gf-complete.html 159 KB
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
<html>

<head>

<link rel="stylesheet" type="text/css" href="style.css">

</head>

<body>

<div id="box">

<h1>
GF-Complete: A Comprehensive Open Source Library for Galois </br>
Field Arithmetic
</h1>

<h1> Version 1.02  </h1>

<h4>James S. Plank* &nbsp&nbsp&nbsp&nbsp&nbsp&nbsp Ethan L. Miller 
Kevin M. Greenan &nbsp&nbsp&nbsp&nbsp&nbsp&nbsp Benjamin A. Arnold<br>
John A. Burnum &nbsp&nbsp&nbsp&nbsp&nbsp&nbsp Adam W. Disney  &nbsp&nbsp&nbsp&nbsp&nbsp&nbsp
Allen C. McBride

</h4> <br>



<a href="">

https://bitbucket.org/jimplank/gf-complete

 </a><br><br>
<a href=""> 
http://web.eecs.utk.edu/~plank/plank/papers/GF-Complete-Manual-1.02.pdf


 </a> <br> <br> 







</div>


<div id="pages_paragraphs_2">

This is a user's manual for GF-Complete, version 1.02. This release supersedes version 0.1 and represents the first
major release of GF-Complete. To our knowledge, this library implements every Galois Field multiplication technique
applicable to erasure coding for storage, which is why we named it GF-Complete. The primary goal of this library is
to allow storage system researchers and implementors to utilize very fast Galois Field arithmetic for Reed-Solomon
coding and the like in their storage installations. The secondary goal is to allow those who want to explore different
ways to perform Galois Field arithmetic to be able to do so effectively.


<p>
If you wish to cite GF-Complete, please cite technical report UT-CS-13-716: [PMG<sup>+</sup>13].

</p>


<h2>If You Use This Library or Document </h2>



Please send me an email to let me know how it goes. Or send me an email just to let me know you are using the
library. One of the ways in which we are evaluated both internally and externally is by the impact of our work, and if
you have found this library and/or this document useful, we would like to be able to document it. Please send mail to
<em>plank@cs.utk.edu.</em> Please send bug reports to that address as well.



<p>
The library itself is protected by the New BSD License. It is free to use and modify within the bounds of this
license. To the authors' knowledge, none of the techniques implemented in this library have been patented, and the
authors are not pursing patents. </p> <br>

 </div>
<div id="footer"> 
 
<span id="footer_bar">&nbsp&nbsp&nbsp&nbsp.*plank@cs.utk.edu (University of Tennessee), el  </span> <em>m@cs.ucsc.edu </em>(UC Santa Cruz), <em>kmgreen2@gmail.com </em> (Box). This material
is based upon work supported by the National Science Foundation under grants CNS-0917396, IIP-0934401 and CSR-1016636, plus REU supplements
CNS-1034216, CSR-1128847 and CSR-1246277. Thanks to Jens Gregor for helping us wade through compilation issues, and for Will
Houston for his initial work on this library.

</div>

<b>Finding the Code </b>
<br><br>
This code is actively maintained on bitbucket:<a href=""> https://bitbucket.org/jimplank/gf-complete. </a> There are
previous versions on my UTK site as a technical report; however, that it too hard to maintain, so the main version is
on bitbucket.<br><br>


<b>Two Related Papers </b> <br><br>

This software acccompanies a large paper that describes these implementation techniques in detail [PGM13a]. We
will refer to this as <em> "The Paper." </em> You do not have to read The Paper to use the software. However, if you want to
start exploring the various implementations, then The Paper is where you'll want to go to learn about the techniques
in detail.



<p>This library implements the techniques described in the paper "Screaming Fast Galois Field Arithmetic Using Intel
SIMD Instructions," [PGM13b]. The Paper describes all of those techniques as well.
</p><br><br>

<b>If You Would Like HelpWith the Software </b><br><br>

Please contact the first author of this manual.<br><br>

<b>Changes from Revision 1.01</b>
<br><br>
The major change is that we are using autoconf to aid with compilation, thus obviating the need for the old <b>flag_tester</b>
code. Additionally, we have added a quick timing tool, and we have modified <b>gf_methods</b> so that it may be used to
run the timing tool and the unit tester.


















<br/>
CONTENT  <span class="aligning_page_number"> 3 </span> 
<h2>Contents </h2>
<div class="index">
1 <span class="aligning_numbers">Introduction </span> <span class="aligning_page_number"> 5 </span>
  <br><br> 
2 <span class="aligning_numbers">Files in the Library </span>	<span class="aligning_page_number"> 6  </span>  <br> </div>

<div class="sub_indices">
2.1 Header files in the directory <b>"include"</b>  . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . <span class="aligning_page_number"> 6 </span>  <br>
2.2 Source files in the <b>"src"</b> directory . . . . . . . . . . . . . . . . . . . . . . .  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  . . . . . . . . . . . . . . . . .<span class="aligning_page_number">   7  </span> <br>
2.3 Library tools files in the <b>"tools"</b> directory  . . . . . . . . . . ..  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  . . . . . . . . . . . . . . . . . .   <span class="aligning_page_number"> 7   </span> <br>
2.4 The unit tester in the <b>"test"</b> directory. . . . . . . . . . . . . . . . . .  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  . . . . . . . . . . . . . . . . <span class="aligning_page_number">  8  </span>  <br>
2.5 Example programs in the <b>"examples"</b> directory . . . .  . . . . . . . . . . . . . . . .  . . . . . . . . . . . . . . . . . . . . .  . . . . . . . . . . . . . . . . . .<span class="aligning_page_number"> 8  </span> 

</div>
<br>
<div class="index">

3 <span class="aligning_numbers">Compilation </span><span class="aligning_page_number">  8 </span>  <br> <br>
4 <span class="aligning_numbers">Some Tools and Examples to Get You Started </span><span class="aligning_page_number">  8 </span> <br><br>  </div> 



<div class="sub_indices">
4.1 Three Simple Command Line Tools: gf mult, gf div and gf add . . . . . . . . . .  . . . .  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . <span class="aligning_page_number"> 8</span>  <br>
4.2 Quick Starting Example #1: Simple multiplication and division . .  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  . . . . . . . . . . . . . . . <span class="aligning_page_number">   9  </span> <br>
4.3 Quick Starting Example #2: Multiplying a region by a constant    . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  . . . . . . . . . . . . . . .  <span class="aligning_page_number"> 10   </span> <br>
4.4 Quick Starting Example #3: Using w = 64 . . . . . . . . . . . . .  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  . . . . . . . . . . . . . . . . <span class="aligning_page_number">  11  </span>  <br>
4.5 Quick Starting Example #4: Using w = 128. . . .  . . . . . . . . . . . . . . . .  . . . . . . . . . . . . . . . . . . . . .  . . . . . . . . . . . . . . . . . . . . . <span class="aligning_page_number"> 11  </span> 
</div>
<br>


<div class="index">
5 <span class="aligning_numbers"> Important Information on Alignment when Multiplying Regions </span><span class="aligning_page_number"> 12</span> <br><br>

6 <span class="aligning_numbers"> The Defaults</span><span class="aligning_page_number"> 13 </span> <br>

</div>

<div class="sub_indices">
6.1 Changing the Defaults . .  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  . . . . . . . .<span class="aligning_page_number">   14  </span> <br>


