aboutsummaryrefslogtreecommitdiffstats
path: root/contrib/libs/llvm14/lib/CodeGen/RegAllocGreedy.cpp
blob: 7870574df5b27a115c72894bcf1e54740937e0da (plain) (blame)
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
//===- RegAllocGreedy.cpp - greedy register allocator ---------------------===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
//
// This file defines the RAGreedy function pass for register allocation in
// optimized builds.
//
//===----------------------------------------------------------------------===//

#include "RegAllocGreedy.h"
#include "AllocationOrder.h"
#include "InterferenceCache.h"
#include "LiveDebugVariables.h"
#include "RegAllocBase.h"
#include "RegAllocEvictionAdvisor.h"
#include "SpillPlacement.h"
#include "SplitKit.h"
#include "llvm/ADT/ArrayRef.h"
#include "llvm/ADT/BitVector.h"
#include "llvm/ADT/DenseMap.h"
#include "llvm/ADT/IndexedMap.h"
#include "llvm/ADT/MapVector.h"
#include "llvm/ADT/SetVector.h"
#include "llvm/ADT/SmallPtrSet.h"
#include "llvm/ADT/SmallSet.h"
#include "llvm/ADT/SmallVector.h"
#include "llvm/ADT/Statistic.h"
#include "llvm/ADT/StringRef.h"
#include "llvm/Analysis/AliasAnalysis.h"
#include "llvm/Analysis/OptimizationRemarkEmitter.h"
#include "llvm/CodeGen/CalcSpillWeights.h"
#include "llvm/CodeGen/EdgeBundles.h"
#include "llvm/CodeGen/LiveInterval.h"
#include "llvm/CodeGen/LiveIntervalUnion.h"
#include "llvm/CodeGen/LiveIntervals.h"
#include "llvm/CodeGen/LiveRangeEdit.h"
#include "llvm/CodeGen/LiveRegMatrix.h"
#include "llvm/CodeGen/LiveStacks.h"
#include "llvm/CodeGen/MachineBasicBlock.h"
#include "llvm/CodeGen/MachineBlockFrequencyInfo.h"
#include "llvm/CodeGen/MachineDominators.h"
#include "llvm/CodeGen/MachineFrameInfo.h"
#include "llvm/CodeGen/MachineFunction.h"
#include "llvm/CodeGen/MachineFunctionPass.h"
#include "llvm/CodeGen/MachineInstr.h"
#include "llvm/CodeGen/MachineLoopInfo.h"
#include "llvm/CodeGen/MachineOperand.h"
#include "llvm/CodeGen/MachineOptimizationRemarkEmitter.h"
#include "llvm/CodeGen/MachineRegisterInfo.h"
#include "llvm/CodeGen/RegAllocRegistry.h"
#include "llvm/CodeGen/RegisterClassInfo.h"
#include "llvm/CodeGen/SlotIndexes.h"
#include "llvm/CodeGen/Spiller.h"
#include "llvm/CodeGen/TargetInstrInfo.h"
#include "llvm/CodeGen/TargetRegisterInfo.h"
#include "llvm/CodeGen/TargetSubtargetInfo.h"
#include "llvm/CodeGen/VirtRegMap.h"
#include "llvm/IR/DebugInfoMetadata.h"
#include "llvm/IR/Function.h"
#include "llvm/IR/LLVMContext.h"
#include "llvm/MC/MCRegisterInfo.h"
#include "llvm/Pass.h"
#include "llvm/Support/BlockFrequency.h"
#include "llvm/Support/BranchProbability.h"
#include "llvm/Support/CommandLine.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/MathExtras.h"
#include "llvm/Support/Timer.h"
#include "llvm/Support/raw_ostream.h"
#include "llvm/Target/TargetMachine.h"
#include <algorithm>
#include <cassert>
#include <cstdint>
#include <memory>
#include <queue>
#include <tuple>
#include <utility>

using namespace llvm;

#define DEBUG_TYPE "regalloc"

STATISTIC(NumGlobalSplits, "Number of split global live ranges");
STATISTIC(NumLocalSplits,  "Number of split local live ranges");
STATISTIC(NumEvicted,      "Number of interferences evicted");

static cl::opt<SplitEditor::ComplementSpillMode> SplitSpillMode(
    "split-spill-mode", cl::Hidden,
    cl::desc("Spill mode for splitting live ranges"),
    cl::values(clEnumValN(SplitEditor::SM_Partition, "default", "Default"),
               clEnumValN(SplitEditor::SM_Size, "size", "Optimize for size"),
               clEnumValN(SplitEditor::SM_Speed, "speed", "Optimize for speed")),
    cl::init(SplitEditor::SM_Speed));

static cl::opt<unsigned>
LastChanceRecoloringMaxDepth("lcr-max-depth", cl::Hidden,
                             cl::desc("Last chance recoloring max depth"),
                             cl::init(5));

static cl::opt<unsigned> LastChanceRecoloringMaxInterference(
    "lcr-max-interf", cl::Hidden,
    cl::desc("Last chance recoloring maximum number of considered"
             " interference at a time"),
    cl::init(8));

static cl::opt<bool> ExhaustiveSearch(
    "exhaustive-register-search", cl::NotHidden,
    cl::desc("Exhaustive Search for registers bypassing the depth "
             "and interference cutoffs of last chance recoloring"),
    cl::Hidden);

static cl::opt<bool> EnableDeferredSpilling(
    "enable-deferred-spilling", cl::Hidden,
    cl::desc("Instead of spilling a variable right away, defer the actual "
             "code insertion to the end of the allocation. That way the "
             "allocator might still find a suitable coloring for this "
             "variable because of other evicted variables."),
    cl::init(false));

// FIXME: Find a good default for this flag and remove the flag.
static cl::opt<unsigned>
CSRFirstTimeCost("regalloc-csr-first-time-cost",
              cl::desc("Cost for first time use of callee-saved register."),
              cl::init(0), cl::Hidden);

static cl::opt<bool> ConsiderLocalIntervalCost(
    "consider-local-interval-cost", cl::Hidden,
    cl::desc("Consider the cost of local intervals created by a split "
             "candidate when choosing the best split candidate."),
    cl::init(false));

static RegisterRegAlloc greedyRegAlloc("greedy", "greedy register allocator",
                                       createGreedyRegisterAllocator);

char RAGreedy::ID = 0;
char &llvm::RAGreedyID = RAGreedy::ID;

INITIALIZE_PASS_BEGIN(RAGreedy, "greedy",
                "Greedy Register Allocator", false, false)
INITIALIZE_PASS_DEPENDENCY(LiveDebugVariables)
INITIALIZE_PASS_DEPENDENCY(SlotIndexes)
INITIALIZE_PASS_DEPENDENCY(LiveIntervals)
INITIALIZE_PASS_DEPENDENCY(RegisterCoalescer)
INITIALIZE_PASS_DEPENDENCY(MachineScheduler)
INITIALIZE_PASS_DEPENDENCY(LiveStacks)
INITIALIZE_PASS_DEPENDENCY(MachineDominatorTree)
INITIALIZE_PASS_DEPENDENCY(MachineLoopInfo)
INITIALIZE_PASS_DEPENDENCY(VirtRegMap)
INITIALIZE_PASS_DEPENDENCY(LiveRegMatrix)
INITIALIZE_PASS_DEPENDENCY(EdgeBundles)
INITIALIZE_PASS_DEPENDENCY(SpillPlacement)
INITIALIZE_PASS_DEPENDENCY(MachineOptimizationRemarkEmitterPass)
INITIALIZE_PASS_DEPENDENCY(RegAllocEvictionAdvisorAnalysis)
INITIALIZE_PASS_END(RAGreedy, "greedy",
                "Greedy Register Allocator", false, false)

#ifndef NDEBUG
const char *const RAGreedy::StageName[] = {
    "RS_New",
    "RS_Assign",
    "RS_Split",
    "RS_Split2",
    "RS_Spill",
    "RS_Memory",
    "RS_Done"
};
#endif

// Hysteresis to use when comparing floats.
// This helps stabilize decisions based on float comparisons.
const float Hysteresis = (2007 / 2048.0f); // 0.97998046875

FunctionPass* llvm::createGreedyRegisterAllocator() {
  return new RAGreedy();
}

namespace llvm {
FunctionPass* createGreedyRegisterAllocator(
  std::function<bool(const TargetRegisterInfo &TRI,
                     const TargetRegisterClass &RC)> Ftor);

}

FunctionPass* llvm::createGreedyRegisterAllocator(
  std::function<bool(const TargetRegisterInfo &TRI,
                     const TargetRegisterClass &RC)> Ftor) {
  return new RAGreedy(Ftor);
}

RAGreedy::RAGreedy(RegClassFilterFunc F):
  MachineFunctionPass(ID),
  RegAllocBase(F) {
}

void RAGreedy::getAnalysisUsage(AnalysisUsage &AU) const {
  AU.setPreservesCFG();
  AU.addRequired<MachineBlockFrequencyInfo>();
  AU.addPreserved<MachineBlockFrequencyInfo>();
  AU.addRequired<AAResultsWrapperPass>();
  AU.addPreserved<AAResultsWrapperPass>();
  AU.addRequired<LiveIntervals>();
  AU.addPreserved<LiveIntervals>();
  AU.addRequired<SlotIndexes>();
  AU.addPreserved<SlotIndexes>();
  AU.addRequired<LiveDebugVariables>();
  AU.addPreserved<LiveDebugVariables>();
  AU.addRequired<LiveStacks>();
  AU.addPreserved<LiveStacks>();
  AU.addRequired<MachineDominatorTree>();
  AU.addPreserved<MachineDominatorTree>();
  AU.addRequired<MachineLoopInfo>();
  AU.addPreserved<MachineLoopInfo>();
  AU.addRequired<VirtRegMap>();
  AU.addPreserved<VirtRegMap>();
  AU.addRequired<LiveRegMatrix>();
  AU.addPreserved<LiveRegMatrix>();
  AU.addRequired<EdgeBundles>();
  AU.addRequired<SpillPlacement>();
  AU.addRequired<MachineOptimizationRemarkEmitterPass>();
  AU.addRequired<RegAllocEvictionAdvisorAnalysis>();
  MachineFunctionPass::getAnalysisUsage(AU);
}

//===----------------------------------------------------------------------===//
//                     LiveRangeEdit delegate methods
//===----------------------------------------------------------------------===//

bool RAGreedy::LRE_CanEraseVirtReg(Register VirtReg) {
  LiveInterval &LI = LIS->getInterval(VirtReg);
  if (VRM->hasPhys(VirtReg)) {
    Matrix->unassign(LI);
    aboutToRemoveInterval(LI);
    return true;
  }
  // Unassigned virtreg is probably in the priority queue.
  // RegAllocBase will erase it after dequeueing.
  // Nonetheless, clear the live-range so that the debug
  // dump will show the right state for that VirtReg.
  LI.clear();
  return false;
}

void RAGreedy::LRE_WillShrinkVirtReg(Register VirtReg) {
  if (!VRM->hasPhys(VirtReg))
    return;

  // Register is assigned, put it back on the queue for reassignment.
  LiveInterval &LI = LIS->getInterval(VirtReg);
  Matrix->unassign(LI);
  RegAllocBase::enqueue(&LI);
}

void RAGreedy::LRE_DidCloneVirtReg(Register New, Register Old) {
  ExtraInfo->LRE_DidCloneVirtReg(New, Old);
}

void RAGreedy::ExtraRegInfo::LRE_DidCloneVirtReg(Register New, Register Old) {
  // Cloning a register we haven't even heard about yet?  Just ignore it.
  if (!Info.inBounds(Old))
    return;

  // LRE may clone a virtual register because dead code elimination causes it to
  // be split into connected components. The new components are much smaller
  // than the original, so they should get a new chance at being assigned.
  // same stage as the parent.
  Info[Old].Stage = RS_Assign;
  Info.grow(New.id());
  Info[New] = Info[Old];
}

void RAGreedy::releaseMemory() {
  SpillerInstance.reset();
  GlobalCand.clear();
}

void RAGreedy::enqueueImpl(LiveInterval *LI) { enqueue(Queue, LI); }

void RAGreedy::enqueue(PQueue &CurQueue, LiveInterval *LI) {
  // Prioritize live ranges by size, assigning larger ranges first.
  // The queue holds (size, reg) pairs.
  const unsigned Size = LI->getSize();
  const Register Reg = LI->reg();
  assert(Reg.isVirtual() && "Can only enqueue virtual registers");
  unsigned Prio;

  auto Stage = ExtraInfo->getOrInitStage(Reg);
  if (Stage == RS_New) {
    Stage = RS_Assign;
    ExtraInfo->setStage(Reg, Stage);
  }
  if (Stage == RS_Split) {
    // Unsplit ranges that couldn't be allocated immediately are deferred until
    // everything else has been allocated.
    Prio = Size;
  } else if (Stage == RS_Memory) {
    // Memory operand should be considered last.
    // Change the priority such that Memory operand are assigned in
    // the reverse order that they came in.
    // TODO: Make this a member variable and probably do something about hints.
    static unsigned MemOp = 0;
    Prio = MemOp++;
  } else {
    // Giant live ranges fall back to the global assignment heuristic, which
    // prevents excessive spilling in pathological cases.
    bool ReverseLocal = TRI->reverseLocalAssignment();
    const TargetRegisterClass &RC = *MRI->getRegClass(Reg);
    bool ForceGlobal = !ReverseLocal &&
      (Size / SlotIndex::InstrDist) > (2 * RCI.getNumAllocatableRegs(&RC));

    if (Stage == RS_Assign && !ForceGlobal && !LI->empty() &&
        LIS->intervalIsInOneMBB(*LI)) {
      // Allocate original local ranges in linear instruction order. Since they
      // are singly defined, this produces optimal coloring in the absence of
      // global interference and other constraints.
      if (!ReverseLocal)
        Prio = LI->beginIndex().getInstrDistance(Indexes->getLastIndex());
      else {
        // Allocating bottom up may allow many short LRGs to be assigned first
        // to one of the cheap registers. This could be much faster for very
        // large blocks on targets with many physical registers.
        Prio = Indexes->getZeroIndex().getInstrDistance(LI->endIndex());
      }
      Prio |= RC.AllocationPriority << 24;
    } else {
      // Allocate global and split ranges in long->short order. Long ranges that
      // don't fit should be spilled (or split) ASAP so they don't create
      // interference.  Mark a bit to prioritize global above local ranges.
      Prio = (1u << 29) + Size;

      Prio |= RC.AllocationPriority << 24;
    }
    // Mark a higher bit to prioritize global and local above RS_Split.
    Prio |= (1u << 31);

    // Boost ranges that have a physical register hint.
    if (VRM->hasKnownPreference(Reg))
      Prio |= (1u << 30);
  }
  // The virtual register number is a tie breaker for same-sized ranges.
  // Give lower vreg numbers higher priority to assign them first.
  CurQueue.push(std::make_pair(Prio, ~Reg));
}

LiveInterval *RAGreedy::dequeue() { return dequeue(Queue); }

LiveInterval *RAGreedy::dequeue(PQueue &CurQueue) {
  if (CurQueue.empty())
    return nullptr;
  LiveInterval *LI = &LIS->getInterval(~CurQueue.top().second);
  CurQueue.pop();
  return LI;
}

//===----------------------------------------------------------------------===//
//                            Direct Assignment
//===----------------------------------------------------------------------===//

/// tryAssign - Try to assign VirtReg to an available register.
MCRegister RAGreedy::tryAssign(LiveInterval &VirtReg,
                             AllocationOrder &Order,
                             SmallVectorImpl<Register> &NewVRegs,
                             const SmallVirtRegSet &FixedRegisters) {
  MCRegister PhysReg;
  for (auto I = Order.begin(), E = Order.end(); I != E && !PhysReg; ++I) {
    assert(*I);
    if (!Matrix->checkInterference(VirtReg, *I)) {
      if (I.isHint())
        return *I;
      else
        PhysReg = *I;
    }
  }
  if (!PhysReg.isValid())
    return PhysReg;

  // PhysReg is available, but there may be a better choice.

