File: jdsample.c

package info (click to toggle)
libjpeg-mmx 0.1.3-3
  • links: PTS
  • area: main
  • in suites: woody
  • size: 2,700 kB
  • ctags: 2,672
  • sloc: ansic: 24,900; sh: 4,565; makefile: 59
file content (1922 lines) | stat: -rw-r--r-- 78,793 bytes parent folder | download | duplicates (2)
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
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
/*
 * jdsample.c
 *
 * Copyright (C) 1991-1996, Thomas G. Lane.
 * This file is part of the Independent JPEG Group's software.
 * For conditions of distribution and use, see the accompanying README file.
 *
 * This file contains upsampling routines.
 *
 * Upsampling input data is counted in "row groups".  A row group
 * is defined to be (v_samp_factor * DCT_scaled_size / min_DCT_scaled_size)
 * sample rows of each component.  Upsampling will normally produce
 * max_v_samp_factor pixel rows from each row group (but this could vary
 * if the upsampler is applying a scale factor of its own).
 *
 * An excellent reference for image resampling is
 *   Digital Image Warping, George Wolberg, 1990.
 *   Pub. by IEEE Computer Society Press, Los Alamitos, CA. ISBN 0-8186-8944-7.
 */

#define JPEG_INTERNALS
#include "jinclude.h"
#include "jpeglib.h"


/* Pointer to routine to upsample a single component */
typedef JMETHOD(void, upsample1_ptr,
		(j_decompress_ptr cinfo, jpeg_component_info * compptr,
		 JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr));

/* Private subobject */

typedef struct {
  struct jpeg_upsampler pub;	/* public fields */

  /* Color conversion buffer.  When using separate upsampling and color
   * conversion steps, this buffer holds one upsampled row group until it
   * has been color converted and output.
   * Note: we do not allocate any storage for component(s) which are full-size,
   * ie do not need rescaling.  The corresponding entry of color_buf[] is
   * simply set to point to the input data array, thereby avoiding copying.
   */
  JSAMPARRAY color_buf[MAX_COMPONENTS];

  /* Per-component upsampling method pointers */
  upsample1_ptr methods[MAX_COMPONENTS];

  int next_row_out;		/* counts rows emitted from color_buf */
  JDIMENSION rows_to_go;	/* counts rows remaining in image */

  /* Height of an input row group for each component. */
  int rowgroup_height[MAX_COMPONENTS];

  /* These arrays save pixel expansion factors so that int_expand need not
   * recompute them each time.  They are unused for other upsampling methods.
   */
  UINT8 h_expand[MAX_COMPONENTS];
  UINT8 v_expand[MAX_COMPONENTS];
} my_upsampler;

typedef my_upsampler * my_upsample_ptr;


/*
 * Initialize for an upsampling pass.
 */

METHODDEF(void)
start_pass_upsample (j_decompress_ptr cinfo)
{
  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;

  /* Mark the conversion buffer empty */
  upsample->next_row_out = cinfo->max_v_samp_factor;
  /* Initialize total-height counter for detecting bottom of image */
  upsample->rows_to_go = cinfo->output_height;
}


/*
 * Control routine to do upsampling (and color conversion).
 *
 * In this version we upsample each component independently.
 * We upsample one row group into the conversion buffer, then apply
 * color conversion a row at a time.
 */

METHODDEF(void)
sep_upsample (j_decompress_ptr cinfo,
	      JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
	      JDIMENSION in_row_groups_avail,
	      JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
	      JDIMENSION out_rows_avail)
{
  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
  int ci;
  jpeg_component_info * compptr;
  JDIMENSION num_rows;

  /* Fill the conversion buffer, if it's empty */
  if (upsample->next_row_out >= cinfo->max_v_samp_factor) {
    for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
	 ci++, compptr++) {
      /* Invoke per-component upsample method.  Notice we pass a POINTER
       * to color_buf[ci], so that fullsize_upsample can change it.
       */
      (*upsample->methods[ci]) (cinfo, compptr,
	input_buf[ci] + (*in_row_group_ctr * upsample->rowgroup_height[ci]),
	upsample->color_buf + ci);
    }
    upsample->next_row_out = 0;
  }

  /* Color-convert and emit rows */

  /* How many we have in the buffer: */
  num_rows = (JDIMENSION) (cinfo->max_v_samp_factor - upsample->next_row_out);
  /* Not more than the distance to the end of the image.  Need this test
   * in case the image height is not a multiple of max_v_samp_factor:
   */
  if (num_rows > upsample->rows_to_go) 
    num_rows = upsample->rows_to_go;
  /* And not more than what the client can accept: */
  out_rows_avail -= *out_row_ctr;
  if (num_rows > out_rows_avail)
    num_rows = out_rows_avail;

  (*cinfo->cconvert->color_convert) (cinfo, upsample->color_buf,
				     (JDIMENSION) upsample->next_row_out,
				     output_buf + *out_row_ctr,
				     (int) num_rows);

  /* Adjust counts */
  *out_row_ctr += num_rows;
  upsample->rows_to_go -= num_rows;
  upsample->next_row_out += num_rows;
  /* When the buffer is emptied, declare this input row group consumed */
  if (upsample->next_row_out >= cinfo->max_v_samp_factor)
    (*in_row_group_ctr)++;
}


/*
 * These are the routines invoked by sep_upsample to upsample pixel values
 * of a single component.  One row group is processed per call.
 */


/*
 * For full-size components, we just make color_buf[ci] point at the
 * input buffer, and thus avoid copying any data.  Note that this is
 * safe only because sep_upsample doesn't declare the input row group
 * "consumed" until we are done color converting and emitting it.
 */

METHODDEF(void)
fullsize_upsample (j_decompress_ptr cinfo, jpeg_component_info * compptr,
		   JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr)
{
  *output_data_ptr = input_data;
}


/*
 * This is a no-op version used for "uninteresting" components.
 * These components will not be referenced by color conversion.
 */

METHODDEF(void)
noop_upsample (j_decompress_ptr cinfo, jpeg_component_info * compptr,
	       JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr)
{
  *output_data_ptr = NULL;	/* safety check */
}


/*
 * This version handles any integral sampling ratios.
 * This is not used for typical JPEG files, so it need not be fast.
 * Nor, for that matter, is it particularly accurate: the algorithm is
 * simple replication of the input pixel onto the corresponding output
 * pixels.  The hi-falutin sampling literature refers to this as a
 * "box filter".  A box filter tends to introduce visible artifacts,
 * so if you are actually going to use 3:1 or 4:1 sampling ratios
 * you would be well advised to improve this code.
 */

METHODDEF(void)
int_upsample (j_decompress_ptr cinfo, jpeg_component_info * compptr,
	      JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr)
{
  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
  JSAMPARRAY output_data = *output_data_ptr;
  register JSAMPROW inptr, outptr;
  register JSAMPLE invalue;
  register int h;
  JSAMPROW outend;
  int h_expand, v_expand;
  int inrow, outrow;

  h_expand = upsample->h_expand[compptr->component_index];
  v_expand = upsample->v_expand[compptr->component_index];

  inrow = outrow = 0;
  while (outrow < cinfo->max_v_samp_factor) {
    /* Generate one output row with proper horizontal expansion */
    inptr = input_data[inrow];
    outptr = output_data[outrow];
    outend = outptr + cinfo->output_width;
    while (outptr < outend) {
      invalue = *inptr++;	/* don't need GETJSAMPLE() here */
      for (h = h_expand; h > 0; h--) {
	*outptr++ = invalue;
      }
    }
    /* Generate any additional output rows by duplicating the first one */
    if (v_expand > 1) {
      jcopy_sample_rows(output_data, outrow, output_data, outrow+1,
			v_expand-1, cinfo->output_width);
    }
    inrow++;
    outrow += v_expand;
  }
}


/*
 * Fast processing for the common case of 2:1 horizontal and 1:1 vertical.
 * It's still a box filter.
 */

METHODDEF(void)
h2v1_upsample (j_decompress_ptr cinfo, jpeg_component_info * compptr,
	       JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr)
{
  JSAMPARRAY output_data = *output_data_ptr;
  register JSAMPROW inptr, outptr;
  register JSAMPLE invalue;
  JSAMPROW outend;
  int inrow;

  for (inrow = 0; inrow < cinfo->max_v_samp_factor; inrow++) {
    inptr = input_data[inrow];
    outptr = output_data[inrow];
    outend = outptr + cinfo->output_width;
    while (outptr < outend) {
      invalue = *inptr++;	/* don't need GETJSAMPLE() here */
      *outptr++ = invalue;
      *outptr++ = invalue;
    }
  }
}


/*
 * Fast processing for the common case of 2:1 horizontal and 2:1 vertical.
 * It's still a box filter.
 */

METHODDEF(void)
h2v2_upsample (j_decompress_ptr cinfo, jpeg_component_info * compptr,
	       JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr)
{
  JSAMPARRAY output_data = *output_data_ptr;
  register JSAMPROW inptr, outptr;
  register JSAMPLE invalue;
  JSAMPROW outend;
  int inrow, outrow;

  inrow = outrow = 0;
  while (outrow < cinfo->max_v_samp_factor) {
    inptr = input_data[inrow];
    outptr = output_data[outrow];
    outend = outptr + cinfo->output_width;
    while (outptr < outend) {
      invalue = *inptr++;	/* don't need GETJSAMPLE() here */
      *outptr++ = invalue;
      *outptr++ = invalue;
    }
    jcopy_sample_rows(output_data, outrow, output_data, outrow+1,
		      1, cinfo->output_width);
    inrow++;
    outrow += 2;
  }
}


/*
 * Fancy processing for the common case of 2:1 horizontal and 1:1 vertical.
 *
 * The upsampling algorithm is linear interpolation between pixel centers,
 * also known as a "triangle filter".  This is a good compromise between
 * speed and visual quality.  The centers of the output pixels are 1/4 and 3/4
 * of the way between input pixel centers.
 *
 * A note about the "bias" calculations: when rounding fractional values to
 * integer, we do not want to always round 0.5 up to the next integer.
 * If we did that, we'd introduce a noticeable bias towards larger values.
 * Instead, this code is arranged so that 0.5 will be rounded up or down at
 * alternate pixel locations (a simple ordered dither pattern).
 */

#if defined(HAVE_MMX_INTEL_MNEMONICS) || defined(HAVE_MMX_ATT_MNEMONICS)   
#define __int64 long long /* This won't work for Intel compilers - tell Gernot to help fixing ! */ 

/* I have no clue why it is written in that strange way, but ok, it works */
union u1 { __int64 q; double align; }
mul3w={0x0003000300030003}, mul9w={0x0009000900090009},
    mul9ws={0x000900090009000c},    mul3ws={0x0003000300030004},
      bias7w={0x0007000700070007},    bias8w={0x0008000800080008},
	bias1w={0x0001000100010001},    bias2w={0x0002000200020002},
	  mask1={0xFF00000000000000},     mask2={0x00000000000000FF},
	    noval = {0}, input0 = {0}, input1 = {0};

/* Silly forward definitions */
METHODDEF(void)
h2v1_fancy_upsample_mmx (j_decompress_ptr cinfo, jpeg_component_info * compptr,
			 JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr);
METHODDEF(void)
h2v1_fancy_upsample_orig (j_decompress_ptr cinfo, jpeg_component_info * compptr,
			  JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr);

METHODDEF(void)
h2v1_fancy_upsample (j_decompress_ptr cinfo, jpeg_component_info * compptr,
		     JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr)
{
  if (MMXAvailable)
    h2v1_fancy_upsample_mmx(cinfo, compptr, input_data, output_data_ptr);
  else
    h2v1_fancy_upsample_orig(cinfo, compptr, input_data, output_data_ptr);
}

