package codex
sectionYPositions = computeSectionYPositions($el), 10)"
x-init="setTimeout(() => sectionYPositions = computeSectionYPositions($el), 10)"
>
The Codex library for building static analysers based on abstract interpretation
Install
dune-project
Dependency
Authors
Maintainers
Sources
1.0-rc4.tar.gz
md5=bc7266a140c6886add673ede90e335d3
sha512=8da42c0ff2c1098c5f9cb2b5b43b306faf7ac93b8f5ae00c176918cee761f249ff45b29309f31a05bbcf6312304f86a0d5a000eb3f1094d3d3c2b9b4c7f5c386
doc/src/codex.framac_ival/ival.ml.html
Source file ival.ml
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(**************************************************************************) (* This file is part of the Codex semantics library. *) (* *) (* Copyright (C) 2013-2025 *) (* CEA (Commissariat à l'énergie atomique et aux énergies *) (* alternatives) *) (* *) (* you can redistribute it and/or modify it under the terms of the GNU *) (* Lesser General Public License as published by the Free Software *) (* Foundation, version 2.1. *) (* *) (* It is distributed in the hope that it will be useful, *) (* but WITHOUT ANY WARRANTY; without even the implied warranty of *) (* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the *) (* GNU Lesser General Public License for more details. *) (* *) (* See the GNU Lesser General Public License version 2.1 *) (* for more details (enclosed in the file LICENSE). *) (* *) (**************************************************************************) module Extlib = struct let the = function | None -> invalid_arg "the" | Some x -> x let may_map f ?dft x = match x, dft with | None, None -> assert false | None, Some dft -> dft | Some x, _ -> f x let opt_map f = function | None -> None | Some x -> Some (f x) end open Abstract_interp (* Make sure all this is synchronized with the default value of -ilevel *) let small_cardinal = ref 8 let small_cardinal_Int = ref (Int.of_int !small_cardinal) let small_cardinal_log = ref 3 let debug_cardinal = false let set_small_cardinal i = assert (2 <= i && i <= 1024); let rec log j p = if i <= p then j else log (j+1) (2*p) in small_cardinal := i; small_cardinal_Int := Int.of_int i; small_cardinal_log := log 1 2 let get_small_cardinal () = !small_cardinal let emitter = Codex_log.register_category "Ival";; let log_imprecision s = Codex_log.imprecision_warning s ;; module Widen_Arithmetic_Value_Set = struct (* include Datatype_sig.Integer.Set *) include Set.Make(Integer) let hash _ = assert false let pretty fmt s = if is_empty s then Format.fprintf fmt "{}" else Pretty_utils.pp_iter ~pre:"@[<hov 1>{" ~suf:"}@]" ~sep:";@ " iter Int.pretty fmt s let of_list l = match l with | [] -> empty | [e] -> singleton e | e :: q -> List.fold_left (fun acc x -> add x acc) (singleton e) q let default_widen_hints = of_list (List.map Int.of_int [-1;0;1]) let nearest_elt_le elt l = match l with | [] -> raise Not_found | hd::rest when Z.gt hd elt -> raise Not_found | hd::rest -> let rec loop candidate = function | [] -> candidate | hd::rest when Z.gt hd elt -> candidate | hd::rest -> loop hd rest in loop hd rest ;; let nearest_elt_le elt l = nearest_elt_le elt @@ elements l let nearest_elt_ge elt l = let rec loop = function | [] -> raise Not_found | hd::rest when Z.geq hd elt -> hd | _::rest -> loop rest in loop l ;; let nearest_elt_ge elt l = nearest_elt_ge elt @@ elements l end exception Infinity let opt2 f m1 m2 = match m1, m2 with None, _ | _, None -> raise Infinity | Some m1, Some m2 -> f m1 m2 let opt1 f m = match m with None -> None | Some m -> Some (f m) module O = Set.Make(Integer) type pre_set = Pre_set of O.t * int | Pre_top of Int.t * Int.t * Int.t type t = | Set of Int.t array | Float of Fval.t | Top of Int.t option * Int.t option * Int.t * Int.t (* Binary abstract operations do not model precisely float/integer operations. It is the responsibility of the callers to have two operands of the same implicit type. The only exception is for [singleton_zero], which is the correct representation of [0.] *) module Widen_Hints = Widen_Arithmetic_Value_Set type size_widen_hint = Integer.t type generic_widen_hint = Widen_Hints.t type widen_hint = size_widen_hint * generic_widen_hint let some_zero = Some Int.zero let bottom = Set (Array.make 0 Int.zero) let top = Top(None, None, Int.zero, Int.one) let hash_v_option v = match v with None -> 97 | Some v -> Int.hash v let hash v = match v with Set s -> Array.fold_left (fun acc v -> 1031 * acc + (Int.hash v)) 17 s | Top(mn,mx,r,m) -> hash_v_option mn + 5501 * (hash_v_option mx) + 59 * (Int.hash r) + 13031 * (Int.hash m) | Float(f) -> 3 + 17 * Fval.hash f let bound_compare x y = match x,y with None, None -> 0 | None, Some _ -> 1 | Some _, None -> -1 | Some x, Some y -> Int.compare x y exception Unequal of int let compare e1 e2 = if e1==e2 then 0 else match e1,e2 with | Set e1,Set e2 -> let l1 = Array.length e1 in let l2 = Array.length e2 in if l1 <> l2 then l1 - l2 (* no overflow here *) else (try for i=0 to l1 -1 do let r = Int.compare e1.(i) e2.(i) in if r <> 0 then raise (Unequal r) done; 0 with Unequal v -> v ) | _, Set _ -> 1 | Set _, _ -> -1 | Top(mn,mx,r,m), Top(mn',mx',r',m') -> let r1 = bound_compare mn mn' in if r1 <> 0 then r1 else let r2 = bound_compare mx mx' in if r2 <> 0 then r2 else let r3 = Int.compare r r' in if r3 <> 0 then r3 else Int.compare m m' | _, Top _ -> 1 | Top _, _ -> -1 | Float(f1), Float(f2) -> Fval.compare f1 f2 (*| _, Float _ -> 1 | Float _, _ -> -1 *) let equal e1 e2 = compare e1 e2 = 0 let pretty fmt t = match t with | Top(mn,mx,r,m) -> let print_bound fmt = function None -> Format.fprintf fmt "--" | Some v -> Int.pretty fmt v in Format.fprintf fmt "[%a..%a]%t" print_bound mn print_bound mx (fun fmt -> if Int.is_zero r && Int.is_one m then Format.fprintf fmt "" else Format.fprintf fmt ",%a%%%a" Int.pretty r Int.pretty m) | Float (f) -> Fval.pretty fmt f | Set s -> if Array.length s = 0 then Format.fprintf fmt "BottomMod" else begin Pretty_utils.pp_iter ~pre:"@[<hov 1>{" ~suf:"}@]" ~sep:";@ " Array.iter Int.pretty fmt s end let min_le_elt min elt = match min with | None -> true | Some m -> Int.le m elt let max_ge_elt max elt = match max with | None -> true | Some m -> Int.ge m elt let all_positives min = match min with | None -> false | Some m -> Int.ge m Int.zero let all_negatives max = match max with | None -> false | Some m -> Int.le m Int.zero let fail min max r modu = let bound fmt = function | None -> Format.fprintf fmt "--" | Some(x) -> Int.pretty fmt x in Codex_log.fatal "Ival: broken Top, min=%a max=%a r=%a modu=%a" bound min bound max Int.pretty r Int.pretty modu let is_safe_modulo r modu = (Int.ge r Int.zero ) && (Int.ge modu Int.one) && (Int.lt r modu) let is_safe_bound bound r modu = match bound with | None -> true | Some m -> Int.equal (Int.e_rem m modu) r (* Sanity check for Top's arguments *) let check min max r modu = if not (is_safe_modulo r modu && is_safe_bound min r modu && is_safe_bound max r modu) then fail min max r modu let cardinal_zero_or_one v = match v with | Top _ -> false | Set s -> Array.length s <= 1 | Float f -> Fval.is_singleton f let is_singleton_int v = match v with | Float _ | Top _ -> false | Set s -> Array.length s = 1 let is_bottom x = x == bottom let o_zero = O.singleton Int.zero let o_one = O.singleton Int.one let o_zero_or_one = O.union o_zero o_one let small_nums = Array.init 33 (fun i -> Set [| (Integer.of_int i) |]) let zero = small_nums.(0) let one = small_nums.(1) let minus_one = Set [| Int.minus_one |] let zero_or_one = Set [| Int.zero ; Int.one |] let float_zeros = Float Fval.zeros let positive_integers = Top(Some Int.zero, None, Int.zero, Int.one) let negative_integers = Top(None, Some Int.zero, Int.zero, Int.one) let is_zero x = x == zero let inject_singleton e = if Int.le Int.zero e && Int.le e Int.thirtytwo then small_nums.(Int.to_int e) else Set [| e |] let o s = if s = 0 then bottom else if s = 1 then begin let e = O.min_elt o in inject_singleton e end else if O.equal o o_zero_or_one then zero_or_one else let a = Array.make s Int.zero in let i = ref 0 in O.iter (fun e -> a.(!i) <- e; incr i) o; assert (!i = s); Set a let a s = if s = 0 then bottom else let e = a.(0) in if s = 1 && Int.le Int.zero e && Int.le e Int.thirtytwo then small_nums.(Int.to_int e) else if s = 2 && Int.is_zero e && Int.is_one a.(1) then zero_or_one else Set a let inject_float f = if Fval.(equal plus_zero f) then zero else Float f let inject_float_interval flow fup = let flow = Fval.F.of_float flow in let fup = Fval.F.of_float fup in (* make sure that zero float is also zero int *) if Fval.F.equal Fval.F.plus_zero flow && Fval.F.equal Fval.F.plus_zero fup then zero else Float (Fval.inject Fval.Double flow fup) (* let minus_zero = Float (Fval.minus_zero, Fval.minus_zero) *) let is_one = equal one let project_float v = if is_zero v then Fval.plus_zero else match v with | Float f -> f | Top _ | Set _ -> assert false (* by hypothesis that it is a float *) let in_interval x min max r modu = Int.equal (Int.e_rem x modu) r && min_le_elt min x && max_ge_elt max x let array_mem v a = let l = Array.length a in let rec c i = if i = l then (-1) else let ae = a.(i) in if Int.equal ae v then i else if Int.gt ae v then (-1) else c (succ i) in c 0 let contains_zero s = match s with | Top(mn,mx,r,m) -> in_interval Int.zero mn mx r m | Set s -> (array_mem Int.zero s)>=0 | Float f -> Fval.contains_a_zero f let contains_non_zero s = match s with | Top _ -> true (* at least two values *) | Set _ -> not (is_zero s || is_bottom s) | Float f -> Fval.contains_non_zero f exception Not_Singleton_Int let project_int v = match v with | Set [| e |] -> e | _ -> raise Not_Singleton_Int let cardinal v = match v with | Top (None,_,_,_) | Top (_,None,_,_) -> None | Top (Some mn, Some mx,_,m) -> Some (Int.succ ((Int.e_div (Int.sub mx mn) m))) | Set s -> Some (Int.of_int (Array.length s)) | Float f -> if Fval.is_singleton f then Some Int.one else None let cardinal_estimate v ~size = match v with | Set s -> Int.of_int (Array.length s) | Top (None, _, _, _) | Top (_, None, _, _) -> Int.two_power size | Top (Some mn, Some mx, _, d) -> Int.