summaryrefslogtreecommitdiff
path: root/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
blob: 6e4a772a89d19f0fade508e58708261e1f76344f (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
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
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
2189
2190
2191
2192
2193
2194
2195
2196
2197
2198
2199
2200
2201
2202
2203
2204
2205
2206
2207
2208
2209
2210
2211
2212
2213
2214
2215
2216
2217
2218
2219
2220
2221
2222
2223
2224
2225
2226
2227
2228
2229
2230
2231
2232
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
2244
2245
2246
2247
2248
2249
2250
2251
2252
2253
2254
2255
2256
2257
2258
2259
2260
2261
2262
2263
2264
2265
2266
2267
2268
2269
2270
2271
2272
2273
2274
2275
2276
2277
2278
2279
2280
2281
2282
2283
2284
2285
2286
2287
2288
2289
2290
2291
2292
2293
2294
2295
2296
2297
2298
2299
2300
2301
2302
2303
2304
2305
2306
2307
2308
2309
2310
2311
2312
2313
2314
2315
2316
2317
2318
2319
2320
2321
2322
2323
2324
2325
2326
2327
2328
2329
2330
2331
2332
2333
2334
2335
2336
2337
2338
2339
2340
2341
2342
2343
2344
2345
2346
2347
2348
2349
2350
2351
2352
2353
2354
2355
2356
2357
2358
2359
2360
2361
2362
2363
2364
2365
2366
2367
2368
2369
2370
2371
2372
2373
2374
2375
2376
2377
2378
2379
2380
2381
2382
2383
2384
2385
2386
2387
2388
2389
2390
2391
2392
2393
2394
2395
2396
2397
2398
2399
2400
2401
2402
2403
2404
2405
2406
2407
2408
2409
2410
2411
2412
2413
2414
2415
2416
2417
2418
2419
2420
2421
2422
2423
2424
2425
2426
2427
2428
2429
2430
2431
2432
2433
2434
2435
2436
2437
2438
2439
2440
2441
2442
2443
2444
2445
2446
2447
2448
2449
2450
2451
2452
2453
2454
2455
2456
2457
2458
2459
2460
2461
2462
2463
2464
2465
2466
2467
2468
2469
2470
2471
2472
2473
2474
2475
2476
2477
2478
2479
2480
2481
2482
2483
2484
2485
2486
2487
2488
2489
2490
2491
2492
2493
2494
2495
2496
2497
2498
2499
2500
2501
2502
2503
2504
2505
2506
2507
2508
2509
2510
2511
2512
2513
2514
2515
2516
2517
2518
2519
2520
2521
2522
2523
2524
2525
2526
2527
2528
2529
2530
2531
2532
2533
2534
2535
2536
2537
2538
2539
2540
2541
2542
2543
2544
2545
2546
2547
2548
2549
2550
2551
2552
2553
2554
2555
2556
2557
2558
2559
2560
2561
2562
2563
2564
2565
2566
2567
2568
2569
2570
2571
2572
2573
2574
2575
2576
2577
2578
2579
2580
2581
2582
2583
2584
2585
2586
2587
2588
2589
2590
2591
2592
2593
2594
2595
2596
2597
2598
2599
2600
2601
2602
2603
2604
2605
2606
2607
2608
2609
2610
2611
2612
2613
2614
2615
2616
2617
2618
2619
2620
2621
2622
2623
2624
2625
2626
2627
2628
2629
2630
2631
2632
2633
2634
2635
2636
2637
2638
2639
2640
2641
2642
2643
2644
2645
2646
2647
2648
2649
2650
2651
2652
2653
2654
2655
2656
2657
2658
2659
2660
2661
2662
2663
2664
2665
2666
2667
2668
2669
2670
2671
2672
2673
2674
2675
2676
2677
2678
2679
2680
2681
2682
2683
2684
2685
2686
2687
2688
2689
2690
2691
2692
2693
2694
2695
2696
2697
2698
2699
2700
2701
2702
2703
2704
2705
2706
2707
2708
2709
2710
2711
2712
2713
2714
2715
2716
2717
2718
2719
2720
2721
2722
2723
2724
2725
2726
2727
2728
2729
2730
2731
2732
2733
2734
2735
2736
2737
2738
2739
2740
2741
2742
2743
2744
2745
2746
2747
2748
2749
2750
2751
2752
2753
2754
2755
2756
2757
2758
2759
2760
2761
2762
2763
2764
2765
2766
2767
2768
2769
2770
2771
2772
2773
2774
2775
2776
2777
2778
2779
2780
2781
2782
2783
2784
2785
2786
2787
2788
2789
2790
2791
2792
2793
2794
2795
2796
2797
2798
2799
2800
2801
2802
2803
2804
2805
2806
2807
2808
2809
2810
2811
2812
2813
2814
2815
2816
2817
2818
2819
2820
2821
2822
2823
2824
2825
2826
2827
2828
2829
2830
2831
2832
2833
2834
2835
2836
2837
2838
2839
2840
2841
2842
2843
2844
2845
2846
2847
2848
2849
2850
2851
2852
2853
2854
2855
2856
2857
2858
2859
2860
2861
2862
2863
2864
2865
2866
2867
2868
2869
2870
2871
2872
2873
2874
2875
2876
2877
2878
2879
2880
2881
2882
2883
2884
2885
2886
2887
2888
2889
2890
2891
2892
2893
2894
2895
2896
2897
2898
2899
2900
2901
2902
2903
2904
2905
2906
2907
2908
2909
2910
2911
2912
2913
2914
2915
2916
2917
2918
2919
2920
2921
2922
2923
2924
2925
2926
2927
2928
2929
2930
2931
2932
2933
2934
2935
2936
2937
2938
2939
2940
2941
2942
2943
2944
2945
2946
2947
2948
2949
2950
2951
2952
2953
2954
2955
2956
2957
2958
2959
2960
2961
2962
2963
2964
2965
2966
2967
2968
2969
2970
2971
2972
2973
2974
2975
2976
2977
2978
2979
2980
2981
2982
2983
2984
2985
2986
2987
2988
2989
2990
2991
2992
2993
2994
2995
2996
2997
2998
2999
3000
3001
3002
3003
3004
3005
3006
3007
3008
3009
3010
3011
3012
3013
3014
3015
3016
3017
3018
3019
3020
3021
3022
3023
3024
3025
3026
3027
3028
3029
3030
3031
3032
3033
3034
3035
3036
3037
3038
3039
3040
3041
3042
3043
3044
3045
3046
3047
3048
3049
3050
3051
3052
3053
3054
3055
3056
3057
3058
3059
3060
3061
3062
3063
3064
3065
3066
3067
3068
3069
3070
3071
3072
3073
3074
3075
3076
3077
3078
3079
3080
3081
3082
3083
3084
3085
3086
3087
3088
3089
3090
3091
3092
3093
3094
3095
3096
3097
3098
3099
3100
3101
3102
3103
3104
3105
3106
3107
3108
3109
3110
3111
3112
3113
3114
3115
3116
3117
3118
3119
3120
3121
3122
3123
3124
3125
3126
3127
3128
3129
3130
3131
3132
3133
3134
3135
3136
3137
3138
3139
3140
3141
3142
3143
3144
3145
3146
3147
3148
3149
3150
3151
3152
3153
3154
3155
3156
3157
3158
3159
3160
3161
3162
3163
3164
3165
3166
3167
3168
3169
3170
3171
3172
3173
3174
3175
3176
3177
3178
3179
3180
3181
3182
3183
3184
3185
3186
3187
3188
3189
3190
3191
3192
3193
3194
3195
3196
3197
3198
3199
3200
3201
3202
3203
3204
3205
3206
3207
3208
3209
3210
3211
3212
3213
3214
3215
3216
3217
3218
3219
3220
3221
3222
3223
3224
3225
3226
3227
3228
3229
3230
3231
3232
3233
3234
3235
3236
3237
3238
3239
3240
3241
3242
3243
3244
3245
3246
3247
3248
3249
3250
3251
3252
3253
3254
3255
3256
3257
3258
3259
3260
3261
3262
3263
3264
3265
3266
3267
3268
3269
3270
3271
3272
3273
3274
3275
3276
3277
3278
3279
3280
3281
3282
3283
3284
3285
3286
3287
3288
3289
3290
3291
3292
3293
3294
3295
3296
3297
3298
3299
3300
3301
3302
3303
3304
3305
3306
3307
3308
3309
3310
3311
3312
3313
3314
3315
3316
3317
3318
3319
3320
3321
3322
3323
3324
3325
3326
3327
3328
3329
3330
3331
3332
3333
3334
3335
3336
3337
3338
3339
3340
3341
3342
3343
3344
3345
3346
3347
3348
3349
3350
3351
3352
3353
3354
3355
3356
3357
3358
3359
3360
3361
3362
3363
3364
3365
3366
3367
3368
3369
3370
3371
3372
3373
3374
3375
3376
3377
3378
3379
3380
3381
3382
3383
3384
3385
3386
3387
3388
3389
3390
3391
3392
3393
3394
3395
3396
3397
3398
3399
3400
3401
3402
3403
3404
3405
3406
3407
3408
3409
3410
3411
3412
3413
3414
3415
3416
3417
3418
3419
3420
3421
3422
3423
3424
3425
3426
3427
3428
3429
3430
3431
3432
3433
3434
3435
3436
3437
3438
3439
3440
3441
3442
3443
3444
3445
3446
3447
3448
3449
3450
3451
3452
3453
3454
3455
3456
3457
3458
3459
3460
3461
3462
3463
3464
3465
3466
3467
3468
3469
3470
3471
3472
3473
3474
3475
3476
3477
3478
3479
3480
3481
3482
3483
3484
3485
3486
3487
3488
3489
3490
3491
3492
3493
3494
3495
3496
3497
3498
3499
3500
3501
3502
3503
3504
3505
3506
3507
3508
3509
3510
3511
3512
3513
3514
3515
3516
3517
3518
3519
3520
3521
3522
3523
3524
3525
3526
3527
3528
3529
3530
3531
3532
3533
3534
3535
3536
3537
3538
3539
3540
3541
3542
3543
3544
3545
3546
3547
3548
3549
3550
3551
3552
3553
3554
3555
3556
3557
3558
3559
3560
3561
3562
3563
3564
3565
3566
3567
3568
3569
3570
3571
3572
3573
3574
3575
3576
3577
3578
3579
3580
3581
3582
3583
3584
3585
3586
3587
3588
3589
3590
3591
3592
3593
3594
3595
3596
3597
3598
3599
3600
3601
3602
3603
3604
3605
3606
3607
3608
3609
3610
3611
3612
3613
3614
3615
3616
3617
3618
3619
3620
3621
3622
3623
3624
3625
3626
3627
3628
3629
3630
3631
3632
3633
3634
3635
3636
3637
3638
3639
3640
3641
3642
3643
3644
3645
3646
3647
3648
3649
3650
3651
3652
3653
3654
3655
3656
3657
3658
3659
3660
3661
3662
3663
3664
3665
3666
3667
3668
3669
3670
3671
3672
3673
3674
3675
3676
3677
3678
3679
3680
3681
3682
3683
3684
3685
3686
3687
3688
3689
3690
3691
3692
3693
3694
3695
3696
3697
3698
3699
3700
3701
3702
3703
3704
3705
3706
3707
3708
3709
3710
3711
3712
3713
3714
3715
3716
3717
3718
3719
3720
3721
3722
3723
3724
3725
3726
3727
3728
3729
3730
3731
3732
3733
3734
3735
3736
3737
3738
3739
3740
3741
3742
3743
3744
3745
3746
3747
3748
3749
3750
3751
3752
3753
3754
3755
3756
3757
3758
3759
3760
3761
3762
3763
3764
3765
3766
3767
3768
3769
3770
3771
3772
3773
3774
3775
3776
3777
3778
3779
3780
3781
3782
3783
3784
3785
3786
3787
3788
3789
3790
3791
3792
3793
3794
3795
3796
3797
3798
3799
3800
3801
3802
3803
3804
3805
3806
3807
3808
3809
3810
3811
3812
3813
3814
3815
3816
3817
3818
3819
3820
3821
3822
3823
3824
3825
3826
3827
3828
3829
3830
3831
3832
3833
3834
3835
3836
3837
3838
3839
3840
3841
3842
3843
3844
3845
3846
3847
3848
3849
3850
3851
3852
3853
3854
3855
3856
3857
3858
3859
3860
3861
3862
3863
3864
3865
3866
3867
3868
3869
3870
3871
3872
3873
3874
3875
3876
3877
3878
3879
3880
3881
3882
3883
3884
3885
3886
3887
3888
3889
3890
3891
3892
3893
3894
3895
3896
3897
3898
3899
3900
3901
3902
3903
3904
3905
3906
3907
3908
3909
3910
3911
3912
3913
3914
3915
3916
3917
3918
3919
3920
3921
3922
3923
3924
3925
3926
3927
3928
3929
3930
3931
3932
3933
3934
3935
3936
3937
3938
3939
3940
3941
3942
3943
3944
3945
3946
3947
3948
3949
3950
3951
3952
3953
3954
3955
3956
3957
3958
3959
3960
3961
3962
3963
3964
3965
3966
3967
3968
3969
3970
3971
3972
3973
3974
3975
3976
3977
3978
3979
3980
3981
3982
3983
3984
3985
3986
3987
3988
3989
3990
3991
3992
3993
3994
3995
3996
3997
3998
3999
4000
4001
4002
4003
4004
4005
4006
4007
4008
4009
4010
4011
4012
4013
4014
4015
4016
4017
4018
4019
4020
4021
4022
4023
4024
4025
4026
4027
4028
4029
4030
4031
4032
4033
4034
4035
4036
4037
4038
4039
4040
4041
4042
4043
4044
4045
4046
4047
4048
4049
4050
4051
4052
4053
4054
4055
4056
4057
4058
4059
4060
4061
4062
4063
4064
4065
4066
4067
4068
4069
4070
4071
4072
4073
4074
4075
4076
4077
4078
4079
4080
4081
4082
4083
4084
4085
4086
4087
4088
4089
4090
4091
4092
4093
4094
4095
4096
4097
4098
4099
4100
4101
4102
4103
4104
4105
4106
4107
4108
4109
4110
4111
4112
4113
4114
4115
4116
4117
4118
4119
4120
4121
4122
4123
4124
4125
4126
4127
4128
4129
4130
4131
4132
4133
4134
4135
4136
4137
4138
4139
4140
4141
4142
4143
4144
4145
4146
4147
4148
4149
4150
4151
4152
4153
4154
4155
4156
4157
4158
4159
4160
4161
4162
4163
4164
4165
4166
4167
4168
4169
4170
4171
4172
4173
4174
4175
4176
4177
4178
4179
4180
4181
4182
4183
4184
4185
4186
4187
4188
4189
4190
4191
4192
4193
4194
4195
4196
4197
4198
4199
4200
4201
4202
4203
4204
4205
4206
4207
4208
4209
4210
4211
4212
4213
4214
4215
4216
4217
4218
4219
4220
4221
4222
4223
4224
4225
4226
4227
4228
4229
4230
4231
4232
4233
4234
4235
4236
4237
4238
4239
4240
4241
4242
4243
4244
4245
4246
4247
4248
4249
4250
4251
4252
4253
4254
4255
4256
4257
4258
4259
4260
4261
4262
4263
4264
4265
4266
4267
4268
4269
4270
4271
4272
4273
4274
4275
4276
4277
4278
4279
4280
4281
4282
4283
4284
4285
4286
4287
4288
4289
4290
4291
4292
4293
4294
4295
4296
4297
4298
4299
4300
4301
4302
4303
4304
4305
4306
4307
4308
4309
4310
4311
4312
4313
4314
4315
4316
4317
4318
4319
4320
4321
4322
4323
4324
4325
4326
4327
4328
4329
4330
4331
4332
4333
4334
4335
4336
4337
4338
4339
4340
4341
4342
4343
4344
4345
4346
4347
4348
4349
4350
4351
4352
4353
4354
4355
4356
4357
4358
4359
4360
4361
4362
4363
4364
4365
4366
4367
4368
4369
4370
4371
4372
4373
4374
4375
4376
4377
4378
4379
4380
4381
4382
4383
4384
4385
4386
4387
4388
4389
4390
4391
4392
4393
4394
4395
4396
4397
4398
4399
4400
4401
4402
4403
4404
4405
4406
4407
4408
4409
4410
4411
4412
4413
4414
4415
4416
4417
4418
4419
4420
4421
4422
4423
4424
4425
4426
4427
4428
4429
4430
4431
4432
4433
4434
4435
4436
4437
4438
4439
4440
4441
4442
4443
4444
4445
4446
4447
4448
4449
4450
4451
4452
4453
4454
4455
4456
4457
4458
4459
4460
4461
4462
4463
4464
4465
4466
4467
4468
4469
4470
4471
4472
4473
4474
4475
4476
4477
4478
4479
4480
4481
4482
4483
4484
4485
4486
4487
4488
4489
4490
4491
4492
4493
4494
4495
4496
4497
4498
4499
4500
4501
4502
4503
4504
4505
4506
4507
4508
4509
4510
4511
4512
4513
4514
4515
4516
4517
4518
4519
4520
4521
4522
4523
4524
4525
4526
4527
4528
4529
4530
4531
4532
4533
4534
4535
4536
4537
4538
4539
4540
4541
4542
4543
4544
4545
4546
4547
4548
4549
4550
4551
4552
4553
4554
4555
4556
4557
4558
4559
4560
4561
4562
4563
4564
4565
4566
4567
4568
4569
4570
4571
4572
4573
4574
4575
4576
4577
4578
4579
4580
4581
4582
4583
4584
4585
4586
4587
4588
4589
4590
4591
4592
4593
4594
4595
4596
4597
4598
4599
4600
4601
4602
4603
4604
4605
4606
4607
4608
4609
4610
4611
4612
4613
4614
4615
4616
4617
4618
4619
4620
4621
4622
4623
4624
4625
4626
4627
4628
4629
4630
4631
4632
4633
4634
4635
4636
4637
4638
4639
4640
4641
4642
4643
4644
4645
4646
4647
4648
4649
4650
4651
4652
4653
4654
4655
4656
4657
4658
4659
4660
4661
4662
4663
4664
4665
4666
4667
4668
4669
4670
4671
4672
4673
4674
4675
4676
4677
4678
4679
4680
4681
4682
4683
4684
4685
4686
4687
4688
4689
4690
4691
4692
4693
4694
4695
4696
4697
4698
4699
4700
4701
4702
4703
4704
4705
4706
4707
4708
4709
4710
4711
4712
4713
4714
4715
4716
4717
4718
4719
4720
4721
4722
4723
4724
4725
4726
4727
4728
4729
4730
4731
4732
4733
4734
4735
4736
4737
4738
4739
4740
4741
4742
4743
4744
4745
4746
4747
4748
4749
4750
4751
4752
4753
4754
4755
4756
4757
4758
4759
4760
4761
4762
4763
4764
4765
4766
4767
4768
4769
4770
4771
4772
4773
4774
4775
4776
4777
4778
4779
4780
4781
4782
4783
4784
4785
4786
4787
4788
4789
4790
4791
4792
4793
4794
4795
4796
4797
4798
4799
4800
4801
4802
4803
4804
4805
4806
4807
4808
4809
4810
4811
4812
4813
4814
4815
4816
4817
4818
4819
4820
4821
4822
4823
4824
4825
4826
4827
4828
4829
4830
4831
4832
4833
4834
4835
4836
4837
4838
4839
4840
4841
4842
4843
4844
4845
4846
4847
4848
4849
4850
4851
4852
4853
4854
4855
4856
4857
4858
4859
4860
4861
4862
4863
4864
4865
4866
4867
4868
4869
4870
4871
4872
4873
4874
4875
4876
4877
4878
4879
4880
4881
4882
4883
4884
4885
4886
4887
4888
4889
4890
4891
4892
4893
4894
4895
4896
4897
4898
4899
4900
4901
4902
4903
4904
4905
4906
4907
4908
4909
4910
4911
4912
4913
4914
4915
4916
4917
4918
4919
4920
4921
4922
4923
4924
4925
4926
4927
4928
4929
4930
4931
4932
4933
4934
4935
4936
4937
4938
4939
4940
4941
4942
4943
4944
4945
4946
4947
4948
4949
4950
4951
4952
4953
4954
4955
4956
4957
4958
4959
4960
4961
4962
4963
4964
4965
4966
4967
4968
4969
4970
4971
4972
4973
4974
4975
4976
4977
4978
4979
4980
4981
4982
4983
4984
4985
4986
4987
4988
4989
4990
4991
4992
4993
4994
4995
4996
4997
4998
4999
5000
5001
5002
5003
5004
5005
5006
5007
5008
5009
5010
5011
5012
5013
5014
5015
5016
5017
5018
5019
5020
5021
5022
5023
5024
5025
5026
5027
5028
5029
5030
5031
5032
5033
5034
5035
5036
5037
5038
5039
5040
5041
5042
5043
5044
5045
5046
5047
5048
5049
5050
5051
5052
5053
5054
5055
5056
5057
5058
5059
5060
5061
5062
5063
5064
5065
5066
5067
5068
5069
5070
5071
5072
5073
5074
5075
5076
5077
5078
5079
5080
5081
5082
5083
5084
5085
5086
5087
5088
5089
5090
5091
5092
5093
5094
5095
5096
5097
5098
5099
5100
5101
5102
5103
5104
5105
5106
5107
5108
5109
5110
5111
5112
5113
5114
5115
5116
5117
5118
5119
5120
5121
5122
5123
5124
5125
5126
5127
5128
5129
5130
5131
5132
5133
5134
5135
5136
5137
5138
5139
5140
5141
5142
5143
5144
5145
5146
5147
5148
5149
5150
5151
5152
5153
5154
5155
5156
5157
5158
5159
5160
5161
5162
5163
5164
5165
5166
5167
5168
5169
5170
5171
5172
5173
5174
5175
5176
5177
5178
5179
5180
5181
5182
5183
5184
5185
5186
5187
5188
5189
5190
5191
5192
5193
5194
5195
5196
5197
5198
5199
5200
5201
5202
5203
5204
5205
5206
5207
5208
5209
5210
5211
5212
5213
5214
5215
5216
5217
5218
5219
5220
5221
5222
5223
5224
5225
5226
5227
5228
5229
5230
5231
5232
5233
5234
5235
5236
5237
5238
5239
5240
5241
5242
5243
5244
5245
5246
5247
5248
5249
5250
5251
5252
5253
5254
5255
5256
5257
5258
5259
5260
5261
5262
5263
5264
5265
5266
5267
5268
5269
5270
5271
5272
5273
5274
5275
5276
5277
5278
5279
5280
5281
5282
5283
5284
5285
5286
5287
5288
5289
5290
5291
5292
5293
5294
5295
5296
5297
5298
5299
5300
5301
5302
5303
5304
5305
5306
5307
5308
5309
5310
5311
5312
5313
5314
5315
5316
5317
5318
5319
5320
5321
5322
5323
5324
5325
5326
5327
5328
5329
5330
5331
5332
5333
5334
5335
5336
5337
5338
5339
5340
5341
5342
5343
5344
5345
5346
5347
5348
5349
5350
5351
5352
5353
5354
5355
5356
5357
5358
5359
5360
5361
5362
5363
5364
5365
5366
5367
5368
5369
5370
5371
5372
5373
5374
5375
5376
5377
5378
5379
5380
5381
5382
5383
5384
5385
5386
5387
5388
5389
5390
5391
5392
5393
5394
5395
5396
5397
5398
5399
5400
5401
5402
5403
5404
5405
5406
5407
5408
5409
5410
5411
5412
5413
5414
5415
5416
5417
5418
5419
5420
5421
5422
5423
5424
5425
5426
5427
5428
5429
5430
5431
5432
5433
5434
5435
5436
5437
5438
5439
5440
5441
5442
5443
5444
5445
5446
5447
5448
5449
5450
5451
5452
5453
5454
5455
5456
5457
5458
5459
5460
5461
5462
5463
5464
5465
5466
5467
5468
5469
5470
5471
5472
5473
5474
5475
5476
5477
5478
5479
5480
5481
5482
5483
5484
5485
5486
5487
5488
5489
5490
5491
5492
5493
5494
5495
5496
5497
5498
5499
5500
5501
5502
5503
5504
5505
5506
5507
5508
5509
5510
5511
5512
5513
5514
5515
5516
5517
5518
5519
5520
5521
5522
5523
5524
5525
5526
5527
5528
5529
5530
5531
5532
5533
5534
5535
5536
5537
5538
5539
5540
5541
5542
5543
5544
5545
5546
5547
5548
5549
5550
5551
5552
5553
5554
5555
5556
5557
5558
5559
5560
5561
5562
5563
5564
5565
5566
5567
5568
5569
5570
5571
5572
5573
5574
5575
5576
5577
5578
5579
5580
5581
5582
5583
5584
5585
5586
5587
5588
5589
5590
5591
5592
5593
5594
5595
5596
5597
5598
5599
5600
5601
5602
5603
5604
5605
5606
5607
5608
5609
5610
5611
5612
5613
5614
5615
5616
5617
5618
5619
5620
5621
5622
5623
5624
5625
5626
5627
5628
5629
5630
5631
5632
5633
5634
5635
5636
5637
5638
5639
5640
5641
5642
5643
5644
5645
5646
5647
5648
5649
5650
5651
5652
5653
5654
5655
5656
5657
5658
5659
5660
5661
5662
5663
5664
5665
5666
5667
5668
5669
5670
5671
5672
5673
5674
5675
5676
5677
5678
5679
5680
5681
5682
5683
5684
5685
5686
5687
5688
5689
5690
5691
5692
5693
5694
5695
5696
5697
5698
5699
5700
5701
5702
5703
5704
5705
5706
5707
5708
5709
5710
5711
5712
5713
5714
5715
5716
5717
5718
5719
5720
5721
5722
5723
5724
5725
5726
5727
5728
5729
5730
5731
5732
5733
5734
5735
5736
5737
5738
5739
5740
5741
5742
5743
5744
5745
5746
5747
5748
5749
5750
5751
5752
5753
5754
5755
5756
5757
5758
5759
5760
5761
5762
5763
5764
5765
5766
5767
5768
5769
5770
5771
5772
5773
5774
5775
5776
5777
5778
5779
5780
5781
5782
5783
5784
5785
5786
5787
5788
5789
5790
5791
5792
5793
5794
5795
5796
5797
5798
5799
5800
5801
5802
5803
5804
5805
5806
5807
5808
5809
5810
5811
5812
5813
5814
5815
5816
5817
5818
5819
5820
5821
5822
5823
5824
5825
5826
5827
5828
5829
5830
5831
5832
5833
5834
5835
5836
5837
5838
5839
5840
5841
5842
5843
5844
5845
5846
5847
5848
5849
5850
5851
5852
5853
5854
5855
5856
5857
5858
5859
5860
5861
5862
5863
5864
5865
5866
5867
5868
5869
5870
5871
5872
5873
5874
5875
5876
5877
5878
5879
5880
5881
5882
5883
5884
5885
5886
5887
5888
5889
5890
5891
5892
5893
5894
5895
5896
5897
5898
5899
5900
5901
5902
5903
5904
5905
5906
5907
5908
5909
5910
5911
5912
5913
5914
5915
5916
5917
5918
5919
5920
5921
5922
5923
5924
5925
5926
5927
5928
5929
5930
5931
5932
5933
5934
5935
5936
5937
5938
5939
5940
5941
5942
5943
5944
5945
5946
5947
5948
5949
5950
5951
5952
5953
5954
5955
5956
5957
5958
5959
5960
5961
5962
5963
5964
5965
5966
5967
5968
5969
5970
5971
5972
5973
5974
5975
5976
5977
5978
5979
5980
5981
5982
5983
5984
5985
5986
5987
5988
5989
5990
5991
5992
5993
5994
5995
5996
5997
5998
5999
6000
6001
6002
6003
6004
6005
6006
6007
6008
6009
6010
6011
6012
6013
6014
6015
6016
6017
6018
6019
6020
6021
6022
6023
6024
6025
6026
6027
6028
6029
6030
6031
6032
6033
6034
6035
6036
6037
6038
6039
6040
6041
6042
6043
6044
6045
6046
6047
6048
6049
6050
6051
6052
6053
6054
6055
6056
6057
6058
6059
6060
6061
6062
6063
6064
6065
6066
6067
6068
6069
6070
6071
6072
6073
6074
6075
6076
6077
6078
6079
6080
6081
6082
6083
6084
6085
6086
6087
6088
6089
6090
6091
6092
6093
6094
6095
6096
6097
6098
6099
6100
6101
6102
6103
6104
6105
6106
6107
6108
6109
6110
6111
6112
6113
6114
6115
6116
6117
6118
6119
6120
6121
6122
6123
6124
6125
6126
6127
6128
6129
6130
6131
6132
6133
6134
6135
6136
6137
6138
6139
6140
6141
6142
6143
6144
6145
6146
6147
6148
6149
6150
6151
6152
6153
6154
6155
6156
6157
6158
6159
6160
6161
6162
6163
6164
6165
6166
6167
6168
6169
6170
6171
6172
6173
6174
6175
6176
6177
6178
6179
6180
6181
6182
6183
6184
6185
6186
6187
6188
6189
6190
6191
6192
6193
6194
6195
6196
6197
6198
6199
6200
6201
6202
6203
6204
6205
6206
6207
6208
6209
6210
6211
6212
6213
6214
6215
6216
6217
6218
6219
6220
6221
6222
6223
6224
6225
6226
6227
6228
6229
6230
6231
6232
6233
6234
6235
6236
6237
6238
6239
6240
6241
6242
6243
6244
6245
6246
6247
6248
6249
6250
6251
6252
6253
6254
6255
6256
6257
6258
6259
6260
6261
6262
6263
6264
6265
6266
6267
6268
6269
6270
6271
6272
6273
6274
6275
6276
6277
6278
6279
6280
6281
6282
6283
6284
6285
6286
6287
6288
6289
6290
6291
6292
6293
6294
6295
6296
6297
6298
6299
6300
6301
6302
6303
6304
6305
6306
6307
6308
6309
6310
6311
6312
6313
6314
6315
6316
6317
6318
6319
6320
6321
6322
6323
6324
6325
6326
6327
6328
6329
6330
6331
6332
6333
6334
6335
6336
6337
6338
6339
6340
6341
6342
6343
6344
6345
6346
6347
6348
6349
6350
6351
6352
6353
6354
6355
6356
6357
6358
6359
6360
6361
6362
6363
6364
6365
6366
6367
6368
6369
6370
6371
6372
6373
6374
6375
6376
6377
6378
6379
6380
6381
6382
6383
6384
6385
6386
6387
6388
6389
6390
6391
6392
6393
6394
6395
6396
6397
6398
6399
6400
6401
6402
6403
6404
6405
6406
6407
6408
6409
6410
6411
6412
6413
6414
6415
6416
6417
6418
6419
6420
6421
6422
6423
6424
6425
6426
6427
6428
6429
6430
6431
6432
6433
6434
6435
6436
6437
6438
6439
6440
6441
6442
6443
6444
6445
6446
6447
6448
6449
6450
6451
6452
6453
6454
6455
6456
6457
6458
6459
6460
6461
6462
6463
6464
6465
6466
6467
6468
6469
6470
6471
6472
6473
6474
6475
6476
6477
6478
6479
6480
6481
6482
6483
6484
6485
6486
6487
6488
6489
6490
6491
6492
6493
6494
6495
6496
6497
6498
6499
6500
6501
6502
6503
6504
6505
6506
6507
6508
6509
6510
6511
6512
6513
6514
6515
6516
6517
6518
6519
6520
6521
6522
6523
6524
6525
6526
6527
6528
6529
6530
6531
6532
6533
6534
6535
6536
6537
6538
6539
6540
6541
6542
6543
6544
6545
6546
6547
6548
6549
6550
6551
6552
6553
6554
6555
6556
6557
6558
6559
6560
6561
6562
6563
6564
6565
6566
6567
6568
6569
6570
6571
6572
6573
6574
6575
6576
6577
6578
6579
6580
6581
6582
6583
6584
6585
6586
6587
6588
6589
6590
6591
6592
6593
6594
6595
6596
6597
6598
6599
6600
6601
6602
6603
6604
6605
6606
6607
6608
6609
6610
6611
6612
6613
6614
6615
6616
6617
6618
6619
6620
6621
6622
6623
6624
6625
6626
6627
6628
6629
6630
6631
6632
6633
6634
6635
6636
6637
6638
6639
6640
6641
6642
6643
6644
6645
6646
6647
6648
6649
6650
6651
6652
6653
6654
6655
6656
6657
6658
6659
6660
6661
6662
6663
6664
6665
6666
6667
6668
6669
6670
6671
6672
6673
6674
6675
6676
6677
6678
6679
6680
6681
6682
6683
6684
6685
6686
6687
6688
6689
6690
6691
6692
6693
6694
6695
6696
6697
6698
6699
6700
6701
6702
6703
6704
6705
6706
6707
6708
6709
6710
6711
6712
6713
6714
6715
6716
6717
6718
6719
6720
6721
6722
6723
6724
6725
6726
6727
6728
6729
6730
6731
6732
6733
6734
6735
6736
6737
6738
6739
6740
6741
6742
6743
6744
6745
6746
6747
6748
6749
6750
6751
6752
6753
6754
6755
6756
6757
6758
6759
6760
6761
6762
6763
6764
6765
6766
6767
6768
6769
6770
6771
6772
6773
6774
6775
6776
6777
6778
6779
6780
6781
6782
6783
6784
6785
6786
6787
6788
6789
6790
6791
6792
6793
6794
6795
6796
6797
6798
6799
6800
6801
6802
6803
6804
6805
6806
6807
6808
6809
6810
6811
6812
6813
6814
6815
6816
6817
6818
6819
6820
6821
6822
6823
6824
6825
6826
6827
6828
6829
6830
6831
6832
6833
6834
6835
6836
6837
6838
6839
6840
6841
6842
6843
6844
6845
6846
6847
6848
6849
6850
6851
6852
6853
6854
6855
6856
6857
6858
6859
6860
6861
6862
6863
6864
6865
6866
6867
6868
6869
6870
6871
6872
6873
6874
6875
6876
6877
6878
6879
6880
6881
6882
6883
6884
6885
6886
6887
6888
6889
6890
6891
6892
6893
6894
6895
6896
6897
6898
6899
6900
6901
6902
6903
6904
6905
6906
6907
6908
6909
6910
6911
6912
6913
6914
6915
6916
6917
6918
6919
6920
6921
6922
6923
6924
6925
6926
6927
6928
6929
6930
6931
6932
6933
6934
6935
6936
6937
6938
6939
6940
6941
6942
6943
6944
6945
6946
6947
6948
6949
6950
6951
6952
6953
6954
6955
6956
6957
6958
6959
6960
6961
6962
6963
6964
6965
6966
6967
6968
6969
6970
6971
6972
6973
6974
6975
6976
6977
6978
6979
6980
6981
6982
6983
6984
6985
6986
6987
6988
6989
6990
6991
6992
6993
6994
6995
6996
6997
6998
6999
7000
7001
7002
7003
7004
7005
7006
7007
7008
7009
7010
7011
7012
7013
7014
7015
7016
7017
7018
7019
7020
7021
7022
7023
7024
7025
7026
7027
7028
7029
7030
7031
7032
7033
7034
7035
7036
7037
7038
7039
7040
7041
7042
7043
7044
7045
7046
7047
7048
7049
7050
7051
7052
7053
7054
7055
7056
7057
7058
7059
7060
7061
7062
7063
7064
7065
7066
7067
7068
7069
7070
7071
7072
7073
7074
7075
7076
7077
7078
7079
7080
7081
7082
7083
7084
7085
7086
7087
7088
7089
7090
7091
7092
7093
7094
7095
7096
7097
7098
7099
7100
7101
7102
7103
7104
7105
7106
7107
7108
7109
7110
7111
7112
7113
7114
7115
7116
7117
7118
7119
7120
7121
7122
7123
7124
7125
7126
7127
7128
7129
7130
7131
7132
7133
7134
7135
7136
7137
7138
7139
7140
7141
7142
7143
7144
7145
7146
7147
7148
7149
7150
7151
7152
7153
7154
7155
7156
7157
7158
7159
7160
7161
7162
7163
7164
7165
7166
7167
7168
7169
7170
7171
7172
7173
7174
7175
7176
7177
7178
7179
7180
7181
7182
7183
7184
7185
7186
7187
7188
7189
7190
7191
7192
7193
7194
7195
7196
7197
7198
7199
7200
7201
7202
7203
7204
7205
7206
7207
7208
7209
7210
7211
7212
7213
7214
7215
7216
7217
7218
7219
7220
7221
7222
7223
7224
7225
7226
7227
7228
7229
7230
7231
7232
7233
7234
7235
7236
7237
7238
7239
7240
7241
7242
7243
7244
7245
7246
7247
7248
7249
7250
7251
7252
7253
7254
7255
7256
7257
7258
7259
7260
7261
7262
7263
7264
7265
7266
7267
7268
7269
7270
7271
7272
7273
7274
7275
7276
7277
7278
7279
7280
7281
7282
7283
7284
7285
7286
7287
7288
7289
7290
7291
7292
7293
7294
7295
7296
7297
7298
7299
7300
7301
7302
7303
7304
7305
7306
7307
7308
7309
7310
7311
7312
7313
7314
7315
7316
7317
7318
7319
7320
7321
7322
7323
7324
7325
7326
7327
7328
7329
7330
7331
7332
7333
7334
7335
7336
7337
7338
7339
7340
7341
7342
7343
7344
7345
7346
7347
7348
7349
7350
7351
7352
7353
7354
7355
7356
7357
7358
7359
7360
7361
7362
7363
7364
7365
7366
7367
7368
7369
7370
7371
7372
7373
7374
7375
7376
7377
7378
7379
7380
7381
7382
7383
7384
7385
7386
7387
7388
7389
7390
7391
7392
7393
7394
7395
7396
7397
7398
7399
7400
7401
7402
7403
7404
7405
7406
7407
7408
7409
7410
7411
7412
7413
7414
7415
7416
7417
7418
7419
7420
7421
7422
7423
7424
7425
7426
7427
7428
7429
7430
7431
7432
7433
7434
7435
7436
7437
7438
7439
7440
7441
7442
7443
7444
7445
7446
7447
7448
7449
7450
7451
7452
7453
7454
7455
7456
7457
7458
7459
7460
7461
7462
7463
7464
7465
7466
7467
7468
7469
7470
7471
7472
7473
7474
7475
7476
7477
7478
7479
7480
7481
7482
7483
7484
7485
7486
7487
7488
7489
7490
7491
7492
7493
7494
7495
7496
7497
7498
7499
7500
7501
7502
7503
7504
7505
7506
7507
7508
7509
7510
7511
7512
7513
7514
7515
7516
7517
7518
7519
7520
7521
7522
7523
7524
7525
7526
7527
7528
7529
7530
7531
7532
7533
7534
7535
7536
7537
7538
7539
7540
7541
7542
7543
7544
7545
7546
7547
7548
7549
7550
7551
7552
7553
7554
7555
7556
7557
7558
7559
7560
7561
7562
7563
7564
7565
7566
7567
7568
7569
7570
7571
7572
7573
7574
7575
7576
7577
7578
7579
7580
7581
7582
7583
7584
7585
7586
7587
7588
7589
7590
7591
7592
7593
7594
7595
7596
7597
7598
7599
7600
7601
7602
7603
7604
7605
7606
7607
7608
7609
7610
7611
7612
7613
7614
7615
7616
7617
7618
7619
7620
7621
7622
7623
7624
7625
7626
7627
7628
7629
7630
7631
7632
7633
7634
7635
7636
7637
7638
7639
7640
7641
7642
7643
7644
7645
7646
7647
7648
7649
7650
7651
7652
7653
7654
7655
7656
7657
7658
7659
7660
7661
7662
7663
7664
7665
7666
7667
7668
7669
7670
7671
7672
7673
7674
7675
7676
7677
7678
7679
7680
7681
7682
7683
7684
7685
7686
7687
7688
7689
7690
7691
7692
7693
7694
7695
7696
7697
7698
7699
7700
7701
7702
7703
7704
7705
7706
7707
7708
7709
7710
7711
7712
7713
7714
7715
7716
7717
7718
7719
7720
7721
7722
7723
7724
7725
7726
7727
7728
7729
7730
7731
7732
7733
7734
7735
7736
7737
7738
7739
7740
7741
7742
7743
7744
7745
7746
7747
7748
7749
7750
7751
7752
7753
7754
7755
7756
7757
7758
7759
7760
7761
7762
7763
7764
7765
7766
7767
7768
7769
7770
7771
7772
7773
7774
7775
7776
7777
7778
7779
7780
7781
7782
7783
7784
7785
7786
7787
7788
7789
7790
7791
7792
7793
7794
7795
7796
7797
7798
7799
7800
7801
7802
7803
7804
7805
7806
7807
7808
7809
7810
7811
7812
7813
7814
7815
7816
7817
7818
7819
7820
7821
7822
7823
7824
7825
7826
7827
7828
7829
7830
7831
7832
7833
7834
7835
7836
7837
7838
7839
7840
7841
7842
7843
7844
7845
7846
7847
7848
7849
7850
7851
7852
7853
7854
7855
7856
7857
7858
7859
7860
7861
7862
7863
7864
7865
7866
7867
7868
7869
7870
7871
7872
7873
7874
7875
7876
7877
7878
7879
7880
7881
7882
7883
7884
7885
7886
7887
7888
7889
7890
7891
7892
7893
7894
7895
7896
7897
7898
7899
7900
7901
7902
7903
7904
7905
7906
7907
7908
7909
7910
7911
7912
7913
7914
7915
7916
7917
7918
7919
7920
7921
7922
7923
7924
7925
7926
7927
7928
7929
7930
7931
7932
7933
7934
7935
7936
7937
7938
7939
7940
7941
7942
7943
7944
7945
7946
7947
7948
7949
7950
7951
7952
7953
7954
7955
7956
7957
7958
7959
7960
7961
7962
7963
7964
7965
7966
7967
7968
7969
7970
7971
7972
7973
7974
7975
7976
7977
7978
7979
7980
7981
7982
7983
7984
7985
7986
7987
7988
7989
7990
7991
7992
7993
7994
7995
7996
7997
7998
7999
8000
8001
8002
8003
8004
8005
8006
8007
8008
8009
8010
8011
8012
8013
8014
8015
8016
8017
8018
8019
8020
8021
8022
8023
8024
8025
8026
8027
8028
8029
8030
8031
8032
8033
8034
8035
8036
8037
8038
8039
8040
8041
8042
8043
8044
8045
8046
8047
8048
8049
8050
8051
8052
8053
8054
8055
8056
8057
8058
8059
8060
8061
8062
8063
8064
8065
8066
8067
8068
8069
8070
8071
8072
8073
8074
8075
8076
8077
8078
8079
8080
8081
8082
8083
8084
8085
8086
8087
8088
8089
8090
8091
8092
8093
8094
8095
8096
8097
8098
8099
8100
8101
8102
8103
8104
8105
8106
8107
8108
8109
8110
8111
8112
8113
8114
8115
8116
8117
8118
8119
8120
8121
8122
8123
8124
8125
8126
8127
8128
8129
8130
8131
8132
8133
8134
8135
8136
8137
8138
8139
8140
8141
8142
8143
8144
8145
8146
8147
8148
8149
8150
8151
8152
8153
8154
8155
8156
8157
8158
8159
8160
8161
8162
8163
8164
8165
8166
8167
8168
8169
8170
8171
8172
8173
8174
8175
8176
8177
8178
8179
8180
8181
8182
8183
8184
8185
8186
8187
8188
8189
8190
8191
8192
8193
8194
8195
8196
8197
8198
8199
8200
8201
8202
8203
8204
8205
8206
8207
8208
8209
8210
8211
8212
8213
8214
8215
8216
8217
8218
8219
8220
8221
8222
8223
8224
8225
8226
8227
8228
8229
8230
8231
8232
8233
8234
8235
8236
8237
8238
8239
8240
8241
8242
8243
8244
8245
8246
8247
8248
8249
8250
8251
8252
8253
8254
8255
8256
8257
8258
8259
8260
8261
8262
8263
8264
8265
8266
8267
8268
8269
8270
8271
8272
8273
8274
8275
8276
8277
8278
8279
8280
8281
8282
8283
8284
8285
8286
8287
8288
8289
8290
8291
8292
8293
8294
8295
8296
8297
8298
8299
8300
8301
8302
8303
8304
8305
8306
8307
8308
8309
8310
8311
8312
8313
8314
8315
8316
8317
8318
8319
8320
8321
8322
8323
8324
8325
8326
8327
8328
8329
8330
8331
8332
8333
8334
8335
8336
8337
8338
8339
8340
8341
8342
8343
8344
8345
8346
8347
8348
8349
8350
8351
8352
8353
8354
8355
8356
8357
8358
8359
8360
8361
8362
8363
8364
8365
8366
8367
8368
8369
8370
8371
8372
8373
8374
8375
8376
8377
8378
8379
8380
8381
8382
8383
8384
8385
8386
8387
8388
8389
8390
8391
8392
8393
8394
8395
8396
8397
8398
8399
8400
8401
8402
8403
8404
8405
8406
8407
8408
8409
8410
8411
8412
8413
8414
8415
8416
8417
8418
8419
8420
8421
8422
8423
8424
8425
8426
8427
8428
8429
8430
8431
8432
8433
8434
8435
8436
8437
8438
8439
8440
8441
8442
8443
8444
8445
8446
8447
8448
8449
8450
8451
8452
8453
8454
8455
8456
8457
8458
8459
8460
8461
8462
8463
8464
8465
8466
8467
8468
8469
8470
8471
8472
8473
8474
8475
8476
8477
8478
8479
8480
8481
8482
8483
8484
8485
8486
8487
8488
8489
8490
8491
8492
8493
8494
8495
8496
8497
8498
8499
8500
8501
8502
8503
8504
8505
8506
8507
8508
8509
8510
8511
8512
8513
8514
8515
8516
8517
8518
8519
8520
8521
8522
8523
8524
8525
8526
8527
8528
8529
8530
8531
8532
8533
8534
8535
8536
8537
8538
8539
8540
8541
8542
8543
8544
8545
8546
8547
8548
8549
8550
8551
8552
8553
8554
8555
8556
8557
8558
8559
8560
8561
8562
8563
8564
8565
8566
8567
8568
8569
8570
8571
8572
8573
8574
8575
8576
8577
8578
8579
8580
8581
8582
8583
8584
8585
8586
8587
8588
8589
8590
8591
8592
8593
8594
8595
8596
8597
8598
8599
8600
8601
8602
8603
8604
8605
8606
8607
8608
8609
8610
8611
8612
8613
8614
8615
8616
8617
8618
8619
8620
8621
8622
8623
8624
8625
8626
8627
8628
8629
8630
8631
8632
8633
8634
8635
8636
8637
8638
8639
8640
8641
8642
8643
8644
8645
8646
8647
8648
8649
8650
8651
8652
8653
8654
8655
8656
8657
8658
8659
8660
8661
8662
8663
8664
8665
8666
8667
8668
8669
8670
8671
8672
8673
8674
8675
8676
8677
8678
8679
8680
8681
8682
8683
8684
8685
8686
8687
8688
8689
8690
8691
8692
8693
8694
8695
8696
8697
8698
8699
8700
8701
8702
8703
8704
8705
8706
8707
8708
8709
8710
8711
8712
8713
8714
8715
8716
8717
8718
8719
8720
8721
8722
8723
8724
8725
8726
8727
8728
8729
8730
8731
8732
8733
8734
8735
8736
8737
8738
8739
8740
8741
8742
8743
8744
8745
8746
8747
8748
8749
8750
8751
8752
8753
8754
8755
8756
8757
8758
8759
8760
8761
8762
8763
8764
8765
8766
8767
8768
8769
8770
8771
8772
8773
8774
8775
8776
8777
8778
8779
8780
8781
8782
8783
8784
8785
8786
8787
8788
8789
8790
8791
8792
8793
8794
8795
8796
8797
8798
8799
8800
8801
8802
8803
8804
8805
8806
8807
8808
8809
8810
8811
8812
8813
8814
8815
8816
8817
8818
8819
8820
8821
8822
8823
8824
8825
8826
8827
8828
8829
8830
8831
8832
8833
8834
8835
8836
8837
8838
8839
8840
8841
8842
8843
8844
8845
8846
8847
8848
8849
8850
8851
8852
8853
8854
8855
8856
8857
8858
8859
8860
8861
8862
8863
8864
8865
8866
8867
8868
8869
8870
8871
8872
8873
8874
8875
8876
8877
8878
8879
8880
8881
8882
8883
8884
8885
8886
8887
8888
8889
8890
8891
8892
8893
8894
8895
8896
8897
8898
8899
8900
8901
8902
8903
8904
8905
8906
8907
8908
8909
8910
8911
8912
8913
8914
8915
8916
8917
8918
8919
8920
8921
8922
8923
8924
8925
8926
8927
8928
8929
8930
8931
8932
8933
8934
8935
8936
8937
8938
8939
8940
8941
8942
8943
8944
8945
8946
8947
8948
8949
8950
8951
8952
8953
8954
8955
8956
8957
8958
8959
8960
8961
8962
8963
8964
8965
8966
8967
8968
8969
8970
8971
8972
8973
8974
8975
8976
8977
8978
8979
8980
8981
8982
8983
8984
8985
8986
8987
8988
8989
8990
8991
8992
8993
8994
8995
8996
8997
8998
8999
9000
9001
9002
9003
9004
9005
9006
9007
9008
9009
9010
9011
9012
9013
9014
9015
9016
9017
9018
9019
9020
9021
9022
9023
9024
9025
9026
9027
9028
9029
9030
9031
9032
9033
9034
9035
9036
9037
9038
9039
9040
9041
9042
9043
9044
9045
9046
9047
9048
9049
9050
9051
9052
9053
9054
9055
9056
9057
9058
9059
9060
9061
9062
9063
9064
9065
9066
9067
9068
9069
9070
9071
9072
9073
9074
9075
9076
9077
9078
9079
9080
9081
9082
9083
9084
9085
9086
9087
9088
9089
9090
9091
9092
9093
9094
9095
9096
9097
9098
9099
9100
9101
9102
9103
9104
9105
9106
9107
9108
9109
9110
9111
9112
9113
9114
9115
9116
9117
9118
9119
9120
9121
9122
9123
9124
9125
9126
9127
9128
9129
9130
9131
9132
9133
9134
9135
9136
9137
9138
9139
9140
9141
9142
9143
9144
9145
9146
9147
9148
9149
9150
9151
9152
9153
9154
9155
9156
9157
9158
9159
9160
9161
9162
9163
9164
9165
9166
9167
9168
9169
9170
9171
9172
9173
9174
9175
9176
9177
9178
9179
9180
9181
9182
9183
9184
9185
9186
9187
9188
9189
9190
9191
9192
9193
9194
9195
9196
9197
9198
9199
9200
9201
9202
9203
9204
9205
9206
9207
9208
9209
9210
9211
9212
9213
9214
9215
9216
9217
9218
9219
9220
9221
9222
9223
9224
9225
9226
9227
9228
9229
9230
9231
9232
9233
9234
9235
9236
9237
9238
9239
9240
9241
9242
9243
9244
9245
9246
9247
9248
9249
9250
9251
9252
9253
9254
9255
9256
9257
9258
9259
9260
9261
9262
9263
9264
9265
9266
9267
9268
9269
9270
9271
9272
9273
9274
9275
9276
9277
9278
9279
9280
9281
9282
9283
9284
9285
9286
9287
9288
9289
9290
9291
9292
9293
9294
9295
9296
9297
9298
9299
9300
9301
9302
9303
9304
9305
9306
9307
9308
9309
9310
9311
9312
9313
9314
9315
9316
9317
9318
9319
9320
9321
9322
9323
9324
9325
9326
9327
9328
9329
9330
9331
9332
9333
9334
9335
9336
9337
9338
9339
9340
9341
9342
9343
9344
9345
9346
9347
9348
9349
9350
9351
9352
9353
9354
9355
9356
9357
9358
9359
9360
9361
9362
9363
9364
9365
9366
9367
9368
9369
9370
9371
9372
9373
9374
9375
9376
9377
9378
9379
9380
9381
9382
9383
9384
9385
9386
9387
9388
9389
9390
9391
9392
9393
9394
9395
9396
9397
9398
9399
9400
9401
9402
9403
9404
9405
9406
9407
9408
9409
9410
9411
9412
9413
9414
9415
9416
9417
9418
9419
9420
9421
9422
9423
9424
9425
9426
9427
9428
9429
9430
9431
9432
9433
9434
9435
9436
9437
9438
9439
9440
9441
9442
9443
9444
9445
9446
9447
9448
9449
9450
9451
9452
9453
9454
9455
9456
9457
9458
9459
9460
9461
9462
9463
9464
9465
9466
9467
9468
9469
9470
9471
9472
9473
9474
9475
9476
9477
9478
9479
9480
9481
9482
9483
9484
9485
9486
9487
9488
9489
9490
9491
9492
9493
9494
9495
9496
9497
9498
9499
9500
9501
9502
9503
9504
9505
9506
9507
9508
9509
9510
9511
9512
9513
9514
9515
9516
9517
9518
9519
9520
9521
9522
9523
9524
9525
9526
9527
9528
9529
9530
9531
9532
9533
9534
9535
9536
9537
9538
9539
9540
9541
9542
9543
9544
9545
9546
9547
9548
9549
9550
9551
9552
9553
9554
9555
9556
9557
9558
9559
9560
9561
9562
9563
9564
9565
9566
9567
9568
9569
9570
9571
9572
9573
9574
9575
9576
9577
9578
9579
9580
9581
9582
9583
9584
9585
9586
9587
9588
9589
9590
9591
9592
9593
9594
9595
9596
9597
9598
9599
9600
9601
9602
9603
9604
9605
9606
9607
9608
9609
9610
9611
9612
9613
9614
9615
9616
9617
9618
9619
9620
9621
9622
9623
9624
9625
9626
9627
9628
9629
9630
9631
9632
9633
9634
9635
9636
9637
9638
9639
9640
9641
9642
9643
9644
9645
9646
9647
9648
9649
9650
9651
9652
9653
9654
9655
9656
9657
9658
9659
9660
9661
9662
9663
9664
9665
9666
9667
9668
9669
9670
9671
9672
9673
9674
9675
9676
9677
9678
9679
9680
9681
9682
9683
9684
9685
9686
9687
9688
9689
9690
9691
9692
9693
9694
9695
9696
9697
9698
9699
9700
9701
9702
9703
//===-- DAGCombiner.cpp - Implement a DAG node combiner -------------------===//
//
//                     The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This pass combines dag nodes to form fewer, simpler DAG nodes.  It can be run
// both before and after the DAG is legalized.
//
// This pass is not a substitute for the LLVM IR instcombine pass. This pass is
// primarily intended to handle simplification opportunities that are implicit
// in the LLVM IR and exposed by the various codegen lowering phases.
//
//===----------------------------------------------------------------------===//

#define DEBUG_TYPE "dagcombine"
#include "llvm/CodeGen/SelectionDAG.h"
#include "llvm/DerivedTypes.h"
#include "llvm/LLVMContext.h"
#include "llvm/CodeGen/MachineFunction.h"
#include "llvm/CodeGen/MachineFrameInfo.h"
#include "llvm/Analysis/AliasAnalysis.h"
#include "llvm/DataLayout.h"
#include "llvm/Target/TargetLowering.h"
#include "llvm/Target/TargetMachine.h"
#include "llvm/Target/TargetOptions.h"
#include "llvm/ADT/SmallPtrSet.h"
#include "llvm/ADT/Statistic.h"
#include "llvm/Support/CommandLine.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/ErrorHandling.h"
#include "llvm/Support/MathExtras.h"
#include "llvm/Support/raw_ostream.h"
#include <algorithm>
using namespace llvm;

STATISTIC(NodesCombined   , "Number of dag nodes combined");
STATISTIC(PreIndexedNodes , "Number of pre-indexed nodes created");
STATISTIC(PostIndexedNodes, "Number of post-indexed nodes created");
STATISTIC(OpsNarrowed     , "Number of load/op/store narrowed");
STATISTIC(LdStFP2Int      , "Number of fp load/store pairs transformed to int");

namespace {
  static cl::opt<bool>
    CombinerAA("combiner-alias-analysis", cl::Hidden,
               cl::desc("Turn on alias analysis during testing"));

  static cl::opt<bool>
    CombinerGlobalAA("combiner-global-alias-analysis", cl::Hidden,
               cl::desc("Include global information in alias analysis"));

//------------------------------ DAGCombiner ---------------------------------//

  class DAGCombiner {
    SelectionDAG &DAG;
    const TargetLowering &TLI;
    CombineLevel Level;
    CodeGenOpt::Level OptLevel;
    bool LegalOperations;
    bool LegalTypes;

    // Worklist of all of the nodes that need to be simplified.
    //
    // This has the semantics that when adding to the worklist,
    // the item added must be next to be processed. It should
    // also only appear once. The naive approach to this takes
    // linear time.
    //
    // To reduce the insert/remove time to logarithmic, we use
    // a set and a vector to maintain our worklist.
    //
    // The set contains the items on the worklist, but does not
    // maintain the order they should be visited.
    //
    // The vector maintains the order nodes should be visited, but may
    // contain duplicate or removed nodes. When choosing a node to
    // visit, we pop off the order stack until we find an item that is
    // also in the contents set. All operations are O(log N).
    SmallPtrSet<SDNode*, 64> WorkListContents;
    SmallVector<SDNode*, 64> WorkListOrder;

    // AA - Used for DAG load/store alias analysis.
    AliasAnalysis &AA;

    /// AddUsersToWorkList - When an instruction is simplified, add all users of
    /// the instruction to the work lists because they might get more simplified
    /// now.
    ///
    void AddUsersToWorkList(SDNode *N) {
      for (SDNode::use_iterator UI = N->use_begin(), UE = N->use_end();
           UI != UE; ++UI)
        AddToWorkList(*UI);
    }

    /// visit - call the node-specific routine that knows how to fold each
    /// particular type of node.
    SDValue visit(SDNode *N);

  public:
    /// AddToWorkList - Add to the work list making sure its instance is at the
    /// back (next to be processed.)
    void AddToWorkList(SDNode *N) {
      WorkListContents.insert(N);
      WorkListOrder.push_back(N);
    }

    /// removeFromWorkList - remove all instances of N from the worklist.
    ///
    void removeFromWorkList(SDNode *N) {
      WorkListContents.erase(N);
    }

    SDValue CombineTo(SDNode *N, const SDValue *To, unsigned NumTo,
                      bool AddTo = true);

    SDValue CombineTo(SDNode *N, SDValue Res, bool AddTo = true) {
      return CombineTo(N, &Res, 1, AddTo);
    }

    SDValue CombineTo(SDNode *N, SDValue Res0, SDValue Res1,
                      bool AddTo = true) {
      SDValue To[] = { Res0, Res1 };
      return CombineTo(N, To, 2, AddTo);
    }

    void CommitTargetLoweringOpt(const TargetLowering::TargetLoweringOpt &TLO);

  private:

    /// SimplifyDemandedBits - Check the specified integer node value to see if
    /// it can be simplified or if things it uses can be simplified by bit
    /// propagation.  If so, return true.
    bool SimplifyDemandedBits(SDValue Op) {
      unsigned BitWidth = Op.getValueType().getScalarType().getSizeInBits();
      APInt Demanded = APInt::getAllOnesValue(BitWidth);
      return SimplifyDemandedBits(Op, Demanded);
    }

    bool SimplifyDemandedBits(SDValue Op, const APInt &Demanded);

    bool CombineToPreIndexedLoadStore(SDNode *N);
    bool CombineToPostIndexedLoadStore(SDNode *N);

    void ReplaceLoadWithPromotedLoad(SDNode *Load, SDNode *ExtLoad);
    SDValue PromoteOperand(SDValue Op, EVT PVT, bool &Replace);
    SDValue SExtPromoteOperand(SDValue Op, EVT PVT);
    SDValue ZExtPromoteOperand(SDValue Op, EVT PVT);
    SDValue PromoteIntBinOp(SDValue Op);
    SDValue PromoteIntShiftOp(SDValue Op);
    SDValue PromoteExtend(SDValue Op);
    bool PromoteLoad(SDValue Op);

    void ExtendSetCCUses(SmallVector<SDNode*, 4> SetCCs,
                         SDValue Trunc, SDValue ExtLoad, DebugLoc DL,
                         ISD::NodeType ExtType);

    /// combine - call the node-specific routine that knows how to fold each
    /// particular type of node. If that doesn't do anything, try the
    /// target-specific DAG combines.
    SDValue combine(SDNode *N);

    // Visitation implementation - Implement dag node combining for different
    // node types.  The semantics are as follows:
    // Return Value:
    //   SDValue.getNode() == 0 - No change was made
    //   SDValue.getNode() == N - N was replaced, is dead and has been handled.
    //   otherwise              - N should be replaced by the returned Operand.
    //
    SDValue visitTokenFactor(SDNode *N);
    SDValue visitMERGE_VALUES(SDNode *N);
    SDValue visitADD(SDNode *N);
    SDValue visitSUB(SDNode *N);
    SDValue visitADDC(SDNode *N);
    SDValue visitSUBC(SDNode *N);
    SDValue visitADDE(SDNode *N);
    SDValue visitSUBE(SDNode *N);
    SDValue visitMUL(SDNode *N);
    SDValue visitSDIV(SDNode *N);
    SDValue visitUDIV(SDNode *N);
    SDValue visitSREM(SDNode *N);
    SDValue visitUREM(SDNode *N);
    SDValue visitMULHU(SDNode *N);
    SDValue visitMULHS(SDNode *N);
    SDValue visitSMUL_LOHI(SDNode *N);
    SDValue visitUMUL_LOHI(SDNode *N);
    SDValue visitSMULO(SDNode *N);
    SDValue visitUMULO(SDNode *N);
    SDValue visitSDIVREM(SDNode *N);
    SDValue visitUDIVREM(SDNode *N);
    SDValue visitAND(SDNode *N);
    SDValue visitOR(SDNode *N);
    SDValue visitXOR(SDNode *N);
    SDValue SimplifyVBinOp(SDNode *N);
    SDValue SimplifyVUnaryOp(SDNode *N);
    SDValue visitSHL(SDNode *N);
    SDValue visitSRA(SDNode *N);
    SDValue visitSRL(SDNode *N);
    SDValue visitCTLZ(SDNode *N);
    SDValue visitCTLZ_ZERO_UNDEF(SDNode *N);
    SDValue visitCTTZ(SDNode *N);
    SDValue visitCTTZ_ZERO_UNDEF(SDNode *N);
    SDValue visitCTPOP(SDNode *N);
    SDValue visitSELECT(SDNode *N);
    SDValue visitSELECT_CC(SDNode *N);
    SDValue visitSETCC(SDNode *N);
    SDValue visitSIGN_EXTEND(SDNode *N);
    SDValue visitZERO_EXTEND(SDNode *N);
    SDValue visitANY_EXTEND(SDNode *N);
    SDValue visitSIGN_EXTEND_INREG(SDNode *N);
    SDValue visitTRUNCATE(SDNode *N);
    SDValue visitBITCAST(SDNode *N);
    SDValue visitBUILD_PAIR(SDNode *N);
    SDValue visitFADD(SDNode *N);
    SDValue visitFSUB(SDNode *N);
    SDValue visitFMUL(SDNode *N);
    SDValue visitFMA(SDNode *N);
    SDValue visitFDIV(SDNode *N);
    SDValue visitFREM(SDNode *N);
    SDValue visitFCOPYSIGN(SDNode *N);
    SDValue visitSINT_TO_FP(SDNode *N);
    SDValue visitUINT_TO_FP(SDNode *N);
    SDValue visitFP_TO_SINT(SDNode *N);
    SDValue visitFP_TO_UINT(SDNode *N);
    SDValue visitFP_ROUND(SDNode *N);
    SDValue visitFP_ROUND_INREG(SDNode *N);
    SDValue visitFP_EXTEND(SDNode *N);
    SDValue visitFNEG(SDNode *N);
    SDValue visitFABS(SDNode *N);
    SDValue visitFCEIL(SDNode *N);
    SDValue visitFTRUNC(SDNode *N);
    SDValue visitFFLOOR(SDNode *N);
    SDValue visitBRCOND(SDNode *N);
    SDValue visitBR_CC(SDNode *N);
    SDValue visitLOAD(SDNode *N);
    SDValue visitSTORE(SDNode *N);
    SDValue visitINSERT_VECTOR_ELT(SDNode *N);
    SDValue visitEXTRACT_VECTOR_ELT(SDNode *N);
    SDValue visitBUILD_VECTOR(SDNode *N);
    SDValue visitCONCAT_VECTORS(SDNode *N);
    SDValue visitEXTRACT_SUBVECTOR(SDNode *N);
    SDValue visitVECTOR_SHUFFLE(SDNode *N);
    SDValue visitMEMBARRIER(SDNode *N);

