Training courses

Kernel and Embedded Linux

Bootlin training courses

Embedded Linux, kernel,
Yocto Project, Buildroot, real-time,
graphics, boot time, debugging...

Bootlin logo

Elixir Cross Referencer

   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
//===- LoopStrengthReduce.cpp - Strength Reduce IVs in Loops --------------===//
//
//                     The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This transformation analyzes and transforms the induction variables (and
// computations derived from them) into forms suitable for efficient execution
// on the target.
//
// This pass performs a strength reduction on array references inside loops that
// have as one or more of their components the loop induction variable, it
// rewrites expressions to take advantage of scaled-index addressing modes
// available on the target, and it performs a variety of other optimizations
// related to loop induction variables.
//
// Terminology note: this code has a lot of handling for "post-increment" or
// "post-inc" users. This is not talking about post-increment addressing modes;
// it is instead talking about code like this:
//
//   %i = phi [ 0, %entry ], [ %i.next, %latch ]
//   ...
//   %i.next = add %i, 1
//   %c = icmp eq %i.next, %n
//
// The SCEV for %i is {0,+,1}<%L>. The SCEV for %i.next is {1,+,1}<%L>, however
// it's useful to think about these as the same register, with some uses using
// the value of the register before the add and some using it after. In this
// example, the icmp is a post-increment user, since it uses %i.next, which is
// the value of the induction variable after the increment. The other common
// case of post-increment users is users outside the loop.
//
// TODO: More sophistication in the way Formulae are generated and filtered.
//
// TODO: Handle multiple loops at a time.
//
// TODO: Should the addressing mode BaseGV be changed to a ConstantExpr instead
//       of a GlobalValue?
//
// TODO: When truncation is free, truncate ICmp users' operands to make it a
//       smaller encoding (on x86 at least).
//
// TODO: When a negated register is used by an add (such as in a list of
//       multiple base registers, or as the increment expression in an addrec),
//       we may not actually need both reg and (-1 * reg) in registers; the
//       negation can be implemented by using a sub instead of an add. The
//       lack of support for taking this into consideration when making
//       register pressure decisions is partly worked around by the "Special"
//       use kind.
//
//===----------------------------------------------------------------------===//

#include "llvm/Transforms/Scalar/LoopStrengthReduce.h"
#include "llvm/ADT/APInt.h"
#include "llvm/ADT/DenseMap.h"
#include "llvm/ADT/DenseSet.h"
#include "llvm/ADT/Hashing.h"
#include "llvm/ADT/PointerIntPair.h"
#include "llvm/ADT/STLExtras.h"
#include "llvm/ADT/SetVector.h"
#include "llvm/ADT/SmallBitVector.h"
#include "llvm/ADT/SmallPtrSet.h"
#include "llvm/ADT/SmallSet.h"
#include "llvm/ADT/SmallVector.h"
#include "llvm/ADT/iterator_range.h"
#include "llvm/Analysis/IVUsers.h"
#include "llvm/Analysis/LoopAnalysisManager.h"
#include "llvm/Analysis/LoopInfo.h"
#include "llvm/Analysis/LoopPass.h"
#include "llvm/Analysis/ScalarEvolution.h"
#include "llvm/Analysis/ScalarEvolutionExpander.h"
#include "llvm/Analysis/ScalarEvolutionExpressions.h"
#include "llvm/Analysis/ScalarEvolutionNormalization.h"
#include "llvm/Analysis/TargetTransformInfo.h"
#include "llvm/IR/BasicBlock.h"
#include "llvm/IR/Constant.h"
#include "llvm/IR/Constants.h"
#include "llvm/IR/DerivedTypes.h"
#include "llvm/IR/Dominators.h"
#include "llvm/IR/GlobalValue.h"
#include "llvm/IR/IRBuilder.h"
#include "llvm/IR/InstrTypes.h"
#include "llvm/IR/Instruction.h"
#include "llvm/IR/Instructions.h"
#include "llvm/IR/IntrinsicInst.h"
#include "llvm/IR/Intrinsics.h"
#include "llvm/IR/Module.h"
#include "llvm/IR/OperandTraits.h"
#include "llvm/IR/Operator.h"
#include "llvm/IR/PassManager.h"
#include "llvm/IR/Type.h"
#include "llvm/IR/Use.h"
#include "llvm/IR/User.h"
#include "llvm/IR/Value.h"
#include "llvm/IR/ValueHandle.h"
#include "llvm/Pass.h"
#include "llvm/Support/Casting.h"
#include "llvm/Support/CommandLine.h"
#include "llvm/Support/Compiler.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/ErrorHandling.h"
#include "llvm/Support/MathExtras.h"
#include "llvm/Support/raw_ostream.h"
#include "llvm/Transforms/Scalar.h"
#include "llvm/Transforms/Utils/BasicBlockUtils.h"
#include "llvm/Transforms/Utils/Local.h"
#include <algorithm>
#include <cassert>
#include <cstddef>
#include <cstdint>
#include <cstdlib>
#include <iterator>
#include <limits>
#include <map>
#include <utility>

using namespace llvm;

#define DEBUG_TYPE "loop-reduce"

/// MaxIVUsers is an arbitrary threshold that provides an early opportunitiy for
/// bail out. This threshold is far beyond the number of users that LSR can
/// conceivably solve, so it should not affect generated code, but catches the
/// worst cases before LSR burns too much compile time and stack space.
static const unsigned MaxIVUsers = 200;

// Temporary flag to cleanup congruent phis after LSR phi expansion.
// It's currently disabled until we can determine whether it's truly useful or
// not. The flag should be removed after the v3.0 release.
// This is now needed for ivchains.
static cl::opt<bool> EnablePhiElim(
  "enable-lsr-phielim", cl::Hidden, cl::init(true),
  cl::desc("Enable LSR phi elimination"));

// The flag adds instruction count to solutions cost comparision.
static cl::opt<bool> InsnsCost(
  "lsr-insns-cost", cl::Hidden, cl::init(true),
  cl::desc("Add instruction count to a LSR cost model"));

// Flag to choose how to narrow complex lsr solution
static cl::opt<bool> LSRExpNarrow(
  "lsr-exp-narrow", cl::Hidden, cl::init(false),
  cl::desc("Narrow LSR complex solution using"
           " expectation of registers number"));

// Flag to narrow search space by filtering non-optimal formulae with
// the same ScaledReg and Scale.
static cl::opt<bool> FilterSameScaledReg(
    "lsr-filter-same-scaled-reg", cl::Hidden, cl::init(true),
    cl::desc("Narrow LSR search space by filtering non-optimal formulae"
             " with the same ScaledReg and Scale"));

#ifndef NDEBUG
// Stress test IV chain generation.
static cl::opt<bool> StressIVChain(
  "stress-ivchain", cl::Hidden, cl::init(false),
  cl::desc("Stress test LSR IV chains"));
#else
static bool StressIVChain = false;
#endif

namespace {

struct MemAccessTy {
  /// Used in situations where the accessed memory type is unknown.
  static const unsigned UnknownAddressSpace =
      std::numeric_limits<unsigned>::max();

  Type *MemTy = nullptr;
  unsigned AddrSpace = UnknownAddressSpace;

  MemAccessTy() = default;
  MemAccessTy(Type *Ty, unsigned AS) : MemTy(Ty), AddrSpace(AS) {}

  bool operator==(MemAccessTy Other) const {
    return MemTy == Other.MemTy && AddrSpace == Other.AddrSpace;
  }

  bool operator!=(MemAccessTy Other) const { return !(*this == Other); }

  static MemAccessTy getUnknown(LLVMContext &Ctx,
                                unsigned AS = UnknownAddressSpace) {
    return MemAccessTy(Type::getVoidTy(Ctx), AS);
  }
};

/// This class holds data which is used to order reuse candidates.
class RegSortData {
public:
  /// This represents the set of LSRUse indices which reference
  /// a particular register.
  SmallBitVector UsedByIndices;

  void print(raw_ostream &OS) const;
  void dump() const;
};

} // end anonymous namespace

#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
void RegSortData::print(raw_ostream &OS) const {
  OS << "[NumUses=" << UsedByIndices.count() << ']';
}

LLVM_DUMP_METHOD void RegSortData::dump() const {
  print(errs()); errs() << '\n';
}
#endif

namespace {

/// Map register candidates to information about how they are used.
class RegUseTracker {
  using RegUsesTy = DenseMap<const SCEV *, RegSortData>;

  RegUsesTy RegUsesMap;
  SmallVector<const SCEV *, 16> RegSequence;

public:
  void countRegister(const SCEV *Reg, size_t LUIdx);
  void dropRegister(const SCEV *Reg, size_t LUIdx);
  void swapAndDropUse(size_t LUIdx, size_t LastLUIdx);

  bool isRegUsedByUsesOtherThan(const SCEV *Reg, size_t LUIdx) const;

  const SmallBitVector &getUsedByIndices(const SCEV *Reg) const;

  void clear();

  using iterator = SmallVectorImpl<const SCEV *>::iterator;
  using const_iterator = SmallVectorImpl<const SCEV *>::const_iterator;

  iterator begin() { return RegSequence.begin(); }
  iterator end()   { return RegSequence.end(); }
  const_iterator begin() const { return RegSequence.begin(); }
  const_iterator end() const   { return RegSequence.end(); }
};

} // end anonymous namespace

void
RegUseTracker::countRegister(const SCEV *Reg, size_t LUIdx) {
  std::pair<RegUsesTy::iterator, bool> Pair =
    RegUsesMap.insert(std::make_pair(Reg, RegSortData()));
  RegSortData &RSD = Pair.first->second;
  if (Pair.second)
    RegSequence.push_back(Reg);
  RSD.UsedByIndices.resize(std::max(RSD.UsedByIndices.size(), LUIdx + 1));
  RSD.UsedByIndices.set(LUIdx);
}

void
RegUseTracker::dropRegister(const SCEV *Reg, size_t LUIdx) {
  RegUsesTy::iterator It = RegUsesMap.find(Reg);
  assert(It != RegUsesMap.end());
  RegSortData &RSD = It->second;
  assert(RSD.UsedByIndices.size() > LUIdx);
  RSD.UsedByIndices.reset(LUIdx);
}

void
RegUseTracker::swapAndDropUse(size_t LUIdx, size_t LastLUIdx) {
  assert(LUIdx <= LastLUIdx);

  // Update RegUses. The data structure is not optimized for this purpose;
  // we must iterate through it and update each of the bit vectors.
  for (auto &Pair : RegUsesMap) {
    SmallBitVector &UsedByIndices = Pair.second.UsedByIndices;
    if (LUIdx < UsedByIndices.size())
      UsedByIndices[LUIdx] =
        LastLUIdx < UsedByIndices.size() ? UsedByIndices[LastLUIdx] : false;
    UsedByIndices.resize(std::min(UsedByIndices.size(), LastLUIdx));
  }
}

bool
RegUseTracker::isRegUsedByUsesOtherThan(const SCEV *Reg, size_t LUIdx) const {
  RegUsesTy::const_iterator I = RegUsesMap.find(Reg);
  if (I == RegUsesMap.end())
    return false;
  const SmallBitVector &UsedByIndices = I->second.UsedByIndices;
  int i = UsedByIndices.find_first();
  if (i == -1) return false;
  if ((size_t)i != LUIdx) return true;
  return UsedByIndices.find_next(i) != -1;
}

const SmallBitVector &RegUseTracker::getUsedByIndices(const SCEV *Reg) const {
  RegUsesTy::const_iterator I = RegUsesMap.find(Reg);
  assert(I != RegUsesMap.end() && "Unknown register!");
  return I->second.UsedByIndices;
}

void RegUseTracker::clear() {
  RegUsesMap.clear();
  RegSequence.clear();
}

namespace {

/// This class holds information that describes a formula for computing
/// satisfying a use. It may include broken-out immediates and scaled registers.
struct Formula {
  /// Global base address used for complex addressing.
  GlobalValue *BaseGV = nullptr;

  /// Base offset for complex addressing.
  int64_t BaseOffset = 0;

  /// Whether any complex addressing has a base register.
  bool HasBaseReg = false;

  /// The scale of any complex addressing.
  int64_t Scale = 0;

  /// The list of "base" registers for this use. When this is non-empty. The
  /// canonical representation of a formula is
  /// 1. BaseRegs.size > 1 implies ScaledReg != NULL and
  /// 2. ScaledReg != NULL implies Scale != 1 || !BaseRegs.empty().
  /// 3. The reg containing recurrent expr related with currect loop in the
  /// formula should be put in the ScaledReg.
  /// #1 enforces that the scaled register is always used when at least two
  /// registers are needed by the formula: e.g., reg1 + reg2 is reg1 + 1 * reg2.
  /// #2 enforces that 1 * reg is reg.
  /// #3 ensures invariant regs with respect to current loop can be combined
  /// together in LSR codegen.
  /// This invariant can be temporarly broken while building a formula.
  /// However, every formula inserted into the LSRInstance must be in canonical
  /// form.
  SmallVector<const SCEV *, 4> BaseRegs;

  /// The 'scaled' register for this use. This should be non-null when Scale is
  /// not zero.
  const SCEV *ScaledReg = nullptr;

  /// An additional constant offset which added near the use. This requires a
  /// temporary register, but the offset itself can live in an add immediate
  /// field rather than a register.
  int64_t UnfoldedOffset = 0;

  Formula() = default;

  void initialMatch(const SCEV *S, Loop *L, ScalarEvolution &SE);

  bool isCanonical(const Loop &L) const;

  void canonicalize(const Loop &L);

  bool unscale();

  bool hasZeroEnd() const;

  size_t getNumRegs() const;
  Type *getType() const;

  void deleteBaseReg(const SCEV *&S);

  bool referencesReg(const SCEV *S) const;
  bool hasRegsUsedByUsesOtherThan(size_t LUIdx,
                                  const RegUseTracker &RegUses) const;

  void print(raw_ostream &OS) const;
  void dump() const;
};

} // end anonymous namespace

/// Recursion helper for initialMatch.
static void DoInitialMatch(const SCEV *S, Loop *L,
                           SmallVectorImpl<const SCEV *> &Good,
                           SmallVectorImpl<const SCEV *> &Bad,
                           ScalarEvolution &SE) {
  // Collect expressions which properly dominate the loop header.
  if (SE.properlyDominates(S, L->getHeader())) {
    Good.push_back(S);
    return;
  }

  // Look at add operands.
  if (const SCEVAddExpr *Add = dyn_cast<SCEVAddExpr>(S)) {
    for (const SCEV *S : Add->operands())
      DoInitialMatch(S, L, Good, Bad, SE);
    return;
  }

  // Look at addrec operands.
  if (const SCEVAddRecExpr *AR = dyn_cast<SCEVAddRecExpr>(S))
    if (!AR->getStart()->isZero() && AR->isAffine()) {
      DoInitialMatch(AR->getStart(), L, Good, Bad, SE);
      DoInitialMatch(SE.getAddRecExpr(SE.getConstant(AR->getType(), 0),
                                      AR->getStepRecurrence(SE),
                                      // FIXME: AR->getNoWrapFlags()
                                      AR->getLoop(), SCEV::FlagAnyWrap),
                     L, Good, Bad, SE);
      return;
    }

  // Handle a multiplication by -1 (negation) if it didn't fold.
  if (const SCEVMulExpr *Mul = dyn_cast<SCEVMulExpr>(S))
    if (Mul->getOperand(0)->isAllOnesValue()) {
      SmallVector<const SCEV *, 4> Ops(Mul->op_begin()+1, Mul->op_end());
      const SCEV *NewMul = SE.getMulExpr(Ops);

      SmallVector<const SCEV *, 4> MyGood;
      SmallVector<const SCEV *, 4> MyBad;
      DoInitialMatch(NewMul, L, MyGood, MyBad, SE);
      const SCEV *NegOne = SE.getSCEV(ConstantInt::getAllOnesValue(
        SE.getEffectiveSCEVType(NewMul->getType())));
      for (const SCEV *S : MyGood)
        Good.push_back(SE.getMulExpr(NegOne, S));
      for (const SCEV *S : MyBad)
        Bad.push_back(SE.getMulExpr(NegOne, S));
      return;
    }

  // Ok, we can't do anything interesting. Just stuff the whole thing into a
  // register and hope for the best.
  Bad.push_back(S);
}

/// Incorporate loop-variant parts of S into this Formula, attempting to keep
/// all loop-invariant and loop-computable values in a single base register.
void Formula::initialMatch(const SCEV *S, Loop *L, ScalarEvolution &SE) {
  SmallVector<const SCEV *, 4> Good;
  SmallVector<const SCEV *, 4> Bad;
  DoInitialMatch(S, L, Good, Bad, SE);
  if (!Good.empty()) {
    const SCEV *Sum = SE.getAddExpr(Good);
    if (!Sum->isZero())
      BaseRegs.push_back(Sum);
    HasBaseReg = true;
  }
  if (!Bad.empty()) {
    const SCEV *Sum = SE.getAddExpr(Bad);
    if (!Sum->isZero())
      BaseRegs.push_back(Sum);
    HasBaseReg = true;
  }
  canonicalize(*L);
}

/// \brief Check whether or not this formula statisfies the canonical
/// representation.
/// \see Formula::BaseRegs.
bool Formula::isCanonical(const Loop &L) const {
  if (!ScaledReg)
    return BaseRegs.size() <= 1;

  if (Scale != 1)
    return true;

  if (Scale == 1 && BaseRegs.empty())
    return false;

  const SCEVAddRecExpr *SAR = dyn_cast<const SCEVAddRecExpr>(ScaledReg);
  if (SAR && SAR->getLoop() == &L)
    return true;

  // If ScaledReg is not a recurrent expr, or it is but its loop is not current
  // loop, meanwhile BaseRegs contains a recurrent expr reg related with current
  // loop, we want to swap the reg in BaseRegs with ScaledReg.
  auto I =
      find_if(make_range(BaseRegs.begin(), BaseRegs.end()), [&](const SCEV *S) {
        return isa<const SCEVAddRecExpr>(S) &&
               (cast<SCEVAddRecExpr>(S)->getLoop() == &L);
      });
  return I == BaseRegs.end();
}

/// \brief Helper method to morph a formula into its canonical representation.
/// \see Formula::BaseRegs.
/// Every formula having more than one base register, must use the ScaledReg
/// field. Otherwise, we would have to do special cases everywhere in LSR
/// to treat reg1 + reg2 + ... the same way as reg1 + 1*reg2 + ...
/// On the other hand, 1*reg should be canonicalized into reg.
void Formula::canonicalize(const Loop &L) {
  if (isCanonical(L))
    return;
  // So far we did not need this case. This is easy to implement but it is
  // useless to maintain dead code. Beside it could hurt compile time.
  assert(!BaseRegs.empty() && "1*reg => reg, should not be needed.");

  // Keep the invariant sum in BaseRegs and one of the variant sum in ScaledReg.
  if (!ScaledReg) {
    ScaledReg = BaseRegs.back();
    BaseRegs.pop_back();
    Scale = 1;
  }

  // If ScaledReg is an invariant with respect to L, find the reg from
  // BaseRegs containing the recurrent expr related with Loop L. Swap the
  // reg with ScaledReg.
  const SCEVAddRecExpr *SAR = dyn_cast<const SCEVAddRecExpr>(ScaledReg);
  if (!SAR || SAR->getLoop() != &L) {
    auto I = find_if(make_range(BaseRegs.begin(), BaseRegs.end()),
                     [&](const SCEV *S) {
                       return isa<const SCEVAddRecExpr>(S) &&
                              (cast<SCEVAddRecExpr>(S)->getLoop() == &L);
                     });
    if (I != BaseRegs.end())
      std::swap(ScaledReg, *I);
  }
}

/// \brief Get rid of the scale in the formula.
/// In other words, this method morphes reg1 + 1*reg2 into reg1 + reg2.
/// \return true if it was possible to get rid of the scale, false otherwise.
/// \note After this operation the formula may not be in the canonical form.
bool Formula::unscale() {
  if (Scale != 1)
    return false;
  Scale = 0;
  BaseRegs.push_back(ScaledReg);
  ScaledReg = nullptr;
  return true;
}

bool Formula::hasZeroEnd() const {
  if (UnfoldedOffset || BaseOffset)
    return false;
  if (BaseRegs.size() != 1 || ScaledReg)
    return false;
  return true;
}

/// Return the total number of register operands used by this formula. This does
/// not include register uses implied by non-constant addrec strides.
size_t Formula::getNumRegs() const {
  return !!ScaledReg + BaseRegs.size();
}

/// Return the type of this formula, if it has one, or null otherwise. This type
/// is meaningless except for the bit size.
Type *Formula::getType() const {
  return !BaseRegs.empty() ? BaseRegs.front()->getType() :
         ScaledReg ? ScaledReg->getType() :
         BaseGV ? BaseGV->getType() :
         nullptr;
}

/// Delete the given base reg from the BaseRegs list.
void Formula::deleteBaseReg(const SCEV *&S) {
  if (&S != &BaseRegs.back())
    std::swap(S, BaseRegs.back());
  BaseRegs.pop_back();
}

/// Test if this formula references the given register.
bool Formula::referencesReg(const SCEV *S) const {
  return S == ScaledReg || is_contained(BaseRegs, S);
}

/// Test whether this formula uses registers which are used by uses other than
/// the use with the given index.
bool Formula::hasRegsUsedByUsesOtherThan(size_t LUIdx,
                                         const RegUseTracker &RegUses) const {
  if (ScaledReg)
    if (RegUses.isRegUsedByUsesOtherThan(ScaledReg, LUIdx))
      return true;
  for (const SCEV *BaseReg : BaseRegs)
    if (RegUses.isRegUsedByUsesOtherThan(BaseReg, LUIdx))
      return true;
  return false;
}

#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
void Formula::print(raw_ostream &OS) const {
  bool First = true;
  if (BaseGV) {
    if (!First) OS << " + "; else First = false;
    BaseGV->printAsOperand(OS, /*PrintType=*/false);
  }
  if (BaseOffset != 0) {
    if (!First) OS << " + "; else First = false;
    OS << BaseOffset;
  }
  for (const SCEV *BaseReg : BaseRegs) {
    if (!First) OS << " + "; else First = false;
    OS << "reg(" << *BaseReg << ')';
  }
  if (HasBaseReg && BaseRegs.empty()) {
    if (!First) OS << " + "; else First = false;
    OS << "**error: HasBaseReg**";
  } else if (!HasBaseReg && !BaseRegs.empty()) {
    if (!First) OS << " + "; else First = false;
    OS << "**error: !HasBaseReg**";
  }
  if (Scale != 0) {
    if (!First) OS << " + "; else First = false;
    OS << Scale << "*reg(";
    if (ScaledReg)
      OS << *ScaledReg;
    else
      OS << "<unknown>";
    OS << ')';
  }
  if (UnfoldedOffset != 0) {
    if (!First) OS << " + ";
    OS << "imm(" << UnfoldedOffset << ')';
  }
}

LLVM_DUMP_METHOD void Formula::dump() const {
  print(errs()); errs() << '\n';
}
#endif

/// Return true if the given addrec can be sign-extended without changing its
/// value.
static bool isAddRecSExtable(const SCEVAddRecExpr *AR, ScalarEvolution &SE) {
  Type *WideTy =
    IntegerType::get(SE.getContext(), SE.getTypeSizeInBits(AR->getType()) + 1);
  return isa<SCEVAddRecExpr>(SE.getSignExtendExpr(AR, WideTy));
}

/// Return true if the given add can be sign-extended without changing its
/// value.
static bool isAddSExtable(const SCEVAddExpr *A, ScalarEvolution &SE) {
  Type *WideTy =
    IntegerType::get(SE.getContext(), SE.getTypeSizeInBits(A->getType()) + 1);
  return isa<SCEVAddExpr>(SE.getSignExtendExpr(A, WideTy));
}

/// Return true if the given mul can be sign-extended without changing its
/// value.
static bool isMulSExtable(const SCEVMulExpr *M, ScalarEvolution &SE) {
  Type *WideTy =
    IntegerType::get(SE.getContext(),
                     SE.getTypeSizeInBits(M->getType()) * M->getNumOperands());
  return isa<SCEVMulExpr>(SE.getSignExtendExpr(M, WideTy));
}

/// Return an expression for LHS /s RHS, if it can be determined and if the
/// remainder is known to be zero, or null otherwise. If IgnoreSignificantBits
/// is true, expressions like (X * Y) /s Y are simplified to Y, ignoring that
/// the multiplication may overflow, which is useful when the result will be
/// used in a context where the most significant bits are ignored.
static const SCEV *getExactSDiv(const SCEV *LHS, const SCEV *RHS,
                                ScalarEvolution &SE,
                                bool IgnoreSignificantBits = false) {
  // Handle the trivial case, which works for any SCEV type.
  if (LHS == RHS)
    return SE.getConstant(LHS->getType(), 1);

  // Handle a few RHS special cases.
  const SCEVConstant *RC = dyn_cast<SCEVConstant>(RHS);
  if (RC) {
    const APInt &RA = RC->getAPInt();
    // Handle x /s -1 as x * -1, to give ScalarEvolution a chance to do
    // some folding.
    if (RA.isAllOnesValue())
      return SE.getMulExpr(LHS, RC);
    // Handle x /s 1 as x.
    if (RA == 1)
      return LHS;
  }

  // Check for a division of a constant by a constant.
  if (const SCEVConstant *C = dyn_cast<SCEVConstant>(LHS)) {
    if (!RC)
      return nullptr;
    const APInt &LA = C->getAPInt();
    const APInt &RA = RC->getAPInt();
    if (LA.srem(RA) != 0)
      return nullptr;
    return SE.getConstant(LA.sdiv(RA));
  }

  // Distribute the sdiv over addrec operands, if the addrec doesn't overflow.
  if (const SCEVAddRecExpr *AR = dyn_cast<SCEVAddRecExpr>(LHS)) {
    if ((IgnoreSignificantBits || isAddRecSExtable(AR, SE)) && AR->isAffine()) {
      const SCEV *Step = getExactSDiv(AR->getStepRecurrence(SE), RHS, SE,
                                      IgnoreSignificantBits);
      if (!Step) return nullptr;
      const SCEV *Start = getExactSDiv(AR->getStart(), RHS, SE,
                                       IgnoreSignificantBits);
      if (!Start) return nullptr;
      // FlagNW is independent of the start value, step direction, and is
      // preserved with smaller magnitude steps.
      // FIXME: AR->getNoWrapFlags(SCEV::FlagNW)
      return SE.getAddRecExpr(Start, Step, AR->getLoop(), SCEV::FlagAnyWrap);
    }
    return nullptr;
  }

  // Distribute the sdiv over add operands, if the add doesn't overflow.
  if (const SCEVAddExpr *Add = dyn_cast<SCEVAddExpr>(LHS)) {
    if (IgnoreSignificantBits || isAddSExtable(Add, SE)) {
      SmallVector<const SCEV *, 8> Ops;
      for (const SCEV *S : Add->operands()) {
        const SCEV *Op = getExactSDiv(S, RHS, SE, IgnoreSignificantBits);
        if (!Op) return nullptr;
        Ops.push_back(Op);
      }
      return SE.getAddExpr(Ops);
    }
    return nullptr;
  }

  // Check for a multiply operand that we can pull RHS out of.
  if (const SCEVMulExpr *Mul = dyn_cast<SCEVMulExpr>(LHS)) {
    if (IgnoreSignificantBits || isMulSExtable(Mul, SE)) {
      SmallVector<const SCEV *, 4> Ops;
      bool Found = false;
      for (const SCEV *S : Mul->operands()) {
        if (!Found)
          if (const SCEV *Q = getExactSDiv(S, RHS, SE,
                                           IgnoreSignificantBits)) {
            S = Q;
            Found = true;
          }
        Ops.push_back(S);
      }
      return Found ? SE.getMulExpr(Ops) : nullptr;
    }
    return nullptr;
  }