(succ (e_div (sub mx mn) d)) | Float f -> if Fval.is_singleton f then Int.one else let bits_of_float = if Integer.(equal size (of_int 32)) then Fval.bits_of_float32_list else if Integer.(equal size (of_int 64)) then Fval.bits_of_float64_list else (fun _ -> [Int.zero, Int.pred (Int.two_power size)]) in let bits_list = bits_of_float f in let count acc (min, max) = Int.add acc (Int.length min max) in List.fold_left count Int.zero bits_list let cardinal_less_than v n = let c = match v with | Top (None,_,_,_) | Top (_,None,_,_) -> raise Not_less_than | Top (Some mn, Some mx,_,m) -> Int.succ ((Int.e_div (Int.sub mx mn) m)) | Set s -> Int.of_int (Array.length s) | Float f -> if Fval.is_singleton f then Int.one else raise Not_less_than in if Int.le c (Int.of_int n) then Int.to_int c (* This is smaller than the original [n] *) else raise Not_less_than let cardinal_is_less_than v n = match cardinal v with | None -> false | Some c -> Int.le c (Int.of_int n) let min max r modu = let r = Top (min, max, r, modu) in if equal r top then top else r let make ~min ~max ~rem ~modu = match min, max with | Some mn, Some mx -> if Int.gt mx mn then let l = Int.succ (Int.e_div (Int.sub mx mn) modu) in if Int.le l !small_cardinal_Int then let l = Int.to_int l in let s = Array.make l Int.zero in let v = ref mn in let i = ref 0 in while (!i < l) do s.(!i) <- !v; v := Int.add modu !v; incr i done; assert (Int.equal !v (Int.add modu mx)); share_array s l else Top (min, max, rem, modu) else if Int.equal mx mn then inject_singleton mn else bottom | _ -> share_top min max rem modu let inject_top min max rem modu = check min max rem modu; make ~min ~max ~rem ~modu let inject_interval ~min ~max ~rem:r ~modu = assert (is_safe_modulo r modu); let fix_bound fix bound = match bound with | None -> None | Some b -> Some (if Int.equal b (Int.e_rem r modu) then b else fix b) in let min = fix_bound (fun min -> Int.round_up_to_r ~min ~r ~modu) min and max = fix_bound (fun max -> Int.round_down_to_r ~max ~r ~modu) max in make ~min ~max ~rem:r ~modu let subdiv_int v = match v with | Float _ -> raise Can_not_subdiv | Set arr -> let len = Array.length arr in assert (len > 0 ); if len <= 1 then raise Can_not_subdiv; let m = len lsr 1 in let lenhi = len - m in let lo = Array.sub arr 0 m in let hi = Array.sub arr m lenhi in share_array lo m, share_array hi lenhi | Top (Some lo, Some hi, rem, modu) -> let mean = Int.e_div (Int.add lo hi) Int.two in let succmean = Int.succ mean in inject_interval ~min:(Some lo) ~max:(Some mean) ~rem ~modu, inject_interval ~min:(Some succmean) ~max:(Some hi) ~rem ~modu | Top _ -> raise Can_not_subdiv let subdivide ~size = function | Float fval -> let fkind = match Integer.to_int size with | 32 -> Fval.Single | 64 -> Fval.Double | _ -> raise Can_not_subdiv (* see Value/Value#105 *) in let f1, f2 = Fval.subdiv_float_interval fkind fval in inject_float f1, inject_float f2 | ival -> subdiv_int ival let inject_range min max = inject_top min max Int.zero Int.one let top_float = Float Fval.top let top_single_precision_float = Float Fval.top let unsafe_make_top_from_set_4 s = if debug_cardinal then assert (O.cardinal s >= 2); let m = O.min_elt s in let modu = O.fold (fun x acc -> if Int.equal x m then acc else Int.pgcd (Int.sub x m) acc) s Int.zero in let r = Int.e_rem m modu in let max = O.max_elt s in let min = m in (min,max,r,modu) let unsafe_make_top_from_array_4 s = let l = Array.length s in assert (l >= 2); let m = s.(0) in let modu = Array.fold_left (fun acc x -> if Int.equal x m then acc else Int.pgcd (Int.sub x m) acc) Int.zero s in let r = Int.e_rem m modu in let max = Some s.(pred l) in let min = Some m in check min max r modu; (min,max,r,modu) let unsafe_make_top_from_array s = let min, max, r, modu = unsafe_make_top_from_array_4 s in share_top min max r modu let empty_ps = Pre_set (O.empty, 0) let add_ps ps x = match ps with | Pre_set(o,s) -> if debug_cardinal then assert (O.cardinal o = s); if (O.mem x o) (* TODO: improve *) then ps else let no = O.add x o in if s < !small_cardinal then begin if debug_cardinal then assert (O.cardinal no = succ s); Pre_set (no, succ s) end else let min, max, _r, modu = unsafe_make_top_from_set_4 no in Pre_top (min, max, modu) | Pre_top (min, max, modu) -> let new_modu = if Int.equal x min then modu else Int.pgcd (Int.sub x min) modu in let new_min = Int.min min x in let new_max = Int.max max x in Pre_top (new_min, new_max, new_modu) let inject_ps ps = match ps with Pre_set(o, s) -> share_set o s | Pre_top (min, max, modu) -> Top(Some min, Some max, Int.e_rem min modu, modu) let min_max_r_mod t = match t with | Set s -> assert (Array.length s >= 2); unsafe_make_top_from_array_4 s | Top (a,b,c,d) -> a,b,c,d | Float _ -> None, None, Int.zero, Int.one let min_and_max t = match t with | Set s -> let l = Array.length s in if l = 0 then raise Error_Bottom else Some s.(0), Some s.(pred l) | Top (a,b,_,_) -> a, b | Float _ -> None, None let min_and_max_float t = match t with | Set _ when is_zero t -> Some (Fval.F.plus_zero, Fval.F.plus_zero), false | Float f -> Fval.min_and_max f | _ -> assert false let is_float = function | Float _ -> true | Set _ | Top _ -> false let has_greater_min_bound t1 t2 = if is_float t1 || is_float t2 then Fval.has_greater_min_bound (project_float t1) (project_float t2) else let m1, _ = min_and_max t1 in let m2, _ = min_and_max t2 in match m1, m2 with | None, None -> 0 | None, Some _ -> -1 | Some _, None -> 1 | Some m1, Some m2 -> Int.compare m1 m2 let has_smaller_max_bound t1 t2 = if is_float t1 || is_float t2 then Fval.has_smaller_max_bound (project_float t1) (project_float t2) else let _, m1 = min_and_max t1 in let _, m2 = min_and_max t2 in match m1, m2 with | None, None -> 0 | None, Some _ -> -1 | Some _, None -> 1 | Some m1, Some m2 -> Int.compare m2 m1 let widen (bitsize,wh) t1 t2 = if equal t1 t2 || cardinal_zero_or_one t1 then t2 else match t2 with | Float f2 -> let f1 = project_float t1 in Float (Fval.widen f1 f2) | Top _ | Set _ -> (* Add possible interval limits deducted from the bitsize *) let wh = (* If bitsize > 128, the values do not correspond to a scalar type. This can (rarely) happen on structures or arrays that have been reinterpreted as one value by the offsetmaps. In this case, do not use limits, and do not create arbitrarily large integers. *) if Integer.gt bitsize (Integer.of_int 128) then Widen_Hints.empty else if Integer.is_zero bitsize then wh else let limits = [ Integer.neg (Integer.two_power (Integer.pred bitsize)); Integer.pred (Integer.two_power (Integer.pred bitsize)); Integer.pred (Integer.two_power bitsize); ] in Widen_Hints.(union wh (of_list limits)) in let (mn2,mx2,r2,m2) = min_max_r_mod t2 in let (mn1,mx1,r1,m1) = min_max_r_mod t1 in let new_mod = Int.pgcd (Int.pgcd m1 m2) (Int.abs (Int.sub r1 r2)) in let new_rem = Int.e_rem r1 new_mod in let new_min = if bound_compare mn1 mn2 = 0 then mn2 else match mn2 with | None -> None | Some mn2 -> try let v = Widen_Hints.nearest_elt_le mn2 wh in Some (Int.round_up_to_r ~r:new_rem ~modu:new_mod ~min:v) with Not_found -> None in let new_max = if bound_compare mx1 mx2 = 0 then mx2 else match mx2 with None -> None | Some mx2 -> try let v = Widen_Hints.nearest_elt_ge mx2 wh in Some (Int.round_down_to_r ~r:new_rem ~modu:new_mod ~max:v) with Not_found -> None in let result = inject_top new_min new_max new_rem new_mod in (* Format.printf "%a -- %a --> %a (thx to %a)@." pretty t1 pretty t2 pretty result Widen_Hints.pretty wh; *) result let compute_first_common mn1 mn2 r modu = if mn1 = None && mn2 = None then None else let m = match (mn1, mn2) with | Some m, None | None, Some m -> m | Some m1, Some m2 -> Int.max m1 m2 | None, None -> assert false (* already tested above *) in Some (Int.round_up_to_r m r modu) let compute_last_common mx1 mx2 r modu = if mx1 = None && mx2 = None then None else let m = match (mx1, mx2) with | Some m, None | None, Some m -> m | Some m1, Some m2 -> Int.min m1 m2 | None, None -> assert false (* already tested above *) in Some (Int.round_down_to_r m r modu) let min_min x y = match x,y with | None,_ | _,None -> None | Some x, Some y -> Some (Int.min x y) let max_max x y = match x,y with | None,_ | _,None -> None | Some x, Some y -> Some (Int.max x y) (* [extended_euclidian_algorithm a b] returns x,y,gcd such that a*x+b*y=gcd(x,y). *) let extended_euclidian_algorithm a b = assert (Int.gt a Int.zero); assert (Int.gt b Int.zero); let a = ref a and b = ref b in let x = ref Int.zero and lastx = ref Int.one in let y = ref Int.one and lasty = ref Int.zero in while not (Int.is_zero !b) do let (q,r) = Int.e_div_rem !a !b in a := !b; b := r; let tmpx = !x in (x:= Int.sub !lastx (Int.mul q !x); lastx := tmpx); let tmpy = !y in (y:= Int.sub !lasty (Int.mul q !y); lasty := tmpy); done; (!lastx,!lasty,!a) (* [JS 2013/05/23] unused right now [modular_inverse a m] returns [x] such that a*x is congruent to 1 mod m. *) let _modular_inverse a m = let (x,_,gcd) = extended_euclidian_algorithm a m in assert (Int.equal Int.one gcd); x (* This function provides solutions to the Chinese remainder theorem, i.e. it finds the solutions x such that: x == r1 mod m1 && x == r2 mod m2. If no such solution exists, it raises Error_Bottom; else it returns (r,m) such