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
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
|
/*
* Copyright (c) 2010 The VP8 project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#include <math.h>
#include <stdlib.h>
#include "vpx_scale/yv12config.h"
#include "pragmas.h"
#define VP8_FILTER_WEIGHT 128
#define VP8_FILTER_SHIFT 7
/* static constants */
__declspec(align(16))
const static short Blur[48] =
{
16, 16, 16, 16, 16, 16, 16, 16,
16, 16, 16, 16, 16, 16, 16, 16,
64, 64, 64, 64, 64, 64, 64, 64,
16, 16, 16, 16, 16, 16, 16, 16,
16, 16, 16, 16, 16, 16, 16, 16,
0, 0, 0, 0, 0, 0, 0, 0,
};
#define RD __declspec(align(16)) __int64 rd = 0x0040004000400040;
#define R4D2 __declspec(align(16)) __int64 rd42[2] = {0x0004000400040004,0x0004000400040004};
#ifndef RELOCATEABLE
const static RD;
const static R4D2;
#endif
/* external references */
extern double vp8_gaussian(double sigma, double mu, double x);
extern short vp8_rv[];
extern int vp8_q2mbl(int x) ;
void vp8_post_proc_down_and_across_mmx
(
unsigned char *src_ptr,
unsigned char *dst_ptr,
int src_pixels_per_line,
int dst_pixels_per_line,
int rows,
int cols,
int flimit
)
{
#ifdef RELOCATEABLE
RD
R4D2
#endif
__asm
{
push ebx
lea ebx, Blur
movd mm2, flimit
punpcklwd mm2, mm2
punpckldq mm2, mm2
mov esi, src_ptr
mov edi, dst_ptr
mov ecx, DWORD PTR rows
mov eax, src_pixels_per_line ;
destination pitch?
pxor mm0, mm0 ;
mm0 = 00000000
nextrow:
xor edx, edx ;
clear out edx for use as loop counter
nextcol:
pxor mm7, mm7 ;
mm7 = 00000000
movq mm6, [ebx + 32 ] ;
mm6 = kernel 2 taps
movq mm3, [esi] ;
mm4 = r0 p0..p7
punpcklbw mm3, mm0 ;
mm3 = p0..p3
movq mm1, mm3 ;
mm1 = p0..p3
pmullw mm3, mm6 ;
mm3 *= kernel 2 modifiers
movq mm6, [ebx + 48] ;
mm6 = kernel 3 taps
movq mm5, [esi + eax] ;
mm4 = r1 p0..p7
punpcklbw mm5, mm0 ;
mm5 = r1 p0..p3
pmullw mm6, mm5 ;
mm6 *= p0..p3 * kernel 3 modifiers
paddusw mm3, mm6 ;
mm3 += mm6
;
thresholding
movq mm7, mm1 ;
mm7 = r0 p0..p3
psubusw mm7, mm5 ;
mm7 = r0 p0..p3 - r1 p0..p3
psubusw mm5, mm1 ;
mm5 = r1 p0..p3 - r0 p0..p3
paddusw mm7, mm5 ;
mm7 = abs(r0 p0..p3 - r1 p0..p3)
pcmpgtw mm7, mm2
movq mm6, [ebx + 64 ] ;
mm6 = kernel 4 modifiers
movq mm5, [esi + 2*eax] ;
mm4 = r2 p0..p7
punpcklbw mm5, mm0 ;
mm5 = r2 p0..p3
pmullw mm6, mm5 ;
mm5 *= kernel 4 modifiers
paddusw mm3, mm6 ;
mm3 += mm5
;
thresholding
movq mm6, mm1 ;
mm6 = r0 p0..p3
psubusw mm6, mm5 ;
mm6 = r0 p0..p3 - r2 p0..p3
psubusw mm5, mm1 ;
mm5 = r2 p0..p3 - r2 p0..p3
paddusw mm6, mm5 ;
mm6 = abs(r0 p0..p3 - r2 p0..p3)
pcmpgtw mm6, mm2
por mm7, mm6 ;
accumulate thresholds
neg eax
movq mm6, [ebx ] ;
kernel 0 taps
movq mm5, [esi+2*eax] ;
mm4 = r-2 p0..p7
punpcklbw mm5, mm0 ;
mm5 = r-2 p0..p3
pmullw mm6, mm5 ;
mm5 *= kernel 0 modifiers
paddusw mm3, mm6 ;
mm3 += mm5
;
thresholding
movq mm6, mm1 ;
mm6 = r0 p0..p3
psubusw mm6, mm5 ;
mm6 = p0..p3 - r-2 p0..p3
psubusw mm5, mm1 ;
mm5 = r-2 p0..p3 - p0..p3
paddusw mm6, mm5 ;
mm6 = abs(r0 p0..p3 - r-2 p0..p3)
pcmpgtw mm6, mm2
por mm7, mm6 ;
accumulate thresholds
movq mm6, [ebx + 16] ;
kernel 1 taps
movq mm4, [esi+eax] ;
mm4 = r-1 p0..p7
punpcklbw mm4, mm0 ;
mm4 = r-1 p0..p3
pmullw mm6, mm4 ;
mm4 *= kernel 1 modifiers.
