1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
2189
2190
2191
2192
2193
2194
2195
2196
2197
2198
2199
2200
2201
2202
2203
2204
2205
2206
2207
2208
2209
2210
2211
2212
2213
2214
2215
2216
2217
2218
2219
2220
2221
2222
2223
2224
2225
2226
2227
2228
2229
2230
2231
2232
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
2244
2245
2246
2247
2248
2249
2250
2251
2252
2253
2254
2255
2256
2257
2258
2259
2260
2261
2262
2263
2264
2265
2266
2267
2268
2269
2270
2271
2272
2273
2274
2275
2276
2277
2278
2279
2280
2281
2282
2283
2284
2285
2286
2287
2288
2289
2290
2291
2292
2293
2294
2295
2296
2297
2298
2299
2300
2301
2302
2303
2304
2305
2306
2307
2308
2309
2310
2311
2312
2313
2314
2315
2316
2317
2318
2319
2320
2321
2322
2323
2324
2325
2326
2327
2328
2329
2330
2331
2332
2333
2334
2335
2336
2337
2338
2339
2340
2341
2342
2343
2344
2345
2346
2347
2348
2349
2350
2351
2352
2353
2354
2355
2356
2357
2358
2359
2360
2361
2362
2363
2364
2365
2366
2367
2368
2369
2370
2371
2372
2373
2374
2375
2376
2377
2378
2379
2380
2381
2382
2383
2384
2385
2386
2387
2388
2389
2390
2391
2392
2393
2394
2395
2396
2397
2398
2399
2400
2401
2402
2403
2404
2405
2406
2407
2408
2409
2410
2411
2412
2413
2414
2415
2416
2417
2418
2419
2420
2421
2422
2423
2424
2425
2426
2427
2428
2429
2430
2431
2432
2433
2434
2435
2436
2437
2438
2439
2440
2441
2442
2443
2444
2445
2446
2447
2448
2449
2450
2451
2452
2453
2454
2455
2456
2457
2458
2459
2460
2461
2462
2463
2464
2465
2466
2467
2468
2469
2470
2471
2472
2473
2474
2475
2476
2477
2478
2479
2480
2481
2482
2483
2484
2485
2486
2487
2488
2489
2490
2491
2492
2493
2494
2495
2496
2497
2498
2499
2500
2501
2502
2503
2504
2505
2506
2507
2508
2509
2510
2511
2512
2513
2514
2515
2516
2517
2518
2519
2520
2521
2522
2523
2524
2525
2526
2527
2528
2529
2530
2531
2532
2533
2534
2535
2536
2537
2538
2539
2540
2541
2542
2543
2544
2545
2546
2547
2548
2549
2550
2551
2552
2553
2554
2555
2556
2557
2558
2559
2560
2561
2562
2563
2564
2565
2566
2567
2568
2569
2570
2571
2572
2573
2574
2575
2576
2577
2578
2579
2580
2581
2582
2583
2584
2585
2586
2587
2588
2589
2590
2591
2592
2593
2594
2595
2596
2597
2598
2599
2600
2601
2602
2603
2604
2605
2606
2607
2608
2609
2610
2611
2612
2613
2614
2615
2616
2617
2618
2619
2620
2621
2622
2623
2624
2625
2626
2627
2628
2629
2630
2631
2632
2633
2634
2635
2636
2637
2638
2639
2640
2641
2642
2643
2644
2645
2646
2647
2648
2649
2650
2651
2652
2653
2654
2655
2656
2657
2658
2659
2660
2661
2662
2663
2664
2665
2666
2667
2668
2669
2670
2671
2672
2673
2674
2675
2676
2677
2678
2679
2680
2681
2682
2683
2684
2685
2686
2687
2688
2689
2690
2691
2692
2693
2694
2695
2696
2697
2698
2699
2700
2701
2702
2703
2704
2705
2706
2707
2708
2709
2710
2711
2712
2713
2714
2715
2716
2717
2718
2719
2720
2721
2722
2723
2724
2725
2726
2727
2728
2729
2730
2731
2732
2733
2734
2735
2736
2737
2738
2739
2740
2741
2742
2743
2744
2745
2746
2747
2748
2749
2750
2751
2752
2753
2754
2755
2756
2757
2758
2759
2760
2761
2762
2763
2764
2765
2766
2767
2768
2769
2770
2771
2772
2773
2774
2775
2776
2777
2778
2779
2780
2781
2782
2783
2784
2785
2786
2787
2788
2789
2790
2791
2792
2793
2794
2795
2796
2797
2798
2799
2800
2801
2802
2803
2804
2805
2806
2807
2808
2809
2810
2811
2812
2813
2814
2815
2816
2817
2818
2819
2820
2821
2822
2823
2824
2825
2826
2827
2828
2829
2830
2831
2832
2833
2834
2835
2836
2837
2838
2839
2840
2841
2842
2843
2844
2845
2846
2847
2848
2849
2850
2851
2852
2853
2854
2855
2856
2857
2858
2859
2860
2861
2862
2863
2864
2865
2866
2867
2868
2869
2870
2871
2872
2873
2874
2875
2876
2877
2878
2879
2880
2881
2882
2883
2884
2885
2886
2887
2888
2889
2890
2891
2892
2893
2894
2895
2896
2897
2898
2899
2900
2901
2902
2903
2904
2905
2906
2907
2908
2909
2910
2911
2912
2913
2914
2915
2916
2917
2918
2919
2920
2921
2922
2923
2924
2925
2926
2927
2928
2929
2930
2931
2932
2933
2934
2935
2936
2937
2938
2939
2940
2941
2942
2943
2944
2945
2946
2947
2948
2949
2950
2951
2952
2953
2954
2955
2956
2957
2958
2959
2960
2961
2962
2963
2964
2965
2966
2967
2968
2969
2970
2971
2972
2973
2974
2975
2976
2977
2978
2979
2980
2981
2982
2983
2984
2985
2986
2987
2988
2989
2990
2991
2992
2993
2994
2995
2996
2997
2998
2999
3000
3001
3002
3003
3004
3005
3006
3007
3008
3009
3010
3011
3012
3013
3014
3015
3016
3017
3018
3019
3020
3021
3022
3023
3024
3025
3026
3027
3028
3029
3030
3031
3032
3033
3034
3035
3036
3037
3038
3039
3040
3041
3042
3043
3044
3045
3046
3047
3048
3049
3050
3051
3052
3053
3054
3055
3056
3057
3058
3059
3060
3061
3062
3063
3064
3065
3066
3067
3068
3069
3070
3071
3072
3073
3074
3075
3076
3077
3078
3079
3080
3081
3082
3083
3084
3085
3086
3087
3088
3089
3090
3091
3092
3093
3094
3095
3096
3097
3098
3099
3100
3101
3102
3103
3104
3105
3106
3107
3108
3109
3110
3111
3112
3113
3114
3115
3116
3117
3118
3119
3120
3121
3122
3123
3124
3125
3126
3127
3128
3129
3130
3131
3132
3133
3134
3135
3136
3137
3138
3139
3140
3141
3142
3143
3144
3145
3146
3147
3148
3149
3150
3151
3152
3153
3154
3155
3156
3157
3158
3159
3160
3161
3162
3163
3164
3165
3166
3167
3168
3169
3170
3171
3172
3173
3174
3175
3176
3177
3178
3179
3180
3181
3182
3183
3184
3185
3186
3187
3188
3189
3190
3191
3192
3193
3194
3195
3196
3197
3198
3199
3200
3201
3202
3203
3204
3205
3206
3207
3208
3209
3210
3211
3212
3213
3214
3215
3216
3217
3218
3219
3220
3221
3222
3223
3224
3225
3226
3227
3228
3229
3230
3231
3232
3233
3234
3235
3236
3237
3238
3239
3240
3241
3242
3243
3244
3245
3246
3247
3248
3249
3250
3251
3252
3253
3254
3255
3256
3257
3258
3259
3260
3261
3262
3263
3264
3265
3266
3267
3268
3269
3270
3271
3272
3273
3274
3275
3276
3277
3278
3279
3280
3281
3282
3283
3284
3285
3286
3287
3288
3289
3290
3291
3292
3293
3294
3295
3296
3297
3298
3299
3300
3301
3302
3303
3304
3305
3306
3307
3308
3309
3310
3311
3312
3313
3314
3315
3316
3317
3318
3319
3320
3321
3322
3323
3324
3325
3326
3327
3328
3329
3330
3331
3332
3333
3334
3335
3336
3337
3338
3339
3340
3341
3342
3343
3344
3345
3346
3347
3348
3349
3350
3351
3352
3353
3354
3355
3356
3357
3358
3359
3360
3361
3362
3363
3364
3365
3366
3367
3368
3369
3370
3371
3372
3373
3374
3375
3376
3377
3378
3379
3380
3381
3382
3383
3384
3385
3386
3387
3388
3389
3390
3391
3392
3393
3394
3395
3396
3397
3398
3399
3400
3401
3402
3403
3404
3405
3406
3407
3408
3409
3410
3411
3412
3413
3414
3415
3416
3417
3418
3419
3420
3421
3422
3423
3424
3425
3426
3427
3428
3429
3430
3431
3432
3433
3434
3435
3436
3437
3438
3439
3440
3441
3442
3443
3444
3445
3446
3447
3448
3449
3450
3451
3452
3453
3454
3455
3456
3457
3458
3459
3460
3461
3462
3463
3464
3465
3466
3467
3468
3469
3470
3471
3472
3473
3474
3475
3476
3477
3478
3479
3480
3481
3482
3483
3484
3485
3486
3487
3488
3489
3490
3491
3492
3493
3494
3495
3496
3497
3498
3499
3500
3501
3502
3503
3504
3505
3506
3507
3508
3509
3510
3511
3512
3513
3514
3515
3516
3517
3518
3519
3520
3521
3522
3523
3524
3525
3526
3527
3528
3529
3530
3531
3532
3533
3534
3535
3536
3537
3538
3539
3540
3541
3542
3543
3544
3545
3546
3547
3548
3549
3550
3551
3552
3553
3554
3555
3556
3557
3558
3559
3560
3561
3562
3563
3564
3565
3566
3567
3568
3569
3570
3571
3572
3573
3574
3575
3576
3577
3578
3579
3580
3581
3582
3583
3584
3585
3586
3587
3588
3589
3590
3591
3592
3593
3594
3595
3596
3597
3598
3599
3600
3601
3602
3603
3604
3605
3606
3607
3608
3609
3610
3611
3612
3613
3614
3615
3616
3617
3618
3619
3620
3621
3622
3623
3624
3625
3626
3627
3628
3629
3630
3631
3632
3633
3634
3635
3636
3637
3638
3639
3640
3641
3642
3643
3644
3645
3646
3647
3648
3649
3650
3651
3652
3653
3654
3655
3656
3657
3658
3659
3660
3661
3662
3663
3664
3665
3666
3667
3668
3669
3670
3671
3672
3673
3674
3675
3676
3677
3678
3679
3680
3681
3682
3683
3684
3685
3686
3687
3688
3689
3690
3691
3692
3693
3694
3695
3696
3697
3698
3699
3700
3701
3702
3703
3704
3705
3706
3707
3708
3709
3710
3711
3712
3713
3714
3715
3716
3717
3718
3719
3720
3721
3722
3723
3724
3725
3726
3727
3728
3729
3730
3731
3732
3733
3734
3735
3736
3737
3738
3739
3740
3741
3742
3743
3744
3745
3746
3747
3748
3749
3750
3751
3752
3753
3754
3755
3756
3757
3758
3759
3760
3761
3762
3763
3764
3765
3766
3767
3768
3769
3770
3771
3772
3773
3774
3775
3776
3777
3778
3779
3780
3781
3782
3783
3784
3785
3786
3787
3788
3789
3790
3791
3792
3793
3794
3795
3796
3797
3798
3799
3800
3801
3802
3803
3804
3805
3806
3807
3808
3809
3810
3811
3812
3813
3814
3815
3816
3817
3818
3819
3820
3821
3822
3823
3824
3825
3826
3827
3828
3829
3830
3831
3832
3833
3834
3835
3836
3837
3838
3839
3840
3841
3842
3843
3844
3845
3846
3847
3848
3849
3850
3851
3852
3853
3854
3855
3856
3857
3858
3859
3860
3861
3862
3863
3864
3865
3866
3867
3868
3869
3870
3871
3872
3873
3874
3875
3876
3877
3878
3879
3880
3881
3882
3883
3884
3885
3886
3887
3888
3889
3890
3891
3892
3893
3894
3895
3896
3897
3898
3899
3900
3901
3902
3903
3904
3905
3906
3907
3908
3909
3910
3911
3912
3913
3914
3915
3916
3917
3918
3919
3920
3921
3922
3923
3924
3925
3926
3927
3928
3929
3930
3931
3932
3933
3934
3935
3936
3937
3938
3939
3940
3941
3942
3943
3944
3945
3946
3947
3948
3949
3950
3951
3952
3953
3954
3955
3956
3957
3958
3959
3960
3961
3962
3963
3964
3965
3966
3967
3968
3969
3970
3971
3972
3973
3974
3975
3976
3977
3978
3979
3980
3981
3982
3983
3984
3985
3986
3987
3988
3989
3990
3991
3992
3993
3994
3995
3996
3997
3998
3999
4000
4001
4002
4003
4004
4005
4006
4007
4008
4009
4010
4011
4012
4013
4014
4015
4016
4017
4018
4019
4020
4021
4022
4023
4024
4025
4026
4027
4028
4029
4030
4031
4032
4033
4034
4035
4036
4037
4038
4039
4040
4041
4042
4043
4044
4045
4046
4047
4048
4049
4050
4051
4052
4053
4054
4055
4056
4057
4058
4059
4060
4061
4062
4063
4064
4065
4066
4067
4068
4069
4070
4071
4072
4073
4074
4075
4076
4077
4078
4079
4080
4081
4082
4083
4084
4085
4086
4087
4088
4089
4090
4091
4092
4093
4094
4095
4096
4097
4098
4099
4100
4101
4102
4103
4104
4105
4106
4107
4108
4109
4110
4111
4112
4113
4114
4115
4116
4117
4118
4119
4120
4121
4122
4123
4124
4125
4126
4127
4128
4129
4130
4131
4132
4133
4134
4135
4136
4137
4138
4139
4140
4141
4142
4143
4144
4145
4146
4147
4148
4149
4150
4151
4152
4153
4154
4155
4156
4157
4158
4159
4160
4161
4162
4163
4164
4165
4166
4167
4168
4169
4170
4171
4172
4173
4174
4175
4176
4177
4178
4179
4180
4181
4182
4183
4184
4185
4186
4187
4188
4189
4190
4191
4192
4193
4194
4195
4196
4197
4198
4199
4200
4201
4202
4203
4204
4205
4206
4207
4208
4209
4210
4211
4212
4213
4214
4215
4216
4217
4218
4219
4220
4221
4222
4223
4224
4225
4226
4227
4228
4229
4230
4231
4232
4233
4234
4235
4236
4237
4238
4239
4240
4241
4242
4243
4244
4245
4246
4247
4248
4249
4250
4251
4252
4253
4254
4255
4256
4257
4258
4259
4260
4261
4262
4263
4264
4265
4266
4267
4268
4269
4270
4271
4272
4273
4274
4275
4276
4277
4278
4279
4280
4281
4282
4283
4284
4285
4286
4287
4288
4289
4290
4291
4292
4293
4294
4295
4296
4297
4298
4299
4300
4301
4302
4303
4304
4305
4306
4307
4308
4309
4310
4311
4312
4313
4314
4315
4316
4317
4318
4319
4320
4321
4322
4323
4324
4325
4326
4327
4328
4329
4330
4331
4332
4333
4334
4335
4336
4337
4338
4339
4340
4341
4342
4343
4344
4345
4346
4347
4348
4349
4350
4351
4352
4353
4354
4355
4356
4357
4358
4359
4360
4361
4362
4363
4364
4365
4366
4367
4368
4369
4370
4371
4372
4373
4374
4375
4376
4377
4378
4379
4380
4381
4382
4383
4384
4385
4386
4387
4388
4389
4390
4391
4392
4393
4394
4395
4396
4397
4398
4399
4400
4401
4402
4403
4404
4405
4406
4407
4408
4409
4410
4411
4412
4413
4414
4415
4416
4417
4418
4419
4420
4421
4422
4423
4424
4425
4426
4427
4428
4429
4430
4431
4432
4433
4434
4435
4436
4437
4438
4439
4440
4441
4442
4443
4444
4445
4446
4447
4448
4449
4450
4451
4452
4453
4454
4455
4456
4457
4458
4459
4460
4461
4462
4463
4464
4465
4466
4467
4468
4469
4470
4471
4472
4473
4474
4475
4476
4477
4478
4479
4480
4481
4482
4483
4484
4485
4486
4487
4488
4489
4490
4491
4492
4493
4494
4495
4496
4497
4498
4499
4500
4501
4502
4503
4504
4505
4506
4507
4508
4509
4510
4511
4512
4513
4514
4515
4516
4517
4518
4519
4520
4521
4522
4523
4524
4525
4526
4527
4528
4529
4530
4531
4532
4533
4534
4535
4536
4537
4538
4539
4540
4541
4542
4543
4544
4545
4546
4547
4548
4549
4550
4551
4552
4553
4554
4555
4556
4557
4558
4559
4560
4561
4562
4563
4564
4565
4566
4567
4568
4569
4570
4571
4572
4573
4574
4575
4576
4577
4578
4579
4580
4581
4582
4583
4584
4585
4586
4587
4588
4589
4590
4591
4592
4593
4594
4595
4596
4597
4598
4599
4600
4601
4602
4603
4604
4605
4606
4607
4608
4609
4610
4611
4612
4613
4614
4615
4616
4617
4618
4619
4620
4621
4622
4623
4624
4625
4626
4627
4628
4629
4630
4631
4632
4633
4634
4635
4636
4637
4638
4639
4640
4641
4642
4643
4644
4645
4646
4647
4648
4649
4650
4651
4652
4653
4654
4655
4656
4657
4658
4659
4660
4661
4662
4663
4664
4665
4666
4667
4668
4669
4670
4671
4672
4673
4674
4675
4676
4677
4678
4679
4680
4681
4682
4683
4684
4685
4686
4687
4688
4689
4690
4691
4692
4693
4694
4695
4696
4697
4698
4699
4700
4701
4702
4703
4704
4705
4706
4707
4708
4709
4710
4711
4712
4713
4714
4715
4716
4717
4718
4719
4720
4721
4722
4723
4724
4725
4726
4727
4728
4729
4730
4731
4732
4733
4734
4735
4736
4737
4738
4739
4740
4741
4742
4743
4744
4745
4746
4747
4748
4749
4750
4751
4752
4753
4754
4755
4756
4757
4758
4759
4760
4761
4762
4763
4764
4765
4766
4767
4768
4769
4770
4771
4772
4773
4774
4775
4776
4777
4778
4779
4780
4781
4782
4783
4784
4785
4786
4787
4788
4789
4790
4791
4792
4793
4794
4795
4796
4797
4798
4799
4800
4801
4802
4803
4804
4805
4806
4807
4808
4809
4810
4811
4812
4813
4814
4815
4816
4817
4818
4819
4820
4821
4822
4823
4824
4825
4826
4827
4828
4829
4830
4831
4832
4833
4834
4835
4836
4837
4838
4839
4840
4841
4842
4843
4844
4845
4846
4847
4848
4849
4850
4851
4852
4853
4854
4855
4856
4857
4858
4859
4860
4861
4862
4863
4864
4865
4866
4867
4868
4869
4870
4871
4872
4873
4874
4875
4876
4877
4878
4879
4880
4881
4882
4883
4884
4885
4886
4887
4888
4889
4890
4891
4892
4893
4894
4895
4896
4897
4898
4899
4900
4901
4902
4903
4904
4905
4906
4907
4908
4909
4910
4911
4912
4913
4914
4915
4916
4917
4918
4919
4920
4921
4922
4923
4924
4925
4926
4927
4928
4929
4930
4931
4932
4933
4934
4935
4936
4937
4938
4939
4940
4941
4942
4943
4944
4945
4946
4947
4948
4949
4950
4951
4952
4953
4954
4955
4956
4957
4958
4959
4960
4961
4962
4963
4964
4965
4966
4967
4968
4969
4970
4971
4972
4973
4974
4975
4976
4977
4978
4979
4980
4981
4982
4983
4984
4985
4986
4987
4988
4989
4990
4991
4992
4993
4994
4995
4996
4997
4998
4999
5000
5001
5002
5003
5004
5005
5006
5007
5008
5009
5010
5011
5012
5013
5014
5015
5016
5017
5018
5019
5020
5021
5022
5023
5024
5025
5026
5027
5028
5029
5030
5031
5032
5033
5034
5035
5036
5037
5038
5039
5040
5041
5042
5043
5044
5045
5046
5047
5048
5049
5050
5051
5052
5053
5054
5055
5056
5057
5058
5059
5060
5061
5062
5063
5064
5065
5066
5067
5068
5069
5070
5071
5072
5073
5074
5075
5076
5077
5078
5079
5080
5081
5082
5083
5084
5085
5086
5087
5088
5089
5090
5091
5092
5093
5094
5095
5096
5097
5098
5099
5100
5101
5102
5103
5104
5105
5106
5107
5108
5109
5110
5111
5112
5113
5114
5115
5116
5117
5118
5119
5120
5121
5122
5123
5124
5125
5126
5127
5128
5129
5130
5131
5132
5133
5134
5135
5136
5137
5138
5139
5140
5141
5142
5143
5144
5145
5146
5147
5148
5149
5150
5151
5152
5153
5154
5155
5156
5157
5158
5159
5160
5161
5162
5163
5164
5165
5166
5167
5168
5169
5170
5171
5172
5173
5174
5175
5176
5177
5178
5179
5180
5181
5182
5183
5184
5185
5186
5187
5188
5189
5190
5191
5192
5193
5194
5195
5196
5197
5198
5199
5200
5201
5202
5203
5204
5205
5206
5207
5208
5209
5210
5211
5212
5213
5214
5215
5216
5217
5218
5219
5220
5221
5222
5223
5224
5225
5226
5227
5228
5229
5230
5231
5232
5233
5234
5235
5236
5237
5238
5239
5240
5241
5242
5243
5244
5245
5246
5247
5248
5249
5250
5251
5252
5253
5254
5255
5256
5257
5258
5259
5260
5261
5262
5263
5264
5265
5266
5267
5268
5269
5270
5271
5272
5273
5274
5275
5276
5277
5278
5279
5280
5281
5282
5283
5284
5285
5286
5287
5288
5289
5290
5291
5292
5293
5294
5295
5296
5297
5298
5299
5300
5301
5302
5303
5304
5305
5306
5307
5308
5309
5310
5311
5312
5313
5314
5315
5316
5317
5318
5319
5320
5321
5322
5323
5324
5325
5326
5327
5328
5329
5330
5331
5332
5333
5334
5335
|
/* Transformation Utilities for Loop Vectorization.
Copyright (C) 2003, 2004, 2005, 2006, 2007 Free Software Foundation, Inc.
Contributed by Dorit Naishlos <dorit@il.ibm.com>
This file is part of GCC.
GCC is free software; you can redistribute it and/or modify it under
the terms of the GNU General Public License as published by the Free
Software Foundation; either version 2, or (at your option) any later
version.
GCC is distributed in the hope that it will be useful, but WITHOUT ANY
WARRANTY; without even the implied warranty of MERCHANTABILITY or
FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
for more details.
You should have received a copy of the GNU General Public License
along with GCC; see the file COPYING. If not, write to the Free
Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA
02110-1301, USA. */
#include "config.h"
#include "system.h"
#include "coretypes.h"
#include "tm.h"
#include "ggc.h"
#include "tree.h"
#include "target.h"
#include "rtl.h"
#include "basic-block.h"
#include "diagnostic.h"
#include "tree-flow.h"
#include "tree-dump.h"
#include "timevar.h"
#include "cfgloop.h"
#include "expr.h"
#include "optabs.h"
#include "params.h"
#include "recog.h"
#include "tree-data-ref.h"
#include "tree-chrec.h"
#include "tree-scalar-evolution.h"
#include "tree-vectorizer.h"
#include "langhooks.h"
#include "tree-pass.h"
#include "toplev.h"
#include "real.h"
/* Utility functions for the code transformation. */
static bool vect_transform_stmt (tree, block_stmt_iterator *, bool *);
static tree vect_create_destination_var (tree, tree);
static tree vect_create_data_ref_ptr
(tree, block_stmt_iterator *, tree, tree *, tree *, bool, tree);
static tree vect_create_addr_base_for_vector_ref (tree, tree *, tree);
static tree vect_setup_realignment (tree, block_stmt_iterator *, tree *);
static tree vect_get_new_vect_var (tree, enum vect_var_kind, const char *);
static tree vect_get_vec_def_for_operand (tree, tree, tree *);
static tree vect_init_vector (tree, tree, tree);
static void vect_finish_stmt_generation
(tree stmt, tree vec_stmt, block_stmt_iterator *bsi);
static bool vect_is_simple_cond (tree, loop_vec_info);
static void update_vuses_to_preheader (tree, struct loop*);
static void vect_create_epilog_for_reduction (tree, tree, enum tree_code, tree);
static tree get_initial_def_for_reduction (tree, tree, tree *);
/* Utility function dealing with loop peeling (not peeling itself). */
static void vect_generate_tmps_on_preheader
(loop_vec_info, tree *, tree *, tree *);
static tree vect_build_loop_niters (loop_vec_info);
static void vect_update_ivs_after_vectorizer (loop_vec_info, tree, edge);
static tree vect_gen_niters_for_prolog_loop (loop_vec_info, tree);
static void vect_update_init_of_dr (struct data_reference *, tree niters);
static void vect_update_inits_of_drs (loop_vec_info, tree);
static int vect_min_worthwhile_factor (enum tree_code);
/* Function vect_get_new_vect_var.
Returns a name for a new variable. The current naming scheme appends the
prefix "vect_" or "vect_p" (depending on the value of VAR_KIND) to
the name of vectorizer generated variables, and appends that to NAME if
provided. */
static tree
vect_get_new_vect_var (tree type, enum vect_var_kind var_kind, const char *name)
{
const char *prefix;
tree new_vect_var;
switch (var_kind)
{
case vect_simple_var:
prefix = "vect_";
break;
case vect_scalar_var:
prefix = "stmp_";
break;
case vect_pointer_var:
prefix = "vect_p";
break;
default:
gcc_unreachable ();
}
if (name)
new_vect_var = create_tmp_var (type, concat (prefix, name, NULL));
else
new_vect_var = create_tmp_var (type, prefix);
/* Mark vector typed variable as a gimple register variable. */
if (TREE_CODE (type) == VECTOR_TYPE)
DECL_GIMPLE_REG_P (new_vect_var) = true;
return new_vect_var;
}
/* Function vect_create_addr_base_for_vector_ref.