    SDValue XformToShuffleWithZero(SDNode *N);
    SDValue ReassociateOps(unsigned Opc, DebugLoc DL, SDValue LHS, SDValue RHS);

    SDValue visitShiftByConstant(SDNode *N, unsigned Amt);

    bool SimplifySelectOps(SDNode *SELECT, SDValue LHS, SDValue RHS);
    SDValue SimplifyBinOpWithSameOpcodeHands(SDNode *N);
    SDValue SimplifySelect(DebugLoc DL, SDValue N0, SDValue N1, SDValue N2);
    SDValue SimplifySelectCC(DebugLoc DL, SDValue N0, SDValue N1, SDValue N2,
                             SDValue N3, ISD::CondCode CC,
                             bool NotExtCompare = false);
    SDValue SimplifySetCC(EVT VT, SDValue N0, SDValue N1, ISD::CondCode Cond,
                          DebugLoc DL, bool foldBooleans = true);
    SDValue SimplifyNodeWithTwoResults(SDNode *N, unsigned LoOp,
                                         unsigned HiOp);
    SDValue CombineConsecutiveLoads(SDNode *N, EVT VT);
    SDValue ConstantFoldBITCASTofBUILD_VECTOR(SDNode *, EVT);
    SDValue BuildSDIV(SDNode *N);
    SDValue BuildUDIV(SDNode *N);
    SDValue MatchBSwapHWordLow(SDNode *N, SDValue N0, SDValue N1,
                               bool DemandHighBits = true);
    SDValue MatchBSwapHWord(SDNode *N, SDValue N0, SDValue N1);
    SDNode *MatchRotate(SDValue LHS, SDValue RHS, DebugLoc DL);
    SDValue ReduceLoadWidth(SDNode *N);
    SDValue ReduceLoadOpStoreWidth(SDNode *N);
    SDValue TransformFPLoadStorePair(SDNode *N);

    SDValue GetDemandedBits(SDValue V, const APInt &Mask);

    /// GatherAllAliases - Walk up chain skipping non-aliasing memory nodes,
    /// looking for aliasing nodes and adding them to the Aliases vector.
    void GatherAllAliases(SDNode *N, SDValue OriginalChain,
                          SmallVector<SDValue, 8> &Aliases);

    /// isAlias - Return true if there is any possibility that the two addresses
    /// overlap.
    bool isAlias(SDValue Ptr1, int64_t Size1,
                 const Value *SrcValue1, int SrcValueOffset1,
                 unsigned SrcValueAlign1,
                 const MDNode *TBAAInfo1,
                 SDValue Ptr2, int64_t Size2,
                 const Value *SrcValue2, int SrcValueOffset2,
                 unsigned SrcValueAlign2,
                 const MDNode *TBAAInfo2) const;

    /// FindAliasInfo - Extracts the relevant alias information from the memory
    /// node.  Returns true if the operand was a load.
    bool FindAliasInfo(SDNode *N,
                       SDValue &Ptr, int64_t &Size,
                       const Value *&SrcValue, int &SrcValueOffset,
                       unsigned &SrcValueAlignment,
                       const MDNode *&TBAAInfo) const;

    /// FindBetterChain - Walk up chain skipping non-aliasing memory nodes,
    /// looking for a better chain (aliasing node.)
    SDValue FindBetterChain(SDNode *N, SDValue Chain);

    /// Merge consecutive store operations into a wide store.
    /// This optimization uses wide integers or vectors when possible.
    /// \return True if some memory operations were changed.
    bool MergeConsecutiveStores(StoreSDNode *N);

  public:
    DAGCombiner(SelectionDAG &D, AliasAnalysis &A, CodeGenOpt::Level OL)
      : DAG(D), TLI(D.getTargetLoweringInfo()), Level(BeforeLegalizeTypes),
        OptLevel(OL), LegalOperations(false), LegalTypes(false), AA(A) {}

    /// Run - runs the dag combiner on all nodes in the work list
    void Run(CombineLevel AtLevel);

    SelectionDAG &getDAG() const { return DAG; }

    /// getShiftAmountTy - Returns a type large enough to hold any valid
    /// shift amount - before type legalization these can be huge.
    EVT getShiftAmountTy(EVT LHSTy) {
      return LegalTypes ? TLI.getShiftAmountTy(LHSTy) : TLI.getPointerTy();
    }

    /// isTypeLegal - This method returns true if we are running before type
    /// legalization or if the specified VT is legal.
    bool isTypeLegal(const EVT &VT) {
      if (!LegalTypes) return true;
      return TLI.isTypeLegal(VT);
    }
  };
}


namespace {
/// WorkListRemover - This class is a DAGUpdateListener that removes any deleted
/// nodes from the worklist.
class WorkListRemover : public SelectionDAG::DAGUpdateListener {
  DAGCombiner &DC;
public:
  explicit WorkListRemover(DAGCombiner &dc)
    : SelectionDAG::DAGUpdateListener(dc.getDAG()), DC(dc) {}

  virtual void NodeDeleted(SDNode *N, SDNode *E) {
    DC.removeFromWorkList(N);
  }
};
}

//===----------------------------------------------------------------------===//
//  TargetLowering::DAGCombinerInfo implementation
//===----------------------------------------------------------------------===//

void TargetLowering::DAGCombinerInfo::AddToWorklist(SDNode *N) {
  ((DAGCombiner*)DC)->AddToWorkList(N);
}

void TargetLowering::DAGCombinerInfo::RemoveFromWorklist(SDNode *N) {
  ((DAGCombiner*)DC)->removeFromWorkList(N);
}

SDValue TargetLowering::DAGCombinerInfo::
CombineTo(SDNode *N, const std::vector<SDValue> &To, bool AddTo) {
  return ((DAGCombiner*)DC)->CombineTo(N, &To[0], To.size(), AddTo);
}

SDValue TargetLowering::DAGCombinerInfo::
CombineTo(SDNode *N, SDValue Res, bool AddTo) {
  return ((DAGCombiner*)DC)->CombineTo(N, Res, AddTo);
}


SDValue TargetLowering::DAGCombinerInfo::
CombineTo(SDNode *N, SDValue Res0, SDValue Res1, bool AddTo) {
  return ((DAGCombiner*)DC)->CombineTo(N, Res0, Res1, AddTo);
}

void TargetLowering::DAGCombinerInfo::
CommitTargetLoweringOpt(const TargetLowering::TargetLoweringOpt &TLO) {
  return ((DAGCombiner*)DC)->CommitTargetLoweringOpt(TLO);
}

//===----------------------------------------------------------------------===//
// Helper Functions
//===----------------------------------------------------------------------===//

/// isNegatibleForFree - Return 1 if we can compute the negated form of the
/// specified expression for the same cost as the expression itself, or 2 if we
/// can compute the negated form more cheaply than the expression itself.
static char isNegatibleForFree(SDValue Op, bool LegalOperations,
                               const TargetLowering &TLI,
                               const TargetOptions *Options,
                               unsigned Depth = 0) {
  // No compile time optimizations on this type.
  if (Op.getValueType() == MVT::ppcf128)
    return 0;

  // fneg is removable even if it has multiple uses.
  if (Op.getOpcode() == ISD::FNEG) return 2;

  // Don't allow anything with multiple uses.
  if (!Op.hasOneUse()) return 0;

  // Don't recurse exponentially.
  if (Depth > 6) return 0;

  switch (Op.getOpcode()) {
  default: return false;
  case ISD::ConstantFP:
    // Don't invert constant FP values after legalize.  The negated constant
    // isn't necessarily legal.
    return LegalOperations ? 0 : 1;
  case ISD::FADD:
    // FIXME: determine better conditions for this xform.
    if (!Options->UnsafeFPMath) return 0;

    // After operation legalization, it might not be legal to create new FSUBs.
    if (LegalOperations &&
        !TLI.isOperationLegalOrCustom(ISD::FSUB,  Op.getValueType()))
      return 0;

    // fold (fneg (fadd A, B)) -> (fsub (fneg A), B)
    if (char V = isNegatibleForFree(Op.getOperand(0), LegalOperations, TLI,
                                    Options, Depth + 1))
      return V;
    // fold (fneg (fadd A, B)) -> (fsub (fneg B), A)
    return isNegatibleForFree(Op.getOperand(1), LegalOperations, TLI, Options,
                              Depth + 1);
  case ISD::FSUB:
    // We can't turn -(A-B) into B-A when we honor signed zeros.
    if (!Options->UnsafeFPMath) return 0;

    // fold (fneg (fsub A, B)) -> (fsub B, A)
    return 1;

  case ISD::FMUL:
  case ISD::FDIV:
    if (Options->HonorSignDependentRoundingFPMath()) return 0;

    // fold (fneg (fmul X, Y)) -> (fmul (fneg X), Y) or (fmul X, (fneg Y))
    if (char V = isNegatibleForFree(Op.getOperand(0), LegalOperations, TLI,
                                    Options, Depth + 1))
      return V;

    return isNegatibleForFree(Op.getOperand(1), LegalOperations, TLI, Options,
                              Depth + 1);

  case ISD::FP_EXTEND:
  case ISD::FP_ROUND:
  case ISD::FSIN:
    return isNegatibleForFree(Op.getOperand(0), LegalOperations, TLI, Options,
                              Depth + 1);
  }
}

/// GetNegatedExpression - If isNegatibleForFree returns true, this function
/// returns the newly negated expression.
static SDValue GetNegatedExpression(SDValue Op, SelectionDAG &DAG,
                                    bool LegalOperations, unsigned Depth = 0) {
  // fneg is removable even if it has multiple uses.
  if (Op.getOpcode() == ISD::FNEG) return Op.getOperand(0);

  // Don't allow anything with multiple uses.
  assert(Op.hasOneUse() && "Unknown reuse!");

  assert(Depth <= 6 && "GetNegatedExpression doesn't match isNegatibleForFree");
  switch (Op.getOpcode()) {
  default: llvm_unreachable("Unknown code");
  case ISD::ConstantFP: {
    APFloat V = cast<ConstantFPSDNode>(Op)->getValueAPF();
    V.changeSign();
    return DAG.getConstantFP(V, Op.getValueType());
  }
  case ISD::FADD:
    // FIXME: determine better conditions for this xform.
    assert(DAG.getTarget().Options.UnsafeFPMath);

    // fold (fneg (fadd A, B)) -> (fsub (fneg A), B)
    if (isNegatibleForFree(Op.getOperand(0), LegalOperations,
                           DAG.getTargetLoweringInfo(),
                           &DAG.getTarget().Options, Depth+1))
      return DAG.getNode(ISD::FSUB, Op.getDebugLoc(), Op.getValueType(),
                         GetNegatedExpression(Op.getOperand(0), DAG,
                                              LegalOperations, Depth+1),
                         Op.getOperand(1));
    // fold (fneg (fadd A, B)) -> (fsub (fneg B), A)
    return DAG.getNode(ISD::FSUB, Op.getDebugLoc(), Op.getValueType(),
                       GetNegatedExpression(Op.getOperand(1), DAG,
                                            LegalOperations, Depth+1),
                       Op.getOperand(0));
  case ISD::FSUB:
    // We can't turn -(A-B) into B-A when we honor signed zeros.
    assert(DAG.getTarget().Options.UnsafeFPMath);

    // fold (fneg (fsub 0, B)) -> B
    if (ConstantFPSDNode *N0CFP = dyn_cast<ConstantFPSDNode>(Op.getOperand(0)))
      if (N0CFP->getValueAPF().isZero())
        return Op.getOperand(1);

    // fold (fneg (fsub A, B)) -> (fsub B, A)
    return DAG.getNode(ISD::FSUB, Op.getDebugLoc(), Op.getValueType(),
                       Op.getOperand(1), Op.getOperand(0));

  case ISD::FMUL:
  case ISD::FDIV:
    assert(!DAG.getTarget().Options.HonorSignDependentRoundingFPMath());

    // fold (fneg (fmul X, Y)) -> (fmul (fneg X), Y)
    if (isNegatibleForFree(Op.getOperand(0), LegalOperations,
                           DAG.getTargetLoweringInfo(),
                           &DAG.getTarget().Options, Depth+1))
      return DAG.getNode(Op.getOpcode(), Op.getDebugLoc(), Op.getValueType(),
                         GetNegatedExpression(Op.getOperand(0), DAG,
                                              LegalOperations, Depth+1),
                         Op.getOperand(1));

    // fold (fneg (fmul X, Y)) -> (fmul X, (fneg Y))
    return DAG.getNode(Op.getOpcode(), Op.getDebugLoc(), Op.getValueType(),
                       Op.getOperand(0),
                       GetNegatedExpression(Op.getOperand(1), DAG,
                                            LegalOperations, Depth+1));

  case ISD::FP_EXTEND:
  case ISD::FSIN:
    return DAG.getNode(Op.getOpcode(), Op.getDebugLoc(), Op.getValueType(),
                       GetNegatedExpression(Op.getOperand(0), DAG,
                                            LegalOperations, Depth+1));
  case ISD::FP_ROUND:
      return DAG.getNode(ISD::FP_ROUND, Op.getDebugLoc(), Op.getValueType(),
                         GetNegatedExpression(Op.getOperand(0), DAG,
                                              LegalOperations, Depth+1),
                         Op.getOperand(1));
  }
}


// isSetCCEquivalent - Return true if this node is a setcc, or is a select_cc
// that selects between the values 1 and 0, making it equivalent to a setcc.
// Also, set the incoming LHS, RHS, and CC references to the appropriate
// nodes based on the type of node we are checking.  This simplifies life a
// bit for the callers.
static bool isSetCCEquivalent(SDValue N, SDValue &LHS, SDValue &RHS,
                              SDValue &CC) {
  if (N.getOpcode() == ISD::SETCC) {
    LHS = N.getOperand(0);
    RHS = N.getOperand(1);
    CC  = N.getOperand(2);
    return true;
  }
  if (N.getOpcode() == ISD::SELECT_CC &&
      N.getOperand(2).getOpcode() == ISD::Constant &&
      N.getOperand(3).getOpcode() == ISD::Constant &&
      cast<ConstantSDNode>(N.getOperand(2))->getAPIntValue() == 1 &&
      cast<ConstantSDNode>(N.getOperand(3))->isNullValue()) {
    LHS = N.getOperand(0);
    RHS = N.getOperand(1);
    CC  = N.getOperand(4);
    return true;
  }
  return false;
}

// isOneUseSetCC - Return true if this is a SetCC-equivalent operation with only
// one use.  If this is true, it allows the users to invert the operation for
// free when it is profitable to do so.
static bool isOneUseSetCC(SDValue N) {
  SDValue N0, N1, N2;
  if (isSetCCEquivalent(N, N0, N1, N2) && N.getNode()->hasOneUse())
    return true;
  return false;
}

SDValue DAGCombiner::ReassociateOps(unsigned Opc, DebugLoc DL,
                                    SDValue N0, SDValue N1) {
  EVT VT = N0.getValueType();
  if (N0.getOpcode() == Opc && isa<ConstantSDNode>(N0.getOperand(1))) {
    if (isa<ConstantSDNode>(N1)) {
      // reassoc. (op (op x, c1), c2) -> (op x, (op c1, c2))
      SDValue OpNode =
        DAG.FoldConstantArithmetic(Opc, VT,
                                   cast<ConstantSDNode>(N0.getOperand(1)),
                                   cast<ConstantSDNode>(N1));
      return DAG.getNode(Opc, DL, VT, N0.getOperand(0), OpNode);
    }
    if (N0.hasOneUse()) {
      // reassoc. (op (op x, c1), y) -> (op (op x, y), c1) iff x+c1 has one use
      SDValue OpNode = DAG.getNode(Opc, N0.getDebugLoc(), VT,
                                   N0.getOperand(0), N1);
      AddToWorkList(OpNode.getNode());
      return DAG.getNode(Opc, DL, VT, OpNode, N0.getOperand(1));
    }
  }

  if (N1.getOpcode() == Opc && isa<ConstantSDNode>(N1.getOperand(1))) {
    if (isa<ConstantSDNode>(N0)) {
      // reassoc. (op c2, (op x, c1)) -> (op x, (op c1, c2))
      SDValue OpNode =
        DAG.FoldConstantArithmetic(Opc, VT,
                                   cast<ConstantSDNode>(N1.getOperand(1)),
                                   cast<ConstantSDNode>(N0));
      return DAG.getNode(Opc, DL, VT, N1.getOperand(0), OpNode);
    }
    if (N1.hasOneUse()) {
      // reassoc. (op y, (op x, c1)) -> (op (op x, y), c1) iff x+c1 has one use
      SDValue OpNode = DAG.getNode(Opc, N0.getDebugLoc(), VT,
                                   N1.getOperand(0), N0);
      AddToWorkList(OpNode.getNode());
      return DAG.getNode(Opc, DL, VT, OpNode, N1.getOperand(1));
    }
  }

  return SDValue();
}

SDValue DAGCombiner::CombineTo(SDNode *N, const SDValue *To, unsigned NumTo,
                               bool AddTo) {
  assert(N->getNumValues() == NumTo && "Broken CombineTo call!");
  ++NodesCombined;
  DEBUG(dbgs() << "\nReplacing.1 ";
        N->dump(&DAG);
        dbgs() << "\nWith: ";
        To[0].getNode()->dump(&DAG);
        dbgs() << " and " << NumTo-1 << " other values\n";
        for (unsigned i = 0, e = NumTo; i != e; ++i)
          assert((!To[i].getNode() ||
                  N->getValueType(i) == To[i].getValueType()) &&
                 "Cannot combine value to value of different type!"));
  WorkListRemover DeadNodes(*this);
  DAG.ReplaceAllUsesWith(N, To);
  if (AddTo) {
    // Push the new nodes and any users onto the worklist
    for (unsigned i = 0, e = NumTo; i != e; ++i) {
      if (To[i].getNode()) {
        AddToWorkList(To[i].getNode());
        AddUsersToWorkList(To[i].getNode());
      }
    }
  }

  // Finally, if the node is now dead, remove it from the graph.  The node
  // may not be dead if the replacement process recursively simplified to
  // something else needing this node.
  if (N->use_empty()) {
    // Nodes can be reintroduced into the worklist.  Make sure we do not
    // process a node that has been replaced.
    removeFromWorkList(N);

    // Finally, since the node is now dead, remove it from the graph.
    DAG.DeleteNode(N);
  }
  return SDValue(N, 0);
}

void DAGCombiner::
CommitTargetLoweringOpt(const TargetLowering::TargetLoweringOpt &TLO) {
  // Replace all uses.  If any nodes become isomorphic to other nodes and
  // are deleted, make sure to remove them from our worklist.
  WorkListRemover DeadNodes(*this);
  DAG.ReplaceAllUsesOfValueWith(TLO.Old, TLO.New);

  // Push the new node and any (possibly new) users onto the worklist.
  AddToWorkList(TLO.New.getNode());
  AddUsersToWorkList(TLO.New.getNode());

  // Finally, if the node is now dead, remove it from the graph.  The node
  // may not be dead if the replacement process recursively simplified to
  // something else needing this node.
  if (TLO.Old.getNode()->use_empty()) {
    removeFromWorkList(TLO.Old.getNode());

    // If the operands of this node are only used by the node, they will now
    // be dead.  Make sure to visit them first to delete dead nodes early.
    for (unsigned i = 0, e = TLO.Old.getNode()->getNumOperands(); i != e; ++i)
      if (TLO.Old.getNode()->getOperand(i).getNode()->hasOneUse())
        AddToWorkList(TLO.Old.getNode()->getOperand(i).getNode());

    DAG.DeleteNode(TLO.Old.getNode());
  }
}

/// SimplifyDemandedBits - Check the specified integer node value to see if
/// it can be simplified or if things it uses can be simplified by bit
/// propagation.  If so, return true.
bool DAGCombiner::SimplifyDemandedBits(SDValue Op, const APInt &Demanded) {
  TargetLowering::TargetLoweringOpt TLO(DAG, LegalTypes, LegalOperations);
  APInt KnownZero, KnownOne;
  if (!TLI.SimplifyDemandedBits(Op, Demanded, KnownZero, KnownOne, TLO))
    return false;

  // Revisit the node.
  AddToWorkList(Op.getNode());

  // Replace the old value with the new one.
  ++NodesCombined;
  DEBUG(dbgs() << "\nReplacing.2 ";
        TLO.Old.getNode()->dump(&DAG);
        dbgs() << "\nWith: ";
        TLO.New.getNode()->dump(&DAG);
        dbgs() << '\n');

  CommitTargetLoweringOpt(TLO);
  return true;
}

void DAGCombiner::ReplaceLoadWithPromotedLoad(SDNode *Load, SDNode *ExtLoad) {
  DebugLoc dl = Load->getDebugLoc();
  EVT VT = Load->getValueType(0);
  SDValue Trunc = DAG.getNode(ISD::TRUNCATE, dl, VT, SDValue(ExtLoad, 0));

  DEBUG(dbgs() << "\nReplacing.9 ";
        Load->dump(&DAG);
        dbgs() << "\nWith: ";
        Trunc.getNode()->dump(&DAG);
        dbgs() << '\n');
  WorkListRemover DeadNodes(*this);
  DAG.ReplaceAllUsesOfValueWith(SDValue(Load, 0), Trunc);
  DAG.ReplaceAllUsesOfValueWith(SDValue(Load, 1), SDValue(ExtLoad, 1));
  removeFromWorkList(Load);
  DAG.DeleteNode(Load);
  AddToWorkList(Trunc.getNode());
}

SDValue DAGCombiner::PromoteOperand(SDValue Op, EVT PVT, bool &Replace) {
  Replace = false;
  DebugLoc dl = Op.getDebugLoc();
  if (LoadSDNode *LD = dyn_cast<LoadSDNode>(Op)) {
    EVT MemVT = LD->getMemoryVT();
    ISD::LoadExtType ExtType = ISD::isNON_EXTLoad(LD)
      ? (TLI.isLoadExtLegal(ISD::ZEXTLOAD, MemVT) ? ISD::ZEXTLOAD
                                                  : ISD::EXTLOAD)
      : LD->getExtensionType();
    Replace = true;
    return DAG.getExtLoad(ExtType, dl, PVT,
                          LD->getChain(), LD->getBasePtr(),
                          LD->getPointerInfo(),
                          MemVT, LD->isVolatile(),
                          LD->isNonTemporal(), LD->getAlignment());
  }

  unsigned Opc = Op.getOpcode();
  switch (Opc) {
  default: break;
  case ISD::AssertSext:
    return DAG.getNode(ISD::AssertSext, dl, PVT,
                       SExtPromoteOperand(Op.getOperand(0), PVT),
                       Op.getOperand(1));
  case ISD::AssertZext:
    return DAG.getNode(ISD::AssertZext, dl, PVT,
                       ZExtPromoteOperand(Op.getOperand(0), PVT),
                       Op.getOperand(1));
  case ISD::Constant: {
    unsigned ExtOpc =
      Op.getValueType().isByteSized() ? ISD::SIGN_EXTEND : ISD::ZERO_EXTEND;
    return DAG.getNode(ExtOpc, dl, PVT, Op);
  }
  }

  if (!TLI.isOperationLegal(ISD::ANY_EXTEND, PVT))
    return SDValue();
  return DAG.getNode(ISD::ANY_EXTEND, dl, PVT, Op);
}

SDValue DAGCombiner::SExtPromoteOperand(SDValue Op, EVT PVT) {
  if (!TLI.isOperationLegal(ISD::SIGN_EXTEND_INREG, PVT))
    return SDValue();
  EVT OldVT = Op.getValueType();
  DebugLoc dl = Op.getDebugLoc();
  bool Replace = false;
  SDValue NewOp = PromoteOperand(Op, PVT, Replace);
  if (NewOp.getNode() == 0)
    return SDValue();
  AddToWorkList(NewOp.getNode());

  if (Replace)
    ReplaceLoadWithPromotedLoad(Op.getNode(), NewOp.getNode());
  return DAG.getNode(ISD::SIGN_EXTEND_INREG, dl, NewOp.getValueType(), NewOp,
                     DAG.getValueType(OldVT));
}

SDValue DAGCombiner::ZExtPromoteOperand(SDValue Op, EVT PVT) {
  EVT OldVT = Op.getValueType();
  DebugLoc dl = Op.getDebugLoc();
  bool Replace = false;
  SDValue NewOp = PromoteOperand(Op, PVT, Replace);
  if (NewOp.getNode() == 0)
    return SDValue();
  AddToWorkList(NewOp.getNode());

  if (Replace)
    ReplaceLoadWithPromotedLoad(Op.getNode(), NewOp.getNode());
  return DAG.getZeroExtendInReg(NewOp, dl, OldVT);
}

/// PromoteIntBinOp - Promote the specified integer binary operation if the
/// target indicates it is beneficial. e.g. On x86, it's usually better to
/// promote i16 operations to i32 since i16 instructions are longer.
SDValue DAGCombiner::PromoteIntBinOp(SDValue Op) {
  if (!LegalOperations)
    return SDValue();

  EVT VT = Op.getValueType();
  if (VT.isVector() || !VT.isInteger())
    return SDValue();

  // If operation type is 'undesirable', e.g. i16 on x86, consider
  // promoting it.
  unsigned Opc = Op.getOpcode();
  if (TLI.isTypeDesirableForOp(Opc, VT))
    return SDValue();

  EVT PVT = VT;
  // Consult target whether it is a good idea to promote this operation and
  // what's the right type to promote it to.
  if (TLI.IsDesirableToPromoteOp(Op, PVT)) {
    assert(PVT != VT && "Don't know what type to promote to!");

    bool Replace0 = false;
    SDValue N0 = Op.getOperand(0);
    SDValue NN0 = PromoteOperand(N0, PVT, Replace0);
    if (NN0.getNode() == 0)
      return SDValue();

    bool Replace1 = false;
    SDValue N1 = Op.getOperand(1);
    SDValue NN1;
    if (N0 == N1)
      NN1 = NN0;
    else {
      NN1 = PromoteOperand(N1, PVT, Replace1);
      if (NN1.getNode() == 0)
        return SDValue();
    }

    AddToWorkList(NN0.getNode());
    if (NN1.getNode())
      AddToWorkList(NN1.getNode());

    if (Replace0)
      ReplaceLoadWithPromotedLoad(N0.getNode(), NN0.getNode());
    if (Replace1)
      ReplaceLoadWithPromotedLoad(N1.getNode(), NN1.getNode());

    DEBUG(dbgs() << "\nPromoting ";
          Op.getNode()->dump(&DAG));
    DebugLoc dl = Op.getDebugLoc();
    return DAG.getNode(ISD::TRUNCATE, dl, VT,
                       DAG.getNode(Opc, dl, PVT, NN0, NN1));
  }
  return SDValue();
}

/// PromoteIntShiftOp - Promote the specified integer shift operation if the
/// target indicates it is beneficial. e.g. On x86, it's usually better to
/// promote i16 operations to i32 since i16 instructions are longer.
SDValue DAGCombiner::PromoteIntShiftOp(SDValue Op) {
  if (!LegalOperations)
    return SDValue();

  EVT VT = Op.getValueType();
  if (VT.isVector() || !VT.isInteger())
    return SDValue();

  // If operation type is 'undesirable', e.g. i16 on x86, consider
  // promoting it.
  unsigned Opc = Op.getOpcode();
  if (TLI.isTypeDesirableForOp(Opc, VT))
    return SDValue();

  EVT PVT = VT;
  // Consult target whether it is a good idea to promote this operation and
  // what's the right type to promote it to.
  if (TLI.IsDesirableToPromoteOp(Op, PVT)) {
    assert(PVT != VT && "Don't know what type to promote to!");

    bool Replace = false;
    SDValue N0 = Op.getOperand(0);
    if (Opc == ISD::SRA)
      N0 = SExtPromoteOperand(Op.getOperand(0), PVT);
    else if (Opc == ISD::SRL)
      N0 = ZExtPromoteOperand(Op.getOperand(0), PVT);
    else
      N0 = PromoteOperand(N0, PVT, Replace);
    if (N0.getNode() == 0)
      return SDValue();

    AddToWorkList(N0.getNode());
    if (Replace)
      ReplaceLoadWithPromotedLoad(Op.getOperand(0).getNode(), N0.getNode());

    DEBUG(dbgs() << "\nPromoting ";
          Op.getNode()->dump(&DAG));
    DebugLoc dl = Op.getDebugLoc();
    return DAG.getNode(ISD::TRUNCATE, dl, VT,
                       DAG.getNode(Opc, dl, PVT, N0, Op.getOperand(1)));
  }
  return SDValue();
}

SDValue DAGCombiner::PromoteExtend(SDValue Op) {
  if (!LegalOperations)
    return SDValue();

  EVT VT = Op.getValueType();
  if (VT.isVector() || !VT.isInteger())
    return SDValue();

  // If operation type is 'undesirable', e.g. i16 on x86, consider
  // promoting it.
  unsigned Opc = Op.getOpcode();
  if (TLI.isTypeDesirableForOp(Opc, VT))
    return SDValue();

  EVT PVT = VT;
  // Consult target whether it is a good idea to promote this operation and
  // what's the right type to promote it to.
  if (TLI.IsDesirableToPromoteOp(Op, PVT)) {
    assert(PVT != VT && "Don't know what type to promote to!");
    // fold (aext (aext x)) -> (aext x)
    // fold (aext (zext x)) -> (zext x)
    // fold (aext (sext x)) -> (sext x)
    DEBUG(dbgs() << "\nPromoting ";
          Op.getNode()->dump(&DAG));
    return DAG.getNode(Op.getOpcode(), Op.getDebugLoc(), VT, Op.getOperand(0));
  }
  return SDValue();
}

bool DAGCombiner::PromoteLoad(SDValue Op) {
  if (!LegalOperations)
    return false;

  EVT VT = Op.getValueType();
  if (VT.isVector() || !VT.isInteger())
    return false;

  // If operation type is 'undesirable', e.g. i16 on x86, consider
  // promoting it.
  unsigned Opc = Op.getOpcode();
  if (TLI.isTypeDesirableForOp(Opc, VT))
    return false;

  EVT PVT = VT;
  // Consult target whether it is a good idea to promote this operation and
  // what's the right type to promote it to.
  if (TLI.IsDesirableToPromoteOp(Op, PVT)) {
    assert(PVT != VT && "Don't know what type to promote to!");

    DebugLoc dl = Op.getDebugLoc();
    SDNode *N = Op.getNode();
    LoadSDNode *LD = cast<LoadSDNode>(N);
    EVT MemVT = LD->getMemoryVT();
    ISD::LoadExtType ExtType = ISD::isNON_EXTLoad(LD)
      ? (TLI.isLoadExtLegal(ISD::ZEXTLOAD, MemVT) ? ISD::ZEXTLOAD
                                                  : ISD::EXTLOAD)
      : LD->getExtensionType();
    SDValue NewLD = DAG.getExtLoad(ExtType, dl, PVT,
                                   LD->getChain(), LD->getBasePtr(),
                                   LD->getPointerInfo(),
                                   MemVT, LD->isVolatile(),
                                   LD->isNonTemporal(), LD->getAlignment());
    SDValue Result = DAG.getNode(ISD::TRUNCATE, dl, VT, NewLD);

    DEBUG(dbgs() << "\nPromoting ";
          N->dump(&DAG);
          dbgs() << "\nTo: ";
          Result.getNode()->dump(&DAG);
          dbgs() << '\n');
    WorkListRemover DeadNodes(*this);
    DAG.ReplaceAllUsesOfValueWith(SDValue(N, 0), Result);
    DAG.ReplaceAllUsesOfValueWith(SDValue(N, 1), NewLD.getValue(1));
    removeFromWorkList(N);
    DAG.DeleteNode(N);
    AddToWorkList(Result.getNode());
    return true;
  }
  return false;
}


//===----------------------------------------------------------------------===//
//  Main DAG Combiner implementation
//===----------------------------------------------------------------------===//

void DAGCombiner::Run(CombineLevel AtLevel) {
  // set the instance variables, so that the various visit routines may use it.
  Level = AtLevel;
  LegalOperations = Level >= AfterLegalizeVectorOps;
  LegalTypes = Level >= AfterLegalizeTypes;

  // Add all the dag nodes to the worklist.
  for (SelectionDAG::allnodes_iterator I = DAG.allnodes_begin(),
       E = DAG.allnodes_end(); I != E; ++I)
    AddToWorkList(I);

  // Create a dummy node (which is not added to allnodes), that adds a reference
  // to the root node, preventing it from being deleted, and tracking any
  // changes of the root.
  HandleSDNode Dummy(DAG.getRoot());

  // The root of the dag may dangle to deleted nodes until the dag combiner is
  // done.  Set it to null to avoid confusion.
  DAG.setRoot(SDValue());

  // while the worklist isn't empty, find a node and
  // try and combine it.
  while (!WorkListContents.empty()) {
    SDNode *N;
    // The WorkListOrder holds the SDNodes in order, but it may contain duplicates.
    // In order to avoid a linear scan, we use a set (O(log N)) to hold what the
    // worklist *should* contain, and check the node we want to visit is should
    // actually be visited.
    do {
      N = WorkListOrder.pop_back_val();
    } while (!WorkListContents.erase(N));

    // If N has no uses, it is dead.  Make sure to revisit all N's operands once
    // N is deleted from the DAG, since they too may now be dead or may have a
    // reduced number of uses, allowing other xforms.
    if (N->use_empty() && N != &Dummy) {
      for (unsigned i = 0, e = N->getNumOperands(); i != e; ++i)
        AddToWorkList(N->getOperand(i).getNode());

      DAG.DeleteNode(N);
      continue;
    }

    SDValue RV = combine(N);

    if (RV.getNode() == 0)
      continue;

    ++NodesCombined;

    // If we get back the same node we passed in, rather than a new node or
    // zero, we know that the node must have defined multiple values and
    // CombineTo was used.  Since CombineTo takes care of the worklist
    // mechanics for us, we have no work to do in this case.
    if (RV.getNode() == N)
      continue;

    assert(N->getOpcode() != ISD::DELETED_NODE &&
           RV.getNode()->getOpcode() != ISD::DELETED_NODE &&
           "Node was deleted but visit returned new node!");

    DEBUG(dbgs() << "\nReplacing.3 ";
          N->dump(&DAG);
          dbgs() << "\nWith: ";
          RV.getNode()->dump(&DAG);
          dbgs() << '\n');

    // Transfer debug value.
    DAG.TransferDbgValues(SDValue(N, 0), RV);
    WorkListRemover DeadNodes(*this);
    if (N->getNumValues() == RV.getNode()->getNumValues())
      DAG.ReplaceAllUsesWith(N, RV.getNode());
    else {
      assert(N->getValueType(0) == RV.getValueType() &&
             N->getNumValues() == 1 && "Type mismatch");
      SDValue OpV = RV;
      DAG.ReplaceAllUsesWith(N, &OpV);
    }

    // Push the new node and any users onto the worklist
    AddToWorkList(RV.getNode());
    AddUsersToWorkList(RV.getNode());

    // Add any uses of the old node to the worklist in case this node is the
    // last one that uses them.  They may become dead after this node is
    // deleted.
    for (unsigned i = 0, e = N->getNumOperands(); i != e; ++i)
      AddToWorkList(N->getOperand(i).getNode());

    // Finally, if the node is now dead, remove it from the graph.  The node
    // may not be dead if the replacement process recursively simplified to
    // something else needing this node.
    if (N->use_empty()) {
      // Nodes can be reintroduced into the worklist.  Make sure we do not
      // process a node that has been replaced.
      removeFromWorkList(N);

      // Finally, since the node is now dead, remove it from the graph.
      DAG.DeleteNode(N);
    }
  }

  // If the root changed (e.g. it was a dead load, update the root).
  DAG.setRoot(Dummy.getValue());
  DAG.RemoveDeadNodes();
}

SDValue DAGCombiner::visit(SDNode *N) {
  switch (N->getOpcode()) {
  default: break;
  case ISD::TokenFactor:        return visitTokenFactor(N);
  case ISD::MERGE_VALUES:       return visitMERGE_VALUES(N);
  case ISD::ADD:                return visitADD(N);
  case ISD::SUB:                return visitSUB(N);
  case ISD::ADDC:               return visitADDC(N);
  case ISD::SUBC:               return visitSUBC(N);
  case ISD::ADDE:               return visitADDE(N);
  case ISD::SUBE:               return visitSUBE(N);
  case ISD::MUL:                return visitMUL(N);
  case ISD::SDIV:               return visitSDIV(N);
  case ISD::UDIV:               return visitUDIV(N);
  case ISD::SREM:               return visitSREM(N);
  case ISD::UREM:               return visitUREM(N);
  case ISD::MULHU:              return visitMULHU(N);
  case ISD::MULHS:              return visitMULHS(N);
  case ISD::SMUL_LOHI:          return visitSMUL_LOHI(N);
  case ISD::UMUL_LOHI:          return visitUMUL_LOHI(N);
  case ISD::SMULO:              return visitSMULO(N);
  case ISD::UMULO:              return visitUMULO(N);
  case ISD::SDIVREM:            return visitSDIVREM(N);
  case ISD::UDIVREM:            return visitUDIVREM(N);
  case ISD::AND:                return visitAND(N);
  case ISD::OR:                 return visitOR(N);
  case ISD::XOR:                return visitXOR(N);
  case ISD::SHL:                return visitSHL(N);
  case ISD::SRA:                return visitSRA(N);
  case ISD::SRL:                return visitSRL(N);
  case ISD::CTLZ:               return visitCTLZ(N);
  case ISD::CTLZ_ZERO_UNDEF:    return visitCTLZ_ZERO_UNDEF(N);
  case ISD::CTTZ:               return visitCTTZ(N);
  case ISD::CTTZ_ZERO_UNDEF:    return visitCTTZ_ZERO_UNDEF(N);
  case ISD::CTPOP:              return visitCTPOP(N);
  case ISD::SELECT:             return visitSELECT(N);
  case ISD::SELECT_CC:          return visitSELECT_CC(N);
  case ISD::SETCC:              return visitSETCC(N);
  case ISD::SIGN_EXTEND:        return visitSIGN_EXTEND(N);
  case ISD::ZERO_EXTEND:        return visitZERO_EXTEND(N);
  case ISD::ANY_EXTEND:         return visitANY_EXTEND(N);
  case ISD::SIGN_EXTEND_INREG:  return visitSIGN_EXTEND_INREG(N);
  case ISD::TRUNCATE:           return visitTRUNCATE(N);
  case ISD::BITCAST:            return visitBITCAST(N);
  case ISD::BUILD_PAIR:         return visitBUILD_PAIR(N);
  case ISD::FADD:               return visitFADD(N);
  case ISD::FSUB:               return visitFSUB(N);
  case ISD::FMUL:               return visitFMUL(N);
  case ISD::FMA:                return visitFMA(N);
  case ISD::FDIV:               return visitFDIV(N);
  case ISD::FREM:               return visitFREM(N);
  case ISD::FCOPYSIGN:          return visitFCOPYSIGN(N);
  case ISD::SINT_TO_FP:         return visitSINT_TO_FP(N);
  case ISD::UINT_TO_FP:         return visitUINT_TO_FP(N);
  case ISD::FP_TO_SINT:         return visitFP_TO_SINT(N);
  case ISD::FP_TO_UINT:         return visitFP_TO_UINT(N);
  case ISD::FP_ROUND:           return visitFP_ROUND(N);
  case ISD::FP_ROUND_INREG:     return visitFP_ROUND_INREG(N);
  case ISD::FP_EXTEND:          return visitFP_EXTEND(N);
  case ISD::FNEG:               return visitFNEG(N);
  case ISD::FABS:               return visitFABS(N);
  case ISD::FFLOOR:             return visitFFLOOR(N);
  case ISD::FCEIL:              return visitFCEIL(N);
  case ISD::FTRUNC:             return visitFTRUNC(N);
  case ISD::BRCOND:             return visitBRCOND(N);
  case ISD::BR_CC:              return visitBR_CC(N);
  case ISD::LOAD:               return visitLOAD(N);
  case ISD::STORE:              return visitSTORE(N);
  case ISD::INSERT_VECTOR_ELT:  return visitINSERT_VECTOR_ELT(N);
  case ISD::EXTRACT_VECTOR_ELT: return visitEXTRACT_VECTOR_ELT(N);
  case ISD::BUILD_VECTOR:       return visitBUILD_VECTOR(N);
  case ISD::CONCAT_VECTORS:     return visitCONCAT_VECTORS(N);
  case ISD::EXTRACT_SUBVECTOR:  return visitEXTRACT_SUBVECTOR(N);
  case ISD::VECTOR_SHUFFLE:     return visitVECTOR_SHUFFLE(N);
  case ISD::MEMBARRIER:         return visitMEMBARRIER(N);
  }
  return SDValue();
}

SDValue DAGCombiner::combine(SDNode *N) {
  SDValue RV = visit(N);

  // If nothing happened, try a target-specific DAG combine.
  if (RV.getNode() == 0) {
    assert(N->getOpcode() != ISD::DELETED_NODE &&
           "Node was deleted but visit returned NULL!");

    if (N->getOpcode() >= ISD::BUILTIN_OP_END ||
        TLI.hasTargetDAGCombine((ISD::NodeType)N->getOpcode())) {

      // Expose the DAG combiner to the target combiner impls.
      TargetLowering::DAGCombinerInfo
        DagCombineInfo(DAG, !LegalTypes, !LegalOperations, false, this);

      RV = TLI.PerformDAGCombine(N, DagCombineInfo);
    }
  }

  // If nothing happened still, try promoting the operation.
  if (RV.getNode() == 0) {
    switch (N->getOpcode()) {
    default: break;
    case ISD::ADD:
    case ISD::SUB:
    case ISD::MUL:
    case ISD::AND:
    case ISD::OR:
    case ISD::XOR:
      RV = PromoteIntBinOp(SDValue(N, 0));
      break;
    case ISD::SHL:
    case ISD::SRA:
    case ISD::SRL:
      RV = PromoteIntShiftOp(SDValue(N, 0));
      break;
    case ISD::SIGN_EXTEND:
    case ISD::ZERO_EXTEND:
    case ISD::ANY_EXTEND:
      RV = PromoteExtend(SDValue(N, 0));
      break;
    case ISD::LOAD:
      if (PromoteLoad(SDValue(N, 0)))
        RV = SDValue(N, 0);
      break;
    }
  }

  // If N is a commutative binary node, try commuting it to enable more
  // sdisel CSE.
  if (RV.getNode() == 0 &&
      SelectionDAG::isCommutativeBinOp(N->getOpcode()) &&
      N->getNumValues() == 1) {
    SDValue N0 = N->getOperand(0);
    SDValue N1 = N->getOperand(1);

    // Constant operands are canonicalized to RHS.
    if (isa<ConstantSDNode>(N0) || !isa<ConstantSDNode>(N1)) {
      SDValue Ops[] = { N1, N0 };
      SDNode *CSENode = DAG.getNodeIfExists(N->getOpcode(), N->getVTList(),
                                            Ops, 2);
      if (CSENode)
        return SDValue(CSENode, 0);
    }
  }

  return RV;
}

/// getInputChainForNode - Given a node, return its input chain if it has one,
/// otherwise return a null sd operand.
static SDValue getInputChainForNode(SDNode *N) {
  if (unsigned NumOps = N->getNumOperands()) {
    if (N->getOperand(0).getValueType() == MVT::Other)
      return N->getOperand(0);
    else if (N->getOperand(NumOps-1).getValueType() == MVT::Other)
      return N->getOperand(NumOps-1);
    for (unsigned i = 1; i < NumOps-1; ++i)
      if (N->getOperand(i).getValueType() == MVT::Other)
        return N->getOperand(i);
  }
  return SDValue();
}

SDValue DAGCombiner::visitTokenFactor(SDNode *N) {
  // If N has two operands, where one has an input chain equal to the other,
  // the 'other' chain is redundant.
  if (N->getNumOperands() == 2) {
    if (getInputChainForNode(N->getOperand(0).getNode()) == N->getOperand(1))
      return N->getOperand(0);
    if (getInputChainForNode(N->getOperand(1).getNode()) == N->getOperand(0))
      return N->getOperand(1);
  }

  SmallVector<SDNode *, 8> TFs;     // List of token factors to visit.
  SmallVector<SDValue, 8> Ops;    // Ops for replacing token factor.
  SmallPtrSet<SDNode*, 16> SeenOps;
  bool Changed = false;             // If we should replace this token factor.

  // Start out with this token factor.
  TFs.push_back(N);

  // Iterate through token factors.  The TFs grows when new token factors are
  // encountered.
  for (unsigned i = 0; i < TFs.size(); ++i) {
    SDNode *TF = TFs[i];

    // Check each of the operands.
    for (unsigned i = 0, ie = TF->getNumOperands(); i != ie; ++i) {
      SDValue Op = TF->getOperand(i);

      switch (Op.getOpcode()) {
      case ISD::EntryToken:
        // Entry tokens don't need to be added to the list. They are
        // rededundant.
        Changed = true;
        break;

      case ISD::TokenFactor:
        if (Op.hasOneUse() &&
            std::find(TFs.begin(), TFs.end(), Op.getNode()) == TFs.end()) {
          // Queue up for processing.
          TFs.push_back(Op.getNode());
          // Clean up in case the token factor is removed.
          AddToWorkList(Op.getNode());
          Changed = true;
          break;
        }
        // Fall thru

      default:
        // Only add if it isn't already in the list.
        if (SeenOps.insert(Op.getNode()))
          Ops.push_back(Op);
        else
          Changed = true;
        break;
      }
    }
  }

  SDValue Result;

  // If we've change things around then replace token factor.
  if (Changed) {
    if (Ops.empty()) {
      // The entry token is the only possible outcome.
      Result = DAG.getEntryNode();
    } else {
      // New and improved token factor.
      Result = DAG.getNode(ISD::TokenFactor, N->getDebugLoc(),
                           MVT::Other, &Ops[0], Ops.size());
    }

    // Don't add users to work list.
    return CombineTo(N, Result, false);
  }

  return Result;
}

/// MERGE_VALUES can always be eliminated.
SDValue DAGCombiner::visitMERGE_VALUES(SDNode *N) {
  WorkListRemover DeadNodes(*this);
  // Replacing results may cause a different MERGE_VALUES to suddenly
  // be CSE'd with N, and carry its uses with it. Iterate until no
  // uses remain, to ensure that the node can be safely deleted.
  // First add the users of this node to the work list so that they
  // can be tried again once they have new operands.
  AddUsersToWorkList(N);
  do {
    for (unsigned i = 0, e = N->getNumOperands(); i != e; ++i)
      DAG.ReplaceAllUsesOfValueWith(SDValue(N, i), N->getOperand(i));
  } while (!N->use_empty());
  removeFromWorkList(N);
  DAG.DeleteNode(N);
  return SDValue(N, 0);   // Return N so it doesn't get rechecked!
}

static
SDValue combineShlAddConstant(DebugLoc DL, SDValue N0, SDValue N1,
                              SelectionDAG &DAG) {
  EVT VT = N0.getValueType();
  SDValue N00 = N0.getOperand(0);
  SDValue N01 = N0.getOperand(1);
  ConstantSDNode *N01C = dyn_cast<ConstantSDNode>(N01);

  if (N01C && N00.getOpcode() == ISD::ADD && N00.getNode()->hasOneUse() &&
      isa<ConstantSDNode>(N00.getOperand(1))) {
    // fold (add (shl (add x, c1), c2), ) -> (add (add (shl x, c2), c1<<c2), )
    N0 = DAG.getNode(ISD::ADD, N0.getDebugLoc(), VT,
                     DAG.getNode(ISD::SHL, N00.getDebugLoc(), VT,
                                 N00.getOperand(0), N01),
                     DAG.getNode(ISD::SHL, N01.getDebugLoc(), VT,
                                 N00.getOperand(1), N01));
    return DAG.getNode(ISD::ADD, DL, VT, N0, N1);
  }

  return SDValue();
}

SDValue DAGCombiner::visitADD(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  SDValue N1 = N->getOperand(1);
  ConstantSDNode *N0C = dyn_cast<ConstantSDNode>(N0);
  ConstantSDNode *N1C = dyn_cast<ConstantSDNode>(N1);
  EVT VT = N0.getValueType();

  // fold vector ops
  if (VT.isVector()) {
    SDValue FoldedVOp = SimplifyVBinOp(N);
    if (FoldedVOp.getNode()) return FoldedVOp;
  }

  // fold (add x, undef) -> undef
  if (N0.getOpcode() == ISD::UNDEF)
    return N0;
  if (N1.getOpcode() == ISD::UNDEF)
    return N1;
  // fold (add c1, c2) -> c1+c2
  if (N0C && N1C)
    return DAG.FoldConstantArithmetic(ISD::ADD, VT, N0C, N1C);
  // canonicalize constant to RHS
  if (N0C && !N1C)
    return DAG.getNode(ISD::ADD, N->getDebugLoc(), VT, N1, N0);
  // fold (add x, 0) -> x
  if (N1C && N1C->isNullValue())
    return N0;
  // fold (add Sym, c) -> Sym+c
  if (GlobalAddressSDNode *GA = dyn_cast<GlobalAddressSDNode>(N0))
    if (!LegalOperations && TLI.isOffsetFoldingLegal(GA) && N1C &&
        GA->getOpcode() == ISD::GlobalAddress)
      return DAG.getGlobalAddress(GA->getGlobal(), N1C->getDebugLoc(), VT,
                                  GA->getOffset() +
                                    (uint64_t)N1C->getSExtValue());
  // fold ((c1-A)+c2) -> (c1+c2)-A
  if (N1C && N0.getOpcode() == ISD::SUB)
    if (ConstantSDNode *N0C = dyn_cast<ConstantSDNode>(N0.getOperand(0)))
      return DAG.getNode(ISD::SUB, N->getDebugLoc(), VT,
                         DAG.getConstant(N1C->getAPIntValue()+
                                         N0C->getAPIntValue(), VT),
                         N0.getOperand(1));
  // reassociate add
  SDValue RADD = ReassociateOps(ISD::ADD, N->getDebugLoc(), N0, N1);
  if (RADD.getNode() != 0)
    return RADD;
  // fold ((0-A) + B) -> B-A
  if (N0.getOpcode() == ISD::SUB && isa<ConstantSDNode>(N0.getOperand(0)) &&
      cast<ConstantSDNode>(N0.getOperand(0))->isNullValue())
    return DAG.getNode(ISD::SUB, N->getDebugLoc(), VT, N1, N0.getOperand(1));
  // fold (A + (0-B)) -> A-B
  if (N1.getOpcode() == ISD::SUB && isa<ConstantSDNode>(N1.getOperand(0)) &&
      cast<ConstantSDNode>(N1.getOperand(0))->isNullValue())
    return DAG.getNode(ISD::SUB, N->getDebugLoc(), VT, N0, N1.getOperand(1));
  // fold (A+(B-A)) -> B
  if (N1.getOpcode() == ISD::SUB && N0 == N1.getOperand(1))
    return N1.getOperand(0);
  // fold ((B-A)+A) -> B
  if (N0.getOpcode() == ISD::SUB && N1 == N0.getOperand(1))
    return N0.getOperand(0);
  // fold (A+(B-(A+C))) to (B-C)
  if (N1.getOpcode() == ISD::SUB && N1.getOperand(1).getOpcode() == ISD::ADD &&
      N0 == N1.getOperand(1).getOperand(0))
    return DAG.getNode(ISD::SUB, N->getDebugLoc(), VT, N1.getOperand(0),
                       N1.getOperand(1).getOperand(1));
  // fold (A+(B-(C+A))) to (B-C)
  if (N1.getOpcode() == ISD::SUB && N1.getOperand(1).getOpcode() == ISD::ADD &&
      N0 == N1.getOperand(1).getOperand(1))
    return DAG.getNode(ISD::SUB, N->getDebugLoc(), VT, N1.getOperand(0),
                       N1.getOperand(1).getOperand(0));
  // fold (A+((B-A)+or-C)) to (B+or-C)
  if ((N1.getOpcode() == ISD::SUB || N1.getOpcode() == ISD::ADD) &&
      N1.getOperand(0).getOpcode() == ISD::SUB &&
      N0 == N1.getOperand(0).getOperand(1))
    return DAG.getNode(N1.getOpcode(), N->getDebugLoc(), VT,
                       N1.getOperand(0).getOperand(0), N1.getOperand(1));

  // fold (A-B)+(C-D) to (A+C)-(B+D) when A or C is constant
  if (N0.getOpcode() == ISD::SUB && N1.getOpcode() == ISD::SUB) {
    SDValue N00 = N0.getOperand(0);
    SDValue N01 = N0.getOperand(1);
    SDValue N10 = N1.getOperand(0);
    SDValue N11 = N1.getOperand(1);

    if (isa<ConstantSDNode>(N00) || isa<ConstantSDNode>(N10))
      return DAG.getNode(ISD::SUB, N->getDebugLoc(), VT,
                         DAG.getNode(ISD::ADD, N0.getDebugLoc(), VT, N00, N10),
                         DAG.getNode(ISD::ADD, N1.getDebugLoc(), VT, N01, N11));
  }

  if (!VT.isVector() && SimplifyDemandedBits(SDValue(N, 0)))
    return SDValue(N, 0);

  // fold (a+b) -> (a|b) iff a and b share no bits.
  if (VT.isInteger() && !VT.isVector()) {
    APInt LHSZero, LHSOne;
    APInt RHSZero, RHSOne;
    DAG.ComputeMaskedBits(N0, LHSZero, LHSOne);

    if (LHSZero.getBoolValue()) {
      DAG.ComputeMaskedBits(N1, RHSZero, RHSOne);

      // If all possibly-set bits on the LHS are clear on the RHS, return an OR.
      // If all possibly-set bits on the RHS are clear on the LHS, return an OR.
      if ((RHSZero & ~LHSZero) == ~LHSZero || (LHSZero & ~RHSZero) == ~RHSZero)
        return DAG.getNode(ISD::OR, N->getDebugLoc(), VT, N0, N1);
    }
  }

  // fold (add (shl (add x, c1), c2), ) -> (add (add (shl x, c2), c1<<c2), )
  if (N0.getOpcode() == ISD::SHL && N0.getNode()->hasOneUse()) {
    SDValue Result = combineShlAddConstant(N->getDebugLoc(), N0, N1, DAG);
    if (Result.getNode()) return Result;
  }
  if (N1.getOpcode() == ISD::SHL && N1.getNode()->hasOneUse()) {
    SDValue Result = combineShlAddConstant(N->getDebugLoc(), N1, N0, DAG);
    if (Result.getNode()) return Result;
  }

  // fold (add x, shl(0 - y, n)) -> sub(x, shl(y, n))
  if (N1.getOpcode() == ISD::SHL &&
      N1.getOperand(0).getOpcode() == ISD::SUB)
    if (ConstantSDNode *C =
          dyn_cast<ConstantSDNode>(N1.getOperand(0).getOperand(0)))
      if (C->getAPIntValue() == 0)
        return DAG.getNode(ISD::SUB, N->getDebugLoc(), VT, N0,
                           DAG.getNode(ISD::SHL, N->getDebugLoc(), VT,
                                       N1.getOperand(0).getOperand(1),
                                       N1.getOperand(1)));
  if (N0.getOpcode() == ISD::SHL &&
      N0.getOperand(0).getOpcode() == ISD::SUB)
    if (ConstantSDNode *C =
          dyn_cast<ConstantSDNode>(N0.getOperand(0).getOperand(0)))
      if (C->getAPIntValue() == 0)
        return DAG.getNode(ISD::SUB, N->getDebugLoc(), VT, N1,
                           DAG.getNode(ISD::SHL, N->getDebugLoc(), VT,
                                       N0.getOperand(0).getOperand(1),
                                       N0.getOperand(1)));

  if (N1.getOpcode() == ISD::AND) {
    SDValue AndOp0 = N1.getOperand(0);
    ConstantSDNode *AndOp1 = dyn_cast<ConstantSDNode>(N1->getOperand(1));
    unsigned NumSignBits = DAG.ComputeNumSignBits(AndOp0);
    unsigned DestBits = VT.getScalarType().getSizeInBits();

    // (add z, (and (sbbl x, x), 1)) -> (sub z, (sbbl x, x))
    // and similar xforms where the inner op is either ~0 or 0.
    if (NumSignBits == DestBits && AndOp1 && AndOp1->isOne()) {
      DebugLoc DL = N->getDebugLoc();
      return DAG.getNode(ISD::SUB, DL, VT, N->getOperand(0), AndOp0);
    }
  }

  // add (sext i1), X -> sub X, (zext i1)
  if (N0.getOpcode() == ISD::SIGN_EXTEND &&
      N0.getOperand(0).getValueType() == MVT::i1 &&
      !TLI.isOperationLegal(ISD::SIGN_EXTEND, MVT::i1)) {
    DebugLoc DL = N->getDebugLoc();
    SDValue ZExt = DAG.getNode(ISD::ZERO_EXTEND, DL, VT, N0.getOperand(0));
    return DAG.getNode(ISD::SUB, DL, VT, N1, ZExt);
  }

  return SDValue();
}

SDValue DAGCombiner::visitADDC(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  SDValue N1 = N->getOperand(1);
  ConstantSDNode *N0C = dyn_cast<ConstantSDNode>(N0);
  ConstantSDNode *N1C = dyn_cast<ConstantSDNode>(N1);
  EVT VT = N0.getValueType();

  // If the flag result is dead, turn this into an ADD.
  if (!N->hasAnyUseOfValue(1))
    return CombineTo(N, DAG.getNode(ISD::ADD, N->getDebugLoc(), VT, N0, N1),
                     DAG.getNode(ISD::CARRY_FALSE,
                                 N->getDebugLoc(), MVT::Glue));

  // canonicalize constant to RHS.
  if (N0C && !N1C)
    return DAG.getNode(ISD::ADDC, N->getDebugLoc(), N->getVTList(), N1, N0);

  // fold (addc x, 0) -> x + no carry out
  if (N1C && N1C->isNullValue())
    return CombineTo(N, N0, DAG.getNode(ISD::CARRY_FALSE,
                                        N->getDebugLoc(), MVT::Glue));

  // fold (addc a, b) -> (or a, b), CARRY_FALSE iff a and b share no bits.
  APInt LHSZero, LHSOne;
  APInt RHSZero, RHSOne;
  DAG.ComputeMaskedBits(N0, LHSZero, LHSOne);

  if (LHSZero.getBoolValue()) {
    DAG.ComputeMaskedBits(N1, RHSZero, RHSOne);

    // If all possibly-set bits on the LHS are clear on the RHS, return an OR.
    // If all possibly-set bits on the RHS are clear on the LHS, return an OR.
    if ((RHSZero & ~LHSZero) == ~LHSZero || (LHSZero & ~RHSZero) == ~RHSZero)
      return CombineTo(N, DAG.getNode(ISD::OR, N->getDebugLoc(), VT, N0, N1),
                       DAG.getNode(ISD::CARRY_FALSE,
                                   N->getDebugLoc(), MVT::Glue));
  }

  return SDValue();
}

SDValue DAGCombiner::visitADDE(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  SDValue N1 = N->getOperand(1);
  SDValue CarryIn = N->getOperand(2);
  ConstantSDNode *N0C = dyn_cast<ConstantSDNode>(N0);
  ConstantSDNode *N1C = dyn_cast<ConstantSDNode>(N1);

  // canonicalize constant to RHS
  if (N0C && !N1C)
    return DAG.getNode(ISD::ADDE, N->getDebugLoc(), N->getVTList(),
                       N1, N0, CarryIn);

  // fold (adde x, y, false) -> (addc x, y)
  if (CarryIn.getOpcode() == ISD::CARRY_FALSE)
    return DAG.getNode(ISD::ADDC, N->getDebugLoc(), N->getVTList(), N0, N1);

  return SDValue();
}

// Since it may not be valid to emit a fold to zero for vector initializers
// check if we can before folding.
static SDValue tryFoldToZero(DebugLoc DL, const TargetLowering &TLI, EVT VT,
                             SelectionDAG &DAG, bool LegalOperations) {
  if (!VT.isVector()) {
    return DAG.getConstant(0, VT);
  }
  if (!LegalOperations || TLI.isOperationLegal(ISD::BUILD_VECTOR, VT)) {
    // Produce a vector of zeros.
    SDValue El = DAG.getConstant(0, VT.getVectorElementType());
    std::vector<SDValue> Ops(VT.getVectorNumElements(), El);
    return DAG.getNode(ISD::BUILD_VECTOR, DL, VT,
      &Ops[0], Ops.size());
  }
  return SDValue();
}

SDValue DAGCombiner::visitSUB(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  SDValue N1 = N->getOperand(1);
  ConstantSDNode *N0C = dyn_cast<ConstantSDNode>(N0.getNode());
  ConstantSDNode *N1C = dyn_cast<ConstantSDNode>(N1.getNode());
  ConstantSDNode *N1C1 = N1.getOpcode() != ISD::ADD ? 0 :
    dyn_cast<ConstantSDNode>(N1.getOperand(1).getNode());
  EVT VT = N0.getValueType();

  // fold vector ops
  if (VT.isVector()) {
    SDValue FoldedVOp = SimplifyVBinOp(N);
    if (FoldedVOp.getNode()) return FoldedVOp;
  }

