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
|
// Copyright 2013 the V8 project authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.
#if V8_TARGET_ARCH_ARM64
#include "src/base/bits.h"
#include "src/base/division-by-constant.h"
#include "src/codegen/assembler.h"
#include "src/codegen/callable.h"
#include "src/codegen/code-factory.h"
#include "src/codegen/external-reference-table.h"
#include "src/codegen/macro-assembler-inl.h"
#include "src/codegen/register-configuration.h"
#include "src/debug/debug.h"
#include "src/deoptimizer/deoptimizer.h"
#include "src/execution/frame-constants.h"
#include "src/execution/frames-inl.h"
#include "src/heap/memory-chunk.h"
#include "src/init/bootstrapper.h"
#include "src/logging/counters.h"
#include "src/runtime/runtime.h"
#include "src/snapshot/embedded/embedded-data.h"
#include "src/snapshot/snapshot.h"
#include "src/wasm/wasm-code-manager.h"
// Satisfy cpplint check, but don't include platform-specific header. It is
// included recursively via macro-assembler.h.
#if 0
#include "src/codegen/arm64/macro-assembler-arm64.h"
#endif
namespace v8 {
namespace internal {
CPURegList TurboAssembler::DefaultTmpList() { return CPURegList(ip0, ip1); }
CPURegList TurboAssembler::DefaultFPTmpList() {
return CPURegList(fp_scratch1, fp_scratch2);
}
int TurboAssembler::RequiredStackSizeForCallerSaved(SaveFPRegsMode fp_mode,
Register exclusion) const {
auto list = kCallerSaved;
list.Remove(exclusion);
list.Align();
int bytes = list.Count() * kXRegSizeInBits / 8;
if (fp_mode == kSaveFPRegs) {
DCHECK_EQ(kCallerSavedV.Count() % 2, 0);
bytes += kCallerSavedV.Count() * kDRegSizeInBits / 8;
}
return bytes;
}
int TurboAssembler::PushCallerSaved(SaveFPRegsMode fp_mode,
Register exclusion) {
auto list = kCallerSaved;
list.Remove(exclusion);
list.Align();
PushCPURegList<kDontStoreLR>(list);
int bytes = list.Count() * kXRegSizeInBits / 8;
if (fp_mode == kSaveFPRegs) {
DCHECK_EQ(kCallerSavedV.Count() % 2, 0);
PushCPURegList(kCallerSavedV);
bytes += kCallerSavedV.Count() * kDRegSizeInBits / 8;
}
return bytes;
}
int TurboAssembler::PopCallerSaved(SaveFPRegsMode fp_mode, Register exclusion) {
int bytes = 0;
if (fp_mode == kSaveFPRegs) {
DCHECK_EQ(kCallerSavedV.Count() % 2, 0);
PopCPURegList(kCallerSavedV);
bytes += kCallerSavedV.Count() * kDRegSizeInBits / 8;
}
auto list = kCallerSaved;
list.Remove(exclusion);
list.Align();
PopCPURegList<kDontLoadLR>(list);
bytes += list.Count() * kXRegSizeInBits / 8;
return bytes;
}
void TurboAssembler::LogicalMacro(const Register& rd, const Register& rn,
const Operand& operand, LogicalOp op) {
UseScratchRegisterScope temps(this);
if (operand.NeedsRelocation(this)) {
Register temp = temps.AcquireX();
Ldr(temp, operand.immediate());
Logical(rd, rn, temp, op);
} else if (operand.IsImmediate()) {
int64_t immediate = operand.ImmediateValue();
unsigned reg_size = rd.SizeInBits();
// If the operation is NOT, invert the operation and immediate.
if ((op & NOT) == NOT) {
op = static_cast<LogicalOp>(op & ~NOT);
immediate = ~immediate;
}
// Ignore the top 32 bits of an immediate if we're moving to a W register.
if (rd.Is32Bits()) {
// Check that the top 32 bits are consistent.
DCHECK(((immediate >> kWRegSizeInBits) == 0) ||
((immediate >> kWRegSizeInBits) == -1));
immediate &= kWRegMask;
}
DCHECK(rd.Is64Bits() || is_uint32(immediate));
// Special cases for all set or all clear immediates.
if (immediate == 0) {
switch (op) {
case AND:
Mov(rd, 0);
return;
case ORR: // Fall through.
case EOR:
Mov(rd, rn);
return;
case ANDS: // Fall through.
case BICS:
break;
default:
UNREACHABLE();
}
} else if ((rd.Is64Bits() && (immediate == -1L)) ||
(rd.Is32Bits() && (immediate == 0xFFFFFFFFL))) {
switch (op) {
case AND:
Mov(rd, rn);
return;
case ORR:
Mov(rd, immediate);
return;
case EOR:
Mvn(rd, rn);
return;
case ANDS: // Fall through.
case BICS:
break;
default:
UNREACHABLE();
}
}
unsigned n, imm_s, imm_r;
if (IsImmLogical(immediate, reg_size, &n, &imm_s, &imm_r)) {
// Immediate can be encoded in the instruction.
LogicalImmediate(rd, rn, n, imm_s, imm_r, op);
} else {
// Immediate can't be encoded: synthesize using move immediate.
Register temp = temps.AcquireSameSizeAs(rn);
// If the left-hand input is the stack pointer, we can't pre-shift the
// immediate, as the encoding won't allow the subsequent post shift.
PreShiftImmMode mode = rn == sp ? kNoShift : kAnyShift;
Operand imm_operand = MoveImmediateForShiftedOp(temp, immediate, mode);
if (rd.IsSP()) {
// If rd is the stack pointer we cannot use it as the destination
// register so we use the temp register as an intermediate again.
Logical(temp, rn, imm_operand, op);
Mov(sp, temp);
} else {
Logical(rd, rn, imm_operand, op);
}
}
} else if (operand.IsExtendedRegister()) {
DCHECK(operand.reg().SizeInBits() <= rd.SizeInBits());
// Add/sub extended supports shift <= 4. We want to support exactly the
// same modes here.
DCHECK_LE(operand.shift_amount(), 4);
DCHECK(operand.reg().Is64Bits() ||
((operand.extend() != UXTX) && (operand.extend() != SXTX)));
Register temp = temps.AcquireSameSizeAs(rn);
EmitExtendShift(temp, operand.reg(), operand.extend(),
operand.shift_amount());
Logical(rd, rn, temp, op);
} else {
// The operand can be encoded in the instruction.
DCHECK(operand.IsShiftedRegister());
Logical(rd, rn, operand, op);
}
}
void TurboAssembler::Mov(const Register& rd, uint64_t imm) {
DCHECK(allow_macro_instructions());
DCHECK(is_uint32(imm) || is_int32(imm) || rd.Is64Bits());
DCHECK(!rd.IsZero());
// TODO(all) extend to support more immediates.
//
// Immediates on Aarch64 can be produced using an initial value, and zero to
// three move keep operations.
//
// Initial values can be generated with:
// 1. 64-bit move zero (movz).
// 2. 32-bit move inverted (movn).
// 3. 64-bit move inverted.
// 4. 32-bit orr immediate.
// 5. 64-bit orr immediate.
// Move-keep may then be used to modify each of the 16-bit half-words.
//
// The code below supports all five initial value generators, and
// applying move-keep operations to move-zero and move-inverted initial
// values.
// Try to move the immediate in one instruction, and if that fails, switch to
// using multiple instructions.
if (!TryOneInstrMoveImmediate(rd, imm)) {
unsigned reg_size = rd.SizeInBits();
// Generic immediate case. Imm will be represented by
// [imm3, imm2, imm1, imm0], where each imm is 16 bits.
// A move-zero or move-inverted is generated for the first non-zero or
// non-0xFFFF immX, and a move-keep for subsequent non-zero immX.
uint64_t ignored_halfword = 0;
bool invert_move = false;
// If the number of 0xFFFF halfwords is greater than the number of 0x0000
// halfwords, it's more efficient to use move-inverted.
if (CountClearHalfWords(~imm, reg_size) >
CountClearHalfWords(imm, reg_size)) {
ignored_halfword = 0xFFFFL;
invert_move = true;
}
// Mov instructions can't move immediate values into the stack pointer, so
// set up a temporary register, if needed.
UseScratchRegisterScope temps(this);
Register temp = rd.IsSP() ? temps.AcquireSameSizeAs(rd) : rd;
// Iterate through the halfwords. Use movn/movz for the first non-ignored
// halfword, and movk for subsequent halfwords.
DCHECK_EQ(reg_size % 16, 0);
bool first_mov_done = false;
for (int i = 0; i < (rd.SizeInBits() / 16); i++) {
uint64_t imm16 = (imm >> (16 * i)) & 0xFFFFL;
if (imm16 != ignored_halfword) {
if (!first_mov_done) {
if (invert_move) {
movn(temp, (~imm16) & 0xFFFFL, 16 * i);
} else {
movz(temp, imm16, 16 * i);
}
first_mov_done = true;
} else {
// Construct a wider constant.
movk(temp, imm16, 16 * i);
}
}
}
DCHECK(first_mov_done);
// Move the temporary if the original destination register was the stack
// pointer.
if (rd.IsSP()) {
mov(rd, temp);
}
}
}
void TurboAssembler::Mov(const Register& rd, const Operand& operand,
DiscardMoveMode discard_mode) {
DCHECK(allow_macro_instructions());
DCHECK(!rd.IsZero());
// Provide a swap register for instructions that need to write into the
// system stack pointer (and can't do this inherently).
UseScratchRegisterScope temps(this);
Register dst = (rd.IsSP()) ? temps.AcquireSameSizeAs(rd) : rd;
if (operand.NeedsRelocation(this)) {
// TODO(jgruber,v8:8887): Also consider a root-relative load when generating
// non-isolate-independent code. In many cases it might be cheaper than
// embedding the relocatable value.
if (root_array_available_ && options().isolate_independent_code) {
if (operand.ImmediateRMode() == RelocInfo::EXTERNAL_REFERENCE) {
Address addr = static_cast<Address>(operand.ImmediateValue());
ExternalReference reference = bit_cast<ExternalReference>(addr);
IndirectLoadExternalReference(rd, reference);
return;
} else if (RelocInfo::IsEmbeddedObjectMode(operand.ImmediateRMode())) {
Handle<HeapObject> x(
reinterpret_cast<Address*>(operand.ImmediateValue()));
// TODO(v8:9706): Fix-it! This load will always uncompress the value
// even when we are loading a compressed embedded object.
IndirectLoadConstant(rd.X(), x);
return;
}
}
Ldr(dst, operand);
} else if (operand.IsImmediate()) {
// Call the macro assembler for generic immediates.
Mov(dst, operand.ImmediateValue());
} else if (operand.IsShiftedRegister() && (operand.shift_amount() != 0)) {
// Emit a shift instruction if moving a shifted register. This operation
// could also be achieved using an orr instruction (like orn used by Mvn),
// but using a shift instruction makes the disassembly clearer.
EmitShift(dst, operand.reg(), operand.shift(), operand.shift_amount());
} else if (operand.IsExtendedRegister()) {
// Emit an extend instruction if moving an extended register. This handles
// extend with post-shift operations, too.
EmitExtendShift(dst, operand.reg(), operand.extend(),
operand.shift_amount());
} else {
// Otherwise, emit a register move only if the registers are distinct, or
// if they are not X registers.
//
// Note that mov(w0, w0) is not a no-op because it clears the top word of
// x0. A flag is provided (kDiscardForSameWReg) if a move between the same W
// registers is not required to clear the top word of the X register. In
// this case, the instruction is discarded.
//
// If sp is an operand, add #0 is emitted, otherwise, orr #0.
if (rd != operand.reg() ||
(rd.Is32Bits() && (discard_mode == kDontDiscardForSameWReg))) {
Assembler::mov(rd, operand.reg());
}
// This case can handle writes into the system stack pointer directly.
dst = rd;
}
// Copy the result to the system stack pointer.
if (dst != rd) {
DCHECK(rd.IsSP());
Assembler::mov(rd, dst);
}
}
void TurboAssembler::Mov(const Register& rd, Smi smi) {
return Mov(rd, Operand(smi));
}
void TurboAssembler::Movi16bitHelper(const VRegister& vd, uint64_t imm) {
DCHECK(is_uint16(imm));
int byte1 = (imm & 0xFF);
int byte2 = ((imm >> 8) & 0xFF);
if (byte1 == byte2) {
movi(vd.Is64Bits() ? vd.V8B() : vd.V16B(), byte1);
} else if (byte1 == 0) {
movi(vd, byte2, LSL, 8);
} else if (byte2 == 0) {
movi(vd, byte1);
} else if (byte1 == 0xFF) {
mvni(vd, ~byte2 & 0xFF, LSL, 8);
} else if (byte2 == 0xFF) {
mvni(vd, ~byte1 & 0xFF);
} else {
UseScratchRegisterScope temps(this);
Register temp = temps.AcquireW();
movz(temp, imm);
dup(vd, temp);
}
}
void TurboAssembler::Movi32bitHelper(const VRegister& vd, uint64_t imm) {
DCHECK(is_uint32(imm));
uint8_t bytes[sizeof(imm)];
memcpy(bytes, &imm, sizeof(imm));
// All bytes are either 0x00 or 0xFF.
{
bool all0orff = true;
for (int i = 0; i < 4; ++i) {
if ((bytes[i] != 0) && (bytes[i] != 0xFF)) {
all0orff = false;
break;
}
}
if (all0orff == true) {
movi(vd.Is64Bits() ? vd.V1D() : vd.V2D(), ((imm << 32) | imm));
return;
}
}
// Of the 4 bytes, only one byte is non-zero.
for (int i = 0; i < 4; i++) {
if ((imm & (0xFF << (i * 8))) == imm) {
movi(vd, bytes[i], LSL, i * 8);
return;
}
}
// Of the 4 bytes, only one byte is not 0xFF.
for (int i = 0; i < 4; i++) {
uint32_t mask = ~(0xFF << (i * 8));
if ((imm & mask) == mask) {
mvni(vd, ~bytes[i] & 0xFF, LSL, i * 8);
return;
}
}
// Immediate is of the form 0x00MMFFFF.
if ((imm & 0xFF00FFFF) == 0x0000FFFF) {
movi(vd, bytes[2], MSL, 16);
return;
}
// Immediate is of the form 0x0000MMFF.
if ((imm & 0xFFFF00FF) == 0x000000FF) {
movi(vd, bytes[1], MSL, 8);
return;
}
// Immediate is of the form 0xFFMM0000.
if ((imm & 0xFF00FFFF) == 0xFF000000) {
mvni(vd, ~bytes[2] & 0xFF, MSL, 16);
return;
}
// Immediate is of the form 0xFFFFMM00.
if ((imm & 0xFFFF00FF) == 0xFFFF0000) {
mvni(vd, ~bytes[1] & 0xFF, MSL, 8);
return;
}
// Top and bottom 16-bits are equal.
if (((imm >> 16) & 0xFFFF) == (imm & 0xFFFF)) {
Movi16bitHelper(vd.Is64Bits() ? vd.V4H() : vd.V8H(), imm & 0xFFFF);
return;
}
// Default case.