Create an expression that computes the address of the first memory location
that will be accessed for a data reference.
Input:
STMT: The statement containing the data reference.
NEW_STMT_LIST: Must be initialized to NULL_TREE or a statement list.
OFFSET: Optional. If supplied, it is be added to the initial address.
Output:
1. Return an SSA_NAME whose value is the address of the memory location of
the first vector of the data reference.
2. If new_stmt_list is not NULL_TREE after return then the caller must insert
these statement(s) which define the returned SSA_NAME.
FORNOW: We are only handling array accesses with step 1. */
static tree
vect_create_addr_base_for_vector_ref (tree stmt,
tree *new_stmt_list,
tree offset)
{
stmt_vec_info stmt_info = vinfo_for_stmt (stmt);
struct data_reference *dr = STMT_VINFO_DATA_REF (stmt_info);
tree data_ref_base = unshare_expr (DR_BASE_ADDRESS (dr));
tree base_name = build_fold_indirect_ref (data_ref_base);
tree vec_stmt;
tree addr_base, addr_expr;
tree dest, new_stmt;
tree base_offset = unshare_expr (DR_OFFSET (dr));
tree init = unshare_expr (DR_INIT (dr));
tree vect_ptr_type, addr_expr2;
/* Create base_offset */
base_offset = size_binop (PLUS_EXPR, base_offset, init);
dest = create_tmp_var (TREE_TYPE (base_offset), "base_off");
add_referenced_var (dest);
base_offset = force_gimple_operand (base_offset, &new_stmt, false, dest);
append_to_statement_list_force (new_stmt, new_stmt_list);
if (offset)
{
tree tmp = create_tmp_var (TREE_TYPE (base_offset), "offset");
tree step;
/* For interleaved access step we divide STEP by the size of the
interleaving group. */
if (DR_GROUP_SIZE (stmt_info))
step = fold_build2 (TRUNC_DIV_EXPR, TREE_TYPE (offset), DR_STEP (dr),
build_int_cst (TREE_TYPE (offset),
DR_GROUP_SIZE (stmt_info)));
else
step = DR_STEP (dr);
add_referenced_var (tmp);
offset = fold_build2 (MULT_EXPR, TREE_TYPE (offset), offset, step);
base_offset = fold_build2 (PLUS_EXPR, TREE_TYPE (base_offset),
base_offset, offset);
base_offset = force_gimple_operand (base_offset, &new_stmt, false, tmp);
append_to_statement_list_force (new_stmt, new_stmt_list);
}
/* base + base_offset */
addr_base = fold_build2 (PLUS_EXPR, TREE_TYPE (data_ref_base), data_ref_base,
base_offset);
vect_ptr_type = build_pointer_type (STMT_VINFO_VECTYPE (stmt_info));
/* addr_expr = addr_base */
addr_expr = vect_get_new_vect_var (vect_ptr_type, vect_pointer_var,
get_name (base_name));
add_referenced_var (addr_expr);
vec_stmt = fold_convert (vect_ptr_type, addr_base);
addr_expr2 = vect_get_new_vect_var (vect_ptr_type, vect_pointer_var,
get_name (base_name));
add_referenced_var (addr_expr2);
vec_stmt = force_gimple_operand (vec_stmt, &new_stmt, false, addr_expr2);
append_to_statement_list_force (new_stmt, new_stmt_list);
if (vect_print_dump_info (REPORT_DETAILS))
{
fprintf (vect_dump, "created ");
print_generic_expr (vect_dump, vec_stmt, TDF_SLIM);
}
return vec_stmt;
}
/* Function vect_create_data_ref_ptr.
Create a new pointer to vector type (vp), that points to the first location
accessed in the loop by STMT, along with the def-use update chain to
appropriately advance the pointer through the loop iterations. Also set
aliasing information for the pointer. This vector pointer is used by the
callers to this function to create a memory reference expression for vector
load/store access.
Input:
1. STMT: a stmt that references memory. Expected to be of the form
GIMPLE_MODIFY_STMT <name, data-ref> or
GIMPLE_MODIFY_STMT <data-ref, name>.
2. BSI: block_stmt_iterator where new stmts can be added.
3. OFFSET (optional): an offset to be added to the initial address accessed
by the data-ref in STMT.
4. ONLY_INIT: indicate if vp is to be updated in the loop, or remain
pointing to the initial address.
5. TYPE: if not NULL indicates the required type of the data-ref
Output:
1. Declare a new ptr to vector_type, and have it point to the base of the
data reference (initial addressed accessed by the data reference).
For example, for vector of type V8HI, the following code is generated:
v8hi *vp;
vp = (v8hi *)initial_address;
if OFFSET is not supplied:
initial_address = &a[init];
if OFFSET is supplied:
initial_address = &a[init + OFFSET];
Return the initial_address in INITIAL_ADDRESS.
2. If ONLY_INIT is true, just return the initial pointer. Otherwise, also
update the pointer in each iteration of the loop.
Return the increment stmt that updates the pointer in PTR_INCR.
3. Return the pointer. */
static tree
vect_create_data_ref_ptr (tree stmt,
block_stmt_iterator *bsi ATTRIBUTE_UNUSED,
tree offset, tree *initial_address, tree *ptr_incr,
bool only_init, tree type)
{
tree base_name;
stmt_vec_info stmt_info = vinfo_for_stmt (stmt);
loop_vec_info loop_vinfo = STMT_VINFO_LOOP_VINFO (stmt_info);
struct loop *loop = LOOP_VINFO_LOOP (loop_vinfo);
tree vectype = STMT_VINFO_VECTYPE (stmt_info);
tree vect_ptr_type;
tree vect_ptr;
tree tag;
tree new_temp;
tree vec_stmt;
tree new_stmt_list = NULL_TREE;
edge pe = loop_preheader_edge (loop);
basic_block new_bb;
tree vect_ptr_init;
struct data_reference *dr = STMT_VINFO_DATA_REF (stmt_info);
base_name = build_fold_indirect_ref (unshare_expr (DR_BASE_ADDRESS (dr)));
if (vect_print_dump_info (REPORT_DETAILS))
{
tree data_ref_base = base_name;
fprintf (vect_dump, "create vector-pointer variable to type: ");
print_generic_expr (vect_dump, vectype, TDF_SLIM);
if (TREE_CODE (data_ref_base) == VAR_DECL)
fprintf (vect_dump, " vectorizing a one dimensional array ref: ");
else if (TREE_CODE (data_ref_base) == ARRAY_REF)
fprintf (vect_dump, " vectorizing a multidimensional array ref: ");
else if (TREE_CODE (data_ref_base) == COMPONENT_REF)
fprintf (vect_dump, " vectorizing a record based array ref: ");
else if (TREE_CODE (data_ref_base) == SSA_NAME)
fprintf (vect_dump, " vectorizing a pointer ref: ");
print_generic_expr (vect_dump, base_name, TDF_SLIM);
}
/** (1) Create the new vector-pointer variable: **/
if (type)
vect_ptr_type = build_pointer_type (type);
else
vect_ptr_type = build_pointer_type (vectype);
vect_ptr = vect_get_new_vect_var (vect_ptr_type, vect_pointer_var,
get_name (base_name));
add_referenced_var (vect_ptr);
/** (2) Add aliasing information to the new vector-pointer:
(The points-to info (DR_PTR_INFO) may be defined later.) **/
tag = DR_MEMTAG (dr);
gcc_assert (tag);
/* If tag is a variable (and NOT_A_TAG) than a new symbol memory
tag must be created with tag added to its may alias list. */
if (!MTAG_P (tag))
new_type_alias (vect_ptr, tag, DR_REF (dr));
else
set_symbol_mem_tag (vect_ptr, tag);
var_ann (vect_ptr)->subvars = DR_SUBVARS (dr);
/** (3) Calculate the initial address the vector-pointer, and set
the vector-pointer to point to it before the loop: **/
/* Create: (&(base[init_val+offset]) in the loop preheader. */
new_temp = vect_create_addr_base_for_vector_ref (stmt, &new_stmt_list,
offset);
pe = loop_preheader_edge (loop);
new_bb = bsi_insert_on_edge_immediate (pe, new_stmt_list);
gcc_assert (!new_bb);
*initial_address = new_temp;
/* Create: p = (vectype *) initial_base */
vec_stmt = fold_convert (vect_ptr_type, new_temp);
vec_stmt = build_gimple_modify_stmt (vect_ptr, vec_stmt);
vect_ptr_init = make_ssa_name (vect_ptr, vec_stmt);
GIMPLE_STMT_OPERAND (vec_stmt, 0) = vect_ptr_init;
new_bb = bsi_insert_on_edge_immediate (pe, vec_stmt);
gcc_assert (!new_bb);
/** (4) Handle the updating of the vector-pointer inside the loop: **/
if (only_init) /* No update in loop is required. */
{
/* Copy the points-to information if it exists. */
if (DR_PTR_INFO (dr))
duplicate_ssa_name_ptr_info (vect_ptr_init, DR_PTR_INFO (dr));
return vect_ptr_init;
}
else
{
block_stmt_iterator incr_bsi;
bool insert_after;
tree indx_before_incr, indx_after_incr;
tree incr;
standard_iv_increment_position (loop, &incr_bsi, &insert_after);
create_iv (vect_ptr_init,
fold_convert (vect_ptr_type, TYPE_SIZE_UNIT (vectype)),
NULL_TREE, loop, &incr_bsi, insert_after,
&indx_before_incr, &indx_after_incr);
incr = bsi_stmt (incr_bsi);
set_stmt_info (stmt_ann (incr),
new_stmt_vec_info (incr, loop_vinfo));
/* Copy the points-to information if it exists. */
if (DR_PTR_INFO (dr))
{
duplicate_ssa_name_ptr_info (indx_before_incr, DR_PTR_INFO (dr));
duplicate_ssa_name_ptr_info (indx_after_incr, DR_PTR_INFO (dr));
}
merge_alias_info (vect_ptr_init, indx_before_incr);
merge_alias_info (vect_ptr_init, indx_after_incr);
if (ptr_incr)
*ptr_incr = incr;
return indx_before_incr;
}
}
/* Function bump_vector_ptr
Increment a pointer (to a vector type) by vector-size. Connect the new
increment stmt to the existing def-use update-chain of the pointer.
The pointer def-use update-chain before this function:
DATAREF_PTR = phi (p_0, p_2)
....
PTR_INCR: p_2 = DATAREF_PTR + step
The pointer def-use update-chain after this function:
DATAREF_PTR = phi (p_0, p_2)
....
NEW_DATAREF_PTR = DATAREF_PTR + vector_size
....
PTR_INCR: p_2 = NEW_DATAREF_PTR + step
Input:
DATAREF_PTR - ssa_name of a pointer (to vector type) that is being updated
in the loop.
PTR_INCR - the stmt that updates the pointer in each iteration of the loop.
The increment amount across iterations is also expected to be
vector_size.
BSI - location where the new update stmt is to be placed.
STMT - the original scalar memory-access stmt that is being vectorized.
Output: Return NEW_DATAREF_PTR as illustrated above.
*/
static tree
bump_vector_ptr (tree dataref_ptr, tree ptr_incr, block_stmt_iterator *bsi,
tree stmt)
{
stmt_vec_info stmt_info = vinfo_for_stmt (stmt);
struct data_reference *dr = STMT_VINFO_DATA_REF (stmt_info);
tree vectype = STMT_VINFO_VECTYPE (stmt_info);
tree vptr_type = TREE_TYPE (dataref_ptr);
tree ptr_var = SSA_NAME_VAR (dataref_ptr);
tree update = fold_convert (vptr_type, TYPE_SIZE_UNIT (vectype));
tree incr_stmt;
ssa_op_iter iter;
use_operand_p use_p;
tree new_dataref_ptr;
incr_stmt = build_gimple_modify_stmt (ptr_var,
build2 (PLUS_EXPR, vptr_type,
dataref_ptr, update));
new_dataref_ptr = make_ssa_name (ptr_var, incr_stmt);
GIMPLE_STMT_OPERAND (incr_stmt, 0) = new_dataref_ptr;
vect_finish_stmt_generation (stmt, incr_stmt, bsi);
/* Update the vector-pointer's cross-iteration increment. */
FOR_EACH_SSA_USE_OPERAND (use_p, ptr_incr, iter, SSA_OP_USE)
{
tree use = USE_FROM_PTR (use_p);
if (use == dataref_ptr)
SET_USE (use_p, new_dataref_ptr);
else
gcc_assert (tree_int_cst_compare (use, update) == 0);
}
/* Copy the points-to information if it exists. */
if (DR_PTR_INFO (dr))
duplicate_ssa_name_ptr_info (new_dataref_ptr, DR_PTR_INFO (dr));
merge_alias_info (new_dataref_ptr, dataref_ptr);
return new_dataref_ptr;
}
/* Function vect_create_destination_var.
Create a new temporary of type VECTYPE. */
static tree
vect_create_destination_var (tree scalar_dest, tree vectype)
{
tree vec_dest;
const char *new_name;
tree type;
enum vect_var_kind kind;
kind = vectype ? vect_simple_var : vect_scalar_var;
type = vectype ? vectype : TREE_TYPE (scalar_dest);
gcc_assert (TREE_CODE (scalar_dest) == SSA_NAME);
new_name = get_name (scalar_dest);
if (!new_name)
new_name = "var_";
vec_dest = vect_get_new_vect_var (type, kind, new_name);
add_referenced_var (vec_dest);
return vec_dest;
}
/* Function vect_init_vector.
Insert a new stmt (INIT_STMT) that initializes a new vector variable with
the vector elements of VECTOR_VAR. Return the DEF of INIT_STMT. It will be
used in the vectorization of STMT. */
static tree
vect_init_vector (tree stmt, tree vector_var, tree vector_type)
{
stmt_vec_info stmt_vinfo = vinfo_for_stmt (stmt);
loop_vec_info loop_vinfo = STMT_VINFO_LOOP_VINFO (stmt_vinfo);
struct loop *loop = LOOP_VINFO_LOOP (loop_vinfo);
tree new_var;
tree init_stmt;
tree vec_oprnd;
edge pe;
tree new_temp;
basic_block new_bb;
new_var = vect_get_new_vect_var (vector_type, vect_simple_var, "cst_");
add_referenced_var (new_var);
init_stmt = build_gimple_modify_stmt (new_var, vector_var);
new_temp = make_ssa_name (new_var, init_stmt);
GIMPLE_STMT_OPERAND (init_stmt, 0) = new_temp;
pe = loop_preheader_edge (loop);
new_bb = bsi_insert_on_edge_immediate (pe, init_stmt);
gcc_assert (!new_bb);
if (vect_print_dump_info (REPORT_DETAILS))
{
fprintf (vect_dump, "created new init_stmt: ");
print_generic_expr (vect_dump, init_stmt, TDF_SLIM);
}
vec_oprnd = GIMPLE_STMT_OPERAND (init_stmt, 0);
return vec_oprnd;
}
/* Function get_initial_def_for_induction
Input:
IV_PHI - the initial value of the induction variable
Output:
Return a vector variable, initialized with the first VF values of
the induction variable. E.g., for an iv with IV_PHI='X' and
evolution S, for a vector of 4 units, we want to return:
[X, X + S, X + 2*S, X + 3*S]. */
static tree
get_initial_def_for_induction (tree iv_phi)
{
stmt_vec_info stmt_vinfo = vinfo_for_stmt (iv_phi);
loop_vec_info loop_vinfo = STMT_VINFO_LOOP_VINFO (stmt_vinfo);
struct loop *loop = LOOP_VINFO_LOOP (loop_vinfo);
tree scalar_type = TREE_TYPE (PHI_RESULT_TREE (iv_phi));
tree vectype = get_vectype_for_scalar_type (scalar_type);
int nunits = TYPE_VECTOR_SUBPARTS (vectype);
edge pe = loop_preheader_edge (loop);
basic_block new_bb;
block_stmt_iterator bsi;
tree vec, vec_init, vec_step, t;
tree access_fn;
tree new_var;
tree new_name;
tree init_stmt;
tree induction_phi, induc_def, new_stmt, vec_def, vec_dest;
tree init_expr, step_expr;
int vf = LOOP_VINFO_VECT_FACTOR (loop_vinfo);
int i;
bool ok;
int ncopies = vf / nunits;
tree expr;
stmt_vec_info phi_info = vinfo_for_stmt (iv_phi);
tree stmts;
tree stmt = NULL_TREE;
block_stmt_iterator si;
basic_block bb = bb_for_stmt (iv_phi);
gcc_assert (phi_info);
gcc_assert (ncopies >= 1);
/* Find the first insertion point in the BB. */
si = bsi_after_labels (bb);
stmt = bsi_stmt (si);
access_fn = analyze_scalar_evolution (loop, PHI_RESULT (iv_phi));
gcc_assert (access_fn);
ok = vect_is_simple_iv_evolution (loop->num, access_fn,
&init_expr, &step_expr);
gcc_assert (ok);
/* Create the vector that holds the initial_value of the induction. */
new_var = vect_get_new_vect_var (scalar_type, vect_scalar_var, "var_");
add_referenced_var (new_var);
new_name = force_gimple_operand (init_expr, &stmts, false, new_var);
if (stmts)
{
new_bb = bsi_insert_on_edge_immediate (pe, stmts);
gcc_assert (!new_bb);
}
t = NULL_TREE;
t = tree_cons (NULL_TREE, new_name, t);
for (i = 1; i < nunits; i++)
{
tree tmp;
/* Create: new_name = new_name + step_expr */
tmp = fold_build2 (PLUS_EXPR, scalar_type, new_name, step_expr);
init_stmt = build_gimple_modify_stmt (new_var, tmp);
new_name = make_ssa_name (new_var, init_stmt);
GIMPLE_STMT_OPERAND (init_stmt, 0) = new_name;
new_bb = bsi_insert_on_edge_immediate (pe, init_stmt);
gcc_assert (!new_bb);
if (vect_print_dump_info (REPORT_DETAILS))
{
fprintf (vect_dump, "created new init_stmt: ");
print_generic_expr (vect_dump, init_stmt, TDF_SLIM);
}
t = tree_cons (NULL_TREE, new_name, t);
}
vec = build_constructor_from_list (vectype, nreverse (t));
vec_init = vect_init_vector (stmt, vec, vectype);
/* Create the vector that holds the step of the induction. */
expr = build_int_cst (scalar_type, vf);
new_name = fold_build2 (MULT_EXPR, scalar_type, expr, step_expr);
t = NULL_TREE;
for (i = 0; i < nunits; i++)
t = tree_cons (NULL_TREE, unshare_expr (new_name), t);
vec = build_constructor_from_list (vectype, t);
vec_step = vect_init_vector (stmt, vec, vectype);
/* Create the following def-use cycle:
loop prolog:
vec_init = [X, X+S, X+2*S, X+3*S]
vec_step = [VF*S, VF*S, VF*S, VF*S]
loop:
vec_iv = PHI <vec_init, vec_loop>
...
STMT
...
vec_loop = vec_iv + vec_step; */
/* Create the induction-phi that defines the induction-operand. */
vec_dest = vect_get_new_vect_var (vectype, vect_simple_var, "vec_iv_");
add_referenced_var (vec_dest);
induction_phi = create_phi_node (vec_dest, loop->header);
set_stmt_info (get_stmt_ann (induction_phi),
new_stmt_vec_info (induction_phi, loop_vinfo));
induc_def = PHI_RESULT (induction_phi);
/* Create the iv update inside the loop */
new_stmt = build_gimple_modify_stmt (NULL_TREE,
build2 (PLUS_EXPR, vectype,
induc_def, vec_step));
vec_def = make_ssa_name (vec_dest, new_stmt);
GIMPLE_STMT_OPERAND (new_stmt, 0) = vec_def;
bsi = bsi_for_stmt (stmt);
vect_finish_stmt_generation (stmt, new_stmt, &bsi);
/* Set the arguments of the phi node: */
add_phi_arg (induction_phi, vec_init, loop_preheader_edge (loop));
add_phi_arg (induction_phi, vec_def, loop_latch_edge (loop));
/* In case the vectorization factor (VF) is bigger than the number
of elements that we can fit in a vectype (nunits), we have to generate
more than one vector stmt - i.e - we need to "unroll" the
vector stmt by a factor VF/nunits. For more details see documentation
in vectorizable_operation. */
if (ncopies > 1)
{
stmt_vec_info prev_stmt_vinfo;
/* Create the vector that holds the step of the induction. */
expr = build_int_cst (scalar_type, nunits);
new_name = fold_build2 (MULT_EXPR, scalar_type, expr, step_expr);
t = NULL_TREE;
for (i = 0; i < nunits; i++)
t = tree_cons (NULL_TREE, unshare_expr (new_name), t);
vec = build_constructor_from_list (vectype, t);
vec_step = vect_init_vector (stmt, vec, vectype);
vec_def = induc_def;
prev_stmt_vinfo = vinfo_for_stmt (induction_phi);
for (i = 1; i < ncopies; i++)
{
tree tmp;
/* vec_i = vec_prev + vec_{step*nunits} */
tmp = build2 (PLUS_EXPR, vectype, vec_def, vec_step);
new_stmt = build_gimple_modify_stmt (NULL_TREE, tmp);
vec_def = make_ssa_name (vec_dest, new_stmt);
GIMPLE_STMT_OPERAND (new_stmt, 0) = vec_def;
bsi = bsi_for_stmt (stmt);
vect_finish_stmt_generation (stmt, new_stmt, &bsi);
STMT_VINFO_RELATED_STMT (prev_stmt_vinfo) = new_stmt;
prev_stmt_vinfo = vinfo_for_stmt (new_stmt);
}
}
if (vect_print_dump_info (REPORT_DETAILS))
{
fprintf (vect_dump, "transform induction: created def-use cycle:");
print_generic_expr (vect_dump, induction_phi, TDF_SLIM);
fprintf (vect_dump, "\n");
print_generic_expr (vect_dump, SSA_NAME_DEF_STMT (vec_def), TDF_SLIM);
}
STMT_VINFO_VEC_STMT (phi_info) = induction_phi;
return induc_def;
}
/* Function vect_get_vec_def_for_operand.
OP is an operand in STMT. This function returns a (vector) def that will be
used in the vectorized stmt for STMT.
In the case that OP is an SSA_NAME which is defined in the loop, then
STMT_VINFO_VEC_STMT of the defining stmt holds the relevant def.
In case OP is an invariant or constant, a new stmt that creates a vector def
needs to be introduced. */
static tree
vect_get_vec_def_for_operand (tree op, tree stmt, tree *scalar_def)
{
tree vec_oprnd;
tree vec_stmt;
tree def_stmt;
stmt_vec_info def_stmt_info = NULL;
stmt_vec_info stmt_vinfo = vinfo_for_stmt (stmt);
tree vectype = STMT_VINFO_VECTYPE (stmt_vinfo);
int nunits = TYPE_VECTOR_SUBPARTS (vectype);
loop_vec_info loop_vinfo = STMT_VINFO_LOOP_VINFO (stmt_vinfo);
struct loop *loop = LOOP_VINFO_LOOP (loop_vinfo);
tree vec_inv;
tree vec_cst;
tree t = NULL_TREE;
tree def;
int i;
enum vect_def_type dt;
bool is_simple_use;
tree vector_type;
if (vect_print_dump_info (REPORT_DETAILS))
{
fprintf (vect_dump, "vect_get_vec_def_for_operand: ");
print_generic_expr (vect_dump, op, TDF_SLIM);
}
is_simple_use = vect_is_simple_use (op, loop_vinfo, &def_stmt, &def, &dt);
gcc_assert (is_simple_use);
if (vect_print_dump_info (REPORT_DETAILS))
{
if (def)
{
fprintf (vect_dump, "def = ");
print_generic_expr (vect_dump, def, TDF_SLIM);
}
if (def_stmt)
{
fprintf (vect_dump, " def_stmt = ");
print_generic_expr (vect_dump, def_stmt, TDF_SLIM);
}
}
switch (dt)
{
/* Case 1: operand is a constant. */
case vect_constant_def:
{
if (scalar_def)
*scalar_def = op;
/* Create 'vect_cst_ = {cst,cst,...,cst}' */
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "Create vector_cst. nunits = %d", nunits);
for (i = nunits - 1; i >= 0; --i)
{
t = tree_cons (NULL_TREE, op, t);
}
vector_type = get_vectype_for_scalar_type (TREE_TYPE (op));
vec_cst = build_vector (vector_type, t);
return vect_init_vector (stmt, vec_cst, vector_type);
}
/* Case 2: operand is defined outside the loop - loop invariant. */
case vect_invariant_def:
{
if (scalar_def)
*scalar_def = def;
/* Create 'vec_inv = {inv,inv,..,inv}' */
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "Create vector_inv.");
for (i = nunits - 1; i >= 0; --i)
{
t = tree_cons (NULL_TREE, def, t);
}
/* FIXME: use build_constructor directly. */
vector_type = get_vectype_for_scalar_type (TREE_TYPE (def));
vec_inv = build_constructor_from_list (vector_type, t);
return vect_init_vector (stmt, vec_inv, vector_type);
}
/* Case 3: operand is defined inside the loop. */
case vect_loop_def:
{
if (scalar_def)
*scalar_def = def_stmt;
/* Get the def from the vectorized stmt. */
def_stmt_info = vinfo_for_stmt (def_stmt);
vec_stmt = STMT_VINFO_VEC_STMT (def_stmt_info);
gcc_assert (vec_stmt);
vec_oprnd = GIMPLE_STMT_OPERAND (vec_stmt, 0);
return vec_oprnd;
}
/* Case 4: operand is defined by a loop header phi - reduction */
case vect_reduction_def:
{
gcc_assert (TREE_CODE (def_stmt) == PHI_NODE);
/* Get the def before the loop */
op = PHI_ARG_DEF_FROM_EDGE (def_stmt, loop_preheader_edge (loop));
return get_initial_def_for_reduction (stmt, op, scalar_def);
}
/* Case 5: operand is defined by loop-header phi - induction. */
case vect_induction_def:
{
gcc_assert (TREE_CODE (def_stmt) == PHI_NODE);
/* Get the def before the loop */
return get_initial_def_for_induction (def_stmt);
}
default:
gcc_unreachable ();
}
}
/* Function vect_get_vec_def_for_stmt_copy
Return a vector-def for an operand. This function is used when the
vectorized stmt to be created (by the caller to this function) is a "copy"
created in case the vectorized result cannot fit in one vector, and several
copies of the vector-stmt are required. In this case the vector-def is
retrieved from the vector stmt recorded in the STMT_VINFO_RELATED_STMT field
of the stmt that defines VEC_OPRND.
DT is the type of the vector def VEC_OPRND.
