1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
2189
2190
2191
2192
2193
2194
2195
2196
2197
2198
2199
2200
2201
2202
2203
2204
2205
2206
2207
2208
2209
2210
2211
2212
2213
2214
2215
2216
2217
2218
2219
2220
2221
2222
2223
2224
2225
2226
2227
2228
2229
2230
2231
2232
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
2244
2245
2246
2247
2248
2249
2250
2251
2252
2253
2254
2255
2256
2257
2258
2259
2260
2261
2262
2263
2264
2265
2266
2267
2268
2269
2270
2271
2272
2273
2274
2275
2276
2277
2278
2279
2280
2281
2282
2283
2284
2285
2286
2287
2288
2289
2290
2291
2292
2293
2294
2295
2296
2297
2298
2299
2300
2301
2302
2303
2304
2305
2306
2307
2308
2309
2310
2311
2312
2313
2314
2315
2316
2317
2318
2319
2320
2321
2322
2323
2324
2325
2326
2327
2328
2329
2330
2331
2332
2333
2334
2335
2336
2337
2338
2339
2340
2341
2342
2343
2344
2345
2346
2347
2348
2349
2350
2351
2352
2353
2354
2355
2356
2357
2358
2359
2360
2361
2362
2363
2364
2365
2366
2367
2368
2369
2370
2371
2372
2373
2374
2375
2376
2377
2378
2379
2380
2381
2382
2383
2384
2385
2386
2387
2388
2389
2390
2391
2392
2393
2394
2395
2396
2397
2398
2399
2400
2401
2402
2403
2404
2405
2406
2407
2408
2409
2410
2411
2412
2413
2414
2415
2416
2417
2418
2419
2420
2421
2422
2423
2424
2425
2426
2427
2428
2429
2430
2431
2432
2433
2434
2435
2436
2437
2438
2439
2440
2441
2442
2443
2444
2445
2446
2447
2448
2449
2450
2451
2452
2453
2454
2455
2456
2457
2458
2459
2460
2461
2462
2463
2464
2465
2466
2467
2468
2469
2470
2471
2472
2473
2474
2475
2476
2477
2478
2479
2480
2481
2482
2483
2484
2485
2486
2487
2488
2489
2490
2491
2492
2493
2494
2495
2496
2497
2498
2499
2500
2501
2502
2503
2504
2505
2506
2507
2508
2509
2510
2511
2512
2513
2514
2515
2516
2517
2518
2519
2520
2521
2522
2523
2524
2525
2526
2527
2528
2529
2530
2531
2532
2533
2534
2535
2536
2537
2538
2539
2540
2541
2542
2543
2544
2545
2546
2547
2548
2549
2550
2551
2552
2553
2554
2555
2556
2557
2558
2559
2560
2561
2562
2563
2564
2565
2566
2567
2568
2569
2570
2571
2572
2573
2574
2575
2576
2577
2578
2579
2580
2581
2582
2583
2584
2585
2586
2587
2588
2589
2590
2591
2592
2593
2594
2595
2596
2597
2598
2599
2600
2601
2602
2603
2604
2605
2606
2607
2608
2609
2610
2611
2612
2613
2614
2615
2616
2617
2618
2619
2620
2621
2622
2623
2624
2625
2626
2627
2628
2629
2630
2631
2632
2633
2634
2635
2636
2637
2638
2639
2640
2641
2642
2643
2644
2645
2646
2647
2648
2649
2650
2651
2652
2653
2654
2655
2656
2657
2658
2659
2660
2661
2662
2663
2664
2665
2666
2667
2668
2669
2670
2671
2672
2673
2674
2675
2676
2677
2678
2679
2680
2681
2682
2683
2684
2685
2686
2687
2688
2689
2690
2691
2692
2693
2694
2695
2696
2697
2698
2699
2700
2701
2702
2703
2704
2705
2706
2707
2708
2709
2710
2711
2712
2713
2714
2715
2716
2717
2718
2719
2720
2721
2722
2723
2724
2725
2726
2727
2728
2729
2730
2731
2732
2733
2734
2735
2736
2737
2738
2739
2740
2741
2742
2743
2744
2745
2746
2747
2748
2749
2750
2751
2752
2753
2754
2755
2756
2757
2758
2759
2760
2761
2762
2763
2764
2765
2766
2767
2768
2769
2770
2771
2772
2773
2774
2775
2776
2777
2778
2779
2780
2781
2782
2783
2784
2785
2786
2787
2788
2789
2790
2791
2792
2793
2794
2795
2796
2797
2798
2799
2800
2801
2802
2803
2804
2805
2806
2807
2808
2809
2810
2811
2812
2813
2814
2815
2816
2817
2818
2819
2820
2821
2822
2823
2824
2825
2826
2827
2828
2829
2830
2831
2832
2833
2834
2835
2836
2837
2838
2839
2840
2841
2842
2843
2844
2845
2846
2847
2848
2849
2850
2851
2852
2853
2854
2855
2856
2857
2858
2859
2860
2861
2862
2863
2864
2865
2866
2867
2868
2869
2870
2871
2872
2873
2874
2875
2876
2877
2878
2879
2880
2881
2882
2883
2884
2885
2886
2887
2888
2889
2890
2891
2892
2893
2894
2895
2896
2897
2898
2899
2900
2901
2902
2903
2904
2905
2906
2907
2908
2909
2910
2911
2912
2913
2914
2915
2916
2917
2918
2919
2920
2921
2922
2923
2924
2925
2926
2927
2928
2929
2930
2931
2932
2933
2934
2935
2936
2937
2938
2939
2940
2941
2942
2943
2944
2945
2946
2947
2948
2949
2950
2951
2952
2953
2954
2955
2956
2957
2958
2959
2960
2961
2962
2963
2964
2965
2966
2967
2968
2969
2970
2971
2972
2973
2974
2975
2976
2977
2978
2979
2980
2981
2982
2983
|
//===- SimplifyCFG.cpp - Code to perform CFG simplification ---------------===//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// Peephole optimize the CFG.
//
//===----------------------------------------------------------------------===//
#define DEBUG_TYPE "simplifycfg"
#include "llvm/Transforms/Utils/Local.h"
#include "llvm/Constants.h"
#include "llvm/DerivedTypes.h"
#include "llvm/GlobalVariable.h"
#include "llvm/Instructions.h"
#include "llvm/IntrinsicInst.h"
#include "llvm/LLVMContext.h"
#include "llvm/Metadata.h"
#include "llvm/Operator.h"
#include "llvm/Type.h"
#include "llvm/Analysis/InstructionSimplify.h"
#include "llvm/Analysis/ValueTracking.h"
#include "llvm/Target/TargetData.h"
#include "llvm/Transforms/Utils/BasicBlockUtils.h"
#include "llvm/ADT/DenseMap.h"
#include "llvm/ADT/SetVector.h"
#include "llvm/ADT/SmallVector.h"
#include "llvm/ADT/SmallPtrSet.h"
#include "llvm/ADT/Statistic.h"
#include "llvm/ADT/STLExtras.h"
#include "llvm/Support/CFG.h"
#include "llvm/Support/CommandLine.h"
#include "llvm/Support/ConstantRange.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/IRBuilder.h"
#include "llvm/Support/NoFolder.h"
#include "llvm/Support/raw_ostream.h"
#include <algorithm>
#include <set>
#include <map>
using namespace llvm;
static cl::opt<unsigned>
PHINodeFoldingThreshold("phi-node-folding-threshold", cl::Hidden, cl::init(1),
cl::desc("Control the amount of phi node folding to perform (default = 1)"));
static cl::opt<bool>
DupRet("simplifycfg-dup-ret", cl::Hidden, cl::init(false),
cl::desc("Duplicate return instructions into unconditional branches"));
STATISTIC(NumSpeculations, "Number of speculative executed instructions");
namespace {
class SimplifyCFGOpt {
const TargetData *const TD;
Value *isValueEqualityComparison(TerminatorInst *TI);
BasicBlock *GetValueEqualityComparisonCases(TerminatorInst *TI,
std::vector<std::pair<ConstantInt*, BasicBlock*> > &Cases);
bool SimplifyEqualityComparisonWithOnlyPredecessor(TerminatorInst *TI,
BasicBlock *Pred,
IRBuilder<> &Builder);
bool FoldValueComparisonIntoPredecessors(TerminatorInst *TI,
IRBuilder<> &Builder);
bool SimplifyReturn(ReturnInst *RI, IRBuilder<> &Builder);
bool SimplifyResume(ResumeInst *RI, IRBuilder<> &Builder);
bool SimplifyUnreachable(UnreachableInst *UI);
bool SimplifySwitch(SwitchInst *SI, IRBuilder<> &Builder);
bool SimplifyIndirectBr(IndirectBrInst *IBI);
bool SimplifyUncondBranch(BranchInst *BI, IRBuilder <> &Builder);
bool SimplifyCondBranch(BranchInst *BI, IRBuilder <>&Builder);
public:
explicit SimplifyCFGOpt(const TargetData *td) : TD(td) {}
bool run(BasicBlock *BB);
};
}
/// SafeToMergeTerminators - Return true if it is safe to merge these two
/// terminator instructions together.
///
static bool SafeToMergeTerminators(TerminatorInst *SI1, TerminatorInst *SI2) {
if (SI1 == SI2) return false; // Can't merge with self!
// It is not safe to merge these two switch instructions if they have a common
// successor, and if that successor has a PHI node, and if *that* PHI node has
// conflicting incoming values from the two switch blocks.
BasicBlock *SI1BB = SI1->getParent();
BasicBlock *SI2BB = SI2->getParent();
SmallPtrSet<BasicBlock*, 16> SI1Succs(succ_begin(SI1BB), succ_end(SI1BB));
for (succ_iterator I = succ_begin(SI2BB), E = succ_end(SI2BB); I != E; ++I)
if (SI1Succs.count(*I))
for (BasicBlock::iterator BBI = (*I)->begin();
isa<PHINode>(BBI); ++BBI) {
PHINode *PN = cast<PHINode>(BBI);
if (PN->getIncomingValueForBlock(SI1BB) !=
PN->getIncomingValueForBlock(SI2BB))
return false;
}
return true;
}
/// AddPredecessorToBlock - Update PHI nodes in Succ to indicate that there will
/// now be entries in it from the 'NewPred' block. The values that will be
/// flowing into the PHI nodes will be the same as those coming in from
/// ExistPred, an existing predecessor of Succ.
static void AddPredecessorToBlock(BasicBlock *Succ, BasicBlock *NewPred,
BasicBlock *ExistPred) {
if (!isa<PHINode>(Succ->begin())) return; // Quick exit if nothing to do
PHINode *PN;
for (BasicBlock::iterator I = Succ->begin();
(PN = dyn_cast<PHINode>(I)); ++I)
PN->addIncoming(PN->getIncomingValueForBlock(ExistPred), NewPred);
}
/// GetIfCondition - Given a basic block (BB) with two predecessors (and at
/// least one PHI node in it), check to see if the merge at this block is due
/// to an "if condition". If so, return the boolean condition that determines
/// which entry into BB will be taken. Also, return by references the block
/// that will be entered from if the condition is true, and the block that will
/// be entered if the condition is false.
///
/// This does no checking to see if the true/false blocks have large or unsavory
/// instructions in them.
static Value *GetIfCondition(BasicBlock *BB, BasicBlock *&IfTrue,
BasicBlock *&IfFalse) {
PHINode *SomePHI = cast<PHINode>(BB->begin());
assert(SomePHI->getNumIncomingValues() == 2 &&
"Function can only handle blocks with 2 predecessors!");
BasicBlock *Pred1 = SomePHI->getIncomingBlock(0);
BasicBlock *Pred2 = SomePHI->getIncomingBlock(1);
// We can only handle branches. Other control flow will be lowered to
// branches if possible anyway.
BranchInst *Pred1Br = dyn_cast<BranchInst>(Pred1->getTerminator());
BranchInst *Pred2Br = dyn_cast<BranchInst>(Pred2->getTerminator());
if (Pred1Br == 0 || Pred2Br == 0)
return 0;
// Eliminate code duplication by ensuring that Pred1Br is conditional if
// either are.
if (Pred2Br->isConditional()) {
// If both branches are conditional, we don't have an "if statement". In
// reality, we could transform this case, but since the condition will be
// required anyway, we stand no chance of eliminating it, so the xform is
// probably not profitable.
if (Pred1Br->isConditional())
return 0;
std::swap(Pred1, Pred2);
std::swap(Pred1Br, Pred2Br);
}
if (Pred1Br->isConditional()) {
// The only thing we have to watch out for here is to make sure that Pred2
// doesn't have incoming edges from other blocks. If it does, the condition
// doesn't dominate BB.
if (Pred2->getSinglePredecessor() == 0)
return 0;
// If we found a conditional branch predecessor, make sure that it branches
// to BB and Pred2Br. If it doesn't, this isn't an "if statement".
if (Pred1Br->getSuccessor(0) == BB &&
Pred1Br->getSuccessor(1) == Pred2) {
IfTrue = Pred1;
IfFalse = Pred2;
} else if (Pred1Br->getSuccessor(0) == Pred2 &&
Pred1Br->getSuccessor(1) == BB) {
IfTrue = Pred2;
IfFalse = Pred1;
} else {
// We know that one arm of the conditional goes to BB, so the other must
// go somewhere unrelated, and this must not be an "if statement".
return 0;
}
return Pred1Br->getCondition();
}
// Ok, if we got here, both predecessors end with an unconditional branch to
// BB. Don't panic! If both blocks only have a single (identical)
// predecessor, and THAT is a conditional branch, then we're all ok!
BasicBlock *CommonPred = Pred1->getSinglePredecessor();
if (CommonPred == 0 || CommonPred != Pred2->getSinglePredecessor())
return 0;
// Otherwise, if this is a conditional branch, then we can use it!
BranchInst *BI = dyn_cast<BranchInst>(CommonPred->getTerminator());
if (BI == 0) return 0;
assert(BI->isConditional() && "Two successors but not conditional?");
if (BI->getSuccessor(0) == Pred1) {
IfTrue = Pred1;
IfFalse = Pred2;
} else {
IfTrue = Pred2;
IfFalse = Pred1;
}
return BI->getCondition();
}
/// ComputeSpeculuationCost - Compute an abstract "cost" of speculating the
/// given instruction, which is assumed to be safe to speculate. 1 means
/// cheap, 2 means less cheap, and UINT_MAX means prohibitively expensive.
static unsigned ComputeSpeculationCost(const User *I) {
assert(isSafeToSpeculativelyExecute(I) &&
"Instruction is not safe to speculatively execute!");
switch (Operator::getOpcode(I)) {
default:
// In doubt, be conservative.
return UINT_MAX;
case Instruction::GetElementPtr:
// GEPs are cheap if all indices are constant.
if (!cast<GEPOperator>(I)->hasAllConstantIndices())
return UINT_MAX;
return 1;
case Instruction::Load:
case Instruction::Add:
case Instruction::Sub:
case Instruction::And:
case Instruction::Or:
case Instruction::Xor:
case Instruction::Shl:
case Instruction::LShr:
case Instruction::AShr:
case Instruction::ICmp:
case Instruction::Trunc:
case Instruction::ZExt:
case Instruction::SExt:
return 1; // These are all cheap.
case Instruction::Call:
case Instruction::Select:
return 2;
}
}
/// DominatesMergePoint - If we have a merge point of an "if condition" as
/// accepted above, return true if the specified value dominates the block. We
/// don't handle the true generality of domination here, just a special case
/// which works well enough for us.
///
/// If AggressiveInsts is non-null, and if V does not dominate BB, we check to
/// see if V (which must be an instruction) and its recursive operands
/// that do not dominate BB have a combined cost lower than CostRemaining and
/// are non-trapping. If both are true, the instruction is inserted into the
/// set and true is returned.
