1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
2189
2190
2191
2192
2193
2194
2195
2196
2197
2198
2199
2200
2201
2202
2203
2204
2205
2206
2207
2208
2209
2210
2211
2212
2213
2214
2215
2216
2217
2218
2219
2220
2221
2222
2223
2224
2225
2226
2227
2228
2229
2230
2231
2232
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
2244
2245
2246
2247
2248
2249
2250
2251
2252
2253
2254
2255
2256
2257
2258
2259
2260
2261
2262
2263
2264
2265
2266
2267
2268
2269
2270
2271
2272
2273
2274
2275
2276
2277
2278
2279
2280
2281
2282
2283
2284
2285
2286
2287
2288
2289
2290
2291
2292
2293
2294
2295
2296
2297
2298
2299
2300
2301
2302
2303
2304
2305
2306
2307
2308
2309
2310
2311
2312
2313
2314
2315
2316
2317
2318
2319
2320
2321
2322
2323
2324
2325
2326
2327
2328
2329
2330
2331
2332
2333
2334
2335
2336
2337
2338
2339
2340
2341
2342
2343
2344
2345
2346
2347
2348
2349
2350
2351
2352
2353
2354
2355
2356
2357
2358
2359
2360
2361
2362
2363
2364
2365
2366
2367
2368
2369
2370
2371
2372
2373
2374
2375
2376
2377
2378
2379
2380
2381
2382
2383
2384
2385
2386
2387
2388
2389
2390
2391
2392
2393
2394
2395
2396
2397
2398
2399
2400
2401
2402
2403
2404
2405
2406
2407
2408
2409
2410
2411
2412
2413
2414
2415
2416
2417
2418
2419
2420
2421
2422
2423
2424
2425
2426
2427
2428
2429
2430
2431
2432
2433
2434
2435
2436
2437
2438
2439
2440
2441
2442
2443
2444
2445
2446
2447
2448
2449
2450
2451
2452
2453
2454
2455
2456
2457
2458
2459
2460
2461
2462
2463
2464
2465
2466
2467
2468
2469
2470
2471
2472
2473
2474
2475
2476
2477
2478
2479
2480
2481
2482
2483
2484
2485
2486
2487
2488
2489
2490
2491
2492
2493
2494
2495
2496
2497
2498
2499
2500
2501
2502
2503
2504
2505
2506
2507
2508
2509
2510
2511
2512
2513
2514
2515
2516
2517
2518
2519
2520
2521
2522
2523
2524
2525
2526
2527
2528
2529
2530
2531
2532
2533
2534
2535
2536
2537
2538
2539
2540
2541
2542
2543
2544
2545
2546
2547
2548
2549
2550
2551
2552
2553
2554
2555
2556
2557
2558
2559
2560
2561
2562
2563
2564
2565
2566
2567
2568
2569
2570
2571
2572
2573
2574
2575
2576
2577
2578
2579
2580
2581
2582
2583
2584
2585
2586
2587
2588
2589
2590
2591
2592
2593
2594
2595
2596
2597
2598
2599
2600
2601
2602
2603
2604
2605
2606
2607
2608
2609
2610
2611
2612
2613
2614
2615
2616
2617
2618
2619
2620
2621
2622
2623
2624
2625
2626
2627
2628
2629
2630
2631
2632
2633
2634
2635
2636
2637
2638
2639
2640
2641
2642
2643
2644
2645
2646
2647
2648
2649
2650
2651
2652
2653
2654
2655
2656
2657
2658
2659
2660
2661
2662
2663
2664
2665
2666
2667
2668
2669
2670
2671
2672
2673
2674
2675
2676
2677
2678
2679
2680
2681
2682
2683
2684
2685
2686
2687
2688
2689
2690
2691
2692
2693
2694
2695
2696
2697
2698
2699
2700
2701
2702
2703
2704
2705
2706
2707
2708
2709
2710
2711
2712
2713
2714
2715
2716
2717
2718
2719
2720
2721
2722
2723
2724
2725
2726
2727
2728
2729
2730
2731
2732
2733
2734
2735
2736
2737
2738
2739
2740
2741
2742
2743
2744
2745
2746
2747
2748
2749
2750
2751
2752
2753
2754
2755
2756
2757
2758
2759
2760
2761
2762
2763
2764
2765
2766
2767
2768
2769
2770
2771
2772
2773
2774
2775
2776
2777
2778
2779
2780
2781
2782
2783
2784
2785
2786
2787
2788
2789
2790
2791
2792
2793
2794
2795
2796
2797
2798
2799
2800
2801
2802
2803
2804
2805
2806
2807
2808
2809
2810
2811
2812
2813
2814
2815
2816
2817
2818
2819
2820
2821
2822
2823
2824
2825
2826
2827
2828
2829
2830
2831
2832
2833
2834
2835
2836
2837
2838
2839
2840
2841
2842
2843
2844
2845
2846
2847
2848
2849
2850
2851
2852
2853
2854
2855
2856
2857
2858
2859
2860
2861
2862
2863
2864
2865
2866
2867
2868
2869
2870
2871
2872
2873
2874
2875
2876
2877
2878
2879
2880
2881
2882
2883
2884
2885
2886
2887
2888
2889
2890
2891
2892
2893
2894
2895
2896
2897
2898
2899
2900
2901
2902
2903
2904
2905
2906
2907
2908
2909
2910
2911
2912
2913
2914
2915
2916
2917
2918
2919
2920
2921
2922
2923
2924
2925
2926
2927
2928
2929
2930
2931
2932
2933
2934
2935
2936
2937
2938
2939
2940
2941
2942
2943
2944
2945
2946
2947
2948
2949
2950
2951
2952
2953
2954
2955
2956
2957
2958
2959
2960
2961
2962
2963
2964
2965
2966
2967
2968
2969
2970
2971
2972
2973
2974
2975
2976
2977
2978
2979
2980
2981
2982
2983
2984
2985
2986
2987
2988
2989
2990
2991
2992
2993
2994
2995
2996
2997
2998
2999
3000
3001
3002
3003
3004
3005
3006
3007
3008
3009
3010
3011
3012
3013
3014
3015
3016
3017
3018
3019
3020
3021
3022
3023
3024
3025
3026
3027
3028
3029
3030
3031
3032
3033
3034
3035
3036
3037
3038
3039
3040
3041
3042
3043
3044
3045
3046
3047
3048
3049
3050
3051
3052
3053
3054
3055
3056
3057
3058
3059
3060
3061
3062
3063
3064
3065
3066
3067
3068
3069
3070
3071
3072
3073
3074
3075
3076
3077
3078
3079
3080
3081
3082
3083
3084
3085
3086
3087
3088
3089
3090
3091
3092
3093
3094
3095
3096
3097
3098
3099
3100
3101
3102
3103
3104
3105
3106
3107
3108
3109
3110
3111
3112
3113
3114
3115
3116
3117
3118
3119
3120
3121
3122
3123
3124
3125
3126
3127
3128
3129
3130
3131
3132
3133
3134
3135
3136
3137
3138
3139
3140
3141
3142
3143
3144
3145
3146
3147
3148
3149
3150
3151
3152
3153
3154
3155
3156
3157
3158
3159
3160
3161
3162
3163
3164
3165
3166
3167
3168
3169
3170
3171
3172
3173
3174
3175
3176
3177
3178
3179
3180
3181
3182
3183
3184
3185
3186
3187
3188
3189
3190
3191
3192
3193
3194
3195
3196
3197
3198
3199
3200
3201
3202
3203
3204
3205
3206
3207
3208
3209
3210
3211
3212
3213
3214
3215
3216
3217
3218
3219
3220
3221
3222
3223
3224
3225
3226
3227
3228
3229
3230
3231
3232
3233
3234
3235
3236
3237
3238
3239
3240
3241
3242
3243
3244
3245
3246
3247
3248
3249
3250
3251
3252
3253
3254
3255
3256
3257
3258
3259
3260
3261
3262
3263
3264
3265
3266
3267
3268
3269
3270
3271
3272
3273
3274
3275
3276
3277
3278
3279
3280
3281
3282
3283
3284
3285
3286
3287
3288
3289
3290
3291
3292
3293
3294
3295
3296
3297
3298
3299
3300
3301
3302
3303
3304
3305
3306
3307
3308
3309
3310
3311
3312
3313
3314
3315
3316
3317
3318
3319
3320
3321
3322
3323
3324
3325
3326
3327
3328
3329
3330
3331
3332
3333
3334
3335
3336
3337
3338
3339
3340
3341
3342
3343
3344
3345
3346
3347
3348
3349
3350
3351
3352
3353
3354
3355
3356
3357
3358
3359
3360
3361
3362
3363
3364
3365
3366
3367
3368
3369
3370
3371
3372
3373
3374
3375
3376
3377
3378
3379
3380
3381
3382
3383
3384
3385
3386
3387
3388
3389
3390
3391
3392
3393
3394
3395
3396
3397
3398
3399
3400
3401
3402
3403
3404
3405
3406
3407
3408
3409
3410
3411
3412
3413
3414
3415
3416
3417
3418
3419
3420
3421
3422
3423
3424
3425
3426
3427
3428
3429
3430
3431
3432
3433
3434
3435
3436
3437
3438
3439
3440
3441
3442
3443
3444
3445
3446
3447
3448
3449
3450
3451
3452
3453
3454
3455
3456
3457
3458
3459
3460
3461
3462
3463
3464
3465
3466
3467
3468
3469
3470
3471
3472
3473
3474
3475
3476
3477
3478
3479
3480
3481
3482
3483
3484
3485
3486
3487
3488
3489
3490
3491
3492
3493
3494
3495
3496
3497
3498
3499
3500
3501
3502
3503
3504
3505
3506
3507
3508
3509
3510
3511
3512
3513
3514
3515
3516
3517
3518
3519
3520
3521
3522
3523
3524
3525
3526
3527
3528
3529
3530
3531
3532
3533
3534
3535
3536
3537
3538
3539
3540
3541
3542
3543
3544
3545
3546
3547
3548
3549
3550
3551
3552
3553
3554
3555
3556
3557
3558
3559
3560
3561
3562
3563
3564
3565
3566
3567
3568
3569
3570
3571
3572
3573
3574
3575
3576
3577
3578
3579
3580
3581
3582
3583
3584
3585
3586
3587
3588
3589
3590
3591
3592
3593
3594
3595
3596
3597
3598
3599
3600
3601
3602
3603
3604
3605
3606
3607
3608
3609
3610
3611
3612
3613
3614
3615
3616
3617
3618
3619
3620
3621
3622
3623
3624
3625
3626
3627
3628
3629
3630
3631
3632
3633
3634
3635
3636
3637
3638
3639
3640
3641
3642
3643
3644
3645
3646
3647
3648
3649
3650
3651
3652
3653
3654
3655
3656
3657
3658
3659
3660
3661
3662
3663
3664
3665
3666
3667
3668
3669
3670
3671
3672
3673
3674
3675
3676
3677
3678
3679
3680
3681
3682
3683
3684
3685
3686
3687
3688
3689
3690
3691
3692
3693
3694
3695
3696
3697
3698
3699
3700
3701
3702
3703
3704
3705
3706
3707
3708
3709
3710
3711
3712
3713
3714
3715
3716
3717
3718
3719
3720
3721
3722
3723
3724
3725
3726
3727
3728
3729
3730
3731
3732
3733
3734
3735
3736
3737
3738
3739
3740
3741
3742
3743
3744
3745
3746
3747
3748
3749
3750
3751
3752
3753
3754
3755
3756
3757
3758
3759
3760
3761
3762
3763
3764
3765
3766
3767
3768
3769
3770
3771
3772
3773
3774
3775
3776
3777
3778
3779
3780
3781
3782
3783
3784
3785
3786
3787
3788
3789
3790
3791
3792
3793
3794
3795
3796
3797
3798
3799
3800
3801
3802
3803
3804
3805
3806
3807
3808
3809
3810
3811
3812
3813
3814
3815
3816
3817
3818
3819
3820
3821
3822
3823
3824
3825
3826
3827
3828
3829
3830
3831
3832
3833
3834
3835
3836
3837
3838
3839
3840
3841
3842
3843
3844
3845
3846
3847
3848
3849
3850
3851
3852
3853
3854
3855
3856
3857
3858
3859
3860
3861
3862
3863
3864
3865
3866
3867
3868
3869
3870
3871
3872
3873
3874
3875
3876
3877
3878
3879
3880
3881
3882
3883
3884
3885
3886
3887
3888
3889
3890
3891
3892
3893
3894
3895
3896
3897
3898
3899
3900
3901
3902
3903
3904
3905
3906
3907
3908
3909
3910
3911
3912
3913
3914
3915
3916
3917
3918
3919
3920
3921
3922
3923
3924
3925
3926
3927
3928
3929
3930
3931
3932
3933
3934
3935
3936
3937
3938
3939
3940
3941
3942
3943
3944
3945
3946
3947
3948
3949
3950
3951
3952
3953
3954
3955
3956
3957
3958
3959
3960
3961
3962
3963
3964
3965
3966
3967
3968
3969
3970
3971
3972
3973
3974
3975
3976
3977
3978
3979
3980
3981
3982
3983
3984
3985
3986
3987
3988
3989
3990
3991
3992
3993
3994
3995
3996
3997
3998
3999
4000
4001
4002
4003
4004
4005
4006
4007
4008
4009
4010
4011
4012
4013
4014
4015
4016
4017
4018
4019
4020
4021
4022
4023
4024
4025
4026
4027
4028
4029
4030
4031
4032
4033
4034
4035
4036
4037
4038
4039
4040
4041
4042
4043
4044
4045
4046
4047
4048
4049
4050
4051
4052
4053
4054
4055
4056
4057
4058
4059
4060
4061
4062
4063
4064
4065
4066
4067
4068
4069
4070
4071
4072
4073
4074
4075
4076
4077
4078
4079
4080
4081
4082
4083
4084
4085
4086
4087
4088
4089
4090
4091
4092
4093
4094
4095
4096
4097
4098
4099
4100
4101
4102
4103
4104
4105
4106
4107
4108
4109
4110
4111
4112
4113
4114
4115
4116
4117
4118
4119
4120
4121
4122
4123
4124
4125
4126
4127
4128
4129
4130
4131
4132
4133
4134
4135
4136
4137
4138
4139
4140
4141
4142
4143
4144
4145
4146
4147
4148
4149
4150
4151
4152
4153
4154
4155
4156
4157
4158
4159
4160
4161
4162
4163
4164
4165
4166
4167
4168
4169
4170
4171
4172
4173
4174
4175
4176
4177
4178
4179
4180
4181
4182
4183
4184
4185
4186
4187
4188
4189
4190
4191
4192
4193
4194
4195
4196
4197
4198
4199
4200
4201
4202
4203
4204
4205
4206
4207
4208
4209
4210
4211
4212
4213
4214
4215
4216
4217
4218
4219
4220
4221
4222
4223
4224
4225
4226
4227
4228
4229
4230
4231
4232
4233
4234
4235
4236
4237
4238
4239
4240
4241
4242
4243
4244
4245
4246
4247
4248
4249
4250
4251
4252
4253
4254
4255
4256
4257
4258
4259
4260
4261
4262
4263
4264
4265
4266
4267
4268
4269
4270
4271
4272
4273
4274
4275
4276
4277
4278
4279
4280
4281
4282
4283
4284
4285
4286
4287
4288
4289
4290
4291
4292
4293
4294
4295
4296
4297
4298
4299
4300
4301
4302
4303
4304
4305
4306
4307
4308
4309
4310
4311
4312
4313
4314
4315
4316
4317
4318
4319
4320
4321
4322
4323
4324
4325
4326
4327
4328
4329
4330
4331
4332
4333
4334
4335
4336
4337
4338
4339
4340
4341
4342
4343
4344
4345
4346
4347
4348
4349
4350
4351
4352
4353
4354
4355
4356
4357
4358
4359
4360
4361
4362
4363
4364
4365
4366
4367
4368
4369
4370
4371
4372
4373
4374
4375
4376
4377
4378
4379
4380
4381
4382
4383
4384
4385
4386
4387
4388
4389
4390
4391
4392
4393
4394
4395
4396
4397
4398
4399
4400
4401
4402
4403
4404
4405
4406
4407
4408
4409
4410
4411
4412
4413
4414
4415
4416
4417
4418
4419
4420
4421
4422
4423
4424
4425
4426
4427
4428
4429
4430
4431
4432
4433
4434
4435
4436
4437
4438
4439
4440
4441
4442
4443
4444
4445
4446
4447
4448
4449
4450
4451
4452
4453
4454
4455
4456
4457
4458
4459
4460
4461
4462
4463
4464
4465
4466
4467
4468
4469
4470
4471
4472
4473
4474
4475
4476
4477
4478
4479
4480
4481
4482
4483
4484
4485
4486
4487
4488
4489
4490
4491
4492
4493
4494
4495
4496
4497
4498
4499
4500
4501
4502
4503
4504
4505
4506
4507
4508
4509
4510
4511
4512
4513
4514
4515
4516
4517
4518
4519
4520
4521
4522
4523
4524
4525
4526
4527
4528
4529
4530
4531
4532
4533
4534
4535
4536
4537
4538
4539
4540
4541
4542
4543
4544
4545
4546
4547
4548
4549
4550
4551
4552
4553
4554
4555
4556
4557
4558
4559
4560
4561
4562
4563
4564
4565
4566
4567
4568
4569
4570
4571
4572
4573
4574
4575
4576
4577
4578
4579
4580
4581
4582
4583
4584
4585
4586
4587
4588
4589
4590
4591
4592
4593
4594
4595
4596
4597
4598
4599
4600
4601
4602
4603
4604
4605
4606
4607
4608
4609
4610
4611
4612
4613
4614
4615
4616
4617
4618
4619
4620
4621
4622
4623
4624
4625
4626
4627
4628
4629
4630
4631
4632
4633
4634
4635
4636
4637
4638
4639
4640
4641
4642
4643
4644
4645
4646
4647
4648
4649
4650
4651
4652
4653
4654
4655
4656
4657
4658
4659
4660
4661
4662
4663
4664
4665
4666
4667
4668
4669
4670
4671
4672
4673
4674
4675
4676
4677
4678
4679
4680
4681
4682
4683
4684
4685
4686
4687
4688
4689
4690
4691
4692
4693
4694
4695
4696
4697
4698
4699
4700
4701
4702
4703
4704
4705
4706
4707
4708
4709
4710
4711
4712
4713
4714
4715
4716
4717
4718
4719
4720
4721
4722
4723
4724
4725
4726
4727
4728
4729
4730
4731
4732
4733
4734
4735
4736
4737
4738
4739
4740
4741
4742
4743
4744
4745
4746
4747
4748
4749
4750
4751
4752
4753
4754
4755
4756
4757
4758
4759
4760
4761
4762
4763
4764
4765
4766
4767
4768
4769
4770
4771
4772
4773
4774
4775
4776
4777
4778
4779
4780
4781
4782
4783
4784
4785
4786
4787
4788
4789
4790
4791
4792
4793
4794
4795
4796
4797
4798
4799
4800
4801
4802
4803
4804
4805
4806
4807
4808
4809
4810
4811
4812
4813
4814
4815
4816
4817
4818
4819
4820
4821
4822
4823
4824
4825
4826
4827
4828
4829
4830
|
//===- InstructionCombining.cpp - Combine multiple instructions -----------===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
//
// InstructionCombining - Combine instructions to form fewer, simple
// instructions. This pass does not modify the CFG. This pass is where
// algebraic simplification happens.
//
// This pass combines things like:
// %Y = add i32 %X, 1
// %Z = add i32 %Y, 1
// into:
// %Z = add i32 %X, 2
//
// This is a simple worklist driven algorithm.
//
// This pass guarantees that the following canonicalizations are performed on
// the program:
// 1. If a binary operator has a constant operand, it is moved to the RHS
// 2. Bitwise operators with constant operands are always grouped so that
// shifts are performed first, then or's, then and's, then xor's.
// 3. Compare instructions are converted from <,>,<=,>= to ==,!= if possible
// 4. All cmp instructions on boolean values are replaced with logical ops
// 5. add X, X is represented as (X*2) => (X << 1)
// 6. Multiplies with a power-of-two constant argument are transformed into
// shifts.
// ... etc.
//
//===----------------------------------------------------------------------===//
#include "InstCombineInternal.h"
#include "llvm/ADT/APInt.h"
#include "llvm/ADT/ArrayRef.h"
#include "llvm/ADT/DenseMap.h"
#include "llvm/ADT/SmallPtrSet.h"
#include "llvm/ADT/SmallVector.h"
#include "llvm/ADT/Statistic.h"
#include "llvm/Analysis/AliasAnalysis.h"
#include "llvm/Analysis/AssumptionCache.h"
#include "llvm/Analysis/BasicAliasAnalysis.h"
#include "llvm/Analysis/BlockFrequencyInfo.h"
#include "llvm/Analysis/CFG.h"
#include "llvm/Analysis/ConstantFolding.h"
#include "llvm/Analysis/GlobalsModRef.h"
#include "llvm/Analysis/InstructionSimplify.h"
#include "llvm/Analysis/LazyBlockFrequencyInfo.h"
#include "llvm/Analysis/LoopInfo.h"
#include "llvm/Analysis/MemoryBuiltins.h"
#include "llvm/Analysis/OptimizationRemarkEmitter.h"
#include "llvm/Analysis/ProfileSummaryInfo.h"
#include "llvm/Analysis/TargetFolder.h"
#include "llvm/Analysis/TargetLibraryInfo.h"
#include "llvm/Analysis/TargetTransformInfo.h"
#include "llvm/Analysis/Utils/Local.h"
#include "llvm/Analysis/ValueTracking.h"
#include "llvm/Analysis/VectorUtils.h"
#include "llvm/IR/BasicBlock.h"
#include "llvm/IR/CFG.h"
#include "llvm/IR/Constant.h"
#include "llvm/IR/Constants.h"
#include "llvm/IR/DIBuilder.h"
#include "llvm/IR/DataLayout.h"
#include "llvm/IR/DebugInfo.h"
#include "llvm/IR/DerivedTypes.h"
#include "llvm/IR/Dominators.h"
#include "llvm/IR/EHPersonalities.h"
#include "llvm/IR/Function.h"
#include "llvm/IR/GetElementPtrTypeIterator.h"
#include "llvm/IR/IRBuilder.h"
#include "llvm/IR/InstrTypes.h"
#include "llvm/IR/Instruction.h"
#include "llvm/IR/Instructions.h"
#include "llvm/IR/IntrinsicInst.h"
#include "llvm/IR/Intrinsics.h"
#include "llvm/IR/Metadata.h"
#include "llvm/IR/Operator.h"
#include "llvm/IR/PassManager.h"
#include "llvm/IR/PatternMatch.h"
#include "llvm/IR/Type.h"
#include "llvm/IR/Use.h"
#include "llvm/IR/User.h"
#include "llvm/IR/Value.h"
#include "llvm/IR/ValueHandle.h"
#include "llvm/InitializePasses.h"
#include "llvm/Support/Casting.h"
#include "llvm/Support/CommandLine.h"
#include "llvm/Support/Compiler.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/DebugCounter.h"
#include "llvm/Support/ErrorHandling.h"
#include "llvm/Support/KnownBits.h"
#include "llvm/Support/raw_ostream.h"
#include "llvm/Transforms/InstCombine/InstCombine.h"
#include "llvm/Transforms/Utils/BasicBlockUtils.h"
#include "llvm/Transforms/Utils/Local.h"
#include <algorithm>
#include <cassert>
#include <cstdint>
#include <memory>
#include <optional>
#include <string>
#include <utility>
#define DEBUG_TYPE "instcombine"
#include "llvm/Transforms/Utils/InstructionWorklist.h"
#include <optional>
using namespace llvm;
using namespace llvm::PatternMatch;
STATISTIC(NumWorklistIterations,
"Number of instruction combining iterations performed");
STATISTIC(NumOneIteration, "Number of functions with one iteration");
STATISTIC(NumTwoIterations, "Number of functions with two iterations");
STATISTIC(NumThreeIterations, "Number of functions with three iterations");
STATISTIC(NumFourOrMoreIterations,
"Number of functions with four or more iterations");
STATISTIC(NumCombined , "Number of insts combined");
STATISTIC(NumConstProp, "Number of constant folds");
STATISTIC(NumDeadInst , "Number of dead inst eliminated");
STATISTIC(NumSunkInst , "Number of instructions sunk");
STATISTIC(NumExpand, "Number of expansions");
STATISTIC(NumFactor , "Number of factorizations");
STATISTIC(NumReassoc , "Number of reassociations");
DEBUG_COUNTER(VisitCounter, "instcombine-visit",
"Controls which instructions are visited");
static cl::opt<bool>
EnableCodeSinking("instcombine-code-sinking", cl::desc("Enable code sinking"),
cl::init(true));
static cl::opt<unsigned> MaxSinkNumUsers(
"instcombine-max-sink-users", cl::init(32),
cl::desc("Maximum number of undroppable users for instruction sinking"));
static cl::opt<unsigned>
MaxArraySize("instcombine-maxarray-size", cl::init(1024),
cl::desc("Maximum array size considered when doing a combine"));
// FIXME: Remove this flag when it is no longer necessary to convert
// llvm.dbg.declare to avoid inaccurate debug info. Setting this to false
// increases variable availability at the cost of accuracy. Variables that
// cannot be promoted by mem2reg or SROA will be described as living in memory
// for their entire lifetime. However, passes like DSE and instcombine can
// delete stores to the alloca, leading to misleading and inaccurate debug
// information. This flag can be removed when those passes are fixed.
static cl::opt<unsigned> ShouldLowerDbgDeclare("instcombine-lower-dbg-declare",
cl::Hidden, cl::init(true));
std::optional<Instruction *>
InstCombiner::targetInstCombineIntrinsic(IntrinsicInst &II) {
// Handle target specific intrinsics
if (II.getCalledFunction()->isTargetIntrinsic()) {
return TTI.instCombineIntrinsic(*this, II);
}
return std::nullopt;
}
std::optional<Value *> InstCombiner::targetSimplifyDemandedUseBitsIntrinsic(
IntrinsicInst &II, APInt DemandedMask, KnownBits &Known,
bool &KnownBitsComputed) {
// Handle target specific intrinsics
if (II.getCalledFunction()->isTargetIntrinsic()) {
return TTI.simplifyDemandedUseBitsIntrinsic(*this, II, DemandedMask, Known,
KnownBitsComputed);
}
return std::nullopt;
}
std::optional<Value *> InstCombiner::targetSimplifyDemandedVectorEltsIntrinsic(
IntrinsicInst &II, APInt DemandedElts, APInt &PoisonElts,
APInt &PoisonElts2, APInt &PoisonElts3,
std::function<void(Instruction *, unsigned, APInt, APInt &)>
SimplifyAndSetOp) {
// Handle target specific intrinsics
if (II.getCalledFunction()->isTargetIntrinsic()) {
return TTI.simplifyDemandedVectorEltsIntrinsic(
*this, II, DemandedElts, PoisonElts, PoisonElts2, PoisonElts3,
SimplifyAndSetOp);
}
return std::nullopt;
}
bool InstCombiner::isValidAddrSpaceCast(unsigned FromAS, unsigned ToAS) const {
return TTI.isValidAddrSpaceCast(FromAS, ToAS);
}
Value *InstCombinerImpl::EmitGEPOffset(User *GEP) {
return llvm::emitGEPOffset(&Builder, DL, GEP);
}
/// Legal integers and common types are considered desirable. This is used to
/// avoid creating instructions with types that may not be supported well by the
/// the backend.
