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
|
/* Instruction scheduling pass. This file computes dependencies between
instructions.
Copyright (C) 1992-2018 Free Software Foundation, Inc.
Contributed by Michael Tiemann (tiemann@cygnus.com) Enhanced by,
and currently maintained by, Jim Wilson (wilson@cygnus.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 3, 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 COPYING3. If not see
<http://www.gnu.org/licenses/>. */
#include "config.h"
#include "system.h"
#include "coretypes.h"
#include "backend.h"
#include "target.h"
#include "rtl.h"
#include "tree.h"
#include "df.h"
#include "insn-config.h"
#include "regs.h"
#include "memmodel.h"
#include "ira.h"
#include "ira-int.h"
#include "insn-attr.h"
#include "cfgbuild.h"
#include "sched-int.h"
#include "params.h"
#include "cselib.h"
#ifdef INSN_SCHEDULING
/* Holds current parameters for the dependency analyzer. */
struct sched_deps_info_def *sched_deps_info;
/* The data is specific to the Haifa scheduler. */
vec<haifa_deps_insn_data_def>
h_d_i_d = vNULL;
/* Return the major type present in the DS. */
enum reg_note
ds_to_dk (ds_t ds)
{
if (ds & DEP_TRUE)
return REG_DEP_TRUE;
if (ds & DEP_OUTPUT)
return REG_DEP_OUTPUT;
if (ds & DEP_CONTROL)
return REG_DEP_CONTROL;
gcc_assert (ds & DEP_ANTI);
return REG_DEP_ANTI;
}
/* Return equivalent dep_status. */
ds_t
dk_to_ds (enum reg_note dk)
{
switch (dk)
{
case REG_DEP_TRUE:
return DEP_TRUE;
case REG_DEP_OUTPUT:
return DEP_OUTPUT;
case REG_DEP_CONTROL:
return DEP_CONTROL;
default:
gcc_assert (dk == REG_DEP_ANTI);
return DEP_ANTI;
}
}
/* Functions to operate with dependence information container - dep_t. */
/* Init DEP with the arguments. */
void
init_dep_1 (dep_t dep, rtx_insn *pro, rtx_insn *con, enum reg_note type, ds_t ds)
{
DEP_PRO (dep) = pro;
DEP_CON (dep) = con;
DEP_TYPE (dep) = type;
DEP_STATUS (dep) = ds;
DEP_COST (dep) = UNKNOWN_DEP_COST;
DEP_NONREG (dep) = 0;
DEP_MULTIPLE (dep) = 0;
DEP_REPLACE (dep) = NULL;
}
/* Init DEP with the arguments.
While most of the scheduler (including targets) only need the major type
of the dependency, it is convenient to hide full dep_status from them. */
void
init_dep (dep_t dep, rtx_insn *pro, rtx_insn *con, enum reg_note kind)
{
ds_t ds;
if ((current_sched_info->flags & USE_DEPS_LIST))
ds = dk_to_ds (kind);
else
ds = 0;
init_dep_1 (dep, pro, con, kind, ds);
}
/* Make a copy of FROM in TO. */
static void
copy_dep (dep_t to, dep_t from)
{
memcpy (to, from, sizeof (*to));
}
static void dump_ds (FILE *, ds_t);
/* Define flags for dump_dep (). */
/* Dump producer of the dependence. */
#define DUMP_DEP_PRO (2)
/* Dump consumer of the dependence. */
#define DUMP_DEP_CON (4)
/* Dump type of the dependence. */
#define DUMP_DEP_TYPE (8)
/* Dump status of the dependence. */
#define DUMP_DEP_STATUS (16)
/* Dump all information about the dependence. */
#define DUMP_DEP_ALL (DUMP_DEP_PRO | DUMP_DEP_CON | DUMP_DEP_TYPE \
|DUMP_DEP_STATUS)
/* Dump DEP to DUMP.
FLAGS is a bit mask specifying what information about DEP needs
to be printed.
If FLAGS has the very first bit set, then dump all information about DEP
and propagate this bit into the callee dump functions. */
static void
dump_dep (FILE *dump, dep_t dep, int flags)
{
if (flags & 1)
flags |= DUMP_DEP_ALL;
fprintf (dump, "<");
if (flags & DUMP_DEP_PRO)
fprintf (dump, "%d; ", INSN_UID (DEP_PRO (dep)));
if (flags & DUMP_DEP_CON)
fprintf (dump, "%d; ", INSN_UID (DEP_CON (dep)));
if (flags & DUMP_DEP_TYPE)
{
char t;
enum reg_note type = DEP_TYPE (dep);
switch (type)
{
case REG_DEP_TRUE:
t = 't';
break;
case REG_DEP_OUTPUT:
t = 'o';
break;
case REG_DEP_CONTROL:
t = 'c';
break;
case REG_DEP_ANTI:
t = 'a';
break;
default:
gcc_unreachable ();
break;
}
fprintf (dump, "%c; ", t);
}
if (flags & DUMP_DEP_STATUS)
{
if (current_sched_info->flags & USE_DEPS_LIST)
dump_ds (dump, DEP_STATUS (dep));
}
fprintf (dump, ">");
}
/* Default flags for dump_dep (). */
static int dump_dep_flags = (DUMP_DEP_PRO | DUMP_DEP_CON);
/* Dump all fields of DEP to STDERR. */
void
sd_debug_dep (dep_t dep)
{
dump_dep (stderr, dep, 1);
fprintf (stderr, "\n");
}
/* Determine whether DEP is a dependency link of a non-debug insn on a
debug insn. */
static inline bool
depl_on_debug_p (dep_link_t dep)
{
return (DEBUG_INSN_P (DEP_LINK_PRO (dep))
&& !DEBUG_INSN_P (DEP_LINK_CON (dep)));
}
/* Functions to operate with a single link from the dependencies lists -
dep_link_t. */
/* Attach L to appear after link X whose &DEP_LINK_NEXT (X) is given by
PREV_NEXT_P. */
static void
attach_dep_link (dep_link_t l, dep_link_t *prev_nextp)
{
dep_link_t next = *prev_nextp;
gcc_assert (DEP_LINK_PREV_NEXTP (l) == NULL
&& DEP_LINK_NEXT (l) == NULL);
/* Init node being inserted. */
DEP_LINK_PREV_NEXTP (l) = prev_nextp;
DEP_LINK_NEXT (l) = next;
/* Fix next node. */
if (next != NULL)
{
gcc_assert (DEP_LINK_PREV_NEXTP (next) == prev_nextp);
DEP_LINK_PREV_NEXTP (next) = &DEP_LINK_NEXT (l);
}
/* Fix prev node. */
*prev_nextp = l;
}
/* Add dep_link LINK to deps_list L. */
static void
add_to_deps_list (dep_link_t link, deps_list_t l)
{
attach_dep_link (link, &DEPS_LIST_FIRST (l));
/* Don't count debug deps. */
if (!depl_on_debug_p (link))
++DEPS_LIST_N_LINKS (l);
}
/* Detach dep_link L from the list. */
static void
detach_dep_link (dep_link_t l)
{
dep_link_t *prev_nextp = DEP_LINK_PREV_NEXTP (l);
dep_link_t next = DEP_LINK_NEXT (l);
*prev_nextp = next;
if (next != NULL)
DEP_LINK_PREV_NEXTP (next) = prev_nextp;
DEP_LINK_PREV_NEXTP (l) = NULL;
DEP_LINK_NEXT (l) = NULL;
}
/* Remove link LINK from list LIST. */
static void
remove_from_deps_list (dep_link_t link, deps_list_t list)
{
detach_dep_link (link);
/* Don't count debug deps. */
if (!depl_on_debug_p (link))
--DEPS_LIST_N_LINKS (list);
}
/* Move link LINK from list FROM to list TO. */
static void
move_dep_link (dep_link_t link, deps_list_t from, deps_list_t to)
{
remove_from_deps_list (link, from);
add_to_deps_list (link, to);
}
/* Return true of LINK is not attached to any list. */
static bool
dep_link_is_detached_p (dep_link_t link)
{
return DEP_LINK_PREV_NEXTP (link) == NULL;
}
/* Pool to hold all dependency nodes (dep_node_t). */
static object_allocator<_dep_node> *dn_pool;
/* Number of dep_nodes out there. */
static int dn_pool_diff = 0;
/* Create a dep_node. */
static dep_node_t
create_dep_node (void)
{
dep_node_t n = dn_pool->allocate ();
dep_link_t back = DEP_NODE_BACK (n);
dep_link_t forw = DEP_NODE_FORW (n);
DEP_LINK_NODE (back) = n;
DEP_LINK_NEXT (back) = NULL;
DEP_LINK_PREV_NEXTP (back) = NULL;
DEP_LINK_NODE (forw) = n;
DEP_LINK_NEXT (forw) = NULL;
DEP_LINK_PREV_NEXTP (forw) = NULL;
++dn_pool_diff;
return n;
}
/* Delete dep_node N. N must not be connected to any deps_list. */
static void
delete_dep_node (dep_node_t n)
{
gcc_assert (dep_link_is_detached_p (DEP_NODE_BACK (n))
&& dep_link_is_detached_p (DEP_NODE_FORW (n)));
XDELETE (DEP_REPLACE (DEP_NODE_DEP (n)));
--dn_pool_diff;
dn_pool->remove (n);
}
/* Pool to hold dependencies lists (deps_list_t). */
static object_allocator<_deps_list> *dl_pool;
/* Number of deps_lists out there. */
static int dl_pool_diff = 0;
/* Functions to operate with dependences lists - deps_list_t. */
/* Return true if list L is empty. */
static bool
deps_list_empty_p (deps_list_t l)
{
return DEPS_LIST_N_LINKS (l) == 0;
}
/* Create a new deps_list. */
static deps_list_t
create_deps_list (void)
{
deps_list_t l = dl_pool->allocate ();
DEPS_LIST_FIRST (l) = NULL;
DEPS_LIST_N_LINKS (l) = 0;
++dl_pool_diff;
return l;
}
/* Free deps_list L. */
static void
free_deps_list (deps_list_t l)
{
gcc_assert (deps_list_empty_p (l));
--dl_pool_diff;
dl_pool->remove (l);
}
/* Return true if there is no dep_nodes and deps_lists out there.
After the region is scheduled all the dependency nodes and lists
should [generally] be returned to pool. */
bool
deps_pools_are_empty_p (void)
{
return dn_pool_diff == 0 && dl_pool_diff == 0;
}
/* Remove all elements from L. */
static void
clear_deps_list (deps_list_t l)
{
do
{
dep_link_t link = DEPS_LIST_FIRST (l);
if (link == NULL)
break;
remove_from_deps_list (link, l);
}
while (1);
}
/* Decide whether a dependency should be treated as a hard or a speculative
dependency. */
static bool
dep_spec_p (dep_t dep)
{
if (current_sched_info->flags & DO_SPECULATION)
{
if (DEP_STATUS (dep) & SPECULATIVE)
return true;
}
if (current_sched_info->flags & DO_PREDICATION)
{
if (DEP_TYPE (dep) == REG_DEP_CONTROL)
return true;
}
if (DEP_REPLACE (dep) != NULL)
return true;
return false;
}
static regset reg_pending_sets;
static regset reg_pending_clobbers;
static regset reg_pending_uses;
static regset reg_pending_control_uses;
static enum reg_pending_barrier_mode reg_pending_barrier;
/* Hard registers implicitly clobbered or used (or may be implicitly
clobbered or used) by the currently analyzed insn. For example,
insn in its constraint has one register class. Even if there is
currently no hard register in the insn, the particular hard
register will be in the insn after reload pass because the
constraint requires it. */
static HARD_REG_SET implicit_reg_pending_clobbers;
static HARD_REG_SET implicit_reg_pending_uses;
/* To speed up the test for duplicate dependency links we keep a
record of dependencies created by add_dependence when the average
number of instructions in a basic block is very large.
Studies have shown that there is typically around 5 instructions between
branches for typical C code. So we can make a guess that the average
basic block is approximately 5 instructions long; we will choose 100X
the average size as a very large basic block.
Each insn has associated bitmaps for its dependencies. Each bitmap
has enough entries to represent a dependency on any other insn in
the insn chain. All bitmap for true dependencies cache is
allocated then the rest two ones are also allocated. */
static bitmap_head *true_dependency_cache = NULL;
static bitmap_head *output_dependency_cache = NULL;
static bitmap_head *anti_dependency_cache = NULL;
static bitmap_head *control_dependency_cache = NULL;
static bitmap_head *spec_dependency_cache = NULL;
static int cache_size;
/* True if we should mark added dependencies as a non-register deps. */
static bool mark_as_hard;
static int deps_may_trap_p (const_rtx);
static void add_dependence_1 (rtx_insn *, rtx_insn *, enum reg_note);
static void add_dependence_list (rtx_insn *, rtx_insn_list *, int,
enum reg_note, bool);
static void add_dependence_list_and_free (struct deps_desc *, rtx_insn *,
rtx_insn_list **, int, enum reg_note,
bool);
static void delete_all_dependences (rtx_insn *);
static void chain_to_prev_insn (rtx_insn *);
static void flush_pending_lists (struct deps_desc *, rtx_insn *, int, int);
static void sched_analyze_1 (struct deps_desc *, rtx, rtx_insn *);
static void sched_analyze_2 (struct deps_desc *, rtx, rtx_insn *);
static void sched_analyze_insn (struct deps_desc *, rtx, rtx_insn *);
static bool sched_has_condition_p (const rtx_insn *);
static int conditions_mutex_p (const_rtx, const_rtx, bool, bool);
static enum DEPS_ADJUST_RESULT maybe_add_or_update_dep_1 (dep_t, bool,
rtx, rtx);
static enum DEPS_ADJUST_RESULT add_or_update_dep_1 (dep_t, bool, rtx, rtx);
static void check_dep (dep_t, bool);
/* Return nonzero if a load of the memory reference MEM can cause a trap. */
static int
deps_may_trap_p (const_rtx mem)
{
const_rtx addr = XEXP (mem, 0);
if (REG_P (addr) && REGNO (addr) >= FIRST_PSEUDO_REGISTER)
{
const_rtx t = get_reg_known_value (REGNO (addr));
if (t)
addr = t;
}
return rtx_addr_can_trap_p (addr);
}
/* Find the condition under which INSN is executed. If REV is not NULL,
it is set to TRUE when the returned comparison should be reversed
to get the actual condition. */
static rtx
sched_get_condition_with_rev_uncached (const rtx_insn *insn, bool *rev)
{
rtx pat = PATTERN (insn);
rtx src;
if (rev)
*rev = false;
if (GET_CODE (pat) == COND_EXEC)
return COND_EXEC_TEST (pat);
if (!any_condjump_p (insn) || !onlyjump_p (insn))
return 0;
src = SET_SRC (pc_set (insn));
if (XEXP (src, 2) == pc_rtx)
return XEXP (src, 0);
else if (XEXP (src, 1) == pc_rtx)
{
rtx cond = XEXP (src, 0);
enum rtx_code revcode = reversed_comparison_code (cond, insn);
if (revcode == UNKNOWN)
return 0;
if (rev)
*rev = true;
return cond;
}
return 0;
}
/* Return the condition under which INSN does not execute (i.e. the
not-taken condition for a conditional branch), or NULL if we cannot
find such a condition. The caller should make a copy of the condition
before using it. */
rtx
sched_get_reverse_condition_uncached (const rtx_insn *insn)
{
bool rev;
rtx cond = sched_get_condition_with_rev_uncached (insn, &rev);
if (cond == NULL_RTX)
return cond;
if (!rev)
{
enum rtx_code revcode = reversed_comparison_code (cond, insn);
cond = gen_rtx_fmt_ee (revcode, GET_MODE (cond),
XEXP (cond, 0),
XEXP (cond, 1));
}
return cond;
}
/* Caching variant of sched_get_condition_with_rev_uncached.
We only do actual work the first time we come here for an insn; the
results are cached in INSN_CACHED_COND and INSN_REVERSE_COND. */
static rtx
sched_get_condition_with_rev (const rtx_insn *insn, bool *rev)
{
bool tmp;
if (INSN_LUID (insn) == 0)
return sched_get_condition_with_rev_uncached (insn, rev);
if (INSN_CACHED_COND (insn) == const_true_rtx)
return NULL_RTX;
if (INSN_CACHED_COND (insn) != NULL_RTX)
{
if (rev)
*rev = INSN_REVERSE_COND (insn);
return INSN_CACHED_COND (insn);
}
INSN_CACHED_COND (insn) = sched_get_condition_with_rev_uncached (insn, &tmp);
INSN_REVERSE_COND (insn) = tmp;
if (INSN_CACHED_COND (insn) == NULL_RTX)
{
INSN_CACHED_COND (insn) = const_true_rtx;
return NULL_RTX;
}
if (rev)
*rev = INSN_REVERSE_COND (insn);
return INSN_CACHED_COND (insn);
}
/* True when we can find a condition under which INSN is executed. */
static bool
sched_has_condition_p (const rtx_insn *insn)
{
return !! sched_get_condition_with_rev (insn, NULL);
}
/* Return nonzero if conditions COND1 and COND2 can never be both true. */
static int
conditions_mutex_p (const_rtx cond1, const_rtx cond2, bool rev1, bool rev2)
{
if (COMPARISON_P (cond1)
&& COMPARISON_P (cond2)
&& GET_CODE (cond1) ==
(rev1==rev2
? reversed_comparison_code (cond2, NULL)
: GET_CODE (cond2))
&& rtx_equal_p (XEXP (cond1, 0), XEXP (cond2, 0))
&& XEXP (cond1, 1) == XEXP (cond2, 1))
return 1;
return 0;
}
/* Return true if insn1 and insn2 can never depend on one another because
the conditions under which they are executed are mutually exclusive. */
bool
sched_insns_conditions_mutex_p (const rtx_insn *insn1, const rtx_insn *insn2)
{
rtx cond1, cond2;
bool rev1 = false, rev2 = false;
/* df doesn't handle conditional lifetimes entirely correctly;
calls mess up the conditional lifetimes. */
if (!CALL_P (insn1) && !CALL_P (insn2))
{
cond1 = sched_get_condition_with_rev (insn1, &rev1);
cond2 = sched_get_condition_with_rev (insn2, &rev2);
if (cond1 && cond2
&& conditions_mutex_p (cond1, cond2, rev1, rev2)
/* Make sure first instruction doesn't affect condition of second
instruction if switched. */
&& !modified_in_p (cond1, insn2)
/* Make sure second instruction doesn't affect condition of first
instruction if switched. */
&& !modified_in_p (cond2, insn1))
return true;
}
return false;
}
/* Return true if INSN can potentially be speculated with type DS. */
bool
sched_insn_is_legitimate_for_speculation_p (const rtx_insn *insn, ds_t ds)
{
if (HAS_INTERNAL_DEP (insn))
return false;
if (!NONJUMP_INSN_P (insn))
return false;
if (SCHED_GROUP_P (insn))
return false;
if (IS_SPECULATION_CHECK_P (CONST_CAST_RTX_INSN (insn)))
return false;
if (side_effects_p (PATTERN (insn)))
return false;
if (ds & BE_IN_SPEC)
/* The following instructions, which depend on a speculatively scheduled
instruction, cannot be speculatively scheduled along. */
{
if (may_trap_or_fault_p (PATTERN (insn)))
/* If instruction might fault, it cannot be speculatively scheduled.
For control speculation it's obvious why and for data speculation
it's because the insn might get wrong input if speculation
wasn't successful. */
return false;
if ((ds & BE_IN_DATA)
&& sched_has_condition_p (insn))
/* If this is a predicated instruction, then it cannot be
speculatively scheduled. See PR35659. */
return false;
}
return true;
}
/* Initialize LIST_PTR to point to one of the lists present in TYPES_PTR,
initialize RESOLVED_P_PTR with true if that list consists of resolved deps,
and remove the type of returned [through LIST_PTR] list from TYPES_PTR.
