aboutsummaryrefslogtreecommitdiffstats
path: root/lib/Target/PowerPC/PPCFastISel.cpp
blob: 4f8e6c1a101bda1b7e82450466b4ed5cfcae995c (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
//===-- PPCFastISel.cpp - PowerPC FastISel implementation -----------------===//
//
//                     The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This file defines the PowerPC-specific support for the FastISel class. Some
// of the target-specific code is generated by tablegen in the file
// PPCGenFastISel.inc, which is #included here.
//
//===----------------------------------------------------------------------===//

#define DEBUG_TYPE "ppcfastisel"
#include "PPC.h"
#include "PPCISelLowering.h"
#include "PPCSubtarget.h"
#include "PPCTargetMachine.h"
#include "MCTargetDesc/PPCPredicates.h"
#include "llvm/ADT/Optional.h"
#include "llvm/CodeGen/CallingConvLower.h"
#include "llvm/CodeGen/FastISel.h"
#include "llvm/CodeGen/FunctionLoweringInfo.h"
#include "llvm/CodeGen/MachineConstantPool.h"
#include "llvm/CodeGen/MachineFrameInfo.h"
#include "llvm/CodeGen/MachineInstrBuilder.h"
#include "llvm/CodeGen/MachineRegisterInfo.h"
#include "llvm/IR/CallingConv.h"
#include "llvm/IR/GlobalAlias.h"
#include "llvm/IR/GlobalVariable.h"
#include "llvm/IR/IntrinsicInst.h"
#include "llvm/IR/Operator.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/GetElementPtrTypeIterator.h"
#include "llvm/Target/TargetLowering.h"
#include "llvm/Target/TargetMachine.h"

//===----------------------------------------------------------------------===//
//
// TBD:
//   FastLowerArguments: Handle simple cases.
//   PPCMaterializeGV: Handle TLS.
//   SelectCall: Handle function pointers.
//   SelectCall: Handle multi-register return values.
//   SelectCall: Optimize away nops for local calls.
//   processCallArgs: Handle bit-converted arguments.
//   finishCall: Handle multi-register return values.
//   PPCComputeAddress: Handle parameter references as FrameIndex's.
//   PPCEmitCmp: Handle immediate as operand 1.
//   SelectCall: Handle small byval arguments.
//   SelectIntrinsicCall: Implement.
//   SelectSelect: Implement.
//   Consider factoring isTypeLegal into the base class.
//   Implement switches and jump tables.
//
//===----------------------------------------------------------------------===//
using namespace llvm;

namespace {

typedef struct Address {
  enum {
    RegBase,
    FrameIndexBase
  } BaseType;

  union {
    unsigned Reg;
    int FI;
  } Base;

  long Offset;

  // Innocuous defaults for our address.
  Address()
   : BaseType(RegBase), Offset(0) {
     Base.Reg = 0;
   }
} Address;

class PPCFastISel : public FastISel {

  const TargetMachine &TM;
  const TargetInstrInfo &TII;
  const TargetLowering &TLI;
  const PPCSubtarget &PPCSubTarget;
  LLVMContext *Context;

  public:
    explicit PPCFastISel(FunctionLoweringInfo &FuncInfo,
                         const TargetLibraryInfo *LibInfo)
    : FastISel(FuncInfo, LibInfo),
      TM(FuncInfo.MF->getTarget()),
      TII(*TM.getInstrInfo()),
      TLI(*TM.getTargetLowering()),
      PPCSubTarget(
       *((static_cast<const PPCTargetMachine *>(&TM))->getSubtargetImpl())
      ),
      Context(&FuncInfo.Fn->getContext()) { }

  // Backend specific FastISel code.
  private:
    virtual bool TargetSelectInstruction(const Instruction *I);
    virtual unsigned TargetMaterializeConstant(const Constant *C);
    virtual unsigned TargetMaterializeAlloca(const AllocaInst *AI);
    virtual bool tryToFoldLoadIntoMI(MachineInstr *MI, unsigned OpNo,
                                     const LoadInst *LI);
    virtual bool FastLowerArguments();
    virtual unsigned FastEmit_i(MVT Ty, MVT RetTy, unsigned Opc, uint64_t Imm);
    virtual unsigned FastEmitInst_ri(unsigned MachineInstOpcode,
                                     const TargetRegisterClass *RC,
                                     unsigned Op0, bool Op0IsKill,
                                     uint64_t Imm);
    virtual unsigned FastEmitInst_r(unsigned MachineInstOpcode,
                                    const TargetRegisterClass *RC,
                                    unsigned Op0, bool Op0IsKill);
    virtual unsigned FastEmitInst_rr(unsigned MachineInstOpcode,
                                     const TargetRegisterClass *RC,
                                     unsigned Op0, bool Op0IsKill,
                                     unsigned Op1, bool Op1IsKill);

  // Instruction selection routines.
  private:
    bool SelectLoad(const Instruction *I);
    bool SelectStore(const Instruction *I);
    bool SelectBranch(const Instruction *I);
    bool SelectIndirectBr(const Instruction *I);
    bool SelectCmp(const Instruction *I);
    bool SelectFPExt(const Instruction *I);
    bool SelectFPTrunc(const Instruction *I);
    bool SelectIToFP(const Instruction *I, bool IsSigned);
    bool SelectFPToI(const Instruction *I, bool IsSigned);
    bool SelectBinaryIntOp(const Instruction *I, unsigned ISDOpcode);
    bool SelectCall(const Instruction *I);
    bool SelectRet(const Instruction *I);
    bool SelectTrunc(const Instruction *I);
    bool SelectIntExt(const Instruction *I);

  // Utility routines.
  private:
    bool isTypeLegal(Type *Ty, MVT &VT);
    bool isLoadTypeLegal(Type *Ty, MVT &VT);
    bool PPCEmitCmp(const Value *Src1Value, const Value *Src2Value,
                    bool isZExt, unsigned DestReg);
    bool PPCEmitLoad(MVT VT, unsigned &ResultReg, Address &Addr,
                     const TargetRegisterClass *RC, bool IsZExt = true,
                     unsigned FP64LoadOpc = PPC::LFD);
    bool PPCEmitStore(MVT VT, unsigned SrcReg, Address &Addr);
    bool PPCComputeAddress(const Value *Obj, Address &Addr);
    void PPCSimplifyAddress(Address &Addr, MVT VT, bool &UseOffset,
                            unsigned &IndexReg);
    bool PPCEmitIntExt(MVT SrcVT, unsigned SrcReg, MVT DestVT,
                           unsigned DestReg, bool IsZExt);
    unsigned PPCMaterializeFP(const ConstantFP *CFP, MVT VT);
    unsigned PPCMaterializeGV(const GlobalValue *GV, MVT VT);
    unsigned PPCMaterializeInt(const Constant *C, MVT VT);
    unsigned PPCMaterialize32BitInt(int64_t Imm,
                                    const TargetRegisterClass *RC);
    unsigned PPCMaterialize64BitInt(int64_t Imm,
                                    const TargetRegisterClass *RC);
    unsigned PPCMoveToIntReg(const Instruction *I, MVT VT,
                             unsigned SrcReg, bool IsSigned);
    unsigned PPCMoveToFPReg(MVT VT, unsigned SrcReg, bool IsSigned);

  // Call handling routines.
  private:
    bool processCallArgs(SmallVectorImpl<Value*> &Args,
                         SmallVectorImpl<unsigned> &ArgRegs,
                         SmallVectorImpl<MVT> &ArgVTs,
                         SmallVectorImpl<ISD::ArgFlagsTy> &ArgFlags,
                         SmallVectorImpl<unsigned> &RegArgs,
                         CallingConv::ID CC,
                         unsigned &NumBytes,
                         bool IsVarArg);
    void finishCall(MVT RetVT, SmallVectorImpl<unsigned> &UsedRegs,
                    const Instruction *I, CallingConv::ID CC,
                    unsigned &NumBytes, bool IsVarArg);
    CCAssignFn *usePPC32CCs(unsigned Flag);

  private:
  #include "PPCGenFastISel.inc"

};

} // end anonymous namespace

#include "PPCGenCallingConv.inc"

// Function whose sole purpose is to kill compiler warnings 
// stemming from unused functions included from PPCGenCallingConv.inc.
CCAssignFn *PPCFastISel::usePPC32CCs(unsigned Flag) {
  if (Flag == 1)
    return CC_PPC32_SVR4;
  else if (Flag == 2)
    return CC_PPC32_SVR4_ByVal;
  else if (Flag == 3)
    return CC_PPC32_SVR4_VarArg;
  else
    return RetCC_PPC;
}

static Optional<PPC::Predicate> getComparePred(CmpInst::Predicate Pred) {
  switch (Pred) {
    // These are not representable with any single compare.
    case CmpInst::FCMP_FALSE:
    case CmpInst::FCMP_UEQ:
    case CmpInst::FCMP_UGT:
    case CmpInst::FCMP_UGE:
    case CmpInst::FCMP_ULT:
    case CmpInst::FCMP_ULE:
    case CmpInst::FCMP_UNE:
    case CmpInst::FCMP_TRUE:
    default:
      return Optional<PPC::Predicate>();

    case CmpInst::FCMP_OEQ:
    case CmpInst::ICMP_EQ:
      return PPC::PRED_EQ;

    case CmpInst::FCMP_OGT:
    case CmpInst::ICMP_UGT:
    case CmpInst::ICMP_SGT:
      return PPC::PRED_GT;

    case CmpInst::FCMP_OGE:
    case CmpInst::ICMP_UGE:
    case CmpInst::ICMP_SGE:
      return PPC::PRED_GE;

    case CmpInst::FCMP_OLT:
    case CmpInst::ICMP_ULT:
    case CmpInst::ICMP_SLT:
      return PPC::PRED_LT;

    case CmpInst::FCMP_OLE:
    case CmpInst::ICMP_ULE:
    case CmpInst::ICMP_SLE:
      return PPC::PRED_LE;

    case CmpInst::FCMP_ONE:
    case CmpInst::ICMP_NE:
      return PPC::PRED_NE;

    case CmpInst::FCMP_ORD:
      return PPC::PRED_NU;

    case CmpInst::FCMP_UNO:
      return PPC::PRED_UN;
  }
}

// Determine whether the type Ty is simple enough to be handled by
// fast-isel, and return its equivalent machine type in VT.
// FIXME: Copied directly from ARM -- factor into base class?
bool PPCFastISel::isTypeLegal(Type *Ty, MVT &VT) {
  EVT Evt = TLI.getValueType(Ty, true);

  // Only handle simple types.
  if (Evt == MVT::Other || !Evt.isSimple()) return false;
  VT = Evt.getSimpleVT();

  // Handle all legal types, i.e. a register that will directly hold this
  // value.
  return TLI.isTypeLegal(VT);
}

// Determine whether the type Ty is simple enough to be handled by
// fast-isel as a load target, and return its equivalent machine type in VT.
bool PPCFastISel::isLoadTypeLegal(Type *Ty, MVT &VT) {
  if (isTypeLegal(Ty, VT)) return true;