<ul style="list-style-type:none;">
<li>6.1.1 Changing the Components of a Galois Field with <b> create_gf_from_argv() </b>   . . . . . . . . . . . . . . . .  . . . . . . . . . . . . . . . .  <span class="aligning_page_number"> 15   </span> <br>
</li>
<li>
6.1.2 Changing the Polynomial. . . . . . . . . . . . .  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  . . . . . . . . . . . . . . . . . . . . . . . . <span class="aligning_page_number">  16  </span>  <br>
</li>
<li>
6.1.3 Changing the Multiplication Technique. . . .  . . . . . . . . . . . . . . . .  . . . . . . . . . . . . . . . . . . . . .  . . . . . . . . . . . . . . . . . . .<span class="aligning_page_number"> 17  </span> 
</li>


<li>
6.1.4 Changing the Division Technique . . . . . .  . . . . . . . . . . . . . . . . . . . . .  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  <span class="aligning_page_number"> 19  </span> 
</li>


<li>
6.1.5 Changing the Region Technique. . . .  . . . . . . . . . . . . . . . .  . . . . . . . . . . . . . . . . . . . . .  . . . . . . . . . . . . . . . . . . . .  . . . ..<span class="aligning_page_number"> 19  </span> 
</li>
</ul>
6.2 Determining Supported Techniques with <b>gf_methods</b> . . . . . . . . . .  . . . .  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . <span class="aligning_page_number"> 20</span>  <br>

6.3 Testing with <b>gf_unit, gf_time,</b> and <b>time_tool.sh </b>. . . . . . . . . .  . . . .  . . . . . . . . . . . . . . . . . . . . . . . . . . .  . . . . . . . . . . . . . . .  . . <span class="aligning_page_number"> 21</span>

<ul style="list-style-type:none;">
<li>
6.3.1 <b>time_tool.sh</b> . . . . . .  . . . . . . . . . . . . . . . . . . . . .  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . .  <span class="aligning_page_number"> 22 </span> 
</li>

<li>
6.3.2 An example of <b>gf_methods</b> and <b>time_tool.sh</b> . . . . . .  . . . . . . . . . . . . . . . . . . . . .  . . . . . . . . . . . . . . . .. . . . . . . .  . .. .  .<span class="aligning_page_number"> 23  </span> 
</li>

</ul>

6.4 Calling <b>gf_init_hard()</b> . . . . . . . . . .  . . . .  . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . .. . . . . . . . .. . . . . . . .  .. . . . . . . .  . . .  <span class="aligning_page_number"> 24</span>  <br>

6.5 <b>gf_size()</b> . . . . . . . . . .  . . . .  . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . .  .. . . . . . . . .. . . . . . . .  .. . . . . . . . .. . . . . . . . . . ..  .  <span class="aligning_page_number"> 26</span>  <br><br>
</div>


<div class="index">
8 <span class="aligning_numbers">  Further Information on Options and Algorithms </span><span class="aligning_page_number">   26 </span> </div> <br><br> </div>
<div class="sub_indices">
7.1 Inlining Single Multiplication and Division for Speed   . . . . . . . . . . . . . . .  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  . . . <span class="aligning_page_number"> 26 </span> <br>
7.2 Using different techniques for single and region multiplication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  .  <span class="aligning_page_number"> 27 </span> <br>
7.3 General <em>w</em> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  . . . . . . . . . . . . . .  . <span class="aligning_page_number"> 28  </span><br>

7.4 Arguments to <b>"SPLIT"</b> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  . . . . . . . . . . . . . . . . . <span class="aligning_page_number"> 28</span>  <br>
7.5 Arguments to <b>"GROUP"</b> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  . . . . .  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  <span class="aligning_page_number">29 </span> <br>
7.6 Considerations with <b>"COMPOSITE"</b> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  . . . . .  . . . . . . . . . . . . . . . . . . . . . . . . . . . .  .  <span class="aligning_page_number">30 </span> <br>
7.7 <b>"CARRY FREE"</b> and the Primitive Polynomial  . . . . . . . . . . . . . . .  . . . . .  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  . .   <span class="aligning_page_number">31 </span> <br>
7.8 More on Primitive Polynomials . .  . . . . . . . . . . . . . . . . . . .  . . . . .  . . . . . . . . . . . . . . . . . . . . . . . . . . . .  . . . . . . . ..   . . . . . . . . .  <span class="aligning_page_number">31 </span> <br>


<ul style="list-style-type:none;">
<li>
7.8.1 Primitive Polynomials that are not Primitive . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  . . . . . . . . . . . . . . . . . . .  . . .  <span class="aligning_page_number"> 31</span>  <br>

</li>
<li>7.8.2 Default Polynomials for Composite Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  . . . . . . . . . . . . . . . . . . . . . . .  <span class="aligning_page_number"> 32</span>  <br>

</li>
</ul>

</div>











<br/>
CONTENT  <span class="aligning_page_number"> 4 </span> 

<div class="sub_indices">
<ul style="list-style-type:none">
<li> 7.8.3 The Program <b>gf_poly</b> for Verifying Irreducibility of Polynomials </span><span class="aligning_page_number">  33 </span> 
</li>
</ul>


7.9<span class="aligning_numbers"><b>"ALTMAP"</b> considerations and <b>extract_word()</b> </span><span class="aligning_page_number">  34 </span>  
<ul style="list-style-type:none">
<li>

7.9.1 Alternate mappings with <b>"SPLIT"</b> . . . . . . . . . .  . . . .  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .<span class="aligning_page_number"> 34</span>  <br>
</li>
<li>
7.9.2 Alternate mappings with <b>"COMPOSITE"</b> . .  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  . . . . . . . . . . . . . . . . . . . . . . . . <span class="aligning_page_number">   36  </span> <br>
</li>
<li>
7.9.3 The mapping of <b>"CAUCHY"</b>    . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  . . . . . . . . . . . . . . . . . .. . . . . .. . . . . . . . . .  . ..  <span class="aligning_page_number"> 37   </span> <br>
</li>
</ul>
</div>


8 <span class="aligning_numbers"><b>Thread Safety </b></span><span class="aligning_page_number">  37 </span> <br><br>  </div> 

9 <span class="aligning_numbers"><b>Listing of Procedures</b> </span><span class="aligning_page_number">  37 </span> <br><br>  </div> 

10 <span class="aligning_numbers"><b>Troubleshooting</b> </span><span class="aligning_page_number">  38 </span> <br><br>  </div> 
11 <span class="aligning_numbers"><b>Timings</b> </span><span class="aligning_page_number">  41 </span> <br><br>  </div> 

<div class="sub_indices">
11.1 Multiply() . . . . . . . . . .  . . . .  . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . .. . . . . . . . .. . . . . . . . .. . . . . . . . . . . . .  . . . .. . . . <span class="aligning_page_number"> 42</span>  <br>
11.2 Divide() . .  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  . . . . . . . . . . . . . . .. . . . . . . . .. . . . . . . . .. . . . . . . . .. . . . . . . . . . . .. . . . . <span class="aligning_page_number">   42  </span> <br>
11.3 Multiply Region()    . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  . . . . . . . . . . . . . . .. . . . . . . . .. . . . . . . . .. . . . . . . . . . . . .  . . . . .  <span class="aligning_page_number"> 43   </span> <br>
</div>






<br/>
INTRODUCTION  <span class="aligning_page_number"> 5 </span> 


<h3>1 Introduction </h3>

Galois Field arithmetic forms the backbone of erasure-coded storage systems, most famously the Reed-Solomon
erasure code. A Galois Field is defined over w-bit words and is termed <em>GF(2<sup>w</sup>).</em> As such, the elements of a Galois
Field are the integers 0, 1, . . ., 2<sup>w</sup> - 1. Galois Field arithmetic defines addition and multiplication over these closed
sets of integers in such a way that they work as you would hope they would work. Specifically, every number has a
unique multiplicative inverse. Moreover, there is a value, typically the value 2, which has the property that you can
enumerate all of the non-zero elements of the field by taking that value to successively higher powers.