  // fold (sub x, x) -> 0
  // FIXME: Refactor this and xor and other similar operations together.
  if (N0 == N1)
    return tryFoldToZero(N->getDebugLoc(), TLI, VT, DAG, LegalOperations);
  // fold (sub c1, c2) -> c1-c2
  if (N0C && N1C)
    return DAG.FoldConstantArithmetic(ISD::SUB, VT, N0C, N1C);
  // fold (sub x, c) -> (add x, -c)
  if (N1C)
    return DAG.getNode(ISD::ADD, N->getDebugLoc(), VT, N0,
                       DAG.getConstant(-N1C->getAPIntValue(), VT));
  // Canonicalize (sub -1, x) -> ~x, i.e. (xor x, -1)
  if (N0C && N0C->isAllOnesValue())
    return DAG.getNode(ISD::XOR, N->getDebugLoc(), VT, N1, N0);
  // fold A-(A-B) -> B
  if (N1.getOpcode() == ISD::SUB && N0 == N1.getOperand(0))
    return N1.getOperand(1);
  // fold (A+B)-A -> B
  if (N0.getOpcode() == ISD::ADD && N0.getOperand(0) == N1)
    return N0.getOperand(1);
  // fold (A+B)-B -> A
  if (N0.getOpcode() == ISD::ADD && N0.getOperand(1) == N1)
    return N0.getOperand(0);
  // fold C2-(A+C1) -> (C2-C1)-A
  if (N1.getOpcode() == ISD::ADD && N0C && N1C1) {
    SDValue NewC = DAG.getConstant(N0C->getAPIntValue() - N1C1->getAPIntValue(),
                                   VT);
    return DAG.getNode(ISD::SUB, N->getDebugLoc(), VT, NewC,
                       N1.getOperand(0));
  }
  // fold ((A+(B+or-C))-B) -> A+or-C
  if (N0.getOpcode() == ISD::ADD &&
      (N0.getOperand(1).getOpcode() == ISD::SUB ||
       N0.getOperand(1).getOpcode() == ISD::ADD) &&
      N0.getOperand(1).getOperand(0) == N1)
    return DAG.getNode(N0.getOperand(1).getOpcode(), N->getDebugLoc(), VT,
                       N0.getOperand(0), N0.getOperand(1).getOperand(1));
  // fold ((A+(C+B))-B) -> A+C
  if (N0.getOpcode() == ISD::ADD &&
      N0.getOperand(1).getOpcode() == ISD::ADD &&
      N0.getOperand(1).getOperand(1) == N1)
    return DAG.getNode(ISD::ADD, N->getDebugLoc(), VT,
                       N0.getOperand(0), N0.getOperand(1).getOperand(0));
  // fold ((A-(B-C))-C) -> A-B
  if (N0.getOpcode() == ISD::SUB &&
      N0.getOperand(1).getOpcode() == ISD::SUB &&
      N0.getOperand(1).getOperand(1) == N1)
    return DAG.getNode(ISD::SUB, N->getDebugLoc(), VT,
                       N0.getOperand(0), N0.getOperand(1).getOperand(0));

  // If either operand of a sub is undef, the result is undef
  if (N0.getOpcode() == ISD::UNDEF)
    return N0;
  if (N1.getOpcode() == ISD::UNDEF)
    return N1;

  // If the relocation model supports it, consider symbol offsets.
  if (GlobalAddressSDNode *GA = dyn_cast<GlobalAddressSDNode>(N0))
    if (!LegalOperations && TLI.isOffsetFoldingLegal(GA)) {
      // fold (sub Sym, c) -> Sym-c
      if (N1C && GA->getOpcode() == ISD::GlobalAddress)
        return DAG.getGlobalAddress(GA->getGlobal(), N1C->getDebugLoc(), VT,
                                    GA->getOffset() -
                                      (uint64_t)N1C->getSExtValue());
      // fold (sub Sym+c1, Sym+c2) -> c1-c2
      if (GlobalAddressSDNode *GB = dyn_cast<GlobalAddressSDNode>(N1))
        if (GA->getGlobal() == GB->getGlobal())
          return DAG.getConstant((uint64_t)GA->getOffset() - GB->getOffset(),
                                 VT);
    }

  return SDValue();
}

SDValue DAGCombiner::visitSUBC(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  SDValue N1 = N->getOperand(1);
  ConstantSDNode *N0C = dyn_cast<ConstantSDNode>(N0);
  ConstantSDNode *N1C = dyn_cast<ConstantSDNode>(N1);
  EVT VT = N0.getValueType();

  // If the flag result is dead, turn this into an SUB.
  if (!N->hasAnyUseOfValue(1))
    return CombineTo(N, DAG.getNode(ISD::SUB, N->getDebugLoc(), VT, N0, N1),
                     DAG.getNode(ISD::CARRY_FALSE, N->getDebugLoc(),
                                 MVT::Glue));

  // fold (subc x, x) -> 0 + no borrow
  if (N0 == N1)
    return CombineTo(N, DAG.getConstant(0, VT),
                     DAG.getNode(ISD::CARRY_FALSE, N->getDebugLoc(),
                                 MVT::Glue));

  // fold (subc x, 0) -> x + no borrow
  if (N1C && N1C->isNullValue())
    return CombineTo(N, N0, DAG.getNode(ISD::CARRY_FALSE, N->getDebugLoc(),
                                        MVT::Glue));

  // Canonicalize (sub -1, x) -> ~x, i.e. (xor x, -1) + no borrow
  if (N0C && N0C->isAllOnesValue())
    return CombineTo(N, DAG.getNode(ISD::XOR, N->getDebugLoc(), VT, N1, N0),
                     DAG.getNode(ISD::CARRY_FALSE, N->getDebugLoc(),
                                 MVT::Glue));

  return SDValue();
}

SDValue DAGCombiner::visitSUBE(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  SDValue N1 = N->getOperand(1);
  SDValue CarryIn = N->getOperand(2);

  // fold (sube x, y, false) -> (subc x, y)
  if (CarryIn.getOpcode() == ISD::CARRY_FALSE)
    return DAG.getNode(ISD::SUBC, N->getDebugLoc(), N->getVTList(), N0, N1);

  return SDValue();
}

SDValue DAGCombiner::visitMUL(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  SDValue N1 = N->getOperand(1);
  ConstantSDNode *N0C = dyn_cast<ConstantSDNode>(N0);
  ConstantSDNode *N1C = dyn_cast<ConstantSDNode>(N1);
  EVT VT = N0.getValueType();

  // fold vector ops
  if (VT.isVector()) {
    SDValue FoldedVOp = SimplifyVBinOp(N);
    if (FoldedVOp.getNode()) return FoldedVOp;
  }

  // fold (mul x, undef) -> 0
  if (N0.getOpcode() == ISD::UNDEF || N1.getOpcode() == ISD::UNDEF)
    return DAG.getConstant(0, VT);
  // fold (mul c1, c2) -> c1*c2
  if (N0C && N1C)
    return DAG.FoldConstantArithmetic(ISD::MUL, VT, N0C, N1C);
  // canonicalize constant to RHS
  if (N0C && !N1C)
    return DAG.getNode(ISD::MUL, N->getDebugLoc(), VT, N1, N0);
  // fold (mul x, 0) -> 0
  if (N1C && N1C->isNullValue())
    return N1;
  // fold (mul x, -1) -> 0-x
  if (N1C && N1C->isAllOnesValue())
    return DAG.getNode(ISD::SUB, N->getDebugLoc(), VT,
                       DAG.getConstant(0, VT), N0);
  // fold (mul x, (1 << c)) -> x << c
  if (N1C && N1C->getAPIntValue().isPowerOf2())
    return DAG.getNode(ISD::SHL, N->getDebugLoc(), VT, N0,
                       DAG.getConstant(N1C->getAPIntValue().logBase2(),
                                       getShiftAmountTy(N0.getValueType())));
  // fold (mul x, -(1 << c)) -> -(x << c) or (-x) << c
  if (N1C && (-N1C->getAPIntValue()).isPowerOf2()) {
    unsigned Log2Val = (-N1C->getAPIntValue()).logBase2();
    // FIXME: If the input is something that is easily negated (e.g. a
    // single-use add), we should put the negate there.
    return DAG.getNode(ISD::SUB, N->getDebugLoc(), VT,
                       DAG.getConstant(0, VT),
                       DAG.getNode(ISD::SHL, N->getDebugLoc(), VT, N0,
                            DAG.getConstant(Log2Val,
                                      getShiftAmountTy(N0.getValueType()))));
  }
  // (mul (shl X, c1), c2) -> (mul X, c2 << c1)
  if (N1C && N0.getOpcode() == ISD::SHL &&
      isa<ConstantSDNode>(N0.getOperand(1))) {
    SDValue C3 = DAG.getNode(ISD::SHL, N->getDebugLoc(), VT,
                             N1, N0.getOperand(1));
    AddToWorkList(C3.getNode());
    return DAG.getNode(ISD::MUL, N->getDebugLoc(), VT,
                       N0.getOperand(0), C3);
  }

  // Change (mul (shl X, C), Y) -> (shl (mul X, Y), C) when the shift has one
  // use.
  {
    SDValue Sh(0,0), Y(0,0);
    // Check for both (mul (shl X, C), Y)  and  (mul Y, (shl X, C)).
    if (N0.getOpcode() == ISD::SHL && isa<ConstantSDNode>(N0.getOperand(1)) &&
        N0.getNode()->hasOneUse()) {
      Sh = N0; Y = N1;
    } else if (N1.getOpcode() == ISD::SHL &&
               isa<ConstantSDNode>(N1.getOperand(1)) &&
               N1.getNode()->hasOneUse()) {
      Sh = N1; Y = N0;
    }

    if (Sh.getNode()) {
      SDValue Mul = DAG.getNode(ISD::MUL, N->getDebugLoc(), VT,
                                Sh.getOperand(0), Y);
      return DAG.getNode(ISD::SHL, N->getDebugLoc(), VT,
                         Mul, Sh.getOperand(1));
    }
  }

  // fold (mul (add x, c1), c2) -> (add (mul x, c2), c1*c2)
  if (N1C && N0.getOpcode() == ISD::ADD && N0.getNode()->hasOneUse() &&
      isa<ConstantSDNode>(N0.getOperand(1)))
    return DAG.getNode(ISD::ADD, N->getDebugLoc(), VT,
                       DAG.getNode(ISD::MUL, N0.getDebugLoc(), VT,
                                   N0.getOperand(0), N1),
                       DAG.getNode(ISD::MUL, N1.getDebugLoc(), VT,
                                   N0.getOperand(1), N1));

  // reassociate mul
  SDValue RMUL = ReassociateOps(ISD::MUL, N->getDebugLoc(), N0, N1);
  if (RMUL.getNode() != 0)
    return RMUL;

  return SDValue();
}

SDValue DAGCombiner::visitSDIV(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  SDValue N1 = N->getOperand(1);
  ConstantSDNode *N0C = dyn_cast<ConstantSDNode>(N0.getNode());
  ConstantSDNode *N1C = dyn_cast<ConstantSDNode>(N1.getNode());
  EVT VT = N->getValueType(0);

  // fold vector ops
  if (VT.isVector()) {
    SDValue FoldedVOp = SimplifyVBinOp(N);
    if (FoldedVOp.getNode()) return FoldedVOp;
  }

  // fold (sdiv c1, c2) -> c1/c2
  if (N0C && N1C && !N1C->isNullValue())
    return DAG.FoldConstantArithmetic(ISD::SDIV, VT, N0C, N1C);
  // fold (sdiv X, 1) -> X
  if (N1C && N1C->getAPIntValue() == 1LL)
    return N0;
  // fold (sdiv X, -1) -> 0-X
  if (N1C && N1C->isAllOnesValue())
    return DAG.getNode(ISD::SUB, N->getDebugLoc(), VT,
                       DAG.getConstant(0, VT), N0);
  // If we know the sign bits of both operands are zero, strength reduce to a
  // udiv instead.  Handles (X&15) /s 4 -> X&15 >> 2
  if (!VT.isVector()) {
    if (DAG.SignBitIsZero(N1) && DAG.SignBitIsZero(N0))
      return DAG.getNode(ISD::UDIV, N->getDebugLoc(), N1.getValueType(),
                         N0, N1);
  }
  // fold (sdiv X, pow2) -> simple ops after legalize
  if (N1C && !N1C->isNullValue() &&
      (N1C->getAPIntValue().isPowerOf2() ||
       (-N1C->getAPIntValue()).isPowerOf2())) {
    // If dividing by powers of two is cheap, then don't perform the following
    // fold.
    if (TLI.isPow2DivCheap())
      return SDValue();

    unsigned lg2 = N1C->getAPIntValue().countTrailingZeros();

    // Splat the sign bit into the register
    SDValue SGN = DAG.getNode(ISD::SRA, N->getDebugLoc(), VT, N0,
                              DAG.getConstant(VT.getSizeInBits()-1,
                                       getShiftAmountTy(N0.getValueType())));
    AddToWorkList(SGN.getNode());

    // Add (N0 < 0) ? abs2 - 1 : 0;
    SDValue SRL = DAG.getNode(ISD::SRL, N->getDebugLoc(), VT, SGN,
                              DAG.getConstant(VT.getSizeInBits() - lg2,
                                       getShiftAmountTy(SGN.getValueType())));
    SDValue ADD = DAG.getNode(ISD::ADD, N->getDebugLoc(), VT, N0, SRL);
    AddToWorkList(SRL.getNode());
    AddToWorkList(ADD.getNode());    // Divide by pow2
    SDValue SRA = DAG.getNode(ISD::SRA, N->getDebugLoc(), VT, ADD,
                  DAG.getConstant(lg2, getShiftAmountTy(ADD.getValueType())));

    // If we're dividing by a positive value, we're done.  Otherwise, we must
    // negate the result.
    if (N1C->getAPIntValue().isNonNegative())
      return SRA;

    AddToWorkList(SRA.getNode());
    return DAG.getNode(ISD::SUB, N->getDebugLoc(), VT,
                       DAG.getConstant(0, VT), SRA);
  }

  // if integer divide is expensive and we satisfy the requirements, emit an
  // alternate sequence.
  if (N1C && !N1C->isNullValue() && !TLI.isIntDivCheap()) {
    SDValue Op = BuildSDIV(N);
    if (Op.getNode()) return Op;
  }

  // undef / X -> 0
  if (N0.getOpcode() == ISD::UNDEF)
    return DAG.getConstant(0, VT);
  // X / undef -> undef
  if (N1.getOpcode() == ISD::UNDEF)
    return N1;

  return SDValue();
}

SDValue DAGCombiner::visitUDIV(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  SDValue N1 = N->getOperand(1);
  ConstantSDNode *N0C = dyn_cast<ConstantSDNode>(N0.getNode());
  ConstantSDNode *N1C = dyn_cast<ConstantSDNode>(N1.getNode());
  EVT VT = N->getValueType(0);

  // fold vector ops
  if (VT.isVector()) {
    SDValue FoldedVOp = SimplifyVBinOp(N);
    if (FoldedVOp.getNode()) return FoldedVOp;
  }

  // fold (udiv c1, c2) -> c1/c2
  if (N0C && N1C && !N1C->isNullValue())
    return DAG.FoldConstantArithmetic(ISD::UDIV, VT, N0C, N1C);
  // fold (udiv x, (1 << c)) -> x >>u c
  if (N1C && N1C->getAPIntValue().isPowerOf2())
    return DAG.getNode(ISD::SRL, N->getDebugLoc(), VT, N0,
                       DAG.getConstant(N1C->getAPIntValue().logBase2(),
                                       getShiftAmountTy(N0.getValueType())));
  // fold (udiv x, (shl c, y)) -> x >>u (log2(c)+y) iff c is power of 2
  if (N1.getOpcode() == ISD::SHL) {
    if (ConstantSDNode *SHC = dyn_cast<ConstantSDNode>(N1.getOperand(0))) {
      if (SHC->getAPIntValue().isPowerOf2()) {
        EVT ADDVT = N1.getOperand(1).getValueType();
        SDValue Add = DAG.getNode(ISD::ADD, N->getDebugLoc(), ADDVT,
                                  N1.getOperand(1),
                                  DAG.getConstant(SHC->getAPIntValue()
                                                                  .logBase2(),
                                                  ADDVT));
        AddToWorkList(Add.getNode());
        return DAG.getNode(ISD::SRL, N->getDebugLoc(), VT, N0, Add);
      }
    }
  }
  // fold (udiv x, c) -> alternate
  if (N1C && !N1C->isNullValue() && !TLI.isIntDivCheap()) {
    SDValue Op = BuildUDIV(N);
    if (Op.getNode()) return Op;
  }

  // undef / X -> 0
  if (N0.getOpcode() == ISD::UNDEF)
    return DAG.getConstant(0, VT);
  // X / undef -> undef
  if (N1.getOpcode() == ISD::UNDEF)
    return N1;

  return SDValue();
}

SDValue DAGCombiner::visitSREM(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  SDValue N1 = N->getOperand(1);
  ConstantSDNode *N0C = dyn_cast<ConstantSDNode>(N0);
  ConstantSDNode *N1C = dyn_cast<ConstantSDNode>(N1);
  EVT VT = N->getValueType(0);

  // fold (srem c1, c2) -> c1%c2
  if (N0C && N1C && !N1C->isNullValue())
    return DAG.FoldConstantArithmetic(ISD::SREM, VT, N0C, N1C);
  // If we know the sign bits of both operands are zero, strength reduce to a
  // urem instead.  Handles (X & 0x0FFFFFFF) %s 16 -> X&15
  if (!VT.isVector()) {
    if (DAG.SignBitIsZero(N1) && DAG.SignBitIsZero(N0))
      return DAG.getNode(ISD::UREM, N->getDebugLoc(), VT, N0, N1);
  }

  // If X/C can be simplified by the division-by-constant logic, lower
  // X%C to the equivalent of X-X/C*C.
  if (N1C && !N1C->isNullValue()) {
    SDValue Div = DAG.getNode(ISD::SDIV, N->getDebugLoc(), VT, N0, N1);
    AddToWorkList(Div.getNode());
    SDValue OptimizedDiv = combine(Div.getNode());
    if (OptimizedDiv.getNode() && OptimizedDiv.getNode() != Div.getNode()) {
      SDValue Mul = DAG.getNode(ISD::MUL, N->getDebugLoc(), VT,
                                OptimizedDiv, N1);
      SDValue Sub = DAG.getNode(ISD::SUB, N->getDebugLoc(), VT, N0, Mul);
      AddToWorkList(Mul.getNode());
      return Sub;
    }
  }

  // undef % X -> 0
  if (N0.getOpcode() == ISD::UNDEF)
    return DAG.getConstant(0, VT);
  // X % undef -> undef
  if (N1.getOpcode() == ISD::UNDEF)
    return N1;

  return SDValue();
}

SDValue DAGCombiner::visitUREM(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  SDValue N1 = N->getOperand(1);
  ConstantSDNode *N0C = dyn_cast<ConstantSDNode>(N0);
  ConstantSDNode *N1C = dyn_cast<ConstantSDNode>(N1);
  EVT VT = N->getValueType(0);

  // fold (urem c1, c2) -> c1%c2
  if (N0C && N1C && !N1C->isNullValue())
    return DAG.FoldConstantArithmetic(ISD::UREM, VT, N0C, N1C);
  // fold (urem x, pow2) -> (and x, pow2-1)
  if (N1C && !N1C->isNullValue() && N1C->getAPIntValue().isPowerOf2())
    return DAG.getNode(ISD::AND, N->getDebugLoc(), VT, N0,
                       DAG.getConstant(N1C->getAPIntValue()-1,VT));
  // fold (urem x, (shl pow2, y)) -> (and x, (add (shl pow2, y), -1))
  if (N1.getOpcode() == ISD::SHL) {
    if (ConstantSDNode *SHC = dyn_cast<ConstantSDNode>(N1.getOperand(0))) {
      if (SHC->getAPIntValue().isPowerOf2()) {
        SDValue Add =
          DAG.getNode(ISD::ADD, N->getDebugLoc(), VT, N1,
                 DAG.getConstant(APInt::getAllOnesValue(VT.getSizeInBits()),
                                 VT));
        AddToWorkList(Add.getNode());
        return DAG.getNode(ISD::AND, N->getDebugLoc(), VT, N0, Add);
      }
    }
  }

  // If X/C can be simplified by the division-by-constant logic, lower
  // X%C to the equivalent of X-X/C*C.
  if (N1C && !N1C->isNullValue()) {
    SDValue Div = DAG.getNode(ISD::UDIV, N->getDebugLoc(), VT, N0, N1);
    AddToWorkList(Div.getNode());
    SDValue OptimizedDiv = combine(Div.getNode());
    if (OptimizedDiv.getNode() && OptimizedDiv.getNode() != Div.getNode()) {
      SDValue Mul = DAG.getNode(ISD::MUL, N->getDebugLoc(), VT,
                                OptimizedDiv, N1);
      SDValue Sub = DAG.getNode(ISD::SUB, N->getDebugLoc(), VT, N0, Mul);
      AddToWorkList(Mul.getNode());
      return Sub;
    }
  }

  // undef % X -> 0
  if (N0.getOpcode() == ISD::UNDEF)
    return DAG.getConstant(0, VT);
  // X % undef -> undef
  if (N1.getOpcode() == ISD::UNDEF)
    return N1;

  return SDValue();
}

SDValue DAGCombiner::visitMULHS(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  SDValue N1 = N->getOperand(1);
  ConstantSDNode *N1C = dyn_cast<ConstantSDNode>(N1);
  EVT VT = N->getValueType(0);
  DebugLoc DL = N->getDebugLoc();

  // fold (mulhs x, 0) -> 0
  if (N1C && N1C->isNullValue())
    return N1;
  // fold (mulhs x, 1) -> (sra x, size(x)-1)
  if (N1C && N1C->getAPIntValue() == 1)
    return DAG.getNode(ISD::SRA, N->getDebugLoc(), N0.getValueType(), N0,
                       DAG.getConstant(N0.getValueType().getSizeInBits() - 1,
                                       getShiftAmountTy(N0.getValueType())));
  // fold (mulhs x, undef) -> 0
  if (N0.getOpcode() == ISD::UNDEF || N1.getOpcode() == ISD::UNDEF)
    return DAG.getConstant(0, VT);

  // If the type twice as wide is legal, transform the mulhs to a wider multiply
  // plus a shift.
  if (VT.isSimple() && !VT.isVector()) {
    MVT Simple = VT.getSimpleVT();
    unsigned SimpleSize = Simple.getSizeInBits();
    EVT NewVT = EVT::getIntegerVT(*DAG.getContext(), SimpleSize*2);
    if (TLI.isOperationLegal(ISD::MUL, NewVT)) {
      N0 = DAG.getNode(ISD::SIGN_EXTEND, DL, NewVT, N0);
      N1 = DAG.getNode(ISD::SIGN_EXTEND, DL, NewVT, N1);
      N1 = DAG.getNode(ISD::MUL, DL, NewVT, N0, N1);
      N1 = DAG.getNode(ISD::SRL, DL, NewVT, N1,
            DAG.getConstant(SimpleSize, getShiftAmountTy(N1.getValueType())));
      return DAG.getNode(ISD::TRUNCATE, DL, VT, N1);
    }
  }

  return SDValue();
}

SDValue DAGCombiner::visitMULHU(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  SDValue N1 = N->getOperand(1);
  ConstantSDNode *N1C = dyn_cast<ConstantSDNode>(N1);
  EVT VT = N->getValueType(0);
  DebugLoc DL = N->getDebugLoc();

  // fold (mulhu x, 0) -> 0
  if (N1C && N1C->isNullValue())
    return N1;
  // fold (mulhu x, 1) -> 0
  if (N1C && N1C->getAPIntValue() == 1)
    return DAG.getConstant(0, N0.getValueType());
  // fold (mulhu x, undef) -> 0
  if (N0.getOpcode() == ISD::UNDEF || N1.getOpcode() == ISD::UNDEF)
    return DAG.getConstant(0, VT);

  // If the type twice as wide is legal, transform the mulhu to a wider multiply
  // plus a shift.
  if (VT.isSimple() && !VT.isVector()) {
    MVT Simple = VT.getSimpleVT();
    unsigned SimpleSize = Simple.getSizeInBits();
    EVT NewVT = EVT::getIntegerVT(*DAG.getContext(), SimpleSize*2);
    if (TLI.isOperationLegal(ISD::MUL, NewVT)) {
      N0 = DAG.getNode(ISD::ZERO_EXTEND, DL, NewVT, N0);
      N1 = DAG.getNode(ISD::ZERO_EXTEND, DL, NewVT, N1);
      N1 = DAG.getNode(ISD::MUL, DL, NewVT, N0, N1);
      N1 = DAG.getNode(ISD::SRL, DL, NewVT, N1,
            DAG.getConstant(SimpleSize, getShiftAmountTy(N1.getValueType())));
      return DAG.getNode(ISD::TRUNCATE, DL, VT, N1);
    }
  }

  return SDValue();
}

/// SimplifyNodeWithTwoResults - Perform optimizations common to nodes that
/// compute two values. LoOp and HiOp give the opcodes for the two computations
/// that are being performed. Return true if a simplification was made.
///
SDValue DAGCombiner::SimplifyNodeWithTwoResults(SDNode *N, unsigned LoOp,
                                                unsigned HiOp) {
  // If the high half is not needed, just compute the low half.
  bool HiExists = N->hasAnyUseOfValue(1);
  if (!HiExists &&
      (!LegalOperations ||
       TLI.isOperationLegal(LoOp, N->getValueType(0)))) {
    SDValue Res = DAG.getNode(LoOp, N->getDebugLoc(), N->getValueType(0),
                              N->op_begin(), N->getNumOperands());
    return CombineTo(N, Res, Res);
  }

  // If the low half is not needed, just compute the high half.
  bool LoExists = N->hasAnyUseOfValue(0);
  if (!LoExists &&
      (!LegalOperations ||
       TLI.isOperationLegal(HiOp, N->getValueType(1)))) {
    SDValue Res = DAG.getNode(HiOp, N->getDebugLoc(), N->getValueType(1),
                              N->op_begin(), N->getNumOperands());
    return CombineTo(N, Res, Res);
  }

  // If both halves are used, return as it is.
  if (LoExists && HiExists)
    return SDValue();

  // If the two computed results can be simplified separately, separate them.
  if (LoExists) {
    SDValue Lo = DAG.getNode(LoOp, N->getDebugLoc(), N->getValueType(0),
                             N->op_begin(), N->getNumOperands());
    AddToWorkList(Lo.getNode());
    SDValue LoOpt = combine(Lo.getNode());
    if (LoOpt.getNode() && LoOpt.getNode() != Lo.getNode() &&
        (!LegalOperations ||
         TLI.isOperationLegal(LoOpt.getOpcode(), LoOpt.getValueType())))
      return CombineTo(N, LoOpt, LoOpt);
  }

  if (HiExists) {
    SDValue Hi = DAG.getNode(HiOp, N->getDebugLoc(), N->getValueType(1),
                             N->op_begin(), N->getNumOperands());
    AddToWorkList(Hi.getNode());
    SDValue HiOpt = combine(Hi.getNode());
    if (HiOpt.getNode() && HiOpt != Hi &&
        (!LegalOperations ||
         TLI.isOperationLegal(HiOpt.getOpcode(), HiOpt.getValueType())))
      return CombineTo(N, HiOpt, HiOpt);
  }

  return SDValue();
}

SDValue DAGCombiner::visitSMUL_LOHI(SDNode *N) {
  SDValue Res = SimplifyNodeWithTwoResults(N, ISD::MUL, ISD::MULHS);
  if (Res.getNode()) return Res;

  EVT VT = N->getValueType(0);
  DebugLoc DL = N->getDebugLoc();

  // If the type twice as wide is legal, transform the mulhu to a wider multiply
  // plus a shift.
  if (VT.isSimple() && !VT.isVector()) {
    MVT Simple = VT.getSimpleVT();
    unsigned SimpleSize = Simple.getSizeInBits();
    EVT NewVT = EVT::getIntegerVT(*DAG.getContext(), SimpleSize*2);
    if (TLI.isOperationLegal(ISD::MUL, NewVT)) {
      SDValue Lo = DAG.getNode(ISD::SIGN_EXTEND, DL, NewVT, N->getOperand(0));
      SDValue Hi = DAG.getNode(ISD::SIGN_EXTEND, DL, NewVT, N->getOperand(1));
      Lo = DAG.getNode(ISD::MUL, DL, NewVT, Lo, Hi);
      // Compute the high part as N1.
      Hi = DAG.getNode(ISD::SRL, DL, NewVT, Lo,
            DAG.getConstant(SimpleSize, getShiftAmountTy(Lo.getValueType())));
      Hi = DAG.getNode(ISD::TRUNCATE, DL, VT, Hi);
      // Compute the low part as N0.
      Lo = DAG.getNode(ISD::TRUNCATE, DL, VT, Lo);
      return CombineTo(N, Lo, Hi);
    }
  }

  return SDValue();
}

SDValue DAGCombiner::visitUMUL_LOHI(SDNode *N) {
  SDValue Res = SimplifyNodeWithTwoResults(N, ISD::MUL, ISD::MULHU);
  if (Res.getNode()) return Res;

  EVT VT = N->getValueType(0);
  DebugLoc DL = N->getDebugLoc();

  // If the type twice as wide is legal, transform the mulhu to a wider multiply
  // plus a shift.
  if (VT.isSimple() && !VT.isVector()) {
    MVT Simple = VT.getSimpleVT();
    unsigned SimpleSize = Simple.getSizeInBits();
    EVT NewVT = EVT::getIntegerVT(*DAG.getContext(), SimpleSize*2);
    if (TLI.isOperationLegal(ISD::MUL, NewVT)) {
      SDValue Lo = DAG.getNode(ISD::ZERO_EXTEND, DL, NewVT, N->getOperand(0));
      SDValue Hi = DAG.getNode(ISD::ZERO_EXTEND, DL, NewVT, N->getOperand(1));
      Lo = DAG.getNode(ISD::MUL, DL, NewVT, Lo, Hi);
      // Compute the high part as N1.
      Hi = DAG.getNode(ISD::SRL, DL, NewVT, Lo,
            DAG.getConstant(SimpleSize, getShiftAmountTy(Lo.getValueType())));
      Hi = DAG.getNode(ISD::TRUNCATE, DL, VT, Hi);
      // Compute the low part as N0.
      Lo = DAG.getNode(ISD::TRUNCATE, DL, VT, Lo);
      return CombineTo(N, Lo, Hi);
    }
  }

  return SDValue();
}

SDValue DAGCombiner::visitSMULO(SDNode *N) {
  // (smulo x, 2) -> (saddo x, x)
  if (ConstantSDNode *C2 = dyn_cast<ConstantSDNode>(N->getOperand(1)))
    if (C2->getAPIntValue() == 2)
      return DAG.getNode(ISD::SADDO, N->getDebugLoc(), N->getVTList(),
                         N->getOperand(0), N->getOperand(0));

  return SDValue();
}

SDValue DAGCombiner::visitUMULO(SDNode *N) {
  // (umulo x, 2) -> (uaddo x, x)
  if (ConstantSDNode *C2 = dyn_cast<ConstantSDNode>(N->getOperand(1)))
    if (C2->getAPIntValue() == 2)
      return DAG.getNode(ISD::UADDO, N->getDebugLoc(), N->getVTList(),
                         N->getOperand(0), N->getOperand(0));

  return SDValue();
}

SDValue DAGCombiner::visitSDIVREM(SDNode *N) {
  SDValue Res = SimplifyNodeWithTwoResults(N, ISD::SDIV, ISD::SREM);
  if (Res.getNode()) return Res;

  return SDValue();
}

SDValue DAGCombiner::visitUDIVREM(SDNode *N) {
  SDValue Res = SimplifyNodeWithTwoResults(N, ISD::UDIV, ISD::UREM);
  if (Res.getNode()) return Res;

  return SDValue();
}

/// SimplifyBinOpWithSameOpcodeHands - If this is a binary operator with
/// two operands of the same opcode, try to simplify it.
SDValue DAGCombiner::SimplifyBinOpWithSameOpcodeHands(SDNode *N) {
  SDValue N0 = N->getOperand(0), N1 = N->getOperand(1);
  EVT VT = N0.getValueType();
  assert(N0.getOpcode() == N1.getOpcode() && "Bad input!");

  // Bail early if none of these transforms apply.
  if (N0.getNode()->getNumOperands() == 0) return SDValue();

  // For each of OP in AND/OR/XOR:
  // fold (OP (zext x), (zext y)) -> (zext (OP x, y))
  // fold (OP (sext x), (sext y)) -> (sext (OP x, y))
  // fold (OP (aext x), (aext y)) -> (aext (OP x, y))
  // fold (OP (trunc x), (trunc y)) -> (trunc (OP x, y)) (if trunc isn't free)
  //
  // do not sink logical op inside of a vector extend, since it may combine
  // into a vsetcc.
  EVT Op0VT = N0.getOperand(0).getValueType();
  if ((N0.getOpcode() == ISD::ZERO_EXTEND ||
       N0.getOpcode() == ISD::SIGN_EXTEND ||
       // Avoid infinite looping with PromoteIntBinOp.
       (N0.getOpcode() == ISD::ANY_EXTEND &&
        (!LegalTypes || TLI.isTypeDesirableForOp(N->getOpcode(), Op0VT))) ||
       (N0.getOpcode() == ISD::TRUNCATE &&
        (!TLI.isZExtFree(VT, Op0VT) ||
         !TLI.isTruncateFree(Op0VT, VT)) &&
        TLI.isTypeLegal(Op0VT))) &&
      !VT.isVector() &&
      Op0VT == N1.getOperand(0).getValueType() &&
      (!LegalOperations || TLI.isOperationLegal(N->getOpcode(), Op0VT))) {
    SDValue ORNode = DAG.getNode(N->getOpcode(), N0.getDebugLoc(),
                                 N0.getOperand(0).getValueType(),
                                 N0.getOperand(0), N1.getOperand(0));
    AddToWorkList(ORNode.getNode());
    return DAG.getNode(N0.getOpcode(), N->getDebugLoc(), VT, ORNode);
  }

  // For each of OP in SHL/SRL/SRA/AND...
  //   fold (and (OP x, z), (OP y, z)) -> (OP (and x, y), z)
  //   fold (or  (OP x, z), (OP y, z)) -> (OP (or  x, y), z)
  //   fold (xor (OP x, z), (OP y, z)) -> (OP (xor x, y), z)
  if ((N0.getOpcode() == ISD::SHL || N0.getOpcode() == ISD::SRL ||
       N0.getOpcode() == ISD::SRA || N0.getOpcode() == ISD::AND) &&
      N0.getOperand(1) == N1.getOperand(1)) {
    SDValue ORNode = DAG.getNode(N->getOpcode(), N0.getDebugLoc(),
                                 N0.getOperand(0).getValueType(),
                                 N0.getOperand(0), N1.getOperand(0));
    AddToWorkList(ORNode.getNode());
    return DAG.getNode(N0.getOpcode(), N->getDebugLoc(), VT,
                       ORNode, N0.getOperand(1));
  }

  // Simplify xor/and/or (bitcast(A), bitcast(B)) -> bitcast(op (A,B))
  // Only perform this optimization after type legalization and before
  // LegalizeVectorOprs. LegalizeVectorOprs promotes vector operations by
  // adding bitcasts. For example (xor v4i32) is promoted to (v2i64), and
  // we don't want to undo this promotion.
  // We also handle SCALAR_TO_VECTOR because xor/or/and operations are cheaper
  // on scalars.
  if ((N0.getOpcode() == ISD::BITCAST ||
       N0.getOpcode() == ISD::SCALAR_TO_VECTOR) &&
      Level == AfterLegalizeTypes) {
    SDValue In0 = N0.getOperand(0);
    SDValue In1 = N1.getOperand(0);
    EVT In0Ty = In0.getValueType();
    EVT In1Ty = In1.getValueType();
    DebugLoc DL = N->getDebugLoc();
    // If both incoming values are integers, and the original types are the
    // same.
    if (In0Ty.isInteger() && In1Ty.isInteger() && In0Ty == In1Ty) {
      SDValue Op = DAG.getNode(N->getOpcode(), DL, In0Ty, In0, In1);
      SDValue BC = DAG.getNode(N0.getOpcode(), DL, VT, Op);
      AddToWorkList(Op.getNode());
      return BC;
    }
  }

  // Xor/and/or are indifferent to the swizzle operation (shuffle of one value).
  // Simplify xor/and/or (shuff(A), shuff(B)) -> shuff(op (A,B))
  // If both shuffles use the same mask, and both shuffle within a single
  // vector, then it is worthwhile to move the swizzle after the operation.
  // The type-legalizer generates this pattern when loading illegal
  // vector types from memory. In many cases this allows additional shuffle
  // optimizations.
  if (N0.getOpcode() == ISD::VECTOR_SHUFFLE && Level < AfterLegalizeDAG &&
      N0.getOperand(1).getOpcode() == ISD::UNDEF &&
      N1.getOperand(1).getOpcode() == ISD::UNDEF) {
    ShuffleVectorSDNode *SVN0 = cast<ShuffleVectorSDNode>(N0);
    ShuffleVectorSDNode *SVN1 = cast<ShuffleVectorSDNode>(N1);

    assert(N0.getOperand(0).getValueType() == N1.getOperand(1).getValueType() &&
           "Inputs to shuffles are not the same type");

    unsigned NumElts = VT.getVectorNumElements();

    // Check that both shuffles use the same mask. The masks are known to be of
    // the same length because the result vector type is the same.
    bool SameMask = true;
    for (unsigned i = 0; i != NumElts; ++i) {
      int Idx0 = SVN0->getMaskElt(i);
      int Idx1 = SVN1->getMaskElt(i);
      if (Idx0 != Idx1) {
        SameMask = false;
        break;
      }
    }

    if (SameMask) {
      SDValue Op = DAG.getNode(N->getOpcode(), N->getDebugLoc(), VT,
                               N0.getOperand(0), N1.getOperand(0));
      AddToWorkList(Op.getNode());
      return DAG.getVectorShuffle(VT, N->getDebugLoc(), Op,
                                  DAG.getUNDEF(VT), &SVN0->getMask()[0]);
    }
  }

  return SDValue();
}

SDValue DAGCombiner::visitAND(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  SDValue N1 = N->getOperand(1);
  SDValue LL, LR, RL, RR, CC0, CC1;
  ConstantSDNode *N0C = dyn_cast<ConstantSDNode>(N0);
  ConstantSDNode *N1C = dyn_cast<ConstantSDNode>(N1);
  EVT VT = N1.getValueType();
  unsigned BitWidth = VT.getScalarType().getSizeInBits();

  // fold vector ops
  if (VT.isVector()) {
    SDValue FoldedVOp = SimplifyVBinOp(N);
    if (FoldedVOp.getNode()) return FoldedVOp;
  }

  // fold (and x, undef) -> 0
  if (N0.getOpcode() == ISD::UNDEF || N1.getOpcode() == ISD::UNDEF)
    return DAG.getConstant(0, VT);
  // fold (and c1, c2) -> c1&c2
  if (N0C && N1C)
    return DAG.FoldConstantArithmetic(ISD::AND, VT, N0C, N1C);
  // canonicalize constant to RHS
  if (N0C && !N1C)
    return DAG.getNode(ISD::AND, N->getDebugLoc(), VT, N1, N0);
  // fold (and x, -1) -> x
  if (N1C && N1C->isAllOnesValue())
    return N0;
  // if (and x, c) is known to be zero, return 0
  if (N1C && DAG.MaskedValueIsZero(SDValue(N, 0),
                                   APInt::getAllOnesValue(BitWidth)))
    return DAG.getConstant(0, VT);
  // reassociate and
  SDValue RAND = ReassociateOps(ISD::AND, N->getDebugLoc(), N0, N1);
  if (RAND.getNode() != 0)
    return RAND;
  // fold (and (or x, C), D) -> D if (C & D) == D
  if (N1C && N0.getOpcode() == ISD::OR)
    if (ConstantSDNode *ORI = dyn_cast<ConstantSDNode>(N0.getOperand(1)))
      if ((ORI->getAPIntValue() & N1C->getAPIntValue()) == N1C->getAPIntValue())
        return N1;
  // fold (and (any_ext V), c) -> (zero_ext V) if 'and' only clears top bits.
  if (N1C && N0.getOpcode() == ISD::ANY_EXTEND) {
    SDValue N0Op0 = N0.getOperand(0);
    APInt Mask = ~N1C->getAPIntValue();
    Mask = Mask.trunc(N0Op0.getValueSizeInBits());
    if (DAG.MaskedValueIsZero(N0Op0, Mask)) {
      SDValue Zext = DAG.getNode(ISD::ZERO_EXTEND, N->getDebugLoc(),
                                 N0.getValueType(), N0Op0);

      // Replace uses of the AND with uses of the Zero extend node.
      CombineTo(N, Zext);

      // We actually want to replace all uses of the any_extend with the
      // zero_extend, to avoid duplicating things.  This will later cause this
      // AND to be folded.
      CombineTo(N0.getNode(), Zext);
      return SDValue(N, 0);   // Return N so it doesn't get rechecked!
    }
  }
  // similarly fold (and (X (load ([non_ext|any_ext|zero_ext] V))), c) -> 
  // (X (load ([non_ext|zero_ext] V))) if 'and' only clears top bits which must
  // already be zero by virtue of the width of the base type of the load.
  //
  // the 'X' node here can either be nothing or an extract_vector_elt to catch
  // more cases.
  if ((N0.getOpcode() == ISD::EXTRACT_VECTOR_ELT &&
       N0.getOperand(0).getOpcode() == ISD::LOAD) ||
      N0.getOpcode() == ISD::LOAD) {
    LoadSDNode *Load = cast<LoadSDNode>( (N0.getOpcode() == ISD::LOAD) ?
                                         N0 : N0.getOperand(0) );

    // Get the constant (if applicable) the zero'th operand is being ANDed with.
    // This can be a pure constant or a vector splat, in which case we treat the
    // vector as a scalar and use the splat value.
    APInt Constant = APInt::getNullValue(1);
    if (const ConstantSDNode *C = dyn_cast<ConstantSDNode>(N1)) {
      Constant = C->getAPIntValue();
    } else if (BuildVectorSDNode *Vector = dyn_cast<BuildVectorSDNode>(N1)) {
      APInt SplatValue, SplatUndef;
      unsigned SplatBitSize;
      bool HasAnyUndefs;
      bool IsSplat = Vector->isConstantSplat(SplatValue, SplatUndef,
                                             SplatBitSize, HasAnyUndefs);
      if (IsSplat) {
        // Undef bits can contribute to a possible optimisation if set, so
        // set them.
        SplatValue |= SplatUndef;

        // The splat value may be something like "0x00FFFFFF", which means 0 for
        // the first vector value and FF for the rest, repeating. We need a mask
        // that will apply equally to all members of the vector, so AND all the
        // lanes of the constant together.
        EVT VT = Vector->getValueType(0);
        unsigned BitWidth = VT.getVectorElementType().getSizeInBits();

        // If the splat value has been compressed to a bitlength lower
        // than the size of the vector lane, we need to re-expand it to
        // the lane size.
        if (BitWidth > SplatBitSize)
          for (SplatValue = SplatValue.zextOrTrunc(BitWidth);
               SplatBitSize < BitWidth;
               SplatBitSize = SplatBitSize * 2)
            SplatValue |= SplatValue.shl(SplatBitSize);

        Constant = APInt::getAllOnesValue(BitWidth);
        for (unsigned i = 0, n = SplatBitSize/BitWidth; i < n; ++i)
          Constant &= SplatValue.lshr(i*BitWidth).zextOrTrunc(BitWidth);
      }
    }

    // If we want to change an EXTLOAD to a ZEXTLOAD, ensure a ZEXTLOAD is
    // actually legal and isn't going to get expanded, else this is a false
    // optimisation.
    bool CanZextLoadProfitably = TLI.isLoadExtLegal(ISD::ZEXTLOAD,
                                                    Load->getMemoryVT());

    // Resize the constant to the same size as the original memory access before
    // extension. If it is still the AllOnesValue then this AND is completely
    // unneeded.
    Constant =
      Constant.zextOrTrunc(Load->getMemoryVT().getScalarType().getSizeInBits());

    bool B;
    switch (Load->getExtensionType()) {
    default: B = false; break;
    case ISD::EXTLOAD: B = CanZextLoadProfitably; break;
    case ISD::ZEXTLOAD:
    case ISD::NON_EXTLOAD: B = true; break;
    }

    if (B && Constant.isAllOnesValue()) {
      // If the load type was an EXTLOAD, convert to ZEXTLOAD in order to
      // preserve semantics once we get rid of the AND.
      SDValue NewLoad(Load, 0);
      if (Load->getExtensionType() == ISD::EXTLOAD) {
        NewLoad = DAG.getLoad(Load->getAddressingMode(), ISD::ZEXTLOAD,
                              Load->getValueType(0), Load->getDebugLoc(),
                              Load->getChain(), Load->getBasePtr(),
                              Load->getOffset(), Load->getMemoryVT(),
                              Load->getMemOperand());
        // Replace uses of the EXTLOAD with the new ZEXTLOAD.
        if (Load->getNumValues() == 3) {
          // PRE/POST_INC loads have 3 values.
          SDValue To[] = { NewLoad.getValue(0), NewLoad.getValue(1),
                           NewLoad.getValue(2) };
          CombineTo(Load, To, 3, true);
        } else {
          CombineTo(Load, NewLoad.getValue(0), NewLoad.getValue(1));
        }
      }

      // Fold the AND away, taking care not to fold to the old load node if we
      // replaced it.
      CombineTo(N, (N0.getNode() == Load) ? NewLoad : N0);

      return SDValue(N, 0); // Return N so it doesn't get rechecked!
    }
  }
  // fold (and (setcc x), (setcc y)) -> (setcc (and x, y))
  if (isSetCCEquivalent(N0, LL, LR, CC0) && isSetCCEquivalent(N1, RL, RR, CC1)){
    ISD::CondCode Op0 = cast<CondCodeSDNode>(CC0)->get();
    ISD::CondCode Op1 = cast<CondCodeSDNode>(CC1)->get();

    if (LR == RR && isa<ConstantSDNode>(LR) && Op0 == Op1 &&
        LL.getValueType().isInteger()) {
      // fold (and (seteq X, 0), (seteq Y, 0)) -> (seteq (or X, Y), 0)
      if (cast<ConstantSDNode>(LR)->isNullValue() && Op1 == ISD::SETEQ) {
        SDValue ORNode = DAG.getNode(ISD::OR, N0.getDebugLoc(),
                                     LR.getValueType(), LL, RL);
        AddToWorkList(ORNode.getNode());
        return DAG.getSetCC(N->getDebugLoc(), VT, ORNode, LR, Op1);
      }
      // fold (and (seteq X, -1), (seteq Y, -1)) -> (seteq (and X, Y), -1)
      if (cast<ConstantSDNode>(LR)->isAllOnesValue() && Op1 == ISD::SETEQ) {
        SDValue ANDNode = DAG.getNode(ISD::AND, N0.getDebugLoc(),
                                      LR.getValueType(), LL, RL);
        AddToWorkList(ANDNode.getNode());
        return DAG.getSetCC(N->getDebugLoc(), VT, ANDNode, LR, Op1);
      }
      // fold (and (setgt X,  -1), (setgt Y,  -1)) -> (setgt (or X, Y), -1)
      if (cast<ConstantSDNode>(LR)->isAllOnesValue() && Op1 == ISD::SETGT) {
        SDValue ORNode = DAG.getNode(ISD::OR, N0.getDebugLoc(),
                                     LR.getValueType(), LL, RL);
        AddToWorkList(ORNode.getNode());
        return DAG.getSetCC(N->getDebugLoc(), VT, ORNode, LR, Op1);
      }
    }
    // canonicalize equivalent to ll == rl
    if (LL == RR && LR == RL) {
      Op1 = ISD::getSetCCSwappedOperands(Op1);
      std::swap(RL, RR);
    }
    if (LL == RL && LR == RR) {
      bool isInteger = LL.getValueType().isInteger();
      ISD::CondCode Result = ISD::getSetCCAndOperation(Op0, Op1, isInteger);
      if (Result != ISD::SETCC_INVALID &&
          (!LegalOperations || TLI.isCondCodeLegal(Result, LL.getValueType())))
        return DAG.getSetCC(N->getDebugLoc(), N0.getValueType(),
                            LL, LR, Result);
    }
  }

  // Simplify: (and (op x...), (op y...))  -> (op (and x, y))
  if (N0.getOpcode() == N1.getOpcode()) {
    SDValue Tmp = SimplifyBinOpWithSameOpcodeHands(N);
    if (Tmp.getNode()) return Tmp;
  }

  // fold (and (sign_extend_inreg x, i16 to i32), 1) -> (and x, 1)
  // fold (and (sra)) -> (and (srl)) when possible.
  if (!VT.isVector() &&
      SimplifyDemandedBits(SDValue(N, 0)))
    return SDValue(N, 0);

  // fold (zext_inreg (extload x)) -> (zextload x)
  if (ISD::isEXTLoad(N0.getNode()) && ISD::isUNINDEXEDLoad(N0.getNode())) {
    LoadSDNode *LN0 = cast<LoadSDNode>(N0);
    EVT MemVT = LN0->getMemoryVT();
    // If we zero all the possible extended bits, then we can turn this into
    // a zextload if we are running before legalize or the operation is legal.
    unsigned BitWidth = N1.getValueType().getScalarType().getSizeInBits();
    if (DAG.MaskedValueIsZero(N1, APInt::getHighBitsSet(BitWidth,
                           BitWidth - MemVT.getScalarType().getSizeInBits())) &&
        ((!LegalOperations && !LN0->isVolatile()) ||
         TLI.isLoadExtLegal(ISD::ZEXTLOAD, MemVT))) {
      SDValue ExtLoad = DAG.getExtLoad(ISD::ZEXTLOAD, N0.getDebugLoc(), VT,
                                       LN0->getChain(), LN0->getBasePtr(),
                                       LN0->getPointerInfo(), MemVT,
                                       LN0->isVolatile(), LN0->isNonTemporal(),
                                       LN0->getAlignment());
      AddToWorkList(N);
      CombineTo(N0.getNode(), ExtLoad, ExtLoad.getValue(1));
      return SDValue(N, 0);   // Return N so it doesn't get rechecked!
    }
  }
  // fold (zext_inreg (sextload x)) -> (zextload x) iff load has one use
  if (ISD::isSEXTLoad(N0.getNode()) && ISD::isUNINDEXEDLoad(N0.getNode()) &&
      N0.hasOneUse()) {
    LoadSDNode *LN0 = cast<LoadSDNode>(N0);
    EVT MemVT = LN0->getMemoryVT();
    // If we zero all the possible extended bits, then we can turn this into
    // a zextload if we are running before legalize or the operation is legal.
    unsigned BitWidth = N1.getValueType().getScalarType().getSizeInBits();
    if (DAG.MaskedValueIsZero(N1, APInt::getHighBitsSet(BitWidth,
                           BitWidth - MemVT.getScalarType().getSizeInBits())) &&
        ((!LegalOperations && !LN0->isVolatile()) ||
         TLI.isLoadExtLegal(ISD::ZEXTLOAD, MemVT))) {
      SDValue ExtLoad = DAG.getExtLoad(ISD::ZEXTLOAD, N0.getDebugLoc(), VT,
                                       LN0->getChain(),
                                       LN0->getBasePtr(), LN0->getPointerInfo(),
                                       MemVT,
                                       LN0->isVolatile(), LN0->isNonTemporal(),
                                       LN0->getAlignment());
      AddToWorkList(N);
      CombineTo(N0.getNode(), ExtLoad, ExtLoad.getValue(1));
      return SDValue(N, 0);   // Return N so it doesn't get rechecked!
    }
  }

  // fold (and (load x), 255) -> (zextload x, i8)
  // fold (and (extload x, i16), 255) -> (zextload x, i8)
  // fold (and (any_ext (extload x, i16)), 255) -> (zextload x, i8)
  if (N1C && (N0.getOpcode() == ISD::LOAD ||
              (N0.getOpcode() == ISD::ANY_EXTEND &&
               N0.getOperand(0).getOpcode() == ISD::LOAD))) {
    bool HasAnyExt = N0.getOpcode() == ISD::ANY_EXTEND;
    LoadSDNode *LN0 = HasAnyExt
      ? cast<LoadSDNode>(N0.getOperand(0))
      : cast<LoadSDNode>(N0);
    if (LN0->getExtensionType() != ISD::SEXTLOAD &&
        LN0->isUnindexed() && N0.hasOneUse() && LN0->hasOneUse()) {
      uint32_t ActiveBits = N1C->getAPIntValue().getActiveBits();
      if (ActiveBits > 0 && APIntOps::isMask(ActiveBits, N1C->getAPIntValue())){
        EVT ExtVT = EVT::getIntegerVT(*DAG.getContext(), ActiveBits);
        EVT LoadedVT = LN0->getMemoryVT();

        if (ExtVT == LoadedVT &&
            (!LegalOperations || TLI.isLoadExtLegal(ISD::ZEXTLOAD, ExtVT))) {
          EVT LoadResultTy = HasAnyExt ? LN0->getValueType(0) : VT;

          SDValue NewLoad =
            DAG.getExtLoad(ISD::ZEXTLOAD, LN0->getDebugLoc(), LoadResultTy,
                           LN0->getChain(), LN0->getBasePtr(),
                           LN0->getPointerInfo(),
                           ExtVT, LN0->isVolatile(), LN0->isNonTemporal(),
                           LN0->getAlignment());
          AddToWorkList(N);
          CombineTo(LN0, NewLoad, NewLoad.getValue(1));
          return SDValue(N, 0);   // Return N so it doesn't get rechecked!
        }

        // Do not change the width of a volatile load.
        // Do not generate loads of non-round integer types since these can
        // be expensive (and would be wrong if the type is not byte sized).
        if (!LN0->isVolatile() && LoadedVT.bitsGT(ExtVT) && ExtVT.isRound() &&
            (!LegalOperations || TLI.isLoadExtLegal(ISD::ZEXTLOAD, ExtVT))) {
          EVT PtrType = LN0->getOperand(1).getValueType();

          unsigned Alignment = LN0->getAlignment();
          SDValue NewPtr = LN0->getBasePtr();

          // For big endian targets, we need to add an offset to the pointer
          // to load the correct bytes.  For little endian systems, we merely
          // need to read fewer bytes from the same pointer.
          if (TLI.isBigEndian()) {
            unsigned LVTStoreBytes = LoadedVT.getStoreSize();
            unsigned EVTStoreBytes = ExtVT.getStoreSize();
            unsigned PtrOff = LVTStoreBytes - EVTStoreBytes;
            NewPtr = DAG.getNode(ISD::ADD, LN0->getDebugLoc(), PtrType,
                                 NewPtr, DAG.getConstant(PtrOff, PtrType));
            Alignment = MinAlign(Alignment, PtrOff);
          }

          AddToWorkList(NewPtr.getNode());

          EVT LoadResultTy = HasAnyExt ? LN0->getValueType(0) : VT;
          SDValue Load =
            DAG.getExtLoad(ISD::ZEXTLOAD, LN0->getDebugLoc(), LoadResultTy,
                           LN0->getChain(), NewPtr,
                           LN0->getPointerInfo(),
                           ExtVT, LN0->isVolatile(), LN0->isNonTemporal(),
                           Alignment);
          AddToWorkList(N);
          CombineTo(LN0, Load, Load.getValue(1));
          return SDValue(N, 0);   // Return N so it doesn't get rechecked!
        }
      }
    }
  }

  if (N0.getOpcode() == ISD::ADD && N1.getOpcode() == ISD::SRL &&
      VT.getSizeInBits() <= 64) {
    if (ConstantSDNode *ADDI = dyn_cast<ConstantSDNode>(N0.getOperand(1))) {
      APInt ADDC = ADDI->getAPIntValue();
      if (!TLI.isLegalAddImmediate(ADDC.getSExtValue())) {
        // Look for (and (add x, c1), (lshr y, c2)). If C1 wasn't a legal
        // immediate for an add, but it is legal if its top c2 bits are set,
        // transform the ADD so the immediate doesn't need to be materialized
        // in a register.
        if (ConstantSDNode *SRLI = dyn_cast<ConstantSDNode>(N1.getOperand(1))) {
          APInt Mask = APInt::getHighBitsSet(VT.getSizeInBits(),
                                             SRLI->getZExtValue());
          if (DAG.MaskedValueIsZero(N0.getOperand(1), Mask)) {
            ADDC |= Mask;
            if (TLI.isLegalAddImmediate(ADDC.getSExtValue())) {
              SDValue NewAdd =
                DAG.getNode(ISD::ADD, N0.getDebugLoc(), VT,
                            N0.getOperand(0), DAG.getConstant(ADDC, VT));
              CombineTo(N0.getNode(), NewAdd);
              return SDValue(N, 0); // Return N so it doesn't get rechecked!
            }
          }
        }
      }
    }
  }
      

  return SDValue();
}

/// MatchBSwapHWord - Match (a >> 8) | (a << 8) as (bswap a) >> 16
///
SDValue DAGCombiner::MatchBSwapHWordLow(SDNode *N, SDValue N0, SDValue N1,
                                        bool DemandHighBits) {
  if (!LegalOperations)
    return SDValue();

  EVT VT = N->getValueType(0);
  if (VT != MVT::i64 && VT != MVT::i32 && VT != MVT::i16)
    return SDValue();
  if (!TLI.isOperationLegal(ISD::BSWAP, VT))
    return SDValue();

  // Recognize (and (shl a, 8), 0xff), (and (srl a, 8), 0xff00)
  bool LookPassAnd0 = false;
  bool LookPassAnd1 = false;
  if (N0.getOpcode() == ISD::AND && N0.getOperand(0).getOpcode() == ISD::SRL)
      std::swap(N0, N1);
  if (N1.getOpcode() == ISD::AND && N1.getOperand(0).getOpcode() == ISD::SHL)
      std::swap(N0, N1);
  if (N0.getOpcode() == ISD::AND) {
    if (!N0.getNode()->hasOneUse())
      return SDValue();
    ConstantSDNode *N01C = dyn_cast<ConstantSDNode>(N0.getOperand(1));
    if (!N01C || N01C->getZExtValue() != 0xFF00)
      return SDValue();
    N0 = N0.getOperand(0);
    LookPassAnd0 = true;
  }

  if (N1.getOpcode() == ISD::AND) {
    if (!N1.getNode()->hasOneUse())
      return SDValue();
    ConstantSDNode *N11C = dyn_cast<ConstantSDNode>(N1.getOperand(1));
    if (!N11C || N11C->getZExtValue() != 0xFF)
      return SDValue();
    N1 = N1.getOperand(0);
    LookPassAnd1 = true;
  }

  if (N0.getOpcode() == ISD::SRL && N1.getOpcode() == ISD::SHL)
    std::swap(N0, N1);
  if (N0.getOpcode() != ISD::SHL || N1.getOpcode() != ISD::SRL)
    return SDValue();
  if (!N0.getNode()->hasOneUse() ||
      !N1.getNode()->hasOneUse())
    return SDValue();

  ConstantSDNode *N01C = dyn_cast<ConstantSDNode>(N0.getOperand(1));
  ConstantSDNode *N11C = dyn_cast<ConstantSDNode>(N1.getOperand(1));
  if (!N01C || !N11C)
    return SDValue();
  if (N01C->getZExtValue() != 8 || N11C->getZExtValue() != 8)
    return SDValue();

  // Look for (shl (and a, 0xff), 8), (srl (and a, 0xff00), 8)
  SDValue N00 = N0->getOperand(0);
  if (!LookPassAnd0 && N00.getOpcode() == ISD::AND) {
    if (!N00.getNode()->hasOneUse())
      return SDValue();
    ConstantSDNode *N001C = dyn_cast<ConstantSDNode>(N00.getOperand(1));
    if (!N001C || N001C->getZExtValue() != 0xFF)
      return SDValue();
    N00 = N00.getOperand(0);
    LookPassAnd0 = true;
  }

  SDValue N10 = N1->getOperand(0);
  if (!LookPassAnd1 && N10.getOpcode() == ISD::AND) {
    if (!N10.getNode()->hasOneUse())
      return SDValue();
    ConstantSDNode *N101C = dyn_cast<ConstantSDNode>(N10.getOperand(1));
    if (!N101C || N101C->getZExtValue() != 0xFF00)
      return SDValue();
    N10 = N10.getOperand(0);
    LookPassAnd1 = true;
  }

  if (N00 != N10)
    return SDValue();

  // Make sure everything beyond the low halfword is zero since the SRL 16
  // will clear the top bits.
  unsigned OpSizeInBits = VT.getSizeInBits();
  if (DemandHighBits && OpSizeInBits > 16 &&
      (!LookPassAnd0 || !LookPassAnd1) &&
      !DAG.MaskedValueIsZero(N10, APInt::getHighBitsSet(OpSizeInBits, 16)))
    return SDValue();

  SDValue Res = DAG.getNode(ISD::BSWAP, N->getDebugLoc(), VT, N00);
  if (OpSizeInBits > 16)
    Res = DAG.getNode(ISD::SRL, N->getDebugLoc(), VT, Res,
                      DAG.getConstant(OpSizeInBits-16, getShiftAmountTy(VT)));
  return Res;
}

/// isBSwapHWordElement - Return true if the specified node is an element
/// that makes up a 32-bit packed halfword byteswap. i.e.
/// ((x&0xff)<<8)|((x&0xff00)>>8)|((x&0x00ff0000)<<8)|((x&0xff000000)>>8)
static bool isBSwapHWordElement(SDValue N, SmallVector<SDNode*,4> &Parts) {
  if (!N.getNode()->hasOneUse())
    return false;

  unsigned Opc = N.getOpcode();
  if (Opc != ISD::AND && Opc != ISD::SHL && Opc != ISD::SRL)
    return false;

  ConstantSDNode *N1C = dyn_cast<ConstantSDNode>(N.getOperand(1));
  if (!N1C)
    return false;

  unsigned Num;
  switch (N1C->getZExtValue()) {
  default:
    return false;
  case 0xFF:       Num = 0; break;
  case 0xFF00:     Num = 1; break;
  case 0xFF0000:   Num = 2; break;
  case 0xFF000000: Num = 3; break;
  }

  // Look for (x & 0xff) << 8 as well as ((x << 8) & 0xff00).
  SDValue N0 = N.getOperand(0);
  if (Opc == ISD::AND) {
    if (Num == 0 || Num == 2) {
      // (x >> 8) & 0xff
      // (x >> 8) & 0xff0000
      if (N0.getOpcode() != ISD::SRL)
        return false;
      ConstantSDNode *C = dyn_cast<ConstantSDNode>(N0.getOperand(1));
      if (!C || C->getZExtValue() != 8)
        return false;
    } else {
      // (x << 8) & 0xff00
      // (x << 8) & 0xff000000
      if (N0.getOpcode() != ISD::SHL)
        return false;
      ConstantSDNode *C = dyn_cast<ConstantSDNode>(N0.getOperand(1));
      if (!C || C->getZExtValue() != 8)
        return false;
    }
  } else if (Opc == ISD::SHL) {
    // (x & 0xff) << 8
    // (x & 0xff0000) << 8
    if (Num != 0 && Num != 2)
      return false;
    ConstantSDNode *C = dyn_cast<ConstantSDNode>(N.getOperand(1));
    if (!C || C->getZExtValue() != 8)
      return false;
  } else { // Opc == ISD::SRL
    // (x & 0xff00) >> 8
    // (x & 0xff000000) >> 8
    if (Num != 1 && Num != 3)
      return false;
    ConstantSDNode *C = dyn_cast<ConstantSDNode>(N.getOperand(1));
    if (!C || C->getZExtValue() != 8)
      return false;
  }

  if (Parts[Num])
    return false;