Context:
In case the vectorization factor (VF) is bigger than the number
of elements that can fit in a vectype (nunits), we have to generate
more than one vector stmt to vectorize the scalar stmt. This situation
arises when there are multiple data-types operated upon in the loop; the
smallest data-type determines the VF, and as a result, when vectorizing
stmts operating on wider types we need to create 'VF/nunits' "copies" of the
vector stmt (each computing a vector of 'nunits' results, and together
computing 'VF' results in each iteration). This function is called when
vectorizing such a stmt (e.g. vectorizing S2 in the illustration below, in
which VF=16 and nunits=4, so the number of copies required is 4):
scalar stmt: vectorized into: STMT_VINFO_RELATED_STMT
S1: x = load VS1.0: vx.0 = memref0 VS1.1
VS1.1: vx.1 = memref1 VS1.2
VS1.2: vx.2 = memref2 VS1.3
VS1.3: vx.3 = memref3
S2: z = x + ... VSnew.0: vz0 = vx.0 + ... VSnew.1
VSnew.1: vz1 = vx.1 + ... VSnew.2
VSnew.2: vz2 = vx.2 + ... VSnew.3
VSnew.3: vz3 = vx.3 + ...
The vectorization of S1 is explained in vectorizable_load.
The vectorization of S2:
To create the first vector-stmt out of the 4 copies - VSnew.0 -
the function 'vect_get_vec_def_for_operand' is called to
get the relevant vector-def for each operand of S2. For operand x it
returns the vector-def 'vx.0'.
To create the remaining copies of the vector-stmt (VSnew.j), this
function is called to get the relevant vector-def for each operand. It is
obtained from the respective VS1.j stmt, which is recorded in the
STMT_VINFO_RELATED_STMT field of the stmt that defines VEC_OPRND.
For example, to obtain the vector-def 'vx.1' in order to create the
vector stmt 'VSnew.1', this function is called with VEC_OPRND='vx.0'.
Given 'vx0' we obtain the stmt that defines it ('VS1.0'); from the
STMT_VINFO_RELATED_STMT field of 'VS1.0' we obtain the next copy - 'VS1.1',
and return its def ('vx.1').
Overall, to create the above sequence this function will be called 3 times:
vx.1 = vect_get_vec_def_for_stmt_copy (dt, vx.0);
vx.2 = vect_get_vec_def_for_stmt_copy (dt, vx.1);
vx.3 = vect_get_vec_def_for_stmt_copy (dt, vx.2); */
static tree
vect_get_vec_def_for_stmt_copy (enum vect_def_type dt, tree vec_oprnd)
{
tree vec_stmt_for_operand;
stmt_vec_info def_stmt_info;
/* Do nothing; can reuse same def. */
if (dt == vect_invariant_def || dt == vect_constant_def )
return vec_oprnd;
vec_stmt_for_operand = SSA_NAME_DEF_STMT (vec_oprnd);
def_stmt_info = vinfo_for_stmt (vec_stmt_for_operand);
gcc_assert (def_stmt_info);
vec_stmt_for_operand = STMT_VINFO_RELATED_STMT (def_stmt_info);
gcc_assert (vec_stmt_for_operand);
vec_oprnd = GIMPLE_STMT_OPERAND (vec_stmt_for_operand, 0);
return vec_oprnd;
}
/* Function vect_finish_stmt_generation.
Insert a new stmt. */
static void
vect_finish_stmt_generation (tree stmt, tree vec_stmt,
block_stmt_iterator *bsi)
{
stmt_vec_info stmt_info = vinfo_for_stmt (stmt);
loop_vec_info loop_vinfo = STMT_VINFO_LOOP_VINFO (stmt_info);
bsi_insert_before (bsi, vec_stmt, BSI_SAME_STMT);
set_stmt_info (get_stmt_ann (vec_stmt),
new_stmt_vec_info (vec_stmt, loop_vinfo));
if (vect_print_dump_info (REPORT_DETAILS))
{
fprintf (vect_dump, "add new stmt: ");
print_generic_expr (vect_dump, vec_stmt, TDF_SLIM);
}
/* Make sure bsi points to the stmt that is being vectorized. */
gcc_assert (stmt == bsi_stmt (*bsi));
#ifdef USE_MAPPED_LOCATION
SET_EXPR_LOCATION (vec_stmt, EXPR_LOCATION (stmt));
#else
SET_EXPR_LOCUS (vec_stmt, EXPR_LOCUS (stmt));
#endif
}
/* Function get_initial_def_for_reduction
Input:
STMT - a stmt that performs a reduction operation in the loop.
INIT_VAL - the initial value of the reduction variable
Output:
ADJUSTMENT_DEF - a tree that holds a value to be added to the final result
of the reduction (used for adjusting the epilog - see below).
Return a vector variable, initialized according to the operation that STMT
performs. This vector will be used as the initial value of the
vector of partial results.
Option1 (adjust in epilog): Initialize the vector as follows:
add: [0,0,...,0,0]
mult: [1,1,...,1,1]
min/max: [init_val,init_val,..,init_val,init_val]
bit and/or: [init_val,init_val,..,init_val,init_val]
and when necessary (e.g. add/mult case) let the caller know
that it needs to adjust the result by init_val.
Option2: Initialize the vector as follows:
add: [0,0,...,0,init_val]
mult: [1,1,...,1,init_val]
min/max: [init_val,init_val,...,init_val]
bit and/or: [init_val,init_val,...,init_val]
and no adjustments are needed.
For example, for the following code:
s = init_val;
for (i=0;i<n;i++)
s = s + a[i];
STMT is 's = s + a[i]', and the reduction variable is 's'.
For a vector of 4 units, we want to return either [0,0,0,init_val],
or [0,0,0,0] and let the caller know that it needs to adjust
the result at the end by 'init_val'.
FORNOW, we are using the 'adjust in epilog' scheme, because this way the
initialization vector is simpler (same element in all entries).
A cost model should help decide between these two schemes. */
static tree
get_initial_def_for_reduction (tree stmt, tree init_val, tree *adjustment_def)
{
stmt_vec_info stmt_vinfo = vinfo_for_stmt (stmt);
tree vectype = STMT_VINFO_VECTYPE (stmt_vinfo);
int nunits = TYPE_VECTOR_SUBPARTS (vectype);
enum tree_code code = TREE_CODE (GIMPLE_STMT_OPERAND (stmt, 1));
tree type = TREE_TYPE (init_val);
tree vecdef;
tree def_for_init;
tree init_def;
tree t = NULL_TREE;
int i;
tree vector_type;
gcc_assert (INTEGRAL_TYPE_P (type) || SCALAR_FLOAT_TYPE_P (type));
vecdef = vect_get_vec_def_for_operand (init_val, stmt, NULL);
switch (code)
{
case WIDEN_SUM_EXPR:
case DOT_PROD_EXPR:
case PLUS_EXPR:
*adjustment_def = init_val;
/* Create a vector of zeros for init_def. */
if (INTEGRAL_TYPE_P (type))
def_for_init = build_int_cst (type, 0);
else
def_for_init = build_real (type, dconst0);
for (i = nunits - 1; i >= 0; --i)
t = tree_cons (NULL_TREE, def_for_init, t);
vector_type = get_vectype_for_scalar_type (TREE_TYPE (def_for_init));
init_def = build_vector (vector_type, t);
break;
case MIN_EXPR:
case MAX_EXPR:
*adjustment_def = NULL_TREE;
init_def = vecdef;
break;
default:
gcc_unreachable ();
}
return init_def;
}
/* Function vect_create_epilog_for_reduction
Create code at the loop-epilog to finalize the result of a reduction
computation.
VECT_DEF is a vector of partial results.
REDUC_CODE is the tree-code for the epilog reduction.
STMT is the scalar reduction stmt that is being vectorized.
REDUCTION_PHI is the phi-node that carries the reduction computation.
This function:
1. Creates the reduction def-use cycle: sets the arguments for
REDUCTION_PHI:
The loop-entry argument is the vectorized initial-value of the reduction.
The loop-latch argument is VECT_DEF - the vector of partial sums.
2. "Reduces" the vector of partial results VECT_DEF into a single result,
by applying the operation specified by REDUC_CODE if available, or by
other means (whole-vector shifts or a scalar loop).
The function also creates a new phi node at the loop exit to preserve
loop-closed form, as illustrated below.
The flow at the entry to this function:
loop:
vec_def = phi <null, null> # REDUCTION_PHI
VECT_DEF = vector_stmt # vectorized form of STMT
s_loop = scalar_stmt # (scalar) STMT
loop_exit:
s_out0 = phi <s_loop> # (scalar) EXIT_PHI
use <s_out0>
use <s_out0>
The above is transformed by this function into:
loop:
vec_def = phi <vec_init, VECT_DEF> # REDUCTION_PHI
VECT_DEF = vector_stmt # vectorized form of STMT
s_loop = scalar_stmt # (scalar) STMT
loop_exit:
s_out0 = phi <s_loop> # (scalar) EXIT_PHI
v_out1 = phi <VECT_DEF> # NEW_EXIT_PHI
v_out2 = reduce <v_out1>
s_out3 = extract_field <v_out2, 0>
s_out4 = adjust_result <s_out3>
use <s_out4>
use <s_out4>
*/
static void
vect_create_epilog_for_reduction (tree vect_def, tree stmt,
enum tree_code reduc_code, tree reduction_phi)
{
stmt_vec_info stmt_info = vinfo_for_stmt (stmt);
tree vectype;
enum machine_mode mode;
loop_vec_info loop_vinfo = STMT_VINFO_LOOP_VINFO (stmt_info);
struct loop *loop = LOOP_VINFO_LOOP (loop_vinfo);
basic_block exit_bb;
tree scalar_dest;
tree scalar_type;
tree new_phi;
block_stmt_iterator exit_bsi;
tree vec_dest;
tree new_temp;
tree new_name;
tree epilog_stmt;
tree new_scalar_dest, exit_phi;
tree bitsize, bitpos, bytesize;
enum tree_code code = TREE_CODE (GIMPLE_STMT_OPERAND (stmt, 1));
tree scalar_initial_def;
tree vec_initial_def;
tree orig_name;
imm_use_iterator imm_iter;
use_operand_p use_p;
bool extract_scalar_result;
tree reduction_op;
tree orig_stmt;
tree use_stmt;
tree operation = GIMPLE_STMT_OPERAND (stmt, 1);
int op_type;
op_type = TREE_OPERAND_LENGTH (operation);
reduction_op = TREE_OPERAND (operation, op_type-1);
vectype = get_vectype_for_scalar_type (TREE_TYPE (reduction_op));
mode = TYPE_MODE (vectype);
/*** 1. Create the reduction def-use cycle ***/
/* 1.1 set the loop-entry arg of the reduction-phi: */
/* For the case of reduction, vect_get_vec_def_for_operand returns
the scalar def before the loop, that defines the initial value
of the reduction variable. */
vec_initial_def = vect_get_vec_def_for_operand (reduction_op, stmt,
&scalar_initial_def);
add_phi_arg (reduction_phi, vec_initial_def, loop_preheader_edge (loop));
/* 1.2 set the loop-latch arg for the reduction-phi: */
add_phi_arg (reduction_phi, vect_def, loop_latch_edge (loop));
if (vect_print_dump_info (REPORT_DETAILS))
{
fprintf (vect_dump, "transform reduction: created def-use cycle:");
print_generic_expr (vect_dump, reduction_phi, TDF_SLIM);
fprintf (vect_dump, "\n");
print_generic_expr (vect_dump, SSA_NAME_DEF_STMT (vect_def), TDF_SLIM);
}
/*** 2. Create epilog code
The reduction epilog code operates across the elements of the vector
of partial results computed by the vectorized loop.
The reduction epilog code consists of:
step 1: compute the scalar result in a vector (v_out2)
step 2: extract the scalar result (s_out3) from the vector (v_out2)
step 3: adjust the scalar result (s_out3) if needed.
Step 1 can be accomplished using one the following three schemes:
(scheme 1) using reduc_code, if available.
(scheme 2) using whole-vector shifts, if available.
(scheme 3) using a scalar loop. In this case steps 1+2 above are
combined.
The overall epilog code looks like this:
s_out0 = phi <s_loop> # original EXIT_PHI
v_out1 = phi <VECT_DEF> # NEW_EXIT_PHI
v_out2 = reduce <v_out1> # step 1
s_out3 = extract_field <v_out2, 0> # step 2
s_out4 = adjust_result <s_out3> # step 3
(step 3 is optional, and step2 1 and 2 may be combined).
Lastly, the uses of s_out0 are replaced by s_out4.
***/
/* 2.1 Create new loop-exit-phi to preserve loop-closed form:
v_out1 = phi <v_loop> */
exit_bb = single_exit (loop)->dest;
new_phi = create_phi_node (SSA_NAME_VAR (vect_def), exit_bb);
SET_PHI_ARG_DEF (new_phi, single_exit (loop)->dest_idx, vect_def);
exit_bsi = bsi_after_labels (exit_bb);
/* 2.2 Get the relevant tree-code to use in the epilog for schemes 2,3
(i.e. when reduc_code is not available) and in the final adjustment
code (if needed). Also get the original scalar reduction variable as
defined in the loop. In case STMT is a "pattern-stmt" (i.e. - it
represents a reduction pattern), the tree-code and scalar-def are
taken from the original stmt that the pattern-stmt (STMT) replaces.
Otherwise (it is a regular reduction) - the tree-code and scalar-def
are taken from STMT. */
orig_stmt = STMT_VINFO_RELATED_STMT (stmt_info);
if (!orig_stmt)
{
/* Regular reduction */
orig_stmt = stmt;
}
else
{
/* Reduction pattern */
stmt_vec_info stmt_vinfo = vinfo_for_stmt (orig_stmt);
gcc_assert (STMT_VINFO_IN_PATTERN_P (stmt_vinfo));
gcc_assert (STMT_VINFO_RELATED_STMT (stmt_vinfo) == stmt);
}
code = TREE_CODE (GIMPLE_STMT_OPERAND (orig_stmt, 1));
scalar_dest = GIMPLE_STMT_OPERAND (orig_stmt, 0);
scalar_type = TREE_TYPE (scalar_dest);
new_scalar_dest = vect_create_destination_var (scalar_dest, NULL);
bitsize = TYPE_SIZE (scalar_type);
bytesize = TYPE_SIZE_UNIT (scalar_type);
/* 2.3 Create the reduction code, using one of the three schemes described
above. */
if (reduc_code < NUM_TREE_CODES)
{
tree tmp;
/*** Case 1: Create:
v_out2 = reduc_expr <v_out1> */
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "Reduce using direct vector reduction.");
vec_dest = vect_create_destination_var (scalar_dest, vectype);
tmp = build1 (reduc_code, vectype, PHI_RESULT (new_phi));
epilog_stmt = build_gimple_modify_stmt (vec_dest, tmp);
new_temp = make_ssa_name (vec_dest, epilog_stmt);
GIMPLE_STMT_OPERAND (epilog_stmt, 0) = new_temp;
bsi_insert_before (&exit_bsi, epilog_stmt, BSI_SAME_STMT);
extract_scalar_result = true;
}
else
{
enum tree_code shift_code = 0;
bool have_whole_vector_shift = true;
int bit_offset;
int element_bitsize = tree_low_cst (bitsize, 1);
int vec_size_in_bits = tree_low_cst (TYPE_SIZE (vectype), 1);
tree vec_temp;
if (vec_shr_optab->handlers[mode].insn_code != CODE_FOR_nothing)
shift_code = VEC_RSHIFT_EXPR;
else
have_whole_vector_shift = false;
/* Regardless of whether we have a whole vector shift, if we're
emulating the operation via tree-vect-generic, we don't want
to use it. Only the first round of the reduction is likely
to still be profitable via emulation. */
/* ??? It might be better to emit a reduction tree code here, so that
tree-vect-generic can expand the first round via bit tricks. */
if (!VECTOR_MODE_P (mode))
have_whole_vector_shift = false;
else
{
optab optab = optab_for_tree_code (code, vectype);
if (optab->handlers[mode].insn_code == CODE_FOR_nothing)
have_whole_vector_shift = false;
}
if (have_whole_vector_shift)
{
/*** Case 2: Create:
for (offset = VS/2; offset >= element_size; offset/=2)
{
Create: va' = vec_shift <va, offset>
Create: va = vop <va, va'>
} */
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "Reduce using vector shifts");
vec_dest = vect_create_destination_var (scalar_dest, vectype);
new_temp = PHI_RESULT (new_phi);
for (bit_offset = vec_size_in_bits/2;
bit_offset >= element_bitsize;
bit_offset /= 2)
{
tree bitpos = size_int (bit_offset);
tree tmp = build2 (shift_code, vectype, new_temp, bitpos);
epilog_stmt = build_gimple_modify_stmt (vec_dest, tmp);
new_name = make_ssa_name (vec_dest, epilog_stmt);
GIMPLE_STMT_OPERAND (epilog_stmt, 0) = new_name;
bsi_insert_before (&exit_bsi, epilog_stmt, BSI_SAME_STMT);
tmp = build2 (code, vectype, new_name, new_temp);
epilog_stmt = build_gimple_modify_stmt (vec_dest, tmp);
new_temp = make_ssa_name (vec_dest, epilog_stmt);
GIMPLE_STMT_OPERAND (epilog_stmt, 0) = new_temp;
bsi_insert_before (&exit_bsi, epilog_stmt, BSI_SAME_STMT);
}
extract_scalar_result = true;
}
else
{
tree rhs;
/*** Case 3: Create:
s = extract_field <v_out2, 0>
for (offset = element_size;
offset < vector_size;
offset += element_size;)
{
Create: s' = extract_field <v_out2, offset>
Create: s = op <s, s'>
} */
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "Reduce using scalar code. ");
vec_temp = PHI_RESULT (new_phi);
vec_size_in_bits = tree_low_cst (TYPE_SIZE (vectype), 1);
rhs = build3 (BIT_FIELD_REF, scalar_type, vec_temp, bitsize,
bitsize_zero_node);
BIT_FIELD_REF_UNSIGNED (rhs) = TYPE_UNSIGNED (scalar_type);
epilog_stmt = build_gimple_modify_stmt (new_scalar_dest, rhs);
new_temp = make_ssa_name (new_scalar_dest, epilog_stmt);
GIMPLE_STMT_OPERAND (epilog_stmt, 0) = new_temp;
bsi_insert_before (&exit_bsi, epilog_stmt, BSI_SAME_STMT);
for (bit_offset = element_bitsize;
bit_offset < vec_size_in_bits;
bit_offset += element_bitsize)
{
tree tmp;
tree bitpos = bitsize_int (bit_offset);
tree rhs = build3 (BIT_FIELD_REF, scalar_type, vec_temp, bitsize,
bitpos);
BIT_FIELD_REF_UNSIGNED (rhs) = TYPE_UNSIGNED (scalar_type);
epilog_stmt = build_gimple_modify_stmt (new_scalar_dest, rhs);
new_name = make_ssa_name (new_scalar_dest, epilog_stmt);
GIMPLE_STMT_OPERAND (epilog_stmt, 0) = new_name;
bsi_insert_before (&exit_bsi, epilog_stmt, BSI_SAME_STMT);
tmp = build2 (code, scalar_type, new_name, new_temp);
epilog_stmt = build_gimple_modify_stmt (new_scalar_dest, tmp);
new_temp = make_ssa_name (new_scalar_dest, epilog_stmt);
GIMPLE_STMT_OPERAND (epilog_stmt, 0) = new_temp;
bsi_insert_before (&exit_bsi, epilog_stmt, BSI_SAME_STMT);
}
extract_scalar_result = false;
}
}
/* 2.4 Extract the final scalar result. Create:
s_out3 = extract_field <v_out2, bitpos> */
if (extract_scalar_result)
{
tree rhs;
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "extract scalar result");
if (BYTES_BIG_ENDIAN)
bitpos = size_binop (MULT_EXPR,
bitsize_int (TYPE_VECTOR_SUBPARTS (vectype) - 1),
TYPE_SIZE (scalar_type));
else
bitpos = bitsize_zero_node;
rhs = build3 (BIT_FIELD_REF, scalar_type, new_temp, bitsize, bitpos);
BIT_FIELD_REF_UNSIGNED (rhs) = TYPE_UNSIGNED (scalar_type);
epilog_stmt = build_gimple_modify_stmt (new_scalar_dest, rhs);
new_temp = make_ssa_name (new_scalar_dest, epilog_stmt);
GIMPLE_STMT_OPERAND (epilog_stmt, 0) = new_temp;
bsi_insert_before (&exit_bsi, epilog_stmt, BSI_SAME_STMT);
}
/* 2.4 Adjust the final result by the initial value of the reduction
variable. (When such adjustment is not needed, then
'scalar_initial_def' is zero).
Create:
s_out4 = scalar_expr <s_out3, scalar_initial_def> */
if (scalar_initial_def)
{
tree tmp = build2 (code, scalar_type, new_temp, scalar_initial_def);
epilog_stmt = build_gimple_modify_stmt (new_scalar_dest, tmp);
new_temp = make_ssa_name (new_scalar_dest, epilog_stmt);
GIMPLE_STMT_OPERAND (epilog_stmt, 0) = new_temp;
bsi_insert_before (&exit_bsi, epilog_stmt, BSI_SAME_STMT);
}
/* 2.6 Replace uses of s_out0 with uses of s_out3 */
/* Find the loop-closed-use at the loop exit of the original scalar result.
(The reduction result is expected to have two immediate uses - one at the
latch block, and one at the loop exit). */
exit_phi = NULL;
FOR_EACH_IMM_USE_FAST (use_p, imm_iter, scalar_dest)
{
if (!flow_bb_inside_loop_p (loop, bb_for_stmt (USE_STMT (use_p))))
{
exit_phi = USE_STMT (use_p);
break;
}
}
/* We expect to have found an exit_phi because of loop-closed-ssa form. */
gcc_assert (exit_phi);
/* Replace the uses: */
orig_name = PHI_RESULT (exit_phi);
FOR_EACH_IMM_USE_STMT (use_stmt, imm_iter, orig_name)
FOR_EACH_IMM_USE_ON_STMT (use_p, imm_iter)
SET_USE (use_p, new_temp);
}
/* Function vectorizable_reduction.
Check if STMT performs a reduction operation that can be vectorized.
If VEC_STMT is also passed, vectorize the STMT: create a vectorized
stmt to replace it, put it in VEC_STMT, and insert it at BSI.
Return FALSE if not a vectorizable STMT, TRUE otherwise.
This function also handles reduction idioms (patterns) that have been
recognized in advance during vect_pattern_recog. In this case, STMT may be
of this form:
X = pattern_expr (arg0, arg1, ..., X)
and it's STMT_VINFO_RELATED_STMT points to the last stmt in the original
sequence that had been detected and replaced by the pattern-stmt (STMT).
In some cases of reduction patterns, the type of the reduction variable X is
different than the type of the other arguments of STMT.
In such cases, the vectype that is used when transforming STMT into a vector
stmt is different than the vectype that is used to determine the
vectorization factor, because it consists of a different number of elements
than the actual number of elements that are being operated upon in parallel.
For example, consider an accumulation of shorts into an int accumulator.
On some targets it's possible to vectorize this pattern operating on 8
shorts at a time (hence, the vectype for purposes of determining the
vectorization factor should be V8HI); on the other hand, the vectype that
is used to create the vector form is actually V4SI (the type of the result).
Upon entry to this function, STMT_VINFO_VECTYPE records the vectype that
indicates what is the actual level of parallelism (V8HI in the example), so
that the right vectorization factor would be derived. This vectype
corresponds to the type of arguments to the reduction stmt, and should *NOT*
be used to create the vectorized stmt. The right vectype for the vectorized
stmt is obtained from the type of the result X:
get_vectype_for_scalar_type (TREE_TYPE (X))
This means that, contrary to "regular" reductions (or "regular" stmts in
general), the following equation:
STMT_VINFO_VECTYPE == get_vectype_for_scalar_type (TREE_TYPE (X))
does *NOT* necessarily hold for reduction patterns. */
bool
vectorizable_reduction (tree stmt, block_stmt_iterator *bsi, tree *vec_stmt)
{
tree vec_dest;
tree scalar_dest;
tree op;
tree loop_vec_def0 = NULL_TREE, loop_vec_def1 = NULL_TREE;
stmt_vec_info stmt_info = vinfo_for_stmt (stmt);
tree vectype = STMT_VINFO_VECTYPE (stmt_info);
loop_vec_info loop_vinfo = STMT_VINFO_LOOP_VINFO (stmt_info);
struct loop *loop = LOOP_VINFO_LOOP (loop_vinfo);
tree operation;
enum tree_code code, orig_code, epilog_reduc_code = 0;
enum machine_mode vec_mode;
int op_type;
optab optab, reduc_optab;
tree new_temp = NULL_TREE;
tree def, def_stmt;
enum vect_def_type dt;
tree new_phi;
tree scalar_type;
bool is_simple_use;
tree orig_stmt;
stmt_vec_info orig_stmt_info;
tree expr = NULL_TREE;
int i;
int nunits = TYPE_VECTOR_SUBPARTS (vectype);
int ncopies = LOOP_VINFO_VECT_FACTOR (loop_vinfo) / nunits;
stmt_vec_info prev_stmt_info;
tree reduc_def;
tree new_stmt = NULL_TREE;
int j;
gcc_assert (ncopies >= 1);
/* 1. Is vectorizable reduction? */
/* Not supportable if the reduction variable is used in the loop. */
if (STMT_VINFO_RELEVANT_P (stmt_info))
return false;
if (!STMT_VINFO_LIVE_P (stmt_info))
return false;
/* Make sure it was already recognized as a reduction computation. */
if (STMT_VINFO_DEF_TYPE (stmt_info) != vect_reduction_def)
return false;
/* 2. Has this been recognized as a reduction pattern?
Check if STMT represents a pattern that has been recognized
in earlier analysis stages. For stmts that represent a pattern,
the STMT_VINFO_RELATED_STMT field records the last stmt in
the original sequence that constitutes the pattern. */
orig_stmt = STMT_VINFO_RELATED_STMT (stmt_info);
if (orig_stmt)
{
orig_stmt_info = vinfo_for_stmt (orig_stmt);
gcc_assert (STMT_VINFO_RELATED_STMT (orig_stmt_info) == stmt);
gcc_assert (STMT_VINFO_IN_PATTERN_P (orig_stmt_info));
gcc_assert (!STMT_VINFO_IN_PATTERN_P (stmt_info));
}
/* 3. Check the operands of the operation. The first operands are defined
inside the loop body. The last operand is the reduction variable,
which is defined by the loop-header-phi. */
gcc_assert (TREE_CODE (stmt) == GIMPLE_MODIFY_STMT);
operation = GIMPLE_STMT_OPERAND (stmt, 1);
code = TREE_CODE (operation);
op_type = TREE_OPERAND_LENGTH (operation);
if (op_type != binary_op && op_type != ternary_op)
return false;
scalar_dest = GIMPLE_STMT_OPERAND (stmt, 0);
scalar_type = TREE_TYPE (scalar_dest);
/* All uses but the last are expected to be defined in the loop.