METHODDEF(void)
h2v1_fancy_upsample_mmx (j_decompress_ptr cinfo, jpeg_component_info * compptr,
		     JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr)
{
  JSAMPARRAY output_data = *output_data_ptr;
#if defined(HAVE_MMX_INTEL_MNEMONICS) /* see in h2v2 for comments */
   register JSAMPROW inptr, outptr;
#else
   JSAMPROW inptr, outptr;
#endif
  int inrow, hsize = compptr->downsampled_width;

  for (inrow = 0; inrow < cinfo->max_v_samp_factor; inrow++) {
    inptr = input_data[inrow];
    outptr = output_data[inrow];

#if defined(HAVE_MMX_INTEL_MNEMONICS) 
	_asm {
			mov ecx, hsize			;// horizontal line size
			mov esi, inptr			;// input buffer pointer
			mov edi, outptr			;// output buffer pointer

			pxor mm6, mm6			;// zero register
			movq mm7, [esi]			;// input register

			;// Special 1st column case - process low 8 bytes of mm7
			movq mm0, mm7			;// move 1st quadword into mm7
			movq mm1, mm7			;// make a copy
			movq mm2, mm7			;// make a copy

			punpcklbw mm0, mm6		;// unpack lower values; inptr[0][1][2][3]
			movq mm3, mm0			;// make a copy
			pmullw mm0, mul3w		;// multiply by 3
			psllq mm1, 8			;// shift 1 byte for previous values; inptr[-1][0][1][2]
			movq mm5, mm7			;// copy original data
			pand mm5, mask2			;// mask out all but lower byte for "previous" state
			paddb mm1, mm5			;// add in byte to quadword
			psrlq mm2, 8			;// shift right for "next" state; inptr[1][2][3][4]
			punpcklbw mm1, mm6		;// unpack 
			punpcklbw mm2, mm6		;// unpack

			paddw mm1, mm0			;// add in result from multiply to "previous" data
			paddw mm1, bias1w		;// add in bias 
			paddw mm2, mm0			;// add in result from multiply to "next" data
			paddw mm2, bias2w		;// add in bias
			psrlw mm1, 2			;// convert from word to byte
			psrlw mm2, 2			;// convert from word to byte
			psllq mm2, 8			;// prepare for interleave
			paddb mm2, mm1			;// do interleave
			movq [edi], mm2			;// write out results

			;// process high 8 bytes of mm7
			movq mm0, mm7			;// copy input data
			movq mm1, mm7			;// copy input data
			movq mm2, mm7			;// copy input data
			movq mm3, mm7			;// copy input data

			punpckhbw mm0, mm6		;// unpack hi data
			pmullw mm0, mul3w		;// multiply by 3
			psllq mm1, 8			;// shift 1 byte for previous values; inptr[-1][0][1][2]
			psrlq mm2, 8			;// shift right for "next" state; inptr[1][2][3][4]
			movq mm7, [esi+8]		;// get next quadword from input buffer
			movq mm5, mm7			;// make copy
			psllq mm5, 56			;// shift left to isolate LSB
			paddb mm2, mm5			;// add in byte for "next" state

			punpckhbw mm1, mm6		;// unpack
			punpckhbw mm2, mm6		;// unpack
			paddw mm1, mm0			;// add in result from multiply to "previous" data
			paddw mm1, bias1w		;// add in bias 
			paddw mm2, mm0			;// add in result from multiply to "next" data
			paddw mm2, bias2w		;// add in bias
			psrlw mm1, 2			;// convert from word to byte
			psrlw mm2, 2			;// convert from word to byte
			psllq mm2, 8			;// prepare for interleave
			paddb mm2, mm1			;// do interleave
			movq [edi+8], mm2		;// write out results

			add edi, 16			;// increment output buffer pointer
			add esi, 8			;// increment input buffer pointer
			sub ecx, 8			;// increment column counter
			cmp ecx, 8			;// cmp with 8
			jle last_col			;// if less that goto last column

			;// Main Loop - process low 8 bytes of mm7
		col_loop:
			movq mm0, mm7			;// copy input data
			movq mm1, mm7			;// copy input data
			movq mm2, mm7			;// copy input data
			punpcklbw mm0, mm6		;// unpack lo data
			pmullw mm0, mul3w		;// multiply by 3; i[0][1][2][3]
			psllq mm1, 8			;// shift left to get previous byte
			movq mm5, mm3			;// retrieve copy of "previous" state
			psrlq mm5, 56			;// shift to get LSB
			paddb mm1, mm5			;// add in byte
			psrlq mm2, 8			;// shift rt for "next" state
			punpcklbw mm1, mm6		;// unpack
			punpcklbw mm2, mm6		;// unpack
			paddw mm1, mm0			;// add in result from multiply to "previous" data
			paddw mm1, bias1w		;// add in bias 
			paddw mm2, mm0			;// add in result from multiply to "next" data
			paddw mm2, bias2w		;// add in bias 
			psrlw mm1, 2			;// convert from word to byte
			psrlw mm2, 2			;// convert from word to byte
			psllq mm2, 8			;// prepare for interleave
			paddb mm2, mm1			;// do interleave

			movq [edi], mm2			;// write out results

			;// process high 8 bytes of mm7
			movq mm0, mm7			;// copy input data
			movq mm1, mm7			;// copy input data
			movq mm2, mm7			;// copy input data
			movq mm3, mm7			;// copy input data

			punpckhbw mm0, mm6		;// unpack hi data
			pmullw mm0, mul3w		;// multiply by 3; i[0][1][2][3]
			psllq mm1, 8			;// shift left to get previous byte		    
			psrlq mm2, 8			;// shift rt for "next" state
			movq mm7, [esi+8]		;// get next quadword from input buffer
			movq mm5, mm7			;// make copy
			psllq mm5, 56			;// shift left for LSB
			paddb mm2, mm5			;// add in byte

			punpckhbw mm1, mm6		;// unpack
			punpckhbw mm2, mm6		;// unpack
			paddw mm1, mm0			;// add in result from multiply to "previous" data
			paddw mm1, bias1w		;// add in bias 
			paddw mm2, mm0			;// add in result from multiply to "next" data
			paddw mm2, bias2w		;// add in bias
			psrlw mm1, 2			;// convert from word to byte
			psrlw mm2, 2			;// convert from word to byte

			psllq mm2, 8			;// prepare for interleave
			paddb mm2, mm1			;// do interleave

			movq [edi+8], mm2		;// write out results

			add edi, 16			;// increment output buffer pointer
			add esi, 8			;// increment input buffer pointer
			sub ecx, 8			;// increment column counter
			cmp ecx, 8			;// cmp with 8
			jg col_loop			;// if &gt; 8 goto main loop

		last_col:
			;// Special last column case - process low 8 bytes of mm7
			movq mm0, mm7			;// copy input data
			movq mm1, mm7			;// copy input data
			movq mm2, mm7			;// copy input data

			punpcklbw mm0, mm6		;// unpack lo data

			pmullw mm0, mul3w		;// multiply by 3; i[0][1][2][3]

			psllq mm1, 8			;// shift left to get previous byte
			
			movq mm5, mm3			;// retrieve copy of "previous" state
			psrlq mm5, 56			;// shift left for MSB
			paddb mm1, mm5			;// add in byte

			psrlq mm2, 8			;// shift rt for "next" state
			punpcklbw mm1, mm6		;// unpack
			punpcklbw mm2, mm6		;// unpack
			paddw mm1, mm0			;// add in result from multiply to "previous" data
			paddw mm1, bias1w		;// add in bias
			paddw mm2, mm0			;// add in result from multiply to "next" data
			paddw mm2, bias2w		;// add in bias
			psrlw mm1, 2			;// convert from word to byte
			psrlw mm2, 2			;// convert from word to byte
			psllq mm2, 8			;// prepare for interleave
			paddb mm2, mm1			;// do interleave
			movq [edi], mm2			;// write out results

			;// Special last column case - process hi 8 bytes of mm7
			movq mm0, mm7			;// copy input data
			movq mm1, mm7			;// copy input data
			movq mm2, mm7			;// copy input data
			punpckhbw mm0, mm6		;// unpack hi data

			pmullw mm0, mul3w		;// multiply by 3; i[0][1][2][3]
			psllq mm1, 8			;// shift left to get previous byte
			psrlq mm2, 8			;// shift rt for "next" state
			pand mm7, mask1			;// mask out all but MSB
			paddb mm2, mm7			;// add in byte
			punpckhbw mm1, mm6		;// unpack
			punpckhbw mm2, mm6		;// unpack
			paddw mm1, mm0			;// add in result from multiply to "previous" data
			paddw mm1, bias1w		;// add in bias
			paddw mm2, mm0			;// add in result from multiply to "next" data
			paddw mm2, bias2w		;// add in bias
			psrlw mm1, 2			;// convert from word to byte
			psrlw mm2, 2			;// convert from word to byte
			psllq mm2, 8			;// prepare for interleave
			paddb mm2, mm1			;// do interleave
			movq [edi+8], mm2		;// write out results
			emms
		}
#endif
#if defined(HAVE_MMX_ATT_MNEMONICS)   
  __asm__ (
    "movl %0, %%ecx          \n\t"  // horizontal line size
    "movl %1, %%esi          \n\t"  // input buffer pointer
    "movl %2, %%edi         \n\t"  // output buffer pointer

    "pxor %%mm6,%%mm6            \n\t"  // zero register
    "movq (%%esi),%%mm7          \n\t"  // input register

    // Special 1st column case - process low 8 bytes of mm7
    "movq %%mm7,%%mm0            \n\t"  // move 1st quadword into mm7
    "movq %%mm7,%%mm1            \n\t"  // make a copy
    "movq %%mm7,%%mm2            \n\t"  // make a copy

    "punpcklbw %%mm6,%%mm0       \n\t" // unpack lower values; inptr[0][1][2][3]
    "movq %%mm0,%%mm3            \n\t"  // make a copy
    "pmullw mul3w,%%mm0         \n\t" // multiply by 3
    "psllq $8,%%mm1              \n\t"  // shift 1 byte for previous values; inptr[-1][0][1][2]
    "movq %%mm7,%%mm5            \n\t"  // copy original data
    "pand mask2,%%mm5           \n\t"  // mask out all but lower byte for "previous" state
    "paddb %%mm5,%%mm1           \n\t"  // add in byte to quadword
    "psrlq $8,%%mm2              \n\t"  // shift right for "next" state; inptr[1][2][3][4]
    "punpcklbw %%mm6,%%mm1       \n\t" // unpack 
    "punpcklbw %%mm6,%%mm2       \n\t" // unpack

    "paddw %%mm0,%%mm1           \n\t"  // add in result from multiply to "previous" data
    "paddw bias1w,%%mm1         \n\t" // add in bias 
    "paddw %%mm0,%%mm2           \n\t"  // add in result from multiply to "next" data
    "paddw bias2w,%%mm2         \n\t" // add in bias
    "psrlw $2,%%mm1              \n\t"  // convert from word to byte
    "psrlw $2,%%mm2              \n\t"  // convert from word to byte
    "psllq $8,%%mm2              \n\t"  // prepare for interleave
    "paddb %%mm1,%%mm2           \n\t"  // do interleave
    "movq %%mm2,(%%edi)          \n\t"  // write out results

    // process high 8 bytes of mm7
    "movq %%mm7,%%mm0            \n\t"  // copy input data
    "movq %%mm7,%%mm1            \n\t"  // copy input data
    "movq %%mm7,%%mm2            \n\t"  // copy input data
    "movq %%mm7,%%mm3            \n\t"  // copy input data