paddusw mm3, mm6 ;
mm3 += mm5
;
thresholding
movq mm6, mm1 ;
mm6 = r0 p0..p3
psubusw mm6, mm4 ;
mm6 = p0..p3 - r-2 p0..p3
psubusw mm4, mm1 ;
mm5 = r-1 p0..p3 - p0..p3
paddusw mm6, mm4 ;
mm6 = abs(r0 p0..p3 - r-1 p0..p3)
pcmpgtw mm6, mm2
por mm7, mm6 ;
accumulate thresholds
paddusw mm3, rd ;
mm3 += round value
psraw mm3, VP8_FILTER_SHIFT ;
mm3 /= 128
pand mm1, mm7 ;
mm1 select vals > thresh from source
pandn mm7, mm3 ;
mm7 select vals < thresh from blurred result
paddusw mm1, mm7 ;
combination
packuswb mm1, mm0 ;
pack to bytes
movd [edi], mm1 ;
neg eax ;
pitch is positive
add esi, 4
add edi, 4
add edx, 4
cmp edx, cols
jl nextcol
// done with the all cols, start the across filtering in place
sub esi, edx
sub edi, edx
push eax
xor edx, edx
mov eax, [edi-4];
acrossnextcol:
pxor mm7, mm7 ;
mm7 = 00000000
movq mm6, [ebx + 32 ] ;
movq mm4, [edi+edx] ;
mm4 = p0..p7
movq mm3, mm4 ;
mm3 = p0..p7
punpcklbw mm3, mm0 ;
mm3 = p0..p3
movq mm1, mm3 ;
mm1 = p0..p3
pmullw mm3, mm6 ;
mm3 *= kernel 2 modifiers
movq mm6, [ebx + 48]
psrlq mm4, 8 ;
mm4 = p1..p7
movq mm5, mm4 ;
mm5 = p1..p7
punpcklbw mm5, mm0 ;
mm5 = p1..p4
pmullw mm6, mm5 ;
mm6 *= p1..p4 * kernel 3 modifiers
paddusw mm3, mm6 ;
mm3 += mm6
;
thresholding
movq mm7, mm1 ;
mm7 = p0..p3
psubusw mm7, mm5 ;
mm7 = p0..p3 - p1..p4
psubusw mm5, mm1 ;
mm5 = p1..p4 - p0..p3
paddusw mm7, mm5 ;
mm7 = abs(p0..p3 - p1..p4)
pcmpgtw mm7, mm2
movq mm6, [ebx + 64 ]
psrlq mm4, 8 ;
mm4 = p2..p7
movq mm5, mm4 ;
mm5 = p2..p7
punpcklbw mm5, mm0 ;
mm5 = p2..p5
pmullw mm6, mm5 ;
mm5 *= kernel 4 modifiers
paddusw mm3, mm6 ;
mm3 += mm5
;
thresholding
movq mm6, mm1 ;
mm6 = p0..p3
psubusw mm6, mm5 ;
mm6 = p0..p3 - p1..p4
psubusw mm5, mm1 ;
mm5 = p1..p4 - p0..p3
paddusw mm6, mm5 ;
mm6 = abs(p0..p3 - p1..p4)
pcmpgtw mm6, mm2
por mm7, mm6 ;
accumulate thresholds
movq mm6, [ebx ]
movq mm4, [edi+edx-2] ;
mm4 = p-2..p5
movq mm5, mm4 ;
mm5 = p-2..p5
punpcklbw mm5, mm0 ;
mm5 = p-2..p1
pmullw mm6, mm5 ;
mm5 *= kernel 0 modifiers
paddusw mm3, mm6 ;
mm3 += mm5
;
thresholding
movq mm6, mm1 ;
mm6 = p0..p3
psubusw mm6, mm5 ;
mm6 = p0..p3 - p1..p4
psubusw mm5, mm1 ;
mm5 = p1..p4 - p0..p3
paddusw mm6, mm5 ;
mm6 = abs(p0..p3 - p1..p4)
pcmpgtw mm6, mm2
por mm7, mm6 ;
accumulate thresholds
movq mm6, [ebx + 16]
psrlq mm4, 8 ;
mm4 = p-1..p5
punpcklbw mm4, mm0 ;
mm4 = p-1..p2
pmullw mm6, mm4 ;
mm4 *= kernel 1 modifiers.
