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
| //===-- ARMBaseInstrInfo.cpp - ARM Instruction Information ----------------===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
//
// This file contains the Base ARM implementation of the TargetInstrInfo class.
//
//===----------------------------------------------------------------------===//
#include "ARMBaseInstrInfo.h"
#include "ARMBaseRegisterInfo.h"
#include "ARMConstantPoolValue.h"
#include "ARMFeatures.h"
#include "ARMHazardRecognizer.h"
#include "ARMMachineFunctionInfo.h"
#include "ARMSubtarget.h"
#include "MCTargetDesc/ARMAddressingModes.h"
#include "MCTargetDesc/ARMBaseInfo.h"
#include "llvm/ADT/DenseMap.h"
#include "llvm/ADT/STLExtras.h"
#include "llvm/ADT/SmallSet.h"
#include "llvm/ADT/SmallVector.h"
#include "llvm/ADT/Triple.h"
#include "llvm/CodeGen/LiveVariables.h"
#include "llvm/CodeGen/MachineBasicBlock.h"
#include "llvm/CodeGen/MachineConstantPool.h"
#include "llvm/CodeGen/MachineFrameInfo.h"
#include "llvm/CodeGen/MachineFunction.h"
#include "llvm/CodeGen/MachineInstr.h"
#include "llvm/CodeGen/MachineInstrBuilder.h"
#include "llvm/CodeGen/MachineMemOperand.h"
#include "llvm/CodeGen/MachineOperand.h"
#include "llvm/CodeGen/MachineRegisterInfo.h"
#include "llvm/CodeGen/ScoreboardHazardRecognizer.h"
#include "llvm/CodeGen/SelectionDAGNodes.h"
#include "llvm/CodeGen/TargetInstrInfo.h"
#include "llvm/CodeGen/TargetRegisterInfo.h"
#include "llvm/CodeGen/TargetSchedule.h"
#include "llvm/IR/Attributes.h"
#include "llvm/IR/Constants.h"
#include "llvm/IR/DebugLoc.h"
#include "llvm/IR/Function.h"
#include "llvm/IR/GlobalValue.h"
#include "llvm/MC/MCAsmInfo.h"
#include "llvm/MC/MCInstrDesc.h"
#include "llvm/MC/MCInstrItineraries.h"
#include "llvm/Support/BranchProbability.h"
#include "llvm/Support/Casting.h"
#include "llvm/Support/CommandLine.h"
#include "llvm/Support/Compiler.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/ErrorHandling.h"
#include "llvm/Support/raw_ostream.h"
#include "llvm/Target/TargetMachine.h"
#include <algorithm>
#include <cassert>
#include <cstdint>
#include <iterator>
#include <new>
#include <utility>
#include <vector>
using namespace llvm;
#define DEBUG_TYPE "arm-instrinfo"
#define GET_INSTRINFO_CTOR_DTOR
#include "ARMGenInstrInfo.inc"
static cl::opt<bool>
EnableARM3Addr("enable-arm-3-addr-conv", cl::Hidden,
cl::desc("Enable ARM 2-addr to 3-addr conv"));
/// ARM_MLxEntry - Record information about MLA / MLS instructions.
struct ARM_MLxEntry {
uint16_t MLxOpc; // MLA / MLS opcode
uint16_t MulOpc; // Expanded multiplication opcode
uint16_t AddSubOpc; // Expanded add / sub opcode
bool NegAcc; // True if the acc is negated before the add / sub.
bool HasLane; // True if instruction has an extra "lane" operand.
};
static const ARM_MLxEntry ARM_MLxTable[] = {
// MLxOpc, MulOpc, AddSubOpc, NegAcc, HasLane
// fp scalar ops
{ ARM::VMLAS, ARM::VMULS, ARM::VADDS, false, false },
{ ARM::VMLSS, ARM::VMULS, ARM::VSUBS, false, false },
{ ARM::VMLAD, ARM::VMULD, ARM::VADDD, false, false },
{ ARM::VMLSD, ARM::VMULD, ARM::VSUBD, false, false },
{ ARM::VNMLAS, ARM::VNMULS, ARM::VSUBS, true, false },
{ ARM::VNMLSS, ARM::VMULS, ARM::VSUBS, true, false },
{ ARM::VNMLAD, ARM::VNMULD, ARM::VSUBD, true, false },
{ ARM::VNMLSD, ARM::VMULD, ARM::VSUBD, true, false },
// fp SIMD ops
{ ARM::VMLAfd, ARM::VMULfd, ARM::VADDfd, false, false },
{ ARM::VMLSfd, ARM::VMULfd, ARM::VSUBfd, false, false },
{ ARM::VMLAfq, ARM::VMULfq, ARM::VADDfq, false, false },
{ ARM::VMLSfq, ARM::VMULfq, ARM::VSUBfq, false, false },
{ ARM::VMLAslfd, ARM::VMULslfd, ARM::VADDfd, false, true },
{ ARM::VMLSslfd, ARM::VMULslfd, ARM::VSUBfd, false, true },
{ ARM::VMLAslfq, ARM::VMULslfq, ARM::VADDfq, false, true },
{ ARM::VMLSslfq, ARM::VMULslfq, ARM::VSUBfq, false, true },
};
ARMBaseInstrInfo::ARMBaseInstrInfo(const ARMSubtarget& STI)
: ARMGenInstrInfo(ARM::ADJCALLSTACKDOWN, ARM::ADJCALLSTACKUP),
Subtarget(STI) {
for (unsigned i = 0, e = array_lengthof(ARM_MLxTable); i != e; ++i) {
if (!MLxEntryMap.insert(std::make_pair(ARM_MLxTable[i].MLxOpc, i)).second)
llvm_unreachable("Duplicated entries?");
MLxHazardOpcodes.insert(ARM_MLxTable[i].AddSubOpc);
MLxHazardOpcodes.insert(ARM_MLxTable[i].MulOpc);
}
}
// Use a ScoreboardHazardRecognizer for prepass ARM scheduling. TargetInstrImpl
// currently defaults to no prepass hazard recognizer.
ScheduleHazardRecognizer *
ARMBaseInstrInfo::CreateTargetHazardRecognizer(const TargetSubtargetInfo *STI,
const ScheduleDAG *DAG) const {
if (usePreRAHazardRecognizer()) {
const InstrItineraryData *II =
static_cast<const ARMSubtarget *>(STI)->getInstrItineraryData();
return new ScoreboardHazardRecognizer(II, DAG, "pre-RA-sched");
}
return TargetInstrInfo::CreateTargetHazardRecognizer(STI, DAG);
}
ScheduleHazardRecognizer *ARMBaseInstrInfo::
CreateTargetPostRAHazardRecognizer(const InstrItineraryData *II,
const ScheduleDAG *DAG) const {
if (Subtarget.isThumb2() || Subtarget.hasVFP2Base())
return (ScheduleHazardRecognizer *)new ARMHazardRecognizer(II, DAG);
return TargetInstrInfo::CreateTargetPostRAHazardRecognizer(II, DAG);
}
MachineInstr *ARMBaseInstrInfo::convertToThreeAddress(
MachineFunction::iterator &MFI, MachineInstr &MI, LiveVariables *LV) const {
// FIXME: Thumb2 support.
if (!EnableARM3Addr)
return nullptr;
MachineFunction &MF = *MI.getParent()->getParent();
uint64_t TSFlags = MI.getDesc().TSFlags;
bool isPre = false;
switch ((TSFlags & ARMII::IndexModeMask) >> ARMII::IndexModeShift) {
default: return nullptr;
case ARMII::IndexModePre:
isPre = true;
break;
case ARMII::IndexModePost:
break;
}
// Try splitting an indexed load/store to an un-indexed one plus an add/sub
// operation.
unsigned MemOpc = getUnindexedOpcode(MI.getOpcode());
if (MemOpc == 0)
return nullptr;
MachineInstr *UpdateMI = nullptr;
MachineInstr *MemMI = nullptr;
unsigned AddrMode = (TSFlags & ARMII::AddrModeMask);
const MCInstrDesc &MCID = MI.getDesc();
unsigned NumOps = MCID.getNumOperands();
bool isLoad = !MI.mayStore();
const MachineOperand &WB = isLoad ? MI.getOperand(1) : MI.getOperand(0);
const MachineOperand &Base = MI.getOperand(2);
const MachineOperand &Offset = MI.getOperand(NumOps - 3);
Register WBReg = WB.getReg();
Register BaseReg = Base.getReg();
Register OffReg = Offset.getReg();
unsigned OffImm = MI.getOperand(NumOps - 2).getImm();
ARMCC::CondCodes Pred = (ARMCC::CondCodes)MI.getOperand(NumOps - 1).getImm();
switch (AddrMode) {
default: llvm_unreachable("Unknown indexed op!");
case ARMII::AddrMode2: {
bool isSub = ARM_AM::getAM2Op(OffImm) == ARM_AM::sub;
unsigned Amt = ARM_AM::getAM2Offset(OffImm);
if (OffReg == 0) {
if (ARM_AM::getSOImmVal(Amt) == -1)
// Can't encode it in a so_imm operand. This transformation will
// add more than 1 instruction. Abandon!
return nullptr;
UpdateMI = BuildMI(MF, MI.getDebugLoc(),
get(isSub ? ARM::SUBri : ARM::ADDri), WBReg)
.addReg(BaseReg)
.addImm(Amt)
.add(predOps(Pred))
.add(condCodeOp());
} else if (Amt != 0) {
ARM_AM::ShiftOpc ShOpc = ARM_AM::getAM2ShiftOpc(OffImm);
unsigned SOOpc = ARM_AM::getSORegOpc(ShOpc, Amt);
UpdateMI = BuildMI(MF, MI.getDebugLoc(),
get(isSub ? ARM::SUBrsi : ARM::ADDrsi), WBReg)
.addReg(BaseReg)
.addReg(OffReg)
.addReg(0)
.addImm(SOOpc)
.add(predOps(Pred))
.add(condCodeOp());
} else
UpdateMI = BuildMI(MF, MI.getDebugLoc(),
get(isSub ? ARM::SUBrr : ARM::ADDrr), WBReg)
.addReg(BaseReg)
.addReg(OffReg)
.add(predOps(Pred))
.add(condCodeOp());
break;
}
case ARMII::AddrMode3 : {
bool isSub = ARM_AM::getAM3Op(OffImm) == ARM_AM::sub;
unsigned Amt = ARM_AM::getAM3Offset(OffImm);
if (OffReg == 0)
// Immediate is 8-bits. It's guaranteed to fit in a so_imm operand.
UpdateMI = BuildMI(MF, MI.getDebugLoc(),
get(isSub ? ARM::SUBri : ARM::ADDri), WBReg)
.addReg(BaseReg)
.addImm(Amt)
.add(predOps(Pred))
.add(condCodeOp());
else
UpdateMI = BuildMI(MF, MI.getDebugLoc(),
get(isSub ? ARM::SUBrr : ARM::ADDrr), WBReg)
.addReg(BaseReg)
.addReg(OffReg)
.add(predOps(Pred))
.add(condCodeOp());
break;
}
}
std::vector<MachineInstr*> NewMIs;
if (isPre) {
if (isLoad)
MemMI =
BuildMI(MF, MI.getDebugLoc(), get(MemOpc), MI.getOperand(0).getReg())
.addReg(WBReg)
.addImm(0)
.addImm(Pred);
else
MemMI = BuildMI(MF, MI.getDebugLoc(), get(MemOpc))
.addReg(MI.getOperand(1).getReg())
.addReg(WBReg)
.addReg(0)
.addImm(0)
.addImm(Pred);
NewMIs.push_back(MemMI);
NewMIs.push_back(UpdateMI);
} else {
if (isLoad)
MemMI =
BuildMI(MF, MI.getDebugLoc(), get(MemOpc), MI.getOperand(0).getReg())
.addReg(BaseReg)
.addImm(0)
.addImm(Pred);
else
MemMI = BuildMI(MF, MI.getDebugLoc(), get(MemOpc))
.addReg(MI.getOperand(1).getReg())
.addReg(BaseReg)
.addReg(0)
.addImm(0)
.addImm(Pred);
if (WB.isDead())
UpdateMI->getOperand(0).setIsDead();
NewMIs.push_back(UpdateMI);
NewMIs.push_back(MemMI);
}
// Transfer LiveVariables states, kill / dead info.
if (LV) {
for (unsigned i = 0, e = MI.getNumOperands(); i != e; ++i) {
MachineOperand &MO = MI.getOperand(i);
if (MO.isReg() && Register::isVirtualRegister(MO.getReg())) {
Register Reg = MO.getReg();
LiveVariables::VarInfo &VI = LV->getVarInfo(Reg);
if (MO.isDef()) {
MachineInstr *NewMI = (Reg == WBReg) ? UpdateMI : MemMI;
if (MO.isDead())
LV->addVirtualRegisterDead(Reg, *NewMI);
}
if (MO.isUse() && MO.isKill()) {
for (unsigned j = 0; j < 2; ++j) {
// Look at the two new MI's in reverse order.
MachineInstr *NewMI = NewMIs[j];
if (!NewMI->readsRegister(Reg))
continue;
LV->addVirtualRegisterKilled(Reg, *NewMI);
if (VI.removeKill(MI))
VI.Kills.push_back(NewMI);
break;
}
}
}
}
}
MachineBasicBlock::iterator MBBI = MI.getIterator();
MFI->insert(MBBI, NewMIs[1]);
MFI->insert(MBBI, NewMIs[0]);
return NewMIs[0];
}
// Branch analysis.
bool ARMBaseInstrInfo::analyzeBranch(MachineBasicBlock &MBB,
MachineBasicBlock *&TBB,
MachineBasicBlock *&FBB,
SmallVectorImpl<MachineOperand> &Cond,
bool AllowModify) const {
TBB = nullptr;
FBB = nullptr;
MachineBasicBlock::iterator I = MBB.end();
if (I == MBB.begin())
return false; // Empty blocks are easy.
--I;
// Walk backwards from the end of the basic block until the branch is
// analyzed or we give up.
while (isPredicated(*I) || I->isTerminator() || I->isDebugValue()) {
// Flag to be raised on unanalyzeable instructions. This is useful in cases
// where we want to clean up on the end of the basic block before we bail
// out.
bool CantAnalyze = false;
// Skip over DEBUG values and predicated nonterminators.
while (I->isDebugInstr() || !I->isTerminator()) {
if (I == MBB.begin())
return false;
--I;
}
if (isIndirectBranchOpcode(I->getOpcode()) ||
isJumpTableBranchOpcode(I->getOpcode())) {
// Indirect branches and jump tables can't be analyzed, but we still want
// to clean up any instructions at the tail of the basic block.
CantAnalyze = true;
} else if (isUncondBranchOpcode(I->getOpcode())) {
TBB = I->getOperand(0).getMBB();
} else if (isCondBranchOpcode(I->getOpcode())) {
// Bail out if we encounter multiple conditional branches.
if (!Cond.empty())
return true;
assert(!FBB && "FBB should have been null.");
FBB = TBB;
TBB = I->getOperand(0).getMBB();
Cond.push_back(I->getOperand(1));
Cond.push_back(I->getOperand(2));
} else if (I->isReturn()) {
// Returns can't be analyzed, but we should run cleanup.
CantAnalyze = !isPredicated(*I);
} else {
// We encountered other unrecognized terminator. Bail out immediately.
return true;
}
// Cleanup code - to be run for unpredicated unconditional branches and
// returns.
if (!isPredicated(*I) &&
(isUncondBranchOpcode(I->getOpcode()) ||
isIndirectBranchOpcode(I->getOpcode()) ||
isJumpTableBranchOpcode(I->getOpcode()) ||
I->isReturn())) {
// Forget any previous condition branch information - it no longer applies.
Cond.clear();
FBB = nullptr;
// If we can modify the function, delete everything below this
// unconditional branch.
if (AllowModify) {
MachineBasicBlock::iterator DI = std::next(I);
while (DI != MBB.end()) {
MachineInstr &InstToDelete = *DI;
++DI;
InstToDelete.eraseFromParent();
}
}
}
if (CantAnalyze)
return true;
if (I == MBB.begin())
return false;
--I;
}
// We made it past the terminators without bailing out - we must have
// analyzed this branch successfully.
return false;
}
unsigned ARMBaseInstrInfo::removeBranch(MachineBasicBlock &MBB,
int *BytesRemoved) const {
assert(!BytesRemoved && "code size not handled");
MachineBasicBlock::iterator I = MBB.getLastNonDebugInstr();
if (I == MBB.end())
return 0;
if (!isUncondBranchOpcode(I->getOpcode()) &&
!isCondBranchOpcode(I->getOpcode()))
return 0;
// Remove the branch.
I->eraseFromParent();
I = MBB.end();
if (I == MBB.begin()) return 1;
--I;
if (!isCondBranchOpcode(I->getOpcode()))
return 1;
// Remove the branch.
I->eraseFromParent();
return 2;
}
unsigned ARMBaseInstrInfo::insertBranch(MachineBasicBlock &MBB,
MachineBasicBlock *TBB,
MachineBasicBlock *FBB,
ArrayRef<MachineOperand> Cond,
const DebugLoc &DL,
int *BytesAdded) const {
assert(!BytesAdded && "code size not handled");
ARMFunctionInfo *AFI = MBB.getParent()->getInfo<ARMFunctionInfo>();
int BOpc = !AFI->isThumbFunction()
? ARM::B : (AFI->isThumb2Function() ? ARM::t2B : ARM::tB);
int BccOpc = !AFI->isThumbFunction()
? ARM::Bcc : (AFI->isThumb2Function() ? ARM::t2Bcc : ARM::tBcc);
bool isThumb = AFI->isThumbFunction() || AFI->isThumb2Function();
// Shouldn't be a fall through.
assert(TBB && "insertBranch must not be told to insert a fallthrough");
assert((Cond.size() == 2 || Cond.size() == 0) &&
"ARM branch conditions have two components!");
// For conditional branches, we use addOperand to preserve CPSR flags.
if (!FBB) {
if (Cond.empty()) { // Unconditional branch?
if (isThumb)
BuildMI(&MBB, DL, get(BOpc)).addMBB(TBB).add(predOps(ARMCC::AL));
else
BuildMI(&MBB, DL, get(BOpc)).addMBB(TBB);
} else
BuildMI(&MBB, DL, get(BccOpc))
.addMBB(TBB)
.addImm(Cond[0].getImm())
.add(Cond[1]);
return 1;
}
// Two-way conditional branch.
BuildMI(&MBB, DL, get(BccOpc))
.addMBB(TBB)
.addImm(Cond[0].getImm())
.add(Cond[1]);
if (isThumb)
BuildMI(&MBB, DL, get(BOpc)).addMBB(FBB).add(predOps(ARMCC::AL));
else
BuildMI(&MBB, DL, get(BOpc)).addMBB(FBB);
return 2;
}
bool ARMBaseInstrInfo::
reverseBranchCondition(SmallVectorImpl<MachineOperand> &Cond) const {
ARMCC::CondCodes CC = (ARMCC::CondCodes)(int)Cond[0].getImm();
Cond[0].setImm(ARMCC::getOppositeCondition(CC));
return false;
}
bool ARMBaseInstrInfo::isPredicated(const MachineInstr &MI) const {
if (MI.isBundle()) {
MachineBasicBlock::const_instr_iterator I = MI.getIterator();
MachineBasicBlock::const_instr_iterator E = MI.getParent()->instr_end();
while (++I != E && I->isInsideBundle()) {
int PIdx = I->findFirstPredOperandIdx();
if (PIdx != -1 && I->getOperand(PIdx).getImm() != ARMCC::AL)
return true;
}
return false;
}
int PIdx = MI.findFirstPredOperandIdx();
return PIdx != -1 && MI.getOperand(PIdx).getImm() != ARMCC::AL;
}
bool ARMBaseInstrInfo::PredicateInstruction(
MachineInstr &MI, ArrayRef<MachineOperand> Pred) const {
unsigned Opc = MI.getOpcode();
if (isUncondBranchOpcode(Opc)) {
MI.setDesc(get(getMatchingCondBranchOpcode(Opc)));
MachineInstrBuilder(*MI.getParent()->getParent(), MI)
.addImm(Pred[0].getImm())
.addReg(Pred[1].getReg());
return true;
}
int PIdx = MI.findFirstPredOperandIdx();
if (PIdx != -1) {
MachineOperand &PMO = MI.getOperand(PIdx);
PMO.setImm(Pred[0].getImm());
MI.getOperand(PIdx+1).setReg(Pred[1].getReg());
return true;
}
return false;
}
bool ARMBaseInstrInfo::SubsumesPredicate(ArrayRef<MachineOperand> Pred1,
ArrayRef<MachineOperand> Pred2) const {
if (Pred1.size() > 2 || Pred2.size() > 2)
return false;
ARMCC::CondCodes CC1 = (ARMCC::CondCodes)Pred1[0].getImm();
ARMCC::CondCodes CC2 = (ARMCC::CondCodes)Pred2[0].getImm();
if (CC1 == CC2)
return true;
switch (CC1) {
default:
return false;
case ARMCC::AL:
return true;
case ARMCC::HS:
return CC2 == ARMCC::HI;
case ARMCC::LS:
return CC2 == ARMCC::LO || CC2 == ARMCC::EQ;
case ARMCC::GE:
return CC2 == ARMCC::GT;
case ARMCC::LE:
return CC2 == ARMCC::LT;
}
}
bool ARMBaseInstrInfo::DefinesPredicate(
MachineInstr &MI, std::vector<MachineOperand> &Pred) const {
bool Found = false;
for (unsigned i = 0, e = MI.getNumOperands(); i != e; ++i) {
const MachineOperand &MO = MI.getOperand(i);
if ((MO.isRegMask() && MO.clobbersPhysReg(ARM::CPSR)) ||
(MO.isReg() && MO.isDef() && MO.getReg() == ARM::CPSR)) {
Pred.push_back(MO);
Found = true;
}
}
return Found;
}
bool ARMBaseInstrInfo::isCPSRDefined(const MachineInstr &MI) {
for (const auto &MO : MI.operands())
if (MO.isReg() && MO.getReg() == ARM::CPSR && MO.isDef() && !MO.isDead())
return true;
return false;
}
bool ARMBaseInstrInfo::isAddrMode3OpImm(const MachineInstr &MI,
unsigned Op) const {
const MachineOperand &Offset = MI.getOperand(Op + 1);
return Offset.getReg() != 0;
}
// Load with negative register offset requires additional 1cyc and +I unit
// for Cortex A57
bool ARMBaseInstrInfo::isAddrMode3OpMinusReg(const MachineInstr &MI,
unsigned Op) const {
const MachineOperand &Offset = MI.getOperand(Op + 1);
const MachineOperand &Opc = MI.getOperand(Op + 2);
assert(Opc.isImm());
assert(Offset.isReg());
int64_t OpcImm = Opc.getImm();
bool isSub = ARM_AM::getAM3Op(OpcImm) == ARM_AM::sub;
return (isSub && Offset.getReg() != 0);
}
bool ARMBaseInstrInfo::isLdstScaledReg(const MachineInstr &MI,
unsigned Op) const {
const MachineOperand &Opc = MI.getOperand(Op + 2);
unsigned OffImm = Opc.getImm();
return ARM_AM::getAM2ShiftOpc(OffImm) != ARM_AM::no_shift;
}
// Load, scaled register offset, not plus LSL2
bool ARMBaseInstrInfo::isLdstScaledRegNotPlusLsl2(const MachineInstr &MI,
unsigned Op) const {
const MachineOperand &Opc = MI.getOperand(Op + 2);
unsigned OffImm = Opc.getImm();
bool isAdd = ARM_AM::getAM2Op(OffImm) == ARM_AM::add;
unsigned Amt = ARM_AM::getAM2Offset(OffImm);
ARM_AM::ShiftOpc ShiftOpc = ARM_AM::getAM2ShiftOpc(OffImm);
if (ShiftOpc == ARM_AM::no_shift) return false; // not scaled
bool SimpleScaled = (isAdd && ShiftOpc == ARM_AM::lsl && Amt == 2);
return !SimpleScaled;
}
// Minus reg for ldstso addr mode
bool ARMBaseInstrInfo::isLdstSoMinusReg(const MachineInstr &MI,
unsigned Op) const {
unsigned OffImm = MI.getOperand(Op + 2).getImm();
return ARM_AM::getAM2Op(OffImm) == ARM_AM::sub;
}
// Load, scaled register offset
bool ARMBaseInstrInfo::isAm2ScaledReg(const MachineInstr &MI,
unsigned Op) const {
unsigned OffImm = MI.getOperand(Op + 2).getImm();
return ARM_AM::getAM2ShiftOpc(OffImm) != ARM_AM::no_shift;
}
static bool isEligibleForITBlock(const MachineInstr *MI) {
switch (MI->getOpcode()) {
default: return true;
case ARM::tADC: // ADC (register) T1
case ARM::tADDi3: // ADD (immediate) T1
case ARM::tADDi8: // ADD (immediate) T2
case ARM::tADDrr: // ADD (register) T1
case ARM::tAND: // AND (register) T1
case ARM::tASRri: // ASR (immediate) T1
case ARM::tASRrr: // ASR (register) T1
case ARM::tBIC: // BIC (register) T1
case ARM::tEOR: // EOR (register) T1
case ARM::tLSLri: // LSL (immediate) T1
case ARM::tLSLrr: // LSL (register) T1
case ARM::tLSRri: // LSR (immediate) T1
case ARM::tLSRrr: // LSR (register) T1
case ARM::tMUL: // MUL T1
case ARM::tMVN: // MVN (register) T1
case ARM::tORR: // ORR (register) T1
case ARM::tROR: // ROR (register) T1
case ARM::tRSB: // RSB (immediate) T1
case ARM::tSBC: // SBC (register) T1
case ARM::tSUBi3: // SUB (immediate) T1
case ARM::tSUBi8: // SUB (immediate) T2
case ARM::tSUBrr: // SUB (register) T1
return !ARMBaseInstrInfo::isCPSRDefined(*MI);
}
}
/// isPredicable - Return true if the specified instruction can be predicated.
/// By default, this returns true for every instruction with a
/// PredicateOperand.
bool ARMBaseInstrInfo::isPredicable(const MachineInstr &MI) const {
if (!MI.isPredicable())
return false;
if (MI.isBundle())
return false;
if (!isEligibleForITBlock(&MI))
return false;
const ARMFunctionInfo *AFI =
MI.getParent()->getParent()->getInfo<ARMFunctionInfo>();
// Neon instructions in Thumb2 IT blocks are deprecated, see ARMARM.
// In their ARM encoding, they can't be encoded in a conditional form.
if ((MI.getDesc().TSFlags & ARMII::DomainMask) == ARMII::DomainNEON)
return false;
if (AFI->isThumb2Function()) {
if (getSubtarget().restrictIT())
return isV8EligibleForIT(&MI);
}
return true;
}
namespace llvm {
template <> bool IsCPSRDead<MachineInstr>(const MachineInstr *MI) {
for (unsigned i = 0, e = MI->getNumOperands(); i != e; ++i) {
const MachineOperand &MO = MI->getOperand(i);
if (!MO.isReg() || MO.isUndef() || MO.isUse())
continue;
if (MO.getReg() != ARM::CPSR)
continue;
if (!MO.isDead())
return false;
}
// all definitions of CPSR are dead
return true;
}
} // end namespace llvm
/// GetInstSize - Return the size of the specified MachineInstr.