    "punpckhbw %%mm6,%%mm0       \n\t" // unpack hi data
    "pmullw mul3w,%%mm0         \n\t" // multiply by 3
    "psllq $8,%%mm1              \n\t"  // shift 1 byte for previous values; inptr[-1][0][1][2]
    "psrlq $8,%%mm2              \n\t"  // shift right for "next" state; inptr[1][2][3][4]
    "movq 8(%%esi),%%mm7         \n\t" // get next quadword from input buffer
    "movq %%mm7,%%mm5            \n\t"  // make copy
    "psllq $56,%%mm5             \n\t"  // shift left to isolate LSB
    "paddb %%mm5,%%mm2           \n\t"  // add in byte for "next" state

    "punpckhbw %%mm6,%%mm1       \n\t" // unpack
    "punpckhbw %%mm6,%%mm2       \n\t" // unpack
    "paddw %%mm0,%%mm1           \n\t"  // add in result from multiply to "previous" data
    "paddw bias1w,%%mm1         \n\t" // add in bias 
    "paddw %%mm0,%%mm2           \n\t"  // add in result from multiply to "next" data
    "paddw bias2w,%%mm2         \n\t" // add in bias
    "psrlw $2,%%mm1              \n\t"  // convert from word to byte
    "psrlw $2,%%mm2              \n\t"  // convert from word to byte
    "psllq $8,%%mm2              \n\t"  // prepare for interleave
    "paddb %%mm1,%%mm2           \n\t"  // do interleave
    "movq %%mm2,8(%%edi)         \n\t" // write out results

    "addl $16,%%edi              \n\t" // increment output buffer pointer
    "addl $8,%%esi               \n\t" // increment input buffer pointer
    "subl $8,%%ecx               \n\t" // increment column counter
    "cmpl $8,%%ecx               \n\t" // cmp with 8
    "jle last_col                \n\t"  // if less that goto last column

    // Main Loop - process low 8 bytes of mm7
  "col_loop_a:                   \n\t"
    "movq %%mm7,%%mm0            \n\t"  // copy input data
    "movq %%mm7,%%mm1            \n\t"  // copy input data
    "movq %%mm7,%%mm2            \n\t"  // copy input data
    "punpcklbw %%mm6,%%mm0       \n\t" // unpack lo data
    "pmullw mul3w,%%mm0         \n\t" // multiply by 3; i[0][1][2][3]
    "psllq $8,%%mm1              \n\t"  // shift left to get previous byte
    "movq %%mm3,%%mm5            \n\t"  // retrieve copy of "previous" state
    "psrlq $56,%%mm5             \n\t"  // shift to get LSB
    "paddb %%mm5,%%mm1           \n\t"  // add in byte
    "psrlq $8,%%mm2              \n\t"  // shift rt for "next" state
    "punpcklbw %%mm6,%%mm1       \n\t" // unpack
    "punpcklbw %%mm6,%%mm2       \n\t" // unpack
    "paddw %%mm0,%%mm1           \n\t"  // add in result from multiply to "previous" data
    "paddw bias1w,%%mm1         \n\t" // add in bias 
    "paddw %%mm0,%%mm2           \n\t"  // add in result from multiply to "next" data
    "paddw bias2w,%%mm2         \n\t" // add in bias 
    "psrlw $2,%%mm1              \n\t"  // convert from word to byte
    "psrlw $2,%%mm2              \n\t"  // convert from word to byte
    "psllq $8,%%mm2              \n\t"  // prepare for interleave
    "paddb %%mm1,%%mm2           \n\t"  // do interleave

    "movq %%mm2,(%%edi)          \n\t"  // write out results

    // process high 8 bytes of mm7
    "movq %%mm7,%%mm0            \n\t"  // copy input data
    "movq %%mm7,%%mm1            \n\t"  // copy input data
    "movq %%mm7,%%mm2            \n\t"  // copy input data
    "movq %%mm7,%%mm3            \n\t"  // copy input data

    "punpckhbw %%mm6,%%mm0       \n\t" // unpack hi data
    "pmullw mul3w,%%mm0         \n\t" // multiply by 3; i[0][1][2][3]
    "psllq $8,%%mm1              \n\t"  // shift left to get previous byte                 
    "psrlq $8,%%mm2              \n\t"  // shift rt for "next" state
    "movq 8(%%esi),%%mm7         \n\t" // get next quadword from input buffer
    "movq %%mm7,%%mm5            \n\t"  // make copy
    "psllq $56,%%mm5             \n\t"  // shift left for LSB
    "paddb %%mm5,%%mm2           \n\t"  // add in byte

    "punpckhbw %%mm6,%%mm1       \n\t" // unpack
    "punpckhbw %%mm6,%%mm2       \n\t" // unpack
    "paddw %%mm0,%%mm1           \n\t"  // add in result from multiply to "previous" data
    "paddw bias1w,%%mm1         \n\t" // add in bias 
    "paddw %%mm0,%%mm2           \n\t"  // add in result from multiply to "next" data
    "paddw bias2w,%%mm2         \n\t" // add in bias
    "psrlw $2,%%mm1              \n\t"  // convert from word to byte
    "psrlw $2,%%mm2              \n\t"  // convert from word to byte

    "psllq $8,%%mm2              \n\t"  // prepare for interleave
    "paddb %%mm1,%%mm2           \n\t"  // do interleave

    "movq %%mm2,8(%%edi)         \n\t" // write out results

    "addl $16,%%edi              \n\t" // increment output buffer pointer
    "addl $8,%%esi               \n\t" // increment input buffer pointer
    "subl $8,%%ecx               \n\t" // increment column counter
    "cmpl $8,%%ecx               \n\t" // cmp with 8
    "jg col_loop_a                 \n\t" // if > 8 goto main loop

  "last_col:                   \n\t"
    // Special last column case - process low 8 bytes of mm7
    "movq %%mm7,%%mm0            \n\t"  // copy input data
    "movq %%mm7,%%mm1            \n\t"  // copy input data
    "movq %%mm7,%%mm2            \n\t"  // copy input data

    "punpcklbw %%mm6,%%mm0       \n\t" // unpack lo data

    "pmullw mul3w,%%mm0         \n\t" // multiply by 3; i[0][1][2][3]

    "psllq $8,%%mm1              \n\t"  // shift left to get previous byte

    "movq %%mm3,%%mm5            \n\t"  // retrieve copy of "previous" state
    "psrlq $56,%%mm5             \n\t"  // shift left for MSB
    "paddb %%mm5,%%mm1           \n\t"  // add in byte

    "psrlq $8,%%mm2              \n\t"  // shift rt for "next" state
    "punpcklbw %%mm6,%%mm1       \n\t" // unpack
    "punpcklbw %%mm6,%%mm2       \n\t" // unpack
    "paddw %%mm0,%%mm1           \n\t"  // add in result from multiply to "previous" data
    "paddw bias1w,%%mm1         \n\t" // add in bias
    "paddw %%mm0,%%mm2           \n\t"  // add in result from multiply to "next" data
    "paddw bias2w,%%mm2         \n\t" // add in bias
    "psrlw $2,%%mm1              \n\t"  // convert from word to byte
    "psrlw $2,%%mm2              \n\t"  // convert from word to byte
    "psllq $8,%%mm2              \n\t"  // prepare for interleave
    "paddb %%mm1,%%mm2           \n\t"  // do interleave
    "movq %%mm2,(%%edi)          \n\t"  // write out results

    // Special last column case - process hi 8 bytes of mm7
    "movq %%mm7,%%mm0            \n\t"  // copy input data
    "movq %%mm7,%%mm1            \n\t"  // copy input data
    "movq %%mm7,%%mm2            \n\t"  // copy input data
    "punpckhbw %%mm6,%%mm0       \n\t" // unpack hi data

    "pmullw mul3w,%%mm0         \n\t" // multiply by 3; i[0][1][2][3]
    "psllq $8,%%mm1              \n\t"  // shift left to get previous byte
    "psrlq $8,%%mm2              \n\t"  // shift rt for "next" state
    "pand mask1,%%mm7           \n\t"  // mask out all but MSB
    "paddb %%mm7,%%mm2           \n\t"  // add in byte
    "punpckhbw %%mm6,%%mm1       \n\t" // unpack
    "punpckhbw %%mm6,%%mm2       \n\t" // unpack
    "paddw %%mm0,%%mm1           \n\t"  // add in result from multiply to "previous" data
    "paddw bias1w,%%mm1         \n\t" // add in bias
    "paddw %%mm0,%%mm2           \n\t"  // add in result from multiply to "next" data
    "paddw bias2w,%%mm2         \n\t" // add in bias
    "psrlw $2,%%mm1              \n\t"  // convert from word to byte
    "psrlw $2,%%mm2              \n\t"  // convert from word to byte
    "psllq $8,%%mm2              \n\t"  // prepare for interleave
    "paddb %%mm1,%%mm2           \n\t"  // do interleave
    "movq %%mm2,8(%%edi)         \n\t" // write out results
    "emms                        \n\t"

	: // no output regs
	//      %0           %1       %2       %3            %4
	: "m"(hsize), "m"(inptr), "m"(outptr)

	: "eax", "ebx", "ecx", "edx", "esi", "edi", "memory", "cc", "st"
      );
#endif
  }
}

METHODDEF(void)
h2v1_fancy_upsample_orig (j_decompress_ptr cinfo, jpeg_component_info * compptr,
		     JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr)
#else
METHODDEF(void)
h2v1_fancy_upsample (j_decompress_ptr cinfo, jpeg_component_info * compptr,
		     JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr)
#endif
{
  JSAMPARRAY output_data = *output_data_ptr;
  register JSAMPROW inptr, outptr;
  register int invalue;
  register JDIMENSION colctr;
  int inrow;

  for (inrow = 0; inrow < cinfo->max_v_samp_factor; inrow++) {
    inptr = input_data[inrow];
    outptr = output_data[inrow];
    /* Special case for first column */
    invalue = GETJSAMPLE(*inptr++);
    *outptr++ = (JSAMPLE) invalue;
    *outptr++ = (JSAMPLE) ((invalue * 3 + GETJSAMPLE(*inptr) + 2) >> 2);

    for (colctr = compptr->downsampled_width - 2; colctr > 0; colctr--) {
      /* General case: 3/4 * nearer pixel + 1/4 * further pixel */
      invalue = GETJSAMPLE(*inptr++) * 3;
      *outptr++ = (JSAMPLE) ((invalue + GETJSAMPLE(inptr[-2]) + 1) >> 2);
      *outptr++ = (JSAMPLE) ((invalue + GETJSAMPLE(*inptr) + 2) >> 2);
    }

    /* Special case for last column */
    invalue = GETJSAMPLE(*inptr);
    *outptr++ = (JSAMPLE) ((invalue * 3 + GETJSAMPLE(inptr[-1]) + 1) >> 2);
    *outptr++ = (JSAMPLE) invalue;
  }
}


/*
 * Fancy processing for the common case of 2:1 horizontal and 2:1 vertical.
 * Again a triangle filter; see comments for h2v1 case, above.
 *
 * It is OK for us to reference the adjacent input rows because we demanded
 * context from the main buffer controller (see initialization code).
 */

#if defined(HAVE_MMX_INTEL_MNEMONICS) || defined(HAVE_MMX_ATT_MNEMONICS)   
/* Silly forward definitions */
METHODDEF(void)
h2v2_fancy_upsample_mmx (j_decompress_ptr cinfo, jpeg_component_info * compptr,
			 JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr);
METHODDEF(void)
h2v2_fancy_upsample_orig (j_decompress_ptr cinfo, jpeg_component_info * compptr,
			  JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr);