paddusw mm3, mm6 ;
mm3 += mm5
;
thresholding
movq mm6, mm1 ;
mm6 = p0..p3
psubusw mm6, mm4 ;
mm6 = p0..p3 - p1..p4
psubusw mm4, mm1 ;
mm5 = p1..p4 - p0..p3
paddusw mm6, mm4 ;
mm6 = abs(p0..p3 - p1..p4)
pcmpgtw mm6, mm2
por mm7, mm6 ;
accumulate thresholds
paddusw mm3, rd ;
mm3 += round value
psraw mm3, VP8_FILTER_SHIFT ;
mm3 /= 128
pand mm1, mm7 ;
mm1 select vals > thresh from source
pandn mm7, mm3 ;
mm7 select vals < thresh from blurred result
paddusw mm1, mm7 ;
combination
packuswb mm1, mm0 ;
pack to bytes
mov DWORD PTR [edi+edx-4], eax ;
store previous four bytes
movd eax, mm1
add edx, 4
cmp edx, cols
jl acrossnextcol;
mov DWORD PTR [edi+edx-4], eax
pop eax
// done with this rwo
add esi, eax ;
next line
mov eax, dst_pixels_per_line ;
destination pitch?
add edi, eax ;
next destination
mov eax, src_pixels_per_line ;
destination pitch?
dec ecx ;
decrement count
jnz nextrow ;
next row
pop ebx
}
}
void vp8_post_proc_down_and_across_xmm
(
unsigned char *src_ptr,
unsigned char *dst_ptr,
int src_pixels_per_line,
int dst_pixels_per_line,
int rows,
int cols,
int flimit
)
{
#ifdef RELOCATEABLE
R4D2
#endif
__asm
{
movd xmm2, flimit
punpcklwd xmm2, xmm2
punpckldq xmm2, xmm2
punpcklqdq xmm2, xmm2
mov esi, src_ptr
mov edi, dst_ptr
mov ecx, DWORD PTR rows
mov eax, src_pixels_per_line ;
destination pitch?
pxor xmm0, xmm0 ;
mm0 = 00000000
nextrow:
xor edx, edx ;
clear out edx for use as loop counter
nextcol:
movq xmm3, QWORD PTR [esi] ;
mm4 = r0 p0..p7
punpcklbw xmm3, xmm0 ;
mm3 = p0..p3
movdqa xmm1, xmm3 ;
mm1 = p0..p3
psllw xmm3, 2 ;
movq xmm5, QWORD PTR [esi + eax] ;
mm4 = r1 p0..p7
punpcklbw xmm5, xmm0 ;
mm5 = r1 p0..p3
paddusw xmm3, xmm5 ;
mm3 += mm6
;
thresholding
movdqa xmm7, xmm1 ;
mm7 = r0 p0..p3
psubusw xmm7, xmm5 ;
mm7 = r0 p0..p3 - r1 p0..p3
psubusw xmm5, xmm1 ;
mm5 = r1 p0..p3 - r0 p0..p3
paddusw xmm7, xmm5 ;
mm7 = abs(r0 p0..p3 - r1 p0..p3)
pcmpgtw xmm7, xmm2
movq xmm5, QWORD PTR [esi + 2*eax] ;
mm4 = r2 p0..p7
punpcklbw xmm5, xmm0 ;
mm5 = r2 p0..p3
paddusw xmm3, xmm5 ;
mm3 += mm5
;
thresholding
movdqa xmm6, xmm1 ;
mm6 = r0 p0..p3
psubusw xmm6, xmm5 ;
mm6 = r0 p0..p3 - r2 p0..p3
psubusw xmm5, xmm1 ;
mm5 = r2 p0..p3 - r2 p0..p3
paddusw xmm6, xmm5 ;
mm6 = abs(r0 p0..p3 - r2 p0..p3)
pcmpgtw xmm6, xmm2
por xmm7, xmm6 ;
accumulate thresholds
neg eax
movq xmm5, QWORD PTR [esi+2*eax] ;
mm4 = r-2 p0..p7
punpcklbw xmm5, xmm0 ;
mm5 = r-2 p0..p3
paddusw xmm3, xmm5 ;
mm3 += mm5
;
thresholding
movdqa xmm6, xmm1 ;
mm6 = r0 p0..p3
psubusw xmm6, xmm5 ;
mm6 = p0..p3 - r-2 p0..p3
psubusw xmm5, xmm1 ;
mm5 = r-2 p0..p3 - p0..p3
paddusw xmm6, xmm5 ;
mm6 = abs(r0 p0..p3 - r-2 p0..p3)
pcmpgtw xmm6, xmm2
por xmm7, xmm6 ;
accumulate thresholds
movq xmm4, QWORD PTR [esi+eax] ;
mm4 = r-1 p0..p7
punpcklbw xmm4, xmm0 ;
mm4 = r-1 p0..p3
paddusw xmm3, xmm4 ;
mm3 += mm5
;
thresholding