  // If we missed a simple hint, try to cheaply evict interference from the
  // preferred register.
  if (Register Hint = MRI->getSimpleHint(VirtReg.reg()))
    if (Order.isHint(Hint)) {
      MCRegister PhysHint = Hint.asMCReg();
      LLVM_DEBUG(dbgs() << "missed hint " << printReg(PhysHint, TRI) << '\n');

      if (EvictAdvisor->canEvictHintInterference(VirtReg, PhysHint,
                                                 FixedRegisters)) {
        evictInterference(VirtReg, PhysHint, NewVRegs);
        return PhysHint;
      }
      // Record the missed hint, we may be able to recover
      // at the end if the surrounding allocation changed.
      SetOfBrokenHints.insert(&VirtReg);
    }

  // Try to evict interference from a cheaper alternative.
  uint8_t Cost = RegCosts[PhysReg];

  // Most registers have 0 additional cost.
  if (!Cost)
    return PhysReg;

  LLVM_DEBUG(dbgs() << printReg(PhysReg, TRI) << " is available at cost "
                    << (unsigned)Cost << '\n');
  MCRegister CheapReg = tryEvict(VirtReg, Order, NewVRegs, Cost, FixedRegisters);
  return CheapReg ? CheapReg : PhysReg;
}

//===----------------------------------------------------------------------===//
//                         Interference eviction
//===----------------------------------------------------------------------===//

Register RegAllocEvictionAdvisor::canReassign(LiveInterval &VirtReg,
                                              Register PrevReg) const {
  auto Order =
      AllocationOrder::create(VirtReg.reg(), *VRM, RegClassInfo, Matrix);
  MCRegister PhysReg;
  for (auto I = Order.begin(), E = Order.end(); I != E && !PhysReg; ++I) {
    if ((*I).id() == PrevReg.id())
      continue;

    MCRegUnitIterator Units(*I, TRI);
    for (; Units.isValid(); ++Units) {
      // Instantiate a "subquery", not to be confused with the Queries array.
      LiveIntervalUnion::Query subQ(VirtReg, Matrix->getLiveUnions()[*Units]);
      if (subQ.checkInterference())
        break;
    }
    // If no units have interference, break out with the current PhysReg.
    if (!Units.isValid())
      PhysReg = *I;
  }
  if (PhysReg)
    LLVM_DEBUG(dbgs() << "can reassign: " << VirtReg << " from "
                      << printReg(PrevReg, TRI) << " to "
                      << printReg(PhysReg, TRI) << '\n');
  return PhysReg;
}

/// Return true if all interferences between VirtReg and PhysReg between
/// Start and End can be evicted.
///
/// \param VirtReg Live range that is about to be assigned.
/// \param PhysReg Desired register for assignment.
/// \param Start   Start of range to look for interferences.
/// \param End     End of range to look for interferences.
/// \param MaxCost Only look for cheaper candidates and update with new cost
///                when returning true.
/// \return True when interference can be evicted cheaper than MaxCost.
bool RAGreedy::canEvictInterferenceInRange(const LiveInterval &VirtReg,
                                           MCRegister PhysReg, SlotIndex Start,
                                           SlotIndex End,
                                           EvictionCost &MaxCost) const {
  EvictionCost Cost;

  for (MCRegUnitIterator Units(PhysReg, TRI); Units.isValid(); ++Units) {
    LiveIntervalUnion::Query &Q = Matrix->query(VirtReg, *Units);

    // Check if any interfering live range is heavier than MaxWeight.
    for (const LiveInterval *Intf : reverse(Q.interferingVRegs())) {
      // Check if interference overlast the segment in interest.
      if (!Intf->overlaps(Start, End))
        continue;

      // Cannot evict non virtual reg interference.
      if (!Register::isVirtualRegister(Intf->reg()))
        return false;
      // Never evict spill products. They cannot split or spill.
      if (ExtraInfo->getStage(*Intf) == RS_Done)
        return false;

      // Would this break a satisfied hint?
      bool BreaksHint = VRM->hasPreferredPhys(Intf->reg());
      // Update eviction cost.
      Cost.BrokenHints += BreaksHint;
      Cost.MaxWeight = std::max(Cost.MaxWeight, Intf->weight());
      // Abort if this would be too expensive.
      if (!(Cost < MaxCost))
        return false;
    }
  }

  if (Cost.MaxWeight == 0)
    return false;

  MaxCost = Cost;
  return true;
}

/// Return the physical register that will be best
/// candidate for eviction by a local split interval that will be created
/// between Start and End.
///
/// \param Order            The allocation order
/// \param VirtReg          Live range that is about to be assigned.
/// \param Start            Start of range to look for interferences
/// \param End              End of range to look for interferences
/// \param BestEvictweight  The eviction cost of that eviction
/// \return The PhysReg which is the best candidate for eviction and the
/// eviction cost in BestEvictweight
MCRegister RAGreedy::getCheapestEvicteeWeight(const AllocationOrder &Order,
                                              const LiveInterval &VirtReg,
                                              SlotIndex Start, SlotIndex End,
                                              float *BestEvictweight) const {
  EvictionCost BestEvictCost;
  BestEvictCost.setMax();
  BestEvictCost.MaxWeight = VirtReg.weight();
  MCRegister BestEvicteePhys;

  // Go over all physical registers and find the best candidate for eviction
  for (MCRegister PhysReg : Order.getOrder()) {

    if (!canEvictInterferenceInRange(VirtReg, PhysReg, Start, End,
                                     BestEvictCost))
      continue;

    // Best so far.
    BestEvicteePhys = PhysReg;
  }
  *BestEvictweight = BestEvictCost.MaxWeight;
  return BestEvicteePhys;
}

/// evictInterference - Evict any interferring registers that prevent VirtReg
/// from being assigned to Physreg. This assumes that canEvictInterference
/// returned true.
void RAGreedy::evictInterference(LiveInterval &VirtReg, MCRegister PhysReg,
                                 SmallVectorImpl<Register> &NewVRegs) {
  // Make sure that VirtReg has a cascade number, and assign that cascade
  // number to every evicted register. These live ranges than then only be
  // evicted by a newer cascade, preventing infinite loops.
  unsigned Cascade = ExtraInfo->getOrAssignNewCascade(VirtReg.reg());

  LLVM_DEBUG(dbgs() << "evicting " << printReg(PhysReg, TRI)
                    << " interference: Cascade " << Cascade << '\n');

  // Collect all interfering virtregs first.
  SmallVector<LiveInterval*, 8> Intfs;
  for (MCRegUnitIterator Units(PhysReg, TRI); Units.isValid(); ++Units) {
    LiveIntervalUnion::Query &Q = Matrix->query(VirtReg, *Units);
    // We usually have the interfering VRegs cached so collectInterferingVRegs()
    // should be fast, we may need to recalculate if when different physregs
    // overlap the same register unit so we had different SubRanges queried
    // against it.
    ArrayRef<LiveInterval*> IVR = Q.interferingVRegs();
    Intfs.append(IVR.begin(), IVR.end());
  }

  // Evict them second. This will invalidate the queries.
  for (LiveInterval *Intf : Intfs) {
    // The same VirtReg may be present in multiple RegUnits. Skip duplicates.
    if (!VRM->hasPhys(Intf->reg()))
      continue;

    LastEvicted.addEviction(PhysReg, VirtReg.reg(), Intf->reg());

    Matrix->unassign(*Intf);
    assert((ExtraInfo->getCascade(Intf->reg()) < Cascade ||
            VirtReg.isSpillable() < Intf->isSpillable()) &&
           "Cannot decrease cascade number, illegal eviction");
    ExtraInfo->setCascade(Intf->reg(), Cascade);
    ++NumEvicted;
    NewVRegs.push_back(Intf->reg());
  }
}

/// Returns true if the given \p PhysReg is a callee saved register and has not
/// been used for allocation yet.
bool RegAllocEvictionAdvisor::isUnusedCalleeSavedReg(MCRegister PhysReg) const {
  MCRegister CSR = RegClassInfo.getLastCalleeSavedAlias(PhysReg);
  if (!CSR)
    return false;

  return !Matrix->isPhysRegUsed(PhysReg);
}

Optional<unsigned>
RegAllocEvictionAdvisor::getOrderLimit(const LiveInterval &VirtReg,
                                       const AllocationOrder &Order,
                                       unsigned CostPerUseLimit) const {
  unsigned OrderLimit = Order.getOrder().size();

  if (CostPerUseLimit < uint8_t(~0u)) {
    // Check of any registers in RC are below CostPerUseLimit.
    const TargetRegisterClass *RC = MRI->getRegClass(VirtReg.reg());
    uint8_t MinCost = RegClassInfo.getMinCost(RC);
    if (MinCost >= CostPerUseLimit) {
      LLVM_DEBUG(dbgs() << TRI->getRegClassName(RC) << " minimum cost = "
                        << MinCost << ", no cheaper registers to be found.\n");
      return None;
    }

    // It is normal for register classes to have a long tail of registers with
    // the same cost. We don't need to look at them if they're too expensive.
    if (RegCosts[Order.getOrder().back()] >= CostPerUseLimit) {
      OrderLimit = RegClassInfo.getLastCostChange(RC);
      LLVM_DEBUG(dbgs() << "Only trying the first " << OrderLimit
                        << " regs.\n");
    }
  }
  return OrderLimit;
}

bool RegAllocEvictionAdvisor::canAllocatePhysReg(unsigned CostPerUseLimit,
                                                 MCRegister PhysReg) const {
  if (RegCosts[PhysReg] >= CostPerUseLimit)
    return false;
  // The first use of a callee-saved register in a function has cost 1.
  // Don't start using a CSR when the CostPerUseLimit is low.
  if (CostPerUseLimit == 1 && isUnusedCalleeSavedReg(PhysReg)) {
    LLVM_DEBUG(
        dbgs() << printReg(PhysReg, TRI) << " would clobber CSR "
               << printReg(RegClassInfo.getLastCalleeSavedAlias(PhysReg), TRI)
               << '\n');
    return false;
  }
  return true;
}

/// tryEvict - Try to evict all interferences for a physreg.
/// @param  VirtReg Currently unassigned virtual register.
/// @param  Order   Physregs to try.
/// @return         Physreg to assign VirtReg, or 0.
MCRegister RAGreedy::tryEvict(LiveInterval &VirtReg, AllocationOrder &Order,
                              SmallVectorImpl<Register> &NewVRegs,
                              uint8_t CostPerUseLimit,
                              const SmallVirtRegSet &FixedRegisters) {
  NamedRegionTimer T("evict", "Evict", TimerGroupName, TimerGroupDescription,
                     TimePassesIsEnabled);

  MCRegister BestPhys = EvictAdvisor->tryFindEvictionCandidate(
      VirtReg, Order, CostPerUseLimit, FixedRegisters);
  if (BestPhys.isValid())
    evictInterference(VirtReg, BestPhys, NewVRegs);
  return BestPhys;
}

//===----------------------------------------------------------------------===//
//                              Region Splitting
//===----------------------------------------------------------------------===//

/// addSplitConstraints - Fill out the SplitConstraints vector based on the
/// interference pattern in Physreg and its aliases. Add the constraints to
/// SpillPlacement and return the static cost of this split in Cost, assuming
/// that all preferences in SplitConstraints are met.
/// Return false if there are no bundles with positive bias.
bool RAGreedy::addSplitConstraints(InterferenceCache::Cursor Intf,
                                   BlockFrequency &Cost) {
  ArrayRef<SplitAnalysis::BlockInfo> UseBlocks = SA->getUseBlocks();

  // Reset interference dependent info.
  SplitConstraints.resize(UseBlocks.size());
  BlockFrequency StaticCost = 0;
  for (unsigned I = 0; I != UseBlocks.size(); ++I) {
    const SplitAnalysis::BlockInfo &BI = UseBlocks[I];
    SpillPlacement::BlockConstraint &BC = SplitConstraints[I];

    BC.Number = BI.MBB->getNumber();
    Intf.moveToBlock(BC.Number);
    BC.Entry = BI.LiveIn ? SpillPlacement::PrefReg : SpillPlacement::DontCare;
    BC.Exit = (BI.LiveOut &&
               !LIS->getInstructionFromIndex(BI.LastInstr)->isImplicitDef())
                  ? SpillPlacement::PrefReg
                  : SpillPlacement::DontCare;
    BC.ChangesValue = BI.FirstDef.isValid();

    if (!Intf.hasInterference())
      continue;

    // Number of spill code instructions to insert.
    unsigned Ins = 0;

    // Interference for the live-in value.
    if (BI.LiveIn) {
      if (Intf.first() <= Indexes->getMBBStartIdx(BC.Number)) {
        BC.Entry = SpillPlacement::MustSpill;
        ++Ins;
      } else if (Intf.first() < BI.FirstInstr) {
        BC.Entry = SpillPlacement::PrefSpill;
        ++Ins;
      } else if (Intf.first() < BI.LastInstr) {
        ++Ins;
      }

      // Abort if the spill cannot be inserted at the MBB' start
      if (((BC.Entry == SpillPlacement::MustSpill) ||
           (BC.Entry == SpillPlacement::PrefSpill)) &&
          SlotIndex::isEarlierInstr(BI.FirstInstr,
                                    SA->getFirstSplitPoint(BC.Number)))
        return false;
    }