///
/// The cost for most non-trapping instructions is defined as 1 except for
/// Select whose cost is 2.
///
/// After this function returns, CostRemaining is decreased by the cost of
/// V plus its non-dominating operands. If that cost is greater than
/// CostRemaining, false is returned and CostRemaining is undefined.
static bool DominatesMergePoint(Value *V, BasicBlock *BB,
SmallPtrSet<Instruction*, 4> *AggressiveInsts,
unsigned &CostRemaining) {
Instruction *I = dyn_cast<Instruction>(V);
if (!I) {
// Non-instructions all dominate instructions, but not all constantexprs
// can be executed unconditionally.
if (ConstantExpr *C = dyn_cast<ConstantExpr>(V))
if (C->canTrap())
return false;
return true;
}
BasicBlock *PBB = I->getParent();
// We don't want to allow weird loops that might have the "if condition" in
// the bottom of this block.
if (PBB == BB) return false;
// If this instruction is defined in a block that contains an unconditional
// branch to BB, then it must be in the 'conditional' part of the "if
// statement". If not, it definitely dominates the region.
BranchInst *BI = dyn_cast<BranchInst>(PBB->getTerminator());
if (BI == 0 || BI->isConditional() || BI->getSuccessor(0) != BB)
return true;
// If we aren't allowing aggressive promotion anymore, then don't consider
// instructions in the 'if region'.
if (AggressiveInsts == 0) return false;
// If we have seen this instruction before, don't count it again.
if (AggressiveInsts->count(I)) return true;
// Okay, it looks like the instruction IS in the "condition". Check to
// see if it's a cheap instruction to unconditionally compute, and if it
// only uses stuff defined outside of the condition. If so, hoist it out.
if (!isSafeToSpeculativelyExecute(I))
return false;
unsigned Cost = ComputeSpeculationCost(I);
if (Cost > CostRemaining)
return false;
CostRemaining -= Cost;
// Okay, we can only really hoist these out if their operands do
// not take us over the cost threshold.
for (User::op_iterator i = I->op_begin(), e = I->op_end(); i != e; ++i)
if (!DominatesMergePoint(*i, BB, AggressiveInsts, CostRemaining))
return false;
// Okay, it's safe to do this! Remember this instruction.
AggressiveInsts->insert(I);
return true;
}
/// GetConstantInt - Extract ConstantInt from value, looking through IntToPtr
/// and PointerNullValue. Return NULL if value is not a constant int.
static ConstantInt *GetConstantInt(Value *V, const TargetData *TD) {
// Normal constant int.
ConstantInt *CI = dyn_cast<ConstantInt>(V);
if (CI || !TD || !isa<Constant>(V) || !V->getType()->isPointerTy())
return CI;
// This is some kind of pointer constant. Turn it into a pointer-sized
// ConstantInt if possible.
IntegerType *PtrTy = TD->getIntPtrType(V->getContext());
// Null pointer means 0, see SelectionDAGBuilder::getValue(const Value*).
if (isa<ConstantPointerNull>(V))
return ConstantInt::get(PtrTy, 0);
// IntToPtr const int.
if (ConstantExpr *CE = dyn_cast<ConstantExpr>(V))
if (CE->getOpcode() == Instruction::IntToPtr)
if (ConstantInt *CI = dyn_cast<ConstantInt>(CE->getOperand(0))) {
// The constant is very likely to have the right type already.
if (CI->getType() == PtrTy)
return CI;
else
return cast<ConstantInt>
(ConstantExpr::getIntegerCast(CI, PtrTy, /*isSigned=*/false));
}
return 0;
}
/// GatherConstantCompares - Given a potentially 'or'd or 'and'd together
/// collection of icmp eq/ne instructions that compare a value against a
/// constant, return the value being compared, and stick the constant into the
/// Values vector.
static Value *
GatherConstantCompares(Value *V, std::vector<ConstantInt*> &Vals, Value *&Extra,
const TargetData *TD, bool isEQ, unsigned &UsedICmps) {
Instruction *I = dyn_cast<Instruction>(V);
if (I == 0) return 0;
// If this is an icmp against a constant, handle this as one of the cases.
if (ICmpInst *ICI = dyn_cast<ICmpInst>(I)) {
if (ConstantInt *C = GetConstantInt(I->getOperand(1), TD)) {
if (ICI->getPredicate() == (isEQ ? ICmpInst::ICMP_EQ:ICmpInst::ICMP_NE)) {
UsedICmps++;
Vals.push_back(C);
return I->getOperand(0);
}
// If we have "x ult 3" comparison, for example, then we can add 0,1,2 to
// the set.
ConstantRange Span =
ConstantRange::makeICmpRegion(ICI->getPredicate(), C->getValue());
// If this is an and/!= check then we want to optimize "x ugt 2" into
// x != 0 && x != 1.
if (!isEQ)
Span = Span.inverse();
// If there are a ton of values, we don't want to make a ginormous switch.
if (Span.getSetSize().ugt(8) || Span.isEmptySet())
return 0;
for (APInt Tmp = Span.getLower(); Tmp != Span.getUpper(); ++Tmp)
Vals.push_back(ConstantInt::get(V->getContext(), Tmp));
UsedICmps++;
return I->getOperand(0);
}
return 0;
}
// Otherwise, we can only handle an | or &, depending on isEQ.
if (I->getOpcode() != (isEQ ? Instruction::Or : Instruction::And))
return 0;
unsigned NumValsBeforeLHS = Vals.size();
unsigned UsedICmpsBeforeLHS = UsedICmps;
if (Value *LHS = GatherConstantCompares(I->getOperand(0), Vals, Extra, TD,
isEQ, UsedICmps)) {
unsigned NumVals = Vals.size();
unsigned UsedICmpsBeforeRHS = UsedICmps;
if (Value *RHS = GatherConstantCompares(I->getOperand(1), Vals, Extra, TD,
isEQ, UsedICmps)) {
if (LHS == RHS)
return LHS;
Vals.resize(NumVals);
UsedICmps = UsedICmpsBeforeRHS;
}
// The RHS of the or/and can't be folded in and we haven't used "Extra" yet,
// set it and return success.
if (Extra == 0 || Extra == I->getOperand(1)) {
Extra = I->getOperand(1);
return LHS;
}
Vals.resize(NumValsBeforeLHS);
UsedICmps = UsedICmpsBeforeLHS;
return 0;
}
// If the LHS can't be folded in, but Extra is available and RHS can, try to
// use LHS as Extra.
if (Extra == 0 || Extra == I->getOperand(0)) {
Value *OldExtra = Extra;
Extra = I->getOperand(0);
if (Value *RHS = GatherConstantCompares(I->getOperand(1), Vals, Extra, TD,
isEQ, UsedICmps))
return RHS;
assert(Vals.size() == NumValsBeforeLHS);
Extra = OldExtra;
}
return 0;
}
static void EraseTerminatorInstAndDCECond(TerminatorInst *TI) {
Instruction *Cond = 0;
if (SwitchInst *SI = dyn_cast<SwitchInst>(TI)) {
Cond = dyn_cast<Instruction>(SI->getCondition());
} else if (BranchInst *BI = dyn_cast<BranchInst>(TI)) {
if (BI->isConditional())
Cond = dyn_cast<Instruction>(BI->getCondition());
} else if (IndirectBrInst *IBI = dyn_cast<IndirectBrInst>(TI)) {
Cond = dyn_cast<Instruction>(IBI->getAddress());
}
TI->eraseFromParent();
if (Cond) RecursivelyDeleteTriviallyDeadInstructions(Cond);
}
/// isValueEqualityComparison - Return true if the specified terminator checks
/// to see if a value is equal to constant integer value.
Value *SimplifyCFGOpt::isValueEqualityComparison(TerminatorInst *TI) {
Value *CV = 0;
if (SwitchInst *SI = dyn_cast<SwitchInst>(TI)) {
// Do not permit merging of large switch instructions into their
// predecessors unless there is only one predecessor.
if (SI->getNumSuccessors()*std::distance(pred_begin(SI->getParent()),
pred_end(SI->getParent())) <= 128)
CV = SI->getCondition();
} else if (BranchInst *BI = dyn_cast<BranchInst>(TI))
if (BI->isConditional() && BI->getCondition()->hasOneUse())
if (ICmpInst *ICI = dyn_cast<ICmpInst>(BI->getCondition()))
if ((ICI->getPredicate() == ICmpInst::ICMP_EQ ||
ICI->getPredicate() == ICmpInst::ICMP_NE) &&
GetConstantInt(ICI->getOperand(1), TD))
CV = ICI->getOperand(0);
// Unwrap any lossless ptrtoint cast.
if (TD && CV && CV->getType() == TD->getIntPtrType(CV->getContext()))
if (PtrToIntInst *PTII = dyn_cast<PtrToIntInst>(CV))
CV = PTII->getOperand(0);
return CV;
}
/// GetValueEqualityComparisonCases - Given a value comparison instruction,
/// decode all of the 'cases' that it represents and return the 'default' block.
BasicBlock *SimplifyCFGOpt::
GetValueEqualityComparisonCases(TerminatorInst *TI,
std::vector<std::pair<ConstantInt*,
BasicBlock*> > &Cases) {
if (SwitchInst *SI = dyn_cast<SwitchInst>(TI)) {
Cases.reserve(SI->getNumCases());
for (unsigned i = 0, e = SI->getNumCases(); i != e; ++i)
Cases.push_back(std::make_pair(SI->getCaseValue(i),
SI->getCaseSuccessor(i)));
return SI->getDefaultDest();
}
BranchInst *BI = cast<BranchInst>(TI);
ICmpInst *ICI = cast<ICmpInst>(BI->getCondition());
Cases.push_back(std::make_pair(GetConstantInt(ICI->getOperand(1), TD),
BI->getSuccessor(ICI->getPredicate() ==
ICmpInst::ICMP_NE)));
return BI->getSuccessor(ICI->getPredicate() == ICmpInst::ICMP_EQ);
}
/// EliminateBlockCases - Given a vector of bb/value pairs, remove any entries
/// in the list that match the specified block.
static void EliminateBlockCases(BasicBlock *BB,
std::vector<std::pair<ConstantInt*, BasicBlock*> > &Cases) {
for (unsigned i = 0, e = Cases.size(); i != e; ++i)
if (Cases[i].second == BB) {
Cases.erase(Cases.begin()+i);
--i; --e;
}
}
/// ValuesOverlap - Return true if there are any keys in C1 that exist in C2 as
/// well.
static bool
ValuesOverlap(std::vector<std::pair<ConstantInt*, BasicBlock*> > &C1,
std::vector<std::pair<ConstantInt*, BasicBlock*> > &C2) {
std::vector<std::pair<ConstantInt*, BasicBlock*> > *V1 = &C1, *V2 = &C2;
// Make V1 be smaller than V2.
if (V1->size() > V2->size())
std::swap(V1, V2);
if (V1->size() == 0) return false;
if (V1->size() == 1) {
// Just scan V2.
ConstantInt *TheVal = (*V1)[0].first;
for (unsigned i = 0, e = V2->size(); i != e; ++i)
if (TheVal == (*V2)[i].first)
return true;
}
// Otherwise, just sort both lists and compare element by element.
array_pod_sort(V1->begin(), V1->end());
array_pod_sort(V2->begin(), V2->end());
unsigned i1 = 0, i2 = 0, e1 = V1->size(), e2 = V2->size();
while (i1 != e1 && i2 != e2) {
if ((*V1)[i1].first == (*V2)[i2].first)
return true;
if ((*V1)[i1].first < (*V2)[i2].first)
++i1;
else
++i2;
}
return false;
}
/// SimplifyEqualityComparisonWithOnlyPredecessor - If TI is known to be a
/// terminator instruction and its block is known to only have a single
/// predecessor block, check to see if that predecessor is also a value
/// comparison with the same value, and if that comparison determines the
/// outcome of this comparison. If so, simplify TI. This does a very limited
/// form of jump threading.
bool SimplifyCFGOpt::
SimplifyEqualityComparisonWithOnlyPredecessor(TerminatorInst *TI,
BasicBlock *Pred,
IRBuilder<> &Builder) {
Value *PredVal = isValueEqualityComparison(Pred->getTerminator());
if (!PredVal) return false; // Not a value comparison in predecessor.
Value *ThisVal = isValueEqualityComparison(TI);
assert(ThisVal && "This isn't a value comparison!!");
if (ThisVal != PredVal) return false; // Different predicates.
// Find out information about when control will move from Pred to TI's block.
std::vector<std::pair<ConstantInt*, BasicBlock*> > PredCases;
BasicBlock *PredDef = GetValueEqualityComparisonCases(Pred->getTerminator(),
PredCases);
EliminateBlockCases(PredDef, PredCases); // Remove default from cases.
// Find information about how control leaves this block.
std::vector<std::pair<ConstantInt*, BasicBlock*> > ThisCases;
BasicBlock *ThisDef = GetValueEqualityComparisonCases(TI, ThisCases);
EliminateBlockCases(ThisDef, ThisCases); // Remove default from cases.
// If TI's block is the default block from Pred's comparison, potentially
// simplify TI based on this knowledge.
if (PredDef == TI->getParent()) {
// If we are here, we know that the value is none of those cases listed in
// PredCases. If there are any cases in ThisCases that are in PredCases, we
// can simplify TI.
if (!ValuesOverlap(PredCases, ThisCases))
return false;
if (isa<BranchInst>(TI)) {
// Okay, one of the successors of this condbr is dead. Convert it to a
// uncond br.
assert(ThisCases.size() == 1 && "Branch can only have one case!");
// Insert the new branch.
Instruction *NI = Builder.CreateBr(ThisDef);
(void) NI;
// Remove PHI node entries for the dead edge.
ThisCases[0].second->removePredecessor(TI->getParent());
DEBUG(dbgs() << "Threading pred instr: " << *Pred->getTerminator()
<< "Through successor TI: " << *TI << "Leaving: " << *NI << "\n");
EraseTerminatorInstAndDCECond(TI);
return true;
}
SwitchInst *SI = cast<SwitchInst>(TI);
// Okay, TI has cases that are statically dead, prune them away.
SmallPtrSet<Constant*, 16> DeadCases;
for (unsigned i = 0, e = PredCases.size(); i != e; ++i)
DeadCases.insert(PredCases[i].first);
DEBUG(dbgs() << "Threading pred instr: " << *Pred->getTerminator()
<< "Through successor TI: " << *TI);
for (unsigned i = SI->getNumCases(); i != 0;) {
--i;
if (DeadCases.count(SI->getCaseValue(i))) {
SI->getCaseSuccessor(i)->removePredecessor(TI->getParent());
SI->removeCase(i);
}
}
DEBUG(dbgs() << "Leaving: " << *TI << "\n");
return true;
}
// Otherwise, TI's block must correspond to some matched value. Find out
// which value (or set of values) this is.
ConstantInt *TIV = 0;
BasicBlock *TIBB = TI->getParent();
for (unsigned i = 0, e = PredCases.size(); i != e; ++i)
if (PredCases[i].second == TIBB) {
if (TIV != 0)
return false; // Cannot handle multiple values coming to this block.
TIV = PredCases[i].first;
}
assert(TIV && "No edge from pred to succ?");
// Okay, we found the one constant that our value can be if we get into TI's
// BB. Find out which successor will unconditionally be branched to.
BasicBlock *TheRealDest = 0;
for (unsigned i = 0, e = ThisCases.size(); i != e; ++i)
if (ThisCases[i].first == TIV) {
TheRealDest = ThisCases[i].second;
break;
}
// If not handled by any explicit cases, it is handled by the default case.
if (TheRealDest == 0) TheRealDest = ThisDef;
// Remove PHI node entries for dead edges.