  Parts[Num] = N0.getOperand(0).getNode();
  return true;
}

/// MatchBSwapHWord - Match a 32-bit packed halfword bswap. That is
/// ((x&0xff)<<8)|((x&0xff00)>>8)|((x&0x00ff0000)<<8)|((x&0xff000000)>>8)
/// => (rotl (bswap x), 16)
SDValue DAGCombiner::MatchBSwapHWord(SDNode *N, SDValue N0, SDValue N1) {
  if (!LegalOperations)
    return SDValue();

  EVT VT = N->getValueType(0);
  if (VT != MVT::i32)
    return SDValue();
  if (!TLI.isOperationLegal(ISD::BSWAP, VT))
    return SDValue();

  SmallVector<SDNode*,4> Parts(4, (SDNode*)0);
  // Look for either
  // (or (or (and), (and)), (or (and), (and)))
  // (or (or (or (and), (and)), (and)), (and))
  if (N0.getOpcode() != ISD::OR)
    return SDValue();
  SDValue N00 = N0.getOperand(0);
  SDValue N01 = N0.getOperand(1);

  if (N1.getOpcode() == ISD::OR) {
    // (or (or (and), (and)), (or (and), (and)))
    SDValue N000 = N00.getOperand(0);
    if (!isBSwapHWordElement(N000, Parts))
      return SDValue();

    SDValue N001 = N00.getOperand(1);
    if (!isBSwapHWordElement(N001, Parts))
      return SDValue();
    SDValue N010 = N01.getOperand(0);
    if (!isBSwapHWordElement(N010, Parts))
      return SDValue();
    SDValue N011 = N01.getOperand(1);
    if (!isBSwapHWordElement(N011, Parts))
      return SDValue();
  } else {
    // (or (or (or (and), (and)), (and)), (and))
    if (!isBSwapHWordElement(N1, Parts))
      return SDValue();
    if (!isBSwapHWordElement(N01, Parts))
      return SDValue();
    if (N00.getOpcode() != ISD::OR)
      return SDValue();
    SDValue N000 = N00.getOperand(0);
    if (!isBSwapHWordElement(N000, Parts))
      return SDValue();
    SDValue N001 = N00.getOperand(1);
    if (!isBSwapHWordElement(N001, Parts))
      return SDValue();
  }

  // Make sure the parts are all coming from the same node.
  if (Parts[0] != Parts[1] || Parts[0] != Parts[2] || Parts[0] != Parts[3])
    return SDValue();

  SDValue BSwap = DAG.getNode(ISD::BSWAP, N->getDebugLoc(), VT,
                              SDValue(Parts[0],0));

  // Result of the bswap should be rotated by 16. If it's not legal, than
  // do  (x << 16) | (x >> 16).
  SDValue ShAmt = DAG.getConstant(16, getShiftAmountTy(VT));
  if (TLI.isOperationLegalOrCustom(ISD::ROTL, VT))
    return DAG.getNode(ISD::ROTL, N->getDebugLoc(), VT, BSwap, ShAmt);
  if (TLI.isOperationLegalOrCustom(ISD::ROTR, VT))
    return DAG.getNode(ISD::ROTR, N->getDebugLoc(), VT, BSwap, ShAmt);
  return DAG.getNode(ISD::OR, N->getDebugLoc(), VT,
                     DAG.getNode(ISD::SHL, N->getDebugLoc(), VT, BSwap, ShAmt),
                     DAG.getNode(ISD::SRL, N->getDebugLoc(), VT, BSwap, ShAmt));
}

SDValue DAGCombiner::visitOR(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  SDValue N1 = N->getOperand(1);
  SDValue LL, LR, RL, RR, CC0, CC1;
  ConstantSDNode *N0C = dyn_cast<ConstantSDNode>(N0);
  ConstantSDNode *N1C = dyn_cast<ConstantSDNode>(N1);
  EVT VT = N1.getValueType();

  // fold vector ops
  if (VT.isVector()) {
    SDValue FoldedVOp = SimplifyVBinOp(N);
    if (FoldedVOp.getNode()) return FoldedVOp;
  }

  // fold (or x, undef) -> -1
  if (!LegalOperations &&
      (N0.getOpcode() == ISD::UNDEF || N1.getOpcode() == ISD::UNDEF)) {
    EVT EltVT = VT.isVector() ? VT.getVectorElementType() : VT;
    return DAG.getConstant(APInt::getAllOnesValue(EltVT.getSizeInBits()), VT);
  }
  // fold (or c1, c2) -> c1|c2
  if (N0C && N1C)
    return DAG.FoldConstantArithmetic(ISD::OR, VT, N0C, N1C);
  // canonicalize constant to RHS
  if (N0C && !N1C)
    return DAG.getNode(ISD::OR, N->getDebugLoc(), VT, N1, N0);
  // fold (or x, 0) -> x
  if (N1C && N1C->isNullValue())
    return N0;
  // fold (or x, -1) -> -1
  if (N1C && N1C->isAllOnesValue())
    return N1;
  // fold (or x, c) -> c iff (x & ~c) == 0
  if (N1C && DAG.MaskedValueIsZero(N0, ~N1C->getAPIntValue()))
    return N1;

  // Recognize halfword bswaps as (bswap + rotl 16) or (bswap + shl 16)
  SDValue BSwap = MatchBSwapHWord(N, N0, N1);
  if (BSwap.getNode() != 0)
    return BSwap;
  BSwap = MatchBSwapHWordLow(N, N0, N1);
  if (BSwap.getNode() != 0)
    return BSwap;

  // reassociate or
  SDValue ROR = ReassociateOps(ISD::OR, N->getDebugLoc(), N0, N1);
  if (ROR.getNode() != 0)
    return ROR;
  // Canonicalize (or (and X, c1), c2) -> (and (or X, c2), c1|c2)
  // iff (c1 & c2) == 0.
  if (N1C && N0.getOpcode() == ISD::AND && N0.getNode()->hasOneUse() &&
             isa<ConstantSDNode>(N0.getOperand(1))) {
    ConstantSDNode *C1 = cast<ConstantSDNode>(N0.getOperand(1));
    if ((C1->getAPIntValue() & N1C->getAPIntValue()) != 0)
      return DAG.getNode(ISD::AND, N->getDebugLoc(), VT,
                         DAG.getNode(ISD::OR, N0.getDebugLoc(), VT,
                                     N0.getOperand(0), N1),
                         DAG.FoldConstantArithmetic(ISD::OR, VT, N1C, C1));
  }
  // fold (or (setcc x), (setcc y)) -> (setcc (or x, y))
  if (isSetCCEquivalent(N0, LL, LR, CC0) && isSetCCEquivalent(N1, RL, RR, CC1)){
    ISD::CondCode Op0 = cast<CondCodeSDNode>(CC0)->get();
    ISD::CondCode Op1 = cast<CondCodeSDNode>(CC1)->get();

    if (LR == RR && isa<ConstantSDNode>(LR) && Op0 == Op1 &&
        LL.getValueType().isInteger()) {
      // fold (or (setne X, 0), (setne Y, 0)) -> (setne (or X, Y), 0)
      // fold (or (setlt X, 0), (setlt Y, 0)) -> (setne (or X, Y), 0)
      if (cast<ConstantSDNode>(LR)->isNullValue() &&
          (Op1 == ISD::SETNE || Op1 == ISD::SETLT)) {
        SDValue ORNode = DAG.getNode(ISD::OR, LR.getDebugLoc(),
                                     LR.getValueType(), LL, RL);
        AddToWorkList(ORNode.getNode());
        return DAG.getSetCC(N->getDebugLoc(), VT, ORNode, LR, Op1);
      }
      // fold (or (setne X, -1), (setne Y, -1)) -> (setne (and X, Y), -1)
      // fold (or (setgt X, -1), (setgt Y  -1)) -> (setgt (and X, Y), -1)
      if (cast<ConstantSDNode>(LR)->isAllOnesValue() &&
          (Op1 == ISD::SETNE || Op1 == ISD::SETGT)) {
        SDValue ANDNode = DAG.getNode(ISD::AND, LR.getDebugLoc(),
                                      LR.getValueType(), LL, RL);
        AddToWorkList(ANDNode.getNode());
        return DAG.getSetCC(N->getDebugLoc(), VT, ANDNode, LR, Op1);
      }
    }
    // canonicalize equivalent to ll == rl
    if (LL == RR && LR == RL) {
      Op1 = ISD::getSetCCSwappedOperands(Op1);
      std::swap(RL, RR);
    }
    if (LL == RL && LR == RR) {
      bool isInteger = LL.getValueType().isInteger();
      ISD::CondCode Result = ISD::getSetCCOrOperation(Op0, Op1, isInteger);
      if (Result != ISD::SETCC_INVALID &&
          (!LegalOperations || TLI.isCondCodeLegal(Result, LL.getValueType())))
        return DAG.getSetCC(N->getDebugLoc(), N0.getValueType(),
                            LL, LR, Result);
    }
  }

  // Simplify: (or (op x...), (op y...))  -> (op (or x, y))
  if (N0.getOpcode() == N1.getOpcode()) {
    SDValue Tmp = SimplifyBinOpWithSameOpcodeHands(N);
    if (Tmp.getNode()) return Tmp;
  }

  // (or (and X, C1), (and Y, C2))  -> (and (or X, Y), C3) if possible.
  if (N0.getOpcode() == ISD::AND &&
      N1.getOpcode() == ISD::AND &&
      N0.getOperand(1).getOpcode() == ISD::Constant &&
      N1.getOperand(1).getOpcode() == ISD::Constant &&
      // Don't increase # computations.
      (N0.getNode()->hasOneUse() || N1.getNode()->hasOneUse())) {
    // We can only do this xform if we know that bits from X that are set in C2
    // but not in C1 are already zero.  Likewise for Y.
    const APInt &LHSMask =
      cast<ConstantSDNode>(N0.getOperand(1))->getAPIntValue();
    const APInt &RHSMask =
      cast<ConstantSDNode>(N1.getOperand(1))->getAPIntValue();

    if (DAG.MaskedValueIsZero(N0.getOperand(0), RHSMask&~LHSMask) &&
        DAG.MaskedValueIsZero(N1.getOperand(0), LHSMask&~RHSMask)) {
      SDValue X = DAG.getNode(ISD::OR, N0.getDebugLoc(), VT,
                              N0.getOperand(0), N1.getOperand(0));
      return DAG.getNode(ISD::AND, N->getDebugLoc(), VT, X,
                         DAG.getConstant(LHSMask | RHSMask, VT));
    }
  }

  // See if this is some rotate idiom.
  if (SDNode *Rot = MatchRotate(N0, N1, N->getDebugLoc()))
    return SDValue(Rot, 0);

  // Simplify the operands using demanded-bits information.
  if (!VT.isVector() &&
      SimplifyDemandedBits(SDValue(N, 0)))
    return SDValue(N, 0);

  return SDValue();
}

/// MatchRotateHalf - Match "(X shl/srl V1) & V2" where V2 may not be present.
static bool MatchRotateHalf(SDValue Op, SDValue &Shift, SDValue &Mask) {
  if (Op.getOpcode() == ISD::AND) {
    if (isa<ConstantSDNode>(Op.getOperand(1))) {
      Mask = Op.getOperand(1);
      Op = Op.getOperand(0);
    } else {
      return false;
    }
  }

  if (Op.getOpcode() == ISD::SRL || Op.getOpcode() == ISD::SHL) {
    Shift = Op;
    return true;
  }

  return false;
}

// MatchRotate - Handle an 'or' of two operands.  If this is one of the many
// idioms for rotate, and if the target supports rotation instructions, generate
// a rot[lr].
SDNode *DAGCombiner::MatchRotate(SDValue LHS, SDValue RHS, DebugLoc DL) {
  // Must be a legal type.  Expanded 'n promoted things won't work with rotates.
  EVT VT = LHS.getValueType();
  if (!TLI.isTypeLegal(VT)) return 0;

  // The target must have at least one rotate flavor.
  bool HasROTL = TLI.isOperationLegalOrCustom(ISD::ROTL, VT);
  bool HasROTR = TLI.isOperationLegalOrCustom(ISD::ROTR, VT);
  if (!HasROTL && !HasROTR) return 0;

  // Match "(X shl/srl V1) & V2" where V2 may not be present.
  SDValue LHSShift;   // The shift.
  SDValue LHSMask;    // AND value if any.
  if (!MatchRotateHalf(LHS, LHSShift, LHSMask))
    return 0; // Not part of a rotate.

  SDValue RHSShift;   // The shift.
  SDValue RHSMask;    // AND value if any.
  if (!MatchRotateHalf(RHS, RHSShift, RHSMask))
    return 0; // Not part of a rotate.

  if (LHSShift.getOperand(0) != RHSShift.getOperand(0))
    return 0;   // Not shifting the same value.

  if (LHSShift.getOpcode() == RHSShift.getOpcode())
    return 0;   // Shifts must disagree.

  // Canonicalize shl to left side in a shl/srl pair.
  if (RHSShift.getOpcode() == ISD::SHL) {
    std::swap(LHS, RHS);
    std::swap(LHSShift, RHSShift);
    std::swap(LHSMask , RHSMask );
  }

  unsigned OpSizeInBits = VT.getSizeInBits();
  SDValue LHSShiftArg = LHSShift.getOperand(0);
  SDValue LHSShiftAmt = LHSShift.getOperand(1);
  SDValue RHSShiftAmt = RHSShift.getOperand(1);

  // fold (or (shl x, C1), (srl x, C2)) -> (rotl x, C1)
  // fold (or (shl x, C1), (srl x, C2)) -> (rotr x, C2)
  if (LHSShiftAmt.getOpcode() == ISD::Constant &&
      RHSShiftAmt.getOpcode() == ISD::Constant) {
    uint64_t LShVal = cast<ConstantSDNode>(LHSShiftAmt)->getZExtValue();
    uint64_t RShVal = cast<ConstantSDNode>(RHSShiftAmt)->getZExtValue();
    if ((LShVal + RShVal) != OpSizeInBits)
      return 0;

    SDValue Rot = DAG.getNode(HasROTL ? ISD::ROTL : ISD::ROTR, DL, VT,
                              LHSShiftArg, HasROTL ? LHSShiftAmt : RHSShiftAmt);

    // If there is an AND of either shifted operand, apply it to the result.
    if (LHSMask.getNode() || RHSMask.getNode()) {
      APInt Mask = APInt::getAllOnesValue(OpSizeInBits);

      if (LHSMask.getNode()) {
        APInt RHSBits = APInt::getLowBitsSet(OpSizeInBits, LShVal);
        Mask &= cast<ConstantSDNode>(LHSMask)->getAPIntValue() | RHSBits;
      }
      if (RHSMask.getNode()) {
        APInt LHSBits = APInt::getHighBitsSet(OpSizeInBits, RShVal);
        Mask &= cast<ConstantSDNode>(RHSMask)->getAPIntValue() | LHSBits;
      }

      Rot = DAG.getNode(ISD::AND, DL, VT, Rot, DAG.getConstant(Mask, VT));
    }

    return Rot.getNode();
  }

  // If there is a mask here, and we have a variable shift, we can't be sure
  // that we're masking out the right stuff.
  if (LHSMask.getNode() || RHSMask.getNode())
    return 0;

  // fold (or (shl x, y), (srl x, (sub 32, y))) -> (rotl x, y)
  // fold (or (shl x, y), (srl x, (sub 32, y))) -> (rotr x, (sub 32, y))
  if (RHSShiftAmt.getOpcode() == ISD::SUB &&
      LHSShiftAmt == RHSShiftAmt.getOperand(1)) {
    if (ConstantSDNode *SUBC =
          dyn_cast<ConstantSDNode>(RHSShiftAmt.getOperand(0))) {
      if (SUBC->getAPIntValue() == OpSizeInBits) {
        return DAG.getNode(HasROTL ? ISD::ROTL : ISD::ROTR, DL, VT, LHSShiftArg,
                           HasROTL ? LHSShiftAmt : RHSShiftAmt).getNode();
      }
    }
  }

  // fold (or (shl x, (sub 32, y)), (srl x, r)) -> (rotr x, y)
  // fold (or (shl x, (sub 32, y)), (srl x, r)) -> (rotl x, (sub 32, y))
  if (LHSShiftAmt.getOpcode() == ISD::SUB &&
      RHSShiftAmt == LHSShiftAmt.getOperand(1)) {
    if (ConstantSDNode *SUBC =
          dyn_cast<ConstantSDNode>(LHSShiftAmt.getOperand(0))) {
      if (SUBC->getAPIntValue() == OpSizeInBits) {
        return DAG.getNode(HasROTR ? ISD::ROTR : ISD::ROTL, DL, VT, LHSShiftArg,
                           HasROTR ? RHSShiftAmt : LHSShiftAmt).getNode();
      }
    }
  }

  // Look for sign/zext/any-extended or truncate cases:
  if ((LHSShiftAmt.getOpcode() == ISD::SIGN_EXTEND ||
       LHSShiftAmt.getOpcode() == ISD::ZERO_EXTEND ||
       LHSShiftAmt.getOpcode() == ISD::ANY_EXTEND ||
       LHSShiftAmt.getOpcode() == ISD::TRUNCATE) &&
      (RHSShiftAmt.getOpcode() == ISD::SIGN_EXTEND ||
       RHSShiftAmt.getOpcode() == ISD::ZERO_EXTEND ||
       RHSShiftAmt.getOpcode() == ISD::ANY_EXTEND ||
       RHSShiftAmt.getOpcode() == ISD::TRUNCATE)) {
    SDValue LExtOp0 = LHSShiftAmt.getOperand(0);
    SDValue RExtOp0 = RHSShiftAmt.getOperand(0);
    if (RExtOp0.getOpcode() == ISD::SUB &&
        RExtOp0.getOperand(1) == LExtOp0) {
      // fold (or (shl x, (*ext y)), (srl x, (*ext (sub 32, y)))) ->
      //   (rotl x, y)
      // fold (or (shl x, (*ext y)), (srl x, (*ext (sub 32, y)))) ->
      //   (rotr x, (sub 32, y))
      if (ConstantSDNode *SUBC =
            dyn_cast<ConstantSDNode>(RExtOp0.getOperand(0))) {
        if (SUBC->getAPIntValue() == OpSizeInBits) {
          return DAG.getNode(HasROTL ? ISD::ROTL : ISD::ROTR, DL, VT,
                             LHSShiftArg,
                             HasROTL ? LHSShiftAmt : RHSShiftAmt).getNode();
        }
      }
    } else if (LExtOp0.getOpcode() == ISD::SUB &&
               RExtOp0 == LExtOp0.getOperand(1)) {
      // fold (or (shl x, (*ext (sub 32, y))), (srl x, (*ext y))) ->
      //   (rotr x, y)
      // fold (or (shl x, (*ext (sub 32, y))), (srl x, (*ext y))) ->
      //   (rotl x, (sub 32, y))
      if (ConstantSDNode *SUBC =
            dyn_cast<ConstantSDNode>(LExtOp0.getOperand(0))) {
        if (SUBC->getAPIntValue() == OpSizeInBits) {
          return DAG.getNode(HasROTR ? ISD::ROTR : ISD::ROTL, DL, VT,
                             LHSShiftArg,
                             HasROTR ? RHSShiftAmt : LHSShiftAmt).getNode();
        }
      }
    }
  }

  return 0;
}

SDValue DAGCombiner::visitXOR(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  SDValue N1 = N->getOperand(1);
  SDValue LHS, RHS, CC;
  ConstantSDNode *N0C = dyn_cast<ConstantSDNode>(N0);
  ConstantSDNode *N1C = dyn_cast<ConstantSDNode>(N1);
  EVT VT = N0.getValueType();

  // fold vector ops
  if (VT.isVector()) {
    SDValue FoldedVOp = SimplifyVBinOp(N);
    if (FoldedVOp.getNode()) return FoldedVOp;
  }

  // fold (xor undef, undef) -> 0. This is a common idiom (misuse).
  if (N0.getOpcode() == ISD::UNDEF && N1.getOpcode() == ISD::UNDEF)
    return DAG.getConstant(0, VT);
  // fold (xor x, undef) -> undef
  if (N0.getOpcode() == ISD::UNDEF)
    return N0;
  if (N1.getOpcode() == ISD::UNDEF)
    return N1;
  // fold (xor c1, c2) -> c1^c2
  if (N0C && N1C)
    return DAG.FoldConstantArithmetic(ISD::XOR, VT, N0C, N1C);
  // canonicalize constant to RHS
  if (N0C && !N1C)
    return DAG.getNode(ISD::XOR, N->getDebugLoc(), VT, N1, N0);
  // fold (xor x, 0) -> x
  if (N1C && N1C->isNullValue())
    return N0;
  // reassociate xor
  SDValue RXOR = ReassociateOps(ISD::XOR, N->getDebugLoc(), N0, N1);
  if (RXOR.getNode() != 0)
    return RXOR;

  // fold !(x cc y) -> (x !cc y)
  if (N1C && N1C->getAPIntValue() == 1 && isSetCCEquivalent(N0, LHS, RHS, CC)) {
    bool isInt = LHS.getValueType().isInteger();
    ISD::CondCode NotCC = ISD::getSetCCInverse(cast<CondCodeSDNode>(CC)->get(),
                                               isInt);

    if (!LegalOperations || TLI.isCondCodeLegal(NotCC, LHS.getValueType())) {
      switch (N0.getOpcode()) {
      default:
        llvm_unreachable("Unhandled SetCC Equivalent!");
      case ISD::SETCC:
        return DAG.getSetCC(N->getDebugLoc(), VT, LHS, RHS, NotCC);
      case ISD::SELECT_CC:
        return DAG.getSelectCC(N->getDebugLoc(), LHS, RHS, N0.getOperand(2),
                               N0.getOperand(3), NotCC);
      }
    }
  }

  // fold (not (zext (setcc x, y))) -> (zext (not (setcc x, y)))
  if (N1C && N1C->getAPIntValue() == 1 && N0.getOpcode() == ISD::ZERO_EXTEND &&
      N0.getNode()->hasOneUse() &&
      isSetCCEquivalent(N0.getOperand(0), LHS, RHS, CC)){
    SDValue V = N0.getOperand(0);
    V = DAG.getNode(ISD::XOR, N0.getDebugLoc(), V.getValueType(), V,
                    DAG.getConstant(1, V.getValueType()));
    AddToWorkList(V.getNode());
    return DAG.getNode(ISD::ZERO_EXTEND, N->getDebugLoc(), VT, V);
  }

  // fold (not (or x, y)) -> (and (not x), (not y)) iff x or y are setcc
  if (N1C && N1C->getAPIntValue() == 1 && VT == MVT::i1 &&
      (N0.getOpcode() == ISD::OR || N0.getOpcode() == ISD::AND)) {
    SDValue LHS = N0.getOperand(0), RHS = N0.getOperand(1);
    if (isOneUseSetCC(RHS) || isOneUseSetCC(LHS)) {
      unsigned NewOpcode = N0.getOpcode() == ISD::AND ? ISD::OR : ISD::AND;
      LHS = DAG.getNode(ISD::XOR, LHS.getDebugLoc(), VT, LHS, N1); // LHS = ~LHS
      RHS = DAG.getNode(ISD::XOR, RHS.getDebugLoc(), VT, RHS, N1); // RHS = ~RHS
      AddToWorkList(LHS.getNode()); AddToWorkList(RHS.getNode());
      return DAG.getNode(NewOpcode, N->getDebugLoc(), VT, LHS, RHS);
    }
  }
  // fold (not (or x, y)) -> (and (not x), (not y)) iff x or y are constants
  if (N1C && N1C->isAllOnesValue() &&
      (N0.getOpcode() == ISD::OR || N0.getOpcode() == ISD::AND)) {
    SDValue LHS = N0.getOperand(0), RHS = N0.getOperand(1);
    if (isa<ConstantSDNode>(RHS) || isa<ConstantSDNode>(LHS)) {
      unsigned NewOpcode = N0.getOpcode() == ISD::AND ? ISD::OR : ISD::AND;
      LHS = DAG.getNode(ISD::XOR, LHS.getDebugLoc(), VT, LHS, N1); // LHS = ~LHS
      RHS = DAG.getNode(ISD::XOR, RHS.getDebugLoc(), VT, RHS, N1); // RHS = ~RHS
      AddToWorkList(LHS.getNode()); AddToWorkList(RHS.getNode());
      return DAG.getNode(NewOpcode, N->getDebugLoc(), VT, LHS, RHS);
    }
  }
  // fold (xor (xor x, c1), c2) -> (xor x, (xor c1, c2))
  if (N1C && N0.getOpcode() == ISD::XOR) {
    ConstantSDNode *N00C = dyn_cast<ConstantSDNode>(N0.getOperand(0));
    ConstantSDNode *N01C = dyn_cast<ConstantSDNode>(N0.getOperand(1));
    if (N00C)
      return DAG.getNode(ISD::XOR, N->getDebugLoc(), VT, N0.getOperand(1),
                         DAG.getConstant(N1C->getAPIntValue() ^
                                         N00C->getAPIntValue(), VT));
    if (N01C)
      return DAG.getNode(ISD::XOR, N->getDebugLoc(), VT, N0.getOperand(0),
                         DAG.getConstant(N1C->getAPIntValue() ^
                                         N01C->getAPIntValue(), VT));
  }
  // fold (xor x, x) -> 0
  if (N0 == N1)
    return tryFoldToZero(N->getDebugLoc(), TLI, VT, DAG, LegalOperations);

  // Simplify: xor (op x...), (op y...)  -> (op (xor x, y))
  if (N0.getOpcode() == N1.getOpcode()) {
    SDValue Tmp = SimplifyBinOpWithSameOpcodeHands(N);
    if (Tmp.getNode()) return Tmp;
  }

  // Simplify the expression using non-local knowledge.
  if (!VT.isVector() &&
      SimplifyDemandedBits(SDValue(N, 0)))
    return SDValue(N, 0);

  return SDValue();
}

/// visitShiftByConstant - Handle transforms common to the three shifts, when
/// the shift amount is a constant.
SDValue DAGCombiner::visitShiftByConstant(SDNode *N, unsigned Amt) {
  SDNode *LHS = N->getOperand(0).getNode();
  if (!LHS->hasOneUse()) return SDValue();

  // We want to pull some binops through shifts, so that we have (and (shift))
  // instead of (shift (and)), likewise for add, or, xor, etc.  This sort of
  // thing happens with address calculations, so it's important to canonicalize
  // it.
  bool HighBitSet = false;  // Can we transform this if the high bit is set?

  switch (LHS->getOpcode()) {
  default: return SDValue();
  case ISD::OR:
  case ISD::XOR:
    HighBitSet = false; // We can only transform sra if the high bit is clear.
    break;
  case ISD::AND:
    HighBitSet = true;  // We can only transform sra if the high bit is set.
    break;
  case ISD::ADD:
    if (N->getOpcode() != ISD::SHL)
      return SDValue(); // only shl(add) not sr[al](add).
    HighBitSet = false; // We can only transform sra if the high bit is clear.
    break;
  }

  // We require the RHS of the binop to be a constant as well.
  ConstantSDNode *BinOpCst = dyn_cast<ConstantSDNode>(LHS->getOperand(1));
  if (!BinOpCst) return SDValue();

  // FIXME: disable this unless the input to the binop is a shift by a constant.
  // If it is not a shift, it pessimizes some common cases like:
  //
  //    void foo(int *X, int i) { X[i & 1235] = 1; }
  //    int bar(int *X, int i) { return X[i & 255]; }
  SDNode *BinOpLHSVal = LHS->getOperand(0).getNode();
  if ((BinOpLHSVal->getOpcode() != ISD::SHL &&
       BinOpLHSVal->getOpcode() != ISD::SRA &&
       BinOpLHSVal->getOpcode() != ISD::SRL) ||
      !isa<ConstantSDNode>(BinOpLHSVal->getOperand(1)))
    return SDValue();

  EVT VT = N->getValueType(0);

  // If this is a signed shift right, and the high bit is modified by the
  // logical operation, do not perform the transformation. The highBitSet
  // boolean indicates the value of the high bit of the constant which would
  // cause it to be modified for this operation.
  if (N->getOpcode() == ISD::SRA) {
    bool BinOpRHSSignSet = BinOpCst->getAPIntValue().isNegative();
    if (BinOpRHSSignSet != HighBitSet)
      return SDValue();
  }

  // Fold the constants, shifting the binop RHS by the shift amount.
  SDValue NewRHS = DAG.getNode(N->getOpcode(), LHS->getOperand(1).getDebugLoc(),
                               N->getValueType(0),
                               LHS->getOperand(1), N->getOperand(1));

  // Create the new shift.
  SDValue NewShift = DAG.getNode(N->getOpcode(),
                                 LHS->getOperand(0).getDebugLoc(),
                                 VT, LHS->getOperand(0), N->getOperand(1));

  // Create the new binop.
  return DAG.getNode(LHS->getOpcode(), N->getDebugLoc(), VT, NewShift, NewRHS);
}

SDValue DAGCombiner::visitSHL(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  SDValue N1 = N->getOperand(1);
  ConstantSDNode *N0C = dyn_cast<ConstantSDNode>(N0);
  ConstantSDNode *N1C = dyn_cast<ConstantSDNode>(N1);
  EVT VT = N0.getValueType();
  unsigned OpSizeInBits = VT.getScalarType().getSizeInBits();

  // fold (shl c1, c2) -> c1<<c2
  if (N0C && N1C)
    return DAG.FoldConstantArithmetic(ISD::SHL, VT, N0C, N1C);
  // fold (shl 0, x) -> 0
  if (N0C && N0C->isNullValue())
    return N0;
  // fold (shl x, c >= size(x)) -> undef
  if (N1C && N1C->getZExtValue() >= OpSizeInBits)
    return DAG.getUNDEF(VT);
  // fold (shl x, 0) -> x
  if (N1C && N1C->isNullValue())
    return N0;
  // fold (shl undef, x) -> 0
  if (N0.getOpcode() == ISD::UNDEF)
    return DAG.getConstant(0, VT);
  // if (shl x, c) is known to be zero, return 0
  if (DAG.MaskedValueIsZero(SDValue(N, 0),
                            APInt::getAllOnesValue(OpSizeInBits)))
    return DAG.getConstant(0, VT);
  // fold (shl x, (trunc (and y, c))) -> (shl x, (and (trunc y), (trunc c))).
  if (N1.getOpcode() == ISD::TRUNCATE &&
      N1.getOperand(0).getOpcode() == ISD::AND &&
      N1.hasOneUse() && N1.getOperand(0).hasOneUse()) {
    SDValue N101 = N1.getOperand(0).getOperand(1);
    if (ConstantSDNode *N101C = dyn_cast<ConstantSDNode>(N101)) {
      EVT TruncVT = N1.getValueType();
      SDValue N100 = N1.getOperand(0).getOperand(0);
      APInt TruncC = N101C->getAPIntValue();
      TruncC = TruncC.trunc(TruncVT.getSizeInBits());
      return DAG.getNode(ISD::SHL, N->getDebugLoc(), VT, N0,
                         DAG.getNode(ISD::AND, N->getDebugLoc(), TruncVT,
                                     DAG.getNode(ISD::TRUNCATE,
                                                 N->getDebugLoc(),
                                                 TruncVT, N100),
                                     DAG.getConstant(TruncC, TruncVT)));
    }
  }

  if (N1C && SimplifyDemandedBits(SDValue(N, 0)))
    return SDValue(N, 0);

  // fold (shl (shl x, c1), c2) -> 0 or (shl x, (add c1, c2))
  if (N1C && N0.getOpcode() == ISD::SHL &&
      N0.getOperand(1).getOpcode() == ISD::Constant) {
    uint64_t c1 = cast<ConstantSDNode>(N0.getOperand(1))->getZExtValue();
    uint64_t c2 = N1C->getZExtValue();
    if (c1 + c2 >= OpSizeInBits)
      return DAG.getConstant(0, VT);
    return DAG.getNode(ISD::SHL, N->getDebugLoc(), VT, N0.getOperand(0),
                       DAG.getConstant(c1 + c2, N1.getValueType()));
  }

  // fold (shl (ext (shl x, c1)), c2) -> (ext (shl x, (add c1, c2)))
  // For this to be valid, the second form must not preserve any of the bits
  // that are shifted out by the inner shift in the first form.  This means
  // the outer shift size must be >= the number of bits added by the ext.
  // As a corollary, we don't care what kind of ext it is.
  if (N1C && (N0.getOpcode() == ISD::ZERO_EXTEND ||
              N0.getOpcode() == ISD::ANY_EXTEND ||
              N0.getOpcode() == ISD::SIGN_EXTEND) &&
      N0.getOperand(0).getOpcode() == ISD::SHL &&
      isa<ConstantSDNode>(N0.getOperand(0)->getOperand(1))) {
    uint64_t c1 =
      cast<ConstantSDNode>(N0.getOperand(0)->getOperand(1))->getZExtValue();
    uint64_t c2 = N1C->getZExtValue();
    EVT InnerShiftVT = N0.getOperand(0).getValueType();
    uint64_t InnerShiftSize = InnerShiftVT.getScalarType().getSizeInBits();
    if (c2 >= OpSizeInBits - InnerShiftSize) {
      if (c1 + c2 >= OpSizeInBits)
        return DAG.getConstant(0, VT);
      return DAG.getNode(ISD::SHL, N0->getDebugLoc(), VT,
                         DAG.getNode(N0.getOpcode(), N0->getDebugLoc(), VT,
                                     N0.getOperand(0)->getOperand(0)),
                         DAG.getConstant(c1 + c2, N1.getValueType()));
    }
  }

  // fold (shl (srl x, c1), c2) -> (and (shl x, (sub c2, c1), MASK) or
  //                               (and (srl x, (sub c1, c2), MASK)
  // Only fold this if the inner shift has no other uses -- if it does, folding
  // this will increase the total number of instructions.
  if (N1C && N0.getOpcode() == ISD::SRL && N0.hasOneUse() &&
      N0.getOperand(1).getOpcode() == ISD::Constant) {
    uint64_t c1 = cast<ConstantSDNode>(N0.getOperand(1))->getZExtValue();
    if (c1 < VT.getSizeInBits()) {
      uint64_t c2 = N1C->getZExtValue();
      APInt Mask = APInt::getHighBitsSet(VT.getSizeInBits(),
                                         VT.getSizeInBits() - c1);
      SDValue Shift;
      if (c2 > c1) {
        Mask = Mask.shl(c2-c1);
        Shift = DAG.getNode(ISD::SHL, N->getDebugLoc(), VT, N0.getOperand(0),
                            DAG.getConstant(c2-c1, N1.getValueType()));
      } else {
        Mask = Mask.lshr(c1-c2);
        Shift = DAG.getNode(ISD::SRL, N->getDebugLoc(), VT, N0.getOperand(0),
                            DAG.getConstant(c1-c2, N1.getValueType()));
      }
      return DAG.getNode(ISD::AND, N0.getDebugLoc(), VT, Shift,
                         DAG.getConstant(Mask, VT));
    }
  }
  // fold (shl (sra x, c1), c1) -> (and x, (shl -1, c1))
  if (N1C && N0.getOpcode() == ISD::SRA && N1 == N0.getOperand(1)) {
    SDValue HiBitsMask =
      DAG.getConstant(APInt::getHighBitsSet(VT.getSizeInBits(),
                                            VT.getSizeInBits() -
                                              N1C->getZExtValue()),
                      VT);
    return DAG.getNode(ISD::AND, N->getDebugLoc(), VT, N0.getOperand(0),
                       HiBitsMask);
  }

  if (N1C) {
    SDValue NewSHL = visitShiftByConstant(N, N1C->getZExtValue());
    if (NewSHL.getNode())
      return NewSHL;
  }

  return SDValue();
}

SDValue DAGCombiner::visitSRA(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  SDValue N1 = N->getOperand(1);
  ConstantSDNode *N0C = dyn_cast<ConstantSDNode>(N0);
  ConstantSDNode *N1C = dyn_cast<ConstantSDNode>(N1);
  EVT VT = N0.getValueType();
  unsigned OpSizeInBits = VT.getScalarType().getSizeInBits();

  // fold (sra c1, c2) -> (sra c1, c2)
  if (N0C && N1C)
    return DAG.FoldConstantArithmetic(ISD::SRA, VT, N0C, N1C);
  // fold (sra 0, x) -> 0
  if (N0C && N0C->isNullValue())
    return N0;
  // fold (sra -1, x) -> -1
  if (N0C && N0C->isAllOnesValue())
    return N0;
  // fold (sra x, (setge c, size(x))) -> undef
  if (N1C && N1C->getZExtValue() >= OpSizeInBits)
    return DAG.getUNDEF(VT);
  // fold (sra x, 0) -> x
  if (N1C && N1C->isNullValue())
    return N0;
  // fold (sra (shl x, c1), c1) -> sext_inreg for some c1 and target supports
  // sext_inreg.
  if (N1C && N0.getOpcode() == ISD::SHL && N1 == N0.getOperand(1)) {
    unsigned LowBits = OpSizeInBits - (unsigned)N1C->getZExtValue();
    EVT ExtVT = EVT::getIntegerVT(*DAG.getContext(), LowBits);
    if (VT.isVector())
      ExtVT = EVT::getVectorVT(*DAG.getContext(),
                               ExtVT, VT.getVectorNumElements());
    if ((!LegalOperations ||
         TLI.isOperationLegal(ISD::SIGN_EXTEND_INREG, ExtVT)))
      return DAG.getNode(ISD::SIGN_EXTEND_INREG, N->getDebugLoc(), VT,
                         N0.getOperand(0), DAG.getValueType(ExtVT));
  }

  // fold (sra (sra x, c1), c2) -> (sra x, (add c1, c2))
  if (N1C && N0.getOpcode() == ISD::SRA) {
    if (ConstantSDNode *C1 = dyn_cast<ConstantSDNode>(N0.getOperand(1))) {
      unsigned Sum = N1C->getZExtValue() + C1->getZExtValue();
      if (Sum >= OpSizeInBits) Sum = OpSizeInBits-1;
      return DAG.getNode(ISD::SRA, N->getDebugLoc(), VT, N0.getOperand(0),
                         DAG.getConstant(Sum, N1C->getValueType(0)));
    }
  }

  // fold (sra (shl X, m), (sub result_size, n))
  // -> (sign_extend (trunc (shl X, (sub (sub result_size, n), m)))) for
  // result_size - n != m.
  // If truncate is free for the target sext(shl) is likely to result in better
  // code.
  if (N0.getOpcode() == ISD::SHL) {
    // Get the two constanst of the shifts, CN0 = m, CN = n.
    const ConstantSDNode *N01C = dyn_cast<ConstantSDNode>(N0.getOperand(1));
    if (N01C && N1C) {
      // Determine what the truncate's result bitsize and type would be.
      EVT TruncVT =
        EVT::getIntegerVT(*DAG.getContext(),
                          OpSizeInBits - N1C->getZExtValue());
      // Determine the residual right-shift amount.
      signed ShiftAmt = N1C->getZExtValue() - N01C->getZExtValue();

      // If the shift is not a no-op (in which case this should be just a sign
      // extend already), the truncated to type is legal, sign_extend is legal
      // on that type, and the truncate to that type is both legal and free,
      // perform the transform.
      if ((ShiftAmt > 0) &&
          TLI.isOperationLegalOrCustom(ISD::SIGN_EXTEND, TruncVT) &&
          TLI.isOperationLegalOrCustom(ISD::TRUNCATE, VT) &&
          TLI.isTruncateFree(VT, TruncVT)) {

          SDValue Amt = DAG.getConstant(ShiftAmt,
              getShiftAmountTy(N0.getOperand(0).getValueType()));
          SDValue Shift = DAG.getNode(ISD::SRL, N0.getDebugLoc(), VT,
                                      N0.getOperand(0), Amt);
          SDValue Trunc = DAG.getNode(ISD::TRUNCATE, N0.getDebugLoc(), TruncVT,
                                      Shift);
          return DAG.getNode(ISD::SIGN_EXTEND, N->getDebugLoc(),
                             N->getValueType(0), Trunc);
      }
    }
  }

  // fold (sra x, (trunc (and y, c))) -> (sra x, (and (trunc y), (trunc c))).
  if (N1.getOpcode() == ISD::TRUNCATE &&
      N1.getOperand(0).getOpcode() == ISD::AND &&
      N1.hasOneUse() && N1.getOperand(0).hasOneUse()) {
    SDValue N101 = N1.getOperand(0).getOperand(1);
    if (ConstantSDNode *N101C = dyn_cast<ConstantSDNode>(N101)) {
      EVT TruncVT = N1.getValueType();
      SDValue N100 = N1.getOperand(0).getOperand(0);
      APInt TruncC = N101C->getAPIntValue();
      TruncC = TruncC.trunc(TruncVT.getScalarType().getSizeInBits());
      return DAG.getNode(ISD::SRA, N->getDebugLoc(), VT, N0,
                         DAG.getNode(ISD::AND, N->getDebugLoc(),
                                     TruncVT,
                                     DAG.getNode(ISD::TRUNCATE,
                                                 N->getDebugLoc(),
                                                 TruncVT, N100),
                                     DAG.getConstant(TruncC, TruncVT)));
    }
  }

  // fold (sra (trunc (sr x, c1)), c2) -> (trunc (sra x, c1+c2))
  //      if c1 is equal to the number of bits the trunc removes
  if (N0.getOpcode() == ISD::TRUNCATE &&
      (N0.getOperand(0).getOpcode() == ISD::SRL ||
       N0.getOperand(0).getOpcode() == ISD::SRA) &&
      N0.getOperand(0).hasOneUse() &&
      N0.getOperand(0).getOperand(1).hasOneUse() &&
      N1C && isa<ConstantSDNode>(N0.getOperand(0).getOperand(1))) {
    EVT LargeVT = N0.getOperand(0).getValueType();
    ConstantSDNode *LargeShiftAmt =
      cast<ConstantSDNode>(N0.getOperand(0).getOperand(1));

    if (LargeVT.getScalarType().getSizeInBits() - OpSizeInBits ==
        LargeShiftAmt->getZExtValue()) {
      SDValue Amt =
        DAG.getConstant(LargeShiftAmt->getZExtValue() + N1C->getZExtValue(),
              getShiftAmountTy(N0.getOperand(0).getOperand(0).getValueType()));
      SDValue SRA = DAG.getNode(ISD::SRA, N->getDebugLoc(), LargeVT,
                                N0.getOperand(0).getOperand(0), Amt);
      return DAG.getNode(ISD::TRUNCATE, N->getDebugLoc(), VT, SRA);
    }
  }

  // Simplify, based on bits shifted out of the LHS.
  if (N1C && SimplifyDemandedBits(SDValue(N, 0)))
    return SDValue(N, 0);


  // If the sign bit is known to be zero, switch this to a SRL.
  if (DAG.SignBitIsZero(N0))
    return DAG.getNode(ISD::SRL, N->getDebugLoc(), VT, N0, N1);

  if (N1C) {
    SDValue NewSRA = visitShiftByConstant(N, N1C->getZExtValue());
    if (NewSRA.getNode())
      return NewSRA;
  }

  return SDValue();
}

SDValue DAGCombiner::visitSRL(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  SDValue N1 = N->getOperand(1);
  ConstantSDNode *N0C = dyn_cast<ConstantSDNode>(N0);
  ConstantSDNode *N1C = dyn_cast<ConstantSDNode>(N1);
  EVT VT = N0.getValueType();
  unsigned OpSizeInBits = VT.getScalarType().getSizeInBits();

  // fold (srl c1, c2) -> c1 >>u c2
  if (N0C && N1C)
    return DAG.FoldConstantArithmetic(ISD::SRL, VT, N0C, N1C);
  // fold (srl 0, x) -> 0
  if (N0C && N0C->isNullValue())
    return N0;
  // fold (srl x, c >= size(x)) -> undef
  if (N1C && N1C->getZExtValue() >= OpSizeInBits)
    return DAG.getUNDEF(VT);
  // fold (srl x, 0) -> x
  if (N1C && N1C->isNullValue())
    return N0;
  // if (srl x, c) is known to be zero, return 0
  if (N1C && DAG.MaskedValueIsZero(SDValue(N, 0),
                                   APInt::getAllOnesValue(OpSizeInBits)))
    return DAG.getConstant(0, VT);

  // fold (srl (srl x, c1), c2) -> 0 or (srl x, (add c1, c2))
  if (N1C && N0.getOpcode() == ISD::SRL &&
      N0.getOperand(1).getOpcode() == ISD::Constant) {
    uint64_t c1 = cast<ConstantSDNode>(N0.getOperand(1))->getZExtValue();
    uint64_t c2 = N1C->getZExtValue();
    if (c1 + c2 >= OpSizeInBits)
      return DAG.getConstant(0, VT);
    return DAG.getNode(ISD::SRL, N->getDebugLoc(), VT, N0.getOperand(0),
                       DAG.getConstant(c1 + c2, N1.getValueType()));
  }

  // fold (srl (trunc (srl x, c1)), c2) -> 0 or (trunc (srl x, (add c1, c2)))
  if (N1C && N0.getOpcode() == ISD::TRUNCATE &&
      N0.getOperand(0).getOpcode() == ISD::SRL &&
      isa<ConstantSDNode>(N0.getOperand(0)->getOperand(1))) {
    uint64_t c1 =
      cast<ConstantSDNode>(N0.getOperand(0)->getOperand(1))->getZExtValue();
    uint64_t c2 = N1C->getZExtValue();
    EVT InnerShiftVT = N0.getOperand(0).getValueType();
    EVT ShiftCountVT = N0.getOperand(0)->getOperand(1).getValueType();
    uint64_t InnerShiftSize = InnerShiftVT.getScalarType().getSizeInBits();
    // This is only valid if the OpSizeInBits + c1 = size of inner shift.
    if (c1 + OpSizeInBits == InnerShiftSize) {
      if (c1 + c2 >= InnerShiftSize)
        return DAG.getConstant(0, VT);
      return DAG.getNode(ISD::TRUNCATE, N0->getDebugLoc(), VT,
                         DAG.getNode(ISD::SRL, N0->getDebugLoc(), InnerShiftVT,
                                     N0.getOperand(0)->getOperand(0),
                                     DAG.getConstant(c1 + c2, ShiftCountVT)));
    }
  }

  // fold (srl (shl x, c), c) -> (and x, cst2)
  if (N1C && N0.getOpcode() == ISD::SHL && N0.getOperand(1) == N1 &&
      N0.getValueSizeInBits() <= 64) {
    uint64_t ShAmt = N1C->getZExtValue()+64-N0.getValueSizeInBits();
    return DAG.getNode(ISD::AND, N->getDebugLoc(), VT, N0.getOperand(0),
                       DAG.getConstant(~0ULL >> ShAmt, VT));
  }


  // fold (srl (anyextend x), c) -> (anyextend (srl x, c))
  if (N1C && N0.getOpcode() == ISD::ANY_EXTEND) {
    // Shifting in all undef bits?
    EVT SmallVT = N0.getOperand(0).getValueType();
    if (N1C->getZExtValue() >= SmallVT.getSizeInBits())
      return DAG.getUNDEF(VT);

    if (!LegalTypes || TLI.isTypeDesirableForOp(ISD::SRL, SmallVT)) {
      uint64_t ShiftAmt = N1C->getZExtValue();
      SDValue SmallShift = DAG.getNode(ISD::SRL, N0.getDebugLoc(), SmallVT,
                                       N0.getOperand(0),
                          DAG.getConstant(ShiftAmt, getShiftAmountTy(SmallVT)));
      AddToWorkList(SmallShift.getNode());
      return DAG.getNode(ISD::ANY_EXTEND, N->getDebugLoc(), VT, SmallShift);
    }
  }

  // fold (srl (sra X, Y), 31) -> (srl X, 31).  This srl only looks at the sign
  // bit, which is unmodified by sra.
  if (N1C && N1C->getZExtValue() + 1 == VT.getSizeInBits()) {
    if (N0.getOpcode() == ISD::SRA)
      return DAG.getNode(ISD::SRL, N->getDebugLoc(), VT, N0.getOperand(0), N1);
  }

  // fold (srl (ctlz x), "5") -> x  iff x has one bit set (the low bit).
  if (N1C && N0.getOpcode() == ISD::CTLZ &&
      N1C->getAPIntValue() == Log2_32(VT.getSizeInBits())) {
    APInt KnownZero, KnownOne;
    DAG.ComputeMaskedBits(N0.getOperand(0), KnownZero, KnownOne);

    // If any of the input bits are KnownOne, then the input couldn't be all
    // zeros, thus the result of the srl will always be zero.
    if (KnownOne.getBoolValue()) return DAG.getConstant(0, VT);

    // If all of the bits input the to ctlz node are known to be zero, then
    // the result of the ctlz is "32" and the result of the shift is one.
    APInt UnknownBits = ~KnownZero;
    if (UnknownBits == 0) return DAG.getConstant(1, VT);

    // Otherwise, check to see if there is exactly one bit input to the ctlz.
    if ((UnknownBits & (UnknownBits - 1)) == 0) {
      // Okay, we know that only that the single bit specified by UnknownBits
      // could be set on input to the CTLZ node. If this bit is set, the SRL
      // will return 0, if it is clear, it returns 1. Change the CTLZ/SRL pair
      // to an SRL/XOR pair, which is likely to simplify more.
      unsigned ShAmt = UnknownBits.countTrailingZeros();
      SDValue Op = N0.getOperand(0);

      if (ShAmt) {
        Op = DAG.getNode(ISD::SRL, N0.getDebugLoc(), VT, Op,
                  DAG.getConstant(ShAmt, getShiftAmountTy(Op.getValueType())));
        AddToWorkList(Op.getNode());
      }

      return DAG.getNode(ISD::XOR, N->getDebugLoc(), VT,
                         Op, DAG.getConstant(1, VT));
    }
  }

  // fold (srl x, (trunc (and y, c))) -> (srl x, (and (trunc y), (trunc c))).
  if (N1.getOpcode() == ISD::TRUNCATE &&
      N1.getOperand(0).getOpcode() == ISD::AND &&
      N1.hasOneUse() && N1.getOperand(0).hasOneUse()) {
    SDValue N101 = N1.getOperand(0).getOperand(1);
    if (ConstantSDNode *N101C = dyn_cast<ConstantSDNode>(N101)) {
      EVT TruncVT = N1.getValueType();
      SDValue N100 = N1.getOperand(0).getOperand(0);
      APInt TruncC = N101C->getAPIntValue();
      TruncC = TruncC.trunc(TruncVT.getSizeInBits());
      return DAG.getNode(ISD::SRL, N->getDebugLoc(), VT, N0,
                         DAG.getNode(ISD::AND, N->getDebugLoc(),
                                     TruncVT,
                                     DAG.getNode(ISD::TRUNCATE,
                                                 N->getDebugLoc(),
                                                 TruncVT, N100),
                                     DAG.getConstant(TruncC, TruncVT)));
    }
  }

  // fold operands of srl based on knowledge that the low bits are not
  // demanded.
  if (N1C && SimplifyDemandedBits(SDValue(N, 0)))
    return SDValue(N, 0);

  if (N1C) {
    SDValue NewSRL = visitShiftByConstant(N, N1C->getZExtValue());
    if (NewSRL.getNode())
      return NewSRL;
  }

  // Attempt to convert a srl of a load into a narrower zero-extending load.
  SDValue NarrowLoad = ReduceLoadWidth(N);
  if (NarrowLoad.getNode())
    return NarrowLoad;

  // Here is a common situation. We want to optimize:
  //
  //   %a = ...
  //   %b = and i32 %a, 2
  //   %c = srl i32 %b, 1
  //   brcond i32 %c ...
  //
  // into
  //
  //   %a = ...
  //   %b = and %a, 2
  //   %c = setcc eq %b, 0
  //   brcond %c ...
  //
  // However when after the source operand of SRL is optimized into AND, the SRL
  // itself may not be optimized further. Look for it and add the BRCOND into
  // the worklist.
  if (N->hasOneUse()) {
    SDNode *Use = *N->use_begin();
    if (Use->getOpcode() == ISD::BRCOND)
      AddToWorkList(Use);
    else if (Use->getOpcode() == ISD::TRUNCATE && Use->hasOneUse()) {
      // Also look pass the truncate.
      Use = *Use->use_begin();
      if (Use->getOpcode() == ISD::BRCOND)
        AddToWorkList(Use);
    }
  }

  return SDValue();
}

SDValue DAGCombiner::visitCTLZ(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  EVT VT = N->getValueType(0);

  // fold (ctlz c1) -> c2
  if (isa<ConstantSDNode>(N0))
    return DAG.getNode(ISD::CTLZ, N->getDebugLoc(), VT, N0);
  return SDValue();
}

SDValue DAGCombiner::visitCTLZ_ZERO_UNDEF(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  EVT VT = N->getValueType(0);

  // fold (ctlz_zero_undef c1) -> c2
  if (isa<ConstantSDNode>(N0))
    return DAG.getNode(ISD::CTLZ_ZERO_UNDEF, N->getDebugLoc(), VT, N0);
  return SDValue();
}

SDValue DAGCombiner::visitCTTZ(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  EVT VT = N->getValueType(0);

  // fold (cttz c1) -> c2
  if (isa<ConstantSDNode>(N0))
    return DAG.getNode(ISD::CTTZ, N->getDebugLoc(), VT, N0);
  return SDValue();
}

SDValue DAGCombiner::visitCTTZ_ZERO_UNDEF(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  EVT VT = N->getValueType(0);

  // fold (cttz_zero_undef c1) -> c2
  if (isa<ConstantSDNode>(N0))
    return DAG.getNode(ISD::CTTZ_ZERO_UNDEF, N->getDebugLoc(), VT, N0);
  return SDValue();
}

SDValue DAGCombiner::visitCTPOP(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  EVT VT = N->getValueType(0);

  // fold (ctpop c1) -> c2
  if (isa<ConstantSDNode>(N0))
    return DAG.getNode(ISD::CTPOP, N->getDebugLoc(), VT, N0);
  return SDValue();
}

SDValue DAGCombiner::visitSELECT(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  SDValue N1 = N->getOperand(1);
  SDValue N2 = N->getOperand(2);
  ConstantSDNode *N0C = dyn_cast<ConstantSDNode>(N0);
  ConstantSDNode *N1C = dyn_cast<ConstantSDNode>(N1);
  ConstantSDNode *N2C = dyn_cast<ConstantSDNode>(N2);
  EVT VT = N->getValueType(0);
  EVT VT0 = N0.getValueType();

  // fold (select C, X, X) -> X
  if (N1 == N2)
    return N1;
  // fold (select true, X, Y) -> X
  if (N0C && !N0C->isNullValue())
    return N1;
  // fold (select false, X, Y) -> Y
  if (N0C && N0C->isNullValue())
    return N2;
  // fold (select C, 1, X) -> (or C, X)
  if (VT == MVT::i1 && N1C && N1C->getAPIntValue() == 1)
    return DAG.getNode(ISD::OR, N->getDebugLoc(), VT, N0, N2);
  // fold (select C, 0, 1) -> (xor C, 1)
  if (VT.isInteger() &&
      (VT0 == MVT::i1 ||
       (VT0.isInteger() &&
        TLI.getBooleanContents(false) ==
        TargetLowering::ZeroOrOneBooleanContent)) &&
      N1C && N2C && N1C->isNullValue() && N2C->getAPIntValue() == 1) {
    SDValue XORNode;
    if (VT == VT0)
      return DAG.getNode(ISD::XOR, N->getDebugLoc(), VT0,
                         N0, DAG.getConstant(1, VT0));
    XORNode = DAG.getNode(ISD::XOR, N0.getDebugLoc(), VT0,
                          N0, DAG.getConstant(1, VT0));
    AddToWorkList(XORNode.getNode());
    if (VT.bitsGT(VT0))
      return DAG.getNode(ISD::ZERO_EXTEND, N->getDebugLoc(), VT, XORNode);
    return DAG.getNode(ISD::TRUNCATE, N->getDebugLoc(), VT, XORNode);
  }
  // fold (select C, 0, X) -> (and (not C), X)
  if (VT == VT0 && VT == MVT::i1 && N1C && N1C->isNullValue()) {
    SDValue NOTNode = DAG.getNOT(N0.getDebugLoc(), N0, VT);
    AddToWorkList(NOTNode.getNode());
    return DAG.getNode(ISD::AND, N->getDebugLoc(), VT, NOTNode, N2);
  }
  // fold (select C, X, 1) -> (or (not C), X)
  if (VT == VT0 && VT == MVT::i1 && N2C && N2C->getAPIntValue() == 1) {
    SDValue NOTNode = DAG.getNOT(N0.getDebugLoc(), N0, VT);
    AddToWorkList(NOTNode.getNode());
    return DAG.getNode(ISD::OR, N->getDebugLoc(), VT, NOTNode, N1);
  }
  // fold (select C, X, 0) -> (and C, X)
  if (VT == MVT::i1 && N2C && N2C->isNullValue())
    return DAG.getNode(ISD::AND, N->getDebugLoc(), VT, N0, N1);
  // fold (select X, X, Y) -> (or X, Y)
  // fold (select X, 1, Y) -> (or X, Y)
  if (VT == MVT::i1 && (N0 == N1 || (N1C && N1C->getAPIntValue() == 1)))
    return DAG.getNode(ISD::OR, N->getDebugLoc(), VT, N0, N2);
  // fold (select X, Y, X) -> (and X, Y)
  // fold (select X, Y, 0) -> (and X, Y)
  if (VT == MVT::i1 && (N0 == N2 || (N2C && N2C->getAPIntValue() == 0)))
    return DAG.getNode(ISD::AND, N->getDebugLoc(), VT, N0, N1);

  // If we can fold this based on the true/false value, do so.
  if (SimplifySelectOps(N, N1, N2))
    return SDValue(N, 0);  // Don't revisit N.

  // fold selects based on a setcc into other things, such as min/max/abs
  if (N0.getOpcode() == ISD::SETCC) {
    // FIXME:
    // Check against MVT::Other for SELECT_CC, which is a workaround for targets
    // having to say they don't support SELECT_CC on every type the DAG knows
    // about, since there is no way to mark an opcode illegal at all value types
    if (TLI.isOperationLegalOrCustom(ISD::SELECT_CC, MVT::Other) &&
        TLI.isOperationLegalOrCustom(ISD::SELECT_CC, VT))
      return DAG.getNode(ISD::SELECT_CC, N->getDebugLoc(), VT,
                         N0.getOperand(0), N0.getOperand(1),
                         N1, N2, N0.getOperand(2));
    return SimplifySelect(N->getDebugLoc(), N0, N1, N2);
  }

  return SDValue();
}

SDValue DAGCombiner::visitSELECT_CC(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  SDValue N1 = N->getOperand(1);
  SDValue N2 = N->getOperand(2);
  SDValue N3 = N->getOperand(3);
  SDValue N4 = N->getOperand(4);
  ISD::CondCode CC = cast<CondCodeSDNode>(N4)->get();

  // fold select_cc lhs, rhs, x, x, cc -> x
  if (N2 == N3)
    return N2;

  // Determine if the condition we're dealing with is constant
  SDValue SCC = SimplifySetCC(TLI.getSetCCResultType(N0.getValueType()),
                              N0, N1, CC, N->getDebugLoc(), false);
  if (SCC.getNode()) AddToWorkList(SCC.getNode());

  if (ConstantSDNode *SCCC = dyn_cast_or_null<ConstantSDNode>(SCC.getNode())) {
    if (!SCCC->isNullValue())
      return N2;    // cond always true -> true val
    else
      return N3;    // cond always false -> false val
  }

  // Fold to a simpler select_cc
  if (SCC.getNode() && SCC.getOpcode() == ISD::SETCC)
    return DAG.getNode(ISD::SELECT_CC, N->getDebugLoc(), N2.getValueType(),
                       SCC.getOperand(0), SCC.getOperand(1), N2, N3,
                       SCC.getOperand(2));

  // If we can fold this based on the true/false value, do so.
  if (SimplifySelectOps(N, N2, N3))
    return SDValue(N, 0);  // Don't revisit N.