/// NOTE: This treats i8, i16 and i32 specially because they are common
/// types in frontend languages.
bool InstCombinerImpl::isDesirableIntType(unsigned BitWidth) const {
switch (BitWidth) {
case 8:
case 16:
case 32:
return true;
default:
return DL.isLegalInteger(BitWidth);
}
}
/// Return true if it is desirable to convert an integer computation from a
/// given bit width to a new bit width.
/// We don't want to convert from a legal or desirable type (like i8) to an
/// illegal type or from a smaller to a larger illegal type. A width of '1'
/// is always treated as a desirable type because i1 is a fundamental type in
/// IR, and there are many specialized optimizations for i1 types.
/// Common/desirable widths are equally treated as legal to convert to, in
/// order to open up more combining opportunities.
bool InstCombinerImpl::shouldChangeType(unsigned FromWidth,
unsigned ToWidth) const {
bool FromLegal = FromWidth == 1 || DL.isLegalInteger(FromWidth);
bool ToLegal = ToWidth == 1 || DL.isLegalInteger(ToWidth);
// Convert to desirable widths even if they are not legal types.
// Only shrink types, to prevent infinite loops.
if (ToWidth < FromWidth && isDesirableIntType(ToWidth))
return true;
// If this is a legal or desiable integer from type, and the result would be
// an illegal type, don't do the transformation.
if ((FromLegal || isDesirableIntType(FromWidth)) && !ToLegal)
return false;
// Otherwise, if both are illegal, do not increase the size of the result. We
// do allow things like i160 -> i64, but not i64 -> i160.
if (!FromLegal && !ToLegal && ToWidth > FromWidth)
return false;
return true;
}
/// Return true if it is desirable to convert a computation from 'From' to 'To'.
/// We don't want to convert from a legal to an illegal type or from a smaller
/// to a larger illegal type. i1 is always treated as a legal type because it is
/// a fundamental type in IR, and there are many specialized optimizations for
/// i1 types.
bool InstCombinerImpl::shouldChangeType(Type *From, Type *To) const {
// TODO: This could be extended to allow vectors. Datalayout changes might be
// needed to properly support that.
if (!From->isIntegerTy() || !To->isIntegerTy())
return false;
unsigned FromWidth = From->getPrimitiveSizeInBits();
unsigned ToWidth = To->getPrimitiveSizeInBits();
return shouldChangeType(FromWidth, ToWidth);
}
// Return true, if No Signed Wrap should be maintained for I.
// The No Signed Wrap flag can be kept if the operation "B (I.getOpcode) C",
// where both B and C should be ConstantInts, results in a constant that does
// not overflow. This function only handles the Add and Sub opcodes. For
// all other opcodes, the function conservatively returns false.
static bool maintainNoSignedWrap(BinaryOperator &I, Value *B, Value *C) {
auto *OBO = dyn_cast<OverflowingBinaryOperator>(&I);
if (!OBO || !OBO->hasNoSignedWrap())
return false;
// We reason about Add and Sub Only.
Instruction::BinaryOps Opcode = I.getOpcode();
if (Opcode != Instruction::Add && Opcode != Instruction::Sub)
return false;
const APInt *BVal, *CVal;
if (!match(B, m_APInt(BVal)) || !match(C, m_APInt(CVal)))
return false;
bool Overflow = false;
if (Opcode == Instruction::Add)
(void)BVal->sadd_ov(*CVal, Overflow);
else
(void)BVal->ssub_ov(*CVal, Overflow);
return !Overflow;
}
static bool hasNoUnsignedWrap(BinaryOperator &I) {
auto *OBO = dyn_cast<OverflowingBinaryOperator>(&I);
return OBO && OBO->hasNoUnsignedWrap();
}
static bool hasNoSignedWrap(BinaryOperator &I) {
auto *OBO = dyn_cast<OverflowingBinaryOperator>(&I);
return OBO && OBO->hasNoSignedWrap();
}
/// Conservatively clears subclassOptionalData after a reassociation or
/// commutation. We preserve fast-math flags when applicable as they can be
/// preserved.
static void ClearSubclassDataAfterReassociation(BinaryOperator &I) {
FPMathOperator *FPMO = dyn_cast<FPMathOperator>(&I);
if (!FPMO) {
I.clearSubclassOptionalData();
return;
}
FastMathFlags FMF = I.getFastMathFlags();
I.clearSubclassOptionalData();
I.setFastMathFlags(FMF);
}
/// Combine constant operands of associative operations either before or after a
/// cast to eliminate one of the associative operations:
/// (op (cast (op X, C2)), C1) --> (cast (op X, op (C1, C2)))
/// (op (cast (op X, C2)), C1) --> (op (cast X), op (C1, C2))
static bool simplifyAssocCastAssoc(BinaryOperator *BinOp1,
InstCombinerImpl &IC) {
auto *Cast = dyn_cast<CastInst>(BinOp1->getOperand(0));
if (!Cast || !Cast->hasOneUse())
return false;
// TODO: Enhance logic for other casts and remove this check.
auto CastOpcode = Cast->getOpcode();
if (CastOpcode != Instruction::ZExt)
return false;
// TODO: Enhance logic for other BinOps and remove this check.
if (!BinOp1->isBitwiseLogicOp())
return false;
auto AssocOpcode = BinOp1->getOpcode();
auto *BinOp2 = dyn_cast<BinaryOperator>(Cast->getOperand(0));
if (!BinOp2 || !BinOp2->hasOneUse() || BinOp2->getOpcode() != AssocOpcode)
return false;
Constant *C1, *C2;
if (!match(BinOp1->getOperand(1), m_Constant(C1)) ||
!match(BinOp2->getOperand(1), m_Constant(C2)))
return false;
// TODO: This assumes a zext cast.
// Eg, if it was a trunc, we'd cast C1 to the source type because casting C2
// to the destination type might lose bits.
// Fold the constants together in the destination type:
// (op (cast (op X, C2)), C1) --> (op (cast X), FoldedC)
const DataLayout &DL = IC.getDataLayout();
Type *DestTy = C1->getType();
Constant *CastC2 = ConstantFoldCastOperand(CastOpcode, C2, DestTy, DL);
if (!CastC2)
return false;
Constant *FoldedC = ConstantFoldBinaryOpOperands(AssocOpcode, C1, CastC2, DL);
if (!FoldedC)
return false;
IC.replaceOperand(*Cast, 0, BinOp2->getOperand(0));
IC.replaceOperand(*BinOp1, 1, FoldedC);
BinOp1->dropPoisonGeneratingFlags();
Cast->dropPoisonGeneratingFlags();
return true;
}
// Simplifies IntToPtr/PtrToInt RoundTrip Cast.
// inttoptr ( ptrtoint (x) ) --> x
Value *InstCombinerImpl::simplifyIntToPtrRoundTripCast(Value *Val) {
auto *IntToPtr = dyn_cast<IntToPtrInst>(Val);
if (IntToPtr && DL.getTypeSizeInBits(IntToPtr->getDestTy()) ==
DL.getTypeSizeInBits(IntToPtr->getSrcTy())) {
auto *PtrToInt = dyn_cast<PtrToIntInst>(IntToPtr->getOperand(0));
Type *CastTy = IntToPtr->getDestTy();
if (PtrToInt &&
CastTy->getPointerAddressSpace() ==
PtrToInt->getSrcTy()->getPointerAddressSpace() &&
DL.getTypeSizeInBits(PtrToInt->getSrcTy()) ==
DL.getTypeSizeInBits(PtrToInt->getDestTy()))
return PtrToInt->getOperand(0);
}
return nullptr;
}
/// This performs a few simplifications for operators that are associative or
/// commutative:
///
/// Commutative operators:
///
/// 1. Order operands such that they are listed from right (least complex) to
/// left (most complex). This puts constants before unary operators before
/// binary operators.
///
/// Associative operators:
///
/// 2. Transform: "(A op B) op C" ==> "A op (B op C)" if "B op C" simplifies.
/// 3. Transform: "A op (B op C)" ==> "(A op B) op C" if "A op B" simplifies.
///
/// Associative and commutative operators:
///
/// 4. Transform: "(A op B) op C" ==> "(C op A) op B" if "C op A" simplifies.
/// 5. Transform: "A op (B op C)" ==> "B op (C op A)" if "C op A" simplifies.
/// 6. Transform: "(A op C1) op (B op C2)" ==> "(A op B) op (C1 op C2)"
/// if C1 and C2 are constants.
bool InstCombinerImpl::SimplifyAssociativeOrCommutative(BinaryOperator &I) {
Instruction::BinaryOps Opcode = I.getOpcode();
bool Changed = false;
do {
// Order operands such that they are listed from right (least complex) to
// left (most complex). This puts constants before unary operators before
// binary operators.
if (I.isCommutative() && getComplexity(I.getOperand(0)) <
getComplexity(I.getOperand(1)))
Changed = !I.swapOperands();
if (I.isCommutative()) {
if (auto Pair = matchSymmetricPair(I.getOperand(0), I.getOperand(1))) {
replaceOperand(I, 0, Pair->first);
replaceOperand(I, 1, Pair->second);
Changed = true;
}
}
BinaryOperator *Op0 = dyn_cast<BinaryOperator>(I.getOperand(0));
BinaryOperator *Op1 = dyn_cast<BinaryOperator>(I.getOperand(1));
if (I.isAssociative()) {
// Transform: "(A op B) op C" ==> "A op (B op C)" if "B op C" simplifies.
if (Op0 && Op0->getOpcode() == Opcode) {
Value *A = Op0->getOperand(0);
Value *B = Op0->getOperand(1);
Value *C = I.getOperand(1);
// Does "B op C" simplify?
if (Value *V = simplifyBinOp(Opcode, B, C, SQ.getWithInstruction(&I))) {
// It simplifies to V. Form "A op V".
replaceOperand(I, 0, A);
replaceOperand(I, 1, V);
bool IsNUW = hasNoUnsignedWrap(I) && hasNoUnsignedWrap(*Op0);
bool IsNSW = maintainNoSignedWrap(I, B, C) && hasNoSignedWrap(*Op0);
// Conservatively clear all optional flags since they may not be
// preserved by the reassociation. Reset nsw/nuw based on the above
// analysis.
ClearSubclassDataAfterReassociation(I);
// Note: this is only valid because SimplifyBinOp doesn't look at
// the operands to Op0.
if (IsNUW)
I.setHasNoUnsignedWrap(true);
if (IsNSW)
I.setHasNoSignedWrap(true);
Changed = true;
++NumReassoc;
continue;
}
}
// Transform: "A op (B op C)" ==> "(A op B) op C" if "A op B" simplifies.
if (Op1 && Op1->getOpcode() == Opcode) {
Value *A = I.getOperand(0);
Value *B = Op1->getOperand(0);
Value *C = Op1->getOperand(1);
// Does "A op B" simplify?
if (Value *V = simplifyBinOp(Opcode, A, B, SQ.getWithInstruction(&I))) {
// It simplifies to V. Form "V op C".
replaceOperand(I, 0, V);
replaceOperand(I, 1, C);
// Conservatively clear the optional flags, since they may not be
// preserved by the reassociation.
ClearSubclassDataAfterReassociation(I);
Changed = true;
++NumReassoc;
continue;
}
}
}
if (I.isAssociative() && I.isCommutative()) {
if (simplifyAssocCastAssoc(&I, *this)) {
Changed = true;
++NumReassoc;
continue;
}
// Transform: "(A op B) op C" ==> "(C op A) op B" if "C op A" simplifies.
if (Op0 && Op0->getOpcode() == Opcode) {
Value *A = Op0->getOperand(0);
Value *B = Op0->getOperand(1);
Value *C = I.getOperand(1);
// Does "C op A" simplify?
if (Value *V = simplifyBinOp(Opcode, C, A, SQ.getWithInstruction(&I))) {
// It simplifies to V. Form "V op B".
replaceOperand(I, 0, V);
replaceOperand(I, 1, B);
// Conservatively clear the optional flags, since they may not be
// preserved by the reassociation.
ClearSubclassDataAfterReassociation(I);
Changed = true;
++NumReassoc;
continue;
}
}
// Transform: "A op (B op C)" ==> "B op (C op A)" if "C op A" simplifies.
if (Op1 && Op1->getOpcode() == Opcode) {
Value *A = I.getOperand(0);
Value *B = Op1->getOperand(0);
Value *C = Op1->getOperand(1);
// Does "C op A" simplify?
if (Value *V = simplifyBinOp(Opcode, C, A, SQ.getWithInstruction(&I))) {
// It simplifies to V. Form "B op V".
replaceOperand(I, 0, B);
replaceOperand(I, 1, V);
// Conservatively clear the optional flags, since they may not be
// preserved by the reassociation.
ClearSubclassDataAfterReassociation(I);
Changed = true;
++NumReassoc;
continue;
}
}
// Transform: "(A op C1) op (B op C2)" ==> "(A op B) op (C1 op C2)"
// if C1 and C2 are constants.
Value *A, *B;
Constant *C1, *C2, *CRes;
if (Op0 && Op1 &&
Op0->getOpcode() == Opcode && Op1->getOpcode() == Opcode &&
match(Op0, m_OneUse(m_BinOp(m_Value(A), m_Constant(C1)))) &&
match(Op1, m_OneUse(m_BinOp(m_Value(B), m_Constant(C2)))) &&
(CRes = ConstantFoldBinaryOpOperands(Opcode, C1, C2, DL))) {
bool IsNUW = hasNoUnsignedWrap(I) &&
hasNoUnsignedWrap(*Op0) &&
hasNoUnsignedWrap(*Op1);
BinaryOperator *NewBO = (IsNUW && Opcode == Instruction::Add) ?
BinaryOperator::CreateNUW(Opcode, A, B) :
BinaryOperator::Create(Opcode, A, B);
if (isa<FPMathOperator>(NewBO)) {
FastMathFlags Flags = I.getFastMathFlags() &
Op0->getFastMathFlags() &
Op1->getFastMathFlags();
NewBO->setFastMathFlags(Flags);
}
InsertNewInstWith(NewBO, I.getIterator());
NewBO->takeName(Op1);
replaceOperand(I, 0, NewBO);
replaceOperand(I, 1, CRes);
// Conservatively clear the optional flags, since they may not be
// preserved by the reassociation.
ClearSubclassDataAfterReassociation(I);
if (IsNUW)
I.setHasNoUnsignedWrap(true);
Changed = true;
continue;
}
}
// No further simplifications.
return Changed;
} while (true);
}
/// Return whether "X LOp (Y ROp Z)" is always equal to
/// "(X LOp Y) ROp (X LOp Z)".
static bool leftDistributesOverRight(Instruction::BinaryOps LOp,
Instruction::BinaryOps ROp) {
// X & (Y | Z) <--> (X & Y) | (X & Z)
// X & (Y ^ Z) <--> (X & Y) ^ (X & Z)
if (LOp == Instruction::And)
return ROp == Instruction::Or || ROp == Instruction::Xor;
// X | (Y & Z) <--> (X | Y) & (X | Z)
if (LOp == Instruction::Or)
return ROp == Instruction::And;
// X * (Y + Z) <--> (X * Y) + (X * Z)
// X * (Y - Z) <--> (X * Y) - (X * Z)
if (LOp == Instruction::Mul)
return ROp == Instruction::Add || ROp == Instruction::Sub;
return false;
}
/// Return whether "(X LOp Y) ROp Z" is always equal to
/// "(X ROp Z) LOp (Y ROp Z)".
static bool rightDistributesOverLeft(Instruction::BinaryOps LOp,
Instruction::BinaryOps ROp) {
if (Instruction::isCommutative(ROp))
return leftDistributesOverRight(ROp, LOp);
// (X {&|^} Y) >> Z <--> (X >> Z) {&|^} (Y >> Z) for all shifts.
return Instruction::isBitwiseLogicOp(LOp) && Instruction::isShift(ROp);
// TODO: It would be nice to handle division, aka "(X + Y)/Z = X/Z + Y/Z",
// but this requires knowing that the addition does not overflow and other
// such subtleties.
}
/// This function returns identity value for given opcode, which can be used to
/// factor patterns like (X * 2) + X ==> (X * 2) + (X * 1) ==> X * (2 + 1).
static Value *getIdentityValue(Instruction::BinaryOps Opcode, Value *V) {
if (isa<Constant>(V))
return nullptr;
return ConstantExpr::getBinOpIdentity(Opcode, V->getType());
}
/// This function predicates factorization using distributive laws. By default,
/// it just returns the 'Op' inputs. But for special-cases like
/// 'add(shl(X, 5), ...)', this function will have TopOpcode == Instruction::Add
/// and Op = shl(X, 5). The 'shl' is treated as the more general 'mul X, 32' to
/// allow more factorization opportunities.
static Instruction::BinaryOps
getBinOpsForFactorization(Instruction::BinaryOps TopOpcode, BinaryOperator *Op,
Value *&LHS, Value *&RHS, BinaryOperator *OtherOp) {
assert(Op && "Expected a binary operator");
LHS = Op->getOperand(0);
RHS = Op->getOperand(1);
if (TopOpcode == Instruction::Add || TopOpcode == Instruction::Sub) {
Constant *C;
if (match(Op, m_Shl(m_Value(), m_Constant(C)))) {
// X << C --> X * (1 << C)
RHS = ConstantExpr::getShl(ConstantInt::get(Op->getType(), 1), C);
return Instruction::Mul;
}
// TODO: We can add other conversions e.g. shr => div etc.
}
if (Instruction::isBitwiseLogicOp(TopOpcode)) {
if (OtherOp && OtherOp->getOpcode() == Instruction::AShr &&
match(Op, m_LShr(m_NonNegative(), m_Value()))) {
// lshr nneg C, X --> ashr nneg C, X
return Instruction::AShr;
}
}
return Op->getOpcode();
}
/// This tries to simplify binary operations by factorizing out common terms
/// (e. g. "(A*B)+(A*C)" -> "A*(B+C)").
static Value *tryFactorization(BinaryOperator &I, const SimplifyQuery &SQ,
InstCombiner::BuilderTy &Builder,
Instruction::BinaryOps InnerOpcode, Value *A,
Value *B, Value *C, Value *D) {
assert(A && B && C && D && "All values must be provided");
Value *V = nullptr;
Value *RetVal = nullptr;
Value *LHS = I.getOperand(0), *RHS = I.getOperand(1);
Instruction::BinaryOps TopLevelOpcode = I.getOpcode();
// Does "X op' Y" always equal "Y op' X"?
bool InnerCommutative = Instruction::isCommutative(InnerOpcode);
// Does "X op' (Y op Z)" always equal "(X op' Y) op (X op' Z)"?
if (leftDistributesOverRight(InnerOpcode, TopLevelOpcode)) {
// Does the instruction have the form "(A op' B) op (A op' D)" or, in the
// commutative case, "(A op' B) op (C op' A)"?
if (A == C || (InnerCommutative && A == D)) {
if (A != C)
std::swap(C, D);
// Consider forming "A op' (B op D)".
// If "B op D" simplifies then it can be formed with no cost.
V = simplifyBinOp(TopLevelOpcode, B, D, SQ.getWithInstruction(&I));
// If "B op D" doesn't simplify then only go on if one of the existing
// operations "A op' B" and "C op' D" will be zapped as no longer used.
if (!V && (LHS->hasOneUse() || RHS->hasOneUse()))
V = Builder.CreateBinOp(TopLevelOpcode, B, D, RHS->getName());
if (V)
RetVal = Builder.CreateBinOp(InnerOpcode, A, V);
}
}
// Does "(X op Y) op' Z" always equal "(X op' Z) op (Y op' Z)"?
if (!RetVal && rightDistributesOverLeft(TopLevelOpcode, InnerOpcode)) {
// Does the instruction have the form "(A op' B) op (C op' B)" or, in the
// commutative case, "(A op' B) op (B op' D)"?
if (B == D || (InnerCommutative && B == C)) {
if (B != D)
std::swap(C, D);
// Consider forming "(A op C) op' B".
// If "A op C" simplifies then it can be formed with no cost.
V = simplifyBinOp(TopLevelOpcode, A, C, SQ.getWithInstruction(&I));
// If "A op C" doesn't simplify then only go on if one of the existing
// operations "A op' B" and "C op' D" will be zapped as no longer used.
if (!V && (LHS->hasOneUse() || RHS->hasOneUse()))
V = Builder.CreateBinOp(TopLevelOpcode, A, C, LHS->getName());
if (V)
RetVal = Builder.CreateBinOp(InnerOpcode, V, B);
}
}
if (!RetVal)
return nullptr;
++NumFactor;
RetVal->takeName(&I);
// Try to add no-overflow flags to the final value.
if (isa<OverflowingBinaryOperator>(RetVal)) {
bool HasNSW = false;
bool HasNUW = false;
if (isa<OverflowingBinaryOperator>(&I)) {
HasNSW = I.hasNoSignedWrap();
HasNUW = I.hasNoUnsignedWrap();
}
if (auto *LOBO = dyn_cast<OverflowingBinaryOperator>(LHS)) {
HasNSW &= LOBO->hasNoSignedWrap();
HasNUW &= LOBO->hasNoUnsignedWrap();
}
if (auto *ROBO = dyn_cast<OverflowingBinaryOperator>(RHS)) {
HasNSW &= ROBO->hasNoSignedWrap();
HasNUW &= ROBO->hasNoUnsignedWrap();
}
if (TopLevelOpcode == Instruction::Add && InnerOpcode == Instruction::Mul) {
// We can propagate 'nsw' if we know that
// %Y = mul nsw i16 %X, C
// %Z = add nsw i16 %Y, %X
// =>
// %Z = mul nsw i16 %X, C+1
//
// iff C+1 isn't INT_MIN
const APInt *CInt;
if (match(V, m_APInt(CInt)) && !CInt->isMinSignedValue())
cast<Instruction>(RetVal)->setHasNoSignedWrap(HasNSW);
// nuw can be propagated with any constant or nuw value.
cast<Instruction>(RetVal)->setHasNoUnsignedWrap(HasNUW);
}
}
return RetVal;
}
// (Binop1 (Binop2 (logic_shift X, C), C1), (logic_shift Y, C))
// IFF
// 1) the logic_shifts match
// 2) either both binops are binops and one is `and` or
// BinOp1 is `and`
// (logic_shift (inv_logic_shift C1, C), C) == C1 or
//
// -> (logic_shift (Binop1 (Binop2 X, inv_logic_shift(C1, C)), Y), C)
//
// (Binop1 (Binop2 (logic_shift X, Amt), Mask), (logic_shift Y, Amt))
// IFF
// 1) the logic_shifts match
// 2) BinOp1 == BinOp2 (if BinOp == `add`, then also requires `shl`).
//
// -> (BinOp (logic_shift (BinOp X, Y)), Mask)
//
// (Binop1 (Binop2 (arithmetic_shift X, Amt), Mask), (arithmetic_shift Y, Amt))
// IFF
// 1) Binop1 is bitwise logical operator `and`, `or` or `xor`
// 2) Binop2 is `not`
//
// -> (arithmetic_shift Binop1((not X), Y), Amt)
Instruction *InstCombinerImpl::foldBinOpShiftWithShift(BinaryOperator &I) {
const DataLayout &DL = I.getModule()->getDataLayout();
auto IsValidBinOpc = [](unsigned Opc) {
switch (Opc) {
default:
return false;
case Instruction::And:
case Instruction::Or:
case Instruction::Xor:
case Instruction::Add:
// Skip Sub as we only match constant masks which will canonicalize to use
// add.
return true;
}
};
// Check if we can distribute binop arbitrarily. `add` + `lshr` has extra
// constraints.
auto IsCompletelyDistributable = [](unsigned BinOpc1, unsigned BinOpc2,
unsigned ShOpc) {
assert(ShOpc != Instruction::AShr);
return (BinOpc1 != Instruction::Add && BinOpc2 != Instruction::Add) ||
ShOpc == Instruction::Shl;
};
auto GetInvShift = [](unsigned ShOpc) {
assert(ShOpc != Instruction::AShr);
return ShOpc == Instruction::LShr ? Instruction::Shl : Instruction::LShr;
};
auto CanDistributeBinops = [&](unsigned BinOpc1, unsigned BinOpc2,
unsigned ShOpc, Constant *CMask,
Constant *CShift) {
// If the BinOp1 is `and` we don't need to check the mask.
if (BinOpc1 == Instruction::And)
return true;
// For all other possible transfers we need complete distributable
// binop/shift (anything but `add` + `lshr`).
if (!IsCompletelyDistributable(BinOpc1, BinOpc2, ShOpc))
return false;
// If BinOp2 is `and`, any mask works (this only really helps for non-splat
// vecs, otherwise the mask will be simplified and the following check will
// handle it).
if (BinOpc2 == Instruction::And)
return true;
// Otherwise, need mask that meets the below requirement.