  // Otherwise we don't know.
  return nullptr;
}

/// If S involves the addition of a constant integer value, return that integer
/// value, and mutate S to point to a new SCEV with that value excluded.
static int64_t ExtractImmediate(const SCEV *&S, ScalarEvolution &SE) {
  if (const SCEVConstant *C = dyn_cast<SCEVConstant>(S)) {
    if (C->getAPInt().getMinSignedBits() <= 64) {
      S = SE.getConstant(C->getType(), 0);
      return C->getValue()->getSExtValue();
    }
  } else if (const SCEVAddExpr *Add = dyn_cast<SCEVAddExpr>(S)) {
    SmallVector<const SCEV *, 8> NewOps(Add->op_begin(), Add->op_end());
    int64_t Result = ExtractImmediate(NewOps.front(), SE);
    if (Result != 0)
      S = SE.getAddExpr(NewOps);
    return Result;
  } else if (const SCEVAddRecExpr *AR = dyn_cast<SCEVAddRecExpr>(S)) {
    SmallVector<const SCEV *, 8> NewOps(AR->op_begin(), AR->op_end());
    int64_t Result = ExtractImmediate(NewOps.front(), SE);
    if (Result != 0)
      S = SE.getAddRecExpr(NewOps, AR->getLoop(),
                           // FIXME: AR->getNoWrapFlags(SCEV::FlagNW)
                           SCEV::FlagAnyWrap);
    return Result;
  }
  return 0;
}

/// If S involves the addition of a GlobalValue address, return that symbol, and
/// mutate S to point to a new SCEV with that value excluded.
static GlobalValue *ExtractSymbol(const SCEV *&S, ScalarEvolution &SE) {
  if (const SCEVUnknown *U = dyn_cast<SCEVUnknown>(S)) {
    if (GlobalValue *GV = dyn_cast<GlobalValue>(U->getValue())) {
      S = SE.getConstant(GV->getType(), 0);
      return GV;
    }
  } else if (const SCEVAddExpr *Add = dyn_cast<SCEVAddExpr>(S)) {
    SmallVector<const SCEV *, 8> NewOps(Add->op_begin(), Add->op_end());
    GlobalValue *Result = ExtractSymbol(NewOps.back(), SE);
    if (Result)
      S = SE.getAddExpr(NewOps);
    return Result;
  } else if (const SCEVAddRecExpr *AR = dyn_cast<SCEVAddRecExpr>(S)) {
    SmallVector<const SCEV *, 8> NewOps(AR->op_begin(), AR->op_end());
    GlobalValue *Result = ExtractSymbol(NewOps.front(), SE);
    if (Result)
      S = SE.getAddRecExpr(NewOps, AR->getLoop(),
                           // FIXME: AR->getNoWrapFlags(SCEV::FlagNW)
                           SCEV::FlagAnyWrap);
    return Result;
  }
  return nullptr;
}

/// Returns true if the specified instruction is using the specified value as an
/// address.
static bool isAddressUse(const TargetTransformInfo &TTI,
                         Instruction *Inst, Value *OperandVal) {
  bool isAddress = isa<LoadInst>(Inst);
  if (StoreInst *SI = dyn_cast<StoreInst>(Inst)) {
    if (SI->getPointerOperand() == OperandVal)
      isAddress = true;
  } else if (IntrinsicInst *II = dyn_cast<IntrinsicInst>(Inst)) {
    // Addressing modes can also be folded into prefetches and a variety
    // of intrinsics.
    switch (II->getIntrinsicID()) {
    case Intrinsic::memset:
    case Intrinsic::prefetch:
      if (II->getArgOperand(0) == OperandVal)
        isAddress = true;
      break;
    case Intrinsic::memmove:
    case Intrinsic::memcpy:
      if (II->getArgOperand(0) == OperandVal ||
          II->getArgOperand(1) == OperandVal)
        isAddress = true;
      break;
    default: {
      MemIntrinsicInfo IntrInfo;
      if (TTI.getTgtMemIntrinsic(II, IntrInfo)) {
        if (IntrInfo.PtrVal == OperandVal)
          isAddress = true;
      }
    }
    }
  } else if (AtomicRMWInst *RMW = dyn_cast<AtomicRMWInst>(Inst)) {
    if (RMW->getPointerOperand() == OperandVal)
      isAddress = true;
  } else if (AtomicCmpXchgInst *CmpX = dyn_cast<AtomicCmpXchgInst>(Inst)) {
    if (CmpX->getPointerOperand() == OperandVal)
      isAddress = true;
  }
  return isAddress;
}

/// Return the type of the memory being accessed.
static MemAccessTy getAccessType(const TargetTransformInfo &TTI,
                                 Instruction *Inst) {
  MemAccessTy AccessTy(Inst->getType(), MemAccessTy::UnknownAddressSpace);
  if (const StoreInst *SI = dyn_cast<StoreInst>(Inst)) {
    AccessTy.MemTy = SI->getOperand(0)->getType();
    AccessTy.AddrSpace = SI->getPointerAddressSpace();
  } else if (const LoadInst *LI = dyn_cast<LoadInst>(Inst)) {
    AccessTy.AddrSpace = LI->getPointerAddressSpace();
  } else if (const AtomicRMWInst *RMW = dyn_cast<AtomicRMWInst>(Inst)) {
    AccessTy.AddrSpace = RMW->getPointerAddressSpace();
  } else if (const AtomicCmpXchgInst *CmpX = dyn_cast<AtomicCmpXchgInst>(Inst)) {
    AccessTy.AddrSpace = CmpX->getPointerAddressSpace();
  } else if (IntrinsicInst *II = dyn_cast<IntrinsicInst>(Inst)) {
    switch (II->getIntrinsicID()) {
    case Intrinsic::prefetch:
      AccessTy.AddrSpace = II->getArgOperand(0)->getType()->getPointerAddressSpace();
      break;
    default: {
      MemIntrinsicInfo IntrInfo;
      if (TTI.getTgtMemIntrinsic(II, IntrInfo) && IntrInfo.PtrVal) {
        AccessTy.AddrSpace
          = IntrInfo.PtrVal->getType()->getPointerAddressSpace();
      }

      break;
    }
    }
  }

  // All pointers have the same requirements, so canonicalize them to an
  // arbitrary pointer type to minimize variation.
  if (PointerType *PTy = dyn_cast<PointerType>(AccessTy.MemTy))
    AccessTy.MemTy = PointerType::get(IntegerType::get(PTy->getContext(), 1),
                                      PTy->getAddressSpace());

  return AccessTy;
}

/// Return true if this AddRec is already a phi in its loop.
static bool isExistingPhi(const SCEVAddRecExpr *AR, ScalarEvolution &SE) {
  for (PHINode &PN : AR->getLoop()->getHeader()->phis()) {
    if (SE.isSCEVable(PN.getType()) &&
        (SE.getEffectiveSCEVType(PN.getType()) ==
         SE.getEffectiveSCEVType(AR->getType())) &&
        SE.getSCEV(&PN) == AR)
      return true;
  }
  return false;
}

/// Check if expanding this expression is likely to incur significant cost. This
/// is tricky because SCEV doesn't track which expressions are actually computed
/// by the current IR.
///
/// We currently allow expansion of IV increments that involve adds,
/// multiplication by constants, and AddRecs from existing phis.
///
/// TODO: Allow UDivExpr if we can find an existing IV increment that is an
/// obvious multiple of the UDivExpr.
static bool isHighCostExpansion(const SCEV *S,
                                SmallPtrSetImpl<const SCEV*> &Processed,
                                ScalarEvolution &SE) {
  // Zero/One operand expressions
  switch (S->getSCEVType()) {
  case scUnknown:
  case scConstant:
    return false;
  case scTruncate:
    return isHighCostExpansion(cast<SCEVTruncateExpr>(S)->getOperand(),
                               Processed, SE);
  case scZeroExtend:
    return isHighCostExpansion(cast<SCEVZeroExtendExpr>(S)->getOperand(),
                               Processed, SE);
  case scSignExtend:
    return isHighCostExpansion(cast<SCEVSignExtendExpr>(S)->getOperand(),
                               Processed, SE);
  }

  if (!Processed.insert(S).second)
    return false;

  if (const SCEVAddExpr *Add = dyn_cast<SCEVAddExpr>(S)) {
    for (const SCEV *S : Add->operands()) {
      if (isHighCostExpansion(S, Processed, SE))
        return true;
    }
    return false;
  }

  if (const SCEVMulExpr *Mul = dyn_cast<SCEVMulExpr>(S)) {
    if (Mul->getNumOperands() == 2) {
      // Multiplication by a constant is ok
      if (isa<SCEVConstant>(Mul->getOperand(0)))
        return isHighCostExpansion(Mul->getOperand(1), Processed, SE);

      // If we have the value of one operand, check if an existing
      // multiplication already generates this expression.
      if (const SCEVUnknown *U = dyn_cast<SCEVUnknown>(Mul->getOperand(1))) {
        Value *UVal = U->getValue();
        for (User *UR : UVal->users()) {
          // If U is a constant, it may be used by a ConstantExpr.
          Instruction *UI = dyn_cast<Instruction>(UR);
          if (UI && UI->getOpcode() == Instruction::Mul &&
              SE.isSCEVable(UI->getType())) {
            return SE.getSCEV(UI) == Mul;
          }
        }
      }
    }
  }

  if (const SCEVAddRecExpr *AR = dyn_cast<SCEVAddRecExpr>(S)) {
    if (isExistingPhi(AR, SE))
      return false;
  }

  // Fow now, consider any other type of expression (div/mul/min/max) high cost.
  return true;
}

/// If any of the instructions is the specified set are trivially dead, delete
/// them and see if this makes any of their operands subsequently dead.
static bool
DeleteTriviallyDeadInstructions(SmallVectorImpl<WeakTrackingVH> &DeadInsts) {
  bool Changed = false;

  while (!DeadInsts.empty()) {
    Value *V = DeadInsts.pop_back_val();
    Instruction *I = dyn_cast_or_null<Instruction>(V);

    if (!I || !isInstructionTriviallyDead(I))
      continue;

    for (Use &O : I->operands())
      if (Instruction *U = dyn_cast<Instruction>(O)) {
        O = nullptr;
        if (U->use_empty())
          DeadInsts.emplace_back(U);
      }

    I->eraseFromParent();
    Changed = true;
  }

  return Changed;
}

namespace {

class LSRUse;

} // end anonymous namespace

/// \brief Check if the addressing mode defined by \p F is completely
/// folded in \p LU at isel time.
/// This includes address-mode folding and special icmp tricks.
/// This function returns true if \p LU can accommodate what \p F
/// defines and up to 1 base + 1 scaled + offset.
/// In other words, if \p F has several base registers, this function may
/// still return true. Therefore, users still need to account for
/// additional base registers and/or unfolded offsets to derive an
/// accurate cost model.
static bool isAMCompletelyFolded(const TargetTransformInfo &TTI,
                                 const LSRUse &LU, const Formula &F);

// Get the cost of the scaling factor used in F for LU.
static unsigned getScalingFactorCost(const TargetTransformInfo &TTI,
                                     const LSRUse &LU, const Formula &F,
                                     const Loop &L);

namespace {

/// This class is used to measure and compare candidate formulae.
class Cost {
  TargetTransformInfo::LSRCost C;

public:
  Cost() {
    C.Insns = 0;
    C.NumRegs = 0;
    C.AddRecCost = 0;
    C.NumIVMuls = 0;
    C.NumBaseAdds = 0;
    C.ImmCost = 0;
    C.SetupCost = 0;
    C.ScaleCost = 0;
  }

  bool isLess(Cost &Other, const TargetTransformInfo &TTI);

  void Lose();

#ifndef NDEBUG
  // Once any of the metrics loses, they must all remain losers.
  bool isValid() {
    return ((C.Insns | C.NumRegs | C.AddRecCost | C.NumIVMuls | C.NumBaseAdds
             | C.ImmCost | C.SetupCost | C.ScaleCost) != ~0u)
      || ((C.Insns & C.NumRegs & C.AddRecCost & C.NumIVMuls & C.NumBaseAdds
           & C.ImmCost & C.SetupCost & C.ScaleCost) == ~0u);
  }
#endif

  bool isLoser() {
    assert(isValid() && "invalid cost");
    return C.NumRegs == ~0u;
  }

  void RateFormula(const TargetTransformInfo &TTI,
                   const Formula &F,
                   SmallPtrSetImpl<const SCEV *> &Regs,
                   const DenseSet<const SCEV *> &VisitedRegs,
                   const Loop *L,
                   ScalarEvolution &SE, DominatorTree &DT,
                   const LSRUse &LU,
                   SmallPtrSetImpl<const SCEV *> *LoserRegs = nullptr);

  void print(raw_ostream &OS) const;
  void dump() const;

private:
  void RateRegister(const SCEV *Reg,
                    SmallPtrSetImpl<const SCEV *> &Regs,
                    const Loop *L,
                    ScalarEvolution &SE, DominatorTree &DT);
  void RatePrimaryRegister(const SCEV *Reg,
                           SmallPtrSetImpl<const SCEV *> &Regs,
                           const Loop *L,
                           ScalarEvolution &SE, DominatorTree &DT,
                           SmallPtrSetImpl<const SCEV *> *LoserRegs);
};

/// An operand value in an instruction which is to be replaced with some
/// equivalent, possibly strength-reduced, replacement.
struct LSRFixup {
  /// The instruction which will be updated.
  Instruction *UserInst = nullptr;

  /// The operand of the instruction which will be replaced. The operand may be
  /// used more than once; every instance will be replaced.
  Value *OperandValToReplace = nullptr;

  /// If this user is to use the post-incremented value of an induction
  /// variable, this set is non-empty and holds the loops associated with the
  /// induction variable.
  PostIncLoopSet PostIncLoops;

  /// A constant offset to be added to the LSRUse expression.  This allows
  /// multiple fixups to share the same LSRUse with different offsets, for
  /// example in an unrolled loop.
  int64_t Offset = 0;

  LSRFixup() = default;

  bool isUseFullyOutsideLoop(const Loop *L) const;

  void print(raw_ostream &OS) const;
  void dump() const;
};

/// A DenseMapInfo implementation for holding DenseMaps and DenseSets of sorted
/// SmallVectors of const SCEV*.
struct UniquifierDenseMapInfo {
  static SmallVector<const SCEV *, 4> getEmptyKey() {
    SmallVector<const SCEV *, 4>  V;
    V.push_back(reinterpret_cast<const SCEV *>(-1));
    return V;
  }

  static SmallVector<const SCEV *, 4> getTombstoneKey() {
    SmallVector<const SCEV *, 4> V;
    V.push_back(reinterpret_cast<const SCEV *>(-2));
    return V;
  }

  static unsigned getHashValue(const SmallVector<const SCEV *, 4> &V) {
    return static_cast<unsigned>(hash_combine_range(V.begin(), V.end()));
  }

  static bool isEqual(const SmallVector<const SCEV *, 4> &LHS,
                      const SmallVector<const SCEV *, 4> &RHS) {
    return LHS == RHS;
  }
};

/// This class holds the state that LSR keeps for each use in IVUsers, as well
/// as uses invented by LSR itself. It includes information about what kinds of
/// things can be folded into the user, information about the user itself, and
/// information about how the use may be satisfied.  TODO: Represent multiple
/// users of the same expression in common?
class LSRUse {
  DenseSet<SmallVector<const SCEV *, 4>, UniquifierDenseMapInfo> Uniquifier;

public:
  /// An enum for a kind of use, indicating what types of scaled and immediate
  /// operands it might support.
  enum KindType {
    Basic,   ///< A normal use, with no folding.
    Special, ///< A special case of basic, allowing -1 scales.
    Address, ///< An address use; folding according to TargetLowering
    ICmpZero ///< An equality icmp with both operands folded into one.
    // TODO: Add a generic icmp too?
  };

  using SCEVUseKindPair = PointerIntPair<const SCEV *, 2, KindType>;

  KindType Kind;
  MemAccessTy AccessTy;

  /// The list of operands which are to be replaced.
  SmallVector<LSRFixup, 8> Fixups;

  /// Keep track of the min and max offsets of the fixups.
  int64_t MinOffset = std::numeric_limits<int64_t>::max();
  int64_t MaxOffset = std::numeric_limits<int64_t>::min();

  /// This records whether all of the fixups using this LSRUse are outside of
  /// the loop, in which case some special-case heuristics may be used.
  bool AllFixupsOutsideLoop = true;

  /// RigidFormula is set to true to guarantee that this use will be associated
  /// with a single formula--the one that initially matched. Some SCEV
  /// expressions cannot be expanded. This allows LSR to consider the registers
  /// used by those expressions without the need to expand them later after
  /// changing the formula.
  bool RigidFormula = false;

  /// This records the widest use type for any fixup using this
  /// LSRUse. FindUseWithSimilarFormula can't consider uses with different max
  /// fixup widths to be equivalent, because the narrower one may be relying on
  /// the implicit truncation to truncate away bogus bits.
  Type *WidestFixupType = nullptr;

  /// A list of ways to build a value that can satisfy this user.  After the
  /// list is populated, one of these is selected heuristically and used to
  /// formulate a replacement for OperandValToReplace in UserInst.
  SmallVector<Formula, 12> Formulae;

  /// The set of register candidates used by all formulae in this LSRUse.
  SmallPtrSet<const SCEV *, 4> Regs;

  LSRUse(KindType K, MemAccessTy AT) : Kind(K), AccessTy(AT) {}

  LSRFixup &getNewFixup() {
    Fixups.push_back(LSRFixup());
    return Fixups.back();
  }

  void pushFixup(LSRFixup &f) {
    Fixups.push_back(f);
    if (f.Offset > MaxOffset)
      MaxOffset = f.Offset;
    if (f.Offset < MinOffset)
      MinOffset = f.Offset;
  }

  bool HasFormulaWithSameRegs(const Formula &F) const;
  float getNotSelectedProbability(const SCEV *Reg) const;
  bool InsertFormula(const Formula &F, const Loop &L);
  void DeleteFormula(Formula &F);
  void RecomputeRegs(size_t LUIdx, RegUseTracker &Reguses);

  void print(raw_ostream &OS) const;
  void dump() const;
};

} // end anonymous namespace

static bool isAMCompletelyFolded(const TargetTransformInfo &TTI,
                                 LSRUse::KindType Kind, MemAccessTy AccessTy,
                                 GlobalValue *BaseGV, int64_t BaseOffset,
                                 bool HasBaseReg, int64_t Scale,
                                 Instruction *Fixup = nullptr);

/// Tally up interesting quantities from the given register.
void Cost::RateRegister(const SCEV *Reg,
                        SmallPtrSetImpl<const SCEV *> &Regs,
                        const Loop *L,
                        ScalarEvolution &SE, DominatorTree &DT) {
  if (const SCEVAddRecExpr *AR = dyn_cast<SCEVAddRecExpr>(Reg)) {
    // If this is an addrec for another loop, it should be an invariant
    // with respect to L since L is the innermost loop (at least
    // for now LSR only handles innermost loops).
    if (AR->getLoop() != L) {
      // If the AddRec exists, consider it's register free and leave it alone.
      if (isExistingPhi(AR, SE))
        return;

      // It is bad to allow LSR for current loop to add induction variables
      // for its sibling loops.
      if (!AR->getLoop()->contains(L)) {
        Lose();
        return;
      }

      // Otherwise, it will be an invariant with respect to Loop L.
      ++C.NumRegs;
      return;
    }
    C.AddRecCost += 1; /// TODO: This should be a function of the stride.

    // Add the step value register, if it needs one.
    // TODO: The non-affine case isn't precisely modeled here.
    if (!AR->isAffine() || !isa<SCEVConstant>(AR->getOperand(1))) {
      if (!Regs.count(AR->getOperand(1))) {
        RateRegister(AR->getOperand(1), Regs, L, SE, DT);
        if (isLoser())
          return;
      }
    }
  }
  ++C.NumRegs;

  // Rough heuristic; favor registers which don't require extra setup
  // instructions in the preheader.
  if (!isa<SCEVUnknown>(Reg) &&
      !isa<SCEVConstant>(Reg) &&
      !(isa<SCEVAddRecExpr>(Reg) &&
        (isa<SCEVUnknown>(cast<SCEVAddRecExpr>(Reg)->getStart()) ||
         isa<SCEVConstant>(cast<SCEVAddRecExpr>(Reg)->getStart()))))
    ++C.SetupCost;

  C.NumIVMuls += isa<SCEVMulExpr>(Reg) &&
               SE.hasComputableLoopEvolution(Reg, L);
}

/// Record this register in the set. If we haven't seen it before, rate
/// it. Optional LoserRegs provides a way to declare any formula that refers to
/// one of those regs an instant loser.
void Cost::RatePrimaryRegister(const SCEV *Reg,
                               SmallPtrSetImpl<const SCEV *> &Regs,
                               const Loop *L,
                               ScalarEvolution &SE, DominatorTree &DT,
                               SmallPtrSetImpl<const SCEV *> *LoserRegs) {
  if (LoserRegs && LoserRegs->count(Reg)) {
    Lose();
    return;
  }
  if (Regs.insert(Reg).second) {
    RateRegister(Reg, Regs, L, SE, DT);
    if (LoserRegs && isLoser())
      LoserRegs->insert(Reg);
  }
}

void Cost::RateFormula(const TargetTransformInfo &TTI,
                       const Formula &F,
                       SmallPtrSetImpl<const SCEV *> &Regs,
                       const DenseSet<const SCEV *> &VisitedRegs,
                       const Loop *L,
                       ScalarEvolution &SE, DominatorTree &DT,
                       const LSRUse &LU,
                       SmallPtrSetImpl<const SCEV *> *LoserRegs) {
  assert(F.isCanonical(*L) && "Cost is accurate only for canonical formula");
  // Tally up the registers.
  unsigned PrevAddRecCost = C.AddRecCost;
  unsigned PrevNumRegs = C.NumRegs;
  unsigned PrevNumBaseAdds = C.NumBaseAdds;
  if (const SCEV *ScaledReg = F.ScaledReg) {
    if (VisitedRegs.count(ScaledReg)) {
      Lose();
      return;
    }
    RatePrimaryRegister(ScaledReg, Regs, L, SE, DT, LoserRegs);
    if (isLoser())
      return;
  }
  for (const SCEV *BaseReg : F.BaseRegs) {
    if (VisitedRegs.count(BaseReg)) {
      Lose();
      return;
    }
    RatePrimaryRegister(BaseReg, Regs, L, SE, DT, LoserRegs);
    if (isLoser())
      return;
  }

  // Determine how many (unfolded) adds we'll need inside the loop.
  size_t NumBaseParts = F.getNumRegs();
  if (NumBaseParts > 1)
    // Do not count the base and a possible second register if the target
    // allows to fold 2 registers.
    C.NumBaseAdds +=
        NumBaseParts - (1 + (F.Scale && isAMCompletelyFolded(TTI, LU, F)));
  C.NumBaseAdds += (F.UnfoldedOffset != 0);

  // Accumulate non-free scaling amounts.
  C.ScaleCost += getScalingFactorCost(TTI, LU, F, *L);

  // Tally up the non-zero immediates.
  for (const LSRFixup &Fixup : LU.Fixups) {
    int64_t O = Fixup.Offset;
    int64_t Offset = (uint64_t)O + F.BaseOffset;
    if (F.BaseGV)
      C.ImmCost += 64; // Handle symbolic values conservatively.
                     // TODO: This should probably be the pointer size.
    else if (Offset != 0)
      C.ImmCost += APInt(64, Offset, true).getMinSignedBits();

    // Check with target if this offset with this instruction is
    // specifically not supported.
    if (LU.Kind == LSRUse::Address && Offset != 0 &&
        !isAMCompletelyFolded(TTI, LSRUse::Address, LU.AccessTy, F.BaseGV,
                              Offset, F.HasBaseReg, F.Scale, Fixup.UserInst))
      C.NumBaseAdds++;
  }

  // If we don't count instruction cost exit here.
  if (!InsnsCost) {
    assert(isValid() && "invalid cost");
    return;
  }

  // Treat every new register that exceeds TTI.getNumberOfRegisters() - 1 as
  // additional instruction (at least fill).
  unsigned TTIRegNum = TTI.getNumberOfRegisters(false) - 1;
  if (C.NumRegs > TTIRegNum) {
    // Cost already exceeded TTIRegNum, then only newly added register can add
    // new instructions.
    if (PrevNumRegs > TTIRegNum)
      C.Insns += (C.NumRegs - PrevNumRegs);
    else
      C.Insns += (C.NumRegs - TTIRegNum);
  }

  // If ICmpZero formula ends with not 0, it could not be replaced by
  // just add or sub. We'll need to compare final result of AddRec.
  // That means we'll need an additional instruction.
  // For -10 + {0, +, 1}:
  // i = i + 1;
  // cmp i, 10
  //
  // For {-10, +, 1}:
  // i = i + 1;
  if (LU.Kind == LSRUse::ICmpZero && !F.hasZeroEnd())
    C.Insns++;
  // Each new AddRec adds 1 instruction to calculation.
  C.Insns += (C.AddRecCost - PrevAddRecCost);

  // BaseAdds adds instructions for unfolded registers.
  if (LU.Kind != LSRUse::ICmpZero)
    C.Insns += C.NumBaseAdds - PrevNumBaseAdds;
  assert(isValid() && "invalid cost");
}

/// Set this cost to a losing value.
void Cost::Lose() {
  C.Insns = std::numeric_limits<unsigned>::max();
  C.NumRegs = std::numeric_limits<unsigned>::max();
  C.AddRecCost = std::numeric_limits<unsigned>::max();
  C.NumIVMuls = std::numeric_limits<unsigned>::max();
  C.NumBaseAdds = std::numeric_limits<unsigned>::max();
  C.ImmCost = std::numeric_limits<unsigned>::max();
  C.SetupCost = std::numeric_limits<unsigned>::max();
  C.ScaleCost = std::numeric_limits<unsigned>::max();
}

/// Choose the lower cost.
bool Cost::isLess(Cost &Other, const TargetTransformInfo &TTI) {
  if (InsnsCost.getNumOccurrences() > 0 && InsnsCost &&
      C.Insns != Other.C.Insns)
    return C.Insns < Other.C.Insns;
  return TTI.isLSRCostLess(C, Other.C);
}

#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
void Cost::print(raw_ostream &OS) const {
  if (InsnsCost)
    OS << C.Insns << " instruction" << (C.Insns == 1 ? " " : "s ");
  OS << C.NumRegs << " reg" << (C.NumRegs == 1 ? "" : "s");
  if (C.AddRecCost != 0)
    OS << ", with addrec cost " << C.AddRecCost;
  if (C.NumIVMuls != 0)
    OS << ", plus " << C.NumIVMuls << " IV mul"
       << (C.NumIVMuls == 1 ? "" : "s");
  if (C.NumBaseAdds != 0)
    OS << ", plus " << C.NumBaseAdds << " base add"
       << (C.NumBaseAdds == 1 ? "" : "s");
  if (C.ScaleCost != 0)
    OS << ", plus " << C.ScaleCost << " scale cost";
  if (C.ImmCost != 0)
    OS << ", plus " << C.ImmCost << " imm cost";
  if (C.SetupCost != 0)
    OS << ", plus " << C.SetupCost << " setup cost";
}

LLVM_DUMP_METHOD void Cost::dump() const {
  print(errs()); errs() << '\n';
}
#endif

/// Test whether this fixup always uses its value outside of the given loop.
bool LSRFixup::isUseFullyOutsideLoop(const Loop *L) const {
  // PHI nodes use their value in their incoming blocks.
  if (const PHINode *PN = dyn_cast<PHINode>(UserInst)) {
    for (unsigned i = 0, e = PN->getNumIncomingValues(); i != e; ++i)
      if (PN->getIncomingValue(i) == OperandValToReplace &&
          L->contains(PN->getIncomingBlock(i)))
        return false;
    return true;
  }

  return !L->contains(UserInst);
}

#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
void LSRFixup::print(raw_ostream &OS) const {
  OS << "UserInst=";
  // Store is common and interesting enough to be worth special-casing.
  if (StoreInst *Store = dyn_cast<StoreInst>(UserInst)) {
    OS << "store ";
    Store->getOperand(0)->printAsOperand(OS, /*PrintType=*/false);
  } else if (UserInst->getType()->isVoidTy())
    OS << UserInst->getOpcodeName();
  else
    UserInst->printAsOperand(OS, /*PrintType=*/false);

  OS << ", OperandValToReplace=";
  OperandValToReplace->printAsOperand(OS, /*PrintType=*/false);

  for (const Loop *PIL : PostIncLoops) {
    OS << ", PostIncLoop=";
    PIL->getHeader()->printAsOperand(OS, /*PrintType=*/false);
  }

  if (Offset != 0)
    OS << ", Offset=" << Offset;
}

LLVM_DUMP_METHOD void LSRFixup::dump() const {
  print(errs()); errs() << '\n';
}
#endif