<p>Addition in a Galois Field is equal to the bitwise exclusive-or operation. That's nice and convenient. Multiplication
is a little more complex, and there are many, many ways to implement it. The Paper describes them all, and the
following references providemore supporting material: [Anv09, GMS08, LHy08, LD00, LBOX12, Pla97]. The intent
of this library is to implement all of the techniques. That way, their performancemay be compared, and their tradeoffs
may be analyzed. <p>




<ol>

When used for erasure codes, there are typically five important operations:<br>
<li> <b>Adding two numbers in </b> GF(2<sup>w</sup>). That's bitwise exclusive-or. </li>
<li> <b>Multiplying two numbers in</b> GF(2<sup>w</sup>). Erasure codes are usually based on matrices in GF(2<sup>w</sup>), and constructing
these matrices requires both addition and multiplication.</li>
<li> <b>Dividing two numbers in </b>GF(2<sup>w</sup>). Sometimes you need to divide to construct matrices (for example, Cauchy
Reed-Solomon codes [BKK<sup>+</sup>95, Rab89]). More often, though, you use division to invert matrices for decoding.
Sometimes it is easier to find a number's inverse than it is to divide. In that case, you can divide by multiplying
by an inverse. </li>

<li><b>adding two regions of numbers in</b> GF(2<sup>w</sup>), which will be explained along with... </li>
<li> <b>Mutiplying a region of numbers in </b>GF(2<sup>w</sup>) by a constant in GF(2<sup>w</sup>). Erasure coding typically boils down
to performing dot products in GF(2<sup>w</sup>). For example, you may define a coding disk using the equation: </li><br>




<center>c<em><sub>0</sub></em>= d<em><sub>0</sub></em> + 2d<em><sub>1</sub></em> + 4d<em><sub>2</sub></em> + 8d<em><sub>3</sub></em>.</sup> </center><br>

That looks like three multiplications and three additions However, the way ' implemented in a disk system
looks as in Figure 1. Large regions of disks are partitioned into w-bit words in GF(2<sup>w</sup>). In the example, let us
suppose that <em>w</em> = 8, and therefore that words are bytes. Then the regions pictured are 1 KB from each disk.
The bytes on disk Di are labeled d<sub>i,0,</sub> d<sub>i,1, . . . ,</sub> d<sub>i,1023,</sub> and the equation above is replicated 1024 times. For
0 &#8804 j < 1024:
<br><br>
<center>c<em><sub>0,j</sub></em> = d<em><sub>0,j</sub></em> + 2d<em><sub>1,j</sub></em> + 4d<em><sub>2,j</sub></em> + 8d<em><sub>3,j</sub></em> . </center>
<br>


While it's possible to implement each of these 1024 equations independently, using the single multiplication
and addition operations above, it is often much more efficient to aggregate. For example, most computer architectures
support bitwise exclusive-or of 64 and 128 bit words. Thus, it makes much more sense to add regions
of numbers in 64 or 128 bit chunks rather than as words in GF(2<sup>w</sup>). Multiplying a region by a constant can
leverage similar optimizations. </ol>


<p>GF-Complete supports multiplication and division of single values for all values of <em>w</em> &#8804 32, plus <em>w</em> = 64 and <em>w</em> =
128. It also supports adding two regions of memory (for any value of <em>w</em>, since addition equals XOR), and multiplying
a region by a constant in <em>GF(2<sup>4</sup>), GF(2<sup>8</sup>), GF(2<sup>16</sup>), GF(2<sup>32</sup>), GF(2<sup>64</sup>) and GF(2<sup>128</sup>).</em> These values are chosen
because words in GF(2<sup>w</sup>) fit into machine words with these values of <em>w.</em> Other values of w don't lend themselves
to efficient multiplication of regions by constants (although see the <b>"CAUCHY"</b> option in section 6.1.5 for a way to
multiply regions for other values of <em>w</em>).</p>






<br/>

2 &nbsp &nbsp  <em>  FILES IN THE LIBRARY     </em>   <span id="index_number">6  </span> <br><br><br>



<div class="image-cell_1"> </div>  <br><br><br>

Figure 1: An example of adding two regions of numbers, and multiplying a region of numbers by a constant
in <em>GF(2<sup>w</sup>) </em>. In this example, <em>w</em> = 8, and each disk is holding a 1KB region. The same coding equation -
c<sub>0,j</sub></b> = d<sub>0,j</sub> + ad<sub>1,j</sub> + a<sup>2</sup>d<sub>2,j</sub> + a<sup>3</sup>d<sub>3,j</sub> is applied 1024 times. However, rather than executing this equation 1024
times, it is more efficient to implement this with three region-constant multiplications and three region-region additions.

<h3>2 &nbsp&nbsp&nbsp Files in the Library </h3>
This section provides an overview of the files that compose GF-Complete. They are partitioned among multiple
directories.

<h4> <b>2.1 &nbsp&nbsp&nbsp Header files in the directory  "include"</b> </h4>

The following header files are part of GF-Complete.
<ul>
<li><b>gf_complete.h:</b> This is the header file that applications should include. It defines the gf_t type, which holds
all of the data that you need to perform the various operations in GF(2<sup>w</sup>). It also defines all of the arithmetic
operations. For an application to use this library, you should include gf_complete.h and then compile with the
library src/libgf_complete.la. </li><br>

<li><b>gf_method.h:</b> If you are wanting to modify the implementation techniques from the defaults, this file provides
a "helper" function so that you can do it from the Unix command line.
</li><br>

<li><b>gf_general.h:</b> This file has helper routines for doing basic Galois Field operations with any legal value of <em>w.</em>
The problem is that <em>w </em> &#8804 32, <em>w </em> = 64 and <em> w </em> = 128 all have different data types, which is a pain. The procedures
in this file try to alleviate that pain. They are used in <b>gf_mult, gf_unit</b> and <b>gf_time.</b> I'm guessing that most
applications won't use them, as most applications use <em>w</em> &#8804 32. </li><br>

<li><b>gf_rand.h:</b> I've learned that <b>srand48()</b> and its kin are not supported in all C installations. Therefore, this file
defines some randomnumber generators to help test the programs. The randomnumber generator is the "Mother
</li>

</ul>







<br/>

2 &nbsp &nbsp  <em>  FILES IN THE LIBRARY     </em>   <span id="index_number">7  </span> <br><br><br>
<ul>

of All" random number generator [Mar94] which we've selected because it has no patent issues. <b>gf_unit</b> and
gf time use these random number generators.<br><br>
<li><b>gf_int.h:</b> This is an internal header file that the various source files use. This is <em>not</em> intended for applications to
include.</li><br>
<li><b>config.xx</b> and <b>stamp-h1</b> are created by autoconf, and should be ignored by applications. </li>
</ul>