movdqa xmm6, xmm1 ;
mm6 = r0 p0..p3
psubusw xmm6, xmm4 ;
mm6 = p0..p3 - r-2 p0..p3
psubusw xmm4, xmm1 ;
mm5 = r-1 p0..p3 - p0..p3
paddusw xmm6, xmm4 ;
mm6 = abs(r0 p0..p3 - r-1 p0..p3)
pcmpgtw xmm6, xmm2
por xmm7, xmm6 ;
accumulate thresholds
paddusw xmm3, rd42 ;
mm3 += round value
psraw xmm3, 3 ;
mm3 /= 8
pand xmm1, xmm7 ;
mm1 select vals > thresh from source
pandn xmm7, xmm3 ;
mm7 select vals < thresh from blurred result
paddusw xmm1, xmm7 ;
combination
packuswb xmm1, xmm0 ;
pack to bytes
movq QWORD PTR [edi], xmm1 ;
neg eax ;
pitch is positive
add esi, 8
add edi, 8
add edx, 8
cmp edx, cols
jl nextcol
// done with the all cols, start the across filtering in place
sub esi, edx
sub edi, edx
xor edx, edx
movq mm0, QWORD PTR [edi-8];
acrossnextcol:
movq xmm7, QWORD PTR [edi +edx -2]
movd xmm4, DWORD PTR [edi +edx +6]
pslldq xmm4, 8
por xmm4, xmm7
movdqa xmm3, xmm4
psrldq xmm3, 2
punpcklbw xmm3, xmm0 ;
mm3 = p0..p3
movdqa xmm1, xmm3 ;
mm1 = p0..p3
psllw xmm3, 2
movdqa xmm5, xmm4
psrldq xmm5, 3
punpcklbw xmm5, xmm0 ;
mm5 = p1..p4
paddusw xmm3, xmm5 ;
mm3 += mm6
;
thresholding
movdqa xmm7, xmm1 ;
mm7 = p0..p3
psubusw xmm7, xmm5 ;
mm7 = p0..p3 - p1..p4
psubusw xmm5, xmm1 ;
mm5 = p1..p4 - p0..p3
paddusw xmm7, xmm5 ;
mm7 = abs(p0..p3 - p1..p4)
pcmpgtw xmm7, xmm2
movdqa xmm5, xmm4
psrldq xmm5, 4
punpcklbw xmm5, xmm0 ;
mm5 = p2..p5
paddusw xmm3, xmm5 ;
mm3 += mm5
;
thresholding
movdqa xmm6, xmm1 ;
mm6 = p0..p3
psubusw xmm6, xmm5 ;
mm6 = p0..p3 - p1..p4
psubusw xmm5, xmm1 ;
mm5 = p1..p4 - p0..p3
paddusw xmm6, xmm5 ;
mm6 = abs(p0..p3 - p1..p4)
pcmpgtw xmm6, xmm2
por xmm7, xmm6 ;
accumulate thresholds
movdqa xmm5, xmm4 ;
mm5 = p-2..p5
punpcklbw xmm5, xmm0 ;
mm5 = p-2..p1
paddusw xmm3, xmm5 ;
mm3 += mm5
;
thresholding
movdqa xmm6, xmm1 ;
mm6 = p0..p3
psubusw xmm6, xmm5 ;
mm6 = p0..p3 - p1..p4
psubusw xmm5, xmm1 ;
mm5 = p1..p4 - p0..p3
paddusw xmm6, xmm5 ;
mm6 = abs(p0..p3 - p1..p4)
pcmpgtw xmm6, xmm2
por xmm7, xmm6 ;
accumulate thresholds
psrldq xmm4, 1 ;
mm4 = p-1..p5
punpcklbw xmm4, xmm0 ;
mm4 = p-1..p2
paddusw xmm3, xmm4 ;
mm3 += mm5
;
thresholding
movdqa xmm6, xmm1 ;
mm6 = p0..p3
psubusw xmm6, xmm4 ;
mm6 = p0..p3 - p1..p4
psubusw xmm4, xmm1 ;
mm5 = p1..p4 - p0..p3
paddusw xmm6, xmm4 ;
mm6 = abs(p0..p3 - p1..p4)
pcmpgtw xmm6, xmm2
por xmm7, xmm6 ;
accumulate thresholds
paddusw xmm3, rd42 ;
mm3 += round value
psraw xmm3, 3 ;
mm3 /= 8
pand xmm1, xmm7 ;
mm1 select vals > thresh from source
pandn xmm7, xmm3 ;
mm7 select vals < thresh from blurred result
paddusw xmm1, xmm7 ;
combination
packuswb xmm1, xmm0 ;
pack to bytes
movq QWORD PTR [edi+edx-8], mm0 ;
store previous four bytes
movdq2q mm0, xmm1
add edx, 8
cmp edx, cols
jl acrossnextcol;
// last 8 pixels
movq QWORD PTR [edi+edx-8], mm0
// done with this rwo
add esi, eax ;
next line
mov eax, dst_pixels_per_line ;
destination pitch?
