1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
2189
2190
2191
2192
2193
2194
2195
2196
2197
2198
2199
2200
2201
2202
2203
2204
2205
2206
2207
2208
2209
2210
2211
2212
2213
2214
2215
2216
2217
2218
2219
2220
2221
2222
2223
2224
2225
2226
2227
2228
2229
2230
2231
2232
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
2244
2245
2246
2247
2248
2249
2250
2251
2252
2253
2254
2255
2256
2257
2258
2259
2260
2261
2262
2263
2264
2265
2266
2267
2268
2269
2270
2271
2272
2273
2274
2275
2276
2277
2278
2279
2280
2281
2282
2283
2284
2285
2286
2287
2288
2289
2290
2291
2292
2293
2294
2295
2296
2297
2298
2299
2300
2301
2302
2303
2304
2305
2306
2307
2308
2309
2310
2311
2312
2313
2314
2315
2316
2317
2318
2319
2320
2321
2322
2323
2324
2325
2326
2327
2328
2329
2330
2331
2332
2333
2334
2335
2336
2337
2338
2339
2340
2341
2342
2343
2344
2345
2346
2347
2348
2349
2350
2351
2352
2353
2354
2355
2356
2357
2358
2359
2360
2361
2362
2363
2364
2365
2366
2367
2368
2369
2370
2371
2372
2373
2374
2375
2376
2377
2378
2379
2380
2381
2382
2383
2384
2385
2386
2387
2388
2389
2390
2391
2392
2393
2394
2395
2396
2397
2398
2399
2400
2401
2402
2403
2404
2405
2406
2407
2408
2409
2410
2411
2412
2413
2414
2415
2416
2417
2418
2419
2420
2421
2422
2423
2424
2425
2426
2427
2428
2429
2430
2431
2432
2433
2434
2435
2436
2437
2438
2439
2440
2441
2442
2443
2444
2445
2446
2447
2448
2449
2450
2451
2452
2453
2454
2455
2456
2457
2458
2459
2460
2461
2462
2463
2464
2465
2466
2467
2468
2469
2470
2471
2472
2473
2474
2475
2476
2477
2478
2479
2480
2481
2482
2483
2484
2485
2486
2487
2488
2489
2490
2491
2492
2493
2494
2495
2496
2497
2498
2499
2500
2501
2502
2503
2504
2505
2506
2507
2508
2509
2510
2511
2512
2513
2514
2515
2516
2517
2518
2519
2520
2521
2522
2523
2524
2525
2526
2527
2528
2529
2530
2531
2532
2533
2534
2535
2536
2537
2538
2539
2540
2541
2542
2543
2544
2545
2546
2547
2548
2549
2550
2551
2552
2553
2554
2555
2556
2557
2558
2559
2560
2561
2562
2563
2564
2565
2566
2567
2568
2569
2570
2571
2572
2573
2574
2575
2576
2577
2578
2579
2580
2581
2582
2583
2584
2585
2586
2587
2588
2589
2590
2591
2592
2593
2594
2595
2596
2597
2598
2599
2600
2601
2602
2603
2604
2605
2606
2607
2608
2609
2610
2611
2612
2613
2614
2615
2616
2617
2618
2619
2620
2621
2622
2623
2624
2625
2626
2627
2628
2629
2630
2631
2632
2633
2634
2635
2636
2637
2638
2639
2640
2641
2642
2643
2644
2645
2646
2647
2648
2649
2650
2651
2652
2653
2654
2655
2656
2657
2658
2659
2660
2661
2662
2663
2664
2665
2666
2667
2668
2669
2670
2671
2672
2673
2674
2675
2676
2677
2678
2679
2680
2681
2682
2683
2684
2685
2686
2687
2688
2689
2690
2691
2692
2693
2694
2695
2696
2697
2698
2699
2700
2701
2702
2703
2704
2705
2706
2707
2708
2709
2710
2711
2712
2713
2714
2715
2716
2717
2718
2719
2720
2721
2722
2723
2724
2725
2726
2727
2728
2729
2730
2731
2732
2733
2734
2735
2736
2737
2738
2739
2740
2741
2742
2743
2744
2745
2746
2747
2748
2749
2750
2751
2752
2753
2754
2755
2756
2757
2758
2759
2760
2761
2762
2763
2764
2765
2766
2767
2768
2769
2770
2771
2772
2773
2774
2775
2776
2777
2778
2779
2780
2781
2782
2783
2784
2785
2786
2787
2788
2789
2790
2791
2792
2793
2794
2795
2796
2797
2798
2799
2800
2801
2802
2803
2804
2805
2806
2807
2808
2809
2810
2811
2812
2813
2814
2815
2816
2817
2818
2819
2820
2821
2822
2823
2824
2825
2826
2827
2828
2829
2830
2831
2832
2833
2834
2835
2836
2837
2838
2839
2840
2841
2842
2843
2844
2845
2846
2847
2848
2849
2850
2851
2852
2853
2854
2855
2856
2857
2858
2859
2860
2861
2862
2863
2864
2865
2866
2867
2868
2869
2870
2871
2872
2873
2874
2875
2876
2877
2878
2879
2880
2881
2882
2883
2884
2885
2886
2887
2888
2889
2890
2891
2892
2893
2894
2895
2896
2897
2898
2899
2900
2901
2902
2903
2904
2905
2906
2907
2908
2909
2910
2911
2912
2913
2914
2915
2916
2917
2918
2919
2920
2921
2922
2923
2924
2925
2926
2927
2928
2929
2930
2931
2932
2933
2934
2935
2936
2937
2938
2939
2940
2941
2942
2943
2944
2945
2946
2947
2948
2949
2950
2951
2952
2953
2954
2955
2956
2957
2958
2959
2960
2961
2962
2963
2964
2965
2966
2967
2968
2969
2970
2971
2972
2973
2974
2975
2976
2977
2978
2979
2980
2981
2982
2983
2984
2985
2986
2987
2988
2989
2990
2991
2992
2993
2994
2995
2996
2997
2998
2999
3000
3001
3002
3003
3004
3005
3006
3007
3008
3009
3010
3011
3012
3013
3014
3015
3016
3017
3018
3019
3020
3021
3022
3023
3024
3025
3026
3027
3028
3029
3030
3031
3032
3033
3034
3035
3036
3037
3038
3039
3040
3041
3042
3043
3044
3045
3046
3047
3048
3049
3050
3051
3052
3053
3054
3055
3056
3057
3058
3059
3060
3061
3062
3063
3064
3065
3066
3067
3068
3069
3070
3071
3072
3073
3074
3075
3076
3077
3078
3079
3080
3081
3082
3083
3084
3085
3086
3087
3088
3089
3090
3091
3092
3093
3094
3095
3096
3097
3098
3099
3100
3101
3102
3103
3104
3105
3106
3107
3108
3109
3110
3111
3112
3113
3114
3115
3116
3117
3118
3119
3120
3121
3122
3123
3124
3125
3126
3127
3128
3129
3130
3131
3132
3133
3134
3135
3136
3137
3138
3139
3140
3141
3142
3143
3144
3145
3146
3147
3148
3149
3150
3151
3152
3153
3154
3155
3156
3157
3158
3159
3160
3161
3162
3163
3164
3165
3166
3167
3168
3169
3170
3171
3172
3173
3174
3175
3176
3177
3178
3179
3180
3181
3182
3183
3184
3185
3186
3187
3188
3189
3190
3191
3192
3193
3194
3195
3196
3197
3198
3199
3200
3201
3202
3203
3204
3205
3206
3207
3208
3209
3210
3211
3212
3213
3214
3215
3216
3217
3218
3219
3220
3221
3222
3223
3224
3225
3226
3227
3228
3229
3230
3231
3232
3233
3234
3235
3236
3237
3238
3239
3240
3241
3242
3243
3244
3245
3246
3247
3248
3249
3250
3251
3252
3253
3254
3255
3256
3257
3258
3259
3260
3261
3262
3263
3264
3265
3266
3267
3268
3269
3270
3271
3272
3273
3274
3275
3276
3277
3278
3279
3280
3281
3282
3283
3284
3285
3286
3287
3288
3289
3290
3291
3292
3293
3294
3295
3296
3297
3298
3299
3300
3301
3302
3303
3304
3305
3306
3307
3308
3309
3310
3311
3312
3313
3314
3315
3316
3317
3318
3319
3320
3321
3322
3323
3324
3325
3326
3327
3328
3329
3330
3331
3332
3333
3334
3335
3336
3337
3338
3339
3340
3341
3342
3343
3344
3345
3346
3347
3348
3349
3350
3351
3352
3353
3354
3355
3356
3357
3358
3359
3360
3361
3362
3363
3364
3365
3366
3367
3368
3369
3370
3371
3372
3373
3374
3375
3376
3377
3378
3379
3380
3381
3382
3383
3384
3385
3386
3387
3388
3389
3390
3391
3392
3393
3394
3395
3396
3397
3398
3399
3400
3401
3402
3403
3404
3405
3406
3407
3408
3409
3410
3411
3412
3413
3414
3415
3416
3417
3418
3419
3420
3421
3422
3423
3424
3425
3426
3427
3428
3429
3430
3431
3432
3433
3434
3435
3436
3437
3438
3439
3440
3441
3442
3443
3444
3445
3446
3447
3448
3449
3450
3451
3452
3453
3454
3455
3456
3457
3458
3459
3460
3461
3462
3463
3464
3465
3466
3467
3468
3469
3470
3471
3472
3473
3474
3475
3476
3477
3478
3479
3480
3481
3482
3483
3484
3485
3486
3487
3488
3489
3490
3491
3492
3493
3494
3495
3496
3497
3498
3499
3500
3501
3502
3503
3504
3505
3506
3507
3508
3509
3510
3511
3512
3513
3514
3515
3516
3517
3518
3519
3520
3521
3522
3523
3524
3525
3526
3527
3528
3529
3530
3531
3532
3533
3534
3535
3536
3537
3538
3539
3540
3541
3542
3543
3544
3545
3546
3547
3548
3549
3550
3551
3552
3553
3554
3555
3556
3557
3558
3559
3560
3561
3562
3563
3564
3565
3566
3567
3568
3569
3570
3571
3572
3573
3574
3575
3576
3577
3578
3579
3580
3581
3582
3583
3584
3585
3586
3587
3588
3589
3590
3591
3592
3593
3594
3595
3596
3597
3598
3599
3600
3601
3602
3603
3604
3605
3606
3607
3608
3609
3610
3611
3612
3613
3614
3615
3616
3617
3618
3619
3620
3621
3622
3623
3624
3625
3626
3627
3628
3629
3630
3631
3632
3633
3634
3635
3636
3637
3638
3639
3640
3641
3642
3643
3644
3645
3646
3647
3648
3649
3650
3651
3652
3653
3654
3655
3656
3657
3658
3659
3660
3661
3662
3663
3664
3665
3666
3667
3668
3669
3670
3671
3672
3673
3674
3675
3676
3677
3678
3679
3680
3681
3682
3683
3684
3685
3686
3687
3688
3689
3690
3691
3692
3693
3694
3695
3696
3697
3698
3699
3700
3701
3702
3703
3704
3705
3706
3707
3708
3709
3710
3711
3712
3713
3714
3715
3716
3717
3718
3719
3720
3721
3722
3723
3724
3725
3726
3727
3728
3729
3730
3731
3732
3733
3734
3735
3736
3737
3738
3739
3740
3741
3742
3743
3744
3745
3746
3747
3748
3749
3750
3751
3752
3753
3754
3755
3756
3757
3758
3759
3760
3761
3762
3763
3764
3765
3766
3767
3768
3769
3770
3771
3772
3773
3774
3775
3776
3777
3778
3779
3780
3781
3782
3783
3784
3785
3786
3787
3788
3789
3790
3791
3792
3793
3794
3795
3796
3797
3798
3799
3800
3801
3802
3803
3804
3805
3806
3807
3808
3809
3810
3811
3812
3813
3814
3815
3816
3817
3818
3819
3820
3821
3822
3823
3824
3825
3826
3827
3828
3829
3830
3831
3832
3833
3834
3835
3836
3837
3838
3839
3840
3841
3842
3843
3844
3845
3846
3847
3848
3849
3850
3851
3852
3853
3854
3855
3856
3857
3858
3859
3860
3861
3862
3863
3864
3865
3866
3867
3868
3869
3870
3871
3872
3873
3874
3875
3876
3877
3878
3879
3880
3881
3882
3883
3884
3885
3886
3887
3888
3889
3890
3891
3892
3893
3894
3895
3896
3897
3898
3899
3900
3901
3902
3903
3904
3905
3906
3907
3908
3909
3910
3911
3912
3913
3914
3915
3916
3917
3918
3919
3920
3921
3922
3923
3924
3925
3926
3927
3928
3929
3930
3931
3932
3933
3934
3935
3936
3937
3938
3939
3940
3941
3942
3943
3944
3945
3946
3947
3948
3949
3950
3951
3952
3953
3954
3955
3956
3957
3958
3959
3960
3961
3962
3963
3964
3965
3966
3967
3968
3969
3970
3971
3972
3973
3974
3975
3976
3977
3978
3979
3980
3981
3982
3983
3984
3985
3986
3987
3988
3989
3990
3991
3992
3993
3994
3995
3996
3997
3998
3999
4000
4001
4002
4003
4004
4005
4006
4007
4008
4009
4010
4011
4012
4013
4014
4015
4016
4017
4018
4019
4020
4021
4022
4023
4024
4025
4026
4027
4028
4029
4030
4031
4032
4033
4034
4035
4036
4037
4038
4039
4040
4041
4042
4043
4044
4045
4046
4047
4048
4049
4050
4051
4052
4053
4054
4055
4056
4057
4058
4059
4060
4061
4062
4063
4064
4065
4066
4067
4068
4069
4070
4071
4072
4073
4074
4075
4076
4077
4078
4079
4080
4081
4082
4083
4084
4085
4086
4087
4088
4089
4090
4091
4092
4093
4094
4095
4096
4097
4098
4099
4100
4101
4102
4103
4104
4105
4106
4107
4108
4109
4110
4111
4112
4113
4114
4115
4116
4117
4118
4119
4120
4121
4122
4123
4124
4125
4126
4127
4128
4129
4130
4131
4132
4133
4134
4135
4136
4137
4138
4139
4140
4141
4142
4143
4144
4145
4146
4147
4148
4149
4150
4151
4152
4153
4154
4155
4156
4157
4158
4159
4160
4161
4162
4163
4164
4165
4166
4167
4168
4169
4170
4171
4172
4173
4174
4175
4176
4177
4178
4179
4180
4181
4182
4183
4184
4185
4186
4187
4188
4189
4190
4191
4192
4193
4194
4195
4196
4197
4198
4199
4200
4201
4202
4203
4204
4205
4206
4207
4208
4209
4210
4211
4212
4213
4214
4215
4216
4217
4218
4219
4220
4221
4222
4223
4224
4225
4226
4227
4228
4229
4230
4231
4232
4233
4234
4235
4236
4237
4238
4239
4240
4241
4242
4243
4244
4245
4246
4247
4248
4249
4250
4251
4252
4253
4254
4255
4256
4257
4258
4259
4260
4261
4262
4263
4264
4265
4266
4267
4268
4269
4270
4271
4272
4273
4274
4275
4276
4277
4278
4279
4280
4281
4282
4283
4284
4285
4286
4287
4288
4289
4290
4291
4292
4293
4294
4295
4296
4297
4298
4299
4300
4301
4302
4303
4304
4305
4306
4307
4308
4309
4310
4311
4312
4313
4314
4315
4316
4317
4318
4319
4320
4321
4322
4323
4324
4325
4326
4327
4328
4329
4330
4331
4332
4333
4334
4335
4336
4337
4338
4339
4340
4341
4342
4343
4344
4345
4346
4347
4348
4349
4350
4351
4352
4353
4354
4355
4356
4357
4358
4359
4360
4361
4362
4363
4364
4365
4366
4367
4368
4369
4370
4371
4372
4373
4374
4375
4376
4377
4378
4379
4380
4381
4382
4383
4384
4385
4386
4387
4388
4389
4390
4391
4392
4393
4394
4395
4396
4397
4398
4399
4400
4401
4402
4403
4404
4405
4406
4407
4408
4409
4410
4411
4412
4413
4414
4415
4416
4417
4418
4419
4420
4421
4422
4423
4424
4425
4426
4427
4428
4429
4430
4431
4432
4433
4434
4435
4436
4437
4438
4439
4440
4441
4442
4443
4444
4445
4446
4447
4448
4449
4450
4451
4452
4453
4454
4455
4456
4457
4458
4459
4460
4461
4462
4463
4464
4465
4466
4467
4468
4469
4470
4471
4472
4473
4474
4475
4476
4477
4478
4479
4480
4481
4482
4483
4484
4485
4486
4487
4488
4489
4490
4491
4492
4493
4494
4495
4496
4497
4498
4499
4500
4501
4502
4503
4504
4505
4506
4507
4508
4509
4510
4511
4512
4513
4514
4515
4516
4517
4518
4519
4520
4521
4522
4523
4524
4525
4526
4527
4528
4529
4530
4531
4532
4533
4534
4535
4536
4537
4538
4539
4540
4541
4542
4543
4544
4545
4546
4547
4548
4549
4550
4551
4552
4553
4554
4555
4556
4557
4558
4559
4560
4561
4562
4563
4564
4565
4566
4567
4568
4569
4570
4571
4572
4573
4574
4575
4576
4577
4578
4579
4580
4581
4582
4583
4584
4585
4586
4587
4588
4589
4590
4591
4592
4593
4594
4595
4596
4597
4598
4599
4600
4601
4602
4603
4604
4605
4606
4607
4608
4609
4610
4611
4612
4613
4614
4615
4616
4617
4618
4619
4620
4621
4622
4623
4624
4625
4626
4627
4628
4629
4630
4631
4632
4633
4634
4635
4636
4637
4638
4639
4640
4641
4642
4643
4644
4645
4646
4647
4648
4649
4650
4651
4652
4653
4654
4655
4656
4657
4658
4659
4660
4661
4662
4663
4664
4665
4666
4667
4668
4669
4670
4671
4672
4673
4674
4675
4676
4677
4678
4679
4680
4681
4682
4683
4684
4685
4686
4687
4688
4689
4690
4691
4692
4693
4694
4695
4696
4697
4698
4699
4700
4701
4702
4703
4704
4705
4706
4707
4708
4709
4710
4711
4712
4713
4714
4715
4716
4717
4718
4719
4720
4721
4722
4723
4724
4725
4726
4727
4728
4729
4730
4731
4732
4733
4734
4735
4736
4737
4738
4739
4740
4741
4742
4743
4744
4745
4746
4747
4748
4749
4750
4751
4752
4753
4754
4755
4756
4757
4758
4759
4760
4761
4762
4763
4764
4765
4766
4767
4768
4769
4770
4771
4772
4773
4774
4775
4776
4777
4778
4779
4780
4781
4782
4783
4784
4785
4786
4787
4788
4789
4790
4791
4792
4793
4794
4795
4796
4797
4798
4799
4800
4801
4802
4803
4804
4805
4806
4807
4808
4809
4810
4811
4812
4813
4814
4815
4816
4817
4818
4819
4820
4821
4822
4823
4824
4825
4826
4827
4828
4829
4830
4831
4832
4833
4834
4835
4836
4837
4838
4839
4840
4841
4842
4843
4844
4845
4846
4847
4848
4849
4850
4851
4852
4853
4854
4855
4856
4857
4858
4859
4860
4861
4862
4863
4864
4865
4866
4867
4868
4869
4870
4871
4872
4873
4874
4875
4876
4877
4878
4879
4880
4881
4882
4883
4884
4885
4886
4887
4888
4889
4890
4891
4892
4893
4894
4895
4896
4897
4898
4899
4900
4901
4902
4903
4904
4905
4906
4907
4908
4909
4910
4911
4912
4913
4914
4915
4916
4917
4918
4919
4920
4921
4922
4923
4924
4925
4926
4927
4928
4929
4930
4931
4932
4933
4934
4935
4936
4937
4938
4939
4940
4941
4942
4943
4944
4945
4946
4947
4948
4949
4950
4951
4952
4953
4954
4955
4956
4957
4958
4959
4960
4961
4962
4963
4964
4965
4966
4967
4968
4969
4970
4971
4972
4973
4974
4975
4976
4977
4978
4979
4980
4981
4982
4983
4984
4985
4986
4987
4988
4989
4990
4991
4992
4993
4994
4995
4996
4997
4998
4999
5000
5001
5002
5003
5004
5005
5006
5007
5008
5009
5010
5011
5012
5013
5014
5015
5016
5017
5018
5019
5020
5021
5022
5023
5024
5025
5026
5027
5028
5029
5030
5031
5032
5033
5034
5035
5036
5037
5038
5039
5040
5041
5042
5043
5044
5045
5046
5047
5048
5049
5050
5051
5052
5053
5054
5055
5056
5057
5058
5059
5060
5061
5062
5063
5064
5065
5066
5067
5068
5069
5070
5071
5072
5073
5074
5075
5076
5077
5078
5079
5080
5081
5082
5083
5084
5085
5086
5087
5088
5089
5090
5091
5092
5093
5094
5095
5096
5097
5098
5099
5100
5101
5102
5103
5104
5105
5106
5107
5108
5109
5110
5111
5112
5113
5114
5115
5116
5117
5118
5119
5120
5121
5122
5123
5124
5125
5126
5127
5128
5129
5130
5131
5132
5133
5134
5135
5136
5137
5138
5139
5140
5141
5142
5143
5144
5145
5146
5147
5148
5149
5150
5151
5152
5153
5154
5155
5156
5157
5158
5159
5160
5161
5162
5163
5164
5165
5166
5167
5168
5169
5170
5171
5172
5173
5174
5175
5176
5177
5178
5179
5180
5181
5182
5183
5184
5185
5186
5187
5188
5189
5190
5191
5192
5193
5194
5195
5196
5197
5198
5199
5200
5201
5202
5203
5204
5205
5206
5207
5208
5209
5210
5211
5212
5213
5214
5215
5216
5217
5218
5219
5220
5221
5222
5223
5224
5225
5226
5227
5228
5229
5230
5231
5232
5233
5234
5235
5236
5237
5238
5239
5240
5241
5242
5243
5244
5245
5246
5247
5248
5249
5250
5251
5252
5253
5254
5255
5256
5257
5258
5259
5260
5261
5262
5263
5264
5265
5266
5267
5268
5269
5270
5271
5272
5273
5274
5275
5276
5277
5278
5279
5280
5281
5282
5283
5284
5285
5286
5287
5288
5289
5290
5291
5292
5293
5294
5295
5296
5297
5298
5299
5300
5301
5302
5303
5304
5305
5306
5307
5308
5309
5310
5311
5312
5313
5314
5315
5316
5317
5318
5319
5320
5321
5322
5323
5324
5325
5326
5327
5328
5329
5330
5331
5332
5333
5334
5335
5336
5337
5338
5339
5340
5341
5342
5343
5344
5345
5346
5347
5348
5349
5350
5351
5352
5353
5354
5355
5356
5357
5358
5359
5360
5361
5362
5363
5364
5365
5366
5367
5368
5369
5370
5371
5372
5373
5374
5375
5376
5377
5378
5379
5380
5381
5382
5383
5384
5385
5386
5387
5388
5389
5390
5391
5392
5393
5394
5395
5396
5397
5398
5399
5400
5401
5402
5403
5404
5405
5406
5407
5408
5409
5410
5411
5412
5413
5414
5415
5416
5417
5418
5419
5420
5421
5422
5423
5424
5425
5426
5427
5428
5429
5430
5431
5432
5433
5434
5435
5436
5437
5438
5439
5440
5441
5442
5443
5444
5445
5446
5447
5448
5449
5450
5451
5452
5453
5454
5455
5456
5457
5458
5459
5460
5461
5462
5463
5464
5465
5466
5467
5468
5469
5470
5471
5472
5473
5474
5475
5476
5477
5478
5479
5480
5481
5482
5483
5484
5485
5486
5487
5488
5489
5490
5491
5492
5493
5494
5495
5496
5497
5498
5499
5500
5501
5502
5503
5504
5505
5506
5507
5508
5509
5510
5511
5512
5513
5514
5515
5516
5517
5518
5519
5520
5521
5522
5523
5524
5525
5526
5527
5528
5529
5530
5531
5532
5533
5534
5535
5536
5537
5538
5539
5540
5541
5542
5543
5544
5545
5546
5547
5548
5549
5550
5551
5552
5553
5554
5555
5556
5557
5558
5559
5560
5561
5562
5563
5564
5565
5566
5567
5568
5569
5570
5571
5572
5573
5574
5575
5576
5577
5578
5579
5580
5581
5582
5583
5584
5585
5586
5587
5588
5589
5590
5591
5592
5593
5594
5595
5596
5597
5598
5599
5600
5601
5602
5603
5604
5605
5606
5607
5608
5609
5610
5611
5612
5613
5614
5615
5616
5617
5618
5619
5620
5621
5622
5623
5624
5625
5626
5627
5628
5629
5630
5631
5632
5633
5634
5635
5636
5637
5638
5639
5640
5641
5642
5643
5644
5645
5646
5647
5648
5649
5650
5651
5652
5653
5654
5655
5656
5657
5658
5659
5660
5661
5662
5663
5664
5665
5666
5667
5668
5669
5670
5671
5672
5673
5674
5675
5676
5677
5678
5679
5680
5681
5682
5683
5684
5685
5686
5687
5688
5689
5690
5691
5692
5693
5694
5695
5696
5697
5698
5699
5700
5701
5702
5703
5704
5705
5706
5707
5708
5709
5710
5711
5712
5713
5714
5715
5716
5717
5718
5719
5720
5721
5722
5723
5724
5725
5726
5727
5728
5729
5730
5731
5732
5733
5734
5735
5736
5737
5738
5739
5740
5741
5742
5743
5744
5745
5746
5747
5748
5749
5750
5751
5752
5753
5754
5755
5756
5757
5758
5759
5760
5761
5762
5763
5764
5765
5766
5767
5768
5769
5770
5771
5772
5773
5774
5775
5776
5777
5778
5779
5780
5781
5782
5783
5784
5785
5786
5787
5788
5789
5790
5791
5792
5793
5794
5795
5796
5797
5798
5799
5800
5801
5802
5803
5804
5805
5806
5807
5808
5809
5810
5811
5812
5813
5814
5815
5816
5817
5818
5819
5820
5821
5822
5823
5824
5825
5826
5827
5828
5829
5830
5831
5832
5833
5834
5835
5836
5837
5838
5839
5840
5841
5842
5843
5844
5845
5846
5847
5848
5849
5850
5851
5852
5853
5854
5855
5856
5857
5858
5859
5860
5861
5862
5863
5864
5865
5866
5867
5868
5869
5870
5871
5872
5873
5874
5875
5876
5877
5878
5879
5880
5881
5882
5883
5884
5885
5886
5887
5888
5889
5890
5891
5892
5893
5894
5895
5896
5897
5898
5899
5900
5901
5902
5903
5904
5905
5906
5907
5908
5909
5910
5911
5912
5913
5914
5915
5916
5917
5918
5919
5920
5921
5922
5923
5924
5925
5926
5927
5928
5929
5930
5931
5932
5933
5934
5935
5936
5937
5938
5939
5940
5941
5942
5943
5944
5945
5946
5947
5948
5949
5950
5951
5952
5953
5954
5955
5956
5957
5958
5959
5960
5961
5962
5963
5964
5965
5966
5967
5968
5969
5970
5971
5972
5973
5974
5975
5976
5977
5978
5979
5980
5981
5982
5983
5984
5985
5986
5987
5988
5989
5990
5991
5992
5993
5994
5995
5996
5997
5998
5999
6000
6001
6002
6003
6004
6005
6006
6007
6008
6009
6010
6011
6012
6013
6014
6015
6016
6017
6018
6019
6020
6021
6022
6023
6024
6025
6026
6027
6028
6029
6030
6031
6032
6033
6034
6035
6036
6037
6038
6039
6040
6041
6042
6043
6044
6045
6046
6047
6048
6049
6050
6051
6052
6053
6054
6055
6056
6057
6058
6059
6060
6061
6062
6063
6064
6065
6066
6067
6068
6069
6070
6071
6072
6073
6074
6075
6076
6077
6078
6079
6080
6081
6082
6083
6084
6085
6086
6087
6088
6089
6090
6091
6092
6093
6094
6095
6096
6097
6098
6099
6100
6101
6102
6103
6104
6105
6106
6107
6108
6109
6110
6111
6112
6113
6114
6115
6116
6117
6118
6119
6120
6121
6122
6123
6124
6125
6126
6127
6128
6129
6130
6131
6132
6133
6134
6135
6136
6137
6138
6139
6140
6141
6142
6143
6144
6145
6146
6147
6148
6149
6150
6151
6152
6153
6154
6155
6156
6157
6158
6159
6160
6161
6162
6163
6164
6165
6166
6167
6168
6169
6170
6171
6172
6173
6174
6175
6176
6177
6178
6179
6180
6181
6182
6183
6184
6185
6186
6187
6188
6189
6190
6191
6192
6193
6194
6195
6196
6197
6198
6199
6200
6201
6202
6203
6204
6205
6206
6207
6208
6209
6210
6211
6212
6213
6214
6215
6216
6217
6218
6219
6220
6221
6222
6223
6224
6225
6226
6227
6228
6229
6230
6231
6232
6233
6234
6235
6236
6237
6238
6239
6240
6241
6242
6243
6244
6245
6246
6247
6248
6249
6250
6251
6252
6253
6254
6255
6256
6257
6258
6259
6260
6261
6262
6263
6264
6265
6266
6267
6268
6269
6270
6271
6272
6273
6274
6275
6276
6277
6278
6279
6280
6281
6282
6283
6284
6285
6286
6287
6288
6289
6290
6291
6292
6293
6294
6295
6296
6297
6298
6299
6300
6301
6302
6303
6304
6305
6306
6307
6308
6309
6310
6311
6312
6313
6314
6315
6316
6317
6318
6319
6320
6321
6322
6323
6324
6325
6326
6327
6328
6329
6330
6331
6332
6333
6334
6335
6336
6337
6338
6339
6340
6341
6342
6343
6344
6345
6346
6347
6348
6349
6350
6351
6352
6353
6354
6355
6356
6357
6358
6359
6360
6361
6362
6363
6364
6365
6366
6367
6368
6369
6370
6371
6372
6373
6374
6375
6376
6377
6378
6379
6380
6381
6382
6383
6384
6385
6386
6387
6388
6389
6390
6391
6392
6393
6394
6395
6396
6397
6398
6399
6400
6401
6402
6403
6404
6405
6406
6407
6408
6409
6410
6411
6412
6413
6414
6415
6416
6417
6418
6419
6420
6421
6422
6423
6424
6425
6426
6427
6428
6429
6430
6431
6432
6433
6434
6435
6436
6437
6438
6439
6440
6441
6442
6443
6444
6445
6446
6447
6448
6449
6450
6451
6452
6453
6454
6455
6456
6457
6458
6459
6460
6461
6462
6463
6464
6465
6466
6467
6468
6469
6470
6471
6472
6473
6474
6475
6476
6477
6478
6479
6480
6481
6482
6483
6484
6485
6486
6487
6488
6489
6490
6491
6492
6493
6494
6495
6496
6497
6498
6499
6500
6501
6502
6503
6504
6505
6506
6507
6508
6509
6510
6511
6512
6513
6514
6515
6516
6517
6518
6519
6520
6521
6522
6523
6524
6525
6526
6527
6528
6529
6530
6531
6532
6533
6534
6535
6536
6537
6538
6539
6540
6541
6542
6543
6544
6545
6546
6547
6548
6549
6550
6551
6552
6553
6554
6555
6556
6557
6558
6559
6560
6561
6562
6563
6564
6565
6566
6567
6568
6569
6570
6571
6572
6573
6574
6575
6576
6577
6578
6579
6580
6581
6582
6583
6584
6585
6586
6587
6588
6589
6590
6591
6592
6593
6594
6595
6596
6597
6598
6599
6600
6601
6602
6603
6604
6605
6606
6607
6608
6609
6610
6611
6612
6613
6614
6615
6616
6617
6618
6619
6620
6621
6622
6623
6624
6625
6626
6627
6628
6629
6630
6631
6632
6633
6634
6635
6636
6637
6638
6639
6640
6641
6642
6643
6644
6645
6646
6647
6648
6649
6650
6651
6652
6653
6654
6655
6656
6657
6658
6659
6660
6661
6662
6663
6664
6665
6666
6667
6668
6669
6670
6671
6672
6673
6674
6675
6676
6677
6678
6679
6680
6681
6682
6683
6684
6685
6686
6687
6688
6689
6690
6691
6692
6693
6694
6695
6696
6697
6698
6699
6700
6701
6702
6703
6704
6705
6706
6707
6708
6709
6710
6711
6712
6713
6714
6715
6716
6717
6718
6719
6720
6721
6722
6723
6724
6725
6726
6727
6728
6729
6730
6731
6732
6733
6734
6735
6736
6737
6738
6739
6740
6741
6742
6743
6744
6745
6746
6747
6748
6749
6750
6751
6752
6753
6754
6755
6756
6757
6758
6759
6760
6761
6762
6763
6764
6765
6766
6767
6768
6769
6770
6771
6772
6773
6774
6775
6776
6777
6778
6779
6780
6781
6782
6783
6784
6785
6786
6787
6788
6789
6790
6791
6792
6793
6794
6795
6796
6797
6798
6799
6800
6801
6802
6803
6804
6805
6806
6807
6808
6809
6810
6811
6812
6813
6814
6815
6816
6817
6818
6819
6820
6821
6822
6823
6824
6825
6826
6827
6828
6829
6830
6831
6832
6833
6834
6835
6836
6837
6838
6839
6840
6841
6842
6843
6844
6845
6846
6847
6848
6849
6850
6851
6852
6853
6854
6855
6856
6857
6858
6859
6860
6861
6862
6863
6864
6865
6866
6867
6868
6869
6870
6871
6872
6873
6874
6875
6876
6877
6878
6879
6880
6881
6882
6883
6884
6885
6886
6887
6888
6889
6890
6891
6892
6893
6894
6895
6896
6897
6898
6899
6900
6901
6902
6903
6904
6905
6906
6907
6908
6909
6910
6911
6912
6913
6914
6915
6916
6917
6918
6919
6920
6921
6922
6923
6924
6925
6926
6927
6928
6929
6930
6931
6932
6933
6934
6935
6936
6937
6938
6939
6940
6941
6942
6943
6944
6945
6946
6947
6948
6949
6950
6951
6952
6953
6954
6955
6956
6957
6958
6959
6960
6961
6962
6963
6964
6965
6966
6967
6968
6969
6970
6971
6972
6973
6974
6975
6976
6977
6978
6979
6980
6981
6982
6983
6984
6985
6986
6987
6988
6989
6990
6991
6992
6993
6994
6995
6996
6997
6998
6999
7000
7001
7002
7003
7004
7005
7006
7007
7008
7009
7010
7011
7012
7013
7014
7015
7016
7017
7018
7019
7020
7021
7022
7023
7024
7025
7026
7027
7028
7029
7030
7031
7032
7033
7034
7035
7036
7037
7038
7039
7040
7041
7042
7043
7044
7045
7046
7047
7048
7049
7050
7051
7052
7053
7054
7055
7056
7057
7058
7059
7060
7061
7062
7063
7064
7065
7066
7067
7068
7069
7070
7071
7072
7073
7074
7075
7076
7077
7078
7079
7080
7081
7082
7083
7084
7085
7086
7087
7088
7089
7090
7091
7092
7093
7094
7095
7096
7097
7098
7099
7100
7101
7102
7103
7104
7105
7106
7107
7108
7109
7110
7111
7112
7113
7114
7115
7116
7117
7118
7119
7120
7121
7122
7123
7124
7125
7126
7127
7128
7129
7130
7131
7132
7133
7134
7135
7136
7137
7138
7139
7140
7141
7142
7143
7144
7145
7146
7147
7148
7149
7150
7151
7152
7153
7154
7155
7156
7157
7158
7159
7160
7161
7162
7163
7164
7165
7166
7167
7168
7169
7170
7171
7172
7173
7174
7175
7176
7177
7178
7179
7180
7181
7182
7183
7184
7185
7186
7187
7188
7189
7190
7191
7192
7193
7194
7195
7196
7197
7198
7199
7200
7201
7202
7203
7204
7205
7206
7207
7208
7209
7210
7211
7212
7213
7214
7215
7216
7217
7218
7219
7220
7221
7222
7223
7224
7225
7226
7227
7228
7229
7230
7231
7232
7233
7234
7235
7236
7237
7238
7239
7240
7241
7242
7243
7244
7245
7246
7247
7248
7249
7250
7251
7252
7253
7254
7255
7256
7257
7258
7259
7260
7261
7262
7263
7264
7265
7266
7267
7268
7269
7270
7271
7272
7273
7274
7275
7276
7277
7278
7279
7280
7281
7282
7283
7284
7285
7286
7287
7288
7289
7290
7291
7292
7293
7294
7295
7296
7297
7298
7299
7300
7301
7302
7303
7304
7305
7306
7307
7308
7309
7310
7311
7312
7313
7314
7315
7316
7317
7318
7319
7320
7321
7322
7323
7324
7325
7326
7327
7328
7329
7330
7331
7332
7333
7334
7335
7336
7337
7338
7339
7340
7341
7342
7343
7344
7345
7346
7347
7348
7349
7350
7351
7352
7353
7354
7355
7356
7357
7358
7359
7360
7361
7362
7363
7364
7365
7366
7367
7368
7369
7370
7371
7372
7373
7374
7375
7376
7377
7378
7379
7380
7381
7382
7383
7384
7385
7386
7387
7388
7389
7390
7391
7392
7393
7394
7395
7396
7397
7398
7399
7400
7401
7402
7403
7404
7405
7406
7407
7408
7409
7410
7411
7412
7413
7414
7415
7416
7417
7418
7419
7420
7421
7422
7423
7424
7425
7426
7427
7428
7429
7430
7431
7432
7433
7434
7435
7436
7437
7438
7439
7440
7441
7442
7443
7444
7445
7446
7447
7448
7449
7450
7451
7452
7453
7454
7455
7456
7457
7458
7459
7460
7461
7462
7463
7464
7465
7466
7467
7468
7469
7470
7471
7472
7473
7474
7475
7476
7477
7478
7479
7480
7481
7482
7483
7484
7485
7486
7487
7488
7489
7490
7491
7492
7493
7494
7495
7496
7497
7498
7499
7500
7501
7502
7503
7504
7505
7506
7507
7508
7509
7510
7511
7512
7513
7514
7515
7516
7517
7518
7519
7520
7521
7522
7523
7524
7525
7526
7527
7528
7529
7530
7531
7532
7533
7534
7535
7536
7537
7538
7539
7540
7541
7542
7543
7544
7545
7546
7547
7548
7549
7550
7551
7552
7553
7554
7555
7556
7557
7558
7559
7560
7561
7562
7563
7564
7565
7566
7567
7568
7569
7570
7571
7572
7573
7574
7575
7576
7577
7578
7579
7580
7581
7582
7583
7584
7585
7586
7587
7588
7589
7590
7591
7592
7593
7594
7595
7596
7597
7598
7599
7600
7601
7602
7603
7604
7605
7606
7607
7608
7609
7610
7611
7612
7613
7614
7615
7616
7617
7618
7619
7620
7621
7622
7623
7624
7625
7626
7627
7628
7629
7630
7631
7632
7633
7634
7635
7636
7637
7638
7639
7640
7641
7642
7643
7644
7645
7646
7647
7648
7649
7650
7651
7652
7653
7654
7655
7656
7657
7658
7659
7660
7661
7662
7663
7664
7665
7666
7667
7668
7669
7670
7671
7672
7673
7674
7675
7676
7677
7678
7679
7680
7681
7682
7683
7684
7685
7686
7687
7688
7689
7690
7691
7692
7693
7694
7695
7696
7697
7698
7699
7700
7701
7702
7703
7704
7705
7706
7707
7708
7709
7710
7711
7712
7713
7714
7715
7716
7717
7718
7719
7720
7721
7722
7723
7724
7725
7726
7727
7728
7729
7730
7731
7732
7733
7734
7735
7736
7737
7738
7739
7740
7741
7742
7743
7744
7745
7746
7747
7748
7749
7750
7751
7752
7753
7754
7755
7756
7757
7758
7759
7760
7761
7762
7763
7764
7765
7766
7767
7768
7769
7770
7771
7772
7773
7774
7775
7776
7777
7778
7779
7780
7781
7782
7783
7784
7785
7786
7787
7788
7789
7790
7791
7792
7793
7794
7795
7796
7797
7798
7799
7800
7801
7802
7803
7804
7805
7806
7807
7808
7809
7810
7811
7812
7813
7814
7815
7816
7817
7818
7819
7820
7821
7822
7823
7824
7825
7826
7827
7828
7829
7830
7831
7832
7833
7834
7835
7836
7837
7838
7839
7840
7841
7842
7843
7844
7845
7846
7847
7848
7849
7850
7851
7852
7853
7854
7855
7856
7857
7858
7859
7860
7861
7862
7863
7864
7865
7866
7867
7868
7869
7870
7871
7872
7873
7874
7875
7876
7877
7878
7879
7880
7881
7882
7883
7884
7885
7886
7887
7888
7889
7890
7891
7892
7893
7894
7895
7896
7897
7898
7899
7900
7901
7902
7903
7904
7905
7906
7907
7908
7909
7910
7911
7912
7913
7914
7915
7916
7917
7918
7919
7920
7921
7922
7923
7924
7925
7926
7927
7928
7929
7930
7931
7932
7933
7934
7935
7936
7937
7938
7939
7940
7941
7942
7943
7944
7945
7946
7947
7948
7949
7950
7951
7952
7953
7954
7955
7956
7957
7958
7959
7960
7961
7962
7963
7964
7965
7966
7967
7968
7969
7970
7971
7972
7973
7974
7975
7976
7977
7978
7979
7980
7981
7982
7983
7984
7985
7986
7987
7988
7989
7990
7991
7992
7993
7994
7995
7996
7997
7998
7999
8000
8001
8002
8003
8004
8005
8006
8007
8008
8009
8010
8011
8012
8013
8014
8015
8016
8017
8018
8019
8020
8021
8022
8023
8024
8025
8026
8027
8028
8029
8030
8031
8032
8033
8034
8035
8036
8037
8038
8039
8040
8041
8042
8043
8044
8045
8046
8047
8048
8049
8050
8051
8052
8053
8054
8055
8056
8057
8058
8059
8060
8061
8062
8063
8064
8065
8066
8067
8068
8069
8070
8071
8072
8073
8074
8075
8076
8077
8078
8079
8080
8081
8082
8083
8084
8085
8086
8087
8088
8089
8090
8091
8092
8093
8094
8095
8096
8097
8098
8099
8100
8101
8102
8103
8104
8105
8106
8107
8108
8109
8110
8111
8112
8113
8114
8115
8116
8117
8118
8119
8120
8121
8122
8123
8124
8125
8126
8127
8128
8129
8130
8131
8132
8133
8134
8135
8136
8137
8138
8139
8140
8141
8142
8143
8144
8145
8146
8147
8148
8149
8150
8151
8152
8153
8154
8155
8156
8157
8158
8159
8160
8161
8162
8163
8164
8165
8166
8167
8168
8169
8170
8171
8172
8173
8174
8175
8176
8177
8178
8179
8180
8181
8182
8183
8184
8185
8186
8187
8188
8189
8190
8191
8192
8193
8194
8195
8196
8197
8198
8199
8200
8201
8202
8203
8204
8205
8206
8207
8208
8209
8210
8211
8212
8213
8214
8215
8216
8217
8218
8219
8220
8221
8222
8223
8224
8225
8226
8227
8228
8229
8230
8231
8232
8233
8234
8235
8236
8237
8238
8239
8240
8241
8242
8243
8244
8245
8246
8247
8248
8249
8250
8251
8252
8253
8254
8255
8256
8257
8258
8259
8260
8261
8262
8263
8264
8265
8266
8267
8268
8269
8270
8271
8272
8273
8274
8275
8276
8277
8278
8279
8280
8281
8282
8283
8284
8285
8286
8287
8288
8289
8290
8291
8292
8293
8294
8295
8296
8297
8298
8299
8300
8301
8302
8303
8304
8305
8306
8307
8308
8309
8310
8311
8312
8313
8314
8315
8316
8317
8318
8319
8320
8321
8322
8323
8324
8325
8326
8327
8328
8329
8330
8331
8332
8333
8334
8335
8336
8337
8338
8339
8340
8341
8342
8343
8344
8345
8346
8347
8348
8349
8350
8351
8352
8353
8354
8355
8356
8357
8358
8359
8360
8361
8362
8363
8364
8365
8366
8367
8368
8369
8370
8371
8372
8373
8374
8375
8376
8377
8378
8379
8380
8381
8382
8383
8384
8385
8386
8387
8388
8389
8390
8391
8392
8393
8394
8395
8396
8397
8398
8399
8400
8401
8402
8403
8404
8405
8406
8407
8408
8409
8410
8411
8412
8413
8414
8415
8416
8417
8418
8419
8420
8421
8422
8423
8424
8425
8426
8427
8428
8429
8430
8431
8432
8433
8434
8435
8436
8437
8438
8439
8440
8441
8442
8443
8444
8445
8446
8447
8448
8449
8450
8451
8452
8453
8454
8455
8456
8457
8458
8459
8460
8461
8462
8463
8464
8465
8466
8467
8468
8469
8470
8471
8472
8473
8474
8475
8476
8477
8478
8479
8480
8481
8482
8483
8484
8485
8486
8487
8488
8489
8490
8491
8492
8493
8494
8495
8496
8497
8498
8499
8500
8501
8502
8503
8504
8505
8506
8507
8508
8509
8510
8511
8512
8513
8514
8515
8516
8517
8518
8519
8520
8521
8522
8523
8524
8525
8526
8527
8528
8529
8530
8531
8532
8533
8534
8535
8536
8537
8538
8539
8540
8541
8542
8543
8544
8545
8546
8547
8548
8549
8550
8551
8552
8553
8554
8555
8556
8557
8558
8559
8560
8561
8562
8563
8564
8565
8566
8567
8568
8569
8570
8571
8572
8573
8574
8575
8576
8577
8578
8579
8580
8581
8582
8583
8584
8585
8586
8587
8588
8589
8590
8591
8592
8593
8594
8595
8596
8597
8598
8599
8600
8601
8602
8603
8604
8605
8606
8607
8608
8609
8610
8611
8612
8613
8614
8615
8616
8617
8618
8619
8620
8621
8622
8623
8624
8625
8626
8627
8628
8629
8630
8631
8632
8633
8634
8635
8636
8637
8638
8639
8640
8641
8642
8643
8644
8645
8646
8647
8648
8649
8650
8651
8652
8653
8654
8655
8656
8657
8658
8659
8660
8661
8662
8663
8664
8665
8666
8667
8668
8669
8670
8671
8672
8673
8674
8675
8676
8677
8678
8679
8680
8681
8682
8683
8684
8685
8686
8687
8688
8689
8690
8691
8692
8693
8694
8695
8696
8697
8698
8699
8700
8701
8702
8703
8704
8705
8706
8707
8708
8709
8710
8711
8712
8713
8714
8715
8716
8717
8718
8719
8720
8721
8722
8723
8724
8725
8726
8727
8728
8729
8730
8731
8732
8733
8734
8735
8736
8737
8738
8739
8740
8741
8742
8743
8744
8745
8746
8747
8748
8749
8750
8751
8752
8753
8754
8755
8756
8757
8758
8759
8760
8761
8762
8763
8764
8765
8766
8767
8768
8769
8770
8771
8772
8773
8774
8775
8776
8777
8778
8779
8780
8781
8782
8783
8784
8785
8786
8787
8788
8789
8790
8791
8792
8793
8794
8795
8796
8797
8798
8799
8800
8801
8802
8803
8804
8805
8806
8807
8808
8809
8810
8811
8812
8813
8814
8815
8816
8817
8818
8819
8820
8821
8822
8823
8824
8825
8826
8827
8828
8829
8830
8831
8832
8833
8834
8835
8836
8837
8838
8839
8840
8841
8842
8843
8844
8845
8846
8847
8848
8849
8850
8851
8852
8853
8854
8855
8856
8857
8858
8859
8860
8861
8862
8863
8864
8865
8866
8867
8868
8869
8870
8871
8872
8873
8874
8875
8876
8877
8878
8879
8880
8881
8882
8883
8884
8885
8886
8887
8888
8889
8890
8891
8892
8893
8894
8895
8896
8897
8898
8899
8900
8901
8902
8903
8904
8905
8906
8907
8908
8909
8910
8911
8912
8913
8914
8915
8916
8917
8918
8919
8920
8921
8922
8923
8924
8925
8926
8927
8928
8929
8930
8931
8932
8933
8934
8935
8936
8937
8938
8939
8940
8941
8942
8943
8944
8945
8946
8947
8948
8949
8950
8951
8952
8953
8954
8955
8956
8957
8958
8959
8960
8961
8962
8963
8964
8965
8966
8967
8968
8969
8970
8971
8972
8973
8974
8975
8976
8977
8978
8979
8980
8981
8982
8983
8984
8985
8986
8987
8988
8989
8990
8991
8992
8993
8994
8995
8996
8997
8998
8999
9000
9001
9002
9003
9004
9005
9006
9007
9008
9009
9010
9011
9012
9013
9014
9015
9016
9017
9018
9019
9020
9021
9022
9023
9024
9025
9026
9027
9028
9029
9030
9031
9032
9033
9034
9035
9036
9037
9038
9039
9040
9041
9042
9043
9044
9045
9046
9047
9048
9049
9050
9051
9052
9053
9054
9055
9056
9057
9058
9059
9060
9061
9062
9063
9064
9065
9066
9067
9068
9069
9070
9071
9072
9073
9074
9075
9076
9077
9078
9079
9080
9081
9082
9083
9084
9085
9086
9087
9088
9089
9090
9091
9092
9093
9094
9095
9096
9097
9098
9099
9100
9101
9102
9103
9104
9105
9106
9107
9108
9109
9110
9111
9112
9113
9114
9115
9116
9117
9118
9119
9120
9121
9122
9123
9124
9125
9126
9127
9128
9129
9130
9131
9132
9133
9134
9135
9136
9137
9138
9139
9140
9141
9142
9143
9144
9145
9146
9147
9148
9149
9150
9151
9152
9153
9154
9155
9156
9157
9158
9159
9160
9161
9162
9163
9164
9165
9166
9167
9168
9169
9170
9171
9172
9173
9174
9175
9176
9177
9178
9179
9180
9181
9182
9183
9184
9185
9186
9187
9188
9189
9190
9191
9192
9193
9194
9195
9196
9197
9198
9199
9200
9201
9202
9203
9204
9205
9206
9207
9208
9209
9210
9211
9212
9213
9214
9215
9216
9217
9218
9219
9220
9221
9222
9223
9224
9225
9226
9227
9228
9229
9230
9231
9232
9233
9234
9235
9236
9237
9238
9239
9240
9241
9242
9243
9244
9245
9246
9247
9248
9249
9250
9251
9252
9253
9254
9255
9256
9257
9258
9259
9260
9261
9262
9263
9264
9265
9266
9267
9268
9269
9270
9271
9272
9273
9274
9275
9276
9277
9278
9279
9280
9281
9282
9283
9284
9285
9286
9287
9288
9289
9290
9291
9292
9293
9294
9295
9296
9297
9298
9299
9300
9301
9302
9303
9304
9305
9306
9307
9308
9309
9310
9311
9312
9313
9314
9315
9316
9317
9318
9319
9320
9321
9322
9323
9324
9325
9326
9327
9328
9329
9330
9331
9332
9333
9334
9335
9336
9337
9338
9339
9340
9341
9342
9343
9344
9345
9346
9347
9348
9349
9350
9351
9352
9353
9354
9355
9356
9357
9358
9359
9360
9361
9362
9363
9364
9365
9366
9367
9368
9369
9370
9371
9372
9373
9374
9375
9376
9377
9378
9379
9380
9381
9382
9383
9384
9385
9386
9387
9388
9389
9390
9391
9392
9393
9394
9395
9396
9397
9398
9399
9400
9401
9402
9403
9404
9405
9406
9407
9408
9409
9410
9411
9412
9413
9414
9415
9416
9417
9418
9419
9420
9421
9422
9423
9424
9425
9426
9427
9428
9429
9430
9431
9432
9433
9434
9435
9436
9437
9438
9439
9440
9441
9442
9443
9444
9445
9446
9447
9448
9449
9450
9451
9452
9453
9454
9455
9456
9457
9458
9459
9460
9461
9462
9463
9464
9465
9466
9467
9468
9469
9470
9471
9472
9473
9474
9475
9476
9477
9478
9479
9480
9481
9482
9483
9484
9485
9486
9487
9488
9489
9490
9491
9492
9493
9494
9495
9496
9497
9498
9499
9500
9501
9502
9503
9504
9505
9506
9507
9508
9509
9510
9511
9512
9513
9514
9515
9516
9517
9518
9519
9520
9521
9522
9523
9524
9525
9526
9527
9528
9529
9530
9531
9532
9533
9534
9535
9536
9537
9538
9539
9540
9541
9542
9543
9544
9545
9546
9547
9548
9549
9550
9551
9552
9553
9554
9555
9556
9557
9558
9559
9560
9561
9562
9563
9564
9565
9566
9567
9568
9569
9570
9571
9572
9573
9574
9575
9576
9577
9578
9579
9580
9581
9582
9583
9584
9585
9586
9587
9588
9589
9590
9591
9592
9593
9594
9595
9596
9597
9598
9599
9600
9601
9602
9603
9604
9605
9606
9607
9608
9609
9610
9611
9612
9613
9614
9615
9616
9617
9618
9619
9620
9621
9622
9623
9624
9625
9626
9627
9628
9629
9630
9631
9632
9633
9634
9635
9636
9637
9638
9639
9640
9641
9642
9643
9644
9645
9646
9647
9648
9649
9650
9651
9652
9653
9654
9655
9656
9657
9658
9659
9660
9661
9662
9663
9664
9665
9666
9667
9668
9669
9670
9671
9672
9673
9674
9675
9676
9677
9678
9679
9680
9681
9682
9683
9684
9685
9686
9687
9688
9689
9690
9691
9692
9693
9694
9695
9696
9697
9698
9699
9700
9701
9702
9703
9704
9705
9706
9707
9708
9709
9710
9711
9712
9713
9714
9715
9716
9717
9718
9719
9720
9721
9722
9723
9724
9725
9726
9727
9728
9729
9730
9731
9732
9733
9734
9735
9736
9737
9738
9739
9740
9741
9742
9743
9744
9745
9746
9747
9748
9749
9750
9751
9752
9753
9754
9755
9756
9757
9758
9759
9760
9761
9762
9763
9764
9765
9766
9767
9768
9769
9770
9771
9772
9773
9774
9775
9776
9777
9778
9779
9780
9781
9782
9783
9784
9785
9786
9787
9788
9789
9790
9791
9792
9793
9794
9795
9796
9797
9798
9799
9800
9801
9802
9803
9804
9805
9806
9807
9808
9809
9810
9811
9812
9813
9814
9815
9816
9817
9818
9819
9820
9821
9822
9823
9824
9825
9826
9827
9828
9829
9830
9831
9832
9833
9834
9835
9836
9837
9838
9839
9840
9841
9842
9843
9844
9845
9846
9847
9848
9849
9850
9851
9852
9853
9854
9855
9856
9857
9858
9859
9860
9861
9862
9863
9864
9865
9866
9867
9868
9869
9870
9871
9872
9873
9874
9875
9876
9877
9878
9879
9880
9881
9882
9883
9884
9885
9886
9887
9888
9889
9890
9891
9892
9893
9894
9895
9896
9897
9898
9899
9900
9901
9902
9903
9904
9905
9906
9907
9908
9909
9910
9911
9912
9913
9914
9915
9916
9917
9918
9919
9920
9921
9922
9923
9924
9925
9926
9927
9928
9929
9930
9931
9932
9933
9934
9935
9936
9937
9938
9939
9940
9941
9942
9943
9944
9945
9946
9947
9948
9949
9950
9951
9952
9953
9954
9955
9956
9957
9958
9959
9960
9961
9962
9963
9964
9965
9966
9967
9968
9969
9970
9971
9972
9973
9974
9975
9976
9977
9978
9979
9980
9981
9982
9983
9984
9985
9986
9987
9988
9989
9990
9991
9992
9993
9994
9995
9996
9997
9998
9999
10000
10001
10002
10003
10004
10005
10006
10007
10008
10009
10010
10011
10012
10013
10014
10015
10016
10017
10018
10019
10020
10021
10022
10023
10024
10025
10026
10027
10028
10029
10030
10031
10032
10033
10034
10035
10036
10037
10038
10039
10040
10041
10042
10043
10044
10045
10046
10047
10048
10049
10050
10051
10052
10053
10054
10055
10056
10057
10058
10059
10060
10061
10062
10063
10064
10065
10066
10067
10068
10069
10070
10071
10072
10073
10074
10075
10076
10077
10078
10079
10080
10081
10082
10083
10084
10085
10086
10087
10088
10089
10090
10091
10092
10093
10094
10095
10096
10097
10098
10099
10100
10101
10102
10103
10104
10105
10106
10107
10108
10109
10110
10111
10112
10113
10114
10115
10116
10117
10118
10119
10120
10121
10122
10123
10124
10125
10126
10127
10128
10129
10130
10131
10132
10133
10134
10135
10136
10137
10138
10139
10140
10141
10142
10143
10144
10145
10146
10147
10148
10149
10150
10151
10152
10153
10154
10155
10156
10157
10158
10159
10160
10161
10162
10163
10164
10165
10166
10167
10168
10169
10170
10171
10172
10173
10174
10175
10176
10177
10178
10179
10180
10181
10182
10183
10184
10185
10186
10187
10188
10189
10190
10191
10192
10193
10194
10195
10196
10197
10198
10199
10200
10201
10202
10203
10204
10205
10206
10207
10208
10209
10210
10211
10212
10213
10214
10215
10216
10217
10218
10219
10220
10221
10222
10223
10224
10225
10226
10227
10228
10229
10230
10231
10232
10233
10234
10235
10236
10237
10238
10239
10240
10241
10242
10243
10244
10245
10246
10247
10248
10249
10250
10251
10252
10253
10254
10255
10256
10257
10258
10259
10260
10261
10262
10263
10264
10265
10266
10267
10268
10269
10270
10271
10272
10273
10274
10275
10276
10277
10278
10279
10280
10281
10282
10283
10284
10285
10286
10287
10288
10289
10290
10291
10292
10293
10294
10295
10296
10297
10298
10299
10300
10301
10302
10303
10304
10305
10306
10307
10308
10309
10310
10311
10312
10313
10314
10315
10316
10317
10318
10319
10320
10321
10322
10323
10324
10325
10326
10327
10328
10329
10330
10331
10332
10333
10334
10335
10336
10337
10338
10339
10340
10341
10342
10343
10344
10345
10346
10347
10348
10349
10350
10351
10352
10353
10354
10355
10356
10357
10358
10359
10360
10361
10362
10363
10364
10365
10366
10367
10368
10369
10370
10371
10372
10373
10374
10375
10376
10377
10378
10379
10380
10381
10382
10383
10384
10385
10386
10387
10388
10389
10390
10391
10392
10393
10394
10395
10396
10397
10398
10399
10400
10401
10402
10403
10404
10405
10406
10407
10408
10409
10410
10411
10412
10413
10414
10415
10416
10417
10418
10419
10420
10421
10422
10423
10424
10425
10426
10427
10428
10429
10430
10431
10432
10433
10434
10435
10436
10437
10438
10439
10440
10441
10442
10443
10444
10445
10446
10447
10448
10449
10450
10451
10452
10453
10454
10455
10456
10457
10458
10459
10460
10461
10462
10463
10464
10465
10466
10467
10468
10469
10470
10471
10472
10473
10474
10475
10476
10477
10478
10479
10480
10481
10482
10483
10484
10485
10486
10487
10488
10489
10490
10491
10492
10493
10494
10495
10496
10497
10498
10499
10500
10501
10502
10503
10504
10505
10506
10507
10508
10509
10510
10511
10512
10513
10514
10515
10516
10517
10518
10519
10520
10521
10522
10523
10524
10525
10526
10527
10528
10529
10530
10531
10532
10533
10534
10535
10536
10537
10538
10539
10540
10541
10542
10543
10544
10545
10546
10547
10548
10549
10550
10551
10552
10553
10554
10555
10556
10557
10558
10559
10560
10561
10562
10563
10564
10565
10566
10567
10568
10569
10570
10571
10572
10573
10574
10575
10576
10577
10578
10579
10580
10581
10582
10583
10584
10585
10586
10587
10588
10589
10590
10591
10592
10593
10594
10595
10596
10597
10598
10599
10600
10601
10602
10603
10604
10605
10606
10607
10608
10609
10610
10611
10612
10613
10614
10615
10616
10617
10618
10619
10620
10621
10622
10623
10624
10625
10626
10627
10628
10629
10630
10631
10632
10633
10634
10635
10636
10637
10638
10639
10640
10641
10642
10643
10644
10645
10646
10647
10648
10649
10650
10651
10652
10653
10654
10655
10656
10657
10658
10659
10660
10661
10662
10663
10664
10665
10666
10667
10668
10669
10670
10671
10672
10673
10674
10675
10676
10677
10678
10679
10680
10681
10682
10683
10684
10685
10686
10687
10688
10689
10690
10691
10692
10693
10694
10695
10696
10697
10698
10699
10700
10701
10702
10703
10704
10705
10706
10707
10708
10709
10710
10711
10712
10713
10714
10715
10716
10717
10718
10719
10720
10721
10722
10723
10724
10725
10726
10727
10728
10729
10730
10731
10732
10733
10734
10735
10736
10737
10738
10739
10740
10741
10742
10743
10744
10745
10746
10747
10748
10749
10750
10751
10752
10753
10754
10755
10756
10757
10758
10759
10760
10761
10762
10763
10764
10765
10766
10767
10768
10769
10770
10771
10772
10773
10774
10775
10776
10777
10778
10779
10780
10781
10782
10783
10784
10785
10786
10787
10788
10789
10790
10791
10792
10793
10794
10795
10796
10797
10798
10799
10800
10801
10802
10803
10804
10805
10806
10807
10808
10809
10810
10811
10812
10813
10814
10815
10816
10817
10818
10819
10820
10821
10822
10823
10824
10825
10826
10827
10828
10829
10830
10831
10832
10833
10834
10835
10836
10837
10838
10839
10840
10841
10842
10843
10844
10845
10846
10847
10848
10849
10850
10851
10852
10853
10854
10855
10856
10857
10858
10859
10860
10861
10862
10863
10864
10865
10866
10867
10868
10869
10870
10871
10872
10873
10874
10875
10876
10877
10878
10879
10880
10881
10882
10883
10884
10885
10886
10887
10888
10889
10890
10891
10892
10893
10894
10895
10896
10897
10898
10899
10900
10901
10902
10903
10904
10905
10906
10907
10908
10909
10910
10911
10912
10913
10914
10915
10916
10917
10918
10919
10920
10921
10922
10923
10924
10925
10926
10927
10928
10929
10930
10931
10932
10933
10934
10935
10936
10937
10938
10939
10940
10941
10942
10943
10944
10945
10946
10947
10948
10949
10950
10951
10952
10953
10954
10955
10956
10957
10958
10959
10960
10961
10962
10963
10964
10965
10966
10967
10968
10969
10970
10971
10972
10973
10974
10975
10976
10977
10978
10979
10980
10981
10982
10983
10984
10985
10986
10987
10988
10989
10990
10991
10992
10993
10994
10995
10996
10997
10998
10999
11000
11001
11002
11003
11004
11005
11006
11007
11008
11009
11010
11011
11012
11013
11014
11015
11016
11017
11018
11019
11020
11021
11022
11023
11024
11025
11026
11027
11028
11029
11030
11031
11032
11033
11034
11035
11036
11037
11038
11039
11040
11041
11042
11043
11044
11045
11046
11047
11048
11049
11050
11051
11052
11053
11054
11055
11056
11057
11058
11059
11060
11061
11062
11063
11064
11065
11066
11067
11068
11069
11070
11071
11072
11073
11074
11075
11076
11077
11078
11079
11080
11081
11082
11083
11084
11085
11086
11087
11088
11089
11090
11091
11092
11093
11094
11095
11096
11097
11098
11099
11100
11101
11102
11103
11104
11105
11106
11107
11108
11109
11110
11111
11112
11113
11114
11115
11116
11117
11118
11119
11120
11121
11122
11123
11124
11125
11126
11127
11128
11129
11130
11131
11132
11133
11134
11135
11136
11137
11138
11139
11140
11141
11142
11143
11144
11145
11146
11147
11148
11149
11150
11151
11152
11153
11154
11155
11156
11157
11158
11159
11160
11161
11162
11163
11164
11165
11166
11167
11168
11169
11170
11171
11172
11173
11174
11175
11176
11177
11178
11179
11180
11181
11182
11183
11184
11185
11186
11187
11188
11189
11190
11191
11192
11193
11194
11195
11196
11197
11198
11199
11200
11201
11202
11203
11204
11205
11206
11207
11208
11209
11210
11211
11212
11213
11214
11215
11216
11217
11218
11219
11220
11221
11222
11223
11224
11225
11226
11227
11228
11229
11230
11231
11232
11233
11234
11235
11236
11237
11238
11239
11240
11241
11242
11243
11244
11245
11246
11247
11248
11249
11250
11251
11252
11253
11254
11255
11256
11257
11258
11259
11260
11261
11262
11263
11264
11265
11266
11267
11268
11269
11270
11271
11272
11273
11274
11275
11276
11277
11278
11279
11280
11281
11282
11283
11284
11285
11286
11287
11288
11289
11290
11291
11292
11293
11294
11295
11296
11297
11298
11299
11300
11301
11302
11303
11304
11305
11306
11307
11308
11309
11310
11311
11312
11313
11314
11315
11316
11317
11318
11319
11320
11321
11322
11323
11324
11325
11326
11327
11328
11329
11330
11331
11332
11333
11334
11335
11336
11337
11338
11339
11340
11341
11342
11343
11344
11345
11346
11347
11348
11349
11350
11351
11352
11353
11354
11355
11356
11357
11358
11359
11360
11361
11362
11363
11364
11365
11366
11367
11368
11369
11370
11371
11372
11373
11374
11375
11376
11377
11378
11379
11380
11381
11382
11383
11384
11385
11386
11387
11388
11389
11390
11391
11392
11393
11394
11395
11396
11397
11398
11399
11400
11401
11402
11403
11404
11405
11406
11407
11408
11409
11410
11411
11412
11413
11414
11415
11416
11417
11418
11419
11420
11421
11422
11423
11424
11425
11426
11427
11428
11429
11430
11431
11432
11433
11434
11435
11436
11437
11438
11439
11440
11441
11442
11443
11444
11445
11446
11447
11448
11449
11450
11451
11452
11453
11454
11455
11456
11457
11458
11459
11460
11461
11462
11463
11464
11465
11466
11467
11468
11469
11470
11471
11472
11473
11474
11475
11476
11477
11478
11479
11480
11481
11482
11483
11484
11485
11486
11487
11488
11489
11490
11491
11492
11493
11494
11495
11496
11497
11498
11499
11500
11501
11502
11503
11504
11505
11506
11507
11508
11509
11510
11511
11512
11513
11514
11515
11516
11517
11518
11519
11520
11521
11522
11523
11524
11525
11526
11527
11528
11529
11530
11531
11532
11533
11534
11535
11536
11537
11538
11539
11540
11541
11542
11543
11544
11545
11546
11547
11548
11549
11550
11551
11552
11553
11554
11555
11556
11557
11558
11559
11560
11561
11562
11563
11564
11565
11566
11567
11568
11569
11570
11571
11572
11573
11574
11575
11576
11577
11578
11579
11580
11581
11582
11583
11584
11585
11586
11587
11588
11589
11590
11591
11592
11593
11594
11595
11596
11597
11598
11599
11600
11601
11602
11603
11604
11605
11606
11607
11608
11609
11610
11611
11612
11613
11614
11615
11616
11617
11618
11619
11620
11621
11622
11623
11624
11625
11626
11627
11628
11629
11630
11631
11632
11633
11634
11635
11636
11637
11638
11639
11640
11641
11642
11643
11644
11645
11646
11647
11648
11649
11650
11651
11652
11653
11654
11655
11656
11657
11658
11659
11660
11661
11662
11663
11664
11665
11666
11667
11668
11669
11670
11671
11672
11673
11674
11675
11676
11677
11678
11679
11680
11681
11682
11683
11684
11685
11686
11687
11688
11689
11690
11691
11692
11693
11694
11695
11696
11697
11698
11699
11700
11701
11702
11703
11704
11705
11706
11707
11708
11709
11710
11711
11712
11713
11714
11715
11716
11717
11718
11719
11720
11721
11722
11723
11724
11725
11726
11727
11728
11729
11730
11731
11732
11733
11734
11735
11736
11737
11738
11739
11740
11741
11742
11743
11744
11745
11746
11747
11748
11749
11750
11751
11752
11753
11754
11755
11756
11757
11758
11759
11760
11761
11762
11763
11764
11765
11766
11767
11768
11769
11770
11771
11772
11773
11774
11775
11776
11777
11778
11779
11780
11781
11782
11783
11784
11785
11786
11787
11788
11789
11790
11791
11792
11793
11794
11795
11796
11797
11798
11799
11800
11801
11802
11803
11804
11805
11806
11807
11808
11809
11810
11811
11812
11813
11814
11815
11816
11817
11818
11819
11820
11821
11822
11823
11824
11825
11826
11827
11828
11829
11830
11831
11832
11833
11834
11835
11836
11837
11838
11839
11840
11841
11842
11843
11844
11845
11846
11847
11848
11849
11850
11851
11852
11853
11854
11855
11856
11857
11858
11859
11860
11861
11862
11863
11864
11865
11866
11867
11868
11869
11870
11871
11872
11873
11874
11875
11876
11877
11878
11879
11880
11881
11882
11883
11884
11885
11886
11887
11888
11889
11890
11891
11892
11893
11894
11895
11896
11897
11898
11899
11900
11901
11902
11903
11904
11905
11906
11907
11908
11909
11910
11911
11912
11913
11914
11915
11916
11917
11918
11919
11920
11921
11922
11923
11924
11925
11926
11927
11928
11929
11930
11931
11932
11933
11934
11935
11936
11937
11938
11939
11940
11941
11942
11943
11944
11945
11946
11947
11948
11949
11950
11951
11952
11953
11954
11955
11956
11957
11958
11959
11960
11961
11962
11963
11964
11965
11966
11967
11968
11969
11970
11971
11972
11973
11974
11975
11976
11977
11978
11979
11980
11981
11982
11983
11984
11985
11986
11987
11988
11989
11990
11991
11992
11993
11994
11995
11996
11997
11998
11999
12000
12001
12002
12003
12004
12005
12006
12007
12008
12009
12010
12011
12012
12013
12014
12015
12016
12017
12018
12019
12020
12021
12022
12023
12024
12025
12026
12027
12028
12029
12030
12031
12032
12033
12034
12035
12036
12037
12038
12039
12040
12041
12042
12043
12044
12045
12046
12047
12048
12049
12050
12051
12052
12053
12054
12055
12056
12057
12058
12059
12060
12061
12062
12063
12064
12065
12066
12067
12068
12069
12070
12071
12072
12073
12074
12075
12076
12077
12078
12079
12080
12081
12082
12083
12084
12085
12086
12087
12088
12089
12090
12091
12092
12093
12094
12095
12096
12097
12098
12099
12100
12101
12102
12103
12104
12105
12106
12107
12108
12109
12110
12111
12112
12113
12114
12115
12116
12117
12118
12119
12120
12121
12122
12123
12124
12125
12126
12127
12128
12129
12130
12131
12132
12133
12134
12135
12136
12137
12138
12139
12140
12141
12142
12143
12144
12145
12146
12147
12148
12149
12150
12151
12152
12153
12154
12155
12156
12157
12158
12159
12160
12161
12162
12163
12164
12165
12166
12167
12168
12169
12170
12171
12172
12173
12174
12175
12176
12177
12178
12179
12180
12181
12182
12183
12184
12185
12186
12187
12188
12189
12190
12191
12192
12193
12194
12195
12196
12197
12198
12199
12200
12201
12202
12203
12204
12205
12206
12207
12208
12209
12210
12211
12212
12213
12214
12215
12216
12217
12218
12219
12220
12221
12222
12223
12224
12225
12226
12227
12228
12229
12230
12231
12232
12233
12234
12235
12236
12237
12238
12239
12240
12241
12242
12243
12244
12245
12246
12247
12248
12249
12250
12251
12252
12253
12254
12255
12256
12257
12258
12259
12260
12261
12262
12263
12264
12265
12266
12267
12268
12269
12270
12271
12272
12273
12274
12275
12276
12277
12278
12279
12280
12281
12282
12283
12284
12285
12286
12287
12288
12289
12290
12291
12292
12293
12294
12295
12296
12297
12298
12299
12300
12301
12302
12303
12304
12305
12306
12307
12308
12309
12310
12311
12312
12313
12314
12315
12316
12317
12318
12319
12320
12321
12322
12323
12324
12325
12326
12327
12328
12329
12330
12331
12332
12333
12334
12335
12336
12337
12338
12339
12340
12341
12342
12343
12344
12345
12346
12347
12348
12349
12350
12351
12352
12353
12354
12355
12356
12357
12358
12359
12360
12361
12362
12363
12364
12365
12366
12367
12368
12369
12370
12371
12372
12373
12374
12375
12376
12377
12378
12379
12380
12381
12382
12383
12384
12385
12386
12387
12388
12389
12390
12391
12392
12393
12394
12395
12396
12397
12398
12399
12400
12401
12402
12403
12404
12405
12406
12407
12408
12409
12410
12411
12412
12413
12414
12415
12416
12417
12418
12419
12420
12421
12422
12423
12424
12425
12426
12427
12428
12429
12430
12431
12432
12433
12434
12435
12436
12437
12438
12439
12440
12441
12442
12443
12444
12445
12446
12447
12448
12449
12450
12451
12452
12453
12454
12455
12456
12457
12458
12459
12460
12461
12462
12463
12464
12465
12466
12467
12468
12469
12470
12471
12472
12473
12474
12475
12476
12477
12478
12479
12480
12481
12482
12483
12484
12485
12486
12487
12488
12489
12490
12491
12492
12493
12494
12495
12496
12497
12498
12499
12500
12501
12502
12503
12504
12505
12506
12507
12508
12509
12510
12511
12512
12513
12514
12515
12516
12517
12518
12519
12520
12521
12522
12523
12524
12525
12526
12527
12528
12529
12530
12531
12532
12533
12534
12535
12536
12537
12538
12539
12540
12541
12542
12543
12544
12545
12546
12547
12548
12549
12550
12551
12552
12553
12554
12555
12556
12557
12558
12559
12560
12561
12562
12563
12564
12565
12566
12567
12568
12569
12570
12571
12572
12573
12574
12575
12576
12577
12578
12579
12580
12581
12582
12583
12584
12585
12586
12587
12588
12589
12590
12591
12592
12593
12594
12595
12596
12597
12598
12599
12600
12601
12602
12603
12604
12605
12606
12607
12608
12609
12610
12611
12612
12613
12614
12615
12616
12617
12618
12619
12620
12621
12622
12623
12624
12625
12626
12627
12628
12629
12630
12631
12632
12633
12634
12635
12636
12637
12638
12639
12640
12641
12642
12643
12644
12645
12646
12647
12648
12649
12650
12651
12652
12653
12654
12655
12656
12657
12658
12659
12660
12661
12662
12663
12664
12665
12666
12667
12668
12669
12670
12671
12672
12673
12674
12675
12676
12677
12678
12679
12680
12681
12682
12683
12684
12685
12686
12687
12688
12689
12690
12691
12692
12693
12694
12695
12696
12697
12698
12699
12700
12701
12702
12703
12704
12705
12706
12707
12708
12709
12710
12711
12712
12713
12714
12715
12716
12717
12718
12719
12720
12721
12722
12723
12724
12725
12726
12727
12728
12729
12730
12731
12732
12733
12734
12735
12736
12737
12738
12739
12740
12741
12742
12743
12744
12745
12746
12747
12748
12749
12750
12751
12752
12753
12754
12755
12756
12757
12758
12759
12760
12761
12762
12763
12764
12765
12766
12767
12768
12769
12770
12771
12772
12773
12774
12775
12776
12777
12778
12779
12780
12781
12782
12783
12784
12785
12786
12787
12788
12789
12790
12791
12792
12793
12794
12795
12796
12797
12798
12799
12800
12801
12802
12803
12804
12805
12806
12807
12808
12809
12810
12811
12812
12813
12814
12815
12816
12817
12818
12819
12820
12821
12822
12823
12824
12825
12826
12827
12828
12829
12830
12831
12832
12833
12834
12835
12836
12837
12838
12839
12840
12841
12842
12843
12844
12845
12846
12847
12848
12849
12850
12851
12852
12853
12854
12855
12856
12857
12858
12859
12860
12861
12862
12863
12864
12865
12866
12867
12868
12869
12870
12871
12872
12873
12874
12875
12876
12877
12878
12879
12880
12881
12882
12883
12884
12885
12886
12887
12888
12889
12890
12891
12892
12893
12894
12895
12896
12897
12898
12899
12900
12901
12902
12903
12904
12905
12906
12907
12908
12909
12910
12911
12912
12913
12914
12915
12916
12917
12918
12919
12920
12921
12922
12923
12924
12925
12926
12927
12928
12929
12930
12931
12932
12933
12934
12935
12936
12937
12938
12939
12940
12941
12942
12943
12944
12945
12946
12947
12948
12949
12950
12951
12952
12953
12954
12955
12956
12957
12958
12959
12960
12961
12962
12963
12964
12965
12966
12967
12968
12969
12970
12971
12972
12973
12974
12975
12976
12977
12978
12979
12980
12981
12982
12983
12984
12985
12986
12987
12988
12989
12990
12991
12992
12993
12994
12995
12996
12997
12998
12999
13000
13001
13002
13003
13004
13005
13006
13007
13008
13009
13010
13011
13012
13013
13014
13015
13016
13017
13018
13019
13020
13021
13022
13023
13024
13025
13026
13027
13028
13029
13030
13031
13032
13033
13034
13035
13036
13037
13038
13039
13040
13041
13042
13043
13044
13045
13046
13047
13048
13049
13050
13051
13052
13053
13054
13055
13056
13057
13058
13059
13060
13061
13062
13063
13064
13065
13066
13067
13068
13069
13070
13071
13072
13073
13074
13075
13076
13077
13078
13079
13080
13081
13082
13083
13084
13085
13086
13087
|
//===- SelectionDAG.cpp - Implement the SelectionDAG data structures ------===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
//
// This implements the SelectionDAG class.