<h3>2.2 &nbsp &nbsp <b> Source files in the "src" directory" </b> </h3>
<ul>
The following C files compose <b>gf_complete.a,</b> and they are in the direcoty src. You shouldn't have to mess with these
files, but we include them in case you have to:<br><br>
<li><b> gf_.c:</b> This implements all of the procedures in both <b>gf_complete.h</b> and <b>gf_int.h.</b> </li><br>
<li><b> gf_w4.c:</b> Procedures specific to <em>w </em> = 4. </li><br>
<li> <b>gf_w8.c:</b> Procedures specific to <em>w </em> = 8</li><br>
<li> <b>gf_w16.c:</b> Procedures specific to <em>w </em> = 16</li><br>
<li> <b>gf_w32.c:</b> Procedures specific to <em>w </em> = 32</li><br>
<li><b>gf_w64.c:</b> Procedures specific to <em>w </em> = 64</li><br>
<li> <b>gf_w128.c:</b> Procedures specific to <em>w </em> = 128</li><br>
<li> <b>gf_wgen.c:</b> Procedures specific to other values of <em>w </em> between 1 and 31</li><br>
<li> <b>gf_general.c:</b> Procedures that let you manipulate general values, regardless of whether <em>w </em> &#8804 32, <em>w </em> = 64
or <em>w </em> = 128. (I.e. the procedures defined in <b>gf_ general.h</b>)</li><br>
<li> <b>gf_method.c:</b> Procedures to help you switch between the various implementation techniques. (I.e. the procedures
defined in <b>gf_method.h</b>)</li><br>
<li> <b>gf_ rand.c:</b>"The Mother of all" random number generator. (I.e. the procedures defined in <b>gf_rand.h</b>)</li><br> </ul>

<h3>2.3 &nbsp &nbsp Library tools files in the "tools" directory </h3>

<ul>
The following are tools to help you with Galois Field arithmetic, and with the library. They are explained in greater
detail elsewhere in this manual.<br><br>
<li> <b>gf_mult.c, gf_ div.c</b> and <b>gf_ add:</b> Command line tools to do multiplication, division and addition by single numbers</li><br>
<li> <b>gf_time.c:</b> A program that times the procedures for given values of <em>w </em> and implementation options</li><br>
<li> <b>time tool.sh:</b> A shell script that helps perform rough timings of the various multiplication, division and region
operations in GF-Complete</li><br>
<li> <b>gf_methods.c:</b> A program that enumerates most of the implementation methods supported by GF-Complete</li><br>
<li> <b> gf_poly.c:</b> A program to identify irreducible polynomials in regular and composite Galois Fields</li><br>

</ul>








<br/>

3 &nbsp &nbsp  <em>  COMPILATION     </em>   <span id="index_number">8  </span> <br><br><br>


<h3>2.4 &nbsp &nbsp The unit tester in the "test" directory </h3>

The test directory contains the proram <b>gf_unit.c,</b> which performs a battery of unit tests on GF-Complete. This is
explained in more detail in section 6.3.


<h3>2.5&nbsp &nbsp Example programs in the "examples" directory </h3>

There are seven example programs to help you understand various facets of GF-Complete. They are in the files
<b>gf_example x.c </b> in the <b>examples</b> directory. They are explained in sections 4.2 through 4.5, and section 7.9.<br><br>

<h2>3 &nbsp &nbsp Compilation </h2>

<em>From revision 1.02 forward, we are using autoconf. The old "flag tester" directory is now gone, as it is no longer in
use. </em><br><br>
To compile and install, you should do the standard operations that you do with most open source Unix code:<br><br>

UNIX> ./configure <br>
... <br>
UNIX> make <br>
... <br>
UNIX> sudo make install <br><br>


<p>If you perform the <b>install,</b> then the header, source, tool, and library files will be moved to system locations. In
particular, you may then compile the library by linking with the flag <b>-lgf_complete,</b> and you may use the tools from a
global executable directory (like <b>/usr/local/bin</b>). </p>

<p>
If you don't perform the install, then the header and tool files will be in their respective directories, and the library
will be in <b>src/libgf_complete.la.</b> </p>
<p>
If your system supports the various Intel SIMD instructions, the compiler will find them, and GF-Complete will
use them by default. </p>



<h2>4 &nbsp &nbsp Some Tools and Examples to Get You Started </h2> 
<h3>4.1 Three Simple Command Line Tools: gf_mult, gf_div and gf_add </h3>


Before delving into the library, it may be helpful to explore Galois Field arithmetic with the command line tools:
<b>gf_mult, gf_div </b> and <b>gf_add.</b> These perform multiplication, division and addition on elements in <em>GF(2<sup>w</sup>).</em> If these are
not installed on your system, then you may find them in the tools directory. Their syntax is:
<ul>
<li><b>gf_mult a b</b> <em>w </em> - Multiplies a and b in <em> GF(2<sup>w</sup>)</em>. </li><br>
<li> <b>gf_div a b </b><em>w </em> - Divides a by b in GF(2<em><sup>w </sup></em>). </li><br>
<li><b>gf_add a b </b> <em>w </em> - Adds a and b in GF(2<em><sup>w </sup> </em>). </li><br>

You may use any value of <em>w </em> from 1 to 32, plus 64 and 128. By default, the values are read and printed in decimal;
however, if you append an 'h' to <em>w </em>, then <em>a, b </em> and the result will be printed in hexadecimal. For <em>w </em> = 128, the 'h' is
mandatory, and all values will be printed in hexadecimal.







<br/>

4 &nbsp &nbsp  <em>   SOME TOOLS AND EXAMPLES TO GET YOU STARTED 9     </em>   <span id="index_number">9  </span> <br><br><br>


<p>Try them out on some examples like the ones below. You of course don't need to know that, for example, 5 * 4 = 7
in <em>GF(2<sup>4 </sup>) </em>; however, once you know that, you know that 7/
5 = 4 and 7/4 = 5. You should be able to verify the <b>gf_add</b>
statements below in your head. As for the other <b>gf_mult's</b>, you can simply verify that division and multiplication work
with each other as you hope they would. </p>
<br><br>
<div id="number_spacing">

UNIX> gf_mult 5 4 4  <br>
7 <br>
UNIX> gf_div 7 5 4 <br>
4 <br>
UNIX> gf_div 7 4 4 <br>
5   <br>
UNIX> gf_mult 8000 2 16h <br>
100b  <br>
UNIX> gf_add f0f0f0f0f0f0f0f0 1313131313131313 64h <br>
e3e3e3e3e3e3e3e3 <br>
UNIX> gf_mult f0f0f0f0f0f0f0f0 1313131313131313 64h <br>
8da08da08da08da0 <br>
UNIX> gf_div 8da08da08da08da0 1313131313131313 64h <br>
f0f0f0f0f0f0f0f0  <br>
UNIX> gf_add f0f0f0f0f0f0f0f01313131313131313 1313131313131313f0f0f0f0f0f0f0f0 128h <br>
e3e3e3e3e3e3e3e3e3e3e3e3e3e3e3e3 <br>
UNIX> gf_mult f0f0f0f0f0f0f0f01313131313131313 1313131313131313f0f0f0f0f0f0f0f0 128h <br>
786278627862784982d782d782d7816e <br>
UNIX> gf_div 786278627862784982d782d782d7816e f0f0f0f0f0f0f0f01313131313131313 128h <br>
1313131313131313f0f0f0f0f0f0f0f0 <br>
UNIX> <br><br>

