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
|
#ifndef EBWT_H_
#define EBWT_H_
#ifdef BOWTIE_MM
#include <sys/mman.h>
#include <sys/shm.h>
#endif
#include <algorithm>
#include <errno.h>
#include <fcntl.h>
#include <fstream>
#include <iostream>
#include <set>
#include <sstream>
#include <stdexcept>
#include <stdint.h>
#include <string.h>
#include <sys/stat.h>
#include "alphabet.h"
#include "assert_helpers.h"
#include "bitpack.h"
#include "bitset.h"
#include "blockwise_sa.h"
#include "ds.h"
#include "endian_swap.h"
#include "hit.h"
#include "mm.h"
#include "random_source.h"
#include "ref_read.h"
#include "reference.h"
#include "shmem.h"
#include "sstring.h"
#include "str_util.h"
#include "threading.h"
#include "timer.h"
#include "word_io.h"
#ifdef POPCNT_CAPABILITY
#include "processor_support.h"
#endif
using namespace std;
#ifndef PREFETCH_LOCALITY
// No locality by default
#define PREFETCH_LOCALITY 2
#endif
// From ccnt_lut.cpp, automatically generated by gen_lookup_tables.pl
extern uint8_t cCntLUT_4[4][4][256];
static const uint64_t c_table[4] = {
0xffffffffffffffffllu,
0xaaaaaaaaaaaaaaaallu,
0x5555555555555555llu,
0x0000000000000000llu
};
#ifndef VMSG_NL
#define VMSG_NL(args...) \
if(this->verbose()) { \
stringstream tmp; \
tmp << args << endl; \
this->verbose(tmp.str()); \
}
#endif
#ifndef VMSG
#define VMSG(args...) \
if(this->verbose()) { \
stringstream tmp; \
tmp << args; \
this->verbose(tmp.str()); \
}
#endif
/**
* Flags describing type of Ebwt.
*/
enum EBWT_FLAGS {
EBWT_COLOR = 2, // true -> Ebwt is colorspace
EBWT_ENTIRE_REV = 4 // true -> reverse Ebwt is the whole
// concatenated string reversed, rather than
// each stretch reversed
};
extern string gLastIOErrMsg;
inline bool is_read_err(int fdesc, ssize_t ret, size_t count){
if (ret < 0) {
std::stringstream sstm;
sstm << "ERRNO: " << errno << " ERR Msg:" << strerror(errno) << std::endl;
gLastIOErrMsg = sstm.str();
return true;
}
return false;
}
/* Checks whether a call to fread() failed or not. */
inline bool is_fread_err(FILE* file_hd, size_t ret, size_t count){
if (ferror(file_hd)) {
gLastIOErrMsg = "Error Reading File!";
return true;
}
return false;
}
/**
* Extended Burrows-Wheeler transform header. This together with the
* actual data arrays and other text-specific parameters defined in
* class Ebwt constitute the entire Ebwt.
*/
class EbwtParams {
public:
EbwtParams() { }
EbwtParams(TIndexOffU len,
int32_t lineRate,
int32_t linesPerSide,
int32_t offRate,
int32_t isaRate,
int32_t ftabChars,
bool entireReverse,
bool isBt2Index)
{
init(len, lineRate, linesPerSide, offRate, isaRate, ftabChars, entireReverse, isBt2Index);
}
EbwtParams(const EbwtParams& eh) {
init(eh._len, eh._lineRate, eh._linesPerSide, eh._offRate,
eh._isaRate, eh._ftabChars, eh._entireReverse, eh._isBt2Index);
}
void init(TIndexOffU len, int32_t lineRate, int32_t linesPerSide,
int32_t offRate, int32_t isaRate, int32_t ftabChars,
bool entireReverse, bool isBt2Index)
{
_isBt2Index = isBt2Index;
_entireReverse = entireReverse;
_len = len;
_bwtLen = _len + 1;
_sz = (len+3)/4;
_bwtSz = (len/4 + 1);
_lineRate = lineRate;
_linesPerSide = _isBt2Index ? 1 : linesPerSide;
_origOffRate = offRate;
_offRate = offRate;
_offMask = OFF_MASK << _offRate;
_isaRate = isaRate;
_isaMask = OFF_MASK << ((_isaRate >= 0) ? _isaRate : 0);
_ftabChars = ftabChars;
_eftabLen = _ftabChars*2;
_eftabSz = _eftabLen*OFF_SIZE;
_ftabLen = (1 << (_ftabChars*2))+1;
_ftabSz = _ftabLen*OFF_SIZE;
_offsLen = (_bwtLen + (1 << _offRate) - 1) >> _offRate;
_offsSz = (uint64_t)_offsLen*OFF_SIZE;
_isaLen = (_isaRate == -1)? 0 : ((_bwtLen + (1 << _isaRate) - 1) >> _isaRate);
_isaSz = _isaLen*OFF_SIZE;
_lineSz = 1 << _lineRate;
_sideSz = _lineSz * _linesPerSide;
_sideBwtSz = _sideSz - 2*OFF_SIZE;
_sideBwtLen = _sideBwtSz*4;
_numSidePairs = (_bwtSz+(2*_sideBwtSz)-1)/(2*_sideBwtSz);
_numSides = _numSidePairs*2;
_numLines = _numSides * _linesPerSide;
_ebwtTotLen = _numSidePairs * (2*_sideSz);
if (_isBt2Index) {
_sideBwtSz = _sideSz - 4*OFF_SIZE;
_sideBwtLen = _sideBwtSz*4;
_numSides = (_bwtSz+(_sideBwtSz)-1)/(_sideBwtSz);
_numSidePairs = _numSides / 2;
_numLines = _numSides * _linesPerSide;
_ebwtTotLen = _numSides * _sideSz;
}
_ebwtTotSz = _ebwtTotLen;
assert(repOk());
}
TIndexOffU len() const { return _len; }
TIndexOffU bwtLen() const { return _bwtLen; }
TIndexOffU sz() const { return _sz; }
TIndexOffU bwtSz() const { return _bwtSz; }
int32_t lineRate() const { return _lineRate; }
int32_t linesPerSide() const { return _linesPerSide; }
int32_t origOffRate() const { return _origOffRate; }
int32_t offRate() const { return _offRate; }
TIndexOffU offMask() const { return _offMask; }
int32_t isaRate() const { return _isaRate; }
uint32_t isaMask() const { return _isaMask; }
int32_t ftabChars() const { return _ftabChars; }
uint32_t eftabLen() const { return _eftabLen; }
uint32_t eftabSz() const { return _eftabSz; }
TIndexOffU ftabLen() const { return _ftabLen; }
TIndexOffU ftabSz() const { return _ftabSz; }
TIndexOffU offsLen() const { return _offsLen; }
uint64_t offsSz() const { return _offsSz; }
TIndexOffU isaLen() const { return _isaLen; }
uint64_t isaSz() const { return _isaSz; }
uint32_t lineSz() const { return _lineSz; }
uint32_t sideSz() const { return _sideSz; }
uint32_t sideBwtSz() const { return _sideBwtSz; }
uint32_t sideBwtLen() const { return _sideBwtLen; }
uint32_t numSidePairs() const { return _numSidePairs; } /* check */
TIndexOffU numSides() const { return _numSides; }
TIndexOffU numLines() const { return _numLines; }
TIndexOffU ebwtTotLen() const { return _ebwtTotLen; }
TIndexOffU ebwtTotSz() const { return _ebwtTotSz; }
bool entireReverse() const { return _entireReverse; }
bool isBt2Index() const { return _isBt2Index; }
/**
* Set a new suffix-array sampling rate, which involves updating
* rate, mask, sample length, and sample size.
*/
void setOffRate(int __offRate) {
_offRate = __offRate;
_offMask = OFF_MASK << _offRate;
_offsLen = (_bwtLen + (1 << _offRate) - 1) >> _offRate;
_offsSz = (uint64_t) _offsLen*OFF_SIZE;
}
/**
* Set a new inverse suffix-array sampling rate, which involves
* updating rate, mask, sample length, and sample size.
*/
void setIsaRate(int __isaRate) {
_isaRate = __isaRate;
_isaMask = OFF_MASK << _isaRate;
_isaLen = (_bwtLen + (1 << _isaRate) - 1) >> _isaRate;
_isaSz = (uint64_t)_isaLen*OFF_SIZE;
}
/// Check that this EbwtParams is internally consistent
bool repOk() const {
assert_gt(_len, 0);
assert_gt(_lineRate, 3);
assert_geq(_offRate, 0);
assert_leq(_ftabChars, 16);
assert_geq(_ftabChars, 1);
assert_lt(_lineRate, 32);
assert_lt(_linesPerSide, 32);
assert_lt(_ftabChars, 32);
assert_eq(0, _ebwtTotSz % (_isBt2Index ? _lineSz : 2*_lineSz));
return true;
}
/**
* Pretty-print the header contents to the given output stream.
*/
void print(ostream& out) const {
out << "Headers:" << endl
<< " len: " << _len << endl
<< " bwtLen: " << _bwtLen << endl
<< " sz: " << _sz << endl
<< " bwtSz: " << _bwtSz << endl
<< " lineRate: " << _lineRate << endl
<< " linesPerSide: " << _linesPerSide << endl
<< " offRate: " << _offRate << endl
<< " offMask: 0x" << hex << _offMask << dec << endl
<< " isaRate: " << _isaRate << endl
<< " isaMask: 0x" << hex << _isaMask << dec << endl
<< " ftabChars: " << _ftabChars << endl
<< " eftabLen: " << _eftabLen << endl
<< " eftabSz: " << _eftabSz << endl
<< " ftabLen: " << _ftabLen << endl
<< " ftabSz: " << _ftabSz << endl
<< " offsLen: " << _offsLen << endl
<< " offsSz: " << _offsSz << endl
<< " isaLen: " << _isaLen << endl
<< " isaSz: " << _isaSz << endl
<< " lineSz: " << _lineSz << endl
<< " sideSz: " << _sideSz << endl
<< " sideBwtSz: " << _sideBwtSz << endl
<< " sideBwtLen: " << _sideBwtLen << endl
<< " numSidePairs: " << _numSidePairs << endl
<< " numSides: " << _numSides << endl
<< " numLines: " << _numLines << endl
<< " ebwtTotLen: " << _ebwtTotLen << endl
<< " ebwtTotSz: " << _ebwtTotSz << endl
<< " reverse: " << _entireReverse << endl;
}
TIndexOffU _len;
TIndexOffU _bwtLen;
TIndexOffU _sz;
TIndexOffU _bwtSz;
int32_t _lineRate;
int32_t _linesPerSide;
int32_t _origOffRate;
int32_t _offRate;
TIndexOffU _offMask;
int32_t _isaRate;
uint32_t _isaMask;
int32_t _ftabChars;
uint32_t _eftabLen;
uint32_t _eftabSz;
TIndexOffU _ftabLen;
TIndexOffU _ftabSz;
TIndexOffU _offsLen;
uint64_t _offsSz;
TIndexOffU _isaLen;
uint64_t _isaSz;
uint32_t _lineSz;
uint32_t _sideSz;
uint32_t _sideBwtSz;
uint32_t _sideBwtLen;
uint32_t _numSidePairs;
TIndexOffU _numSides;
TIndexOffU _numLines;
TIndexOffU _ebwtTotLen;
TIndexOffU _ebwtTotSz;
bool _entireReverse;
bool _isBt2Index;
};
/**
* Exception to throw when a file-realted error occurs.
*/
class EbwtFileOpenException : public std::runtime_error {
public:
EbwtFileOpenException(const std::string& msg = "") :
std::runtime_error(msg) { }
};
/**
* Calculate size of file with given name.
*/
static inline int64_t fileSize(const char* name) {
std::ifstream f;
f.open(name, std::ios_base::binary | std::ios_base::in);
if (!f.good() || f.eof() || !f.is_open()) { return 0; }
f.seekg(0, std::ios_base::beg);
std::ifstream::pos_type begin_pos = f.tellg();
f.seekg(0, std::ios_base::end);
return static_cast<int64_t>(f.tellg() - begin_pos);
}
// Forward declarations for Ebwt class
struct SideLocus;
class EbwtSearchParams;
/**
* Extended Burrows-Wheeler transform data.
*
* An Ebwt may be transferred to and from RAM with calls to
* evictFromMemory() and loadIntoMemory(). By default, a newly-created
* Ebwt is not loaded into memory; if the user would like to use a
* newly-created Ebwt to answer queries, they must first call
* loadIntoMemory().
*/
class Ebwt {
public:
#define Ebwt_INITS \
_toBigEndian(currentlyBigEndian()), \
_overrideOffRate(__overrideOffRate), \
_overrideIsaRate(__overrideIsaRate), \
_verbose(verbose), \
_passMemExc(passMemExc), \
_sanity(sanityCheck), \
_isBt2Index(isBt2Index), \
_fw(__fw), \
_in1(NULL), \
_in2(NULL), \
_zOff(OFF_MASK), \
_zEbwtByteOff(OFF_MASK), \
_zEbwtBpOff(-1), \
_nPat(0), \
_nFrag(0), \
_plen(NULL), \
_rstarts(NULL), \
_fchr(NULL), \
_ftab(NULL), \
_eftab(NULL), \
_offs(NULL), \
_isa(NULL), \
_ebwt(NULL), \
_useMm(false), \
useShmem_(false), \
_refnames(), \
mmFile1_(NULL), \
mmFile2_(NULL)
#ifdef EBWT_STATS
#define Ebwt_STAT_INITS \
,mapLFExs_(0llu), \
mapLFs_(0llu), \
mapLFcs_(0llu), \
mapLF1cs_(0llu), \
mapLF1s_(0llu)
#else
#define Ebwt_STAT_INITS
#endif
/// Construct an Ebwt from the given input file
Ebwt(const string& in,
int needEntireReverse,
bool __fw,
int32_t __overrideOffRate = -1,
int32_t __overrideIsaRate = -1,
bool useMm = false,
bool useShmem = false,
bool mmSweep = false,
bool loadNames = false,
bool verbose = false,
bool startVerbose = false,
bool passMemExc = false,
bool sanityCheck = false,
bool isBt2Index = false) :
Ebwt_INITS
Ebwt_STAT_INITS
{
assert(!useMm || !useShmem);
#ifdef POPCNT_CAPABILITY
ProcessorSupport ps;
_usePOPCNTinstruction = ps.POPCNTenabled();
#endif
_packed = false;
_useMm = useMm;
useShmem_ = useShmem;
_in1Str = in + ".1." + gEbwt_ext;
_in2Str = in + ".2." + gEbwt_ext;
readIntoMemory(
__fw ? -1 : needEntireReverse, // need REF_READ_REVERSE
true, // stop after loading the header portion?
&_eh, // params structure to fill in
mmSweep, // mmSweep
loadNames, // loadNames
startVerbose); // startVerbose
// If the offRate has been overridden, reflect that in the
// _eh._offRate field
if(_overrideOffRate > _eh._offRate) {
_eh.setOffRate(_overrideOffRate);
assert_eq(_overrideOffRate, _eh._offRate);
}
// Same with isaRate
if(_overrideIsaRate > _eh._isaRate) {
_eh.setIsaRate(_overrideIsaRate);
assert_eq(_overrideIsaRate, _eh._isaRate);
}
assert(repOk());
}
/// Construct an Ebwt from the given header parameters and string
/// vector, optionally using a blockwise suffix sorter with the
/// given 'bmax' and 'dcv' parameters. The string vector is
/// ultimately joined and the joined string is passed to buildToDisk().
template<typename TStr>
Ebwt(TStr exampleStr,
bool packed,
int32_t lineRate,
int32_t linesPerSide,
int32_t offRate,
int32_t isaRate,
int32_t ftabChars,
int nthreads,
const string& file, // base filename for EBWT files
bool __fw,
bool useBlockwise,
TIndexOffU bmax,
TIndexOffU bmaxSqrtMult,
TIndexOffU bmaxDivN,
int dcv,
EList<FileBuf*>& is,
EList<RefRecord>& szs,
EList<uint32_t>& plens,
TIndexOffU sztot,
const RefReadInParams& refparams,
uint32_t seed,
int32_t __overrideOffRate = -1,
int32_t __overrideIsaRate = -1,
bool verbose = false,
bool passMemExc = false,
bool sanityCheck = false,
bool isBt2Index = false) :
Ebwt_INITS
Ebwt_STAT_INITS,
_eh(joinedLen(szs),
lineRate,
linesPerSide,
offRate,
isaRate,
ftabChars,
refparams.reverse == REF_READ_REVERSE,
isBt2Index)
{
_packed = packed;
#ifdef POPCNT_CAPABILITY
ProcessorSupport ps;
_usePOPCNTinstruction = ps.POPCNTenabled();
#endif
_in1Str = file + ".1." + gEbwt_ext;
_in2Str = file + ".2." + gEbwt_ext;
// Open output files
ofstream fout1(_in1Str.c_str(), ios::binary);
if(!fout1.good()) {
cerr << "Could not open index file for writing: \"" << _in1Str << "\"" << endl
<< "Please make sure the directory exists and that permissions allow writing by" << endl
<< "Bowtie." << endl;
throw 1;
}
ofstream fout2(_in2Str.c_str(), ios::binary);
if(!fout2.good()) {
cerr << "Could not open index file for writing: \"" << _in2Str << "\"" << endl
<< "Please make sure the directory exists and that permissions allow writing by" << endl
<< "Bowtie." << endl;
throw 1;
}
// Build
initFromVector<TStr>(
is,
szs,
plens,
sztot,
refparams,
fout1,
fout2,
file,
nthreads,
useBlockwise,
bmax,
bmaxSqrtMult,
bmaxDivN,
dcv,
seed);
// Close output files
fout1.flush();
int64_t tellpSz1 = (int64_t)fout1.tellp();
VMSG_NL("Wrote " << fout1.tellp() << " bytes to primary EBWT file: " << _in1Str);
fout1.close();
bool err = false;
if(tellpSz1 > fileSize(_in1Str.c_str())) {
err = true;
cerr << "Index is corrupt: File size for " << _in1Str << " should have been " << tellpSz1
<< " but is actually " << fileSize(_in1Str.c_str()) << "." << endl;
}
fout2.flush();
int64_t tellpSz2 = (int64_t)fout2.tellp();
VMSG_NL("Wrote " << fout2.tellp() << " bytes to secondary EBWT file: " << _in2Str);
fout2.close();
if(tellpSz2 > fileSize(_in2Str.c_str())) {
err = true;
cerr << "Index is corrupt: File size for " << _in2Str << " should have been " << tellpSz2
<< " but is actually " << fileSize(_in2Str.c_str()) << "." << endl;
}
if(err) {
cerr << "Please check if there is a problem with the disk or if disk is full." << endl;
throw 1;
}
// Reopen as input streams
VMSG_NL("Re-opening _in1 and _in2 as input streams");
if(_sanity) {
VMSG_NL("Sanity-checking Ebwt");
assert(!isInMemory());
readIntoMemory(
__fw ? -1 : refparams.reverse == REF_READ_REVERSE,
false,
NULL,
false,
true,
false);
sanityCheckAll(refparams.reverse);
evictFromMemory();
assert(!isInMemory());
}
VMSG_NL("Returning from Ebwt constructor");
}
bool isPacked() {
return _packed;
};
/**
* Write the rstarts array given the szs array for the reference.
*/
void szsToDisk(const EList<RefRecord>& szs, ostream& os, int reverse) {
TIndexOffU seq = 0;
TIndexOffU off = 0;
TIndexOffU totlen = 0;
for(unsigned int i = 0; i < szs.size(); i++) {
if(szs[i].len == 0) continue;
if(szs[i].first) off = 0;
off += szs[i].off;
#ifdef ACCOUNT_FOR_ALL_GAP_REFS
if(szs[i].first && szs[i].len > 0) seq++;
#else
if(szs[i].first) seq++;
#endif
TIndexOffU seqm1 = seq-1;
assert_lt(seqm1, _nPat);
TIndexOffU fwoff = off;
if(reverse == REF_READ_REVERSE) {
// Invert pattern idxs
seqm1 = _nPat - seqm1 - 1;
// Invert pattern idxs
assert_leq(off + szs[i].len, _plen[seqm1]);
fwoff = _plen[seqm1] - (off + szs[i].len);
}
writeU<TIndexOffU>(os, totlen, this->toBe()); // offset from beginning of joined string
writeU<TIndexOffU>(os, seqm1, this->toBe()); // sequence id
writeU<TIndexOffU>(os, fwoff, this->toBe()); // offset into sequence
totlen += szs[i].len;
off += szs[i].len;
}
}
/**
* Helper for the constructors above. Takes a vector of text
* strings and joins them into a single string with a call to
* joinToDisk, which does a join (with padding) and writes some of
* the resulting data directly to disk rather than keep it in
* memory. It then constructs a suffix-array producer (what kind
* depends on 'useBlockwise') for the resulting sequence. The
* suffix-array producer can then be used to obtain chunks of the
* joined string's suffix array.