//
//===----------------------------------------------------------------------===//
#include "llvm/CodeGen/SelectionDAG.h"
#include "SDNodeDbgValue.h"
#include "llvm/ADT/APFloat.h"
#include "llvm/ADT/APInt.h"
#include "llvm/ADT/APSInt.h"
#include "llvm/ADT/ArrayRef.h"
#include "llvm/ADT/BitVector.h"
#include "llvm/ADT/DenseSet.h"
#include "llvm/ADT/FoldingSet.h"
#include "llvm/ADT/STLExtras.h"
#include "llvm/ADT/SmallPtrSet.h"
#include "llvm/ADT/SmallVector.h"
#include "llvm/ADT/Twine.h"
#include "llvm/Analysis/AliasAnalysis.h"
#include "llvm/Analysis/MemoryLocation.h"
#include "llvm/Analysis/ValueTracking.h"
#include "llvm/Analysis/VectorUtils.h"
#include "llvm/BinaryFormat/Dwarf.h"
#include "llvm/CodeGen/Analysis.h"
#include "llvm/CodeGen/FunctionLoweringInfo.h"
#include "llvm/CodeGen/ISDOpcodes.h"
#include "llvm/CodeGen/MachineBasicBlock.h"
#include "llvm/CodeGen/MachineConstantPool.h"
#include "llvm/CodeGen/MachineFrameInfo.h"
#include "llvm/CodeGen/MachineFunction.h"
#include "llvm/CodeGen/MachineMemOperand.h"
#include "llvm/CodeGen/RuntimeLibcalls.h"
#include "llvm/CodeGen/SelectionDAGAddressAnalysis.h"
#include "llvm/CodeGen/SelectionDAGNodes.h"
#include "llvm/CodeGen/SelectionDAGTargetInfo.h"
#include "llvm/CodeGen/TargetFrameLowering.h"
#include "llvm/CodeGen/TargetLowering.h"
#include "llvm/CodeGen/TargetRegisterInfo.h"
#include "llvm/CodeGen/TargetSubtargetInfo.h"
#include "llvm/CodeGen/ValueTypes.h"
#include "llvm/CodeGenTypes/MachineValueType.h"
#include "llvm/IR/Constant.h"
#include "llvm/IR/ConstantRange.h"
#include "llvm/IR/Constants.h"
#include "llvm/IR/DataLayout.h"
#include "llvm/IR/DebugInfoMetadata.h"
#include "llvm/IR/DebugLoc.h"
#include "llvm/IR/DerivedTypes.h"
#include "llvm/IR/Function.h"
#include "llvm/IR/GlobalValue.h"
#include "llvm/IR/Metadata.h"
#include "llvm/IR/Type.h"
#include "llvm/Support/Casting.h"
#include "llvm/Support/CodeGen.h"
#include "llvm/Support/Compiler.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/ErrorHandling.h"
#include "llvm/Support/KnownBits.h"
#include "llvm/Support/MathExtras.h"
#include "llvm/Support/Mutex.h"
#include "llvm/Support/raw_ostream.h"
#include "llvm/Target/TargetMachine.h"
#include "llvm/Target/TargetOptions.h"
#include "llvm/TargetParser/Triple.h"
#include "llvm/Transforms/Utils/SizeOpts.h"
#include <algorithm>
#include <cassert>
#include <cstdint>
#include <cstdlib>
#include <limits>
#include <set>
#include <string>
#include <utility>
#include <vector>
using namespace llvm;
/// makeVTList - Return an instance of the SDVTList struct initialized with the
/// specified members.
static SDVTList makeVTList(const EVT *VTs, unsigned NumVTs) {
SDVTList Res = {VTs, NumVTs};
return Res;
}
// Default null implementations of the callbacks.
void SelectionDAG::DAGUpdateListener::NodeDeleted(SDNode*, SDNode*) {}
void SelectionDAG::DAGUpdateListener::NodeUpdated(SDNode*) {}
void SelectionDAG::DAGUpdateListener::NodeInserted(SDNode *) {}
void SelectionDAG::DAGNodeDeletedListener::anchor() {}
void SelectionDAG::DAGNodeInsertedListener::anchor() {}
#define DEBUG_TYPE "selectiondag"
static cl::opt<bool> EnableMemCpyDAGOpt("enable-memcpy-dag-opt",
cl::Hidden, cl::init(true),
cl::desc("Gang up loads and stores generated by inlining of memcpy"));
static cl::opt<int> MaxLdStGlue("ldstmemcpy-glue-max",
cl::desc("Number limit for gluing ld/st of memcpy."),
cl::Hidden, cl::init(0));
static void NewSDValueDbgMsg(SDValue V, StringRef Msg, SelectionDAG *G) {
LLVM_DEBUG(dbgs() << Msg; V.getNode()->dump(G););
}
//===----------------------------------------------------------------------===//
// ConstantFPSDNode Class
//===----------------------------------------------------------------------===//
/// isExactlyValue - We don't rely on operator== working on double values, as
/// it returns true for things that are clearly not equal, like -0.0 and 0.0.
/// As such, this method can be used to do an exact bit-for-bit comparison of
/// two floating point values.
bool ConstantFPSDNode::isExactlyValue(const APFloat& V) const {
return getValueAPF().bitwiseIsEqual(V);
}
bool ConstantFPSDNode::isValueValidForType(EVT VT,
const APFloat& Val) {
assert(VT.isFloatingPoint() && "Can only convert between FP types");
// convert modifies in place, so make a copy.
APFloat Val2 = APFloat(Val);
bool losesInfo;
(void) Val2.convert(SelectionDAG::EVTToAPFloatSemantics(VT),
APFloat::rmNearestTiesToEven,
&losesInfo);
return !losesInfo;
}
//===----------------------------------------------------------------------===//
// ISD Namespace
//===----------------------------------------------------------------------===//
bool ISD::isConstantSplatVector(const SDNode *N, APInt &SplatVal) {
if (N->getOpcode() == ISD::SPLAT_VECTOR) {
unsigned EltSize =
N->getValueType(0).getVectorElementType().getSizeInBits();
if (auto *Op0 = dyn_cast<ConstantSDNode>(N->getOperand(0))) {
SplatVal = Op0->getAPIntValue().trunc(EltSize);
return true;
}
if (auto *Op0 = dyn_cast<ConstantFPSDNode>(N->getOperand(0))) {
SplatVal = Op0->getValueAPF().bitcastToAPInt().trunc(EltSize);
return true;
}
}
auto *BV = dyn_cast<BuildVectorSDNode>(N);
if (!BV)
return false;
APInt SplatUndef;
unsigned SplatBitSize;
bool HasUndefs;
unsigned EltSize = N->getValueType(0).getVectorElementType().getSizeInBits();
// Endianness does not matter here. We are checking for a splat given the
// element size of the vector, and if we find such a splat for little endian
// layout, then that should be valid also for big endian (as the full vector
// size is known to be a multiple of the element size).
const bool IsBigEndian = false;
return BV->isConstantSplat(SplatVal, SplatUndef, SplatBitSize, HasUndefs,
EltSize, IsBigEndian) &&
EltSize == SplatBitSize;
}
// FIXME: AllOnes and AllZeros duplicate a lot of code. Could these be
// specializations of the more general isConstantSplatVector()?
bool ISD::isConstantSplatVectorAllOnes(const SDNode *N, bool BuildVectorOnly) {
// Look through a bit convert.
while (N->getOpcode() == ISD::BITCAST)
N = N->getOperand(0).getNode();
if (!BuildVectorOnly && N->getOpcode() == ISD::SPLAT_VECTOR) {
APInt SplatVal;
return isConstantSplatVector(N, SplatVal) && SplatVal.isAllOnes();
}
if (N->getOpcode() != ISD::BUILD_VECTOR) return false;
unsigned i = 0, e = N->getNumOperands();
// Skip over all of the undef values.
while (i != e && N->getOperand(i).isUndef())
++i;
// Do not accept an all-undef vector.
if (i == e) return false;
// Do not accept build_vectors that aren't all constants or which have non-~0
// elements. We have to be a bit careful here, as the type of the constant
// may not be the same as the type of the vector elements due to type
// legalization (the elements are promoted to a legal type for the target and
// a vector of a type may be legal when the base element type is not).
// We only want to check enough bits to cover the vector elements, because
// we care if the resultant vector is all ones, not whether the individual
// constants are.
SDValue NotZero = N->getOperand(i);
unsigned EltSize = N->getValueType(0).getScalarSizeInBits();
if (ConstantSDNode *CN = dyn_cast<ConstantSDNode>(NotZero)) {
if (CN->getAPIntValue().countr_one() < EltSize)
return false;
} else if (ConstantFPSDNode *CFPN = dyn_cast<ConstantFPSDNode>(NotZero)) {
if (CFPN->getValueAPF().bitcastToAPInt().countr_one() < EltSize)
return false;
} else
return false;
// Okay, we have at least one ~0 value, check to see if the rest match or are
// undefs. Even with the above element type twiddling, this should be OK, as
// the same type legalization should have applied to all the elements.
for (++i; i != e; ++i)
if (N->getOperand(i) != NotZero && !N->getOperand(i).isUndef())
return false;
return true;
}
bool ISD::isConstantSplatVectorAllZeros(const SDNode *N, bool BuildVectorOnly) {
// Look through a bit convert.
while (N->getOpcode() == ISD::BITCAST)
N = N->getOperand(0).getNode();
if (!BuildVectorOnly && N->getOpcode() == ISD::SPLAT_VECTOR) {
APInt SplatVal;
return isConstantSplatVector(N, SplatVal) && SplatVal.isZero();
}
if (N->getOpcode() != ISD::BUILD_VECTOR) return false;
bool IsAllUndef = true;
for (const SDValue &Op : N->op_values()) {
if (Op.isUndef())
continue;
IsAllUndef = false;
// Do not accept build_vectors that aren't all constants or which have non-0
// elements. We have to be a bit careful here, as the type of the constant
// may not be the same as the type of the vector elements due to type
// legalization (the elements are promoted to a legal type for the target
// and a vector of a type may be legal when the base element type is not).
// We only want to check enough bits to cover the vector elements, because
// we care if the resultant vector is all zeros, not whether the individual
// constants are.
unsigned EltSize = N->getValueType(0).getScalarSizeInBits();
if (ConstantSDNode *CN = dyn_cast<ConstantSDNode>(Op)) {
if (CN->getAPIntValue().countr_zero() < EltSize)
return false;
} else if (ConstantFPSDNode *CFPN = dyn_cast<ConstantFPSDNode>(Op)) {
if (CFPN->getValueAPF().bitcastToAPInt().countr_zero() < EltSize)
return false;
} else
return false;
}
// Do not accept an all-undef vector.
if (IsAllUndef)
return false;
return true;
}
bool ISD::isBuildVectorAllOnes(const SDNode *N) {
return isConstantSplatVectorAllOnes(N, /*BuildVectorOnly*/ true);
}
bool ISD::isBuildVectorAllZeros(const SDNode *N) {
return isConstantSplatVectorAllZeros(N, /*BuildVectorOnly*/ true);
}
bool ISD::isBuildVectorOfConstantSDNodes(const SDNode *N) {
if (N->getOpcode() != ISD::BUILD_VECTOR)
return false;
for (const SDValue &Op : N->op_values()) {
if (Op.isUndef())
continue;
if (!isa<ConstantSDNode>(Op))
return false;
}
return true;
}
bool ISD::isBuildVectorOfConstantFPSDNodes(const SDNode *N) {
if (N->getOpcode() != ISD::BUILD_VECTOR)
return false;
for (const SDValue &Op : N->op_values()) {
if (Op.isUndef())
continue;
if (!isa<ConstantFPSDNode>(Op))
return false;
}
return true;
}
bool ISD::isVectorShrinkable(const SDNode *N, unsigned NewEltSize,
bool Signed) {
assert(N->getValueType(0).isVector() && "Expected a vector!");
unsigned EltSize = N->getValueType(0).getScalarSizeInBits();
if (EltSize <= NewEltSize)
return false;
if (N->getOpcode() == ISD::ZERO_EXTEND) {
return (N->getOperand(0).getValueType().getScalarSizeInBits() <=
NewEltSize) &&
!Signed;
}
if (N->getOpcode() == ISD::SIGN_EXTEND) {
return (N->getOperand(0).getValueType().getScalarSizeInBits() <=
NewEltSize) &&
Signed;
}
if (N->getOpcode() != ISD::BUILD_VECTOR)
return false;
for (const SDValue &Op : N->op_values()) {
if (Op.isUndef())
continue;
if (!isa<ConstantSDNode>(Op))
return false;
APInt C = Op->getAsAPIntVal().trunc(EltSize);
if (Signed && C.trunc(NewEltSize).sext(EltSize) != C)
return false;
if (!Signed && C.trunc(NewEltSize).zext(EltSize) != C)
return false;
}
return true;
}
bool ISD::allOperandsUndef(const SDNode *N) {
// Return false if the node has no operands.
// This is "logically inconsistent" with the definition of "all" but
// is probably the desired behavior.
if (N->getNumOperands() == 0)
return false;
return all_of(N->op_values(), [](SDValue Op) { return Op.isUndef(); });
}
bool ISD::isFreezeUndef(const SDNode *N) {
return N->getOpcode() == ISD::FREEZE && N->getOperand(0).isUndef();
}
template <typename ConstNodeType>
bool ISD::matchUnaryPredicateImpl(SDValue Op,
std::function<bool(ConstNodeType *)> Match,
bool AllowUndefs) {
// FIXME: Add support for scalar UNDEF cases?
if (auto *C = dyn_cast<ConstNodeType>(Op))
return Match(C);
// FIXME: Add support for vector UNDEF cases?
if (ISD::BUILD_VECTOR != Op.getOpcode() &&
ISD::SPLAT_VECTOR != Op.getOpcode())
return false;
EVT SVT = Op.getValueType().getScalarType();
for (unsigned i = 0, e = Op.getNumOperands(); i != e; ++i) {
if (AllowUndefs && Op.getOperand(i).isUndef()) {
if (!Match(nullptr))
return false;
continue;
}
auto *Cst = dyn_cast<ConstNodeType>(Op.getOperand(i));
if (!Cst || Cst->getValueType(0) != SVT || !Match(Cst))
return false;
}
return true;
}
// Build used template types.
template bool ISD::matchUnaryPredicateImpl<ConstantSDNode>(
SDValue, std::function<bool(ConstantSDNode *)>, bool);
template bool ISD::matchUnaryPredicateImpl<ConstantFPSDNode>(
SDValue, std::function<bool(ConstantFPSDNode *)>, bool);
bool ISD::matchBinaryPredicate(
SDValue LHS, SDValue RHS,
std::function<bool(ConstantSDNode *, ConstantSDNode *)> Match,
bool AllowUndefs, bool AllowTypeMismatch) {
if (!AllowTypeMismatch && LHS.getValueType() != RHS.getValueType())
return false;
// TODO: Add support for scalar UNDEF cases?
if (auto *LHSCst = dyn_cast<ConstantSDNode>(LHS))
if (auto *RHSCst = dyn_cast<ConstantSDNode>(RHS))
return Match(LHSCst, RHSCst);
// TODO: Add support for vector UNDEF cases?
if (LHS.getOpcode() != RHS.getOpcode() ||
(LHS.getOpcode() != ISD::BUILD_VECTOR &&
LHS.getOpcode() != ISD::SPLAT_VECTOR))
return false;
EVT SVT = LHS.getValueType().getScalarType();
for (unsigned i = 0, e = LHS.getNumOperands(); i != e; ++i) {
SDValue LHSOp = LHS.getOperand(i);
SDValue RHSOp = RHS.getOperand(i);
bool LHSUndef = AllowUndefs && LHSOp.isUndef();
bool RHSUndef = AllowUndefs && RHSOp.isUndef();
auto *LHSCst = dyn_cast<ConstantSDNode>(LHSOp);
auto *RHSCst = dyn_cast<ConstantSDNode>(RHSOp);
if ((!LHSCst && !LHSUndef) || (!RHSCst && !RHSUndef))
return false;
if (!AllowTypeMismatch && (LHSOp.getValueType() != SVT ||
LHSOp.getValueType() != RHSOp.getValueType()))
return false;
if (!Match(LHSCst, RHSCst))
return false;
}
return true;
}
ISD::NodeType ISD::getVecReduceBaseOpcode(unsigned VecReduceOpcode) {
switch (VecReduceOpcode) {
default:
llvm_unreachable("Expected VECREDUCE opcode");
case ISD::VECREDUCE_FADD:
case ISD::VECREDUCE_SEQ_FADD:
case ISD::VP_REDUCE_FADD:
case ISD::VP_REDUCE_SEQ_FADD:
return ISD::FADD;
case ISD::VECREDUCE_FMUL:
case ISD::VECREDUCE_SEQ_FMUL:
case ISD::VP_REDUCE_FMUL:
case ISD::VP_REDUCE_SEQ_FMUL:
return ISD::FMUL;
case ISD::VECREDUCE_ADD:
case ISD::VP_REDUCE_ADD:
return ISD::ADD;
case ISD::VECREDUCE_MUL:
case ISD::VP_REDUCE_MUL:
return ISD::MUL;
case ISD::VECREDUCE_AND:
case ISD::VP_REDUCE_AND:
return ISD::AND;
case ISD::VECREDUCE_OR:
case ISD::VP_REDUCE_OR:
return ISD::OR;
case ISD::VECREDUCE_XOR:
case ISD::VP_REDUCE_XOR:
return ISD::XOR;
case ISD::VECREDUCE_SMAX:
case ISD::VP_REDUCE_SMAX:
return ISD::SMAX;
case ISD::VECREDUCE_SMIN:
case ISD::VP_REDUCE_SMIN:
return ISD::SMIN;
case ISD::VECREDUCE_UMAX:
case ISD::VP_REDUCE_UMAX:
return ISD::UMAX;
case ISD::VECREDUCE_UMIN:
case ISD::VP_REDUCE_UMIN:
return ISD::UMIN;
case ISD::VECREDUCE_FMAX:
case ISD::VP_REDUCE_FMAX:
return ISD::FMAXNUM;
case ISD::VECREDUCE_FMIN:
case ISD::VP_REDUCE_FMIN:
return ISD::FMINNUM;
case ISD::VECREDUCE_FMAXIMUM:
return ISD::FMAXIMUM;
case ISD::VECREDUCE_FMINIMUM:
return ISD::FMINIMUM;
}
}
bool ISD::isVPOpcode(unsigned Opcode) {
switch (Opcode) {
default:
return false;
#define BEGIN_REGISTER_VP_SDNODE(VPSD, ...) \
case ISD::VPSD: \
return true;
#include "llvm/IR/VPIntrinsics.def"
}
}
bool ISD::isVPBinaryOp(unsigned Opcode) {
switch (Opcode) {
default:
break;
#define BEGIN_REGISTER_VP_SDNODE(VPSD, ...) case ISD::VPSD:
#define VP_PROPERTY_BINARYOP return true;
#define END_REGISTER_VP_SDNODE(VPSD) break;
#include "llvm/IR/VPIntrinsics.def"
}
return false;
}
bool ISD::isVPReduction(unsigned Opcode) {
switch (Opcode) {
default:
break;
#define BEGIN_REGISTER_VP_SDNODE(VPSD, ...) case ISD::VPSD:
#define VP_PROPERTY_REDUCTION(STARTPOS, ...) return true;
#define END_REGISTER_VP_SDNODE(VPSD) break;
#include "llvm/IR/VPIntrinsics.def"
}
return false;
}
/// The operand position of the vector mask.
std::optional<unsigned> ISD::getVPMaskIdx(unsigned Opcode) {
switch (Opcode) {
default:
return std::nullopt;
#define BEGIN_REGISTER_VP_SDNODE(VPSD, LEGALPOS, TDNAME, MASKPOS, ...) \
case ISD::VPSD: \
return MASKPOS;
#include "llvm/IR/VPIntrinsics.def"
}
}
/// The operand position of the explicit vector length parameter.
std::optional<unsigned> ISD::getVPExplicitVectorLengthIdx(unsigned Opcode) {
switch (Opcode) {
default:
return std::nullopt;
#define BEGIN_REGISTER_VP_SDNODE(VPSD, LEGALPOS, TDNAME, MASKPOS, EVLPOS) \
case ISD::VPSD: \
return EVLPOS;
#include "llvm/IR/VPIntrinsics.def"
}
}
std::optional<unsigned> ISD::getBaseOpcodeForVP(unsigned VPOpcode,
bool hasFPExcept) {
// FIXME: Return strict opcodes in case of fp exceptions.
switch (VPOpcode) {
default:
return std::nullopt;
#define BEGIN_REGISTER_VP_SDNODE(VPOPC, ...) case ISD::VPOPC:
#define VP_PROPERTY_FUNCTIONAL_SDOPC(SDOPC) return ISD::SDOPC;
#define END_REGISTER_VP_SDNODE(VPOPC) break;
#include "llvm/IR/VPIntrinsics.def"
}
return std::nullopt;
}
unsigned ISD::getVPForBaseOpcode(unsigned Opcode) {
switch (Opcode) {
default:
llvm_unreachable("can not translate this Opcode to VP.");
#define BEGIN_REGISTER_VP_SDNODE(VPOPC, ...) break;
#define VP_PROPERTY_FUNCTIONAL_SDOPC(SDOPC) case ISD::SDOPC:
#define END_REGISTER_VP_SDNODE(VPOPC) return ISD::VPOPC;
#include "llvm/IR/VPIntrinsics.def"
}
}
ISD::NodeType ISD::getExtForLoadExtType(bool IsFP, ISD::LoadExtType ExtType) {
switch (ExtType) {
case ISD::EXTLOAD:
return IsFP ? ISD::FP_EXTEND : ISD::ANY_EXTEND;
case ISD::SEXTLOAD:
return ISD::SIGN_EXTEND;
case ISD::ZEXTLOAD:
return ISD::ZERO_EXTEND;
default:
break;
}
llvm_unreachable("Invalid LoadExtType");
}
ISD::CondCode ISD::getSetCCSwappedOperands(ISD::CondCode Operation) {
// To perform this operation, we just need to swap the L and G bits of the
// operation.
unsigned OldL = (Operation >> 2) & 1;
unsigned OldG = (Operation >> 1) & 1;
return ISD::CondCode((Operation & ~6) | // Keep the N, U, E bits
(OldL << 1) | // New G bit
(OldG << 2)); // New L bit.
}
static ISD::CondCode getSetCCInverseImpl(ISD::CondCode Op, bool isIntegerLike) {
unsigned Operation = Op;
if (isIntegerLike)
Operation ^= 7; // Flip L, G, E bits, but not U.
else
Operation ^= 15; // Flip all of the condition bits.
if (Operation > ISD::SETTRUE2)
Operation &= ~8; // Don't let N and U bits get set.
return ISD::CondCode(Operation);
}
ISD::CondCode ISD::getSetCCInverse(ISD::CondCode Op, EVT Type) {
return getSetCCInverseImpl(Op, Type.isInteger());
}
ISD::CondCode ISD::GlobalISel::getSetCCInverse(ISD::CondCode Op,
bool isIntegerLike) {
return getSetCCInverseImpl(Op, isIntegerLike);
}
/// For an integer comparison, return 1 if the comparison is a signed operation
/// and 2 if the result is an unsigned comparison. Return zero if the operation
/// does not depend on the sign of the input (setne and seteq).
static int isSignedOp(ISD::CondCode Opcode) {
switch (Opcode) {
default: llvm_unreachable("Illegal integer setcc operation!");
case ISD::SETEQ:
case ISD::SETNE: return 0;
case ISD::SETLT:
case ISD::SETLE:
case ISD::SETGT:
case ISD::SETGE: return 1;
case ISD::SETULT:
case ISD::SETULE:
case ISD::SETUGT:
case ISD::SETUGE: return 2;
}
}
ISD::CondCode ISD::getSetCCOrOperation(ISD::CondCode Op1, ISD::CondCode Op2,
EVT Type) {
bool IsInteger = Type.isInteger();
if (IsInteger && (isSignedOp(Op1) | isSignedOp(Op2)) == 3)
// Cannot fold a signed integer setcc with an unsigned integer setcc.
return ISD::SETCC_INVALID;
unsigned Op = Op1 | Op2; // Combine all of the condition bits.
// If the N and U bits get set, then the resultant comparison DOES suddenly
// care about orderedness, and it is true when ordered.
if (Op > ISD::SETTRUE2)
Op &= ~16; // Clear the U bit if the N bit is set.
// Canonicalize illegal integer setcc's.
if (IsInteger && Op == ISD::SETUNE) // e.g. SETUGT | SETULT
Op = ISD::SETNE;
return ISD::CondCode(Op);
}
ISD::CondCode ISD::getSetCCAndOperation(ISD::CondCode Op1, ISD::CondCode Op2,
EVT Type) {
bool IsInteger = Type.isInteger();
if (IsInteger && (isSignedOp(Op1) | isSignedOp(Op2)) == 3)
// Cannot fold a signed setcc with an unsigned setcc.
return ISD::SETCC_INVALID;
// Combine all of the condition bits.
ISD::CondCode Result = ISD::CondCode(Op1 & Op2);
// Canonicalize illegal integer setcc's.
if (IsInteger) {
switch (Result) {
default: break;
case ISD::SETUO : Result = ISD::SETFALSE; break; // SETUGT & SETULT
case ISD::SETOEQ: // SETEQ & SETU[LG]E
case ISD::SETUEQ: Result = ISD::SETEQ ; break; // SETUGE & SETULE
case ISD::SETOLT: Result = ISD::SETULT ; break; // SETULT & SETNE
case ISD::SETOGT: Result = ISD::SETUGT ; break; // SETUGT & SETNE
}
}
return Result;
}
//===----------------------------------------------------------------------===//
// SDNode Profile Support
//===----------------------------------------------------------------------===//
/// AddNodeIDOpcode - Add the node opcode to the NodeID data.
static void AddNodeIDOpcode(FoldingSetNodeID &ID, unsigned OpC) {
ID.AddInteger(OpC);
}
/// AddNodeIDValueTypes - Value type lists are intern'd so we can represent them
/// solely with their pointer.
static void AddNodeIDValueTypes(FoldingSetNodeID &ID, SDVTList VTList) {
ID.AddPointer(VTList.VTs);
}
/// AddNodeIDOperands - Various routines for adding operands to the NodeID data.
static void AddNodeIDOperands(FoldingSetNodeID &ID,
ArrayRef<SDValue> Ops) {
for (const auto &Op : Ops) {
ID.AddPointer(Op.getNode());
ID.AddInteger(Op.getResNo());
}
}
/// AddNodeIDOperands - Various routines for adding operands to the NodeID data.
static void AddNodeIDOperands(FoldingSetNodeID &ID,
ArrayRef<SDUse> Ops) {
for (const auto &Op : Ops) {
ID.AddPointer(Op.getNode());
ID.AddInteger(Op.getResNo());
}
}
static void AddNodeIDNode(FoldingSetNodeID &ID, unsigned OpC,
SDVTList VTList, ArrayRef<SDValue> OpList) {
AddNodeIDOpcode(ID, OpC);
AddNodeIDValueTypes(ID, VTList);
AddNodeIDOperands(ID, OpList);
}
/// If this is an SDNode with special info, add this info to the NodeID data.
static void AddNodeIDCustom(FoldingSetNodeID &ID, const SDNode *N) {
switch (N->getOpcode()) {
case ISD::TargetExternalSymbol:
case ISD::ExternalSymbol:
case ISD::MCSymbol:
llvm_unreachable("Should only be used on nodes with operands");
default: break; // Normal nodes don't need extra info.
case ISD::TargetConstant:
case ISD::Constant: {
const ConstantSDNode *C = cast<ConstantSDNode>(N);
ID.AddPointer(C->getConstantIntValue());
ID.AddBoolean(C->isOpaque());
break;
}
case ISD::TargetConstantFP:
case ISD::ConstantFP:
ID.AddPointer(cast<ConstantFPSDNode>(N)->getConstantFPValue());
break;
case ISD::TargetGlobalAddress:
case ISD::GlobalAddress:
case ISD::TargetGlobalTLSAddress:
case ISD::GlobalTLSAddress: {
const GlobalAddressSDNode *GA = cast<GlobalAddressSDNode>(N);
ID.AddPointer(GA->getGlobal());
ID.AddInteger(GA->getOffset());
ID.AddInteger(GA->getTargetFlags());
break;
}
case ISD::BasicBlock:
ID.AddPointer(cast<BasicBlockSDNode>(N)->getBasicBlock());
break;
case ISD::Register:
ID.AddInteger(cast<RegisterSDNode>(N)->getReg());
break;
case ISD::RegisterMask:
ID.AddPointer(cast<RegisterMaskSDNode>(N)->getRegMask());
break;
case ISD::SRCVALUE:
ID.AddPointer(cast<SrcValueSDNode>(N)->getValue());
break;
case ISD::FrameIndex:
case ISD::TargetFrameIndex:
ID.AddInteger(cast<FrameIndexSDNode>(N)->getIndex());
break;
case ISD::LIFETIME_START:
case ISD::LIFETIME_END:
if (cast<LifetimeSDNode>(N)->hasOffset()) {
ID.AddInteger(cast<LifetimeSDNode>(N)->getSize());
ID.AddInteger(cast<LifetimeSDNode>(N)->getOffset());
}
break;
case ISD::PSEUDO_PROBE:
ID.AddInteger(cast<PseudoProbeSDNode>(N)->getGuid());
ID.AddInteger(cast<PseudoProbeSDNode>(N)->getIndex());
ID.AddInteger(cast<PseudoProbeSDNode>(N)->getAttributes());
break;
case ISD::JumpTable:
case ISD::TargetJumpTable:
ID.AddInteger(cast<JumpTableSDNode>(N)->getIndex());
ID.AddInteger(cast<JumpTableSDNode>(N)->getTargetFlags());
break;
case ISD::ConstantPool:
case ISD::TargetConstantPool: {
const ConstantPoolSDNode *CP = cast<ConstantPoolSDNode>(N);
ID.AddInteger(CP->getAlign().value());
ID.AddInteger(CP->getOffset());
if (CP->isMachineConstantPoolEntry())
CP->getMachineCPVal()->addSelectionDAGCSEId(ID);
else
ID.AddPointer(CP->getConstVal());
ID.AddInteger(CP->getTargetFlags());
break;
}
case ISD::TargetIndex: {
const TargetIndexSDNode *TI = cast<TargetIndexSDNode>(N);
ID.AddInteger(TI->getIndex());
ID.AddInteger(TI->getOffset());
ID.AddInteger(TI->getTargetFlags());
break;
}
case ISD::LOAD: {
const LoadSDNode *LD = cast<LoadSDNode>(N);
ID.AddInteger(LD->getMemoryVT().getRawBits());
ID.AddInteger(LD->getRawSubclassData());
ID.AddInteger(LD->getPointerInfo().getAddrSpace());
ID.AddInteger(LD->getMemOperand()->getFlags());
break;
}
case ISD::STORE: {
const StoreSDNode *ST = cast<StoreSDNode>(N);
ID.AddInteger(ST->getMemoryVT().getRawBits());
ID.AddInteger(ST->getRawSubclassData());
ID.AddInteger(ST->getPointerInfo().getAddrSpace());
ID.AddInteger(ST->getMemOperand()->getFlags());
break;
}
case ISD::VP_LOAD: {
const VPLoadSDNode *ELD = cast<VPLoadSDNode>(N);
ID.AddInteger(ELD->getMemoryVT().getRawBits());
ID.AddInteger(ELD->getRawSubclassData());
ID.AddInteger(ELD->getPointerInfo().getAddrSpace());
ID.AddInteger(ELD->getMemOperand()->getFlags());
break;
}
case ISD::VP_STORE: {
const VPStoreSDNode *EST = cast<VPStoreSDNode>(N);
ID.AddInteger(EST->getMemoryVT().getRawBits());
ID.AddInteger(EST->getRawSubclassData());
ID.AddInteger(EST->getPointerInfo().getAddrSpace());
ID.AddInteger(EST->getMemOperand()->getFlags());
break;
}
case ISD::EXPERIMENTAL_VP_STRIDED_LOAD: {
const VPStridedLoadSDNode *SLD = cast<VPStridedLoadSDNode>(N);
ID.AddInteger(SLD->getMemoryVT().getRawBits());
ID.AddInteger(SLD->getRawSubclassData());
ID.AddInteger(SLD->getPointerInfo().getAddrSpace());
break;
}
case ISD::EXPERIMENTAL_VP_STRIDED_STORE: {
const VPStridedStoreSDNode *SST = cast<VPStridedStoreSDNode>(N);
ID.AddInteger(SST->getMemoryVT().getRawBits());
ID.AddInteger(SST->getRawSubclassData());
ID.AddInteger(SST->getPointerInfo().getAddrSpace());
break;
}
case ISD::VP_GATHER: {
const VPGatherSDNode *EG = cast<VPGatherSDNode>(N);
ID.AddInteger(EG->getMemoryVT().getRawBits());
ID.AddInteger(EG->getRawSubclassData());
ID.AddInteger(EG->getPointerInfo().getAddrSpace());
ID.AddInteger(EG->getMemOperand()->getFlags());
break;
}
case ISD::VP_SCATTER: {
const VPScatterSDNode *ES = cast<VPScatterSDNode>(N);
ID.AddInteger(ES->getMemoryVT().getRawBits());
ID.AddInteger(ES->getRawSubclassData());
ID.AddInteger(ES->getPointerInfo().getAddrSpace());
ID.AddInteger(ES->getMemOperand()->getFlags());
break;
}
case ISD::MLOAD: {
const MaskedLoadSDNode *MLD = cast<MaskedLoadSDNode>(N);
ID.AddInteger(MLD->getMemoryVT().getRawBits());
ID.AddInteger(MLD->getRawSubclassData());
ID.AddInteger(MLD->getPointerInfo().getAddrSpace());
ID.AddInteger(MLD->getMemOperand()->getFlags());
break;
}
case ISD::MSTORE: {
const MaskedStoreSDNode *MST = cast<MaskedStoreSDNode>(N);
ID.AddInteger(MST->getMemoryVT().getRawBits());
ID.AddInteger(MST->getRawSubclassData());
ID.AddInteger(MST->getPointerInfo().getAddrSpace());
ID.AddInteger(MST->getMemOperand()->getFlags());
break;
}
case ISD::MGATHER: {
const MaskedGatherSDNode *MG = cast<MaskedGatherSDNode>(N);
ID.AddInteger(MG->getMemoryVT().getRawBits());
ID.AddInteger(MG->getRawSubclassData());
ID.AddInteger(MG->getPointerInfo().getAddrSpace());
ID.AddInteger(MG->getMemOperand()->getFlags());
break;
}
case ISD::MSCATTER: {
const MaskedScatterSDNode *MS = cast<MaskedScatterSDNode>(N);
ID.AddInteger(MS->getMemoryVT().getRawBits());
ID.AddInteger(MS->getRawSubclassData());
ID.AddInteger(MS->getPointerInfo().getAddrSpace());
ID.AddInteger(MS->getMemOperand()->getFlags());
break;
}
case ISD::ATOMIC_CMP_SWAP:
case ISD::ATOMIC_CMP_SWAP_WITH_SUCCESS:
case ISD::ATOMIC_SWAP:
case ISD::ATOMIC_LOAD_ADD:
case ISD::ATOMIC_LOAD_SUB:
case ISD::ATOMIC_LOAD_AND:
case ISD::ATOMIC_LOAD_CLR:
case ISD::ATOMIC_LOAD_OR:
case ISD::ATOMIC_LOAD_XOR:
case ISD::ATOMIC_LOAD_NAND:
case ISD::ATOMIC_LOAD_MIN:
case ISD::ATOMIC_LOAD_MAX:
case ISD::ATOMIC_LOAD_UMIN:
case ISD::ATOMIC_LOAD_UMAX:
case ISD::ATOMIC_LOAD:
case ISD::ATOMIC_STORE: {
const AtomicSDNode *AT = cast<AtomicSDNode>(N);
ID.AddInteger(AT->getMemoryVT().getRawBits());
ID.AddInteger(AT->getRawSubclassData());
ID.AddInteger(AT->getPointerInfo().getAddrSpace());
ID.AddInteger(AT->getMemOperand()->getFlags());
break;
}
case ISD::VECTOR_SHUFFLE: {
ArrayRef<int> Mask = cast<ShuffleVectorSDNode>(N)->getMask();
for (int M : Mask)
ID.AddInteger(M);
break;
}
case ISD::TargetBlockAddress:
case ISD::BlockAddress: {
const BlockAddressSDNode *BA = cast<BlockAddressSDNode>(N);
ID.AddPointer(BA->getBlockAddress());
ID.AddInteger(BA->getOffset());
ID.AddInteger(BA->getTargetFlags());
break;
}
case ISD::AssertAlign:
ID.AddInteger(cast<AssertAlignSDNode>(N)->getAlign().value());
break;
case ISD::PREFETCH:
case ISD::INTRINSIC_VOID:
case ISD::INTRINSIC_W_CHAIN:
// Handled by MemIntrinsicSDNode check after the switch.
break;
} // end switch (N->getOpcode())
// MemIntrinsic nodes could also have subclass data, address spaces, and flags
// to check.
if (auto *MN = dyn_cast<MemIntrinsicSDNode>(N)) {
ID.AddInteger(MN->getRawSubclassData());
ID.AddInteger(MN->getPointerInfo().getAddrSpace());
ID.AddInteger(MN->getMemOperand()->getFlags());
ID.AddInteger(MN->getMemoryVT().getRawBits());
}
}
/// AddNodeIDNode - Generic routine for adding a nodes info to the NodeID
/// data.
static void AddNodeIDNode(FoldingSetNodeID &ID, const SDNode *N) {
AddNodeIDOpcode(ID, N->getOpcode());
// Add the return value info.
AddNodeIDValueTypes(ID, N->getVTList());
// Add the operand info.
AddNodeIDOperands(ID, N->ops());
// Handle SDNode leafs with special info.
AddNodeIDCustom(ID, N);
}
//===----------------------------------------------------------------------===//
// SelectionDAG Class
//===----------------------------------------------------------------------===//
/// doNotCSE - Return true if CSE should not be performed for this node.
static bool doNotCSE(SDNode *N) {
if (N->getValueType(0) == MVT::Glue)
return true; // Never CSE anything that produces a glue result.
switch (N->getOpcode()) {
default: break;
case ISD::HANDLENODE:
case ISD::EH_LABEL:
return true; // Never CSE these nodes.
}
// Check that remaining values produced are not flags.
for (unsigned i = 1, e = N->getNumValues(); i != e; ++i)
if (N->getValueType(i) == MVT::Glue)
return true; // Never CSE anything that produces a glue result.
return false;
}
/// RemoveDeadNodes - This method deletes all unreachable nodes in the
/// SelectionDAG.
void SelectionDAG::RemoveDeadNodes() {
// Create a dummy node (which is not added to allnodes), that adds a reference
// to the root node, preventing it from being deleted.
HandleSDNode Dummy(getRoot());
SmallVector<SDNode*, 128> DeadNodes;
// Add all obviously-dead nodes to the DeadNodes worklist.
for (SDNode &Node : allnodes())
if (Node.use_empty())
DeadNodes.push_back(&Node);
RemoveDeadNodes(DeadNodes);
// If the root changed (e.g. it was a dead load, update the root).
setRoot(Dummy.getValue());
}
/// RemoveDeadNodes - This method deletes the unreachable nodes in the
/// given list, and any nodes that become unreachable as a result.
void SelectionDAG::RemoveDeadNodes(SmallVectorImpl<SDNode *> &DeadNodes) {
// Process the worklist, deleting the nodes and adding their uses to the
// worklist.
while (!DeadNodes.empty()) {
SDNode *N = DeadNodes.pop_back_val();
// Skip to next node if we've already managed to delete the node. This could
// happen if replacing a node causes a node previously added to the node to
// be deleted.
if (N->getOpcode() == ISD::DELETED_NODE)
continue;
for (DAGUpdateListener *DUL = UpdateListeners; DUL; DUL = DUL->Next)
DUL->NodeDeleted(N, nullptr);
// Take the node out of the appropriate CSE map.
RemoveNodeFromCSEMaps(N);
// Next, brutally remove the operand list. This is safe to do, as there are
// no cycles in the graph.
for (SDNode::op_iterator I = N->op_begin(), E = N->op_end(); I != E; ) {
SDUse &Use = *I++;
SDNode *Operand = Use.getNode();
Use.set(SDValue());
// Now that we removed this operand, see if there are no uses of it left.
if (Operand->use_empty())
DeadNodes.push_back(Operand);
}
DeallocateNode(N);
}
}
void SelectionDAG::RemoveDeadNode(SDNode *N){
SmallVector<SDNode*, 16> DeadNodes(1, N);
// Create a dummy node that adds a reference to the root node, preventing
// it from being deleted. (This matters if the root is an operand of the
// dead node.)
HandleSDNode Dummy(getRoot());
RemoveDeadNodes(DeadNodes);
}
void SelectionDAG::DeleteNode(SDNode *N) {
// First take this out of the appropriate CSE map.
RemoveNodeFromCSEMaps(N);
// Finally, remove uses due to operands of this node, remove from the
// AllNodes list, and delete the node.
DeleteNodeNotInCSEMaps(N);
}
void SelectionDAG::DeleteNodeNotInCSEMaps(SDNode *N) {
assert(N->getIterator() != AllNodes.begin() &&
"Cannot delete the entry node!");
assert(N->use_empty() && "Cannot delete a node that is not dead!");
// Drop all of the operands and decrement used node's use counts.
N->DropOperands();
DeallocateNode(N);
}
void SDDbgInfo::add(SDDbgValue *V, bool isParameter) {
assert(!(V->isVariadic() && isParameter));
if (isParameter)
ByvalParmDbgValues.push_back(V);
else
DbgValues.push_back(V);
for (const SDNode *Node : V->getSDNodes())
if (Node)
DbgValMap[Node].push_back(V);
}
void SDDbgInfo::erase(const SDNode *Node) {
DbgValMapType::iterator I = DbgValMap.find(Node);
if (I == DbgValMap.end())
return;
for (auto &Val: I->second)
Val->setIsInvalidated();
DbgValMap.erase(I);
}
void SelectionDAG::DeallocateNode(SDNode *N) {
// If we have operands, deallocate them.
removeOperands(N);
NodeAllocator.Deallocate(AllNodes.remove(N));
// Set the opcode to DELETED_NODE to help catch bugs when node
// memory is reallocated.
// FIXME: There are places in SDag that have grown a dependency on the opcode
// value in the released node.
__asan_unpoison_memory_region(&N->NodeType, sizeof(N->NodeType));
N->NodeType = ISD::DELETED_NODE;
// If any of the SDDbgValue nodes refer to this SDNode, invalidate
// them and forget about that node.
DbgInfo->erase(N);
// Invalidate extra info.
SDEI.erase(N);
}
#ifndef NDEBUG
/// VerifySDNode - Check the given SDNode. Aborts if it is invalid.
static void VerifySDNode(SDNode *N) {
switch (N->getOpcode()) {
default:
break;
case ISD::BUILD_PAIR: {
EVT VT = N->getValueType(0);
assert(N->getNumValues() == 1 && "Too many results!");
assert(!VT.isVector() && (VT.isInteger() || VT.isFloatingPoint()) &&
"Wrong return type!");
assert(N->getNumOperands() == 2 && "Wrong number of operands!");
assert(N->getOperand(0).getValueType() == N->getOperand(1).getValueType() &&
"Mismatched operand types!");
assert(N->getOperand(0).getValueType().isInteger() == VT.isInteger() &&
"Wrong operand type!");
assert(VT.getSizeInBits() == 2 * N->getOperand(0).getValueSizeInBits() &&
"Wrong return type size");
break;
}
case ISD::BUILD_VECTOR: {
assert(N->getNumValues() == 1 && "Too many results!");
assert(N->getValueType(0).isVector() && "Wrong return type!");
assert(N->getNumOperands() == N->getValueType(0).getVectorNumElements() &&
"Wrong number of operands!");
EVT EltVT = N->getValueType(0).getVectorElementType();
for (const SDUse &Op : N->ops()) {
assert((Op.getValueType() == EltVT ||
(EltVT.isInteger() && Op.getValueType().isInteger() &&
EltVT.bitsLE(Op.getValueType()))) &&
"Wrong operand type!");
assert(Op.getValueType() == N->getOperand(0).getValueType() &&
"Operands must all have the same type");
}
break;
}
}
}
#endif // NDEBUG
/// Insert a newly allocated node into the DAG.
///
/// Handles insertion into the all nodes list and CSE map, as well as
/// verification and other common operations when a new node is allocated.
void SelectionDAG::InsertNode(SDNode *N) {
AllNodes.push_back(N);
#ifndef NDEBUG
N->PersistentId = NextPersistentId++;
VerifySDNode(N);
#endif
for (DAGUpdateListener *DUL = UpdateListeners; DUL; DUL = DUL->Next)
DUL->NodeInserted(N);
}
/// RemoveNodeFromCSEMaps - Take the specified node out of the CSE map that
/// correspond to it. This is useful when we're about to delete or repurpose
/// the node. We don't want future request for structurally identical nodes
/// to return N anymore.
bool SelectionDAG::RemoveNodeFromCSEMaps(SDNode *N) {
bool Erased = false;
switch (N->getOpcode()) {
case ISD::HANDLENODE: return false; // noop.
case ISD::CONDCODE:
assert(CondCodeNodes[cast<CondCodeSDNode>(N)->get()] &&
"Cond code doesn't exist!");
Erased = CondCodeNodes[cast<CondCodeSDNode>(N)->get()] != nullptr;
CondCodeNodes[cast<CondCodeSDNode>(N)->get()] = nullptr;
break;
case ISD::ExternalSymbol:
Erased = ExternalSymbols.erase(cast<ExternalSymbolSDNode>(N)->getSymbol());
break;
case ISD::TargetExternalSymbol: {
ExternalSymbolSDNode *ESN = cast<ExternalSymbolSDNode>(N);
Erased = TargetExternalSymbols.erase(std::pair<std::string, unsigned>(
ESN->getSymbol(), ESN->getTargetFlags()));
break;
}
case ISD::MCSymbol: {
auto *MCSN = cast<MCSymbolSDNode>(N);
Erased = MCSymbols.erase(MCSN->getMCSymbol());
break;
}
case ISD::VALUETYPE: {
EVT VT = cast<VTSDNode>(N)->getVT();
if (VT.isExtended()) {
Erased = ExtendedValueTypeNodes.erase(VT);
} else {
Erased = ValueTypeNodes[VT.getSimpleVT().SimpleTy] != nullptr;
ValueTypeNodes[VT.getSimpleVT().SimpleTy] = nullptr;
}
break;
}
default:
// Remove it from the CSE Map.
assert(N->getOpcode() != ISD::DELETED_NODE && "DELETED_NODE in CSEMap!");
assert(N->getOpcode() != ISD::EntryToken && "EntryToken in CSEMap!");
Erased = CSEMap.RemoveNode(N);
break;
}
#ifndef NDEBUG
// Verify that the node was actually in one of the CSE maps, unless it has a
// glue result (which cannot be CSE'd) or is one of the special cases that are
// not subject to CSE.
if (!Erased && N->getValueType(N->getNumValues()-1) != MVT::Glue &&
!N->isMachineOpcode() && !doNotCSE(N)) {
N->dump(this);
dbgs() << "\n";
llvm_unreachable("Node is not in map!");
}
#endif
return Erased;
}
/// AddModifiedNodeToCSEMaps - The specified node has been removed from the CSE
/// maps and modified in place. Add it back to the CSE maps, unless an identical
/// node already exists, in which case transfer all its users to the existing
/// node. This transfer can potentially trigger recursive merging.
void
SelectionDAG::AddModifiedNodeToCSEMaps(SDNode *N) {
// For node types that aren't CSE'd, just act as if no identical node
// already exists.
if (!doNotCSE(N)) {
SDNode *Existing = CSEMap.GetOrInsertNode(N);
if (Existing != N) {
// If there was already an existing matching node, use ReplaceAllUsesWith
// to replace the dead one with the existing one. This can cause
// recursive merging of other unrelated nodes down the line.
ReplaceAllUsesWith(N, Existing);
// N is now dead. Inform the listeners and delete it.
for (DAGUpdateListener *DUL = UpdateListeners; DUL; DUL = DUL->Next)
DUL->NodeDeleted(N, Existing);
DeleteNodeNotInCSEMaps(N);
return;
}
}
// If the node doesn't already exist, we updated it. Inform listeners.
for (DAGUpdateListener *DUL = UpdateListeners; DUL; DUL = DUL->Next)
DUL->NodeUpdated(N);
}
/// FindModifiedNodeSlot - Find a slot for the specified node if its operands
/// were replaced with those specified. If this node is never memoized,
/// return null, otherwise return a pointer to the slot it would take. If a
/// node already exists with these operands, the slot will be non-null.
SDNode *SelectionDAG::FindModifiedNodeSlot(SDNode *N, SDValue Op,
void *&InsertPos) {
if (doNotCSE(N))
return nullptr;
SDValue Ops[] = { Op };
FoldingSetNodeID ID;
AddNodeIDNode(ID, N->getOpcode(), N->getVTList(), Ops);
AddNodeIDCustom(ID, N);
SDNode *Node = FindNodeOrInsertPos(ID, SDLoc(N), InsertPos);
if (Node)
Node->intersectFlagsWith(N->getFlags());
return Node;
}
/// FindModifiedNodeSlot - Find a slot for the specified node if its operands
/// were replaced with those specified. If this node is never memoized,
/// return null, otherwise return a pointer to the slot it would take. If a
/// node already exists with these operands, the slot will be non-null.
SDNode *SelectionDAG::FindModifiedNodeSlot(SDNode *N,
SDValue Op1, SDValue Op2,
void *&InsertPos) {
if (doNotCSE(N))
return nullptr;
SDValue Ops[] = { Op1, Op2 };
FoldingSetNodeID ID;
AddNodeIDNode(ID, N->getOpcode(), N->getVTList(), Ops);
AddNodeIDCustom(ID, N);
SDNode *Node = FindNodeOrInsertPos(ID, SDLoc(N), InsertPos);
if (Node)
Node->intersectFlagsWith(N->getFlags());
return Node;
}
/// FindModifiedNodeSlot - Find a slot for the specified node if its operands
/// were replaced with those specified. If this node is never memoized,
/// return null, otherwise return a pointer to the slot it would take. If a
/// node already exists with these operands, the slot will be non-null.
SDNode *SelectionDAG::FindModifiedNodeSlot(SDNode *N, ArrayRef<SDValue> Ops,
void *&InsertPos) {
if (doNotCSE(N))
return nullptr;
FoldingSetNodeID ID;
AddNodeIDNode(ID, N->getOpcode(), N->getVTList(), Ops);
AddNodeIDCustom(ID, N);
SDNode *Node = FindNodeOrInsertPos(ID, SDLoc(N), InsertPos);
if (Node)
Node->intersectFlagsWith(N->getFlags());
return Node;
}
Align SelectionDAG::getEVTAlign(EVT VT) const {
Type *Ty = VT == MVT::iPTR ? PointerType::get(*getContext(), 0)
: VT.getTypeForEVT(*getContext());
return getDataLayout().getABITypeAlign(Ty);
}
// EntryNode could meaningfully have debug info if we can find it...
SelectionDAG::SelectionDAG(const TargetMachine &tm, CodeGenOptLevel OL)
: TM(tm), OptLevel(OL), EntryNode(ISD::EntryToken, 0, DebugLoc(),
getVTList(MVT::Other, MVT::Glue)),
Root(getEntryNode()) {
InsertNode(&EntryNode);
DbgInfo = new SDDbgInfo();
}
void SelectionDAG::init(MachineFunction &NewMF,
OptimizationRemarkEmitter &NewORE, Pass *PassPtr,
const TargetLibraryInfo *LibraryInfo,
UniformityInfo *NewUA, ProfileSummaryInfo *PSIin,
BlockFrequencyInfo *BFIin,
FunctionVarLocs const *VarLocs) {
MF = &NewMF;
SDAGISelPass = PassPtr;
ORE = &NewORE;
TLI = getSubtarget().getTargetLowering();
TSI = getSubtarget().getSelectionDAGInfo();
LibInfo = LibraryInfo;
Context = &MF->getFunction().getContext();
UA = NewUA;
PSI = PSIin;
BFI = BFIin;
FnVarLocs = VarLocs;
}
SelectionDAG::~SelectionDAG() {
assert(!UpdateListeners && "Dangling registered DAGUpdateListeners");
allnodes_clear();
OperandRecycler.clear(OperandAllocator);
delete DbgInfo;
}
bool SelectionDAG::shouldOptForSize() const {
return MF->getFunction().hasOptSize() ||
llvm::shouldOptimizeForSize(FLI->MBB->getBasicBlock(), PSI, BFI);
}
void SelectionDAG::allnodes_clear() {
assert(&*AllNodes.begin() == &EntryNode);
AllNodes.remove(AllNodes.begin());
while (!AllNodes.empty())
DeallocateNode(&AllNodes.front());
#ifndef NDEBUG
NextPersistentId = 0;
#endif
}
SDNode *SelectionDAG::FindNodeOrInsertPos(const FoldingSetNodeID &ID,
void *&InsertPos) {
SDNode *N = CSEMap.FindNodeOrInsertPos(ID, InsertPos);
if (N) {
switch (N->getOpcode()) {
default: break;
case ISD::Constant:
case ISD::ConstantFP:
llvm_unreachable("Querying for Constant and ConstantFP nodes requires "
"debug location. Use another overload.");
}
}
return N;
}
SDNode *SelectionDAG::FindNodeOrInsertPos(const FoldingSetNodeID &ID,
const SDLoc &DL, void *&InsertPos) {
SDNode *N = CSEMap.FindNodeOrInsertPos(ID, InsertPos);
if (N) {
switch (N->getOpcode()) {
case ISD::Constant:
case ISD::ConstantFP:
// Erase debug location from the node if the node is used at several
// different places. Do not propagate one location to all uses as it
// will cause a worse single stepping debugging experience.
if (N->getDebugLoc() != DL.getDebugLoc())
N->setDebugLoc(DebugLoc());
break;
default:
// When the node's point of use is located earlier in the instruction
// sequence than its prior point of use, update its debug info to the
// earlier location.
if (DL.getIROrder() && DL.getIROrder() < N->getIROrder())
N->setDebugLoc(DL.getDebugLoc());
break;
}
}
return N;
}
void SelectionDAG::clear() {
allnodes_clear();
OperandRecycler.clear(OperandAllocator);
OperandAllocator.Reset();
CSEMap.clear();
ExtendedValueTypeNodes.clear();
ExternalSymbols.clear();
TargetExternalSymbols.clear();
MCSymbols.clear();
SDEI.clear();
std::fill(CondCodeNodes.begin(), CondCodeNodes.end(),
static_cast<CondCodeSDNode*>(nullptr));
std::fill(ValueTypeNodes.begin(), ValueTypeNodes.end(),
static_cast<SDNode*>(nullptr));
EntryNode.UseList = nullptr;
InsertNode(&EntryNode);
Root = getEntryNode();
DbgInfo->clear();
}
SDValue SelectionDAG::getFPExtendOrRound(SDValue Op, const SDLoc &DL, EVT VT) {
return VT.bitsGT(Op.getValueType())
? getNode(ISD::FP_EXTEND, DL, VT, Op)
: getNode(ISD::FP_ROUND, DL, VT, Op,
getIntPtrConstant(0, DL, /*isTarget=*/true));
}
std::pair<SDValue, SDValue>
SelectionDAG::getStrictFPExtendOrRound(SDValue Op, SDValue Chain,
const SDLoc &DL, EVT VT) {
assert(!VT.bitsEq(Op.getValueType()) &&
"Strict no-op FP extend/round not allowed.");
SDValue Res =
VT.bitsGT(Op.getValueType())
? getNode(ISD::STRICT_FP_EXTEND, DL, {VT, MVT::Other}, {Chain, Op})
: getNode(ISD::STRICT_FP_ROUND, DL, {VT, MVT::Other},
{Chain, Op, getIntPtrConstant(0, DL)});
return std::pair<SDValue, SDValue>(Res, SDValue(Res.getNode(), 1));
}
SDValue SelectionDAG::getAnyExtOrTrunc(SDValue Op, const SDLoc &DL, EVT VT) {
return VT.bitsGT(Op.getValueType()) ?
getNode(ISD::ANY_EXTEND, DL, VT, Op) :
getNode(ISD::TRUNCATE, DL, VT, Op);
}
SDValue SelectionDAG::getSExtOrTrunc(SDValue Op, const SDLoc &DL, EVT VT) {
return VT.bitsGT(Op.getValueType()) ?
getNode(ISD::SIGN_EXTEND, DL, VT, Op) :
getNode(ISD::TRUNCATE, DL, VT, Op);
}
SDValue SelectionDAG::getZExtOrTrunc(SDValue Op, const SDLoc &DL, EVT VT) {
return VT.bitsGT(Op.getValueType()) ?
getNode(ISD::ZERO_EXTEND, DL, VT, Op) :
getNode(ISD::TRUNCATE, DL, VT, Op);
}
SDValue SelectionDAG::getBitcastedAnyExtOrTrunc(SDValue Op, const SDLoc &DL,
EVT VT) {
assert(!VT.isVector());
auto Type = Op.getValueType();
SDValue DestOp;
if (Type == VT)
return Op;
auto Size = Op.getValueSizeInBits();
DestOp = getBitcast(MVT::getIntegerVT(Size), Op);
if (DestOp.getValueType() == VT)
return DestOp;
return getAnyExtOrTrunc(DestOp, DL, VT);
}
SDValue SelectionDAG::getBitcastedSExtOrTrunc(SDValue Op, const SDLoc &DL,
EVT VT) {
assert(!VT.isVector());
auto Type = Op.getValueType();
SDValue DestOp;
if (Type == VT)
return Op;
auto Size = Op.getValueSizeInBits();
DestOp = getBitcast(MVT::getIntegerVT(Size), Op);
if (DestOp.getValueType() == VT)
return DestOp;
return getSExtOrTrunc(DestOp, DL, VT);
}
SDValue SelectionDAG::getBitcastedZExtOrTrunc(SDValue Op, const SDLoc &DL,
EVT VT) {
assert(!VT.isVector());
auto Type = Op.getValueType();
SDValue DestOp;
if (Type == VT)
return Op;
auto Size = Op.getValueSizeInBits();
DestOp = getBitcast(MVT::getIntegerVT(Size), Op);
if (DestOp.getValueType() == VT)
return DestOp;
return getZExtOrTrunc(DestOp, DL, VT);
}
SDValue SelectionDAG::getBoolExtOrTrunc(SDValue Op, const SDLoc &SL, EVT VT,
EVT OpVT) {
if (VT.bitsLE(Op.getValueType()))
return getNode(ISD::TRUNCATE, SL, VT, Op);
TargetLowering::BooleanContent BType = TLI->getBooleanContents(OpVT);
return getNode(TLI->getExtendForContent(BType), SL, VT, Op);
}
SDValue SelectionDAG::getZeroExtendInReg(SDValue Op, const SDLoc &DL, EVT VT) {
EVT OpVT = Op.getValueType();
assert(VT.isInteger() && OpVT.isInteger() &&
"Cannot getZeroExtendInReg FP types");
assert(VT.isVector() == OpVT.isVector() &&
"getZeroExtendInReg type should be vector iff the operand "
"type is vector!");
assert((!VT.isVector() ||
VT.getVectorElementCount() == OpVT.getVectorElementCount()) &&
"Vector element counts must match in getZeroExtendInReg");
assert(VT.bitsLE(OpVT) && "Not extending!");
if (OpVT == VT)
return Op;
APInt Imm = APInt::getLowBitsSet(OpVT.getScalarSizeInBits(),
VT.getScalarSizeInBits());
return getNode(ISD::AND, DL, OpVT, Op, getConstant(Imm, DL, OpVT));
}
SDValue SelectionDAG::getPtrExtOrTrunc(SDValue Op, const SDLoc &DL, EVT VT) {
// Only unsigned pointer semantics are supported right now. In the future this
// might delegate to TLI to check pointer signedness.
return getZExtOrTrunc(Op, DL, VT);
}
SDValue SelectionDAG::getPtrExtendInReg(SDValue Op, const SDLoc &DL, EVT VT) {
// Only unsigned pointer semantics are supported right now. In the future this
// might delegate to TLI to check pointer signedness.
return getZeroExtendInReg(Op, DL, VT);
}
SDValue SelectionDAG::getNegative(SDValue Val, const SDLoc &DL, EVT VT) {
return getNode(ISD::SUB, DL, VT, getConstant(0, DL, VT), Val);
}
/// getNOT - Create a bitwise NOT operation as (XOR Val, -1).
SDValue SelectionDAG::getNOT(const SDLoc &DL, SDValue Val, EVT VT) {
return getNode(ISD::XOR, DL, VT, Val, getAllOnesConstant(DL, VT));
}
SDValue SelectionDAG::getLogicalNOT(const SDLoc &DL, SDValue Val, EVT VT) {
SDValue TrueValue = getBoolConstant(true, DL, VT, VT);
return getNode(ISD::XOR, DL, VT, Val, TrueValue);
}
SDValue SelectionDAG::getVPLogicalNOT(const SDLoc &DL, SDValue Val,
SDValue Mask, SDValue EVL, EVT VT) {
SDValue TrueValue = getBoolConstant(true, DL, VT, VT);
return getNode(ISD::VP_XOR, DL, VT, Val, TrueValue, Mask, EVL);
}
SDValue SelectionDAG::getVPPtrExtOrTrunc(const SDLoc &DL, EVT VT, SDValue Op,
SDValue Mask, SDValue EVL) {
return getVPZExtOrTrunc(DL, VT, Op, Mask, EVL);
}
SDValue SelectionDAG::getVPZExtOrTrunc(const SDLoc &DL, EVT VT, SDValue Op,
SDValue Mask, SDValue EVL) {
if (VT.bitsGT(Op.getValueType()))
return getNode(ISD::VP_ZERO_EXTEND, DL, VT, Op, Mask, EVL);
if (VT.bitsLT(Op.getValueType()))
return getNode(ISD::VP_TRUNCATE, DL, VT, Op, Mask, EVL);
return Op;
}
SDValue SelectionDAG::getBoolConstant(bool V, const SDLoc &DL, EVT VT,
EVT OpVT) {
if (!V)
return getConstant(0, DL, VT);
switch (TLI->getBooleanContents(OpVT)) {
case TargetLowering::ZeroOrOneBooleanContent:
case TargetLowering::UndefinedBooleanContent:
return getConstant(1, DL, VT);
case TargetLowering::ZeroOrNegativeOneBooleanContent:
return getAllOnesConstant(DL, VT);
}
llvm_unreachable("Unexpected boolean content enum!");
}
SDValue SelectionDAG::getConstant(uint64_t Val, const SDLoc &DL, EVT VT,
bool isT, bool isO) {
EVT EltVT = VT.getScalarType();
assert((EltVT.getSizeInBits() >= 64 ||
(uint64_t)((int64_t)Val >> EltVT.getSizeInBits()) + 1 < 2) &&
"getConstant with a uint64_t value that doesn't fit in the type!");
return getConstant(APInt(EltVT.getSizeInBits(), Val), DL, VT, isT, isO);
}
SDValue SelectionDAG::getConstant(const APInt &Val, const SDLoc &DL, EVT VT,
bool isT, bool isO) {
return getConstant(*ConstantInt::get(*Context, Val), DL, VT, isT, isO);
}
SDValue SelectionDAG::getConstant(const ConstantInt &Val, const SDLoc &DL,
EVT VT, bool isT, bool isO) {
assert(VT.isInteger() && "Cannot create FP integer constant!");
EVT EltVT = VT.getScalarType();
const ConstantInt *Elt = &Val;
// In some cases the vector type is legal but the element type is illegal and
// needs to be promoted, for example v8i8 on ARM. In this case, promote the
// inserted value (the type does not need to match the vector element type).
// Any extra bits introduced will be truncated away.
if (VT.isVector() && TLI->getTypeAction(*getContext(), EltVT) ==
TargetLowering::TypePromoteInteger) {
EltVT = TLI->getTypeToTransformTo(*getContext(), EltVT);
APInt NewVal;
if (TLI->isSExtCheaperThanZExt(VT.getScalarType(), EltVT))
NewVal = Elt->getValue().sextOrTrunc(EltVT.getSizeInBits());
else
NewVal = Elt->getValue().zextOrTrunc(EltVT.getSizeInBits());
Elt = ConstantInt::get(*getContext(), NewVal);
}
// In other cases the element type is illegal and needs to be expanded, for
// example v2i64 on MIPS32. In this case, find the nearest legal type, split
// the value into n parts and use a vector type with n-times the elements.
// Then bitcast to the type requested.
// Legalizing constants too early makes the DAGCombiner's job harder so we
// only legalize if the DAG tells us we must produce legal types.
else if (NewNodesMustHaveLegalTypes && VT.isVector() &&
TLI->getTypeAction(*getContext(), EltVT) ==
TargetLowering::TypeExpandInteger) {
const APInt &NewVal = Elt->getValue();
EVT ViaEltVT = TLI->getTypeToTransformTo(*getContext(), EltVT);
unsigned ViaEltSizeInBits = ViaEltVT.getSizeInBits();
// For scalable vectors, try to use a SPLAT_VECTOR_PARTS node.
if (VT.isScalableVector() ||
TLI->isOperationLegal(ISD::SPLAT_VECTOR, VT)) {
assert(EltVT.getSizeInBits() % ViaEltSizeInBits == 0 &&
"Can only handle an even split!");
unsigned Parts = EltVT.getSizeInBits() / ViaEltSizeInBits;
SmallVector<SDValue, 2> ScalarParts;
for (unsigned i = 0; i != Parts; ++i)
ScalarParts.push_back(getConstant(
NewVal.extractBits(ViaEltSizeInBits, i * ViaEltSizeInBits), DL,
ViaEltVT, isT, isO));
return getNode(ISD::SPLAT_VECTOR_PARTS, DL, VT, ScalarParts);
}
unsigned ViaVecNumElts = VT.getSizeInBits() / ViaEltSizeInBits;
EVT ViaVecVT = EVT::getVectorVT(*getContext(), ViaEltVT, ViaVecNumElts);
// Check the temporary vector is the correct size. If this fails then
// getTypeToTransformTo() probably returned a type whose size (in bits)
// isn't a power-of-2 factor of the requested type size.
assert(ViaVecVT.getSizeInBits() == VT.getSizeInBits());
SmallVector<SDValue, 2> EltParts;
for (unsigned i = 0; i < ViaVecNumElts / VT.getVectorNumElements(); ++i)
EltParts.push_back(getConstant(
NewVal.extractBits(ViaEltSizeInBits, i * ViaEltSizeInBits), DL,
ViaEltVT, isT, isO));
// EltParts is currently in little endian order. If we actually want
// big-endian order then reverse it now.
if (getDataLayout().isBigEndian())
std::reverse(EltParts.begin(), EltParts.end());
// The elements must be reversed when the element order is different
// to the endianness of the elements (because the BITCAST is itself a
// vector shuffle in this situation). However, we do not need any code to
// perform this reversal because getConstant() is producing a vector
// splat.
// This situation occurs in MIPS MSA.
SmallVector<SDValue, 8> Ops;
for (unsigned i = 0, e = VT.getVectorNumElements(); i != e; ++i)
llvm::append_range(Ops, EltParts);
SDValue V =
getNode(ISD::BITCAST, DL, VT, getBuildVector(ViaVecVT, DL, Ops));
return V;
}
assert(Elt->getBitWidth() == EltVT.getSizeInBits() &&
"APInt size does not match type size!");
unsigned Opc = isT ? ISD::TargetConstant : ISD::Constant;
FoldingSetNodeID ID;
AddNodeIDNode(ID, Opc, getVTList(EltVT), std::nullopt);
ID.AddPointer(Elt);
ID.AddBoolean(isO);
void *IP = nullptr;
SDNode *N = nullptr;
if ((N = FindNodeOrInsertPos(ID, DL, IP)))
if (!VT.isVector())
return SDValue(N, 0);
if (!N) {
N = newSDNode<ConstantSDNode>(isT, isO, Elt, EltVT);
CSEMap.InsertNode(N, IP);
InsertNode(N);
NewSDValueDbgMsg(SDValue(N, 0), "Creating constant: ", this);
}
SDValue Result(N, 0);
if (VT.isVector())
Result = getSplat(VT, DL, Result);
return Result;
}
SDValue SelectionDAG::getIntPtrConstant(uint64_t Val, const SDLoc &DL,
bool isTarget) {
return getConstant(Val, DL, TLI->getPointerTy(getDataLayout()), isTarget);
}
SDValue SelectionDAG::getShiftAmountConstant(uint64_t Val, EVT VT,
const SDLoc &DL, bool LegalTypes) {
assert(VT.isInteger() && "Shift amount is not an integer type!");
EVT ShiftVT = TLI->getShiftAmountTy(VT, getDataLayout(), LegalTypes);
return getConstant(Val, DL, ShiftVT);
}
SDValue SelectionDAG::getShiftAmountConstant(const APInt &Val, EVT VT,
const SDLoc &DL, bool LegalTypes) {
assert(Val.ult(VT.getScalarSizeInBits()) && "Out of range shift");
return getShiftAmountConstant(Val.getZExtValue(), VT, DL, LegalTypes);
}
SDValue SelectionDAG::getVectorIdxConstant(uint64_t Val, const SDLoc &DL,
bool isTarget) {
return getConstant(Val, DL, TLI->getVectorIdxTy(getDataLayout()), isTarget);
}
SDValue SelectionDAG::getConstantFP(const APFloat &V, const SDLoc &DL, EVT VT,
bool isTarget) {
return getConstantFP(*ConstantFP::get(*getContext(), V), DL, VT, isTarget);
}
SDValue SelectionDAG::getConstantFP(const ConstantFP &V, const SDLoc &DL,
EVT VT, bool isTarget) {
assert(VT.isFloatingPoint() && "Cannot create integer FP constant!");
EVT EltVT = VT.getScalarType();
// Do the map lookup using the actual bit pattern for the floating point
// value, so that we don't have problems with 0.0 comparing equal to -0.0, and
// we don't have issues with SNANs.
unsigned Opc = isTarget ? ISD::TargetConstantFP : ISD::ConstantFP;
FoldingSetNodeID ID;
AddNodeIDNode(ID, Opc, getVTList(EltVT), std::nullopt);
ID.AddPointer(&V);
void *IP = nullptr;
SDNode *N = nullptr;
if ((N = FindNodeOrInsertPos(ID, DL, IP)))
if (!VT.isVector())
return SDValue(N, 0);
if (!N) {
N = newSDNode<ConstantFPSDNode>(isTarget, &V, EltVT);
CSEMap.InsertNode(N, IP);
InsertNode(N);
}
SDValue Result(N, 0);
if (VT.isVector())
Result = getSplat(VT, DL, Result);
NewSDValueDbgMsg(Result, "Creating fp constant: ", this);
return Result;
}
SDValue SelectionDAG::getConstantFP(double Val, const SDLoc &DL, EVT VT,
bool isTarget) {
EVT EltVT = VT.getScalarType();
if (EltVT == MVT::f32)
return getConstantFP(APFloat((float)Val), DL, VT, isTarget);
if (EltVT == MVT::f64)
return getConstantFP(APFloat(Val), DL, VT, isTarget);
if (EltVT == MVT::f80 || EltVT == MVT::f128 || EltVT == MVT::ppcf128 ||
EltVT == MVT::f16 || EltVT == MVT::bf16) {
bool Ignored;
APFloat APF = APFloat(Val);
APF.convert(EVTToAPFloatSemantics(EltVT), APFloat::rmNearestTiesToEven,
&Ignored);
return getConstantFP(APF, DL, VT, isTarget);
}
llvm_unreachable("Unsupported type in getConstantFP");
}
SDValue SelectionDAG::getGlobalAddress(const GlobalValue *GV, const SDLoc &DL,
EVT VT, int64_t Offset, bool isTargetGA,
unsigned TargetFlags) {
assert((TargetFlags == 0 || isTargetGA) &&
"Cannot set target flags on target-independent globals");
// Truncate (with sign-extension) the offset value to the pointer size.
unsigned BitWidth = getDataLayout().getPointerTypeSizeInBits(GV->getType());
if (BitWidth < 64)
Offset = SignExtend64(Offset, BitWidth);
unsigned Opc;
if (GV->isThreadLocal())
Opc = isTargetGA ? ISD::TargetGlobalTLSAddress : ISD::GlobalTLSAddress;
else
Opc = isTargetGA ? ISD::TargetGlobalAddress : ISD::GlobalAddress;
FoldingSetNodeID ID;
AddNodeIDNode(ID, Opc, getVTList(VT), std::nullopt);
ID.AddPointer(GV);
ID.AddInteger(Offset);
ID.AddInteger(TargetFlags);
void *IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, DL, IP))
return SDValue(E, 0);
auto *N = newSDNode<GlobalAddressSDNode>(
Opc, DL.getIROrder(), DL.getDebugLoc(), GV, VT, Offset, TargetFlags);
CSEMap.InsertNode(N, IP);
InsertNode(N);
return SDValue(N, 0);
}
SDValue SelectionDAG::getFrameIndex(int FI, EVT VT, bool isTarget) {
unsigned Opc = isTarget ? ISD::TargetFrameIndex : ISD::FrameIndex;
FoldingSetNodeID ID;
AddNodeIDNode(ID, Opc, getVTList(VT), std::nullopt);
ID.AddInteger(FI);
void *IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, IP))
return SDValue(E, 0);
auto *N = newSDNode<FrameIndexSDNode>(FI, VT, isTarget);
CSEMap.InsertNode(N, IP);
InsertNode(N);
return SDValue(N, 0);
}
SDValue SelectionDAG::getJumpTable(int JTI, EVT VT, bool isTarget,
unsigned TargetFlags) {
assert((TargetFlags == 0 || isTarget) &&
"Cannot set target flags on target-independent jump tables");
unsigned Opc = isTarget ? ISD::TargetJumpTable : ISD::JumpTable;
FoldingSetNodeID ID;
AddNodeIDNode(ID, Opc, getVTList(VT), std::nullopt);
ID.AddInteger(JTI);
ID.AddInteger(TargetFlags);
void *IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, IP))
return SDValue(E, 0);
auto *N = newSDNode<JumpTableSDNode>(JTI, VT, isTarget, TargetFlags);
CSEMap.InsertNode(N, IP);
InsertNode(N);
return SDValue(N, 0);
}
SDValue SelectionDAG::getJumpTableDebugInfo(int JTI, SDValue Chain,
const SDLoc &DL) {
EVT PTy = getTargetLoweringInfo().getPointerTy(getDataLayout());
return getNode(ISD::JUMP_TABLE_DEBUG_INFO, DL, MVT::Glue, Chain,
getTargetConstant(static_cast<uint64_t>(JTI), DL, PTy, true));
}
SDValue SelectionDAG::getConstantPool(const Constant *C, EVT VT,
MaybeAlign Alignment, int Offset,
bool isTarget, unsigned TargetFlags) {
assert((TargetFlags == 0 || isTarget) &&
"Cannot set target flags on target-independent globals");
if (!Alignment)
Alignment = shouldOptForSize()
? getDataLayout().getABITypeAlign(C->getType())
: getDataLayout().getPrefTypeAlign(C->getType());
unsigned Opc = isTarget ? ISD::TargetConstantPool : ISD::ConstantPool;
FoldingSetNodeID ID;
AddNodeIDNode(ID, Opc, getVTList(VT), std::nullopt);
ID.AddInteger(Alignment->value());
ID.AddInteger(Offset);
ID.AddPointer(C);
ID.AddInteger(TargetFlags);
void *IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, IP))
return SDValue(E, 0);
auto *N = newSDNode<ConstantPoolSDNode>(isTarget, C, VT, Offset, *Alignment,
TargetFlags);
CSEMap.InsertNode(N, IP);
InsertNode(N);
SDValue V = SDValue(N, 0);
NewSDValueDbgMsg(V, "Creating new constant pool: ", this);
return V;
}
SDValue SelectionDAG::getConstantPool(MachineConstantPoolValue *C, EVT VT,
MaybeAlign Alignment, int Offset,
bool isTarget, unsigned TargetFlags) {
assert((TargetFlags == 0 || isTarget) &&
"Cannot set target flags on target-independent globals");
if (!Alignment)
Alignment = getDataLayout().getPrefTypeAlign(C->getType());
unsigned Opc = isTarget ? ISD::TargetConstantPool : ISD::ConstantPool;
FoldingSetNodeID ID;
AddNodeIDNode(ID, Opc, getVTList(VT), std::nullopt);
ID.AddInteger(Alignment->value());
ID.AddInteger(Offset);
C->addSelectionDAGCSEId(ID);
ID.AddInteger(TargetFlags);
void *IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, IP))
return SDValue(E, 0);
auto *N = newSDNode<ConstantPoolSDNode>(isTarget, C, VT, Offset, *Alignment,
TargetFlags);
CSEMap.InsertNode(N, IP);
InsertNode(N);
return SDValue(N, 0);
}
SDValue SelectionDAG::getBasicBlock(MachineBasicBlock *MBB) {
FoldingSetNodeID ID;
AddNodeIDNode(ID, ISD::BasicBlock, getVTList(MVT::Other), std::nullopt);
ID.AddPointer(MBB);
void *IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, IP))
return SDValue(E, 0);
auto *N = newSDNode<BasicBlockSDNode>(MBB);
CSEMap.InsertNode(N, IP);
InsertNode(N);
return SDValue(N, 0);
}
SDValue SelectionDAG::getValueType(EVT VT) {
if (VT.isSimple() && (unsigned)VT.getSimpleVT().SimpleTy >=
ValueTypeNodes.size())
ValueTypeNodes.resize(VT.getSimpleVT().SimpleTy+1);
SDNode *&N = VT.isExtended() ?