This function is used to switch sd_iterator to the next list.
!!! For internal use only. Might consider moving it to sched-int.h. */
void
sd_next_list (const_rtx insn, sd_list_types_def *types_ptr,
deps_list_t *list_ptr, bool *resolved_p_ptr)
{
sd_list_types_def types = *types_ptr;
if (types & SD_LIST_HARD_BACK)
{
*list_ptr = INSN_HARD_BACK_DEPS (insn);
*resolved_p_ptr = false;
*types_ptr = types & ~SD_LIST_HARD_BACK;
}
else if (types & SD_LIST_SPEC_BACK)
{
*list_ptr = INSN_SPEC_BACK_DEPS (insn);
*resolved_p_ptr = false;
*types_ptr = types & ~SD_LIST_SPEC_BACK;
}
else if (types & SD_LIST_FORW)
{
*list_ptr = INSN_FORW_DEPS (insn);
*resolved_p_ptr = false;
*types_ptr = types & ~SD_LIST_FORW;
}
else if (types & SD_LIST_RES_BACK)
{
*list_ptr = INSN_RESOLVED_BACK_DEPS (insn);
*resolved_p_ptr = true;
*types_ptr = types & ~SD_LIST_RES_BACK;
}
else if (types & SD_LIST_RES_FORW)
{
*list_ptr = INSN_RESOLVED_FORW_DEPS (insn);
*resolved_p_ptr = true;
*types_ptr = types & ~SD_LIST_RES_FORW;
}
else
{
*list_ptr = NULL;
*resolved_p_ptr = false;
*types_ptr = SD_LIST_NONE;
}
}
/* Return the summary size of INSN's lists defined by LIST_TYPES. */
int
sd_lists_size (const_rtx insn, sd_list_types_def list_types)
{
int size = 0;
while (list_types != SD_LIST_NONE)
{
deps_list_t list;
bool resolved_p;
sd_next_list (insn, &list_types, &list, &resolved_p);
if (list)
size += DEPS_LIST_N_LINKS (list);
}
return size;
}
/* Return true if INSN's lists defined by LIST_TYPES are all empty. */
bool
sd_lists_empty_p (const_rtx insn, sd_list_types_def list_types)
{
while (list_types != SD_LIST_NONE)
{
deps_list_t list;
bool resolved_p;
sd_next_list (insn, &list_types, &list, &resolved_p);
if (!deps_list_empty_p (list))
return false;
}
return true;
}
/* Initialize data for INSN. */
void
sd_init_insn (rtx_insn *insn)
{
INSN_HARD_BACK_DEPS (insn) = create_deps_list ();
INSN_SPEC_BACK_DEPS (insn) = create_deps_list ();
INSN_RESOLVED_BACK_DEPS (insn) = create_deps_list ();
INSN_FORW_DEPS (insn) = create_deps_list ();
INSN_RESOLVED_FORW_DEPS (insn) = create_deps_list ();
/* ??? It would be nice to allocate dependency caches here. */
}
/* Free data for INSN. */
void
sd_finish_insn (rtx_insn *insn)
{
/* ??? It would be nice to deallocate dependency caches here. */
free_deps_list (INSN_HARD_BACK_DEPS (insn));
INSN_HARD_BACK_DEPS (insn) = NULL;
free_deps_list (INSN_SPEC_BACK_DEPS (insn));
INSN_SPEC_BACK_DEPS (insn) = NULL;
free_deps_list (INSN_RESOLVED_BACK_DEPS (insn));
INSN_RESOLVED_BACK_DEPS (insn) = NULL;
free_deps_list (INSN_FORW_DEPS (insn));
INSN_FORW_DEPS (insn) = NULL;
free_deps_list (INSN_RESOLVED_FORW_DEPS (insn));
INSN_RESOLVED_FORW_DEPS (insn) = NULL;
}
/* Find a dependency between producer PRO and consumer CON.
Search through resolved dependency lists if RESOLVED_P is true.
If no such dependency is found return NULL,
otherwise return the dependency and initialize SD_IT_PTR [if it is nonnull]
with an iterator pointing to it. */
static dep_t
sd_find_dep_between_no_cache (rtx pro, rtx con, bool resolved_p,
sd_iterator_def *sd_it_ptr)
{
sd_list_types_def pro_list_type;
sd_list_types_def con_list_type;
sd_iterator_def sd_it;
dep_t dep;
bool found_p = false;
if (resolved_p)
{
pro_list_type = SD_LIST_RES_FORW;
con_list_type = SD_LIST_RES_BACK;
}
else
{
pro_list_type = SD_LIST_FORW;
con_list_type = SD_LIST_BACK;
}
/* Walk through either back list of INSN or forw list of ELEM
depending on which one is shorter. */
if (sd_lists_size (con, con_list_type) < sd_lists_size (pro, pro_list_type))
{
/* Find the dep_link with producer PRO in consumer's back_deps. */
FOR_EACH_DEP (con, con_list_type, sd_it, dep)
if (DEP_PRO (dep) == pro)
{
found_p = true;
break;
}
}
else
{
/* Find the dep_link with consumer CON in producer's forw_deps. */
FOR_EACH_DEP (pro, pro_list_type, sd_it, dep)
if (DEP_CON (dep) == con)
{
found_p = true;
break;
}
}
if (found_p)
{
if (sd_it_ptr != NULL)
*sd_it_ptr = sd_it;
return dep;
}
return NULL;
}
/* Find a dependency between producer PRO and consumer CON.
Use dependency [if available] to check if dependency is present at all.
Search through resolved dependency lists if RESOLVED_P is true.
If the dependency or NULL if none found. */
dep_t
sd_find_dep_between (rtx pro, rtx con, bool resolved_p)
{
if (true_dependency_cache != NULL)
/* Avoiding the list walk below can cut compile times dramatically
for some code. */
{
int elem_luid = INSN_LUID (pro);
int insn_luid = INSN_LUID (con);
if (!bitmap_bit_p (&true_dependency_cache[insn_luid], elem_luid)
&& !bitmap_bit_p (&output_dependency_cache[insn_luid], elem_luid)
&& !bitmap_bit_p (&anti_dependency_cache[insn_luid], elem_luid)
&& !bitmap_bit_p (&control_dependency_cache[insn_luid], elem_luid))
return NULL;
}
return sd_find_dep_between_no_cache (pro, con, resolved_p, NULL);
}
/* Add or update a dependence described by DEP.
MEM1 and MEM2, if non-null, correspond to memory locations in case of
data speculation.
The function returns a value indicating if an old entry has been changed
or a new entry has been added to insn's backward deps.
This function merely checks if producer and consumer is the same insn
and doesn't create a dep in this case. Actual manipulation of
dependence data structures is performed in add_or_update_dep_1. */
static enum DEPS_ADJUST_RESULT
maybe_add_or_update_dep_1 (dep_t dep, bool resolved_p, rtx mem1, rtx mem2)
{
rtx_insn *elem = DEP_PRO (dep);
rtx_insn *insn = DEP_CON (dep);
gcc_assert (INSN_P (insn) && INSN_P (elem));
/* Don't depend an insn on itself. */
if (insn == elem)
{
if (sched_deps_info->generate_spec_deps)
/* INSN has an internal dependence, which we can't overcome. */
HAS_INTERNAL_DEP (insn) = 1;
return DEP_NODEP;
}
return add_or_update_dep_1 (dep, resolved_p, mem1, mem2);
}
/* Ask dependency caches what needs to be done for dependence DEP.
Return DEP_CREATED if new dependence should be created and there is no
need to try to find one searching the dependencies lists.
Return DEP_PRESENT if there already is a dependence described by DEP and
hence nothing is to be done.
Return DEP_CHANGED if there already is a dependence, but it should be
updated to incorporate additional information from DEP. */
static enum DEPS_ADJUST_RESULT
ask_dependency_caches (dep_t dep)
{
int elem_luid = INSN_LUID (DEP_PRO (dep));
int insn_luid = INSN_LUID (DEP_CON (dep));
gcc_assert (true_dependency_cache != NULL
&& output_dependency_cache != NULL
&& anti_dependency_cache != NULL
&& control_dependency_cache != NULL);
if (!(current_sched_info->flags & USE_DEPS_LIST))
{
enum reg_note present_dep_type;
if (bitmap_bit_p (&true_dependency_cache[insn_luid], elem_luid))
present_dep_type = REG_DEP_TRUE;
else if (bitmap_bit_p (&output_dependency_cache[insn_luid], elem_luid))
present_dep_type = REG_DEP_OUTPUT;
else if (bitmap_bit_p (&anti_dependency_cache[insn_luid], elem_luid))
present_dep_type = REG_DEP_ANTI;
else if (bitmap_bit_p (&control_dependency_cache[insn_luid], elem_luid))
present_dep_type = REG_DEP_CONTROL;
else
/* There is no existing dep so it should be created. */
return DEP_CREATED;
if ((int) DEP_TYPE (dep) >= (int) present_dep_type)
/* DEP does not add anything to the existing dependence. */
return DEP_PRESENT;
}
else
{
ds_t present_dep_types = 0;
if (bitmap_bit_p (&true_dependency_cache[insn_luid], elem_luid))
present_dep_types |= DEP_TRUE;
if (bitmap_bit_p (&output_dependency_cache[insn_luid], elem_luid))
present_dep_types |= DEP_OUTPUT;
if (bitmap_bit_p (&anti_dependency_cache[insn_luid], elem_luid))
present_dep_types |= DEP_ANTI;
if (bitmap_bit_p (&control_dependency_cache[insn_luid], elem_luid))
present_dep_types |= DEP_CONTROL;
if (present_dep_types == 0)
/* There is no existing dep so it should be created. */
return DEP_CREATED;
if (!(current_sched_info->flags & DO_SPECULATION)
|| !bitmap_bit_p (&spec_dependency_cache[insn_luid], elem_luid))
{
if ((present_dep_types | (DEP_STATUS (dep) & DEP_TYPES))
== present_dep_types)
/* DEP does not add anything to the existing dependence. */
return DEP_PRESENT;
}
else
{
/* Only true dependencies can be data speculative and
only anti dependencies can be control speculative. */
gcc_assert ((present_dep_types & (DEP_TRUE | DEP_ANTI))
== present_dep_types);
/* if (DEP is SPECULATIVE) then
..we should update DEP_STATUS
else
..we should reset existing dep to non-speculative. */
}
}
return DEP_CHANGED;
}
/* Set dependency caches according to DEP. */
static void
set_dependency_caches (dep_t dep)
{
int elem_luid = INSN_LUID (DEP_PRO (dep));
int insn_luid = INSN_LUID (DEP_CON (dep));
if (!(current_sched_info->flags & USE_DEPS_LIST))
{
switch (DEP_TYPE (dep))
{
case REG_DEP_TRUE:
bitmap_set_bit (&true_dependency_cache[insn_luid], elem_luid);
break;
case REG_DEP_OUTPUT:
bitmap_set_bit (&output_dependency_cache[insn_luid], elem_luid);
break;
case REG_DEP_ANTI:
bitmap_set_bit (&anti_dependency_cache[insn_luid], elem_luid);
break;
case REG_DEP_CONTROL:
bitmap_set_bit (&control_dependency_cache[insn_luid], elem_luid);
break;
default:
gcc_unreachable ();
}
}
else
{
ds_t ds = DEP_STATUS (dep);
if (ds & DEP_TRUE)
bitmap_set_bit (&true_dependency_cache[insn_luid], elem_luid);
if (ds & DEP_OUTPUT)
bitmap_set_bit (&output_dependency_cache[insn_luid], elem_luid);
if (ds & DEP_ANTI)
bitmap_set_bit (&anti_dependency_cache[insn_luid], elem_luid);
if (ds & DEP_CONTROL)
bitmap_set_bit (&control_dependency_cache[insn_luid], elem_luid);
if (ds & SPECULATIVE)
{
gcc_assert (current_sched_info->flags & DO_SPECULATION);
bitmap_set_bit (&spec_dependency_cache[insn_luid], elem_luid);
}
}
}
/* Type of dependence DEP have changed from OLD_TYPE. Update dependency
caches accordingly. */
static void
update_dependency_caches (dep_t dep, enum reg_note old_type)
{
int elem_luid = INSN_LUID (DEP_PRO (dep));
int insn_luid = INSN_LUID (DEP_CON (dep));
/* Clear corresponding cache entry because type of the link
may have changed. Keep them if we use_deps_list. */
if (!(current_sched_info->flags & USE_DEPS_LIST))
{
switch (old_type)
{
case REG_DEP_OUTPUT:
bitmap_clear_bit (&output_dependency_cache[insn_luid], elem_luid);
break;
case REG_DEP_ANTI:
bitmap_clear_bit (&anti_dependency_cache[insn_luid], elem_luid);
break;
case REG_DEP_CONTROL:
bitmap_clear_bit (&control_dependency_cache[insn_luid], elem_luid);
break;
default:
gcc_unreachable ();
}
}
set_dependency_caches (dep);
}
/* Convert a dependence pointed to by SD_IT to be non-speculative. */
static void
change_spec_dep_to_hard (sd_iterator_def sd_it)
{
dep_node_t node = DEP_LINK_NODE (*sd_it.linkp);
dep_link_t link = DEP_NODE_BACK (node);
dep_t dep = DEP_NODE_DEP (node);
rtx_insn *elem = DEP_PRO (dep);
rtx_insn *insn = DEP_CON (dep);
move_dep_link (link, INSN_SPEC_BACK_DEPS (insn), INSN_HARD_BACK_DEPS (insn));
DEP_STATUS (dep) &= ~SPECULATIVE;
if (true_dependency_cache != NULL)
/* Clear the cache entry. */
bitmap_clear_bit (&spec_dependency_cache[INSN_LUID (insn)],
INSN_LUID (elem));
}
/* Update DEP to incorporate information from NEW_DEP.
SD_IT points to DEP in case it should be moved to another list.
MEM1 and MEM2, if nonnull, correspond to memory locations in case if
data-speculative dependence should be updated. */
static enum DEPS_ADJUST_RESULT
update_dep (dep_t dep, dep_t new_dep,
sd_iterator_def sd_it ATTRIBUTE_UNUSED,
rtx mem1 ATTRIBUTE_UNUSED,
rtx mem2 ATTRIBUTE_UNUSED)
{
enum DEPS_ADJUST_RESULT res = DEP_PRESENT;
enum reg_note old_type = DEP_TYPE (dep);
bool was_spec = dep_spec_p (dep);
DEP_NONREG (dep) |= DEP_NONREG (new_dep);
DEP_MULTIPLE (dep) = 1;
/* If this is a more restrictive type of dependence than the
existing one, then change the existing dependence to this
type. */
if ((int) DEP_TYPE (new_dep) < (int) old_type)
{
DEP_TYPE (dep) = DEP_TYPE (new_dep);
res = DEP_CHANGED;
}
if (current_sched_info->flags & USE_DEPS_LIST)
/* Update DEP_STATUS. */
{
ds_t dep_status = DEP_STATUS (dep);
ds_t ds = DEP_STATUS (new_dep);
ds_t new_status = ds | dep_status;
if (new_status & SPECULATIVE)
{
/* Either existing dep or a dep we're adding or both are
speculative. */
if (!(ds & SPECULATIVE)
|| !(dep_status & SPECULATIVE))
/* The new dep can't be speculative. */
new_status &= ~SPECULATIVE;
else
{
/* Both are speculative. Merge probabilities. */
if (mem1 != NULL)
{
dw_t dw;
dw = estimate_dep_weak (mem1, mem2);
ds = set_dep_weak (ds, BEGIN_DATA, dw);
}
new_status = ds_merge (dep_status, ds);
}
}
ds = new_status;
if (dep_status != ds)
{
DEP_STATUS (dep) = ds;
res = DEP_CHANGED;
}
}
if (was_spec && !dep_spec_p (dep))
/* The old dep was speculative, but now it isn't. */
change_spec_dep_to_hard (sd_it);
if (true_dependency_cache != NULL
&& res == DEP_CHANGED)
update_dependency_caches (dep, old_type);
return res;
}
/* Add or update a dependence described by DEP.
MEM1 and MEM2, if non-null, correspond to memory locations in case of
data speculation.
The function returns a value indicating if an old entry has been changed
or a new entry has been added to insn's backward deps or nothing has
been updated at all. */
static enum DEPS_ADJUST_RESULT
add_or_update_dep_1 (dep_t new_dep, bool resolved_p,
rtx mem1 ATTRIBUTE_UNUSED, rtx mem2 ATTRIBUTE_UNUSED)
{
bool maybe_present_p = true;
bool present_p = false;
gcc_assert (INSN_P (DEP_PRO (new_dep)) && INSN_P (DEP_CON (new_dep))
&& DEP_PRO (new_dep) != DEP_CON (new_dep));
if (flag_checking)
check_dep (new_dep, mem1 != NULL);
if (true_dependency_cache != NULL)
{
switch (ask_dependency_caches (new_dep))
{
case DEP_PRESENT:
dep_t present_dep;
sd_iterator_def sd_it;
present_dep = sd_find_dep_between_no_cache (DEP_PRO (new_dep),
DEP_CON (new_dep),
resolved_p, &sd_it);
DEP_MULTIPLE (present_dep) = 1;
return DEP_PRESENT;
case DEP_CHANGED:
maybe_present_p = true;
present_p = true;
break;
case DEP_CREATED:
maybe_present_p = false;
present_p = false;
break;
default:
gcc_unreachable ();
break;
}
}
/* Check that we don't already have this dependence. */
if (maybe_present_p)
{
dep_t present_dep;
sd_iterator_def sd_it;
gcc_assert (true_dependency_cache == NULL || present_p);
present_dep = sd_find_dep_between_no_cache (DEP_PRO (new_dep),
DEP_CON (new_dep),
resolved_p, &sd_it);
if (present_dep != NULL)
/* We found an existing dependency between ELEM and INSN. */
return update_dep (present_dep, new_dep, sd_it, mem1, mem2);
else
/* We didn't find a dep, it shouldn't present in the cache. */
gcc_assert (!present_p);
}
/* Might want to check one level of transitivity to save conses.
This check should be done in maybe_add_or_update_dep_1.
Since we made it to add_or_update_dep_1, we must create
(or update) a link. */
if (mem1 != NULL_RTX)
{
gcc_assert (sched_deps_info->generate_spec_deps);
DEP_STATUS (new_dep) = set_dep_weak (DEP_STATUS (new_dep), BEGIN_DATA,
estimate_dep_weak (mem1, mem2));
}
sd_add_dep (new_dep, resolved_p);
return DEP_CREATED;
}
/* Initialize BACK_LIST_PTR with consumer's backward list and
FORW_LIST_PTR with producer's forward list. If RESOLVED_P is true
initialize with lists that hold resolved deps. */
static void
get_back_and_forw_lists (dep_t dep, bool resolved_p,
deps_list_t *back_list_ptr,
deps_list_t *forw_list_ptr)
{
rtx_insn *con = DEP_CON (dep);
if (!resolved_p)
{
if (dep_spec_p (dep))
*back_list_ptr = INSN_SPEC_BACK_DEPS (con);
else
*back_list_ptr = INSN_HARD_BACK_DEPS (con);
*forw_list_ptr = INSN_FORW_DEPS (DEP_PRO (dep));
}
else
{
*back_list_ptr = INSN_RESOLVED_BACK_DEPS (con);
*forw_list_ptr = INSN_RESOLVED_FORW_DEPS (DEP_PRO (dep));
}
}
/* Add dependence described by DEP.