*/
template<typename TStr>
void initFromVector(
EList<FileBuf*>& is,
EList<RefRecord>& szs,
EList<uint32_t>& plens,
TIndexOffU sztot,
const RefReadInParams& refparams,
ofstream& out1,
ofstream& out2,
const string& outfile,
int nthreads,
bool useBlockwise,
TIndexOffU bmax,
TIndexOffU bmaxSqrtMult,
TIndexOffU bmaxDivN,
int dcv,
uint32_t seed)
{
// Compose text strings into single string
VMSG_NL("Calculating joined length");
TStr s; // holds the entire joined reference after call to joinToDisk
TIndexOffU jlen;
jlen = joinedLen(szs);
assert_geq(jlen, sztot);
VMSG_NL("Writing header");
writeFromMemory(true, out1, out2);
try {
VMSG_NL("Reserving space for joined string");
s.resize(jlen);
VMSG_NL("Joining reference sequences");
if(refparams.reverse == REF_READ_REVERSE) {
{
Timer timer(cout, " Time to join reference sequences: ", _verbose);
joinToDisk(is, szs, plens, sztot, refparams, s, out1, out2, seed);
} {
Timer timer(cout, " Time to reverse reference sequence: ", _verbose);
EList<RefRecord> tmp;
s.reverse();
reverseRefRecords(szs, tmp, false, false);
szsToDisk(tmp, out1, refparams.reverse);
}
} else {
Timer timer(cout, " Time to join reference sequences: ", _verbose);
joinToDisk(is, szs, plens, sztot, refparams, s, out1, out2, seed);
szsToDisk(szs, out1, refparams.reverse);
}
// Joined reference sequence now in 's'
} catch(bad_alloc& e) {
// If we throw an allocation exception in the try block,
// that means that the joined version of the reference
// string itself is too larger to fit in memory. The only
// alternatives are to tell the user to give us more memory
// or to try again with a packed representation of the
// reference (if we haven't tried that already).
cerr << "Could not allocate space for a joined string of " << jlen << " elements." << endl;
if(!isPacked() && _passMemExc) {
// Pass the exception up so that we can retry using a
// packed string representation
throw e;
}
// There's no point passing this exception on. The fact
// that we couldn't allocate the joined string means that
// --bmax is irrelevant - the user should re-run with
// ebwt-build-packed
if(isPacked()) {
cerr << "Please try running bowtie-build on a computer with more memory." << endl;
} else {
cerr << "Please try running bowtie-build in packed mode (-p/--packed) or in automatic" << endl
<< "mode (-a/--auto), or try again on a computer with more memory." << endl;
}
if(sizeof(void*) == 4) {
cerr << "If this computer has more than 4 GB of memory, try using a 64-bit executable;" << endl
<< "this executable is 32-bit." << endl;
}
throw 1;
}
// Succesfully obtained joined reference string
assert_geq(s.length(), jlen);
if(bmax != OFF_MASK) {
VMSG_NL("bmax according to bmax setting: " << bmax);
}
else if(bmaxSqrtMult != OFF_MASK) {
bmax *= bmaxSqrtMult;
VMSG_NL("bmax according to bmaxSqrtMult setting: " << bmax);
}
else if(bmaxDivN != OFF_MASK) {
bmax = max<TIndexOffU>(jlen / bmaxDivN, 1);
VMSG_NL("bmax according to bmaxDivN setting: " << bmax);
}
else {
bmax = (TIndexOffU)sqrt(s.length());
VMSG_NL("bmax defaulted to: " << bmax);
}
int iter = 0;
bool first = true;
// Look for bmax/dcv parameters that work.
while(true) {
if(!first && bmax < 40 && _passMemExc) {
cerr << "Could not find approrpiate bmax/dcv settings for building this index." << endl;
if(!isPacked()) {
// Throw an exception exception so that we can
// retry using a packed string representation
throw bad_alloc();
} else {
cerr << "Already tried a packed string representation." << endl;
}
cerr << "Please try indexing this reference on a computer with more memory." << endl;
if(sizeof(void*) == 4) {
cerr << "If this computer has more than 4 GB of memory, try using a 64-bit executable;" << endl
<< "this executable is 32-bit." << endl;
}
throw 1;
}
if(dcv > 4096) dcv = 4096;
if((iter % 6) == 5 && dcv < 4096 && dcv != 0) {
dcv <<= 1; // double difference-cover period
} else {
bmax -= (bmax >> 2); // reduce by 25%
}
VMSG("Using parameters --bmax " << bmax);
if(dcv == 0) {
VMSG_NL(" and *no difference cover*");
} else {
VMSG_NL(" --dcv " << dcv);
}
iter++;
try {
{
VMSG_NL(" Doing ahead-of-time memory usage test");
// Make a quick-and-dirty attempt to force a bad_alloc iff
// we would have thrown one eventually as part of
// constructing the DifferenceCoverSample
dcv <<= 1;
TIndexOffU sz = (TIndexOffU)DifferenceCoverSample<TStr>::simulateAllocs(s, dcv >> 1);
if(nthreads > 1) sz *= (nthreads + 1);
AutoArray<uint8_t> tmp(sz);
dcv >>= 1;
// Likewise with the KarkkainenBlockwiseSA
sz = (TIndexOffU)KarkkainenBlockwiseSA<TStr>::simulateAllocs(s, bmax);
AutoArray<uint8_t> tmp2(sz);
// Now throw in the 'ftab' and 'isaSample' structures
// that we'll eventually allocate in buildToDisk
AutoArray<TIndexOffU> ftab(_eh._ftabLen * 2);
AutoArray<uint8_t> side(_eh._sideSz);
// Grab another 20 MB out of caution
AutoArray<uint32_t> extra(20*1024*1024);
// If we made it here without throwing bad_alloc, then we
// passed the memory-usage stress test
VMSG(" Passed! Constructing with these parameters: --bmax " << bmax << " --dcv " << dcv);
if(isPacked()) {
VMSG(" --packed");
}
VMSG_NL("");
}
VMSG_NL("Constructing suffix-array element generator");
KarkkainenBlockwiseSA<TStr> bsa(s, bmax, nthreads, dcv, seed, _sanity, _passMemExc, _verbose, outfile);
assert(bsa.suffixItrIsReset());
assert_eq(bsa.size(), s.length()+1);
VMSG_NL("Converting suffix-array elements to index image");
buildToDisk(bsa, s, out1, out2);
out1.flush(); out2.flush();
if(out1.fail() || out2.fail()) {
cerr << "An error occurred writing the index to disk. Please check if the disk is full." << endl;
throw 1;
}
break;
} catch(bad_alloc& e) {
if(_passMemExc) {
VMSG_NL(" Ran out of memory; automatically trying more memory-economical parameters.");
} else {
cerr << "Out of memory while constructing suffix array. Please try using a smaller" << endl
<< "number of blocks by specifying a smaller --bmax or a larger --bmaxdivn" << endl;
throw 1;
}
}
first = false;
}
assert(repOk());
// Now write reference sequence names on the end
#ifdef ACCOUNT_FOR_ALL_GAP_REFS
assert_geq(this->_refnames.size(), this->_nPat);
#else
assert_eq(this->_refnames.size(), this->_nPat);
#endif
for(TIndexOffU i = 0; i < this->_refnames.size(); i++) {
out1 << this->_refnames[i] << endl;
}
out1 << '\0';
out1.flush(); out2.flush();
if(out1.fail() || out2.fail()) {
cerr << "An error occurred writing the index to disk. Please check if the disk is full." << endl;
throw 1;
}
VMSG_NL("Returning from initFromVector");
}
/**
* Return the length that the joined string of the given string
* list will have. Note that this is indifferent to how the text
* fragments correspond to input sequences - it just cares about
* the lengths of the fragments.
*/
TIndexOffU joinedLen(EList<RefRecord>& szs) {
TIndexOffU ret = 0;
for(unsigned int i = 0; i < szs.size(); i++) {
ret += szs[i].len;
}
return ret;
}
/// Destruct an Ebwt
~Ebwt() {
// Only free buffers if we're *not* using memory-mapped files
if(!_useMm) {
// Delete everything that was allocated in read(false, ...)
if(_fchr != NULL) delete[] _fchr;
if(_ftab != NULL) delete[] _ftab;
if(_eftab != NULL) delete[] _eftab;
if(_offs != NULL && !useShmem_)
delete[] _offs;
else if(_offs != NULL && useShmem_)
FREE_SHARED(_offs);
if(_isa != NULL) delete[] _isa;
if(_plen != NULL) delete[] _plen;
if(_rstarts != NULL) delete[] _rstarts;
if(_ebwt != NULL && !useShmem_)
delete[] _ebwt;
else if(_ebwt != NULL && useShmem_)
FREE_SHARED(_ebwt);
}
if (_in1 != NULL) fclose(_in1);
if (_in2 != NULL) fclose(_in2);
#ifdef EBWT_STATS
cout << (_fw ? "Forward index:" : "Mirror index:") << endl;
cout << " mapLFEx: " << mapLFExs_ << endl;
cout << " mapLF: " << mapLFs_ << endl;
cout << " mapLF(c): " << mapLFcs_ << endl;
cout << " mapLF1(c): " << mapLF1cs_ << endl;
cout << " mapLF(c): " << mapLF1s_ << endl;
#endif
}
/// Accessors
const EbwtParams& eh() const { return _eh; }
TIndexOffU zOff() const { return _zOff; }
TIndexOffU zEbwtByteOff() const { return _zEbwtByteOff; }
TIndexOff zEbwtBpOff() const { return _zEbwtBpOff; }
TIndexOffU nPat() const { return _nPat; }
TIndexOffU nFrag() const { return _nFrag; }
TIndexOffU* fchr() const { return _fchr; }
TIndexOffU* ftab() const { return _ftab; }
TIndexOffU* eftab() const { return _eftab; }
TIndexOffU* offs() const { return _offs; }
TIndexOffU* isa() const { return _isa; } /* check */
TIndexOffU* plen() const { return _plen; }
TIndexOffU* rstarts() const { return _rstarts; }
uint8_t* ebwt() const { return _ebwt; }
bool toBe() const { return _toBigEndian; }
bool verbose() const { return _verbose; }
bool sanityCheck() const { return _sanity; }
EList<string>& refnames() { return _refnames; }
bool fw() const { return _fw; }
#ifdef POPCNT_CAPABILITY
bool _usePOPCNTinstruction;
#endif
/// Return true iff the Ebwt is currently in memory
bool isInMemory() const {
if(_ebwt != NULL) {
assert(_eh.repOk());
assert(_ftab != NULL);
assert(_eftab != NULL);
assert(_fchr != NULL);
assert(_offs != NULL);
assert(_isa != NULL);
assert(_rstarts != NULL);
assert_neq(_zEbwtByteOff, OFF_MASK);
assert_neq(_zEbwtBpOff, -1);
return true;
} else {
assert(_ftab == NULL);
assert(_eftab == NULL);
assert(_fchr == NULL);
assert(_offs == NULL);
assert(_rstarts == NULL);
assert_eq(_zEbwtByteOff, OFF_MASK);
assert_eq(_zEbwtBpOff, -1);
return false;
}
}
/// Return true iff the Ebwt is currently stored on disk
bool isEvicted() const {
return !isInMemory();
}
/**
* Load this Ebwt into memory by reading it in from the _in1 and
* _in2 streams.
*/
void loadIntoMemory(
int needEntireReverse,
bool loadNames,
bool verbose)
{
readIntoMemory(
needEntireReverse, // require reverse index to be concatenated reference reversed
false, // stop after loading the header portion?
NULL, // params
false, // mmSweep
loadNames, // loadNames
verbose); // startVerbose
}
/**
* Frees memory associated with the Ebwt.
*/
void evictFromMemory() {
assert(isInMemory());
if(!_useMm) {
delete[] _fchr;
delete[] _ftab;
delete[] _eftab;
if(!useShmem_) delete[] _offs;
delete[] _isa;
// Keep plen; it's small and the client may want to query it
// even when the others are evicted.
//delete[] _plen;
delete[] _rstarts;
if(!useShmem_) delete[] _ebwt;
}
_fchr = NULL;
_ftab = NULL;
_eftab = NULL;
_offs = NULL;
_isa = NULL;
// Keep plen; it's small and the client may want to query it
// even when the others are evicted.
//_plen = NULL;
_rstarts = NULL;
_ebwt = NULL;
_zEbwtByteOff = OFF_MASK;
_zEbwtBpOff = -1;
}
/**
* Non-static facade for static function ftabHi.
*/
TIndexOffU ftabHi(TIndexOffU i) const {
return Ebwt::ftabHi(_ftab, _eftab, _eh._len, _eh._ftabLen,
_eh._eftabLen, i);
}
/**
* Get "high interpretation" of ftab entry at index i. The high
* interpretation of a regular ftab entry is just the entry
* itself. The high interpretation of an extended entry is the
* second correpsonding ui32 in the eftab.
*
* It's a static member because it's convenient to ask this
* question before the Ebwt is fully initialized.
*/
static TIndexOffU ftabHi(TIndexOffU *ftab,
TIndexOffU *eftab,
TIndexOffU len,
TIndexOffU ftabLen,
TIndexOffU eftabLen,
TIndexOffU i)
{
assert_lt(i, ftabLen);
if(ftab[i] <= len) {
return ftab[i];
} else {
TIndexOffU efIdx = ftab[i] ^ OFF_MASK;
assert_lt(efIdx*2+1, eftabLen);
return eftab[efIdx*2+1];
}
}
/**
* Non-static facade for static function ftabLo.
*/
TIndexOffU ftabLo(TIndexOffU i) const {
return Ebwt::ftabLo(_ftab, _eftab, _eh._len, _eh._ftabLen,
_eh._eftabLen, i);
}
/**
* Get "low interpretation" of ftab entry at index i. The low
* interpretation of a regular ftab entry is just the entry
* itself. The low interpretation of an extended entry is the
* first correpsonding ui32 in the eftab.
*
* It's a static member because it's convenient to ask this
* question before the Ebwt is fully initialized.
*/
static TIndexOffU ftabLo(TIndexOffU *ftab,
TIndexOffU *eftab,
TIndexOffU len,
TIndexOffU ftabLen,
TIndexOffU eftabLen,
TIndexOffU i)
{
assert_lt(i, ftabLen);
if(ftab[i] <= len) {
return ftab[i];
} else {
TIndexOffU efIdx = ftab[i] ^ OFF_MASK;
assert_lt(efIdx*2+1, eftabLen);
return eftab[efIdx*2];
}
}
/**
* When using read() to create an Ebwt, we have to set a couple of
* additional fields in the Ebwt object that aren't part of the
* parameter list and are not stored explicitly in the file. Right
* now, this just involves initializing _zEbwtByteOff and
* _zEbwtBpOff from _zOff.
*/
void postReadInit(EbwtParams& eh) {
TIndexOffU sideNum = _zOff / eh._sideBwtLen;
TIndexOffU sideCharOff = _zOff % eh._sideBwtLen;
TIndexOffU sideByteOff = sideNum * eh._sideSz;
_zEbwtByteOff = sideCharOff >> 2;
assert_lt(_zEbwtByteOff, eh._sideBwtSz);
_zEbwtBpOff = sideCharOff & 3;
assert_lt(_zEbwtBpOff, 4);
if(!eh._isBt2Index && (sideNum & 1) == 0) {
// This is an even (backward) side
_zEbwtByteOff = eh._sideBwtSz - _zEbwtByteOff - 1;
_zEbwtBpOff = 3 - _zEbwtBpOff;
assert_lt(_zEbwtBpOff, 4);
}
_zEbwtByteOff += sideByteOff;
assert(repOk(eh)); // Ebwt should be fully initialized now
}
/**
* Pretty-print the Ebwt to the given output stream.
*/
void print(ostream& out) const {
print(out, _eh);
}
/**
* Pretty-print the Ebwt and given EbwtParams to the given output
* stream.
*/
void print(ostream& out, const EbwtParams& eh) const {
eh.print(out); // print params
out << "Ebwt (" << (isInMemory()? "memory" : "disk") << "):" << endl
<< " zOff: " << _zOff << endl
<< " zEbwtByteOff: " << _zEbwtByteOff << endl
<< " zEbwtBpOff: " << _zEbwtBpOff << endl
<< " nPat: " << _nPat << endl
<< " plen: ";
if(_plen == NULL) {
out << "NULL" << endl;
} else {
out << "non-NULL, [0] = " << _plen[0] << endl;
}
out << " rstarts: ";
if(_rstarts == NULL) {
out << "NULL" << endl;
} else {
out << "non-NULL, [0] = " << _rstarts[0] << endl;
}
out << " ebwt: ";
if(_ebwt == NULL) {
out << "NULL" << endl;
} else {
out << "non-NULL, [0] = " << _ebwt[0] << endl;
}
out << " fchr: ";
if(_fchr == NULL) {
out << "NULL" << endl;
} else {
out << "non-NULL, [0] = " << _fchr[0] << endl;
}
out << " ftab: ";
if(_ftab == NULL) {
out << "NULL" << endl;
} else {
out << "non-NULL, [0] = " << _ftab[0] << endl;
}
out << " eftab: ";
if(_eftab == NULL) {
out << "NULL" << endl;
} else {
out << "non-NULL, [0] = " << _eftab[0] << endl;
}
out << " offs: ";
if(_offs == NULL) {
out << "NULL" << endl;
} else {
out << "non-NULL, [0] = " << _offs[0] << endl;
}
}
// Building
template <typename TStr> static TStr join(EList<TStr>& l, uint32_t seed);
template <typename TStr> static TStr join(EList<FileBuf*>& l, EList<RefRecord>& szs, TIndexOffU sztot, const RefReadInParams& refparams, uint32_t seed);
template <typename TStr> void joinToDisk(EList<FileBuf*>& l, EList<RefRecord>& szs, EList<uint32_t>& plens, TIndexOffU sztot, const RefReadInParams& refparams, TStr& ret, ostream& out1, ostream& out2, uint32_t seed = 0);
template <typename TStr> void buildToDisk(InorderBlockwiseSA<TStr>& sa, const TStr& s, ostream& out1, ostream& out2);
// I/O
void readIntoMemory(int needEntireReverse, bool justHeader, EbwtParams *params, bool mmSweep, bool loadNames, bool startVerbose);
void writeFromMemory(bool justHeader, ostream& out1, ostream& out2) const;
void writeFromMemory(bool justHeader, const string& out1, const string& out2) const;
// Sanity checking
void printRangeFw(uint32_t begin, uint32_t end) const;
void printRangeBw(uint32_t begin, uint32_t end) const;
void sanityCheckUpToSide(TIndexOff upToSide) const;
void sanityCheckAll(int reverse) const;
void restore(BTRefString& s) const;
void checkOrigs(const EList<BTRefString >& os, bool mirror) const;
// Searching and reporting
void joinedToTextOff(TIndexOffU qlen, TIndexOffU off, TIndexOffU& tidx, TIndexOffU& textoff, TIndexOffU& tlen) const;
inline bool report(const BTDnaString& query, BTString* quals, BTString* name, const EList<TIndexOffU>& mmui32, const EList<uint8_t>& refcs, size_t numMms, TIndexOffU off, TIndexOffU top, TIndexOffU bot, uint32_t qlen, int stratum, uint16_t cost, uint32_t patid, uint32_t seed, const EbwtSearchParams& params) const;
inline bool reportChaseOne(const BTDnaString& query, BTString* quals, BTString* name, const EList<TIndexOffU>& mmui32, const EList<uint8_t>& refcs, size_t numMms, TIndexOffU i, TIndexOffU top, TIndexOffU bot, uint32_t qlen, int stratum, uint16_t cost, uint32_t patid, uint32_t seed, const EbwtSearchParams& params, SideLocus *l = NULL) const;
inline int rowL(const SideLocus& l) const;
inline TIndexOffU countUpTo(const SideLocus& l, int c) const;
inline void countUpToEx(const SideLocus& l, TIndexOffU* pairs) const;
inline TIndexOffU countFwSide(const SideLocus& l, int c) const;
inline void countFwSideEx(const SideLocus& l, TIndexOffU *pairs) const;
inline TIndexOffU countBwSide(const SideLocus& l, int c) const;
inline void countBwSideEx(const SideLocus& l, TIndexOffU *pairs) const;
inline TIndexOffU countBt2Side(const SideLocus& l, int c) const;
inline void countBt2SideEx(const SideLocus& l, TIndexOffU *pairs) const;
inline TIndexOffU mapLF(const SideLocus& l ASSERT_ONLY(, bool overrideSanity = false)) const;
inline void mapLFEx(const SideLocus& l, TIndexOffU *pairs ASSERT_ONLY(, bool overrideSanity = false)) const;
inline void mapLFEx(const SideLocus& ltop, const SideLocus& lbot, TIndexOffU *tops, TIndexOffU *bots ASSERT_ONLY(, bool overrideSanity = false)) const;
inline TIndexOffU mapLF(const SideLocus& l, int c ASSERT_ONLY(, bool overrideSanity = false)) const;
inline TIndexOffU mapLF1(TIndexOffU row, const SideLocus& l, int c ASSERT_ONLY(, bool overrideSanity = false)) const;
inline int mapLF1(TIndexOffU& row, const SideLocus& l ASSERT_ONLY(, bool overrideSanity = false)) const;
/// Check that in-memory Ebwt is internally consistent with respect
/// to given EbwtParams; assert if not
bool inMemoryRepOk(const EbwtParams& eh) const {
// assert_leq(ValueSize<TAlphabet>::VALUE, 4);
assert_geq(_zEbwtBpOff, 0);
assert_lt(_zEbwtBpOff, 4);
assert_lt(_zEbwtByteOff, eh._ebwtTotSz);
assert_lt(_zOff, eh._bwtLen);
assert(_rstarts != NULL);
assert_geq(_nFrag, _nPat);
return true;
}
/// Check that in-memory Ebwt is internally consistent; assert if
/// not
bool inMemoryRepOk() const {
return repOk(_eh);
}
/// Check that Ebwt is internally consistent with respect to given
/// EbwtParams; assert if not
bool repOk(const EbwtParams& eh) const {
assert(_eh.repOk());
if(isInMemory()) {
return inMemoryRepOk(eh);
}
return true;
}
/// Check that Ebwt is internally consistent; assert if not
bool repOk() const {
return repOk(_eh);
}
bool _toBigEndian;
int32_t _overrideOffRate;
int32_t _overrideIsaRate;
bool _verbose;
bool _passMemExc;
bool _sanity;
bool _isBt2Index;
bool _fw; // true iff this is a forward index
FILE *_in1; // input fd for primary index file
FILE *_in2; // input fd for secondary index file
string _in1Str; // filename for primary index file
string _in2Str; // filename for secondary index file
TIndexOffU _zOff;
TIndexOffU _zEbwtByteOff;
TIndexOff _zEbwtBpOff;
TIndexOffU _nPat; /// number of reference texts
TIndexOffU _nFrag; /// number of fragments
TIndexOffU* _plen;
TIndexOffU* _rstarts; // starting offset of fragments / text indexes
// _fchr, _ftab and _eftab are expected to be relatively small
// (usually < 1MB, perhaps a few MB if _fchr is particularly large
// - like, say, 11). For this reason, we don't bother with writing
// them to disk through separate output streams; we
TIndexOffU* _fchr;
TIndexOffU* _ftab;
TIndexOffU* _eftab; // "extended" entries for _ftab
// _offs may be extremely large. E.g. for DNA w/ offRate=4 (one
// offset every 16 rows), the total size of _offs is the same as
// the total size of the input sequence
TIndexOffU* _offs;
TIndexOffU* _isa;
// _ebwt is the Extended Burrows-Wheeler Transform itself, and thus
// is at least as large as the input sequence.