ExtendedValueTypeNodes[VT] : ValueTypeNodes[VT.getSimpleVT().SimpleTy];
if (N) return SDValue(N, 0);
N = newSDNode<VTSDNode>(VT);
InsertNode(N);
return SDValue(N, 0);
}
SDValue SelectionDAG::getExternalSymbol(const char *Sym, EVT VT) {
SDNode *&N = ExternalSymbols[Sym];
if (N) return SDValue(N, 0);
N = newSDNode<ExternalSymbolSDNode>(false, Sym, 0, VT);
InsertNode(N);
return SDValue(N, 0);
}
SDValue SelectionDAG::getMCSymbol(MCSymbol *Sym, EVT VT) {
SDNode *&N = MCSymbols[Sym];
if (N)
return SDValue(N, 0);
N = newSDNode<MCSymbolSDNode>(Sym, VT);
InsertNode(N);
return SDValue(N, 0);
}
SDValue SelectionDAG::getTargetExternalSymbol(const char *Sym, EVT VT,
unsigned TargetFlags) {
SDNode *&N =
TargetExternalSymbols[std::pair<std::string, unsigned>(Sym, TargetFlags)];
if (N) return SDValue(N, 0);
N = newSDNode<ExternalSymbolSDNode>(true, Sym, TargetFlags, VT);
InsertNode(N);
return SDValue(N, 0);
}
SDValue SelectionDAG::getCondCode(ISD::CondCode Cond) {
if ((unsigned)Cond >= CondCodeNodes.size())
CondCodeNodes.resize(Cond+1);
if (!CondCodeNodes[Cond]) {
auto *N = newSDNode<CondCodeSDNode>(Cond);
CondCodeNodes[Cond] = N;
InsertNode(N);
}
return SDValue(CondCodeNodes[Cond], 0);
}
SDValue SelectionDAG::getVScale(const SDLoc &DL, EVT VT, APInt MulImm,
bool ConstantFold) {
assert(MulImm.getBitWidth() == VT.getSizeInBits() &&
"APInt size does not match type size!");
if (MulImm == 0)
return getConstant(0, DL, VT);
if (ConstantFold) {
const MachineFunction &MF = getMachineFunction();
const Function &F = MF.getFunction();
ConstantRange CR = getVScaleRange(&F, 64);
if (const APInt *C = CR.getSingleElement())
return getConstant(MulImm * C->getZExtValue(), DL, VT);
}
return getNode(ISD::VSCALE, DL, VT, getConstant(MulImm, DL, VT));
}
SDValue SelectionDAG::getElementCount(const SDLoc &DL, EVT VT, ElementCount EC,
bool ConstantFold) {
if (EC.isScalable())
return getVScale(DL, VT,
APInt(VT.getSizeInBits(), EC.getKnownMinValue()));
return getConstant(EC.getKnownMinValue(), DL, VT);
}
SDValue SelectionDAG::getStepVector(const SDLoc &DL, EVT ResVT) {
APInt One(ResVT.getScalarSizeInBits(), 1);
return getStepVector(DL, ResVT, One);
}
SDValue SelectionDAG::getStepVector(const SDLoc &DL, EVT ResVT, APInt StepVal) {
assert(ResVT.getScalarSizeInBits() == StepVal.getBitWidth());
if (ResVT.isScalableVector())
return getNode(
ISD::STEP_VECTOR, DL, ResVT,
getTargetConstant(StepVal, DL, ResVT.getVectorElementType()));
SmallVector<SDValue, 16> OpsStepConstants;
for (uint64_t i = 0; i < ResVT.getVectorNumElements(); i++)
OpsStepConstants.push_back(
getConstant(StepVal * i, DL, ResVT.getVectorElementType()));
return getBuildVector(ResVT, DL, OpsStepConstants);
}
/// Swaps the values of N1 and N2. Swaps all indices in the shuffle mask M that
/// point at N1 to point at N2 and indices that point at N2 to point at N1.
static void commuteShuffle(SDValue &N1, SDValue &N2, MutableArrayRef<int> M) {
std::swap(N1, N2);
ShuffleVectorSDNode::commuteMask(M);
}
SDValue SelectionDAG::getVectorShuffle(EVT VT, const SDLoc &dl, SDValue N1,
SDValue N2, ArrayRef<int> Mask) {
assert(VT.getVectorNumElements() == Mask.size() &&
"Must have the same number of vector elements as mask elements!");
assert(VT == N1.getValueType() && VT == N2.getValueType() &&
"Invalid VECTOR_SHUFFLE");
// Canonicalize shuffle undef, undef -> undef
if (N1.isUndef() && N2.isUndef())
return getUNDEF(VT);
// Validate that all indices in Mask are within the range of the elements
// input to the shuffle.
int NElts = Mask.size();
assert(llvm::all_of(Mask,
[&](int M) { return M < (NElts * 2) && M >= -1; }) &&
"Index out of range");
// Copy the mask so we can do any needed cleanup.
SmallVector<int, 8> MaskVec(Mask);
// Canonicalize shuffle v, v -> v, undef
if (N1 == N2) {
N2 = getUNDEF(VT);
for (int i = 0; i != NElts; ++i)
if (MaskVec[i] >= NElts) MaskVec[i] -= NElts;
}
// Canonicalize shuffle undef, v -> v, undef. Commute the shuffle mask.
if (N1.isUndef())
commuteShuffle(N1, N2, MaskVec);
if (TLI->hasVectorBlend()) {
// If shuffling a splat, try to blend the splat instead. We do this here so
// that even when this arises during lowering we don't have to re-handle it.
auto BlendSplat = [&](BuildVectorSDNode *BV, int Offset) {
BitVector UndefElements;
SDValue Splat = BV->getSplatValue(&UndefElements);
if (!Splat)
return;
for (int i = 0; i < NElts; ++i) {
if (MaskVec[i] < Offset || MaskVec[i] >= (Offset + NElts))
continue;
// If this input comes from undef, mark it as such.
if (UndefElements[MaskVec[i] - Offset]) {
MaskVec[i] = -1;
continue;
}
// If we can blend a non-undef lane, use that instead.
if (!UndefElements[i])
MaskVec[i] = i + Offset;
}
};
if (auto *N1BV = dyn_cast<BuildVectorSDNode>(N1))
BlendSplat(N1BV, 0);
if (auto *N2BV = dyn_cast<BuildVectorSDNode>(N2))
BlendSplat(N2BV, NElts);
}
// Canonicalize all index into lhs, -> shuffle lhs, undef
// Canonicalize all index into rhs, -> shuffle rhs, undef
bool AllLHS = true, AllRHS = true;
bool N2Undef = N2.isUndef();
for (int i = 0; i != NElts; ++i) {
if (MaskVec[i] >= NElts) {
if (N2Undef)
MaskVec[i] = -1;
else
AllLHS = false;
} else if (MaskVec[i] >= 0) {
AllRHS = false;
}
}
if (AllLHS && AllRHS)
return getUNDEF(VT);
if (AllLHS && !N2Undef)
N2 = getUNDEF(VT);
if (AllRHS) {
N1 = getUNDEF(VT);
commuteShuffle(N1, N2, MaskVec);
}
// Reset our undef status after accounting for the mask.
N2Undef = N2.isUndef();
// Re-check whether both sides ended up undef.
if (N1.isUndef() && N2Undef)
return getUNDEF(VT);
// If Identity shuffle return that node.
bool Identity = true, AllSame = true;
for (int i = 0; i != NElts; ++i) {
if (MaskVec[i] >= 0 && MaskVec[i] != i) Identity = false;
if (MaskVec[i] != MaskVec[0]) AllSame = false;
}
if (Identity && NElts)
return N1;
// Shuffling a constant splat doesn't change the result.
if (N2Undef) {
SDValue V = N1;
// Look through any bitcasts. We check that these don't change the number
// (and size) of elements and just changes their types.
while (V.getOpcode() == ISD::BITCAST)
V = V->getOperand(0);
// A splat should always show up as a build vector node.
if (auto *BV = dyn_cast<BuildVectorSDNode>(V)) {
BitVector UndefElements;
SDValue Splat = BV->getSplatValue(&UndefElements);
// If this is a splat of an undef, shuffling it is also undef.
if (Splat && Splat.isUndef())
return getUNDEF(VT);
bool SameNumElts =
V.getValueType().getVectorNumElements() == VT.getVectorNumElements();
// We only have a splat which can skip shuffles if there is a splatted
// value and no undef lanes rearranged by the shuffle.
if (Splat && UndefElements.none()) {
// Splat of <x, x, ..., x>, return <x, x, ..., x>, provided that the
// number of elements match or the value splatted is a zero constant.
if (SameNumElts || isNullConstant(Splat))
return N1;
}
// If the shuffle itself creates a splat, build the vector directly.
if (AllSame && SameNumElts) {
EVT BuildVT = BV->getValueType(0);
const SDValue &Splatted = BV->getOperand(MaskVec[0]);
SDValue NewBV = getSplatBuildVector(BuildVT, dl, Splatted);
// We may have jumped through bitcasts, so the type of the
// BUILD_VECTOR may not match the type of the shuffle.
if (BuildVT != VT)
NewBV = getNode(ISD::BITCAST, dl, VT, NewBV);
return NewBV;
}
}
}
FoldingSetNodeID ID;
SDValue Ops[2] = { N1, N2 };
AddNodeIDNode(ID, ISD::VECTOR_SHUFFLE, getVTList(VT), Ops);
for (int i = 0; i != NElts; ++i)
ID.AddInteger(MaskVec[i]);
void* IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, dl, IP))
return SDValue(E, 0);
// Allocate the mask array for the node out of the BumpPtrAllocator, since
// SDNode doesn't have access to it. This memory will be "leaked" when
// the node is deallocated, but recovered when the NodeAllocator is released.
int *MaskAlloc = OperandAllocator.Allocate<int>(NElts);
llvm::copy(MaskVec, MaskAlloc);
auto *N = newSDNode<ShuffleVectorSDNode>(VT, dl.getIROrder(),
dl.getDebugLoc(), MaskAlloc);
createOperands(N, Ops);
CSEMap.InsertNode(N, IP);
InsertNode(N);
SDValue V = SDValue(N, 0);
NewSDValueDbgMsg(V, "Creating new node: ", this);
return V;
}
SDValue SelectionDAG::getCommutedVectorShuffle(const ShuffleVectorSDNode &SV) {
EVT VT = SV.getValueType(0);
SmallVector<int, 8> MaskVec(SV.getMask());
ShuffleVectorSDNode::commuteMask(MaskVec);
SDValue Op0 = SV.getOperand(0);
SDValue Op1 = SV.getOperand(1);
return getVectorShuffle(VT, SDLoc(&SV), Op1, Op0, MaskVec);
}
SDValue SelectionDAG::getRegister(unsigned RegNo, EVT VT) {
FoldingSetNodeID ID;
AddNodeIDNode(ID, ISD::Register, getVTList(VT), std::nullopt);
ID.AddInteger(RegNo);
void *IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, IP))
return SDValue(E, 0);
auto *N = newSDNode<RegisterSDNode>(RegNo, VT);
N->SDNodeBits.IsDivergent = TLI->isSDNodeSourceOfDivergence(N, FLI, UA);
CSEMap.InsertNode(N, IP);
InsertNode(N);
return SDValue(N, 0);
}
SDValue SelectionDAG::getRegisterMask(const uint32_t *RegMask) {
FoldingSetNodeID ID;
AddNodeIDNode(ID, ISD::RegisterMask, getVTList(MVT::Untyped), std::nullopt);
ID.AddPointer(RegMask);
void *IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, IP))
return SDValue(E, 0);
auto *N = newSDNode<RegisterMaskSDNode>(RegMask);
CSEMap.InsertNode(N, IP);
InsertNode(N);
return SDValue(N, 0);
}
SDValue SelectionDAG::getEHLabel(const SDLoc &dl, SDValue Root,
MCSymbol *Label) {
return getLabelNode(ISD::EH_LABEL, dl, Root, Label);
}
SDValue SelectionDAG::getLabelNode(unsigned Opcode, const SDLoc &dl,
SDValue Root, MCSymbol *Label) {
FoldingSetNodeID ID;
SDValue Ops[] = { Root };
AddNodeIDNode(ID, Opcode, getVTList(MVT::Other), Ops);
ID.AddPointer(Label);
void *IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, IP))
return SDValue(E, 0);
auto *N =
newSDNode<LabelSDNode>(Opcode, dl.getIROrder(), dl.getDebugLoc(), Label);
createOperands(N, Ops);
CSEMap.InsertNode(N, IP);
InsertNode(N);
return SDValue(N, 0);
}
SDValue SelectionDAG::getBlockAddress(const BlockAddress *BA, EVT VT,
int64_t Offset, bool isTarget,
unsigned TargetFlags) {
unsigned Opc = isTarget ? ISD::TargetBlockAddress : ISD::BlockAddress;
FoldingSetNodeID ID;
AddNodeIDNode(ID, Opc, getVTList(VT), std::nullopt);
ID.AddPointer(BA);
ID.AddInteger(Offset);
ID.AddInteger(TargetFlags);
void *IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, IP))
return SDValue(E, 0);
auto *N = newSDNode<BlockAddressSDNode>(Opc, VT, BA, Offset, TargetFlags);
CSEMap.InsertNode(N, IP);
InsertNode(N);
return SDValue(N, 0);
}
SDValue SelectionDAG::getSrcValue(const Value *V) {
FoldingSetNodeID ID;
AddNodeIDNode(ID, ISD::SRCVALUE, getVTList(MVT::Other), std::nullopt);
ID.AddPointer(V);
void *IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, IP))
return SDValue(E, 0);
auto *N = newSDNode<SrcValueSDNode>(V);
CSEMap.InsertNode(N, IP);
InsertNode(N);
return SDValue(N, 0);
}
SDValue SelectionDAG::getMDNode(const MDNode *MD) {
FoldingSetNodeID ID;
AddNodeIDNode(ID, ISD::MDNODE_SDNODE, getVTList(MVT::Other), std::nullopt);
ID.AddPointer(MD);
void *IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, IP))
return SDValue(E, 0);
auto *N = newSDNode<MDNodeSDNode>(MD);
CSEMap.InsertNode(N, IP);
InsertNode(N);
return SDValue(N, 0);
}
SDValue SelectionDAG::getBitcast(EVT VT, SDValue V) {
if (VT == V.getValueType())
return V;
return getNode(ISD::BITCAST, SDLoc(V), VT, V);
}
SDValue SelectionDAG::getAddrSpaceCast(const SDLoc &dl, EVT VT, SDValue Ptr,
unsigned SrcAS, unsigned DestAS) {
SDValue Ops[] = {Ptr};
FoldingSetNodeID ID;
AddNodeIDNode(ID, ISD::ADDRSPACECAST, getVTList(VT), Ops);
ID.AddInteger(SrcAS);
ID.AddInteger(DestAS);
void *IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, dl, IP))
return SDValue(E, 0);
auto *N = newSDNode<AddrSpaceCastSDNode>(dl.getIROrder(), dl.getDebugLoc(),
VT, SrcAS, DestAS);
createOperands(N, Ops);
CSEMap.InsertNode(N, IP);
InsertNode(N);
return SDValue(N, 0);
}
SDValue SelectionDAG::getFreeze(SDValue V) {
return getNode(ISD::FREEZE, SDLoc(V), V.getValueType(), V);
}
/// getShiftAmountOperand - Return the specified value casted to
/// the target's desired shift amount type.
SDValue SelectionDAG::getShiftAmountOperand(EVT LHSTy, SDValue Op) {
EVT OpTy = Op.getValueType();
EVT ShTy = TLI->getShiftAmountTy(LHSTy, getDataLayout());
if (OpTy == ShTy || OpTy.isVector()) return Op;
return getZExtOrTrunc(Op, SDLoc(Op), ShTy);
}
SDValue SelectionDAG::expandVAArg(SDNode *Node) {
SDLoc dl(Node);
const TargetLowering &TLI = getTargetLoweringInfo();
const Value *V = cast<SrcValueSDNode>(Node->getOperand(2))->getValue();
EVT VT = Node->getValueType(0);
SDValue Tmp1 = Node->getOperand(0);
SDValue Tmp2 = Node->getOperand(1);
const MaybeAlign MA(Node->getConstantOperandVal(3));
SDValue VAListLoad = getLoad(TLI.getPointerTy(getDataLayout()), dl, Tmp1,
Tmp2, MachinePointerInfo(V));
SDValue VAList = VAListLoad;
if (MA && *MA > TLI.getMinStackArgumentAlignment()) {
VAList = getNode(ISD::ADD, dl, VAList.getValueType(), VAList,
getConstant(MA->value() - 1, dl, VAList.getValueType()));
VAList =
getNode(ISD::AND, dl, VAList.getValueType(), VAList,
getConstant(-(int64_t)MA->value(), dl, VAList.getValueType()));
}
// Increment the pointer, VAList, to the next vaarg
Tmp1 = getNode(ISD::ADD, dl, VAList.getValueType(), VAList,
getConstant(getDataLayout().getTypeAllocSize(
VT.getTypeForEVT(*getContext())),
dl, VAList.getValueType()));
// Store the incremented VAList to the legalized pointer
Tmp1 =
getStore(VAListLoad.getValue(1), dl, Tmp1, Tmp2, MachinePointerInfo(V));
// Load the actual argument out of the pointer VAList
return getLoad(VT, dl, Tmp1, VAList, MachinePointerInfo());
}
SDValue SelectionDAG::expandVACopy(SDNode *Node) {
SDLoc dl(Node);
const TargetLowering &TLI = getTargetLoweringInfo();
// This defaults to loading a pointer from the input and storing it to the
// output, returning the chain.
const Value *VD = cast<SrcValueSDNode>(Node->getOperand(3))->getValue();
const Value *VS = cast<SrcValueSDNode>(Node->getOperand(4))->getValue();
SDValue Tmp1 =
getLoad(TLI.getPointerTy(getDataLayout()), dl, Node->getOperand(0),
Node->getOperand(2), MachinePointerInfo(VS));
return getStore(Tmp1.getValue(1), dl, Tmp1, Node->getOperand(1),
MachinePointerInfo(VD));
}
Align SelectionDAG::getReducedAlign(EVT VT, bool UseABI) {
const DataLayout &DL = getDataLayout();
Type *Ty = VT.getTypeForEVT(*getContext());
Align RedAlign = UseABI ? DL.getABITypeAlign(Ty) : DL.getPrefTypeAlign(Ty);
if (TLI->isTypeLegal(VT) || !VT.isVector())
return RedAlign;
const TargetFrameLowering *TFI = MF->getSubtarget().getFrameLowering();
const Align StackAlign = TFI->getStackAlign();
// See if we can choose a smaller ABI alignment in cases where it's an
// illegal vector type that will get broken down.
if (RedAlign > StackAlign) {
EVT IntermediateVT;
MVT RegisterVT;
unsigned NumIntermediates;
TLI->getVectorTypeBreakdown(*getContext(), VT, IntermediateVT,
NumIntermediates, RegisterVT);
Ty = IntermediateVT.getTypeForEVT(*getContext());
Align RedAlign2 = UseABI ? DL.getABITypeAlign(Ty) : DL.getPrefTypeAlign(Ty);
if (RedAlign2 < RedAlign)
RedAlign = RedAlign2;
}
return RedAlign;
}
SDValue SelectionDAG::CreateStackTemporary(TypeSize Bytes, Align Alignment) {
MachineFrameInfo &MFI = MF->getFrameInfo();
const TargetFrameLowering *TFI = MF->getSubtarget().getFrameLowering();
int StackID = 0;
if (Bytes.isScalable())
StackID = TFI->getStackIDForScalableVectors();
// The stack id gives an indication of whether the object is scalable or
// not, so it's safe to pass in the minimum size here.
int FrameIdx = MFI.CreateStackObject(Bytes.getKnownMinValue(), Alignment,
false, nullptr, StackID);
return getFrameIndex(FrameIdx, TLI->getFrameIndexTy(getDataLayout()));
}
SDValue SelectionDAG::CreateStackTemporary(EVT VT, unsigned minAlign) {
Type *Ty = VT.getTypeForEVT(*getContext());
Align StackAlign =
std::max(getDataLayout().getPrefTypeAlign(Ty), Align(minAlign));
return CreateStackTemporary(VT.getStoreSize(), StackAlign);
}
SDValue SelectionDAG::CreateStackTemporary(EVT VT1, EVT VT2) {
TypeSize VT1Size = VT1.getStoreSize();
TypeSize VT2Size = VT2.getStoreSize();
assert(VT1Size.isScalable() == VT2Size.isScalable() &&
"Don't know how to choose the maximum size when creating a stack "
"temporary");
TypeSize Bytes = VT1Size.getKnownMinValue() > VT2Size.getKnownMinValue()
? VT1Size
: VT2Size;
Type *Ty1 = VT1.getTypeForEVT(*getContext());
Type *Ty2 = VT2.getTypeForEVT(*getContext());
const DataLayout &DL = getDataLayout();
Align Align = std::max(DL.getPrefTypeAlign(Ty1), DL.getPrefTypeAlign(Ty2));
return CreateStackTemporary(Bytes, Align);
}
SDValue SelectionDAG::FoldSetCC(EVT VT, SDValue N1, SDValue N2,
ISD::CondCode Cond, const SDLoc &dl) {
EVT OpVT = N1.getValueType();
auto GetUndefBooleanConstant = [&]() {
if (VT.getScalarType() == MVT::i1 ||
TLI->getBooleanContents(OpVT) ==
TargetLowering::UndefinedBooleanContent)
return getUNDEF(VT);
// ZeroOrOne / ZeroOrNegative require specific values for the high bits,
// so we cannot use getUNDEF(). Return zero instead.
return getConstant(0, dl, VT);
};
// These setcc operations always fold.
switch (Cond) {
default: break;
case ISD::SETFALSE:
case ISD::SETFALSE2: return getBoolConstant(false, dl, VT, OpVT);
case ISD::SETTRUE:
case ISD::SETTRUE2: return getBoolConstant(true, dl, VT, OpVT);
case ISD::SETOEQ:
case ISD::SETOGT:
case ISD::SETOGE:
case ISD::SETOLT:
case ISD::SETOLE:
case ISD::SETONE:
case ISD::SETO:
case ISD::SETUO:
case ISD::SETUEQ:
case ISD::SETUNE:
assert(!OpVT.isInteger() && "Illegal setcc for integer!");
break;
}
if (OpVT.isInteger()) {
// For EQ and NE, we can always pick a value for the undef to make the
// predicate pass or fail, so we can return undef.
// Matches behavior in llvm::ConstantFoldCompareInstruction.
// icmp eq/ne X, undef -> undef.
if ((N1.isUndef() || N2.isUndef()) &&
(Cond == ISD::SETEQ || Cond == ISD::SETNE))
return GetUndefBooleanConstant();
// If both operands are undef, we can return undef for int comparison.
// icmp undef, undef -> undef.
if (N1.isUndef() && N2.isUndef())
return GetUndefBooleanConstant();
// icmp X, X -> true/false
// icmp X, undef -> true/false because undef could be X.
if (N1.isUndef() || N2.isUndef() || N1 == N2)
return getBoolConstant(ISD::isTrueWhenEqual(Cond), dl, VT, OpVT);
}
if (ConstantSDNode *N2C = dyn_cast<ConstantSDNode>(N2)) {
const APInt &C2 = N2C->getAPIntValue();
if (ConstantSDNode *N1C = dyn_cast<ConstantSDNode>(N1)) {
const APInt &C1 = N1C->getAPIntValue();
return getBoolConstant(ICmpInst::compare(C1, C2, getICmpCondCode(Cond)),
dl, VT, OpVT);
}
}
auto *N1CFP = dyn_cast<ConstantFPSDNode>(N1);
auto *N2CFP = dyn_cast<ConstantFPSDNode>(N2);
if (N1CFP && N2CFP) {
APFloat::cmpResult R = N1CFP->getValueAPF().compare(N2CFP->getValueAPF());
switch (Cond) {
default: break;
case ISD::SETEQ: if (R==APFloat::cmpUnordered)
return GetUndefBooleanConstant();
[[fallthrough]];
case ISD::SETOEQ: return getBoolConstant(R==APFloat::cmpEqual, dl, VT,
OpVT);
case ISD::SETNE: if (R==APFloat::cmpUnordered)
return GetUndefBooleanConstant();
[[fallthrough]];
case ISD::SETONE: return getBoolConstant(R==APFloat::cmpGreaterThan ||
R==APFloat::cmpLessThan, dl, VT,
OpVT);
case ISD::SETLT: if (R==APFloat::cmpUnordered)
return GetUndefBooleanConstant();
[[fallthrough]];
case ISD::SETOLT: return getBoolConstant(R==APFloat::cmpLessThan, dl, VT,
OpVT);
case ISD::SETGT: if (R==APFloat::cmpUnordered)
return GetUndefBooleanConstant();
[[fallthrough]];
case ISD::SETOGT: return getBoolConstant(R==APFloat::cmpGreaterThan, dl,
VT, OpVT);
case ISD::SETLE: if (R==APFloat::cmpUnordered)
return GetUndefBooleanConstant();
[[fallthrough]];
case ISD::SETOLE: return getBoolConstant(R==APFloat::cmpLessThan ||
R==APFloat::cmpEqual, dl, VT,
OpVT);
case ISD::SETGE: if (R==APFloat::cmpUnordered)
return GetUndefBooleanConstant();
[[fallthrough]];
case ISD::SETOGE: return getBoolConstant(R==APFloat::cmpGreaterThan ||
R==APFloat::cmpEqual, dl, VT, OpVT);
case ISD::SETO: return getBoolConstant(R!=APFloat::cmpUnordered, dl, VT,
OpVT);
case ISD::SETUO: return getBoolConstant(R==APFloat::cmpUnordered, dl, VT,
OpVT);
case ISD::SETUEQ: return getBoolConstant(R==APFloat::cmpUnordered ||
R==APFloat::cmpEqual, dl, VT,
OpVT);
case ISD::SETUNE: return getBoolConstant(R!=APFloat::cmpEqual, dl, VT,
OpVT);
case ISD::SETULT: return getBoolConstant(R==APFloat::cmpUnordered ||
R==APFloat::cmpLessThan, dl, VT,
OpVT);
case ISD::SETUGT: return getBoolConstant(R==APFloat::cmpGreaterThan ||
R==APFloat::cmpUnordered, dl, VT,
OpVT);
case ISD::SETULE: return getBoolConstant(R!=APFloat::cmpGreaterThan, dl,
VT, OpVT);
case ISD::SETUGE: return getBoolConstant(R!=APFloat::cmpLessThan, dl, VT,
OpVT);
}
} else if (N1CFP && OpVT.isSimple() && !N2.isUndef()) {
// Ensure that the constant occurs on the RHS.
ISD::CondCode SwappedCond = ISD::getSetCCSwappedOperands(Cond);
if (!TLI->isCondCodeLegal(SwappedCond, OpVT.getSimpleVT()))
return SDValue();
return getSetCC(dl, VT, N2, N1, SwappedCond);
} else if ((N2CFP && N2CFP->getValueAPF().isNaN()) ||
(OpVT.isFloatingPoint() && (N1.isUndef() || N2.isUndef()))) {
// If an operand is known to be a nan (or undef that could be a nan), we can
// fold it.
// Choosing NaN for the undef will always make unordered comparison succeed
// and ordered comparison fails.
// Matches behavior in llvm::ConstantFoldCompareInstruction.
switch (ISD::getUnorderedFlavor(Cond)) {
default:
llvm_unreachable("Unknown flavor!");
case 0: // Known false.
return getBoolConstant(false, dl, VT, OpVT);
case 1: // Known true.
return getBoolConstant(true, dl, VT, OpVT);
case 2: // Undefined.
return GetUndefBooleanConstant();
}
}
// Could not fold it.
return SDValue();
}
/// SignBitIsZero - Return true if the sign bit of Op is known to be zero. We
/// use this predicate to simplify operations downstream.
bool SelectionDAG::SignBitIsZero(SDValue Op, unsigned Depth) const {
unsigned BitWidth = Op.getScalarValueSizeInBits();
return MaskedValueIsZero(Op, APInt::getSignMask(BitWidth), Depth);
}
/// MaskedValueIsZero - Return true if 'V & Mask' is known to be zero. We use
/// this predicate to simplify operations downstream. Mask is known to be zero
/// for bits that V cannot have.
bool SelectionDAG::MaskedValueIsZero(SDValue V, const APInt &Mask,
unsigned Depth) const {
return Mask.isSubsetOf(computeKnownBits(V, Depth).Zero);
}
/// MaskedValueIsZero - Return true if 'V & Mask' is known to be zero in
/// DemandedElts. We use this predicate to simplify operations downstream.
/// Mask is known to be zero for bits that V cannot have.
bool SelectionDAG::MaskedValueIsZero(SDValue V, const APInt &Mask,
const APInt &DemandedElts,
unsigned Depth) const {
return Mask.isSubsetOf(computeKnownBits(V, DemandedElts, Depth).Zero);
}
/// MaskedVectorIsZero - Return true if 'Op' is known to be zero in
/// DemandedElts. We use this predicate to simplify operations downstream.
bool SelectionDAG::MaskedVectorIsZero(SDValue V, const APInt &DemandedElts,
unsigned Depth /* = 0 */) const {
return computeKnownBits(V, DemandedElts, Depth).isZero();
}
/// MaskedValueIsAllOnes - Return true if '(Op & Mask) == Mask'.
bool SelectionDAG::MaskedValueIsAllOnes(SDValue V, const APInt &Mask,
unsigned Depth) const {
return Mask.isSubsetOf(computeKnownBits(V, Depth).One);
}
APInt SelectionDAG::computeVectorKnownZeroElements(SDValue Op,
const APInt &DemandedElts,
unsigned Depth) const {
EVT VT = Op.getValueType();
assert(VT.isVector() && !VT.isScalableVector() && "Only for fixed vectors!");
unsigned NumElts = VT.getVectorNumElements();
assert(DemandedElts.getBitWidth() == NumElts && "Unexpected demanded mask.");
APInt KnownZeroElements = APInt::getZero(NumElts);
for (unsigned EltIdx = 0; EltIdx != NumElts; ++EltIdx) {
if (!DemandedElts[EltIdx])
continue; // Don't query elements that are not demanded.
APInt Mask = APInt::getOneBitSet(NumElts, EltIdx);
if (MaskedVectorIsZero(Op, Mask, Depth))
KnownZeroElements.setBit(EltIdx);
}
return KnownZeroElements;
}
/// isSplatValue - Return true if the vector V has the same value
/// across all DemandedElts. For scalable vectors, we don't know the
/// number of lanes at compile time. Instead, we use a 1 bit APInt
/// to represent a conservative value for all lanes; that is, that
/// one bit value is implicitly splatted across all lanes.
bool SelectionDAG::isSplatValue(SDValue V, const APInt &DemandedElts,
APInt &UndefElts, unsigned Depth) const {
unsigned Opcode = V.getOpcode();
EVT VT = V.getValueType();
assert(VT.isVector() && "Vector type expected");
assert((!VT.isScalableVector() || DemandedElts.getBitWidth() == 1) &&
"scalable demanded bits are ignored");
if (!DemandedElts)
return false; // No demanded elts, better to assume we don't know anything.
if (Depth >= MaxRecursionDepth)
return false; // Limit search depth.
// Deal with some common cases here that work for both fixed and scalable
// vector types.
switch (Opcode) {
case ISD::SPLAT_VECTOR:
UndefElts = V.getOperand(0).isUndef()
? APInt::getAllOnes(DemandedElts.getBitWidth())
: APInt(DemandedElts.getBitWidth(), 0);
return true;
case ISD::ADD:
case ISD::SUB:
case ISD::AND:
case ISD::XOR:
case ISD::OR: {
APInt UndefLHS, UndefRHS;
SDValue LHS = V.getOperand(0);
SDValue RHS = V.getOperand(1);
if (isSplatValue(LHS, DemandedElts, UndefLHS, Depth + 1) &&
isSplatValue(RHS, DemandedElts, UndefRHS, Depth + 1)) {
UndefElts = UndefLHS | UndefRHS;
return true;
}
return false;
}
case ISD::ABS:
case ISD::TRUNCATE:
case ISD::SIGN_EXTEND:
case ISD::ZERO_EXTEND:
return isSplatValue(V.getOperand(0), DemandedElts, UndefElts, Depth + 1);
default:
if (Opcode >= ISD::BUILTIN_OP_END || Opcode == ISD::INTRINSIC_WO_CHAIN ||
Opcode == ISD::INTRINSIC_W_CHAIN || Opcode == ISD::INTRINSIC_VOID)
return TLI->isSplatValueForTargetNode(V, DemandedElts, UndefElts, *this,
Depth);
break;
}
// We don't support other cases than those above for scalable vectors at
// the moment.
if (VT.isScalableVector())
return false;
unsigned NumElts = VT.getVectorNumElements();
assert(NumElts == DemandedElts.getBitWidth() && "Vector size mismatch");
UndefElts = APInt::getZero(NumElts);
switch (Opcode) {
case ISD::BUILD_VECTOR: {
SDValue Scl;
for (unsigned i = 0; i != NumElts; ++i) {
SDValue Op = V.getOperand(i);
if (Op.isUndef()) {
UndefElts.setBit(i);
continue;
}
if (!DemandedElts[i])
continue;
if (Scl && Scl != Op)
return false;
Scl = Op;
}
return true;
}
case ISD::VECTOR_SHUFFLE: {
// Check if this is a shuffle node doing a splat or a shuffle of a splat.
APInt DemandedLHS = APInt::getZero(NumElts);
APInt DemandedRHS = APInt::getZero(NumElts);
ArrayRef<int> Mask = cast<ShuffleVectorSDNode>(V)->getMask();
for (int i = 0; i != (int)NumElts; ++i) {
int M = Mask[i];
if (M < 0) {
UndefElts.setBit(i);
continue;
}
if (!DemandedElts[i])
continue;
if (M < (int)NumElts)
DemandedLHS.setBit(M);
else
DemandedRHS.setBit(M - NumElts);
}
// If we aren't demanding either op, assume there's no splat.
// If we are demanding both ops, assume there's no splat.
if ((DemandedLHS.isZero() && DemandedRHS.isZero()) ||
(!DemandedLHS.isZero() && !DemandedRHS.isZero()))
return false;
// See if the demanded elts of the source op is a splat or we only demand
// one element, which should always be a splat.
// TODO: Handle source ops splats with undefs.
auto CheckSplatSrc = [&](SDValue Src, const APInt &SrcElts) {
APInt SrcUndefs;
return (SrcElts.popcount() == 1) ||
(isSplatValue(Src, SrcElts, SrcUndefs, Depth + 1) &&
(SrcElts & SrcUndefs).isZero());
};
if (!DemandedLHS.isZero())
return CheckSplatSrc(V.getOperand(0), DemandedLHS);
return CheckSplatSrc(V.getOperand(1), DemandedRHS);
}
case ISD::EXTRACT_SUBVECTOR: {
// Offset the demanded elts by the subvector index.
SDValue Src = V.getOperand(0);
// We don't support scalable vectors at the moment.
if (Src.getValueType().isScalableVector())
return false;
uint64_t Idx = V.getConstantOperandVal(1);
unsigned NumSrcElts = Src.getValueType().getVectorNumElements();
APInt UndefSrcElts;
APInt DemandedSrcElts = DemandedElts.zext(NumSrcElts).shl(Idx);
if (isSplatValue(Src, DemandedSrcElts, UndefSrcElts, Depth + 1)) {
UndefElts = UndefSrcElts.extractBits(NumElts, Idx);
return true;
}
break;
}
case ISD::ANY_EXTEND_VECTOR_INREG:
case ISD::SIGN_EXTEND_VECTOR_INREG:
case ISD::ZERO_EXTEND_VECTOR_INREG: {
// Widen the demanded elts by the src element count.
SDValue Src = V.getOperand(0);
// We don't support scalable vectors at the moment.
if (Src.getValueType().isScalableVector())
return false;
unsigned NumSrcElts = Src.getValueType().getVectorNumElements();
APInt UndefSrcElts;
APInt DemandedSrcElts = DemandedElts.zext(NumSrcElts);
if (isSplatValue(Src, DemandedSrcElts, UndefSrcElts, Depth + 1)) {
UndefElts = UndefSrcElts.trunc(NumElts);
return true;
}
break;
}
case ISD::BITCAST: {
SDValue Src = V.getOperand(0);
EVT SrcVT = Src.getValueType();
unsigned SrcBitWidth = SrcVT.getScalarSizeInBits();
unsigned BitWidth = VT.getScalarSizeInBits();
// Ignore bitcasts from unsupported types.
// TODO: Add fp support?
if (!SrcVT.isVector() || !SrcVT.isInteger() || !VT.isInteger())
break;
// Bitcast 'small element' vector to 'large element' vector.
if ((BitWidth % SrcBitWidth) == 0) {
// See if each sub element is a splat.
unsigned Scale = BitWidth / SrcBitWidth;
unsigned NumSrcElts = SrcVT.getVectorNumElements();
APInt ScaledDemandedElts =
APIntOps::ScaleBitMask(DemandedElts, NumSrcElts);
for (unsigned I = 0; I != Scale; ++I) {
APInt SubUndefElts;
APInt SubDemandedElt = APInt::getOneBitSet(Scale, I);
APInt SubDemandedElts = APInt::getSplat(NumSrcElts, SubDemandedElt);
SubDemandedElts &= ScaledDemandedElts;
if (!isSplatValue(Src, SubDemandedElts, SubUndefElts, Depth + 1))
return false;
// TODO: Add support for merging sub undef elements.
if (!SubUndefElts.isZero())
return false;
}
return true;
}
break;
}
}
return false;
}
/// Helper wrapper to main isSplatValue function.
bool SelectionDAG::isSplatValue(SDValue V, bool AllowUndefs) const {
EVT VT = V.getValueType();
assert(VT.isVector() && "Vector type expected");
APInt UndefElts;
// Since the number of lanes in a scalable vector is unknown at compile time,
// we track one bit which is implicitly broadcast to all lanes. This means
// that all lanes in a scalable vector are considered demanded.
APInt DemandedElts
= APInt::getAllOnes(VT.isScalableVector() ? 1 : VT.getVectorNumElements());
return isSplatValue(V, DemandedElts, UndefElts) &&
(AllowUndefs || !UndefElts);
}
SDValue SelectionDAG::getSplatSourceVector(SDValue V, int &SplatIdx) {
V = peekThroughExtractSubvectors(V);
EVT VT = V.getValueType();
unsigned Opcode = V.getOpcode();
switch (Opcode) {
default: {
APInt UndefElts;
// Since the number of lanes in a scalable vector is unknown at compile time,
// we track one bit which is implicitly broadcast to all lanes. This means
// that all lanes in a scalable vector are considered demanded.
APInt DemandedElts
= APInt::getAllOnes(VT.isScalableVector() ? 1 : VT.getVectorNumElements());
if (isSplatValue(V, DemandedElts, UndefElts)) {
if (VT.isScalableVector()) {
// DemandedElts and UndefElts are ignored for scalable vectors, since
// the only supported cases are SPLAT_VECTOR nodes.
SplatIdx = 0;
} else {
// Handle case where all demanded elements are UNDEF.
if (DemandedElts.isSubsetOf(UndefElts)) {
SplatIdx = 0;
return getUNDEF(VT);
}
SplatIdx = (UndefElts & DemandedElts).countr_one();
}
return V;
}
break;
}
case ISD::SPLAT_VECTOR:
SplatIdx = 0;
return V;
case ISD::VECTOR_SHUFFLE: {
assert(!VT.isScalableVector());
// Check if this is a shuffle node doing a splat.
// TODO - remove this and rely purely on SelectionDAG::isSplatValue,
// getTargetVShiftNode currently struggles without the splat source.
auto *SVN = cast<ShuffleVectorSDNode>(V);
if (!SVN->isSplat())
break;
int Idx = SVN->getSplatIndex();
int NumElts = V.getValueType().getVectorNumElements();
SplatIdx = Idx % NumElts;
return V.getOperand(Idx / NumElts);
}
}
return SDValue();
}
SDValue SelectionDAG::getSplatValue(SDValue V, bool LegalTypes) {
int SplatIdx;
if (SDValue SrcVector = getSplatSourceVector(V, SplatIdx)) {
EVT SVT = SrcVector.getValueType().getScalarType();
EVT LegalSVT = SVT;
if (LegalTypes && !TLI->isTypeLegal(SVT)) {
if (!SVT.isInteger())
return SDValue();
LegalSVT = TLI->getTypeToTransformTo(*getContext(), LegalSVT);
if (LegalSVT.bitsLT(SVT))
return SDValue();
}
return getNode(ISD::EXTRACT_VECTOR_ELT, SDLoc(V), LegalSVT, SrcVector,
getVectorIdxConstant(SplatIdx, SDLoc(V)));
}
return SDValue();
}
const APInt *
SelectionDAG::getValidShiftAmountConstant(SDValue V,
const APInt &DemandedElts) const {
assert((V.getOpcode() == ISD::SHL || V.getOpcode() == ISD::SRL ||
V.getOpcode() == ISD::SRA) &&
"Unknown shift node");
unsigned BitWidth = V.getScalarValueSizeInBits();
if (ConstantSDNode *SA = isConstOrConstSplat(V.getOperand(1), DemandedElts)) {
// Shifting more than the bitwidth is not valid.
const APInt &ShAmt = SA->getAPIntValue();
if (ShAmt.ult(BitWidth))
return &ShAmt;
}
return nullptr;
}
const APInt *SelectionDAG::getValidMinimumShiftAmountConstant(
SDValue V, const APInt &DemandedElts) const {
assert((V.getOpcode() == ISD::SHL || V.getOpcode() == ISD::SRL ||
V.getOpcode() == ISD::SRA) &&
"Unknown shift node");
if (const APInt *ValidAmt = getValidShiftAmountConstant(V, DemandedElts))
return ValidAmt;
unsigned BitWidth = V.getScalarValueSizeInBits();
auto *BV = dyn_cast<BuildVectorSDNode>(V.getOperand(1));
if (!BV)
return nullptr;
const APInt *MinShAmt = nullptr;
for (unsigned i = 0, e = BV->getNumOperands(); i != e; ++i) {
if (!DemandedElts[i])
continue;
auto *SA = dyn_cast<ConstantSDNode>(BV->getOperand(i));
if (!SA)
return nullptr;
// Shifting more than the bitwidth is not valid.
const APInt &ShAmt = SA->getAPIntValue();
if (ShAmt.uge(BitWidth))
return nullptr;
if (MinShAmt && MinShAmt->ule(ShAmt))
continue;
MinShAmt = &ShAmt;
}
return MinShAmt;
}
const APInt *SelectionDAG::getValidMaximumShiftAmountConstant(
SDValue V, const APInt &DemandedElts) const {
assert((V.getOpcode() == ISD::SHL || V.getOpcode() == ISD::SRL ||
V.getOpcode() == ISD::SRA) &&
"Unknown shift node");
if (const APInt *ValidAmt = getValidShiftAmountConstant(V, DemandedElts))
return ValidAmt;
unsigned BitWidth = V.getScalarValueSizeInBits();
auto *BV = dyn_cast<BuildVectorSDNode>(V.getOperand(1));
if (!BV)
return nullptr;
const APInt *MaxShAmt = nullptr;
for (unsigned i = 0, e = BV->getNumOperands(); i != e; ++i) {
if (!DemandedElts[i])
continue;
auto *SA = dyn_cast<ConstantSDNode>(BV->getOperand(i));
if (!SA)
return nullptr;
// Shifting more than the bitwidth is not valid.
const APInt &ShAmt = SA->getAPIntValue();
if (ShAmt.uge(BitWidth))
return nullptr;
if (MaxShAmt && MaxShAmt->uge(ShAmt))
continue;
MaxShAmt = &ShAmt;
}
return MaxShAmt;
}
/// Determine which bits of Op are known to be either zero or one and return
/// them in Known. For vectors, the known bits are those that are shared by
/// every vector element.
KnownBits SelectionDAG::computeKnownBits(SDValue Op, unsigned Depth) const {
EVT VT = Op.getValueType();
// Since the number of lanes in a scalable vector is unknown at compile time,
// we track one bit which is implicitly broadcast to all lanes. This means
// that all lanes in a scalable vector are considered demanded.
APInt DemandedElts = VT.isFixedLengthVector()
? APInt::getAllOnes(VT.getVectorNumElements())
: APInt(1, 1);
return computeKnownBits(Op, DemandedElts, Depth);
}
/// Determine which bits of Op are known to be either zero or one and return
/// them in Known. The DemandedElts argument allows us to only collect the known
/// bits that are shared by the requested vector elements.
KnownBits SelectionDAG::computeKnownBits(SDValue Op, const APInt &DemandedElts,
unsigned Depth) const {
unsigned BitWidth = Op.getScalarValueSizeInBits();
KnownBits Known(BitWidth); // Don't know anything.
if (auto *C = dyn_cast<ConstantSDNode>(Op)) {
// We know all of the bits for a constant!
return KnownBits::makeConstant(C->getAPIntValue());
}
if (auto *C = dyn_cast<ConstantFPSDNode>(Op)) {
// We know all of the bits for a constant fp!
return KnownBits::makeConstant(C->getValueAPF().bitcastToAPInt());
}
if (Depth >= MaxRecursionDepth)
return Known; // Limit search depth.
KnownBits Known2;
unsigned NumElts = DemandedElts.getBitWidth();
assert((!Op.getValueType().isFixedLengthVector() ||
NumElts == Op.getValueType().getVectorNumElements()) &&
"Unexpected vector size");
if (!DemandedElts)
return Known; // No demanded elts, better to assume we don't know anything.
unsigned Opcode = Op.getOpcode();
switch (Opcode) {
case ISD::MERGE_VALUES:
return computeKnownBits(Op.getOperand(Op.getResNo()), DemandedElts,
Depth + 1);
case ISD::SPLAT_VECTOR: {
SDValue SrcOp = Op.getOperand(0);
assert(SrcOp.getValueSizeInBits() >= BitWidth &&
"Expected SPLAT_VECTOR implicit truncation");
// Implicitly truncate the bits to match the official semantics of
// SPLAT_VECTOR.
Known = computeKnownBits(SrcOp, Depth + 1).trunc(BitWidth);
break;
}
case ISD::SPLAT_VECTOR_PARTS: {
unsigned ScalarSize = Op.getOperand(0).getScalarValueSizeInBits();
assert(ScalarSize * Op.getNumOperands() == BitWidth &&
"Expected SPLAT_VECTOR_PARTS scalars to cover element width");
for (auto [I, SrcOp] : enumerate(Op->ops())) {
Known.insertBits(computeKnownBits(SrcOp, Depth + 1), ScalarSize * I);
}
break;
}
case ISD::STEP_VECTOR: {
const APInt &Step = Op.getConstantOperandAPInt(0);
if (Step.isPowerOf2())
Known.Zero.setLowBits(Step.logBase2());
const Function &F = getMachineFunction().getFunction();
if (!isUIntN(BitWidth, Op.getValueType().getVectorMinNumElements()))
break;
const APInt MinNumElts =
APInt(BitWidth, Op.getValueType().getVectorMinNumElements());
bool Overflow;
const APInt MaxNumElts = getVScaleRange(&F, BitWidth)
.getUnsignedMax()
.umul_ov(MinNumElts, Overflow);
if (Overflow)
break;
const APInt MaxValue = (MaxNumElts - 1).umul_ov(Step, Overflow);
if (Overflow)
break;
Known.Zero.setHighBits(MaxValue.countl_zero());
break;
}
case ISD::BUILD_VECTOR:
assert(!Op.getValueType().isScalableVector());
// Collect the known bits that are shared by every demanded vector element.
Known.Zero.setAllBits(); Known.One.setAllBits();
for (unsigned i = 0, e = Op.getNumOperands(); i != e; ++i) {
if (!DemandedElts[i])
continue;
SDValue SrcOp = Op.getOperand(i);
Known2 = computeKnownBits(SrcOp, Depth + 1);
// BUILD_VECTOR can implicitly truncate sources, we must handle this.
if (SrcOp.getValueSizeInBits() != BitWidth) {
assert(SrcOp.getValueSizeInBits() > BitWidth &&
"Expected BUILD_VECTOR implicit truncation");
Known2 = Known2.trunc(BitWidth);
}
// Known bits are the values that are shared by every demanded element.
Known = Known.intersectWith(Known2);
// If we don't know any bits, early out.
if (Known.isUnknown())
break;
}
break;
case ISD::VECTOR_SHUFFLE: {
assert(!Op.getValueType().isScalableVector());
// Collect the known bits that are shared by every vector element referenced
// by the shuffle.
APInt DemandedLHS, DemandedRHS;
const ShuffleVectorSDNode *SVN = cast<ShuffleVectorSDNode>(Op);
assert(NumElts == SVN->getMask().size() && "Unexpected vector size");
if (!getShuffleDemandedElts(NumElts, SVN->getMask(), DemandedElts,
DemandedLHS, DemandedRHS))
break;
// Known bits are the values that are shared by every demanded element.
Known.Zero.setAllBits(); Known.One.setAllBits();
if (!!DemandedLHS) {
SDValue LHS = Op.getOperand(0);
Known2 = computeKnownBits(LHS, DemandedLHS, Depth + 1);
Known = Known.intersectWith(Known2);
}
// If we don't know any bits, early out.
if (Known.isUnknown())
break;
if (!!DemandedRHS) {
SDValue RHS = Op.getOperand(1);
Known2 = computeKnownBits(RHS, DemandedRHS, Depth + 1);
Known = Known.intersectWith(Known2);
}
break;
}
case ISD::VSCALE: {
const Function &F = getMachineFunction().getFunction();
const APInt &Multiplier = Op.getConstantOperandAPInt(0);
Known = getVScaleRange(&F, BitWidth).multiply(Multiplier).toKnownBits();
break;
}
case ISD::CONCAT_VECTORS: {
if (Op.getValueType().isScalableVector())
break;
// Split DemandedElts and test each of the demanded subvectors.
Known.Zero.setAllBits(); Known.One.setAllBits();
EVT SubVectorVT = Op.getOperand(0).getValueType();
unsigned NumSubVectorElts = SubVectorVT.getVectorNumElements();
unsigned NumSubVectors = Op.getNumOperands();
for (unsigned i = 0; i != NumSubVectors; ++i) {
APInt DemandedSub =
DemandedElts.extractBits(NumSubVectorElts, i * NumSubVectorElts);
if (!!DemandedSub) {
SDValue Sub = Op.getOperand(i);
Known2 = computeKnownBits(Sub, DemandedSub, Depth + 1);
Known = Known.intersectWith(Known2);
}
// If we don't know any bits, early out.
if (Known.isUnknown())
break;
}
break;
}
case ISD::INSERT_SUBVECTOR: {
if (Op.getValueType().isScalableVector())
break;
// Demand any elements from the subvector and the remainder from the src its
// inserted into.
SDValue Src = Op.getOperand(0);
SDValue Sub = Op.getOperand(1);
uint64_t Idx = Op.getConstantOperandVal(2);
unsigned NumSubElts = Sub.getValueType().getVectorNumElements();
APInt DemandedSubElts = DemandedElts.extractBits(NumSubElts, Idx);
APInt DemandedSrcElts = DemandedElts;
DemandedSrcElts.insertBits(APInt::getZero(NumSubElts), Idx);
Known.One.setAllBits();
Known.Zero.setAllBits();
if (!!DemandedSubElts) {
Known = computeKnownBits(Sub, DemandedSubElts, Depth + 1);
if (Known.isUnknown())
break; // early-out.
}
if (!!DemandedSrcElts) {
Known2 = computeKnownBits(Src, DemandedSrcElts, Depth + 1);
Known = Known.intersectWith(Known2);
}
break;
}
case ISD::EXTRACT_SUBVECTOR: {
// Offset the demanded elts by the subvector index.
SDValue Src = Op.getOperand(0);
// Bail until we can represent demanded elements for scalable vectors.
if (Op.getValueType().isScalableVector() || Src.getValueType().isScalableVector())
break;
uint64_t Idx = Op.getConstantOperandVal(1);
unsigned NumSrcElts = Src.getValueType().getVectorNumElements();
APInt DemandedSrcElts = DemandedElts.zext(NumSrcElts).shl(Idx);
Known = computeKnownBits(Src, DemandedSrcElts, Depth + 1);
break;
}
case ISD::SCALAR_TO_VECTOR: {
if (Op.getValueType().isScalableVector())
break;
// We know about scalar_to_vector as much as we know about it source,
// which becomes the first element of otherwise unknown vector.
if (DemandedElts != 1)
break;
SDValue N0 = Op.getOperand(0);
Known = computeKnownBits(N0, Depth + 1);
if (N0.getValueSizeInBits() != BitWidth)
Known = Known.trunc(BitWidth);
break;
}
case ISD::BITCAST: {
if (Op.getValueType().isScalableVector())
break;
SDValue N0 = Op.getOperand(0);
EVT SubVT = N0.getValueType();
unsigned SubBitWidth = SubVT.getScalarSizeInBits();
// Ignore bitcasts from unsupported types.
if (!(SubVT.isInteger() || SubVT.isFloatingPoint()))
break;
// Fast handling of 'identity' bitcasts.
if (BitWidth == SubBitWidth) {
Known = computeKnownBits(N0, DemandedElts, Depth + 1);
break;
}
bool IsLE = getDataLayout().isLittleEndian();
// Bitcast 'small element' vector to 'large element' scalar/vector.
if ((BitWidth % SubBitWidth) == 0) {
assert(N0.getValueType().isVector() && "Expected bitcast from vector");
// Collect known bits for the (larger) output by collecting the known
// bits from each set of sub elements and shift these into place.
// We need to separately call computeKnownBits for each set of
// sub elements as the knownbits for each is likely to be different.
unsigned SubScale = BitWidth / SubBitWidth;
APInt SubDemandedElts(NumElts * SubScale, 0);
for (unsigned i = 0; i != NumElts; ++i)
if (DemandedElts[i])
SubDemandedElts.setBit(i * SubScale);
for (unsigned i = 0; i != SubScale; ++i) {
Known2 = computeKnownBits(N0, SubDemandedElts.shl(i),
Depth + 1);
unsigned Shifts = IsLE ? i : SubScale - 1 - i;
Known.insertBits(Known2, SubBitWidth * Shifts);
}
}
// Bitcast 'large element' scalar/vector to 'small element' vector.
if ((SubBitWidth % BitWidth) == 0) {
assert(Op.getValueType().isVector() && "Expected bitcast to vector");
// Collect known bits for the (smaller) output by collecting the known
// bits from the overlapping larger input elements and extracting the
// sub sections we actually care about.
unsigned SubScale = SubBitWidth / BitWidth;
APInt SubDemandedElts =
APIntOps::ScaleBitMask(DemandedElts, NumElts / SubScale);
Known2 = computeKnownBits(N0, SubDemandedElts, Depth + 1);
Known.Zero.setAllBits(); Known.One.setAllBits();
for (unsigned i = 0; i != NumElts; ++i)
if (DemandedElts[i]) {
unsigned Shifts = IsLE ? i : NumElts - 1 - i;
unsigned Offset = (Shifts % SubScale) * BitWidth;
Known = Known.intersectWith(Known2.extractBits(BitWidth, Offset));
// If we don't know any bits, early out.
if (Known.isUnknown())
break;
}
}
break;
}
case ISD::AND:
Known = computeKnownBits(Op.getOperand(1), DemandedElts, Depth + 1);
Known2 = computeKnownBits(Op.getOperand(0), DemandedElts, Depth + 1);
Known &= Known2;
break;
case ISD::OR:
Known = computeKnownBits(Op.getOperand(1), DemandedElts, Depth + 1);
Known2 = computeKnownBits(Op.getOperand(0), DemandedElts, Depth + 1);
Known |= Known2;
break;
case ISD::XOR:
Known = computeKnownBits(Op.getOperand(1), DemandedElts, Depth + 1);
Known2 = computeKnownBits(Op.getOperand(0), DemandedElts, Depth + 1);
Known ^= Known2;
break;
case ISD::MUL: {
Known = computeKnownBits(Op.getOperand(1), DemandedElts, Depth + 1);
Known2 = computeKnownBits(Op.getOperand(0), DemandedElts, Depth + 1);
bool SelfMultiply = Op.getOperand(0) == Op.getOperand(1);
// TODO: SelfMultiply can be poison, but not undef.
if (SelfMultiply)
SelfMultiply &= isGuaranteedNotToBeUndefOrPoison(
Op.getOperand(0), DemandedElts, false, Depth + 1);
Known = KnownBits::mul(Known, Known2, SelfMultiply);
// If the multiplication is known not to overflow, the product of a number
// with itself is non-negative. Only do this if we didn't already computed
// the opposite value for the sign bit.
if (Op->getFlags().hasNoSignedWrap() &&
Op.getOperand(0) == Op.getOperand(1) &&
!Known.isNegative())
Known.makeNonNegative();
break;
}
case ISD::MULHU: {
Known = computeKnownBits(Op.getOperand(1), DemandedElts, Depth + 1);
Known2 = computeKnownBits(Op.getOperand(0), DemandedElts, Depth + 1);
Known = KnownBits::mulhu(Known, Known2);
break;
}
case ISD::MULHS: {
Known = computeKnownBits(Op.getOperand(1), DemandedElts, Depth + 1);
Known2 = computeKnownBits(Op.getOperand(0), DemandedElts, Depth + 1);
Known = KnownBits::mulhs(Known, Known2);
break;
}
case ISD::UMUL_LOHI: {
assert((Op.getResNo() == 0 || Op.getResNo() == 1) && "Unknown result");
Known = computeKnownBits(Op.getOperand(1), DemandedElts, Depth + 1);
Known2 = computeKnownBits(Op.getOperand(0), DemandedElts, Depth + 1);
bool SelfMultiply = Op.getOperand(0) == Op.getOperand(1);
if (Op.getResNo() == 0)
Known = KnownBits::mul(Known, Known2, SelfMultiply);
else
Known = KnownBits::mulhu(Known, Known2);
break;
}
case ISD::SMUL_LOHI: {
assert((Op.getResNo() == 0 || Op.getResNo() == 1) && "Unknown result");
Known = computeKnownBits(Op.getOperand(1), DemandedElts, Depth + 1);
Known2 = computeKnownBits(Op.getOperand(0), DemandedElts, Depth + 1);
bool SelfMultiply = Op.getOperand(0) == Op.getOperand(1);
if (Op.getResNo() == 0)
Known = KnownBits::mul(Known, Known2, SelfMultiply);
else
Known = KnownBits::mulhs(Known, Known2);
break;
}
case ISD::AVGCEILU: {
Known = computeKnownBits(Op.getOperand(0), DemandedElts, Depth + 1);
Known2 = computeKnownBits(Op.getOperand(1), DemandedElts, Depth + 1);
Known = Known.zext(BitWidth + 1);
Known2 = Known2.zext(BitWidth + 1);
KnownBits One = KnownBits::makeConstant(APInt(1, 1));
Known = KnownBits::computeForAddCarry(Known, Known2, One);
Known = Known.extractBits(BitWidth, 1);
break;
}
case ISD::SELECT:
case ISD::VSELECT:
Known = computeKnownBits(Op.getOperand(2), DemandedElts, Depth+1);
// If we don't know any bits, early out.
if (Known.isUnknown())
break;
Known2 = computeKnownBits(Op.getOperand(1), DemandedElts, Depth+1);
// Only known if known in both the LHS and RHS.
Known = Known.intersectWith(Known2);
break;
case ISD::SELECT_CC:
Known = computeKnownBits(Op.getOperand(3), DemandedElts, Depth+1);
// If we don't know any bits, early out.
if (Known.isUnknown())
break;
Known2 = computeKnownBits(Op.getOperand(2), DemandedElts, Depth+1);
// Only known if known in both the LHS and RHS.
Known = Known.intersectWith(Known2);
break;
case ISD::SMULO:
case ISD::UMULO:
if (Op.getResNo() != 1)
break;
// The boolean result conforms to getBooleanContents.
// If we know the result of a setcc has the top bits zero, use this info.
// We know that we have an integer-based boolean since these operations
// are only available for integer.
if (TLI->getBooleanContents(Op.getValueType().isVector(), false) ==
TargetLowering::ZeroOrOneBooleanContent &&
BitWidth > 1)
Known.Zero.setBitsFrom(1);
break;
case ISD::SETCC:
case ISD::SETCCCARRY:
case ISD::STRICT_FSETCC:
case ISD::STRICT_FSETCCS: {
unsigned OpNo = Op->isStrictFPOpcode() ? 1 : 0;
// If we know the result of a setcc has the top bits zero, use this info.
if (TLI->getBooleanContents(Op.getOperand(OpNo).getValueType()) ==
TargetLowering::ZeroOrOneBooleanContent &&
BitWidth > 1)
Known.Zero.setBitsFrom(1);
break;
}
case ISD::SHL:
Known = computeKnownBits(Op.getOperand(0), DemandedElts, Depth + 1);
Known2 = computeKnownBits(Op.getOperand(1), DemandedElts, Depth + 1);
Known = KnownBits::shl(Known, Known2);
// Minimum shift low bits are known zero.
if (const APInt *ShMinAmt =
getValidMinimumShiftAmountConstant(Op, DemandedElts))
Known.Zero.setLowBits(ShMinAmt->getZExtValue());
break;
case ISD::SRL:
Known = computeKnownBits(Op.getOperand(0), DemandedElts, Depth + 1);
Known2 = computeKnownBits(Op.getOperand(1), DemandedElts, Depth + 1);
Known = KnownBits::lshr(Known, Known2);
// Minimum shift high bits are known zero.
if (const APInt *ShMinAmt =
getValidMinimumShiftAmountConstant(Op, DemandedElts))
Known.Zero.setHighBits(ShMinAmt->getZExtValue());
break;
case ISD::SRA:
Known = computeKnownBits(Op.getOperand(0), DemandedElts, Depth + 1);
Known2 = computeKnownBits(Op.getOperand(1), DemandedElts, Depth + 1);
Known = KnownBits::ashr(Known, Known2);
break;
case ISD::FSHL:
case ISD::FSHR:
if (ConstantSDNode *C = isConstOrConstSplat(Op.getOperand(2), DemandedElts)) {
unsigned Amt = C->getAPIntValue().urem(BitWidth);
// For fshl, 0-shift returns the 1st arg.
// For fshr, 0-shift returns the 2nd arg.
if (Amt == 0) {
Known = computeKnownBits(Op.getOperand(Opcode == ISD::FSHL ? 0 : 1),
DemandedElts, Depth + 1);
break;
}
// fshl: (X << (Z % BW)) | (Y >> (BW - (Z % BW)))
// fshr: (X << (BW - (Z % BW))) | (Y >> (Z % BW))
Known = computeKnownBits(Op.getOperand(0), DemandedElts, Depth + 1);
Known2 = computeKnownBits(Op.getOperand(1), DemandedElts, Depth + 1);
if (Opcode == ISD::FSHL) {
Known.One <<= Amt;
Known.Zero <<= Amt;
Known2.One.lshrInPlace(BitWidth - Amt);
Known2.Zero.lshrInPlace(BitWidth - Amt);
} else {
Known.One <<= BitWidth - Amt;
Known.Zero <<= BitWidth - Amt;
Known2.One.lshrInPlace(Amt);
Known2.Zero.lshrInPlace(Amt);
}
Known = Known.unionWith(Known2);
}
break;
case ISD::SHL_PARTS:
case ISD::SRA_PARTS:
case ISD::SRL_PARTS: {
assert((Op.getResNo() == 0 || Op.getResNo() == 1) && "Unknown result");
// Collect lo/hi source values and concatenate.
unsigned LoBits = Op.getOperand(0).getScalarValueSizeInBits();
unsigned HiBits = Op.getOperand(1).getScalarValueSizeInBits();
Known = computeKnownBits(Op.getOperand(0), DemandedElts, Depth + 1);
Known2 = computeKnownBits(Op.getOperand(1), DemandedElts, Depth + 1);
Known = Known2.concat(Known);
// Collect shift amount.
Known2 = computeKnownBits(Op.getOperand(2), DemandedElts, Depth + 1);
if (Opcode == ISD::SHL_PARTS)
Known = KnownBits::shl(Known, Known2);
else if (Opcode == ISD::SRA_PARTS)
Known = KnownBits::ashr(Known, Known2);
else // if (Opcode == ISD::SRL_PARTS)
Known = KnownBits::lshr(Known, Known2);
// TODO: Minimum shift low/high bits are known zero.
if (Op.getResNo() == 0)
Known = Known.extractBits(LoBits, 0);
else
Known = Known.extractBits(HiBits, LoBits);
break;
}
case ISD::SIGN_EXTEND_INREG: {
Known = computeKnownBits(Op.getOperand(0), DemandedElts, Depth + 1);
EVT EVT = cast<VTSDNode>(Op.getOperand(1))->getVT();
Known = Known.sextInReg(EVT.getScalarSizeInBits());
break;
}
case ISD::CTTZ:
case ISD::CTTZ_ZERO_UNDEF: {
Known2 = computeKnownBits(Op.getOperand(0), DemandedElts, Depth + 1);
// If we have a known 1, its position is our upper bound.
unsigned PossibleTZ = Known2.countMaxTrailingZeros();
unsigned LowBits = llvm::bit_width(PossibleTZ);
Known.Zero.setBitsFrom(LowBits);
break;
}
case ISD::CTLZ:
case ISD::CTLZ_ZERO_UNDEF: {
Known2 = computeKnownBits(Op.getOperand(0), DemandedElts, Depth + 1);
// If we have a known 1, its position is our upper bound.
unsigned PossibleLZ = Known2.countMaxLeadingZeros();
unsigned LowBits = llvm::bit_width(PossibleLZ);
Known.Zero.setBitsFrom(LowBits);
break;
}
case ISD::CTPOP: {
Known2 = computeKnownBits(Op.getOperand(0), DemandedElts, Depth + 1);
// If we know some of the bits are zero, they can't be one.
unsigned PossibleOnes = Known2.countMaxPopulation();
Known.Zero.setBitsFrom(llvm::bit_width(PossibleOnes));
break;
}
case ISD::PARITY: {
// Parity returns 0 everywhere but the LSB.
Known.Zero.setBitsFrom(1);
break;
}
case ISD::LOAD: {
LoadSDNode *LD = cast<LoadSDNode>(Op);
const Constant *Cst = TLI->getTargetConstantFromLoad(LD);
if (ISD::isNON_EXTLoad(LD) && Cst) {
// Determine any common known bits from the loaded constant pool value.
Type *CstTy = Cst->getType();
if ((NumElts * BitWidth) == CstTy->getPrimitiveSizeInBits() &&
!Op.getValueType().isScalableVector()) {
// If its a vector splat, then we can (quickly) reuse the scalar path.
// NOTE: We assume all elements match and none are UNDEF.
if (CstTy->isVectorTy()) {
if (const Constant *Splat = Cst->getSplatValue()) {
Cst = Splat;
CstTy = Cst->getType();
}
}
// TODO - do we need to handle different bitwidths?
if (CstTy->isVectorTy() && BitWidth == CstTy->getScalarSizeInBits()) {
// Iterate across all vector elements finding common known bits.
Known.One.setAllBits();
Known.Zero.setAllBits();
for (unsigned i = 0; i != NumElts; ++i) {
if (!DemandedElts[i])
continue;
if (Constant *Elt = Cst->getAggregateElement(i)) {
if (auto *CInt = dyn_cast<ConstantInt>(Elt)) {
const APInt &Value = CInt->getValue();
Known.One &= Value;
Known.Zero &= ~Value;
continue;
}
if (auto *CFP = dyn_cast<ConstantFP>(Elt)) {
APInt Value = CFP->getValueAPF().bitcastToAPInt();
Known.One &= Value;
Known.Zero &= ~Value;
continue;
}
}
Known.One.clearAllBits();
Known.Zero.clearAllBits();
break;
}
} else if (BitWidth == CstTy->getPrimitiveSizeInBits()) {
if (auto *CInt = dyn_cast<ConstantInt>(Cst)) {
Known = KnownBits::makeConstant(CInt->getValue());
} else if (auto *CFP = dyn_cast<ConstantFP>(Cst)) {
Known =
KnownBits::makeConstant(CFP->getValueAPF().bitcastToAPInt());
}
}
}
} else if (ISD::isZEXTLoad(Op.getNode()) && Op.getResNo() == 0) {
// If this is a ZEXTLoad and we are looking at the loaded value.
EVT VT = LD->getMemoryVT();
unsigned MemBits = VT.getScalarSizeInBits();
Known.Zero.setBitsFrom(MemBits);
} else if (const MDNode *Ranges = LD->getRanges()) {
EVT VT = LD->getValueType(0);
// TODO: Handle for extending loads
if (LD->getExtensionType() == ISD::NON_EXTLOAD) {
if (VT.isVector()) {
// Handle truncation to the first demanded element.
// TODO: Figure out which demanded elements are covered
if (DemandedElts != 1 || !getDataLayout().isLittleEndian())
break;
// Handle the case where a load has a vector type, but scalar memory
// with an attached range.
EVT MemVT = LD->getMemoryVT();
KnownBits KnownFull(MemVT.getSizeInBits());
computeKnownBitsFromRangeMetadata(*Ranges, KnownFull);
Known = KnownFull.trunc(BitWidth);
} else
computeKnownBitsFromRangeMetadata(*Ranges, Known);
}
}
break;
}
case ISD::ZERO_EXTEND_VECTOR_INREG: {
if (Op.getValueType().isScalableVector())
break;
EVT InVT = Op.getOperand(0).getValueType();
APInt InDemandedElts = DemandedElts.zext(InVT.getVectorNumElements());
Known = computeKnownBits(Op.getOperand(0), InDemandedElts, Depth + 1);
Known = Known.zext(BitWidth);
break;
}
case ISD::ZERO_EXTEND: {
Known = computeKnownBits(Op.getOperand(0), DemandedElts, Depth + 1);
Known = Known.zext(BitWidth);
break;
}
case ISD::SIGN_EXTEND_VECTOR_INREG: {
if (Op.getValueType().isScalableVector())
break;
EVT InVT = Op.getOperand(0).getValueType();
APInt InDemandedElts = DemandedElts.zext(InVT.getVectorNumElements());
Known = computeKnownBits(Op.getOperand(0), InDemandedElts, Depth + 1);
// If the sign bit is known to be zero or one, then sext will extend
// it to the top bits, else it will just zext.
Known = Known.sext(BitWidth);
break;
}
case ISD::SIGN_EXTEND: {
Known = computeKnownBits(Op.getOperand(0), DemandedElts, Depth + 1);
// If the sign bit is known to be zero or one, then sext will extend
// it to the top bits, else it will just zext.
Known = Known.sext(BitWidth);
break;
}
case ISD::ANY_EXTEND_VECTOR_INREG: {
if (Op.getValueType().isScalableVector())
break;
EVT InVT = Op.getOperand(0).getValueType();
APInt InDemandedElts = DemandedElts.zext(InVT.getVectorNumElements());
Known = computeKnownBits(Op.getOperand(0), InDemandedElts, Depth + 1);
Known = Known.anyext(BitWidth);
break;
}
case ISD::ANY_EXTEND: {
Known = computeKnownBits(Op.getOperand(0), DemandedElts, Depth + 1);
Known = Known.anyext(BitWidth);
break;
}
case ISD::TRUNCATE: {
Known = computeKnownBits(Op.getOperand(0), DemandedElts, Depth + 1);
Known = Known.trunc(BitWidth);
break;
}
case ISD::AssertZext: {
EVT VT = cast<VTSDNode>(Op.getOperand(1))->getVT();
APInt InMask = APInt::getLowBitsSet(BitWidth, VT.getSizeInBits());
Known = computeKnownBits(Op.getOperand(0), DemandedElts, Depth + 1);
Known.Zero |= (~InMask);
Known.One &= (~Known.Zero);
break;
}
case ISD::AssertAlign: {
unsigned LogOfAlign = Log2(cast<AssertAlignSDNode>(Op)->getAlign());
assert(LogOfAlign != 0);
// TODO: Should use maximum with source
// If a node is guaranteed to be aligned, set low zero bits accordingly as
// well as clearing one bits.
Known.Zero.setLowBits(LogOfAlign);
Known.One.clearLowBits(LogOfAlign);
break;
}
case ISD::FGETSIGN:
// All bits are zero except the low bit.
Known.Zero.setBitsFrom(1);
break;
case ISD::ADD:
case ISD::SUB: {
SDNodeFlags Flags = Op.getNode()->getFlags();
Known = computeKnownBits(Op.getOperand(0), DemandedElts, Depth + 1);
Known2 = computeKnownBits(Op.getOperand(1), DemandedElts, Depth + 1);
Known = KnownBits::computeForAddSub(Op.getOpcode() == ISD::ADD,
Flags.hasNoSignedWrap(), Known, Known2);
break;
}
case ISD::USUBO:
case ISD::SSUBO:
case ISD::USUBO_CARRY:
case ISD::SSUBO_CARRY:
if (Op.getResNo() == 1) {
// If we know the result of a setcc has the top bits zero, use this info.
if (TLI->getBooleanContents(Op.getOperand(0).getValueType()) ==
TargetLowering::ZeroOrOneBooleanContent &&
BitWidth > 1)
Known.Zero.setBitsFrom(1);
break;
}
[[fallthrough]];
case ISD::SUBC: {
assert(Op.getResNo() == 0 &&
"We only compute knownbits for the difference here.");
// With USUBO_CARRY and SSUBO_CARRY a borrow bit may be added in.
KnownBits Borrow(1);
if (Opcode == ISD::USUBO_CARRY || Opcode == ISD::SSUBO_CARRY) {
Borrow = computeKnownBits(Op.getOperand(2), DemandedElts, Depth + 1);
// Borrow has bit width 1
Borrow = Borrow.trunc(1);
} else {
Borrow.setAllZero();
}
Known = computeKnownBits(Op.getOperand(0), DemandedElts, Depth + 1);
Known2 = computeKnownBits(Op.getOperand(1), DemandedElts, Depth + 1);
Known = KnownBits::computeForSubBorrow(Known, Known2, Borrow);
break;
}
case ISD::UADDO:
case ISD::SADDO:
case ISD::UADDO_CARRY:
case ISD::SADDO_CARRY:
if (Op.getResNo() == 1) {
// If we know the result of a setcc has the top bits zero, use this info.
if (TLI->getBooleanContents(Op.getOperand(0).getValueType()) ==
TargetLowering::ZeroOrOneBooleanContent &&
BitWidth > 1)
Known.Zero.setBitsFrom(1);
break;
}
[[fallthrough]];
case ISD::ADDC:
case ISD::ADDE: {
assert(Op.getResNo() == 0 && "We only compute knownbits for the sum here.");
// With ADDE and UADDO_CARRY, a carry bit may be added in.
KnownBits Carry(1);
if (Opcode == ISD::ADDE)
// Can't track carry from glue, set carry to unknown.