// (logic_shift (inv_logic_shift Mask, ShAmt), ShAmt) == Mask
Constant *MaskInvShift =
ConstantFoldBinaryOpOperands(GetInvShift(ShOpc), CMask, CShift, DL);
return ConstantFoldBinaryOpOperands(ShOpc, MaskInvShift, CShift, DL) ==
CMask;
};
auto MatchBinOp = [&](unsigned ShOpnum) -> Instruction * {
Constant *CMask, *CShift;
Value *X, *Y, *ShiftedX, *Mask, *Shift;
if (!match(I.getOperand(ShOpnum),
m_OneUse(m_Shift(m_Value(Y), m_Value(Shift)))))
return nullptr;
if (!match(I.getOperand(1 - ShOpnum),
m_BinOp(m_Value(ShiftedX), m_Value(Mask))))
return nullptr;
if (!match(ShiftedX, m_OneUse(m_Shift(m_Value(X), m_Specific(Shift)))))
return nullptr;
// Make sure we are matching instruction shifts and not ConstantExpr
auto *IY = dyn_cast<Instruction>(I.getOperand(ShOpnum));
auto *IX = dyn_cast<Instruction>(ShiftedX);
if (!IY || !IX)
return nullptr;
// LHS and RHS need same shift opcode
unsigned ShOpc = IY->getOpcode();
if (ShOpc != IX->getOpcode())
return nullptr;
// Make sure binop is real instruction and not ConstantExpr
auto *BO2 = dyn_cast<Instruction>(I.getOperand(1 - ShOpnum));
if (!BO2)
return nullptr;
unsigned BinOpc = BO2->getOpcode();
// Make sure we have valid binops.
if (!IsValidBinOpc(I.getOpcode()) || !IsValidBinOpc(BinOpc))
return nullptr;
if (ShOpc == Instruction::AShr) {
if (Instruction::isBitwiseLogicOp(I.getOpcode()) &&
BinOpc == Instruction::Xor && match(Mask, m_AllOnes())) {
Value *NotX = Builder.CreateNot(X);
Value *NewBinOp = Builder.CreateBinOp(I.getOpcode(), Y, NotX);
return BinaryOperator::Create(
static_cast<Instruction::BinaryOps>(ShOpc), NewBinOp, Shift);
}
return nullptr;
}
// If BinOp1 == BinOp2 and it's bitwise or shl with add, then just
// distribute to drop the shift irrelevant of constants.
if (BinOpc == I.getOpcode() &&
IsCompletelyDistributable(I.getOpcode(), BinOpc, ShOpc)) {
Value *NewBinOp2 = Builder.CreateBinOp(I.getOpcode(), X, Y);
Value *NewBinOp1 = Builder.CreateBinOp(
static_cast<Instruction::BinaryOps>(ShOpc), NewBinOp2, Shift);
return BinaryOperator::Create(I.getOpcode(), NewBinOp1, Mask);
}
// Otherwise we can only distribute by constant shifting the mask, so
// ensure we have constants.
if (!match(Shift, m_ImmConstant(CShift)))
return nullptr;
if (!match(Mask, m_ImmConstant(CMask)))
return nullptr;
// Check if we can distribute the binops.
if (!CanDistributeBinops(I.getOpcode(), BinOpc, ShOpc, CMask, CShift))
return nullptr;
Constant *NewCMask =
ConstantFoldBinaryOpOperands(GetInvShift(ShOpc), CMask, CShift, DL);
Value *NewBinOp2 = Builder.CreateBinOp(
static_cast<Instruction::BinaryOps>(BinOpc), X, NewCMask);
Value *NewBinOp1 = Builder.CreateBinOp(I.getOpcode(), Y, NewBinOp2);
return BinaryOperator::Create(static_cast<Instruction::BinaryOps>(ShOpc),
NewBinOp1, CShift);
};
if (Instruction *R = MatchBinOp(0))
return R;
return MatchBinOp(1);
}
// (Binop (zext C), (select C, T, F))
// -> (select C, (binop 1, T), (binop 0, F))
//
// (Binop (sext C), (select C, T, F))
// -> (select C, (binop -1, T), (binop 0, F))
//
// Attempt to simplify binary operations into a select with folded args, when
// one operand of the binop is a select instruction and the other operand is a
// zext/sext extension, whose value is the select condition.
Instruction *
InstCombinerImpl::foldBinOpOfSelectAndCastOfSelectCondition(BinaryOperator &I) {
// TODO: this simplification may be extended to any speculatable instruction,
// not just binops, and would possibly be handled better in FoldOpIntoSelect.
Instruction::BinaryOps Opc = I.getOpcode();
Value *LHS = I.getOperand(0), *RHS = I.getOperand(1);
Value *A, *CondVal, *TrueVal, *FalseVal;
Value *CastOp;
auto MatchSelectAndCast = [&](Value *CastOp, Value *SelectOp) {
return match(CastOp, m_ZExtOrSExt(m_Value(A))) &&
A->getType()->getScalarSizeInBits() == 1 &&
match(SelectOp, m_Select(m_Value(CondVal), m_Value(TrueVal),
m_Value(FalseVal)));
};
// Make sure one side of the binop is a select instruction, and the other is a
// zero/sign extension operating on a i1.
if (MatchSelectAndCast(LHS, RHS))
CastOp = LHS;
else if (MatchSelectAndCast(RHS, LHS))
CastOp = RHS;
else
return nullptr;
auto NewFoldedConst = [&](bool IsTrueArm, Value *V) {
bool IsCastOpRHS = (CastOp == RHS);
bool IsZExt = isa<ZExtInst>(CastOp);
Constant *C;
if (IsTrueArm) {
C = Constant::getNullValue(V->getType());
} else if (IsZExt) {
unsigned BitWidth = V->getType()->getScalarSizeInBits();
C = Constant::getIntegerValue(V->getType(), APInt(BitWidth, 1));
} else {
C = Constant::getAllOnesValue(V->getType());
}
return IsCastOpRHS ? Builder.CreateBinOp(Opc, V, C)
: Builder.CreateBinOp(Opc, C, V);
};
// If the value used in the zext/sext is the select condition, or the negated
// of the select condition, the binop can be simplified.
if (CondVal == A) {
Value *NewTrueVal = NewFoldedConst(false, TrueVal);
return SelectInst::Create(CondVal, NewTrueVal,
NewFoldedConst(true, FalseVal));
}
if (match(A, m_Not(m_Specific(CondVal)))) {
Value *NewTrueVal = NewFoldedConst(true, TrueVal);
return SelectInst::Create(CondVal, NewTrueVal,
NewFoldedConst(false, FalseVal));
}
return nullptr;
}
Value *InstCombinerImpl::tryFactorizationFolds(BinaryOperator &I) {
Value *LHS = I.getOperand(0), *RHS = I.getOperand(1);
BinaryOperator *Op0 = dyn_cast<BinaryOperator>(LHS);
BinaryOperator *Op1 = dyn_cast<BinaryOperator>(RHS);
Instruction::BinaryOps TopLevelOpcode = I.getOpcode();
Value *A, *B, *C, *D;
Instruction::BinaryOps LHSOpcode, RHSOpcode;
if (Op0)
LHSOpcode = getBinOpsForFactorization(TopLevelOpcode, Op0, A, B, Op1);
if (Op1)
RHSOpcode = getBinOpsForFactorization(TopLevelOpcode, Op1, C, D, Op0);
// The instruction has the form "(A op' B) op (C op' D)". Try to factorize
// a common term.
if (Op0 && Op1 && LHSOpcode == RHSOpcode)
if (Value *V = tryFactorization(I, SQ, Builder, LHSOpcode, A, B, C, D))
return V;
// The instruction has the form "(A op' B) op (C)". Try to factorize common
// term.
if (Op0)
if (Value *Ident = getIdentityValue(LHSOpcode, RHS))
if (Value *V =
tryFactorization(I, SQ, Builder, LHSOpcode, A, B, RHS, Ident))
return V;
// The instruction has the form "(B) op (C op' D)". Try to factorize common
// term.
if (Op1)
if (Value *Ident = getIdentityValue(RHSOpcode, LHS))
if (Value *V =
tryFactorization(I, SQ, Builder, RHSOpcode, LHS, Ident, C, D))
return V;
return nullptr;
}
/// This tries to simplify binary operations which some other binary operation
/// distributes over either by factorizing out common terms
/// (eg "(A*B)+(A*C)" -> "A*(B+C)") or expanding out if this results in
/// simplifications (eg: "A & (B | C) -> (A&B) | (A&C)" if this is a win).
/// Returns the simplified value, or null if it didn't simplify.
Value *InstCombinerImpl::foldUsingDistributiveLaws(BinaryOperator &I) {
Value *LHS = I.getOperand(0), *RHS = I.getOperand(1);
BinaryOperator *Op0 = dyn_cast<BinaryOperator>(LHS);
BinaryOperator *Op1 = dyn_cast<BinaryOperator>(RHS);
Instruction::BinaryOps TopLevelOpcode = I.getOpcode();
// Factorization.
if (Value *R = tryFactorizationFolds(I))
return R;
// Expansion.
if (Op0 && rightDistributesOverLeft(Op0->getOpcode(), TopLevelOpcode)) {
// The instruction has the form "(A op' B) op C". See if expanding it out
// to "(A op C) op' (B op C)" results in simplifications.
Value *A = Op0->getOperand(0), *B = Op0->getOperand(1), *C = RHS;
Instruction::BinaryOps InnerOpcode = Op0->getOpcode(); // op'
// Disable the use of undef because it's not safe to distribute undef.
auto SQDistributive = SQ.getWithInstruction(&I).getWithoutUndef();
Value *L = simplifyBinOp(TopLevelOpcode, A, C, SQDistributive);
Value *R = simplifyBinOp(TopLevelOpcode, B, C, SQDistributive);
// Do "A op C" and "B op C" both simplify?
if (L && R) {
// They do! Return "L op' R".
++NumExpand;
C = Builder.CreateBinOp(InnerOpcode, L, R);
C->takeName(&I);
return C;
}
// Does "A op C" simplify to the identity value for the inner opcode?
if (L && L == ConstantExpr::getBinOpIdentity(InnerOpcode, L->getType())) {
// They do! Return "B op C".
++NumExpand;
C = Builder.CreateBinOp(TopLevelOpcode, B, C);
C->takeName(&I);
return C;
}
// Does "B op C" simplify to the identity value for the inner opcode?
if (R && R == ConstantExpr::getBinOpIdentity(InnerOpcode, R->getType())) {
// They do! Return "A op C".
++NumExpand;
C = Builder.CreateBinOp(TopLevelOpcode, A, C);
C->takeName(&I);
return C;
}
}
if (Op1 && leftDistributesOverRight(TopLevelOpcode, Op1->getOpcode())) {
// The instruction has the form "A op (B op' C)". See if expanding it out
// to "(A op B) op' (A op C)" results in simplifications.
Value *A = LHS, *B = Op1->getOperand(0), *C = Op1->getOperand(1);
Instruction::BinaryOps InnerOpcode = Op1->getOpcode(); // op'
// Disable the use of undef because it's not safe to distribute undef.
auto SQDistributive = SQ.getWithInstruction(&I).getWithoutUndef();
Value *L = simplifyBinOp(TopLevelOpcode, A, B, SQDistributive);
Value *R = simplifyBinOp(TopLevelOpcode, A, C, SQDistributive);
// Do "A op B" and "A op C" both simplify?
if (L && R) {
// They do! Return "L op' R".
++NumExpand;
A = Builder.CreateBinOp(InnerOpcode, L, R);
A->takeName(&I);
return A;
}
// Does "A op B" simplify to the identity value for the inner opcode?
if (L && L == ConstantExpr::getBinOpIdentity(InnerOpcode, L->getType())) {
// They do! Return "A op C".
++NumExpand;
A = Builder.CreateBinOp(TopLevelOpcode, A, C);
A->takeName(&I);
return A;
}
// Does "A op C" simplify to the identity value for the inner opcode?
if (R && R == ConstantExpr::getBinOpIdentity(InnerOpcode, R->getType())) {
// They do! Return "A op B".
++NumExpand;
A = Builder.CreateBinOp(TopLevelOpcode, A, B);
A->takeName(&I);
return A;
}
}
return SimplifySelectsFeedingBinaryOp(I, LHS, RHS);
}
static std::optional<std::pair<Value *, Value *>>
matchSymmetricPhiNodesPair(PHINode *LHS, PHINode *RHS) {
if (LHS->getParent() != RHS->getParent())
return std::nullopt;
if (LHS->getNumIncomingValues() < 2)
return std::nullopt;
if (!equal(LHS->blocks(), RHS->blocks()))
return std::nullopt;
Value *L0 = LHS->getIncomingValue(0);
Value *R0 = RHS->getIncomingValue(0);
for (unsigned I = 1, E = LHS->getNumIncomingValues(); I != E; ++I) {
Value *L1 = LHS->getIncomingValue(I);
Value *R1 = RHS->getIncomingValue(I);
if ((L0 == L1 && R0 == R1) || (L0 == R1 && R0 == L1))
continue;
return std::nullopt;
}
return std::optional(std::pair(L0, R0));
}
std::optional<std::pair<Value *, Value *>>
InstCombinerImpl::matchSymmetricPair(Value *LHS, Value *RHS) {
Instruction *LHSInst = dyn_cast<Instruction>(LHS);
Instruction *RHSInst = dyn_cast<Instruction>(RHS);
if (!LHSInst || !RHSInst || LHSInst->getOpcode() != RHSInst->getOpcode())
return std::nullopt;
switch (LHSInst->getOpcode()) {
case Instruction::PHI:
return matchSymmetricPhiNodesPair(cast<PHINode>(LHS), cast<PHINode>(RHS));
case Instruction::Select: {
Value *Cond = LHSInst->getOperand(0);
Value *TrueVal = LHSInst->getOperand(1);
Value *FalseVal = LHSInst->getOperand(2);
if (Cond == RHSInst->getOperand(0) && TrueVal == RHSInst->getOperand(2) &&
FalseVal == RHSInst->getOperand(1))
return std::pair(TrueVal, FalseVal);
return std::nullopt;
}
case Instruction::Call: {
// Match min(a, b) and max(a, b)
MinMaxIntrinsic *LHSMinMax = dyn_cast<MinMaxIntrinsic>(LHSInst);
MinMaxIntrinsic *RHSMinMax = dyn_cast<MinMaxIntrinsic>(RHSInst);
if (LHSMinMax && RHSMinMax &&
LHSMinMax->getPredicate() ==
ICmpInst::getSwappedPredicate(RHSMinMax->getPredicate()) &&
((LHSMinMax->getLHS() == RHSMinMax->getLHS() &&
LHSMinMax->getRHS() == RHSMinMax->getRHS()) ||
(LHSMinMax->getLHS() == RHSMinMax->getRHS() &&
LHSMinMax->getRHS() == RHSMinMax->getLHS())))
return std::pair(LHSMinMax->getLHS(), LHSMinMax->getRHS());
return std::nullopt;
}
default:
return std::nullopt;
}
}
Value *InstCombinerImpl::SimplifySelectsFeedingBinaryOp(BinaryOperator &I,
Value *LHS,
Value *RHS) {
Value *A, *B, *C, *D, *E, *F;
bool LHSIsSelect = match(LHS, m_Select(m_Value(A), m_Value(B), m_Value(C)));
bool RHSIsSelect = match(RHS, m_Select(m_Value(D), m_Value(E), m_Value(F)));
if (!LHSIsSelect && !RHSIsSelect)
return nullptr;
FastMathFlags FMF;
BuilderTy::FastMathFlagGuard Guard(Builder);
if (isa<FPMathOperator>(&I)) {
FMF = I.getFastMathFlags();
Builder.setFastMathFlags(FMF);
}
Instruction::BinaryOps Opcode = I.getOpcode();
SimplifyQuery Q = SQ.getWithInstruction(&I);
Value *Cond, *True = nullptr, *False = nullptr;
// Special-case for add/negate combination. Replace the zero in the negation
// with the trailing add operand:
// (Cond ? TVal : -N) + Z --> Cond ? True : (Z - N)
// (Cond ? -N : FVal) + Z --> Cond ? (Z - N) : False
auto foldAddNegate = [&](Value *TVal, Value *FVal, Value *Z) -> Value * {
// We need an 'add' and exactly 1 arm of the select to have been simplified.
if (Opcode != Instruction::Add || (!True && !False) || (True && False))
return nullptr;
Value *N;
if (True && match(FVal, m_Neg(m_Value(N)))) {
Value *Sub = Builder.CreateSub(Z, N);
return Builder.CreateSelect(Cond, True, Sub, I.getName());
}
if (False && match(TVal, m_Neg(m_Value(N)))) {
Value *Sub = Builder.CreateSub(Z, N);
return Builder.CreateSelect(Cond, Sub, False, I.getName());
}
return nullptr;
};
if (LHSIsSelect && RHSIsSelect && A == D) {
// (A ? B : C) op (A ? E : F) -> A ? (B op E) : (C op F)
Cond = A;
True = simplifyBinOp(Opcode, B, E, FMF, Q);
False = simplifyBinOp(Opcode, C, F, FMF, Q);
if (LHS->hasOneUse() && RHS->hasOneUse()) {
if (False && !True)
True = Builder.CreateBinOp(Opcode, B, E);
else if (True && !False)
False = Builder.CreateBinOp(Opcode, C, F);
}
} else if (LHSIsSelect && LHS->hasOneUse()) {
// (A ? B : C) op Y -> A ? (B op Y) : (C op Y)
Cond = A;
True = simplifyBinOp(Opcode, B, RHS, FMF, Q);
False = simplifyBinOp(Opcode, C, RHS, FMF, Q);
if (Value *NewSel = foldAddNegate(B, C, RHS))
return NewSel;
} else if (RHSIsSelect && RHS->hasOneUse()) {
// X op (D ? E : F) -> D ? (X op E) : (X op F)
Cond = D;
True = simplifyBinOp(Opcode, LHS, E, FMF, Q);
False = simplifyBinOp(Opcode, LHS, F, FMF, Q);
if (Value *NewSel = foldAddNegate(E, F, LHS))
return NewSel;
}
if (!True || !False)
return nullptr;
Value *SI = Builder.CreateSelect(Cond, True, False);
SI->takeName(&I);
return SI;
}
/// Freely adapt every user of V as-if V was changed to !V.
/// WARNING: only if canFreelyInvertAllUsersOf() said this can be done.
void InstCombinerImpl::freelyInvertAllUsersOf(Value *I, Value *IgnoredUser) {
assert(!isa<Constant>(I) && "Shouldn't invert users of constant");
for (User *U : make_early_inc_range(I->users())) {
if (U == IgnoredUser)
continue; // Don't consider this user.
switch (cast<Instruction>(U)->getOpcode()) {
case Instruction::Select: {
auto *SI = cast<SelectInst>(U);
SI->swapValues();
SI->swapProfMetadata();
break;
}
case Instruction::Br:
cast<BranchInst>(U)->swapSuccessors(); // swaps prof metadata too
break;
case Instruction::Xor:
replaceInstUsesWith(cast<Instruction>(*U), I);
// Add to worklist for DCE.
addToWorklist(cast<Instruction>(U));
break;
default:
llvm_unreachable("Got unexpected user - out of sync with "
"canFreelyInvertAllUsersOf() ?");
}
}
}
/// Given a 'sub' instruction, return the RHS of the instruction if the LHS is a
/// constant zero (which is the 'negate' form).
Value *InstCombinerImpl::dyn_castNegVal(Value *V) const {
Value *NegV;
if (match(V, m_Neg(m_Value(NegV))))
return NegV;
// Constants can be considered to be negated values if they can be folded.
if (ConstantInt *C = dyn_cast<ConstantInt>(V))
return ConstantExpr::getNeg(C);
if (ConstantDataVector *C = dyn_cast<ConstantDataVector>(V))
if (C->getType()->getElementType()->isIntegerTy())
return ConstantExpr::getNeg(C);
if (ConstantVector *CV = dyn_cast<ConstantVector>(V)) {
for (unsigned i = 0, e = CV->getNumOperands(); i != e; ++i) {
Constant *Elt = CV->getAggregateElement(i);
if (!Elt)
return nullptr;
if (isa<UndefValue>(Elt))
continue;
if (!isa<ConstantInt>(Elt))
return nullptr;
}
return ConstantExpr::getNeg(CV);
}
// Negate integer vector splats.
if (auto *CV = dyn_cast<Constant>(V))
if (CV->getType()->isVectorTy() &&
CV->getType()->getScalarType()->isIntegerTy() && CV->getSplatValue())
return ConstantExpr::getNeg(CV);
return nullptr;
}
/// A binop with a constant operand and a sign-extended boolean operand may be
/// converted into a select of constants by applying the binary operation to
/// the constant with the two possible values of the extended boolean (0 or -1).
Instruction *InstCombinerImpl::foldBinopOfSextBoolToSelect(BinaryOperator &BO) {
// TODO: Handle non-commutative binop (constant is operand 0).
// TODO: Handle zext.
// TODO: Peek through 'not' of cast.
Value *BO0 = BO.getOperand(0);
Value *BO1 = BO.getOperand(1);
Value *X;
Constant *C;
if (!match(BO0, m_SExt(m_Value(X))) || !match(BO1, m_ImmConstant(C)) ||
!X->getType()->isIntOrIntVectorTy(1))
return nullptr;
// bo (sext i1 X), C --> select X, (bo -1, C), (bo 0, C)
Constant *Ones = ConstantInt::getAllOnesValue(BO.getType());
Constant *Zero = ConstantInt::getNullValue(BO.getType());
Value *TVal = Builder.CreateBinOp(BO.getOpcode(), Ones, C);
Value *FVal = Builder.CreateBinOp(BO.getOpcode(), Zero, C);
return SelectInst::Create(X, TVal, FVal);
}
static Constant *constantFoldOperationIntoSelectOperand(Instruction &I,
SelectInst *SI,
bool IsTrueArm) {
SmallVector<Constant *> ConstOps;
for (Value *Op : I.operands()) {
CmpInst::Predicate Pred;
Constant *C = nullptr;
if (Op == SI) {
C = dyn_cast<Constant>(IsTrueArm ? SI->getTrueValue()
: SI->getFalseValue());
} else if (match(SI->getCondition(),
m_ICmp(Pred, m_Specific(Op), m_Constant(C))) &&
Pred == (IsTrueArm ? ICmpInst::ICMP_EQ : ICmpInst::ICMP_NE) &&
isGuaranteedNotToBeUndefOrPoison(C)) {
// Pass
} else {
C = dyn_cast<Constant>(Op);
}
if (C == nullptr)
return nullptr;
ConstOps.push_back(C);
}
return ConstantFoldInstOperands(&I, ConstOps, I.getModule()->getDataLayout());
}
static Value *foldOperationIntoSelectOperand(Instruction &I, SelectInst *SI,
Value *NewOp, InstCombiner &IC) {
Instruction *Clone = I.clone();
Clone->replaceUsesOfWith(SI, NewOp);
IC.InsertNewInstBefore(Clone, SI->getIterator());
return Clone;
}
Instruction *InstCombinerImpl::FoldOpIntoSelect(Instruction &Op, SelectInst *SI,
bool FoldWithMultiUse) {
// Don't modify shared select instructions unless set FoldWithMultiUse
if (!SI->hasOneUse() && !FoldWithMultiUse)
return nullptr;
Value *TV = SI->getTrueValue();
Value *FV = SI->getFalseValue();
if (!(isa<Constant>(TV) || isa<Constant>(FV)))
return nullptr;
// Bool selects with constant operands can be folded to logical ops.
if (SI->getType()->isIntOrIntVectorTy(1))
return nullptr;
// If it's a bitcast involving vectors, make sure it has the same number of
// elements on both sides.
if (auto *BC = dyn_cast<BitCastInst>(&Op)) {
VectorType *DestTy = dyn_cast<VectorType>(BC->getDestTy());
VectorType *SrcTy = dyn_cast<VectorType>(BC->getSrcTy());
// Verify that either both or neither are vectors.
if ((SrcTy == nullptr) != (DestTy == nullptr))
return nullptr;
// If vectors, verify that they have the same number of elements.
if (SrcTy && SrcTy->getElementCount() != DestTy->getElementCount())
return nullptr;
}
// Test if a FCmpInst instruction is used exclusively by a select as
// part of a minimum or maximum operation. If so, refrain from doing
// any other folding. This helps out other analyses which understand
// non-obfuscated minimum and maximum idioms. And in this case, at
// least one of the comparison operands has at least one user besides
// the compare (the select), which would often largely negate the
// benefit of folding anyway.
if (auto *CI = dyn_cast<FCmpInst>(SI->getCondition())) {
if (CI->hasOneUse()) {
Value *Op0 = CI->getOperand(0), *Op1 = CI->getOperand(1);
if ((TV == Op0 && FV == Op1) || (FV == Op0 && TV == Op1))
return nullptr;
}
}
// Make sure that one of the select arms constant folds successfully.
Value *NewTV = constantFoldOperationIntoSelectOperand(Op, SI, /*IsTrueArm*/ true);
Value *NewFV = constantFoldOperationIntoSelectOperand(Op, SI, /*IsTrueArm*/ false);
if (!NewTV && !NewFV)
return nullptr;
// Create an instruction for the arm that did not fold.
if (!NewTV)
NewTV = foldOperationIntoSelectOperand(Op, SI, TV, *this);
if (!NewFV)
NewFV = foldOperationIntoSelectOperand(Op, SI, FV, *this);
return SelectInst::Create(SI->getCondition(), NewTV, NewFV, "", nullptr, SI);
}
static Value *simplifyInstructionWithPHI(Instruction &I, PHINode *PN,
Value *InValue, BasicBlock *InBB,
const DataLayout &DL,
const SimplifyQuery SQ) {
// NB: It is a precondition of this transform that the operands be
// phi translatable! This is usually trivially satisfied by limiting it
// to constant ops, and for selects we do a more sophisticated check.
SmallVector<Value *> Ops;
for (Value *Op : I.operands()) {
if (Op == PN)
Ops.push_back(InValue);
else
Ops.push_back(Op->DoPHITranslation(PN->getParent(), InBB));
}
// Don't consider the simplification successful if we get back a constant
// expression. That's just an instruction in hiding.
// Also reject the case where we simplify back to the phi node. We wouldn't
// be able to remove it in that case.
Value *NewVal = simplifyInstructionWithOperands(
&I, Ops, SQ.getWithInstruction(InBB->getTerminator()));
if (NewVal && NewVal != PN && !match(NewVal, m_ConstantExpr()))
return NewVal;
// Check if incoming PHI value can be replaced with constant
// based on implied condition.
BranchInst *TerminatorBI = dyn_cast<BranchInst>(InBB->getTerminator());
const ICmpInst *ICmp = dyn_cast<ICmpInst>(&I);
if (TerminatorBI && TerminatorBI->isConditional() &&
TerminatorBI->getSuccessor(0) != TerminatorBI->getSuccessor(1) && ICmp) {
bool LHSIsTrue = TerminatorBI->getSuccessor(0) == PN->getParent();
std::optional<bool> ImpliedCond =
isImpliedCondition(TerminatorBI->getCondition(), ICmp->getPredicate(),
Ops[0], Ops[1], DL, LHSIsTrue);
if (ImpliedCond)
return ConstantInt::getBool(I.getType(), ImpliedCond.value());
}
return nullptr;
}
Instruction *InstCombinerImpl::foldOpIntoPhi(Instruction &I, PHINode *PN) {
unsigned NumPHIValues = PN->getNumIncomingValues();
if (NumPHIValues == 0)
return nullptr;
// We normally only transform phis with a single use. However, if a PHI has
// multiple uses and they are all the same operation, we can fold *all* of the
// uses into the PHI.
if (!PN->hasOneUse()) {
// Walk the use list for the instruction, comparing them to I.
for (User *U : PN->users()) {
Instruction *UI = cast<Instruction>(U);
if (UI != &I && !I.isIdenticalTo(UI))
return nullptr;
}
// Otherwise, we can replace *all* users with the new PHI we form.
}
// Check to see whether the instruction can be folded into each phi operand.
// If there is one operand that does not fold, remember the BB it is in.
// If there is more than one or if *it* is a PHI, bail out.