If RESOLVED_P is true treat the dependence as a resolved one. */
void
sd_add_dep (dep_t dep, bool resolved_p)
{
dep_node_t n = create_dep_node ();
deps_list_t con_back_deps;
deps_list_t pro_forw_deps;
rtx_insn *elem = DEP_PRO (dep);
rtx_insn *insn = DEP_CON (dep);
gcc_assert (INSN_P (insn) && INSN_P (elem) && insn != elem);
if ((current_sched_info->flags & DO_SPECULATION) == 0
|| !sched_insn_is_legitimate_for_speculation_p (insn, DEP_STATUS (dep)))
DEP_STATUS (dep) &= ~SPECULATIVE;
copy_dep (DEP_NODE_DEP (n), dep);
get_back_and_forw_lists (dep, resolved_p, &con_back_deps, &pro_forw_deps);
add_to_deps_list (DEP_NODE_BACK (n), con_back_deps);
if (flag_checking)
check_dep (dep, false);
add_to_deps_list (DEP_NODE_FORW (n), pro_forw_deps);
/* If we are adding a dependency to INSN's LOG_LINKs, then note that
in the bitmap caches of dependency information. */
if (true_dependency_cache != NULL)
set_dependency_caches (dep);
}
/* Add or update backward dependence between INSN and ELEM
with given type DEP_TYPE and dep_status DS.
This function is a convenience wrapper. */
enum DEPS_ADJUST_RESULT
sd_add_or_update_dep (dep_t dep, bool resolved_p)
{
return add_or_update_dep_1 (dep, resolved_p, NULL_RTX, NULL_RTX);
}
/* Resolved dependence pointed to by SD_IT.
SD_IT will advance to the next element. */
void
sd_resolve_dep (sd_iterator_def sd_it)
{
dep_node_t node = DEP_LINK_NODE (*sd_it.linkp);
dep_t dep = DEP_NODE_DEP (node);
rtx_insn *pro = DEP_PRO (dep);
rtx_insn *con = DEP_CON (dep);
if (dep_spec_p (dep))
move_dep_link (DEP_NODE_BACK (node), INSN_SPEC_BACK_DEPS (con),
INSN_RESOLVED_BACK_DEPS (con));
else
move_dep_link (DEP_NODE_BACK (node), INSN_HARD_BACK_DEPS (con),
INSN_RESOLVED_BACK_DEPS (con));
move_dep_link (DEP_NODE_FORW (node), INSN_FORW_DEPS (pro),
INSN_RESOLVED_FORW_DEPS (pro));
}
/* Perform the inverse operation of sd_resolve_dep. Restore the dependence
pointed to by SD_IT to unresolved state. */
void
sd_unresolve_dep (sd_iterator_def sd_it)
{
dep_node_t node = DEP_LINK_NODE (*sd_it.linkp);
dep_t dep = DEP_NODE_DEP (node);
rtx_insn *pro = DEP_PRO (dep);
rtx_insn *con = DEP_CON (dep);
if (dep_spec_p (dep))
move_dep_link (DEP_NODE_BACK (node), INSN_RESOLVED_BACK_DEPS (con),
INSN_SPEC_BACK_DEPS (con));
else
move_dep_link (DEP_NODE_BACK (node), INSN_RESOLVED_BACK_DEPS (con),
INSN_HARD_BACK_DEPS (con));
move_dep_link (DEP_NODE_FORW (node), INSN_RESOLVED_FORW_DEPS (pro),
INSN_FORW_DEPS (pro));
}
/* Make TO depend on all the FROM's producers.
If RESOLVED_P is true add dependencies to the resolved lists. */
void
sd_copy_back_deps (rtx_insn *to, rtx_insn *from, bool resolved_p)
{
sd_list_types_def list_type;
sd_iterator_def sd_it;
dep_t dep;
list_type = resolved_p ? SD_LIST_RES_BACK : SD_LIST_BACK;
FOR_EACH_DEP (from, list_type, sd_it, dep)
{
dep_def _new_dep, *new_dep = &_new_dep;
copy_dep (new_dep, dep);
DEP_CON (new_dep) = to;
sd_add_dep (new_dep, resolved_p);
}
}
/* Remove a dependency referred to by SD_IT.
SD_IT will point to the next dependence after removal. */
void
sd_delete_dep (sd_iterator_def sd_it)
{
dep_node_t n = DEP_LINK_NODE (*sd_it.linkp);
dep_t dep = DEP_NODE_DEP (n);
rtx_insn *pro = DEP_PRO (dep);
rtx_insn *con = DEP_CON (dep);
deps_list_t con_back_deps;
deps_list_t pro_forw_deps;
if (true_dependency_cache != NULL)
{
int elem_luid = INSN_LUID (pro);
int insn_luid = INSN_LUID (con);
bitmap_clear_bit (&true_dependency_cache[insn_luid], elem_luid);
bitmap_clear_bit (&anti_dependency_cache[insn_luid], elem_luid);
bitmap_clear_bit (&control_dependency_cache[insn_luid], elem_luid);
bitmap_clear_bit (&output_dependency_cache[insn_luid], elem_luid);
if (current_sched_info->flags & DO_SPECULATION)
bitmap_clear_bit (&spec_dependency_cache[insn_luid], elem_luid);
}
get_back_and_forw_lists (dep, sd_it.resolved_p,
&con_back_deps, &pro_forw_deps);
remove_from_deps_list (DEP_NODE_BACK (n), con_back_deps);
remove_from_deps_list (DEP_NODE_FORW (n), pro_forw_deps);
delete_dep_node (n);
}
/* Dump size of the lists. */
#define DUMP_LISTS_SIZE (2)
/* Dump dependencies of the lists. */
#define DUMP_LISTS_DEPS (4)
/* Dump all information about the lists. */
#define DUMP_LISTS_ALL (DUMP_LISTS_SIZE | DUMP_LISTS_DEPS)
/* Dump deps_lists of INSN specified by TYPES to DUMP.
FLAGS is a bit mask specifying what information about the lists needs
to be printed.
If FLAGS has the very first bit set, then dump all information about
the lists and propagate this bit into the callee dump functions. */
static void
dump_lists (FILE *dump, rtx insn, sd_list_types_def types, int flags)
{
sd_iterator_def sd_it;
dep_t dep;
int all;
all = (flags & 1);
if (all)
flags |= DUMP_LISTS_ALL;
fprintf (dump, "[");
if (flags & DUMP_LISTS_SIZE)
fprintf (dump, "%d; ", sd_lists_size (insn, types));
if (flags & DUMP_LISTS_DEPS)
{
FOR_EACH_DEP (insn, types, sd_it, dep)
{
dump_dep (dump, dep, dump_dep_flags | all);
fprintf (dump, " ");
}
}
}
/* Dump all information about deps_lists of INSN specified by TYPES
to STDERR. */
void
sd_debug_lists (rtx insn, sd_list_types_def types)
{
dump_lists (stderr, insn, types, 1);
fprintf (stderr, "\n");
}
/* A wrapper around add_dependence_1, to add a dependence of CON on
PRO, with type DEP_TYPE. This function implements special handling
for REG_DEP_CONTROL dependencies. For these, we optionally promote
the type to REG_DEP_ANTI if we can determine that predication is
impossible; otherwise we add additional true dependencies on the
INSN_COND_DEPS list of the jump (which PRO must be). */
void
add_dependence (rtx_insn *con, rtx_insn *pro, enum reg_note dep_type)
{
if (dep_type == REG_DEP_CONTROL
&& !(current_sched_info->flags & DO_PREDICATION))
dep_type = REG_DEP_ANTI;
/* A REG_DEP_CONTROL dependence may be eliminated through predication,
so we must also make the insn dependent on the setter of the
condition. */
if (dep_type == REG_DEP_CONTROL)
{
rtx_insn *real_pro = pro;
rtx_insn *other = real_insn_for_shadow (real_pro);
rtx cond;
if (other != NULL_RTX)
real_pro = other;
cond = sched_get_reverse_condition_uncached (real_pro);
/* Verify that the insn does not use a different value in
the condition register than the one that was present at
the jump. */
if (cond == NULL_RTX)
dep_type = REG_DEP_ANTI;
else if (INSN_CACHED_COND (real_pro) == const_true_rtx)
{
HARD_REG_SET uses;
CLEAR_HARD_REG_SET (uses);
note_uses (&PATTERN (con), record_hard_reg_uses, &uses);
if (TEST_HARD_REG_BIT (uses, REGNO (XEXP (cond, 0))))
dep_type = REG_DEP_ANTI;
}
if (dep_type == REG_DEP_CONTROL)
{
if (sched_verbose >= 5)
fprintf (sched_dump, "making DEP_CONTROL for %d\n",
INSN_UID (real_pro));
add_dependence_list (con, INSN_COND_DEPS (real_pro), 0,
REG_DEP_TRUE, false);
}
}
add_dependence_1 (con, pro, dep_type);
}
/* A convenience wrapper to operate on an entire list. HARD should be
true if DEP_NONREG should be set on newly created dependencies. */
static void
add_dependence_list (rtx_insn *insn, rtx_insn_list *list, int uncond,
enum reg_note dep_type, bool hard)
{
mark_as_hard = hard;
for (; list; list = list->next ())
{
if (uncond || ! sched_insns_conditions_mutex_p (insn, list->insn ()))
add_dependence (insn, list->insn (), dep_type);
}
mark_as_hard = false;
}
/* Similar, but free *LISTP at the same time, when the context
is not readonly. HARD should be true if DEP_NONREG should be set on
newly created dependencies. */
static void
add_dependence_list_and_free (struct deps_desc *deps, rtx_insn *insn,
rtx_insn_list **listp,
int uncond, enum reg_note dep_type, bool hard)
{
add_dependence_list (insn, *listp, uncond, dep_type, hard);
/* We don't want to short-circuit dependencies involving debug
insns, because they may cause actual dependencies to be
disregarded. */
if (deps->readonly || DEBUG_INSN_P (insn))
return;
free_INSN_LIST_list (listp);
}
/* Remove all occurrences of INSN from LIST. Return the number of
occurrences removed. */
static int
remove_from_dependence_list (rtx_insn *insn, rtx_insn_list **listp)
{
int removed = 0;
while (*listp)
{
if ((*listp)->insn () == insn)
{
remove_free_INSN_LIST_node (listp);
removed++;
continue;
}
listp = (rtx_insn_list **)&XEXP (*listp, 1);
}
return removed;
}
/* Same as above, but process two lists at once. */
static int
remove_from_both_dependence_lists (rtx_insn *insn,
rtx_insn_list **listp,
rtx_expr_list **exprp)
{
int removed = 0;
while (*listp)
{
if (XEXP (*listp, 0) == insn)
{
remove_free_INSN_LIST_node (listp);
remove_free_EXPR_LIST_node (exprp);
removed++;
continue;
}
listp = (rtx_insn_list **)&XEXP (*listp, 1);
exprp = (rtx_expr_list **)&XEXP (*exprp, 1);
}
return removed;
}
/* Clear all dependencies for an insn. */
static void
delete_all_dependences (rtx_insn *insn)
{
sd_iterator_def sd_it;
dep_t dep;
/* The below cycle can be optimized to clear the caches and back_deps
in one call but that would provoke duplication of code from
delete_dep (). */
for (sd_it = sd_iterator_start (insn, SD_LIST_BACK);
sd_iterator_cond (&sd_it, &dep);)
sd_delete_dep (sd_it);
}
/* All insns in a scheduling group except the first should only have
dependencies on the previous insn in the group. So we find the
first instruction in the scheduling group by walking the dependence
chains backwards. Then we add the dependencies for the group to
the previous nonnote insn. */
static void
chain_to_prev_insn (rtx_insn *insn)
{
sd_iterator_def sd_it;
dep_t dep;
rtx_insn *prev_nonnote;
FOR_EACH_DEP (insn, SD_LIST_BACK, sd_it, dep)
{
rtx_insn *i = insn;
rtx_insn *pro = DEP_PRO (dep);
do
{
i = prev_nonnote_insn (i);
if (pro == i)
goto next_link;
} while (SCHED_GROUP_P (i) || DEBUG_INSN_P (i));
if (! sched_insns_conditions_mutex_p (i, pro))
add_dependence (i, pro, DEP_TYPE (dep));
next_link:;
}
delete_all_dependences (insn);
prev_nonnote = prev_nonnote_nondebug_insn (insn);
if (BLOCK_FOR_INSN (insn) == BLOCK_FOR_INSN (prev_nonnote)
&& ! sched_insns_conditions_mutex_p (insn, prev_nonnote))
add_dependence (insn, prev_nonnote, REG_DEP_ANTI);
}
/* Process an insn's memory dependencies. There are four kinds of
dependencies:
(0) read dependence: read follows read
(1) true dependence: read follows write
(2) output dependence: write follows write
(3) anti dependence: write follows read
We are careful to build only dependencies which actually exist, and
use transitivity to avoid building too many links. */
/* Add an INSN and MEM reference pair to a pending INSN_LIST and MEM_LIST.
The MEM is a memory reference contained within INSN, which we are saving
so that we can do memory aliasing on it. */
static void
add_insn_mem_dependence (struct deps_desc *deps, bool read_p,
rtx_insn *insn, rtx mem)
{
rtx_insn_list **insn_list;
rtx_insn_list *insn_node;
rtx_expr_list **mem_list;
rtx_expr_list *mem_node;
gcc_assert (!deps->readonly);
if (read_p)
{
insn_list = &deps->pending_read_insns;
mem_list = &deps->pending_read_mems;
if (!DEBUG_INSN_P (insn))
deps->pending_read_list_length++;
}
else
{
insn_list = &deps->pending_write_insns;
mem_list = &deps->pending_write_mems;
deps->pending_write_list_length++;
}
insn_node = alloc_INSN_LIST (insn, *insn_list);
*insn_list = insn_node;
if (sched_deps_info->use_cselib)
{
mem = shallow_copy_rtx (mem);
XEXP (mem, 0) = cselib_subst_to_values_from_insn (XEXP (mem, 0),
GET_MODE (mem), insn);
}
mem_node = alloc_EXPR_LIST (VOIDmode, canon_rtx (mem), *mem_list);
*mem_list = mem_node;
}
/* Make a dependency between every memory reference on the pending lists
and INSN, thus flushing the pending lists. FOR_READ is true if emitting
dependencies for a read operation, similarly with FOR_WRITE. */
static void
flush_pending_lists (struct deps_desc *deps, rtx_insn *insn, int for_read,
int for_write)
{
if (for_write)
{
add_dependence_list_and_free (deps, insn, &deps->pending_read_insns,
1, REG_DEP_ANTI, true);
if (!deps->readonly)
{
free_EXPR_LIST_list (&deps->pending_read_mems);
deps->pending_read_list_length = 0;
}
}
add_dependence_list_and_free (deps, insn, &deps->pending_write_insns, 1,
for_read ? REG_DEP_ANTI : REG_DEP_OUTPUT,
true);
add_dependence_list_and_free (deps, insn,
&deps->last_pending_memory_flush, 1,
for_read ? REG_DEP_ANTI : REG_DEP_OUTPUT,
true);
add_dependence_list_and_free (deps, insn, &deps->pending_jump_insns, 1,
REG_DEP_ANTI, true);
if (DEBUG_INSN_P (insn))
{
if (for_write)
free_INSN_LIST_list (&deps->pending_read_insns);
free_INSN_LIST_list (&deps->pending_write_insns);
free_INSN_LIST_list (&deps->last_pending_memory_flush);
free_INSN_LIST_list (&deps->pending_jump_insns);
}
if (!deps->readonly)
{
free_EXPR_LIST_list (&deps->pending_write_mems);
deps->pending_write_list_length = 0;
deps->last_pending_memory_flush = alloc_INSN_LIST (insn, NULL_RTX);
deps->pending_flush_length = 1;
}
mark_as_hard = false;
}
/* Instruction which dependencies we are analyzing. */
static rtx_insn *cur_insn = NULL;
/* Implement hooks for haifa scheduler. */
static void
haifa_start_insn (rtx_insn *insn)
{
gcc_assert (insn && !cur_insn);
cur_insn = insn;
}
static void
haifa_finish_insn (void)
{
cur_insn = NULL;
}
void
haifa_note_reg_set (int regno)
{
SET_REGNO_REG_SET (reg_pending_sets, regno);
}
void
haifa_note_reg_clobber (int regno)
{
SET_REGNO_REG_SET (reg_pending_clobbers, regno);
}
void
haifa_note_reg_use (int regno)
{
SET_REGNO_REG_SET (reg_pending_uses, regno);
}
static void
haifa_note_mem_dep (rtx mem, rtx pending_mem, rtx_insn *pending_insn, ds_t ds)
{
if (!(ds & SPECULATIVE))
{
mem = NULL_RTX;
pending_mem = NULL_RTX;
}
else
gcc_assert (ds & BEGIN_DATA);
{
dep_def _dep, *dep = &_dep;
init_dep_1 (dep, pending_insn, cur_insn, ds_to_dt (ds),
current_sched_info->flags & USE_DEPS_LIST ? ds : 0);
DEP_NONREG (dep) = 1;
maybe_add_or_update_dep_1 (dep, false, pending_mem, mem);
}
}
static void
haifa_note_dep (rtx_insn *elem, ds_t ds)
{
dep_def _dep;
dep_t dep = &_dep;
init_dep (dep, elem, cur_insn, ds_to_dt (ds));
if (mark_as_hard)
DEP_NONREG (dep) = 1;
maybe_add_or_update_dep_1 (dep, false, NULL_RTX, NULL_RTX);
}
static void
note_reg_use (int r)
{
if (sched_deps_info->note_reg_use)
sched_deps_info->note_reg_use (r);
}
static void
note_reg_set (int r)
{
if (sched_deps_info->note_reg_set)
sched_deps_info->note_reg_set (r);
}
static void
note_reg_clobber (int r)
{
if (sched_deps_info->note_reg_clobber)
sched_deps_info->note_reg_clobber (r);
}
static void
note_mem_dep (rtx m1, rtx m2, rtx_insn *e, ds_t ds)
{
if (sched_deps_info->note_mem_dep)
sched_deps_info->note_mem_dep (m1, m2, e, ds);
}
static void
note_dep (rtx_insn *e, ds_t ds)
{
if (sched_deps_info->note_dep)
sched_deps_info->note_dep (e, ds);
}
/* Return corresponding to DS reg_note. */
enum reg_note
ds_to_dt (ds_t ds)
{
if (ds & DEP_TRUE)
return REG_DEP_TRUE;
else if (ds & DEP_OUTPUT)
return REG_DEP_OUTPUT;
else if (ds & DEP_ANTI)
return REG_DEP_ANTI;
else
{
gcc_assert (ds & DEP_CONTROL);
return REG_DEP_CONTROL;
}
}
/* Functions for computation of info needed for register pressure
sensitive insn scheduling. */
/* Allocate and return reg_use_data structure for REGNO and INSN. */
static struct reg_use_data *
create_insn_reg_use (int regno, rtx_insn *insn)
{
struct reg_use_data *use;
use = (struct reg_use_data *) xmalloc (sizeof (struct reg_use_data));
use->regno = regno;
use->insn = insn;
use->next_insn_use = INSN_REG_USE_LIST (insn);
INSN_REG_USE_LIST (insn) = use;
return use;
}
/* Allocate reg_set_data structure for REGNO and INSN. */
static void
create_insn_reg_set (int regno, rtx insn)
{
struct reg_set_data *set;
set = (struct reg_set_data *) xmalloc (sizeof (struct reg_set_data));
set->regno = regno;
set->insn = insn;
set->next_insn_set = INSN_REG_SET_LIST (insn);
INSN_REG_SET_LIST (insn) = set;
}
/* Set up insn register uses for INSN and dependency context DEPS. */
static void
setup_insn_reg_uses (struct deps_desc *deps, rtx_insn *insn)
{
unsigned i;
reg_set_iterator rsi;
struct reg_use_data *use, *use2, *next;
struct deps_reg *reg_last;
EXECUTE_IF_SET_IN_REG_SET (reg_pending_uses, 0, i, rsi)
{
if (i < FIRST_PSEUDO_REGISTER
&& TEST_HARD_REG_BIT (ira_no_alloc_regs, i))
continue;
if (find_regno_note (insn, REG_DEAD, i) == NULL_RTX
&& ! REGNO_REG_SET_P (reg_pending_sets, i)
&& ! REGNO_REG_SET_P (reg_pending_clobbers, i))
/* Ignore use which is not dying. */
continue;
use = create_insn_reg_use (i, insn);
use->next_regno_use = use;
reg_last = &deps->reg_last[i];
/* Create the cycle list of uses. */
for (rtx_insn_list *list = reg_last->uses; list; list = list->next ())
{
use2 = create_insn_reg_use (i, list->insn ());
next = use->next_regno_use;
use->next_regno_use = use2;
use2->next_regno_use = next;
}
}
}
/* Register pressure info for the currently processed insn. */
static struct reg_pressure_data reg_pressure_info[N_REG_CLASSES];
/* Return TRUE if INSN has the use structure for REGNO. */
static bool
insn_use_p (rtx insn, int regno)
{
struct reg_use_data *use;
for (use = INSN_REG_USE_LIST (insn); use != NULL; use = use->next_insn_use)
if (use->regno == regno)
return true;
return false;
}
/* Update the register pressure info after birth of pseudo register REGNO
in INSN. Arguments CLOBBER_P and UNUSED_P say correspondingly that