///
unsigned ARMBaseInstrInfo::getInstSizeInBytes(const MachineInstr &MI) const {
const MachineBasicBlock &MBB = *MI.getParent();
const MachineFunction *MF = MBB.getParent();
const MCAsmInfo *MAI = MF->getTarget().getMCAsmInfo();
const MCInstrDesc &MCID = MI.getDesc();
if (MCID.getSize())
return MCID.getSize();
switch (MI.getOpcode()) {
default:
// pseudo-instruction sizes are zero.
return 0;
case TargetOpcode::BUNDLE:
return getInstBundleLength(MI);
case ARM::MOVi16_ga_pcrel:
case ARM::MOVTi16_ga_pcrel:
case ARM::t2MOVi16_ga_pcrel:
case ARM::t2MOVTi16_ga_pcrel:
return 4;
case ARM::MOVi32imm:
case ARM::t2MOVi32imm:
return 8;
case ARM::CONSTPOOL_ENTRY:
case ARM::JUMPTABLE_INSTS:
case ARM::JUMPTABLE_ADDRS:
case ARM::JUMPTABLE_TBB:
case ARM::JUMPTABLE_TBH:
// If this machine instr is a constant pool entry, its size is recorded as
// operand #2.
return MI.getOperand(2).getImm();
case ARM::Int_eh_sjlj_longjmp:
return 16;
case ARM::tInt_eh_sjlj_longjmp:
return 10;
case ARM::tInt_WIN_eh_sjlj_longjmp:
return 12;
case ARM::Int_eh_sjlj_setjmp:
case ARM::Int_eh_sjlj_setjmp_nofp:
return 20;
case ARM::tInt_eh_sjlj_setjmp:
case ARM::t2Int_eh_sjlj_setjmp:
case ARM::t2Int_eh_sjlj_setjmp_nofp:
return 12;
case ARM::SPACE:
return MI.getOperand(1).getImm();
case ARM::INLINEASM:
case ARM::INLINEASM_BR: {
// If this machine instr is an inline asm, measure it.
unsigned Size = getInlineAsmLength(MI.getOperand(0).getSymbolName(), *MAI);
if (!MF->getInfo<ARMFunctionInfo>()->isThumbFunction())
Size = alignTo(Size, 4);
return Size;
}
}
}
unsigned ARMBaseInstrInfo::getInstBundleLength(const MachineInstr &MI) const {
unsigned Size = 0;
MachineBasicBlock::const_instr_iterator I = MI.getIterator();
MachineBasicBlock::const_instr_iterator E = MI.getParent()->instr_end();
while (++I != E && I->isInsideBundle()) {
assert(!I->isBundle() && "No nested bundle!");
Size += getInstSizeInBytes(*I);
}
return Size;
}
void ARMBaseInstrInfo::copyFromCPSR(MachineBasicBlock &MBB,
MachineBasicBlock::iterator I,
unsigned DestReg, bool KillSrc,
const ARMSubtarget &Subtarget) const {
unsigned Opc = Subtarget.isThumb()
? (Subtarget.isMClass() ? ARM::t2MRS_M : ARM::t2MRS_AR)
: ARM::MRS;
MachineInstrBuilder MIB =
BuildMI(MBB, I, I->getDebugLoc(), get(Opc), DestReg);
// There is only 1 A/R class MRS instruction, and it always refers to
// APSR. However, there are lots of other possibilities on M-class cores.
if (Subtarget.isMClass())
MIB.addImm(0x800);
MIB.add(predOps(ARMCC::AL))
.addReg(ARM::CPSR, RegState::Implicit | getKillRegState(KillSrc));
}
void ARMBaseInstrInfo::copyToCPSR(MachineBasicBlock &MBB,
MachineBasicBlock::iterator I,
unsigned SrcReg, bool KillSrc,
const ARMSubtarget &Subtarget) const {
unsigned Opc = Subtarget.isThumb()
? (Subtarget.isMClass() ? ARM::t2MSR_M : ARM::t2MSR_AR)
: ARM::MSR;
MachineInstrBuilder MIB = BuildMI(MBB, I, I->getDebugLoc(), get(Opc));
if (Subtarget.isMClass())
MIB.addImm(0x800);
else
MIB.addImm(8);
MIB.addReg(SrcReg, getKillRegState(KillSrc))
.add(predOps(ARMCC::AL))
.addReg(ARM::CPSR, RegState::Implicit | RegState::Define);
}
void llvm::addUnpredicatedMveVpredNOp(MachineInstrBuilder &MIB) {
MIB.addImm(ARMVCC::None);
MIB.addReg(0);
}
void llvm::addUnpredicatedMveVpredROp(MachineInstrBuilder &MIB,
unsigned DestReg) {
addUnpredicatedMveVpredNOp(MIB);
MIB.addReg(DestReg, RegState::Undef);
}
void llvm::addPredicatedMveVpredNOp(MachineInstrBuilder &MIB, unsigned Cond) {
MIB.addImm(Cond);
MIB.addReg(ARM::VPR, RegState::Implicit);
}
void llvm::addPredicatedMveVpredROp(MachineInstrBuilder &MIB,
unsigned Cond, unsigned Inactive) {
addPredicatedMveVpredNOp(MIB, Cond);
MIB.addReg(Inactive);
}
void ARMBaseInstrInfo::copyPhysReg(MachineBasicBlock &MBB,
MachineBasicBlock::iterator I,
const DebugLoc &DL, unsigned DestReg,
unsigned SrcReg, bool KillSrc) const {
bool GPRDest = ARM::GPRRegClass.contains(DestReg);
bool GPRSrc = ARM::GPRRegClass.contains(SrcReg);
if (GPRDest && GPRSrc) {
BuildMI(MBB, I, DL, get(ARM::MOVr), DestReg)
.addReg(SrcReg, getKillRegState(KillSrc))
.add(predOps(ARMCC::AL))
.add(condCodeOp());
return;
}
bool SPRDest = ARM::SPRRegClass.contains(DestReg);
bool SPRSrc = ARM::SPRRegClass.contains(SrcReg);
unsigned Opc = 0;
if (SPRDest && SPRSrc)
Opc = ARM::VMOVS;
else if (GPRDest && SPRSrc)
Opc = ARM::VMOVRS;
else if (SPRDest && GPRSrc)
Opc = ARM::VMOVSR;
else if (ARM::DPRRegClass.contains(DestReg, SrcReg) && Subtarget.hasFP64())
Opc = ARM::VMOVD;
else if (ARM::QPRRegClass.contains(DestReg, SrcReg))
Opc = Subtarget.hasNEON() ? ARM::VORRq : ARM::MVE_VORR;
if (Opc) {
MachineInstrBuilder MIB = BuildMI(MBB, I, DL, get(Opc), DestReg);
MIB.addReg(SrcReg, getKillRegState(KillSrc));
if (Opc == ARM::VORRq || Opc == ARM::MVE_VORR)
MIB.addReg(SrcReg, getKillRegState(KillSrc));
if (Opc == ARM::MVE_VORR)
addUnpredicatedMveVpredROp(MIB, DestReg);
else
MIB.add(predOps(ARMCC::AL));
return;
}
// Handle register classes that require multiple instructions.
unsigned BeginIdx = 0;
unsigned SubRegs = 0;
int Spacing = 1;
// Use VORRq when possible.
if (ARM::QQPRRegClass.contains(DestReg, SrcReg)) {
Opc = Subtarget.hasNEON() ? ARM::VORRq : ARM::MVE_VORR;
BeginIdx = ARM::qsub_0;
SubRegs = 2;
} else if (ARM::QQQQPRRegClass.contains(DestReg, SrcReg)) {
Opc = Subtarget.hasNEON() ? ARM::VORRq : ARM::MVE_VORR;
BeginIdx = ARM::qsub_0;
SubRegs = 4;
// Fall back to VMOVD.
} else if (ARM::DPairRegClass.contains(DestReg, SrcReg)) {
Opc = ARM::VMOVD;
BeginIdx = ARM::dsub_0;
SubRegs = 2;
} else if (ARM::DTripleRegClass.contains(DestReg, SrcReg)) {
Opc = ARM::VMOVD;
BeginIdx = ARM::dsub_0;
SubRegs = 3;
} else if (ARM::DQuadRegClass.contains(DestReg, SrcReg)) {
Opc = ARM::VMOVD;
BeginIdx = ARM::dsub_0;
SubRegs = 4;
} else if (ARM::GPRPairRegClass.contains(DestReg, SrcReg)) {
Opc = Subtarget.isThumb2() ? ARM::tMOVr : ARM::MOVr;
BeginIdx = ARM::gsub_0;
SubRegs = 2;
} else if (ARM::DPairSpcRegClass.contains(DestReg, SrcReg)) {
Opc = ARM::VMOVD;
BeginIdx = ARM::dsub_0;
SubRegs = 2;
Spacing = 2;
} else if (ARM::DTripleSpcRegClass.contains(DestReg, SrcReg)) {
Opc = ARM::VMOVD;
BeginIdx = ARM::dsub_0;
SubRegs = 3;
Spacing = 2;
} else if (ARM::DQuadSpcRegClass.contains(DestReg, SrcReg)) {
Opc = ARM::VMOVD;
BeginIdx = ARM::dsub_0;
SubRegs = 4;
Spacing = 2;
} else if (ARM::DPRRegClass.contains(DestReg, SrcReg) &&
!Subtarget.hasFP64()) {
Opc = ARM::VMOVS;
BeginIdx = ARM::ssub_0;
SubRegs = 2;
} else if (SrcReg == ARM::CPSR) {
copyFromCPSR(MBB, I, DestReg, KillSrc, Subtarget);
return;
} else if (DestReg == ARM::CPSR) {
copyToCPSR(MBB, I, SrcReg, KillSrc, Subtarget);
return;
} else if (DestReg == ARM::VPR) {
assert(ARM::GPRRegClass.contains(SrcReg));
BuildMI(MBB, I, I->getDebugLoc(), get(ARM::VMSR_P0), DestReg)
.addReg(SrcReg, getKillRegState(KillSrc))
.add(predOps(ARMCC::AL));
return;
} else if (SrcReg == ARM::VPR) {
assert(ARM::GPRRegClass.contains(DestReg));
BuildMI(MBB, I, I->getDebugLoc(), get(ARM::VMRS_P0), DestReg)
.addReg(SrcReg, getKillRegState(KillSrc))
.add(predOps(ARMCC::AL));
return;
} else if (DestReg == ARM::FPSCR_NZCV) {
assert(ARM::GPRRegClass.contains(SrcReg));
BuildMI(MBB, I, I->getDebugLoc(), get(ARM::VMSR_FPSCR_NZCVQC), DestReg)
.addReg(SrcReg, getKillRegState(KillSrc))
.add(predOps(ARMCC::AL));
return;
} else if (SrcReg == ARM::FPSCR_NZCV) {
assert(ARM::GPRRegClass.contains(DestReg));
BuildMI(MBB, I, I->getDebugLoc(), get(ARM::VMRS_FPSCR_NZCVQC), DestReg)
.addReg(SrcReg, getKillRegState(KillSrc))
.add(predOps(ARMCC::AL));
return;
}
assert(Opc && "Impossible reg-to-reg copy");
const TargetRegisterInfo *TRI = &getRegisterInfo();
MachineInstrBuilder Mov;
// Copy register tuples backward when the first Dest reg overlaps with SrcReg.
if (TRI->regsOverlap(SrcReg, TRI->getSubReg(DestReg, BeginIdx))) {
BeginIdx = BeginIdx + ((SubRegs - 1) * Spacing);
Spacing = -Spacing;
}
#ifndef NDEBUG
SmallSet<unsigned, 4> DstRegs;
#endif
for (unsigned i = 0; i != SubRegs; ++i) {
Register Dst = TRI->getSubReg(DestReg, BeginIdx + i * Spacing);
Register Src = TRI->getSubReg(SrcReg, BeginIdx + i * Spacing);
assert(Dst && Src && "Bad sub-register");
#ifndef NDEBUG
assert(!DstRegs.count(Src) && "destructive vector copy");
DstRegs.insert(Dst);
#endif
Mov = BuildMI(MBB, I, I->getDebugLoc(), get(Opc), Dst).addReg(Src);
// VORR (NEON or MVE) takes two source operands.
if (Opc == ARM::VORRq || Opc == ARM::MVE_VORR) {
Mov.addReg(Src);
}
// MVE VORR takes predicate operands in place of an ordinary condition.
if (Opc == ARM::MVE_VORR)
addUnpredicatedMveVpredROp(Mov, Dst);
else
Mov = Mov.add(predOps(ARMCC::AL));
// MOVr can set CC.
if (Opc == ARM::MOVr)
Mov = Mov.add(condCodeOp());
}
// Add implicit super-register defs and kills to the last instruction.
Mov->addRegisterDefined(DestReg, TRI);
if (KillSrc)
Mov->addRegisterKilled(SrcReg, TRI);
}
bool ARMBaseInstrInfo::isCopyInstrImpl(const MachineInstr &MI,
const MachineOperand *&Src,
const MachineOperand *&Dest) const {
// VMOVRRD is also a copy instruction but it requires
// special way of handling. It is more complex copy version
// and since that we are not considering it. For recognition
// of such instruction isExtractSubregLike MI interface fuction
// could be used.
// VORRq is considered as a move only if two inputs are
// the same register.
if (!MI.isMoveReg() ||
(MI.getOpcode() == ARM::VORRq &&
MI.getOperand(1).getReg() != MI.getOperand(2).getReg()))
return false;
Dest = &MI.getOperand(0);
Src = &MI.getOperand(1);
return true;
}
const MachineInstrBuilder &
ARMBaseInstrInfo::AddDReg(MachineInstrBuilder &MIB, unsigned Reg,
unsigned SubIdx, unsigned State,
const TargetRegisterInfo *TRI) const {
if (!SubIdx)
return MIB.addReg(Reg, State);
if (Register::isPhysicalRegister(Reg))
return MIB.addReg(TRI->getSubReg(Reg, SubIdx), State);
return MIB.addReg(Reg, State, SubIdx);
}
void ARMBaseInstrInfo::
storeRegToStackSlot(MachineBasicBlock &MBB, MachineBasicBlock::iterator I,
unsigned SrcReg, bool isKill, int FI,
const TargetRegisterClass *RC,
const TargetRegisterInfo *TRI) const {
MachineFunction &MF = *MBB.getParent();
MachineFrameInfo &MFI = MF.getFrameInfo();
unsigned Align = MFI.getObjectAlignment(FI);
MachineMemOperand *MMO = MF.getMachineMemOperand(
MachinePointerInfo::getFixedStack(MF, FI), MachineMemOperand::MOStore,
MFI.getObjectSize(FI), Align);
switch (TRI->getSpillSize(*RC)) {
case 2:
if (ARM::HPRRegClass.hasSubClassEq(RC)) {
BuildMI(MBB, I, DebugLoc(), get(ARM::VSTRH))
.addReg(SrcReg, getKillRegState(isKill))
.addFrameIndex(FI)
.addImm(0)
.addMemOperand(MMO)
.add(predOps(ARMCC::AL));
} else
llvm_unreachable("Unknown reg class!");
break;
case 4:
if (ARM::GPRRegClass.hasSubClassEq(RC)) {
BuildMI(MBB, I, DebugLoc(), get(ARM::STRi12))
.addReg(SrcReg, getKillRegState(isKill))
.addFrameIndex(FI)
.addImm(0)
.addMemOperand(MMO)
.add(predOps(ARMCC::AL));
} else if (ARM::SPRRegClass.hasSubClassEq(RC)) {
BuildMI(MBB, I, DebugLoc(), get(ARM::VSTRS))
.addReg(SrcReg, getKillRegState(isKill))
.addFrameIndex(FI)
.addImm(0)
.addMemOperand(MMO)
.add(predOps(ARMCC::AL));
} else if (ARM::VCCRRegClass.hasSubClassEq(RC)) {
BuildMI(MBB, I, DebugLoc(), get(ARM::VSTR_P0_off))
.addReg(SrcReg, getKillRegState(isKill))
.addFrameIndex(FI)
.addImm(0)
.addMemOperand(MMO)
.add(predOps(ARMCC::AL));
} else
llvm_unreachable("Unknown reg class!");
break;
case 8:
if (ARM::DPRRegClass.hasSubClassEq(RC)) {
BuildMI(MBB, I, DebugLoc(), get(ARM::VSTRD))
.addReg(SrcReg, getKillRegState(isKill))
.addFrameIndex(FI)
.addImm(0)
.addMemOperand(MMO)
.add(predOps(ARMCC::AL));
} else if (ARM::GPRPairRegClass.hasSubClassEq(RC)) {
if (Subtarget.hasV5TEOps()) {
MachineInstrBuilder MIB = BuildMI(MBB, I, DebugLoc(), get(ARM::STRD));
AddDReg(MIB, SrcReg, ARM::gsub_0, getKillRegState(isKill), TRI);
AddDReg(MIB, SrcReg, ARM::gsub_1, 0, TRI);
MIB.addFrameIndex(FI).addReg(0).addImm(0).addMemOperand(MMO)
.add(predOps(ARMCC::AL));
} else {
// Fallback to STM instruction, which has existed since the dawn of
// time.
MachineInstrBuilder MIB = BuildMI(MBB, I, DebugLoc(), get(ARM::STMIA))
.addFrameIndex(FI)
.addMemOperand(MMO)
.add(predOps(ARMCC::AL));
AddDReg(MIB, SrcReg, ARM::gsub_0, getKillRegState(isKill), TRI);
AddDReg(MIB, SrcReg, ARM::gsub_1, 0, TRI);
}
} else
llvm_unreachable("Unknown reg class!");
break;
case 16:
if (ARM::DPairRegClass.hasSubClassEq(RC) && Subtarget.hasNEON()) {
// Use aligned spills if the stack can be realigned.
if (Align >= 16 && getRegisterInfo().canRealignStack(MF)) {
BuildMI(MBB, I, DebugLoc(), get(ARM::VST1q64))
.addFrameIndex(FI)
.addImm(16)
.addReg(SrcReg, getKillRegState(isKill))
.addMemOperand(MMO)
.add(predOps(ARMCC::AL));
} else {
BuildMI(MBB, I, DebugLoc(), get(ARM::VSTMQIA))
.addReg(SrcReg, getKillRegState(isKill))
.addFrameIndex(FI)
.addMemOperand(MMO)
.add(predOps(ARMCC::AL));
}
} else if (ARM::QPRRegClass.hasSubClassEq(RC) &&
Subtarget.hasMVEIntegerOps()) {
auto MIB = BuildMI(MBB, I, DebugLoc(), get(ARM::MVE_VSTRWU32));
MIB.addReg(SrcReg, getKillRegState(isKill))
.addFrameIndex(FI)
.addImm(0)
.addMemOperand(MMO);
addUnpredicatedMveVpredNOp(MIB);
} else
llvm_unreachable("Unknown reg class!");
break;
case 24:
if (ARM::DTripleRegClass.hasSubClassEq(RC)) {
// Use aligned spills if the stack can be realigned.
if (Align >= 16 && getRegisterInfo().canRealignStack(MF) &&
Subtarget.hasNEON()) {
BuildMI(MBB, I, DebugLoc(), get(ARM::VST1d64TPseudo))
.addFrameIndex(FI)
.addImm(16)
.addReg(SrcReg, getKillRegState(isKill))
.addMemOperand(MMO)
.add(predOps(ARMCC::AL));
} else {
MachineInstrBuilder MIB = BuildMI(MBB, I, DebugLoc(),
get(ARM::VSTMDIA))
.addFrameIndex(FI)
.add(predOps(ARMCC::AL))
.addMemOperand(MMO);
MIB = AddDReg(MIB, SrcReg, ARM::dsub_0, getKillRegState(isKill), TRI);
MIB = AddDReg(MIB, SrcReg, ARM::dsub_1, 0, TRI);
AddDReg(MIB, SrcReg, ARM::dsub_2, 0, TRI);
}
} else
llvm_unreachable("Unknown reg class!");
break;
case 32:
if (ARM::QQPRRegClass.hasSubClassEq(RC) || ARM::DQuadRegClass.hasSubClassEq(RC)) {
if (Align >= 16 && getRegisterInfo().canRealignStack(MF) &&
Subtarget.hasNEON()) {
// FIXME: It's possible to only store part of the QQ register if the
// spilled def has a sub-register index.
BuildMI(MBB, I, DebugLoc(), get(ARM::VST1d64QPseudo))
.addFrameIndex(FI)
.addImm(16)
.addReg(SrcReg, getKillRegState(isKill))
.addMemOperand(MMO)
.add(predOps(ARMCC::AL));
} else {
MachineInstrBuilder MIB = BuildMI(MBB, I, DebugLoc(),
get(ARM::VSTMDIA))
.addFrameIndex(FI)
.add(predOps(ARMCC::AL))
.addMemOperand(MMO);
MIB = AddDReg(MIB, SrcReg, ARM::dsub_0, getKillRegState(isKill), TRI);
MIB = AddDReg(MIB, SrcReg, ARM::dsub_1, 0, TRI);
MIB = AddDReg(MIB, SrcReg, ARM::dsub_2, 0, TRI);
AddDReg(MIB, SrcReg, ARM::dsub_3, 0, TRI);
}
} else
llvm_unreachable("Unknown reg class!");
break;
case 64:
if (ARM::QQQQPRRegClass.hasSubClassEq(RC)) {
MachineInstrBuilder MIB = BuildMI(MBB, I, DebugLoc(), get(ARM::VSTMDIA))
.addFrameIndex(FI)
.add(predOps(ARMCC::AL))
.addMemOperand(MMO);
MIB = AddDReg(MIB, SrcReg, ARM::dsub_0, getKillRegState(isKill), TRI);
MIB = AddDReg(MIB, SrcReg, ARM::dsub_1, 0, TRI);
MIB = AddDReg(MIB, SrcReg, ARM::dsub_2, 0, TRI);
MIB = AddDReg(MIB, SrcReg, ARM::dsub_3, 0, TRI);
MIB = AddDReg(MIB, SrcReg, ARM::dsub_4, 0, TRI);
MIB = AddDReg(MIB, SrcReg, ARM::dsub_5, 0, TRI);
MIB = AddDReg(MIB, SrcReg, ARM::dsub_6, 0, TRI);
AddDReg(MIB, SrcReg, ARM::dsub_7, 0, TRI);
} else
llvm_unreachable("Unknown reg class!");
break;
default:
llvm_unreachable("Unknown reg class!");
}
}
unsigned ARMBaseInstrInfo::isStoreToStackSlot(const MachineInstr &MI,
int &FrameIndex) const {
switch (MI.getOpcode()) {
default: break;
case ARM::STRrs:
case ARM::t2STRs: // FIXME: don't use t2STRs to access frame.
if (MI.getOperand(1).isFI() && MI.getOperand(2).isReg() &&
MI.getOperand(3).isImm() && MI.getOperand(2).getReg() == 0 &&
MI.getOperand(3).getImm() == 0) {
FrameIndex = MI.getOperand(1).getIndex();
return MI.getOperand(0).getReg();
}
break;
case ARM::STRi12:
case ARM::t2STRi12:
case ARM::tSTRspi:
case ARM::VSTRD:
case ARM::VSTRS:
if (MI.getOperand(1).isFI() && MI.getOperand(2).isImm() &&
MI.getOperand(2).getImm() == 0) {
FrameIndex = MI.getOperand(1).getIndex();
return MI.getOperand(0).getReg();
}
break;
case ARM::VSTR_P0_off:
if (MI.getOperand(0).isFI() && MI.getOperand(1).isImm() &&
MI.getOperand(1).getImm() == 0) {
FrameIndex = MI.getOperand(0).getIndex();
return ARM::P0;
}
break;
case ARM::VST1q64:
case ARM::VST1d64TPseudo:
case ARM::VST1d64QPseudo:
if (MI.getOperand(0).isFI() && MI.getOperand(2).getSubReg() == 0) {
FrameIndex = MI.getOperand(0).getIndex();
return MI.getOperand(2).getReg();
}
break;
case ARM::VSTMQIA:
if (MI.getOperand(1).isFI() && MI.getOperand(0).getSubReg() == 0) {
FrameIndex = MI.getOperand(1).getIndex();
return MI.getOperand(0).getReg();
}
break;
}
return 0;
}
unsigned ARMBaseInstrInfo::isStoreToStackSlotPostFE(const MachineInstr &MI,
int &FrameIndex) const {
SmallVector<const MachineMemOperand *, 1> Accesses;
if (MI.mayStore() && hasStoreToStackSlot(MI, Accesses) &&
Accesses.size() == 1) {
FrameIndex =
cast<FixedStackPseudoSourceValue>(Accesses.front()->getPseudoValue())
->getFrameIndex();
return true;
}
return false;
}
void ARMBaseInstrInfo::
loadRegFromStackSlot(MachineBasicBlock &MBB, MachineBasicBlock::iterator I,
unsigned DestReg, int FI,
const TargetRegisterClass *RC,
const TargetRegisterInfo *TRI) const {
DebugLoc DL;
if (I != MBB.end()) DL = I->getDebugLoc();
MachineFunction &MF = *MBB.getParent();
MachineFrameInfo &MFI = MF.getFrameInfo();
unsigned Align = MFI.getObjectAlignment(FI);
MachineMemOperand *MMO = MF.getMachineMemOperand(
MachinePointerInfo::getFixedStack(MF, FI), MachineMemOperand::MOLoad,
MFI.getObjectSize(FI), Align);
switch (TRI->getSpillSize(*RC)) {
case 2:
if (ARM::HPRRegClass.hasSubClassEq(RC)) {
BuildMI(MBB, I, DL, get(ARM::VLDRH), DestReg)
.addFrameIndex(FI)
.addImm(0)
.addMemOperand(MMO)
.add(predOps(ARMCC::AL));
} else
llvm_unreachable("Unknown reg class!");
break;
case 4:
if (ARM::GPRRegClass.hasSubClassEq(RC)) {
BuildMI(MBB, I, DL, get(ARM::LDRi12), DestReg)
.addFrameIndex(FI)
.addImm(0)
.addMemOperand(MMO)
.add(predOps(ARMCC::AL));
} else if (ARM::SPRRegClass.hasSubClassEq(RC)) {
BuildMI(MBB, I, DL, get(ARM::VLDRS), DestReg)
.addFrameIndex(FI)
.addImm(0)
.addMemOperand(MMO)
.add(predOps(ARMCC::AL));
} else if (ARM::VCCRRegClass.hasSubClassEq(RC)) {
BuildMI(MBB, I, DL, get(ARM::VLDR_P0_off), DestReg)
.addFrameIndex(FI)
.addImm(0)
.addMemOperand(MMO)
.add(predOps(ARMCC::AL));
} else
llvm_unreachable("Unknown reg class!");
break;
case 8:
if (ARM::DPRRegClass.hasSubClassEq(RC)) {
BuildMI(MBB, I, DL, get(ARM::VLDRD), DestReg)
.addFrameIndex(FI)
.addImm(0)
.addMemOperand(MMO)
.add(predOps(ARMCC::AL));
} else if (ARM::GPRPairRegClass.hasSubClassEq(RC)) {
MachineInstrBuilder MIB;
if (Subtarget.hasV5TEOps()) {
MIB = BuildMI(MBB, I, DL, get(ARM::LDRD));
AddDReg(MIB, DestReg, ARM::gsub_0, RegState::DefineNoRead, TRI);
AddDReg(MIB, DestReg, ARM::gsub_1, RegState::DefineNoRead, TRI);
MIB.addFrameIndex(FI).addReg(0).addImm(0).addMemOperand(MMO)
.add(predOps(ARMCC::AL));
} else {
// Fallback to LDM instruction, which has existed since the dawn of
// time.
