xm2, xm3
psrldq xm4, xm0, 4 ; y=-2,x=[+0,+5]
psrldq xm5, xm0, 6 ; y=-2,x=[+1,+5]
psrldq xm0, 8 ; y=-2,x=[+2,+5]
punpcklwd xm4, xm5
punpcklwd xm0, xm1
psrldq xm3, xm1, 6 ; y=-1,x=[+1,+5]
psrldq xm1, xm1, 8 ; y=-1,x=[+2,+5]
punpcklwd xm3, xm1
pmaddwd xm4, xm9
pmaddwd xm0, xm10
pmaddwd xm3, xm12
paddd xm4, xm0
paddd xm2, xm3
paddd xm2, xm4
%if %2
movq xm0, [bufyq+xq*2]
%if %3
movq xm3, [bufyq+xq*2+82]
%endif
pmaddubsw xm0, xm6, xm0
%if %3
pmaddubsw xm3, xm6, xm3
paddw xm0, xm3
%endif
pmulhrsw xm0, xm7
%else
pmovsxbw xm0, [bufyq+xq]
%endif
punpcklwd xm0, xm15
pmaddwd xm0, xm14
paddd xm2, xm0
movq xm0, [bufq+xq-2] ; y=0,x=[-2,+5]
.x_loop_ar2_inner:
pmovsxbw xm0, xm0
pmaddwd xm3, xm0, xm13
paddd xm3, xm2
psrldq xm2, 4 ; shift top to next pixel
psrad xm3, [fg_dataq+FGData.ar_coeff_shift]
pslldq xm3, 2
psrldq xm0, 2
paddw xm3, xm0
vpblendw xm0, xm3, 00000010b
packsswb xm0, xm0
pextrb [bufq+xq], xm0, 1
inc xq
jz .x_loop_ar2_end
test xq, 3
jnz .x_loop_ar2_inner
jmp .x_loop_ar2
.x_loop_ar2_end:
add bufq, 82
add bufyq, 82<<%3
dec hd
jg .y_loop_ar2
RET
.ar3:
DEFINE_ARGS buf, bufy, fg_data, uv, unused, shift
SUB rsp, 16*12
%assign stack_size_padded (stack_size_padded+16*12)
%assign stack_size (stack_size+16*12)
mov shiftd, [fg_dataq+FGData.ar_coeff_shift]
imul uvd, 28
vpbroadcastw xm14, [base+round_vals-12+shiftq*2]
pmovsxbw xm0, [fg_dataq+FGData.ar_coeffs_uv+uvq+ 0] ; cf0-7
pmovsxbw xm1, [fg_dataq+FGData.ar_coeffs_uv+uvq+ 8] ; cf8-15
pmovsxbw xm2, [fg_dataq+FGData.ar_coeffs_uv+uvq+16] ; cf16-23
pmovsxbw xm5, [fg_dataq+FGData.ar_coeffs_uv+uvq+24] ; cf24 [luma]
pshufd xm9, xm0, q1111
pshufd xm10, xm0, q2222
pshufd xm11, xm0, q3333
pshufd xm0, xm0, q0000
pshufd xm6, xm1, q1111
pshufd xm7, xm1, q2222
pshufd xm8, xm1, q3333
pshufd xm1, xm1, q0000
pshufd xm3, xm2, q1111
pshufd xm4, xm2, q2222
vpbroadcastw xm5, xm5
vpblendw xm4, xm5, 10101010b ; interleave luma cf
psrldq xm5, xm2, 10
pshufd xm2, xm2, q0000
pinsrw xm5, [base+round_vals+shiftq*2-10], 3
pmovzxwd xm14, xm14
mova [rsp+ 0*16], xm0
mova [rsp+ 1*16], xm9
mova [rsp+ 2*16], xm10
mova [rsp+ 3*16], xm11
mova [rsp+ 4*16], xm1
mova [rsp+ 5*16], xm6
mova [rsp+ 6*16], xm7
mova [rsp+ 7*16], xm8
mova [rsp+ 8*16], xm2
mova [rsp+ 9*16], xm3
mova [rsp+10*16], xm4
mova [rsp+11*16], xm5
%if %2
vpbroadcastd xm13, [base+pb_1]
vpbroadcastw xm15, [base+hmul_bits+2+%3*2]
%endif
DEFINE_ARGS buf, bufy, fg_data, h, unused, x
%if %2
sub bufq, 82*(73-35*%3)+44-(82*3+41)
%else
sub bufq, 82*70-(82-3)
%endif
add bufyq, 79+82*3
mov hd, 70-35*%3
.y_loop_ar3:
mov xq, -(76>>%2)
.x_loop_ar3:
movu xm0, [bufq+xq-82*3-3] ; y=-3,x=[-3,+12]
movu xm1, [bufq+xq-82*2-3] ; y=-2,x=[-3,+12]
movu xm2, [bufq+xq-82*1-3] ; y=-1,x=[-3,+12]
pxor xm3, xm3
pcmpgtb xm6, xm3, xm2