/// Test whether this use as a formula which has the same registers as the given
/// formula.
bool LSRUse::HasFormulaWithSameRegs(const Formula &F) const {
  SmallVector<const SCEV *, 4> Key = F.BaseRegs;
  if (F.ScaledReg) Key.push_back(F.ScaledReg);
  // Unstable sort by host order ok, because this is only used for uniquifying.
  std::sort(Key.begin(), Key.end());
  return Uniquifier.count(Key);
}

/// The function returns a probability of selecting formula without Reg.
float LSRUse::getNotSelectedProbability(const SCEV *Reg) const {
  unsigned FNum = 0;
  for (const Formula &F : Formulae)
    if (F.referencesReg(Reg))
      FNum++;
  return ((float)(Formulae.size() - FNum)) / Formulae.size();
}

/// If the given formula has not yet been inserted, add it to the list, and
/// return true. Return false otherwise.  The formula must be in canonical form.
bool LSRUse::InsertFormula(const Formula &F, const Loop &L) {
  assert(F.isCanonical(L) && "Invalid canonical representation");

  if (!Formulae.empty() && RigidFormula)
    return false;

  SmallVector<const SCEV *, 4> Key = F.BaseRegs;
  if (F.ScaledReg) Key.push_back(F.ScaledReg);
  // Unstable sort by host order ok, because this is only used for uniquifying.
  std::sort(Key.begin(), Key.end());

  if (!Uniquifier.insert(Key).second)
    return false;

  // Using a register to hold the value of 0 is not profitable.
  assert((!F.ScaledReg || !F.ScaledReg->isZero()) &&
         "Zero allocated in a scaled register!");
#ifndef NDEBUG
  for (const SCEV *BaseReg : F.BaseRegs)
    assert(!BaseReg->isZero() && "Zero allocated in a base register!");
#endif

  // Add the formula to the list.
  Formulae.push_back(F);

  // Record registers now being used by this use.
  Regs.insert(F.BaseRegs.begin(), F.BaseRegs.end());
  if (F.ScaledReg)
    Regs.insert(F.ScaledReg);

  return true;
}

/// Remove the given formula from this use's list.
void LSRUse::DeleteFormula(Formula &F) {
  if (&F != &Formulae.back())
    std::swap(F, Formulae.back());
  Formulae.pop_back();
}

/// Recompute the Regs field, and update RegUses.
void LSRUse::RecomputeRegs(size_t LUIdx, RegUseTracker &RegUses) {
  // Now that we've filtered out some formulae, recompute the Regs set.
  SmallPtrSet<const SCEV *, 4> OldRegs = std::move(Regs);
  Regs.clear();
  for (const Formula &F : Formulae) {
    if (F.ScaledReg) Regs.insert(F.ScaledReg);
    Regs.insert(F.BaseRegs.begin(), F.BaseRegs.end());
  }

  // Update the RegTracker.
  for (const SCEV *S : OldRegs)
    if (!Regs.count(S))
      RegUses.dropRegister(S, LUIdx);
}

#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
void LSRUse::print(raw_ostream &OS) const {
  OS << "LSR Use: Kind=";
  switch (Kind) {
  case Basic:    OS << "Basic"; break;
  case Special:  OS << "Special"; break;
  case ICmpZero: OS << "ICmpZero"; break;
  case Address:
    OS << "Address of ";
    if (AccessTy.MemTy->isPointerTy())
      OS << "pointer"; // the full pointer type could be really verbose
    else {
      OS << *AccessTy.MemTy;
    }

    OS << " in addrspace(" << AccessTy.AddrSpace << ')';
  }

  OS << ", Offsets={";
  bool NeedComma = false;
  for (const LSRFixup &Fixup : Fixups) {
    if (NeedComma) OS << ',';
    OS << Fixup.Offset;
    NeedComma = true;
  }
  OS << '}';

  if (AllFixupsOutsideLoop)
    OS << ", all-fixups-outside-loop";

  if (WidestFixupType)
    OS << ", widest fixup type: " << *WidestFixupType;
}

LLVM_DUMP_METHOD void LSRUse::dump() const {
  print(errs()); errs() << '\n';
}
#endif

static bool isAMCompletelyFolded(const TargetTransformInfo &TTI,
                                 LSRUse::KindType Kind, MemAccessTy AccessTy,
                                 GlobalValue *BaseGV, int64_t BaseOffset,
                                 bool HasBaseReg, int64_t Scale,
                                 Instruction *Fixup/*= nullptr*/) {
  switch (Kind) {
  case LSRUse::Address:
    return TTI.isLegalAddressingMode(AccessTy.MemTy, BaseGV, BaseOffset,
                                     HasBaseReg, Scale, AccessTy.AddrSpace, Fixup);

  case LSRUse::ICmpZero:
    // There's not even a target hook for querying whether it would be legal to
    // fold a GV into an ICmp.
    if (BaseGV)
      return false;

    // ICmp only has two operands; don't allow more than two non-trivial parts.
    if (Scale != 0 && HasBaseReg && BaseOffset != 0)
      return false;

    // ICmp only supports no scale or a -1 scale, as we can "fold" a -1 scale by
    // putting the scaled register in the other operand of the icmp.
    if (Scale != 0 && Scale != -1)
      return false;

    // If we have low-level target information, ask the target if it can fold an
    // integer immediate on an icmp.
    if (BaseOffset != 0) {
      // We have one of:
      // ICmpZero     BaseReg + BaseOffset => ICmp BaseReg, -BaseOffset
      // ICmpZero -1*ScaleReg + BaseOffset => ICmp ScaleReg, BaseOffset
      // Offs is the ICmp immediate.
      if (Scale == 0)
        // The cast does the right thing with
        // std::numeric_limits<int64_t>::min().
        BaseOffset = -(uint64_t)BaseOffset;
      return TTI.isLegalICmpImmediate(BaseOffset);
    }

    // ICmpZero BaseReg + -1*ScaleReg => ICmp BaseReg, ScaleReg
    return true;

  case LSRUse::Basic:
    // Only handle single-register values.
    return !BaseGV && Scale == 0 && BaseOffset == 0;

  case LSRUse::Special:
    // Special case Basic to handle -1 scales.
    return !BaseGV && (Scale == 0 || Scale == -1) && BaseOffset == 0;
  }

  llvm_unreachable("Invalid LSRUse Kind!");
}

static bool isAMCompletelyFolded(const TargetTransformInfo &TTI,
                                 int64_t MinOffset, int64_t MaxOffset,
                                 LSRUse::KindType Kind, MemAccessTy AccessTy,
                                 GlobalValue *BaseGV, int64_t BaseOffset,
                                 bool HasBaseReg, int64_t Scale) {
  // Check for overflow.
  if (((int64_t)((uint64_t)BaseOffset + MinOffset) > BaseOffset) !=
      (MinOffset > 0))
    return false;
  MinOffset = (uint64_t)BaseOffset + MinOffset;
  if (((int64_t)((uint64_t)BaseOffset + MaxOffset) > BaseOffset) !=
      (MaxOffset > 0))
    return false;
  MaxOffset = (uint64_t)BaseOffset + MaxOffset;

  return isAMCompletelyFolded(TTI, Kind, AccessTy, BaseGV, MinOffset,
                              HasBaseReg, Scale) &&
         isAMCompletelyFolded(TTI, Kind, AccessTy, BaseGV, MaxOffset,
                              HasBaseReg, Scale);
}

static bool isAMCompletelyFolded(const TargetTransformInfo &TTI,
                                 int64_t MinOffset, int64_t MaxOffset,
                                 LSRUse::KindType Kind, MemAccessTy AccessTy,
                                 const Formula &F, const Loop &L) {
  // For the purpose of isAMCompletelyFolded either having a canonical formula
  // or a scale not equal to zero is correct.
  // Problems may arise from non canonical formulae having a scale == 0.
  // Strictly speaking it would best to just rely on canonical formulae.
  // However, when we generate the scaled formulae, we first check that the
  // scaling factor is profitable before computing the actual ScaledReg for
  // compile time sake.
  assert((F.isCanonical(L) || F.Scale != 0));
  return isAMCompletelyFolded(TTI, MinOffset, MaxOffset, Kind, AccessTy,
                              F.BaseGV, F.BaseOffset, F.HasBaseReg, F.Scale);
}

/// Test whether we know how to expand the current formula.
static bool isLegalUse(const TargetTransformInfo &TTI, int64_t MinOffset,
                       int64_t MaxOffset, LSRUse::KindType Kind,
                       MemAccessTy AccessTy, GlobalValue *BaseGV,
                       int64_t BaseOffset, bool HasBaseReg, int64_t Scale) {
  // We know how to expand completely foldable formulae.
  return isAMCompletelyFolded(TTI, MinOffset, MaxOffset, Kind, AccessTy, BaseGV,
                              BaseOffset, HasBaseReg, Scale) ||
         // Or formulae that use a base register produced by a sum of base
         // registers.
         (Scale == 1 &&
          isAMCompletelyFolded(TTI, MinOffset, MaxOffset, Kind, AccessTy,
                               BaseGV, BaseOffset, true, 0));
}

static bool isLegalUse(const TargetTransformInfo &TTI, int64_t MinOffset,
                       int64_t MaxOffset, LSRUse::KindType Kind,
                       MemAccessTy AccessTy, const Formula &F) {
  return isLegalUse(TTI, MinOffset, MaxOffset, Kind, AccessTy, F.BaseGV,
                    F.BaseOffset, F.HasBaseReg, F.Scale);
}

static bool isAMCompletelyFolded(const TargetTransformInfo &TTI,
                                 const LSRUse &LU, const Formula &F) {
  // Target may want to look at the user instructions.
  if (LU.Kind == LSRUse::Address && TTI.LSRWithInstrQueries()) {
    for (const LSRFixup &Fixup : LU.Fixups)
      if (!isAMCompletelyFolded(TTI, LSRUse::Address, LU.AccessTy, F.BaseGV,
                                (F.BaseOffset + Fixup.Offset), F.HasBaseReg,
                                F.Scale, Fixup.UserInst))
        return false;
    return true;
  }

  return isAMCompletelyFolded(TTI, LU.MinOffset, LU.MaxOffset, LU.Kind,
                              LU.AccessTy, F.BaseGV, F.BaseOffset, F.HasBaseReg,
                              F.Scale);
}

static unsigned getScalingFactorCost(const TargetTransformInfo &TTI,
                                     const LSRUse &LU, const Formula &F,
                                     const Loop &L) {
  if (!F.Scale)
    return 0;

  // If the use is not completely folded in that instruction, we will have to
  // pay an extra cost only for scale != 1.
  if (!isAMCompletelyFolded(TTI, LU.MinOffset, LU.MaxOffset, LU.Kind,
                            LU.AccessTy, F, L))
    return F.Scale != 1;

  switch (LU.Kind) {
  case LSRUse::Address: {
    // Check the scaling factor cost with both the min and max offsets.
    int ScaleCostMinOffset = TTI.getScalingFactorCost(
        LU.AccessTy.MemTy, F.BaseGV, F.BaseOffset + LU.MinOffset, F.HasBaseReg,
        F.Scale, LU.AccessTy.AddrSpace);
    int ScaleCostMaxOffset = TTI.getScalingFactorCost(
        LU.AccessTy.MemTy, F.BaseGV, F.BaseOffset + LU.MaxOffset, F.HasBaseReg,
        F.Scale, LU.AccessTy.AddrSpace);

    assert(ScaleCostMinOffset >= 0 && ScaleCostMaxOffset >= 0 &&
           "Legal addressing mode has an illegal cost!");
    return std::max(ScaleCostMinOffset, ScaleCostMaxOffset);
  }
  case LSRUse::ICmpZero:
  case LSRUse::Basic:
  case LSRUse::Special:
    // The use is completely folded, i.e., everything is folded into the
    // instruction.
    return 0;
  }

  llvm_unreachable("Invalid LSRUse Kind!");
}

static bool isAlwaysFoldable(const TargetTransformInfo &TTI,
                             LSRUse::KindType Kind, MemAccessTy AccessTy,
                             GlobalValue *BaseGV, int64_t BaseOffset,
                             bool HasBaseReg) {
  // Fast-path: zero is always foldable.
  if (BaseOffset == 0 && !BaseGV) return true;

  // Conservatively, create an address with an immediate and a
  // base and a scale.
  int64_t Scale = Kind == LSRUse::ICmpZero ? -1 : 1;

  // Canonicalize a scale of 1 to a base register if the formula doesn't
  // already have a base register.
  if (!HasBaseReg && Scale == 1) {
    Scale = 0;
    HasBaseReg = true;
  }

  return isAMCompletelyFolded(TTI, Kind, AccessTy, BaseGV, BaseOffset,
                              HasBaseReg, Scale);
}

static bool isAlwaysFoldable(const TargetTransformInfo &TTI,
                             ScalarEvolution &SE, int64_t MinOffset,
                             int64_t MaxOffset, LSRUse::KindType Kind,
                             MemAccessTy AccessTy, const SCEV *S,
                             bool HasBaseReg) {
  // Fast-path: zero is always foldable.
  if (S->isZero()) return true;

  // Conservatively, create an address with an immediate and a
  // base and a scale.
  int64_t BaseOffset = ExtractImmediate(S, SE);
  GlobalValue *BaseGV = ExtractSymbol(S, SE);

  // If there's anything else involved, it's not foldable.
  if (!S->isZero()) return false;

  // Fast-path: zero is always foldable.
  if (BaseOffset == 0 && !BaseGV) return true;

  // Conservatively, create an address with an immediate and a
  // base and a scale.
  int64_t Scale = Kind == LSRUse::ICmpZero ? -1 : 1;

  return isAMCompletelyFolded(TTI, MinOffset, MaxOffset, Kind, AccessTy, BaseGV,
                              BaseOffset, HasBaseReg, Scale);
}

namespace {

/// An individual increment in a Chain of IV increments.  Relate an IV user to
/// an expression that computes the IV it uses from the IV used by the previous
/// link in the Chain.
///
/// For the head of a chain, IncExpr holds the absolute SCEV expression for the
/// original IVOperand. The head of the chain's IVOperand is only valid during
/// chain collection, before LSR replaces IV users. During chain generation,
/// IncExpr can be used to find the new IVOperand that computes the same
/// expression.
struct IVInc {
  Instruction *UserInst;
  Value* IVOperand;
  const SCEV *IncExpr;

  IVInc(Instruction *U, Value *O, const SCEV *E)
      : UserInst(U), IVOperand(O), IncExpr(E) {}
};

// The list of IV increments in program order.  We typically add the head of a
// chain without finding subsequent links.
struct IVChain {
  SmallVector<IVInc, 1> Incs;
  const SCEV *ExprBase = nullptr;

  IVChain() = default;
  IVChain(const IVInc &Head, const SCEV *Base)
      : Incs(1, Head), ExprBase(Base) {}

  using const_iterator = SmallVectorImpl<IVInc>::const_iterator;

  // Return the first increment in the chain.
  const_iterator begin() const {
    assert(!Incs.empty());
    return std::next(Incs.begin());
  }
  const_iterator end() const {
    return Incs.end();
  }

  // Returns true if this chain contains any increments.
  bool hasIncs() const { return Incs.size() >= 2; }

  // Add an IVInc to the end of this chain.
  void add(const IVInc &X) { Incs.push_back(X); }

  // Returns the last UserInst in the chain.
  Instruction *tailUserInst() const { return Incs.back().UserInst; }

  // Returns true if IncExpr can be profitably added to this chain.
  bool isProfitableIncrement(const SCEV *OperExpr,
                             const SCEV *IncExpr,
                             ScalarEvolution&);
};

/// Helper for CollectChains to track multiple IV increment uses.  Distinguish
/// between FarUsers that definitely cross IV increments and NearUsers that may
/// be used between IV increments.
struct ChainUsers {
  SmallPtrSet<Instruction*, 4> FarUsers;
  SmallPtrSet<Instruction*, 4> NearUsers;
};

/// This class holds state for the main loop strength reduction logic.
class LSRInstance {
  IVUsers &IU;
  ScalarEvolution &SE;
  DominatorTree &DT;
  LoopInfo &LI;
  const TargetTransformInfo &TTI;
  Loop *const L;
  bool Changed = false;

  /// This is the insert position that the current loop's induction variable
  /// increment should be placed. In simple loops, this is the latch block's
  /// terminator. But in more complicated cases, this is a position which will
  /// dominate all the in-loop post-increment users.
  Instruction *IVIncInsertPos = nullptr;

  /// Interesting factors between use strides.
  ///
  /// We explicitly use a SetVector which contains a SmallSet, instead of the
  /// default, a SmallDenseSet, because we need to use the full range of
  /// int64_ts, and there's currently no good way of doing that with
  /// SmallDenseSet.
  SetVector<int64_t, SmallVector<int64_t, 8>, SmallSet<int64_t, 8>> Factors;

  /// Interesting use types, to facilitate truncation reuse.
  SmallSetVector<Type *, 4> Types;

  /// The list of interesting uses.
  SmallVector<LSRUse, 16> Uses;

  /// Track which uses use which register candidates.
  RegUseTracker RegUses;

  // Limit the number of chains to avoid quadratic behavior. We don't expect to
  // have more than a few IV increment chains in a loop. Missing a Chain falls
  // back to normal LSR behavior for those uses.
  static const unsigned MaxChains = 8;

  /// IV users can form a chain of IV increments.
  SmallVector<IVChain, MaxChains> IVChainVec;

  /// IV users that belong to profitable IVChains.
  SmallPtrSet<Use*, MaxChains> IVIncSet;

  void OptimizeShadowIV();
  bool FindIVUserForCond(ICmpInst *Cond, IVStrideUse *&CondUse);
  ICmpInst *OptimizeMax(ICmpInst *Cond, IVStrideUse* &CondUse);
  void OptimizeLoopTermCond();

  void ChainInstruction(Instruction *UserInst, Instruction *IVOper,
                        SmallVectorImpl<ChainUsers> &ChainUsersVec);
  void FinalizeChain(IVChain &Chain);
  void CollectChains();
  void GenerateIVChain(const IVChain &Chain, SCEVExpander &Rewriter,
                       SmallVectorImpl<WeakTrackingVH> &DeadInsts);

  void CollectInterestingTypesAndFactors();
  void CollectFixupsAndInitialFormulae();

  // Support for sharing of LSRUses between LSRFixups.
  using UseMapTy = DenseMap<LSRUse::SCEVUseKindPair, size_t>;
  UseMapTy UseMap;

  bool reconcileNewOffset(LSRUse &LU, int64_t NewOffset, bool HasBaseReg,
                          LSRUse::KindType Kind, MemAccessTy AccessTy);

  std::pair<size_t, int64_t> getUse(const SCEV *&Expr, LSRUse::KindType Kind,
                                    MemAccessTy AccessTy);

  void DeleteUse(LSRUse &LU, size_t LUIdx);

  LSRUse *FindUseWithSimilarFormula(const Formula &F, const LSRUse &OrigLU);

  void InsertInitialFormula(const SCEV *S, LSRUse &LU, size_t LUIdx);
  void InsertSupplementalFormula(const SCEV *S, LSRUse &LU, size_t LUIdx);
  void CountRegisters(const Formula &F, size_t LUIdx);
  bool InsertFormula(LSRUse &LU, unsigned LUIdx, const Formula &F);

  void CollectLoopInvariantFixupsAndFormulae();

  void GenerateReassociations(LSRUse &LU, unsigned LUIdx, Formula Base,
                              unsigned Depth = 0);

  void GenerateReassociationsImpl(LSRUse &LU, unsigned LUIdx,
                                  const Formula &Base, unsigned Depth,
                                  size_t Idx, bool IsScaledReg = false);
  void GenerateCombinations(LSRUse &LU, unsigned LUIdx, Formula Base);
  void GenerateSymbolicOffsetsImpl(LSRUse &LU, unsigned LUIdx,
                                   const Formula &Base, size_t Idx,
                                   bool IsScaledReg = false);
  void GenerateSymbolicOffsets(LSRUse &LU, unsigned LUIdx, Formula Base);
  void GenerateConstantOffsetsImpl(LSRUse &LU, unsigned LUIdx,
                                   const Formula &Base,
                                   const SmallVectorImpl<int64_t> &Worklist,
                                   size_t Idx, bool IsScaledReg = false);
  void GenerateConstantOffsets(LSRUse &LU, unsigned LUIdx, Formula Base);
  void GenerateICmpZeroScales(LSRUse &LU, unsigned LUIdx, Formula Base);
  void GenerateScales(LSRUse &LU, unsigned LUIdx, Formula Base);
  void GenerateTruncates(LSRUse &LU, unsigned LUIdx, Formula Base);
  void GenerateCrossUseConstantOffsets();
  void GenerateAllReuseFormulae();

  void FilterOutUndesirableDedicatedRegisters();

  size_t EstimateSearchSpaceComplexity() const;
  void NarrowSearchSpaceByDetectingSupersets();
  void NarrowSearchSpaceByCollapsingUnrolledCode();
  void NarrowSearchSpaceByRefilteringUndesirableDedicatedRegisters();
  void NarrowSearchSpaceByFilterFormulaWithSameScaledReg();
  void NarrowSearchSpaceByDeletingCostlyFormulas();
  void NarrowSearchSpaceByPickingWinnerRegs();
  void NarrowSearchSpaceUsingHeuristics();

  void SolveRecurse(SmallVectorImpl<const Formula *> &Solution,
                    Cost &SolutionCost,
                    SmallVectorImpl<const Formula *> &Workspace,
                    const Cost &CurCost,
                    const SmallPtrSet<const SCEV *, 16> &CurRegs,
                    DenseSet<const SCEV *> &VisitedRegs) const;
  void Solve(SmallVectorImpl<const Formula *> &Solution) const;

  BasicBlock::iterator
    HoistInsertPosition(BasicBlock::iterator IP,
                        const SmallVectorImpl<Instruction *> &Inputs) const;
  BasicBlock::iterator
    AdjustInsertPositionForExpand(BasicBlock::iterator IP,
                                  const LSRFixup &LF,
                                  const LSRUse &LU,
                                  SCEVExpander &Rewriter) const;

  Value *Expand(const LSRUse &LU, const LSRFixup &LF, const Formula &F,
                BasicBlock::iterator IP, SCEVExpander &Rewriter,
                SmallVectorImpl<WeakTrackingVH> &DeadInsts) const;
  void RewriteForPHI(PHINode *PN, const LSRUse &LU, const LSRFixup &LF,
                     const Formula &F, SCEVExpander &Rewriter,
                     SmallVectorImpl<WeakTrackingVH> &DeadInsts) const;
  void Rewrite(const LSRUse &LU, const LSRFixup &LF, const Formula &F,
               SCEVExpander &Rewriter,
               SmallVectorImpl<WeakTrackingVH> &DeadInsts) const;
  void ImplementSolution(const SmallVectorImpl<const Formula *> &Solution);

public:
  LSRInstance(Loop *L, IVUsers &IU, ScalarEvolution &SE, DominatorTree &DT,
              LoopInfo &LI, const TargetTransformInfo &TTI);

  bool getChanged() const { return Changed; }

  void print_factors_and_types(raw_ostream &OS) const;
  void print_fixups(raw_ostream &OS) const;
  void print_uses(raw_ostream &OS) const;
  void print(raw_ostream &OS) const;
  void dump() const;
};

} // end anonymous namespace

/// If IV is used in a int-to-float cast inside the loop then try to eliminate
/// the cast operation.
void LSRInstance::OptimizeShadowIV() {
  const SCEV *BackedgeTakenCount = SE.getBackedgeTakenCount(L);
  if (isa<SCEVCouldNotCompute>(BackedgeTakenCount))
    return;

  for (IVUsers::const_iterator UI = IU.begin(), E = IU.end();
       UI != E; /* empty */) {
    IVUsers::const_iterator CandidateUI = UI;
    ++UI;
    Instruction *ShadowUse = CandidateUI->getUser();
    Type *DestTy = nullptr;
    bool IsSigned = false;

    /* If shadow use is a int->float cast then insert a second IV
       to eliminate this cast.

         for (unsigned i = 0; i < n; ++i)
           foo((double)i);

       is transformed into

         double d = 0.0;
         for (unsigned i = 0; i < n; ++i, ++d)
           foo(d);
    */
    if (UIToFPInst *UCast = dyn_cast<UIToFPInst>(CandidateUI->getUser())) {
      IsSigned = false;
      DestTy = UCast->getDestTy();
    }
    else if (SIToFPInst *SCast = dyn_cast<SIToFPInst>(CandidateUI->getUser())) {
      IsSigned = true;
      DestTy = SCast->getDestTy();
    }
    if (!DestTy) continue;

    // If target does not support DestTy natively then do not apply
    // this transformation.
    if (!TTI.isTypeLegal(DestTy)) continue;

    PHINode *PH = dyn_cast<PHINode>(ShadowUse->getOperand(0));
    if (!PH) continue;
    if (PH->getNumIncomingValues() != 2) continue;

    // If the calculation in integers overflows, the result in FP type will
    // differ. So we only can do this transformation if we are guaranteed to not
    // deal with overflowing values
    const SCEVAddRecExpr *AR = dyn_cast<SCEVAddRecExpr>(SE.getSCEV(PH));
    if (!AR) continue;
    if (IsSigned && !AR->hasNoSignedWrap()) continue;
    if (!IsSigned && !AR->hasNoUnsignedWrap()) continue;

    Type *SrcTy = PH->getType();
    int Mantissa = DestTy->getFPMantissaWidth();
    if (Mantissa == -1) continue;
    if ((int)SE.getTypeSizeInBits(SrcTy) > Mantissa)
      continue;

    unsigned Entry, Latch;
    if (PH->getIncomingBlock(0) == L->getLoopPreheader()) {
      Entry = 0;
      Latch = 1;
    } else {
      Entry = 1;
      Latch = 0;
    }

    ConstantInt *Init = dyn_cast<ConstantInt>(PH->getIncomingValue(Entry));
    if (!Init) continue;
    Constant *NewInit = ConstantFP::get(DestTy, IsSigned ?
                                        (double)Init->getSExtValue() :
                                        (double)Init->getZExtValue());

    BinaryOperator *Incr =
      dyn_cast<BinaryOperator>(PH->getIncomingValue(Latch));
    if (!Incr) continue;
    if (Incr->getOpcode() != Instruction::Add
        && Incr->getOpcode() != Instruction::Sub)
      continue;

    /* Initialize new IV, double d = 0.0 in above example. */
    ConstantInt *C = nullptr;
    if (Incr->getOperand(0) == PH)
      C = dyn_cast<ConstantInt>(Incr->getOperand(1));
    else if (Incr->getOperand(1) == PH)
      C = dyn_cast<ConstantInt>(Incr->getOperand(0));
    else
      continue;

    if (!C) continue;

    // Ignore negative constants, as the code below doesn't handle them
    // correctly. TODO: Remove this restriction.
    if (!C->getValue().isStrictlyPositive()) continue;

    /* Add new PHINode. */
    PHINode *NewPH = PHINode::Create(DestTy, 2, "IV.S.", PH);

    /* create new increment. '++d' in above example. */
    Constant *CFP = ConstantFP::get(DestTy, C->getZExtValue());
    BinaryOperator *NewIncr =
      BinaryOperator::Create(Incr->getOpcode() == Instruction::Add ?
                               Instruction::FAdd : Instruction::FSub,
                             NewPH, CFP, "IV.S.next.", Incr);

    NewPH->addIncoming(NewInit, PH->getIncomingBlock(Entry));
    NewPH->addIncoming(NewIncr, PH->getIncomingBlock(Latch));

    /* Remove cast operation */
    ShadowUse->replaceAllUsesWith(NewPH);
    ShadowUse->eraseFromParent();
    Changed = true;
    break;
  }
}

/// If Cond has an operand that is an expression of an IV, set the IV user and
/// stride information and return true, otherwise return false.
bool LSRInstance::FindIVUserForCond(ICmpInst *Cond, IVStrideUse *&CondUse) {
  for (IVStrideUse &U : IU)
    if (U.getUser() == Cond) {
      // NOTE: we could handle setcc instructions with multiple uses here, but
      // InstCombine does it as well for simple uses, it's not clear that it
      // occurs enough in real life to handle.
      CondUse = &U;
      return true;
    }
  return false;
}