BasicBlock *CheckEdge = TheRealDest;
for (succ_iterator SI = succ_begin(TIBB), e = succ_end(TIBB); SI != e; ++SI)
if (*SI != CheckEdge)
(*SI)->removePredecessor(TIBB);
else
CheckEdge = 0;
// Insert the new branch.
Instruction *NI = Builder.CreateBr(TheRealDest);
(void) NI;
DEBUG(dbgs() << "Threading pred instr: " << *Pred->getTerminator()
<< "Through successor TI: " << *TI << "Leaving: " << *NI << "\n");
EraseTerminatorInstAndDCECond(TI);
return true;
}
namespace {
/// ConstantIntOrdering - This class implements a stable ordering of constant
/// integers that does not depend on their address. This is important for
/// applications that sort ConstantInt's to ensure uniqueness.
struct ConstantIntOrdering {
bool operator()(const ConstantInt *LHS, const ConstantInt *RHS) const {
return LHS->getValue().ult(RHS->getValue());
}
};
}
static int ConstantIntSortPredicate(const void *P1, const void *P2) {
const ConstantInt *LHS = *(const ConstantInt**)P1;
const ConstantInt *RHS = *(const ConstantInt**)P2;
if (LHS->getValue().ult(RHS->getValue()))
return 1;
if (LHS->getValue() == RHS->getValue())
return 0;
return -1;
}
/// FoldValueComparisonIntoPredecessors - The specified terminator is a value
/// equality comparison instruction (either a switch or a branch on "X == c").
/// See if any of the predecessors of the terminator block are value comparisons
/// on the same value. If so, and if safe to do so, fold them together.
bool SimplifyCFGOpt::FoldValueComparisonIntoPredecessors(TerminatorInst *TI,
IRBuilder<> &Builder) {
BasicBlock *BB = TI->getParent();
Value *CV = isValueEqualityComparison(TI); // CondVal
assert(CV && "Not a comparison?");
bool Changed = false;
SmallVector<BasicBlock*, 16> Preds(pred_begin(BB), pred_end(BB));
while (!Preds.empty()) {
BasicBlock *Pred = Preds.pop_back_val();
// See if the predecessor is a comparison with the same value.
TerminatorInst *PTI = Pred->getTerminator();
Value *PCV = isValueEqualityComparison(PTI); // PredCondVal
if (PCV == CV && SafeToMergeTerminators(TI, PTI)) {
// Figure out which 'cases' to copy from SI to PSI.
std::vector<std::pair<ConstantInt*, BasicBlock*> > BBCases;
BasicBlock *BBDefault = GetValueEqualityComparisonCases(TI, BBCases);
std::vector<std::pair<ConstantInt*, BasicBlock*> > PredCases;
BasicBlock *PredDefault = GetValueEqualityComparisonCases(PTI, PredCases);
// Based on whether the default edge from PTI goes to BB or not, fill in
// PredCases and PredDefault with the new switch cases we would like to
// build.
SmallVector<BasicBlock*, 8> NewSuccessors;
if (PredDefault == BB) {
// If this is the default destination from PTI, only the edges in TI
// that don't occur in PTI, or that branch to BB will be activated.
std::set<ConstantInt*, ConstantIntOrdering> PTIHandled;
for (unsigned i = 0, e = PredCases.size(); i != e; ++i)
if (PredCases[i].second != BB)
PTIHandled.insert(PredCases[i].first);
else {
// The default destination is BB, we don't need explicit targets.
std::swap(PredCases[i], PredCases.back());
PredCases.pop_back();
--i; --e;
}
// Reconstruct the new switch statement we will be building.
if (PredDefault != BBDefault) {
PredDefault->removePredecessor(Pred);
PredDefault = BBDefault;
NewSuccessors.push_back(BBDefault);
}
for (unsigned i = 0, e = BBCases.size(); i != e; ++i)
if (!PTIHandled.count(BBCases[i].first) &&
BBCases[i].second != BBDefault) {
PredCases.push_back(BBCases[i]);
NewSuccessors.push_back(BBCases[i].second);
}
} else {
// If this is not the default destination from PSI, only the edges
// in SI that occur in PSI with a destination of BB will be
// activated.
std::set<ConstantInt*, ConstantIntOrdering> PTIHandled;
for (unsigned i = 0, e = PredCases.size(); i != e; ++i)
if (PredCases[i].second == BB) {
PTIHandled.insert(PredCases[i].first);
std::swap(PredCases[i], PredCases.back());
PredCases.pop_back();
--i; --e;
}
// Okay, now we know which constants were sent to BB from the
// predecessor. Figure out where they will all go now.
for (unsigned i = 0, e = BBCases.size(); i != e; ++i)
if (PTIHandled.count(BBCases[i].first)) {
// If this is one we are capable of getting...
PredCases.push_back(BBCases[i]);
NewSuccessors.push_back(BBCases[i].second);
PTIHandled.erase(BBCases[i].first);// This constant is taken care of
}
// If there are any constants vectored to BB that TI doesn't handle,
// they must go to the default destination of TI.
for (std::set<ConstantInt*, ConstantIntOrdering>::iterator I =
PTIHandled.begin(),
E = PTIHandled.end(); I != E; ++I) {
PredCases.push_back(std::make_pair(*I, BBDefault));
NewSuccessors.push_back(BBDefault);
}
}
// Okay, at this point, we know which new successor Pred will get. Make
// sure we update the number of entries in the PHI nodes for these
// successors.
for (unsigned i = 0, e = NewSuccessors.size(); i != e; ++i)
AddPredecessorToBlock(NewSuccessors[i], Pred, BB);
Builder.SetInsertPoint(PTI);
// Convert pointer to int before we switch.
if (CV->getType()->isPointerTy()) {
assert(TD && "Cannot switch on pointer without TargetData");
CV = Builder.CreatePtrToInt(CV, TD->getIntPtrType(CV->getContext()),
"magicptr");
}
// Now that the successors are updated, create the new Switch instruction.
SwitchInst *NewSI = Builder.CreateSwitch(CV, PredDefault,
PredCases.size());
NewSI->setDebugLoc(PTI->getDebugLoc());
for (unsigned i = 0, e = PredCases.size(); i != e; ++i)
NewSI->addCase(PredCases[i].first, PredCases[i].second);
EraseTerminatorInstAndDCECond(PTI);
// Okay, last check. If BB is still a successor of PSI, then we must
// have an infinite loop case. If so, add an infinitely looping block
// to handle the case to preserve the behavior of the code.
BasicBlock *InfLoopBlock = 0;
for (unsigned i = 0, e = NewSI->getNumSuccessors(); i != e; ++i)
if (NewSI->getSuccessor(i) == BB) {
if (InfLoopBlock == 0) {
// Insert it at the end of the function, because it's either code,
// or it won't matter if it's hot. :)
InfLoopBlock = BasicBlock::Create(BB->getContext(),
"infloop", BB->getParent());
BranchInst::Create(InfLoopBlock, InfLoopBlock);
}
NewSI->setSuccessor(i, InfLoopBlock);
}
Changed = true;
}
}
return Changed;
}
// isSafeToHoistInvoke - If we would need to insert a select that uses the
// value of this invoke (comments in HoistThenElseCodeToIf explain why we
// would need to do this), we can't hoist the invoke, as there is nowhere
// to put the select in this case.
static bool isSafeToHoistInvoke(BasicBlock *BB1, BasicBlock *BB2,
Instruction *I1, Instruction *I2) {
for (succ_iterator SI = succ_begin(BB1), E = succ_end(BB1); SI != E; ++SI) {
PHINode *PN;
for (BasicBlock::iterator BBI = SI->begin();
(PN = dyn_cast<PHINode>(BBI)); ++BBI) {
Value *BB1V = PN->getIncomingValueForBlock(BB1);
Value *BB2V = PN->getIncomingValueForBlock(BB2);
if (BB1V != BB2V && (BB1V==I1 || BB2V==I2)) {
return false;
}
}
}
return true;
}
/// HoistThenElseCodeToIf - Given a conditional branch that goes to BB1 and
/// BB2, hoist any common code in the two blocks up into the branch block. The
/// caller of this function guarantees that BI's block dominates BB1 and BB2.
static bool HoistThenElseCodeToIf(BranchInst *BI) {
// This does very trivial matching, with limited scanning, to find identical
// instructions in the two blocks. In particular, we don't want to get into
// O(M*N) situations here where M and N are the sizes of BB1 and BB2. As
// such, we currently just scan for obviously identical instructions in an
// identical order.
BasicBlock *BB1 = BI->getSuccessor(0); // The true destination.
BasicBlock *BB2 = BI->getSuccessor(1); // The false destination
BasicBlock::iterator BB1_Itr = BB1->begin();
BasicBlock::iterator BB2_Itr = BB2->begin();
Instruction *I1 = BB1_Itr++, *I2 = BB2_Itr++;
// Skip debug info if it is not identical.
DbgInfoIntrinsic *DBI1 = dyn_cast<DbgInfoIntrinsic>(I1);
DbgInfoIntrinsic *DBI2 = dyn_cast<DbgInfoIntrinsic>(I2);
if (!DBI1 || !DBI2 || !DBI1->isIdenticalToWhenDefined(DBI2)) {
while (isa<DbgInfoIntrinsic>(I1))
I1 = BB1_Itr++;
while (isa<DbgInfoIntrinsic>(I2))
I2 = BB2_Itr++;
}
if (isa<PHINode>(I1) || !I1->isIdenticalToWhenDefined(I2) ||
(isa<InvokeInst>(I1) && !isSafeToHoistInvoke(BB1, BB2, I1, I2)))
return false;
// If we get here, we can hoist at least one instruction.
BasicBlock *BIParent = BI->getParent();
do {
// If we are hoisting the terminator instruction, don't move one (making a
// broken BB), instead clone it, and remove BI.
if (isa<TerminatorInst>(I1))
goto HoistTerminator;
// For a normal instruction, we just move one to right before the branch,
// then replace all uses of the other with the first. Finally, we remove
// the now redundant second instruction.
BIParent->getInstList().splice(BI, BB1->getInstList(), I1);
if (!I2->use_empty())
I2->replaceAllUsesWith(I1);
I1->intersectOptionalDataWith(I2);
I2->eraseFromParent();
I1 = BB1_Itr++;
I2 = BB2_Itr++;
// Skip debug info if it is not identical.
DbgInfoIntrinsic *DBI1 = dyn_cast<DbgInfoIntrinsic>(I1);
DbgInfoIntrinsic *DBI2 = dyn_cast<DbgInfoIntrinsic>(I2);
if (!DBI1 || !DBI2 || !DBI1->isIdenticalToWhenDefined(DBI2)) {
while (isa<DbgInfoIntrinsic>(I1))
I1 = BB1_Itr++;
while (isa<DbgInfoIntrinsic>(I2))
I2 = BB2_Itr++;
}
} while (I1->isIdenticalToWhenDefined(I2));
return true;
HoistTerminator:
// It may not be possible to hoist an invoke.
if (isa<InvokeInst>(I1) && !isSafeToHoistInvoke(BB1, BB2, I1, I2))
return true;
// Okay, it is safe to hoist the terminator.
Instruction *NT = I1->clone();
BIParent->getInstList().insert(BI, NT);
if (!NT->getType()->isVoidTy()) {
I1->replaceAllUsesWith(NT);
I2->replaceAllUsesWith(NT);
NT->takeName(I1);
}
IRBuilder<true, NoFolder> Builder(NT);
// Hoisting one of the terminators from our successor is a great thing.
// Unfortunately, the successors of the if/else blocks may have PHI nodes in
// them. If they do, all PHI entries for BB1/BB2 must agree for all PHI
// nodes, so we insert select instruction to compute the final result.
std::map<std::pair<Value*,Value*>, SelectInst*> InsertedSelects;
for (succ_iterator SI = succ_begin(BB1), E = succ_end(BB1); SI != E; ++SI) {
PHINode *PN;
for (BasicBlock::iterator BBI = SI->begin();
(PN = dyn_cast<PHINode>(BBI)); ++BBI) {
Value *BB1V = PN->getIncomingValueForBlock(BB1);
Value *BB2V = PN->getIncomingValueForBlock(BB2);
if (BB1V == BB2V) continue;
// These values do not agree. Insert a select instruction before NT
// that determines the right value.
SelectInst *&SI = InsertedSelects[std::make_pair(BB1V, BB2V)];
if (SI == 0)
SI = cast<SelectInst>
(Builder.CreateSelect(BI->getCondition(), BB1V, BB2V,
BB1V->getName()+"."+BB2V->getName()));
// Make the PHI node use the select for all incoming values for BB1/BB2
for (unsigned i = 0, e = PN->getNumIncomingValues(); i != e; ++i)
if (PN->getIncomingBlock(i) == BB1 || PN->getIncomingBlock(i) == BB2)
PN->setIncomingValue(i, SI);
}
}
// Update any PHI nodes in our new successors.
for (succ_iterator SI = succ_begin(BB1), E = succ_end(BB1); SI != E; ++SI)
AddPredecessorToBlock(*SI, BIParent, BB1);
EraseTerminatorInstAndDCECond(BI);
return true;
}
/// SpeculativelyExecuteBB - Given a conditional branch that goes to BB1
/// and an BB2 and the only successor of BB1 is BB2, hoist simple code
/// (for now, restricted to a single instruction that's side effect free) from
/// the BB1 into the branch block to speculatively execute it.
///
/// Turn
/// BB:
/// %t1 = icmp
/// br i1 %t1, label %BB1, label %BB2
/// BB1:
/// %t3 = add %t2, c
/// br label BB2
/// BB2:
/// =>
/// BB:
/// %t1 = icmp
/// %t4 = add %t2, c
/// %t3 = select i1 %t1, %t2, %t3
static bool SpeculativelyExecuteBB(BranchInst *BI, BasicBlock *BB1) {
// Only speculatively execution a single instruction (not counting the
// terminator) for now.
Instruction *HInst = NULL;
Instruction *Term = BB1->getTerminator();
for (BasicBlock::iterator BBI = BB1->begin(), BBE = BB1->end();
BBI != BBE; ++BBI) {
Instruction *I = BBI;
// Skip debug info.
if (isa<DbgInfoIntrinsic>(I)) continue;
if (I == Term) break;
if (HInst)
return false;
HInst = I;
}
BasicBlock *BIParent = BI->getParent();
// Check the instruction to be hoisted, if there is one.
if (HInst) {
// Don't hoist the instruction if it's unsafe or expensive.
if (!isSafeToSpeculativelyExecute(HInst))
return false;
if (ComputeSpeculationCost(HInst) > PHINodeFoldingThreshold)
return false;
// Do not hoist the instruction if any of its operands are defined but not
// used in this BB. The transformation will prevent the operand from
// being sunk into the use block.
for (User::op_iterator i = HInst->op_begin(), e = HInst->op_end();
i != e; ++i) {
Instruction *OpI = dyn_cast<Instruction>(*i);
if (OpI && OpI->getParent() == BIParent &&
!OpI->mayHaveSideEffects() &&
!OpI->isUsedInBasicBlock(BIParent))
return false;
}
}
// Be conservative for now. FP select instruction can often be expensive.
Value *BrCond = BI->getCondition();
if (isa<FCmpInst>(BrCond))
return false;
// If BB1 is actually on the false edge of the conditional branch, remember
// to swap the select operands later.
bool Invert = false;
if (BB1 != BI->getSuccessor(0)) {
assert(BB1 == BI->getSuccessor(1) && "No edge from 'if' block?");
Invert = true;
}
// Collect interesting PHIs, and scan for hazards.