Carry.resetAll();
else if (Opcode == ISD::UADDO_CARRY || Opcode == ISD::SADDO_CARRY) {
Carry = computeKnownBits(Op.getOperand(2), DemandedElts, Depth + 1);
// Carry has bit width 1
Carry = Carry.trunc(1);
} else {
Carry.setAllZero();
}
Known = computeKnownBits(Op.getOperand(0), DemandedElts, Depth + 1);
Known2 = computeKnownBits(Op.getOperand(1), DemandedElts, Depth + 1);
Known = KnownBits::computeForAddCarry(Known, Known2, Carry);
break;
}
case ISD::UDIV: {
Known = computeKnownBits(Op.getOperand(0), DemandedElts, Depth + 1);
Known2 = computeKnownBits(Op.getOperand(1), DemandedElts, Depth + 1);
Known = KnownBits::udiv(Known, Known2, Op->getFlags().hasExact());
break;
}
case ISD::SDIV: {
Known = computeKnownBits(Op.getOperand(0), DemandedElts, Depth + 1);
Known2 = computeKnownBits(Op.getOperand(1), DemandedElts, Depth + 1);
Known = KnownBits::sdiv(Known, Known2, Op->getFlags().hasExact());
break;
}
case ISD::SREM: {
Known = computeKnownBits(Op.getOperand(0), DemandedElts, Depth + 1);
Known2 = computeKnownBits(Op.getOperand(1), DemandedElts, Depth + 1);
Known = KnownBits::srem(Known, Known2);
break;
}
case ISD::UREM: {
Known = computeKnownBits(Op.getOperand(0), DemandedElts, Depth + 1);
Known2 = computeKnownBits(Op.getOperand(1), DemandedElts, Depth + 1);
Known = KnownBits::urem(Known, Known2);
break;
}
case ISD::EXTRACT_ELEMENT: {
Known = computeKnownBits(Op.getOperand(0), Depth+1);
const unsigned Index = Op.getConstantOperandVal(1);
const unsigned EltBitWidth = Op.getValueSizeInBits();
// Remove low part of known bits mask
Known.Zero = Known.Zero.getHiBits(Known.getBitWidth() - Index * EltBitWidth);
Known.One = Known.One.getHiBits(Known.getBitWidth() - Index * EltBitWidth);
// Remove high part of known bit mask
Known = Known.trunc(EltBitWidth);
break;
}
case ISD::EXTRACT_VECTOR_ELT: {
SDValue InVec = Op.getOperand(0);
SDValue EltNo = Op.getOperand(1);
EVT VecVT = InVec.getValueType();
// computeKnownBits not yet implemented for scalable vectors.
if (VecVT.isScalableVector())
break;
const unsigned EltBitWidth = VecVT.getScalarSizeInBits();
const unsigned NumSrcElts = VecVT.getVectorNumElements();
// If BitWidth > EltBitWidth the value is anyext:ed. So we do not know
// anything about the extended bits.
if (BitWidth > EltBitWidth)
Known = Known.trunc(EltBitWidth);
// If we know the element index, just demand that vector element, else for
// an unknown element index, ignore DemandedElts and demand them all.
APInt DemandedSrcElts = APInt::getAllOnes(NumSrcElts);
auto *ConstEltNo = dyn_cast<ConstantSDNode>(EltNo);
if (ConstEltNo && ConstEltNo->getAPIntValue().ult(NumSrcElts))
DemandedSrcElts =
APInt::getOneBitSet(NumSrcElts, ConstEltNo->getZExtValue());
Known = computeKnownBits(InVec, DemandedSrcElts, Depth + 1);
if (BitWidth > EltBitWidth)
Known = Known.anyext(BitWidth);
break;
}
case ISD::INSERT_VECTOR_ELT: {
if (Op.getValueType().isScalableVector())
break;
// If we know the element index, split the demand between the
// source vector and the inserted element, otherwise assume we need
// the original demanded vector elements and the value.
SDValue InVec = Op.getOperand(0);
SDValue InVal = Op.getOperand(1);
SDValue EltNo = Op.getOperand(2);
bool DemandedVal = true;
APInt DemandedVecElts = DemandedElts;
auto *CEltNo = dyn_cast<ConstantSDNode>(EltNo);
if (CEltNo && CEltNo->getAPIntValue().ult(NumElts)) {
unsigned EltIdx = CEltNo->getZExtValue();
DemandedVal = !!DemandedElts[EltIdx];
DemandedVecElts.clearBit(EltIdx);
}
Known.One.setAllBits();
Known.Zero.setAllBits();
if (DemandedVal) {
Known2 = computeKnownBits(InVal, Depth + 1);
Known = Known.intersectWith(Known2.zextOrTrunc(BitWidth));
}
if (!!DemandedVecElts) {
Known2 = computeKnownBits(InVec, DemandedVecElts, Depth + 1);
Known = Known.intersectWith(Known2);
}
break;
}
case ISD::BITREVERSE: {
Known2 = computeKnownBits(Op.getOperand(0), DemandedElts, Depth + 1);
Known = Known2.reverseBits();
break;
}
case ISD::BSWAP: {
Known2 = computeKnownBits(Op.getOperand(0), DemandedElts, Depth + 1);
Known = Known2.byteSwap();
break;
}
case ISD::ABS: {
Known2 = computeKnownBits(Op.getOperand(0), DemandedElts, Depth + 1);
Known = Known2.abs();
break;
}
case ISD::USUBSAT: {
Known = computeKnownBits(Op.getOperand(0), DemandedElts, Depth + 1);
Known2 = computeKnownBits(Op.getOperand(1), DemandedElts, Depth + 1);
Known = KnownBits::usub_sat(Known, Known2);
break;
}
case ISD::UMIN: {
Known = computeKnownBits(Op.getOperand(0), DemandedElts, Depth + 1);
Known2 = computeKnownBits(Op.getOperand(1), DemandedElts, Depth + 1);
Known = KnownBits::umin(Known, Known2);
break;
}
case ISD::UMAX: {
Known = computeKnownBits(Op.getOperand(0), DemandedElts, Depth + 1);
Known2 = computeKnownBits(Op.getOperand(1), DemandedElts, Depth + 1);
Known = KnownBits::umax(Known, Known2);
break;
}
case ISD::SMIN:
case ISD::SMAX: {
// If we have a clamp pattern, we know that the number of sign bits will be
// the minimum of the clamp min/max range.
bool IsMax = (Opcode == ISD::SMAX);
ConstantSDNode *CstLow = nullptr, *CstHigh = nullptr;
if ((CstLow = isConstOrConstSplat(Op.getOperand(1), DemandedElts)))
if (Op.getOperand(0).getOpcode() == (IsMax ? ISD::SMIN : ISD::SMAX))
CstHigh =
isConstOrConstSplat(Op.getOperand(0).getOperand(1), DemandedElts);
if (CstLow && CstHigh) {
if (!IsMax)
std::swap(CstLow, CstHigh);
const APInt &ValueLow = CstLow->getAPIntValue();
const APInt &ValueHigh = CstHigh->getAPIntValue();
if (ValueLow.sle(ValueHigh)) {
unsigned LowSignBits = ValueLow.getNumSignBits();
unsigned HighSignBits = ValueHigh.getNumSignBits();
unsigned MinSignBits = std::min(LowSignBits, HighSignBits);
if (ValueLow.isNegative() && ValueHigh.isNegative()) {
Known.One.setHighBits(MinSignBits);
break;
}
if (ValueLow.isNonNegative() && ValueHigh.isNonNegative()) {
Known.Zero.setHighBits(MinSignBits);
break;
}
}
}
Known = computeKnownBits(Op.getOperand(0), DemandedElts, Depth + 1);
Known2 = computeKnownBits(Op.getOperand(1), DemandedElts, Depth + 1);
if (IsMax)
Known = KnownBits::smax(Known, Known2);
else
Known = KnownBits::smin(Known, Known2);
// For SMAX, if CstLow is non-negative we know the result will be
// non-negative and thus all sign bits are 0.
// TODO: There's an equivalent of this for smin with negative constant for
// known ones.
if (IsMax && CstLow) {
const APInt &ValueLow = CstLow->getAPIntValue();
if (ValueLow.isNonNegative()) {
unsigned SignBits = ComputeNumSignBits(Op.getOperand(0), Depth + 1);
Known.Zero.setHighBits(std::min(SignBits, ValueLow.getNumSignBits()));
}
}
break;
}
case ISD::FP_TO_UINT_SAT: {
// FP_TO_UINT_SAT produces an unsigned value that fits in the saturating VT.
EVT VT = cast<VTSDNode>(Op.getOperand(1))->getVT();
Known.Zero |= APInt::getBitsSetFrom(BitWidth, VT.getScalarSizeInBits());
break;
}
case ISD::ATOMIC_CMP_SWAP_WITH_SUCCESS:
if (Op.getResNo() == 1) {
// The boolean result conforms to getBooleanContents.
// If we know the result of a setcc has the top bits zero, use this info.
// We know that we have an integer-based boolean since these operations
// are only available for integer.
if (TLI->getBooleanContents(Op.getValueType().isVector(), false) ==
TargetLowering::ZeroOrOneBooleanContent &&
BitWidth > 1)
Known.Zero.setBitsFrom(1);
break;
}
[[fallthrough]];
case ISD::ATOMIC_CMP_SWAP:
case ISD::ATOMIC_SWAP:
case ISD::ATOMIC_LOAD_ADD:
case ISD::ATOMIC_LOAD_SUB:
case ISD::ATOMIC_LOAD_AND:
case ISD::ATOMIC_LOAD_CLR:
case ISD::ATOMIC_LOAD_OR:
case ISD::ATOMIC_LOAD_XOR:
case ISD::ATOMIC_LOAD_NAND:
case ISD::ATOMIC_LOAD_MIN:
case ISD::ATOMIC_LOAD_MAX:
case ISD::ATOMIC_LOAD_UMIN:
case ISD::ATOMIC_LOAD_UMAX:
case ISD::ATOMIC_LOAD: {
unsigned MemBits =
cast<AtomicSDNode>(Op)->getMemoryVT().getScalarSizeInBits();
// If we are looking at the loaded value.
if (Op.getResNo() == 0) {
if (TLI->getExtendForAtomicOps() == ISD::ZERO_EXTEND)
Known.Zero.setBitsFrom(MemBits);
}
break;
}
case ISD::FrameIndex:
case ISD::TargetFrameIndex:
TLI->computeKnownBitsForFrameIndex(cast<FrameIndexSDNode>(Op)->getIndex(),
Known, getMachineFunction());
break;
default:
if (Opcode < ISD::BUILTIN_OP_END)
break;
[[fallthrough]];
case ISD::INTRINSIC_WO_CHAIN:
case ISD::INTRINSIC_W_CHAIN:
case ISD::INTRINSIC_VOID:
// TODO: Probably okay to remove after audit; here to reduce change size
// in initial enablement patch for scalable vectors
if (Op.getValueType().isScalableVector())
break;
// Allow the target to implement this method for its nodes.
TLI->computeKnownBitsForTargetNode(Op, Known, DemandedElts, *this, Depth);
break;
}
assert(!Known.hasConflict() && "Bits known to be one AND zero?");
return Known;
}
/// Convert ConstantRange OverflowResult into SelectionDAG::OverflowKind.
static SelectionDAG::OverflowKind mapOverflowResult(ConstantRange::OverflowResult OR) {
switch (OR) {
case ConstantRange::OverflowResult::MayOverflow:
return SelectionDAG::OFK_Sometime;
case ConstantRange::OverflowResult::AlwaysOverflowsLow:
case ConstantRange::OverflowResult::AlwaysOverflowsHigh:
return SelectionDAG::OFK_Always;
case ConstantRange::OverflowResult::NeverOverflows:
return SelectionDAG::OFK_Never;
}
llvm_unreachable("Unknown OverflowResult");
}
SelectionDAG::OverflowKind
SelectionDAG::computeOverflowForSignedAdd(SDValue N0, SDValue N1) const {
// X + 0 never overflow
if (isNullConstant(N1))
return OFK_Never;
// If both operands each have at least two sign bits, the addition
// cannot overflow.
if (ComputeNumSignBits(N0) > 1 && ComputeNumSignBits(N1) > 1)
return OFK_Never;
// TODO: Add ConstantRange::signedAddMayOverflow handling.
return OFK_Sometime;
}
SelectionDAG::OverflowKind
SelectionDAG::computeOverflowForUnsignedAdd(SDValue N0, SDValue N1) const {
// X + 0 never overflow
if (isNullConstant(N1))
return OFK_Never;
// mulhi + 1 never overflow
KnownBits N1Known = computeKnownBits(N1);
if (N0.getOpcode() == ISD::UMUL_LOHI && N0.getResNo() == 1 &&
N1Known.getMaxValue().ult(2))
return OFK_Never;
KnownBits N0Known = computeKnownBits(N0);
if (N1.getOpcode() == ISD::UMUL_LOHI && N1.getResNo() == 1 &&
N0Known.getMaxValue().ult(2))
return OFK_Never;
// Fallback to ConstantRange::unsignedAddMayOverflow handling.
ConstantRange N0Range = ConstantRange::fromKnownBits(N0Known, false);
ConstantRange N1Range = ConstantRange::fromKnownBits(N1Known, false);
return mapOverflowResult(N0Range.unsignedAddMayOverflow(N1Range));
}
SelectionDAG::OverflowKind
SelectionDAG::computeOverflowForSignedSub(SDValue N0, SDValue N1) const {
// X - 0 never overflow
if (isNullConstant(N1))
return OFK_Never;
// If both operands each have at least two sign bits, the subtraction
// cannot overflow.
if (ComputeNumSignBits(N0) > 1 && ComputeNumSignBits(N1) > 1)
return OFK_Never;
KnownBits N0Known = computeKnownBits(N0);
KnownBits N1Known = computeKnownBits(N1);
ConstantRange N0Range = ConstantRange::fromKnownBits(N0Known, true);
ConstantRange N1Range = ConstantRange::fromKnownBits(N1Known, true);
return mapOverflowResult(N0Range.signedSubMayOverflow(N1Range));
}
SelectionDAG::OverflowKind
SelectionDAG::computeOverflowForUnsignedSub(SDValue N0, SDValue N1) const {
// X - 0 never overflow
if (isNullConstant(N1))
return OFK_Never;
KnownBits N0Known = computeKnownBits(N0);
KnownBits N1Known = computeKnownBits(N1);
ConstantRange N0Range = ConstantRange::fromKnownBits(N0Known, false);
ConstantRange N1Range = ConstantRange::fromKnownBits(N1Known, false);
return mapOverflowResult(N0Range.unsignedSubMayOverflow(N1Range));
}
SelectionDAG::OverflowKind
SelectionDAG::computeOverflowForUnsignedMul(SDValue N0, SDValue N1) const {
// X * 0 and X * 1 never overflow.
if (isNullConstant(N1) || isOneConstant(N1))
return OFK_Never;
KnownBits N0Known = computeKnownBits(N0);
KnownBits N1Known = computeKnownBits(N1);
ConstantRange N0Range = ConstantRange::fromKnownBits(N0Known, false);
ConstantRange N1Range = ConstantRange::fromKnownBits(N1Known, false);
return mapOverflowResult(N0Range.unsignedMulMayOverflow(N1Range));
}
SelectionDAG::OverflowKind
SelectionDAG::computeOverflowForSignedMul(SDValue N0, SDValue N1) const {
// X * 0 and X * 1 never overflow.
if (isNullConstant(N1) || isOneConstant(N1))
return OFK_Never;
// Get the size of the result.
unsigned BitWidth = N0.getScalarValueSizeInBits();
// Sum of the sign bits.
unsigned SignBits = ComputeNumSignBits(N0) + ComputeNumSignBits(N1);
// If we have enough sign bits, then there's no overflow.
if (SignBits > BitWidth + 1)
return OFK_Never;
if (SignBits == BitWidth + 1) {
// The overflow occurs when the true multiplication of the
// the operands is the minimum negative number.
KnownBits N0Known = computeKnownBits(N0);
KnownBits N1Known = computeKnownBits(N1);
// If one of the operands is non-negative, then there's no
// overflow.
if (N0Known.isNonNegative() || N1Known.isNonNegative())
return OFK_Never;
}
return OFK_Sometime;
}
bool SelectionDAG::isKnownToBeAPowerOfTwo(SDValue Val, unsigned Depth) const {
if (Depth >= MaxRecursionDepth)
return false; // Limit search depth.
EVT OpVT = Val.getValueType();
unsigned BitWidth = OpVT.getScalarSizeInBits();
// Is the constant a known power of 2?
if (ISD::matchUnaryPredicate(Val, [BitWidth](ConstantSDNode *C) {
return C->getAPIntValue().zextOrTrunc(BitWidth).isPowerOf2();
}))
return true;
// A left-shift of a constant one will have exactly one bit set because
// shifting the bit off the end is undefined.
if (Val.getOpcode() == ISD::SHL) {
auto *C = isConstOrConstSplat(Val.getOperand(0));
if (C && C->getAPIntValue() == 1)
return true;
return isKnownToBeAPowerOfTwo(Val.getOperand(0), Depth + 1) &&
isKnownNeverZero(Val, Depth);
}
// Similarly, a logical right-shift of a constant sign-bit will have exactly
// one bit set.
if (Val.getOpcode() == ISD::SRL) {
auto *C = isConstOrConstSplat(Val.getOperand(0));
if (C && C->getAPIntValue().isSignMask())
return true;
return isKnownToBeAPowerOfTwo(Val.getOperand(0), Depth + 1) &&
isKnownNeverZero(Val, Depth);
}
if (Val.getOpcode() == ISD::ROTL || Val.getOpcode() == ISD::ROTR)
return isKnownToBeAPowerOfTwo(Val.getOperand(0), Depth + 1);
// Are all operands of a build vector constant powers of two?
if (Val.getOpcode() == ISD::BUILD_VECTOR)
if (llvm::all_of(Val->ops(), [BitWidth](SDValue E) {
if (ConstantSDNode *C = dyn_cast<ConstantSDNode>(E))
return C->getAPIntValue().zextOrTrunc(BitWidth).isPowerOf2();
return false;
}))
return true;
// Is the operand of a splat vector a constant power of two?
if (Val.getOpcode() == ISD::SPLAT_VECTOR)
if (ConstantSDNode *C = dyn_cast<ConstantSDNode>(Val->getOperand(0)))
if (C->getAPIntValue().zextOrTrunc(BitWidth).isPowerOf2())
return true;
// vscale(power-of-two) is a power-of-two for some targets
if (Val.getOpcode() == ISD::VSCALE &&
getTargetLoweringInfo().isVScaleKnownToBeAPowerOfTwo() &&
isKnownToBeAPowerOfTwo(Val.getOperand(0), Depth + 1))
return true;
if (Val.getOpcode() == ISD::SMIN || Val.getOpcode() == ISD::SMAX ||
Val.getOpcode() == ISD::UMIN || Val.getOpcode() == ISD::UMAX)
return isKnownToBeAPowerOfTwo(Val.getOperand(1), Depth + 1) &&
isKnownToBeAPowerOfTwo(Val.getOperand(0), Depth + 1);
if (Val.getOpcode() == ISD::SELECT || Val.getOpcode() == ISD::VSELECT)
return isKnownToBeAPowerOfTwo(Val.getOperand(2), Depth + 1) &&
isKnownToBeAPowerOfTwo(Val.getOperand(1), Depth + 1);
if (Val.getOpcode() == ISD::AND) {
// Looking for `x & -x` pattern:
// If x == 0:
// x & -x -> 0
// If x != 0:
// x & -x -> non-zero pow2
// so if we find the pattern return whether we know `x` is non-zero.
for (unsigned OpIdx = 0; OpIdx < 2; ++OpIdx) {
SDValue NegOp = Val.getOperand(OpIdx);
if (NegOp.getOpcode() == ISD::SUB &&
NegOp.getOperand(1) == Val.getOperand(1 - OpIdx) &&
isNullOrNullSplat(NegOp.getOperand(0)))
return isKnownNeverZero(Val.getOperand(1 - OpIdx), Depth);
}
}
if (Val.getOpcode() == ISD::ZERO_EXTEND)
return isKnownToBeAPowerOfTwo(Val.getOperand(0), Depth + 1);
// More could be done here, though the above checks are enough
// to handle some common cases.
return false;
}
unsigned SelectionDAG::ComputeNumSignBits(SDValue Op, unsigned Depth) const {
EVT VT = Op.getValueType();
// Since the number of lanes in a scalable vector is unknown at compile time,
// we track one bit which is implicitly broadcast to all lanes. This means
// that all lanes in a scalable vector are considered demanded.
APInt DemandedElts = VT.isFixedLengthVector()
? APInt::getAllOnes(VT.getVectorNumElements())
: APInt(1, 1);
return ComputeNumSignBits(Op, DemandedElts, Depth);
}
unsigned SelectionDAG::ComputeNumSignBits(SDValue Op, const APInt &DemandedElts,
unsigned Depth) const {
EVT VT = Op.getValueType();
assert((VT.isInteger() || VT.isFloatingPoint()) && "Invalid VT!");
unsigned VTBits = VT.getScalarSizeInBits();
unsigned NumElts = DemandedElts.getBitWidth();
unsigned Tmp, Tmp2;
unsigned FirstAnswer = 1;
if (auto *C = dyn_cast<ConstantSDNode>(Op)) {
const APInt &Val = C->getAPIntValue();
return Val.getNumSignBits();
}
if (Depth >= MaxRecursionDepth)
return 1; // Limit search depth.
if (!DemandedElts)
return 1; // No demanded elts, better to assume we don't know anything.
unsigned Opcode = Op.getOpcode();
switch (Opcode) {
default: break;
case ISD::AssertSext:
Tmp = cast<VTSDNode>(Op.getOperand(1))->getVT().getSizeInBits();
return VTBits-Tmp+1;
case ISD::AssertZext:
Tmp = cast<VTSDNode>(Op.getOperand(1))->getVT().getSizeInBits();
return VTBits-Tmp;
case ISD::MERGE_VALUES:
return ComputeNumSignBits(Op.getOperand(Op.getResNo()), DemandedElts,
Depth + 1);
case ISD::SPLAT_VECTOR: {
// Check if the sign bits of source go down as far as the truncated value.
unsigned NumSrcBits = Op.getOperand(0).getValueSizeInBits();
unsigned NumSrcSignBits = ComputeNumSignBits(Op.getOperand(0), Depth + 1);
if (NumSrcSignBits > (NumSrcBits - VTBits))
return NumSrcSignBits - (NumSrcBits - VTBits);
break;
}
case ISD::BUILD_VECTOR:
assert(!VT.isScalableVector());
Tmp = VTBits;
for (unsigned i = 0, e = Op.getNumOperands(); (i < e) && (Tmp > 1); ++i) {
if (!DemandedElts[i])
continue;
SDValue SrcOp = Op.getOperand(i);
// BUILD_VECTOR can implicitly truncate sources, we handle this specially
// for constant nodes to ensure we only look at the sign bits.
if (ConstantSDNode *C = dyn_cast<ConstantSDNode>(SrcOp)) {
APInt T = C->getAPIntValue().trunc(VTBits);
Tmp2 = T.getNumSignBits();
} else {
Tmp2 = ComputeNumSignBits(SrcOp, Depth + 1);
if (SrcOp.getValueSizeInBits() != VTBits) {
assert(SrcOp.getValueSizeInBits() > VTBits &&
"Expected BUILD_VECTOR implicit truncation");
unsigned ExtraBits = SrcOp.getValueSizeInBits() - VTBits;
Tmp2 = (Tmp2 > ExtraBits ? Tmp2 - ExtraBits : 1);
}
}
Tmp = std::min(Tmp, Tmp2);
}
return Tmp;
case ISD::VECTOR_SHUFFLE: {
// Collect the minimum number of sign bits that are shared by every vector
// element referenced by the shuffle.
APInt DemandedLHS, DemandedRHS;
const ShuffleVectorSDNode *SVN = cast<ShuffleVectorSDNode>(Op);
assert(NumElts == SVN->getMask().size() && "Unexpected vector size");
if (!getShuffleDemandedElts(NumElts, SVN->getMask(), DemandedElts,
DemandedLHS, DemandedRHS))
return 1;
Tmp = std::numeric_limits<unsigned>::max();
if (!!DemandedLHS)
Tmp = ComputeNumSignBits(Op.getOperand(0), DemandedLHS, Depth + 1);
if (!!DemandedRHS) {
Tmp2 = ComputeNumSignBits(Op.getOperand(1), DemandedRHS, Depth + 1);
Tmp = std::min(Tmp, Tmp2);
}
// If we don't know anything, early out and try computeKnownBits fall-back.
if (Tmp == 1)
break;
assert(Tmp <= VTBits && "Failed to determine minimum sign bits");
return Tmp;
}
case ISD::BITCAST: {
if (VT.isScalableVector())
break;
SDValue N0 = Op.getOperand(0);
EVT SrcVT = N0.getValueType();
unsigned SrcBits = SrcVT.getScalarSizeInBits();
// Ignore bitcasts from unsupported types..
if (!(SrcVT.isInteger() || SrcVT.isFloatingPoint()))
break;
// Fast handling of 'identity' bitcasts.
if (VTBits == SrcBits)
return ComputeNumSignBits(N0, DemandedElts, Depth + 1);
bool IsLE = getDataLayout().isLittleEndian();
// Bitcast 'large element' scalar/vector to 'small element' vector.
if ((SrcBits % VTBits) == 0) {
assert(VT.isVector() && "Expected bitcast to vector");
unsigned Scale = SrcBits / VTBits;
APInt SrcDemandedElts =
APIntOps::ScaleBitMask(DemandedElts, NumElts / Scale);
// Fast case - sign splat can be simply split across the small elements.
Tmp = ComputeNumSignBits(N0, SrcDemandedElts, Depth + 1);
if (Tmp == SrcBits)
return VTBits;
// Slow case - determine how far the sign extends into each sub-element.
Tmp2 = VTBits;
for (unsigned i = 0; i != NumElts; ++i)
if (DemandedElts[i]) {
unsigned SubOffset = i % Scale;
SubOffset = (IsLE ? ((Scale - 1) - SubOffset) : SubOffset);
SubOffset = SubOffset * VTBits;
if (Tmp <= SubOffset)
return 1;
Tmp2 = std::min(Tmp2, Tmp - SubOffset);
}
return Tmp2;
}
break;
}
case ISD::FP_TO_SINT_SAT:
// FP_TO_SINT_SAT produces a signed value that fits in the saturating VT.
Tmp = cast<VTSDNode>(Op.getOperand(1))->getVT().getScalarSizeInBits();
return VTBits - Tmp + 1;
case ISD::SIGN_EXTEND:
Tmp = VTBits - Op.getOperand(0).getScalarValueSizeInBits();
return ComputeNumSignBits(Op.getOperand(0), DemandedElts, Depth+1) + Tmp;
case ISD::SIGN_EXTEND_INREG:
// Max of the input and what this extends.
Tmp = cast<VTSDNode>(Op.getOperand(1))->getVT().getScalarSizeInBits();
Tmp = VTBits-Tmp+1;
Tmp2 = ComputeNumSignBits(Op.getOperand(0), DemandedElts, Depth+1);
return std::max(Tmp, Tmp2);
case ISD::SIGN_EXTEND_VECTOR_INREG: {
if (VT.isScalableVector())
break;
SDValue Src = Op.getOperand(0);
EVT SrcVT = Src.getValueType();
APInt DemandedSrcElts = DemandedElts.zext(SrcVT.getVectorNumElements());
Tmp = VTBits - SrcVT.getScalarSizeInBits();
return ComputeNumSignBits(Src, DemandedSrcElts, Depth+1) + Tmp;
}
case ISD::SRA:
Tmp = ComputeNumSignBits(Op.getOperand(0), DemandedElts, Depth + 1);
// SRA X, C -> adds C sign bits.
if (const APInt *ShAmt =
getValidMinimumShiftAmountConstant(Op, DemandedElts))
Tmp = std::min<uint64_t>(Tmp + ShAmt->getZExtValue(), VTBits);
return Tmp;
case ISD::SHL:
if (const APInt *ShAmt =
getValidMaximumShiftAmountConstant(Op, DemandedElts)) {
// shl destroys sign bits, ensure it doesn't shift out all sign bits.
Tmp = ComputeNumSignBits(Op.getOperand(0), DemandedElts, Depth + 1);
if (ShAmt->ult(Tmp))
return Tmp - ShAmt->getZExtValue();
}
break;
case ISD::AND:
case ISD::OR:
case ISD::XOR: // NOT is handled here.
// Logical binary ops preserve the number of sign bits at the worst.
Tmp = ComputeNumSignBits(Op.getOperand(0), DemandedElts, Depth+1);
if (Tmp != 1) {
Tmp2 = ComputeNumSignBits(Op.getOperand(1), DemandedElts, Depth+1);
FirstAnswer = std::min(Tmp, Tmp2);
// We computed what we know about the sign bits as our first
// answer. Now proceed to the generic code that uses
// computeKnownBits, and pick whichever answer is better.
}
break;
case ISD::SELECT:
case ISD::VSELECT:
Tmp = ComputeNumSignBits(Op.getOperand(1), DemandedElts, Depth+1);
if (Tmp == 1) return 1; // Early out.
Tmp2 = ComputeNumSignBits(Op.getOperand(2), DemandedElts, Depth+1);
return std::min(Tmp, Tmp2);
case ISD::SELECT_CC:
Tmp = ComputeNumSignBits(Op.getOperand(2), DemandedElts, Depth+1);
if (Tmp == 1) return 1; // Early out.
Tmp2 = ComputeNumSignBits(Op.getOperand(3), DemandedElts, Depth+1);
return std::min(Tmp, Tmp2);
case ISD::SMIN:
case ISD::SMAX: {
// If we have a clamp pattern, we know that the number of sign bits will be
// the minimum of the clamp min/max range.
bool IsMax = (Opcode == ISD::SMAX);
ConstantSDNode *CstLow = nullptr, *CstHigh = nullptr;
if ((CstLow = isConstOrConstSplat(Op.getOperand(1), DemandedElts)))
if (Op.getOperand(0).getOpcode() == (IsMax ? ISD::SMIN : ISD::SMAX))
CstHigh =
isConstOrConstSplat(Op.getOperand(0).getOperand(1), DemandedElts);
if (CstLow && CstHigh) {
if (!IsMax)
std::swap(CstLow, CstHigh);
if (CstLow->getAPIntValue().sle(CstHigh->getAPIntValue())) {
Tmp = CstLow->getAPIntValue().getNumSignBits();
Tmp2 = CstHigh->getAPIntValue().getNumSignBits();
return std::min(Tmp, Tmp2);
}
}
// Fallback - just get the minimum number of sign bits of the operands.
Tmp = ComputeNumSignBits(Op.getOperand(0), DemandedElts, Depth + 1);
if (Tmp == 1)
return 1; // Early out.
Tmp2 = ComputeNumSignBits(Op.getOperand(1), DemandedElts, Depth + 1);
return std::min(Tmp, Tmp2);
}
case ISD::UMIN:
case ISD::UMAX:
Tmp = ComputeNumSignBits(Op.getOperand(0), DemandedElts, Depth + 1);
if (Tmp == 1)
return 1; // Early out.
Tmp2 = ComputeNumSignBits(Op.getOperand(1), DemandedElts, Depth + 1);
return std::min(Tmp, Tmp2);
case ISD::SADDO:
case ISD::UADDO:
case ISD::SADDO_CARRY:
case ISD::UADDO_CARRY:
case ISD::SSUBO:
case ISD::USUBO:
case ISD::SSUBO_CARRY:
case ISD::USUBO_CARRY:
case ISD::SMULO:
case ISD::UMULO:
if (Op.getResNo() != 1)
break;
// The boolean result conforms to getBooleanContents. Fall through.
// If setcc returns 0/-1, all bits are sign bits.
// We know that we have an integer-based boolean since these operations
// are only available for integer.
if (TLI->getBooleanContents(VT.isVector(), false) ==
TargetLowering::ZeroOrNegativeOneBooleanContent)
return VTBits;
break;
case ISD::SETCC:
case ISD::SETCCCARRY:
case ISD::STRICT_FSETCC:
case ISD::STRICT_FSETCCS: {
unsigned OpNo = Op->isStrictFPOpcode() ? 1 : 0;
// If setcc returns 0/-1, all bits are sign bits.
if (TLI->getBooleanContents(Op.getOperand(OpNo).getValueType()) ==
TargetLowering::ZeroOrNegativeOneBooleanContent)
return VTBits;
break;
}
case ISD::ROTL:
case ISD::ROTR:
Tmp = ComputeNumSignBits(Op.getOperand(0), DemandedElts, Depth + 1);
// If we're rotating an 0/-1 value, then it stays an 0/-1 value.
if (Tmp == VTBits)
return VTBits;
if (ConstantSDNode *C =
isConstOrConstSplat(Op.getOperand(1), DemandedElts)) {
unsigned RotAmt = C->getAPIntValue().urem(VTBits);
// Handle rotate right by N like a rotate left by 32-N.
if (Opcode == ISD::ROTR)
RotAmt = (VTBits - RotAmt) % VTBits;
// If we aren't rotating out all of the known-in sign bits, return the
// number that are left. This handles rotl(sext(x), 1) for example.
if (Tmp > (RotAmt + 1)) return (Tmp - RotAmt);
}
break;
case ISD::ADD:
case ISD::ADDC:
// Add can have at most one carry bit. Thus we know that the output
// is, at worst, one more bit than the inputs.
Tmp = ComputeNumSignBits(Op.getOperand(0), DemandedElts, Depth + 1);
if (Tmp == 1) return 1; // Early out.
// Special case decrementing a value (ADD X, -1):
if (ConstantSDNode *CRHS =
isConstOrConstSplat(Op.getOperand(1), DemandedElts))
if (CRHS->isAllOnes()) {
KnownBits Known =
computeKnownBits(Op.getOperand(0), DemandedElts, Depth + 1);
// If the input is known to be 0 or 1, the output is 0/-1, which is all
// sign bits set.
if ((Known.Zero | 1).isAllOnes())
return VTBits;
// If we are subtracting one from a positive number, there is no carry
// out of the result.
if (Known.isNonNegative())
return Tmp;
}
Tmp2 = ComputeNumSignBits(Op.getOperand(1), DemandedElts, Depth + 1);
if (Tmp2 == 1) return 1; // Early out.
return std::min(Tmp, Tmp2) - 1;
case ISD::SUB:
Tmp2 = ComputeNumSignBits(Op.getOperand(1), DemandedElts, Depth + 1);
if (Tmp2 == 1) return 1; // Early out.
// Handle NEG.
if (ConstantSDNode *CLHS =
isConstOrConstSplat(Op.getOperand(0), DemandedElts))
if (CLHS->isZero()) {
KnownBits Known =
computeKnownBits(Op.getOperand(1), DemandedElts, Depth + 1);
// If the input is known to be 0 or 1, the output is 0/-1, which is all
// sign bits set.
if ((Known.Zero | 1).isAllOnes())
return VTBits;
// If the input is known to be positive (the sign bit is known clear),
// the output of the NEG has the same number of sign bits as the input.
if (Known.isNonNegative())
return Tmp2;
// Otherwise, we treat this like a SUB.
}
// Sub can have at most one carry bit. Thus we know that the output
// is, at worst, one more bit than the inputs.
Tmp = ComputeNumSignBits(Op.getOperand(0), DemandedElts, Depth + 1);
if (Tmp == 1) return 1; // Early out.
return std::min(Tmp, Tmp2) - 1;
case ISD::MUL: {
// The output of the Mul can be at most twice the valid bits in the inputs.
unsigned SignBitsOp0 = ComputeNumSignBits(Op.getOperand(0), Depth + 1);
if (SignBitsOp0 == 1)
break;
unsigned SignBitsOp1 = ComputeNumSignBits(Op.getOperand(1), Depth + 1);
if (SignBitsOp1 == 1)
break;
unsigned OutValidBits =
(VTBits - SignBitsOp0 + 1) + (VTBits - SignBitsOp1 + 1);
return OutValidBits > VTBits ? 1 : VTBits - OutValidBits + 1;
}
case ISD::SREM:
// The sign bit is the LHS's sign bit, except when the result of the
// remainder is zero. The magnitude of the result should be less than or
// equal to the magnitude of the LHS. Therefore, the result should have
// at least as many sign bits as the left hand side.
return ComputeNumSignBits(Op.getOperand(0), DemandedElts, Depth + 1);
case ISD::TRUNCATE: {
// Check if the sign bits of source go down as far as the truncated value.
unsigned NumSrcBits = Op.getOperand(0).getScalarValueSizeInBits();
unsigned NumSrcSignBits = ComputeNumSignBits(Op.getOperand(0), Depth + 1);
if (NumSrcSignBits > (NumSrcBits - VTBits))
return NumSrcSignBits - (NumSrcBits - VTBits);
break;
}
case ISD::EXTRACT_ELEMENT: {
if (VT.isScalableVector())
break;
const int KnownSign = ComputeNumSignBits(Op.getOperand(0), Depth+1);
const int BitWidth = Op.getValueSizeInBits();
const int Items = Op.getOperand(0).getValueSizeInBits() / BitWidth;
// Get reverse index (starting from 1), Op1 value indexes elements from
// little end. Sign starts at big end.
const int rIndex = Items - 1 - Op.getConstantOperandVal(1);
// If the sign portion ends in our element the subtraction gives correct
// result. Otherwise it gives either negative or > bitwidth result
return std::clamp(KnownSign - rIndex * BitWidth, 0, BitWidth);
}
case ISD::INSERT_VECTOR_ELT: {
if (VT.isScalableVector())
break;
// If we know the element index, split the demand between the
// source vector and the inserted element, otherwise assume we need
// the original demanded vector elements and the value.
SDValue InVec = Op.getOperand(0);
SDValue InVal = Op.getOperand(1);
SDValue EltNo = Op.getOperand(2);
bool DemandedVal = true;
APInt DemandedVecElts = DemandedElts;
auto *CEltNo = dyn_cast<ConstantSDNode>(EltNo);
if (CEltNo && CEltNo->getAPIntValue().ult(NumElts)) {
unsigned EltIdx = CEltNo->getZExtValue();
DemandedVal = !!DemandedElts[EltIdx];
DemandedVecElts.clearBit(EltIdx);
}
Tmp = std::numeric_limits<unsigned>::max();
if (DemandedVal) {
// TODO - handle implicit truncation of inserted elements.
if (InVal.getScalarValueSizeInBits() != VTBits)
break;
Tmp2 = ComputeNumSignBits(InVal, Depth + 1);
Tmp = std::min(Tmp, Tmp2);
}
if (!!DemandedVecElts) {
Tmp2 = ComputeNumSignBits(InVec, DemandedVecElts, Depth + 1);
Tmp = std::min(Tmp, Tmp2);
}
assert(Tmp <= VTBits && "Failed to determine minimum sign bits");
return Tmp;
}
case ISD::EXTRACT_VECTOR_ELT: {
assert(!VT.isScalableVector());
SDValue InVec = Op.getOperand(0);
SDValue EltNo = Op.getOperand(1);
EVT VecVT = InVec.getValueType();
// ComputeNumSignBits not yet implemented for scalable vectors.
if (VecVT.isScalableVector())
break;
const unsigned BitWidth = Op.getValueSizeInBits();
const unsigned EltBitWidth = Op.getOperand(0).getScalarValueSizeInBits();
const unsigned NumSrcElts = VecVT.getVectorNumElements();
// If BitWidth > EltBitWidth the value is anyext:ed, and we do not know
// anything about sign bits. But if the sizes match we can derive knowledge
// about sign bits from the vector operand.
if (BitWidth != EltBitWidth)
break;
// If we know the element index, just demand that vector element, else for
// an unknown element index, ignore DemandedElts and demand them all.
APInt DemandedSrcElts = APInt::getAllOnes(NumSrcElts);
auto *ConstEltNo = dyn_cast<ConstantSDNode>(EltNo);
if (ConstEltNo && ConstEltNo->getAPIntValue().ult(NumSrcElts))
DemandedSrcElts =
APInt::getOneBitSet(NumSrcElts, ConstEltNo->getZExtValue());
return ComputeNumSignBits(InVec, DemandedSrcElts, Depth + 1);
}
case ISD::EXTRACT_SUBVECTOR: {
// Offset the demanded elts by the subvector index.
SDValue Src = Op.getOperand(0);
// Bail until we can represent demanded elements for scalable vectors.
if (Src.getValueType().isScalableVector())
break;
uint64_t Idx = Op.getConstantOperandVal(1);
unsigned NumSrcElts = Src.getValueType().getVectorNumElements();
APInt DemandedSrcElts = DemandedElts.zext(NumSrcElts).shl(Idx);
return ComputeNumSignBits(Src, DemandedSrcElts, Depth + 1);
}
case ISD::CONCAT_VECTORS: {
if (VT.isScalableVector())
break;
// Determine the minimum number of sign bits across all demanded
// elts of the input vectors. Early out if the result is already 1.
Tmp = std::numeric_limits<unsigned>::max();
EVT SubVectorVT = Op.getOperand(0).getValueType();
unsigned NumSubVectorElts = SubVectorVT.getVectorNumElements();
unsigned NumSubVectors = Op.getNumOperands();
for (unsigned i = 0; (i < NumSubVectors) && (Tmp > 1); ++i) {
APInt DemandedSub =
DemandedElts.extractBits(NumSubVectorElts, i * NumSubVectorElts);
if (!DemandedSub)
continue;
Tmp2 = ComputeNumSignBits(Op.getOperand(i), DemandedSub, Depth + 1);
Tmp = std::min(Tmp, Tmp2);
}
assert(Tmp <= VTBits && "Failed to determine minimum sign bits");
return Tmp;
}
case ISD::INSERT_SUBVECTOR: {
if (VT.isScalableVector())
break;
// Demand any elements from the subvector and the remainder from the src its
// inserted into.
SDValue Src = Op.getOperand(0);
SDValue Sub = Op.getOperand(1);
uint64_t Idx = Op.getConstantOperandVal(2);
unsigned NumSubElts = Sub.getValueType().getVectorNumElements();
APInt DemandedSubElts = DemandedElts.extractBits(NumSubElts, Idx);
APInt DemandedSrcElts = DemandedElts;
DemandedSrcElts.insertBits(APInt::getZero(NumSubElts), Idx);
Tmp = std::numeric_limits<unsigned>::max();
if (!!DemandedSubElts) {
Tmp = ComputeNumSignBits(Sub, DemandedSubElts, Depth + 1);
if (Tmp == 1)
return 1; // early-out
}
if (!!DemandedSrcElts) {
Tmp2 = ComputeNumSignBits(Src, DemandedSrcElts, Depth + 1);
Tmp = std::min(Tmp, Tmp2);
}
assert(Tmp <= VTBits && "Failed to determine minimum sign bits");
return Tmp;
}
case ISD::LOAD: {
LoadSDNode *LD = cast<LoadSDNode>(Op);
if (const MDNode *Ranges = LD->getRanges()) {
if (DemandedElts != 1)
break;
ConstantRange CR = getConstantRangeFromMetadata(*Ranges);
if (VTBits > CR.getBitWidth()) {
switch (LD->getExtensionType()) {
case ISD::SEXTLOAD:
CR = CR.signExtend(VTBits);
break;
case ISD::ZEXTLOAD:
CR = CR.zeroExtend(VTBits);
break;
default:
break;
}
}
if (VTBits != CR.getBitWidth())
break;
return std::min(CR.getSignedMin().getNumSignBits(),
CR.getSignedMax().getNumSignBits());
}
break;
}
case ISD::ATOMIC_CMP_SWAP:
case ISD::ATOMIC_CMP_SWAP_WITH_SUCCESS:
case ISD::ATOMIC_SWAP:
case ISD::ATOMIC_LOAD_ADD:
case ISD::ATOMIC_LOAD_SUB:
case ISD::ATOMIC_LOAD_AND:
case ISD::ATOMIC_LOAD_CLR:
case ISD::ATOMIC_LOAD_OR:
case ISD::ATOMIC_LOAD_XOR:
case ISD::ATOMIC_LOAD_NAND:
case ISD::ATOMIC_LOAD_MIN:
case ISD::ATOMIC_LOAD_MAX:
case ISD::ATOMIC_LOAD_UMIN:
case ISD::ATOMIC_LOAD_UMAX:
case ISD::ATOMIC_LOAD: {
Tmp = cast<AtomicSDNode>(Op)->getMemoryVT().getScalarSizeInBits();
// If we are looking at the loaded value.
if (Op.getResNo() == 0) {
if (Tmp == VTBits)
return 1; // early-out
if (TLI->getExtendForAtomicOps() == ISD::SIGN_EXTEND)
return VTBits - Tmp + 1;
if (TLI->getExtendForAtomicOps() == ISD::ZERO_EXTEND)
return VTBits - Tmp;
}
break;
}
}
// If we are looking at the loaded value of the SDNode.
if (Op.getResNo() == 0) {
// Handle LOADX separately here. EXTLOAD case will fallthrough.
if (LoadSDNode *LD = dyn_cast<LoadSDNode>(Op)) {
unsigned ExtType = LD->getExtensionType();
switch (ExtType) {
default: break;
case ISD::SEXTLOAD: // e.g. i16->i32 = '17' bits known.
Tmp = LD->getMemoryVT().getScalarSizeInBits();
return VTBits - Tmp + 1;
case ISD::ZEXTLOAD: // e.g. i16->i32 = '16' bits known.
Tmp = LD->getMemoryVT().getScalarSizeInBits();
return VTBits - Tmp;
case ISD::NON_EXTLOAD:
if (const Constant *Cst = TLI->getTargetConstantFromLoad(LD)) {
// We only need to handle vectors - computeKnownBits should handle
// scalar cases.
Type *CstTy = Cst->getType();
if (CstTy->isVectorTy() && !VT.isScalableVector() &&
(NumElts * VTBits) == CstTy->getPrimitiveSizeInBits() &&
VTBits == CstTy->getScalarSizeInBits()) {
Tmp = VTBits;
for (unsigned i = 0; i != NumElts; ++i) {
if (!DemandedElts[i])
continue;
if (Constant *Elt = Cst->getAggregateElement(i)) {
if (auto *CInt = dyn_cast<ConstantInt>(Elt)) {
const APInt &Value = CInt->getValue();
Tmp = std::min(Tmp, Value.getNumSignBits());
continue;
}
if (auto *CFP = dyn_cast<ConstantFP>(Elt)) {
APInt Value = CFP->getValueAPF().bitcastToAPInt();
Tmp = std::min(Tmp, Value.getNumSignBits());
continue;
}
}
// Unknown type. Conservatively assume no bits match sign bit.
return 1;
}
return Tmp;
}
}
break;
}
}
}
// Allow the target to implement this method for its nodes.
if (Opcode >= ISD::BUILTIN_OP_END ||
Opcode == ISD::INTRINSIC_WO_CHAIN ||
Opcode == ISD::INTRINSIC_W_CHAIN ||
Opcode == ISD::INTRINSIC_VOID) {
// TODO: This can probably be removed once target code is audited. This
// is here purely to reduce patch size and review complexity.
if (!VT.isScalableVector()) {
unsigned NumBits =
TLI->ComputeNumSignBitsForTargetNode(Op, DemandedElts, *this, Depth);
if (NumBits > 1)
FirstAnswer = std::max(FirstAnswer, NumBits);
}
}
// Finally, if we can prove that the top bits of the result are 0's or 1's,
// use this information.
KnownBits Known = computeKnownBits(Op, DemandedElts, Depth);
return std::max(FirstAnswer, Known.countMinSignBits());
}
unsigned SelectionDAG::ComputeMaxSignificantBits(SDValue Op,
unsigned Depth) const {
unsigned SignBits = ComputeNumSignBits(Op, Depth);
return Op.getScalarValueSizeInBits() - SignBits + 1;
}
unsigned SelectionDAG::ComputeMaxSignificantBits(SDValue Op,
const APInt &DemandedElts,
unsigned Depth) const {
unsigned SignBits = ComputeNumSignBits(Op, DemandedElts, Depth);
return Op.getScalarValueSizeInBits() - SignBits + 1;
}
bool SelectionDAG::isGuaranteedNotToBeUndefOrPoison(SDValue Op, bool PoisonOnly,
unsigned Depth) const {
// Early out for FREEZE.
if (Op.getOpcode() == ISD::FREEZE)
return true;
// TODO: Assume we don't know anything for now.
EVT VT = Op.getValueType();
if (VT.isScalableVector())
return false;
APInt DemandedElts = VT.isVector()
? APInt::getAllOnes(VT.getVectorNumElements())
: APInt(1, 1);
return isGuaranteedNotToBeUndefOrPoison(Op, DemandedElts, PoisonOnly, Depth);
}
bool SelectionDAG::isGuaranteedNotToBeUndefOrPoison(SDValue Op,
const APInt &DemandedElts,
bool PoisonOnly,
unsigned Depth) const {
unsigned Opcode = Op.getOpcode();
// Early out for FREEZE.
if (Opcode == ISD::FREEZE)
return true;
if (Depth >= MaxRecursionDepth)
return false; // Limit search depth.
if (isIntOrFPConstant(Op))
return true;
switch (Opcode) {
case ISD::VALUETYPE:
case ISD::FrameIndex:
case ISD::TargetFrameIndex:
return true;
case ISD::UNDEF:
return PoisonOnly;
case ISD::BUILD_VECTOR:
// NOTE: BUILD_VECTOR has implicit truncation of wider scalar elements -
// this shouldn't affect the result.
for (unsigned i = 0, e = Op.getNumOperands(); i < e; ++i) {
if (!DemandedElts[i])
continue;
if (!isGuaranteedNotToBeUndefOrPoison(Op.getOperand(i), PoisonOnly,
Depth + 1))
return false;
}
return true;
// TODO: Search for noundef attributes from library functions.
// TODO: Pointers dereferenced by ISD::LOAD/STORE ops are noundef.
default:
// Allow the target to implement this method for its nodes.
if (Opcode >= ISD::BUILTIN_OP_END || Opcode == ISD::INTRINSIC_WO_CHAIN ||
Opcode == ISD::INTRINSIC_W_CHAIN || Opcode == ISD::INTRINSIC_VOID)
return TLI->isGuaranteedNotToBeUndefOrPoisonForTargetNode(
Op, DemandedElts, *this, PoisonOnly, Depth);
break;
}
// If Op can't create undef/poison and none of its operands are undef/poison
// then Op is never undef/poison.
// NOTE: TargetNodes should handle this in themselves in
// isGuaranteedNotToBeUndefOrPoisonForTargetNode.
return !canCreateUndefOrPoison(Op, PoisonOnly, /*ConsiderFlags*/ true,
Depth) &&
all_of(Op->ops(), [&](SDValue V) {
return isGuaranteedNotToBeUndefOrPoison(V, PoisonOnly, Depth + 1);
});
}
bool SelectionDAG::canCreateUndefOrPoison(SDValue Op, bool PoisonOnly,
bool ConsiderFlags,
unsigned Depth) const {
// TODO: Assume we don't know anything for now.
EVT VT = Op.getValueType();
if (VT.isScalableVector())
return true;
APInt DemandedElts = VT.isVector()
? APInt::getAllOnes(VT.getVectorNumElements())
: APInt(1, 1);
return canCreateUndefOrPoison(Op, DemandedElts, PoisonOnly, ConsiderFlags,
Depth);
}
bool SelectionDAG::canCreateUndefOrPoison(SDValue Op, const APInt &DemandedElts,
bool PoisonOnly, bool ConsiderFlags,
unsigned Depth) const {
// TODO: Assume we don't know anything for now.
EVT VT = Op.getValueType();
if (VT.isScalableVector())
return true;
unsigned Opcode = Op.getOpcode();
switch (Opcode) {
case ISD::FREEZE:
case ISD::CONCAT_VECTORS:
case ISD::INSERT_SUBVECTOR:
case ISD::AND:
case ISD::XOR:
case ISD::ROTL:
case ISD::ROTR:
case ISD::FSHL:
case ISD::FSHR:
case ISD::BSWAP:
case ISD::CTPOP:
case ISD::BITREVERSE:
case ISD::PARITY:
case ISD::SIGN_EXTEND:
case ISD::TRUNCATE:
case ISD::SIGN_EXTEND_INREG:
case ISD::SIGN_EXTEND_VECTOR_INREG:
case ISD::ZERO_EXTEND_VECTOR_INREG:
case ISD::BITCAST:
case ISD::BUILD_VECTOR:
case ISD::BUILD_PAIR:
return false;
// Matches hasPoisonGeneratingFlags().
case ISD::ZERO_EXTEND:
return ConsiderFlags && Op->getFlags().hasNonNeg();
case ISD::ADD:
case ISD::SUB:
case ISD::MUL:
// Matches hasPoisonGeneratingFlags().
return ConsiderFlags && (Op->getFlags().hasNoSignedWrap() ||
Op->getFlags().hasNoUnsignedWrap());
case ISD::SHL:
// If the max shift amount isn't in range, then the shift can create poison.
if (!getValidMaximumShiftAmountConstant(Op, DemandedElts))
return true;
// Matches hasPoisonGeneratingFlags().
return ConsiderFlags && (Op->getFlags().hasNoSignedWrap() ||
Op->getFlags().hasNoUnsignedWrap());
// Matches hasPoisonGeneratingFlags().
case ISD::OR:
return ConsiderFlags && Op->getFlags().hasDisjoint();
case ISD::INSERT_VECTOR_ELT:{
// Ensure that the element index is in bounds.
EVT VecVT = Op.getOperand(0).getValueType();
KnownBits KnownIdx = computeKnownBits(Op.getOperand(2), Depth + 1);
return KnownIdx.getMaxValue().uge(VecVT.getVectorMinNumElements());
}
default:
// Allow the target to implement this method for its nodes.
if (Opcode >= ISD::BUILTIN_OP_END || Opcode == ISD::INTRINSIC_WO_CHAIN ||
Opcode == ISD::INTRINSIC_W_CHAIN || Opcode == ISD::INTRINSIC_VOID)
return TLI->canCreateUndefOrPoisonForTargetNode(
Op, DemandedElts, *this, PoisonOnly, ConsiderFlags, Depth);
break;
}
// Be conservative and return true.
return true;
}
bool SelectionDAG::isADDLike(SDValue Op) const {
unsigned Opcode = Op.getOpcode();
if (Opcode == ISD::OR)
return Op->getFlags().hasDisjoint() ||
haveNoCommonBitsSet(Op.getOperand(0), Op.getOperand(1));
if (Opcode == ISD::XOR)
return isMinSignedConstant(Op.getOperand(1));
return false;
}
bool SelectionDAG::isBaseWithConstantOffset(SDValue Op) const {
if ((Op.getOpcode() != ISD::ADD && Op.getOpcode() != ISD::OR) ||
!isa<ConstantSDNode>(Op.getOperand(1)))
return false;
if (Op.getOpcode() == ISD::OR &&
!MaskedValueIsZero(Op.getOperand(0), Op.getConstantOperandAPInt(1)))
return false;
return true;
}
bool SelectionDAG::isKnownNeverNaN(SDValue Op, bool SNaN, unsigned Depth) const {
// If we're told that NaNs won't happen, assume they won't.
if (getTarget().Options.NoNaNsFPMath || Op->getFlags().hasNoNaNs())
return true;
if (Depth >= MaxRecursionDepth)
return false; // Limit search depth.
// If the value is a constant, we can obviously see if it is a NaN or not.
if (const ConstantFPSDNode *C = dyn_cast<ConstantFPSDNode>(Op)) {
return !C->getValueAPF().isNaN() ||
(SNaN && !C->getValueAPF().isSignaling());
}
unsigned Opcode = Op.getOpcode();
switch (Opcode) {
case ISD::FADD:
case ISD::FSUB:
case ISD::FMUL:
case ISD::FDIV:
case ISD::FREM:
case ISD::FSIN:
case ISD::FCOS:
case ISD::FMA:
case ISD::FMAD: {
if (SNaN)
return true;
// TODO: Need isKnownNeverInfinity
return false;
}
case ISD::FCANONICALIZE:
case ISD::FEXP:
case ISD::FEXP2:
case ISD::FEXP10:
case ISD::FTRUNC:
case ISD::FFLOOR:
case ISD::FCEIL:
case ISD::FROUND:
case ISD::FROUNDEVEN:
case ISD::FRINT:
case ISD::LRINT:
case ISD::LLRINT:
case ISD::FNEARBYINT:
case ISD::FLDEXP: {
if (SNaN)
return true;
return isKnownNeverNaN(Op.getOperand(0), SNaN, Depth + 1);
}
case ISD::FABS:
case ISD::FNEG:
case ISD::FCOPYSIGN: {
return isKnownNeverNaN(Op.getOperand(0), SNaN, Depth + 1);
}
case ISD::SELECT:
return isKnownNeverNaN(Op.getOperand(1), SNaN, Depth + 1) &&
isKnownNeverNaN(Op.getOperand(2), SNaN, Depth + 1);
case ISD::FP_EXTEND:
case ISD::FP_ROUND: {
if (SNaN)
return true;
return isKnownNeverNaN(Op.getOperand(0), SNaN, Depth + 1);
}
case ISD::SINT_TO_FP:
case ISD::UINT_TO_FP:
return true;
case ISD::FSQRT: // Need is known positive
case ISD::FLOG:
case ISD::FLOG2:
case ISD::FLOG10:
case ISD::FPOWI:
case ISD::FPOW: {
if (SNaN)
return true;
// TODO: Refine on operand
return false;
}
case ISD::FMINNUM:
case ISD::FMAXNUM: {
// Only one needs to be known not-nan, since it will be returned if the
// other ends up being one.
return isKnownNeverNaN(Op.getOperand(0), SNaN, Depth + 1) ||
isKnownNeverNaN(Op.getOperand(1), SNaN, Depth + 1);
}
case ISD::FMINNUM_IEEE:
case ISD::FMAXNUM_IEEE: {
if (SNaN)
return true;
// This can return a NaN if either operand is an sNaN, or if both operands
// are NaN.
return (isKnownNeverNaN(Op.getOperand(0), false, Depth + 1) &&
isKnownNeverSNaN(Op.getOperand(1), Depth + 1)) ||
(isKnownNeverNaN(Op.getOperand(1), false, Depth + 1) &&
isKnownNeverSNaN(Op.getOperand(0), Depth + 1));
}
case ISD::FMINIMUM:
case ISD::FMAXIMUM: {
// TODO: Does this quiet or return the origina NaN as-is?
return isKnownNeverNaN(Op.getOperand(0), SNaN, Depth + 1) &&
isKnownNeverNaN(Op.getOperand(1), SNaN, Depth + 1);
}
case ISD::EXTRACT_VECTOR_ELT: {
return isKnownNeverNaN(Op.getOperand(0), SNaN, Depth + 1);
}
case ISD::BUILD_VECTOR: {
for (const SDValue &Opnd : Op->ops())
if (!isKnownNeverNaN(Opnd, SNaN, Depth + 1))
return false;
return true;
}
default:
if (Opcode >= ISD::BUILTIN_OP_END ||
Opcode == ISD::INTRINSIC_WO_CHAIN ||
Opcode == ISD::INTRINSIC_W_CHAIN ||
Opcode == ISD::INTRINSIC_VOID) {
return TLI->isKnownNeverNaNForTargetNode(Op, *this, SNaN, Depth);
}
return false;
}
}
bool SelectionDAG::isKnownNeverZeroFloat(SDValue Op) const {
assert(Op.getValueType().isFloatingPoint() &&
"Floating point type expected");
// If the value is a constant, we can obviously see if it is a zero or not.
return ISD::matchUnaryFpPredicate(
Op, [](ConstantFPSDNode *C) { return !C->isZero(); });
}
bool SelectionDAG::isKnownNeverZero(SDValue Op, unsigned Depth) const {
if (Depth >= MaxRecursionDepth)
return false; // Limit search depth.
assert(!Op.getValueType().isFloatingPoint() &&
"Floating point types unsupported - use isKnownNeverZeroFloat");
// If the value is a constant, we can obviously see if it is a zero or not.
if (ISD::matchUnaryPredicate(Op,
[](ConstantSDNode *C) { return !C->isZero(); }))
return true;
// TODO: Recognize more cases here. Most of the cases are also incomplete to
// some degree.
switch (Op.getOpcode()) {
default:
break;
case ISD::OR:
return isKnownNeverZero(Op.getOperand(1), Depth + 1) ||
isKnownNeverZero(Op.getOperand(0), Depth + 1);
case ISD::VSELECT:
case ISD::SELECT:
return isKnownNeverZero(Op.getOperand(1), Depth + 1) &&
isKnownNeverZero(Op.getOperand(2), Depth + 1);
case ISD::SHL: {
if (Op->getFlags().hasNoSignedWrap() || Op->getFlags().hasNoUnsignedWrap())
return isKnownNeverZero(Op.getOperand(0), Depth + 1);
KnownBits ValKnown = computeKnownBits(Op.getOperand(0), Depth + 1);
// 1 << X is never zero.
if (ValKnown.One[0])
return true;
// If max shift cnt of known ones is non-zero, result is non-zero.
APInt MaxCnt = computeKnownBits(Op.getOperand(1), Depth + 1).getMaxValue();
if (MaxCnt.ult(ValKnown.getBitWidth()) &&
!ValKnown.One.shl(MaxCnt).isZero())
return true;
break;
}
case ISD::UADDSAT:
case ISD::UMAX:
return isKnownNeverZero(Op.getOperand(1), Depth + 1) ||
isKnownNeverZero(Op.getOperand(0), Depth + 1);
// TODO for smin/smax: If either operand is known negative/positive
// respectively we don't need the other to be known at all.
case ISD::SMAX:
case ISD::SMIN:
case ISD::UMIN:
return isKnownNeverZero(Op.getOperand(1), Depth + 1) &&
isKnownNeverZero(Op.getOperand(0), Depth + 1);
case ISD::ROTL:
case ISD::ROTR:
case ISD::BITREVERSE:
case ISD::BSWAP:
case ISD::CTPOP:
case ISD::ABS:
return isKnownNeverZero(Op.getOperand(0), Depth + 1);
case ISD::SRA:
case ISD::SRL: {
if (Op->getFlags().hasExact())
return isKnownNeverZero(Op.getOperand(0), Depth + 1);
KnownBits ValKnown = computeKnownBits(Op.getOperand(0), Depth + 1);
if (ValKnown.isNegative())
return true;
// If max shift cnt of known ones is non-zero, result is non-zero.
APInt MaxCnt = computeKnownBits(Op.getOperand(1), Depth + 1).getMaxValue();
if (MaxCnt.ult(ValKnown.getBitWidth()) &&
!ValKnown.One.lshr(MaxCnt).isZero())
return true;
break;
}
case ISD::UDIV:
case ISD::SDIV:
// div exact can only produce a zero if the dividend is zero.
// TODO: For udiv this is also true if Op1 u<= Op0
if (Op->getFlags().hasExact())
return isKnownNeverZero(Op.getOperand(0), Depth + 1);
break;
case ISD::ADD:
if (Op->getFlags().hasNoUnsignedWrap())
if (isKnownNeverZero(Op.getOperand(1), Depth + 1) ||
isKnownNeverZero(Op.getOperand(0), Depth + 1))
return true;
// TODO: There are a lot more cases we can prove for add.
break;
case ISD::SUB: {
if (isNullConstant(Op.getOperand(0)))
return isKnownNeverZero(Op.getOperand(1), Depth + 1);
std::optional<bool> ne =
KnownBits::ne(computeKnownBits(Op.getOperand(0), Depth + 1),
computeKnownBits(Op.getOperand(1), Depth + 1));
return ne && *ne;
}
case ISD::MUL:
if (Op->getFlags().hasNoSignedWrap() || Op->getFlags().hasNoUnsignedWrap())
if (isKnownNeverZero(Op.getOperand(1), Depth + 1) &&
isKnownNeverZero(Op.getOperand(0), Depth + 1))
return true;
break;
case ISD::ZERO_EXTEND:
case ISD::SIGN_EXTEND:
return isKnownNeverZero(Op.getOperand(0), Depth + 1);
}
return computeKnownBits(Op, Depth).isNonZero();
}
bool SelectionDAG::isEqualTo(SDValue A, SDValue B) const {
// Check the obvious case.
if (A == B) return true;
// For negative and positive zero.
if (const ConstantFPSDNode *CA = dyn_cast<ConstantFPSDNode>(A))
if (const ConstantFPSDNode *CB = dyn_cast<ConstantFPSDNode>(B))
if (CA->isZero() && CB->isZero()) return true;
// Otherwise they may not be equal.
return false;
}
// Only bits set in Mask must be negated, other bits may be arbitrary.
SDValue llvm::getBitwiseNotOperand(SDValue V, SDValue Mask, bool AllowUndefs) {
if (isBitwiseNot(V, AllowUndefs))
return V.getOperand(0);
// Handle any_extend (not (truncate X)) pattern, where Mask only sets
// bits in the non-extended part.
ConstantSDNode *MaskC = isConstOrConstSplat(Mask);
if (!MaskC || V.getOpcode() != ISD::ANY_EXTEND)
return SDValue();
SDValue ExtArg = V.getOperand(0);
if (ExtArg.getScalarValueSizeInBits() >=
MaskC->getAPIntValue().getActiveBits() &&
isBitwiseNot(ExtArg, AllowUndefs) &&
ExtArg.getOperand(0).getOpcode() == ISD::TRUNCATE &&
ExtArg.getOperand(0).getOperand(0).getValueType() == V.getValueType())
return ExtArg.getOperand(0).getOperand(0);
return SDValue();
}
static bool haveNoCommonBitsSetCommutative(SDValue A, SDValue B) {
// Match masked merge pattern (X & ~M) op (Y & M)
// Including degenerate case (X & ~M) op M
auto MatchNoCommonBitsPattern = [&](SDValue Not, SDValue Mask,
SDValue Other) {
if (SDValue NotOperand =
getBitwiseNotOperand(Not, Mask, /* AllowUndefs */ true)) {
if (NotOperand->getOpcode() == ISD::ZERO_EXTEND ||
NotOperand->getOpcode() == ISD::TRUNCATE)
NotOperand = NotOperand->getOperand(0);
if (Other == NotOperand)
return true;
if (Other->getOpcode() == ISD::AND)
return NotOperand == Other->getOperand(0) ||
NotOperand == Other->getOperand(1);
}
return false;
};
if (A->getOpcode() == ISD::ZERO_EXTEND || A->getOpcode() == ISD::TRUNCATE)
A = A->getOperand(0);
if (B->getOpcode() == ISD::ZERO_EXTEND || B->getOpcode() == ISD::TRUNCATE)
B = B->getOperand(0);
if (A->getOpcode() == ISD::AND)
return MatchNoCommonBitsPattern(A->getOperand(0), A->getOperand(1), B) ||
MatchNoCommonBitsPattern(A->getOperand(1), A->getOperand(0), B);
return false;
}
// FIXME: unify with llvm::haveNoCommonBitsSet.
bool SelectionDAG::haveNoCommonBitsSet(SDValue A, SDValue B) const {
assert(A.getValueType() == B.getValueType() &&
"Values must have the same type");
if (haveNoCommonBitsSetCommutative(A, B) ||
haveNoCommonBitsSetCommutative(B, A))
return true;
return KnownBits::haveNoCommonBitsSet(computeKnownBits(A),
computeKnownBits(B));
}
static SDValue FoldSTEP_VECTOR(const SDLoc &DL, EVT VT, SDValue Step,
SelectionDAG &DAG) {
if (cast<ConstantSDNode>(Step)->isZero())
return DAG.getConstant(0, DL, VT);
return SDValue();
}
static SDValue FoldBUILD_VECTOR(const SDLoc &DL, EVT VT,
ArrayRef<SDValue> Ops,
SelectionDAG &DAG) {
int NumOps = Ops.size();
assert(NumOps != 0 && "Can't build an empty vector!");
assert(!VT.isScalableVector() &&
"BUILD_VECTOR cannot be used with scalable types");
assert(VT.getVectorNumElements() == (unsigned)NumOps &&
"Incorrect element count in BUILD_VECTOR!");
// BUILD_VECTOR of UNDEFs is UNDEF.
if (llvm::all_of(Ops, [](SDValue Op) { return Op.isUndef(); }))
return DAG.getUNDEF(VT);
// BUILD_VECTOR of seq extract/insert from the same vector + type is Identity.
SDValue IdentitySrc;
bool IsIdentity = true;
for (int i = 0; i != NumOps; ++i) {
if (Ops[i].getOpcode() != ISD::EXTRACT_VECTOR_ELT ||
Ops[i].getOperand(0).getValueType() != VT ||
(IdentitySrc && Ops[i].getOperand(0) != IdentitySrc) ||
!isa<ConstantSDNode>(Ops[i].getOperand(1)) ||
Ops[i].getConstantOperandAPInt(1) != i) {
IsIdentity = false;
break;
}
IdentitySrc = Ops[i].getOperand(0);
}
if (IsIdentity)
return IdentitySrc;
return SDValue();
}
/// Try to simplify vector concatenation to an input value, undef, or build
/// vector.
static SDValue foldCONCAT_VECTORS(const SDLoc &DL, EVT VT,
ArrayRef<SDValue> Ops,
SelectionDAG &DAG) {
assert(!Ops.empty() && "Can't concatenate an empty list of vectors!");
assert(llvm::all_of(Ops,
[Ops](SDValue Op) {
return Ops[0].getValueType() == Op.getValueType();
}) &&
"Concatenation of vectors with inconsistent value types!");
assert((Ops[0].getValueType().getVectorElementCount() * Ops.size()) ==
VT.getVectorElementCount() &&
"Incorrect element count in vector concatenation!");
if (Ops.size() == 1)
return Ops[0];
// Concat of UNDEFs is UNDEF.
if (llvm::all_of(Ops, [](SDValue Op) { return Op.isUndef(); }))
return DAG.getUNDEF(VT);
// Scan the operands and look for extract operations from a single source
// that correspond to insertion at the same location via this concatenation:
// concat (extract X, 0*subvec_elts), (extract X, 1*subvec_elts), ...
SDValue IdentitySrc;
bool IsIdentity = true;
for (unsigned i = 0, e = Ops.size(); i != e; ++i) {
SDValue Op = Ops[i];
unsigned IdentityIndex = i * Op.getValueType().getVectorMinNumElements();
if (Op.getOpcode() != ISD::EXTRACT_SUBVECTOR ||
Op.getOperand(0).getValueType() != VT ||
(IdentitySrc && Op.getOperand(0) != IdentitySrc) ||
Op.getConstantOperandVal(1) != IdentityIndex) {
IsIdentity = false;
break;
}
assert((!IdentitySrc || IdentitySrc == Op.getOperand(0)) &&
"Unexpected identity source vector for concat of extracts");
IdentitySrc = Op.getOperand(0);
}
if (IsIdentity) {
assert(IdentitySrc && "Failed to set source vector of extracts");
return IdentitySrc;
}
// The code below this point is only designed to work for fixed width
// vectors, so we bail out for now.
if (VT.isScalableVector())
return SDValue();
// A CONCAT_VECTOR with all UNDEF/BUILD_VECTOR operands can be
// simplified to one big BUILD_VECTOR.
// FIXME: Add support for SCALAR_TO_VECTOR as well.
EVT SVT = VT.getScalarType();
SmallVector<SDValue, 16> Elts;
for (SDValue Op : Ops) {
EVT OpVT = Op.getValueType();
if (Op.isUndef())
Elts.append(OpVT.getVectorNumElements(), DAG.getUNDEF(SVT));
else if (Op.getOpcode() == ISD::BUILD_VECTOR)
Elts.append(Op->op_begin(), Op->op_end());
else
return SDValue();
}
// BUILD_VECTOR requires all inputs to be of the same type, find the
// maximum type and extend them all.
for (SDValue Op : Elts)
SVT = (SVT.bitsLT(Op.getValueType()) ? Op.getValueType() : SVT);
if (SVT.bitsGT(VT.getScalarType())) {
for (SDValue &Op : Elts) {
if (Op.isUndef())
Op = DAG.getUNDEF(SVT);
else
Op = DAG.getTargetLoweringInfo().isZExtFree(Op.getValueType(), SVT)
? DAG.getZExtOrTrunc(Op, DL, SVT)
: DAG.getSExtOrTrunc(Op, DL, SVT);
}
}
SDValue V = DAG.getBuildVector(VT, DL, Elts);
NewSDValueDbgMsg(V, "New node fold concat vectors: ", &DAG);
return V;
}
/// Gets or creates the specified node.
SDValue SelectionDAG::getNode(unsigned Opcode, const SDLoc &DL, EVT VT) {
FoldingSetNodeID ID;
AddNodeIDNode(ID, Opcode, getVTList(VT), std::nullopt);
void *IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, DL, IP))
return SDValue(E, 0);
auto *N = newSDNode<SDNode>(Opcode, DL.getIROrder(), DL.getDebugLoc(),
getVTList(VT));
CSEMap.InsertNode(N, IP);
InsertNode(N);
SDValue V = SDValue(N, 0);
NewSDValueDbgMsg(V, "Creating new node: ", this);
return V;
}
SDValue SelectionDAG::getNode(unsigned Opcode, const SDLoc &DL, EVT VT,
SDValue N1) {
SDNodeFlags Flags;
if (Inserter)
Flags = Inserter->getFlags();
return getNode(Opcode, DL, VT, N1, Flags);
}
SDValue SelectionDAG::getNode(unsigned Opcode, const SDLoc &DL, EVT VT,
SDValue N1, const SDNodeFlags Flags) {
assert(N1.getOpcode() != ISD::DELETED_NODE && "Operand is DELETED_NODE!");
// Constant fold unary operations with a vector integer or float operand.
switch (Opcode) {
default:
// FIXME: Entirely reasonable to perform folding of other unary
// operations here as the need arises.
break;
case ISD::FNEG:
case ISD::FABS:
case ISD::FCEIL:
case ISD::FTRUNC:
case ISD::FFLOOR:
case ISD::FP_EXTEND:
case ISD::FP_TO_SINT:
case ISD::FP_TO_UINT:
case ISD::FP_TO_FP16:
case ISD::FP_TO_BF16:
case ISD::TRUNCATE:
case ISD::ANY_EXTEND:
case ISD::ZERO_EXTEND:
case ISD::SIGN_EXTEND:
case ISD::UINT_TO_FP:
case ISD::SINT_TO_FP:
case ISD::FP16_TO_FP:
case ISD::BF16_TO_FP:
case ISD::BITCAST:
case ISD::ABS:
case ISD::BITREVERSE:
case ISD::BSWAP:
case ISD::CTLZ:
case ISD::CTLZ_ZERO_UNDEF:
case ISD::CTTZ:
case ISD::CTTZ_ZERO_UNDEF:
case ISD::CTPOP:
case ISD::STEP_VECTOR: {
SDValue Ops = {N1};
if (SDValue Fold = FoldConstantArithmetic(Opcode, DL, VT, Ops))
return Fold;
}
}
unsigned OpOpcode = N1.getNode()->getOpcode();
switch (Opcode) {
case ISD::STEP_VECTOR:
assert(VT.isScalableVector() &&
"STEP_VECTOR can only be used with scalable types");
assert(OpOpcode == ISD::TargetConstant &&
VT.getVectorElementType() == N1.getValueType() &&
"Unexpected step operand");
break;
case ISD::FREEZE:
assert(VT == N1.getValueType() && "Unexpected VT!");
if (isGuaranteedNotToBeUndefOrPoison(N1, /*PoisonOnly*/ false,
/*Depth*/ 1))
return N1;
break;
case ISD::TokenFactor:
case ISD::MERGE_VALUES:
case ISD::CONCAT_VECTORS:
return N1; // Factor, merge or concat of one node? No need.
case ISD::BUILD_VECTOR: {
// Attempt to simplify BUILD_VECTOR.
SDValue Ops[] = {N1};
if (SDValue V = FoldBUILD_VECTOR(DL, VT, Ops, *this))
return V;
break;
}
case ISD::FP_ROUND: llvm_unreachable("Invalid method to make FP_ROUND node");
case ISD::FP_EXTEND:
assert(VT.isFloatingPoint() && N1.getValueType().isFloatingPoint() &&
"Invalid FP cast!");
if (N1.getValueType() == VT) return N1; // noop conversion.
assert((!VT.isVector() || VT.getVectorElementCount() ==
N1.getValueType().getVectorElementCount()) &&
"Vector element count mismatch!");
assert(N1.getValueType().bitsLT(VT) && "Invalid fpext node, dst < src!");
if (N1.isUndef())
return getUNDEF(VT);
break;
case ISD::FP_TO_SINT:
case ISD::FP_TO_UINT:
if (N1.isUndef())
return getUNDEF(VT);
break;
case ISD::SINT_TO_FP:
case ISD::UINT_TO_FP:
// [us]itofp(undef) = 0, because the result value is bounded.
if (N1.isUndef())
return getConstantFP(0.0, DL, VT);
break;
case ISD::SIGN_EXTEND:
assert(VT.isInteger() && N1.getValueType().isInteger() &&
"Invalid SIGN_EXTEND!");
assert(VT.isVector() == N1.getValueType().isVector() &&
"SIGN_EXTEND result type type should be vector iff the operand "
"type is vector!");
if (N1.getValueType() == VT) return N1; // noop extension
assert((!VT.isVector() || VT.getVectorElementCount() ==
N1.getValueType().getVectorElementCount()) &&
"Vector element count mismatch!");
assert(N1.getValueType().bitsLT(VT) && "Invalid sext node, dst < src!");
if (OpOpcode == ISD::SIGN_EXTEND || OpOpcode == ISD::ZERO_EXTEND) {
SDNodeFlags Flags;
if (OpOpcode == ISD::ZERO_EXTEND)
Flags.setNonNeg(N1->getFlags().hasNonNeg());
return getNode(OpOpcode, DL, VT, N1.getOperand(0), Flags);
}
if (OpOpcode == ISD::UNDEF)
// sext(undef) = 0, because the top bits will all be the same.
return getConstant(0, DL, VT);
break;
case ISD::ZERO_EXTEND:
assert(VT.isInteger() && N1.getValueType().isInteger() &&
"Invalid ZERO_EXTEND!");
assert(VT.isVector() == N1.getValueType().isVector() &&
"ZERO_EXTEND result type type should be vector iff the operand "
"type is vector!");
if (N1.getValueType() == VT) return N1; // noop extension
assert((!VT.isVector() || VT.getVectorElementCount() ==
N1.getValueType().getVectorElementCount()) &&
"Vector element count mismatch!");
assert(N1.getValueType().bitsLT(VT) && "Invalid zext node, dst < src!");
if (OpOpcode == ISD::ZERO_EXTEND) { // (zext (zext x)) -> (zext x)
SDNodeFlags Flags;
Flags.setNonNeg(N1->getFlags().hasNonNeg());
return getNode(ISD::ZERO_EXTEND, DL, VT, N1.getOperand(0), Flags);
}
if (OpOpcode == ISD::UNDEF)
// zext(undef) = 0, because the top bits will be zero.
return getConstant(0, DL, VT);
// Skip unnecessary zext_inreg pattern:
// (zext (trunc x)) -> x iff the upper bits are known zero.