uint8_t* _ebwt;
bool _useMm; /// use memory-mapped files to hold the index
bool useShmem_; /// use shared memory to hold large parts of the index
EList<string> _refnames; /// names of the reference sequences
char *mmFile1_;
char *mmFile2_;
EbwtParams _eh;
bool _packed;
#ifdef BOWTIE_64BIT_INDEX
static const int default_lineRate = 7;
#else
static const int default_lineRate = 6;
#endif
#ifdef EBWT_STATS
uint64_t mapLFExs_;
uint64_t mapLFs_;
uint64_t mapLFcs_;
#endif
private:
ostream& log() const {
return cout; // TODO: turn this into a parameter
}
/// Print a verbose message and flush (flushing is helpful for
/// debugging)
void verbose(const string& s) const {
if(this->verbose()) {
this->log() << s;
this->log().flush();
}
}
};
/**
* Structure encapsulating search parameters, such as whether and how
* to backtrack and how to deal with multiple equally-good hits.
*/
class EbwtSearchParams {
public:
EbwtSearchParams(HitSinkPerThread& sink,
const EList<BTRefString >& texts,
bool fw = true,
bool ebwtFw = true) :
_sink(sink),
_texts(texts),
_patid(0xffffffff),
_fw(fw) { }
HitSinkPerThread& sink() const { return _sink; }
void setPatId(uint32_t patid) { _patid = patid; }
uint32_t patId() const { return _patid; }
void setFw(bool fw) { _fw = fw; }
bool fw() const { return _fw; }
/**
* Report a hit. Returns true iff caller can call off the search.
*/
bool reportHit(const BTDnaString& query, // read sequence
BTString* quals, // read quality values
BTString* name, // read name
bool ebwtFw, // whether index is forward (true) or mirror (false)
const EList<TIndexOffU>& mmui32, // mismatch list
const EList<uint8_t>& refcs, // reference characters
size_t numMms, // # mismatches
UPair h, // ref coords
UPair mh, // mate's ref coords
bool mfw, // mate's orientation
uint16_t mlen, // mate length
UPair a, // arrow pair
uint32_t tlen, // length of text
uint32_t qlen, // length of query
int stratum, // alignment stratum
uint16_t cost, // cost of alignment
uint32_t oms, // approx. # other valid alignments
uint32_t patid,
uint32_t seed,
uint8_t mate) const
{
#ifndef NDEBUG
// Check that no two elements of the mms array are the same
for(size_t i = 0; i < numMms; i++) {
for(size_t j = i+1; j < numMms; j++) {
assert_neq(mmui32[i], mmui32[j]);
}
}
#endif
// If ebwtFw is true, then 'query' and 'quals' are reversed
// If _fw is false, then 'query' and 'quals' are reverse complemented
assert(quals != NULL);
assert(name != NULL);
assert_eq(mmui32.size(), refcs.size());
assert_leq(numMms, mmui32.size());
assert_gt(qlen, 0);
Hit hit;
hit.stratum = stratum;
hit.cost = cost;
hit.patSeq = query;
hit.quals = *quals;
if(!ebwtFw) {
// Re-reverse the pattern and the quals back to how they
// appeared in the read file
hit.patSeq.reverse();
hit.quals.reverse();
}
// Turn the mmui32 and refcs arrays into the mm FixedBitset and
// the refc vector
hit.refcs.resize(qlen);
hit.refcs.fillZero();
for(size_t i = 0; i < numMms; i++) {
if (ebwtFw != _fw) {
// The 3' end is on the left but the mm vector encodes
// mismatches w/r/t the 5' end, so we flip
uint32_t off = qlen - mmui32[i] - 1;
hit.mms.set(off);
hit.refcs[off] = refcs[i];
} else {
hit.mms.set(mmui32[i]);
hit.refcs[mmui32[i]] = refcs[i];
}
}
// Check the hit against the original text, if it's available
if(_texts.size() > 0) {
assert_lt(h.first, _texts.size());
FixedBitset<1024> diffs;
// This type of check assumes that only mismatches are
// possible. If indels are possible, then we either need
// the caller to provide information about indel locations,
// or we need to extend this to a more complicated check.
assert_leq(h.second + qlen, _texts[h.first].length());
for(size_t i = 0; i < qlen; i++) {
assert_neq(4, (int)_texts[h.first][h.second + i]);
// Forward pattern appears at h
if((int)hit.patSeq[i] != (int)_texts[h.first][h.second + i]) {
uint32_t qoff = (uint32_t)i;
// if ebwtFw != _fw the 3' end is on on the
// left end of the pattern, but the diff vector
// should encode mismatches w/r/t the 5' end,
// so we flip
if (_fw) diffs.set(qoff);
else diffs.set(qlen - qoff - 1);
}
}
if(diffs != hit.mms) {
// Oops, mismatches were not where we expected them;
// print a diagnostic message before asserting
cerr << "Expected " << hit.mms.str() << " mismatches, got " << diffs.str() << endl;
cerr << " Pat: " << hit.patSeq << endl;
cerr << " Tseg: ";
for(size_t i = 0; i < qlen; i++) {
cerr << _texts[h.first][h.second + i];
}
cerr << endl;
cerr << " mmui32: ";
for(size_t i = 0; i < numMms; i++) {
cerr << mmui32[i] << " ";
}
cerr << endl;
cerr << " FW: " << _fw << endl;
cerr << " Ebwt FW: " << ebwtFw << endl;
}
if(diffs != hit.mms) assert(false);
}
hit.h = h;
hit.patId = ((patid == 0xffffffff) ? _patid : patid);
hit.patName = *name;
hit.mh = mh;
hit.fw = _fw;
hit.mfw = mfw;
hit.mlen = mlen;
hit.oms = oms;
hit.mate = mate;
hit.seed = seed;
assert(hit.repOk());
return sink().reportHit(hit, stratum);
}
private:
HitSinkPerThread& _sink;
const EList<BTRefString >& _texts; // original texts, if available (if not
// available, _texts.size() == 0)
uint32_t _patid; // id of current read
bool _fw; // current read is forward-oriented
};
/**
* Encapsulates a location in the bwt text in terms of the side it
* occurs in and its offset within the side.
*/
struct SideLocus {
SideLocus() :
_sideByteOff(0),
_sideNum(0),
_charOff(0),
_fw(true),
_by(-1),
_bp(-1) { }
/**
* Construct from row and other relevant information about the Ebwt.
*/
SideLocus(TIndexOffU row, const EbwtParams& ep, const uint8_t* ebwt) {
initFromRow(row, ep, ebwt);
}
/**
* Init two SideLocus objects from a top/bot pair, using the result
* from one call to initFromRow to possibly avoid a second call.
*/
static void initFromTopBot(TIndexOffU top,
TIndexOffU bot,
const EbwtParams& ep,
const uint8_t* ebwt,
SideLocus& ltop,
SideLocus& lbot)
{
const TIndexOffU sideBwtLen = ep._sideBwtLen;
const uint32_t sideBwtSz = ep._sideBwtSz;
assert_gt(bot, top);
ltop.initFromRow(top, ep, ebwt);
TIndexOffU spread = bot - top;
if(ltop._charOff + spread < sideBwtLen) {
lbot._charOff = (uint32_t)(ltop._charOff + spread);
lbot._sideNum = ltop._sideNum;
lbot._sideByteOff = ltop._sideByteOff;
lbot._fw = ep._isBt2Index ? true : ltop._fw;
lbot._by = lbot._charOff >> 2;
assert_lt(lbot._by, (int)sideBwtSz);
if(!lbot._fw) lbot._by = sideBwtSz - lbot._by - 1;
lbot._bp = lbot._charOff & 3;
if(!lbot._fw) lbot._bp ^= 3;
} else {
lbot.initFromRow(bot, ep, ebwt);
}
}
/**
* Calculate SideLocus based on a row and other relevant
* information about the shape of the Ebwt.
*/
void initFromRow(TIndexOffU row, const EbwtParams& ep, const uint8_t* ebwt) {
const uint32_t sideSz = ep._sideSz;
// Side length is hard-coded for now; this allows the compiler
// to do clever things to accelerate / and %.
if (ep._isBt2Index) {
_sideNum = row / (48*OFF_SIZE);
_charOff = row % (48*OFF_SIZE);
} else {
_sideNum = row / (56*OFF_SIZE);
_charOff = row % (56*OFF_SIZE);
}
_sideByteOff = _sideNum * sideSz;
assert_leq(row, ep._len);
assert_leq(_sideByteOff + sideSz, ep._ebwtTotSz);
#ifndef NO_PREFETCH
__builtin_prefetch((const void *)(ebwt + _sideByteOff),
0 /* prepare for read */,
PREFETCH_LOCALITY);
#endif
// prefetch this side too
_fw = ep._isBt2Index ? true : ((_sideNum & 1) != 0); // odd-numbered sides are forward
_by = _charOff >> 2; // byte within side
assert_lt(_by, (int)ep._sideBwtSz);
_bp = _charOff & 3; // bit-pair within byte
if(!_fw) {
_by = ep._sideBwtSz - _by - 1;
_bp ^= 3;
}
}
/// Return true iff this is an initialized SideLocus
bool valid() {
return _bp != -1;
}
/// Make this look like an invalid SideLocus
void invalidate() {
_bp = -1;
}
const uint8_t *side(const uint8_t* ebwt) const {
return ebwt + _sideByteOff;
}
const uint8_t *oside(const uint8_t* ebwt) const {
return ebwt + _sideByteOff + (_fw? (-128) : (128));
}
TIndexOffU _sideByteOff; // offset of top side within ebwt[]
TIndexOffU _sideNum; // index of side
uint16_t _charOff; // character offset within side
bool _fw; // side is forward or backward?
int16_t _by; // byte within side (not adjusted for bw sides)
int8_t _bp; // bitpair within byte (not adjusted for bw sides)
};
#include "ebwt_search_backtrack.h"
///////////////////////////////////////////////////////////////////////
//
// Functions for printing and sanity-checking Ebwts
//
///////////////////////////////////////////////////////////////////////
/**
* Given a range of positions in the EBWT array within the BWT portion
* of a forward side, print the characters at those positions along
* with a summary occ[] array.
*/
void Ebwt::printRangeFw(uint32_t begin, uint32_t end) const {
assert(isInMemory());
uint32_t occ[] = {0, 0, 0, 0};
assert_gt(end, begin);
for(uint32_t i = begin; i < end; i++) {
uint8_t by = this->_ebwt[i];
for(int j = 0; j < 4; j++) {
// Unpack from lowest to highest bit pair
int twoBit = unpack_2b_from_8b(by, j);
occ[twoBit]++;
cout << "ACGT"[twoBit];
}
assert_eq(0, (occ[0] + occ[1] + occ[2] + occ[3]) & 3);
}
cout << ":{" << occ[0] << "," << occ[1] << "," << occ[2] << "," << occ[3] << "}" << endl;
}
/**
* Given a range of positions in the EBWT array within the BWT portion
* of a backward side, print the characters at those positions along
* with a summary occ[] array.
*/
void Ebwt::printRangeBw(uint32_t begin, uint32_t end) const {
assert(isInMemory());
uint32_t occ[] = {0, 0, 0, 0};
assert_gt(end, begin);
for(uint32_t i = end-1; i >= begin; i--) {
uint8_t by = this->_ebwt[i];
for(int j = 3; j >= 0; j--) {
// Unpack from lowest to highest bit pair
int twoBit = unpack_2b_from_8b(by, j);
occ[twoBit]++;
cout << "ACGT"[twoBit];
}
assert_eq(0, (occ[0] + occ[1] + occ[2] + occ[3]) & 3);
if(i == 0) break;
}
cout << ":{" << occ[0] << "," << occ[1] << "," << occ[2] << "," << occ[3] << "}" << endl;
}
/**
* Check that the ebwt array is internally consistent up to (and not
* including) the given side index by re-counting the chars and
* comparing against the embedded occ[] arrays.
*/
void Ebwt::sanityCheckUpToSide(TIndexOff upToSide) const {
assert(isInMemory());
TIndexOffU occ[] = {0, 0, 0, 0};
ASSERT_ONLY(TIndexOffU occ_save[] = {0, 0});
TIndexOffU cur = 0; // byte pointer
const EbwtParams& eh = this->_eh;
bool fw = false;
while(cur < (TIndexOffU)(upToSide * eh._sideSz)) {
assert_leq(cur + eh._sideSz, eh._ebwtTotLen);
for(uint32_t i = 0; i < eh._sideBwtSz; i++) {
uint8_t by = this->_ebwt[cur + (fw ? i : eh._sideBwtSz-i-1)];
for(int j = 0; j < 4; j++) {
// Unpack from lowest to highest bit pair
int twoBit = unpack_2b_from_8b(by, fw ? j : 3-j);
occ[twoBit]++;
//if(_verbose) cout << "ACGT"[twoBit];
}
assert_eq(0, (occ[0] + occ[1] + occ[2] + occ[3]) % 4);
}
assert_eq(0, (occ[0] + occ[1] + occ[2] + occ[3]) % eh._sideBwtLen);
if(fw) {
// Finished forward bucket; check saved [G] and [T]
// against the two uint32_ts encoded here
ASSERT_ONLY(TIndexOffU *u32ebwt = reinterpret_cast<TIndexOffU*>(&this->_ebwt[cur + eh._sideBwtSz]));
ASSERT_ONLY(TIndexOffU gs = u32ebwt[0]);
ASSERT_ONLY(TIndexOffU ts = u32ebwt[1]);
assert_eq(gs, occ_save[0]);
assert_eq(ts, occ_save[1]);
fw = false;
} else {
// Finished backward bucket; check current [A] and [C]
// against the two uint32_ts encoded here
ASSERT_ONLY(TIndexOffU *u32ebwt = reinterpret_cast<TIndexOffU*>(&this->_ebwt[cur + eh._sideBwtSz]));
ASSERT_ONLY(TIndexOffU as = u32ebwt[0]);
ASSERT_ONLY(TIndexOffU cs = u32ebwt[1]);
assert(as == occ[0] || as == occ[0]-1); // one 'a' is a skipped '$' and doesn't count toward occ[]
assert_eq(cs, occ[1]);
ASSERT_ONLY(occ_save[0] = occ[2]); // save gs
ASSERT_ONLY(occ_save[1] = occ[3]); // save ts
fw = true;
}
cur += eh._sideSz;
}
}
/**
* Sanity-check various pieces of the Ebwt
*/
void Ebwt::sanityCheckAll(int reverse) const {
const EbwtParams& eh = this->_eh;
assert(isInMemory());
// Check ftab
for(TIndexOffU i = 1; i < eh._ftabLen; i++) {
assert_geq(this->ftabHi(i), this->ftabLo(i-1));
assert_geq(this->ftabLo(i), this->ftabHi(i-1));
assert_leq(this->ftabHi(i), eh._bwtLen+1);
}
assert_eq(this->ftabHi(eh._ftabLen-1), eh._bwtLen);
// Check offs
TIndexOff seenLen = (eh._bwtLen + 31) >> ((TIndexOffU)5);
TIndexOff *seen;
try {
seen = new TIndexOff[seenLen]; // bitvector marking seen offsets
} catch(bad_alloc& e) {
cerr << "Out of memory allocating seen[] at " << __FILE__ << ":" << __LINE__ << endl;
throw e;
}
memset(seen, 0, OFF_SIZE * seenLen);
TIndexOffU offsLen = eh._offsLen;
for(TIndexOffU i = 0; i < offsLen; i++) {
assert_lt(this->_offs[i], eh._bwtLen);
TIndexOff w = this->_offs[i] >> 5;
TIndexOff r = this->_offs[i] & 31;
assert_eq(0, (seen[w] >> r) & 1); // shouldn't have been seen before
seen[w] |= (1 << r);
}
delete[] seen;
// Check nPat
assert_gt(this->_nPat, 0);
// Check plen, flen
for(TIndexOffU i = 0; i < this->_nPat; i++) {
assert_geq(this->_plen[i], 0);
}
// Check rstarts
for(TIndexOffU i = 0; i < this->_nFrag-1; i++) {
assert_gt(this->_rstarts[(i+1)*3], this->_rstarts[i*3]);
if(reverse == REF_READ_REVERSE) {
assert(this->_rstarts[(i*3)+1] >= this->_rstarts[((i+1)*3)+1]);
} else {
assert(this->_rstarts[(i*3)+1] <= this->_rstarts[((i+1)*3)+1]);
}
}
// Check ebwt
sanityCheckUpToSide(eh._numSides);
VMSG_NL("Ebwt::sanityCheck passed");
}
///////////////////////////////////////////////////////////////////////
//
// Functions for searching Ebwts
//
///////////////////////////////////////////////////////////////////////
/**
* Return the final character in row i (i.e. the i'th character in the
* BWT transform). Note that the 'L' in the name of the function
* stands for 'last', as in the literature.
*/
inline int Ebwt::rowL(const SideLocus& l) const {
// Extract and return appropriate bit-pair
#ifdef SIXTY4_FORMAT
return (((uint64_t*)l.side(this->_ebwt))[l._by >> 3] >> ((((l._by & 7) << 2) + l._bp) << 1)) & 3;
#else
return unpack_2b_from_8b(l.side(this->_ebwt)[l._by], l._bp);
#endif
}
/**
* Inline-function version of the above. This does not always seem to
* be inlined
*/
#if 0
// Use gcc's intrinsic popcountll. I don't recommend it because it
// seems to be somewhat slower than the bit-bashing pop64 routine both
// on an AMD server and on an Intel workstation. On the other hand,
// perhaps when the builtin is used GCC is smart enough to insert a
// pop-count instruction on architectures that have one (e.g. Itanium).