/// Rewrite the loop's terminating condition if it uses a max computation.
///
/// This is a narrow solution to a specific, but acute, problem. For loops
/// like this:
///
///   i = 0;
///   do {
///     p[i] = 0.0;
///   } while (++i < n);
///
/// the trip count isn't just 'n', because 'n' might not be positive. And
/// unfortunately this can come up even for loops where the user didn't use
/// a C do-while loop. For example, seemingly well-behaved top-test loops
/// will commonly be lowered like this:
///
///   if (n > 0) {
///     i = 0;
///     do {
///       p[i] = 0.0;
///     } while (++i < n);
///   }
///
/// and then it's possible for subsequent optimization to obscure the if
/// test in such a way that indvars can't find it.
///
/// When indvars can't find the if test in loops like this, it creates a
/// max expression, which allows it to give the loop a canonical
/// induction variable:
///
///   i = 0;
///   max = n < 1 ? 1 : n;
///   do {
///     p[i] = 0.0;
///   } while (++i != max);
///
/// Canonical induction variables are necessary because the loop passes
/// are designed around them. The most obvious example of this is the
/// LoopInfo analysis, which doesn't remember trip count values. It
/// expects to be able to rediscover the trip count each time it is
/// needed, and it does this using a simple analysis that only succeeds if
/// the loop has a canonical induction variable.
///
/// However, when it comes time to generate code, the maximum operation
/// can be quite costly, especially if it's inside of an outer loop.
///
/// This function solves this problem by detecting this type of loop and
/// rewriting their conditions from ICMP_NE back to ICMP_SLT, and deleting
/// the instructions for the maximum computation.
ICmpInst *LSRInstance::OptimizeMax(ICmpInst *Cond, IVStrideUse* &CondUse) {
  // Check that the loop matches the pattern we're looking for.
  if (Cond->getPredicate() != CmpInst::ICMP_EQ &&
      Cond->getPredicate() != CmpInst::ICMP_NE)
    return Cond;

  SelectInst *Sel = dyn_cast<SelectInst>(Cond->getOperand(1));
  if (!Sel || !Sel->hasOneUse()) return Cond;

  const SCEV *BackedgeTakenCount = SE.getBackedgeTakenCount(L);
  if (isa<SCEVCouldNotCompute>(BackedgeTakenCount))
    return Cond;
  const SCEV *One = SE.getConstant(BackedgeTakenCount->getType(), 1);

  // Add one to the backedge-taken count to get the trip count.
  const SCEV *IterationCount = SE.getAddExpr(One, BackedgeTakenCount);
  if (IterationCount != SE.getSCEV(Sel)) return Cond;

  // Check for a max calculation that matches the pattern. There's no check
  // for ICMP_ULE here because the comparison would be with zero, which
  // isn't interesting.
  CmpInst::Predicate Pred = ICmpInst::BAD_ICMP_PREDICATE;
  const SCEVNAryExpr *Max = nullptr;
  if (const SCEVSMaxExpr *S = dyn_cast<SCEVSMaxExpr>(BackedgeTakenCount)) {
    Pred = ICmpInst::ICMP_SLE;
    Max = S;
  } else if (const SCEVSMaxExpr *S = dyn_cast<SCEVSMaxExpr>(IterationCount)) {
    Pred = ICmpInst::ICMP_SLT;
    Max = S;
  } else if (const SCEVUMaxExpr *U = dyn_cast<SCEVUMaxExpr>(IterationCount)) {
    Pred = ICmpInst::ICMP_ULT;
    Max = U;
  } else {
    // No match; bail.
    return Cond;
  }

  // To handle a max with more than two operands, this optimization would
  // require additional checking and setup.
  if (Max->getNumOperands() != 2)
    return Cond;

  const SCEV *MaxLHS = Max->getOperand(0);
  const SCEV *MaxRHS = Max->getOperand(1);

  // ScalarEvolution canonicalizes constants to the left. For < and >, look
  // for a comparison with 1. For <= and >=, a comparison with zero.
  if (!MaxLHS ||
      (ICmpInst::isTrueWhenEqual(Pred) ? !MaxLHS->isZero() : (MaxLHS != One)))
    return Cond;

  // Check the relevant induction variable for conformance to
  // the pattern.
  const SCEV *IV = SE.getSCEV(Cond->getOperand(0));
  const SCEVAddRecExpr *AR = dyn_cast<SCEVAddRecExpr>(IV);
  if (!AR || !AR->isAffine() ||
      AR->getStart() != One ||
      AR->getStepRecurrence(SE) != One)
    return Cond;

  assert(AR->getLoop() == L &&
         "Loop condition operand is an addrec in a different loop!");

  // Check the right operand of the select, and remember it, as it will
  // be used in the new comparison instruction.
  Value *NewRHS = nullptr;
  if (ICmpInst::isTrueWhenEqual(Pred)) {
    // Look for n+1, and grab n.
    if (AddOperator *BO = dyn_cast<AddOperator>(Sel->getOperand(1)))
      if (ConstantInt *BO1 = dyn_cast<ConstantInt>(BO->getOperand(1)))
         if (BO1->isOne() && SE.getSCEV(BO->getOperand(0)) == MaxRHS)
           NewRHS = BO->getOperand(0);
    if (AddOperator *BO = dyn_cast<AddOperator>(Sel->getOperand(2)))
      if (ConstantInt *BO1 = dyn_cast<ConstantInt>(BO->getOperand(1)))
        if (BO1->isOne() && SE.getSCEV(BO->getOperand(0)) == MaxRHS)
          NewRHS = BO->getOperand(0);
    if (!NewRHS)
      return Cond;
  } else if (SE.getSCEV(Sel->getOperand(1)) == MaxRHS)
    NewRHS = Sel->getOperand(1);
  else if (SE.getSCEV(Sel->getOperand(2)) == MaxRHS)
    NewRHS = Sel->getOperand(2);
  else if (const SCEVUnknown *SU = dyn_cast<SCEVUnknown>(MaxRHS))
    NewRHS = SU->getValue();
  else
    // Max doesn't match expected pattern.
    return Cond;

  // Determine the new comparison opcode. It may be signed or unsigned,
  // and the original comparison may be either equality or inequality.
  if (Cond->getPredicate() == CmpInst::ICMP_EQ)
    Pred = CmpInst::getInversePredicate(Pred);

  // Ok, everything looks ok to change the condition into an SLT or SGE and
  // delete the max calculation.
  ICmpInst *NewCond =
    new ICmpInst(Cond, Pred, Cond->getOperand(0), NewRHS, "scmp");

  // Delete the max calculation instructions.
  Cond->replaceAllUsesWith(NewCond);
  CondUse->setUser(NewCond);
  Instruction *Cmp = cast<Instruction>(Sel->getOperand(0));
  Cond->eraseFromParent();
  Sel->eraseFromParent();
  if (Cmp->use_empty())
    Cmp->eraseFromParent();
  return NewCond;
}

/// Change loop terminating condition to use the postinc iv when possible.
void
LSRInstance::OptimizeLoopTermCond() {
  SmallPtrSet<Instruction *, 4> PostIncs;

  // We need a different set of heuristics for rotated and non-rotated loops.
  // If a loop is rotated then the latch is also the backedge, so inserting
  // post-inc expressions just before the latch is ideal. To reduce live ranges
  // it also makes sense to rewrite terminating conditions to use post-inc
  // expressions.
  //
  // If the loop is not rotated then the latch is not a backedge; the latch
  // check is done in the loop head. Adding post-inc expressions before the
  // latch will cause overlapping live-ranges of pre-inc and post-inc expressions
  // in the loop body. In this case we do *not* want to use post-inc expressions
  // in the latch check, and we want to insert post-inc expressions before
  // the backedge.
  BasicBlock *LatchBlock = L->getLoopLatch();
  SmallVector<BasicBlock*, 8> ExitingBlocks;
  L->getExitingBlocks(ExitingBlocks);
  if (llvm::all_of(ExitingBlocks, [&LatchBlock](const BasicBlock *BB) {
        return LatchBlock != BB;
      })) {
    // The backedge doesn't exit the loop; treat this as a head-tested loop.
    IVIncInsertPos = LatchBlock->getTerminator();
    return;
  }

  // Otherwise treat this as a rotated loop.
  for (BasicBlock *ExitingBlock : ExitingBlocks) {
    // Get the terminating condition for the loop if possible.  If we
    // can, we want to change it to use a post-incremented version of its
    // induction variable, to allow coalescing the live ranges for the IV into
    // one register value.

    BranchInst *TermBr = dyn_cast<BranchInst>(ExitingBlock->getTerminator());
    if (!TermBr)
      continue;
    // FIXME: Overly conservative, termination condition could be an 'or' etc..
    if (TermBr->isUnconditional() || !isa<ICmpInst>(TermBr->getCondition()))
      continue;

    // Search IVUsesByStride to find Cond's IVUse if there is one.
    IVStrideUse *CondUse = nullptr;
    ICmpInst *Cond = cast<ICmpInst>(TermBr->getCondition());
    if (!FindIVUserForCond(Cond, CondUse))
      continue;

    // If the trip count is computed in terms of a max (due to ScalarEvolution
    // being unable to find a sufficient guard, for example), change the loop
    // comparison to use SLT or ULT instead of NE.
    // One consequence of doing this now is that it disrupts the count-down
    // optimization. That's not always a bad thing though, because in such
    // cases it may still be worthwhile to avoid a max.
    Cond = OptimizeMax(Cond, CondUse);

    // If this exiting block dominates the latch block, it may also use
    // the post-inc value if it won't be shared with other uses.
    // Check for dominance.
    if (!DT.dominates(ExitingBlock, LatchBlock))
      continue;

    // Conservatively avoid trying to use the post-inc value in non-latch
    // exits if there may be pre-inc users in intervening blocks.
    if (LatchBlock != ExitingBlock)
      for (IVUsers::const_iterator UI = IU.begin(), E = IU.end(); UI != E; ++UI)
        // Test if the use is reachable from the exiting block. This dominator
        // query is a conservative approximation of reachability.
        if (&*UI != CondUse &&
            !DT.properlyDominates(UI->getUser()->getParent(), ExitingBlock)) {
          // Conservatively assume there may be reuse if the quotient of their
          // strides could be a legal scale.
          const SCEV *A = IU.getStride(*CondUse, L);
          const SCEV *B = IU.getStride(*UI, L);
          if (!A || !B) continue;
          if (SE.getTypeSizeInBits(A->getType()) !=
              SE.getTypeSizeInBits(B->getType())) {
            if (SE.getTypeSizeInBits(A->getType()) >
                SE.getTypeSizeInBits(B->getType()))
              B = SE.getSignExtendExpr(B, A->getType());
            else
              A = SE.getSignExtendExpr(A, B->getType());
          }
          if (const SCEVConstant *D =
                dyn_cast_or_null<SCEVConstant>(getExactSDiv(B, A, SE))) {
            const ConstantInt *C = D->getValue();
            // Stride of one or negative one can have reuse with non-addresses.
            if (C->isOne() || C->isMinusOne())
              goto decline_post_inc;
            // Avoid weird situations.
            if (C->getValue().getMinSignedBits() >= 64 ||
                C->getValue().isMinSignedValue())
              goto decline_post_inc;
            // Check for possible scaled-address reuse.
            MemAccessTy AccessTy = getAccessType(TTI, UI->getUser());
            int64_t Scale = C->getSExtValue();
            if (TTI.isLegalAddressingMode(AccessTy.MemTy, /*BaseGV=*/nullptr,
                                          /*BaseOffset=*/0,
                                          /*HasBaseReg=*/false, Scale,
                                          AccessTy.AddrSpace))
              goto decline_post_inc;
            Scale = -Scale;
            if (TTI.isLegalAddressingMode(AccessTy.MemTy, /*BaseGV=*/nullptr,
                                          /*BaseOffset=*/0,
                                          /*HasBaseReg=*/false, Scale,
                                          AccessTy.AddrSpace))
              goto decline_post_inc;
          }
        }

    DEBUG(dbgs() << "  Change loop exiting icmp to use postinc iv: "
                 << *Cond << '\n');

    // It's possible for the setcc instruction to be anywhere in the loop, and
    // possible for it to have multiple users.  If it is not immediately before
    // the exiting block branch, move it.
    if (&*++BasicBlock::iterator(Cond) != TermBr) {
      if (Cond->hasOneUse()) {
        Cond->moveBefore(TermBr);
      } else {
        // Clone the terminating condition and insert into the loopend.
        ICmpInst *OldCond = Cond;
        Cond = cast<ICmpInst>(Cond->clone());
        Cond->setName(L->getHeader()->getName() + ".termcond");
        ExitingBlock->getInstList().insert(TermBr->getIterator(), Cond);

        // Clone the IVUse, as the old use still exists!
        CondUse = &IU.AddUser(Cond, CondUse->getOperandValToReplace());
        TermBr->replaceUsesOfWith(OldCond, Cond);
      }
    }

    // If we get to here, we know that we can transform the setcc instruction to
    // use the post-incremented version of the IV, allowing us to coalesce the
    // live ranges for the IV correctly.
    CondUse->transformToPostInc(L);
    Changed = true;

    PostIncs.insert(Cond);
  decline_post_inc:;
  }

  // Determine an insertion point for the loop induction variable increment. It
  // must dominate all the post-inc comparisons we just set up, and it must
  // dominate the loop latch edge.
  IVIncInsertPos = L->getLoopLatch()->getTerminator();
  for (Instruction *Inst : PostIncs) {
    BasicBlock *BB =
      DT.findNearestCommonDominator(IVIncInsertPos->getParent(),
                                    Inst->getParent());
    if (BB == Inst->getParent())
      IVIncInsertPos = Inst;
    else if (BB != IVIncInsertPos->getParent())
      IVIncInsertPos = BB->getTerminator();
  }
}

/// Determine if the given use can accommodate a fixup at the given offset and
/// other details. If so, update the use and return true.
bool LSRInstance::reconcileNewOffset(LSRUse &LU, int64_t NewOffset,
                                     bool HasBaseReg, LSRUse::KindType Kind,
                                     MemAccessTy AccessTy) {
  int64_t NewMinOffset = LU.MinOffset;
  int64_t NewMaxOffset = LU.MaxOffset;
  MemAccessTy NewAccessTy = AccessTy;

  // Check for a mismatched kind. It's tempting to collapse mismatched kinds to
  // something conservative, however this can pessimize in the case that one of
  // the uses will have all its uses outside the loop, for example.
  if (LU.Kind != Kind)
    return false;

  // Check for a mismatched access type, and fall back conservatively as needed.
  // TODO: Be less conservative when the type is similar and can use the same
  // addressing modes.
  if (Kind == LSRUse::Address) {
    if (AccessTy.MemTy != LU.AccessTy.MemTy) {
      NewAccessTy = MemAccessTy::getUnknown(AccessTy.MemTy->getContext(),
                                            AccessTy.AddrSpace);
    }
  }

  // Conservatively assume HasBaseReg is true for now.
  if (NewOffset < LU.MinOffset) {
    if (!isAlwaysFoldable(TTI, Kind, NewAccessTy, /*BaseGV=*/nullptr,
                          LU.MaxOffset - NewOffset, HasBaseReg))
      return false;
    NewMinOffset = NewOffset;
  } else if (NewOffset > LU.MaxOffset) {
    if (!isAlwaysFoldable(TTI, Kind, NewAccessTy, /*BaseGV=*/nullptr,
                          NewOffset - LU.MinOffset, HasBaseReg))
      return false;
    NewMaxOffset = NewOffset;
  }

  // Update the use.
  LU.MinOffset = NewMinOffset;
  LU.MaxOffset = NewMaxOffset;
  LU.AccessTy = NewAccessTy;
  return true;
}

/// Return an LSRUse index and an offset value for a fixup which needs the given
/// expression, with the given kind and optional access type.  Either reuse an
/// existing use or create a new one, as needed.
std::pair<size_t, int64_t> LSRInstance::getUse(const SCEV *&Expr,
                                               LSRUse::KindType Kind,
                                               MemAccessTy AccessTy) {
  const SCEV *Copy = Expr;
  int64_t Offset = ExtractImmediate(Expr, SE);

  // Basic uses can't accept any offset, for example.
  if (!isAlwaysFoldable(TTI, Kind, AccessTy, /*BaseGV=*/ nullptr,
                        Offset, /*HasBaseReg=*/ true)) {
    Expr = Copy;
    Offset = 0;
  }

  std::pair<UseMapTy::iterator, bool> P =
    UseMap.insert(std::make_pair(LSRUse::SCEVUseKindPair(Expr, Kind), 0));
  if (!P.second) {
    // A use already existed with this base.
    size_t LUIdx = P.first->second;
    LSRUse &LU = Uses[LUIdx];
    if (reconcileNewOffset(LU, Offset, /*HasBaseReg=*/true, Kind, AccessTy))
      // Reuse this use.
      return std::make_pair(LUIdx, Offset);
  }

  // Create a new use.
  size_t LUIdx = Uses.size();
  P.first->second = LUIdx;
  Uses.push_back(LSRUse(Kind, AccessTy));
  LSRUse &LU = Uses[LUIdx];

  LU.MinOffset = Offset;
  LU.MaxOffset = Offset;
  return std::make_pair(LUIdx, Offset);
}

/// Delete the given use from the Uses list.
void LSRInstance::DeleteUse(LSRUse &LU, size_t LUIdx) {
  if (&LU != &Uses.back())
    std::swap(LU, Uses.back());
  Uses.pop_back();

  // Update RegUses.
  RegUses.swapAndDropUse(LUIdx, Uses.size());
}

/// Look for a use distinct from OrigLU which is has a formula that has the same
/// registers as the given formula.
LSRUse *
LSRInstance::FindUseWithSimilarFormula(const Formula &OrigF,
                                       const LSRUse &OrigLU) {
  // Search all uses for the formula. This could be more clever.
  for (size_t LUIdx = 0, NumUses = Uses.size(); LUIdx != NumUses; ++LUIdx) {
    LSRUse &LU = Uses[LUIdx];
    // Check whether this use is close enough to OrigLU, to see whether it's
    // worthwhile looking through its formulae.
    // Ignore ICmpZero uses because they may contain formulae generated by
    // GenerateICmpZeroScales, in which case adding fixup offsets may
    // be invalid.
    if (&LU != &OrigLU &&
        LU.Kind != LSRUse::ICmpZero &&
        LU.Kind == OrigLU.Kind && OrigLU.AccessTy == LU.AccessTy &&
        LU.WidestFixupType == OrigLU.WidestFixupType &&
        LU.HasFormulaWithSameRegs(OrigF)) {
      // Scan through this use's formulae.
      for (const Formula &F : LU.Formulae) {
        // Check to see if this formula has the same registers and symbols
        // as OrigF.
        if (F.BaseRegs == OrigF.BaseRegs &&
            F.ScaledReg == OrigF.ScaledReg &&
            F.BaseGV == OrigF.BaseGV &&
            F.Scale == OrigF.Scale &&
            F.UnfoldedOffset == OrigF.UnfoldedOffset) {
          if (F.BaseOffset == 0)
            return &LU;
          // This is the formula where all the registers and symbols matched;
          // there aren't going to be any others. Since we declined it, we
          // can skip the rest of the formulae and proceed to the next LSRUse.
          break;
        }
      }
    }
  }

  // Nothing looked good.
  return nullptr;
}

void LSRInstance::CollectInterestingTypesAndFactors() {
  SmallSetVector<const SCEV *, 4> Strides;

  // Collect interesting types and strides.
  SmallVector<const SCEV *, 4> Worklist;
  for (const IVStrideUse &U : IU) {
    const SCEV *Expr = IU.getExpr(U);

    // Collect interesting types.
    Types.insert(SE.getEffectiveSCEVType(Expr->getType()));

    // Add strides for mentioned loops.
    Worklist.push_back(Expr);
    do {
      const SCEV *S = Worklist.pop_back_val();
      if (const SCEVAddRecExpr *AR = dyn_cast<SCEVAddRecExpr>(S)) {
        if (AR->getLoop() == L)
          Strides.insert(AR->getStepRecurrence(SE));
        Worklist.push_back(AR->getStart());
      } else if (const SCEVAddExpr *Add = dyn_cast<SCEVAddExpr>(S)) {
        Worklist.append(Add->op_begin(), Add->op_end());
      }
    } while (!Worklist.empty());
  }

  // Compute interesting factors from the set of interesting strides.
  for (SmallSetVector<const SCEV *, 4>::const_iterator
       I = Strides.begin(), E = Strides.end(); I != E; ++I)
    for (SmallSetVector<const SCEV *, 4>::const_iterator NewStrideIter =
         std::next(I); NewStrideIter != E; ++NewStrideIter) {
      const SCEV *OldStride = *I;
      const SCEV *NewStride = *NewStrideIter;

      if (SE.getTypeSizeInBits(OldStride->getType()) !=
          SE.getTypeSizeInBits(NewStride->getType())) {
        if (SE.getTypeSizeInBits(OldStride->getType()) >
            SE.getTypeSizeInBits(NewStride->getType()))
          NewStride = SE.getSignExtendExpr(NewStride, OldStride->getType());
        else
          OldStride = SE.getSignExtendExpr(OldStride, NewStride->getType());
      }
      if (const SCEVConstant *Factor =
            dyn_cast_or_null<SCEVConstant>(getExactSDiv(NewStride, OldStride,
                                                        SE, true))) {
        if (Factor->getAPInt().getMinSignedBits() <= 64)
          Factors.insert(Factor->getAPInt().getSExtValue());
      } else if (const SCEVConstant *Factor =
                   dyn_cast_or_null<SCEVConstant>(getExactSDiv(OldStride,
                                                               NewStride,
                                                               SE, true))) {
        if (Factor->getAPInt().getMinSignedBits() <= 64)
          Factors.insert(Factor->getAPInt().getSExtValue());
      }
    }

  // If all uses use the same type, don't bother looking for truncation-based
  // reuse.
  if (Types.size() == 1)
    Types.clear();

  DEBUG(print_factors_and_types(dbgs()));
}

/// Helper for CollectChains that finds an IV operand (computed by an AddRec in
/// this loop) within [OI,OE) or returns OE. If IVUsers mapped Instructions to
/// IVStrideUses, we could partially skip this.
static User::op_iterator
findIVOperand(User::op_iterator OI, User::op_iterator OE,
              Loop *L, ScalarEvolution &SE) {
  for(; OI != OE; ++OI) {
    if (Instruction *Oper = dyn_cast<Instruction>(*OI)) {
      if (!SE.isSCEVable(Oper->getType()))
        continue;

      if (const SCEVAddRecExpr *AR =
          dyn_cast<SCEVAddRecExpr>(SE.getSCEV(Oper))) {
        if (AR->getLoop() == L)
          break;
      }
    }
  }
  return OI;
}

/// IVChain logic must consistenctly peek base TruncInst operands, so wrap it in
/// a convenient helper.
static Value *getWideOperand(Value *Oper) {
  if (TruncInst *Trunc = dyn_cast<TruncInst>(Oper))
    return Trunc->getOperand(0);
  return Oper;
}

/// Return true if we allow an IV chain to include both types.
static bool isCompatibleIVType(Value *LVal, Value *RVal) {
  Type *LType = LVal->getType();
  Type *RType = RVal->getType();
  return (LType == RType) || (LType->isPointerTy() && RType->isPointerTy() &&
                              // Different address spaces means (possibly)
                              // different types of the pointer implementation,
                              // e.g. i16 vs i32 so disallow that.
                              (LType->getPointerAddressSpace() ==
                               RType->getPointerAddressSpace()));
}

/// Return an approximation of this SCEV expression's "base", or NULL for any
/// constant. Returning the expression itself is conservative. Returning a
/// deeper subexpression is more precise and valid as long as it isn't less
/// complex than another subexpression. For expressions involving multiple
/// unscaled values, we need to return the pointer-type SCEVUnknown. This avoids
/// forming chains across objects, such as: PrevOper==a[i], IVOper==b[i],
/// IVInc==b-a.
///
/// Since SCEVUnknown is the rightmost type, and pointers are the rightmost
/// SCEVUnknown, we simply return the rightmost SCEV operand.
static const SCEV *getExprBase(const SCEV *S) {
  switch (S->getSCEVType()) {
  default: // uncluding scUnknown.
    return S;
  case scConstant:
    return nullptr;
  case scTruncate:
    return getExprBase(cast<SCEVTruncateExpr>(S)->getOperand());
  case scZeroExtend:
    return getExprBase(cast<SCEVZeroExtendExpr>(S)->getOperand());
  case scSignExtend:
    return getExprBase(cast<SCEVSignExtendExpr>(S)->getOperand());
  case scAddExpr: {
    // Skip over scaled operands (scMulExpr) to follow add operands as long as
    // there's nothing more complex.
    // FIXME: not sure if we want to recognize negation.
    const SCEVAddExpr *Add = cast<SCEVAddExpr>(S);
    for (std::reverse_iterator<SCEVAddExpr::op_iterator> I(Add->op_end()),
           E(Add->op_begin()); I != E; ++I) {
      const SCEV *SubExpr = *I;
      if (SubExpr->getSCEVType() == scAddExpr)
        return getExprBase(SubExpr);

      if (SubExpr->getSCEVType() != scMulExpr)
        return SubExpr;
    }
    return S; // all operands are scaled, be conservative.
  }
  case scAddRecExpr:
    return getExprBase(cast<SCEVAddRecExpr>(S)->getStart());
  }
}

/// Return true if the chain increment is profitable to expand into a loop
/// invariant value, which may require its own register. A profitable chain
/// increment will be an offset relative to the same base. We allow such offsets
/// to potentially be used as chain increment as long as it's not obviously
/// expensive to expand using real instructions.
bool IVChain::isProfitableIncrement(const SCEV *OperExpr,
                                    const SCEV *IncExpr,
                                    ScalarEvolution &SE) {
  // Aggressively form chains when -stress-ivchain.
  if (StressIVChain)
    return true;

  // Do not replace a constant offset from IV head with a nonconstant IV
  // increment.
  if (!isa<SCEVConstant>(IncExpr)) {
    const SCEV *HeadExpr = SE.getSCEV(getWideOperand(Incs[0].IVOperand));
    if (isa<SCEVConstant>(SE.getMinusSCEV(OperExpr, HeadExpr)))
      return false;
  }

  SmallPtrSet<const SCEV*, 8> Processed;
  return !isHighCostExpansion(IncExpr, Processed, SE);
}

/// Return true if the number of registers needed for the chain is estimated to
/// be less than the number required for the individual IV users. First prohibit
/// any IV users that keep the IV live across increments (the Users set should
/// be empty). Next count the number and type of increments in the chain.
///
/// Chaining IVs can lead to considerable code bloat if ISEL doesn't
/// effectively use postinc addressing modes. Only consider it profitable it the
/// increments can be computed in fewer registers when chained.
///
/// TODO: Consider IVInc free if it's already used in another chains.
static bool
isProfitableChain(IVChain &Chain, SmallPtrSetImpl<Instruction*> &Users,
                  ScalarEvolution &SE, const TargetTransformInfo &TTI) {
  if (StressIVChain)
    return true;

  if (!Chain.hasIncs())
    return false;

  if (!Users.empty()) {
    DEBUG(dbgs() << "Chain: " << *Chain.Incs[0].UserInst << " users:\n";
          for (Instruction *Inst : Users) {
            dbgs() << "  " << *Inst << "\n";
          });
    return false;
  }
  assert(!Chain.Incs.empty() && "empty IV chains are not allowed");

  // The chain itself may require a register, so intialize cost to 1.
  int cost = 1;

  // A complete chain likely eliminates the need for keeping the original IV in
  // a register. LSR does not currently know how to form a complete chain unless
  // the header phi already exists.
  if (isa<PHINode>(Chain.tailUserInst())
      && SE.getSCEV(Chain.tailUserInst()) == Chain.Incs[0].IncExpr) {
    --cost;
  }
  const SCEV *LastIncExpr = nullptr;
  unsigned NumConstIncrements = 0;
  unsigned NumVarIncrements = 0;
  unsigned NumReusedIncrements = 0;
  for (const IVInc &Inc : Chain) {
    if (Inc.IncExpr->isZero())
      continue;

    // Incrementing by zero or some constant is neutral. We assume constants can
    // be folded into an addressing mode or an add's immediate operand.
    if (isa<SCEVConstant>(Inc.IncExpr)) {
      ++NumConstIncrements;
      continue;
    }

    if (Inc.IncExpr == LastIncExpr)
      ++NumReusedIncrements;
    else
      ++NumVarIncrements;

    LastIncExpr = Inc.IncExpr;
  }
  // An IV chain with a single increment is handled by LSR's postinc
  // uses. However, a chain with multiple increments requires keeping the IV's
  // value live longer than it needs to be if chained.
  if (NumConstIncrements > 1)
    --cost;

  // Materializing increment expressions in the preheader that didn't exist in
  // the original code may cost a register. For example, sign-extended array
  // indices can produce ridiculous increments like this:
  // IV + ((sext i32 (2 * %s) to i64) + (-1 * (sext i32 %s to i64)))
  cost += NumVarIncrements;