that all solutions x are such that x == r mod m. *) let compute_r_common r1 m1 r2 m2 = (* (E1) x == r1 mod m1 && x == r2 mod m2 <=> \E k1,k2: x = r1 + k1*m1 && x = r2 + k2*m2 <=> \E k1,k2: x = r1 + k1*m1 && k1*m1 - k2*m2 = r2 - r1 Let r = r2 - r1. The equation (E2): k1*m1 - k2*m2 = r is diophantine; there are solutions x to (E1) iff there are solutions (k1,k2) to (E2). Let d = pgcd(m1,m2). There are solutions to (E2) only if d divides r (because d divides k1*m1 - k2*m2). Else we raise [Error_Bottom]. *) let (x1,_,pgcd) = extended_euclidian_algorithm m1 m2 in let r = Int.sub r2 r1 in let r_div,r_rem = Int.e_div_rem r pgcd in if not (Int.equal r_rem Int.zero) then raise Error_Bottom (* The extended euclidian algorithm has provided solutions x1,x2 to the Bezout identity x1*m1 + x2*m2 = d. x1*m1 + x2*m2 = d ==> x1*(r/d)*m1 + x2*(r/d)*m2 = d*(r/d). Thus, k1 = x1*(r/d), k2=-x2*(r/d) are solutions to (E2) Thus, x = r1 + x1*(r/d)*m1 is a particular solution to (E1). *) else let k1 = Int.mul x1 r_div in let x = Int.add r1 (Int.mul k1 m1) in (* If two solutions x and y exist, they are equal modulo ppcm(m1,m2). We have x == r1 mod m1 && y == r1 mod m1 ==> \E k1: x - y = k1*m1 x == r2 mod m2 && y == r2 mod m2 ==> \E k2: x - y = k2*m2 Thus k1*m1 = k2*m2 is a multiple of m1 and m2, i.e. is a multiple of ppcm(m1,m2). Thus x = y mod ppcm(m1,m2). *) let ppcm = Integer.ppcm m1 m2 in (* x may be bigger than the ppcm, we normalize it. *) (Int.e_rem x ppcm, ppcm) ;; let array_truncate r i = if i = 0 then bottom else if i = 1 then inject_singleton r.(0) else begin let r = Array.sub r 0 i in (* (Obj.truncate (Obj.repr r) i); *) assert (Array.length r = i); Set r end let array_filter (f : Int.t -> bool) (a : Int.t array) : t = let l = Array.length a in let r = Array.make l Int.zero in let j = ref 0 in for i = 0 to l - 1 do let x = a.(i) in if f x then begin r.(!j) <- x; incr j; end done; array_truncate r !j let array_map_reduce (f : 'a -> 'b) (g : 'b -> 'b -> 'b) (set : 'a array) : 'b = if Array.length set <= 0 then raise Error_Bottom else let acc = ref (f set.(0)) in for i = 1 to Array.length set - 1 do acc := g !acc (f set.(i)) done; !acc let array_inter a1 a2 = let l1 = Array.length a1 in let l2 = Array.length a2 in let lr_max = min l1 l2 in let r = Array.make lr_max Int.zero in let rec c i i1 i2 = if i1 = l1 || i2 = l2 then array_truncate r i else let e1 = a1.(i1) in let e2 = a2.(i2) in if Int.equal e1 e2 then begin r.(i) <- e1; c (succ i) (succ i1) (succ i2) end else if Int.lt e1 e2 then c i (succ i1) i2 else c i i1 (succ i2) in c 0 0 0 (* Do the two arrays have an integer in common *) let arrays_intersect a1 a2 = let l1 = Array.length a1 in let l2 = Array.length a2 in let rec aux i1 i2 = if i1 = l1 || i2 = l2 then false else let e1 = a1.(i1) in let e2 = a2.(i2) in if Int.equal e1 e2 then true else if Int.lt e1 e2 then aux (succ i1) i2 else aux i1 (succ i2) in aux 0 0 let meet v1 v2 = if v1 == v2 then v1 else let result = match v1,v2 with | Top(min1,max1,r1,modu1), Top(min2,max2,r2,modu2) -> begin try let r,modu = compute_r_common r1 modu1 r2 modu2 in inject_top (compute_first_common min1 min2 r modu) (compute_last_common max1 max2 r modu) r modu with Error_Bottom -> (*Format.printf "meet to bottom: %a /\\ %a@\n" pretty v1 pretty v2;*) bottom end | Set s1 , Set s2 -> array_inter s1 s2 | Set s, Top(min, max, rm, modu) | Top(min, max, rm, modu), Set s -> let l = Array.length s in let r = Array.make l Int.zero in let rec c i j = if i = l then array_truncate r j else let si = succ i in let x = s.(i) in if in_interval x min max rm modu then begin r.(j) <- x; c si (succ j) end else c si j in c 0 0 | Float(f1), Float(f2) -> begin match Fval.meet f1 f2 with | `Value f -> inject_float f | `Bottom -> bottom end | (Float f as ff), (Top _ | Set _ as o) | (Top _ | Set _ as o), (Float f as ff) -> if equal o top then ff else if contains_zero o && Fval.contains_plus_zero f then zero else bottom in (* Format.printf "meet: %a /\\ %a -> %a@\n" pretty v1 pretty v2 pretty result;*) result let intersects v1 v2 = v1 == v2 || match v1, v2 with | Top _, Top _ -> not (is_bottom (meet v1 v2)) (* YYY: slightly inefficient *) | Set s1 , Set s2 -> arrays_intersect s1 s2 | Set s, Top (min, max, rm, modu) | Top (min, max, rm, modu), Set s -> Array.exists (fun x -> in_interval x min max rm modu) s | Float f1, Float f2 -> begin match Fval.forward_comp Comp.Eq f1 f2 with | Comp.False -> false | Comp.True | Comp.Unknown -> true end | Float f, other | other, Float f -> equal top other || (Fval.contains_plus_zero f && contains_zero other) let narrow v1 v2 = match v1, v2 with | _, Set [||] | Set [||], _ -> bottom | Float _, Float _ | (Top _| Set _), (Top _ | Set _) -> meet v1 v2 (* meet is exact *) | v, (Top _ as t) when equal t top -> v | (Top _ as t), v when equal t top -> v | Float f, (Set _ as s) | (Set _ as s), Float f when is_zero s -> begin match Fval.narrow f Fval.zeros with | `Value f -> inject_float f | `Bottom -> bottom end | Float _, (Set _ | Top _) | (Set _ | Top _), Float _ -> (* ill-typed case. It is better to keep the operation symmetric *) top (* Given a set of elements that is an under-approximation, returns an ival (while maintaining the ival invariants that the "Set" constructor is used only for small sets of elements. *) let set_to_ival_under set = let card = Int.Set.cardinal set in if card <= !small_cardinal then (let a = Array.make card Int.zero in ignore(Int.Set.fold (fun elt i -> Array.set a i elt; i + 1) set 0); share_array a card) else (* If by chance the set is contiguous. *) if (Int.equal (Int.sub (Int.Set.max_elt set) (Int.Set.min_elt set)) (Int.of_int (card - 1))) then Top( Some(Int.Set.min_elt set), Some(Int.Set.max_elt set), Int.one, Int.zero) (* Else: arbitrarily drop some elements of the under approximation. *) else let a = Array.make !small_cardinal Int.zero in log_imprecision "Ival.set_to_ival_under"; try ignore(Int.Set.fold (fun elt i -> if i = !small_cardinal then raise Exit; Array.set a i elt; i + 1) set 0); assert false with Exit -> Set a ;; let link v1 v2 = match v1, v2 with | Set a1, Set a2 -> let s1 = Array.fold_right Int.Set.add a1 Int.Set.empty in let s2 = Array.fold_right Int.Set.add a2 s1 in set_to_ival_under s2 | Top(mn1,mx1,r1,m1), Top(mn2,mx2,r2,m2) -> if Int.equal r1 r2 && Int.equal m1 m2 then let min = match mn1,mn2 with | Some(a), Some(b) -> Some(Int.min a b) | _ -> None in let max = match mx1,mx2 with | Some(a), Some(b) -> Some(Int.max a b) | _ -> None in inject_top min max r1 m1 else v1 (* No best abstraction anyway. *) | Top(mn,mx,r,m), Set s | Set s, Top(mn,mx,r,m) -> let max = match mx with | None -> None | Some(max) -> let curmax = ref max in for i = 0 to (Array.length s) - 1 do let elt = s.(i) in if Int.equal elt (Int.add !curmax m) then curmax := elt done; Some(!curmax) in let min = match mn with | None -> None | Some(min) -> let curmin = ref min in for i = (Array.length s) - 1 downto 0 do let elt = s.(i) in if Int.equal elt (Int.sub !curmin m) then curmin := elt done; Some(!curmin) in inject_top min max r m | _ -> bottom ;; let join v1 v2 = let result = if v1 == v2 then v1 else match v1,v2 with | Top(mn1,mx1,r1,m1), Top(mn2,mx2,r2,m2) -> check mn1 mx1 r1 m1; check mn2 mx2 r2 m2; let modu = Int.pgcd (Int.pgcd m1 m2) (Int.abs(Int.sub r1 r2)) in let r = Int.e_rem r1 modu in let min = min_min mn1 mn2 in let max = max_max mx1 mx2 in let r = inject_top min max r modu in r | Set s, (Top(min, max, r, modu) as t) | (Top(min, max, r, modu) as t), Set s -> let l = Array.length s in if l = 0 then t else let f modu elt = Int.pgcd modu (Int.abs(Int.sub r elt)) in let new_modu = Array.fold_left f modu s in let new_r = Int.e_rem r new_modu in let new_min = match min with None -> None | Some m -> Some (Int.min m s.(0)) in let new_max = match max with None -> None | Some m -> Some (Int.max m s.(pred l)) in check new_min new_max new_r new_modu; share_top new_min new_max new_r new_modu | Set s1 , Set s2 -> let l1 = Array.length s1 in if l1 = 0 then v2 else let l2 = Array.length s2 in if l2 = 0 then v1 else (* second pass: make a set or make a top *) let second uniq = if uniq <= !small_cardinal then let r = Array.make uniq Int.zero in let rec c i i1 i2 = if i1 = l1 then begin Array.blit s2 i2 r i (l2 - i2); share_array r uniq end else if i2 = l2 then begin Array.blit s1 i1 r i (l1 - i1); share_array r uniq end else let si = succ i in let e1 = s1.(i1) in let e2 = s2.(i2) in if Int.lt e2 e1 then begin r.(i) <- e2; c si i1 (succ i2) end else begin r.(i) <- e1; let si1 = succ i1 in if Int.equal e1 e2 then begin c si si1 (succ i2) end else begin c si si1 i2 end end in c 0 0 0 else begin let m = Int.min s1.(0) s2.(0) in let accum acc x = if Int.equal x m then acc else Int.pgcd (Int.sub x m) acc in let modu = ref Int.zero in for j = 0 to pred l1 do modu := accum !modu s1.(j) done; for j = 0 to pred l2 do modu := accum !modu s2.(j) done; inject_ps (Pre_top (m, Int.max s1.(pred l1) s2.(pred l2), !modu)) end in (* first pass: count unique elements and detect inclusions *) let rec first i1 i2 uniq inc1 inc2 = let finished1 = i1 = l1 in if finished1 then begin if inc2 then v2 else second (uniq + l2 - i2) end else let finished2 = i2 = l2 in if finished2 then begin if inc1 then v1 else second (uniq + l1 - i1) end else let e1 = s1.(i1) in let e2 = s2.