  // fold select_cc into other things, such as min/max/abs
  return SimplifySelectCC(N->getDebugLoc(), N0, N1, N2, N3, CC);
}

SDValue DAGCombiner::visitSETCC(SDNode *N) {
  return SimplifySetCC(N->getValueType(0), N->getOperand(0), N->getOperand(1),
                       cast<CondCodeSDNode>(N->getOperand(2))->get(),
                       N->getDebugLoc());
}

// ExtendUsesToFormExtLoad - Trying to extend uses of a load to enable this:
// "fold ({s|z|a}ext (load x)) -> ({s|z|a}ext (truncate ({s|z|a}extload x)))"
// transformation. Returns true if extension are possible and the above
// mentioned transformation is profitable.
static bool ExtendUsesToFormExtLoad(SDNode *N, SDValue N0,
                                    unsigned ExtOpc,
                                    SmallVector<SDNode*, 4> &ExtendNodes,
                                    const TargetLowering &TLI) {
  bool HasCopyToRegUses = false;
  bool isTruncFree = TLI.isTruncateFree(N->getValueType(0), N0.getValueType());
  for (SDNode::use_iterator UI = N0.getNode()->use_begin(),
                            UE = N0.getNode()->use_end();
       UI != UE; ++UI) {
    SDNode *User = *UI;
    if (User == N)
      continue;
    if (UI.getUse().getResNo() != N0.getResNo())
      continue;
    // FIXME: Only extend SETCC N, N and SETCC N, c for now.
    if (ExtOpc != ISD::ANY_EXTEND && User->getOpcode() == ISD::SETCC) {
      ISD::CondCode CC = cast<CondCodeSDNode>(User->getOperand(2))->get();
      if (ExtOpc == ISD::ZERO_EXTEND && ISD::isSignedIntSetCC(CC))
        // Sign bits will be lost after a zext.
        return false;
      bool Add = false;
      for (unsigned i = 0; i != 2; ++i) {
        SDValue UseOp = User->getOperand(i);
        if (UseOp == N0)
          continue;
        if (!isa<ConstantSDNode>(UseOp))
          return false;
        Add = true;
      }
      if (Add)
        ExtendNodes.push_back(User);
      continue;
    }
    // If truncates aren't free and there are users we can't
    // extend, it isn't worthwhile.
    if (!isTruncFree)
      return false;
    // Remember if this value is live-out.
    if (User->getOpcode() == ISD::CopyToReg)
      HasCopyToRegUses = true;
  }

  if (HasCopyToRegUses) {
    bool BothLiveOut = false;
    for (SDNode::use_iterator UI = N->use_begin(), UE = N->use_end();
         UI != UE; ++UI) {
      SDUse &Use = UI.getUse();
      if (Use.getResNo() == 0 && Use.getUser()->getOpcode() == ISD::CopyToReg) {
        BothLiveOut = true;
        break;
      }
    }
    if (BothLiveOut)
      // Both unextended and extended values are live out. There had better be
      // a good reason for the transformation.
      return ExtendNodes.size();
  }
  return true;
}

void DAGCombiner::ExtendSetCCUses(SmallVector<SDNode*, 4> SetCCs,
                                  SDValue Trunc, SDValue ExtLoad, DebugLoc DL,
                                  ISD::NodeType ExtType) {
  // Extend SetCC uses if necessary.
  for (unsigned i = 0, e = SetCCs.size(); i != e; ++i) {
    SDNode *SetCC = SetCCs[i];
    SmallVector<SDValue, 4> Ops;

    for (unsigned j = 0; j != 2; ++j) {
      SDValue SOp = SetCC->getOperand(j);
      if (SOp == Trunc)
        Ops.push_back(ExtLoad);
      else
        Ops.push_back(DAG.getNode(ExtType, DL, ExtLoad->getValueType(0), SOp));
    }

    Ops.push_back(SetCC->getOperand(2));
    CombineTo(SetCC, DAG.getNode(ISD::SETCC, DL, SetCC->getValueType(0),
                                 &Ops[0], Ops.size()));
  }
}

SDValue DAGCombiner::visitSIGN_EXTEND(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  EVT VT = N->getValueType(0);

  // fold (sext c1) -> c1
  if (isa<ConstantSDNode>(N0))
    return DAG.getNode(ISD::SIGN_EXTEND, N->getDebugLoc(), VT, N0);

  // fold (sext (sext x)) -> (sext x)
  // fold (sext (aext x)) -> (sext x)
  if (N0.getOpcode() == ISD::SIGN_EXTEND || N0.getOpcode() == ISD::ANY_EXTEND)
    return DAG.getNode(ISD::SIGN_EXTEND, N->getDebugLoc(), VT,
                       N0.getOperand(0));

  if (N0.getOpcode() == ISD::TRUNCATE) {
    // fold (sext (truncate (load x))) -> (sext (smaller load x))
    // fold (sext (truncate (srl (load x), c))) -> (sext (smaller load (x+c/n)))
    SDValue NarrowLoad = ReduceLoadWidth(N0.getNode());
    if (NarrowLoad.getNode()) {
      SDNode* oye = N0.getNode()->getOperand(0).getNode();
      if (NarrowLoad.getNode() != N0.getNode()) {
        CombineTo(N0.getNode(), NarrowLoad);
        // CombineTo deleted the truncate, if needed, but not what's under it.
        AddToWorkList(oye);
      }
      return SDValue(N, 0);   // Return N so it doesn't get rechecked!
    }

    // See if the value being truncated is already sign extended.  If so, just
    // eliminate the trunc/sext pair.
    SDValue Op = N0.getOperand(0);
    unsigned OpBits   = Op.getValueType().getScalarType().getSizeInBits();
    unsigned MidBits  = N0.getValueType().getScalarType().getSizeInBits();
    unsigned DestBits = VT.getScalarType().getSizeInBits();
    unsigned NumSignBits = DAG.ComputeNumSignBits(Op);

    if (OpBits == DestBits) {
      // Op is i32, Mid is i8, and Dest is i32.  If Op has more than 24 sign
      // bits, it is already ready.
      if (NumSignBits > DestBits-MidBits)
        return Op;
    } else if (OpBits < DestBits) {
      // Op is i32, Mid is i8, and Dest is i64.  If Op has more than 24 sign
      // bits, just sext from i32.
      if (NumSignBits > OpBits-MidBits)
        return DAG.getNode(ISD::SIGN_EXTEND, N->getDebugLoc(), VT, Op);
    } else {
      // Op is i64, Mid is i8, and Dest is i32.  If Op has more than 56 sign
      // bits, just truncate to i32.
      if (NumSignBits > OpBits-MidBits)
        return DAG.getNode(ISD::TRUNCATE, N->getDebugLoc(), VT, Op);
    }

    // fold (sext (truncate x)) -> (sextinreg x).
    if (!LegalOperations || TLI.isOperationLegal(ISD::SIGN_EXTEND_INREG,
                                                 N0.getValueType())) {
      if (OpBits < DestBits)
        Op = DAG.getNode(ISD::ANY_EXTEND, N0.getDebugLoc(), VT, Op);
      else if (OpBits > DestBits)
        Op = DAG.getNode(ISD::TRUNCATE, N0.getDebugLoc(), VT, Op);
      return DAG.getNode(ISD::SIGN_EXTEND_INREG, N->getDebugLoc(), VT, Op,
                         DAG.getValueType(N0.getValueType()));
    }
  }

  // fold (sext (load x)) -> (sext (truncate (sextload x)))
  // None of the supported targets knows how to perform load and sign extend
  // on vectors in one instruction.  We only perform this transformation on
  // scalars.
  if (ISD::isNON_EXTLoad(N0.getNode()) && !VT.isVector() &&
      ((!LegalOperations && !cast<LoadSDNode>(N0)->isVolatile()) ||
       TLI.isLoadExtLegal(ISD::SEXTLOAD, N0.getValueType()))) {
    bool DoXform = true;
    SmallVector<SDNode*, 4> SetCCs;
    if (!N0.hasOneUse())
      DoXform = ExtendUsesToFormExtLoad(N, N0, ISD::SIGN_EXTEND, SetCCs, TLI);
    if (DoXform) {
      LoadSDNode *LN0 = cast<LoadSDNode>(N0);
      SDValue ExtLoad = DAG.getExtLoad(ISD::SEXTLOAD, N->getDebugLoc(), VT,
                                       LN0->getChain(),
                                       LN0->getBasePtr(), LN0->getPointerInfo(),
                                       N0.getValueType(),
                                       LN0->isVolatile(), LN0->isNonTemporal(),
                                       LN0->getAlignment());
      CombineTo(N, ExtLoad);
      SDValue Trunc = DAG.getNode(ISD::TRUNCATE, N0.getDebugLoc(),
                                  N0.getValueType(), ExtLoad);
      CombineTo(N0.getNode(), Trunc, ExtLoad.getValue(1));
      ExtendSetCCUses(SetCCs, Trunc, ExtLoad, N->getDebugLoc(),
                      ISD::SIGN_EXTEND);
      return SDValue(N, 0);   // Return N so it doesn't get rechecked!
    }
  }

  // fold (sext (sextload x)) -> (sext (truncate (sextload x)))
  // fold (sext ( extload x)) -> (sext (truncate (sextload x)))
  if ((ISD::isSEXTLoad(N0.getNode()) || ISD::isEXTLoad(N0.getNode())) &&
      ISD::isUNINDEXEDLoad(N0.getNode()) && N0.hasOneUse()) {
    LoadSDNode *LN0 = cast<LoadSDNode>(N0);
    EVT MemVT = LN0->getMemoryVT();
    if ((!LegalOperations && !LN0->isVolatile()) ||
        TLI.isLoadExtLegal(ISD::SEXTLOAD, MemVT)) {
      SDValue ExtLoad = DAG.getExtLoad(ISD::SEXTLOAD, N->getDebugLoc(), VT,
                                       LN0->getChain(),
                                       LN0->getBasePtr(), LN0->getPointerInfo(),
                                       MemVT,
                                       LN0->isVolatile(), LN0->isNonTemporal(),
                                       LN0->getAlignment());
      CombineTo(N, ExtLoad);
      CombineTo(N0.getNode(),
                DAG.getNode(ISD::TRUNCATE, N0.getDebugLoc(),
                            N0.getValueType(), ExtLoad),
                ExtLoad.getValue(1));
      return SDValue(N, 0);   // Return N so it doesn't get rechecked!
    }
  }

  // fold (sext (and/or/xor (load x), cst)) ->
  //      (and/or/xor (sextload x), (sext cst))
  if ((N0.getOpcode() == ISD::AND || N0.getOpcode() == ISD::OR ||
       N0.getOpcode() == ISD::XOR) &&
      isa<LoadSDNode>(N0.getOperand(0)) &&
      N0.getOperand(1).getOpcode() == ISD::Constant &&
      TLI.isLoadExtLegal(ISD::SEXTLOAD, N0.getValueType()) &&
      (!LegalOperations && TLI.isOperationLegal(N0.getOpcode(), VT))) {
    LoadSDNode *LN0 = cast<LoadSDNode>(N0.getOperand(0));
    if (LN0->getExtensionType() != ISD::ZEXTLOAD) {
      bool DoXform = true;
      SmallVector<SDNode*, 4> SetCCs;
      if (!N0.hasOneUse())
        DoXform = ExtendUsesToFormExtLoad(N, N0.getOperand(0), ISD::SIGN_EXTEND,
                                          SetCCs, TLI);
      if (DoXform) {
        SDValue ExtLoad = DAG.getExtLoad(ISD::SEXTLOAD, LN0->getDebugLoc(), VT,
                                         LN0->getChain(), LN0->getBasePtr(),
                                         LN0->getPointerInfo(),
                                         LN0->getMemoryVT(),
                                         LN0->isVolatile(),
                                         LN0->isNonTemporal(),
                                         LN0->getAlignment());
        APInt Mask = cast<ConstantSDNode>(N0.getOperand(1))->getAPIntValue();
        Mask = Mask.sext(VT.getSizeInBits());
        SDValue And = DAG.getNode(N0.getOpcode(), N->getDebugLoc(), VT,
                                  ExtLoad, DAG.getConstant(Mask, VT));
        SDValue Trunc = DAG.getNode(ISD::TRUNCATE,
                                    N0.getOperand(0).getDebugLoc(),
                                    N0.getOperand(0).getValueType(), ExtLoad);
        CombineTo(N, And);
        CombineTo(N0.getOperand(0).getNode(), Trunc, ExtLoad.getValue(1));
        ExtendSetCCUses(SetCCs, Trunc, ExtLoad, N->getDebugLoc(),
                        ISD::SIGN_EXTEND);
        return SDValue(N, 0);   // Return N so it doesn't get rechecked!
      }
    }
  }

  if (N0.getOpcode() == ISD::SETCC) {
    // sext(setcc) -> sext_in_reg(vsetcc) for vectors.
    // Only do this before legalize for now.
    if (VT.isVector() && !LegalOperations) {
      EVT N0VT = N0.getOperand(0).getValueType();
      // On some architectures (such as SSE/NEON/etc) the SETCC result type is
      // of the same size as the compared operands. Only optimize sext(setcc())
      // if this is the case.
      EVT SVT = TLI.getSetCCResultType(N0VT);

      // We know that the # elements of the results is the same as the
      // # elements of the compare (and the # elements of the compare result
      // for that matter).  Check to see that they are the same size.  If so,
      // we know that the element size of the sext'd result matches the
      // element size of the compare operands.
      if (VT.getSizeInBits() == SVT.getSizeInBits())
        return DAG.getSetCC(N->getDebugLoc(), VT, N0.getOperand(0),
                             N0.getOperand(1),
                             cast<CondCodeSDNode>(N0.getOperand(2))->get());
      // If the desired elements are smaller or larger than the source
      // elements we can use a matching integer vector type and then
      // truncate/sign extend
      EVT MatchingElementType =
        EVT::getIntegerVT(*DAG.getContext(),
                          N0VT.getScalarType().getSizeInBits());
      EVT MatchingVectorType =
        EVT::getVectorVT(*DAG.getContext(), MatchingElementType,
                         N0VT.getVectorNumElements());

      if (SVT == MatchingVectorType) {
        SDValue VsetCC = DAG.getSetCC(N->getDebugLoc(), MatchingVectorType,
                               N0.getOperand(0), N0.getOperand(1),
                               cast<CondCodeSDNode>(N0.getOperand(2))->get());
        return DAG.getSExtOrTrunc(VsetCC, N->getDebugLoc(), VT);
      }
    }

    // sext(setcc x, y, cc) -> (select_cc x, y, -1, 0, cc)
    unsigned ElementWidth = VT.getScalarType().getSizeInBits();
    SDValue NegOne =
      DAG.getConstant(APInt::getAllOnesValue(ElementWidth), VT);
    SDValue SCC =
      SimplifySelectCC(N->getDebugLoc(), N0.getOperand(0), N0.getOperand(1),
                       NegOne, DAG.getConstant(0, VT),
                       cast<CondCodeSDNode>(N0.getOperand(2))->get(), true);
    if (SCC.getNode()) return SCC;
    if (!LegalOperations ||
        TLI.isOperationLegal(ISD::SETCC, TLI.getSetCCResultType(VT)))
      return DAG.getNode(ISD::SELECT, N->getDebugLoc(), VT,
                         DAG.getSetCC(N->getDebugLoc(),
                                      TLI.getSetCCResultType(VT),
                                      N0.getOperand(0), N0.getOperand(1),
                                 cast<CondCodeSDNode>(N0.getOperand(2))->get()),
                         NegOne, DAG.getConstant(0, VT));
  }

  // fold (sext x) -> (zext x) if the sign bit is known zero.
  if ((!LegalOperations || TLI.isOperationLegal(ISD::ZERO_EXTEND, VT)) &&
      DAG.SignBitIsZero(N0))
    return DAG.getNode(ISD::ZERO_EXTEND, N->getDebugLoc(), VT, N0);

  return SDValue();
}

// isTruncateOf - If N is a truncate of some other value, return true, record
// the value being truncated in Op and which of Op's bits are zero in KnownZero.
// This function computes KnownZero to avoid a duplicated call to
// ComputeMaskedBits in the caller.
static bool isTruncateOf(SelectionDAG &DAG, SDValue N, SDValue &Op,
                         APInt &KnownZero) {
  APInt KnownOne;
  if (N->getOpcode() == ISD::TRUNCATE) {
    Op = N->getOperand(0);
    DAG.ComputeMaskedBits(Op, KnownZero, KnownOne);
    return true;
  }

  if (N->getOpcode() != ISD::SETCC || N->getValueType(0) != MVT::i1 ||
      cast<CondCodeSDNode>(N->getOperand(2))->get() != ISD::SETNE)
    return false;

  SDValue Op0 = N->getOperand(0);
  SDValue Op1 = N->getOperand(1);
  assert(Op0.getValueType() == Op1.getValueType());

  ConstantSDNode *COp0 = dyn_cast<ConstantSDNode>(Op0);
  ConstantSDNode *COp1 = dyn_cast<ConstantSDNode>(Op1);
  if (COp0 && COp0->isNullValue())
    Op = Op1;
  else if (COp1 && COp1->isNullValue())
    Op = Op0;
  else
    return false;

  DAG.ComputeMaskedBits(Op, KnownZero, KnownOne);

  if (!(KnownZero | APInt(Op.getValueSizeInBits(), 1)).isAllOnesValue())
    return false;

  return true;
}

SDValue DAGCombiner::visitZERO_EXTEND(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  EVT VT = N->getValueType(0);

  // fold (zext c1) -> c1
  if (isa<ConstantSDNode>(N0))
    return DAG.getNode(ISD::ZERO_EXTEND, N->getDebugLoc(), VT, N0);
  // fold (zext (zext x)) -> (zext x)
  // fold (zext (aext x)) -> (zext x)
  if (N0.getOpcode() == ISD::ZERO_EXTEND || N0.getOpcode() == ISD::ANY_EXTEND)
    return DAG.getNode(ISD::ZERO_EXTEND, N->getDebugLoc(), VT,
                       N0.getOperand(0));

  // fold (zext (truncate x)) -> (zext x) or
  //      (zext (truncate x)) -> (truncate x)
  // This is valid when the truncated bits of x are already zero.
  // FIXME: We should extend this to work for vectors too.
  SDValue Op;
  APInt KnownZero;
  if (!VT.isVector() && isTruncateOf(DAG, N0, Op, KnownZero)) {
    APInt TruncatedBits =
      (Op.getValueSizeInBits() == N0.getValueSizeInBits()) ?
      APInt(Op.getValueSizeInBits(), 0) :
      APInt::getBitsSet(Op.getValueSizeInBits(),
                        N0.getValueSizeInBits(),
                        std::min(Op.getValueSizeInBits(),
                                 VT.getSizeInBits()));
    if (TruncatedBits == (KnownZero & TruncatedBits)) {
      if (VT.bitsGT(Op.getValueType()))
        return DAG.getNode(ISD::ZERO_EXTEND, N->getDebugLoc(), VT, Op);
      if (VT.bitsLT(Op.getValueType()))
        return DAG.getNode(ISD::TRUNCATE, N->getDebugLoc(), VT, Op);

      return Op;
    }
  }

  // fold (zext (truncate (load x))) -> (zext (smaller load x))
  // fold (zext (truncate (srl (load x), c))) -> (zext (small load (x+c/n)))
  if (N0.getOpcode() == ISD::TRUNCATE) {
    SDValue NarrowLoad = ReduceLoadWidth(N0.getNode());
    if (NarrowLoad.getNode()) {
      SDNode* oye = N0.getNode()->getOperand(0).getNode();
      if (NarrowLoad.getNode() != N0.getNode()) {
        CombineTo(N0.getNode(), NarrowLoad);
        // CombineTo deleted the truncate, if needed, but not what's under it.
        AddToWorkList(oye);
      }
      return SDValue(N, 0);   // Return N so it doesn't get rechecked!
    }
  }

  // fold (zext (truncate x)) -> (and x, mask)
  if (N0.getOpcode() == ISD::TRUNCATE &&
      (!LegalOperations || TLI.isOperationLegal(ISD::AND, VT))) {

    // fold (zext (truncate (load x))) -> (zext (smaller load x))
    // fold (zext (truncate (srl (load x), c))) -> (zext (smaller load (x+c/n)))
    SDValue NarrowLoad = ReduceLoadWidth(N0.getNode());
    if (NarrowLoad.getNode()) {
      SDNode* oye = N0.getNode()->getOperand(0).getNode();
      if (NarrowLoad.getNode() != N0.getNode()) {
        CombineTo(N0.getNode(), NarrowLoad);
        // CombineTo deleted the truncate, if needed, but not what's under it.
        AddToWorkList(oye);
      }
      return SDValue(N, 0);   // Return N so it doesn't get rechecked!
    }

    SDValue Op = N0.getOperand(0);
    if (Op.getValueType().bitsLT(VT)) {
      Op = DAG.getNode(ISD::ANY_EXTEND, N->getDebugLoc(), VT, Op);
      AddToWorkList(Op.getNode());
    } else if (Op.getValueType().bitsGT(VT)) {
      Op = DAG.getNode(ISD::TRUNCATE, N->getDebugLoc(), VT, Op);
      AddToWorkList(Op.getNode());
    }
    return DAG.getZeroExtendInReg(Op, N->getDebugLoc(),
                                  N0.getValueType().getScalarType());
  }

  // Fold (zext (and (trunc x), cst)) -> (and x, cst),
  // if either of the casts is not free.
  if (N0.getOpcode() == ISD::AND &&
      N0.getOperand(0).getOpcode() == ISD::TRUNCATE &&
      N0.getOperand(1).getOpcode() == ISD::Constant &&
      (!TLI.isTruncateFree(N0.getOperand(0).getOperand(0).getValueType(),
                           N0.getValueType()) ||
       !TLI.isZExtFree(N0.getValueType(), VT))) {
    SDValue X = N0.getOperand(0).getOperand(0);
    if (X.getValueType().bitsLT(VT)) {
      X = DAG.getNode(ISD::ANY_EXTEND, X.getDebugLoc(), VT, X);
    } else if (X.getValueType().bitsGT(VT)) {
      X = DAG.getNode(ISD::TRUNCATE, X.getDebugLoc(), VT, X);
    }
    APInt Mask = cast<ConstantSDNode>(N0.getOperand(1))->getAPIntValue();
    Mask = Mask.zext(VT.getSizeInBits());
    return DAG.getNode(ISD::AND, N->getDebugLoc(), VT,
                       X, DAG.getConstant(Mask, VT));
  }

  // fold (zext (load x)) -> (zext (truncate (zextload x)))
  // None of the supported targets knows how to perform load and vector_zext
  // on vectors in one instruction.  We only perform this transformation on
  // scalars.
  if (ISD::isNON_EXTLoad(N0.getNode()) && !VT.isVector() &&
      ((!LegalOperations && !cast<LoadSDNode>(N0)->isVolatile()) ||
       TLI.isLoadExtLegal(ISD::ZEXTLOAD, N0.getValueType()))) {
    bool DoXform = true;
    SmallVector<SDNode*, 4> SetCCs;
    if (!N0.hasOneUse())
      DoXform = ExtendUsesToFormExtLoad(N, N0, ISD::ZERO_EXTEND, SetCCs, TLI);
    if (DoXform) {
      LoadSDNode *LN0 = cast<LoadSDNode>(N0);
      SDValue ExtLoad = DAG.getExtLoad(ISD::ZEXTLOAD, N->getDebugLoc(), VT,
                                       LN0->getChain(),
                                       LN0->getBasePtr(), LN0->getPointerInfo(),
                                       N0.getValueType(),
                                       LN0->isVolatile(), LN0->isNonTemporal(),
                                       LN0->getAlignment());
      CombineTo(N, ExtLoad);
      SDValue Trunc = DAG.getNode(ISD::TRUNCATE, N0.getDebugLoc(),
                                  N0.getValueType(), ExtLoad);
      CombineTo(N0.getNode(), Trunc, ExtLoad.getValue(1));

      ExtendSetCCUses(SetCCs, Trunc, ExtLoad, N->getDebugLoc(),
                      ISD::ZERO_EXTEND);
      return SDValue(N, 0);   // Return N so it doesn't get rechecked!
    }
  }

  // fold (zext (and/or/xor (load x), cst)) ->
  //      (and/or/xor (zextload x), (zext cst))
  if ((N0.getOpcode() == ISD::AND || N0.getOpcode() == ISD::OR ||
       N0.getOpcode() == ISD::XOR) &&
      isa<LoadSDNode>(N0.getOperand(0)) &&
      N0.getOperand(1).getOpcode() == ISD::Constant &&
      TLI.isLoadExtLegal(ISD::ZEXTLOAD, N0.getValueType()) &&
      (!LegalOperations && TLI.isOperationLegal(N0.getOpcode(), VT))) {
    LoadSDNode *LN0 = cast<LoadSDNode>(N0.getOperand(0));
    if (LN0->getExtensionType() != ISD::SEXTLOAD) {
      bool DoXform = true;
      SmallVector<SDNode*, 4> SetCCs;
      if (!N0.hasOneUse())
        DoXform = ExtendUsesToFormExtLoad(N, N0.getOperand(0), ISD::ZERO_EXTEND,
                                          SetCCs, TLI);
      if (DoXform) {
        SDValue ExtLoad = DAG.getExtLoad(ISD::ZEXTLOAD, LN0->getDebugLoc(), VT,
                                         LN0->getChain(), LN0->getBasePtr(),
                                         LN0->getPointerInfo(),
                                         LN0->getMemoryVT(),
                                         LN0->isVolatile(),
                                         LN0->isNonTemporal(),
                                         LN0->getAlignment());
        APInt Mask = cast<ConstantSDNode>(N0.getOperand(1))->getAPIntValue();
        Mask = Mask.zext(VT.getSizeInBits());
        SDValue And = DAG.getNode(N0.getOpcode(), N->getDebugLoc(), VT,
                                  ExtLoad, DAG.getConstant(Mask, VT));
        SDValue Trunc = DAG.getNode(ISD::TRUNCATE,
                                    N0.getOperand(0).getDebugLoc(),
                                    N0.getOperand(0).getValueType(), ExtLoad);
        CombineTo(N, And);
        CombineTo(N0.getOperand(0).getNode(), Trunc, ExtLoad.getValue(1));
        ExtendSetCCUses(SetCCs, Trunc, ExtLoad, N->getDebugLoc(),
                        ISD::ZERO_EXTEND);
        return SDValue(N, 0);   // Return N so it doesn't get rechecked!
      }
    }
  }

  // fold (zext (zextload x)) -> (zext (truncate (zextload x)))
  // fold (zext ( extload x)) -> (zext (truncate (zextload x)))
  if ((ISD::isZEXTLoad(N0.getNode()) || ISD::isEXTLoad(N0.getNode())) &&
      ISD::isUNINDEXEDLoad(N0.getNode()) && N0.hasOneUse()) {
    LoadSDNode *LN0 = cast<LoadSDNode>(N0);
    EVT MemVT = LN0->getMemoryVT();
    if ((!LegalOperations && !LN0->isVolatile()) ||
        TLI.isLoadExtLegal(ISD::ZEXTLOAD, MemVT)) {
      SDValue ExtLoad = DAG.getExtLoad(ISD::ZEXTLOAD, N->getDebugLoc(), VT,
                                       LN0->getChain(),
                                       LN0->getBasePtr(), LN0->getPointerInfo(),
                                       MemVT,
                                       LN0->isVolatile(), LN0->isNonTemporal(),
                                       LN0->getAlignment());
      CombineTo(N, ExtLoad);
      CombineTo(N0.getNode(),
                DAG.getNode(ISD::TRUNCATE, N0.getDebugLoc(), N0.getValueType(),
                            ExtLoad),
                ExtLoad.getValue(1));
      return SDValue(N, 0);   // Return N so it doesn't get rechecked!
    }
  }

  if (N0.getOpcode() == ISD::SETCC) {
    if (!LegalOperations && VT.isVector()) {
      // zext(setcc) -> (and (vsetcc), (1, 1, ...) for vectors.
      // Only do this before legalize for now.
      EVT N0VT = N0.getOperand(0).getValueType();
      EVT EltVT = VT.getVectorElementType();
      SmallVector<SDValue,8> OneOps(VT.getVectorNumElements(),
                                    DAG.getConstant(1, EltVT));
      if (VT.getSizeInBits() == N0VT.getSizeInBits())
        // We know that the # elements of the results is the same as the
        // # elements of the compare (and the # elements of the compare result
        // for that matter).  Check to see that they are the same size.  If so,
        // we know that the element size of the sext'd result matches the
        // element size of the compare operands.
        return DAG.getNode(ISD::AND, N->getDebugLoc(), VT,
                           DAG.getSetCC(N->getDebugLoc(), VT, N0.getOperand(0),
                                         N0.getOperand(1),
                                 cast<CondCodeSDNode>(N0.getOperand(2))->get()),
                           DAG.getNode(ISD::BUILD_VECTOR, N->getDebugLoc(), VT,
                                       &OneOps[0], OneOps.size()));

      // If the desired elements are smaller or larger than the source
      // elements we can use a matching integer vector type and then
      // truncate/sign extend
      EVT MatchingElementType =
        EVT::getIntegerVT(*DAG.getContext(),
                          N0VT.getScalarType().getSizeInBits());
      EVT MatchingVectorType =
        EVT::getVectorVT(*DAG.getContext(), MatchingElementType,
                         N0VT.getVectorNumElements());
      SDValue VsetCC =
        DAG.getSetCC(N->getDebugLoc(), MatchingVectorType, N0.getOperand(0),
                      N0.getOperand(1),
                      cast<CondCodeSDNode>(N0.getOperand(2))->get());
      return DAG.getNode(ISD::AND, N->getDebugLoc(), VT,
                         DAG.getSExtOrTrunc(VsetCC, N->getDebugLoc(), VT),
                         DAG.getNode(ISD::BUILD_VECTOR, N->getDebugLoc(), VT,
                                     &OneOps[0], OneOps.size()));
    }

    // zext(setcc x,y,cc) -> select_cc x, y, 1, 0, cc
    SDValue SCC =
      SimplifySelectCC(N->getDebugLoc(), N0.getOperand(0), N0.getOperand(1),
                       DAG.getConstant(1, VT), DAG.getConstant(0, VT),
                       cast<CondCodeSDNode>(N0.getOperand(2))->get(), true);
    if (SCC.getNode()) return SCC;
  }

  // (zext (shl (zext x), cst)) -> (shl (zext x), cst)
  if ((N0.getOpcode() == ISD::SHL || N0.getOpcode() == ISD::SRL) &&
      isa<ConstantSDNode>(N0.getOperand(1)) &&
      N0.getOperand(0).getOpcode() == ISD::ZERO_EXTEND &&
      N0.hasOneUse()) {
    SDValue ShAmt = N0.getOperand(1);
    unsigned ShAmtVal = cast<ConstantSDNode>(ShAmt)->getZExtValue();
    if (N0.getOpcode() == ISD::SHL) {
      SDValue InnerZExt = N0.getOperand(0);
      // If the original shl may be shifting out bits, do not perform this
      // transformation.
      unsigned KnownZeroBits = InnerZExt.getValueType().getSizeInBits() -
        InnerZExt.getOperand(0).getValueType().getSizeInBits();
      if (ShAmtVal > KnownZeroBits)
        return SDValue();
    }

    DebugLoc DL = N->getDebugLoc();

    // Ensure that the shift amount is wide enough for the shifted value.
    if (VT.getSizeInBits() >= 256)
      ShAmt = DAG.getNode(ISD::ZERO_EXTEND, DL, MVT::i32, ShAmt);

    return DAG.getNode(N0.getOpcode(), DL, VT,
                       DAG.getNode(ISD::ZERO_EXTEND, DL, VT, N0.getOperand(0)),
                       ShAmt);
  }

  return SDValue();
}

SDValue DAGCombiner::visitANY_EXTEND(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  EVT VT = N->getValueType(0);

  // fold (aext c1) -> c1
  if (isa<ConstantSDNode>(N0))
    return DAG.getNode(ISD::ANY_EXTEND, N->getDebugLoc(), VT, N0);
  // fold (aext (aext x)) -> (aext x)
  // fold (aext (zext x)) -> (zext x)
  // fold (aext (sext x)) -> (sext x)
  if (N0.getOpcode() == ISD::ANY_EXTEND  ||
      N0.getOpcode() == ISD::ZERO_EXTEND ||
      N0.getOpcode() == ISD::SIGN_EXTEND)
    return DAG.getNode(N0.getOpcode(), N->getDebugLoc(), VT, N0.getOperand(0));

  // fold (aext (truncate (load x))) -> (aext (smaller load x))
  // fold (aext (truncate (srl (load x), c))) -> (aext (small load (x+c/n)))
  if (N0.getOpcode() == ISD::TRUNCATE) {
    SDValue NarrowLoad = ReduceLoadWidth(N0.getNode());
    if (NarrowLoad.getNode()) {
      SDNode* oye = N0.getNode()->getOperand(0).getNode();
      if (NarrowLoad.getNode() != N0.getNode()) {
        CombineTo(N0.getNode(), NarrowLoad);
        // CombineTo deleted the truncate, if needed, but not what's under it.
        AddToWorkList(oye);
      }
      return SDValue(N, 0);   // Return N so it doesn't get rechecked!
    }
  }

  // fold (aext (truncate x))
  if (N0.getOpcode() == ISD::TRUNCATE) {
    SDValue TruncOp = N0.getOperand(0);
    if (TruncOp.getValueType() == VT)
      return TruncOp; // x iff x size == zext size.
    if (TruncOp.getValueType().bitsGT(VT))
      return DAG.getNode(ISD::TRUNCATE, N->getDebugLoc(), VT, TruncOp);
    return DAG.getNode(ISD::ANY_EXTEND, N->getDebugLoc(), VT, TruncOp);
  }

  // Fold (aext (and (trunc x), cst)) -> (and x, cst)
  // if the trunc is not free.
  if (N0.getOpcode() == ISD::AND &&
      N0.getOperand(0).getOpcode() == ISD::TRUNCATE &&
      N0.getOperand(1).getOpcode() == ISD::Constant &&
      !TLI.isTruncateFree(N0.getOperand(0).getOperand(0).getValueType(),
                          N0.getValueType())) {
    SDValue X = N0.getOperand(0).getOperand(0);
    if (X.getValueType().bitsLT(VT)) {
      X = DAG.getNode(ISD::ANY_EXTEND, N->getDebugLoc(), VT, X);
    } else if (X.getValueType().bitsGT(VT)) {
      X = DAG.getNode(ISD::TRUNCATE, N->getDebugLoc(), VT, X);
    }
    APInt Mask = cast<ConstantSDNode>(N0.getOperand(1))->getAPIntValue();
    Mask = Mask.zext(VT.getSizeInBits());
    return DAG.getNode(ISD::AND, N->getDebugLoc(), VT,
                       X, DAG.getConstant(Mask, VT));
  }

  // fold (aext (load x)) -> (aext (truncate (extload x)))
  // None of the supported targets knows how to perform load and any_ext
  // on vectors in one instruction.  We only perform this transformation on
  // scalars.
  if (ISD::isNON_EXTLoad(N0.getNode()) && !VT.isVector() &&
      ((!LegalOperations && !cast<LoadSDNode>(N0)->isVolatile()) ||
       TLI.isLoadExtLegal(ISD::EXTLOAD, N0.getValueType()))) {
    bool DoXform = true;
    SmallVector<SDNode*, 4> SetCCs;
    if (!N0.hasOneUse())
      DoXform = ExtendUsesToFormExtLoad(N, N0, ISD::ANY_EXTEND, SetCCs, TLI);
    if (DoXform) {
      LoadSDNode *LN0 = cast<LoadSDNode>(N0);
      SDValue ExtLoad = DAG.getExtLoad(ISD::EXTLOAD, N->getDebugLoc(), VT,
                                       LN0->getChain(),
                                       LN0->getBasePtr(), LN0->getPointerInfo(),
                                       N0.getValueType(),
                                       LN0->isVolatile(), LN0->isNonTemporal(),
                                       LN0->getAlignment());
      CombineTo(N, ExtLoad);
      SDValue Trunc = DAG.getNode(ISD::TRUNCATE, N0.getDebugLoc(),
                                  N0.getValueType(), ExtLoad);
      CombineTo(N0.getNode(), Trunc, ExtLoad.getValue(1));
      ExtendSetCCUses(SetCCs, Trunc, ExtLoad, N->getDebugLoc(),
                      ISD::ANY_EXTEND);
      return SDValue(N, 0);   // Return N so it doesn't get rechecked!
    }
  }

  // fold (aext (zextload x)) -> (aext (truncate (zextload x)))
  // fold (aext (sextload x)) -> (aext (truncate (sextload x)))
  // fold (aext ( extload x)) -> (aext (truncate (extload  x)))
  if (N0.getOpcode() == ISD::LOAD &&
      !ISD::isNON_EXTLoad(N0.getNode()) && ISD::isUNINDEXEDLoad(N0.getNode()) &&
      N0.hasOneUse()) {
    LoadSDNode *LN0 = cast<LoadSDNode>(N0);
    EVT MemVT = LN0->getMemoryVT();
    SDValue ExtLoad = DAG.getExtLoad(LN0->getExtensionType(), N->getDebugLoc(),
                                     VT, LN0->getChain(), LN0->getBasePtr(),
                                     LN0->getPointerInfo(), MemVT,
                                     LN0->isVolatile(), LN0->isNonTemporal(),
                                     LN0->getAlignment());
    CombineTo(N, ExtLoad);
    CombineTo(N0.getNode(),
              DAG.getNode(ISD::TRUNCATE, N0.getDebugLoc(),
                          N0.getValueType(), ExtLoad),
              ExtLoad.getValue(1));
    return SDValue(N, 0);   // Return N so it doesn't get rechecked!
  }

  if (N0.getOpcode() == ISD::SETCC) {
    // aext(setcc) -> sext_in_reg(vsetcc) for vectors.
    // Only do this before legalize for now.
    if (VT.isVector() && !LegalOperations) {
      EVT N0VT = N0.getOperand(0).getValueType();
        // We know that the # elements of the results is the same as the
        // # elements of the compare (and the # elements of the compare result
        // for that matter).  Check to see that they are the same size.  If so,
        // we know that the element size of the sext'd result matches the
        // element size of the compare operands.
      if (VT.getSizeInBits() == N0VT.getSizeInBits())
        return DAG.getSetCC(N->getDebugLoc(), VT, N0.getOperand(0),
                             N0.getOperand(1),
                             cast<CondCodeSDNode>(N0.getOperand(2))->get());
      // If the desired elements are smaller or larger than the source
      // elements we can use a matching integer vector type and then
      // truncate/sign extend
      else {
        EVT MatchingElementType =
          EVT::getIntegerVT(*DAG.getContext(),
                            N0VT.getScalarType().getSizeInBits());
        EVT MatchingVectorType =
          EVT::getVectorVT(*DAG.getContext(), MatchingElementType,
                           N0VT.getVectorNumElements());
        SDValue VsetCC =
          DAG.getSetCC(N->getDebugLoc(), MatchingVectorType, N0.getOperand(0),
                        N0.getOperand(1),
                        cast<CondCodeSDNode>(N0.getOperand(2))->get());
        return DAG.getSExtOrTrunc(VsetCC, N->getDebugLoc(), VT);
      }
    }

    // aext(setcc x,y,cc) -> select_cc x, y, 1, 0, cc
    SDValue SCC =
      SimplifySelectCC(N->getDebugLoc(), N0.getOperand(0), N0.getOperand(1),
                       DAG.getConstant(1, VT), DAG.getConstant(0, VT),
                       cast<CondCodeSDNode>(N0.getOperand(2))->get(), true);
    if (SCC.getNode())
      return SCC;
  }

  return SDValue();
}

/// GetDemandedBits - See if the specified operand can be simplified with the
/// knowledge that only the bits specified by Mask are used.  If so, return the
/// simpler operand, otherwise return a null SDValue.
SDValue DAGCombiner::GetDemandedBits(SDValue V, const APInt &Mask) {
  switch (V.getOpcode()) {
  default: break;
  case ISD::Constant: {
    const ConstantSDNode *CV = cast<ConstantSDNode>(V.getNode());
    assert(CV != 0 && "Const value should be ConstSDNode.");
    const APInt &CVal = CV->getAPIntValue();
    APInt NewVal = CVal & Mask;
    if (NewVal != CVal) {
      return DAG.getConstant(NewVal, V.getValueType());
    }
    break;
  }
  case ISD::OR:
  case ISD::XOR:
    // If the LHS or RHS don't contribute bits to the or, drop them.
    if (DAG.MaskedValueIsZero(V.getOperand(0), Mask))
      return V.getOperand(1);
    if (DAG.MaskedValueIsZero(V.getOperand(1), Mask))
      return V.getOperand(0);
    break;
  case ISD::SRL:
    // Only look at single-use SRLs.
    if (!V.getNode()->hasOneUse())
      break;
    if (ConstantSDNode *RHSC = dyn_cast<ConstantSDNode>(V.getOperand(1))) {
      // See if we can recursively simplify the LHS.
      unsigned Amt = RHSC->getZExtValue();

      // Watch out for shift count overflow though.
      if (Amt >= Mask.getBitWidth()) break;
      APInt NewMask = Mask << Amt;
      SDValue SimplifyLHS = GetDemandedBits(V.getOperand(0), NewMask);
      if (SimplifyLHS.getNode())
        return DAG.getNode(ISD::SRL, V.getDebugLoc(), V.getValueType(),
                           SimplifyLHS, V.getOperand(1));
    }
  }
  return SDValue();
}

/// ReduceLoadWidth - If the result of a wider load is shifted to right of N
/// bits and then truncated to a narrower type and where N is a multiple
/// of number of bits of the narrower type, transform it to a narrower load
/// from address + N / num of bits of new type. If the result is to be
/// extended, also fold the extension to form a extending load.
SDValue DAGCombiner::ReduceLoadWidth(SDNode *N) {
  unsigned Opc = N->getOpcode();

  ISD::LoadExtType ExtType = ISD::NON_EXTLOAD;
  SDValue N0 = N->getOperand(0);
  EVT VT = N->getValueType(0);
  EVT ExtVT = VT;

  // This transformation isn't valid for vector loads.
  if (VT.isVector())
    return SDValue();

  // Special case: SIGN_EXTEND_INREG is basically truncating to ExtVT then
  // extended to VT.
  if (Opc == ISD::SIGN_EXTEND_INREG) {
    ExtType = ISD::SEXTLOAD;
    ExtVT = cast<VTSDNode>(N->getOperand(1))->getVT();
  } else if (Opc == ISD::SRL) {
    // Another special-case: SRL is basically zero-extending a narrower value.
    ExtType = ISD::ZEXTLOAD;
    N0 = SDValue(N, 0);
    ConstantSDNode *N01 = dyn_cast<ConstantSDNode>(N0.getOperand(1));
    if (!N01) return SDValue();
    ExtVT = EVT::getIntegerVT(*DAG.getContext(),
                              VT.getSizeInBits() - N01->getZExtValue());
  }
  if (LegalOperations && !TLI.isLoadExtLegal(ExtType, ExtVT))
    return SDValue();

  unsigned EVTBits = ExtVT.getSizeInBits();

  // Do not generate loads of non-round integer types since these can
  // be expensive (and would be wrong if the type is not byte sized).
  if (!ExtVT.isRound())
    return SDValue();

  unsigned ShAmt = 0;
  if (N0.getOpcode() == ISD::SRL && N0.hasOneUse()) {
    if (ConstantSDNode *N01 = dyn_cast<ConstantSDNode>(N0.getOperand(1))) {
      ShAmt = N01->getZExtValue();
      // Is the shift amount a multiple of size of VT?
      if ((ShAmt & (EVTBits-1)) == 0) {
        N0 = N0.getOperand(0);
        // Is the load width a multiple of size of VT?
        if ((N0.getValueType().getSizeInBits() & (EVTBits-1)) != 0)
          return SDValue();
      }

      // At this point, we must have a load or else we can't do the transform.
      if (!isa<LoadSDNode>(N0)) return SDValue();

      // If the shift amount is larger than the input type then we're not
      // accessing any of the loaded bytes.  If the load was a zextload/extload
      // then the result of the shift+trunc is zero/undef (handled elsewhere).
      // If the load was a sextload then the result is a splat of the sign bit
      // of the extended byte.  This is not worth optimizing for.
      if (ShAmt >= cast<LoadSDNode>(N0)->getMemoryVT().getSizeInBits())
        return SDValue();
    }
  }

  // If the load is shifted left (and the result isn't shifted back right),
  // we can fold the truncate through the shift.
  unsigned ShLeftAmt = 0;
  if (ShAmt == 0 && N0.getOpcode() == ISD::SHL && N0.hasOneUse() &&
      ExtVT == VT && TLI.isNarrowingProfitable(N0.getValueType(), VT)) {
    if (ConstantSDNode *N01 = dyn_cast<ConstantSDNode>(N0.getOperand(1))) {
      ShLeftAmt = N01->getZExtValue();
      N0 = N0.getOperand(0);
    }
  }

  // If we haven't found a load, we can't narrow it.  Don't transform one with
  // multiple uses, this would require adding a new load.
  if (!isa<LoadSDNode>(N0) || !N0.hasOneUse() ||
      // Don't change the width of a volatile load.
      cast<LoadSDNode>(N0)->isVolatile())
    return SDValue();

  // Verify that we are actually reducing a load width here.
  if (cast<LoadSDNode>(N0)->getMemoryVT().getSizeInBits() < EVTBits)
    return SDValue();

  LoadSDNode *LN0 = cast<LoadSDNode>(N0);
  EVT PtrType = N0.getOperand(1).getValueType();

  if (PtrType == MVT::Untyped || PtrType.isExtended())
    // It's not possible to generate a constant of extended or untyped type.
    return SDValue();

  // For big endian targets, we need to adjust the offset to the pointer to
  // load the correct bytes.
  if (TLI.isBigEndian()) {
    unsigned LVTStoreBits = LN0->getMemoryVT().getStoreSizeInBits();
    unsigned EVTStoreBits = ExtVT.getStoreSizeInBits();
    ShAmt = LVTStoreBits - EVTStoreBits - ShAmt;
  }

  uint64_t PtrOff = ShAmt / 8;
  unsigned NewAlign = MinAlign(LN0->getAlignment(), PtrOff);
  SDValue NewPtr = DAG.getNode(ISD::ADD, LN0->getDebugLoc(),
                               PtrType, LN0->getBasePtr(),
                               DAG.getConstant(PtrOff, PtrType));
  AddToWorkList(NewPtr.getNode());

  SDValue Load;
  if (ExtType == ISD::NON_EXTLOAD)
    Load =  DAG.getLoad(VT, N0.getDebugLoc(), LN0->getChain(), NewPtr,
                        LN0->getPointerInfo().getWithOffset(PtrOff),
                        LN0->isVolatile(), LN0->isNonTemporal(),
                        LN0->isInvariant(), NewAlign);
  else
    Load = DAG.getExtLoad(ExtType, N0.getDebugLoc(), VT, LN0->getChain(),NewPtr,
                          LN0->getPointerInfo().getWithOffset(PtrOff),
                          ExtVT, LN0->isVolatile(), LN0->isNonTemporal(),
                          NewAlign);

  // Replace the old load's chain with the new load's chain.
  WorkListRemover DeadNodes(*this);
  DAG.ReplaceAllUsesOfValueWith(N0.getValue(1), Load.getValue(1));

  // Shift the result left, if we've swallowed a left shift.
  SDValue Result = Load;
  if (ShLeftAmt != 0) {
    EVT ShImmTy = getShiftAmountTy(Result.getValueType());
    if (!isUIntN(ShImmTy.getSizeInBits(), ShLeftAmt))
      ShImmTy = VT;
    Result = DAG.getNode(ISD::SHL, N0.getDebugLoc(), VT,
                         Result, DAG.getConstant(ShLeftAmt, ShImmTy));
  }

  // Return the new loaded value.
  return Result;
}

SDValue DAGCombiner::visitSIGN_EXTEND_INREG(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  SDValue N1 = N->getOperand(1);
  EVT VT = N->getValueType(0);
  EVT EVT = cast<VTSDNode>(N1)->getVT();
  unsigned VTBits = VT.getScalarType().getSizeInBits();
  unsigned EVTBits = EVT.getScalarType().getSizeInBits();

  // fold (sext_in_reg c1) -> c1
  if (isa<ConstantSDNode>(N0) || N0.getOpcode() == ISD::UNDEF)
    return DAG.getNode(ISD::SIGN_EXTEND_INREG, N->getDebugLoc(), VT, N0, N1);

  // If the input is already sign extended, just drop the extension.
  if (DAG.ComputeNumSignBits(N0) >= VTBits-EVTBits+1)
    return N0;

  // fold (sext_in_reg (sext_in_reg x, VT2), VT1) -> (sext_in_reg x, minVT) pt2
  if (N0.getOpcode() == ISD::SIGN_EXTEND_INREG &&
      EVT.bitsLT(cast<VTSDNode>(N0.getOperand(1))->getVT())) {
    return DAG.getNode(ISD::SIGN_EXTEND_INREG, N->getDebugLoc(), VT,
                       N0.getOperand(0), N1);
  }

  // fold (sext_in_reg (sext x)) -> (sext x)
  // fold (sext_in_reg (aext x)) -> (sext x)
  // if x is small enough.
  if (N0.getOpcode() == ISD::SIGN_EXTEND || N0.getOpcode() == ISD::ANY_EXTEND) {
    SDValue N00 = N0.getOperand(0);
    if (N00.getValueType().getScalarType().getSizeInBits() <= EVTBits &&
        (!LegalOperations || TLI.isOperationLegal(ISD::SIGN_EXTEND, VT)))
      return DAG.getNode(ISD::SIGN_EXTEND, N->getDebugLoc(), VT, N00, N1);
  }

  // fold (sext_in_reg x) -> (zext_in_reg x) if the sign bit is known zero.
  if (DAG.MaskedValueIsZero(N0, APInt::getBitsSet(VTBits, EVTBits-1, EVTBits)))
    return DAG.getZeroExtendInReg(N0, N->getDebugLoc(), EVT);

  // fold operands of sext_in_reg based on knowledge that the top bits are not
  // demanded.
  if (SimplifyDemandedBits(SDValue(N, 0)))
    return SDValue(N, 0);

  // fold (sext_in_reg (load x)) -> (smaller sextload x)
  // fold (sext_in_reg (srl (load x), c)) -> (smaller sextload (x+c/evtbits))
  SDValue NarrowLoad = ReduceLoadWidth(N);
  if (NarrowLoad.getNode())
    return NarrowLoad;

  // fold (sext_in_reg (srl X, 24), i8) -> (sra X, 24)
  // fold (sext_in_reg (srl X, 23), i8) -> (sra X, 23) iff possible.
  // We already fold "(sext_in_reg (srl X, 25), i8) -> srl X, 25" above.
  if (N0.getOpcode() == ISD::SRL) {
    if (ConstantSDNode *ShAmt = dyn_cast<ConstantSDNode>(N0.getOperand(1)))
      if (ShAmt->getZExtValue()+EVTBits <= VTBits) {
        // We can turn this into an SRA iff the input to the SRL is already sign
        // extended enough.
        unsigned InSignBits = DAG.ComputeNumSignBits(N0.getOperand(0));
        if (VTBits-(ShAmt->getZExtValue()+EVTBits) < InSignBits)
          return DAG.getNode(ISD::SRA, N->getDebugLoc(), VT,
                             N0.getOperand(0), N0.getOperand(1));
      }
  }

  // fold (sext_inreg (extload x)) -> (sextload x)
  if (ISD::isEXTLoad(N0.getNode()) &&
      ISD::isUNINDEXEDLoad(N0.getNode()) &&
      EVT == cast<LoadSDNode>(N0)->getMemoryVT() &&
      ((!LegalOperations && !cast<LoadSDNode>(N0)->isVolatile()) ||
       TLI.isLoadExtLegal(ISD::SEXTLOAD, EVT))) {
    LoadSDNode *LN0 = cast<LoadSDNode>(N0);
    SDValue ExtLoad = DAG.getExtLoad(ISD::SEXTLOAD, N->getDebugLoc(), VT,
                                     LN0->getChain(),
                                     LN0->getBasePtr(), LN0->getPointerInfo(),
                                     EVT,
                                     LN0->isVolatile(), LN0->isNonTemporal(),
                                     LN0->getAlignment());
    CombineTo(N, ExtLoad);
    CombineTo(N0.getNode(), ExtLoad, ExtLoad.getValue(1));
    return SDValue(N, 0);   // Return N so it doesn't get rechecked!
  }
  // fold (sext_inreg (zextload x)) -> (sextload x) iff load has one use
  if (ISD::isZEXTLoad(N0.getNode()) && ISD::isUNINDEXEDLoad(N0.getNode()) &&
      N0.hasOneUse() &&
      EVT == cast<LoadSDNode>(N0)->getMemoryVT() &&
      ((!LegalOperations && !cast<LoadSDNode>(N0)->isVolatile()) ||
       TLI.isLoadExtLegal(ISD::SEXTLOAD, EVT))) {
    LoadSDNode *LN0 = cast<LoadSDNode>(N0);
    SDValue ExtLoad = DAG.getExtLoad(ISD::SEXTLOAD, N->getDebugLoc(), VT,
                                     LN0->getChain(),
                                     LN0->getBasePtr(), LN0->getPointerInfo(),
                                     EVT,
                                     LN0->isVolatile(), LN0->isNonTemporal(),
                                     LN0->getAlignment());
    CombineTo(N, ExtLoad);
    CombineTo(N0.getNode(), ExtLoad, ExtLoad.getValue(1));
    return SDValue(N, 0);   // Return N so it doesn't get rechecked!
  }

  // Form (sext_inreg (bswap >> 16)) or (sext_inreg (rotl (bswap) 16))
  if (EVTBits <= 16 && N0.getOpcode() == ISD::OR) {
    SDValue BSwap = MatchBSwapHWordLow(N0.getNode(), N0.getOperand(0),
                                       N0.getOperand(1), false);
    if (BSwap.getNode() != 0)
      return DAG.getNode(ISD::SIGN_EXTEND_INREG, N->getDebugLoc(), VT,
                         BSwap, N1);
  }

  return SDValue();
}

SDValue DAGCombiner::visitTRUNCATE(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  EVT VT = N->getValueType(0);
  bool isLE = TLI.isLittleEndian();

  // noop truncate
  if (N0.getValueType() == N->getValueType(0))
    return N0;
  // fold (truncate c1) -> c1
  if (isa<ConstantSDNode>(N0))
    return DAG.getNode(ISD::TRUNCATE, N->getDebugLoc(), VT, N0);
  // fold (truncate (truncate x)) -> (truncate x)
  if (N0.getOpcode() == ISD::TRUNCATE)
    return DAG.getNode(ISD::TRUNCATE, N->getDebugLoc(), VT, N0.getOperand(0));
  // fold (truncate (ext x)) -> (ext x) or (truncate x) or x
  if (N0.getOpcode() == ISD::ZERO_EXTEND ||
      N0.getOpcode() == ISD::SIGN_EXTEND ||
      N0.getOpcode() == ISD::ANY_EXTEND) {
    if (N0.getOperand(0).getValueType().bitsLT(VT))
      // if the source is smaller than the dest, we still need an extend
      return DAG.getNode(N0.getOpcode(), N->getDebugLoc(), VT,
                         N0.getOperand(0));
    if (N0.getOperand(0).getValueType().bitsGT(VT))
      // if the source is larger than the dest, than we just need the truncate
      return DAG.getNode(ISD::TRUNCATE, N->getDebugLoc(), VT, N0.getOperand(0));
    // if the source and dest are the same type, we can drop both the extend
    // and the truncate.
    return N0.getOperand(0);
  }

  // Fold extract-and-trunc into a narrow extract. For example:
  //   i64 x = EXTRACT_VECTOR_ELT(v2i64 val, i32 1)
  //   i32 y = TRUNCATE(i64 x)
  //        -- becomes --
  //   v16i8 b = BITCAST (v2i64 val)
  //   i8 x = EXTRACT_VECTOR_ELT(v16i8 b, i32 8)
  //
  // Note: We only run this optimization after type legalization (which often
  // creates this pattern) and before operation legalization after which
  // we need to be more careful about the vector instructions that we generate.
  if (N0.getOpcode() == ISD::EXTRACT_VECTOR_ELT &&
      LegalTypes && !LegalOperations && N0->hasOneUse()) {

    EVT VecTy = N0.getOperand(0).getValueType();
    EVT ExTy = N0.getValueType();
    EVT TrTy = N->getValueType(0);

    unsigned NumElem = VecTy.getVectorNumElements();
    unsigned SizeRatio = ExTy.getSizeInBits()/TrTy.getSizeInBits();

    EVT NVT = EVT::getVectorVT(*DAG.getContext(), TrTy, SizeRatio * NumElem);
    assert(NVT.getSizeInBits() == VecTy.getSizeInBits() && "Invalid Size");

    SDValue EltNo = N0->getOperand(1);
    if (isa<ConstantSDNode>(EltNo) && isTypeLegal(NVT)) {
      int Elt = cast<ConstantSDNode>(EltNo)->getZExtValue();
      EVT IndexTy = N0->getOperand(1).getValueType();
      int Index = isLE ? (Elt*SizeRatio) : (Elt*SizeRatio + (SizeRatio-1));

      SDValue V = DAG.getNode(ISD::BITCAST, N->getDebugLoc(),
                              NVT, N0.getOperand(0));

      return DAG.getNode(ISD::EXTRACT_VECTOR_ELT,
                         N->getDebugLoc(), TrTy, V,
                         DAG.getConstant(Index, IndexTy));
    }
  }

  // See if we can simplify the input to this truncate through knowledge that
  // only the low bits are being used.
  // For example "trunc (or (shl x, 8), y)" // -> trunc y
  // Currently we only perform this optimization on scalars because vectors
  // may have different active low bits.
  if (!VT.isVector()) {
    SDValue Shorter =
      GetDemandedBits(N0, APInt::getLowBitsSet(N0.getValueSizeInBits(),
                                               VT.getSizeInBits()));
    if (Shorter.getNode())
      return DAG.getNode(ISD::TRUNCATE, N->getDebugLoc(), VT, Shorter);
  }
  // fold (truncate (load x)) -> (smaller load x)
  // fold (truncate (srl (load x), c)) -> (smaller load (x+c/evtbits))
  if (!LegalTypes || TLI.isTypeDesirableForOp(N0.getOpcode(), VT)) {
    SDValue Reduced = ReduceLoadWidth(N);
    if (Reduced.getNode())
      return Reduced;
  }
  // fold (trunc (fptoXi x)) -> (smaller fptoXi x)
  if ((N0.getOpcode() == ISD::FP_TO_UINT ||
       N0.getOpcode() == ISD::FP_TO_SINT) && !LegalTypes)
    return DAG.getNode(N0.getOpcode(), N->getDebugLoc(), VT, N0.getOperand(0));
  // fold (trunc (concat ... x ...)) -> (concat ..., (trunc x), ...)),
  // where ... are all 'undef'.
  if (N0.getOpcode() == ISD::CONCAT_VECTORS && !LegalTypes) {
    SmallVector<EVT, 8> VTs;
    SDValue V;
    unsigned Idx = 0;
    unsigned NumDefs = 0;

    for (unsigned i = 0, e = N0.getNumOperands(); i != e; ++i) {
      SDValue X = N0.getOperand(i);
      if (X.getOpcode() != ISD::UNDEF) {
        V = X;
        Idx = i;
        NumDefs++;
      }
      // Stop if more than one members are non-undef.
      if (NumDefs > 1)
        break;
      VTs.push_back(EVT::getVectorVT(*DAG.getContext(),
                                     VT.getVectorElementType(),
                                     X.getValueType().getVectorNumElements()));
    }

    if (NumDefs == 0)
      return DAG.getUNDEF(VT);

    if (NumDefs == 1) {
      assert(V.getNode() && "The single defined operand is empty!");
      SmallVector<SDValue, 8> Opnds;
      for (unsigned i = 0, e = VTs.size(); i != e; ++i) {
        if (i != Idx) {
          Opnds.push_back(DAG.getUNDEF(VTs[i]));
          continue;
        }
        SDValue NV = DAG.getNode(ISD::TRUNCATE, V.getDebugLoc(), VTs[i], V);
        AddToWorkList(NV.getNode());
        Opnds.push_back(NV);
      }
      return DAG.getNode(ISD::CONCAT_VECTORS, N->getDebugLoc(), VT,
                         &Opnds[0], Opnds.size());
    }
  }

  // Simplify the operands using demanded-bits information.
  if (!VT.isVector() &&
      SimplifyDemandedBits(SDValue(N, 0)))
    return SDValue(N, 0);

  return SDValue();
}

static SDNode *getBuildPairElt(SDNode *N, unsigned i) {
  SDValue Elt = N->getOperand(i);
  if (Elt.getOpcode() != ISD::MERGE_VALUES)
    return Elt.getNode();
  return Elt.getOperand(Elt.getResNo()).getNode();
}

/// CombineConsecutiveLoads - build_pair (load, load) -> load
/// if load locations are consecutive.
SDValue DAGCombiner::CombineConsecutiveLoads(SDNode *N, EVT VT) {
  assert(N->getOpcode() == ISD::BUILD_PAIR);

  LoadSDNode *LD1 = dyn_cast<LoadSDNode>(getBuildPairElt(N, 0));
  LoadSDNode *LD2 = dyn_cast<LoadSDNode>(getBuildPairElt(N, 1));
  if (!LD1 || !LD2 || !ISD::isNON_EXTLoad(LD1) || !LD1->hasOneUse() ||
      LD1->getPointerInfo().getAddrSpace() !=
         LD2->getPointerInfo().getAddrSpace())
    return SDValue();
  EVT LD1VT = LD1->getValueType(0);

  if (ISD::isNON_EXTLoad(LD2) &&
      LD2->hasOneUse() &&
      // If both are volatile this would reduce the number of volatile loads.
      // If one is volatile it might be ok, but play conservative and bail out.
      !LD1->isVolatile() &&
      !LD2->isVolatile() &&
      DAG.isConsecutiveLoad(LD2, LD1, LD1VT.getSizeInBits()/8, 1)) {
    unsigned Align = LD1->getAlignment();
    unsigned NewAlign = TLI.getDataLayout()->
      getABITypeAlignment(VT.getTypeForEVT(*DAG.getContext()));

    if (NewAlign <= Align &&
        (!LegalOperations || TLI.isOperationLegal(ISD::LOAD, VT)))
      return DAG.getLoad(VT, N->getDebugLoc(), LD1->getChain(),
                         LD1->getBasePtr(), LD1->getPointerInfo(),
                         false, false, false, Align);
  }

  return SDValue();
}

SDValue DAGCombiner::visitBITCAST(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  EVT VT = N->getValueType(0);

  // If the input is a BUILD_VECTOR with all constant elements, fold this now.
  // Only do this before legalize, since afterward the target may be depending
  // on the bitconvert.
  // First check to see if this is all constant.
  if (!LegalTypes &&
      N0.getOpcode() == ISD::BUILD_VECTOR && N0.getNode()->hasOneUse() &&
      VT.isVector()) {
    bool isSimple = true;
    for (unsigned i = 0, e = N0.getNumOperands(); i != e; ++i)
      if (N0.getOperand(i).getOpcode() != ISD::UNDEF &&
          N0.getOperand(i).getOpcode() != ISD::Constant &&
          N0.getOperand(i).getOpcode() != ISD::ConstantFP) {
        isSimple = false;
        break;
      }

    EVT DestEltVT = N->getValueType(0).getVectorElementType();
    assert(!DestEltVT.isVector() &&
           "Element type of vector ValueType must not be vector!");
    if (isSimple)
      return ConstantFoldBITCASTofBUILD_VECTOR(N0.getNode(), DestEltVT);
  }

  // If the input is a constant, let getNode fold it.
  if (isa<ConstantSDNode>(N0) || isa<ConstantFPSDNode>(N0)) {
    SDValue Res = DAG.getNode(ISD::BITCAST, N->getDebugLoc(), VT, N0);
    if (Res.getNode() != N) {
      if (!LegalOperations ||
          TLI.isOperationLegal(Res.getNode()->getOpcode(), VT))
        return Res;

      // Folding it resulted in an illegal node, and it's too late to
      // do that. Clean up the old node and forego the transformation.
      // Ideally this won't happen very often, because instcombine
      // and the earlier dagcombine runs (where illegal nodes are
      // permitted) should have folded most of them already.
      DAG.DeleteNode(Res.getNode());
    }
  }

  // (conv (conv x, t1), t2) -> (conv x, t2)
  if (N0.getOpcode() == ISD::BITCAST)
    return DAG.getNode(ISD::BITCAST, N->getDebugLoc(), VT,
                       N0.getOperand(0));

  // fold (conv (load x)) -> (load (conv*)x)
  // If the resultant load doesn't need a higher alignment than the original!
  if (ISD::isNormalLoad(N0.getNode()) && N0.hasOneUse() &&
      // Do not change the width of a volatile load.
      !cast<LoadSDNode>(N0)->isVolatile() &&
      (!LegalOperations || TLI.isOperationLegal(ISD::LOAD, VT))) {
    LoadSDNode *LN0 = cast<LoadSDNode>(N0);
    unsigned Align = TLI.getDataLayout()->
      getABITypeAlignment(VT.getTypeForEVT(*DAG.getContext()));
    unsigned OrigAlign = LN0->getAlignment();

    if (Align <= OrigAlign) {
      SDValue Load = DAG.getLoad(VT, N->getDebugLoc(), LN0->getChain(),
                                 LN0->getBasePtr(), LN0->getPointerInfo(),
                                 LN0->isVolatile(), LN0->isNonTemporal(),
                                 LN0->isInvariant(), OrigAlign);
      AddToWorkList(N);
      CombineTo(N0.getNode(),
                DAG.getNode(ISD::BITCAST, N0.getDebugLoc(),
                            N0.getValueType(), Load),
                Load.getValue(1));
      return Load;
    }
  }

  // fold (bitconvert (fneg x)) -> (xor (bitconvert x), signbit)
  // fold (bitconvert (fabs x)) -> (and (bitconvert x), (not signbit))
  // This often reduces constant pool loads.
  if (((N0.getOpcode() == ISD::FNEG && !TLI.isFNegFree(VT)) ||
       (N0.getOpcode() == ISD::FABS && !TLI.isFAbsFree(VT))) &&
      N0.getNode()->hasOneUse() && VT.isInteger() &&
      !VT.isVector() && !N0.getValueType().isVector()) {
    SDValue NewConv = DAG.getNode(ISD::BITCAST, N0.getDebugLoc(), VT,
                                  N0.getOperand(0));
    AddToWorkList(NewConv.getNode());