the register is in clobber or unused after the insn. */
static void
mark_insn_pseudo_birth (rtx insn, int regno, bool clobber_p, bool unused_p)
{
int incr, new_incr;
enum reg_class cl;
gcc_assert (regno >= FIRST_PSEUDO_REGISTER);
cl = sched_regno_pressure_class[regno];
if (cl != NO_REGS)
{
incr = ira_reg_class_max_nregs[cl][PSEUDO_REGNO_MODE (regno)];
if (clobber_p)
{
new_incr = reg_pressure_info[cl].clobber_increase + incr;
reg_pressure_info[cl].clobber_increase = new_incr;
}
else if (unused_p)
{
new_incr = reg_pressure_info[cl].unused_set_increase + incr;
reg_pressure_info[cl].unused_set_increase = new_incr;
}
else
{
new_incr = reg_pressure_info[cl].set_increase + incr;
reg_pressure_info[cl].set_increase = new_incr;
if (! insn_use_p (insn, regno))
reg_pressure_info[cl].change += incr;
create_insn_reg_set (regno, insn);
}
gcc_assert (new_incr < (1 << INCREASE_BITS));
}
}
/* Like mark_insn_pseudo_regno_birth except that NREGS saying how many
hard registers involved in the birth. */
static void
mark_insn_hard_regno_birth (rtx insn, int regno, int nregs,
bool clobber_p, bool unused_p)
{
enum reg_class cl;
int new_incr, last = regno + nregs;
while (regno < last)
{
gcc_assert (regno < FIRST_PSEUDO_REGISTER);
if (! TEST_HARD_REG_BIT (ira_no_alloc_regs, regno))
{
cl = sched_regno_pressure_class[regno];
if (cl != NO_REGS)
{
if (clobber_p)
{
new_incr = reg_pressure_info[cl].clobber_increase + 1;
reg_pressure_info[cl].clobber_increase = new_incr;
}
else if (unused_p)
{
new_incr = reg_pressure_info[cl].unused_set_increase + 1;
reg_pressure_info[cl].unused_set_increase = new_incr;
}
else
{
new_incr = reg_pressure_info[cl].set_increase + 1;
reg_pressure_info[cl].set_increase = new_incr;
if (! insn_use_p (insn, regno))
reg_pressure_info[cl].change += 1;
create_insn_reg_set (regno, insn);
}
gcc_assert (new_incr < (1 << INCREASE_BITS));
}
}
regno++;
}
}
/* Update the register pressure info after birth of pseudo or hard
register REG in INSN. Arguments CLOBBER_P and UNUSED_P say
correspondingly that the register is in clobber or unused after the
insn. */
static void
mark_insn_reg_birth (rtx insn, rtx reg, bool clobber_p, bool unused_p)
{
int regno;
if (GET_CODE (reg) == SUBREG)
reg = SUBREG_REG (reg);
if (! REG_P (reg))
return;
regno = REGNO (reg);
if (regno < FIRST_PSEUDO_REGISTER)
mark_insn_hard_regno_birth (insn, regno, REG_NREGS (reg),
clobber_p, unused_p);
else
mark_insn_pseudo_birth (insn, regno, clobber_p, unused_p);
}
/* Update the register pressure info after death of pseudo register
REGNO. */
static void
mark_pseudo_death (int regno)
{
int incr;
enum reg_class cl;
gcc_assert (regno >= FIRST_PSEUDO_REGISTER);
cl = sched_regno_pressure_class[regno];
if (cl != NO_REGS)
{
incr = ira_reg_class_max_nregs[cl][PSEUDO_REGNO_MODE (regno)];
reg_pressure_info[cl].change -= incr;
}
}
/* Like mark_pseudo_death except that NREGS saying how many hard
registers involved in the death. */
static void
mark_hard_regno_death (int regno, int nregs)
{
enum reg_class cl;
int last = regno + nregs;
while (regno < last)
{
gcc_assert (regno < FIRST_PSEUDO_REGISTER);
if (! TEST_HARD_REG_BIT (ira_no_alloc_regs, regno))
{
cl = sched_regno_pressure_class[regno];
if (cl != NO_REGS)
reg_pressure_info[cl].change -= 1;
}
regno++;
}
}
/* Update the register pressure info after death of pseudo or hard
register REG. */
static void
mark_reg_death (rtx reg)
{
int regno;
if (GET_CODE (reg) == SUBREG)
reg = SUBREG_REG (reg);
if (! REG_P (reg))
return;
regno = REGNO (reg);
if (regno < FIRST_PSEUDO_REGISTER)
mark_hard_regno_death (regno, REG_NREGS (reg));
else
mark_pseudo_death (regno);
}
/* Process SETTER of REG. DATA is an insn containing the setter. */
static void
mark_insn_reg_store (rtx reg, const_rtx setter, void *data)
{
if (setter != NULL_RTX && GET_CODE (setter) != SET)
return;
mark_insn_reg_birth
((rtx) data, reg, false,
find_reg_note ((const_rtx) data, REG_UNUSED, reg) != NULL_RTX);
}
/* Like mark_insn_reg_store except notice just CLOBBERs; ignore SETs. */
static void
mark_insn_reg_clobber (rtx reg, const_rtx setter, void *data)
{
if (GET_CODE (setter) == CLOBBER)
mark_insn_reg_birth ((rtx) data, reg, true, false);
}
/* Set up reg pressure info related to INSN. */
void
init_insn_reg_pressure_info (rtx_insn *insn)
{
int i, len;
enum reg_class cl;
static struct reg_pressure_data *pressure_info;
rtx link;
gcc_assert (sched_pressure != SCHED_PRESSURE_NONE);
if (! INSN_P (insn))
return;
for (i = 0; i < ira_pressure_classes_num; i++)
{
cl = ira_pressure_classes[i];
reg_pressure_info[cl].clobber_increase = 0;
reg_pressure_info[cl].set_increase = 0;
reg_pressure_info[cl].unused_set_increase = 0;
reg_pressure_info[cl].change = 0;
}
note_stores (PATTERN (insn), mark_insn_reg_clobber, insn);
note_stores (PATTERN (insn), mark_insn_reg_store, insn);
if (AUTO_INC_DEC)
for (link = REG_NOTES (insn); link; link = XEXP (link, 1))
if (REG_NOTE_KIND (link) == REG_INC)
mark_insn_reg_store (XEXP (link, 0), NULL_RTX, insn);
for (link = REG_NOTES (insn); link; link = XEXP (link, 1))
if (REG_NOTE_KIND (link) == REG_DEAD)
mark_reg_death (XEXP (link, 0));
len = sizeof (struct reg_pressure_data) * ira_pressure_classes_num;
pressure_info
= INSN_REG_PRESSURE (insn) = (struct reg_pressure_data *) xmalloc (len);
if (sched_pressure == SCHED_PRESSURE_WEIGHTED)
INSN_MAX_REG_PRESSURE (insn) = (int *) xcalloc (ira_pressure_classes_num
* sizeof (int), 1);
for (i = 0; i < ira_pressure_classes_num; i++)
{
cl = ira_pressure_classes[i];
pressure_info[i].clobber_increase
= reg_pressure_info[cl].clobber_increase;
pressure_info[i].set_increase = reg_pressure_info[cl].set_increase;
pressure_info[i].unused_set_increase
= reg_pressure_info[cl].unused_set_increase;
pressure_info[i].change = reg_pressure_info[cl].change;
}
}
/* Internal variable for sched_analyze_[12] () functions.
If it is nonzero, this means that sched_analyze_[12] looks
at the most toplevel SET. */
static bool can_start_lhs_rhs_p;
/* Extend reg info for the deps context DEPS given that
we have just generated a register numbered REGNO. */
static void
extend_deps_reg_info (struct deps_desc *deps, int regno)
{
int max_regno = regno + 1;
gcc_assert (!reload_completed);
/* In a readonly context, it would not hurt to extend info,
but it should not be needed. */
if (reload_completed && deps->readonly)
{
deps->max_reg = max_regno;
return;
}
if (max_regno > deps->max_reg)
{
deps->reg_last = XRESIZEVEC (struct deps_reg, deps->reg_last,
max_regno);
memset (&deps->reg_last[deps->max_reg],
0, (max_regno - deps->max_reg)
* sizeof (struct deps_reg));
deps->max_reg = max_regno;
}
}
/* Extends REG_INFO_P if needed. */
void
maybe_extend_reg_info_p (void)
{
/* Extend REG_INFO_P, if needed. */
if ((unsigned int)max_regno - 1 >= reg_info_p_size)
{
size_t new_reg_info_p_size = max_regno + 128;
gcc_assert (!reload_completed && sel_sched_p ());
reg_info_p = (struct reg_info_t *) xrecalloc (reg_info_p,
new_reg_info_p_size,
reg_info_p_size,
sizeof (*reg_info_p));
reg_info_p_size = new_reg_info_p_size;
}
}
/* Analyze a single reference to register (reg:MODE REGNO) in INSN.
The type of the reference is specified by REF and can be SET,
CLOBBER, PRE_DEC, POST_DEC, PRE_INC, POST_INC or USE. */
static void
sched_analyze_reg (struct deps_desc *deps, int regno, machine_mode mode,
enum rtx_code ref, rtx_insn *insn)
{
/* We could emit new pseudos in renaming. Extend the reg structures. */
if (!reload_completed && sel_sched_p ()
&& (regno >= max_reg_num () - 1 || regno >= deps->max_reg))
extend_deps_reg_info (deps, regno);
maybe_extend_reg_info_p ();
/* A hard reg in a wide mode may really be multiple registers.
If so, mark all of them just like the first. */
if (regno < FIRST_PSEUDO_REGISTER)
{
int i = hard_regno_nregs (regno, mode);
if (ref == SET)
{
while (--i >= 0)
note_reg_set (regno + i);
}
else if (ref == USE)
{
while (--i >= 0)
note_reg_use (regno + i);
}
else
{
while (--i >= 0)
note_reg_clobber (regno + i);
}
}
/* ??? Reload sometimes emits USEs and CLOBBERs of pseudos that
it does not reload. Ignore these as they have served their
purpose already. */
else if (regno >= deps->max_reg)
{
enum rtx_code code = GET_CODE (PATTERN (insn));
gcc_assert (code == USE || code == CLOBBER);
}
else
{
if (ref == SET)
note_reg_set (regno);
else if (ref == USE)
note_reg_use (regno);
else
note_reg_clobber (regno);
/* Pseudos that are REG_EQUIV to something may be replaced
by that during reloading. We need only add dependencies for
the address in the REG_EQUIV note. */
if (!reload_completed && get_reg_known_equiv_p (regno))
{
rtx t = get_reg_known_value (regno);
if (MEM_P (t))
sched_analyze_2 (deps, XEXP (t, 0), insn);
}
/* Don't let it cross a call after scheduling if it doesn't
already cross one. */
if (REG_N_CALLS_CROSSED (regno) == 0)
{
if (!deps->readonly && ref == USE && !DEBUG_INSN_P (insn))
deps->sched_before_next_call
= alloc_INSN_LIST (insn, deps->sched_before_next_call);
else
add_dependence_list (insn, deps->last_function_call, 1,
REG_DEP_ANTI, false);
}
}
}
/* Analyze a single SET, CLOBBER, PRE_DEC, POST_DEC, PRE_INC or POST_INC
rtx, X, creating all dependencies generated by the write to the
destination of X, and reads of everything mentioned. */
static void
sched_analyze_1 (struct deps_desc *deps, rtx x, rtx_insn *insn)
{
rtx dest = XEXP (x, 0);
enum rtx_code code = GET_CODE (x);
bool cslr_p = can_start_lhs_rhs_p;
can_start_lhs_rhs_p = false;
gcc_assert (dest);
if (dest == 0)
return;
if (cslr_p && sched_deps_info->start_lhs)
sched_deps_info->start_lhs (dest);
if (GET_CODE (dest) == PARALLEL)
{
int i;
for (i = XVECLEN (dest, 0) - 1; i >= 0; i--)
if (XEXP (XVECEXP (dest, 0, i), 0) != 0)
sched_analyze_1 (deps,
gen_rtx_CLOBBER (VOIDmode,
XEXP (XVECEXP (dest, 0, i), 0)),
insn);
if (cslr_p && sched_deps_info->finish_lhs)
sched_deps_info->finish_lhs ();
if (code == SET)
{
can_start_lhs_rhs_p = cslr_p;
sched_analyze_2 (deps, SET_SRC (x), insn);
can_start_lhs_rhs_p = false;
}
return;
}
while (GET_CODE (dest) == STRICT_LOW_PART || GET_CODE (dest) == SUBREG
|| GET_CODE (dest) == ZERO_EXTRACT)
{
if (GET_CODE (dest) == STRICT_LOW_PART
|| GET_CODE (dest) == ZERO_EXTRACT
|| read_modify_subreg_p (dest))
{
/* These both read and modify the result. We must handle
them as writes to get proper dependencies for following
instructions. We must handle them as reads to get proper
dependencies from this to previous instructions.
Thus we need to call sched_analyze_2. */
sched_analyze_2 (deps, XEXP (dest, 0), insn);
}
if (GET_CODE (dest) == ZERO_EXTRACT)
{
/* The second and third arguments are values read by this insn. */
sched_analyze_2 (deps, XEXP (dest, 1), insn);
sched_analyze_2 (deps, XEXP (dest, 2), insn);
}
dest = XEXP (dest, 0);
}
if (REG_P (dest))
{
int regno = REGNO (dest);
machine_mode mode = GET_MODE (dest);
sched_analyze_reg (deps, regno, mode, code, insn);
#ifdef STACK_REGS
/* Treat all writes to a stack register as modifying the TOS. */
if (regno >= FIRST_STACK_REG && regno <= LAST_STACK_REG)
{
/* Avoid analyzing the same register twice. */
if (regno != FIRST_STACK_REG)
sched_analyze_reg (deps, FIRST_STACK_REG, mode, code, insn);
add_to_hard_reg_set (&implicit_reg_pending_uses, mode,
FIRST_STACK_REG);
}
#endif
}
else if (MEM_P (dest))
{
/* Writing memory. */
rtx t = dest;
if (sched_deps_info->use_cselib)
{
machine_mode address_mode = get_address_mode (dest);
t = shallow_copy_rtx (dest);
cselib_lookup_from_insn (XEXP (t, 0), address_mode, 1,
GET_MODE (t), insn);
XEXP (t, 0)
= cselib_subst_to_values_from_insn (XEXP (t, 0), GET_MODE (t),
insn);
}
t = canon_rtx (t);
/* Pending lists can't get larger with a readonly context. */
if (!deps->readonly
&& ((deps->pending_read_list_length + deps->pending_write_list_length)
>= MAX_PENDING_LIST_LENGTH))
{
/* Flush all pending reads and writes to prevent the pending lists
from getting any larger. Insn scheduling runs too slowly when
these lists get long. When compiling GCC with itself,
this flush occurs 8 times for sparc, and 10 times for m88k using
the default value of 32. */
flush_pending_lists (deps, insn, false, true);
}
else
{
rtx_insn_list *pending;
rtx_expr_list *pending_mem;
pending = deps->pending_read_insns;
pending_mem = deps->pending_read_mems;
while (pending)
{
if (anti_dependence (pending_mem->element (), t)
&& ! sched_insns_conditions_mutex_p (insn, pending->insn ()))
note_mem_dep (t, pending_mem->element (), pending->insn (),
DEP_ANTI);
pending = pending->next ();
pending_mem = pending_mem->next ();
}
pending = deps->pending_write_insns;
pending_mem = deps->pending_write_mems;
while (pending)
{
if (output_dependence (pending_mem->element (), t)
&& ! sched_insns_conditions_mutex_p (insn, pending->insn ()))
note_mem_dep (t, pending_mem->element (),
pending->insn (),
DEP_OUTPUT);
pending = pending->next ();
pending_mem = pending_mem-> next ();
}
add_dependence_list (insn, deps->last_pending_memory_flush, 1,
REG_DEP_ANTI, true);
add_dependence_list (insn, deps->pending_jump_insns, 1,
REG_DEP_CONTROL, true);
if (!deps->readonly)
add_insn_mem_dependence (deps, false, insn, dest);
}
sched_analyze_2 (deps, XEXP (dest, 0), insn);
}
if (cslr_p && sched_deps_info->finish_lhs)
sched_deps_info->finish_lhs ();
/* Analyze reads. */
if (GET_CODE (x) == SET)
{
can_start_lhs_rhs_p = cslr_p;
sched_analyze_2 (deps, SET_SRC (x), insn);
can_start_lhs_rhs_p = false;
}
}
/* Analyze the uses of memory and registers in rtx X in INSN. */
static void
sched_analyze_2 (struct deps_desc *deps, rtx x, rtx_insn *insn)
{
int i;
int j;
enum rtx_code code;
const char *fmt;
bool cslr_p = can_start_lhs_rhs_p;
can_start_lhs_rhs_p = false;
gcc_assert (x);
if (x == 0)
return;
if (cslr_p && sched_deps_info->start_rhs)
sched_deps_info->start_rhs (x);
code = GET_CODE (x);
switch (code)
{
CASE_CONST_ANY:
case SYMBOL_REF:
case CONST:
case LABEL_REF:
/* Ignore constants. */
if (cslr_p && sched_deps_info->finish_rhs)
sched_deps_info->finish_rhs ();
return;
case CC0:
if (!HAVE_cc0)
gcc_unreachable ();
/* User of CC0 depends on immediately preceding insn. */
SCHED_GROUP_P (insn) = 1;
/* Don't move CC0 setter to another block (it can set up the
same flag for previous CC0 users which is safe). */
CANT_MOVE (prev_nonnote_insn (insn)) = 1;
if (cslr_p && sched_deps_info->finish_rhs)
sched_deps_info->finish_rhs ();
return;
case REG:
{
int regno = REGNO (x);
machine_mode mode = GET_MODE (x);
sched_analyze_reg (deps, regno, mode, USE, insn);
#ifdef STACK_REGS
/* Treat all reads of a stack register as modifying the TOS. */
if (regno >= FIRST_STACK_REG && regno <= LAST_STACK_REG)
{
/* Avoid analyzing the same register twice. */
if (regno != FIRST_STACK_REG)
sched_analyze_reg (deps, FIRST_STACK_REG, mode, USE, insn);
sched_analyze_reg (deps, FIRST_STACK_REG, mode, SET, insn);
}
#endif
if (cslr_p && sched_deps_info->finish_rhs)
sched_deps_info->finish_rhs ();
return;
}
case MEM:
{
/* Reading memory. */
rtx_insn_list *u;
rtx_insn_list *pending;
rtx_expr_list *pending_mem;
rtx t = x;
if (sched_deps_info->use_cselib)
{
machine_mode address_mode = get_address_mode (t);
t = shallow_copy_rtx (t);
cselib_lookup_from_insn (XEXP (t, 0), address_mode, 1,
GET_MODE (t), insn);
XEXP (t, 0)
= cselib_subst_to_values_from_insn (XEXP (t, 0), GET_MODE (t),
insn);
}
if (!DEBUG_INSN_P (insn))
{
t = canon_rtx (t);
pending = deps->pending_read_insns;
pending_mem = deps->pending_read_mems;
while (pending)
{
if (read_dependence (pending_mem->element (), t)
&& ! sched_insns_conditions_mutex_p (insn,
pending->insn ()))
note_mem_dep (t, pending_mem->element (),
pending->insn (),
DEP_ANTI);
pending = pending->next ();
pending_mem = pending_mem->next ();
}
pending = deps->pending_write_insns;
pending_mem = deps->pending_write_mems;
while (pending)
{
if (true_dependence (pending_mem->element (), VOIDmode, t)
&& ! sched_insns_conditions_mutex_p (insn,
pending->insn ()))
note_mem_dep (t, pending_mem->element (),
pending->insn (),
sched_deps_info->generate_spec_deps
? BEGIN_DATA | DEP_TRUE : DEP_TRUE);
pending = pending->next ();
pending_mem = pending_mem->next ();
}
for (u = deps->last_pending_memory_flush; u; u = u->next ())
add_dependence (insn, u->insn (), REG_DEP_ANTI);
for (u = deps->pending_jump_insns; u; u = u->next ())
if (deps_may_trap_p (x))
{
if ((sched_deps_info->generate_spec_deps)
&& sel_sched_p () && (spec_info->mask & BEGIN_CONTROL))
{
ds_t ds = set_dep_weak (DEP_ANTI, BEGIN_CONTROL,
MAX_DEP_WEAK);
note_dep (u->insn (), ds);
}
else
add_dependence (insn, u->insn (), REG_DEP_CONTROL);
}
}
/* Always add these dependencies to pending_reads, since
this insn may be followed by a write. */
if (!deps->readonly)
{
if ((deps->pending_read_list_length
+ deps->pending_write_list_length)
>= MAX_PENDING_LIST_LENGTH
&& !DEBUG_INSN_P (insn))
flush_pending_lists (deps, insn, true, true);
add_insn_mem_dependence (deps, true, insn, x);
}
sched_analyze_2 (deps, XEXP (x, 0), insn);
if (cslr_p && sched_deps_info->finish_rhs)
sched_deps_info->finish_rhs ();
return;
}
/* Force pending stores to memory in case a trap handler needs them.