// TODO: Remove (zext (trunc (and x, c))) exception which some targets
// use to recognise zext_inreg patterns.
if (OpOpcode == ISD::TRUNCATE) {
SDValue OpOp = N1.getOperand(0);
if (OpOp.getValueType() == VT) {
if (OpOp.getOpcode() != ISD::AND) {
APInt HiBits = APInt::getBitsSetFrom(VT.getScalarSizeInBits(),
N1.getScalarValueSizeInBits());
if (MaskedValueIsZero(OpOp, HiBits)) {
transferDbgValues(N1, OpOp);
return OpOp;
}
}
}
}
break;
case ISD::ANY_EXTEND:
assert(VT.isInteger() && N1.getValueType().isInteger() &&
"Invalid ANY_EXTEND!");
assert(VT.isVector() == N1.getValueType().isVector() &&
"ANY_EXTEND result type type should be vector iff the operand "
"type is vector!");
if (N1.getValueType() == VT) return N1; // noop extension
assert((!VT.isVector() || VT.getVectorElementCount() ==
N1.getValueType().getVectorElementCount()) &&
"Vector element count mismatch!");
assert(N1.getValueType().bitsLT(VT) && "Invalid anyext node, dst < src!");
if (OpOpcode == ISD::ZERO_EXTEND || OpOpcode == ISD::SIGN_EXTEND ||
OpOpcode == ISD::ANY_EXTEND) {
SDNodeFlags Flags;
if (OpOpcode == ISD::ZERO_EXTEND)
Flags.setNonNeg(N1->getFlags().hasNonNeg());
// (ext (zext x)) -> (zext x) and (ext (sext x)) -> (sext x)
return getNode(OpOpcode, DL, VT, N1.getOperand(0), Flags);
}
if (OpOpcode == ISD::UNDEF)
return getUNDEF(VT);
// (ext (trunc x)) -> x
if (OpOpcode == ISD::TRUNCATE) {
SDValue OpOp = N1.getOperand(0);
if (OpOp.getValueType() == VT) {
transferDbgValues(N1, OpOp);
return OpOp;
}
}
break;
case ISD::TRUNCATE:
assert(VT.isInteger() && N1.getValueType().isInteger() &&
"Invalid TRUNCATE!");
assert(VT.isVector() == N1.getValueType().isVector() &&
"TRUNCATE result type type should be vector iff the operand "
"type is vector!");
if (N1.getValueType() == VT) return N1; // noop truncate
assert((!VT.isVector() || VT.getVectorElementCount() ==
N1.getValueType().getVectorElementCount()) &&
"Vector element count mismatch!");
assert(N1.getValueType().bitsGT(VT) && "Invalid truncate node, src < dst!");
if (OpOpcode == ISD::TRUNCATE)
return getNode(ISD::TRUNCATE, DL, VT, N1.getOperand(0));
if (OpOpcode == ISD::ZERO_EXTEND || OpOpcode == ISD::SIGN_EXTEND ||
OpOpcode == ISD::ANY_EXTEND) {
// If the source is smaller than the dest, we still need an extend.
if (N1.getOperand(0).getValueType().getScalarType().bitsLT(
VT.getScalarType()))
return getNode(OpOpcode, DL, VT, N1.getOperand(0));
if (N1.getOperand(0).getValueType().bitsGT(VT))
return getNode(ISD::TRUNCATE, DL, VT, N1.getOperand(0));
return N1.getOperand(0);
}
if (OpOpcode == ISD::UNDEF)
return getUNDEF(VT);
if (OpOpcode == ISD::VSCALE && !NewNodesMustHaveLegalTypes)
return getVScale(DL, VT,
N1.getConstantOperandAPInt(0).trunc(VT.getSizeInBits()));
break;
case ISD::ANY_EXTEND_VECTOR_INREG:
case ISD::ZERO_EXTEND_VECTOR_INREG:
case ISD::SIGN_EXTEND_VECTOR_INREG:
assert(VT.isVector() && "This DAG node is restricted to vector types.");
assert(N1.getValueType().bitsLE(VT) &&
"The input must be the same size or smaller than the result.");
assert(VT.getVectorMinNumElements() <
N1.getValueType().getVectorMinNumElements() &&
"The destination vector type must have fewer lanes than the input.");
break;
case ISD::ABS:
assert(VT.isInteger() && VT == N1.getValueType() && "Invalid ABS!");
if (OpOpcode == ISD::UNDEF)
return getConstant(0, DL, VT);
break;
case ISD::BSWAP:
assert(VT.isInteger() && VT == N1.getValueType() && "Invalid BSWAP!");
assert((VT.getScalarSizeInBits() % 16 == 0) &&
"BSWAP types must be a multiple of 16 bits!");
if (OpOpcode == ISD::UNDEF)
return getUNDEF(VT);
// bswap(bswap(X)) -> X.
if (OpOpcode == ISD::BSWAP)
return N1.getOperand(0);
break;
case ISD::BITREVERSE:
assert(VT.isInteger() && VT == N1.getValueType() && "Invalid BITREVERSE!");
if (OpOpcode == ISD::UNDEF)
return getUNDEF(VT);
break;
case ISD::BITCAST:
assert(VT.getSizeInBits() == N1.getValueSizeInBits() &&
"Cannot BITCAST between types of different sizes!");
if (VT == N1.getValueType()) return N1; // noop conversion.
if (OpOpcode == ISD::BITCAST) // bitconv(bitconv(x)) -> bitconv(x)
return getNode(ISD::BITCAST, DL, VT, N1.getOperand(0));
if (OpOpcode == ISD::UNDEF)
return getUNDEF(VT);
break;
case ISD::SCALAR_TO_VECTOR:
assert(VT.isVector() && !N1.getValueType().isVector() &&
(VT.getVectorElementType() == N1.getValueType() ||
(VT.getVectorElementType().isInteger() &&
N1.getValueType().isInteger() &&
VT.getVectorElementType().bitsLE(N1.getValueType()))) &&
"Illegal SCALAR_TO_VECTOR node!");
if (OpOpcode == ISD::UNDEF)
return getUNDEF(VT);
// scalar_to_vector(extract_vector_elt V, 0) -> V, top bits are undefined.
if (OpOpcode == ISD::EXTRACT_VECTOR_ELT &&
isa<ConstantSDNode>(N1.getOperand(1)) &&
N1.getConstantOperandVal(1) == 0 &&
N1.getOperand(0).getValueType() == VT)
return N1.getOperand(0);
break;
case ISD::FNEG:
// Negation of an unknown bag of bits is still completely undefined.
if (OpOpcode == ISD::UNDEF)
return getUNDEF(VT);
if (OpOpcode == ISD::FNEG) // --X -> X
return N1.getOperand(0);
break;
case ISD::FABS:
if (OpOpcode == ISD::FNEG) // abs(-X) -> abs(X)
return getNode(ISD::FABS, DL, VT, N1.getOperand(0));
break;
case ISD::VSCALE:
assert(VT == N1.getValueType() && "Unexpected VT!");
break;
case ISD::CTPOP:
if (N1.getValueType().getScalarType() == MVT::i1)
return N1;
break;
case ISD::CTLZ:
case ISD::CTTZ:
if (N1.getValueType().getScalarType() == MVT::i1)
return getNOT(DL, N1, N1.getValueType());
break;
case ISD::VECREDUCE_ADD:
if (N1.getValueType().getScalarType() == MVT::i1)
return getNode(ISD::VECREDUCE_XOR, DL, VT, N1);
break;
case ISD::VECREDUCE_SMIN:
case ISD::VECREDUCE_UMAX:
if (N1.getValueType().getScalarType() == MVT::i1)
return getNode(ISD::VECREDUCE_OR, DL, VT, N1);
break;
case ISD::VECREDUCE_SMAX:
case ISD::VECREDUCE_UMIN:
if (N1.getValueType().getScalarType() == MVT::i1)
return getNode(ISD::VECREDUCE_AND, DL, VT, N1);
break;
}
SDNode *N;
SDVTList VTs = getVTList(VT);
SDValue Ops[] = {N1};
if (VT != MVT::Glue) { // Don't CSE glue producing nodes
FoldingSetNodeID ID;
AddNodeIDNode(ID, Opcode, VTs, Ops);
void *IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, DL, IP)) {
E->intersectFlagsWith(Flags);
return SDValue(E, 0);
}
N = newSDNode<SDNode>(Opcode, DL.getIROrder(), DL.getDebugLoc(), VTs);
N->setFlags(Flags);
createOperands(N, Ops);
CSEMap.InsertNode(N, IP);
} else {
N = newSDNode<SDNode>(Opcode, DL.getIROrder(), DL.getDebugLoc(), VTs);
createOperands(N, Ops);
}
InsertNode(N);
SDValue V = SDValue(N, 0);
NewSDValueDbgMsg(V, "Creating new node: ", this);
return V;
}
static std::optional<APInt> FoldValue(unsigned Opcode, const APInt &C1,
const APInt &C2) {
switch (Opcode) {
case ISD::ADD: return C1 + C2;
case ISD::SUB: return C1 - C2;
case ISD::MUL: return C1 * C2;
case ISD::AND: return C1 & C2;
case ISD::OR: return C1 | C2;
case ISD::XOR: return C1 ^ C2;
case ISD::SHL: return C1 << C2;
case ISD::SRL: return C1.lshr(C2);
case ISD::SRA: return C1.ashr(C2);
case ISD::ROTL: return C1.rotl(C2);
case ISD::ROTR: return C1.rotr(C2);
case ISD::SMIN: return C1.sle(C2) ? C1 : C2;
case ISD::SMAX: return C1.sge(C2) ? C1 : C2;
case ISD::UMIN: return C1.ule(C2) ? C1 : C2;
case ISD::UMAX: return C1.uge(C2) ? C1 : C2;
case ISD::SADDSAT: return C1.sadd_sat(C2);
case ISD::UADDSAT: return C1.uadd_sat(C2);
case ISD::SSUBSAT: return C1.ssub_sat(C2);
case ISD::USUBSAT: return C1.usub_sat(C2);
case ISD::SSHLSAT: return C1.sshl_sat(C2);
case ISD::USHLSAT: return C1.ushl_sat(C2);
case ISD::UDIV:
if (!C2.getBoolValue())
break;
return C1.udiv(C2);
case ISD::UREM:
if (!C2.getBoolValue())
break;
return C1.urem(C2);
case ISD::SDIV:
if (!C2.getBoolValue())
break;
return C1.sdiv(C2);
case ISD::SREM:
if (!C2.getBoolValue())
break;
return C1.srem(C2);
case ISD::MULHS: {
unsigned FullWidth = C1.getBitWidth() * 2;
APInt C1Ext = C1.sext(FullWidth);
APInt C2Ext = C2.sext(FullWidth);
return (C1Ext * C2Ext).extractBits(C1.getBitWidth(), C1.getBitWidth());
}
case ISD::MULHU: {
unsigned FullWidth = C1.getBitWidth() * 2;
APInt C1Ext = C1.zext(FullWidth);
APInt C2Ext = C2.zext(FullWidth);
return (C1Ext * C2Ext).extractBits(C1.getBitWidth(), C1.getBitWidth());
}
case ISD::AVGFLOORS: {
unsigned FullWidth = C1.getBitWidth() + 1;
APInt C1Ext = C1.sext(FullWidth);
APInt C2Ext = C2.sext(FullWidth);
return (C1Ext + C2Ext).extractBits(C1.getBitWidth(), 1);
}
case ISD::AVGFLOORU: {
unsigned FullWidth = C1.getBitWidth() + 1;
APInt C1Ext = C1.zext(FullWidth);
APInt C2Ext = C2.zext(FullWidth);
return (C1Ext + C2Ext).extractBits(C1.getBitWidth(), 1);
}
case ISD::AVGCEILS: {
unsigned FullWidth = C1.getBitWidth() + 1;
APInt C1Ext = C1.sext(FullWidth);
APInt C2Ext = C2.sext(FullWidth);
return (C1Ext + C2Ext + 1).extractBits(C1.getBitWidth(), 1);
}
case ISD::AVGCEILU: {
unsigned FullWidth = C1.getBitWidth() + 1;
APInt C1Ext = C1.zext(FullWidth);
APInt C2Ext = C2.zext(FullWidth);
return (C1Ext + C2Ext + 1).extractBits(C1.getBitWidth(), 1);
}
case ISD::ABDS:
return APIntOps::smax(C1, C2) - APIntOps::smin(C1, C2);
case ISD::ABDU:
return APIntOps::umax(C1, C2) - APIntOps::umin(C1, C2);
}
return std::nullopt;
}
// Handle constant folding with UNDEF.
// TODO: Handle more cases.
static std::optional<APInt> FoldValueWithUndef(unsigned Opcode, const APInt &C1,
bool IsUndef1, const APInt &C2,
bool IsUndef2) {
if (!(IsUndef1 || IsUndef2))
return FoldValue(Opcode, C1, C2);
// Fold and(x, undef) -> 0
// Fold mul(x, undef) -> 0
if (Opcode == ISD::AND || Opcode == ISD::MUL)
return APInt::getZero(C1.getBitWidth());
return std::nullopt;
}
SDValue SelectionDAG::FoldSymbolOffset(unsigned Opcode, EVT VT,
const GlobalAddressSDNode *GA,
const SDNode *N2) {
if (GA->getOpcode() != ISD::GlobalAddress)
return SDValue();
if (!TLI->isOffsetFoldingLegal(GA))
return SDValue();
auto *C2 = dyn_cast<ConstantSDNode>(N2);
if (!C2)
return SDValue();
int64_t Offset = C2->getSExtValue();
switch (Opcode) {
case ISD::ADD: break;
case ISD::SUB: Offset = -uint64_t(Offset); break;
default: return SDValue();
}
return getGlobalAddress(GA->getGlobal(), SDLoc(C2), VT,
GA->getOffset() + uint64_t(Offset));
}
bool SelectionDAG::isUndef(unsigned Opcode, ArrayRef<SDValue> Ops) {
switch (Opcode) {
case ISD::SDIV:
case ISD::UDIV:
case ISD::SREM:
case ISD::UREM: {
// If a divisor is zero/undef or any element of a divisor vector is
// zero/undef, the whole op is undef.
assert(Ops.size() == 2 && "Div/rem should have 2 operands");
SDValue Divisor = Ops[1];
if (Divisor.isUndef() || isNullConstant(Divisor))
return true;
return ISD::isBuildVectorOfConstantSDNodes(Divisor.getNode()) &&
llvm::any_of(Divisor->op_values(),
[](SDValue V) { return V.isUndef() ||
isNullConstant(V); });
// TODO: Handle signed overflow.
}
// TODO: Handle oversized shifts.
default:
return false;
}
}
SDValue SelectionDAG::FoldConstantArithmetic(unsigned Opcode, const SDLoc &DL,
EVT VT, ArrayRef<SDValue> Ops) {
// If the opcode is a target-specific ISD node, there's nothing we can
// do here and the operand rules may not line up with the below, so
// bail early.
// We can't create a scalar CONCAT_VECTORS so skip it. It will break
// for concats involving SPLAT_VECTOR. Concats of BUILD_VECTORS are handled by
// foldCONCAT_VECTORS in getNode before this is called.
if (Opcode >= ISD::BUILTIN_OP_END || Opcode == ISD::CONCAT_VECTORS)
return SDValue();
unsigned NumOps = Ops.size();
if (NumOps == 0)
return SDValue();
if (isUndef(Opcode, Ops))
return getUNDEF(VT);
// Handle unary special cases.
if (NumOps == 1) {
SDValue N1 = Ops[0];
// Constant fold unary operations with an integer constant operand. Even
// opaque constant will be folded, because the folding of unary operations
// doesn't create new constants with different values. Nevertheless, the
// opaque flag is preserved during folding to prevent future folding with
// other constants.
if (auto *C = dyn_cast<ConstantSDNode>(N1)) {
const APInt &Val = C->getAPIntValue();
switch (Opcode) {
case ISD::SIGN_EXTEND:
return getConstant(Val.sextOrTrunc(VT.getSizeInBits()), DL, VT,
C->isTargetOpcode(), C->isOpaque());
case ISD::TRUNCATE:
if (C->isOpaque())
break;
[[fallthrough]];
case ISD::ZERO_EXTEND:
return getConstant(Val.zextOrTrunc(VT.getSizeInBits()), DL, VT,
C->isTargetOpcode(), C->isOpaque());
case ISD::ANY_EXTEND:
// Some targets like RISCV prefer to sign extend some types.
if (TLI->isSExtCheaperThanZExt(N1.getValueType(), VT))
return getConstant(Val.sextOrTrunc(VT.getSizeInBits()), DL, VT,
C->isTargetOpcode(), C->isOpaque());
return getConstant(Val.zextOrTrunc(VT.getSizeInBits()), DL, VT,
C->isTargetOpcode(), C->isOpaque());
case ISD::ABS:
return getConstant(Val.abs(), DL, VT, C->isTargetOpcode(),
C->isOpaque());
case ISD::BITREVERSE:
return getConstant(Val.reverseBits(), DL, VT, C->isTargetOpcode(),
C->isOpaque());
case ISD::BSWAP:
return getConstant(Val.byteSwap(), DL, VT, C->isTargetOpcode(),
C->isOpaque());
case ISD::CTPOP:
return getConstant(Val.popcount(), DL, VT, C->isTargetOpcode(),
C->isOpaque());
case ISD::CTLZ:
case ISD::CTLZ_ZERO_UNDEF:
return getConstant(Val.countl_zero(), DL, VT, C->isTargetOpcode(),
C->isOpaque());
case ISD::CTTZ:
case ISD::CTTZ_ZERO_UNDEF:
return getConstant(Val.countr_zero(), DL, VT, C->isTargetOpcode(),
C->isOpaque());
case ISD::UINT_TO_FP:
case ISD::SINT_TO_FP: {
APFloat apf(EVTToAPFloatSemantics(VT),
APInt::getZero(VT.getSizeInBits()));
(void)apf.convertFromAPInt(Val, Opcode == ISD::SINT_TO_FP,
APFloat::rmNearestTiesToEven);
return getConstantFP(apf, DL, VT);
}
case ISD::FP16_TO_FP:
case ISD::BF16_TO_FP: {
bool Ignored;
APFloat FPV(Opcode == ISD::FP16_TO_FP ? APFloat::IEEEhalf()
: APFloat::BFloat(),
(Val.getBitWidth() == 16) ? Val : Val.trunc(16));
// This can return overflow, underflow, or inexact; we don't care.
// FIXME need to be more flexible about rounding mode.
(void)FPV.convert(EVTToAPFloatSemantics(VT),
APFloat::rmNearestTiesToEven, &Ignored);
return getConstantFP(FPV, DL, VT);
}
case ISD::STEP_VECTOR:
if (SDValue V = FoldSTEP_VECTOR(DL, VT, N1, *this))
return V;
break;
case ISD::BITCAST:
if (VT == MVT::f16 && C->getValueType(0) == MVT::i16)
return getConstantFP(APFloat(APFloat::IEEEhalf(), Val), DL, VT);
if (VT == MVT::f32 && C->getValueType(0) == MVT::i32)
return getConstantFP(APFloat(APFloat::IEEEsingle(), Val), DL, VT);
if (VT == MVT::f64 && C->getValueType(0) == MVT::i64)
return getConstantFP(APFloat(APFloat::IEEEdouble(), Val), DL, VT);
if (VT == MVT::f128 && C->getValueType(0) == MVT::i128)
return getConstantFP(APFloat(APFloat::IEEEquad(), Val), DL, VT);
break;
}
}
// Constant fold unary operations with a floating point constant operand.
if (auto *C = dyn_cast<ConstantFPSDNode>(N1)) {
APFloat V = C->getValueAPF(); // make copy
switch (Opcode) {
case ISD::FNEG:
V.changeSign();
return getConstantFP(V, DL, VT);
case ISD::FABS:
V.clearSign();
return getConstantFP(V, DL, VT);
case ISD::FCEIL: {
APFloat::opStatus fs = V.roundToIntegral(APFloat::rmTowardPositive);
if (fs == APFloat::opOK || fs == APFloat::opInexact)
return getConstantFP(V, DL, VT);
return SDValue();
}
case ISD::FTRUNC: {
APFloat::opStatus fs = V.roundToIntegral(APFloat::rmTowardZero);
if (fs == APFloat::opOK || fs == APFloat::opInexact)
return getConstantFP(V, DL, VT);
return SDValue();
}
case ISD::FFLOOR: {
APFloat::opStatus fs = V.roundToIntegral(APFloat::rmTowardNegative);
if (fs == APFloat::opOK || fs == APFloat::opInexact)
return getConstantFP(V, DL, VT);
return SDValue();
}
case ISD::FP_EXTEND: {
bool ignored;
// This can return overflow, underflow, or inexact; we don't care.
// FIXME need to be more flexible about rounding mode.
(void)V.convert(EVTToAPFloatSemantics(VT), APFloat::rmNearestTiesToEven,
&ignored);
return getConstantFP(V, DL, VT);
}
case ISD::FP_TO_SINT:
case ISD::FP_TO_UINT: {
bool ignored;
APSInt IntVal(VT.getSizeInBits(), Opcode == ISD::FP_TO_UINT);
// FIXME need to be more flexible about rounding mode.
APFloat::opStatus s =
V.convertToInteger(IntVal, APFloat::rmTowardZero, &ignored);
if (s == APFloat::opInvalidOp) // inexact is OK, in fact usual
break;
return getConstant(IntVal, DL, VT);
}
case ISD::FP_TO_FP16:
case ISD::FP_TO_BF16: {
bool Ignored;
// This can return overflow, underflow, or inexact; we don't care.
// FIXME need to be more flexible about rounding mode.
(void)V.convert(Opcode == ISD::FP_TO_FP16 ? APFloat::IEEEhalf()
: APFloat::BFloat(),
APFloat::rmNearestTiesToEven, &Ignored);
return getConstant(V.bitcastToAPInt().getZExtValue(), DL, VT);
}
case ISD::BITCAST:
if (VT == MVT::i16 && C->getValueType(0) == MVT::f16)
return getConstant((uint16_t)V.bitcastToAPInt().getZExtValue(), DL,
VT);
if (VT == MVT::i16 && C->getValueType(0) == MVT::bf16)
return getConstant((uint16_t)V.bitcastToAPInt().getZExtValue(), DL,
VT);
if (VT == MVT::i32 && C->getValueType(0) == MVT::f32)
return getConstant((uint32_t)V.bitcastToAPInt().getZExtValue(), DL,
VT);
if (VT == MVT::i64 && C->getValueType(0) == MVT::f64)
return getConstant(V.bitcastToAPInt().getZExtValue(), DL, VT);
break;
}
}
// Early-out if we failed to constant fold a bitcast.
if (Opcode == ISD::BITCAST)
return SDValue();
}
// Handle binops special cases.
if (NumOps == 2) {
if (SDValue CFP = foldConstantFPMath(Opcode, DL, VT, Ops))
return CFP;
if (auto *C1 = dyn_cast<ConstantSDNode>(Ops[0])) {
if (auto *C2 = dyn_cast<ConstantSDNode>(Ops[1])) {
if (C1->isOpaque() || C2->isOpaque())
return SDValue();
std::optional<APInt> FoldAttempt =
FoldValue(Opcode, C1->getAPIntValue(), C2->getAPIntValue());
if (!FoldAttempt)
return SDValue();
SDValue Folded = getConstant(*FoldAttempt, DL, VT);
assert((!Folded || !VT.isVector()) &&
"Can't fold vectors ops with scalar operands");
return Folded;
}
}
// fold (add Sym, c) -> Sym+c
if (GlobalAddressSDNode *GA = dyn_cast<GlobalAddressSDNode>(Ops[0]))
return FoldSymbolOffset(Opcode, VT, GA, Ops[1].getNode());
if (TLI->isCommutativeBinOp(Opcode))
if (GlobalAddressSDNode *GA = dyn_cast<GlobalAddressSDNode>(Ops[1]))
return FoldSymbolOffset(Opcode, VT, GA, Ops[0].getNode());
}
// This is for vector folding only from here on.
if (!VT.isVector())
return SDValue();
ElementCount NumElts = VT.getVectorElementCount();
// See if we can fold through bitcasted integer ops.
if (NumOps == 2 && VT.isFixedLengthVector() && VT.isInteger() &&
Ops[0].getValueType() == VT && Ops[1].getValueType() == VT &&
Ops[0].getOpcode() == ISD::BITCAST &&
Ops[1].getOpcode() == ISD::BITCAST) {
SDValue N1 = peekThroughBitcasts(Ops[0]);
SDValue N2 = peekThroughBitcasts(Ops[1]);
auto *BV1 = dyn_cast<BuildVectorSDNode>(N1);
auto *BV2 = dyn_cast<BuildVectorSDNode>(N2);
EVT BVVT = N1.getValueType();
if (BV1 && BV2 && BVVT.isInteger() && BVVT == N2.getValueType()) {
bool IsLE = getDataLayout().isLittleEndian();
unsigned EltBits = VT.getScalarSizeInBits();
SmallVector<APInt> RawBits1, RawBits2;
BitVector UndefElts1, UndefElts2;
if (BV1->getConstantRawBits(IsLE, EltBits, RawBits1, UndefElts1) &&
BV2->getConstantRawBits(IsLE, EltBits, RawBits2, UndefElts2)) {
SmallVector<APInt> RawBits;
for (unsigned I = 0, E = NumElts.getFixedValue(); I != E; ++I) {
std::optional<APInt> Fold = FoldValueWithUndef(
Opcode, RawBits1[I], UndefElts1[I], RawBits2[I], UndefElts2[I]);
if (!Fold)
break;
RawBits.push_back(*Fold);
}
if (RawBits.size() == NumElts.getFixedValue()) {
// We have constant folded, but we need to cast this again back to
// the original (possibly legalized) type.
SmallVector<APInt> DstBits;
BitVector DstUndefs;
BuildVectorSDNode::recastRawBits(IsLE, BVVT.getScalarSizeInBits(),
DstBits, RawBits, DstUndefs,
BitVector(RawBits.size(), false));
EVT BVEltVT = BV1->getOperand(0).getValueType();
unsigned BVEltBits = BVEltVT.getSizeInBits();
SmallVector<SDValue> Ops(DstBits.size(), getUNDEF(BVEltVT));
for (unsigned I = 0, E = DstBits.size(); I != E; ++I) {
if (DstUndefs[I])
continue;
Ops[I] = getConstant(DstBits[I].sext(BVEltBits), DL, BVEltVT);
}
return getBitcast(VT, getBuildVector(BVVT, DL, Ops));
}
}
}
}
// Fold (mul step_vector(C0), C1) to (step_vector(C0 * C1)).
// (shl step_vector(C0), C1) -> (step_vector(C0 << C1))
if ((Opcode == ISD::MUL || Opcode == ISD::SHL) &&
Ops[0].getOpcode() == ISD::STEP_VECTOR) {
APInt RHSVal;
if (ISD::isConstantSplatVector(Ops[1].getNode(), RHSVal)) {
APInt NewStep = Opcode == ISD::MUL
? Ops[0].getConstantOperandAPInt(0) * RHSVal
: Ops[0].getConstantOperandAPInt(0) << RHSVal;
return getStepVector(DL, VT, NewStep);
}
}
auto IsScalarOrSameVectorSize = [NumElts](const SDValue &Op) {
return !Op.getValueType().isVector() ||
Op.getValueType().getVectorElementCount() == NumElts;
};
auto IsBuildVectorSplatVectorOrUndef = [](const SDValue &Op) {
return Op.isUndef() || Op.getOpcode() == ISD::CONDCODE ||
Op.getOpcode() == ISD::BUILD_VECTOR ||
Op.getOpcode() == ISD::SPLAT_VECTOR;
};
// All operands must be vector types with the same number of elements as
// the result type and must be either UNDEF or a build/splat vector
// or UNDEF scalars.
if (!llvm::all_of(Ops, IsBuildVectorSplatVectorOrUndef) ||
!llvm::all_of(Ops, IsScalarOrSameVectorSize))
return SDValue();
// If we are comparing vectors, then the result needs to be a i1 boolean that
// is then extended back to the legal result type depending on how booleans
// are represented.
EVT SVT = (Opcode == ISD::SETCC ? MVT::i1 : VT.getScalarType());
ISD::NodeType ExtendCode =
(Opcode == ISD::SETCC && SVT != VT.getScalarType())
? TargetLowering::getExtendForContent(TLI->getBooleanContents(VT))
: ISD::SIGN_EXTEND;
// Find legal integer scalar type for constant promotion and
// ensure that its scalar size is at least as large as source.
EVT LegalSVT = VT.getScalarType();
if (NewNodesMustHaveLegalTypes && LegalSVT.isInteger()) {
LegalSVT = TLI->getTypeToTransformTo(*getContext(), LegalSVT);
if (LegalSVT.bitsLT(VT.getScalarType()))
return SDValue();
}
// For scalable vector types we know we're dealing with SPLAT_VECTORs. We
// only have one operand to check. For fixed-length vector types we may have
// a combination of BUILD_VECTOR and SPLAT_VECTOR.
unsigned NumVectorElts = NumElts.isScalable() ? 1 : NumElts.getFixedValue();
// Constant fold each scalar lane separately.
SmallVector<SDValue, 4> ScalarResults;
for (unsigned I = 0; I != NumVectorElts; I++) {
SmallVector<SDValue, 4> ScalarOps;
for (SDValue Op : Ops) {
EVT InSVT = Op.getValueType().getScalarType();
if (Op.getOpcode() != ISD::BUILD_VECTOR &&
Op.getOpcode() != ISD::SPLAT_VECTOR) {
if (Op.isUndef())
ScalarOps.push_back(getUNDEF(InSVT));
else
ScalarOps.push_back(Op);
continue;
}
SDValue ScalarOp =
Op.getOperand(Op.getOpcode() == ISD::SPLAT_VECTOR ? 0 : I);
EVT ScalarVT = ScalarOp.getValueType();
// Build vector (integer) scalar operands may need implicit
// truncation - do this before constant folding.
if (ScalarVT.isInteger() && ScalarVT.bitsGT(InSVT)) {
// Don't create illegally-typed nodes unless they're constants or undef
// - if we fail to constant fold we can't guarantee the (dead) nodes
// we're creating will be cleaned up before being visited for
// legalization.
if (NewNodesMustHaveLegalTypes && !ScalarOp.isUndef() &&
!isa<ConstantSDNode>(ScalarOp) &&
TLI->getTypeAction(*getContext(), InSVT) !=
TargetLowering::TypeLegal)
return SDValue();
ScalarOp = getNode(ISD::TRUNCATE, DL, InSVT, ScalarOp);
}
ScalarOps.push_back(ScalarOp);
}
// Constant fold the scalar operands.
SDValue ScalarResult = getNode(Opcode, DL, SVT, ScalarOps);
// Legalize the (integer) scalar constant if necessary.
if (LegalSVT != SVT)
ScalarResult = getNode(ExtendCode, DL, LegalSVT, ScalarResult);
// Scalar folding only succeeded if the result is a constant or UNDEF.
if (!ScalarResult.isUndef() && ScalarResult.getOpcode() != ISD::Constant &&
ScalarResult.getOpcode() != ISD::ConstantFP)
return SDValue();
ScalarResults.push_back(ScalarResult);
}
SDValue V = NumElts.isScalable() ? getSplatVector(VT, DL, ScalarResults[0])
: getBuildVector(VT, DL, ScalarResults);
NewSDValueDbgMsg(V, "New node fold constant vector: ", this);
return V;
}
SDValue SelectionDAG::foldConstantFPMath(unsigned Opcode, const SDLoc &DL,
EVT VT, ArrayRef<SDValue> Ops) {
// TODO: Add support for unary/ternary fp opcodes.
if (Ops.size() != 2)
return SDValue();
// TODO: We don't do any constant folding for strict FP opcodes here, but we
// should. That will require dealing with a potentially non-default
// rounding mode, checking the "opStatus" return value from the APFloat
// math calculations, and possibly other variations.
SDValue N1 = Ops[0];
SDValue N2 = Ops[1];
ConstantFPSDNode *N1CFP = isConstOrConstSplatFP(N1, /*AllowUndefs*/ false);
ConstantFPSDNode *N2CFP = isConstOrConstSplatFP(N2, /*AllowUndefs*/ false);
if (N1CFP && N2CFP) {
APFloat C1 = N1CFP->getValueAPF(); // make copy
const APFloat &C2 = N2CFP->getValueAPF();
switch (Opcode) {
case ISD::FADD:
C1.add(C2, APFloat::rmNearestTiesToEven);
return getConstantFP(C1, DL, VT);
case ISD::FSUB:
C1.subtract(C2, APFloat::rmNearestTiesToEven);
return getConstantFP(C1, DL, VT);
case ISD::FMUL:
C1.multiply(C2, APFloat::rmNearestTiesToEven);
return getConstantFP(C1, DL, VT);
case ISD::FDIV:
C1.divide(C2, APFloat::rmNearestTiesToEven);
return getConstantFP(C1, DL, VT);
case ISD::FREM:
C1.mod(C2);
return getConstantFP(C1, DL, VT);
case ISD::FCOPYSIGN:
C1.copySign(C2);
return getConstantFP(C1, DL, VT);
case ISD::FMINNUM:
return getConstantFP(minnum(C1, C2), DL, VT);
case ISD::FMAXNUM:
return getConstantFP(maxnum(C1, C2), DL, VT);
case ISD::FMINIMUM:
return getConstantFP(minimum(C1, C2), DL, VT);
case ISD::FMAXIMUM:
return getConstantFP(maximum(C1, C2), DL, VT);
default: break;
}
}
if (N1CFP && Opcode == ISD::FP_ROUND) {
APFloat C1 = N1CFP->getValueAPF(); // make copy
bool Unused;
// This can return overflow, underflow, or inexact; we don't care.
// FIXME need to be more flexible about rounding mode.
(void) C1.convert(EVTToAPFloatSemantics(VT), APFloat::rmNearestTiesToEven,
&Unused);
return getConstantFP(C1, DL, VT);
}
switch (Opcode) {
case ISD::FSUB:
// -0.0 - undef --> undef (consistent with "fneg undef")
if (ConstantFPSDNode *N1C = isConstOrConstSplatFP(N1, /*AllowUndefs*/ true))
if (N1C && N1C->getValueAPF().isNegZero() && N2.isUndef())
return getUNDEF(VT);
[[fallthrough]];
case ISD::FADD:
case ISD::FMUL:
case ISD::FDIV:
case ISD::FREM:
// If both operands are undef, the result is undef. If 1 operand is undef,
// the result is NaN. This should match the behavior of the IR optimizer.
if (N1.isUndef() && N2.isUndef())
return getUNDEF(VT);
if (N1.isUndef() || N2.isUndef())
return getConstantFP(APFloat::getNaN(EVTToAPFloatSemantics(VT)), DL, VT);
}
return SDValue();
}
SDValue SelectionDAG::getAssertAlign(const SDLoc &DL, SDValue Val, Align A) {
assert(Val.getValueType().isInteger() && "Invalid AssertAlign!");
// There's no need to assert on a byte-aligned pointer. All pointers are at
// least byte aligned.
if (A == Align(1))
return Val;
FoldingSetNodeID ID;
AddNodeIDNode(ID, ISD::AssertAlign, getVTList(Val.getValueType()), {Val});
ID.AddInteger(A.value());
void *IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, DL, IP))
return SDValue(E, 0);
auto *N = newSDNode<AssertAlignSDNode>(DL.getIROrder(), DL.getDebugLoc(),
Val.getValueType(), A);
createOperands(N, {Val});
CSEMap.InsertNode(N, IP);
InsertNode(N);
SDValue V(N, 0);
NewSDValueDbgMsg(V, "Creating new node: ", this);
return V;
}
SDValue SelectionDAG::getNode(unsigned Opcode, const SDLoc &DL, EVT VT,
SDValue N1, SDValue N2) {
SDNodeFlags Flags;
if (Inserter)
Flags = Inserter->getFlags();
return getNode(Opcode, DL, VT, N1, N2, Flags);
}
void SelectionDAG::canonicalizeCommutativeBinop(unsigned Opcode, SDValue &N1,
SDValue &N2) const {
if (!TLI->isCommutativeBinOp(Opcode))
return;
// Canonicalize:
// binop(const, nonconst) -> binop(nonconst, const)
SDNode *N1C = isConstantIntBuildVectorOrConstantInt(N1);
SDNode *N2C = isConstantIntBuildVectorOrConstantInt(N2);
SDNode *N1CFP = isConstantFPBuildVectorOrConstantFP(N1);
SDNode *N2CFP = isConstantFPBuildVectorOrConstantFP(N2);
if ((N1C && !N2C) || (N1CFP && !N2CFP))
std::swap(N1, N2);
// Canonicalize:
// binop(splat(x), step_vector) -> binop(step_vector, splat(x))
else if (N1.getOpcode() == ISD::SPLAT_VECTOR &&
N2.getOpcode() == ISD::STEP_VECTOR)
std::swap(N1, N2);
}
SDValue SelectionDAG::getNode(unsigned Opcode, const SDLoc &DL, EVT VT,
SDValue N1, SDValue N2, const SDNodeFlags Flags) {
assert(N1.getOpcode() != ISD::DELETED_NODE &&
N2.getOpcode() != ISD::DELETED_NODE &&
"Operand is DELETED_NODE!");
canonicalizeCommutativeBinop(Opcode, N1, N2);
auto *N1C = dyn_cast<ConstantSDNode>(N1);
auto *N2C = dyn_cast<ConstantSDNode>(N2);
// Don't allow undefs in vector splats - we might be returning N2 when folding
// to zero etc.
ConstantSDNode *N2CV =
isConstOrConstSplat(N2, /*AllowUndefs*/ false, /*AllowTruncation*/ true);
switch (Opcode) {
default: break;
case ISD::TokenFactor:
assert(VT == MVT::Other && N1.getValueType() == MVT::Other &&
N2.getValueType() == MVT::Other && "Invalid token factor!");
// Fold trivial token factors.
if (N1.getOpcode() == ISD::EntryToken) return N2;
if (N2.getOpcode() == ISD::EntryToken) return N1;
if (N1 == N2) return N1;
break;
case ISD::BUILD_VECTOR: {
// Attempt to simplify BUILD_VECTOR.
SDValue Ops[] = {N1, N2};
if (SDValue V = FoldBUILD_VECTOR(DL, VT, Ops, *this))
return V;
break;
}
case ISD::CONCAT_VECTORS: {
SDValue Ops[] = {N1, N2};
if (SDValue V = foldCONCAT_VECTORS(DL, VT, Ops, *this))
return V;
break;
}
case ISD::AND:
assert(VT.isInteger() && "This operator does not apply to FP types!");
assert(N1.getValueType() == N2.getValueType() &&
N1.getValueType() == VT && "Binary operator types must match!");
// (X & 0) -> 0. This commonly occurs when legalizing i64 values, so it's
// worth handling here.
if (N2CV && N2CV->isZero())
return N2;
if (N2CV && N2CV->isAllOnes()) // X & -1 -> X
return N1;
break;
case ISD::OR:
case ISD::XOR:
case ISD::ADD:
case ISD::SUB:
assert(VT.isInteger() && "This operator does not apply to FP types!");
assert(N1.getValueType() == N2.getValueType() &&
N1.getValueType() == VT && "Binary operator types must match!");
// (X ^|+- 0) -> X. This commonly occurs when legalizing i64 values, so
// it's worth handling here.
if (N2CV && N2CV->isZero())
return N1;
if ((Opcode == ISD::ADD || Opcode == ISD::SUB) && VT.isVector() &&
VT.getVectorElementType() == MVT::i1)
return getNode(ISD::XOR, DL, VT, N1, N2);
break;
case ISD::MUL:
assert(VT.isInteger() && "This operator does not apply to FP types!");
assert(N1.getValueType() == N2.getValueType() &&
N1.getValueType() == VT && "Binary operator types must match!");
if (VT.isVector() && VT.getVectorElementType() == MVT::i1)
return getNode(ISD::AND, DL, VT, N1, N2);
if (N2C && (N1.getOpcode() == ISD::VSCALE) && Flags.hasNoSignedWrap()) {
const APInt &MulImm = N1->getConstantOperandAPInt(0);
const APInt &N2CImm = N2C->getAPIntValue();
return getVScale(DL, VT, MulImm * N2CImm);
}
break;
case ISD::UDIV:
case ISD::UREM:
case ISD::MULHU:
case ISD::MULHS:
case ISD::SDIV:
case ISD::SREM:
case ISD::SADDSAT:
case ISD::SSUBSAT:
case ISD::UADDSAT:
case ISD::USUBSAT:
assert(VT.isInteger() && "This operator does not apply to FP types!");
assert(N1.getValueType() == N2.getValueType() &&
N1.getValueType() == VT && "Binary operator types must match!");
if (VT.isVector() && VT.getVectorElementType() == MVT::i1) {
// fold (add_sat x, y) -> (or x, y) for bool types.
if (Opcode == ISD::SADDSAT || Opcode == ISD::UADDSAT)
return getNode(ISD::OR, DL, VT, N1, N2);
// fold (sub_sat x, y) -> (and x, ~y) for bool types.
if (Opcode == ISD::SSUBSAT || Opcode == ISD::USUBSAT)
return getNode(ISD::AND, DL, VT, N1, getNOT(DL, N2, VT));
}
break;
case ISD::ABDS:
case ISD::ABDU:
assert(VT.isInteger() && "This operator does not apply to FP types!");
assert(N1.getValueType() == N2.getValueType() &&
N1.getValueType() == VT && "Binary operator types must match!");
break;
case ISD::SMIN:
case ISD::UMAX:
assert(VT.isInteger() && "This operator does not apply to FP types!");
assert(N1.getValueType() == N2.getValueType() &&
N1.getValueType() == VT && "Binary operator types must match!");
if (VT.isVector() && VT.getVectorElementType() == MVT::i1)
return getNode(ISD::OR, DL, VT, N1, N2);
break;
case ISD::SMAX:
case ISD::UMIN:
assert(VT.isInteger() && "This operator does not apply to FP types!");
assert(N1.getValueType() == N2.getValueType() &&
N1.getValueType() == VT && "Binary operator types must match!");
if (VT.isVector() && VT.getVectorElementType() == MVT::i1)
return getNode(ISD::AND, DL, VT, N1, N2);
break;
case ISD::FADD:
case ISD::FSUB:
case ISD::FMUL:
case ISD::FDIV:
case ISD::FREM:
assert(VT.isFloatingPoint() && "This operator only applies to FP types!");
assert(N1.getValueType() == N2.getValueType() &&
N1.getValueType() == VT && "Binary operator types must match!");
if (SDValue V = simplifyFPBinop(Opcode, N1, N2, Flags))
return V;
break;
case ISD::FCOPYSIGN: // N1 and result must match. N1/N2 need not match.
assert(N1.getValueType() == VT &&
N1.getValueType().isFloatingPoint() &&
N2.getValueType().isFloatingPoint() &&
"Invalid FCOPYSIGN!");
break;
case ISD::SHL:
if (N2C && (N1.getOpcode() == ISD::VSCALE) && Flags.hasNoSignedWrap()) {
const APInt &MulImm = N1->getConstantOperandAPInt(0);
const APInt &ShiftImm = N2C->getAPIntValue();
return getVScale(DL, VT, MulImm << ShiftImm);
}
[[fallthrough]];
case ISD::SRA:
case ISD::SRL:
if (SDValue V = simplifyShift(N1, N2))
return V;
[[fallthrough]];
case ISD::ROTL:
case ISD::ROTR:
assert(VT == N1.getValueType() &&
"Shift operators return type must be the same as their first arg");
assert(VT.isInteger() && N2.getValueType().isInteger() &&
"Shifts only work on integers");
assert((!VT.isVector() || VT == N2.getValueType()) &&
"Vector shift amounts must be in the same as their first arg");
// Verify that the shift amount VT is big enough to hold valid shift
// amounts. This catches things like trying to shift an i1024 value by an
// i8, which is easy to fall into in generic code that uses
// TLI.getShiftAmount().
assert(N2.getValueType().getScalarSizeInBits() >=
Log2_32_Ceil(VT.getScalarSizeInBits()) &&
"Invalid use of small shift amount with oversized value!");
// Always fold shifts of i1 values so the code generator doesn't need to
// handle them. Since we know the size of the shift has to be less than the
// size of the value, the shift/rotate count is guaranteed to be zero.
if (VT == MVT::i1)
return N1;
if (N2CV && N2CV->isZero())
return N1;
break;
case ISD::FP_ROUND:
assert(VT.isFloatingPoint() &&
N1.getValueType().isFloatingPoint() &&
VT.bitsLE(N1.getValueType()) &&
N2C && (N2C->getZExtValue() == 0 || N2C->getZExtValue() == 1) &&
"Invalid FP_ROUND!");
if (N1.getValueType() == VT) return N1; // noop conversion.
break;
case ISD::AssertSext:
case ISD::AssertZext: {
EVT EVT = cast<VTSDNode>(N2)->getVT();
assert(VT == N1.getValueType() && "Not an inreg extend!");
assert(VT.isInteger() && EVT.isInteger() &&
"Cannot *_EXTEND_INREG FP types");
assert(!EVT.isVector() &&
"AssertSExt/AssertZExt type should be the vector element type "
"rather than the vector type!");
assert(EVT.bitsLE(VT.getScalarType()) && "Not extending!");
if (VT.getScalarType() == EVT) return N1; // noop assertion.
break;
}
case ISD::SIGN_EXTEND_INREG: {
EVT EVT = cast<VTSDNode>(N2)->getVT();
assert(VT == N1.getValueType() && "Not an inreg extend!");
assert(VT.isInteger() && EVT.isInteger() &&
"Cannot *_EXTEND_INREG FP types");
assert(EVT.isVector() == VT.isVector() &&
"SIGN_EXTEND_INREG type should be vector iff the operand "
"type is vector!");
assert((!EVT.isVector() ||
EVT.getVectorElementCount() == VT.getVectorElementCount()) &&
"Vector element counts must match in SIGN_EXTEND_INREG");
assert(EVT.bitsLE(VT) && "Not extending!");
if (EVT == VT) return N1; // Not actually extending
auto SignExtendInReg = [&](APInt Val, llvm::EVT ConstantVT) {
unsigned FromBits = EVT.getScalarSizeInBits();
Val <<= Val.getBitWidth() - FromBits;
Val.ashrInPlace(Val.getBitWidth() - FromBits);
return getConstant(Val, DL, ConstantVT);
};
if (N1C) {
const APInt &Val = N1C->getAPIntValue();
return SignExtendInReg(Val, VT);
}
if (ISD::isBuildVectorOfConstantSDNodes(N1.getNode())) {
SmallVector<SDValue, 8> Ops;
llvm::EVT OpVT = N1.getOperand(0).getValueType();
for (int i = 0, e = VT.getVectorNumElements(); i != e; ++i) {
SDValue Op = N1.getOperand(i);
if (Op.isUndef()) {
Ops.push_back(getUNDEF(OpVT));
continue;
}
ConstantSDNode *C = cast<ConstantSDNode>(Op);
APInt Val = C->getAPIntValue();
Ops.push_back(SignExtendInReg(Val, OpVT));
}
return getBuildVector(VT, DL, Ops);
}
if (N1.getOpcode() == ISD::SPLAT_VECTOR &&
isa<ConstantSDNode>(N1.getOperand(0)))
return getNode(
ISD::SPLAT_VECTOR, DL, VT,
SignExtendInReg(N1.getConstantOperandAPInt(0),
N1.getOperand(0).getValueType()));
break;
}
case ISD::FP_TO_SINT_SAT:
case ISD::FP_TO_UINT_SAT: {
assert(VT.isInteger() && cast<VTSDNode>(N2)->getVT().isInteger() &&
N1.getValueType().isFloatingPoint() && "Invalid FP_TO_*INT_SAT");
assert(N1.getValueType().isVector() == VT.isVector() &&
"FP_TO_*INT_SAT type should be vector iff the operand type is "
"vector!");
assert((!VT.isVector() || VT.getVectorElementCount() ==
N1.getValueType().getVectorElementCount()) &&
"Vector element counts must match in FP_TO_*INT_SAT");
assert(!cast<VTSDNode>(N2)->getVT().isVector() &&
"Type to saturate to must be a scalar.");
assert(cast<VTSDNode>(N2)->getVT().bitsLE(VT.getScalarType()) &&
"Not extending!");
break;
}
case ISD::EXTRACT_VECTOR_ELT:
assert(VT.getSizeInBits() >= N1.getValueType().getScalarSizeInBits() &&
"The result of EXTRACT_VECTOR_ELT must be at least as wide as the \
element type of the vector.");
// Extract from an undefined value or using an undefined index is undefined.
if (N1.isUndef() || N2.isUndef())
return getUNDEF(VT);
// EXTRACT_VECTOR_ELT of out-of-bounds element is an UNDEF for fixed length
// vectors. For scalable vectors we will provide appropriate support for
// dealing with arbitrary indices.
if (N2C && N1.getValueType().isFixedLengthVector() &&
N2C->getAPIntValue().uge(N1.getValueType().getVectorNumElements()))
return getUNDEF(VT);
// EXTRACT_VECTOR_ELT of CONCAT_VECTORS is often formed while lowering is
// expanding copies of large vectors from registers. This only works for
// fixed length vectors, since we need to know the exact number of
// elements.
if (N2C && N1.getOpcode() == ISD::CONCAT_VECTORS &&
N1.getOperand(0).getValueType().isFixedLengthVector()) {
unsigned Factor =
N1.getOperand(0).getValueType().getVectorNumElements();
return getNode(ISD::EXTRACT_VECTOR_ELT, DL, VT,
N1.getOperand(N2C->getZExtValue() / Factor),
getVectorIdxConstant(N2C->getZExtValue() % Factor, DL));
}
// EXTRACT_VECTOR_ELT of BUILD_VECTOR or SPLAT_VECTOR is often formed while
// lowering is expanding large vector constants.
if (N2C && (N1.getOpcode() == ISD::BUILD_VECTOR ||
N1.getOpcode() == ISD::SPLAT_VECTOR)) {
assert((N1.getOpcode() != ISD::BUILD_VECTOR ||
N1.getValueType().isFixedLengthVector()) &&
"BUILD_VECTOR used for scalable vectors");
unsigned Index =
N1.getOpcode() == ISD::BUILD_VECTOR ? N2C->getZExtValue() : 0;
SDValue Elt = N1.getOperand(Index);
if (VT != Elt.getValueType())
// If the vector element type is not legal, the BUILD_VECTOR operands
// are promoted and implicitly truncated, and the result implicitly
// extended. Make that explicit here.
Elt = getAnyExtOrTrunc(Elt, DL, VT);
return Elt;
}
// EXTRACT_VECTOR_ELT of INSERT_VECTOR_ELT is often formed when vector
// operations are lowered to scalars.
if (N1.getOpcode() == ISD::INSERT_VECTOR_ELT) {
// If the indices are the same, return the inserted element else
// if the indices are known different, extract the element from
// the original vector.
SDValue N1Op2 = N1.getOperand(2);
ConstantSDNode *N1Op2C = dyn_cast<ConstantSDNode>(N1Op2);
if (N1Op2C && N2C) {
if (N1Op2C->getZExtValue() == N2C->getZExtValue()) {
if (VT == N1.getOperand(1).getValueType())
return N1.getOperand(1);
if (VT.isFloatingPoint()) {
assert(VT.getSizeInBits() > N1.getOperand(1).getValueType().getSizeInBits());
return getFPExtendOrRound(N1.getOperand(1), DL, VT);
}
return getSExtOrTrunc(N1.getOperand(1), DL, VT);
}
return getNode(ISD::EXTRACT_VECTOR_ELT, DL, VT, N1.getOperand(0), N2);
}
}
// EXTRACT_VECTOR_ELT of v1iX EXTRACT_SUBVECTOR could be formed
// when vector types are scalarized and v1iX is legal.
// vextract (v1iX extract_subvector(vNiX, Idx)) -> vextract(vNiX,Idx).
// Here we are completely ignoring the extract element index (N2),
// which is fine for fixed width vectors, since any index other than 0
// is undefined anyway. However, this cannot be ignored for scalable
// vectors - in theory we could support this, but we don't want to do this
// without a profitability check.
if (N1.getOpcode() == ISD::EXTRACT_SUBVECTOR &&
N1.getValueType().isFixedLengthVector() &&
N1.getValueType().getVectorNumElements() == 1) {
return getNode(ISD::EXTRACT_VECTOR_ELT, DL, VT, N1.getOperand(0),
N1.getOperand(1));
}
break;
case ISD::EXTRACT_ELEMENT:
assert(N2C && (unsigned)N2C->getZExtValue() < 2 && "Bad EXTRACT_ELEMENT!");
assert(!N1.getValueType().isVector() && !VT.isVector() &&
(N1.getValueType().isInteger() == VT.isInteger()) &&
N1.getValueType() != VT &&
"Wrong types for EXTRACT_ELEMENT!");
// EXTRACT_ELEMENT of BUILD_PAIR is often formed while legalize is expanding
// 64-bit integers into 32-bit parts. Instead of building the extract of
// the BUILD_PAIR, only to have legalize rip it apart, just do it now.
if (N1.getOpcode() == ISD::BUILD_PAIR)
return N1.getOperand(N2C->getZExtValue());
// EXTRACT_ELEMENT of a constant int is also very common.
if (N1C) {
unsigned ElementSize = VT.getSizeInBits();
unsigned Shift = ElementSize * N2C->getZExtValue();
const APInt &Val = N1C->getAPIntValue();
return getConstant(Val.extractBits(ElementSize, Shift), DL, VT);
}
break;
case ISD::EXTRACT_SUBVECTOR: {
EVT N1VT = N1.getValueType();
assert(VT.isVector() && N1VT.isVector() &&
"Extract subvector VTs must be vectors!");
assert(VT.getVectorElementType() == N1VT.getVectorElementType() &&
"Extract subvector VTs must have the same element type!");
assert((VT.isFixedLengthVector() || N1VT.isScalableVector()) &&
"Cannot extract a scalable vector from a fixed length vector!");
assert((VT.isScalableVector() != N1VT.isScalableVector() ||
VT.getVectorMinNumElements() <= N1VT.getVectorMinNumElements()) &&
"Extract subvector must be from larger vector to smaller vector!");
assert(N2C && "Extract subvector index must be a constant");
assert((VT.isScalableVector() != N1VT.isScalableVector() ||
(VT.getVectorMinNumElements() + N2C->getZExtValue()) <=
N1VT.getVectorMinNumElements()) &&
"Extract subvector overflow!");
assert(N2C->getAPIntValue().getBitWidth() ==
TLI->getVectorIdxTy(getDataLayout()).getFixedSizeInBits() &&
"Constant index for EXTRACT_SUBVECTOR has an invalid size");
// Trivial extraction.
if (VT == N1VT)
return N1;
// EXTRACT_SUBVECTOR of an UNDEF is an UNDEF.
if (N1.isUndef())
return getUNDEF(VT);
// EXTRACT_SUBVECTOR of CONCAT_VECTOR can be simplified if the pieces of
// the concat have the same type as the extract.
if (N1.getOpcode() == ISD::CONCAT_VECTORS &&
VT == N1.getOperand(0).getValueType()) {
unsigned Factor = VT.getVectorMinNumElements();
return N1.getOperand(N2C->getZExtValue() / Factor);
}
// EXTRACT_SUBVECTOR of INSERT_SUBVECTOR is often created
// during shuffle legalization.
if (N1.getOpcode() == ISD::INSERT_SUBVECTOR && N2 == N1.getOperand(2) &&
VT == N1.getOperand(1).getValueType())
return N1.getOperand(1);
break;
}
}
// Perform trivial constant folding.
if (SDValue SV = FoldConstantArithmetic(Opcode, DL, VT, {N1, N2}))
return SV;
// Canonicalize an UNDEF to the RHS, even over a constant.
if (N1.isUndef()) {
if (TLI->isCommutativeBinOp(Opcode)) {
std::swap(N1, N2);
} else {
switch (Opcode) {
case ISD::SUB:
return getUNDEF(VT); // fold op(undef, arg2) -> undef
case ISD::SIGN_EXTEND_INREG:
case ISD::UDIV:
case ISD::SDIV:
case ISD::UREM:
case ISD::SREM:
case ISD::SSUBSAT:
case ISD::USUBSAT:
return getConstant(0, DL, VT); // fold op(undef, arg2) -> 0
}
}
}
// Fold a bunch of operators when the RHS is undef.
if (N2.isUndef()) {
switch (Opcode) {
case ISD::XOR:
if (N1.isUndef())
// Handle undef ^ undef -> 0 special case. This is a common
// idiom (misuse).
return getConstant(0, DL, VT);
[[fallthrough]];
case ISD::ADD:
case ISD::SUB:
case ISD::UDIV:
case ISD::SDIV:
case ISD::UREM:
case ISD::SREM:
return getUNDEF(VT); // fold op(arg1, undef) -> undef
case ISD::MUL:
case ISD::AND:
case ISD::SSUBSAT:
case ISD::USUBSAT:
return getConstant(0, DL, VT); // fold op(arg1, undef) -> 0
case ISD::OR:
case ISD::SADDSAT:
case ISD::UADDSAT:
return getAllOnesConstant(DL, VT);
}
}
// Memoize this node if possible.
SDNode *N;
SDVTList VTs = getVTList(VT);
SDValue Ops[] = {N1, N2};
if (VT != MVT::Glue) {
FoldingSetNodeID ID;
AddNodeIDNode(ID, Opcode, VTs, Ops);
void *IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, DL, IP)) {
E->intersectFlagsWith(Flags);
return SDValue(E, 0);
}
N = newSDNode<SDNode>(Opcode, DL.getIROrder(), DL.getDebugLoc(), VTs);
N->setFlags(Flags);
createOperands(N, Ops);
CSEMap.InsertNode(N, IP);
} else {
N = newSDNode<SDNode>(Opcode, DL.getIROrder(), DL.getDebugLoc(), VTs);
createOperands(N, Ops);
}
InsertNode(N);
SDValue V = SDValue(N, 0);
NewSDValueDbgMsg(V, "Creating new node: ", this);
return V;
}
SDValue SelectionDAG::getNode(unsigned Opcode, const SDLoc &DL, EVT VT,
SDValue N1, SDValue N2, SDValue N3) {
SDNodeFlags Flags;
if (Inserter)
Flags = Inserter->getFlags();
return getNode(Opcode, DL, VT, N1, N2, N3, Flags);
}
SDValue SelectionDAG::getNode(unsigned Opcode, const SDLoc &DL, EVT VT,
SDValue N1, SDValue N2, SDValue N3,
const SDNodeFlags Flags) {
assert(N1.getOpcode() != ISD::DELETED_NODE &&
N2.getOpcode() != ISD::DELETED_NODE &&
N3.getOpcode() != ISD::DELETED_NODE &&
"Operand is DELETED_NODE!");
// Perform various simplifications.
switch (Opcode) {
case ISD::FMA:
case ISD::FMAD: {
assert(VT.isFloatingPoint() && "This operator only applies to FP types!");
assert(N1.getValueType() == VT && N2.getValueType() == VT &&
N3.getValueType() == VT && "FMA types must match!");
ConstantFPSDNode *N1CFP = dyn_cast<ConstantFPSDNode>(N1);
ConstantFPSDNode *N2CFP = dyn_cast<ConstantFPSDNode>(N2);
ConstantFPSDNode *N3CFP = dyn_cast<ConstantFPSDNode>(N3);
if (N1CFP && N2CFP && N3CFP) {
APFloat V1 = N1CFP->getValueAPF();
const APFloat &V2 = N2CFP->getValueAPF();
const APFloat &V3 = N3CFP->getValueAPF();
if (Opcode == ISD::FMAD) {
V1.multiply(V2, APFloat::rmNearestTiesToEven);
V1.add(V3, APFloat::rmNearestTiesToEven);
} else
V1.fusedMultiplyAdd(V2, V3, APFloat::rmNearestTiesToEven);
return getConstantFP(V1, DL, VT);
}
break;
}
case ISD::BUILD_VECTOR: {
// Attempt to simplify BUILD_VECTOR.
SDValue Ops[] = {N1, N2, N3};
if (SDValue V = FoldBUILD_VECTOR(DL, VT, Ops, *this))
return V;
break;
}
case ISD::CONCAT_VECTORS: {
SDValue Ops[] = {N1, N2, N3};
if (SDValue V = foldCONCAT_VECTORS(DL, VT, Ops, *this))
return V;
break;
}
case ISD::SETCC: {
assert(VT.isInteger() && "SETCC result type must be an integer!");
assert(N1.getValueType() == N2.getValueType() &&
"SETCC operands must have the same type!");
assert(VT.isVector() == N1.getValueType().isVector() &&
"SETCC type should be vector iff the operand type is vector!");
assert((!VT.isVector() || VT.getVectorElementCount() ==
N1.getValueType().getVectorElementCount()) &&
"SETCC vector element counts must match!");
// Use FoldSetCC to simplify SETCC's.
if (SDValue V = FoldSetCC(VT, N1, N2, cast<CondCodeSDNode>(N3)->get(), DL))
return V;
// Vector constant folding.
SDValue Ops[] = {N1, N2, N3};
if (SDValue V = FoldConstantArithmetic(Opcode, DL, VT, Ops)) {
NewSDValueDbgMsg(V, "New node vector constant folding: ", this);
return V;
}
break;
}
case ISD::SELECT:
case ISD::VSELECT:
if (SDValue V = simplifySelect(N1, N2, N3))
return V;
break;
case ISD::VECTOR_SHUFFLE:
llvm_unreachable("should use getVectorShuffle constructor!");
case ISD::VECTOR_SPLICE: {
if (cast<ConstantSDNode>(N3)->isZero())
return N1;
break;
}
case ISD::INSERT_VECTOR_ELT: {
ConstantSDNode *N3C = dyn_cast<ConstantSDNode>(N3);
// INSERT_VECTOR_ELT into out-of-bounds element is an UNDEF, except
// for scalable vectors where we will generate appropriate code to
// deal with out-of-bounds cases correctly.
if (N3C && N1.getValueType().isFixedLengthVector() &&
N3C->getZExtValue() >= N1.getValueType().getVectorNumElements())
return getUNDEF(VT);
// Undefined index can be assumed out-of-bounds, so that's UNDEF too.
if (N3.isUndef())
return getUNDEF(VT);
// If the inserted element is an UNDEF, just use the input vector.
if (N2.isUndef())
return N1;
break;
}
case ISD::INSERT_SUBVECTOR: {
// Inserting undef into undef is still undef.
if (N1.isUndef() && N2.isUndef())
return getUNDEF(VT);
EVT N2VT = N2.getValueType();
assert(VT == N1.getValueType() &&
"Dest and insert subvector source types must match!");
assert(VT.isVector() && N2VT.isVector() &&
"Insert subvector VTs must be vectors!");
assert(VT.getVectorElementType() == N2VT.getVectorElementType() &&
"Insert subvector VTs must have the same element type!");
assert((VT.isScalableVector() || N2VT.isFixedLengthVector()) &&
"Cannot insert a scalable vector into a fixed length vector!");
assert((VT.isScalableVector() != N2VT.isScalableVector() ||
VT.getVectorMinNumElements() >= N2VT.getVectorMinNumElements()) &&
"Insert subvector must be from smaller vector to larger vector!");
assert(isa<ConstantSDNode>(N3) &&
"Insert subvector index must be constant");
assert((VT.isScalableVector() != N2VT.isScalableVector() ||
(N2VT.getVectorMinNumElements() + N3->getAsZExtVal()) <=
VT.getVectorMinNumElements()) &&
"Insert subvector overflow!");
assert(N3->getAsAPIntVal().getBitWidth() ==
TLI->getVectorIdxTy(getDataLayout()).getFixedSizeInBits() &&
"Constant index for INSERT_SUBVECTOR has an invalid size");
// Trivial insertion.
if (VT == N2VT)
return N2;
// If this is an insert of an extracted vector into an undef vector, we
// can just use the input to the extract.
if (N1.isUndef() && N2.getOpcode() == ISD::EXTRACT_SUBVECTOR &&
N2.getOperand(1) == N3 && N2.getOperand(0).getValueType() == VT)
return N2.getOperand(0);
break;
}
case ISD::BITCAST:
// Fold bit_convert nodes from a type to themselves.
if (N1.getValueType() == VT)
return N1;
break;
case ISD::VP_TRUNCATE:
case ISD::VP_SIGN_EXTEND:
case ISD::VP_ZERO_EXTEND:
// Don't create noop casts.
if (N1.getValueType() == VT)
return N1;
break;
}
// Memoize node if it doesn't produce a glue result.
SDNode *N;
SDVTList VTs = getVTList(VT);
SDValue Ops[] = {N1, N2, N3};
if (VT != MVT::Glue) {
FoldingSetNodeID ID;
AddNodeIDNode(ID, Opcode, VTs, Ops);
void *IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, DL, IP)) {
E->intersectFlagsWith(Flags);
return SDValue(E, 0);
}
N = newSDNode<SDNode>(Opcode, DL.getIROrder(), DL.getDebugLoc(), VTs);
N->setFlags(Flags);
createOperands(N, Ops);
CSEMap.InsertNode(N, IP);
} else {
N = newSDNode<SDNode>(Opcode, DL.getIROrder(), DL.getDebugLoc(), VTs);
createOperands(N, Ops);
}
InsertNode(N);
SDValue V = SDValue(N, 0);
NewSDValueDbgMsg(V, "Creating new node: ", this);
return V;
}
SDValue SelectionDAG::getNode(unsigned Opcode, const SDLoc &DL, EVT VT,
SDValue N1, SDValue N2, SDValue N3, SDValue N4) {
SDValue Ops[] = { N1, N2, N3, N4 };
return getNode(Opcode, DL, VT, Ops);
}
SDValue SelectionDAG::getNode(unsigned Opcode, const SDLoc &DL, EVT VT,
SDValue N1, SDValue N2, SDValue N3, SDValue N4,
SDValue N5) {
SDValue Ops[] = { N1, N2, N3, N4, N5 };
return getNode(Opcode, DL, VT, Ops);
}
/// getStackArgumentTokenFactor - Compute a TokenFactor to force all
/// the incoming stack arguments to be loaded from the stack.
SDValue SelectionDAG::getStackArgumentTokenFactor(SDValue Chain) {
SmallVector<SDValue, 8> ArgChains;
// Include the original chain at the beginning of the list. When this is
// used by target LowerCall hooks, this helps legalize find the
// CALLSEQ_BEGIN node.
ArgChains.push_back(Chain);
// Add a chain value for each stack argument.
for (SDNode *U : getEntryNode().getNode()->uses())
if (LoadSDNode *L = dyn_cast<LoadSDNode>(U))
if (FrameIndexSDNode *FI = dyn_cast<FrameIndexSDNode>(L->getBasePtr()))
if (FI->getIndex() < 0)
ArgChains.push_back(SDValue(L, 1));
// Build a tokenfactor for all the chains.
return getNode(ISD::TokenFactor, SDLoc(Chain), MVT::Other, ArgChains);
}
/// getMemsetValue - Vectorized representation of the memset value
/// operand.
static SDValue getMemsetValue(SDValue Value, EVT VT, SelectionDAG &DAG,
const SDLoc &dl) {
assert(!Value.isUndef());
unsigned NumBits = VT.getScalarSizeInBits();
if (ConstantSDNode *C = dyn_cast<ConstantSDNode>(Value)) {
assert(C->getAPIntValue().getBitWidth() == 8);
APInt Val = APInt::getSplat(NumBits, C->getAPIntValue());
if (VT.isInteger()) {
bool IsOpaque = VT.getSizeInBits() > 64 ||
!DAG.getTargetLoweringInfo().isLegalStoreImmediate(C->getSExtValue());
return DAG.getConstant(Val, dl, VT, false, IsOpaque);
}
return DAG.getConstantFP(APFloat(DAG.EVTToAPFloatSemantics(VT), Val), dl,
VT);
}
assert(Value.getValueType() == MVT::i8 && "memset with non-byte fill value?");
EVT IntVT = VT.getScalarType();
if (!IntVT.isInteger())
IntVT = EVT::getIntegerVT(*DAG.getContext(), IntVT.getSizeInBits());
Value = DAG.getNode(ISD::ZERO_EXTEND, dl, IntVT, Value);
if (NumBits > 8) {
// Use a multiplication with 0x010101... to extend the input to the
// required length.
APInt Magic = APInt::getSplat(NumBits, APInt(8, 0x01));
Value = DAG.getNode(ISD::MUL, dl, IntVT, Value,
DAG.getConstant(Magic, dl, IntVT));
}
if (VT != Value.getValueType() && !VT.isInteger())
Value = DAG.getBitcast(VT.getScalarType(), Value);
if (VT != Value.getValueType())
Value = DAG.getSplatBuildVector(VT, dl, Value);
return Value;
}
/// getMemsetStringVal - Similar to getMemsetValue. Except this is only
/// used when a memcpy is turned into a memset when the source is a constant
/// string ptr.
static SDValue getMemsetStringVal(EVT VT, const SDLoc &dl, SelectionDAG &DAG,
const TargetLowering &TLI,
const ConstantDataArraySlice &Slice) {
// Handle vector with all elements zero.
if (Slice.Array == nullptr) {
if (VT.isInteger())
return DAG.getConstant(0, dl, VT);
if (VT == MVT::f32 || VT == MVT::f64 || VT == MVT::f128)
return DAG.getConstantFP(0.0, dl, VT);
if (VT.isVector()) {
unsigned NumElts = VT.getVectorNumElements();
MVT EltVT = (VT.getVectorElementType() == MVT::f32) ? MVT::i32 : MVT::i64;
return DAG.getNode(ISD::BITCAST, dl, VT,
DAG.getConstant(0, dl,
EVT::getVectorVT(*DAG.getContext(),
EltVT, NumElts)));
}
llvm_unreachable("Expected type!");
}
assert(!VT.isVector() && "Can't handle vector type here!");
unsigned NumVTBits = VT.getSizeInBits();
unsigned NumVTBytes = NumVTBits / 8;
unsigned NumBytes = std::min(NumVTBytes, unsigned(Slice.Length));
APInt Val(NumVTBits, 0);
if (DAG.getDataLayout().isLittleEndian()) {
for (unsigned i = 0; i != NumBytes; ++i)
Val |= (uint64_t)(unsigned char)Slice[i] << i*8;
} else {
for (unsigned i = 0; i != NumBytes; ++i)
Val |= (uint64_t)(unsigned char)Slice[i] << (NumVTBytes-i-1)*8;
}
// If the "cost" of materializing the integer immediate is less than the cost
// of a load, then it is cost effective to turn the load into the immediate.
Type *Ty = VT.getTypeForEVT(*DAG.getContext());
if (TLI.shouldConvertConstantLoadToIntImm(Val, Ty))
return DAG.getConstant(Val, dl, VT);
return SDValue();
}
SDValue SelectionDAG::getMemBasePlusOffset(SDValue Base, TypeSize Offset,
const SDLoc &DL,
const SDNodeFlags Flags) {
EVT VT = Base.getValueType();
SDValue Index;
if (Offset.isScalable())
Index = getVScale(DL, Base.getValueType(),
APInt(Base.getValueSizeInBits().getFixedValue(),
Offset.getKnownMinValue()));
else
Index = getConstant(Offset.getFixedValue(), DL, VT);
return getMemBasePlusOffset(Base, Index, DL, Flags);
}
SDValue SelectionDAG::getMemBasePlusOffset(SDValue Ptr, SDValue Offset,
const SDLoc &DL,
const SDNodeFlags Flags) {
assert(Offset.getValueType().isInteger());
EVT BasePtrVT = Ptr.getValueType();
return getNode(ISD::ADD, DL, BasePtrVT, Ptr, Offset, Flags);
}
/// Returns true if memcpy source is constant data.
static bool isMemSrcFromConstant(SDValue Src, ConstantDataArraySlice &Slice) {
uint64_t SrcDelta = 0;
GlobalAddressSDNode *G = nullptr;
if (Src.getOpcode() == ISD::GlobalAddress)
G = cast<GlobalAddressSDNode>(Src);
else if (Src.getOpcode() == ISD::ADD &&
Src.getOperand(0).getOpcode() == ISD::GlobalAddress &&
Src.getOperand(1).getOpcode() == ISD::Constant) {
G = cast<GlobalAddressSDNode>(Src.getOperand(0));
SrcDelta = Src.getConstantOperandVal(1);
}
if (!G)
return false;
return getConstantDataArrayInfo(G->getGlobal(), Slice, 8,
SrcDelta + G->getOffset());
}
static bool shouldLowerMemFuncForSize(const MachineFunction &MF,
SelectionDAG &DAG) {
// On Darwin, -Os means optimize for size without hurting performance, so
// only really optimize for size when -Oz (MinSize) is used.
if (MF.getTarget().getTargetTriple().isOSDarwin())
return MF.getFunction().hasMinSize();
return DAG.shouldOptForSize();
}
static void chainLoadsAndStoresForMemcpy(SelectionDAG &DAG, const SDLoc &dl,
SmallVector<SDValue, 32> &OutChains, unsigned From,
unsigned To, SmallVector<SDValue, 16> &OutLoadChains,
SmallVector<SDValue, 16> &OutStoreChains) {
assert(OutLoadChains.size() && "Missing loads in memcpy inlining");
assert(OutStoreChains.size() && "Missing stores in memcpy inlining");
SmallVector<SDValue, 16> GluedLoadChains;
for (unsigned i = From; i < To; ++i) {
OutChains.push_back(OutLoadChains[i]);
GluedLoadChains.push_back(OutLoadChains[i]);
}
// Chain for all loads.
SDValue LoadToken = DAG.getNode(ISD::TokenFactor, dl, MVT::Other,
GluedLoadChains);
for (unsigned i = From; i < To; ++i) {
StoreSDNode *ST = dyn_cast<StoreSDNode>(OutStoreChains[i]);
SDValue NewStore = DAG.getTruncStore(LoadToken, dl, ST->getValue(),
ST->getBasePtr(), ST->getMemoryVT(),
ST->getMemOperand());
OutChains.push_back(NewStore);
}
}
static SDValue getMemcpyLoadsAndStores(SelectionDAG &DAG, const SDLoc &dl,
SDValue Chain, SDValue Dst, SDValue Src,
uint64_t Size, Align Alignment,
bool isVol, bool AlwaysInline,
MachinePointerInfo DstPtrInfo,
MachinePointerInfo SrcPtrInfo,
const AAMDNodes &AAInfo, AAResults *AA) {
// Turn a memcpy of undef to nop.
// FIXME: We need to honor volatile even is Src is undef.
if (Src.isUndef())
return Chain;
// Expand memcpy to a series of load and store ops if the size operand falls
// below a certain threshold.
// TODO: In the AlwaysInline case, if the size is big then generate a loop
// rather than maybe a humongous number of loads and stores.
const TargetLowering &TLI = DAG.getTargetLoweringInfo();
const DataLayout &DL = DAG.getDataLayout();
LLVMContext &C = *DAG.getContext();
std::vector<EVT> MemOps;
bool DstAlignCanChange = false;
MachineFunction &MF = DAG.getMachineFunction();
MachineFrameInfo &MFI = MF.getFrameInfo();
bool OptSize = shouldLowerMemFuncForSize(MF, DAG);
FrameIndexSDNode *FI = dyn_cast<FrameIndexSDNode>(Dst);
if (FI && !MFI.isFixedObjectIndex(FI->getIndex()))
DstAlignCanChange = true;
MaybeAlign SrcAlign = DAG.InferPtrAlign(Src);
if (!SrcAlign || Alignment > *SrcAlign)
SrcAlign = Alignment;
assert(SrcAlign && "SrcAlign must be set");
ConstantDataArraySlice Slice;
// If marked as volatile, perform a copy even when marked as constant.
bool CopyFromConstant = !isVol && isMemSrcFromConstant(Src, Slice);
bool isZeroConstant = CopyFromConstant && Slice.Array == nullptr;
unsigned Limit = AlwaysInline ? ~0U : TLI.getMaxStoresPerMemcpy(OptSize);
const MemOp Op = isZeroConstant
? MemOp::Set(Size, DstAlignCanChange, Alignment,
/*IsZeroMemset*/ true, isVol)
: MemOp::Copy(Size, DstAlignCanChange, Alignment,
*SrcAlign, isVol, CopyFromConstant);
if (!TLI.findOptimalMemOpLowering(
MemOps, Limit, Op, DstPtrInfo.getAddrSpace(),
SrcPtrInfo.getAddrSpace(), MF.getFunction().getAttributes()))
return SDValue();
if (DstAlignCanChange) {
Type *Ty = MemOps[0].getTypeForEVT(C);
Align NewAlign = DL.getABITypeAlign(Ty);
// Don't promote to an alignment that would require dynamic stack
// realignment which may conflict with optimizations such as tail call
// optimization.
const TargetRegisterInfo *TRI = MF.getSubtarget().getRegisterInfo();
if (!TRI->hasStackRealignment(MF))
while (NewAlign > Alignment && DL.exceedsNaturalStackAlignment(NewAlign))
NewAlign = NewAlign.previous();
if (NewAlign > Alignment) {
// Give the stack frame object a larger alignment if needed.
if (MFI.getObjectAlign(FI->getIndex()) < NewAlign)
MFI.setObjectAlignment(FI->getIndex(), NewAlign);
Alignment = NewAlign;
}
}
// Prepare AAInfo for loads/stores after lowering this memcpy.