    APInt SignBit = APInt::getSignBit(VT.getSizeInBits());
    if (N0.getOpcode() == ISD::FNEG)
      return DAG.getNode(ISD::XOR, N->getDebugLoc(), VT,
                         NewConv, DAG.getConstant(SignBit, VT));
    assert(N0.getOpcode() == ISD::FABS);
    return DAG.getNode(ISD::AND, N->getDebugLoc(), VT,
                       NewConv, DAG.getConstant(~SignBit, VT));
  }

  // fold (bitconvert (fcopysign cst, x)) ->
  //         (or (and (bitconvert x), sign), (and cst, (not sign)))
  // Note that we don't handle (copysign x, cst) because this can always be
  // folded to an fneg or fabs.
  if (N0.getOpcode() == ISD::FCOPYSIGN && N0.getNode()->hasOneUse() &&
      isa<ConstantFPSDNode>(N0.getOperand(0)) &&
      VT.isInteger() && !VT.isVector()) {
    unsigned OrigXWidth = N0.getOperand(1).getValueType().getSizeInBits();
    EVT IntXVT = EVT::getIntegerVT(*DAG.getContext(), OrigXWidth);
    if (isTypeLegal(IntXVT)) {
      SDValue X = DAG.getNode(ISD::BITCAST, N0.getDebugLoc(),
                              IntXVT, N0.getOperand(1));
      AddToWorkList(X.getNode());

      // If X has a different width than the result/lhs, sext it or truncate it.
      unsigned VTWidth = VT.getSizeInBits();
      if (OrigXWidth < VTWidth) {
        X = DAG.getNode(ISD::SIGN_EXTEND, N->getDebugLoc(), VT, X);
        AddToWorkList(X.getNode());
      } else if (OrigXWidth > VTWidth) {
        // To get the sign bit in the right place, we have to shift it right
        // before truncating.
        X = DAG.getNode(ISD::SRL, X.getDebugLoc(),
                        X.getValueType(), X,
                        DAG.getConstant(OrigXWidth-VTWidth, X.getValueType()));
        AddToWorkList(X.getNode());
        X = DAG.getNode(ISD::TRUNCATE, X.getDebugLoc(), VT, X);
        AddToWorkList(X.getNode());
      }

      APInt SignBit = APInt::getSignBit(VT.getSizeInBits());
      X = DAG.getNode(ISD::AND, X.getDebugLoc(), VT,
                      X, DAG.getConstant(SignBit, VT));
      AddToWorkList(X.getNode());

      SDValue Cst = DAG.getNode(ISD::BITCAST, N0.getDebugLoc(),
                                VT, N0.getOperand(0));
      Cst = DAG.getNode(ISD::AND, Cst.getDebugLoc(), VT,
                        Cst, DAG.getConstant(~SignBit, VT));
      AddToWorkList(Cst.getNode());

      return DAG.getNode(ISD::OR, N->getDebugLoc(), VT, X, Cst);
    }
  }

  // bitconvert(build_pair(ld, ld)) -> ld iff load locations are consecutive.
  if (N0.getOpcode() == ISD::BUILD_PAIR) {
    SDValue CombineLD = CombineConsecutiveLoads(N0.getNode(), VT);
    if (CombineLD.getNode())
      return CombineLD;
  }

  return SDValue();
}

SDValue DAGCombiner::visitBUILD_PAIR(SDNode *N) {
  EVT VT = N->getValueType(0);
  return CombineConsecutiveLoads(N, VT);
}

/// ConstantFoldBITCASTofBUILD_VECTOR - We know that BV is a build_vector
/// node with Constant, ConstantFP or Undef operands.  DstEltVT indicates the
/// destination element value type.
SDValue DAGCombiner::
ConstantFoldBITCASTofBUILD_VECTOR(SDNode *BV, EVT DstEltVT) {
  EVT SrcEltVT = BV->getValueType(0).getVectorElementType();

  // If this is already the right type, we're done.
  if (SrcEltVT == DstEltVT) return SDValue(BV, 0);

  unsigned SrcBitSize = SrcEltVT.getSizeInBits();
  unsigned DstBitSize = DstEltVT.getSizeInBits();

  // If this is a conversion of N elements of one type to N elements of another
  // type, convert each element.  This handles FP<->INT cases.
  if (SrcBitSize == DstBitSize) {
    EVT VT = EVT::getVectorVT(*DAG.getContext(), DstEltVT,
                              BV->getValueType(0).getVectorNumElements());

    // Due to the FP element handling below calling this routine recursively,
    // we can end up with a scalar-to-vector node here.
    if (BV->getOpcode() == ISD::SCALAR_TO_VECTOR)
      return DAG.getNode(ISD::SCALAR_TO_VECTOR, BV->getDebugLoc(), VT,
                         DAG.getNode(ISD::BITCAST, BV->getDebugLoc(),
                                     DstEltVT, BV->getOperand(0)));

    SmallVector<SDValue, 8> Ops;
    for (unsigned i = 0, e = BV->getNumOperands(); i != e; ++i) {
      SDValue Op = BV->getOperand(i);
      // If the vector element type is not legal, the BUILD_VECTOR operands
      // are promoted and implicitly truncated.  Make that explicit here.
      if (Op.getValueType() != SrcEltVT)
        Op = DAG.getNode(ISD::TRUNCATE, BV->getDebugLoc(), SrcEltVT, Op);
      Ops.push_back(DAG.getNode(ISD::BITCAST, BV->getDebugLoc(),
                                DstEltVT, Op));
      AddToWorkList(Ops.back().getNode());
    }
    return DAG.getNode(ISD::BUILD_VECTOR, BV->getDebugLoc(), VT,
                       &Ops[0], Ops.size());
  }

  // Otherwise, we're growing or shrinking the elements.  To avoid having to
  // handle annoying details of growing/shrinking FP values, we convert them to
  // int first.
  if (SrcEltVT.isFloatingPoint()) {
    // Convert the input float vector to a int vector where the elements are the
    // same sizes.
    assert((SrcEltVT == MVT::f32 || SrcEltVT == MVT::f64) && "Unknown FP VT!");
    EVT IntVT = EVT::getIntegerVT(*DAG.getContext(), SrcEltVT.getSizeInBits());
    BV = ConstantFoldBITCASTofBUILD_VECTOR(BV, IntVT).getNode();
    SrcEltVT = IntVT;
  }

  // Now we know the input is an integer vector.  If the output is a FP type,
  // convert to integer first, then to FP of the right size.
  if (DstEltVT.isFloatingPoint()) {
    assert((DstEltVT == MVT::f32 || DstEltVT == MVT::f64) && "Unknown FP VT!");
    EVT TmpVT = EVT::getIntegerVT(*DAG.getContext(), DstEltVT.getSizeInBits());
    SDNode *Tmp = ConstantFoldBITCASTofBUILD_VECTOR(BV, TmpVT).getNode();

    // Next, convert to FP elements of the same size.
    return ConstantFoldBITCASTofBUILD_VECTOR(Tmp, DstEltVT);
  }

  // Okay, we know the src/dst types are both integers of differing types.
  // Handling growing first.
  assert(SrcEltVT.isInteger() && DstEltVT.isInteger());
  if (SrcBitSize < DstBitSize) {
    unsigned NumInputsPerOutput = DstBitSize/SrcBitSize;

    SmallVector<SDValue, 8> Ops;
    for (unsigned i = 0, e = BV->getNumOperands(); i != e;
         i += NumInputsPerOutput) {
      bool isLE = TLI.isLittleEndian();
      APInt NewBits = APInt(DstBitSize, 0);
      bool EltIsUndef = true;
      for (unsigned j = 0; j != NumInputsPerOutput; ++j) {
        // Shift the previously computed bits over.
        NewBits <<= SrcBitSize;
        SDValue Op = BV->getOperand(i+ (isLE ? (NumInputsPerOutput-j-1) : j));
        if (Op.getOpcode() == ISD::UNDEF) continue;
        EltIsUndef = false;

        NewBits |= cast<ConstantSDNode>(Op)->getAPIntValue().
                   zextOrTrunc(SrcBitSize).zext(DstBitSize);
      }

      if (EltIsUndef)
        Ops.push_back(DAG.getUNDEF(DstEltVT));
      else
        Ops.push_back(DAG.getConstant(NewBits, DstEltVT));
    }

    EVT VT = EVT::getVectorVT(*DAG.getContext(), DstEltVT, Ops.size());
    return DAG.getNode(ISD::BUILD_VECTOR, BV->getDebugLoc(), VT,
                       &Ops[0], Ops.size());
  }

  // Finally, this must be the case where we are shrinking elements: each input
  // turns into multiple outputs.
  bool isS2V = ISD::isScalarToVector(BV);
  unsigned NumOutputsPerInput = SrcBitSize/DstBitSize;
  EVT VT = EVT::getVectorVT(*DAG.getContext(), DstEltVT,
                            NumOutputsPerInput*BV->getNumOperands());
  SmallVector<SDValue, 8> Ops;

  for (unsigned i = 0, e = BV->getNumOperands(); i != e; ++i) {
    if (BV->getOperand(i).getOpcode() == ISD::UNDEF) {
      for (unsigned j = 0; j != NumOutputsPerInput; ++j)
        Ops.push_back(DAG.getUNDEF(DstEltVT));
      continue;
    }

    APInt OpVal = cast<ConstantSDNode>(BV->getOperand(i))->
                  getAPIntValue().zextOrTrunc(SrcBitSize);

    for (unsigned j = 0; j != NumOutputsPerInput; ++j) {
      APInt ThisVal = OpVal.trunc(DstBitSize);
      Ops.push_back(DAG.getConstant(ThisVal, DstEltVT));
      if (isS2V && i == 0 && j == 0 && ThisVal.zext(SrcBitSize) == OpVal)
        // Simply turn this into a SCALAR_TO_VECTOR of the new type.
        return DAG.getNode(ISD::SCALAR_TO_VECTOR, BV->getDebugLoc(), VT,
                           Ops[0]);
      OpVal = OpVal.lshr(DstBitSize);
    }

    // For big endian targets, swap the order of the pieces of each element.
    if (TLI.isBigEndian())
      std::reverse(Ops.end()-NumOutputsPerInput, Ops.end());
  }

  return DAG.getNode(ISD::BUILD_VECTOR, BV->getDebugLoc(), VT,
                     &Ops[0], Ops.size());
}

SDValue DAGCombiner::visitFADD(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  SDValue N1 = N->getOperand(1);
  ConstantFPSDNode *N0CFP = dyn_cast<ConstantFPSDNode>(N0);
  ConstantFPSDNode *N1CFP = dyn_cast<ConstantFPSDNode>(N1);
  EVT VT = N->getValueType(0);

  // fold vector ops
  if (VT.isVector()) {
    SDValue FoldedVOp = SimplifyVBinOp(N);
    if (FoldedVOp.getNode()) return FoldedVOp;
  }

  // fold (fadd c1, c2) -> c1 + c2
  if (N0CFP && N1CFP && VT != MVT::ppcf128)
    return DAG.getNode(ISD::FADD, N->getDebugLoc(), VT, N0, N1);
  // canonicalize constant to RHS
  if (N0CFP && !N1CFP)
    return DAG.getNode(ISD::FADD, N->getDebugLoc(), VT, N1, N0);
  // fold (fadd A, 0) -> A
  if (DAG.getTarget().Options.UnsafeFPMath && N1CFP &&
      N1CFP->getValueAPF().isZero())
    return N0;
  // fold (fadd A, (fneg B)) -> (fsub A, B)
  if ((!LegalOperations || TLI.isOperationLegalOrCustom(ISD::FSUB, VT)) &&
    isNegatibleForFree(N1, LegalOperations, TLI, &DAG.getTarget().Options) == 2)
    return DAG.getNode(ISD::FSUB, N->getDebugLoc(), VT, N0,
                       GetNegatedExpression(N1, DAG, LegalOperations));
  // fold (fadd (fneg A), B) -> (fsub B, A)
  if ((!LegalOperations || TLI.isOperationLegalOrCustom(ISD::FSUB, VT)) &&
    isNegatibleForFree(N0, LegalOperations, TLI, &DAG.getTarget().Options) == 2)
    return DAG.getNode(ISD::FSUB, N->getDebugLoc(), VT, N1,
                       GetNegatedExpression(N0, DAG, LegalOperations));

  // If allowed, fold (fadd (fadd x, c1), c2) -> (fadd x, (fadd c1, c2))
  if (DAG.getTarget().Options.UnsafeFPMath && N1CFP &&
      N0.getOpcode() == ISD::FADD && N0.getNode()->hasOneUse() &&
      isa<ConstantFPSDNode>(N0.getOperand(1)))
    return DAG.getNode(ISD::FADD, N->getDebugLoc(), VT, N0.getOperand(0),
                       DAG.getNode(ISD::FADD, N->getDebugLoc(), VT,
                                   N0.getOperand(1), N1));

  // In unsafe math mode, we can fold chains of FADD's of the same value
  // into multiplications.  This transform is not safe in general because
  // we are reducing the number of rounding steps.
  if (DAG.getTarget().Options.UnsafeFPMath &&
      TLI.isOperationLegalOrCustom(ISD::FMUL, VT) &&
      !N0CFP && !N1CFP) {
    if (N0.getOpcode() == ISD::FMUL) {
      ConstantFPSDNode *CFP00 = dyn_cast<ConstantFPSDNode>(N0.getOperand(0));
      ConstantFPSDNode *CFP01 = dyn_cast<ConstantFPSDNode>(N0.getOperand(1));

      // (fadd (fmul c, x), x) -> (fmul c+1, x)
      if (CFP00 && !CFP01 && N0.getOperand(1) == N1) {
        SDValue NewCFP = DAG.getNode(ISD::FADD, N->getDebugLoc(), VT,
                                     SDValue(CFP00, 0),
                                     DAG.getConstantFP(1.0, VT));
        return DAG.getNode(ISD::FMUL, N->getDebugLoc(), VT,
                           N1, NewCFP);
      }

      // (fadd (fmul x, c), x) -> (fmul c+1, x)
      if (CFP01 && !CFP00 && N0.getOperand(0) == N1) {
        SDValue NewCFP = DAG.getNode(ISD::FADD, N->getDebugLoc(), VT,
                                     SDValue(CFP01, 0),
                                     DAG.getConstantFP(1.0, VT));
        return DAG.getNode(ISD::FMUL, N->getDebugLoc(), VT,
                           N1, NewCFP);
      }

      // (fadd (fadd x, x), x) -> (fmul 3.0, x)
      if (!CFP00 && !CFP01 && N0.getOperand(0) == N0.getOperand(1) &&
          N0.getOperand(0) == N1) {
        return DAG.getNode(ISD::FMUL, N->getDebugLoc(), VT,
                           N1, DAG.getConstantFP(3.0, VT));
      }

      // (fadd (fmul c, x), (fadd x, x)) -> (fmul c+2, x)
      if (CFP00 && !CFP01 && N1.getOpcode() == ISD::FADD &&
          N1.getOperand(0) == N1.getOperand(1) &&
          N0.getOperand(1) == N1.getOperand(0)) {
        SDValue NewCFP = DAG.getNode(ISD::FADD, N->getDebugLoc(), VT,
                                     SDValue(CFP00, 0),
                                     DAG.getConstantFP(2.0, VT));
        return DAG.getNode(ISD::FMUL, N->getDebugLoc(), VT,
                           N0.getOperand(1), NewCFP);
      }

      // (fadd (fmul x, c), (fadd x, x)) -> (fmul c+2, x)
      if (CFP01 && !CFP00 && N1.getOpcode() == ISD::FADD &&
          N1.getOperand(0) == N1.getOperand(1) &&
          N0.getOperand(0) == N1.getOperand(0)) {
        SDValue NewCFP = DAG.getNode(ISD::FADD, N->getDebugLoc(), VT,
                                     SDValue(CFP01, 0),
                                     DAG.getConstantFP(2.0, VT));
        return DAG.getNode(ISD::FMUL, N->getDebugLoc(), VT,
                           N0.getOperand(0), NewCFP);
      }
    }

    if (N1.getOpcode() == ISD::FMUL) {
      ConstantFPSDNode *CFP10 = dyn_cast<ConstantFPSDNode>(N1.getOperand(0));
      ConstantFPSDNode *CFP11 = dyn_cast<ConstantFPSDNode>(N1.getOperand(1));

      // (fadd x, (fmul c, x)) -> (fmul c+1, x)
      if (CFP10 && !CFP11 && N1.getOperand(1) == N0) {
        SDValue NewCFP = DAG.getNode(ISD::FADD, N->getDebugLoc(), VT,
                                     SDValue(CFP10, 0),
                                     DAG.getConstantFP(1.0, VT));
        return DAG.getNode(ISD::FMUL, N->getDebugLoc(), VT,
                           N0, NewCFP);
      }

      // (fadd x, (fmul x, c)) -> (fmul c+1, x)
      if (CFP11 && !CFP10 && N1.getOperand(0) == N0) {
        SDValue NewCFP = DAG.getNode(ISD::FADD, N->getDebugLoc(), VT,
                                     SDValue(CFP11, 0),
                                     DAG.getConstantFP(1.0, VT));
        return DAG.getNode(ISD::FMUL, N->getDebugLoc(), VT,
                           N0, NewCFP);
      }

      // (fadd x, (fadd x, x)) -> (fmul 3.0, x)
      if (!CFP10 && !CFP11 && N1.getOperand(0) == N1.getOperand(1) &&
          N1.getOperand(0) == N0) {
        return DAG.getNode(ISD::FMUL, N->getDebugLoc(), VT,
                           N0, DAG.getConstantFP(3.0, VT));
      }

      // (fadd (fadd x, x), (fmul c, x)) -> (fmul c+2, x)
      if (CFP10 && !CFP11 && N1.getOpcode() == ISD::FADD &&
          N1.getOperand(0) == N1.getOperand(1) &&
          N0.getOperand(1) == N1.getOperand(0)) {
        SDValue NewCFP = DAG.getNode(ISD::FADD, N->getDebugLoc(), VT,
                                     SDValue(CFP10, 0),
                                     DAG.getConstantFP(2.0, VT));
        return DAG.getNode(ISD::FMUL, N->getDebugLoc(), VT,
                           N0.getOperand(1), NewCFP);
      }

      // (fadd (fadd x, x), (fmul x, c)) -> (fmul c+2, x)
      if (CFP11 && !CFP10 && N1.getOpcode() == ISD::FADD &&
          N1.getOperand(0) == N1.getOperand(1) &&
          N0.getOperand(0) == N1.getOperand(0)) {
        SDValue NewCFP = DAG.getNode(ISD::FADD, N->getDebugLoc(), VT,
                                     SDValue(CFP11, 0),
                                     DAG.getConstantFP(2.0, VT));
        return DAG.getNode(ISD::FMUL, N->getDebugLoc(), VT,
                           N0.getOperand(0), NewCFP);
      }
    }

    // (fadd (fadd x, x), (fadd x, x)) -> (fmul 4.0, x)
    if (N0.getOpcode() == ISD::FADD && N1.getOpcode() == ISD::FADD &&
        N0.getOperand(0) == N0.getOperand(1) &&
        N1.getOperand(0) == N1.getOperand(1) &&
        N0.getOperand(0) == N1.getOperand(0)) {
      return DAG.getNode(ISD::FMUL, N->getDebugLoc(), VT,
                         N0.getOperand(0),
                         DAG.getConstantFP(4.0, VT));
    }
  }

  // FADD -> FMA combines:
  if ((DAG.getTarget().Options.AllowFPOpFusion == FPOpFusion::Fast ||
       DAG.getTarget().Options.UnsafeFPMath) &&
      DAG.getTarget().getTargetLowering()->isFMAFasterThanMulAndAdd(VT) &&
      TLI.isOperationLegalOrCustom(ISD::FMA, VT)) {

    // fold (fadd (fmul x, y), z) -> (fma x, y, z)
    if (N0.getOpcode() == ISD::FMUL && N0->hasOneUse()) {
      return DAG.getNode(ISD::FMA, N->getDebugLoc(), VT,
                         N0.getOperand(0), N0.getOperand(1), N1);
    }

    // fold (fadd x, (fmul y, z)) -> (fma y, z, x)
    // Note: Commutes FADD operands.
    if (N1.getOpcode() == ISD::FMUL && N1->hasOneUse()) {
      return DAG.getNode(ISD::FMA, N->getDebugLoc(), VT,
                         N1.getOperand(0), N1.getOperand(1), N0);
    }
  }

  return SDValue();
}

SDValue DAGCombiner::visitFSUB(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  SDValue N1 = N->getOperand(1);
  ConstantFPSDNode *N0CFP = dyn_cast<ConstantFPSDNode>(N0);
  ConstantFPSDNode *N1CFP = dyn_cast<ConstantFPSDNode>(N1);
  EVT VT = N->getValueType(0);
  DebugLoc dl = N->getDebugLoc();

  // fold vector ops
  if (VT.isVector()) {
    SDValue FoldedVOp = SimplifyVBinOp(N);
    if (FoldedVOp.getNode()) return FoldedVOp;
  }

  // fold (fsub c1, c2) -> c1-c2
  if (N0CFP && N1CFP && VT != MVT::ppcf128)
    return DAG.getNode(ISD::FSUB, N->getDebugLoc(), VT, N0, N1);
  // fold (fsub A, 0) -> A
  if (DAG.getTarget().Options.UnsafeFPMath &&
      N1CFP && N1CFP->getValueAPF().isZero())
    return N0;
  // fold (fsub 0, B) -> -B
  if (DAG.getTarget().Options.UnsafeFPMath &&
      N0CFP && N0CFP->getValueAPF().isZero()) {
    if (isNegatibleForFree(N1, LegalOperations, TLI, &DAG.getTarget().Options))
      return GetNegatedExpression(N1, DAG, LegalOperations);
    if (!LegalOperations || TLI.isOperationLegal(ISD::FNEG, VT))
      return DAG.getNode(ISD::FNEG, dl, VT, N1);
  }
  // fold (fsub A, (fneg B)) -> (fadd A, B)
  if (isNegatibleForFree(N1, LegalOperations, TLI, &DAG.getTarget().Options))
    return DAG.getNode(ISD::FADD, dl, VT, N0,
                       GetNegatedExpression(N1, DAG, LegalOperations));

  // If 'unsafe math' is enabled, fold
  //    (fsub x, x) -> 0.0 &
  //    (fsub x, (fadd x, y)) -> (fneg y) &
  //    (fsub x, (fadd y, x)) -> (fneg y)
  if (DAG.getTarget().Options.UnsafeFPMath) {
    if (N0 == N1)
      return DAG.getConstantFP(0.0f, VT);

    if (N1.getOpcode() == ISD::FADD) {
      SDValue N10 = N1->getOperand(0);
      SDValue N11 = N1->getOperand(1);

      if (N10 == N0 && isNegatibleForFree(N11, LegalOperations, TLI,
                                          &DAG.getTarget().Options))
        return GetNegatedExpression(N11, DAG, LegalOperations);
      else if (N11 == N0 && isNegatibleForFree(N10, LegalOperations, TLI,
                                               &DAG.getTarget().Options))
        return GetNegatedExpression(N10, DAG, LegalOperations);
    }
  }

  // FSUB -> FMA combines:
  if ((DAG.getTarget().Options.AllowFPOpFusion == FPOpFusion::Fast ||
       DAG.getTarget().Options.UnsafeFPMath) &&
      DAG.getTarget().getTargetLowering()->isFMAFasterThanMulAndAdd(VT) &&
      TLI.isOperationLegalOrCustom(ISD::FMA, VT)) {

    // fold (fsub (fmul x, y), z) -> (fma x, y, (fneg z))
    if (N0.getOpcode() == ISD::FMUL && N0->hasOneUse()) {
      return DAG.getNode(ISD::FMA, dl, VT,
                         N0.getOperand(0), N0.getOperand(1),
                         DAG.getNode(ISD::FNEG, dl, VT, N1));
    }

    // fold (fsub x, (fmul y, z)) -> (fma (fneg y), z, x)
    // Note: Commutes FSUB operands.
    if (N1.getOpcode() == ISD::FMUL && N1->hasOneUse()) {
      return DAG.getNode(ISD::FMA, dl, VT,
                         DAG.getNode(ISD::FNEG, dl, VT,
                         N1.getOperand(0)),
                         N1.getOperand(1), N0);
    }

    // fold (fsub (-(fmul, x, y)), z) -> (fma (fneg x), y, (fneg z))
    if (N0.getOpcode() == ISD::FNEG && 
        N0.getOperand(0).getOpcode() == ISD::FMUL &&
        N0->hasOneUse() && N0.getOperand(0).hasOneUse()) {
      SDValue N00 = N0.getOperand(0).getOperand(0);
      SDValue N01 = N0.getOperand(0).getOperand(1);
      return DAG.getNode(ISD::FMA, dl, VT,
                         DAG.getNode(ISD::FNEG, dl, VT, N00), N01,
                         DAG.getNode(ISD::FNEG, dl, VT, N1));
    }
  }

  return SDValue();
}

SDValue DAGCombiner::visitFMUL(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  SDValue N1 = N->getOperand(1);
  ConstantFPSDNode *N0CFP = dyn_cast<ConstantFPSDNode>(N0);
  ConstantFPSDNode *N1CFP = dyn_cast<ConstantFPSDNode>(N1);
  EVT VT = N->getValueType(0);
  const TargetLowering &TLI = DAG.getTargetLoweringInfo();

  // fold vector ops
  if (VT.isVector()) {
    SDValue FoldedVOp = SimplifyVBinOp(N);
    if (FoldedVOp.getNode()) return FoldedVOp;
  }

  // fold (fmul c1, c2) -> c1*c2
  if (N0CFP && N1CFP && VT != MVT::ppcf128)
    return DAG.getNode(ISD::FMUL, N->getDebugLoc(), VT, N0, N1);
  // canonicalize constant to RHS
  if (N0CFP && !N1CFP)
    return DAG.getNode(ISD::FMUL, N->getDebugLoc(), VT, N1, N0);
  // fold (fmul A, 0) -> 0
  if (DAG.getTarget().Options.UnsafeFPMath &&
      N1CFP && N1CFP->getValueAPF().isZero())
    return N1;
  // fold (fmul A, 0) -> 0, vector edition.
  if (DAG.getTarget().Options.UnsafeFPMath &&
      ISD::isBuildVectorAllZeros(N1.getNode()))
    return N1;
  // fold (fmul A, 1.0) -> A
  if (N1CFP && N1CFP->isExactlyValue(1.0))
    return N0;
  // fold (fmul X, 2.0) -> (fadd X, X)
  if (N1CFP && N1CFP->isExactlyValue(+2.0))
    return DAG.getNode(ISD::FADD, N->getDebugLoc(), VT, N0, N0);
  // fold (fmul X, -1.0) -> (fneg X)
  if (N1CFP && N1CFP->isExactlyValue(-1.0))
    if (!LegalOperations || TLI.isOperationLegal(ISD::FNEG, VT))
      return DAG.getNode(ISD::FNEG, N->getDebugLoc(), VT, N0);

  // fold (fmul (fneg X), (fneg Y)) -> (fmul X, Y)
  if (char LHSNeg = isNegatibleForFree(N0, LegalOperations, TLI,
                                       &DAG.getTarget().Options)) {
    if (char RHSNeg = isNegatibleForFree(N1, LegalOperations, TLI, 
                                         &DAG.getTarget().Options)) {
      // Both can be negated for free, check to see if at least one is cheaper
      // negated.
      if (LHSNeg == 2 || RHSNeg == 2)
        return DAG.getNode(ISD::FMUL, N->getDebugLoc(), VT,
                           GetNegatedExpression(N0, DAG, LegalOperations),
                           GetNegatedExpression(N1, DAG, LegalOperations));
    }
  }

  // If allowed, fold (fmul (fmul x, c1), c2) -> (fmul x, (fmul c1, c2))
  if (DAG.getTarget().Options.UnsafeFPMath &&
      N1CFP && N0.getOpcode() == ISD::FMUL &&
      N0.getNode()->hasOneUse() && isa<ConstantFPSDNode>(N0.getOperand(1)))
    return DAG.getNode(ISD::FMUL, N->getDebugLoc(), VT, N0.getOperand(0),
                       DAG.getNode(ISD::FMUL, N->getDebugLoc(), VT,
                                   N0.getOperand(1), N1));

  return SDValue();
}

SDValue DAGCombiner::visitFMA(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  SDValue N1 = N->getOperand(1);
  SDValue N2 = N->getOperand(2);
  ConstantFPSDNode *N0CFP = dyn_cast<ConstantFPSDNode>(N0);
  ConstantFPSDNode *N1CFP = dyn_cast<ConstantFPSDNode>(N1);
  EVT VT = N->getValueType(0);
  DebugLoc dl = N->getDebugLoc();

  if (N0CFP && N0CFP->isExactlyValue(1.0))
    return DAG.getNode(ISD::FADD, N->getDebugLoc(), VT, N1, N2);
  if (N1CFP && N1CFP->isExactlyValue(1.0))
    return DAG.getNode(ISD::FADD, N->getDebugLoc(), VT, N0, N2);

  // Canonicalize (fma c, x, y) -> (fma x, c, y)
  if (N0CFP && !N1CFP)
    return DAG.getNode(ISD::FMA, N->getDebugLoc(), VT, N1, N0, N2);

  // (fma x, c1, (fmul x, c2)) -> (fmul x, c1+c2)
  if (DAG.getTarget().Options.UnsafeFPMath && N1CFP &&
      N2.getOpcode() == ISD::FMUL &&
      N0 == N2.getOperand(0) &&
      N2.getOperand(1).getOpcode() == ISD::ConstantFP) {
    return DAG.getNode(ISD::FMUL, dl, VT, N0,
                       DAG.getNode(ISD::FADD, dl, VT, N1, N2.getOperand(1)));
  }


  // (fma (fmul x, c1), c2, y) -> (fma x, c1*c2, y)
  if (DAG.getTarget().Options.UnsafeFPMath &&
      N0.getOpcode() == ISD::FMUL && N1CFP &&
      N0.getOperand(1).getOpcode() == ISD::ConstantFP) {
    return DAG.getNode(ISD::FMA, dl, VT,
                       N0.getOperand(0),
                       DAG.getNode(ISD::FMUL, dl, VT, N1, N0.getOperand(1)),
                       N2);
  }

  // (fma x, 1, y) -> (fadd x, y)
  // (fma x, -1, y) -> (fadd (fneg x), y)
  if (N1CFP) {
    if (N1CFP->isExactlyValue(1.0))
      return DAG.getNode(ISD::FADD, dl, VT, N0, N2);

    if (N1CFP->isExactlyValue(-1.0) &&
        (!LegalOperations || TLI.isOperationLegal(ISD::FNEG, VT))) {
      SDValue RHSNeg = DAG.getNode(ISD::FNEG, dl, VT, N0);
      AddToWorkList(RHSNeg.getNode());
      return DAG.getNode(ISD::FADD, dl, VT, N2, RHSNeg);
    }
  }

  // (fma x, c, x) -> (fmul x, (c+1))
  if (DAG.getTarget().Options.UnsafeFPMath && N1CFP && N0 == N2) {
    return DAG.getNode(ISD::FMUL, dl, VT,
                       N0,
                       DAG.getNode(ISD::FADD, dl, VT,
                                   N1, DAG.getConstantFP(1.0, VT)));
  }

  // (fma x, c, (fneg x)) -> (fmul x, (c-1))
  if (DAG.getTarget().Options.UnsafeFPMath && N1CFP &&
      N2.getOpcode() == ISD::FNEG && N2.getOperand(0) == N0) {
    return DAG.getNode(ISD::FMUL, dl, VT,
                       N0,
                       DAG.getNode(ISD::FADD, dl, VT,
                                   N1, DAG.getConstantFP(-1.0, VT)));
  }


  return SDValue();
}

SDValue DAGCombiner::visitFDIV(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  SDValue N1 = N->getOperand(1);
  ConstantFPSDNode *N0CFP = dyn_cast<ConstantFPSDNode>(N0);
  ConstantFPSDNode *N1CFP = dyn_cast<ConstantFPSDNode>(N1);
  EVT VT = N->getValueType(0);
  const TargetLowering &TLI = DAG.getTargetLoweringInfo();

  // fold vector ops
  if (VT.isVector()) {
    SDValue FoldedVOp = SimplifyVBinOp(N);
    if (FoldedVOp.getNode()) return FoldedVOp;
  }

  // fold (fdiv c1, c2) -> c1/c2
  if (N0CFP && N1CFP && VT != MVT::ppcf128)
    return DAG.getNode(ISD::FDIV, N->getDebugLoc(), VT, N0, N1);

  // fold (fdiv X, c2) -> fmul X, 1/c2 if losing precision is acceptable.
  if (N1CFP && VT != MVT::ppcf128 && DAG.getTarget().Options.UnsafeFPMath) {
    // Compute the reciprocal 1.0 / c2.
    APFloat N1APF = N1CFP->getValueAPF();
    APFloat Recip(N1APF.getSemantics(), 1); // 1.0
    APFloat::opStatus st = Recip.divide(N1APF, APFloat::rmNearestTiesToEven);
    // Only do the transform if the reciprocal is a legal fp immediate that
    // isn't too nasty (eg NaN, denormal, ...).
    if ((st == APFloat::opOK || st == APFloat::opInexact) && // Not too nasty
        (!LegalOperations ||
         // FIXME: custom lowering of ConstantFP might fail (see e.g. ARM
         // backend)... we should handle this gracefully after Legalize.
         // TLI.isOperationLegalOrCustom(llvm::ISD::ConstantFP, VT) ||
         TLI.isOperationLegal(llvm::ISD::ConstantFP, VT) ||
         TLI.isFPImmLegal(Recip, VT)))
      return DAG.getNode(ISD::FMUL, N->getDebugLoc(), VT, N0,
                         DAG.getConstantFP(Recip, VT));
  }

  // (fdiv (fneg X), (fneg Y)) -> (fdiv X, Y)
  if (char LHSNeg = isNegatibleForFree(N0, LegalOperations, TLI,
                                       &DAG.getTarget().Options)) {
    if (char RHSNeg = isNegatibleForFree(N1, LegalOperations, TLI,
                                         &DAG.getTarget().Options)) {
      // Both can be negated for free, check to see if at least one is cheaper
      // negated.
      if (LHSNeg == 2 || RHSNeg == 2)
        return DAG.getNode(ISD::FDIV, N->getDebugLoc(), VT,
                           GetNegatedExpression(N0, DAG, LegalOperations),
                           GetNegatedExpression(N1, DAG, LegalOperations));
    }
  }

  return SDValue();
}

SDValue DAGCombiner::visitFREM(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  SDValue N1 = N->getOperand(1);
  ConstantFPSDNode *N0CFP = dyn_cast<ConstantFPSDNode>(N0);
  ConstantFPSDNode *N1CFP = dyn_cast<ConstantFPSDNode>(N1);
  EVT VT = N->getValueType(0);

  // fold (frem c1, c2) -> fmod(c1,c2)
  if (N0CFP && N1CFP && VT != MVT::ppcf128)
    return DAG.getNode(ISD::FREM, N->getDebugLoc(), VT, N0, N1);

  return SDValue();
}

SDValue DAGCombiner::visitFCOPYSIGN(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  SDValue N1 = N->getOperand(1);
  ConstantFPSDNode *N0CFP = dyn_cast<ConstantFPSDNode>(N0);
  ConstantFPSDNode *N1CFP = dyn_cast<ConstantFPSDNode>(N1);
  EVT VT = N->getValueType(0);

  if (N0CFP && N1CFP && VT != MVT::ppcf128)  // Constant fold
    return DAG.getNode(ISD::FCOPYSIGN, N->getDebugLoc(), VT, N0, N1);

  if (N1CFP) {
    const APFloat& V = N1CFP->getValueAPF();
    // copysign(x, c1) -> fabs(x)       iff ispos(c1)
    // copysign(x, c1) -> fneg(fabs(x)) iff isneg(c1)
    if (!V.isNegative()) {
      if (!LegalOperations || TLI.isOperationLegal(ISD::FABS, VT))
        return DAG.getNode(ISD::FABS, N->getDebugLoc(), VT, N0);
    } else {
      if (!LegalOperations || TLI.isOperationLegal(ISD::FNEG, VT))
        return DAG.getNode(ISD::FNEG, N->getDebugLoc(), VT,
                           DAG.getNode(ISD::FABS, N0.getDebugLoc(), VT, N0));
    }
  }

  // copysign(fabs(x), y) -> copysign(x, y)
  // copysign(fneg(x), y) -> copysign(x, y)
  // copysign(copysign(x,z), y) -> copysign(x, y)
  if (N0.getOpcode() == ISD::FABS || N0.getOpcode() == ISD::FNEG ||
      N0.getOpcode() == ISD::FCOPYSIGN)
    return DAG.getNode(ISD::FCOPYSIGN, N->getDebugLoc(), VT,
                       N0.getOperand(0), N1);

  // copysign(x, abs(y)) -> abs(x)
  if (N1.getOpcode() == ISD::FABS)
    return DAG.getNode(ISD::FABS, N->getDebugLoc(), VT, N0);

  // copysign(x, copysign(y,z)) -> copysign(x, z)
  if (N1.getOpcode() == ISD::FCOPYSIGN)
    return DAG.getNode(ISD::FCOPYSIGN, N->getDebugLoc(), VT,
                       N0, N1.getOperand(1));

  // copysign(x, fp_extend(y)) -> copysign(x, y)
  // copysign(x, fp_round(y)) -> copysign(x, y)
  if (N1.getOpcode() == ISD::FP_EXTEND || N1.getOpcode() == ISD::FP_ROUND)
    return DAG.getNode(ISD::FCOPYSIGN, N->getDebugLoc(), VT,
                       N0, N1.getOperand(0));

  return SDValue();
}

SDValue DAGCombiner::visitSINT_TO_FP(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  ConstantSDNode *N0C = dyn_cast<ConstantSDNode>(N0);
  EVT VT = N->getValueType(0);
  EVT OpVT = N0.getValueType();

  // fold (sint_to_fp c1) -> c1fp
  if (N0C && OpVT != MVT::ppcf128 &&
      // ...but only if the target supports immediate floating-point values
      (!LegalOperations ||
       TLI.isOperationLegalOrCustom(llvm::ISD::ConstantFP, VT)))
    return DAG.getNode(ISD::SINT_TO_FP, N->getDebugLoc(), VT, N0);

  // If the input is a legal type, and SINT_TO_FP is not legal on this target,
  // but UINT_TO_FP is legal on this target, try to convert.
  if (!TLI.isOperationLegalOrCustom(ISD::SINT_TO_FP, OpVT) &&
      TLI.isOperationLegalOrCustom(ISD::UINT_TO_FP, OpVT)) {
    // If the sign bit is known to be zero, we can change this to UINT_TO_FP.
    if (DAG.SignBitIsZero(N0))
      return DAG.getNode(ISD::UINT_TO_FP, N->getDebugLoc(), VT, N0);
  }

  // The next optimizations are desireable only if SELECT_CC can be lowered.
  // Check against MVT::Other for SELECT_CC, which is a workaround for targets
  // having to say they don't support SELECT_CC on every type the DAG knows
  // about, since there is no way to mark an opcode illegal at all value types
  // (See also visitSELECT)
  if (TLI.isOperationLegalOrCustom(ISD::SELECT_CC, MVT::Other)) {
    // fold (sint_to_fp (setcc x, y, cc)) -> (select_cc x, y, -1.0, 0.0,, cc)
    if (N0.getOpcode() == ISD::SETCC && N0.getValueType() == MVT::i1 &&
        !VT.isVector() &&
        (!LegalOperations ||
         TLI.isOperationLegalOrCustom(llvm::ISD::ConstantFP, VT))) {
      SDValue Ops[] =
        { N0.getOperand(0), N0.getOperand(1),
          DAG.getConstantFP(-1.0, VT) , DAG.getConstantFP(0.0, VT),
          N0.getOperand(2) };
      return DAG.getNode(ISD::SELECT_CC, N->getDebugLoc(), VT, Ops, 5);
    }

    // fold (sint_to_fp (zext (setcc x, y, cc))) ->
    //      (select_cc x, y, 1.0, 0.0,, cc)
    if (N0.getOpcode() == ISD::ZERO_EXTEND &&
        N0.getOperand(0).getOpcode() == ISD::SETCC &&!VT.isVector() &&
        (!LegalOperations ||
         TLI.isOperationLegalOrCustom(llvm::ISD::ConstantFP, VT))) {
      SDValue Ops[] =
        { N0.getOperand(0).getOperand(0), N0.getOperand(0).getOperand(1),
          DAG.getConstantFP(1.0, VT) , DAG.getConstantFP(0.0, VT),
          N0.getOperand(0).getOperand(2) };
      return DAG.getNode(ISD::SELECT_CC, N->getDebugLoc(), VT, Ops, 5);
    }
  }

  return SDValue();
}

SDValue DAGCombiner::visitUINT_TO_FP(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  ConstantSDNode *N0C = dyn_cast<ConstantSDNode>(N0);
  EVT VT = N->getValueType(0);
  EVT OpVT = N0.getValueType();

  // fold (uint_to_fp c1) -> c1fp
  if (N0C && OpVT != MVT::ppcf128 &&
      // ...but only if the target supports immediate floating-point values
      (!LegalOperations ||
       TLI.isOperationLegalOrCustom(llvm::ISD::ConstantFP, VT)))
    return DAG.getNode(ISD::UINT_TO_FP, N->getDebugLoc(), VT, N0);

  // If the input is a legal type, and UINT_TO_FP is not legal on this target,
  // but SINT_TO_FP is legal on this target, try to convert.
  if (!TLI.isOperationLegalOrCustom(ISD::UINT_TO_FP, OpVT) &&
      TLI.isOperationLegalOrCustom(ISD::SINT_TO_FP, OpVT)) {
    // If the sign bit is known to be zero, we can change this to SINT_TO_FP.
    if (DAG.SignBitIsZero(N0))
      return DAG.getNode(ISD::SINT_TO_FP, N->getDebugLoc(), VT, N0);
  }

  // The next optimizations are desireable only if SELECT_CC can be lowered.
  // Check against MVT::Other for SELECT_CC, which is a workaround for targets
  // having to say they don't support SELECT_CC on every type the DAG knows
  // about, since there is no way to mark an opcode illegal at all value types
  // (See also visitSELECT)
  if (TLI.isOperationLegalOrCustom(ISD::SELECT_CC, MVT::Other)) {
    // fold (uint_to_fp (setcc x, y, cc)) -> (select_cc x, y, -1.0, 0.0,, cc)

    if (N0.getOpcode() == ISD::SETCC && !VT.isVector() &&
        (!LegalOperations ||
         TLI.isOperationLegalOrCustom(llvm::ISD::ConstantFP, VT))) {
      SDValue Ops[] =
        { N0.getOperand(0), N0.getOperand(1),
          DAG.getConstantFP(1.0, VT),  DAG.getConstantFP(0.0, VT),
          N0.getOperand(2) };
      return DAG.getNode(ISD::SELECT_CC, N->getDebugLoc(), VT, Ops, 5);
    }
  }

  return SDValue();
}

SDValue DAGCombiner::visitFP_TO_SINT(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  ConstantFPSDNode *N0CFP = dyn_cast<ConstantFPSDNode>(N0);
  EVT VT = N->getValueType(0);

  // fold (fp_to_sint c1fp) -> c1
  if (N0CFP)
    return DAG.getNode(ISD::FP_TO_SINT, N->getDebugLoc(), VT, N0);

  return SDValue();
}

SDValue DAGCombiner::visitFP_TO_UINT(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  ConstantFPSDNode *N0CFP = dyn_cast<ConstantFPSDNode>(N0);
  EVT VT = N->getValueType(0);

  // fold (fp_to_uint c1fp) -> c1
  if (N0CFP && VT != MVT::ppcf128)
    return DAG.getNode(ISD::FP_TO_UINT, N->getDebugLoc(), VT, N0);

  return SDValue();
}

SDValue DAGCombiner::visitFP_ROUND(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  SDValue N1 = N->getOperand(1);
  ConstantFPSDNode *N0CFP = dyn_cast<ConstantFPSDNode>(N0);
  EVT VT = N->getValueType(0);

  // fold (fp_round c1fp) -> c1fp
  if (N0CFP && N0.getValueType() != MVT::ppcf128)
    return DAG.getNode(ISD::FP_ROUND, N->getDebugLoc(), VT, N0, N1);

  // fold (fp_round (fp_extend x)) -> x
  if (N0.getOpcode() == ISD::FP_EXTEND && VT == N0.getOperand(0).getValueType())
    return N0.getOperand(0);

  // fold (fp_round (fp_round x)) -> (fp_round x)
  if (N0.getOpcode() == ISD::FP_ROUND) {
    // This is a value preserving truncation if both round's are.
    bool IsTrunc = N->getConstantOperandVal(1) == 1 &&
                   N0.getNode()->getConstantOperandVal(1) == 1;
    return DAG.getNode(ISD::FP_ROUND, N->getDebugLoc(), VT, N0.getOperand(0),
                       DAG.getIntPtrConstant(IsTrunc));
  }

  // fold (fp_round (copysign X, Y)) -> (copysign (fp_round X), Y)
  if (N0.getOpcode() == ISD::FCOPYSIGN && N0.getNode()->hasOneUse()) {
    SDValue Tmp = DAG.getNode(ISD::FP_ROUND, N0.getDebugLoc(), VT,
                              N0.getOperand(0), N1);
    AddToWorkList(Tmp.getNode());
    return DAG.getNode(ISD::FCOPYSIGN, N->getDebugLoc(), VT,
                       Tmp, N0.getOperand(1));
  }

  return SDValue();
}

SDValue DAGCombiner::visitFP_ROUND_INREG(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  EVT VT = N->getValueType(0);
  EVT EVT = cast<VTSDNode>(N->getOperand(1))->getVT();
  ConstantFPSDNode *N0CFP = dyn_cast<ConstantFPSDNode>(N0);

  // fold (fp_round_inreg c1fp) -> c1fp
  if (N0CFP && isTypeLegal(EVT)) {
    SDValue Round = DAG.getConstantFP(*N0CFP->getConstantFPValue(), EVT);
    return DAG.getNode(ISD::FP_EXTEND, N->getDebugLoc(), VT, Round);
  }

  return SDValue();
}

SDValue DAGCombiner::visitFP_EXTEND(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  ConstantFPSDNode *N0CFP = dyn_cast<ConstantFPSDNode>(N0);
  EVT VT = N->getValueType(0);

  // If this is fp_round(fpextend), don't fold it, allow ourselves to be folded.
  if (N->hasOneUse() &&
      N->use_begin()->getOpcode() == ISD::FP_ROUND)
    return SDValue();

  // fold (fp_extend c1fp) -> c1fp
  if (N0CFP && VT != MVT::ppcf128)
    return DAG.getNode(ISD::FP_EXTEND, N->getDebugLoc(), VT, N0);

  // Turn fp_extend(fp_round(X, 1)) -> x since the fp_round doesn't affect the
  // value of X.
  if (N0.getOpcode() == ISD::FP_ROUND
      && N0.getNode()->getConstantOperandVal(1) == 1) {
    SDValue In = N0.getOperand(0);
    if (In.getValueType() == VT) return In;
    if (VT.bitsLT(In.getValueType()))
      return DAG.getNode(ISD::FP_ROUND, N->getDebugLoc(), VT,
                         In, N0.getOperand(1));
    return DAG.getNode(ISD::FP_EXTEND, N->getDebugLoc(), VT, In);
  }

  // fold (fpext (load x)) -> (fpext (fptrunc (extload x)))
  if (ISD::isNON_EXTLoad(N0.getNode()) && N0.hasOneUse() &&
      ((!LegalOperations && !cast<LoadSDNode>(N0)->isVolatile()) ||
       TLI.isLoadExtLegal(ISD::EXTLOAD, N0.getValueType()))) {
    LoadSDNode *LN0 = cast<LoadSDNode>(N0);
    SDValue ExtLoad = DAG.getExtLoad(ISD::EXTLOAD, N->getDebugLoc(), VT,
                                     LN0->getChain(),
                                     LN0->getBasePtr(), LN0->getPointerInfo(),
                                     N0.getValueType(),
                                     LN0->isVolatile(), LN0->isNonTemporal(),
                                     LN0->getAlignment());
    CombineTo(N, ExtLoad);
    CombineTo(N0.getNode(),
              DAG.getNode(ISD::FP_ROUND, N0.getDebugLoc(),
                          N0.getValueType(), ExtLoad, DAG.getIntPtrConstant(1)),
              ExtLoad.getValue(1));
    return SDValue(N, 0);   // Return N so it doesn't get rechecked!
  }

  return SDValue();
}

SDValue DAGCombiner::visitFNEG(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  EVT VT = N->getValueType(0);

  if (VT.isVector()) {
    SDValue FoldedVOp = SimplifyVUnaryOp(N);
    if (FoldedVOp.getNode()) return FoldedVOp;
  }

  if (isNegatibleForFree(N0, LegalOperations, DAG.getTargetLoweringInfo(),
                         &DAG.getTarget().Options))
    return GetNegatedExpression(N0, DAG, LegalOperations);

  // Transform fneg(bitconvert(x)) -> bitconvert(x^sign) to avoid loading
  // constant pool values.
  if (!TLI.isFNegFree(VT) && N0.getOpcode() == ISD::BITCAST &&
      !VT.isVector() &&
      N0.getNode()->hasOneUse() &&
      N0.getOperand(0).getValueType().isInteger()) {
    SDValue Int = N0.getOperand(0);
    EVT IntVT = Int.getValueType();
    if (IntVT.isInteger() && !IntVT.isVector()) {
      Int = DAG.getNode(ISD::XOR, N0.getDebugLoc(), IntVT, Int,
              DAG.getConstant(APInt::getSignBit(IntVT.getSizeInBits()), IntVT));
      AddToWorkList(Int.getNode());
      return DAG.getNode(ISD::BITCAST, N->getDebugLoc(),
                         VT, Int);
    }
  }

  // (fneg (fmul c, x)) -> (fmul -c, x)
  if (N0.getOpcode() == ISD::FMUL) {
    ConstantFPSDNode *CFP1 = dyn_cast<ConstantFPSDNode>(N0.getOperand(1));
    if (CFP1) {
      return DAG.getNode(ISD::FMUL, N->getDebugLoc(), VT,
                         N0.getOperand(0),
                         DAG.getNode(ISD::FNEG, N->getDebugLoc(), VT,
                                     N0.getOperand(1)));
    }
  }

  return SDValue();
}

SDValue DAGCombiner::visitFCEIL(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  ConstantFPSDNode *N0CFP = dyn_cast<ConstantFPSDNode>(N0);
  EVT VT = N->getValueType(0);

  // fold (fceil c1) -> fceil(c1)
  if (N0CFP && VT != MVT::ppcf128)
    return DAG.getNode(ISD::FCEIL, N->getDebugLoc(), VT, N0);

  return SDValue();
}

SDValue DAGCombiner::visitFTRUNC(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  ConstantFPSDNode *N0CFP = dyn_cast<ConstantFPSDNode>(N0);
  EVT VT = N->getValueType(0);

  // fold (ftrunc c1) -> ftrunc(c1)
  if (N0CFP && VT != MVT::ppcf128)
    return DAG.getNode(ISD::FTRUNC, N->getDebugLoc(), VT, N0);

  return SDValue();
}

SDValue DAGCombiner::visitFFLOOR(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  ConstantFPSDNode *N0CFP = dyn_cast<ConstantFPSDNode>(N0);
  EVT VT = N->getValueType(0);

  // fold (ffloor c1) -> ffloor(c1)
  if (N0CFP && VT != MVT::ppcf128)
    return DAG.getNode(ISD::FFLOOR, N->getDebugLoc(), VT, N0);

  return SDValue();
}

SDValue DAGCombiner::visitFABS(SDNode *N) {
  SDValue N0 = N->getOperand(0);
  ConstantFPSDNode *N0CFP = dyn_cast<ConstantFPSDNode>(N0);
  EVT VT = N->getValueType(0);

  if (VT.isVector()) {
    SDValue FoldedVOp = SimplifyVUnaryOp(N);
    if (FoldedVOp.getNode()) return FoldedVOp;
  }

  // fold (fabs c1) -> fabs(c1)
  if (N0CFP && VT != MVT::ppcf128)
    return DAG.getNode(ISD::FABS, N->getDebugLoc(), VT, N0);
  // fold (fabs (fabs x)) -> (fabs x)
  if (N0.getOpcode() == ISD::FABS)
    return N->getOperand(0);
  // fold (fabs (fneg x)) -> (fabs x)
  // fold (fabs (fcopysign x, y)) -> (fabs x)
  if (N0.getOpcode() == ISD::FNEG || N0.getOpcode() == ISD::FCOPYSIGN)
    return DAG.getNode(ISD::FABS, N->getDebugLoc(), VT, N0.getOperand(0));

  // Transform fabs(bitconvert(x)) -> bitconvert(x&~sign) to avoid loading
  // constant pool values.
  if (!TLI.isFAbsFree(VT) && 
      N0.getOpcode() == ISD::BITCAST && N0.getNode()->hasOneUse() &&
      N0.getOperand(0).getValueType().isInteger() &&
      !N0.getOperand(0).getValueType().isVector()) {
    SDValue Int = N0.getOperand(0);
    EVT IntVT = Int.getValueType();
    if (IntVT.isInteger() && !IntVT.isVector()) {
      Int = DAG.getNode(ISD::AND, N0.getDebugLoc(), IntVT, Int,
             DAG.getConstant(~APInt::getSignBit(IntVT.getSizeInBits()), IntVT));
      AddToWorkList(Int.getNode());
      return DAG.getNode(ISD::BITCAST, N->getDebugLoc(),
                         N->getValueType(0), Int);
    }
  }

  return SDValue();
}

SDValue DAGCombiner::visitBRCOND(SDNode *N) {
  SDValue Chain = N->getOperand(0);
  SDValue N1 = N->getOperand(1);
  SDValue N2 = N->getOperand(2);

  // If N is a constant we could fold this into a fallthrough or unconditional
  // branch. However that doesn't happen very often in normal code, because
  // Instcombine/SimplifyCFG should have handled the available opportunities.
  // If we did this folding here, it would be necessary to update the
  // MachineBasicBlock CFG, which is awkward.

  // fold a brcond with a setcc condition into a BR_CC node if BR_CC is legal
  // on the target.
  if (N1.getOpcode() == ISD::SETCC &&
      TLI.isOperationLegalOrCustom(ISD::BR_CC, MVT::Other)) {
    return DAG.getNode(ISD::BR_CC, N->getDebugLoc(), MVT::Other,
                       Chain, N1.getOperand(2),
                       N1.getOperand(0), N1.getOperand(1), N2);
  }

  if ((N1.hasOneUse() && N1.getOpcode() == ISD::SRL) ||
      ((N1.getOpcode() == ISD::TRUNCATE && N1.hasOneUse()) &&
       (N1.getOperand(0).hasOneUse() &&
        N1.getOperand(0).getOpcode() == ISD::SRL))) {
    SDNode *Trunc = 0;
    if (N1.getOpcode() == ISD::TRUNCATE) {
      // Look pass the truncate.
      Trunc = N1.getNode();
      N1 = N1.getOperand(0);
    }

    // Match this pattern so that we can generate simpler code:
    //
    //   %a = ...
    //   %b = and i32 %a, 2
    //   %c = srl i32 %b, 1
    //   brcond i32 %c ...
    //
    // into
    //
    //   %a = ...
    //   %b = and i32 %a, 2
    //   %c = setcc eq %b, 0
    //   brcond %c ...
    //
    // This applies only when the AND constant value has one bit set and the
    // SRL constant is equal to the log2 of the AND constant. The back-end is
    // smart enough to convert the result into a TEST/JMP sequence.
    SDValue Op0 = N1.getOperand(0);
    SDValue Op1 = N1.getOperand(1);

    if (Op0.getOpcode() == ISD::AND &&
        Op1.getOpcode() == ISD::Constant) {
      SDValue AndOp1 = Op0.getOperand(1);

      if (AndOp1.getOpcode() == ISD::Constant) {
        const APInt &AndConst = cast<ConstantSDNode>(AndOp1)->getAPIntValue();

        if (AndConst.isPowerOf2() &&
            cast<ConstantSDNode>(Op1)->getAPIntValue()==AndConst.logBase2()) {
          SDValue SetCC =
            DAG.getSetCC(N->getDebugLoc(),
                         TLI.getSetCCResultType(Op0.getValueType()),
                         Op0, DAG.getConstant(0, Op0.getValueType()),
                         ISD::SETNE);

          SDValue NewBRCond = DAG.getNode(ISD::BRCOND, N->getDebugLoc(),
                                          MVT::Other, Chain, SetCC, N2);
          // Don't add the new BRCond into the worklist or else SimplifySelectCC
          // will convert it back to (X & C1) >> C2.
          CombineTo(N, NewBRCond, false);
          // Truncate is dead.
          if (Trunc) {
            removeFromWorkList(Trunc);
            DAG.DeleteNode(Trunc);
          }
          // Replace the uses of SRL with SETCC
          WorkListRemover DeadNodes(*this);
          DAG.ReplaceAllUsesOfValueWith(N1, SetCC);
          removeFromWorkList(N1.getNode());
          DAG.DeleteNode(N1.getNode());
          return SDValue(N, 0);   // Return N so it doesn't get rechecked!
        }
      }
    }

    if (Trunc)
      // Restore N1 if the above transformation doesn't match.
      N1 = N->getOperand(1);
  }

  // Transform br(xor(x, y)) -> br(x != y)
  // Transform br(xor(xor(x,y), 1)) -> br (x == y)
  if (N1.hasOneUse() && N1.getOpcode() == ISD::XOR) {
    SDNode *TheXor = N1.getNode();
    SDValue Op0 = TheXor->getOperand(0);
    SDValue Op1 = TheXor->getOperand(1);
    if (Op0.getOpcode() == Op1.getOpcode()) {
      // Avoid missing important xor optimizations.
      SDValue Tmp = visitXOR(TheXor);
      if (Tmp.getNode() && Tmp.getNode() != TheXor) {
        DEBUG(dbgs() << "\nReplacing.8 ";
              TheXor->dump(&DAG);
              dbgs() << "\nWith: ";
              Tmp.getNode()->dump(&DAG);
              dbgs() << '\n');
        WorkListRemover DeadNodes(*this);
        DAG.ReplaceAllUsesOfValueWith(N1, Tmp);
        removeFromWorkList(TheXor);
        DAG.DeleteNode(TheXor);
        return DAG.getNode(ISD::BRCOND, N->getDebugLoc(),
                           MVT::Other, Chain, Tmp, N2);
      }
    }

    if (Op0.getOpcode() != ISD::SETCC && Op1.getOpcode() != ISD::SETCC) {
      bool Equal = false;
      if (ConstantSDNode *RHSCI = dyn_cast<ConstantSDNode>(Op0))
        if (RHSCI->getAPIntValue() == 1 && Op0.hasOneUse() &&
            Op0.getOpcode() == ISD::XOR) {
          TheXor = Op0.getNode();
          Equal = true;
        }

      EVT SetCCVT = N1.getValueType();
      if (LegalTypes)
        SetCCVT = TLI.getSetCCResultType(SetCCVT);
      SDValue SetCC = DAG.getSetCC(TheXor->getDebugLoc(),
                                   SetCCVT,
                                   Op0, Op1,
                                   Equal ? ISD::SETEQ : ISD::SETNE);
      // Replace the uses of XOR with SETCC
      WorkListRemover DeadNodes(*this);
      DAG.ReplaceAllUsesOfValueWith(N1, SetCC);
      removeFromWorkList(N1.getNode());
      DAG.DeleteNode(N1.getNode());
      return DAG.getNode(ISD::BRCOND, N->getDebugLoc(),
                         MVT::Other, Chain, SetCC, N2);
    }
  }

  return SDValue();
}

// Operand List for BR_CC: Chain, CondCC, CondLHS, CondRHS, DestBB.
//
SDValue DAGCombiner::visitBR_CC(SDNode *N) {
  CondCodeSDNode *CC = cast<CondCodeSDNode>(N->getOperand(1));
  SDValue CondLHS = N->getOperand(2), CondRHS = N->getOperand(3);

  // If N is a constant we could fold this into a fallthrough or unconditional
  // branch. However that doesn't happen very often in normal code, because
  // Instcombine/SimplifyCFG should have handled the available opportunities.
  // If we did this folding here, it would be necessary to update the
  // MachineBasicBlock CFG, which is awkward.

  // Use SimplifySetCC to simplify SETCC's.
  SDValue Simp = SimplifySetCC(TLI.getSetCCResultType(CondLHS.getValueType()),
                               CondLHS, CondRHS, CC->get(), N->getDebugLoc(),
                               false);
  if (Simp.getNode()) AddToWorkList(Simp.getNode());

  // fold to a simpler setcc
  if (Simp.getNode() && Simp.getOpcode() == ISD::SETCC)
    return DAG.getNode(ISD::BR_CC, N->getDebugLoc(), MVT::Other,
                       N->getOperand(0), Simp.getOperand(2),
                       Simp.getOperand(0), Simp.getOperand(1),
                       N->getOperand(4));

  return SDValue();
}

/// canFoldInAddressingMode - Return true if 'Use' is a load or a store that
/// uses N as its base pointer and that N may be folded in the load / store
/// addressing mode.
static bool canFoldInAddressingMode(SDNode *N, SDNode *Use,
                                    SelectionDAG &DAG,
                                    const TargetLowering &TLI) {
  EVT VT;
  if (LoadSDNode *LD  = dyn_cast<LoadSDNode>(Use)) {
    if (LD->isIndexed() || LD->getBasePtr().getNode() != N)
      return false;
    VT = Use->getValueType(0);
  } else if (StoreSDNode *ST  = dyn_cast<StoreSDNode>(Use)) {
    if (ST->isIndexed() || ST->getBasePtr().getNode() != N)
      return false;
    VT = ST->getValue().getValueType();
  } else
    return false;

  AddrMode AM;
  if (N->getOpcode() == ISD::ADD) {
    ConstantSDNode *Offset = dyn_cast<ConstantSDNode>(N->getOperand(1));
    if (Offset)
      // [reg +/- imm]
      AM.BaseOffs = Offset->getSExtValue();
    else
      // [reg +/- reg]
      AM.Scale = 1;
  } else if (N->getOpcode() == ISD::SUB) {
    ConstantSDNode *Offset = dyn_cast<ConstantSDNode>(N->getOperand(1));
    if (Offset)
      // [reg +/- imm]
      AM.BaseOffs = -Offset->getSExtValue();
    else
      // [reg +/- reg]
      AM.Scale = 1;
  } else
    return false;

  return TLI.isLegalAddressingMode(AM, VT.getTypeForEVT(*DAG.getContext()));
}

/// CombineToPreIndexedLoadStore - Try turning a load / store into a
/// pre-indexed load / store when the base pointer is an add or subtract
/// and it has other uses besides the load / store. After the
/// transformation, the new indexed load / store has effectively folded
/// the add / subtract in and all of its other uses are redirected to the
/// new load / store.
bool DAGCombiner::CombineToPreIndexedLoadStore(SDNode *N) {
  if (Level < AfterLegalizeDAG)
    return false;

  bool isLoad = true;
  SDValue Ptr;
  EVT VT;
  if (LoadSDNode *LD  = dyn_cast<LoadSDNode>(N)) {
    if (LD->isIndexed())
      return false;
    VT = LD->getMemoryVT();
    if (!TLI.isIndexedLoadLegal(ISD::PRE_INC, VT) &&
        !TLI.isIndexedLoadLegal(ISD::PRE_DEC, VT))
      return false;
    Ptr = LD->getBasePtr();
  } else if (StoreSDNode *ST  = dyn_cast<StoreSDNode>(N)) {
    if (ST->isIndexed())
      return false;
    VT = ST->getMemoryVT();
    if (!TLI.isIndexedStoreLegal(ISD::PRE_INC, VT) &&
        !TLI.isIndexedStoreLegal(ISD::PRE_DEC, VT))
      return false;
    Ptr = ST->getBasePtr();
    isLoad = false;
  } else {
    return false;
  }

  // If the pointer is not an add/sub, or if it doesn't have multiple uses, bail
  // out.  There is no reason to make this a preinc/predec.
  if ((Ptr.getOpcode() != ISD::ADD && Ptr.getOpcode() != ISD::SUB) ||
      Ptr.getNode()->hasOneUse())
    return false;

  // Ask the target to do addressing mode selection.
  SDValue BasePtr;
  SDValue Offset;
  ISD::MemIndexedMode AM = ISD::UNINDEXED;
  if (!TLI.getPreIndexedAddressParts(N, BasePtr, Offset, AM, DAG))
    return false;
  // Don't create a indexed load / store with zero offset.
  if (isa<ConstantSDNode>(Offset) &&
      cast<ConstantSDNode>(Offset)->isNullValue())
    return false;

  // Try turning it into a pre-indexed load / store except when:
  // 1) The new base ptr is a frame index.
  // 2) If N is a store and the new base ptr is either the same as or is a
  //    predecessor of the value being stored.
  // 3) Another use of old base ptr is a predecessor of N. If ptr is folded
  //    that would create a cycle.
  // 4) All uses are load / store ops that use it as old base ptr.