The last use is the reduction variable. */
for (i = 0; i < op_type-1; i++)
{
op = TREE_OPERAND (operation, i);
is_simple_use = vect_is_simple_use (op, loop_vinfo, &def_stmt, &def, &dt);
gcc_assert (is_simple_use);
if (dt != vect_loop_def
&& dt != vect_invariant_def
&& dt != vect_constant_def
&& dt != vect_induction_def)
return false;
}
op = TREE_OPERAND (operation, i);
is_simple_use = vect_is_simple_use (op, loop_vinfo, &def_stmt, &def, &dt);
gcc_assert (is_simple_use);
gcc_assert (dt == vect_reduction_def);
gcc_assert (TREE_CODE (def_stmt) == PHI_NODE);
if (orig_stmt)
gcc_assert (orig_stmt == vect_is_simple_reduction (loop, def_stmt));
else
gcc_assert (stmt == vect_is_simple_reduction (loop, def_stmt));
if (STMT_VINFO_LIVE_P (vinfo_for_stmt (def_stmt)))
return false;
/* 4. Supportable by target? */
/* 4.1. check support for the operation in the loop */
optab = optab_for_tree_code (code, vectype);
if (!optab)
{
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "no optab.");
return false;
}
vec_mode = TYPE_MODE (vectype);
if (optab->handlers[(int) vec_mode].insn_code == CODE_FOR_nothing)
{
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "op not supported by target.");
if (GET_MODE_SIZE (vec_mode) != UNITS_PER_WORD
|| LOOP_VINFO_VECT_FACTOR (loop_vinfo)
< vect_min_worthwhile_factor (code))
return false;
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "proceeding using word mode.");
}
/* Worthwhile without SIMD support? */
if (!VECTOR_MODE_P (TYPE_MODE (vectype))
&& LOOP_VINFO_VECT_FACTOR (loop_vinfo)
< vect_min_worthwhile_factor (code))
{
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "not worthwhile without SIMD support.");
return false;
}
/* 4.2. Check support for the epilog operation.
If STMT represents a reduction pattern, then the type of the
reduction variable may be different than the type of the rest
of the arguments. For example, consider the case of accumulation
of shorts into an int accumulator; The original code:
S1: int_a = (int) short_a;
orig_stmt-> S2: int_acc = plus <int_a ,int_acc>;
was replaced with:
STMT: int_acc = widen_sum <short_a, int_acc>
This means that:
1. The tree-code that is used to create the vector operation in the
epilog code (that reduces the partial results) is not the
tree-code of STMT, but is rather the tree-code of the original
stmt from the pattern that STMT is replacing. I.e, in the example
above we want to use 'widen_sum' in the loop, but 'plus' in the
epilog.
2. The type (mode) we use to check available target support
for the vector operation to be created in the *epilog*, is
determined by the type of the reduction variable (in the example
above we'd check this: plus_optab[vect_int_mode]).
However the type (mode) we use to check available target support
for the vector operation to be created *inside the loop*, is
determined by the type of the other arguments to STMT (in the
example we'd check this: widen_sum_optab[vect_short_mode]).
This is contrary to "regular" reductions, in which the types of all
the arguments are the same as the type of the reduction variable.
For "regular" reductions we can therefore use the same vector type
(and also the same tree-code) when generating the epilog code and
when generating the code inside the loop. */
if (orig_stmt)
{
/* This is a reduction pattern: get the vectype from the type of the
reduction variable, and get the tree-code from orig_stmt. */
orig_code = TREE_CODE (GIMPLE_STMT_OPERAND (orig_stmt, 1));
vectype = get_vectype_for_scalar_type (TREE_TYPE (def));
vec_mode = TYPE_MODE (vectype);
}
else
{
/* Regular reduction: use the same vectype and tree-code as used for
the vector code inside the loop can be used for the epilog code. */
orig_code = code;
}
if (!reduction_code_for_scalar_code (orig_code, &epilog_reduc_code))
return false;
reduc_optab = optab_for_tree_code (epilog_reduc_code, vectype);
if (!reduc_optab)
{
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "no optab for reduction.");
epilog_reduc_code = NUM_TREE_CODES;
}
if (reduc_optab->handlers[(int) vec_mode].insn_code == CODE_FOR_nothing)
{
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "reduc op not supported by target.");
epilog_reduc_code = NUM_TREE_CODES;
}
if (!vec_stmt) /* transformation not required. */
{
STMT_VINFO_TYPE (stmt_info) = reduc_vec_info_type;
return true;
}
/** Transform. **/
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "transform reduction.");
/* Create the destination vector */
vec_dest = vect_create_destination_var (scalar_dest, vectype);
/* Create the reduction-phi that defines the reduction-operand. */
new_phi = create_phi_node (vec_dest, loop->header);
/* In case the vectorization factor (VF) is bigger than the number
of elements that we can fit in a vectype (nunits), we have to generate
more than one vector stmt - i.e - we need to "unroll" the
vector stmt by a factor VF/nunits. For more details see documentation
in vectorizable_operation. */
prev_stmt_info = NULL;
for (j = 0; j < ncopies; j++)
{
/* Handle uses. */
if (j == 0)
{
op = TREE_OPERAND (operation, 0);
loop_vec_def0 = vect_get_vec_def_for_operand (op, stmt, NULL);
if (op_type == ternary_op)
{
op = TREE_OPERAND (operation, 1);
loop_vec_def1 = vect_get_vec_def_for_operand (op, stmt, NULL);
}
/* Get the vector def for the reduction variable from the phi node */
reduc_def = PHI_RESULT (new_phi);
}
else
{
enum vect_def_type dt = vect_unknown_def_type; /* Dummy */
loop_vec_def0 = vect_get_vec_def_for_stmt_copy (dt, loop_vec_def0);
if (op_type == ternary_op)
loop_vec_def1 = vect_get_vec_def_for_stmt_copy (dt, loop_vec_def1);
/* Get the vector def for the reduction variable from the vectorized
reduction operation generated in the previous iteration (j-1) */
reduc_def = GIMPLE_STMT_OPERAND (new_stmt ,0);
}
/* Arguments are ready. create the new vector stmt. */
if (op_type == binary_op)
expr = build2 (code, vectype, loop_vec_def0, reduc_def);
else
expr = build3 (code, vectype, loop_vec_def0, loop_vec_def1,
reduc_def);
new_stmt = build_gimple_modify_stmt (vec_dest, expr);
new_temp = make_ssa_name (vec_dest, new_stmt);
GIMPLE_STMT_OPERAND (new_stmt, 0) = new_temp;
vect_finish_stmt_generation (stmt, new_stmt, bsi);
if (j == 0)
STMT_VINFO_VEC_STMT (stmt_info) = *vec_stmt = new_stmt;
else
STMT_VINFO_RELATED_STMT (prev_stmt_info) = new_stmt;
prev_stmt_info = vinfo_for_stmt (new_stmt);
}
/* Finalize the reduction-phi (set it's arguments) and create the
epilog reduction code. */
vect_create_epilog_for_reduction (new_temp, stmt, epilog_reduc_code, new_phi);
return true;
}
/* Checks if CALL can be vectorized in type VECTYPE. Returns
a function declaration if the target has a vectorized version
of the function, or NULL_TREE if the function cannot be vectorized. */
tree
vectorizable_function (tree call, tree vectype_out, tree vectype_in)
{
tree fndecl = get_callee_fndecl (call);
enum built_in_function code;
/* We only handle functions that do not read or clobber memory -- i.e.
const or novops ones. */
if (!(call_expr_flags (call) & (ECF_CONST | ECF_NOVOPS)))
return NULL_TREE;
if (!fndecl
|| TREE_CODE (fndecl) != FUNCTION_DECL
|| !DECL_BUILT_IN (fndecl))
return NULL_TREE;
code = DECL_FUNCTION_CODE (fndecl);
return targetm.vectorize.builtin_vectorized_function (code, vectype_out,
vectype_in);
}
/* Function vectorizable_call.
Check if STMT performs a function call that can be vectorized.
If VEC_STMT is also passed, vectorize the STMT: create a vectorized
stmt to replace it, put it in VEC_STMT, and insert it at BSI.
Return FALSE if not a vectorizable STMT, TRUE otherwise. */
bool
vectorizable_call (tree stmt, block_stmt_iterator *bsi, tree *vec_stmt)
{
tree vec_dest;
tree scalar_dest;
tree operation;
tree op, type;
stmt_vec_info stmt_info = vinfo_for_stmt (stmt), prev_stmt_info;
tree vectype_out, vectype_in;
loop_vec_info loop_vinfo = STMT_VINFO_LOOP_VINFO (stmt_info);
tree fndecl, rhs, new_temp, def, def_stmt, rhs_type, lhs_type;
enum vect_def_type dt[2];
int ncopies, j, nargs;
call_expr_arg_iterator iter;
if (!STMT_VINFO_RELEVANT_P (stmt_info))
return false;
if (STMT_VINFO_DEF_TYPE (stmt_info) != vect_loop_def)
return false;
/* FORNOW: not yet supported. */
if (STMT_VINFO_LIVE_P (stmt_info))
{
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "value used after loop.");
return false;
}
/* Is STMT a vectorizable call? */
if (TREE_CODE (stmt) != GIMPLE_MODIFY_STMT)
return false;
if (TREE_CODE (GIMPLE_STMT_OPERAND (stmt, 0)) != SSA_NAME)
return false;
operation = GIMPLE_STMT_OPERAND (stmt, 1);
if (TREE_CODE (operation) != CALL_EXPR)
return false;
/* Process function arguments. */
rhs_type = NULL_TREE;
nargs = 0;
FOR_EACH_CALL_EXPR_ARG (op, iter, operation)
{
++nargs;
/* Bail out if the function has more than two arguments, we
do not have interesting builtin functions to vectorize with
more than two arguments. */
if (nargs > 2)
return false;
/* We can only handle calls with arguments of the same type. */
if (rhs_type
&& rhs_type != TREE_TYPE (op))
{
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "argument types differ.");
return false;
}
rhs_type = TREE_TYPE (op);
if (!vect_is_simple_use (op, loop_vinfo, &def_stmt, &def, &dt[nargs-1]))
{
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "use not simple.");
return false;
}
}
/* No arguments is also not good. */
if (nargs == 0)
return false;
vectype_in = get_vectype_for_scalar_type (rhs_type);
lhs_type = TREE_TYPE (GIMPLE_STMT_OPERAND (stmt, 0));
vectype_out = get_vectype_for_scalar_type (lhs_type);
/* Only handle the case of vectors with the same number of elements.
FIXME: We need a way to handle for example the SSE2 cvtpd2dq
instruction which converts V2DFmode to V4SImode but only
using the lower half of the V4SImode result. */
if (TYPE_VECTOR_SUBPARTS (vectype_in) != TYPE_VECTOR_SUBPARTS (vectype_out))
return false;
/* For now, we only vectorize functions if a target specific builtin
is available. TODO -- in some cases, it might be profitable to
insert the calls for pieces of the vector, in order to be able
to vectorize other operations in the loop. */
fndecl = vectorizable_function (operation, vectype_out, vectype_in);
if (fndecl == NULL_TREE)
{
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "function is not vectorizable.");
return false;
}
gcc_assert (ZERO_SSA_OPERANDS (stmt, SSA_OP_ALL_VIRTUALS));
if (!vec_stmt) /* transformation not required. */
{
STMT_VINFO_TYPE (stmt_info) = call_vec_info_type;
return true;
}
/** Transform. **/
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "transform operation.");
ncopies = (LOOP_VINFO_VECT_FACTOR (loop_vinfo)
/ TYPE_VECTOR_SUBPARTS (vectype_out));
gcc_assert (ncopies >= 1);
/* Handle def. */
scalar_dest = GIMPLE_STMT_OPERAND (stmt, 0);
vec_dest = vect_create_destination_var (scalar_dest, vectype_out);
prev_stmt_info = NULL;
for (j = 0; j < ncopies; ++j)
{
tree new_stmt, vargs;
tree vec_oprnd[2];
int n;
/* Build argument list for the vectorized call. */
/* FIXME: Rewrite this so that it doesn't construct a temporary
list. */
vargs = NULL_TREE;
n = -1;
FOR_EACH_CALL_EXPR_ARG (op, iter, operation)
{
++n;
if (j == 0)
vec_oprnd[n] = vect_get_vec_def_for_operand (op, stmt, NULL);
else
vec_oprnd[n] = vect_get_vec_def_for_stmt_copy (dt[n], vec_oprnd[n]);
vargs = tree_cons (NULL_TREE, vec_oprnd[n], vargs);
}
vargs = nreverse (vargs);
rhs = build_function_call_expr (fndecl, vargs);
new_stmt = build_gimple_modify_stmt (vec_dest, rhs);
new_temp = make_ssa_name (vec_dest, new_stmt);
GIMPLE_STMT_OPERAND (new_stmt, 0) = new_temp;
vect_finish_stmt_generation (stmt, new_stmt, bsi);
if (j == 0)
STMT_VINFO_VEC_STMT (stmt_info) = *vec_stmt = new_stmt;
else
STMT_VINFO_RELATED_STMT (prev_stmt_info) = new_stmt;
prev_stmt_info = vinfo_for_stmt (new_stmt);
}
/* The call in STMT might prevent it from being removed in dce. We however
cannot remove it here, due to the way the ssa name it defines is mapped
to the new definition. So just replace rhs of the statement with something
harmless. */
type = TREE_TYPE (scalar_dest);
GIMPLE_STMT_OPERAND (stmt, 1) = fold_convert (type, integer_zero_node);
return true;
}
/* Function vectorizable_conversion.
Check if STMT performs a conversion operation, that can be vectorized.
If VEC_STMT is also passed, vectorize the STMT: create a vectorized
stmt to replace it, put it in VEC_STMT, and insert it at BSI.
Return FALSE if not a vectorizable STMT, TRUE otherwise. */
bool
vectorizable_conversion (tree stmt, block_stmt_iterator * bsi,
tree * vec_stmt)
{
tree vec_dest;
tree scalar_dest;
tree operation;
tree op0;
tree vec_oprnd0 = NULL_TREE;
stmt_vec_info stmt_info = vinfo_for_stmt (stmt);
loop_vec_info loop_vinfo = STMT_VINFO_LOOP_VINFO (stmt_info);
enum tree_code code;
tree new_temp;
tree def, def_stmt;
enum vect_def_type dt0;
tree new_stmt;
int nunits_in;
int nunits_out;
int ncopies, j;
tree vectype_out, vectype_in;
tree rhs_type, lhs_type;
tree builtin_decl;
stmt_vec_info prev_stmt_info;
/* Is STMT a vectorizable conversion? */
if (!STMT_VINFO_RELEVANT_P (stmt_info))
return false;
if (STMT_VINFO_DEF_TYPE (stmt_info) != vect_loop_def)
return false;
if (STMT_VINFO_LIVE_P (stmt_info))
{
/* FORNOW: not yet supported. */
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "value used after loop.");
return false;
}
if (TREE_CODE (stmt) != GIMPLE_MODIFY_STMT)
return false;
if (TREE_CODE (GIMPLE_STMT_OPERAND (stmt, 0)) != SSA_NAME)
return false;
operation = GIMPLE_STMT_OPERAND (stmt, 1);
code = TREE_CODE (operation);
if (code != FIX_TRUNC_EXPR && code != FLOAT_EXPR)
return false;
/* Check types of lhs and rhs */
op0 = TREE_OPERAND (operation, 0);
rhs_type = TREE_TYPE (op0);
vectype_in = get_vectype_for_scalar_type (rhs_type);
nunits_in = TYPE_VECTOR_SUBPARTS (vectype_in);
scalar_dest = GIMPLE_STMT_OPERAND (stmt, 0);
lhs_type = TREE_TYPE (scalar_dest);
vectype_out = get_vectype_for_scalar_type (lhs_type);
gcc_assert (STMT_VINFO_VECTYPE (stmt_info) == vectype_out);
nunits_out = TYPE_VECTOR_SUBPARTS (vectype_out);
/* FORNOW: need to extend to support short<->float conversions as well. */
if (nunits_out != nunits_in)
return false;
/* Bail out if the types are both integral or non-integral */
if ((INTEGRAL_TYPE_P (rhs_type) && INTEGRAL_TYPE_P (lhs_type))
|| (!INTEGRAL_TYPE_P (rhs_type) && !INTEGRAL_TYPE_P (lhs_type)))
return false;
/* Sanity check: make sure that at least one copy of the vectorized stmt
needs to be generated. */
ncopies = LOOP_VINFO_VECT_FACTOR (loop_vinfo) / nunits_in;
gcc_assert (ncopies >= 1);
if (!vect_is_simple_use (op0, loop_vinfo, &def_stmt, &def, &dt0))
{
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "use not simple.");
return false;
}
/* Supportable by target? */
if (!targetm.vectorize.builtin_conversion (code, vectype_in))
{
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "op not supported by target.");
return false;
}
if (!vec_stmt) /* transformation not required. */
{
STMT_VINFO_TYPE (stmt_info) = type_conversion_vec_info_type;
return true;
}
/** Transform. **/
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "transform conversion.");
/* Handle def. */
vec_dest = vect_create_destination_var (scalar_dest, vectype_out);
prev_stmt_info = NULL;
for (j = 0; j < ncopies; j++)
{
tree sym;
ssa_op_iter iter;
if (j == 0)
vec_oprnd0 = vect_get_vec_def_for_operand (op0, stmt, NULL);
else
vec_oprnd0 = vect_get_vec_def_for_stmt_copy (dt0, vec_oprnd0);
builtin_decl =
targetm.vectorize.builtin_conversion (code, vectype_in);
new_stmt = build_call_expr (builtin_decl, 1, vec_oprnd0);
/* Arguments are ready. create the new vector stmt. */
new_stmt = build_gimple_modify_stmt (vec_dest, new_stmt);
new_temp = make_ssa_name (vec_dest, new_stmt);
GIMPLE_STMT_OPERAND (new_stmt, 0) = new_temp;
vect_finish_stmt_generation (stmt, new_stmt, bsi);
FOR_EACH_SSA_TREE_OPERAND (sym, new_stmt, iter, SSA_OP_ALL_VIRTUALS)
{
if (TREE_CODE (sym) == SSA_NAME)
sym = SSA_NAME_VAR (sym);
mark_sym_for_renaming (sym);
}
if (j == 0)
STMT_VINFO_VEC_STMT (stmt_info) = *vec_stmt = new_stmt;
else
STMT_VINFO_RELATED_STMT (prev_stmt_info) = new_stmt;
prev_stmt_info = vinfo_for_stmt (new_stmt);
}
return true;
}
/* Function vectorizable_assignment.
Check if STMT performs an assignment (copy) that can be vectorized.
If VEC_STMT is also passed, vectorize the STMT: create a vectorized
stmt to replace it, put it in VEC_STMT, and insert it at BSI.
Return FALSE if not a vectorizable STMT, TRUE otherwise. */
bool
vectorizable_assignment (tree stmt, block_stmt_iterator *bsi, tree *vec_stmt)
{
tree vec_dest;
tree scalar_dest;
tree op;
tree vec_oprnd;
stmt_vec_info stmt_info = vinfo_for_stmt (stmt);
tree vectype = STMT_VINFO_VECTYPE (stmt_info);
loop_vec_info loop_vinfo = STMT_VINFO_LOOP_VINFO (stmt_info);
tree new_temp;
tree def, def_stmt;
enum vect_def_type dt;
int nunits = TYPE_VECTOR_SUBPARTS (vectype);
int ncopies = LOOP_VINFO_VECT_FACTOR (loop_vinfo) / nunits;
gcc_assert (ncopies >= 1);
if (ncopies > 1)
return false; /* FORNOW */
if (!STMT_VINFO_RELEVANT_P (stmt_info))
return false;
if (STMT_VINFO_DEF_TYPE (stmt_info) != vect_loop_def)
return false;
/* FORNOW: not yet supported. */
if (STMT_VINFO_LIVE_P (stmt_info))
{
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "value used after loop.");
return false;
}
/* Is vectorizable assignment? */
if (TREE_CODE (stmt) != GIMPLE_MODIFY_STMT)
return false;
scalar_dest = GIMPLE_STMT_OPERAND (stmt, 0);
if (TREE_CODE (scalar_dest) != SSA_NAME)
return false;
op = GIMPLE_STMT_OPERAND (stmt, 1);
if (!vect_is_simple_use (op, loop_vinfo, &def_stmt, &def, &dt))
{
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "use not simple.");
return false;
}
if (!vec_stmt) /* transformation not required. */
{
STMT_VINFO_TYPE (stmt_info) = assignment_vec_info_type;
return true;
}
/** Transform. **/
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "transform assignment.");
/* Handle def. */
vec_dest = vect_create_destination_var (scalar_dest, vectype);
/* Handle use. */
op = GIMPLE_STMT_OPERAND (stmt, 1);
vec_oprnd = vect_get_vec_def_for_operand (op, stmt, NULL);
/* Arguments are ready. create the new vector stmt. */
*vec_stmt = build_gimple_modify_stmt (vec_dest, vec_oprnd);
new_temp = make_ssa_name (vec_dest, *vec_stmt);
GIMPLE_STMT_OPERAND (*vec_stmt, 0) = new_temp;
vect_finish_stmt_generation (stmt, *vec_stmt, bsi);
return true;
}
/* Function vect_min_worthwhile_factor.
For a loop where we could vectorize the operation indicated by CODE,
return the minimum vectorization factor that makes it worthwhile
to use generic vectors. */
static int
vect_min_worthwhile_factor (enum tree_code code)
{
switch (code)
{
case PLUS_EXPR:
case MINUS_EXPR:
case NEGATE_EXPR:
return 4;
case BIT_AND_EXPR:
case BIT_IOR_EXPR:
case BIT_XOR_EXPR:
case BIT_NOT_EXPR:
return 2;
default:
return INT_MAX;
}
}
/* Function vectorizable_induction
Check if PHI performs an induction computation that can be vectorized.
If VEC_STMT is also passed, vectorize the induction PHI: create a vectorized
phi to replace it, put it in VEC_STMT, and add it to the same basic block.
Return FALSE if not a vectorizable STMT, TRUE otherwise. */
bool
vectorizable_induction (tree phi, block_stmt_iterator *bsi ATTRIBUTE_UNUSED,
tree *vec_stmt)
{
stmt_vec_info stmt_info = vinfo_for_stmt (phi);
tree vectype = STMT_VINFO_VECTYPE (stmt_info);
loop_vec_info loop_vinfo = STMT_VINFO_LOOP_VINFO (stmt_info);
int nunits = TYPE_VECTOR_SUBPARTS (vectype);
int ncopies = LOOP_VINFO_VECT_FACTOR (loop_vinfo) / nunits;
tree vec_def;
gcc_assert (ncopies >= 1);
if (!STMT_VINFO_RELEVANT_P (stmt_info))
return false;
gcc_assert (STMT_VINFO_DEF_TYPE (stmt_info) == vect_induction_def);
if (STMT_VINFO_LIVE_P (stmt_info))
{
/* FORNOW: not yet supported. */
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "value used after loop.");
return false;
}
if (TREE_CODE (phi) != PHI_NODE)
return false;
if (!vec_stmt) /* transformation not required. */
{
STMT_VINFO_TYPE (stmt_info) = induc_vec_info_type;
return true;
}
/** Transform. **/
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "transform induction phi.");
vec_def = get_initial_def_for_induction (phi);
*vec_stmt = SSA_NAME_DEF_STMT (vec_def);
return true;
}
/* Function vectorizable_operation.
Check if STMT performs a binary or unary operation that can be vectorized.
If VEC_STMT is also passed, vectorize the STMT: create a vectorized
stmt to replace it, put it in VEC_STMT, and insert it at BSI.