  // If this is a type than can be sign or zero-extended to a basic operation
  // go ahead and accept it now.
  if (VT == MVT::i8 || VT == MVT::i16 || VT == MVT::i32) {
    return true;
  }

  return false;
}

// Given a value Obj, create an Address object Addr that represents its
// address.  Return false if we can't handle it.
bool PPCFastISel::PPCComputeAddress(const Value *Obj, Address &Addr) {
  const User *U = NULL;
  unsigned Opcode = Instruction::UserOp1;
  if (const Instruction *I = dyn_cast<Instruction>(Obj)) {
    // Don't walk into other basic blocks unless the object is an alloca from
    // another block, otherwise it may not have a virtual register assigned.
    if (FuncInfo.StaticAllocaMap.count(static_cast<const AllocaInst *>(Obj)) ||
        FuncInfo.MBBMap[I->getParent()] == FuncInfo.MBB) {
      Opcode = I->getOpcode();
      U = I;
    }
  } else if (const ConstantExpr *C = dyn_cast<ConstantExpr>(Obj)) {
    Opcode = C->getOpcode();
    U = C;
  }

  switch (Opcode) {
    default:
      break;
    case Instruction::BitCast:
      // Look through bitcasts.
      return PPCComputeAddress(U->getOperand(0), Addr);
    case Instruction::IntToPtr:
      // Look past no-op inttoptrs.
      if (TLI.getValueType(U->getOperand(0)->getType()) == TLI.getPointerTy())
        return PPCComputeAddress(U->getOperand(0), Addr);
      break;
    case Instruction::PtrToInt:
      // Look past no-op ptrtoints.
      if (TLI.getValueType(U->getType()) == TLI.getPointerTy())
        return PPCComputeAddress(U->getOperand(0), Addr);
      break;
    case Instruction::GetElementPtr: {
      Address SavedAddr = Addr;
      long TmpOffset = Addr.Offset;

      // Iterate through the GEP folding the constants into offsets where
      // we can.
      gep_type_iterator GTI = gep_type_begin(U);
      for (User::const_op_iterator II = U->op_begin() + 1, IE = U->op_end();
           II != IE; ++II, ++GTI) {
        const Value *Op = *II;
        if (StructType *STy = dyn_cast<StructType>(*GTI)) {
          const StructLayout *SL = TD.getStructLayout(STy);
          unsigned Idx = cast<ConstantInt>(Op)->getZExtValue();
          TmpOffset += SL->getElementOffset(Idx);
        } else {
          uint64_t S = TD.getTypeAllocSize(GTI.getIndexedType());
          for (;;) {
            if (const ConstantInt *CI = dyn_cast<ConstantInt>(Op)) {
              // Constant-offset addressing.
              TmpOffset += CI->getSExtValue() * S;
              break;
            }
            if (isa<AddOperator>(Op) &&
                (!isa<Instruction>(Op) ||
                 FuncInfo.MBBMap[cast<Instruction>(Op)->getParent()]
                 == FuncInfo.MBB) &&
                isa<ConstantInt>(cast<AddOperator>(Op)->getOperand(1))) {
              // An add (in the same block) with a constant operand. Fold the
              // constant.
              ConstantInt *CI =
              cast<ConstantInt>(cast<AddOperator>(Op)->getOperand(1));
              TmpOffset += CI->getSExtValue() * S;
              // Iterate on the other operand.
              Op = cast<AddOperator>(Op)->getOperand(0);
              continue;
            }
            // Unsupported
            goto unsupported_gep;
          }
        }
      }

      // Try to grab the base operand now.
      Addr.Offset = TmpOffset;
      if (PPCComputeAddress(U->getOperand(0), Addr)) return true;

      // We failed, restore everything and try the other options.
      Addr = SavedAddr;

      unsupported_gep:
      break;
    }
    case Instruction::Alloca: {
      const AllocaInst *AI = cast<AllocaInst>(Obj);
      DenseMap<const AllocaInst*, int>::iterator SI =
        FuncInfo.StaticAllocaMap.find(AI);
      if (SI != FuncInfo.StaticAllocaMap.end()) {
        Addr.BaseType = Address::FrameIndexBase;
        Addr.Base.FI = SI->second;
        return true;
      }
      break;
    }
  }

  // FIXME: References to parameters fall through to the behavior
  // below.  They should be able to reference a frame index since
  // they are stored to the stack, so we can get "ld rx, offset(r1)"
  // instead of "addi ry, r1, offset / ld rx, 0(ry)".  Obj will
  // just contain the parameter.  Try to handle this with a FI.

  // Try to get this in a register if nothing else has worked.
  if (Addr.Base.Reg == 0)
    Addr.Base.Reg = getRegForValue(Obj);

  // Prevent assignment of base register to X0, which is inappropriate
  // for loads and stores alike.
  if (Addr.Base.Reg != 0)
    MRI.setRegClass(Addr.Base.Reg, &PPC::G8RC_and_G8RC_NOX0RegClass);

  return Addr.Base.Reg != 0;
}

// Fix up some addresses that can't be used directly.  For example, if
// an offset won't fit in an instruction field, we may need to move it
// into an index register.
void PPCFastISel::PPCSimplifyAddress(Address &Addr, MVT VT, bool &UseOffset,
                                     unsigned &IndexReg) {

  // Check whether the offset fits in the instruction field.
  if (!isInt<16>(Addr.Offset))
    UseOffset = false;

  // If this is a stack pointer and the offset needs to be simplified then
  // put the alloca address into a register, set the base type back to
  // register and continue. This should almost never happen.
  if (!UseOffset && Addr.BaseType == Address::FrameIndexBase) {
    unsigned ResultReg = createResultReg(&PPC::G8RC_and_G8RC_NOX0RegClass);
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL, TII.get(PPC::ADDI8),
            ResultReg).addFrameIndex(Addr.Base.FI).addImm(0);
    Addr.Base.Reg = ResultReg;
    Addr.BaseType = Address::RegBase;
  }

  if (!UseOffset) {
    IntegerType *OffsetTy = ((VT == MVT::i32) ? Type::getInt32Ty(*Context)
                             : Type::getInt64Ty(*Context));
    const ConstantInt *Offset =
      ConstantInt::getSigned(OffsetTy, (int64_t)(Addr.Offset));
    IndexReg = PPCMaterializeInt(Offset, MVT::i64);
    assert(IndexReg && "Unexpected error in PPCMaterializeInt!");
  }
}

// Emit a load instruction if possible, returning true if we succeeded,
// otherwise false.  See commentary below for how the register class of
// the load is determined. 
bool PPCFastISel::PPCEmitLoad(MVT VT, unsigned &ResultReg, Address &Addr,
                              const TargetRegisterClass *RC,
                              bool IsZExt, unsigned FP64LoadOpc) {
  unsigned Opc;
  bool UseOffset = true;

  // If ResultReg is given, it determines the register class of the load.
  // Otherwise, RC is the register class to use.  If the result of the
  // load isn't anticipated in this block, both may be zero, in which
  // case we must make a conservative guess.  In particular, don't assign
  // R0 or X0 to the result register, as the result may be used in a load,
  // store, add-immediate, or isel that won't permit this.  (Though
  // perhaps the spill and reload of live-exit values would handle this?)
  const TargetRegisterClass *UseRC =
    (ResultReg ? MRI.getRegClass(ResultReg) :
     (RC ? RC :
      (VT == MVT::f64 ? &PPC::F8RCRegClass :
       (VT == MVT::f32 ? &PPC::F4RCRegClass :
        (VT == MVT::i64 ? &PPC::G8RC_and_G8RC_NOX0RegClass :
         &PPC::GPRC_and_GPRC_NOR0RegClass)))));

  bool Is32BitInt = UseRC->hasSuperClassEq(&PPC::GPRCRegClass);

  switch (VT.SimpleTy) {
    default: // e.g., vector types not handled
      return false;
    case MVT::i8:
      Opc = Is32BitInt ? PPC::LBZ : PPC::LBZ8;
      break;
    case MVT::i16:
      Opc = (IsZExt ?
             (Is32BitInt ? PPC::LHZ : PPC::LHZ8) : 
             (Is32BitInt ? PPC::LHA : PPC::LHA8));
      break;
    case MVT::i32:
      Opc = (IsZExt ? 
             (Is32BitInt ? PPC::LWZ : PPC::LWZ8) :
             (Is32BitInt ? PPC::LWA_32 : PPC::LWA));
      if ((Opc == PPC::LWA || Opc == PPC::LWA_32) && ((Addr.Offset & 3) != 0))
        UseOffset = false;
      break;
    case MVT::i64:
      Opc = PPC::LD;
      assert(UseRC->hasSuperClassEq(&PPC::G8RCRegClass) && 
             "64-bit load with 32-bit target??");
      UseOffset = ((Addr.Offset & 3) == 0);
      break;
    case MVT::f32:
      Opc = PPC::LFS;
      break;
    case MVT::f64:
      Opc = FP64LoadOpc;
      break;
  }

  // If necessary, materialize the offset into a register and use
  // the indexed form.  Also handle stack pointers with special needs.
  unsigned IndexReg = 0;
  PPCSimplifyAddress(Addr, VT, UseOffset, IndexReg);
  if (ResultReg == 0)
    ResultReg = createResultReg(UseRC);

  // Note: If we still have a frame index here, we know the offset is
  // in range, as otherwise PPCSimplifyAddress would have converted it
  // into a RegBase.
  if (Addr.BaseType == Address::FrameIndexBase) {

    MachineMemOperand *MMO =
      FuncInfo.MF->getMachineMemOperand(
        MachinePointerInfo::getFixedStack(Addr.Base.FI, Addr.Offset),
        MachineMemOperand::MOLoad, MFI.getObjectSize(Addr.Base.FI),
        MFI.getObjectAlignment(Addr.Base.FI));

    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL, TII.get(Opc), ResultReg)
      .addImm(Addr.Offset).addFrameIndex(Addr.Base.FI).addMemOperand(MMO);

  // Base reg with offset in range.
  } else if (UseOffset) {

    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL, TII.get(Opc), ResultReg)
      .addImm(Addr.Offset).addReg(Addr.Base.Reg);

  // Indexed form.
  } else {
    // Get the RR opcode corresponding to the RI one.  FIXME: It would be
    // preferable to use the ImmToIdxMap from PPCRegisterInfo.cpp, but it
    // is hard to get at.
    switch (Opc) {
      default:        llvm_unreachable("Unexpected opcode!");
      case PPC::LBZ:    Opc = PPC::LBZX;    break;
      case PPC::LBZ8:   Opc = PPC::LBZX8;   break;
      case PPC::LHZ:    Opc = PPC::LHZX;    break;
      case PPC::LHZ8:   Opc = PPC::LHZX8;   break;
      case PPC::LHA:    Opc = PPC::LHAX;    break;
      case PPC::LHA8:   Opc = PPC::LHAX8;   break;
      case PPC::LWZ:    Opc = PPC::LWZX;    break;
      case PPC::LWZ8:   Opc = PPC::LWZX8;   break;
      case PPC::LWA:    Opc = PPC::LWAX;    break;
      case PPC::LWA_32: Opc = PPC::LWAX_32; break;
      case PPC::LD:     Opc = PPC::LDX;     break;
      case PPC::LFS:    Opc = PPC::LFSX;    break;
      case PPC::LFD:    Opc = PPC::LFDX;    break;
    }
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL, TII.get(Opc), ResultReg)
      .addReg(Addr.Base.Reg).addReg(IndexReg);
  }

  return true;
}

// Attempt to fast-select a load instruction.
bool PPCFastISel::SelectLoad(const Instruction *I) {
  // FIXME: No atomic loads are supported.
  if (cast<LoadInst>(I)->isAtomic())
    return false;