MIB = BuildMI(MBB, I, DL, get(ARM::LDMIA))
.addFrameIndex(FI)
.addMemOperand(MMO)
.add(predOps(ARMCC::AL));
MIB = AddDReg(MIB, DestReg, ARM::gsub_0, RegState::DefineNoRead, TRI);
MIB = AddDReg(MIB, DestReg, ARM::gsub_1, RegState::DefineNoRead, TRI);
}
if (Register::isPhysicalRegister(DestReg))
MIB.addReg(DestReg, RegState::ImplicitDefine);
} else
llvm_unreachable("Unknown reg class!");
break;
case 16:
if (ARM::DPairRegClass.hasSubClassEq(RC) && Subtarget.hasNEON()) {
if (Align >= 16 && getRegisterInfo().canRealignStack(MF)) {
BuildMI(MBB, I, DL, get(ARM::VLD1q64), DestReg)
.addFrameIndex(FI)
.addImm(16)
.addMemOperand(MMO)
.add(predOps(ARMCC::AL));
} else {
BuildMI(MBB, I, DL, get(ARM::VLDMQIA), DestReg)
.addFrameIndex(FI)
.addMemOperand(MMO)
.add(predOps(ARMCC::AL));
}
} else if (ARM::QPRRegClass.hasSubClassEq(RC) &&
Subtarget.hasMVEIntegerOps()) {
auto MIB = BuildMI(MBB, I, DL, get(ARM::MVE_VLDRWU32), DestReg);
MIB.addFrameIndex(FI)
.addImm(0)
.addMemOperand(MMO);
addUnpredicatedMveVpredNOp(MIB);
} else
llvm_unreachable("Unknown reg class!");
break;
case 24:
if (ARM::DTripleRegClass.hasSubClassEq(RC)) {
if (Align >= 16 && getRegisterInfo().canRealignStack(MF) &&
Subtarget.hasNEON()) {
BuildMI(MBB, I, DL, get(ARM::VLD1d64TPseudo), DestReg)
.addFrameIndex(FI)
.addImm(16)
.addMemOperand(MMO)
.add(predOps(ARMCC::AL));
} else {
MachineInstrBuilder MIB = BuildMI(MBB, I, DL, get(ARM::VLDMDIA))
.addFrameIndex(FI)
.addMemOperand(MMO)
.add(predOps(ARMCC::AL));
MIB = AddDReg(MIB, DestReg, ARM::dsub_0, RegState::DefineNoRead, TRI);
MIB = AddDReg(MIB, DestReg, ARM::dsub_1, RegState::DefineNoRead, TRI);
MIB = AddDReg(MIB, DestReg, ARM::dsub_2, RegState::DefineNoRead, TRI);
if (Register::isPhysicalRegister(DestReg))
MIB.addReg(DestReg, RegState::ImplicitDefine);
}
} else
llvm_unreachable("Unknown reg class!");
break;
case 32:
if (ARM::QQPRRegClass.hasSubClassEq(RC) || ARM::DQuadRegClass.hasSubClassEq(RC)) {
if (Align >= 16 && getRegisterInfo().canRealignStack(MF) &&
Subtarget.hasNEON()) {
BuildMI(MBB, I, DL, get(ARM::VLD1d64QPseudo), DestReg)
.addFrameIndex(FI)
.addImm(16)
.addMemOperand(MMO)
.add(predOps(ARMCC::AL));
} else {
MachineInstrBuilder MIB = BuildMI(MBB, I, DL, get(ARM::VLDMDIA))
.addFrameIndex(FI)
.add(predOps(ARMCC::AL))
.addMemOperand(MMO);
MIB = AddDReg(MIB, DestReg, ARM::dsub_0, RegState::DefineNoRead, TRI);
MIB = AddDReg(MIB, DestReg, ARM::dsub_1, RegState::DefineNoRead, TRI);
MIB = AddDReg(MIB, DestReg, ARM::dsub_2, RegState::DefineNoRead, TRI);
MIB = AddDReg(MIB, DestReg, ARM::dsub_3, RegState::DefineNoRead, TRI);
if (Register::isPhysicalRegister(DestReg))
MIB.addReg(DestReg, RegState::ImplicitDefine);
}
} else
llvm_unreachable("Unknown reg class!");
break;
case 64:
if (ARM::QQQQPRRegClass.hasSubClassEq(RC)) {
MachineInstrBuilder MIB = BuildMI(MBB, I, DL, get(ARM::VLDMDIA))
.addFrameIndex(FI)
.add(predOps(ARMCC::AL))
.addMemOperand(MMO);
MIB = AddDReg(MIB, DestReg, ARM::dsub_0, RegState::DefineNoRead, TRI);
MIB = AddDReg(MIB, DestReg, ARM::dsub_1, RegState::DefineNoRead, TRI);
MIB = AddDReg(MIB, DestReg, ARM::dsub_2, RegState::DefineNoRead, TRI);
MIB = AddDReg(MIB, DestReg, ARM::dsub_3, RegState::DefineNoRead, TRI);
MIB = AddDReg(MIB, DestReg, ARM::dsub_4, RegState::DefineNoRead, TRI);
MIB = AddDReg(MIB, DestReg, ARM::dsub_5, RegState::DefineNoRead, TRI);
MIB = AddDReg(MIB, DestReg, ARM::dsub_6, RegState::DefineNoRead, TRI);
MIB = AddDReg(MIB, DestReg, ARM::dsub_7, RegState::DefineNoRead, TRI);
if (Register::isPhysicalRegister(DestReg))
MIB.addReg(DestReg, RegState::ImplicitDefine);
} else
llvm_unreachable("Unknown reg class!");
break;
default:
llvm_unreachable("Unknown regclass!");
}
}
unsigned ARMBaseInstrInfo::isLoadFromStackSlot(const MachineInstr &MI,
int &FrameIndex) const {
switch (MI.getOpcode()) {
default: break;
case ARM::LDRrs:
case ARM::t2LDRs: // FIXME: don't use t2LDRs to access frame.
if (MI.getOperand(1).isFI() && MI.getOperand(2).isReg() &&
MI.getOperand(3).isImm() && MI.getOperand(2).getReg() == 0 &&
MI.getOperand(3).getImm() == 0) {
FrameIndex = MI.getOperand(1).getIndex();
return MI.getOperand(0).getReg();
}
break;
case ARM::LDRi12:
case ARM::t2LDRi12:
case ARM::tLDRspi:
case ARM::VLDRD:
case ARM::VLDRS:
if (MI.getOperand(1).isFI() && MI.getOperand(2).isImm() &&
MI.getOperand(2).getImm() == 0) {
FrameIndex = MI.getOperand(1).getIndex();
return MI.getOperand(0).getReg();
}
break;
case ARM::VLDR_P0_off:
if (MI.getOperand(0).isFI() && MI.getOperand(1).isImm() &&
MI.getOperand(1).getImm() == 0) {
FrameIndex = MI.getOperand(0).getIndex();
return ARM::P0;
}
break;
case ARM::VLD1q64:
case ARM::VLD1d8TPseudo:
case ARM::VLD1d16TPseudo:
case ARM::VLD1d32TPseudo:
case ARM::VLD1d64TPseudo:
case ARM::VLD1d8QPseudo:
case ARM::VLD1d16QPseudo:
case ARM::VLD1d32QPseudo:
case ARM::VLD1d64QPseudo:
if (MI.getOperand(1).isFI() && MI.getOperand(0).getSubReg() == 0) {
FrameIndex = MI.getOperand(1).getIndex();
return MI.getOperand(0).getReg();
}
break;
case ARM::VLDMQIA:
if (MI.getOperand(1).isFI() && MI.getOperand(0).getSubReg() == 0) {
FrameIndex = MI.getOperand(1).getIndex();
return MI.getOperand(0).getReg();
}
break;
}
return 0;
}
unsigned ARMBaseInstrInfo::isLoadFromStackSlotPostFE(const MachineInstr &MI,
int &FrameIndex) const {
SmallVector<const MachineMemOperand *, 1> Accesses;
if (MI.mayLoad() && hasLoadFromStackSlot(MI, Accesses) &&
Accesses.size() == 1) {
FrameIndex =
cast<FixedStackPseudoSourceValue>(Accesses.front()->getPseudoValue())
->getFrameIndex();
return true;
}
return false;
}
/// Expands MEMCPY to either LDMIA/STMIA or LDMIA_UPD/STMID_UPD
/// depending on whether the result is used.
void ARMBaseInstrInfo::expandMEMCPY(MachineBasicBlock::iterator MI) const {
bool isThumb1 = Subtarget.isThumb1Only();
bool isThumb2 = Subtarget.isThumb2();
const ARMBaseInstrInfo *TII = Subtarget.getInstrInfo();
DebugLoc dl = MI->getDebugLoc();
MachineBasicBlock *BB = MI->getParent();
MachineInstrBuilder LDM, STM;
if (isThumb1 || !MI->getOperand(1).isDead()) {
MachineOperand LDWb(MI->getOperand(1));
LDM = BuildMI(*BB, MI, dl, TII->get(isThumb2 ? ARM::t2LDMIA_UPD
: isThumb1 ? ARM::tLDMIA_UPD
: ARM::LDMIA_UPD))
.add(LDWb);
} else {
LDM = BuildMI(*BB, MI, dl, TII->get(isThumb2 ? ARM::t2LDMIA : ARM::LDMIA));
}
if (isThumb1 || !MI->getOperand(0).isDead()) {
MachineOperand STWb(MI->getOperand(0));
STM = BuildMI(*BB, MI, dl, TII->get(isThumb2 ? ARM::t2STMIA_UPD
: isThumb1 ? ARM::tSTMIA_UPD
: ARM::STMIA_UPD))
.add(STWb);
} else {
STM = BuildMI(*BB, MI, dl, TII->get(isThumb2 ? ARM::t2STMIA : ARM::STMIA));
}
MachineOperand LDBase(MI->getOperand(3));
LDM.add(LDBase).add(predOps(ARMCC::AL));
MachineOperand STBase(MI->getOperand(2));
STM.add(STBase).add(predOps(ARMCC::AL));
// Sort the scratch registers into ascending order.
const TargetRegisterInfo &TRI = getRegisterInfo();
SmallVector<unsigned, 6> ScratchRegs;
for(unsigned I = 5; I < MI->getNumOperands(); ++I)
ScratchRegs.push_back(MI->getOperand(I).getReg());
llvm::sort(ScratchRegs,
[&TRI](const unsigned &Reg1, const unsigned &Reg2) -> bool {
return TRI.getEncodingValue(Reg1) <
TRI.getEncodingValue(Reg2);
});
for (const auto &Reg : ScratchRegs) {
LDM.addReg(Reg, RegState::Define);
STM.addReg(Reg, RegState::Kill);
}
BB->erase(MI);
}
bool ARMBaseInstrInfo::expandPostRAPseudo(MachineInstr &MI) const {
if (MI.getOpcode() == TargetOpcode::LOAD_STACK_GUARD) {
assert(getSubtarget().getTargetTriple().isOSBinFormatMachO() &&
"LOAD_STACK_GUARD currently supported only for MachO.");
expandLoadStackGuard(MI);
MI.getParent()->erase(MI);
return true;
}
if (MI.getOpcode() == ARM::MEMCPY) {
expandMEMCPY(MI);
return true;
}
// This hook gets to expand COPY instructions before they become
// copyPhysReg() calls. Look for VMOVS instructions that can legally be
// widened to VMOVD. We prefer the VMOVD when possible because it may be
// changed into a VORR that can go down the NEON pipeline.
if (!MI.isCopy() || Subtarget.dontWidenVMOVS() || !Subtarget.hasFP64())
return false;
// Look for a copy between even S-registers. That is where we keep floats
// when using NEON v2f32 instructions for f32 arithmetic.
Register DstRegS = MI.getOperand(0).getReg();
Register SrcRegS = MI.getOperand(1).getReg();
if (!ARM::SPRRegClass.contains(DstRegS, SrcRegS))
return false;
const TargetRegisterInfo *TRI = &getRegisterInfo();
unsigned DstRegD = TRI->getMatchingSuperReg(DstRegS, ARM::ssub_0,
&ARM::DPRRegClass);
unsigned SrcRegD = TRI->getMatchingSuperReg(SrcRegS, ARM::ssub_0,
&ARM::DPRRegClass);
if (!DstRegD || !SrcRegD)
return false;
// We want to widen this into a DstRegD = VMOVD SrcRegD copy. This is only
// legal if the COPY already defines the full DstRegD, and it isn't a
// sub-register insertion.
if (!MI.definesRegister(DstRegD, TRI) || MI.readsRegister(DstRegD, TRI))
return false;
// A dead copy shouldn't show up here, but reject it just in case.
if (MI.getOperand(0).isDead())
return false;
// All clear, widen the COPY.
LLVM_DEBUG(dbgs() << "widening: " << MI);
MachineInstrBuilder MIB(*MI.getParent()->getParent(), MI);
// Get rid of the old implicit-def of DstRegD. Leave it if it defines a Q-reg
// or some other super-register.
int ImpDefIdx = MI.findRegisterDefOperandIdx(DstRegD);
if (ImpDefIdx != -1)
MI.RemoveOperand(ImpDefIdx);
// Change the opcode and operands.
MI.setDesc(get(ARM::VMOVD));
MI.getOperand(0).setReg(DstRegD);
MI.getOperand(1).setReg(SrcRegD);
MIB.add(predOps(ARMCC::AL));
// We are now reading SrcRegD instead of SrcRegS. This may upset the
// register scavenger and machine verifier, so we need to indicate that we
// are reading an undefined value from SrcRegD, but a proper value from
// SrcRegS.
MI.getOperand(1).setIsUndef();
MIB.addReg(SrcRegS, RegState::Implicit);
// SrcRegD may actually contain an unrelated value in the ssub_1
// sub-register. Don't kill it. Only kill the ssub_0 sub-register.
if (MI.getOperand(1).isKill()) {
MI.getOperand(1).setIsKill(false);
MI.addRegisterKilled(SrcRegS, TRI, true);
}
LLVM_DEBUG(dbgs() << "replaced by: " << MI);
return true;
}
/// Create a copy of a const pool value. Update CPI to the new index and return
/// the label UID.
static unsigned duplicateCPV(MachineFunction &MF, unsigned &CPI) {
MachineConstantPool *MCP = MF.getConstantPool();
ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
const MachineConstantPoolEntry &MCPE = MCP->getConstants()[CPI];
assert(MCPE.isMachineConstantPoolEntry() &&
"Expecting a machine constantpool entry!");
ARMConstantPoolValue *ACPV =
static_cast<ARMConstantPoolValue*>(MCPE.Val.MachineCPVal);
unsigned PCLabelId = AFI->createPICLabelUId();
ARMConstantPoolValue *NewCPV = nullptr;
// FIXME: The below assumes PIC relocation model and that the function
// is Thumb mode (t1 or t2). PCAdjustment would be 8 for ARM mode PIC, and
// zero for non-PIC in ARM or Thumb. The callers are all of thumb LDR
// instructions, so that's probably OK, but is PIC always correct when
// we get here?
if (ACPV->isGlobalValue())
NewCPV = ARMConstantPoolConstant::Create(
cast<ARMConstantPoolConstant>(ACPV)->getGV(), PCLabelId, ARMCP::CPValue,
4, ACPV->getModifier(), ACPV->mustAddCurrentAddress());
else if (ACPV->isExtSymbol())
NewCPV = ARMConstantPoolSymbol::
Create(MF.getFunction().getContext(),
cast<ARMConstantPoolSymbol>(ACPV)->getSymbol(), PCLabelId, 4);
else if (ACPV->isBlockAddress())
NewCPV = ARMConstantPoolConstant::
Create(cast<ARMConstantPoolConstant>(ACPV)->getBlockAddress(), PCLabelId,
ARMCP::CPBlockAddress, 4);
else if (ACPV->isLSDA())
NewCPV = ARMConstantPoolConstant::Create(&MF.getFunction(), PCLabelId,
ARMCP::CPLSDA, 4);
else if (ACPV->isMachineBasicBlock())
NewCPV = ARMConstantPoolMBB::
Create(MF.getFunction().getContext(),
cast<ARMConstantPoolMBB>(ACPV)->getMBB(), PCLabelId, 4);
else
llvm_unreachable("Unexpected ARM constantpool value type!!");
CPI = MCP->getConstantPoolIndex(NewCPV, MCPE.getAlignment());
return PCLabelId;
}
void ARMBaseInstrInfo::reMaterialize(MachineBasicBlock &MBB,
MachineBasicBlock::iterator I,
unsigned DestReg, unsigned SubIdx,
const MachineInstr &Orig,
const TargetRegisterInfo &TRI) const {
unsigned Opcode = Orig.getOpcode();
switch (Opcode) {
default: {
MachineInstr *MI = MBB.getParent()->CloneMachineInstr(&Orig);
MI->substituteRegister(Orig.getOperand(0).getReg(), DestReg, SubIdx, TRI);
MBB.insert(I, MI);
break;
}
case ARM::tLDRpci_pic:
case ARM::t2LDRpci_pic: {
MachineFunction &MF = *MBB.getParent();
unsigned CPI = Orig.getOperand(1).getIndex();
unsigned PCLabelId = duplicateCPV(MF, CPI);
BuildMI(MBB, I, Orig.getDebugLoc(), get(Opcode), DestReg)
.addConstantPoolIndex(CPI)
.addImm(PCLabelId)
.cloneMemRefs(Orig);
break;
}
}
}
MachineInstr &
ARMBaseInstrInfo::duplicate(MachineBasicBlock &MBB,
MachineBasicBlock::iterator InsertBefore,
const MachineInstr &Orig) const {
MachineInstr &Cloned = TargetInstrInfo::duplicate(MBB, InsertBefore, Orig);
MachineBasicBlock::instr_iterator I = Cloned.getIterator();
for (;;) {
switch (I->getOpcode()) {
case ARM::tLDRpci_pic:
case ARM::t2LDRpci_pic: {
MachineFunction &MF = *MBB.getParent();
unsigned CPI = I->getOperand(1).getIndex();
unsigned PCLabelId = duplicateCPV(MF, CPI);
I->getOperand(1).setIndex(CPI);
I->getOperand(2).setImm(PCLabelId);
break;
}
}
if (!I->isBundledWithSucc())
break;
++I;
}
return Cloned;
}
bool ARMBaseInstrInfo::produceSameValue(const MachineInstr &MI0,
const MachineInstr &MI1,
const MachineRegisterInfo *MRI) const {
unsigned Opcode = MI0.getOpcode();
if (Opcode == ARM::t2LDRpci ||
Opcode == ARM::t2LDRpci_pic ||
Opcode == ARM::tLDRpci ||
Opcode == ARM::tLDRpci_pic ||
Opcode == ARM::LDRLIT_ga_pcrel ||
Opcode == ARM::LDRLIT_ga_pcrel_ldr ||
Opcode == ARM::tLDRLIT_ga_pcrel ||
Opcode == ARM::MOV_ga_pcrel ||
Opcode == ARM::MOV_ga_pcrel_ldr ||
Opcode == ARM::t2MOV_ga_pcrel) {
if (MI1.getOpcode() != Opcode)
return false;
if (MI0.getNumOperands() != MI1.getNumOperands())
return false;
const MachineOperand &MO0 = MI0.getOperand(1);
const MachineOperand &MO1 = MI1.getOperand(1);
if (MO0.getOffset() != MO1.getOffset())
return false;
if (Opcode == ARM::LDRLIT_ga_pcrel ||
Opcode == ARM::LDRLIT_ga_pcrel_ldr ||
Opcode == ARM::tLDRLIT_ga_pcrel ||
Opcode == ARM::MOV_ga_pcrel ||
Opcode == ARM::MOV_ga_pcrel_ldr ||
Opcode == ARM::t2MOV_ga_pcrel)
// Ignore the PC labels.
return MO0.getGlobal() == MO1.getGlobal();
const MachineFunction *MF = MI0.getParent()->getParent();
const MachineConstantPool *MCP = MF->getConstantPool();
int CPI0 = MO0.getIndex();
int CPI1 = MO1.getIndex();
const MachineConstantPoolEntry &MCPE0 = MCP->getConstants()[CPI0];
const MachineConstantPoolEntry &MCPE1 = MCP->getConstants()[CPI1];
bool isARMCP0 = MCPE0.isMachineConstantPoolEntry();
bool isARMCP1 = MCPE1.isMachineConstantPoolEntry();
if (isARMCP0 && isARMCP1) {
ARMConstantPoolValue *ACPV0 =
static_cast<ARMConstantPoolValue*>(MCPE0.Val.MachineCPVal);
ARMConstantPoolValue *ACPV1 =
static_cast<ARMConstantPoolValue*>(MCPE1.Val.MachineCPVal);
return ACPV0->hasSameValue(ACPV1);
} else if (!isARMCP0 && !isARMCP1) {
return MCPE0.Val.ConstVal == MCPE1.Val.ConstVal;
}
return false;
} else if (Opcode == ARM::PICLDR) {
if (MI1.getOpcode() != Opcode)
return false;
if (MI0.getNumOperands() != MI1.getNumOperands())
return false;
Register Addr0 = MI0.getOperand(1).getReg();
Register Addr1 = MI1.getOperand(1).getReg();
if (Addr0 != Addr1) {
if (!MRI || !Register::isVirtualRegister(Addr0) ||
!Register::isVirtualRegister(Addr1))
return false;
// This assumes SSA form.
MachineInstr *Def0 = MRI->getVRegDef(Addr0);
MachineInstr *Def1 = MRI->getVRegDef(Addr1);
// Check if the loaded value, e.g. a constantpool of a global address, are
// the same.
if (!produceSameValue(*Def0, *Def1, MRI))
return false;
}
for (unsigned i = 3, e = MI0.getNumOperands(); i != e; ++i) {
// %12 = PICLDR %11, 0, 14, %noreg
const MachineOperand &MO0 = MI0.getOperand(i);
const MachineOperand &MO1 = MI1.getOperand(i);
if (!MO0.isIdenticalTo(MO1))
return false;
}
return true;
}
return MI0.isIdenticalTo(MI1, MachineInstr::IgnoreVRegDefs);
}
/// areLoadsFromSameBasePtr - This is used by the pre-regalloc scheduler to
/// determine if two loads are loading from the same base address. It should
/// only return true if the base pointers are the same and the only differences
/// between the two addresses is the offset. It also returns the offsets by
/// reference.
///
/// FIXME: remove this in favor of the MachineInstr interface once pre-RA-sched
/// is permanently disabled.
bool ARMBaseInstrInfo::areLoadsFromSameBasePtr(SDNode *Load1, SDNode *Load2,
int64_t &Offset1,
int64_t &Offset2) const {
// Don't worry about Thumb: just ARM and Thumb2.
if (Subtarget.isThumb1Only()) return false;
if (!Load1->isMachineOpcode() || !Load2->isMachineOpcode())
return false;
switch (Load1->getMachineOpcode()) {
default:
return false;
case ARM::LDRi12:
case ARM::LDRBi12:
case ARM::LDRD:
case ARM::LDRH:
case ARM::LDRSB:
case ARM::LDRSH:
case ARM::VLDRD:
case ARM::VLDRS:
case ARM::t2LDRi8:
case ARM::t2LDRBi8:
case ARM::t2LDRDi8:
case ARM::t2LDRSHi8:
case ARM::t2LDRi12:
case ARM::t2LDRBi12:
case ARM::t2LDRSHi12:
break;
}
switch (Load2->getMachineOpcode()) {
default:
return false;
case ARM::LDRi12:
case ARM::LDRBi12:
case ARM::LDRD:
case ARM::LDRH:
case ARM::LDRSB:
case ARM::LDRSH:
case ARM::VLDRD:
case ARM::VLDRS:
case ARM::t2LDRi8:
case ARM::t2LDRBi8:
case ARM::t2LDRSHi8:
case ARM::t2LDRi12:
case ARM::t2LDRBi12:
case ARM::t2LDRSHi12:
break;
}
// Check if base addresses and chain operands match.
if (Load1->getOperand(0) != Load2->getOperand(0) ||
Load1->getOperand(4) != Load2->getOperand(4))
return false;
// Index should be Reg0.
if (Load1->getOperand(3) != Load2->getOperand(3))
return false;
// Determine the offsets.
if (isa<ConstantSDNode>(Load1->getOperand(1)) &&
isa<ConstantSDNode>(Load2->getOperand(1))) {
Offset1 = cast<ConstantSDNode>(Load1->getOperand(1))->getSExtValue();
Offset2 = cast<ConstantSDNode>(Load2->getOperand(1))->getSExtValue();
return true;
}
return false;
}
/// shouldScheduleLoadsNear - This is a used by the pre-regalloc scheduler to
/// determine (in conjunction with areLoadsFromSameBasePtr) if two loads should
/// be scheduled togther. On some targets if two loads are loading from
/// addresses in the same cache line, it's better if they are scheduled
/// together. This function takes two integers that represent the load offsets
/// from the common base address. It returns true if it decides it's desirable
/// to schedule the two loads together. "NumLoads" is the number of loads that
/// have already been scheduled after Load1.
///
/// FIXME: remove this in favor of the MachineInstr interface once pre-RA-sched
/// is permanently disabled.
bool ARMBaseInstrInfo::shouldScheduleLoadsNear(SDNode *Load1, SDNode *Load2,
int64_t Offset1, int64_t Offset2,
unsigned NumLoads) const {
// Don't worry about Thumb: just ARM and Thumb2.
if (Subtarget.isThumb1Only()) return false;
assert(Offset2 > Offset1);
if ((Offset2 - Offset1) / 8 > 64)
return false;
// Check if the machine opcodes are different. If they are different
// then we consider them to not be of the same base address,
// EXCEPT in the case of Thumb2 byte loads where one is LDRBi8 and the other LDRBi12.
// In this case, they are considered to be the same because they are different
// encoding forms of the same basic instruction.
if ((Load1->getMachineOpcode() != Load2->getMachineOpcode()) &&
!((Load1->getMachineOpcode() == ARM::t2LDRBi8 &&
Load2->getMachineOpcode() == ARM::t2LDRBi12) ||
(Load1->getMachineOpcode() == ARM::t2LDRBi12 &&
Load2->getMachineOpcode() == ARM::t2LDRBi8)))
return false; // FIXME: overly conservative?