(i2) in if Int.lt e2 e1 then begin first i1 (succ i2) (succ uniq) false inc2 end else if Int.gt e2 e1 then begin first (succ i1) i2 (succ uniq) inc1 false end else first (succ i1) (succ i2) (succ uniq) inc1 inc2 in first 0 0 0 true true | Float(f1), Float(f2) -> inject_float (Fval.join f1 f2) | Float (f) as ff, other | other, (Float (f) as ff) -> if is_zero other then inject_float (Fval.join Fval.plus_zero f) else if is_bottom other then ff else top in (* Format.printf "mod_join %a %a -> %a@." pretty v1 pretty v2 pretty result; *) result let fold_int f v acc = match v with Top(None,_,_,_) | Top(_,None,_,_) | Float _ -> raise Error_Top | Top(Some inf, Some sup, _, step) -> Int.fold f ~inf ~sup ~step acc | Set s -> Array.fold_left (fun acc x -> f x acc) acc s let fold_int_decrease f v acc = match v with Top(None,_,_,_) | Top(_,None,_,_) | Float _ -> raise Error_Top | Top(Some inf, Some sup, _, step) -> Int.fold f ~inf ~sup ~step:(Int.neg step) acc | Set s -> Array.fold_right (fun x acc -> f x acc) s acc let fold_enum f v acc = match v with | Float fl when Fval.is_singleton fl -> f v acc | Float _ -> raise Error_Top | Set _ | Top _ -> fold_int (fun x acc -> f (inject_singleton x) acc) v acc (** [min_is_lower mn1 mn2] is true iff mn1 is a lower min than mn2 *) let min_is_lower mn1 mn2 = match mn1, mn2 with None, _ -> true | _, None -> false | Some m1, Some m2 -> Int.le m1 m2 (** [max_is_greater mx1 mx2] is true iff mx1 is a greater max than mx2 *) let max_is_greater mx1 mx2 = match mx1, mx2 with None, _ -> true | _, None -> false | Some m1, Some m2 -> Int.ge m1 m2 let rem_is_included r1 m1 r2 m2 = (Int.is_zero (Int.e_rem m1 m2)) && (Int.equal (Int.e_rem r1 m2) r2) let array_for_all f (a : Integer.t array) = let l = Array.length a in let rec c i = i = l || ((f a.(i)) && c (succ i)) in c 0 let array_subset a1 a2 = let l1 = Array.length a1 in let l2 = Array.length a2 in if l1 > l2 then false else let rec c i1 i2 = if i1 = l1 then true else if i2 = l2 then false else let e1 = a1.(i1) in let e2 = a2.(i2) in let si2 = succ i2 in if Int.equal e1 e2 then c (succ i1) si2 else if Int.lt e1 e2 then false else c i1 si2 (* TODO: improve by not reading a1.(i1) all the time *) in c 0 0 let is_included t1 t2 = (t1 == t2) || match t1,t2 with | Set [||], _ -> true | Top(mn1,mx1,r1,m1), Top(mn2,mx2,r2,m2) -> (min_is_lower mn2 mn1) && (max_is_greater mx2 mx1) && rem_is_included r1 m1 r2 m2 | Top _, Set _ -> false (* Top _ represents more elements than can be represented by Set _ *) | Set s, Top(min, max, r, modu) -> (* Inclusion of bounds is needed for the entire inclusion *) min_le_elt min s.(0) && max_ge_elt max s.(Array.length s-1) && (Int.equal Int.one modu || (*Top side contains all integers, we're done*) array_for_all (fun x -> Int.equal (Int.e_rem x modu) r) s) | Set s1, Set s2 -> array_subset s1 s2 | Float f1, Float f2 -> Fval.is_included f1 f2 | Float _, _ -> equal t2 top | Set _, Float f -> is_zero t1 && Fval.contains_plus_zero f | Top _, Float _ -> false let map_set_exnsafe_acc f acc (s : Integer.t array) = Array.fold_left (fun acc v -> add_ps acc (f v)) acc s let map_set_exnsafe f (s : Integer.t array) = inject_ps (map_set_exnsafe_acc f empty_ps s) let apply2_notzero f (s1 : Integer.t array) s2 = inject_ps (Array.fold_left (fun acc v1 -> Array.fold_left (fun acc v2 -> if Int.is_zero v2 then acc else add_ps acc (f v1 v2)) acc s2) empty_ps s1) let apply2_n f (s1 : Integer.t array) (s2 : Integer.t array) = let ps = ref empty_ps in let l1 = Array.length s1 in let l2 = Array.length s2 in for i1 = 0 to pred l1 do let e1 = s1.(i1) in for i2 = 0 to pred l2 do ps := add_ps !ps (f e1 s2.(i2)) done done; inject_ps !ps let apply_set f v1 v2 = match v1,v2 with | Set s1, Set s2 -> apply2_n f s1 s2 | _ -> top let apply_set_unary f v = match v with | Set s -> map_set_exnsafe f s | _ -> top let apply_bin_1_strict_incr f x (s : Integer.t array) = let l = Array.length s in let r = Array.make l Int.zero in let rec c i = if i = l then share_array r l else let v = f x s.(i) in r.(i) <- v; c (succ i) in c 0 let apply_bin_1_strict_decr f x (s : Integer.t array) = let l = Array.length s in let r = Array.make l Int.zero in let rec c i = if i = l then share_array r l else let v = f x s.(i) in r.(l - i - 1) <- v; c (succ i) in c 0 let map_set_strict_decr f (s : Integer.t array) = let l = Array.length s in let r = Array.make l Int.zero in let rec c i = if i = l then share_array r l else let v = f s.(i) in r.(l - i - 1) <- v; c (succ i) in c 0 let map_set_decr f (s : Integer.t array) = let l = Array.length s in if l = 0 then bottom else let r = Array.make l Int.zero in let rec c srcindex dstindex last = if srcindex < 0 then begin r.(dstindex) <- last; array_truncate r (succ dstindex) end else let v = f s.(srcindex) in if Int.equal v last then c (pred srcindex) dstindex last else begin r.(dstindex) <- last; c (pred srcindex) (succ dstindex) v end in c (l-2) 0 (f s.(pred l)) let map_set_incr f (s : Integer.t array) = let l = Array.length s in if l = 0 then bottom else let r = Array.make l Int.zero in let rec c srcindex dstindex last = if srcindex = l then begin r.(dstindex) <- last; array_truncate r (succ dstindex) end else let v = f s.(srcindex) in if Int.equal v last then c (succ srcindex) dstindex last else begin r.(dstindex) <- last; c (succ srcindex) (succ dstindex) v end in c 1 0 (f s.(0)) let add_singleton_int i v = match v with | Float _ -> assert false | Set s -> apply_bin_1_strict_incr Int.add i s | Top (mn, mx, r, m) -> let incr v = Int.add i v in let new_mn = opt1 incr mn in let new_mx = opt1 incr mx in let new_r = Int.e_rem (incr r) m in share_top new_mn new_mx new_r m let rec add_int v1 v2 = match v1,v2 with | Float _, _ | _, Float _ -> assert false | Set [| x |], Set s | Set s, Set [| x |]-> apply_bin_1_strict_incr Int.add x s | Set s1, Set s2 -> apply2_n Int.add s1 s2 | Top(mn1,mx1,r1,m1), Top(mn2,mx2,r2,m2) -> let m = Int.pgcd m1 m2 in let r = Int.e_rem (Int.add r1 r2) m in let mn = try Some (Int.round_up_to_r (opt2 Int.add mn1 mn2) r m) with Infinity -> None in let mx = try Some (Int.round_down_to_r (opt2 Int.add mx1 mx2) r m) with Infinity -> None in inject_top mn mx r m | Set s, (Top _ as t) | (Top _ as t), Set s -> let l = Array.length s in if l = 0 then bottom else if l = 1 then (* only one element *) add_singleton_int s.(0) t else add_int t (unsafe_make_top_from_array s) let add_int_under v1 v2 = match v1,v2 with | Float _, _ | _, Float _ -> assert false | Set [| x |], Set s | Set s, Set [| x |]-> apply_bin_1_strict_incr Int.add x s | Set s1, Set s2 -> let set = Array.fold_left (fun acc i1 -> Array.fold_left (fun acc i2 -> Int.Set.add (Int.add i1 i2) acc) acc s2) Int.Set.empty s1 in set_to_ival_under set | Top(min1,max1,r1,modu1) , Top(min2,max2,r2,modu2) when Int.equal modu1 modu2 -> (* Note: min1+min2 % modu = max1 + max2 % modu = r1 + r2 % modu; no need to trim the bounds here. *) let r = Int.e_rem (Int.add r1 r2) modu1 in let min = match min1, min2 with | Some min1, Some min2 -> Some (Int.add min1 min2) | _ -> None in let max = match max1, max2 with | Some max1, Some max2 -> Some (Int.add max1 max2) | _ -> None in inject_top min max r modu1 (* In many cases, there is no best abstraction; for instance when modu1 divides modu2, a part of the resulting interval is congruent to modu1, and a larger part is congruent to modu2. In general, one can take the intersection. In any case, this code should be rarely called. *) | Top _, Top _ -> bottom | Set s, (Top _ as t) | (Top _ as t), Set s -> let l = Array.length s in if l = 0 then bottom else if l = 1 then (* only one element: precise. *) add_singleton_int s.(0) t else begin log_imprecision "Ival.add_int_under"; (* Not worse than another computation. *) add_singleton_int s.(0) t end ;; let neg_int v = match v with | Float _ -> assert false | Set s -> map_set_strict_decr Int.neg s | Top(mn,mx,r,m) -> share_top (opt1 Int.neg mx) (opt1 Int.neg mn) (Int.e_rem (Int.neg r) m) m let sub_int v1 v2 = add_int v1 (neg_int v2) let sub_int_under v1 v2 = add_int_under v1 (neg_int v2) type ext_value = Ninf | Pinf | Val of Int.t let inject_min = function None -> Ninf | Some m -> Val m let inject_max = function None -> Pinf | Some m -> Val m let ext_neg = function Ninf -> Pinf | Pinf -> Ninf | Val v -> Val (Int.neg v) let ext_mul x y = match x, y with | Ninf, Ninf | Pinf, Pinf -> Pinf | Ninf, Pinf | Pinf, Ninf -> Ninf | Val v1, Val v2 -> Val (Int.mul v1 v2) | (Ninf | Pinf as x), Val v | Val v, (Ninf | Pinf as x) -> if Int.gt v Int.zero then x else if Int.lt v Int.zero then ext_neg x else Val Int.zero let ext_min x y = match x,y with Ninf, _ | _, Ninf -> Ninf | Pinf, x | x, Pinf -> x | Val x, Val y -> Val(Int.min x y) let ext_max x y = match x,y with Pinf, _ | _, Pinf -> Pinf | Ninf, x | x, Ninf -> x | Val x, Val y -> Val(Int.max x y) let ext_proj = function Val x -> Some x | _ -> None let min_int s = match s with | Top (min,_,_,_) -> min | Set s -> if Array.length s = 0 then raise Error_Bottom else Some s.(0) | Float _ -> None let max_int s = match s with | Top (_,max,_,_) -> max | Set s -> let l = Array.length s in if l = 0 then raise Error_Bottom else Some s.