add edi, eax ;
next destination
mov eax, src_pixels_per_line ;
destination pitch?
dec ecx ;
decrement count
jnz nextrow ;
next row
}
}
void vp8_mbpost_proc_down_mmx(unsigned char *dst, int pitch, int rows, int cols, int flimit)
{
int c, i;
__declspec(align(16))
int flimit2[2];
__declspec(align(16))
unsigned char d[16][8];
flimit = vp8_q2mbl(flimit);
for (i = 0; i < 2; i++)
flimit2[i] = flimit;
rows += 8;
for (c = 0; c < cols; c += 4)
{
unsigned char *s = &dst[c];
__asm
{
mov esi, s ;
pxor mm0, mm0 ;
mov eax, pitch ;
neg eax // eax = -pitch
lea esi, [esi + eax*8]; // edi = s[-pitch*8]
neg eax
pxor mm5, mm5
pxor mm6, mm6 ;
pxor mm7, mm7 ;
mov edi, esi
mov ecx, 15 ;
loop_initvar:
movd mm1, DWORD PTR [edi];
punpcklbw mm1, mm0 ;
paddw mm5, mm1 ;
pmullw mm1, mm1 ;
movq mm2, mm1 ;
punpcklwd mm1, mm0 ;
punpckhwd mm2, mm0 ;
paddd mm6, mm1 ;
paddd mm7, mm2 ;
lea edi, [edi+eax] ;
dec ecx
jne loop_initvar
//save the var and sum
xor edx, edx
loop_row:
movd mm1, DWORD PTR [esi] // [s-pitch*8]
movd mm2, DWORD PTR [edi] // [s+pitch*7]
punpcklbw mm1, mm0
punpcklbw mm2, mm0
paddw mm5, mm2
psubw mm5, mm1
pmullw mm2, mm2
movq mm4, mm2
punpcklwd mm2, mm0
punpckhwd mm4, mm0
paddd mm6, mm2
paddd mm7, mm4
pmullw mm1, mm1
movq mm2, mm1
punpcklwd mm1, mm0
psubd mm6, mm1
punpckhwd mm2, mm0
psubd mm7, mm2
movq mm3, mm6
pslld mm3, 4
psubd mm3, mm6
movq mm1, mm5
movq mm4, mm5
pmullw mm1, mm1
pmulhw mm4, mm4
movq mm2, mm1
punpcklwd mm1, mm4
punpckhwd mm2, mm4
movq mm4, mm7
pslld mm4, 4
psubd mm4, mm7
psubd mm3, mm1
psubd mm4, mm2
psubd mm3, flimit2
psubd mm4, flimit2
psrad mm3, 31
psrad mm4, 31
packssdw mm3, mm4
packsswb mm3, mm0
movd mm1, DWORD PTR [esi+eax*8]
movq mm2, mm1
punpcklbw mm1, mm0
paddw mm1, mm5
mov ecx, edx
and ecx, 127
movq mm4, vp8_rv[ecx*2]
paddw mm1, mm4
//paddw xmm1, eight8s
psraw mm1, 4
packuswb mm1, mm0
pand mm1, mm3
pandn mm3, mm2
por mm1, mm3
and ecx, 15
movd DWORD PTR d[ecx*4], mm1
mov ecx, edx
sub ecx, 8
and ecx, 15
movd mm1, DWORD PTR d[ecx*4]
movd [esi], mm1
lea esi, [esi+eax]
lea edi, [edi+eax]
add edx, 1
cmp edx, rows
jl loop_row
}
}
}
void vp8_mbpost_proc_down_xmm(unsigned char *dst, int pitch, int rows, int cols, int flimit)
{
int c, i;
__declspec(align(16))
int flimit4[4];
__declspec(align(16))
unsigned char d[16][8];
flimit = vp8_q2mbl(flimit);
for (i = 0; i < 4; i++)
flimit4[i] = flimit;
rows += 8;
for (c = 0; c < cols; c += 8)
{
unsigned char *s = &dst[c];
__asm
{
mov esi, s ;
pxor xmm0, xmm0 ;
mov eax, pitch ;
neg eax // eax = -pitch
lea esi, [esi + eax*8]; // edi = s[-pitch*8]
neg eax
pxor xmm5, xmm5
pxor xmm6, xmm6 ;
pxor xmm7, xmm7 ;
mov edi, esi
mov ecx, 15 ;