METHODDEF(void)
h2v2_fancy_upsample (j_decompress_ptr cinfo, jpeg_component_info * compptr,
		     JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr)
{
  if (MMXAvailable)
    h2v2_fancy_upsample_mmx(cinfo, compptr, input_data, output_data_ptr);
  else
    h2v2_fancy_upsample_orig(cinfo, compptr, input_data, output_data_ptr);
}

METHODDEF(void)
h2v2_fancy_upsample_mmx (j_decompress_ptr cinfo, jpeg_component_info * compptr,
		     JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr)

{
 JSAMPARRAY output_data = *output_data_ptr;

#if defined(HAVE_MMX_INTEL_MNEMONICS) /* I think that's needed for an Intel compiler */
 register JSAMPROW inptr0, inptr1, inptr2, outptr, outptr2, save_val;    /* pointers to unsigned char */
#else /* but you get warnings on a GNU compiler */
 JSAMPROW inptr0, inptr1, inptr2, outptr, outptr2, save_val;    /* pointers to unsigned char */
#endif
 int inrow, outrow, vsamp = cinfo->max_v_samp_factor, c = 0,
   dsamp = compptr->downsampled_width, out_offset = dsamp * 4, 
   dsamp_2xw = dsamp * 2;
 
 inrow = outrow = 0;
 
 while (outrow < cinfo->max_v_samp_factor) {
   /* inptr0 points to nearest input row, inptr1 points to next nearest */
		inptr0 = input_data[inrow];
		inptr1 = input_data[inrow-1];
		inptr2 = input_data[inrow+1];
		outptr = output_data[outrow++];
		save_val = outptr + out_offset;
		outptr2 = output_data[outrow++];

#if defined(HAVE_MMX_INTEL_MNEMONICS) 
		_asm {
		  /* This is what we are trying to accomplish here

	  		mm0	     mm2	    mm1	     mm3

		o1 = (9 * i0[0] + 3 * i1[0] + 3 * i0[-1] + i1[-1] + 8) >> 4
		o3 = (9 * i0[1] + 3 * i1[1] + 3 * i0[0]  + i1[0]  + 8) >> 4
		o5 = (9 * i0[2] + 3 * i1[2] + 3 * i0[1]  + i1[1]  + 8) >> 4
		o7 = (9 * i0[3] + 3 * i1[3] + 3 * i0[2]  + i1[2]  + 8) >> 4

	  		mm0	     mm2	    mm1	     mm3

		o2 = (9 * i0[0] + 3 * i1[0] + 3 * i0[1] + i1[1] + 7) >> 4
		o4 = (9 * i0[1] + 3 * i1[1] + 3 * i0[2] + i1[2] + 7) >> 4
		o6 = (9 * i0[2] + 3 * i1[2] + 3 * i0[3] + i1[3] + 7) >> 4
		o8 = (9 * i0[3] + 3 * i1[3] + 3 * i0[4] + i1[4] + 7) >> 4

		output_buf = [o1 o2 o3 o4 o5 o6 o7 o8]

		NOTE:  for special first and last column cases
			o1 = (12 * i0[0] + 4 * i1[0] + 3 * 0 + 0 + 8) >> 4
		*/

			;// Part 1 of the output - process lo data for o1 o3 o5 o7
			mov ecx, dsamp        ;// columns to process
			mov edx, inptr0		  ;// input row1
			mov esi, inptr1		  ;// input row2
			mov edi, outptr		  ;// output buffer
			mov eax, save_val

			movq mm0, [edx]       ;// get data from input row 0
			movq mm2, [esi]       ;// get data from input row 1
			movq mm4, mm0         ;// save to process hi half of input0
			movq mm5, mm2         ;// save to process hi half of input1

			punpcklbw mm0, noval  ;// process inptr0
			movq   mm1, mm0       ;// copy inptr0
			psllq  mm1, 16        ;// shift for first column special case i0[-1]
			pmullw mm0, mul9ws    ;// multiply by special case constant
			pmullw mm1, mul3w     ;// multiply input1 by 3
			punpcklbw mm2, noval  ;// process inptr1
			movq  mm3, mm2        ;// copy inptr0
			psllq mm3, 16         ;// shift for first column special case i1[-1]
			pmullw mm2, mul3ws    ;// multiply by special case constant
			paddw mm1, mm0        ;// Add up results for
			movq [eax], mm1
			movq mm6, mm1        ;// with the next results
			paddw mm3, mm2        ;// final o1 o3 o5 o7
			paddw mm6, mm3        ;// output to be interleaved
			paddw mm6, bias8w     ;// Add even bias
			psrlw mm6, 4          ;// convert from word to byte (truncate)

			;// Part 2 of the output - process lo data for o2 o4 o6 o8
			movq mm0, mm4         ;// get data from input row 0
			movq mm2, mm5         ;// get data from input row 1
			movq mm1, mm0         ;// copy inptr0 for unpack
			movq mm3, mm2         ;// copy inptr1 for unpack

			punpcklbw mm0, noval  ;// process inptr1
			psrlq  mm1, 8         ;// shift right for i0[1][2][3][4]
			punpcklbw mm1, noval  ;// process inptr1
			pmullw mm0, mul9w     ;// multiply by nearest point constant
			pmullw mm1, mul3w     ;// multiply by next nearest constant

			punpcklbw mm2, noval  ;// process inptr1
			psrlq  mm3, 8         ;// shift right for i1[1][2][3][4]
			punpcklbw mm3, noval  ;// process inptr1
			pmullw mm2, mul3w     ;// multiply by next nearest constant

			paddw mm0, mm1        ;// Add up results for final o2 o4 o6 o8
			movq [eax+8], mm0
			paddw mm0, mm3        ;// previous results for o1 o3 o5 o7
			paddw mm0, bias7w     ;// Add odd bias
			paddw mm0, mm2        ;// output to be interleaved with the

			psrlw mm0, 4          ;// convert back to byte (with truncate)
			psllq mm0, 8          ;// prepare to interleave output results
			paddw mm6, mm0        ;// interleave results
			movq [edi], mm6       ;// write to output buffer

			add edi, 8            ;// increment output pointer 
			add eax, 16
			sub ecx, 8
			cmp ecx, 0
			jle last_column

			;// End of special case.  Now for generic loop
		col_loop:

			;// Part 2 of the output
			movq mm0, mm4         ;// get data from input row 0
			movq mm2, mm5         ;// get data from input row 1
			movq mm1, mm0         ;// copy inptr0 for unpack
			movq mm3, mm2         ;// copy inptr1 for unpack
			movq input0, mm0
			movq input1, mm2

			punpckhbw mm0, noval  ;// process inptr1[0]
			psllq  mm1, 8         ;// shift for inptr0[-1]
			punpckhbw mm1, noval  ;// process inptr1[1]
			pmullw mm0, mul9w     ;// multiply by special case constant
			pmullw mm1, mul3w     ;// multiply inptr1 by 3
			punpckhbw mm2, noval  ;// process inptr1[0]
			psllq mm3, 8          ;// shift for inptr1[-1]
			punpckhbw mm3, noval  ;// process inptr1
			pmullw mm2, mul3w     ;// multiply by special case constant

			paddw mm1, mm0        ;// Add up results for
			movq [eax], mm1
			movq mm6, mm1        ;// with the next results
			paddw mm6, bias8w     ;// Add even bias
			paddw mm3, mm2        ;// final o1 o3 o5 o7
			paddw  mm6, mm3        ;// output to be interleaved
			psrlw mm6, 4          ;// convert from word to byte (truncate)

			;// process hi data for  o2 o4 o6 o8
			movq mm1, mm4         ;// get data from input row 0
			movq mm3, mm5         ;// copy inptr1 for unpack

			psrlq  mm1, 8         ;// shift right for i0[1][2][3][4]
			movq  mm4, [edx + 8]  ;// need to add in a byte from the next column
			                      ;// load next inptr0 to mm4 for future use

			movq mm7, mm4
			psllq mm7, 56         ;// shift for MSB
			paddb  mm1, mm7		  ;// add in MSB from next input0 column
			punpckhbw mm1, noval  ;// process inptr0
			pmullw mm1, mul3w     ;// multiply by next nearest constant

			psrlq  mm3, 8         ;// shift right for i1[1][2][3][4]
			movq  mm5, [esi + 8]  ;// need to add in a byte from the next column
			                      ;// load next inptr1 to mm5 for future use

			movq  mm7, mm5
			psllq mm7, 56         ;// shift for MSB
			paddb  mm3, mm7		  ;// add in MSB from next input1 column
			punpckhbw mm3, noval  ;// process inptr1

			paddw mm0, mm1        ;// Add odd bias
			movq [eax+8], mm0
			paddw mm3, bias7w      ;// Add up results for final o2 o4 o6 o8
			paddw mm0, mm3        ;// output to be interleaved with the
			paddw mm0, mm2        ;// previous results for o1 o3 o5 o7
			psrlw mm0, 4          ;// convert back to byte (with truncate)

			psllq mm0, 8          ;// prepare to interleave output results
			paddw mm6, mm0        ;// interleave results
			movq [edi], mm6       ;// write to output buffer

			add edi, 8

			;// Part 1 of the output - process lo data for o1 o3 o5 o7

			movq mm0, mm4         ;// get data from input row 0
			movq mm2, mm5         ;// get data from input row 1

			punpcklbw mm0, noval  ;// process inptr0

			movq   mm1, mm0       ;// copy inptr0
			psllq  mm1, 16        ;// shift for first column special case i0[-1]
			movq   mm7, input0
			psrlq mm7, 56
			paddw mm1, mm7
			pmullw mm0, mul9w     ;// multiply by special case constant
			pmullw mm1, mul3w     ;// multiply input1 by 3
			punpcklbw mm2, noval  ;// process intr1
			movq  mm3, mm2        ;// copy inptr0
			psllq mm3, 16         ;// shift for first column special case i1[-1]
			movq   mm7, input1
			psrlq mm7, 56
			paddw mm3, mm7

			pmullw mm2, mul3w     ;// multiply by special case constant

			paddw mm1, mm0        ;// Add up results for
			movq [eax+16], mm1
			movq mm6, mm1        ;// with the next results
			paddw mm6, bias8w     ;// Add even bias
			paddw mm3, mm2        ;// final o1 o3 o5 o7
			paddw  mm6, mm3        ;// output to be interleaved
			psrlw mm6, 4          ;// convert from word to byte (truncate)

			;// Process lo data for o2 o4 o6 o8
			movq mm1, mm4         ;// copy inptr0 for unpack
			movq mm3, mm5         ;// copy inptr1 for unpack

			psrlq  mm1, 8         ;// shift right for i0[1][2][3][4]
			punpcklbw mm1, noval  ;// process inptr1
			pmullw mm1, mul3w     ;// multiply by next nearest constant

			psrlq  mm3, 8         ;// shift right for i1[1][2][3][4]
			punpcklbw mm3, noval  ;// process inptr1

			paddw mm0, mm1        ;// Add up results for final o2 o4 o6 o8
			movq [eax+24], mm0
			paddw mm0, mm3        ;// previous results for o1 o3 o5 o7
			paddw mm0, bias7w     ;// Add odd bias
			paddw mm0, mm2        ;// output to be interleaved with the

			psrlw mm0, 4          ;// convert back to byte (with truncate)

			psllq mm0, 8          ;// prepare to interleave output results
			paddw mm6, mm0        ;// interleave results
			movq [edi], mm6       ;// write to output buffer

			add edi, 8            ;// increment output pointer
			add edx, 8            ;// increment input0 pointer
			add esi, 8            ;// increment input1 pointer
			add eax, 32

			sub ecx, 8
			cmp ecx, 0
			jg col_loop

		last_column:
			;// Special for last column - process hi data for o1 o3 o5 o7
			movq mm0, mm4         ;// get data from input row 0
			movq mm1, mm0         ;// copy inptr0 for unpack
			movq mm3, mm5         ;// copy inptr1 for unpack

			punpckhbw mm0, noval  ;// process inptr1[0]
			psllq  mm1, 8         ;// shift for inptr0[-1]
			punpckhbw mm1, noval  ;// process inptr1[1]
			pmullw mm0, mul9w     ;// multiply by special case constant
			pmullw mm1, mul3w     ;// multiply inptr1 by 3