SmallSetVector<std::pair<Value *, Value *>, 4> PHIs;
BasicBlock *BB2 = BB1->getTerminator()->getSuccessor(0);
for (BasicBlock::iterator I = BB2->begin();
PHINode *PN = dyn_cast<PHINode>(I); ++I) {
Value *BB1V = PN->getIncomingValueForBlock(BB1);
Value *BIParentV = PN->getIncomingValueForBlock(BIParent);
// Skip PHIs which are trivial.
if (BB1V == BIParentV)
continue;
// Check for saftey.
if (ConstantExpr *CE = dyn_cast<ConstantExpr>(BB1V)) {
// An unfolded ConstantExpr could end up getting expanded into
// Instructions. Don't speculate this and another instruction at
// the same time.
if (HInst)
return false;
if (!isSafeToSpeculativelyExecute(CE))
return false;
if (ComputeSpeculationCost(CE) > PHINodeFoldingThreshold)
return false;
}
// Ok, we may insert a select for this PHI.
PHIs.insert(std::make_pair(BB1V, BIParentV));
}
// If there are no PHIs to process, bail early. This helps ensure idempotence
// as well.
if (PHIs.empty())
return false;
// If we get here, we can hoist the instruction and if-convert.
DEBUG(dbgs() << "SPECULATIVELY EXECUTING BB" << *BB1 << "\n";);
// Hoist the instruction.
if (HInst)
BIParent->getInstList().splice(BI, BB1->getInstList(), HInst);
// Insert selects and rewrite the PHI operands.
IRBuilder<true, NoFolder> Builder(BI);
for (unsigned i = 0, e = PHIs.size(); i != e; ++i) {
Value *TrueV = PHIs[i].first;
Value *FalseV = PHIs[i].second;
// Create a select whose true value is the speculatively executed value and
// false value is the previously determined FalseV.
SelectInst *SI;
if (Invert)
SI = cast<SelectInst>
(Builder.CreateSelect(BrCond, FalseV, TrueV,
FalseV->getName() + "." + TrueV->getName()));
else
SI = cast<SelectInst>
(Builder.CreateSelect(BrCond, TrueV, FalseV,
TrueV->getName() + "." + FalseV->getName()));
// Make the PHI node use the select for all incoming values for "then" and
// "if" blocks.
for (BasicBlock::iterator I = BB2->begin();
PHINode *PN = dyn_cast<PHINode>(I); ++I) {
unsigned BB1I = PN->getBasicBlockIndex(BB1);
unsigned BIParentI = PN->getBasicBlockIndex(BIParent);
Value *BB1V = PN->getIncomingValue(BB1I);
Value *BIParentV = PN->getIncomingValue(BIParentI);
if (TrueV == BB1V && FalseV == BIParentV) {
PN->setIncomingValue(BB1I, SI);
PN->setIncomingValue(BIParentI, SI);
}
}
}
++NumSpeculations;
return true;
}
/// BlockIsSimpleEnoughToThreadThrough - Return true if we can thread a branch
/// across this block.
static bool BlockIsSimpleEnoughToThreadThrough(BasicBlock *BB) {
BranchInst *BI = cast<BranchInst>(BB->getTerminator());
unsigned Size = 0;
for (BasicBlock::iterator BBI = BB->begin(); &*BBI != BI; ++BBI) {
if (isa<DbgInfoIntrinsic>(BBI))
continue;
if (Size > 10) return false; // Don't clone large BB's.
++Size;
// We can only support instructions that do not define values that are
// live outside of the current basic block.
for (Value::use_iterator UI = BBI->use_begin(), E = BBI->use_end();
UI != E; ++UI) {
Instruction *U = cast<Instruction>(*UI);
if (U->getParent() != BB || isa<PHINode>(U)) return false;
}
// Looks ok, continue checking.
}
return true;
}
/// FoldCondBranchOnPHI - If we have a conditional branch on a PHI node value
/// that is defined in the same block as the branch and if any PHI entries are
/// constants, thread edges corresponding to that entry to be branches to their
/// ultimate destination.
static bool FoldCondBranchOnPHI(BranchInst *BI, const TargetData *TD) {
BasicBlock *BB = BI->getParent();
PHINode *PN = dyn_cast<PHINode>(BI->getCondition());
// NOTE: we currently cannot transform this case if the PHI node is used
// outside of the block.
if (!PN || PN->getParent() != BB || !PN->hasOneUse())
return false;
// Degenerate case of a single entry PHI.
if (PN->getNumIncomingValues() == 1) {
FoldSingleEntryPHINodes(PN->getParent());
return true;
}
// Now we know that this block has multiple preds and two succs.
if (!BlockIsSimpleEnoughToThreadThrough(BB)) return false;
// Okay, this is a simple enough basic block. See if any phi values are
// constants.
for (unsigned i = 0, e = PN->getNumIncomingValues(); i != e; ++i) {
ConstantInt *CB = dyn_cast<ConstantInt>(PN->getIncomingValue(i));
if (CB == 0 || !CB->getType()->isIntegerTy(1)) continue;
// Okay, we now know that all edges from PredBB should be revectored to
// branch to RealDest.
BasicBlock *PredBB = PN->getIncomingBlock(i);
BasicBlock *RealDest = BI->getSuccessor(!CB->getZExtValue());
if (RealDest == BB) continue; // Skip self loops.
// Skip if the predecessor's terminator is an indirect branch.
if (isa<IndirectBrInst>(PredBB->getTerminator())) continue;
// The dest block might have PHI nodes, other predecessors and other
// difficult cases. Instead of being smart about this, just insert a new
// block that jumps to the destination block, effectively splitting
// the edge we are about to create.
BasicBlock *EdgeBB = BasicBlock::Create(BB->getContext(),
RealDest->getName()+".critedge",
RealDest->getParent(), RealDest);
BranchInst::Create(RealDest, EdgeBB);
// Update PHI nodes.
AddPredecessorToBlock(RealDest, EdgeBB, BB);
// BB may have instructions that are being threaded over. Clone these
// instructions into EdgeBB. We know that there will be no uses of the
// cloned instructions outside of EdgeBB.
BasicBlock::iterator InsertPt = EdgeBB->begin();
DenseMap<Value*, Value*> TranslateMap; // Track translated values.
for (BasicBlock::iterator BBI = BB->begin(); &*BBI != BI; ++BBI) {
if (PHINode *PN = dyn_cast<PHINode>(BBI)) {
TranslateMap[PN] = PN->getIncomingValueForBlock(PredBB);
continue;
}
// Clone the instruction.
Instruction *N = BBI->clone();
if (BBI->hasName()) N->setName(BBI->getName()+".c");
// Update operands due to translation.
for (User::op_iterator i = N->op_begin(), e = N->op_end();
i != e; ++i) {
DenseMap<Value*, Value*>::iterator PI = TranslateMap.find(*i);
if (PI != TranslateMap.end())
*i = PI->second;
}
// Check for trivial simplification.
if (Value *V = SimplifyInstruction(N, TD)) {
TranslateMap[BBI] = V;
delete N; // Instruction folded away, don't need actual inst
} else {
// Insert the new instruction into its new home.
EdgeBB->getInstList().insert(InsertPt, N);
if (!BBI->use_empty())
TranslateMap[BBI] = N;
}
}
// Loop over all of the edges from PredBB to BB, changing them to branch
// to EdgeBB instead.
TerminatorInst *PredBBTI = PredBB->getTerminator();
for (unsigned i = 0, e = PredBBTI->getNumSuccessors(); i != e; ++i)
if (PredBBTI->getSuccessor(i) == BB) {
BB->removePredecessor(PredBB);
PredBBTI->setSuccessor(i, EdgeBB);
}
// Recurse, simplifying any other constants.
return FoldCondBranchOnPHI(BI, TD) | true;
}
return false;
}
/// FoldTwoEntryPHINode - Given a BB that starts with the specified two-entry
/// PHI node, see if we can eliminate it.
static bool FoldTwoEntryPHINode(PHINode *PN, const TargetData *TD) {
// Ok, this is a two entry PHI node. Check to see if this is a simple "if
// statement", which has a very simple dominance structure. Basically, we
// are trying to find the condition that is being branched on, which
// subsequently causes this merge to happen. We really want control
// dependence information for this check, but simplifycfg can't keep it up
// to date, and this catches most of the cases we care about anyway.
BasicBlock *BB = PN->getParent();
BasicBlock *IfTrue, *IfFalse;
Value *IfCond = GetIfCondition(BB, IfTrue, IfFalse);
if (!IfCond ||
// Don't bother if the branch will be constant folded trivially.
isa<ConstantInt>(IfCond))
return false;
// Okay, we found that we can merge this two-entry phi node into a select.
// Doing so would require us to fold *all* two entry phi nodes in this block.
// At some point this becomes non-profitable (particularly if the target
// doesn't support cmov's). Only do this transformation if there are two or
// fewer PHI nodes in this block.
unsigned NumPhis = 0;
for (BasicBlock::iterator I = BB->begin(); isa<PHINode>(I); ++NumPhis, ++I)
if (NumPhis > 2)
return false;
// Loop over the PHI's seeing if we can promote them all to select
// instructions. While we are at it, keep track of the instructions
// that need to be moved to the dominating block.
SmallPtrSet<Instruction*, 4> AggressiveInsts;
unsigned MaxCostVal0 = PHINodeFoldingThreshold,
MaxCostVal1 = PHINodeFoldingThreshold;
for (BasicBlock::iterator II = BB->begin(); isa<PHINode>(II);) {
PHINode *PN = cast<PHINode>(II++);
if (Value *V = SimplifyInstruction(PN, TD)) {
PN->replaceAllUsesWith(V);
PN->eraseFromParent();
continue;
}
if (!DominatesMergePoint(PN->getIncomingValue(0), BB, &AggressiveInsts,
MaxCostVal0) ||
!DominatesMergePoint(PN->getIncomingValue(1), BB, &AggressiveInsts,
MaxCostVal1))
return false;
}
// If we folded the the first phi, PN dangles at this point. Refresh it. If
// we ran out of PHIs then we simplified them all.
PN = dyn_cast<PHINode>(BB->begin());
if (PN == 0) return true;
// Don't fold i1 branches on PHIs which contain binary operators. These can
// often be turned into switches and other things.
if (PN->getType()->isIntegerTy(1) &&
(isa<BinaryOperator>(PN->getIncomingValue(0)) ||
isa<BinaryOperator>(PN->getIncomingValue(1)) ||
isa<BinaryOperator>(IfCond)))
return false;
// If we all PHI nodes are promotable, check to make sure that all
// instructions in the predecessor blocks can be promoted as well. If
// not, we won't be able to get rid of the control flow, so it's not
// worth promoting to select instructions.
BasicBlock *DomBlock = 0;
BasicBlock *IfBlock1 = PN->getIncomingBlock(0);
BasicBlock *IfBlock2 = PN->getIncomingBlock(1);
if (cast<BranchInst>(IfBlock1->getTerminator())->isConditional()) {
IfBlock1 = 0;
} else {
DomBlock = *pred_begin(IfBlock1);
for (BasicBlock::iterator I = IfBlock1->begin();!isa<TerminatorInst>(I);++I)
if (!AggressiveInsts.count(I) && !isa<DbgInfoIntrinsic>(I)) {
// This is not an aggressive instruction that we can promote.
// Because of this, we won't be able to get rid of the control
// flow, so the xform is not worth it.
return false;
}
}
if (cast<BranchInst>(IfBlock2->getTerminator())->isConditional()) {
IfBlock2 = 0;
} else {
DomBlock = *pred_begin(IfBlock2);
for (BasicBlock::iterator I = IfBlock2->begin();!isa<TerminatorInst>(I);++I)
if (!AggressiveInsts.count(I) && !isa<DbgInfoIntrinsic>(I)) {
// This is not an aggressive instruction that we can promote.
// Because of this, we won't be able to get rid of the control
// flow, so the xform is not worth it.
return false;
}
}
DEBUG(dbgs() << "FOUND IF CONDITION! " << *IfCond << " T: "
<< IfTrue->getName() << " F: " << IfFalse->getName() << "\n");
// If we can still promote the PHI nodes after this gauntlet of tests,
// do all of the PHI's now.
Instruction *InsertPt = DomBlock->getTerminator();
IRBuilder<true, NoFolder> Builder(InsertPt);
// Move all 'aggressive' instructions, which are defined in the
// conditional parts of the if's up to the dominating block.
if (IfBlock1)
DomBlock->getInstList().splice(InsertPt,
IfBlock1->getInstList(), IfBlock1->begin(),
IfBlock1->getTerminator());
if (IfBlock2)
DomBlock->getInstList().splice(InsertPt,
IfBlock2->getInstList(), IfBlock2->begin(),
IfBlock2->getTerminator());
while (PHINode *PN = dyn_cast<PHINode>(BB->begin())) {
// Change the PHI node into a select instruction.
Value *TrueVal = PN->getIncomingValue(PN->getIncomingBlock(0) == IfFalse);
Value *FalseVal = PN->getIncomingValue(PN->getIncomingBlock(0) == IfTrue);
SelectInst *NV =
cast<SelectInst>(Builder.CreateSelect(IfCond, TrueVal, FalseVal, ""));
PN->replaceAllUsesWith(NV);
NV->takeName(PN);
PN->eraseFromParent();
}
// At this point, IfBlock1 and IfBlock2 are both empty, so our if statement
// has been flattened. Change DomBlock to jump directly to our new block to
// avoid other simplifycfg's kicking in on the diamond.
TerminatorInst *OldTI = DomBlock->getTerminator();
Builder.SetInsertPoint(OldTI);
Builder.CreateBr(BB);
OldTI->eraseFromParent();
return true;
}
/// SimplifyCondBranchToTwoReturns - If we found a conditional branch that goes
/// to two returning blocks, try to merge them together into one return,
/// introducing a select if the return values disagree.
static bool SimplifyCondBranchToTwoReturns(BranchInst *BI,
IRBuilder<> &Builder) {
assert(BI->isConditional() && "Must be a conditional branch");
BasicBlock *TrueSucc = BI->getSuccessor(0);
BasicBlock *FalseSucc = BI->getSuccessor(1);
ReturnInst *TrueRet = cast<ReturnInst>(TrueSucc->getTerminator());
ReturnInst *FalseRet = cast<ReturnInst>(FalseSucc->getTerminator());
// Check to ensure both blocks are empty (just a return) or optionally empty
// with PHI nodes. If there are other instructions, merging would cause extra
// computation on one path or the other.
if (!TrueSucc->getFirstNonPHIOrDbg()->isTerminator())
return false;
if (!FalseSucc->getFirstNonPHIOrDbg()->isTerminator())
return false;
Builder.SetInsertPoint(BI);
// Okay, we found a branch that is going to two return nodes. If
// there is no return value for this function, just change the
// branch into a return.
if (FalseRet->getNumOperands() == 0) {
TrueSucc->removePredecessor(BI->getParent());
FalseSucc->removePredecessor(BI->getParent());
Builder.CreateRetVoid();
EraseTerminatorInstAndDCECond(BI);
return true;
}
// Otherwise, figure out what the true and false return values are
// so we can insert a new select instruction.
Value *TrueValue = TrueRet->getReturnValue();
Value *FalseValue = FalseRet->getReturnValue();
// Unwrap any PHI nodes in the return blocks.
if (PHINode *TVPN = dyn_cast_or_null<PHINode>(TrueValue))
if (TVPN->getParent() == TrueSucc)
TrueValue = TVPN->getIncomingValueForBlock(BI->getParent());
if (PHINode *FVPN = dyn_cast_or_null<PHINode>(FalseValue))
if (FVPN->getParent() == FalseSucc)
FalseValue = FVPN->getIncomingValueForBlock(BI->getParent());
// In order for this transformation to be safe, we must be able to
// unconditionally execute both operands to the return. This is
// normally the case, but we could have a potentially-trapping
// constant expression that prevents this transformation from being
// safe.
if (ConstantExpr *TCV = dyn_cast_or_null<ConstantExpr>(TrueValue))
if (TCV->canTrap())
return false;
if (ConstantExpr *FCV = dyn_cast_or_null<ConstantExpr>(FalseValue))
if (FCV->canTrap())
return false;
// Okay, we collected all the mapped values and checked them for sanity, and
// defined to really do this transformation. First, update the CFG.