{
UseScratchRegisterScope temps(this);
Register temp = temps.AcquireW();
Mov(temp, imm);
dup(vd, temp);
}
}
void TurboAssembler::Movi64bitHelper(const VRegister& vd, uint64_t imm) {
// All bytes are either 0x00 or 0xFF.
{
bool all0orff = true;
for (int i = 0; i < 8; ++i) {
int byteval = (imm >> (i * 8)) & 0xFF;
if (byteval != 0 && byteval != 0xFF) {
all0orff = false;
break;
}
}
if (all0orff == true) {
movi(vd, imm);
return;
}
}
// Top and bottom 32-bits are equal.
if (((imm >> 32) & 0xFFFFFFFF) == (imm & 0xFFFFFFFF)) {
Movi32bitHelper(vd.Is64Bits() ? vd.V2S() : vd.V4S(), imm & 0xFFFFFFFF);
return;
}
// Default case.
{
UseScratchRegisterScope temps(this);
Register temp = temps.AcquireX();
Mov(temp, imm);
if (vd.Is1D()) {
mov(vd.D(), 0, temp);
} else {
dup(vd.V2D(), temp);
}
}
}
void TurboAssembler::Movi(const VRegister& vd, uint64_t imm, Shift shift,
int shift_amount) {
DCHECK(allow_macro_instructions());
if (shift_amount != 0 || shift != LSL) {
movi(vd, imm, shift, shift_amount);
} else if (vd.Is8B() || vd.Is16B()) {
// 8-bit immediate.
DCHECK(is_uint8(imm));
movi(vd, imm);
} else if (vd.Is4H() || vd.Is8H()) {
// 16-bit immediate.
Movi16bitHelper(vd, imm);
} else if (vd.Is2S() || vd.Is4S()) {
// 32-bit immediate.
Movi32bitHelper(vd, imm);
} else {
// 64-bit immediate.
Movi64bitHelper(vd, imm);
}
}
void TurboAssembler::Movi(const VRegister& vd, uint64_t hi, uint64_t lo) {
// TODO(all): Move 128-bit values in a more efficient way.
DCHECK(vd.Is128Bits());
UseScratchRegisterScope temps(this);
Movi(vd.V2D(), lo);
Register temp = temps.AcquireX();
Mov(temp, hi);
Ins(vd.V2D(), 1, temp);
}
void TurboAssembler::Mvn(const Register& rd, const Operand& operand) {
DCHECK(allow_macro_instructions());
if (operand.NeedsRelocation(this)) {
Ldr(rd, operand.immediate());
mvn(rd, rd);
} else if (operand.IsImmediate()) {
// Call the macro assembler for generic immediates.
Mov(rd, ~operand.ImmediateValue());
} else if (operand.IsExtendedRegister()) {
// Emit two instructions for the extend case. This differs from Mov, as
// the extend and invert can't be achieved in one instruction.
EmitExtendShift(rd, operand.reg(), operand.extend(),
operand.shift_amount());
mvn(rd, rd);
} else {
mvn(rd, operand);
}
}
unsigned TurboAssembler::CountClearHalfWords(uint64_t imm, unsigned reg_size) {
DCHECK_EQ(reg_size % 8, 0);
int count = 0;
for (unsigned i = 0; i < (reg_size / 16); i++) {
if ((imm & 0xFFFF) == 0) {
count++;
}
imm >>= 16;
}
return count;
}
// The movz instruction can generate immediates containing an arbitrary 16-bit
// half-word, with remaining bits clear, eg. 0x00001234, 0x0000123400000000.
bool TurboAssembler::IsImmMovz(uint64_t imm, unsigned reg_size) {
DCHECK((reg_size == kXRegSizeInBits) || (reg_size == kWRegSizeInBits));
return CountClearHalfWords(imm, reg_size) >= ((reg_size / 16) - 1);
}
// The movn instruction can generate immediates containing an arbitrary 16-bit
// half-word, with remaining bits set, eg. 0xFFFF1234, 0xFFFF1234FFFFFFFF.
bool TurboAssembler::IsImmMovn(uint64_t imm, unsigned reg_size) {
return IsImmMovz(~imm, reg_size);
}
void TurboAssembler::ConditionalCompareMacro(const Register& rn,
const Operand& operand,
StatusFlags nzcv, Condition cond,
ConditionalCompareOp op) {
DCHECK((cond != al) && (cond != nv));
if (operand.NeedsRelocation(this)) {
UseScratchRegisterScope temps(this);
Register temp = temps.AcquireX();
Ldr(temp, operand.immediate());
ConditionalCompareMacro(rn, temp, nzcv, cond, op);
} else if ((operand.IsShiftedRegister() && (operand.shift_amount() == 0)) ||
(operand.IsImmediate() &&
IsImmConditionalCompare(operand.ImmediateValue()))) {
// The immediate can be encoded in the instruction, or the operand is an
// unshifted register: call the assembler.
ConditionalCompare(rn, operand, nzcv, cond, op);
} else {
// The operand isn't directly supported by the instruction: perform the
// operation on a temporary register.
UseScratchRegisterScope temps(this);
Register temp = temps.AcquireSameSizeAs(rn);
Mov(temp, operand);
ConditionalCompare(rn, temp, nzcv, cond, op);
}
}
void TurboAssembler::Csel(const Register& rd, const Register& rn,
const Operand& operand, Condition cond) {
DCHECK(allow_macro_instructions());
DCHECK(!rd.IsZero());
DCHECK((cond != al) && (cond != nv));
if (operand.IsImmediate()) {
// Immediate argument. Handle special cases of 0, 1 and -1 using zero
// register.
int64_t imm = operand.ImmediateValue();
Register zr = AppropriateZeroRegFor(rn);
if (imm == 0) {
csel(rd, rn, zr, cond);
} else if (imm == 1) {
csinc(rd, rn, zr, cond);
} else if (imm == -1) {
csinv(rd, rn, zr, cond);
} else {
UseScratchRegisterScope temps(this);
Register temp = temps.AcquireSameSizeAs(rn);
Mov(temp, imm);
csel(rd, rn, temp, cond);
}
} else if (operand.IsShiftedRegister() && (operand.shift_amount() == 0)) {
// Unshifted register argument.
csel(rd, rn, operand.reg(), cond);
} else {
// All other arguments.
UseScratchRegisterScope temps(this);
Register temp = temps.AcquireSameSizeAs(rn);
Mov(temp, operand);
csel(rd, rn, temp, cond);
}
}
bool TurboAssembler::TryOneInstrMoveImmediate(const Register& dst,
int64_t imm) {
unsigned n, imm_s, imm_r;
int reg_size = dst.SizeInBits();
if (IsImmMovz(imm, reg_size) && !dst.IsSP()) {
// Immediate can be represented in a move zero instruction. Movz can't write
// to the stack pointer.
movz(dst, imm);
return true;
} else if (IsImmMovn(imm, reg_size) && !dst.IsSP()) {
// Immediate can be represented in a move not instruction. Movn can't write
// to the stack pointer.
movn(dst, dst.Is64Bits() ? ~imm : (~imm & kWRegMask));
return true;
} else if (IsImmLogical(imm, reg_size, &n, &imm_s, &imm_r)) {
// Immediate can be represented in a logical orr instruction.
LogicalImmediate(dst, AppropriateZeroRegFor(dst), n, imm_s, imm_r, ORR);
return true;
}
return false;
}
Operand TurboAssembler::MoveImmediateForShiftedOp(const Register& dst,
int64_t imm,
PreShiftImmMode mode) {
int reg_size = dst.SizeInBits();
// Encode the immediate in a single move instruction, if possible.
if (TryOneInstrMoveImmediate(dst, imm)) {
// The move was successful; nothing to do here.
} else {
// Pre-shift the immediate to the least-significant bits of the register.
int shift_low;
if (reg_size == 64) {
shift_low = base::bits::CountTrailingZeros(imm);
} else {
DCHECK_EQ(reg_size, 32);
shift_low = base::bits::CountTrailingZeros(static_cast<uint32_t>(imm));
}
if (mode == kLimitShiftForSP) {
// When applied to the stack pointer, the subsequent arithmetic operation
// can use the extend form to shift left by a maximum of four bits. Right
// shifts are not allowed, so we filter them out later before the new
// immediate is tested.
shift_low = std::min(shift_low, 4);
}
int64_t imm_low = imm >> shift_low;
// Pre-shift the immediate to the most-significant bits of the register. We
// insert set bits in the least-significant bits, as this creates a
// different immediate that may be encodable using movn or orr-immediate.
// If this new immediate is encodable, the set bits will be eliminated by
// the post shift on the following instruction.
int shift_high = CountLeadingZeros(imm, reg_size);
int64_t imm_high = (imm << shift_high) | ((INT64_C(1) << shift_high) - 1);
if ((mode != kNoShift) && TryOneInstrMoveImmediate(dst, imm_low)) {
// The new immediate has been moved into the destination's low bits:
// return a new leftward-shifting operand.
return Operand(dst, LSL, shift_low);
} else if ((mode == kAnyShift) && TryOneInstrMoveImmediate(dst, imm_high)) {
// The new immediate has been moved into the destination's high bits:
// return a new rightward-shifting operand.
return Operand(dst, LSR, shift_high);
} else {
// Use the generic move operation to set up the immediate.
Mov(dst, imm);
}
}
return Operand(dst);
}
void TurboAssembler::AddSubMacro(const Register& rd, const Register& rn,
const Operand& operand, FlagsUpdate S,
AddSubOp op) {
if (operand.IsZero() && rd == rn && rd.Is64Bits() && rn.Is64Bits() &&
!operand.NeedsRelocation(this) && (S == LeaveFlags)) {
// The instruction would be a nop. Avoid generating useless code.
return;
}
if (operand.NeedsRelocation(this)) {
UseScratchRegisterScope temps(this);
Register temp = temps.AcquireX();
Ldr(temp, operand.immediate());
AddSubMacro(rd, rn, temp, S, op);
} else if ((operand.IsImmediate() &&
!IsImmAddSub(operand.ImmediateValue())) ||
(rn.IsZero() && !operand.IsShiftedRegister()) ||
(operand.IsShiftedRegister() && (operand.shift() == ROR))) {
UseScratchRegisterScope temps(this);
Register temp = temps.AcquireSameSizeAs(rn);
if (operand.IsImmediate()) {
PreShiftImmMode mode = kAnyShift;
// If the destination or source register is the stack pointer, we can
// only pre-shift the immediate right by values supported in the add/sub
// extend encoding.
if (rd == sp) {
// If the destination is SP and flags will be set, we can't pre-shift
// the immediate at all.
mode = (S == SetFlags) ? kNoShift : kLimitShiftForSP;
} else if (rn == sp) {
mode = kLimitShiftForSP;
}
Operand imm_operand =
MoveImmediateForShiftedOp(temp, operand.ImmediateValue(), mode);
AddSub(rd, rn, imm_operand, S, op);
} else {
Mov(temp, operand);
AddSub(rd, rn, temp, S, op);
}
} else {
AddSub(rd, rn, operand, S, op);
}
}
void TurboAssembler::AddSubWithCarryMacro(const Register& rd,
const Register& rn,
const Operand& operand, FlagsUpdate S,
AddSubWithCarryOp op) {
DCHECK(rd.SizeInBits() == rn.SizeInBits());
UseScratchRegisterScope temps(this);
if (operand.NeedsRelocation(this)) {
Register temp = temps.AcquireX();
Ldr(temp, operand.immediate());
AddSubWithCarryMacro(rd, rn, temp, S, op);
} else if (operand.IsImmediate() ||
(operand.IsShiftedRegister() && (operand.shift() == ROR))) {
// Add/sub with carry (immediate or ROR shifted register.)
Register temp = temps.AcquireSameSizeAs(rn);
Mov(temp, operand);
AddSubWithCarry(rd, rn, temp, S, op);
} else if (operand.IsShiftedRegister() && (operand.shift_amount() != 0)) {
// Add/sub with carry (shifted register).
DCHECK(operand.reg().SizeInBits() == rd.SizeInBits());
DCHECK(operand.shift() != ROR);
DCHECK(is_uintn(operand.shift_amount(), rd.SizeInBits() == kXRegSizeInBits
? kXRegSizeInBitsLog2
: kWRegSizeInBitsLog2));
Register temp = temps.AcquireSameSizeAs(rn);
EmitShift(temp, operand.reg(), operand.shift(), operand.shift_amount());
AddSubWithCarry(rd, rn, temp, S, op);
} else if (operand.IsExtendedRegister()) {
// Add/sub with carry (extended register).
DCHECK(operand.reg().SizeInBits() <= rd.SizeInBits());
// Add/sub extended supports a shift <= 4. We want to support exactly the
// same modes.
DCHECK_LE(operand.shift_amount(), 4);
DCHECK(operand.reg().Is64Bits() ||
((operand.extend() != UXTX) && (operand.extend() != SXTX)));
Register temp = temps.AcquireSameSizeAs(rn);
EmitExtendShift(temp, operand.reg(), operand.extend(),
operand.shift_amount());
AddSubWithCarry(rd, rn, temp, S, op);
} else {
// The addressing mode is directly supported by the instruction.
AddSubWithCarry(rd, rn, operand, S, op);
}
}
void TurboAssembler::LoadStoreMacro(const CPURegister& rt,
const MemOperand& addr, LoadStoreOp op) {
int64_t offset = addr.offset();
unsigned size = CalcLSDataSize(op);
// Check if an immediate offset fits in the immediate field of the
// appropriate instruction. If not, emit two instructions to perform
// the operation.
if (addr.IsImmediateOffset() && !IsImmLSScaled(offset, size) &&
!IsImmLSUnscaled(offset)) {
// Immediate offset that can't be encoded using unsigned or unscaled
// addressing modes.
UseScratchRegisterScope temps(this);
Register temp = temps.AcquireSameSizeAs(addr.base());
Mov(temp, addr.offset());
LoadStore(rt, MemOperand(addr.base(), temp), op);
} else if (addr.IsPostIndex() && !IsImmLSUnscaled(offset)) {
// Post-index beyond unscaled addressing range.
LoadStore(rt, MemOperand(addr.base()), op);
add(addr.base(), addr.base(), offset);
} else if (addr.IsPreIndex() && !IsImmLSUnscaled(offset)) {
// Pre-index beyond unscaled addressing range.
add(addr.base(), addr.base(), offset);
LoadStore(rt, MemOperand(addr.base()), op);
} else {
// Encodable in one load/store instruction.