  // Check #1.  Preinc'ing a frame index would require copying the stack pointer
  // (plus the implicit offset) to a register to preinc anyway.
  if (isa<FrameIndexSDNode>(BasePtr) || isa<RegisterSDNode>(BasePtr))
    return false;

  // Check #2.
  if (!isLoad) {
    SDValue Val = cast<StoreSDNode>(N)->getValue();
    if (Val == BasePtr || BasePtr.getNode()->isPredecessorOf(Val.getNode()))
      return false;
  }

  // Now check for #3 and #4.
  bool RealUse = false;

  // Caches for hasPredecessorHelper
  SmallPtrSet<const SDNode *, 32> Visited;
  SmallVector<const SDNode *, 16> Worklist;

  for (SDNode::use_iterator I = Ptr.getNode()->use_begin(),
         E = Ptr.getNode()->use_end(); I != E; ++I) {
    SDNode *Use = *I;
    if (Use == N)
      continue;
    if (N->hasPredecessorHelper(Use, Visited, Worklist))
      return false;

    // If Ptr may be folded in addressing mode of other use, then it's
    // not profitable to do this transformation.
    if (!canFoldInAddressingMode(Ptr.getNode(), Use, DAG, TLI))
      RealUse = true;
  }

  if (!RealUse)
    return false;

  SDValue Result;
  if (isLoad)
    Result = DAG.getIndexedLoad(SDValue(N,0), N->getDebugLoc(),
                                BasePtr, Offset, AM);
  else
    Result = DAG.getIndexedStore(SDValue(N,0), N->getDebugLoc(),
                                 BasePtr, Offset, AM);
  ++PreIndexedNodes;
  ++NodesCombined;
  DEBUG(dbgs() << "\nReplacing.4 ";
        N->dump(&DAG);
        dbgs() << "\nWith: ";
        Result.getNode()->dump(&DAG);
        dbgs() << '\n');
  WorkListRemover DeadNodes(*this);
  if (isLoad) {
    DAG.ReplaceAllUsesOfValueWith(SDValue(N, 0), Result.getValue(0));
    DAG.ReplaceAllUsesOfValueWith(SDValue(N, 1), Result.getValue(2));
  } else {
    DAG.ReplaceAllUsesOfValueWith(SDValue(N, 0), Result.getValue(1));
  }

  // Finally, since the node is now dead, remove it from the graph.
  DAG.DeleteNode(N);

  // Replace the uses of Ptr with uses of the updated base value.
  DAG.ReplaceAllUsesOfValueWith(Ptr, Result.getValue(isLoad ? 1 : 0));
  removeFromWorkList(Ptr.getNode());
  DAG.DeleteNode(Ptr.getNode());

  return true;
}

/// CombineToPostIndexedLoadStore - Try to combine a load / store with a
/// add / sub of the base pointer node into a post-indexed load / store.
/// The transformation folded the add / subtract into the new indexed
/// load / store effectively and all of its uses are redirected to the
/// new load / store.
bool DAGCombiner::CombineToPostIndexedLoadStore(SDNode *N) {
  if (Level < AfterLegalizeDAG)
    return false;

  bool isLoad = true;
  SDValue Ptr;
  EVT VT;
  if (LoadSDNode *LD  = dyn_cast<LoadSDNode>(N)) {
    if (LD->isIndexed())
      return false;
    VT = LD->getMemoryVT();
    if (!TLI.isIndexedLoadLegal(ISD::POST_INC, VT) &&
        !TLI.isIndexedLoadLegal(ISD::POST_DEC, VT))
      return false;
    Ptr = LD->getBasePtr();
  } else if (StoreSDNode *ST  = dyn_cast<StoreSDNode>(N)) {
    if (ST->isIndexed())
      return false;
    VT = ST->getMemoryVT();
    if (!TLI.isIndexedStoreLegal(ISD::POST_INC, VT) &&
        !TLI.isIndexedStoreLegal(ISD::POST_DEC, VT))
      return false;
    Ptr = ST->getBasePtr();
    isLoad = false;
  } else {
    return false;
  }

  if (Ptr.getNode()->hasOneUse())
    return false;

  for (SDNode::use_iterator I = Ptr.getNode()->use_begin(),
         E = Ptr.getNode()->use_end(); I != E; ++I) {
    SDNode *Op = *I;
    if (Op == N ||
        (Op->getOpcode() != ISD::ADD && Op->getOpcode() != ISD::SUB))
      continue;

    SDValue BasePtr;
    SDValue Offset;
    ISD::MemIndexedMode AM = ISD::UNINDEXED;
    if (TLI.getPostIndexedAddressParts(N, Op, BasePtr, Offset, AM, DAG)) {
      // Don't create a indexed load / store with zero offset.
      if (isa<ConstantSDNode>(Offset) &&
          cast<ConstantSDNode>(Offset)->isNullValue())
        continue;

      // Try turning it into a post-indexed load / store except when
      // 1) All uses are load / store ops that use it as base ptr (and
      //    it may be folded as addressing mmode).
      // 2) Op must be independent of N, i.e. Op is neither a predecessor
      //    nor a successor of N. Otherwise, if Op is folded that would
      //    create a cycle.

      if (isa<FrameIndexSDNode>(BasePtr) || isa<RegisterSDNode>(BasePtr))
        continue;

      // Check for #1.
      bool TryNext = false;
      for (SDNode::use_iterator II = BasePtr.getNode()->use_begin(),
             EE = BasePtr.getNode()->use_end(); II != EE; ++II) {
        SDNode *Use = *II;
        if (Use == Ptr.getNode())
          continue;

        // If all the uses are load / store addresses, then don't do the
        // transformation.
        if (Use->getOpcode() == ISD::ADD || Use->getOpcode() == ISD::SUB){
          bool RealUse = false;
          for (SDNode::use_iterator III = Use->use_begin(),
                 EEE = Use->use_end(); III != EEE; ++III) {
            SDNode *UseUse = *III;
            if (!canFoldInAddressingMode(Use, UseUse, DAG, TLI)) 
              RealUse = true;
          }

          if (!RealUse) {
            TryNext = true;
            break;
          }
        }
      }

      if (TryNext)
        continue;

      // Check for #2
      if (!Op->isPredecessorOf(N) && !N->isPredecessorOf(Op)) {
        SDValue Result = isLoad
          ? DAG.getIndexedLoad(SDValue(N,0), N->getDebugLoc(),
                               BasePtr, Offset, AM)
          : DAG.getIndexedStore(SDValue(N,0), N->getDebugLoc(),
                                BasePtr, Offset, AM);
        ++PostIndexedNodes;
        ++NodesCombined;
        DEBUG(dbgs() << "\nReplacing.5 ";
              N->dump(&DAG);
              dbgs() << "\nWith: ";
              Result.getNode()->dump(&DAG);
              dbgs() << '\n');
        WorkListRemover DeadNodes(*this);
        if (isLoad) {
          DAG.ReplaceAllUsesOfValueWith(SDValue(N, 0), Result.getValue(0));
          DAG.ReplaceAllUsesOfValueWith(SDValue(N, 1), Result.getValue(2));
        } else {
          DAG.ReplaceAllUsesOfValueWith(SDValue(N, 0), Result.getValue(1));
        }

        // Finally, since the node is now dead, remove it from the graph.
        DAG.DeleteNode(N);

        // Replace the uses of Use with uses of the updated base value.
        DAG.ReplaceAllUsesOfValueWith(SDValue(Op, 0),
                                      Result.getValue(isLoad ? 1 : 0));
        removeFromWorkList(Op);
        DAG.DeleteNode(Op);
        return true;
      }
    }
  }

  return false;
}

SDValue DAGCombiner::visitLOAD(SDNode *N) {
  LoadSDNode *LD  = cast<LoadSDNode>(N);
  SDValue Chain = LD->getChain();
  SDValue Ptr   = LD->getBasePtr();

  // If load is not volatile and there are no uses of the loaded value (and
  // the updated indexed value in case of indexed loads), change uses of the
  // chain value into uses of the chain input (i.e. delete the dead load).
  if (!LD->isVolatile()) {
    if (N->getValueType(1) == MVT::Other) {
      // Unindexed loads.
      if (!N->hasAnyUseOfValue(0)) {
        // It's not safe to use the two value CombineTo variant here. e.g.
        // v1, chain2 = load chain1, loc
        // v2, chain3 = load chain2, loc
        // v3         = add v2, c
        // Now we replace use of chain2 with chain1.  This makes the second load
        // isomorphic to the one we are deleting, and thus makes this load live.
        DEBUG(dbgs() << "\nReplacing.6 ";
              N->dump(&DAG);
              dbgs() << "\nWith chain: ";
              Chain.getNode()->dump(&DAG);
              dbgs() << "\n");
        WorkListRemover DeadNodes(*this);
        DAG.ReplaceAllUsesOfValueWith(SDValue(N, 1), Chain);

        if (N->use_empty()) {
          removeFromWorkList(N);
          DAG.DeleteNode(N);
        }

        return SDValue(N, 0);   // Return N so it doesn't get rechecked!
      }
    } else {
      // Indexed loads.
      assert(N->getValueType(2) == MVT::Other && "Malformed indexed loads?");
      if (!N->hasAnyUseOfValue(0) && !N->hasAnyUseOfValue(1)) {
        SDValue Undef = DAG.getUNDEF(N->getValueType(0));
        DEBUG(dbgs() << "\nReplacing.7 ";
              N->dump(&DAG);
              dbgs() << "\nWith: ";
              Undef.getNode()->dump(&DAG);
              dbgs() << " and 2 other values\n");
        WorkListRemover DeadNodes(*this);
        DAG.ReplaceAllUsesOfValueWith(SDValue(N, 0), Undef);
        DAG.ReplaceAllUsesOfValueWith(SDValue(N, 1),
                                      DAG.getUNDEF(N->getValueType(1)));
        DAG.ReplaceAllUsesOfValueWith(SDValue(N, 2), Chain);
        removeFromWorkList(N);
        DAG.DeleteNode(N);
        return SDValue(N, 0);   // Return N so it doesn't get rechecked!
      }
    }
  }

  // If this load is directly stored, replace the load value with the stored
  // value.
  // TODO: Handle store large -> read small portion.
  // TODO: Handle TRUNCSTORE/LOADEXT
  if (ISD::isNormalLoad(N) && !LD->isVolatile()) {
    if (ISD::isNON_TRUNCStore(Chain.getNode())) {
      StoreSDNode *PrevST = cast<StoreSDNode>(Chain);
      if (PrevST->getBasePtr() == Ptr &&
          PrevST->getValue().getValueType() == N->getValueType(0))
      return CombineTo(N, Chain.getOperand(1), Chain);
    }
  }

  // Try to infer better alignment information than the load already has.
  if (OptLevel != CodeGenOpt::None && LD->isUnindexed()) {
    if (unsigned Align = DAG.InferPtrAlignment(Ptr)) {
      if (Align > LD->getAlignment())
        return DAG.getExtLoad(LD->getExtensionType(), N->getDebugLoc(),
                              LD->getValueType(0),
                              Chain, Ptr, LD->getPointerInfo(),
                              LD->getMemoryVT(),
                              LD->isVolatile(), LD->isNonTemporal(), Align);
    }
  }

  if (CombinerAA) {
    // Walk up chain skipping non-aliasing memory nodes.
    SDValue BetterChain = FindBetterChain(N, Chain);

    // If there is a better chain.
    if (Chain != BetterChain) {
      SDValue ReplLoad;

      // Replace the chain to void dependency.
      if (LD->getExtensionType() == ISD::NON_EXTLOAD) {
        ReplLoad = DAG.getLoad(N->getValueType(0), LD->getDebugLoc(),
                               BetterChain, Ptr, LD->getPointerInfo(),
                               LD->isVolatile(), LD->isNonTemporal(),
                               LD->isInvariant(), LD->getAlignment());
      } else {
        ReplLoad = DAG.getExtLoad(LD->getExtensionType(), LD->getDebugLoc(),
                                  LD->getValueType(0),
                                  BetterChain, Ptr, LD->getPointerInfo(),
                                  LD->getMemoryVT(),
                                  LD->isVolatile(),
                                  LD->isNonTemporal(),
                                  LD->getAlignment());
      }

      // Create token factor to keep old chain connected.
      SDValue Token = DAG.getNode(ISD::TokenFactor, N->getDebugLoc(),
                                  MVT::Other, Chain, ReplLoad.getValue(1));

      // Make sure the new and old chains are cleaned up.
      AddToWorkList(Token.getNode());

      // Replace uses with load result and token factor. Don't add users
      // to work list.
      return CombineTo(N, ReplLoad.getValue(0), Token, false);
    }
  }

  // Try transforming N to an indexed load.
  if (CombineToPreIndexedLoadStore(N) || CombineToPostIndexedLoadStore(N))
    return SDValue(N, 0);

  return SDValue();
}

/// CheckForMaskedLoad - Check to see if V is (and load (ptr), imm), where the
/// load is having specific bytes cleared out.  If so, return the byte size
/// being masked out and the shift amount.
static std::pair<unsigned, unsigned>
CheckForMaskedLoad(SDValue V, SDValue Ptr, SDValue Chain) {
  std::pair<unsigned, unsigned> Result(0, 0);

  // Check for the structure we're looking for.
  if (V->getOpcode() != ISD::AND ||
      !isa<ConstantSDNode>(V->getOperand(1)) ||
      !ISD::isNormalLoad(V->getOperand(0).getNode()))
    return Result;

  // Check the chain and pointer.
  LoadSDNode *LD = cast<LoadSDNode>(V->getOperand(0));
  if (LD->getBasePtr() != Ptr) return Result;  // Not from same pointer.

  // The store should be chained directly to the load or be an operand of a
  // tokenfactor.
  if (LD == Chain.getNode())
    ; // ok.
  else if (Chain->getOpcode() != ISD::TokenFactor)
    return Result; // Fail.
  else {
    bool isOk = false;
    for (unsigned i = 0, e = Chain->getNumOperands(); i != e; ++i)
      if (Chain->getOperand(i).getNode() == LD) {
        isOk = true;
        break;
      }
    if (!isOk) return Result;
  }

  // This only handles simple types.
  if (V.getValueType() != MVT::i16 &&
      V.getValueType() != MVT::i32 &&
      V.getValueType() != MVT::i64)
    return Result;

  // Check the constant mask.  Invert it so that the bits being masked out are
  // 0 and the bits being kept are 1.  Use getSExtValue so that leading bits
  // follow the sign bit for uniformity.
  uint64_t NotMask = ~cast<ConstantSDNode>(V->getOperand(1))->getSExtValue();
  unsigned NotMaskLZ = CountLeadingZeros_64(NotMask);
  if (NotMaskLZ & 7) return Result;  // Must be multiple of a byte.
  unsigned NotMaskTZ = CountTrailingZeros_64(NotMask);
  if (NotMaskTZ & 7) return Result;  // Must be multiple of a byte.
  if (NotMaskLZ == 64) return Result;  // All zero mask.

  // See if we have a continuous run of bits.  If so, we have 0*1+0*
  if (CountTrailingOnes_64(NotMask >> NotMaskTZ)+NotMaskTZ+NotMaskLZ != 64)
    return Result;

  // Adjust NotMaskLZ down to be from the actual size of the int instead of i64.
  if (V.getValueType() != MVT::i64 && NotMaskLZ)
    NotMaskLZ -= 64-V.getValueSizeInBits();

  unsigned MaskedBytes = (V.getValueSizeInBits()-NotMaskLZ-NotMaskTZ)/8;
  switch (MaskedBytes) {
  case 1:
  case 2:
  case 4: break;
  default: return Result; // All one mask, or 5-byte mask.
  }

  // Verify that the first bit starts at a multiple of mask so that the access
  // is aligned the same as the access width.
  if (NotMaskTZ && NotMaskTZ/8 % MaskedBytes) return Result;

  Result.first = MaskedBytes;
  Result.second = NotMaskTZ/8;
  return Result;
}


/// ShrinkLoadReplaceStoreWithStore - Check to see if IVal is something that
/// provides a value as specified by MaskInfo.  If so, replace the specified
/// store with a narrower store of truncated IVal.
static SDNode *
ShrinkLoadReplaceStoreWithStore(const std::pair<unsigned, unsigned> &MaskInfo,
                                SDValue IVal, StoreSDNode *St,
                                DAGCombiner *DC) {
  unsigned NumBytes = MaskInfo.first;
  unsigned ByteShift = MaskInfo.second;
  SelectionDAG &DAG = DC->getDAG();

  // Check to see if IVal is all zeros in the part being masked in by the 'or'
  // that uses this.  If not, this is not a replacement.
  APInt Mask = ~APInt::getBitsSet(IVal.getValueSizeInBits(),
                                  ByteShift*8, (ByteShift+NumBytes)*8);
  if (!DAG.MaskedValueIsZero(IVal, Mask)) return 0;

  // Check that it is legal on the target to do this.  It is legal if the new
  // VT we're shrinking to (i8/i16/i32) is legal or we're still before type
  // legalization.
  MVT VT = MVT::getIntegerVT(NumBytes*8);
  if (!DC->isTypeLegal(VT))
    return 0;

  // Okay, we can do this!  Replace the 'St' store with a store of IVal that is
  // shifted by ByteShift and truncated down to NumBytes.
  if (ByteShift)
    IVal = DAG.getNode(ISD::SRL, IVal->getDebugLoc(), IVal.getValueType(), IVal,
                       DAG.getConstant(ByteShift*8,
                                    DC->getShiftAmountTy(IVal.getValueType())));

  // Figure out the offset for the store and the alignment of the access.
  unsigned StOffset;
  unsigned NewAlign = St->getAlignment();

  if (DAG.getTargetLoweringInfo().isLittleEndian())
    StOffset = ByteShift;
  else
    StOffset = IVal.getValueType().getStoreSize() - ByteShift - NumBytes;

  SDValue Ptr = St->getBasePtr();
  if (StOffset) {
    Ptr = DAG.getNode(ISD::ADD, IVal->getDebugLoc(), Ptr.getValueType(),
                      Ptr, DAG.getConstant(StOffset, Ptr.getValueType()));
    NewAlign = MinAlign(NewAlign, StOffset);
  }

  // Truncate down to the new size.
  IVal = DAG.getNode(ISD::TRUNCATE, IVal->getDebugLoc(), VT, IVal);

  ++OpsNarrowed;
  return DAG.getStore(St->getChain(), St->getDebugLoc(), IVal, Ptr,
                      St->getPointerInfo().getWithOffset(StOffset),
                      false, false, NewAlign).getNode();
}


/// ReduceLoadOpStoreWidth - Look for sequence of load / op / store where op is
/// one of 'or', 'xor', and 'and' of immediates. If 'op' is only touching some
/// of the loaded bits, try narrowing the load and store if it would end up
/// being a win for performance or code size.
SDValue DAGCombiner::ReduceLoadOpStoreWidth(SDNode *N) {
  StoreSDNode *ST  = cast<StoreSDNode>(N);
  if (ST->isVolatile())
    return SDValue();

  SDValue Chain = ST->getChain();
  SDValue Value = ST->getValue();
  SDValue Ptr   = ST->getBasePtr();
  EVT VT = Value.getValueType();

  if (ST->isTruncatingStore() || VT.isVector() || !Value.hasOneUse())
    return SDValue();

  unsigned Opc = Value.getOpcode();

  // If this is "store (or X, Y), P" and X is "(and (load P), cst)", where cst
  // is a byte mask indicating a consecutive number of bytes, check to see if
  // Y is known to provide just those bytes.  If so, we try to replace the
  // load + replace + store sequence with a single (narrower) store, which makes
  // the load dead.
  if (Opc == ISD::OR) {
    std::pair<unsigned, unsigned> MaskedLoad;
    MaskedLoad = CheckForMaskedLoad(Value.getOperand(0), Ptr, Chain);
    if (MaskedLoad.first)
      if (SDNode *NewST = ShrinkLoadReplaceStoreWithStore(MaskedLoad,
                                                  Value.getOperand(1), ST,this))
        return SDValue(NewST, 0);

    // Or is commutative, so try swapping X and Y.
    MaskedLoad = CheckForMaskedLoad(Value.getOperand(1), Ptr, Chain);
    if (MaskedLoad.first)
      if (SDNode *NewST = ShrinkLoadReplaceStoreWithStore(MaskedLoad,
                                                  Value.getOperand(0), ST,this))
        return SDValue(NewST, 0);
  }

  if ((Opc != ISD::OR && Opc != ISD::XOR && Opc != ISD::AND) ||
      Value.getOperand(1).getOpcode() != ISD::Constant)
    return SDValue();

  SDValue N0 = Value.getOperand(0);
  if (ISD::isNormalLoad(N0.getNode()) && N0.hasOneUse() &&
      Chain == SDValue(N0.getNode(), 1)) {
    LoadSDNode *LD = cast<LoadSDNode>(N0);
    if (LD->getBasePtr() != Ptr ||
        LD->getPointerInfo().getAddrSpace() !=
        ST->getPointerInfo().getAddrSpace())
      return SDValue();

    // Find the type to narrow it the load / op / store to.
    SDValue N1 = Value.getOperand(1);
    unsigned BitWidth = N1.getValueSizeInBits();
    APInt Imm = cast<ConstantSDNode>(N1)->getAPIntValue();
    if (Opc == ISD::AND)
      Imm ^= APInt::getAllOnesValue(BitWidth);
    if (Imm == 0 || Imm.isAllOnesValue())
      return SDValue();
    unsigned ShAmt = Imm.countTrailingZeros();
    unsigned MSB = BitWidth - Imm.countLeadingZeros() - 1;
    unsigned NewBW = NextPowerOf2(MSB - ShAmt);
    EVT NewVT = EVT::getIntegerVT(*DAG.getContext(), NewBW);
    while (NewBW < BitWidth &&
           !(TLI.isOperationLegalOrCustom(Opc, NewVT) &&
             TLI.isNarrowingProfitable(VT, NewVT))) {
      NewBW = NextPowerOf2(NewBW);
      NewVT = EVT::getIntegerVT(*DAG.getContext(), NewBW);
    }
    if (NewBW >= BitWidth)
      return SDValue();

    // If the lsb changed does not start at the type bitwidth boundary,
    // start at the previous one.
    if (ShAmt % NewBW)
      ShAmt = (((ShAmt + NewBW - 1) / NewBW) * NewBW) - NewBW;
    APInt Mask = APInt::getBitsSet(BitWidth, ShAmt, ShAmt + NewBW);
    if ((Imm & Mask) == Imm) {
      APInt NewImm = (Imm & Mask).lshr(ShAmt).trunc(NewBW);
      if (Opc == ISD::AND)
        NewImm ^= APInt::getAllOnesValue(NewBW);
      uint64_t PtrOff = ShAmt / 8;
      // For big endian targets, we need to adjust the offset to the pointer to
      // load the correct bytes.
      if (TLI.isBigEndian())
        PtrOff = (BitWidth + 7 - NewBW) / 8 - PtrOff;

      unsigned NewAlign = MinAlign(LD->getAlignment(), PtrOff);
      Type *NewVTTy = NewVT.getTypeForEVT(*DAG.getContext());
      if (NewAlign < TLI.getDataLayout()->getABITypeAlignment(NewVTTy))
        return SDValue();

      SDValue NewPtr = DAG.getNode(ISD::ADD, LD->getDebugLoc(),
                                   Ptr.getValueType(), Ptr,
                                   DAG.getConstant(PtrOff, Ptr.getValueType()));
      SDValue NewLD = DAG.getLoad(NewVT, N0.getDebugLoc(),
                                  LD->getChain(), NewPtr,
                                  LD->getPointerInfo().getWithOffset(PtrOff),
                                  LD->isVolatile(), LD->isNonTemporal(),
                                  LD->isInvariant(), NewAlign);
      SDValue NewVal = DAG.getNode(Opc, Value.getDebugLoc(), NewVT, NewLD,
                                   DAG.getConstant(NewImm, NewVT));
      SDValue NewST = DAG.getStore(Chain, N->getDebugLoc(),
                                   NewVal, NewPtr,
                                   ST->getPointerInfo().getWithOffset(PtrOff),
                                   false, false, NewAlign);

      AddToWorkList(NewPtr.getNode());
      AddToWorkList(NewLD.getNode());
      AddToWorkList(NewVal.getNode());
      WorkListRemover DeadNodes(*this);
      DAG.ReplaceAllUsesOfValueWith(N0.getValue(1), NewLD.getValue(1));
      ++OpsNarrowed;
      return NewST;
    }
  }

  return SDValue();
}

/// TransformFPLoadStorePair - For a given floating point load / store pair,
/// if the load value isn't used by any other operations, then consider
/// transforming the pair to integer load / store operations if the target
/// deems the transformation profitable.
SDValue DAGCombiner::TransformFPLoadStorePair(SDNode *N) {
  StoreSDNode *ST  = cast<StoreSDNode>(N);
  SDValue Chain = ST->getChain();
  SDValue Value = ST->getValue();
  if (ISD::isNormalStore(ST) && ISD::isNormalLoad(Value.getNode()) &&
      Value.hasOneUse() &&
      Chain == SDValue(Value.getNode(), 1)) {
    LoadSDNode *LD = cast<LoadSDNode>(Value);
    EVT VT = LD->getMemoryVT();
    if (!VT.isFloatingPoint() ||
        VT != ST->getMemoryVT() ||
        LD->isNonTemporal() ||
        ST->isNonTemporal() ||
        LD->getPointerInfo().getAddrSpace() != 0 ||
        ST->getPointerInfo().getAddrSpace() != 0)
      return SDValue();

    EVT IntVT = EVT::getIntegerVT(*DAG.getContext(), VT.getSizeInBits());
    if (!TLI.isOperationLegal(ISD::LOAD, IntVT) ||
        !TLI.isOperationLegal(ISD::STORE, IntVT) ||
        !TLI.isDesirableToTransformToIntegerOp(ISD::LOAD, VT) ||
        !TLI.isDesirableToTransformToIntegerOp(ISD::STORE, VT))
      return SDValue();

    unsigned LDAlign = LD->getAlignment();
    unsigned STAlign = ST->getAlignment();
    Type *IntVTTy = IntVT.getTypeForEVT(*DAG.getContext());
    unsigned ABIAlign = TLI.getDataLayout()->getABITypeAlignment(IntVTTy);
    if (LDAlign < ABIAlign || STAlign < ABIAlign)
      return SDValue();

    SDValue NewLD = DAG.getLoad(IntVT, Value.getDebugLoc(),
                                LD->getChain(), LD->getBasePtr(),
                                LD->getPointerInfo(),
                                false, false, false, LDAlign);

    SDValue NewST = DAG.getStore(NewLD.getValue(1), N->getDebugLoc(),
                                 NewLD, ST->getBasePtr(),
                                 ST->getPointerInfo(),
                                 false, false, STAlign);

    AddToWorkList(NewLD.getNode());
    AddToWorkList(NewST.getNode());
    WorkListRemover DeadNodes(*this);
    DAG.ReplaceAllUsesOfValueWith(Value.getValue(1), NewLD.getValue(1));
    ++LdStFP2Int;
    return NewST;
  }

  return SDValue();
}

/// Returns the base pointer and an integer offset from that object.
static std::pair<SDValue, int64_t> GetPointerBaseAndOffset(SDValue Ptr) {
  if (Ptr->getOpcode() == ISD::ADD && isa<ConstantSDNode>(Ptr->getOperand(1))) {
    int64_t Offset = cast<ConstantSDNode>(Ptr->getOperand(1))->getSExtValue();
    SDValue Base = Ptr->getOperand(0);
    return std::make_pair(Base, Offset);
  }

  return std::make_pair(Ptr, 0);
}

/// Holds a pointer to an LSBaseSDNode as well as information on where it
/// is located in a sequence of memory operations connected by a chain.
struct MemOpLink {
  MemOpLink (LSBaseSDNode *N, int64_t Offset, unsigned Seq):
    MemNode(N), OffsetFromBase(Offset), SequenceNum(Seq) { }
  // Ptr to the mem node.
  LSBaseSDNode *MemNode;
  // Offset from the base ptr.
  int64_t OffsetFromBase;
  // What is the sequence number of this mem node.
  // Lowest mem operand in the DAG starts at zero.
  unsigned SequenceNum;
};

/// Sorts store nodes in a link according to their offset from a shared
// base ptr.
struct ConsecutiveMemoryChainSorter {
  bool operator()(MemOpLink LHS, MemOpLink RHS) {
    return LHS.OffsetFromBase < RHS.OffsetFromBase;
  }
};

bool DAGCombiner::MergeConsecutiveStores(StoreSDNode* St) {
  EVT MemVT = St->getMemoryVT();
  int64_t ElementSizeBytes = MemVT.getSizeInBits()/8;

  // Don't merge vectors into wider inputs.
  if (MemVT.isVector() || !MemVT.isSimple())
    return false;

  // Perform an early exit check. Do not bother looking at stored values that
  // are not constants or loads.
  SDValue StoredVal = St->getValue();
  bool IsLoadSrc = isa<LoadSDNode>(StoredVal);
  if (!isa<ConstantSDNode>(StoredVal) && !isa<ConstantFPSDNode>(StoredVal) &&
      !IsLoadSrc)
    return false;

  // Only look at ends of store sequences.
  SDValue Chain = SDValue(St, 1);
  if (Chain->hasOneUse() && Chain->use_begin()->getOpcode() == ISD::STORE)
    return false;

  // This holds the base pointer and the offset in bytes from the base pointer.
  std::pair<SDValue, int64_t> BasePtr =
      GetPointerBaseAndOffset(St->getBasePtr());

  // We must have a base and an offset.
  if (!BasePtr.first.getNode())
    return false;

  // Do not handle stores to undef base pointers.
  if (BasePtr.first.getOpcode() == ISD::UNDEF)
    return false;

  SmallVector<MemOpLink, 8> StoreNodes;
  // Walk up the chain and look for nodes with offsets from the same
  // base pointer. Stop when reaching an instruction with a different kind
  // or instruction which has a different base pointer.
  unsigned Seq = 0;
  StoreSDNode *Index = St;
  while (Index) {
    // If the chain has more than one use, then we can't reorder the mem ops.
    if (Index != St && !SDValue(Index, 1)->hasOneUse())
      break;

    // Find the base pointer and offset for this memory node.
    std::pair<SDValue, int64_t> Ptr =
      GetPointerBaseAndOffset(Index->getBasePtr());

    // Check that the base pointer is the same as the original one.
    if (Ptr.first.getNode() != BasePtr.first.getNode())
      break;

    // Check that the alignment is the same.
    if (Index->getAlignment() != St->getAlignment())
      break;

    // The memory operands must not be volatile.
    if (Index->isVolatile() || Index->isIndexed())
      break;

    // No truncation.
    if (StoreSDNode *St = dyn_cast<StoreSDNode>(Index))
      if (St->isTruncatingStore())
        break;

    // The stored memory type must be the same.
    if (Index->getMemoryVT() != MemVT)
      break;

    // We do not allow unaligned stores because we want to prevent overriding
    // stores.
    if (Index->getAlignment()*8 != MemVT.getSizeInBits())
      break;

    // We found a potential memory operand to merge.
    StoreNodes.push_back(MemOpLink(Index, Ptr.second, Seq++));

    // Move up the chain to the next memory operation.
    Index = dyn_cast<StoreSDNode>(Index->getChain().getNode());
  }

  // Check if there is anything to merge.
  if (StoreNodes.size() < 2)
    return false;

  // Sort the memory operands according to their distance from the base pointer.
  std::sort(StoreNodes.begin(), StoreNodes.end(),
            ConsecutiveMemoryChainSorter());

  // Scan the memory operations on the chain and find the first non-consecutive
  // store memory address.
  unsigned LastConsecutiveStore = 0;
  int64_t StartAddress = StoreNodes[0].OffsetFromBase;
  for (unsigned i=1; i<StoreNodes.size(); ++i) {
    int64_t CurrAddress = StoreNodes[i].OffsetFromBase;
    if (CurrAddress - StartAddress != (ElementSizeBytes * i))
      break;

    // Mark this node as useful.
    LastConsecutiveStore = i;
  }

  // The node with the lowest store address.
  LSBaseSDNode *FirstInChain = StoreNodes[0].MemNode;

  // Store the constants into memory as one consecutive store.
  if (!IsLoadSrc) {
    unsigned LastLegalType = 0;
    unsigned LastLegalVectorType = 0;
    bool NonZero = false;
    for (unsigned i=0; i<LastConsecutiveStore+1; ++i) {
      StoreSDNode *St  = cast<StoreSDNode>(StoreNodes[i].MemNode);
      SDValue StoredVal = St->getValue();

      if (ConstantSDNode *C = dyn_cast<ConstantSDNode>(StoredVal)) {
        NonZero |= !C->isNullValue();
      } else if (ConstantFPSDNode *C = dyn_cast<ConstantFPSDNode>(StoredVal)) {
        NonZero |= !C->getConstantFPValue()->isNullValue();
      } else {
        // Non constant.
        break;
      }

      // Find a legal type for the constant store.
      unsigned StoreBW = (i+1) * ElementSizeBytes * 8;
      EVT StoreTy = EVT::getIntegerVT(*DAG.getContext(), StoreBW);
      if (TLI.isTypeLegal(StoreTy))
        LastLegalType = i+1;

      // Find a legal type for the vector store.
      EVT Ty = EVT::getVectorVT(*DAG.getContext(), MemVT, i+1);
      if (TLI.isTypeLegal(Ty))
        LastLegalVectorType = i + 1;
    }

    // We only use vectors if the constant is known to be zero.
    if (NonZero)
      LastLegalVectorType = 0;

    // Check if we found a legal integer type to store.
    if (LastLegalType == 0 && LastLegalVectorType == 0)
      return false;

    bool UseVector = LastLegalVectorType > LastLegalType;
    unsigned NumElem = UseVector ? LastLegalVectorType : LastLegalType;

    // Make sure we have something to merge.
    if (NumElem < 2)
      return false;

    unsigned EarliestNodeUsed = 0;
    for (unsigned i=0; i < NumElem; ++i) {
      // Find a chain for the new wide-store operand. Notice that some
      // of the store nodes that we found may not be selected for inclusion
      // in the wide store. The chain we use needs to be the chain of the
      // earliest store node which is *used* and replaced by the wide store.
      if (StoreNodes[i].SequenceNum > StoreNodes[EarliestNodeUsed].SequenceNum)
        EarliestNodeUsed = i;
    }

    // The earliest Node in the DAG.
    LSBaseSDNode *EarliestOp = StoreNodes[EarliestNodeUsed].MemNode;
    DebugLoc DL = StoreNodes[0].MemNode->getDebugLoc();

    SDValue StoredVal;
    if (UseVector) {
      // Find a legal type for the vector store.
      EVT Ty = EVT::getVectorVT(*DAG.getContext(), MemVT, NumElem);
      assert(TLI.isTypeLegal(Ty) && "Illegal vector store");
      StoredVal = DAG.getConstant(0, Ty);
    } else {
      unsigned StoreBW = NumElem * ElementSizeBytes * 8;
      APInt StoreInt(StoreBW, 0);

      // Construct a single integer constant which is made of the smaller
      // constant inputs.
      bool IsLE = TLI.isLittleEndian();
      for (unsigned i = 0; i < NumElem ; ++i) {
        unsigned Idx = IsLE ?(NumElem - 1 - i) : i;
        StoreSDNode *St  = cast<StoreSDNode>(StoreNodes[Idx].MemNode);
        SDValue Val = St->getValue();
        StoreInt<<=ElementSizeBytes*8;
        if (ConstantSDNode *C = dyn_cast<ConstantSDNode>(Val)) {
          StoreInt|=C->getAPIntValue().zext(StoreBW);
        } else if (ConstantFPSDNode *C = dyn_cast<ConstantFPSDNode>(Val)) {
          StoreInt|= C->getValueAPF().bitcastToAPInt().zext(StoreBW);
        } else {
          assert(false && "Invalid constant element type");
        }
      }

      // Create the new Load and Store operations.
      EVT StoreTy = EVT::getIntegerVT(*DAG.getContext(), StoreBW);
      StoredVal = DAG.getConstant(StoreInt, StoreTy);
    }

    SDValue NewStore = DAG.getStore(EarliestOp->getChain(), DL, StoredVal,
                                    FirstInChain->getBasePtr(),
                                    FirstInChain->getPointerInfo(),
                                    false, false,
                                    FirstInChain->getAlignment());

    // Replace the first store with the new store
    CombineTo(EarliestOp, NewStore);
    // Erase all other stores.
    for (unsigned i = 0; i < NumElem ; ++i) {
      if (StoreNodes[i].MemNode == EarliestOp)
        continue;
      StoreSDNode *St = cast<StoreSDNode>(StoreNodes[i].MemNode);
      DAG.ReplaceAllUsesOfValueWith(SDValue(St, 0), St->getChain());
      removeFromWorkList(St);
      DAG.DeleteNode(St);
    }

    return true;
  }

  // Below we handle the case of multiple consecutive stores that
  // come from multiple consecutive loads. We merge them into a single
  // wide load and a single wide store.

  // Look for load nodes which are used by the stored values.
  SmallVector<MemOpLink, 8> LoadNodes;

  // Find acceptable loads. Loads need to have the same chain (token factor),
  // must not be zext, volatile, indexed, and they must be consecutive.
  SDValue LdBasePtr;
  for (unsigned i=0; i<LastConsecutiveStore+1; ++i) {
    StoreSDNode *St  = cast<StoreSDNode>(StoreNodes[i].MemNode);
    LoadSDNode *Ld = dyn_cast<LoadSDNode>(St->getValue());
    if (!Ld) break;

    // Loads must only have one use.
    if (!Ld->hasNUsesOfValue(1, 0))
      break;

    // Check that the alignment is the same as the stores.
    if (Ld->getAlignment() != St->getAlignment())
      break;

    // The memory operands must not be volatile.
    if (Ld->isVolatile() || Ld->isIndexed())
      break;

    // We do not accept ext loads.
    if (Ld->getExtensionType() != ISD::NON_EXTLOAD)
      break;

    // The stored memory type must be the same.
    if (Ld->getMemoryVT() != MemVT)
      break;

    std::pair<SDValue, int64_t> LdPtr =
    GetPointerBaseAndOffset(Ld->getBasePtr());

    // If this is not the first ptr that we check.
    if (LdBasePtr.getNode()) {
      // The base ptr must be the same.
      if (LdPtr.first != LdBasePtr)
        break;
    } else {
      // Check that all other base pointers are the same as this one.
      LdBasePtr = LdPtr.first;
    }

    // We found a potential memory operand to merge.
    LoadNodes.push_back(MemOpLink(Ld, LdPtr.second, 0));
  }

  if (LoadNodes.size() < 2)
    return false;

  // Scan the memory operations on the chain and find the first non-consecutive
  // load memory address. These variables hold the index in the store node
  // array.
  unsigned LastConsecutiveLoad = 0;
  // This variable refers to the size and not index in the array.
  unsigned LastLegalVectorType = 0;
  unsigned LastLegalIntegerType = 0;
  StartAddress = LoadNodes[0].OffsetFromBase;
  SDValue FirstChain = LoadNodes[0].MemNode->getChain();
  for (unsigned i = 1; i < LoadNodes.size(); ++i) {
    // All loads much share the same chain.
    if (LoadNodes[i].MemNode->getChain() != FirstChain)
      break;
    
    int64_t CurrAddress = LoadNodes[i].OffsetFromBase;
    if (CurrAddress - StartAddress != (ElementSizeBytes * i))
      break;
    LastConsecutiveLoad = i;

    // Find a legal type for the vector store.
    EVT StoreTy = EVT::getVectorVT(*DAG.getContext(), MemVT, i+1);
    if (TLI.isTypeLegal(StoreTy))
      LastLegalVectorType = i + 1;

    // Find a legal type for the integer store.
    unsigned StoreBW = (i+1) * ElementSizeBytes * 8;
    StoreTy = EVT::getIntegerVT(*DAG.getContext(), StoreBW);
    if (TLI.isTypeLegal(StoreTy))
      LastLegalIntegerType = i + 1;
  }

  // Only use vector types if the vector type is larger than the integer type.
  // If they are the same, use integers.
  bool UseVectorTy = LastLegalVectorType > LastLegalIntegerType;
  unsigned LastLegalType = std::max(LastLegalVectorType, LastLegalIntegerType);

  // We add +1 here because the LastXXX variables refer to location while
  // the NumElem refers to array/index size.
  unsigned NumElem = std::min(LastConsecutiveStore, LastConsecutiveLoad) + 1;
  NumElem = std::min(LastLegalType, NumElem);

  if (NumElem < 2)
    return false;

  // The earliest Node in the DAG.
  unsigned EarliestNodeUsed = 0;
  LSBaseSDNode *EarliestOp = StoreNodes[EarliestNodeUsed].MemNode;
  for (unsigned i=1; i<NumElem; ++i) {
    // Find a chain for the new wide-store operand. Notice that some
    // of the store nodes that we found may not be selected for inclusion
    // in the wide store. The chain we use needs to be the chain of the
    // earliest store node which is *used* and replaced by the wide store.
    if (StoreNodes[i].SequenceNum > StoreNodes[EarliestNodeUsed].SequenceNum)
      EarliestNodeUsed = i;
  }

  // Find if it is better to use vectors or integers to load and store
  // to memory.
  EVT JointMemOpVT;
  if (UseVectorTy) {
    JointMemOpVT = EVT::getVectorVT(*DAG.getContext(), MemVT, NumElem);
  } else {
    unsigned StoreBW = NumElem * ElementSizeBytes * 8;
    JointMemOpVT = EVT::getIntegerVT(*DAG.getContext(), StoreBW);
  }

  DebugLoc LoadDL = LoadNodes[0].MemNode->getDebugLoc();
  DebugLoc StoreDL = StoreNodes[0].MemNode->getDebugLoc();

  LoadSDNode *FirstLoad = cast<LoadSDNode>(LoadNodes[0].MemNode);
  SDValue NewLoad = DAG.getLoad(JointMemOpVT, LoadDL,
                                FirstLoad->getChain(),
                                FirstLoad->getBasePtr(),
                                FirstLoad->getPointerInfo(),
                                false, false, false,
                                FirstLoad->getAlignment());

  SDValue NewStore = DAG.getStore(EarliestOp->getChain(), StoreDL, NewLoad,
                                  FirstInChain->getBasePtr(),
                                  FirstInChain->getPointerInfo(), false, false,
                                  FirstInChain->getAlignment());

  // Replace one of the loads with the new load.
  LoadSDNode *Ld = cast<LoadSDNode>(LoadNodes[0].MemNode);
  DAG.ReplaceAllUsesOfValueWith(SDValue(Ld, 1),
                                SDValue(NewLoad.getNode(), 1));

  // Remove the rest of the load chains.
  for (unsigned i = 1; i < NumElem ; ++i) {
    // Replace all chain users of the old load nodes with the chain of the new
    // load node.
    LoadSDNode *Ld = cast<LoadSDNode>(LoadNodes[i].MemNode);
    DAG.ReplaceAllUsesOfValueWith(SDValue(Ld, 1), Ld->getChain());
  }

  // Replace the first store with the new store.
  CombineTo(EarliestOp, NewStore);
  // Erase all other stores.
  for (unsigned i = 0; i < NumElem ; ++i) {
    // Remove all Store nodes.
    if (StoreNodes[i].MemNode == EarliestOp)
      continue;
    StoreSDNode *St = cast<StoreSDNode>(StoreNodes[i].MemNode);
    DAG.ReplaceAllUsesOfValueWith(SDValue(St, 0), St->getChain());
    removeFromWorkList(St);
    DAG.DeleteNode(St);
  }

  return true;
}

SDValue DAGCombiner::visitSTORE(SDNode *N) {
  StoreSDNode *ST  = cast<StoreSDNode>(N);
  SDValue Chain = ST->getChain();
  SDValue Value = ST->getValue();
  SDValue Ptr   = ST->getBasePtr();

  // If this is a store of a bit convert, store the input value if the
  // resultant store does not need a higher alignment than the original.
  if (Value.getOpcode() == ISD::BITCAST && !ST->isTruncatingStore() &&
      ST->isUnindexed()) {
    unsigned OrigAlign = ST->getAlignment();
    EVT SVT = Value.getOperand(0).getValueType();
    unsigned Align = TLI.getDataLayout()->
      getABITypeAlignment(SVT.getTypeForEVT(*DAG.getContext()));
    if (Align <= OrigAlign &&
        ((!LegalOperations && !ST->isVolatile()) ||
         TLI.isOperationLegalOrCustom(ISD::STORE, SVT)))
      return DAG.getStore(Chain, N->getDebugLoc(), Value.getOperand(0),
                          Ptr, ST->getPointerInfo(), ST->isVolatile(),
                          ST->isNonTemporal(), OrigAlign);
  }

  // Turn 'store undef, Ptr' -> nothing.
  if (Value.getOpcode() == ISD::UNDEF && ST->isUnindexed())
    return Chain;

  // Turn 'store float 1.0, Ptr' -> 'store int 0x12345678, Ptr'
  if (ConstantFPSDNode *CFP = dyn_cast<ConstantFPSDNode>(Value)) {
    // NOTE: If the original store is volatile, this transform must not increase
    // the number of stores.  For example, on x86-32 an f64 can be stored in one
    // processor operation but an i64 (which is not legal) requires two.  So the
    // transform should not be done in this case.
    if (Value.getOpcode() != ISD::TargetConstantFP) {
      SDValue Tmp;
      switch (CFP->getValueType(0).getSimpleVT().SimpleTy) {
      default: llvm_unreachable("Unknown FP type");
      case MVT::f16:    // We don't do this for these yet.
      case MVT::f80:
      case MVT::f128:
      case MVT::ppcf128:
        break;
      case MVT::f32:
        if ((isTypeLegal(MVT::i32) && !LegalOperations && !ST->isVolatile()) ||
            TLI.isOperationLegalOrCustom(ISD::STORE, MVT::i32)) {
          Tmp = DAG.getConstant((uint32_t)CFP->getValueAPF().
                              bitcastToAPInt().getZExtValue(), MVT::i32);
          return DAG.getStore(Chain, N->getDebugLoc(), Tmp,
                              Ptr, ST->getPointerInfo(), ST->isVolatile(),
                              ST->isNonTemporal(), ST->getAlignment());
        }
        break;
      case MVT::f64:
        if ((TLI.isTypeLegal(MVT::i64) && !LegalOperations &&
             !ST->isVolatile()) ||
            TLI.isOperationLegalOrCustom(ISD::STORE, MVT::i64)) {
          Tmp = DAG.getConstant(CFP->getValueAPF().bitcastToAPInt().
                                getZExtValue(), MVT::i64);
          return DAG.getStore(Chain, N->getDebugLoc(), Tmp,
                              Ptr, ST->getPointerInfo(), ST->isVolatile(),
                              ST->isNonTemporal(), ST->getAlignment());
        }

        if (!ST->isVolatile() &&
            TLI.isOperationLegalOrCustom(ISD::STORE, MVT::i32)) {
          // Many FP stores are not made apparent until after legalize, e.g. for
          // argument passing.  Since this is so common, custom legalize the
          // 64-bit integer store into two 32-bit stores.
          uint64_t Val = CFP->getValueAPF().bitcastToAPInt().getZExtValue();
          SDValue Lo = DAG.getConstant(Val & 0xFFFFFFFF, MVT::i32);
          SDValue Hi = DAG.getConstant(Val >> 32, MVT::i32);
          if (TLI.isBigEndian()) std::swap(Lo, Hi);

          unsigned Alignment = ST->getAlignment();
          bool isVolatile = ST->isVolatile();
          bool isNonTemporal = ST->isNonTemporal();

          SDValue St0 = DAG.getStore(Chain, ST->getDebugLoc(), Lo,
                                     Ptr, ST->getPointerInfo(),
                                     isVolatile, isNonTemporal,
                                     ST->getAlignment());
          Ptr = DAG.getNode(ISD::ADD, N->getDebugLoc(), Ptr.getValueType(), Ptr,
                            DAG.getConstant(4, Ptr.getValueType()));
          Alignment = MinAlign(Alignment, 4U);
          SDValue St1 = DAG.getStore(Chain, ST->getDebugLoc(), Hi,
                                     Ptr, ST->getPointerInfo().getWithOffset(4),
                                     isVolatile, isNonTemporal,
                                     Alignment);
          return DAG.getNode(ISD::TokenFactor, N->getDebugLoc(), MVT::Other,
                             St0, St1);
        }

        break;
      }
    }
  }

  // Try to infer better alignment information than the store already has.
  if (OptLevel != CodeGenOpt::None && ST->isUnindexed()) {
    if (unsigned Align = DAG.InferPtrAlignment(Ptr)) {
      if (Align > ST->getAlignment())
        return DAG.getTruncStore(Chain, N->getDebugLoc(), Value,
                                 Ptr, ST->getPointerInfo(), ST->getMemoryVT(),
                                 ST->isVolatile(), ST->isNonTemporal(), Align);
    }
  }

  // Try transforming a pair floating point load / store ops to integer
  // load / store ops.
  SDValue NewST = TransformFPLoadStorePair(N);
  if (NewST.getNode())
    return NewST;

  if (CombinerAA) {
    // Walk up chain skipping non-aliasing memory nodes.
    SDValue BetterChain = FindBetterChain(N, Chain);

    // If there is a better chain.
    if (Chain != BetterChain) {
      SDValue ReplStore;

      // Replace the chain to avoid dependency.
      if (ST->isTruncatingStore()) {
        ReplStore = DAG.getTruncStore(BetterChain, N->getDebugLoc(), Value, Ptr,
                                      ST->getPointerInfo(),
                                      ST->getMemoryVT(), ST->isVolatile(),
                                      ST->isNonTemporal(), ST->getAlignment());
      } else {
        ReplStore = DAG.getStore(BetterChain, N->getDebugLoc(), Value, Ptr,
                                 ST->getPointerInfo(),
                                 ST->isVolatile(), ST->isNonTemporal(),
                                 ST->getAlignment());
      }

      // Create token to keep both nodes around.
      SDValue Token = DAG.getNode(ISD::TokenFactor, N->getDebugLoc(),
                                  MVT::Other, Chain, ReplStore);

      // Make sure the new and old chains are cleaned up.
      AddToWorkList(Token.getNode());

      // Don't add users to work list.
      return CombineTo(N, Token, false);
    }
  }

  // Try transforming N to an indexed store.
  if (CombineToPreIndexedLoadStore(N) || CombineToPostIndexedLoadStore(N))
    return SDValue(N, 0);

  // FIXME: is there such a thing as a truncating indexed store?
  if (ST->isTruncatingStore() && ST->isUnindexed() &&
      Value.getValueType().isInteger()) {
    // See if we can simplify the input to this truncstore with knowledge that
    // only the low bits are being used.  For example:
    // "truncstore (or (shl x, 8), y), i8"  -> "truncstore y, i8"
    SDValue Shorter =
      GetDemandedBits(Value,
                      APInt::getLowBitsSet(
                        Value.getValueType().getScalarType().getSizeInBits(),
                        ST->getMemoryVT().getScalarType().getSizeInBits()));
    AddToWorkList(Value.getNode());
    if (Shorter.getNode())
      return DAG.getTruncStore(Chain, N->getDebugLoc(), Shorter,
                               Ptr, ST->getPointerInfo(), ST->getMemoryVT(),
                               ST->isVolatile(), ST->isNonTemporal(),
                               ST->getAlignment());

    // Otherwise, see if we can simplify the operation with
    // SimplifyDemandedBits, which only works if the value has a single use.
    if (SimplifyDemandedBits(Value,
                        APInt::getLowBitsSet(
                          Value.getValueType().getScalarType().getSizeInBits(),
                          ST->getMemoryVT().getScalarType().getSizeInBits())))
      return SDValue(N, 0);
  }

  // If this is a load followed by a store to the same location, then the store
  // is dead/noop.
  if (LoadSDNode *Ld = dyn_cast<LoadSDNode>(Value)) {
    if (Ld->getBasePtr() == Ptr && ST->getMemoryVT() == Ld->getMemoryVT() &&
        ST->isUnindexed() && !ST->isVolatile() &&
        // There can't be any side effects between the load and store, such as
        // a call or store.
        Chain.reachesChainWithoutSideEffects(SDValue(Ld, 1))) {
      // The store is dead, remove it.
      return Chain;
    }
  }

  // If this is an FP_ROUND or TRUNC followed by a store, fold this into a
  // truncating store.  We can do this even if this is already a truncstore.
  if ((Value.getOpcode() == ISD::FP_ROUND || Value.getOpcode() == ISD::TRUNCATE)
      && Value.getNode()->hasOneUse() && ST->isUnindexed() &&
      TLI.isTruncStoreLegal(Value.getOperand(0).getValueType(),
                            ST->getMemoryVT())) {
    return DAG.getTruncStore(Chain, N->getDebugLoc(), Value.getOperand(0),
                             Ptr, ST->getPointerInfo(), ST->getMemoryVT(),
                             ST->isVolatile(), ST->isNonTemporal(),
                             ST->getAlignment());
  }

  // Only perform this optimization before the types are legal, because we
  // don't want to perform this optimization on every DAGCombine invocation.
  if (!LegalTypes && MergeConsecutiveStores(ST))
    return SDValue(N, 0);

  return ReduceLoadOpStoreWidth(N);
}

SDValue DAGCombiner::visitINSERT_VECTOR_ELT(SDNode *N) {
  SDValue InVec = N->getOperand(0);
  SDValue InVal = N->getOperand(1);
  SDValue EltNo = N->getOperand(2);
  DebugLoc dl = N->getDebugLoc();

  // If the inserted element is an UNDEF, just use the input vector.
  if (InVal.getOpcode() == ISD::UNDEF)
    return InVec;

  EVT VT = InVec.getValueType();

  // If we can't generate a legal BUILD_VECTOR, exit
  if (LegalOperations && !TLI.isOperationLegal(ISD::BUILD_VECTOR, VT))
    return SDValue();

  // Check that we know which element is being inserted
  if (!isa<ConstantSDNode>(EltNo))
    return SDValue();
  unsigned Elt = cast<ConstantSDNode>(EltNo)->getZExtValue();

  // Check that the operand is a BUILD_VECTOR (or UNDEF, which can essentially
  // be converted to a BUILD_VECTOR).  Fill in the Ops vector with the
  // vector elements.
  SmallVector<SDValue, 8> Ops;
  if (InVec.getOpcode() == ISD::BUILD_VECTOR) {
    Ops.append(InVec.getNode()->op_begin(),
               InVec.getNode()->op_end());
  } else if (InVec.getOpcode() == ISD::UNDEF) {
    unsigned NElts = VT.getVectorNumElements();
    Ops.append(NElts, DAG.getUNDEF(InVal.getValueType()));
  } else {
    return SDValue();
  }

  // Insert the element
  if (Elt < Ops.size()) {
    // All the operands of BUILD_VECTOR must have the same type;
    // we enforce that here.
    EVT OpVT = Ops[0].getValueType();
    if (InVal.getValueType() != OpVT)
      InVal = OpVT.bitsGT(InVal.getValueType()) ?
                DAG.getNode(ISD::ANY_EXTEND, dl, OpVT, InVal) :
                DAG.getNode(ISD::TRUNCATE, dl, OpVT, InVal);
    Ops[Elt] = InVal;
  }

  // Return the new vector
  return DAG.getNode(ISD::BUILD_VECTOR, dl,
                     VT, &Ops[0], Ops.size());
}

SDValue DAGCombiner::visitEXTRACT_VECTOR_ELT(SDNode *N) {
  // (vextract (scalar_to_vector val, 0) -> val
  SDValue InVec = N->getOperand(0);
  EVT VT = InVec.getValueType();
  EVT NVT = N->getValueType(0);

  if (InVec.getOpcode() == ISD::SCALAR_TO_VECTOR) {
    // Check if the result type doesn't match the inserted element type. A
    // SCALAR_TO_VECTOR may truncate the inserted element and the
    // EXTRACT_VECTOR_ELT may widen the extracted vector.
    SDValue InOp = InVec.getOperand(0);
    if (InOp.getValueType() != NVT) {
      assert(InOp.getValueType().isInteger() && NVT.isInteger());
      return DAG.getSExtOrTrunc(InOp, InVec.getDebugLoc(), NVT);
    }
    return InOp;
  }

  SDValue EltNo = N->getOperand(1);
  bool ConstEltNo = isa<ConstantSDNode>(EltNo);

  // Transform: (EXTRACT_VECTOR_ELT( VECTOR_SHUFFLE )) -> EXTRACT_VECTOR_ELT.
  // We only perform this optimization before the op legalization phase because
  // we may introduce new vector instructions which are not backed by TD
  // patterns. For example on AVX, extracting elements from a wide vector
  // without using extract_subvector.
  if (InVec.getOpcode() == ISD::VECTOR_SHUFFLE
      && ConstEltNo && !LegalOperations) {
    int Elt = cast<ConstantSDNode>(EltNo)->getZExtValue();
    int NumElem = VT.getVectorNumElements();
    ShuffleVectorSDNode *SVOp = cast<ShuffleVectorSDNode>(InVec);
    // Find the new index to extract from.
    int OrigElt = SVOp->getMaskElt(Elt);

    // Extracting an undef index is undef.
    if (OrigElt == -1)
      return DAG.getUNDEF(NVT);

    // Select the right vector half to extract from.
    if (OrigElt < NumElem) {
      InVec = InVec->getOperand(0);
    } else {
      InVec = InVec->getOperand(1);
      OrigElt -= NumElem;
    }

    EVT IndexTy = N->getOperand(1).getValueType();
    return DAG.getNode(ISD::EXTRACT_VECTOR_ELT, N->getDebugLoc(), NVT,
                       InVec, DAG.getConstant(OrigElt, IndexTy));
  }

  // Perform only after legalization to ensure build_vector / vector_shuffle
  // optimizations have already been done.
  if (!LegalOperations) return SDValue();

  // (vextract (v4f32 load $addr), c) -> (f32 load $addr+c*size)
  // (vextract (v4f32 s2v (f32 load $addr)), c) -> (f32 load $addr+c*size)
  // (vextract (v4f32 shuffle (load $addr), <1,u,u,u>), 0) -> (f32 load $addr)

  if (ConstEltNo) {
    int Elt = cast<ConstantSDNode>(EltNo)->getZExtValue();
    bool NewLoad = false;
    bool BCNumEltsChanged = false;
    EVT ExtVT = VT.getVectorElementType();
    EVT LVT = ExtVT;

    // If the result of load has to be truncated, then it's not necessarily
    // profitable.
    if (NVT.bitsLT(LVT) && !TLI.isTruncateFree(LVT, NVT))
      return SDValue();

    if (InVec.getOpcode() == ISD::BITCAST) {
      // Don't duplicate a load with other uses.
      if (!InVec.hasOneUse())
        return SDValue();

      EVT BCVT = InVec.getOperand(0).getValueType();
      if (!BCVT.isVector() || ExtVT.bitsGT(BCVT.getVectorElementType()))
        return SDValue();
      if (VT.getVectorNumElements() != BCVT.getVectorNumElements())
        BCNumEltsChanged = true;
      InVec = InVec.getOperand(0);
      ExtVT = BCVT.getVectorElementType();
      NewLoad = true;
    }

    LoadSDNode *LN0 = NULL;
    const ShuffleVectorSDNode *SVN = NULL;
    if (ISD::isNormalLoad(InVec.getNode())) {
      LN0 = cast<LoadSDNode>(InVec);
    } else if (InVec.getOpcode() == ISD::SCALAR_TO_VECTOR &&
               InVec.getOperand(0).getValueType() == ExtVT &&
               ISD::isNormalLoad(InVec.getOperand(0).getNode())) {
      // Don't duplicate a load with other uses.
      if (!InVec.hasOneUse())
        return SDValue();

      LN0 = cast<LoadSDNode>(InVec.getOperand(0));
    } else if ((SVN = dyn_cast<ShuffleVectorSDNode>(InVec))) {
      // (vextract (vector_shuffle (load $addr), v2, <1, u, u, u>), 1)
      // =>
      // (load $addr+1*size)

      // Don't duplicate a load with other uses.
      if (!InVec.hasOneUse())
        return SDValue();

      // If the bit convert changed the number of elements, it is unsafe
      // to examine the mask.
      if (BCNumEltsChanged)
        return SDValue();

      // Select the input vector, guarding against out of range extract vector.
      unsigned NumElems = VT.getVectorNumElements();
      int Idx = (Elt > (int)NumElems) ? -1 : SVN->getMaskElt(Elt);
      InVec = (Idx < (int)NumElems) ? InVec.getOperand(0) : InVec.getOperand(1);

      if (InVec.getOpcode() == ISD::BITCAST) {
        // Don't duplicate a load with other uses.
        if (!InVec.hasOneUse())
          return SDValue();

        InVec = InVec.getOperand(0);
      }
      if (ISD::isNormalLoad(InVec.getNode())) {
        LN0 = cast<LoadSDNode>(InVec);
        Elt = (Idx < (int)NumElems) ? Idx : Idx - (int)NumElems;
      }
    }

    // Make sure we found a non-volatile load and the extractelement is
    // the only use.
    if (!LN0 || !LN0->hasNUsesOfValue(1,0) || LN0->isVolatile())
      return SDValue();

    // If Idx was -1 above, Elt is going to be -1, so just return undef.
    if (Elt == -1)
      return DAG.getUNDEF(LVT);

    unsigned Align = LN0->getAlignment();
    if (NewLoad) {
      // Check the resultant load doesn't need a higher alignment than the
      // original load.
      unsigned NewAlign =
        TLI.getDataLayout()
            ->getABITypeAlignment(LVT.getTypeForEVT(*DAG.getContext()));

      if (NewAlign > Align || !TLI.isOperationLegalOrCustom(ISD::LOAD, LVT))
        return SDValue();

      Align = NewAlign;
    }

    SDValue NewPtr = LN0->getBasePtr();
    unsigned PtrOff = 0;

    if (Elt) {
      PtrOff = LVT.getSizeInBits() * Elt / 8;
      EVT PtrType = NewPtr.getValueType();
      if (TLI.isBigEndian())
        PtrOff = VT.getSizeInBits() / 8 - PtrOff;
      NewPtr = DAG.getNode(ISD::ADD, N->getDebugLoc(), PtrType, NewPtr,
                           DAG.getConstant(PtrOff, PtrType));
    }

    // The replacement we need to do here is a little tricky: we need to
    // replace an extractelement of a load with a load.
    // Use ReplaceAllUsesOfValuesWith to do the replacement.
    // Note that this replacement assumes that the extractvalue is the only
    // use of the load; that's okay because we don't want to perform this
    // transformation in other cases anyway.
    SDValue Load;
    SDValue Chain;
    if (NVT.bitsGT(LVT)) {
      // If the result type of vextract is wider than the load, then issue an
      // extending load instead.
      ISD::LoadExtType ExtType = TLI.isLoadExtLegal(ISD::ZEXTLOAD, LVT)
        ? ISD::ZEXTLOAD : ISD::EXTLOAD;
      Load = DAG.getExtLoad(ExtType, N->getDebugLoc(), NVT, LN0->getChain(),
                            NewPtr, LN0->getPointerInfo().getWithOffset(PtrOff),
                            LVT, LN0->isVolatile(), LN0->isNonTemporal(),Align);
      Chain = Load.getValue(1);
    } else {
      Load = DAG.getLoad(LVT, N->getDebugLoc(), LN0->getChain(), NewPtr,
                         LN0->getPointerInfo().getWithOffset(PtrOff),
                         LN0->isVolatile(), LN0->isNonTemporal(), 
                         LN0->isInvariant(), Align);
      Chain = Load.getValue(1);
      if (NVT.bitsLT(LVT))
        Load = DAG.getNode(ISD::TRUNCATE, N->getDebugLoc(), NVT, Load);
      else
        Load = DAG.getNode(ISD::BITCAST, N->getDebugLoc(), NVT, Load);
    }
    WorkListRemover DeadNodes(*this);
    SDValue From[] = { SDValue(N, 0), SDValue(LN0,1) };
    SDValue To[] = { Load, Chain };
    DAG.ReplaceAllUsesOfValuesWith(From, To, 2);
    // Since we're explcitly calling ReplaceAllUses, add the new node to the
    // worklist explicitly as well.
    AddToWorkList(Load.getNode());
    AddUsersToWorkList(Load.getNode()); // Add users too
    // Make sure to revisit this node to clean it up; it will usually be dead.
    AddToWorkList(N);
    return SDValue(N, 0);
  }

  return SDValue();
}

SDValue DAGCombiner::visitBUILD_VECTOR(SDNode *N) {
  unsigned NumInScalars = N->getNumOperands();
  DebugLoc dl = N->getDebugLoc();
  EVT VT = N->getValueType(0);

  // A vector built entirely of undefs is undef.
  if (ISD::allOperandsUndef(N))
    return DAG.getUNDEF(VT);

  // Check to see if this is a BUILD_VECTOR of a bunch of values
  // which come from any_extend or zero_extend nodes. If so, we can create
  // a new BUILD_VECTOR using bit-casts which may enable other BUILD_VECTOR
  // optimizations. We do not handle sign-extend because we can't fill the sign
  // using shuffles.
  EVT SourceType = MVT::Other;
  bool AllAnyExt = true;

  for (unsigned i = 0; i != NumInScalars; ++i) {
    SDValue In = N->getOperand(i);
    // Ignore undef inputs.
    if (In.getOpcode() == ISD::UNDEF) continue;

    bool AnyExt  = In.getOpcode() == ISD::ANY_EXTEND;
    bool ZeroExt = In.getOpcode() == ISD::ZERO_EXTEND;

    // Abort if the element is not an extension.
    if (!ZeroExt && !AnyExt) {
      SourceType = MVT::Other;
      break;
    }

    // The input is a ZeroExt or AnyExt. Check the original type.
    EVT InTy = In.getOperand(0).getValueType();

    // Check that all of the widened source types are the same.
    if (SourceType == MVT::Other)
      // First time.
      SourceType = InTy;
    else if (InTy != SourceType) {
      // Multiple income types. Abort.
      SourceType = MVT::Other;
      break;
    }

    // Check if all of the extends are ANY_EXTENDs.
    AllAnyExt &= AnyExt;
  }

  // In order to have valid types, all of the inputs must be extended from the
  // same source type and all of the inputs must be any or zero extend.
  // Scalar sizes must be a power of two.
  EVT OutScalarTy = N->getValueType(0).getScalarType();
  bool ValidTypes = SourceType != MVT::Other &&
                 isPowerOf2_32(OutScalarTy.getSizeInBits()) &&
                 isPowerOf2_32(SourceType.getSizeInBits());

  // We perform this optimization post type-legalization because
  // the type-legalizer often scalarizes integer-promoted vectors.
  // Performing this optimization before may create bit-casts which
  // will be type-legalized to complex code sequences.
  // We perform this optimization only before the operation legalizer because we
  // may introduce illegal operations.
  // Create a new simpler BUILD_VECTOR sequence which other optimizations can
  // turn into a single shuffle instruction.
  if ((Level == AfterLegalizeVectorOps || Level == AfterLegalizeTypes) &&
      ValidTypes) {
    bool isLE = TLI.isLittleEndian();
    unsigned ElemRatio = OutScalarTy.getSizeInBits()/SourceType.getSizeInBits();
    assert(ElemRatio > 1 && "Invalid element size ratio");
    SDValue Filler = AllAnyExt ? DAG.getUNDEF(SourceType):
                                 DAG.getConstant(0, SourceType);

    unsigned NewBVElems = ElemRatio * N->getValueType(0).getVectorNumElements();
    SmallVector<SDValue, 8> Ops(NewBVElems, Filler);

    // Populate the new build_vector
    for (unsigned i=0; i < N->getNumOperands(); ++i) {
      SDValue Cast = N->getOperand(i);
      assert((Cast.getOpcode() == ISD::ANY_EXTEND ||
              Cast.getOpcode() == ISD::ZERO_EXTEND ||
              Cast.getOpcode() == ISD::UNDEF) && "Invalid cast opcode");
      SDValue In;
      if (Cast.getOpcode() == ISD::UNDEF)
        In = DAG.getUNDEF(SourceType);
      else
        In = Cast->getOperand(0);
      unsigned Index = isLE ? (i * ElemRatio) :
                              (i * ElemRatio + (ElemRatio - 1));

      assert(Index < Ops.size() && "Invalid index");
      Ops[Index] = In;
    }

    // The type of the new BUILD_VECTOR node.
    EVT VecVT = EVT::getVectorVT(*DAG.getContext(), SourceType, NewBVElems);
    assert(VecVT.getSizeInBits() == N->getValueType(0).getSizeInBits() &&
           "Invalid vector size");
    // Check if the new vector type is legal.
    if (!isTypeLegal(VecVT)) return SDValue();

    // Make the new BUILD_VECTOR.
    SDValue BV = DAG.getNode(ISD::BUILD_VECTOR, N->getDebugLoc(),
                                 VecVT, &Ops[0], Ops.size());

    // The new BUILD_VECTOR node has the potential to be further optimized.
    AddToWorkList(BV.getNode());
    // Bitcast to the desired type.
    return DAG.getNode(ISD::BITCAST, dl, N->getValueType(0), BV);
  }

  // Check to see if this is a BUILD_VECTOR of a bunch of EXTRACT_VECTOR_ELT
  // operations.  If so, and if the EXTRACT_VECTOR_ELT vector inputs come from
  // at most two distinct vectors, turn this into a shuffle node.