    // Interference for the live-out value.
    if (BI.LiveOut) {
      if (Intf.last() >= SA->getLastSplitPoint(BC.Number)) {
        BC.Exit = SpillPlacement::MustSpill;
        ++Ins;
      } else if (Intf.last() > BI.LastInstr) {
        BC.Exit = SpillPlacement::PrefSpill;
        ++Ins;
      } else if (Intf.last() > BI.FirstInstr) {
        ++Ins;
      }
    }

    // Accumulate the total frequency of inserted spill code.
    while (Ins--)
      StaticCost += SpillPlacer->getBlockFrequency(BC.Number);
  }
  Cost = StaticCost;

  // Add constraints for use-blocks. Note that these are the only constraints
  // that may add a positive bias, it is downhill from here.
  SpillPlacer->addConstraints(SplitConstraints);
  return SpillPlacer->scanActiveBundles();
}

/// addThroughConstraints - Add constraints and links to SpillPlacer from the
/// live-through blocks in Blocks.
bool RAGreedy::addThroughConstraints(InterferenceCache::Cursor Intf,
                                     ArrayRef<unsigned> Blocks) {
  const unsigned GroupSize = 8;
  SpillPlacement::BlockConstraint BCS[GroupSize];
  unsigned TBS[GroupSize];
  unsigned B = 0, T = 0;

  for (unsigned Number : Blocks) {
    Intf.moveToBlock(Number);

    if (!Intf.hasInterference()) {
      assert(T < GroupSize && "Array overflow");
      TBS[T] = Number;
      if (++T == GroupSize) {
        SpillPlacer->addLinks(makeArrayRef(TBS, T));
        T = 0;
      }
      continue;
    }

    assert(B < GroupSize && "Array overflow");
    BCS[B].Number = Number;

    // Abort if the spill cannot be inserted at the MBB' start
    MachineBasicBlock *MBB = MF->getBlockNumbered(Number);
    auto FirstNonDebugInstr = MBB->getFirstNonDebugInstr();
    if (FirstNonDebugInstr != MBB->end() &&
        SlotIndex::isEarlierInstr(LIS->getInstructionIndex(*FirstNonDebugInstr),
                                  SA->getFirstSplitPoint(Number)))
      return false;
    // Interference for the live-in value.
    if (Intf.first() <= Indexes->getMBBStartIdx(Number))
      BCS[B].Entry = SpillPlacement::MustSpill;
    else
      BCS[B].Entry = SpillPlacement::PrefSpill;

    // Interference for the live-out value.
    if (Intf.last() >= SA->getLastSplitPoint(Number))
      BCS[B].Exit = SpillPlacement::MustSpill;
    else
      BCS[B].Exit = SpillPlacement::PrefSpill;

    if (++B == GroupSize) {
      SpillPlacer->addConstraints(makeArrayRef(BCS, B));
      B = 0;
    }
  }

  SpillPlacer->addConstraints(makeArrayRef(BCS, B));
  SpillPlacer->addLinks(makeArrayRef(TBS, T));
  return true;
}

bool RAGreedy::growRegion(GlobalSplitCandidate &Cand) {
  // Keep track of through blocks that have not been added to SpillPlacer.
  BitVector Todo = SA->getThroughBlocks();
  SmallVectorImpl<unsigned> &ActiveBlocks = Cand.ActiveBlocks;
  unsigned AddedTo = 0;
#ifndef NDEBUG
  unsigned Visited = 0;
#endif

  while (true) {
    ArrayRef<unsigned> NewBundles = SpillPlacer->getRecentPositive();
    // Find new through blocks in the periphery of PrefRegBundles.
    for (unsigned Bundle : NewBundles) {
      // Look at all blocks connected to Bundle in the full graph.
      ArrayRef<unsigned> Blocks = Bundles->getBlocks(Bundle);
      for (unsigned Block : Blocks) {
        if (!Todo.test(Block))
          continue;
        Todo.reset(Block);
        // This is a new through block. Add it to SpillPlacer later.
        ActiveBlocks.push_back(Block);
#ifndef NDEBUG
        ++Visited;
#endif
      }
    }
    // Any new blocks to add?
    if (ActiveBlocks.size() == AddedTo)
      break;

    // Compute through constraints from the interference, or assume that all
    // through blocks prefer spilling when forming compact regions.
    auto NewBlocks = makeArrayRef(ActiveBlocks).slice(AddedTo);
    if (Cand.PhysReg) {
      if (!addThroughConstraints(Cand.Intf, NewBlocks))
        return false;
    } else
      // Provide a strong negative bias on through blocks to prevent unwanted
      // liveness on loop backedges.
      SpillPlacer->addPrefSpill(NewBlocks, /* Strong= */ true);
    AddedTo = ActiveBlocks.size();

    // Perhaps iterating can enable more bundles?
    SpillPlacer->iterate();
  }
  LLVM_DEBUG(dbgs() << ", v=" << Visited);
  return true;
}

/// calcCompactRegion - Compute the set of edge bundles that should be live
/// when splitting the current live range into compact regions.  Compact
/// regions can be computed without looking at interference.  They are the
/// regions formed by removing all the live-through blocks from the live range.
///
/// Returns false if the current live range is already compact, or if the
/// compact regions would form single block regions anyway.
bool RAGreedy::calcCompactRegion(GlobalSplitCandidate &Cand) {
  // Without any through blocks, the live range is already compact.
  if (!SA->getNumThroughBlocks())
    return false;

  // Compact regions don't correspond to any physreg.
  Cand.reset(IntfCache, MCRegister::NoRegister);

  LLVM_DEBUG(dbgs() << "Compact region bundles");

  // Use the spill placer to determine the live bundles. GrowRegion pretends
  // that all the through blocks have interference when PhysReg is unset.
  SpillPlacer->prepare(Cand.LiveBundles);

  // The static split cost will be zero since Cand.Intf reports no interference.
  BlockFrequency Cost;
  if (!addSplitConstraints(Cand.Intf, Cost)) {
    LLVM_DEBUG(dbgs() << ", none.\n");
    return false;
  }

  if (!growRegion(Cand)) {
    LLVM_DEBUG(dbgs() << ", cannot spill all interferences.\n");
    return false;
  }

  SpillPlacer->finish();

  if (!Cand.LiveBundles.any()) {
    LLVM_DEBUG(dbgs() << ", none.\n");
    return false;
  }

  LLVM_DEBUG({
    for (int I : Cand.LiveBundles.set_bits())
      dbgs() << " EB#" << I;
    dbgs() << ".\n";
  });
  return true;
}

/// calcSpillCost - Compute how expensive it would be to split the live range in
/// SA around all use blocks instead of forming bundle regions.
BlockFrequency RAGreedy::calcSpillCost() {
  BlockFrequency Cost = 0;
  ArrayRef<SplitAnalysis::BlockInfo> UseBlocks = SA->getUseBlocks();
  for (const SplitAnalysis::BlockInfo &BI : UseBlocks) {
    unsigned Number = BI.MBB->getNumber();
    // We normally only need one spill instruction - a load or a store.
    Cost += SpillPlacer->getBlockFrequency(Number);

    // Unless the value is redefined in the block.
    if (BI.LiveIn && BI.LiveOut && BI.FirstDef)
      Cost += SpillPlacer->getBlockFrequency(Number);
  }
  return Cost;
}

/// Check if splitting Evictee will create a local split interval in
/// basic block number BBNumber that may cause a bad eviction chain. This is
/// intended to prevent bad eviction sequences like:
/// movl	%ebp, 8(%esp)           # 4-byte Spill
/// movl	%ecx, %ebp
/// movl	%ebx, %ecx
/// movl	%edi, %ebx
/// movl	%edx, %edi
/// cltd
/// idivl	%esi
/// movl	%edi, %edx
/// movl	%ebx, %edi
/// movl	%ecx, %ebx
/// movl	%ebp, %ecx
/// movl	16(%esp), %ebp          # 4 - byte Reload
///
/// Such sequences are created in 2 scenarios:
///
/// Scenario #1:
/// %0 is evicted from physreg0 by %1.
/// Evictee %0 is intended for region splitting with split candidate
/// physreg0 (the reg %0 was evicted from).
/// Region splitting creates a local interval because of interference with the
/// evictor %1 (normally region splitting creates 2 interval, the "by reg"
/// and "by stack" intervals and local interval created when interference
/// occurs).
/// One of the split intervals ends up evicting %2 from physreg1.
/// Evictee %2 is intended for region splitting with split candidate
/// physreg1.
/// One of the split intervals ends up evicting %3 from physreg2, etc.
///
/// Scenario #2
/// %0 is evicted from physreg0 by %1.
/// %2 is evicted from physreg2 by %3 etc.
/// Evictee %0 is intended for region splitting with split candidate
/// physreg1.
/// Region splitting creates a local interval because of interference with the
/// evictor %1.
/// One of the split intervals ends up evicting back original evictor %1
/// from physreg0 (the reg %0 was evicted from).
/// Another evictee %2 is intended for region splitting with split candidate
/// physreg1.
/// One of the split intervals ends up evicting %3 from physreg2, etc.
///
/// \param Evictee  The register considered to be split.
/// \param Cand     The split candidate that determines the physical register
///                 we are splitting for and the interferences.
/// \param BBNumber The number of a BB for which the region split process will
///                 create a local split interval.
/// \param Order    The physical registers that may get evicted by a split
///                 artifact of Evictee.
/// \return True if splitting Evictee may cause a bad eviction chain, false
/// otherwise.
bool RAGreedy::splitCanCauseEvictionChain(Register Evictee,
                                          GlobalSplitCandidate &Cand,
                                          unsigned BBNumber,
                                          const AllocationOrder &Order) {
  EvictionTrack::EvictorInfo VregEvictorInfo = LastEvicted.getEvictor(Evictee);
  unsigned Evictor = VregEvictorInfo.first;
  MCRegister PhysReg = VregEvictorInfo.second;

  // No actual evictor.
  if (!Evictor || !PhysReg)
    return false;

  float MaxWeight = 0;
  MCRegister FutureEvictedPhysReg =
      getCheapestEvicteeWeight(Order, LIS->getInterval(Evictee),
                               Cand.Intf.first(), Cand.Intf.last(), &MaxWeight);

  // The bad eviction chain occurs when either the split candidate is the
  // evicting reg or one of the split artifact will evict the evicting reg.
  if ((PhysReg != Cand.PhysReg) && (PhysReg != FutureEvictedPhysReg))
    return false;

  Cand.Intf.moveToBlock(BBNumber);

  // Check to see if the Evictor contains interference (with Evictee) in the
  // given BB. If so, this interference caused the eviction of Evictee from
  // PhysReg. This suggest that we will create a local interval during the
  // region split to avoid this interference This local interval may cause a bad
  // eviction chain.
  if (!LIS->hasInterval(Evictor))
    return false;
  LiveInterval &EvictorLI = LIS->getInterval(Evictor);
  if (EvictorLI.FindSegmentContaining(Cand.Intf.first()) == EvictorLI.end())
    return false;

  // Now, check to see if the local interval we will create is going to be
  // expensive enough to evict somebody If so, this may cause a bad eviction
  // chain.
  float splitArtifactWeight =
      VRAI->futureWeight(LIS->getInterval(Evictee),
                         Cand.Intf.first().getPrevIndex(), Cand.Intf.last());
  if (splitArtifactWeight >= 0 && splitArtifactWeight < MaxWeight)
    return false;

  return true;
}

/// Check if splitting VirtRegToSplit will create a local split interval
/// in basic block number BBNumber that may cause a spill.
///
/// \param VirtRegToSplit The register considered to be split.
/// \param Cand           The split candidate that determines the physical
///                       register we are splitting for and the interferences.
/// \param BBNumber       The number of a BB for which the region split process
///                       will create a local split interval.
/// \param Order          The physical registers that may get evicted by a
///                       split artifact of VirtRegToSplit.
/// \return True if splitting VirtRegToSplit may cause a spill, false
/// otherwise.
bool RAGreedy::splitCanCauseLocalSpill(unsigned VirtRegToSplit,
                                       GlobalSplitCandidate &Cand,
                                       unsigned BBNumber,
                                       const AllocationOrder &Order) {
  Cand.Intf.moveToBlock(BBNumber);

  // Check if the local interval will find a non interfereing assignment.
  for (auto PhysReg : Order.getOrder()) {
    if (!Matrix->checkInterference(Cand.Intf.first().getPrevIndex(),
                                   Cand.Intf.last(), PhysReg))
      return false;
  }

  // The local interval is not able to find non interferencing assignment
  // and not able to evict a less worthy interval, therfore, it can cause a
  // spill.
  return true;
}

/// calcGlobalSplitCost - Return the global split cost of following the split
/// pattern in LiveBundles. This cost should be added to the local cost of the
/// interference pattern in SplitConstraints.
///
BlockFrequency RAGreedy::calcGlobalSplitCost(GlobalSplitCandidate &Cand,
                                             const AllocationOrder &Order,
                                             bool *CanCauseEvictionChain) {
  BlockFrequency GlobalCost = 0;
  const BitVector &LiveBundles = Cand.LiveBundles;
  Register VirtRegToSplit = SA->getParent().reg();
  ArrayRef<SplitAnalysis::BlockInfo> UseBlocks = SA->getUseBlocks();
  for (unsigned I = 0; I != UseBlocks.size(); ++I) {
    const SplitAnalysis::BlockInfo &BI = UseBlocks[I];
    SpillPlacement::BlockConstraint &BC = SplitConstraints[I];
    bool RegIn  = LiveBundles[Bundles->getBundle(BC.Number, false)];
    bool RegOut = LiveBundles[Bundles->getBundle(BC.Number, true)];
    unsigned Ins = 0;

    Cand.Intf.moveToBlock(BC.Number);
    // Check wheather a local interval is going to be created during the region
    // split. Calculate adavanced spilt cost (cost of local intervals) if option
    // is enabled.
    if (EnableAdvancedRASplitCost && Cand.Intf.hasInterference() && BI.LiveIn &&
        BI.LiveOut && RegIn && RegOut) {

      if (CanCauseEvictionChain &&
          splitCanCauseEvictionChain(VirtRegToSplit, Cand, BC.Number, Order)) {
        // This interference causes our eviction from this assignment, we might
        // evict somebody else and eventually someone will spill, add that cost.
        // See splitCanCauseEvictionChain for detailed description of scenarios.
        GlobalCost += SpillPlacer->getBlockFrequency(BC.Number);
        GlobalCost += SpillPlacer->getBlockFrequency(BC.Number);

        *CanCauseEvictionChain = true;