  // Verify we have a legal type before going any further.
  MVT VT;
  if (!isLoadTypeLegal(I->getType(), VT))
    return false;

  // See if we can handle this address.
  Address Addr;
  if (!PPCComputeAddress(I->getOperand(0), Addr))
    return false;

  // Look at the currently assigned register for this instruction
  // to determine the required register class.  This is necessary
  // to constrain RA from using R0/X0 when this is not legal.
  unsigned AssignedReg = FuncInfo.ValueMap[I];
  const TargetRegisterClass *RC =
    AssignedReg ? MRI.getRegClass(AssignedReg) : 0;

  unsigned ResultReg = 0;
  if (!PPCEmitLoad(VT, ResultReg, Addr, RC))
    return false;
  UpdateValueMap(I, ResultReg);
  return true;
}

// Emit a store instruction to store SrcReg at Addr.
bool PPCFastISel::PPCEmitStore(MVT VT, unsigned SrcReg, Address &Addr) {
  assert(SrcReg && "Nothing to store!");
  unsigned Opc;
  bool UseOffset = true;

  const TargetRegisterClass *RC = MRI.getRegClass(SrcReg);
  bool Is32BitInt = RC->hasSuperClassEq(&PPC::GPRCRegClass);

  switch (VT.SimpleTy) {
    default: // e.g., vector types not handled
      return false;
    case MVT::i8:
      Opc = Is32BitInt ? PPC::STB : PPC::STB8;
      break;
    case MVT::i16:
      Opc = Is32BitInt ? PPC::STH : PPC::STH8;
      break;
    case MVT::i32:
      assert(Is32BitInt && "Not GPRC for i32??");
      Opc = PPC::STW;
      break;
    case MVT::i64:
      Opc = PPC::STD;
      UseOffset = ((Addr.Offset & 3) == 0);
      break;
    case MVT::f32:
      Opc = PPC::STFS;
      break;
    case MVT::f64:
      Opc = PPC::STFD;
      break;
  }

  // If necessary, materialize the offset into a register and use
  // the indexed form.  Also handle stack pointers with special needs.
  unsigned IndexReg = 0;
  PPCSimplifyAddress(Addr, VT, UseOffset, IndexReg);

  // Note: If we still have a frame index here, we know the offset is
  // in range, as otherwise PPCSimplifyAddress would have converted it
  // into a RegBase.
  if (Addr.BaseType == Address::FrameIndexBase) {
    MachineMemOperand *MMO =
      FuncInfo.MF->getMachineMemOperand(
        MachinePointerInfo::getFixedStack(Addr.Base.FI, Addr.Offset),
        MachineMemOperand::MOStore, MFI.getObjectSize(Addr.Base.FI),
        MFI.getObjectAlignment(Addr.Base.FI));

    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL, TII.get(Opc)).addReg(SrcReg)
      .addImm(Addr.Offset).addFrameIndex(Addr.Base.FI).addMemOperand(MMO);

  // Base reg with offset in range.
  } else if (UseOffset)
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL, TII.get(Opc))
      .addReg(SrcReg).addImm(Addr.Offset).addReg(Addr.Base.Reg);

  // Indexed form.
  else {
    // Get the RR opcode corresponding to the RI one.  FIXME: It would be
    // preferable to use the ImmToIdxMap from PPCRegisterInfo.cpp, but it
    // is hard to get at.
    switch (Opc) {
      default:        llvm_unreachable("Unexpected opcode!");
      case PPC::STB:  Opc = PPC::STBX;  break;
      case PPC::STH : Opc = PPC::STHX;  break;
      case PPC::STW : Opc = PPC::STWX;  break;
      case PPC::STB8: Opc = PPC::STBX8; break;
      case PPC::STH8: Opc = PPC::STHX8; break;
      case PPC::STW8: Opc = PPC::STWX8; break;
      case PPC::STD:  Opc = PPC::STDX;  break;
      case PPC::STFS: Opc = PPC::STFSX; break;
      case PPC::STFD: Opc = PPC::STFDX; break;
    }
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL, TII.get(Opc))
      .addReg(SrcReg).addReg(Addr.Base.Reg).addReg(IndexReg);
  }

  return true;
}

// Attempt to fast-select a store instruction.
bool PPCFastISel::SelectStore(const Instruction *I) {
  Value *Op0 = I->getOperand(0);
  unsigned SrcReg = 0;

  // FIXME: No atomics loads are supported.
  if (cast<StoreInst>(I)->isAtomic())
    return false;

  // Verify we have a legal type before going any further.
  MVT VT;
  if (!isLoadTypeLegal(Op0->getType(), VT))
    return false;

  // Get the value to be stored into a register.
  SrcReg = getRegForValue(Op0);
  if (SrcReg == 0)
    return false;

  // See if we can handle this address.
  Address Addr;
  if (!PPCComputeAddress(I->getOperand(1), Addr))
    return false;

  if (!PPCEmitStore(VT, SrcReg, Addr))
    return false;

  return true;
}

// Attempt to fast-select a branch instruction.
bool PPCFastISel::SelectBranch(const Instruction *I) {
  const BranchInst *BI = cast<BranchInst>(I);
  MachineBasicBlock *BrBB = FuncInfo.MBB;
  MachineBasicBlock *TBB = FuncInfo.MBBMap[BI->getSuccessor(0)];
  MachineBasicBlock *FBB = FuncInfo.MBBMap[BI->getSuccessor(1)];

  // For now, just try the simplest case where it's fed by a compare.
  if (const CmpInst *CI = dyn_cast<CmpInst>(BI->getCondition())) {
    Optional<PPC::Predicate> OptPPCPred = getComparePred(CI->getPredicate());
    if (!OptPPCPred)
      return false;

    PPC::Predicate PPCPred = OptPPCPred.getValue();

    // Take advantage of fall-through opportunities.
    if (FuncInfo.MBB->isLayoutSuccessor(TBB)) {
      std::swap(TBB, FBB);
      PPCPred = PPC::InvertPredicate(PPCPred);
    }

    unsigned CondReg = createResultReg(&PPC::CRRCRegClass);

    if (!PPCEmitCmp(CI->getOperand(0), CI->getOperand(1), CI->isUnsigned(),
                    CondReg))
      return false;

    BuildMI(*BrBB, FuncInfo.InsertPt, DL, TII.get(PPC::BCC))
      .addImm(PPCPred).addReg(CondReg).addMBB(TBB);
    FastEmitBranch(FBB, DL);
    FuncInfo.MBB->addSuccessor(TBB);
    return true;

  } else if (const ConstantInt *CI =
             dyn_cast<ConstantInt>(BI->getCondition())) {
    uint64_t Imm = CI->getZExtValue();
    MachineBasicBlock *Target = (Imm == 0) ? FBB : TBB;
    FastEmitBranch(Target, DL);
    return true;
  }

  // FIXME: ARM looks for a case where the block containing the compare
  // has been split from the block containing the branch.  If this happens,
  // there is a vreg available containing the result of the compare.  I'm
  // not sure we can do much, as we've lost the predicate information with
  // the compare instruction -- we have a 4-bit CR but don't know which bit
  // to test here.
  return false;
}

// Attempt to emit a compare of the two source values.  Signed and unsigned
// comparisons are supported.  Return false if we can't handle it.
bool PPCFastISel::PPCEmitCmp(const Value *SrcValue1, const Value *SrcValue2,
                             bool IsZExt, unsigned DestReg) {
  Type *Ty = SrcValue1->getType();
  EVT SrcEVT = TLI.getValueType(Ty, true);
  if (!SrcEVT.isSimple())
    return false;
  MVT SrcVT = SrcEVT.getSimpleVT();

  // See if operand 2 is an immediate encodeable in the compare.
  // FIXME: Operands are not in canonical order at -O0, so an immediate
  // operand in position 1 is a lost opportunity for now.  We are
  // similar to ARM in this regard.
  long Imm = 0;
  bool UseImm = false;

  // Only 16-bit integer constants can be represented in compares for 
  // PowerPC.  Others will be materialized into a register.
  if (const ConstantInt *ConstInt = dyn_cast<ConstantInt>(SrcValue2)) {
    if (SrcVT == MVT::i64 || SrcVT == MVT::i32 || SrcVT == MVT::i16 ||
        SrcVT == MVT::i8 || SrcVT == MVT::i1) {
      const APInt &CIVal = ConstInt->getValue();
      Imm = (IsZExt) ? (long)CIVal.getZExtValue() : (long)CIVal.getSExtValue();
      if ((IsZExt && isUInt<16>(Imm)) || (!IsZExt && isInt<16>(Imm)))
        UseImm = true;
    }
  }

  unsigned CmpOpc;
  bool NeedsExt = false;
  switch (SrcVT.SimpleTy) {
    default: return false;
    case MVT::f32:
      CmpOpc = PPC::FCMPUS;
      break;
    case MVT::f64:
      CmpOpc = PPC::FCMPUD;
      break;
    case MVT::i1:
    case MVT::i8:
    case MVT::i16:
      NeedsExt = true;
      // Intentional fall-through.
    case MVT::i32:
      if (!UseImm)
        CmpOpc = IsZExt ? PPC::CMPLW : PPC::CMPW;
      else
        CmpOpc = IsZExt ? PPC::CMPLWI : PPC::CMPWI;
      break;
    case MVT::i64:
      if (!UseImm)
        CmpOpc = IsZExt ? PPC::CMPLD : PPC::CMPD;
      else
        CmpOpc = IsZExt ? PPC::CMPLDI : PPC::CMPDI;
      break;
  }

  unsigned SrcReg1 = getRegForValue(SrcValue1);
  if (SrcReg1 == 0)
    return false;

  unsigned SrcReg2 = 0;
  if (!UseImm) {
    SrcReg2 = getRegForValue(SrcValue2);
    if (SrcReg2 == 0)
      return false;
  }

  if (NeedsExt) {
    unsigned ExtReg = createResultReg(&PPC::GPRCRegClass);
    if (!PPCEmitIntExt(SrcVT, SrcReg1, MVT::i32, ExtReg, IsZExt))
      return false;
    SrcReg1 = ExtReg;

    if (!UseImm) {
      unsigned ExtReg = createResultReg(&PPC::GPRCRegClass);
      if (!PPCEmitIntExt(SrcVT, SrcReg2, MVT::i32, ExtReg, IsZExt))
        return false;
      SrcReg2 = ExtReg;
    }
  }

  if (!UseImm)
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL, TII.get(CmpOpc), DestReg)
      .addReg(SrcReg1).addReg(SrcReg2);
  else
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL, TII.get(CmpOpc), DestReg)
      .addReg(SrcReg1).addImm(Imm);

  return true;
}

// Attempt to fast-select a floating-point extend instruction.
bool PPCFastISel::SelectFPExt(const Instruction *I) {
  Value *Src  = I->getOperand(0);
  EVT SrcVT  = TLI.getValueType(Src->getType(), true);
  EVT DestVT = TLI.getValueType(I->getType(), true);

  if (SrcVT != MVT::f32 || DestVT != MVT::f64)
    return false;

  unsigned SrcReg = getRegForValue(Src);
  if (!SrcReg)
    return false;