  // May only combine to shuffle after legalize if shuffle is legal.
  if (LegalOperations &&
      !TLI.isOperationLegalOrCustom(ISD::VECTOR_SHUFFLE, VT))
    return SDValue();

  SDValue VecIn1, VecIn2;
  for (unsigned i = 0; i != NumInScalars; ++i) {
    // Ignore undef inputs.
    if (N->getOperand(i).getOpcode() == ISD::UNDEF) continue;

    // If this input is something other than a EXTRACT_VECTOR_ELT with a
    // constant index, bail out.
    if (N->getOperand(i).getOpcode() != ISD::EXTRACT_VECTOR_ELT ||
        !isa<ConstantSDNode>(N->getOperand(i).getOperand(1))) {
      VecIn1 = VecIn2 = SDValue(0, 0);
      break;
    }

    // We allow up to two distinct input vectors.
    SDValue ExtractedFromVec = N->getOperand(i).getOperand(0);
    if (ExtractedFromVec == VecIn1 || ExtractedFromVec == VecIn2)
      continue;

    if (VecIn1.getNode() == 0) {
      VecIn1 = ExtractedFromVec;
    } else if (VecIn2.getNode() == 0) {
      VecIn2 = ExtractedFromVec;
    } else {
      // Too many inputs.
      VecIn1 = VecIn2 = SDValue(0, 0);
      break;
    }
  }

    // If everything is good, we can make a shuffle operation.
  if (VecIn1.getNode()) {
    SmallVector<int, 8> Mask;
    for (unsigned i = 0; i != NumInScalars; ++i) {
      if (N->getOperand(i).getOpcode() == ISD::UNDEF) {
        Mask.push_back(-1);
        continue;
      }

      // If extracting from the first vector, just use the index directly.
      SDValue Extract = N->getOperand(i);
      SDValue ExtVal = Extract.getOperand(1);
      if (Extract.getOperand(0) == VecIn1) {
        unsigned ExtIndex = cast<ConstantSDNode>(ExtVal)->getZExtValue();
        if (ExtIndex > VT.getVectorNumElements())
          return SDValue();

        Mask.push_back(ExtIndex);
        continue;
      }

      // Otherwise, use InIdx + VecSize
      unsigned Idx = cast<ConstantSDNode>(ExtVal)->getZExtValue();
      Mask.push_back(Idx+NumInScalars);
    }

    // We can't generate a shuffle node with mismatched input and output types.
    // Attempt to transform a single input vector to the correct type.
    if ((VT != VecIn1.getValueType())) {
      // We don't support shuffeling between TWO values of different types.
      if (VecIn2.getNode() != 0)
        return SDValue();

      // We only support widening of vectors which are half the size of the
      // output registers. For example XMM->YMM widening on X86 with AVX.
      if (VecIn1.getValueType().getSizeInBits()*2 != VT.getSizeInBits())
        return SDValue();

      // If the input vector type has a different base type to the output
      // vector type, bail out.
      if (VecIn1.getValueType().getVectorElementType() !=
          VT.getVectorElementType())
        return SDValue();

      // Widen the input vector by adding undef values.
      VecIn1 = DAG.getNode(ISD::CONCAT_VECTORS, N->getDebugLoc(), VT,
                           VecIn1, DAG.getUNDEF(VecIn1.getValueType()));
    }

    // If VecIn2 is unused then change it to undef.
    VecIn2 = VecIn2.getNode() ? VecIn2 : DAG.getUNDEF(VT);

    // Check that we were able to transform all incoming values to the same
    // type.
    if (VecIn2.getValueType() != VecIn1.getValueType() ||
        VecIn1.getValueType() != VT)
          return SDValue();

    // Only type-legal BUILD_VECTOR nodes are converted to shuffle nodes.
    if (!isTypeLegal(VT))
      return SDValue();

    // Return the new VECTOR_SHUFFLE node.
    SDValue Ops[2];
    Ops[0] = VecIn1;
    Ops[1] = VecIn2;
    return DAG.getVectorShuffle(VT, N->getDebugLoc(), Ops[0], Ops[1], &Mask[0]);
  }

  return SDValue();
}

SDValue DAGCombiner::visitCONCAT_VECTORS(SDNode *N) {
  // TODO: Check to see if this is a CONCAT_VECTORS of a bunch of
  // EXTRACT_SUBVECTOR operations.  If so, and if the EXTRACT_SUBVECTOR vector
  // inputs come from at most two distinct vectors, turn this into a shuffle
  // node.

  // If we only have one input vector, we don't need to do any concatenation.
  if (N->getNumOperands() == 1)
    return N->getOperand(0);

  // Check if all of the operands are undefs.
  if (ISD::allOperandsUndef(N))
    return DAG.getUNDEF(N->getValueType(0));

  return SDValue();
}

SDValue DAGCombiner::visitEXTRACT_SUBVECTOR(SDNode* N) {
  EVT NVT = N->getValueType(0);
  SDValue V = N->getOperand(0);

  if (V->getOpcode() == ISD::INSERT_SUBVECTOR) {
    // Handle only simple case where vector being inserted and vector
    // being extracted are of same type, and are half size of larger vectors.
    EVT BigVT = V->getOperand(0).getValueType();
    EVT SmallVT = V->getOperand(1).getValueType();
    if (NVT != SmallVT || NVT.getSizeInBits()*2 != BigVT.getSizeInBits())
      return SDValue();

    // Only handle cases where both indexes are constants with the same type.
    ConstantSDNode *InsIdx = dyn_cast<ConstantSDNode>(N->getOperand(1));
    ConstantSDNode *ExtIdx = dyn_cast<ConstantSDNode>(V->getOperand(2));

    if (InsIdx && ExtIdx &&
        InsIdx->getValueType(0).getSizeInBits() <= 64 &&
        ExtIdx->getValueType(0).getSizeInBits() <= 64) {
      // Combine:
      //    (extract_subvec (insert_subvec V1, V2, InsIdx), ExtIdx)
      // Into:
      //    indices are equal => V1
      //    otherwise => (extract_subvec V1, ExtIdx)
      if (InsIdx->getZExtValue() == ExtIdx->getZExtValue())
        return V->getOperand(1);
      return DAG.getNode(ISD::EXTRACT_SUBVECTOR, N->getDebugLoc(), NVT,
                         V->getOperand(0), N->getOperand(1));
    }
  }

  return SDValue();
}

SDValue DAGCombiner::visitVECTOR_SHUFFLE(SDNode *N) {
  EVT VT = N->getValueType(0);
  unsigned NumElts = VT.getVectorNumElements();

  SDValue N0 = N->getOperand(0);
  SDValue N1 = N->getOperand(1);

  assert(N0.getValueType() == VT && "Vector shuffle must be normalized in DAG");

  // Canonicalize shuffle undef, undef -> undef
  if (N0.getOpcode() == ISD::UNDEF && N1.getOpcode() == ISD::UNDEF)
    return DAG.getUNDEF(VT);

  ShuffleVectorSDNode *SVN = cast<ShuffleVectorSDNode>(N);

  // Canonicalize shuffle v, v -> v, undef
  if (N0 == N1) {
    SmallVector<int, 8> NewMask;
    for (unsigned i = 0; i != NumElts; ++i) {
      int Idx = SVN->getMaskElt(i);
      if (Idx >= (int)NumElts) Idx -= NumElts;
      NewMask.push_back(Idx);
    }
    return DAG.getVectorShuffle(VT, N->getDebugLoc(), N0, DAG.getUNDEF(VT),
                                &NewMask[0]);
  }

  // Canonicalize shuffle undef, v -> v, undef.  Commute the shuffle mask.
  if (N0.getOpcode() == ISD::UNDEF) {
    SmallVector<int, 8> NewMask;
    for (unsigned i = 0; i != NumElts; ++i) {
      int Idx = SVN->getMaskElt(i);
      if (Idx >= 0) {
        if (Idx < (int)NumElts)
          Idx += NumElts;
        else
          Idx -= NumElts;
      }
      NewMask.push_back(Idx);
    }
    return DAG.getVectorShuffle(VT, N->getDebugLoc(), N1, DAG.getUNDEF(VT),
                                &NewMask[0]);
  }

  // Remove references to rhs if it is undef
  if (N1.getOpcode() == ISD::UNDEF) {
    bool Changed = false;
    SmallVector<int, 8> NewMask;
    for (unsigned i = 0; i != NumElts; ++i) {
      int Idx = SVN->getMaskElt(i);
      if (Idx >= (int)NumElts) {
        Idx = -1;
        Changed = true;
      }
      NewMask.push_back(Idx);
    }
    if (Changed)
      return DAG.getVectorShuffle(VT, N->getDebugLoc(), N0, N1, &NewMask[0]);
  }

  // If it is a splat, check if the argument vector is another splat or a
  // build_vector with all scalar elements the same.
  if (SVN->isSplat() && SVN->getSplatIndex() < (int)NumElts) {
    SDNode *V = N0.getNode();

    // If this is a bit convert that changes the element type of the vector but
    // not the number of vector elements, look through it.  Be careful not to
    // look though conversions that change things like v4f32 to v2f64.
    if (V->getOpcode() == ISD::BITCAST) {
      SDValue ConvInput = V->getOperand(0);
      if (ConvInput.getValueType().isVector() &&
          ConvInput.getValueType().getVectorNumElements() == NumElts)
        V = ConvInput.getNode();
    }

    if (V->getOpcode() == ISD::BUILD_VECTOR) {
      assert(V->getNumOperands() == NumElts &&
             "BUILD_VECTOR has wrong number of operands");
      SDValue Base;
      bool AllSame = true;
      for (unsigned i = 0; i != NumElts; ++i) {
        if (V->getOperand(i).getOpcode() != ISD::UNDEF) {
          Base = V->getOperand(i);
          break;
        }
      }
      // Splat of <u, u, u, u>, return <u, u, u, u>
      if (!Base.getNode())
        return N0;
      for (unsigned i = 0; i != NumElts; ++i) {
        if (V->getOperand(i) != Base) {
          AllSame = false;
          break;
        }
      }
      // Splat of <x, x, x, x>, return <x, x, x, x>
      if (AllSame)
        return N0;
    }
  }

  // If this shuffle node is simply a swizzle of another shuffle node,
  // and it reverses the swizzle of the previous shuffle then we can
  // optimize shuffle(shuffle(x, undef), undef) -> x.
  if (N0.getOpcode() == ISD::VECTOR_SHUFFLE && Level < AfterLegalizeDAG &&
      N1.getOpcode() == ISD::UNDEF) {

    ShuffleVectorSDNode *OtherSV = cast<ShuffleVectorSDNode>(N0);

    // Shuffle nodes can only reverse shuffles with a single non-undef value.
    if (N0.getOperand(1).getOpcode() != ISD::UNDEF)
      return SDValue();

    // The incoming shuffle must be of the same type as the result of the
    // current shuffle.
    assert(OtherSV->getOperand(0).getValueType() == VT &&
           "Shuffle types don't match");

    for (unsigned i = 0; i != NumElts; ++i) {
      int Idx = SVN->getMaskElt(i);
      assert(Idx < (int)NumElts && "Index references undef operand");
      // Next, this index comes from the first value, which is the incoming
      // shuffle. Adopt the incoming index.
      if (Idx >= 0)
        Idx = OtherSV->getMaskElt(Idx);

      // The combined shuffle must map each index to itself.
      if (Idx >= 0 && (unsigned)Idx != i)
        return SDValue();
    }

    return OtherSV->getOperand(0);
  }

  return SDValue();
}

SDValue DAGCombiner::visitMEMBARRIER(SDNode* N) {
  if (!TLI.getShouldFoldAtomicFences())
    return SDValue();

  SDValue atomic = N->getOperand(0);
  switch (atomic.getOpcode()) {
    case ISD::ATOMIC_CMP_SWAP:
    case ISD::ATOMIC_SWAP:
    case ISD::ATOMIC_LOAD_ADD:
    case ISD::ATOMIC_LOAD_SUB:
    case ISD::ATOMIC_LOAD_AND:
    case ISD::ATOMIC_LOAD_OR:
    case ISD::ATOMIC_LOAD_XOR:
    case ISD::ATOMIC_LOAD_NAND:
    case ISD::ATOMIC_LOAD_MIN:
    case ISD::ATOMIC_LOAD_MAX:
    case ISD::ATOMIC_LOAD_UMIN:
    case ISD::ATOMIC_LOAD_UMAX:
      break;
    default:
      return SDValue();
  }

  SDValue fence = atomic.getOperand(0);
  if (fence.getOpcode() != ISD::MEMBARRIER)
    return SDValue();

  switch (atomic.getOpcode()) {
    case ISD::ATOMIC_CMP_SWAP:
      return SDValue(DAG.UpdateNodeOperands(atomic.getNode(),
                                    fence.getOperand(0),
                                    atomic.getOperand(1), atomic.getOperand(2),
                                    atomic.getOperand(3)), atomic.getResNo());
    case ISD::ATOMIC_SWAP:
    case ISD::ATOMIC_LOAD_ADD:
    case ISD::ATOMIC_LOAD_SUB:
    case ISD::ATOMIC_LOAD_AND:
    case ISD::ATOMIC_LOAD_OR:
    case ISD::ATOMIC_LOAD_XOR:
    case ISD::ATOMIC_LOAD_NAND:
    case ISD::ATOMIC_LOAD_MIN:
    case ISD::ATOMIC_LOAD_MAX:
    case ISD::ATOMIC_LOAD_UMIN:
    case ISD::ATOMIC_LOAD_UMAX:
      return SDValue(DAG.UpdateNodeOperands(atomic.getNode(),
                                    fence.getOperand(0),
                                    atomic.getOperand(1), atomic.getOperand(2)),
                     atomic.getResNo());
    default:
      return SDValue();
  }
}

/// XformToShuffleWithZero - Returns a vector_shuffle if it able to transform
/// an AND to a vector_shuffle with the destination vector and a zero vector.
/// e.g. AND V, <0xffffffff, 0, 0xffffffff, 0>. ==>
///      vector_shuffle V, Zero, <0, 4, 2, 4>
SDValue DAGCombiner::XformToShuffleWithZero(SDNode *N) {
  EVT VT = N->getValueType(0);
  DebugLoc dl = N->getDebugLoc();
  SDValue LHS = N->getOperand(0);
  SDValue RHS = N->getOperand(1);
  if (N->getOpcode() == ISD::AND) {
    if (RHS.getOpcode() == ISD::BITCAST)
      RHS = RHS.getOperand(0);
    if (RHS.getOpcode() == ISD::BUILD_VECTOR) {
      SmallVector<int, 8> Indices;
      unsigned NumElts = RHS.getNumOperands();
      for (unsigned i = 0; i != NumElts; ++i) {
        SDValue Elt = RHS.getOperand(i);
        if (!isa<ConstantSDNode>(Elt))
          return SDValue();

        if (cast<ConstantSDNode>(Elt)->isAllOnesValue())
          Indices.push_back(i);
        else if (cast<ConstantSDNode>(Elt)->isNullValue())
          Indices.push_back(NumElts);
        else
          return SDValue();
      }

      // Let's see if the target supports this vector_shuffle.
      EVT RVT = RHS.getValueType();
      if (!TLI.isVectorClearMaskLegal(Indices, RVT))
        return SDValue();

      // Return the new VECTOR_SHUFFLE node.
      EVT EltVT = RVT.getVectorElementType();
      SmallVector<SDValue,8> ZeroOps(RVT.getVectorNumElements(),
                                     DAG.getConstant(0, EltVT));
      SDValue Zero = DAG.getNode(ISD::BUILD_VECTOR, N->getDebugLoc(),
                                 RVT, &ZeroOps[0], ZeroOps.size());
      LHS = DAG.getNode(ISD::BITCAST, dl, RVT, LHS);
      SDValue Shuf = DAG.getVectorShuffle(RVT, dl, LHS, Zero, &Indices[0]);
      return DAG.getNode(ISD::BITCAST, dl, VT, Shuf);
    }
  }

  return SDValue();
}

/// SimplifyVBinOp - Visit a binary vector operation, like ADD.
SDValue DAGCombiner::SimplifyVBinOp(SDNode *N) {
  // After legalize, the target may be depending on adds and other
  // binary ops to provide legal ways to construct constants or other
  // things. Simplifying them may result in a loss of legality.
  if (LegalOperations) return SDValue();

  assert(N->getValueType(0).isVector() &&
         "SimplifyVBinOp only works on vectors!");

  SDValue LHS = N->getOperand(0);
  SDValue RHS = N->getOperand(1);
  SDValue Shuffle = XformToShuffleWithZero(N);
  if (Shuffle.getNode()) return Shuffle;

  // If the LHS and RHS are BUILD_VECTOR nodes, see if we can constant fold
  // this operation.
  if (LHS.getOpcode() == ISD::BUILD_VECTOR &&
      RHS.getOpcode() == ISD::BUILD_VECTOR) {
    SmallVector<SDValue, 8> Ops;
    for (unsigned i = 0, e = LHS.getNumOperands(); i != e; ++i) {
      SDValue LHSOp = LHS.getOperand(i);
      SDValue RHSOp = RHS.getOperand(i);
      // If these two elements can't be folded, bail out.
      if ((LHSOp.getOpcode() != ISD::UNDEF &&
           LHSOp.getOpcode() != ISD::Constant &&
           LHSOp.getOpcode() != ISD::ConstantFP) ||
          (RHSOp.getOpcode() != ISD::UNDEF &&
           RHSOp.getOpcode() != ISD::Constant &&
           RHSOp.getOpcode() != ISD::ConstantFP))
        break;

      // Can't fold divide by zero.
      if (N->getOpcode() == ISD::SDIV || N->getOpcode() == ISD::UDIV ||
          N->getOpcode() == ISD::FDIV) {
        if ((RHSOp.getOpcode() == ISD::Constant &&
             cast<ConstantSDNode>(RHSOp.getNode())->isNullValue()) ||
            (RHSOp.getOpcode() == ISD::ConstantFP &&
             cast<ConstantFPSDNode>(RHSOp.getNode())->getValueAPF().isZero()))
          break;
      }

      EVT VT = LHSOp.getValueType();
      EVT RVT = RHSOp.getValueType();
      if (RVT != VT) {
        // Integer BUILD_VECTOR operands may have types larger than the element
        // size (e.g., when the element type is not legal).  Prior to type
        // legalization, the types may not match between the two BUILD_VECTORS.
        // Truncate one of the operands to make them match.
        if (RVT.getSizeInBits() > VT.getSizeInBits()) {
          RHSOp = DAG.getNode(ISD::TRUNCATE, N->getDebugLoc(), VT, RHSOp);
        } else {
          LHSOp = DAG.getNode(ISD::TRUNCATE, N->getDebugLoc(), RVT, LHSOp);
          VT = RVT;
        }
      }
      SDValue FoldOp = DAG.getNode(N->getOpcode(), LHS.getDebugLoc(), VT,
                                   LHSOp, RHSOp);
      if (FoldOp.getOpcode() != ISD::UNDEF &&
          FoldOp.getOpcode() != ISD::Constant &&
          FoldOp.getOpcode() != ISD::ConstantFP)
        break;
      Ops.push_back(FoldOp);
      AddToWorkList(FoldOp.getNode());
    }

    if (Ops.size() == LHS.getNumOperands())
      return DAG.getNode(ISD::BUILD_VECTOR, N->getDebugLoc(),
                         LHS.getValueType(), &Ops[0], Ops.size());
  }

  return SDValue();
}

/// SimplifyVUnaryOp - Visit a binary vector operation, like FABS/FNEG.
SDValue DAGCombiner::SimplifyVUnaryOp(SDNode *N) {
  // After legalize, the target may be depending on adds and other
  // binary ops to provide legal ways to construct constants or other
  // things. Simplifying them may result in a loss of legality.
  if (LegalOperations) return SDValue();

  assert(N->getValueType(0).isVector() &&
         "SimplifyVUnaryOp only works on vectors!");

  SDValue N0 = N->getOperand(0);

  if (N0.getOpcode() != ISD::BUILD_VECTOR)
    return SDValue();

  // Operand is a BUILD_VECTOR node, see if we can constant fold it.
  SmallVector<SDValue, 8> Ops;
  for (unsigned i = 0, e = N0.getNumOperands(); i != e; ++i) {
    SDValue Op = N0.getOperand(i);
    if (Op.getOpcode() != ISD::UNDEF &&
        Op.getOpcode() != ISD::ConstantFP)
      break;
    EVT EltVT = Op.getValueType();
    SDValue FoldOp = DAG.getNode(N->getOpcode(), N0.getDebugLoc(), EltVT, Op);
    if (FoldOp.getOpcode() != ISD::UNDEF &&
        FoldOp.getOpcode() != ISD::ConstantFP)
      break;
    Ops.push_back(FoldOp);
    AddToWorkList(FoldOp.getNode());
  }

  if (Ops.size() != N0.getNumOperands())
    return SDValue();

  return DAG.getNode(ISD::BUILD_VECTOR, N->getDebugLoc(),
                     N0.getValueType(), &Ops[0], Ops.size());
}

SDValue DAGCombiner::SimplifySelect(DebugLoc DL, SDValue N0,
                                    SDValue N1, SDValue N2){
  assert(N0.getOpcode() ==ISD::SETCC && "First argument must be a SetCC node!");

  SDValue SCC = SimplifySelectCC(DL, N0.getOperand(0), N0.getOperand(1), N1, N2,
                                 cast<CondCodeSDNode>(N0.getOperand(2))->get());

  // If we got a simplified select_cc node back from SimplifySelectCC, then
  // break it down into a new SETCC node, and a new SELECT node, and then return
  // the SELECT node, since we were called with a SELECT node.
  if (SCC.getNode()) {
    // Check to see if we got a select_cc back (to turn into setcc/select).
    // Otherwise, just return whatever node we got back, like fabs.
    if (SCC.getOpcode() == ISD::SELECT_CC) {
      SDValue SETCC = DAG.getNode(ISD::SETCC, N0.getDebugLoc(),
                                  N0.getValueType(),
                                  SCC.getOperand(0), SCC.getOperand(1),
                                  SCC.getOperand(4));
      AddToWorkList(SETCC.getNode());
      return DAG.getNode(ISD::SELECT, SCC.getDebugLoc(), SCC.getValueType(),
                         SCC.getOperand(2), SCC.getOperand(3), SETCC);
    }

    return SCC;
  }
  return SDValue();
}

/// SimplifySelectOps - Given a SELECT or a SELECT_CC node, where LHS and RHS
/// are the two values being selected between, see if we can simplify the
/// select.  Callers of this should assume that TheSelect is deleted if this
/// returns true.  As such, they should return the appropriate thing (e.g. the
/// node) back to the top-level of the DAG combiner loop to avoid it being
/// looked at.
bool DAGCombiner::SimplifySelectOps(SDNode *TheSelect, SDValue LHS,
                                    SDValue RHS) {

  // Cannot simplify select with vector condition
  if (TheSelect->getOperand(0).getValueType().isVector()) return false;

  // If this is a select from two identical things, try to pull the operation
  // through the select.
  if (LHS.getOpcode() != RHS.getOpcode() ||
      !LHS.hasOneUse() || !RHS.hasOneUse())
    return false;

  // If this is a load and the token chain is identical, replace the select
  // of two loads with a load through a select of the address to load from.
  // This triggers in things like "select bool X, 10.0, 123.0" after the FP
  // constants have been dropped into the constant pool.
  if (LHS.getOpcode() == ISD::LOAD) {
    LoadSDNode *LLD = cast<LoadSDNode>(LHS);
    LoadSDNode *RLD = cast<LoadSDNode>(RHS);

    // Token chains must be identical.
    if (LHS.getOperand(0) != RHS.getOperand(0) ||
        // Do not let this transformation reduce the number of volatile loads.
        LLD->isVolatile() || RLD->isVolatile() ||
        // If this is an EXTLOAD, the VT's must match.
        LLD->getMemoryVT() != RLD->getMemoryVT() ||
        // If this is an EXTLOAD, the kind of extension must match.
        (LLD->getExtensionType() != RLD->getExtensionType() &&
         // The only exception is if one of the extensions is anyext.
         LLD->getExtensionType() != ISD::EXTLOAD &&
         RLD->getExtensionType() != ISD::EXTLOAD) ||
        // FIXME: this discards src value information.  This is
        // over-conservative. It would be beneficial to be able to remember
        // both potential memory locations.  Since we are discarding
        // src value info, don't do the transformation if the memory
        // locations are not in the default address space.
        LLD->getPointerInfo().getAddrSpace() != 0 ||
        RLD->getPointerInfo().getAddrSpace() != 0)
      return false;

    // Check that the select condition doesn't reach either load.  If so,
    // folding this will induce a cycle into the DAG.  If not, this is safe to
    // xform, so create a select of the addresses.
    SDValue Addr;
    if (TheSelect->getOpcode() == ISD::SELECT) {
      SDNode *CondNode = TheSelect->getOperand(0).getNode();
      if ((LLD->hasAnyUseOfValue(1) && LLD->isPredecessorOf(CondNode)) ||
          (RLD->hasAnyUseOfValue(1) && RLD->isPredecessorOf(CondNode)))
        return false;
      Addr = DAG.getNode(ISD::SELECT, TheSelect->getDebugLoc(),
                         LLD->getBasePtr().getValueType(),
                         TheSelect->getOperand(0), LLD->getBasePtr(),
                         RLD->getBasePtr());
    } else {  // Otherwise SELECT_CC
      SDNode *CondLHS = TheSelect->getOperand(0).getNode();
      SDNode *CondRHS = TheSelect->getOperand(1).getNode();

      if ((LLD->hasAnyUseOfValue(1) &&
           (LLD->isPredecessorOf(CondLHS) || LLD->isPredecessorOf(CondRHS))) ||
          (RLD->hasAnyUseOfValue(1) &&
           (RLD->isPredecessorOf(CondLHS) || RLD->isPredecessorOf(CondRHS))))
        return false;

      Addr = DAG.getNode(ISD::SELECT_CC, TheSelect->getDebugLoc(),
                         LLD->getBasePtr().getValueType(),
                         TheSelect->getOperand(0),
                         TheSelect->getOperand(1),
                         LLD->getBasePtr(), RLD->getBasePtr(),
                         TheSelect->getOperand(4));
    }

    SDValue Load;
    if (LLD->getExtensionType() == ISD::NON_EXTLOAD) {
      Load = DAG.getLoad(TheSelect->getValueType(0),
                         TheSelect->getDebugLoc(),
                         // FIXME: Discards pointer info.
                         LLD->getChain(), Addr, MachinePointerInfo(),
                         LLD->isVolatile(), LLD->isNonTemporal(),
                         LLD->isInvariant(), LLD->getAlignment());
    } else {
      Load = DAG.getExtLoad(LLD->getExtensionType() == ISD::EXTLOAD ?
                            RLD->getExtensionType() : LLD->getExtensionType(),
                            TheSelect->getDebugLoc(),
                            TheSelect->getValueType(0),
                            // FIXME: Discards pointer info.
                            LLD->getChain(), Addr, MachinePointerInfo(),
                            LLD->getMemoryVT(), LLD->isVolatile(),
                            LLD->isNonTemporal(), LLD->getAlignment());
    }

    // Users of the select now use the result of the load.
    CombineTo(TheSelect, Load);

    // Users of the old loads now use the new load's chain.  We know the
    // old-load value is dead now.
    CombineTo(LHS.getNode(), Load.getValue(0), Load.getValue(1));
    CombineTo(RHS.getNode(), Load.getValue(0), Load.getValue(1));
    return true;
  }

  return false;
}

/// SimplifySelectCC - Simplify an expression of the form (N0 cond N1) ? N2 : N3
/// where 'cond' is the comparison specified by CC.
SDValue DAGCombiner::SimplifySelectCC(DebugLoc DL, SDValue N0, SDValue N1,
                                      SDValue N2, SDValue N3,
                                      ISD::CondCode CC, bool NotExtCompare) {
  // (x ? y : y) -> y.
  if (N2 == N3) return N2;

  EVT VT = N2.getValueType();
  ConstantSDNode *N1C = dyn_cast<ConstantSDNode>(N1.getNode());
  ConstantSDNode *N2C = dyn_cast<ConstantSDNode>(N2.getNode());
  ConstantSDNode *N3C = dyn_cast<ConstantSDNode>(N3.getNode());

  // Determine if the condition we're dealing with is constant
  SDValue SCC = SimplifySetCC(TLI.getSetCCResultType(N0.getValueType()),
                              N0, N1, CC, DL, false);
  if (SCC.getNode()) AddToWorkList(SCC.getNode());
  ConstantSDNode *SCCC = dyn_cast_or_null<ConstantSDNode>(SCC.getNode());

  // fold select_cc true, x, y -> x
  if (SCCC && !SCCC->isNullValue())
    return N2;
  // fold select_cc false, x, y -> y
  if (SCCC && SCCC->isNullValue())
    return N3;

  // Check to see if we can simplify the select into an fabs node
  if (ConstantFPSDNode *CFP = dyn_cast<ConstantFPSDNode>(N1)) {
    // Allow either -0.0 or 0.0
    if (CFP->getValueAPF().isZero()) {
      // select (setg[te] X, +/-0.0), X, fneg(X) -> fabs
      if ((CC == ISD::SETGE || CC == ISD::SETGT) &&
          N0 == N2 && N3.getOpcode() == ISD::FNEG &&
          N2 == N3.getOperand(0))
        return DAG.getNode(ISD::FABS, DL, VT, N0);

      // select (setl[te] X, +/-0.0), fneg(X), X -> fabs
      if ((CC == ISD::SETLT || CC == ISD::SETLE) &&
          N0 == N3 && N2.getOpcode() == ISD::FNEG &&
          N2.getOperand(0) == N3)
        return DAG.getNode(ISD::FABS, DL, VT, N3);
    }
  }

  // Turn "(a cond b) ? 1.0f : 2.0f" into "load (tmp + ((a cond b) ? 0 : 4)"
  // where "tmp" is a constant pool entry containing an array with 1.0 and 2.0
  // in it.  This is a win when the constant is not otherwise available because
  // it replaces two constant pool loads with one.  We only do this if the FP
  // type is known to be legal, because if it isn't, then we are before legalize
  // types an we want the other legalization to happen first (e.g. to avoid
  // messing with soft float) and if the ConstantFP is not legal, because if
  // it is legal, we may not need to store the FP constant in a constant pool.
  if (ConstantFPSDNode *TV = dyn_cast<ConstantFPSDNode>(N2))
    if (ConstantFPSDNode *FV = dyn_cast<ConstantFPSDNode>(N3)) {
      if (TLI.isTypeLegal(N2.getValueType()) &&
          (TLI.getOperationAction(ISD::ConstantFP, N2.getValueType()) !=
           TargetLowering::Legal) &&
          // If both constants have multiple uses, then we won't need to do an
          // extra load, they are likely around in registers for other users.
          (TV->hasOneUse() || FV->hasOneUse())) {
        Constant *Elts[] = {
          const_cast<ConstantFP*>(FV->getConstantFPValue()),
          const_cast<ConstantFP*>(TV->getConstantFPValue())
        };
        Type *FPTy = Elts[0]->getType();
        const DataLayout &TD = *TLI.getDataLayout();

        // Create a ConstantArray of the two constants.
        Constant *CA = ConstantArray::get(ArrayType::get(FPTy, 2), Elts);
        SDValue CPIdx = DAG.getConstantPool(CA, TLI.getPointerTy(),
                                            TD.getPrefTypeAlignment(FPTy));
        unsigned Alignment = cast<ConstantPoolSDNode>(CPIdx)->getAlignment();

        // Get the offsets to the 0 and 1 element of the array so that we can
        // select between them.
        SDValue Zero = DAG.getIntPtrConstant(0);
        unsigned EltSize = (unsigned)TD.getTypeAllocSize(Elts[0]->getType());
        SDValue One = DAG.getIntPtrConstant(EltSize);

        SDValue Cond = DAG.getSetCC(DL,
                                    TLI.getSetCCResultType(N0.getValueType()),
                                    N0, N1, CC);
        AddToWorkList(Cond.getNode());
        SDValue CstOffset = DAG.getNode(ISD::SELECT, DL, Zero.getValueType(),
                                        Cond, One, Zero);
        AddToWorkList(CstOffset.getNode());
        CPIdx = DAG.getNode(ISD::ADD, DL, TLI.getPointerTy(), CPIdx,
                            CstOffset);
        AddToWorkList(CPIdx.getNode());
        return DAG.getLoad(TV->getValueType(0), DL, DAG.getEntryNode(), CPIdx,
                           MachinePointerInfo::getConstantPool(), false,
                           false, false, Alignment);

      }
    }

  // Check to see if we can perform the "gzip trick", transforming
  // (select_cc setlt X, 0, A, 0) -> (and (sra X, (sub size(X), 1), A)
  if (N1C && N3C && N3C->isNullValue() && CC == ISD::SETLT &&
      (N1C->isNullValue() ||                         // (a < 0) ? b : 0
       (N1C->getAPIntValue() == 1 && N0 == N2))) {   // (a < 1) ? a : 0
    EVT XType = N0.getValueType();
    EVT AType = N2.getValueType();
    if (XType.bitsGE(AType)) {
      // and (sra X, size(X)-1, A) -> "and (srl X, C2), A" iff A is a
      // single-bit constant.
      if (N2C && ((N2C->getAPIntValue() & (N2C->getAPIntValue()-1)) == 0)) {
        unsigned ShCtV = N2C->getAPIntValue().logBase2();
        ShCtV = XType.getSizeInBits()-ShCtV-1;
        SDValue ShCt = DAG.getConstant(ShCtV,
                                       getShiftAmountTy(N0.getValueType()));
        SDValue Shift = DAG.getNode(ISD::SRL, N0.getDebugLoc(),
                                    XType, N0, ShCt);
        AddToWorkList(Shift.getNode());

        if (XType.bitsGT(AType)) {
          Shift = DAG.getNode(ISD::TRUNCATE, DL, AType, Shift);
          AddToWorkList(Shift.getNode());
        }

        return DAG.getNode(ISD::AND, DL, AType, Shift, N2);
      }

      SDValue Shift = DAG.getNode(ISD::SRA, N0.getDebugLoc(),
                                  XType, N0,
                                  DAG.getConstant(XType.getSizeInBits()-1,
                                         getShiftAmountTy(N0.getValueType())));
      AddToWorkList(Shift.getNode());

      if (XType.bitsGT(AType)) {
        Shift = DAG.getNode(ISD::TRUNCATE, DL, AType, Shift);
        AddToWorkList(Shift.getNode());
      }

      return DAG.getNode(ISD::AND, DL, AType, Shift, N2);
    }
  }

  // fold (select_cc seteq (and x, y), 0, 0, A) -> (and (shr (shl x)) A)
  // where y is has a single bit set.
  // A plaintext description would be, we can turn the SELECT_CC into an AND
  // when the condition can be materialized as an all-ones register.  Any
  // single bit-test can be materialized as an all-ones register with
  // shift-left and shift-right-arith.
  if (CC == ISD::SETEQ && N0->getOpcode() == ISD::AND &&
      N0->getValueType(0) == VT &&
      N1C && N1C->isNullValue() &&
      N2C && N2C->isNullValue()) {
    SDValue AndLHS = N0->getOperand(0);
    ConstantSDNode *ConstAndRHS = dyn_cast<ConstantSDNode>(N0->getOperand(1));
    if (ConstAndRHS && ConstAndRHS->getAPIntValue().countPopulation() == 1) {
      // Shift the tested bit over the sign bit.
      APInt AndMask = ConstAndRHS->getAPIntValue();
      SDValue ShlAmt =
        DAG.getConstant(AndMask.countLeadingZeros(),
                        getShiftAmountTy(AndLHS.getValueType()));
      SDValue Shl = DAG.getNode(ISD::SHL, N0.getDebugLoc(), VT, AndLHS, ShlAmt);

      // Now arithmetic right shift it all the way over, so the result is either
      // all-ones, or zero.
      SDValue ShrAmt =
        DAG.getConstant(AndMask.getBitWidth()-1,
                        getShiftAmountTy(Shl.getValueType()));
      SDValue Shr = DAG.getNode(ISD::SRA, N0.getDebugLoc(), VT, Shl, ShrAmt);

      return DAG.getNode(ISD::AND, DL, VT, Shr, N3);
    }
  }

  // fold select C, 16, 0 -> shl C, 4
  if (N2C && N3C && N3C->isNullValue() && N2C->getAPIntValue().isPowerOf2() &&
    TLI.getBooleanContents(N0.getValueType().isVector()) ==
      TargetLowering::ZeroOrOneBooleanContent) {

    // If the caller doesn't want us to simplify this into a zext of a compare,
    // don't do it.
    if (NotExtCompare && N2C->getAPIntValue() == 1)
      return SDValue();

    // Get a SetCC of the condition
    // FIXME: Should probably make sure that setcc is legal if we ever have a
    // target where it isn't.
    SDValue Temp, SCC;
    // cast from setcc result type to select result type
    if (LegalTypes) {
      SCC  = DAG.getSetCC(DL, TLI.getSetCCResultType(N0.getValueType()),
                          N0, N1, CC);
      if (N2.getValueType().bitsLT(SCC.getValueType()))
        Temp = DAG.getZeroExtendInReg(SCC, N2.getDebugLoc(), N2.getValueType());
      else
        Temp = DAG.getNode(ISD::ZERO_EXTEND, N2.getDebugLoc(),
                           N2.getValueType(), SCC);
    } else {
      SCC  = DAG.getSetCC(N0.getDebugLoc(), MVT::i1, N0, N1, CC);
      Temp = DAG.getNode(ISD::ZERO_EXTEND, N2.getDebugLoc(),
                         N2.getValueType(), SCC);
    }

    AddToWorkList(SCC.getNode());
    AddToWorkList(Temp.getNode());

    if (N2C->getAPIntValue() == 1)
      return Temp;

    // shl setcc result by log2 n2c
    return DAG.getNode(ISD::SHL, DL, N2.getValueType(), Temp,
                       DAG.getConstant(N2C->getAPIntValue().logBase2(),
                                       getShiftAmountTy(Temp.getValueType())));
  }

  // Check to see if this is the equivalent of setcc
  // FIXME: Turn all of these into setcc if setcc if setcc is legal
  // otherwise, go ahead with the folds.
  if (0 && N3C && N3C->isNullValue() && N2C && (N2C->getAPIntValue() == 1ULL)) {
    EVT XType = N0.getValueType();
    if (!LegalOperations ||
        TLI.isOperationLegal(ISD::SETCC, TLI.getSetCCResultType(XType))) {
      SDValue Res = DAG.getSetCC(DL, TLI.getSetCCResultType(XType), N0, N1, CC);
      if (Res.getValueType() != VT)
        Res = DAG.getNode(ISD::ZERO_EXTEND, DL, VT, Res);
      return Res;
    }

    // fold (seteq X, 0) -> (srl (ctlz X, log2(size(X))))
    if (N1C && N1C->isNullValue() && CC == ISD::SETEQ &&
        (!LegalOperations ||
         TLI.isOperationLegal(ISD::CTLZ, XType))) {
      SDValue Ctlz = DAG.getNode(ISD::CTLZ, N0.getDebugLoc(), XType, N0);
      return DAG.getNode(ISD::SRL, DL, XType, Ctlz,
                         DAG.getConstant(Log2_32(XType.getSizeInBits()),
                                       getShiftAmountTy(Ctlz.getValueType())));
    }
    // fold (setgt X, 0) -> (srl (and (-X, ~X), size(X)-1))
    if (N1C && N1C->isNullValue() && CC == ISD::SETGT) {
      SDValue NegN0 = DAG.getNode(ISD::SUB, N0.getDebugLoc(),
                                  XType, DAG.getConstant(0, XType), N0);
      SDValue NotN0 = DAG.getNOT(N0.getDebugLoc(), N0, XType);
      return DAG.getNode(ISD::SRL, DL, XType,
                         DAG.getNode(ISD::AND, DL, XType, NegN0, NotN0),
                         DAG.getConstant(XType.getSizeInBits()-1,
                                         getShiftAmountTy(XType)));
    }
    // fold (setgt X, -1) -> (xor (srl (X, size(X)-1), 1))
    if (N1C && N1C->isAllOnesValue() && CC == ISD::SETGT) {
      SDValue Sign = DAG.getNode(ISD::SRL, N0.getDebugLoc(), XType, N0,
                                 DAG.getConstant(XType.getSizeInBits()-1,
                                         getShiftAmountTy(N0.getValueType())));
      return DAG.getNode(ISD::XOR, DL, XType, Sign, DAG.getConstant(1, XType));
    }
  }

  // Check to see if this is an integer abs.
  // select_cc setg[te] X,  0,  X, -X ->
  // select_cc setgt    X, -1,  X, -X ->
  // select_cc setl[te] X,  0, -X,  X ->
  // select_cc setlt    X,  1, -X,  X ->
  // Y = sra (X, size(X)-1); xor (add (X, Y), Y)
  if (N1C) {
    ConstantSDNode *SubC = NULL;
    if (((N1C->isNullValue() && (CC == ISD::SETGT || CC == ISD::SETGE)) ||
         (N1C->isAllOnesValue() && CC == ISD::SETGT)) &&
        N0 == N2 && N3.getOpcode() == ISD::SUB && N0 == N3.getOperand(1))
      SubC = dyn_cast<ConstantSDNode>(N3.getOperand(0));
    else if (((N1C->isNullValue() && (CC == ISD::SETLT || CC == ISD::SETLE)) ||
              (N1C->isOne() && CC == ISD::SETLT)) &&
             N0 == N3 && N2.getOpcode() == ISD::SUB && N0 == N2.getOperand(1))
      SubC = dyn_cast<ConstantSDNode>(N2.getOperand(0));

    EVT XType = N0.getValueType();
    if (SubC && SubC->isNullValue() && XType.isInteger()) {
      SDValue Shift = DAG.getNode(ISD::SRA, N0.getDebugLoc(), XType,
                                  N0,
                                  DAG.getConstant(XType.getSizeInBits()-1,
                                         getShiftAmountTy(N0.getValueType())));
      SDValue Add = DAG.getNode(ISD::ADD, N0.getDebugLoc(),
                                XType, N0, Shift);
      AddToWorkList(Shift.getNode());
      AddToWorkList(Add.getNode());
      return DAG.getNode(ISD::XOR, DL, XType, Add, Shift);
    }
  }

  return SDValue();
}

/// SimplifySetCC - This is a stub for TargetLowering::SimplifySetCC.
SDValue DAGCombiner::SimplifySetCC(EVT VT, SDValue N0,
                                   SDValue N1, ISD::CondCode Cond,
                                   DebugLoc DL, bool foldBooleans) {
  TargetLowering::DAGCombinerInfo
    DagCombineInfo(DAG, !LegalTypes, !LegalOperations, false, this);
  return TLI.SimplifySetCC(VT, N0, N1, Cond, foldBooleans, DagCombineInfo, DL);
}

/// BuildSDIVSequence - Given an ISD::SDIV node expressing a divide by constant,
/// return a DAG expression to select that will generate the same value by
/// multiplying by a magic number.  See:
/// <http://the.wall.riscom.net/books/proc/ppc/cwg/code2.html>
SDValue DAGCombiner::BuildSDIV(SDNode *N) {
  std::vector<SDNode*> Built;
  SDValue S = TLI.BuildSDIV(N, DAG, LegalOperations, &Built);

  for (std::vector<SDNode*>::iterator ii = Built.begin(), ee = Built.end();
       ii != ee; ++ii)
    AddToWorkList(*ii);
  return S;
}

/// BuildUDIVSequence - Given an ISD::UDIV node expressing a divide by constant,
/// return a DAG expression to select that will generate the same value by
/// multiplying by a magic number.  See:
/// <http://the.wall.riscom.net/books/proc/ppc/cwg/code2.html>
SDValue DAGCombiner::BuildUDIV(SDNode *N) {
  std::vector<SDNode*> Built;
  SDValue S = TLI.BuildUDIV(N, DAG, LegalOperations, &Built);

  for (std::vector<SDNode*>::iterator ii = Built.begin(), ee = Built.end();
       ii != ee; ++ii)
    AddToWorkList(*ii);
  return S;
}

/// FindBaseOffset - Return true if base is a frame index, which is known not
// to alias with anything but itself.  Provides base object and offset as
// results.
static bool FindBaseOffset(SDValue Ptr, SDValue &Base, int64_t &Offset,
                           const GlobalValue *&GV, const void *&CV) {
  // Assume it is a primitive operation.
  Base = Ptr; Offset = 0; GV = 0; CV = 0;

  // If it's an adding a simple constant then integrate the offset.
  if (Base.getOpcode() == ISD::ADD) {
    if (ConstantSDNode *C = dyn_cast<ConstantSDNode>(Base.getOperand(1))) {
      Base = Base.getOperand(0);
      Offset += C->getZExtValue();
    }
  }

  // Return the underlying GlobalValue, and update the Offset.  Return false
  // for GlobalAddressSDNode since the same GlobalAddress may be represented
  // by multiple nodes with different offsets.
  if (GlobalAddressSDNode *G = dyn_cast<GlobalAddressSDNode>(Base)) {
    GV = G->getGlobal();
    Offset += G->getOffset();
    return false;
  }

  // Return the underlying Constant value, and update the Offset.  Return false
  // for ConstantSDNodes since the same constant pool entry may be represented
  // by multiple nodes with different offsets.
  if (ConstantPoolSDNode *C = dyn_cast<ConstantPoolSDNode>(Base)) {
    CV = C->isMachineConstantPoolEntry() ? (const void *)C->getMachineCPVal()
                                         : (const void *)C->getConstVal();
    Offset += C->getOffset();
    return false;
  }
  // If it's any of the following then it can't alias with anything but itself.
  return isa<FrameIndexSDNode>(Base);
}

/// isAlias - Return true if there is any possibility that the two addresses
/// overlap.
bool DAGCombiner::isAlias(SDValue Ptr1, int64_t Size1,
                          const Value *SrcValue1, int SrcValueOffset1,
                          unsigned SrcValueAlign1,
                          const MDNode *TBAAInfo1,
                          SDValue Ptr2, int64_t Size2,
                          const Value *SrcValue2, int SrcValueOffset2,
                          unsigned SrcValueAlign2,
                          const MDNode *TBAAInfo2) const {
  // If they are the same then they must be aliases.
  if (Ptr1 == Ptr2) return true;

  // Gather base node and offset information.
  SDValue Base1, Base2;
  int64_t Offset1, Offset2;
  const GlobalValue *GV1, *GV2;
  const void *CV1, *CV2;
  bool isFrameIndex1 = FindBaseOffset(Ptr1, Base1, Offset1, GV1, CV1);
  bool isFrameIndex2 = FindBaseOffset(Ptr2, Base2, Offset2, GV2, CV2);

  // If they have a same base address then check to see if they overlap.
  if (Base1 == Base2 || (GV1 && (GV1 == GV2)) || (CV1 && (CV1 == CV2)))
    return !((Offset1 + Size1) <= Offset2 || (Offset2 + Size2) <= Offset1);

  // It is possible for different frame indices to alias each other, mostly
  // when tail call optimization reuses return address slots for arguments.
  // To catch this case, look up the actual index of frame indices to compute
  // the real alias relationship.
  if (isFrameIndex1 && isFrameIndex2) {
    MachineFrameInfo *MFI = DAG.getMachineFunction().getFrameInfo();
    Offset1 += MFI->getObjectOffset(cast<FrameIndexSDNode>(Base1)->getIndex());
    Offset2 += MFI->getObjectOffset(cast<FrameIndexSDNode>(Base2)->getIndex());
    return !((Offset1 + Size1) <= Offset2 || (Offset2 + Size2) <= Offset1);
  }

  // Otherwise, if we know what the bases are, and they aren't identical, then
  // we know they cannot alias.
  if ((isFrameIndex1 || CV1 || GV1) && (isFrameIndex2 || CV2 || GV2))
    return false;

  // If we know required SrcValue1 and SrcValue2 have relatively large alignment
  // compared to the size and offset of the access, we may be able to prove they
  // do not alias.  This check is conservative for now to catch cases created by
  // splitting vector types.
  if ((SrcValueAlign1 == SrcValueAlign2) &&
      (SrcValueOffset1 != SrcValueOffset2) &&
      (Size1 == Size2) && (SrcValueAlign1 > Size1)) {
    int64_t OffAlign1 = SrcValueOffset1 % SrcValueAlign1;
    int64_t OffAlign2 = SrcValueOffset2 % SrcValueAlign1;

    // There is no overlap between these relatively aligned accesses of similar
    // size, return no alias.
    if ((OffAlign1 + Size1) <= OffAlign2 || (OffAlign2 + Size2) <= OffAlign1)
      return false;
  }

  if (CombinerGlobalAA) {
    // Use alias analysis information.
    int64_t MinOffset = std::min(SrcValueOffset1, SrcValueOffset2);
    int64_t Overlap1 = Size1 + SrcValueOffset1 - MinOffset;
    int64_t Overlap2 = Size2 + SrcValueOffset2 - MinOffset;
    AliasAnalysis::AliasResult AAResult =
      AA.alias(AliasAnalysis::Location(SrcValue1, Overlap1, TBAAInfo1),
               AliasAnalysis::Location(SrcValue2, Overlap2, TBAAInfo2));
    if (AAResult == AliasAnalysis::NoAlias)
      return false;
  }

  // Otherwise we have to assume they alias.
  return true;
}

/// FindAliasInfo - Extracts the relevant alias information from the memory
/// node.  Returns true if the operand was a load.
bool DAGCombiner::FindAliasInfo(SDNode *N,
                                SDValue &Ptr, int64_t &Size,
                                const Value *&SrcValue,
                                int &SrcValueOffset,
                                unsigned &SrcValueAlign,
                                const MDNode *&TBAAInfo) const {
  LSBaseSDNode *LS = cast<LSBaseSDNode>(N);

  Ptr = LS->getBasePtr();
  Size = LS->getMemoryVT().getSizeInBits() >> 3;
  SrcValue = LS->getSrcValue();
  SrcValueOffset = LS->getSrcValueOffset();
  SrcValueAlign = LS->getOriginalAlignment();
  TBAAInfo = LS->getTBAAInfo();
  return isa<LoadSDNode>(LS);
}

/// GatherAllAliases - Walk up chain skipping non-aliasing memory nodes,
/// looking for aliasing nodes and adding them to the Aliases vector.
void DAGCombiner::GatherAllAliases(SDNode *N, SDValue OriginalChain,
                                   SmallVector<SDValue, 8> &Aliases) {
  SmallVector<SDValue, 8> Chains;     // List of chains to visit.
  SmallPtrSet<SDNode *, 16> Visited;  // Visited node set.

  // Get alias information for node.
  SDValue Ptr;
  int64_t Size;
  const Value *SrcValue;
  int SrcValueOffset;
  unsigned SrcValueAlign;
  const MDNode *SrcTBAAInfo;
  bool IsLoad = FindAliasInfo(N, Ptr, Size, SrcValue, SrcValueOffset,
                              SrcValueAlign, SrcTBAAInfo);

  // Starting off.
  Chains.push_back(OriginalChain);
  unsigned Depth = 0;

  // Look at each chain and determine if it is an alias.  If so, add it to the
  // aliases list.  If not, then continue up the chain looking for the next
  // candidate.
  while (!Chains.empty()) {
    SDValue Chain = Chains.back();
    Chains.pop_back();

    // For TokenFactor nodes, look at each operand and only continue up the
    // chain until we find two aliases.  If we've seen two aliases, assume we'll
    // find more and revert to original chain since the xform is unlikely to be
    // profitable.
    //
    // FIXME: The depth check could be made to return the last non-aliasing
    // chain we found before we hit a tokenfactor rather than the original
    // chain.
    if (Depth > 6 || Aliases.size() == 2) {
      Aliases.clear();
      Aliases.push_back(OriginalChain);
      break;
    }

    // Don't bother if we've been before.
    if (!Visited.insert(Chain.getNode()))
      continue;

    switch (Chain.getOpcode()) {
    case ISD::EntryToken:
      // Entry token is ideal chain operand, but handled in FindBetterChain.
      break;

    case ISD::LOAD:
    case ISD::STORE: {
      // Get alias information for Chain.
      SDValue OpPtr;
      int64_t OpSize;
      const Value *OpSrcValue;
      int OpSrcValueOffset;
      unsigned OpSrcValueAlign;
      const MDNode *OpSrcTBAAInfo;
      bool IsOpLoad = FindAliasInfo(Chain.getNode(), OpPtr, OpSize,
                                    OpSrcValue, OpSrcValueOffset,
                                    OpSrcValueAlign,
                                    OpSrcTBAAInfo);

      // If chain is alias then stop here.
      if (!(IsLoad && IsOpLoad) &&
          isAlias(Ptr, Size, SrcValue, SrcValueOffset, SrcValueAlign,
                  SrcTBAAInfo,
                  OpPtr, OpSize, OpSrcValue, OpSrcValueOffset,
                  OpSrcValueAlign, OpSrcTBAAInfo)) {
        Aliases.push_back(Chain);
      } else {
        // Look further up the chain.
        Chains.push_back(Chain.getOperand(0));
        ++Depth;
      }
      break;
    }

    case ISD::TokenFactor:
      // We have to check each of the operands of the token factor for "small"
      // token factors, so we queue them up.  Adding the operands to the queue
      // (stack) in reverse order maintains the original order and increases the
      // likelihood that getNode will find a matching token factor (CSE.)
      if (Chain.getNumOperands() > 16) {
        Aliases.push_back(Chain);
        break;
      }
      for (unsigned n = Chain.getNumOperands(); n;)
        Chains.push_back(Chain.getOperand(--n));
      ++Depth;
      break;

    default:
      // For all other instructions we will just have to take what we can get.
      Aliases.push_back(Chain);
      break;
    }
  }
}

/// FindBetterChain - Walk up chain skipping non-aliasing memory nodes, looking
/// for a better chain (aliasing node.)
SDValue DAGCombiner::FindBetterChain(SDNode *N, SDValue OldChain) {
  SmallVector<SDValue, 8> Aliases;  // Ops for replacing token factor.

  // Accumulate all the aliases to this node.
  GatherAllAliases(N, OldChain, Aliases);

  // If no operands then chain to entry token.
  if (Aliases.size() == 0)
    return DAG.getEntryNode();

  // If a single operand then chain to it.  We don't need to revisit it.
  if (Aliases.size() == 1)
    return Aliases[0];

  // Construct a custom tailored token factor.
  return DAG.getNode(ISD::TokenFactor, N->getDebugLoc(), MVT::Other,
                     &Aliases[0], Aliases.size());
}

// SelectionDAG::Combine - This is the entry point for the file.
//
void SelectionDAG::Combine(CombineLevel Level, AliasAnalysis &AA,
                           CodeGenOpt::Level OptLevel) {
  /// run - This is the main entry point to this class.
  ///
  DAGCombiner(*this, AA, OptLevel).Run(Level);
}