TrueSucc->removePredecessor(BI->getParent());
FalseSucc->removePredecessor(BI->getParent());
// Insert select instructions where needed.
Value *BrCond = BI->getCondition();
if (TrueValue) {
// Insert a select if the results differ.
if (TrueValue == FalseValue || isa<UndefValue>(FalseValue)) {
} else if (isa<UndefValue>(TrueValue)) {
TrueValue = FalseValue;
} else {
TrueValue = Builder.CreateSelect(BrCond, TrueValue,
FalseValue, "retval");
}
}
Value *RI = !TrueValue ?
Builder.CreateRetVoid() : Builder.CreateRet(TrueValue);
(void) RI;
DEBUG(dbgs() << "\nCHANGING BRANCH TO TWO RETURNS INTO SELECT:"
<< "\n " << *BI << "NewRet = " << *RI
<< "TRUEBLOCK: " << *TrueSucc << "FALSEBLOCK: "<< *FalseSucc);
EraseTerminatorInstAndDCECond(BI);
return true;
}
/// ExtractBranchMetadata - Given a conditional BranchInstruction, retrieve the
/// probabilities of the branch taking each edge. Fills in the two APInt
/// parameters and return true, or returns false if no or invalid metadata was
/// found.
static bool ExtractBranchMetadata(BranchInst *BI,
APInt &ProbTrue, APInt &ProbFalse) {
assert(BI->isConditional() &&
"Looking for probabilities on unconditional branch?");
MDNode *ProfileData = BI->getMetadata(LLVMContext::MD_prof);
if (!ProfileData || ProfileData->getNumOperands() != 3) return false;
ConstantInt *CITrue = dyn_cast<ConstantInt>(ProfileData->getOperand(1));
ConstantInt *CIFalse = dyn_cast<ConstantInt>(ProfileData->getOperand(2));
if (!CITrue || !CIFalse) return false;
ProbTrue = CITrue->getValue();
ProbFalse = CIFalse->getValue();
assert(ProbTrue.getBitWidth() == 32 && ProbFalse.getBitWidth() == 32 &&
"Branch probability metadata must be 32-bit integers");
return true;
}
/// MultiplyAndLosePrecision - Multiplies A and B, then returns the result. In
/// the event of overflow, logically-shifts all four inputs right until the
/// multiply fits.
static APInt MultiplyAndLosePrecision(APInt &A, APInt &B, APInt &C, APInt &D,
unsigned &BitsLost) {
BitsLost = 0;
bool Overflow = false;
APInt Result = A.umul_ov(B, Overflow);
if (Overflow) {
APInt MaxB = APInt::getMaxValue(A.getBitWidth()).udiv(A);
do {
B = B.lshr(1);
++BitsLost;
} while (B.ugt(MaxB));
A = A.lshr(BitsLost);
C = C.lshr(BitsLost);
D = D.lshr(BitsLost);
Result = A * B;
}
return Result;
}
/// FoldBranchToCommonDest - If this basic block is simple enough, and if a
/// predecessor branches to us and one of our successors, fold the block into
/// the predecessor and use logical operations to pick the right destination.
bool llvm::FoldBranchToCommonDest(BranchInst *BI) {
BasicBlock *BB = BI->getParent();
Instruction *Cond = dyn_cast<Instruction>(BI->getCondition());
if (Cond == 0 || (!isa<CmpInst>(Cond) && !isa<BinaryOperator>(Cond)) ||
Cond->getParent() != BB || !Cond->hasOneUse())
return false;
// Only allow this if the condition is a simple instruction that can be
// executed unconditionally. It must be in the same block as the branch, and
// must be at the front of the block.
BasicBlock::iterator FrontIt = BB->front();
// Ignore dbg intrinsics.
while (isa<DbgInfoIntrinsic>(FrontIt)) ++FrontIt;
// Allow a single instruction to be hoisted in addition to the compare
// that feeds the branch. We later ensure that any values that _it_ uses
// were also live in the predecessor, so that we don't unnecessarily create
// register pressure or inhibit out-of-order execution.
Instruction *BonusInst = 0;
if (&*FrontIt != Cond &&
FrontIt->hasOneUse() && *FrontIt->use_begin() == Cond &&
isSafeToSpeculativelyExecute(FrontIt)) {
BonusInst = &*FrontIt;
++FrontIt;
// Ignore dbg intrinsics.
while (isa<DbgInfoIntrinsic>(FrontIt)) ++FrontIt;
}
// Only a single bonus inst is allowed.
if (&*FrontIt != Cond)
return false;
// Make sure the instruction after the condition is the cond branch.
BasicBlock::iterator CondIt = Cond; ++CondIt;
// Ingore dbg intrinsics.
while (isa<DbgInfoIntrinsic>(CondIt)) ++CondIt;
if (&*CondIt != BI)
return false;
// Cond is known to be a compare or binary operator. Check to make sure that
// neither operand is a potentially-trapping constant expression.
if (ConstantExpr *CE = dyn_cast<ConstantExpr>(Cond->getOperand(0)))
if (CE->canTrap())
return false;
if (ConstantExpr *CE = dyn_cast<ConstantExpr>(Cond->getOperand(1)))
if (CE->canTrap())
return false;
// Finally, don't infinitely unroll conditional loops.
BasicBlock *TrueDest = BI->getSuccessor(0);
BasicBlock *FalseDest = BI->getSuccessor(1);
if (TrueDest == BB || FalseDest == BB)
return false;
for (pred_iterator PI = pred_begin(BB), E = pred_end(BB); PI != E; ++PI) {
BasicBlock *PredBlock = *PI;
BranchInst *PBI = dyn_cast<BranchInst>(PredBlock->getTerminator());
// Check that we have two conditional branches. If there is a PHI node in
// the common successor, verify that the same value flows in from both
// blocks.
if (PBI == 0 || PBI->isUnconditional() || !SafeToMergeTerminators(BI, PBI))
continue;
// Determine if the two branches share a common destination.
Instruction::BinaryOps Opc;
bool InvertPredCond = false;
if (PBI->getSuccessor(0) == TrueDest)
Opc = Instruction::Or;
else if (PBI->getSuccessor(1) == FalseDest)
Opc = Instruction::And;
else if (PBI->getSuccessor(0) == FalseDest)
Opc = Instruction::And, InvertPredCond = true;
else if (PBI->getSuccessor(1) == TrueDest)
Opc = Instruction::Or, InvertPredCond = true;
else
continue;
// Ensure that any values used in the bonus instruction are also used
// by the terminator of the predecessor. This means that those values
// must already have been resolved, so we won't be inhibiting the
// out-of-order core by speculating them earlier.
if (BonusInst) {
// Collect the values used by the bonus inst
SmallPtrSet<Value*, 4> UsedValues;
for (Instruction::op_iterator OI = BonusInst->op_begin(),
OE = BonusInst->op_end(); OI != OE; ++OI) {
Value *V = *OI;
if (!isa<Constant>(V))
UsedValues.insert(V);
}
SmallVector<std::pair<Value*, unsigned>, 4> Worklist;
Worklist.push_back(std::make_pair(PBI->getOperand(0), 0));
// Walk up to four levels back up the use-def chain of the predecessor's
// terminator to see if all those values were used. The choice of four
// levels is arbitrary, to provide a compile-time-cost bound.
while (!Worklist.empty()) {
std::pair<Value*, unsigned> Pair = Worklist.back();
Worklist.pop_back();
if (Pair.second >= 4) continue;
UsedValues.erase(Pair.first);
if (UsedValues.empty()) break;
if (Instruction *I = dyn_cast<Instruction>(Pair.first)) {
for (Instruction::op_iterator OI = I->op_begin(), OE = I->op_end();
OI != OE; ++OI)
Worklist.push_back(std::make_pair(OI->get(), Pair.second+1));
}
}
if (!UsedValues.empty()) return false;
}
DEBUG(dbgs() << "FOLDING BRANCH TO COMMON DEST:\n" << *PBI << *BB);
IRBuilder<> Builder(PBI);
// If we need to invert the condition in the pred block to match, do so now.
if (InvertPredCond) {
Value *NewCond = PBI->getCondition();
if (NewCond->hasOneUse() && isa<CmpInst>(NewCond)) {
CmpInst *CI = cast<CmpInst>(NewCond);
CI->setPredicate(CI->getInversePredicate());
} else {
NewCond = Builder.CreateNot(NewCond,
PBI->getCondition()->getName()+".not");
}
PBI->setCondition(NewCond);
PBI->swapSuccessors();
}
// If we have a bonus inst, clone it into the predecessor block.
Instruction *NewBonus = 0;
if (BonusInst) {
NewBonus = BonusInst->clone();
PredBlock->getInstList().insert(PBI, NewBonus);
NewBonus->takeName(BonusInst);
BonusInst->setName(BonusInst->getName()+".old");
}
// Clone Cond into the predecessor basic block, and or/and the
// two conditions together.
Instruction *New = Cond->clone();
if (BonusInst) New->replaceUsesOfWith(BonusInst, NewBonus);
PredBlock->getInstList().insert(PBI, New);
New->takeName(Cond);
Cond->setName(New->getName()+".old");
Instruction *NewCond =
cast<Instruction>(Builder.CreateBinOp(Opc, PBI->getCondition(),
New, "or.cond"));
PBI->setCondition(NewCond);
if (PBI->getSuccessor(0) == BB) {
AddPredecessorToBlock(TrueDest, PredBlock, BB);
PBI->setSuccessor(0, TrueDest);
}
if (PBI->getSuccessor(1) == BB) {
AddPredecessorToBlock(FalseDest, PredBlock, BB);
PBI->setSuccessor(1, FalseDest);
}
// TODO: If BB is reachable from all paths through PredBlock, then we
// could replace PBI's branch probabilities with BI's.
// Merge probability data into PredBlock's branch.
APInt A, B, C, D;
if (ExtractBranchMetadata(PBI, C, D) && ExtractBranchMetadata(BI, A, B)) {
// Given IR which does:
// bbA:
// br i1 %x, label %bbB, label %bbC
// bbB:
// br i1 %y, label %bbD, label %bbC
// Let's call the probability that we take the edge from %bbA to %bbB
// 'a', from %bbA to %bbC, 'b', from %bbB to %bbD 'c' and from %bbB to
// %bbC probability 'd'.
//
// We transform the IR into:
// bbA:
// br i1 %z, label %bbD, label %bbC
// where the probability of going to %bbD is (a*c) and going to bbC is
// (b+a*d).
//
// Probabilities aren't stored as ratios directly. Using branch weights,
// we get:
// (a*c)% = A*C, (b+(a*d))% = A*D+B*C+B*D.
// In the event of overflow, we want to drop the LSB of the input
// probabilities.
unsigned BitsLost;
// Ignore overflow result on ProbTrue.
APInt ProbTrue = MultiplyAndLosePrecision(A, C, B, D, BitsLost);
APInt Tmp1 = MultiplyAndLosePrecision(B, D, A, C, BitsLost);
if (BitsLost) {
ProbTrue = ProbTrue.lshr(BitsLost*2);
}
APInt Tmp2 = MultiplyAndLosePrecision(A, D, C, B, BitsLost);
if (BitsLost) {
ProbTrue = ProbTrue.lshr(BitsLost*2);
Tmp1 = Tmp1.lshr(BitsLost*2);
}
APInt Tmp3 = MultiplyAndLosePrecision(B, C, A, D, BitsLost);
if (BitsLost) {
ProbTrue = ProbTrue.lshr(BitsLost*2);
Tmp1 = Tmp1.lshr(BitsLost*2);
Tmp2 = Tmp2.lshr(BitsLost*2);
}
bool Overflow1 = false, Overflow2 = false;
APInt Tmp4 = Tmp2.uadd_ov(Tmp3, Overflow1);
APInt ProbFalse = Tmp4.uadd_ov(Tmp1, Overflow2);
if (Overflow1 || Overflow2) {
ProbTrue = ProbTrue.lshr(1);
Tmp1 = Tmp1.lshr(1);
Tmp2 = Tmp2.lshr(1);
Tmp3 = Tmp3.lshr(1);
Tmp4 = Tmp2 + Tmp3;
ProbFalse = Tmp4 + Tmp1;
}
// The sum of branch weights must fit in 32-bits.
if (ProbTrue.isNegative() && ProbFalse.isNegative()) {
ProbTrue = ProbTrue.lshr(1);
ProbFalse = ProbFalse.lshr(1);
}
if (ProbTrue != ProbFalse) {
// Normalize the result.
APInt GCD = APIntOps::GreatestCommonDivisor(ProbTrue, ProbFalse);
ProbTrue = ProbTrue.udiv(GCD);
ProbFalse = ProbFalse.udiv(GCD);
LLVMContext &Context = BI->getContext();
Value *Ops[3];
Ops[0] = BI->getMetadata(LLVMContext::MD_prof)->getOperand(0);
Ops[1] = ConstantInt::get(Context, ProbTrue);
Ops[2] = ConstantInt::get(Context, ProbFalse);
PBI->setMetadata(LLVMContext::MD_prof, MDNode::get(Context, Ops));
} else {
PBI->setMetadata(LLVMContext::MD_prof, NULL);
}
} else {
PBI->setMetadata(LLVMContext::MD_prof, NULL);
}
// Copy any debug value intrinsics into the end of PredBlock.
for (BasicBlock::iterator I = BB->begin(), E = BB->end(); I != E; ++I)
if (isa<DbgInfoIntrinsic>(*I))
I->clone()->insertBefore(PBI);
return true;
}
return false;
}
/// SimplifyCondBranchToCondBranch - If we have a conditional branch as a
/// predecessor of another block, this function tries to simplify it. We know
/// that PBI and BI are both conditional branches, and BI is in one of the
/// successor blocks of PBI - PBI branches to BI.
static bool SimplifyCondBranchToCondBranch(BranchInst *PBI, BranchInst *BI) {
assert(PBI->isConditional() && BI->isConditional());
BasicBlock *BB = BI->getParent();
// If this block ends with a branch instruction, and if there is a
// predecessor that ends on a branch of the same condition, make
// this conditional branch redundant.
if (PBI->getCondition() == BI->getCondition() &&
PBI->getSuccessor(0) != PBI->getSuccessor(1)) {
// Okay, the outcome of this conditional branch is statically
// knowable. If this block had a single pred, handle specially.
if (BB->getSinglePredecessor()) {
// Turn this into a branch on constant.
bool CondIsTrue = PBI->getSuccessor(0) == BB;
BI->setCondition(ConstantInt::get(Type::getInt1Ty(BB->getContext()),
CondIsTrue));
return true; // Nuke the branch on constant.
}
// Otherwise, if there are multiple predecessors, insert a PHI that merges
// in the constant and simplify the block result. Subsequent passes of
// simplifycfg will thread the block.
if (BlockIsSimpleEnoughToThreadThrough(BB)) {
pred_iterator PB = pred_begin(BB), PE = pred_end(BB);
PHINode *NewPN = PHINode::Create(Type::getInt1Ty(BB->getContext()),
std::distance(PB, PE),
BI->getCondition()->getName() + ".pr",
BB->begin());
// Okay, we're going to insert the PHI node. Since PBI is not the only
// predecessor, compute the PHI'd conditional value for all of the preds.
// Any predecessor where the condition is not computable we keep symbolic.
for (pred_iterator PI = PB; PI != PE; ++PI) {
BasicBlock *P = *PI;
if ((PBI = dyn_cast<BranchInst>(P->getTerminator())) &&
PBI != BI && PBI->isConditional() &&
PBI->getCondition() == BI->getCondition() &&
PBI->getSuccessor(0) != PBI->getSuccessor(1)) {
bool CondIsTrue = PBI->getSuccessor(0) == BB;
NewPN->addIncoming(ConstantInt::get(Type::getInt1Ty(BB->getContext()),
CondIsTrue), P);
} else {
NewPN->addIncoming(BI->getCondition(), P);
}
}
BI->setCondition(NewPN);
return true;
}
}
// If this is a conditional branch in an empty block, and if any
// predecessors is a conditional branch to one of our destinations,
// fold the conditions into logical ops and one cond br.