// Four loads in a row should be sufficient.
if (NumLoads >= 3)
return false;
return true;
}
bool ARMBaseInstrInfo::isSchedulingBoundary(const MachineInstr &MI,
const MachineBasicBlock *MBB,
const MachineFunction &MF) const {
// Debug info is never a scheduling boundary. It's necessary to be explicit
// due to the special treatment of IT instructions below, otherwise a
// dbg_value followed by an IT will result in the IT instruction being
// considered a scheduling hazard, which is wrong. It should be the actual
// instruction preceding the dbg_value instruction(s), just like it is
// when debug info is not present.
if (MI.isDebugInstr())
return false;
// Terminators and labels can't be scheduled around.
if (MI.isTerminator() || MI.isPosition())
return true;
// Treat the start of the IT block as a scheduling boundary, but schedule
// t2IT along with all instructions following it.
// FIXME: This is a big hammer. But the alternative is to add all potential
// true and anti dependencies to IT block instructions as implicit operands
// to the t2IT instruction. The added compile time and complexity does not
// seem worth it.
MachineBasicBlock::const_iterator I = MI;
// Make sure to skip any debug instructions
while (++I != MBB->end() && I->isDebugInstr())
;
if (I != MBB->end() && I->getOpcode() == ARM::t2IT)
return true;
// Don't attempt to schedule around any instruction that defines
// a stack-oriented pointer, as it's unlikely to be profitable. This
// saves compile time, because it doesn't require every single
// stack slot reference to depend on the instruction that does the
// modification.
// Calls don't actually change the stack pointer, even if they have imp-defs.
// No ARM calling conventions change the stack pointer. (X86 calling
// conventions sometimes do).
if (!MI.isCall() && MI.definesRegister(ARM::SP))
return true;
return false;
}
bool ARMBaseInstrInfo::
isProfitableToIfCvt(MachineBasicBlock &MBB,
unsigned NumCycles, unsigned ExtraPredCycles,
BranchProbability Probability) const {
if (!NumCycles)
return false;
// If we are optimizing for size, see if the branch in the predecessor can be
// lowered to cbn?z by the constant island lowering pass, and return false if
// so. This results in a shorter instruction sequence.
if (MBB.getParent()->getFunction().hasOptSize()) {
MachineBasicBlock *Pred = *MBB.pred_begin();
if (!Pred->empty()) {
MachineInstr *LastMI = &*Pred->rbegin();
if (LastMI->getOpcode() == ARM::t2Bcc) {
const TargetRegisterInfo *TRI = &getRegisterInfo();
MachineInstr *CmpMI = findCMPToFoldIntoCBZ(LastMI, TRI);
if (CmpMI)
return false;
}
}
}
return isProfitableToIfCvt(MBB, NumCycles, ExtraPredCycles,
MBB, 0, 0, Probability);
}
bool ARMBaseInstrInfo::
isProfitableToIfCvt(MachineBasicBlock &TBB,
unsigned TCycles, unsigned TExtra,
MachineBasicBlock &FBB,
unsigned FCycles, unsigned FExtra,
BranchProbability Probability) const {
if (!TCycles)
return false;
// In thumb code we often end up trading one branch for a IT block, and
// if we are cloning the instruction can increase code size. Prevent
// blocks with multiple predecesors from being ifcvted to prevent this
// cloning.
if (Subtarget.isThumb2() && TBB.getParent()->getFunction().hasMinSize()) {
if (TBB.pred_size() != 1 || FBB.pred_size() != 1)
return false;
}
// Attempt to estimate the relative costs of predication versus branching.
// Here we scale up each component of UnpredCost to avoid precision issue when
// scaling TCycles/FCycles by Probability.
const unsigned ScalingUpFactor = 1024;
unsigned PredCost = (TCycles + FCycles + TExtra + FExtra) * ScalingUpFactor;
unsigned UnpredCost;
if (!Subtarget.hasBranchPredictor()) {
// When we don't have a branch predictor it's always cheaper to not take a
// branch than take it, so we have to take that into account.
unsigned NotTakenBranchCost = 1;
unsigned TakenBranchCost = Subtarget.getMispredictionPenalty();
unsigned TUnpredCycles, FUnpredCycles;
if (!FCycles) {
// Triangle: TBB is the fallthrough
TUnpredCycles = TCycles + NotTakenBranchCost;
FUnpredCycles = TakenBranchCost;
} else {
// Diamond: TBB is the block that is branched to, FBB is the fallthrough
TUnpredCycles = TCycles + TakenBranchCost;
FUnpredCycles = FCycles + NotTakenBranchCost;
// The branch at the end of FBB will disappear when it's predicated, so
// discount it from PredCost.
PredCost -= 1 * ScalingUpFactor;
}
// The total cost is the cost of each path scaled by their probabilites
unsigned TUnpredCost = Probability.scale(TUnpredCycles * ScalingUpFactor);
unsigned FUnpredCost = Probability.getCompl().scale(FUnpredCycles * ScalingUpFactor);
UnpredCost = TUnpredCost + FUnpredCost;
// When predicating assume that the first IT can be folded away but later
// ones cost one cycle each
if (Subtarget.isThumb2() && TCycles + FCycles > 4) {
PredCost += ((TCycles + FCycles - 4) / 4) * ScalingUpFactor;
}
} else {
unsigned TUnpredCost = Probability.scale(TCycles * ScalingUpFactor);
unsigned FUnpredCost =
Probability.getCompl().scale(FCycles * ScalingUpFactor);
UnpredCost = TUnpredCost + FUnpredCost;
UnpredCost += 1 * ScalingUpFactor; // The branch itself
UnpredCost += Subtarget.getMispredictionPenalty() * ScalingUpFactor / 10;
}
return PredCost <= UnpredCost;
}
unsigned
ARMBaseInstrInfo::extraSizeToPredicateInstructions(const MachineFunction &MF,
unsigned NumInsts) const {
// Thumb2 needs a 2-byte IT instruction to predicate up to 4 instructions.
// ARM has a condition code field in every predicable instruction, using it
// doesn't change code size.
return Subtarget.isThumb2() ? divideCeil(NumInsts, 4) * 2 : 0;
}
unsigned
ARMBaseInstrInfo::predictBranchSizeForIfCvt(MachineInstr &MI) const {
// If this branch is likely to be folded into the comparison to form a
// CB(N)Z, then removing it won't reduce code size at all, because that will
// just replace the CB(N)Z with a CMP.
if (MI.getOpcode() == ARM::t2Bcc &&
findCMPToFoldIntoCBZ(&MI, &getRegisterInfo()))
return 0;
unsigned Size = getInstSizeInBytes(MI);
// For Thumb2, all branches are 32-bit instructions during the if conversion
// pass, but may be replaced with 16-bit instructions during size reduction.
// Since the branches considered by if conversion tend to be forward branches
// over small basic blocks, they are very likely to be in range for the
// narrow instructions, so we assume the final code size will be half what it
// currently is.
if (Subtarget.isThumb2())
Size /= 2;
return Size;
}
bool
ARMBaseInstrInfo::isProfitableToUnpredicate(MachineBasicBlock &TMBB,
MachineBasicBlock &FMBB) const {
// Reduce false anti-dependencies to let the target's out-of-order execution
// engine do its thing.
return Subtarget.isProfitableToUnpredicate();
}
/// getInstrPredicate - If instruction is predicated, returns its predicate
/// condition, otherwise returns AL. It also returns the condition code
/// register by reference.
ARMCC::CondCodes llvm::getInstrPredicate(const MachineInstr &MI,
unsigned &PredReg) {
int PIdx = MI.findFirstPredOperandIdx();
if (PIdx == -1) {
PredReg = 0;
return ARMCC::AL;
}
PredReg = MI.getOperand(PIdx+1).getReg();
return (ARMCC::CondCodes)MI.getOperand(PIdx).getImm();
}
unsigned llvm::getMatchingCondBranchOpcode(unsigned Opc) {
if (Opc == ARM::B)
return ARM::Bcc;
if (Opc == ARM::tB)
return ARM::tBcc;
if (Opc == ARM::t2B)
return ARM::t2Bcc;
llvm_unreachable("Unknown unconditional branch opcode!");
}
MachineInstr *ARMBaseInstrInfo::commuteInstructionImpl(MachineInstr &MI,
bool NewMI,
unsigned OpIdx1,
unsigned OpIdx2) const {
switch (MI.getOpcode()) {
case ARM::MOVCCr:
case ARM::t2MOVCCr: {
// MOVCC can be commuted by inverting the condition.
unsigned PredReg = 0;
ARMCC::CondCodes CC = getInstrPredicate(MI, PredReg);
// MOVCC AL can't be inverted. Shouldn't happen.
if (CC == ARMCC::AL || PredReg != ARM::CPSR)
return nullptr;
MachineInstr *CommutedMI =
TargetInstrInfo::commuteInstructionImpl(MI, NewMI, OpIdx1, OpIdx2);
if (!CommutedMI)
return nullptr;
// After swapping the MOVCC operands, also invert the condition.
CommutedMI->getOperand(CommutedMI->findFirstPredOperandIdx())
.setImm(ARMCC::getOppositeCondition(CC));
return CommutedMI;
}
}
return TargetInstrInfo::commuteInstructionImpl(MI, NewMI, OpIdx1, OpIdx2);
}
/// Identify instructions that can be folded into a MOVCC instruction, and
/// return the defining instruction.
MachineInstr *
ARMBaseInstrInfo::canFoldIntoMOVCC(unsigned Reg, const MachineRegisterInfo &MRI,
const TargetInstrInfo *TII) const {
if (!Register::isVirtualRegister(Reg))
return nullptr;
if (!MRI.hasOneNonDBGUse(Reg))
return nullptr;
MachineInstr *MI = MRI.getVRegDef(Reg);
if (!MI)
return nullptr;
// Check if MI can be predicated and folded into the MOVCC.
if (!isPredicable(*MI))
return nullptr;
// Check if MI has any non-dead defs or physreg uses. This also detects
// predicated instructions which will be reading CPSR.
for (unsigned i = 1, e = MI->getNumOperands(); i != e; ++i) {
const MachineOperand &MO = MI->getOperand(i);
// Reject frame index operands, PEI can't handle the predicated pseudos.
if (MO.isFI() || MO.isCPI() || MO.isJTI())
return nullptr;
if (!MO.isReg())
continue;
// MI can't have any tied operands, that would conflict with predication.
if (MO.isTied())
return nullptr;
if (Register::isPhysicalRegister(MO.getReg()))
return nullptr;
if (MO.isDef() && !MO.isDead())
return nullptr;
}
bool DontMoveAcrossStores = true;
if (!MI->isSafeToMove(/* AliasAnalysis = */ nullptr, DontMoveAcrossStores))
return nullptr;
return MI;
}
bool ARMBaseInstrInfo::analyzeSelect(const MachineInstr &MI,
SmallVectorImpl<MachineOperand> &Cond,
unsigned &TrueOp, unsigned &FalseOp,
bool &Optimizable) const {
assert((MI.getOpcode() == ARM::MOVCCr || MI.getOpcode() == ARM::t2MOVCCr) &&
"Unknown select instruction");
// MOVCC operands:
// 0: Def.
// 1: True use.
// 2: False use.
// 3: Condition code.
// 4: CPSR use.
TrueOp = 1;
FalseOp = 2;
Cond.push_back(MI.getOperand(3));
Cond.push_back(MI.getOperand(4));
// We can always fold a def.
Optimizable = true;
return false;
}
MachineInstr *
ARMBaseInstrInfo::optimizeSelect(MachineInstr &MI,
SmallPtrSetImpl<MachineInstr *> &SeenMIs,
bool PreferFalse) const {
assert((MI.getOpcode() == ARM::MOVCCr || MI.getOpcode() == ARM::t2MOVCCr) &&
"Unknown select instruction");
MachineRegisterInfo &MRI = MI.getParent()->getParent()->getRegInfo();
MachineInstr *DefMI = canFoldIntoMOVCC(MI.getOperand(2).getReg(), MRI, this);
bool Invert = !DefMI;
if (!DefMI)
DefMI = canFoldIntoMOVCC(MI.getOperand(1).getReg(), MRI, this);
if (!DefMI)
return nullptr;
// Find new register class to use.
MachineOperand FalseReg = MI.getOperand(Invert ? 2 : 1);
Register DestReg = MI.getOperand(0).getReg();
const TargetRegisterClass *PreviousClass = MRI.getRegClass(FalseReg.getReg());
if (!MRI.constrainRegClass(DestReg, PreviousClass))
return nullptr;
// Create a new predicated version of DefMI.
// Rfalse is the first use.
MachineInstrBuilder NewMI =
BuildMI(*MI.getParent(), MI, MI.getDebugLoc(), DefMI->getDesc(), DestReg);
// Copy all the DefMI operands, excluding its (null) predicate.
const MCInstrDesc &DefDesc = DefMI->getDesc();
for (unsigned i = 1, e = DefDesc.getNumOperands();
i != e && !DefDesc.OpInfo[i].isPredicate(); ++i)
NewMI.add(DefMI->getOperand(i));
unsigned CondCode = MI.getOperand(3).getImm();
if (Invert)
NewMI.addImm(ARMCC::getOppositeCondition(ARMCC::CondCodes(CondCode)));
else
NewMI.addImm(CondCode);
NewMI.add(MI.getOperand(4));
// DefMI is not the -S version that sets CPSR, so add an optional %noreg.
if (NewMI->hasOptionalDef())
NewMI.add(condCodeOp());
// The output register value when the predicate is false is an implicit
// register operand tied to the first def.
// The tie makes the register allocator ensure the FalseReg is allocated the
// same register as operand 0.
FalseReg.setImplicit();
NewMI.add(FalseReg);
NewMI->tieOperands(0, NewMI->getNumOperands() - 1);
// Update SeenMIs set: register newly created MI and erase removed DefMI.
SeenMIs.insert(NewMI);
SeenMIs.erase(DefMI);
// If MI is inside a loop, and DefMI is outside the loop, then kill flags on
// DefMI would be invalid when tranferred inside the loop. Checking for a
// loop is expensive, but at least remove kill flags if they are in different
// BBs.
if (DefMI->getParent() != MI.getParent())
NewMI->clearKillInfo();
// The caller will erase MI, but not DefMI.
DefMI->eraseFromParent();
return NewMI;
}
/// Map pseudo instructions that imply an 'S' bit onto real opcodes. Whether the
/// instruction is encoded with an 'S' bit is determined by the optional CPSR
/// def operand.
///
/// This will go away once we can teach tblgen how to set the optional CPSR def
/// operand itself.
struct AddSubFlagsOpcodePair {
uint16_t PseudoOpc;
uint16_t MachineOpc;
};
static const AddSubFlagsOpcodePair AddSubFlagsOpcodeMap[] = {
{ARM::ADDSri, ARM::ADDri},
{ARM::ADDSrr, ARM::ADDrr},
{ARM::ADDSrsi, ARM::ADDrsi},
{ARM::ADDSrsr, ARM::ADDrsr},
{ARM::SUBSri, ARM::SUBri},
{ARM::SUBSrr, ARM::SUBrr},
{ARM::SUBSrsi, ARM::SUBrsi},
{ARM::SUBSrsr, ARM::SUBrsr},
{ARM::RSBSri, ARM::RSBri},
{ARM::RSBSrsi, ARM::RSBrsi},
{ARM::RSBSrsr, ARM::RSBrsr},
{ARM::tADDSi3, ARM::tADDi3},
{ARM::tADDSi8, ARM::tADDi8},
{ARM::tADDSrr, ARM::tADDrr},
{ARM::tADCS, ARM::tADC},
{ARM::tSUBSi3, ARM::tSUBi3},
{ARM::tSUBSi8, ARM::tSUBi8},
{ARM::tSUBSrr, ARM::tSUBrr},
{ARM::tSBCS, ARM::tSBC},
{ARM::tRSBS, ARM::tRSB},
{ARM::tLSLSri, ARM::tLSLri},
{ARM::t2ADDSri, ARM::t2ADDri},
{ARM::t2ADDSrr, ARM::t2ADDrr},
{ARM::t2ADDSrs, ARM::t2ADDrs},
{ARM::t2SUBSri, ARM::t2SUBri},
{ARM::t2SUBSrr, ARM::t2SUBrr},
{ARM::t2SUBSrs, ARM::t2SUBrs},
{ARM::t2RSBSri, ARM::t2RSBri},
{ARM::t2RSBSrs, ARM::t2RSBrs},
};
unsigned llvm::convertAddSubFlagsOpcode(unsigned OldOpc) {
for (unsigned i = 0, e = array_lengthof(AddSubFlagsOpcodeMap); i != e; ++i)
if (OldOpc == AddSubFlagsOpcodeMap[i].PseudoOpc)
return AddSubFlagsOpcodeMap[i].MachineOpc;
return 0;
}
void llvm::emitARMRegPlusImmediate(MachineBasicBlock &MBB,
MachineBasicBlock::iterator &MBBI,
const DebugLoc &dl, unsigned DestReg,
unsigned BaseReg, int NumBytes,
ARMCC::CondCodes Pred, unsigned PredReg,
const ARMBaseInstrInfo &TII,
unsigned MIFlags) {
if (NumBytes == 0 && DestReg != BaseReg) {
BuildMI(MBB, MBBI, dl, TII.get(ARM::MOVr), DestReg)
.addReg(BaseReg, RegState::Kill)
.add(predOps(Pred, PredReg))
.add(condCodeOp())
.setMIFlags(MIFlags);
return;
}
bool isSub = NumBytes < 0;
if (isSub) NumBytes = -NumBytes;
while (NumBytes) {
unsigned RotAmt = ARM_AM::getSOImmValRotate(NumBytes);
unsigned ThisVal = NumBytes & ARM_AM::rotr32(0xFF, RotAmt);
assert(ThisVal && "Didn't extract field correctly");
// We will handle these bits from offset, clear them.
NumBytes &= ~ThisVal;
assert(ARM_AM::getSOImmVal(ThisVal) != -1 && "Bit extraction didn't work?");
// Build the new ADD / SUB.
unsigned Opc = isSub ? ARM::SUBri : ARM::ADDri;
BuildMI(MBB, MBBI, dl, TII.get(Opc), DestReg)
.addReg(BaseReg, RegState::Kill)
.addImm(ThisVal)
.add(predOps(Pred, PredReg))
.add(condCodeOp())
.setMIFlags(MIFlags);
BaseReg = DestReg;
}
}
bool llvm::tryFoldSPUpdateIntoPushPop(const ARMSubtarget &Subtarget,
MachineFunction &MF, MachineInstr *MI,
unsigned NumBytes) {
// This optimisation potentially adds lots of load and store
// micro-operations, it's only really a great benefit to code-size.
if (!Subtarget.hasMinSize())
return false;
// If only one register is pushed/popped, LLVM can use an LDR/STR
// instead. We can't modify those so make sure we're dealing with an
// instruction we understand.
bool IsPop = isPopOpcode(MI->getOpcode());
bool IsPush = isPushOpcode(MI->getOpcode());
if (!IsPush && !IsPop)
return false;
bool IsVFPPushPop = MI->getOpcode() == ARM::VSTMDDB_UPD ||
MI->getOpcode() == ARM::VLDMDIA_UPD;
bool IsT1PushPop = MI->getOpcode() == ARM::tPUSH ||
MI->getOpcode() == ARM::tPOP ||
MI->getOpcode() == ARM::tPOP_RET;
assert((IsT1PushPop || (MI->getOperand(0).getReg() == ARM::SP &&
MI->getOperand(1).getReg() == ARM::SP)) &&
"trying to fold sp update into non-sp-updating push/pop");
// The VFP push & pop act on D-registers, so we can only fold an adjustment
// by a multiple of 8 bytes in correctly. Similarly rN is 4-bytes. Don't try
// if this is violated.
if (NumBytes % (IsVFPPushPop ? 8 : 4) != 0)
return false;
// ARM and Thumb2 push/pop insts have explicit "sp, sp" operands (+
// pred) so the list starts at 4. Thumb1 starts after the predicate.
int RegListIdx = IsT1PushPop ? 2 : 4;
// Calculate the space we'll need in terms of registers.
unsigned RegsNeeded;
const TargetRegisterClass *RegClass;
if (IsVFPPushPop) {
RegsNeeded = NumBytes / 8;
RegClass = &ARM::DPRRegClass;
} else {
RegsNeeded = NumBytes / 4;
RegClass = &ARM::GPRRegClass;
}
// We're going to have to strip all list operands off before
// re-adding them since the order matters, so save the existing ones
// for later.
SmallVector<MachineOperand, 4> RegList;
// We're also going to need the first register transferred by this
// instruction, which won't necessarily be the first register in the list.
unsigned FirstRegEnc = -1;
const TargetRegisterInfo *TRI = MF.getRegInfo().getTargetRegisterInfo();
for (int i = MI->getNumOperands() - 1; i >= RegListIdx; --i) {
MachineOperand &MO = MI->getOperand(i);
RegList.push_back(MO);
if (MO.isReg() && !MO.isImplicit() &&
TRI->getEncodingValue(MO.getReg()) < FirstRegEnc)
FirstRegEnc = TRI->getEncodingValue(MO.getReg());
}
const MCPhysReg *CSRegs = TRI->getCalleeSavedRegs(&MF);
// Now try to find enough space in the reglist to allocate NumBytes.
for (int CurRegEnc = FirstRegEnc - 1; CurRegEnc >= 0 && RegsNeeded;
--CurRegEnc) {
unsigned CurReg = RegClass->getRegister(CurRegEnc);
if (IsT1PushPop && CurRegEnc > TRI->getEncodingValue(ARM::R7))
continue;
if (!IsPop) {
// Pushing any register is completely harmless, mark the register involved
// as undef since we don't care about its value and must not restore it
// during stack unwinding.
RegList.push_back(MachineOperand::CreateReg(CurReg, false, false,
false, false, true));
--RegsNeeded;
continue;
}
// However, we can only pop an extra register if it's not live. For
// registers live within the function we might clobber a return value
// register; the other way a register can be live here is if it's
// callee-saved.
if (isCalleeSavedRegister(CurReg, CSRegs) ||
MI->getParent()->computeRegisterLiveness(TRI, CurReg, MI) !=
MachineBasicBlock::LQR_Dead) {
// VFP pops don't allow holes in the register list, so any skip is fatal
// for our transformation. GPR pops do, so we should just keep looking.
if (IsVFPPushPop)
return false;
else
continue;
}
// Mark the unimportant registers as <def,dead> in the POP.
RegList.push_back(MachineOperand::CreateReg(CurReg, true, false, false,
true));
--RegsNeeded;
}
if (RegsNeeded > 0)
return false;
// Finally we know we can profitably perform the optimisation so go
// ahead: strip all existing registers off and add them back again
// in the right order.
for (int i = MI->getNumOperands() - 1; i >= RegListIdx; --i)
MI->RemoveOperand(i);
// Add the complete list back in.
MachineInstrBuilder MIB(MF, &*MI);
for (int i = RegList.size() - 1; i >= 0; --i)
MIB.add(RegList[i]);
return true;
}
bool llvm::rewriteARMFrameIndex(MachineInstr &MI, unsigned FrameRegIdx,
unsigned FrameReg, int &Offset,
const ARMBaseInstrInfo &TII) {
unsigned Opcode = MI.getOpcode();
const MCInstrDesc &Desc = MI.getDesc();
unsigned AddrMode = (Desc.TSFlags & ARMII::AddrModeMask);
bool isSub = false;
// Memory operands in inline assembly always use AddrMode2.
if (Opcode == ARM::INLINEASM || Opcode == ARM::INLINEASM_BR)
AddrMode = ARMII::AddrMode2;
if (Opcode == ARM::ADDri) {
Offset += MI.getOperand(FrameRegIdx+1).getImm();
if (Offset == 0) {
// Turn it into a move.
MI.setDesc(TII.get(ARM::MOVr));
MI.getOperand(FrameRegIdx).ChangeToRegister(FrameReg, false);
MI.RemoveOperand(FrameRegIdx+1);
Offset = 0;
return true;
} else if (Offset < 0) {
Offset = -Offset;
isSub = true;
MI.setDesc(TII.get(ARM::SUBri));
}
// Common case: small offset, fits into instruction.
if (ARM_AM::getSOImmVal(Offset) != -1) {
// Replace the FrameIndex with sp / fp
MI.getOperand(FrameRegIdx).ChangeToRegister(FrameReg, false);
MI.getOperand(FrameRegIdx+1).ChangeToImmediate(Offset);
Offset = 0;
return true;
}
// Otherwise, pull as much of the immedidate into this ADDri/SUBri
// as possible.
unsigned RotAmt = ARM_AM::getSOImmValRotate(Offset);
unsigned ThisImmVal = Offset & ARM_AM::rotr32(0xFF, RotAmt);
// We will handle these bits from offset, clear them.
Offset &= ~ThisImmVal;
// Get the properly encoded SOImmVal field.
assert(ARM_AM::getSOImmVal(ThisImmVal) != -1 &&
"Bit extraction didn't work?");
MI.getOperand(FrameRegIdx+1).ChangeToImmediate(ThisImmVal);
} else {
unsigned ImmIdx = 0;
int InstrOffs = 0;
unsigned NumBits = 0;
unsigned Scale = 1;
switch (AddrMode) {
case ARMII::AddrMode_i12:
ImmIdx = FrameRegIdx + 1;
InstrOffs = MI.getOperand(ImmIdx).getImm();
NumBits = 12;
break;
case ARMII::AddrMode2:
ImmIdx = FrameRegIdx+2;
InstrOffs = ARM_AM::getAM2Offset(MI.getOperand(ImmIdx).getImm());
if (ARM_AM::getAM2Op(MI.getOperand(ImmIdx).getImm()) == ARM_AM::sub)
InstrOffs *= -1;
NumBits = 12;
break;
case ARMII::AddrMode3:
ImmIdx = FrameRegIdx+2;
InstrOffs = ARM_AM::getAM3Offset(MI.getOperand(ImmIdx).getImm());
if (ARM_AM::getAM3Op(MI.getOperand(ImmIdx).getImm()) == ARM_AM::sub)
InstrOffs *= -1;
NumBits = 8;
break;
case ARMII::AddrMode4:
case ARMII::AddrMode6:
// Can't fold any offset even if it's zero.
return false;
case ARMII::AddrMode5:
ImmIdx = FrameRegIdx+1;
InstrOffs = ARM_AM::getAM5Offset(MI.getOperand(ImmIdx).getImm());
if (ARM_AM::getAM5Op(MI.getOperand(ImmIdx).getImm()) == ARM_AM::sub)
InstrOffs *= -1;
NumBits = 8;
Scale = 4;
break;
case ARMII::AddrMode5FP16:
ImmIdx = FrameRegIdx+1;
InstrOffs = ARM_AM::getAM5Offset(MI.getOperand(ImmIdx).getImm());
if (ARM_AM::getAM5Op(MI.getOperand(ImmIdx).getImm()) == ARM_AM::sub)
InstrOffs *= -1;
NumBits = 8;
Scale = 2;
break;
case ARMII::AddrModeT2_i7:
case ARMII::AddrModeT2_i7s2:
case ARMII::AddrModeT2_i7s4:
ImmIdx = FrameRegIdx+1;
InstrOffs = MI.getOperand(ImmIdx).getImm();
NumBits = 7;
Scale = (AddrMode == ARMII::AddrModeT2_i7s2 ? 2 :
AddrMode == ARMII::AddrModeT2_i7s4 ? 4 : 1);
break;
default:
llvm_unreachable("Unsupported addressing mode!");
}
Offset += InstrOffs * Scale;
assert((Offset & (Scale-1)) == 0 && "Can't encode this offset!");
if (Offset < 0) {
Offset = -Offset;
isSub = true;
}
// Attempt to fold address comp. if opcode has offset bits
if (NumBits > 0) {
// Common case: small offset, fits into instruction.