</div>


Don't bother trying to read the source code of these programs yet. Start with some simpler examples  like the ones
below. <br><br>

<h3>4.2 Quick Starting Example #1: Simple multiplication and division </h3>

The source files for these examples are in the examples directory.
<p>These two examples are intended for those who just want to use the library without getting too complex. The
first example is <b>gf_example 1,</b> and it takes one command line argument - w, which must be between 1 and 32. It
generates two random non-zero numbers in <em>GF(2<sup>w </sup>) </em> and multiplies them. After doing that, it divides the product by
each number. </p>
<p>
To perform multiplication and division in <em>GF(2<sup>w </sup>) </em>, you must declare an instance of the gf_t type, and then initialize
it for <em>GF(2<sup>w </sup>) </em> by calling <b>gf_init_easy().</b> This is done in <b>gf_example 1.c</b> with the following lines: </p><br><br>

gf_t gf; <br><br>r
... <br><br>
if (!gf_init_easy(&gf, w)) { <br>
fprintf(stderr, "Couldn't initialize GF structure.\n"); <br>
exit(0); <br>
}  <br>






<br/>

4 &nbsp &nbsp  <em>   SOME TOOLS AND EXAMPLES TO GET YOU STARTED      </em>   <span id="index_number">10  </span> <br><br><br>

<p>Once <b>gf</b> is initialized, you may use it for multiplication and division with the function pointers <b>multiply.w32</b> and
<b>divide.w32.</b> These work for any element of <em>GF(2<sup>w</sup>)</em> so long as w &#8804 32. </p> <br><br>

<div id="number_spacing">
<div style="padding-left:54px">
c = gf.multiply.w32(&gf, a, b);<br>
printf("%u * %u = %u\n", a, b, c);<br><br>
printf("%u / %u = %u\n", c, a, gf.divide.w32(&gf, c, a));<br>
printf("%u / %u = %u\n", c, b, gf.divide.w32(&gf, c, b));<br>


</div> </div>
<br><br>
Go ahead and test this program out. You can use <b>gf_mult</b> and <b>gf_div</b> to verify the results:<br><br>

<div id="number_spacing">
UNIX> gf_example_1 4 <br>
12 * 4 = 5  <br>
5 / 12 = 4  <br>
5 / 4 = 12  <br>
UNIX> gf_mult 12 4 4 <br>
5  <br>
UNIX> gf_example_1 16 <br>
14411 * 60911 = 44568 <br>
44568 / 14411 = 60911 <br>
44568 / 60911 = 14411  <br>
UNIX> gf_mult 14411 60911 16 <br>
44568 <br>
UNIX>  <br><br>
</div>

<b>gf_init_easy()</b> (and <b>later_gf_init_hard()</b>) do call <b>malloc()</b> to implement internal structures. To release memory, call
<b>gf_free().</b> Please see section 6.4 to see how to call <b>gf_init_hard()</b> in such a way that it doesn't call <b>malloc().</b> <br><br>



<h3>4.3 &nbsp &nbsp &nbspQuick Starting Example #2: Multiplying a region by a constant </h3>


The program <b>gf_example</b> 2 expands on <b>gf_example</b> 1. If <em>w</em> is equal to 4, 8, 16 or 32, it performs a region multiply
operation. It allocates two sixteen byte regions, <b>r1</b> and <b>r2,</b> and then multiples <b>r1</b> by a and puts the result in <b>r2</b> using
the <b>multiply_region.w32</b> function pointer: <br><br>

<div style="padding-left:52px">
gf.multiply_region.w32 (&gf, r1, r2, a, 16, 0); <br><br>
</div>

That last argument specifies whether to simply place the product into r2 or to XOR it with the contents that are already
in r2. Zero means to place the product there. When we run it, it prints the results of the <b>multiply_region.w32</b> in
hexadecimal. Again, you can verify it using gf mult:<br><br>
<div id="number_spacing">
UNIX> gf_example_2 4 <br>
12 * 2 = 11 <br>
11 / 12 = 2 <br>
11 / 2 = 12 <br><br>
multiply_region by 0xc (12) <br><br>
R1 (the source): 0 2 d 9 d 6 8 a 8 d b 3 5 c 1 8 8 e b 0 6 1 5 a 2 c 4 b 3 9 3 6 <br>
R2 (the product): 0 b 3 6 3 e a 1 a 3 d 7 9 f c a a 4 d 0 e c 9 1 b f 5 d 7 6 7 e <br>

</div>










<br/>

4 &nbsp &nbsp  <em>   SOME TOOLS AND EXAMPLES TO GET YOU STARTED      </em>   <span id="index_number">11  </span> <br><br><br>

<div id="number_spacing">
<table cellpadding="6">
<tr><td>UNIX></td> <td colspan="4"> gf_example_2 16 </td> </tr>

<tr>

<td>49598</td> <td> * </td> <td> 35999</td> <td> = </td> <td>19867 </td> </tr>

<tr><td>19867 </td><td>/ </td> <td> 49598 </td> <td> =  </td> <td>35999 </td> </tr>
<tr><td>19867</td><td> /</td> <td> 35999 </td> <td> = </td> <td> 49598 </td> </tr>  </table><br>


&nbsp multiply_region by 0xc1be (49598) <br><br>


<table cellpadding="6" >
<tr>
<td>R1 (the source):</td> <td> 8c9f </td> <td> b30e </td> <td> 5bf3 </td> <td> 7cbb </td> <td>16a9 </td> <td> 105d </td> <td> 9368 </td> <td> 4bbe </td> </tr>
<td>R2 (the product):</td> <td> 4d9b</td> <td> 992d </td> <td> 02f2 </td> <td> c95c </td> <td> 228e </td> <td> ec82 </td> <td> 324e </td> <td> 35e4 </td></tr>
</table>
</div>
<div id="number_spacing">
<div style="padding-left:9px">
UNIX> gf_mult c1be 8c9f 16h<br>
4d9b <br>
UNIX> gf_mult c1be b30e 16h <br>
992d <br>
UNIX> <br><br>
</div>
</div>

<h3>4.4 &nbsp &nbsp &nbsp Quick Starting Example #3: Using <em>w </em>= 64 </h3>
The program in <b>gf_example 3.c </b> is identical to the previous program, except it uses <em> GF(2<sup>64 </sup>). </em> Now <em>a, b</em> and <em> c </em> are
<b>uint64 t'</b>s, and you have to use the function pointers that have <b>w64</b> extensions so that the larger types may be employed.
<br><br>
<div id="number_spacing">

UNIX> gf_example_31 
<table cellpadding="6">
<tr>

<td>a9af3adef0d23242 </td> <td> * </td> <td> 61fd8433b25fe7cd</td> <td> = </td> <td>bf5acdde4c41ee0c </td> </tr>