LoadStore(rt, addr, op);
}
}
void TurboAssembler::LoadStorePairMacro(const CPURegister& rt,
const CPURegister& rt2,
const MemOperand& addr,
LoadStorePairOp op) {
// TODO(all): Should we support register offset for load-store-pair?
DCHECK(!addr.IsRegisterOffset());
int64_t offset = addr.offset();
unsigned size = CalcLSPairDataSize(op);
// Check if the offset fits in the immediate field of the appropriate
// instruction. If not, emit two instructions to perform the operation.
if (IsImmLSPair(offset, size)) {
// Encodable in one load/store pair instruction.
LoadStorePair(rt, rt2, addr, op);
} else {
Register base = addr.base();
if (addr.IsImmediateOffset()) {
UseScratchRegisterScope temps(this);
Register temp = temps.AcquireSameSizeAs(base);
Add(temp, base, offset);
LoadStorePair(rt, rt2, MemOperand(temp), op);
} else if (addr.IsPostIndex()) {
LoadStorePair(rt, rt2, MemOperand(base), op);
Add(base, base, offset);
} else {
DCHECK(addr.IsPreIndex());
Add(base, base, offset);
LoadStorePair(rt, rt2, MemOperand(base), op);
}
}
}
bool TurboAssembler::NeedExtraInstructionsOrRegisterBranch(
Label* label, ImmBranchType b_type) {
bool need_longer_range = false;
// There are two situations in which we care about the offset being out of
// range:
// - The label is bound but too far away.
// - The label is not bound but linked, and the previous branch
// instruction in the chain is too far away.
if (label->is_bound() || label->is_linked()) {
need_longer_range =
!Instruction::IsValidImmPCOffset(b_type, label->pos() - pc_offset());
}
if (!need_longer_range && !label->is_bound()) {
int max_reachable_pc = pc_offset() + Instruction::ImmBranchRange(b_type);
unresolved_branches_.insert(std::pair<int, FarBranchInfo>(
max_reachable_pc, FarBranchInfo(pc_offset(), label)));
// Also maintain the next pool check.
next_veneer_pool_check_ = Min(
next_veneer_pool_check_, max_reachable_pc - kVeneerDistanceCheckMargin);
}
return need_longer_range;
}
void TurboAssembler::Adr(const Register& rd, Label* label, AdrHint hint) {
DCHECK(allow_macro_instructions());
DCHECK(!rd.IsZero());
if (hint == kAdrNear) {
adr(rd, label);
return;
}
DCHECK_EQ(hint, kAdrFar);
if (label->is_bound()) {
int label_offset = label->pos() - pc_offset();
if (Instruction::IsValidPCRelOffset(label_offset)) {
adr(rd, label);
} else {
DCHECK_LE(label_offset, 0);
int min_adr_offset = -(1 << (Instruction::ImmPCRelRangeBitwidth - 1));
adr(rd, min_adr_offset);
Add(rd, rd, label_offset - min_adr_offset);
}
} else {
UseScratchRegisterScope temps(this);
Register scratch = temps.AcquireX();
InstructionAccurateScope scope(this,
PatchingAssembler::kAdrFarPatchableNInstrs);
adr(rd, label);
for (int i = 0; i < PatchingAssembler::kAdrFarPatchableNNops; ++i) {
nop(ADR_FAR_NOP);
}
movz(scratch, 0);
}
}
void TurboAssembler::B(Label* label, BranchType type, Register reg, int bit) {
DCHECK((reg == NoReg || type >= kBranchTypeFirstUsingReg) &&
(bit == -1 || type >= kBranchTypeFirstUsingBit));
if (kBranchTypeFirstCondition <= type && type <= kBranchTypeLastCondition) {
B(static_cast<Condition>(type), label);
} else {
switch (type) {
case always:
B(label);
break;
case never:
break;
case reg_zero:
Cbz(reg, label);
break;
case reg_not_zero:
Cbnz(reg, label);
break;
case reg_bit_clear:
Tbz(reg, bit, label);
break;
case reg_bit_set:
Tbnz(reg, bit, label);
break;
default:
UNREACHABLE();
}
}
}
void TurboAssembler::B(Label* label, Condition cond) {
DCHECK(allow_macro_instructions());
DCHECK((cond != al) && (cond != nv));
Label done;
bool need_extra_instructions =
NeedExtraInstructionsOrRegisterBranch(label, CondBranchType);
if (need_extra_instructions) {
b(&done, NegateCondition(cond));
B(label);
} else {
b(label, cond);
}
bind(&done);
}
void TurboAssembler::Tbnz(const Register& rt, unsigned bit_pos, Label* label) {
DCHECK(allow_macro_instructions());
Label done;
bool need_extra_instructions =
NeedExtraInstructionsOrRegisterBranch(label, TestBranchType);
if (need_extra_instructions) {
tbz(rt, bit_pos, &done);
B(label);
} else {
tbnz(rt, bit_pos, label);
}
bind(&done);
}
void TurboAssembler::Tbz(const Register& rt, unsigned bit_pos, Label* label) {
DCHECK(allow_macro_instructions());
Label done;
bool need_extra_instructions =
NeedExtraInstructionsOrRegisterBranch(label, TestBranchType);
if (need_extra_instructions) {
tbnz(rt, bit_pos, &done);
B(label);
} else {
tbz(rt, bit_pos, label);
}
bind(&done);
}
void TurboAssembler::Cbnz(const Register& rt, Label* label) {
DCHECK(allow_macro_instructions());
Label done;
bool need_extra_instructions =
NeedExtraInstructionsOrRegisterBranch(label, CompareBranchType);
if (need_extra_instructions) {
cbz(rt, &done);
B(label);
} else {
cbnz(rt, label);
}
bind(&done);
}
void TurboAssembler::Cbz(const Register& rt, Label* label) {
DCHECK(allow_macro_instructions());
Label done;
bool need_extra_instructions =
NeedExtraInstructionsOrRegisterBranch(label, CompareBranchType);
if (need_extra_instructions) {
cbnz(rt, &done);
B(label);
} else {
cbz(rt, label);
}
bind(&done);
}
// Pseudo-instructions.
void TurboAssembler::Abs(const Register& rd, const Register& rm,
Label* is_not_representable, Label* is_representable) {
DCHECK(allow_macro_instructions());
DCHECK(AreSameSizeAndType(rd, rm));
Cmp(rm, 1);
Cneg(rd, rm, lt);
// If the comparison sets the v flag, the input was the smallest value
// representable by rm, and the mathematical result of abs(rm) is not
// representable using two's complement.
if ((is_not_representable != nullptr) && (is_representable != nullptr)) {
B(is_not_representable, vs);
B(is_representable);
} else if (is_not_representable != nullptr) {
B(is_not_representable, vs);
} else if (is_representable != nullptr) {
B(is_representable, vc);
}
}
// Abstracted stack operations.
void TurboAssembler::Push(const CPURegister& src0, const CPURegister& src1,
const CPURegister& src2, const CPURegister& src3,
const CPURegister& src4, const CPURegister& src5,
const CPURegister& src6, const CPURegister& src7) {
DCHECK(AreSameSizeAndType(src0, src1, src2, src3, src4, src5, src6, src7));
int count = 5 + src5.is_valid() + src6.is_valid() + src6.is_valid();
int size = src0.SizeInBytes();
DCHECK_EQ(0, (size * count) % 16);
PushHelper(4, size, src0, src1, src2, src3);
PushHelper(count - 4, size, src4, src5, src6, src7);
}
void TurboAssembler::Pop(const CPURegister& dst0, const CPURegister& dst1,
const CPURegister& dst2, const CPURegister& dst3,
const CPURegister& dst4, const CPURegister& dst5,
const CPURegister& dst6, const CPURegister& dst7) {
// It is not valid to pop into the same register more than once in one
// instruction, not even into the zero register.
DCHECK(!AreAliased(dst0, dst1, dst2, dst3, dst4, dst5, dst6, dst7));
DCHECK(AreSameSizeAndType(dst0, dst1, dst2, dst3, dst4, dst5, dst6, dst7));
DCHECK(dst0.is_valid());
int count = 5 + dst5.is_valid() + dst6.is_valid() + dst7.is_valid();
int size = dst0.SizeInBytes();
DCHECK_EQ(0, (size * count) % 16);
PopHelper(4, size, dst0, dst1, dst2, dst3);
PopHelper(count - 4, size, dst4, dst5, dst6, dst7);
}
void MacroAssembler::PushMultipleTimes(CPURegister src, Register count) {
UseScratchRegisterScope temps(this);
Register temp = temps.AcquireSameSizeAs(count);
Label loop, leftover2, leftover1, done;
Subs(temp, count, 4);
B(mi, &leftover2);
// Push groups of four first.
Bind(&loop);
Subs(temp, temp, 4);
PushHelper(4, src.SizeInBytes(), src, src, src, src);
B(pl, &loop);
// Push groups of two.
Bind(&leftover2);
Tbz(count, 1, &leftover1);
PushHelper(2, src.SizeInBytes(), src, src, NoReg, NoReg);
// Push the last one (if required).
Bind(&leftover1);
Tbz(count, 0, &done);
PushHelper(1, src.SizeInBytes(), src, NoReg, NoReg, NoReg);
Bind(&done);
}
void TurboAssembler::PushHelper(int count, int size, const CPURegister& src0,
const CPURegister& src1,
const CPURegister& src2,
const CPURegister& src3) {
// Ensure that we don't unintentially modify scratch or debug registers.
InstructionAccurateScope scope(this);
DCHECK(AreSameSizeAndType(src0, src1, src2, src3));
DCHECK(size == src0.SizeInBytes());
// When pushing multiple registers, the store order is chosen such that
// Push(a, b) is equivalent to Push(a) followed by Push(b).
switch (count) {
case 1:
DCHECK(src1.IsNone() && src2.IsNone() && src3.IsNone());
str(src0, MemOperand(sp, -1 * size, PreIndex));
break;
case 2:
DCHECK(src2.IsNone() && src3.IsNone());
stp(src1, src0, MemOperand(sp, -2 * size, PreIndex));
break;
case 3:
DCHECK(src3.IsNone());
stp(src2, src1, MemOperand(sp, -3 * size, PreIndex));
str(src0, MemOperand(sp, 2 * size));
break;
case 4:
// Skip over 4 * size, then fill in the gap. This allows four W registers
// to be pushed using sp, whilst maintaining 16-byte alignment for sp
// at all times.
stp(src3, src2, MemOperand(sp, -4 * size, PreIndex));
stp(src1, src0, MemOperand(sp, 2 * size));
break;
default:
UNREACHABLE();
}
}
void TurboAssembler::PopHelper(int count, int size, const CPURegister& dst0,
const CPURegister& dst1, const CPURegister& dst2,
const CPURegister& dst3) {
// Ensure that we don't unintentially modify scratch or debug registers.
InstructionAccurateScope scope(this);
DCHECK(AreSameSizeAndType(dst0, dst1, dst2, dst3));
DCHECK(size == dst0.SizeInBytes());
// When popping multiple registers, the load order is chosen such that
// Pop(a, b) is equivalent to Pop(a) followed by Pop(b).
switch (count) {
case 1:
DCHECK(dst1.IsNone() && dst2.IsNone() && dst3.IsNone());
ldr(dst0, MemOperand(sp, 1 * size, PostIndex));
break;
case 2:
DCHECK(dst2.IsNone() && dst3.IsNone());
ldp(dst0, dst1, MemOperand(sp, 2 * size, PostIndex));
break;
case 3:
DCHECK(dst3.IsNone());
ldr(dst2, MemOperand(sp, 2 * size));
ldp(dst0, dst1, MemOperand(sp, 3 * size, PostIndex));
break;
case 4:
// Load the higher addresses first, then load the lower addresses and
// skip the whole block in the second instruction. This allows four W
// registers to be popped using sp, whilst maintaining 16-byte alignment
// for sp at all times.
ldp(dst2, dst3, MemOperand(sp, 2 * size));
ldp(dst0, dst1, MemOperand(sp, 4 * size, PostIndex));
break;
default:
UNREACHABLE();
}
}
void TurboAssembler::PokePair(const CPURegister& src1, const CPURegister& src2,
int offset) {
DCHECK(AreSameSizeAndType(src1, src2));
DCHECK((offset >= 0) && ((offset % src1.SizeInBytes()) == 0));
Stp(src1, src2, MemOperand(sp, offset));
}
void MacroAssembler::PeekPair(const CPURegister& dst1, const CPURegister& dst2,
int offset) {
DCHECK(AreSameSizeAndType(dst1, dst2));
DCHECK((offset >= 0) && ((offset % dst1.SizeInBytes()) == 0));
Ldp(dst1, dst2, MemOperand(sp, offset));
}
void MacroAssembler::PushCalleeSavedRegisters() {
#ifdef V8_ENABLE_CONTROL_FLOW_INTEGRITY
Paciasp();
#endif
{
// Ensure that the macro-assembler doesn't use any scratch registers.
InstructionAccurateScope scope(this);
MemOperand tos(sp, -2 * static_cast<int>(kXRegSize), PreIndex);
stp(d14, d15, tos);
stp(d12, d13, tos);
stp(d10, d11, tos);
stp(d8, d9, tos);
STATIC_ASSERT(
EntryFrameConstants::kCalleeSavedRegisterBytesPushedBeforeFpLrPair ==
8 * kSystemPointerSize);
stp(x29, x30, tos); // fp, lr
STATIC_ASSERT(
EntryFrameConstants::kCalleeSavedRegisterBytesPushedAfterFpLrPair ==
10 * kSystemPointerSize);
stp(x27, x28, tos);
stp(x25, x26, tos);
stp(x23, x24, tos);
stp(x21, x22, tos);
stp(x19, x20, tos);
}
}
void MacroAssembler::PopCalleeSavedRegisters() {
{
// Ensure that the macro-assembler doesn't use any scratch registers.
InstructionAccurateScope scope(this);
MemOperand tos(sp, 2 * kXRegSize, PostIndex);
ldp(x19, x20, tos);
ldp(x21, x22, tos);
ldp(x23, x24, tos);
ldp(x25, x26, tos);
ldp(x27, x28, tos);
ldp(x29, x30, tos);
ldp(d8, d9, tos);
ldp(d10, d11, tos);
ldp(d12, d13, tos);
ldp(d14, d15, tos);
}
#ifdef V8_ENABLE_CONTROL_FLOW_INTEGRITY
Autiasp();
#endif
}
void TurboAssembler::AssertSpAligned() {
if (emit_debug_code()) {
HardAbortScope hard_abort(this); // Avoid calls to Abort.
// Arm64 requires the stack pointer to be 16-byte aligned prior to address
// calculation.