(pred l) | Float _ -> None (* TODO: rename this function to scale_int *) let scale f v = if Int.is_zero f then zero else match v with | Float _ -> top | Top(mn1,mx1,r1,m1) -> let incr = Int.mul f in if Int.gt f Int.zero then let modu = incr m1 in share_top (opt1 incr mn1) (opt1 incr mx1) (Int.e_rem (incr r1) modu) modu else let modu = Int.neg (incr m1) in share_top (opt1 incr mx1) (opt1 incr mn1) (Int.e_rem (incr r1) modu) modu | Set s -> if Int.ge f Int.zero then apply_bin_1_strict_incr Int.mul f s else apply_bin_1_strict_decr Int.mul f s let scale_div_common ~pos f v degenerate_ival degenerate_float = assert (not (Int.is_zero f)); let div_f = if pos then fun a -> Int.e_div a f else fun a -> Int.c_div a f in match v with | Top(mn1,mx1,r1,m1) -> let r, modu = let negative = max_is_greater (some_zero) mx1 in if (negative (* all negative *) || pos (* good div *) || (min_is_lower (some_zero) mn1) (* all positive *) || (Int.is_zero (Int.e_rem r1 f)) (* exact *) ) && (Int.is_zero (Int.e_rem m1 f)) then let modu = Int.abs (div_f m1) in let r = if negative then Int.sub r1 m1 else r1 in (Int.e_rem (div_f r) modu), modu else (* degeneration*) degenerate_ival r1 m1 in let divf_mn1 = opt1 div_f mn1 in let divf_mx1 = opt1 div_f mx1 in let mn, mx = if Int.gt f Int.zero then divf_mn1, divf_mx1 else divf_mx1, divf_mn1 in inject_top mn mx r modu | Set s -> if Int.lt f Int.zero then map_set_decr div_f s else map_set_incr div_f s | Float _ -> degenerate_float let scale_div ~pos f v = scale_div_common ~pos f v (fun _ _ -> Int.zero, Int.one) top ;; let scale_div_under ~pos f v = try (* TODO: a more precise result could be obtained by transforming Top(min,max,r,m) into Top(min,max,r/f,m/gcd(m,f)). But this is more complex to implement when pos or f is negative. *) scale_div_common ~pos f v (fun _r _m -> raise Exit) bottom with Exit -> bottom ;; let div_set x sy = Array.fold_left (fun acc elt -> if Int.is_zero elt then acc else join acc (scale_div ~pos:false elt x)) bottom sy (* ymin and ymax must be the same sign *) let div_range x ymn ymx = match min_and_max x with | Some xmn, Some xmx -> let c1 = Int.c_div xmn ymn in let c2 = Int.c_div xmx ymn in let c3 = Int.c_div xmn ymx in let c4 = Int.c_div xmx ymx in let min = Int.min (Int.min c1 c2) (Int.min c3 c4) in let max = Int.max (Int.max c1 c2) (Int.max c3 c4) in (* Format.printf "div: %a %a %a %a@." Int.pretty mn Int.pretty mx Int.pretty xmn Int.pretty xmx; *) inject_range (Some min) (Some max) | _ -> log_imprecision "Ival.div_range"; top let div x y = (*if (intersects y negative || intersects x negative) then ignore (CilE.warn_once "using 'round towards zero' semantics for '/', which only became specified in C99."); *) match y with Set sy -> div_set x sy | Top (Some mn,Some mx, r, modu) -> let result_pos = if Int.gt mx Int.zero then let lpos = if Int.gt mn Int.zero then mn else Int.round_up_to_r ~min:Int.one ~r ~modu in div_range x lpos mx else bottom in let result_neg = if Int.lt mn Int.zero then let gneg = if Int.lt mx Int.zero then mx else Int.round_down_to_r ~max:Int.minus_one ~r ~modu in div_range x mn gneg else bottom in join result_neg result_pos | Float _ -> assert false | Top (None, _, _, _) | Top (_, None, _, _) -> log_imprecision "Ival.div"; top (* [scale_rem ~pos:false f v] is an over-approximation of the set of elements [x mod f] for [x] in [v]. [scale_rem ~pos:true f v] is an over-approximation of the set of elements [x e_rem f] for [x] in [v]. *) let scale_rem ~pos f v = (* Format.printf "scale_rem %b %a %a@." pos Int.pretty f pretty v; *) if Int.is_zero f then bottom else let f = if Int.lt f Int.zero then Int.neg f else f in let rem_f a = if pos then Int.e_rem a f else Int.c_rem a f in match v with | Top(mn,mx,r,m) -> let modu = Int.pgcd f m in let rr = Int.e_rem r modu in let binf,bsup = if pos then (Int.round_up_to_r ~min:Int.zero ~r:rr ~modu), (Int.round_down_to_r ~max:(Int.pred f) ~r:rr ~modu) else let min = if all_positives mn then Int.zero else Int.neg (Int.pred f) in let max = if all_negatives mx then Int.zero else Int.pred f in (Int.round_up_to_r ~min ~r:rr ~modu, Int.round_down_to_r ~max ~r:rr ~modu) in let mn_rem,mx_rem = match mn,mx with | Some mn,Some mx -> let div_f a = if pos then Int.e_div a f else Int.c_div a f in (* See if [mn..mx] is included in [k*f..(k+1)*f] for some [k]. In this case, [%] is monotonic and [mn%f .. mx%f] is a more precise result. *) if Int.equal (div_f mn) (div_f mx) then rem_f mn, rem_f mx else binf,bsup | _ -> binf,bsup in inject_top (Some mn_rem) (Some mx_rem) rr modu | Set s -> map_set_exnsafe rem_f s | Float _ -> top let c_rem x y = match y with | Top (None, _, _, _) | Top (_, None, _, _) | Float _ -> top | Top (Some mn, Some mx, _, _) -> if Int.equal mx Int.zero then bottom (* completely undefined. *) else (* Result is of the sign of x. Also, compute |x| to bound the result *) let neg, pos, max_x = match x with | Float _ -> true, true, None | Set set -> let s = Array.length set in if s = 0 then (* Bottom *) false, false, None else Int.le set.(0) Int.minus_one, Int.ge set.(s-1) Int.one, Some (Int.max (Int.abs set.(0)) (Int.abs set.(s-1))) | Top (mn, mx, _, _) -> min_le_elt mn Int.minus_one, max_ge_elt mx Int.one, (match mn, mx with | Some mn, Some mx -> Some (Int.max (Int.abs mn) (Int.abs mx)) | _ -> None) in (* Bound the result: no more than |x|, and no more than |y|-1 *) let pos_rem = Integer.max (Int.abs mn) (Int.abs mx) in let bound = Int.pred pos_rem in let bound = Extlib.may_map (Int.min bound) ~dft:bound max_x in (* Compute result bounds using sign information *) let mn = if neg then Some (Int.neg bound) else Some Int.zero in let mx = if pos then Some bound else Some Int.zero in inject_top mn mx Int.zero Int.one | Set yy -> ( match x with Set xx -> apply2_notzero Int.c_rem xx yy | Float _ -> top | Top _ -> let f acc y = join (scale_rem ~pos:false y x) acc in Array.fold_left f bottom yy) module AllValueHashtbl = Hashtbl.Make (struct type t = Int.t * bool * int let equal (a,b,c:t) (d,e,f:t) = b=e && c=f && Int.equal a d let hash (a,b,c:t) = 257 * (Hashtbl.hash b) + 17 * (Hashtbl.hash c) + Int.hash a end) let all_values_table = AllValueHashtbl.create 7 let create_all_values_modu ~modu ~signed ~size = let t = modu, signed, size in try AllValueHashtbl.find all_values_table t with Not_found -> let mn, mx = if signed then let b = Int.two_power_of_int (size-1) in (Int.round_up_to_r ~min:(Int.neg b) ~modu ~r:Int.zero, Int.round_down_to_r ~max:(Int.pred b) ~modu ~r:Int.zero) else let b = Int.two_power_of_int size in Int.zero, Int.round_down_to_r ~max:(Int.pred b) ~modu ~r:Int.zero in let r = inject_top (Some mn) (Some mx) Int.zero modu in AllValueHashtbl.add all_values_table t r; r let create_all_values ~signed ~size = if size <= !small_cardinal_log then (* We may need to create a set. Use slow path *) create_all_values_modu ~signed ~size ~modu:Int.one else if signed then let b = Int.two_power_of_int (size-1) in Top (Some (Int.neg b), Some (Int.pred b), Int.zero, Int.one) else let b = Int.two_power_of_int size in Top (Some Int.zero, Some (Int.pred b), Int.zero, Int.one) let big_int_64 = Int.of_int 64 let big_int_32 = Int.thirtytwo let cast_int_to_int ~size ~signed value = if equal top value then create_all_values ~size:(Int.to_int size) ~signed else let result = let factor = Int.two_power size in let mask = Int.two_power (Int.pred size) in let rem_f value = Int.cast ~size ~signed ~value in let not_p_factor = Int.neg factor in let best_effort r m = let modu = Int.pgcd factor m in let rr = Int.e_rem r modu in let min_val = Some (if signed then Int.round_up_to_r ~min:(Int.neg mask) ~r:rr ~modu else Int.round_up_to_r ~min:Int.zero ~r:rr ~modu) in let max_val = Some (if signed then Int.round_down_to_r ~max:(Int.pred mask) ~r:rr ~modu else Int.round_down_to_r ~max:(Int.pred factor) ~r:rr ~modu) in inject_top min_val max_val rr modu in match value with | Top(Some mn,Some mx,r,m) -> let highbits_mn,highbits_mx = if signed then Int.logand (Int.add mn mask) not_p_factor, Int.logand (Int.add mx mask) not_p_factor else Int.logand mn not_p_factor, Int.logand mx not_p_factor in if Int.equal highbits_mn highbits_mx then if Int.is_zero highbits_mn then value else let new_min = rem_f mn in let new_r = Int.e_rem new_min m in inject_top (Some new_min) (Some (rem_f mx)) new_r m else best_effort r m | Top (_,_,r,m) -> best_effort r m | Set s -> begin let all = create_all_values ~size:(Int.to_int size) ~signed in if is_included value all then value else map_set_exnsafe rem_f s end | Float _ -> assert false in (* If sharing is no longer preserved, please change Cvalue.V.cast *) if equal result value then value else result let reinterpret_float_as_int ~signed ~size f = let reinterpret_list l = let reinterpret_one (b, e) = let i = inject_range (Some b) (Some e) in cast_int_to_int ~size ~signed i in let l = List.map reinterpret_one l in List.fold_left join bottom l in if Int.equal size big_int_64 then let itvs = Fval.bits_of_float64_list f in reinterpret_list itvs else if Int.equal size big_int_32 then let itvs = Fval.bits_of_float32_list f in reinterpret_list itvs else top let reinterpret_as_int ~size ~signed i = match i with | Set _ | Top _ -> (* On integers, cast and reinterpretation are the same operation *) cast_int_to_int ~signed ~size i | Float f -> reinterpret_float_as_int ~signed ~size f let cast_float_to_float fkind v = match v with | Float f -> begin match fkind with | Fval.Real | Fval.Long_Double | Fval.Double -> v | Fval.Single -> inject_float (Fval.round_to_single_precision_float f) end | Set _ when is_zero v -> zero | Set _ | Top _ -> top_float (* TODO rename to mul_int *) let rec mul v1 v2 = (* Format.printf "mul. Args: '%a' '%a'@\n" pretty v1 pretty v2; *) let result = if is_one v1 then v2 else if is_zero v2 || is_zero v1 then zero else if is_one v2 then v1 else match v1,v2 with | Float _, _ | _, Float _ -> top | Set s1, Set [| x |] | Set [| x |], Set s1 -> if Int.ge x Int.zero then apply_bin_1_strict_incr Int.mul x s1 else apply_bin_1_strict_decr Int.mul x s1 | Set s1, Set s2 -> apply2_n Int.mul s1 s2 | Top(mn1,mx1,r1,m1), Top(mn2,mx2,r2,m2) -> check mn1 mx1 r1 m1; check mn2 mx2 r2 m2; let mn1 = inject_min mn1 in let mx1 = inject_max mx1 in let mn2 = inject_min mn2 in let mx2 = inject_max mx2 in let a = ext_mul mn1 mn2 in let b = ext_mul mn1 mx2 in let c = ext_mul mx1 mn2 in let d = ext_mul mx1 mx2 in let min = ext_min (ext_min a b) (ext_min c d) in let max = ext_max (ext_max a b) (ext_max c d) in (* let multipl1 = Int.pgcd m1 r1 in let multipl2 = Int.pgcd m2 r2 in let modu1 = Int.pgcd m1 m2 in let modu2 = Int.mul multipl1 multipl2 in let modu = Int.ppcm modu1 modu2 in *) let modu = Int.