<td>bf5acdde4c41ee0c </td> <td> / </td> <td> a9af3adef0d23242 </td> <td> = </td> <td>61fd8433b25fe7cd </td> </tr>
<td>bf5acdde4c41ee0c </td> <td> / </td> <td> 61fd8433b25fe7cd  </td> <td>= </td> <td>a9af3adef0d23242 </td> </tr>
</table><br><br>

&nbsp multiply_region by a9af3adef0d23242<br><br>
<table cellpadding="6" >
<tr>
<td>R1 (the source): </td> <td> 61fd8433b25fe7cd </td> <td>272d5d4b19ca44b7 </td> <td> 3870bf7e63c3451a </td> <td> 08992149b3e2f8b7 </td> </tr>
<tr><td>R2 (the product): </td> <td> bf5acdde4c41ee0c </td> <td> ad2d786c6e4d66b7 </td> <td> 43a7d857503fd261 </td> <td> d3d29c7be46b1f7c </td> </tr>
</table>

<div style="padding-left:9px">

UNIX> gf_mult a9af3adef0d23242 61fd8433b25fe7cd 64h <br>
bf5acdde4c41ee0c<br>
UNIX><br><br>
</div>
</div>
<h3>4.5 &nbsp &nbsp &nbsp Quick Starting Example #4: Using <em>w </em>= 128 </h3>
Finally, the program in <b>gf_example_4.c</b> uses  <em>GF(2<sup>128</sup>).</em> Since there is not universal support for uint128 t, the library
represents 128-bit numbers as arrays of two uint64 t's. The function pointers for multiplication, division and region
multiplication now accept the return values as arguments:<br><br>

gf.multiply.w128(&gf, a, b, c); <br><br>

Again, we can use <b>gf_mult </b> and <b>gf_div </b>to verify the results:<br><br>
<div id="number_spacing">
<div style="padding-left:9px">
UNIX> gf_example_4 </div>
<table cellpadding="6" >
<tr>

<td>e252d9c145c0bf29b85b21a1ae2921fa </td> <td> * </td> <td> b23044e7f45daf4d70695fb7bf249432 </td> <td> = </td> </tr>
<tr><td>7883669ef3001d7fabf83784d52eb414 </td> </tr>

</table>

</div>








<br/>

4 &nbsp &nbsp  <em>   IMPORTANT INFORMATION ON ALIGNMENT WHEN MULTIPLYING REGIONS      </em>   <span id="index_number">12  </span> <br><br><br>

<div id="number_spacing">
multiply_region by e252d9c145c0bf29b85b21a1ae2921fa <br>
R1 (the source): f4f56f08fa92494c5faa57ddcd874149 b4c06a61adbbec2f4b0ffc68e43008cb <br>
R2 (the product): b1e34d34b031660676965b868b892043 382f12719ffe3978385f5d97540a13a1 <br>
UNIX> gf_mult e252d9c145c0bf29b85b21a1ae2921fa f4f56f08fa92494c5faa57ddcd874149 128h <br>
b1e34d34b031660676965b868b892043 <br>
UNIX> gf_div 382f12719ffe3978385f5d97540a13a1 b4c06a61adbbec2f4b0ffc68e43008cb 128h<br>
e252d9c145c0bf29b85b21a1ae2921fa<br>
UNIX><br><br>

</div>


<h2>5 &nbsp &nbsp &nbspImportant Information on Alignment when Multiplying Regions </h2>



In order to make multiplication of regions fast, we often employ 64 and 128 bit instructions. This has ramifications
for pointer alignment, because we want to avoid bus errors, and because on many machines, loading and manipulating
aligned quantities is much faster than unalinged quantities.<br><br>


When you perform multiply_region.wxx(<em>gf, source, dest, value, size, add </em>), there are three requirements:
<ol>
<li>
 The pointers <em>source</em> and <em>dest </em> must be aligned for <em>w</em>-bit words. For <em>w </em> = 4 and <em>w </em> = 8, there is no restriction;
however for <em>w </em> = 16, the pointers must be multiples of 2, for <em>w </em> = 32, they must be multiples of 4, and for
<em>w </em> &#1013; {64, 128}, they must be multiples of 8. </li><br>

<li> The <em>size</em> must be a multiple of &#91; <em>w /
</em> 
8 .&#93;
 With <em>w </em> = 4 and <em>w </em> = 8, <em>w/ </em>
8  = 1 and there is no restriction. The other
sizes must be multiples of <em>w </em>/
8  because you have to be multiplying whole elements of <em> GF(2<sup>w </sup>) </em>. </li><br>

<li> The <b>source</b> and <b>dest</b> pointers must be aligned identically with respect to each other for the implementation
chosen. This is subtle, and we explain it in detail in the next few paragraphs. However, if you'd rather not figure
it out, the following recommendation will <em>always </em> work in GF-Complete: </li>

</ol>



<div style="padding-left:100px">
<b>If you want to be safe, make sure that source and dest are both multiples of 16. That is not a
strict requirement, but it will always work! </b> <br><br>
</div>


If you want to relax the above recommendation, please read further.
<p>When performing <b>multiply_region.wxx() </b>, the implementation is typically optimized for a region of bytes whose
size must be a multiple of a variable <em>s </em> ,, and which must be aligned to a multiple of another variable <em>t </em>. For example,
when doing <b>multiply_region.w32() </b> in <em> GF(2<sup>16 </sup>) </em> with SSE enabled, the implementation is optimized for regions of
32 bytes, which must be aligned on a 16-byte quantity. Thus, <em>s </em> = 32 and <em>t</em> = 16. However, we don't want <b>multiply_
region.w32() </b> to be too restrictive, so instead of requiring <em>source</em> and <em> dest </em> to be aligned to 16-byte regions, we
require that (<em>source </em> mod 16) equal (<em>dest</em> mod 16). Or, in general, that (<em>source</em> mod t) equal (<em>dest</em> mod <em>t</em>). </p>


<p>
Then, <b>multiply_region.wxx()</b> proceeds in three phases. In the first phase,<b> multiply.wxx()</b> is called on successive
words until (<em>source</em> mod <em>t</em>) equals zero. The second phase then performs the optimized region multiplication on
chunks of <em> s  </em>bytes, until the remaining part of the region is less than s bytes. At that point, the third phase calls
<em>multiply.wxx() </em> on the last part of the region. </p>

A detailed example helps to illustrate. Suppose we make the following call in <em>GF(2<sup>16</sup>) </em> with SSE enabled:<br><br>
<center><b>multiply region.w32(gf, 0x10006, 0x20006, a, 274, 0)</b> </center>







<br/>

2 &nbsp &nbsp  <em>  FILES IN THE LIBRARY     </em>   <span id="index_number">13  </span> <br><br><br>



<div class="image-cell_2"> </div>  <br><br><br>

Figure 2: Example of multiplying a region of 274 bytes in GF(216) when (source mod 16) = (dest mod 16) = 6. The
alignment parameters are s = 32 and t = 16. The multiplication is in three phases, which correspond to the initial
unaligned region (10 bytes), the aligned region of s-byte chunks (256 bytes), and the final leftover region (8 bytes).