loop_initvar:
movq xmm1, QWORD PTR [edi];
punpcklbw xmm1, xmm0 ;
paddw xmm5, xmm1 ;
pmullw xmm1, xmm1 ;
movdqa xmm2, xmm1 ;
punpcklwd xmm1, xmm0 ;
punpckhwd xmm2, xmm0 ;
paddd xmm6, xmm1 ;
paddd xmm7, xmm2 ;
lea edi, [edi+eax] ;
dec ecx
jne loop_initvar
//save the var and sum
xor edx, edx
loop_row:
movq xmm1, QWORD PTR [esi] // [s-pitch*8]
movq xmm2, QWORD PTR [edi] // [s+pitch*7]
punpcklbw xmm1, xmm0
punpcklbw xmm2, xmm0
paddw xmm5, xmm2
psubw xmm5, xmm1
pmullw xmm2, xmm2
movdqa xmm4, xmm2
punpcklwd xmm2, xmm0
punpckhwd xmm4, xmm0
paddd xmm6, xmm2
paddd xmm7, xmm4
pmullw xmm1, xmm1
movdqa xmm2, xmm1
punpcklwd xmm1, xmm0
psubd xmm6, xmm1
punpckhwd xmm2, xmm0
psubd xmm7, xmm2
movdqa xmm3, xmm6
pslld xmm3, 4
psubd xmm3, xmm6
movdqa xmm1, xmm5
movdqa xmm4, xmm5
pmullw xmm1, xmm1
pmulhw xmm4, xmm4
movdqa xmm2, xmm1
punpcklwd xmm1, xmm4
punpckhwd xmm2, xmm4
movdqa xmm4, xmm7
pslld xmm4, 4
psubd xmm4, xmm7
psubd xmm3, xmm1
psubd xmm4, xmm2
psubd xmm3, flimit4
psubd xmm4, flimit4
psrad xmm3, 31
psrad xmm4, 31
packssdw xmm3, xmm4
packsswb xmm3, xmm0
movq xmm1, QWORD PTR [esi+eax*8]
movq xmm2, xmm1
punpcklbw xmm1, xmm0
paddw xmm1, xmm5
mov ecx, edx
and ecx, 127
movdqu xmm4, vp8_rv[ecx*2]
paddw xmm1, xmm4
//paddw xmm1, eight8s
psraw xmm1, 4
packuswb xmm1, xmm0
pand xmm1, xmm3
pandn xmm3, xmm2
por xmm1, xmm3
and ecx, 15
movq QWORD PTR d[ecx*8], xmm1
mov ecx, edx
sub ecx, 8
and ecx, 15
movq mm0, d[ecx*8]
movq [esi], mm0
lea esi, [esi+eax]
lea edi, [edi+eax]
add edx, 1
cmp edx, rows
jl loop_row
}
}
}
#if 0
/****************************************************************************
*
* ROUTINE : plane_add_noise_wmt
*
* INPUTS : unsigned char *Start starting address of buffer to add gaussian
* noise to
* unsigned int Width width of plane
* unsigned int Height height of plane
* int Pitch distance between subsequent lines of frame
* int q quantizer used to determine amount of noise
* to add
*
* OUTPUTS : None.
*
* RETURNS : void.
*
* FUNCTION : adds gaussian noise to a plane of pixels
*
* SPECIAL NOTES : None.
*
****************************************************************************/
void vp8_plane_add_noise_wmt(unsigned char *Start, unsigned int Width, unsigned int Height, int Pitch, int q, int a)
{
unsigned int i;
__declspec(align(16)) unsigned char blackclamp[16];
__declspec(align(16)) unsigned char whiteclamp[16];
__declspec(align(16)) unsigned char bothclamp[16];
char char_dist[300];
char Rand[2048];
double sigma;
// return;
__asm emms
sigma = a + .5 + .6 * (63 - q) / 63.0;
// set up a lookup table of 256 entries that matches
// a gaussian distribution with sigma determined by q.