  // Reusing variable increments likely saves a register to hold the multiple of
  // the stride.
  cost -= NumReusedIncrements;

  DEBUG(dbgs() << "Chain: " << *Chain.Incs[0].UserInst << " Cost: " << cost
               << "\n");

  return cost < 0;
}

/// Add this IV user to an existing chain or make it the head of a new chain.
void LSRInstance::ChainInstruction(Instruction *UserInst, Instruction *IVOper,
                                   SmallVectorImpl<ChainUsers> &ChainUsersVec) {
  // When IVs are used as types of varying widths, they are generally converted
  // to a wider type with some uses remaining narrow under a (free) trunc.
  Value *const NextIV = getWideOperand(IVOper);
  const SCEV *const OperExpr = SE.getSCEV(NextIV);
  const SCEV *const OperExprBase = getExprBase(OperExpr);

  // Visit all existing chains. Check if its IVOper can be computed as a
  // profitable loop invariant increment from the last link in the Chain.
  unsigned ChainIdx = 0, NChains = IVChainVec.size();
  const SCEV *LastIncExpr = nullptr;
  for (; ChainIdx < NChains; ++ChainIdx) {
    IVChain &Chain = IVChainVec[ChainIdx];

    // Prune the solution space aggressively by checking that both IV operands
    // are expressions that operate on the same unscaled SCEVUnknown. This
    // "base" will be canceled by the subsequent getMinusSCEV call. Checking
    // first avoids creating extra SCEV expressions.
    if (!StressIVChain && Chain.ExprBase != OperExprBase)
      continue;

    Value *PrevIV = getWideOperand(Chain.Incs.back().IVOperand);
    if (!isCompatibleIVType(PrevIV, NextIV))
      continue;

    // A phi node terminates a chain.
    if (isa<PHINode>(UserInst) && isa<PHINode>(Chain.tailUserInst()))
      continue;

    // The increment must be loop-invariant so it can be kept in a register.
    const SCEV *PrevExpr = SE.getSCEV(PrevIV);
    const SCEV *IncExpr = SE.getMinusSCEV(OperExpr, PrevExpr);
    if (!SE.isLoopInvariant(IncExpr, L))
      continue;

    if (Chain.isProfitableIncrement(OperExpr, IncExpr, SE)) {
      LastIncExpr = IncExpr;
      break;
    }
  }
  // If we haven't found a chain, create a new one, unless we hit the max. Don't
  // bother for phi nodes, because they must be last in the chain.
  if (ChainIdx == NChains) {
    if (isa<PHINode>(UserInst))
      return;
    if (NChains >= MaxChains && !StressIVChain) {
      DEBUG(dbgs() << "IV Chain Limit\n");
      return;
    }
    LastIncExpr = OperExpr;
    // IVUsers may have skipped over sign/zero extensions. We don't currently
    // attempt to form chains involving extensions unless they can be hoisted
    // into this loop's AddRec.
    if (!isa<SCEVAddRecExpr>(LastIncExpr))
      return;
    ++NChains;
    IVChainVec.push_back(IVChain(IVInc(UserInst, IVOper, LastIncExpr),
                                 OperExprBase));
    ChainUsersVec.resize(NChains);
    DEBUG(dbgs() << "IV Chain#" << ChainIdx << " Head: (" << *UserInst
                 << ") IV=" << *LastIncExpr << "\n");
  } else {
    DEBUG(dbgs() << "IV Chain#" << ChainIdx << "  Inc: (" << *UserInst
                 << ") IV+" << *LastIncExpr << "\n");
    // Add this IV user to the end of the chain.
    IVChainVec[ChainIdx].add(IVInc(UserInst, IVOper, LastIncExpr));
  }
  IVChain &Chain = IVChainVec[ChainIdx];

  SmallPtrSet<Instruction*,4> &NearUsers = ChainUsersVec[ChainIdx].NearUsers;
  // This chain's NearUsers become FarUsers.
  if (!LastIncExpr->isZero()) {
    ChainUsersVec[ChainIdx].FarUsers.insert(NearUsers.begin(),
                                            NearUsers.end());
    NearUsers.clear();
  }

  // All other uses of IVOperand become near uses of the chain.
  // We currently ignore intermediate values within SCEV expressions, assuming
  // they will eventually be used be the current chain, or can be computed
  // from one of the chain increments. To be more precise we could
  // transitively follow its user and only add leaf IV users to the set.
  for (User *U : IVOper->users()) {
    Instruction *OtherUse = dyn_cast<Instruction>(U);
    if (!OtherUse)
      continue;
    // Uses in the chain will no longer be uses if the chain is formed.
    // Include the head of the chain in this iteration (not Chain.begin()).
    IVChain::const_iterator IncIter = Chain.Incs.begin();
    IVChain::const_iterator IncEnd = Chain.Incs.end();
    for( ; IncIter != IncEnd; ++IncIter) {
      if (IncIter->UserInst == OtherUse)
        break;
    }
    if (IncIter != IncEnd)
      continue;

    if (SE.isSCEVable(OtherUse->getType())
        && !isa<SCEVUnknown>(SE.getSCEV(OtherUse))
        && IU.isIVUserOrOperand(OtherUse)) {
      continue;
    }
    NearUsers.insert(OtherUse);
  }

  // Since this user is part of the chain, it's no longer considered a use
  // of the chain.
  ChainUsersVec[ChainIdx].FarUsers.erase(UserInst);
}

/// Populate the vector of Chains.
///
/// This decreases ILP at the architecture level. Targets with ample registers,
/// multiple memory ports, and no register renaming probably don't want
/// this. However, such targets should probably disable LSR altogether.
///
/// The job of LSR is to make a reasonable choice of induction variables across
/// the loop. Subsequent passes can easily "unchain" computation exposing more
/// ILP *within the loop* if the target wants it.
///
/// Finding the best IV chain is potentially a scheduling problem. Since LSR
/// will not reorder memory operations, it will recognize this as a chain, but
/// will generate redundant IV increments. Ideally this would be corrected later
/// by a smart scheduler:
///        = A[i]
///        = A[i+x]
/// A[i]   =
/// A[i+x] =
///
/// TODO: Walk the entire domtree within this loop, not just the path to the
/// loop latch. This will discover chains on side paths, but requires
/// maintaining multiple copies of the Chains state.
void LSRInstance::CollectChains() {
  DEBUG(dbgs() << "Collecting IV Chains.\n");
  SmallVector<ChainUsers, 8> ChainUsersVec;

  SmallVector<BasicBlock *,8> LatchPath;
  BasicBlock *LoopHeader = L->getHeader();
  for (DomTreeNode *Rung = DT.getNode(L->getLoopLatch());
       Rung->getBlock() != LoopHeader; Rung = Rung->getIDom()) {
    LatchPath.push_back(Rung->getBlock());
  }
  LatchPath.push_back(LoopHeader);

  // Walk the instruction stream from the loop header to the loop latch.
  for (BasicBlock *BB : reverse(LatchPath)) {
    for (Instruction &I : *BB) {
      // Skip instructions that weren't seen by IVUsers analysis.
      if (isa<PHINode>(I) || !IU.isIVUserOrOperand(&I))
        continue;

      // Ignore users that are part of a SCEV expression. This way we only
      // consider leaf IV Users. This effectively rediscovers a portion of
      // IVUsers analysis but in program order this time.
      if (SE.isSCEVable(I.getType()) && !isa<SCEVUnknown>(SE.getSCEV(&I)))
          continue;

      // Remove this instruction from any NearUsers set it may be in.
      for (unsigned ChainIdx = 0, NChains = IVChainVec.size();
           ChainIdx < NChains; ++ChainIdx) {
        ChainUsersVec[ChainIdx].NearUsers.erase(&I);
      }
      // Search for operands that can be chained.
      SmallPtrSet<Instruction*, 4> UniqueOperands;
      User::op_iterator IVOpEnd = I.op_end();
      User::op_iterator IVOpIter = findIVOperand(I.op_begin(), IVOpEnd, L, SE);
      while (IVOpIter != IVOpEnd) {
        Instruction *IVOpInst = cast<Instruction>(*IVOpIter);
        if (UniqueOperands.insert(IVOpInst).second)
          ChainInstruction(&I, IVOpInst, ChainUsersVec);
        IVOpIter = findIVOperand(std::next(IVOpIter), IVOpEnd, L, SE);
      }
    } // Continue walking down the instructions.
  } // Continue walking down the domtree.
  // Visit phi backedges to determine if the chain can generate the IV postinc.
  for (PHINode &PN : L->getHeader()->phis()) {
    if (!SE.isSCEVable(PN.getType()))
      continue;

    Instruction *IncV =
        dyn_cast<Instruction>(PN.getIncomingValueForBlock(L->getLoopLatch()));
    if (IncV)
      ChainInstruction(&PN, IncV, ChainUsersVec);
  }
  // Remove any unprofitable chains.
  unsigned ChainIdx = 0;
  for (unsigned UsersIdx = 0, NChains = IVChainVec.size();
       UsersIdx < NChains; ++UsersIdx) {
    if (!isProfitableChain(IVChainVec[UsersIdx],
                           ChainUsersVec[UsersIdx].FarUsers, SE, TTI))
      continue;
    // Preserve the chain at UsesIdx.
    if (ChainIdx != UsersIdx)
      IVChainVec[ChainIdx] = IVChainVec[UsersIdx];
    FinalizeChain(IVChainVec[ChainIdx]);
    ++ChainIdx;
  }
  IVChainVec.resize(ChainIdx);
}

void LSRInstance::FinalizeChain(IVChain &Chain) {
  assert(!Chain.Incs.empty() && "empty IV chains are not allowed");
  DEBUG(dbgs() << "Final Chain: " << *Chain.Incs[0].UserInst << "\n");

  for (const IVInc &Inc : Chain) {
    DEBUG(dbgs() << "        Inc: " << *Inc.UserInst << "\n");
    auto UseI = find(Inc.UserInst->operands(), Inc.IVOperand);
    assert(UseI != Inc.UserInst->op_end() && "cannot find IV operand");
    IVIncSet.insert(UseI);
  }
}

/// Return true if the IVInc can be folded into an addressing mode.
static bool canFoldIVIncExpr(const SCEV *IncExpr, Instruction *UserInst,
                             Value *Operand, const TargetTransformInfo &TTI) {
  const SCEVConstant *IncConst = dyn_cast<SCEVConstant>(IncExpr);
  if (!IncConst || !isAddressUse(TTI, UserInst, Operand))
    return false;

  if (IncConst->getAPInt().getMinSignedBits() > 64)
    return false;

  MemAccessTy AccessTy = getAccessType(TTI, UserInst);
  int64_t IncOffset = IncConst->getValue()->getSExtValue();
  if (!isAlwaysFoldable(TTI, LSRUse::Address, AccessTy, /*BaseGV=*/nullptr,
                        IncOffset, /*HaseBaseReg=*/false))
    return false;

  return true;
}

/// Generate an add or subtract for each IVInc in a chain to materialize the IV
/// user's operand from the previous IV user's operand.
void LSRInstance::GenerateIVChain(const IVChain &Chain, SCEVExpander &Rewriter,
                                  SmallVectorImpl<WeakTrackingVH> &DeadInsts) {
  // Find the new IVOperand for the head of the chain. It may have been replaced
  // by LSR.
  const IVInc &Head = Chain.Incs[0];
  User::op_iterator IVOpEnd = Head.UserInst->op_end();
  // findIVOperand returns IVOpEnd if it can no longer find a valid IV user.
  User::op_iterator IVOpIter = findIVOperand(Head.UserInst->op_begin(),
                                             IVOpEnd, L, SE);
  Value *IVSrc = nullptr;
  while (IVOpIter != IVOpEnd) {
    IVSrc = getWideOperand(*IVOpIter);

    // If this operand computes the expression that the chain needs, we may use
    // it. (Check this after setting IVSrc which is used below.)
    //
    // Note that if Head.IncExpr is wider than IVSrc, then this phi is too
    // narrow for the chain, so we can no longer use it. We do allow using a
    // wider phi, assuming the LSR checked for free truncation. In that case we
    // should already have a truncate on this operand such that
    // getSCEV(IVSrc) == IncExpr.
    if (SE.getSCEV(*IVOpIter) == Head.IncExpr
        || SE.getSCEV(IVSrc) == Head.IncExpr) {
      break;
    }
    IVOpIter = findIVOperand(std::next(IVOpIter), IVOpEnd, L, SE);
  }
  if (IVOpIter == IVOpEnd) {
    // Gracefully give up on this chain.
    DEBUG(dbgs() << "Concealed chain head: " << *Head.UserInst << "\n");
    return;
  }

  DEBUG(dbgs() << "Generate chain at: " << *IVSrc << "\n");
  Type *IVTy = IVSrc->getType();
  Type *IntTy = SE.getEffectiveSCEVType(IVTy);
  const SCEV *LeftOverExpr = nullptr;
  for (const IVInc &Inc : Chain) {
    Instruction *InsertPt = Inc.UserInst;
    if (isa<PHINode>(InsertPt))
      InsertPt = L->getLoopLatch()->getTerminator();

    // IVOper will replace the current IV User's operand. IVSrc is the IV
    // value currently held in a register.
    Value *IVOper = IVSrc;
    if (!Inc.IncExpr->isZero()) {
      // IncExpr was the result of subtraction of two narrow values, so must
      // be signed.
      const SCEV *IncExpr = SE.getNoopOrSignExtend(Inc.IncExpr, IntTy);
      LeftOverExpr = LeftOverExpr ?
        SE.getAddExpr(LeftOverExpr, IncExpr) : IncExpr;
    }
    if (LeftOverExpr && !LeftOverExpr->isZero()) {
      // Expand the IV increment.
      Rewriter.clearPostInc();
      Value *IncV = Rewriter.expandCodeFor(LeftOverExpr, IntTy, InsertPt);
      const SCEV *IVOperExpr = SE.getAddExpr(SE.getUnknown(IVSrc),
                                             SE.getUnknown(IncV));
      IVOper = Rewriter.expandCodeFor(IVOperExpr, IVTy, InsertPt);

      // If an IV increment can't be folded, use it as the next IV value.
      if (!canFoldIVIncExpr(LeftOverExpr, Inc.UserInst, Inc.IVOperand, TTI)) {
        assert(IVTy == IVOper->getType() && "inconsistent IV increment type");
        IVSrc = IVOper;
        LeftOverExpr = nullptr;
      }
    }
    Type *OperTy = Inc.IVOperand->getType();
    if (IVTy != OperTy) {
      assert(SE.getTypeSizeInBits(IVTy) >= SE.getTypeSizeInBits(OperTy) &&
             "cannot extend a chained IV");
      IRBuilder<> Builder(InsertPt);
      IVOper = Builder.CreateTruncOrBitCast(IVOper, OperTy, "lsr.chain");
    }
    Inc.UserInst->replaceUsesOfWith(Inc.IVOperand, IVOper);
    DeadInsts.emplace_back(Inc.IVOperand);
  }
  // If LSR created a new, wider phi, we may also replace its postinc. We only
  // do this if we also found a wide value for the head of the chain.
  if (isa<PHINode>(Chain.tailUserInst())) {
    for (PHINode &Phi : L->getHeader()->phis()) {
      if (!isCompatibleIVType(&Phi, IVSrc))
        continue;
      Instruction *PostIncV = dyn_cast<Instruction>(
          Phi.getIncomingValueForBlock(L->getLoopLatch()));
      if (!PostIncV || (SE.getSCEV(PostIncV) != SE.getSCEV(IVSrc)))
        continue;
      Value *IVOper = IVSrc;
      Type *PostIncTy = PostIncV->getType();
      if (IVTy != PostIncTy) {
        assert(PostIncTy->isPointerTy() && "mixing int/ptr IV types");
        IRBuilder<> Builder(L->getLoopLatch()->getTerminator());
        Builder.SetCurrentDebugLocation(PostIncV->getDebugLoc());
        IVOper = Builder.CreatePointerCast(IVSrc, PostIncTy, "lsr.chain");
      }
      Phi.replaceUsesOfWith(PostIncV, IVOper);
      DeadInsts.emplace_back(PostIncV);
    }
  }
}

void LSRInstance::CollectFixupsAndInitialFormulae() {
  for (const IVStrideUse &U : IU) {
    Instruction *UserInst = U.getUser();
    // Skip IV users that are part of profitable IV Chains.
    User::op_iterator UseI =
        find(UserInst->operands(), U.getOperandValToReplace());
    assert(UseI != UserInst->op_end() && "cannot find IV operand");
    if (IVIncSet.count(UseI)) {
      DEBUG(dbgs() << "Use is in profitable chain: " << **UseI << '\n');
      continue;
    }

    LSRUse::KindType Kind = LSRUse::Basic;
    MemAccessTy AccessTy;
    if (isAddressUse(TTI, UserInst, U.getOperandValToReplace())) {
      Kind = LSRUse::Address;
      AccessTy = getAccessType(TTI, UserInst);
    }

    const SCEV *S = IU.getExpr(U);
    PostIncLoopSet TmpPostIncLoops = U.getPostIncLoops();

    // Equality (== and !=) ICmps are special. We can rewrite (i == N) as
    // (N - i == 0), and this allows (N - i) to be the expression that we work
    // with rather than just N or i, so we can consider the register
    // requirements for both N and i at the same time. Limiting this code to
    // equality icmps is not a problem because all interesting loops use
    // equality icmps, thanks to IndVarSimplify.
    if (ICmpInst *CI = dyn_cast<ICmpInst>(UserInst))
      if (CI->isEquality()) {
        // Swap the operands if needed to put the OperandValToReplace on the
        // left, for consistency.
        Value *NV = CI->getOperand(1);
        if (NV == U.getOperandValToReplace()) {
          CI->setOperand(1, CI->getOperand(0));
          CI->setOperand(0, NV);
          NV = CI->getOperand(1);
          Changed = true;
        }

        // x == y  -->  x - y == 0
        const SCEV *N = SE.getSCEV(NV);
        if (SE.isLoopInvariant(N, L) && isSafeToExpand(N, SE)) {
          // S is normalized, so normalize N before folding it into S
          // to keep the result normalized.
          N = normalizeForPostIncUse(N, TmpPostIncLoops, SE);
          Kind = LSRUse::ICmpZero;
          S = SE.getMinusSCEV(N, S);
        }

        // -1 and the negations of all interesting strides (except the negation
        // of -1) are now also interesting.
        for (size_t i = 0, e = Factors.size(); i != e; ++i)
          if (Factors[i] != -1)
            Factors.insert(-(uint64_t)Factors[i]);
        Factors.insert(-1);
      }

    // Get or create an LSRUse.
    std::pair<size_t, int64_t> P = getUse(S, Kind, AccessTy);
    size_t LUIdx = P.first;
    int64_t Offset = P.second;
    LSRUse &LU = Uses[LUIdx];

    // Record the fixup.
    LSRFixup &LF = LU.getNewFixup();
    LF.UserInst = UserInst;
    LF.OperandValToReplace = U.getOperandValToReplace();
    LF.PostIncLoops = TmpPostIncLoops;
    LF.Offset = Offset;
    LU.AllFixupsOutsideLoop &= LF.isUseFullyOutsideLoop(L);

    if (!LU.WidestFixupType ||
        SE.getTypeSizeInBits(LU.WidestFixupType) <
        SE.getTypeSizeInBits(LF.OperandValToReplace->getType()))
      LU.WidestFixupType = LF.OperandValToReplace->getType();

    // If this is the first use of this LSRUse, give it a formula.
    if (LU.Formulae.empty()) {
      InsertInitialFormula(S, LU, LUIdx);
      CountRegisters(LU.Formulae.back(), LUIdx);
    }
  }

  DEBUG(print_fixups(dbgs()));
}

/// Insert a formula for the given expression into the given use, separating out
/// loop-variant portions from loop-invariant and loop-computable portions.
void
LSRInstance::InsertInitialFormula(const SCEV *S, LSRUse &LU, size_t LUIdx) {
  // Mark uses whose expressions cannot be expanded.
  if (!isSafeToExpand(S, SE))
    LU.RigidFormula = true;

  Formula F;
  F.initialMatch(S, L, SE);
  bool Inserted = InsertFormula(LU, LUIdx, F);
  assert(Inserted && "Initial formula already exists!"); (void)Inserted;
}

/// Insert a simple single-register formula for the given expression into the
/// given use.
void
LSRInstance::InsertSupplementalFormula(const SCEV *S,
                                       LSRUse &LU, size_t LUIdx) {
  Formula F;
  F.BaseRegs.push_back(S);
  F.HasBaseReg = true;
  bool Inserted = InsertFormula(LU, LUIdx, F);
  assert(Inserted && "Supplemental formula already exists!"); (void)Inserted;
}

/// Note which registers are used by the given formula, updating RegUses.
void LSRInstance::CountRegisters(const Formula &F, size_t LUIdx) {
  if (F.ScaledReg)
    RegUses.countRegister(F.ScaledReg, LUIdx);
  for (const SCEV *BaseReg : F.BaseRegs)
    RegUses.countRegister(BaseReg, LUIdx);
}

/// If the given formula has not yet been inserted, add it to the list, and
/// return true. Return false otherwise.
bool LSRInstance::InsertFormula(LSRUse &LU, unsigned LUIdx, const Formula &F) {
  // Do not insert formula that we will not be able to expand.
  assert(isLegalUse(TTI, LU.MinOffset, LU.MaxOffset, LU.Kind, LU.AccessTy, F) &&
         "Formula is illegal");

  if (!LU.InsertFormula(F, *L))
    return false;

  CountRegisters(F, LUIdx);
  return true;
}

/// Check for other uses of loop-invariant values which we're tracking. These
/// other uses will pin these values in registers, making them less profitable
/// for elimination.
/// TODO: This currently misses non-constant addrec step registers.
/// TODO: Should this give more weight to users inside the loop?
void
LSRInstance::CollectLoopInvariantFixupsAndFormulae() {
  SmallVector<const SCEV *, 8> Worklist(RegUses.begin(), RegUses.end());
  SmallPtrSet<const SCEV *, 32> Visited;

  while (!Worklist.empty()) {
    const SCEV *S = Worklist.pop_back_val();

    // Don't process the same SCEV twice
    if (!Visited.insert(S).second)
      continue;

    if (const SCEVNAryExpr *N = dyn_cast<SCEVNAryExpr>(S))
      Worklist.append(N->op_begin(), N->op_end());
    else if (const SCEVCastExpr *C = dyn_cast<SCEVCastExpr>(S))
      Worklist.push_back(C->getOperand());
    else if (const SCEVUDivExpr *D = dyn_cast<SCEVUDivExpr>(S)) {
      Worklist.push_back(D->getLHS());
      Worklist.push_back(D->getRHS());
    } else if (const SCEVUnknown *US = dyn_cast<SCEVUnknown>(S)) {
      const Value *V = US->getValue();
      if (const Instruction *Inst = dyn_cast<Instruction>(V)) {
        // Look for instructions defined outside the loop.
        if (L->contains(Inst)) continue;
      } else if (isa<UndefValue>(V))
        // Undef doesn't have a live range, so it doesn't matter.
        continue;
      for (const Use &U : V->uses()) {
        const Instruction *UserInst = dyn_cast<Instruction>(U.getUser());
        // Ignore non-instructions.
        if (!UserInst)
          continue;
        // Ignore instructions in other functions (as can happen with
        // Constants).
        if (UserInst->getParent()->getParent() != L->getHeader()->getParent())
          continue;
        // Ignore instructions not dominated by the loop.
        const BasicBlock *UseBB = !isa<PHINode>(UserInst) ?
          UserInst->getParent() :
          cast<PHINode>(UserInst)->getIncomingBlock(
            PHINode::getIncomingValueNumForOperand(U.getOperandNo()));
        if (!DT.dominates(L->getHeader(), UseBB))
          continue;
        // Don't bother if the instruction is in a BB which ends in an EHPad.
        if (UseBB->getTerminator()->isEHPad())
          continue;
        // Don't bother rewriting PHIs in catchswitch blocks.
        if (isa<CatchSwitchInst>(UserInst->getParent()->getTerminator()))
          continue;
        // Ignore uses which are part of other SCEV expressions, to avoid
        // analyzing them multiple times.
        if (SE.isSCEVable(UserInst->getType())) {
          const SCEV *UserS = SE.getSCEV(const_cast<Instruction *>(UserInst));
          // If the user is a no-op, look through to its uses.
          if (!isa<SCEVUnknown>(UserS))
            continue;
          if (UserS == US) {
            Worklist.push_back(
              SE.getUnknown(const_cast<Instruction *>(UserInst)));
            continue;
          }
        }
        // Ignore icmp instructions which are already being analyzed.
        if (const ICmpInst *ICI = dyn_cast<ICmpInst>(UserInst)) {
          unsigned OtherIdx = !U.getOperandNo();
          Value *OtherOp = const_cast<Value *>(ICI->getOperand(OtherIdx));
          if (SE.hasComputableLoopEvolution(SE.getSCEV(OtherOp), L))
            continue;
        }

        std::pair<size_t, int64_t> P = getUse(
            S, LSRUse::Basic, MemAccessTy());
        size_t LUIdx = P.first;
        int64_t Offset = P.second;
        LSRUse &LU = Uses[LUIdx];
        LSRFixup &LF = LU.getNewFixup();
        LF.UserInst = const_cast<Instruction *>(UserInst);
        LF.OperandValToReplace = U;
        LF.Offset = Offset;
        LU.AllFixupsOutsideLoop &= LF.isUseFullyOutsideLoop(L);
        if (!LU.WidestFixupType ||
            SE.getTypeSizeInBits(LU.WidestFixupType) <
            SE.getTypeSizeInBits(LF.OperandValToReplace->getType()))
          LU.WidestFixupType = LF.OperandValToReplace->getType();
        InsertSupplementalFormula(US, LU, LUIdx);
        CountRegisters(LU.Formulae.back(), Uses.size() - 1);
        break;
      }
    }
  }
}

/// Split S into subexpressions which can be pulled out into separate
/// registers. If C is non-null, multiply each subexpression by C.
///
/// Return remainder expression after factoring the subexpressions captured by
/// Ops. If Ops is complete, return NULL.
static const SCEV *CollectSubexprs(const SCEV *S, const SCEVConstant *C,
                                   SmallVectorImpl<const SCEV *> &Ops,
                                   const Loop *L,
                                   ScalarEvolution &SE,
                                   unsigned Depth = 0) {
  // Arbitrarily cap recursion to protect compile time.
  if (Depth >= 3)
    return S;

  if (const SCEVAddExpr *Add = dyn_cast<SCEVAddExpr>(S)) {
    // Break out add operands.
    for (const SCEV *S : Add->operands()) {
      const SCEV *Remainder = CollectSubexprs(S, C, Ops, L, SE, Depth+1);
      if (Remainder)
        Ops.push_back(C ? SE.getMulExpr(C, Remainder) : Remainder);
    }
    return nullptr;
  } else if (const SCEVAddRecExpr *AR = dyn_cast<SCEVAddRecExpr>(S)) {
    // Split a non-zero base out of an addrec.
    if (AR->getStart()->isZero() || !AR->isAffine())
      return S;

    const SCEV *Remainder = CollectSubexprs(AR->getStart(),
                                            C, Ops, L, SE, Depth+1);
    // Split the non-zero AddRec unless it is part of a nested recurrence that
    // does not pertain to this loop.
    if (Remainder && (AR->getLoop() == L || !isa<SCEVAddRecExpr>(Remainder))) {
      Ops.push_back(C ? SE.getMulExpr(C, Remainder) : Remainder);
      Remainder = nullptr;
    }
    if (Remainder != AR->getStart()) {
      if (!Remainder)
        Remainder = SE.getConstant(AR->getType(), 0);
      return SE.getAddRecExpr(Remainder,
                              AR->getStepRecurrence(SE),
                              AR->getLoop(),
                              //FIXME: AR->getNoWrapFlags(SCEV::FlagNW)
                              SCEV::FlagAnyWrap);
    }
  } else if (const SCEVMulExpr *Mul = dyn_cast<SCEVMulExpr>(S)) {
    // Break (C * (a + b + c)) into C*a + C*b + C*c.
    if (Mul->getNumOperands() != 2)
      return S;
    if (const SCEVConstant *Op0 =
        dyn_cast<SCEVConstant>(Mul->getOperand(0))) {
      C = C ? cast<SCEVConstant>(SE.getMulExpr(C, Op0)) : Op0;
      const SCEV *Remainder =
        CollectSubexprs(Mul->getOperand(1), C, Ops, L, SE, Depth+1);
      if (Remainder)
        Ops.push_back(SE.getMulExpr(C, Remainder));
      return nullptr;
    }
  }
  return S;
}