(pgcd (pgcd (mul m1 m2) (mul r1 m2)) (mul r2 m1)) in let r = Int.e_rem (Int.mul r1 r2) modu in (* let t = Top (ext_proj min, ext_proj max, r, modu) in Format.printf "mul. Result: '%a'@\n" pretty t; *) inject_top (ext_proj min) (ext_proj max) r modu | Set s, (Top(_,_,_,_) as t) | (Top(_,_,_,_) as t), Set s -> let l = Array.length s in if l = 0 then bottom else if l = 1 then (* only one element *) scale s.(0) t else mul t (unsafe_make_top_from_array s) in (* Format.printf "mul. result : %a@\n" pretty result;*) result (** Computes [x (op) ({y >= 0} * 2^n)], as an auxiliary function for [shift_left] and [shift_right]. [op] and [scale] must verify [scale a b == op (inject_singleton a) b] *) let shift_aux scale op (x: t) (y: t) = let y = narrow (inject_range (Some Int.zero) None) y in try match y with | Set s -> Array.fold_left (fun acc n -> join acc (scale (Int.two_power n) x)) bottom s | _ -> let min_factor = Extlib.opt_map Int.two_power (min_int y) in let max_factor = Extlib.opt_map Int.two_power (max_int y) in let modu = match min_factor with None -> Int.one | Some m -> m in let factor = inject_top min_factor max_factor Int.zero modu in op x factor with Z.Overflow -> Codex_log.imprecision_warning "Ival.shift_aux"; (* We only preserve the sign of the result *) if is_included x positive_integers then positive_integers else if is_included x negative_integers then negative_integers else top let shift_right x y = shift_aux (scale_div ~pos:true) div x y let shift_left x y = shift_aux scale mul x y let interp_boolean ~contains_zero ~contains_non_zero = match contains_zero, contains_non_zero with | true, true -> zero_or_one | true, false -> zero | false, true -> one | false, false -> bottom module Infty = struct let lt0 = function | None -> true | Some a -> Int.lt a Int.zero let div a b = match a with | None -> None | Some a -> match b with | None -> Some Int.zero | Some b -> Some (Int.e_div a b) let neg = function | Some a -> Some (Int.neg a) | None -> None end let backward_mult_pos_left min_right max_right result = let min_res, max_res = min_and_max result in let min_left = Infty.div min_res (if Infty.lt0 min_res then Some min_right else max_right) and max_left = Infty.div max_res (if Infty.lt0 max_res then max_right else Some min_right) in inject_range min_left max_left let backward_mult_neg_left min_right max_right result = backward_mult_pos_left (Integer.neg max_right) (Infty.neg min_right) (neg_int result) let backward_mult_int_left ~right ~result = match min_and_max right with | None, None -> `Value None | Some a, Some b when a > b -> `Bottom | Some a, Some b when a = Int.zero && b = Int.zero -> if contains_zero result then `Value None else `Bottom | Some a, max when a > Int.zero -> `Value (Some (backward_mult_pos_left a max result)) | Some a, max when a >= Int.zero -> if contains_zero result then `Value None else `Value (Some (backward_mult_pos_left Int.one max result)) | min, Some b when b < Int.zero -> `Value (Some (backward_mult_neg_left min b result)) | min, Some b when b = Int.zero -> if contains_zero result then `Value None else `Value (Some (backward_mult_neg_left min Int.minus_one result)) | min, max -> if contains_zero result then `Value None else `Value (Some (join (backward_mult_pos_left Int.one max result) (backward_mult_neg_left min Int.one result))) let backward_le_int max v = match v with | Float _ -> v | Set _ | Top _ -> narrow v (Top(None,max,Int.zero,Int.one)) let backward_ge_int min v = match v with | Float _ -> v | Set _ | Top _ -> narrow v (Top(min,None,Int.zero,Int.one)) let backward_lt_int max v = backward_le_int (opt1 Int.pred max) v let backward_gt_int min v = backward_ge_int (opt1 Int.succ min) v let diff_if_one value rem = match rem, value with | Set [| v |], Set a -> let index = array_mem v a in if index >= 0 then let l = Array.length a in let pl = pred l in let r = Array.make pl Int.zero in Array.blit a 0 r 0 index; Array.blit a (succ index) r index (pl-index); share_array r pl else value | Set [| v |], Top (Some mn, mx, r, m) when Int.equal v mn -> inject_top (Some (Int.add mn m)) mx r m | Set [| v |], Top (mn, Some mx, r, m) when Int.equal v mx -> inject_top mn (Some (Int.sub mx m)) r m | Set [| v |], Top ((Some mn as min), (Some mx as max), r, m) when Int.equal (Int.sub mx mn) (Int.mul m !small_cardinal_Int) && in_interval v min max r m -> let r = ref mn in Set (Array.init !small_cardinal (fun _ -> let c = !r in let corrected_c = if Int.equal c v then Int.add c m else c in r := Int.add corrected_c m; corrected_c)) | _ -> value (* TODO: more cases: Float *) let diff value rem = log_imprecision "Ival.diff"; diff_if_one value rem (* This function is an iterator, but it needs [diff_if_one] just above. *) let fold_int_bounds f v acc = match v with | Float _ -> f v acc | Set _ | Top _ -> if cardinal_zero_or_one v then f v acc else (* apply [f] to [b] and reduce [v] if [b] is finite, or return [v] and [acc] unchanged *) let on_bound b v acc = match b with | None -> v, acc | Some b -> let b = inject_singleton b in diff_if_one v b, f b acc in let min, max = min_and_max v in (* [v] has cardinal at least 2, so [min] and [max] are distinct *) let v, acc = on_bound min v acc in let v, acc = on_bound max v acc in (* but if the cardinal was 2, then this [v] may be bottom *) if equal v bottom then acc else f v acc let backward_comp_int_left op l r = let open Comp in try match op with | Le -> backward_le_int (max_int r) l | Ge -> backward_ge_int (min_int r) l | Lt -> backward_lt_int (max_int r) l | Gt -> backward_gt_int (min_int r) l | Eq -> narrow l r | Ne -> diff_if_one l r with Error_Bottom (* raised by max_int *) -> bottom let backward_comp_float_left_true op fkind f1 f2 = let f1 = project_float f1 in let f2 = project_float f2 in begin match Fval.backward_comp_left_true op fkind f1 f2 with | `Value f -> inject_float f | `Bottom -> bottom end let backward_comp_float_left_false op fkind f1 f2 = let f1 = project_float f1 in let f2 = project_float f2 in begin match Fval.backward_comp_left_false op fkind f1 f2 with | `Value f -> inject_float f | `Bottom -> bottom end let rec extract_bits ~start ~stop ~size v = match v with | Set s -> inject_ps (Array.fold_left (fun acc elt -> add_ps acc (Int.extract_bits ~start ~stop elt)) empty_ps s) | Float f -> let inject (b, e) = inject_range (Some b) (Some e) in let itvs = if Int.equal size big_int_64 then List.map inject (Fval.bits_of_float64_list f) else if Int.equal size big_int_32 then List.map inject (Fval.bits_of_float32_list f) else (* long double *) [top] in let bits = List.map (extract_bits ~start ~stop ~size) itvs in List.fold_left join bottom bits | Top(_,_,_,_) as d -> try let dived = scale_div ~pos:true (Int.two_power start) d in scale_rem ~pos:true (Int.two_power (Int.length start stop)) dived with Z.Overflow -> Codex_log.imprecision_warning "Ival.extract_bits"; top ;; let all_values ~size v = if Int.lt big_int_64 size then false (* values of this size cannot be enumerated anyway in C. They may occur while initializing large blocks of arrays. *) else match v with | Float _ -> false | Top (None,_,_,modu) | Top (_,None,_,modu) -> Int.is_one modu | Top (Some mn, Some mx,_,modu) -> Int.is_one modu && Int.le (Int.two_power size) (Int.length mn mx) | Set s -> let siz = Int.to_int size in Array.length s >= 1 lsl siz && equal (cast_int_to_int ~size ~signed:false v) (create_all_values ~size:siz ~signed:false) let compare_min_max min max = match min, max with | None,_ -> -1 | _,None -> -1 | Some min, Some max -> Int.compare min max let compare_max_min max min = match max, min with | None,_ -> 1 | _,None -> 1 | Some max, Some min -> Int.compare max min let forward_le_int i1 i2 = if compare_max_min (max_int i1) (min_int i2) <= 0 then Comp.True else if compare_min_max (min_int i1) (max_int i2) > 0 then Comp.False else Comp.Unknown let forward_lt_int i1 i2 = if compare_max_min (max_int i1) (min_int i2) < 0 then Comp.True else if compare_min_max (min_int i1) (max_int i2) >= 0 then Comp.False else Comp.Unknown let forward_eq_int i1 i2 = if cardinal_zero_or_one i1 && equal i1 i2 then Comp.True else if intersects i2 i2 then Comp.Unknown else Comp.False let forward_comp_int op i1 i2 = let open Abstract_interp.Comp in match op with | Le -> forward_le_int i1 i2 | Ge -> forward_le_int i2 i1 | Lt -> forward_lt_int i1 i2 | Gt -> forward_lt_int i2 i1 | Eq -> forward_eq_int i1 i2 | Ne -> inv_truth (forward_eq_int i1 i2) let rehash x = match x with | Set a -> share_array a (Array.length a) | _ -> x (* let scale_int_base factor v = match factor with * | Int_Base.Top -> top * | Int_Base.Value f -> scale f v *) type overflow_float_to_int = | FtI_Ok of Int.t (* Value in range *) | FtI_Overflow of Floating_point.sign (* Overflow in the corresponding direction *) let cast_float_to_int_non_nan ~signed ~size (min, max) = let all = create_all_values ~size ~signed in let min_all = Extlib.the (min_int all) in let max_all = Extlib.the (max_int all) in let conv f = try (* truncate_to_integer returns an integer that fits in a 64 bits integer, but might not fit in [size, sized] *) let i = Floating_point.truncate_to_integer f in if Int.ge i min_all then if Int.le i max_all then FtI_Ok i else FtI_Overflow Floating_point.Pos else FtI_Overflow Floating_point.Neg with Floating_point.Float_Non_representable_as_Int64 sign -> FtI_Overflow sign in let min_int = conv (Fval.F.to_float min) in let max_int = conv (Fval.F.to_float max) in match min_int, max_int with | FtI_Ok min_int, FtI_Ok max_int -> (* no overflow *) inject_range (Some min_int) (Some max_int) | FtI_Overflow Floating_point.Neg, FtI_Ok