SmallVector<Value *> NewPhiValues;
BasicBlock *NonSimplifiedBB = nullptr;
Value *NonSimplifiedInVal = nullptr;
for (unsigned i = 0; i != NumPHIValues; ++i) {
Value *InVal = PN->getIncomingValue(i);
BasicBlock *InBB = PN->getIncomingBlock(i);
if (auto *NewVal = simplifyInstructionWithPHI(I, PN, InVal, InBB, DL, SQ)) {
NewPhiValues.push_back(NewVal);
continue;
}
if (NonSimplifiedBB) return nullptr; // More than one non-simplified value.
NonSimplifiedBB = InBB;
NonSimplifiedInVal = InVal;
NewPhiValues.push_back(nullptr);
// If the InVal is an invoke at the end of the pred block, then we can't
// insert a computation after it without breaking the edge.
if (isa<InvokeInst>(InVal))
if (cast<Instruction>(InVal)->getParent() == NonSimplifiedBB)
return nullptr;
// If the incoming non-constant value is reachable from the phis block,
// we'll push the operation across a loop backedge. This could result in
// an infinite combine loop, and is generally non-profitable (especially
// if the operation was originally outside the loop).
if (isPotentiallyReachable(PN->getParent(), NonSimplifiedBB, nullptr, &DT,
LI))
return nullptr;
}
// If there is exactly one non-simplified value, we can insert a copy of the
// operation in that block. However, if this is a critical edge, we would be
// inserting the computation on some other paths (e.g. inside a loop). Only
// do this if the pred block is unconditionally branching into the phi block.
// Also, make sure that the pred block is not dead code.
if (NonSimplifiedBB != nullptr) {
BranchInst *BI = dyn_cast<BranchInst>(NonSimplifiedBB->getTerminator());
if (!BI || !BI->isUnconditional() ||
!DT.isReachableFromEntry(NonSimplifiedBB))
return nullptr;
}
// Okay, we can do the transformation: create the new PHI node.
PHINode *NewPN = PHINode::Create(I.getType(), PN->getNumIncomingValues());
InsertNewInstBefore(NewPN, PN->getIterator());
NewPN->takeName(PN);
NewPN->setDebugLoc(PN->getDebugLoc());
// If we are going to have to insert a new computation, do so right before the
// predecessor's terminator.
Instruction *Clone = nullptr;
if (NonSimplifiedBB) {
Clone = I.clone();
for (Use &U : Clone->operands()) {
if (U == PN)
U = NonSimplifiedInVal;
else
U = U->DoPHITranslation(PN->getParent(), NonSimplifiedBB);
}
InsertNewInstBefore(Clone, NonSimplifiedBB->getTerminator()->getIterator());
}
for (unsigned i = 0; i != NumPHIValues; ++i) {
if (NewPhiValues[i])
NewPN->addIncoming(NewPhiValues[i], PN->getIncomingBlock(i));
else
NewPN->addIncoming(Clone, PN->getIncomingBlock(i));
}
for (User *U : make_early_inc_range(PN->users())) {
Instruction *User = cast<Instruction>(U);
if (User == &I) continue;
replaceInstUsesWith(*User, NewPN);
eraseInstFromFunction(*User);
}
replaceAllDbgUsesWith(const_cast<PHINode &>(*PN),
const_cast<PHINode &>(*NewPN),
const_cast<PHINode &>(*PN), DT);
return replaceInstUsesWith(I, NewPN);
}
Instruction *InstCombinerImpl::foldBinopWithPhiOperands(BinaryOperator &BO) {
// TODO: This should be similar to the incoming values check in foldOpIntoPhi:
// we are guarding against replicating the binop in >1 predecessor.
// This could miss matching a phi with 2 constant incoming values.
auto *Phi0 = dyn_cast<PHINode>(BO.getOperand(0));
auto *Phi1 = dyn_cast<PHINode>(BO.getOperand(1));
if (!Phi0 || !Phi1 || !Phi0->hasOneUse() || !Phi1->hasOneUse() ||
Phi0->getNumOperands() != Phi1->getNumOperands())
return nullptr;
// TODO: Remove the restriction for binop being in the same block as the phis.
if (BO.getParent() != Phi0->getParent() ||
BO.getParent() != Phi1->getParent())
return nullptr;
// Fold if there is at least one specific constant value in phi0 or phi1's
// incoming values that comes from the same block and this specific constant
// value can be used to do optimization for specific binary operator.
// For example:
// %phi0 = phi i32 [0, %bb0], [%i, %bb1]
// %phi1 = phi i32 [%j, %bb0], [0, %bb1]
// %add = add i32 %phi0, %phi1
// ==>
// %add = phi i32 [%j, %bb0], [%i, %bb1]
Constant *C = ConstantExpr::getBinOpIdentity(BO.getOpcode(), BO.getType(),
/*AllowRHSConstant*/ false);
if (C) {
SmallVector<Value *, 4> NewIncomingValues;
auto CanFoldIncomingValuePair = [&](std::tuple<Use &, Use &> T) {
auto &Phi0Use = std::get<0>(T);
auto &Phi1Use = std::get<1>(T);
if (Phi0->getIncomingBlock(Phi0Use) != Phi1->getIncomingBlock(Phi1Use))
return false;
Value *Phi0UseV = Phi0Use.get();
Value *Phi1UseV = Phi1Use.get();
if (Phi0UseV == C)
NewIncomingValues.push_back(Phi1UseV);
else if (Phi1UseV == C)
NewIncomingValues.push_back(Phi0UseV);
else
return false;
return true;
};
if (all_of(zip(Phi0->operands(), Phi1->operands()),
CanFoldIncomingValuePair)) {
PHINode *NewPhi =
PHINode::Create(Phi0->getType(), Phi0->getNumOperands());
assert(NewIncomingValues.size() == Phi0->getNumOperands() &&
"The number of collected incoming values should equal the number "
"of the original PHINode operands!");
for (unsigned I = 0; I < Phi0->getNumOperands(); I++)
NewPhi->addIncoming(NewIncomingValues[I], Phi0->getIncomingBlock(I));
return NewPhi;
}
}
if (Phi0->getNumOperands() != 2 || Phi1->getNumOperands() != 2)
return nullptr;
// Match a pair of incoming constants for one of the predecessor blocks.
BasicBlock *ConstBB, *OtherBB;
Constant *C0, *C1;
if (match(Phi0->getIncomingValue(0), m_ImmConstant(C0))) {
ConstBB = Phi0->getIncomingBlock(0);
OtherBB = Phi0->getIncomingBlock(1);
} else if (match(Phi0->getIncomingValue(1), m_ImmConstant(C0))) {
ConstBB = Phi0->getIncomingBlock(1);
OtherBB = Phi0->getIncomingBlock(0);
} else {
return nullptr;
}
if (!match(Phi1->getIncomingValueForBlock(ConstBB), m_ImmConstant(C1)))
return nullptr;
// The block that we are hoisting to must reach here unconditionally.
// Otherwise, we could be speculatively executing an expensive or
// non-speculative op.
auto *PredBlockBranch = dyn_cast<BranchInst>(OtherBB->getTerminator());
if (!PredBlockBranch || PredBlockBranch->isConditional() ||
!DT.isReachableFromEntry(OtherBB))
return nullptr;
// TODO: This check could be tightened to only apply to binops (div/rem) that
// are not safe to speculatively execute. But that could allow hoisting
// potentially expensive instructions (fdiv for example).
for (auto BBIter = BO.getParent()->begin(); &*BBIter != &BO; ++BBIter)
if (!isGuaranteedToTransferExecutionToSuccessor(&*BBIter))
return nullptr;
// Fold constants for the predecessor block with constant incoming values.
Constant *NewC = ConstantFoldBinaryOpOperands(BO.getOpcode(), C0, C1, DL);
if (!NewC)
return nullptr;
// Make a new binop in the predecessor block with the non-constant incoming
// values.
Builder.SetInsertPoint(PredBlockBranch);
Value *NewBO = Builder.CreateBinOp(BO.getOpcode(),
Phi0->getIncomingValueForBlock(OtherBB),
Phi1->getIncomingValueForBlock(OtherBB));
if (auto *NotFoldedNewBO = dyn_cast<BinaryOperator>(NewBO))
NotFoldedNewBO->copyIRFlags(&BO);
// Replace the binop with a phi of the new values. The old phis are dead.
PHINode *NewPhi = PHINode::Create(BO.getType(), 2);
NewPhi->addIncoming(NewBO, OtherBB);
NewPhi->addIncoming(NewC, ConstBB);
return NewPhi;
}
Instruction *InstCombinerImpl::foldBinOpIntoSelectOrPhi(BinaryOperator &I) {
if (!isa<Constant>(I.getOperand(1)))
return nullptr;
if (auto *Sel = dyn_cast<SelectInst>(I.getOperand(0))) {
if (Instruction *NewSel = FoldOpIntoSelect(I, Sel))
return NewSel;
} else if (auto *PN = dyn_cast<PHINode>(I.getOperand(0))) {
if (Instruction *NewPhi = foldOpIntoPhi(I, PN))
return NewPhi;
}
return nullptr;
}
static bool shouldMergeGEPs(GEPOperator &GEP, GEPOperator &Src) {
// If this GEP has only 0 indices, it is the same pointer as
// Src. If Src is not a trivial GEP too, don't combine
// the indices.
if (GEP.hasAllZeroIndices() && !Src.hasAllZeroIndices() &&
!Src.hasOneUse())
return false;
return true;
}
Instruction *InstCombinerImpl::foldVectorBinop(BinaryOperator &Inst) {
if (!isa<VectorType>(Inst.getType()))
return nullptr;
BinaryOperator::BinaryOps Opcode = Inst.getOpcode();
Value *LHS = Inst.getOperand(0), *RHS = Inst.getOperand(1);
assert(cast<VectorType>(LHS->getType())->getElementCount() ==
cast<VectorType>(Inst.getType())->getElementCount());
assert(cast<VectorType>(RHS->getType())->getElementCount() ==
cast<VectorType>(Inst.getType())->getElementCount());
// If both operands of the binop are vector concatenations, then perform the
// narrow binop on each pair of the source operands followed by concatenation
// of the results.
Value *L0, *L1, *R0, *R1;
ArrayRef<int> Mask;
if (match(LHS, m_Shuffle(m_Value(L0), m_Value(L1), m_Mask(Mask))) &&
match(RHS, m_Shuffle(m_Value(R0), m_Value(R1), m_SpecificMask(Mask))) &&
LHS->hasOneUse() && RHS->hasOneUse() &&
cast<ShuffleVectorInst>(LHS)->isConcat() &&
cast<ShuffleVectorInst>(RHS)->isConcat()) {
// This transform does not have the speculative execution constraint as
// below because the shuffle is a concatenation. The new binops are
// operating on exactly the same elements as the existing binop.
// TODO: We could ease the mask requirement to allow different undef lanes,
// but that requires an analysis of the binop-with-undef output value.
Value *NewBO0 = Builder.CreateBinOp(Opcode, L0, R0);
if (auto *BO = dyn_cast<BinaryOperator>(NewBO0))
BO->copyIRFlags(&Inst);
Value *NewBO1 = Builder.CreateBinOp(Opcode, L1, R1);
if (auto *BO = dyn_cast<BinaryOperator>(NewBO1))
BO->copyIRFlags(&Inst);
return new ShuffleVectorInst(NewBO0, NewBO1, Mask);
}
auto createBinOpReverse = [&](Value *X, Value *Y) {
Value *V = Builder.CreateBinOp(Opcode, X, Y, Inst.getName());
if (auto *BO = dyn_cast<BinaryOperator>(V))
BO->copyIRFlags(&Inst);
Module *M = Inst.getModule();
Function *F = Intrinsic::getDeclaration(
M, Intrinsic::experimental_vector_reverse, V->getType());
return CallInst::Create(F, V);
};
// NOTE: Reverse shuffles don't require the speculative execution protection
// below because they don't affect which lanes take part in the computation.
Value *V1, *V2;
if (match(LHS, m_VecReverse(m_Value(V1)))) {
// Op(rev(V1), rev(V2)) -> rev(Op(V1, V2))
if (match(RHS, m_VecReverse(m_Value(V2))) &&
(LHS->hasOneUse() || RHS->hasOneUse() ||
(LHS == RHS && LHS->hasNUses(2))))
return createBinOpReverse(V1, V2);
// Op(rev(V1), RHSSplat)) -> rev(Op(V1, RHSSplat))
if (LHS->hasOneUse() && isSplatValue(RHS))
return createBinOpReverse(V1, RHS);
}
// Op(LHSSplat, rev(V2)) -> rev(Op(LHSSplat, V2))
else if (isSplatValue(LHS) && match(RHS, m_OneUse(m_VecReverse(m_Value(V2)))))
return createBinOpReverse(LHS, V2);
// It may not be safe to reorder shuffles and things like div, urem, etc.
// because we may trap when executing those ops on unknown vector elements.
// See PR20059.
if (!isSafeToSpeculativelyExecute(&Inst))
return nullptr;
auto createBinOpShuffle = [&](Value *X, Value *Y, ArrayRef<int> M) {
Value *XY = Builder.CreateBinOp(Opcode, X, Y);
if (auto *BO = dyn_cast<BinaryOperator>(XY))
BO->copyIRFlags(&Inst);
return new ShuffleVectorInst(XY, M);
};
// If both arguments of the binary operation are shuffles that use the same
// mask and shuffle within a single vector, move the shuffle after the binop.
if (match(LHS, m_Shuffle(m_Value(V1), m_Poison(), m_Mask(Mask))) &&
match(RHS, m_Shuffle(m_Value(V2), m_Poison(), m_SpecificMask(Mask))) &&
V1->getType() == V2->getType() &&
(LHS->hasOneUse() || RHS->hasOneUse() || LHS == RHS)) {
// Op(shuffle(V1, Mask), shuffle(V2, Mask)) -> shuffle(Op(V1, V2), Mask)
return createBinOpShuffle(V1, V2, Mask);
}
// If both arguments of a commutative binop are select-shuffles that use the
// same mask with commuted operands, the shuffles are unnecessary.
if (Inst.isCommutative() &&
match(LHS, m_Shuffle(m_Value(V1), m_Value(V2), m_Mask(Mask))) &&
match(RHS,
m_Shuffle(m_Specific(V2), m_Specific(V1), m_SpecificMask(Mask)))) {
auto *LShuf = cast<ShuffleVectorInst>(LHS);
auto *RShuf = cast<ShuffleVectorInst>(RHS);
// TODO: Allow shuffles that contain undefs in the mask?
// That is legal, but it reduces undef knowledge.
// TODO: Allow arbitrary shuffles by shuffling after binop?
// That might be legal, but we have to deal with poison.
if (LShuf->isSelect() &&
!is_contained(LShuf->getShuffleMask(), PoisonMaskElem) &&
RShuf->isSelect() &&
!is_contained(RShuf->getShuffleMask(), PoisonMaskElem)) {
// Example:
// LHS = shuffle V1, V2, <0, 5, 6, 3>
// RHS = shuffle V2, V1, <0, 5, 6, 3>
// LHS + RHS --> (V10+V20, V21+V11, V22+V12, V13+V23) --> V1 + V2
Instruction *NewBO = BinaryOperator::Create(Opcode, V1, V2);
NewBO->copyIRFlags(&Inst);
return NewBO;
}
}
// If one argument is a shuffle within one vector and the other is a constant,
// try moving the shuffle after the binary operation. This canonicalization
// intends to move shuffles closer to other shuffles and binops closer to
// other binops, so they can be folded. It may also enable demanded elements
// transforms.
Constant *C;
auto *InstVTy = dyn_cast<FixedVectorType>(Inst.getType());
if (InstVTy &&
match(&Inst, m_c_BinOp(m_OneUse(m_Shuffle(m_Value(V1), m_Poison(),
m_Mask(Mask))),
m_ImmConstant(C))) &&
cast<FixedVectorType>(V1->getType())->getNumElements() <=
InstVTy->getNumElements()) {
assert(InstVTy->getScalarType() == V1->getType()->getScalarType() &&
"Shuffle should not change scalar type");
// Find constant NewC that has property:
// shuffle(NewC, ShMask) = C
// If such constant does not exist (example: ShMask=<0,0> and C=<1,2>)
// reorder is not possible. A 1-to-1 mapping is not required. Example:
// ShMask = <1,1,2,2> and C = <5,5,6,6> --> NewC = <undef,5,6,undef>
bool ConstOp1 = isa<Constant>(RHS);
ArrayRef<int> ShMask = Mask;
unsigned SrcVecNumElts =
cast<FixedVectorType>(V1->getType())->getNumElements();
PoisonValue *PoisonScalar = PoisonValue::get(C->getType()->getScalarType());
SmallVector<Constant *, 16> NewVecC(SrcVecNumElts, PoisonScalar);
bool MayChange = true;
unsigned NumElts = InstVTy->getNumElements();
for (unsigned I = 0; I < NumElts; ++I) {
Constant *CElt = C->getAggregateElement(I);
if (ShMask[I] >= 0) {
assert(ShMask[I] < (int)NumElts && "Not expecting narrowing shuffle");
Constant *NewCElt = NewVecC[ShMask[I]];
// Bail out if:
// 1. The constant vector contains a constant expression.
// 2. The shuffle needs an element of the constant vector that can't
// be mapped to a new constant vector.
// 3. This is a widening shuffle that copies elements of V1 into the
// extended elements (extending with poison is allowed).
if (!CElt || (!isa<PoisonValue>(NewCElt) && NewCElt != CElt) ||
I >= SrcVecNumElts) {
MayChange = false;
break;
}
NewVecC[ShMask[I]] = CElt;
}
// If this is a widening shuffle, we must be able to extend with poison
// elements. If the original binop does not produce a poison in the high
// lanes, then this transform is not safe.
// Similarly for poison lanes due to the shuffle mask, we can only
// transform binops that preserve poison.
// TODO: We could shuffle those non-poison constant values into the
// result by using a constant vector (rather than an poison vector)
// as operand 1 of the new binop, but that might be too aggressive
// for target-independent shuffle creation.
if (I >= SrcVecNumElts || ShMask[I] < 0) {
Constant *MaybePoison =
ConstOp1
? ConstantFoldBinaryOpOperands(Opcode, PoisonScalar, CElt, DL)
: ConstantFoldBinaryOpOperands(Opcode, CElt, PoisonScalar, DL);
if (!MaybePoison || !isa<PoisonValue>(MaybePoison)) {
MayChange = false;
break;
}
}
}
if (MayChange) {
Constant *NewC = ConstantVector::get(NewVecC);
// It may not be safe to execute a binop on a vector with poison elements
// because the entire instruction can be folded to undef or create poison
// that did not exist in the original code.
// TODO: The shift case should not be necessary.
if (Inst.isIntDivRem() || (Inst.isShift() && ConstOp1))
NewC = getSafeVectorConstantForBinop(Opcode, NewC, ConstOp1);
// Op(shuffle(V1, Mask), C) -> shuffle(Op(V1, NewC), Mask)
// Op(C, shuffle(V1, Mask)) -> shuffle(Op(NewC, V1), Mask)
Value *NewLHS = ConstOp1 ? V1 : NewC;
Value *NewRHS = ConstOp1 ? NewC : V1;
return createBinOpShuffle(NewLHS, NewRHS, Mask);
}
}
// Try to reassociate to sink a splat shuffle after a binary operation.
if (Inst.isAssociative() && Inst.isCommutative()) {
// Canonicalize shuffle operand as LHS.
if (isa<ShuffleVectorInst>(RHS))
std::swap(LHS, RHS);
Value *X;
ArrayRef<int> MaskC;
int SplatIndex;
Value *Y, *OtherOp;
if (!match(LHS,
m_OneUse(m_Shuffle(m_Value(X), m_Undef(), m_Mask(MaskC)))) ||
!match(MaskC, m_SplatOrUndefMask(SplatIndex)) ||
X->getType() != Inst.getType() ||
!match(RHS, m_OneUse(m_BinOp(Opcode, m_Value(Y), m_Value(OtherOp)))))
return nullptr;
// FIXME: This may not be safe if the analysis allows undef elements. By
// moving 'Y' before the splat shuffle, we are implicitly assuming
// that it is not undef/poison at the splat index.
if (isSplatValue(OtherOp, SplatIndex)) {
std::swap(Y, OtherOp);
} else if (!isSplatValue(Y, SplatIndex)) {
return nullptr;
}
// X and Y are splatted values, so perform the binary operation on those
// values followed by a splat followed by the 2nd binary operation:
// bo (splat X), (bo Y, OtherOp) --> bo (splat (bo X, Y)), OtherOp
Value *NewBO = Builder.CreateBinOp(Opcode, X, Y);
SmallVector<int, 8> NewMask(MaskC.size(), SplatIndex);
Value *NewSplat = Builder.CreateShuffleVector(NewBO, NewMask);
Instruction *R = BinaryOperator::Create(Opcode, NewSplat, OtherOp);
// Intersect FMF on both new binops. Other (poison-generating) flags are
// dropped to be safe.
if (isa<FPMathOperator>(R)) {
R->copyFastMathFlags(&Inst);
R->andIRFlags(RHS);
}
if (auto *NewInstBO = dyn_cast<BinaryOperator>(NewBO))
NewInstBO->copyIRFlags(R);
return R;
}
return nullptr;
}
/// Try to narrow the width of a binop if at least 1 operand is an extend of
/// of a value. This requires a potentially expensive known bits check to make
/// sure the narrow op does not overflow.
Instruction *InstCombinerImpl::narrowMathIfNoOverflow(BinaryOperator &BO) {
// We need at least one extended operand.
Value *Op0 = BO.getOperand(0), *Op1 = BO.getOperand(1);
// If this is a sub, we swap the operands since we always want an extension
// on the RHS. The LHS can be an extension or a constant.
if (BO.getOpcode() == Instruction::Sub)
std::swap(Op0, Op1);
Value *X;
bool IsSext = match(Op0, m_SExt(m_Value(X)));
if (!IsSext && !match(Op0, m_ZExt(m_Value(X))))
return nullptr;
// If both operands are the same extension from the same source type and we
// can eliminate at least one (hasOneUse), this might work.
CastInst::CastOps CastOpc = IsSext ? Instruction::SExt : Instruction::ZExt;
Value *Y;
if (!(match(Op1, m_ZExtOrSExt(m_Value(Y))) && X->getType() == Y->getType() &&
cast<Operator>(Op1)->getOpcode() == CastOpc &&
(Op0->hasOneUse() || Op1->hasOneUse()))) {
// If that did not match, see if we have a suitable constant operand.
// Truncating and extending must produce the same constant.
Constant *WideC;
if (!Op0->hasOneUse() || !match(Op1, m_Constant(WideC)))
return nullptr;
Constant *NarrowC = getLosslessTrunc(WideC, X->getType(), CastOpc);
if (!NarrowC)
return nullptr;
Y = NarrowC;
}
// Swap back now that we found our operands.
if (BO.getOpcode() == Instruction::Sub)
std::swap(X, Y);
// Both operands have narrow versions. Last step: the math must not overflow
// in the narrow width.
if (!willNotOverflow(BO.getOpcode(), X, Y, BO, IsSext))
return nullptr;
// bo (ext X), (ext Y) --> ext (bo X, Y)
// bo (ext X), C --> ext (bo X, C')
Value *NarrowBO = Builder.CreateBinOp(BO.getOpcode(), X, Y, "narrow");
if (auto *NewBinOp = dyn_cast<BinaryOperator>(NarrowBO)) {
if (IsSext)
NewBinOp->setHasNoSignedWrap();
else
NewBinOp->setHasNoUnsignedWrap();
}
return CastInst::Create(CastOpc, NarrowBO, BO.getType());
}
static bool isMergedGEPInBounds(GEPOperator &GEP1, GEPOperator &GEP2) {
// At least one GEP must be inbounds.
if (!GEP1.isInBounds() && !GEP2.isInBounds())
return false;
return (GEP1.isInBounds() || GEP1.hasAllZeroIndices()) &&
(GEP2.isInBounds() || GEP2.hasAllZeroIndices());
}
/// Thread a GEP operation with constant indices through the constant true/false
/// arms of a select.
static Instruction *foldSelectGEP(GetElementPtrInst &GEP,
InstCombiner::BuilderTy &Builder) {
if (!GEP.hasAllConstantIndices())
return nullptr;
Instruction *Sel;
Value *Cond;
Constant *TrueC, *FalseC;
if (!match(GEP.getPointerOperand(), m_Instruction(Sel)) ||
!match(Sel,
m_Select(m_Value(Cond), m_Constant(TrueC), m_Constant(FalseC))))
return nullptr;
// gep (select Cond, TrueC, FalseC), IndexC --> select Cond, TrueC', FalseC'
// Propagate 'inbounds' and metadata from existing instructions.
// Note: using IRBuilder to create the constants for efficiency.
SmallVector<Value *, 4> IndexC(GEP.indices());
bool IsInBounds = GEP.isInBounds();
Type *Ty = GEP.getSourceElementType();
Value *NewTrueC = Builder.CreateGEP(Ty, TrueC, IndexC, "", IsInBounds);
Value *NewFalseC = Builder.CreateGEP(Ty, FalseC, IndexC, "", IsInBounds);
return SelectInst::Create(Cond, NewTrueC, NewFalseC, "", nullptr, Sel);
}
Instruction *InstCombinerImpl::visitGEPOfGEP(GetElementPtrInst &GEP,
GEPOperator *Src) {
// Combine Indices - If the source pointer to this getelementptr instruction
// is a getelementptr instruction with matching element type, combine the
// indices of the two getelementptr instructions into a single instruction.
if (!shouldMergeGEPs(*cast<GEPOperator>(&GEP), *Src))
return nullptr;
// For constant GEPs, use a more general offset-based folding approach.
Type *PtrTy = Src->getType()->getScalarType();
if (GEP.hasAllConstantIndices() &&
(Src->hasOneUse() || Src->hasAllConstantIndices())) {
// Split Src into a variable part and a constant suffix.
gep_type_iterator GTI = gep_type_begin(*Src);
Type *BaseType = GTI.getIndexedType();
bool IsFirstType = true;
unsigned NumVarIndices = 0;
for (auto Pair : enumerate(Src->indices())) {
if (!isa<ConstantInt>(Pair.value())) {
BaseType = GTI.getIndexedType();
IsFirstType = false;
NumVarIndices = Pair.index() + 1;
}
++GTI;
}
// Determine the offset for the constant suffix of Src.
APInt Offset(DL.getIndexTypeSizeInBits(PtrTy), 0);
if (NumVarIndices != Src->getNumIndices()) {
// FIXME: getIndexedOffsetInType() does not handled scalable vectors.
if (BaseType->isScalableTy())
return nullptr;
SmallVector<Value *> ConstantIndices;
if (!IsFirstType)
ConstantIndices.push_back(
Constant::getNullValue(Type::getInt32Ty(GEP.getContext())));
append_range(ConstantIndices, drop_begin(Src->indices(), NumVarIndices));
Offset += DL.getIndexedOffsetInType(BaseType, ConstantIndices);
}
// Add the offset for GEP (which is fully constant).
if (!GEP.accumulateConstantOffset(DL, Offset))
return nullptr;
APInt OffsetOld = Offset;
// Convert the total offset back into indices.