UseScratchRegisterScope scope(this);
Register temp = scope.AcquireX();
Mov(temp, sp);
Tst(temp, 15);
Check(eq, AbortReason::kUnexpectedStackPointer);
}
}
void TurboAssembler::CopySlots(int dst, Register src, Register slot_count) {
DCHECK(!src.IsZero());
UseScratchRegisterScope scope(this);
Register dst_reg = scope.AcquireX();
SlotAddress(dst_reg, dst);
SlotAddress(src, src);
CopyDoubleWords(dst_reg, src, slot_count);
}
void TurboAssembler::CopySlots(Register dst, Register src,
Register slot_count) {
DCHECK(!dst.IsZero() && !src.IsZero());
SlotAddress(dst, dst);
SlotAddress(src, src);
CopyDoubleWords(dst, src, slot_count);
}
void TurboAssembler::CopyDoubleWords(Register dst, Register src, Register count,
CopyDoubleWordsMode mode) {
DCHECK(!AreAliased(dst, src, count));
if (emit_debug_code()) {
Register pointer1 = dst;
Register pointer2 = src;
if (mode == kSrcLessThanDst) {
pointer1 = src;
pointer2 = dst;
}
// Copy requires pointer1 < pointer2 || (pointer1 - pointer2) >= count.
Label pointer1_below_pointer2;
Subs(pointer1, pointer1, pointer2);
B(lt, &pointer1_below_pointer2);
Cmp(pointer1, count);
Check(ge, AbortReason::kOffsetOutOfRange);
Bind(&pointer1_below_pointer2);
Add(pointer1, pointer1, pointer2);
}
static_assert(kSystemPointerSize == kDRegSize,
"pointers must be the same size as doubles");
if (mode == kDstLessThanSrcAndReverse) {
Add(src, src, Operand(count, LSL, kSystemPointerSizeLog2));
Sub(src, src, kSystemPointerSize);
}
int src_direction = (mode == kDstLessThanSrc) ? 1 : -1;
int dst_direction = (mode == kSrcLessThanDst) ? -1 : 1;
UseScratchRegisterScope scope(this);
VRegister temp0 = scope.AcquireD();
VRegister temp1 = scope.AcquireD();
Label pairs, loop, done;
Tbz(count, 0, &pairs);
Ldr(temp0, MemOperand(src, src_direction * kSystemPointerSize, PostIndex));
Sub(count, count, 1);
Str(temp0, MemOperand(dst, dst_direction * kSystemPointerSize, PostIndex));
Bind(&pairs);
if (mode == kSrcLessThanDst) {
// Adjust pointers for post-index ldp/stp with negative offset:
Sub(dst, dst, kSystemPointerSize);
Sub(src, src, kSystemPointerSize);
} else if (mode == kDstLessThanSrcAndReverse) {
Sub(src, src, kSystemPointerSize);
}
Bind(&loop);
Cbz(count, &done);
Ldp(temp0, temp1,
MemOperand(src, 2 * src_direction * kSystemPointerSize, PostIndex));
Sub(count, count, 2);
if (mode == kDstLessThanSrcAndReverse) {
Stp(temp1, temp0,
MemOperand(dst, 2 * dst_direction * kSystemPointerSize, PostIndex));
} else {
Stp(temp0, temp1,
MemOperand(dst, 2 * dst_direction * kSystemPointerSize, PostIndex));
}
B(&loop);
// TODO(all): large copies may benefit from using temporary Q registers
// to copy four double words per iteration.
Bind(&done);
}
void TurboAssembler::SlotAddress(Register dst, int slot_offset) {
Add(dst, sp, slot_offset << kSystemPointerSizeLog2);
}
void TurboAssembler::SlotAddress(Register dst, Register slot_offset) {
Add(dst, sp, Operand(slot_offset, LSL, kSystemPointerSizeLog2));
}
void TurboAssembler::AssertFPCRState(Register fpcr) {
if (emit_debug_code()) {
Label unexpected_mode, done;
UseScratchRegisterScope temps(this);
if (fpcr.IsNone()) {
fpcr = temps.AcquireX();
Mrs(fpcr, FPCR);
}
// Settings left to their default values:
// - Assert that flush-to-zero is not set.
Tbnz(fpcr, FZ_offset, &unexpected_mode);
// - Assert that the rounding mode is nearest-with-ties-to-even.
STATIC_ASSERT(FPTieEven == 0);
Tst(fpcr, RMode_mask);
B(eq, &done);
Bind(&unexpected_mode);
Abort(AbortReason::kUnexpectedFPCRMode);
Bind(&done);
}
}
void TurboAssembler::CanonicalizeNaN(const VRegister& dst,
const VRegister& src) {
AssertFPCRState();
// Subtracting 0.0 preserves all inputs except for signalling NaNs, which
// become quiet NaNs. We use fsub rather than fadd because fsub preserves -0.0
// inputs: -0.0 + 0.0 = 0.0, but -0.0 - 0.0 = -0.0.
Fsub(dst, src, fp_zero);
}
void TurboAssembler::LoadRoot(Register destination, RootIndex index) {
// TODO(jbramley): Most root values are constants, and can be synthesized
// without a load. Refer to the ARM back end for details.
Ldr(destination,
MemOperand(kRootRegister, RootRegisterOffsetForRootIndex(index)));
}
void TurboAssembler::Move(Register dst, Smi src) { Mov(dst, src); }
void TurboAssembler::MovePair(Register dst0, Register src0, Register dst1,
Register src1) {
DCHECK_NE(dst0, dst1);
if (dst0 != src1) {
Mov(dst0, src0);
Mov(dst1, src1);
} else if (dst1 != src0) {
// Swap the order of the moves to resolve the overlap.
Mov(dst1, src1);
Mov(dst0, src0);
} else {
// Worse case scenario, this is a swap.
Swap(dst0, src0);
}
}
void TurboAssembler::Swap(Register lhs, Register rhs) {
DCHECK(lhs.IsSameSizeAndType(rhs));
DCHECK_NE(lhs, rhs);
UseScratchRegisterScope temps(this);
Register temp = temps.AcquireX();
Mov(temp, rhs);
Mov(rhs, lhs);
Mov(lhs, temp);
}
void TurboAssembler::Swap(VRegister lhs, VRegister rhs) {
DCHECK(lhs.IsSameSizeAndType(rhs));
DCHECK_NE(lhs, rhs);
UseScratchRegisterScope temps(this);
VRegister temp = VRegister::no_reg();
if (lhs.IsS()) {
temp = temps.AcquireS();
} else if (lhs.IsD()) {
temp = temps.AcquireD();
} else {
DCHECK(lhs.IsQ());
temp = temps.AcquireQ();
}
Mov(temp, rhs);
Mov(rhs, lhs);
Mov(lhs, temp);
}
void TurboAssembler::AssertSmi(Register object, AbortReason reason) {
if (emit_debug_code()) {
STATIC_ASSERT(kSmiTag == 0);
Tst(object, kSmiTagMask);
Check(eq, reason);
}
}
void MacroAssembler::AssertNotSmi(Register object, AbortReason reason) {
if (emit_debug_code()) {
STATIC_ASSERT(kSmiTag == 0);
Tst(object, kSmiTagMask);
Check(ne, reason);
}
}
void MacroAssembler::AssertConstructor(Register object) {
if (emit_debug_code()) {
AssertNotSmi(object, AbortReason::kOperandIsASmiAndNotAConstructor);
UseScratchRegisterScope temps(this);
Register temp = temps.AcquireX();
LoadMap(temp, object);
Ldrb(temp, FieldMemOperand(temp, Map::kBitFieldOffset));
Tst(temp, Operand(Map::Bits1::IsConstructorBit::kMask));
Check(ne, AbortReason::kOperandIsNotAConstructor);
}
}
void MacroAssembler::AssertFunction(Register object) {
if (emit_debug_code()) {
AssertNotSmi(object, AbortReason::kOperandIsASmiAndNotAFunction);
UseScratchRegisterScope temps(this);
Register temp = temps.AcquireX();
CompareObjectType(object, temp, temp, JS_FUNCTION_TYPE);
Check(eq, AbortReason::kOperandIsNotAFunction);
}
}
void MacroAssembler::AssertBoundFunction(Register object) {
if (emit_debug_code()) {
AssertNotSmi(object, AbortReason::kOperandIsASmiAndNotABoundFunction);
UseScratchRegisterScope temps(this);
Register temp = temps.AcquireX();
CompareObjectType(object, temp, temp, JS_BOUND_FUNCTION_TYPE);
Check(eq, AbortReason::kOperandIsNotABoundFunction);
}
}
void MacroAssembler::AssertGeneratorObject(Register object) {
if (!emit_debug_code()) return;
AssertNotSmi(object, AbortReason::kOperandIsASmiAndNotAGeneratorObject);
// Load map
UseScratchRegisterScope temps(this);
Register temp = temps.AcquireX();
LoadMap(temp, object);
Label do_check;
// Load instance type and check if JSGeneratorObject
CompareInstanceType(temp, temp, JS_GENERATOR_OBJECT_TYPE);
B(eq, &do_check);
// Check if JSAsyncFunctionObject
Cmp(temp, JS_ASYNC_FUNCTION_OBJECT_TYPE);
B(eq, &do_check);
// Check if JSAsyncGeneratorObject
Cmp(temp, JS_ASYNC_GENERATOR_OBJECT_TYPE);
bind(&do_check);
// Restore generator object to register and perform assertion
Check(eq, AbortReason::kOperandIsNotAGeneratorObject);
}
void MacroAssembler::AssertUndefinedOrAllocationSite(Register object) {
if (emit_debug_code()) {
UseScratchRegisterScope temps(this);
Register scratch = temps.AcquireX();
Label done_checking;
AssertNotSmi(object);
JumpIfRoot(object, RootIndex::kUndefinedValue, &done_checking);
LoadMap(scratch, object);
CompareInstanceType(scratch, scratch, ALLOCATION_SITE_TYPE);
Assert(eq, AbortReason::kExpectedUndefinedOrCell);
Bind(&done_checking);
}
}
void TurboAssembler::AssertPositiveOrZero(Register value) {
if (emit_debug_code()) {
Label done;
int sign_bit = value.Is64Bits() ? kXSignBit : kWSignBit;
Tbz(value, sign_bit, &done);
Abort(AbortReason::kUnexpectedNegativeValue);
Bind(&done);
}
}
void MacroAssembler::CallRuntime(const Runtime::Function* f, int num_arguments,
SaveFPRegsMode save_doubles) {
// All arguments must be on the stack before this function is called.
// x0 holds the return value after the call.
// Check that the number of arguments matches what the function expects.
// If f->nargs is -1, the function can accept a variable number of arguments.
CHECK(f->nargs < 0 || f->nargs == num_arguments);
// Place the necessary arguments.
Mov(x0, num_arguments);
Mov(x1, ExternalReference::Create(f));
Handle<Code> code =
CodeFactory::CEntry(isolate(), f->result_size, save_doubles);
Call(code, RelocInfo::CODE_TARGET);
}
void MacroAssembler::JumpToExternalReference(const ExternalReference& builtin,
bool builtin_exit_frame) {
Mov(x1, builtin);
Handle<Code> code = CodeFactory::CEntry(isolate(), 1, kDontSaveFPRegs,
kArgvOnStack, builtin_exit_frame);
Jump(code, RelocInfo::CODE_TARGET);
}
void MacroAssembler::JumpToInstructionStream(Address entry) {
Ldr(kOffHeapTrampolineRegister, Operand(entry, RelocInfo::OFF_HEAP_TARGET));
Br(kOffHeapTrampolineRegister);
}
void MacroAssembler::TailCallRuntime(Runtime::FunctionId fid) {
const Runtime::Function* function = Runtime::FunctionForId(fid);
DCHECK_EQ(1, function->result_size);
if (function->nargs >= 0) {
// TODO(1236192): Most runtime routines don't need the number of
// arguments passed in because it is constant. At some point we
// should remove this need and make the runtime routine entry code
// smarter.
Mov(x0, function->nargs);
}
JumpToExternalReference(ExternalReference::Create(fid));
}
int TurboAssembler::ActivationFrameAlignment() {
#if V8_HOST_ARCH_ARM64
// Running on the real platform. Use the alignment as mandated by the local
// environment.
// Note: This will break if we ever start generating snapshots on one ARM
// platform for another ARM platform with a different alignment.
return base::OS::ActivationFrameAlignment();
#else // V8_HOST_ARCH_ARM64
// If we are using the simulator then we should always align to the expected
// alignment. As the simulator is used to generate snapshots we do not know
// if the target platform will need alignment, so this is controlled from a
// flag.
return FLAG_sim_stack_alignment;
#endif // V8_HOST_ARCH_ARM64
}
void TurboAssembler::CallCFunction(ExternalReference function,
int num_of_reg_args) {
CallCFunction(function, num_of_reg_args, 0);
}
void TurboAssembler::CallCFunction(ExternalReference function,
int num_of_reg_args,
int num_of_double_args) {
UseScratchRegisterScope temps(this);
Register temp = temps.AcquireX();
Mov(temp, function);
CallCFunction(temp, num_of_reg_args, num_of_double_args);
}
static const int kRegisterPassedArguments = 8;
void TurboAssembler::CallCFunction(Register function, int num_of_reg_args,
int num_of_double_args) {
DCHECK_LE(num_of_reg_args + num_of_double_args, kMaxCParameters);
DCHECK(has_frame());
// If we're passing doubles, we're limited to the following prototypes
// (defined by ExternalReference::Type):
// BUILTIN_COMPARE_CALL: int f(double, double)
// BUILTIN_FP_FP_CALL: double f(double, double)
// BUILTIN_FP_CALL: double f(double)
// BUILTIN_FP_INT_CALL: double f(double, int)
if (num_of_double_args > 0) {
DCHECK_LE(num_of_reg_args, 1);
DCHECK_LE(num_of_double_args + num_of_reg_args, 2);
}
// Save the frame pointer and PC so that the stack layout remains iterable,
// even without an ExitFrame which normally exists between JS and C frames.
Register pc_scratch = x4;
Register addr_scratch = x5;
Push(pc_scratch, addr_scratch);
Label get_pc;
Bind(&get_pc);
Adr(pc_scratch, &get_pc);
// See x64 code for reasoning about how to address the isolate data fields.
if (root_array_available()) {
Str(pc_scratch,
MemOperand(kRootRegister, IsolateData::fast_c_call_caller_pc_offset()));
Str(fp,
MemOperand(kRootRegister, IsolateData::fast_c_call_caller_fp_offset()));
} else {
DCHECK_NOT_NULL(isolate());
Mov(addr_scratch,
ExternalReference::fast_c_call_caller_pc_address(isolate()));
Str(pc_scratch, MemOperand(addr_scratch));
Mov(addr_scratch,
ExternalReference::fast_c_call_caller_fp_address(isolate()));
Str(fp, MemOperand(addr_scratch));
}
Pop(addr_scratch, pc_scratch);
// Call directly. The function called cannot cause a GC, or allow preemption,
// so the return address in the link register stays correct.