MachineOperand &ImmOp = MI.getOperand(ImmIdx);
int ImmedOffset = Offset / Scale;
unsigned Mask = (1 << NumBits) - 1;
if ((unsigned)Offset <= Mask * Scale) {
// Replace the FrameIndex with sp
MI.getOperand(FrameRegIdx).ChangeToRegister(FrameReg, false);
// FIXME: When addrmode2 goes away, this will simplify (like the
// T2 version), as the LDR.i12 versions don't need the encoding
// tricks for the offset value.
if (isSub) {
if (AddrMode == ARMII::AddrMode_i12)
ImmedOffset = -ImmedOffset;
else
ImmedOffset |= 1 << NumBits;
}
ImmOp.ChangeToImmediate(ImmedOffset);
Offset = 0;
return true;
}
// Otherwise, it didn't fit. Pull in what we can to simplify the immed.
ImmedOffset = ImmedOffset & Mask;
if (isSub) {
if (AddrMode == ARMII::AddrMode_i12)
ImmedOffset = -ImmedOffset;
else
ImmedOffset |= 1 << NumBits;
}
ImmOp.ChangeToImmediate(ImmedOffset);
Offset &= ~(Mask*Scale);
}
}
Offset = (isSub) ? -Offset : Offset;
return Offset == 0;
}
/// analyzeCompare - For a comparison instruction, return the source registers
/// in SrcReg and SrcReg2 if having two register operands, and the value it
/// compares against in CmpValue. Return true if the comparison instruction
/// can be analyzed.
bool ARMBaseInstrInfo::analyzeCompare(const MachineInstr &MI, unsigned &SrcReg,
unsigned &SrcReg2, int &CmpMask,
int &CmpValue) const {
switch (MI.getOpcode()) {
default: break;
case ARM::CMPri:
case ARM::t2CMPri:
case ARM::tCMPi8:
SrcReg = MI.getOperand(0).getReg();
SrcReg2 = 0;
CmpMask = ~0;
CmpValue = MI.getOperand(1).getImm();
return true;
case ARM::CMPrr:
case ARM::t2CMPrr:
case ARM::tCMPr:
SrcReg = MI.getOperand(0).getReg();
SrcReg2 = MI.getOperand(1).getReg();
CmpMask = ~0;
CmpValue = 0;
return true;
case ARM::TSTri:
case ARM::t2TSTri:
SrcReg = MI.getOperand(0).getReg();
SrcReg2 = 0;
CmpMask = MI.getOperand(1).getImm();
CmpValue = 0;
return true;
}
return false;
}
/// isSuitableForMask - Identify a suitable 'and' instruction that
/// operates on the given source register and applies the same mask
/// as a 'tst' instruction. Provide a limited look-through for copies.
/// When successful, MI will hold the found instruction.
static bool isSuitableForMask(MachineInstr *&MI, unsigned SrcReg,
int CmpMask, bool CommonUse) {
switch (MI->getOpcode()) {
case ARM::ANDri:
case ARM::t2ANDri:
if (CmpMask != MI->getOperand(2).getImm())
return false;
if (SrcReg == MI->getOperand(CommonUse ? 1 : 0).getReg())
return true;
break;
}
return false;
}
/// getSwappedCondition - assume the flags are set by MI(a,b), return
/// the condition code if we modify the instructions such that flags are
/// set by MI(b,a).
inline static ARMCC::CondCodes getSwappedCondition(ARMCC::CondCodes CC) {
switch (CC) {
default: return ARMCC::AL;
case ARMCC::EQ: return ARMCC::EQ;
case ARMCC::NE: return ARMCC::NE;
case ARMCC::HS: return ARMCC::LS;
case ARMCC::LO: return ARMCC::HI;
case ARMCC::HI: return ARMCC::LO;
case ARMCC::LS: return ARMCC::HS;
case ARMCC::GE: return ARMCC::LE;
case ARMCC::LT: return ARMCC::GT;
case ARMCC::GT: return ARMCC::LT;
case ARMCC::LE: return ARMCC::GE;
}
}
/// getCmpToAddCondition - assume the flags are set by CMP(a,b), return
/// the condition code if we modify the instructions such that flags are
/// set by ADD(a,b,X).
inline static ARMCC::CondCodes getCmpToAddCondition(ARMCC::CondCodes CC) {
switch (CC) {
default: return ARMCC::AL;
case ARMCC::HS: return ARMCC::LO;
case ARMCC::LO: return ARMCC::HS;
case ARMCC::VS: return ARMCC::VS;
case ARMCC::VC: return ARMCC::VC;
}
}
/// isRedundantFlagInstr - check whether the first instruction, whose only
/// purpose is to update flags, can be made redundant.
/// CMPrr can be made redundant by SUBrr if the operands are the same.
/// CMPri can be made redundant by SUBri if the operands are the same.
/// CMPrr(r0, r1) can be made redundant by ADDr[ri](r0, r1, X).
/// This function can be extended later on.
inline static bool isRedundantFlagInstr(const MachineInstr *CmpI,
unsigned SrcReg, unsigned SrcReg2,
int ImmValue, const MachineInstr *OI,
bool &IsThumb1) {
if ((CmpI->getOpcode() == ARM::CMPrr || CmpI->getOpcode() == ARM::t2CMPrr) &&
(OI->getOpcode() == ARM::SUBrr || OI->getOpcode() == ARM::t2SUBrr) &&
((OI->getOperand(1).getReg() == SrcReg &&
OI->getOperand(2).getReg() == SrcReg2) ||
(OI->getOperand(1).getReg() == SrcReg2 &&
OI->getOperand(2).getReg() == SrcReg))) {
IsThumb1 = false;
return true;
}
if (CmpI->getOpcode() == ARM::tCMPr && OI->getOpcode() == ARM::tSUBrr &&
((OI->getOperand(2).getReg() == SrcReg &&
OI->getOperand(3).getReg() == SrcReg2) ||
(OI->getOperand(2).getReg() == SrcReg2 &&
OI->getOperand(3).getReg() == SrcReg))) {
IsThumb1 = true;
return true;
}
if ((CmpI->getOpcode() == ARM::CMPri || CmpI->getOpcode() == ARM::t2CMPri) &&
(OI->getOpcode() == ARM::SUBri || OI->getOpcode() == ARM::t2SUBri) &&
OI->getOperand(1).getReg() == SrcReg &&
OI->getOperand(2).getImm() == ImmValue) {
IsThumb1 = false;
return true;
}
if (CmpI->getOpcode() == ARM::tCMPi8 &&
(OI->getOpcode() == ARM::tSUBi8 || OI->getOpcode() == ARM::tSUBi3) &&
OI->getOperand(2).getReg() == SrcReg &&
OI->getOperand(3).getImm() == ImmValue) {
IsThumb1 = true;
return true;
}
if ((CmpI->getOpcode() == ARM::CMPrr || CmpI->getOpcode() == ARM::t2CMPrr) &&
(OI->getOpcode() == ARM::ADDrr || OI->getOpcode() == ARM::t2ADDrr ||
OI->getOpcode() == ARM::ADDri || OI->getOpcode() == ARM::t2ADDri) &&
OI->getOperand(0).isReg() && OI->getOperand(1).isReg() &&
OI->getOperand(0).getReg() == SrcReg &&
OI->getOperand(1).getReg() == SrcReg2) {
IsThumb1 = false;
return true;
}
if (CmpI->getOpcode() == ARM::tCMPr &&
(OI->getOpcode() == ARM::tADDi3 || OI->getOpcode() == ARM::tADDi8 ||
OI->getOpcode() == ARM::tADDrr) &&
OI->getOperand(0).getReg() == SrcReg &&
OI->getOperand(2).getReg() == SrcReg2) {
IsThumb1 = true;
return true;
}
return false;
}
static bool isOptimizeCompareCandidate(MachineInstr *MI, bool &IsThumb1) {
switch (MI->getOpcode()) {
default: return false;
case ARM::tLSLri:
case ARM::tLSRri:
case ARM::tLSLrr:
case ARM::tLSRrr:
case ARM::tSUBrr:
case ARM::tADDrr:
case ARM::tADDi3:
case ARM::tADDi8:
case ARM::tSUBi3:
case ARM::tSUBi8:
case ARM::tMUL:
case ARM::tADC:
case ARM::tSBC:
case ARM::tRSB:
case ARM::tAND:
case ARM::tORR:
case ARM::tEOR:
case ARM::tBIC:
case ARM::tMVN:
case ARM::tASRri:
case ARM::tASRrr:
case ARM::tROR:
IsThumb1 = true;
LLVM_FALLTHROUGH;
case ARM::RSBrr:
case ARM::RSBri:
case ARM::RSCrr:
case ARM::RSCri:
case ARM::ADDrr:
case ARM::ADDri:
case ARM::ADCrr:
case ARM::ADCri:
case ARM::SUBrr:
case ARM::SUBri:
case ARM::SBCrr:
case ARM::SBCri:
case ARM::t2RSBri:
case ARM::t2ADDrr:
case ARM::t2ADDri:
case ARM::t2ADCrr:
case ARM::t2ADCri:
case ARM::t2SUBrr:
case ARM::t2SUBri:
case ARM::t2SBCrr:
case ARM::t2SBCri:
case ARM::ANDrr:
case ARM::ANDri:
case ARM::t2ANDrr:
case ARM::t2ANDri:
case ARM::ORRrr:
case ARM::ORRri:
case ARM::t2ORRrr:
case ARM::t2ORRri:
case ARM::EORrr:
case ARM::EORri:
case ARM::t2EORrr:
case ARM::t2EORri:
case ARM::t2LSRri:
case ARM::t2LSRrr:
case ARM::t2LSLri:
case ARM::t2LSLrr:
return true;
}
}
/// optimizeCompareInstr - Convert the instruction supplying the argument to the
/// comparison into one that sets the zero bit in the flags register;
/// Remove a redundant Compare instruction if an earlier instruction can set the
/// flags in the same way as Compare.
/// E.g. SUBrr(r1,r2) and CMPrr(r1,r2). We also handle the case where two
/// operands are swapped: SUBrr(r1,r2) and CMPrr(r2,r1), by updating the
/// condition code of instructions which use the flags.
bool ARMBaseInstrInfo::optimizeCompareInstr(
MachineInstr &CmpInstr, unsigned SrcReg, unsigned SrcReg2, int CmpMask,
int CmpValue, const MachineRegisterInfo *MRI) const {
// Get the unique definition of SrcReg.
MachineInstr *MI = MRI->getUniqueVRegDef(SrcReg);
if (!MI) return false;
// Masked compares sometimes use the same register as the corresponding 'and'.
if (CmpMask != ~0) {
if (!isSuitableForMask(MI, SrcReg, CmpMask, false) || isPredicated(*MI)) {
MI = nullptr;
for (MachineRegisterInfo::use_instr_iterator
UI = MRI->use_instr_begin(SrcReg), UE = MRI->use_instr_end();
UI != UE; ++UI) {
if (UI->getParent() != CmpInstr.getParent())
continue;
MachineInstr *PotentialAND = &*UI;
if (!isSuitableForMask(PotentialAND, SrcReg, CmpMask, true) ||
isPredicated(*PotentialAND))
continue;
MI = PotentialAND;
break;
}
if (!MI) return false;
}
}
// Get ready to iterate backward from CmpInstr.
MachineBasicBlock::iterator I = CmpInstr, E = MI,
B = CmpInstr.getParent()->begin();
// Early exit if CmpInstr is at the beginning of the BB.
if (I == B) return false;
// There are two possible candidates which can be changed to set CPSR:
// One is MI, the other is a SUB or ADD instruction.
// For CMPrr(r1,r2), we are looking for SUB(r1,r2), SUB(r2,r1), or
// ADDr[ri](r1, r2, X).
// For CMPri(r1, CmpValue), we are looking for SUBri(r1, CmpValue).
MachineInstr *SubAdd = nullptr;
if (SrcReg2 != 0)
// MI is not a candidate for CMPrr.
MI = nullptr;
else if (MI->getParent() != CmpInstr.getParent() || CmpValue != 0) {
// Conservatively refuse to convert an instruction which isn't in the same
// BB as the comparison.
// For CMPri w/ CmpValue != 0, a SubAdd may still be a candidate.
// Thus we cannot return here.
if (CmpInstr.getOpcode() == ARM::CMPri ||
CmpInstr.getOpcode() == ARM::t2CMPri ||
CmpInstr.getOpcode() == ARM::tCMPi8)
MI = nullptr;
else
return false;
}
bool IsThumb1 = false;
if (MI && !isOptimizeCompareCandidate(MI, IsThumb1))
return false;
// We also want to do this peephole for cases like this: if (a*b == 0),
// and optimise away the CMP instruction from the generated code sequence:
// MULS, MOVS, MOVS, CMP. Here the MOVS instructions load the boolean values
// resulting from the select instruction, but these MOVS instructions for
// Thumb1 (V6M) are flag setting and are thus preventing this optimisation.
// However, if we only have MOVS instructions in between the CMP and the
// other instruction (the MULS in this example), then the CPSR is dead so we
// can safely reorder the sequence into: MOVS, MOVS, MULS, CMP. We do this
// reordering and then continue the analysis hoping we can eliminate the
// CMP. This peephole works on the vregs, so is still in SSA form. As a
// consequence, the movs won't redefine/kill the MUL operands which would
// make this reordering illegal.
const TargetRegisterInfo *TRI = &getRegisterInfo();
if (MI && IsThumb1) {
--I;
if (I != E && !MI->readsRegister(ARM::CPSR, TRI)) {
bool CanReorder = true;
for (; I != E; --I) {
if (I->getOpcode() != ARM::tMOVi8) {
CanReorder = false;
break;
}
}
if (CanReorder) {
MI = MI->removeFromParent();
E = CmpInstr;
CmpInstr.getParent()->insert(E, MI);
}
}
I = CmpInstr;
E = MI;
}
// Check that CPSR isn't set between the comparison instruction and the one we
// want to change. At the same time, search for SubAdd.
bool SubAddIsThumb1 = false;
do {
const MachineInstr &Instr = *--I;
// Check whether CmpInstr can be made redundant by the current instruction.
if (isRedundantFlagInstr(&CmpInstr, SrcReg, SrcReg2, CmpValue, &Instr,
SubAddIsThumb1)) {
SubAdd = &*I;
break;
}
// Allow E (which was initially MI) to be SubAdd but do not search before E.
if (I == E)
break;
if (Instr.modifiesRegister(ARM::CPSR, TRI) ||
Instr.readsRegister(ARM::CPSR, TRI))
// This instruction modifies or uses CPSR after the one we want to
// change. We can't do this transformation.
return false;
if (I == B) {
// In some cases, we scan the use-list of an instruction for an AND;
// that AND is in the same BB, but may not be scheduled before the
// corresponding TST. In that case, bail out.
//
// FIXME: We could try to reschedule the AND.
return false;
}
} while (true);
// Return false if no candidates exist.
if (!MI && !SubAdd)
return false;
// If we found a SubAdd, use it as it will be closer to the CMP
if (SubAdd) {
MI = SubAdd;
IsThumb1 = SubAddIsThumb1;
}
// We can't use a predicated instruction - it doesn't always write the flags.
if (isPredicated(*MI))
return false;
// Scan forward for the use of CPSR
// When checking against MI: if it's a conditional code that requires
// checking of the V bit or C bit, then this is not safe to do.
// It is safe to remove CmpInstr if CPSR is redefined or killed.
// If we are done with the basic block, we need to check whether CPSR is
// live-out.
SmallVector<std::pair<MachineOperand*, ARMCC::CondCodes>, 4>
OperandsToUpdate;
bool isSafe = false;
I = CmpInstr;
E = CmpInstr.getParent()->end();
while (!isSafe && ++I != E) {
const MachineInstr &Instr = *I;
for (unsigned IO = 0, EO = Instr.getNumOperands();
!isSafe && IO != EO; ++IO) {
const MachineOperand &MO = Instr.getOperand(IO);
if (MO.isRegMask() && MO.clobbersPhysReg(ARM::CPSR)) {
isSafe = true;
break;
}
if (!MO.isReg() || MO.getReg() != ARM::CPSR)
continue;
if (MO.isDef()) {
isSafe = true;
break;
}
// Condition code is after the operand before CPSR except for VSELs.
ARMCC::CondCodes CC;
bool IsInstrVSel = true;
switch (Instr.getOpcode()) {
default:
IsInstrVSel = false;
CC = (ARMCC::CondCodes)Instr.getOperand(IO - 1).getImm();
break;
case ARM::VSELEQD:
case ARM::VSELEQS:
case ARM::VSELEQH:
CC = ARMCC::EQ;
break;
case ARM::VSELGTD:
case ARM::VSELGTS:
case ARM::VSELGTH:
CC = ARMCC::GT;
break;
case ARM::VSELGED:
case ARM::VSELGES:
case ARM::VSELGEH:
CC = ARMCC::GE;
break;
case ARM::VSELVSD:
case ARM::VSELVSS:
case ARM::VSELVSH:
CC = ARMCC::VS;
break;
}
if (SubAdd) {
// If we have SUB(r1, r2) and CMP(r2, r1), the condition code based
// on CMP needs to be updated to be based on SUB.
// If we have ADD(r1, r2, X) and CMP(r1, r2), the condition code also
// needs to be modified.
// Push the condition code operands to OperandsToUpdate.
// If it is safe to remove CmpInstr, the condition code of these
// operands will be modified.
unsigned Opc = SubAdd->getOpcode();
bool IsSub = Opc == ARM::SUBrr || Opc == ARM::t2SUBrr ||
Opc == ARM::SUBri || Opc == ARM::t2SUBri ||
Opc == ARM::tSUBrr || Opc == ARM::tSUBi3 ||
Opc == ARM::tSUBi8;
unsigned OpI = Opc != ARM::tSUBrr ? 1 : 2;
if (!IsSub ||
(SrcReg2 != 0 && SubAdd->getOperand(OpI).getReg() == SrcReg2 &&
SubAdd->getOperand(OpI + 1).getReg() == SrcReg)) {
// VSel doesn't support condition code update.
if (IsInstrVSel)
return false;
// Ensure we can swap the condition.
ARMCC::CondCodes NewCC = (IsSub ? getSwappedCondition(CC) : getCmpToAddCondition(CC));
if (NewCC == ARMCC::AL)
return false;
OperandsToUpdate.push_back(
std::make_pair(&((*I).getOperand(IO - 1)), NewCC));
}
} else {
// No SubAdd, so this is x = <op> y, z; cmp x, 0.
switch (CC) {
case ARMCC::EQ: // Z
case ARMCC::NE: // Z
case ARMCC::MI: // N
case ARMCC::PL: // N
case ARMCC::AL: // none
// CPSR can be used multiple times, we should continue.
break;
case ARMCC::HS: // C
case ARMCC::LO: // C
case ARMCC::VS: // V
case ARMCC::VC: // V
case ARMCC::HI: // C Z
case ARMCC::LS: // C Z
case ARMCC::GE: // N V
case ARMCC::LT: // N V
case ARMCC::GT: // Z N V
case ARMCC::LE: // Z N V
// The instruction uses the V bit or C bit which is not safe.
return false;
}
}
}
}
// If CPSR is not killed nor re-defined, we should check whether it is
// live-out. If it is live-out, do not optimize.
if (!isSafe) {
MachineBasicBlock *MBB = CmpInstr.getParent();
for (MachineBasicBlock::succ_iterator SI = MBB->succ_begin(),
SE = MBB->succ_end(); SI != SE; ++SI)
if ((*SI)->isLiveIn(ARM::CPSR))
return false;
}
// Toggle the optional operand to CPSR (if it exists - in Thumb1 we always
// set CPSR so this is represented as an explicit output)
if (!IsThumb1) {
MI->getOperand(5).setReg(ARM::CPSR);
MI->getOperand(5).setIsDef(true);
}
assert(!isPredicated(*MI) && "Can't use flags from predicated instruction");
CmpInstr.eraseFromParent();
// Modify the condition code of operands in OperandsToUpdate.
// Since we have SUB(r1, r2) and CMP(r2, r1), the condition code needs to
// be changed from r2 > r1 to r1 < r2, from r2 < r1 to r1 > r2, etc.
for (unsigned i = 0, e = OperandsToUpdate.size(); i < e; i++)
OperandsToUpdate[i].first->setImm(OperandsToUpdate[i].second);
MI->clearRegisterDeads(ARM::CPSR);
return true;
}
bool ARMBaseInstrInfo::shouldSink(const MachineInstr &MI) const {
// Do not sink MI if it might be used to optimize a redundant compare.
// We heuristically only look at the instruction immediately following MI to
// avoid potentially searching the entire basic block.
if (isPredicated(MI))
return true;
MachineBasicBlock::const_iterator Next = &MI;
++Next;
unsigned SrcReg, SrcReg2;
int CmpMask, CmpValue;
bool IsThumb1;
if (Next != MI.getParent()->end() &&
analyzeCompare(*Next, SrcReg, SrcReg2, CmpMask, CmpValue) &&
isRedundantFlagInstr(&*Next, SrcReg, SrcReg2, CmpValue, &MI, IsThumb1))
return false;
return true;
}
bool ARMBaseInstrInfo::FoldImmediate(MachineInstr &UseMI, MachineInstr &DefMI,
unsigned Reg,
MachineRegisterInfo *MRI) const {
// Fold large immediates into add, sub, or, xor.
unsigned DefOpc = DefMI.getOpcode();
if (DefOpc != ARM::t2MOVi32imm && DefOpc != ARM::MOVi32imm)
return false;
if (!DefMI.getOperand(1).isImm())
// Could be t2MOVi32imm @xx
return false;
if (!MRI->hasOneNonDBGUse(Reg))
return false;
const MCInstrDesc &DefMCID = DefMI.getDesc();
if (DefMCID.hasOptionalDef()) {
unsigned NumOps = DefMCID.getNumOperands();
const MachineOperand &MO = DefMI.getOperand(NumOps - 1);
if (MO.getReg() == ARM::CPSR && !MO.isDead())
// If DefMI defines CPSR and it is not dead, it's obviously not safe
// to delete DefMI.
return false;
}
const MCInstrDesc &UseMCID = UseMI.getDesc();
if (UseMCID.hasOptionalDef()) {
unsigned NumOps = UseMCID.getNumOperands();
if (UseMI.getOperand(NumOps - 1).getReg() == ARM::CPSR)
// If the instruction sets the flag, do not attempt this optimization
// since it may change the semantics of the code.
return false;
}
unsigned UseOpc = UseMI.getOpcode();
unsigned NewUseOpc = 0;
uint32_t ImmVal = (uint32_t)DefMI.getOperand(1).getImm();
uint32_t SOImmValV1 = 0, SOImmValV2 = 0;
bool Commute = false;
switch (UseOpc) {
default: return false;
case ARM::SUBrr:
case ARM::ADDrr:
case ARM::ORRrr:
case ARM::EORrr:
case ARM::t2SUBrr:
case ARM::t2ADDrr:
case ARM::t2ORRrr:
case ARM::t2EORrr: {
Commute = UseMI.getOperand(2).getReg() != Reg;
switch (UseOpc) {
default: break;
case ARM::ADDrr:
case ARM::SUBrr:
if (UseOpc == ARM::SUBrr && Commute)
return false;
// ADD/SUB are special because they're essentially the same operation, so
// we can handle a larger range of immediates.
if (ARM_AM::isSOImmTwoPartVal(ImmVal))
NewUseOpc = UseOpc == ARM::ADDrr ? ARM::ADDri : ARM::SUBri;
else if (ARM_AM::isSOImmTwoPartVal(-ImmVal)) {
ImmVal = -ImmVal;
NewUseOpc = UseOpc == ARM::ADDrr ? ARM::SUBri : ARM::ADDri;
} else
return false;
SOImmValV1 = (uint32_t)ARM_AM::getSOImmTwoPartFirst(ImmVal);
SOImmValV2 = (uint32_t)ARM_AM::getSOImmTwoPartSecond(ImmVal);
break;
case ARM::ORRrr:
case ARM::EORrr:
if (!ARM_AM::isSOImmTwoPartVal(ImmVal))
return false;
SOImmValV1 = (uint32_t)ARM_AM::getSOImmTwoPartFirst(ImmVal);
SOImmValV2 = (uint32_t)ARM_AM::getSOImmTwoPartSecond(ImmVal);
switch (UseOpc) {
default: break;
case ARM::ORRrr: NewUseOpc = ARM::ORRri; break;
case ARM::EORrr: NewUseOpc = ARM::EORri; break;
}
break;
case ARM::t2ADDrr:
case ARM::t2SUBrr:
if (UseOpc == ARM::t2SUBrr && Commute)
return false;
// ADD/SUB are special because they're essentially the same operation, so
// we can handle a larger range of immediates.