Return FALSE if not a vectorizable STMT, TRUE otherwise. */
bool
vectorizable_operation (tree stmt, block_stmt_iterator *bsi, tree *vec_stmt)
{
tree vec_dest;
tree scalar_dest;
tree operation;
tree op0, op1 = NULL;
tree vec_oprnd0 = NULL_TREE, vec_oprnd1 = NULL_TREE;
stmt_vec_info stmt_info = vinfo_for_stmt (stmt);
tree vectype = STMT_VINFO_VECTYPE (stmt_info);
loop_vec_info loop_vinfo = STMT_VINFO_LOOP_VINFO (stmt_info);
enum tree_code code;
enum machine_mode vec_mode;
tree new_temp;
int op_type;
optab optab;
int icode;
enum machine_mode optab_op2_mode;
tree def, def_stmt;
enum vect_def_type dt0, dt1;
tree new_stmt;
stmt_vec_info prev_stmt_info;
int nunits_in = TYPE_VECTOR_SUBPARTS (vectype);
int nunits_out;
tree vectype_out;
int ncopies = LOOP_VINFO_VECT_FACTOR (loop_vinfo) / nunits_in;
int j;
gcc_assert (ncopies >= 1);
if (!STMT_VINFO_RELEVANT_P (stmt_info))
return false;
if (STMT_VINFO_DEF_TYPE (stmt_info) != vect_loop_def)
return false;
/* FORNOW: not yet supported. */
if (STMT_VINFO_LIVE_P (stmt_info))
{
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "value used after loop.");
return false;
}
/* Is STMT a vectorizable binary/unary operation? */
if (TREE_CODE (stmt) != GIMPLE_MODIFY_STMT)
return false;
if (TREE_CODE (GIMPLE_STMT_OPERAND (stmt, 0)) != SSA_NAME)
return false;
scalar_dest = GIMPLE_STMT_OPERAND (stmt, 0);
vectype_out = get_vectype_for_scalar_type (TREE_TYPE (scalar_dest));
nunits_out = TYPE_VECTOR_SUBPARTS (vectype_out);
if (nunits_out != nunits_in)
return false;
operation = GIMPLE_STMT_OPERAND (stmt, 1);
code = TREE_CODE (operation);
optab = optab_for_tree_code (code, vectype);
/* Support only unary or binary operations. */
op_type = TREE_OPERAND_LENGTH (operation);
if (op_type != unary_op && op_type != binary_op)
{
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "num. args = %d (not unary/binary op).", op_type);
return false;
}
op0 = TREE_OPERAND (operation, 0);
if (!vect_is_simple_use (op0, loop_vinfo, &def_stmt, &def, &dt0))
{
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "use not simple.");
return false;
}
if (op_type == binary_op)
{
op1 = TREE_OPERAND (operation, 1);
if (!vect_is_simple_use (op1, loop_vinfo, &def_stmt, &def, &dt1))
{
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "use not simple.");
return false;
}
}
/* Supportable by target? */
if (!optab)
{
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "no optab.");
return false;
}
vec_mode = TYPE_MODE (vectype);
icode = (int) optab->handlers[(int) vec_mode].insn_code;
if (icode == CODE_FOR_nothing)
{
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "op not supported by target.");
if (GET_MODE_SIZE (vec_mode) != UNITS_PER_WORD
|| LOOP_VINFO_VECT_FACTOR (loop_vinfo)
< vect_min_worthwhile_factor (code))
return false;
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "proceeding using word mode.");
}
/* Worthwhile without SIMD support? */
if (!VECTOR_MODE_P (TYPE_MODE (vectype))
&& LOOP_VINFO_VECT_FACTOR (loop_vinfo)
< vect_min_worthwhile_factor (code))
{
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "not worthwhile without SIMD support.");
return false;
}
if (code == LSHIFT_EXPR || code == RSHIFT_EXPR)
{
/* FORNOW: not yet supported. */
if (!VECTOR_MODE_P (vec_mode))
return false;
/* Invariant argument is needed for a vector shift
by a scalar shift operand. */
optab_op2_mode = insn_data[icode].operand[2].mode;
if (! (VECTOR_MODE_P (optab_op2_mode)
|| dt1 == vect_constant_def
|| dt1 == vect_invariant_def))
{
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "operand mode requires invariant argument.");
return false;
}
}
if (!vec_stmt) /* transformation not required. */
{
STMT_VINFO_TYPE (stmt_info) = op_vec_info_type;
return true;
}
/** Transform. **/
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "transform binary/unary operation.");
/* Handle def. */
vec_dest = vect_create_destination_var (scalar_dest, vectype);
/* In case the vectorization factor (VF) is bigger than the number
of elements that we can fit in a vectype (nunits), we have to generate
more than one vector stmt - i.e - we need to "unroll" the
vector stmt by a factor VF/nunits. In doing so, we record a pointer
from one copy of the vector stmt to the next, in the field
STMT_VINFO_RELATED_STMT. This is necessary in order to allow following
stages to find the correct vector defs to be used when vectorizing
stmts that use the defs of the current stmt. The example below illustrates
the vectorization process when VF=16 and nunits=4 (i.e - we need to create
4 vectorized stmts):
before vectorization:
RELATED_STMT VEC_STMT
S1: x = memref - -
S2: z = x + 1 - -
step 1: vectorize stmt S1 (done in vectorizable_load. See more details
there):
RELATED_STMT VEC_STMT
VS1_0: vx0 = memref0 VS1_1 -
VS1_1: vx1 = memref1 VS1_2 -
VS1_2: vx2 = memref2 VS1_3 -
VS1_3: vx3 = memref3 - -
S1: x = load - VS1_0
S2: z = x + 1 - -
step2: vectorize stmt S2 (done here):
To vectorize stmt S2 we first need to find the relevant vector
def for the first operand 'x'. This is, as usual, obtained from
the vector stmt recorded in the STMT_VINFO_VEC_STMT of the stmt
that defines 'x' (S1). This way we find the stmt VS1_0, and the
relevant vector def 'vx0'. Having found 'vx0' we can generate
the vector stmt VS2_0, and as usual, record it in the
STMT_VINFO_VEC_STMT of stmt S2.
When creating the second copy (VS2_1), we obtain the relevant vector
def from the vector stmt recorded in the STMT_VINFO_RELATED_STMT of
stmt VS1_0. This way we find the stmt VS1_1 and the relevant
vector def 'vx1'. Using 'vx1' we create stmt VS2_1 and record a
pointer to it in the STMT_VINFO_RELATED_STMT of the vector stmt VS2_0.
Similarly when creating stmts VS2_2 and VS2_3. This is the resulting
chain of stmts and pointers:
RELATED_STMT VEC_STMT
VS1_0: vx0 = memref0 VS1_1 -
VS1_1: vx1 = memref1 VS1_2 -
VS1_2: vx2 = memref2 VS1_3 -
VS1_3: vx3 = memref3 - -
S1: x = load - VS1_0
VS2_0: vz0 = vx0 + v1 VS2_1 -
VS2_1: vz1 = vx1 + v1 VS2_2 -
VS2_2: vz2 = vx2 + v1 VS2_3 -
VS2_3: vz3 = vx3 + v1 - -
S2: z = x + 1 - VS2_0 */
prev_stmt_info = NULL;
for (j = 0; j < ncopies; j++)
{
/* Handle uses. */
if (j == 0)
{
vec_oprnd0 = vect_get_vec_def_for_operand (op0, stmt, NULL);
if (op_type == binary_op)
{
if (code == LSHIFT_EXPR || code == RSHIFT_EXPR)
{
/* Vector shl and shr insn patterns can be defined with
scalar operand 2 (shift operand). In this case, use
constant or loop invariant op1 directly, without
extending it to vector mode first. */
optab_op2_mode = insn_data[icode].operand[2].mode;
if (!VECTOR_MODE_P (optab_op2_mode))
{
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "operand 1 using scalar mode.");
vec_oprnd1 = op1;
}
}
if (!vec_oprnd1)
vec_oprnd1 = vect_get_vec_def_for_operand (op1, stmt, NULL);
}
}
else
{
vec_oprnd0 = vect_get_vec_def_for_stmt_copy (dt0, vec_oprnd0);
if (op_type == binary_op)
vec_oprnd1 = vect_get_vec_def_for_stmt_copy (dt1, vec_oprnd1);
}
/* Arguments are ready. create the new vector stmt. */
if (op_type == binary_op)
new_stmt = build_gimple_modify_stmt (vec_dest,
build2 (code, vectype, vec_oprnd0, vec_oprnd1));
else
new_stmt = build_gimple_modify_stmt (vec_dest,
build1 (code, vectype, vec_oprnd0));
new_temp = make_ssa_name (vec_dest, new_stmt);
GIMPLE_STMT_OPERAND (new_stmt, 0) = new_temp;
vect_finish_stmt_generation (stmt, new_stmt, bsi);
if (j == 0)
STMT_VINFO_VEC_STMT (stmt_info) = *vec_stmt = new_stmt;
else
STMT_VINFO_RELATED_STMT (prev_stmt_info) = new_stmt;
prev_stmt_info = vinfo_for_stmt (new_stmt);
}
return true;
}
/* Function vectorizable_type_demotion
Check if STMT performs a binary or unary operation that involves
type demotion, and if it can be vectorized.
If VEC_STMT is also passed, vectorize the STMT: create a vectorized
stmt to replace it, put it in VEC_STMT, and insert it at BSI.
Return FALSE if not a vectorizable STMT, TRUE otherwise. */
bool
vectorizable_type_demotion (tree stmt, block_stmt_iterator *bsi,
tree *vec_stmt)
{
tree vec_dest;
tree scalar_dest;
tree operation;
tree op0;
tree vec_oprnd0=NULL, vec_oprnd1=NULL;
stmt_vec_info stmt_info = vinfo_for_stmt (stmt);
loop_vec_info loop_vinfo = STMT_VINFO_LOOP_VINFO (stmt_info);
enum tree_code code;
tree new_temp;
tree def, def_stmt;
enum vect_def_type dt0;
tree new_stmt;
stmt_vec_info prev_stmt_info;
int nunits_in;
int nunits_out;
tree vectype_out;
int ncopies;
int j;
tree expr;
tree vectype_in;
tree scalar_type;
optab optab;
enum machine_mode vec_mode;
if (!STMT_VINFO_RELEVANT_P (stmt_info))
return false;
if (STMT_VINFO_DEF_TYPE (stmt_info) != vect_loop_def)
return false;
/* FORNOW: not yet supported. */
if (STMT_VINFO_LIVE_P (stmt_info))
{
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "value used after loop.");
return false;
}
/* Is STMT a vectorizable type-demotion operation? */
if (TREE_CODE (stmt) != GIMPLE_MODIFY_STMT)
return false;
if (TREE_CODE (GIMPLE_STMT_OPERAND (stmt, 0)) != SSA_NAME)
return false;
operation = GIMPLE_STMT_OPERAND (stmt, 1);
code = TREE_CODE (operation);
if (code != NOP_EXPR && code != CONVERT_EXPR)
return false;
op0 = TREE_OPERAND (operation, 0);
vectype_in = get_vectype_for_scalar_type (TREE_TYPE (op0));
nunits_in = TYPE_VECTOR_SUBPARTS (vectype_in);
scalar_dest = GIMPLE_STMT_OPERAND (stmt, 0);
scalar_type = TREE_TYPE (scalar_dest);
vectype_out = get_vectype_for_scalar_type (scalar_type);
nunits_out = TYPE_VECTOR_SUBPARTS (vectype_out);
if (nunits_in != nunits_out / 2) /* FORNOW */
return false;
ncopies = LOOP_VINFO_VECT_FACTOR (loop_vinfo) / nunits_out;
gcc_assert (ncopies >= 1);
if (! ((INTEGRAL_TYPE_P (TREE_TYPE (scalar_dest))
&& INTEGRAL_TYPE_P (TREE_TYPE (op0)))
|| (SCALAR_FLOAT_TYPE_P (TREE_TYPE (scalar_dest))
&& SCALAR_FLOAT_TYPE_P (TREE_TYPE (op0))
&& (code == NOP_EXPR || code == CONVERT_EXPR))))
return false;
/* Check the operands of the operation. */
if (!vect_is_simple_use (op0, loop_vinfo, &def_stmt, &def, &dt0))
{
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "use not simple.");
return false;
}
/* Supportable by target? */
code = VEC_PACK_TRUNC_EXPR;
optab = optab_for_tree_code (code, vectype_in);
if (!optab)
return false;
vec_mode = TYPE_MODE (vectype_in);
if (optab->handlers[(int) vec_mode].insn_code == CODE_FOR_nothing)
return false;
STMT_VINFO_VECTYPE (stmt_info) = vectype_in;
if (!vec_stmt) /* transformation not required. */
{
STMT_VINFO_TYPE (stmt_info) = type_demotion_vec_info_type;
return true;
}
/** Transform. **/
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "transform type demotion operation. ncopies = %d.",
ncopies);
/* Handle def. */
vec_dest = vect_create_destination_var (scalar_dest, vectype_out);
/* In case the vectorization factor (VF) is bigger than the number
of elements that we can fit in a vectype (nunits), we have to generate
more than one vector stmt - i.e - we need to "unroll" the
vector stmt by a factor VF/nunits. */
prev_stmt_info = NULL;
for (j = 0; j < ncopies; j++)
{
/* Handle uses. */
if (j == 0)
{
vec_oprnd0 = vect_get_vec_def_for_operand (op0, stmt, NULL);
vec_oprnd1 = vect_get_vec_def_for_stmt_copy (dt0, vec_oprnd0);
}
else
{
vec_oprnd0 = vect_get_vec_def_for_stmt_copy (dt0, vec_oprnd1);
vec_oprnd1 = vect_get_vec_def_for_stmt_copy (dt0, vec_oprnd0);
}
/* Arguments are ready. Create the new vector stmt. */
expr = build2 (code, vectype_out, vec_oprnd0, vec_oprnd1);
new_stmt = build_gimple_modify_stmt (vec_dest, expr);
new_temp = make_ssa_name (vec_dest, new_stmt);
GIMPLE_STMT_OPERAND (new_stmt, 0) = new_temp;
vect_finish_stmt_generation (stmt, new_stmt, bsi);
if (j == 0)
STMT_VINFO_VEC_STMT (stmt_info) = new_stmt;
else
STMT_VINFO_RELATED_STMT (prev_stmt_info) = new_stmt;
prev_stmt_info = vinfo_for_stmt (new_stmt);
}
*vec_stmt = STMT_VINFO_VEC_STMT (stmt_info);
return true;
}
/* Function vect_gen_widened_results_half
Create a vector stmt whose code, type, number of arguments, and result
variable are CODE, VECTYPE, OP_TYPE, and VEC_DEST, and its arguments are
VEC_OPRND0 and VEC_OPRND1. The new vector stmt is to be inserted at BSI.
In the case that CODE is a CALL_EXPR, this means that a call to DECL
needs to be created (DECL is a function-decl of a target-builtin).
STMT is the original scalar stmt that we are vectorizing. */
static tree
vect_gen_widened_results_half (enum tree_code code, tree vectype, tree decl,
tree vec_oprnd0, tree vec_oprnd1, int op_type,
tree vec_dest, block_stmt_iterator *bsi,
tree stmt)
{
tree expr;
tree new_stmt;
tree new_temp;
tree sym;
ssa_op_iter iter;
/* Generate half of the widened result: */
if (code == CALL_EXPR)
{
/* Target specific support */
if (op_type == binary_op)
expr = build_call_expr (decl, 2, vec_oprnd0, vec_oprnd1);
else
expr = build_call_expr (decl, 1, vec_oprnd0);
}
else
{
/* Generic support */
gcc_assert (op_type == TREE_CODE_LENGTH (code));
if (op_type == binary_op)
expr = build2 (code, vectype, vec_oprnd0, vec_oprnd1);
else
expr = build1 (code, vectype, vec_oprnd0);
}
new_stmt = build_gimple_modify_stmt (vec_dest, expr);
new_temp = make_ssa_name (vec_dest, new_stmt);
GIMPLE_STMT_OPERAND (new_stmt, 0) = new_temp;
vect_finish_stmt_generation (stmt, new_stmt, bsi);
if (code == CALL_EXPR)
{
FOR_EACH_SSA_TREE_OPERAND (sym, new_stmt, iter, SSA_OP_ALL_VIRTUALS)
{
if (TREE_CODE (sym) == SSA_NAME)
sym = SSA_NAME_VAR (sym);
mark_sym_for_renaming (sym);
}
}
return new_stmt;
}
/* Function vectorizable_type_promotion
Check if STMT performs a binary or unary operation that involves
type promotion, and if it can be vectorized.
If VEC_STMT is also passed, vectorize the STMT: create a vectorized
stmt to replace it, put it in VEC_STMT, and insert it at BSI.
Return FALSE if not a vectorizable STMT, TRUE otherwise. */
bool
vectorizable_type_promotion (tree stmt, block_stmt_iterator *bsi,
tree *vec_stmt)
{
tree vec_dest;
tree scalar_dest;
tree operation;
tree op0, op1 = NULL;
tree vec_oprnd0=NULL, vec_oprnd1=NULL;
stmt_vec_info stmt_info = vinfo_for_stmt (stmt);
loop_vec_info loop_vinfo = STMT_VINFO_LOOP_VINFO (stmt_info);
enum tree_code code, code1 = CODE_FOR_nothing, code2 = CODE_FOR_nothing;
tree decl1 = NULL_TREE, decl2 = NULL_TREE;
int op_type;
tree def, def_stmt;
enum vect_def_type dt0, dt1;
tree new_stmt;
stmt_vec_info prev_stmt_info;
int nunits_in;
int nunits_out;
tree vectype_out;
int ncopies;
int j;
tree vectype_in;
if (!STMT_VINFO_RELEVANT_P (stmt_info))
return false;
if (STMT_VINFO_DEF_TYPE (stmt_info) != vect_loop_def)
return false;
/* FORNOW: not yet supported. */
if (STMT_VINFO_LIVE_P (stmt_info))
{
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "value used after loop.");
return false;
}
/* Is STMT a vectorizable type-promotion operation? */
if (TREE_CODE (stmt) != GIMPLE_MODIFY_STMT)
return false;
if (TREE_CODE (GIMPLE_STMT_OPERAND (stmt, 0)) != SSA_NAME)
return false;
operation = GIMPLE_STMT_OPERAND (stmt, 1);
code = TREE_CODE (operation);
if (code != NOP_EXPR && code != WIDEN_MULT_EXPR)
return false;
op0 = TREE_OPERAND (operation, 0);
vectype_in = get_vectype_for_scalar_type (TREE_TYPE (op0));
nunits_in = TYPE_VECTOR_SUBPARTS (vectype_in);
ncopies = LOOP_VINFO_VECT_FACTOR (loop_vinfo) / nunits_in;
gcc_assert (ncopies >= 1);
scalar_dest = GIMPLE_STMT_OPERAND (stmt, 0);
vectype_out = get_vectype_for_scalar_type (TREE_TYPE (scalar_dest));
nunits_out = TYPE_VECTOR_SUBPARTS (vectype_out);
if (nunits_out != nunits_in / 2) /* FORNOW */
return false;
if (! ((INTEGRAL_TYPE_P (TREE_TYPE (scalar_dest))
&& INTEGRAL_TYPE_P (TREE_TYPE (op0)))
|| (SCALAR_FLOAT_TYPE_P (TREE_TYPE (scalar_dest))
&& SCALAR_FLOAT_TYPE_P (TREE_TYPE (op0))
&& (code == CONVERT_EXPR || code == NOP_EXPR))))
return false;
/* Check the operands of the operation. */
if (!vect_is_simple_use (op0, loop_vinfo, &def_stmt, &def, &dt0))
{
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "use not simple.");
return false;
}
op_type = TREE_CODE_LENGTH (code);
if (op_type == binary_op)
{
op1 = TREE_OPERAND (operation, 1);
if (!vect_is_simple_use (op1, loop_vinfo, &def_stmt, &def, &dt1))
{
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "use not simple.");
return false;
}
}
/* Supportable by target? */
if (!supportable_widening_operation (code, stmt, vectype_in,
&decl1, &decl2, &code1, &code2))
return false;
STMT_VINFO_VECTYPE (stmt_info) = vectype_in;
if (!vec_stmt) /* transformation not required. */
{
STMT_VINFO_TYPE (stmt_info) = type_promotion_vec_info_type;
return true;
}
/** Transform. **/
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "transform type promotion operation. ncopies = %d.",
ncopies);
/* Handle def. */
vec_dest = vect_create_destination_var (scalar_dest, vectype_out);
/* In case the vectorization factor (VF) is bigger than the number
of elements that we can fit in a vectype (nunits), we have to generate
more than one vector stmt - i.e - we need to "unroll" the
vector stmt by a factor VF/nunits. */
prev_stmt_info = NULL;
for (j = 0; j < ncopies; j++)
{
/* Handle uses. */
if (j == 0)
{
vec_oprnd0 = vect_get_vec_def_for_operand (op0, stmt, NULL);
if (op_type == binary_op)
vec_oprnd1 = vect_get_vec_def_for_operand (op1, stmt, NULL);
}
else
{
vec_oprnd0 = vect_get_vec_def_for_stmt_copy (dt0, vec_oprnd0);
if (op_type == binary_op)
vec_oprnd1 = vect_get_vec_def_for_stmt_copy (dt1, vec_oprnd1);
}
/* Arguments are ready. Create the new vector stmt. We are creating
two vector defs because the widened result does not fit in one vector.
The vectorized stmt can be expressed as a call to a taregt builtin,
or a using a tree-code. */
/* Generate first half of the widened result: */
new_stmt = vect_gen_widened_results_half (code1, vectype_out, decl1,
vec_oprnd0, vec_oprnd1, op_type, vec_dest, bsi, stmt);
if (j == 0)
STMT_VINFO_VEC_STMT (stmt_info) = new_stmt;
else
STMT_VINFO_RELATED_STMT (prev_stmt_info) = new_stmt;
prev_stmt_info = vinfo_for_stmt (new_stmt);
/* Generate second half of the widened result: */
new_stmt = vect_gen_widened_results_half (code2, vectype_out, decl2,
vec_oprnd0, vec_oprnd1, op_type, vec_dest, bsi, stmt);
STMT_VINFO_RELATED_STMT (prev_stmt_info) = new_stmt;
prev_stmt_info = vinfo_for_stmt (new_stmt);
}
*vec_stmt = STMT_VINFO_VEC_STMT (stmt_info);
return true;
}
/* Function vect_strided_store_supported.
Returns TRUE is INTERLEAVE_HIGH and INTERLEAVE_LOW operations are supported,
and FALSE otherwise. */
static bool
vect_strided_store_supported (tree vectype)
{
optab interleave_high_optab, interleave_low_optab;
int mode;
mode = (int) TYPE_MODE (vectype);
/* Check that the operation is supported. */
interleave_high_optab = optab_for_tree_code (VEC_INTERLEAVE_HIGH_EXPR,
vectype);
interleave_low_optab = optab_for_tree_code (VEC_INTERLEAVE_LOW_EXPR,
vectype);
if (!interleave_high_optab || !interleave_low_optab)
{
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "no optab for interleave.");
return false;
}
if (interleave_high_optab->handlers[(int) mode].insn_code
== CODE_FOR_nothing
|| interleave_low_optab->handlers[(int) mode].insn_code
== CODE_FOR_nothing)
{
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "interleave op not supported by target.");
return false;
}
return true;
}
/* Function vect_permute_store_chain.
Given a chain of interleaved stores in DR_CHAIN of LENGTH that must be
a power of 2, generate interleave_high/low stmts to reorder the data
correctly for the stores. Return the final references for stores in
RESULT_CHAIN.
E.g., LENGTH is 4 and the scalar type is short, i.e., VF is 8.
The input is 4 vectors each containing 8 elements. We assign a number to each
element, the input sequence is:
1st vec: 0 1 2 3 4 5 6 7
2nd vec: 8 9 10 11 12 13 14 15
3rd vec: 16 17 18 19 20 21 22 23
4th vec: 24 25 26 27 28 29 30 31
The output sequence should be:
1st vec: 0 8 16 24 1 9 17 25
2nd vec: 2 10 18 26 3 11 19 27
3rd vec: 4 12 20 28 5 13 21 30
4th vec: 6 14 22 30 7 15 23 31
i.e., we interleave the contents of the four vectors in their order.
We use interleave_high/low instructions to create such output. The input of
each interleave_high/low operation is two vectors:
1st vec 2nd vec
0 1 2 3 4 5 6 7
the even elements of the result vector are obtained left-to-right from the
high/low elements of the first vector. The odd elements of the result are
obtained left-to-right from the high/low elements of the second vector.
The output of interleave_high will be: 0 4 1 5
and of interleave_low: 2 6 3 7
The permutation is done in log LENGTH stages. In each stage interleave_high
and interleave_low stmts are created for each pair of vectors in DR_CHAIN,
where the first argument is taken from the first half of DR_CHAIN and the
second argument from it's second half.
In our example,
I1: interleave_high (1st vec, 3rd vec)
I2: interleave_low (1st vec, 3rd vec)
I3: interleave_high (2nd vec, 4th vec)
I4: interleave_low (2nd vec, 4th vec)
The output for the first stage is:
I1: 0 16 1 17 2 18 3 19
I2: 4 20 5 21 6 22 7 23
I3: 8 24 9 25 10 26 11 27
I4: 12 28 13 29 14 30 15 31
The output of the second stage, i.e. the final result is:
I1: 0 8 16 24 1 9 17 25
I2: 2 10 18 26 3 11 19 27
I3: 4 12 20 28 5 13 21 30
I4: 6 14 22 30 7 15 23 31. */
static bool
vect_permute_store_chain (VEC(tree,heap) *dr_chain,
unsigned int length,
tree stmt,
block_stmt_iterator *bsi,
VEC(tree,heap) **result_chain)
{
tree perm_dest, perm_stmt, vect1, vect2, high, low;
tree vectype = STMT_VINFO_VECTYPE (vinfo_for_stmt (stmt));
tree scalar_dest, tmp;
int i;
unsigned int j;
VEC(tree,heap) *first, *second;
scalar_dest = GIMPLE_STMT_OPERAND (stmt, 0);
first = VEC_alloc (tree, heap, length/2);
second = VEC_alloc (tree, heap, length/2);
/* Check that the operation is supported. */
if (!vect_strided_store_supported (vectype))
return false;
*result_chain = VEC_copy (tree, heap, dr_chain);
for (i = 0; i < exact_log2 (length); i++)
{
for (j = 0; j < length/2; j++)
{
vect1 = VEC_index (tree, dr_chain, j);
vect2 = VEC_index (tree, dr_chain, j+length/2);
/* Create interleaving stmt:
in the case of big endian:
high = interleave_high (vect1, vect2)
and in the case of little endian:
high = interleave_low (vect1, vect2). */
perm_dest = create_tmp_var (vectype, "vect_inter_high");
DECL_GIMPLE_REG_P (perm_dest) = 1;
add_referenced_var (perm_dest);
if (BYTES_BIG_ENDIAN)
tmp = build2 (VEC_INTERLEAVE_HIGH_EXPR, vectype, vect1, vect2);
else
tmp = build2 (VEC_INTERLEAVE_LOW_EXPR, vectype, vect1, vect2);
perm_stmt = build_gimple_modify_stmt (perm_dest, tmp);
high = make_ssa_name (perm_dest, perm_stmt);
GIMPLE_STMT_OPERAND (perm_stmt, 0) = high;
vect_finish_stmt_generation (stmt, perm_stmt, bsi);
VEC_replace (tree, *result_chain, 2*j, high);
/* Create interleaving stmt:
in the case of big endian:
low = interleave_low (vect1, vect2)
and in the case of little endian:
low = interleave_high (vect1, vect2). */
perm_dest = create_tmp_var (vectype, "vect_inter_low");
DECL_GIMPLE_REG_P (perm_dest) = 1;
add_referenced_var (perm_dest);
if (BYTES_BIG_ENDIAN)
tmp = build2 (VEC_INTERLEAVE_LOW_EXPR, vectype, vect1, vect2);
else
tmp = build2 (VEC_INTERLEAVE_HIGH_EXPR, vectype, vect1, vect2);
perm_stmt = build_gimple_modify_stmt (perm_dest, tmp);
low = make_ssa_name (perm_dest, perm_stmt);
GIMPLE_STMT_OPERAND (perm_stmt, 0) = low;
vect_finish_stmt_generation (stmt, perm_stmt, bsi);
VEC_replace (tree, *result_chain, 2*j+1, low);
}
dr_chain = VEC_copy (tree, heap, *result_chain);
}
return true;
}
/* Function vectorizable_store.
Check if STMT defines a non scalar data-ref (array/pointer/structure) that
can be vectorized.
If VEC_STMT is also passed, vectorize the STMT: create a vectorized
stmt to replace it, put it in VEC_STMT, and insert it at BSI.