;//			punpckhbw mm2, noval  ;// process inptr1[0]
			psllq mm3, 8          ;// shift for inptr1[-1]
			punpckhbw mm3, noval  ;// process inptr1
;//			pmullw mm2, mul3w     ;// multiply by special case constant

			paddw mm1, mm0        ;// Add up results for
			movq [eax], mm1
			movq mm6, mm1        ;// with the next results
			paddw mm6, bias8w     ;// Add even bias
			paddw mm3, mm2        ;// final o1 o3 o5 o7
			paddw  mm6, mm3        ;// output to be interleaved
			psrlw mm6, 4          ;// convert from word to byte (truncate)

			;// Part 4 of the output - process hi data for  o2 o4 o6 o8
;//			movq mm2, mm5         ;// get data from input row 1
			movq mm1, mm4         ;// copy inptr0 for unpack
			movq mm3, mm5         ;// copy inptr1 for unpack

			psrlq  mm1, 8         ;// shift right for i0[1][2][3][4]
			                      ;// load next inptr0 to mm4 for future use
			pand  mm4, mask1
			paddb  mm1, mm4		  ;// add in MSB from next input0 column
			punpckhbw mm1, noval  ;// process inptr0
			pmullw mm1, mul3w     ;// multiply by next nearest constan

			psrlq  mm3, 8         ;// shift right for i1[1][2][3][4]
			                      ;// load next inptr1 to mm5 for future use
			pand  mm5, mask1
			paddb  mm3, mm5		  ;// add in MSB from next input1 column
			punpckhbw mm3, noval  ;// process inptr1

			paddw mm0, mm1	      ;// Add odd bias
			movq [eax+8], mm0
			paddw mm3, bias7w     ;// Add up results for final o2 o4 o6 o8
			paddw mm0, mm3        ;// output to be interleaved with the
			paddw mm0, mm2        ;// previous results for o1 o3 o5 o7

			psrlw mm0, 4          ;// convert back to byte (with truncate)

			psllq mm0, 8          ;// prepare to interleave output results
			paddw mm6, mm0        ;// interleave results
			movq [edi], mm6       ;// write to output buffer

;//			add edi, 8            ;// increment output pointer
			add edx, 8            ;// increment input0 pointer
;//			add esi, 8            ;// increment input1 pointer

/*************  For v = 1 *****************/

			mov ecx, dsamp        ;// columns to process
			mov esi, inptr2		  ;// input row2
			mov edi, outptr2	  ;// output buffer
			mov edx, inptr0
			mov eax, save_val

			movq mm2, [esi]       ;// get data from input row 1
			movq mm5, mm2         ;// save to process hi half of input1

			punpcklbw mm2, noval  ;// process inptr1
			movq  mm3, mm2        ;// copy inptr0
			psllq mm3, 16         ;// shift for first column special case i1[-1]

			pmullw mm2, mul3ws    ;// multiply by special case constant

			movq mm6, [eax]        ;// Add up results for
			paddw mm3, mm2        ;// final o1 o3 o5 o7
			paddw mm6, mm3        ;// output to be interleaved
			paddw mm6, bias8w     ;// Add even bias
			psrlw mm6, 4          ;// convert from word to byte (truncate)

			;// Part 2 of the output - process lo data for o2 o4 o6 o8
			movq mm2, mm5         ;// get data from input row 1
			movq mm3, mm2         ;// copy inptr1 for unpack

			punpcklbw mm2, noval  ;// process inptr1
			psrlq  mm3, 8         ;// shift right for i1[1][2][3][4]
			punpcklbw mm3, noval  ;// process inptr1
			pmullw mm2, mul3w     ;// multiply by next nearest constant

			movq  mm0, [eax+8]        ;// Add up results for final o2 o4 o6 o8
			paddw mm0, mm3        ;// previous results for o1 o3 o5 o7
			paddw mm0, bias7w     ;// Add odd bias
			paddw mm0, mm2        ;// output to be interleaved with the

			psrlw mm0, 4          ;// convert back to byte (with truncate)

			psllq mm0, 8          ;// prepare to interleave output results
			paddw mm6, mm0        ;// interleave results
			movq [edi], mm6       ;// write to output buffer

			add edi, 8            ;// increment output pointer 
			add eax, 16
			sub ecx, 8
			cmp ecx, 0
			jle last_column2

			;// End of special case.  Now for generic loop
		col_loop2:

			;// Part 2 of the output
			movq mm2, mm5         ;// get data from input row 1
			movq mm3, mm2         ;// copy inptr1 for unpack
			movq mm1, mm2

			punpckhbw mm2, noval  ;// process inptr1[0]
			psllq mm3, 8          ;// shift for inptr1[-1]
			punpckhbw mm3, noval  ;// process inptr1
			pmullw mm2, mul3w     ;// multiply by special case constant

			movq mm6, [eax]        ;// with the next results
			paddw mm6, bias8w     ;// Add even bias
			paddw mm3, mm2        ;// final o1 o3 o5 o7
			paddw  mm6, mm3        ;// output to be interleaved
			psrlw mm6, 4          ;// convert from word to byte (truncate)

			;// process hi data for  o2 o4 o6 o8
			movq mm2, mm5         ;// get data from input row 1
			movq mm3, mm2         ;// copy inptr1 for unpack

			punpckhbw mm2, noval  ;// process inptr1
			psrlq  mm3, 8         ;// shift right for i1[1][2][3][4]
			movq  mm5, [esi + 8]  ;// need to add in a byte from the next column
			                      ;// load next inptr1 to mm5 for future use
			movq  mm7, mm5
			psllq mm7, 56         ;// shift for MSB
			paddb  mm3, mm7		  ;// add in MSB from next input1 column
			punpckhbw mm3, noval  ;// process inptr1
			pmullw mm2, mul3w     ;// multiply by next nearest constant

			movq mm0, [eax+8]        ;// Add odd bias
			paddw mm3, bias7w      ;// Add up results for final o2 o4 o6 o8
			paddw mm0, mm3        ;// output to be interleaved with the
			paddw mm0, mm2        ;// previous results for o1 o3 o5 o7
			psrlw mm0, 4          ;// convert back to byte (with truncate)

			psllq mm0, 8          ;// prepare to interleave output results
			paddw mm6, mm0        ;// interleave results
			movq [edi], mm6       ;// write to output buffer

			add edi, 8

			;// Part 1 of the output - process lo data for o1 o3 o5 o7
			movq mm2, mm5         ;// get data from input row 1

			punpcklbw mm2, noval  ;// process inptr1
			movq  mm3, mm2        ;// copy inptr0
			psllq mm3, 16         ;// shift for first column special case i1[-1]
			movq   mm7, mm1
			psrlq mm7, 56
			paddw mm3, mm7

			pmullw mm2, mul3w     ;// multiply by special case constant

			movq mm6, [eax+16]        ;// Add up results for
			paddw mm6, bias8w     ;// Add even bias
			paddw mm3, mm2        ;// final o1 o3 o5 o7
			paddw  mm6, mm3        ;// output to be interleaved
			psrlw mm6, 4          ;// convert from word to byte (truncate)

			;// Process lo data for o2 o4 o6 o8
			movq mm2, mm5         ;// get data from input row 1
			movq mm3, mm2         ;// copy inptr1 for unpack

			punpcklbw mm2, noval  ;// process inptr1
			psrlq  mm3, 8         ;// shift right for i1[1][2][3][4]
			punpcklbw mm3, noval  ;// process inptr1
			pmullw mm2, mul3w     ;// multiply by next nearest constant

			movq mm0, [eax+24]        ;// Add up results for final o2 o4 o6 o8
			paddw mm0, mm3        ;// previous results for o1 o3 o5 o7
			paddw mm0, bias7w     ;// Add odd bias
			paddw mm0, mm2        ;// output to be interleaved with the

			psrlw mm0, 4          ;// convert back to byte (with truncate)

			psllq mm0, 8          ;// prepare to interleave output results
			paddw mm6, mm0        ;// interleave results
			movq [edi], mm6       ;// write to output buffer

			add edi, 8            ;// increment output pointer
			add edx, 8            ;// increment input0 pointer
			add esi, 8            ;// increment input1 pointer
			add eax, 32

			movq mm4, [edx]
;//			movq mm5, [esi]

			sub ecx, 8
			cmp ecx, 0
			jg col_loop2

		last_column2:
			;// Special for last column - process hi data for o1 o3 o5 o7
			movq mm2, mm5         ;// get data from input row 1
			movq mm3, mm2         ;// copy inptr1 for unpack

			punpckhbw mm2, noval  ;// process inptr1[0]
			psllq mm3, 8          ;// shift for inptr1[-1]
			punpckhbw mm3, noval  ;// process inptr1
			pmullw mm2, mul3w     ;// multiply by special case constant

			movq mm6, [eax]        ;// with the next results
			paddw mm6, bias8w     ;// Add even bias
			paddw mm3, mm2        ;// final o1 o3 o5 o7
			paddw  mm6, mm3        ;// output to be interleaved
			psrlw mm6, 4          ;// convert from word to byte (truncate)

			;// Part 4 of the output - process hi data for  o2 o4 o6 o8
			movq mm2, mm5         ;// get data from input row 1
			movq mm3, mm2         ;// copy inptr1 for unpack

			punpckhbw mm2, noval  ;// process inptr1
			psrlq  mm3, 8         ;// shift right for i1[1][2][3][4]
			                      ;// load next inptr1 to mm5 for future use
			pand  mm5, mask1
			paddb  mm3, mm5		  ;// add in MSB from next input1 column
			punpckhbw mm3, noval  ;// process inptr1
			pmullw mm2, mul3w     ;// multiply by next nearest constant

			movq mm0, [eax+8]	      ;// Add odd bias
			paddw mm3, bias7w     ;// Add up results for final o2 o4 o6 o8
			paddw mm0, mm3        ;// output to be interleaved with the
			paddw mm0, mm2        ;// previous results for o1 o3 o5 o7
			psrlw mm0, 4          ;// convert back to byte (with truncate)
			psllq mm0, 8          ;// prepare to interleave output results
			paddw mm6, mm0        ;// interleave results
			movq [edi], mm6       ;// write to output buffer
			add edi, 8            ;// increment output pointer
			add edx, 8            ;// increment input0 pointer
			add esi, 8            ;// increment input1 pointer
		}
#endif
#if defined(HAVE_MMX_ATT_MNEMONICS)
		__asm__ (
			 /* This is what we are trying to accomplish here

			    mm0          mm2            mm1      mm3

			    o1 = (9 * i0[0] + 3 * i1[0] + 3 * i0[-1] + i1[-1] + 8) >> 4
			    o3 = (9 * i0[1] + 3 * i1[1] + 3 * i0[0]  + i1[0]  + 8) >> 4
			    o5 = (9 * i0[2] + 3 * i1[2] + 3 * i0[1]  + i1[1]  + 8) >> 4
			    o7 = (9 * i0[3] + 3 * i1[3] + 3 * i0[2]  + i1[2]  + 8) >> 4

			    mm0          mm2            mm1      mm3

			    o2 = (9 * i0[0] + 3 * i1[0] + 3 * i0[1] + i1[1] + 7) >> 4
			    o4 = (9 * i0[1] + 3 * i1[1] + 3 * i0[2] + i1[2] + 7) >> 4
			    o6 = (9 * i0[2] + 3 * i1[2] + 3 * i0[3] + i1[3] + 7) >> 4
			    o8 = (9 * i0[3] + 3 * i1[3] + 3 * i0[4] + i1[4] + 7) >> 4

			    output_buf = [o1 o2 o3 o4 o5 o6 o7 o8]