Call(function);
// We don't unset the PC; the FP is the source of truth.
if (root_array_available()) {
Str(xzr,
MemOperand(kRootRegister, IsolateData::fast_c_call_caller_fp_offset()));
} else {
DCHECK_NOT_NULL(isolate());
Push(addr_scratch, xzr);
Mov(addr_scratch,
ExternalReference::fast_c_call_caller_fp_address(isolate()));
Str(xzr, MemOperand(addr_scratch));
Pop(xzr, addr_scratch);
}
if (num_of_reg_args > kRegisterPassedArguments) {
// Drop the register passed arguments.
int claim_slots = RoundUp(num_of_reg_args - kRegisterPassedArguments, 2);
Drop(claim_slots);
}
}
void TurboAssembler::LoadFromConstantsTable(Register destination,
int constant_index) {
DCHECK(RootsTable::IsImmortalImmovable(RootIndex::kBuiltinsConstantsTable));
LoadRoot(destination, RootIndex::kBuiltinsConstantsTable);
LoadTaggedPointerField(
destination, FieldMemOperand(destination, FixedArray::OffsetOfElementAt(
constant_index)));
}
void TurboAssembler::LoadRootRelative(Register destination, int32_t offset) {
Ldr(destination, MemOperand(kRootRegister, offset));
}
void TurboAssembler::LoadRootRegisterOffset(Register destination,
intptr_t offset) {
if (offset == 0) {
Mov(destination, kRootRegister);
} else {
Add(destination, kRootRegister, offset);
}
}
void TurboAssembler::Jump(Register target, Condition cond) {
if (cond == nv) return;
Label done;
if (cond != al) B(NegateCondition(cond), &done);
Br(target);
Bind(&done);
}
void TurboAssembler::JumpHelper(int64_t offset, RelocInfo::Mode rmode,
Condition cond) {
if (cond == nv) return;
Label done;
if (cond != al) B(NegateCondition(cond), &done);
if (CanUseNearCallOrJump(rmode)) {
DCHECK(IsNearCallOffset(offset));
near_jump(static_cast<int>(offset), rmode);
} else {
UseScratchRegisterScope temps(this);
Register temp = temps.AcquireX();
uint64_t imm = reinterpret_cast<uint64_t>(pc_) + offset * kInstrSize;
Mov(temp, Immediate(imm, rmode));
Br(temp);
}
Bind(&done);
}
namespace {
// The calculated offset is either:
// * the 'target' input unmodified if this is a Wasm call, or
// * the offset of the target from the current PC, in instructions, for any
// other type of call.
static int64_t CalculateTargetOffset(Address target, RelocInfo::Mode rmode,
byte* pc) {
int64_t offset = static_cast<int64_t>(target);
// The target of WebAssembly calls is still an index instead of an actual
// address at this point, and needs to be encoded as-is.
if (rmode != RelocInfo::WASM_CALL && rmode != RelocInfo::WASM_STUB_CALL) {
offset -= reinterpret_cast<int64_t>(pc);
DCHECK_EQ(offset % kInstrSize, 0);
offset = offset / static_cast<int>(kInstrSize);
}
return offset;
}
} // namespace
void TurboAssembler::Jump(Address target, RelocInfo::Mode rmode,
Condition cond) {
JumpHelper(CalculateTargetOffset(target, rmode, pc_), rmode, cond);
}
void TurboAssembler::Jump(Handle<Code> code, RelocInfo::Mode rmode,
Condition cond) {
DCHECK(RelocInfo::IsCodeTarget(rmode));
DCHECK_IMPLIES(options().isolate_independent_code,
Builtins::IsIsolateIndependentBuiltin(*code));
if (options().inline_offheap_trampolines) {
int builtin_index = Builtins::kNoBuiltinId;
if (isolate()->builtins()->IsBuiltinHandle(code, &builtin_index) &&
Builtins::IsIsolateIndependent(builtin_index)) {
// Inline the trampoline.
RecordCommentForOffHeapTrampoline(builtin_index);
CHECK_NE(builtin_index, Builtins::kNoBuiltinId);
UseScratchRegisterScope temps(this);
Register scratch = temps.AcquireX();
EmbeddedData d = EmbeddedData::FromBlob();
Address entry = d.InstructionStartOfBuiltin(builtin_index);
Ldr(scratch, Operand(entry, RelocInfo::OFF_HEAP_TARGET));
Jump(scratch, cond);
return;
}
}
if (CanUseNearCallOrJump(rmode)) {
EmbeddedObjectIndex index = AddEmbeddedObject(code);
DCHECK(is_int32(index));
JumpHelper(static_cast<int64_t>(index), rmode, cond);
} else {
Jump(code.address(), rmode, cond);
}
}
void TurboAssembler::Jump(const ExternalReference& reference) {
UseScratchRegisterScope temps(this);
Register scratch = temps.AcquireX();
Mov(scratch, reference);
Jump(scratch);
}
void TurboAssembler::Call(Register target) {
BlockPoolsScope scope(this);
Blr(target);
}
void TurboAssembler::Call(Address target, RelocInfo::Mode rmode) {
BlockPoolsScope scope(this);
if (CanUseNearCallOrJump(rmode)) {
int64_t offset = CalculateTargetOffset(target, rmode, pc_);
DCHECK(IsNearCallOffset(offset));
near_call(static_cast<int>(offset), rmode);
} else {
IndirectCall(target, rmode);
}
}
void TurboAssembler::Call(Handle<Code> code, RelocInfo::Mode rmode) {
DCHECK_IMPLIES(options().isolate_independent_code,
Builtins::IsIsolateIndependentBuiltin(*code));
BlockPoolsScope scope(this);
if (options().inline_offheap_trampolines) {
int builtin_index = Builtins::kNoBuiltinId;
if (isolate()->builtins()->IsBuiltinHandle(code, &builtin_index) &&
Builtins::IsIsolateIndependent(builtin_index)) {
// Inline the trampoline.
CallBuiltin(builtin_index);
return;
}
}
DCHECK(code->IsExecutable());
if (CanUseNearCallOrJump(rmode)) {
EmbeddedObjectIndex index = AddEmbeddedObject(code);
DCHECK(is_int32(index));
near_call(static_cast<int32_t>(index), rmode);
} else {
IndirectCall(code.address(), rmode);
}
}
void TurboAssembler::Call(ExternalReference target) {
UseScratchRegisterScope temps(this);
Register temp = temps.AcquireX();
Mov(temp, target);
Call(temp);
}
void TurboAssembler::LoadEntryFromBuiltinIndex(Register builtin_index) {
// The builtin_index register contains the builtin index as a Smi.
// Untagging is folded into the indexing operand below.
if (SmiValuesAre32Bits()) {
Asr(builtin_index, builtin_index, kSmiShift - kSystemPointerSizeLog2);
Add(builtin_index, builtin_index,
IsolateData::builtin_entry_table_offset());
Ldr(builtin_index, MemOperand(kRootRegister, builtin_index));
} else {
DCHECK(SmiValuesAre31Bits());
if (COMPRESS_POINTERS_BOOL) {
Add(builtin_index, kRootRegister,
Operand(builtin_index.W(), SXTW, kSystemPointerSizeLog2 - kSmiShift));
} else {
Add(builtin_index, kRootRegister,
Operand(builtin_index, LSL, kSystemPointerSizeLog2 - kSmiShift));
}
Ldr(builtin_index,
MemOperand(builtin_index, IsolateData::builtin_entry_table_offset()));
}
}
void TurboAssembler::CallBuiltinByIndex(Register builtin_index) {
LoadEntryFromBuiltinIndex(builtin_index);
Call(builtin_index);
}
void TurboAssembler::CallBuiltin(int builtin_index) {
DCHECK(Builtins::IsBuiltinId(builtin_index));
RecordCommentForOffHeapTrampoline(builtin_index);
CHECK_NE(builtin_index, Builtins::kNoBuiltinId);
UseScratchRegisterScope temps(this);
Register scratch = temps.AcquireX();
EmbeddedData d = EmbeddedData::FromBlob();
Address entry = d.InstructionStartOfBuiltin(builtin_index);
Ldr(scratch, Operand(entry, RelocInfo::OFF_HEAP_TARGET));
Call(scratch);
}
void TurboAssembler::LoadCodeObjectEntry(Register destination,
Register code_object) {
// Code objects are called differently depending on whether we are generating
// builtin code (which will later be embedded into the binary) or compiling
// user JS code at runtime.
// * Builtin code runs in --jitless mode and thus must not call into on-heap
// Code targets. Instead, we dispatch through the builtins entry table.
// * Codegen at runtime does not have this restriction and we can use the
// shorter, branchless instruction sequence. The assumption here is that
// targets are usually generated code and not builtin Code objects.
if (options().isolate_independent_code) {
DCHECK(root_array_available());
Label if_code_is_off_heap, out;
UseScratchRegisterScope temps(this);
Register scratch = temps.AcquireX();
DCHECK(!AreAliased(destination, scratch));
DCHECK(!AreAliased(code_object, scratch));
// Check whether the Code object is an off-heap trampoline. If so, call its
// (off-heap) entry point directly without going through the (on-heap)
// trampoline. Otherwise, just call the Code object as always.
Ldrsw(scratch, FieldMemOperand(code_object, Code::kFlagsOffset));
Tst(scratch, Operand(Code::IsOffHeapTrampoline::kMask));
B(ne, &if_code_is_off_heap);
// Not an off-heap trampoline object, the entry point is at
// Code::raw_instruction_start().
Add(destination, code_object, Code::kHeaderSize - kHeapObjectTag);
B(&out);
// An off-heap trampoline, the entry point is loaded from the builtin entry
// table.
bind(&if_code_is_off_heap);
Ldrsw(scratch, FieldMemOperand(code_object, Code::kBuiltinIndexOffset));
Lsl(destination, scratch, kSystemPointerSizeLog2);
Add(destination, destination, kRootRegister);
Ldr(destination,
MemOperand(destination, IsolateData::builtin_entry_table_offset()));
bind(&out);
} else {
Add(destination, code_object, Code::kHeaderSize - kHeapObjectTag);
}
}
void TurboAssembler::CallCodeObject(Register code_object) {
LoadCodeObjectEntry(code_object, code_object);
Call(code_object);
}
void TurboAssembler::JumpCodeObject(Register code_object) {
LoadCodeObjectEntry(code_object, code_object);
Jump(code_object);
}
void TurboAssembler::StoreReturnAddressAndCall(Register target) {
// This generates the final instruction sequence for calls to C functions
// once an exit frame has been constructed.
//
// Note that this assumes the caller code (i.e. the Code object currently
// being generated) is immovable or that the callee function cannot trigger
// GC, since the callee function will return to it.
UseScratchRegisterScope temps(this);
temps.Exclude(x16, x17);
Label return_location;
Adr(x17, &return_location);
#ifdef V8_ENABLE_CONTROL_FLOW_INTEGRITY
Add(x16, sp, kSystemPointerSize);
Pacia1716();
#endif
Poke(x17, 0);
if (emit_debug_code()) {
// Verify that the slot below fp[kSPOffset]-8 points to the signed return
// location.
Ldr(x16, MemOperand(fp, ExitFrameConstants::kSPOffset));
Ldr(x16, MemOperand(x16, -static_cast<int64_t>(kXRegSize)));
Cmp(x16, x17);
Check(eq, AbortReason::kReturnAddressNotFoundInFrame);
}
Blr(target);
Bind(&return_location);
}
void TurboAssembler::IndirectCall(Address target, RelocInfo::Mode rmode) {
UseScratchRegisterScope temps(this);
Register temp = temps.AcquireX();
Mov(temp, Immediate(target, rmode));
Blr(temp);
}
bool TurboAssembler::IsNearCallOffset(int64_t offset) {
return is_int26(offset);
}
void TurboAssembler::CallForDeoptimization(Address target, int deopt_id,
Label* exit, DeoptimizeKind kind) {
BlockPoolsScope scope(this);
int64_t offset = static_cast<int64_t>(target) -
static_cast<int64_t>(options().code_range_start);
DCHECK_EQ(offset % kInstrSize, 0);
offset = offset / static_cast<int>(kInstrSize);
DCHECK(IsNearCallOffset(offset));
near_call(static_cast<int>(offset), RelocInfo::RUNTIME_ENTRY);
DCHECK_EQ(SizeOfCodeGeneratedSince(exit),
(kind == DeoptimizeKind::kLazy)
? Deoptimizer::kLazyDeoptExitSize
: Deoptimizer::kNonLazyDeoptExitSize);
USE(exit, kind);
}
void TurboAssembler::PrepareForTailCall(Register callee_args_count,
Register caller_args_count,
Register scratch0, Register scratch1) {
DCHECK(!AreAliased(callee_args_count, caller_args_count, scratch0, scratch1));
// Calculate the end of destination area where we will put the arguments
// after we drop current frame. We add kSystemPointerSize to count the
// receiver argument which is not included into formal parameters count.
Register dst_reg = scratch0;
Add(dst_reg, fp, Operand(caller_args_count, LSL, kSystemPointerSizeLog2));
Add(dst_reg, dst_reg,
StandardFrameConstants::kCallerSPOffset + kSystemPointerSize);
// Round dst_reg up to a multiple of 16 bytes, so that we overwrite any
// potential padding.
Add(dst_reg, dst_reg, 15);
Bic(dst_reg, dst_reg, 15);
Register src_reg = caller_args_count;
// Calculate the end of source area. +kSystemPointerSize is for the receiver.
Add(src_reg, sp, Operand(callee_args_count, LSL, kSystemPointerSizeLog2));
Add(src_reg, src_reg, kSystemPointerSize);
// Round src_reg up to a multiple of 16 bytes, so we include any potential
// padding in the copy.
Add(src_reg, src_reg, 15);
Bic(src_reg, src_reg, 15);
if (FLAG_debug_code) {
Cmp(src_reg, dst_reg);
Check(lo, AbortReason::kStackAccessBelowStackPointer);
}
// Restore caller's frame pointer and return address now as they will be
// overwritten by the copying loop.
RestoreFPAndLR();
// Now copy callee arguments to the caller frame going backwards to avoid
// callee arguments corruption (source and destination areas could overlap).
// Both src_reg and dst_reg are pointing to the word after the one to copy,
// so they must be pre-decremented in the loop.
Register tmp_reg = scratch1;
Label loop, entry;
B(&entry);
bind(&loop);
Ldr(tmp_reg, MemOperand(src_reg, -kSystemPointerSize, PreIndex));
Str(tmp_reg, MemOperand(dst_reg, -kSystemPointerSize, PreIndex));
bind(&entry);
Cmp(sp, src_reg);
B(ne, &loop);
// Leave current frame.
Mov(sp, dst_reg);
}
void MacroAssembler::InvokePrologue(Register expected_parameter_count,
Register actual_parameter_count,
Label* done, InvokeFlag flag) {
Label regular_invoke;
// Check whether the expected and actual arguments count match. The registers
// are set up according to contract with ArgumentsAdaptorTrampoline:
// x0: actual arguments count.