  // No code is generated for a FP extend.
  UpdateValueMap(I, SrcReg);
  return true;
}

// Attempt to fast-select a floating-point truncate instruction.
bool PPCFastISel::SelectFPTrunc(const Instruction *I) {
  Value *Src  = I->getOperand(0);
  EVT SrcVT  = TLI.getValueType(Src->getType(), true);
  EVT DestVT = TLI.getValueType(I->getType(), true);

  if (SrcVT != MVT::f64 || DestVT != MVT::f32)
    return false;

  unsigned SrcReg = getRegForValue(Src);
  if (!SrcReg)
    return false;

  // Round the result to single precision.
  unsigned DestReg = createResultReg(&PPC::F4RCRegClass);
  BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL, TII.get(PPC::FRSP), DestReg)
    .addReg(SrcReg);

  UpdateValueMap(I, DestReg);
  return true;
}

// Move an i32 or i64 value in a GPR to an f64 value in an FPR.
// FIXME: When direct register moves are implemented (see PowerISA 2.08),
// those should be used instead of moving via a stack slot when the
// subtarget permits.
// FIXME: The code here is sloppy for the 4-byte case.  Can use a 4-byte
// stack slot and 4-byte store/load sequence.  Or just sext the 4-byte
// case to 8 bytes which produces tighter code but wastes stack space.
unsigned PPCFastISel::PPCMoveToFPReg(MVT SrcVT, unsigned SrcReg,
                                     bool IsSigned) {

  // If necessary, extend 32-bit int to 64-bit.
  if (SrcVT == MVT::i32) {
    unsigned TmpReg = createResultReg(&PPC::G8RCRegClass);
    if (!PPCEmitIntExt(MVT::i32, SrcReg, MVT::i64, TmpReg, !IsSigned))
      return 0;
    SrcReg = TmpReg;
  }

  // Get a stack slot 8 bytes wide, aligned on an 8-byte boundary.
  Address Addr;
  Addr.BaseType = Address::FrameIndexBase;
  Addr.Base.FI = MFI.CreateStackObject(8, 8, false);

  // Store the value from the GPR.
  if (!PPCEmitStore(MVT::i64, SrcReg, Addr))
    return 0;

  // Load the integer value into an FPR.  The kind of load used depends
  // on a number of conditions.
  unsigned LoadOpc = PPC::LFD;

  if (SrcVT == MVT::i32) {
    Addr.Offset = 4;
    if (!IsSigned)
      LoadOpc = PPC::LFIWZX;
    else if (PPCSubTarget.hasLFIWAX())
      LoadOpc = PPC::LFIWAX;
  }

  const TargetRegisterClass *RC = &PPC::F8RCRegClass;
  unsigned ResultReg = 0;
  if (!PPCEmitLoad(MVT::f64, ResultReg, Addr, RC, !IsSigned, LoadOpc))
    return 0;

  return ResultReg;
}

// Attempt to fast-select an integer-to-floating-point conversion.
bool PPCFastISel::SelectIToFP(const Instruction *I, bool IsSigned) {
  MVT DstVT;
  Type *DstTy = I->getType();
  if (!isTypeLegal(DstTy, DstVT))
    return false;

  if (DstVT != MVT::f32 && DstVT != MVT::f64)
    return false;

  Value *Src = I->getOperand(0);
  EVT SrcEVT = TLI.getValueType(Src->getType(), true);
  if (!SrcEVT.isSimple())
    return false;

  MVT SrcVT = SrcEVT.getSimpleVT();

  if (SrcVT != MVT::i8  && SrcVT != MVT::i16 &&
      SrcVT != MVT::i32 && SrcVT != MVT::i64)
    return false;

  unsigned SrcReg = getRegForValue(Src);
  if (SrcReg == 0)
    return false;

  // We can only lower an unsigned convert if we have the newer
  // floating-point conversion operations.
  if (!IsSigned && !PPCSubTarget.hasFPCVT())
    return false;

  // FIXME: For now we require the newer floating-point conversion operations
  // (which are present only on P7 and A2 server models) when converting
  // to single-precision float.  Otherwise we have to generate a lot of
  // fiddly code to avoid double rounding.  If necessary, the fiddly code
  // can be found in PPCTargetLowering::LowerINT_TO_FP().
  if (DstVT == MVT::f32 && !PPCSubTarget.hasFPCVT())
    return false;

  // Extend the input if necessary.
  if (SrcVT == MVT::i8 || SrcVT == MVT::i16) {
    unsigned TmpReg = createResultReg(&PPC::G8RCRegClass);
    if (!PPCEmitIntExt(SrcVT, SrcReg, MVT::i64, TmpReg, !IsSigned))
      return false;
    SrcVT = MVT::i64;
    SrcReg = TmpReg;
  }

  // Move the integer value to an FPR.
  unsigned FPReg = PPCMoveToFPReg(SrcVT, SrcReg, IsSigned);
  if (FPReg == 0)
    return false;

  // Determine the opcode for the conversion.
  const TargetRegisterClass *RC = &PPC::F8RCRegClass;
  unsigned DestReg = createResultReg(RC);
  unsigned Opc;

  if (DstVT == MVT::f32)
    Opc = IsSigned ? PPC::FCFIDS : PPC::FCFIDUS;
  else
    Opc = IsSigned ? PPC::FCFID : PPC::FCFIDU;

  // Generate the convert.
  BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL, TII.get(Opc), DestReg)
    .addReg(FPReg);

  UpdateValueMap(I, DestReg);
  return true;
}

// Move the floating-point value in SrcReg into an integer destination
// register, and return the register (or zero if we can't handle it).
// FIXME: When direct register moves are implemented (see PowerISA 2.08),
// those should be used instead of moving via a stack slot when the
// subtarget permits.
unsigned PPCFastISel::PPCMoveToIntReg(const Instruction *I, MVT VT,
                                      unsigned SrcReg, bool IsSigned) {
  // Get a stack slot 8 bytes wide, aligned on an 8-byte boundary.
  // Note that if have STFIWX available, we could use a 4-byte stack
  // slot for i32, but this being fast-isel we'll just go with the
  // easiest code gen possible.
  Address Addr;
  Addr.BaseType = Address::FrameIndexBase;
  Addr.Base.FI = MFI.CreateStackObject(8, 8, false);

  // Store the value from the FPR.
  if (!PPCEmitStore(MVT::f64, SrcReg, Addr))
    return 0;

  // Reload it into a GPR.  If we want an i32, modify the address
  // to have a 4-byte offset so we load from the right place.
  if (VT == MVT::i32)
    Addr.Offset = 4;

  // Look at the currently assigned register for this instruction
  // to determine the required register class.
  unsigned AssignedReg = FuncInfo.ValueMap[I];
  const TargetRegisterClass *RC =
    AssignedReg ? MRI.getRegClass(AssignedReg) : 0;

  unsigned ResultReg = 0;
  if (!PPCEmitLoad(VT, ResultReg, Addr, RC, !IsSigned))
    return 0;

  return ResultReg;
}

// Attempt to fast-select a floating-point-to-integer conversion.
bool PPCFastISel::SelectFPToI(const Instruction *I, bool IsSigned) {
  MVT DstVT, SrcVT;
  Type *DstTy = I->getType();
  if (!isTypeLegal(DstTy, DstVT))
    return false;

  if (DstVT != MVT::i32 && DstVT != MVT::i64)
    return false;

  Value *Src = I->getOperand(0);
  Type *SrcTy = Src->getType();
  if (!isTypeLegal(SrcTy, SrcVT))
    return false;

  if (SrcVT != MVT::f32 && SrcVT != MVT::f64)
    return false;

  unsigned SrcReg = getRegForValue(Src);
  if (SrcReg == 0)
    return false;

  // Convert f32 to f64 if necessary.  This is just a meaningless copy
  // to get the register class right.  COPY_TO_REGCLASS is needed since
  // a COPY from F4RC to F8RC is converted to a F4RC-F4RC copy downstream.
  const TargetRegisterClass *InRC = MRI.getRegClass(SrcReg);
  if (InRC == &PPC::F4RCRegClass) {
    unsigned TmpReg = createResultReg(&PPC::F8RCRegClass);
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL,
            TII.get(TargetOpcode::COPY_TO_REGCLASS), TmpReg)
      .addReg(SrcReg).addImm(PPC::F8RCRegClassID);
    SrcReg = TmpReg;
  }

  // Determine the opcode for the conversion, which takes place
  // entirely within FPRs.
  unsigned DestReg = createResultReg(&PPC::F8RCRegClass);
  unsigned Opc;

  if (DstVT == MVT::i32)
    if (IsSigned)
      Opc = PPC::FCTIWZ;
    else
      Opc = PPCSubTarget.hasFPCVT() ? PPC::FCTIWUZ : PPC::FCTIDZ;
  else
    Opc = IsSigned ? PPC::FCTIDZ : PPC::FCTIDUZ;

  // Generate the convert.
  BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL, TII.get(Opc), DestReg)
    .addReg(SrcReg);

  // Now move the integer value from a float register to an integer register.
  unsigned IntReg = PPCMoveToIntReg(I, DstVT, DestReg, IsSigned);
  if (IntReg == 0)
    return false;

  UpdateValueMap(I, IntReg);
  return true;
}

// Attempt to fast-select a binary integer operation that isn't already
// handled automatically.
bool PPCFastISel::SelectBinaryIntOp(const Instruction *I, unsigned ISDOpcode) {
  EVT DestVT  = TLI.getValueType(I->getType(), true);

  // We can get here in the case when we have a binary operation on a non-legal
  // type and the target independent selector doesn't know how to handle it.
  if (DestVT != MVT::i16 && DestVT != MVT::i8)
    return false;

  // Look at the currently assigned register for this instruction
  // to determine the required register class.  If there is no register,
  // make a conservative choice (don't assign R0).
  unsigned AssignedReg = FuncInfo.ValueMap[I];
  const TargetRegisterClass *RC =
    (AssignedReg ? MRI.getRegClass(AssignedReg) :
     &PPC::GPRC_and_GPRC_NOR0RegClass);
  bool IsGPRC = RC->hasSuperClassEq(&PPC::GPRCRegClass);

  unsigned Opc;
  switch (ISDOpcode) {
    default: return false;
    case ISD::ADD:
      Opc = IsGPRC ? PPC::ADD4 : PPC::ADD8;
      break;
    case ISD::OR:
      Opc = IsGPRC ? PPC::OR : PPC::OR8;
      break;
    case ISD::SUB:
      Opc = IsGPRC ? PPC::SUBF : PPC::SUBF8;
      break;
  }

  unsigned ResultReg = createResultReg(RC ? RC : &PPC::G8RCRegClass);
  unsigned SrcReg1 = getRegForValue(I->getOperand(0));
  if (SrcReg1 == 0) return false;