<p>First, note that <em>source</em> and <em>dest</em> are aligned on two-byte quantities, which they must be in <em>GF(2<sup>16</sup>).</em> Second, note
that size is a multiple of &#91; 16/
8 &#93 = 2. And last, note that (<em>source</em> mod 16) equals (<em>dest</em> mod 16). We illustrate the three
phases of region multiplication in Figure 2. Because (<em>source</em> mod 16) = 6, there are 10 bytes of unaligned words that
are multiplied with five calls to <b>multiply.w32()</b> in the first phase. The second phase multiplies 256 bytes (eight chunks
of <em>s</em> = 32 bytes) using the SSE instructions. That leaves 8 bytes remaining for the third phase.
</p>

<p>
When we describe the defaults and the various implementation options, we specify s and t as "alignment parameters."
</p>
<p>
One of the advanced region options is using an alternate mapping of words to memory ("ALTMAP"). These interact
in a more subtle manner with alignment. Please see Section 7.9 for details.
</p>

<h3> 6 &nbsp &nbspThe Defaults </h3>


GF-Complete implements a wide variety of techniques for multiplication, division and region multiplication. We have
set the defaults with three considerations in mind:
<ol>
<li>
<b>Speed:</b> Obviously, we want the implementations to be fast. Therefore, we choose the fastest implementations
that don’t violate the other considerations. The compilation environment is considered. For example, if SSE is
enabled, region multiplication in <em> GF(2<sup>4 </sup>) </em> employs a single multiplication table. If SSE is not enabled, then a
"double" table is employed that performs table lookup two bytes at a time. </li><br>
<li>
<b>Memory Consumption:</b> We try to keep the memory footprint of GF-Complete low. For example, the fastest
way to perform <b>multiply.w32()</b> in <em>GF(2<sup>32</sup>) </em> is to employ 1.75 MB of multiplication tables (see Section 7.4
below). We do not include this as a default, however, because we want to keep the default memory consumption
of GF-Complete low.
</li>

</ul>






<br/>

6 &nbsp &nbsp  <em>  THE DEFAULTS     </em>   <span id="index_number">14  </span> <br><br><br>

<ul>

3. &nbsp <b>Compatibility with "standard" implementations:</b> While there is no <em>de facto</em> standard of Galois Field arithmetic,
most libraries implement the same fields. For that reason, we have not selected composite fields, alternate
polynomials or memory layouts for the defaults, even though these would be faster. Again, see section 7.7 for
more information.

</ul>

<p>Table 1 shows the default methods used for each power-of-two word size, their alignment parameters <em>s</em> and <em> t,</em> their
memory consumption and their rough performance. The performance tests are on an Intel Core i7-3770 running at
3.40 GHz, and are included solely to give a flavor of performance on a standard microprocessor. Some processors
will be faster with some techniques and others will be slower, so we only put numbers in so that you can ballpark it.
For other values of <em>w</em> between 1 and 31, we use table lookup when w &#8804 8, discrete logarithms when w &#8804 16 and
"Bytwop" for w &#8804 32. </p>
<br><br>
<center> With SSE 
<div id="data1">
<table cellpadding="6" cellspacing="0">
<tr>
<th>w </th><th class="double_border" >Memory <br> Usage </br> </th><th>multiply() <br> Implementation</th><th>Performance <br>(Mega Ops / s) </th><th>multiply region() <br> Implementation </th>
<th>s </th> <th>t </th> <th> Performance <br>(MB/s)</th>
</tr>
<tr>
<td>4 </td><td class="double_border"><1K </td><td>Table</td><td>501</td><td>Table</td>
<td>16 </td><td>16 </td> <td>11,659</td> </tr>

<tr>
<td>8 </td><td class="double_border">136K </td><td>Table</td><td>501</td><td>Split Table (8,4)</td>
<td>16 </td><td>16 </td> <td>11,824</td> </tr>

<tr>
<td>16 </td><td class="double_border">896K </td><td>Log</td><td>260</td><td>Split Table (16,4)</td>
<td>32 </td><td>16 </td> <td>7,749</td> </tr>

<tr>
<td>32 </td><td class="double_border"><1K </td><td>Carry-Free</td><td>48</td><td>Split Table (32,4)</td>
<td>64 </td><td>16 </td> <td>5,011</td> </tr>

<tr>
<td>64 </td><td class="double_border">2K </td><td>Carry-Free</td><td>84</td><td>Split Table (64,4)</td>
<td>128 </td><td>16 </td> <td>2,402</td> </tr>

<tr>
<td>128 </td><td class="double_border">64K </td><td>Carry-Free</td><td>48</td><td>Split Table (128,4)</td>
<td>16 </td><td>16 </td> <td>833</td> </tr>
</table></div>


<div id="data1">
<center>Without SE </center>
<table cellpadding="6" cellspacing="0">
<tr>
<th>w </th><th>Memory <br> Usage </br> </th><th>multiply() <br> Implementation</th><th>Performance <br>(Mega Ops / s) </th><th>multiply region() <br> Implementation </th>
<th>s </th> <th>t </th> <th> Performance <br>(MB/s)</th>
</tr>
<tr>
<td>4 </td><td>4K </td><td>Table</td><td>501</td><td>Double Table</td>
<td>16 </td><td>16 </td> <td>11,659</td> </tr>

<tr>
<td>8 </td><td>128K </td><td>Table</td><td>501</td><td>Table</td>
<td>1 </td><td>1 </td> <td>1,397</td> </tr>

<tr>
<td>16 </td><td>896K </td><td>Log</td><td>266</td><td>Split Table (16,8)</td>
<td>32 </td><td>16 </td> <td>2,135</td> </tr>

<tr>
<td>32 </td><td>4K </td><td>Bytwop</td><td>19</td><td>Split Table (32,4)</td>
<td>4 </td><td>4 </td> <td>1,149</td> </tr>

<tr>
<td>64 </td><td>16K </td><td>Bytwop</td><td>9</td><td>Split Table (64,4)</td>
<td>8 </td><td>8 </td> <td>987</td> </tr>

<tr>
<td>128 </td><td>64K </td><td>Bytwop</td><td>1.4</td><td>Split Table (128,4)</td>
<td>16 </td><td>8 </td> <td>833</td> </tr>
</table>
</div>
</center>
<br><br>
Table 1: The default implementations, memory consumption and rough performance when w is a power of two. The
variables s and t are alignment variables described in Section 5.
<p>
A few comments on Table 1 are in order. First, with SSE, the performance of <b>multiply()</b> is faster when <em> w </em> = 64
than when<em> w </em> = 32. That is because the primitive polynomial for <em> w  </em>= 32, that has historically been used in Galois
Field implementations, is sub-ideal for using carry-free multiplication (PCLMUL). You can change this polynomial
(see section 7.7) so that the performance matches <em>w </em> = 64. </p>
<p>
The region operations for <em> w  </em>= 4 and <em>w </em>= 8 without SSE have been selected to have a low memory footprint. There
are better options that consume more memory, or that only work on large memory regions (see section 6.1.5).
</p>

There are times that you may want to stray from the defaults. For example:
<ul>
<li>
You may want better performance.
</li>

</ul>










<br/>

6 &nbsp &nbsp  <em>  THE DEFAULTS     </em>   <span id="index_number">15  </span> <br><br><br>

<ul>
<li>You may want a lower memory footprint.</li>
<li>You may want to use a different Galois Field or even a ring.</li>
<li>You only care about multiplying a region by the value two.</li>