			    NOTE:  for special first and last column cases
			    o1 = (12 * i0[0] + 4 * i1[0] + 3 * 0 + 0 + 8) >> 4
			 */

                       // Part 1 of the output - process lo data for o1 o3 o5 o7
			 "movl %0, %%ecx          \n\t"// columns to process
			 "movl %1, %%edx         \n\t"// input row1
			 "movl %2, %%esi         \n\t"// input row2
			 "movl %3, %%edi         \n\t"// output buffer
			 "movl %4, %%eax       \n\t"

			 "movq (%%edx),%%mm0          \n\t"// get data from input row 0
			 "movq (%%esi),%%mm2          \n\t"// get data from input row 1
			 "movq %%mm0,%%mm4            \n\t"// save to process hi half of input0
			 "movq %%mm2,%%mm5            \n\t"// save to process hi half of input1

			 "punpcklbw noval,%%mm0      \n\t"// process inptr0
			 "movq   %%mm0,%%mm1          \n\t"// copy inptr0
			 "psllq  $16,%%mm1            \n\t"// shift for first column special case i0[-1]
			 "pmullw mul9ws,%%mm0        \n\t"// multiply by special case constant
			 "pmullw mul3w,%%mm1         \n\t"// multiply input1 by 3
			 "punpcklbw noval,%%mm2      \n\t"// process inptr1
			 "movq  %%mm2,%%mm3           \n\t"// copy inptr0
			 "psllq $16,%%mm3             \n\t"// shift for first column special case i1[-1]
			 "pmullw mul3ws,%%mm2        \n\t"// multiply by special case constant
			 "paddw %%mm0,%%mm1           \n\t"// Add up results for
			 "movq %%mm1,(%%eax)          \n\t"
			 "movq %%mm1,%%mm6            \n\t"// with the next results
			 "paddw %%mm2,%%mm3           \n\t"// final o1 o3 o5 o7
			 "paddw %%mm3,%%mm6           \n\t"// output to be interleaved
			 "paddw bias8w,%%mm6         \n\t"// Add even bias
			 "psrlw $4,%%mm6              \n\t"// convert from word to byte (truncate)

			 // Part 2 of the output - process lo data for o2 o4 o6 o8
			 "movq %%mm4,%%mm0            \n\t"// get data from input row 0
			 "movq %%mm5,%%mm2            \n\t"// get data from input row 1
			 "movq %%mm0,%%mm1            \n\t"// copy inptr0 for unpack
			 "movq %%mm2,%%mm3            \n\t"// copy inptr1 for unpack

			 "punpcklbw noval,%%mm0      \n\t"// process inptr1
			 "psrlq  $8,%%mm1             \n\t"// shift right for i0[1][2][3][4]
			 "punpcklbw noval,%%mm1      \n\t"// process inptr1
			 "pmullw mul9w,%%mm0         \n\t"// multiply by nearest point constant
			 "pmullw mul3w,%%mm1         \n\t"// multiply by next nearest constant

			 "punpcklbw noval,%%mm2      \n\t"// process inptr1
			 "psrlq  $8,%%mm3             \n\t"// shift right for i1[1][2][3][4]
			 "punpcklbw noval,%%mm3      \n\t"// process inptr1
			 "pmullw mul3w,%%mm2         \n\t"// multiply by next nearest constant

			 "paddw %%mm1,%%mm0           \n\t"// Add up results for final o2 o4 o6 o8
			 "movq %%mm0,8(%%eax)         \n\t"
			 "paddw %%mm3,%%mm0           \n\t"// previous results for o1 o3 o5 o7
			 "paddw bias7w,%%mm0         \n\t"// Add odd bias
			 "paddw %%mm2,%%mm0           \n\t"// output to be interleaved with the

			 "psrlw $4,%%mm0              \n\t"// convert back to byte (with truncate)
			 "psllq $8,%%mm0              \n\t"// prepare to interleave output results
			 "paddw %%mm0,%%mm6           \n\t"// interleave results
			 "movq %%mm6,(%%edi)          \n\t"// write to output buffer

			 "addl $8,%%edi               \n\t"// increment output pointer 
			 "addl $16,%%eax              \n\t"
			 "subl $8,%%ecx               \n\t"
			 "cmpl $0,%%ecx               \n\t"
			 "jle last_column             \n\t"

			 // End of special case.  Now for generic loop
			 "col_loop:                   \n\t"

			 // Part 2 of the output
			 "movq %%mm4,%%mm0            \n\t"// get data from input row 0
			 "movq %%mm5,%%mm2            \n\t"// get data from input row 1
			 "movq %%mm0,%%mm1            \n\t"// copy inptr0 for unpack
			 "movq %%mm2,%%mm3            \n\t"// copy inptr1 for unpack
			 "movq %%mm0, input0         \n\t"
			 "movq %%mm2, input1         \n\t"

			 "punpckhbw noval,%%mm0      \n\t"// process inptr1[0]
			 "psllq  $8,%%mm1             \n\t"// shift for inptr0[-1]
			 "punpckhbw noval,%%mm1      \n\t"// process inptr1[1]
			 "pmullw mul9w,%%mm0         \n\t"// multiply by special case constant
			 "pmullw mul3w,%%mm1         \n\t"// multiply inptr1 by 3
			 "punpckhbw noval,%%mm2      \n\t"// process inptr1[0]
			 "psllq $8,%%mm3              \n\t"// shift for inptr1[-1]
			 "punpckhbw noval,%%mm3      \n\t"// process inptr1
			 "pmullw mul3w,%%mm2         \n\t"// multiply by special case constant

			 "paddw %%mm0,%%mm1           \n\t"// Add up results for
			 "movq %%mm1,(%%eax)          \n\t"
			 "movq %%mm1,%%mm6            \n\t"// with the next results
			 "paddw bias8w,%%mm6         \n\t"// Add even bias
			 "paddw %%mm2,%%mm3           \n\t"// final o1 o3 o5 o7
			 "paddw  %%mm3,%%mm6          \n\t"// output to be interleaved
			 "psrlw $4,%%mm6              \n\t"// convert from word to byte (truncate)

			 // process hi data for  o2 o4 o6 o8
			 "movq %%mm4,%%mm1            \n\t"// get data from input row 0
			 "movq %%mm5,%%mm3            \n\t"// copy inptr1 for unpack

			 "psrlq  $8,%%mm1             \n\t"// shift right for i0[1][2][3][4]
			 "movq  8(%%edx),%%mm4        \n\t"// need to add in a byte from the next column
			 // load next inptr0 to mm4 for future use

			 "movq %%mm4,%%mm7            \n\t"
			 "psllq $56,%%mm7             \n\t"// shift for MSB
			 "paddb  %%mm7,%%mm1          \n\t"// add in MSB from next input0 column
			 "punpckhbw noval,%%mm1      \n\t"// process inptr0
			 "pmullw mul3w,%%mm1         \n\t"// multiply by next nearest constant

			 "psrlq  $8,%%mm3             \n\t"// shift right for i1[1][2][3][4]
			 "movq  8(%%esi),%%mm5        \n\t"// need to add in a byte from the next column
			 // load next inptr1 to mm5 for future use

			 "movq  %%mm5,%%mm7           \n\t"
			 "psllq $56,%%mm7             \n\t"// shift for MSB
			 "paddb  %%mm7,%%mm3          \n\t"// add in MSB from next input1 column
			 "punpckhbw noval,%%mm3      \n\t"// process inptr1

			 "paddw %%mm1,%%mm0           \n\t"// Add odd bias
			 "movq %%mm0,8(%%eax)         \n\t"
			 "paddw bias7w,%%mm3         \n\t"// Add up results for final o2 o4 o6 o8
			 "paddw %%mm3,%%mm0           \n\t"// output to be interleaved with the
			 "paddw %%mm2,%%mm0           \n\t"// previous results for o1 o3 o5 o7
			 "psrlw $4,%%mm0              \n\t"// convert back to byte (with truncate)

			 "psllq $8,%%mm0              \n\t"// prepare to interleave output results
			 "paddw %%mm0,%%mm6           \n\t"// interleave results
			 "movq %%mm6,(%%edi)          \n\t"// write to output buffer

			 "addl $8,%%edi               \n\t"

			 // Part 1 of the output - process lo data for o1 o3 o5 o7

			 "movq %%mm4,%%mm0            \n\t"// get data from input row 0
			 "movq %%mm5,%%mm2            \n\t"// get data from input row 1

			 "punpcklbw noval,%%mm0      \n\t"// process inptr0

			 "movq   %%mm0,%%mm1          \n\t"// copy inptr0
			 "psllq  $16,%%mm1            \n\t"// shift for first column special case i0[-1]
			 "movq   input0,%%mm7        \n\t"
			 "psrlq $56,%%mm7             \n\t"
			 "paddw %%mm7,%%mm1           \n\t"
			 "pmullw mul9w,%%mm0         \n\t"// multiply by special case constant
			 "pmullw mul3w,%%mm1         \n\t"// multiply input1 by 3
			 "punpcklbw noval,%%mm2      \n\t"// process intr1
			 "movq  %%mm2,%%mm3           \n\t"// copy inptr0
			 "psllq $16,%%mm3             \n\t"// shift for first column special case i1[-1]
			 "movq   input1,%%mm7        \n\t"
			 "psrlq $56,%%mm7             \n\t"
			 "paddw %%mm7,%%mm3           \n\t"

			 "pmullw mul3w,%%mm2         \n\t"// multiply by special case constant

			 "paddw %%mm0,%%mm1           \n\t"// Add up results for
			 "movq %%mm1,16(%%eax)        \n\t"
			 "movq %%mm1,%%mm6            \n\t"// with the next results
			 "paddw bias8w,%%mm6         \n\t"// Add even bias
			 "paddw %%mm2,%%mm3           \n\t"// final o1 o3 o5 o7
			 "paddw  %%mm3,%%mm6          \n\t"// output to be interleaved
			 "psrlw $4,%%mm6              \n\t"// convert from word to byte (truncate)

			 // Process lo data for o2 o4 o6 o8
			 "movq %%mm4,%%mm1            \n\t"// copy inptr0 for unpack
			 "movq %%mm5,%%mm3            \n\t"// copy inptr1 for unpack

			 "psrlq  $8,%%mm1             \n\t"// shift right for i0[1][2][3][4]
			 "punpcklbw noval,%%mm1      \n\t"// process inptr1
			 "pmullw mul3w,%%mm1         \n\t"// multiply by next nearest constant

			 "psrlq  $8,%%mm3             \n\t"// shift right for i1[1][2][3][4]
			 "punpcklbw noval,%%mm3      \n\t"// process inptr1