BasicBlock::iterator BBI = BB->begin();
// Ignore dbg intrinsics.
while (isa<DbgInfoIntrinsic>(BBI))
++BBI;
if (&*BBI != BI)
return false;
if (ConstantExpr *CE = dyn_cast<ConstantExpr>(BI->getCondition()))
if (CE->canTrap())
return false;
int PBIOp, BIOp;
if (PBI->getSuccessor(0) == BI->getSuccessor(0))
PBIOp = BIOp = 0;
else if (PBI->getSuccessor(0) == BI->getSuccessor(1))
PBIOp = 0, BIOp = 1;
else if (PBI->getSuccessor(1) == BI->getSuccessor(0))
PBIOp = 1, BIOp = 0;
else if (PBI->getSuccessor(1) == BI->getSuccessor(1))
PBIOp = BIOp = 1;
else
return false;
// Check to make sure that the other destination of this branch
// isn't BB itself. If so, this is an infinite loop that will
// keep getting unwound.
if (PBI->getSuccessor(PBIOp) == BB)
return false;
// Do not perform this transformation if it would require
// insertion of a large number of select instructions. For targets
// without predication/cmovs, this is a big pessimization.
BasicBlock *CommonDest = PBI->getSuccessor(PBIOp);
unsigned NumPhis = 0;
for (BasicBlock::iterator II = CommonDest->begin();
isa<PHINode>(II); ++II, ++NumPhis)
if (NumPhis > 2) // Disable this xform.
return false;
// Finally, if everything is ok, fold the branches to logical ops.
BasicBlock *OtherDest = BI->getSuccessor(BIOp ^ 1);
DEBUG(dbgs() << "FOLDING BRs:" << *PBI->getParent()
<< "AND: " << *BI->getParent());
// If OtherDest *is* BB, then BB is a basic block with a single conditional
// branch in it, where one edge (OtherDest) goes back to itself but the other
// exits. We don't *know* that the program avoids the infinite loop
// (even though that seems likely). If we do this xform naively, we'll end up
// recursively unpeeling the loop. Since we know that (after the xform is
// done) that the block *is* infinite if reached, we just make it an obviously
// infinite loop with no cond branch.
if (OtherDest == BB) {
// Insert it at the end of the function, because it's either code,
// or it won't matter if it's hot. :)
BasicBlock *InfLoopBlock = BasicBlock::Create(BB->getContext(),
"infloop", BB->getParent());
BranchInst::Create(InfLoopBlock, InfLoopBlock);
OtherDest = InfLoopBlock;
}
DEBUG(dbgs() << *PBI->getParent()->getParent());
// BI may have other predecessors. Because of this, we leave
// it alone, but modify PBI.
// Make sure we get to CommonDest on True&True directions.
Value *PBICond = PBI->getCondition();
IRBuilder<true, NoFolder> Builder(PBI);
if (PBIOp)
PBICond = Builder.CreateNot(PBICond, PBICond->getName()+".not");
Value *BICond = BI->getCondition();
if (BIOp)
BICond = Builder.CreateNot(BICond, BICond->getName()+".not");
// Merge the conditions.
Value *Cond = Builder.CreateOr(PBICond, BICond, "brmerge");
// Modify PBI to branch on the new condition to the new dests.
PBI->setCondition(Cond);
PBI->setSuccessor(0, CommonDest);
PBI->setSuccessor(1, OtherDest);
// OtherDest may have phi nodes. If so, add an entry from PBI's
// block that are identical to the entries for BI's block.
AddPredecessorToBlock(OtherDest, PBI->getParent(), BB);
// We know that the CommonDest already had an edge from PBI to
// it. If it has PHIs though, the PHIs may have different
// entries for BB and PBI's BB. If so, insert a select to make
// them agree.
PHINode *PN;
for (BasicBlock::iterator II = CommonDest->begin();
(PN = dyn_cast<PHINode>(II)); ++II) {
Value *BIV = PN->getIncomingValueForBlock(BB);
unsigned PBBIdx = PN->getBasicBlockIndex(PBI->getParent());
Value *PBIV = PN->getIncomingValue(PBBIdx);
if (BIV != PBIV) {
// Insert a select in PBI to pick the right value.
Value *NV = cast<SelectInst>
(Builder.CreateSelect(PBICond, PBIV, BIV, PBIV->getName()+".mux"));
PN->setIncomingValue(PBBIdx, NV);
}
}
DEBUG(dbgs() << "INTO: " << *PBI->getParent());
DEBUG(dbgs() << *PBI->getParent()->getParent());
// This basic block is probably dead. We know it has at least
// one fewer predecessor.
return true;
}
// SimplifyTerminatorOnSelect - Simplifies a terminator by replacing it with a
// branch to TrueBB if Cond is true or to FalseBB if Cond is false.
// Takes care of updating the successors and removing the old terminator.
// Also makes sure not to introduce new successors by assuming that edges to
// non-successor TrueBBs and FalseBBs aren't reachable.
static bool SimplifyTerminatorOnSelect(TerminatorInst *OldTerm, Value *Cond,
BasicBlock *TrueBB, BasicBlock *FalseBB){
// Remove any superfluous successor edges from the CFG.
// First, figure out which successors to preserve.
// If TrueBB and FalseBB are equal, only try to preserve one copy of that
// successor.
BasicBlock *KeepEdge1 = TrueBB;
BasicBlock *KeepEdge2 = TrueBB != FalseBB ? FalseBB : 0;
// Then remove the rest.
for (unsigned I = 0, E = OldTerm->getNumSuccessors(); I != E; ++I) {
BasicBlock *Succ = OldTerm->getSuccessor(I);
// Make sure only to keep exactly one copy of each edge.
if (Succ == KeepEdge1)
KeepEdge1 = 0;
else if (Succ == KeepEdge2)
KeepEdge2 = 0;
else
Succ->removePredecessor(OldTerm->getParent());
}
IRBuilder<> Builder(OldTerm);
Builder.SetCurrentDebugLocation(OldTerm->getDebugLoc());
// Insert an appropriate new terminator.
if ((KeepEdge1 == 0) && (KeepEdge2 == 0)) {
if (TrueBB == FalseBB)
// We were only looking for one successor, and it was present.
// Create an unconditional branch to it.
Builder.CreateBr(TrueBB);
else
// We found both of the successors we were looking for.
// Create a conditional branch sharing the condition of the select.
Builder.CreateCondBr(Cond, TrueBB, FalseBB);
} else if (KeepEdge1 && (KeepEdge2 || TrueBB == FalseBB)) {
// Neither of the selected blocks were successors, so this
// terminator must be unreachable.
new UnreachableInst(OldTerm->getContext(), OldTerm);
} else {
// One of the selected values was a successor, but the other wasn't.
// Insert an unconditional branch to the one that was found;
// the edge to the one that wasn't must be unreachable.
if (KeepEdge1 == 0)
// Only TrueBB was found.
Builder.CreateBr(TrueBB);
else
// Only FalseBB was found.
Builder.CreateBr(FalseBB);
}
EraseTerminatorInstAndDCECond(OldTerm);
return true;
}
// SimplifySwitchOnSelect - Replaces
// (switch (select cond, X, Y)) on constant X, Y
// with a branch - conditional if X and Y lead to distinct BBs,
// unconditional otherwise.
static bool SimplifySwitchOnSelect(SwitchInst *SI, SelectInst *Select) {
// Check for constant integer values in the select.
ConstantInt *TrueVal = dyn_cast<ConstantInt>(Select->getTrueValue());
ConstantInt *FalseVal = dyn_cast<ConstantInt>(Select->getFalseValue());
if (!TrueVal || !FalseVal)
return false;
// Find the relevant condition and destinations.
Value *Condition = Select->getCondition();
unsigned TrueCase = SI->findCaseValue(TrueVal);
unsigned FalseCase = SI->findCaseValue(FalseVal);
BasicBlock *TrueBB = SI->getSuccessor(SI->resolveSuccessorIndex(TrueCase));
BasicBlock *FalseBB = SI->getSuccessor(SI->resolveSuccessorIndex(FalseCase));
// Perform the actual simplification.
return SimplifyTerminatorOnSelect(SI, Condition, TrueBB, FalseBB);
}
// SimplifyIndirectBrOnSelect - Replaces
// (indirectbr (select cond, blockaddress(@fn, BlockA),
// blockaddress(@fn, BlockB)))
// with
// (br cond, BlockA, BlockB).
static bool SimplifyIndirectBrOnSelect(IndirectBrInst *IBI, SelectInst *SI) {
// Check that both operands of the select are block addresses.
BlockAddress *TBA = dyn_cast<BlockAddress>(SI->getTrueValue());
BlockAddress *FBA = dyn_cast<BlockAddress>(SI->getFalseValue());
if (!TBA || !FBA)
return false;
// Extract the actual blocks.
BasicBlock *TrueBB = TBA->getBasicBlock();
BasicBlock *FalseBB = FBA->getBasicBlock();
// Perform the actual simplification.
return SimplifyTerminatorOnSelect(IBI, SI->getCondition(), TrueBB, FalseBB);
}
/// TryToSimplifyUncondBranchWithICmpInIt - This is called when we find an icmp
/// instruction (a seteq/setne with a constant) as the only instruction in a
/// block that ends with an uncond branch. We are looking for a very specific
/// pattern that occurs when "A == 1 || A == 2 || A == 3" gets simplified. In
/// this case, we merge the first two "or's of icmp" into a switch, but then the
/// default value goes to an uncond block with a seteq in it, we get something
/// like:
///
/// switch i8 %A, label %DEFAULT [ i8 1, label %end i8 2, label %end ]
/// DEFAULT:
/// %tmp = icmp eq i8 %A, 92
/// br label %end
/// end:
/// ... = phi i1 [ true, %entry ], [ %tmp, %DEFAULT ], [ true, %entry ]
///
/// We prefer to split the edge to 'end' so that there is a true/false entry to
/// the PHI, merging the third icmp into the switch.
static bool TryToSimplifyUncondBranchWithICmpInIt(ICmpInst *ICI,
const TargetData *TD,
IRBuilder<> &Builder) {
BasicBlock *BB = ICI->getParent();
// If the block has any PHIs in it or the icmp has multiple uses, it is too
// complex.
if (isa<PHINode>(BB->begin()) || !ICI->hasOneUse()) return false;
Value *V = ICI->getOperand(0);
ConstantInt *Cst = cast<ConstantInt>(ICI->getOperand(1));
// The pattern we're looking for is where our only predecessor is a switch on
// 'V' and this block is the default case for the switch. In this case we can
// fold the compared value into the switch to simplify things.
BasicBlock *Pred = BB->getSinglePredecessor();
if (Pred == 0 || !isa<SwitchInst>(Pred->getTerminator())) return false;
SwitchInst *SI = cast<SwitchInst>(Pred->getTerminator());
if (SI->getCondition() != V)
return false;
// If BB is reachable on a non-default case, then we simply know the value of
// V in this block. Substitute it and constant fold the icmp instruction
// away.
if (SI->getDefaultDest() != BB) {
ConstantInt *VVal = SI->findCaseDest(BB);
assert(VVal && "Should have a unique destination value");
ICI->setOperand(0, VVal);
if (Value *V = SimplifyInstruction(ICI, TD)) {
ICI->replaceAllUsesWith(V);
ICI->eraseFromParent();
}
// BB is now empty, so it is likely to simplify away.
return SimplifyCFG(BB) | true;
}
// Ok, the block is reachable from the default dest. If the constant we're
// comparing exists in one of the other edges, then we can constant fold ICI
// and zap it.
if (SI->findCaseValue(Cst) != SwitchInst::ErrorIndex) {
Value *V;
if (ICI->getPredicate() == ICmpInst::ICMP_EQ)
V = ConstantInt::getFalse(BB->getContext());
else
V = ConstantInt::getTrue(BB->getContext());
ICI->replaceAllUsesWith(V);
ICI->eraseFromParent();
// BB is now empty, so it is likely to simplify away.
return SimplifyCFG(BB) | true;
}
// The use of the icmp has to be in the 'end' block, by the only PHI node in
// the block.
BasicBlock *SuccBlock = BB->getTerminator()->getSuccessor(0);
PHINode *PHIUse = dyn_cast<PHINode>(ICI->use_back());
if (PHIUse == 0 || PHIUse != &SuccBlock->front() ||
isa<PHINode>(++BasicBlock::iterator(PHIUse)))
return false;
// If the icmp is a SETEQ, then the default dest gets false, the new edge gets
// true in the PHI.
Constant *DefaultCst = ConstantInt::getTrue(BB->getContext());
Constant *NewCst = ConstantInt::getFalse(BB->getContext());
if (ICI->getPredicate() == ICmpInst::ICMP_EQ)
std::swap(DefaultCst, NewCst);
// Replace ICI (which is used by the PHI for the default value) with true or
// false depending on if it is EQ or NE.
ICI->replaceAllUsesWith(DefaultCst);
ICI->eraseFromParent();
// Okay, the switch goes to this block on a default value. Add an edge from
// the switch to the merge point on the compared value.
BasicBlock *NewBB = BasicBlock::Create(BB->getContext(), "switch.edge",
BB->getParent(), BB);
SI->addCase(Cst, NewBB);
// NewBB branches to the phi block, add the uncond branch and the phi entry.
Builder.SetInsertPoint(NewBB);
Builder.SetCurrentDebugLocation(SI->getDebugLoc());
Builder.CreateBr(SuccBlock);
PHIUse->addIncoming(NewCst, NewBB);
return true;
}
/// SimplifyBranchOnICmpChain - The specified branch is a conditional branch.
/// Check to see if it is branching on an or/and chain of icmp instructions, and
/// fold it into a switch instruction if so.
static bool SimplifyBranchOnICmpChain(BranchInst *BI, const TargetData *TD,
IRBuilder<> &Builder) {
Instruction *Cond = dyn_cast<Instruction>(BI->getCondition());
if (Cond == 0) return false;
// Change br (X == 0 | X == 1), T, F into a switch instruction.
// If this is a bunch of seteq's or'd together, or if it's a bunch of
// 'setne's and'ed together, collect them.
Value *CompVal = 0;
std::vector<ConstantInt*> Values;
bool TrueWhenEqual = true;
Value *ExtraCase = 0;
unsigned UsedICmps = 0;
if (Cond->getOpcode() == Instruction::Or) {
CompVal = GatherConstantCompares(Cond, Values, ExtraCase, TD, true,
UsedICmps);
} else if (Cond->getOpcode() == Instruction::And) {
CompVal = GatherConstantCompares(Cond, Values, ExtraCase, TD, false,
UsedICmps);
TrueWhenEqual = false;
}
// If we didn't have a multiply compared value, fail.
if (CompVal == 0) return false;
// Avoid turning single icmps into a switch.
if (UsedICmps <= 1)
return false;
// There might be duplicate constants in the list, which the switch
// instruction can't handle, remove them now.
array_pod_sort(Values.begin(), Values.end(), ConstantIntSortPredicate);
Values.erase(std::unique(Values.begin(), Values.end()), Values.end());
// If Extra was used, we require at least two switch values to do the
// transformation. A switch with one value is just an cond branch.
if (ExtraCase && Values.size() < 2) return false;
// Figure out which block is which destination.