/// \brief Helper function for LSRInstance::GenerateReassociations.
void LSRInstance::GenerateReassociationsImpl(LSRUse &LU, unsigned LUIdx,
                                             const Formula &Base,
                                             unsigned Depth, size_t Idx,
                                             bool IsScaledReg) {
  const SCEV *BaseReg = IsScaledReg ? Base.ScaledReg : Base.BaseRegs[Idx];
  SmallVector<const SCEV *, 8> AddOps;
  const SCEV *Remainder = CollectSubexprs(BaseReg, nullptr, AddOps, L, SE);
  if (Remainder)
    AddOps.push_back(Remainder);

  if (AddOps.size() == 1)
    return;

  for (SmallVectorImpl<const SCEV *>::const_iterator J = AddOps.begin(),
                                                     JE = AddOps.end();
       J != JE; ++J) {
    // Loop-variant "unknown" values are uninteresting; we won't be able to
    // do anything meaningful with them.
    if (isa<SCEVUnknown>(*J) && !SE.isLoopInvariant(*J, L))
      continue;

    // Don't pull a constant into a register if the constant could be folded
    // into an immediate field.
    if (isAlwaysFoldable(TTI, SE, LU.MinOffset, LU.MaxOffset, LU.Kind,
                         LU.AccessTy, *J, Base.getNumRegs() > 1))
      continue;

    // Collect all operands except *J.
    SmallVector<const SCEV *, 8> InnerAddOps(
        ((const SmallVector<const SCEV *, 8> &)AddOps).begin(), J);
    InnerAddOps.append(std::next(J),
                       ((const SmallVector<const SCEV *, 8> &)AddOps).end());

    // Don't leave just a constant behind in a register if the constant could
    // be folded into an immediate field.
    if (InnerAddOps.size() == 1 &&
        isAlwaysFoldable(TTI, SE, LU.MinOffset, LU.MaxOffset, LU.Kind,
                         LU.AccessTy, InnerAddOps[0], Base.getNumRegs() > 1))
      continue;

    const SCEV *InnerSum = SE.getAddExpr(InnerAddOps);
    if (InnerSum->isZero())
      continue;
    Formula F = Base;

    // Add the remaining pieces of the add back into the new formula.
    const SCEVConstant *InnerSumSC = dyn_cast<SCEVConstant>(InnerSum);
    if (InnerSumSC && SE.getTypeSizeInBits(InnerSumSC->getType()) <= 64 &&
        TTI.isLegalAddImmediate((uint64_t)F.UnfoldedOffset +
                                InnerSumSC->getValue()->getZExtValue())) {
      F.UnfoldedOffset =
          (uint64_t)F.UnfoldedOffset + InnerSumSC->getValue()->getZExtValue();
      if (IsScaledReg)
        F.ScaledReg = nullptr;
      else
        F.BaseRegs.erase(F.BaseRegs.begin() + Idx);
    } else if (IsScaledReg)
      F.ScaledReg = InnerSum;
    else
      F.BaseRegs[Idx] = InnerSum;

    // Add J as its own register, or an unfolded immediate.
    const SCEVConstant *SC = dyn_cast<SCEVConstant>(*J);
    if (SC && SE.getTypeSizeInBits(SC->getType()) <= 64 &&
        TTI.isLegalAddImmediate((uint64_t)F.UnfoldedOffset +
                                SC->getValue()->getZExtValue()))
      F.UnfoldedOffset =
          (uint64_t)F.UnfoldedOffset + SC->getValue()->getZExtValue();
    else
      F.BaseRegs.push_back(*J);
    // We may have changed the number of register in base regs, adjust the
    // formula accordingly.
    F.canonicalize(*L);

    if (InsertFormula(LU, LUIdx, F))
      // If that formula hadn't been seen before, recurse to find more like
      // it.
      GenerateReassociations(LU, LUIdx, LU.Formulae.back(), Depth + 1);
  }
}

/// Split out subexpressions from adds and the bases of addrecs.
void LSRInstance::GenerateReassociations(LSRUse &LU, unsigned LUIdx,
                                         Formula Base, unsigned Depth) {
  assert(Base.isCanonical(*L) && "Input must be in the canonical form");
  // Arbitrarily cap recursion to protect compile time.
  if (Depth >= 3)
    return;

  for (size_t i = 0, e = Base.BaseRegs.size(); i != e; ++i)
    GenerateReassociationsImpl(LU, LUIdx, Base, Depth, i);

  if (Base.Scale == 1)
    GenerateReassociationsImpl(LU, LUIdx, Base, Depth,
                               /* Idx */ -1, /* IsScaledReg */ true);
}

///  Generate a formula consisting of all of the loop-dominating registers added
/// into a single register.
void LSRInstance::GenerateCombinations(LSRUse &LU, unsigned LUIdx,
                                       Formula Base) {
  // This method is only interesting on a plurality of registers.
  if (Base.BaseRegs.size() + (Base.Scale == 1) <= 1)
    return;

  // Flatten the representation, i.e., reg1 + 1*reg2 => reg1 + reg2, before
  // processing the formula.
  Base.unscale();
  Formula F = Base;
  F.BaseRegs.clear();
  SmallVector<const SCEV *, 4> Ops;
  for (const SCEV *BaseReg : Base.BaseRegs) {
    if (SE.properlyDominates(BaseReg, L->getHeader()) &&
        !SE.hasComputableLoopEvolution(BaseReg, L))
      Ops.push_back(BaseReg);
    else
      F.BaseRegs.push_back(BaseReg);
  }
  if (Ops.size() > 1) {
    const SCEV *Sum = SE.getAddExpr(Ops);
    // TODO: If Sum is zero, it probably means ScalarEvolution missed an
    // opportunity to fold something. For now, just ignore such cases
    // rather than proceed with zero in a register.
    if (!Sum->isZero()) {
      F.BaseRegs.push_back(Sum);
      F.canonicalize(*L);
      (void)InsertFormula(LU, LUIdx, F);
    }
  }
}

/// \brief Helper function for LSRInstance::GenerateSymbolicOffsets.
void LSRInstance::GenerateSymbolicOffsetsImpl(LSRUse &LU, unsigned LUIdx,
                                              const Formula &Base, size_t Idx,
                                              bool IsScaledReg) {
  const SCEV *G = IsScaledReg ? Base.ScaledReg : Base.BaseRegs[Idx];
  GlobalValue *GV = ExtractSymbol(G, SE);
  if (G->isZero() || !GV)
    return;
  Formula F = Base;
  F.BaseGV = GV;
  if (!isLegalUse(TTI, LU.MinOffset, LU.MaxOffset, LU.Kind, LU.AccessTy, F))
    return;
  if (IsScaledReg)
    F.ScaledReg = G;
  else
    F.BaseRegs[Idx] = G;
  (void)InsertFormula(LU, LUIdx, F);
}

/// Generate reuse formulae using symbolic offsets.
void LSRInstance::GenerateSymbolicOffsets(LSRUse &LU, unsigned LUIdx,
                                          Formula Base) {
  // We can't add a symbolic offset if the address already contains one.
  if (Base.BaseGV) return;

  for (size_t i = 0, e = Base.BaseRegs.size(); i != e; ++i)
    GenerateSymbolicOffsetsImpl(LU, LUIdx, Base, i);
  if (Base.Scale == 1)
    GenerateSymbolicOffsetsImpl(LU, LUIdx, Base, /* Idx */ -1,
                                /* IsScaledReg */ true);
}

/// \brief Helper function for LSRInstance::GenerateConstantOffsets.
void LSRInstance::GenerateConstantOffsetsImpl(
    LSRUse &LU, unsigned LUIdx, const Formula &Base,
    const SmallVectorImpl<int64_t> &Worklist, size_t Idx, bool IsScaledReg) {
  const SCEV *G = IsScaledReg ? Base.ScaledReg : Base.BaseRegs[Idx];
  for (int64_t Offset : Worklist) {
    Formula F = Base;
    F.BaseOffset = (uint64_t)Base.BaseOffset - Offset;
    if (isLegalUse(TTI, LU.MinOffset - Offset, LU.MaxOffset - Offset, LU.Kind,
                   LU.AccessTy, F)) {
      // Add the offset to the base register.
      const SCEV *NewG = SE.getAddExpr(SE.getConstant(G->getType(), Offset), G);
      // If it cancelled out, drop the base register, otherwise update it.
      if (NewG->isZero()) {
        if (IsScaledReg) {
          F.Scale = 0;
          F.ScaledReg = nullptr;
        } else
          F.deleteBaseReg(F.BaseRegs[Idx]);
        F.canonicalize(*L);
      } else if (IsScaledReg)
        F.ScaledReg = NewG;
      else
        F.BaseRegs[Idx] = NewG;

      (void)InsertFormula(LU, LUIdx, F);
    }
  }

  int64_t Imm = ExtractImmediate(G, SE);
  if (G->isZero() || Imm == 0)
    return;
  Formula F = Base;
  F.BaseOffset = (uint64_t)F.BaseOffset + Imm;
  if (!isLegalUse(TTI, LU.MinOffset, LU.MaxOffset, LU.Kind, LU.AccessTy, F))
    return;
  if (IsScaledReg)
    F.ScaledReg = G;
  else
    F.BaseRegs[Idx] = G;
  (void)InsertFormula(LU, LUIdx, F);
}

/// GenerateConstantOffsets - Generate reuse formulae using symbolic offsets.
void LSRInstance::GenerateConstantOffsets(LSRUse &LU, unsigned LUIdx,
                                          Formula Base) {
  // TODO: For now, just add the min and max offset, because it usually isn't
  // worthwhile looking at everything inbetween.
  SmallVector<int64_t, 2> Worklist;
  Worklist.push_back(LU.MinOffset);
  if (LU.MaxOffset != LU.MinOffset)
    Worklist.push_back(LU.MaxOffset);

  for (size_t i = 0, e = Base.BaseRegs.size(); i != e; ++i)
    GenerateConstantOffsetsImpl(LU, LUIdx, Base, Worklist, i);
  if (Base.Scale == 1)
    GenerateConstantOffsetsImpl(LU, LUIdx, Base, Worklist, /* Idx */ -1,
                                /* IsScaledReg */ true);
}

/// For ICmpZero, check to see if we can scale up the comparison. For example, x
/// == y -> x*c == y*c.
void LSRInstance::GenerateICmpZeroScales(LSRUse &LU, unsigned LUIdx,
                                         Formula Base) {
  if (LU.Kind != LSRUse::ICmpZero) return;

  // Determine the integer type for the base formula.
  Type *IntTy = Base.getType();
  if (!IntTy) return;
  if (SE.getTypeSizeInBits(IntTy) > 64) return;

  // Don't do this if there is more than one offset.
  if (LU.MinOffset != LU.MaxOffset) return;

  // Check if transformation is valid. It is illegal to multiply pointer.
  if (Base.ScaledReg && Base.ScaledReg->getType()->isPointerTy())
    return;
  for (const SCEV *BaseReg : Base.BaseRegs)
    if (BaseReg->getType()->isPointerTy())
      return;
  assert(!Base.BaseGV && "ICmpZero use is not legal!");

  // Check each interesting stride.
  for (int64_t Factor : Factors) {
    // Check that the multiplication doesn't overflow.
    if (Base.BaseOffset == std::numeric_limits<int64_t>::min() && Factor == -1)
      continue;
    int64_t NewBaseOffset = (uint64_t)Base.BaseOffset * Factor;
    if (NewBaseOffset / Factor != Base.BaseOffset)
      continue;
    // If the offset will be truncated at this use, check that it is in bounds.
    if (!IntTy->isPointerTy() &&
        !ConstantInt::isValueValidForType(IntTy, NewBaseOffset))
      continue;

    // Check that multiplying with the use offset doesn't overflow.
    int64_t Offset = LU.MinOffset;
    if (Offset == std::numeric_limits<int64_t>::min() && Factor == -1)
      continue;
    Offset = (uint64_t)Offset * Factor;
    if (Offset / Factor != LU.MinOffset)
      continue;
    // If the offset will be truncated at this use, check that it is in bounds.
    if (!IntTy->isPointerTy() &&
        !ConstantInt::isValueValidForType(IntTy, Offset))
      continue;

    Formula F = Base;
    F.BaseOffset = NewBaseOffset;

    // Check that this scale is legal.
    if (!isLegalUse(TTI, Offset, Offset, LU.Kind, LU.AccessTy, F))
      continue;

    // Compensate for the use having MinOffset built into it.
    F.BaseOffset = (uint64_t)F.BaseOffset + Offset - LU.MinOffset;

    const SCEV *FactorS = SE.getConstant(IntTy, Factor);

    // Check that multiplying with each base register doesn't overflow.
    for (size_t i = 0, e = F.BaseRegs.size(); i != e; ++i) {
      F.BaseRegs[i] = SE.getMulExpr(F.BaseRegs[i], FactorS);
      if (getExactSDiv(F.BaseRegs[i], FactorS, SE) != Base.BaseRegs[i])
        goto next;
    }

    // Check that multiplying with the scaled register doesn't overflow.
    if (F.ScaledReg) {
      F.ScaledReg = SE.getMulExpr(F.ScaledReg, FactorS);
      if (getExactSDiv(F.ScaledReg, FactorS, SE) != Base.ScaledReg)
        continue;
    }

    // Check that multiplying with the unfolded offset doesn't overflow.
    if (F.UnfoldedOffset != 0) {
      if (F.UnfoldedOffset == std::numeric_limits<int64_t>::min() &&
          Factor == -1)
        continue;
      F.UnfoldedOffset = (uint64_t)F.UnfoldedOffset * Factor;
      if (F.UnfoldedOffset / Factor != Base.UnfoldedOffset)
        continue;
      // If the offset will be truncated, check that it is in bounds.
      if (!IntTy->isPointerTy() &&
          !ConstantInt::isValueValidForType(IntTy, F.UnfoldedOffset))
        continue;
    }

    // If we make it here and it's legal, add it.
    (void)InsertFormula(LU, LUIdx, F);
  next:;
  }
}

/// Generate stride factor reuse formulae by making use of scaled-offset address
/// modes, for example.
void LSRInstance::GenerateScales(LSRUse &LU, unsigned LUIdx, Formula Base) {
  // Determine the integer type for the base formula.
  Type *IntTy = Base.getType();
  if (!IntTy) return;

  // If this Formula already has a scaled register, we can't add another one.
  // Try to unscale the formula to generate a better scale.
  if (Base.Scale != 0 && !Base.unscale())
    return;

  assert(Base.Scale == 0 && "unscale did not did its job!");

  // Check each interesting stride.
  for (int64_t Factor : Factors) {
    Base.Scale = Factor;
    Base.HasBaseReg = Base.BaseRegs.size() > 1;
    // Check whether this scale is going to be legal.
    if (!isLegalUse(TTI, LU.MinOffset, LU.MaxOffset, LU.Kind, LU.AccessTy,
                    Base)) {
      // As a special-case, handle special out-of-loop Basic users specially.
      // TODO: Reconsider this special case.
      if (LU.Kind == LSRUse::Basic &&
          isLegalUse(TTI, LU.MinOffset, LU.MaxOffset, LSRUse::Special,
                     LU.AccessTy, Base) &&
          LU.AllFixupsOutsideLoop)
        LU.Kind = LSRUse::Special;
      else
        continue;
    }
    // For an ICmpZero, negating a solitary base register won't lead to
    // new solutions.
    if (LU.Kind == LSRUse::ICmpZero &&
        !Base.HasBaseReg && Base.BaseOffset == 0 && !Base.BaseGV)
      continue;
    // For each addrec base reg, if its loop is current loop, apply the scale.
    for (size_t i = 0, e = Base.BaseRegs.size(); i != e; ++i) {
      const SCEVAddRecExpr *AR = dyn_cast<SCEVAddRecExpr>(Base.BaseRegs[i]);
      if (AR && (AR->getLoop() == L || LU.AllFixupsOutsideLoop)) {
        const SCEV *FactorS = SE.getConstant(IntTy, Factor);
        if (FactorS->isZero())
          continue;
        // Divide out the factor, ignoring high bits, since we'll be
        // scaling the value back up in the end.
        if (const SCEV *Quotient = getExactSDiv(AR, FactorS, SE, true)) {
          // TODO: This could be optimized to avoid all the copying.
          Formula F = Base;
          F.ScaledReg = Quotient;
          F.deleteBaseReg(F.BaseRegs[i]);
          // The canonical representation of 1*reg is reg, which is already in
          // Base. In that case, do not try to insert the formula, it will be
          // rejected anyway.
          if (F.Scale == 1 && (F.BaseRegs.empty() ||
                               (AR->getLoop() != L && LU.AllFixupsOutsideLoop)))
            continue;
          // If AllFixupsOutsideLoop is true and F.Scale is 1, we may generate
          // non canonical Formula with ScaledReg's loop not being L.
          if (F.Scale == 1 && LU.AllFixupsOutsideLoop)
            F.canonicalize(*L);
          (void)InsertFormula(LU, LUIdx, F);
        }
      }
    }
  }
}

/// Generate reuse formulae from different IV types.
void LSRInstance::GenerateTruncates(LSRUse &LU, unsigned LUIdx, Formula Base) {
  // Don't bother truncating symbolic values.
  if (Base.BaseGV) return;

  // Determine the integer type for the base formula.
  Type *DstTy = Base.getType();
  if (!DstTy) return;
  DstTy = SE.getEffectiveSCEVType(DstTy);

  for (Type *SrcTy : Types) {
    if (SrcTy != DstTy && TTI.isTruncateFree(SrcTy, DstTy)) {
      Formula F = Base;

      if (F.ScaledReg) F.ScaledReg = SE.getAnyExtendExpr(F.ScaledReg, SrcTy);
      for (const SCEV *&BaseReg : F.BaseRegs)
        BaseReg = SE.getAnyExtendExpr(BaseReg, SrcTy);

      // TODO: This assumes we've done basic processing on all uses and
      // have an idea what the register usage is.
      if (!F.hasRegsUsedByUsesOtherThan(LUIdx, RegUses))
        continue;

      F.canonicalize(*L);
      (void)InsertFormula(LU, LUIdx, F);
    }
  }
}

namespace {

/// Helper class for GenerateCrossUseConstantOffsets. It's used to defer
/// modifications so that the search phase doesn't have to worry about the data
/// structures moving underneath it.
struct WorkItem {
  size_t LUIdx;
  int64_t Imm;
  const SCEV *OrigReg;

  WorkItem(size_t LI, int64_t I, const SCEV *R)
      : LUIdx(LI), Imm(I), OrigReg(R) {}

  void print(raw_ostream &OS) const;
  void dump() const;
};

} // end anonymous namespace

#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
void WorkItem::print(raw_ostream &OS) const {
  OS << "in formulae referencing " << *OrigReg << " in use " << LUIdx
     << " , add offset " << Imm;
}

LLVM_DUMP_METHOD void WorkItem::dump() const {
  print(errs()); errs() << '\n';
}
#endif

/// Look for registers which are a constant distance apart and try to form reuse
/// opportunities between them.
void LSRInstance::GenerateCrossUseConstantOffsets() {
  // Group the registers by their value without any added constant offset.
  using ImmMapTy = std::map<int64_t, const SCEV *>;

  DenseMap<const SCEV *, ImmMapTy> Map;
  DenseMap<const SCEV *, SmallBitVector> UsedByIndicesMap;
  SmallVector<const SCEV *, 8> Sequence;
  for (const SCEV *Use : RegUses) {
    const SCEV *Reg = Use; // Make a copy for ExtractImmediate to modify.
    int64_t Imm = ExtractImmediate(Reg, SE);
    auto Pair = Map.insert(std::make_pair(Reg, ImmMapTy()));
    if (Pair.second)
      Sequence.push_back(Reg);
    Pair.first->second.insert(std::make_pair(Imm, Use));
    UsedByIndicesMap[Reg] |= RegUses.getUsedByIndices(Use);
  }

  // Now examine each set of registers with the same base value. Build up
  // a list of work to do and do the work in a separate step so that we're
  // not adding formulae and register counts while we're searching.
  SmallVector<WorkItem, 32> WorkItems;
  SmallSet<std::pair<size_t, int64_t>, 32> UniqueItems;
  for (const SCEV *Reg : Sequence) {
    const ImmMapTy &Imms = Map.find(Reg)->second;

    // It's not worthwhile looking for reuse if there's only one offset.
    if (Imms.size() == 1)
      continue;

    DEBUG(dbgs() << "Generating cross-use offsets for " << *Reg << ':';
          for (const auto &Entry : Imms)
            dbgs() << ' ' << Entry.first;
          dbgs() << '\n');

    // Examine each offset.
    for (ImmMapTy::const_iterator J = Imms.begin(), JE = Imms.end();
         J != JE; ++J) {
      const SCEV *OrigReg = J->second;

      int64_t JImm = J->first;
      const SmallBitVector &UsedByIndices = RegUses.getUsedByIndices(OrigReg);

      if (!isa<SCEVConstant>(OrigReg) &&
          UsedByIndicesMap[Reg].count() == 1) {
        DEBUG(dbgs() << "Skipping cross-use reuse for " << *OrigReg << '\n');
        continue;
      }

      // Conservatively examine offsets between this orig reg a few selected
      // other orig regs.
      ImmMapTy::const_iterator OtherImms[] = {
        Imms.begin(), std::prev(Imms.end()),
        Imms.lower_bound((Imms.begin()->first + std::prev(Imms.end())->first) /
                         2)
      };
      for (size_t i = 0, e = array_lengthof(OtherImms); i != e; ++i) {
        ImmMapTy::const_iterator M = OtherImms[i];
        if (M == J || M == JE) continue;

        // Compute the difference between the two.
        int64_t Imm = (uint64_t)JImm - M->first;
        for (unsigned LUIdx : UsedByIndices.set_bits())
          // Make a memo of this use, offset, and register tuple.
          if (UniqueItems.insert(std::make_pair(LUIdx, Imm)).second)
            WorkItems.push_back(WorkItem(LUIdx, Imm, OrigReg));
      }
    }
  }

  Map.clear();
  Sequence.clear();
  UsedByIndicesMap.clear();
  UniqueItems.clear();

  // Now iterate through the worklist and add new formulae.
  for (const WorkItem &WI : WorkItems) {
    size_t LUIdx = WI.LUIdx;
    LSRUse &LU = Uses[LUIdx];
    int64_t Imm = WI.Imm;
    const SCEV *OrigReg = WI.OrigReg;

    Type *IntTy = SE.getEffectiveSCEVType(OrigReg->getType());
    const SCEV *NegImmS = SE.getSCEV(ConstantInt::get(IntTy, -(uint64_t)Imm));
    unsigned BitWidth = SE.getTypeSizeInBits(IntTy);

    // TODO: Use a more targeted data structure.
    for (size_t L = 0, LE = LU.Formulae.size(); L != LE; ++L) {
      Formula F = LU.Formulae[L];
      // FIXME: The code for the scaled and unscaled registers looks
      // very similar but slightly different. Investigate if they
      // could be merged. That way, we would not have to unscale the
      // Formula.
      F.unscale();
      // Use the immediate in the scaled register.
      if (F.ScaledReg == OrigReg) {
        int64_t Offset = (uint64_t)F.BaseOffset + Imm * (uint64_t)F.Scale;
        // Don't create 50 + reg(-50).
        if (F.referencesReg(SE.getSCEV(
                   ConstantInt::get(IntTy, -(uint64_t)Offset))))
          continue;
        Formula NewF = F;
        NewF.BaseOffset = Offset;
        if (!isLegalUse(TTI, LU.MinOffset, LU.MaxOffset, LU.Kind, LU.AccessTy,
                        NewF))
          continue;
        NewF.ScaledReg = SE.getAddExpr(NegImmS, NewF.ScaledReg);

        // If the new scale is a constant in a register, and adding the constant
        // value to the immediate would produce a value closer to zero than the
        // immediate itself, then the formula isn't worthwhile.
        if (const SCEVConstant *C = dyn_cast<SCEVConstant>(NewF.ScaledReg))
          if (C->getValue()->isNegative() != (NewF.BaseOffset < 0) &&
              (C->getAPInt().abs() * APInt(BitWidth, F.Scale))
                  .ule(std::abs(NewF.BaseOffset)))
            continue;

        // OK, looks good.
        NewF.canonicalize(*this->L);
        (void)InsertFormula(LU, LUIdx, NewF);
      } else {
        // Use the immediate in a base register.
        for (size_t N = 0, NE = F.BaseRegs.size(); N != NE; ++N) {
          const SCEV *BaseReg = F.BaseRegs[N];
          if (BaseReg != OrigReg)
            continue;
          Formula NewF = F;
          NewF.BaseOffset = (uint64_t)NewF.BaseOffset + Imm;
          if (!isLegalUse(TTI, LU.MinOffset, LU.MaxOffset,
                          LU.Kind, LU.AccessTy, NewF)) {
            if (!TTI.isLegalAddImmediate((uint64_t)NewF.UnfoldedOffset + Imm))
              continue;
            NewF = F;
            NewF.UnfoldedOffset = (uint64_t)NewF.UnfoldedOffset + Imm;
          }
          NewF.BaseRegs[N] = SE.getAddExpr(NegImmS, BaseReg);

          // If the new formula has a constant in a register, and adding the
          // constant value to the immediate would produce a value closer to
          // zero than the immediate itself, then the formula isn't worthwhile.
          for (const SCEV *NewReg : NewF.BaseRegs)
            if (const SCEVConstant *C = dyn_cast<SCEVConstant>(NewReg))
              if ((C->getAPInt() + NewF.BaseOffset)
                      .abs()
                      .slt(std::abs(NewF.BaseOffset)) &&
                  (C->getAPInt() + NewF.BaseOffset).countTrailingZeros() >=
                      countTrailingZeros<uint64_t>(NewF.BaseOffset))
                goto skip_formula;

          // Ok, looks good.
          NewF.canonicalize(*this->L);
          (void)InsertFormula(LU, LUIdx, NewF);
          break;
        skip_formula:;
        }
      }
    }
  }
}

/// Generate formulae for each use.
void
LSRInstance::GenerateAllReuseFormulae() {
  // This is split into multiple loops so that hasRegsUsedByUsesOtherThan
  // queries are more precise.
  for (size_t LUIdx = 0, NumUses = Uses.size(); LUIdx != NumUses; ++LUIdx) {
    LSRUse &LU = Uses[LUIdx];
    for (size_t i = 0, f = LU.Formulae.size(); i != f; ++i)
      GenerateReassociations(LU, LUIdx, LU.Formulae[i]);
    for (size_t i = 0, f = LU.Formulae.size(); i != f; ++i)
      GenerateCombinations(LU, LUIdx, LU.Formulae[i]);
  }
  for (size_t LUIdx = 0, NumUses = Uses.size(); LUIdx != NumUses; ++LUIdx) {
    LSRUse &LU = Uses[LUIdx];
    for (size_t i = 0, f = LU.Formulae.size(); i != f; ++i)
      GenerateSymbolicOffsets(LU, LUIdx, LU.Formulae[i]);
    for (size_t i = 0, f = LU.Formulae.size(); i != f; ++i)
      GenerateConstantOffsets(LU, LUIdx, LU.Formulae[i]);
    for (size_t i = 0, f = LU.Formulae.size(); i != f; ++i)
      GenerateICmpZeroScales(LU, LUIdx, LU.Formulae[i]);
    for (size_t i = 0, f = LU.Formulae.size(); i != f; ++i)
      GenerateScales(LU, LUIdx, LU.Formulae[i]);
  }
  for (size_t LUIdx = 0, NumUses = Uses.size(); LUIdx != NumUses; ++LUIdx) {
    LSRUse &LU = Uses[LUIdx];
    for (size_t i = 0, f = LU.Formulae.size(); i != f; ++i)
      GenerateTruncates(LU, LUIdx, LU.Formulae[i]);
  }

  GenerateCrossUseConstantOffsets();

  DEBUG(dbgs() << "\n"
                  "After generating reuse formulae:\n";
        print_uses(dbgs()));
}

/// If there are multiple formulae with the same set of registers used
/// by other uses, pick the best one and delete the others.
void LSRInstance::FilterOutUndesirableDedicatedRegisters() {
  DenseSet<const SCEV *> VisitedRegs;
  SmallPtrSet<const SCEV *, 16> Regs;
  SmallPtrSet<const SCEV *, 16> LoserRegs;
#ifndef NDEBUG
  bool ChangedFormulae = false;
#endif