if (ARM_AM::isT2SOImmTwoPartVal(ImmVal))
NewUseOpc = UseOpc == ARM::t2ADDrr ? ARM::t2ADDri : ARM::t2SUBri;
else if (ARM_AM::isT2SOImmTwoPartVal(-ImmVal)) {
ImmVal = -ImmVal;
NewUseOpc = UseOpc == ARM::t2ADDrr ? ARM::t2SUBri : ARM::t2ADDri;
} else
return false;
SOImmValV1 = (uint32_t)ARM_AM::getT2SOImmTwoPartFirst(ImmVal);
SOImmValV2 = (uint32_t)ARM_AM::getT2SOImmTwoPartSecond(ImmVal);
break;
case ARM::t2ORRrr:
case ARM::t2EORrr:
if (!ARM_AM::isT2SOImmTwoPartVal(ImmVal))
return false;
SOImmValV1 = (uint32_t)ARM_AM::getT2SOImmTwoPartFirst(ImmVal);
SOImmValV2 = (uint32_t)ARM_AM::getT2SOImmTwoPartSecond(ImmVal);
switch (UseOpc) {
default: break;
case ARM::t2ORRrr: NewUseOpc = ARM::t2ORRri; break;
case ARM::t2EORrr: NewUseOpc = ARM::t2EORri; break;
}
break;
}
}
}
unsigned OpIdx = Commute ? 2 : 1;
Register Reg1 = UseMI.getOperand(OpIdx).getReg();
bool isKill = UseMI.getOperand(OpIdx).isKill();
Register NewReg = MRI->createVirtualRegister(MRI->getRegClass(Reg));
BuildMI(*UseMI.getParent(), UseMI, UseMI.getDebugLoc(), get(NewUseOpc),
NewReg)
.addReg(Reg1, getKillRegState(isKill))
.addImm(SOImmValV1)
.add(predOps(ARMCC::AL))
.add(condCodeOp());
UseMI.setDesc(get(NewUseOpc));
UseMI.getOperand(1).setReg(NewReg);
UseMI.getOperand(1).setIsKill();
UseMI.getOperand(2).ChangeToImmediate(SOImmValV2);
DefMI.eraseFromParent();
return true;
}
static unsigned getNumMicroOpsSwiftLdSt(const InstrItineraryData *ItinData,
const MachineInstr &MI) {
switch (MI.getOpcode()) {
default: {
const MCInstrDesc &Desc = MI.getDesc();
int UOps = ItinData->getNumMicroOps(Desc.getSchedClass());
assert(UOps >= 0 && "bad # UOps");
return UOps;
}
case ARM::LDRrs:
case ARM::LDRBrs:
case ARM::STRrs:
case ARM::STRBrs: {
unsigned ShOpVal = MI.getOperand(3).getImm();
bool isSub = ARM_AM::getAM2Op(ShOpVal) == ARM_AM::sub;
unsigned ShImm = ARM_AM::getAM2Offset(ShOpVal);
if (!isSub &&
(ShImm == 0 ||
((ShImm == 1 || ShImm == 2 || ShImm == 3) &&
ARM_AM::getAM2ShiftOpc(ShOpVal) == ARM_AM::lsl)))
return 1;
return 2;
}
case ARM::LDRH:
case ARM::STRH: {
if (!MI.getOperand(2).getReg())
return 1;
unsigned ShOpVal = MI.getOperand(3).getImm();
bool isSub = ARM_AM::getAM2Op(ShOpVal) == ARM_AM::sub;
unsigned ShImm = ARM_AM::getAM2Offset(ShOpVal);
if (!isSub &&
(ShImm == 0 ||
((ShImm == 1 || ShImm == 2 || ShImm == 3) &&
ARM_AM::getAM2ShiftOpc(ShOpVal) == ARM_AM::lsl)))
return 1;
return 2;
}
case ARM::LDRSB:
case ARM::LDRSH:
return (ARM_AM::getAM3Op(MI.getOperand(3).getImm()) == ARM_AM::sub) ? 3 : 2;
case ARM::LDRSB_POST:
case ARM::LDRSH_POST: {
Register Rt = MI.getOperand(0).getReg();
Register Rm = MI.getOperand(3).getReg();
return (Rt == Rm) ? 4 : 3;
}
case ARM::LDR_PRE_REG:
case ARM::LDRB_PRE_REG: {
Register Rt = MI.getOperand(0).getReg();
Register Rm = MI.getOperand(3).getReg();
if (Rt == Rm)
return 3;
unsigned ShOpVal = MI.getOperand(4).getImm();
bool isSub = ARM_AM::getAM2Op(ShOpVal) == ARM_AM::sub;
unsigned ShImm = ARM_AM::getAM2Offset(ShOpVal);
if (!isSub &&
(ShImm == 0 ||
((ShImm == 1 || ShImm == 2 || ShImm == 3) &&
ARM_AM::getAM2ShiftOpc(ShOpVal) == ARM_AM::lsl)))
return 2;
return 3;
}
case ARM::STR_PRE_REG:
case ARM::STRB_PRE_REG: {
unsigned ShOpVal = MI.getOperand(4).getImm();
bool isSub = ARM_AM::getAM2Op(ShOpVal) == ARM_AM::sub;
unsigned ShImm = ARM_AM::getAM2Offset(ShOpVal);
if (!isSub &&
(ShImm == 0 ||
((ShImm == 1 || ShImm == 2 || ShImm == 3) &&
ARM_AM::getAM2ShiftOpc(ShOpVal) == ARM_AM::lsl)))
return 2;
return 3;
}
case ARM::LDRH_PRE:
case ARM::STRH_PRE: {
Register Rt = MI.getOperand(0).getReg();
Register Rm = MI.getOperand(3).getReg();
if (!Rm)
return 2;
if (Rt == Rm)
return 3;
return (ARM_AM::getAM3Op(MI.getOperand(4).getImm()) == ARM_AM::sub) ? 3 : 2;
}
case ARM::LDR_POST_REG:
case ARM::LDRB_POST_REG:
case ARM::LDRH_POST: {
Register Rt = MI.getOperand(0).getReg();
Register Rm = MI.getOperand(3).getReg();
return (Rt == Rm) ? 3 : 2;
}
case ARM::LDR_PRE_IMM:
case ARM::LDRB_PRE_IMM:
case ARM::LDR_POST_IMM:
case ARM::LDRB_POST_IMM:
case ARM::STRB_POST_IMM:
case ARM::STRB_POST_REG:
case ARM::STRB_PRE_IMM:
case ARM::STRH_POST:
case ARM::STR_POST_IMM:
case ARM::STR_POST_REG:
case ARM::STR_PRE_IMM:
return 2;
case ARM::LDRSB_PRE:
case ARM::LDRSH_PRE: {
Register Rm = MI.getOperand(3).getReg();
if (Rm == 0)
return 3;
Register Rt = MI.getOperand(0).getReg();
if (Rt == Rm)
return 4;
unsigned ShOpVal = MI.getOperand(4).getImm();
bool isSub = ARM_AM::getAM2Op(ShOpVal) == ARM_AM::sub;
unsigned ShImm = ARM_AM::getAM2Offset(ShOpVal);
if (!isSub &&
(ShImm == 0 ||
((ShImm == 1 || ShImm == 2 || ShImm == 3) &&
ARM_AM::getAM2ShiftOpc(ShOpVal) == ARM_AM::lsl)))
return 3;
return 4;
}
case ARM::LDRD: {
Register Rt = MI.getOperand(0).getReg();
Register Rn = MI.getOperand(2).getReg();
Register Rm = MI.getOperand(3).getReg();
if (Rm)
return (ARM_AM::getAM3Op(MI.getOperand(4).getImm()) == ARM_AM::sub) ? 4
: 3;
return (Rt == Rn) ? 3 : 2;
}
case ARM::STRD: {
Register Rm = MI.getOperand(3).getReg();
if (Rm)
return (ARM_AM::getAM3Op(MI.getOperand(4).getImm()) == ARM_AM::sub) ? 4
: 3;
return 2;
}
case ARM::LDRD_POST:
case ARM::t2LDRD_POST:
return 3;
case ARM::STRD_POST:
case ARM::t2STRD_POST:
return 4;
case ARM::LDRD_PRE: {
Register Rt = MI.getOperand(0).getReg();
Register Rn = MI.getOperand(3).getReg();
Register Rm = MI.getOperand(4).getReg();
if (Rm)
return (ARM_AM::getAM3Op(MI.getOperand(5).getImm()) == ARM_AM::sub) ? 5
: 4;
return (Rt == Rn) ? 4 : 3;
}
case ARM::t2LDRD_PRE: {
Register Rt = MI.getOperand(0).getReg();
Register Rn = MI.getOperand(3).getReg();
return (Rt == Rn) ? 4 : 3;
}
case ARM::STRD_PRE: {
Register Rm = MI.getOperand(4).getReg();
if (Rm)
return (ARM_AM::getAM3Op(MI.getOperand(5).getImm()) == ARM_AM::sub) ? 5
: 4;
return 3;
}
case ARM::t2STRD_PRE:
return 3;
case ARM::t2LDR_POST:
case ARM::t2LDRB_POST:
case ARM::t2LDRB_PRE:
case ARM::t2LDRSBi12:
case ARM::t2LDRSBi8:
case ARM::t2LDRSBpci:
case ARM::t2LDRSBs:
case ARM::t2LDRH_POST:
case ARM::t2LDRH_PRE:
case ARM::t2LDRSBT:
case ARM::t2LDRSB_POST:
case ARM::t2LDRSB_PRE:
case ARM::t2LDRSH_POST:
case ARM::t2LDRSH_PRE:
case ARM::t2LDRSHi12:
case ARM::t2LDRSHi8:
case ARM::t2LDRSHpci:
case ARM::t2LDRSHs:
return 2;
case ARM::t2LDRDi8: {
Register Rt = MI.getOperand(0).getReg();
Register Rn = MI.getOperand(2).getReg();
return (Rt == Rn) ? 3 : 2;
}
case ARM::t2STRB_POST:
case ARM::t2STRB_PRE:
case ARM::t2STRBs:
case ARM::t2STRDi8:
case ARM::t2STRH_POST:
case ARM::t2STRH_PRE:
case ARM::t2STRHs:
case ARM::t2STR_POST:
case ARM::t2STR_PRE:
case ARM::t2STRs:
return 2;
}
}
// Return the number of 32-bit words loaded by LDM or stored by STM. If this
// can't be easily determined return 0 (missing MachineMemOperand).
//
// FIXME: The current MachineInstr design does not support relying on machine
// mem operands to determine the width of a memory access. Instead, we expect
// the target to provide this information based on the instruction opcode and
// operands. However, using MachineMemOperand is the best solution now for
// two reasons:
//
// 1) getNumMicroOps tries to infer LDM memory width from the total number of MI
// operands. This is much more dangerous than using the MachineMemOperand
// sizes because CodeGen passes can insert/remove optional machine operands. In
// fact, it's totally incorrect for preRA passes and appears to be wrong for
// postRA passes as well.
//
// 2) getNumLDMAddresses is only used by the scheduling machine model and any
// machine model that calls this should handle the unknown (zero size) case.
//
// Long term, we should require a target hook that verifies MachineMemOperand
// sizes during MC lowering. That target hook should be local to MC lowering
// because we can't ensure that it is aware of other MI forms. Doing this will
// ensure that MachineMemOperands are correctly propagated through all passes.
unsigned ARMBaseInstrInfo::getNumLDMAddresses(const MachineInstr &MI) const {
unsigned Size = 0;
for (MachineInstr::mmo_iterator I = MI.memoperands_begin(),
E = MI.memoperands_end();
I != E; ++I) {
Size += (*I)->getSize();
}
// FIXME: The scheduler currently can't handle values larger than 16. But
// the values can actually go up to 32 for floating-point load/store
// multiple (VLDMIA etc.). Also, the way this code is reasoning about memory
// operations isn't right; we could end up with "extra" memory operands for
// various reasons, like tail merge merging two memory operations.
return std::min(Size / 4, 16U);
}
static unsigned getNumMicroOpsSingleIssuePlusExtras(unsigned Opc,
unsigned NumRegs) {
unsigned UOps = 1 + NumRegs; // 1 for address computation.
switch (Opc) {
default:
break;
case ARM::VLDMDIA_UPD:
case ARM::VLDMDDB_UPD:
case ARM::VLDMSIA_UPD:
case ARM::VLDMSDB_UPD:
case ARM::VSTMDIA_UPD:
case ARM::VSTMDDB_UPD:
case ARM::VSTMSIA_UPD:
case ARM::VSTMSDB_UPD:
case ARM::LDMIA_UPD:
case ARM::LDMDA_UPD:
case ARM::LDMDB_UPD:
case ARM::LDMIB_UPD:
case ARM::STMIA_UPD:
case ARM::STMDA_UPD:
case ARM::STMDB_UPD:
case ARM::STMIB_UPD:
case ARM::tLDMIA_UPD:
case ARM::tSTMIA_UPD:
case ARM::t2LDMIA_UPD:
case ARM::t2LDMDB_UPD:
case ARM::t2STMIA_UPD:
case ARM::t2STMDB_UPD:
++UOps; // One for base register writeback.
break;
case ARM::LDMIA_RET:
case ARM::tPOP_RET:
case ARM::t2LDMIA_RET:
UOps += 2; // One for base reg wb, one for write to pc.
break;
}
return UOps;
}
unsigned ARMBaseInstrInfo::getNumMicroOps(const InstrItineraryData *ItinData,
const MachineInstr &MI) const {
if (!ItinData || ItinData->isEmpty())
return 1;
const MCInstrDesc &Desc = MI.getDesc();
unsigned Class = Desc.getSchedClass();
int ItinUOps = ItinData->getNumMicroOps(Class);
if (ItinUOps >= 0) {
if (Subtarget.isSwift() && (Desc.mayLoad() || Desc.mayStore()))
return getNumMicroOpsSwiftLdSt(ItinData, MI);
return ItinUOps;
}
unsigned Opc = MI.getOpcode();
switch (Opc) {
default:
llvm_unreachable("Unexpected multi-uops instruction!");
case ARM::VLDMQIA:
case ARM::VSTMQIA:
return 2;
// The number of uOps for load / store multiple are determined by the number
// registers.
//
// On Cortex-A8, each pair of register loads / stores can be scheduled on the
// same cycle. The scheduling for the first load / store must be done
// separately by assuming the address is not 64-bit aligned.
//
// On Cortex-A9, the formula is simply (#reg / 2) + (#reg % 2). If the address
// is not 64-bit aligned, then AGU would take an extra cycle. For VFP / NEON
// load / store multiple, the formula is (#reg / 2) + (#reg % 2) + 1.
case ARM::VLDMDIA:
case ARM::VLDMDIA_UPD:
case ARM::VLDMDDB_UPD:
case ARM::VLDMSIA:
case ARM::VLDMSIA_UPD:
case ARM::VLDMSDB_UPD:
case ARM::VSTMDIA:
case ARM::VSTMDIA_UPD:
case ARM::VSTMDDB_UPD:
case ARM::VSTMSIA:
case ARM::VSTMSIA_UPD:
case ARM::VSTMSDB_UPD: {
unsigned NumRegs = MI.getNumOperands() - Desc.getNumOperands();
return (NumRegs / 2) + (NumRegs % 2) + 1;
}
case ARM::LDMIA_RET:
case ARM::LDMIA:
case ARM::LDMDA:
case ARM::LDMDB:
case ARM::LDMIB:
case ARM::LDMIA_UPD:
case ARM::LDMDA_UPD:
case ARM::LDMDB_UPD:
case ARM::LDMIB_UPD:
case ARM::STMIA:
case ARM::STMDA:
case ARM::STMDB:
case ARM::STMIB:
case ARM::STMIA_UPD:
case ARM::STMDA_UPD:
case ARM::STMDB_UPD:
case ARM::STMIB_UPD:
case ARM::tLDMIA:
case ARM::tLDMIA_UPD:
case ARM::tSTMIA_UPD:
case ARM::tPOP_RET:
case ARM::tPOP:
case ARM::tPUSH:
case ARM::t2LDMIA_RET:
case ARM::t2LDMIA:
case ARM::t2LDMDB:
case ARM::t2LDMIA_UPD:
case ARM::t2LDMDB_UPD:
case ARM::t2STMIA:
case ARM::t2STMDB:
case ARM::t2STMIA_UPD:
case ARM::t2STMDB_UPD: {
unsigned NumRegs = MI.getNumOperands() - Desc.getNumOperands() + 1;
switch (Subtarget.getLdStMultipleTiming()) {
case ARMSubtarget::SingleIssuePlusExtras:
return getNumMicroOpsSingleIssuePlusExtras(Opc, NumRegs);
case ARMSubtarget::SingleIssue:
// Assume the worst.
return NumRegs;
case ARMSubtarget::DoubleIssue: {
if (NumRegs < 4)
return 2;
// 4 registers would be issued: 2, 2.
// 5 registers would be issued: 2, 2, 1.
unsigned UOps = (NumRegs / 2);
if (NumRegs % 2)
++UOps;
return UOps;
}
case ARMSubtarget::DoubleIssueCheckUnalignedAccess: {
unsigned UOps = (NumRegs / 2);
// If there are odd number of registers or if it's not 64-bit aligned,
// then it takes an extra AGU (Address Generation Unit) cycle.
if ((NumRegs % 2) || !MI.hasOneMemOperand() ||
(*MI.memoperands_begin())->getAlignment() < 8)
++UOps;
return UOps;
}
}
}
}
llvm_unreachable("Didn't find the number of microops");
}
int
ARMBaseInstrInfo::getVLDMDefCycle(const InstrItineraryData *ItinData,
const MCInstrDesc &DefMCID,
unsigned DefClass,
unsigned DefIdx, unsigned DefAlign) const {
int RegNo = (int)(DefIdx+1) - DefMCID.getNumOperands() + 1;
if (RegNo <= 0)
// Def is the address writeback.
return ItinData->getOperandCycle(DefClass, DefIdx);
int DefCycle;
if (Subtarget.isCortexA8() || Subtarget.isCortexA7()) {
// (regno / 2) + (regno % 2) + 1
DefCycle = RegNo / 2 + 1;
if (RegNo % 2)
++DefCycle;
} else if (Subtarget.isLikeA9() || Subtarget.isSwift()) {
DefCycle = RegNo;
bool isSLoad = false;
switch (DefMCID.getOpcode()) {
default: break;
case ARM::VLDMSIA:
case ARM::VLDMSIA_UPD:
case ARM::VLDMSDB_UPD:
isSLoad = true;
break;
}
// If there are odd number of 'S' registers or if it's not 64-bit aligned,
// then it takes an extra cycle.
if ((isSLoad && (RegNo % 2)) || DefAlign < 8)
++DefCycle;
} else {
// Assume the worst.
DefCycle = RegNo + 2;
}
return DefCycle;
}
bool ARMBaseInstrInfo::isLDMBaseRegInList(const MachineInstr &MI) const {
Register BaseReg = MI.getOperand(0).getReg();
for (unsigned i = 1, sz = MI.getNumOperands(); i < sz; ++i) {
const auto &Op = MI.getOperand(i);
if (Op.isReg() && Op.getReg() == BaseReg)
return true;
}
return false;
}
unsigned
ARMBaseInstrInfo::getLDMVariableDefsSize(const MachineInstr &MI) const {
// ins GPR:$Rn, $p (2xOp), reglist:$regs, variable_ops
// (outs GPR:$wb), (ins GPR:$Rn, $p (2xOp), reglist:$regs, variable_ops)
return MI.getNumOperands() + 1 - MI.getDesc().getNumOperands();
}
int
ARMBaseInstrInfo::getLDMDefCycle(const InstrItineraryData *ItinData,
const MCInstrDesc &DefMCID,
unsigned DefClass,
unsigned DefIdx, unsigned DefAlign) const {
int RegNo = (int)(DefIdx+1) - DefMCID.getNumOperands() + 1;
if (RegNo <= 0)
// Def is the address writeback.
return ItinData->getOperandCycle(DefClass, DefIdx);
int DefCycle;
if (Subtarget.isCortexA8() || Subtarget.isCortexA7()) {
// 4 registers would be issued: 1, 2, 1.
// 5 registers would be issued: 1, 2, 2.
DefCycle = RegNo / 2;
if (DefCycle < 1)
DefCycle = 1;
// Result latency is issue cycle + 2: E2.
DefCycle += 2;
} else if (Subtarget.isLikeA9() || Subtarget.isSwift()) {
DefCycle = (RegNo / 2);
// If there are odd number of registers or if it's not 64-bit aligned,
// then it takes an extra AGU (Address Generation Unit) cycle.
if ((RegNo % 2) || DefAlign < 8)
++DefCycle;
// Result latency is AGU cycles + 2.
DefCycle += 2;
} else {
// Assume the worst.
DefCycle = RegNo + 2;
}
return DefCycle;
}
int
ARMBaseInstrInfo::getVSTMUseCycle(const InstrItineraryData *ItinData,
const MCInstrDesc &UseMCID,
unsigned UseClass,
unsigned UseIdx, unsigned UseAlign) const {
int RegNo = (int)(UseIdx+1) - UseMCID.getNumOperands() + 1;
if (RegNo <= 0)
return ItinData->getOperandCycle(UseClass, UseIdx);
int UseCycle;
if (Subtarget.isCortexA8() || Subtarget.isCortexA7()) {
// (regno / 2) + (regno % 2) + 1
UseCycle = RegNo / 2 + 1;
if (RegNo % 2)
++UseCycle;
} else if (Subtarget.isLikeA9() || Subtarget.isSwift()) {
UseCycle = RegNo;
bool isSStore = false;
switch (UseMCID.getOpcode()) {
default: break;
case ARM::VSTMSIA:
case ARM::VSTMSIA_UPD:
case ARM::VSTMSDB_UPD:
isSStore = true;
break;
}
// If there are odd number of 'S' registers or if it's not 64-bit aligned,
// then it takes an extra cycle.
if ((isSStore && (RegNo % 2)) || UseAlign < 8)
++UseCycle;
} else {
// Assume the worst.
UseCycle = RegNo + 2;
}
return UseCycle;
}
int
ARMBaseInstrInfo::getSTMUseCycle(const InstrItineraryData *ItinData,
const MCInstrDesc &UseMCID,
unsigned UseClass,
unsigned UseIdx, unsigned UseAlign) const {
int RegNo = (int)(UseIdx+1) - UseMCID.getNumOperands() + 1;
if (RegNo <= 0)
return ItinData->getOperandCycle(UseClass, UseIdx);
int UseCycle;
if (Subtarget.isCortexA8() || Subtarget.isCortexA7()) {
UseCycle = RegNo / 2;
if (UseCycle < 2)
UseCycle = 2;
// Read in E3.
UseCycle += 2;
} else if (Subtarget.isLikeA9() || Subtarget.isSwift()) {
UseCycle = (RegNo / 2);
// If there are odd number of registers or if it's not 64-bit aligned,
// then it takes an extra AGU (Address Generation Unit) cycle.
if ((RegNo % 2) || UseAlign < 8)
++UseCycle;
} else {
// Assume the worst.
UseCycle = 1;
}
return UseCycle;
}
int
ARMBaseInstrInfo::getOperandLatency(const InstrItineraryData *ItinData,
const MCInstrDesc &DefMCID,
unsigned DefIdx, unsigned DefAlign,
const MCInstrDesc &UseMCID,
unsigned UseIdx, unsigned UseAlign) const {
unsigned DefClass = DefMCID.getSchedClass();
unsigned UseClass = UseMCID.getSchedClass();
if (DefIdx < DefMCID.getNumDefs() && UseIdx < UseMCID.getNumOperands())
return ItinData->getOperandLatency(DefClass, DefIdx, UseClass, UseIdx);
// This may be a def / use of a variable_ops instruction, the operand
// latency might be determinable dynamically. Let the target try to
// figure it out.
int DefCycle = -1;
bool LdmBypass = false;
switch (DefMCID.getOpcode()) {
default:
DefCycle = ItinData->getOperandCycle(DefClass, DefIdx);
break;
case ARM::VLDMDIA:
case ARM::VLDMDIA_UPD:
case ARM::VLDMDDB_UPD:
case ARM::VLDMSIA:
case ARM::VLDMSIA_UPD:
case ARM::VLDMSDB_UPD:
DefCycle = getVLDMDefCycle(ItinData, DefMCID, DefClass, DefIdx, DefAlign);
break;
case ARM::LDMIA_RET:
case ARM::LDMIA:
case ARM::LDMDA:
case ARM::LDMDB:
case ARM::LDMIB:
case ARM::LDMIA_UPD:
case ARM::LDMDA_UPD:
case ARM::LDMDB_UPD:
case ARM::LDMIB_UPD:
case ARM::tLDMIA:
case ARM::tLDMIA_UPD:
case ARM::tPUSH:
case ARM::t2LDMIA_RET:
case ARM::t2LDMIA:
case ARM::t2LDMDB:
case ARM::t2LDMIA_UPD:
case ARM::t2LDMDB_UPD:
LdmBypass = true;
DefCycle = getLDMDefCycle(ItinData, DefMCID, DefClass, DefIdx, DefAlign);
break;
}
if (DefCycle == -1)
// We can't seem to determine the result latency of the def, assume it's 2.
DefCycle = 2;
int UseCycle = -1;
switch (UseMCID.getOpcode()) {
default:
UseCycle = ItinData->getOperandCycle(UseClass, UseIdx);
break;
case ARM::VSTMDIA:
case ARM::VSTMDIA_UPD:
case ARM::VSTMDDB_UPD:
case ARM::VSTMSIA:
case ARM::VSTMSIA_UPD:
case ARM::VSTMSDB_UPD:
UseCycle = getVSTMUseCycle(ItinData, UseMCID, UseClass, UseIdx, UseAlign);
break;
case ARM::STMIA:
case ARM::STMDA:
case ARM::STMDB:
case ARM::STMIB:
case ARM::STMIA_UPD:
case ARM::STMDA_UPD:
case ARM::STMDB_UPD:
case ARM::STMIB_UPD:
case ARM::tSTMIA_UPD:
case ARM::tPOP_RET:
case ARM::tPOP:
case ARM::t2STMIA:
case ARM::t2STMDB:
case ARM::t2STMIA_UPD:
case ARM::t2STMDB_UPD:
UseCycle = getSTMUseCycle(ItinData, UseMCID, UseClass, UseIdx, UseAlign);
break;
}
if (UseCycle == -1)
// Assume it's read in the first stage.