      } else if (splitCanCauseLocalSpill(VirtRegToSplit, Cand, BC.Number,
                                         Order)) {
        // This interference causes local interval to spill, add that cost.
        GlobalCost += SpillPlacer->getBlockFrequency(BC.Number);
        GlobalCost += SpillPlacer->getBlockFrequency(BC.Number);
      }
    }

    if (BI.LiveIn)
      Ins += RegIn != (BC.Entry == SpillPlacement::PrefReg);
    if (BI.LiveOut)
      Ins += RegOut != (BC.Exit == SpillPlacement::PrefReg);
    while (Ins--)
      GlobalCost += SpillPlacer->getBlockFrequency(BC.Number);
  }

  for (unsigned Number : Cand.ActiveBlocks) {
    bool RegIn  = LiveBundles[Bundles->getBundle(Number, false)];
    bool RegOut = LiveBundles[Bundles->getBundle(Number, true)];
    if (!RegIn && !RegOut)
      continue;
    if (RegIn && RegOut) {
      // We need double spill code if this block has interference.
      Cand.Intf.moveToBlock(Number);
      if (Cand.Intf.hasInterference()) {
        GlobalCost += SpillPlacer->getBlockFrequency(Number);
        GlobalCost += SpillPlacer->getBlockFrequency(Number);

        // Check wheather a local interval is going to be created during the
        // region split.
        if (EnableAdvancedRASplitCost && CanCauseEvictionChain &&
            splitCanCauseEvictionChain(VirtRegToSplit, Cand, Number, Order)) {
          // This interference cause our eviction from this assignment, we might
          // evict somebody else, add that cost.
          // See splitCanCauseEvictionChain for detailed description of
          // scenarios.
          GlobalCost += SpillPlacer->getBlockFrequency(Number);
          GlobalCost += SpillPlacer->getBlockFrequency(Number);

          *CanCauseEvictionChain = true;
        }
      }
      continue;
    }
    // live-in / stack-out or stack-in live-out.
    GlobalCost += SpillPlacer->getBlockFrequency(Number);
  }
  return GlobalCost;
}

/// splitAroundRegion - Split the current live range around the regions
/// determined by BundleCand and GlobalCand.
///
/// Before calling this function, GlobalCand and BundleCand must be initialized
/// so each bundle is assigned to a valid candidate, or NoCand for the
/// stack-bound bundles.  The shared SA/SE SplitAnalysis and SplitEditor
/// objects must be initialized for the current live range, and intervals
/// created for the used candidates.
///
/// @param LREdit    The LiveRangeEdit object handling the current split.
/// @param UsedCands List of used GlobalCand entries. Every BundleCand value
///                  must appear in this list.
void RAGreedy::splitAroundRegion(LiveRangeEdit &LREdit,
                                 ArrayRef<unsigned> UsedCands) {
  // These are the intervals created for new global ranges. We may create more
  // intervals for local ranges.
  const unsigned NumGlobalIntvs = LREdit.size();
  LLVM_DEBUG(dbgs() << "splitAroundRegion with " << NumGlobalIntvs
                    << " globals.\n");
  assert(NumGlobalIntvs && "No global intervals configured");

  // Isolate even single instructions when dealing with a proper sub-class.
  // That guarantees register class inflation for the stack interval because it
  // is all copies.
  Register Reg = SA->getParent().reg();
  bool SingleInstrs = RegClassInfo.isProperSubClass(MRI->getRegClass(Reg));

  // First handle all the blocks with uses.
  ArrayRef<SplitAnalysis::BlockInfo> UseBlocks = SA->getUseBlocks();
  for (const SplitAnalysis::BlockInfo &BI : UseBlocks) {
    unsigned Number = BI.MBB->getNumber();
    unsigned IntvIn = 0, IntvOut = 0;
    SlotIndex IntfIn, IntfOut;
    if (BI.LiveIn) {
      unsigned CandIn = BundleCand[Bundles->getBundle(Number, false)];
      if (CandIn != NoCand) {
        GlobalSplitCandidate &Cand = GlobalCand[CandIn];
        IntvIn = Cand.IntvIdx;
        Cand.Intf.moveToBlock(Number);
        IntfIn = Cand.Intf.first();
      }
    }
    if (BI.LiveOut) {
      unsigned CandOut = BundleCand[Bundles->getBundle(Number, true)];
      if (CandOut != NoCand) {
        GlobalSplitCandidate &Cand = GlobalCand[CandOut];
        IntvOut = Cand.IntvIdx;
        Cand.Intf.moveToBlock(Number);
        IntfOut = Cand.Intf.last();
      }
    }

    // Create separate intervals for isolated blocks with multiple uses.
    if (!IntvIn && !IntvOut) {
      LLVM_DEBUG(dbgs() << printMBBReference(*BI.MBB) << " isolated.\n");
      if (SA->shouldSplitSingleBlock(BI, SingleInstrs))
        SE->splitSingleBlock(BI);
      continue;
    }

    if (IntvIn && IntvOut)
      SE->splitLiveThroughBlock(Number, IntvIn, IntfIn, IntvOut, IntfOut);
    else if (IntvIn)
      SE->splitRegInBlock(BI, IntvIn, IntfIn);
    else
      SE->splitRegOutBlock(BI, IntvOut, IntfOut);
  }

  // Handle live-through blocks. The relevant live-through blocks are stored in
  // the ActiveBlocks list with each candidate. We need to filter out
  // duplicates.
  BitVector Todo = SA->getThroughBlocks();
  for (unsigned UsedCand : UsedCands) {
    ArrayRef<unsigned> Blocks = GlobalCand[UsedCand].ActiveBlocks;
    for (unsigned Number : Blocks) {
      if (!Todo.test(Number))
        continue;
      Todo.reset(Number);

      unsigned IntvIn = 0, IntvOut = 0;
      SlotIndex IntfIn, IntfOut;

      unsigned CandIn = BundleCand[Bundles->getBundle(Number, false)];
      if (CandIn != NoCand) {
        GlobalSplitCandidate &Cand = GlobalCand[CandIn];
        IntvIn = Cand.IntvIdx;
        Cand.Intf.moveToBlock(Number);
        IntfIn = Cand.Intf.first();
      }

      unsigned CandOut = BundleCand[Bundles->getBundle(Number, true)];
      if (CandOut != NoCand) {
        GlobalSplitCandidate &Cand = GlobalCand[CandOut];
        IntvOut = Cand.IntvIdx;
        Cand.Intf.moveToBlock(Number);
        IntfOut = Cand.Intf.last();
      }
      if (!IntvIn && !IntvOut)
        continue;
      SE->splitLiveThroughBlock(Number, IntvIn, IntfIn, IntvOut, IntfOut);
    }
  }

  ++NumGlobalSplits;

  SmallVector<unsigned, 8> IntvMap;
  SE->finish(&IntvMap);
  DebugVars->splitRegister(Reg, LREdit.regs(), *LIS);

  unsigned OrigBlocks = SA->getNumLiveBlocks();

  // Sort out the new intervals created by splitting. We get four kinds:
  // - Remainder intervals should not be split again.
  // - Candidate intervals can be assigned to Cand.PhysReg.
  // - Block-local splits are candidates for local splitting.
  // - DCE leftovers should go back on the queue.
  for (unsigned I = 0, E = LREdit.size(); I != E; ++I) {
    const LiveInterval &Reg = LIS->getInterval(LREdit.get(I));

    // Ignore old intervals from DCE.
    if (ExtraInfo->getOrInitStage(Reg.reg()) != RS_New)
      continue;

    // Remainder interval. Don't try splitting again, spill if it doesn't
    // allocate.
    if (IntvMap[I] == 0) {
      ExtraInfo->setStage(Reg, RS_Spill);
      continue;
    }

    // Global intervals. Allow repeated splitting as long as the number of live
    // blocks is strictly decreasing.
    if (IntvMap[I] < NumGlobalIntvs) {
      if (SA->countLiveBlocks(&Reg) >= OrigBlocks) {
        LLVM_DEBUG(dbgs() << "Main interval covers the same " << OrigBlocks
                          << " blocks as original.\n");
        // Don't allow repeated splitting as a safe guard against looping.
        ExtraInfo->setStage(Reg, RS_Split2);
      }
      continue;
    }

    // Other intervals are treated as new. This includes local intervals created
    // for blocks with multiple uses, and anything created by DCE.
  }

  if (VerifyEnabled)
    MF->verify(this, "After splitting live range around region");
}

MCRegister RAGreedy::tryRegionSplit(LiveInterval &VirtReg,
                                    AllocationOrder &Order,
                                    SmallVectorImpl<Register> &NewVRegs) {
  if (!TRI->shouldRegionSplitForVirtReg(*MF, VirtReg))
    return MCRegister::NoRegister;
  unsigned NumCands = 0;
  BlockFrequency SpillCost = calcSpillCost();
  BlockFrequency BestCost;

  // Check if we can split this live range around a compact region.
  bool HasCompact = calcCompactRegion(GlobalCand.front());
  if (HasCompact) {
    // Yes, keep GlobalCand[0] as the compact region candidate.
    NumCands = 1;
    BestCost = BlockFrequency::getMaxFrequency();
  } else {
    // No benefit from the compact region, our fallback will be per-block
    // splitting. Make sure we find a solution that is cheaper than spilling.
    BestCost = SpillCost;
    LLVM_DEBUG(dbgs() << "Cost of isolating all blocks = ";
               MBFI->printBlockFreq(dbgs(), BestCost) << '\n');
  }

  bool CanCauseEvictionChain = false;
  unsigned BestCand =
      calculateRegionSplitCost(VirtReg, Order, BestCost, NumCands,
                               false /*IgnoreCSR*/, &CanCauseEvictionChain);

  // Split candidates with compact regions can cause a bad eviction sequence.
  // See splitCanCauseEvictionChain for detailed description of scenarios.
  // To avoid it, we need to comapre the cost with the spill cost and not the
  // current max frequency.
  if (HasCompact && (BestCost > SpillCost) && (BestCand != NoCand) &&
    CanCauseEvictionChain) {
    return MCRegister::NoRegister;
  }

  // No solutions found, fall back to single block splitting.
  if (!HasCompact && BestCand == NoCand)
    return MCRegister::NoRegister;

  return doRegionSplit(VirtReg, BestCand, HasCompact, NewVRegs);
}

unsigned RAGreedy::calculateRegionSplitCost(LiveInterval &VirtReg,
                                            AllocationOrder &Order,
                                            BlockFrequency &BestCost,
                                            unsigned &NumCands, bool IgnoreCSR,
                                            bool *CanCauseEvictionChain) {
  unsigned BestCand = NoCand;
  for (MCPhysReg PhysReg : Order) {
    assert(PhysReg);
    if (IgnoreCSR && EvictAdvisor->isUnusedCalleeSavedReg(PhysReg))
      continue;

    // Discard bad candidates before we run out of interference cache cursors.
    // This will only affect register classes with a lot of registers (>32).
    if (NumCands == IntfCache.getMaxCursors()) {
      unsigned WorstCount = ~0u;
      unsigned Worst = 0;
      for (unsigned CandIndex = 0; CandIndex != NumCands; ++CandIndex) {
        if (CandIndex == BestCand || !GlobalCand[CandIndex].PhysReg)
          continue;
        unsigned Count = GlobalCand[CandIndex].LiveBundles.count();
        if (Count < WorstCount) {
          Worst = CandIndex;
          WorstCount = Count;
        }
      }
      --NumCands;
      GlobalCand[Worst] = GlobalCand[NumCands];
      if (BestCand == NumCands)
        BestCand = Worst;
    }

    if (GlobalCand.size() <= NumCands)
      GlobalCand.resize(NumCands+1);
    GlobalSplitCandidate &Cand = GlobalCand[NumCands];
    Cand.reset(IntfCache, PhysReg);

    SpillPlacer->prepare(Cand.LiveBundles);
    BlockFrequency Cost;
    if (!addSplitConstraints(Cand.Intf, Cost)) {
      LLVM_DEBUG(dbgs() << printReg(PhysReg, TRI) << "\tno positive bundles\n");
      continue;
    }
    LLVM_DEBUG(dbgs() << printReg(PhysReg, TRI) << "\tstatic = ";
               MBFI->printBlockFreq(dbgs(), Cost));
    if (Cost >= BestCost) {
      LLVM_DEBUG({
        if (BestCand == NoCand)
          dbgs() << " worse than no bundles\n";
        else
          dbgs() << " worse than "
                 << printReg(GlobalCand[BestCand].PhysReg, TRI) << '\n';
      });
      continue;
    }
    if (!growRegion(Cand)) {
      LLVM_DEBUG(dbgs() << ", cannot spill all interferences.\n");
      continue;
    }

    SpillPlacer->finish();

    // No live bundles, defer to splitSingleBlocks().
    if (!Cand.LiveBundles.any()) {
      LLVM_DEBUG(dbgs() << " no bundles.\n");
      continue;
    }

    bool HasEvictionChain = false;
    Cost += calcGlobalSplitCost(Cand, Order, &HasEvictionChain);
    LLVM_DEBUG({
      dbgs() << ", total = ";
      MBFI->printBlockFreq(dbgs(), Cost) << " with bundles";
      for (int I : Cand.LiveBundles.set_bits())
        dbgs() << " EB#" << I;
      dbgs() << ".\n";
    });
    if (Cost < BestCost) {
      BestCand = NumCands;
      BestCost = Cost;
      // See splitCanCauseEvictionChain for detailed description of bad
      // eviction chain scenarios.
      if (CanCauseEvictionChain)
        *CanCauseEvictionChain = HasEvictionChain;
    }
    ++NumCands;
  }

  if (CanCauseEvictionChain && BestCand != NoCand) {
    // See splitCanCauseEvictionChain for detailed description of bad
    // eviction chain scenarios.
    LLVM_DEBUG(dbgs() << "Best split candidate of vreg "
                      << printReg(VirtReg.reg(), TRI) << "  may ");
    if (!(*CanCauseEvictionChain))
      LLVM_DEBUG(dbgs() << "not ");
    LLVM_DEBUG(dbgs() << "cause bad eviction chain\n");
  }

  return BestCand;
}

unsigned RAGreedy::doRegionSplit(LiveInterval &VirtReg, unsigned BestCand,
                                 bool HasCompact,
                                 SmallVectorImpl<Register> &NewVRegs) {
  SmallVector<unsigned, 8> UsedCands;
  // Prepare split editor.
  LiveRangeEdit LREdit(&VirtReg, NewVRegs, *MF, *LIS, VRM, this, &DeadRemats);
  SE->reset(LREdit, SplitSpillMode);

  // Assign all edge bundles to the preferred candidate, or NoCand.
  BundleCand.assign(Bundles->getNumBundles(), NoCand);