BasicBlock *DefaultBB = BI->getSuccessor(1);
BasicBlock *EdgeBB = BI->getSuccessor(0);
if (!TrueWhenEqual) std::swap(DefaultBB, EdgeBB);
BasicBlock *BB = BI->getParent();
DEBUG(dbgs() << "Converting 'icmp' chain with " << Values.size()
<< " cases into SWITCH. BB is:\n" << *BB);
// If there are any extra values that couldn't be folded into the switch
// then we evaluate them with an explicit branch first. Split the block
// right before the condbr to handle it.
if (ExtraCase) {
BasicBlock *NewBB = BB->splitBasicBlock(BI, "switch.early.test");
// Remove the uncond branch added to the old block.
TerminatorInst *OldTI = BB->getTerminator();
Builder.SetInsertPoint(OldTI);
if (TrueWhenEqual)
Builder.CreateCondBr(ExtraCase, EdgeBB, NewBB);
else
Builder.CreateCondBr(ExtraCase, NewBB, EdgeBB);
OldTI->eraseFromParent();
// If there are PHI nodes in EdgeBB, then we need to add a new entry to them
// for the edge we just added.
AddPredecessorToBlock(EdgeBB, BB, NewBB);
DEBUG(dbgs() << " ** 'icmp' chain unhandled condition: " << *ExtraCase
<< "\nEXTRABB = " << *BB);
BB = NewBB;
}
Builder.SetInsertPoint(BI);
// Convert pointer to int before we switch.
if (CompVal->getType()->isPointerTy()) {
assert(TD && "Cannot switch on pointer without TargetData");
CompVal = Builder.CreatePtrToInt(CompVal,
TD->getIntPtrType(CompVal->getContext()),
"magicptr");
}
// Create the new switch instruction now.
SwitchInst *New = Builder.CreateSwitch(CompVal, DefaultBB, Values.size());
// Add all of the 'cases' to the switch instruction.
for (unsigned i = 0, e = Values.size(); i != e; ++i)
New->addCase(Values[i], EdgeBB);
// We added edges from PI to the EdgeBB. As such, if there were any
// PHI nodes in EdgeBB, they need entries to be added corresponding to
// the number of edges added.
for (BasicBlock::iterator BBI = EdgeBB->begin();
isa<PHINode>(BBI); ++BBI) {
PHINode *PN = cast<PHINode>(BBI);
Value *InVal = PN->getIncomingValueForBlock(BB);
for (unsigned i = 0, e = Values.size()-1; i != e; ++i)
PN->addIncoming(InVal, BB);
}
// Erase the old branch instruction.
EraseTerminatorInstAndDCECond(BI);
DEBUG(dbgs() << " ** 'icmp' chain result is:\n" << *BB << '\n');
return true;
}
bool SimplifyCFGOpt::SimplifyResume(ResumeInst *RI, IRBuilder<> &Builder) {
// If this is a trivial landing pad that just continues unwinding the caught
// exception then zap the landing pad, turning its invokes into calls.
BasicBlock *BB = RI->getParent();
LandingPadInst *LPInst = dyn_cast<LandingPadInst>(BB->getFirstNonPHI());
if (RI->getValue() != LPInst)
// Not a landing pad, or the resume is not unwinding the exception that
// caused control to branch here.
return false;
// Check that there are no other instructions except for debug intrinsics.
BasicBlock::iterator I = LPInst, E = RI;
while (++I != E)
if (!isa<DbgInfoIntrinsic>(I))
return false;
// Turn all invokes that unwind here into calls and delete the basic block.
for (pred_iterator PI = pred_begin(BB), PE = pred_end(BB); PI != PE;) {
InvokeInst *II = cast<InvokeInst>((*PI++)->getTerminator());
SmallVector<Value*, 8> Args(II->op_begin(), II->op_end() - 3);
// Insert a call instruction before the invoke.
CallInst *Call = CallInst::Create(II->getCalledValue(), Args, "", II);
Call->takeName(II);
Call->setCallingConv(II->getCallingConv());
Call->setAttributes(II->getAttributes());
Call->setDebugLoc(II->getDebugLoc());
// Anything that used the value produced by the invoke instruction now uses
// the value produced by the call instruction. Note that we do this even
// for void functions and calls with no uses so that the callgraph edge is
// updated.
II->replaceAllUsesWith(Call);
BB->removePredecessor(II->getParent());
// Insert a branch to the normal destination right before the invoke.
BranchInst::Create(II->getNormalDest(), II);
// Finally, delete the invoke instruction!
II->eraseFromParent();
}
// The landingpad is now unreachable. Zap it.
BB->eraseFromParent();
return true;
}
bool SimplifyCFGOpt::SimplifyReturn(ReturnInst *RI, IRBuilder<> &Builder) {
BasicBlock *BB = RI->getParent();
if (!BB->getFirstNonPHIOrDbg()->isTerminator()) return false;
// Find predecessors that end with branches.
SmallVector<BasicBlock*, 8> UncondBranchPreds;
SmallVector<BranchInst*, 8> CondBranchPreds;
for (pred_iterator PI = pred_begin(BB), E = pred_end(BB); PI != E; ++PI) {
BasicBlock *P = *PI;
TerminatorInst *PTI = P->getTerminator();
if (BranchInst *BI = dyn_cast<BranchInst>(PTI)) {
if (BI->isUnconditional())
UncondBranchPreds.push_back(P);
else
CondBranchPreds.push_back(BI);
}
}
// If we found some, do the transformation!
if (!UncondBranchPreds.empty() && DupRet) {
while (!UncondBranchPreds.empty()) {
BasicBlock *Pred = UncondBranchPreds.pop_back_val();
DEBUG(dbgs() << "FOLDING: " << *BB
<< "INTO UNCOND BRANCH PRED: " << *Pred);
(void)FoldReturnIntoUncondBranch(RI, BB, Pred);
}
// If we eliminated all predecessors of the block, delete the block now.
if (pred_begin(BB) == pred_end(BB))
// We know there are no successors, so just nuke the block.
BB->eraseFromParent();
return true;
}
// Check out all of the conditional branches going to this return
// instruction. If any of them just select between returns, change the
// branch itself into a select/return pair.
while (!CondBranchPreds.empty()) {
BranchInst *BI = CondBranchPreds.pop_back_val();
// Check to see if the non-BB successor is also a return block.
if (isa<ReturnInst>(BI->getSuccessor(0)->getTerminator()) &&
isa<ReturnInst>(BI->getSuccessor(1)->getTerminator()) &&
SimplifyCondBranchToTwoReturns(BI, Builder))
return true;
}
return false;
}
bool SimplifyCFGOpt::SimplifyUnreachable(UnreachableInst *UI) {
BasicBlock *BB = UI->getParent();
bool Changed = false;
// If there are any instructions immediately before the unreachable that can
// be removed, do so.
while (UI != BB->begin()) {
BasicBlock::iterator BBI = UI;
--BBI;
// Do not delete instructions that can have side effects which might cause
// the unreachable to not be reachable; specifically, calls and volatile
// operations may have this effect.
if (isa<CallInst>(BBI) && !isa<DbgInfoIntrinsic>(BBI)) break;
if (BBI->mayHaveSideEffects()) {
if (StoreInst *SI = dyn_cast<StoreInst>(BBI)) {
if (SI->isVolatile())
break;
} else if (LoadInst *LI = dyn_cast<LoadInst>(BBI)) {
if (LI->isVolatile())
break;
} else if (AtomicRMWInst *RMWI = dyn_cast<AtomicRMWInst>(BBI)) {
if (RMWI->isVolatile())
break;
} else if (AtomicCmpXchgInst *CXI = dyn_cast<AtomicCmpXchgInst>(BBI)) {
if (CXI->isVolatile())
break;
} else if (!isa<FenceInst>(BBI) && !isa<VAArgInst>(BBI) &&
!isa<LandingPadInst>(BBI)) {
break;
}
// Note that deleting LandingPad's here is in fact okay, although it
// involves a bit of subtle reasoning. If this inst is a LandingPad,
// all the predecessors of this block will be the unwind edges of Invokes,
// and we can therefore guarantee this block will be erased.
}
// Delete this instruction (any uses are guaranteed to be dead)
if (!BBI->use_empty())
BBI->replaceAllUsesWith(UndefValue::get(BBI->getType()));
BBI->eraseFromParent();
Changed = true;
}
// If the unreachable instruction is the first in the block, take a gander
// at all of the predecessors of this instruction, and simplify them.
if (&BB->front() != UI) return Changed;
SmallVector<BasicBlock*, 8> Preds(pred_begin(BB), pred_end(BB));
for (unsigned i = 0, e = Preds.size(); i != e; ++i) {
TerminatorInst *TI = Preds[i]->getTerminator();
IRBuilder<> Builder(TI);
if (BranchInst *BI = dyn_cast<BranchInst>(TI)) {
if (BI->isUnconditional()) {
if (BI->getSuccessor(0) == BB) {
new UnreachableInst(TI->getContext(), TI);
TI->eraseFromParent();
Changed = true;
}
} else {
if (BI->getSuccessor(0) == BB) {
Builder.CreateBr(BI->getSuccessor(1));
EraseTerminatorInstAndDCECond(BI);
} else if (BI->getSuccessor(1) == BB) {
Builder.CreateBr(BI->getSuccessor(0));
EraseTerminatorInstAndDCECond(BI);
Changed = true;
}
}
} else if (SwitchInst *SI = dyn_cast<SwitchInst>(TI)) {
for (unsigned i = 0, e = SI->getNumCases(); i != e; ++i)
if (SI->getCaseSuccessor(i) == BB) {
BB->removePredecessor(SI->getParent());
SI->removeCase(i);
--i; --e;
Changed = true;
}
// If the default value is unreachable, figure out the most popular
// destination and make it the default.
if (SI->getDefaultDest() == BB) {
std::map<BasicBlock*, std::pair<unsigned, unsigned> > Popularity;
for (unsigned i = 0, e = SI->getNumCases(); i != e; ++i) {
std::pair<unsigned, unsigned> &entry =
Popularity[SI->getCaseSuccessor(i)];
if (entry.first == 0) {
entry.first = 1;
entry.second = i;
} else {
entry.first++;
}
}
// Find the most popular block.
unsigned MaxPop = 0;
unsigned MaxIndex = 0;
BasicBlock *MaxBlock = 0;
for (std::map<BasicBlock*, std::pair<unsigned, unsigned> >::iterator
I = Popularity.begin(), E = Popularity.end(); I != E; ++I) {
if (I->second.first > MaxPop ||
(I->second.first == MaxPop && MaxIndex > I->second.second)) {
MaxPop = I->second.first;
MaxIndex = I->second.second;
MaxBlock = I->first;
}
}
if (MaxBlock) {
// Make this the new default, allowing us to delete any explicit
// edges to it.
SI->setDefaultDest(MaxBlock);
Changed = true;
// If MaxBlock has phinodes in it, remove MaxPop-1 entries from
// it.
if (isa<PHINode>(MaxBlock->begin()))
for (unsigned i = 0; i != MaxPop-1; ++i)
MaxBlock->removePredecessor(SI->getParent());
for (unsigned i = 0, e = SI->getNumCases(); i != e; ++i)
if (SI->getCaseSuccessor(i) == MaxBlock) {
SI->removeCase(i);
--i; --e;
}
}
}
} else if (InvokeInst *II = dyn_cast<InvokeInst>(TI)) {
if (II->getUnwindDest() == BB) {
// Convert the invoke to a call instruction. This would be a good
// place to note that the call does not throw though.
BranchInst *BI = Builder.CreateBr(II->getNormalDest());
II->removeFromParent(); // Take out of symbol table
// Insert the call now...
SmallVector<Value*, 8> Args(II->op_begin(), II->op_end()-3);
Builder.SetInsertPoint(BI);
CallInst *CI = Builder.CreateCall(II->getCalledValue(),
Args, II->getName());
CI->setCallingConv(II->getCallingConv());
CI->setAttributes(II->getAttributes());
// If the invoke produced a value, the call does now instead.
II->replaceAllUsesWith(CI);
delete II;
Changed = true;
}
}
}
// If this block is now dead, remove it.
if (pred_begin(BB) == pred_end(BB) &&
BB != &BB->getParent()->getEntryBlock()) {
// We know there are no successors, so just nuke the block.
BB->eraseFromParent();
return true;
}
return Changed;
}
/// TurnSwitchRangeIntoICmp - Turns a switch with that contains only a
/// integer range comparison into a sub, an icmp and a branch.
static bool TurnSwitchRangeIntoICmp(SwitchInst *SI, IRBuilder<> &Builder) {
assert(SI->getNumCases() > 1 && "Degenerate switch?");
// Make sure all cases point to the same destination and gather the values.
SmallVector<ConstantInt *, 16> Cases;
Cases.push_back(SI->getCaseValue(0));
for (unsigned I = 1, E = SI->getNumCases(); I != E; ++I) {
if (SI->getCaseSuccessor(I-1) != SI->getCaseSuccessor(I))
return false;
Cases.push_back(SI->getCaseValue(I));
}
assert(Cases.size() == SI->getNumCases() && "Not all cases gathered");
// Sort the case values, then check if they form a range we can transform.
array_pod_sort(Cases.begin(), Cases.end(), ConstantIntSortPredicate);
for (unsigned I = 1, E = Cases.size(); I != E; ++I) {
if (Cases[I-1]->getValue() != Cases[I]->getValue()+1)
return false;
}
Constant *Offset = ConstantExpr::getNeg(Cases.back());
Constant *NumCases = ConstantInt::get(Offset->getType(), SI->getNumCases());
Value *Sub = SI->getCondition();
if (!Offset->isNullValue())
Sub = Builder.CreateAdd(Sub, Offset, Sub->getName()+".off");
Value *Cmp = Builder.CreateICmpULT(Sub, NumCases, "switch");
Builder.CreateCondBr(Cmp, SI->getCaseSuccessor(0), SI->getDefaultDest());
// Prune obsolete incoming values off the successor's PHI nodes.
for (BasicBlock::iterator BBI = SI->getCaseSuccessor(0)->begin();
isa<PHINode>(BBI); ++BBI) {
for (unsigned I = 0, E = SI->getNumCases()-1; I != E; ++I)
cast<PHINode>(BBI)->removeIncomingValue(SI->getParent());
}
SI->eraseFromParent();
return true;
}
/// EliminateDeadSwitchCases - Compute masked bits for the condition of a switch
/// and use it to remove dead cases.
static bool EliminateDeadSwitchCases(SwitchInst *SI) {
Value *Cond = SI->getCondition();
unsigned Bits = cast<IntegerType>(Cond->getType())->getBitWidth();
APInt KnownZero(Bits, 0), KnownOne(Bits, 0);
ComputeMaskedBits(Cond, APInt::getAllOnesValue(Bits), KnownZero, KnownOne);
// Gather dead cases.
SmallVector<ConstantInt*, 8> DeadCases;
for (unsigned I = 0, E = SI->getNumCases(); I != E; ++I) {
if ((SI->getCaseValue(I)->getValue() & KnownZero) != 0 ||
(SI->getCaseValue(I)->getValue() & KnownOne) != KnownOne) {
DeadCases.push_back(SI->getCaseValue(I));
DEBUG(dbgs() << "SimplifyCFG: switch case '"
<< SI->getCaseValue(I)->getValue() << "' is dead.\n");
}
}
// Remove dead cases from the switch.
for (unsigned I = 0, E = DeadCases.size(); I != E; ++I) {
unsigned Case = SI->findCaseValue(DeadCases[I]);
assert(Case != SwitchInst::ErrorIndex &&
"Case was not found. Probably mistake in DeadCases forming.");
// Prune unused values from PHI nodes.