			 "paddw %%mm1,%%mm0           \n\t"// Add up results for final o2 o4 o6 o8
			 "movq %%mm0,24(%%eax)        \n\t"
			 "paddw %%mm3,%%mm0           \n\t"// previous results for o1 o3 o5 o7
			 "paddw bias7w,%%mm0         \n\t"// Add odd bias
			 "paddw %%mm2,%%mm0           \n\t"// output to be interleaved with the

			 "psrlw $4,%%mm0              \n\t"// convert back to byte (with truncate)

			 "psllq $8,%%mm0              \n\t"// prepare to interleave output results
			 "paddw %%mm0,%%mm6           \n\t"// interleave results
			 "movq %%mm6,(%%edi)          \n\t"// write to output buffer

			 "addl $8,%%edi               \n\t"// increment output pointer
			 "addl $8,%%edx               \n\t"// increment input0 pointer
			 "addl $8,%%esi               \n\t"// increment input1 pointer
			 "addl $32,%%eax              \n\t"

			 "subl $8,%%ecx               \n\t"
			 "cmpl $0,%%ecx               \n\t"
			 "jg col_loop                 \n\t"

			 "last_column:                \n\t"
			 // Special for last column - process hi data for o1 o3 o5 o7
			 "movq %%mm4,%%mm0            \n\t"// get data from input row 0
			 "movq %%mm0,%%mm1            \n\t"// copy inptr0 for unpack
			 "movq %%mm5,%%mm3            \n\t"// copy inptr1 for unpack

			 "punpckhbw noval,%%mm0      \n\t"// process inptr1[0]
			 "psllq  $8,%%mm1             \n\t"// shift for inptr0[-1]
			 "punpckhbw noval,%%mm1      \n\t"// process inptr1[1]
			 "pmullw mul9w,%%mm0         \n\t"// multiply by special case constant
			 "pmullw mul3w,%%mm1         \n\t"// multiply inptr1 by 3

			 //                     punpckhbw mm2, noval  ;// process inptr1[0]
			 "psllq $8,%%mm3              \n\t"// shift for inptr1[-1]
			 "punpckhbw noval,%%mm3      \n\t"// process inptr1
			 //                     pmullw mm2, mul3w     ;// multiply by special case constant

			 "paddw %%mm0,%%mm1           \n\t"// Add up results for
			 "movq %%mm1,(%%eax)          \n\t"
			 "movq %%mm1,%%mm6            \n\t"// with the next results
			 "paddw bias8w,%%mm6         \n\t"// Add even bias
			 "paddw %%mm2,%%mm3           \n\t"// final o1 o3 o5 o7
			 "paddw  %%mm3,%%mm6          \n\t"// output to be interleaved
			 "psrlw $4,%%mm6              \n\t"// convert from word to byte (truncate)

			 // Part 4 of the output - process hi data for  o2 o4 o6 o8
			 //                     movq mm2, mm5         ;// get data from input row 1
			 "movq %%mm4,%%mm1            \n\t"// copy inptr0 for unpack
			 "movq %%mm5,%%mm3            \n\t"// copy inptr1 for unpack

			 "psrlq  $8,%%mm1             \n\t"// shift right for i0[1][2][3][4]
			 // load next inptr0 to mm4 for future use
			 "pand  mask1,%%mm4          \n\t"
			 "paddb  %%mm4,%%mm1          \n\t"// add in MSB from next input0 column
			 "punpckhbw noval,%%mm1      \n\t"// process inptr0
			 "pmullw mul3w,%%mm1         \n\t"// multiply by next nearest constan

			 "psrlq  $8,%%mm3             \n\t"// shift right for i1[1][2][3][4]
			 // load next inptr1 to mm5 for future use
			 "pand  mask1,%%mm5          \n\t"
			 "paddb  %%mm5,%%mm3          \n\t"// add in MSB from next input1 column
			 "punpckhbw noval,%%mm3      \n\t"// process inptr1

			 "paddw %%mm1,%%mm0           \n\t"// Add odd bias
			 "movq %%mm0,8(%%eax)         \n\t"
			 "paddw bias7w,%%mm3         \n\t"// Add up results for final o2 o4 o6 o8
			 "paddw %%mm3,%%mm0           \n\t"// output to be interleaved with the
			 "paddw %%mm2,%%mm0           \n\t"// previous results for o1 o3 o5 o7

			 "psrlw $4,%%mm0              \n\t"// convert back to byte (with truncate)

			 "psllq $8,%%mm0              \n\t"// prepare to interleave output results
			 "paddw %%mm0,%%mm6           \n\t"// interleave results
			 "movq %%mm6,(%%edi)          \n\t"// write to output buffer

			 //                     add edi, 8            ;// increment output pointer
			 "addl $8,%%edx               \n\t"// increment input0 pointer
			 //                     add esi, 8            ;// increment input1 pointer

	   /*************  For v = 1 *****************/
			 "movl %0, %%ecx          \n\t"// columns to process
			 "movl %5, %%esi         \n\t"// input row2
			 "movl %6, %%edi        \n\t"// output buffer
			 "movl %1, %%edx         \n\t"
			 "movl %4, %%eax       \n\t"

			 "movq (%%esi),%%mm2          \n\t"// get data from input row 1
			 "movq %%mm2,%%mm5            \n\t"// save to process hi half of input1

			 "punpcklbw noval,%%mm2      \n\t"// process inptr1
			 "movq  %%mm2,%%mm3           \n\t"// copy inptr0
			 "psllq $16,%%mm3             \n\t"// shift for first column special case i1[-1]

			 "pmullw mul3ws,%%mm2        \n\t"// multiply by special case constant

			 "movq (%%eax),%%mm6          \n\t"// Add up results for
			 "paddw %%mm2,%%mm3           \n\t"// final o1 o3 o5 o7
			 "paddw %%mm3,%%mm6           \n\t"// output to be interleaved
			 "paddw bias8w,%%mm6         \n\t"// Add even bias
			 "psrlw $4,%%mm6              \n\t"// convert from word to byte (truncate)

			 // Part 2 of the output - process lo data for o2 o4 o6 o8
			 "movq %%mm5,%%mm2            \n\t"// get data from input row 1
			 "movq %%mm2,%%mm3            \n\t"// copy inptr1 for unpack

			 "punpcklbw noval,%%mm2      \n\t"// process inptr1
			 "psrlq  $8,%%mm3             \n\t"// shift right for i1[1][2][3][4]
			 "punpcklbw noval,%%mm3      \n\t"// process inptr1
			 "pmullw mul3w,%%mm2         \n\t"// multiply by next nearest constant

			 "movq  8(%%eax),%%mm0        \n\t"// Add up results for final o2 o4 o6 o8
			 "paddw %%mm3,%%mm0           \n\t"// previous results for o1 o3 o5 o7
			 "paddw bias7w,%%mm0         \n\t"// Add odd bias
			 "paddw %%mm2,%%mm0           \n\t"// output to be interleaved with the

			 "psrlw $4,%%mm0              \n\t"// convert back to byte (with truncate)

			 "psllq $8,%%mm0              \n\t"// prepare to interleave output results
			 "paddw %%mm0,%%mm6           \n\t"// interleave results
			 "movq %%mm6,(%%edi)          \n\t"// write to output buffer

			 "addl $8,%%edi               \n\t"// increment output pointer 
			 "addl $16,%%eax              \n\t"
			 "subl $8,%%ecx               \n\t"
			 "cmpl $0,%%ecx               \n\t"
			 "jle last_column2            \n\t"

			 // End of special case.  Now for generic loop
			 "col_loop2:                  \n\t"

			 // Part 2 of the output
			 "movq %%mm5,%%mm2            \n\t"// get data from input row 1
			 "movq %%mm2,%%mm3            \n\t"// copy inptr1 for unpack
			 "movq %%mm2,%%mm1            \n\t"

			 "punpckhbw noval,%%mm2      \n\t"// process inptr1[0]
			 "psllq $8,%%mm3              \n\t"// shift for inptr1[-1]
			 "punpckhbw noval,%%mm3      \n\t"// process inptr1
			 "pmullw mul3w,%%mm2         \n\t"// multiply by special case constant

			 "movq (%%eax),%%mm6          \n\t"// with the next results
			 "paddw bias8w,%%mm6         \n\t"// Add even bias
			 "paddw %%mm2,%%mm3           \n\t"// final o1 o3 o5 o7
			 "paddw  %%mm3,%%mm6          \n\t"// output to be interleaved
			 "psrlw $4,%%mm6              \n\t"// convert from word to byte (truncate)

			 // process hi data for  o2 o4 o6 o8
			 "movq %%mm5,%%mm2            \n\t"// get data from input row 1
			 "movq %%mm2,%%mm3            \n\t"// copy inptr1 for unpack

			 "punpckhbw noval,%%mm2      \n\t"// process inptr1
			 "psrlq  $8,%%mm3             \n\t"// shift right for i1[1][2][3][4]
			 "movq  8(%%esi),%%mm5        \n\t"// need to add in a byte from the next column
			 // load next inptr1 to mm5 for future use
			 "movq  %%mm5,%%mm7           \n\t"
			 "psllq $56,%%mm7             \n\t"// shift for MSB
			 "paddb  %%mm7,%%mm3          \n\t"// add in MSB from next input1 column
			 "punpckhbw noval,%%mm3      \n\t"// process inptr1
			 "pmullw mul3w,%%mm2         \n\t"// multiply by next nearest constant

			 "movq 8(%%eax),%%mm0         \n\t"// Add odd bias
			 "paddw bias7w,%%mm3         \n\t"// Add up results for final o2 o4 o6 o8
			 "paddw %%mm3,%%mm0           \n\t"// output to be interleaved with the
			 "paddw %%mm2,%%mm0           \n\t"// previous results for o1 o3 o5 o7
			 "psrlw $4,%%mm0              \n\t"// convert back to byte (with truncate)

			 "psllq $8,%%mm0              \n\t"// prepare to interleave output results
			 "paddw %%mm0,%%mm6           \n\t"// interleave results
			 "movq %%mm6,(%%edi)          \n\t"// write to output buffer

			 "addl $8,%%edi               \n\t"

			 // Part 1 of the output - process lo data for o1 o3 o5 o7
			 "movq %%mm5,%%mm2            \n\t"// get data from input row 1

			 "punpcklbw noval,%%mm2      \n\t"// process inptr1
			 "movq  %%mm2,%%mm3           \n\t"// copy inptr0
			 "psllq $16,%%mm3             \n\t"// shift for first column special case i1[-1]
			 "movq   %%mm1,%%mm7          \n\t"
			 "psrlq $56,%%mm7             \n\t"
			 "paddw %%mm7,%%mm3           \n\t"

			 "pmullw mul3w,%%mm2         \n\t"// multiply by special case constant

			 "movq 16(%%eax),%%mm6        \n\t"// Add up results for
			 "paddw bias8w,%%mm6         \n\t"// Add even bias
			 "paddw %%mm2,%%mm3           \n\t"// final o1 o3 o5 o7
			 "paddw  %%mm3,%%mm6          \n\t"// output to be interleaved
			 "psrlw $4,%%mm6              \n\t"// convert from word to byte (truncate)

			 // Process lo data for o2 o4 o6 o8
			 "movq %%mm5,%%mm2            \n\t"// get data from input row 1
			 "movq %%mm2,%%mm3            \n\t"// copy inptr1 for unpack

			 "punpcklbw noval,%%mm2      \n\t"// process inptr1
			 "psrlq  $8,%%mm3             \n\t"// shift right for i1[1][2][3][4]
			 "punpcklbw noval,%%mm3      \n\t"// process inptr1
			 "pmullw mul3w,%%mm2         \n\t"// multiply by next nearest constant