// For now, it's disabled.
#define pop64(x) __builtin_popcountll(x)
#elif 0
__declspec naked int __stdcall pop64
(uint64_t v)
{
static const uint64_t C55 = 0x5555555555555555ll;
static const uint64_t C33 = 0x3333333333333333ll;
static const uint64_t C0F = 0x0F0F0F0F0F0F0F0Fll;
__asm {
MOVD MM0, [ESP+4] ;v_low
PUNPCKLDQ MM0, [ESP+8] ;v
MOVQ MM1, MM0 ;v
PSRLD MM0, 1 ;v >> 1
PAND MM0, [C55] ;(v >> 1) & 0x55555555
PSUBD MM1, MM0 ;w = v - ((v >> 1) & 0x55555555)
MOVQ MM0, MM1 ;w
PSRLD MM1, 2 ;w >> 2
PAND MM0, [C33] ;w & 0x33333333
PAND MM1, [C33] ;(w >> 2) & 0x33333333
PADDD MM0, MM1 ;x = (w & 0x33333333) +
; ((w >> 2) & 0x33333333)
MOVQ MM1, MM0 ;x
PSRLD MM0, 4 ;x >> 4
PADDD MM0, MM1 ;x + (x >> 4)
PAND MM0, [C0F] ;y = (x + (x >> 4) & 0x0F0F0F0F)
PXOR MM1, MM1 ;0
PSADBW (MM0, MM1) ;sum across all 8 bytes
MOVD EAX, MM0 ;result in EAX per calling
; convention
EMMS ;clear MMX state
RET 8 ;pop 8-byte argument off stack
; and return
}
}
#elif 0
// Use a bytewise LUT version of popcount. This is slower than the
// bit-bashing pop64 routine both on an AMD server and on an Intel
// workstation. It seems to be about the same speed as the GCC builtin
// on Intel, and a bit faster than it on AMD. For now, it's disabled.
const int popcntU8Table[256] = {
0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,1,2,2,3,2,3,3,4,2,3,3,4,3,4,4,5,
1,2,2,3,2,3,3,4,2,3,3,4,3,4,4,5,2,3,3,4,3,4,4,5,3,4,4,5,4,5,5,6,
1,2,2,3,2,3,3,4,2,3,3,4,3,4,4,5,2,3,3,4,3,4,4,5,3,4,4,5,4,5,5,6,
2,3,3,4,3,4,4,5,3,4,4,5,4,5,5,6,3,4,4,5,4,5,5,6,4,5,5,6,5,6,6,7,
1,2,2,3,2,3,3,4,2,3,3,4,3,4,4,5,2,3,3,4,3,4,4,5,3,4,4,5,4,5,5,6,
2,3,3,4,3,4,4,5,3,4,4,5,4,5,5,6,3,4,4,5,4,5,5,6,4,5,5,6,5,6,6,7,
2,3,3,4,3,4,4,5,3,4,4,5,4,5,5,6,3,4,4,5,4,5,5,6,4,5,5,6,5,6,6,7,
3,4,4,5,4,5,5,6,4,5,5,6,5,6,6,7,4,5,5,6,5,6,6,7,5,6,6,7,6,7,7,8
};
// Use this bytewise population count table
inline static int pop64(uint64_t x) {
const unsigned char * p = (const unsigned char *) &x;
return popcntU8Table[p[0]] +
popcntU8Table[p[1]] +
popcntU8Table[p[2]] +
popcntU8Table[p[3]] +
popcntU8Table[p[4]] +
popcntU8Table[p[5]] +
popcntU8Table[p[6]] +
popcntU8Table[p[7]];
}
#else
#ifdef POPCNT_CAPABILITY // wrapping of "struct"
struct USE_POPCNT_GENERIC {
#endif
// Use this standard bit-bashing population count
inline static int pop64(uint64_t x) {
x = x - ((x >> 1) & 0x5555555555555555llu);
x = (x & 0x3333333333333333llu) + ((x >> 2) & 0x3333333333333333llu);
x = (x + (x >> 4)) & 0x0F0F0F0F0F0F0F0Fllu;
x = x + (x >> 8);
x = x + (x >> 16);
x = x + (x >> 32);
return x & 0x3F;
}
#ifdef POPCNT_CAPABILITY // wrapping a "struct"
};
#endif
#ifdef POPCNT_CAPABILITY
struct USE_POPCNT_INSTRUCTION {
inline static int pop64(uint64_t x) {
int64_t count;
asm ("popcntq %[x],%[count]\n": [count] "=&r" (count): [x] "r" (x));
return count;
}
};
#endif
#endif
/**
* Tricky-bit-bashing bitpair counting for given two-bit value (0-3)
* within a 64-bit argument.
*/
#ifdef POPCNT_CAPABILITY
template<typename Operation>
#endif
inline static int countInU64(int c, uint64_t dw) {
uint64_t c0 = c_table[c];
uint64_t x0 = dw ^ c0;
uint64_t x1 = (x0 >> 1);
uint64_t x2 = x1 & (0x5555555555555555llu);
uint64_t x3 = x0 & x2;
#ifdef POPCNT_CAPABILITY
uint64_t tmp = Operation().pop64(x3);
#else
uint64_t tmp = pop64(x3);
#endif
return (int) tmp;
}
/**
* Tricky-bit-bashing bitpair counting for given two-bit value (0-3)
* within a 64-bit argument.
*
* Function gets 2.32% in profile
*/
#ifdef POPCNT_CAPABILITY
template<typename Operation>
#endif
inline static void countInU64Ex(uint64_t dw, TIndexOffU* arrs) {
uint64_t c0 = c_table[0];
uint64_t x0 = dw ^ c0;
uint64_t x1 = (x0 >> 1);
uint64_t x2 = x1 & (0x5555555555555555llu);
uint64_t x3 = x0 & x2;
#ifdef POPCNT_CAPABILITY
uint64_t tmp = Operation().pop64(x3);
#else
uint64_t tmp = pop64(x3);
#endif
arrs[0] += (uint32_t) tmp;
c0 = c_table[1];
x0 = dw ^ c0;
x1 = (x0 >> 1);
x2 = x1 & (0x5555555555555555llu);
x3 = x0 & x2;
#ifdef POPCNT_CAPABILITY
tmp = Operation().pop64(x3);
#else
tmp = pop64(x3);
#endif
arrs[1] += (uint32_t) tmp;
c0 = c_table[2];
x0 = dw ^ c0;
x1 = (x0 >> 1);
x2 = x1 & (0x5555555555555555llu);
x3 = x0 & x2;
#ifdef POPCNT_CAPABILITY
tmp = Operation().pop64(x3);
#else
tmp = pop64(x3);
#endif
arrs[2] += (uint32_t) tmp;
c0 = c_table[3];
x0 = dw ^ c0;
x1 = (x0 >> 1);
x2 = x1 & (0x5555555555555555llu);
x3 = x0 & x2;
#ifdef POPCNT_CAPABILITY
tmp = Operation().pop64(x3);
#else
tmp = pop64(x3);
#endif
arrs[3] += (uint32_t) tmp;
}
/**
* Counts the number of occurrences of character 'c' in the given Ebwt
* side up to (but not including) the given byte/bitpair (by/bp).
*
* This is a performance-critical function. This is the top search-
* related hit in the time profile.
*
* Function gets 11.09% in profile
*/
inline TIndexOffU Ebwt::countUpTo(const SideLocus& l, int c) const {
// Count occurrences of c in each 64-bit (using bit trickery);
// Someday countInU64() and pop() functions should be
// vectorized/SSE-ized in case that helps.
TIndexOffU cCnt = 0;
const uint8_t *side = l.side(this->_ebwt);
int i = 0;
#if 1
#ifdef POPCNT_CAPABILITY
if ( _usePOPCNTinstruction) {
for(; i + 7 < l._by; i += 8) {
cCnt += countInU64<USE_POPCNT_INSTRUCTION>(c, *(uint64_t*)&side[i]);
}
}
else {
for(; i + 7 < l._by; i += 8) {
cCnt += countInU64<USE_POPCNT_GENERIC>(c, *(uint64_t*)&side[i]);
}
}
#else
for(; i + 7 < l._by; i += 8) {
cCnt += countInU64(c, *(uint64_t*)&side[i]);
}
#endif
#else
for(; i + 2 < l._by; i += 2) {
cCnt += cCntLUT_16b_4[c][*(uint16_t*)&side[i]];
}
#endif
#ifdef SIXTY4_FORMAT
// Calculate number of bit pairs to shift off the end
const int bpShiftoff = 32 - (((l._by & 7) << 2) + l._bp);
if(bpShiftoff < 32) {
assert_lt(bpShiftoff, 32);
const uint64_t sw = (*(uint64_t*)&side[i]) << (bpShiftoff << 1);
#ifdef POPCNT_CAPABILITY
if (_usePOPCNTinstruction) {
cCnt += countInU64<USE_POPCNT_INSTRUCTION>(c, sw);
}
else {
cCnt += countInU64<USE_POPCNT_GENERIC>(c, sw);
}
#else
cCnt += countInU64(c, sw);
#endif
if(c == 0) cCnt -= bpShiftoff; // we turned these into As
}
#else
// Count occurences of c in the rest of the side (using LUT)
for(; i < l._by; i++) {
cCnt += cCntLUT_4[0][c][side[i]];
}
// Count occurences of c in the rest of the byte
if(l._bp > 0) {
cCnt += cCntLUT_4[(int)l._bp][c][side[i]];
}
#endif
return cCnt;
}
/**
* Counts the number of occurrences of character 'c' in the given Ebwt
* side up to (but not including) the given byte/bitpair (by/bp).
*/
inline void Ebwt::countUpToEx(const SideLocus& l, TIndexOffU* arrs) const {
int i = 0;
// Count occurrences of c in each 64-bit (using bit trickery);
// note: this seems does not seem to lend a significant boost to
// performance. If you comment out this whole loop (which won't
// affect correctness - it will just cause the following loop to
// take up the slack) then runtime does not change noticeably.
// Someday the countInU64() and pop() functions should be
// vectorized/SSE-ized in case that helps.
const uint8_t *side = l.side(this->_ebwt);
#ifdef POPCNT_CAPABILITY
if (_usePOPCNTinstruction) {
for(; i+7 < l._by; i += 8) {
countInU64Ex<USE_POPCNT_INSTRUCTION>(*(uint64_t*)&side[i], arrs);
}
}
else {
for(; i+7 < l._by; i += 8) {
countInU64Ex<USE_POPCNT_GENERIC>(*(uint64_t*)&side[i], arrs);
}
}
#else
for(; i+7 < l._by; i += 8) {
countInU64Ex(*(uint64_t*)&side[i], arrs);
}
#endif
#ifdef SIXTY4_FORMAT
// Calculate number of bit pairs to shift off the end
const int bpShiftoff = 32 - (((l._by & 7) << 2) + l._bp);
assert_leq(bpShiftoff, 32);
if(bpShiftoff < 32) {
const uint64_t sw = (*(uint64_t*)&l.side(this->_ebwt)[i]) << (bpShiftoff << 1);
#ifdef POPCNT_CAPABILITY
if (_usePOPCNTinstruction) {
countInU64Ex<USE_POPCNT_INSTRUCTION>(sw, arrs);
}
else{
countInU64Ex<USE_POPCNT_GENERIC>(sw, arrs);
}
#else
countInU64Ex(sw, arrs);
#endif
arrs[0] -= bpShiftoff;
}
#else
// Count occurences of c in the rest of the side (using LUT)
for(; i < l._by; i++) {
arrs[0] += cCntLUT_4[0][0][side[i]];
arrs[1] += cCntLUT_4[0][1][side[i]];
arrs[2] += cCntLUT_4[0][2][side[i]];
arrs[3] += cCntLUT_4[0][3][side[i]];
}
// Count occurences of c in the rest of the byte
if(l._bp > 0) {
arrs[0] += cCntLUT_4[(int)l._bp][0][side[i]];
arrs[1] += cCntLUT_4[(int)l._bp][1][side[i]];
arrs[2] += cCntLUT_4[(int)l._bp][2][side[i]];
arrs[3] += cCntLUT_4[(int)l._bp][3][side[i]];
}
#endif
}
/**
* Count all occurrences of character c from the beginning of the
* forward side to <by,bp> and add in the occ[] count up to the side
* break just prior to the side.
*/
inline TIndexOffU Ebwt::countFwSide(const SideLocus& l, int c) const { /* check */
assert_lt(c, 4);
assert_geq(c, 0);
assert_lt(l._by, (int)this->_eh._sideBwtSz);
assert_geq(l._by, 0);
assert_lt(l._bp, 4);
assert_geq(l._bp, 0);
const uint8_t *side = l.side(this->_ebwt);
TIndexOffU cCnt = countUpTo(l, c);
assert_leq(cCnt, this->_eh._sideBwtLen);
if(c == 0 && l._sideByteOff <= _zEbwtByteOff && l._sideByteOff + l._by >= _zEbwtByteOff) {
// Adjust for the fact that we represented $ with an 'A', but
// shouldn't count it as an 'A' here
if((l._sideByteOff + l._by > _zEbwtByteOff) ||
(l._sideByteOff + l._by == _zEbwtByteOff && l._bp > _zEbwtBpOff))
{
cCnt--; // Adjust for '$' looking like an 'A'
}
}
TIndexOffU ret;
// Now factor in the occ[] count at the side break
if(c < 2) {
const TIndexOffU *ac = reinterpret_cast<const TIndexOffU*>(side - 2*OFF_SIZE);
assert_leq(ac[0], this->_eh._numSides * this->_eh._sideBwtLen); // b/c it's used as padding
assert_leq(ac[1], this->_eh._len);
ret = ac[c] + cCnt + this->_fchr[c];
} else {
const TIndexOffU *gt = reinterpret_cast<const TIndexOffU*>(side + this->_eh._sideSz - 2*OFF_SIZE); // next
assert_leq(gt[0], this->_eh._len); assert_leq(gt[1], this->_eh._len);
ret = gt[c-2] + cCnt + this->_fchr[c];
}
#ifndef NDEBUG
assert_leq(ret, this->_fchr[c+1]); // can't have jumpded into next char's section
if(c == 0) {
assert_leq(cCnt, this->_eh._sideBwtLen);
} else {
assert_leq(ret, this->_eh._bwtLen);
}
#endif
return ret;
}
/**
* Count all occurrences of character c from the beginning of the
* forward side to <by,bp> and add in the occ[] count up to the side
* break just prior to the side.
*/
inline void Ebwt::countFwSideEx(const SideLocus& l, TIndexOffU* arrs) const
{
assert_lt(l._by, (int)this->_eh._sideBwtSz);
assert_geq(l._by, 0);
assert_lt(l._bp, 4);
assert_geq(l._bp, 0);
countUpToEx(l, arrs);
#ifndef NDEBUG
assert_leq(arrs[0], this->_fchr[1]); // can't have jumped into next char's section
assert_leq(arrs[1], this->_fchr[2]); // can't have jumped into next char's section
assert_leq(arrs[2], this->_fchr[3]); // can't have jumped into next char's section
assert_leq(arrs[3], this->_fchr[4]); // can't have jumped into next char's section
#endif
assert_leq(arrs[0], this->_eh._sideBwtLen);
assert_leq(arrs[1], this->_eh._sideBwtLen);
assert_leq(arrs[2], this->_eh._sideBwtLen);
assert_leq(arrs[3], this->_eh._sideBwtLen);
const uint8_t *side = l.side(this->_ebwt);
if(l._sideByteOff <= _zEbwtByteOff && l._sideByteOff + l._by >= _zEbwtByteOff) {
// Adjust for the fact that we represented $ with an 'A', but
// shouldn't count it as an 'A' here
if((l._sideByteOff + l._by > _zEbwtByteOff) ||
(l._sideByteOff + l._by == _zEbwtByteOff && l._bp > _zEbwtBpOff))
{
arrs[0]--; // Adjust for '$' looking like an 'A'
}
}
// Now factor in the occ[] count at the side break
const TIndexOffU *ac = reinterpret_cast<const TIndexOffU*>(side - 2*OFF_SIZE);
const TIndexOffU *gt = reinterpret_cast<const TIndexOffU*>(side + this->_eh._sideSz - 2*OFF_SIZE);
#ifndef NDEBUG
assert_leq(ac[0], this->_fchr[1] + this->_eh.sideBwtLen());
assert_leq(ac[1], this->_fchr[2]-this->_fchr[1]);
assert_leq(gt[0], this->_fchr[3]-this->_fchr[2]);
assert_leq(gt[1], this->_fchr[4]-this->_fchr[3]);
#endif
assert_leq(ac[0], this->_eh._len + this->_eh.sideBwtLen()); assert_leq(ac[1], this->_eh._len);
assert_leq(gt[0], this->_eh._len); assert_leq(gt[1], this->_eh._len);
arrs[0] += (ac[0] + this->_fchr[0]);
arrs[1] += (ac[1] + this->_fchr[1]);
arrs[2] += (gt[0] + this->_fchr[2]);
arrs[3] += (gt[1] + this->_fchr[3]);
#ifndef NDEBUG
assert_leq(arrs[0], this->_fchr[1]); // can't have jumped into next char's section
assert_leq(arrs[1], this->_fchr[2]); // can't have jumped into next char's section
assert_leq(arrs[2], this->_fchr[3]); // can't have jumped into next char's section
assert_leq(arrs[3], this->_fchr[4]); // can't have jumped into next char's section
#endif
}
/**
* Count all instances of character c from <by,bp> to the logical end
* (actual beginning) of the backward side, and subtract that from the
* occ[] count up to the side break.
*/
inline TIndexOffU Ebwt::countBwSide(const SideLocus& l, int c) const {
assert_lt(c, 4);
assert_geq(c, 0);
assert_lt(l._by, (int)this->_eh._sideBwtSz);
assert_geq(l._by, 0);
assert_lt(l._bp, 4);
assert_geq(l._bp, 0);
const uint8_t *side = l.side(this->_ebwt);
TIndexOffU cCnt = countUpTo(l, c);
if(rowL(l) == c) cCnt++;
assert_leq(cCnt, this->_eh._sideBwtLen);
if(c == 0 && l._sideByteOff <= _zEbwtByteOff && l._sideByteOff + l._by >= _zEbwtByteOff) {
// Adjust for the fact that we represented $ with an 'A', but
// shouldn't count it as an 'A' here
if((l._sideByteOff + l._by > _zEbwtByteOff) ||
(l._sideByteOff + l._by == _zEbwtByteOff && l._bp >= _zEbwtBpOff))
{
cCnt--;
}
}
TIndexOffU ret;
// Now factor in the occ[] count at the side break
if(c < 2) {
const TIndexOffU *ac = reinterpret_cast<const TIndexOffU*>(side + this->_eh._sideSz - 2*OFF_SIZE);
assert_leq(ac[0], this->_eh._numSides * this->_eh._sideBwtLen); // b/c it's used as padding
assert_leq(ac[1], this->_eh._len);
ret = ac[c] - cCnt + this->_fchr[c];
} else {
const TIndexOffU *gt = reinterpret_cast<const TIndexOffU*>(side + (2*this->_eh._sideSz) - 2*OFF_SIZE); // next
assert_leq(gt[0], this->_eh._len); assert_leq(gt[1], this->_eh._len);
ret = gt[c-2] - cCnt + this->_fchr[c];
}
#ifndef NDEBUG
assert_leq(ret, this->_fchr[c+1]); // can't have jumped into next char's section
if(c == 0) {
assert_leq(cCnt, this->_eh._sideBwtLen);
} else {
assert_lt(ret, this->_eh._bwtLen);
}
#endif
return ret;
}
/**
* Count all instances of character c from <by,bp> to the logical end
* (actual beginning) of the backward side, and subtract that from the
* occ[] count up to the side break.