// x1: function (passed through to callee).
// x2: expected arguments count.
// The code below is made a lot easier because the calling code already sets
// up actual and expected registers according to the contract.
DCHECK_EQ(actual_parameter_count, x0);
DCHECK_EQ(expected_parameter_count, x2);
// If actual == expected perform a regular invocation.
Cmp(expected_parameter_count, actual_parameter_count);
B(eq, ®ular_invoke);
// The argument counts mismatch, generate a call to the argument adaptor.
Handle<Code> adaptor = BUILTIN_CODE(isolate(), ArgumentsAdaptorTrampoline);
if (flag == CALL_FUNCTION) {
Call(adaptor);
// If the arg counts don't match, no extra code is emitted by
// MAsm::InvokeFunctionCode and we can just fall through.
B(done);
} else {
Jump(adaptor, RelocInfo::CODE_TARGET);
}
Bind(®ular_invoke);
}
void MacroAssembler::CallDebugOnFunctionCall(Register fun, Register new_target,
Register expected_parameter_count,
Register actual_parameter_count) {
// Load receiver to pass it later to DebugOnFunctionCall hook.
Peek(x4, ReceiverOperand(actual_parameter_count));
FrameScope frame(this, has_frame() ? StackFrame::NONE : StackFrame::INTERNAL);
if (!new_target.is_valid()) new_target = padreg;
// Save values on stack.
SmiTag(expected_parameter_count);
SmiTag(actual_parameter_count);
Push(expected_parameter_count, actual_parameter_count, new_target, fun);
Push(fun, x4);
CallRuntime(Runtime::kDebugOnFunctionCall);
// Restore values from stack.
Pop(fun, new_target, actual_parameter_count, expected_parameter_count);
SmiUntag(actual_parameter_count);
SmiUntag(expected_parameter_count);
}
void MacroAssembler::InvokeFunctionCode(Register function, Register new_target,
Register expected_parameter_count,
Register actual_parameter_count,
InvokeFlag flag) {
// You can't call a function without a valid frame.
DCHECK(flag == JUMP_FUNCTION || has_frame());
DCHECK_EQ(function, x1);
DCHECK_IMPLIES(new_target.is_valid(), new_target == x3);
// On function call, call into the debugger if necessary.
Label debug_hook, continue_after_hook;
{
Mov(x4, ExternalReference::debug_hook_on_function_call_address(isolate()));
Ldrsb(x4, MemOperand(x4));
Cbnz(x4, &debug_hook);
}
bind(&continue_after_hook);
// Clear the new.target register if not given.
if (!new_target.is_valid()) {
LoadRoot(x3, RootIndex::kUndefinedValue);
}
Label done;
InvokePrologue(expected_parameter_count, actual_parameter_count, &done, flag);
// If actual != expected, InvokePrologue will have handled the call through
// the argument adaptor mechanism.
// The called function expects the call kind in x5.
// We call indirectly through the code field in the function to
// allow recompilation to take effect without changing any of the
// call sites.
Register code = kJavaScriptCallCodeStartRegister;
LoadTaggedPointerField(code,
FieldMemOperand(function, JSFunction::kCodeOffset));
if (flag == CALL_FUNCTION) {
CallCodeObject(code);
} else {
DCHECK(flag == JUMP_FUNCTION);
JumpCodeObject(code);
}
B(&done);
// Deferred debug hook.
bind(&debug_hook);
CallDebugOnFunctionCall(function, new_target, expected_parameter_count,
actual_parameter_count);
B(&continue_after_hook);
// Continue here if InvokePrologue does handle the invocation due to
// mismatched parameter counts.
Bind(&done);
}
Operand MacroAssembler::ReceiverOperand(Register arg_count) {
#ifdef V8_REVERSE_JSARGS
return Operand(0);
#else
return Operand(arg_count, LSL, kXRegSizeLog2);
#endif
}
void MacroAssembler::InvokeFunctionWithNewTarget(
Register function, Register new_target, Register actual_parameter_count,
InvokeFlag flag) {
// You can't call a function without a valid frame.
DCHECK(flag == JUMP_FUNCTION || has_frame());
// Contract with called JS functions requires that function is passed in x1.
// (See FullCodeGenerator::Generate().)
DCHECK_EQ(function, x1);
Register expected_parameter_count = x2;
LoadTaggedPointerField(cp,
FieldMemOperand(function, JSFunction::kContextOffset));
// The number of arguments is stored as an int32_t, and -1 is a marker
// (kDontAdaptArgumentsSentinel), so we need sign
// extension to correctly handle it.
LoadTaggedPointerField(
expected_parameter_count,
FieldMemOperand(function, JSFunction::kSharedFunctionInfoOffset));
Ldrh(expected_parameter_count,
FieldMemOperand(expected_parameter_count,
SharedFunctionInfo::kFormalParameterCountOffset));
InvokeFunctionCode(function, new_target, expected_parameter_count,
actual_parameter_count, flag);
}
void MacroAssembler::InvokeFunction(Register function,
Register expected_parameter_count,
Register actual_parameter_count,
InvokeFlag flag) {
// You can't call a function without a valid frame.
DCHECK(flag == JUMP_FUNCTION || has_frame());
// Contract with called JS functions requires that function is passed in x1.
// (See FullCodeGenerator::Generate().)
DCHECK_EQ(function, x1);
// Set up the context.
LoadTaggedPointerField(cp,
FieldMemOperand(function, JSFunction::kContextOffset));
InvokeFunctionCode(function, no_reg, expected_parameter_count,
actual_parameter_count, flag);
}
void TurboAssembler::TryConvertDoubleToInt64(Register result,
DoubleRegister double_input,
Label* done) {
// Try to convert with an FPU convert instruction. It's trivial to compute
// the modulo operation on an integer register so we convert to a 64-bit
// integer.
//
// Fcvtzs will saturate to INT64_MIN (0x800...00) or INT64_MAX (0x7FF...FF)
// when the double is out of range. NaNs and infinities will be converted to 0
// (as ECMA-262 requires).
Fcvtzs(result.X(), double_input);
// The values INT64_MIN (0x800...00) or INT64_MAX (0x7FF...FF) are not
// representable using a double, so if the result is one of those then we know
// that saturation occurred, and we need to manually handle the conversion.
//
// It is easy to detect INT64_MIN and INT64_MAX because adding or subtracting
// 1 will cause signed overflow.
Cmp(result.X(), 1);
Ccmp(result.X(), -1, VFlag, vc);
B(vc, done);
}
void TurboAssembler::TruncateDoubleToI(Isolate* isolate, Zone* zone,
Register result,
DoubleRegister double_input,
StubCallMode stub_mode,
LinkRegisterStatus lr_status) {
Label done;
// Try to convert the double to an int64. If successful, the bottom 32 bits
// contain our truncated int32 result.
TryConvertDoubleToInt64(result, double_input, &done);
// If we fell through then inline version didn't succeed - call stub instead.
if (lr_status == kLRHasNotBeenSaved) {
Push<TurboAssembler::kSignLR>(lr, double_input);
} else {
Push<TurboAssembler::kDontStoreLR>(xzr, double_input);
}
// DoubleToI preserves any registers it needs to clobber.
if (stub_mode == StubCallMode::kCallWasmRuntimeStub) {
Call(wasm::WasmCode::kDoubleToI, RelocInfo::WASM_STUB_CALL);
} else if (options().inline_offheap_trampolines) {
CallBuiltin(Builtins::kDoubleToI);
} else {
Call(BUILTIN_CODE(isolate, DoubleToI), RelocInfo::CODE_TARGET);
}
Ldr(result, MemOperand(sp, 0));
DCHECK_EQ(xzr.SizeInBytes(), double_input.SizeInBytes());
if (lr_status == kLRHasNotBeenSaved) {
// Pop into xzr here to drop the double input on the stack:
Pop<TurboAssembler::kAuthLR>(xzr, lr);
} else {
Drop(2);
}
Bind(&done);
// Keep our invariant that the upper 32 bits are zero.
Uxtw(result.W(), result.W());
}
void TurboAssembler::Prologue() {
Push<TurboAssembler::kSignLR>(lr, fp, cp, x1);
Add(fp, sp, StandardFrameConstants::kFixedFrameSizeFromFp);
}
void TurboAssembler::EnterFrame(StackFrame::Type type) {
UseScratchRegisterScope temps(this);
if (type == StackFrame::INTERNAL || type == StackFrame::WASM_DEBUG_BREAK) {
Register type_reg = temps.AcquireX();
Mov(type_reg, StackFrame::TypeToMarker(type));
Push<TurboAssembler::kSignLR>(lr, fp, type_reg, padreg);
const int kFrameSize =
TypedFrameConstants::kFixedFrameSizeFromFp + kSystemPointerSize;
Add(fp, sp, kFrameSize);
// sp[3] : lr
// sp[2] : fp
// sp[1] : type
// sp[0] : for alignment
} else if (type == StackFrame::WASM ||
type == StackFrame::WASM_COMPILE_LAZY ||
type == StackFrame::WASM_EXIT) {
Register type_reg = temps.AcquireX();
Mov(type_reg, StackFrame::TypeToMarker(type));
Push<TurboAssembler::kSignLR>(lr, fp);
Mov(fp, sp);
Push(type_reg, padreg);
// sp[3] : lr
// sp[2] : fp
// sp[1] : type
// sp[0] : for alignment
} else {
DCHECK_EQ(type, StackFrame::CONSTRUCT);
Register type_reg = temps.AcquireX();
Mov(type_reg, StackFrame::TypeToMarker(type));
// Users of this frame type push a context pointer after the type field,
// so do it here to keep the stack pointer aligned.
Push<TurboAssembler::kSignLR>(lr, fp, type_reg, cp);
// The context pointer isn't part of the fixed frame, so add an extra slot
// to account for it.
Add(fp, sp,
TypedFrameConstants::kFixedFrameSizeFromFp + kSystemPointerSize);
// sp[3] : lr
// sp[2] : fp
// sp[1] : type
// sp[0] : cp
}
}
void TurboAssembler::LeaveFrame(StackFrame::Type type) {
// Drop the execution stack down to the frame pointer and restore
// the caller frame pointer and return address.
Mov(sp, fp);
Pop<TurboAssembler::kAuthLR>(fp, lr);
}
void MacroAssembler::ExitFramePreserveFPRegs() {
DCHECK_EQ(kCallerSavedV.Count() % 2, 0);
PushCPURegList(kCallerSavedV);
}
void MacroAssembler::ExitFrameRestoreFPRegs() {
// Read the registers from the stack without popping them. The stack pointer
// will be reset as part of the unwinding process.
CPURegList saved_fp_regs = kCallerSavedV;
DCHECK_EQ(saved_fp_regs.Count() % 2, 0);
int offset = ExitFrameConstants::kLastExitFrameField;
while (!saved_fp_regs.IsEmpty()) {
const CPURegister& dst0 = saved_fp_regs.PopHighestIndex();
const CPURegister& dst1 = saved_fp_regs.PopHighestIndex();
offset -= 2 * kDRegSize;
Ldp(dst1, dst0, MemOperand(fp, offset));
}
}
void MacroAssembler::EnterExitFrame(bool save_doubles, const Register& scratch,
int extra_space,
StackFrame::Type frame_type) {
DCHECK(frame_type == StackFrame::EXIT ||
frame_type == StackFrame::BUILTIN_EXIT);
// Set up the new stack frame.
Push<TurboAssembler::kSignLR>(lr, fp);
Mov(fp, sp);
Mov(scratch, StackFrame::TypeToMarker(frame_type));
Push(scratch, xzr);
// fp[8]: CallerPC (lr)
// fp -> fp[0]: CallerFP (old fp)
// fp[-8]: STUB marker
// sp -> fp[-16]: Space reserved for SPOffset.
STATIC_ASSERT((2 * kSystemPointerSize) ==
ExitFrameConstants::kCallerSPOffset);
STATIC_ASSERT((1 * kSystemPointerSize) ==
ExitFrameConstants::kCallerPCOffset);
STATIC_ASSERT((0 * kSystemPointerSize) ==
ExitFrameConstants::kCallerFPOffset);
STATIC_ASSERT((-2 * kSystemPointerSize) == ExitFrameConstants::kSPOffset);
// Save the frame pointer and context pointer in the top frame.
Mov(scratch,
ExternalReference::Create(IsolateAddressId::kCEntryFPAddress, isolate()));
Str(fp, MemOperand(scratch));
Mov(scratch,
ExternalReference::Create(IsolateAddressId::kContextAddress, isolate()));
Str(cp, MemOperand(scratch));
STATIC_ASSERT((-2 * kSystemPointerSize) ==
ExitFrameConstants::kLastExitFrameField);
if (save_doubles) {
ExitFramePreserveFPRegs();
}
// Round the number of space we need to claim to a multiple of two.
int slots_to_claim = RoundUp(extra_space + 1, 2);
// Reserve space for the return address and for user requested memory.
// We do this before aligning to make sure that we end up correctly
// aligned with the minimum of wasted space.
Claim(slots_to_claim, kXRegSize);
// fp[8]: CallerPC (lr)
// fp -> fp[0]: CallerFP (old fp)
// fp[-8]: STUB marker
// fp[-16]: Space reserved for SPOffset.
// fp[-16 - fp_size]: Saved doubles (if save_doubles is true).
// sp[8]: Extra space reserved for caller (if extra_space != 0).
// sp -> sp[0]: Space reserved for the return address.
// ExitFrame::GetStateForFramePointer expects to find the return address at
// the memory address immediately below the pointer stored in SPOffset.
// It is not safe to derive much else from SPOffset, because the size of the
// padding can vary.
Add(scratch, sp, kXRegSize);
Str(scratch, MemOperand(fp, ExitFrameConstants::kSPOffset));
}
// Leave the current exit frame.
void MacroAssembler::LeaveExitFrame(bool restore_doubles,
const Register& scratch,
const Register& scratch2) {
if (restore_doubles) {
ExitFrameRestoreFPRegs();
}
// Restore the context pointer from the top frame.
Mov(scratch,
ExternalReference::Create(IsolateAddressId::kContextAddress, isolate()));
Ldr(cp, MemOperand(scratch));
if (emit_debug_code()) {
// Also emit debug code to clear the cp in the top frame.
Mov(scratch2, Operand(Context::kInvalidContext));
Mov(scratch, ExternalReference::Create(IsolateAddressId::kContextAddress,
isolate()));
Str(scratch2, MemOperand(scratch));
}
// Clear the frame pointer from the top frame.
Mov(scratch,
ExternalReference::Create(IsolateAddressId::kCEntryFPAddress, isolate()));
Str(xzr, MemOperand(scratch));
// Pop the exit frame.