SI->getCaseSuccessor(Case)->removePredecessor(SI->getParent());
SI->removeCase(Case);
}
return !DeadCases.empty();
}
/// FindPHIForConditionForwarding - If BB would be eligible for simplification
/// by TryToSimplifyUncondBranchFromEmptyBlock (i.e. it is empty and terminated
/// by an unconditional branch), look at the phi node for BB in the successor
/// block and see if the incoming value is equal to CaseValue. If so, return
/// the phi node, and set PhiIndex to BB's index in the phi node.
static PHINode *FindPHIForConditionForwarding(ConstantInt *CaseValue,
BasicBlock *BB,
int *PhiIndex) {
if (BB->getFirstNonPHIOrDbg() != BB->getTerminator())
return NULL; // BB must be empty to be a candidate for simplification.
if (!BB->getSinglePredecessor())
return NULL; // BB must be dominated by the switch.
BranchInst *Branch = dyn_cast<BranchInst>(BB->getTerminator());
if (!Branch || !Branch->isUnconditional())
return NULL; // Terminator must be unconditional branch.
BasicBlock *Succ = Branch->getSuccessor(0);
BasicBlock::iterator I = Succ->begin();
while (PHINode *PHI = dyn_cast<PHINode>(I++)) {
int Idx = PHI->getBasicBlockIndex(BB);
assert(Idx >= 0 && "PHI has no entry for predecessor?");
Value *InValue = PHI->getIncomingValue(Idx);
if (InValue != CaseValue) continue;
*PhiIndex = Idx;
return PHI;
}
return NULL;
}
/// ForwardSwitchConditionToPHI - Try to forward the condition of a switch
/// instruction to a phi node dominated by the switch, if that would mean that
/// some of the destination blocks of the switch can be folded away.
/// Returns true if a change is made.
static bool ForwardSwitchConditionToPHI(SwitchInst *SI) {
typedef DenseMap<PHINode*, SmallVector<int,4> > ForwardingNodesMap;
ForwardingNodesMap ForwardingNodes;
for (unsigned I = 0; I < SI->getNumCases(); ++I) { // 0 is the default case.
ConstantInt *CaseValue = SI->getCaseValue(I);
BasicBlock *CaseDest = SI->getCaseSuccessor(I);
int PhiIndex;
PHINode *PHI = FindPHIForConditionForwarding(CaseValue, CaseDest,
&PhiIndex);
if (!PHI) continue;
ForwardingNodes[PHI].push_back(PhiIndex);
}
bool Changed = false;
for (ForwardingNodesMap::iterator I = ForwardingNodes.begin(),
E = ForwardingNodes.end(); I != E; ++I) {
PHINode *Phi = I->first;
SmallVector<int,4> &Indexes = I->second;
if (Indexes.size() < 2) continue;
for (size_t I = 0, E = Indexes.size(); I != E; ++I)
Phi->setIncomingValue(Indexes[I], SI->getCondition());
Changed = true;
}
return Changed;
}
bool SimplifyCFGOpt::SimplifySwitch(SwitchInst *SI, IRBuilder<> &Builder) {
// If this switch is too complex to want to look at, ignore it.
if (!isValueEqualityComparison(SI))
return false;
BasicBlock *BB = SI->getParent();
// If we only have one predecessor, and if it is a branch on this value,
// see if that predecessor totally determines the outcome of this switch.
if (BasicBlock *OnlyPred = BB->getSinglePredecessor())
if (SimplifyEqualityComparisonWithOnlyPredecessor(SI, OnlyPred, Builder))
return SimplifyCFG(BB) | true;
Value *Cond = SI->getCondition();
if (SelectInst *Select = dyn_cast<SelectInst>(Cond))
if (SimplifySwitchOnSelect(SI, Select))
return SimplifyCFG(BB) | true;
// If the block only contains the switch, see if we can fold the block
// away into any preds.
BasicBlock::iterator BBI = BB->begin();
// Ignore dbg intrinsics.
while (isa<DbgInfoIntrinsic>(BBI))
++BBI;
if (SI == &*BBI)
if (FoldValueComparisonIntoPredecessors(SI, Builder))
return SimplifyCFG(BB) | true;
// Try to transform the switch into an icmp and a branch.
if (TurnSwitchRangeIntoICmp(SI, Builder))
return SimplifyCFG(BB) | true;
// Remove unreachable cases.
if (EliminateDeadSwitchCases(SI))
return SimplifyCFG(BB) | true;
if (ForwardSwitchConditionToPHI(SI))
return SimplifyCFG(BB) | true;
return false;
}
bool SimplifyCFGOpt::SimplifyIndirectBr(IndirectBrInst *IBI) {
BasicBlock *BB = IBI->getParent();
bool Changed = false;
// Eliminate redundant destinations.
SmallPtrSet<Value *, 8> Succs;
for (unsigned i = 0, e = IBI->getNumDestinations(); i != e; ++i) {
BasicBlock *Dest = IBI->getDestination(i);
if (!Dest->hasAddressTaken() || !Succs.insert(Dest)) {
Dest->removePredecessor(BB);
IBI->removeDestination(i);
--i; --e;
Changed = true;
}
}
if (IBI->getNumDestinations() == 0) {
// If the indirectbr has no successors, change it to unreachable.
new UnreachableInst(IBI->getContext(), IBI);
EraseTerminatorInstAndDCECond(IBI);
return true;
}
if (IBI->getNumDestinations() == 1) {
// If the indirectbr has one successor, change it to a direct branch.
BranchInst::Create(IBI->getDestination(0), IBI);
EraseTerminatorInstAndDCECond(IBI);
return true;
}
if (SelectInst *SI = dyn_cast<SelectInst>(IBI->getAddress())) {
if (SimplifyIndirectBrOnSelect(IBI, SI))
return SimplifyCFG(BB) | true;
}
return Changed;
}
bool SimplifyCFGOpt::SimplifyUncondBranch(BranchInst *BI, IRBuilder<> &Builder){
BasicBlock *BB = BI->getParent();
// If the Terminator is the only non-phi instruction, simplify the block.
BasicBlock::iterator I = BB->getFirstNonPHIOrDbgOrLifetime();
if (I->isTerminator() && BB != &BB->getParent()->getEntryBlock() &&
TryToSimplifyUncondBranchFromEmptyBlock(BB))
return true;
// If the only instruction in the block is a seteq/setne comparison
// against a constant, try to simplify the block.
if (ICmpInst *ICI = dyn_cast<ICmpInst>(I))
if (ICI->isEquality() && isa<ConstantInt>(ICI->getOperand(1))) {
for (++I; isa<DbgInfoIntrinsic>(I); ++I)
;
if (I->isTerminator() &&
TryToSimplifyUncondBranchWithICmpInIt(ICI, TD, Builder))
return true;
}
return false;
}
bool SimplifyCFGOpt::SimplifyCondBranch(BranchInst *BI, IRBuilder<> &Builder) {
BasicBlock *BB = BI->getParent();
// Conditional branch
if (isValueEqualityComparison(BI)) {
// If we only have one predecessor, and if it is a branch on this value,
// see if that predecessor totally determines the outcome of this
// switch.
if (BasicBlock *OnlyPred = BB->getSinglePredecessor())
if (SimplifyEqualityComparisonWithOnlyPredecessor(BI, OnlyPred, Builder))
return SimplifyCFG(BB) | true;
// This block must be empty, except for the setcond inst, if it exists.
// Ignore dbg intrinsics.
BasicBlock::iterator I = BB->begin();
// Ignore dbg intrinsics.
while (isa<DbgInfoIntrinsic>(I))
++I;
if (&*I == BI) {
if (FoldValueComparisonIntoPredecessors(BI, Builder))
return SimplifyCFG(BB) | true;
} else if (&*I == cast<Instruction>(BI->getCondition())){
++I;
// Ignore dbg intrinsics.
while (isa<DbgInfoIntrinsic>(I))
++I;
if (&*I == BI && FoldValueComparisonIntoPredecessors(BI, Builder))
return SimplifyCFG(BB) | true;
}
}
// Try to turn "br (X == 0 | X == 1), T, F" into a switch instruction.
if (SimplifyBranchOnICmpChain(BI, TD, Builder))
return true;
// If this basic block is ONLY a compare and a branch, and if a predecessor
// branches to us and one of our successors, fold the comparison into the
// predecessor and use logical operations to pick the right destination.
if (FoldBranchToCommonDest(BI))
return SimplifyCFG(BB) | true;
// We have a conditional branch to two blocks that are only reachable
// from BI. We know that the condbr dominates the two blocks, so see if
// there is any identical code in the "then" and "else" blocks. If so, we
// can hoist it up to the branching block.
if (BI->getSuccessor(0)->getSinglePredecessor() != 0) {
if (BI->getSuccessor(1)->getSinglePredecessor() != 0) {
if (HoistThenElseCodeToIf(BI))
return SimplifyCFG(BB) | true;
} else {
// If Successor #1 has multiple preds, we may be able to conditionally
// execute Successor #0 if it branches to successor #1.
TerminatorInst *Succ0TI = BI->getSuccessor(0)->getTerminator();
if (Succ0TI->getNumSuccessors() == 1 &&
Succ0TI->getSuccessor(0) == BI->getSuccessor(1))
if (SpeculativelyExecuteBB(BI, BI->getSuccessor(0)))
return SimplifyCFG(BB) | true;
}
} else if (BI->getSuccessor(1)->getSinglePredecessor() != 0) {
// If Successor #0 has multiple preds, we may be able to conditionally
// execute Successor #1 if it branches to successor #0.
TerminatorInst *Succ1TI = BI->getSuccessor(1)->getTerminator();
if (Succ1TI->getNumSuccessors() == 1 &&
Succ1TI->getSuccessor(0) == BI->getSuccessor(0))
if (SpeculativelyExecuteBB(BI, BI->getSuccessor(1)))
return SimplifyCFG(BB) | true;
}
// If this is a branch on a phi node in the current block, thread control
// through this block if any PHI node entries are constants.
if (PHINode *PN = dyn_cast<PHINode>(BI->getCondition()))
if (PN->getParent() == BI->getParent())
if (FoldCondBranchOnPHI(BI, TD))
return SimplifyCFG(BB) | true;
// Scan predecessor blocks for conditional branches.
for (pred_iterator PI = pred_begin(BB), E = pred_end(BB); PI != E; ++PI)
if (BranchInst *PBI = dyn_cast<BranchInst>((*PI)->getTerminator()))
if (PBI != BI && PBI->isConditional())
if (SimplifyCondBranchToCondBranch(PBI, BI))
return SimplifyCFG(BB) | true;
return false;
}
/// Check if passing a value to an instruction will cause undefined behavior.
static bool passingValueIsAlwaysUndefined(Value *V, Instruction *I) {
Constant *C = dyn_cast<Constant>(V);
if (!C)
return false;
if (!I->hasOneUse()) // Only look at single-use instructions, for compile time
return false;
if (C->isNullValue()) {
Instruction *Use = I->use_back();
// Now make sure that there are no instructions in between that can alter
// control flow (eg. calls)
for (BasicBlock::iterator i = ++BasicBlock::iterator(I); &*i != Use; ++i)
if (i == I->getParent()->end() || i->mayHaveSideEffects())
return false;
// Look through GEPs. A load from a GEP derived from NULL is still undefined
if (GetElementPtrInst *GEP = dyn_cast<GetElementPtrInst>(Use))
if (GEP->getPointerOperand() == I)
return passingValueIsAlwaysUndefined(V, GEP);
// Look through bitcasts.
if (BitCastInst *BC = dyn_cast<BitCastInst>(Use))
return passingValueIsAlwaysUndefined(V, BC);
// Load from null is undefined.
if (LoadInst *LI = dyn_cast<LoadInst>(Use))
return LI->getPointerAddressSpace() == 0;
// Store to null is undefined.
if (StoreInst *SI = dyn_cast<StoreInst>(Use))
return SI->getPointerAddressSpace() == 0 && SI->getPointerOperand() == I;
}
return false;
}
/// If BB has an incoming value that will always trigger undefined behavior
/// (eg. null pointer dereference), remove the branch leading here.
static bool removeUndefIntroducingPredecessor(BasicBlock *BB) {
for (BasicBlock::iterator i = BB->begin();
PHINode *PHI = dyn_cast<PHINode>(i); ++i)
for (unsigned i = 0, e = PHI->getNumIncomingValues(); i != e; ++i)
if (passingValueIsAlwaysUndefined(PHI->getIncomingValue(i), PHI)) {
TerminatorInst *T = PHI->getIncomingBlock(i)->getTerminator();
IRBuilder<> Builder(T);
if (BranchInst *BI = dyn_cast<BranchInst>(T)) {
BB->removePredecessor(PHI->getIncomingBlock(i));
// Turn uncoditional branches into unreachables and remove the dead
// destination from conditional branches.
if (BI->isUnconditional())
Builder.CreateUnreachable();
else
Builder.CreateBr(BI->getSuccessor(0) == BB ? BI->getSuccessor(1) :
BI->getSuccessor(0));
BI->eraseFromParent();
return true;
}
// TODO: SwitchInst.
}
return false;
}
bool SimplifyCFGOpt::run(BasicBlock *BB) {
bool Changed = false;
assert(BB && BB->getParent() && "Block not embedded in function!");
assert(BB->getTerminator() && "Degenerate basic block encountered!");
// Remove basic blocks that have no predecessors (except the entry block)...
// or that just have themself as a predecessor. These are unreachable.
if ((pred_begin(BB) == pred_end(BB) &&
BB != &BB->getParent()->getEntryBlock()) ||
BB->getSinglePredecessor() == BB) {
DEBUG(dbgs() << "Removing BB: \n" << *BB);
DeleteDeadBlock(BB);
return true;
}
// Check to see if we can constant propagate this terminator instruction
// away...
Changed |= ConstantFoldTerminator(BB, true);
// Check for and eliminate duplicate PHI nodes in this block.
Changed |= EliminateDuplicatePHINodes(BB);
// Check for and remove branches that will always cause undefined behavior.
Changed |= removeUndefIntroducingPredecessor(BB);
// Merge basic blocks into their predecessor if there is only one distinct
// pred, and if there is only one distinct successor of the predecessor, and
// if there are no PHI nodes.
//
if (MergeBlockIntoPredecessor(BB))
return true;
IRBuilder<> Builder(BB);
// If there is a trivial two-entry PHI node in this basic block, and we can
// eliminate it, do so now.
if (PHINode *PN = dyn_cast<PHINode>(BB->begin()))
if (PN->getNumIncomingValues() == 2)
Changed |= FoldTwoEntryPHINode(PN, TD);
Builder.SetInsertPoint(BB->getTerminator());
if (BranchInst *BI = dyn_cast<BranchInst>(BB->getTerminator())) {
if (BI->isUnconditional()) {
if (SimplifyUncondBranch(BI, Builder)) return true;
} else {
if (SimplifyCondBranch(BI, Builder)) return true;
}
} else if (ReturnInst *RI = dyn_cast<ReturnInst>(BB->getTerminator())) {
if (SimplifyReturn(RI, Builder)) return true;
} else if (ResumeInst *RI = dyn_cast<ResumeInst>(BB->getTerminator())) {
if (SimplifyResume(RI, Builder)) return true;
} else if (SwitchInst *SI = dyn_cast<SwitchInst>(BB->getTerminator())) {
if (SimplifySwitch(SI, Builder)) return true;
} else if (UnreachableInst *UI =
dyn_cast<UnreachableInst>(BB->getTerminator())) {
if (SimplifyUnreachable(UI)) return true;
} else if (IndirectBrInst *IBI =
dyn_cast<IndirectBrInst>(BB->getTerminator())) {
if (SimplifyIndirectBr(IBI)) return true;
}
return Changed;
}
/// SimplifyCFG - This function is used to do simplification of a CFG. For
/// example, it adjusts branches to branches to eliminate the extra hop, it
/// eliminates unreachable basic blocks, and does other "peephole" optimization
/// of the CFG. It returns true if a modification was made.
///
bool llvm::SimplifyCFG(BasicBlock *BB, const TargetData *TD) {
return SimplifyCFGOpt(TD).run(BB);
}
|