pcmpgtb xm5, xm3, xm1
pcmpgtb xm4, xm3, xm0
punpckhbw xm3, xm0, xm4
punpcklbw xm0, xm4
punpckhbw xm4, xm1, xm5
punpcklbw xm1, xm5
punpckhbw xm5, xm2, xm6
punpcklbw xm2, xm6
psrldq xm6, xm0, 2
psrldq xm7, xm0, 4
psrldq xm8, xm0, 6
psrldq xm9, xm0, 8
palignr xm10, xm3, xm0, 10
palignr xm11, xm3, xm0, 12
punpcklwd xm0, xm6
punpcklwd xm7, xm8
punpcklwd xm9, xm10
punpcklwd xm11, xm1
pmaddwd xm0, [rsp+ 0*16]
pmaddwd xm7, [rsp+ 1*16]
pmaddwd xm9, [rsp+ 2*16]
pmaddwd xm11, [rsp+ 3*16]
paddd xm0, xm7
paddd xm9, xm11
paddd xm0, xm9
psrldq xm6, xm1, 2
psrldq xm7, xm1, 4
psrldq xm8, xm1, 6
psrldq xm9, xm1, 8
palignr xm10, xm4, xm1, 10
palignr xm11, xm4, xm1, 12
psrldq xm12, xm2, 2
punpcklwd xm6, xm7
punpcklwd xm8, xm9
punpcklwd xm10, xm11
punpcklwd xm12, xm2, xm12
pmaddwd xm6, [rsp+ 4*16]
pmaddwd xm8, [rsp+ 5*16]
pmaddwd xm10, [rsp+ 6*16]
pmaddwd xm12, [rsp+ 7*16]
paddd xm6, xm8
paddd xm10, xm12
paddd xm6, xm10
paddd xm0, xm6
psrldq xm6, xm2, 4
psrldq xm7, xm2, 6
psrldq xm8, xm2, 8
palignr xm9, xm5, xm2, 10
palignr xm5, xm5, xm2, 12
%if %2
movq xm1, [bufyq+xq*2]
%if %3
movq xm2, [bufyq+xq*2+82]
%endif
pmaddubsw xm1, xm13, xm1
%if %3
pmaddubsw xm2, xm13, xm2
paddw xm1, xm2
%endif
pmulhrsw xm1, xm15
%else
pmovsxbw xm1, [bufyq+xq]
%endif
punpcklwd xm6, xm7
punpcklwd xm8, xm9
punpcklwd xm5, xm1
pmaddwd xm6, [rsp+ 8*16]
pmaddwd xm8, [rsp+ 9*16]
pmaddwd xm5, [rsp+10*16]
paddd xm0, xm6
paddd xm8, xm5
paddd xm0, xm8
paddd xm0, xm14
movq xm1, [bufq+xq-3] ; y=0,x=[-3,+4]
.x_loop_ar3_inner:
pmovsxbw xm1, xm1
pmaddwd xm2, xm1, [rsp+16*11]
pshufd xm3, xm2, q1111
paddd xm2, xm3 ; left+cur
paddd xm2, xm0 ; add top
psrldq xm0, 4
psrad xm2, [fg_dataq+FGData.ar_coeff_shift]
; don't packssdw, we only care about one value
pslldq xm2, 6
vpblendw xm1, xm2, 1000b
packsswb xm1, xm1
pextrb [bufq+xq], xm1, 3
psrldq xm1, 1
inc xq
jz .x_loop_ar3_end
test xq, 3
jnz .x_loop_ar3_inner
jmp .x_loop_ar3
.x_loop_ar3_end:
add bufq, 82
add bufyq, 82<<%3
dec hd
jg .y_loop_ar3
RET
%endmacro
generate_grain_uv_fn 420, 1, 1
generate_grain_uv_fn 422, 1, 0
generate_grain_uv_fn 444, 0, 0
INIT_YMM avx2
cglobal fgy_32x32xn, 6, 13, 16, dst, src, stride, fg_data, w, scaling, grain_lut
pcmpeqw m10, m10
psrld m10, 24
mov r7d, [fg_dataq+FGData.scaling_shift]
lea r8, [pb_mask]
%define base r8-pb_mask
vpbroadcastw m11, [base+mul_bits+r7*2-14]
mov r7d, [fg_dataq+FGData.clip_to_restricted_range]
vpbroadcastw m12, [base+max+r7*4]
vpbroadcastw m13, [base+min+r7*2]
DEFINE_ARGS dst, src, stride, fg_data, w, scaling, grain_lut, unused, sby, see, overlap
mov overlapd, [fg_dataq+FGData.overlap_flag]
movifnidn sbyd, sbym
test sbyd, sbyd
setnz r7b
test r7b, overlapb
jnz .vertical_overlap
imul seed, sbyd, (173 << 24) | 37
add seed, (105 << 24) | 178
rol seed, 8