*/
inline void Ebwt::countBwSideEx(const SideLocus& l, TIndexOffU* arrs) const {
assert_lt(l._by, (int)this->_eh._sideBwtSz);
assert_geq(l._by, 0);
assert_lt(l._bp, 4);
assert_geq(l._bp, 0);
const uint8_t *side = l.side(this->_ebwt);
countUpToEx(l, arrs);
arrs[rowL(l)]++;
assert_leq(arrs[0], this->_eh._sideBwtLen);
assert_leq(arrs[1], this->_eh._sideBwtLen);
assert_leq(arrs[2], this->_eh._sideBwtLen);
assert_leq(arrs[3], this->_eh._sideBwtLen);
if(l._sideByteOff <= _zEbwtByteOff && l._sideByteOff + l._by >= _zEbwtByteOff) {
// Adjust for the fact that we represented $ with an 'A', but
// shouldn't count it as an 'A' here
if((l._sideByteOff + l._by > _zEbwtByteOff) ||
(l._sideByteOff + l._by == _zEbwtByteOff && l._bp >= _zEbwtBpOff))
{
arrs[0]--; // Adjust for '$' looking like an 'A'
}
}
// Now factor in the occ[] count at the side break
const TIndexOffU *ac = reinterpret_cast<const TIndexOffU*>(side + this->_eh._sideSz - 2*OFF_SIZE);
const TIndexOffU *gt = reinterpret_cast<const TIndexOffU*>(side + (2*this->_eh._sideSz) - 2*OFF_SIZE);
#ifndef NDEBUG
assert_leq(ac[0], this->_fchr[1] + this->_eh.sideBwtLen());
assert_leq(ac[1], this->_fchr[2]-this->_fchr[1]);
assert_leq(gt[0], this->_fchr[3]-this->_fchr[2]);
assert_leq(gt[1], this->_fchr[4]-this->_fchr[3]);
#endif
assert_leq(ac[0], this->_eh._len + this->_eh.sideBwtLen()); assert_leq(ac[1], this->_eh._len);
assert_leq(gt[0], this->_eh._len); assert_leq(gt[1], this->_eh._len);
arrs[0] = (ac[0] - arrs[0] + this->_fchr[0]);
arrs[1] = (ac[1] - arrs[1] + this->_fchr[1]);
arrs[2] = (gt[0] - arrs[2] + this->_fchr[2]);
arrs[3] = (gt[1] - arrs[3] + this->_fchr[3]);
#ifndef NDEBUG
assert_leq(arrs[0], this->_fchr[1]); // can't have jumped into next char's section
assert_leq(arrs[1], this->_fchr[2]); // can't have jumped into next char's section
assert_leq(arrs[2], this->_fchr[3]); // can't have jumped into next char's section
assert_leq(arrs[3], this->_fchr[4]); // can't have jumped into next char's section
#endif
}
#define WITHIN_BWT_LEN(x) \
assert_leq(x[0], this->_eh._sideBwtLen); \
assert_leq(x[1], this->_eh._sideBwtLen); \
assert_leq(x[2], this->_eh._sideBwtLen); \
assert_leq(x[3], this->_eh._sideBwtLen)
#define WITHIN_FCHR(x) \
assert_leq(x[0], this->fchr()[1]); \
assert_leq(x[1], this->fchr()[2]); \
assert_leq(x[2], this->fchr()[3]); \
assert_leq(x[3], this->fchr()[4])
inline TIndexOffU Ebwt::countBt2Side(const SideLocus& l, int c) const {
assert_range(0, 3, c);
assert_range(0, (int)this->_eh._sideBwtSz-1, (int)l._by);
assert_range(0, 3, (int)l._bp);
const uint8_t *side = l.side(this->ebwt());
TIndexOffU cCnt = countUpTo(l, c);
assert_leq(cCnt, this->_eh._sideBwtLen);
if(c == 0 && l._sideByteOff <= _zEbwtByteOff && l._sideByteOff + l._by >= _zEbwtByteOff) {
// Adjust for the fact that we represented $ with an 'A', but
// shouldn't count it as an 'A' here
if((l._sideByteOff + l._by > _zEbwtByteOff) ||
(l._sideByteOff + l._by == _zEbwtByteOff && l._bp > _zEbwtBpOff))
{
cCnt--; // Adjust for '$' looking like an 'A'
}
}
TIndexOffU ret;
// Now factor in the occ[] count at the side break
const uint8_t *acgt8 = side + _eh._sideBwtSz;
const TIndexOffU *acgt = reinterpret_cast<const TIndexOffU*>(acgt8);
assert_leq(acgt[0], this->_eh._numSides * this->_eh._sideBwtLen); // b/c it's used as padding
assert_leq(acgt[1], this->_eh._len);
assert_leq(acgt[2], this->_eh._len);
assert_leq(acgt[3], this->_eh._len);
ret = acgt[c] + cCnt + this->fchr()[c];
#ifndef NDEBUG
assert_leq(ret, this->fchr()[c+1]); // can't have jumpded into next char's section
if(c == 0) {
assert_leq(cCnt, this->_eh._sideBwtLen);
} else {
assert_leq(ret, this->_eh._bwtLen);
}
#endif
return ret;
}
/**
* Count all occurrences of character c from the beginning of the
* forward side to <by,bp> and add in the occ[] count up to the side
* break just prior to the side.
*
* A forward side is shaped like:
*
* [A] [C] XXXXXXXXXXXXXXXX
* -4- -4- --------56------ (numbers in bytes)
* ^
* Side ptr (result from SideLocus.side())
*
* And following it is a reverse side shaped like:
*
* [G] [T] XXXXXXXXXXXXXXXX
* -4- -4- --------56------ (numbers in bytes)
* ^
* Side ptr (result from SideLocus.side())
*
*/
inline void Ebwt::countBt2SideEx(const SideLocus& l, TIndexOffU* arrs) const {
assert_range(0, (int)this->_eh._sideBwtSz-1, (int)l._by);
assert_range(0, 3, (int)l._bp);
countUpToEx(l, arrs);
if(l._sideByteOff <= _zEbwtByteOff && l._sideByteOff + l._by >= _zEbwtByteOff) {
// Adjust for the fact that we represented $ with an 'A', but
// shouldn't count it as an 'A' here
if((l._sideByteOff + l._by > _zEbwtByteOff) ||
(l._sideByteOff + l._by == _zEbwtByteOff && l._bp > _zEbwtBpOff))
{
arrs[0]--; // Adjust for '$' looking like an 'A'
}
}
WITHIN_FCHR(arrs);
WITHIN_BWT_LEN(arrs);
// Now factor in the occ[] count at the side break
const uint8_t *side = l.side(this->ebwt());
const uint8_t *acgt16 = side + this->_eh._sideSz - OFF_SIZE*4;
const TIndexOffU *acgt = reinterpret_cast<const TIndexOffU*>(acgt16);
assert_leq(acgt[0], this->fchr()[1] + this->_eh.sideBwtLen());
assert_leq(acgt[1], this->fchr()[2]-this->fchr()[1]);
assert_leq(acgt[2], this->fchr()[3]-this->fchr()[2]);
assert_leq(acgt[3], this->fchr()[4]-this->fchr()[3]);
assert_leq(acgt[0], this->_eh._len + this->_eh.sideBwtLen());
assert_leq(acgt[1], this->_eh._len);
assert_leq(acgt[2], this->_eh._len);
assert_leq(acgt[3], this->_eh._len);
arrs[0] += (acgt[0] + this->fchr()[0]);
arrs[1] += (acgt[1] + this->fchr()[1]);
arrs[2] += (acgt[2] + this->fchr()[2]);
arrs[3] += (acgt[3] + this->fchr()[3]);
WITHIN_FCHR(arrs);
}
/**
* Given top and bot loci, calculate counts of all four DNA chars up to
* those loci. Used for more advanced backtracking-search.
*/
inline void Ebwt::mapLFEx(const SideLocus& ltop,
const SideLocus& lbot,
TIndexOffU *tops,
TIndexOffU *bots
ASSERT_ONLY(, bool overrideSanity)
) const
{
// TODO: Where there's overlap, reuse the count for the overlapping
// portion
#ifdef EBWT_STATS
const_cast<Ebwt*>(this)->mapLFExs_++;
#endif
assert_eq(0, tops[0]); assert_eq(0, bots[0]);
assert_eq(0, tops[1]); assert_eq(0, bots[1]);
assert_eq(0, tops[2]); assert_eq(0, bots[2]);
assert_eq(0, tops[3]); assert_eq(0, bots[3]);
if(ltop._fw) {
// Forward side
!_eh._isBt2Index ? countFwSideEx(ltop, tops)
: countBt2SideEx(ltop, tops);
} else {
countBwSideEx(ltop, tops); // Backward side
}
if(lbot._fw) {
// Forward side
!_eh._isBt2Index ? countFwSideEx(lbot, bots)
: countBt2SideEx(lbot, bots);
} else {
countBwSideEx(lbot, bots); // Backward side
}
#ifndef NDEBUG
if(_sanity && !overrideSanity) {
// Make sure results match up with individual calls to mapLF;
// be sure to override sanity-checking in the callee, or we'll
// have infinite recursion
assert_eq(mapLF(ltop, 0, true), tops[0]);
assert_eq(mapLF(ltop, 1, true), tops[1]);
assert_eq(mapLF(ltop, 2, true), tops[2]);
assert_eq(mapLF(ltop, 3, true), tops[3]);
assert_eq(mapLF(lbot, 0, true), bots[0]);
assert_eq(mapLF(lbot, 1, true), bots[1]);
assert_eq(mapLF(lbot, 2, true), bots[2]);
assert_eq(mapLF(lbot, 3, true), bots[3]);
}
#endif
}
#ifndef NDEBUG
/**
* Given top and bot loci, calculate counts of all four DNA chars up to
* those loci. Used for more advanced backtracking-search.
*/
inline void Ebwt::mapLFEx(const SideLocus& l,
TIndexOffU *arrs
ASSERT_ONLY(, bool overrideSanity)
) const
{
assert_eq(0, arrs[0]);
assert_eq(0, arrs[1]);
assert_eq(0, arrs[2]);
assert_eq(0, arrs[3]);
if(l._fw) {
// Forward side
!_eh._isBt2Index ? countFwSideEx(l, arrs)
: countBt2SideEx(l, arrs);
} else {
countBwSideEx(l, arrs); // Backward side
}
#ifndef NDEBUG
if(_sanity && !overrideSanity) {
// Make sure results match up with individual calls to mapLF;
// be sure to override sanity-checking in the callee, or we'll
// have infinite recursion
assert_eq(mapLF(l, 0, true), arrs[0]);
assert_eq(mapLF(l, 1, true), arrs[1]);
assert_eq(mapLF(l, 2, true), arrs[2]);
assert_eq(mapLF(l, 3, true), arrs[3]);
}
#endif
}
#endif
/**
* Given row i, return the row that the LF mapping maps i to.
*/
inline TIndexOffU Ebwt::mapLF(const SideLocus& l
ASSERT_ONLY(, bool overrideSanity)
) const
{
#ifdef EBWT_STATS
const_cast<Ebwt*>(this)->mapLFs_++;
#endif
TIndexOffU ret;
assert(l.side(this->_ebwt) != NULL);
int c = rowL(l);
assert_lt(c, 4);
assert_geq(c, 0);
if(l._fw) {
// Forward side
ret = !_eh._isBt2Index ? countFwSide(l, c)
: countBt2Side(l, c);
}
else {
ret = countBwSide(l, c); // Backward side
}
assert_lt(ret, this->_eh._bwtLen);
#ifndef NDEBUG
if(_sanity && !overrideSanity) {
// Make sure results match up with results from mapLFEx;
// be sure to override sanity-checking in the callee, or we'll
// have infinite recursion
TIndexOffU arrs[] = { 0, 0, 0, 0 };
mapLFEx(l, arrs, true);
assert_eq(arrs[c], ret);
}
#endif
return ret;
}
/**
* Given row i and character c, return the row that the LF mapping maps
* i to on character c.
*/
inline TIndexOffU Ebwt::mapLF(const SideLocus& l, int c
ASSERT_ONLY(, bool overrideSanity)
) const
{
#ifdef EBWT_STATS
const_cast<Ebwt*>(this)->mapLFcs_++;
#endif
TIndexOffU ret;
assert_lt(c, 4);
assert_geq(c, 0);
if(l._fw) {
// Forward side
ret = !_eh._isBt2Index ? countFwSide(l, c)
: countBt2Side(l, c);
}
else {
ret = countBwSide(l, c); // Backward side
}
assert_lt(ret, this->_eh._bwtLen);
#ifndef NDEBUG
if(_sanity && !overrideSanity) {
// Make sure results match up with results from mapLFEx;
// be sure to override sanity-checking in the callee, or we'll
// have infinite recursion
TIndexOffU arrs[] = { 0, 0, 0, 0 };
mapLFEx(l, arrs, true);
assert_eq(arrs[c], ret);
}
#endif
return ret;
}
/**
* Given row i and character c, return the row that the LF mapping maps
* i to on character c.
*/
inline TIndexOffU Ebwt::mapLF1(TIndexOffU row, const SideLocus& l, int c
ASSERT_ONLY(, bool overrideSanity)
) const
{
#ifdef EBWT_STATS
const_cast<Ebwt*>(this)->mapLF1cs_++;
#endif
if(rowL(l) != c || row == _zOff) return OFF_MASK;
TIndexOffU ret;
assert_lt(c, 4);
assert_geq(c, 0);
if(l._fw) {
// Forward side
ret = !_eh._isBt2Index ? countFwSide(l, c)
: countBt2Side(l, c);
} else {
ret = countBwSide(l, c); // Backward side
}
assert_lt(ret, this->_eh._bwtLen);
#ifndef NDEBUG
if(_sanity && !overrideSanity) {
// Make sure results match up with results from mapLFEx;
// be sure to override sanity-checking in the callee, or we'll
// have infinite recursion
TIndexOffU arrs[] = { 0, 0, 0, 0 };
mapLFEx(l, arrs, true);
assert_eq(arrs[c], ret);
}
#endif
return ret;
}
/**
* Given row i and character c, return the row that the LF mapping maps
* i to on character c.
*/
inline int Ebwt::mapLF1(TIndexOffU& row, const SideLocus& l
ASSERT_ONLY(, bool overrideSanity)
) const
{
#ifdef EBWT_STATS
const_cast<Ebwt*>(this)->mapLF1s_++;
#endif
if(row == _zOff) return -1;
int c = rowL(l);
assert_lt(c, 4);
assert_geq(c, 0);
if(l._fw) {
// Forward side
row = !_eh._isBt2Index ? countFwSide(l, c)
: countBt2Side(l, c);
} else {
row = countBwSide(l, c); // Backward side
}
assert_lt(row, this->_eh._bwtLen);
#ifndef NDEBUG
if(_sanity && !overrideSanity) {
// Make sure results match up with results from mapLFEx;
// be sure to override sanity-checking in the callee, or we'll
// have infinite recursion
TIndexOffU arrs[] = { 0, 0, 0, 0 };
mapLFEx(l, arrs, true);
assert_eq(arrs[c], row);
}
#endif
return c;
}
/**
* Take an offset into the joined text and translate it into the
* reference of the index it falls on, the offset into the reference,
* and the length of the reference. Use a binary search through the
* sorted list of reference fragment ranges t
*/
void Ebwt::joinedToTextOff(TIndexOffU qlen, TIndexOffU off,
TIndexOffU& tidx,
TIndexOffU& textoff,
TIndexOffU& tlen) const
{
TIndexOffU top = 0;
TIndexOffU bot = _nFrag; // 1 greater than largest addressable element
TIndexOffU elt = OFF_MASK;
// Begin binary search
while(true) {
ASSERT_ONLY(TIndexOffU oldelt = elt);
elt = top + ((bot - top) >> 1);
assert_neq(oldelt, elt); // must have made progress
TIndexOffU lower = _rstarts[elt*3];
TIndexOffU upper;
if(elt == _nFrag-1) {
upper = _eh._len;
} else {
upper = _rstarts[((elt+1)*3)];
}
assert_gt(upper, lower);
TIndexOffU fraglen = upper - lower;
if(lower <= off) {
if(upper > off) { // not last element, but it's within
// off is in this range; check if it falls off
if(off + qlen > upper) {
// it falls off; signal no-go and return
tidx = OFF_MASK;
assert_lt(elt, _nFrag-1);
return;
}
tidx = _rstarts[(elt*3)+1];
assert_lt(tidx, this->_nPat);
assert_leq(fraglen, this->_plen[tidx]);
// it doesn't fall off; now calculate textoff.
// Initially it's the number of characters that precede
// the alignment in the fragment
uint32_t fragoff = off - _rstarts[(elt*3)];
if(!this->_fw) {
fragoff = fraglen - fragoff - 1;
fragoff -= (qlen-1);
}
// Add the alignment's offset into the fragment
// ('fragoff') to the fragment's offset within the text
textoff = fragoff + _rstarts[(elt*3)+2];
assert_lt(textoff, this->_plen[tidx]);
break; // done with binary search
} else {
// 'off' belongs somewhere in the region between elt
// and bot
top = elt;
}
} else {
// 'off' belongs somewhere in the region between top and
// elt
bot = elt;
}
// continue with binary search
}
tlen = this->_plen[tidx];
}
/**
* Report a potential match at offset 'off' with pattern length
* 'qlen'. Filter out spurious matches that span texts.
*/
inline bool Ebwt::report(const BTDnaString& query,
BTString* quals,
BTString* name,
const EList<TIndexOffU>& mmui32,
const EList<uint8_t>& refcs,
size_t numMms,
TIndexOffU off,
TIndexOffU top,
TIndexOffU bot,
uint32_t qlen,
int stratum,
uint16_t cost,
uint32_t patid,
uint32_t seed,
const EbwtSearchParams& params) const
{
VMSG_NL("In report");
assert_geq(cost, (uint32_t)(stratum << 14));
assert_lt(off, this->_eh._len);
TIndexOffU tidx;
TIndexOffU textoff;
TIndexOffU tlen;
joinedToTextOff(qlen, off, tidx, textoff, tlen);
if(tidx == OFF_MASK) {
return false;
}
return params.reportHit(
query, // read sequence
quals, // read quality values
name, // read name
_fw, // true = index is forward; false = mirror
mmui32, // mismatch positions
refcs, // reference characters for mms
numMms, // # mismatches
make_pair(tidx, textoff), // position
make_pair<TIndexOffU,TIndexOffU>(0, 0), // (bogus) mate position
true, // (bogus) mate orientation
0, // (bogus) mate length
make_pair(top, bot), // arrows
tlen, // textlen
qlen, // qlen
stratum, // alignment stratum
cost, // cost, including stratum & quality penalty
bot-top-1, // # other hits
patid, // pattern id
seed, // pseudo-random seed
0); // mate (0 = unpaired)
}
#include "row_chaser.h"
/**
* Report a result. Involves walking backwards along the original
* string by way of the LF-mapping until we reach a marked SA row or
* the row corresponding to the 0th suffix. A marked row's offset
* into the original string can be read directly from the this->_offs[]
* array.
*/
inline bool Ebwt::reportChaseOne(const BTDnaString& query,
BTString* quals,
BTString* name,
const EList<TIndexOffU>& mmui32,
const EList<uint8_t>& refcs,
size_t numMms,
TIndexOffU i,
TIndexOffU top,
TIndexOffU bot,
uint32_t qlen,
int stratum,
uint16_t cost,
uint32_t patid,
uint32_t seed,
const EbwtSearchParams& params,
SideLocus *l) const
{
VMSG_NL("In reportChaseOne");
TIndexOffU off;
uint32_t jumps = 0;
ASSERT_ONLY(uint32_t origi = i);
SideLocus myl;
const TIndexOffU offMask = this->_eh._offMask;
const uint32_t offRate = this->_eh._offRate;
const TIndexOffU* offs = this->_offs;
// If the caller didn't give us a pre-calculated (and prefetched)
// locus, then we have to do that now
if(l == NULL) {
l = &myl;
l->initFromRow(i, this->_eh, this->_ebwt);
}
assert(l != NULL);
assert(l->valid());
// Walk along until we reach the next marked row to the left
while(((i & offMask) != i) && i != _zOff) {
// Not a marked row; walk left one more char
TIndexOffU newi = mapLF(*l); // calc next row
assert_neq(newi, i);
i = newi; // update row
l->initFromRow(i, this->_eh, this->_ebwt); // update locus
jumps++;
}
// This is a marked row
if(i == _zOff) {
// Special case: it's the row corresponding to the
// lexicographically smallest suffix, which is implicitly
// marked 0
off = jumps;
VMSG_NL("reportChaseOne found zoff off=" << off << " (jumps=" << jumps << ")");
} else {
// Normal marked row, calculate offset of row i
off = offs[i >> offRate] + jumps;
VMSG_NL("reportChaseOne found off=" << off << " (jumps=" << jumps << ")");
}
#ifndef NDEBUG
{
uint32_t rcoff = RowChaser::toFlatRefOff(this, qlen, origi);
assert_eq(rcoff, off);
}
#endif
return report(query, quals, name, mmui32, refcs, numMms, off, top, bot,
qlen, stratum, cost, patid, seed, params);
}
/**
* Transform this Ebwt into the original string in linear time by using
* the LF mapping to walk backwards starting at the row correpsonding
* to the end of the string. The result is written to s. The Ebwt
* must be in memory.