Also force pending loads from memory; loads and stores can segfault
and the signal handler won't be triggered if the trap insn was moved
above load or store insn. */
case TRAP_IF:
flush_pending_lists (deps, insn, true, true);
break;
case PREFETCH:
if (PREFETCH_SCHEDULE_BARRIER_P (x))
reg_pending_barrier = TRUE_BARRIER;
/* Prefetch insn contains addresses only. So if the prefetch
address has no registers, there will be no dependencies on
the prefetch insn. This is wrong with result code
correctness point of view as such prefetch can be moved below
a jump insn which usually generates MOVE_BARRIER preventing
to move insns containing registers or memories through the
barrier. It is also wrong with generated code performance
point of view as prefetch withouth dependecies will have a
tendency to be issued later instead of earlier. It is hard
to generate accurate dependencies for prefetch insns as
prefetch has only the start address but it is better to have
something than nothing. */
if (!deps->readonly)
{
rtx x = gen_rtx_MEM (Pmode, XEXP (PATTERN (insn), 0));
if (sched_deps_info->use_cselib)
cselib_lookup_from_insn (x, Pmode, true, VOIDmode, insn);
add_insn_mem_dependence (deps, true, insn, x);
}
break;
case UNSPEC_VOLATILE:
flush_pending_lists (deps, insn, true, true);
/* FALLTHRU */
case ASM_OPERANDS:
case ASM_INPUT:
{
/* Traditional and volatile asm instructions must be considered to use
and clobber all hard registers, all pseudo-registers and all of
memory. So must TRAP_IF and UNSPEC_VOLATILE operations.
Consider for instance a volatile asm that changes the fpu rounding
mode. An insn should not be moved across this even if it only uses
pseudo-regs because it might give an incorrectly rounded result. */
if ((code != ASM_OPERANDS || MEM_VOLATILE_P (x))
&& !DEBUG_INSN_P (insn))
reg_pending_barrier = TRUE_BARRIER;
/* For all ASM_OPERANDS, we must traverse the vector of input operands.
We can not just fall through here since then we would be confused
by the ASM_INPUT rtx inside ASM_OPERANDS, which do not indicate
traditional asms unlike their normal usage. */
if (code == ASM_OPERANDS)
{
for (j = 0; j < ASM_OPERANDS_INPUT_LENGTH (x); j++)
sched_analyze_2 (deps, ASM_OPERANDS_INPUT (x, j), insn);
if (cslr_p && sched_deps_info->finish_rhs)
sched_deps_info->finish_rhs ();
return;
}
break;
}
case PRE_DEC:
case POST_DEC:
case PRE_INC:
case POST_INC:
/* These both read and modify the result. We must handle them as writes
to get proper dependencies for following instructions. We must handle
them as reads to get proper dependencies from this to previous
instructions. Thus we need to pass them to both sched_analyze_1
and sched_analyze_2. We must call sched_analyze_2 first in order
to get the proper antecedent for the read. */
sched_analyze_2 (deps, XEXP (x, 0), insn);
sched_analyze_1 (deps, x, insn);
if (cslr_p && sched_deps_info->finish_rhs)
sched_deps_info->finish_rhs ();
return;
case POST_MODIFY:
case PRE_MODIFY:
/* op0 = op0 + op1 */
sched_analyze_2 (deps, XEXP (x, 0), insn);
sched_analyze_2 (deps, XEXP (x, 1), insn);
sched_analyze_1 (deps, x, insn);
if (cslr_p && sched_deps_info->finish_rhs)
sched_deps_info->finish_rhs ();
return;
default:
break;
}
/* Other cases: walk the insn. */
fmt = GET_RTX_FORMAT (code);
for (i = GET_RTX_LENGTH (code) - 1; i >= 0; i--)
{
if (fmt[i] == 'e')
sched_analyze_2 (deps, XEXP (x, i), insn);
else if (fmt[i] == 'E')
for (j = 0; j < XVECLEN (x, i); j++)
sched_analyze_2 (deps, XVECEXP (x, i, j), insn);
}
if (cslr_p && sched_deps_info->finish_rhs)
sched_deps_info->finish_rhs ();
}
/* Try to group two fusible insns together to prevent scheduler
from scheduling them apart. */
static void
sched_macro_fuse_insns (rtx_insn *insn)
{
rtx_insn *prev;
/* No target hook would return true for debug insn as any of the
hook operand, and with very large sequences of only debug insns
where on each we call sched_macro_fuse_insns it has quadratic
compile time complexity. */
if (DEBUG_INSN_P (insn))
return;
prev = prev_nonnote_nondebug_insn (insn);
if (!prev)
return;
if (any_condjump_p (insn))
{
unsigned int condreg1, condreg2;
rtx cc_reg_1;
targetm.fixed_condition_code_regs (&condreg1, &condreg2);
cc_reg_1 = gen_rtx_REG (CCmode, condreg1);
if (reg_referenced_p (cc_reg_1, PATTERN (insn))
&& modified_in_p (cc_reg_1, prev))
{
if (targetm.sched.macro_fusion_pair_p (prev, insn))
SCHED_GROUP_P (insn) = 1;
return;
}
}
if (single_set (insn) && single_set (prev))
{
if (targetm.sched.macro_fusion_pair_p (prev, insn))
SCHED_GROUP_P (insn) = 1;
}
}
/* Get the implicit reg pending clobbers for INSN and save them in TEMP. */
void
get_implicit_reg_pending_clobbers (HARD_REG_SET *temp, rtx_insn *insn)
{
extract_insn (insn);
preprocess_constraints (insn);
alternative_mask preferred = get_preferred_alternatives (insn);
ira_implicitly_set_insn_hard_regs (temp, preferred);
AND_COMPL_HARD_REG_SET (*temp, ira_no_alloc_regs);
}
/* Analyze an INSN with pattern X to find all dependencies. */
static void
sched_analyze_insn (struct deps_desc *deps, rtx x, rtx_insn *insn)
{
RTX_CODE code = GET_CODE (x);
rtx link;
unsigned i;
reg_set_iterator rsi;
if (! reload_completed)
{
HARD_REG_SET temp;
get_implicit_reg_pending_clobbers (&temp, insn);
IOR_HARD_REG_SET (implicit_reg_pending_clobbers, temp);
}
can_start_lhs_rhs_p = (NONJUMP_INSN_P (insn)
&& code == SET);
/* Group compare and branch insns for macro-fusion. */
if (!deps->readonly
&& targetm.sched.macro_fusion_p
&& targetm.sched.macro_fusion_p ())
sched_macro_fuse_insns (insn);
if (may_trap_p (x))
/* Avoid moving trapping instructions across function calls that might
not always return. */
add_dependence_list (insn, deps->last_function_call_may_noreturn,
1, REG_DEP_ANTI, true);
/* We must avoid creating a situation in which two successors of the
current block have different unwind info after scheduling. If at any
point the two paths re-join this leads to incorrect unwind info. */
/* ??? There are certain situations involving a forced frame pointer in
which, with extra effort, we could fix up the unwind info at a later
CFG join. However, it seems better to notice these cases earlier
during prologue generation and avoid marking the frame pointer setup
as frame-related at all. */
if (RTX_FRAME_RELATED_P (insn))
{
/* Make sure prologue insn is scheduled before next jump. */
deps->sched_before_next_jump
= alloc_INSN_LIST (insn, deps->sched_before_next_jump);
/* Make sure epilogue insn is scheduled after preceding jumps. */
add_dependence_list (insn, deps->last_pending_memory_flush, 1,
REG_DEP_ANTI, true);
add_dependence_list (insn, deps->pending_jump_insns, 1, REG_DEP_ANTI,
true);
}
if (code == COND_EXEC)
{
sched_analyze_2 (deps, COND_EXEC_TEST (x), insn);
/* ??? Should be recording conditions so we reduce the number of
false dependencies. */
x = COND_EXEC_CODE (x);
code = GET_CODE (x);
}
if (code == SET || code == CLOBBER)
{
sched_analyze_1 (deps, x, insn);
/* Bare clobber insns are used for letting life analysis, reg-stack
and others know that a value is dead. Depend on the last call
instruction so that reg-stack won't get confused. */
if (code == CLOBBER)
add_dependence_list (insn, deps->last_function_call, 1,
REG_DEP_OUTPUT, true);
}
else if (code == PARALLEL)
{
for (i = XVECLEN (x, 0); i--;)
{
rtx sub = XVECEXP (x, 0, i);
code = GET_CODE (sub);
if (code == COND_EXEC)
{
sched_analyze_2 (deps, COND_EXEC_TEST (sub), insn);
sub = COND_EXEC_CODE (sub);
code = GET_CODE (sub);
}
if (code == SET || code == CLOBBER)
sched_analyze_1 (deps, sub, insn);
else
sched_analyze_2 (deps, sub, insn);
}
}
else
sched_analyze_2 (deps, x, insn);
/* Mark registers CLOBBERED or used by called function. */
if (CALL_P (insn))
{
for (link = CALL_INSN_FUNCTION_USAGE (insn); link; link = XEXP (link, 1))
{
if (GET_CODE (XEXP (link, 0)) == CLOBBER)
sched_analyze_1 (deps, XEXP (link, 0), insn);
else if (GET_CODE (XEXP (link, 0)) != SET)
sched_analyze_2 (deps, XEXP (link, 0), insn);
}
/* Don't schedule anything after a tail call, tail call needs
to use at least all call-saved registers. */
if (SIBLING_CALL_P (insn))
reg_pending_barrier = TRUE_BARRIER;
else if (find_reg_note (insn, REG_SETJMP, NULL))
reg_pending_barrier = MOVE_BARRIER;
}
if (JUMP_P (insn))
{
rtx_insn *next = next_nonnote_nondebug_insn (insn);
if (next && BARRIER_P (next))
reg_pending_barrier = MOVE_BARRIER;
else
{
rtx_insn_list *pending;
rtx_expr_list *pending_mem;
if (sched_deps_info->compute_jump_reg_dependencies)
{
(*sched_deps_info->compute_jump_reg_dependencies)
(insn, reg_pending_control_uses);
/* Make latency of jump equal to 0 by using anti-dependence. */
EXECUTE_IF_SET_IN_REG_SET (reg_pending_control_uses, 0, i, rsi)
{
struct deps_reg *reg_last = &deps->reg_last[i];
add_dependence_list (insn, reg_last->sets, 0, REG_DEP_ANTI,
false);
add_dependence_list (insn, reg_last->implicit_sets,
0, REG_DEP_ANTI, false);
add_dependence_list (insn, reg_last->clobbers, 0,
REG_DEP_ANTI, false);
}
}
/* All memory writes and volatile reads must happen before the
jump. Non-volatile reads must happen before the jump iff
the result is needed by the above register used mask. */
pending = deps->pending_write_insns;
pending_mem = deps->pending_write_mems;
while (pending)
{
if (! sched_insns_conditions_mutex_p (insn, pending->insn ()))
add_dependence (insn, pending->insn (),
REG_DEP_OUTPUT);
pending = pending->next ();
pending_mem = pending_mem->next ();
}
pending = deps->pending_read_insns;
pending_mem = deps->pending_read_mems;
while (pending)
{
if (MEM_VOLATILE_P (pending_mem->element ())
&& ! sched_insns_conditions_mutex_p (insn, pending->insn ()))
add_dependence (insn, pending->insn (),
REG_DEP_OUTPUT);
pending = pending->next ();
pending_mem = pending_mem->next ();
}
add_dependence_list (insn, deps->last_pending_memory_flush, 1,
REG_DEP_ANTI, true);
add_dependence_list (insn, deps->pending_jump_insns, 1,
REG_DEP_ANTI, true);
}
}
/* If this instruction can throw an exception, then moving it changes
where block boundaries fall. This is mighty confusing elsewhere.
Therefore, prevent such an instruction from being moved. Same for
non-jump instructions that define block boundaries.