  // Handle case of small immediate operand.
  if (const ConstantInt *ConstInt = dyn_cast<ConstantInt>(I->getOperand(1))) {
    const APInt &CIVal = ConstInt->getValue();
    int Imm = (int)CIVal.getSExtValue();
    bool UseImm = true;
    if (isInt<16>(Imm)) {
      switch (Opc) {
        default:
          llvm_unreachable("Missing case!");
        case PPC::ADD4:
          Opc = PPC::ADDI;
          MRI.setRegClass(SrcReg1, &PPC::GPRC_and_GPRC_NOR0RegClass);
          break;
        case PPC::ADD8:
          Opc = PPC::ADDI8;
          MRI.setRegClass(SrcReg1, &PPC::G8RC_and_G8RC_NOX0RegClass);
          break;
        case PPC::OR:
          Opc = PPC::ORI;
          break;
        case PPC::OR8:
          Opc = PPC::ORI8;
          break;
        case PPC::SUBF:
          if (Imm == -32768)
            UseImm = false;
          else {
            Opc = PPC::ADDI;
            MRI.setRegClass(SrcReg1, &PPC::GPRC_and_GPRC_NOR0RegClass);
            Imm = -Imm;
          }
          break;
        case PPC::SUBF8:
          if (Imm == -32768)
            UseImm = false;
          else {
            Opc = PPC::ADDI8;
            MRI.setRegClass(SrcReg1, &PPC::G8RC_and_G8RC_NOX0RegClass);
            Imm = -Imm;
          }
          break;
      }

      if (UseImm) {
        BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL, TII.get(Opc), ResultReg)
          .addReg(SrcReg1).addImm(Imm);
        UpdateValueMap(I, ResultReg);
        return true;
      }
    }
  }

  // Reg-reg case.
  unsigned SrcReg2 = getRegForValue(I->getOperand(1));
  if (SrcReg2 == 0) return false;

  // Reverse operands for subtract-from.
  if (ISDOpcode == ISD::SUB)
    std::swap(SrcReg1, SrcReg2);

  BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL, TII.get(Opc), ResultReg)
    .addReg(SrcReg1).addReg(SrcReg2);
  UpdateValueMap(I, ResultReg);
  return true;
}

// Handle arguments to a call that we're attempting to fast-select.
// Return false if the arguments are too complex for us at the moment.
bool PPCFastISel::processCallArgs(SmallVectorImpl<Value*> &Args,
                                  SmallVectorImpl<unsigned> &ArgRegs,
                                  SmallVectorImpl<MVT> &ArgVTs,
                                  SmallVectorImpl<ISD::ArgFlagsTy> &ArgFlags,
                                  SmallVectorImpl<unsigned> &RegArgs,
                                  CallingConv::ID CC,
                                  unsigned &NumBytes,
                                  bool IsVarArg) {
  SmallVector<CCValAssign, 16> ArgLocs;
  CCState CCInfo(CC, IsVarArg, *FuncInfo.MF, TM, ArgLocs, *Context);
  CCInfo.AnalyzeCallOperands(ArgVTs, ArgFlags, CC_PPC64_ELF_FIS);

  // Bail out if we can't handle any of the arguments.
  for (unsigned I = 0, E = ArgLocs.size(); I != E; ++I) {
    CCValAssign &VA = ArgLocs[I];
    MVT ArgVT = ArgVTs[VA.getValNo()];

    // Skip vector arguments for now, as well as long double and
    // uint128_t, and anything that isn't passed in a register.
    if (ArgVT.isVector() || ArgVT.getSizeInBits() > 64 ||
        !VA.isRegLoc() || VA.needsCustom())
      return false;

    // Skip bit-converted arguments for now.
    if (VA.getLocInfo() == CCValAssign::BCvt)
      return false;
  }

  // Get a count of how many bytes are to be pushed onto the stack.
  NumBytes = CCInfo.getNextStackOffset();

  // Issue CALLSEQ_START.
  BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL,
          TII.get(TII.getCallFrameSetupOpcode()))
    .addImm(NumBytes);

  // Prepare to assign register arguments.  Every argument uses up a
  // GPR protocol register even if it's passed in a floating-point
  // register.
  unsigned NextGPR = PPC::X3;
  unsigned NextFPR = PPC::F1;

  // Process arguments.
  for (unsigned I = 0, E = ArgLocs.size(); I != E; ++I) {
    CCValAssign &VA = ArgLocs[I];
    unsigned Arg = ArgRegs[VA.getValNo()];
    MVT ArgVT = ArgVTs[VA.getValNo()];

    // Handle argument promotion and bitcasts.
    switch (VA.getLocInfo()) {
      default:
        llvm_unreachable("Unknown loc info!");
      case CCValAssign::Full:
        break;
      case CCValAssign::SExt: {
        MVT DestVT = VA.getLocVT();
        const TargetRegisterClass *RC =
          (DestVT == MVT::i64) ? &PPC::G8RCRegClass : &PPC::GPRCRegClass;
        unsigned TmpReg = createResultReg(RC);
        if (!PPCEmitIntExt(ArgVT, Arg, DestVT, TmpReg, /*IsZExt*/false))
          llvm_unreachable("Failed to emit a sext!");
        ArgVT = DestVT;
        Arg = TmpReg;
        break;
      }
      case CCValAssign::AExt:
      case CCValAssign::ZExt: {
        MVT DestVT = VA.getLocVT();
        const TargetRegisterClass *RC =
          (DestVT == MVT::i64) ? &PPC::G8RCRegClass : &PPC::GPRCRegClass;
        unsigned TmpReg = createResultReg(RC);
        if (!PPCEmitIntExt(ArgVT, Arg, DestVT, TmpReg, /*IsZExt*/true))
          llvm_unreachable("Failed to emit a zext!");
        ArgVT = DestVT;
        Arg = TmpReg;
        break;
      }
      case CCValAssign::BCvt: {
        // FIXME: Not yet handled.
        llvm_unreachable("Should have bailed before getting here!");
        break;
      }
    }

    // Copy this argument to the appropriate register.
    unsigned ArgReg;
    if (ArgVT == MVT::f32 || ArgVT == MVT::f64) {
      ArgReg = NextFPR++;
      ++NextGPR;
    } else
      ArgReg = NextGPR++;
      
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL, TII.get(TargetOpcode::COPY),
            ArgReg).addReg(Arg);
    RegArgs.push_back(ArgReg);
  }

  return true;
}

// For a call that we've determined we can fast-select, finish the
// call sequence and generate a copy to obtain the return value (if any).
void PPCFastISel::finishCall(MVT RetVT, SmallVectorImpl<unsigned> &UsedRegs,
                             const Instruction *I, CallingConv::ID CC,
                             unsigned &NumBytes, bool IsVarArg) {
  // Issue CallSEQ_END.
  BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL,
          TII.get(TII.getCallFrameDestroyOpcode()))
    .addImm(NumBytes).addImm(0);

  // Next, generate a copy to obtain the return value.
  // FIXME: No multi-register return values yet, though I don't foresee
  // any real difficulties there.
  if (RetVT != MVT::isVoid) {
    SmallVector<CCValAssign, 16> RVLocs;
    CCState CCInfo(CC, IsVarArg, *FuncInfo.MF, TM, RVLocs, *Context);
    CCInfo.AnalyzeCallResult(RetVT, RetCC_PPC64_ELF_FIS);
    CCValAssign &VA = RVLocs[0];
    assert(RVLocs.size() == 1 && "No support for multi-reg return values!");
    assert(VA.isRegLoc() && "Can only return in registers!");

    MVT DestVT = VA.getValVT();
    MVT CopyVT = DestVT;

    // Ints smaller than a register still arrive in a full 64-bit
    // register, so make sure we recognize this.
    if (RetVT == MVT::i8 || RetVT == MVT::i16 || RetVT == MVT::i32)
      CopyVT = MVT::i64;

    unsigned SourcePhysReg = VA.getLocReg();
    unsigned ResultReg = 0;

    if (RetVT == CopyVT) {
      const TargetRegisterClass *CpyRC = TLI.getRegClassFor(CopyVT);
      ResultReg = createResultReg(CpyRC);

      BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL,
              TII.get(TargetOpcode::COPY), ResultReg)
        .addReg(SourcePhysReg);

    // If necessary, round the floating result to single precision.
    } else if (CopyVT == MVT::f64) {
      ResultReg = createResultReg(TLI.getRegClassFor(RetVT));
      BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL, TII.get(PPC::FRSP),
              ResultReg).addReg(SourcePhysReg);

    // If only the low half of a general register is needed, generate
    // a GPRC copy instead of a G8RC copy.  (EXTRACT_SUBREG can't be
    // used along the fast-isel path (not lowered), and downstream logic
    // also doesn't like a direct subreg copy on a physical reg.)
    } else if (RetVT == MVT::i8 || RetVT == MVT::i16 || RetVT == MVT::i32) {
      ResultReg = createResultReg(&PPC::GPRCRegClass);
      // Convert physical register from G8RC to GPRC.
      SourcePhysReg -= PPC::X0 - PPC::R0;
      BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL,
              TII.get(TargetOpcode::COPY), ResultReg)
        .addReg(SourcePhysReg);
    }

    assert(ResultReg && "ResultReg unset!");
    UsedRegs.push_back(SourcePhysReg);
    UpdateValueMap(I, ResultReg);
  }
}

// Attempt to fast-select a call instruction.
bool PPCFastISel::SelectCall(const Instruction *I) {
  const CallInst *CI = cast<CallInst>(I);
  const Value *Callee = CI->getCalledValue();

  // Can't handle inline asm.
  if (isa<InlineAsm>(Callee))
    return false;

  // Allow SelectionDAG isel to handle tail calls.
  if (CI->isTailCall())
    return false;

  // Obtain calling convention.
  ImmutableCallSite CS(CI);
  CallingConv::ID CC = CS.getCallingConv();

  PointerType *PT = cast<PointerType>(CS.getCalledValue()->getType());
  FunctionType *FTy = cast<FunctionType>(PT->getElementType());
  bool IsVarArg = FTy->isVarArg();

  // Not ready for varargs yet.
  if (IsVarArg)
    return false;

  // Handle simple calls for now, with legal return types and
  // those that can be extended.
  Type *RetTy = I->getType();
  MVT RetVT;
  if (RetTy->isVoidTy())
    RetVT = MVT::isVoid;
  else if (!isTypeLegal(RetTy, RetVT) && RetVT != MVT::i16 &&
           RetVT != MVT::i8)
    return false;

  // FIXME: No multi-register return values yet.
  if (RetVT != MVT::isVoid && RetVT != MVT::i8 && RetVT != MVT::i16 &&
      RetVT != MVT::i32 && RetVT != MVT::i64 && RetVT != MVT::f32 &&
      RetVT != MVT::f64) {
    SmallVector<CCValAssign, 16> RVLocs;
    CCState CCInfo(CC, IsVarArg, *FuncInfo.MF, TM, RVLocs, *Context);
    CCInfo.AnalyzeCallResult(RetVT, RetCC_PPC64_ELF_FIS);
    if (RVLocs.size() > 1)
      return false;
  }

  // Bail early if more than 8 arguments, as we only currently
  // handle arguments passed in registers.
  unsigned NumArgs = CS.arg_size();
  if (NumArgs > 8)
    return false;