SmallVector<APInt> ConstIndices =
DL.getGEPIndicesForOffset(BaseType, Offset);
if (!Offset.isZero() || (!IsFirstType && !ConstIndices[0].isZero())) {
// If both GEP are constant-indexed, and cannot be merged in either way,
// convert them to a GEP of i8.
if (Src->hasAllConstantIndices())
return replaceInstUsesWith(
GEP, Builder.CreateGEP(
Builder.getInt8Ty(), Src->getOperand(0),
Builder.getInt(OffsetOld), "",
isMergedGEPInBounds(*Src, *cast<GEPOperator>(&GEP))));
return nullptr;
}
bool IsInBounds = isMergedGEPInBounds(*Src, *cast<GEPOperator>(&GEP));
SmallVector<Value *> Indices;
append_range(Indices, drop_end(Src->indices(),
Src->getNumIndices() - NumVarIndices));
for (const APInt &Idx : drop_begin(ConstIndices, !IsFirstType)) {
Indices.push_back(ConstantInt::get(GEP.getContext(), Idx));
// Even if the total offset is inbounds, we may end up representing it
// by first performing a larger negative offset, and then a smaller
// positive one. The large negative offset might go out of bounds. Only
// preserve inbounds if all signs are the same.
IsInBounds &= Idx.isNonNegative() == ConstIndices[0].isNonNegative();
}
return replaceInstUsesWith(
GEP, Builder.CreateGEP(Src->getSourceElementType(), Src->getOperand(0),
Indices, "", IsInBounds));
}
if (Src->getResultElementType() != GEP.getSourceElementType())
return nullptr;
SmallVector<Value*, 8> Indices;
// Find out whether the last index in the source GEP is a sequential idx.
bool EndsWithSequential = false;
for (gep_type_iterator I = gep_type_begin(*Src), E = gep_type_end(*Src);
I != E; ++I)
EndsWithSequential = I.isSequential();
// Can we combine the two pointer arithmetics offsets?
if (EndsWithSequential) {
// Replace: gep (gep %P, long B), long A, ...
// With: T = long A+B; gep %P, T, ...
Value *SO1 = Src->getOperand(Src->getNumOperands()-1);
Value *GO1 = GEP.getOperand(1);
// If they aren't the same type, then the input hasn't been processed
// by the loop above yet (which canonicalizes sequential index types to
// intptr_t). Just avoid transforming this until the input has been
// normalized.
if (SO1->getType() != GO1->getType())
return nullptr;
Value *Sum =
simplifyAddInst(GO1, SO1, false, false, SQ.getWithInstruction(&GEP));
// Only do the combine when we are sure the cost after the
// merge is never more than that before the merge.
if (Sum == nullptr)
return nullptr;
// Update the GEP in place if possible.
if (Src->getNumOperands() == 2) {
GEP.setIsInBounds(isMergedGEPInBounds(*Src, *cast<GEPOperator>(&GEP)));
replaceOperand(GEP, 0, Src->getOperand(0));
replaceOperand(GEP, 1, Sum);
return &GEP;
}
Indices.append(Src->op_begin()+1, Src->op_end()-1);
Indices.push_back(Sum);
Indices.append(GEP.op_begin()+2, GEP.op_end());
} else if (isa<Constant>(*GEP.idx_begin()) &&
cast<Constant>(*GEP.idx_begin())->isNullValue() &&
Src->getNumOperands() != 1) {
// Otherwise we can do the fold if the first index of the GEP is a zero
Indices.append(Src->op_begin()+1, Src->op_end());
Indices.append(GEP.idx_begin()+1, GEP.idx_end());
}
if (!Indices.empty())
return replaceInstUsesWith(
GEP, Builder.CreateGEP(
Src->getSourceElementType(), Src->getOperand(0), Indices, "",
isMergedGEPInBounds(*Src, *cast<GEPOperator>(&GEP))));
return nullptr;
}
Value *InstCombiner::getFreelyInvertedImpl(Value *V, bool WillInvertAllUses,
BuilderTy *Builder,
bool &DoesConsume, unsigned Depth) {
static Value *const NonNull = reinterpret_cast<Value *>(uintptr_t(1));
// ~(~(X)) -> X.
Value *A, *B;
if (match(V, m_Not(m_Value(A)))) {
DoesConsume = true;
return A;
}
Constant *C;
// Constants can be considered to be not'ed values.
if (match(V, m_ImmConstant(C)))
return ConstantExpr::getNot(C);
if (Depth++ >= MaxAnalysisRecursionDepth)
return nullptr;
// The rest of the cases require that we invert all uses so don't bother
// doing the analysis if we know we can't use the result.
if (!WillInvertAllUses)
return nullptr;
// Compares can be inverted if all of their uses are being modified to use
// the ~V.
if (auto *I = dyn_cast<CmpInst>(V)) {
if (Builder != nullptr)
return Builder->CreateCmp(I->getInversePredicate(), I->getOperand(0),
I->getOperand(1));
return NonNull;
}
// If `V` is of the form `A + B` then `-1 - V` can be folded into
// `(-1 - B) - A` if we are willing to invert all of the uses.
if (match(V, m_Add(m_Value(A), m_Value(B)))) {
if (auto *BV = getFreelyInvertedImpl(B, B->hasOneUse(), Builder,
DoesConsume, Depth))
return Builder ? Builder->CreateSub(BV, A) : NonNull;
if (auto *AV = getFreelyInvertedImpl(A, A->hasOneUse(), Builder,
DoesConsume, Depth))
return Builder ? Builder->CreateSub(AV, B) : NonNull;
return nullptr;
}
// If `V` is of the form `A ^ ~B` then `~(A ^ ~B)` can be folded
// into `A ^ B` if we are willing to invert all of the uses.
if (match(V, m_Xor(m_Value(A), m_Value(B)))) {
if (auto *BV = getFreelyInvertedImpl(B, B->hasOneUse(), Builder,
DoesConsume, Depth))
return Builder ? Builder->CreateXor(A, BV) : NonNull;
if (auto *AV = getFreelyInvertedImpl(A, A->hasOneUse(), Builder,
DoesConsume, Depth))
return Builder ? Builder->CreateXor(AV, B) : NonNull;
return nullptr;
}
// If `V` is of the form `B - A` then `-1 - V` can be folded into
// `A + (-1 - B)` if we are willing to invert all of the uses.
if (match(V, m_Sub(m_Value(A), m_Value(B)))) {
if (auto *AV = getFreelyInvertedImpl(A, A->hasOneUse(), Builder,
DoesConsume, Depth))
return Builder ? Builder->CreateAdd(AV, B) : NonNull;
return nullptr;
}
// If `V` is of the form `(~A) s>> B` then `~((~A) s>> B)` can be folded
// into `A s>> B` if we are willing to invert all of the uses.
if (match(V, m_AShr(m_Value(A), m_Value(B)))) {
if (auto *AV = getFreelyInvertedImpl(A, A->hasOneUse(), Builder,
DoesConsume, Depth))
return Builder ? Builder->CreateAShr(AV, B) : NonNull;
return nullptr;
}
Value *Cond;
// LogicOps are special in that we canonicalize them at the cost of an
// instruction.
bool IsSelect = match(V, m_Select(m_Value(Cond), m_Value(A), m_Value(B))) &&
!shouldAvoidAbsorbingNotIntoSelect(*cast<SelectInst>(V));
// Selects/min/max with invertible operands are freely invertible
if (IsSelect || match(V, m_MaxOrMin(m_Value(A), m_Value(B)))) {
if (!getFreelyInvertedImpl(B, B->hasOneUse(), /*Builder*/ nullptr,
DoesConsume, Depth))
return nullptr;
if (Value *NotA = getFreelyInvertedImpl(A, A->hasOneUse(), Builder,
DoesConsume, Depth)) {
if (Builder != nullptr) {
Value *NotB = getFreelyInvertedImpl(B, B->hasOneUse(), Builder,
DoesConsume, Depth);
assert(NotB != nullptr &&
"Unable to build inverted value for known freely invertable op");
if (auto *II = dyn_cast<IntrinsicInst>(V))
return Builder->CreateBinaryIntrinsic(
getInverseMinMaxIntrinsic(II->getIntrinsicID()), NotA, NotB);
return Builder->CreateSelect(Cond, NotA, NotB);
}
return NonNull;
}
}
return nullptr;
}
Instruction *InstCombinerImpl::visitGetElementPtrInst(GetElementPtrInst &GEP) {
Value *PtrOp = GEP.getOperand(0);
SmallVector<Value *, 8> Indices(GEP.indices());
Type *GEPType = GEP.getType();
Type *GEPEltType = GEP.getSourceElementType();
bool IsGEPSrcEleScalable = GEPEltType->isScalableTy();
if (Value *V = simplifyGEPInst(GEPEltType, PtrOp, Indices, GEP.isInBounds(),
SQ.getWithInstruction(&GEP)))
return replaceInstUsesWith(GEP, V);
// For vector geps, use the generic demanded vector support.
// Skip if GEP return type is scalable. The number of elements is unknown at
// compile-time.
if (auto *GEPFVTy = dyn_cast<FixedVectorType>(GEPType)) {
auto VWidth = GEPFVTy->getNumElements();
APInt PoisonElts(VWidth, 0);
APInt AllOnesEltMask(APInt::getAllOnes(VWidth));
if (Value *V = SimplifyDemandedVectorElts(&GEP, AllOnesEltMask,
PoisonElts)) {
if (V != &GEP)
return replaceInstUsesWith(GEP, V);
return &GEP;
}
// TODO: 1) Scalarize splat operands, 2) scalarize entire instruction if
// possible (decide on canonical form for pointer broadcast), 3) exploit
// undef elements to decrease demanded bits
}
// Eliminate unneeded casts for indices, and replace indices which displace
// by multiples of a zero size type with zero.
bool MadeChange = false;
// Index width may not be the same width as pointer width.
// Data layout chooses the right type based on supported integer types.
Type *NewScalarIndexTy =
DL.getIndexType(GEP.getPointerOperandType()->getScalarType());
gep_type_iterator GTI = gep_type_begin(GEP);
for (User::op_iterator I = GEP.op_begin() + 1, E = GEP.op_end(); I != E;
++I, ++GTI) {
// Skip indices into struct types.
if (GTI.isStruct())
continue;
Type *IndexTy = (*I)->getType();
Type *NewIndexType =
IndexTy->isVectorTy()
? VectorType::get(NewScalarIndexTy,
cast<VectorType>(IndexTy)->getElementCount())
: NewScalarIndexTy;
// If the element type has zero size then any index over it is equivalent
// to an index of zero, so replace it with zero if it is not zero already.
Type *EltTy = GTI.getIndexedType();
if (EltTy->isSized() && DL.getTypeAllocSize(EltTy).isZero())
if (!isa<Constant>(*I) || !match(I->get(), m_Zero())) {
*I = Constant::getNullValue(NewIndexType);
MadeChange = true;
}
if (IndexTy != NewIndexType) {
// If we are using a wider index than needed for this platform, shrink
// it to what we need. If narrower, sign-extend it to what we need.
// This explicit cast can make subsequent optimizations more obvious.
*I = Builder.CreateIntCast(*I, NewIndexType, true);
MadeChange = true;
}
}
if (MadeChange)
return &GEP;
// Check to see if the inputs to the PHI node are getelementptr instructions.
if (auto *PN = dyn_cast<PHINode>(PtrOp)) {
auto *Op1 = dyn_cast<GetElementPtrInst>(PN->getOperand(0));
if (!Op1)
return nullptr;
// Don't fold a GEP into itself through a PHI node. This can only happen
// through the back-edge of a loop. Folding a GEP into itself means that
// the value of the previous iteration needs to be stored in the meantime,
// thus requiring an additional register variable to be live, but not
// actually achieving anything (the GEP still needs to be executed once per
// loop iteration).
if (Op1 == &GEP)
return nullptr;
int DI = -1;
for (auto I = PN->op_begin()+1, E = PN->op_end(); I !=E; ++I) {
auto *Op2 = dyn_cast<GetElementPtrInst>(*I);
if (!Op2 || Op1->getNumOperands() != Op2->getNumOperands() ||
Op1->getSourceElementType() != Op2->getSourceElementType())
return nullptr;
// As for Op1 above, don't try to fold a GEP into itself.
if (Op2 == &GEP)
return nullptr;
// Keep track of the type as we walk the GEP.
Type *CurTy = nullptr;
for (unsigned J = 0, F = Op1->getNumOperands(); J != F; ++J) {
if (Op1->getOperand(J)->getType() != Op2->getOperand(J)->getType())
return nullptr;
if (Op1->getOperand(J) != Op2->getOperand(J)) {
if (DI == -1) {
// We have not seen any differences yet in the GEPs feeding the
// PHI yet, so we record this one if it is allowed to be a
// variable.
// The first two arguments can vary for any GEP, the rest have to be
// static for struct slots
if (J > 1) {
assert(CurTy && "No current type?");
if (CurTy->isStructTy())
return nullptr;
}
DI = J;
} else {
// The GEP is different by more than one input. While this could be
// extended to support GEPs that vary by more than one variable it
// doesn't make sense since it greatly increases the complexity and
// would result in an R+R+R addressing mode which no backend
// directly supports and would need to be broken into several
// simpler instructions anyway.
return nullptr;
}
}
// Sink down a layer of the type for the next iteration.
if (J > 0) {
if (J == 1) {
CurTy = Op1->getSourceElementType();
} else {
CurTy =
GetElementPtrInst::getTypeAtIndex(CurTy, Op1->getOperand(J));
}
}
}
}
// If not all GEPs are identical we'll have to create a new PHI node.
// Check that the old PHI node has only one use so that it will get
// removed.
if (DI != -1 && !PN->hasOneUse())
return nullptr;
auto *NewGEP = cast<GetElementPtrInst>(Op1->clone());
if (DI == -1) {
// All the GEPs feeding the PHI are identical. Clone one down into our
// BB so that it can be merged with the current GEP.
} else {
// All the GEPs feeding the PHI differ at a single offset. Clone a GEP
// into the current block so it can be merged, and create a new PHI to
// set that index.
PHINode *NewPN;
{
IRBuilderBase::InsertPointGuard Guard(Builder);
Builder.SetInsertPoint(PN);
NewPN = Builder.CreatePHI(Op1->getOperand(DI)->getType(),
PN->getNumOperands());
}
for (auto &I : PN->operands())
NewPN->addIncoming(cast<GEPOperator>(I)->getOperand(DI),
PN->getIncomingBlock(I));
NewGEP->setOperand(DI, NewPN);
}
NewGEP->insertBefore(*GEP.getParent(), GEP.getParent()->getFirstInsertionPt());
return replaceOperand(GEP, 0, NewGEP);
}
if (auto *Src = dyn_cast<GEPOperator>(PtrOp))
if (Instruction *I = visitGEPOfGEP(GEP, Src))
return I;
// Skip if GEP source element type is scalable. The type alloc size is unknown
// at compile-time.
if (GEP.getNumIndices() == 1 && !IsGEPSrcEleScalable) {
unsigned AS = GEP.getPointerAddressSpace();
if (GEP.getOperand(1)->getType()->getScalarSizeInBits() ==
DL.getIndexSizeInBits(AS)) {
uint64_t TyAllocSize = DL.getTypeAllocSize(GEPEltType).getFixedValue();
if (TyAllocSize == 1) {
// Canonicalize (gep i8* X, (ptrtoint Y)-(ptrtoint X)) to (bitcast Y),
// but only if the result pointer is only used as if it were an integer,
// or both point to the same underlying object (otherwise provenance is
// not necessarily retained).
Value *X = GEP.getPointerOperand();
Value *Y;
if (match(GEP.getOperand(1),
m_Sub(m_PtrToInt(m_Value(Y)), m_PtrToInt(m_Specific(X)))) &&
GEPType == Y->getType()) {
bool HasSameUnderlyingObject =
getUnderlyingObject(X) == getUnderlyingObject(Y);
bool Changed = false;
GEP.replaceUsesWithIf(Y, [&](Use &U) {
bool ShouldReplace = HasSameUnderlyingObject ||
isa<ICmpInst>(U.getUser()) ||
isa<PtrToIntInst>(U.getUser());
Changed |= ShouldReplace;
return ShouldReplace;
});
return Changed ? &GEP : nullptr;
}
} else {
// Canonicalize (gep T* X, V / sizeof(T)) to (gep i8* X, V)
Value *V;
if ((has_single_bit(TyAllocSize) &&
match(GEP.getOperand(1),
m_Exact(m_AShr(m_Value(V),
m_SpecificInt(countr_zero(TyAllocSize)))))) ||
match(GEP.getOperand(1),
m_Exact(m_SDiv(m_Value(V), m_SpecificInt(TyAllocSize))))) {
GetElementPtrInst *NewGEP = GetElementPtrInst::Create(
Builder.getInt8Ty(), GEP.getPointerOperand(), V);
NewGEP->setIsInBounds(GEP.isInBounds());
return NewGEP;
}
}
}
}
// We do not handle pointer-vector geps here.
if (GEPType->isVectorTy())
return nullptr;
if (GEP.getNumIndices() == 1) {
// Try to replace ADD + GEP with GEP + GEP.
Value *Idx1, *Idx2;
if (match(GEP.getOperand(1),
m_OneUse(m_Add(m_Value(Idx1), m_Value(Idx2))))) {
// %idx = add i64 %idx1, %idx2
// %gep = getelementptr i32, ptr %ptr, i64 %idx
// as:
// %newptr = getelementptr i32, ptr %ptr, i64 %idx1
// %newgep = getelementptr i32, ptr %newptr, i64 %idx2
auto *NewPtr = Builder.CreateGEP(GEP.getResultElementType(),
GEP.getPointerOperand(), Idx1);
return GetElementPtrInst::Create(GEP.getResultElementType(), NewPtr,
Idx2);
}
ConstantInt *C;
if (match(GEP.getOperand(1), m_OneUse(m_SExtLike(m_OneUse(m_NSWAdd(
m_Value(Idx1), m_ConstantInt(C))))))) {
// %add = add nsw i32 %idx1, idx2
// %sidx = sext i32 %add to i64
// %gep = getelementptr i32, ptr %ptr, i64 %sidx
// as:
// %newptr = getelementptr i32, ptr %ptr, i32 %idx1
// %newgep = getelementptr i32, ptr %newptr, i32 idx2
auto *NewPtr = Builder.CreateGEP(
GEP.getResultElementType(), GEP.getPointerOperand(),
Builder.CreateSExt(Idx1, GEP.getOperand(1)->getType()));
return GetElementPtrInst::Create(
GEP.getResultElementType(), NewPtr,
Builder.CreateSExt(C, GEP.getOperand(1)->getType()));
}
}
if (!GEP.isInBounds()) {
unsigned IdxWidth =
DL.getIndexSizeInBits(PtrOp->getType()->getPointerAddressSpace());
APInt BasePtrOffset(IdxWidth, 0);
Value *UnderlyingPtrOp =
PtrOp->stripAndAccumulateInBoundsConstantOffsets(DL,
BasePtrOffset);
bool CanBeNull, CanBeFreed;
uint64_t DerefBytes = UnderlyingPtrOp->getPointerDereferenceableBytes(
DL, CanBeNull, CanBeFreed);
if (!CanBeNull && !CanBeFreed && DerefBytes != 0) {
if (GEP.accumulateConstantOffset(DL, BasePtrOffset) &&
BasePtrOffset.isNonNegative()) {
APInt AllocSize(IdxWidth, DerefBytes);
if (BasePtrOffset.ule(AllocSize)) {
return GetElementPtrInst::CreateInBounds(
GEP.getSourceElementType(), PtrOp, Indices, GEP.getName());
}
}
}
}
if (Instruction *R = foldSelectGEP(GEP, Builder))
return R;
return nullptr;
}
static bool isNeverEqualToUnescapedAlloc(Value *V, const TargetLibraryInfo &TLI,
Instruction *AI) {
if (isa<ConstantPointerNull>(V))
return true;
if (auto *LI = dyn_cast<LoadInst>(V))
return isa<GlobalVariable>(LI->getPointerOperand());
// Two distinct allocations will never be equal.
return isAllocLikeFn(V, &TLI) && V != AI;
}
/// Given a call CB which uses an address UsedV, return true if we can prove the
/// call's only possible effect is storing to V.
static bool isRemovableWrite(CallBase &CB, Value *UsedV,
const TargetLibraryInfo &TLI) {
if (!CB.use_empty())
// TODO: add recursion if returned attribute is present
return false;
if (CB.isTerminator())
// TODO: remove implementation restriction
return false;
if (!CB.willReturn() || !CB.doesNotThrow())
return false;
// If the only possible side effect of the call is writing to the alloca,
// and the result isn't used, we can safely remove any reads implied by the
// call including those which might read the alloca itself.
std::optional<MemoryLocation> Dest = MemoryLocation::getForDest(&CB, TLI);
return Dest && Dest->Ptr == UsedV;
}
static bool isAllocSiteRemovable(Instruction *AI,
SmallVectorImpl<WeakTrackingVH> &Users,
const TargetLibraryInfo &TLI) {
SmallVector<Instruction*, 4> Worklist;
const std::optional<StringRef> Family = getAllocationFamily(AI, &TLI);
Worklist.push_back(AI);
do {
Instruction *PI = Worklist.pop_back_val();
for (User *U : PI->users()) {
Instruction *I = cast<Instruction>(U);
switch (I->getOpcode()) {
default:
// Give up the moment we see something we can't handle.
return false;
case Instruction::AddrSpaceCast:
case Instruction::BitCast:
case Instruction::GetElementPtr:
Users.emplace_back(I);
Worklist.push_back(I);
continue;
case Instruction::ICmp: {
ICmpInst *ICI = cast<ICmpInst>(I);
// We can fold eq/ne comparisons with null to false/true, respectively.
// We also fold comparisons in some conditions provided the alloc has
// not escaped (see isNeverEqualToUnescapedAlloc).
if (!ICI->isEquality())
return false;
unsigned OtherIndex = (ICI->getOperand(0) == PI) ? 1 : 0;
if (!isNeverEqualToUnescapedAlloc(ICI->getOperand(OtherIndex), TLI, AI))
return false;
// Do not fold compares to aligned_alloc calls, as they may have to
// return null in case the required alignment cannot be satisfied,
// unless we can prove that both alignment and size are valid.
auto AlignmentAndSizeKnownValid = [](CallBase *CB) {
// Check if alignment and size of a call to aligned_alloc is valid,
// that is alignment is a power-of-2 and the size is a multiple of the
// alignment.
const APInt *Alignment;
const APInt *Size;
return match(CB->getArgOperand(0), m_APInt(Alignment)) &&
match(CB->getArgOperand(1), m_APInt(Size)) &&
Alignment->isPowerOf2() && Size->urem(*Alignment).isZero();
};
auto *CB = dyn_cast<CallBase>(AI);
LibFunc TheLibFunc;
if (CB && TLI.getLibFunc(*CB->getCalledFunction(), TheLibFunc) &&
TLI.has(TheLibFunc) && TheLibFunc == LibFunc_aligned_alloc &&
!AlignmentAndSizeKnownValid(CB))
return false;
Users.emplace_back(I);
continue;
}
case Instruction::Call:
// Ignore no-op and store intrinsics.
if (IntrinsicInst *II = dyn_cast<IntrinsicInst>(I)) {
switch (II->getIntrinsicID()) {
default:
return false;
case Intrinsic::memmove:
case Intrinsic::memcpy:
case Intrinsic::memset: {
MemIntrinsic *MI = cast<MemIntrinsic>(II);
if (MI->isVolatile() || MI->getRawDest() != PI)
return false;
[[fallthrough]];
}
case Intrinsic::assume:
case Intrinsic::invariant_start:
case Intrinsic::invariant_end:
case Intrinsic::lifetime_start:
case Intrinsic::lifetime_end:
case Intrinsic::objectsize:
Users.emplace_back(I);
continue;
case Intrinsic::launder_invariant_group:
case Intrinsic::strip_invariant_group:
Users.emplace_back(I);
Worklist.push_back(I);
continue;
}
}
if (isRemovableWrite(*cast<CallBase>(I), PI, TLI)) {
Users.emplace_back(I);
continue;
}
if (getFreedOperand(cast<CallBase>(I), &TLI) == PI &&
getAllocationFamily(I, &TLI) == Family) {
assert(Family);
Users.emplace_back(I);
continue;
}
if (getReallocatedOperand(cast<CallBase>(I)) == PI &&
getAllocationFamily(I, &TLI) == Family) {
assert(Family);
Users.emplace_back(I);
Worklist.push_back(I);
continue;
}
return false;
case Instruction::Store: {
StoreInst *SI = cast<StoreInst>(I);
if (SI->isVolatile() || SI->getPointerOperand() != PI)
return false;
Users.emplace_back(I);
continue;
}
}
llvm_unreachable("missing a return?");
}
} while (!Worklist.empty());
return true;
}
Instruction *InstCombinerImpl::visitAllocSite(Instruction &MI) {
assert(isa<AllocaInst>(MI) || isRemovableAlloc(&cast<CallBase>(MI), &TLI));
// If we have a malloc call which is only used in any amount of comparisons to
// null and free calls, delete the calls and replace the comparisons with true
// or false as appropriate.
// This is based on the principle that we can substitute our own allocation
// function (which will never return null) rather than knowledge of the
// specific function being called. In some sense this can change the permitted
// outputs of a program (when we convert a malloc to an alloca, the fact that
// the allocation is now on the stack is potentially visible, for example),
// but we believe in a permissible manner.
SmallVector<WeakTrackingVH, 64> Users;
// If we are removing an alloca with a dbg.declare, insert dbg.value calls
// before each store.
SmallVector<DbgVariableIntrinsic *, 8> DVIs;
SmallVector<DPValue *, 8> DPVs;
std::unique_ptr<DIBuilder> DIB;
if (isa<AllocaInst>(MI)) {
findDbgUsers(DVIs, &MI, &DPVs);
DIB.reset(new DIBuilder(*MI.getModule(), /*AllowUnresolved=*/false));
}
if (isAllocSiteRemovable(&MI, Users, TLI)) {
for (unsigned i = 0, e = Users.size(); i != e; ++i) {
// Lowering all @llvm.objectsize calls first because they may
// use a bitcast/GEP of the alloca we are removing.
if (!Users[i])
continue;
Instruction *I = cast<Instruction>(&*Users[i]);
if (IntrinsicInst *II = dyn_cast<IntrinsicInst>(I)) {
if (II->getIntrinsicID() == Intrinsic::objectsize) {
SmallVector<Instruction *> InsertedInstructions;
Value *Result = lowerObjectSizeCall(
II, DL, &TLI, AA, /*MustSucceed=*/true, &InsertedInstructions);
for (Instruction *Inserted : InsertedInstructions)
Worklist.add(Inserted);
replaceInstUsesWith(*I, Result);
eraseInstFromFunction(*I);
Users[i] = nullptr; // Skip examining in the next loop.