//
{
double i;
int next, j;
next = 0;
for (i = -32; i < 32; i++)
{
double g = 256 * vp8_gaussian(sigma, 0, 1.0 * i);
int a = (int)(g + .5);
if (a)
{
for (j = 0; j < a; j++)
{
char_dist[next+j] = (char) i;
}
next = next + j;
}
}
for (next = next; next < 256; next++)
char_dist[next] = 0;
}
for (i = 0; i < 2048; i++)
{
Rand[i] = char_dist[rand() & 0xff];
}
for (i = 0; i < 16; i++)
{
blackclamp[i] = -char_dist[0];
whiteclamp[i] = -char_dist[0];
bothclamp[i] = -2 * char_dist[0];
}
for (i = 0; i < Height; i++)
{
unsigned char *Pos = Start + i * Pitch;
char *Ref = Rand + (rand() & 0xff);
__asm
{
mov ecx, [Width]
mov esi, Pos
mov edi, Ref
xor eax, eax
nextset:
movdqu xmm1, [esi+eax] // get the source
psubusb xmm1, blackclamp // clamp both sides so we don't outrange adding noise
paddusb xmm1, bothclamp
psubusb xmm1, whiteclamp
movdqu xmm2, [edi+eax] // get the noise for this line
paddb xmm1, xmm2 // add it in
movdqu [esi+eax], xmm1 // store the result
add eax, 16 // move to the next line
cmp eax, ecx
jl nextset
}
}
}
#endif
__declspec(align(16))
static const int four8s[4] = { 8, 8, 8, 8};
void vp8_mbpost_proc_across_ip_xmm(unsigned char *src, int pitch, int rows, int cols, int flimit)
{
int r, i;
__declspec(align(16))
int flimit4[4];
unsigned char *s = src;
int sumsq;
int sum;
flimit = vp8_q2mbl(flimit);
flimit4[0] =
flimit4[1] =
flimit4[2] =
flimit4[3] = flimit;
for (r = 0; r < rows; r++)
{
sumsq = 0;
sum = 0;
for (i = -8; i <= 6; i++)
{
sumsq += s[i] * s[i];
sum += s[i];
}
__asm
{
mov eax, sumsq
movd xmm7, eax
mov eax, sum
movd xmm6, eax
mov esi, s
xor ecx, ecx
mov edx, cols
add edx, 8
pxor mm0, mm0
pxor mm1, mm1
pxor xmm0, xmm0
nextcol4:
movd xmm1, DWORD PTR [esi+ecx-8] // -8 -7 -6 -5
movd xmm2, DWORD PTR [esi+ecx+7] // +7 +8 +9 +10
punpcklbw xmm1, xmm0 // expanding
punpcklbw xmm2, xmm0 // expanding
punpcklwd xmm1, xmm0 // expanding to dwords
punpcklwd xmm2, xmm0 // expanding to dwords
psubd xmm2, xmm1 // 7--8 8--7 9--6 10--5
paddd xmm1, xmm1 // -8*2 -7*2 -6*2 -5*2
paddd xmm1, xmm2 // 7+-8 8+-7 9+-6 10+-5
pmaddwd xmm1, xmm2 // squared of 7+-8 8+-7 9+-6 10+-5
paddd xmm6, xmm2
paddd xmm7, xmm1
pshufd xmm6, xmm6, 0 // duplicate the last ones
pshufd xmm7, xmm7, 0 // duplicate the last ones
psrldq xmm1, 4 // 8--7 9--6 10--5 0000
psrldq xmm2, 4 // 8--7 9--6 10--5 0000
pshufd xmm3, xmm1, 3 // 0000 8--7 8--7 8--7 squared
pshufd xmm4, xmm2, 3 // 0000 8--7 8--7 8--7 squared
paddd xmm6, xmm4
paddd xmm7, xmm3
pshufd xmm3, xmm1, 01011111b // 0000 0000 9--6 9--6 squared
pshufd xmm4, xmm2, 01011111b // 0000 0000 9--6 9--6 squared
paddd xmm7, xmm3
paddd xmm6, xmm4
pshufd xmm3, xmm1, 10111111b // 0000 0000 8--7 8--7 squared
pshufd xmm4, xmm2, 10111111b // 0000 0000 8--7 8--7 squared
paddd xmm7, xmm3
paddd xmm6, xmm4
movdqa xmm3, xmm6
pmaddwd xmm3, xmm3
movdqa xmm5, xmm7
pslld xmm5, 4
psubd xmm5, xmm7
psubd xmm5, xmm3
psubd xmm5, flimit4
psrad xmm5, 31
packssdw xmm5, xmm0
packsswb xmm5, xmm0
movd xmm1, DWORD PTR [esi+ecx]
movq xmm2, xmm1
punpcklbw xmm1, xmm0
punpcklwd xmm1, xmm0
paddd xmm1, xmm6
paddd xmm1, four8s
psrad xmm1, 4
packssdw xmm1, xmm0
packuswb xmm1, xmm0
pand xmm1, xmm5
pandn xmm5, xmm2
por xmm5, xmm1
movd [esi+ecx-8], mm0
movq mm0, mm1
movdq2q mm1, xmm5
psrldq xmm7, 12
psrldq xmm6, 12
add ecx, 4
cmp ecx, edx
jl nextcol4
}
s += pitch;
}
}
#if 0
/****************************************************************************
*
* ROUTINE : plane_add_noise_mmx
*
* INPUTS : unsigned char *Start starting address of buffer to add gaussian
* noise to
* unsigned int Width width of plane
* unsigned int Height height of plane
* int Pitch distance between subsequent lines of frame
* int q quantizer used to determine amount of noise
* to add
*
* OUTPUTS : None.
*
* RETURNS : void.
*
* FUNCTION : adds gaussian noise to a plane of pixels
*
* SPECIAL NOTES : None.