  // Set up the argument vectors.
  SmallVector<Value*, 8> Args;
  SmallVector<unsigned, 8> ArgRegs;
  SmallVector<MVT, 8> ArgVTs;
  SmallVector<ISD::ArgFlagsTy, 8> ArgFlags;

  Args.reserve(NumArgs);
  ArgRegs.reserve(NumArgs);
  ArgVTs.reserve(NumArgs);
  ArgFlags.reserve(NumArgs);

  for (ImmutableCallSite::arg_iterator II = CS.arg_begin(), IE = CS.arg_end();
       II != IE; ++II) {
    // FIXME: ARM does something for intrinsic calls here, check into that.

    unsigned AttrIdx = II - CS.arg_begin() + 1;
    
    // Only handle easy calls for now.  It would be reasonably easy
    // to handle <= 8-byte structures passed ByVal in registers, but we
    // have to ensure they are right-justified in the register.
    if (CS.paramHasAttr(AttrIdx, Attribute::InReg) ||
        CS.paramHasAttr(AttrIdx, Attribute::StructRet) ||
        CS.paramHasAttr(AttrIdx, Attribute::Nest) ||
        CS.paramHasAttr(AttrIdx, Attribute::ByVal))
      return false;

    ISD::ArgFlagsTy Flags;
    if (CS.paramHasAttr(AttrIdx, Attribute::SExt))
      Flags.setSExt();
    if (CS.paramHasAttr(AttrIdx, Attribute::ZExt))
      Flags.setZExt();

    Type *ArgTy = (*II)->getType();
    MVT ArgVT;
    if (!isTypeLegal(ArgTy, ArgVT) && ArgVT != MVT::i16 && ArgVT != MVT::i8)
      return false;

    if (ArgVT.isVector())
      return false;

    unsigned Arg = getRegForValue(*II);
    if (Arg == 0)
      return false;

    unsigned OriginalAlignment = TD.getABITypeAlignment(ArgTy);
    Flags.setOrigAlign(OriginalAlignment);

    Args.push_back(*II);
    ArgRegs.push_back(Arg);
    ArgVTs.push_back(ArgVT);
    ArgFlags.push_back(Flags);
  }

  // Process the arguments.
  SmallVector<unsigned, 8> RegArgs;
  unsigned NumBytes;

  if (!processCallArgs(Args, ArgRegs, ArgVTs, ArgFlags,
                       RegArgs, CC, NumBytes, IsVarArg))
    return false;

  // FIXME: No handling for function pointers yet.  This requires
  // implementing the function descriptor (OPD) setup.
  const GlobalValue *GV = dyn_cast<GlobalValue>(Callee);
  if (!GV)
    return false;

  // Build direct call with NOP for TOC restore.
  // FIXME: We can and should optimize away the NOP for local calls.
  MachineInstrBuilder MIB = BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL,
                                    TII.get(PPC::BL8_NOP));
  // Add callee.
  MIB.addGlobalAddress(GV);

  // Add implicit physical register uses to the call.
  for (unsigned II = 0, IE = RegArgs.size(); II != IE; ++II)
    MIB.addReg(RegArgs[II], RegState::Implicit);

  // Add a register mask with the call-preserved registers.  Proper
  // defs for return values will be added by setPhysRegsDeadExcept().
  MIB.addRegMask(TRI.getCallPreservedMask(CC));

  // Finish off the call including any return values.
  SmallVector<unsigned, 4> UsedRegs;
  finishCall(RetVT, UsedRegs, I, CC, NumBytes, IsVarArg);

  // Set all unused physregs defs as dead.
  static_cast<MachineInstr *>(MIB)->setPhysRegsDeadExcept(UsedRegs, TRI);

  return true;
}

// Attempt to fast-select a return instruction.
bool PPCFastISel::SelectRet(const Instruction *I) {

  if (!FuncInfo.CanLowerReturn)
    return false;

  const ReturnInst *Ret = cast<ReturnInst>(I);
  const Function &F = *I->getParent()->getParent();

  // Build a list of return value registers.
  SmallVector<unsigned, 4> RetRegs;
  CallingConv::ID CC = F.getCallingConv();

  if (Ret->getNumOperands() > 0) {
    SmallVector<ISD::OutputArg, 4> Outs;
    GetReturnInfo(F.getReturnType(), F.getAttributes(), Outs, TLI);

    // Analyze operands of the call, assigning locations to each operand.
    SmallVector<CCValAssign, 16> ValLocs;
    CCState CCInfo(CC, F.isVarArg(), *FuncInfo.MF, TM, ValLocs, *Context);
    CCInfo.AnalyzeReturn(Outs, RetCC_PPC64_ELF_FIS);
    const Value *RV = Ret->getOperand(0);
    
    // FIXME: Only one output register for now.
    if (ValLocs.size() > 1)
      return false;

    // Special case for returning a constant integer of any size.
    // Materialize the constant as an i64 and copy it to the return
    // register.  This avoids an unnecessary extend or truncate.
    if (isa<ConstantInt>(*RV)) {
      const Constant *C = cast<Constant>(RV);
      unsigned SrcReg = PPCMaterializeInt(C, MVT::i64);
      unsigned RetReg = ValLocs[0].getLocReg();
      BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL, TII.get(TargetOpcode::COPY),
              RetReg).addReg(SrcReg);
      RetRegs.push_back(RetReg);

    } else {
      unsigned Reg = getRegForValue(RV);

      if (Reg == 0)
        return false;

      // Copy the result values into the output registers.
      for (unsigned i = 0; i < ValLocs.size(); ++i) {

        CCValAssign &VA = ValLocs[i];
        assert(VA.isRegLoc() && "Can only return in registers!");
        RetRegs.push_back(VA.getLocReg());
        unsigned SrcReg = Reg + VA.getValNo();

        EVT RVEVT = TLI.getValueType(RV->getType());
        if (!RVEVT.isSimple())
          return false;
        MVT RVVT = RVEVT.getSimpleVT();
        MVT DestVT = VA.getLocVT();

        if (RVVT != DestVT && RVVT != MVT::i8 &&
            RVVT != MVT::i16 && RVVT != MVT::i32)
          return false;
      
        if (RVVT != DestVT) {
          switch (VA.getLocInfo()) {
            default:
              llvm_unreachable("Unknown loc info!");
            case CCValAssign::Full:
              llvm_unreachable("Full value assign but types don't match?");
            case CCValAssign::AExt:
            case CCValAssign::ZExt: {
              const TargetRegisterClass *RC =
                (DestVT == MVT::i64) ? &PPC::G8RCRegClass : &PPC::GPRCRegClass;
              unsigned TmpReg = createResultReg(RC);
              if (!PPCEmitIntExt(RVVT, SrcReg, DestVT, TmpReg, true))
                return false;
              SrcReg = TmpReg;
              break;
            }
            case CCValAssign::SExt: {
              const TargetRegisterClass *RC =
                (DestVT == MVT::i64) ? &PPC::G8RCRegClass : &PPC::GPRCRegClass;
              unsigned TmpReg = createResultReg(RC);
              if (!PPCEmitIntExt(RVVT, SrcReg, DestVT, TmpReg, false))
                return false;
              SrcReg = TmpReg;
              break;
            }
          }
        }

        BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL,
                TII.get(TargetOpcode::COPY), RetRegs[i])
          .addReg(SrcReg);
      }
    }
  }

  MachineInstrBuilder MIB = BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL,
                                    TII.get(PPC::BLR));

  for (unsigned i = 0, e = RetRegs.size(); i != e; ++i)
    MIB.addReg(RetRegs[i], RegState::Implicit);

  return true;
}

// Attempt to emit an integer extend of SrcReg into DestReg.  Both
// signed and zero extensions are supported.  Return false if we
// can't handle it.
bool PPCFastISel::PPCEmitIntExt(MVT SrcVT, unsigned SrcReg, MVT DestVT,
                                unsigned DestReg, bool IsZExt) {
  if (DestVT != MVT::i32 && DestVT != MVT::i64)
    return false;
  if (SrcVT != MVT::i8 && SrcVT != MVT::i16 && SrcVT != MVT::i32)
    return false;

  // Signed extensions use EXTSB, EXTSH, EXTSW.
  if (!IsZExt) {
    unsigned Opc;
    if (SrcVT == MVT::i8)
      Opc = (DestVT == MVT::i32) ? PPC::EXTSB : PPC::EXTSB8_32_64;
    else if (SrcVT == MVT::i16)
      Opc = (DestVT == MVT::i32) ? PPC::EXTSH : PPC::EXTSH8_32_64;
    else {
      assert(DestVT == MVT::i64 && "Signed extend from i32 to i32??");
      Opc = PPC::EXTSW_32_64;
    }
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL, TII.get(Opc), DestReg)
      .addReg(SrcReg);

  // Unsigned 32-bit extensions use RLWINM.
  } else if (DestVT == MVT::i32) {
    unsigned MB;
    if (SrcVT == MVT::i8)
      MB = 24;
    else {
      assert(SrcVT == MVT::i16 && "Unsigned extend from i32 to i32??");
      MB = 16;
    }
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL, TII.get(PPC::RLWINM),
            DestReg)
      .addReg(SrcReg).addImm(/*SH=*/0).addImm(MB).addImm(/*ME=*/31);

  // Unsigned 64-bit extensions use RLDICL (with a 32-bit source).
  } else {
    unsigned MB;
    if (SrcVT == MVT::i8)
      MB = 56;
    else if (SrcVT == MVT::i16)
      MB = 48;
    else
      MB = 32;
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL,
            TII.get(PPC::RLDICL_32_64), DestReg)
      .addReg(SrcReg).addImm(/*SH=*/0).addImm(MB);
  }

  return true;
}

// Attempt to fast-select an indirect branch instruction.
bool PPCFastISel::SelectIndirectBr(const Instruction *I) {
  unsigned AddrReg = getRegForValue(I->getOperand(0));
  if (AddrReg == 0)
    return false;

  BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL, TII.get(PPC::MTCTR8))
    .addReg(AddrReg);
  BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL, TII.get(PPC::BCTR8));

  const IndirectBrInst *IB = cast<IndirectBrInst>(I);
  for (unsigned i = 0, e = IB->getNumSuccessors(); i != e; ++i)
    FuncInfo.MBB->addSuccessor(FuncInfo.MBBMap[IB->getSuccessor(i)]);

  return true;
}

// Attempt to fast-select an integer truncate instruction.
bool PPCFastISel::SelectTrunc(const Instruction *I) {
  Value *Src  = I->getOperand(0);
  EVT SrcVT  = TLI.getValueType(Src->getType(), true);
  EVT DestVT = TLI.getValueType(I->getType(), true);

  if (SrcVT != MVT::i64 && SrcVT != MVT::i32 && SrcVT != MVT::i16)
    return false;

  if (DestVT != MVT::i32 && DestVT != MVT::i16 && DestVT != MVT::i8)
    return false;

  unsigned SrcReg = getRegForValue(Src);
  if (!SrcReg)
    return false;