Return FALSE if not a vectorizable STMT, TRUE otherwise. */
bool
vectorizable_store (tree stmt, block_stmt_iterator *bsi, tree *vec_stmt)
{
tree scalar_dest;
tree data_ref;
tree op;
tree vec_oprnd = NULL_TREE;
stmt_vec_info stmt_info = vinfo_for_stmt (stmt);
struct data_reference *dr = STMT_VINFO_DATA_REF (stmt_info), *first_dr = NULL;
tree vectype = STMT_VINFO_VECTYPE (stmt_info);
loop_vec_info loop_vinfo = STMT_VINFO_LOOP_VINFO (stmt_info);
enum machine_mode vec_mode;
tree dummy;
enum dr_alignment_support alignment_support_cheme;
ssa_op_iter iter;
def_operand_p def_p;
tree def, def_stmt;
enum vect_def_type dt;
stmt_vec_info prev_stmt_info = NULL;
tree dataref_ptr = NULL_TREE;
int nunits = TYPE_VECTOR_SUBPARTS (vectype);
int ncopies = LOOP_VINFO_VECT_FACTOR (loop_vinfo) / nunits;
int j;
tree next_stmt, first_stmt;
bool strided_store = false;
unsigned int group_size, i;
VEC(tree,heap) *dr_chain = NULL, *oprnds = NULL, *result_chain = NULL;
gcc_assert (ncopies >= 1);
if (!STMT_VINFO_RELEVANT_P (stmt_info))
return false;
if (STMT_VINFO_DEF_TYPE (stmt_info) != vect_loop_def)
return false;
if (STMT_VINFO_LIVE_P (stmt_info))
{
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "value used after loop.");
return false;
}
/* Is vectorizable store? */
if (TREE_CODE (stmt) != GIMPLE_MODIFY_STMT)
return false;
scalar_dest = GIMPLE_STMT_OPERAND (stmt, 0);
if (TREE_CODE (scalar_dest) != ARRAY_REF
&& TREE_CODE (scalar_dest) != INDIRECT_REF
&& !DR_GROUP_FIRST_DR (stmt_info))
return false;
op = GIMPLE_STMT_OPERAND (stmt, 1);
if (!vect_is_simple_use (op, loop_vinfo, &def_stmt, &def, &dt))
{
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "use not simple.");
return false;
}
vec_mode = TYPE_MODE (vectype);
/* FORNOW. In some cases can vectorize even if data-type not supported
(e.g. - array initialization with 0). */
if (mov_optab->handlers[(int)vec_mode].insn_code == CODE_FOR_nothing)
return false;
if (!STMT_VINFO_DATA_REF (stmt_info))
return false;
if (DR_GROUP_FIRST_DR (stmt_info))
{
strided_store = true;
if (!vect_strided_store_supported (vectype))
return false;
}
if (!vec_stmt) /* transformation not required. */
{
STMT_VINFO_TYPE (stmt_info) = store_vec_info_type;
return true;
}
/** Transform. **/
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "transform store. ncopies = %d",ncopies);
if (strided_store)
{
first_stmt = DR_GROUP_FIRST_DR (stmt_info);
first_dr = STMT_VINFO_DATA_REF (vinfo_for_stmt (first_stmt));
group_size = DR_GROUP_SIZE (vinfo_for_stmt (first_stmt));
DR_GROUP_STORE_COUNT (vinfo_for_stmt (first_stmt))++;
/* We vectorize all the stmts of the interleaving group when we
reach the last stmt in the group. */
if (DR_GROUP_STORE_COUNT (vinfo_for_stmt (first_stmt))
< DR_GROUP_SIZE (vinfo_for_stmt (first_stmt)))
{
*vec_stmt = NULL_TREE;
return true;
}
}
else
{
first_stmt = stmt;
first_dr = dr;
group_size = 1;
}
dr_chain = VEC_alloc (tree, heap, group_size);
oprnds = VEC_alloc (tree, heap, group_size);
alignment_support_cheme = vect_supportable_dr_alignment (first_dr);
gcc_assert (alignment_support_cheme);
gcc_assert (alignment_support_cheme == dr_aligned); /* FORNOW */
/* In case the vectorization factor (VF) is bigger than the number
of elements that we can fit in a vectype (nunits), we have to generate
more than one vector stmt - i.e - we need to "unroll" the
vector stmt by a factor VF/nunits. For more details see documentation in
vect_get_vec_def_for_copy_stmt. */
/* In case of interleaving (non-unit strided access):
S1: &base + 2 = x2
S2: &base = x0
S3: &base + 1 = x1
S4: &base + 3 = x3
We create vectorized stores starting from base address (the access of the
first stmt in the chain (S2 in the above example), when the last store stmt
of the chain (S4) is reached:
VS1: &base = vx2
VS2: &base + vec_size*1 = vx0
VS3: &base + vec_size*2 = vx1
VS4: &base + vec_size*3 = vx3
Then permutation statements are generated:
VS5: vx5 = VEC_INTERLEAVE_HIGH_EXPR < vx0, vx3 >
VS6: vx6 = VEC_INTERLEAVE_LOW_EXPR < vx0, vx3 >
...
And they are put in STMT_VINFO_VEC_STMT of the corresponding scalar stmts
(the order of the data-refs in the output of vect_permute_store_chain
corresponds to the order of scalar stmts in the interleaving chain - see
the documentation of vect_permute_store_chain()).
In case of both multiple types and interleaving, above vector stores and
permutation stmts are created for every copy. The result vector stmts are
put in STMT_VINFO_VEC_STMT for the first copy and in the corresponding
STMT_VINFO_RELATED_STMT for the next copies.
*/
prev_stmt_info = NULL;
for (j = 0; j < ncopies; j++)
{
tree new_stmt;
tree ptr_incr;
if (j == 0)
{
/* For interleaved stores we collect vectorized defs for all the
stores in the group in DR_CHAIN and OPRNDS. DR_CHAIN is then used
as an input to vect_permute_store_chain(), and OPRNDS as an input
to vect_get_vec_def_for_stmt_copy() for the next copy.
If the store is not strided, GROUP_SIZE is 1, and DR_CHAIN and
OPRNDS are of size 1. */
next_stmt = first_stmt;
for (i = 0; i < group_size; i++)
{
/* Since gaps are not supported for interleaved stores, GROUP_SIZE
is the exact number of stmts in the chain. Therefore, NEXT_STMT
can't be NULL_TREE. In case that there is no interleaving,
GROUP_SIZE is 1, and only one iteration of the loop will be
executed. */
gcc_assert (next_stmt);
op = GIMPLE_STMT_OPERAND (next_stmt, 1);
vec_oprnd = vect_get_vec_def_for_operand (op, next_stmt, NULL);
VEC_quick_push(tree, dr_chain, vec_oprnd);
VEC_quick_push(tree, oprnds, vec_oprnd);
next_stmt = DR_GROUP_NEXT_DR (vinfo_for_stmt (next_stmt));
}
dataref_ptr = vect_create_data_ref_ptr (first_stmt, bsi, NULL_TREE,
&dummy, &ptr_incr, false,
TREE_TYPE (vec_oprnd));
}
else
{
/* For interleaved stores we created vectorized defs for all the
defs stored in OPRNDS in the previous iteration (previous copy).
DR_CHAIN is then used as an input to vect_permute_store_chain(),
and OPRNDS as an input to vect_get_vec_def_for_stmt_copy() for the
next copy.
If the store is not strided, GROUP_SIZE is 1, and DR_CHAIN and
OPRNDS are of size 1. */
for (i = 0; i < group_size; i++)
{
vec_oprnd = vect_get_vec_def_for_stmt_copy (dt,
VEC_index (tree, oprnds, i));
VEC_replace(tree, dr_chain, i, vec_oprnd);
VEC_replace(tree, oprnds, i, vec_oprnd);
}
dataref_ptr = bump_vector_ptr (dataref_ptr, ptr_incr, bsi, stmt);
}
if (strided_store)
{
result_chain = VEC_alloc (tree, heap, group_size);
/* Permute. */
if (!vect_permute_store_chain (dr_chain, group_size, stmt, bsi,
&result_chain))
return false;
}
next_stmt = first_stmt;
for (i = 0; i < group_size; i++)
{
/* For strided stores vectorized defs are interleaved in
vect_permute_store_chain(). */
if (strided_store)
vec_oprnd = VEC_index(tree, result_chain, i);
data_ref = build_fold_indirect_ref (dataref_ptr);
/* Arguments are ready. Create the new vector stmt. */
new_stmt = build_gimple_modify_stmt (data_ref, vec_oprnd);
vect_finish_stmt_generation (stmt, new_stmt, bsi);
/* Set the VDEFs for the vector pointer. If this virtual def
has a use outside the loop and a loop peel is performed
then the def may be renamed by the peel. Mark it for
renaming so the later use will also be renamed. */
copy_virtual_operands (new_stmt, next_stmt);
if (j == 0)
{
/* The original store is deleted so the same SSA_NAMEs
can be used. */
FOR_EACH_SSA_TREE_OPERAND (def, next_stmt, iter, SSA_OP_VDEF)
{
SSA_NAME_DEF_STMT (def) = new_stmt;
mark_sym_for_renaming (SSA_NAME_VAR (def));
}
STMT_VINFO_VEC_STMT (stmt_info) = *vec_stmt = new_stmt;
}
else
{
/* Create new names for all the definitions created by COPY and
add replacement mappings for each new name. */
FOR_EACH_SSA_DEF_OPERAND (def_p, new_stmt, iter, SSA_OP_VDEF)
{
create_new_def_for (DEF_FROM_PTR (def_p), new_stmt, def_p);
mark_sym_for_renaming (SSA_NAME_VAR (DEF_FROM_PTR (def_p)));
}
STMT_VINFO_RELATED_STMT (prev_stmt_info) = new_stmt;
}
prev_stmt_info = vinfo_for_stmt (new_stmt);
next_stmt = DR_GROUP_NEXT_DR (vinfo_for_stmt (next_stmt));
if (!next_stmt)
break;
/* Bump the vector pointer. */
dataref_ptr = bump_vector_ptr (dataref_ptr, ptr_incr, bsi, stmt);
}
}
return true;
}
/* Function vect_setup_realignment
This function is called when vectorizing an unaligned load using
the dr_unaligned_software_pipeline scheme.
This function generates the following code at the loop prolog:
p = initial_addr;
msq_init = *(floor(p)); # prolog load
realignment_token = call target_builtin;
loop:
msq = phi (msq_init, ---)
The code above sets up a new (vector) pointer, pointing to the first
location accessed by STMT, and a "floor-aligned" load using that pointer.
It also generates code to compute the "realignment-token" (if the relevant
target hook was defined), and creates a phi-node at the loop-header bb
whose arguments are the result of the prolog-load (created by this
function) and the result of a load that takes place in the loop (to be
created by the caller to this function).
The caller to this function uses the phi-result (msq) to create the
realignment code inside the loop, and sets up the missing phi argument,
as follows:
loop:
msq = phi (msq_init, lsq)
lsq = *(floor(p')); # load in loop
result = realign_load (msq, lsq, realignment_token);
Input:
STMT - (scalar) load stmt to be vectorized. This load accesses
a memory location that may be unaligned.
BSI - place where new code is to be inserted.
Output:
REALIGNMENT_TOKEN - the result of a call to the builtin_mask_for_load
target hook, if defined.
Return value - the result of the loop-header phi node. */
static tree
vect_setup_realignment (tree stmt, block_stmt_iterator *bsi,
tree *realignment_token)
{
stmt_vec_info stmt_info = vinfo_for_stmt (stmt);
tree vectype = STMT_VINFO_VECTYPE (stmt_info);
loop_vec_info loop_vinfo = STMT_VINFO_LOOP_VINFO (stmt_info);
struct loop *loop = LOOP_VINFO_LOOP (loop_vinfo);
edge pe = loop_preheader_edge (loop);
tree scalar_dest = GIMPLE_STMT_OPERAND (stmt, 0);
tree vec_dest;
tree init_addr;
tree inc;
tree ptr;
tree data_ref;
tree new_stmt;
basic_block new_bb;
tree msq_init;
tree new_temp;
tree phi_stmt;
tree msq;
/* 1. Create msq_init = *(floor(p1)) in the loop preheader */
vec_dest = vect_create_destination_var (scalar_dest, vectype);
ptr = vect_create_data_ref_ptr (stmt, bsi, NULL_TREE, &init_addr, &inc, true,
NULL_TREE);
data_ref = build1 (ALIGN_INDIRECT_REF, vectype, ptr);
new_stmt = build_gimple_modify_stmt (vec_dest, data_ref);
new_temp = make_ssa_name (vec_dest, new_stmt);
GIMPLE_STMT_OPERAND (new_stmt, 0) = new_temp;
new_bb = bsi_insert_on_edge_immediate (pe, new_stmt);
gcc_assert (!new_bb);
msq_init = GIMPLE_STMT_OPERAND (new_stmt, 0);
copy_virtual_operands (new_stmt, stmt);
update_vuses_to_preheader (new_stmt, loop);
/* 2. Create permutation mask, if required, in loop preheader. */
if (targetm.vectorize.builtin_mask_for_load)
{
tree builtin_decl;
builtin_decl = targetm.vectorize.builtin_mask_for_load ();
new_stmt = build_call_expr (builtin_decl, 1, init_addr);
vec_dest = vect_create_destination_var (scalar_dest,
TREE_TYPE (new_stmt));
new_stmt = build_gimple_modify_stmt (vec_dest, new_stmt);
new_temp = make_ssa_name (vec_dest, new_stmt);
GIMPLE_STMT_OPERAND (new_stmt, 0) = new_temp;
new_bb = bsi_insert_on_edge_immediate (pe, new_stmt);
gcc_assert (!new_bb);
*realignment_token = GIMPLE_STMT_OPERAND (new_stmt, 0);
/* The result of the CALL_EXPR to this builtin is determined from
the value of the parameter and no global variables are touched
which makes the builtin a "const" function. Requiring the
builtin to have the "const" attribute makes it unnecessary
to call mark_call_clobbered. */
gcc_assert (TREE_READONLY (builtin_decl));
}
/* 3. Create msq = phi <msq_init, lsq> in loop */
vec_dest = vect_create_destination_var (scalar_dest, vectype);
msq = make_ssa_name (vec_dest, NULL_TREE);
phi_stmt = create_phi_node (msq, loop->header);
SSA_NAME_DEF_STMT (msq) = phi_stmt;
add_phi_arg (phi_stmt, msq_init, loop_preheader_edge (loop));
return msq;
}
/* Function vect_strided_load_supported.
Returns TRUE is EXTRACT_EVEN and EXTRACT_ODD operations are supported,
and FALSE otherwise. */
static bool
vect_strided_load_supported (tree vectype)
{
optab perm_even_optab, perm_odd_optab;
int mode;
mode = (int) TYPE_MODE (vectype);
perm_even_optab = optab_for_tree_code (VEC_EXTRACT_EVEN_EXPR, vectype);
if (!perm_even_optab)
{
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "no optab for perm_even.");
return false;
}
if (perm_even_optab->handlers[mode].insn_code == CODE_FOR_nothing)
{
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "perm_even op not supported by target.");
return false;
}
perm_odd_optab = optab_for_tree_code (VEC_EXTRACT_ODD_EXPR, vectype);
if (!perm_odd_optab)
{
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "no optab for perm_odd.");
return false;
}
if (perm_odd_optab->handlers[mode].insn_code == CODE_FOR_nothing)
{
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "perm_odd op not supported by target.");
return false;
}
return true;
}
/* Function vect_permute_load_chain.
Given a chain of interleaved loads in DR_CHAIN of LENGTH that must be
a power of 2, generate extract_even/odd stmts to reorder the input data
correctly. Return the final references for loads in RESULT_CHAIN.
E.g., LENGTH is 4 and the scalar type is short, i.e., VF is 8.
The input is 4 vectors each containing 8 elements. We assign a number to each
element, the input sequence is:
1st vec: 0 1 2 3 4 5 6 7
2nd vec: 8 9 10 11 12 13 14 15
3rd vec: 16 17 18 19 20 21 22 23
4th vec: 24 25 26 27 28 29 30 31
The output sequence should be:
1st vec: 0 4 8 12 16 20 24 28
2nd vec: 1 5 9 13 17 21 25 29
3rd vec: 2 6 10 14 18 22 26 30
4th vec: 3 7 11 15 19 23 27 31
i.e., the first output vector should contain the first elements of each
interleaving group, etc.
We use extract_even/odd instructions to create such output. The input of each
extract_even/odd operation is two vectors
1st vec 2nd vec
0 1 2 3 4 5 6 7
and the output is the vector of extracted even/odd elements. The output of
extract_even will be: 0 2 4 6
and of extract_odd: 1 3 5 7
The permutation is done in log LENGTH stages. In each stage extract_even and
extract_odd stmts are created for each pair of vectors in DR_CHAIN in their
order. In our example,
E1: extract_even (1st vec, 2nd vec)
E2: extract_odd (1st vec, 2nd vec)
E3: extract_even (3rd vec, 4th vec)
E4: extract_odd (3rd vec, 4th vec)
The output for the first stage will be:
E1: 0 2 4 6 8 10 12 14
E2: 1 3 5 7 9 11 13 15
E3: 16 18 20 22 24 26 28 30
E4: 17 19 21 23 25 27 29 31
In order to proceed and create the correct sequence for the next stage (or
for the correct output, if the second stage is the last one, as in our
example), we first put the output of extract_even operation and then the
output of extract_odd in RESULT_CHAIN (which is then copied to DR_CHAIN).
The input for the second stage is:
1st vec (E1): 0 2 4 6 8 10 12 14
2nd vec (E3): 16 18 20 22 24 26 28 30
3rd vec (E2): 1 3 5 7 9 11 13 15
4th vec (E4): 17 19 21 23 25 27 29 31
The output of the second stage:
E1: 0 4 8 12 16 20 24 28
E2: 2 6 10 14 18 22 26 30
E3: 1 5 9 13 17 21 25 29
E4: 3 7 11 15 19 23 27 31
And RESULT_CHAIN after reordering:
1st vec (E1): 0 4 8 12 16 20 24 28
2nd vec (E3): 1 5 9 13 17 21 25 29
3rd vec (E2): 2 6 10 14 18 22 26 30
4th vec (E4): 3 7 11 15 19 23 27 31. */
static bool
vect_permute_load_chain (VEC(tree,heap) *dr_chain,
unsigned int length,
tree stmt,
block_stmt_iterator *bsi,
VEC(tree,heap) **result_chain)
{
tree perm_dest, perm_stmt, data_ref, first_vect, second_vect;
tree vectype = STMT_VINFO_VECTYPE (vinfo_for_stmt (stmt));
tree tmp;
int i;
unsigned int j;
/* Check that the operation is supported. */
if (!vect_strided_load_supported (vectype))
return false;
*result_chain = VEC_copy (tree, heap, dr_chain);
for (i = 0; i < exact_log2 (length); i++)
{
for (j = 0; j < length; j +=2)
{
first_vect = VEC_index (tree, dr_chain, j);
second_vect = VEC_index (tree, dr_chain, j+1);
/* data_ref = permute_even (first_data_ref, second_data_ref); */
perm_dest = create_tmp_var (vectype, "vect_perm_even");
DECL_GIMPLE_REG_P (perm_dest) = 1;
add_referenced_var (perm_dest);
tmp = build2 (VEC_EXTRACT_EVEN_EXPR, vectype,
first_vect, second_vect);
perm_stmt = build_gimple_modify_stmt (perm_dest, tmp);
data_ref = make_ssa_name (perm_dest, perm_stmt);
GIMPLE_STMT_OPERAND (perm_stmt, 0) = data_ref;
vect_finish_stmt_generation (stmt, perm_stmt, bsi);
mark_symbols_for_renaming (perm_stmt);
VEC_replace (tree, *result_chain, j/2, data_ref);
/* data_ref = permute_odd (first_data_ref, second_data_ref); */
perm_dest = create_tmp_var (vectype, "vect_perm_odd");
DECL_GIMPLE_REG_P (perm_dest) = 1;
add_referenced_var (perm_dest);
tmp = build2 (VEC_EXTRACT_ODD_EXPR, vectype,
first_vect, second_vect);
perm_stmt = build_gimple_modify_stmt (perm_dest, tmp);
data_ref = make_ssa_name (perm_dest, perm_stmt);
GIMPLE_STMT_OPERAND (perm_stmt, 0) = data_ref;
vect_finish_stmt_generation (stmt, perm_stmt, bsi);
mark_symbols_for_renaming (perm_stmt);
VEC_replace (tree, *result_chain, j/2+length/2, data_ref);
}
dr_chain = VEC_copy (tree, heap, *result_chain);
}
return true;
}
/* Function vect_transform_strided_load.
Given a chain of input interleaved data-refs (in DR_CHAIN), build statements
to perform their permutation and ascribe the result vectorized statements to
the scalar statements.
*/
static bool
vect_transform_strided_load (tree stmt, VEC(tree,heap) *dr_chain, int size,
block_stmt_iterator *bsi)
{
stmt_vec_info stmt_info = vinfo_for_stmt (stmt);
tree first_stmt = DR_GROUP_FIRST_DR (stmt_info);
tree next_stmt, new_stmt;
VEC(tree,heap) *result_chain = NULL;
unsigned int i, gap_count;
tree tmp_data_ref;
/* DR_CHAIN contains input data-refs that are a part of the interleaving.
RESULT_CHAIN is the output of vect_permute_load_chain, it contains permuted
vectors, that are ready for vector computation. */
result_chain = VEC_alloc (tree, heap, size);
/* Permute. */
if (!vect_permute_load_chain (dr_chain, size, stmt, bsi, &result_chain))
return false;
/* Put a permuted data-ref in the VECTORIZED_STMT field.
Since we scan the chain starting from it's first node, their order
corresponds the order of data-refs in RESULT_CHAIN. */
next_stmt = first_stmt;
gap_count = 1;
for (i = 0; VEC_iterate(tree, result_chain, i, tmp_data_ref); i++)
{
if (!next_stmt)
break;
/* Skip the gaps. Loads created for the gaps will be removed by dead
code elimination pass later.
DR_GROUP_GAP is the number of steps in elements from the previous
access (if there is no gap DR_GROUP_GAP is 1). We skip loads that
correspond to the gaps.
*/
if (gap_count < DR_GROUP_GAP (vinfo_for_stmt (next_stmt)))
{
gap_count++;
continue;
}
while (next_stmt)
{
new_stmt = SSA_NAME_DEF_STMT (tmp_data_ref);
/* We assume that if VEC_STMT is not NULL, this is a case of multiple
copies, and we put the new vector statement in the first available
RELATED_STMT. */
if (!STMT_VINFO_VEC_STMT (vinfo_for_stmt (next_stmt)))
STMT_VINFO_VEC_STMT (vinfo_for_stmt (next_stmt)) = new_stmt;
else
{
tree prev_stmt = STMT_VINFO_VEC_STMT (vinfo_for_stmt (next_stmt));
tree rel_stmt = STMT_VINFO_RELATED_STMT (
vinfo_for_stmt (prev_stmt));
while (rel_stmt)
{
prev_stmt = rel_stmt;
rel_stmt = STMT_VINFO_RELATED_STMT (vinfo_for_stmt (rel_stmt));
}
STMT_VINFO_RELATED_STMT (vinfo_for_stmt (prev_stmt)) = new_stmt;
}
next_stmt = DR_GROUP_NEXT_DR (vinfo_for_stmt (next_stmt));
gap_count = 1;
/* If NEXT_STMT accesses the same DR as the previous statement,
put the same TMP_DATA_REF as its vectorized statement; otherwise
get the next data-ref from RESULT_CHAIN. */
if (!next_stmt || !DR_GROUP_SAME_DR_STMT (vinfo_for_stmt (next_stmt)))
break;
}
}
return true;
}
/* vectorizable_load.
Check if STMT reads a non scalar data-ref (array/pointer/structure) that
can be vectorized.
If VEC_STMT is also passed, vectorize the STMT: create a vectorized
stmt to replace it, put it in VEC_STMT, and insert it at BSI.
Return FALSE if not a vectorizable STMT, TRUE otherwise. */
bool
vectorizable_load (tree stmt, block_stmt_iterator *bsi, tree *vec_stmt)
{
tree scalar_dest;
tree vec_dest = NULL;
tree data_ref = NULL;
tree op;
stmt_vec_info stmt_info = vinfo_for_stmt (stmt);
stmt_vec_info prev_stmt_info;
loop_vec_info loop_vinfo = STMT_VINFO_LOOP_VINFO (stmt_info);
struct loop *loop = LOOP_VINFO_LOOP (loop_vinfo);
struct data_reference *dr = STMT_VINFO_DATA_REF (stmt_info), *first_dr;
tree vectype = STMT_VINFO_VECTYPE (stmt_info);
tree new_temp;
int mode;
tree new_stmt = NULL_TREE;
tree dummy;
enum dr_alignment_support alignment_support_cheme;
tree dataref_ptr = NULL_TREE;
tree ptr_incr;
int nunits = TYPE_VECTOR_SUBPARTS (vectype);
int ncopies = LOOP_VINFO_VECT_FACTOR (loop_vinfo) / nunits;
int i, j, group_size;
tree msq = NULL_TREE, lsq;
tree offset = NULL_TREE;
tree realignment_token = NULL_TREE;
tree phi_stmt = NULL_TREE;
VEC(tree,heap) *dr_chain = NULL;
bool strided_load = false;
tree first_stmt;
if (!STMT_VINFO_RELEVANT_P (stmt_info))
return false;
if (STMT_VINFO_DEF_TYPE (stmt_info) != vect_loop_def)
return false;
/* FORNOW: not yet supported. */
if (STMT_VINFO_LIVE_P (stmt_info))
{
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "value used after loop.");
return false;
}
/* Is vectorizable load? */
if (TREE_CODE (stmt) != GIMPLE_MODIFY_STMT)
return false;
scalar_dest = GIMPLE_STMT_OPERAND (stmt, 0);
if (TREE_CODE (scalar_dest) != SSA_NAME)
return false;
op = GIMPLE_STMT_OPERAND (stmt, 1);
if (TREE_CODE (op) != ARRAY_REF
&& TREE_CODE (op) != INDIRECT_REF
&& !DR_GROUP_FIRST_DR (stmt_info))
return false;
if (!STMT_VINFO_DATA_REF (stmt_info))
return false;
mode = (int) TYPE_MODE (vectype);
/* FORNOW. In some cases can vectorize even if data-type not supported
(e.g. - data copies). */
if (mov_optab->handlers[mode].insn_code == CODE_FOR_nothing)
{
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "Aligned load, but unsupported type.");
return false;
}
/* Check if the load is a part of an interleaving chain. */
if (DR_GROUP_FIRST_DR (stmt_info))
{
strided_load = true;
/* Check if interleaving is supported. */
if (!vect_strided_load_supported (vectype))
return false;
}
if (!vec_stmt) /* transformation not required. */
{
STMT_VINFO_TYPE (stmt_info) = load_vec_info_type;
return true;
}
/** Transform. **/
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "transform load.");
if (strided_load)
{
first_stmt = DR_GROUP_FIRST_DR (stmt_info);
/* Check if the chain of loads is already vectorized. */
if (STMT_VINFO_VEC_STMT (vinfo_for_stmt (first_stmt)))
{
*vec_stmt = STMT_VINFO_VEC_STMT (stmt_info);
return true;
}
first_dr = STMT_VINFO_DATA_REF (vinfo_for_stmt (first_stmt));
group_size = DR_GROUP_SIZE (vinfo_for_stmt (first_stmt));
dr_chain = VEC_alloc (tree, heap, group_size);
}
else
{
first_stmt = stmt;
first_dr = dr;
group_size = 1;
}
alignment_support_cheme = vect_supportable_dr_alignment (first_dr);
gcc_assert (alignment_support_cheme);
/* In case the vectorization factor (VF) is bigger than the number
of elements that we can fit in a vectype (nunits), we have to generate
more than one vector stmt - i.e - we need to "unroll" the
vector stmt by a factor VF/nunits. In doing so, we record a pointer
from one copy of the vector stmt to the next, in the field
STMT_VINFO_RELATED_STMT. This is necessary in order to allow following
stages to find the correct vector defs to be used when vectorizing
stmts that use the defs of the current stmt. The example below illustrates
the vectorization process when VF=16 and nunits=4 (i.e - we need to create
4 vectorized stmts):
before vectorization:
RELATED_STMT VEC_STMT
S1: x = memref - -
S2: z = x + 1 - -
step 1: vectorize stmt S1:
We first create the vector stmt VS1_0, and, as usual, record a
pointer to it in the STMT_VINFO_VEC_STMT of the scalar stmt S1.