AAMDNodes NewAAInfo = AAInfo;
NewAAInfo.TBAA = NewAAInfo.TBAAStruct = nullptr;
const Value *SrcVal = dyn_cast_if_present<const Value *>(SrcPtrInfo.V);
bool isConstant =
AA && SrcVal &&
AA->pointsToConstantMemory(MemoryLocation(SrcVal, Size, AAInfo));
MachineMemOperand::Flags MMOFlags =
isVol ? MachineMemOperand::MOVolatile : MachineMemOperand::MONone;
SmallVector<SDValue, 16> OutLoadChains;
SmallVector<SDValue, 16> OutStoreChains;
SmallVector<SDValue, 32> OutChains;
unsigned NumMemOps = MemOps.size();
uint64_t SrcOff = 0, DstOff = 0;
for (unsigned i = 0; i != NumMemOps; ++i) {
EVT VT = MemOps[i];
unsigned VTSize = VT.getSizeInBits() / 8;
SDValue Value, Store;
if (VTSize > Size) {
// Issuing an unaligned load / store pair that overlaps with the previous
// pair. Adjust the offset accordingly.
assert(i == NumMemOps-1 && i != 0);
SrcOff -= VTSize - Size;
DstOff -= VTSize - Size;
}
if (CopyFromConstant &&
(isZeroConstant || (VT.isInteger() && !VT.isVector()))) {
// It's unlikely a store of a vector immediate can be done in a single
// instruction. It would require a load from a constantpool first.
// We only handle zero vectors here.
// FIXME: Handle other cases where store of vector immediate is done in
// a single instruction.
ConstantDataArraySlice SubSlice;
if (SrcOff < Slice.Length) {
SubSlice = Slice;
SubSlice.move(SrcOff);
} else {
// This is an out-of-bounds access and hence UB. Pretend we read zero.
SubSlice.Array = nullptr;
SubSlice.Offset = 0;
SubSlice.Length = VTSize;
}
Value = getMemsetStringVal(VT, dl, DAG, TLI, SubSlice);
if (Value.getNode()) {
Store = DAG.getStore(
Chain, dl, Value,
DAG.getMemBasePlusOffset(Dst, TypeSize::getFixed(DstOff), dl),
DstPtrInfo.getWithOffset(DstOff), Alignment, MMOFlags, NewAAInfo);
OutChains.push_back(Store);
}
}
if (!Store.getNode()) {
// The type might not be legal for the target. This should only happen
// if the type is smaller than a legal type, as on PPC, so the right
// thing to do is generate a LoadExt/StoreTrunc pair. These simplify
// to Load/Store if NVT==VT.
// FIXME does the case above also need this?
EVT NVT = TLI.getTypeToTransformTo(C, VT);
assert(NVT.bitsGE(VT));
bool isDereferenceable =
SrcPtrInfo.getWithOffset(SrcOff).isDereferenceable(VTSize, C, DL);
MachineMemOperand::Flags SrcMMOFlags = MMOFlags;
if (isDereferenceable)
SrcMMOFlags |= MachineMemOperand::MODereferenceable;
if (isConstant)
SrcMMOFlags |= MachineMemOperand::MOInvariant;
Value = DAG.getExtLoad(
ISD::EXTLOAD, dl, NVT, Chain,
DAG.getMemBasePlusOffset(Src, TypeSize::getFixed(SrcOff), dl),
SrcPtrInfo.getWithOffset(SrcOff), VT,
commonAlignment(*SrcAlign, SrcOff), SrcMMOFlags, NewAAInfo);
OutLoadChains.push_back(Value.getValue(1));
Store = DAG.getTruncStore(
Chain, dl, Value,
DAG.getMemBasePlusOffset(Dst, TypeSize::getFixed(DstOff), dl),
DstPtrInfo.getWithOffset(DstOff), VT, Alignment, MMOFlags, NewAAInfo);
OutStoreChains.push_back(Store);
}
SrcOff += VTSize;
DstOff += VTSize;
Size -= VTSize;
}
unsigned GluedLdStLimit = MaxLdStGlue == 0 ?
TLI.getMaxGluedStoresPerMemcpy() : MaxLdStGlue;
unsigned NumLdStInMemcpy = OutStoreChains.size();
if (NumLdStInMemcpy) {
// It may be that memcpy might be converted to memset if it's memcpy
// of constants. In such a case, we won't have loads and stores, but
// just stores. In the absence of loads, there is nothing to gang up.
if ((GluedLdStLimit <= 1) || !EnableMemCpyDAGOpt) {
// If target does not care, just leave as it.
for (unsigned i = 0; i < NumLdStInMemcpy; ++i) {
OutChains.push_back(OutLoadChains[i]);
OutChains.push_back(OutStoreChains[i]);
}
} else {
// Ld/St less than/equal limit set by target.
if (NumLdStInMemcpy <= GluedLdStLimit) {
chainLoadsAndStoresForMemcpy(DAG, dl, OutChains, 0,
NumLdStInMemcpy, OutLoadChains,
OutStoreChains);
} else {
unsigned NumberLdChain = NumLdStInMemcpy / GluedLdStLimit;
unsigned RemainingLdStInMemcpy = NumLdStInMemcpy % GluedLdStLimit;
unsigned GlueIter = 0;
for (unsigned cnt = 0; cnt < NumberLdChain; ++cnt) {
unsigned IndexFrom = NumLdStInMemcpy - GlueIter - GluedLdStLimit;
unsigned IndexTo = NumLdStInMemcpy - GlueIter;
chainLoadsAndStoresForMemcpy(DAG, dl, OutChains, IndexFrom, IndexTo,
OutLoadChains, OutStoreChains);
GlueIter += GluedLdStLimit;
}
// Residual ld/st.
if (RemainingLdStInMemcpy) {
chainLoadsAndStoresForMemcpy(DAG, dl, OutChains, 0,
RemainingLdStInMemcpy, OutLoadChains,
OutStoreChains);
}
}
}
}
return DAG.getNode(ISD::TokenFactor, dl, MVT::Other, OutChains);
}
static SDValue getMemmoveLoadsAndStores(SelectionDAG &DAG, const SDLoc &dl,
SDValue Chain, SDValue Dst, SDValue Src,
uint64_t Size, Align Alignment,
bool isVol, bool AlwaysInline,
MachinePointerInfo DstPtrInfo,
MachinePointerInfo SrcPtrInfo,
const AAMDNodes &AAInfo) {
// Turn a memmove of undef to nop.
// FIXME: We need to honor volatile even is Src is undef.
if (Src.isUndef())
return Chain;
// Expand memmove to a series of load and store ops if the size operand falls
// below a certain threshold.
const TargetLowering &TLI = DAG.getTargetLoweringInfo();
const DataLayout &DL = DAG.getDataLayout();
LLVMContext &C = *DAG.getContext();
std::vector<EVT> MemOps;
bool DstAlignCanChange = false;
MachineFunction &MF = DAG.getMachineFunction();
MachineFrameInfo &MFI = MF.getFrameInfo();
bool OptSize = shouldLowerMemFuncForSize(MF, DAG);
FrameIndexSDNode *FI = dyn_cast<FrameIndexSDNode>(Dst);
if (FI && !MFI.isFixedObjectIndex(FI->getIndex()))
DstAlignCanChange = true;
MaybeAlign SrcAlign = DAG.InferPtrAlign(Src);
if (!SrcAlign || Alignment > *SrcAlign)
SrcAlign = Alignment;
assert(SrcAlign && "SrcAlign must be set");
unsigned Limit = AlwaysInline ? ~0U : TLI.getMaxStoresPerMemmove(OptSize);
if (!TLI.findOptimalMemOpLowering(
MemOps, Limit,
MemOp::Copy(Size, DstAlignCanChange, Alignment, *SrcAlign,
/*IsVolatile*/ true),
DstPtrInfo.getAddrSpace(), SrcPtrInfo.getAddrSpace(),
MF.getFunction().getAttributes()))
return SDValue();
if (DstAlignCanChange) {
Type *Ty = MemOps[0].getTypeForEVT(C);
Align NewAlign = DL.getABITypeAlign(Ty);
// Don't promote to an alignment that would require dynamic stack
// realignment which may conflict with optimizations such as tail call
// optimization.
const TargetRegisterInfo *TRI = MF.getSubtarget().getRegisterInfo();
if (!TRI->hasStackRealignment(MF))
while (NewAlign > Alignment && DL.exceedsNaturalStackAlignment(NewAlign))
NewAlign = NewAlign.previous();
if (NewAlign > Alignment) {
// Give the stack frame object a larger alignment if needed.
if (MFI.getObjectAlign(FI->getIndex()) < NewAlign)
MFI.setObjectAlignment(FI->getIndex(), NewAlign);
Alignment = NewAlign;
}
}
// Prepare AAInfo for loads/stores after lowering this memmove.
AAMDNodes NewAAInfo = AAInfo;
NewAAInfo.TBAA = NewAAInfo.TBAAStruct = nullptr;
MachineMemOperand::Flags MMOFlags =
isVol ? MachineMemOperand::MOVolatile : MachineMemOperand::MONone;
uint64_t SrcOff = 0, DstOff = 0;
SmallVector<SDValue, 8> LoadValues;
SmallVector<SDValue, 8> LoadChains;
SmallVector<SDValue, 8> OutChains;
unsigned NumMemOps = MemOps.size();
for (unsigned i = 0; i < NumMemOps; i++) {
EVT VT = MemOps[i];
unsigned VTSize = VT.getSizeInBits() / 8;
SDValue Value;
bool isDereferenceable =
SrcPtrInfo.getWithOffset(SrcOff).isDereferenceable(VTSize, C, DL);
MachineMemOperand::Flags SrcMMOFlags = MMOFlags;
if (isDereferenceable)
SrcMMOFlags |= MachineMemOperand::MODereferenceable;
Value = DAG.getLoad(
VT, dl, Chain,
DAG.getMemBasePlusOffset(Src, TypeSize::getFixed(SrcOff), dl),
SrcPtrInfo.getWithOffset(SrcOff), *SrcAlign, SrcMMOFlags, NewAAInfo);
LoadValues.push_back(Value);
LoadChains.push_back(Value.getValue(1));
SrcOff += VTSize;
}
Chain = DAG.getNode(ISD::TokenFactor, dl, MVT::Other, LoadChains);
OutChains.clear();
for (unsigned i = 0; i < NumMemOps; i++) {
EVT VT = MemOps[i];
unsigned VTSize = VT.getSizeInBits() / 8;
SDValue Store;
Store = DAG.getStore(
Chain, dl, LoadValues[i],
DAG.getMemBasePlusOffset(Dst, TypeSize::getFixed(DstOff), dl),
DstPtrInfo.getWithOffset(DstOff), Alignment, MMOFlags, NewAAInfo);
OutChains.push_back(Store);
DstOff += VTSize;
}
return DAG.getNode(ISD::TokenFactor, dl, MVT::Other, OutChains);
}
/// Lower the call to 'memset' intrinsic function into a series of store
/// operations.
///
/// \param DAG Selection DAG where lowered code is placed.
/// \param dl Link to corresponding IR location.
/// \param Chain Control flow dependency.
/// \param Dst Pointer to destination memory location.
/// \param Src Value of byte to write into the memory.
/// \param Size Number of bytes to write.
/// \param Alignment Alignment of the destination in bytes.
/// \param isVol True if destination is volatile.
/// \param AlwaysInline Makes sure no function call is generated.
/// \param DstPtrInfo IR information on the memory pointer.
/// \returns New head in the control flow, if lowering was successful, empty
/// SDValue otherwise.
///
/// The function tries to replace 'llvm.memset' intrinsic with several store
/// operations and value calculation code. This is usually profitable for small
/// memory size or when the semantic requires inlining.
static SDValue getMemsetStores(SelectionDAG &DAG, const SDLoc &dl,
SDValue Chain, SDValue Dst, SDValue Src,
uint64_t Size, Align Alignment, bool isVol,
bool AlwaysInline, MachinePointerInfo DstPtrInfo,
const AAMDNodes &AAInfo) {
// Turn a memset of undef to nop.
// FIXME: We need to honor volatile even is Src is undef.
if (Src.isUndef())
return Chain;
// Expand memset to a series of load/store ops if the size operand
// falls below a certain threshold.
const TargetLowering &TLI = DAG.getTargetLoweringInfo();
std::vector<EVT> MemOps;
bool DstAlignCanChange = false;
MachineFunction &MF = DAG.getMachineFunction();
MachineFrameInfo &MFI = MF.getFrameInfo();
bool OptSize = shouldLowerMemFuncForSize(MF, DAG);
FrameIndexSDNode *FI = dyn_cast<FrameIndexSDNode>(Dst);
if (FI && !MFI.isFixedObjectIndex(FI->getIndex()))
DstAlignCanChange = true;
bool IsZeroVal = isNullConstant(Src);
unsigned Limit = AlwaysInline ? ~0 : TLI.getMaxStoresPerMemset(OptSize);
if (!TLI.findOptimalMemOpLowering(
MemOps, Limit,
MemOp::Set(Size, DstAlignCanChange, Alignment, IsZeroVal, isVol),
DstPtrInfo.getAddrSpace(), ~0u, MF.getFunction().getAttributes()))
return SDValue();
if (DstAlignCanChange) {
Type *Ty = MemOps[0].getTypeForEVT(*DAG.getContext());
const DataLayout &DL = DAG.getDataLayout();
Align NewAlign = DL.getABITypeAlign(Ty);
// Don't promote to an alignment that would require dynamic stack
// realignment which may conflict with optimizations such as tail call
// optimization.
const TargetRegisterInfo *TRI = MF.getSubtarget().getRegisterInfo();
if (!TRI->hasStackRealignment(MF))
while (NewAlign > Alignment && DL.exceedsNaturalStackAlignment(NewAlign))
NewAlign = NewAlign.previous();
if (NewAlign > Alignment) {
// Give the stack frame object a larger alignment if needed.
if (MFI.getObjectAlign(FI->getIndex()) < NewAlign)
MFI.setObjectAlignment(FI->getIndex(), NewAlign);
Alignment = NewAlign;
}
}
SmallVector<SDValue, 8> OutChains;
uint64_t DstOff = 0;
unsigned NumMemOps = MemOps.size();
// Find the largest store and generate the bit pattern for it.
EVT LargestVT = MemOps[0];
for (unsigned i = 1; i < NumMemOps; i++)
if (MemOps[i].bitsGT(LargestVT))
LargestVT = MemOps[i];
SDValue MemSetValue = getMemsetValue(Src, LargestVT, DAG, dl);
// Prepare AAInfo for loads/stores after lowering this memset.
AAMDNodes NewAAInfo = AAInfo;
NewAAInfo.TBAA = NewAAInfo.TBAAStruct = nullptr;
for (unsigned i = 0; i < NumMemOps; i++) {
EVT VT = MemOps[i];
unsigned VTSize = VT.getSizeInBits() / 8;
if (VTSize > Size) {
// Issuing an unaligned load / store pair that overlaps with the previous
// pair. Adjust the offset accordingly.
assert(i == NumMemOps-1 && i != 0);
DstOff -= VTSize - Size;
}
// If this store is smaller than the largest store see whether we can get
// the smaller value for free with a truncate or extract vector element and
// then store.
SDValue Value = MemSetValue;
if (VT.bitsLT(LargestVT)) {
unsigned Index;
unsigned NElts = LargestVT.getSizeInBits() / VT.getSizeInBits();
EVT SVT = EVT::getVectorVT(*DAG.getContext(), VT.getScalarType(), NElts);
if (!LargestVT.isVector() && !VT.isVector() &&
TLI.isTruncateFree(LargestVT, VT))
Value = DAG.getNode(ISD::TRUNCATE, dl, VT, MemSetValue);
else if (LargestVT.isVector() && !VT.isVector() &&
TLI.shallExtractConstSplatVectorElementToStore(
LargestVT.getTypeForEVT(*DAG.getContext()),
VT.getSizeInBits(), Index) &&
TLI.isTypeLegal(SVT) &&
LargestVT.getSizeInBits() == SVT.getSizeInBits()) {
// Target which can combine store(extractelement VectorTy, Idx) can get
// the smaller value for free.
SDValue TailValue = DAG.getNode(ISD::BITCAST, dl, SVT, MemSetValue);
Value = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, dl, VT, TailValue,
DAG.getVectorIdxConstant(Index, dl));
} else
Value = getMemsetValue(Src, VT, DAG, dl);
}
assert(Value.getValueType() == VT && "Value with wrong type.");
SDValue Store = DAG.getStore(
Chain, dl, Value,
DAG.getMemBasePlusOffset(Dst, TypeSize::getFixed(DstOff), dl),
DstPtrInfo.getWithOffset(DstOff), Alignment,
isVol ? MachineMemOperand::MOVolatile : MachineMemOperand::MONone,
NewAAInfo);
OutChains.push_back(Store);
DstOff += VT.getSizeInBits() / 8;
Size -= VTSize;
}
return DAG.getNode(ISD::TokenFactor, dl, MVT::Other, OutChains);
}
static void checkAddrSpaceIsValidForLibcall(const TargetLowering *TLI,
unsigned AS) {
// Lowering memcpy / memset / memmove intrinsics to calls is only valid if all
// pointer operands can be losslessly bitcasted to pointers of address space 0
if (AS != 0 && !TLI->getTargetMachine().isNoopAddrSpaceCast(AS, 0)) {
report_fatal_error("cannot lower memory intrinsic in address space " +
Twine(AS));
}
}
SDValue SelectionDAG::getMemcpy(SDValue Chain, const SDLoc &dl, SDValue Dst,
SDValue Src, SDValue Size, Align Alignment,
bool isVol, bool AlwaysInline, bool isTailCall,
MachinePointerInfo DstPtrInfo,
MachinePointerInfo SrcPtrInfo,
const AAMDNodes &AAInfo, AAResults *AA) {
// Check to see if we should lower the memcpy to loads and stores first.
// For cases within the target-specified limits, this is the best choice.
ConstantSDNode *ConstantSize = dyn_cast<ConstantSDNode>(Size);
if (ConstantSize) {
// Memcpy with size zero? Just return the original chain.
if (ConstantSize->isZero())
return Chain;
SDValue Result = getMemcpyLoadsAndStores(
*this, dl, Chain, Dst, Src, ConstantSize->getZExtValue(), Alignment,
isVol, false, DstPtrInfo, SrcPtrInfo, AAInfo, AA);
if (Result.getNode())
return Result;
}
// Then check to see if we should lower the memcpy with target-specific
// code. If the target chooses to do this, this is the next best.
if (TSI) {
SDValue Result = TSI->EmitTargetCodeForMemcpy(
*this, dl, Chain, Dst, Src, Size, Alignment, isVol, AlwaysInline,
DstPtrInfo, SrcPtrInfo);
if (Result.getNode())
return Result;
}
// If we really need inline code and the target declined to provide it,
// use a (potentially long) sequence of loads and stores.
if (AlwaysInline) {
assert(ConstantSize && "AlwaysInline requires a constant size!");
return getMemcpyLoadsAndStores(
*this, dl, Chain, Dst, Src, ConstantSize->getZExtValue(), Alignment,
isVol, true, DstPtrInfo, SrcPtrInfo, AAInfo, AA);
}
checkAddrSpaceIsValidForLibcall(TLI, DstPtrInfo.getAddrSpace());
checkAddrSpaceIsValidForLibcall(TLI, SrcPtrInfo.getAddrSpace());
// FIXME: If the memcpy is volatile (isVol), lowering it to a plain libc
// memcpy is not guaranteed to be safe. libc memcpys aren't required to
// respect volatile, so they may do things like read or write memory
// beyond the given memory regions. But fixing this isn't easy, and most
// people don't care.
// Emit a library call.
TargetLowering::ArgListTy Args;
TargetLowering::ArgListEntry Entry;
Entry.Ty = PointerType::getUnqual(*getContext());
Entry.Node = Dst; Args.push_back(Entry);
Entry.Node = Src; Args.push_back(Entry);
Entry.Ty = getDataLayout().getIntPtrType(*getContext());
Entry.Node = Size; Args.push_back(Entry);
// FIXME: pass in SDLoc
TargetLowering::CallLoweringInfo CLI(*this);
CLI.setDebugLoc(dl)
.setChain(Chain)
.setLibCallee(TLI->getLibcallCallingConv(RTLIB::MEMCPY),
Dst.getValueType().getTypeForEVT(*getContext()),
getExternalSymbol(TLI->getLibcallName(RTLIB::MEMCPY),
TLI->getPointerTy(getDataLayout())),
std::move(Args))
.setDiscardResult()
.setTailCall(isTailCall);
std::pair<SDValue,SDValue> CallResult = TLI->LowerCallTo(CLI);
return CallResult.second;
}
SDValue SelectionDAG::getAtomicMemcpy(SDValue Chain, const SDLoc &dl,
SDValue Dst, SDValue Src, SDValue Size,
Type *SizeTy, unsigned ElemSz,
bool isTailCall,
MachinePointerInfo DstPtrInfo,
MachinePointerInfo SrcPtrInfo) {
// Emit a library call.
TargetLowering::ArgListTy Args;
TargetLowering::ArgListEntry Entry;
Entry.Ty = getDataLayout().getIntPtrType(*getContext());
Entry.Node = Dst;
Args.push_back(Entry);
Entry.Node = Src;
Args.push_back(Entry);
Entry.Ty = SizeTy;
Entry.Node = Size;
Args.push_back(Entry);
RTLIB::Libcall LibraryCall =
RTLIB::getMEMCPY_ELEMENT_UNORDERED_ATOMIC(ElemSz);
if (LibraryCall == RTLIB::UNKNOWN_LIBCALL)
report_fatal_error("Unsupported element size");
TargetLowering::CallLoweringInfo CLI(*this);
CLI.setDebugLoc(dl)
.setChain(Chain)
.setLibCallee(TLI->getLibcallCallingConv(LibraryCall),
Type::getVoidTy(*getContext()),
getExternalSymbol(TLI->getLibcallName(LibraryCall),
TLI->getPointerTy(getDataLayout())),
std::move(Args))
.setDiscardResult()
.setTailCall(isTailCall);
std::pair<SDValue, SDValue> CallResult = TLI->LowerCallTo(CLI);
return CallResult.second;
}
SDValue SelectionDAG::getMemmove(SDValue Chain, const SDLoc &dl, SDValue Dst,
SDValue Src, SDValue Size, Align Alignment,
bool isVol, bool isTailCall,
MachinePointerInfo DstPtrInfo,
MachinePointerInfo SrcPtrInfo,
const AAMDNodes &AAInfo, AAResults *AA) {
// Check to see if we should lower the memmove to loads and stores first.
// For cases within the target-specified limits, this is the best choice.
ConstantSDNode *ConstantSize = dyn_cast<ConstantSDNode>(Size);
if (ConstantSize) {
// Memmove with size zero? Just return the original chain.
if (ConstantSize->isZero())
return Chain;
SDValue Result = getMemmoveLoadsAndStores(
*this, dl, Chain, Dst, Src, ConstantSize->getZExtValue(), Alignment,
isVol, false, DstPtrInfo, SrcPtrInfo, AAInfo);
if (Result.getNode())
return Result;
}
// Then check to see if we should lower the memmove with target-specific
// code. If the target chooses to do this, this is the next best.
if (TSI) {
SDValue Result =
TSI->EmitTargetCodeForMemmove(*this, dl, Chain, Dst, Src, Size,
Alignment, isVol, DstPtrInfo, SrcPtrInfo);
if (Result.getNode())
return Result;
}
checkAddrSpaceIsValidForLibcall(TLI, DstPtrInfo.getAddrSpace());
checkAddrSpaceIsValidForLibcall(TLI, SrcPtrInfo.getAddrSpace());
// FIXME: If the memmove is volatile, lowering it to plain libc memmove may
// not be safe. See memcpy above for more details.
// Emit a library call.
TargetLowering::ArgListTy Args;
TargetLowering::ArgListEntry Entry;
Entry.Ty = PointerType::getUnqual(*getContext());
Entry.Node = Dst; Args.push_back(Entry);
Entry.Node = Src; Args.push_back(Entry);
Entry.Ty = getDataLayout().getIntPtrType(*getContext());
Entry.Node = Size; Args.push_back(Entry);
// FIXME: pass in SDLoc
TargetLowering::CallLoweringInfo CLI(*this);
CLI.setDebugLoc(dl)
.setChain(Chain)
.setLibCallee(TLI->getLibcallCallingConv(RTLIB::MEMMOVE),
Dst.getValueType().getTypeForEVT(*getContext()),
getExternalSymbol(TLI->getLibcallName(RTLIB::MEMMOVE),
TLI->getPointerTy(getDataLayout())),
std::move(Args))
.setDiscardResult()
.setTailCall(isTailCall);
std::pair<SDValue,SDValue> CallResult = TLI->LowerCallTo(CLI);
return CallResult.second;
}
SDValue SelectionDAG::getAtomicMemmove(SDValue Chain, const SDLoc &dl,
SDValue Dst, SDValue Src, SDValue Size,
Type *SizeTy, unsigned ElemSz,
bool isTailCall,
MachinePointerInfo DstPtrInfo,
MachinePointerInfo SrcPtrInfo) {
// Emit a library call.
TargetLowering::ArgListTy Args;
TargetLowering::ArgListEntry Entry;
Entry.Ty = getDataLayout().getIntPtrType(*getContext());
Entry.Node = Dst;
Args.push_back(Entry);
Entry.Node = Src;
Args.push_back(Entry);
Entry.Ty = SizeTy;
Entry.Node = Size;
Args.push_back(Entry);
RTLIB::Libcall LibraryCall =
RTLIB::getMEMMOVE_ELEMENT_UNORDERED_ATOMIC(ElemSz);
if (LibraryCall == RTLIB::UNKNOWN_LIBCALL)
report_fatal_error("Unsupported element size");
TargetLowering::CallLoweringInfo CLI(*this);
CLI.setDebugLoc(dl)
.setChain(Chain)
.setLibCallee(TLI->getLibcallCallingConv(LibraryCall),
Type::getVoidTy(*getContext()),
getExternalSymbol(TLI->getLibcallName(LibraryCall),
TLI->getPointerTy(getDataLayout())),
std::move(Args))
.setDiscardResult()
.setTailCall(isTailCall);
std::pair<SDValue, SDValue> CallResult = TLI->LowerCallTo(CLI);
return CallResult.second;
}
SDValue SelectionDAG::getMemset(SDValue Chain, const SDLoc &dl, SDValue Dst,
SDValue Src, SDValue Size, Align Alignment,
bool isVol, bool AlwaysInline, bool isTailCall,
MachinePointerInfo DstPtrInfo,
const AAMDNodes &AAInfo) {
// Check to see if we should lower the memset to stores first.
// For cases within the target-specified limits, this is the best choice.
ConstantSDNode *ConstantSize = dyn_cast<ConstantSDNode>(Size);
if (ConstantSize) {
// Memset with size zero? Just return the original chain.
if (ConstantSize->isZero())
return Chain;
SDValue Result = getMemsetStores(*this, dl, Chain, Dst, Src,
ConstantSize->getZExtValue(), Alignment,
isVol, false, DstPtrInfo, AAInfo);
if (Result.getNode())
return Result;
}
// Then check to see if we should lower the memset with target-specific
// code. If the target chooses to do this, this is the next best.
if (TSI) {
SDValue Result = TSI->EmitTargetCodeForMemset(
*this, dl, Chain, Dst, Src, Size, Alignment, isVol, AlwaysInline, DstPtrInfo);
if (Result.getNode())
return Result;
}
// If we really need inline code and the target declined to provide it,
// use a (potentially long) sequence of loads and stores.
if (AlwaysInline) {
assert(ConstantSize && "AlwaysInline requires a constant size!");
SDValue Result = getMemsetStores(*this, dl, Chain, Dst, Src,
ConstantSize->getZExtValue(), Alignment,
isVol, true, DstPtrInfo, AAInfo);
assert(Result &&
"getMemsetStores must return a valid sequence when AlwaysInline");
return Result;
}
checkAddrSpaceIsValidForLibcall(TLI, DstPtrInfo.getAddrSpace());
// Emit a library call.
auto &Ctx = *getContext();
const auto& DL = getDataLayout();
TargetLowering::CallLoweringInfo CLI(*this);
// FIXME: pass in SDLoc
CLI.setDebugLoc(dl).setChain(Chain);
const char *BzeroName = getTargetLoweringInfo().getLibcallName(RTLIB::BZERO);
// Helper function to create an Entry from Node and Type.
const auto CreateEntry = [](SDValue Node, Type *Ty) {
TargetLowering::ArgListEntry Entry;
Entry.Node = Node;
Entry.Ty = Ty;
return Entry;
};
// If zeroing out and bzero is present, use it.
if (isNullConstant(Src) && BzeroName) {
TargetLowering::ArgListTy Args;
Args.push_back(CreateEntry(Dst, PointerType::getUnqual(Ctx)));
Args.push_back(CreateEntry(Size, DL.getIntPtrType(Ctx)));
CLI.setLibCallee(
TLI->getLibcallCallingConv(RTLIB::BZERO), Type::getVoidTy(Ctx),
getExternalSymbol(BzeroName, TLI->getPointerTy(DL)), std::move(Args));
} else {
TargetLowering::ArgListTy Args;
Args.push_back(CreateEntry(Dst, PointerType::getUnqual(Ctx)));
Args.push_back(CreateEntry(Src, Src.getValueType().getTypeForEVT(Ctx)));
Args.push_back(CreateEntry(Size, DL.getIntPtrType(Ctx)));
CLI.setLibCallee(TLI->getLibcallCallingConv(RTLIB::MEMSET),
Dst.getValueType().getTypeForEVT(Ctx),
getExternalSymbol(TLI->getLibcallName(RTLIB::MEMSET),
TLI->getPointerTy(DL)),
std::move(Args));
}
CLI.setDiscardResult().setTailCall(isTailCall);
std::pair<SDValue, SDValue> CallResult = TLI->LowerCallTo(CLI);
return CallResult.second;
}
SDValue SelectionDAG::getAtomicMemset(SDValue Chain, const SDLoc &dl,
SDValue Dst, SDValue Value, SDValue Size,
Type *SizeTy, unsigned ElemSz,
bool isTailCall,
MachinePointerInfo DstPtrInfo) {
// Emit a library call.
TargetLowering::ArgListTy Args;
TargetLowering::ArgListEntry Entry;
Entry.Ty = getDataLayout().getIntPtrType(*getContext());
Entry.Node = Dst;
Args.push_back(Entry);
Entry.Ty = Type::getInt8Ty(*getContext());
Entry.Node = Value;
Args.push_back(Entry);
Entry.Ty = SizeTy;
Entry.Node = Size;
Args.push_back(Entry);
RTLIB::Libcall LibraryCall =
RTLIB::getMEMSET_ELEMENT_UNORDERED_ATOMIC(ElemSz);
if (LibraryCall == RTLIB::UNKNOWN_LIBCALL)
report_fatal_error("Unsupported element size");
TargetLowering::CallLoweringInfo CLI(*this);
CLI.setDebugLoc(dl)
.setChain(Chain)
.setLibCallee(TLI->getLibcallCallingConv(LibraryCall),
Type::getVoidTy(*getContext()),
getExternalSymbol(TLI->getLibcallName(LibraryCall),
TLI->getPointerTy(getDataLayout())),
std::move(Args))
.setDiscardResult()
.setTailCall(isTailCall);
std::pair<SDValue, SDValue> CallResult = TLI->LowerCallTo(CLI);
return CallResult.second;
}
SDValue SelectionDAG::getAtomic(unsigned Opcode, const SDLoc &dl, EVT MemVT,
SDVTList VTList, ArrayRef<SDValue> Ops,
MachineMemOperand *MMO) {
FoldingSetNodeID ID;
ID.AddInteger(MemVT.getRawBits());
AddNodeIDNode(ID, Opcode, VTList, Ops);
ID.AddInteger(MMO->getPointerInfo().getAddrSpace());
ID.AddInteger(MMO->getFlags());
void* IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, dl, IP)) {
cast<AtomicSDNode>(E)->refineAlignment(MMO);
return SDValue(E, 0);
}
auto *N = newSDNode<AtomicSDNode>(Opcode, dl.getIROrder(), dl.getDebugLoc(),
VTList, MemVT, MMO);
createOperands(N, Ops);
CSEMap.InsertNode(N, IP);
InsertNode(N);
return SDValue(N, 0);
}
SDValue SelectionDAG::getAtomicCmpSwap(unsigned Opcode, const SDLoc &dl,
EVT MemVT, SDVTList VTs, SDValue Chain,
SDValue Ptr, SDValue Cmp, SDValue Swp,
MachineMemOperand *MMO) {
assert(Opcode == ISD::ATOMIC_CMP_SWAP ||
Opcode == ISD::ATOMIC_CMP_SWAP_WITH_SUCCESS);
assert(Cmp.getValueType() == Swp.getValueType() && "Invalid Atomic Op Types");
SDValue Ops[] = {Chain, Ptr, Cmp, Swp};
return getAtomic(Opcode, dl, MemVT, VTs, Ops, MMO);
}
SDValue SelectionDAG::getAtomic(unsigned Opcode, const SDLoc &dl, EVT MemVT,
SDValue Chain, SDValue Ptr, SDValue Val,
MachineMemOperand *MMO) {
assert((Opcode == ISD::ATOMIC_LOAD_ADD ||
Opcode == ISD::ATOMIC_LOAD_SUB ||
Opcode == ISD::ATOMIC_LOAD_AND ||
Opcode == ISD::ATOMIC_LOAD_CLR ||
Opcode == ISD::ATOMIC_LOAD_OR ||
Opcode == ISD::ATOMIC_LOAD_XOR ||
Opcode == ISD::ATOMIC_LOAD_NAND ||
Opcode == ISD::ATOMIC_LOAD_MIN ||
Opcode == ISD::ATOMIC_LOAD_MAX ||
Opcode == ISD::ATOMIC_LOAD_UMIN ||
Opcode == ISD::ATOMIC_LOAD_UMAX ||
Opcode == ISD::ATOMIC_LOAD_FADD ||
Opcode == ISD::ATOMIC_LOAD_FSUB ||
Opcode == ISD::ATOMIC_LOAD_FMAX ||
Opcode == ISD::ATOMIC_LOAD_FMIN ||
Opcode == ISD::ATOMIC_LOAD_UINC_WRAP ||
Opcode == ISD::ATOMIC_LOAD_UDEC_WRAP ||
Opcode == ISD::ATOMIC_SWAP ||
Opcode == ISD::ATOMIC_STORE) &&
"Invalid Atomic Op");
EVT VT = Val.getValueType();
SDVTList VTs = Opcode == ISD::ATOMIC_STORE ? getVTList(MVT::Other) :
getVTList(VT, MVT::Other);
SDValue Ops[] = {Chain, Ptr, Val};
return getAtomic(Opcode, dl, MemVT, VTs, Ops, MMO);
}
SDValue SelectionDAG::getAtomic(unsigned Opcode, const SDLoc &dl, EVT MemVT,
EVT VT, SDValue Chain, SDValue Ptr,
MachineMemOperand *MMO) {
assert(Opcode == ISD::ATOMIC_LOAD && "Invalid Atomic Op");
SDVTList VTs = getVTList(VT, MVT::Other);
SDValue Ops[] = {Chain, Ptr};
return getAtomic(Opcode, dl, MemVT, VTs, Ops, MMO);
}
/// getMergeValues - Create a MERGE_VALUES node from the given operands.
SDValue SelectionDAG::getMergeValues(ArrayRef<SDValue> Ops, const SDLoc &dl) {
if (Ops.size() == 1)
return Ops[0];
SmallVector<EVT, 4> VTs;
VTs.reserve(Ops.size());
for (const SDValue &Op : Ops)
VTs.push_back(Op.getValueType());
return getNode(ISD::MERGE_VALUES, dl, getVTList(VTs), Ops);
}
SDValue SelectionDAG::getMemIntrinsicNode(
unsigned Opcode, const SDLoc &dl, SDVTList VTList, ArrayRef<SDValue> Ops,
EVT MemVT, MachinePointerInfo PtrInfo, Align Alignment,
MachineMemOperand::Flags Flags, uint64_t Size, const AAMDNodes &AAInfo) {
if (!Size && MemVT.isScalableVector())
Size = MemoryLocation::UnknownSize;
else if (!Size)
Size = MemVT.getStoreSize();
MachineFunction &MF = getMachineFunction();
MachineMemOperand *MMO =
MF.getMachineMemOperand(PtrInfo, Flags, Size, Alignment, AAInfo);
return getMemIntrinsicNode(Opcode, dl, VTList, Ops, MemVT, MMO);
}
SDValue SelectionDAG::getMemIntrinsicNode(unsigned Opcode, const SDLoc &dl,
SDVTList VTList,
ArrayRef<SDValue> Ops, EVT MemVT,
MachineMemOperand *MMO) {
assert((Opcode == ISD::INTRINSIC_VOID ||
Opcode == ISD::INTRINSIC_W_CHAIN ||
Opcode == ISD::PREFETCH ||
(Opcode <= (unsigned)std::numeric_limits<int>::max() &&
(int)Opcode >= ISD::FIRST_TARGET_MEMORY_OPCODE)) &&
"Opcode is not a memory-accessing opcode!");
// Memoize the node unless it returns a glue result.
MemIntrinsicSDNode *N;
if (VTList.VTs[VTList.NumVTs-1] != MVT::Glue) {
FoldingSetNodeID ID;
AddNodeIDNode(ID, Opcode, VTList, Ops);
ID.AddInteger(getSyntheticNodeSubclassData<MemIntrinsicSDNode>(
Opcode, dl.getIROrder(), VTList, MemVT, MMO));
ID.AddInteger(MMO->getPointerInfo().getAddrSpace());
ID.AddInteger(MMO->getFlags());
ID.AddInteger(MemVT.getRawBits());
void *IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, dl, IP)) {
cast<MemIntrinsicSDNode>(E)->refineAlignment(MMO);
return SDValue(E, 0);
}
N = newSDNode<MemIntrinsicSDNode>(Opcode, dl.getIROrder(), dl.getDebugLoc(),
VTList, MemVT, MMO);
createOperands(N, Ops);
CSEMap.InsertNode(N, IP);
} else {
N = newSDNode<MemIntrinsicSDNode>(Opcode, dl.getIROrder(), dl.getDebugLoc(),
VTList, MemVT, MMO);
createOperands(N, Ops);
}
InsertNode(N);
SDValue V(N, 0);
NewSDValueDbgMsg(V, "Creating new node: ", this);
return V;
}
SDValue SelectionDAG::getLifetimeNode(bool IsStart, const SDLoc &dl,
SDValue Chain, int FrameIndex,
int64_t Size, int64_t Offset) {
const unsigned Opcode = IsStart ? ISD::LIFETIME_START : ISD::LIFETIME_END;
const auto VTs = getVTList(MVT::Other);
SDValue Ops[2] = {
Chain,
getFrameIndex(FrameIndex,
getTargetLoweringInfo().getFrameIndexTy(getDataLayout()),
true)};
FoldingSetNodeID ID;
AddNodeIDNode(ID, Opcode, VTs, Ops);
ID.AddInteger(FrameIndex);
ID.AddInteger(Size);
ID.AddInteger(Offset);
void *IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, dl, IP))
return SDValue(E, 0);
LifetimeSDNode *N = newSDNode<LifetimeSDNode>(
Opcode, dl.getIROrder(), dl.getDebugLoc(), VTs, Size, Offset);
createOperands(N, Ops);
CSEMap.InsertNode(N, IP);
InsertNode(N);
SDValue V(N, 0);
NewSDValueDbgMsg(V, "Creating new node: ", this);
return V;
}
SDValue SelectionDAG::getPseudoProbeNode(const SDLoc &Dl, SDValue Chain,
uint64_t Guid, uint64_t Index,
uint32_t Attr) {
const unsigned Opcode = ISD::PSEUDO_PROBE;
const auto VTs = getVTList(MVT::Other);
SDValue Ops[] = {Chain};
FoldingSetNodeID ID;
AddNodeIDNode(ID, Opcode, VTs, Ops);
ID.AddInteger(Guid);
ID.AddInteger(Index);
void *IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, Dl, IP))
return SDValue(E, 0);
auto *N = newSDNode<PseudoProbeSDNode>(
Opcode, Dl.getIROrder(), Dl.getDebugLoc(), VTs, Guid, Index, Attr);
createOperands(N, Ops);
CSEMap.InsertNode(N, IP);
InsertNode(N);
SDValue V(N, 0);
NewSDValueDbgMsg(V, "Creating new node: ", this);
return V;
}
/// InferPointerInfo - If the specified ptr/offset is a frame index, infer a
/// MachinePointerInfo record from it. This is particularly useful because the
/// code generator has many cases where it doesn't bother passing in a
/// MachinePointerInfo to getLoad or getStore when it has "FI+Cst".
static MachinePointerInfo InferPointerInfo(const MachinePointerInfo &Info,
SelectionDAG &DAG, SDValue Ptr,
int64_t Offset = 0) {
// If this is FI+Offset, we can model it.
if (const FrameIndexSDNode *FI = dyn_cast<FrameIndexSDNode>(Ptr))
return MachinePointerInfo::getFixedStack(DAG.getMachineFunction(),
FI->getIndex(), Offset);
// If this is (FI+Offset1)+Offset2, we can model it.
if (Ptr.getOpcode() != ISD::ADD ||
!isa<ConstantSDNode>(Ptr.getOperand(1)) ||
!isa<FrameIndexSDNode>(Ptr.getOperand(0)))
return Info;
int FI = cast<FrameIndexSDNode>(Ptr.getOperand(0))->getIndex();
return MachinePointerInfo::getFixedStack(
DAG.getMachineFunction(), FI,
Offset + cast<ConstantSDNode>(Ptr.getOperand(1))->getSExtValue());
}
/// InferPointerInfo - If the specified ptr/offset is a frame index, infer a
/// MachinePointerInfo record from it. This is particularly useful because the
/// code generator has many cases where it doesn't bother passing in a
/// MachinePointerInfo to getLoad or getStore when it has "FI+Cst".
static MachinePointerInfo InferPointerInfo(const MachinePointerInfo &Info,
SelectionDAG &DAG, SDValue Ptr,
SDValue OffsetOp) {
// If the 'Offset' value isn't a constant, we can't handle this.
if (ConstantSDNode *OffsetNode = dyn_cast<ConstantSDNode>(OffsetOp))
return InferPointerInfo(Info, DAG, Ptr, OffsetNode->getSExtValue());
if (OffsetOp.isUndef())
return InferPointerInfo(Info, DAG, Ptr);
return Info;
}
SDValue SelectionDAG::getLoad(ISD::MemIndexedMode AM, ISD::LoadExtType ExtType,
EVT VT, const SDLoc &dl, SDValue Chain,
SDValue Ptr, SDValue Offset,
MachinePointerInfo PtrInfo, EVT MemVT,
Align Alignment,
MachineMemOperand::Flags MMOFlags,
const AAMDNodes &AAInfo, const MDNode *Ranges) {
assert(Chain.getValueType() == MVT::Other &&
"Invalid chain type");
MMOFlags |= MachineMemOperand::MOLoad;
assert((MMOFlags & MachineMemOperand::MOStore) == 0);
// If we don't have a PtrInfo, infer the trivial frame index case to simplify
// clients.
if (PtrInfo.V.isNull())
PtrInfo = InferPointerInfo(PtrInfo, *this, Ptr, Offset);
uint64_t Size = MemoryLocation::getSizeOrUnknown(MemVT.getStoreSize());
MachineFunction &MF = getMachineFunction();
MachineMemOperand *MMO = MF.getMachineMemOperand(PtrInfo, MMOFlags, Size,
Alignment, AAInfo, Ranges);
return getLoad(AM, ExtType, VT, dl, Chain, Ptr, Offset, MemVT, MMO);
}
SDValue SelectionDAG::getLoad(ISD::MemIndexedMode AM, ISD::LoadExtType ExtType,
EVT VT, const SDLoc &dl, SDValue Chain,
SDValue Ptr, SDValue Offset, EVT MemVT,
MachineMemOperand *MMO) {
if (VT == MemVT) {
ExtType = ISD::NON_EXTLOAD;
} else if (ExtType == ISD::NON_EXTLOAD) {
assert(VT == MemVT && "Non-extending load from different memory type!");
} else {
// Extending load.
assert(MemVT.getScalarType().bitsLT(VT.getScalarType()) &&
"Should only be an extending load, not truncating!");
assert(VT.isInteger() == MemVT.isInteger() &&
"Cannot convert from FP to Int or Int -> FP!");
assert(VT.isVector() == MemVT.isVector() &&
"Cannot use an ext load to convert to or from a vector!");
assert((!VT.isVector() ||
VT.getVectorElementCount() == MemVT.getVectorElementCount()) &&
"Cannot use an ext load to change the number of vector elements!");
}
bool Indexed = AM != ISD::UNINDEXED;
assert((Indexed || Offset.isUndef()) && "Unindexed load with an offset!");
SDVTList VTs = Indexed ?
getVTList(VT, Ptr.getValueType(), MVT::Other) : getVTList(VT, MVT::Other);
SDValue Ops[] = { Chain, Ptr, Offset };
FoldingSetNodeID ID;
AddNodeIDNode(ID, ISD::LOAD, VTs, Ops);
ID.AddInteger(MemVT.getRawBits());
ID.AddInteger(getSyntheticNodeSubclassData<LoadSDNode>(
dl.getIROrder(), VTs, AM, ExtType, MemVT, MMO));
ID.AddInteger(MMO->getPointerInfo().getAddrSpace());
ID.AddInteger(MMO->getFlags());
void *IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, dl, IP)) {
cast<LoadSDNode>(E)->refineAlignment(MMO);
return SDValue(E, 0);
}
auto *N = newSDNode<LoadSDNode>(dl.getIROrder(), dl.getDebugLoc(), VTs, AM,
ExtType, MemVT, MMO);
createOperands(N, Ops);
CSEMap.InsertNode(N, IP);
InsertNode(N);
SDValue V(N, 0);
NewSDValueDbgMsg(V, "Creating new node: ", this);
return V;
}
SDValue SelectionDAG::getLoad(EVT VT, const SDLoc &dl, SDValue Chain,
SDValue Ptr, MachinePointerInfo PtrInfo,
MaybeAlign Alignment,
MachineMemOperand::Flags MMOFlags,
const AAMDNodes &AAInfo, const MDNode *Ranges) {
SDValue Undef = getUNDEF(Ptr.getValueType());
return getLoad(ISD::UNINDEXED, ISD::NON_EXTLOAD, VT, dl, Chain, Ptr, Undef,
PtrInfo, VT, Alignment, MMOFlags, AAInfo, Ranges);
}
SDValue SelectionDAG::getLoad(EVT VT, const SDLoc &dl, SDValue Chain,
SDValue Ptr, MachineMemOperand *MMO) {
SDValue Undef = getUNDEF(Ptr.getValueType());
return getLoad(ISD::UNINDEXED, ISD::NON_EXTLOAD, VT, dl, Chain, Ptr, Undef,
VT, MMO);
}
SDValue SelectionDAG::getExtLoad(ISD::LoadExtType ExtType, const SDLoc &dl,
EVT VT, SDValue Chain, SDValue Ptr,
MachinePointerInfo PtrInfo, EVT MemVT,
MaybeAlign Alignment,
MachineMemOperand::Flags MMOFlags,
const AAMDNodes &AAInfo) {
SDValue Undef = getUNDEF(Ptr.getValueType());
return getLoad(ISD::UNINDEXED, ExtType, VT, dl, Chain, Ptr, Undef, PtrInfo,
MemVT, Alignment, MMOFlags, AAInfo);
}
SDValue SelectionDAG::getExtLoad(ISD::LoadExtType ExtType, const SDLoc &dl,
EVT VT, SDValue Chain, SDValue Ptr, EVT MemVT,
MachineMemOperand *MMO) {
SDValue Undef = getUNDEF(Ptr.getValueType());
return getLoad(ISD::UNINDEXED, ExtType, VT, dl, Chain, Ptr, Undef,
MemVT, MMO);
}
SDValue SelectionDAG::getIndexedLoad(SDValue OrigLoad, const SDLoc &dl,
SDValue Base, SDValue Offset,
ISD::MemIndexedMode AM) {
LoadSDNode *LD = cast<LoadSDNode>(OrigLoad);
assert(LD->getOffset().isUndef() && "Load is already a indexed load!");
// Don't propagate the invariant or dereferenceable flags.
auto MMOFlags =
LD->getMemOperand()->getFlags() &
~(MachineMemOperand::MOInvariant | MachineMemOperand::MODereferenceable);
return getLoad(AM, LD->getExtensionType(), OrigLoad.getValueType(), dl,
LD->getChain(), Base, Offset, LD->getPointerInfo(),
LD->getMemoryVT(), LD->getAlign(), MMOFlags, LD->getAAInfo());
}
SDValue SelectionDAG::getStore(SDValue Chain, const SDLoc &dl, SDValue Val,
SDValue Ptr, MachinePointerInfo PtrInfo,
Align Alignment,
MachineMemOperand::Flags MMOFlags,
const AAMDNodes &AAInfo) {
assert(Chain.getValueType() == MVT::Other && "Invalid chain type");
MMOFlags |= MachineMemOperand::MOStore;
assert((MMOFlags & MachineMemOperand::MOLoad) == 0);
if (PtrInfo.V.isNull())
PtrInfo = InferPointerInfo(PtrInfo, *this, Ptr);
MachineFunction &MF = getMachineFunction();
uint64_t Size =
MemoryLocation::getSizeOrUnknown(Val.getValueType().getStoreSize());
MachineMemOperand *MMO =
MF.getMachineMemOperand(PtrInfo, MMOFlags, Size, Alignment, AAInfo);
return getStore(Chain, dl, Val, Ptr, MMO);
}
SDValue SelectionDAG::getStore(SDValue Chain, const SDLoc &dl, SDValue Val,
SDValue Ptr, MachineMemOperand *MMO) {
assert(Chain.getValueType() == MVT::Other &&
"Invalid chain type");
EVT VT = Val.getValueType();
SDVTList VTs = getVTList(MVT::Other);
SDValue Undef = getUNDEF(Ptr.getValueType());
SDValue Ops[] = { Chain, Val, Ptr, Undef };
FoldingSetNodeID ID;
AddNodeIDNode(ID, ISD::STORE, VTs, Ops);
ID.AddInteger(VT.getRawBits());
ID.AddInteger(getSyntheticNodeSubclassData<StoreSDNode>(
dl.getIROrder(), VTs, ISD::UNINDEXED, false, VT, MMO));
ID.AddInteger(MMO->getPointerInfo().getAddrSpace());
ID.AddInteger(MMO->getFlags());
void *IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, dl, IP)) {
cast<StoreSDNode>(E)->refineAlignment(MMO);
return SDValue(E, 0);
}
auto *N = newSDNode<StoreSDNode>(dl.getIROrder(), dl.getDebugLoc(), VTs,
ISD::UNINDEXED, false, VT, MMO);
createOperands(N, Ops);
CSEMap.InsertNode(N, IP);
InsertNode(N);
SDValue V(N, 0);
NewSDValueDbgMsg(V, "Creating new node: ", this);
return V;
}
SDValue SelectionDAG::getTruncStore(SDValue Chain, const SDLoc &dl, SDValue Val,
SDValue Ptr, MachinePointerInfo PtrInfo,
EVT SVT, Align Alignment,
MachineMemOperand::Flags MMOFlags,
const AAMDNodes &AAInfo) {
assert(Chain.getValueType() == MVT::Other &&
"Invalid chain type");
MMOFlags |= MachineMemOperand::MOStore;
assert((MMOFlags & MachineMemOperand::MOLoad) == 0);
if (PtrInfo.V.isNull())
PtrInfo = InferPointerInfo(PtrInfo, *this, Ptr);
MachineFunction &MF = getMachineFunction();
MachineMemOperand *MMO = MF.getMachineMemOperand(
PtrInfo, MMOFlags, MemoryLocation::getSizeOrUnknown(SVT.getStoreSize()),
Alignment, AAInfo);
return getTruncStore(Chain, dl, Val, Ptr, SVT, MMO);
}
SDValue SelectionDAG::getTruncStore(SDValue Chain, const SDLoc &dl, SDValue Val,
SDValue Ptr, EVT SVT,
MachineMemOperand *MMO) {
EVT VT = Val.getValueType();
assert(Chain.getValueType() == MVT::Other &&
"Invalid chain type");
if (VT == SVT)
return getStore(Chain, dl, Val, Ptr, MMO);
assert(SVT.getScalarType().bitsLT(VT.getScalarType()) &&
"Should only be a truncating store, not extending!");
assert(VT.isInteger() == SVT.isInteger() &&
"Can't do FP-INT conversion!");
assert(VT.isVector() == SVT.isVector() &&
"Cannot use trunc store to convert to or from a vector!");
assert((!VT.isVector() ||
VT.getVectorElementCount() == SVT.getVectorElementCount()) &&
"Cannot use trunc store to change the number of vector elements!");
SDVTList VTs = getVTList(MVT::Other);
SDValue Undef = getUNDEF(Ptr.getValueType());
SDValue Ops[] = { Chain, Val, Ptr, Undef };
FoldingSetNodeID ID;
AddNodeIDNode(ID, ISD::STORE, VTs, Ops);
ID.AddInteger(SVT.getRawBits());
ID.AddInteger(getSyntheticNodeSubclassData<StoreSDNode>(
dl.getIROrder(), VTs, ISD::UNINDEXED, true, SVT, MMO));
ID.AddInteger(MMO->getPointerInfo().getAddrSpace());
ID.AddInteger(MMO->getFlags());
void *IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, dl, IP)) {
cast<StoreSDNode>(E)->refineAlignment(MMO);
return SDValue(E, 0);
}
auto *N = newSDNode<StoreSDNode>(dl.getIROrder(), dl.getDebugLoc(), VTs,
ISD::UNINDEXED, true, SVT, MMO);
createOperands(N, Ops);
CSEMap.InsertNode(N, IP);
InsertNode(N);
SDValue V(N, 0);
NewSDValueDbgMsg(V, "Creating new node: ", this);
return V;
}
SDValue SelectionDAG::getIndexedStore(SDValue OrigStore, const SDLoc &dl,
SDValue Base, SDValue Offset,
ISD::MemIndexedMode AM) {
StoreSDNode *ST = cast<StoreSDNode>(OrigStore);
assert(ST->getOffset().isUndef() && "Store is already a indexed store!");
SDVTList VTs = getVTList(Base.getValueType(), MVT::Other);
SDValue Ops[] = { ST->getChain(), ST->getValue(), Base, Offset };
FoldingSetNodeID ID;
AddNodeIDNode(ID, ISD::STORE, VTs, Ops);
ID.AddInteger(ST->getMemoryVT().getRawBits());
ID.AddInteger(ST->getRawSubclassData());
ID.AddInteger(ST->getPointerInfo().getAddrSpace());
ID.AddInteger(ST->getMemOperand()->getFlags());
void *IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, dl, IP))
return SDValue(E, 0);
auto *N = newSDNode<StoreSDNode>(dl.getIROrder(), dl.getDebugLoc(), VTs, AM,
ST->isTruncatingStore(), ST->getMemoryVT(),
ST->getMemOperand());
createOperands(N, Ops);
CSEMap.InsertNode(N, IP);
InsertNode(N);
SDValue V(N, 0);
NewSDValueDbgMsg(V, "Creating new node: ", this);
return V;
}
SDValue SelectionDAG::getLoadVP(
ISD::MemIndexedMode AM, ISD::LoadExtType ExtType, EVT VT, const SDLoc &dl,
SDValue Chain, SDValue Ptr, SDValue Offset, SDValue Mask, SDValue EVL,
MachinePointerInfo PtrInfo, EVT MemVT, Align Alignment,
MachineMemOperand::Flags MMOFlags, const AAMDNodes &AAInfo,
const MDNode *Ranges, bool IsExpanding) {
assert(Chain.getValueType() == MVT::Other && "Invalid chain type");
MMOFlags |= MachineMemOperand::MOLoad;
assert((MMOFlags & MachineMemOperand::MOStore) == 0);
// If we don't have a PtrInfo, infer the trivial frame index case to simplify
// clients.
if (PtrInfo.V.isNull())
PtrInfo = InferPointerInfo(PtrInfo, *this, Ptr, Offset);
uint64_t Size = MemoryLocation::getSizeOrUnknown(MemVT.getStoreSize());
MachineFunction &MF = getMachineFunction();
MachineMemOperand *MMO = MF.getMachineMemOperand(PtrInfo, MMOFlags, Size,
Alignment, AAInfo, Ranges);
return getLoadVP(AM, ExtType, VT, dl, Chain, Ptr, Offset, Mask, EVL, MemVT,
MMO, IsExpanding);
}
SDValue SelectionDAG::getLoadVP(ISD::MemIndexedMode AM,
ISD::LoadExtType ExtType, EVT VT,
const SDLoc &dl, SDValue Chain, SDValue Ptr,
SDValue Offset, SDValue Mask, SDValue EVL,
EVT MemVT, MachineMemOperand *MMO,
bool IsExpanding) {
bool Indexed = AM != ISD::UNINDEXED;
assert((Indexed || Offset.isUndef()) && "Unindexed load with an offset!");
SDVTList VTs = Indexed ? getVTList(VT, Ptr.getValueType(), MVT::Other)
: getVTList(VT, MVT::Other);
SDValue Ops[] = {Chain, Ptr, Offset, Mask, EVL};
FoldingSetNodeID ID;
AddNodeIDNode(ID, ISD::VP_LOAD, VTs, Ops);
ID.AddInteger(MemVT.getRawBits());
ID.AddInteger(getSyntheticNodeSubclassData<VPLoadSDNode>(
dl.getIROrder(), VTs, AM, ExtType, IsExpanding, MemVT, MMO));
ID.AddInteger(MMO->getPointerInfo().getAddrSpace());
ID.AddInteger(MMO->getFlags());
void *IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, dl, IP)) {
cast<VPLoadSDNode>(E)->refineAlignment(MMO);
return SDValue(E, 0);
}
auto *N = newSDNode<VPLoadSDNode>(dl.getIROrder(), dl.getDebugLoc(), VTs, AM,
ExtType, IsExpanding, MemVT, MMO);
createOperands(N, Ops);
CSEMap.InsertNode(N, IP);
InsertNode(N);
SDValue V(N, 0);
NewSDValueDbgMsg(V, "Creating new node: ", this);
return V;
}
SDValue SelectionDAG::getLoadVP(EVT VT, const SDLoc &dl, SDValue Chain,
SDValue Ptr, SDValue Mask, SDValue EVL,
MachinePointerInfo PtrInfo,
MaybeAlign Alignment,
MachineMemOperand::Flags MMOFlags,
const AAMDNodes &AAInfo, const MDNode *Ranges,
bool IsExpanding) {
SDValue Undef = getUNDEF(Ptr.getValueType());
return getLoadVP(ISD::UNINDEXED, ISD::NON_EXTLOAD, VT, dl, Chain, Ptr, Undef,
Mask, EVL, PtrInfo, VT, Alignment, MMOFlags, AAInfo, Ranges,
IsExpanding);
}
SDValue SelectionDAG::getLoadVP(EVT VT, const SDLoc &dl, SDValue Chain,
SDValue Ptr, SDValue Mask, SDValue EVL,
MachineMemOperand *MMO, bool IsExpanding) {
SDValue Undef = getUNDEF(Ptr.getValueType());
return getLoadVP(ISD::UNINDEXED, ISD::NON_EXTLOAD, VT, dl, Chain, Ptr, Undef,
Mask, EVL, VT, MMO, IsExpanding);
}
SDValue SelectionDAG::getExtLoadVP(ISD::LoadExtType ExtType, const SDLoc &dl,
EVT VT, SDValue Chain, SDValue Ptr,
SDValue Mask, SDValue EVL,
MachinePointerInfo PtrInfo, EVT MemVT,
MaybeAlign Alignment,
MachineMemOperand::Flags MMOFlags,
const AAMDNodes &AAInfo, bool IsExpanding) {
SDValue Undef = getUNDEF(Ptr.getValueType());
return getLoadVP(ISD::UNINDEXED, ExtType, VT, dl, Chain, Ptr, Undef, Mask,
EVL, PtrInfo, MemVT, Alignment, MMOFlags, AAInfo, nullptr,
IsExpanding);
}
SDValue SelectionDAG::getExtLoadVP(ISD::LoadExtType ExtType, const SDLoc &dl,
EVT VT, SDValue Chain, SDValue Ptr,
SDValue Mask, SDValue EVL, EVT MemVT,
MachineMemOperand *MMO, bool IsExpanding) {
SDValue Undef = getUNDEF(Ptr.getValueType());
return getLoadVP(ISD::UNINDEXED, ExtType, VT, dl, Chain, Ptr, Undef, Mask,
EVL, MemVT, MMO, IsExpanding);
}
SDValue SelectionDAG::getIndexedLoadVP(SDValue OrigLoad, const SDLoc &dl,
SDValue Base, SDValue Offset,
ISD::MemIndexedMode AM) {
auto *LD = cast<VPLoadSDNode>(OrigLoad);
assert(LD->getOffset().isUndef() && "Load is already a indexed load!");
// Don't propagate the invariant or dereferenceable flags.
auto MMOFlags =
LD->getMemOperand()->getFlags() &
~(MachineMemOperand::MOInvariant | MachineMemOperand::MODereferenceable);
return getLoadVP(AM, LD->getExtensionType(), OrigLoad.getValueType(), dl,
LD->getChain(), Base, Offset, LD->getMask(),
LD->getVectorLength(), LD->getPointerInfo(),
LD->getMemoryVT(), LD->getAlign(), MMOFlags, LD->getAAInfo(),
nullptr, LD->isExpandingLoad());
}
SDValue SelectionDAG::getStoreVP(SDValue Chain, const SDLoc &dl, SDValue Val,
SDValue Ptr, SDValue Offset, SDValue Mask,
SDValue EVL, EVT MemVT, MachineMemOperand *MMO,
ISD::MemIndexedMode AM, bool IsTruncating,
bool IsCompressing) {
assert(Chain.getValueType() == MVT::Other && "Invalid chain type");
bool Indexed = AM != ISD::UNINDEXED;
assert((Indexed || Offset.isUndef()) && "Unindexed vp_store with an offset!");
SDVTList VTs = Indexed ? getVTList(Ptr.getValueType(), MVT::Other)
: getVTList(MVT::Other);
SDValue Ops[] = {Chain, Val, Ptr, Offset, Mask, EVL};
FoldingSetNodeID ID;
AddNodeIDNode(ID, ISD::VP_STORE, VTs, Ops);
ID.AddInteger(MemVT.getRawBits());
ID.AddInteger(getSyntheticNodeSubclassData<VPStoreSDNode>(
dl.getIROrder(), VTs, AM, IsTruncating, IsCompressing, MemVT, MMO));
ID.AddInteger(MMO->getPointerInfo().getAddrSpace());
ID.AddInteger(MMO->getFlags());
void *IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, dl, IP)) {
cast<VPStoreSDNode>(E)->refineAlignment(MMO);
return SDValue(E, 0);
}
auto *N = newSDNode<VPStoreSDNode>(dl.getIROrder(), dl.getDebugLoc(), VTs, AM,
IsTruncating, IsCompressing, MemVT, MMO);
createOperands(N, Ops);
CSEMap.InsertNode(N, IP);
InsertNode(N);
SDValue V(N, 0);
NewSDValueDbgMsg(V, "Creating new node: ", this);
return V;
}
SDValue SelectionDAG::getTruncStoreVP(SDValue Chain, const SDLoc &dl,
SDValue Val, SDValue Ptr, SDValue Mask,
SDValue EVL, MachinePointerInfo PtrInfo,
EVT SVT, Align Alignment,
MachineMemOperand::Flags MMOFlags,
const AAMDNodes &AAInfo,
bool IsCompressing) {
assert(Chain.getValueType() == MVT::Other && "Invalid chain type");
MMOFlags |= MachineMemOperand::MOStore;
assert((MMOFlags & MachineMemOperand::MOLoad) == 0);
if (PtrInfo.V.isNull())
PtrInfo = InferPointerInfo(PtrInfo, *this, Ptr);
MachineFunction &MF = getMachineFunction();
MachineMemOperand *MMO = MF.getMachineMemOperand(
PtrInfo, MMOFlags, MemoryLocation::getSizeOrUnknown(SVT.getStoreSize()),
Alignment, AAInfo);
return getTruncStoreVP(Chain, dl, Val, Ptr, Mask, EVL, SVT, MMO,
IsCompressing);
}
SDValue SelectionDAG::getTruncStoreVP(SDValue Chain, const SDLoc &dl,
SDValue Val, SDValue Ptr, SDValue Mask,
SDValue EVL, EVT SVT,
MachineMemOperand *MMO,
bool IsCompressing) {
EVT VT = Val.getValueType();
assert(Chain.getValueType() == MVT::Other && "Invalid chain type");
if (VT == SVT)
return getStoreVP(Chain, dl, Val, Ptr, getUNDEF(Ptr.getValueType()), Mask,
EVL, VT, MMO, ISD::UNINDEXED,
/*IsTruncating*/ false, IsCompressing);
assert(SVT.getScalarType().bitsLT(VT.getScalarType()) &&
"Should only be a truncating store, not extending!");
assert(VT.isInteger() == SVT.isInteger() && "Can't do FP-INT conversion!");
assert(VT.isVector() == SVT.isVector() &&
"Cannot use trunc store to convert to or from a vector!");
assert((!VT.isVector() ||
VT.getVectorElementCount() == SVT.getVectorElementCount()) &&
"Cannot use trunc store to change the number of vector elements!");
SDVTList VTs = getVTList(MVT::Other);
SDValue Undef = getUNDEF(Ptr.getValueType());
SDValue Ops[] = {Chain, Val, Ptr, Undef, Mask, EVL};
FoldingSetNodeID ID;
AddNodeIDNode(ID, ISD::VP_STORE, VTs, Ops);
ID.AddInteger(SVT.getRawBits());
ID.AddInteger(getSyntheticNodeSubclassData<VPStoreSDNode>(
dl.getIROrder(), VTs, ISD::UNINDEXED, true, IsCompressing, SVT, MMO));
ID.AddInteger(MMO->getPointerInfo().getAddrSpace());
ID.AddInteger(MMO->getFlags());
void *IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, dl, IP)) {
cast<VPStoreSDNode>(E)->refineAlignment(MMO);
return SDValue(E, 0);
}
auto *N =
newSDNode<VPStoreSDNode>(dl.getIROrder(), dl.getDebugLoc(), VTs,
ISD::UNINDEXED, true, IsCompressing, SVT, MMO);
createOperands(N, Ops);
CSEMap.InsertNode(N, IP);
InsertNode(N);
SDValue V(N, 0);
NewSDValueDbgMsg(V, "Creating new node: ", this);
return V;
}
SDValue SelectionDAG::getIndexedStoreVP(SDValue OrigStore, const SDLoc &dl,
SDValue Base, SDValue Offset,
ISD::MemIndexedMode AM) {
auto *ST = cast<VPStoreSDNode>(OrigStore);
assert(ST->getOffset().isUndef() && "Store is already an indexed store!");
SDVTList VTs = getVTList(Base.getValueType(), MVT::Other);
SDValue Ops[] = {ST->getChain(), ST->getValue(), Base,
Offset, ST->getMask(), ST->getVectorLength()};
FoldingSetNodeID ID;
AddNodeIDNode(ID, ISD::VP_STORE, VTs, Ops);
ID.AddInteger(ST->getMemoryVT().getRawBits());
ID.AddInteger(ST->getRawSubclassData());
ID.AddInteger(ST->getPointerInfo().getAddrSpace());
ID.AddInteger(ST->getMemOperand()->getFlags());
void *IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, dl, IP))
return SDValue(E, 0);
auto *N = newSDNode<VPStoreSDNode>(
dl.getIROrder(), dl.getDebugLoc(), VTs, AM, ST->isTruncatingStore(),
ST->isCompressingStore(), ST->getMemoryVT(), ST->getMemOperand());
createOperands(N, Ops);
CSEMap.InsertNode(N, IP);
InsertNode(N);
SDValue V(N, 0);
NewSDValueDbgMsg(V, "Creating new node: ", this);
return V;
}
SDValue SelectionDAG::getStridedLoadVP(
ISD::MemIndexedMode AM, ISD::LoadExtType ExtType, EVT VT, const SDLoc &DL,
SDValue Chain, SDValue Ptr, SDValue Offset, SDValue Stride, SDValue Mask,
SDValue EVL, MachinePointerInfo PtrInfo, EVT MemVT, Align Alignment,
MachineMemOperand::Flags MMOFlags, const AAMDNodes &AAInfo,
const MDNode *Ranges, bool IsExpanding) {
assert(Chain.getValueType() == MVT::Other && "Invalid chain type");
MMOFlags |= MachineMemOperand::MOLoad;
assert((MMOFlags & MachineMemOperand::MOStore) == 0);
// If we don't have a PtrInfo, infer the trivial frame index case to simplify
// clients.
if (PtrInfo.V.isNull())
PtrInfo = InferPointerInfo(PtrInfo, *this, Ptr, Offset);
uint64_t Size = MemoryLocation::UnknownSize;
MachineFunction &MF = getMachineFunction();
MachineMemOperand *MMO = MF.getMachineMemOperand(PtrInfo, MMOFlags, Size,
Alignment, AAInfo, Ranges);
return getStridedLoadVP(AM, ExtType, VT, DL, Chain, Ptr, Offset, Stride, Mask,
EVL, MemVT, MMO, IsExpanding);
}
SDValue SelectionDAG::getStridedLoadVP(
ISD::MemIndexedMode AM, ISD::LoadExtType ExtType, EVT VT, const SDLoc &DL,
SDValue Chain, SDValue Ptr, SDValue Offset, SDValue Stride, SDValue Mask,
SDValue EVL, EVT MemVT, MachineMemOperand *MMO, bool IsExpanding) {
bool Indexed = AM != ISD::UNINDEXED;
assert((Indexed || Offset.isUndef()) && "Unindexed load with an offset!");
SDValue Ops[] = {Chain, Ptr, Offset, Stride, Mask, EVL};
SDVTList VTs = Indexed ? getVTList(VT, Ptr.getValueType(), MVT::Other)
: getVTList(VT, MVT::Other);
FoldingSetNodeID ID;
AddNodeIDNode(ID, ISD::EXPERIMENTAL_VP_STRIDED_LOAD, VTs, Ops);
ID.AddInteger(VT.getRawBits());
ID.AddInteger(getSyntheticNodeSubclassData<VPStridedLoadSDNode>(
DL.getIROrder(), VTs, AM, ExtType, IsExpanding, MemVT, MMO));
ID.AddInteger(MMO->getPointerInfo().getAddrSpace());
void *IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, DL, IP)) {
cast<VPStridedLoadSDNode>(E)->refineAlignment(MMO);
return SDValue(E, 0);
}
auto *N =
newSDNode<VPStridedLoadSDNode>(DL.getIROrder(), DL.getDebugLoc(), VTs, AM,
ExtType, IsExpanding, MemVT, MMO);
createOperands(N, Ops);
CSEMap.InsertNode(N, IP);
InsertNode(N);
SDValue V(N, 0);
NewSDValueDbgMsg(V, "Creating new node: ", this);
return V;
}
SDValue SelectionDAG::getStridedLoadVP(
EVT VT, const SDLoc &DL, SDValue Chain, SDValue Ptr, SDValue Stride,
SDValue Mask, SDValue EVL, MachinePointerInfo PtrInfo, MaybeAlign Alignment,
MachineMemOperand::Flags MMOFlags, const AAMDNodes &AAInfo,
const MDNode *Ranges, bool IsExpanding) {
SDValue Undef = getUNDEF(Ptr.getValueType());
return getStridedLoadVP(ISD::UNINDEXED, ISD::NON_EXTLOAD, VT, DL, Chain, Ptr,
Undef, Stride, Mask, EVL, PtrInfo, VT, Alignment,
MMOFlags, AAInfo, Ranges, IsExpanding);
}
SDValue SelectionDAG::getStridedLoadVP(EVT VT, const SDLoc &DL, SDValue Chain,
SDValue Ptr, SDValue Stride,
SDValue Mask, SDValue EVL,
MachineMemOperand *MMO,
bool IsExpanding) {
SDValue Undef = getUNDEF(Ptr.getValueType());
return getStridedLoadVP(ISD::UNINDEXED, ISD::NON_EXTLOAD, VT, DL, Chain, Ptr,
Undef, Stride, Mask, EVL, VT, MMO, IsExpanding);
}
SDValue SelectionDAG::getExtStridedLoadVP(
ISD::LoadExtType ExtType, const SDLoc &DL, EVT VT, SDValue Chain,
SDValue Ptr, SDValue Stride, SDValue Mask, SDValue EVL,
MachinePointerInfo PtrInfo, EVT MemVT, MaybeAlign Alignment,
MachineMemOperand::Flags MMOFlags, const AAMDNodes &AAInfo,
bool IsExpanding) {
SDValue Undef = getUNDEF(Ptr.getValueType());
return getStridedLoadVP(ISD::UNINDEXED, ExtType, VT, DL, Chain, Ptr, Undef,
Stride, Mask, EVL, PtrInfo, MemVT, Alignment,
MMOFlags, AAInfo, nullptr, IsExpanding);
}
SDValue SelectionDAG::getExtStridedLoadVP(
ISD::LoadExtType ExtType, const SDLoc &DL, EVT VT, SDValue Chain,
SDValue Ptr, SDValue Stride, SDValue Mask, SDValue EVL, EVT MemVT,
MachineMemOperand *MMO, bool IsExpanding) {
SDValue Undef = getUNDEF(Ptr.getValueType());
return getStridedLoadVP(ISD::UNINDEXED, ExtType, VT, DL, Chain, Ptr, Undef,
Stride, Mask, EVL, MemVT, MMO, IsExpanding);
}
SDValue SelectionDAG::getIndexedStridedLoadVP(SDValue OrigLoad, const SDLoc &DL,
SDValue Base, SDValue Offset,
ISD::MemIndexedMode AM) {
auto *SLD = cast<VPStridedLoadSDNode>(OrigLoad);
assert(SLD->getOffset().isUndef() &&
"Strided load is already a indexed load!");
// Don't propagate the invariant or dereferenceable flags.