*/
void Ebwt::restore(BTRefString& s) const {
assert(isInMemory());
s.resize(this->_eh._len);
TIndexOffU jumps = 0;
TIndexOffU i = this->_eh._len; // should point to final SA elt (starting with '$')
SideLocus l(i, this->_eh, this->_ebwt);
while(i != _zOff) {
assert_lt(jumps, this->_eh._len);
//if(_verbose) cout << "restore: i: " << i << endl;
// Not a marked row; go back a char in the original string
TIndexOffU newi = mapLF(l);
assert_neq(newi, i);
s[this->_eh._len - jumps - 1] = rowL(l);
i = newi;
l.initFromRow(i, this->_eh, this->_ebwt);
jumps++;
}
assert_eq(jumps, this->_eh._len);
}
/**
* Check that this Ebwt, when restored via restore(), matches up with
* the given array of reference sequences. For sanity checking.
*/
void Ebwt::checkOrigs(const EList<BTRefString >& os, bool mirror) const
{
BTRefString rest;
restore(rest);
TIndexOffU restOff = 0;
size_t i = 0, j = 0;
if(mirror) {
// TODO: FIXME
return;
}
while(i < os.size()) {
size_t olen = os[i].length();
for(; j < olen; j++) {
size_t joff = j;
if(mirror) joff = olen - j - 1;
if((int)os[i][joff] == 4) {
// Skip over Ns
if(!mirror) {
while(j < olen && (int)os[i][j] == 4) j++;
} else {
while(j < olen && (int)os[i][olen-j-1] == 4) j++;
}
j--;
continue;
}
assert_eq(os[i][joff], rest[restOff]);
restOff++;
}
if(j == os[i].length()) {
// Moved to next sequence
i++;
j = 0;
} else {
// Just jumped over a gap
}
}
}
///////////////////////////////////////////////////////////////////////
//
// Functions for reading and writing Ebwts
//
///////////////////////////////////////////////////////////////////////
/**
* Read an Ebwt from file with given filename.
*/
void Ebwt::readIntoMemory(
int needEntireRev,
bool justHeader,
EbwtParams *params,
bool mmSweep,
bool loadNames,
bool startVerbose)
{
bool switchEndian; // dummy; caller doesn't care
#ifdef BOWTIE_MM
char *mmFile[] = { NULL, NULL };
#endif
if(_in1Str.length() > 0) {
if(_verbose || startVerbose) {
cerr << " About to open input files: ";
logTime(cerr);
}
// Initialize our primary and secondary input-stream fields
if(_in1 != NULL) fclose(_in1);
if(_verbose || startVerbose) cerr << "Opening \"" << _in1Str << "\"" << endl;
if((_in1 = fopen(_in1Str.c_str(), "rb")) == NULL) {
cerr << "Could not open index file " << _in1Str << endl;
}
if(_in2 != NULL) fclose(_in2);
if(_verbose || startVerbose) cerr << "Opening \"" << _in2Str << "\"" << endl;
if((_in2 = fopen(_in2Str.c_str(), "rb")) == NULL) {
cerr << "Could not open index file " << _in2Str << endl;
}
if(_verbose || startVerbose) {
cerr << " Finished opening input files: ";
logTime(cerr);
}
#ifdef BOWTIE_MM
if(_useMm /*&& !justHeader*/) {
const char *names[] = {_in1Str.c_str(), _in2Str.c_str()};
int fds[] = { fileno(_in1), fileno(_in2) };
for(int i = 0; i < 2; i++) {
if(_verbose || startVerbose) {
cerr << " Memory-mapping input file " << (i+1) << ": ";
logTime(cerr);
}
struct stat sbuf;
if (stat(names[i], &sbuf) == -1) {
perror("stat");
cerr << "Error: Could not stat index file " << names[i] << " prior to memory-mapping" << endl;
throw 1;
}
mmFile[i] = (char*)mmap((void *)0, sbuf.st_size,
PROT_READ, MAP_SHARED, fds[i], 0);
if(mmFile[i] == (void *)(-1)) {
perror("mmap");
cerr << "Error: Could not memory-map the index file " << names[i] << endl;
throw 1;
}
if(mmSweep) {
int sum = 0;
for(off_t j = 0; j < sbuf.st_size; j += 1024) {
sum += (int) mmFile[i][j];
}
if(startVerbose) {
cerr << " Swept the memory-mapped ebwt index file 1; checksum: " << sum << ": ";
logTime(cerr);
}
}
}
mmFile1_ = mmFile[0];
mmFile2_ = mmFile[1];
}
#endif
}
#ifdef BOWTIE_MM
else if(_useMm && !justHeader) {
mmFile[0] = mmFile1_;
mmFile[1] = mmFile2_;
}
if(_useMm && !justHeader) {
assert(mmFile[0] == mmFile1_);
assert(mmFile[1] == mmFile2_);
}
#endif
if(_verbose || startVerbose) {
cerr << " Reading header: ";
logTime(cerr);
}
// Read endianness hints from both streams
uint64_t bytesRead = 0;
switchEndian = false;
uint32_t one = readU<uint32_t>(_in1, switchEndian); // 1st word of primary stream
bytesRead += 4;
#ifndef NDEBUG
assert_eq(one, readU<uint32_t>(_in2, switchEndian)); // should match!
#else
readU<uint32_t>(_in2, switchEndian);
#endif
if(one != 1) {
assert_eq((1u<<24), one);
assert_eq(1, endianSwapU32(one));
switchEndian = true;
}
// Can't switch endianness and use memory-mapped files; in order to
// support this, someone has to modify the file to switch
// endiannesses appropriately, and we can't do this inside Bowtie
// or we might be setting up a race condition with other processes.
if(switchEndian && _useMm) {
cerr << "Error: Can't use memory-mapped files when the index is the opposite endianness" << endl;
throw 1;
}
// Reads header entries one by one from primary stream
TIndexOffU len = readU<TIndexOffU>(_in1, switchEndian);
bytesRead += OFF_SIZE;
int32_t lineRate = readI<int32_t>(_in1, switchEndian);
bytesRead += 4;
int32_t linesPerSide = readI<int32_t>(_in1, switchEndian);
bytesRead += 4;
int32_t offRate = readI<int32_t>(_in1, switchEndian);
bytesRead += 4;
// TODO: add isaRate to the actual file format (right now, the
// user has to tell us whether there's an ISA sample and what the
// sampling rate is.
int32_t isaRate = _overrideIsaRate;
int32_t ftabChars = readI<int32_t>(_in1, switchEndian);
bytesRead += 4;
// chunkRate was deprecated in an earlier version of Bowtie; now
// we use it to hold flags.
int32_t flags = readI<int32_t>(_in1, switchEndian);
bool entireRev = false;
if(flags < 0 && (((-flags) & EBWT_ENTIRE_REV) == 0)) {
if(needEntireRev != -1 && needEntireRev != 0) {
cerr << "Error: This index is not compatible with this version of bowtie. Please use a" << endl
<< "current version of bowtie-build." << endl;
throw 1;
}
} else entireRev = true;
bytesRead += 4;
// Create a new EbwtParams from the entries read from primary stream
EbwtParams *eh;
bool deleteEh = false;
if(params != NULL) {
params->init(len, lineRate, linesPerSide, offRate, isaRate, ftabChars, entireRev, _isBt2Index);
if(_verbose || startVerbose) params->print(cerr);
eh = params;
} else {
eh = new EbwtParams(len, lineRate, linesPerSide, offRate, isaRate, ftabChars, entireRev, _isBt2Index);
deleteEh = true;
}
// Set up overridden suffix-array-sample parameters
TIndexOffU offsLen = eh->_offsLen;
uint64_t offsSz = eh->_offsSz;
TIndexOffU offRateDiff = 0;
TIndexOffU offsLenSampled = offsLen;
if(_overrideOffRate > offRate) {
offRateDiff = _overrideOffRate - offRate;
}
if(offRateDiff > 0) {
offsLenSampled >>= offRateDiff;
if((offsLen & ~(OFF_MASK << offRateDiff)) != 0) {
offsLenSampled++;
}
}
// Set up overridden inverted-suffix-array-sample parameters
TIndexOffU isaLen = eh->_isaLen;
TIndexOffU isaRateDiff = 0;
TIndexOffU isaLenSampled = isaLen;
if(_overrideIsaRate > isaRate) {
isaRateDiff = _overrideIsaRate - isaRate;
}
if(isaRateDiff > 0) {
isaLenSampled >>= isaRateDiff;
if((isaLen & ~(OFF_MASK << isaRateDiff)) != 0) {
isaLenSampled++;
}
}
// Can't override the offrate or isarate and use memory-mapped
// files; ultimately, all processes need to copy the sparser sample
// into their own memory spaces.
if(_useMm && (offRateDiff || isaRateDiff)) {
cerr << "Error: Can't use memory-mapped files when the offrate or isarate is overridden" << endl;
throw 1;
}
// Read nPat from primary stream
this->_nPat = readI<TIndexOffU>(_in1, switchEndian);
bytesRead += OFF_SIZE;
if(this->_plen != NULL && !_useMm) {
// Delete it so that we can re-read it
delete[] this->_plen;
this->_plen = NULL;
}
// Read plen from primary stream
if(_useMm) {
#ifdef BOWTIE_MM
this->_plen = (TIndexOffU*)(mmFile[0] + bytesRead);
bytesRead += this->_nPat*OFF_SIZE;
fseeko(_in1, this->_nPat*OFF_SIZE, SEEK_CUR);
#endif
} else {
try {
if(_verbose || startVerbose) {
cerr << "Reading plen (" << this->_nPat << "): ";
logTime(cerr);
}
this->_plen = new TIndexOffU[this->_nPat];
if(switchEndian) {
for(TIndexOffU i = 0; i < this->_nPat; i++) {
this->_plen[i] = readU<TIndexOffU>(_in1, switchEndian);
}
} else {
size_t r = MM_READ(_in1, (void*)this->_plen, this->_nPat*OFF_SIZE);
if(r != (size_t)(this->_nPat*OFF_SIZE)) {
cerr << "Error reading _plen[] array: " << r << ", " << (this->_nPat*OFF_SIZE) << endl;
throw 1;
}
}
} catch(bad_alloc& e) {
cerr << "Out of memory allocating plen[] in Ebwt::read()"
<< " at " << __FILE__ << ":" << __LINE__ << endl;
throw e;
}
}
bool shmemLeader;
// TODO: I'm not consistent on what "header" means. Here I'm using
// "header" to mean everything that would exist in memory if we
// started to build the Ebwt but stopped short of the build*() step
// (i.e. everything up to and including join()).
if(justHeader) goto done;
this->_nFrag = readU<TIndexOffU>(_in1, switchEndian);
bytesRead += OFF_SIZE;
if(_verbose || startVerbose) {
cerr << "Reading rstarts (" << this->_nFrag*3 << "): ";
logTime(cerr);
}
assert_geq(this->_nFrag, this->_nPat);
if(_useMm) {
#ifdef BOWTIE_MM
this->_rstarts = (TIndexOffU*)(mmFile[0] + bytesRead);
bytesRead += this->_nFrag*OFF_SIZE*3;
fseeko(_in1, this->_nFrag*OFF_SIZE*3, SEEK_CUR);
#endif
} else {
this->_rstarts = new TIndexOffU[this->_nFrag*3];
if(switchEndian) {
for(TIndexOffU i = 0; i < this->_nFrag*3; i += 3) {
// fragment starting position in joined reference
// string, text id, and fragment offset within text
this->_rstarts[i] = readU<TIndexOffU>(_in1, switchEndian);
this->_rstarts[i+1] = readU<TIndexOffU>(_in1, switchEndian);
this->_rstarts[i+2] = readU<TIndexOffU>(_in1, switchEndian);
}
} else {
size_t r = MM_READ(_in1, (void *)this->_rstarts, this->_nFrag*OFF_SIZE*3);
if(r != (size_t)(this->_nFrag*OFF_SIZE*3)) {
cerr << "Error reading _rstarts[] array: " << r << ", " << (this->_nFrag*OFF_SIZE*3) << endl;
throw 1;
}
}
}
if(_useMm) {
#ifdef BOWTIE_MM
this->_ebwt = (uint8_t*)(mmFile[0] + bytesRead);
bytesRead += eh->_ebwtTotLen;
fseeko(_in1, eh->_ebwtTotLen, SEEK_CUR);
#endif
} else {
// Allocate ebwt (big allocation)
if(_verbose || startVerbose) {
cerr << "Reading ebwt (" << eh->_ebwtTotLen << "): ";
logTime(cerr);
}
bool shmemLeader = true;
if(useShmem_) {
shmemLeader = ALLOC_SHARED_U8(
(_in1Str + "[ebwt]"), eh->_ebwtTotLen, &this->_ebwt,
"ebwt[]", (_verbose || startVerbose));
if(_verbose || startVerbose) {
cerr << " shared-mem " << (shmemLeader ? "leader" : "follower") << endl;
}
} else {
try {
this->_ebwt = new uint8_t[eh->_ebwtTotLen];
} catch(bad_alloc& e) {
cerr << "Out of memory allocating the ebwt[] array for the Bowtie index. Please try" << endl
<< "again on a computer with more memory." << endl;
throw 1;
}
}
if(shmemLeader) {
// Read ebwt from primary stream
uint64_t bytesLeft = eh->_ebwtTotLen;
char *pebwt = (char*)this->ebwt();
while (bytesLeft>0){
size_t r = MM_READ(_in1, (void *)pebwt, bytesLeft);
if(MM_IS_IO_ERR(_in1,r,bytesLeft)) {
cerr << "Error reading ebwt array: returned " << r << ", length was " << (eh->_ebwtTotLen) << endl
<< "Your index files may be corrupt; please try re-building or re-downloading." << endl
<< "A complete index consists of 6 files: XYZ.1.ebwt, XYZ.2.ebwt, XYZ.3.ebwt," << endl
<< "XYZ.4.ebwt, XYZ.rev.1.ebwt, and XYZ.rev.2.ebwt. The XYZ.1.ebwt and " << endl
<< "XYZ.rev.1.ebwt files should have the same size, as should the XYZ.2.ebwt and" << endl
<< "XYZ.rev.2.ebwt files." << endl;
throw 1;
}
pebwt += r;
bytesLeft -= r;
}
if(switchEndian) {
uint8_t *side = this->_ebwt;
for(size_t i = 0; i < eh->_numSides; i++) {
TIndexOffU *cums = reinterpret_cast<TIndexOffU*>(side + eh->_sideSz - 2*OFF_SIZE);
cums[0] = endianSwapU(cums[0]);
cums[1] = endianSwapU(cums[1]);
side += this->_eh._sideSz;
}
}
if(useShmem_) NOTIFY_SHARED(this->_ebwt, eh->_ebwtTotLen);
} else {
// Seek past the data and wait until master is finished
fseeko(_in1, eh->_ebwtTotLen, SEEK_CUR);
if(useShmem_) WAIT_SHARED(this->_ebwt, eh->_ebwtTotLen);
}
}
// Read zOff from primary stream
_zOff = readU<TIndexOffU>(_in1, switchEndian);
bytesRead += OFF_SIZE;
assert_lt(_zOff, len);
try {
// Read fchr from primary stream
if(_verbose || startVerbose) cerr << "Reading fchr (5)" << endl;
if(_useMm) {
#ifdef BOWTIE_MM
this->_fchr = (TIndexOffU*)(mmFile[0] + bytesRead);
bytesRead += 5*OFF_SIZE;
fseeko(_in1, 5*OFF_SIZE, SEEK_CUR);
#endif
} else {
this->_fchr = new TIndexOffU[5];
for(int i = 0; i < 5; i++) {
this->_fchr[i] = readU<TIndexOffU>(_in1, switchEndian);
assert_leq(this->_fchr[i], len);
if(i > 0) assert_geq(this->_fchr[i], this->_fchr[i-1]);
}
}
assert_gt(this->_fchr[4], this->_fchr[0]);
// Read ftab from primary stream
if(_verbose || startVerbose) {
cerr << "Reading ftab (" << eh->_ftabLen << "): ";
logTime(cerr);
}
if(_useMm) {
#ifdef BOWTIE_MM
this->_ftab = (TIndexOffU*)(mmFile[0] + bytesRead);
bytesRead += eh->_ftabLen*OFF_SIZE;
fseeko(_in1, eh->_ftabLen*OFF_SIZE, SEEK_CUR);
#endif
} else {
this->_ftab = new TIndexOffU[eh->_ftabLen];
if(switchEndian) {
for(TIndexOffU i = 0; i < eh->_ftabLen; i++)
this->_ftab[i] = readU<TIndexOffU>(_in1, switchEndian);
} else {
size_t r = MM_READ(_in1, (void *)this->_ftab, eh->_ftabLen*OFF_SIZE);
if(r != (size_t)(eh->_ftabLen*OFF_SIZE)) {
cerr << "Error reading _ftab[] array: " << r << ", " << (eh->_ftabLen*OFF_SIZE) << endl;
throw 1;
}
}
}
// Read etab from primary stream
if(_verbose || startVerbose) {
cerr << "Reading eftab (" << eh->_eftabLen << "): ";
logTime(cerr);
}
if(_useMm) {
#ifdef BOWTIE_MM
this->_eftab = (TIndexOffU*)(mmFile[0] + bytesRead);
bytesRead += eh->_eftabLen*OFF_SIZE;
fseeko(_in1, eh->_eftabLen*OFF_SIZE, SEEK_CUR);
#endif
} else {
this->_eftab = new TIndexOffU[eh->_eftabLen];
if(switchEndian) {
for(TIndexOffU i = 0; i < eh->_eftabLen; i++)
this->_eftab[i] = readU<TIndexOffU>(_in1, switchEndian);
} else {
size_t r = MM_READ(_in1, (void *)this->_eftab, eh->_eftabLen*OFF_SIZE);
if(r != (size_t)(eh->_eftabLen*OFF_SIZE)) {
cerr << "Error reading _eftab[] array: " << r << ", " << (eh->_eftabLen*OFF_SIZE) << endl;
throw 1;
}
}
}
for(TIndexOffU i = 0; i < eh->_eftabLen; i++) {
if(i > 0 && this->_eftab[i] > 0) {
assert_geq(this->_eftab[i], this->_eftab[i-1]);
} else if(i > 0 && this->_eftab[i-1] == 0) {
assert_eq(0, this->_eftab[i]);
}
}
} catch(bad_alloc& e) {
cerr << "Out of memory allocating fchr[], ftab[] or eftab[] arrays for the Bowtie index." << endl
<< "Please try again on a computer with more memory." << endl;
throw 1;
}
// Read reference sequence names from primary index file (or not,
// if --refidx is specified)
if(loadNames) {
while(true) {
char c = '\0';
if(MM_READ(_in1, (void *)(&c), (size_t)1) != (size_t)1) break;
bytesRead++;
if(c == '\0') break;
else if(c == '\n') {
this->_refnames.push_back("");
} else {
if(this->_refnames.size() == 0) {
this->_refnames.push_back("");
}
this->_refnames.back().push_back(c);
}
}
}
bytesRead = 4; // reset for secondary index file (already read 1-sentinel)
shmemLeader = true;
if(_verbose || startVerbose) {
cerr << "Reading offs (" << offsLenSampled << " 32-bit words): ";
logTime(cerr);
}
if(!_useMm) {
if(!useShmem_) {
// Allocate offs_
try {
this->_offs = new TIndexOffU[offsLenSampled];
} catch(bad_alloc& e) {
cerr << "Out of memory allocating the offs[] array for the Bowtie index." << endl
<< "Please try again on a computer with more memory." << endl;
throw 1;
}
} else {
shmemLeader = ALLOC_SHARED_U(
(_in2Str + "[offs]"), offsLenSampled*OFF_SIZE, &this->_offs,
"offs", (_verbose || startVerbose));
}
}
if(_overrideOffRate < 32) {
if(shmemLeader) {
// Allocate offs (big allocation)
if(switchEndian || offRateDiff > 0) {
assert(!_useMm);
const TIndexOffU blockMaxSz = (2 * 1024 * 1024); // 2 MB block size
const TIndexOffU blockMaxSzU = (blockMaxSz >> (OFF_SIZE/4 +1)); // # U32s per block
char *buf = new char[blockMaxSz];
for(TIndexOffU i = 0; i < offsLen; i += blockMaxSzU) {
TIndexOffU block = min<TIndexOffU>(blockMaxSzU, offsLen - i);
size_t r = MM_READ(_in2, (void *)buf, block << (OFF_SIZE/4 + 1));
if(r != (size_t)(block << (OFF_SIZE/4 + 1))) {
cerr << "Error reading block of offs array: " << r << ", " << (block << (OFF_SIZE/4 + 1)) << endl
<< "Your index files may be corrupt; please try re-building or re-downloading." << endl
<< "A complete index consists of 6 files: XYZ.1.ebwt, XYZ.2.ebwt, XYZ.3.ebwt," << endl
<< "XYZ.4.ebwt, XYZ.rev.1.ebwt, and XYZ.rev.2.ebwt. The XYZ.1.ebwt and " << endl
<< "XYZ.rev.1.ebwt files should have the same size, as should the XYZ.2.ebwt and" << endl
<< "XYZ.rev.2.ebwt files." << endl;
throw 1;
}
TIndexOffU idx = i >> offRateDiff;
for(TIndexOffU j = 0; j < block; j += (1 << offRateDiff)) {
assert_lt(idx, offsLenSampled);
this->_offs[idx] = ((TIndexOffU*)buf)[j];
if(switchEndian) {
this->_offs[idx] = endianSwapU(this->_offs[idx]);
}
idx++;
}
}
delete[] buf;
} else {
if(_useMm) {
#ifdef BOWTIE_MM
this->_offs = (TIndexOffU*)(mmFile[1] + bytesRead);
bytesRead += offsSz;
// Argument to lseek can be 64 bits if compiled with
// _FILE_OFFSET_BITS
fseeko(_in2, offsSz, SEEK_CUR);
#endif
} else {
// If any of the high two bits are set
// Workaround for small-index mode where MM_READ may
// not be able to handle read amounts greater than 2^32
// bytes.