  // Collect the best formula for each unique set of shared registers. This
  // is reset for each use.
  using BestFormulaeTy =
      DenseMap<SmallVector<const SCEV *, 4>, size_t, UniquifierDenseMapInfo>;

  BestFormulaeTy BestFormulae;

  for (size_t LUIdx = 0, NumUses = Uses.size(); LUIdx != NumUses; ++LUIdx) {
    LSRUse &LU = Uses[LUIdx];
    DEBUG(dbgs() << "Filtering for use "; LU.print(dbgs()); dbgs() << '\n');

    bool Any = false;
    for (size_t FIdx = 0, NumForms = LU.Formulae.size();
         FIdx != NumForms; ++FIdx) {
      Formula &F = LU.Formulae[FIdx];

      // Some formulas are instant losers. For example, they may depend on
      // nonexistent AddRecs from other loops. These need to be filtered
      // immediately, otherwise heuristics could choose them over others leading
      // to an unsatisfactory solution. Passing LoserRegs into RateFormula here
      // avoids the need to recompute this information across formulae using the
      // same bad AddRec. Passing LoserRegs is also essential unless we remove
      // the corresponding bad register from the Regs set.
      Cost CostF;
      Regs.clear();
      CostF.RateFormula(TTI, F, Regs, VisitedRegs, L, SE, DT, LU, &LoserRegs);
      if (CostF.isLoser()) {
        // During initial formula generation, undesirable formulae are generated
        // by uses within other loops that have some non-trivial address mode or
        // use the postinc form of the IV. LSR needs to provide these formulae
        // as the basis of rediscovering the desired formula that uses an AddRec
        // corresponding to the existing phi. Once all formulae have been
        // generated, these initial losers may be pruned.
        DEBUG(dbgs() << "  Filtering loser "; F.print(dbgs());
              dbgs() << "\n");
      }
      else {
        SmallVector<const SCEV *, 4> Key;
        for (const SCEV *Reg : F.BaseRegs) {
          if (RegUses.isRegUsedByUsesOtherThan(Reg, LUIdx))
            Key.push_back(Reg);
        }
        if (F.ScaledReg &&
            RegUses.isRegUsedByUsesOtherThan(F.ScaledReg, LUIdx))
          Key.push_back(F.ScaledReg);
        // Unstable sort by host order ok, because this is only used for
        // uniquifying.
        std::sort(Key.begin(), Key.end());

        std::pair<BestFormulaeTy::const_iterator, bool> P =
          BestFormulae.insert(std::make_pair(Key, FIdx));
        if (P.second)
          continue;

        Formula &Best = LU.Formulae[P.first->second];

        Cost CostBest;
        Regs.clear();
        CostBest.RateFormula(TTI, Best, Regs, VisitedRegs, L, SE, DT, LU);
        if (CostF.isLess(CostBest, TTI))
          std::swap(F, Best);
        DEBUG(dbgs() << "  Filtering out formula "; F.print(dbgs());
              dbgs() << "\n"
                        "    in favor of formula "; Best.print(dbgs());
              dbgs() << '\n');
      }
#ifndef NDEBUG
      ChangedFormulae = true;
#endif
      LU.DeleteFormula(F);
      --FIdx;
      --NumForms;
      Any = true;
    }

    // Now that we've filtered out some formulae, recompute the Regs set.
    if (Any)
      LU.RecomputeRegs(LUIdx, RegUses);

    // Reset this to prepare for the next use.
    BestFormulae.clear();
  }

  DEBUG(if (ChangedFormulae) {
          dbgs() << "\n"
                    "After filtering out undesirable candidates:\n";
          print_uses(dbgs());
        });
}

// This is a rough guess that seems to work fairly well.
static const size_t ComplexityLimit = std::numeric_limits<uint16_t>::max();

/// Estimate the worst-case number of solutions the solver might have to
/// consider. It almost never considers this many solutions because it prune the
/// search space, but the pruning isn't always sufficient.
size_t LSRInstance::EstimateSearchSpaceComplexity() const {
  size_t Power = 1;
  for (const LSRUse &LU : Uses) {
    size_t FSize = LU.Formulae.size();
    if (FSize >= ComplexityLimit) {
      Power = ComplexityLimit;
      break;
    }
    Power *= FSize;
    if (Power >= ComplexityLimit)
      break;
  }
  return Power;
}

/// When one formula uses a superset of the registers of another formula, it
/// won't help reduce register pressure (though it may not necessarily hurt
/// register pressure); remove it to simplify the system.
void LSRInstance::NarrowSearchSpaceByDetectingSupersets() {
  if (EstimateSearchSpaceComplexity() >= ComplexityLimit) {
    DEBUG(dbgs() << "The search space is too complex.\n");

    DEBUG(dbgs() << "Narrowing the search space by eliminating formulae "
                    "which use a superset of registers used by other "
                    "formulae.\n");

    for (size_t LUIdx = 0, NumUses = Uses.size(); LUIdx != NumUses; ++LUIdx) {
      LSRUse &LU = Uses[LUIdx];
      bool Any = false;
      for (size_t i = 0, e = LU.Formulae.size(); i != e; ++i) {
        Formula &F = LU.Formulae[i];
        // Look for a formula with a constant or GV in a register. If the use
        // also has a formula with that same value in an immediate field,
        // delete the one that uses a register.
        for (SmallVectorImpl<const SCEV *>::const_iterator
             I = F.BaseRegs.begin(), E = F.BaseRegs.end(); I != E; ++I) {
          if (const SCEVConstant *C = dyn_cast<SCEVConstant>(*I)) {
            Formula NewF = F;
            NewF.BaseOffset += C->getValue()->getSExtValue();
            NewF.BaseRegs.erase(NewF.BaseRegs.begin() +
                                (I - F.BaseRegs.begin()));
            if (LU.HasFormulaWithSameRegs(NewF)) {
              DEBUG(dbgs() << "  Deleting "; F.print(dbgs()); dbgs() << '\n');
              LU.DeleteFormula(F);
              --i;
              --e;
              Any = true;
              break;
            }
          } else if (const SCEVUnknown *U = dyn_cast<SCEVUnknown>(*I)) {
            if (GlobalValue *GV = dyn_cast<GlobalValue>(U->getValue()))
              if (!F.BaseGV) {
                Formula NewF = F;
                NewF.BaseGV = GV;
                NewF.BaseRegs.erase(NewF.BaseRegs.begin() +
                                    (I - F.BaseRegs.begin()));
                if (LU.HasFormulaWithSameRegs(NewF)) {
                  DEBUG(dbgs() << "  Deleting "; F.print(dbgs());
                        dbgs() << '\n');
                  LU.DeleteFormula(F);
                  --i;
                  --e;
                  Any = true;
                  break;
                }
              }
          }
        }
      }
      if (Any)
        LU.RecomputeRegs(LUIdx, RegUses);
    }

    DEBUG(dbgs() << "After pre-selection:\n";
          print_uses(dbgs()));
  }
}

/// When there are many registers for expressions like A, A+1, A+2, etc.,
/// allocate a single register for them.
void LSRInstance::NarrowSearchSpaceByCollapsingUnrolledCode() {
  if (EstimateSearchSpaceComplexity() < ComplexityLimit)
    return;

  DEBUG(dbgs() << "The search space is too complex.\n"
                  "Narrowing the search space by assuming that uses separated "
                  "by a constant offset will use the same registers.\n");

  // This is especially useful for unrolled loops.

  for (size_t LUIdx = 0, NumUses = Uses.size(); LUIdx != NumUses; ++LUIdx) {
    LSRUse &LU = Uses[LUIdx];
    for (const Formula &F : LU.Formulae) {
      if (F.BaseOffset == 0 || (F.Scale != 0 && F.Scale != 1))
        continue;

      LSRUse *LUThatHas = FindUseWithSimilarFormula(F, LU);
      if (!LUThatHas)
        continue;

      if (!reconcileNewOffset(*LUThatHas, F.BaseOffset, /*HasBaseReg=*/ false,
                              LU.Kind, LU.AccessTy))
        continue;

      DEBUG(dbgs() << "  Deleting use "; LU.print(dbgs()); dbgs() << '\n');

      LUThatHas->AllFixupsOutsideLoop &= LU.AllFixupsOutsideLoop;

      // Transfer the fixups of LU to LUThatHas.
      for (LSRFixup &Fixup : LU.Fixups) {
        Fixup.Offset += F.BaseOffset;
        LUThatHas->pushFixup(Fixup);
        DEBUG(dbgs() << "New fixup has offset " << Fixup.Offset << '\n');
      }

      // Delete formulae from the new use which are no longer legal.
      bool Any = false;
      for (size_t i = 0, e = LUThatHas->Formulae.size(); i != e; ++i) {
        Formula &F = LUThatHas->Formulae[i];
        if (!isLegalUse(TTI, LUThatHas->MinOffset, LUThatHas->MaxOffset,
                        LUThatHas->Kind, LUThatHas->AccessTy, F)) {
          DEBUG(dbgs() << "  Deleting "; F.print(dbgs());
                dbgs() << '\n');
          LUThatHas->DeleteFormula(F);
          --i;
          --e;
          Any = true;
        }
      }

      if (Any)
        LUThatHas->RecomputeRegs(LUThatHas - &Uses.front(), RegUses);

      // Delete the old use.
      DeleteUse(LU, LUIdx);
      --LUIdx;
      --NumUses;
      break;
    }
  }

  DEBUG(dbgs() << "After pre-selection:\n"; print_uses(dbgs()));
}

/// Call FilterOutUndesirableDedicatedRegisters again, if necessary, now that
/// we've done more filtering, as it may be able to find more formulae to
/// eliminate.
void LSRInstance::NarrowSearchSpaceByRefilteringUndesirableDedicatedRegisters(){
  if (EstimateSearchSpaceComplexity() >= ComplexityLimit) {
    DEBUG(dbgs() << "The search space is too complex.\n");

    DEBUG(dbgs() << "Narrowing the search space by re-filtering out "
                    "undesirable dedicated registers.\n");

    FilterOutUndesirableDedicatedRegisters();

    DEBUG(dbgs() << "After pre-selection:\n";
          print_uses(dbgs()));
  }
}

/// If a LSRUse has multiple formulae with the same ScaledReg and Scale.
/// Pick the best one and delete the others.
/// This narrowing heuristic is to keep as many formulae with different
/// Scale and ScaledReg pair as possible while narrowing the search space.
/// The benefit is that it is more likely to find out a better solution
/// from a formulae set with more Scale and ScaledReg variations than
/// a formulae set with the same Scale and ScaledReg. The picking winner
/// reg heurstic will often keep the formulae with the same Scale and
/// ScaledReg and filter others, and we want to avoid that if possible.
void LSRInstance::NarrowSearchSpaceByFilterFormulaWithSameScaledReg() {
  if (EstimateSearchSpaceComplexity() < ComplexityLimit)
    return;

  DEBUG(dbgs() << "The search space is too complex.\n"
                  "Narrowing the search space by choosing the best Formula "
                  "from the Formulae with the same Scale and ScaledReg.\n");

  // Map the "Scale * ScaledReg" pair to the best formula of current LSRUse.
  using BestFormulaeTy = DenseMap<std::pair<const SCEV *, int64_t>, size_t>;

  BestFormulaeTy BestFormulae;
#ifndef NDEBUG
  bool ChangedFormulae = false;
#endif
  DenseSet<const SCEV *> VisitedRegs;
  SmallPtrSet<const SCEV *, 16> Regs;

  for (size_t LUIdx = 0, NumUses = Uses.size(); LUIdx != NumUses; ++LUIdx) {
    LSRUse &LU = Uses[LUIdx];
    DEBUG(dbgs() << "Filtering for use "; LU.print(dbgs()); dbgs() << '\n');

    // Return true if Formula FA is better than Formula FB.
    auto IsBetterThan = [&](Formula &FA, Formula &FB) {
      // First we will try to choose the Formula with fewer new registers.
      // For a register used by current Formula, the more the register is
      // shared among LSRUses, the less we increase the register number
      // counter of the formula.
      size_t FARegNum = 0;
      for (const SCEV *Reg : FA.BaseRegs) {
        const SmallBitVector &UsedByIndices = RegUses.getUsedByIndices(Reg);
        FARegNum += (NumUses - UsedByIndices.count() + 1);
      }
      size_t FBRegNum = 0;
      for (const SCEV *Reg : FB.BaseRegs) {
        const SmallBitVector &UsedByIndices = RegUses.getUsedByIndices(Reg);
        FBRegNum += (NumUses - UsedByIndices.count() + 1);
      }
      if (FARegNum != FBRegNum)
        return FARegNum < FBRegNum;

      // If the new register numbers are the same, choose the Formula with
      // less Cost.
      Cost CostFA, CostFB;
      Regs.clear();
      CostFA.RateFormula(TTI, FA, Regs, VisitedRegs, L, SE, DT, LU);
      Regs.clear();
      CostFB.RateFormula(TTI, FB, Regs, VisitedRegs, L, SE, DT, LU);
      return CostFA.isLess(CostFB, TTI);
    };

    bool Any = false;
    for (size_t FIdx = 0, NumForms = LU.Formulae.size(); FIdx != NumForms;
         ++FIdx) {
      Formula &F = LU.Formulae[FIdx];
      if (!F.ScaledReg)
        continue;
      auto P = BestFormulae.insert({{F.ScaledReg, F.Scale}, FIdx});
      if (P.second)
        continue;

      Formula &Best = LU.Formulae[P.first->second];
      if (IsBetterThan(F, Best))
        std::swap(F, Best);
      DEBUG(dbgs() << "  Filtering out formula "; F.print(dbgs());
            dbgs() << "\n"
                      "    in favor of formula ";
            Best.print(dbgs()); dbgs() << '\n');
#ifndef NDEBUG
      ChangedFormulae = true;
#endif
      LU.DeleteFormula(F);
      --FIdx;
      --NumForms;
      Any = true;
    }
    if (Any)
      LU.RecomputeRegs(LUIdx, RegUses);

    // Reset this to prepare for the next use.
    BestFormulae.clear();
  }

  DEBUG(if (ChangedFormulae) {
    dbgs() << "\n"
              "After filtering out undesirable candidates:\n";
    print_uses(dbgs());
  });
}

/// The function delete formulas with high registers number expectation.
/// Assuming we don't know the value of each formula (already delete
/// all inefficient), generate probability of not selecting for each
/// register.
/// For example,
/// Use1:
///  reg(a) + reg({0,+,1})
///  reg(a) + reg({-1,+,1}) + 1
///  reg({a,+,1})
/// Use2:
///  reg(b) + reg({0,+,1})
///  reg(b) + reg({-1,+,1}) + 1
///  reg({b,+,1})
/// Use3:
///  reg(c) + reg(b) + reg({0,+,1})
///  reg(c) + reg({b,+,1})
///
/// Probability of not selecting
///                 Use1   Use2    Use3
/// reg(a)         (1/3) *   1   *   1
/// reg(b)           1   * (1/3) * (1/2)
/// reg({0,+,1})   (2/3) * (2/3) * (1/2)
/// reg({-1,+,1})  (2/3) * (2/3) *   1
/// reg({a,+,1})   (2/3) *   1   *   1
/// reg({b,+,1})     1   * (2/3) * (2/3)
/// reg(c)           1   *   1   *   0
///
/// Now count registers number mathematical expectation for each formula:
/// Note that for each use we exclude probability if not selecting for the use.
/// For example for Use1 probability for reg(a) would be just 1 * 1 (excluding
/// probabilty 1/3 of not selecting for Use1).
/// Use1:
///  reg(a) + reg({0,+,1})          1 + 1/3       -- to be deleted
///  reg(a) + reg({-1,+,1}) + 1     1 + 4/9       -- to be deleted
///  reg({a,+,1})                   1
/// Use2:
///  reg(b) + reg({0,+,1})          1/2 + 1/3     -- to be deleted
///  reg(b) + reg({-1,+,1}) + 1     1/2 + 2/3     -- to be deleted
///  reg({b,+,1})                   2/3
/// Use3:
///  reg(c) + reg(b) + reg({0,+,1}) 1 + 1/3 + 4/9 -- to be deleted
///  reg(c) + reg({b,+,1})          1 + 2/3
void LSRInstance::NarrowSearchSpaceByDeletingCostlyFormulas() {
  if (EstimateSearchSpaceComplexity() < ComplexityLimit)
    return;
  // Ok, we have too many of formulae on our hands to conveniently handle.
  // Use a rough heuristic to thin out the list.

  // Set of Regs wich will be 100% used in final solution.
  // Used in each formula of a solution (in example above this is reg(c)).
  // We can skip them in calculations.
  SmallPtrSet<const SCEV *, 4> UniqRegs;
  DEBUG(dbgs() << "The search space is too complex.\n");

  // Map each register to probability of not selecting
  DenseMap <const SCEV *, float> RegNumMap;
  for (const SCEV *Reg : RegUses) {
    if (UniqRegs.count(Reg))
      continue;
    float PNotSel = 1;
    for (const LSRUse &LU : Uses) {
      if (!LU.Regs.count(Reg))
        continue;
      float P = LU.getNotSelectedProbability(Reg);
      if (P != 0.0)
        PNotSel *= P;
      else
        UniqRegs.insert(Reg);
    }
    RegNumMap.insert(std::make_pair(Reg, PNotSel));
  }

  DEBUG(dbgs() << "Narrowing the search space by deleting costly formulas\n");

  // Delete formulas where registers number expectation is high.
  for (size_t LUIdx = 0, NumUses = Uses.size(); LUIdx != NumUses; ++LUIdx) {
    LSRUse &LU = Uses[LUIdx];
    // If nothing to delete - continue.
    if (LU.Formulae.size() < 2)
      continue;
    // This is temporary solution to test performance. Float should be
    // replaced with round independent type (based on integers) to avoid
    // different results for different target builds.
    float FMinRegNum = LU.Formulae[0].getNumRegs();
    float FMinARegNum = LU.Formulae[0].getNumRegs();
    size_t MinIdx = 0;
    for (size_t i = 0, e = LU.Formulae.size(); i != e; ++i) {
      Formula &F = LU.Formulae[i];
      float FRegNum = 0;
      float FARegNum = 0;
      for (const SCEV *BaseReg : F.BaseRegs) {
        if (UniqRegs.count(BaseReg))
          continue;
        FRegNum += RegNumMap[BaseReg] / LU.getNotSelectedProbability(BaseReg);
        if (isa<SCEVAddRecExpr>(BaseReg))
          FARegNum +=
              RegNumMap[BaseReg] / LU.getNotSelectedProbability(BaseReg);
      }
      if (const SCEV *ScaledReg = F.ScaledReg) {
        if (!UniqRegs.count(ScaledReg)) {
          FRegNum +=
              RegNumMap[ScaledReg] / LU.getNotSelectedProbability(ScaledReg);
          if (isa<SCEVAddRecExpr>(ScaledReg))
            FARegNum +=
                RegNumMap[ScaledReg] / LU.getNotSelectedProbability(ScaledReg);
        }
      }
      if (FMinRegNum > FRegNum ||
          (FMinRegNum == FRegNum && FMinARegNum > FARegNum)) {
        FMinRegNum = FRegNum;
        FMinARegNum = FARegNum;
        MinIdx = i;
      }
    }
    DEBUG(dbgs() << "  The formula "; LU.Formulae[MinIdx].print(dbgs());
          dbgs() << " with min reg num " << FMinRegNum << '\n');
    if (MinIdx != 0)
      std::swap(LU.Formulae[MinIdx], LU.Formulae[0]);
    while (LU.Formulae.size() != 1) {
      DEBUG(dbgs() << "  Deleting "; LU.Formulae.back().print(dbgs());
            dbgs() << '\n');
      LU.Formulae.pop_back();
    }
    LU.RecomputeRegs(LUIdx, RegUses);
    assert(LU.Formulae.size() == 1 && "Should be exactly 1 min regs formula");
    Formula &F = LU.Formulae[0];
    DEBUG(dbgs() << "  Leaving only "; F.print(dbgs()); dbgs() << '\n');
    // When we choose the formula, the regs become unique.
    UniqRegs.insert(F.BaseRegs.begin(), F.BaseRegs.end());
    if (F.ScaledReg)
      UniqRegs.insert(F.ScaledReg);
  }
  DEBUG(dbgs() << "After pre-selection:\n";
  print_uses(dbgs()));
}

/// Pick a register which seems likely to be profitable, and then in any use
/// which has any reference to that register, delete all formulae which do not
/// reference that register.
void LSRInstance::NarrowSearchSpaceByPickingWinnerRegs() {
  // With all other options exhausted, loop until the system is simple
  // enough to handle.
  SmallPtrSet<const SCEV *, 4> Taken;
  while (EstimateSearchSpaceComplexity() >= ComplexityLimit) {
    // Ok, we have too many of formulae on our hands to conveniently handle.
    // Use a rough heuristic to thin out the list.
    DEBUG(dbgs() << "The search space is too complex.\n");

    // Pick the register which is used by the most LSRUses, which is likely
    // to be a good reuse register candidate.
    const SCEV *Best = nullptr;
    unsigned BestNum = 0;
    for (const SCEV *Reg : RegUses) {
      if (Taken.count(Reg))
        continue;
      if (!Best) {
        Best = Reg;
        BestNum = RegUses.getUsedByIndices(Reg).count();
      } else {
        unsigned Count = RegUses.getUsedByIndices(Reg).count();
        if (Count > BestNum) {
          Best = Reg;
          BestNum = Count;
        }
      }
    }

    DEBUG(dbgs() << "Narrowing the search space by assuming " << *Best
                 << " will yield profitable reuse.\n");
    Taken.insert(Best);

    // In any use with formulae which references this register, delete formulae
    // which don't reference it.
    for (size_t LUIdx = 0, NumUses = Uses.size(); LUIdx != NumUses; ++LUIdx) {
      LSRUse &LU = Uses[LUIdx];
      if (!LU.Regs.count(Best)) continue;

      bool Any = false;
      for (size_t i = 0, e = LU.Formulae.size(); i != e; ++i) {
        Formula &F = LU.Formulae[i];
        if (!F.referencesReg(Best)) {
          DEBUG(dbgs() << "  Deleting "; F.print(dbgs()); dbgs() << '\n');
          LU.DeleteFormula(F);
          --e;
          --i;
          Any = true;
          assert(e != 0 && "Use has no formulae left! Is Regs inconsistent?");
          continue;
        }
      }

      if (Any)
        LU.RecomputeRegs(LUIdx, RegUses);
    }

    DEBUG(dbgs() << "After pre-selection:\n";
          print_uses(dbgs()));
  }
}

/// If there are an extraordinary number of formulae to choose from, use some
/// rough heuristics to prune down the number of formulae. This keeps the main
/// solver from taking an extraordinary amount of time in some worst-case
/// scenarios.
void LSRInstance::NarrowSearchSpaceUsingHeuristics() {
  NarrowSearchSpaceByDetectingSupersets();
  NarrowSearchSpaceByCollapsingUnrolledCode();
  NarrowSearchSpaceByRefilteringUndesirableDedicatedRegisters();
  if (FilterSameScaledReg)
    NarrowSearchSpaceByFilterFormulaWithSameScaledReg();
  if (LSRExpNarrow)
    NarrowSearchSpaceByDeletingCostlyFormulas();
  else
    NarrowSearchSpaceByPickingWinnerRegs();
}

/// This is the recursive solver.
void LSRInstance::SolveRecurse(SmallVectorImpl<const Formula *> &Solution,
                               Cost &SolutionCost,
                               SmallVectorImpl<const Formula *> &Workspace,
                               const Cost &CurCost,
                               const SmallPtrSet<const SCEV *, 16> &CurRegs,
                               DenseSet<const SCEV *> &VisitedRegs) const {
  // Some ideas:
  //  - prune more:
  //    - use more aggressive filtering
  //    - sort the formula so that the most profitable solutions are found first
  //    - sort the uses too
  //  - search faster:
  //    - don't compute a cost, and then compare. compare while computing a cost
  //      and bail early.
  //    - track register sets with SmallBitVector

  const LSRUse &LU = Uses[Workspace.size()];

  // If this use references any register that's already a part of the
  // in-progress solution, consider it a requirement that a formula must
  // reference that register in order to be considered. This prunes out
  // unprofitable searching.
  SmallSetVector<const SCEV *, 4> ReqRegs;
  for (const SCEV *S : CurRegs)
    if (LU.Regs.count(S))
      ReqRegs.insert(S);

  SmallPtrSet<const SCEV *, 16> NewRegs;
  Cost NewCost;
  for (const Formula &F : LU.Formulae) {
    // Ignore formulae which may not be ideal in terms of register reuse of
    // ReqRegs.  The formula should use all required registers before
    // introducing new ones.
    int NumReqRegsToFind = std::min(F.getNumRegs(), ReqRegs.size());
    for (const SCEV *Reg : ReqRegs) {
      if ((F.ScaledReg && F.ScaledReg == Reg) ||
          is_contained(F.BaseRegs, Reg)) {
        --NumReqRegsToFind;
        if (NumReqRegsToFind == 0)
          break;
      }
    }
    if (NumReqRegsToFind != 0) {
      // If none of the formulae satisfied the required registers, then we could
      // clear ReqRegs and try again. Currently, we simply give up in this case.
      continue;
    }

    // Evaluate the cost of the current formula. If it's already worse than
    // the current best, prune the search at that point.
    NewCost = CurCost;
    NewRegs = CurRegs;
    NewCost.RateFormula(TTI, F, NewRegs, VisitedRegs, L, SE, DT, LU);
    if (NewCost.isLess(SolutionCost, TTI)) {
      Workspace.push_back(&F);
      if (Workspace.size() != Uses.size()) {
        SolveRecurse(Solution, SolutionCost, Workspace, NewCost,
                     NewRegs, VisitedRegs);
        if (F.getNumRegs() == 1 && Workspace.size() == 1)
          VisitedRegs.insert(F.ScaledReg ? F.ScaledReg : F.BaseRegs[0]);
      } else {
        DEBUG(dbgs() << "New best at "; NewCost.print(dbgs());
              dbgs() << ".\n Regs:";
              for (const SCEV *S : NewRegs)
                dbgs() << ' ' << *S;
              dbgs() << '\n');

        SolutionCost = NewCost;
        Solution = Workspace;
      }
      Workspace.pop_back();
    }
  }
}

/// Choose one formula from each use. Return the results in the given Solution
/// vector.
void LSRInstance::Solve(SmallVectorImpl<const Formula *> &Solution) const {
  SmallVector<const Formula *, 8> Workspace;
  Cost SolutionCost;
  SolutionCost.Lose();
  Cost CurCost;
  SmallPtrSet<const SCEV *, 16> CurRegs;
  DenseSet<const SCEV *> VisitedRegs;
  Workspace.reserve(Uses.size());

  // SolveRecurse does all the work.
  SolveRecurse(Solution, SolutionCost, Workspace, CurCost,
               CurRegs, VisitedRegs);
  if (Solution.empty()) {
    DEBUG(dbgs() << "\nNo Satisfactory Solution\n");
    return;
  }

  // Ok, we've now made all our decisions.
  DEBUG(dbgs() << "\n"
                  "The chosen solution requires "; SolutionCost.print(dbgs());
        dbgs() << ":\n";
        for (size_t i = 0, e = Uses.size(); i != e; ++i) {
          dbgs() << "  ";
          Uses[i].print(dbgs());
          dbgs() << "\n"
                    "    ";
          Solution[i]->print(dbgs());
          dbgs() << '\n';
        });

  assert(Solution.size() == Uses.size() && "Malformed solution!");
}

/// Helper for AdjustInsertPositionForExpand. Climb up the dominator tree far as
/// we can go while still being dominated by the input positions. This helps
/// canonicalize the insert position, which encourages sharing.
BasicBlock::iterator
LSRInstance::HoistInsertPosition(BasicBlock::iterator IP,
                                 const SmallVectorImpl<Instruction *> &Inputs)
                                                                         const {
  Instruction *Tentative = &*IP;
  while (true) {
    bool AllDominate = true;
    Instruction *BetterPos = nullptr;
    // Don't bother attempting to insert before a catchswitch, their basic block
    // cannot have other non-PHI instructions.
    if (isa<CatchSwitchInst>(Tentative))
      return IP;

    for (Instruction *Inst : Inputs) {
      if (Inst == Tentative || !DT.dominates(Inst, Tentative)) {
        AllDominate = false;
        break;
      }
      // Attempt to find an insert position in the middle of the block,
      // instead of at the end, so that it can be used for other expansions.
      if (Tentative->getParent() == Inst->getParent() &&
          (!BetterPos || !DT.dominates(Inst, BetterPos)))
        BetterPos = &*std::next(BasicBlock::iterator(Inst));
    }
    if (!AllDominate)
      break;
    if (BetterPos)
      IP = BetterPos->getIterator();
    else
      IP = Tentative->getIterator();

    const Loop *IPLoop = LI.getLoopFor(IP->getParent());
    unsigned IPLoopDepth = IPLoop ? IPLoop->getLoopDepth() : 0;