max_int -> (* one overflow *) inject_range (Some min_all) (Some max_int) | FtI_Ok min_int, FtI_Overflow Floating_point.Pos -> (* one overflow *) inject_range (Some min_int) (Some max_all) (* two overflows *) | FtI_Overflow Floating_point.Neg, FtI_Overflow Floating_point.Pos -> inject_range (Some min_all) (Some max_all) (* Completely out of range *) | FtI_Overflow Floating_point.Pos, FtI_Overflow Floating_point.Pos | FtI_Overflow Floating_point.Neg, FtI_Overflow Floating_point.Neg -> bottom | FtI_Overflow Floating_point.Pos, FtI_Overflow Floating_point.Neg | FtI_Overflow Floating_point.Pos, FtI_Ok _ | FtI_Ok _, FtI_Overflow Floating_point.Neg -> assert false (* impossible if min-max are correct *) let cast_float_to_int ~signed ~size iv = match Fval.min_and_max (project_float iv) with | Some (min, max), _nan -> cast_float_to_int_non_nan ~signed ~size (min, max) | None, _ -> bottom (* means NaN *) (* These are the bounds of the range of integers that can be represented exactly as 64 bits double values *) let double_min_exact_integer = Int.neg (Int.two_power_of_int 53) let double_max_exact_integer = Int.two_power_of_int 53 (* same with 32 bits single values *) let single_min_exact_integer = Int.neg (Int.two_power_of_int 24) let single_max_exact_integer = Int.two_power_of_int 24 (* Same values expressed as double *) let double_min_exact_integer_d = -. (2. ** 53.) let double_max_exact_integer_d = 2. ** 53. let single_min_exact_integer_d = -. (2. ** 24.) let single_max_exact_integer_d = 2. ** 24. (* finds all floating-point values [f] such that casting [f] to an integer type returns [i]. *) let cast_float_to_int_inverse ~single_precision i = let exact_min, exact_max = if single_precision then single_min_exact_integer, single_max_exact_integer else double_min_exact_integer, double_max_exact_integer in let fkind = if single_precision then Fval.Single else Fval.Double in match min_and_max i with | Some min, Some max when Int.lt exact_min min && Int.lt max exact_max -> let minf = if Int.le min Int.zero then (* min is negative. We want to return [(float)((real)(min-1)+epsilon)], as converting this number to int will truncate all the fractional part (C99 6.3.1.4). Given [exact_min] and [exact_max], 1ulp is at most 1 here, so adding 1ulp will at most cancel the -1. Hence, we can use [next_float]. *) (* This float is finite because min is small enough *) Fval.F.next_float fkind (Int.to_float (Int.pred min)) else (* min is positive. Since casting truncates towards 0, [(int)((real)min-epsilon)] would return [min-1]. Hence, we can simply return the float corresponding to [min] -- which can be represented precisely given [exact_min] and [exact_max]. *) Int.to_float min in (* All operations are dual w.r.t. the min bound. *) let maxf = if Int.le Int.zero max then (* This float is finite because max is big enough *) Fval.F.prev_float fkind (Int.to_float (Int.succ max)) else Int.to_float max in assert (Fval.F.is_finite (Fval.F.of_float minf)); assert (Fval.F.is_finite (Fval.F.of_float maxf)); Float (Fval.inject fkind (Fval.F.of_float minf) (Fval.F.of_float maxf)) | _ -> if single_precision then top_single_precision_float else top_float let cast_int_to_float_inverse_not_nan ~single_precision (min, max) = (* We restrict ourselves to (min,max) \in [exact_min, exact_max]. Outside of this range, the conversion int -> float is not exact, and the operation is more involved. *) let exact_min, exact_max = if single_precision then single_min_exact_integer_d, single_max_exact_integer_d else double_min_exact_integer_d, double_max_exact_integer_d in (* We find the integer range included in [f] *) let min = Fval.F.to_float min in let max = Fval.F.to_float max in if exact_min <= min && max <= exact_max then (* Round to integers in the proper direction: discard the non-floating-point values on each extremity. *) let min = ceil min in let max = floor max in let conv f = try Some (Integer.of_float f) with Z.Overflow -> None in let r = inject_range (conv min) (conv max) in (* Kernel.result "Cast I->F inv: %a -> %a@." pretty f pretty r; *) r else top (* Approximate *) let cast_int_to_float_inverse ~single_precision f = match min_and_max_float f with | None, _ -> (* NaN *) bottom (* a cast of NaN to int is fully undefined *) | Some (min, max), _ (*we can disregard the NaN boolean for the same reason *) -> cast_int_to_float_inverse_not_nan ~single_precision (min, max) let of_int i = inject_singleton (Int.of_int i) let of_int64 i = inject_singleton (Int.of_int64 i) (* This function always succeeds without alarms for C integers, because they always fit within a float32. *) let cast_int_to_float fkind v = let min,max = min_and_max v in inject_float (Fval.cast_int_to_float fkind min max) let reinterpret_as_float kind i = match i with | Float _ -> i | Set _ when is_zero i || is_bottom i -> i | Top _ | Set _ -> (* Reinterpret a range of integers as a range of floats. Float are ordered this way : if [min_i], [max_i] are the bounds of the signed integer type that has the same number of bits as the floating point type, and [min_f] [max_f] are the integer representation of the most negative and most positive finite float of the type, and < is signed integer comparison, we have: min_i < min_f < min_f+1 < -1 < 0 < max_f < max_f+1 < max_i | | | | | | | | --finite-- -not finite- -finite- -not finite- | | |<---------> | | |<---------> -0. -max -inf NaNs +0. max inf NaNs The float are of the same sign as the integer they convert into. Furthermore, the conversion function is increasing on the positive interval, and decreasing on the negative one. *) let reinterpret size kind conv min_f max_f = let size = Integer.of_int size in let i = cast_int_to_int ~size ~signed:true i in (* Intersect [i'] with [i], and return the (finite) bounds directly. *) let bounds_narrow i' = let r = narrow i i' in if is_bottom r then `Bottom else match min_and_max r with | None, _ | _, None -> assert false (* i is finite thanks to cast *) | Some b, Some e -> `Value (b, e) in let s_max_f = Int.succ max_f (* neg inf *) in let s_min_f = Int.succ min_f (* pos inf *) in let s_s_max_f = Int.succ s_max_f (* first 'positive' NaN *) in let s_s_min_f = Int.succ s_min_f (* first 'negative' NaN *) in (* positive floats *) let f_pos = inject_range (Some Integer.zero) (Some s_max_f) in (* negative floats *) let f_neg = inject_range None (Some s_min_f) in (* 'positive' NaNs *) let nan_pos = inject_range (Some s_s_max_f) None in (* 'negative' NaNs *) let nan_neg = inject_range (Some s_s_min_f) (Some Int.minus_one) in let nan = (* at least one NaN somewhere *) if intersects i nan_pos || intersects i nan_neg then [`Value Fval.nan] else [] in let open Bottom in let range mn mx = Fval.inject kind (conv mn) (conv mx) in (* convert positive floats; increasing on positive range *) let pos = bounds_narrow f_pos >>-: (fun (b, e) -> range b e) in (* convert negative floats; decreasing on negative range *) let neg = bounds_narrow f_neg >>-: (fun (b, e) -> range e b) in let f = Bottom.join_list Fval.join (pos :: neg :: nan) in inject_float (Bottom.non_bottom f) in let open Floating_point in match kind with (* | Cil_types.FDouble -> * let conv v = Fval.F.of_float (Int64.float_of_bits (Int.to_int64 v)) in * reinterpret * 64 Fval.Double conv bits_of_most_negative_double bits_of_max_double * | Cil_types.FFloat -> * let conv v = Fval.F.of_float(Int32.float_of_bits (Int.to_int32 v)) in * reinterpret * 32 Fval.Single conv bits_of_most_negative_float bits_of_max_float * | Cil_types.FLongDouble -> *) (* currently always imprecise *) | _ -> top_float let overlaps ~partial ~size t1 t2 = let diff = sub_int t1 t2 in match diff with | Set array -> not (array_for_all (fun i -> Int.ge (Int.abs i) size || (partial && Int.is_zero i)) array) | Top (min, max, _r, _modu) -> let pred_size = Int.pred size in min_le_elt min pred_size && max_ge_elt max (Int.neg pred_size) | Float _ -> assert false (* ------------------------------------------------------------------------ *) (* --- Bitwise operators --- *) (* ------------------------------------------------------------------------ *) (* --- Bit lattice --- *) type bit_value = On | Off | Both module Bit = struct type t = bit_value let to_string = function | Off -> "0" | On -> "1" | Both -> "T" let _pretty (fmt : Format.formatter) (b :t) = Format.pp_print_string fmt (to_string b) let union (b1 : t) (b2 : t) : t = if b1 = b2 then b1 else Both let not : t -> t = function | On -> Off | Off -> On | Both -> Both end (* --- Bit operators --- *) module type BitOperator = sig (* Printable version of the operator *) val representation : string (* forward is given here as the lifted function of some bit operator op where op 1. is assumed to be commutative (backward functions do not assume the position of the arguments) 2. must ensure 0 op 0 = 0 as otherwise applying op on a sign bit may produce a negative result from two positive operands; but we don't want to produce a negative result when the operation is unsigned which we don't know unless one of the operands is negative; 3. is not constant, otherwise nothing of all of this makes sense. forward is defined as forward b1 b2 = { x1 op x2 | x1 \in b1, x2 \in b2 } *) val forward : bit_value -> bit_value -> bit_value (* backward_off b = { x | \exist y \in b . x op y = y op x = 1 } *) val backward_off : bit_value -> bit_value (* backward_on b = { x | \exist y \in b . x op y = y op x = 0 } *) val backward_on : bit_value -> bit_value end module And : BitOperator = struct let representation = "&" let forward v1 v2 = match v1 with | Off -> Off | On -> v2 | Both -> if v2 = Off then Off else Both let backward_off = function | (Off | Both) -> Both | On -> Off let backward_on = function | Off -> assert false | (On | Both) -> On end module Or : BitOperator = struct let representation = "|" let forward v1 v2 = match v1 with | On -> On | Off -> v2 | Both -> if v2 = On then On else Both let backward_off = function | On -> assert false | (Off | Both) -> Off let backward_on = function | (On | Both) -> Both | Off -> On end module Xor : BitOperator = struct let representation = "^" let forward v1 v2 = match v1 with | Both -> Both | Off -> v2 | On -> Bit.not v2 let backward_on v = Bit.not v let backward_off v = v end (* --- Bit extraction and mutation --- *) let significant_bits (v : t) : int option = match min_and_max v with | None, _ | _, None -> None | Some l, Some u -> Some (max (Z.numbits l) (Z.numbits u)) let extract_sign (v : t) : bit_value = match min_and_max v with | _, Some u when Int.