}
}
}
for (unsigned i = 0, e = Users.size(); i != e; ++i) {
if (!Users[i])
continue;
Instruction *I = cast<Instruction>(&*Users[i]);
if (ICmpInst *C = dyn_cast<ICmpInst>(I)) {
replaceInstUsesWith(*C,
ConstantInt::get(Type::getInt1Ty(C->getContext()),
C->isFalseWhenEqual()));
} else if (auto *SI = dyn_cast<StoreInst>(I)) {
for (auto *DVI : DVIs)
if (DVI->isAddressOfVariable())
ConvertDebugDeclareToDebugValue(DVI, SI, *DIB);
for (auto *DPV : DPVs)
if (DPV->isAddressOfVariable())
ConvertDebugDeclareToDebugValue(DPV, SI, *DIB);
} else {
// Casts, GEP, or anything else: we're about to delete this instruction,
// so it can not have any valid uses.
replaceInstUsesWith(*I, PoisonValue::get(I->getType()));
}
eraseInstFromFunction(*I);
}
if (InvokeInst *II = dyn_cast<InvokeInst>(&MI)) {
// Replace invoke with a NOP intrinsic to maintain the original CFG
Module *M = II->getModule();
Function *F = Intrinsic::getDeclaration(M, Intrinsic::donothing);
InvokeInst::Create(F, II->getNormalDest(), II->getUnwindDest(),
std::nullopt, "", II->getParent());
}
// Remove debug intrinsics which describe the value contained within the
// alloca. In addition to removing dbg.{declare,addr} which simply point to
// the alloca, remove dbg.value(<alloca>, ..., DW_OP_deref)'s as well, e.g.:
//
// ```
// define void @foo(i32 %0) {
// %a = alloca i32 ; Deleted.
// store i32 %0, i32* %a
// dbg.value(i32 %0, "arg0") ; Not deleted.
// dbg.value(i32* %a, "arg0", DW_OP_deref) ; Deleted.
// call void @trivially_inlinable_no_op(i32* %a)
// ret void
// }
// ```
//
// This may not be required if we stop describing the contents of allocas
// using dbg.value(<alloca>, ..., DW_OP_deref), but we currently do this in
// the LowerDbgDeclare utility.
//
// If there is a dead store to `%a` in @trivially_inlinable_no_op, the
// "arg0" dbg.value may be stale after the call. However, failing to remove
// the DW_OP_deref dbg.value causes large gaps in location coverage.
//
// FIXME: the Assignment Tracking project has now likely made this
// redundant (and it's sometimes harmful).
for (auto *DVI : DVIs)
if (DVI->isAddressOfVariable() || DVI->getExpression()->startsWithDeref())
DVI->eraseFromParent();
for (auto *DPV : DPVs)
if (DPV->isAddressOfVariable() || DPV->getExpression()->startsWithDeref())
DPV->eraseFromParent();
return eraseInstFromFunction(MI);
}
return nullptr;
}
/// Move the call to free before a NULL test.
///
/// Check if this free is accessed after its argument has been test
/// against NULL (property 0).
/// If yes, it is legal to move this call in its predecessor block.
///
/// The move is performed only if the block containing the call to free
/// will be removed, i.e.:
/// 1. it has only one predecessor P, and P has two successors
/// 2. it contains the call, noops, and an unconditional branch
/// 3. its successor is the same as its predecessor's successor
///
/// The profitability is out-of concern here and this function should
/// be called only if the caller knows this transformation would be
/// profitable (e.g., for code size).
static Instruction *tryToMoveFreeBeforeNullTest(CallInst &FI,
const DataLayout &DL) {
Value *Op = FI.getArgOperand(0);
BasicBlock *FreeInstrBB = FI.getParent();
BasicBlock *PredBB = FreeInstrBB->getSinglePredecessor();
// Validate part of constraint #1: Only one predecessor
// FIXME: We can extend the number of predecessor, but in that case, we
// would duplicate the call to free in each predecessor and it may
// not be profitable even for code size.
if (!PredBB)
return nullptr;
// Validate constraint #2: Does this block contains only the call to
// free, noops, and an unconditional branch?
BasicBlock *SuccBB;
Instruction *FreeInstrBBTerminator = FreeInstrBB->getTerminator();
if (!match(FreeInstrBBTerminator, m_UnconditionalBr(SuccBB)))
return nullptr;
// If there are only 2 instructions in the block, at this point,
// this is the call to free and unconditional.
// If there are more than 2 instructions, check that they are noops
// i.e., they won't hurt the performance of the generated code.
if (FreeInstrBB->size() != 2) {
for (const Instruction &Inst : FreeInstrBB->instructionsWithoutDebug()) {
if (&Inst == &FI || &Inst == FreeInstrBBTerminator)
continue;
auto *Cast = dyn_cast<CastInst>(&Inst);
if (!Cast || !Cast->isNoopCast(DL))
return nullptr;
}
}
// Validate the rest of constraint #1 by matching on the pred branch.
Instruction *TI = PredBB->getTerminator();
BasicBlock *TrueBB, *FalseBB;
ICmpInst::Predicate Pred;
if (!match(TI, m_Br(m_ICmp(Pred,
m_CombineOr(m_Specific(Op),
m_Specific(Op->stripPointerCasts())),
m_Zero()),
TrueBB, FalseBB)))
return nullptr;
if (Pred != ICmpInst::ICMP_EQ && Pred != ICmpInst::ICMP_NE)
return nullptr;
// Validate constraint #3: Ensure the null case just falls through.
if (SuccBB != (Pred == ICmpInst::ICMP_EQ ? TrueBB : FalseBB))
return nullptr;
assert(FreeInstrBB == (Pred == ICmpInst::ICMP_EQ ? FalseBB : TrueBB) &&
"Broken CFG: missing edge from predecessor to successor");
// At this point, we know that everything in FreeInstrBB can be moved
// before TI.
for (Instruction &Instr : llvm::make_early_inc_range(*FreeInstrBB)) {
if (&Instr == FreeInstrBBTerminator)
break;
Instr.moveBeforePreserving(TI);
}
assert(FreeInstrBB->size() == 1 &&
"Only the branch instruction should remain");
// Now that we've moved the call to free before the NULL check, we have to
// remove any attributes on its parameter that imply it's non-null, because
// those attributes might have only been valid because of the NULL check, and
// we can get miscompiles if we keep them. This is conservative if non-null is
// also implied by something other than the NULL check, but it's guaranteed to
// be correct, and the conservativeness won't matter in practice, since the
// attributes are irrelevant for the call to free itself and the pointer
// shouldn't be used after the call.
AttributeList Attrs = FI.getAttributes();
Attrs = Attrs.removeParamAttribute(FI.getContext(), 0, Attribute::NonNull);
Attribute Dereferenceable = Attrs.getParamAttr(0, Attribute::Dereferenceable);
if (Dereferenceable.isValid()) {
uint64_t Bytes = Dereferenceable.getDereferenceableBytes();
Attrs = Attrs.removeParamAttribute(FI.getContext(), 0,
Attribute::Dereferenceable);
Attrs = Attrs.addDereferenceableOrNullParamAttr(FI.getContext(), 0, Bytes);
}
FI.setAttributes(Attrs);
return &FI;
}
Instruction *InstCombinerImpl::visitFree(CallInst &FI, Value *Op) {
// free undef -> unreachable.
if (isa<UndefValue>(Op)) {
// Leave a marker since we can't modify the CFG here.
CreateNonTerminatorUnreachable(&FI);
return eraseInstFromFunction(FI);
}
// If we have 'free null' delete the instruction. This can happen in stl code
// when lots of inlining happens.
if (isa<ConstantPointerNull>(Op))
return eraseInstFromFunction(FI);
// If we had free(realloc(...)) with no intervening uses, then eliminate the
// realloc() entirely.
CallInst *CI = dyn_cast<CallInst>(Op);
if (CI && CI->hasOneUse())
if (Value *ReallocatedOp = getReallocatedOperand(CI))
return eraseInstFromFunction(*replaceInstUsesWith(*CI, ReallocatedOp));
// If we optimize for code size, try to move the call to free before the null
// test so that simplify cfg can remove the empty block and dead code
// elimination the branch. I.e., helps to turn something like:
// if (foo) free(foo);
// into
// free(foo);
//
// Note that we can only do this for 'free' and not for any flavor of
// 'operator delete'; there is no 'operator delete' symbol for which we are
// permitted to invent a call, even if we're passing in a null pointer.
if (MinimizeSize) {
LibFunc Func;
if (TLI.getLibFunc(FI, Func) && TLI.has(Func) && Func == LibFunc_free)
if (Instruction *I = tryToMoveFreeBeforeNullTest(FI, DL))
return I;
}
return nullptr;
}
Instruction *InstCombinerImpl::visitReturnInst(ReturnInst &RI) {
// Nothing for now.
return nullptr;
}
// WARNING: keep in sync with SimplifyCFGOpt::simplifyUnreachable()!
bool InstCombinerImpl::removeInstructionsBeforeUnreachable(Instruction &I) {
// Try to remove the previous instruction if it must lead to unreachable.
// This includes instructions like stores and "llvm.assume" that may not get
// removed by simple dead code elimination.
bool Changed = false;
while (Instruction *Prev = I.getPrevNonDebugInstruction()) {
// While we theoretically can erase EH, that would result in a block that
// used to start with an EH no longer starting with EH, which is invalid.
// To make it valid, we'd need to fixup predecessors to no longer refer to
// this block, but that changes CFG, which is not allowed in InstCombine.
if (Prev->isEHPad())
break; // Can not drop any more instructions. We're done here.
if (!isGuaranteedToTransferExecutionToSuccessor(Prev))
break; // Can not drop any more instructions. We're done here.
// Otherwise, this instruction can be freely erased,
// even if it is not side-effect free.
// A value may still have uses before we process it here (for example, in
// another unreachable block), so convert those to poison.
replaceInstUsesWith(*Prev, PoisonValue::get(Prev->getType()));
eraseInstFromFunction(*Prev);
Changed = true;
}
return Changed;
}
Instruction *InstCombinerImpl::visitUnreachableInst(UnreachableInst &I) {
removeInstructionsBeforeUnreachable(I);
return nullptr;
}
Instruction *InstCombinerImpl::visitUnconditionalBranchInst(BranchInst &BI) {
assert(BI.isUnconditional() && "Only for unconditional branches.");
// If this store is the second-to-last instruction in the basic block
// (excluding debug info and bitcasts of pointers) and if the block ends with
// an unconditional branch, try to move the store to the successor block.
auto GetLastSinkableStore = [](BasicBlock::iterator BBI) {
auto IsNoopInstrForStoreMerging = [](BasicBlock::iterator BBI) {
return BBI->isDebugOrPseudoInst() ||
(isa<BitCastInst>(BBI) && BBI->getType()->isPointerTy());
};
BasicBlock::iterator FirstInstr = BBI->getParent()->begin();
do {
if (BBI != FirstInstr)
--BBI;
} while (BBI != FirstInstr && IsNoopInstrForStoreMerging(BBI));
return dyn_cast<StoreInst>(BBI);
};
if (StoreInst *SI = GetLastSinkableStore(BasicBlock::iterator(BI)))
if (mergeStoreIntoSuccessor(*SI))
return &BI;
return nullptr;
}
void InstCombinerImpl::addDeadEdge(BasicBlock *From, BasicBlock *To,
SmallVectorImpl<BasicBlock *> &Worklist) {
if (!DeadEdges.insert({From, To}).second)
return;
// Replace phi node operands in successor with poison.
for (PHINode &PN : To->phis())
for (Use &U : PN.incoming_values())
if (PN.getIncomingBlock(U) == From && !isa<PoisonValue>(U)) {
replaceUse(U, PoisonValue::get(PN.getType()));
addToWorklist(&PN);
MadeIRChange = true;
}
Worklist.push_back(To);
}
// Under the assumption that I is unreachable, remove it and following
// instructions. Changes are reported directly to MadeIRChange.
void InstCombinerImpl::handleUnreachableFrom(
Instruction *I, SmallVectorImpl<BasicBlock *> &Worklist) {
BasicBlock *BB = I->getParent();
for (Instruction &Inst : make_early_inc_range(
make_range(std::next(BB->getTerminator()->getReverseIterator()),
std::next(I->getReverseIterator())))) {
if (!Inst.use_empty() && !Inst.getType()->isTokenTy()) {
replaceInstUsesWith(Inst, PoisonValue::get(Inst.getType()));
MadeIRChange = true;
}
if (Inst.isEHPad() || Inst.getType()->isTokenTy())
continue;
// RemoveDIs: erase debug-info on this instruction manually.
Inst.dropDbgValues();
eraseInstFromFunction(Inst);
MadeIRChange = true;
}
// RemoveDIs: to match behaviour in dbg.value mode, drop debug-info on
// terminator too.
BB->getTerminator()->dropDbgValues();
// Handle potentially dead successors.
for (BasicBlock *Succ : successors(BB))
addDeadEdge(BB, Succ, Worklist);
}
void InstCombinerImpl::handlePotentiallyDeadBlocks(
SmallVectorImpl<BasicBlock *> &Worklist) {
while (!Worklist.empty()) {
BasicBlock *BB = Worklist.pop_back_val();
if (!all_of(predecessors(BB), [&](BasicBlock *Pred) {
return DeadEdges.contains({Pred, BB}) || DT.dominates(BB, Pred);
}))
continue;
handleUnreachableFrom(&BB->front(), Worklist);
}
}
void InstCombinerImpl::handlePotentiallyDeadSuccessors(BasicBlock *BB,
BasicBlock *LiveSucc) {
SmallVector<BasicBlock *> Worklist;
for (BasicBlock *Succ : successors(BB)) {
// The live successor isn't dead.
if (Succ == LiveSucc)
continue;
addDeadEdge(BB, Succ, Worklist);
}
handlePotentiallyDeadBlocks(Worklist);
}
Instruction *InstCombinerImpl::visitBranchInst(BranchInst &BI) {
if (BI.isUnconditional())
return visitUnconditionalBranchInst(BI);
// Change br (not X), label True, label False to: br X, label False, True
Value *Cond = BI.getCondition();
Value *X;
if (match(Cond, m_Not(m_Value(X))) && !isa<Constant>(X)) {
// Swap Destinations and condition...
BI.swapSuccessors();
return replaceOperand(BI, 0, X);
}
// Canonicalize logical-and-with-invert as logical-or-with-invert.
// This is done by inverting the condition and swapping successors:
// br (X && !Y), T, F --> br !(X && !Y), F, T --> br (!X || Y), F, T
Value *Y;
if (isa<SelectInst>(Cond) &&
match(Cond,
m_OneUse(m_LogicalAnd(m_Value(X), m_OneUse(m_Not(m_Value(Y))))))) {
Value *NotX = Builder.CreateNot(X, "not." + X->getName());
Value *Or = Builder.CreateLogicalOr(NotX, Y);
BI.swapSuccessors();
return replaceOperand(BI, 0, Or);
}
// If the condition is irrelevant, remove the use so that other
// transforms on the condition become more effective.
if (!isa<ConstantInt>(Cond) && BI.getSuccessor(0) == BI.getSuccessor(1))
return replaceOperand(BI, 0, ConstantInt::getFalse(Cond->getType()));
// Canonicalize, for example, fcmp_one -> fcmp_oeq.
CmpInst::Predicate Pred;
if (match(Cond, m_OneUse(m_FCmp(Pred, m_Value(), m_Value()))) &&
!isCanonicalPredicate(Pred)) {
// Swap destinations and condition.
auto *Cmp = cast<CmpInst>(Cond);
Cmp->setPredicate(CmpInst::getInversePredicate(Pred));
BI.swapSuccessors();
Worklist.push(Cmp);
return &BI;
}
if (isa<UndefValue>(Cond)) {
handlePotentiallyDeadSuccessors(BI.getParent(), /*LiveSucc*/ nullptr);
return nullptr;
}
if (auto *CI = dyn_cast<ConstantInt>(Cond)) {
handlePotentiallyDeadSuccessors(BI.getParent(),
BI.getSuccessor(!CI->getZExtValue()));
return nullptr;
}
DC.registerBranch(&BI);
return nullptr;
}
Instruction *InstCombinerImpl::visitSwitchInst(SwitchInst &SI) {
Value *Cond = SI.getCondition();
Value *Op0;
ConstantInt *AddRHS;
if (match(Cond, m_Add(m_Value(Op0), m_ConstantInt(AddRHS)))) {
// Change 'switch (X+4) case 1:' into 'switch (X) case -3'.
for (auto Case : SI.cases()) {
Constant *NewCase = ConstantExpr::getSub(Case.getCaseValue(), AddRHS);
assert(isa<ConstantInt>(NewCase) &&
"Result of expression should be constant");
Case.setValue(cast<ConstantInt>(NewCase));
}
return replaceOperand(SI, 0, Op0);
}
ConstantInt *SubLHS;
if (match(Cond, m_Sub(m_ConstantInt(SubLHS), m_Value(Op0)))) {
// Change 'switch (1-X) case 1:' into 'switch (X) case 0'.
for (auto Case : SI.cases()) {
Constant *NewCase = ConstantExpr::getSub(SubLHS, Case.getCaseValue());
assert(isa<ConstantInt>(NewCase) &&
"Result of expression should be constant");
Case.setValue(cast<ConstantInt>(NewCase));
}
return replaceOperand(SI, 0, Op0);
}
uint64_t ShiftAmt;
if (match(Cond, m_Shl(m_Value(Op0), m_ConstantInt(ShiftAmt))) &&
ShiftAmt < Op0->getType()->getScalarSizeInBits() &&
all_of(SI.cases(), [&](const auto &Case) {
return Case.getCaseValue()->getValue().countr_zero() >= ShiftAmt;
})) {
// Change 'switch (X << 2) case 4:' into 'switch (X) case 1:'.
OverflowingBinaryOperator *Shl = cast<OverflowingBinaryOperator>(Cond);
if (Shl->hasNoUnsignedWrap() || Shl->hasNoSignedWrap() ||
Shl->hasOneUse()) {
Value *NewCond = Op0;
if (!Shl->hasNoUnsignedWrap() && !Shl->hasNoSignedWrap()) {
// If the shift may wrap, we need to mask off the shifted bits.
unsigned BitWidth = Op0->getType()->getScalarSizeInBits();
NewCond = Builder.CreateAnd(
Op0, APInt::getLowBitsSet(BitWidth, BitWidth - ShiftAmt));
}
for (auto Case : SI.cases()) {
const APInt &CaseVal = Case.getCaseValue()->getValue();
APInt ShiftedCase = Shl->hasNoSignedWrap() ? CaseVal.ashr(ShiftAmt)
: CaseVal.lshr(ShiftAmt);
Case.setValue(ConstantInt::get(SI.getContext(), ShiftedCase));
}
return replaceOperand(SI, 0, NewCond);
}
}
KnownBits Known = computeKnownBits(Cond, 0, &SI);
unsigned LeadingKnownZeros = Known.countMinLeadingZeros();
unsigned LeadingKnownOnes = Known.countMinLeadingOnes();
// Compute the number of leading bits we can ignore.
// TODO: A better way to determine this would use ComputeNumSignBits().
for (const auto &C : SI.cases()) {
LeadingKnownZeros =
std::min(LeadingKnownZeros, C.getCaseValue()->getValue().countl_zero());
LeadingKnownOnes =
std::min(LeadingKnownOnes, C.getCaseValue()->getValue().countl_one());
}
unsigned NewWidth = Known.getBitWidth() - std::max(LeadingKnownZeros, LeadingKnownOnes);
// Shrink the condition operand if the new type is smaller than the old type.
// But do not shrink to a non-standard type, because backend can't generate
// good code for that yet.
// TODO: We can make it aggressive again after fixing PR39569.
if (NewWidth > 0 && NewWidth < Known.getBitWidth() &&
shouldChangeType(Known.getBitWidth(), NewWidth)) {
IntegerType *Ty = IntegerType::get(SI.getContext(), NewWidth);
Builder.SetInsertPoint(&SI);
Value *NewCond = Builder.CreateTrunc(Cond, Ty, "trunc");
for (auto Case : SI.cases()) {
APInt TruncatedCase = Case.getCaseValue()->getValue().trunc(NewWidth);
Case.setValue(ConstantInt::get(SI.getContext(), TruncatedCase));
}
return replaceOperand(SI, 0, NewCond);
}
if (isa<UndefValue>(Cond)) {
handlePotentiallyDeadSuccessors(SI.getParent(), /*LiveSucc*/ nullptr);
return nullptr;
}
if (auto *CI = dyn_cast<ConstantInt>(Cond)) {
handlePotentiallyDeadSuccessors(SI.getParent(),
SI.findCaseValue(CI)->getCaseSuccessor());
return nullptr;
}
return nullptr;
}
Instruction *
InstCombinerImpl::foldExtractOfOverflowIntrinsic(ExtractValueInst &EV) {
auto *WO = dyn_cast<WithOverflowInst>(EV.getAggregateOperand());
if (!WO)
return nullptr;
Intrinsic::ID OvID = WO->getIntrinsicID();
const APInt *C = nullptr;
if (match(WO->getRHS(), m_APIntAllowUndef(C))) {
if (*EV.idx_begin() == 0 && (OvID == Intrinsic::smul_with_overflow ||
OvID == Intrinsic::umul_with_overflow)) {
// extractvalue (any_mul_with_overflow X, -1), 0 --> -X
if (C->isAllOnes())
return BinaryOperator::CreateNeg(WO->getLHS());
// extractvalue (any_mul_with_overflow X, 2^n), 0 --> X << n
if (C->isPowerOf2()) {
return BinaryOperator::CreateShl(
WO->getLHS(),
ConstantInt::get(WO->getLHS()->getType(), C->logBase2()));
}
}
}
// We're extracting from an overflow intrinsic. See if we're the only user.
// That allows us to simplify multiple result intrinsics to simpler things
// that just get one value.
if (!WO->hasOneUse())
return nullptr;
// Check if we're grabbing only the result of a 'with overflow' intrinsic
// and replace it with a traditional binary instruction.
if (*EV.idx_begin() == 0) {
Instruction::BinaryOps BinOp = WO->getBinaryOp();
Value *LHS = WO->getLHS(), *RHS = WO->getRHS();
// Replace the old instruction's uses with poison.
replaceInstUsesWith(*WO, PoisonValue::get(WO->getType()));
eraseInstFromFunction(*WO);
return BinaryOperator::Create(BinOp, LHS, RHS);
}
assert(*EV.idx_begin() == 1 && "Unexpected extract index for overflow inst");
// (usub LHS, RHS) overflows when LHS is unsigned-less-than RHS.
if (OvID == Intrinsic::usub_with_overflow)
return new ICmpInst(ICmpInst::ICMP_ULT, WO->getLHS(), WO->getRHS());
// smul with i1 types overflows when both sides are set: -1 * -1 == +1, but
// +1 is not possible because we assume signed values.
if (OvID == Intrinsic::smul_with_overflow &&
WO->getLHS()->getType()->isIntOrIntVectorTy(1))
return BinaryOperator::CreateAnd(WO->getLHS(), WO->getRHS());
// If only the overflow result is used, and the right hand side is a
// constant (or constant splat), we can remove the intrinsic by directly
// checking for overflow.
if (C) {
// Compute the no-wrap range for LHS given RHS=C, then construct an
// equivalent icmp, potentially using an offset.
ConstantRange NWR = ConstantRange::makeExactNoWrapRegion(
WO->getBinaryOp(), *C, WO->getNoWrapKind());
CmpInst::Predicate Pred;
APInt NewRHSC, Offset;
NWR.getEquivalentICmp(Pred, NewRHSC, Offset);
auto *OpTy = WO->getRHS()->getType();
auto *NewLHS = WO->getLHS();
if (Offset != 0)
NewLHS = Builder.CreateAdd(NewLHS, ConstantInt::get(OpTy, Offset));
return new ICmpInst(ICmpInst::getInversePredicate(Pred), NewLHS,
ConstantInt::get(OpTy, NewRHSC));
}
return nullptr;
}
Instruction *InstCombinerImpl::visitExtractValueInst(ExtractValueInst &EV) {
Value *Agg = EV.getAggregateOperand();
if (!EV.hasIndices())
return replaceInstUsesWith(EV, Agg);
if (Value *V = simplifyExtractValueInst(Agg, EV.getIndices(),
SQ.getWithInstruction(&EV)))
return replaceInstUsesWith(EV, V);
if (InsertValueInst *IV = dyn_cast<InsertValueInst>(Agg)) {
// We're extracting from an insertvalue instruction, compare the indices
const unsigned *exti, *exte, *insi, *inse;
for (exti = EV.idx_begin(), insi = IV->idx_begin(),
exte = EV.idx_end(), inse = IV->idx_end();
exti != exte && insi != inse;
++exti, ++insi) {
if (*insi != *exti)
// The insert and extract both reference distinctly different elements.
// This means the extract is not influenced by the insert, and we can
// replace the aggregate operand of the extract with the aggregate
// operand of the insert. i.e., replace
// %I = insertvalue { i32, { i32 } } %A, { i32 } { i32 42 }, 1
// %E = extractvalue { i32, { i32 } } %I, 0
// with
// %E = extractvalue { i32, { i32 } } %A, 0
return ExtractValueInst::Create(IV->getAggregateOperand(),
EV.getIndices());
}
if (exti == exte && insi == inse)
// Both iterators are at the end: Index lists are identical. Replace
// %B = insertvalue { i32, { i32 } } %A, i32 42, 1, 0
// %C = extractvalue { i32, { i32 } } %B, 1, 0
// with "i32 42"
return replaceInstUsesWith(EV, IV->getInsertedValueOperand());
if (exti == exte) {
// The extract list is a prefix of the insert list. i.e. replace
// %I = insertvalue { i32, { i32 } } %A, i32 42, 1, 0
// %E = extractvalue { i32, { i32 } } %I, 1
// with
// %X = extractvalue { i32, { i32 } } %A, 1
// %E = insertvalue { i32 } %X, i32 42, 0
// by switching the order of the insert and extract (though the
// insertvalue should be left in, since it may have other uses).
Value *NewEV = Builder.CreateExtractValue(IV->getAggregateOperand(),
EV.getIndices());
return InsertValueInst::Create(NewEV, IV->getInsertedValueOperand(),
ArrayRef(insi, inse));
}
if (insi == inse)
// The insert list is a prefix of the extract list
// We can simply remove the common indices from the extract and make it
// operate on the inserted value instead of the insertvalue result.