??? Unclear whether this is still necessary in EBB mode. If not,
add_branch_dependences should be adjusted for RGN mode instead. */
if (((CALL_P (insn) || JUMP_P (insn)) && can_throw_internal (insn))
|| (NONJUMP_INSN_P (insn) && control_flow_insn_p (insn)))
reg_pending_barrier = MOVE_BARRIER;
if (sched_pressure != SCHED_PRESSURE_NONE)
{
setup_insn_reg_uses (deps, insn);
init_insn_reg_pressure_info (insn);
}
/* Add register dependencies for insn. */
if (DEBUG_INSN_P (insn))
{
rtx_insn *prev = deps->last_debug_insn;
rtx_insn_list *u;
if (!deps->readonly)
deps->last_debug_insn = insn;
if (prev)
add_dependence (insn, prev, REG_DEP_ANTI);
add_dependence_list (insn, deps->last_function_call, 1,
REG_DEP_ANTI, false);
if (!sel_sched_p ())
for (u = deps->last_pending_memory_flush; u; u = u->next ())
add_dependence (insn, u->insn (), REG_DEP_ANTI);
EXECUTE_IF_SET_IN_REG_SET (reg_pending_uses, 0, i, rsi)
{
struct deps_reg *reg_last = &deps->reg_last[i];
add_dependence_list (insn, reg_last->sets, 1, REG_DEP_ANTI, false);
/* There's no point in making REG_DEP_CONTROL dependencies for
debug insns. */
add_dependence_list (insn, reg_last->clobbers, 1, REG_DEP_ANTI,
false);
if (!deps->readonly)
reg_last->uses = alloc_INSN_LIST (insn, reg_last->uses);
}
CLEAR_REG_SET (reg_pending_uses);
/* Quite often, a debug insn will refer to stuff in the
previous instruction, but the reason we want this
dependency here is to make sure the scheduler doesn't
gratuitously move a debug insn ahead. This could dirty
DF flags and cause additional analysis that wouldn't have
occurred in compilation without debug insns, and such
additional analysis can modify the generated code. */
prev = PREV_INSN (insn);
if (prev && NONDEBUG_INSN_P (prev))
add_dependence (insn, prev, REG_DEP_ANTI);
}
else
{
regset_head set_or_clobbered;
EXECUTE_IF_SET_IN_REG_SET (reg_pending_uses, 0, i, rsi)
{
struct deps_reg *reg_last = &deps->reg_last[i];
add_dependence_list (insn, reg_last->sets, 0, REG_DEP_TRUE, false);
add_dependence_list (insn, reg_last->implicit_sets, 0, REG_DEP_ANTI,
false);
add_dependence_list (insn, reg_last->clobbers, 0, REG_DEP_TRUE,
false);
if (!deps->readonly)
{
reg_last->uses = alloc_INSN_LIST (insn, reg_last->uses);
reg_last->uses_length++;
}
}
for (i = 0; i < FIRST_PSEUDO_REGISTER; i++)
if (TEST_HARD_REG_BIT (implicit_reg_pending_uses, i))
{
struct deps_reg *reg_last = &deps->reg_last[i];
add_dependence_list (insn, reg_last->sets, 0, REG_DEP_TRUE, false);
add_dependence_list (insn, reg_last->implicit_sets, 0,
REG_DEP_ANTI, false);
add_dependence_list (insn, reg_last->clobbers, 0, REG_DEP_TRUE,
false);
if (!deps->readonly)
{
reg_last->uses = alloc_INSN_LIST (insn, reg_last->uses);
reg_last->uses_length++;
}
}
if (targetm.sched.exposed_pipeline)
{
INIT_REG_SET (&set_or_clobbered);
bitmap_ior (&set_or_clobbered, reg_pending_clobbers,
reg_pending_sets);
EXECUTE_IF_SET_IN_REG_SET (&set_or_clobbered, 0, i, rsi)
{
struct deps_reg *reg_last = &deps->reg_last[i];
rtx list;
for (list = reg_last->uses; list; list = XEXP (list, 1))
{
rtx other = XEXP (list, 0);
if (INSN_CACHED_COND (other) != const_true_rtx
&& refers_to_regno_p (i, INSN_CACHED_COND (other)))
INSN_CACHED_COND (other) = const_true_rtx;
}
}
}
/* If the current insn is conditional, we can't free any
of the lists. */
if (sched_has_condition_p (insn))
{
EXECUTE_IF_SET_IN_REG_SET (reg_pending_clobbers, 0, i, rsi)
{
struct deps_reg *reg_last = &deps->reg_last[i];
add_dependence_list (insn, reg_last->sets, 0, REG_DEP_OUTPUT,
false);
add_dependence_list (insn, reg_last->implicit_sets, 0,
REG_DEP_ANTI, false);
add_dependence_list (insn, reg_last->uses, 0, REG_DEP_ANTI,
false);
add_dependence_list (insn, reg_last->control_uses, 0,
REG_DEP_CONTROL, false);
if (!deps->readonly)
{
reg_last->clobbers
= alloc_INSN_LIST (insn, reg_last->clobbers);
reg_last->clobbers_length++;
}
}
EXECUTE_IF_SET_IN_REG_SET (reg_pending_sets, 0, i, rsi)
{
struct deps_reg *reg_last = &deps->reg_last[i];
add_dependence_list (insn, reg_last->sets, 0, REG_DEP_OUTPUT,
false);
add_dependence_list (insn, reg_last->implicit_sets, 0,
REG_DEP_ANTI, false);
add_dependence_list (insn, reg_last->clobbers, 0, REG_DEP_OUTPUT,
false);
add_dependence_list (insn, reg_last->uses, 0, REG_DEP_ANTI,
false);
add_dependence_list (insn, reg_last->control_uses, 0,
REG_DEP_CONTROL, false);
if (!deps->readonly)
reg_last->sets = alloc_INSN_LIST (insn, reg_last->sets);
}
}
else
{
EXECUTE_IF_SET_IN_REG_SET (reg_pending_clobbers, 0, i, rsi)
{
struct deps_reg *reg_last = &deps->reg_last[i];
if (reg_last->uses_length >= MAX_PENDING_LIST_LENGTH
|| reg_last->clobbers_length >= MAX_PENDING_LIST_LENGTH)
{
add_dependence_list_and_free (deps, insn, ®_last->sets, 0,
REG_DEP_OUTPUT, false);
add_dependence_list_and_free (deps, insn,
®_last->implicit_sets, 0,
REG_DEP_ANTI, false);
add_dependence_list_and_free (deps, insn, ®_last->uses, 0,
REG_DEP_ANTI, false);
add_dependence_list_and_free (deps, insn,
®_last->control_uses, 0,
REG_DEP_ANTI, false);
add_dependence_list_and_free (deps, insn,
®_last->clobbers, 0,
REG_DEP_OUTPUT, false);
if (!deps->readonly)
{
reg_last->sets = alloc_INSN_LIST (insn, reg_last->sets);
reg_last->clobbers_length = 0;
reg_last->uses_length = 0;
}
}
else
{
add_dependence_list (insn, reg_last->sets, 0, REG_DEP_OUTPUT,
false);
add_dependence_list (insn, reg_last->implicit_sets, 0,
REG_DEP_ANTI, false);
add_dependence_list (insn, reg_last->uses, 0, REG_DEP_ANTI,
false);
add_dependence_list (insn, reg_last->control_uses, 0,
REG_DEP_CONTROL, false);
}
if (!deps->readonly)
{
reg_last->clobbers_length++;
reg_last->clobbers
= alloc_INSN_LIST (insn, reg_last->clobbers);
}
}
EXECUTE_IF_SET_IN_REG_SET (reg_pending_sets, 0, i, rsi)
{
struct deps_reg *reg_last = &deps->reg_last[i];
add_dependence_list_and_free (deps, insn, ®_last->sets, 0,
REG_DEP_OUTPUT, false);
add_dependence_list_and_free (deps, insn,
®_last->implicit_sets,
0, REG_DEP_ANTI, false);
add_dependence_list_and_free (deps, insn, ®_last->clobbers, 0,
REG_DEP_OUTPUT, false);
add_dependence_list_and_free (deps, insn, ®_last->uses, 0,
REG_DEP_ANTI, false);
add_dependence_list (insn, reg_last->control_uses, 0,
REG_DEP_CONTROL, false);
if (!deps->readonly)
{
reg_last->sets = alloc_INSN_LIST (insn, reg_last->sets);
reg_last->uses_length = 0;
reg_last->clobbers_length = 0;
}
}
}
if (!deps->readonly)
{
EXECUTE_IF_SET_IN_REG_SET (reg_pending_control_uses, 0, i, rsi)
{
struct deps_reg *reg_last = &deps->reg_last[i];
reg_last->control_uses
= alloc_INSN_LIST (insn, reg_last->control_uses);
}
}
}
for (i = 0; i < FIRST_PSEUDO_REGISTER; i++)
if (TEST_HARD_REG_BIT (implicit_reg_pending_clobbers, i))
{
struct deps_reg *reg_last = &deps->reg_last[i];
add_dependence_list (insn, reg_last->sets, 0, REG_DEP_ANTI, false);
add_dependence_list (insn, reg_last->clobbers, 0, REG_DEP_ANTI, false);
add_dependence_list (insn, reg_last->uses, 0, REG_DEP_ANTI, false);
add_dependence_list (insn, reg_last->control_uses, 0, REG_DEP_ANTI,
false);
if (!deps->readonly)
reg_last->implicit_sets
= alloc_INSN_LIST (insn, reg_last->implicit_sets);
}
if (!deps->readonly)
{
IOR_REG_SET (&deps->reg_last_in_use, reg_pending_uses);
IOR_REG_SET (&deps->reg_last_in_use, reg_pending_clobbers);
IOR_REG_SET (&deps->reg_last_in_use, reg_pending_sets);
for (i = 0; i < FIRST_PSEUDO_REGISTER; i++)
if (TEST_HARD_REG_BIT (implicit_reg_pending_uses, i)
|| TEST_HARD_REG_BIT (implicit_reg_pending_clobbers, i))
SET_REGNO_REG_SET (&deps->reg_last_in_use, i);
/* Set up the pending barrier found. */
deps->last_reg_pending_barrier = reg_pending_barrier;
}
CLEAR_REG_SET (reg_pending_uses);
CLEAR_REG_SET (reg_pending_clobbers);
CLEAR_REG_SET (reg_pending_sets);
CLEAR_REG_SET (reg_pending_control_uses);
CLEAR_HARD_REG_SET (implicit_reg_pending_clobbers);
CLEAR_HARD_REG_SET (implicit_reg_pending_uses);
/* Add dependencies if a scheduling barrier was found. */
if (reg_pending_barrier)
{
/* In the case of barrier the most added dependencies are not
real, so we use anti-dependence here. */
if (sched_has_condition_p (insn))
{
EXECUTE_IF_SET_IN_REG_SET (&deps->reg_last_in_use, 0, i, rsi)
{
struct deps_reg *reg_last = &deps->reg_last[i];
add_dependence_list (insn, reg_last->uses, 0, REG_DEP_ANTI,
true);
add_dependence_list (insn, reg_last->sets, 0,
reg_pending_barrier == TRUE_BARRIER
? REG_DEP_TRUE : REG_DEP_ANTI, true);
add_dependence_list (insn, reg_last->implicit_sets, 0,
REG_DEP_ANTI, true);
add_dependence_list (insn, reg_last->clobbers, 0,
reg_pending_barrier == TRUE_BARRIER
? REG_DEP_TRUE : REG_DEP_ANTI, true);
}
}
else
{
EXECUTE_IF_SET_IN_REG_SET (&deps->reg_last_in_use, 0, i, rsi)
{
struct deps_reg *reg_last = &deps->reg_last[i];
add_dependence_list_and_free (deps, insn, ®_last->uses, 0,
REG_DEP_ANTI, true);
add_dependence_list_and_free (deps, insn,
®_last->control_uses, 0,
REG_DEP_CONTROL, true);
add_dependence_list_and_free (deps, insn, ®_last->sets, 0,
reg_pending_barrier == TRUE_BARRIER
? REG_DEP_TRUE : REG_DEP_ANTI,
true);
add_dependence_list_and_free (deps, insn,
®_last->implicit_sets, 0,
REG_DEP_ANTI, true);
add_dependence_list_and_free (deps, insn, ®_last->clobbers, 0,
reg_pending_barrier == TRUE_BARRIER
? REG_DEP_TRUE : REG_DEP_ANTI,
true);
if (!deps->readonly)
{
reg_last->uses_length = 0;
reg_last->clobbers_length = 0;
}
}
}
if (!deps->readonly)
for (i = 0; i < (unsigned)deps->max_reg; i++)
{
struct deps_reg *reg_last = &deps->reg_last[i];
reg_last->sets = alloc_INSN_LIST (insn, reg_last->sets);
SET_REGNO_REG_SET (&deps->reg_last_in_use, i);
}
/* Don't flush pending lists on speculative checks for
selective scheduling. */
if (!sel_sched_p () || !sel_insn_is_speculation_check (insn))
flush_pending_lists (deps, insn, true, true);
reg_pending_barrier = NOT_A_BARRIER;
}
/* If a post-call group is still open, see if it should remain so.
This insn must be a simple move of a hard reg to a pseudo or
vice-versa.
We must avoid moving these insns for correctness on targets
with small register classes, and for special registers like
PIC_OFFSET_TABLE_REGNUM. For simplicity, extend this to all
hard regs for all targets. */
if (deps->in_post_call_group_p)
{
rtx tmp, set = single_set (insn);
int src_regno, dest_regno;
if (set == NULL)
{
if (DEBUG_INSN_P (insn))
/* We don't want to mark debug insns as part of the same
sched group. We know they really aren't, but if we use
debug insns to tell that a call group is over, we'll
get different code if debug insns are not there and
instructions that follow seem like they should be part
of the call group.
Also, if we did, chain_to_prev_insn would move the
deps of the debug insn to the call insn, modifying
non-debug post-dependency counts of the debug insn
dependencies and otherwise messing with the scheduling
order.
Instead, let such debug insns be scheduled freely, but
keep the call group open in case there are insns that
should be part of it afterwards. Since we grant debug
insns higher priority than even sched group insns, it
will all turn out all right. */
goto debug_dont_end_call_group;
else
goto end_call_group;
}
tmp = SET_DEST (set);
if (GET_CODE (tmp) == SUBREG)
tmp = SUBREG_REG (tmp);
if (REG_P (tmp))
dest_regno = REGNO (tmp);
else
goto end_call_group;
tmp = SET_SRC (set);
if (GET_CODE (tmp) == SUBREG)
tmp = SUBREG_REG (tmp);
if ((GET_CODE (tmp) == PLUS
|| GET_CODE (tmp) == MINUS)
&& REG_P (XEXP (tmp, 0))
&& REGNO (XEXP (tmp, 0)) == STACK_POINTER_REGNUM
&& dest_regno == STACK_POINTER_REGNUM)
src_regno = STACK_POINTER_REGNUM;
else if (REG_P (tmp))
src_regno = REGNO (tmp);
else
goto end_call_group;
if (src_regno < FIRST_PSEUDO_REGISTER
|| dest_regno < FIRST_PSEUDO_REGISTER)
{
if (!deps->readonly
&& deps->in_post_call_group_p == post_call_initial)
deps->in_post_call_group_p = post_call;
if (!sel_sched_p () || sched_emulate_haifa_p)
{
SCHED_GROUP_P (insn) = 1;
CANT_MOVE (insn) = 1;
}
}
else
{
end_call_group:
if (!deps->readonly)
deps->in_post_call_group_p = not_post_call;
}
}
debug_dont_end_call_group:
if ((current_sched_info->flags & DO_SPECULATION)
&& !sched_insn_is_legitimate_for_speculation_p (insn, 0))
/* INSN has an internal dependency (e.g. r14 = [r14]) and thus cannot
be speculated. */
{
if (sel_sched_p ())
sel_mark_hard_insn (insn);
else
{
sd_iterator_def sd_it;
dep_t dep;
for (sd_it = sd_iterator_start (insn, SD_LIST_SPEC_BACK);
sd_iterator_cond (&sd_it, &dep);)
change_spec_dep_to_hard (sd_it);
}
}
/* We do not yet have code to adjust REG_ARGS_SIZE, therefore we must
honor their original ordering. */
if (find_reg_note (insn, REG_ARGS_SIZE, NULL))
{
if (deps->last_args_size)
add_dependence (insn, deps->last_args_size, REG_DEP_OUTPUT);
if (!deps->readonly)
deps->last_args_size = insn;
}
/* We must not mix prologue and epilogue insns. See PR78029. */
if (prologue_contains (insn))
{
add_dependence_list (insn, deps->last_epilogue, true, REG_DEP_ANTI, true);
if (!deps->readonly)
{
if (deps->last_logue_was_epilogue)
free_INSN_LIST_list (&deps->last_prologue);
deps->last_prologue = alloc_INSN_LIST (insn, deps->last_prologue);
deps->last_logue_was_epilogue = false;
}
}
if (epilogue_contains (insn))
{
add_dependence_list (insn, deps->last_prologue, true, REG_DEP_ANTI, true);
if (!deps->readonly)
{
if (!deps->last_logue_was_epilogue)
free_INSN_LIST_list (&deps->last_epilogue);
deps->last_epilogue = alloc_INSN_LIST (insn, deps->last_epilogue);
deps->last_logue_was_epilogue = true;
}
}
}
/* Return TRUE if INSN might not always return normally (e.g. call exit,
longjmp, loop forever, ...). */
/* FIXME: Why can't this function just use flags_from_decl_or_type and
test for ECF_NORETURN? */
static bool
call_may_noreturn_p (rtx_insn *insn)
{
rtx call;
/* const or pure calls that aren't looping will always return. */
if (RTL_CONST_OR_PURE_CALL_P (insn)
&& !RTL_LOOPING_CONST_OR_PURE_CALL_P (insn))
return false;
call = get_call_rtx_from (insn);
if (call && GET_CODE (XEXP (XEXP (call, 0), 0)) == SYMBOL_REF)
{
rtx symbol = XEXP (XEXP (call, 0), 0);
if (SYMBOL_REF_DECL (symbol)
&& TREE_CODE (SYMBOL_REF_DECL (symbol)) == FUNCTION_DECL)
{
if (DECL_BUILT_IN_CLASS (SYMBOL_REF_DECL (symbol))
== BUILT_IN_NORMAL)
switch (DECL_FUNCTION_CODE (SYMBOL_REF_DECL (symbol)))
{
case BUILT_IN_BCMP:
case BUILT_IN_BCOPY:
case BUILT_IN_BZERO:
case BUILT_IN_INDEX:
case BUILT_IN_MEMCHR:
case BUILT_IN_MEMCMP:
case BUILT_IN_MEMCPY:
case BUILT_IN_MEMMOVE:
case BUILT_IN_MEMPCPY:
case BUILT_IN_MEMSET:
case BUILT_IN_RINDEX:
case BUILT_IN_STPCPY:
case BUILT_IN_STPNCPY:
case BUILT_IN_STRCAT:
case BUILT_IN_STRCHR:
case BUILT_IN_STRCMP:
case BUILT_IN_STRCPY:
case BUILT_IN_STRCSPN:
case BUILT_IN_STRLEN:
case BUILT_IN_STRNCAT:
case BUILT_IN_STRNCMP:
case BUILT_IN_STRNCPY:
case BUILT_IN_STRPBRK:
case BUILT_IN_STRRCHR:
case BUILT_IN_STRSPN:
case BUILT_IN_STRSTR:
/* Assume certain string/memory builtins always return. */
return false;
default:
break;
}
}
}
/* For all other calls assume that they might not always return. */
return true;
}
/* Return true if INSN should be made dependent on the previous instruction
group, and if all INSN's dependencies should be moved to the first
instruction of that group. */
static bool
chain_to_prev_insn_p (rtx_insn *insn)
{
/* INSN forms a group with the previous instruction. */
if (SCHED_GROUP_P (insn))
return true;
/* If the previous instruction clobbers a register R and this one sets
part of R, the clobber was added specifically to help us track the
liveness of R. There's no point scheduling the clobber and leaving
INSN behind, especially if we move the clobber to another block. */
rtx_insn *prev = prev_nonnote_nondebug_insn (insn);
if (prev
&& INSN_P (prev)
&& BLOCK_FOR_INSN (prev) == BLOCK_FOR_INSN (insn)
&& GET_CODE (PATTERN (prev)) == CLOBBER)
{
rtx x = XEXP (PATTERN (prev), 0);
if (set_of (x, insn))
return true;
}
return false;
}
/* Analyze INSN with DEPS as a context. */
void
deps_analyze_insn (struct deps_desc *deps, rtx_insn *insn)
{
if (sched_deps_info->start_insn)
sched_deps_info->start_insn (insn);
/* Record the condition for this insn. */
if (NONDEBUG_INSN_P (insn))
{
rtx t;
sched_get_condition_with_rev (insn, NULL);
t = INSN_CACHED_COND (insn);
INSN_COND_DEPS (insn) = NULL;
if (reload_completed
&& (current_sched_info->flags & DO_PREDICATION)
&& COMPARISON_P (t)
&& REG_P (XEXP (t, 0))
&& CONSTANT_P (XEXP (t, 1)))
{
unsigned int regno;
int nregs;
rtx_insn_list *cond_deps = NULL;
t = XEXP (t, 0);
regno = REGNO (t);
nregs = REG_NREGS (t);
while (nregs-- > 0)
{
struct deps_reg *reg_last = &deps->reg_last[regno + nregs];
cond_deps = concat_INSN_LIST (reg_last->sets, cond_deps);
cond_deps = concat_INSN_LIST (reg_last->clobbers, cond_deps);
cond_deps = concat_INSN_LIST (reg_last->implicit_sets, cond_deps);
}
INSN_COND_DEPS (insn) = cond_deps;
}
}
if (JUMP_P (insn))
{
/* Make each JUMP_INSN (but not a speculative check)
a scheduling barrier for memory references. */
if (!deps->readonly
&& !(sel_sched_p ()
&& sel_insn_is_speculation_check (insn)))
{
/* Keep the list a reasonable size. */
if (deps->pending_flush_length++ >= MAX_PENDING_LIST_LENGTH)
flush_pending_lists (deps, insn, true, true);
else
deps->pending_jump_insns
= alloc_INSN_LIST (insn, deps->pending_jump_insns);
}
/* For each insn which shouldn't cross a jump, add a dependence. */
add_dependence_list_and_free (deps, insn,
&deps->sched_before_next_jump, 1,
REG_DEP_ANTI, true);
sched_analyze_insn (deps, PATTERN (insn), insn);
}
else if (NONJUMP_INSN_P (insn) || DEBUG_INSN_P (insn))
{
sched_analyze_insn (deps, PATTERN (insn), insn);
}
else if (CALL_P (insn))
{
int i;
CANT_MOVE (insn) = 1;
if (find_reg_note (insn, REG_SETJMP, NULL))
{
/* This is setjmp. Assume that all registers, not just
hard registers, may be clobbered by this call. */
reg_pending_barrier = MOVE_BARRIER;
}
else
{
for (i = 0; i < FIRST_PSEUDO_REGISTER; i++)
/* A call may read and modify global register variables. */
if (global_regs[i])
{
SET_REGNO_REG_SET (reg_pending_sets, i);
SET_HARD_REG_BIT (implicit_reg_pending_uses, i);
}
/* Other call-clobbered hard regs may be clobbered.