auto MMOFlags =
SLD->getMemOperand()->getFlags() &
~(MachineMemOperand::MOInvariant | MachineMemOperand::MODereferenceable);
return getStridedLoadVP(
AM, SLD->getExtensionType(), OrigLoad.getValueType(), DL, SLD->getChain(),
Base, Offset, SLD->getStride(), SLD->getMask(), SLD->getVectorLength(),
SLD->getPointerInfo(), SLD->getMemoryVT(), SLD->getAlign(), MMOFlags,
SLD->getAAInfo(), nullptr, SLD->isExpandingLoad());
}
SDValue SelectionDAG::getStridedStoreVP(SDValue Chain, const SDLoc &DL,
SDValue Val, SDValue Ptr,
SDValue Offset, SDValue Stride,
SDValue Mask, SDValue EVL, EVT MemVT,
MachineMemOperand *MMO,
ISD::MemIndexedMode AM,
bool IsTruncating, bool IsCompressing) {
assert(Chain.getValueType() == MVT::Other && "Invalid chain type");
bool Indexed = AM != ISD::UNINDEXED;
assert((Indexed || Offset.isUndef()) && "Unindexed vp_store with an offset!");
SDVTList VTs = Indexed ? getVTList(Ptr.getValueType(), MVT::Other)
: getVTList(MVT::Other);
SDValue Ops[] = {Chain, Val, Ptr, Offset, Stride, Mask, EVL};
FoldingSetNodeID ID;
AddNodeIDNode(ID, ISD::EXPERIMENTAL_VP_STRIDED_STORE, VTs, Ops);
ID.AddInteger(MemVT.getRawBits());
ID.AddInteger(getSyntheticNodeSubclassData<VPStridedStoreSDNode>(
DL.getIROrder(), VTs, AM, IsTruncating, IsCompressing, MemVT, MMO));
ID.AddInteger(MMO->getPointerInfo().getAddrSpace());
void *IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, DL, IP)) {
cast<VPStridedStoreSDNode>(E)->refineAlignment(MMO);
return SDValue(E, 0);
}
auto *N = newSDNode<VPStridedStoreSDNode>(DL.getIROrder(), DL.getDebugLoc(),
VTs, AM, IsTruncating,
IsCompressing, MemVT, MMO);
createOperands(N, Ops);
CSEMap.InsertNode(N, IP);
InsertNode(N);
SDValue V(N, 0);
NewSDValueDbgMsg(V, "Creating new node: ", this);
return V;
}
SDValue SelectionDAG::getTruncStridedStoreVP(
SDValue Chain, const SDLoc &DL, SDValue Val, SDValue Ptr, SDValue Stride,
SDValue Mask, SDValue EVL, MachinePointerInfo PtrInfo, EVT SVT,
Align Alignment, MachineMemOperand::Flags MMOFlags, const AAMDNodes &AAInfo,
bool IsCompressing) {
assert(Chain.getValueType() == MVT::Other && "Invalid chain type");
MMOFlags |= MachineMemOperand::MOStore;
assert((MMOFlags & MachineMemOperand::MOLoad) == 0);
if (PtrInfo.V.isNull())
PtrInfo = InferPointerInfo(PtrInfo, *this, Ptr);
MachineFunction &MF = getMachineFunction();
MachineMemOperand *MMO = MF.getMachineMemOperand(
PtrInfo, MMOFlags, MemoryLocation::UnknownSize, Alignment, AAInfo);
return getTruncStridedStoreVP(Chain, DL, Val, Ptr, Stride, Mask, EVL, SVT,
MMO, IsCompressing);
}
SDValue SelectionDAG::getTruncStridedStoreVP(SDValue Chain, const SDLoc &DL,
SDValue Val, SDValue Ptr,
SDValue Stride, SDValue Mask,
SDValue EVL, EVT SVT,
MachineMemOperand *MMO,
bool IsCompressing) {
EVT VT = Val.getValueType();
assert(Chain.getValueType() == MVT::Other && "Invalid chain type");
if (VT == SVT)
return getStridedStoreVP(Chain, DL, Val, Ptr, getUNDEF(Ptr.getValueType()),
Stride, Mask, EVL, VT, MMO, ISD::UNINDEXED,
/*IsTruncating*/ false, IsCompressing);
assert(SVT.getScalarType().bitsLT(VT.getScalarType()) &&
"Should only be a truncating store, not extending!");
assert(VT.isInteger() == SVT.isInteger() && "Can't do FP-INT conversion!");
assert(VT.isVector() == SVT.isVector() &&
"Cannot use trunc store to convert to or from a vector!");
assert((!VT.isVector() ||
VT.getVectorElementCount() == SVT.getVectorElementCount()) &&
"Cannot use trunc store to change the number of vector elements!");
SDVTList VTs = getVTList(MVT::Other);
SDValue Undef = getUNDEF(Ptr.getValueType());
SDValue Ops[] = {Chain, Val, Ptr, Undef, Stride, Mask, EVL};
FoldingSetNodeID ID;
AddNodeIDNode(ID, ISD::EXPERIMENTAL_VP_STRIDED_STORE, VTs, Ops);
ID.AddInteger(SVT.getRawBits());
ID.AddInteger(getSyntheticNodeSubclassData<VPStridedStoreSDNode>(
DL.getIROrder(), VTs, ISD::UNINDEXED, true, IsCompressing, SVT, MMO));
ID.AddInteger(MMO->getPointerInfo().getAddrSpace());
void *IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, DL, IP)) {
cast<VPStridedStoreSDNode>(E)->refineAlignment(MMO);
return SDValue(E, 0);
}
auto *N = newSDNode<VPStridedStoreSDNode>(DL.getIROrder(), DL.getDebugLoc(),
VTs, ISD::UNINDEXED, true,
IsCompressing, SVT, MMO);
createOperands(N, Ops);
CSEMap.InsertNode(N, IP);
InsertNode(N);
SDValue V(N, 0);
NewSDValueDbgMsg(V, "Creating new node: ", this);
return V;
}
SDValue SelectionDAG::getIndexedStridedStoreVP(SDValue OrigStore,
const SDLoc &DL, SDValue Base,
SDValue Offset,
ISD::MemIndexedMode AM) {
auto *SST = cast<VPStridedStoreSDNode>(OrigStore);
assert(SST->getOffset().isUndef() &&
"Strided store is already an indexed store!");
SDVTList VTs = getVTList(Base.getValueType(), MVT::Other);
SDValue Ops[] = {
SST->getChain(), SST->getValue(), Base, Offset, SST->getStride(),
SST->getMask(), SST->getVectorLength()};
FoldingSetNodeID ID;
AddNodeIDNode(ID, ISD::EXPERIMENTAL_VP_STRIDED_STORE, VTs, Ops);
ID.AddInteger(SST->getMemoryVT().getRawBits());
ID.AddInteger(SST->getRawSubclassData());
ID.AddInteger(SST->getPointerInfo().getAddrSpace());
void *IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, DL, IP))
return SDValue(E, 0);
auto *N = newSDNode<VPStridedStoreSDNode>(
DL.getIROrder(), DL.getDebugLoc(), VTs, AM, SST->isTruncatingStore(),
SST->isCompressingStore(), SST->getMemoryVT(), SST->getMemOperand());
createOperands(N, Ops);
CSEMap.InsertNode(N, IP);
InsertNode(N);
SDValue V(N, 0);
NewSDValueDbgMsg(V, "Creating new node: ", this);
return V;
}
SDValue SelectionDAG::getGatherVP(SDVTList VTs, EVT VT, const SDLoc &dl,
ArrayRef<SDValue> Ops, MachineMemOperand *MMO,
ISD::MemIndexType IndexType) {
assert(Ops.size() == 6 && "Incompatible number of operands");
FoldingSetNodeID ID;
AddNodeIDNode(ID, ISD::VP_GATHER, VTs, Ops);
ID.AddInteger(VT.getRawBits());
ID.AddInteger(getSyntheticNodeSubclassData<VPGatherSDNode>(
dl.getIROrder(), VTs, VT, MMO, IndexType));
ID.AddInteger(MMO->getPointerInfo().getAddrSpace());
ID.AddInteger(MMO->getFlags());
void *IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, dl, IP)) {
cast<VPGatherSDNode>(E)->refineAlignment(MMO);
return SDValue(E, 0);
}
auto *N = newSDNode<VPGatherSDNode>(dl.getIROrder(), dl.getDebugLoc(), VTs,
VT, MMO, IndexType);
createOperands(N, Ops);
assert(N->getMask().getValueType().getVectorElementCount() ==
N->getValueType(0).getVectorElementCount() &&
"Vector width mismatch between mask and data");
assert(N->getIndex().getValueType().getVectorElementCount().isScalable() ==
N->getValueType(0).getVectorElementCount().isScalable() &&
"Scalable flags of index and data do not match");
assert(ElementCount::isKnownGE(
N->getIndex().getValueType().getVectorElementCount(),
N->getValueType(0).getVectorElementCount()) &&
"Vector width mismatch between index and data");
assert(isa<ConstantSDNode>(N->getScale()) &&
N->getScale()->getAsAPIntVal().isPowerOf2() &&
"Scale should be a constant power of 2");
CSEMap.InsertNode(N, IP);
InsertNode(N);
SDValue V(N, 0);
NewSDValueDbgMsg(V, "Creating new node: ", this);
return V;
}
SDValue SelectionDAG::getScatterVP(SDVTList VTs, EVT VT, const SDLoc &dl,
ArrayRef<SDValue> Ops,
MachineMemOperand *MMO,
ISD::MemIndexType IndexType) {
assert(Ops.size() == 7 && "Incompatible number of operands");
FoldingSetNodeID ID;
AddNodeIDNode(ID, ISD::VP_SCATTER, VTs, Ops);
ID.AddInteger(VT.getRawBits());
ID.AddInteger(getSyntheticNodeSubclassData<VPScatterSDNode>(
dl.getIROrder(), VTs, VT, MMO, IndexType));
ID.AddInteger(MMO->getPointerInfo().getAddrSpace());
ID.AddInteger(MMO->getFlags());
void *IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, dl, IP)) {
cast<VPScatterSDNode>(E)->refineAlignment(MMO);
return SDValue(E, 0);
}
auto *N = newSDNode<VPScatterSDNode>(dl.getIROrder(), dl.getDebugLoc(), VTs,
VT, MMO, IndexType);
createOperands(N, Ops);
assert(N->getMask().getValueType().getVectorElementCount() ==
N->getValue().getValueType().getVectorElementCount() &&
"Vector width mismatch between mask and data");
assert(
N->getIndex().getValueType().getVectorElementCount().isScalable() ==
N->getValue().getValueType().getVectorElementCount().isScalable() &&
"Scalable flags of index and data do not match");
assert(ElementCount::isKnownGE(
N->getIndex().getValueType().getVectorElementCount(),
N->getValue().getValueType().getVectorElementCount()) &&
"Vector width mismatch between index and data");
assert(isa<ConstantSDNode>(N->getScale()) &&
N->getScale()->getAsAPIntVal().isPowerOf2() &&
"Scale should be a constant power of 2");
CSEMap.InsertNode(N, IP);
InsertNode(N);
SDValue V(N, 0);
NewSDValueDbgMsg(V, "Creating new node: ", this);
return V;
}
SDValue SelectionDAG::getMaskedLoad(EVT VT, const SDLoc &dl, SDValue Chain,
SDValue Base, SDValue Offset, SDValue Mask,
SDValue PassThru, EVT MemVT,
MachineMemOperand *MMO,
ISD::MemIndexedMode AM,
ISD::LoadExtType ExtTy, bool isExpanding) {
bool Indexed = AM != ISD::UNINDEXED;
assert((Indexed || Offset.isUndef()) &&
"Unindexed masked load with an offset!");
SDVTList VTs = Indexed ? getVTList(VT, Base.getValueType(), MVT::Other)
: getVTList(VT, MVT::Other);
SDValue Ops[] = {Chain, Base, Offset, Mask, PassThru};
FoldingSetNodeID ID;
AddNodeIDNode(ID, ISD::MLOAD, VTs, Ops);
ID.AddInteger(MemVT.getRawBits());
ID.AddInteger(getSyntheticNodeSubclassData<MaskedLoadSDNode>(
dl.getIROrder(), VTs, AM, ExtTy, isExpanding, MemVT, MMO));
ID.AddInteger(MMO->getPointerInfo().getAddrSpace());
ID.AddInteger(MMO->getFlags());
void *IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, dl, IP)) {
cast<MaskedLoadSDNode>(E)->refineAlignment(MMO);
return SDValue(E, 0);
}
auto *N = newSDNode<MaskedLoadSDNode>(dl.getIROrder(), dl.getDebugLoc(), VTs,
AM, ExtTy, isExpanding, MemVT, MMO);
createOperands(N, Ops);
CSEMap.InsertNode(N, IP);
InsertNode(N);
SDValue V(N, 0);
NewSDValueDbgMsg(V, "Creating new node: ", this);
return V;
}
SDValue SelectionDAG::getIndexedMaskedLoad(SDValue OrigLoad, const SDLoc &dl,
SDValue Base, SDValue Offset,
ISD::MemIndexedMode AM) {
MaskedLoadSDNode *LD = cast<MaskedLoadSDNode>(OrigLoad);
assert(LD->getOffset().isUndef() && "Masked load is already a indexed load!");
return getMaskedLoad(OrigLoad.getValueType(), dl, LD->getChain(), Base,
Offset, LD->getMask(), LD->getPassThru(),
LD->getMemoryVT(), LD->getMemOperand(), AM,
LD->getExtensionType(), LD->isExpandingLoad());
}
SDValue SelectionDAG::getMaskedStore(SDValue Chain, const SDLoc &dl,
SDValue Val, SDValue Base, SDValue Offset,
SDValue Mask, EVT MemVT,
MachineMemOperand *MMO,
ISD::MemIndexedMode AM, bool IsTruncating,
bool IsCompressing) {
assert(Chain.getValueType() == MVT::Other &&
"Invalid chain type");
bool Indexed = AM != ISD::UNINDEXED;
assert((Indexed || Offset.isUndef()) &&
"Unindexed masked store with an offset!");
SDVTList VTs = Indexed ? getVTList(Base.getValueType(), MVT::Other)
: getVTList(MVT::Other);
SDValue Ops[] = {Chain, Val, Base, Offset, Mask};
FoldingSetNodeID ID;
AddNodeIDNode(ID, ISD::MSTORE, VTs, Ops);
ID.AddInteger(MemVT.getRawBits());
ID.AddInteger(getSyntheticNodeSubclassData<MaskedStoreSDNode>(
dl.getIROrder(), VTs, AM, IsTruncating, IsCompressing, MemVT, MMO));
ID.AddInteger(MMO->getPointerInfo().getAddrSpace());
ID.AddInteger(MMO->getFlags());
void *IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, dl, IP)) {
cast<MaskedStoreSDNode>(E)->refineAlignment(MMO);
return SDValue(E, 0);
}
auto *N =
newSDNode<MaskedStoreSDNode>(dl.getIROrder(), dl.getDebugLoc(), VTs, AM,
IsTruncating, IsCompressing, MemVT, MMO);
createOperands(N, Ops);
CSEMap.InsertNode(N, IP);
InsertNode(N);
SDValue V(N, 0);
NewSDValueDbgMsg(V, "Creating new node: ", this);
return V;
}
SDValue SelectionDAG::getIndexedMaskedStore(SDValue OrigStore, const SDLoc &dl,
SDValue Base, SDValue Offset,
ISD::MemIndexedMode AM) {
MaskedStoreSDNode *ST = cast<MaskedStoreSDNode>(OrigStore);
assert(ST->getOffset().isUndef() &&
"Masked store is already a indexed store!");
return getMaskedStore(ST->getChain(), dl, ST->getValue(), Base, Offset,
ST->getMask(), ST->getMemoryVT(), ST->getMemOperand(),
AM, ST->isTruncatingStore(), ST->isCompressingStore());
}
SDValue SelectionDAG::getMaskedGather(SDVTList VTs, EVT MemVT, const SDLoc &dl,
ArrayRef<SDValue> Ops,
MachineMemOperand *MMO,
ISD::MemIndexType IndexType,
ISD::LoadExtType ExtTy) {
assert(Ops.size() == 6 && "Incompatible number of operands");
FoldingSetNodeID ID;
AddNodeIDNode(ID, ISD::MGATHER, VTs, Ops);
ID.AddInteger(MemVT.getRawBits());
ID.AddInteger(getSyntheticNodeSubclassData<MaskedGatherSDNode>(
dl.getIROrder(), VTs, MemVT, MMO, IndexType, ExtTy));
ID.AddInteger(MMO->getPointerInfo().getAddrSpace());
ID.AddInteger(MMO->getFlags());
void *IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, dl, IP)) {
cast<MaskedGatherSDNode>(E)->refineAlignment(MMO);
return SDValue(E, 0);
}
auto *N = newSDNode<MaskedGatherSDNode>(dl.getIROrder(), dl.getDebugLoc(),
VTs, MemVT, MMO, IndexType, ExtTy);
createOperands(N, Ops);
assert(N->getPassThru().getValueType() == N->getValueType(0) &&
"Incompatible type of the PassThru value in MaskedGatherSDNode");
assert(N->getMask().getValueType().getVectorElementCount() ==
N->getValueType(0).getVectorElementCount() &&
"Vector width mismatch between mask and data");
assert(N->getIndex().getValueType().getVectorElementCount().isScalable() ==
N->getValueType(0).getVectorElementCount().isScalable() &&
"Scalable flags of index and data do not match");
assert(ElementCount::isKnownGE(
N->getIndex().getValueType().getVectorElementCount(),
N->getValueType(0).getVectorElementCount()) &&
"Vector width mismatch between index and data");
assert(isa<ConstantSDNode>(N->getScale()) &&
N->getScale()->getAsAPIntVal().isPowerOf2() &&
"Scale should be a constant power of 2");
CSEMap.InsertNode(N, IP);
InsertNode(N);
SDValue V(N, 0);
NewSDValueDbgMsg(V, "Creating new node: ", this);
return V;
}
SDValue SelectionDAG::getMaskedScatter(SDVTList VTs, EVT MemVT, const SDLoc &dl,
ArrayRef<SDValue> Ops,
MachineMemOperand *MMO,
ISD::MemIndexType IndexType,
bool IsTrunc) {
assert(Ops.size() == 6 && "Incompatible number of operands");
FoldingSetNodeID ID;
AddNodeIDNode(ID, ISD::MSCATTER, VTs, Ops);
ID.AddInteger(MemVT.getRawBits());
ID.AddInteger(getSyntheticNodeSubclassData<MaskedScatterSDNode>(
dl.getIROrder(), VTs, MemVT, MMO, IndexType, IsTrunc));
ID.AddInteger(MMO->getPointerInfo().getAddrSpace());
ID.AddInteger(MMO->getFlags());
void *IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, dl, IP)) {
cast<MaskedScatterSDNode>(E)->refineAlignment(MMO);
return SDValue(E, 0);
}
auto *N = newSDNode<MaskedScatterSDNode>(dl.getIROrder(), dl.getDebugLoc(),
VTs, MemVT, MMO, IndexType, IsTrunc);
createOperands(N, Ops);
assert(N->getMask().getValueType().getVectorElementCount() ==
N->getValue().getValueType().getVectorElementCount() &&
"Vector width mismatch between mask and data");
assert(
N->getIndex().getValueType().getVectorElementCount().isScalable() ==
N->getValue().getValueType().getVectorElementCount().isScalable() &&
"Scalable flags of index and data do not match");
assert(ElementCount::isKnownGE(
N->getIndex().getValueType().getVectorElementCount(),
N->getValue().getValueType().getVectorElementCount()) &&
"Vector width mismatch between index and data");
assert(isa<ConstantSDNode>(N->getScale()) &&
N->getScale()->getAsAPIntVal().isPowerOf2() &&
"Scale should be a constant power of 2");
CSEMap.InsertNode(N, IP);
InsertNode(N);
SDValue V(N, 0);
NewSDValueDbgMsg(V, "Creating new node: ", this);
return V;
}
SDValue SelectionDAG::getGetFPEnv(SDValue Chain, const SDLoc &dl, SDValue Ptr,
EVT MemVT, MachineMemOperand *MMO) {
assert(Chain.getValueType() == MVT::Other && "Invalid chain type");
SDVTList VTs = getVTList(MVT::Other);
SDValue Ops[] = {Chain, Ptr};
FoldingSetNodeID ID;
AddNodeIDNode(ID, ISD::GET_FPENV_MEM, VTs, Ops);
ID.AddInteger(MemVT.getRawBits());
ID.AddInteger(getSyntheticNodeSubclassData<FPStateAccessSDNode>(
ISD::GET_FPENV_MEM, dl.getIROrder(), VTs, MemVT, MMO));
ID.AddInteger(MMO->getPointerInfo().getAddrSpace());
ID.AddInteger(MMO->getFlags());
void *IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, dl, IP))
return SDValue(E, 0);
auto *N = newSDNode<FPStateAccessSDNode>(ISD::GET_FPENV_MEM, dl.getIROrder(),
dl.getDebugLoc(), VTs, MemVT, MMO);
createOperands(N, Ops);
CSEMap.InsertNode(N, IP);
InsertNode(N);
SDValue V(N, 0);
NewSDValueDbgMsg(V, "Creating new node: ", this);
return V;
}
SDValue SelectionDAG::getSetFPEnv(SDValue Chain, const SDLoc &dl, SDValue Ptr,
EVT MemVT, MachineMemOperand *MMO) {
assert(Chain.getValueType() == MVT::Other && "Invalid chain type");
SDVTList VTs = getVTList(MVT::Other);
SDValue Ops[] = {Chain, Ptr};
FoldingSetNodeID ID;
AddNodeIDNode(ID, ISD::SET_FPENV_MEM, VTs, Ops);
ID.AddInteger(MemVT.getRawBits());
ID.AddInteger(getSyntheticNodeSubclassData<FPStateAccessSDNode>(
ISD::SET_FPENV_MEM, dl.getIROrder(), VTs, MemVT, MMO));
ID.AddInteger(MMO->getPointerInfo().getAddrSpace());
ID.AddInteger(MMO->getFlags());
void *IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, dl, IP))
return SDValue(E, 0);
auto *N = newSDNode<FPStateAccessSDNode>(ISD::SET_FPENV_MEM, dl.getIROrder(),
dl.getDebugLoc(), VTs, MemVT, MMO);
createOperands(N, Ops);
CSEMap.InsertNode(N, IP);
InsertNode(N);
SDValue V(N, 0);
NewSDValueDbgMsg(V, "Creating new node: ", this);
return V;
}
SDValue SelectionDAG::simplifySelect(SDValue Cond, SDValue T, SDValue F) {
// select undef, T, F --> T (if T is a constant), otherwise F
// select, ?, undef, F --> F
// select, ?, T, undef --> T
if (Cond.isUndef())
return isConstantValueOfAnyType(T) ? T : F;
if (T.isUndef())
return F;
if (F.isUndef())
return T;
// select true, T, F --> T
// select false, T, F --> F
if (auto *CondC = dyn_cast<ConstantSDNode>(Cond))
return CondC->isZero() ? F : T;
// TODO: This should simplify VSELECT with non-zero constant condition using
// something like this (but check boolean contents to be complete?):
if (ConstantSDNode *CondC = isConstOrConstSplat(Cond, /*AllowUndefs*/ false,
/*AllowTruncation*/ true))
if (CondC->isZero())
return F;
// select ?, T, T --> T
if (T == F)
return T;
return SDValue();
}
SDValue SelectionDAG::simplifyShift(SDValue X, SDValue Y) {
// shift undef, Y --> 0 (can always assume that the undef value is 0)
if (X.isUndef())
return getConstant(0, SDLoc(X.getNode()), X.getValueType());
// shift X, undef --> undef (because it may shift by the bitwidth)
if (Y.isUndef())
return getUNDEF(X.getValueType());
// shift 0, Y --> 0
// shift X, 0 --> X
if (isNullOrNullSplat(X) || isNullOrNullSplat(Y))
return X;
// shift X, C >= bitwidth(X) --> undef
// All vector elements must be too big (or undef) to avoid partial undefs.
auto isShiftTooBig = [X](ConstantSDNode *Val) {
return !Val || Val->getAPIntValue().uge(X.getScalarValueSizeInBits());
};
if (ISD::matchUnaryPredicate(Y, isShiftTooBig, true))
return getUNDEF(X.getValueType());
return SDValue();
}
SDValue SelectionDAG::simplifyFPBinop(unsigned Opcode, SDValue X, SDValue Y,
SDNodeFlags Flags) {
// If this operation has 'nnan' or 'ninf' and at least 1 disallowed operand
// (an undef operand can be chosen to be Nan/Inf), then the result of this
// operation is poison. That result can be relaxed to undef.
ConstantFPSDNode *XC = isConstOrConstSplatFP(X, /* AllowUndefs */ true);
ConstantFPSDNode *YC = isConstOrConstSplatFP(Y, /* AllowUndefs */ true);
bool HasNan = (XC && XC->getValueAPF().isNaN()) ||
(YC && YC->getValueAPF().isNaN());
bool HasInf = (XC && XC->getValueAPF().isInfinity()) ||
(YC && YC->getValueAPF().isInfinity());
if (Flags.hasNoNaNs() && (HasNan || X.isUndef() || Y.isUndef()))
return getUNDEF(X.getValueType());
if (Flags.hasNoInfs() && (HasInf || X.isUndef() || Y.isUndef()))
return getUNDEF(X.getValueType());
if (!YC)
return SDValue();
// X + -0.0 --> X
if (Opcode == ISD::FADD)
if (YC->getValueAPF().isNegZero())
return X;
// X - +0.0 --> X
if (Opcode == ISD::FSUB)
if (YC->getValueAPF().isPosZero())
return X;
// X * 1.0 --> X
// X / 1.0 --> X
if (Opcode == ISD::FMUL || Opcode == ISD::FDIV)
if (YC->getValueAPF().isExactlyValue(1.0))
return X;
// X * 0.0 --> 0.0
if (Opcode == ISD::FMUL && Flags.hasNoNaNs() && Flags.hasNoSignedZeros())
if (YC->getValueAPF().isZero())
return getConstantFP(0.0, SDLoc(Y), Y.getValueType());
return SDValue();
}
SDValue SelectionDAG::getVAArg(EVT VT, const SDLoc &dl, SDValue Chain,
SDValue Ptr, SDValue SV, unsigned Align) {
SDValue Ops[] = { Chain, Ptr, SV, getTargetConstant(Align, dl, MVT::i32) };
return getNode(ISD::VAARG, dl, getVTList(VT, MVT::Other), Ops);
}
SDValue SelectionDAG::getNode(unsigned Opcode, const SDLoc &DL, EVT VT,
ArrayRef<SDUse> Ops) {
switch (Ops.size()) {
case 0: return getNode(Opcode, DL, VT);
case 1: return getNode(Opcode, DL, VT, static_cast<const SDValue>(Ops[0]));
case 2: return getNode(Opcode, DL, VT, Ops[0], Ops[1]);
case 3: return getNode(Opcode, DL, VT, Ops[0], Ops[1], Ops[2]);
default: break;
}
// Copy from an SDUse array into an SDValue array for use with
// the regular getNode logic.
SmallVector<SDValue, 8> NewOps(Ops.begin(), Ops.end());
return getNode(Opcode, DL, VT, NewOps);
}
SDValue SelectionDAG::getNode(unsigned Opcode, const SDLoc &DL, EVT VT,
ArrayRef<SDValue> Ops) {
SDNodeFlags Flags;
if (Inserter)
Flags = Inserter->getFlags();
return getNode(Opcode, DL, VT, Ops, Flags);
}
SDValue SelectionDAG::getNode(unsigned Opcode, const SDLoc &DL, EVT VT,
ArrayRef<SDValue> Ops, const SDNodeFlags Flags) {
unsigned NumOps = Ops.size();
switch (NumOps) {
case 0: return getNode(Opcode, DL, VT);
case 1: return getNode(Opcode, DL, VT, Ops[0], Flags);
case 2: return getNode(Opcode, DL, VT, Ops[0], Ops[1], Flags);
case 3: return getNode(Opcode, DL, VT, Ops[0], Ops[1], Ops[2], Flags);
default: break;
}
#ifndef NDEBUG
for (const auto &Op : Ops)
assert(Op.getOpcode() != ISD::DELETED_NODE &&
"Operand is DELETED_NODE!");
#endif
switch (Opcode) {
default: break;
case ISD::BUILD_VECTOR:
// Attempt to simplify BUILD_VECTOR.
if (SDValue V = FoldBUILD_VECTOR(DL, VT, Ops, *this))
return V;
break;
case ISD::CONCAT_VECTORS:
if (SDValue V = foldCONCAT_VECTORS(DL, VT, Ops, *this))
return V;
break;
case ISD::SELECT_CC:
assert(NumOps == 5 && "SELECT_CC takes 5 operands!");
assert(Ops[0].getValueType() == Ops[1].getValueType() &&
"LHS and RHS of condition must have same type!");
assert(Ops[2].getValueType() == Ops[3].getValueType() &&
"True and False arms of SelectCC must have same type!");
assert(Ops[2].getValueType() == VT &&
"select_cc node must be of same type as true and false value!");
assert((!Ops[0].getValueType().isVector() ||
Ops[0].getValueType().getVectorElementCount() ==
VT.getVectorElementCount()) &&
"Expected select_cc with vector result to have the same sized "
"comparison type!");
break;
case ISD::BR_CC:
assert(NumOps == 5 && "BR_CC takes 5 operands!");
assert(Ops[2].getValueType() == Ops[3].getValueType() &&
"LHS/RHS of comparison should match types!");
break;
case ISD::VP_ADD:
case ISD::VP_SUB:
// If it is VP_ADD/VP_SUB mask operation then turn it to VP_XOR
if (VT.isVector() && VT.getVectorElementType() == MVT::i1)
Opcode = ISD::VP_XOR;
break;
case ISD::VP_MUL:
// If it is VP_MUL mask operation then turn it to VP_AND
if (VT.isVector() && VT.getVectorElementType() == MVT::i1)
Opcode = ISD::VP_AND;
break;
case ISD::VP_REDUCE_MUL:
// If it is VP_REDUCE_MUL mask operation then turn it to VP_REDUCE_AND
if (VT == MVT::i1)
Opcode = ISD::VP_REDUCE_AND;
break;
case ISD::VP_REDUCE_ADD:
// If it is VP_REDUCE_ADD mask operation then turn it to VP_REDUCE_XOR
if (VT == MVT::i1)
Opcode = ISD::VP_REDUCE_XOR;
break;
case ISD::VP_REDUCE_SMAX:
case ISD::VP_REDUCE_UMIN:
// If it is VP_REDUCE_SMAX/VP_REDUCE_UMIN mask operation then turn it to
// VP_REDUCE_AND.
if (VT == MVT::i1)
Opcode = ISD::VP_REDUCE_AND;
break;
case ISD::VP_REDUCE_SMIN:
case ISD::VP_REDUCE_UMAX:
// If it is VP_REDUCE_SMIN/VP_REDUCE_UMAX mask operation then turn it to
// VP_REDUCE_OR.
if (VT == MVT::i1)
Opcode = ISD::VP_REDUCE_OR;
break;
}
// Memoize nodes.
SDNode *N;
SDVTList VTs = getVTList(VT);
if (VT != MVT::Glue) {
FoldingSetNodeID ID;
AddNodeIDNode(ID, Opcode, VTs, Ops);
void *IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, DL, IP))
return SDValue(E, 0);
N = newSDNode<SDNode>(Opcode, DL.getIROrder(), DL.getDebugLoc(), VTs);
createOperands(N, Ops);
CSEMap.InsertNode(N, IP);
} else {
N = newSDNode<SDNode>(Opcode, DL.getIROrder(), DL.getDebugLoc(), VTs);
createOperands(N, Ops);
}
N->setFlags(Flags);
InsertNode(N);
SDValue V(N, 0);
NewSDValueDbgMsg(V, "Creating new node: ", this);
return V;
}
SDValue SelectionDAG::getNode(unsigned Opcode, const SDLoc &DL,
ArrayRef<EVT> ResultTys, ArrayRef<SDValue> Ops) {
return getNode(Opcode, DL, getVTList(ResultTys), Ops);
}
SDValue SelectionDAG::getNode(unsigned Opcode, const SDLoc &DL, SDVTList VTList,
ArrayRef<SDValue> Ops) {
SDNodeFlags Flags;
if (Inserter)
Flags = Inserter->getFlags();
return getNode(Opcode, DL, VTList, Ops, Flags);
}
SDValue SelectionDAG::getNode(unsigned Opcode, const SDLoc &DL, SDVTList VTList,
ArrayRef<SDValue> Ops, const SDNodeFlags Flags) {
if (VTList.NumVTs == 1)
return getNode(Opcode, DL, VTList.VTs[0], Ops, Flags);
#ifndef NDEBUG
for (const auto &Op : Ops)
assert(Op.getOpcode() != ISD::DELETED_NODE &&
"Operand is DELETED_NODE!");
#endif
switch (Opcode) {
case ISD::SADDO:
case ISD::UADDO:
case ISD::SSUBO:
case ISD::USUBO: {
assert(VTList.NumVTs == 2 && Ops.size() == 2 &&
"Invalid add/sub overflow op!");
assert(VTList.VTs[0].isInteger() && VTList.VTs[1].isInteger() &&
Ops[0].getValueType() == Ops[1].getValueType() &&
Ops[0].getValueType() == VTList.VTs[0] &&
"Binary operator types must match!");
SDValue N1 = Ops[0], N2 = Ops[1];
canonicalizeCommutativeBinop(Opcode, N1, N2);
// (X +- 0) -> X with zero-overflow.
ConstantSDNode *N2CV = isConstOrConstSplat(N2, /*AllowUndefs*/ false,
/*AllowTruncation*/ true);
if (N2CV && N2CV->isZero()) {
SDValue ZeroOverFlow = getConstant(0, DL, VTList.VTs[1]);
return getNode(ISD::MERGE_VALUES, DL, VTList, {N1, ZeroOverFlow}, Flags);
}
if (VTList.VTs[0].isVector() &&
VTList.VTs[0].getVectorElementType() == MVT::i1 &&
VTList.VTs[1].getVectorElementType() == MVT::i1) {
SDValue F1 = getFreeze(N1);
SDValue F2 = getFreeze(N2);
// {vXi1,vXi1} (u/s)addo(vXi1 x, vXi1y) -> {xor(x,y),and(x,y)}
if (Opcode == ISD::UADDO || Opcode == ISD::SADDO)
return getNode(ISD::MERGE_VALUES, DL, VTList,
{getNode(ISD::XOR, DL, VTList.VTs[0], F1, F2),
getNode(ISD::AND, DL, VTList.VTs[1], F1, F2)},
Flags);
// {vXi1,vXi1} (u/s)subo(vXi1 x, vXi1y) -> {xor(x,y),and(~x,y)}
if (Opcode == ISD::USUBO || Opcode == ISD::SSUBO) {
SDValue NotF1 = getNOT(DL, F1, VTList.VTs[0]);
return getNode(ISD::MERGE_VALUES, DL, VTList,
{getNode(ISD::XOR, DL, VTList.VTs[0], F1, F2),
getNode(ISD::AND, DL, VTList.VTs[1], NotF1, F2)},
Flags);
}
}
break;
}
case ISD::SMUL_LOHI:
case ISD::UMUL_LOHI: {
assert(VTList.NumVTs == 2 && Ops.size() == 2 && "Invalid mul lo/hi op!");
assert(VTList.VTs[0].isInteger() && VTList.VTs[0] == VTList.VTs[1] &&
VTList.VTs[0] == Ops[0].getValueType() &&
VTList.VTs[0] == Ops[1].getValueType() &&
"Binary operator types must match!");
// Constant fold.
ConstantSDNode *LHS = dyn_cast<ConstantSDNode>(Ops[0]);
ConstantSDNode *RHS = dyn_cast<ConstantSDNode>(Ops[1]);
if (LHS && RHS) {
unsigned Width = VTList.VTs[0].getScalarSizeInBits();
unsigned OutWidth = Width * 2;
APInt Val = LHS->getAPIntValue();
APInt Mul = RHS->getAPIntValue();
if (Opcode == ISD::SMUL_LOHI) {
Val = Val.sext(OutWidth);
Mul = Mul.sext(OutWidth);
} else {
Val = Val.zext(OutWidth);
Mul = Mul.zext(OutWidth);
}
Val *= Mul;
SDValue Hi =
getConstant(Val.extractBits(Width, Width), DL, VTList.VTs[0]);
SDValue Lo = getConstant(Val.trunc(Width), DL, VTList.VTs[0]);
return getNode(ISD::MERGE_VALUES, DL, VTList, {Lo, Hi}, Flags);
}
break;
}
case ISD::FFREXP: {
assert(VTList.NumVTs == 2 && Ops.size() == 1 && "Invalid ffrexp op!");
assert(VTList.VTs[0].isFloatingPoint() && VTList.VTs[1].isInteger() &&
VTList.VTs[0] == Ops[0].getValueType() && "frexp type mismatch");
if (const ConstantFPSDNode *C = dyn_cast<ConstantFPSDNode>(Ops[0])) {
int FrexpExp;
APFloat FrexpMant =
frexp(C->getValueAPF(), FrexpExp, APFloat::rmNearestTiesToEven);
SDValue Result0 = getConstantFP(FrexpMant, DL, VTList.VTs[0]);
SDValue Result1 =
getConstant(FrexpMant.isFinite() ? FrexpExp : 0, DL, VTList.VTs[1]);
return getNode(ISD::MERGE_VALUES, DL, VTList, {Result0, Result1}, Flags);
}
break;
}
case ISD::STRICT_FP_EXTEND:
assert(VTList.NumVTs == 2 && Ops.size() == 2 &&
"Invalid STRICT_FP_EXTEND!");
assert(VTList.VTs[0].isFloatingPoint() &&
Ops[1].getValueType().isFloatingPoint() && "Invalid FP cast!");
assert(VTList.VTs[0].isVector() == Ops[1].getValueType().isVector() &&
"STRICT_FP_EXTEND result type should be vector iff the operand "
"type is vector!");
assert((!VTList.VTs[0].isVector() ||
VTList.VTs[0].getVectorElementCount() ==
Ops[1].getValueType().getVectorElementCount()) &&
"Vector element count mismatch!");
assert(Ops[1].getValueType().bitsLT(VTList.VTs[0]) &&
"Invalid fpext node, dst <= src!");
break;
case ISD::STRICT_FP_ROUND:
assert(VTList.NumVTs == 2 && Ops.size() == 3 && "Invalid STRICT_FP_ROUND!");
assert(VTList.VTs[0].isVector() == Ops[1].getValueType().isVector() &&
"STRICT_FP_ROUND result type should be vector iff the operand "
"type is vector!");
assert((!VTList.VTs[0].isVector() ||
VTList.VTs[0].getVectorElementCount() ==
Ops[1].getValueType().getVectorElementCount()) &&
"Vector element count mismatch!");
assert(VTList.VTs[0].isFloatingPoint() &&
Ops[1].getValueType().isFloatingPoint() &&
VTList.VTs[0].bitsLT(Ops[1].getValueType()) &&
isa<ConstantSDNode>(Ops[2]) &&
(Ops[2]->getAsZExtVal() == 0 || Ops[2]->getAsZExtVal() == 1) &&
"Invalid STRICT_FP_ROUND!");
break;
#if 0
// FIXME: figure out how to safely handle things like
// int foo(int x) { return 1 << (x & 255); }
// int bar() { return foo(256); }
case ISD::SRA_PARTS:
case ISD::SRL_PARTS:
case ISD::SHL_PARTS:
if (N3.getOpcode() == ISD::SIGN_EXTEND_INREG &&
cast<VTSDNode>(N3.getOperand(1))->getVT() != MVT::i1)
return getNode(Opcode, DL, VT, N1, N2, N3.getOperand(0));
else if (N3.getOpcode() == ISD::AND)
if (ConstantSDNode *AndRHS = dyn_cast<ConstantSDNode>(N3.getOperand(1))) {
// If the and is only masking out bits that cannot effect the shift,
// eliminate the and.
unsigned NumBits = VT.getScalarSizeInBits()*2;
if ((AndRHS->getValue() & (NumBits-1)) == NumBits-1)
return getNode(Opcode, DL, VT, N1, N2, N3.getOperand(0));
}
break;
#endif
}
// Memoize the node unless it returns a glue result.
SDNode *N;
if (VTList.VTs[VTList.NumVTs-1] != MVT::Glue) {
FoldingSetNodeID ID;
AddNodeIDNode(ID, Opcode, VTList, Ops);
void *IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, DL, IP))
return SDValue(E, 0);
N = newSDNode<SDNode>(Opcode, DL.getIROrder(), DL.getDebugLoc(), VTList);
createOperands(N, Ops);
CSEMap.InsertNode(N, IP);
} else {
N = newSDNode<SDNode>(Opcode, DL.getIROrder(), DL.getDebugLoc(), VTList);
createOperands(N, Ops);
}
N->setFlags(Flags);
InsertNode(N);
SDValue V(N, 0);
NewSDValueDbgMsg(V, "Creating new node: ", this);
return V;
}
SDValue SelectionDAG::getNode(unsigned Opcode, const SDLoc &DL,
SDVTList VTList) {
return getNode(Opcode, DL, VTList, std::nullopt);
}
SDValue SelectionDAG::getNode(unsigned Opcode, const SDLoc &DL, SDVTList VTList,
SDValue N1) {
SDValue Ops[] = { N1 };
return getNode(Opcode, DL, VTList, Ops);
}
SDValue SelectionDAG::getNode(unsigned Opcode, const SDLoc &DL, SDVTList VTList,
SDValue N1, SDValue N2) {
SDValue Ops[] = { N1, N2 };
return getNode(Opcode, DL, VTList, Ops);
}
SDValue SelectionDAG::getNode(unsigned Opcode, const SDLoc &DL, SDVTList VTList,
SDValue N1, SDValue N2, SDValue N3) {
SDValue Ops[] = { N1, N2, N3 };
return getNode(Opcode, DL, VTList, Ops);
}
SDValue SelectionDAG::getNode(unsigned Opcode, const SDLoc &DL, SDVTList VTList,
SDValue N1, SDValue N2, SDValue N3, SDValue N4) {
SDValue Ops[] = { N1, N2, N3, N4 };
return getNode(Opcode, DL, VTList, Ops);
}
SDValue SelectionDAG::getNode(unsigned Opcode, const SDLoc &DL, SDVTList VTList,
SDValue N1, SDValue N2, SDValue N3, SDValue N4,
SDValue N5) {
SDValue Ops[] = { N1, N2, N3, N4, N5 };
return getNode(Opcode, DL, VTList, Ops);
}
SDVTList SelectionDAG::getVTList(EVT VT) {
return makeVTList(SDNode::getValueTypeList(VT), 1);
}
SDVTList SelectionDAG::getVTList(EVT VT1, EVT VT2) {
FoldingSetNodeID ID;
ID.AddInteger(2U);
ID.AddInteger(VT1.getRawBits());
ID.AddInteger(VT2.getRawBits());
void *IP = nullptr;
SDVTListNode *Result = VTListMap.FindNodeOrInsertPos(ID, IP);
if (!Result) {
EVT *Array = Allocator.Allocate<EVT>(2);
Array[0] = VT1;
Array[1] = VT2;
Result = new (Allocator) SDVTListNode(ID.Intern(Allocator), Array, 2);
VTListMap.InsertNode(Result, IP);
}
return Result->getSDVTList();
}
SDVTList SelectionDAG::getVTList(EVT VT1, EVT VT2, EVT VT3) {
FoldingSetNodeID ID;
ID.AddInteger(3U);
ID.AddInteger(VT1.getRawBits());
ID.AddInteger(VT2.getRawBits());
ID.AddInteger(VT3.getRawBits());
void *IP = nullptr;
SDVTListNode *Result = VTListMap.FindNodeOrInsertPos(ID, IP);
if (!Result) {
EVT *Array = Allocator.Allocate<EVT>(3);
Array[0] = VT1;
Array[1] = VT2;
Array[2] = VT3;
Result = new (Allocator) SDVTListNode(ID.Intern(Allocator), Array, 3);
VTListMap.InsertNode(Result, IP);
}
return Result->getSDVTList();
}
SDVTList SelectionDAG::getVTList(EVT VT1, EVT VT2, EVT VT3, EVT VT4) {
FoldingSetNodeID ID;
ID.AddInteger(4U);
ID.AddInteger(VT1.getRawBits());
ID.AddInteger(VT2.getRawBits());
ID.AddInteger(VT3.getRawBits());
ID.AddInteger(VT4.getRawBits());
void *IP = nullptr;
SDVTListNode *Result = VTListMap.FindNodeOrInsertPos(ID, IP);
if (!Result) {
EVT *Array = Allocator.Allocate<EVT>(4);
Array[0] = VT1;
Array[1] = VT2;
Array[2] = VT3;
Array[3] = VT4;
Result = new (Allocator) SDVTListNode(ID.Intern(Allocator), Array, 4);
VTListMap.InsertNode(Result, IP);
}
return Result->getSDVTList();
}
SDVTList SelectionDAG::getVTList(ArrayRef<EVT> VTs) {
unsigned NumVTs = VTs.size();
FoldingSetNodeID ID;
ID.AddInteger(NumVTs);
for (unsigned index = 0; index < NumVTs; index++) {
ID.AddInteger(VTs[index].getRawBits());
}
void *IP = nullptr;
SDVTListNode *Result = VTListMap.FindNodeOrInsertPos(ID, IP);
if (!Result) {
EVT *Array = Allocator.Allocate<EVT>(NumVTs);
llvm::copy(VTs, Array);
Result = new (Allocator) SDVTListNode(ID.Intern(Allocator), Array, NumVTs);
VTListMap.InsertNode(Result, IP);
}
return Result->getSDVTList();
}
/// UpdateNodeOperands - *Mutate* the specified node in-place to have the
/// specified operands. If the resultant node already exists in the DAG,
/// this does not modify the specified node, instead it returns the node that
/// already exists. If the resultant node does not exist in the DAG, the
/// input node is returned. As a degenerate case, if you specify the same
/// input operands as the node already has, the input node is returned.
SDNode *SelectionDAG::UpdateNodeOperands(SDNode *N, SDValue Op) {
assert(N->getNumOperands() == 1 && "Update with wrong number of operands");
// Check to see if there is no change.
if (Op == N->getOperand(0)) return N;
// See if the modified node already exists.
void *InsertPos = nullptr;
if (SDNode *Existing = FindModifiedNodeSlot(N, Op, InsertPos))
return Existing;
// Nope it doesn't. Remove the node from its current place in the maps.
if (InsertPos)
if (!RemoveNodeFromCSEMaps(N))
InsertPos = nullptr;
// Now we update the operands.
N->OperandList[0].set(Op);
updateDivergence(N);
// If this gets put into a CSE map, add it.
if (InsertPos) CSEMap.InsertNode(N, InsertPos);
return N;
}
SDNode *SelectionDAG::UpdateNodeOperands(SDNode *N, SDValue Op1, SDValue Op2) {
assert(N->getNumOperands() == 2 && "Update with wrong number of operands");
// Check to see if there is no change.
if (Op1 == N->getOperand(0) && Op2 == N->getOperand(1))
return N; // No operands changed, just return the input node.
// See if the modified node already exists.
void *InsertPos = nullptr;
if (SDNode *Existing = FindModifiedNodeSlot(N, Op1, Op2, InsertPos))
return Existing;
// Nope it doesn't. Remove the node from its current place in the maps.
if (InsertPos)
if (!RemoveNodeFromCSEMaps(N))
InsertPos = nullptr;
// Now we update the operands.
if (N->OperandList[0] != Op1)
N->OperandList[0].set(Op1);
if (N->OperandList[1] != Op2)
N->OperandList[1].set(Op2);
updateDivergence(N);
// If this gets put into a CSE map, add it.
if (InsertPos) CSEMap.InsertNode(N, InsertPos);
return N;
}
SDNode *SelectionDAG::
UpdateNodeOperands(SDNode *N, SDValue Op1, SDValue Op2, SDValue Op3) {
SDValue Ops[] = { Op1, Op2, Op3 };
return UpdateNodeOperands(N, Ops);
}
SDNode *SelectionDAG::
UpdateNodeOperands(SDNode *N, SDValue Op1, SDValue Op2,
SDValue Op3, SDValue Op4) {
SDValue Ops[] = { Op1, Op2, Op3, Op4 };
return UpdateNodeOperands(N, Ops);
}
SDNode *SelectionDAG::
UpdateNodeOperands(SDNode *N, SDValue Op1, SDValue Op2,
SDValue Op3, SDValue Op4, SDValue Op5) {
SDValue Ops[] = { Op1, Op2, Op3, Op4, Op5 };
return UpdateNodeOperands(N, Ops);
}
SDNode *SelectionDAG::
UpdateNodeOperands(SDNode *N, ArrayRef<SDValue> Ops) {
unsigned NumOps = Ops.size();
assert(N->getNumOperands() == NumOps &&
"Update with wrong number of operands");
// If no operands changed just return the input node.
if (std::equal(Ops.begin(), Ops.end(), N->op_begin()))
return N;
// See if the modified node already exists.
void *InsertPos = nullptr;
if (SDNode *Existing = FindModifiedNodeSlot(N, Ops, InsertPos))
return Existing;
// Nope it doesn't. Remove the node from its current place in the maps.
if (InsertPos)
if (!RemoveNodeFromCSEMaps(N))
InsertPos = nullptr;
// Now we update the operands.
for (unsigned i = 0; i != NumOps; ++i)
if (N->OperandList[i] != Ops[i])
N->OperandList[i].set(Ops[i]);
updateDivergence(N);
// If this gets put into a CSE map, add it.
if (InsertPos) CSEMap.InsertNode(N, InsertPos);
return N;
}
/// DropOperands - Release the operands and set this node to have
/// zero operands.
void SDNode::DropOperands() {
// Unlike the code in MorphNodeTo that does this, we don't need to
// watch for dead nodes here.
for (op_iterator I = op_begin(), E = op_end(); I != E; ) {
SDUse &Use = *I++;
Use.set(SDValue());
}
}
void SelectionDAG::setNodeMemRefs(MachineSDNode *N,
ArrayRef<MachineMemOperand *> NewMemRefs) {
if (NewMemRefs.empty()) {
N->clearMemRefs();
return;
}
// Check if we can avoid allocating by storing a single reference directly.
if (NewMemRefs.size() == 1) {
N->MemRefs = NewMemRefs[0];
N->NumMemRefs = 1;
return;
}
MachineMemOperand **MemRefsBuffer =
Allocator.template Allocate<MachineMemOperand *>(NewMemRefs.size());
llvm::copy(NewMemRefs, MemRefsBuffer);
N->MemRefs = MemRefsBuffer;
N->NumMemRefs = static_cast<int>(NewMemRefs.size());
}
/// SelectNodeTo - These are wrappers around MorphNodeTo that accept a
/// machine opcode.
///
SDNode *SelectionDAG::SelectNodeTo(SDNode *N, unsigned MachineOpc,
EVT VT) {
SDVTList VTs = getVTList(VT);
return SelectNodeTo(N, MachineOpc, VTs, std::nullopt);
}
SDNode *SelectionDAG::SelectNodeTo(SDNode *N, unsigned MachineOpc,
EVT VT, SDValue Op1) {
SDVTList VTs = getVTList(VT);
SDValue Ops[] = { Op1 };
return SelectNodeTo(N, MachineOpc, VTs, Ops);
}
SDNode *SelectionDAG::SelectNodeTo(SDNode *N, unsigned MachineOpc,
EVT VT, SDValue Op1,
SDValue Op2) {
SDVTList VTs = getVTList(VT);
SDValue Ops[] = { Op1, Op2 };
return SelectNodeTo(N, MachineOpc, VTs, Ops);
}
SDNode *SelectionDAG::SelectNodeTo(SDNode *N, unsigned MachineOpc,
EVT VT, SDValue Op1,
SDValue Op2, SDValue Op3) {
SDVTList VTs = getVTList(VT);
SDValue Ops[] = { Op1, Op2, Op3 };
return SelectNodeTo(N, MachineOpc, VTs, Ops);
}
SDNode *SelectionDAG::SelectNodeTo(SDNode *N, unsigned MachineOpc,
EVT VT, ArrayRef<SDValue> Ops) {
SDVTList VTs = getVTList(VT);
return SelectNodeTo(N, MachineOpc, VTs, Ops);
}
SDNode *SelectionDAG::SelectNodeTo(SDNode *N, unsigned MachineOpc,
EVT VT1, EVT VT2, ArrayRef<SDValue> Ops) {
SDVTList VTs = getVTList(VT1, VT2);
return SelectNodeTo(N, MachineOpc, VTs, Ops);
}
SDNode *SelectionDAG::SelectNodeTo(SDNode *N, unsigned MachineOpc,
EVT VT1, EVT VT2) {
SDVTList VTs = getVTList(VT1, VT2);
return SelectNodeTo(N, MachineOpc, VTs, std::nullopt);
}
SDNode *SelectionDAG::SelectNodeTo(SDNode *N, unsigned MachineOpc,
EVT VT1, EVT VT2, EVT VT3,
ArrayRef<SDValue> Ops) {
SDVTList VTs = getVTList(VT1, VT2, VT3);
return SelectNodeTo(N, MachineOpc, VTs, Ops);
}
SDNode *SelectionDAG::SelectNodeTo(SDNode *N, unsigned MachineOpc,
EVT VT1, EVT VT2,
SDValue Op1, SDValue Op2) {
SDVTList VTs = getVTList(VT1, VT2);
SDValue Ops[] = { Op1, Op2 };
return SelectNodeTo(N, MachineOpc, VTs, Ops);
}
SDNode *SelectionDAG::SelectNodeTo(SDNode *N, unsigned MachineOpc,
SDVTList VTs,ArrayRef<SDValue> Ops) {
SDNode *New = MorphNodeTo(N, ~MachineOpc, VTs, Ops);
// Reset the NodeID to -1.
New->setNodeId(-1);
if (New != N) {
ReplaceAllUsesWith(N, New);
RemoveDeadNode(N);
}
return New;
}
/// UpdateSDLocOnMergeSDNode - If the opt level is -O0 then it throws away
/// the line number information on the merged node since it is not possible to
/// preserve the information that operation is associated with multiple lines.
/// This will make the debugger working better at -O0, were there is a higher
/// probability having other instructions associated with that line.
///
/// For IROrder, we keep the smaller of the two
SDNode *SelectionDAG::UpdateSDLocOnMergeSDNode(SDNode *N, const SDLoc &OLoc) {
DebugLoc NLoc = N->getDebugLoc();
if (NLoc && OptLevel == CodeGenOptLevel::None && OLoc.getDebugLoc() != NLoc) {
N->setDebugLoc(DebugLoc());
}
unsigned Order = std::min(N->getIROrder(), OLoc.getIROrder());
N->setIROrder(Order);
return N;
}
/// MorphNodeTo - This *mutates* the specified node to have the specified
/// return type, opcode, and operands.
///
/// Note that MorphNodeTo returns the resultant node. If there is already a
/// node of the specified opcode and operands, it returns that node instead of
/// the current one. Note that the SDLoc need not be the same.
///
/// Using MorphNodeTo is faster than creating a new node and swapping it in
/// with ReplaceAllUsesWith both because it often avoids allocating a new
/// node, and because it doesn't require CSE recalculation for any of
/// the node's users.
///
/// However, note that MorphNodeTo recursively deletes dead nodes from the DAG.
/// As a consequence it isn't appropriate to use from within the DAG combiner or
/// the legalizer which maintain worklists that would need to be updated when
/// deleting things.
SDNode *SelectionDAG::MorphNodeTo(SDNode *N, unsigned Opc,
SDVTList VTs, ArrayRef<SDValue> Ops) {
// If an identical node already exists, use it.
void *IP = nullptr;
if (VTs.VTs[VTs.NumVTs-1] != MVT::Glue) {
FoldingSetNodeID ID;
AddNodeIDNode(ID, Opc, VTs, Ops);
if (SDNode *ON = FindNodeOrInsertPos(ID, SDLoc(N), IP))
return UpdateSDLocOnMergeSDNode(ON, SDLoc(N));
}
if (!RemoveNodeFromCSEMaps(N))
IP = nullptr;
// Start the morphing.
N->NodeType = Opc;
N->ValueList = VTs.VTs;
N->NumValues = VTs.NumVTs;
// Clear the operands list, updating used nodes to remove this from their
// use list. Keep track of any operands that become dead as a result.
SmallPtrSet<SDNode*, 16> DeadNodeSet;
for (SDNode::op_iterator I = N->op_begin(), E = N->op_end(); I != E; ) {
SDUse &Use = *I++;
SDNode *Used = Use.getNode();
Use.set(SDValue());
if (Used->use_empty())
DeadNodeSet.insert(Used);
}
// For MachineNode, initialize the memory references information.
if (MachineSDNode *MN = dyn_cast<MachineSDNode>(N))
MN->clearMemRefs();
// Swap for an appropriately sized array from the recycler.
removeOperands(N);
createOperands(N, Ops);
// Delete any nodes that are still dead after adding the uses for the
// new operands.
if (!DeadNodeSet.empty()) {
SmallVector<SDNode *, 16> DeadNodes;
for (SDNode *N : DeadNodeSet)
if (N->use_empty())
DeadNodes.push_back(N);
RemoveDeadNodes(DeadNodes);
}
if (IP)
CSEMap.InsertNode(N, IP); // Memoize the new node.
return N;
}
SDNode* SelectionDAG::mutateStrictFPToFP(SDNode *Node) {
unsigned OrigOpc = Node->getOpcode();
unsigned NewOpc;
switch (OrigOpc) {
default:
llvm_unreachable("mutateStrictFPToFP called with unexpected opcode!");
#define DAG_INSTRUCTION(NAME, NARG, ROUND_MODE, INTRINSIC, DAGN) \
case ISD::STRICT_##DAGN: NewOpc = ISD::DAGN; break;
#define CMP_INSTRUCTION(NAME, NARG, ROUND_MODE, INTRINSIC, DAGN) \
case ISD::STRICT_##DAGN: NewOpc = ISD::SETCC; break;
#include "llvm/IR/ConstrainedOps.def"
}
assert(Node->getNumValues() == 2 && "Unexpected number of results!");
// We're taking this node out of the chain, so we need to re-link things.
SDValue InputChain = Node->getOperand(0);
SDValue OutputChain = SDValue(Node, 1);
ReplaceAllUsesOfValueWith(OutputChain, InputChain);
SmallVector<SDValue, 3> Ops;
for (unsigned i = 1, e = Node->getNumOperands(); i != e; ++i)
Ops.push_back(Node->getOperand(i));
SDVTList VTs = getVTList(Node->getValueType(0));
SDNode *Res = MorphNodeTo(Node, NewOpc, VTs, Ops);
// MorphNodeTo can operate in two ways: if an existing node with the
// specified operands exists, it can just return it. Otherwise, it
// updates the node in place to have the requested operands.
if (Res == Node) {
// If we updated the node in place, reset the node ID. To the isel,
// this should be just like a newly allocated machine node.
Res->setNodeId(-1);
} else {
ReplaceAllUsesWith(Node, Res);
RemoveDeadNode(Node);
}
return Res;
}
/// getMachineNode - These are used for target selectors to create a new node
/// with specified return type(s), MachineInstr opcode, and operands.
///
/// Note that getMachineNode returns the resultant node. If there is already a
/// node of the specified opcode and operands, it returns that node instead of
/// the current one.
MachineSDNode *SelectionDAG::getMachineNode(unsigned Opcode, const SDLoc &dl,
EVT VT) {
SDVTList VTs = getVTList(VT);
return getMachineNode(Opcode, dl, VTs, std::nullopt);
}
MachineSDNode *SelectionDAG::getMachineNode(unsigned Opcode, const SDLoc &dl,
EVT VT, SDValue Op1) {
SDVTList VTs = getVTList(VT);
SDValue Ops[] = { Op1 };
return getMachineNode(Opcode, dl, VTs, Ops);
}
MachineSDNode *SelectionDAG::getMachineNode(unsigned Opcode, const SDLoc &dl,
EVT VT, SDValue Op1, SDValue Op2) {
SDVTList VTs = getVTList(VT);
SDValue Ops[] = { Op1, Op2 };
return getMachineNode(Opcode, dl, VTs, Ops);
}
MachineSDNode *SelectionDAG::getMachineNode(unsigned Opcode, const SDLoc &dl,
EVT VT, SDValue Op1, SDValue Op2,
SDValue Op3) {
SDVTList VTs = getVTList(VT);
SDValue Ops[] = { Op1, Op2, Op3 };
return getMachineNode(Opcode, dl, VTs, Ops);
}
MachineSDNode *SelectionDAG::getMachineNode(unsigned Opcode, const SDLoc &dl,
EVT VT, ArrayRef<SDValue> Ops) {
SDVTList VTs = getVTList(VT);
return getMachineNode(Opcode, dl, VTs, Ops);
}
MachineSDNode *SelectionDAG::getMachineNode(unsigned Opcode, const SDLoc &dl,
EVT VT1, EVT VT2, SDValue Op1,
SDValue Op2) {
SDVTList VTs = getVTList(VT1, VT2);
SDValue Ops[] = { Op1, Op2 };
return getMachineNode(Opcode, dl, VTs, Ops);
}
MachineSDNode *SelectionDAG::getMachineNode(unsigned Opcode, const SDLoc &dl,
EVT VT1, EVT VT2, SDValue Op1,
SDValue Op2, SDValue Op3) {
SDVTList VTs = getVTList(VT1, VT2);
SDValue Ops[] = { Op1, Op2, Op3 };
return getMachineNode(Opcode, dl, VTs, Ops);
}
MachineSDNode *SelectionDAG::getMachineNode(unsigned Opcode, const SDLoc &dl,
EVT VT1, EVT VT2,
ArrayRef<SDValue> Ops) {
SDVTList VTs = getVTList(VT1, VT2);
return getMachineNode(Opcode, dl, VTs, Ops);
}
MachineSDNode *SelectionDAG::getMachineNode(unsigned Opcode, const SDLoc &dl,
EVT VT1, EVT VT2, EVT VT3,
SDValue Op1, SDValue Op2) {
SDVTList VTs = getVTList(VT1, VT2, VT3);
SDValue Ops[] = { Op1, Op2 };
return getMachineNode(Opcode, dl, VTs, Ops);
}
MachineSDNode *SelectionDAG::getMachineNode(unsigned Opcode, const SDLoc &dl,
EVT VT1, EVT VT2, EVT VT3,
SDValue Op1, SDValue Op2,
SDValue Op3) {
SDVTList VTs = getVTList(VT1, VT2, VT3);
SDValue Ops[] = { Op1, Op2, Op3 };
return getMachineNode(Opcode, dl, VTs, Ops);
}
MachineSDNode *SelectionDAG::getMachineNode(unsigned Opcode, const SDLoc &dl,
EVT VT1, EVT VT2, EVT VT3,
ArrayRef<SDValue> Ops) {
SDVTList VTs = getVTList(VT1, VT2, VT3);
return getMachineNode(Opcode, dl, VTs, Ops);
}
MachineSDNode *SelectionDAG::getMachineNode(unsigned Opcode, const SDLoc &dl,
ArrayRef<EVT> ResultTys,
ArrayRef<SDValue> Ops) {
SDVTList VTs = getVTList(ResultTys);
return getMachineNode(Opcode, dl, VTs, Ops);
}
MachineSDNode *SelectionDAG::getMachineNode(unsigned Opcode, const SDLoc &DL,
SDVTList VTs,
ArrayRef<SDValue> Ops) {
bool DoCSE = VTs.VTs[VTs.NumVTs-1] != MVT::Glue;
MachineSDNode *N;
void *IP = nullptr;
if (DoCSE) {
FoldingSetNodeID ID;
AddNodeIDNode(ID, ~Opcode, VTs, Ops);
IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, DL, IP)) {
return cast<MachineSDNode>(UpdateSDLocOnMergeSDNode(E, DL));
}
}
// Allocate a new MachineSDNode.
N = newSDNode<MachineSDNode>(~Opcode, DL.getIROrder(), DL.getDebugLoc(), VTs);
createOperands(N, Ops);
if (DoCSE)
CSEMap.InsertNode(N, IP);
InsertNode(N);
NewSDValueDbgMsg(SDValue(N, 0), "Creating new machine node: ", this);
return N;
}
/// getTargetExtractSubreg - A convenience function for creating
/// TargetOpcode::EXTRACT_SUBREG nodes.
SDValue SelectionDAG::getTargetExtractSubreg(int SRIdx, const SDLoc &DL, EVT VT,
SDValue Operand) {
SDValue SRIdxVal = getTargetConstant(SRIdx, DL, MVT::i32);
SDNode *Subreg = getMachineNode(TargetOpcode::EXTRACT_SUBREG, DL,
VT, Operand, SRIdxVal);
return SDValue(Subreg, 0);
}
/// getTargetInsertSubreg - A convenience function for creating
/// TargetOpcode::INSERT_SUBREG nodes.
SDValue SelectionDAG::getTargetInsertSubreg(int SRIdx, const SDLoc &DL, EVT VT,
SDValue Operand, SDValue Subreg) {
SDValue SRIdxVal = getTargetConstant(SRIdx, DL, MVT::i32);
SDNode *Result = getMachineNode(TargetOpcode::INSERT_SUBREG, DL,
VT, Operand, Subreg, SRIdxVal);
return SDValue(Result, 0);
}
/// getNodeIfExists - Get the specified node if it's already available, or
/// else return NULL.
SDNode *SelectionDAG::getNodeIfExists(unsigned Opcode, SDVTList VTList,
ArrayRef<SDValue> Ops) {
SDNodeFlags Flags;
if (Inserter)
Flags = Inserter->getFlags();
return getNodeIfExists(Opcode, VTList, Ops, Flags);
}
SDNode *SelectionDAG::getNodeIfExists(unsigned Opcode, SDVTList VTList,
ArrayRef<SDValue> Ops,
const SDNodeFlags Flags) {
if (VTList.VTs[VTList.NumVTs - 1] != MVT::Glue) {
FoldingSetNodeID ID;
AddNodeIDNode(ID, Opcode, VTList, Ops);
void *IP = nullptr;
if (SDNode *E = FindNodeOrInsertPos(ID, SDLoc(), IP)) {
E->intersectFlagsWith(Flags);
return E;
}
}
return nullptr;
}
/// doesNodeExist - Check if a node exists without modifying its flags.
bool SelectionDAG::doesNodeExist(unsigned Opcode, SDVTList VTList,
ArrayRef<SDValue> Ops) {
if (VTList.VTs[VTList.NumVTs - 1] != MVT::Glue) {
FoldingSetNodeID ID;
AddNodeIDNode(ID, Opcode, VTList, Ops);
void *IP = nullptr;
if (FindNodeOrInsertPos(ID, SDLoc(), IP))
return true;
}
return false;
}
/// getDbgValue - Creates a SDDbgValue node.
///
/// SDNode
SDDbgValue *SelectionDAG::getDbgValue(DIVariable *Var, DIExpression *Expr,
SDNode *N, unsigned R, bool IsIndirect,
const DebugLoc &DL, unsigned O) {
assert(cast<DILocalVariable>(Var)->isValidLocationForIntrinsic(DL) &&
"Expected inlined-at fields to agree");
return new (DbgInfo->getAlloc())
SDDbgValue(DbgInfo->getAlloc(), Var, Expr, SDDbgOperand::fromNode(N, R),
{}, IsIndirect, DL, O,
/*IsVariadic=*/false);
}
/// Constant
SDDbgValue *SelectionDAG::getConstantDbgValue(DIVariable *Var,
DIExpression *Expr,
const Value *C,
const DebugLoc &DL, unsigned O) {
assert(cast<DILocalVariable>(Var)->isValidLocationForIntrinsic(DL) &&
"Expected inlined-at fields to agree");
return new (DbgInfo->getAlloc())
SDDbgValue(DbgInfo->getAlloc(), Var, Expr, SDDbgOperand::fromConst(C), {},
/*IsIndirect=*/false, DL, O,
/*IsVariadic=*/false);
}
/// FrameIndex
SDDbgValue *SelectionDAG::getFrameIndexDbgValue(DIVariable *Var,
DIExpression *Expr, unsigned FI,
bool IsIndirect,
const DebugLoc &DL,
unsigned O) {
assert(cast<DILocalVariable>(Var)->isValidLocationForIntrinsic(DL) &&
"Expected inlined-at fields to agree");
return getFrameIndexDbgValue(Var, Expr, FI, {}, IsIndirect, DL, O);
}
/// FrameIndex with dependencies
SDDbgValue *SelectionDAG::getFrameIndexDbgValue(DIVariable *Var,
DIExpression *Expr, unsigned FI,
ArrayRef<SDNode *> Dependencies,
bool IsIndirect,
const DebugLoc &DL,
unsigned O) {
assert(cast<DILocalVariable>(Var)->isValidLocationForIntrinsic(DL) &&
"Expected inlined-at fields to agree");
return new (DbgInfo->getAlloc())
SDDbgValue(DbgInfo->getAlloc(), Var, Expr, SDDbgOperand::fromFrameIdx(FI),
Dependencies, IsIndirect, DL, O,
/*IsVariadic=*/false);
}
/// VReg
SDDbgValue *SelectionDAG::getVRegDbgValue(DIVariable *Var, DIExpression *Expr,
unsigned VReg, bool IsIndirect,
const DebugLoc &DL, unsigned O) {
assert(cast<DILocalVariable>(Var)->isValidLocationForIntrinsic(DL) &&
"Expected inlined-at fields to agree");
return new (DbgInfo->getAlloc())
SDDbgValue(DbgInfo->getAlloc(), Var, Expr, SDDbgOperand::fromVReg(VReg),
{}, IsIndirect, DL, O,
/*IsVariadic=*/false);
}
SDDbgValue *SelectionDAG::getDbgValueList(DIVariable *Var, DIExpression *Expr,
ArrayRef<SDDbgOperand> Locs,
ArrayRef<SDNode *> Dependencies,
bool IsIndirect, const DebugLoc &DL,
unsigned O, bool IsVariadic) {
assert(cast<DILocalVariable>(Var)->isValidLocationForIntrinsic(DL) &&
"Expected inlined-at fields to agree");
return new (DbgInfo->getAlloc())
SDDbgValue(DbgInfo->getAlloc(), Var, Expr, Locs, Dependencies, IsIndirect,
DL, O, IsVariadic);
}
void SelectionDAG::transferDbgValues(SDValue From, SDValue To,
unsigned OffsetInBits, unsigned SizeInBits,
bool InvalidateDbg) {
SDNode *FromNode = From.getNode();
SDNode *ToNode = To.getNode();
assert(FromNode && ToNode && "Can't modify dbg values");
// PR35338
// TODO: assert(From != To && "Redundant dbg value transfer");
// TODO: assert(FromNode != ToNode && "Intranode dbg value transfer");
if (From == To || FromNode == ToNode)
return;
if (!FromNode->getHasDebugValue())
return;
SDDbgOperand FromLocOp =
SDDbgOperand::fromNode(From.getNode(), From.getResNo());
SDDbgOperand ToLocOp = SDDbgOperand::fromNode(To.getNode(), To.getResNo());
SmallVector<SDDbgValue *, 2> ClonedDVs;
for (SDDbgValue *Dbg : GetDbgValues(FromNode)) {
if (Dbg->isInvalidated())
continue;
// TODO: assert(!Dbg->isInvalidated() && "Transfer of invalid dbg value");
// Create a new location ops vector that is equal to the old vector, but
// with each instance of FromLocOp replaced with ToLocOp.
bool Changed = false;
auto NewLocOps = Dbg->copyLocationOps();
std::replace_if(
NewLocOps.begin(), NewLocOps.end(),
[&Changed, FromLocOp](const SDDbgOperand &Op) {
bool Match = Op == FromLocOp;
Changed |= Match;
return Match;
},
ToLocOp);
// Ignore this SDDbgValue if we didn't find a matching location.
if (!Changed)
continue;
DIVariable *Var = Dbg->getVariable();
auto *Expr = Dbg->getExpression();
// If a fragment is requested, update the expression.
if (SizeInBits) {
// When splitting a larger (e.g., sign-extended) value whose
// lower bits are described with an SDDbgValue, do not attempt
// to transfer the SDDbgValue to the upper bits.
if (auto FI = Expr->getFragmentInfo())
if (OffsetInBits + SizeInBits > FI->SizeInBits)
continue;
auto Fragment = DIExpression::createFragmentExpression(Expr, OffsetInBits,
SizeInBits);
if (!Fragment)
continue;
Expr = *Fragment;
}
auto AdditionalDependencies = Dbg->getAdditionalDependencies();
// Clone the SDDbgValue and move it to To.
SDDbgValue *Clone = getDbgValueList(
Var, Expr, NewLocOps, AdditionalDependencies, Dbg->isIndirect(),
Dbg->getDebugLoc(), std::max(ToNode->getIROrder(), Dbg->getOrder()),
Dbg->isVariadic());
ClonedDVs.push_back(Clone);
if (InvalidateDbg) {
// Invalidate value and indicate the SDDbgValue should not be emitted.
Dbg->setIsInvalidated();
Dbg->setIsEmitted();
}
}
for (SDDbgValue *Dbg : ClonedDVs) {
assert(is_contained(Dbg->getSDNodes(), ToNode) &&
"Transferred DbgValues should depend on the new SDNode");
AddDbgValue(Dbg, false);
}
}
void SelectionDAG::salvageDebugInfo(SDNode &N) {
if (!N.getHasDebugValue())
return;
SmallVector<SDDbgValue *, 2> ClonedDVs;
for (auto *DV : GetDbgValues(&N)) {
if (DV->isInvalidated())
continue;
switch (N.getOpcode()) {
default:
break;
case ISD::ADD: {
SDValue N0 = N.getOperand(0);
SDValue N1 = N.getOperand(1);
if (!isa<ConstantSDNode>(N0)) {
bool RHSConstant = isa<ConstantSDNode>(N1);
uint64_t Offset;
if (RHSConstant)
Offset = N.getConstantOperandVal(1);
// We are not allowed to turn indirect debug values variadic, so
// don't salvage those.
if (!RHSConstant && DV->isIndirect())
continue;
// Rewrite an ADD constant node into a DIExpression. Since we are
// performing arithmetic to compute the variable's *value* in the
// DIExpression, we need to mark the expression with a
// DW_OP_stack_value.
auto *DIExpr = DV->getExpression();
auto NewLocOps = DV->copyLocationOps();
bool Changed = false;
size_t OrigLocOpsSize = NewLocOps.size();
for (size_t i = 0; i < OrigLocOpsSize; ++i) {
// We're not given a ResNo to compare against because the whole
// node is going away. We know that any ISD::ADD only has one
// result, so we can assume any node match is using the result.
if (NewLocOps[i].getKind() != SDDbgOperand::SDNODE ||
NewLocOps[i].getSDNode() != &N)
continue;
NewLocOps[i] = SDDbgOperand::fromNode(N0.getNode(), N0.getResNo());
if (RHSConstant) {
SmallVector<uint64_t, 3> ExprOps;
DIExpression::appendOffset(ExprOps, Offset);
DIExpr = DIExpression::appendOpsToArg(DIExpr, ExprOps, i, true);
} else {
// Convert to a variadic expression (if not already).
// convertToVariadicExpression() returns a const pointer, so we use
// a temporary const variable here.
const auto *TmpDIExpr =
DIExpression::convertToVariadicExpression(DIExpr);
SmallVector<uint64_t, 3> ExprOps;
ExprOps.push_back(dwarf::DW_OP_LLVM_arg);
ExprOps.push_back(NewLocOps.size());
ExprOps.push_back(dwarf::DW_OP_plus);
SDDbgOperand RHS =
SDDbgOperand::fromNode(N1.getNode(), N1.getResNo());
NewLocOps.push_back(RHS);
DIExpr = DIExpression::appendOpsToArg(TmpDIExpr, ExprOps, i, true);
}
Changed = true;
}
(void)Changed;
assert(Changed && "Salvage target doesn't use N");
bool IsVariadic =
DV->isVariadic() || OrigLocOpsSize != NewLocOps.size();
auto AdditionalDependencies = DV->getAdditionalDependencies();
SDDbgValue *Clone = getDbgValueList(
DV->getVariable(), DIExpr, NewLocOps, AdditionalDependencies,
DV->isIndirect(), DV->getDebugLoc(), DV->getOrder(), IsVariadic);
ClonedDVs.push_back(Clone);
DV->setIsInvalidated();
DV->setIsEmitted();
LLVM_DEBUG(dbgs() << "SALVAGE: Rewriting";
N0.getNode()->dumprFull(this);
dbgs() << " into " << *DIExpr << '\n');
}
break;
}
case ISD::TRUNCATE: {
SDValue N0 = N.getOperand(0);
TypeSize FromSize = N0.getValueSizeInBits();
TypeSize ToSize = N.getValueSizeInBits(0);
DIExpression *DbgExpression = DV->getExpression();
auto ExtOps = DIExpression::getExtOps(FromSize, ToSize, false);
auto NewLocOps = DV->copyLocationOps();
bool Changed = false;
for (size_t i = 0; i < NewLocOps.size(); ++i) {
if (NewLocOps[i].getKind() != SDDbgOperand::SDNODE ||
NewLocOps[i].getSDNode() != &N)
continue;
NewLocOps[i] = SDDbgOperand::fromNode(N0.getNode(), N0.getResNo());
DbgExpression = DIExpression::appendOpsToArg(DbgExpression, ExtOps, i);
Changed = true;
}
assert(Changed && "Salvage target doesn't use N");
(void)Changed;
SDDbgValue *Clone =
getDbgValueList(DV->getVariable(), DbgExpression, NewLocOps,
DV->getAdditionalDependencies(), DV->isIndirect(),
DV->getDebugLoc(), DV->getOrder(), DV->isVariadic());
ClonedDVs.push_back(Clone);
DV->setIsInvalidated();
DV->setIsEmitted();
LLVM_DEBUG(dbgs() << "SALVAGE: Rewriting"; N0.getNode()->dumprFull(this);
dbgs() << " into " << *DbgExpression << '\n');
break;
}
}
}
for (SDDbgValue *Dbg : ClonedDVs) {
assert(!Dbg->getSDNodes().empty() &&
"Salvaged DbgValue should depend on a new SDNode");
AddDbgValue(Dbg, false);
}
}
/// Creates a SDDbgLabel node.
SDDbgLabel *SelectionDAG::getDbgLabel(DILabel *Label,
const DebugLoc &DL, unsigned O) {
assert(cast<DILabel>(Label)->isValidLocationForIntrinsic(DL) &&
"Expected inlined-at fields to agree");
return new (DbgInfo->getAlloc()) SDDbgLabel(Label, DL, O);
}
namespace {
/// RAUWUpdateListener - Helper for ReplaceAllUsesWith - When the node
/// pointed to by a use iterator is deleted, increment the use iterator
/// so that it doesn't dangle.
///
class RAUWUpdateListener : public SelectionDAG::DAGUpdateListener {
SDNode::use_iterator &UI;
SDNode::use_iterator &UE;
void NodeDeleted(SDNode *N, SDNode *E) override {
// Increment the iterator as needed.
while (UI != UE && N == *UI)
++UI;
}
public:
RAUWUpdateListener(SelectionDAG &d,
SDNode::use_iterator &ui,
SDNode::use_iterator &ue)
: SelectionDAG::DAGUpdateListener(d), UI(ui), UE(ue) {}
};
} // end anonymous namespace
/// ReplaceAllUsesWith - Modify anything using 'From' to use 'To' instead.
/// This can cause recursive merging of nodes in the DAG.
///
/// This version assumes From has a single result value.
///
void SelectionDAG::ReplaceAllUsesWith(SDValue FromN, SDValue To) {
SDNode *From = FromN.getNode();
assert(From->getNumValues() == 1 && FromN.getResNo() == 0 &&
"Cannot replace with this method!");
assert(From != To.getNode() && "Cannot replace uses of with self");
// Preserve Debug Values
transferDbgValues(FromN, To);
// Preserve extra info.
copyExtraInfo(From, To.getNode());
// Iterate over all the existing uses of From. New uses will be added
// to the beginning of the use list, which we avoid visiting.
// This specifically avoids visiting uses of From that arise while the
// replacement is happening, because any such uses would be the result
// of CSE: If an existing node looks like From after one of its operands
// is replaced by To, we don't want to replace of all its users with To
// too. See PR3018 for more info.
SDNode::use_iterator UI = From->use_begin(), UE = From->use_end();
RAUWUpdateListener Listener(*this, UI, UE);
while (UI != UE) {
SDNode *User = *UI;
// This node is about to morph, remove its old self from the CSE maps.
RemoveNodeFromCSEMaps(User);
// A user can appear in a use list multiple times, and when this
// happens the uses are usually next to each other in the list.
// To help reduce the number of CSE recomputations, process all
// the uses of this user that we can find this way.
do {
SDUse &Use = UI.getUse();
++UI;
Use.set(To);
if (To->isDivergent() != From->isDivergent())
updateDivergence(User);
} while (UI != UE && *UI == User);
// Now that we have modified User, add it back to the CSE maps. If it
// already exists there, recursively merge the results together.
AddModifiedNodeToCSEMaps(User);
}
// If we just RAUW'd the root, take note.
if (FromN == getRoot())
setRoot(To);
}
/// ReplaceAllUsesWith - Modify anything using 'From' to use 'To' instead.