  // Assign bundles for the best candidate region.
  if (BestCand != NoCand) {
    GlobalSplitCandidate &Cand = GlobalCand[BestCand];
    if (unsigned B = Cand.getBundles(BundleCand, BestCand)) {
      UsedCands.push_back(BestCand);
      Cand.IntvIdx = SE->openIntv();
      LLVM_DEBUG(dbgs() << "Split for " << printReg(Cand.PhysReg, TRI) << " in "
                        << B << " bundles, intv " << Cand.IntvIdx << ".\n");
      (void)B;
    }
  }

  // Assign bundles for the compact region.
  if (HasCompact) {
    GlobalSplitCandidate &Cand = GlobalCand.front();
    assert(!Cand.PhysReg && "Compact region has no physreg");
    if (unsigned B = Cand.getBundles(BundleCand, 0)) {
      UsedCands.push_back(0);
      Cand.IntvIdx = SE->openIntv();
      LLVM_DEBUG(dbgs() << "Split for compact region in " << B
                        << " bundles, intv " << Cand.IntvIdx << ".\n");
      (void)B;
    }
  }

  splitAroundRegion(LREdit, UsedCands);
  return 0;
}

//===----------------------------------------------------------------------===//
//                            Per-Block Splitting
//===----------------------------------------------------------------------===//

/// tryBlockSplit - Split a global live range around every block with uses. This
/// creates a lot of local live ranges, that will be split by tryLocalSplit if
/// they don't allocate.
unsigned RAGreedy::tryBlockSplit(LiveInterval &VirtReg, AllocationOrder &Order,
                                 SmallVectorImpl<Register> &NewVRegs) {
  assert(&SA->getParent() == &VirtReg && "Live range wasn't analyzed");
  Register Reg = VirtReg.reg();
  bool SingleInstrs = RegClassInfo.isProperSubClass(MRI->getRegClass(Reg));
  LiveRangeEdit LREdit(&VirtReg, NewVRegs, *MF, *LIS, VRM, this, &DeadRemats);
  SE->reset(LREdit, SplitSpillMode);
  ArrayRef<SplitAnalysis::BlockInfo> UseBlocks = SA->getUseBlocks();
  for (const SplitAnalysis::BlockInfo &BI : UseBlocks) {
    if (SA->shouldSplitSingleBlock(BI, SingleInstrs))
      SE->splitSingleBlock(BI);
  }
  // No blocks were split.
  if (LREdit.empty())
    return 0;

  // We did split for some blocks.
  SmallVector<unsigned, 8> IntvMap;
  SE->finish(&IntvMap);

  // Tell LiveDebugVariables about the new ranges.
  DebugVars->splitRegister(Reg, LREdit.regs(), *LIS);

  // Sort out the new intervals created by splitting. The remainder interval
  // goes straight to spilling, the new local ranges get to stay RS_New.
  for (unsigned I = 0, E = LREdit.size(); I != E; ++I) {
    const LiveInterval &LI = LIS->getInterval(LREdit.get(I));
    if (ExtraInfo->getOrInitStage(LI.reg()) == RS_New && IntvMap[I] == 0)
      ExtraInfo->setStage(LI, RS_Spill);
  }

  if (VerifyEnabled)
    MF->verify(this, "After splitting live range around basic blocks");
  return 0;
}

//===----------------------------------------------------------------------===//
//                         Per-Instruction Splitting
//===----------------------------------------------------------------------===//

/// Get the number of allocatable registers that match the constraints of \p Reg
/// on \p MI and that are also in \p SuperRC.
static unsigned getNumAllocatableRegsForConstraints(
    const MachineInstr *MI, Register Reg, const TargetRegisterClass *SuperRC,
    const TargetInstrInfo *TII, const TargetRegisterInfo *TRI,
    const RegisterClassInfo &RCI) {
  assert(SuperRC && "Invalid register class");

  const TargetRegisterClass *ConstrainedRC =
      MI->getRegClassConstraintEffectForVReg(Reg, SuperRC, TII, TRI,
                                             /* ExploreBundle */ true);
  if (!ConstrainedRC)
    return 0;
  return RCI.getNumAllocatableRegs(ConstrainedRC);
}

/// tryInstructionSplit - Split a live range around individual instructions.
/// This is normally not worthwhile since the spiller is doing essentially the
/// same thing. However, when the live range is in a constrained register
/// class, it may help to insert copies such that parts of the live range can
/// be moved to a larger register class.
///
/// This is similar to spilling to a larger register class.
unsigned
RAGreedy::tryInstructionSplit(LiveInterval &VirtReg, AllocationOrder &Order,
                              SmallVectorImpl<Register> &NewVRegs) {
  const TargetRegisterClass *CurRC = MRI->getRegClass(VirtReg.reg());
  // There is no point to this if there are no larger sub-classes.
  if (!RegClassInfo.isProperSubClass(CurRC))
    return 0;

  // Always enable split spill mode, since we're effectively spilling to a
  // register.
  LiveRangeEdit LREdit(&VirtReg, NewVRegs, *MF, *LIS, VRM, this, &DeadRemats);
  SE->reset(LREdit, SplitEditor::SM_Size);

  ArrayRef<SlotIndex> Uses = SA->getUseSlots();
  if (Uses.size() <= 1)
    return 0;

  LLVM_DEBUG(dbgs() << "Split around " << Uses.size()
                    << " individual instrs.\n");

  const TargetRegisterClass *SuperRC =
      TRI->getLargestLegalSuperClass(CurRC, *MF);
  unsigned SuperRCNumAllocatableRegs = RCI.getNumAllocatableRegs(SuperRC);
  // Split around every non-copy instruction if this split will relax
  // the constraints on the virtual register.
  // Otherwise, splitting just inserts uncoalescable copies that do not help
  // the allocation.
  for (const SlotIndex Use : Uses) {
    if (const MachineInstr *MI = Indexes->getInstructionFromIndex(Use))
      if (MI->isFullCopy() ||
          SuperRCNumAllocatableRegs ==
              getNumAllocatableRegsForConstraints(MI, VirtReg.reg(), SuperRC,
                                                  TII, TRI, RCI)) {
        LLVM_DEBUG(dbgs() << "    skip:\t" << Use << '\t' << *MI);
        continue;
      }
    SE->openIntv();
    SlotIndex SegStart = SE->enterIntvBefore(Use);
    SlotIndex SegStop = SE->leaveIntvAfter(Use);
    SE->useIntv(SegStart, SegStop);
  }

  if (LREdit.empty()) {
    LLVM_DEBUG(dbgs() << "All uses were copies.\n");
    return 0;
  }

  SmallVector<unsigned, 8> IntvMap;
  SE->finish(&IntvMap);
  DebugVars->splitRegister(VirtReg.reg(), LREdit.regs(), *LIS);
  // Assign all new registers to RS_Spill. This was the last chance.
  ExtraInfo->setStage(LREdit.begin(), LREdit.end(), RS_Spill);
  return 0;
}

//===----------------------------------------------------------------------===//
//                             Local Splitting
//===----------------------------------------------------------------------===//

/// calcGapWeights - Compute the maximum spill weight that needs to be evicted
/// in order to use PhysReg between two entries in SA->UseSlots.
///
/// GapWeight[I] represents the gap between UseSlots[I] and UseSlots[I + 1].
///
void RAGreedy::calcGapWeights(MCRegister PhysReg,
                              SmallVectorImpl<float> &GapWeight) {
  assert(SA->getUseBlocks().size() == 1 && "Not a local interval");
  const SplitAnalysis::BlockInfo &BI = SA->getUseBlocks().front();
  ArrayRef<SlotIndex> Uses = SA->getUseSlots();
  const unsigned NumGaps = Uses.size()-1;

  // Start and end points for the interference check.
  SlotIndex StartIdx =
    BI.LiveIn ? BI.FirstInstr.getBaseIndex() : BI.FirstInstr;
  SlotIndex StopIdx =
    BI.LiveOut ? BI.LastInstr.getBoundaryIndex() : BI.LastInstr;

  GapWeight.assign(NumGaps, 0.0f);

  // Add interference from each overlapping register.
  for (MCRegUnitIterator Units(PhysReg, TRI); Units.isValid(); ++Units) {
    if (!Matrix->query(const_cast<LiveInterval&>(SA->getParent()), *Units)
          .checkInterference())
      continue;

    // We know that VirtReg is a continuous interval from FirstInstr to
    // LastInstr, so we don't need InterferenceQuery.
    //
    // Interference that overlaps an instruction is counted in both gaps
    // surrounding the instruction. The exception is interference before
    // StartIdx and after StopIdx.
    //
    LiveIntervalUnion::SegmentIter IntI =
      Matrix->getLiveUnions()[*Units] .find(StartIdx);
    for (unsigned Gap = 0; IntI.valid() && IntI.start() < StopIdx; ++IntI) {
      // Skip the gaps before IntI.
      while (Uses[Gap+1].getBoundaryIndex() < IntI.start())
        if (++Gap == NumGaps)
          break;
      if (Gap == NumGaps)
        break;

      // Update the gaps covered by IntI.
      const float weight = IntI.value()->weight();
      for (; Gap != NumGaps; ++Gap) {
        GapWeight[Gap] = std::max(GapWeight[Gap], weight);
        if (Uses[Gap+1].getBaseIndex() >= IntI.stop())
          break;
      }
      if (Gap == NumGaps)
        break;
    }
  }

  // Add fixed interference.
  for (MCRegUnitIterator Units(PhysReg, TRI); Units.isValid(); ++Units) {
    const LiveRange &LR = LIS->getRegUnit(*Units);
    LiveRange::const_iterator I = LR.find(StartIdx);
    LiveRange::const_iterator E = LR.end();

    // Same loop as above. Mark any overlapped gaps as HUGE_VALF.
    for (unsigned Gap = 0; I != E && I->start < StopIdx; ++I) {
      while (Uses[Gap+1].getBoundaryIndex() < I->start)
        if (++Gap == NumGaps)
          break;
      if (Gap == NumGaps)
        break;

      for (; Gap != NumGaps; ++Gap) {
        GapWeight[Gap] = huge_valf;
        if (Uses[Gap+1].getBaseIndex() >= I->end)
          break;
      }
      if (Gap == NumGaps)
        break;
    }
  }
}

/// tryLocalSplit - Try to split VirtReg into smaller intervals inside its only
/// basic block.
///
unsigned RAGreedy::tryLocalSplit(LiveInterval &VirtReg, AllocationOrder &Order,
                                 SmallVectorImpl<Register> &NewVRegs) {
  // TODO: the function currently only handles a single UseBlock; it should be
  // possible to generalize.
  if (SA->getUseBlocks().size() != 1)
    return 0;

  const SplitAnalysis::BlockInfo &BI = SA->getUseBlocks().front();

  // Note that it is possible to have an interval that is live-in or live-out
  // while only covering a single block - A phi-def can use undef values from
  // predecessors, and the block could be a single-block loop.
  // We don't bother doing anything clever about such a case, we simply assume
  // that the interval is continuous from FirstInstr to LastInstr. We should
  // make sure that we don't do anything illegal to such an interval, though.

  ArrayRef<SlotIndex> Uses = SA->getUseSlots();
  if (Uses.size() <= 2)
    return 0;
  const unsigned NumGaps = Uses.size()-1;

  LLVM_DEBUG({
    dbgs() << "tryLocalSplit: ";
    for (const auto &Use : Uses)
      dbgs() << ' ' << Use;
    dbgs() << '\n';
  });

  // If VirtReg is live across any register mask operands, compute a list of
  // gaps with register masks.
  SmallVector<unsigned, 8> RegMaskGaps;
  if (Matrix->checkRegMaskInterference(VirtReg)) {
    // Get regmask slots for the whole block.
    ArrayRef<SlotIndex> RMS = LIS->getRegMaskSlotsInBlock(BI.MBB->getNumber());
    LLVM_DEBUG(dbgs() << RMS.size() << " regmasks in block:");
    // Constrain to VirtReg's live range.
    unsigned RI =
        llvm::lower_bound(RMS, Uses.front().getRegSlot()) - RMS.begin();
    unsigned RE = RMS.size();
    for (unsigned I = 0; I != NumGaps && RI != RE; ++I) {
      // Look for Uses[I] <= RMS <= Uses[I + 1].
      assert(!SlotIndex::isEarlierInstr(RMS[RI], Uses[I]));
      if (SlotIndex::isEarlierInstr(Uses[I + 1], RMS[RI]))
        continue;
      // Skip a regmask on the same instruction as the last use. It doesn't
      // overlap the live range.
      if (SlotIndex::isSameInstr(Uses[I + 1], RMS[RI]) && I + 1 == NumGaps)
        break;
      LLVM_DEBUG(dbgs() << ' ' << RMS[RI] << ':' << Uses[I] << '-'
                        << Uses[I + 1]);
      RegMaskGaps.push_back(I);
      // Advance ri to the next gap. A regmask on one of the uses counts in
      // both gaps.
      while (RI != RE && SlotIndex::isEarlierInstr(RMS[RI], Uses[I + 1]))
        ++RI;
    }
    LLVM_DEBUG(dbgs() << '\n');
  }

  // Since we allow local split results to be split again, there is a risk of
  // creating infinite loops. It is tempting to require that the new live
  // ranges have less instructions than the original. That would guarantee
  // convergence, but it is too strict. A live range with 3 instructions can be
  // split 2+3 (including the COPY), and we want to allow that.
  //
  // Instead we use these rules:
  //
  // 1. Allow any split for ranges with getStage() < RS_Split2. (Except for the
  //    noop split, of course).
  // 2. Require progress be made for ranges with getStage() == RS_Split2. All
  //    the new ranges must have fewer instructions than before the split.
  // 3. New ranges with the same number of instructions are marked RS_Split2,
  //    smaller ranges are marked RS_New.
  //
  // These rules allow a 3 -> 2+3 split once, which we need. They also prevent
  // excessive splitting and infinite loops.
  //
  bool ProgressRequired = ExtraInfo->getStage(VirtReg) >= RS_Split2;

  // Best split candidate.
  unsigned BestBefore = NumGaps;
  unsigned BestAfter = 0;
  float BestDiff = 0;

  const float blockFreq =
    SpillPlacer->getBlockFrequency(BI.MBB->getNumber()).getFrequency() *
    (1.0f / MBFI->getEntryFreq());
  SmallVector<float, 8> GapWeight;

  for (MCPhysReg PhysReg : Order) {
    assert(PhysReg);
    // Keep track of the largest spill weight that would need to be evicted in
    // order to make use of PhysReg between UseSlots[I] and UseSlots[I + 1].
    calcGapWeights(PhysReg, GapWeight);

    // Remove any gaps with regmask clobbers.
    if (Matrix->checkRegMaskInterference(VirtReg, PhysReg))
      for (unsigned I = 0, E = RegMaskGaps.size(); I != E; ++I)
        GapWeight[RegMaskGaps[I]] = huge_valf;

    // Try to find the best sequence of gaps to close.
    // The new spill weight must be larger than any gap interference.