    BasicBlock *IDom;
    for (DomTreeNode *Rung = DT.getNode(IP->getParent()); ; ) {
      if (!Rung) return IP;
      Rung = Rung->getIDom();
      if (!Rung) return IP;
      IDom = Rung->getBlock();

      // Don't climb into a loop though.
      const Loop *IDomLoop = LI.getLoopFor(IDom);
      unsigned IDomDepth = IDomLoop ? IDomLoop->getLoopDepth() : 0;
      if (IDomDepth <= IPLoopDepth &&
          (IDomDepth != IPLoopDepth || IDomLoop == IPLoop))
        break;
    }

    Tentative = IDom->getTerminator();
  }

  return IP;
}

/// Determine an input position which will be dominated by the operands and
/// which will dominate the result.
BasicBlock::iterator
LSRInstance::AdjustInsertPositionForExpand(BasicBlock::iterator LowestIP,
                                           const LSRFixup &LF,
                                           const LSRUse &LU,
                                           SCEVExpander &Rewriter) const {
  // Collect some instructions which must be dominated by the
  // expanding replacement. These must be dominated by any operands that
  // will be required in the expansion.
  SmallVector<Instruction *, 4> Inputs;
  if (Instruction *I = dyn_cast<Instruction>(LF.OperandValToReplace))
    Inputs.push_back(I);
  if (LU.Kind == LSRUse::ICmpZero)
    if (Instruction *I =
          dyn_cast<Instruction>(cast<ICmpInst>(LF.UserInst)->getOperand(1)))
      Inputs.push_back(I);
  if (LF.PostIncLoops.count(L)) {
    if (LF.isUseFullyOutsideLoop(L))
      Inputs.push_back(L->getLoopLatch()->getTerminator());
    else
      Inputs.push_back(IVIncInsertPos);
  }
  // The expansion must also be dominated by the increment positions of any
  // loops it for which it is using post-inc mode.
  for (const Loop *PIL : LF.PostIncLoops) {
    if (PIL == L) continue;

    // Be dominated by the loop exit.
    SmallVector<BasicBlock *, 4> ExitingBlocks;
    PIL->getExitingBlocks(ExitingBlocks);
    if (!ExitingBlocks.empty()) {
      BasicBlock *BB = ExitingBlocks[0];
      for (unsigned i = 1, e = ExitingBlocks.size(); i != e; ++i)
        BB = DT.findNearestCommonDominator(BB, ExitingBlocks[i]);
      Inputs.push_back(BB->getTerminator());
    }
  }

  assert(!isa<PHINode>(LowestIP) && !LowestIP->isEHPad()
         && !isa<DbgInfoIntrinsic>(LowestIP) &&
         "Insertion point must be a normal instruction");

  // Then, climb up the immediate dominator tree as far as we can go while
  // still being dominated by the input positions.
  BasicBlock::iterator IP = HoistInsertPosition(LowestIP, Inputs);

  // Don't insert instructions before PHI nodes.
  while (isa<PHINode>(IP)) ++IP;

  // Ignore landingpad instructions.
  while (IP->isEHPad()) ++IP;

  // Ignore debug intrinsics.
  while (isa<DbgInfoIntrinsic>(IP)) ++IP;

  // Set IP below instructions recently inserted by SCEVExpander. This keeps the
  // IP consistent across expansions and allows the previously inserted
  // instructions to be reused by subsequent expansion.
  while (Rewriter.isInsertedInstruction(&*IP) && IP != LowestIP)
    ++IP;

  return IP;
}

/// Emit instructions for the leading candidate expression for this LSRUse (this
/// is called "expanding").
Value *LSRInstance::Expand(const LSRUse &LU, const LSRFixup &LF,
                           const Formula &F, BasicBlock::iterator IP,
                           SCEVExpander &Rewriter,
                           SmallVectorImpl<WeakTrackingVH> &DeadInsts) const {
  if (LU.RigidFormula)
    return LF.OperandValToReplace;

  // Determine an input position which will be dominated by the operands and
  // which will dominate the result.
  IP = AdjustInsertPositionForExpand(IP, LF, LU, Rewriter);
  Rewriter.setInsertPoint(&*IP);

  // Inform the Rewriter if we have a post-increment use, so that it can
  // perform an advantageous expansion.
  Rewriter.setPostInc(LF.PostIncLoops);

  // This is the type that the user actually needs.
  Type *OpTy = LF.OperandValToReplace->getType();
  // This will be the type that we'll initially expand to.
  Type *Ty = F.getType();
  if (!Ty)
    // No type known; just expand directly to the ultimate type.
    Ty = OpTy;
  else if (SE.getEffectiveSCEVType(Ty) == SE.getEffectiveSCEVType(OpTy))
    // Expand directly to the ultimate type if it's the right size.
    Ty = OpTy;
  // This is the type to do integer arithmetic in.
  Type *IntTy = SE.getEffectiveSCEVType(Ty);

  // Build up a list of operands to add together to form the full base.
  SmallVector<const SCEV *, 8> Ops;

  // Expand the BaseRegs portion.
  for (const SCEV *Reg : F.BaseRegs) {
    assert(!Reg->isZero() && "Zero allocated in a base register!");

    // If we're expanding for a post-inc user, make the post-inc adjustment.
    Reg = denormalizeForPostIncUse(Reg, LF.PostIncLoops, SE);
    Ops.push_back(SE.getUnknown(Rewriter.expandCodeFor(Reg, nullptr)));
  }

  // Expand the ScaledReg portion.
  Value *ICmpScaledV = nullptr;
  if (F.Scale != 0) {
    const SCEV *ScaledS = F.ScaledReg;

    // If we're expanding for a post-inc user, make the post-inc adjustment.
    PostIncLoopSet &Loops = const_cast<PostIncLoopSet &>(LF.PostIncLoops);
    ScaledS = denormalizeForPostIncUse(ScaledS, Loops, SE);

    if (LU.Kind == LSRUse::ICmpZero) {
      // Expand ScaleReg as if it was part of the base regs.
      if (F.Scale == 1)
        Ops.push_back(
            SE.getUnknown(Rewriter.expandCodeFor(ScaledS, nullptr)));
      else {
        // An interesting way of "folding" with an icmp is to use a negated
        // scale, which we'll implement by inserting it into the other operand
        // of the icmp.
        assert(F.Scale == -1 &&
               "The only scale supported by ICmpZero uses is -1!");
        ICmpScaledV = Rewriter.expandCodeFor(ScaledS, nullptr);
      }
    } else {
      // Otherwise just expand the scaled register and an explicit scale,
      // which is expected to be matched as part of the address.

      // Flush the operand list to suppress SCEVExpander hoisting address modes.
      // Unless the addressing mode will not be folded.
      if (!Ops.empty() && LU.Kind == LSRUse::Address &&
          isAMCompletelyFolded(TTI, LU, F)) {
        Value *FullV = Rewriter.expandCodeFor(SE.getAddExpr(Ops), Ty);
        Ops.clear();
        Ops.push_back(SE.getUnknown(FullV));
      }
      ScaledS = SE.getUnknown(Rewriter.expandCodeFor(ScaledS, nullptr));
      if (F.Scale != 1)
        ScaledS =
            SE.getMulExpr(ScaledS, SE.getConstant(ScaledS->getType(), F.Scale));
      Ops.push_back(ScaledS);
    }
  }

  // Expand the GV portion.
  if (F.BaseGV) {
    // Flush the operand list to suppress SCEVExpander hoisting.
    if (!Ops.empty()) {
      Value *FullV = Rewriter.expandCodeFor(SE.getAddExpr(Ops), Ty);
      Ops.clear();
      Ops.push_back(SE.getUnknown(FullV));
    }
    Ops.push_back(SE.getUnknown(F.BaseGV));
  }

  // Flush the operand list to suppress SCEVExpander hoisting of both folded and
  // unfolded offsets. LSR assumes they both live next to their uses.
  if (!Ops.empty()) {
    Value *FullV = Rewriter.expandCodeFor(SE.getAddExpr(Ops), Ty);
    Ops.clear();
    Ops.push_back(SE.getUnknown(FullV));
  }

  // Expand the immediate portion.
  int64_t Offset = (uint64_t)F.BaseOffset + LF.Offset;
  if (Offset != 0) {
    if (LU.Kind == LSRUse::ICmpZero) {
      // The other interesting way of "folding" with an ICmpZero is to use a
      // negated immediate.
      if (!ICmpScaledV)
        ICmpScaledV = ConstantInt::get(IntTy, -(uint64_t)Offset);
      else {
        Ops.push_back(SE.getUnknown(ICmpScaledV));
        ICmpScaledV = ConstantInt::get(IntTy, Offset);
      }
    } else {
      // Just add the immediate values. These again are expected to be matched
      // as part of the address.
      Ops.push_back(SE.getUnknown(ConstantInt::getSigned(IntTy, Offset)));
    }
  }

  // Expand the unfolded offset portion.
  int64_t UnfoldedOffset = F.UnfoldedOffset;
  if (UnfoldedOffset != 0) {
    // Just add the immediate values.
    Ops.push_back(SE.getUnknown(ConstantInt::getSigned(IntTy,
                                                       UnfoldedOffset)));
  }

  // Emit instructions summing all the operands.
  const SCEV *FullS = Ops.empty() ?
                      SE.getConstant(IntTy, 0) :
                      SE.getAddExpr(Ops);
  Value *FullV = Rewriter.expandCodeFor(FullS, Ty);

  // We're done expanding now, so reset the rewriter.
  Rewriter.clearPostInc();

  // An ICmpZero Formula represents an ICmp which we're handling as a
  // comparison against zero. Now that we've expanded an expression for that
  // form, update the ICmp's other operand.
  if (LU.Kind == LSRUse::ICmpZero) {
    ICmpInst *CI = cast<ICmpInst>(LF.UserInst);
    DeadInsts.emplace_back(CI->getOperand(1));
    assert(!F.BaseGV && "ICmp does not support folding a global value and "
                           "a scale at the same time!");
    if (F.Scale == -1) {
      if (ICmpScaledV->getType() != OpTy) {
        Instruction *Cast =
          CastInst::Create(CastInst::getCastOpcode(ICmpScaledV, false,
                                                   OpTy, false),
                           ICmpScaledV, OpTy, "tmp", CI);
        ICmpScaledV = Cast;
      }
      CI->setOperand(1, ICmpScaledV);
    } else {
      // A scale of 1 means that the scale has been expanded as part of the
      // base regs.
      assert((F.Scale == 0 || F.Scale == 1) &&
             "ICmp does not support folding a global value and "
             "a scale at the same time!");
      Constant *C = ConstantInt::getSigned(SE.getEffectiveSCEVType(OpTy),
                                           -(uint64_t)Offset);
      if (C->getType() != OpTy)
        C = ConstantExpr::getCast(CastInst::getCastOpcode(C, false,
                                                          OpTy, false),
                                  C, OpTy);

      CI->setOperand(1, C);
    }
  }

  return FullV;
}

/// Helper for Rewrite. PHI nodes are special because the use of their operands
/// effectively happens in their predecessor blocks, so the expression may need
/// to be expanded in multiple places.
void LSRInstance::RewriteForPHI(
    PHINode *PN, const LSRUse &LU, const LSRFixup &LF, const Formula &F,
    SCEVExpander &Rewriter, SmallVectorImpl<WeakTrackingVH> &DeadInsts) const {
  DenseMap<BasicBlock *, Value *> Inserted;
  for (unsigned i = 0, e = PN->getNumIncomingValues(); i != e; ++i)
    if (PN->getIncomingValue(i) == LF.OperandValToReplace) {
      BasicBlock *BB = PN->getIncomingBlock(i);

      // If this is a critical edge, split the edge so that we do not insert
      // the code on all predecessor/successor paths.  We do this unless this
      // is the canonical backedge for this loop, which complicates post-inc
      // users.
      if (e != 1 && BB->getTerminator()->getNumSuccessors() > 1 &&
          !isa<IndirectBrInst>(BB->getTerminator()) &&
          !isa<CatchSwitchInst>(BB->getTerminator())) {
        BasicBlock *Parent = PN->getParent();
        Loop *PNLoop = LI.getLoopFor(Parent);
        if (!PNLoop || Parent != PNLoop->getHeader()) {
          // Split the critical edge.
          BasicBlock *NewBB = nullptr;
          if (!Parent->isLandingPad()) {
            NewBB = SplitCriticalEdge(BB, Parent,
                                      CriticalEdgeSplittingOptions(&DT, &LI)
                                          .setMergeIdenticalEdges()
                                          .setDontDeleteUselessPHIs());
          } else {
            SmallVector<BasicBlock*, 2> NewBBs;
            SplitLandingPadPredecessors(Parent, BB, "", "", NewBBs, &DT, &LI);
            NewBB = NewBBs[0];
          }
          // If NewBB==NULL, then SplitCriticalEdge refused to split because all
          // phi predecessors are identical. The simple thing to do is skip
          // splitting in this case rather than complicate the API.
          if (NewBB) {
            // If PN is outside of the loop and BB is in the loop, we want to
            // move the block to be immediately before the PHI block, not
            // immediately after BB.
            if (L->contains(BB) && !L->contains(PN))
              NewBB->moveBefore(PN->getParent());

            // Splitting the edge can reduce the number of PHI entries we have.
            e = PN->getNumIncomingValues();
            BB = NewBB;
            i = PN->getBasicBlockIndex(BB);
          }
        }
      }

      std::pair<DenseMap<BasicBlock *, Value *>::iterator, bool> Pair =
        Inserted.insert(std::make_pair(BB, static_cast<Value *>(nullptr)));
      if (!Pair.second)
        PN->setIncomingValue(i, Pair.first->second);
      else {
        Value *FullV = Expand(LU, LF, F, BB->getTerminator()->getIterator(),
                              Rewriter, DeadInsts);

        // If this is reuse-by-noop-cast, insert the noop cast.
        Type *OpTy = LF.OperandValToReplace->getType();
        if (FullV->getType() != OpTy)
          FullV =
            CastInst::Create(CastInst::getCastOpcode(FullV, false,
                                                     OpTy, false),
                             FullV, LF.OperandValToReplace->getType(),
                             "tmp", BB->getTerminator());

        PN->setIncomingValue(i, FullV);
        Pair.first->second = FullV;
      }
    }
}

/// Emit instructions for the leading candidate expression for this LSRUse (this
/// is called "expanding"), and update the UserInst to reference the newly
/// expanded value.
void LSRInstance::Rewrite(const LSRUse &LU, const LSRFixup &LF,
                          const Formula &F, SCEVExpander &Rewriter,
                          SmallVectorImpl<WeakTrackingVH> &DeadInsts) const {
  // First, find an insertion point that dominates UserInst. For PHI nodes,
  // find the nearest block which dominates all the relevant uses.
  if (PHINode *PN = dyn_cast<PHINode>(LF.UserInst)) {
    RewriteForPHI(PN, LU, LF, F, Rewriter, DeadInsts);
  } else {
    Value *FullV =
      Expand(LU, LF, F, LF.UserInst->getIterator(), Rewriter, DeadInsts);

    // If this is reuse-by-noop-cast, insert the noop cast.
    Type *OpTy = LF.OperandValToReplace->getType();
    if (FullV->getType() != OpTy) {
      Instruction *Cast =
        CastInst::Create(CastInst::getCastOpcode(FullV, false, OpTy, false),
                         FullV, OpTy, "tmp", LF.UserInst);
      FullV = Cast;
    }

    // Update the user. ICmpZero is handled specially here (for now) because
    // Expand may have updated one of the operands of the icmp already, and
    // its new value may happen to be equal to LF.OperandValToReplace, in
    // which case doing replaceUsesOfWith leads to replacing both operands
    // with the same value. TODO: Reorganize this.
    if (LU.Kind == LSRUse::ICmpZero)
      LF.UserInst->setOperand(0, FullV);
    else
      LF.UserInst->replaceUsesOfWith(LF.OperandValToReplace, FullV);
  }

  DeadInsts.emplace_back(LF.OperandValToReplace);
}

/// Rewrite all the fixup locations with new values, following the chosen
/// solution.
void LSRInstance::ImplementSolution(
    const SmallVectorImpl<const Formula *> &Solution) {
  // Keep track of instructions we may have made dead, so that
  // we can remove them after we are done working.
  SmallVector<WeakTrackingVH, 16> DeadInsts;

  SCEVExpander Rewriter(SE, L->getHeader()->getModule()->getDataLayout(),
                        "lsr");
#ifndef NDEBUG
  Rewriter.setDebugType(DEBUG_TYPE);
#endif
  Rewriter.disableCanonicalMode();
  Rewriter.enableLSRMode();
  Rewriter.setIVIncInsertPos(L, IVIncInsertPos);

  // Mark phi nodes that terminate chains so the expander tries to reuse them.
  for (const IVChain &Chain : IVChainVec) {
    if (PHINode *PN = dyn_cast<PHINode>(Chain.tailUserInst()))
      Rewriter.setChainedPhi(PN);
  }

  // Expand the new value definitions and update the users.
  for (size_t LUIdx = 0, NumUses = Uses.size(); LUIdx != NumUses; ++LUIdx)
    for (const LSRFixup &Fixup : Uses[LUIdx].Fixups) {
      Rewrite(Uses[LUIdx], Fixup, *Solution[LUIdx], Rewriter, DeadInsts);
      Changed = true;
    }

  for (const IVChain &Chain : IVChainVec) {
    GenerateIVChain(Chain, Rewriter, DeadInsts);
    Changed = true;
  }
  // Clean up after ourselves. This must be done before deleting any
  // instructions.
  Rewriter.clear();

  Changed |= DeleteTriviallyDeadInstructions(DeadInsts);
}

LSRInstance::LSRInstance(Loop *L, IVUsers &IU, ScalarEvolution &SE,
                         DominatorTree &DT, LoopInfo &LI,
                         const TargetTransformInfo &TTI)
    : IU(IU), SE(SE), DT(DT), LI(LI), TTI(TTI), L(L) {
  // If LoopSimplify form is not available, stay out of trouble.
  if (!L->isLoopSimplifyForm())
    return;

  // If there's no interesting work to be done, bail early.
  if (IU.empty()) return;

  // If there's too much analysis to be done, bail early. We won't be able to
  // model the problem anyway.
  unsigned NumUsers = 0;
  for (const IVStrideUse &U : IU) {
    if (++NumUsers > MaxIVUsers) {
      (void)U;
      DEBUG(dbgs() << "LSR skipping loop, too many IV Users in " << U << "\n");
      return;
    }
    // Bail out if we have a PHI on an EHPad that gets a value from a
    // CatchSwitchInst.  Because the CatchSwitchInst cannot be split, there is
    // no good place to stick any instructions.
    if (auto *PN = dyn_cast<PHINode>(U.getUser())) {
       auto *FirstNonPHI = PN->getParent()->getFirstNonPHI();
       if (isa<FuncletPadInst>(FirstNonPHI) ||
           isa<CatchSwitchInst>(FirstNonPHI))
         for (BasicBlock *PredBB : PN->blocks())
           if (isa<CatchSwitchInst>(PredBB->getFirstNonPHI()))
             return;
    }
  }

#ifndef NDEBUG
  // All dominating loops must have preheaders, or SCEVExpander may not be able
  // to materialize an AddRecExpr whose Start is an outer AddRecExpr.
  //
  // IVUsers analysis should only create users that are dominated by simple loop
  // headers. Since this loop should dominate all of its users, its user list
  // should be empty if this loop itself is not within a simple loop nest.
  for (DomTreeNode *Rung = DT.getNode(L->getLoopPreheader());
       Rung; Rung = Rung->getIDom()) {
    BasicBlock *BB = Rung->getBlock();
    const Loop *DomLoop = LI.getLoopFor(BB);
    if (DomLoop && DomLoop->getHeader() == BB) {
      assert(DomLoop->getLoopPreheader() && "LSR needs a simplified loop nest");
    }
  }
#endif // DEBUG

  DEBUG(dbgs() << "\nLSR on loop ";
        L->getHeader()->printAsOperand(dbgs(), /*PrintType=*/false);
        dbgs() << ":\n");

  // First, perform some low-level loop optimizations.
  OptimizeShadowIV();
  OptimizeLoopTermCond();

  // If loop preparation eliminates all interesting IV users, bail.
  if (IU.empty()) return;

  // Skip nested loops until we can model them better with formulae.
  if (!L->empty()) {
    DEBUG(dbgs() << "LSR skipping outer loop " << *L << "\n");
    return;
  }

  // Start collecting data and preparing for the solver.
  CollectChains();
  CollectInterestingTypesAndFactors();
  CollectFixupsAndInitialFormulae();
  CollectLoopInvariantFixupsAndFormulae();

  assert(!Uses.empty() && "IVUsers reported at least one use");
  DEBUG(dbgs() << "LSR found " << Uses.size() << " uses:\n";
        print_uses(dbgs()));

  // Now use the reuse data to generate a bunch of interesting ways
  // to formulate the values needed for the uses.
  GenerateAllReuseFormulae();

  FilterOutUndesirableDedicatedRegisters();
  NarrowSearchSpaceUsingHeuristics();

  SmallVector<const Formula *, 8> Solution;
  Solve(Solution);

  // Release memory that is no longer needed.
  Factors.clear();
  Types.clear();
  RegUses.clear();

  if (Solution.empty())
    return;

#ifndef NDEBUG
  // Formulae should be legal.
  for (const LSRUse &LU : Uses) {
    for (const Formula &F : LU.Formulae)
      assert(isLegalUse(TTI, LU.MinOffset, LU.MaxOffset, LU.Kind, LU.AccessTy,
                        F) && "Illegal formula generated!");
  };
#endif

  // Now that we've decided what we want, make it so.
  ImplementSolution(Solution);
}

#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
void LSRInstance::print_factors_and_types(raw_ostream &OS) const {
  if (Factors.empty() && Types.empty()) return;

  OS << "LSR has identified the following interesting factors and types: ";
  bool First = true;

  for (int64_t Factor : Factors) {
    if (!First) OS << ", ";
    First = false;
    OS << '*' << Factor;
  }

  for (Type *Ty : Types) {
    if (!First) OS << ", ";
    First = false;
    OS << '(' << *Ty << ')';
  }
  OS << '\n';
}

void LSRInstance::print_fixups(raw_ostream &OS) const {
  OS << "LSR is examining the following fixup sites:\n";
  for (const LSRUse &LU : Uses)
    for (const LSRFixup &LF : LU.Fixups) {
      dbgs() << "  ";
      LF.print(OS);
      OS << '\n';
    }
}

void LSRInstance::print_uses(raw_ostream &OS) const {
  OS << "LSR is examining the following uses:\n";
  for (const LSRUse &LU : Uses) {
    dbgs() << "  ";
    LU.print(OS);
    OS << '\n';
    for (const Formula &F : LU.Formulae) {
      OS << "    ";
      F.print(OS);
      OS << '\n';
    }
  }
}

void LSRInstance::print(raw_ostream &OS) const {
  print_factors_and_types(OS);
  print_fixups(OS);
  print_uses(OS);
}

LLVM_DUMP_METHOD void LSRInstance::dump() const {
  print(errs()); errs() << '\n';
}
#endif

namespace {

class LoopStrengthReduce : public LoopPass {
public:
  static char ID; // Pass ID, replacement for typeid

  LoopStrengthReduce();

private:
  bool runOnLoop(Loop *L, LPPassManager &LPM) override;
  void getAnalysisUsage(AnalysisUsage &AU) const override;
};

} // end anonymous namespace

LoopStrengthReduce::LoopStrengthReduce() : LoopPass(ID) {
  initializeLoopStrengthReducePass(*PassRegistry::getPassRegistry());
}

void LoopStrengthReduce::getAnalysisUsage(AnalysisUsage &AU) const {
  // We split critical edges, so we change the CFG.  However, we do update
  // many analyses if they are around.
  AU.addPreservedID(LoopSimplifyID);

  AU.addRequired<LoopInfoWrapperPass>();
  AU.addPreserved<LoopInfoWrapperPass>();
  AU.addRequiredID(LoopSimplifyID);
  AU.addRequired<DominatorTreeWrapperPass>();
  AU.addPreserved<DominatorTreeWrapperPass>();
  AU.addRequired<ScalarEvolutionWrapperPass>();
  AU.addPreserved<ScalarEvolutionWrapperPass>();
  // Requiring LoopSimplify a second time here prevents IVUsers from running
  // twice, since LoopSimplify was invalidated by running ScalarEvolution.
  AU.addRequiredID(LoopSimplifyID);
  AU.addRequired<IVUsersWrapperPass>();
  AU.addPreserved<IVUsersWrapperPass>();
  AU.addRequired<TargetTransformInfoWrapperPass>();
}

static bool ReduceLoopStrength(Loop *L, IVUsers &IU, ScalarEvolution &SE,
                               DominatorTree &DT, LoopInfo &LI,
                               const TargetTransformInfo &TTI) {
  bool Changed = false;

  // Run the main LSR transformation.
  Changed |= LSRInstance(L, IU, SE, DT, LI, TTI).getChanged();

  // Remove any extra phis created by processing inner loops.
  Changed |= DeleteDeadPHIs(L->getHeader());
  if (EnablePhiElim && L->isLoopSimplifyForm()) {
    SmallVector<WeakTrackingVH, 16> DeadInsts;
    const DataLayout &DL = L->getHeader()->getModule()->getDataLayout();
    SCEVExpander Rewriter(SE, DL, "lsr");
#ifndef NDEBUG
    Rewriter.setDebugType(DEBUG_TYPE);
#endif
    unsigned numFolded = Rewriter.replaceCongruentIVs(L, &DT, DeadInsts, &TTI);
    if (numFolded) {
      Changed = true;
      DeleteTriviallyDeadInstructions(DeadInsts);
      DeleteDeadPHIs(L->getHeader());
    }
  }
  return Changed;
}

bool LoopStrengthReduce::runOnLoop(Loop *L, LPPassManager & /*LPM*/) {
  if (skipLoop(L))
    return false;

  auto &IU = getAnalysis<IVUsersWrapperPass>().getIU();
  auto &SE = getAnalysis<ScalarEvolutionWrapperPass>().getSE();
  auto &DT = getAnalysis<DominatorTreeWrapperPass>().getDomTree();
  auto &LI = getAnalysis<LoopInfoWrapperPass>().getLoopInfo();
  const auto &TTI = getAnalysis<TargetTransformInfoWrapperPass>().getTTI(
      *L->getHeader()->getParent());
  return ReduceLoopStrength(L, IU, SE, DT, LI, TTI);
}

PreservedAnalyses LoopStrengthReducePass::run(Loop &L, LoopAnalysisManager &AM,
                                              LoopStandardAnalysisResults &AR,
                                              LPMUpdater &) {
  if (!ReduceLoopStrength(&L, AM.getResult<IVUsersAnalysis>(L, AR), AR.SE,
                          AR.DT, AR.LI, AR.TTI))
    return PreservedAnalyses::all();

  return getLoopPassPreservedAnalyses();
}

char LoopStrengthReduce::ID = 0;

INITIALIZE_PASS_BEGIN(LoopStrengthReduce, "loop-reduce",
                      "Loop Strength Reduction", false, false)
INITIALIZE_PASS_DEPENDENCY(TargetTransformInfoWrapperPass)
INITIALIZE_PASS_DEPENDENCY(DominatorTreeWrapperPass)
INITIALIZE_PASS_DEPENDENCY(ScalarEvolutionWrapperPass)
INITIALIZE_PASS_DEPENDENCY(IVUsersWrapperPass)
INITIALIZE_PASS_DEPENDENCY(LoopInfoWrapperPass)
INITIALIZE_PASS_DEPENDENCY(LoopSimplify)
INITIALIZE_PASS_END(LoopStrengthReduce, "loop-reduce",
                    "Loop Strength Reduction", false, false)

Pass *llvm::createLoopStrengthReducePass() { return new LoopStrengthReduce(); }