// i.e., replace
// %I = insertvalue { i32, { i32 } } %A, { i32 } { i32 42 }, 1
// %E = extractvalue { i32, { i32 } } %I, 1, 0
// with
// %E extractvalue { i32 } { i32 42 }, 0
return ExtractValueInst::Create(IV->getInsertedValueOperand(),
ArrayRef(exti, exte));
}
if (Instruction *R = foldExtractOfOverflowIntrinsic(EV))
return R;
if (LoadInst *L = dyn_cast<LoadInst>(Agg)) {
// Bail out if the aggregate contains scalable vector type
if (auto *STy = dyn_cast<StructType>(Agg->getType());
STy && STy->containsScalableVectorType())
return nullptr;
// If the (non-volatile) load only has one use, we can rewrite this to a
// load from a GEP. This reduces the size of the load. If a load is used
// only by extractvalue instructions then this either must have been
// optimized before, or it is a struct with padding, in which case we
// don't want to do the transformation as it loses padding knowledge.
if (L->isSimple() && L->hasOneUse()) {
// extractvalue has integer indices, getelementptr has Value*s. Convert.
SmallVector<Value*, 4> Indices;
// Prefix an i32 0 since we need the first element.
Indices.push_back(Builder.getInt32(0));
for (unsigned Idx : EV.indices())
Indices.push_back(Builder.getInt32(Idx));
// We need to insert these at the location of the old load, not at that of
// the extractvalue.
Builder.SetInsertPoint(L);
Value *GEP = Builder.CreateInBoundsGEP(L->getType(),
L->getPointerOperand(), Indices);
Instruction *NL = Builder.CreateLoad(EV.getType(), GEP);
// Whatever aliasing information we had for the orignal load must also
// hold for the smaller load, so propagate the annotations.
NL->setAAMetadata(L->getAAMetadata());
// Returning the load directly will cause the main loop to insert it in
// the wrong spot, so use replaceInstUsesWith().
return replaceInstUsesWith(EV, NL);
}
}
if (auto *PN = dyn_cast<PHINode>(Agg))
if (Instruction *Res = foldOpIntoPhi(EV, PN))
return Res;
// We could simplify extracts from other values. Note that nested extracts may
// already be simplified implicitly by the above: extract (extract (insert) )
// will be translated into extract ( insert ( extract ) ) first and then just
// the value inserted, if appropriate. Similarly for extracts from single-use
// loads: extract (extract (load)) will be translated to extract (load (gep))
// and if again single-use then via load (gep (gep)) to load (gep).
// However, double extracts from e.g. function arguments or return values
// aren't handled yet.
return nullptr;
}
/// Return 'true' if the given typeinfo will match anything.
static bool isCatchAll(EHPersonality Personality, Constant *TypeInfo) {
switch (Personality) {
case EHPersonality::GNU_C:
case EHPersonality::GNU_C_SjLj:
case EHPersonality::Rust:
// The GCC C EH and Rust personality only exists to support cleanups, so
// it's not clear what the semantics of catch clauses are.
return false;
case EHPersonality::Unknown:
return false;
case EHPersonality::GNU_Ada:
// While __gnat_all_others_value will match any Ada exception, it doesn't
// match foreign exceptions (or didn't, before gcc-4.7).
return false;
case EHPersonality::GNU_CXX:
case EHPersonality::GNU_CXX_SjLj:
case EHPersonality::GNU_ObjC:
case EHPersonality::MSVC_X86SEH:
case EHPersonality::MSVC_TableSEH:
case EHPersonality::MSVC_CXX:
case EHPersonality::CoreCLR:
case EHPersonality::Wasm_CXX:
case EHPersonality::XL_CXX:
return TypeInfo->isNullValue();
}
llvm_unreachable("invalid enum");
}
static bool shorter_filter(const Value *LHS, const Value *RHS) {
return
cast<ArrayType>(LHS->getType())->getNumElements()
<
cast<ArrayType>(RHS->getType())->getNumElements();
}
Instruction *InstCombinerImpl::visitLandingPadInst(LandingPadInst &LI) {
// The logic here should be correct for any real-world personality function.
// However if that turns out not to be true, the offending logic can always
// be conditioned on the personality function, like the catch-all logic is.
EHPersonality Personality =
classifyEHPersonality(LI.getParent()->getParent()->getPersonalityFn());
// Simplify the list of clauses, eg by removing repeated catch clauses
// (these are often created by inlining).
bool MakeNewInstruction = false; // If true, recreate using the following:
SmallVector<Constant *, 16> NewClauses; // - Clauses for the new instruction;
bool CleanupFlag = LI.isCleanup(); // - The new instruction is a cleanup.
SmallPtrSet<Value *, 16> AlreadyCaught; // Typeinfos known caught already.
for (unsigned i = 0, e = LI.getNumClauses(); i != e; ++i) {
bool isLastClause = i + 1 == e;
if (LI.isCatch(i)) {
// A catch clause.
Constant *CatchClause = LI.getClause(i);
Constant *TypeInfo = CatchClause->stripPointerCasts();
// If we already saw this clause, there is no point in having a second
// copy of it.
if (AlreadyCaught.insert(TypeInfo).second) {
// This catch clause was not already seen.
NewClauses.push_back(CatchClause);
} else {
// Repeated catch clause - drop the redundant copy.
MakeNewInstruction = true;
}
// If this is a catch-all then there is no point in keeping any following
// clauses or marking the landingpad as having a cleanup.
if (isCatchAll(Personality, TypeInfo)) {
if (!isLastClause)
MakeNewInstruction = true;
CleanupFlag = false;
break;
}
} else {
// A filter clause. If any of the filter elements were already caught
// then they can be dropped from the filter. It is tempting to try to
// exploit the filter further by saying that any typeinfo that does not
// occur in the filter can't be caught later (and thus can be dropped).
// However this would be wrong, since typeinfos can match without being
// equal (for example if one represents a C++ class, and the other some
// class derived from it).
assert(LI.isFilter(i) && "Unsupported landingpad clause!");
Constant *FilterClause = LI.getClause(i);
ArrayType *FilterType = cast<ArrayType>(FilterClause->getType());
unsigned NumTypeInfos = FilterType->getNumElements();
// An empty filter catches everything, so there is no point in keeping any
// following clauses or marking the landingpad as having a cleanup. By
// dealing with this case here the following code is made a bit simpler.
if (!NumTypeInfos) {
NewClauses.push_back(FilterClause);
if (!isLastClause)
MakeNewInstruction = true;
CleanupFlag = false;
break;
}
bool MakeNewFilter = false; // If true, make a new filter.
SmallVector<Constant *, 16> NewFilterElts; // New elements.
if (isa<ConstantAggregateZero>(FilterClause)) {
// Not an empty filter - it contains at least one null typeinfo.
assert(NumTypeInfos > 0 && "Should have handled empty filter already!");
Constant *TypeInfo =
Constant::getNullValue(FilterType->getElementType());
// If this typeinfo is a catch-all then the filter can never match.
if (isCatchAll(Personality, TypeInfo)) {
// Throw the filter away.
MakeNewInstruction = true;
continue;
}
// There is no point in having multiple copies of this typeinfo, so
// discard all but the first copy if there is more than one.
NewFilterElts.push_back(TypeInfo);
if (NumTypeInfos > 1)
MakeNewFilter = true;
} else {
ConstantArray *Filter = cast<ConstantArray>(FilterClause);
SmallPtrSet<Value *, 16> SeenInFilter; // For uniquing the elements.
NewFilterElts.reserve(NumTypeInfos);
// Remove any filter elements that were already caught or that already
// occurred in the filter. While there, see if any of the elements are
// catch-alls. If so, the filter can be discarded.
bool SawCatchAll = false;
for (unsigned j = 0; j != NumTypeInfos; ++j) {
Constant *Elt = Filter->getOperand(j);
Constant *TypeInfo = Elt->stripPointerCasts();
if (isCatchAll(Personality, TypeInfo)) {
// This element is a catch-all. Bail out, noting this fact.
SawCatchAll = true;
break;
}
// Even if we've seen a type in a catch clause, we don't want to
// remove it from the filter. An unexpected type handler may be
// set up for a call site which throws an exception of the same
// type caught. In order for the exception thrown by the unexpected
// handler to propagate correctly, the filter must be correctly
// described for the call site.
//
// Example:
//
// void unexpected() { throw 1;}
// void foo() throw (int) {
// std::set_unexpected(unexpected);
// try {
// throw 2.0;
// } catch (int i) {}
// }
// There is no point in having multiple copies of the same typeinfo in
// a filter, so only add it if we didn't already.
if (SeenInFilter.insert(TypeInfo).second)
NewFilterElts.push_back(cast<Constant>(Elt));
}
// A filter containing a catch-all cannot match anything by definition.
if (SawCatchAll) {
// Throw the filter away.
MakeNewInstruction = true;
continue;
}
// If we dropped something from the filter, make a new one.
if (NewFilterElts.size() < NumTypeInfos)
MakeNewFilter = true;
}
if (MakeNewFilter) {
FilterType = ArrayType::get(FilterType->getElementType(),
NewFilterElts.size());
FilterClause = ConstantArray::get(FilterType, NewFilterElts);
MakeNewInstruction = true;
}
NewClauses.push_back(FilterClause);
// If the new filter is empty then it will catch everything so there is
// no point in keeping any following clauses or marking the landingpad
// as having a cleanup. The case of the original filter being empty was
// already handled above.
if (MakeNewFilter && !NewFilterElts.size()) {
assert(MakeNewInstruction && "New filter but not a new instruction!");
CleanupFlag = false;
break;
}
}
}
// If several filters occur in a row then reorder them so that the shortest
// filters come first (those with the smallest number of elements). This is
// advantageous because shorter filters are more likely to match, speeding up
// unwinding, but mostly because it increases the effectiveness of the other
// filter optimizations below.
for (unsigned i = 0, e = NewClauses.size(); i + 1 < e; ) {
unsigned j;
// Find the maximal 'j' s.t. the range [i, j) consists entirely of filters.
for (j = i; j != e; ++j)
if (!isa<ArrayType>(NewClauses[j]->getType()))
break;
// Check whether the filters are already sorted by length. We need to know
// if sorting them is actually going to do anything so that we only make a
// new landingpad instruction if it does.
for (unsigned k = i; k + 1 < j; ++k)
if (shorter_filter(NewClauses[k+1], NewClauses[k])) {
// Not sorted, so sort the filters now. Doing an unstable sort would be
// correct too but reordering filters pointlessly might confuse users.
std::stable_sort(NewClauses.begin() + i, NewClauses.begin() + j,
shorter_filter);
MakeNewInstruction = true;
break;
}
// Look for the next batch of filters.
i = j + 1;
}
// If typeinfos matched if and only if equal, then the elements of a filter L
// that occurs later than a filter F could be replaced by the intersection of
// the elements of F and L. In reality two typeinfos can match without being
// equal (for example if one represents a C++ class, and the other some class
// derived from it) so it would be wrong to perform this transform in general.
// However the transform is correct and useful if F is a subset of L. In that
// case L can be replaced by F, and thus removed altogether since repeating a
// filter is pointless. So here we look at all pairs of filters F and L where
// L follows F in the list of clauses, and remove L if every element of F is
// an element of L. This can occur when inlining C++ functions with exception
// specifications.
for (unsigned i = 0; i + 1 < NewClauses.size(); ++i) {
// Examine each filter in turn.
Value *Filter = NewClauses[i];
ArrayType *FTy = dyn_cast<ArrayType>(Filter->getType());
if (!FTy)
// Not a filter - skip it.
continue;
unsigned FElts = FTy->getNumElements();
// Examine each filter following this one. Doing this backwards means that
// we don't have to worry about filters disappearing under us when removed.
for (unsigned j = NewClauses.size() - 1; j != i; --j) {
Value *LFilter = NewClauses[j];
ArrayType *LTy = dyn_cast<ArrayType>(LFilter->getType());
if (!LTy)
// Not a filter - skip it.
continue;
// If Filter is a subset of LFilter, i.e. every element of Filter is also
// an element of LFilter, then discard LFilter.
SmallVectorImpl<Constant *>::iterator J = NewClauses.begin() + j;
// If Filter is empty then it is a subset of LFilter.
if (!FElts) {
// Discard LFilter.
NewClauses.erase(J);
MakeNewInstruction = true;
// Move on to the next filter.
continue;
}
unsigned LElts = LTy->getNumElements();
// If Filter is longer than LFilter then it cannot be a subset of it.
if (FElts > LElts)
// Move on to the next filter.
continue;
// At this point we know that LFilter has at least one element.
if (isa<ConstantAggregateZero>(LFilter)) { // LFilter only contains zeros.
// Filter is a subset of LFilter iff Filter contains only zeros (as we
// already know that Filter is not longer than LFilter).
if (isa<ConstantAggregateZero>(Filter)) {
assert(FElts <= LElts && "Should have handled this case earlier!");
// Discard LFilter.
NewClauses.erase(J);
MakeNewInstruction = true;
}
// Move on to the next filter.
continue;
}
ConstantArray *LArray = cast<ConstantArray>(LFilter);
if (isa<ConstantAggregateZero>(Filter)) { // Filter only contains zeros.
// Since Filter is non-empty and contains only zeros, it is a subset of
// LFilter iff LFilter contains a zero.
assert(FElts > 0 && "Should have eliminated the empty filter earlier!");
for (unsigned l = 0; l != LElts; ++l)
if (LArray->getOperand(l)->isNullValue()) {
// LFilter contains a zero - discard it.
NewClauses.erase(J);
MakeNewInstruction = true;
break;
}
// Move on to the next filter.
continue;
}
// At this point we know that both filters are ConstantArrays. Loop over
// operands to see whether every element of Filter is also an element of
// LFilter. Since filters tend to be short this is probably faster than
// using a method that scales nicely.
ConstantArray *FArray = cast<ConstantArray>(Filter);
bool AllFound = true;
for (unsigned f = 0; f != FElts; ++f) {
Value *FTypeInfo = FArray->getOperand(f)->stripPointerCasts();
AllFound = false;
for (unsigned l = 0; l != LElts; ++l) {
Value *LTypeInfo = LArray->getOperand(l)->stripPointerCasts();
if (LTypeInfo == FTypeInfo) {
AllFound = true;
break;
}
}
if (!AllFound)
break;
}
if (AllFound) {
// Discard LFilter.
NewClauses.erase(J);
MakeNewInstruction = true;
}
// Move on to the next filter.
}
}
// If we changed any of the clauses, replace the old landingpad instruction
// with a new one.
if (MakeNewInstruction) {
LandingPadInst *NLI = LandingPadInst::Create(LI.getType(),
NewClauses.size());
for (unsigned i = 0, e = NewClauses.size(); i != e; ++i)
NLI->addClause(NewClauses[i]);
// A landing pad with no clauses must have the cleanup flag set. It is
// theoretically possible, though highly unlikely, that we eliminated all
// clauses. If so, force the cleanup flag to true.
if (NewClauses.empty())
CleanupFlag = true;
NLI->setCleanup(CleanupFlag);
return NLI;
}
// Even if none of the clauses changed, we may nonetheless have understood
// that the cleanup flag is pointless. Clear it if so.
if (LI.isCleanup() != CleanupFlag) {
assert(!CleanupFlag && "Adding a cleanup, not removing one?!");
LI.setCleanup(CleanupFlag);
return &LI;
}
return nullptr;
}
Value *
InstCombinerImpl::pushFreezeToPreventPoisonFromPropagating(FreezeInst &OrigFI) {
// Try to push freeze through instructions that propagate but don't produce
// poison as far as possible. If an operand of freeze follows three
// conditions 1) one-use, 2) does not produce poison, and 3) has all but one
// guaranteed-non-poison operands then push the freeze through to the one
// operand that is not guaranteed non-poison. The actual transform is as
// follows.
// Op1 = ... ; Op1 can be posion
// Op0 = Inst(Op1, NonPoisonOps...) ; Op0 has only one use and only have
// ; single guaranteed-non-poison operands
// ... = Freeze(Op0)
// =>
// Op1 = ...
// Op1.fr = Freeze(Op1)
// ... = Inst(Op1.fr, NonPoisonOps...)
auto *OrigOp = OrigFI.getOperand(0);
auto *OrigOpInst = dyn_cast<Instruction>(OrigOp);
// While we could change the other users of OrigOp to use freeze(OrigOp), that
// potentially reduces their optimization potential, so let's only do this iff
// the OrigOp is only used by the freeze.
if (!OrigOpInst || !OrigOpInst->hasOneUse() || isa<PHINode>(OrigOp))
return nullptr;
// We can't push the freeze through an instruction which can itself create
// poison. If the only source of new poison is flags, we can simply
// strip them (since we know the only use is the freeze and nothing can
// benefit from them.)
if (canCreateUndefOrPoison(cast<Operator>(OrigOp),
/*ConsiderFlagsAndMetadata*/ false))
return nullptr;
// If operand is guaranteed not to be poison, there is no need to add freeze
// to the operand. So we first find the operand that is not guaranteed to be
// poison.
Use *MaybePoisonOperand = nullptr;
for (Use &U : OrigOpInst->operands()) {
if (isa<MetadataAsValue>(U.get()) ||
isGuaranteedNotToBeUndefOrPoison(U.get()))
continue;
if (!MaybePoisonOperand)
MaybePoisonOperand = &U;
else
return nullptr;
}
OrigOpInst->dropPoisonGeneratingFlagsAndMetadata();
// If all operands are guaranteed to be non-poison, we can drop freeze.
if (!MaybePoisonOperand)
return OrigOp;
Builder.SetInsertPoint(OrigOpInst);
auto *FrozenMaybePoisonOperand = Builder.CreateFreeze(
MaybePoisonOperand->get(), MaybePoisonOperand->get()->getName() + ".fr");
replaceUse(*MaybePoisonOperand, FrozenMaybePoisonOperand);
return OrigOp;
}
Instruction *InstCombinerImpl::foldFreezeIntoRecurrence(FreezeInst &FI,
PHINode *PN) {
// Detect whether this is a recurrence with a start value and some number of
// backedge values. We'll check whether we can push the freeze through the
// backedge values (possibly dropping poison flags along the way) until we
// reach the phi again. In that case, we can move the freeze to the start
// value.
Use *StartU = nullptr;
SmallVector<Value *> Worklist;
for (Use &U : PN->incoming_values()) {
if (DT.dominates(PN->getParent(), PN->getIncomingBlock(U))) {
// Add backedge value to worklist.
Worklist.push_back(U.get());
continue;
}
// Don't bother handling multiple start values.
if (StartU)
return nullptr;
StartU = &U;
}
if (!StartU || Worklist.empty())
return nullptr; // Not a recurrence.
Value *StartV = StartU->get();
BasicBlock *StartBB = PN->getIncomingBlock(*StartU);
bool StartNeedsFreeze = !isGuaranteedNotToBeUndefOrPoison(StartV);
// We can't insert freeze if the start value is the result of the
// terminator (e.g. an invoke).
if (StartNeedsFreeze && StartBB->getTerminator() == StartV)
return nullptr;
SmallPtrSet<Value *, 32> Visited;
SmallVector<Instruction *> DropFlags;
while (!Worklist.empty()) {
Value *V = Worklist.pop_back_val();
if (!Visited.insert(V).second)
continue;
if (Visited.size() > 32)
return nullptr; // Limit the total number of values we inspect.
// Assume that PN is non-poison, because it will be after the transform.
if (V == PN || isGuaranteedNotToBeUndefOrPoison(V))
continue;
Instruction *I = dyn_cast<Instruction>(V);
if (!I || canCreateUndefOrPoison(cast<Operator>(I),
/*ConsiderFlagsAndMetadata*/ false))
return nullptr;
DropFlags.push_back(I);
append_range(Worklist, I->operands());
}
for (Instruction *I : DropFlags)
I->dropPoisonGeneratingFlagsAndMetadata();
if (StartNeedsFreeze) {
Builder.SetInsertPoint(StartBB->getTerminator());
Value *FrozenStartV = Builder.CreateFreeze(StartV,
StartV->getName() + ".fr");
replaceUse(*StartU, FrozenStartV);
}
return replaceInstUsesWith(FI, PN);
}
bool InstCombinerImpl::freezeOtherUses(FreezeInst &FI) {
Value *Op = FI.getOperand(0);
if (isa<Constant>(Op) || Op->hasOneUse())
return false;
// Move the freeze directly after the definition of its operand, so that
// it dominates the maximum number of uses. Note that it may not dominate
// *all* uses if the operand is an invoke/callbr and the use is in a phi on
// the normal/default destination. This is why the domination check in the
// replacement below is still necessary.
BasicBlock::iterator MoveBefore;
if (isa<Argument>(Op)) {
MoveBefore =
FI.getFunction()->getEntryBlock().getFirstNonPHIOrDbgOrAlloca();
} else {
auto MoveBeforeOpt = cast<Instruction>(Op)->getInsertionPointAfterDef();
if (!MoveBeforeOpt)
return false;
MoveBefore = *MoveBeforeOpt;
}
// Don't move to the position of a debug intrinsic.
if (isa<DbgInfoIntrinsic>(MoveBefore))
MoveBefore = MoveBefore->getNextNonDebugInstruction()->getIterator();
// Re-point iterator to come after any debug-info records, if we're
// running in "RemoveDIs" mode
MoveBefore.setHeadBit(false);
bool Changed = false;
if (&FI != &*MoveBefore) {
FI.moveBefore(*MoveBefore->getParent(), MoveBefore);
Changed = true;
}
Op->replaceUsesWithIf(&FI, [&](Use &U) -> bool {
bool Dominates = DT.dominates(&FI, U);
Changed |= Dominates;
return Dominates;
});
return Changed;
}
// Check if any direct or bitcast user of this value is a shuffle instruction.
static bool isUsedWithinShuffleVector(Value *V) {
for (auto *U : V->users()) {
if (isa<ShuffleVectorInst>(U))
return true;
else if (match(U, m_BitCast(m_Specific(V))) && isUsedWithinShuffleVector(U))
return true;
}
return false;
}
Instruction *InstCombinerImpl::visitFreeze(FreezeInst &I) {
Value *Op0 = I.getOperand(0);
if (Value *V = simplifyFreezeInst(Op0, SQ.getWithInstruction(&I)))
return replaceInstUsesWith(I, V);
// freeze (phi const, x) --> phi const, (freeze x)
if (auto *PN = dyn_cast<PHINode>(Op0)) {
if (Instruction *NV = foldOpIntoPhi(I, PN))
return NV;
if (Instruction *NV = foldFreezeIntoRecurrence(I, PN))
return NV;
}
if (Value *NI = pushFreezeToPreventPoisonFromPropagating(I))
return replaceInstUsesWith(I, NI);
// If I is freeze(undef), check its uses and fold it to a fixed constant.
// - or: pick -1
// - select's condition: if the true value is constant, choose it by making
// the condition true.
// - default: pick 0
//
// Note that this transform is intentionally done here rather than
// via an analysis in InstSimplify or at individual user sites. That is
// because we must produce the same value for all uses of the freeze -
// it's the reason "freeze" exists!
//
// TODO: This could use getBinopAbsorber() / getBinopIdentity() to avoid
// duplicating logic for binops at least.
auto getUndefReplacement = [&I](Type *Ty) {
Constant *BestValue = nullptr;
Constant *NullValue = Constant::getNullValue(Ty);
for (const auto *U : I.users()) {
Constant *C = NullValue;
if (match(U, m_Or(m_Value(), m_Value())))
C = ConstantInt::getAllOnesValue(Ty);
else if (match(U, m_Select(m_Specific(&I), m_Constant(), m_Value())))
C = ConstantInt::getTrue(Ty);
if (!BestValue)
BestValue = C;
else if (BestValue != C)
BestValue = NullValue;
}
assert(BestValue && "Must have at least one use");
return BestValue;
};
if (match(Op0, m_Undef())) {
// Don't fold freeze(undef/poison) if it's used as a vector operand in
// a shuffle. This may improve codegen for shuffles that allow
// unspecified inputs.
if (isUsedWithinShuffleVector(&I))
return nullptr;
return replaceInstUsesWith(I, getUndefReplacement(I.getType()));
}
Constant *C;
if (match(Op0, m_Constant(C)) && C->containsUndefOrPoisonElement()) {
Constant *ReplaceC = getUndefReplacement(I.getType()->getScalarType());
return replaceInstUsesWith(I, Constant::replaceUndefsWith(C, ReplaceC));
}
// Replace uses of Op with freeze(Op).
if (freezeOtherUses(I))
return &I;
return nullptr;
}
/// Check for case where the call writes to an otherwise dead alloca. This
/// shows up for unused out-params in idiomatic C/C++ code. Note that this
/// helper *only* analyzes the write; doesn't check any other legality aspect.
static bool SoleWriteToDeadLocal(Instruction *I, TargetLibraryInfo &TLI) {
auto *CB = dyn_cast<CallBase>(I);
if (!CB)
// TODO: handle e.g. store to alloca here - only worth doing if we extend
// to allow reload along used path as described below. Otherwise, this
// is simply a store to a dead allocation which will be removed.
return false;
std::optional<MemoryLocation> Dest = MemoryLocation::getForDest(CB, TLI);
if (!Dest)
return false;
auto *AI = dyn_cast<AllocaInst>(getUnderlyingObject(Dest->Ptr));
if (!AI)
// TODO: allow malloc?
return false;
// TODO: allow memory access dominated by move point? Note that since AI
// could have a reference to itself captured by the call, we would need to
// account for cycles in doing so.
SmallVector<const User *> AllocaUsers;
SmallPtrSet<const User *, 4> Visited;
auto pushUsers = [&](const Instruction &I) {
for (const User *U : I.users()) {
if (Visited.insert(U).second)
AllocaUsers.push_back(U);
}
};
pushUsers(*AI);
while (!AllocaUsers.empty()) {
auto *UserI = cast<Instruction>(AllocaUsers.pop_back_val());
if (isa<BitCastInst>(UserI) || isa<GetElementPtrInst>(UserI) ||
isa<AddrSpaceCastInst>(UserI)) {
pushUsers(*UserI);
continue;
}
if (UserI == CB)
continue;
// TODO: support lifetime.start/end here
return false;
}
return true;
}
/// Try to move the specified instruction from its current block into the
/// beginning of DestBlock, which can only happen if it's safe to move the
/// instruction past all of the instructions between it and the end of its
/// block.
bool InstCombinerImpl::tryToSinkInstruction(Instruction *I,
BasicBlock *DestBlock) {
BasicBlock *SrcBlock = I->getParent();
// Cannot move control-flow-involving, volatile loads, vaarg, etc.
if (isa<PHINode>(I) || I->isEHPad() || I->mayThrow() || !I->willReturn() ||
I->isTerminator())
return false;
// Do not sink static or dynamic alloca instructions. Static allocas must
// remain in the entry block, and dynamic allocas must not be sunk in between
// a stacksave / stackrestore pair, which would incorrectly shorten its
// lifetime.
if (isa<AllocaInst>(I))
return false;
// Do not sink into catchswitch blocks.
if (isa<CatchSwitchInst>(DestBlock->getTerminator()))
return false;
// Do not sink convergent call instructions.
if (auto *CI = dyn_cast<CallInst>(I)) {
if (CI->isConvergent())
return false;
}
// Unless we can prove that the memory write isn't visibile except on the
// path we're sinking to, we must bail.
if (I->mayWriteToMemory()) {
if (!SoleWriteToDeadLocal(I, TLI))
return false;
}
// We can only sink load instructions if there is nothing between the load and
// the end of block that could change the value.
if (I->mayReadFromMemory()) {
// We don't want to do any sophisticated alias analysis, so we only check
// the instructions after I in I's parent block if we try to sink to its
// successor block.
if (DestBlock->getUniquePredecessor() != I->getParent())
return false;
for (BasicBlock::iterator Scan = std::next(I->getIterator()),
E = I->getParent()->end();
Scan != E; ++Scan)
if (Scan->mayWriteToMemory())
return false;
}
I->dropDroppableUses([&](const Use *U) {
auto *I = dyn_cast<Instruction>(U->getUser());
if (I && I->getParent() != DestBlock) {
Worklist.add(I);
return true;
}
return false;
});
/// FIXME: We could remove droppable uses that are not dominated by
/// the new position.