UseCycle = 1;
UseCycle = DefCycle - UseCycle + 1;
if (UseCycle > 0) {
if (LdmBypass) {
// It's a variable_ops instruction so we can't use DefIdx here. Just use
// first def operand.
if (ItinData->hasPipelineForwarding(DefClass, DefMCID.getNumOperands()-1,
UseClass, UseIdx))
--UseCycle;
} else if (ItinData->hasPipelineForwarding(DefClass, DefIdx,
UseClass, UseIdx)) {
--UseCycle;
}
}
return UseCycle;
}
static const MachineInstr *getBundledDefMI(const TargetRegisterInfo *TRI,
const MachineInstr *MI, unsigned Reg,
unsigned &DefIdx, unsigned &Dist) {
Dist = 0;
MachineBasicBlock::const_iterator I = MI; ++I;
MachineBasicBlock::const_instr_iterator II = std::prev(I.getInstrIterator());
assert(II->isInsideBundle() && "Empty bundle?");
int Idx = -1;
while (II->isInsideBundle()) {
Idx = II->findRegisterDefOperandIdx(Reg, false, true, TRI);
if (Idx != -1)
break;
--II;
++Dist;
}
assert(Idx != -1 && "Cannot find bundled definition!");
DefIdx = Idx;
return &*II;
}
static const MachineInstr *getBundledUseMI(const TargetRegisterInfo *TRI,
const MachineInstr &MI, unsigned Reg,
unsigned &UseIdx, unsigned &Dist) {
Dist = 0;
MachineBasicBlock::const_instr_iterator II = ++MI.getIterator();
assert(II->isInsideBundle() && "Empty bundle?");
MachineBasicBlock::const_instr_iterator E = MI.getParent()->instr_end();
// FIXME: This doesn't properly handle multiple uses.
int Idx = -1;
while (II != E && II->isInsideBundle()) {
Idx = II->findRegisterUseOperandIdx(Reg, false, TRI);
if (Idx != -1)
break;
if (II->getOpcode() != ARM::t2IT)
++Dist;
++II;
}
if (Idx == -1) {
Dist = 0;
return nullptr;
}
UseIdx = Idx;
return &*II;
}
/// Return the number of cycles to add to (or subtract from) the static
/// itinerary based on the def opcode and alignment. The caller will ensure that
/// adjusted latency is at least one cycle.
static int adjustDefLatency(const ARMSubtarget &Subtarget,
const MachineInstr &DefMI,
const MCInstrDesc &DefMCID, unsigned DefAlign) {
int Adjust = 0;
if (Subtarget.isCortexA8() || Subtarget.isLikeA9() || Subtarget.isCortexA7()) {
// FIXME: Shifter op hack: no shift (i.e. [r +/- r]) or [r + r << 2]
// variants are one cycle cheaper.
switch (DefMCID.getOpcode()) {
default: break;
case ARM::LDRrs:
case ARM::LDRBrs: {
unsigned ShOpVal = DefMI.getOperand(3).getImm();
unsigned ShImm = ARM_AM::getAM2Offset(ShOpVal);
if (ShImm == 0 ||
(ShImm == 2 && ARM_AM::getAM2ShiftOpc(ShOpVal) == ARM_AM::lsl))
--Adjust;
break;
}
case ARM::t2LDRs:
case ARM::t2LDRBs:
case ARM::t2LDRHs:
case ARM::t2LDRSHs: {
// Thumb2 mode: lsl only.
unsigned ShAmt = DefMI.getOperand(3).getImm();
if (ShAmt == 0 || ShAmt == 2)
--Adjust;
break;
}
}
} else if (Subtarget.isSwift()) {
// FIXME: Properly handle all of the latency adjustments for address
// writeback.
switch (DefMCID.getOpcode()) {
default: break;
case ARM::LDRrs:
case ARM::LDRBrs: {
unsigned ShOpVal = DefMI.getOperand(3).getImm();
bool isSub = ARM_AM::getAM2Op(ShOpVal) == ARM_AM::sub;
unsigned ShImm = ARM_AM::getAM2Offset(ShOpVal);
if (!isSub &&
(ShImm == 0 ||
((ShImm == 1 || ShImm == 2 || ShImm == 3) &&
ARM_AM::getAM2ShiftOpc(ShOpVal) == ARM_AM::lsl)))
Adjust -= 2;
else if (!isSub &&
ShImm == 1 && ARM_AM::getAM2ShiftOpc(ShOpVal) == ARM_AM::lsr)
--Adjust;
break;
}
case ARM::t2LDRs:
case ARM::t2LDRBs:
case ARM::t2LDRHs:
case ARM::t2LDRSHs: {
// Thumb2 mode: lsl only.
unsigned ShAmt = DefMI.getOperand(3).getImm();
if (ShAmt == 0 || ShAmt == 1 || ShAmt == 2 || ShAmt == 3)
Adjust -= 2;
break;
}
}
}
if (DefAlign < 8 && Subtarget.checkVLDnAccessAlignment()) {
switch (DefMCID.getOpcode()) {
default: break;
case ARM::VLD1q8:
case ARM::VLD1q16:
case ARM::VLD1q32:
case ARM::VLD1q64:
case ARM::VLD1q8wb_fixed:
case ARM::VLD1q16wb_fixed:
case ARM::VLD1q32wb_fixed:
case ARM::VLD1q64wb_fixed:
case ARM::VLD1q8wb_register:
case ARM::VLD1q16wb_register:
case ARM::VLD1q32wb_register:
case ARM::VLD1q64wb_register:
case ARM::VLD2d8:
case ARM::VLD2d16:
case ARM::VLD2d32:
case ARM::VLD2q8:
case ARM::VLD2q16:
case ARM::VLD2q32:
case ARM::VLD2d8wb_fixed:
case ARM::VLD2d16wb_fixed:
case ARM::VLD2d32wb_fixed:
case ARM::VLD2q8wb_fixed:
case ARM::VLD2q16wb_fixed:
case ARM::VLD2q32wb_fixed:
case ARM::VLD2d8wb_register:
case ARM::VLD2d16wb_register:
case ARM::VLD2d32wb_register:
case ARM::VLD2q8wb_register:
case ARM::VLD2q16wb_register:
case ARM::VLD2q32wb_register:
case ARM::VLD3d8:
case ARM::VLD3d16:
case ARM::VLD3d32:
case ARM::VLD1d64T:
case ARM::VLD3d8_UPD:
case ARM::VLD3d16_UPD:
case ARM::VLD3d32_UPD:
case ARM::VLD1d64Twb_fixed:
case ARM::VLD1d64Twb_register:
case ARM::VLD3q8_UPD:
case ARM::VLD3q16_UPD:
case ARM::VLD3q32_UPD:
case ARM::VLD4d8:
case ARM::VLD4d16:
case ARM::VLD4d32:
case ARM::VLD1d64Q:
case ARM::VLD4d8_UPD:
case ARM::VLD4d16_UPD:
case ARM::VLD4d32_UPD:
case ARM::VLD1d64Qwb_fixed:
case ARM::VLD1d64Qwb_register:
case ARM::VLD4q8_UPD:
case ARM::VLD4q16_UPD:
case ARM::VLD4q32_UPD:
case ARM::VLD1DUPq8:
case ARM::VLD1DUPq16:
case ARM::VLD1DUPq32:
case ARM::VLD1DUPq8wb_fixed:
case ARM::VLD1DUPq16wb_fixed:
case ARM::VLD1DUPq32wb_fixed:
case ARM::VLD1DUPq8wb_register:
case ARM::VLD1DUPq16wb_register:
case ARM::VLD1DUPq32wb_register:
case ARM::VLD2DUPd8:
case ARM::VLD2DUPd16:
case ARM::VLD2DUPd32:
case ARM::VLD2DUPd8wb_fixed:
case ARM::VLD2DUPd16wb_fixed:
case ARM::VLD2DUPd32wb_fixed:
case ARM::VLD2DUPd8wb_register:
case ARM::VLD2DUPd16wb_register:
case ARM::VLD2DUPd32wb_register:
case ARM::VLD4DUPd8:
case ARM::VLD4DUPd16:
case ARM::VLD4DUPd32:
case ARM::VLD4DUPd8_UPD:
case ARM::VLD4DUPd16_UPD:
case ARM::VLD4DUPd32_UPD:
case ARM::VLD1LNd8:
case ARM::VLD1LNd16:
case ARM::VLD1LNd32:
case ARM::VLD1LNd8_UPD:
case ARM::VLD1LNd16_UPD:
case ARM::VLD1LNd32_UPD:
case ARM::VLD2LNd8:
case ARM::VLD2LNd16:
case ARM::VLD2LNd32:
case ARM::VLD2LNq16:
case ARM::VLD2LNq32:
case ARM::VLD2LNd8_UPD:
case ARM::VLD2LNd16_UPD:
case ARM::VLD2LNd32_UPD:
case ARM::VLD2LNq16_UPD:
case ARM::VLD2LNq32_UPD:
case ARM::VLD4LNd8:
case ARM::VLD4LNd16:
case ARM::VLD4LNd32:
case ARM::VLD4LNq16:
case ARM::VLD4LNq32:
case ARM::VLD4LNd8_UPD:
case ARM::VLD4LNd16_UPD:
case ARM::VLD4LNd32_UPD:
case ARM::VLD4LNq16_UPD:
case ARM::VLD4LNq32_UPD:
// If the address is not 64-bit aligned, the latencies of these
// instructions increases by one.
++Adjust;
break;
}
}
return Adjust;
}
int ARMBaseInstrInfo::getOperandLatency(const InstrItineraryData *ItinData,
const MachineInstr &DefMI,
unsigned DefIdx,
const MachineInstr &UseMI,
unsigned UseIdx) const {
// No operand latency. The caller may fall back to getInstrLatency.
if (!ItinData || ItinData->isEmpty())
return -1;
const MachineOperand &DefMO = DefMI.getOperand(DefIdx);
Register Reg = DefMO.getReg();
const MachineInstr *ResolvedDefMI = &DefMI;
unsigned DefAdj = 0;
if (DefMI.isBundle())
ResolvedDefMI =
getBundledDefMI(&getRegisterInfo(), &DefMI, Reg, DefIdx, DefAdj);
if (ResolvedDefMI->isCopyLike() || ResolvedDefMI->isInsertSubreg() ||
ResolvedDefMI->isRegSequence() || ResolvedDefMI->isImplicitDef()) {
return 1;
}
const MachineInstr *ResolvedUseMI = &UseMI;
unsigned UseAdj = 0;
if (UseMI.isBundle()) {
ResolvedUseMI =
getBundledUseMI(&getRegisterInfo(), UseMI, Reg, UseIdx, UseAdj);
if (!ResolvedUseMI)
return -1;
}
return getOperandLatencyImpl(
ItinData, *ResolvedDefMI, DefIdx, ResolvedDefMI->getDesc(), DefAdj, DefMO,
Reg, *ResolvedUseMI, UseIdx, ResolvedUseMI->getDesc(), UseAdj);
}
int ARMBaseInstrInfo::getOperandLatencyImpl(
const InstrItineraryData *ItinData, const MachineInstr &DefMI,
unsigned DefIdx, const MCInstrDesc &DefMCID, unsigned DefAdj,
const MachineOperand &DefMO, unsigned Reg, const MachineInstr &UseMI,
unsigned UseIdx, const MCInstrDesc &UseMCID, unsigned UseAdj) const {
if (Reg == ARM::CPSR) {
if (DefMI.getOpcode() == ARM::FMSTAT) {
// fpscr -> cpsr stalls over 20 cycles on A8 (and earlier?)
return Subtarget.isLikeA9() ? 1 : 20;
}
// CPSR set and branch can be paired in the same cycle.
if (UseMI.isBranch())
return 0;
// Otherwise it takes the instruction latency (generally one).
unsigned Latency = getInstrLatency(ItinData, DefMI);
// For Thumb2 and -Os, prefer scheduling CPSR setting instruction close to
// its uses. Instructions which are otherwise scheduled between them may
// incur a code size penalty (not able to use the CPSR setting 16-bit
// instructions).
if (Latency > 0 && Subtarget.isThumb2()) {
const MachineFunction *MF = DefMI.getParent()->getParent();
// FIXME: Use Function::hasOptSize().
if (MF->getFunction().hasFnAttribute(Attribute::OptimizeForSize))
--Latency;
}
return Latency;
}
if (DefMO.isImplicit() || UseMI.getOperand(UseIdx).isImplicit())
return -1;
unsigned DefAlign = DefMI.hasOneMemOperand()
? (*DefMI.memoperands_begin())->getAlignment()
: 0;
unsigned UseAlign = UseMI.hasOneMemOperand()
? (*UseMI.memoperands_begin())->getAlignment()
: 0;
// Get the itinerary's latency if possible, and handle variable_ops.
int Latency = getOperandLatency(ItinData, DefMCID, DefIdx, DefAlign, UseMCID,
UseIdx, UseAlign);
// Unable to find operand latency. The caller may resort to getInstrLatency.
if (Latency < 0)
return Latency;
// Adjust for IT block position.
int Adj = DefAdj + UseAdj;
// Adjust for dynamic def-side opcode variants not captured by the itinerary.
Adj += adjustDefLatency(Subtarget, DefMI, DefMCID, DefAlign);
if (Adj >= 0 || (int)Latency > -Adj) {
return Latency + Adj;
}
// Return the itinerary latency, which may be zero but not less than zero.
return Latency;
}
int
ARMBaseInstrInfo::getOperandLatency(const InstrItineraryData *ItinData,
SDNode *DefNode, unsigned DefIdx,
SDNode *UseNode, unsigned UseIdx) const {
if (!DefNode->isMachineOpcode())
return 1;
const MCInstrDesc &DefMCID = get(DefNode->getMachineOpcode());
if (isZeroCost(DefMCID.Opcode))
return 0;
if (!ItinData || ItinData->isEmpty())
return DefMCID.mayLoad() ? 3 : 1;
if (!UseNode->isMachineOpcode()) {
int Latency = ItinData->getOperandCycle(DefMCID.getSchedClass(), DefIdx);
int Adj = Subtarget.getPreISelOperandLatencyAdjustment();
int Threshold = 1 + Adj;
return Latency <= Threshold ? 1 : Latency - Adj;
}
const MCInstrDesc &UseMCID = get(UseNode->getMachineOpcode());
auto *DefMN = cast<MachineSDNode>(DefNode);
unsigned DefAlign = !DefMN->memoperands_empty()
? (*DefMN->memoperands_begin())->getAlignment() : 0;
auto *UseMN = cast<MachineSDNode>(UseNode);
unsigned UseAlign = !UseMN->memoperands_empty()
? (*UseMN->memoperands_begin())->getAlignment() : 0;
int Latency = getOperandLatency(ItinData, DefMCID, DefIdx, DefAlign,
UseMCID, UseIdx, UseAlign);
if (Latency > 1 &&
(Subtarget.isCortexA8() || Subtarget.isLikeA9() ||
Subtarget.isCortexA7())) {
// FIXME: Shifter op hack: no shift (i.e. [r +/- r]) or [r + r << 2]
// variants are one cycle cheaper.
switch (DefMCID.getOpcode()) {
default: break;
case ARM::LDRrs:
case ARM::LDRBrs: {
unsigned ShOpVal =
cast<ConstantSDNode>(DefNode->getOperand(2))->getZExtValue();
unsigned ShImm = ARM_AM::getAM2Offset(ShOpVal);
if (ShImm == 0 ||
(ShImm == 2 && ARM_AM::getAM2ShiftOpc(ShOpVal) == ARM_AM::lsl))
--Latency;
break;
}
case ARM::t2LDRs:
case ARM::t2LDRBs:
case ARM::t2LDRHs:
case ARM::t2LDRSHs: {
// Thumb2 mode: lsl only.
unsigned ShAmt =
cast<ConstantSDNode>(DefNode->getOperand(2))->getZExtValue();
if (ShAmt == 0 || ShAmt == 2)
--Latency;
break;
}
}
} else if (DefIdx == 0 && Latency > 2 && Subtarget.isSwift()) {
// FIXME: Properly handle all of the latency adjustments for address
// writeback.
switch (DefMCID.getOpcode()) {
default: break;
case ARM::LDRrs:
case ARM::LDRBrs: {
unsigned ShOpVal =
cast<ConstantSDNode>(DefNode->getOperand(2))->getZExtValue();
unsigned ShImm = ARM_AM::getAM2Offset(ShOpVal);
if (ShImm == 0 ||
((ShImm == 1 || ShImm == 2 || ShImm == 3) &&
ARM_AM::getAM2ShiftOpc(ShOpVal) == ARM_AM::lsl))
Latency -= 2;
else if (ShImm == 1 && ARM_AM::getAM2ShiftOpc(ShOpVal) == ARM_AM::lsr)
--Latency;
break;
}
case ARM::t2LDRs:
case ARM::t2LDRBs:
case ARM::t2LDRHs:
case ARM::t2LDRSHs:
// Thumb2 mode: lsl 0-3 only.
Latency -= 2;
break;
}
}
if (DefAlign < 8 && Subtarget.checkVLDnAccessAlignment())
switch (DefMCID.getOpcode()) {
default: break;
case ARM::VLD1q8:
case ARM::VLD1q16:
case ARM::VLD1q32:
case ARM::VLD1q64:
case ARM::VLD1q8wb_register:
case ARM::VLD1q16wb_register:
case ARM::VLD1q32wb_register:
case ARM::VLD1q64wb_register:
case ARM::VLD1q8wb_fixed:
case ARM::VLD1q16wb_fixed:
case ARM::VLD1q32wb_fixed:
case ARM::VLD1q64wb_fixed:
case ARM::VLD2d8:
case ARM::VLD2d16:
case ARM::VLD2d32:
case ARM::VLD2q8Pseudo:
case ARM::VLD2q16Pseudo:
case ARM::VLD2q32Pseudo:
case ARM::VLD2d8wb_fixed:
case ARM::VLD2d16wb_fixed:
case ARM::VLD2d32wb_fixed:
case ARM::VLD2q8PseudoWB_fixed:
case ARM::VLD2q16PseudoWB_fixed:
case ARM::VLD2q32PseudoWB_fixed:
case ARM::VLD2d8wb_register:
case ARM::VLD2d16wb_register:
case ARM::VLD2d32wb_register:
case ARM::VLD2q8PseudoWB_register:
case ARM::VLD2q16PseudoWB_register:
case ARM::VLD2q32PseudoWB_register:
case ARM::VLD3d8Pseudo:
case ARM::VLD3d16Pseudo:
case ARM::VLD3d32Pseudo:
case ARM::VLD1d8TPseudo:
case ARM::VLD1d16TPseudo:
case ARM::VLD1d32TPseudo:
case ARM::VLD1d64TPseudo:
case ARM::VLD1d64TPseudoWB_fixed:
case ARM::VLD1d64TPseudoWB_register:
case ARM::VLD3d8Pseudo_UPD:
case ARM::VLD3d16Pseudo_UPD:
case ARM::VLD3d32Pseudo_UPD:
case ARM::VLD3q8Pseudo_UPD:
case ARM::VLD3q16Pseudo_UPD:
case ARM::VLD3q32Pseudo_UPD:
case ARM::VLD3q8oddPseudo:
case ARM::VLD3q16oddPseudo:
case ARM::VLD3q32oddPseudo:
case ARM::VLD3q8oddPseudo_UPD:
case ARM::VLD3q16oddPseudo_UPD:
case ARM::VLD3q32oddPseudo_UPD:
case ARM::VLD4d8Pseudo:
case ARM::VLD4d16Pseudo:
case ARM::VLD4d32Pseudo:
case ARM::VLD1d8QPseudo:
case ARM::VLD1d16QPseudo:
case ARM::VLD1d32QPseudo:
case ARM::VLD1d64QPseudo:
case ARM::VLD1d64QPseudoWB_fixed:
case ARM::VLD1d64QPseudoWB_register:
case ARM::VLD1q8HighQPseudo:
case ARM::VLD1q8LowQPseudo_UPD:
case ARM::VLD1q8HighTPseudo:
case ARM::VLD1q8LowTPseudo_UPD:
case ARM::VLD1q16HighQPseudo:
case ARM::VLD1q16LowQPseudo_UPD:
case ARM::VLD1q16HighTPseudo:
case ARM::VLD1q16LowTPseudo_UPD:
case ARM::VLD1q32HighQPseudo:
case ARM::VLD1q32LowQPseudo_UPD:
case ARM::VLD1q32HighTPseudo:
case ARM::VLD1q32LowTPseudo_UPD:
case ARM::VLD1q64HighQPseudo:
case ARM::VLD1q64LowQPseudo_UPD:
case ARM::VLD1q64HighTPseudo:
case ARM::VLD1q64LowTPseudo_UPD:
case ARM::VLD4d8Pseudo_UPD:
case ARM::VLD4d16Pseudo_UPD:
case ARM::VLD4d32Pseudo_UPD:
case ARM::VLD4q8Pseudo_UPD:
case ARM::VLD4q16Pseudo_UPD:
case ARM::VLD4q32Pseudo_UPD:
case ARM::VLD4q8oddPseudo:
case ARM::VLD4q16oddPseudo:
case ARM::VLD4q32oddPseudo:
case ARM::VLD4q8oddPseudo_UPD:
case ARM::VLD4q16oddPseudo_UPD:
case ARM::VLD4q32oddPseudo_UPD:
case ARM::VLD1DUPq8:
case ARM::VLD1DUPq16:
case ARM::VLD1DUPq32:
case ARM::VLD1DUPq8wb_fixed:
case ARM::VLD1DUPq16wb_fixed:
case ARM::VLD1DUPq32wb_fixed:
case ARM::VLD1DUPq8wb_register:
case ARM::VLD1DUPq16wb_register:
case ARM::VLD1DUPq32wb_register:
case ARM::VLD2DUPd8:
case ARM::VLD2DUPd16:
case ARM::VLD2DUPd32:
case ARM::VLD2DUPd8wb_fixed:
case ARM::VLD2DUPd16wb_fixed:
case ARM::VLD2DUPd32wb_fixed:
case ARM::VLD2DUPd8wb_register:
case ARM::VLD2DUPd16wb_register:
case ARM::VLD2DUPd32wb_register:
case ARM::VLD2DUPq8EvenPseudo:
case ARM::VLD2DUPq8OddPseudo:
case ARM::VLD2DUPq16EvenPseudo:
case ARM::VLD2DUPq16OddPseudo:
case ARM::VLD2DUPq32EvenPseudo:
case ARM::VLD2DUPq32OddPseudo:
case ARM::VLD3DUPq8EvenPseudo:
case ARM::VLD3DUPq8OddPseudo:
case ARM::VLD3DUPq16EvenPseudo:
case ARM::VLD3DUPq16OddPseudo:
case ARM::VLD3DUPq32EvenPseudo:
case ARM::VLD3DUPq32OddPseudo:
case ARM::VLD4DUPd8Pseudo:
case ARM::VLD4DUPd16Pseudo:
case ARM::VLD4DUPd32Pseudo:
case ARM::VLD4DUPd8Pseudo_UPD:
case ARM::VLD4DUPd16Pseudo_UPD:
case ARM::VLD4DUPd32Pseudo_UPD:
case ARM::VLD4DUPq8EvenPseudo:
case ARM::VLD4DUPq8OddPseudo:
case ARM::VLD4DUPq16EvenPseudo:
case ARM::VLD4DUPq16OddPseudo:
case ARM::VLD4DUPq32EvenPseudo:
case ARM::VLD4DUPq32OddPseudo:
case ARM::VLD1LNq8Pseudo:
case ARM::VLD1LNq16Pseudo:
case ARM::VLD1LNq32Pseudo:
case ARM::VLD1LNq8Pseudo_UPD:
case ARM::VLD1LNq16Pseudo_UPD:
case ARM::VLD1LNq32Pseudo_UPD:
case ARM::VLD2LNd8Pseudo:
case ARM::VLD2LNd16Pseudo:
case ARM::VLD2LNd32Pseudo:
case ARM::VLD2LNq16Pseudo:
case ARM::VLD2LNq32Pseudo:
case ARM::VLD2LNd8Pseudo_UPD:
case ARM::VLD2LNd16Pseudo_UPD:
case ARM::VLD2LNd32Pseudo_UPD:
case ARM::VLD2LNq16Pseudo_UPD:
case ARM::VLD2LNq32Pseudo_UPD:
case ARM::VLD4LNd8Pseudo:
case ARM::VLD4LNd16Pseudo:
case ARM::VLD4LNd32Pseudo:
case ARM::VLD4LNq16Pseudo:
case ARM::VLD4LNq32Pseudo:
case ARM::VLD4LNd8Pseudo_UPD:
case ARM::VLD4LNd16Pseudo_UPD:
case ARM::VLD4LNd32Pseudo_UPD:
case ARM::VLD4LNq16Pseudo_UPD:
case ARM::VLD4LNq32Pseudo_UPD:
// If the address is not 64-bit aligned, the latencies of these
// instructions increases by one.
++Latency;
break;
}
return Latency;
}
unsigned ARMBaseInstrInfo::getPredicationCost(const MachineInstr &MI) const {
if (MI.isCopyLike() || MI.isInsertSubreg() || MI.isRegSequence() ||
MI.isImplicitDef())
return 0;
if (MI.isBundle())
return 0;
const MCInstrDesc &MCID = MI.getDesc();
if (MCID.isCall() || (MCID.hasImplicitDefOfPhysReg(ARM::CPSR) &&
!Subtarget.cheapPredicableCPSRDef())) {
// When predicated, CPSR is an additional source operand for CPSR updating
// instructions, this apparently increases their latencies.
return 1;
}
return 0;
}
unsigned ARMBaseInstrInfo::getInstrLatency(const InstrItineraryData *ItinData,
const MachineInstr &MI,
unsigned *PredCost) const {
if (MI.isCopyLike() || MI.isInsertSubreg() || MI.isRegSequence() ||
MI.isImplicitDef())
return 1;
// An instruction scheduler typically runs on unbundled instructions, however
// other passes may query the latency of a bundled instruction.
if (MI.isBundle()) {
unsigned Latency = 0;
MachineBasicBlock::const_instr_iterator I = MI.getIterator();
MachineBasicBlock::const_instr_iterator E = MI.getParent()->instr_end();
while (++I != E && I->isInsideBundle()) {
if (I->getOpcode() != ARM::t2IT)
Latency += getInstrLatency(ItinData, *I, PredCost);
}
return Latency;
}
const MCInstrDesc &MCID = MI.getDesc();
if (PredCost && (MCID.isCall() || (MCID.hasImplicitDefOfPhysReg(ARM::CPSR) &&
!Subtarget.cheapPredicableCPSRDef()))) {
// When predicated, CPSR is an additional source operand for CPSR updating
// instructions, this apparently increases their latencies.
*PredCost = 1;
}
// Be sure to call getStageLatency for an empty itinerary in case it has a
// valid MinLatency property.
if (!ItinData)
return MI.mayLoad() ? 3 : 1;
unsigned Class = MCID.getSchedClass();
// For instructions with variable uops, use uops as latency.
if (!ItinData->isEmpty() && ItinData->getNumMicroOps(Class) < 0)
return getNumMicroOps(ItinData, MI);
// For the common case, fall back on the itinerary's latency.
unsigned Latency = ItinData->getStageLatency(Class);
// Adjust for dynamic def-side opcode variants not captured by the itinerary.
unsigned DefAlign =
MI.hasOneMemOperand() ? (*MI.memoperands_begin())->getAlignment() : 0;
int Adj = adjustDefLatency(Subtarget, MI, MCID, DefAlign);
if (Adj >= 0 || (int)Latency > -Adj) {
return Latency + Adj;
}
return Latency;
}
int ARMBaseInstrInfo::getInstrLatency(const InstrItineraryData *ItinData,
SDNode *Node) const {
if (!Node->isMachineOpcode())
return 1;
if (!ItinData || ItinData->isEmpty())
return 1;
unsigned Opcode = Node->getMachineOpcode();
switch (Opcode) {
default:
return ItinData->getStageLatency(get(Opcode).getSchedClass());
case ARM::VLDMQIA:
case ARM::VSTMQIA:
return 2;
}
}
bool ARMBaseInstrInfo::hasHighOperandLatency(const TargetSchedModel &SchedModel,
const MachineRegisterInfo *MRI,
const MachineInstr &DefMI,
unsigned DefIdx,
const MachineInstr &UseMI,
unsigned UseIdx) const {
unsigned DDomain = DefMI.getDesc().TSFlags & ARMII::DomainMask;
unsigned UDomain = UseMI.getDesc().TSFlags & ARMII::DomainMask;
if (Subtarget.nonpipelinedVFP() &&
(DDomain == ARMII::DomainVFP || UDomain == ARMII::DomainVFP))
return true;
// Hoist VFP / NEON instructions with 4 or higher latency.
unsigned Latency =
SchedModel.computeOperandLatency(&DefMI, DefIdx, &UseMI, UseIdx);
if (Latency <= 3)
return false;
return DDomain == ARMII::DomainVFP || DDomain == ARMII::DomainNEON ||
UDomain == ARMII::DomainVFP || UDomain == ARMII::DomainNEON;
}
bool ARMBaseInstrInfo::hasLowDefLatency(const TargetSchedModel &SchedModel,
const MachineInstr &DefMI,
unsigned DefIdx) const {
const InstrItineraryData *ItinData = SchedModel.getInstrItineraries();
if (!ItinData || ItinData->isEmpty())
return false;
unsigned DDomain = DefMI.getDesc().TSFlags & ARMII::DomainMask;
if (DDomain == ARMII::DomainGeneral) {
unsigned DefClass = DefMI.getDesc().getSchedClass();
int DefCycle = ItinData->getOperandCycle(DefClass, DefIdx);
return (DefCycle != -1 && DefCycle <= 2);
}
return false;
}
bool ARMBaseInstrInfo::verifyInstruction(const MachineInstr &MI,
StringRef &ErrInfo) const {
if (convertAddSubFlagsOpcode(MI.getOpcode())) {
ErrInfo = "Pseudo flag setting opcodes only exist in Selection DAG";
return false;
}
if (MI.getOpcode() == ARM::tMOVr && !Subtarget.hasV6Ops()) {
// Make sure we don't generate a lo-lo mov that isn't supported.
if (!ARM::hGPRRegClass.contains(MI.getOperand(0).getReg()) &&
!ARM::hGPRRegClass.contains(MI.getOperand(1).getReg())) {
ErrInfo = "Non-flag-setting Thumb1 mov is v6-only";
return false;
}
}
if (MI.getOpcode() == ARM::tPUSH ||
MI.getOpcode() == ARM::tPOP ||
MI.getOpcode() == ARM::tPOP_RET) {
for (int i = 2, e = MI.getNumOperands(); i < e; ++i) {
if (MI.getOperand(i).isImplicit() ||
!MI.getOperand(i).isReg())
continue;
Register Reg = MI.getOperand(i).getReg();
if (Reg < ARM::R0 || Reg > ARM::R7) {
if (!(MI.getOpcode() == ARM::tPUSH && Reg == ARM::LR) &&
!(MI.getOpcode() == ARM::tPOP_RET && Reg == ARM::PC)) {
ErrInfo = "Unsupported register in Thumb1 push/pop";
return false;
}
}
}
}
return true;
}
// LoadStackGuard has so far only been implemented for MachO. Different code
// sequence is needed for other targets.
void ARMBaseInstrInfo::expandLoadStackGuardBase(MachineBasicBlock::iterator MI,
unsigned LoadImmOpc,
unsigned LoadOpc) const {
assert(!Subtarget.isROPI() && !Subtarget.isRWPI() &&
"ROPI/RWPI not currently supported with stack guard");
MachineBasicBlock &MBB = *MI->getParent();
DebugLoc DL = MI->getDebugLoc();
Register Reg = MI->getOperand(0).getReg();
const GlobalValue *GV =
cast<GlobalValue>((*MI->memoperands_begin())->getValue());
MachineInstrBuilder MIB;
BuildMI(MBB, MI, DL, get(LoadImmOpc), Reg)
.addGlobalAddress(GV, 0, ARMII::MO_NONLAZY);
if (Subtarget.isGVIndirectSymbol(GV)) {
MIB = BuildMI(MBB, MI, DL, get(LoadOpc), Reg);
MIB.addReg(Reg, RegState::Kill).addImm(0);
auto Flags = MachineMemOperand::MOLoad |
MachineMemOperand::MODereferenceable |
MachineMemOperand::MOInvariant;
MachineMemOperand *MMO = MBB.getParent()->getMachineMemOperand(
MachinePointerInfo::getGOT(*MBB.getParent()), Flags, 4, 4);
MIB.addMemOperand(MMO).add(predOps(ARMCC::AL));
}
MIB = BuildMI(MBB, MI, DL, get(LoadOpc), Reg);
MIB.addReg(Reg, RegState::Kill)
.addImm(0)
.cloneMemRefs(*MI)
.add(predOps(ARMCC::AL));
}
bool
ARMBaseInstrInfo::isFpMLxInstruction(unsigned Opcode, unsigned &MulOpc,
unsigned &AddSubOpc,
bool &NegAcc, bool &HasLane) const {
DenseMap<unsigned, unsigned>::const_iterator I = MLxEntryMap.find(Opcode);
if (I == MLxEntryMap.end())
return false;
const ARM_MLxEntry &Entry = ARM_MLxTable[I->second];
MulOpc = Entry.MulOpc;
AddSubOpc = Entry.AddSubOpc;
NegAcc = Entry.NegAcc;
HasLane = Entry.HasLane;
return true;
}
//===----------------------------------------------------------------------===//
// Execution domains.