(lt u zero) -> On | Some l, _ when Int.(ge l zero) -> Off | _, _ -> Both let extract_bit (i : int) (v : t) : bit_value = let bit_value x = if Z.testbit x i then On else Off in match v with | Float _ -> Both | Set s -> array_map_reduce bit_value Bit.union s | Top (None, _, _r, _m) | Top (_, None, _r, _m) -> Both | Top (Some l, Some u, _r, _m) -> (* It does not take modulo into account *) if Int.(ge (sub u l) (two_power_of_int i)) (* u - l >= mask *) then Both else Bit.union (bit_value l) (bit_value u) let reduce_sign (v : t) (b : bit_value) : t = match b with | Both -> v | On -> begin match v with | Float _ -> v | Set s -> array_filter Int.(gt zero) s | Top (_l, Some u, _r, _modu) when Int.(lt u zero) -> v | Top (l, _u, r, modu) -> let u = Some Int.(round_down_to_r ~max:minus_one ~r ~modu) in inject_top l u r modu end | Off -> begin match v with | Float _ -> v | Set s -> array_filter Int.(le zero) s | Top (Some l, _u, _r, _modu) when Int.(ge l zero) -> v | Top (_l, u, r, modu) -> let l = Some Int.(round_up_to_r ~min:zero ~r ~modu) in inject_top l u r modu end let reduce_bit (i : int) (v : t) (b : bit_value) : t = let bit_value x = if Z.testbit x i then On else Off in if b = Both then v else match v with | Float _ -> v | Set s -> array_filter (fun x -> bit_value x = b) s | Top (l, u, r, modu) -> let power = Int.(two_power_of_int i) in (* 001000 *) let mask = Int.(pred (two_power_of_int (i+1))) in (* 001111 *) (* Reduce bounds to the nearest satisfying bound *) let l' = match l with | Some l when bit_value l <> b -> let min = match b with | On -> Int.(logor (logand l (lognot mask)) power) (* ll1000 *) | Off -> Int.(succ (logor l mask)) (* ll1111 + 1 *) | Both -> assert false in Some (Int.round_up_to_r ~min ~r ~modu) | _ -> l and u' = match u with | Some u when bit_value u <> b -> let max = match b with | On -> Int.(pred (logand u (lognot mask))) (* uu0000 - 1 *) | Off -> Int.(logand (logor u mask) (lognot power)) (* uu0111 *) | Both -> assert false in Some (Int.round_down_to_r ~max ~r ~modu) | _ -> u in inject_top l' u' r modu type bit = Sign | Bit of int let extract_bit = function | Sign -> extract_sign | Bit i -> extract_bit i let set_bit_on ~size bit = let mask = match bit with | Sign -> Int.(neg (two_power_of_int size)) | Bit i -> Int.(two_power_of_int i) in fun v -> Int.logor mask v let reduce_bit = function | Sign -> reduce_sign | Bit i -> reduce_bit i (* --- Bitwise binary operators --- *) module BitwiseOperator (Op : BitOperator) = struct let backward (b : bit_value) = function | On -> Op.backward_on b | Off -> Op.backward_off b | Both -> assert false (** Bit masks are composed of an array of significant bit values where index 0 represents the lowest bit, and a single bit_value to represent the possible leading bits. *) type bit_mask = bit_value array * bit_value (* Converts an integer [x] into a bit array of size [n]. *) let int_to_bit_array n (x : Int.t) = let make i = if Z.testbit x i then On else Off in Array.init n make (* Computes a bit_mask for the lowest bits of an ival, using the modulo information for non singleton values. *) let low_bit_mask : t -> bit_mask = function | Set [| |] -> raise Error_Bottom | Set [| x |] -> (* singleton : build a full mask *) let n = Z.numbits x in int_to_bit_array n x, if Int.(ge x zero) then Off else On | v -> let _,_,r,modu = min_max_r_mod v in (* requires cardinal > 1 *) (* Find how much [modu] can be divided by two. *) let n = Z.trailing_zeros modu in int_to_bit_array n r, Both (* Computes a remainder and modulo for the result of [v1 op v2]. *) let compute_modulo v1 v2 = let b1, s1 = low_bit_mask v1 and b2, s2 = low_bit_mask v2 in let size = max (Array.length b1) (Array.length b2) in (* Sets the [i] nth bits of [rem] until an uncertainty appears. *) let rec step i rem = let b1 = try b1.(i) with _ -> s1 and b2 = try b2.(i) with _ -> s2 in let b = Op.forward b1 b2 in if i >= size || b = Both then rem, Int.two_power_of_int i else (* [rem] starts at 0, so we only need to turn on the 1 bits. *) let rem = if b = On then set_bit_on ~size (Bit i) rem else rem in step (i+1) rem in step 0 Int.zero (* The number of bits on which the result should be significant *) let result_size (v1 : t) (v2 : t) : int option = let n1 = significant_bits v1 and n2 = significant_bits v2 in let n1_greater = match n1, n2 with | None, _ -> true | _, None -> false | Some n1, Some n2 -> n1 >= n2 in (* whether n1 or n2 is greater, look if the sign bit oped with anything is not constant. If it is constant, then the highest bits are irrelevant. *) if n1_greater then if Op.forward Both (extract_sign v2) = Both then n1 else n2 else if Op.forward (extract_sign v1) Both = Both then n2 else n1 exception Do_not_fit_small_sets (* Try to build a small set. It is basically enumerating the possible results, by choosing the possible bits from left to right. This function aborts if it ever exceeds the small set size. The algorithm is probably not complete, as it is not always possible to reduce the operands leading to a result (without an exponential cost) meaning that sometimes small sets can be obtained but the algorithm will fail to find them. *) let compute_small_set ~size (v1 : t) (v2 : t) (r : Int.t) (modu : Int.t) = let set_bit i acc (r, v1, v2) = let b1 = extract_bit i v1 and b2 = extract_bit i v2 in match Op.forward b1 b2 with | On -> (set_bit_on ~size i r, v1, v2) :: acc | Off -> (r, v1, v2) :: acc | Both -> let v1_off = reduce_bit i v1 (Op.backward_off b2) and v2_off = reduce_bit i v2 (Op.backward_off b1) in let v1_on = reduce_bit i v1 (Op.backward_on b2) and v2_on = reduce_bit i v2 (Op.backward_on b1) in (set_bit_on ~size i r, v1_on, v2_on) :: (r, v1_off, v2_off) :: acc in let acc = ref (set_bit Sign [] (r, v1, v2)) in for i = size - 1 downto Z.numbits modu - 1 do acc := List.fold_left (set_bit (Bit i)) [] !acc; if List.length !acc > !small_cardinal then raise Do_not_fit_small_sets done; let o = List.fold_left (fun o (r,_,_) -> O.add r o) O.empty !acc in share_set o (O.cardinal o) (* If lower is true (resp. false), compute the lower (resp. upper) bound of the result interval when applying the bitwise operator to [v1] and [v2]. [size] is the number of bits of the result. This function should be exact when the operands are small sets or tops with modulo 1. Otherwise, it is an overapproximation of the bound. *) let compute_bound ~size v1 v2 lower = (* Sets the [i]-nth bit of the currently computed bound [r] of [v1 op v2]. If possible, reduces [v1] and [v2] accordingly. *) let set_bit i (r, v1, v2) = let b1 = extract_bit i v1 and b2 = extract_bit i v2 in let b, v1, v2 = match Op.forward b1 b2 with | On | Off as b -> b, v1, v2 (* Constant bit, no reduction. *) | Both -> (* Choose the best bit for the searched bound, and reduces [v1] and [v2] accordingly. *) let b = match i with | Sign -> if lower then On else Off | Bit _ -> if lower then Off else On in let v1 = reduce_bit i v1 (backward b2 b) and v2 = reduce_bit i v2 (backward b1 b) in b, v1, v2 in (* Only sets 1 bit, as [r] is 0 at the beginning. *) let r = if b = On then set_bit_on ~size i r else r in r, v1, v2 in (* The result is 0 at the beginning, and [set_bit] turns on the 1 bits. *) let r = ref (Int.zero, v1, v2) in (* Sets the sign bit, and then the bits from size to 0. *) r := set_bit Sign !r; for i = (size - 1) downto 0 do r := set_bit (Bit i) !r; done; let bound, _v1, _v2 = !r in bound let bitwise_forward (v1 : t) (v2 : t) : t = try let r, modu = compute_modulo v1 v2 in match result_size v1 v2 with | None -> (* We could do better here, as one of the bound may be finite. However, this case should occur rarely or not at all. *) inject_interval None None r modu | Some size -> try compute_small_set ~size v1 v2 r modu with Do_not_fit_small_sets -> let min = compute_bound ~size v1 v2 true and max = compute_bound ~size v1 v2 false in inject_interval (Some min) (Some max) r modu with Error_Bottom -> bottom end let bitwise_or = let module M = BitwiseOperator (Or) in M.bitwise_forward let bitwise_and = let module M = BitwiseOperator (And) in M.bitwise_forward let bitwise_xor = let module M = BitwiseOperator (Xor) in M.bitwise_forward (* --- Bitwise not --- *) let bitwise_signed_not v = match v with | Float _ -> assert false | Top _ -> add_int (neg_int v) minus_one (* [-v - 1] *) | Set s -> map_set_strict_decr Int.lognot s let bitwise_unsigned_not ~size v = let size = Int.of_int size in cast_int_to_int ~size ~signed:false (bitwise_signed_not v) let bitwise_not ~size ~signed v = if signed then bitwise_signed_not v else bitwise_unsigned_not ~size v let pretty_debug = pretty let name = "ival" (* Local Variables: compile-command: "make -C ../../.." End: *)
sectionYPositions = computeSectionYPositions($el), 10)"
x-init="setTimeout(() => sectionYPositions = computeSectionYPositions($el), 10)"
>