uint64_t bytesLeft = offsSz;
char *offs = (char *)this->offs();
while(bytesLeft > 0) {
size_t r = MM_READ(_in2, (void*)offs, bytesLeft);
if(MM_IS_IO_ERR(_in2,r,bytesLeft)) {
cerr << "Error reading block of _offs[] array: "
<< r << ", " << bytesLeft << gLastIOErrMsg << endl;
throw 1;
}
offs += r;
bytesLeft -= r;
}
}
}
{
ASSERT_ONLY(Bitset offsSeen(len+1));
for(TIndexOffU i = 0; i < offsLenSampled; i++) {
assert(!offsSeen.test(this->_offs[i]));
ASSERT_ONLY(offsSeen.set(this->_offs[i]));
assert_leq(this->_offs[i], len);
}
}
if(useShmem_) NOTIFY_SHARED(this->_offs, offsLenSampled*OFF_SIZE);
} else {
// Not the shmem leader
fseeko(_in2, offsLenSampled*OFF_SIZE, SEEK_CUR);
if(useShmem_) WAIT_SHARED(this->_offs, offsLenSampled*OFF_SIZE);
}
}
// Allocate _isa[] (big allocation)
if(_verbose || startVerbose) {
cerr << "Reading isa (" << isaLenSampled << "): ";
logTime(cerr);
}
if(!_useMm) {
try {
this->_isa = new TIndexOffU[isaLenSampled];
} catch(bad_alloc& e) {
cerr << "Out of memory allocating the isa[] array for the Bowtie index." << endl
<< "Please try again on a computer with more memory." << endl;
throw 1;
}
}
// Read _isa[]
if(switchEndian || isaRateDiff > 0) {
assert(!_useMm);
for(TIndexOffU i = 0; i < isaLen; i++) {
if((i & ~(OFF_MASK << isaRateDiff)) != 0) {
char tmp[OFF_SIZE];
size_t r = MM_READ(_in2, (void *)tmp, OFF_SIZE);
if(r != (size_t)OFF_SIZE) {
cerr << "Error reading a word of the _isa[] array: " << r << ", 4" << endl;
throw 1;
}
} else {
TIndexOffU idx = i >> isaRateDiff;
assert_lt(idx, isaLenSampled);
this->_isa[idx] = readU<TIndexOffU>(_in2, switchEndian);
}
}
} else {
if(_useMm) {
#ifdef BOWTIE_MM
this->_isa = (TIndexOffU*)(mmFile[1] + bytesRead);
bytesRead += (isaLen << 2);
fseeko(_in2, (isaLen << 2), SEEK_CUR);
#endif
} else {
size_t r = MM_READ(_in2, (void *)this->_isa, isaLen*OFF_SIZE);
if(r != (size_t)(isaLen*OFF_SIZE)) {
cerr << "Error reading _isa[] array: " << r << ", " << (isaLen*OFF_SIZE) << endl;
throw 1;
}
}
}
{
ASSERT_ONLY(Bitset isasSeen(len+1));
for(TIndexOffU i = 0; i < isaLenSampled; i++) {
assert(!isasSeen.test(this->_isa[i]));
ASSERT_ONLY(isasSeen.set(this->_isa[i]));
assert_leq(this->_isa[i], len);
}
}
this->postReadInit(*eh); // Initialize fields of Ebwt not read from file
if(_verbose || startVerbose) print(cerr, *eh);
// The fact that _ebwt and friends actually point to something
// (other than NULL) now signals to other member functions that the
// Ebwt is loaded into memory.
done: // Exit hatch for both justHeader and !justHeader
// Be kind
if(deleteEh) delete eh;
if (_in1 != NULL) rewind(_in1);
if (_in2 != NULL) rewind(_in2);
}
/**
* Read reference names from an input stream 'in' for an Ebwt primary
* file and store them in 'refnames'.
* TODO: revisit this function
*/
static inline void
readEbwtRefnames(FILE* fin, EList<string>& refnames) {
// _in1 must already be open with the get cursor at the
// beginning and no error flags set.
assert(fin != NULL);
assert_eq(ftello(fin), 0);
// Read endianness hints from both streams
bool switchEndian = false;
uint32_t one = readU<uint32_t>(fin, switchEndian); // 1st word of primary stream
if(one != 1) {
assert_eq((1u<<24), one);
switchEndian = true;
}
// Reads header entries one by one from primary stream
TIndexOffU len = readU<TIndexOffU>(fin, switchEndian);
int32_t lineRate = readI<int32_t>(fin, switchEndian);
int32_t linesPerSide = readI<int32_t>(fin, switchEndian);
int32_t offRate = readI<int32_t>(fin, switchEndian);
int32_t ftabChars = readI<int32_t>(fin, switchEndian);
// BTL: chunkRate is now deprecated
int32_t flags = readI<int32_t>(fin, switchEndian);
bool entireReverse = false;
if(flags < 0) {
entireReverse = (((-flags) & EBWT_ENTIRE_REV) != 0);
}
// Create a new EbwtParams from the entries read from primary stream
bool isBt2Index = false;
if (gEbwt_ext == "bt2" || gEbwt_ext == "bt2") {
isBt2Index = true;
}
EbwtParams eh(len, lineRate, linesPerSide, offRate, -1, ftabChars, entireReverse, isBt2Index);
TIndexOffU nPat = readI<TIndexOffU>(fin, switchEndian); // nPat
fseeko(fin, nPat*OFF_SIZE, SEEK_CUR);
// Skip rstarts
TIndexOffU nFrag = readU<TIndexOffU>(fin, switchEndian);
fseeko(fin, nFrag*OFF_SIZE*3, SEEK_CUR);
// Skip ebwt
fseeko(fin, eh._ebwtTotLen, SEEK_CUR);
// Skip zOff from primary stream
readU<TIndexOffU>(fin, switchEndian);
// Skip fchr
fseeko(fin, 5 * OFF_SIZE, SEEK_CUR);
// Skip ftab
fseeko(fin, eh._ftabLen*OFF_SIZE, SEEK_CUR);
// Skip eftab
fseeko(fin, eh._eftabLen*OFF_SIZE, SEEK_CUR);
// Read reference sequence names from primary index file
while(true) {
char c = '\0';
int read_value = 0;
read_value = fgetc(fin);
if(read_value == EOF) break;
c = read_value;
if(c == '\0') break;
else if(c == '\n') {
refnames.push_back("");
} else {
if(refnames.size() == 0) {
refnames.push_back("");
}
refnames.back().push_back(c);
}
}
if(refnames.back().empty()) {
refnames.pop_back();
}
// Be kind
fseeko(fin, 0, SEEK_SET);
assert(ferror(fin) == 0);
}
/**
* Read reference names from the index with basename 'in' and store
* them in 'refnames'.
*/
static inline void
readEbwtRefnames(const string& instr, EList<string>& refnames) {
FILE* fin;
// Initialize our primary and secondary input-stream fields
fin = fopen((instr + ".1." + gEbwt_ext).c_str(),"rb");
if(fin == NULL) {
throw EbwtFileOpenException("Cannot open file " + instr);
}
assert_eq(ftello(fin), 0);
readEbwtRefnames(fin, refnames);
fclose(fin);
}
/**
* Read just enough of the Ebwt's header to get its flags
*/
static inline int32_t readFlags(const string& instr) {
ifstream in;
// Initialize our primary and secondary input-stream fields
in.open((instr + ".1." + gEbwt_ext).c_str(), ios_base::in | ios::binary);
if(!in.is_open()) {
throw EbwtFileOpenException("Cannot open file " + instr);
}
assert(in.is_open());
assert(in.good());
bool switchEndian = false;
uint32_t one = readU<uint32_t>(in, switchEndian); // 1st word of primary stream
if(one != 1) {
assert_eq((1u<<24), one);
assert_eq(1, endianSwapU32(one));
switchEndian = true;
}
readU<TIndexOffU>(in, switchEndian);
readI<int32_t>(in, switchEndian);
readI<int32_t>(in, switchEndian);
readI<int32_t>(in, switchEndian);
readI<int32_t>(in, switchEndian);
int32_t flags = readI<int32_t>(in, switchEndian);
return flags;
}
/**
* Read just enough of the Ebwt's header to determine whether it's
* entirely reversed.
*/
static inline bool
readEntireReverse(const string& instr) {
int32_t flags = readFlags(instr);
if(flags < 0 && (((-flags) & EBWT_ENTIRE_REV) != 0)) {
return true;
} else {
return false;
}
}
/**
* Write an extended Burrows-Wheeler transform to a pair of output
* streams.
*
* @param out1 output stream to primary file
* @param out2 output stream to secondary file
* @param be write in big endian?
*/
void Ebwt::writeFromMemory(bool justHeader,
ostream& out1,
ostream& out2) const
{
const EbwtParams& eh = this->_eh;
assert(eh.repOk());
uint32_t be = this->toBe();
assert(out1.good());
assert(out2.good());
// When building an Ebwt, these header parameters are known
// "up-front", i.e., they can be written to disk immediately,
// before we join() or buildToDisk()
writeI<int32_t>(out1, 1, be); // endian hint for priamry stream
writeI<int32_t>(out2, 1, be); // endian hint for secondary stream
writeU<TIndexOffU>(out1, eh._len, be); // length of string (and bwt and suffix array)
writeI<int32_t>(out1, eh._lineRate, be); // 2^lineRate = size in bytes of 1 line
writeI<int32_t>(out1, eh._linesPerSide, be); // not used
writeI<int32_t>(out1, eh._offRate, be); // every 2^offRate chars is "marked"
writeI<int32_t>(out1, eh._ftabChars, be); // number of 2-bit chars used to address ftab
int32_t flags = 1;
if(eh._entireReverse) flags |= EBWT_ENTIRE_REV;
writeI<int32_t>(out1, -flags, be); // BTL: chunkRate is now deprecated
if(!justHeader) {
assert(isInMemory());
// These Ebwt parameters are known after the inputs strings have
// been joined() but before they have been built(). These can
// written to the disk next and then discarded from memory.
writeU<TIndexOffU>(out1, this->_nPat, be);
for(TIndexOffU i = 0; i < this->_nPat; i++)
writeU<TIndexOffU>(out1, this->_plen[i], be);
assert_geq(this->_nFrag, this->_nPat);
writeU<TIndexOffU>(out1, this->_nFrag, be);
for(TIndexOffU i = 0; i < this->_nFrag*3; i++)
writeU<TIndexOffU>(out1, this->_rstarts[i], be);
// These Ebwt parameters are discovered only as the Ebwt is being
// built (in buildToDisk()). Of these, only 'offs' and 'ebwt' are
// terribly large. 'ebwt' is written to the primary file and then
// discarded from memory as it is built; 'offs' is similarly
// written to the secondary file and discarded.
out1.write((const char *)this->ebwt(), eh._ebwtTotLen);
writeU<TIndexOffU>(out1, this->zOff(), be);
TIndexOffU offsLen = eh._offsLen;
for(TIndexOffU i = 0; i < offsLen; i++)
writeU<TIndexOffU>(out2, this->_offs[i], be);
uint32_t isaLen = eh._isaLen;
for(TIndexOffU i = 0; i < isaLen; i++)
writeU<TIndexOffU>(out2, this->_isa[i], be);
// 'fchr', 'ftab' and 'eftab' are not fully determined until the
// loop is finished, so they are written to the primary file after
// all of 'ebwt' has already been written and only then discarded
// from memory.
for(int i = 0; i < 5; i++)
writeU<TIndexOffU>(out1, this->_fchr[i], be);
for(TIndexOffU i = 0; i < eh._ftabLen; i++)
writeU<TIndexOffU>(out1, this->ftab()[i], be);
for(TIndexOffU i = 0; i < eh._eftabLen; i++)
writeU<TIndexOffU>(out1, this->eftab()[i], be);
}
}
/**
* Given a pair of strings representing output filenames, and assuming
* this Ebwt object is currently in memory, write out this Ebwt to the
* specified files.
*
* If sanity-checking is enabled, then once the streams have been
* fully written and closed, we reopen them and read them into a
* (hopefully) exact copy of this Ebwt. We then assert that the
* current Ebwt and the copy match in all of their fields.
*/
void Ebwt::writeFromMemory(bool justHeader,
const string& out1,
const string& out2) const
{
assert(isInMemory());
assert(this->_eh.repOk());
ofstream fout1(out1.c_str(), ios::binary);
ofstream fout2(out2.c_str(), ios::binary);
writeFromMemory(justHeader, fout1, fout2);
fout1.close();
fout2.close();
// Read the file back in and assert that all components match
// TODO: revisit this block
if(_sanity) {
#if 0
if(_verbose)
cout << "Re-reading \"" << out1 << "\"/\"" << out2 << "\" for sanity check" << endl;
Ebwt copy(out1, out2, _verbose, _sanity);
assert(!isInMemory());
copy.loadIntoMemory(eh._color ? 1 : 0, -1, false, false);
assert(isInMemory());
assert_eq(eh._lineRate, copy.eh()._lineRate);
assert_eq(eh._linesPerSide, copy.eh()._linesPerSide);
assert_eq(eh._offRate, copy.eh()._offRate);
assert_eq(eh._isaRate, copy.eh()._isaRate);
assert_eq(eh._ftabChars, copy.eh()._ftabChars);
assert_eq(eh._len, copy.eh()._len);
assert_eq(_zOff, copy.zOff());
assert_eq(_zEbwtBpOff, copy.zEbwtBpOff());
assert_eq(_zEbwtByteOff, copy.zEbwtByteOff());
assert_eq(_nPat, copy.nPat());
for(TIndexOffU i = 0; i < _nPat; i++)
assert_eq(this->_plen[i], copy.plen()[i]);
assert_eq(this->_nFrag, copy.nFrag());
for(TIndexOffU i = 0; i < this->_nFrag*3; i++) {
assert_eq(this->_rstarts[i], copy.rstarts()[i]);
}
for(uint32_t i = 0; i < 5; i++)
assert_eq(this->_fchr[i], copy.fchr()[i]);
for(TIndexOffU i = 0; i < eh._ftabLen; i++)
assert_eq(this->ftab()[i], copy.ftab()[i]);
for(TIndexOffU i = 0; i < eh._eftabLen; i++)
assert_eq(this->eftab()[i], copy.eftab()[i]);
for(TIndexOffU i = 0; i < eh._offsLen; i++)
assert_eq(this->_offs[i], copy.offs()[i]);
for(TIndexOffU i = 0; i < eh._isaLen; i++)
assert_eq(this->_isa[i], copy.isa()[i]);
for(TIndexOffU i = 0; i < eh._ebwtTotLen; i++)
assert_eq(this->ebwt()[i], copy.ebwt()[i]);
//copy.sanityCheckAll();
if(_verbose)
cout << "Read-in check passed for \"" << out1 << "\"/\"" << out2 << "\"" << endl;
#endif
}
}
///////////////////////////////////////////////////////////////////////
//
// Functions for building Ebwts
//
///////////////////////////////////////////////////////////////////////
/**
* Join several text strings together in a way that's compatible with
* the text-chunking scheme dictated by chunkRate parameter.
*
* The non-static member Ebwt::join additionally builds auxilliary
* arrays that maintain a mapping between chunks in the joined string
* and the original text strings.
*/
template <typename TStr>
TStr Ebwt::join(EList<TStr>& l, uint32_t seed) {
RandomSource rand; // reproducible given same seed
rand.init(seed);
TStr ret;
size_t guessLen = 0;
for(size_t i = 0; i < l.size(); i++) {
guessLen += l[i].length();
}
ret.reserve(guessLen);
for(size_t i = 0; i < l.size(); i++) {
TStr& s = l[i];
assert_gt(s.length(), 0);
ret.push_back(s);
}
return ret;
}
/**
* Join several text strings together in a way that's compatible with
* the text-chunking scheme dictated by chunkRate parameter.
*
* The non-static member Ebwt::join additionally builds auxilliary
* arrays that maintain a mapping between chunks in the joined string
* and the original text strings.
*/
template<typename TStr>
TStr Ebwt::join(EList<FileBuf*>& l,
EList<RefRecord>& szs,
TIndexOffU sztot,
const RefReadInParams& refparams,
uint32_t seed)
{
RandomSource rand; // reproducible given same seed
rand.init(seed);
RefReadInParams rpcp = refparams;
TStr ret;
size_t guessLen = sztot;
ret.resize(guessLen);
TIndexOffU dstoff = 0;
ASSERT_ONLY(size_t szsi = 0);
for(size_t i = 0; i < l.size(); i++) {
// For each sequence we can pull out of istream l[i]...
assert(!l[i]->eof());
bool first = true;
while(!l[i]->eof()) {
RefRecord rec = fastaRefReadAppend(*l[i], first, ret, dstoff, rpcp);
#ifndef ACCOUNT_FOR_ALL_GAP_REFS
if(rec.first && rec.len == 0) rec.first = false;
#endif
first = false;
size_t bases = rec.len;
assert_eq(rec.off, szs[szsi].off);
assert_eq(rec.len, szs[szsi].len);
assert_eq(rec.first, szs[szsi].first);
ASSERT_ONLY(szsi++);
if(bases == 0) continue;
}
}
return ret;
}
/**
* Join several text strings together according to the text-chunking
* scheme specified in the EbwtParams. Ebwt fields calculated in this
* function are written directly to disk.
*
* It is assumed, but not required, that the header values have already
* been written to 'out1' before this function is called.
*
* The static member Ebwt::join just returns a joined version of a
* list of strings without building any of the auxiliary arrays.
* Because the pseudo-random number generator is the same, we expect
* this function and the static function to give the same result given
* the same seed.