  // The only interesting case is when we need to switch register classes.
  if (SrcVT == MVT::i64) {
    unsigned ResultReg = createResultReg(&PPC::GPRCRegClass);
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL, TII.get(TargetOpcode::COPY),
            ResultReg).addReg(SrcReg, 0, PPC::sub_32);
    SrcReg = ResultReg;
  }

  UpdateValueMap(I, SrcReg);
  return true;
}

// Attempt to fast-select an integer extend instruction.
bool PPCFastISel::SelectIntExt(const Instruction *I) {
  Type *DestTy = I->getType();
  Value *Src = I->getOperand(0);
  Type *SrcTy = Src->getType();

  bool IsZExt = isa<ZExtInst>(I);
  unsigned SrcReg = getRegForValue(Src);
  if (!SrcReg) return false;

  EVT SrcEVT, DestEVT;
  SrcEVT = TLI.getValueType(SrcTy, true);
  DestEVT = TLI.getValueType(DestTy, true);
  if (!SrcEVT.isSimple())
    return false;
  if (!DestEVT.isSimple())
    return false;

  MVT SrcVT = SrcEVT.getSimpleVT();
  MVT DestVT = DestEVT.getSimpleVT();

  // If we know the register class needed for the result of this
  // instruction, use it.  Otherwise pick the register class of the
  // correct size that does not contain X0/R0, since we don't know
  // whether downstream uses permit that assignment.
  unsigned AssignedReg = FuncInfo.ValueMap[I];
  const TargetRegisterClass *RC =
    (AssignedReg ? MRI.getRegClass(AssignedReg) :
     (DestVT == MVT::i64 ? &PPC::G8RC_and_G8RC_NOX0RegClass :
      &PPC::GPRC_and_GPRC_NOR0RegClass));
  unsigned ResultReg = createResultReg(RC);

  if (!PPCEmitIntExt(SrcVT, SrcReg, DestVT, ResultReg, IsZExt))
    return false;

  UpdateValueMap(I, ResultReg);
  return true;
}

// Attempt to fast-select an instruction that wasn't handled by
// the table-generated machinery.
bool PPCFastISel::TargetSelectInstruction(const Instruction *I) {

  switch (I->getOpcode()) {
    case Instruction::Load:
      return SelectLoad(I);
    case Instruction::Store:
      return SelectStore(I);
    case Instruction::Br:
      return SelectBranch(I);
    case Instruction::IndirectBr:
      return SelectIndirectBr(I);
    case Instruction::FPExt:
      return SelectFPExt(I);
    case Instruction::FPTrunc:
      return SelectFPTrunc(I);
    case Instruction::SIToFP:
      return SelectIToFP(I, /*IsSigned*/ true);
    case Instruction::UIToFP:
      return SelectIToFP(I, /*IsSigned*/ false);
    case Instruction::FPToSI:
      return SelectFPToI(I, /*IsSigned*/ true);
    case Instruction::FPToUI:
      return SelectFPToI(I, /*IsSigned*/ false);
    case Instruction::Add:
      return SelectBinaryIntOp(I, ISD::ADD);
    case Instruction::Or:
      return SelectBinaryIntOp(I, ISD::OR);
    case Instruction::Sub:
      return SelectBinaryIntOp(I, ISD::SUB);
    case Instruction::Call:
      if (dyn_cast<IntrinsicInst>(I))
        return false;
      return SelectCall(I);
    case Instruction::Ret:
      return SelectRet(I);
    case Instruction::Trunc:
      return SelectTrunc(I);
    case Instruction::ZExt:
    case Instruction::SExt:
      return SelectIntExt(I);
    // Here add other flavors of Instruction::XXX that automated
    // cases don't catch.  For example, switches are terminators
    // that aren't yet handled.
    default:
      break;
  }
  return false;
}

// Materialize a floating-point constant into a register, and return
// the register number (or zero if we failed to handle it).
unsigned PPCFastISel::PPCMaterializeFP(const ConstantFP *CFP, MVT VT) {
  // No plans to handle long double here.
  if (VT != MVT::f32 && VT != MVT::f64)
    return 0;

  // All FP constants are loaded from the constant pool.
  unsigned Align = TD.getPrefTypeAlignment(CFP->getType());
  assert(Align > 0 && "Unexpectedly missing alignment information!");
  unsigned Idx = MCP.getConstantPoolIndex(cast<Constant>(CFP), Align);
  unsigned DestReg = createResultReg(TLI.getRegClassFor(VT));
  CodeModel::Model CModel = TM.getCodeModel();

  MachineMemOperand *MMO =
    FuncInfo.MF->getMachineMemOperand(
      MachinePointerInfo::getConstantPool(), MachineMemOperand::MOLoad,
      (VT == MVT::f32) ? 4 : 8, Align);

  unsigned Opc = (VT == MVT::f32) ? PPC::LFS : PPC::LFD;
  unsigned TmpReg = createResultReg(&PPC::G8RC_and_G8RC_NOX0RegClass);

  // For small code model, generate a LF[SD](0, LDtocCPT(Idx, X2)).
  if (CModel == CodeModel::Small || CModel == CodeModel::JITDefault) {
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL, TII.get(PPC::LDtocCPT),
            TmpReg)
      .addConstantPoolIndex(Idx).addReg(PPC::X2);
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL, TII.get(Opc), DestReg)
      .addImm(0).addReg(TmpReg).addMemOperand(MMO);
  } else {
    // Otherwise we generate LF[SD](Idx[lo], ADDIStocHA(X2, Idx)).
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL, TII.get(PPC::ADDIStocHA),
            TmpReg).addReg(PPC::X2).addConstantPoolIndex(Idx);
    // But for large code model, we must generate a LDtocL followed
    // by the LF[SD].
    if (CModel == CodeModel::Large) {
      unsigned TmpReg2 = createResultReg(&PPC::G8RC_and_G8RC_NOX0RegClass);
      BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL, TII.get(PPC::LDtocL),
              TmpReg2).addConstantPoolIndex(Idx).addReg(TmpReg);
      BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL, TII.get(Opc), DestReg)
        .addImm(0).addReg(TmpReg2);
    } else 
      BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL, TII.get(Opc), DestReg)
        .addConstantPoolIndex(Idx, 0, PPCII::MO_TOC_LO)
        .addReg(TmpReg)
        .addMemOperand(MMO);
  }

  return DestReg;
}

// Materialize the address of a global value into a register, and return
// the register number (or zero if we failed to handle it).
unsigned PPCFastISel::PPCMaterializeGV(const GlobalValue *GV, MVT VT) {
  assert(VT == MVT::i64 && "Non-address!");
  const TargetRegisterClass *RC = &PPC::G8RC_and_G8RC_NOX0RegClass;
  unsigned DestReg = createResultReg(RC);

  // Global values may be plain old object addresses, TLS object
  // addresses, constant pool entries, or jump tables.  How we generate
  // code for these may depend on small, medium, or large code model.
  CodeModel::Model CModel = TM.getCodeModel();

  // FIXME: Jump tables are not yet required because fast-isel doesn't
  // handle switches; if that changes, we need them as well.  For now,
  // what follows assumes everything's a generic (or TLS) global address.
  const GlobalVariable *GVar = dyn_cast<GlobalVariable>(GV);
  if (!GVar) {
    // If GV is an alias, use the aliasee for determining thread-locality.
    if (const GlobalAlias *GA = dyn_cast<GlobalAlias>(GV))
      GVar = dyn_cast_or_null<GlobalVariable>(GA->resolveAliasedGlobal(false));
  }

  // FIXME: We don't yet handle the complexity of TLS.
  bool IsTLS = GVar && GVar->isThreadLocal();
  if (IsTLS)
    return 0;

  // For small code model, generate a simple TOC load.
  if (CModel == CodeModel::Small || CModel == CodeModel::JITDefault)
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL, TII.get(PPC::LDtoc), DestReg)
      .addGlobalAddress(GV).addReg(PPC::X2);
  else {
    // If the address is an externally defined symbol, a symbol with
    // common or externally available linkage, a function address, or a
    // jump table address (not yet needed), or if we are generating code
    // for large code model, we generate:
    //       LDtocL(GV, ADDIStocHA(%X2, GV))
    // Otherwise we generate:
    //       ADDItocL(ADDIStocHA(%X2, GV), GV)
    // Either way, start with the ADDIStocHA:
    unsigned HighPartReg = createResultReg(RC);
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL, TII.get(PPC::ADDIStocHA),
            HighPartReg).addReg(PPC::X2).addGlobalAddress(GV);

    // !GVar implies a function address.  An external variable is one
    // without an initializer.
    // If/when switches are implemented, jump tables should be handled
    // on the "if" path here.
    if (CModel == CodeModel::Large || !GVar || !GVar->hasInitializer() ||
        GVar->hasCommonLinkage() || GVar->hasAvailableExternallyLinkage())
      BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL, TII.get(PPC::LDtocL),
              DestReg).addGlobalAddress(GV).addReg(HighPartReg);
    else
      // Otherwise generate the ADDItocL.
      BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL, TII.get(PPC::ADDItocL),
              DestReg).addReg(HighPartReg).addGlobalAddress(GV);
  }

  return DestReg;
}

// Materialize a 32-bit integer constant into a register, and return
// the register number (or zero if we failed to handle it).
unsigned PPCFastISel::PPCMaterialize32BitInt(int64_t Imm,
                                             const TargetRegisterClass *RC) {
  unsigned Lo = Imm & 0xFFFF;
  unsigned Hi = (Imm >> 16) & 0xFFFF;

  unsigned ResultReg = createResultReg(RC);
  bool IsGPRC = RC->hasSuperClassEq(&PPC::GPRCRegClass);

  if (isInt<16>(Imm))
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL,
            TII.get(IsGPRC ? PPC::LI : PPC::LI8), ResultReg)
      .addImm(Imm);
  else if (Lo) {
    // Both Lo and Hi have nonzero bits.
    unsigned TmpReg = createResultReg(RC);
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL,
            TII.get(IsGPRC ? PPC::LIS : PPC::LIS8), TmpReg)
      .addImm(Hi);
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL,
            TII.get(IsGPRC ? PPC::ORI : PPC::ORI8), ResultReg)
      .addReg(TmpReg).addImm(Lo);
  } else
    // Just Hi bits.
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL,
            TII.get(IsGPRC ? PPC::LIS : PPC::LIS8), ResultReg)
      .addImm(Hi);
  
  return ResultReg;
}

// Materialize a 64-bit integer constant into a register, and return
// the register number (or zero if we failed to handle it).
unsigned PPCFastISel::PPCMaterialize64BitInt(int64_t Imm,
                                             const TargetRegisterClass *RC) {
  unsigned Remainder = 0;
  unsigned Shift = 0;

  // If the value doesn't fit in 32 bits, see if we can shift it
  // so that it fits in 32 bits.
  if (!isInt<32>(Imm)) {
    Shift = countTrailingZeros<uint64_t>(Imm);
    int64_t ImmSh = static_cast<uint64_t>(Imm) >> Shift;

    if (isInt<32>(ImmSh))
      Imm = ImmSh;
    else {
      Remainder = Imm;
      Shift = 32;
      Imm >>= 32;
    }
  }

  // Handle the high-order 32 bits (if shifted) or the whole 32 bits
  // (if not shifted).
  unsigned TmpReg1 = PPCMaterialize32BitInt(Imm, RC);
  if (!Shift)
    return TmpReg1;