//===----------------------------------------------------------------------===//
//
// Some instructions go down the NEON pipeline, some go down the VFP pipeline,
// and some can go down both. The vmov instructions go down the VFP pipeline,
// but they can be changed to vorr equivalents that are executed by the NEON
// pipeline.
//
// We use the following execution domain numbering:
//
enum ARMExeDomain {
ExeGeneric = 0,
ExeVFP = 1,
ExeNEON = 2
};
//
// Also see ARMInstrFormats.td and Domain* enums in ARMBaseInfo.h
//
std::pair<uint16_t, uint16_t>
ARMBaseInstrInfo::getExecutionDomain(const MachineInstr &MI) const {
// If we don't have access to NEON instructions then we won't be able
// to swizzle anything to the NEON domain. Check to make sure.
if (Subtarget.hasNEON()) {
// VMOVD, VMOVRS and VMOVSR are VFP instructions, but can be changed to NEON
// if they are not predicated.
if (MI.getOpcode() == ARM::VMOVD && !isPredicated(MI))
return std::make_pair(ExeVFP, (1 << ExeVFP) | (1 << ExeNEON));
// CortexA9 is particularly picky about mixing the two and wants these
// converted.
if (Subtarget.useNEONForFPMovs() && !isPredicated(MI) &&
(MI.getOpcode() == ARM::VMOVRS || MI.getOpcode() == ARM::VMOVSR ||
MI.getOpcode() == ARM::VMOVS))
return std::make_pair(ExeVFP, (1 << ExeVFP) | (1 << ExeNEON));
}
// No other instructions can be swizzled, so just determine their domain.
unsigned Domain = MI.getDesc().TSFlags & ARMII::DomainMask;
if (Domain & ARMII::DomainNEON)
return std::make_pair(ExeNEON, 0);
// Certain instructions can go either way on Cortex-A8.
// Treat them as NEON instructions.
if ((Domain & ARMII::DomainNEONA8) && Subtarget.isCortexA8())
return std::make_pair(ExeNEON, 0);
if (Domain & ARMII::DomainVFP)
return std::make_pair(ExeVFP, 0);
return std::make_pair(ExeGeneric, 0);
}
static unsigned getCorrespondingDRegAndLane(const TargetRegisterInfo *TRI,
unsigned SReg, unsigned &Lane) {
unsigned DReg = TRI->getMatchingSuperReg(SReg, ARM::ssub_0, &ARM::DPRRegClass);
Lane = 0;
if (DReg != ARM::NoRegister)
return DReg;
Lane = 1;
DReg = TRI->getMatchingSuperReg(SReg, ARM::ssub_1, &ARM::DPRRegClass);
assert(DReg && "S-register with no D super-register?");
return DReg;
}
/// getImplicitSPRUseForDPRUse - Given a use of a DPR register and lane,
/// set ImplicitSReg to a register number that must be marked as implicit-use or
/// zero if no register needs to be defined as implicit-use.
///
/// If the function cannot determine if an SPR should be marked implicit use or
/// not, it returns false.
///
/// This function handles cases where an instruction is being modified from taking
/// an SPR to a DPR[Lane]. A use of the DPR is being added, which may conflict
/// with an earlier def of an SPR corresponding to DPR[Lane^1] (i.e. the other
/// lane of the DPR).
///
/// If the other SPR is defined, an implicit-use of it should be added. Else,
/// (including the case where the DPR itself is defined), it should not.
///
static bool getImplicitSPRUseForDPRUse(const TargetRegisterInfo *TRI,
MachineInstr &MI, unsigned DReg,
unsigned Lane, unsigned &ImplicitSReg) {
// If the DPR is defined or used already, the other SPR lane will be chained
// correctly, so there is nothing to be done.
if (MI.definesRegister(DReg, TRI) || MI.readsRegister(DReg, TRI)) {
ImplicitSReg = 0;
return true;
}
// Otherwise we need to go searching to see if the SPR is set explicitly.
ImplicitSReg = TRI->getSubReg(DReg,
(Lane & 1) ? ARM::ssub_0 : ARM::ssub_1);
MachineBasicBlock::LivenessQueryResult LQR =
MI.getParent()->computeRegisterLiveness(TRI, ImplicitSReg, MI);
if (LQR == MachineBasicBlock::LQR_Live)
return true;
else if (LQR == MachineBasicBlock::LQR_Unknown)
return false;
// If the register is known not to be live, there is no need to add an
// implicit-use.
ImplicitSReg = 0;
return true;
}
void ARMBaseInstrInfo::setExecutionDomain(MachineInstr &MI,
unsigned Domain) const {
unsigned DstReg, SrcReg, DReg;
unsigned Lane;
MachineInstrBuilder MIB(*MI.getParent()->getParent(), MI);
const TargetRegisterInfo *TRI = &getRegisterInfo();
switch (MI.getOpcode()) {
default:
llvm_unreachable("cannot handle opcode!");
break;
case ARM::VMOVD:
if (Domain != ExeNEON)
break;
// Zap the predicate operands.
assert(!isPredicated(MI) && "Cannot predicate a VORRd");
// Make sure we've got NEON instructions.
assert(Subtarget.hasNEON() && "VORRd requires NEON");
// Source instruction is %DDst = VMOVD %DSrc, 14, %noreg (; implicits)
DstReg = MI.getOperand(0).getReg();
SrcReg = MI.getOperand(1).getReg();
for (unsigned i = MI.getDesc().getNumOperands(); i; --i)
MI.RemoveOperand(i - 1);
// Change to a %DDst = VORRd %DSrc, %DSrc, 14, %noreg (; implicits)
MI.setDesc(get(ARM::VORRd));
MIB.addReg(DstReg, RegState::Define)
.addReg(SrcReg)
.addReg(SrcReg)
.add(predOps(ARMCC::AL));
break;
case ARM::VMOVRS:
if (Domain != ExeNEON)
break;
assert(!isPredicated(MI) && "Cannot predicate a VGETLN");
// Source instruction is %RDst = VMOVRS %SSrc, 14, %noreg (; implicits)
DstReg = MI.getOperand(0).getReg();
SrcReg = MI.getOperand(1).getReg();
for (unsigned i = MI.getDesc().getNumOperands(); i; --i)
MI.RemoveOperand(i - 1);
DReg = getCorrespondingDRegAndLane(TRI, SrcReg, Lane);
// Convert to %RDst = VGETLNi32 %DSrc, Lane, 14, %noreg (; imps)
// Note that DSrc has been widened and the other lane may be undef, which
// contaminates the entire register.
MI.setDesc(get(ARM::VGETLNi32));
MIB.addReg(DstReg, RegState::Define)
.addReg(DReg, RegState::Undef)
.addImm(Lane)
.add(predOps(ARMCC::AL));
// The old source should be an implicit use, otherwise we might think it
// was dead before here.
MIB.addReg(SrcReg, RegState::Implicit);
break;
case ARM::VMOVSR: {
if (Domain != ExeNEON)
break;
assert(!isPredicated(MI) && "Cannot predicate a VSETLN");
// Source instruction is %SDst = VMOVSR %RSrc, 14, %noreg (; implicits)
DstReg = MI.getOperand(0).getReg();
SrcReg = MI.getOperand(1).getReg();
DReg = getCorrespondingDRegAndLane(TRI, DstReg, Lane);
unsigned ImplicitSReg;
if (!getImplicitSPRUseForDPRUse(TRI, MI, DReg, Lane, ImplicitSReg))
break;
for (unsigned i = MI.getDesc().getNumOperands(); i; --i)
MI.RemoveOperand(i - 1);
// Convert to %DDst = VSETLNi32 %DDst, %RSrc, Lane, 14, %noreg (; imps)
// Again DDst may be undefined at the beginning of this instruction.
MI.setDesc(get(ARM::VSETLNi32));
MIB.addReg(DReg, RegState::Define)
.addReg(DReg, getUndefRegState(!MI.readsRegister(DReg, TRI)))
.addReg(SrcReg)
.addImm(Lane)
.add(predOps(ARMCC::AL));
// The narrower destination must be marked as set to keep previous chains
// in place.
MIB.addReg(DstReg, RegState::Define | RegState::Implicit);
if (ImplicitSReg != 0)
MIB.addReg(ImplicitSReg, RegState::Implicit);
break;
}
case ARM::VMOVS: {
if (Domain != ExeNEON)
break;
// Source instruction is %SDst = VMOVS %SSrc, 14, %noreg (; implicits)
DstReg = MI.getOperand(0).getReg();
SrcReg = MI.getOperand(1).getReg();
unsigned DstLane = 0, SrcLane = 0, DDst, DSrc;
DDst = getCorrespondingDRegAndLane(TRI, DstReg, DstLane);
DSrc = getCorrespondingDRegAndLane(TRI, SrcReg, SrcLane);
unsigned ImplicitSReg;
if (!getImplicitSPRUseForDPRUse(TRI, MI, DSrc, SrcLane, ImplicitSReg))
break;
for (unsigned i = MI.getDesc().getNumOperands(); i; --i)
MI.RemoveOperand(i - 1);
if (DSrc == DDst) {
// Destination can be:
// %DDst = VDUPLN32d %DDst, Lane, 14, %noreg (; implicits)
MI.setDesc(get(ARM::VDUPLN32d));
MIB.addReg(DDst, RegState::Define)
.addReg(DDst, getUndefRegState(!MI.readsRegister(DDst, TRI)))
.addImm(SrcLane)
.add(predOps(ARMCC::AL));
// Neither the source or the destination are naturally represented any
// more, so add them in manually.
MIB.addReg(DstReg, RegState::Implicit | RegState::Define);
MIB.addReg(SrcReg, RegState::Implicit);
if (ImplicitSReg != 0)
MIB.addReg(ImplicitSReg, RegState::Implicit);
break;
}
// In general there's no single instruction that can perform an S <-> S
// move in NEON space, but a pair of VEXT instructions *can* do the
// job. It turns out that the VEXTs needed will only use DSrc once, with
// the position based purely on the combination of lane-0 and lane-1
// involved. For example
// vmov s0, s2 -> vext.32 d0, d0, d1, #1 vext.32 d0, d0, d0, #1
// vmov s1, s3 -> vext.32 d0, d1, d0, #1 vext.32 d0, d0, d0, #1
// vmov s0, s3 -> vext.32 d0, d0, d0, #1 vext.32 d0, d1, d0, #1
// vmov s1, s2 -> vext.32 d0, d0, d0, #1 vext.32 d0, d0, d1, #1
//
// Pattern of the MachineInstrs is:
// %DDst = VEXTd32 %DSrc1, %DSrc2, Lane, 14, %noreg (;implicits)
MachineInstrBuilder NewMIB;
NewMIB = BuildMI(*MI.getParent(), MI, MI.getDebugLoc(), get(ARM::VEXTd32),
DDst);
// On the first instruction, both DSrc and DDst may be undef if present.
// Specifically when the original instruction didn't have them as an
// <imp-use>.
unsigned CurReg = SrcLane == 1 && DstLane == 1 ? DSrc : DDst;
bool CurUndef = !MI.readsRegister(CurReg, TRI);
NewMIB.addReg(CurReg, getUndefRegState(CurUndef));
CurReg = SrcLane == 0 && DstLane == 0 ? DSrc : DDst;
CurUndef = !MI.readsRegister(CurReg, TRI);
NewMIB.addReg(CurReg, getUndefRegState(CurUndef))
.addImm(1)
.add(predOps(ARMCC::AL));
if (SrcLane == DstLane)
NewMIB.addReg(SrcReg, RegState::Implicit);
MI.setDesc(get(ARM::VEXTd32));
MIB.addReg(DDst, RegState::Define);
// On the second instruction, DDst has definitely been defined above, so
// it is not undef. DSrc, if present, can be undef as above.
CurReg = SrcLane == 1 && DstLane == 0 ? DSrc : DDst;
CurUndef = CurReg == DSrc && !MI.readsRegister(CurReg, TRI);
MIB.addReg(CurReg, getUndefRegState(CurUndef));
CurReg = SrcLane == 0 && DstLane == 1 ? DSrc : DDst;
CurUndef = CurReg == DSrc && !MI.readsRegister(CurReg, TRI);
MIB.addReg(CurReg, getUndefRegState(CurUndef))
.addImm(1)
.add(predOps(ARMCC::AL));
if (SrcLane != DstLane)
MIB.addReg(SrcReg, RegState::Implicit);
// As before, the original destination is no longer represented, add it
// implicitly.
MIB.addReg(DstReg, RegState::Define | RegState::Implicit);
if (ImplicitSReg != 0)
MIB.addReg(ImplicitSReg, RegState::Implicit);
break;
}
}
}
//===----------------------------------------------------------------------===//
// Partial register updates
//===----------------------------------------------------------------------===//
//
// Swift renames NEON registers with 64-bit granularity. That means any
// instruction writing an S-reg implicitly reads the containing D-reg. The
// problem is mostly avoided by translating f32 operations to v2f32 operations
// on D-registers, but f32 loads are still a problem.
//
// These instructions can load an f32 into a NEON register:
//
// VLDRS - Only writes S, partial D update.
// VLD1LNd32 - Writes all D-regs, explicit partial D update, 2 uops.
// VLD1DUPd32 - Writes all D-regs, no partial reg update, 2 uops.
//
// FCONSTD can be used as a dependency-breaking instruction.
unsigned ARMBaseInstrInfo::getPartialRegUpdateClearance(
const MachineInstr &MI, unsigned OpNum,
const TargetRegisterInfo *TRI) const {
auto PartialUpdateClearance = Subtarget.getPartialUpdateClearance();
if (!PartialUpdateClearance)
return 0;
assert(TRI && "Need TRI instance");
const MachineOperand &MO = MI.getOperand(OpNum);
if (MO.readsReg())
return 0;
Register Reg = MO.getReg();
int UseOp = -1;
switch (MI.getOpcode()) {
// Normal instructions writing only an S-register.
case ARM::VLDRS:
case ARM::FCONSTS:
case ARM::VMOVSR:
case ARM::VMOVv8i8:
case ARM::VMOVv4i16:
case ARM::VMOVv2i32:
case ARM::VMOVv2f32:
case ARM::VMOVv1i64:
UseOp = MI.findRegisterUseOperandIdx(Reg, false, TRI);
break;
// Explicitly reads the dependency.
case ARM::VLD1LNd32:
UseOp = 3;
break;
default:
return 0;
}
// If this instruction actually reads a value from Reg, there is no unwanted
// dependency.
if (UseOp != -1 && MI.getOperand(UseOp).readsReg())
return 0;
// We must be able to clobber the whole D-reg.
if (Register::isVirtualRegister(Reg)) {
// Virtual register must be a def undef foo:ssub_0 operand.
if (!MO.getSubReg() || MI.readsVirtualRegister(Reg))
return 0;
} else if (ARM::SPRRegClass.contains(Reg)) {
// Physical register: MI must define the full D-reg.
unsigned DReg = TRI->getMatchingSuperReg(Reg, ARM::ssub_0,
&ARM::DPRRegClass);
if (!DReg || !MI.definesRegister(DReg, TRI))
return 0;
}
// MI has an unwanted D-register dependency.
// Avoid defs in the previous N instructrions.
return PartialUpdateClearance;
}
// Break a partial register dependency after getPartialRegUpdateClearance
// returned non-zero.
void ARMBaseInstrInfo::breakPartialRegDependency(
MachineInstr &MI, unsigned OpNum, const TargetRegisterInfo *TRI) const {
assert(OpNum < MI.getDesc().getNumDefs() && "OpNum is not a def");
assert(TRI && "Need TRI instance");
const MachineOperand &MO = MI.getOperand(OpNum);
Register Reg = MO.getReg();
assert(Register::isPhysicalRegister(Reg) &&
"Can't break virtual register dependencies.");
unsigned DReg = Reg;
// If MI defines an S-reg, find the corresponding D super-register.
if (ARM::SPRRegClass.contains(Reg)) {
DReg = ARM::D0 + (Reg - ARM::S0) / 2;
assert(TRI->isSuperRegister(Reg, DReg) && "Register enums broken");
}
assert(ARM::DPRRegClass.contains(DReg) && "Can only break D-reg deps");
assert(MI.definesRegister(DReg, TRI) && "MI doesn't clobber full D-reg");
// FIXME: In some cases, VLDRS can be changed to a VLD1DUPd32 which defines
// the full D-register by loading the same value to both lanes. The
// instruction is micro-coded with 2 uops, so don't do this until we can
// properly schedule micro-coded instructions. The dispatcher stalls cause
// too big regressions.
// Insert the dependency-breaking FCONSTD before MI.
// 96 is the encoding of 0.5, but the actual value doesn't matter here.
BuildMI(*MI.getParent(), MI, MI.getDebugLoc(), get(ARM::FCONSTD), DReg)
.addImm(96)
.add(predOps(ARMCC::AL));
MI.addRegisterKilled(DReg, TRI, true);
}
bool ARMBaseInstrInfo::hasNOP() const {
return Subtarget.getFeatureBits()[ARM::HasV6KOps];
}
bool ARMBaseInstrInfo::isSwiftFastImmShift(const MachineInstr *MI) const {
if (MI->getNumOperands() < 4)
return true;
unsigned ShOpVal = MI->getOperand(3).getImm();
unsigned ShImm = ARM_AM::getSORegOffset(ShOpVal);
// Swift supports faster shifts for: lsl 2, lsl 1, and lsr 1.
if ((ShImm == 1 && ARM_AM::getSORegShOp(ShOpVal) == ARM_AM::lsr) ||
((ShImm == 1 || ShImm == 2) &&
ARM_AM::getSORegShOp(ShOpVal) == ARM_AM::lsl))
return true;
return false;
}
bool ARMBaseInstrInfo::getRegSequenceLikeInputs(
const MachineInstr &MI, unsigned DefIdx,
SmallVectorImpl<RegSubRegPairAndIdx> &InputRegs) const {
assert(DefIdx < MI.getDesc().getNumDefs() && "Invalid definition index");
assert(MI.isRegSequenceLike() && "Invalid kind of instruction");
switch (MI.getOpcode()) {
case ARM::VMOVDRR:
// dX = VMOVDRR rY, rZ
// is the same as:
// dX = REG_SEQUENCE rY, ssub_0, rZ, ssub_1
// Populate the InputRegs accordingly.
// rY
const MachineOperand *MOReg = &MI.getOperand(1);
if (!MOReg->isUndef())
InputRegs.push_back(RegSubRegPairAndIdx(MOReg->getReg(),
MOReg->getSubReg(), ARM::ssub_0));
// rZ
MOReg = &MI.getOperand(2);
if (!MOReg->isUndef())
InputRegs.push_back(RegSubRegPairAndIdx(MOReg->getReg(),
MOReg->getSubReg(), ARM::ssub_1));
return true;
}
llvm_unreachable("Target dependent opcode missing");
}
bool ARMBaseInstrInfo::getExtractSubregLikeInputs(
const MachineInstr &MI, unsigned DefIdx,
RegSubRegPairAndIdx &InputReg) const {
assert(DefIdx < MI.getDesc().getNumDefs() && "Invalid definition index");
assert(MI.isExtractSubregLike() && "Invalid kind of instruction");
switch (MI.getOpcode()) {
case ARM::VMOVRRD:
// rX, rY = VMOVRRD dZ
// is the same as:
// rX = EXTRACT_SUBREG dZ, ssub_0
// rY = EXTRACT_SUBREG dZ, ssub_1
const MachineOperand &MOReg = MI.getOperand(2);
if (MOReg.isUndef())
return false;
InputReg.Reg = MOReg.getReg();
InputReg.SubReg = MOReg.getSubReg();
InputReg.SubIdx = DefIdx == 0 ? ARM::ssub_0 : ARM::ssub_1;
return true;
}
llvm_unreachable("Target dependent opcode missing");
}
bool ARMBaseInstrInfo::getInsertSubregLikeInputs(
const MachineInstr &MI, unsigned DefIdx, RegSubRegPair &BaseReg,
RegSubRegPairAndIdx &InsertedReg) const {
assert(DefIdx < MI.getDesc().getNumDefs() && "Invalid definition index");
assert(MI.isInsertSubregLike() && "Invalid kind of instruction");
switch (MI.getOpcode()) {
case ARM::VSETLNi32:
// dX = VSETLNi32 dY, rZ, imm
const MachineOperand &MOBaseReg = MI.getOperand(1);
const MachineOperand &MOInsertedReg = MI.getOperand(2);
if (MOInsertedReg.isUndef())
return false;
const MachineOperand &MOIndex = MI.getOperand(3);
BaseReg.Reg = MOBaseReg.getReg();
BaseReg.SubReg = MOBaseReg.getSubReg();
InsertedReg.Reg = MOInsertedReg.getReg();
InsertedReg.SubReg = MOInsertedReg.getSubReg();
InsertedReg.SubIdx = MOIndex.getImm() == 0 ? ARM::ssub_0 : ARM::ssub_1;
return true;
}
llvm_unreachable("Target dependent opcode missing");
}
std::pair<unsigned, unsigned>
ARMBaseInstrInfo::decomposeMachineOperandsTargetFlags(unsigned TF) const {
const unsigned Mask = ARMII::MO_OPTION_MASK;
return std::make_pair(TF & Mask, TF & ~Mask);
}
ArrayRef<std::pair<unsigned, const char *>>
ARMBaseInstrInfo::getSerializableDirectMachineOperandTargetFlags() const {
using namespace ARMII;
static const std::pair<unsigned, const char *> TargetFlags[] = {
{MO_LO16, "arm-lo16"}, {MO_HI16, "arm-hi16"}};
return makeArrayRef(TargetFlags);
}
ArrayRef<std::pair<unsigned, const char *>>
ARMBaseInstrInfo::getSerializableBitmaskMachineOperandTargetFlags() const {
using namespace ARMII;
static const std::pair<unsigned, const char *> TargetFlags[] = {
{MO_COFFSTUB, "arm-coffstub"},
{MO_GOT, "arm-got"},
{MO_SBREL, "arm-sbrel"},
{MO_DLLIMPORT, "arm-dllimport"},
{MO_SECREL, "arm-secrel"},
{MO_NONLAZY, "arm-nonlazy"}};
return makeArrayRef(TargetFlags);
}
bool llvm::registerDefinedBetween(unsigned Reg,
MachineBasicBlock::iterator From,
MachineBasicBlock::iterator To,
const TargetRegisterInfo *TRI) {
for (auto I = From; I != To; ++I)
if (I->modifiesRegister(Reg, TRI))
return true;
return false;
}
MachineInstr *llvm::findCMPToFoldIntoCBZ(MachineInstr *Br,
const TargetRegisterInfo *TRI) {
// Search backwards to the instruction that defines CSPR. This may or not
// be a CMP, we check that after this loop. If we find another instruction
// that reads cpsr, we return nullptr.
MachineBasicBlock::iterator CmpMI = Br;
while (CmpMI != Br->getParent()->begin()) {
--CmpMI;
if (CmpMI->modifiesRegister(ARM::CPSR, TRI))
break;
if (CmpMI->readsRegister(ARM::CPSR, TRI))
break;
}
// Check that this inst is a CMP r[0-7], #0 and that the register
// is not redefined between the cmp and the br.
if (CmpMI->getOpcode() != ARM::tCMPi8 && CmpMI->getOpcode() != ARM::t2CMPri)
return nullptr;
Register Reg = CmpMI->getOperand(0).getReg();
unsigned PredReg = 0;
ARMCC::CondCodes Pred = getInstrPredicate(*CmpMI, PredReg);
if (Pred != ARMCC::AL || CmpMI->getOperand(1).getImm() != 0)
return nullptr;
if (!isARMLowRegister(Reg))
return nullptr;
if (registerDefinedBetween(Reg, CmpMI->getNextNode(), Br, TRI))
return nullptr;
return &*CmpMI;
}
unsigned llvm::ConstantMaterializationCost(unsigned Val,
const ARMSubtarget *Subtarget,
bool ForCodesize) {
if (Subtarget->isThumb()) {
if (Val <= 255) // MOV
return |