    // We will split before Uses[SplitBefore] and after Uses[SplitAfter].
    unsigned SplitBefore = 0, SplitAfter = 1;

    // MaxGap should always be max(GapWeight[SplitBefore..SplitAfter-1]).
    // It is the spill weight that needs to be evicted.
    float MaxGap = GapWeight[0];

    while (true) {
      // Live before/after split?
      const bool LiveBefore = SplitBefore != 0 || BI.LiveIn;
      const bool LiveAfter = SplitAfter != NumGaps || BI.LiveOut;

      LLVM_DEBUG(dbgs() << printReg(PhysReg, TRI) << ' ' << Uses[SplitBefore]
                        << '-' << Uses[SplitAfter] << " I=" << MaxGap);

      // Stop before the interval gets so big we wouldn't be making progress.
      if (!LiveBefore && !LiveAfter) {
        LLVM_DEBUG(dbgs() << " all\n");
        break;
      }
      // Should the interval be extended or shrunk?
      bool Shrink = true;

      // How many gaps would the new range have?
      unsigned NewGaps = LiveBefore + SplitAfter - SplitBefore + LiveAfter;

      // Legally, without causing looping?
      bool Legal = !ProgressRequired || NewGaps < NumGaps;

      if (Legal && MaxGap < huge_valf) {
        // Estimate the new spill weight. Each instruction reads or writes the
        // register. Conservatively assume there are no read-modify-write
        // instructions.
        //
        // Try to guess the size of the new interval.
        const float EstWeight = normalizeSpillWeight(
            blockFreq * (NewGaps + 1),
            Uses[SplitBefore].distance(Uses[SplitAfter]) +
                (LiveBefore + LiveAfter) * SlotIndex::InstrDist,
            1);
        // Would this split be possible to allocate?
        // Never allocate all gaps, we wouldn't be making progress.
        LLVM_DEBUG(dbgs() << " w=" << EstWeight);
        if (EstWeight * Hysteresis >= MaxGap) {
          Shrink = false;
          float Diff = EstWeight - MaxGap;
          if (Diff > BestDiff) {
            LLVM_DEBUG(dbgs() << " (best)");
            BestDiff = Hysteresis * Diff;
            BestBefore = SplitBefore;
            BestAfter = SplitAfter;
          }
        }
      }

      // Try to shrink.
      if (Shrink) {
        if (++SplitBefore < SplitAfter) {
          LLVM_DEBUG(dbgs() << " shrink\n");
          // Recompute the max when necessary.
          if (GapWeight[SplitBefore - 1] >= MaxGap) {
            MaxGap = GapWeight[SplitBefore];
            for (unsigned I = SplitBefore + 1; I != SplitAfter; ++I)
              MaxGap = std::max(MaxGap, GapWeight[I]);
          }
          continue;
        }
        MaxGap = 0;
      }

      // Try to extend the interval.
      if (SplitAfter >= NumGaps) {
        LLVM_DEBUG(dbgs() << " end\n");
        break;
      }

      LLVM_DEBUG(dbgs() << " extend\n");
      MaxGap = std::max(MaxGap, GapWeight[SplitAfter++]);
    }
  }

  // Didn't find any candidates?
  if (BestBefore == NumGaps)
    return 0;

  LLVM_DEBUG(dbgs() << "Best local split range: " << Uses[BestBefore] << '-'
                    << Uses[BestAfter] << ", " << BestDiff << ", "
                    << (BestAfter - BestBefore + 1) << " instrs\n");

  LiveRangeEdit LREdit(&VirtReg, NewVRegs, *MF, *LIS, VRM, this, &DeadRemats);
  SE->reset(LREdit);

  SE->openIntv();
  SlotIndex SegStart = SE->enterIntvBefore(Uses[BestBefore]);
  SlotIndex SegStop  = SE->leaveIntvAfter(Uses[BestAfter]);
  SE->useIntv(SegStart, SegStop);
  SmallVector<unsigned, 8> IntvMap;
  SE->finish(&IntvMap);
  DebugVars->splitRegister(VirtReg.reg(), LREdit.regs(), *LIS);
  // If the new range has the same number of instructions as before, mark it as
  // RS_Split2 so the next split will be forced to make progress. Otherwise,
  // leave the new intervals as RS_New so they can compete.
  bool LiveBefore = BestBefore != 0 || BI.LiveIn;
  bool LiveAfter = BestAfter != NumGaps || BI.LiveOut;
  unsigned NewGaps = LiveBefore + BestAfter - BestBefore + LiveAfter;
  if (NewGaps >= NumGaps) {
    LLVM_DEBUG(dbgs() << "Tagging non-progress ranges:");
    assert(!ProgressRequired && "Didn't make progress when it was required.");
    for (unsigned I = 0, E = IntvMap.size(); I != E; ++I)
      if (IntvMap[I] == 1) {
        ExtraInfo->setStage(LIS->getInterval(LREdit.get(I)), RS_Split2);
        LLVM_DEBUG(dbgs() << ' ' << printReg(LREdit.get(I)));
      }
    LLVM_DEBUG(dbgs() << '\n');
  }
  ++NumLocalSplits;

  return 0;
}

//===----------------------------------------------------------------------===//
//                          Live Range Splitting
//===----------------------------------------------------------------------===//

/// trySplit - Try to split VirtReg or one of its interferences, making it
/// assignable.
/// @return Physreg when VirtReg may be assigned and/or new NewVRegs.
unsigned RAGreedy::trySplit(LiveInterval &VirtReg, AllocationOrder &Order,
                            SmallVectorImpl<Register> &NewVRegs,
                            const SmallVirtRegSet &FixedRegisters) {
  // Ranges must be Split2 or less.
  if (ExtraInfo->getStage(VirtReg) >= RS_Spill)
    return 0;

  // Local intervals are handled separately.
  if (LIS->intervalIsInOneMBB(VirtReg)) {
    NamedRegionTimer T("local_split", "Local Splitting", TimerGroupName,
                       TimerGroupDescription, TimePassesIsEnabled);
    SA->analyze(&VirtReg);
    Register PhysReg = tryLocalSplit(VirtReg, Order, NewVRegs);
    if (PhysReg || !NewVRegs.empty())
      return PhysReg;
    return tryInstructionSplit(VirtReg, Order, NewVRegs);
  }

  NamedRegionTimer T("global_split", "Global Splitting", TimerGroupName,
                     TimerGroupDescription, TimePassesIsEnabled);

  SA->analyze(&VirtReg);

  // First try to split around a region spanning multiple blocks. RS_Split2
  // ranges already made dubious progress with region splitting, so they go
  // straight to single block splitting.
  if (ExtraInfo->getStage(VirtReg) < RS_Split2) {
    MCRegister PhysReg = tryRegionSplit(VirtReg, Order, NewVRegs);
    if (PhysReg || !NewVRegs.empty())
      return PhysReg;
  }

  // Then isolate blocks.
  return tryBlockSplit(VirtReg, Order, NewVRegs);
}

//===----------------------------------------------------------------------===//
//                          Last Chance Recoloring
//===----------------------------------------------------------------------===//

/// Return true if \p reg has any tied def operand.
static bool hasTiedDef(MachineRegisterInfo *MRI, unsigned reg) {
  for (const MachineOperand &MO : MRI->def_operands(reg))
    if (MO.isTied())
      return true;

  return false;
}

/// mayRecolorAllInterferences - Check if the virtual registers that
/// interfere with \p VirtReg on \p PhysReg (or one of its aliases) may be
/// recolored to free \p PhysReg.
/// When true is returned, \p RecoloringCandidates has been augmented with all
/// the live intervals that need to be recolored in order to free \p PhysReg
/// for \p VirtReg.
/// \p FixedRegisters contains all the virtual registers that cannot be
/// recolored.
bool RAGreedy::mayRecolorAllInterferences(
    MCRegister PhysReg, LiveInterval &VirtReg, SmallLISet &RecoloringCandidates,
    const SmallVirtRegSet &FixedRegisters) {
  const TargetRegisterClass *CurRC = MRI->getRegClass(VirtReg.reg());

  for (MCRegUnitIterator Units(PhysReg, TRI); Units.isValid(); ++Units) {
    LiveIntervalUnion::Query &Q = Matrix->query(VirtReg, *Units);
    // If there is LastChanceRecoloringMaxInterference or more interferences,
    // chances are one would not be recolorable.
    if (Q.interferingVRegs(LastChanceRecoloringMaxInterference).size() >=
            LastChanceRecoloringMaxInterference &&
        !ExhaustiveSearch) {
      LLVM_DEBUG(dbgs() << "Early abort: too many interferences.\n");
      CutOffInfo |= CO_Interf;
      return false;
    }
    for (LiveInterval *Intf : reverse(Q.interferingVRegs())) {
      // If Intf is done and sit on the same register class as VirtReg,
      // it would not be recolorable as it is in the same state as VirtReg.
      // However, if VirtReg has tied defs and Intf doesn't, then
      // there is still a point in examining if it can be recolorable.
      if (((ExtraInfo->getStage(*Intf) == RS_Done &&
            MRI->getRegClass(Intf->reg()) == CurRC) &&
           !(hasTiedDef(MRI, VirtReg.reg()) &&
             !hasTiedDef(MRI, Intf->reg()))) ||
          FixedRegisters.count(Intf->reg())) {
        LLVM_DEBUG(
            dbgs() << "Early abort: the interference is not recolorable.\n");
        return false;
      }
      RecoloringCandidates.insert(Intf);
    }
  }
  return true;
}

/// tryLastChanceRecoloring - Try to assign a color to \p VirtReg by recoloring
/// its interferences.
/// Last chance recoloring chooses a color for \p VirtReg and recolors every
/// virtual register that was using it. The recoloring process may recursively
/// use the last chance recoloring. Therefore, when a virtual register has been
/// assigned a color by this mechanism, it is marked as Fixed, i.e., it cannot
/// be last-chance-recolored again during this recoloring "session".
/// E.g.,
/// Let
/// vA can use {R1, R2    }
/// vB can use {    R2, R3}
/// vC can use {R1        }
/// Where vA, vB, and vC cannot be split anymore (they are reloads for
/// instance) and they all interfere.
///
/// vA is assigned R1
/// vB is assigned R2
/// vC tries to evict vA but vA is already done.
/// Regular register allocation fails.
///
/// Last chance recoloring kicks in:
/// vC does as if vA was evicted => vC uses R1.
/// vC is marked as fixed.
/// vA needs to find a color.
/// None are available.
/// vA cannot evict vC: vC is a fixed virtual register now.
/// vA does as if vB was evicted => vA uses R2.
/// vB needs to find a color.
/// R3 is available.
/// Recoloring => vC = R1, vA = R2, vB = R3
///
/// \p Order defines the preferred allocation order for \p VirtReg.
/// \p NewRegs will contain any new virtual register that have been created
/// (split, spill) during the process and that must be assigned.
/// \p FixedRegisters contains all the virtual registers that cannot be
/// recolored.
/// \p Depth gives the current depth of the last chance recoloring.
/// \return a physical register that can be used for VirtReg or ~0u if none
/// exists.
unsigned RAGreedy::tryLastChanceRecoloring(LiveInterval &VirtReg,
                                           AllocationOrder &Order,
                                           SmallVectorImpl<Register> &NewVRegs,
                                           SmallVirtRegSet &FixedRegisters,
                                           unsigned Depth) {
  if (!TRI->shouldUseLastChanceRecoloringForVirtReg(*MF, VirtReg))
    return ~0u;

  LLVM_DEBUG(dbgs() << "Try last chance recoloring for " << VirtReg << '\n');
  // Ranges must be Done.
  assert((ExtraInfo->getStage(VirtReg) >= RS_Done || !VirtReg.isSpillable()) &&
         "Last chance recoloring should really be last chance");
  // Set the max depth to LastChanceRecoloringMaxDepth.
  // We may want to reconsider that if we end up with a too large search space
  // for target with hundreds of registers.
  // Indeed, in that case we may want to cut the search space earlier.
  if (Depth >= LastChanceRecoloringMaxDepth && !ExhaustiveSearch) {
    LLVM_DEBUG(dbgs() << "Abort because max depth has been reached.\n");
    CutOffInfo |= CO_Depth;
    return ~0u;
  }

  // Set of Live intervals that will need to be recolored.
  SmallLISet RecoloringCandidates;
  // Record the original mapping virtual register to physical register in case
  // the recoloring fails.
  DenseMap<Register, MCRegister> VirtRegToPhysReg;
  // Mark VirtReg as fixed, i.e., it will not be recolored pass this point in
  // this recoloring "session".
  assert(!FixedRegisters.count(VirtReg.reg()));
  FixedRegisters.insert(VirtReg.reg());
  SmallVector<Register, 4> CurrentNewVRegs;

  for (MCRegister PhysReg : Order) {
    assert(PhysReg.isValid());
    LLVM_DEBUG(dbgs() << "Try to assign: " << VirtReg << " to "
                      << printReg(PhysReg, TRI) << '\n');
    RecoloringCandidates.clear();
    VirtRegToPhysReg.clear();
    CurrentNewVRegs.clear();

    // It is only possible to recolor virtual register interference.
    if (Matrix->checkInterference(VirtReg, PhysReg) >
        LiveRegMatrix::IK_VirtReg) {
      LLVM_DEBUG(
          dbgs() << "Some interferences are not with virtual registers.\n");

      continue;
    }

    // Early give up on this PhysReg if it is obvious we cannot recolor all
    // the interferences.
    if (!mayRecolorAllInterferences(PhysReg, VirtReg, RecoloringCandidates,
                                    FixedRegisters)) {
      LLVM_DEBUG(dbgs() << "Some interferences cannot be recolored.\n");
      continue;
    }

    // RecoloringCandidates contains all the virtual registers that interfer
    // with VirtReg on PhysReg (or one of its aliases).
    // Enqueue them for recoloring and perform the actual recoloring.
    PQueue RecoloringQueue;
    for (LiveInterval *RC : RecoloringCandidates) {
      Register ItVirtReg = RC->reg();
      enqueue(RecoloringQueue, RC);
      assert(VRM->hasPhys(ItVirtReg) &&
             "Interferences are supposed to be with allocated variables");

      // Record the current allocation.
      VirtRegToPhysReg[ItVirtReg] = VRM->getPhys(ItVirtReg);
      // unset the related struct.
      Matrix->unassign(*RC);
    }