BasicBlock::iterator InsertPos = DestBlock->getFirstInsertionPt();
I->moveBefore(*DestBlock, InsertPos);
++NumSunkInst;
// Also sink all related debug uses from the source basic block. Otherwise we
// get debug use before the def. Attempt to salvage debug uses first, to
// maximise the range variables have location for. If we cannot salvage, then
// mark the location undef: we know it was supposed to receive a new location
// here, but that computation has been sunk.
SmallVector<DbgVariableIntrinsic *, 2> DbgUsers;
findDbgUsers(DbgUsers, I);
// For all debug values in the destination block, the sunk instruction
// will still be available, so they do not need to be dropped.
SmallVector<DbgVariableIntrinsic *, 2> DbgUsersToSalvage;
SmallVector<DPValue *, 2> DPValuesToSalvage;
for (auto &DbgUser : DbgUsers)
if (DbgUser->getParent() != DestBlock)
DbgUsersToSalvage.push_back(DbgUser);
// Process the sinking DbgUsersToSalvage in reverse order, as we only want
// to clone the last appearing debug intrinsic for each given variable.
SmallVector<DbgVariableIntrinsic *, 2> DbgUsersToSink;
for (DbgVariableIntrinsic *DVI : DbgUsersToSalvage)
if (DVI->getParent() == SrcBlock)
DbgUsersToSink.push_back(DVI);
llvm::sort(DbgUsersToSink,
[](auto *A, auto *B) { return B->comesBefore(A); });
SmallVector<DbgVariableIntrinsic *, 2> DIIClones;
SmallSet<DebugVariable, 4> SunkVariables;
for (auto *User : DbgUsersToSink) {
// A dbg.declare instruction should not be cloned, since there can only be
// one per variable fragment. It should be left in the original place
// because the sunk instruction is not an alloca (otherwise we could not be
// here).
if (isa<DbgDeclareInst>(User))
continue;
DebugVariable DbgUserVariable =
DebugVariable(User->getVariable(), User->getExpression(),
User->getDebugLoc()->getInlinedAt());
if (!SunkVariables.insert(DbgUserVariable).second)
continue;
// Leave dbg.assign intrinsics in their original positions and there should
// be no need to insert a clone.
if (isa<DbgAssignIntrinsic>(User))
continue;
DIIClones.emplace_back(cast<DbgVariableIntrinsic>(User->clone()));
if (isa<DbgDeclareInst>(User) && isa<CastInst>(I))
DIIClones.back()->replaceVariableLocationOp(I, I->getOperand(0));
LLVM_DEBUG(dbgs() << "CLONE: " << *DIIClones.back() << '\n');
}
// Perform salvaging without the clones, then sink the clones.
if (!DIIClones.empty()) {
// RemoveDIs: pass in empty vector of DPValues until we get to instrumenting
// this pass.
SmallVector<DPValue *, 1> DummyDPValues;
salvageDebugInfoForDbgValues(*I, DbgUsersToSalvage, DummyDPValues);
// The clones are in reverse order of original appearance, reverse again to
// maintain the original order.
for (auto &DIIClone : llvm::reverse(DIIClones)) {
DIIClone->insertBefore(&*InsertPos);
LLVM_DEBUG(dbgs() << "SINK: " << *DIIClone << '\n');
}
}
return true;
}
bool InstCombinerImpl::run() {
while (!Worklist.isEmpty()) {
// Walk deferred instructions in reverse order, and push them to the
// worklist, which means they'll end up popped from the worklist in-order.
while (Instruction *I = Worklist.popDeferred()) {
// Check to see if we can DCE the instruction. We do this already here to
// reduce the number of uses and thus allow other folds to trigger.
// Note that eraseInstFromFunction() may push additional instructions on
// the deferred worklist, so this will DCE whole instruction chains.
if (isInstructionTriviallyDead(I, &TLI)) {
eraseInstFromFunction(*I);
++NumDeadInst;
continue;
}
Worklist.push(I);
}
Instruction *I = Worklist.removeOne();
if (I == nullptr) continue; // skip null values.
// Check to see if we can DCE the instruction.
if (isInstructionTriviallyDead(I, &TLI)) {
eraseInstFromFunction(*I);
++NumDeadInst;
continue;
}
if (!DebugCounter::shouldExecute(VisitCounter))
continue;
// See if we can trivially sink this instruction to its user if we can
// prove that the successor is not executed more frequently than our block.
// Return the UserBlock if successful.
auto getOptionalSinkBlockForInst =
[this](Instruction *I) -> std::optional<BasicBlock *> {
if (!EnableCodeSinking)
return std::nullopt;
BasicBlock *BB = I->getParent();
BasicBlock *UserParent = nullptr;
unsigned NumUsers = 0;
for (auto *U : I->users()) {
if (U->isDroppable())
continue;
if (NumUsers > MaxSinkNumUsers)
return std::nullopt;
Instruction *UserInst = cast<Instruction>(U);
// Special handling for Phi nodes - get the block the use occurs in.
if (PHINode *PN = dyn_cast<PHINode>(UserInst)) {
for (unsigned i = 0; i < PN->getNumIncomingValues(); i++) {
if (PN->getIncomingValue(i) == I) {
// Bail out if we have uses in different blocks. We don't do any
// sophisticated analysis (i.e finding NearestCommonDominator of
// these use blocks).
if (UserParent && UserParent != PN->getIncomingBlock(i))
return std::nullopt;
UserParent = PN->getIncomingBlock(i);
}
}
assert(UserParent && "expected to find user block!");
} else {
if (UserParent && UserParent != UserInst->getParent())
return std::nullopt;
UserParent = UserInst->getParent();
}
// Make sure these checks are done only once, naturally we do the checks
// the first time we get the userparent, this will save compile time.
if (NumUsers == 0) {
// Try sinking to another block. If that block is unreachable, then do
// not bother. SimplifyCFG should handle it.
if (UserParent == BB || !DT.isReachableFromEntry(UserParent))
return std::nullopt;
auto *Term = UserParent->getTerminator();
// See if the user is one of our successors that has only one
// predecessor, so that we don't have to split the critical edge.
// Another option where we can sink is a block that ends with a
// terminator that does not pass control to other block (such as
// return or unreachable or resume). In this case:
// - I dominates the User (by SSA form);
// - the User will be executed at most once.
// So sinking I down to User is always profitable or neutral.
if (UserParent->getUniquePredecessor() != BB && !succ_empty(Term))
return std::nullopt;
assert(DT.dominates(BB, UserParent) && "Dominance relation broken?");
}
NumUsers++;
}
// No user or only has droppable users.
if (!UserParent)
return std::nullopt;
return UserParent;
};
auto OptBB = getOptionalSinkBlockForInst(I);
if (OptBB) {
auto *UserParent = *OptBB;
// Okay, the CFG is simple enough, try to sink this instruction.
if (tryToSinkInstruction(I, UserParent)) {
LLVM_DEBUG(dbgs() << "IC: Sink: " << *I << '\n');
MadeIRChange = true;
// We'll add uses of the sunk instruction below, but since
// sinking can expose opportunities for it's *operands* add
// them to the worklist
for (Use &U : I->operands())
if (Instruction *OpI = dyn_cast<Instruction>(U.get()))
Worklist.push(OpI);
}
}
// Now that we have an instruction, try combining it to simplify it.
Builder.SetInsertPoint(I);
Builder.CollectMetadataToCopy(
I, {LLVMContext::MD_dbg, LLVMContext::MD_annotation});
#ifndef NDEBUG
std::string OrigI;
#endif
LLVM_DEBUG(raw_string_ostream SS(OrigI); I->print(SS); OrigI = SS.str(););
LLVM_DEBUG(dbgs() << "IC: Visiting: " << OrigI << '\n');
if (Instruction *Result = visit(*I)) {
++NumCombined;
// Should we replace the old instruction with a new one?
if (Result != I) {
LLVM_DEBUG(dbgs() << "IC: Old = " << *I << '\n'
<< " New = " << *Result << '\n');
Result->copyMetadata(*I,
{LLVMContext::MD_dbg, LLVMContext::MD_annotation});
// Everything uses the new instruction now.
I->replaceAllUsesWith(Result);
// Move the name to the new instruction first.
Result->takeName(I);
// Insert the new instruction into the basic block...
BasicBlock *InstParent = I->getParent();
BasicBlock::iterator InsertPos = I->getIterator();
// Are we replace a PHI with something that isn't a PHI, or vice versa?
if (isa<PHINode>(Result) != isa<PHINode>(I)) {
// We need to fix up the insertion point.
if (isa<PHINode>(I)) // PHI -> Non-PHI
InsertPos = InstParent->getFirstInsertionPt();
else // Non-PHI -> PHI
InsertPos = InstParent->getFirstNonPHI()->getIterator();
}
Result->insertInto(InstParent, InsertPos);
// Push the new instruction and any users onto the worklist.
Worklist.pushUsersToWorkList(*Result);
Worklist.push(Result);
eraseInstFromFunction(*I);
} else {
LLVM_DEBUG(dbgs() << "IC: Mod = " << OrigI << '\n'
<< " New = " << *I << '\n');
// If the instruction was modified, it's possible that it is now dead.
// if so, remove it.
if (isInstructionTriviallyDead(I, &TLI)) {
eraseInstFromFunction(*I);
} else {
Worklist.pushUsersToWorkList(*I);
Worklist.push(I);
}
}
MadeIRChange = true;
}
}
Worklist.zap();
return MadeIRChange;
}
// Track the scopes used by !alias.scope and !noalias. In a function, a
// @llvm.experimental.noalias.scope.decl is only useful if that scope is used
// by both sets. If not, the declaration of the scope can be safely omitted.
// The MDNode of the scope can be omitted as well for the instructions that are
// part of this function. We do not do that at this point, as this might become
// too time consuming to do.
class AliasScopeTracker {
SmallPtrSet<const MDNode *, 8> UsedAliasScopesAndLists;
SmallPtrSet<const MDNode *, 8> UsedNoAliasScopesAndLists;
public:
void analyse(Instruction *I) {
// This seems to be faster than checking 'mayReadOrWriteMemory()'.
if (!I->hasMetadataOtherThanDebugLoc())
return;
auto Track = [](Metadata *ScopeList, auto &Container) {
const auto *MDScopeList = dyn_cast_or_null<MDNode>(ScopeList);
if (!MDScopeList || !Container.insert(MDScopeList).second)
return;
for (const auto &MDOperand : MDScopeList->operands())
if (auto *MDScope = dyn_cast<MDNode>(MDOperand))
Container.insert(MDScope);
};
Track(I->getMetadata(LLVMContext::MD_alias_scope), UsedAliasScopesAndLists);
Track(I->getMetadata(LLVMContext::MD_noalias), UsedNoAliasScopesAndLists);
}
bool isNoAliasScopeDeclDead(Instruction *Inst) {
NoAliasScopeDeclInst *Decl = dyn_cast<NoAliasScopeDeclInst>(Inst);
if (!Decl)
return false;
assert(Decl->use_empty() &&
"llvm.experimental.noalias.scope.decl in use ?");
const MDNode *MDSL = Decl->getScopeList();
assert(MDSL->getNumOperands() == 1 &&
"llvm.experimental.noalias.scope should refer to a single scope");
auto &MDOperand = MDSL->getOperand(0);
if (auto *MD = dyn_cast<MDNode>(MDOperand))
return !UsedAliasScopesAndLists.contains(MD) ||
!UsedNoAliasScopesAndLists.contains(MD);
// Not an MDNode ? throw away.
return true;
}
};
/// Populate the IC worklist from a function, by walking it in reverse
/// post-order and adding all reachable code to the worklist.
///
/// This has a couple of tricks to make the code faster and more powerful. In
/// particular, we constant fold and DCE instructions as we go, to avoid adding
/// them to the worklist (this significantly speeds up instcombine on code where
/// many instructions are dead or constant). Additionally, if we find a branch
/// whose condition is a known constant, we only visit the reachable successors.
bool InstCombinerImpl::prepareWorklist(
Function &F, ReversePostOrderTraversal<BasicBlock *> &RPOT) {
bool MadeIRChange = false;
SmallPtrSet<BasicBlock *, 32> LiveBlocks;
SmallVector<Instruction *, 128> InstrsForInstructionWorklist;
DenseMap<Constant *, Constant *> FoldedConstants;
AliasScopeTracker SeenAliasScopes;
auto HandleOnlyLiveSuccessor = [&](BasicBlock *BB, BasicBlock *LiveSucc) {
for (BasicBlock *Succ : successors(BB))
if (Succ != LiveSucc && DeadEdges.insert({BB, Succ}).second)
for (PHINode &PN : Succ->phis())
for (Use &U : PN.incoming_values())
if (PN.getIncomingBlock(U) == BB && !isa<PoisonValue>(U)) {
U.set(PoisonValue::get(PN.getType()));
MadeIRChange = true;
}
};
for (BasicBlock *BB : RPOT) {
if (!BB->isEntryBlock() && all_of(predecessors(BB), [&](BasicBlock *Pred) {
return DeadEdges.contains({Pred, BB}) || DT.dominates(BB, Pred);
})) {
HandleOnlyLiveSuccessor(BB, nullptr);
continue;
}
LiveBlocks.insert(BB);
for (Instruction &Inst : llvm::make_early_inc_range(*BB)) {
// ConstantProp instruction if trivially constant.
if (!Inst.use_empty() &&
(Inst.getNumOperands() == 0 || isa<Constant>(Inst.getOperand(0))))
if (Constant *C = ConstantFoldInstruction(&Inst, DL, &TLI)) {
LLVM_DEBUG(dbgs() << "IC: ConstFold to: " << *C << " from: " << Inst
<< '\n');
Inst.replaceAllUsesWith(C);
++NumConstProp;
if (isInstructionTriviallyDead(&Inst, &TLI))
Inst.eraseFromParent();
MadeIRChange = true;
continue;
}
// See if we can constant fold its operands.
for (Use &U : Inst.operands()) {
if (!isa<ConstantVector>(U) && !isa<ConstantExpr>(U))
continue;
auto *C = cast<Constant>(U);
Constant *&FoldRes = FoldedConstants[C];
if (!FoldRes)
FoldRes = ConstantFoldConstant(C, DL, &TLI);
if (FoldRes != C) {
LLVM_DEBUG(dbgs() << "IC: ConstFold operand of: " << Inst
<< "\n Old = " << *C
<< "\n New = " << *FoldRes << '\n');
U = FoldRes;
MadeIRChange = true;
}
}
// Skip processing debug and pseudo intrinsics in InstCombine. Processing
// these call instructions consumes non-trivial amount of time and
// provides no value for the optimization.
if (!Inst.isDebugOrPseudoInst()) {
InstrsForInstructionWorklist.push_back(&Inst);
SeenAliasScopes.analyse(&Inst);
}
}
// If this is a branch or switch on a constant, mark only the single
// live successor. Otherwise assume all successors are live.
Instruction *TI = BB->getTerminator();
if (BranchInst *BI = dyn_cast<BranchInst>(TI); BI && BI->isConditional()) {
if (isa<UndefValue>(BI->getCondition())) {
// Branch on undef is UB.
HandleOnlyLiveSuccessor(BB, nullptr);
continue;
}
if (auto *Cond = dyn_cast<ConstantInt>(BI->getCondition())) {
bool CondVal = Cond->getZExtValue();
HandleOnlyLiveSuccessor(BB, BI->getSuccessor(!CondVal));
continue;
}
} else if (SwitchInst *SI = dyn_cast<SwitchInst>(TI)) {
if (isa<UndefValue>(SI->getCondition())) {
// Switch on undef is UB.
HandleOnlyLiveSuccessor(BB, nullptr);
continue;
}
if (auto *Cond = dyn_cast<ConstantInt>(SI->getCondition())) {
HandleOnlyLiveSuccessor(BB,
SI->findCaseValue(Cond)->getCaseSuccessor());
continue;
}
}
}
// Remove instructions inside unreachable blocks. This prevents the
// instcombine code from having to deal with some bad special cases, and
// reduces use counts of instructions.
for (BasicBlock &BB : F) {
if (LiveBlocks.count(&BB))
continue;
unsigned NumDeadInstInBB;
unsigned NumDeadDbgInstInBB;
std::tie(NumDeadInstInBB, NumDeadDbgInstInBB) =
removeAllNonTerminatorAndEHPadInstructions(&BB);
MadeIRChange |= NumDeadInstInBB + NumDeadDbgInstInBB > 0;
NumDeadInst += NumDeadInstInBB;
}
// Once we've found all of the instructions to add to instcombine's worklist,
// add them in reverse order. This way instcombine will visit from the top
// of the function down. This jives well with the way that it adds all uses
// of instructions to the worklist after doing a transformation, thus avoiding
// some N^2 behavior in pathological cases.
Worklist.reserve(InstrsForInstructionWorklist.size());
for (Instruction *Inst : reverse(InstrsForInstructionWorklist)) {
// DCE instruction if trivially dead. As we iterate in reverse program
// order here, we will clean up whole chains of dead instructions.
if (isInstructionTriviallyDead(Inst, &TLI) ||
SeenAliasScopes.isNoAliasScopeDeclDead(Inst)) {
++NumDeadInst;
LLVM_DEBUG(dbgs() << "IC: DCE: " << *Inst << '\n');
salvageDebugInfo(*Inst);
Inst->eraseFromParent();
MadeIRChange = true;
continue;
}
Worklist.push(Inst);
}
return MadeIRChange;
}
static bool combineInstructionsOverFunction(
Function &F, InstructionWorklist &Worklist, AliasAnalysis *AA,
AssumptionCache &AC, TargetLibraryInfo &TLI, TargetTransformInfo &TTI,
DominatorTree &DT, OptimizationRemarkEmitter &ORE, BlockFrequencyInfo *BFI,
ProfileSummaryInfo *PSI, LoopInfo *LI, const InstCombineOptions &Opts) {
auto &DL = F.getParent()->getDataLayout();
/// Builder - This is an IRBuilder that automatically inserts new
/// instructions into the worklist when they are created.
IRBuilder<TargetFolder, IRBuilderCallbackInserter> Builder(
F.getContext(), TargetFolder(DL),
IRBuilderCallbackInserter([&Worklist, &AC](Instruction *I) {
Worklist.add(I);
if (auto *Assume = dyn_cast<AssumeInst>(I))
AC.registerAssumption(Assume);
}));
ReversePostOrderTraversal<BasicBlock *> RPOT(&F.front());
// Lower dbg.declare intrinsics otherwise their value may be clobbered
// by instcombiner.
bool MadeIRChange = false;
if (ShouldLowerDbgDeclare)
MadeIRChange = LowerDbgDeclare(F);
// Iterate while there is work to do.
unsigned Iteration = 0;
while (true) {
++Iteration;
if (Iteration > Opts.MaxIterations && !Opts.VerifyFixpoint) {
LLVM_DEBUG(dbgs() << "\n\n[IC] Iteration limit #" << Opts.MaxIterations
<< " on " << F.getName()
<< " reached; stopping without verifying fixpoint\n");
break;
}
++NumWorklistIterations;
LLVM_DEBUG(dbgs() << "\n\nINSTCOMBINE ITERATION #" << Iteration << " on "
<< F.getName() << "\n");
InstCombinerImpl IC(Worklist, Builder, F.hasMinSize(), AA, AC, TLI, TTI, DT,
ORE, BFI, PSI, DL, LI);
IC.MaxArraySizeForCombine = MaxArraySize;
bool MadeChangeInThisIteration = IC.prepareWorklist(F, RPOT);
MadeChangeInThisIteration |= IC.run();
if (!MadeChangeInThisIteration)
break;
MadeIRChange = true;
if (Iteration > Opts.MaxIterations) {
report_fatal_error(
"Instruction Combining did not reach a fixpoint after " +
Twine(Opts.MaxIterations) + " iterations");
}
}
if (Iteration == 1)
++NumOneIteration;
else if (Iteration == 2)
++NumTwoIterations;
else if (Iteration == 3)
++NumThreeIterations;
else
++NumFourOrMoreIterations;
return MadeIRChange;
}
InstCombinePass::InstCombinePass(InstCombineOptions Opts) : Options(Opts) {}
void InstCombinePass::printPipeline(
raw_ostream &OS, function_ref<StringRef(StringRef)> MapClassName2PassName) {
static_cast<PassInfoMixin<InstCombinePass> *>(this)->printPipeline(
OS, MapClassName2PassName);
OS << '<';
OS << "max-iterations=" << Options.MaxIterations << ";";
OS << (Options.UseLoopInfo ? "" : "no-") << "use-loop-info;";
OS << (Options.VerifyFixpoint ? "" : "no-") << "verify-fixpoint";
OS << '>';
}
PreservedAnalyses InstCombinePass::run(Function &F,
FunctionAnalysisManager &AM) {
auto &AC = AM.getResult<AssumptionAnalysis>(F);
auto &DT = AM.getResult<DominatorTreeAnalysis>(F);
auto &TLI = AM.getResult<TargetLibraryAnalysis>(F);
auto &ORE = AM.getResult<OptimizationRemarkEmitterAnalysis>(F);
auto &TTI = AM.getResult<TargetIRAnalysis>(F);
// TODO: Only use LoopInfo when the option is set. This requires that the
// callers in the pass pipeline explicitly set the option.
auto *LI = AM.getCachedResult<LoopAnalysis>(F);
if (!LI && Options.UseLoopInfo)
LI = &AM.getResult<LoopAnalysis>(F);
auto *AA = &AM.getResult<AAManager>(F);
auto &MAMProxy = AM.getResult<ModuleAnalysisManagerFunctionProxy>(F);
ProfileSummaryInfo *PSI =
MAMProxy.getCachedResult<ProfileSummaryAnalysis>(*F.getParent());
auto *BFI = (PSI && PSI->hasProfileSummary()) ?
&AM.getResult<BlockFrequencyAnalysis>(F) : nullptr;
if (!combineInstructionsOverFunction(F, Worklist, AA, AC, TLI, TTI, DT, ORE,
BFI, PSI, LI, Options))
// No changes, all analyses are preserved.
return PreservedAnalyses::all();
// Mark all the analyses that instcombine updates as preserved.
PreservedAnalyses PA;
PA.preserveSet<CFGAnalyses>();
return PA;
}
void InstructionCombiningPass::getAnalysisUsage(AnalysisUsage &AU) const {
AU.setPreservesCFG();
AU.addRequired<AAResultsWrapperPass>();
AU.addRequired<AssumptionCacheTracker>();
AU.addRequired<TargetLibraryInfoWrapperPass>();
AU.addRequired<TargetTransformInfoWrapperPass>();
AU.addRequired<DominatorTreeWrapperPass>();
AU.addRequired<OptimizationRemarkEmitterWrapperPass>();
AU.addPreserved<DominatorTreeWrapperPass>();
AU.addPreserved<AAResultsWrapperPass>();
AU.addPreserved<BasicAAWrapperPass>();
AU.addPreserved<GlobalsAAWrapperPass>();
AU.addRequired<ProfileSummaryInfoWrapperPass>();
LazyBlockFrequencyInfoPass::getLazyBFIAnalysisUsage(AU);
}
bool InstructionCombiningPass::runOnFunction(Function &F) {
if (skipFunction(F))
return false;
// Required analyses.
auto AA = &getAnalysis<AAResultsWrapperPass>().getAAResults();
auto &AC = getAnalysis<AssumptionCacheTracker>().getAssumptionCache(F);
auto &TLI = getAnalysis<TargetLibraryInfoWrapperPass>().getTLI(F);
auto &TTI = getAnalysis<TargetTransformInfoWrapperPass>().getTTI(F);
auto &DT = getAnalysis<DominatorTreeWrapperPass>().getDomTree();
auto &ORE = getAnalysis<OptimizationRemarkEmitterWrapperPass>().getORE();
// Optional analyses.
auto *LIWP = getAnalysisIfAvailable<LoopInfoWrapperPass>();
auto *LI = LIWP ? &LIWP->getLoopInfo() : nullptr;
ProfileSummaryInfo *PSI =
&getAnalysis<ProfileSummaryInfoWrapperPass>().getPSI();
BlockFrequencyInfo *BFI =
(PSI && PSI->hasProfileSummary()) ?
&getAnalysis<LazyBlockFrequencyInfoPass>().getBFI() :
nullptr;
return combineInstructionsOverFunction(F, Worklist, AA, AC, TLI, TTI, DT, ORE,
BFI, PSI, LI, InstCombineOptions());
}
char InstructionCombiningPass::ID = 0;
InstructionCombiningPass::InstructionCombiningPass() : FunctionPass(ID) {
initializeInstructionCombiningPassPass(*PassRegistry::getPassRegistry());
}
INITIALIZE_PASS_BEGIN(InstructionCombiningPass, "instcombine",
"Combine redundant instructions", false, false)
INITIALIZE_PASS_DEPENDENCY(AssumptionCacheTracker)
INITIALIZE_PASS_DEPENDENCY(TargetLibraryInfoWrapperPass)
INITIALIZE_PASS_DEPENDENCY(TargetTransformInfoWrapperPass)
INITIALIZE_PASS_DEPENDENCY(DominatorTreeWrapperPass)
INITIALIZE_PASS_DEPENDENCY(AAResultsWrapperPass)
INITIALIZE_PASS_DEPENDENCY(GlobalsAAWrapperPass)
INITIALIZE_PASS_DEPENDENCY(OptimizationRemarkEmitterWrapperPass)
INITIALIZE_PASS_DEPENDENCY(LazyBlockFrequencyInfoPass)
INITIALIZE_PASS_DEPENDENCY(ProfileSummaryInfoWrapperPass)
INITIALIZE_PASS_END(InstructionCombiningPass, "instcombine",
"Combine redundant instructions", false, false)
// Initialization Routines
void llvm::initializeInstCombine(PassRegistry &Registry) {
initializeInstructionCombiningPassPass(Registry);
}
FunctionPass *llvm::createInstructionCombiningPass() {
return new InstructionCombiningPass();
}
|