*/
template<typename TStr>
void Ebwt::joinToDisk(
EList<FileBuf*>& l,
EList<RefRecord>& szs,
EList<uint32_t>& plens,
TIndexOffU sztot,
const RefReadInParams& refparams,
TStr& ret,
ostream& out1,
ostream& out2,
uint32_t seed)
{
RandomSource rand; // reproducible given same seed
rand.init(seed);
RefReadInParams rpcp = refparams;
assert_gt(szs.size(), 0);
assert_gt(l.size(), 0);
assert_gt(sztot, 0);
// Not every fragment represents a distinct sequence - many
// fragments may correspond to a single sequence. Count the
// number of sequences here by counting the number of "first"
// fragments.
this->_nPat = 0;
this->_nFrag = 0;
#ifdef ACCOUNT_FOR_ALL_GAP_REFS
int nGapFrag = 0;
#endif
for(size_t i = 0; i < szs.size(); i++) {
if(szs[i].len > 0) this->_nFrag++;
#ifdef ACCOUNT_FOR_ALL_GAP_REFS
if(szs[i].len == 0 && szs[i].off > 0) nGapFrag++;
if(szs[i].first && szs[i].len > 0) this->_nPat++;
#else
// For all records where len=0 and first=1, set first=0
assert(szs[i].len > 0 || !szs[i].first);
if(szs[i].first) this->_nPat++;
#endif
}
assert_gt(this->_nPat, 0);
assert_geq(this->_nFrag, this->_nPat);
this->_rstarts = NULL;
writeU<TIndexOffU>(out1, this->_nPat, this->toBe());
assert_eq(plens.size(), this->_nPat);
// Allocate plen[]
try {
this->_plen = new TIndexOffU[this->_nPat];
} catch(bad_alloc& e) {
cerr << "Out of memory allocating plen[] in Ebwt::join()"
<< " at " << __FILE__ << ":" << __LINE__ << endl;
throw e;
}
// For each pattern, set plen
for(TIndexOffU i = 0; i < plens.size(); i++) {
this->_plen[i] = plens[i];
writeU<TIndexOffU>(out1, this->_plen[i], this->toBe());
}
// Write the number of fragments
writeU<TIndexOffU>(out1, this->_nFrag, this->toBe());
TIndexOffU seqsRead = 0;
ASSERT_ONLY(TIndexOffU szsi = 0);
ASSERT_ONLY(TIndexOffU entsWritten = 0);
TIndexOffU dstoff = 0;
// For each filebuf
for(unsigned int i = 0; i < l.size(); i++) {
assert(!l[i]->eof());
bool first = true;
TIndexOffU patoff = 0;
// For each *fragment* (not necessary an entire sequence) we
// can pull out of istream l[i]...
while(!l[i]->eof()) {
string name;
// Push a new name onto our vector
_refnames.push_back("");
//uint32_t oldRetLen = length(ret);
RefRecord rec = fastaRefReadAppend(*l[i], first, ret, dstoff, rpcp, &_refnames.back());
#ifndef ACCOUNT_FOR_ALL_GAP_REFS
if(rec.first && rec.len == 0) rec.first = false;
#endif
first = false;
if(rec.first) {
if(_refnames.back().length() == 0) {
// If name was empty, replace with an index
ostringstream stm;
stm << (_refnames.size()-1);
_refnames.back() = stm.str();
}
} else {
// This record didn't actually start a new sequence so
// no need to add a name
//assert_eq(0, _refnames.back().length());
_refnames.pop_back();
}
assert_lt(szsi, szs.size());
assert(szs[szsi].first == 0 || szs[szsi].first == 1);
assert_eq(rec.off, szs[szsi].off);
assert_eq(rec.len, szs[szsi].len);
// szs[szsi].first == 2 sometimes?!?! g++ is unable to do
// the following correctly, regardless of how I write it
//assert((rec.first == 0) == (szs[szsi].first == 0));
assert(rec.first || rec.off > 0);
ASSERT_ONLY(szsi++);
#ifdef ACCOUNT_FOR_ALL_GAP_REFS
if(rec.len == 0) continue;
if(rec.first && rec.len > 0) seqsRead++;
assert_leq(rec.len, this->_plen[seqsRead-1]);
#else
if(rec.first) seqsRead++;
if(rec.len == 0) continue;
assert_leq(rec.len, this->_plen[seqsRead-1]);
#endif
// Reset the patoff if this is the first fragment
if(rec.first) patoff = 0;
patoff += rec.off; // add fragment's offset from end of last frag.
// Adjust rpcps
//uint32_t seq = seqsRead-1;
ASSERT_ONLY(entsWritten++);
// This is where rstarts elements are written to the output stream
//writeU32(out1, oldRetLen, this->toBe()); // offset from beginning of joined string
//writeU32(out1, seq, this->toBe()); // sequence id
//writeU32(out1, patoff, this->toBe()); // offset into sequence
patoff += rec.len;
}
assert_gt(szsi, 0);
l[i]->reset();
assert(!l[i]->eof());
#ifndef NDEBUG
int c = l[i]->get();
assert_eq('>', c);
assert(!l[i]->eof());
l[i]->reset();
assert(!l[i]->eof());
#endif
}
assert_eq(entsWritten, this->_nFrag);
}
/**
* Build an Ebwt from a string 's' and its suffix array 'sa' (which
* might actually be a suffix array *builder* that builds blocks of the
* array on demand). The bulk of the Ebwt, i.e. the ebwt and offs
* arrays, is written directly to disk. This is by design: keeping
* those arrays in memory needlessly increases the footprint of the
* building process. Instead, we prefer to build the Ebwt directly
* "to disk" and then read it back into memory later as necessary.
*
* It is assumed that the header values and join-related values (nPat,
* plen) have already been written to 'out1' before this function
* is called. When this function is finished, it will have
* additionally written ebwt, zOff, fchr, ftab and eftab to the primary
* file and offs to the secondary file.
*
* Assume DNA/RNA/any alphabet with 4 or fewer elements.
* Assume occ array entries are 32 bits each.
*
* @param sa the suffix array to convert to a Ebwt
* @param buildISA whether to output an ISA sample into out2 after
* the SA sample
* @param s the original string
* @param out
*/
template<typename TStr>
void Ebwt::buildToDisk(InorderBlockwiseSA<TStr>& sa,
const TStr& s,
ostream& out1,
ostream& out2)
{
const EbwtParams& eh = this->_eh;
assert(eh.repOk());
assert_eq(s.length()+1, sa.size());
assert_eq(s.length(), eh._len);
assert_gt(eh._lineRate, 3);
assert(sa.suffixItrIsReset());
// assert_leq((int)ValueSize<Dna>::VALUE, 4);
TIndexOffU len = eh._len;
TIndexOffU ftabLen = eh._ftabLen;
TIndexOffU sideSz = eh._sideSz;
TIndexOffU ebwtTotSz = eh._ebwtTotSz;
TIndexOffU fchr[] = {0, 0, 0, 0, 0};
TIndexOffU* ftab = NULL;
TIndexOffU zOff = OFF_MASK;
// Save # of occurrences of each character as we walk along the bwt
TIndexOffU occ[4] = {0, 0, 0, 0};
// Save 'G' and 'T' occurrences between backward and forward buckets
TIndexOffU occSave[2] = {0, 0};
// Record rows that should "absorb" adjacent rows in the ftab.
// The absorbed rows represent suffixes shorter than the ftabChars
// cutoff.
uint8_t absorbCnt = 0;
uint8_t *absorbFtab;
try {
VMSG_NL("Allocating ftab, absorbFtab");
ftab = new TIndexOffU[ftabLen];
memset(ftab, 0, OFF_SIZE * ftabLen);
absorbFtab = new uint8_t[ftabLen];
memset(absorbFtab, 0, ftabLen);
} catch(bad_alloc &e) {
cerr << "Out of memory allocating ftab[] or absorbFtab[] "
<< "in Ebwt::buildToDisk() at " << __FILE__ << ":"
<< __LINE__ << endl;
throw e;
}
assert(ftab != NULL);
assert(absorbFtab != NULL);
// Allocate the side buffer; holds a single side as its being
// constructed and then written to disk. Reused across all sides.
#ifdef SIXTY4_FORMAT
uint64_t *ebwtSide = NULL;
#else
uint8_t *ebwtSide = NULL;
#endif
try {
#ifdef SIXTY4_FORMAT
ebwtSide = new uint64_t[sideSz >> 3];
#else
ebwtSide = new uint8_t[sideSz];
#endif
} catch(bad_alloc &e) {
cerr << "Out of memory allocating ebwtSide[] in "
<< "Ebwt::buildToDisk() at " << __FILE__ << ":"
<< __LINE__ << endl;
throw e;
}
assert(ebwtSide != NULL);
// Allocate a buffer to hold the ISA sample, which we accumulate in
// the loop and then output at the end. We can't write output the
// ISA right away because the order in which we calculate its
// elements is based on the suffix array, which we only see bit by
// bit
uint32_t *isaSample = NULL;
if(eh._isaRate >= 0) {
try {
isaSample = new uint32_t[eh._isaLen];
} catch(bad_alloc &e) {
cerr << "Out of memory allocating isaSample[] in "
<< "Ebwt::buildToDisk() at " << __FILE__ << ":"
<< __LINE__ << endl;
throw e;
}
assert(isaSample != NULL);
}
// Points to the base offset within ebwt for the side currently
// being written
TIndexOffU side = 0;
// Points to a byte offset from 'side' within ebwt[] where next
// char should be written
#ifdef SIXTY4_FORMAT
TIndexOff sideCur = (eh._sideBwtSz >> 3) - 1;
#else
TIndexOff sideCur = eh._sideBwtSz - 1;
#endif
// Whether we're assembling a forward or a reverse bucket
bool fw = false;
// Did we just finish writing a forward bucket? (Must be true when
// we exit the loop.)
ASSERT_ONLY(bool wroteFwBucket = false);
// Have we skipped the '$' in the last column yet?
ASSERT_ONLY(bool dollarSkipped = false);
TIndexOffU si = 0; // string offset (chars)
ASSERT_ONLY(TIndexOffU lastSufInt = 0);
ASSERT_ONLY(bool inSA = true); // true iff saI still points inside suffix
// array (as opposed to the padding at the
// end)
// Iterate over packed bwt bytes
VMSG_NL("Entering Ebwt loop");
ASSERT_ONLY(TIndexOffU beforeEbwtOff = (uint32_t)out1.tellp());
while(side < ebwtTotSz) {
ASSERT_ONLY(wroteFwBucket = false);
// Sanity-check our cursor into the side buffer
assert_geq(sideCur, 0);
assert_lt(sideCur, (int)eh._sideBwtSz);
assert_eq(0, side % sideSz); // 'side' must be on side boundary
ebwtSide[sideCur] = 0; // clear
assert_lt(side + sideCur, ebwtTotSz);
// Iterate over bit-pairs in the si'th character of the BWT
#ifdef SIXTY4_FORMAT
for(int bpi = 0; bpi < 32; bpi++, si++)
#else
for(int bpi = 0; bpi < 4; bpi++, si++)
#endif
{
int bwtChar;
bool count = true;
if(si <= len) {
// Still in the SA; extract the bwtChar
TIndexOffU saElt = sa.nextSuffix();
// (that might have triggered sa to calc next suf block)
if(isaSample != NULL && (saElt & eh._isaMask) == saElt) {
// This element belongs in the ISA sample. Add
// an entry mapping the text offset to the offset
// into the suffix array that holds the suffix
// beginning with the character at that text offset
assert_lt((saElt >> eh._isaRate), eh._isaLen);
isaSample[saElt >> eh._isaRate] = si;
}
if(saElt == 0) {
// Don't add the '$' in the last column to the BWT
// transform; we can't encode a $ (only A C T or G)
// and counting it as, say, an A, will mess up the
// LR mapping
bwtChar = 0; count = false;
ASSERT_ONLY(dollarSkipped = true);
zOff = si; // remember the SA row that
// corresponds to the 0th suffix
} else {
bwtChar = (int)(s[saElt-1]);
assert_lt(bwtChar, 4);
// Update the fchr
fchr[bwtChar]++;
}
// Update ftab
if((len-saElt) >= (TIndexOffU)eh._ftabChars) {
// Turn the first ftabChars characters of the
// suffix into an integer index into ftab
TIndexOffU sufInt = 0;
for(int i = 0; i < eh._ftabChars; i++) {
sufInt <<= 2;
assert_lt((TIndexOffU)i, len-saElt);
sufInt |= (unsigned char)(s[saElt+i]);
}
// Assert that this prefix-of-suffix is greater
// than or equal to the last one (true b/c the
// suffix array is sorted)
#ifndef NDEBUG
if(lastSufInt > 0) assert_geq(sufInt, lastSufInt);
lastSufInt = sufInt;
#endif
// Update ftab
assert_lt(sufInt+1, ftabLen);
ftab[sufInt+1]++;
if(absorbCnt > 0) {
// Absorb all short suffixes since the last
// transition into this transition
absorbFtab[sufInt] = absorbCnt;
absorbCnt = 0;
}
} else {
// Otherwise if suffix is fewer than ftabChars
// characters long, then add it to the 'absorbCnt';
// it will be absorbed into the next transition
assert_lt(absorbCnt, 255);
absorbCnt++;
}
// Suffix array offset boundary? - update offset array
if((si & eh._offMask) == si) {
assert_lt((si >> eh._offRate), eh._offsLen);
// Write offsets directly to the secondary output
// stream, thereby avoiding keeping them in memory
writeU<TIndexOffU>(out2, saElt, this->toBe());
}
} else {
// Strayed off the end of the SA, now we're just
// padding out a bucket
#ifndef NDEBUG
if(inSA) {
// Assert that we wrote all the characters in the
// string before now
assert_eq(si, len+1);
inSA = false;
}
#endif
// 'A' used for padding; important that padding be
// counted in the occ[] array
bwtChar = 0;
}
if(count) occ[bwtChar]++;
// Append BWT char to bwt section of current side
if(fw) {
// Forward bucket: fill from least to most
#ifdef SIXTY4_FORMAT
ebwtSide[sideCur] |= ((uint64_t)bwtChar << (bpi << 1));
if(bwtChar > 0) assert_gt(ebwtSide[sideCur], 0);
#else
pack_2b_in_8b(bwtChar, ebwtSide[sideCur], bpi);
assert_eq((ebwtSide[sideCur] >> (bpi*2)) & 3, bwtChar);
#endif
} else {
// Backward bucket: fill from most to least
#ifdef SIXTY4_FORMAT
ebwtSide[sideCur] |= ((uint64_t)bwtChar << ((31 - bpi) << 1));
if(bwtChar > 0) assert_gt(ebwtSide[sideCur], 0);
#else
pack_2b_in_8b(bwtChar, ebwtSide[sideCur], 3-bpi);
assert_eq((ebwtSide[sideCur] >> ((3-bpi)*2)) & 3, bwtChar);
#endif
}
} // end loop over bit-pairs
assert_eq(dollarSkipped ? 3 : 0, (occ[0] + occ[1] + occ[2] + occ[3]) & 3);
#ifdef SIXTY4_FORMAT
assert_eq(0, si & 31);
#else
assert_eq(0, si & 3);
#endif
if(fw) sideCur++;
else sideCur--;
#ifdef SIXTY4_FORMAT
if(sideCur == (int)eh._sideBwtSz >> 3)
#else
if(sideCur == (int)eh._sideBwtSz)
#endif
{
// Forward side boundary
assert_eq(0, si % eh._sideBwtLen);
#ifdef SIXTY4_FORMAT
sideCur = (eh._sideBwtSz >> 3) - 1;
#else
sideCur = eh._sideBwtSz - 1;
#endif
assert(fw); fw = false;
ASSERT_ONLY(wroteFwBucket = true);
// Write 'G' and 'T'
assert_leq(occSave[0], occ[2]);
assert_leq(occSave[1], occ[3]);
TIndexOffU *u32side = reinterpret_cast<TIndexOffU*>(ebwtSide);
side += sideSz;
assert_leq(side, eh._ebwtTotSz);
#ifdef BOWTIE_64BIT_INDEX
u32side[(sideSz >> 3)-2] = endianizeU<TIndexOffU>(occSave[0], this->toBe());
u32side[(sideSz >> 3)-1] = endianizeU<TIndexOffU>(occSave[1], this->toBe());
#else
u32side[(sideSz >> 2)-2] = endianizeU<TIndexOffU>(occSave[0], this->toBe());
u32side[(sideSz >> 2)-1] = endianizeU<TIndexOffU>(occSave[1], this->toBe());
#endif
// Write forward side to primary file
out1.write((const char *)ebwtSide, sideSz);
} else if (sideCur == -1) {
// Backward side boundary
assert_eq(0, si % eh._sideBwtLen);
sideCur = 0;
assert(!fw); fw = true;
// Write 'A' and 'C'
TIndexOffU *u32side = reinterpret_cast<TIndexOffU*>(ebwtSide);
side += sideSz;
assert_leq(side, eh._ebwtTotSz);
#ifdef BOWTIE_64BIT_INDEX
u32side[(sideSz >> 3)-2] = endianizeU<TIndexOffU>(occ[0], this->toBe());
u32side[(sideSz >> 3)-1] = endianizeU<TIndexOffU>(occ[1], this->toBe());
#else
u32side[(sideSz >> 2)-2] = endianizeU<TIndexOffU>(occ[0], this->toBe());
u32side[(sideSz >> 2)-1] = endianizeU<TIndexOffU>(occ[1], this->toBe());
#endif
occSave[0] = occ[2]; // save 'G' count
occSave[1] = occ[3]; // save 'T' count
// Write backward side to primary file
out1.write((const char *)ebwtSide, sideSz);
}
}
VMSG_NL("Exited Ebwt loop");
assert(ftab != NULL);
assert_neq(zOff, OFF_MASK);
if(absorbCnt > 0) {
// Absorb any trailing, as-yet-unabsorbed short suffixes into
// the last element of ftab
absorbFtab[ftabLen-1] = absorbCnt;
}
// Assert that our loop counter got incremented right to the end
assert_eq(side, eh._ebwtTotSz);
// Assert that we wrote the expected amount to out1
assert_eq(((TIndexOffU)out1.tellp() - beforeEbwtOff), eh._ebwtTotSz);
// assert that the last thing we did was write a forward bucket
assert(wroteFwBucket);
//
// Write zOff to primary stream
//
writeU<TIndexOffU>(out1, zOff, this->toBe());
//
// Finish building fchr
//
// Exclusive prefix sum on fchr
for(int i = 1; i < 4; i++) {
fchr[i] += fchr[i-1];
}
assert_eq(fchr[3], len);
// Shift everybody up by one
for(int i = 4; i >= 1; i--) {
fchr[i] = fchr[i-1];
}
fchr[0] = 0;
if(_verbose) {
for(int i = 0; i < 5; i++)
cout << "fchr[" << "ACGT$"[i] << "]: " << fchr[i] << endl;
}
// Write fchr to primary file
for(int i = 0; i < 5; i++) {
writeU<TIndexOffU>(out1, fchr[i], this->toBe());
}
//
// Finish building ftab and build eftab
//
// Prefix sum on ftable
TIndexOffU eftabLen = 0;
assert_eq(0, absorbFtab[0]);
for(TIndexOffU i = 1; i < ftabLen; i++) {
if(absorbFtab[i] > 0) eftabLen += 2;
}
assert_leq(eftabLen, (TIndexOffU)eh._ftabChars*2);
eftabLen = eh._ftabChars*2;
TIndexOffU *eftab = NULL;
try {
eftab = new TIndexOffU[eftabLen];
memset(eftab, 0, OFF_SIZE * eftabLen);
} catch(bad_alloc &e) {
cerr << "Out of memory allocating eftab[] "
<< "in Ebwt::buildToDisk() at " << __FILE__ << ":"
<< __LINE__ << endl;
throw e;
}
assert(eftab != NULL);
TIndexOffU eftabCur = 0;
for(TIndexOffU i = 1; i < ftabLen; i++) {
TIndexOffU lo = ftab[i] + Ebwt::ftabHi(ftab, eftab, len, ftabLen, eftabLen, i-1);
if(absorbFtab[i] > 0) {
// Skip a number of short pattern indicated by absorbFtab[i]
TIndexOffU hi = lo + absorbFtab[i];
assert_lt(eftabCur*2+1, eftabLen);
eftab[eftabCur*2] = lo;
eftab[eftabCur*2+1] = hi;
ftab[i] = (eftabCur++) ^ OFF_MASK; // insert pointer into eftab
assert_eq(lo, Ebwt::ftabLo(ftab, eftab, len, ftabLen, eftabLen, i));
assert_eq(hi, Ebwt::ftabHi(ftab, eftab, len, ftabLen, eftabLen, i));
} else {
ftab[i] = lo;
}
}
assert_eq(Ebwt::ftabHi(ftab, eftab, len, ftabLen, eftabLen, ftabLen-1), len+1);
// Write ftab to primary file
for(TIndexOffU i = 0; i < ftabLen; i++) {
writeU<TIndexOffU>(out1, ftab[i], this->toBe());
}
// Write eftab to primary file
for(TIndexOffU i = 0; i < eftabLen; i++) {
writeU<TIndexOffU>(out1, eftab[i], this->toBe());
}
// Write isa to primary file
if(isaSample != NULL) {
ASSERT_ONLY(Bitset sawISA(eh._len+1));
for(TIndexOffU i = 0; i < eh._isaLen; i++) {
TIndexOffU s = isaSample[i];
assert_leq(s, eh._len);
assert(!sawISA.test(s));
ASSERT_ONLY(sawISA.set(s));
writeU<TIndexOffU>(out2, s, this->toBe());
}
delete[] isaSample;
}
delete[] ftab;
delete[] eftab;
delete[] absorbFtab;
// Note: if you'd like to sanity-check the Ebwt, you'll have to
// read it back into memory first!
assert(!isInMemory());
VMSG_NL("Exiting Ebwt::buildToDisk()");
}
/**
* Try to find the Bowtie index specified by the user. First try the
* exact path given by the user. Then try the user-provided string
* appended onto the path of the "indexes" subdirectory below this
* executable, then try the provided string appended onto
* "$BOWTIE_INDEXES/".
*/
string adjustEbwtBase(const string& cmdline,
const string& ebwtFileBase,
bool verbose = false);
#endif /*EBWT_H_*/
|