// fp[8]: CallerPC (lr)
// fp -> fp[0]: CallerFP (old fp)
// fp[...]: The rest of the frame.
Mov(sp, fp);
Pop<TurboAssembler::kAuthLR>(fp, lr);
}
void MacroAssembler::LoadGlobalProxy(Register dst) {
LoadNativeContextSlot(Context::GLOBAL_PROXY_INDEX, dst);
}
void MacroAssembler::LoadWeakValue(Register out, Register in,
Label* target_if_cleared) {
CompareAndBranch(in.W(), Operand(kClearedWeakHeapObjectLower32), eq,
target_if_cleared);
and_(out, in, Operand(~kWeakHeapObjectMask));
}
void MacroAssembler::IncrementCounter(StatsCounter* counter, int value,
Register scratch1, Register scratch2) {
DCHECK_NE(value, 0);
if (FLAG_native_code_counters && counter->Enabled()) {
// This operation has to be exactly 32-bit wide in case the external
// reference table redirects the counter to a uint32_t dummy_stats_counter_
// field.
Mov(scratch2, ExternalReference::Create(counter));
Ldr(scratch1.W(), MemOperand(scratch2));
Add(scratch1.W(), scratch1.W(), value);
Str(scratch1.W(), MemOperand(scratch2));
}
}
void MacroAssembler::DecrementCounter(StatsCounter* counter, int value,
Register scratch1, Register scratch2) {
IncrementCounter(counter, -value, scratch1, scratch2);
}
void MacroAssembler::MaybeDropFrames() {
// Check whether we need to drop frames to restart a function on the stack.
Mov(x1, ExternalReference::debug_restart_fp_address(isolate()));
Ldr(x1, MemOperand(x1));
Tst(x1, x1);
Jump(BUILTIN_CODE(isolate(), FrameDropperTrampoline), RelocInfo::CODE_TARGET,
ne);
}
void MacroAssembler::JumpIfObjectType(Register object, Register map,
Register type_reg, InstanceType type,
Label* if_cond_pass, Condition cond) {
CompareObjectType(object, map, type_reg, type);
B(cond, if_cond_pass);
}
// Sets condition flags based on comparison, and returns type in type_reg.
void MacroAssembler::CompareObjectType(Register object, Register map,
Register type_reg, InstanceType type) {
LoadMap(map, object);
CompareInstanceType(map, type_reg, type);
}
void MacroAssembler::LoadMap(Register dst, Register object) {
LoadTaggedPointerField(dst, FieldMemOperand(object, HeapObject::kMapOffset));
}
// Sets condition flags based on comparison, and returns type in type_reg.
void MacroAssembler::CompareInstanceType(Register map, Register type_reg,
InstanceType type) {
Ldrh(type_reg, FieldMemOperand(map, Map::kInstanceTypeOffset));
Cmp(type_reg, type);
}
void MacroAssembler::LoadElementsKindFromMap(Register result, Register map) {
// Load the map's "bit field 2".
Ldrb(result, FieldMemOperand(map, Map::kBitField2Offset));
// Retrieve elements_kind from bit field 2.
DecodeField<Map::Bits2::ElementsKindBits>(result);
}
void MacroAssembler::CompareRoot(const Register& obj, RootIndex index) {
UseScratchRegisterScope temps(this);
Register temp = temps.AcquireX();
DCHECK(!AreAliased(obj, temp));
LoadRoot(temp, index);
CmpTagged(obj, temp);
}
void MacroAssembler::JumpIfRoot(const Register& obj, RootIndex index,
Label* if_equal) {
CompareRoot(obj, index);
B(eq, if_equal);
}
void MacroAssembler::JumpIfNotRoot(const Register& obj, RootIndex index,
Label* if_not_equal) {
CompareRoot(obj, index);
B(ne, if_not_equal);
}
void MacroAssembler::JumpIfIsInRange(const Register& value,
unsigned lower_limit,
unsigned higher_limit,
Label* on_in_range) {
if (lower_limit != 0) {
UseScratchRegisterScope temps(this);
Register scratch = temps.AcquireW();
Sub(scratch, value, Operand(lower_limit));
CompareAndBranch(scratch, Operand(higher_limit - lower_limit), ls,
on_in_range);
} else {
CompareAndBranch(value, Operand(higher_limit - lower_limit), ls,
on_in_range);
}
}
void TurboAssembler::LoadTaggedPointerField(const Register& destination,
const MemOperand& field_operand) {
if (COMPRESS_POINTERS_BOOL) {
DecompressTaggedPointer(destination, field_operand);
} else {
Ldr(destination, field_operand);
}
}
void TurboAssembler::LoadAnyTaggedField(const Register& destination,
const MemOperand& field_operand) {
if (COMPRESS_POINTERS_BOOL) {
DecompressAnyTagged(destination, field_operand);
} else {
Ldr(destination, field_operand);
}
}
void TurboAssembler::SmiUntagField(Register dst, const MemOperand& src) {
SmiUntag(dst, src);
}
void TurboAssembler::StoreTaggedField(const Register& value,
const MemOperand& dst_field_operand) {
if (COMPRESS_POINTERS_BOOL) {
Str(value.W(), dst_field_operand);
} else {
Str(value, dst_field_operand);
}
}
void TurboAssembler::DecompressTaggedSigned(const Register& destination,
const MemOperand& field_operand) {
RecordComment("[ DecompressTaggedSigned");
Ldr(destination.W(), field_operand);
if (FLAG_debug_code) {
// Corrupt the top 32 bits. Made up of 16 fixed bits and 16 pc offset bits.
Add(destination, destination,
((kDebugZapValue << 16) | (pc_offset() & 0xffff)) << 32);
}
RecordComment("]");
}
void TurboAssembler::DecompressTaggedPointer(const Register& destination,
const MemOperand& field_operand) {
RecordComment("[ DecompressTaggedPointer");
Ldr(destination.W(), field_operand);
Add(destination, kRootRegister, destination);
RecordComment("]");
}
void TurboAssembler::DecompressTaggedPointer(const Register& destination,
const Register& source) {
RecordComment("[ DecompressTaggedPointer");
Add(destination, kRootRegister, Operand(source, UXTW));
RecordComment("]");
}
void TurboAssembler::DecompressAnyTagged(const Register& destination,
const MemOperand& field_operand) {
RecordComment("[ DecompressAnyTagged");
Ldr(destination.W(), field_operand);
Add(destination, kRootRegister, destination);
RecordComment("]");
}
void TurboAssembler::CheckPageFlag(const Register& object, int mask,
Condition cc, Label* condition_met) {
UseScratchRegisterScope temps(this);
Register scratch = temps.AcquireX();
And(scratch, object, ~kPageAlignmentMask);
Ldr(scratch, MemOperand(scratch, MemoryChunk::kFlagsOffset));
if (cc == eq) {
TestAndBranchIfAnySet(scratch, mask, condition_met);
} else {
DCHECK_EQ(cc, ne);
TestAndBranchIfAllClear(scratch, mask, condition_met);
}
}
void MacroAssembler::RecordWriteField(Register object, int offset,
Register value,
LinkRegisterStatus lr_status,
SaveFPRegsMode save_fp,
RememberedSetAction remembered_set_action,
SmiCheck smi_check) {
// First, check if a write barrier is even needed. The tests below
// catch stores of Smis.
Label done;
// Skip the barrier if writing a smi.
if (smi_check == INLINE_SMI_CHECK) {
JumpIfSmi(value, &done);
}
// Although the object register is tagged, the offset is relative to the start
// of the object, so offset must be a multiple of kTaggedSize.
DCHECK(IsAligned(offset, kTaggedSize));
if (emit_debug_code()) {
Label ok;
UseScratchRegisterScope temps(this);
Register scratch = temps.AcquireX();
Add(scratch, object, offset - kHeapObjectTag);
Tst(scratch, kTaggedSize - 1);
B(eq, &ok);
Abort(AbortReason::kUnalignedCellInWriteBarrier);
Bind(&ok);
}
RecordWrite(object, Operand(offset - kHeapObjectTag), value, lr_status,
save_fp, remembered_set_action, OMIT_SMI_CHECK);
Bind(&done);
}
void TurboAssembler::SaveRegisters(RegList registers) {
DCHECK_GT(NumRegs(registers), 0);
CPURegList regs(CPURegister::kRegister, kXRegSizeInBits, registers);
// If we were saving LR, we might need to sign it.
DCHECK(!regs.IncludesAliasOf(lr));
regs.Align();
PushCPURegList(regs);
}
void TurboAssembler::RestoreRegisters(RegList registers) {
DCHECK_GT(NumRegs(registers), 0);
CPURegList regs(CPURegister::kRegister, kXRegSizeInBits, registers);
// If we were saving LR, we might need to sign it.
DCHECK(!regs.IncludesAliasOf(lr));
regs.Align();
PopCPURegList(regs);
}
void TurboAssembler::CallEphemeronKeyBarrier(Register object, Operand offset,
SaveFPRegsMode fp_mode) {
EphemeronKeyBarrierDescriptor descriptor;
RegList registers = descriptor.allocatable_registers();
SaveRegisters(registers);
Register object_parameter(
descriptor.GetRegisterParameter(EphemeronKeyBarrierDescriptor::kObject));
Register slot_parameter(descriptor.GetRegisterParameter(
EphemeronKeyBarrierDescriptor::kSlotAddress));
Register fp_mode_parameter(
descriptor.GetRegisterParameter(EphemeronKeyBarrierDescriptor::kFPMode));
MoveObjectAndSlot(object_parameter, slot_parameter, object, offset);
Mov(fp_mode_parameter, Smi::FromEnum(fp_mode));
Call(isolate()->builtins()->builtin_handle(Builtins::kEphemeronKeyBarrier),
RelocInfo::CODE_TARGET);
RestoreRegisters(registers);
}
void TurboAssembler::CallRecordWriteStub(
Register object, Operand offset, RememberedSetAction remembered_set_action,
SaveFPRegsMode fp_mode) {
CallRecordWriteStub(
object, offset, remembered_set_action, fp_mode,
isolate()->builtins()->builtin_handle(Builtins::kRecordWrite),
kNullAddress);
}
void TurboAssembler::CallRecordWriteStub(
Register object, Operand offset, RememberedSetAction remembered_set_action,
SaveFPRegsMode fp_mode, Address wasm_target) {
CallRecordWriteStub(object, offset, remembered_set_action, fp_mode,
Handle<Code>::null(), wasm_target);
}
void TurboAssembler::CallRecordWriteStub(
Register object, Operand offset, RememberedSetAction remembered_set_action,
SaveFPRegsMode fp_mode, Handle<Code> code_target, Address wasm_target) {
DCHECK_NE(code_target.is_null(), wasm_target == kNullAddress);
// TODO(albertnetymk): For now we ignore remembered_set_action and fp_mode,
// i.e. always emit remember set and save FP registers in RecordWriteStub. If
// large performance regression is observed, we should use these values to
// avoid unnecessary work.
RecordWriteDescriptor descriptor;
RegList registers = descriptor.allocatable_registers();
SaveRegisters(registers);
Register object_parameter(
descriptor.GetRegisterParameter(RecordWriteDescriptor::kObject));
Register slot_parameter(
descriptor.GetRegisterParameter(RecordWriteDescriptor::kSlot));
Register remembered_set_parameter(
descriptor.GetRegisterParameter(RecordWriteDescriptor::kRememberedSet));
Register fp_mode_parameter(
descriptor.GetRegisterParameter(RecordWriteDescriptor::kFPMode));
MoveObjectAndSlot(object_parameter, slot_parameter, object, offset);
Mov(remembered_set_parameter, Smi::FromEnum(remembered_set_action));
Mov(fp_mode_parameter, Smi::FromEnum(fp_mode));
if (code_target.is_null()) {
Call(wasm_target, RelocInfo::WASM_STUB_CALL);
} else {
Call(code_target, RelocInfo::CODE_TARGET);
}
RestoreRegisters(registers);
}
void TurboAssembler::MoveObjectAndSlot(Register dst_object, Register dst_slot,
Register object, Operand offset) {
DCHECK_NE(dst_object, dst_slot);
// If `offset` is a register, it cannot overlap with `object`.
DCHECK_IMPLIES(!offset.IsImmediate(), offset.reg() != object);
// If the slot register does not overlap with the object register, we can
// overwrite it.
if (dst_slot != object) {
Add(dst_slot, object, offset);
Mov(dst_object, object);
return;
}
DCHECK_EQ(dst_slot, object);
// If the destination object register does not overlap with the offset
// register, we can overwrite it.
if (offset.IsImmediate() || (offset.reg() != dst_object)) {
Mov(dst_object, dst_slot);
Add(dst_slot, dst_slot, offset);
return;
}
DCHECK_EQ(dst_object, offset.reg());
// We only have `dst_slot` and `dst_object` left as distinct registers so we
// have to swap them. We write this as a add+sub sequence to avoid using a
// scratch register.
Add(dst_slot, dst_slot, dst_object);
Sub(dst_object, dst_slot, dst_object);
}
// If lr_status is kLRHasBeenSaved, lr will be clobbered.
//
// The register 'object' contains a heap object pointer. The heap object tag is
// shifted away.
void MacroAssembler::RecordWrite(Register object, Operand offset,
Register value, LinkRegisterStatus lr_status,
SaveFPRegsMode fp_mode,
RememberedSetAction remembered_set_action,
SmiCheck smi_check) {
ASM_LOCATION_IN_ASSEMBLER("MacroAssembler::RecordWrite");
DCHECK(!AreAliased(object, value));
if (emit_debug_code()) {
UseScratchRegisterScope temps(this);
Register temp = temps.AcquireX();
Add(temp, object, offset);
LoadTaggedPointerField(temp, MemOperand(temp));
Cmp(temp, value);
Check(eq, AbortReason::kWrongAddressOrValuePassedToRecordWrite);
}
if ((remembered_set_action == OMIT_REMEMBERED_SET &&
!FLAG_incremental_marking) ||
FLAG_disable_write_barriers) {
return;
}
// First, check if a write barrier is even needed. The tests below
// catch stores of smis and stores into the young generation.
Label done;
if (smi_check == INLINE_SMI_CHECK) {
DCHECK_EQ(0, kSmiTag);
JumpIfSmi(value, &done);
}
CheckPageFlag(value, MemoryChunk::kPointersToHereAreInterestingMask, ne,
&done);
CheckPageFlag(object, MemoryChunk::kPointersFromHereAreInterestingMask, ne,
&done);
// Record the actual write.
if (lr_status == kLRHasNotBeenSaved) {
Push<TurboAssembler::kSignLR>(padreg, lr);
}
CallRecordWriteStub(object, offset, remembered_set_action, fp_mode);
if (lr_status == kLRHasNotBeenSaved) {
Pop<TurboAssembler::kAuthLR>(lr, padreg);
}
Bind(&done);
}
void TurboAssembler::Assert(Condition cond, AbortReason reason) {
if (emit_debug_code()) {
Check(cond, reason);
}
}
void TurboAssembler::AssertUnreachable(AbortReason reason) {
if (emit_debug_code()) Abort(reason);
}
void TurboAssembler::Check(Condition cond, AbortReason reason) {
Label ok;
B(cond, &ok);
Abort(reason);
// Will not return here.