Since we only have a choice between 'might be clobbered'
and 'definitely not clobbered', we must include all
partly call-clobbered registers here. */
else if (targetm.hard_regno_call_part_clobbered (i,
reg_raw_mode[i])
|| TEST_HARD_REG_BIT (regs_invalidated_by_call, i))
SET_REGNO_REG_SET (reg_pending_clobbers, i);
/* We don't know what set of fixed registers might be used
by the function, but it is certain that the stack pointer
is among them, but be conservative. */
else if (fixed_regs[i])
SET_HARD_REG_BIT (implicit_reg_pending_uses, i);
/* The frame pointer is normally not used by the function
itself, but by the debugger. */
/* ??? MIPS o32 is an exception. It uses the frame pointer
in the macro expansion of jal but does not represent this
fact in the call_insn rtl. */
else if (i == FRAME_POINTER_REGNUM
|| (i == HARD_FRAME_POINTER_REGNUM
&& (! reload_completed || frame_pointer_needed)))
SET_HARD_REG_BIT (implicit_reg_pending_uses, i);
}
/* For each insn which shouldn't cross a call, add a dependence
between that insn and this call insn. */
add_dependence_list_and_free (deps, insn,
&deps->sched_before_next_call, 1,
REG_DEP_ANTI, true);
sched_analyze_insn (deps, PATTERN (insn), insn);
/* If CALL would be in a sched group, then this will violate
convention that sched group insns have dependencies only on the
previous instruction.
Of course one can say: "Hey! What about head of the sched group?"
And I will answer: "Basic principles (one dep per insn) are always
the same." */
gcc_assert (!SCHED_GROUP_P (insn));
/* In the absence of interprocedural alias analysis, we must flush
all pending reads and writes, and start new dependencies starting
from here. But only flush writes for constant calls (which may
be passed a pointer to something we haven't written yet). */
flush_pending_lists (deps, insn, true, ! RTL_CONST_OR_PURE_CALL_P (insn));
if (!deps->readonly)
{
/* Remember the last function call for limiting lifetimes. */
free_INSN_LIST_list (&deps->last_function_call);
deps->last_function_call = alloc_INSN_LIST (insn, NULL_RTX);
if (call_may_noreturn_p (insn))
{
/* Remember the last function call that might not always return
normally for limiting moves of trapping insns. */
free_INSN_LIST_list (&deps->last_function_call_may_noreturn);
deps->last_function_call_may_noreturn
= alloc_INSN_LIST (insn, NULL_RTX);
}
/* Before reload, begin a post-call group, so as to keep the
lifetimes of hard registers correct. */
if (! reload_completed)
deps->in_post_call_group_p = post_call;
}
}
if (sched_deps_info->use_cselib)
cselib_process_insn (insn);
if (sched_deps_info->finish_insn)
sched_deps_info->finish_insn ();
/* Fixup the dependencies in the sched group. */
if ((NONJUMP_INSN_P (insn) || JUMP_P (insn))
&& chain_to_prev_insn_p (insn)
&& !sel_sched_p ())
chain_to_prev_insn (insn);
}
/* Initialize DEPS for the new block beginning with HEAD. */
void
deps_start_bb (struct deps_desc *deps, rtx_insn *head)
{
gcc_assert (!deps->readonly);
/* Before reload, if the previous block ended in a call, show that
we are inside a post-call group, so as to keep the lifetimes of
hard registers correct. */
if (! reload_completed && !LABEL_P (head))
{
rtx_insn *insn = prev_nonnote_nondebug_insn (head);
if (insn && CALL_P (insn))
deps->in_post_call_group_p = post_call_initial;
}
}
/* Analyze every insn between HEAD and TAIL inclusive, creating backward
dependencies for each insn. */
void
sched_analyze (struct deps_desc *deps, rtx_insn *head, rtx_insn *tail)
{
rtx_insn *insn;
if (sched_deps_info->use_cselib)
cselib_init (CSELIB_RECORD_MEMORY);
deps_start_bb (deps, head);
for (insn = head;; insn = NEXT_INSN (insn))
{
if (INSN_P (insn))
{
/* And initialize deps_lists. */
sd_init_insn (insn);
/* Clean up SCHED_GROUP_P which may be set by last
scheduler pass. */
if (SCHED_GROUP_P (insn))
SCHED_GROUP_P (insn) = 0;
}
deps_analyze_insn (deps, insn);
if (insn == tail)
{
if (sched_deps_info->use_cselib)
cselib_finish ();
return;
}
}
gcc_unreachable ();
}
/* Helper for sched_free_deps ().
Delete INSN's (RESOLVED_P) backward dependencies. */
static void
delete_dep_nodes_in_back_deps (rtx_insn *insn, bool resolved_p)
{
sd_iterator_def sd_it;
dep_t dep;
sd_list_types_def types;
if (resolved_p)
types = SD_LIST_RES_BACK;
else
types = SD_LIST_BACK;
for (sd_it = sd_iterator_start (insn, types);
sd_iterator_cond (&sd_it, &dep);)
{
dep_link_t link = *sd_it.linkp;
dep_node_t node = DEP_LINK_NODE (link);
deps_list_t back_list;
deps_list_t forw_list;
get_back_and_forw_lists (dep, resolved_p, &back_list, &forw_list);
remove_from_deps_list (link, back_list);
delete_dep_node (node);
}
}
/* Delete (RESOLVED_P) dependencies between HEAD and TAIL together with
deps_lists. */
void
sched_free_deps (rtx_insn *head, rtx_insn *tail, bool resolved_p)
{
rtx_insn *insn;
rtx_insn *next_tail = NEXT_INSN (tail);
/* We make two passes since some insns may be scheduled before their
dependencies are resolved. */
for (insn = head; insn != next_tail; insn = NEXT_INSN (insn))
if (INSN_P (insn) && INSN_LUID (insn) > 0)
{
/* Clear forward deps and leave the dep_nodes to the
corresponding back_deps list. */
if (resolved_p)
clear_deps_list (INSN_RESOLVED_FORW_DEPS (insn));
else
clear_deps_list (INSN_FORW_DEPS (insn));
}
for (insn = head; insn != next_tail; insn = NEXT_INSN (insn))
if (INSN_P (insn) && INSN_LUID (insn) > 0)
{
/* Clear resolved back deps together with its dep_nodes. */
delete_dep_nodes_in_back_deps (insn, resolved_p);
sd_finish_insn (insn);
}
}
/* Initialize variables for region data dependence analysis.
When LAZY_REG_LAST is true, do not allocate reg_last array
of struct deps_desc immediately. */
void
init_deps (struct deps_desc *deps, bool lazy_reg_last)
{
int max_reg = (reload_completed ? FIRST_PSEUDO_REGISTER : max_reg_num ());
deps->max_reg = max_reg;
if (lazy_reg_last)
deps->reg_last = NULL;
else
deps->reg_last = XCNEWVEC (struct deps_reg, max_reg);
INIT_REG_SET (&deps->reg_last_in_use);
deps->pending_read_insns = 0;
deps->pending_read_mems = 0;
deps->pending_write_insns = 0;
deps->pending_write_mems = 0;
deps->pending_jump_insns = 0;
deps->pending_read_list_length = 0;
deps->pending_write_list_length = 0;
deps->pending_flush_length = 0;
deps->last_pending_memory_flush = 0;
deps->last_function_call = 0;
deps->last_function_call_may_noreturn = 0;
deps->sched_before_next_call = 0;
deps->sched_before_next_jump = 0;
deps->in_post_call_group_p = not_post_call;
deps->last_debug_insn = 0;
deps->last_args_size = 0;
deps->last_prologue = 0;
deps->last_epilogue = 0;
deps->last_logue_was_epilogue = false;
deps->last_reg_pending_barrier = NOT_A_BARRIER;
deps->readonly = 0;
}
/* Init only reg_last field of DEPS, which was not allocated before as
we inited DEPS lazily. */
void
init_deps_reg_last (struct deps_desc *deps)
{
gcc_assert (deps && deps->max_reg > 0);
gcc_assert (deps->reg_last == NULL);
deps->reg_last = XCNEWVEC (struct deps_reg, deps->max_reg);
}
/* Free insn lists found in DEPS. */
void
free_deps (struct deps_desc *deps)
{
unsigned i;
reg_set_iterator rsi;
/* We set max_reg to 0 when this context was already freed. */
if (deps->max_reg == 0)
{
gcc_assert (deps->reg_last == NULL);
return;
}
deps->max_reg = 0;
free_INSN_LIST_list (&deps->pending_read_insns);
free_EXPR_LIST_list (&deps->pending_read_mems);
free_INSN_LIST_list (&deps->pending_write_insns);
free_EXPR_LIST_list (&deps->pending_write_mems);
free_INSN_LIST_list (&deps->last_pending_memory_flush);
/* Without the EXECUTE_IF_SET, this loop is executed max_reg * nr_regions
times. For a testcase with 42000 regs and 8000 small basic blocks,
this loop accounted for nearly 60% (84 sec) of the total -O2 runtime. */
EXECUTE_IF_SET_IN_REG_SET (&deps->reg_last_in_use, 0, i, rsi)
{
struct deps_reg *reg_last = &deps->reg_last[i];
if (reg_last->uses)
free_INSN_LIST_list (®_last->uses);
if (reg_last->sets)
free_INSN_LIST_list (®_last->sets);
if (reg_last->implicit_sets)
free_INSN_LIST_list (®_last->implicit_sets);
if (reg_last->control_uses)
free_INSN_LIST_list (®_last->control_uses);
if (reg_last->clobbers)
free_INSN_LIST_list (®_last->clobbers);
}
CLEAR_REG_SET (&deps->reg_last_in_use);
/* As we initialize reg_last lazily, it is possible that we didn't allocate
it at all. */
free (deps->reg_last);
deps->reg_last = NULL;
deps = NULL;
}
/* Remove INSN from dependence contexts DEPS. */
void
remove_from_deps (struct deps_desc *deps, rtx_insn *insn)
{
int removed;
unsigned i;
reg_set_iterator rsi;
removed = remove_from_both_dependence_lists (insn, &deps->pending_read_insns,
&deps->pending_read_mems);
if (!DEBUG_INSN_P (insn))
deps->pending_read_list_length -= removed;
removed = remove_from_both_dependence_lists (insn, &deps->pending_write_insns,
&deps->pending_write_mems);
deps->pending_write_list_length -= removed;
removed = remove_from_dependence_list (insn, &deps->pending_jump_insns);
deps->pending_flush_length -= removed;
removed = remove_from_dependence_list (insn, &deps->last_pending_memory_flush);
deps->pending_flush_length -= removed;
unsigned to_clear = -1U;
EXECUTE_IF_SET_IN_REG_SET (&deps->reg_last_in_use, 0, i, rsi)
{
if (to_clear != -1U)
{
CLEAR_REGNO_REG_SET (&deps->reg_last_in_use, to_clear);
to_clear = -1U;
}
struct deps_reg *reg_last = &deps->reg_last[i];
if (reg_last->uses)
remove_from_dependence_list (insn, ®_last->uses);
if (reg_last->sets)
remove_from_dependence_list (insn, ®_last->sets);
if (reg_last->implicit_sets)
remove_from_dependence_list (insn, ®_last->implicit_sets);
if (reg_last->clobbers)
remove_from_dependence_list (insn, ®_last->clobbers);
if (!reg_last->uses && !reg_last->sets && !reg_last->implicit_sets
&& !reg_last->clobbers)
to_clear = i;
}
if (to_clear != -1U)
CLEAR_REGNO_REG_SET (&deps->reg_last_in_use, to_clear);
if (CALL_P (insn))
{
remove_from_dependence_list (insn, &deps->last_function_call);
remove_from_dependence_list (insn,
&deps->last_function_call_may_noreturn);
}
remove_from_dependence_list (insn, &deps->sched_before_next_call);
}
/* Init deps data vector. */
static void
init_deps_data_vector (void)
{
int reserve = (sched_max_luid + 1 - h_d_i_d.length ());
if (reserve > 0 && ! h_d_i_d.space (reserve))
h_d_i_d.safe_grow_cleared (3 * sched_max_luid / 2);
}
/* If it is profitable to use them, initialize or extend (depending on
GLOBAL_P) dependency data. */
void
sched_deps_init (bool global_p)
{
/* Average number of insns in the basic block.
'+ 1' is used to make it nonzero. */
int insns_in_block = sched_max_luid / n_basic_blocks_for_fn (cfun) + 1;
init_deps_data_vector ();
/* We use another caching mechanism for selective scheduling, so
we don't use this one. */
if (!sel_sched_p () && global_p && insns_in_block > 100 * 5)
{
/* ?!? We could save some memory by computing a per-region luid mapping
which could reduce both the number of vectors in the cache and the
size of each vector. Instead we just avoid the cache entirely unless
the average number of instructions in a basic block is very high. See
the comment before the declaration of true_dependency_cache for
what we consider "very high". */
cache_size = 0;
extend_dependency_caches (sched_max_luid, true);
}
if (global_p)
{
dl_pool = new object_allocator<_deps_list> ("deps_list");
/* Allocate lists for one block at a time. */
dn_pool = new object_allocator<_dep_node> ("dep_node");
/* Allocate nodes for one block at a time. */
}
}
/* Create or extend (depending on CREATE_P) dependency caches to
size N. */
void
extend_dependency_caches (int n, bool create_p)
{
if (create_p || true_dependency_cache)
{
int i, luid = cache_size + n;
true_dependency_cache = XRESIZEVEC (bitmap_head, true_dependency_cache,
luid);
output_dependency_cache = XRESIZEVEC (bitmap_head,
output_dependency_cache, luid);
anti_dependency_cache = XRESIZEVEC (bitmap_head, anti_dependency_cache,
luid);
control_dependency_cache = XRESIZEVEC (bitmap_head, control_dependency_cache,
luid);
if (current_sched_info->flags & DO_SPECULATION)
spec_dependency_cache = XRESIZEVEC (bitmap_head, spec_dependency_cache,
luid);
for (i = cache_size; i < luid; i++)
{
bitmap_initialize (&true_dependency_cache[i], 0);
bitmap_initialize (&output_dependency_cache[i], 0);
bitmap_initialize (&anti_dependency_cache[i], 0);
bitmap_initialize (&control_dependency_cache[i], 0);
if (current_sched_info->flags & DO_SPECULATION)
bitmap_initialize (&spec_dependency_cache[i], 0);
}
cache_size = luid;
}
}
/* Finalize dependency information for the whole function. */
void
sched_deps_finish (void)
{
gcc_assert (deps_pools_are_empty_p ());
delete dn_pool;
delete dl_pool;
dn_pool = NULL;
dl_pool = NULL;
h_d_i_d.release ();
cache_size = 0;
if (true_dependency_cache)
{
int i;
for (i = 0; i < cache_size; i++)
{
bitmap_clear (&true_dependency_cache[i]);
bitmap_clear (&output_dependency_cache[i]);
bitmap_clear (&anti_dependency_cache[i]);
bitmap_clear (&control_dependency_cache[i]);
if (sched_deps_info->generate_spec_deps)
bitmap_clear (&spec_dependency_cache[i]);
}
free (true_dependency_cache);
true_dependency_cache = NULL;
free (output_dependency_cache);
output_dependency_cache = NULL;
free (anti_dependency_cache);
anti_dependency_cache = NULL;
free (control_dependency_cache);
control_dependency_cache = NULL;
if (sched_deps_info->generate_spec_deps)
{
free (spec_dependency_cache);
spec_dependency_cache = NULL;
}
}
}
/* Initialize some global variables needed by the dependency analysis
code. */
void
init_deps_global (void)
{
CLEAR_HARD_REG_SET (implicit_reg_pending_clobbers);
CLEAR_HARD_REG_SET (implicit_reg_pending_uses);
reg_pending_sets = ALLOC_REG_SET (®_obstack);
reg_pending_clobbers = ALLOC_REG_SET (®_obstack);
reg_pending_uses = ALLOC_REG_SET (®_obstack);
reg_pending_control_uses = ALLOC_REG_SET (®_obstack);
reg_pending_barrier = NOT_A_BARRIER;
if (!sel_sched_p () || sched_emulate_haifa_p)
{
sched_deps_info->start_insn = haifa_start_insn;
sched_deps_info->finish_insn = haifa_finish_insn;
sched_deps_info->note_reg_set = haifa_note_reg_set;
sched_deps_info->note_reg_clobber = haifa_note_reg_clobber;
sched_deps_info->note_reg_use = haifa_note_reg_use;
sched_deps_info->note_mem_dep = haifa_note_mem_dep;
sched_deps_info->note_dep = haifa_note_dep;
}
}
/* Free everything used by the dependency analysis code. */
void
finish_deps_global (void)
{
FREE_REG_SET (reg_pending_sets);
FREE_REG_SET (reg_pending_clobbers);
FREE_REG_SET (reg_pending_uses);
FREE_REG_SET (reg_pending_control_uses);
}
/* Estimate the weakness of dependence between MEM1 and MEM2. */
dw_t
estimate_dep_weak (rtx mem1, rtx mem2)
{
if (mem1 == mem2)
/* MEMs are the same - don't speculate. */
return MIN_DEP_WEAK;
rtx r1 = XEXP (mem1, 0);
rtx r2 = XEXP (mem2, 0);
if (sched_deps_info->use_cselib)
{
/* We cannot call rtx_equal_for_cselib_p because the VALUEs might be
dangling at this point, since we never preserve them. Instead we
canonicalize manually to get stable VALUEs out of hashing. */
if (GET_CODE (r1) == VALUE && CSELIB_VAL_PTR (r1))
r1 = canonical_cselib_val (CSELIB_VAL_PTR (r1))->val_rtx;
if (GET_CODE (r2) == VALUE && CSELIB_VAL_PTR (r2))
r2 = canonical_cselib_val (CSELIB_VAL_PTR (r2))->val_rtx;
}
if (r1 == r2
|| (REG_P (r1) && REG_P (r2) && REGNO (r1) == REGNO (r2)))
/* Again, MEMs are the same. */
return MIN_DEP_WEAK;
else if ((REG_P (r1) && !REG_P (r2)) || (!REG_P (r1) && REG_P (r2)))
/* Different addressing modes - reason to be more speculative,
than usual. */
return NO_DEP_WEAK - (NO_DEP_WEAK - UNCERTAIN_DEP_WEAK) / 2;
else
/* We can't say anything about the dependence. */
return UNCERTAIN_DEP_WEAK;
}
/* Add or update backward dependence between INSN and ELEM with type DEP_TYPE.
This function can handle same INSN and ELEM (INSN == ELEM).
It is a convenience wrapper. */
static void
add_dependence_1 (rtx_insn *insn, rtx_insn *elem, enum reg_note dep_type)
{
ds_t ds;
bool internal;
if (dep_type == REG_DEP_TRUE)
ds = DEP_TRUE;
else if (dep_type == REG_DEP_OUTPUT)
ds = DEP_OUTPUT;
else if (dep_type == REG_DEP_CONTROL)
ds = DEP_CONTROL;
else
{
gcc_assert (dep_type == REG_DEP_ANTI);
ds = DEP_ANTI;
}
/* When add_dependence is called from inside sched-deps.c, we expect
cur_insn to be non-null. */
internal = cur_insn != NULL;
if (internal)
gcc_assert (insn == cur_insn);
else
cur_insn = insn;
note_dep (elem, ds);
if (!internal)
cur_insn = NULL;
}
/* Return weakness of speculative type TYPE in the dep_status DS,
without checking to prevent ICEs on malformed input. */
static dw_t
get_dep_weak_1 (ds_t ds, ds_t type)
{
ds = ds & type;
switch (type)
{
case BEGIN_DATA: ds >>= BEGIN_DATA_BITS_OFFSET; break;
case BE_IN_DATA: ds >>= BE_IN_DATA_BITS_OFFSET; break;
case BEGIN_CONTROL: ds >>= BEGIN_CONTROL_BITS_OFFSET; break;
case BE_IN_CONTROL: ds >>= BE_IN_CONTROL_BITS_OFFSET; break;
default: gcc_unreachable ();
}
return (dw_t) ds;
}
/* Return weakness of speculative type TYPE in the dep_status DS. */
dw_t
get_dep_weak (ds_t ds, ds_t type)
{
dw_t dw = get_dep_weak_1 (ds, type);
gcc_assert (MIN_DEP_WEAK <= dw && dw <= MAX_DEP_WEAK);
return dw;
}
/* Return the dep_status, which has the same parameters as DS, except for
speculative type TYPE, that will have weakness DW. */
ds_t
set_dep_weak (ds_t ds, ds_t type, dw_t dw)
{
gcc_assert (MIN_DEP_WEAK <= dw && dw <= MAX_DEP_WEAK);
ds &= ~type;
switch (type)
{
case BEGIN_DATA: ds |= ((ds_t) dw) << BEGIN_DATA_BITS_OFFSET; break;
case BE_IN_DATA: ds |= ((ds_t) dw) << BE_IN_DATA_BITS_OFFSET; break;
case BEGIN_CONTROL: ds |= ((ds_t) dw) << BEGIN_CONTROL_BITS_OFFSET; break;
case BE_IN_CONTROL: ds |= ((ds_t) dw) << BE_IN_CONTROL_BITS_OFFSET; break;
default: gcc_unreachable ();
}
return ds;
}
/* Return the join of two dep_statuses DS1 and DS2.