    // Do as if VirtReg was assigned to PhysReg so that the underlying
    // recoloring has the right information about the interferes and
    // available colors.
    Matrix->assign(VirtReg, PhysReg);

    // Save the current recoloring state.
    // If we cannot recolor all the interferences, we will have to start again
    // at this point for the next physical register.
    SmallVirtRegSet SaveFixedRegisters(FixedRegisters);
    if (tryRecoloringCandidates(RecoloringQueue, CurrentNewVRegs,
                                FixedRegisters, Depth)) {
      // Push the queued vregs into the main queue.
      for (Register NewVReg : CurrentNewVRegs)
        NewVRegs.push_back(NewVReg);
      // Do not mess up with the global assignment process.
      // I.e., VirtReg must be unassigned.
      Matrix->unassign(VirtReg);
      return PhysReg;
    }

    LLVM_DEBUG(dbgs() << "Fail to assign: " << VirtReg << " to "
                      << printReg(PhysReg, TRI) << '\n');

    // The recoloring attempt failed, undo the changes.
    FixedRegisters = SaveFixedRegisters;
    Matrix->unassign(VirtReg);

    // For a newly created vreg which is also in RecoloringCandidates,
    // don't add it to NewVRegs because its physical register will be restored
    // below. Other vregs in CurrentNewVRegs are created by calling
    // selectOrSplit and should be added into NewVRegs.
    for (Register &R : CurrentNewVRegs) {
      if (RecoloringCandidates.count(&LIS->getInterval(R)))
        continue;
      NewVRegs.push_back(R);
    }

    for (LiveInterval *RC : RecoloringCandidates) {
      Register ItVirtReg = RC->reg();
      if (VRM->hasPhys(ItVirtReg))
        Matrix->unassign(*RC);
      MCRegister ItPhysReg = VirtRegToPhysReg[ItVirtReg];
      Matrix->assign(*RC, ItPhysReg);
    }
  }

  // Last chance recoloring did not worked either, give up.
  return ~0u;
}

/// tryRecoloringCandidates - Try to assign a new color to every register
/// in \RecoloringQueue.
/// \p NewRegs will contain any new virtual register created during the
/// recoloring process.
/// \p FixedRegisters[in/out] contains all the registers that have been
/// recolored.
/// \return true if all virtual registers in RecoloringQueue were successfully
/// recolored, false otherwise.
bool RAGreedy::tryRecoloringCandidates(PQueue &RecoloringQueue,
                                       SmallVectorImpl<Register> &NewVRegs,
                                       SmallVirtRegSet &FixedRegisters,
                                       unsigned Depth) {
  while (!RecoloringQueue.empty()) {
    LiveInterval *LI = dequeue(RecoloringQueue);
    LLVM_DEBUG(dbgs() << "Try to recolor: " << *LI << '\n');
    MCRegister PhysReg =
        selectOrSplitImpl(*LI, NewVRegs, FixedRegisters, Depth + 1);
    // When splitting happens, the live-range may actually be empty.
    // In that case, this is okay to continue the recoloring even
    // if we did not find an alternative color for it. Indeed,
    // there will not be anything to color for LI in the end.
    if (PhysReg == ~0u || (!PhysReg && !LI->empty()))
      return false;

    if (!PhysReg) {
      assert(LI->empty() && "Only empty live-range do not require a register");
      LLVM_DEBUG(dbgs() << "Recoloring of " << *LI
                        << " succeeded. Empty LI.\n");
      continue;
    }
    LLVM_DEBUG(dbgs() << "Recoloring of " << *LI
                      << " succeeded with: " << printReg(PhysReg, TRI) << '\n');

    Matrix->assign(*LI, PhysReg);
    FixedRegisters.insert(LI->reg());
  }
  return true;
}

//===----------------------------------------------------------------------===//
//                            Main Entry Point
//===----------------------------------------------------------------------===//

MCRegister RAGreedy::selectOrSplit(LiveInterval &VirtReg,
                                   SmallVectorImpl<Register> &NewVRegs) {
  CutOffInfo = CO_None;
  LLVMContext &Ctx = MF->getFunction().getContext();
  SmallVirtRegSet FixedRegisters;
  MCRegister Reg = selectOrSplitImpl(VirtReg, NewVRegs, FixedRegisters);
  if (Reg == ~0U && (CutOffInfo != CO_None)) {
    uint8_t CutOffEncountered = CutOffInfo & (CO_Depth | CO_Interf);
    if (CutOffEncountered == CO_Depth)
      Ctx.emitError("register allocation failed: maximum depth for recoloring "
                    "reached. Use -fexhaustive-register-search to skip "
                    "cutoffs");
    else if (CutOffEncountered == CO_Interf)
      Ctx.emitError("register allocation failed: maximum interference for "
                    "recoloring reached. Use -fexhaustive-register-search "
                    "to skip cutoffs");
    else if (CutOffEncountered == (CO_Depth | CO_Interf))
      Ctx.emitError("register allocation failed: maximum interference and "
                    "depth for recoloring reached. Use "
                    "-fexhaustive-register-search to skip cutoffs");
  }
  return Reg;
}

/// Using a CSR for the first time has a cost because it causes push|pop
/// to be added to prologue|epilogue. Splitting a cold section of the live
/// range can have lower cost than using the CSR for the first time;
/// Spilling a live range in the cold path can have lower cost than using
/// the CSR for the first time. Returns the physical register if we decide
/// to use the CSR; otherwise return 0.
MCRegister
RAGreedy::tryAssignCSRFirstTime(LiveInterval &VirtReg, AllocationOrder &Order,
                                MCRegister PhysReg, uint8_t &CostPerUseLimit,
                                SmallVectorImpl<Register> &NewVRegs) {
  if (ExtraInfo->getStage(VirtReg) == RS_Spill && VirtReg.isSpillable()) {
    // We choose spill over using the CSR for the first time if the spill cost
    // is lower than CSRCost.
    SA->analyze(&VirtReg);
    if (calcSpillCost() >= CSRCost)
      return PhysReg;

    // We are going to spill, set CostPerUseLimit to 1 to make sure that
    // we will not use a callee-saved register in tryEvict.
    CostPerUseLimit = 1;
    return 0;
  }
  if (ExtraInfo->getStage(VirtReg) < RS_Split) {
    // We choose pre-splitting over using the CSR for the first time if
    // the cost of splitting is lower than CSRCost.
    SA->analyze(&VirtReg);
    unsigned NumCands = 0;
    BlockFrequency BestCost = CSRCost; // Don't modify CSRCost.
    unsigned BestCand = calculateRegionSplitCost(VirtReg, Order, BestCost,
                                                 NumCands, true /*IgnoreCSR*/);
    if (BestCand == NoCand)
      // Use the CSR if we can't find a region split below CSRCost.
      return PhysReg;

    // Perform the actual pre-splitting.
    doRegionSplit(VirtReg, BestCand, false/*HasCompact*/, NewVRegs);
    return 0;
  }
  return PhysReg;
}

void RAGreedy::aboutToRemoveInterval(LiveInterval &LI) {
  // Do not keep invalid information around.
  SetOfBrokenHints.remove(&LI);
}

void RAGreedy::initializeCSRCost() {
  // We use the larger one out of the command-line option and the value report
  // by TRI.
  CSRCost = BlockFrequency(
      std::max((unsigned)CSRFirstTimeCost, TRI->getCSRFirstUseCost()));
  if (!CSRCost.getFrequency())
    return;

  // Raw cost is relative to Entry == 2^14; scale it appropriately.
  uint64_t ActualEntry = MBFI->getEntryFreq();
  if (!ActualEntry) {
    CSRCost = 0;
    return;
  }
  uint64_t FixedEntry = 1 << 14;
  if (ActualEntry < FixedEntry)
    CSRCost *= BranchProbability(ActualEntry, FixedEntry);
  else if (ActualEntry <= UINT32_MAX)
    // Invert the fraction and divide.
    CSRCost /= BranchProbability(FixedEntry, ActualEntry);
  else
    // Can't use BranchProbability in general, since it takes 32-bit numbers.
    CSRCost = CSRCost.getFrequency() * (ActualEntry / FixedEntry);
}

/// Collect the hint info for \p Reg.
/// The results are stored into \p Out.
/// \p Out is not cleared before being populated.
void RAGreedy::collectHintInfo(Register Reg, HintsInfo &Out) {
  for (const MachineInstr &Instr : MRI->reg_nodbg_instructions(Reg)) {
    if (!Instr.isFullCopy())
      continue;
    // Look for the other end of the copy.
    Register OtherReg = Instr.getOperand(0).getReg();
    if (OtherReg == Reg) {
      OtherReg = Instr.getOperand(1).getReg();
      if (OtherReg == Reg)
        continue;
    }
    // Get the current assignment.
    MCRegister OtherPhysReg =
        OtherReg.isPhysical() ? OtherReg.asMCReg() : VRM->getPhys(OtherReg);
    // Push the collected information.
    Out.push_back(HintInfo(MBFI->getBlockFreq(Instr.getParent()), OtherReg,
                           OtherPhysReg));
  }
}

/// Using the given \p List, compute the cost of the broken hints if
/// \p PhysReg was used.
/// \return The cost of \p List for \p PhysReg.
BlockFrequency RAGreedy::getBrokenHintFreq(const HintsInfo &List,
                                           MCRegister PhysReg) {
  BlockFrequency Cost = 0;
  for (const HintInfo &Info : List) {
    if (Info.PhysReg != PhysReg)
      Cost += Info.Freq;
  }
  return Cost;
}

/// Using the register assigned to \p VirtReg, try to recolor
/// all the live ranges that are copy-related with \p VirtReg.
/// The recoloring is then propagated to all the live-ranges that have
/// been recolored and so on, until no more copies can be coalesced or
/// it is not profitable.
/// For a given live range, profitability is determined by the sum of the
/// frequencies of the non-identity copies it would introduce with the old
/// and new register.
void RAGreedy::tryHintRecoloring(LiveInterval &VirtReg) {
  // We have a broken hint, check if it is possible to fix it by
  // reusing PhysReg for the copy-related live-ranges. Indeed, we evicted
  // some register and PhysReg may be available for the other live-ranges.
  SmallSet<Register, 4> Visited;
  SmallVector<unsigned, 2> RecoloringCandidates;
  HintsInfo Info;
  Register Reg = VirtReg.reg();
  MCRegister PhysReg = VRM->getPhys(Reg);
  // Start the recoloring algorithm from the input live-interval, then
  // it will propagate to the ones that are copy-related with it.
  Visited.insert(Reg);
  RecoloringCandidates.push_back(Reg);

  LLVM_DEBUG(dbgs() << "Trying to reconcile hints for: " << printReg(Reg, TRI)
                    << '(' << printReg(PhysReg, TRI) << ")\n");

  do {
    Reg = RecoloringCandidates.pop_back_val();

    // We cannot recolor physical register.
    if (Register::isPhysicalRegister(Reg))
      continue;

    // This may be a skipped class
    if (!VRM->hasPhys(Reg)) {
      assert(!ShouldAllocateClass(*TRI, *MRI->getRegClass(Reg)) &&
             "We have an unallocated variable which should have been handled");
      continue;
    }

    // Get the live interval mapped with this virtual register to be able
    // to check for the interference with the new color.
    LiveInterval &LI = LIS->getInterval(Reg);
    MCRegister CurrPhys = VRM->getPhys(Reg);
    // Check that the new color matches the register class constraints and
    // that it is free for this live range.
    if (CurrPhys != PhysReg && (!MRI->getRegClass(Reg)->contains(PhysReg) ||
                                Matrix->checkInterference(LI, PhysReg)))
      continue;

    LLVM_DEBUG(dbgs() << printReg(Reg, TRI) << '(' << printReg(CurrPhys, TRI)
                      << ") is recolorable.\n");

    // Gather the hint info.
    Info.clear();
    collectHintInfo(Reg, Info);
    // Check if recoloring the live-range will increase the cost of the
    // non-identity copies.
    if (CurrPhys != PhysReg) {
      LLVM_DEBUG(dbgs() << "Checking profitability:\n");
      BlockFrequency OldCopiesCost = getBrokenHintFreq(Info, CurrPhys);
      BlockFrequency NewCopiesCost = getBrokenHintFreq(Info, PhysReg);
      LLVM_DEBUG(dbgs() << "Old Cost: " << OldCopiesCost.getFrequency()
                        << "\nNew Cost: " << NewCopiesCost.getFrequency()
                        << '\n');
      if (OldCopiesCost < NewCopiesCost) {
        LLVM_DEBUG(dbgs() << "=> Not profitable.\n");
        continue;
      }
      // At this point, the cost is either cheaper or equal. If it is
      // equal, we consider this is profitable because it may expose
      // more recoloring opportunities.
      LLVM_DEBUG(dbgs() << "=> Profitable.\n");
      // Recolor the live-range.
      Matrix->unassign(LI);
      Matrix->assign(LI, PhysReg);
    }
    // Push all copy-related live-ranges to keep reconciling the broken
    // hints.
    for (const HintInfo &HI : Info) {
      if (Visited.insert(HI.Reg).second)
        RecoloringCandidates.push_back(HI.Reg);
    }
  } while (!RecoloringCandidates.empty());
}

/// Try to recolor broken hints.
/// Broken hints may be repaired by recoloring when an evicted variable
/// freed up a register for a larger live-range.
/// Consider the following example:
/// BB1:
///   a =
///   b =
/// BB2:
///   ...
///   = b
///   = a
/// Let us assume b gets split:
/// BB1:
///   a =
///   b =
/// BB2:
///   c = b
///   ...
///   d = c
///   = d
///   = a
/// Because of how the allocation work, b, c, and d may be assigned different
/// colors. Now, if a gets evicted later:
/// BB1:
///   a =
///   st a, SpillSlot
///   b =
/// BB2:
///   c = b
///   ...
///   d = c
///   = d
///   e = ld SpillSlot
///   = e
/// This is likely that we can assign the same register for b, c, and d,
/// getting rid of 2 copies.
void RAGreedy::tryHintsRecoloring() {
  for (LiveInterval *LI : SetOfBrokenHints) {
    assert(Register::isVirtualRegister(LI->reg()) &&
           "Recoloring is possible only for virtual registers");
    // Some dead defs may be around (e.g., because of debug uses).
    // Ignore those.
    if (!VRM->hasPhys(LI->reg()))
      continue;
    tryHintRecoloring(*LI);
  }
}