Bind(&ok);
}
void TurboAssembler::Trap() { Brk(0); }
void TurboAssembler::DebugBreak() { Debug("DebugBreak", 0, BREAK); }
void TurboAssembler::Abort(AbortReason reason) {
#ifdef DEBUG
RecordComment("Abort message: ");
RecordComment(GetAbortReason(reason));
#endif
// Avoid emitting call to builtin if requested.
if (trap_on_abort()) {
Brk(0);
return;
}
// We need some scratch registers for the MacroAssembler, so make sure we have
// some. This is safe here because Abort never returns.
RegList old_tmp_list = TmpList()->list();
TmpList()->Combine(MacroAssembler::DefaultTmpList());
if (should_abort_hard()) {
// We don't care if we constructed a frame. Just pretend we did.
FrameScope assume_frame(this, StackFrame::NONE);
Mov(w0, static_cast<int>(reason));
Call(ExternalReference::abort_with_reason());
return;
}
// Avoid infinite recursion; Push contains some assertions that use Abort.
HardAbortScope hard_aborts(this);
Mov(x1, Smi::FromInt(static_cast<int>(reason)));
if (!has_frame_) {
// We don't actually want to generate a pile of code for this, so just
// claim there is a stack frame, without generating one.
FrameScope scope(this, StackFrame::NONE);
Call(BUILTIN_CODE(isolate(), Abort), RelocInfo::CODE_TARGET);
} else {
Call(BUILTIN_CODE(isolate(), Abort), RelocInfo::CODE_TARGET);
}
TmpList()->set_list(old_tmp_list);
}
void MacroAssembler::LoadNativeContextSlot(int index, Register dst) {
LoadMap(dst, cp);
LoadTaggedPointerField(
dst, FieldMemOperand(
dst, Map::kConstructorOrBackPointerOrNativeContextOffset));
LoadTaggedPointerField(dst, MemOperand(dst, Context::SlotOffset(index)));
}
// This is the main Printf implementation. All other Printf variants call
// PrintfNoPreserve after setting up one or more PreserveRegisterScopes.
void TurboAssembler::PrintfNoPreserve(const char* format,
const CPURegister& arg0,
const CPURegister& arg1,
const CPURegister& arg2,
const CPURegister& arg3) {
// We cannot handle a caller-saved stack pointer. It doesn't make much sense
// in most cases anyway, so this restriction shouldn't be too serious.
DCHECK(!kCallerSaved.IncludesAliasOf(sp));
// The provided arguments, and their proper procedure-call standard registers.
CPURegister args[kPrintfMaxArgCount] = {arg0, arg1, arg2, arg3};
CPURegister pcs[kPrintfMaxArgCount] = {NoReg, NoReg, NoReg, NoReg};
int arg_count = kPrintfMaxArgCount;
// The PCS varargs registers for printf. Note that x0 is used for the printf
// format string.
static const CPURegList kPCSVarargs =
CPURegList(CPURegister::kRegister, kXRegSizeInBits, 1, arg_count);
static const CPURegList kPCSVarargsFP =
CPURegList(CPURegister::kVRegister, kDRegSizeInBits, 0, arg_count - 1);
// We can use caller-saved registers as scratch values, except for the
// arguments and the PCS registers where they might need to go.
CPURegList tmp_list = kCallerSaved;
tmp_list.Remove(x0); // Used to pass the format string.
tmp_list.Remove(kPCSVarargs);
tmp_list.Remove(arg0, arg1, arg2, arg3);
CPURegList fp_tmp_list = kCallerSavedV;
fp_tmp_list.Remove(kPCSVarargsFP);
fp_tmp_list.Remove(arg0, arg1, arg2, arg3);
// Override the TurboAssembler's scratch register list. The lists will be
// reset automatically at the end of the UseScratchRegisterScope.
UseScratchRegisterScope temps(this);
TmpList()->set_list(tmp_list.list());
FPTmpList()->set_list(fp_tmp_list.list());
// Copies of the printf vararg registers that we can pop from.
CPURegList pcs_varargs = kPCSVarargs;
#ifndef V8_OS_WIN
CPURegList pcs_varargs_fp = kPCSVarargsFP;
#endif
// Place the arguments. There are lots of clever tricks and optimizations we
// could use here, but Printf is a debug tool so instead we just try to keep
// it simple: Move each input that isn't already in the right place to a
// scratch register, then move everything back.
for (unsigned i = 0; i < kPrintfMaxArgCount; i++) {
// Work out the proper PCS register for this argument.
if (args[i].IsRegister()) {
pcs[i] = pcs_varargs.PopLowestIndex().X();
// We might only need a W register here. We need to know the size of the
// argument so we can properly encode it for the simulator call.
if (args[i].Is32Bits()) pcs[i] = pcs[i].W();
} else if (args[i].IsVRegister()) {
// In C, floats are always cast to doubles for varargs calls.
#ifdef V8_OS_WIN
// In case of variadic functions SIMD and Floating-point registers
// aren't used. The general x0-x7 should be used instead.
// https://docs.microsoft.com/en-us/cpp/build/arm64-windows-abi-conventions
pcs[i] = pcs_varargs.PopLowestIndex().X();
#else
pcs[i] = pcs_varargs_fp.PopLowestIndex().D();
#endif
} else {
DCHECK(args[i].IsNone());
arg_count = i;
break;
}
// If the argument is already in the right place, leave it where it is.
if (args[i].Aliases(pcs[i])) continue;
// Otherwise, if the argument is in a PCS argument register, allocate an
// appropriate scratch register and then move it out of the way.
if (kPCSVarargs.IncludesAliasOf(args[i]) ||
kPCSVarargsFP.IncludesAliasOf(args[i])) {
if (args[i].IsRegister()) {
Register old_arg = args[i].Reg();
Register new_arg = temps.AcquireSameSizeAs(old_arg);
Mov(new_arg, old_arg);
args[i] = new_arg;
} else {
VRegister old_arg = args[i].VReg();
VRegister new_arg = temps.AcquireSameSizeAs(old_arg);
Fmov(new_arg, old_arg);
args[i] = new_arg;
}
}
}
// Do a second pass to move values into their final positions and perform any
// conversions that may be required.
for (int i = 0; i < arg_count; i++) {
#ifdef V8_OS_WIN
if (args[i].IsVRegister()) {
if (pcs[i].SizeInBytes() != args[i].SizeInBytes()) {
// If the argument is half- or single-precision
// converts to double-precision before that is
// moved into the one of X scratch register.
VRegister temp0 = temps.AcquireD();
Fcvt(temp0.VReg(), args[i].VReg());
Fmov(pcs[i].Reg(), temp0);
} else {
Fmov(pcs[i].Reg(), args[i].VReg());
}
} else {
Mov(pcs[i].Reg(), args[i].Reg(), kDiscardForSameWReg);
}
#else
DCHECK(pcs[i].type() == args[i].type());
if (pcs[i].IsRegister()) {
Mov(pcs[i].Reg(), args[i].Reg(), kDiscardForSameWReg);
} else {
DCHECK(pcs[i].IsVRegister());
if (pcs[i].SizeInBytes() == args[i].SizeInBytes()) {
Fmov(pcs[i].VReg(), args[i].VReg());
} else {
Fcvt(pcs[i].VReg(), args[i].VReg());
}
}
#endif
}
// Load the format string into x0, as per the procedure-call standard.
//
// To make the code as portable as possible, the format string is encoded
// directly in the instruction stream. It might be cleaner to encode it in a
// literal pool, but since Printf is usually used for debugging, it is
// beneficial for it to be minimally dependent on other features.
Label format_address;
Adr(x0, &format_address);
// Emit the format string directly in the instruction stream.
{
BlockPoolsScope scope(this);
Label after_data;
B(&after_data);
Bind(&format_address);
EmitStringData(format);
Unreachable();
Bind(&after_data);
}
CallPrintf(arg_count, pcs);
}
void TurboAssembler::CallPrintf(int arg_count, const CPURegister* args) {
// A call to printf needs special handling for the simulator, since the system
// printf function will use a different instruction set and the procedure-call
// standard will not be compatible.
if (options().enable_simulator_code) {
InstructionAccurateScope scope(this, kPrintfLength / kInstrSize);
hlt(kImmExceptionIsPrintf);
dc32(arg_count); // kPrintfArgCountOffset
// Determine the argument pattern.
uint32_t arg_pattern_list = 0;
for (int i = 0; i < arg_count; i++) {
uint32_t arg_pattern;
if (args[i].IsRegister()) {
arg_pattern = args[i].Is32Bits() ? kPrintfArgW : kPrintfArgX;
} else {
DCHECK(args[i].Is64Bits());
arg_pattern = kPrintfArgD;
}
DCHECK(arg_pattern < (1 << kPrintfArgPatternBits));
arg_pattern_list |= (arg_pattern << (kPrintfArgPatternBits * i));
}
dc32(arg_pattern_list); // kPrintfArgPatternListOffset
return;
}
Call(ExternalReference::printf_function());
}
void TurboAssembler::Printf(const char* format, CPURegister arg0,
CPURegister arg1, CPURegister arg2,
CPURegister arg3) {
// Printf is expected to preserve all registers, so make sure that none are
// available as scratch registers until we've preserved them.
RegList old_tmp_list = TmpList()->list();
RegList old_fp_tmp_list = FPTmpList()->list();
TmpList()->set_list(0);
FPTmpList()->set_list(0);
CPURegList saved_registers = kCallerSaved;
saved_registers.Align();
// Preserve all caller-saved registers as well as NZCV.
// PushCPURegList asserts that the size of each list is a multiple of 16
// bytes.
PushCPURegList<kDontStoreLR>(saved_registers);
PushCPURegList(kCallerSavedV);
// We can use caller-saved registers as scratch values (except for argN).
CPURegList tmp_list = saved_registers;
CPURegList fp_tmp_list = kCallerSavedV;
tmp_list.Remove(arg0, arg1, arg2, arg3);
fp_tmp_list.Remove(arg0, arg1, arg2, arg3);
TmpList()->set_list(tmp_list.list());
FPTmpList()->set_list(fp_tmp_list.list());
{
UseScratchRegisterScope temps(this);
// If any of the arguments are the current stack pointer, allocate a new
// register for them, and adjust the value to compensate for pushing the
// caller-saved registers.
bool arg0_sp = arg0.is_valid() && sp.Aliases(arg0);
bool arg1_sp = arg1.is_valid() && sp.Aliases(arg1);
bool arg2_sp = arg2.is_valid() && sp.Aliases(arg2);
bool arg3_sp = arg3.is_valid() && sp.Aliases(arg3);
if (arg0_sp || arg1_sp || arg2_sp || arg3_sp) {
// Allocate a register to hold the original stack pointer value, to pass
// to PrintfNoPreserve as an argument.
Register arg_sp = temps.AcquireX();
Add(arg_sp, sp,
saved_registers.TotalSizeInBytes() +
kCallerSavedV.TotalSizeInBytes());
if (arg0_sp) arg0 = Register::Create(arg_sp.code(), arg0.SizeInBits());
if (arg1_sp) arg1 = Register::Create(arg_sp.code(), arg1.SizeInBits());
if (arg2_sp) arg2 = Register::Create(arg_sp.code(), arg2.SizeInBits());
if (arg3_sp) arg3 = Register::Create(arg_sp.code(), arg3.SizeInBits());
}
// Preserve NZCV.
{
UseScratchRegisterScope temps(this);
Register tmp = temps.AcquireX();
Mrs(tmp, NZCV);
Push(tmp, xzr);
}
PrintfNoPreserve(format, arg0, arg1, arg2, arg3);
// Restore NZCV.
{
UseScratchRegisterScope temps(this);
Register tmp = temps.AcquireX();
Pop(xzr, tmp);
Msr(NZCV, tmp);
}
}
PopCPURegList(kCallerSavedV);
PopCPURegList<kDontLoadLR>(saved_registers);
TmpList()->set_list(old_tmp_list);
FPTmpList()->set_list(old_fp_tmp_list);
}
UseScratchRegisterScope::~UseScratchRegisterScope() {
available_->set_list(old_available_);
availablefp_->set_list(old_availablefp_);
}
Register UseScratchRegisterScope::AcquireSameSizeAs(const Register& reg) {
int code = AcquireNextAvailable(available_).code();
return Register::Create(code, reg.SizeInBits());
}
VRegister UseScratchRegisterScope::AcquireSameSizeAs(const VRegister& reg) {
int code = AcquireNextAvailable(availablefp_).code();
return VRegister::Create(code, reg.SizeInBits());
}
CPURegister UseScratchRegisterScope::AcquireNextAvailable(
CPURegList* available) {
CHECK(!available->IsEmpty());
CPURegister result = available->PopLowestIndex();
DCHECK(!AreAliased(result, xzr, sp));
return result;
}
void TurboAssembler::ComputeCodeStartAddress(const Register& rd) {
// We can use adr to load a pc relative location.
adr(rd, -pc_offset());
}
void TurboAssembler::ResetSpeculationPoisonRegister() {
Mov(kSpeculationPoisonRegister, -1);
}
void TurboAssembler::RestoreFPAndLR() {
static_assert(StandardFrameConstants::kCallerFPOffset + kSystemPointerSize ==
StandardFrameConstants::kCallerPCOffset,
"Offsets must be consecutive for ldp!");
#ifdef V8_ENABLE_CONTROL_FLOW_INTEGRITY
// Make sure we can use x16 and x17.
UseScratchRegisterScope temps(this);
temps.Exclude(x16, x17);
// We can load the return address directly into x17.
Add(x16, fp, StandardFrameConstants::kCallerSPOffset);
Ldp(fp, x17, MemOperand(fp, StandardFrameConstants::kCallerFPOffset));
Autia1716();
Mov(lr, x17);
#else
Ldp(fp, lr, MemOperand(fp, StandardFrameConstants::kCallerFPOffset));
#endif
}
void TurboAssembler::StoreReturnAddressInWasmExitFrame(Label* return_location) {
UseScratchRegisterScope temps(this);
temps.Exclude(x16, x17);
Adr(x17, return_location);
#ifdef V8_ENABLE_CONTROL_FLOW_INTEGRITY
Add(x16, fp, WasmExitFrameConstants::kCallingPCOffset + kSystemPointerSize);
Pacia1716();
#endif
Str(x17, MemOperand(fp, WasmExitFrameConstants::kCallingPCOffset));
}
} // namespace internal
} // namespace v8
#endif // V8_TARGET_ARCH_ARM64
|