  // If upper 32 bits were not zero, we've built them and need to shift
  // them into place.
  unsigned TmpReg2;
  if (Imm) {
    TmpReg2 = createResultReg(RC);
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL, TII.get(PPC::RLDICR),
            TmpReg2).addReg(TmpReg1).addImm(Shift).addImm(63 - Shift);
  } else
    TmpReg2 = TmpReg1;

  unsigned TmpReg3, Hi, Lo;
  if ((Hi = (Remainder >> 16) & 0xFFFF)) {
    TmpReg3 = createResultReg(RC);
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL, TII.get(PPC::ORIS8),
            TmpReg3).addReg(TmpReg2).addImm(Hi);
  } else
    TmpReg3 = TmpReg2;

  if ((Lo = Remainder & 0xFFFF)) {
    unsigned ResultReg = createResultReg(RC);
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL, TII.get(PPC::ORI8),
            ResultReg).addReg(TmpReg3).addImm(Lo);
    return ResultReg;
  }

  return TmpReg3;
}


// Materialize an integer constant into a register, and return
// the register number (or zero if we failed to handle it).
unsigned PPCFastISel::PPCMaterializeInt(const Constant *C, MVT VT) {

  if (VT != MVT::i64 && VT != MVT::i32 && VT != MVT::i16 &&
      VT != MVT::i8 && VT != MVT::i1) 
    return 0;

  const TargetRegisterClass *RC = ((VT == MVT::i64) ? &PPC::G8RCRegClass :
                                   &PPC::GPRCRegClass);

  // If the constant is in range, use a load-immediate.
  const ConstantInt *CI = cast<ConstantInt>(C);
  if (isInt<16>(CI->getSExtValue())) {
    unsigned Opc = (VT == MVT::i64) ? PPC::LI8 : PPC::LI;
    unsigned ImmReg = createResultReg(RC);
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL, TII.get(Opc), ImmReg)
      .addImm(CI->getSExtValue());
    return ImmReg;
  }

  // Construct the constant piecewise.
  int64_t Imm = CI->getZExtValue();

  if (VT == MVT::i64)
    return PPCMaterialize64BitInt(Imm, RC);
  else if (VT == MVT::i32)
    return PPCMaterialize32BitInt(Imm, RC);

  return 0;
}

// Materialize a constant into a register, and return the register
// number (or zero if we failed to handle it).
unsigned PPCFastISel::TargetMaterializeConstant(const Constant *C) {
  EVT CEVT = TLI.getValueType(C->getType(), true);

  // Only handle simple types.
  if (!CEVT.isSimple()) return 0;
  MVT VT = CEVT.getSimpleVT();

  if (const ConstantFP *CFP = dyn_cast<ConstantFP>(C))
    return PPCMaterializeFP(CFP, VT);
  else if (const GlobalValue *GV = dyn_cast<GlobalValue>(C))
    return PPCMaterializeGV(GV, VT);
  else if (isa<ConstantInt>(C))
    return PPCMaterializeInt(C, VT);

  return 0;
}

// Materialize the address created by an alloca into a register, and
// return the register number (or zero if we failed to handle it).
unsigned PPCFastISel::TargetMaterializeAlloca(const AllocaInst *AI) {
  // Don't handle dynamic allocas.
  if (!FuncInfo.StaticAllocaMap.count(AI)) return 0;

  MVT VT;
  if (!isLoadTypeLegal(AI->getType(), VT)) return 0;

  DenseMap<const AllocaInst*, int>::iterator SI =
    FuncInfo.StaticAllocaMap.find(AI);

  if (SI != FuncInfo.StaticAllocaMap.end()) {
    unsigned ResultReg = createResultReg(&PPC::G8RC_and_G8RC_NOX0RegClass);
    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL, TII.get(PPC::ADDI8),
            ResultReg).addFrameIndex(SI->second).addImm(0);
    return ResultReg;
  }

  return 0;
}

// Fold loads into extends when possible.
// FIXME: We can have multiple redundant extend/trunc instructions
// following a load.  The folding only picks up one.  Extend this
// to check subsequent instructions for the same pattern and remove
// them.  Thus ResultReg should be the def reg for the last redundant
// instruction in a chain, and all intervening instructions can be
// removed from parent.  Change test/CodeGen/PowerPC/fast-isel-fold.ll
// to add ELF64-NOT: rldicl to the appropriate tests when this works.
bool PPCFastISel::tryToFoldLoadIntoMI(MachineInstr *MI, unsigned OpNo,
                                      const LoadInst *LI) {
  // Verify we have a legal type before going any further.
  MVT VT;
  if (!isLoadTypeLegal(LI->getType(), VT))
    return false;

  // Combine load followed by zero- or sign-extend.
  bool IsZExt = false;
  switch(MI->getOpcode()) {
    default:
      return false;

    case PPC::RLDICL:
    case PPC::RLDICL_32_64: {
      IsZExt = true;
      unsigned MB = MI->getOperand(3).getImm();
      if ((VT == MVT::i8 && MB <= 56) ||
          (VT == MVT::i16 && MB <= 48) ||
          (VT == MVT::i32 && MB <= 32))
        break;
      return false;
    }

    case PPC::RLWINM:
    case PPC::RLWINM8: {
      IsZExt = true;
      unsigned MB = MI->getOperand(3).getImm();
      if ((VT == MVT::i8 && MB <= 24) ||
          (VT == MVT::i16 && MB <= 16))
        break;
      return false;
    }

    case PPC::EXTSB:
    case PPC::EXTSB8:
    case PPC::EXTSB8_32_64:
      /* There is no sign-extending load-byte instruction. */
      return false;

    case PPC::EXTSH:
    case PPC::EXTSH8:
    case PPC::EXTSH8_32_64: {
      if (VT != MVT::i16 && VT != MVT::i8)
        return false;
      break;
    }

    case PPC::EXTSW:
    case PPC::EXTSW_32_64: {
      if (VT != MVT::i32 && VT != MVT::i16 && VT != MVT::i8)
        return false;
      break;
    }
  }

  // See if we can handle this address.
  Address Addr;
  if (!PPCComputeAddress(LI->getOperand(0), Addr))
    return false;

  unsigned ResultReg = MI->getOperand(0).getReg();

  if (!PPCEmitLoad(VT, ResultReg, Addr, 0, IsZExt))
    return false;

  MI->eraseFromParent();
  return true;
}

// Attempt to lower call arguments in a faster way than done by
// the selection DAG code.
bool PPCFastISel::FastLowerArguments() {
  // Defer to normal argument lowering for now.  It's reasonably
  // efficient.  Consider doing something like ARM to handle the
  // case where all args fit in registers, no varargs, no float
  // or vector args.
  return false;
}

// Handle materializing integer constants into a register.  This is not
// automatically generated for PowerPC, so must be explicitly created here.
unsigned PPCFastISel::FastEmit_i(MVT Ty, MVT VT, unsigned Opc, uint64_t Imm) {
  
  if (Opc != ISD::Constant)
    return 0;

  if (VT != MVT::i64 && VT != MVT::i32 && VT != MVT::i16 &&
      VT != MVT::i8 && VT != MVT::i1) 
    return 0;

  const TargetRegisterClass *RC = ((VT == MVT::i64) ? &PPC::G8RCRegClass :
                                   &PPC::GPRCRegClass);
  if (VT == MVT::i64)
    return PPCMaterialize64BitInt(Imm, RC);
  else
    return PPCMaterialize32BitInt(Imm, RC);
}

// Override for ADDI and ADDI8 to set the correct register class
// on RHS operand 0.  The automatic infrastructure naively assumes
// GPRC for i32 and G8RC for i64; the concept of "no R0" is lost
// for these cases.  At the moment, none of the other automatically
// generated RI instructions require special treatment.  However, once
// SelectSelect is implemented, "isel" requires similar handling.
//
// Also be conservative about the output register class.  Avoid
// assigning R0 or X0 to the output register for GPRC and G8RC
// register classes, as any such result could be used in ADDI, etc.,
// where those regs have another meaning.
unsigned PPCFastISel::FastEmitInst_ri(unsigned MachineInstOpcode,
                                      const TargetRegisterClass *RC,
                                      unsigned Op0, bool Op0IsKill,
                                      uint64_t Imm) {
  if (MachineInstOpcode == PPC::ADDI)
    MRI.setRegClass(Op0, &PPC::GPRC_and_GPRC_NOR0RegClass);
  else if (MachineInstOpcode == PPC::ADDI8)
    MRI.setRegClass(Op0, &PPC::G8RC_and_G8RC_NOX0RegClass);

  const TargetRegisterClass *UseRC =
    (RC == &PPC::GPRCRegClass ? &PPC::GPRC_and_GPRC_NOR0RegClass :
     (RC == &PPC::G8RCRegClass ? &PPC::G8RC_and_G8RC_NOX0RegClass : RC));

  return FastISel::FastEmitInst_ri(MachineInstOpcode, UseRC,
                                   Op0, Op0IsKill, Imm);
}

// Override for instructions with one register operand to avoid use of
// R0/X0.  The automatic infrastructure isn't aware of the context so
// we must be conservative.
unsigned PPCFastISel::FastEmitInst_r(unsigned MachineInstOpcode,
                                     const TargetRegisterClass* RC,
                                     unsigned Op0, bool Op0IsKill) {
  const TargetRegisterClass *UseRC =
    (RC == &PPC::GPRCRegClass ? &PPC::GPRC_and_GPRC_NOR0RegClass :
     (RC == &PPC::G8RCRegClass ? &PPC::G8RC_and_G8RC_NOX0RegClass : RC));

  return FastISel::FastEmitInst_r(MachineInstOpcode, UseRC, Op0, Op0IsKill);
}

// Override for instructions with two register operands to avoid use
// of R0/X0.  The automatic infrastructure isn't aware of the context
// so we must be conservative.
unsigned PPCFastISel::FastEmitInst_rr(unsigned MachineInstOpcode,
                                      const TargetRegisterClass* RC,
                                      unsigned Op0, bool Op0IsKill,
                                      unsigned Op1, bool Op1IsKill) {
  const TargetRegisterClass *UseRC =
    (RC == &PPC::GPRCRegClass ? &PPC::GPRC_and_GPRC_NOR0RegClass :
     (RC == &PPC::G8RCRegClass ? &PPC::G8RC_and_G8RC_NOX0RegClass : RC));

  return FastISel::FastEmitInst_rr(MachineInstOpcode, UseRC, Op0, Op0IsKill,
                                   Op1, Op1IsKill);
}

namespace llvm {
  // Create the fast instruction selector for PowerPC64 ELF.
  FastISel *PPC::createFastISel(FunctionLoweringInfo &FuncInfo,
                                const TargetLibraryInfo *LibInfo) {
    const TargetMachine &TM = FuncInfo.MF->getTarget();

    // Only available on 64-bit ELF for now.
    const PPCSubtarget *Subtarget = &TM.getSubtarget<PPCSubtarget>();
    if (Subtarget->isPPC64() && Subtarget->isSVR4ABI())
      return new PPCFastISel(FuncInfo, LibInfo);

    return 0;
  }
}