MCRegister RAGreedy::selectOrSplitImpl(LiveInterval &VirtReg,
                                       SmallVectorImpl<Register> &NewVRegs,
                                       SmallVirtRegSet &FixedRegisters,
                                       unsigned Depth) {
  uint8_t CostPerUseLimit = uint8_t(~0u);
  // First try assigning a free register.
  auto Order =
      AllocationOrder::create(VirtReg.reg(), *VRM, RegClassInfo, Matrix);
  if (MCRegister PhysReg =
          tryAssign(VirtReg, Order, NewVRegs, FixedRegisters)) {
    // If VirtReg got an assignment, the eviction info is no longer relevant.
    LastEvicted.clearEvicteeInfo(VirtReg.reg());
    // When NewVRegs is not empty, we may have made decisions such as evicting
    // a virtual register, go with the earlier decisions and use the physical
    // register.
    if (CSRCost.getFrequency() &&
        EvictAdvisor->isUnusedCalleeSavedReg(PhysReg) && NewVRegs.empty()) {
      MCRegister CSRReg = tryAssignCSRFirstTime(VirtReg, Order, PhysReg,
                                                CostPerUseLimit, NewVRegs);
      if (CSRReg || !NewVRegs.empty())
        // Return now if we decide to use a CSR or create new vregs due to
        // pre-splitting.
        return CSRReg;
    } else
      return PhysReg;
  }

  LiveRangeStage Stage = ExtraInfo->getStage(VirtReg);
  LLVM_DEBUG(dbgs() << StageName[Stage] << " Cascade "
                    << ExtraInfo->getCascade(VirtReg.reg()) << '\n');

  // Try to evict a less worthy live range, but only for ranges from the primary
  // queue. The RS_Split ranges already failed to do this, and they should not
  // get a second chance until they have been split.
  if (Stage != RS_Split)
    if (Register PhysReg =
            tryEvict(VirtReg, Order, NewVRegs, CostPerUseLimit,
                     FixedRegisters)) {
      Register Hint = MRI->getSimpleHint(VirtReg.reg());
      // If VirtReg has a hint and that hint is broken record this
      // virtual register as a recoloring candidate for broken hint.
      // Indeed, since we evicted a variable in its neighborhood it is
      // likely we can at least partially recolor some of the
      // copy-related live-ranges.
      if (Hint && Hint != PhysReg)
        SetOfBrokenHints.insert(&VirtReg);
      // If VirtReg eviction someone, the eviction info for it as an evictee is
      // no longer relevant.
      LastEvicted.clearEvicteeInfo(VirtReg.reg());
      return PhysReg;
    }

  assert((NewVRegs.empty() || Depth) && "Cannot append to existing NewVRegs");

  // The first time we see a live range, don't try to split or spill.
  // Wait until the second time, when all smaller ranges have been allocated.
  // This gives a better picture of the interference to split around.
  if (Stage < RS_Split) {
    ExtraInfo->setStage(VirtReg, RS_Split);
    LLVM_DEBUG(dbgs() << "wait for second round\n");
    NewVRegs.push_back(VirtReg.reg());
    return 0;
  }

  if (Stage < RS_Spill) {
    // Try splitting VirtReg or interferences.
    unsigned NewVRegSizeBefore = NewVRegs.size();
    Register PhysReg = trySplit(VirtReg, Order, NewVRegs, FixedRegisters);
    if (PhysReg || (NewVRegs.size() - NewVRegSizeBefore)) {
      // If VirtReg got split, the eviction info is no longer relevant.
      LastEvicted.clearEvicteeInfo(VirtReg.reg());
      return PhysReg;
    }
  }

  // If we couldn't allocate a register from spilling, there is probably some
  // invalid inline assembly. The base class will report it.
  if (Stage >= RS_Done || !VirtReg.isSpillable())
    return tryLastChanceRecoloring(VirtReg, Order, NewVRegs, FixedRegisters,
                                   Depth);

  // Finally spill VirtReg itself.
  if ((EnableDeferredSpilling ||
       TRI->shouldUseDeferredSpillingForVirtReg(*MF, VirtReg)) &&
      ExtraInfo->getStage(VirtReg) < RS_Memory) {
    // TODO: This is experimental and in particular, we do not model
    // the live range splitting done by spilling correctly.
    // We would need a deep integration with the spiller to do the
    // right thing here. Anyway, that is still good for early testing.
    ExtraInfo->setStage(VirtReg, RS_Memory);
    LLVM_DEBUG(dbgs() << "Do as if this register is in memory\n");
    NewVRegs.push_back(VirtReg.reg());
  } else {
    NamedRegionTimer T("spill", "Spiller", TimerGroupName,
                       TimerGroupDescription, TimePassesIsEnabled);
    LiveRangeEdit LRE(&VirtReg, NewVRegs, *MF, *LIS, VRM, this, &DeadRemats);
    spiller().spill(LRE);
    ExtraInfo->setStage(NewVRegs.begin(), NewVRegs.end(), RS_Done);

    // Tell LiveDebugVariables about the new ranges. Ranges not being covered by
    // the new regs are kept in LDV (still mapping to the old register), until
    // we rewrite spilled locations in LDV at a later stage.
    DebugVars->splitRegister(VirtReg.reg(), LRE.regs(), *LIS);

    if (VerifyEnabled)
      MF->verify(this, "After spilling");
  }

  // The live virtual register requesting allocation was spilled, so tell
  // the caller not to allocate anything during this round.
  return 0;
}

void RAGreedy::RAGreedyStats::report(MachineOptimizationRemarkMissed &R) {
  using namespace ore;
  if (Spills) {
    R << NV("NumSpills", Spills) << " spills ";
    R << NV("TotalSpillsCost", SpillsCost) << " total spills cost ";
  }
  if (FoldedSpills) {
    R << NV("NumFoldedSpills", FoldedSpills) << " folded spills ";
    R << NV("TotalFoldedSpillsCost", FoldedSpillsCost)
      << " total folded spills cost ";
  }
  if (Reloads) {
    R << NV("NumReloads", Reloads) << " reloads ";
    R << NV("TotalReloadsCost", ReloadsCost) << " total reloads cost ";
  }
  if (FoldedReloads) {
    R << NV("NumFoldedReloads", FoldedReloads) << " folded reloads ";
    R << NV("TotalFoldedReloadsCost", FoldedReloadsCost)
      << " total folded reloads cost ";
  }
  if (ZeroCostFoldedReloads)
    R << NV("NumZeroCostFoldedReloads", ZeroCostFoldedReloads)
      << " zero cost folded reloads ";
  if (Copies) {
    R << NV("NumVRCopies", Copies) << " virtual registers copies ";
    R << NV("TotalCopiesCost", CopiesCost) << " total copies cost ";
  }
}

RAGreedy::RAGreedyStats RAGreedy::computeStats(MachineBasicBlock &MBB) {
  RAGreedyStats Stats;
  const MachineFrameInfo &MFI = MF->getFrameInfo();
  int FI;

  auto isSpillSlotAccess = [&MFI](const MachineMemOperand *A) {
    return MFI.isSpillSlotObjectIndex(cast<FixedStackPseudoSourceValue>(
        A->getPseudoValue())->getFrameIndex());
  };
  auto isPatchpointInstr = [](const MachineInstr &MI) {
    return MI.getOpcode() == TargetOpcode::PATCHPOINT ||
           MI.getOpcode() == TargetOpcode::STACKMAP ||
           MI.getOpcode() == TargetOpcode::STATEPOINT;
  };
  for (MachineInstr &MI : MBB) {
    if (MI.isCopy()) {
      MachineOperand &Dest = MI.getOperand(0);
      MachineOperand &Src = MI.getOperand(1);
      if (Dest.isReg() && Src.isReg() && Dest.getReg().isVirtual() &&
          Src.getReg().isVirtual())
        ++Stats.Copies;
      continue;
    }

    SmallVector<const MachineMemOperand *, 2> Accesses;
    if (TII->isLoadFromStackSlot(MI, FI) && MFI.isSpillSlotObjectIndex(FI)) {
      ++Stats.Reloads;
      continue;
    }
    if (TII->isStoreToStackSlot(MI, FI) && MFI.isSpillSlotObjectIndex(FI)) {
      ++Stats.Spills;
      continue;
    }
    if (TII->hasLoadFromStackSlot(MI, Accesses) &&
        llvm::any_of(Accesses, isSpillSlotAccess)) {
      if (!isPatchpointInstr(MI)) {
        Stats.FoldedReloads += Accesses.size();
        continue;
      }
      // For statepoint there may be folded and zero cost folded stack reloads.
      std::pair<unsigned, unsigned> NonZeroCostRange =
          TII->getPatchpointUnfoldableRange(MI);
      SmallSet<unsigned, 16> FoldedReloads;
      SmallSet<unsigned, 16> ZeroCostFoldedReloads;
      for (unsigned Idx = 0, E = MI.getNumOperands(); Idx < E; ++Idx) {
        MachineOperand &MO = MI.getOperand(Idx);
        if (!MO.isFI() || !MFI.isSpillSlotObjectIndex(MO.getIndex()))
          continue;
        if (Idx >= NonZeroCostRange.first && Idx < NonZeroCostRange.second)
          FoldedReloads.insert(MO.getIndex());
        else
          ZeroCostFoldedReloads.insert(MO.getIndex());
      }
      // If stack slot is used in folded reload it is not zero cost then.
      for (unsigned Slot : FoldedReloads)
        ZeroCostFoldedReloads.erase(Slot);
      Stats.FoldedReloads += FoldedReloads.size();
      Stats.ZeroCostFoldedReloads += ZeroCostFoldedReloads.size();
      continue;
    }
    Accesses.clear();
    if (TII->hasStoreToStackSlot(MI, Accesses) &&
        llvm::any_of(Accesses, isSpillSlotAccess)) {
      Stats.FoldedSpills += Accesses.size();
    }
  }
  // Set cost of collected statistic by multiplication to relative frequency of
  // this basic block.
  float RelFreq = MBFI->getBlockFreqRelativeToEntryBlock(&MBB);
  Stats.ReloadsCost = RelFreq * Stats.Reloads;
  Stats.FoldedReloadsCost = RelFreq * Stats.FoldedReloads;
  Stats.SpillsCost = RelFreq * Stats.Spills;
  Stats.FoldedSpillsCost = RelFreq * Stats.FoldedSpills;
  Stats.CopiesCost = RelFreq * Stats.Copies;
  return Stats;
}

RAGreedy::RAGreedyStats RAGreedy::reportStats(MachineLoop *L) {
  RAGreedyStats Stats;

  // Sum up the spill and reloads in subloops.
  for (MachineLoop *SubLoop : *L)
    Stats.add(reportStats(SubLoop));

  for (MachineBasicBlock *MBB : L->getBlocks())
    // Handle blocks that were not included in subloops.
    if (Loops->getLoopFor(MBB) == L)
      Stats.add(computeStats(*MBB));

  if (!Stats.isEmpty()) {
    using namespace ore;

    ORE->emit([&]() {
      MachineOptimizationRemarkMissed R(DEBUG_TYPE, "LoopSpillReloadCopies",
                                        L->getStartLoc(), L->getHeader());
      Stats.report(R);
      R << "generated in loop";
      return R;
    });
  }
  return Stats;
}

void RAGreedy::reportStats() {
  if (!ORE->allowExtraAnalysis(DEBUG_TYPE))
    return;
  RAGreedyStats Stats;
  for (MachineLoop *L : *Loops)
    Stats.add(reportStats(L));
  // Process non-loop blocks.
  for (MachineBasicBlock &MBB : *MF)
    if (!Loops->getLoopFor(&MBB))
      Stats.add(computeStats(MBB));
  if (!Stats.isEmpty()) {
    using namespace ore;

    ORE->emit([&]() {
      DebugLoc Loc;
      if (auto *SP = MF->getFunction().getSubprogram())
        Loc = DILocation::get(SP->getContext(), SP->getLine(), 1, SP);
      MachineOptimizationRemarkMissed R(DEBUG_TYPE, "SpillReloadCopies", Loc,
                                        &MF->front());
      Stats.report(R);
      R << "generated in function";
      return R;
    });
  }
}

bool RAGreedy::runOnMachineFunction(MachineFunction &mf) {
  LLVM_DEBUG(dbgs() << "********** GREEDY REGISTER ALLOCATION **********\n"
                    << "********** Function: " << mf.getName() << '\n');

  MF = &mf;
  TRI = MF->getSubtarget().getRegisterInfo();
  TII = MF->getSubtarget().getInstrInfo();
  RCI.runOnMachineFunction(mf);

  EnableAdvancedRASplitCost =
      ConsiderLocalIntervalCost.getNumOccurrences()
          ? ConsiderLocalIntervalCost
          : MF->getSubtarget().enableAdvancedRASplitCost();

  if (VerifyEnabled)
    MF->verify(this, "Before greedy register allocator");

  RegAllocBase::init(getAnalysis<VirtRegMap>(),
                     getAnalysis<LiveIntervals>(),
                     getAnalysis<LiveRegMatrix>());
  Indexes = &getAnalysis<SlotIndexes>();
  MBFI = &getAnalysis<MachineBlockFrequencyInfo>();
  DomTree = &getAnalysis<MachineDominatorTree>();
  ORE = &getAnalysis<MachineOptimizationRemarkEmitterPass>().getORE();
  Loops = &getAnalysis<MachineLoopInfo>();
  Bundles = &getAnalysis<EdgeBundles>();
  SpillPlacer = &getAnalysis<SpillPlacement>();
  DebugVars = &getAnalysis<LiveDebugVariables>();
  AA = &getAnalysis<AAResultsWrapperPass>().getAAResults();

  initializeCSRCost();

  RegCosts = TRI->getRegisterCosts(*MF);

  ExtraInfo.emplace();
  EvictAdvisor =
      getAnalysis<RegAllocEvictionAdvisorAnalysis>().getAdvisor(*MF, *this);

  VRAI = std::make_unique<VirtRegAuxInfo>(*MF, *LIS, *VRM, *Loops, *MBFI);
  SpillerInstance.reset(createInlineSpiller(*this, *MF, *VRM, *VRAI));

  VRAI->calculateSpillWeightsAndHints();

  LLVM_DEBUG(LIS->dump());

  SA.reset(new SplitAnalysis(*VRM, *LIS, *Loops));
  SE.reset(new SplitEditor(*SA, *AA, *LIS, *VRM, *DomTree, *MBFI, *VRAI));

  IntfCache.init(MF, Matrix->getLiveUnions(), Indexes, LIS, TRI);
  GlobalCand.resize(32);  // This will grow as needed.
  SetOfBrokenHints.clear();
  LastEvicted.clear();

  allocatePhysRegs();
  tryHintsRecoloring();

  if (VerifyEnabled)
    MF->verify(this, "Before post optimization");
  postOptimization();
  reportStats();

  releaseMemory();
  return true;
}