Next, we create the vector stmt VS1_1, and record a pointer to
it in the STMT_VINFO_RELATED_STMT of the vector stmt VS1_0.
Similarly, for VS1_2 and VS1_3. This is the resulting chain of
stmts and pointers:
RELATED_STMT VEC_STMT
VS1_0: vx0 = memref0 VS1_1 -
VS1_1: vx1 = memref1 VS1_2 -
VS1_2: vx2 = memref2 VS1_3 -
VS1_3: vx3 = memref3 - -
S1: x = load - VS1_0
S2: z = x + 1 - -
See in documentation in vect_get_vec_def_for_stmt_copy for how the
information we recorded in RELATED_STMT field is used to vectorize
stmt S2. */
/* In case of interleaving (non-unit strided access):
S1: x2 = &base + 2
S2: x0 = &base
S3: x1 = &base + 1
S4: x3 = &base + 3
Vectorized loads are created in the order of memory accesses
starting from the access of the first stmt of the chain:
VS1: vx0 = &base
VS2: vx1 = &base + vec_size*1
VS3: vx3 = &base + vec_size*2
VS4: vx4 = &base + vec_size*3
Then permutation statements are generated:
VS5: vx5 = VEC_EXTRACT_EVEN_EXPR < vx0, vx1 >
VS6: vx6 = VEC_EXTRACT_ODD_EXPR < vx0, vx1 >
...
And they are put in STMT_VINFO_VEC_STMT of the corresponding scalar stmts
(the order of the data-refs in the output of vect_permute_load_chain
corresponds to the order of scalar stmts in the interleaving chain - see
the documentation of vect_permute_load_chain()).
The generation of permutation stmts and recording them in
STMT_VINFO_VEC_STMT is done in vect_transform_strided_load().
In case of both multiple types and interleaving, the vector loads and
permutation stmts above are created for every copy. The result vector stmts
are put in STMT_VINFO_VEC_STMT for the first copy and in the corresponding
STMT_VINFO_RELATED_STMT for the next copies. */
/* If the data reference is aligned (dr_aligned) or potentially unaligned
on a target that supports unaligned accesses (dr_unaligned_supported)
we generate the following code:
p = initial_addr;
indx = 0;
loop {
p = p + indx * vectype_size;
vec_dest = *(p);
indx = indx + 1;
}
Otherwise, the data reference is potentially unaligned on a target that
does not support unaligned accesses (dr_unaligned_software_pipeline) -
then generate the following code, in which the data in each iteration is
obtained by two vector loads, one from the previous iteration, and one
from the current iteration:
p1 = initial_addr;
msq_init = *(floor(p1))
p2 = initial_addr + VS - 1;
realignment_token = call target_builtin;
indx = 0;
loop {
p2 = p2 + indx * vectype_size
lsq = *(floor(p2))
vec_dest = realign_load (msq, lsq, realignment_token)
indx = indx + 1;
msq = lsq;
} */
if (alignment_support_cheme == dr_unaligned_software_pipeline)
{
msq = vect_setup_realignment (first_stmt, bsi, &realignment_token);
phi_stmt = SSA_NAME_DEF_STMT (msq);
offset = size_int (TYPE_VECTOR_SUBPARTS (vectype) - 1);
}
prev_stmt_info = NULL;
for (j = 0; j < ncopies; j++)
{
/* 1. Create the vector pointer update chain. */
if (j == 0)
dataref_ptr = vect_create_data_ref_ptr (first_stmt, bsi, offset, &dummy,
&ptr_incr, false, NULL_TREE);
else
dataref_ptr = bump_vector_ptr (dataref_ptr, ptr_incr, bsi, stmt);
for (i = 0; i < group_size; i++)
{
/* 2. Create the vector-load in the loop. */
switch (alignment_support_cheme)
{
case dr_aligned:
gcc_assert (aligned_access_p (first_dr));
data_ref = build_fold_indirect_ref (dataref_ptr);
break;
case dr_unaligned_supported:
{
int mis = DR_MISALIGNMENT (first_dr);
tree tmis = (mis == -1 ? size_zero_node : size_int (mis));
gcc_assert (!aligned_access_p (first_dr));
tmis = size_binop (MULT_EXPR, tmis, size_int(BITS_PER_UNIT));
data_ref =
build2 (MISALIGNED_INDIRECT_REF, vectype, dataref_ptr, tmis);
break;
}
case dr_unaligned_software_pipeline:
gcc_assert (!aligned_access_p (first_dr));
data_ref = build1 (ALIGN_INDIRECT_REF, vectype, dataref_ptr);
break;
default:
gcc_unreachable ();
}
vec_dest = vect_create_destination_var (scalar_dest, vectype);
new_stmt = build_gimple_modify_stmt (vec_dest, data_ref);
new_temp = make_ssa_name (vec_dest, new_stmt);
GIMPLE_STMT_OPERAND (new_stmt, 0) = new_temp;
vect_finish_stmt_generation (stmt, new_stmt, bsi);
copy_virtual_operands (new_stmt, stmt);
mark_symbols_for_renaming (new_stmt);
/* 3. Handle explicit realignment if necessary/supported. */
if (alignment_support_cheme == dr_unaligned_software_pipeline)
{
/* Create in loop:
<vec_dest = realign_load (msq, lsq, realignment_token)> */
lsq = GIMPLE_STMT_OPERAND (new_stmt, 0);
if (!realignment_token)
realignment_token = dataref_ptr;
vec_dest = vect_create_destination_var (scalar_dest, vectype);
new_stmt =
build3 (REALIGN_LOAD_EXPR, vectype, msq, lsq, realignment_token);
new_stmt = build_gimple_modify_stmt (vec_dest, new_stmt);
new_temp = make_ssa_name (vec_dest, new_stmt);
GIMPLE_STMT_OPERAND (new_stmt, 0) = new_temp;
vect_finish_stmt_generation (stmt, new_stmt, bsi);
if (i == group_size - 1 && j == ncopies - 1)
add_phi_arg (phi_stmt, lsq, loop_latch_edge (loop));
msq = lsq;
}
if (strided_load)
VEC_quick_push (tree, dr_chain, new_temp);
if (i < group_size - 1)
dataref_ptr = bump_vector_ptr (dataref_ptr, ptr_incr, bsi, stmt);
}
if (strided_load)
{
if (!vect_transform_strided_load (stmt, dr_chain, group_size, bsi))
return false;
*vec_stmt = STMT_VINFO_VEC_STMT (stmt_info);
dr_chain = VEC_alloc (tree, heap, group_size);
}
else
{
if (j == 0)
STMT_VINFO_VEC_STMT (stmt_info) = *vec_stmt = new_stmt;
else
STMT_VINFO_RELATED_STMT (prev_stmt_info) = new_stmt;
prev_stmt_info = vinfo_for_stmt (new_stmt);
}
}
return true;
}
/* Function vectorizable_live_operation.
STMT computes a value that is used outside the loop. Check if
it can be supported. */
bool
vectorizable_live_operation (tree stmt,
block_stmt_iterator *bsi ATTRIBUTE_UNUSED,
tree *vec_stmt ATTRIBUTE_UNUSED)
{
tree operation;
stmt_vec_info stmt_info = vinfo_for_stmt (stmt);
loop_vec_info loop_vinfo = STMT_VINFO_LOOP_VINFO (stmt_info);
int i;
int op_type;
tree op;
tree def, def_stmt;
enum vect_def_type dt;
gcc_assert (STMT_VINFO_LIVE_P (stmt_info));
if (STMT_VINFO_DEF_TYPE (stmt_info) == vect_reduction_def)
return false;
if (TREE_CODE (stmt) != GIMPLE_MODIFY_STMT)
return false;
if (TREE_CODE (GIMPLE_STMT_OPERAND (stmt, 0)) != SSA_NAME)
return false;
operation = GIMPLE_STMT_OPERAND (stmt, 1);
op_type = TREE_OPERAND_LENGTH (operation);
/* FORNOW: support only if all uses are invariant. This means
that the scalar operations can remain in place, unvectorized.
The original last scalar value that they compute will be used. */
for (i = 0; i < op_type; i++)
{
op = TREE_OPERAND (operation, i);
if (!vect_is_simple_use (op, loop_vinfo, &def_stmt, &def, &dt))
{
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "use not simple.");
return false;
}
if (dt != vect_invariant_def && dt != vect_constant_def)
return false;
}
/* No transformation is required for the cases we currently support. */
return true;
}
/* Function vect_is_simple_cond.
Input:
LOOP - the loop that is being vectorized.
COND - Condition that is checked for simple use.
Returns whether a COND can be vectorized. Checks whether
condition operands are supportable using vec_is_simple_use. */
static bool
vect_is_simple_cond (tree cond, loop_vec_info loop_vinfo)
{
tree lhs, rhs;
tree def;
enum vect_def_type dt;
if (!COMPARISON_CLASS_P (cond))
return false;
lhs = TREE_OPERAND (cond, 0);
rhs = TREE_OPERAND (cond, 1);
if (TREE_CODE (lhs) == SSA_NAME)
{
tree lhs_def_stmt = SSA_NAME_DEF_STMT (lhs);
if (!vect_is_simple_use (lhs, loop_vinfo, &lhs_def_stmt, &def, &dt))
return false;
}
else if (TREE_CODE (lhs) != INTEGER_CST && TREE_CODE (lhs) != REAL_CST)
return false;
if (TREE_CODE (rhs) == SSA_NAME)
{
tree rhs_def_stmt = SSA_NAME_DEF_STMT (rhs);
if (!vect_is_simple_use (rhs, loop_vinfo, &rhs_def_stmt, &def, &dt))
return false;
}
else if (TREE_CODE (rhs) != INTEGER_CST && TREE_CODE (rhs) != REAL_CST)
return false;
return true;
}
/* vectorizable_condition.
Check if STMT is conditional modify expression that can be vectorized.
If VEC_STMT is also passed, vectorize the STMT: create a vectorized
stmt using VEC_COND_EXPR to replace it, put it in VEC_STMT, and insert it
at BSI.
Return FALSE if not a vectorizable STMT, TRUE otherwise. */
bool
vectorizable_condition (tree stmt, block_stmt_iterator *bsi, tree *vec_stmt)
{
tree scalar_dest = NULL_TREE;
tree vec_dest = NULL_TREE;
tree op = NULL_TREE;
tree cond_expr, then_clause, else_clause;
stmt_vec_info stmt_info = vinfo_for_stmt (stmt);
tree vectype = STMT_VINFO_VECTYPE (stmt_info);
tree vec_cond_lhs, vec_cond_rhs, vec_then_clause, vec_else_clause;
tree vec_compare, vec_cond_expr;
tree new_temp;
loop_vec_info loop_vinfo = STMT_VINFO_LOOP_VINFO (stmt_info);
enum machine_mode vec_mode;
tree def;
enum vect_def_type dt;
int nunits = TYPE_VECTOR_SUBPARTS (vectype);
int ncopies = LOOP_VINFO_VECT_FACTOR (loop_vinfo) / nunits;
gcc_assert (ncopies >= 1);
if (ncopies > 1)
return false; /* FORNOW */
if (!STMT_VINFO_RELEVANT_P (stmt_info))
return false;
if (STMT_VINFO_DEF_TYPE (stmt_info) != vect_loop_def)
return false;
/* FORNOW: not yet supported. */
if (STMT_VINFO_LIVE_P (stmt_info))
{
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "value used after loop.");
return false;
}
/* Is vectorizable conditional operation? */
if (TREE_CODE (stmt) != GIMPLE_MODIFY_STMT)
return false;
op = GIMPLE_STMT_OPERAND (stmt, 1);
if (TREE_CODE (op) != COND_EXPR)
return false;
cond_expr = TREE_OPERAND (op, 0);
then_clause = TREE_OPERAND (op, 1);
else_clause = TREE_OPERAND (op, 2);
if (!vect_is_simple_cond (cond_expr, loop_vinfo))
return false;
/* We do not handle two different vector types for the condition
and the values. */
if (TREE_TYPE (TREE_OPERAND (cond_expr, 0)) != TREE_TYPE (vectype))
return false;
if (TREE_CODE (then_clause) == SSA_NAME)
{
tree then_def_stmt = SSA_NAME_DEF_STMT (then_clause);
if (!vect_is_simple_use (then_clause, loop_vinfo,
&then_def_stmt, &def, &dt))
return false;
}
else if (TREE_CODE (then_clause) != INTEGER_CST
&& TREE_CODE (then_clause) != REAL_CST)
return false;
if (TREE_CODE (else_clause) == SSA_NAME)
{
tree else_def_stmt = SSA_NAME_DEF_STMT (else_clause);
if (!vect_is_simple_use (else_clause, loop_vinfo,
&else_def_stmt, &def, &dt))
return false;
}
else if (TREE_CODE (else_clause) != INTEGER_CST
&& TREE_CODE (else_clause) != REAL_CST)
return false;
vec_mode = TYPE_MODE (vectype);
if (!vec_stmt)
{
STMT_VINFO_TYPE (stmt_info) = condition_vec_info_type;
return expand_vec_cond_expr_p (op, vec_mode);
}
/* Transform */
/* Handle def. */
scalar_dest = GIMPLE_STMT_OPERAND (stmt, 0);
vec_dest = vect_create_destination_var (scalar_dest, vectype);
/* Handle cond expr. */
vec_cond_lhs =
vect_get_vec_def_for_operand (TREE_OPERAND (cond_expr, 0), stmt, NULL);
vec_cond_rhs =
vect_get_vec_def_for_operand (TREE_OPERAND (cond_expr, 1), stmt, NULL);
vec_then_clause = vect_get_vec_def_for_operand (then_clause, stmt, NULL);
vec_else_clause = vect_get_vec_def_for_operand (else_clause, stmt, NULL);
/* Arguments are ready. create the new vector stmt. */
vec_compare = build2 (TREE_CODE (cond_expr), vectype,
vec_cond_lhs, vec_cond_rhs);
vec_cond_expr = build3 (VEC_COND_EXPR, vectype,
vec_compare, vec_then_clause, vec_else_clause);
*vec_stmt = build_gimple_modify_stmt (vec_dest, vec_cond_expr);
new_temp = make_ssa_name (vec_dest, *vec_stmt);
GIMPLE_STMT_OPERAND (*vec_stmt, 0) = new_temp;
vect_finish_stmt_generation (stmt, *vec_stmt, bsi);
return true;
}
/* Function vect_transform_stmt.
Create a vectorized stmt to replace STMT, and insert it at BSI. */
bool
vect_transform_stmt (tree stmt, block_stmt_iterator *bsi, bool *strided_store)
{
bool is_store = false;
tree vec_stmt = NULL_TREE;
stmt_vec_info stmt_info = vinfo_for_stmt (stmt);
tree orig_stmt_in_pattern;
bool done;
switch (STMT_VINFO_TYPE (stmt_info))
{
case type_demotion_vec_info_type:
done = vectorizable_type_demotion (stmt, bsi, &vec_stmt);
gcc_assert (done);
break;
case type_promotion_vec_info_type:
done = vectorizable_type_promotion (stmt, bsi, &vec_stmt);
gcc_assert (done);
break;
case type_conversion_vec_info_type:
done = vectorizable_conversion (stmt, bsi, &vec_stmt);
gcc_assert (done);
break;
case induc_vec_info_type:
done = vectorizable_induction (stmt, bsi, &vec_stmt);
gcc_assert (done);
break;
case op_vec_info_type:
done = vectorizable_operation (stmt, bsi, &vec_stmt);
gcc_assert (done);
break;
case assignment_vec_info_type:
done = vectorizable_assignment (stmt, bsi, &vec_stmt);
gcc_assert (done);
break;
case load_vec_info_type:
done = vectorizable_load (stmt, bsi, &vec_stmt);
gcc_assert (done);
break;
case store_vec_info_type:
done = vectorizable_store (stmt, bsi, &vec_stmt);
gcc_assert (done);
if (DR_GROUP_FIRST_DR (stmt_info))
{
/* In case of interleaving, the whole chain is vectorized when the
last store in the chain is reached. Store stmts before the last
one are skipped, and there vec_stmt_info shouldn't be freed
meanwhile. */
*strided_store = true;
if (STMT_VINFO_VEC_STMT (stmt_info))
is_store = true;
}
else
is_store = true;
break;
case condition_vec_info_type:
done = vectorizable_condition (stmt, bsi, &vec_stmt);
gcc_assert (done);
break;
case call_vec_info_type:
done = vectorizable_call (stmt, bsi, &vec_stmt);
break;
case reduc_vec_info_type:
done = vectorizable_reduction (stmt, bsi, &vec_stmt);
gcc_assert (done);
break;
default:
if (!STMT_VINFO_LIVE_P (stmt_info))
{
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "stmt not supported.");
gcc_unreachable ();
}
}
if (STMT_VINFO_LIVE_P (stmt_info)
&& STMT_VINFO_TYPE (stmt_info) != reduc_vec_info_type)
{
done = vectorizable_live_operation (stmt, bsi, &vec_stmt);
gcc_assert (done);
}
if (vec_stmt)
{
STMT_VINFO_VEC_STMT (stmt_info) = vec_stmt;
orig_stmt_in_pattern = STMT_VINFO_RELATED_STMT (stmt_info);
if (orig_stmt_in_pattern)
{
stmt_vec_info stmt_vinfo = vinfo_for_stmt (orig_stmt_in_pattern);
/* STMT was inserted by the vectorizer to replace a computation idiom.
ORIG_STMT_IN_PATTERN is a stmt in the original sequence that
computed this idiom. We need to record a pointer to VEC_STMT in
the stmt_info of ORIG_STMT_IN_PATTERN. See more details in the
documentation of vect_pattern_recog. */
if (STMT_VINFO_IN_PATTERN_P (stmt_vinfo))
{
gcc_assert (STMT_VINFO_RELATED_STMT (stmt_vinfo) == stmt);
STMT_VINFO_VEC_STMT (stmt_vinfo) = vec_stmt;
}
}
}
return is_store;
}
/* This function builds ni_name = number of iterations loop executes
on the loop preheader. */
static tree
vect_build_loop_niters (loop_vec_info loop_vinfo)
{
tree ni_name, stmt, var;
edge pe;
struct loop *loop = LOOP_VINFO_LOOP (loop_vinfo);
tree ni = unshare_expr (LOOP_VINFO_NITERS (loop_vinfo));
var = create_tmp_var (TREE_TYPE (ni), "niters");
add_referenced_var (var);
ni_name = force_gimple_operand (ni, &stmt, false, var);
pe = loop_preheader_edge (loop);
if (stmt)
{
basic_block new_bb = bsi_insert_on_edge_immediate (pe, stmt);
gcc_assert (!new_bb);
}
return ni_name;
}
/* This function generates the following statements:
ni_name = number of iterations loop executes
ratio = ni_name / vf
ratio_mult_vf_name = ratio * vf
and places them at the loop preheader edge. */
static void
vect_generate_tmps_on_preheader (loop_vec_info loop_vinfo,
tree *ni_name_ptr,
tree *ratio_mult_vf_name_ptr,
tree *ratio_name_ptr)
{
edge pe;
basic_block new_bb;
tree stmt, ni_name;
tree var;
tree ratio_name;
tree ratio_mult_vf_name;
struct loop *loop = LOOP_VINFO_LOOP (loop_vinfo);
tree ni = LOOP_VINFO_NITERS (loop_vinfo);
int vf = LOOP_VINFO_VECT_FACTOR (loop_vinfo);
tree log_vf;
pe = loop_preheader_edge (loop);
/* Generate temporary variable that contains
number of iterations loop executes. */
ni_name = vect_build_loop_niters (loop_vinfo);
log_vf = build_int_cst (TREE_TYPE (ni), exact_log2 (vf));
/* Create: ratio = ni >> log2(vf) */
ratio_name = fold_build2 (RSHIFT_EXPR, TREE_TYPE (ni_name), ni_name, log_vf);
if (!is_gimple_val (ratio_name))
{
var = create_tmp_var (TREE_TYPE (ni), "bnd");
add_referenced_var (var);
ratio_name = force_gimple_operand (ratio_name, &stmt, true, var);
pe = loop_preheader_edge (loop);
new_bb = bsi_insert_on_edge_immediate (pe, stmt);
gcc_assert (!new_bb);
}
/* Create: ratio_mult_vf = ratio << log2 (vf). */
ratio_mult_vf_name = fold_build2 (LSHIFT_EXPR, TREE_TYPE (ratio_name),
ratio_name, log_vf);
if (!is_gimple_val (ratio_mult_vf_name))
{
var = create_tmp_var (TREE_TYPE (ni), "ratio_mult_vf");
add_referenced_var (var);
ratio_mult_vf_name = force_gimple_operand (ratio_mult_vf_name, &stmt,
true, var);
pe = loop_preheader_edge (loop);
new_bb = bsi_insert_on_edge_immediate (pe, stmt);
gcc_assert (!new_bb);
}
*ni_name_ptr = ni_name;
*ratio_mult_vf_name_ptr = ratio_mult_vf_name;
*ratio_name_ptr = ratio_name;
return;
}
/* Function update_vuses_to_preheader.
Input:
STMT - a statement with potential VUSEs.
LOOP - the loop whose preheader will contain STMT.
It's possible to vectorize a loop even though an SSA_NAME from a VUSE
appears to be defined in a VDEF in another statement in a loop.
One such case is when the VUSE is at the dereference of a __restricted__
pointer in a load and the VDEF is at the dereference of a different
__restricted__ pointer in a store. Vectorization may result in
copy_virtual_uses being called to copy the problematic VUSE to a new
statement that is being inserted in the loop preheader. This procedure
is called to change the SSA_NAME in the new statement's VUSE from the
SSA_NAME updated in the loop to the related SSA_NAME available on the
path entering the loop.
When this function is called, we have the following situation:
# vuse <name1>
S1: vload
do {
# name1 = phi < name0 , name2>
# vuse <name1>
S2: vload
# name2 = vdef <name1>
S3: vstore
}while...
Stmt S1 was created in the loop preheader block as part of misaligned-load
handling. This function fixes the name of the vuse of S1 from 'name1' to
'name0'. */
static void
update_vuses_to_preheader (tree stmt, struct loop *loop)
{
basic_block header_bb = loop->header;
edge preheader_e = loop_preheader_edge (loop);
ssa_op_iter iter;
use_operand_p use_p;
FOR_EACH_SSA_USE_OPERAND (use_p, stmt, iter, SSA_OP_VUSE)
{
tree ssa_name = USE_FROM_PTR (use_p);
tree def_stmt = SSA_NAME_DEF_STMT (ssa_name);
tree name_var = SSA_NAME_VAR (ssa_name);
basic_block bb = bb_for_stmt (def_stmt);
/* For a use before any definitions, def_stmt is a NOP_EXPR. */
if (!IS_EMPTY_STMT (def_stmt)
&& flow_bb_inside_loop_p (loop, bb))
{
/* If the block containing the statement defining the SSA_NAME
is in the loop then it's necessary to find the definition
outside the loop using the PHI nodes of the header. */
tree phi;
bool updated = false;
for (phi = phi_nodes (header_bb); phi; phi = PHI_CHAIN (phi))
{
if (SSA_NAME_VAR (PHI_RESULT (phi)) == name_var)
{
SET_USE (use_p, PHI_ARG_DEF (phi, preheader_e->dest_idx));
updated = true;
break;
}
}
gcc_assert (updated);
}
}
}
/* Function vect_update_ivs_after_vectorizer.
"Advance" the induction variables of LOOP to the value they should take
after the execution of LOOP. This is currently necessary because the
vectorizer does not handle induction variables that are used after the
loop. Such a situation occurs when the last iterations of LOOP are
peeled, because:
1. We introduced new uses after LOOP for IVs that were not originally used
after LOOP: the IVs of LOOP are now used by an epilog loop.
2. LOOP is going to be vectorized; this means that it will iterate N/VF
times, whereas the loop IVs should be bumped N times.
Input:
- LOOP - a loop that is going to be vectorized. The last few iterations
of LOOP were peeled.
- NITERS - the number of iterations that LOOP executes (before it is
vectorized). i.e, the number of times the ivs should be bumped.
- UPDATE_E - a successor edge of LOOP->exit that is on the (only) path
coming out from LOOP on which there are uses of the LOOP ivs
(this is the path from LOOP->exit to epilog_loop->preheader).
The new definitions of the ivs are placed in LOOP->exit.
The phi args associated with the edge UPDATE_E in the bb
UPDATE_E->dest are updated accordingly.
Assumption 1: Like the rest of the vectorizer, this function assumes
a single loop exit that has a single predecessor.
Assumption 2: The phi nodes in the LOOP header and in update_bb are
organized in the same order.
Assumption 3: The access function of the ivs is simple enough (see
vect_can_advance_ivs_p). This assumption will be relaxed in the future.
Assumption 4: Exactly one of the successors of LOOP exit-bb is on a path
coming out of LOOP on which the ivs of LOOP are used (this is the path
that leads to the epilog loop; other paths skip the epilog loop). This
path starts with the edge UPDATE_E, and its destination (denoted update_bb)
needs to have its phis updated.