*
****************************************************************************/
void vp8_plane_add_noise_mmx(unsigned char *Start, unsigned int Width, unsigned int Height, int Pitch, int q, int a)
{
unsigned int i;
int Pitch4 = Pitch * 4;
const int noise_amount = 2;
const int noise_adder = 2 * noise_amount + 1;
__declspec(align(16)) unsigned char blackclamp[16];
__declspec(align(16)) unsigned char whiteclamp[16];
__declspec(align(16)) unsigned char bothclamp[16];
char char_dist[300];
char Rand[2048];
double sigma;
__asm emms
sigma = a + .5 + .6 * (63 - q) / 63.0;
// set up a lookup table of 256 entries that matches
// a gaussian distribution with sigma determined by q.
//
{
double i, sum = 0;
int next, j;
next = 0;
for (i = -32; i < 32; i++)
{
int a = (int)(.5 + 256 * vp8_gaussian(sigma, 0, i));
if (a)
{
for (j = 0; j < a; j++)
{
char_dist[next+j] = (char) i;
}
next = next + j;
}
}
for (next = next; next < 256; next++)
char_dist[next] = 0;
}
for (i = 0; i < 2048; i++)
{
Rand[i] = char_dist[rand() & 0xff];
}
for (i = 0; i < 16; i++)
{
blackclamp[i] = -char_dist[0];
whiteclamp[i] = -char_dist[0];
bothclamp[i] = -2 * char_dist[0];
}
for (i = 0; i < Height; i++)
{
unsigned char *Pos = Start + i * Pitch;
char *Ref = Rand + (rand() & 0xff);
__asm
{
mov ecx, [Width]
mov esi, Pos
mov edi, Ref
xor eax, eax
nextset:
movq mm1, [esi+eax] // get the source
psubusb mm1, blackclamp // clamp both sides so we don't outrange adding noise
paddusb mm1, bothclamp
psubusb mm1, whiteclamp
movq mm2, [edi+eax] // get the noise for this line
paddb mm1, mm2 // add it in
movq [esi+eax], mm1 // store the result
add eax, 8 // move to the next line
cmp eax, ecx
jl nextset
}
}
}
#else
extern char an[8][64][3072];
extern int cd[8][64];
void vp8_plane_add_noise_mmx(unsigned char *Start, unsigned int Width, unsigned int Height, int Pitch, int q, int a)
{
unsigned int i;
__declspec(align(16)) unsigned char blackclamp[16];
__declspec(align(16)) unsigned char whiteclamp[16];
__declspec(align(16)) unsigned char bothclamp[16];
__asm emms
for (i = 0; i < 16; i++)
{
blackclamp[i] = -cd[a][q];
whiteclamp[i] = -cd[a][q];
bothclamp[i] = -2 * cd[a][q];
}
for (i = 0; i < Height; i++)
{
unsigned char *Pos = Start + i * Pitch;
char *Ref = an[a][q] + (rand() & 0xff);
__asm
{
mov ecx, [Width]
mov esi, Pos
mov edi, Ref
xor eax, eax
nextset:
movq mm1, [esi+eax] // get the source
psubusb mm1, blackclamp // clamp both sides so we don't outrange adding noise
paddusb mm1, bothclamp
psubusb mm1, whiteclamp
movq mm2, [edi+eax] // get the noise for this line
paddb mm1, mm2 // add it in
movq [esi+eax], mm1 // store the result
add eax, 8 // move to the next line
cmp eax, ecx
jl nextset
}
}
}
void vp8_plane_add_noise_wmt(unsigned char *Start, unsigned int Width, unsigned int Height, int Pitch, int q, int a)
{
unsigned int i;
__declspec(align(16)) unsigned char blackclamp[16];
__declspec(align(16)) unsigned char whiteclamp[16];
__declspec(align(16)) unsigned char bothclamp[16];
__asm emms
for (i = 0; i < 16; i++)
{
blackclamp[i] = -cd[a][q];
whiteclamp[i] = -cd[a][q];
bothclamp[i] = -2 * cd[a][q];
}
for (i = 0; i < Height; i++)
{
unsigned char *Pos = Start + i * Pitch;
char *Ref = an[a][q] + (rand() & 0xff);
__asm
{
mov ecx, [Width]
mov esi, Pos
mov edi, Ref
xor eax, eax
nextset:
movdqu xmm1, [esi+eax] // get the source
psubusb xmm1, blackclamp // clamp both sides so we don't outrange adding noise
paddusb xmm1, bothclamp
psubusb xmm1, whiteclamp
movdqu xmm2, [edi+eax] // get the noise for this line
paddb xmm1, xmm2 // add it in
movdqu [esi+eax], xmm1 // store the result
add eax, 16 // move to the next line
cmp eax, ecx
jl nextset
}
}
}
#endif
|