/// This can cause recursive merging of nodes in the DAG.
///
/// This version assumes that for each value of From, there is a
/// corresponding value in To in the same position with the same type.
///
void SelectionDAG::ReplaceAllUsesWith(SDNode *From, SDNode *To) {
#ifndef NDEBUG
for (unsigned i = 0, e = From->getNumValues(); i != e; ++i)
assert((!From->hasAnyUseOfValue(i) ||
From->getValueType(i) == To->getValueType(i)) &&
"Cannot use this version of ReplaceAllUsesWith!");
#endif
// Handle the trivial case.
if (From == To)
return;
// Preserve Debug Info. Only do this if there's a use.
for (unsigned i = 0, e = From->getNumValues(); i != e; ++i)
if (From->hasAnyUseOfValue(i)) {
assert((i < To->getNumValues()) && "Invalid To location");
transferDbgValues(SDValue(From, i), SDValue(To, i));
}
// Preserve extra info.
copyExtraInfo(From, To);
// Iterate over just the existing users of From. See the comments in
// the ReplaceAllUsesWith above.
SDNode::use_iterator UI = From->use_begin(), UE = From->use_end();
RAUWUpdateListener Listener(*this, UI, UE);
while (UI != UE) {
SDNode *User = *UI;
// This node is about to morph, remove its old self from the CSE maps.
RemoveNodeFromCSEMaps(User);
// A user can appear in a use list multiple times, and when this
// happens the uses are usually next to each other in the list.
// To help reduce the number of CSE recomputations, process all
// the uses of this user that we can find this way.
do {
SDUse &Use = UI.getUse();
++UI;
Use.setNode(To);
if (To->isDivergent() != From->isDivergent())
updateDivergence(User);
} while (UI != UE && *UI == User);
// Now that we have modified User, add it back to the CSE maps. If it
// already exists there, recursively merge the results together.
AddModifiedNodeToCSEMaps(User);
}
// If we just RAUW'd the root, take note.
if (From == getRoot().getNode())
setRoot(SDValue(To, getRoot().getResNo()));
}
/// ReplaceAllUsesWith - Modify anything using 'From' to use 'To' instead.
/// This can cause recursive merging of nodes in the DAG.
///
/// This version can replace From with any result values. To must match the
/// number and types of values returned by From.
void SelectionDAG::ReplaceAllUsesWith(SDNode *From, const SDValue *To) {
if (From->getNumValues() == 1) // Handle the simple case efficiently.
return ReplaceAllUsesWith(SDValue(From, 0), To[0]);
for (unsigned i = 0, e = From->getNumValues(); i != e; ++i) {
// Preserve Debug Info.
transferDbgValues(SDValue(From, i), To[i]);
// Preserve extra info.
copyExtraInfo(From, To[i].getNode());
}
// Iterate over just the existing users of From. See the comments in
// the ReplaceAllUsesWith above.
SDNode::use_iterator UI = From->use_begin(), UE = From->use_end();
RAUWUpdateListener Listener(*this, UI, UE);
while (UI != UE) {
SDNode *User = *UI;
// This node is about to morph, remove its old self from the CSE maps.
RemoveNodeFromCSEMaps(User);
// A user can appear in a use list multiple times, and when this happens the
// uses are usually next to each other in the list. To help reduce the
// number of CSE and divergence recomputations, process all the uses of this
// user that we can find this way.
bool To_IsDivergent = false;
do {
SDUse &Use = UI.getUse();
const SDValue &ToOp = To[Use.getResNo()];
++UI;
Use.set(ToOp);
To_IsDivergent |= ToOp->isDivergent();
} while (UI != UE && *UI == User);
if (To_IsDivergent != From->isDivergent())
updateDivergence(User);
// Now that we have modified User, add it back to the CSE maps. If it
// already exists there, recursively merge the results together.
AddModifiedNodeToCSEMaps(User);
}
// If we just RAUW'd the root, take note.
if (From == getRoot().getNode())
setRoot(SDValue(To[getRoot().getResNo()]));
}
/// ReplaceAllUsesOfValueWith - Replace any uses of From with To, leaving
/// uses of other values produced by From.getNode() alone. The Deleted
/// vector is handled the same way as for ReplaceAllUsesWith.
void SelectionDAG::ReplaceAllUsesOfValueWith(SDValue From, SDValue To){
// Handle the really simple, really trivial case efficiently.
if (From == To) return;
// Handle the simple, trivial, case efficiently.
if (From.getNode()->getNumValues() == 1) {
ReplaceAllUsesWith(From, To);
return;
}
// Preserve Debug Info.
transferDbgValues(From, To);
copyExtraInfo(From.getNode(), To.getNode());
// Iterate over just the existing users of From. See the comments in
// the ReplaceAllUsesWith above.
SDNode::use_iterator UI = From.getNode()->use_begin(),
UE = From.getNode()->use_end();
RAUWUpdateListener Listener(*this, UI, UE);
while (UI != UE) {
SDNode *User = *UI;
bool UserRemovedFromCSEMaps = false;
// A user can appear in a use list multiple times, and when this
// happens the uses are usually next to each other in the list.
// To help reduce the number of CSE recomputations, process all
// the uses of this user that we can find this way.
do {
SDUse &Use = UI.getUse();
// Skip uses of different values from the same node.
if (Use.getResNo() != From.getResNo()) {
++UI;
continue;
}
// If this node hasn't been modified yet, it's still in the CSE maps,
// so remove its old self from the CSE maps.
if (!UserRemovedFromCSEMaps) {
RemoveNodeFromCSEMaps(User);
UserRemovedFromCSEMaps = true;
}
++UI;
Use.set(To);
if (To->isDivergent() != From->isDivergent())
updateDivergence(User);
} while (UI != UE && *UI == User);
// We are iterating over all uses of the From node, so if a use
// doesn't use the specific value, no changes are made.
if (!UserRemovedFromCSEMaps)
continue;
// Now that we have modified User, add it back to the CSE maps. If it
// already exists there, recursively merge the results together.
AddModifiedNodeToCSEMaps(User);
}
// If we just RAUW'd the root, take note.
if (From == getRoot())
setRoot(To);
}
namespace {
/// UseMemo - This class is used by SelectionDAG::ReplaceAllUsesOfValuesWith
/// to record information about a use.
struct UseMemo {
SDNode *User;
unsigned Index;
SDUse *Use;
};
/// operator< - Sort Memos by User.
bool operator<(const UseMemo &L, const UseMemo &R) {
return (intptr_t)L.User < (intptr_t)R.User;
}
/// RAUOVWUpdateListener - Helper for ReplaceAllUsesOfValuesWith - When the node
/// pointed to by a UseMemo is deleted, set the User to nullptr to indicate that
/// the node already has been taken care of recursively.
class RAUOVWUpdateListener : public SelectionDAG::DAGUpdateListener {
SmallVector<UseMemo, 4> &Uses;
void NodeDeleted(SDNode *N, SDNode *E) override {
for (UseMemo &Memo : Uses)
if (Memo.User == N)
Memo.User = nullptr;
}
public:
RAUOVWUpdateListener(SelectionDAG &d, SmallVector<UseMemo, 4> &uses)
: SelectionDAG::DAGUpdateListener(d), Uses(uses) {}
};
} // end anonymous namespace
bool SelectionDAG::calculateDivergence(SDNode *N) {
if (TLI->isSDNodeAlwaysUniform(N)) {
assert(!TLI->isSDNodeSourceOfDivergence(N, FLI, UA) &&
"Conflicting divergence information!");
return false;
}
if (TLI->isSDNodeSourceOfDivergence(N, FLI, UA))
return true;
for (const auto &Op : N->ops()) {
if (Op.Val.getValueType() != MVT::Other && Op.getNode()->isDivergent())
return true;
}
return false;
}
void SelectionDAG::updateDivergence(SDNode *N) {
SmallVector<SDNode *, 16> Worklist(1, N);
do {
N = Worklist.pop_back_val();
bool IsDivergent = calculateDivergence(N);
if (N->SDNodeBits.IsDivergent != IsDivergent) {
N->SDNodeBits.IsDivergent = IsDivergent;
llvm::append_range(Worklist, N->uses());
}
} while (!Worklist.empty());
}
void SelectionDAG::CreateTopologicalOrder(std::vector<SDNode *> &Order) {
DenseMap<SDNode *, unsigned> Degree;
Order.reserve(AllNodes.size());
for (auto &N : allnodes()) {
unsigned NOps = N.getNumOperands();
Degree[&N] = NOps;
if (0 == NOps)
Order.push_back(&N);
}
for (size_t I = 0; I != Order.size(); ++I) {
SDNode *N = Order[I];
for (auto *U : N->uses()) {
unsigned &UnsortedOps = Degree[U];
if (0 == --UnsortedOps)
Order.push_back(U);
}
}
}
#ifndef NDEBUG
void SelectionDAG::VerifyDAGDivergence() {
std::vector<SDNode *> TopoOrder;
CreateTopologicalOrder(TopoOrder);
for (auto *N : TopoOrder) {
assert(calculateDivergence(N) == N->isDivergent() &&
"Divergence bit inconsistency detected");
}
}
#endif
/// ReplaceAllUsesOfValuesWith - Replace any uses of From with To, leaving
/// uses of other values produced by From.getNode() alone. The same value
/// may appear in both the From and To list. The Deleted vector is
/// handled the same way as for ReplaceAllUsesWith.
void SelectionDAG::ReplaceAllUsesOfValuesWith(const SDValue *From,
const SDValue *To,
unsigned Num){
// Handle the simple, trivial case efficiently.
if (Num == 1)
return ReplaceAllUsesOfValueWith(*From, *To);
transferDbgValues(*From, *To);
copyExtraInfo(From->getNode(), To->getNode());
// Read up all the uses and make records of them. This helps
// processing new uses that are introduced during the
// replacement process.
SmallVector<UseMemo, 4> Uses;
for (unsigned i = 0; i != Num; ++i) {
unsigned FromResNo = From[i].getResNo();
SDNode *FromNode = From[i].getNode();
for (SDNode::use_iterator UI = FromNode->use_begin(),
E = FromNode->use_end(); UI != E; ++UI) {
SDUse &Use = UI.getUse();
if (Use.getResNo() == FromResNo) {
UseMemo Memo = { *UI, i, &Use };
Uses.push_back(Memo);
}
}
}
// Sort the uses, so that all the uses from a given User are together.
llvm::sort(Uses);
RAUOVWUpdateListener Listener(*this, Uses);
for (unsigned UseIndex = 0, UseIndexEnd = Uses.size();
UseIndex != UseIndexEnd; ) {
// We know that this user uses some value of From. If it is the right
// value, update it.
SDNode *User = Uses[UseIndex].User;
// If the node has been deleted by recursive CSE updates when updating
// another node, then just skip this entry.
if (User == nullptr) {
++UseIndex;
continue;
}
// This node is about to morph, remove its old self from the CSE maps.
RemoveNodeFromCSEMaps(User);
// The Uses array is sorted, so all the uses for a given User
// are next to each other in the list.
// To help reduce the number of CSE recomputations, process all
// the uses of this user that we can find this way.
do {
unsigned i = Uses[UseIndex].Index;
SDUse &Use = *Uses[UseIndex].Use;
++UseIndex;
Use.set(To[i]);
} while (UseIndex != UseIndexEnd && Uses[UseIndex].User == User);
// Now that we have modified User, add it back to the CSE maps. If it
// already exists there, recursively merge the results together.
AddModifiedNodeToCSEMaps(User);
}
}
/// AssignTopologicalOrder - Assign a unique node id for each node in the DAG
/// based on their topological order. It returns the maximum id and a vector
/// of the SDNodes* in assigned order by reference.
unsigned SelectionDAG::AssignTopologicalOrder() {
unsigned DAGSize = 0;
// SortedPos tracks the progress of the algorithm. Nodes before it are
// sorted, nodes after it are unsorted. When the algorithm completes
// it is at the end of the list.
allnodes_iterator SortedPos = allnodes_begin();
// Visit all the nodes. Move nodes with no operands to the front of
// the list immediately. Annotate nodes that do have operands with their
// operand count. Before we do this, the Node Id fields of the nodes
// may contain arbitrary values. After, the Node Id fields for nodes
// before SortedPos will contain the topological sort index, and the
// Node Id fields for nodes At SortedPos and after will contain the
// count of outstanding operands.
for (SDNode &N : llvm::make_early_inc_range(allnodes())) {
checkForCycles(&N, this);
unsigned Degree = N.getNumOperands();
if (Degree == 0) {
// A node with no uses, add it to the result array immediately.
N.setNodeId(DAGSize++);
allnodes_iterator Q(&N);
if (Q != SortedPos)
SortedPos = AllNodes.insert(SortedPos, AllNodes.remove(Q));
assert(SortedPos != AllNodes.end() && "Overran node list");
++SortedPos;
} else {
// Temporarily use the Node Id as scratch space for the degree count.
N.setNodeId(Degree);
}
}
// Visit all the nodes. As we iterate, move nodes into sorted order,
// such that by the time the end is reached all nodes will be sorted.
for (SDNode &Node : allnodes()) {
SDNode *N = &Node;
checkForCycles(N, this);
// N is in sorted position, so all its uses have one less operand
// that needs to be sorted.
for (SDNode *P : N->uses()) {
unsigned Degree = P->getNodeId();
assert(Degree != 0 && "Invalid node degree");
--Degree;
if (Degree == 0) {
// All of P's operands are sorted, so P may sorted now.
P->setNodeId(DAGSize++);
if (P->getIterator() != SortedPos)
SortedPos = AllNodes.insert(SortedPos, AllNodes.remove(P));
assert(SortedPos != AllNodes.end() && "Overran node list");
++SortedPos;
} else {
// Update P's outstanding operand count.
P->setNodeId(Degree);
}
}
if (Node.getIterator() == SortedPos) {
#ifndef NDEBUG
allnodes_iterator I(N);
SDNode *S = &*++I;
dbgs() << "Overran sorted position:\n";
S->dumprFull(this); dbgs() << "\n";
dbgs() << "Checking if this is due to cycles\n";
checkForCycles(this, true);
#endif
llvm_unreachable(nullptr);
}
}
assert(SortedPos == AllNodes.end() &&
"Topological sort incomplete!");
assert(AllNodes.front().getOpcode() == ISD::EntryToken &&
"First node in topological sort is not the entry token!");
assert(AllNodes.front().getNodeId() == 0 &&
"First node in topological sort has non-zero id!");
assert(AllNodes.front().getNumOperands() == 0 &&
"First node in topological sort has operands!");
assert(AllNodes.back().getNodeId() == (int)DAGSize-1 &&
"Last node in topologic sort has unexpected id!");
assert(AllNodes.back().use_empty() &&
"Last node in topologic sort has users!");
assert(DAGSize == allnodes_size() && "Node count mismatch!");
return DAGSize;
}
/// AddDbgValue - Add a dbg_value SDNode. If SD is non-null that means the
/// value is produced by SD.
void SelectionDAG::AddDbgValue(SDDbgValue *DB, bool isParameter) {
for (SDNode *SD : DB->getSDNodes()) {
if (!SD)
continue;
assert(DbgInfo->getSDDbgValues(SD).empty() || SD->getHasDebugValue());
SD->setHasDebugValue(true);
}
DbgInfo->add(DB, isParameter);
}
void SelectionDAG::AddDbgLabel(SDDbgLabel *DB) { DbgInfo->add(DB); }
SDValue SelectionDAG::makeEquivalentMemoryOrdering(SDValue OldChain,
SDValue NewMemOpChain) {
assert(isa<MemSDNode>(NewMemOpChain) && "Expected a memop node");
assert(NewMemOpChain.getValueType() == MVT::Other && "Expected a token VT");
// The new memory operation must have the same position as the old load in
// terms of memory dependency. Create a TokenFactor for the old load and new
// memory operation and update uses of the old load's output chain to use that
// TokenFactor.
if (OldChain == NewMemOpChain || OldChain.use_empty())
return NewMemOpChain;
SDValue TokenFactor = getNode(ISD::TokenFactor, SDLoc(OldChain), MVT::Other,
OldChain, NewMemOpChain);
ReplaceAllUsesOfValueWith(OldChain, TokenFactor);
UpdateNodeOperands(TokenFactor.getNode(), OldChain, NewMemOpChain);
return TokenFactor;
}
SDValue SelectionDAG::makeEquivalentMemoryOrdering(LoadSDNode *OldLoad,
SDValue NewMemOp) {
assert(isa<MemSDNode>(NewMemOp.getNode()) && "Expected a memop node");
SDValue OldChain = SDValue(OldLoad, 1);
SDValue NewMemOpChain = NewMemOp.getValue(1);
return makeEquivalentMemoryOrdering(OldChain, NewMemOpChain);
}
SDValue SelectionDAG::getSymbolFunctionGlobalAddress(SDValue Op,
Function **OutFunction) {
assert(isa<ExternalSymbolSDNode>(Op) && "Node should be an ExternalSymbol");
auto *Symbol = cast<ExternalSymbolSDNode>(Op)->getSymbol();
auto *Module = MF->getFunction().getParent();
auto *Function = Module->getFunction(Symbol);
if (OutFunction != nullptr)
*OutFunction = Function;
if (Function != nullptr) {
auto PtrTy = TLI->getPointerTy(getDataLayout(), Function->getAddressSpace());
return getGlobalAddress(Function, SDLoc(Op), PtrTy);
}
std::string ErrorStr;
raw_string_ostream ErrorFormatter(ErrorStr);
ErrorFormatter << "Undefined external symbol ";
ErrorFormatter << '"' << Symbol << '"';
report_fatal_error(Twine(ErrorFormatter.str()));
}
//===----------------------------------------------------------------------===//
// SDNode Class
//===----------------------------------------------------------------------===//
bool llvm::isNullConstant(SDValue V) {
ConstantSDNode *Const = dyn_cast<ConstantSDNode>(V);
return Const != nullptr && Const->isZero();
}
bool llvm::isNullFPConstant(SDValue V) {
ConstantFPSDNode *Const = dyn_cast<ConstantFPSDNode>(V);
return Const != nullptr && Const->isZero() && !Const->isNegative();
}
bool llvm::isAllOnesConstant(SDValue V) {
ConstantSDNode *Const = dyn_cast<ConstantSDNode>(V);
return Const != nullptr && Const->isAllOnes();
}
bool llvm::isOneConstant(SDValue V) {
ConstantSDNode *Const = dyn_cast<ConstantSDNode>(V);
return Const != nullptr && Const->isOne();
}
bool llvm::isMinSignedConstant(SDValue V) {
ConstantSDNode *Const = dyn_cast<ConstantSDNode>(V);
return Const != nullptr && Const->isMinSignedValue();
}
bool llvm::isNeutralConstant(unsigned Opcode, SDNodeFlags Flags, SDValue V,
unsigned OperandNo) {
// NOTE: The cases should match with IR's ConstantExpr::getBinOpIdentity().
// TODO: Target-specific opcodes could be added.
if (auto *Const = isConstOrConstSplat(V)) {
switch (Opcode) {
case ISD::ADD:
case ISD::OR:
case ISD::XOR:
case ISD::UMAX:
return Const->isZero();
case ISD::MUL:
return Const->isOne();
case ISD::AND:
case ISD::UMIN:
return Const->isAllOnes();
case ISD::SMAX:
return Const->isMinSignedValue();
case ISD::SMIN:
return Const->isMaxSignedValue();
case ISD::SUB:
case ISD::SHL:
case ISD::SRA:
case ISD::SRL:
return OperandNo == 1 && Const->isZero();
case ISD::UDIV:
case ISD::SDIV:
return OperandNo == 1 && Const->isOne();
}
} else if (auto *ConstFP = isConstOrConstSplatFP(V)) {
switch (Opcode) {
case ISD::FADD:
return ConstFP->isZero() &&
(Flags.hasNoSignedZeros() || ConstFP->isNegative());
case ISD::FSUB:
return OperandNo == 1 && ConstFP->isZero() &&
(Flags.hasNoSignedZeros() || !ConstFP->isNegative());
case ISD::FMUL:
return ConstFP->isExactlyValue(1.0);
case ISD::FDIV:
return OperandNo == 1 && ConstFP->isExactlyValue(1.0);
case ISD::FMINNUM:
case ISD::FMAXNUM: {
// Neutral element for fminnum is NaN, Inf or FLT_MAX, depending on FMF.
EVT VT = V.getValueType();
const fltSemantics &Semantics = SelectionDAG::EVTToAPFloatSemantics(VT);
APFloat NeutralAF = !Flags.hasNoNaNs()
? APFloat::getQNaN(Semantics)
: !Flags.hasNoInfs()
? APFloat::getInf(Semantics)
: APFloat::getLargest(Semantics);
if (Opcode == ISD::FMAXNUM)
NeutralAF.changeSign();
return ConstFP->isExactlyValue(NeutralAF);
}
}
}
return false;
}
SDValue llvm::peekThroughBitcasts(SDValue V) {
while (V.getOpcode() == ISD::BITCAST)
V = V.getOperand(0);
return V;
}
SDValue llvm::peekThroughOneUseBitcasts(SDValue V) {
while (V.getOpcode() == ISD::BITCAST && V.getOperand(0).hasOneUse())
V = V.getOperand(0);
return V;
}
SDValue llvm::peekThroughExtractSubvectors(SDValue V) {
while (V.getOpcode() == ISD::EXTRACT_SUBVECTOR)
V = V.getOperand(0);
return V;
}
SDValue llvm::peekThroughTruncates(SDValue V) {
while (V.getOpcode() == ISD::TRUNCATE)
V = V.getOperand(0);
return V;
}
bool llvm::isBitwiseNot(SDValue V, bool AllowUndefs) {
if (V.getOpcode() != ISD::XOR)
return false;
V = peekThroughBitcasts(V.getOperand(1));
unsigned NumBits = V.getScalarValueSizeInBits();
ConstantSDNode *C =
isConstOrConstSplat(V, AllowUndefs, /*AllowTruncation*/ true);
return C && (C->getAPIntValue().countr_one() >= NumBits);
}
ConstantSDNode *llvm::isConstOrConstSplat(SDValue N, bool AllowUndefs,
bool AllowTruncation) {
EVT VT = N.getValueType();
APInt DemandedElts = VT.isFixedLengthVector()
? APInt::getAllOnes(VT.getVectorMinNumElements())
: APInt(1, 1);
return isConstOrConstSplat(N, DemandedElts, AllowUndefs, AllowTruncation);
}
ConstantSDNode *llvm::isConstOrConstSplat(SDValue N, const APInt &DemandedElts,
bool AllowUndefs,
bool AllowTruncation) {
if (ConstantSDNode *CN = dyn_cast<ConstantSDNode>(N))
return CN;
// SplatVectors can truncate their operands. Ignore that case here unless
// AllowTruncation is set.
if (N->getOpcode() == ISD::SPLAT_VECTOR) {
EVT VecEltVT = N->getValueType(0).getVectorElementType();
if (auto *CN = dyn_cast<ConstantSDNode>(N->getOperand(0))) {
EVT CVT = CN->getValueType(0);
assert(CVT.bitsGE(VecEltVT) && "Illegal splat_vector element extension");
if (AllowTruncation || CVT == VecEltVT)
return CN;
}
}
if (BuildVectorSDNode *BV = dyn_cast<BuildVectorSDNode>(N)) {
BitVector UndefElements;
ConstantSDNode *CN = BV->getConstantSplatNode(DemandedElts, &UndefElements);
// BuildVectors can truncate their operands. Ignore that case here unless
// AllowTruncation is set.
// TODO: Look into whether we should allow UndefElements in non-DemandedElts
if (CN && (UndefElements.none() || AllowUndefs)) {
EVT CVT = CN->getValueType(0);
EVT NSVT = N.getValueType().getScalarType();
assert(CVT.bitsGE(NSVT) && "Illegal build vector element extension");
if (AllowTruncation || (CVT == NSVT))
return CN;
}
}
return nullptr;
}
ConstantFPSDNode *llvm::isConstOrConstSplatFP(SDValue N, bool AllowUndefs) {
EVT VT = N.getValueType();
APInt DemandedElts = VT.isFixedLengthVector()
? APInt::getAllOnes(VT.getVectorMinNumElements())
: APInt(1, 1);
return isConstOrConstSplatFP(N, DemandedElts, AllowUndefs);
}
ConstantFPSDNode *llvm::isConstOrConstSplatFP(SDValue N,
const APInt &DemandedElts,
bool AllowUndefs) {
if (ConstantFPSDNode *CN = dyn_cast<ConstantFPSDNode>(N))
return CN;
if (BuildVectorSDNode *BV = dyn_cast<BuildVectorSDNode>(N)) {
BitVector UndefElements;
ConstantFPSDNode *CN =
BV->getConstantFPSplatNode(DemandedElts, &UndefElements);
// TODO: Look into whether we should allow UndefElements in non-DemandedElts
if (CN && (UndefElements.none() || AllowUndefs))
return CN;
}
if (N.getOpcode() == ISD::SPLAT_VECTOR)
if (ConstantFPSDNode *CN = dyn_cast<ConstantFPSDNode>(N.getOperand(0)))
return CN;
return nullptr;
}
bool llvm::isNullOrNullSplat(SDValue N, bool AllowUndefs) {
// TODO: may want to use peekThroughBitcast() here.
ConstantSDNode *C =
isConstOrConstSplat(N, AllowUndefs, /*AllowTruncation=*/true);
return C && C->isZero();
}
bool llvm::isOneOrOneSplat(SDValue N, bool AllowUndefs) {
ConstantSDNode *C =
isConstOrConstSplat(N, AllowUndefs, /*AllowTruncation*/ true);
return C && C->isOne();
}
bool llvm::isAllOnesOrAllOnesSplat(SDValue N, bool AllowUndefs) {
N = peekThroughBitcasts(N);
unsigned BitWidth = N.getScalarValueSizeInBits();
ConstantSDNode *C = isConstOrConstSplat(N, AllowUndefs);
return C && C->isAllOnes() && C->getValueSizeInBits(0) == BitWidth;
}
HandleSDNode::~HandleSDNode() {
DropOperands();
}
GlobalAddressSDNode::GlobalAddressSDNode(unsigned Opc, unsigned Order,
const DebugLoc &DL,
const GlobalValue *GA, EVT VT,
int64_t o, unsigned TF)
: SDNode(Opc, Order, DL, getSDVTList(VT)), Offset(o), TargetFlags(TF) {
TheGlobal = GA;
}
AddrSpaceCastSDNode::AddrSpaceCastSDNode(unsigned Order, const DebugLoc &dl,
EVT VT, unsigned SrcAS,
unsigned DestAS)
: SDNode(ISD::ADDRSPACECAST, Order, dl, getSDVTList(VT)),
SrcAddrSpace(SrcAS), DestAddrSpace(DestAS) {}
MemSDNode::MemSDNode(unsigned Opc, unsigned Order, const DebugLoc &dl,
SDVTList VTs, EVT memvt, MachineMemOperand *mmo)
: SDNode(Opc, Order, dl, VTs), MemoryVT(memvt), MMO(mmo) {
MemSDNodeBits.IsVolatile = MMO->isVolatile();
MemSDNodeBits.IsNonTemporal = MMO->isNonTemporal();
MemSDNodeBits.IsDereferenceable = MMO->isDereferenceable();
MemSDNodeBits.IsInvariant = MMO->isInvariant();
// We check here that the size of the memory operand fits within the size of
// the MMO. This is because the MMO might indicate only a possible address
// range instead of specifying the affected memory addresses precisely.
// TODO: Make MachineMemOperands aware of scalable vectors.
assert(memvt.getStoreSize().getKnownMinValue() <= MMO->getSize() &&
"Size mismatch!");
}
/// Profile - Gather unique data for the node.
///
void SDNode::Profile(FoldingSetNodeID &ID) const {
AddNodeIDNode(ID, this);
}
namespace {
struct EVTArray {
std::vector<EVT> VTs;
EVTArray() {
VTs.reserve(MVT::VALUETYPE_SIZE);
for (unsigned i = 0; i < MVT::VALUETYPE_SIZE; ++i)
VTs.push_back(MVT((MVT::SimpleValueType)i));
}
};
} // end anonymous namespace
/// getValueTypeList - Return a pointer to the specified value type.
///
const EVT *SDNode::getValueTypeList(EVT VT) {
static std::set<EVT, EVT::compareRawBits> EVTs;
static EVTArray SimpleVTArray;
static sys::SmartMutex<true> VTMutex;
if (VT.isExtended()) {
sys::SmartScopedLock<true> Lock(VTMutex);
return &(*EVTs.insert(VT).first);
}
assert(VT.getSimpleVT() < MVT::VALUETYPE_SIZE && "Value type out of range!");
return &SimpleVTArray.VTs[VT.getSimpleVT().SimpleTy];
}
/// hasNUsesOfValue - Return true if there are exactly NUSES uses of the
/// indicated value. This method ignores uses of other values defined by this
/// operation.
bool SDNode::hasNUsesOfValue(unsigned NUses, unsigned Value) const {
assert(Value < getNumValues() && "Bad value!");
// TODO: Only iterate over uses of a given value of the node
for (SDNode::use_iterator UI = use_begin(), E = use_end(); UI != E; ++UI) {
if (UI.getUse().getResNo() == Value) {
if (NUses == 0)
return false;
--NUses;
}
}
// Found exactly the right number of uses?
return NUses == 0;
}
/// hasAnyUseOfValue - Return true if there are any use of the indicated
/// value. This method ignores uses of other values defined by this operation.
bool SDNode::hasAnyUseOfValue(unsigned Value) const {
assert(Value < getNumValues() && "Bad value!");
for (SDNode::use_iterator UI = use_begin(), E = use_end(); UI != E; ++UI)
if (UI.getUse().getResNo() == Value)
return true;
return false;
}
/// isOnlyUserOf - Return true if this node is the only use of N.
bool SDNode::isOnlyUserOf(const SDNode *N) const {
bool Seen = false;
for (const SDNode *User : N->uses()) {
if (User == this)
Seen = true;
else
return false;
}
return Seen;
}
/// Return true if the only users of N are contained in Nodes.
bool SDNode::areOnlyUsersOf(ArrayRef<const SDNode *> Nodes, const SDNode *N) {
bool Seen = false;
for (const SDNode *User : N->uses()) {
if (llvm::is_contained(Nodes, User))
Seen = true;
else
return false;
}
return Seen;
}
/// isOperand - Return true if this node is an operand of N.
bool SDValue::isOperandOf(const SDNode *N) const {
return is_contained(N->op_values(), *this);
}
bool SDNode::isOperandOf(const SDNode *N) const {
return any_of(N->op_values(),
[this](SDValue Op) { return this == Op.getNode(); });
}
/// reachesChainWithoutSideEffects - Return true if this operand (which must
/// be a chain) reaches the specified operand without crossing any
/// side-effecting instructions on any chain path. In practice, this looks
/// through token factors and non-volatile loads. In order to remain efficient,
/// this only looks a couple of nodes in, it does not do an exhaustive search.
///
/// Note that we only need to examine chains when we're searching for
/// side-effects; SelectionDAG requires that all side-effects are represented
/// by chains, even if another operand would force a specific ordering. This
/// constraint is necessary to allow transformations like splitting loads.
bool SDValue::reachesChainWithoutSideEffects(SDValue Dest,
unsigned Depth) const {
if (*this == Dest) return true;
// Don't search too deeply, we just want to be able to see through
// TokenFactor's etc.
if (Depth == 0) return false;
// If this is a token factor, all inputs to the TF happen in parallel.
if (getOpcode() == ISD::TokenFactor) {
// First, try a shallow search.
if (is_contained((*this)->ops(), Dest)) {
// We found the chain we want as an operand of this TokenFactor.
// Essentially, we reach the chain without side-effects if we could
// serialize the TokenFactor into a simple chain of operations with
// Dest as the last operation. This is automatically true if the
// chain has one use: there are no other ordering constraints.
// If the chain has more than one use, we give up: some other
// use of Dest might force a side-effect between Dest and the current
// node.
if (Dest.hasOneUse())
return true;
}
// Next, try a deep search: check whether every operand of the TokenFactor
// reaches Dest.
return llvm::all_of((*this)->ops(), [=](SDValue Op) {
return Op.reachesChainWithoutSideEffects(Dest, Depth - 1);
});
}
// Loads don't have side effects, look through them.
if (LoadSDNode *Ld = dyn_cast<LoadSDNode>(*this)) {
if (Ld->isUnordered())
return Ld->getChain().reachesChainWithoutSideEffects(Dest, Depth-1);
}
return false;
}
bool SDNode::hasPredecessor(const SDNode *N) const {
SmallPtrSet<const SDNode *, 32> Visited;
SmallVector<const SDNode *, 16> Worklist;
Worklist.push_back(this);
return hasPredecessorHelper(N, Visited, Worklist);
}
void SDNode::intersectFlagsWith(const SDNodeFlags Flags) {
this->Flags.intersectWith(Flags);
}
SDValue
SelectionDAG::matchBinOpReduction(SDNode *Extract, ISD::NodeType &BinOp,
ArrayRef<ISD::NodeType> CandidateBinOps,
bool AllowPartials) {
// The pattern must end in an extract from index 0.
if (Extract->getOpcode() != ISD::EXTRACT_VECTOR_ELT ||
!isNullConstant(Extract->getOperand(1)))
return SDValue();
// Match against one of the candidate binary ops.
SDValue Op = Extract->getOperand(0);
if (llvm::none_of(CandidateBinOps, [Op](ISD::NodeType BinOp) {
return Op.getOpcode() == unsigned(BinOp);
}))
return SDValue();
// Floating-point reductions may require relaxed constraints on the final step
// of the reduction because they may reorder intermediate operations.
unsigned CandidateBinOp = Op.getOpcode();
if (Op.getValueType().isFloatingPoint()) {
SDNodeFlags Flags = Op->getFlags();
switch (CandidateBinOp) {
case ISD::FADD:
if (!Flags.hasNoSignedZeros() || !Flags.hasAllowReassociation())
return SDValue();
break;
default:
llvm_unreachable("Unhandled FP opcode for binop reduction");
}
}
// Matching failed - attempt to see if we did enough stages that a partial
// reduction from a subvector is possible.
auto PartialReduction = [&](SDValue Op, unsigned NumSubElts) {
if (!AllowPartials || !Op)
return SDValue();
EVT OpVT = Op.getValueType();
EVT OpSVT = OpVT.getScalarType();
EVT SubVT = EVT::getVectorVT(*getContext(), OpSVT, NumSubElts);
if (!TLI->isExtractSubvectorCheap(SubVT, OpVT, 0))
return SDValue();
BinOp = (ISD::NodeType)CandidateBinOp;
return getNode(ISD::EXTRACT_SUBVECTOR, SDLoc(Op), SubVT, Op,
getVectorIdxConstant(0, SDLoc(Op)));
};
// At each stage, we're looking for something that looks like:
// %s = shufflevector <8 x i32> %op, <8 x i32> undef,
// <8 x i32> <i32 2, i32 3, i32 undef, i32 undef,
// i32 undef, i32 undef, i32 undef, i32 undef>
// %a = binop <8 x i32> %op, %s
// Where the mask changes according to the stage. E.g. for a 3-stage pyramid,
// we expect something like:
// <4,5,6,7,u,u,u,u>
// <2,3,u,u,u,u,u,u>
// <1,u,u,u,u,u,u,u>
// While a partial reduction match would be:
// <2,3,u,u,u,u,u,u>
// <1,u,u,u,u,u,u,u>
unsigned Stages = Log2_32(Op.getValueType().getVectorNumElements());
SDValue PrevOp;
for (unsigned i = 0; i < Stages; ++i) {
unsigned MaskEnd = (1 << i);
if (Op.getOpcode() != CandidateBinOp)
return PartialReduction(PrevOp, MaskEnd);
SDValue Op0 = Op.getOperand(0);
SDValue Op1 = Op.getOperand(1);
ShuffleVectorSDNode *Shuffle = dyn_cast<ShuffleVectorSDNode>(Op0);
if (Shuffle) {
Op = Op1;
} else {
Shuffle = dyn_cast<ShuffleVectorSDNode>(Op1);
Op = Op0;
}
// The first operand of the shuffle should be the same as the other operand
// of the binop.
if (!Shuffle || Shuffle->getOperand(0) != Op)
return PartialReduction(PrevOp, MaskEnd);
// Verify the shuffle has the expected (at this stage of the pyramid) mask.
for (int Index = 0; Index < (int)MaskEnd; ++Index)
if (Shuffle->getMaskElt(Index) != (int)(MaskEnd + Index))
return PartialReduction(PrevOp, MaskEnd);
PrevOp = Op;
}
// Handle subvector reductions, which tend to appear after the shuffle
// reduction stages.
while (Op.getOpcode() == CandidateBinOp) {
unsigned NumElts = Op.getValueType().getVectorNumElements();
SDValue Op0 = Op.getOperand(0);
SDValue Op1 = Op.getOperand(1);
if (Op0.getOpcode() != ISD::EXTRACT_SUBVECTOR ||
Op1.getOpcode() != ISD::EXTRACT_SUBVECTOR ||
Op0.getOperand(0) != Op1.getOperand(0))
break;
SDValue Src = Op0.getOperand(0);
unsigned NumSrcElts = Src.getValueType().getVectorNumElements();
if (NumSrcElts != (2 * NumElts))
break;
if (!(Op0.getConstantOperandAPInt(1) == 0 &&
Op1.getConstantOperandAPInt(1) == NumElts) &&
!(Op1.getConstantOperandAPInt(1) == 0 &&
Op0.getConstantOperandAPInt(1) == NumElts))
break;
Op = Src;
}
BinOp = (ISD::NodeType)CandidateBinOp;
return Op;
}
SDValue SelectionDAG::UnrollVectorOp(SDNode *N, unsigned ResNE) {
EVT VT = N->getValueType(0);
EVT EltVT = VT.getVectorElementType();
unsigned NE = VT.getVectorNumElements();
SDLoc dl(N);
// If ResNE is 0, fully unroll the vector op.
if (ResNE == 0)
ResNE = NE;
else if (NE > ResNE)
NE = ResNE;
if (N->getNumValues() == 2) {
SmallVector<SDValue, 8> Scalars0, Scalars1;
SmallVector<SDValue, 4> Operands(N->getNumOperands());
EVT VT1 = N->getValueType(1);
EVT EltVT1 = VT1.getVectorElementType();
unsigned i;
for (i = 0; i != NE; ++i) {
for (unsigned j = 0, e = N->getNumOperands(); j != e; ++j) {
SDValue Operand = N->getOperand(j);
EVT OperandVT = Operand.getValueType();
// A vector operand; extract a single element.
EVT OperandEltVT = OperandVT.getVectorElementType();
Operands[j] = getNode(ISD::EXTRACT_VECTOR_ELT, dl, OperandEltVT,
Operand, getVectorIdxConstant(i, dl));
}
SDValue EltOp = getNode(N->getOpcode(), dl, {EltVT, EltVT1}, Operands);
Scalars0.push_back(EltOp);
Scalars1.push_back(EltOp.getValue(1));
}
SDValue Vec0 = getBuildVector(VT, dl, Scalars0);
SDValue Vec1 = getBuildVector(VT1, dl, Scalars1);
return getMergeValues({Vec0, Vec1}, dl);
}
assert(N->getNumValues() == 1 &&
"Can't unroll a vector with multiple results!");
SmallVector<SDValue, 8> Scalars;
SmallVector<SDValue, 4> Operands(N->getNumOperands());
unsigned i;
for (i= 0; i != NE; ++i) {
for (unsigned j = 0, e = N->getNumOperands(); j != e; ++j) {
SDValue Operand = N->getOperand(j);
EVT OperandVT = Operand.getValueType();
if (OperandVT.isVector()) {
// A vector operand; extract a single element.
EVT OperandEltVT = OperandVT.getVectorElementType();
Operands[j] = getNode(ISD::EXTRACT_VECTOR_ELT, dl, OperandEltVT,
Operand, getVectorIdxConstant(i, dl));
} else {
// A scalar operand; just use it as is.
Operands[j] = Operand;
}
}
switch (N->getOpcode()) {
default: {
Scalars.push_back(getNode(N->getOpcode(), dl, EltVT, Operands,
N->getFlags()));
break;
}
case ISD::VSELECT:
Scalars.push_back(getNode(ISD::SELECT, dl, EltVT, Operands));
break;
case ISD::SHL:
case ISD::SRA:
case ISD::SRL:
case ISD::ROTL:
case ISD::ROTR:
Scalars.push_back(getNode(N->getOpcode(), dl, EltVT, Operands[0],
getShiftAmountOperand(Operands[0].getValueType(),
Operands[1])));
break;
case ISD::SIGN_EXTEND_INREG: {
EVT ExtVT = cast<VTSDNode>(Operands[1])->getVT().getVectorElementType();
Scalars.push_back(getNode(N->getOpcode(), dl, EltVT,
Operands[0],
getValueType(ExtVT)));
}
}
}
for (; i < ResNE; ++i)
Scalars.push_back(getUNDEF(EltVT));
EVT VecVT = EVT::getVectorVT(*getContext(), EltVT, ResNE);
return getBuildVector(VecVT, dl, Scalars);
}
std::pair<SDValue, SDValue> SelectionDAG::UnrollVectorOverflowOp(
SDNode *N, unsigned ResNE) {
unsigned Opcode = N->getOpcode();
assert((Opcode == ISD::UADDO || Opcode == ISD::SADDO ||
Opcode == ISD::USUBO || Opcode == ISD::SSUBO ||
Opcode == ISD::UMULO || Opcode == ISD::SMULO) &&
"Expected an overflow opcode");
EVT ResVT = N->getValueType(0);
EVT OvVT = N->getValueType(1);
EVT ResEltVT = ResVT.getVectorElementType();
EVT OvEltVT = OvVT.getVectorElementType();
SDLoc dl(N);
// If ResNE is 0, fully unroll the vector op.
unsigned NE = ResVT.getVectorNumElements();
if (ResNE == 0)
ResNE = NE;
else if (NE > ResNE)
NE = ResNE;
SmallVector<SDValue, 8> LHSScalars;
SmallVector<SDValue, 8> RHSScalars;
ExtractVectorElements(N->getOperand(0), LHSScalars, 0, NE);
ExtractVectorElements(N->getOperand(1), RHSScalars, 0, NE);
EVT SVT = TLI->getSetCCResultType(getDataLayout(), *getContext(), ResEltVT);
SDVTList VTs = getVTList(ResEltVT, SVT);
SmallVector<SDValue, 8> ResScalars;
SmallVector<SDValue, 8> OvScalars;
for (unsigned i = 0; i < NE; ++i) {
SDValue Res = getNode(Opcode, dl, VTs, LHSScalars[i], RHSScalars[i]);
SDValue Ov =
getSelect(dl, OvEltVT, Res.getValue(1),
getBoolConstant(true, dl, OvEltVT, ResVT),
getConstant(0, dl, OvEltVT));
ResScalars.push_back(Res);
OvScalars.push_back(Ov);
}
ResScalars.append(ResNE - NE, getUNDEF(ResEltVT));
OvScalars.append(ResNE - NE, getUNDEF(OvEltVT));
EVT NewResVT = EVT::getVectorVT(*getContext(), ResEltVT, ResNE);
EVT NewOvVT = EVT::getVectorVT(*getContext(), OvEltVT, ResNE);
return std::make_pair(getBuildVector(NewResVT, dl, ResScalars),
getBuildVector(NewOvVT, dl, OvScalars));
}
bool SelectionDAG::areNonVolatileConsecutiveLoads(LoadSDNode *LD,
LoadSDNode *Base,
unsigned Bytes,
int Dist) const {
if (LD->isVolatile() || Base->isVolatile())
return false;
// TODO: probably too restrictive for atomics, revisit
if (!LD->isSimple())
return false;
if (LD->isIndexed() || Base->isIndexed())
return false;
if (LD->getChain() != Base->getChain())
return false;
EVT VT = LD->getMemoryVT();
if (VT.getSizeInBits() / 8 != Bytes)
return false;
auto BaseLocDecomp = BaseIndexOffset::match(Base, *this);
auto LocDecomp = BaseIndexOffset::match(LD, *this);
int64_t Offset = 0;
if (BaseLocDecomp.equalBaseIndex(LocDecomp, *this, Offset))
return (Dist * (int64_t)Bytes == Offset);
return false;
}
/// InferPtrAlignment - Infer alignment of a load / store address. Return
/// std::nullopt if it cannot be inferred.
MaybeAlign SelectionDAG::InferPtrAlign(SDValue Ptr) const {
// If this is a GlobalAddress + cst, return the alignment.
const GlobalValue *GV = nullptr;
int64_t GVOffset = 0;
if (TLI->isGAPlusOffset(Ptr.getNode(), GV, GVOffset)) {
unsigned PtrWidth = getDataLayout().getPointerTypeSizeInBits(GV->getType());
KnownBits Known(PtrWidth);
llvm::computeKnownBits(GV, Known, getDataLayout());
unsigned AlignBits = Known.countMinTrailingZeros();
if (AlignBits)
return commonAlignment(Align(1ull << std::min(31U, AlignBits)), GVOffset);
}
// If this is a direct reference to a stack slot, use information about the
// stack slot's alignment.
int FrameIdx = INT_MIN;
int64_t FrameOffset = 0;
if (FrameIndexSDNode *FI = dyn_cast<FrameIndexSDNode>(Ptr)) {
FrameIdx = FI->getIndex();
} else if (isBaseWithConstantOffset(Ptr) &&
isa<FrameIndexSDNode>(Ptr.getOperand(0))) {
// Handle FI+Cst
FrameIdx = cast<FrameIndexSDNode>(Ptr.getOperand(0))->getIndex();
FrameOffset = Ptr.getConstantOperandVal(1);
}
if (FrameIdx != INT_MIN) {
const MachineFrameInfo &MFI = getMachineFunction().getFrameInfo();
return commonAlignment(MFI.getObjectAlign(FrameIdx), FrameOffset);
}
return std::nullopt;
}
/// Split the scalar node with EXTRACT_ELEMENT using the provided
/// VTs and return the low/high part.
std::pair<SDValue, SDValue> SelectionDAG::SplitScalar(const SDValue &N,
const SDLoc &DL,
const EVT &LoVT,
const EVT &HiVT) {
assert(!LoVT.isVector() && !HiVT.isVector() && !N.getValueType().isVector() &&
"Split node must be a scalar type");
SDValue Lo =
getNode(ISD::EXTRACT_ELEMENT, DL, LoVT, N, getIntPtrConstant(0, DL));
SDValue Hi =
getNode(ISD::EXTRACT_ELEMENT, DL, HiVT, N, getIntPtrConstant(1, DL));
return std::make_pair(Lo, Hi);
}
/// GetSplitDestVTs - Compute the VTs needed for the low/hi parts of a type
/// which is split (or expanded) into two not necessarily identical pieces.
std::pair<EVT, EVT> SelectionDAG::GetSplitDestVTs(const EVT &VT) const {
// Currently all types are split in half.
EVT LoVT, HiVT;
if (!VT.isVector())
LoVT = HiVT = TLI->getTypeToTransformTo(*getContext(), VT);
else
LoVT = HiVT = VT.getHalfNumVectorElementsVT(*getContext());
return std::make_pair(LoVT, HiVT);
}
/// GetDependentSplitDestVTs - Compute the VTs needed for the low/hi parts of a
/// type, dependent on an enveloping VT that has been split into two identical
/// pieces. Sets the HiIsEmpty flag when hi type has zero storage size.
std::pair<EVT, EVT>
SelectionDAG::GetDependentSplitDestVTs(const EVT &VT, const EVT &EnvVT,
bool *HiIsEmpty) const {
EVT EltTp = VT.getVectorElementType();
// Examples:
// custom VL=8 with enveloping VL=8/8 yields 8/0 (hi empty)
// custom VL=9 with enveloping VL=8/8 yields 8/1
// custom VL=10 with enveloping VL=8/8 yields 8/2
// etc.
ElementCount VTNumElts = VT.getVectorElementCount();
ElementCount EnvNumElts = EnvVT.getVectorElementCount();
assert(VTNumElts.isScalable() == EnvNumElts.isScalable() &&
"Mixing fixed width and scalable vectors when enveloping a type");
EVT LoVT, HiVT;
if (VTNumElts.getKnownMinValue() > EnvNumElts.getKnownMinValue()) {
LoVT = EVT::getVectorVT(*getContext(), EltTp, EnvNumElts);
HiVT = EVT::getVectorVT(*getContext(), EltTp, VTNumElts - EnvNumElts);
*HiIsEmpty = false;
} else {
// Flag that hi type has zero storage size, but return split envelop type
// (this would be easier if vector types with zero elements were allowed).
LoVT = EVT::getVectorVT(*getContext(), EltTp, VTNumElts);
HiVT = EVT::getVectorVT(*getContext(), EltTp, EnvNumElts);
*HiIsEmpty = true;
}
return std::make_pair(LoVT, HiVT);
}
/// SplitVector - Split the vector with EXTRACT_SUBVECTOR and return the
/// low/high part.
std::pair<SDValue, SDValue>
SelectionDAG::SplitVector(const SDValue &N, const SDLoc &DL, const EVT &LoVT,
const EVT &HiVT) {
assert(LoVT.isScalableVector() == HiVT.isScalableVector() &&
LoVT.isScalableVector() == N.getValueType().isScalableVector() &&
"Splitting vector with an invalid mixture of fixed and scalable "
"vector types");
assert(LoVT.getVectorMinNumElements() + HiVT.getVectorMinNumElements() <=
N.getValueType().getVectorMinNumElements() &&
"More vector elements requested than available!");
SDValue Lo, Hi;
Lo =
getNode(ISD::EXTRACT_SUBVECTOR, DL, LoVT, N, getVectorIdxConstant(0, DL));
// For scalable vectors it is safe to use LoVT.getVectorMinNumElements()
// (rather than having to use ElementCount), because EXTRACT_SUBVECTOR scales
// IDX with the runtime scaling factor of the result vector type. For
// fixed-width result vectors, that runtime scaling factor is 1.
Hi = getNode(ISD::EXTRACT_SUBVECTOR, DL, HiVT, N,
getVectorIdxConstant(LoVT.getVectorMinNumElements(), DL));
return std::make_pair(Lo, Hi);
}
std::pair<SDValue, SDValue> SelectionDAG::SplitEVL(SDValue N, EVT VecVT,
const SDLoc &DL) {
// Split the vector length parameter.
// %evl -> umin(%evl, %halfnumelts) and usubsat(%evl - %halfnumelts).
EVT VT = N.getValueType();
assert(VecVT.getVectorElementCount().isKnownEven() &&
"Expecting the mask to be an evenly-sized vector");
unsigned HalfMinNumElts = VecVT.getVectorMinNumElements() / 2;
SDValue HalfNumElts =
VecVT.isFixedLengthVector()
? getConstant(HalfMinNumElts, DL, VT)
: getVScale(DL, VT, APInt(VT.getScalarSizeInBits(), HalfMinNumElts));
SDValue Lo = getNode(ISD::UMIN, DL, VT, N, HalfNumElts);
SDValue Hi = getNode(ISD::USUBSAT, DL, VT, N, HalfNumElts);
return std::make_pair(Lo, Hi);
}
/// Widen the vector up to the next power of two using INSERT_SUBVECTOR.
SDValue SelectionDAG::WidenVector(const SDValue &N, const SDLoc &DL) {
EVT VT = N.getValueType();
EVT WideVT = EVT::getVectorVT(*getContext(), VT.getVectorElementType(),
NextPowerOf2(VT.getVectorNumElements()));
return getNode(ISD::INSERT_SUBVECTOR, DL, WideVT, getUNDEF(WideVT), N,
getVectorIdxConstant(0, DL));
}
void SelectionDAG::ExtractVectorElements(SDValue Op,
SmallVectorImpl<SDValue> &Args,
unsigned Start, unsigned Count,
EVT EltVT) {
EVT VT = Op.getValueType();
if (Count == 0)
Count = VT.getVectorNumElements();
if (EltVT == EVT())
EltVT = VT.getVectorElementType();
SDLoc SL(Op);
for (unsigned i = Start, e = Start + Count; i != e; ++i) {
Args.push_back(getNode(ISD::EXTRACT_VECTOR_ELT, SL, EltVT, Op,
getVectorIdxConstant(i, SL)));
}
}
// getAddressSpace - Return the address space this GlobalAddress belongs to.
unsigned GlobalAddressSDNode::getAddressSpace() const {
return getGlobal()->getType()->getAddressSpace();
}
Type *ConstantPoolSDNode::getType() const {
if (isMachineConstantPoolEntry())
return Val.MachineCPVal->getType();
return Val.ConstVal->getType();
}
bool BuildVectorSDNode::isConstantSplat(APInt &SplatValue, APInt &SplatUndef,
unsigned &SplatBitSize,
bool &HasAnyUndefs,
unsigned MinSplatBits,
bool IsBigEndian) const {
EVT VT = getValueType(0);
assert(VT.isVector() && "Expected a vector type");
unsigned VecWidth = VT.getSizeInBits();
if (MinSplatBits > VecWidth)
return false;
// FIXME: The widths are based on this node's type, but build vectors can
// truncate their operands.
SplatValue = APInt(VecWidth, 0);
SplatUndef = APInt(VecWidth, 0);
// Get the bits. Bits with undefined values (when the corresponding element
// of the vector is an ISD::UNDEF value) are set in SplatUndef and cleared
// in SplatValue. If any of the values are not constant, give up and return
// false.
unsigned int NumOps = getNumOperands();
assert(NumOps > 0 && "isConstantSplat has 0-size build vector");
unsigned EltWidth = VT.getScalarSizeInBits();
for (unsigned j = 0; j < NumOps; ++j) {
unsigned i = IsBigEndian ? NumOps - 1 - j : j;
SDValue OpVal = getOperand(i);
unsigned BitPos = j * EltWidth;
if (OpVal.isUndef())
SplatUndef.setBits(BitPos, BitPos + EltWidth);
else if (auto *CN = dyn_cast<ConstantSDNode>(OpVal))
SplatValue.insertBits(CN->getAPIntValue().zextOrTrunc(EltWidth), BitPos);
else if (auto *CN = dyn_cast<ConstantFPSDNode>(OpVal))
SplatValue.insertBits(CN->getValueAPF().bitcastToAPInt(), BitPos);
else
return false;
}
// The build_vector is all constants or undefs. Find the smallest element
// size that splats the vector.
HasAnyUndefs = (SplatUndef != 0);
// FIXME: This does not work for vectors with elements less than 8 bits.
while (VecWidth > 8) {
// If we can't split in half, stop here.
if (VecWidth & 1)
break;
unsigned HalfSize = VecWidth / 2;
APInt HighValue = SplatValue.extractBits(HalfSize, HalfSize);
APInt LowValue = SplatValue.extractBits(HalfSize, 0);
APInt HighUndef = SplatUndef.extractBits(HalfSize, HalfSize);
APInt LowUndef = SplatUndef.extractBits(HalfSize, 0);
// If the two halves do not match (ignoring undef bits), stop here.
if ((HighValue & ~LowUndef) != (LowValue & ~HighUndef) ||
MinSplatBits > HalfSize)
break;
SplatValue = HighValue | LowValue;
SplatUndef = HighUndef & LowUndef;
VecWidth = HalfSize;
}
// FIXME: The loop above only tries to split in halves. But if the input
// vector for example is <3 x i16> it wouldn't be able to detect a
// SplatBitSize of 16. No idea if that is a design flaw currently limiting
// optimizations. I guess that back in the days when this helper was created
// vectors normally was power-of-2 sized.
SplatBitSize = VecWidth;
return true;
}
SDValue BuildVectorSDNode::getSplatValue(const APInt &DemandedElts,
BitVector *UndefElements) const {
unsigned NumOps = getNumOperands();
if (UndefElements) {
UndefElements->clear();
UndefElements->resize(NumOps);
}
assert(NumOps == DemandedElts.getBitWidth() && "Unexpected vector size");
if (!DemandedElts)
return SDValue();
SDValue Splatted;
for (unsigned i = 0; i != NumOps; ++i) {
if (!DemandedElts[i])
continue;
SDValue Op = getOperand(i);
if (Op.isUndef()) {
if (UndefElements)
(*UndefElements)[i] = true;
} else if (!Splatted) {
Splatted = Op;
} else if (Splatted != Op) {
return SDValue();
}
}
if (!Splatted) {
unsigned FirstDemandedIdx = DemandedElts.countr_zero();
assert(getOperand(FirstDemandedIdx).isUndef() &&
"Can only have a splat without a constant for all undefs.");
return getOperand(FirstDemandedIdx);
}
return Splatted;
}
SDValue BuildVectorSDNode::getSplatValue(BitVector *UndefElements) const {
APInt DemandedElts = APInt::getAllOnes(getNumOperands());
return getSplatValue(DemandedElts, UndefElements);
}
bool BuildVectorSDNode::getRepeatedSequence(const APInt &DemandedElts,
SmallVectorImpl<SDValue> &Sequence,
BitVector *UndefElements) const {
unsigned NumOps = getNumOperands();
Sequence.clear();
if (UndefElements) {
UndefElements->clear();
UndefElements->resize(NumOps);
}
assert(NumOps == DemandedElts.getBitWidth() && "Unexpected vector size");
if (!DemandedElts || NumOps < 2 || !isPowerOf2_32(NumOps))
return false;
// Set the undefs even if we don't find a sequence (like getSplatValue).
if (UndefElements)
for (unsigned I = 0; I != NumOps; ++I)
if (DemandedElts[I] && getOperand(I).isUndef())
(*UndefElements)[I] = true;
// Iteratively widen the sequence length looking for repetitions.
for (unsigned SeqLen = 1; SeqLen < NumOps; SeqLen *= 2) {
Sequence.append(SeqLen, SDValue());
for (unsigned I = 0; I != NumOps; ++I) {
if (!DemandedElts[I])
continue;
SDValue &SeqOp = Sequence[I % SeqLen];
SDValue Op = getOperand(I);
if (Op.isUndef()) {
if (!SeqOp)
SeqOp = Op;
continue;
}
if (SeqOp && !SeqOp.isUndef() && SeqOp != Op) {
Sequence.clear();
break;
}
SeqOp = Op;
}
if (!Sequence.empty())
return true;
}
assert(Sequence.empty() && "Failed to empty non-repeating sequence pattern");
return false;
}
bool BuildVectorSDNode::getRepeatedSequence(SmallVectorImpl<SDValue> &Sequence,
BitVector *UndefElements) const {
APInt DemandedElts = APInt::getAllOnes(getNumOperands());
return getRepeatedSequence(DemandedElts, Sequence, UndefElements);
}
ConstantSDNode *
BuildVectorSDNode::getConstantSplatNode(const APInt &DemandedElts,
BitVector *UndefElements) const {
return dyn_cast_or_null<ConstantSDNode>(
getSplatValue(DemandedElts, UndefElements));
}
ConstantSDNode *
BuildVectorSDNode::getConstantSplatNode(BitVector *UndefElements) const {
return dyn_cast_or_null<ConstantSDNode>(getSplatValue(UndefElements));
}
ConstantFPSDNode *
BuildVectorSDNode::getConstantFPSplatNode(const APInt &DemandedElts,
BitVector *UndefElements) const {
return dyn_cast_or_null<ConstantFPSDNode>(
getSplatValue(DemandedElts, UndefElements));
}
ConstantFPSDNode *
BuildVectorSDNode::getConstantFPSplatNode(BitVector *UndefElements) const {
return dyn_cast_or_null<ConstantFPSDNode>(getSplatValue(UndefElements));
}
int32_t
BuildVectorSDNode::getConstantFPSplatPow2ToLog2Int(BitVector *UndefElements,
uint32_t BitWidth) const {
if (ConstantFPSDNode *CN =
dyn_cast_or_null<ConstantFPSDNode>(getSplatValue(UndefElements))) {
bool IsExact;
APSInt IntVal(BitWidth);
const APFloat &APF = CN->getValueAPF();
if (APF.convertToInteger(IntVal, APFloat::rmTowardZero, &IsExact) !=
APFloat::opOK ||
!IsExact)
return -1;
return IntVal.exactLogBase2();
}
return -1;
}
bool BuildVectorSDNode::getConstantRawBits(
bool IsLittleEndian, unsigned DstEltSizeInBits,
SmallVectorImpl<APInt> &RawBitElements, BitVector &UndefElements) const {
// Early-out if this contains anything but Undef/Constant/ConstantFP.
if (!isConstant())
return false;
unsigned NumSrcOps = getNumOperands();
unsigned SrcEltSizeInBits = getValueType(0).getScalarSizeInBits();
assert(((NumSrcOps * SrcEltSizeInBits) % DstEltSizeInBits) == 0 &&
"Invalid bitcast scale");
// Extract raw src bits.
SmallVector<APInt> SrcBitElements(NumSrcOps,
APInt::getZero(SrcEltSizeInBits));
BitVector SrcUndeElements(NumSrcOps, false);
for (unsigned I = 0; I != NumSrcOps; ++I) {
SDValue Op = getOperand(I);
if (Op.isUndef()) {
SrcUndeElements.set(I);
continue;
}
auto *CInt = dyn_cast<ConstantSDNode>(Op);
auto *CFP = dyn_cast<ConstantFPSDNode>(Op);
assert((CInt || CFP) && "Unknown constant");
SrcBitElements[I] = CInt ? CInt->getAPIntValue().trunc(SrcEltSizeInBits)
: CFP->getValueAPF().bitcastToAPInt();
}
// Recast to dst width.
recastRawBits(IsLittleEndian, DstEltSizeInBits, RawBitElements,
SrcBitElements, UndefElements, SrcUndeElements);
return true;
}
void BuildVectorSDNode::recastRawBits(bool IsLittleEndian,
unsigned DstEltSizeInBits,
SmallVectorImpl<APInt> &DstBitElements,
ArrayRef<APInt> SrcBitElements,
BitVector &DstUndefElements,
const BitVector &SrcUndefElements) {
unsigned NumSrcOps = SrcBitElements.size();
unsigned SrcEltSizeInBits = SrcBitElements[0].getBitWidth();
assert(((NumSrcOps * SrcEltSizeInBits) % DstEltSizeInBits) == 0 &&
"Invalid bitcast scale");
assert(NumSrcOps == SrcUndefElements.size() &&
"Vector size mismatch");
unsigned NumDstOps = (NumSrcOps * SrcEltSizeInBits) / DstEltSizeInBits;
DstUndefElements.clear();
DstUndefElements.resize(NumDstOps, false);
DstBitElements.assign(NumDstOps, APInt::getZero(DstEltSizeInBits));
// Concatenate src elements constant bits together into dst element.
if (SrcEltSizeInBits <= DstEltSizeInBits) {
unsigned Scale = DstEltSizeInBits / SrcEltSizeInBits;
for (unsigned I = 0; I != NumDstOps; ++I) {
DstUndefElements.set(I);
APInt &DstBits = DstBitElements[I];
for (unsigned J = 0; J != Scale; ++J) {
unsigned Idx = (I * Scale) + (IsLittleEndian ? J : (Scale - J - 1));
if (SrcUndefElements[Idx])
continue;
DstUndefElements.reset(I);
const APInt &SrcBits = SrcBitElements[Idx];
assert(SrcBits.getBitWidth() == SrcEltSizeInBits &&
"Illegal constant bitwidths");
DstBits.insertBits(SrcBits, J * SrcEltSizeInBits);
}
}
return;
}
// Split src element constant bits into dst elements.
unsigned Scale = SrcEltSizeInBits / DstEltSizeInBits;
for (unsigned I = 0; I != NumSrcOps; ++I) {
if (SrcUndefElements[I]) {
DstUndefElements.set(I * Scale, (I + 1) * Scale);
continue;
}
const APInt &SrcBits = SrcBitElements[I];
for (unsigned J = 0; J != Scale; ++J) {
unsigned Idx = (I * Scale) + (IsLittleEndian ? J : (Scale - J - 1));
APInt &DstBits = DstBitElements[Idx];
DstBits = SrcBits.extractBits(DstEltSizeInBits, J * DstEltSizeInBits);
}
}
}
bool BuildVectorSDNode::isConstant() const {
for (const SDValue &Op : op_values()) {
unsigned Opc = Op.getOpcode();
if (Opc != ISD::UNDEF && Opc != ISD::Constant && Opc != ISD::ConstantFP)
return false;
}
return true;
}
std::optional<std::pair<APInt, APInt>>
BuildVectorSDNode::isConstantSequence() const {
unsigned NumOps = getNumOperands();
if (NumOps < 2)
return std::nullopt;
if (!isa<ConstantSDNode>(getOperand(0)) ||
!isa<ConstantSDNode>(getOperand(1)))
return std::nullopt;
unsigned EltSize = getValueType(0).getScalarSizeInBits();
APInt Start = getConstantOperandAPInt(0).trunc(EltSize);
APInt Stride = getConstantOperandAPInt(1).trunc(EltSize) - Start;
if (Stride.isZero())
return std::nullopt;
for (unsigned i = 2; i < NumOps; ++i) {
if (!isa<ConstantSDNode>(getOperand(i)))
return std::nullopt;
APInt Val = getConstantOperandAPInt(i).trunc(EltSize);
if (Val != (Start + (Stride * i)))
return std::nullopt;
}
return std::make_pair(Start, Stride);
}
bool ShuffleVectorSDNode::isSplatMask(const int *Mask, EVT VT) {
// Find the first non-undef value in the shuffle mask.
unsigned i, e;
for (i = 0, e = VT.getVectorNumElements(); i != e && Mask[i] < 0; ++i)
/* search */;
// If all elements are undefined, this shuffle can be considered a splat
// (although it should eventually get simplified away completely).
if (i == e)
return true;
// Make sure all remaining elements are either undef or the same as the first
// non-undef value.
for (int Idx = Mask[i]; i != e; ++i)
if (Mask[i] >= 0 && Mask[i] != Idx)
return false;
return true;
}
// Returns the SDNode if it is a constant integer BuildVector
// or constant integer.
SDNode *SelectionDAG::isConstantIntBuildVectorOrConstantInt(SDValue N) const {
if (isa<ConstantSDNode>(N))
return N.getNode();
if (ISD::isBuildVectorOfConstantSDNodes(N.getNode()))
return N.getNode();
// Treat a GlobalAddress supporting constant offset folding as a
// constant integer.
if (GlobalAddressSDNode *GA = dyn_cast<GlobalAddressSDNode>(N))
if (GA->getOpcode() == ISD::GlobalAddress &&
TLI->isOffsetFoldingLegal(GA))
return GA;
if ((N.getOpcode() == ISD::SPLAT_VECTOR) &&
isa<ConstantSDNode>(N.getOperand(0)))
return N.getNode();
return nullptr;
}
// Returns the SDNode if it is a constant float BuildVector
// or constant float.
SDNode *SelectionDAG::isConstantFPBuildVectorOrConstantFP(SDValue N) const {
if (isa<ConstantFPSDNode>(N))
return N.getNode();
if (ISD::isBuildVectorOfConstantFPSDNodes(N.getNode()))
return N.getNode();
if ((N.getOpcode() == ISD::SPLAT_VECTOR) &&
isa<ConstantFPSDNode>(N.getOperand(0)))
return N.getNode();
return nullptr;
}
void SelectionDAG::createOperands(SDNode *Node, ArrayRef<SDValue> Vals) {
assert(!Node->OperandList && "Node already has operands");
assert(SDNode::getMaxNumOperands() >= Vals.size() &&
"too many operands to fit into SDNode");
SDUse *Ops = OperandRecycler.allocate(
ArrayRecycler<SDUse>::Capacity::get(Vals.size()), OperandAllocator);
bool IsDivergent = false;
for (unsigned I = 0; I != Vals.size(); ++I) {
Ops[I].setUser(Node);
Ops[I].setInitial(Vals[I]);
if (Ops[I].Val.getValueType() != MVT::Other) // Skip Chain. It does not carry divergence.
IsDivergent |= Ops[I].getNode()->isDivergent();
}
Node->NumOperands = Vals.size();
Node->OperandList = Ops;
if (!TLI->isSDNodeAlwaysUniform(Node)) {
IsDivergent |= TLI->isSDNodeSourceOfDivergence(Node, FLI, UA);
Node->SDNodeBits.IsDivergent = IsDivergent;
}
checkForCycles(Node);
}
SDValue SelectionDAG::getTokenFactor(const SDLoc &DL,
SmallVectorImpl<SDValue> &Vals) {
size_t Limit = SDNode::getMaxNumOperands();
while (Vals.size() > Limit) {
unsigned SliceIdx = Vals.size() - Limit;
auto ExtractedTFs = ArrayRef<SDValue>(Vals).slice(SliceIdx, Limit);
SDValue NewTF = getNode(ISD::TokenFactor, DL, MVT::Other, ExtractedTFs);
Vals.erase(Vals.begin() + SliceIdx, Vals.end());
Vals.emplace_back(NewTF);
}
return getNode(ISD::TokenFactor, DL, MVT::Other, Vals);
}
SDValue SelectionDAG::getNeutralElement(unsigned Opcode, const SDLoc &DL,
EVT VT, SDNodeFlags Flags) {
switch (Opcode) {
default:
return SDValue();
case ISD::ADD:
case ISD::OR:
case ISD::XOR:
case ISD::UMAX:
return getConstant(0, DL, VT);
case ISD::MUL:
return getConstant(1, DL, VT);
case ISD::AND:
case ISD::UMIN:
return getAllOnesConstant(DL, VT);
case ISD::SMAX:
return getConstant(APInt::getSignedMinValue(VT.getSizeInBits()), DL, VT);
case ISD::SMIN:
return getConstant(APInt::getSignedMaxValue(VT.getSizeInBits()), DL, VT);
case ISD::FADD:
return getConstantFP(-0.0, DL, VT);
case ISD::FMUL:
return getConstantFP(1.0, DL, VT);
case ISD::FMINNUM:
case ISD::FMAXNUM: {
// Neutral element for fminnum is NaN, Inf or FLT_MAX, depending on FMF.
const fltSemantics &Semantics = EVTToAPFloatSemantics(VT);
APFloat NeutralAF = !Flags.hasNoNaNs() ? APFloat::getQNaN(Semantics) :
!Flags.hasNoInfs() ? APFloat::getInf(Semantics) :
APFloat::getLargest(Semantics);
if (Opcode == ISD::FMAXNUM)
NeutralAF.changeSign();
return getConstantFP(NeutralAF, DL, VT);
}
case ISD::FMINIMUM:
case ISD::FMAXIMUM: {
// Neutral element for fminimum is Inf or FLT_MAX, depending on FMF.
const fltSemantics &Semantics = EVTToAPFloatSemantics(VT);
APFloat NeutralAF = !Flags.hasNoInfs() ? APFloat::getInf(Semantics)
: APFloat::getLargest(Semantics);
if (Opcode == ISD::FMAXIMUM)
NeutralAF.changeSign();
return getConstantFP(NeutralAF, DL, VT);
}
}
}
/// Helper used to make a call to a library function that has one argument of
/// pointer type.
///
/// Such functions include 'fegetmode', 'fesetenv' and some others, which are
/// used to get or set floating-point state. They have one argument of pointer
/// type, which points to the memory region containing bits of the
/// floating-point state. The value returned by such function is ignored in the
/// created call.
///
/// \param LibFunc Reference to library function (value of RTLIB::Libcall).
/// \param Ptr Pointer used to save/load state.
/// \param InChain Ingoing token chain.
/// \returns Outgoing chain token.
SDValue SelectionDAG::makeStateFunctionCall(unsigned LibFunc, SDValue Ptr,
SDValue InChain,
const SDLoc &DLoc) {
assert(InChain.getValueType() == MVT::Other && "Expected token chain");
TargetLowering::ArgListTy Args;
TargetLowering::ArgListEntry Entry;
Entry.Node = Ptr;
Entry.Ty = Ptr.getValueType().getTypeForEVT(*getContext());
Args.push_back(Entry);
RTLIB::Libcall LC = static_cast<RTLIB::Libcall>(LibFunc);
SDValue Callee = getExternalSymbol(TLI->getLibcallName(LC),
TLI->getPointerTy(getDataLayout()));
TargetLowering::CallLoweringInfo CLI(*this);
CLI.setDebugLoc(DLoc).setChain(InChain).setLibCallee(
TLI->getLibcallCallingConv(LC), Type::getVoidTy(*getContext()), Callee,
std::move(Args));
return TLI->LowerCallTo(CLI).second;
}
void SelectionDAG::copyExtraInfo(SDNode *From, SDNode *To) {
assert(From && To && "Invalid SDNode; empty source SDValue?");
auto I = SDEI.find(From);
if (I == SDEI.end())
return;
// Use of operator[] on the DenseMap may cause an insertion, which invalidates
// the iterator, hence the need to make a copy to prevent a use-after-free.
NodeExtraInfo NEI = I->second;
if (LLVM_LIKELY(!NEI.PCSections)) {
// No deep copy required for the types of extra info set.
//
// FIXME: Investigate if other types of extra info also need deep copy. This
// depends on the types of nodes they can be attached to: if some extra info
// is only ever attached to nodes where a replacement To node is always the
// node where later use and propagation of the extra info has the intended
// semantics, no deep copy is required.
SDEI[To] = std::move(NEI);
return;
}
// We need to copy NodeExtraInfo to all _new_ nodes that are being introduced
// through the replacement of From with To. Otherwise, replacements of a node
// (From) with more complex nodes (To and its operands) may result in lost
// extra info where the root node (To) is insignificant in further propagating
// and using extra info when further lowering to MIR.
//
// In the first step pre-populate the visited set with the nodes reachable
// from the old From node. This avoids copying NodeExtraInfo to parts of the
// DAG that is not new and should be left untouched.
SmallVector<const SDNode *> Leafs{From}; // Leafs reachable with VisitFrom.
DenseSet<const SDNode *> FromReach; // The set of nodes reachable from From.
auto VisitFrom = [&](auto &&Self, const SDNode *N, int MaxDepth) {
if (MaxDepth == 0) {
// Remember this node in case we need to increase MaxDepth and continue
// populating FromReach from this node.
Leafs.emplace_back(N);
return;
}
if (!FromReach.insert(N).second)
return;
for (const SDValue &Op : N->op_values())
Self(Self, Op.getNode(), MaxDepth - 1);
};
// Copy extra info to To and all its transitive operands (that are new).
SmallPtrSet<const SDNode *, 8> Visited;
auto DeepCopyTo = [&](auto &&Self, const SDNode *N) {
if (FromReach.contains(N))
return true;
if (!Visited.insert(N).second)
return true;
if (getEntryNode().getNode() == N)
return false;
for (const SDValue &Op : N->op_values()) {
if (!Self(Self, Op.getNode()))
return false;
}
// Copy only if entry node was not reached.
SDEI[N] = NEI;
return true;
};
// We first try with a lower MaxDepth, assuming that the path to common
// operands between From and To is relatively short. This significantly
// improves performance in the common case. The initial MaxDepth is big
// enough to avoid retry in the common case; the last MaxDepth is large
// enough to avoid having to use the fallback below (and protects from
// potential stack exhaustion from recursion).
for (int PrevDepth = 0, MaxDepth = 16; MaxDepth <= 1024;
PrevDepth = MaxDepth, MaxDepth *= 2, Visited.clear()) {
// StartFrom is the previous (or initial) set of leafs reachable at the
// previous maximum depth.
SmallVector<const SDNode *> StartFrom;
std::swap(StartFrom, Leafs);
for (const SDNode *N : StartFrom)
VisitFrom(VisitFrom, N, MaxDepth - PrevDepth);
if (LLVM_LIKELY(DeepCopyTo(DeepCopyTo, To)))
return;
// This should happen very rarely (reached the entry node).
LLVM_DEBUG(dbgs() << __func__ << ": MaxDepth=" << MaxDepth << " too low\n");
assert(!Leafs.empty());
}
// This should not happen - but if it did, that means the subgraph reachable
// from From has depth greater or equal to maximum MaxDepth, and VisitFrom()
// could not visit all reachable common operands. Consequently, we were able
// to reach the entry node.
errs() << "warning: incomplete propagation of SelectionDAG::NodeExtraInfo\n";
assert(false && "From subgraph too complex - increase max. MaxDepth?");
// Best-effort fallback if assertions disabled.
SDEI[To] = std::move(NEI);
}
#ifndef NDEBUG
static void checkForCyclesHelper(const SDNode *N,
SmallPtrSetImpl<const SDNode*> &Visited,
SmallPtrSetImpl<const SDNode*> &Checked,
const llvm::SelectionDAG *DAG) {
// If this node has already been checked, don't check it again.
if (Checked.count(N))
return;
// If a node has already been visited on this depth-first walk, reject it as
// a cycle.
if (!Visited.insert(N).second) {
errs() << "Detected cycle in SelectionDAG\n";
dbgs() << "Offending node:\n";
N->dumprFull(DAG); dbgs() << "\n";
abort();
}
for (const SDValue &Op : N->op_values())
checkForCyclesHelper(Op.getNode(), Visited, Checked, DAG);
Checked.insert(N);
Visited.erase(N);
}
#endif
void llvm::checkForCycles(const llvm::SDNode *N,
const llvm::SelectionDAG *DAG,
bool force) {
#ifndef NDEBUG
bool check = force;
#ifdef EXPENSIVE_CHECKS
check = true;
#endif // EXPENSIVE_CHECKS
if (check) {
assert(N && "Checking nonexistent SDNode");
SmallPtrSet<const SDNode*, 32> visited;
SmallPtrSet<const SDNode*, 32> checked;
checkForCyclesHelper(N, visited, checked, DAG);
}
#endif // !NDEBUG
}
void llvm::checkForCycles(const llvm::SelectionDAG *DAG, bool force) {
checkForCycles(DAG->getRoot().getNode(), DAG, force);
}
|