*/
static void
vect_update_ivs_after_vectorizer (loop_vec_info loop_vinfo, tree niters,
edge update_e)
{
struct loop *loop = LOOP_VINFO_LOOP (loop_vinfo);
basic_block exit_bb = single_exit (loop)->dest;
tree phi, phi1;
basic_block update_bb = update_e->dest;
/* gcc_assert (vect_can_advance_ivs_p (loop_vinfo)); */
/* Make sure there exists a single-predecessor exit bb: */
gcc_assert (single_pred_p (exit_bb));
for (phi = phi_nodes (loop->header), phi1 = phi_nodes (update_bb);
phi && phi1;
phi = PHI_CHAIN (phi), phi1 = PHI_CHAIN (phi1))
{
tree access_fn = NULL;
tree evolution_part;
tree init_expr;
tree step_expr;
tree var, stmt, ni, ni_name;
block_stmt_iterator last_bsi;
if (vect_print_dump_info (REPORT_DETAILS))
{
fprintf (vect_dump, "vect_update_ivs_after_vectorizer: phi: ");
print_generic_expr (vect_dump, phi, TDF_SLIM);
}
/* Skip virtual phi's. */
if (!is_gimple_reg (SSA_NAME_VAR (PHI_RESULT (phi))))
{
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "virtual phi. skip.");
continue;
}
/* Skip reduction phis. */
if (STMT_VINFO_DEF_TYPE (vinfo_for_stmt (phi)) == vect_reduction_def)
{
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "reduc phi. skip.");
continue;
}
access_fn = analyze_scalar_evolution (loop, PHI_RESULT (phi));
gcc_assert (access_fn);
evolution_part =
unshare_expr (evolution_part_in_loop_num (access_fn, loop->num));
gcc_assert (evolution_part != NULL_TREE);
/* FORNOW: We do not support IVs whose evolution function is a polynomial
of degree >= 2 or exponential. */
gcc_assert (!tree_is_chrec (evolution_part));
step_expr = evolution_part;
init_expr = unshare_expr (initial_condition_in_loop_num (access_fn,
loop->num));
ni = fold_build2 (PLUS_EXPR, TREE_TYPE (init_expr),
fold_build2 (MULT_EXPR, TREE_TYPE (init_expr),
fold_convert (TREE_TYPE (init_expr),
niters),
step_expr),
init_expr);
var = create_tmp_var (TREE_TYPE (init_expr), "tmp");
add_referenced_var (var);
ni_name = force_gimple_operand (ni, &stmt, false, var);
/* Insert stmt into exit_bb. */
last_bsi = bsi_last (exit_bb);
if (stmt)
bsi_insert_before (&last_bsi, stmt, BSI_SAME_STMT);
/* Fix phi expressions in the successor bb. */
SET_PHI_ARG_DEF (phi1, update_e->dest_idx, ni_name);
}
}
/* Function vect_do_peeling_for_loop_bound
Peel the last iterations of the loop represented by LOOP_VINFO.
The peeled iterations form a new epilog loop. Given that the loop now
iterates NITERS times, the new epilog loop iterates
NITERS % VECTORIZATION_FACTOR times.
The original loop will later be made to iterate
NITERS / VECTORIZATION_FACTOR times (this value is placed into RATIO). */
static void
vect_do_peeling_for_loop_bound (loop_vec_info loop_vinfo, tree *ratio)
{
tree ni_name, ratio_mult_vf_name;
struct loop *loop = LOOP_VINFO_LOOP (loop_vinfo);
struct loop *new_loop;
edge update_e;
basic_block preheader;
int loop_num;
unsigned int th;
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "=== vect_do_peeling_for_loop_bound ===");
initialize_original_copy_tables ();
/* Generate the following variables on the preheader of original loop:
ni_name = number of iteration the original loop executes
ratio = ni_name / vf
ratio_mult_vf_name = ratio * vf */
vect_generate_tmps_on_preheader (loop_vinfo, &ni_name,
&ratio_mult_vf_name, ratio);
loop_num = loop->num;
/* Threshold for vectorized loop. */
th = (PARAM_VALUE (PARAM_MIN_VECT_LOOP_BOUND)) *
LOOP_VINFO_VECT_FACTOR (loop_vinfo);
new_loop = slpeel_tree_peel_loop_to_edge (loop, single_exit (loop),
ratio_mult_vf_name, ni_name, false, th);
gcc_assert (new_loop);
gcc_assert (loop_num == loop->num);
#ifdef ENABLE_CHECKING
slpeel_verify_cfg_after_peeling (loop, new_loop);
#endif
/* A guard that controls whether the new_loop is to be executed or skipped
is placed in LOOP->exit. LOOP->exit therefore has two successors - one
is the preheader of NEW_LOOP, where the IVs from LOOP are used. The other
is a bb after NEW_LOOP, where these IVs are not used. Find the edge that
is on the path where the LOOP IVs are used and need to be updated. */
preheader = loop_preheader_edge (new_loop)->src;
if (EDGE_PRED (preheader, 0)->src == single_exit (loop)->dest)
update_e = EDGE_PRED (preheader, 0);
else
update_e = EDGE_PRED (preheader, 1);
/* Update IVs of original loop as if they were advanced
by ratio_mult_vf_name steps. */
vect_update_ivs_after_vectorizer (loop_vinfo, ratio_mult_vf_name, update_e);
/* After peeling we have to reset scalar evolution analyzer. */
scev_reset ();
free_original_copy_tables ();
}
/* Function vect_gen_niters_for_prolog_loop
Set the number of iterations for the loop represented by LOOP_VINFO
to the minimum between LOOP_NITERS (the original iteration count of the loop)
and the misalignment of DR - the data reference recorded in
LOOP_VINFO_UNALIGNED_DR (LOOP_VINFO). As a result, after the execution of
this loop, the data reference DR will refer to an aligned location.
The following computation is generated:
If the misalignment of DR is known at compile time:
addr_mis = int mis = DR_MISALIGNMENT (dr);
Else, compute address misalignment in bytes:
addr_mis = addr & (vectype_size - 1)
prolog_niters = min ( LOOP_NITERS , (VF - addr_mis/elem_size)&(VF-1) )
(elem_size = element type size; an element is the scalar element
whose type is the inner type of the vectype)
For interleaving,
prolog_niters = min ( LOOP_NITERS ,
(VF/group_size - addr_mis/elem_size)&(VF/group_size-1) )
where group_size is the size of the interleaved group.
The above formulas assume that VF == number of elements in the vector. This
may not hold when there are multiple-types in the loop.
In this case, for some data-references in the loop the VF does not represent
the number of elements that fit in the vector. Therefore, instead of VF we
use TYPE_VECTOR_SUBPARTS. */
static tree
vect_gen_niters_for_prolog_loop (loop_vec_info loop_vinfo, tree loop_niters)
{
struct data_reference *dr = LOOP_VINFO_UNALIGNED_DR (loop_vinfo);
struct loop *loop = LOOP_VINFO_LOOP (loop_vinfo);
tree var, stmt;
tree iters, iters_name;
edge pe;
basic_block new_bb;
tree dr_stmt = DR_STMT (dr);
stmt_vec_info stmt_info = vinfo_for_stmt (dr_stmt);
tree vectype = STMT_VINFO_VECTYPE (stmt_info);
int vectype_align = TYPE_ALIGN (vectype) / BITS_PER_UNIT;
tree niters_type = TREE_TYPE (loop_niters);
int group_size = 1;
int element_size = GET_MODE_SIZE (TYPE_MODE (TREE_TYPE (DR_REF (dr))));
int nelements = TYPE_VECTOR_SUBPARTS (vectype);
if (DR_GROUP_FIRST_DR (stmt_info))
{
/* For interleaved access element size must be multiplied by the size of
the interleaved group. */
group_size = DR_GROUP_SIZE (vinfo_for_stmt (
DR_GROUP_FIRST_DR (stmt_info)));
element_size *= group_size;
}
pe = loop_preheader_edge (loop);
if (LOOP_PEELING_FOR_ALIGNMENT (loop_vinfo) > 0)
{
int byte_misalign = LOOP_PEELING_FOR_ALIGNMENT (loop_vinfo);
int elem_misalign = byte_misalign / element_size;
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "known alignment = %d.", byte_misalign);
iters = build_int_cst (niters_type,
(nelements - elem_misalign)&(nelements/group_size-1));
}
else
{
tree new_stmts = NULL_TREE;
tree start_addr =
vect_create_addr_base_for_vector_ref (dr_stmt, &new_stmts, NULL_TREE);
tree ptr_type = TREE_TYPE (start_addr);
tree size = TYPE_SIZE (ptr_type);
tree type = lang_hooks.types.type_for_size (tree_low_cst (size, 1), 1);
tree vectype_size_minus_1 = build_int_cst (type, vectype_align - 1);
tree elem_size_log =
build_int_cst (type, exact_log2 (vectype_align/nelements));
tree nelements_minus_1 = build_int_cst (type, nelements - 1);
tree nelements_tree = build_int_cst (type, nelements);
tree byte_misalign;
tree elem_misalign;
new_bb = bsi_insert_on_edge_immediate (pe, new_stmts);
gcc_assert (!new_bb);
/* Create: byte_misalign = addr & (vectype_size - 1) */
byte_misalign =
fold_build2 (BIT_AND_EXPR, type, start_addr, vectype_size_minus_1);
/* Create: elem_misalign = byte_misalign / element_size */
elem_misalign =
fold_build2 (RSHIFT_EXPR, type, byte_misalign, elem_size_log);
/* Create: (niters_type) (nelements - elem_misalign)&(nelements - 1) */
iters = fold_build2 (MINUS_EXPR, type, nelements_tree, elem_misalign);
iters = fold_build2 (BIT_AND_EXPR, type, iters, nelements_minus_1);
iters = fold_convert (niters_type, iters);
}
/* Create: prolog_loop_niters = min (iters, loop_niters) */
/* If the loop bound is known at compile time we already verified that it is
greater than vf; since the misalignment ('iters') is at most vf, there's
no need to generate the MIN_EXPR in this case. */
if (TREE_CODE (loop_niters) != INTEGER_CST)
iters = fold_build2 (MIN_EXPR, niters_type, iters, loop_niters);
if (vect_print_dump_info (REPORT_DETAILS))
{
fprintf (vect_dump, "niters for prolog loop: ");
print_generic_expr (vect_dump, iters, TDF_SLIM);
}
var = create_tmp_var (niters_type, "prolog_loop_niters");
add_referenced_var (var);
iters_name = force_gimple_operand (iters, &stmt, false, var);
/* Insert stmt on loop preheader edge. */
if (stmt)
{
basic_block new_bb = bsi_insert_on_edge_immediate (pe, stmt);
gcc_assert (!new_bb);
}
return iters_name;
}
/* Function vect_update_init_of_dr
NITERS iterations were peeled from LOOP. DR represents a data reference
in LOOP. This function updates the information recorded in DR to
account for the fact that the first NITERS iterations had already been
executed. Specifically, it updates the OFFSET field of DR. */
static void
vect_update_init_of_dr (struct data_reference *dr, tree niters)
{
tree offset = DR_OFFSET (dr);
niters = fold_build2 (MULT_EXPR, TREE_TYPE (niters), niters, DR_STEP (dr));
offset = fold_build2 (PLUS_EXPR, TREE_TYPE (offset), offset, niters);
DR_OFFSET (dr) = offset;
}
/* Function vect_update_inits_of_drs
NITERS iterations were peeled from the loop represented by LOOP_VINFO.
This function updates the information recorded for the data references in
the loop to account for the fact that the first NITERS iterations had
already been executed. Specifically, it updates the initial_condition of the
access_function of all the data_references in the loop. */
static void
vect_update_inits_of_drs (loop_vec_info loop_vinfo, tree niters)
{
unsigned int i;
VEC (data_reference_p, heap) *datarefs = LOOP_VINFO_DATAREFS (loop_vinfo);
struct data_reference *dr;
if (vect_dump && (dump_flags & TDF_DETAILS))
fprintf (vect_dump, "=== vect_update_inits_of_dr ===");
for (i = 0; VEC_iterate (data_reference_p, datarefs, i, dr); i++)
vect_update_init_of_dr (dr, niters);
}
/* Function vect_do_peeling_for_alignment
Peel the first 'niters' iterations of the loop represented by LOOP_VINFO.
'niters' is set to the misalignment of one of the data references in the
loop, thereby forcing it to refer to an aligned location at the beginning
of the execution of this loop. The data reference for which we are
peeling is recorded in LOOP_VINFO_UNALIGNED_DR. */
static void
vect_do_peeling_for_alignment (loop_vec_info loop_vinfo)
{
struct loop *loop = LOOP_VINFO_LOOP (loop_vinfo);
tree niters_of_prolog_loop, ni_name;
tree n_iters;
struct loop *new_loop;
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "=== vect_do_peeling_for_alignment ===");
initialize_original_copy_tables ();
ni_name = vect_build_loop_niters (loop_vinfo);
niters_of_prolog_loop = vect_gen_niters_for_prolog_loop (loop_vinfo, ni_name);
/* Peel the prolog loop and iterate it niters_of_prolog_loop. */
new_loop =
slpeel_tree_peel_loop_to_edge (loop, loop_preheader_edge (loop),
niters_of_prolog_loop, ni_name, true, 0);
gcc_assert (new_loop);
#ifdef ENABLE_CHECKING
slpeel_verify_cfg_after_peeling (new_loop, loop);
#endif
/* Update number of times loop executes. */
n_iters = LOOP_VINFO_NITERS (loop_vinfo);
LOOP_VINFO_NITERS (loop_vinfo) = fold_build2 (MINUS_EXPR,
TREE_TYPE (n_iters), n_iters, niters_of_prolog_loop);
/* Update the init conditions of the access functions of all data refs. */
vect_update_inits_of_drs (loop_vinfo, niters_of_prolog_loop);
/* After peeling we have to reset scalar evolution analyzer. */
scev_reset ();
free_original_copy_tables ();
}
/* Function vect_create_cond_for_align_checks.
Create a conditional expression that represents the alignment checks for
all of data references (array element references) whose alignment must be
checked at runtime.
Input:
LOOP_VINFO - two fields of the loop information are used.
LOOP_VINFO_PTR_MASK is the mask used to check the alignment.
LOOP_VINFO_MAY_MISALIGN_STMTS contains the refs to be checked.
Output:
COND_EXPR_STMT_LIST - statements needed to construct the conditional
expression.
The returned value is the conditional expression to be used in the if
statement that controls which version of the loop gets executed at runtime.
The algorithm makes two assumptions:
1) The number of bytes "n" in a vector is a power of 2.
2) An address "a" is aligned if a%n is zero and that this
test can be done as a&(n-1) == 0. For example, for 16
byte vectors the test is a&0xf == 0. */
static tree
vect_create_cond_for_align_checks (loop_vec_info loop_vinfo,
tree *cond_expr_stmt_list)
{
VEC(tree,heap) *may_misalign_stmts
= LOOP_VINFO_MAY_MISALIGN_STMTS (loop_vinfo);
tree ref_stmt, tmp;
int mask = LOOP_VINFO_PTR_MASK (loop_vinfo);
tree mask_cst;
unsigned int i;
tree psize;
tree int_ptrsize_type;
char tmp_name[20];
tree or_tmp_name = NULL_TREE;
tree and_tmp, and_tmp_name, and_stmt;
tree ptrsize_zero;
/* Check that mask is one less than a power of 2, i.e., mask is
all zeros followed by all ones. */
gcc_assert ((mask != 0) && ((mask & (mask+1)) == 0));
/* CHECKME: what is the best integer or unsigned type to use to hold a
cast from a pointer value? */
psize = TYPE_SIZE (ptr_type_node);
int_ptrsize_type
= lang_hooks.types.type_for_size (tree_low_cst (psize, 1), 0);
/* Create expression (mask & (dr_1 || ... || dr_n)) where dr_i is the address
of the first vector of the i'th data reference. */
for (i = 0; VEC_iterate (tree, may_misalign_stmts, i, ref_stmt); i++)
{
tree new_stmt_list = NULL_TREE;
tree addr_base;
tree addr_tmp, addr_tmp_name, addr_stmt;
tree or_tmp, new_or_tmp_name, or_stmt;
/* create: addr_tmp = (int)(address_of_first_vector) */
addr_base = vect_create_addr_base_for_vector_ref (ref_stmt,
&new_stmt_list,
NULL_TREE);
if (new_stmt_list != NULL_TREE)
append_to_statement_list_force (new_stmt_list, cond_expr_stmt_list);
sprintf (tmp_name, "%s%d", "addr2int", i);
addr_tmp = create_tmp_var (int_ptrsize_type, tmp_name);
add_referenced_var (addr_tmp);
addr_tmp_name = make_ssa_name (addr_tmp, NULL_TREE);
addr_stmt = fold_convert (int_ptrsize_type, addr_base);
addr_stmt = build_gimple_modify_stmt (addr_tmp_name, addr_stmt);
SSA_NAME_DEF_STMT (addr_tmp_name) = addr_stmt;
append_to_statement_list_force (addr_stmt, cond_expr_stmt_list);
/* The addresses are OR together. */
if (or_tmp_name != NULL_TREE)
{
/* create: or_tmp = or_tmp | addr_tmp */
sprintf (tmp_name, "%s%d", "orptrs", i);
or_tmp = create_tmp_var (int_ptrsize_type, tmp_name);
add_referenced_var (or_tmp);
new_or_tmp_name = make_ssa_name (or_tmp, NULL_TREE);
tmp = build2 (BIT_IOR_EXPR, int_ptrsize_type,
or_tmp_name, addr_tmp_name);
or_stmt = build_gimple_modify_stmt (new_or_tmp_name, tmp);
SSA_NAME_DEF_STMT (new_or_tmp_name) = or_stmt;
append_to_statement_list_force (or_stmt, cond_expr_stmt_list);
or_tmp_name = new_or_tmp_name;
}
else
or_tmp_name = addr_tmp_name;
} /* end for i */
mask_cst = build_int_cst (int_ptrsize_type, mask);
/* create: and_tmp = or_tmp & mask */
and_tmp = create_tmp_var (int_ptrsize_type, "andmask" );
add_referenced_var (and_tmp);
and_tmp_name = make_ssa_name (and_tmp, NULL_TREE);
tmp = build2 (BIT_AND_EXPR, int_ptrsize_type, or_tmp_name, mask_cst);
and_stmt = build_gimple_modify_stmt (and_tmp_name, tmp);
SSA_NAME_DEF_STMT (and_tmp_name) = and_stmt;
append_to_statement_list_force (and_stmt, cond_expr_stmt_list);
/* Make and_tmp the left operand of the conditional test against zero.
if and_tmp has a nonzero bit then some address is unaligned. */
ptrsize_zero = build_int_cst (int_ptrsize_type, 0);
return build2 (EQ_EXPR, boolean_type_node,
and_tmp_name, ptrsize_zero);
}
/* Function vect_transform_loop.
The analysis phase has determined that the loop is vectorizable.
Vectorize the loop - created vectorized stmts to replace the scalar
stmts in the loop, and update the loop exit condition. */
void
vect_transform_loop (loop_vec_info loop_vinfo)
{
struct loop *loop = LOOP_VINFO_LOOP (loop_vinfo);
basic_block *bbs = LOOP_VINFO_BBS (loop_vinfo);
int nbbs = loop->num_nodes;
block_stmt_iterator si, next_si;
int i;
tree ratio = NULL;
int vectorization_factor = LOOP_VINFO_VECT_FACTOR (loop_vinfo);
bool strided_store;
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "=== vec_transform_loop ===");
/* If the loop has data references that may or may not be aligned then
two versions of the loop need to be generated, one which is vectorized
and one which isn't. A test is then generated to control which of the
loops is executed. The test checks for the alignment of all of the
data references that may or may not be aligned. */
if (VEC_length (tree, LOOP_VINFO_MAY_MISALIGN_STMTS (loop_vinfo)))
{
struct loop *nloop;
tree cond_expr;
tree cond_expr_stmt_list = NULL_TREE;
basic_block condition_bb;
block_stmt_iterator cond_exp_bsi;
basic_block merge_bb;
basic_block new_exit_bb;
edge new_exit_e, e;
tree orig_phi, new_phi, arg;
unsigned prob = 4 * REG_BR_PROB_BASE / 5;
cond_expr = vect_create_cond_for_align_checks (loop_vinfo,
&cond_expr_stmt_list);
initialize_original_copy_tables ();
nloop = loop_version (loop, cond_expr, &condition_bb,
prob, prob, REG_BR_PROB_BASE - prob, true);
free_original_copy_tables();
/** Loop versioning violates an assumption we try to maintain during
vectorization - that the loop exit block has a single predecessor.
After versioning, the exit block of both loop versions is the same
basic block (i.e. it has two predecessors). Just in order to simplify
following transformations in the vectorizer, we fix this situation
here by adding a new (empty) block on the exit-edge of the loop,
with the proper loop-exit phis to maintain loop-closed-form. **/
merge_bb = single_exit (loop)->dest;
gcc_assert (EDGE_COUNT (merge_bb->preds) == 2);
new_exit_bb = split_edge (single_exit (loop));
new_exit_e = single_exit (loop);
e = EDGE_SUCC (new_exit_bb, 0);
for (orig_phi = phi_nodes (merge_bb); orig_phi;
orig_phi = PHI_CHAIN (orig_phi))
{
new_phi = create_phi_node (SSA_NAME_VAR (PHI_RESULT (orig_phi)),
new_exit_bb);
arg = PHI_ARG_DEF_FROM_EDGE (orig_phi, e);
add_phi_arg (new_phi, arg, new_exit_e);
SET_PHI_ARG_DEF (orig_phi, e->dest_idx, PHI_RESULT (new_phi));
}
/** end loop-exit-fixes after versioning **/
update_ssa (TODO_update_ssa);
cond_exp_bsi = bsi_last (condition_bb);
bsi_insert_before (&cond_exp_bsi, cond_expr_stmt_list, BSI_SAME_STMT);
}
/* CHECKME: we wouldn't need this if we called update_ssa once
for all loops. */
bitmap_zero (vect_memsyms_to_rename);
/* Peel the loop if there are data refs with unknown alignment.
Only one data ref with unknown store is allowed. */
if (LOOP_PEELING_FOR_ALIGNMENT (loop_vinfo))
vect_do_peeling_for_alignment (loop_vinfo);
/* If the loop has a symbolic number of iterations 'n' (i.e. it's not a
compile time constant), or it is a constant that doesn't divide by the
vectorization factor, then an epilog loop needs to be created.
We therefore duplicate the loop: the original loop will be vectorized,
and will compute the first (n/VF) iterations. The second copy of the loop
will remain scalar and will compute the remaining (n%VF) iterations.
(VF is the vectorization factor). */
if (!LOOP_VINFO_NITERS_KNOWN_P (loop_vinfo)
|| (LOOP_VINFO_NITERS_KNOWN_P (loop_vinfo)
&& LOOP_VINFO_INT_NITERS (loop_vinfo) % vectorization_factor != 0))
vect_do_peeling_for_loop_bound (loop_vinfo, &ratio);
else
ratio = build_int_cst (TREE_TYPE (LOOP_VINFO_NITERS (loop_vinfo)),
LOOP_VINFO_INT_NITERS (loop_vinfo) / vectorization_factor);
/* 1) Make sure the loop header has exactly two entries
2) Make sure we have a preheader basic block. */
gcc_assert (EDGE_COUNT (loop->header->preds) == 2);
split_edge (loop_preheader_edge (loop));
/* FORNOW: the vectorizer supports only loops which body consist
of one basic block (header + empty latch). When the vectorizer will
support more involved loop forms, the order by which the BBs are
traversed need to be reconsidered. */
for (i = 0; i < nbbs; i++)
{
basic_block bb = bbs[i];
stmt_vec_info stmt_info;
tree phi;
for (phi = phi_nodes (bb); phi; phi = PHI_CHAIN (phi))
{
if (vect_print_dump_info (REPORT_DETAILS))
{
fprintf (vect_dump, "------>vectorizing phi: ");
print_generic_expr (vect_dump, phi, TDF_SLIM);
}
stmt_info = vinfo_for_stmt (phi);
if (!stmt_info)
continue;
if (!STMT_VINFO_RELEVANT_P (stmt_info)
&& !STMT_VINFO_LIVE_P (stmt_info))
continue;
if ((TYPE_VECTOR_SUBPARTS (STMT_VINFO_VECTYPE (stmt_info))
!= (unsigned HOST_WIDE_INT) vectorization_factor)
&& vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "multiple-types.");
if (STMT_VINFO_DEF_TYPE (stmt_info) == vect_induction_def)
{
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "transform phi.");
vect_transform_stmt (phi, NULL, NULL);
}
}
for (si = bsi_start (bb); !bsi_end_p (si);)
{
tree stmt = bsi_stmt (si);
bool is_store;
if (vect_print_dump_info (REPORT_DETAILS))
{
fprintf (vect_dump, "------>vectorizing statement: ");
print_generic_expr (vect_dump, stmt, TDF_SLIM);
}
stmt_info = vinfo_for_stmt (stmt);
gcc_assert (stmt_info);
if (!STMT_VINFO_RELEVANT_P (stmt_info)
&& !STMT_VINFO_LIVE_P (stmt_info))
{
bsi_next (&si);
continue;
}
gcc_assert (STMT_VINFO_VECTYPE (stmt_info));
if ((TYPE_VECTOR_SUBPARTS (STMT_VINFO_VECTYPE (stmt_info))
!= (unsigned HOST_WIDE_INT) vectorization_factor)
&& vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "multiple-types.");
/* -------- vectorize statement ------------ */
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "transform statement.");
strided_store = false;
is_store = vect_transform_stmt (stmt, &si, &strided_store);
if (is_store)
{
stmt_ann_t ann;
if (DR_GROUP_FIRST_DR (stmt_info))
{
/* Interleaving. If IS_STORE is TRUE, the vectorization of the
interleaving chain was completed - free all the stores in
the chain. */
tree next = DR_GROUP_FIRST_DR (stmt_info);
tree tmp;
stmt_vec_info next_stmt_info;
while (next)
{
next_si = bsi_for_stmt (next);
next_stmt_info = vinfo_for_stmt (next);
/* Free the attached stmt_vec_info and remove the stmt. */
ann = stmt_ann (next);
tmp = DR_GROUP_NEXT_DR (next_stmt_info);
free (next_stmt_info);
set_stmt_info (ann, NULL);
bsi_remove (&next_si, true);
next = tmp;
}
bsi_remove (&si, true);
continue;
}
else
{
/* Free the attached stmt_vec_info and remove the stmt. */
ann = stmt_ann (stmt);
free (stmt_info);
set_stmt_info (ann, NULL);
bsi_remove (&si, true);
continue;
}
}
bsi_next (&si);
} /* stmts in BB */
} /* BBs in loop */
slpeel_make_loop_iterate_ntimes (loop, ratio);
mark_set_for_renaming (vect_memsyms_to_rename);
/* The memory tags and pointers in vectorized statements need to
have their SSA forms updated. FIXME, why can't this be delayed
until all the loops have been transformed? */
update_ssa (TODO_update_ssa);
if (vect_print_dump_info (REPORT_VECTORIZED_LOOPS))
fprintf (vect_dump, "LOOP VECTORIZED.");
}
|