			 "movq 24(%%eax),%%mm0        \n\t"// Add up results for final o2 o4 o6 o8
			 "paddw %%mm3,%%mm0           \n\t"// previous results for o1 o3 o5 o7
			 "paddw bias7w,%%mm0         \n\t"// Add odd bias
			 "paddw %%mm2,%%mm0           \n\t"// output to be interleaved with the

			 "psrlw $4,%%mm0              \n\t"// convert back to byte (with truncate)

			 "psllq $8,%%mm0              \n\t"// prepare to interleave output results
			 "paddw %%mm0,%%mm6           \n\t"// interleave results
			 "movq %%mm6,(%%edi)          \n\t"// write to output buffer

			 "addl $8,%%edi               \n\t"// increment output pointer
			 "addl $8,%%edx               \n\t"// increment input0 pointer
			 "addl $8,%%esi               \n\t"// increment input1 pointer
			 "addl $32,%%eax              \n\t"

			 "movq (%%edx),%%mm4          \n\t"
			 //                     movq mm5, [esi]

			 "subl $8,%%ecx               \n\t"
			 "cmpl $0,%%ecx               \n\t"
			 "jg col_loop2                \n\t"

			 "last_column2:               \n\t"
			 // Special for last column - process hi data for o1 o3 o5 o7
			 "movq %%mm5,%%mm2            \n\t"// get data from input row 1
			 "movq %%mm2,%%mm3            \n\t"// copy inptr1 for unpack

			 "punpckhbw noval,%%mm2      \n\t"// process inptr1[0]
			 "psllq $8,%%mm3              \n\t"// shift for inptr1[-1]
			 "punpckhbw noval,%%mm3      \n\t"// process inptr1
			 "pmullw mul3w,%%mm2         \n\t"// multiply by special case constant

			 "movq (%%eax),%%mm6          \n\t"// with the next results
			 "paddw bias8w,%%mm6         \n\t"// Add even bias
			 "paddw %%mm2,%%mm3           \n\t"// final o1 o3 o5 o7
			 "paddw  %%mm3,%%mm6          \n\t"// output to be interleaved
			 "psrlw $4,%%mm6              \n\t"// convert from word to byte (truncate)

			 // Part 4 of the output - process hi data for  o2 o4 o6 o8
			 "movq %%mm5,%%mm2            \n\t"// get data from input row 1
			 "movq %%mm2,%%mm3            \n\t"// copy inptr1 for unpack

			 "punpckhbw noval,%%mm2      \n\t"// process inptr1
			 "psrlq  $8,%%mm3             \n\t"// shift right for i1[1][2][3][4]
			 // load next inptr1 to mm5 for future use
			 "pand  mask1,%%mm5          \n\t"
			 "paddb  %%mm5,%%mm3          \n\t"// add in MSB from next input1 column
			 "punpckhbw noval,%%mm3      \n\t"// process inptr1
			 "pmullw mul3w,%%mm2         \n\t"// multiply by next nearest constant

			 "movq 8(%%eax),%%mm0         \n\t"// Add odd bias
			 "paddw bias7w,%%mm3         \n\t"// Add up results for final o2 o4 o6 o8
			 "paddw %%mm3,%%mm0           \n\t"// output to be interleaved with the
			 "paddw %%mm2,%%mm0           \n\t"// previous results for o1 o3 o5 o7
			 "psrlw $4,%%mm0              \n\t"// convert back to byte (with truncate)
			 "psllq $8,%%mm0              \n\t"// prepare to interleave output results
			 "paddw %%mm0,%%mm6           \n\t"// interleave results
			 "movq %%mm6,(%%edi)          \n\t"// write to output buffer
			 "addl $8,%%edi               \n\t"// increment output pointer
			 "addl $8,%%edx               \n\t"// increment input0 pointer
			 "addl $8,%%esi               \n\t"// increment input1 pointer

			 : // no output regs
			 //      %0           %1       %2       %3              %4                 %5
			 : "m"(dsamp), "m"(inptr0), "m"(inptr1), "m"(outptr),  "m"(save_val), "m"(inptr2),
			 "m"(outptr2) /* %6 */

			 : "eax", "ebx", "ecx", "edx", "esi", "edi", "memory", "cc", "st"
			 );
#endif
    inrow++;

  }

#if defined(HAVE_MMX_INTEL_MNEMONICS)
  __asm emms
#endif
#if defined(HAVE_MMX_ATT_MNEMONICS)
    __asm__("emms");
#endif
}

METHODDEF(void)
h2v2_fancy_upsample_orig (j_decompress_ptr cinfo, jpeg_component_info * compptr,
			  JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr)
#else
METHODDEF(void)
h2v2_fancy_upsample (j_decompress_ptr cinfo, jpeg_component_info * compptr,
		     JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr)
#endif
{
  JSAMPARRAY output_data = *output_data_ptr;
  register JSAMPROW inptr0, inptr1, outptr;
#if BITS_IN_JSAMPLE == 8
  register int thiscolsum, lastcolsum, nextcolsum;
#else
  register INT32 thiscolsum, lastcolsum, nextcolsum;
#endif
  register JDIMENSION colctr;
  int inrow, outrow, v;

  inrow = outrow = 0;
  while (outrow < cinfo->max_v_samp_factor) {
    for (v = 0; v < 2; v++) {
      /* inptr0 points to nearest input row, inptr1 points to next nearest */
      inptr0 = input_data[inrow];
      if (v == 0)		/* next nearest is row above */
	inptr1 = input_data[inrow-1];
      else			/* next nearest is row below */
	inptr1 = input_data[inrow+1];
      outptr = output_data[outrow++];

      /* Special case for first column */
      thiscolsum = GETJSAMPLE(*inptr0++) * 3 + GETJSAMPLE(*inptr1++);
      nextcolsum = GETJSAMPLE(*inptr0++) * 3 + GETJSAMPLE(*inptr1++);
      *outptr++ = (JSAMPLE) ((thiscolsum * 4 + 8) >> 4);
      *outptr++ = (JSAMPLE) ((thiscolsum * 3 + nextcolsum + 7) >> 4);
      lastcolsum = thiscolsum; thiscolsum = nextcolsum;

      for (colctr = compptr->downsampled_width - 2; colctr > 0; colctr--) {
	/* General case: 3/4 * nearer pixel + 1/4 * further pixel in each */
	/* dimension, thus 9/16, 3/16, 3/16, 1/16 overall */
	nextcolsum = GETJSAMPLE(*inptr0++) * 3 + GETJSAMPLE(*inptr1++);
	*outptr++ = (JSAMPLE) ((thiscolsum * 3 + lastcolsum + 8) >> 4);
	*outptr++ = (JSAMPLE) ((thiscolsum * 3 + nextcolsum + 7) >> 4);
	lastcolsum = thiscolsum; thiscolsum = nextcolsum;
      }

      /* Special case for last column */
      *outptr++ = (JSAMPLE) ((thiscolsum * 3 + lastcolsum + 8) >> 4);
      *outptr++ = (JSAMPLE) ((thiscolsum * 4 + 7) >> 4);
    }
    inrow++;
  }
}




/*
 * Module initialization routine for upsampling.
 */

GLOBAL(void)
jinit_upsampler (j_decompress_ptr cinfo)
{
  my_upsample_ptr upsample;
  int ci;
  jpeg_component_info * compptr;
  boolean need_buffer, do_fancy;
  int h_in_group, v_in_group, h_out_group, v_out_group;

  upsample = (my_upsample_ptr)
    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
				SIZEOF(my_upsampler));
  cinfo->upsample = (struct jpeg_upsampler *) upsample;
  upsample->pub.start_pass = start_pass_upsample;
  upsample->pub.upsample = sep_upsample;
  upsample->pub.need_context_rows = FALSE; /* until we find out differently */

  if (cinfo->CCIR601_sampling)	/* this isn't supported */
    ERREXIT(cinfo, JERR_CCIR601_NOTIMPL);

  /* jdmainct.c doesn't support context rows when min_DCT_scaled_size = 1,
   * so don't ask for it.
   */
  do_fancy = cinfo->do_fancy_upsampling && cinfo->min_DCT_scaled_size > 1;

  /* Verify we can handle the sampling factors, select per-component methods,
   * and create storage as needed.
   */
  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
       ci++, compptr++) {
    /* Compute size of an "input group" after IDCT scaling.  This many samples
     * are to be converted to max_h_samp_factor * max_v_samp_factor pixels.
     */
    h_in_group = (compptr->h_samp_factor * compptr->DCT_scaled_size) /
		 cinfo->min_DCT_scaled_size;
    v_in_group = (compptr->v_samp_factor * compptr->DCT_scaled_size) /
		 cinfo->min_DCT_scaled_size;
    h_out_group = cinfo->max_h_samp_factor;
    v_out_group = cinfo->max_v_samp_factor;
    upsample->rowgroup_height[ci] = v_in_group; /* save for use later */
    need_buffer = TRUE;
    if (! compptr->component_needed) {
      /* Don't bother to upsample an uninteresting component. */
      upsample->methods[ci] = noop_upsample;
      need_buffer = FALSE;
    } else if (h_in_group == h_out_group && v_in_group == v_out_group) {
      /* Fullsize components can be processed without any work. */
      upsample->methods[ci] = fullsize_upsample;
      need_buffer = FALSE;
    } else if (h_in_group * 2 == h_out_group &&
	       v_in_group == v_out_group) {
      /* Special cases for 2h1v upsampling */
      if (do_fancy && compptr->downsampled_width > 2)
	upsample->methods[ci] = h2v1_fancy_upsample;
      else
	upsample->methods[ci] = h2v1_upsample;
    } else if (h_in_group * 2 == h_out_group &&
	       v_in_group * 2 == v_out_group) {
      /* Special cases for 2h2v upsampling */
      if (do_fancy && compptr->downsampled_width > 2) {
	upsample->methods[ci] = h2v2_fancy_upsample;
	upsample->pub.need_context_rows = TRUE;
      } else
	upsample->methods[ci] = h2v2_upsample;
    } else if ((h_out_group % h_in_group) == 0 &&
	       (v_out_group % v_in_group) == 0) {
      /* Generic integral-factors upsampling method */
      upsample->methods[ci] = int_upsample;
      upsample->h_expand[ci] = (UINT8) (h_out_group / h_in_group);
      upsample->v_expand[ci] = (UINT8) (v_out_group / v_in_group);
    } else
      ERREXIT(cinfo, JERR_FRACT_SAMPLE_NOTIMPL);
    if (need_buffer) {
#if defined(HAVE_MMX_INTEL_MNEMONICS) || defined(HAVE_MMX_ATT_MNEMONICS)
      int multiply_factor = (upsample->pub.need_context_rows == TRUE) ? 3 : 1;
      
      if (MMXAvailable)
	/* Increase memory allocation for h2v2_fancy_upsampling */ 
	/* for saving reusable data */
	upsample->color_buf[ci] = (*cinfo->mem->alloc_sarray)
	  ((j_common_ptr) cinfo, JPOOL_IMAGE,
	   (JDIMENSION) jround_up((long) cinfo->output_width,
				  (long) cinfo->max_h_samp_factor),
	   (JDIMENSION) cinfo->max_v_samp_factor*multiply_factor);  
      else
#endif
	upsample->color_buf[ci] = (*cinfo->mem->alloc_sarray)
	  ((j_common_ptr) cinfo, JPOOL_IMAGE,
	   (JDIMENSION) jround_up((long) cinfo->output_width,
				  (long) cinfo->max_h_samp_factor),
	   (JDIMENSION) cinfo->max_v_samp_factor);
    }
  }
}