If MAX_P is true then choose the greater probability,
otherwise multiply probabilities.
This function assumes that both DS1 and DS2 contain speculative bits. */
static ds_t
ds_merge_1 (ds_t ds1, ds_t ds2, bool max_p)
{
ds_t ds, t;
gcc_assert ((ds1 & SPECULATIVE) && (ds2 & SPECULATIVE));
ds = (ds1 & DEP_TYPES) | (ds2 & DEP_TYPES);
t = FIRST_SPEC_TYPE;
do
{
if ((ds1 & t) && !(ds2 & t))
ds |= ds1 & t;
else if (!(ds1 & t) && (ds2 & t))
ds |= ds2 & t;
else if ((ds1 & t) && (ds2 & t))
{
dw_t dw1 = get_dep_weak (ds1, t);
dw_t dw2 = get_dep_weak (ds2, t);
ds_t dw;
if (!max_p)
{
dw = ((ds_t) dw1) * ((ds_t) dw2);
dw /= MAX_DEP_WEAK;
if (dw < MIN_DEP_WEAK)
dw = MIN_DEP_WEAK;
}
else
{
if (dw1 >= dw2)
dw = dw1;
else
dw = dw2;
}
ds = set_dep_weak (ds, t, (dw_t) dw);
}
if (t == LAST_SPEC_TYPE)
break;
t <<= SPEC_TYPE_SHIFT;
}
while (1);
return ds;
}
/* Return the join of two dep_statuses DS1 and DS2.
This function assumes that both DS1 and DS2 contain speculative bits. */
ds_t
ds_merge (ds_t ds1, ds_t ds2)
{
return ds_merge_1 (ds1, ds2, false);
}
/* Return the join of two dep_statuses DS1 and DS2. */
ds_t
ds_full_merge (ds_t ds, ds_t ds2, rtx mem1, rtx mem2)
{
ds_t new_status = ds | ds2;
if (new_status & SPECULATIVE)
{
if ((ds && !(ds & SPECULATIVE))
|| (ds2 && !(ds2 & SPECULATIVE)))
/* Then this dep can't be speculative. */
new_status &= ~SPECULATIVE;
else
{
/* Both are speculative. Merging probabilities. */
if (mem1)
{
dw_t dw;
dw = estimate_dep_weak (mem1, mem2);
ds = set_dep_weak (ds, BEGIN_DATA, dw);
}
if (!ds)
new_status = ds2;
else if (!ds2)
new_status = ds;
else
new_status = ds_merge (ds2, ds);
}
}
return new_status;
}
/* Return the join of DS1 and DS2. Use maximum instead of multiplying
probabilities. */
ds_t
ds_max_merge (ds_t ds1, ds_t ds2)
{
if (ds1 == 0 && ds2 == 0)
return 0;
if (ds1 == 0 && ds2 != 0)
return ds2;
if (ds1 != 0 && ds2 == 0)
return ds1;
return ds_merge_1 (ds1, ds2, true);
}
/* Return the probability of speculation success for the speculation
status DS. */
dw_t
ds_weak (ds_t ds)
{
ds_t res = 1, dt;
int n = 0;
dt = FIRST_SPEC_TYPE;
do
{
if (ds & dt)
{
res *= (ds_t) get_dep_weak (ds, dt);
n++;
}
if (dt == LAST_SPEC_TYPE)
break;
dt <<= SPEC_TYPE_SHIFT;
}
while (1);
gcc_assert (n);
while (--n)
res /= MAX_DEP_WEAK;
if (res < MIN_DEP_WEAK)
res = MIN_DEP_WEAK;
gcc_assert (res <= MAX_DEP_WEAK);
return (dw_t) res;
}
/* Return a dep status that contains all speculation types of DS. */
ds_t
ds_get_speculation_types (ds_t ds)
{
if (ds & BEGIN_DATA)
ds |= BEGIN_DATA;
if (ds & BE_IN_DATA)
ds |= BE_IN_DATA;
if (ds & BEGIN_CONTROL)
ds |= BEGIN_CONTROL;
if (ds & BE_IN_CONTROL)
ds |= BE_IN_CONTROL;
return ds & SPECULATIVE;
}
/* Return a dep status that contains maximal weakness for each speculation
type present in DS. */
ds_t
ds_get_max_dep_weak (ds_t ds)
{
if (ds & BEGIN_DATA)
ds = set_dep_weak (ds, BEGIN_DATA, MAX_DEP_WEAK);
if (ds & BE_IN_DATA)
ds = set_dep_weak (ds, BE_IN_DATA, MAX_DEP_WEAK);
if (ds & BEGIN_CONTROL)
ds = set_dep_weak (ds, BEGIN_CONTROL, MAX_DEP_WEAK);
if (ds & BE_IN_CONTROL)
ds = set_dep_weak (ds, BE_IN_CONTROL, MAX_DEP_WEAK);
return ds;
}
/* Dump information about the dependence status S. */
static void
dump_ds (FILE *f, ds_t s)
{
fprintf (f, "{");
if (s & BEGIN_DATA)
fprintf (f, "BEGIN_DATA: %d; ", get_dep_weak_1 (s, BEGIN_DATA));
if (s & BE_IN_DATA)
fprintf (f, "BE_IN_DATA: %d; ", get_dep_weak_1 (s, BE_IN_DATA));
if (s & BEGIN_CONTROL)
fprintf (f, "BEGIN_CONTROL: %d; ", get_dep_weak_1 (s, BEGIN_CONTROL));
if (s & BE_IN_CONTROL)
fprintf (f, "BE_IN_CONTROL: %d; ", get_dep_weak_1 (s, BE_IN_CONTROL));
if (s & HARD_DEP)
fprintf (f, "HARD_DEP; ");
if (s & DEP_TRUE)
fprintf (f, "DEP_TRUE; ");
if (s & DEP_OUTPUT)
fprintf (f, "DEP_OUTPUT; ");
if (s & DEP_ANTI)
fprintf (f, "DEP_ANTI; ");
if (s & DEP_CONTROL)
fprintf (f, "DEP_CONTROL; ");
fprintf (f, "}");
}
DEBUG_FUNCTION void
debug_ds (ds_t s)
{
dump_ds (stderr, s);
fprintf (stderr, "\n");
}
/* Verify that dependence type and status are consistent.
If RELAXED_P is true, then skip dep_weakness checks. */
static void
check_dep (dep_t dep, bool relaxed_p)
{
enum reg_note dt = DEP_TYPE (dep);
ds_t ds = DEP_STATUS (dep);
gcc_assert (DEP_PRO (dep) != DEP_CON (dep));
if (!(current_sched_info->flags & USE_DEPS_LIST))
{
gcc_assert (ds == 0);
return;
}
/* Check that dependence type contains the same bits as the status. */
if (dt == REG_DEP_TRUE)
gcc_assert (ds & DEP_TRUE);
else if (dt == REG_DEP_OUTPUT)
gcc_assert ((ds & DEP_OUTPUT)
&& !(ds & DEP_TRUE));
else if (dt == REG_DEP_ANTI)
gcc_assert ((ds & DEP_ANTI)
&& !(ds & (DEP_OUTPUT | DEP_TRUE)));
else
gcc_assert (dt == REG_DEP_CONTROL
&& (ds & DEP_CONTROL)
&& !(ds & (DEP_OUTPUT | DEP_ANTI | DEP_TRUE)));
/* HARD_DEP can not appear in dep_status of a link. */
gcc_assert (!(ds & HARD_DEP));
/* Check that dependence status is set correctly when speculation is not
supported. */
if (!sched_deps_info->generate_spec_deps)
gcc_assert (!(ds & SPECULATIVE));
else if (ds & SPECULATIVE)
{
if (!relaxed_p)
{
ds_t type = FIRST_SPEC_TYPE;
/* Check that dependence weakness is in proper range. */
do
{
if (ds & type)
get_dep_weak (ds, type);
if (type == LAST_SPEC_TYPE)
break;
type <<= SPEC_TYPE_SHIFT;
}
while (1);
}
if (ds & BEGIN_SPEC)
{
/* Only true dependence can be data speculative. */
if (ds & BEGIN_DATA)
gcc_assert (ds & DEP_TRUE);
/* Control dependencies in the insn scheduler are represented by
anti-dependencies, therefore only anti dependence can be
control speculative. */
if (ds & BEGIN_CONTROL)
gcc_assert (ds & DEP_ANTI);
}
else
{
/* Subsequent speculations should resolve true dependencies. */
gcc_assert ((ds & DEP_TYPES) == DEP_TRUE);
}
/* Check that true and anti dependencies can't have other speculative
statuses. */
if (ds & DEP_TRUE)
gcc_assert (ds & (BEGIN_DATA | BE_IN_SPEC));
/* An output dependence can't be speculative at all. */
gcc_assert (!(ds & DEP_OUTPUT));
if (ds & DEP_ANTI)
gcc_assert (ds & BEGIN_CONTROL);
}
}
/* The following code discovers opportunities to switch a memory reference
and an increment by modifying the address. We ensure that this is done
only for dependencies that are only used to show a single register
dependence (using DEP_NONREG and DEP_MULTIPLE), and so that every memory
instruction involved is subject to only one dep that can cause a pattern
change.
When we discover a suitable dependency, we fill in the dep_replacement
structure to show how to modify the memory reference. */
/* Holds information about a pair of memory reference and register increment
insns which depend on each other, but could possibly be interchanged. */
struct mem_inc_info
{
rtx_insn *inc_insn;
rtx_insn *mem_insn;
rtx *mem_loc;
/* A register occurring in the memory address for which we wish to break
the dependence. This must be identical to the destination register of
the increment. */
rtx mem_reg0;
/* Any kind of index that is added to that register. */
rtx mem_index;
/* The constant offset used in the memory address. */
HOST_WIDE_INT mem_constant;
/* The constant added in the increment insn. Negated if the increment is
after the memory address. */
HOST_WIDE_INT inc_constant;
/* The source register used in the increment. May be different from mem_reg0
if the increment occurs before the memory address. */
rtx inc_input;
};
/* Verify that the memory location described in MII can be replaced with
one using NEW_ADDR. Return the new memory reference or NULL_RTX. The
insn remains unchanged by this function. */
static rtx
attempt_change (struct mem_inc_info *mii, rtx new_addr)
{
rtx mem = *mii->mem_loc;
rtx new_mem;
/* Jump through a lot of hoops to keep the attributes up to date. We
do not want to call one of the change address variants that take
an offset even though we know the offset in many cases. These
assume you are changing where the address is pointing by the
offset. */
new_mem = replace_equiv_address_nv (mem, new_addr);
if (! validate_change (mii->mem_insn, mii->mem_loc, new_mem, 0))
{
if (sched_verbose >= 5)
fprintf (sched_dump, "validation failure\n");
return NULL_RTX;
}
/* Put back the old one. */
validate_change (mii->mem_insn, mii->mem_loc, mem, 0);
return new_mem;
}
/* Return true if INSN is of a form "a = b op c" where a and b are
regs. op is + if c is a reg and +|- if c is a const. Fill in
informantion in MII about what is found.
BEFORE_MEM indicates whether the increment is found before or after
a corresponding memory reference. */
static bool
parse_add_or_inc (struct mem_inc_info *mii, rtx_insn *insn, bool before_mem)
{
rtx pat = single_set (insn);
rtx src, cst;
bool regs_equal;
if (RTX_FRAME_RELATED_P (insn) || !pat)
return false;
/* Do not allow breaking data dependencies for insns that are marked
with REG_STACK_CHECK. */
if (find_reg_note (insn, REG_STACK_CHECK, NULL))
return false;
/* Result must be single reg. */
if (!REG_P (SET_DEST (pat)))
return false;
if (GET_CODE (SET_SRC (pat)) != PLUS)
return false;
mii->inc_insn = insn;
src = SET_SRC (pat);
mii->inc_input = XEXP (src, 0);
if (!REG_P (XEXP (src, 0)))
return false;
if (!rtx_equal_p (SET_DEST (pat), mii->mem_reg0))
return false;
cst = XEXP (src, 1);
if (!CONST_INT_P (cst))
return false;
mii->inc_constant = INTVAL (cst);
regs_equal = rtx_equal_p (mii->inc_input, mii->mem_reg0);
if (!before_mem)
{
mii->inc_constant = -mii->inc_constant;
if (!regs_equal)
return false;
}
if (regs_equal && REGNO (SET_DEST (pat)) == STACK_POINTER_REGNUM)
{
/* Note that the sign has already been reversed for !before_mem. */
if (STACK_GROWS_DOWNWARD)
return mii->inc_constant > 0;
else
return mii->inc_constant < 0;
}
return true;
}
/* Once a suitable mem reference has been found and the corresponding data
in MII has been filled in, this function is called to find a suitable
add or inc insn involving the register we found in the memory
reference. */
static bool
find_inc (struct mem_inc_info *mii, bool backwards)
{
sd_iterator_def sd_it;
dep_t dep;
sd_it = sd_iterator_start (mii->mem_insn,
backwards ? SD_LIST_HARD_BACK : SD_LIST_FORW);
while (sd_iterator_cond (&sd_it, &dep))
{
dep_node_t node = DEP_LINK_NODE (*sd_it.linkp);
rtx_insn *pro = DEP_PRO (dep);
rtx_insn *con = DEP_CON (dep);
rtx_insn *inc_cand = backwards ? pro : con;
if (DEP_NONREG (dep) || DEP_MULTIPLE (dep))
goto next;
if (parse_add_or_inc (mii, inc_cand, backwards))
{
struct dep_replacement *desc;
df_ref def;
rtx newaddr, newmem;
if (sched_verbose >= 5)
fprintf (sched_dump, "candidate mem/inc pair: %d %d\n",
INSN_UID (mii->mem_insn), INSN_UID (inc_cand));
/* Need to assure that none of the operands of the inc
instruction are assigned to by the mem insn. */
FOR_EACH_INSN_DEF (def, mii->mem_insn)
if (reg_overlap_mentioned_p (DF_REF_REG (def), mii->inc_input)
|| reg_overlap_mentioned_p (DF_REF_REG (def), mii->mem_reg0))
{
if (sched_verbose >= 5)
fprintf (sched_dump,
"inc conflicts with store failure.\n");
goto next;
}
newaddr = mii->inc_input;
if (mii->mem_index != NULL_RTX)
newaddr = gen_rtx_PLUS (GET_MODE (newaddr), newaddr,
mii->mem_index);
newaddr = plus_constant (GET_MODE (newaddr), newaddr,
mii->mem_constant + mii->inc_constant);
newmem = attempt_change (mii, newaddr);
if (newmem == NULL_RTX)
goto next;
if (sched_verbose >= 5)
fprintf (sched_dump, "successful address replacement\n");
desc = XCNEW (struct dep_replacement);
DEP_REPLACE (dep) = desc;
desc->loc = mii->mem_loc;
desc->newval = newmem;
desc->orig = *desc->loc;
desc->insn = mii->mem_insn;
move_dep_link (DEP_NODE_BACK (node), INSN_HARD_BACK_DEPS (con),
INSN_SPEC_BACK_DEPS (con));
if (backwards)
{
FOR_EACH_DEP (mii->inc_insn, SD_LIST_BACK, sd_it, dep)
add_dependence_1 (mii->mem_insn, DEP_PRO (dep),
REG_DEP_TRUE);
}
else
{
FOR_EACH_DEP (mii->inc_insn, SD_LIST_FORW, sd_it, dep)
add_dependence_1 (DEP_CON (dep), mii->mem_insn,
REG_DEP_ANTI);
}
return true;
}
next:
sd_iterator_next (&sd_it);
}
return false;
}
/* A recursive function that walks ADDRESS_OF_X to find memory references
which could be modified during scheduling. We call find_inc for each
one we find that has a recognizable form. MII holds information about
the pair of memory/increment instructions.
We ensure that every instruction with a memory reference (which will be
the location of the replacement) is assigned at most one breakable
dependency. */
static bool
find_mem (struct mem_inc_info *mii, rtx *address_of_x)
{
rtx x = *address_of_x;
enum rtx_code code = GET_CODE (x);
const char *const fmt = GET_RTX_FORMAT (code);
int i;
if (code == MEM)
{
rtx reg0 = XEXP (x, 0);
mii->mem_loc = address_of_x;
mii->mem_index = NULL_RTX;
mii->mem_constant = 0;
if (GET_CODE (reg0) == PLUS && CONST_INT_P (XEXP (reg0, 1)))
{
mii->mem_constant = INTVAL (XEXP (reg0, 1));
reg0 = XEXP (reg0, 0);
}
if (GET_CODE (reg0) == PLUS)
{
mii->mem_index = XEXP (reg0, 1);
reg0 = XEXP (reg0, 0);
}
if (REG_P (reg0))
{
df_ref use;
int occurrences = 0;
/* Make sure this reg appears only once in this insn. Can't use
count_occurrences since that only works for pseudos. */
FOR_EACH_INSN_USE (use, mii->mem_insn)
if (reg_overlap_mentioned_p (reg0, DF_REF_REG (use)))
if (++occurrences > 1)
{
if (sched_verbose >= 5)
fprintf (sched_dump, "mem count failure\n");
return false;
}
mii->mem_reg0 = reg0;
return find_inc (mii, true) || find_inc (mii, false);
}
return false;
}
if (code == SIGN_EXTRACT || code == ZERO_EXTRACT)
{
/* If REG occurs inside a MEM used in a bit-field reference,
that is unacceptable. */
return false;
}
/* Time for some deep diving. */
for (i = GET_RTX_LENGTH (code) - 1; i >= 0; i--)
{
if (fmt[i] == 'e')
{
if (find_mem (mii, &XEXP (x, i)))
return true;
}
else if (fmt[i] == 'E')
{
int j;
for (j = XVECLEN (x, i) - 1; j >= 0; j--)
if (find_mem (mii, &XVECEXP (x, i, j)))
return true;
}
}
return false;
}
/* Examine the instructions between HEAD and TAIL and try to find
dependencies that can be broken by modifying one of the patterns. */
void
find_modifiable_mems (rtx_insn *head, rtx_insn *tail)
{
rtx_insn *insn, *next_tail = NEXT_INSN (tail);
int success_in_block = 0;
for (insn = head; insn != next_tail; insn = NEXT_INSN (insn))
{
struct mem_inc_info mii;
if (!NONDEBUG_INSN_P (insn) || RTX_FRAME_RELATED_P (insn))
continue;
mii.mem_insn = insn;
if (find_mem (&mii, &PATTERN (insn)))
success_in_block++;
}
if (success_in_block && sched_verbose >= 5)
fprintf (sched_dump, "%d candidates for address modification found.\n",
success_in_block);
}
#endif /* INSN_SCHEDULING */
|