</ul>


<p>
Our command line tools allow you to deviate from the defaults, and we have two C functions <b>-gf_init_hard()</b>
and <b>create_gf_from_argv()</b> that can be called from application code to override the default methods. There are six
command-line tools that can be used to explore the many techniques implemented in GF-Complete: </p>

<ul><br>

<li> <b>gf_methods</b> is a tool that enumerates most of the possible command-line arguments that can be sent to the other
tools</li><br>
<li> <b>gf_mult</b> and <b>gf_div</b> are explained above. You may change the multiplication and division technique in these
tools if you desire</li><br>
<li> <b>gf_unit</b> performs unit tests on a set of techniques to verify correctness</li><br>
<li> <b> gf_time measures </b> the performance of a particular set of techniques</li><br>
<li> <b>time_tool.sh </b> makes some quick calls to <b>gf_time</b> so that you may gauge rough performance.</li><br>
<li> <b>gf_poly</b> tests the irreducibility of polynomials in a Galois Field</li><br>
</ul>


<p>To change the default behavior in application code, you need to call <b>gf_init_hard()</b> rather than <b>gf_init_easy().</b>
Alternatively, you can use <b>create_g_from_argv(),</b> included from <b>gf_method.h,</b> which uses an <b>argv</b>-style array of
strings to specify the options that you want. The procedure in <b>gf_method.c</b> parses the array and makes the proper
<b>gf_init_hard()</b> procedure call. This is the technique used to parse the command line in <b> gf_mult, gf_div, gf_unit </b><em>et al.</em> </p>


<h2>6.1.1 Changing the Components of a Galois Field with create <b>gf_from_argv()</b> </h2>
There are five main components to every Galois Field instance:
<ul>
<li> <em>w </em> </li>
<li> Multiplication technique </li>
<li> Division technique  </li>
<li> Region technique(s) </li>
<li> Polynomial </li>
</ul>

<p>The procedures <b>gf_init_hard()</b> and <b> create_gf_from_argv()</b> allow you to specify these parameters when you create
your Galois Field instance. We focus first on <b>create_gf_from_argv(),</b> because that is how the tools allow you to specify
the components. The prototype of <b>create_gf_from_argv()</b> is as follows: </p><br>

<div id="number_spacing">
int create_gf_from_argv(gf_t *gf, int w, int argc, char **argv, int starting);<br><br> </div>

You pass it a pointer to a gf_t, which it will initialize. You specify the word size with the parameter <em><b>w,</b></em> and then you
pass it an <b>argc/argv</b> pair as in any C or C++ program. You also specify a <b>starting</b> argument, which is where in <b>argv</b>
the specifications begin. If it successfully parses <b>argc</b> and <b>argv,</b> then it creates the <b>gf_t</b> using <b>gf_init_hard()</b> (described
below in section 6.4). It returns one past the last index of <b>argv</b> that it considered when creating the <b>gf_t.</b> If it fails, then
it returns zero, and the <b>gf_t</b> is unmodified.



<p>For example, <b>gf_mult.c</b> calls create gf_from_argv() by simply passing <b>argc</b> and <b>argv</b> from its <b>main()</b> declaration,
and setting starting to 4.</p>








<br/>

6 &nbsp &nbsp  <em>  THE DEFAULTS     </em>   <span id="index_number">16  </span> <br><br><br>

<p>
To choose defaults, <b>argv[starting]</b> should equal "-". Otherwise, you specify the component that you are changing
with "-m" for multiplication technique, "-d" for division technique, "-r" for region technique, and "-p" for the
polynomial. You may change multiple components. You end your specification with a single dash. For example, the
following call multiplies 6 and 5 in <em>GF(2<sup>4</sup>)</em> with polynomial 0x19 using the "SHIFT" technique for multiplication
(we'll explain these parameters later):
</p><br><br>

<div id="number_spacing">
UNIX> ./gf_mult 6 5 4 -p 0x19 -m SHIFT -<br>
7 <br>
UNIX> <br><br>
</div>

<p>If <b>create_gf_from_argv()</b> fails, then you can call the procedure <b>gf_error(),</b> which prints out the reason why <b>create_
gf_from_argv()</b> failed. </p>


<h2>6.1.2 Changing the Polynomial </h2>

Galois Fields are typically implemented by representing numbers as polynomials with binary coefficients, and then
using the properties of polynomials to define addition and multiplication. You do not need to understand any of that to
use this library. However, if you want to learn more about polynomial representations and how they construct fields,
please refer to The Paper.

<p>Multiplication is based on a special polynomial that we will refer to here as the "defining polynomial." This
polynomial has binary coefficients and is of degree <em> w.</em> You may change the polynomial with "-p" and then a number
in hexadecimal (the leading "0x" is optional). It is assumed that the <em>w</em>-th bit of the polynomial is set - you may include
it or omit it. For example, if you wish to set the polynomial for GF(2<sup>16</sup>) to x<sup>16</sup> + x<sup>5</sup> + x<sup>3</sup> + x<sup>2</sup> + 1, rather than its
default of x<sup>16</sup> + x<sup>12</sup> + x<sup>3</sup> + x + 1, you may say "-p 0x1002d," "-p 1002d," "-p 0x2d" or "-p 2d."
We discuss changing the polynomial for three reasons in other sections: </p>
<ul>
<li>Leveraging carry-free multiplication (section 7.7). </li>
<li>Defining composite fields (section 7.6). </li>
<li>Implementing rings (section 7.8.1). </li>

</ul>

<p>
Some words about nomenclature with respect to the polynomial. A Galois Field requires the polynomial to be
<em>irreducible </em>.. That means that it cannot be factored. For example, when the coefficients are binary, the polynomial x<sup>5</sup>+
x<sup>4</sup>+x+1 may be factored as (x<sup>4</sup>+1)(x+1). Therefore it is not irreducible and cannot be used to define a Galois Field.
It may, however, be used to define a ring. Please see section 7.8.1 for a discussion of ring support in GF-Complete. </p>
<p>
There is a subset of irreducible polynomials called primitive. These have an important property that one may enumerate
all of the elements of the field by raising 2 to successive posers. All of the default polynomials in GF-Complete 
are primitive. However, so long as a polynomial is irreducible, it defines a Galois Field. Please see section 7.7 for a
further discussion of the polynomial. </p>

<p>
One thing that we want to stress here is that changing the polynomial changes the field, so fields with different
polynomialsmay not be used interchangeably. So long as the polynomial is irreducible, it generates a Galois Field that
is isomorphic to all other Galois Fields; however the multiplication and division of elements will differ. For example,
the polynomials 0x13 (the default) and 0x19 in <em>GF(2<sup>4</sup>) </em> are both irreducible, so both generate valid Galois Fields.
However, their multiplication differs: </p><br>

<div id="number_spacing">
UNIX> gf_mult 8 2 4 -p 0x13 - <br>
3 <br>
UNIX> gf_mult 8 2 4 -p 0x19 - <br>
9 <br>
</div>









<br/>

6 &nbsp &nbsp  <em>  THE DEFAULTS     </em>   <span id="index_number">17  </span> <br><br><br>

<div id="number_spacing">
UNIX> gf_div 3 8 4 -p 0x13 -<br>
2 <br>
UNIX> gf_div 9 8 4 -p 0x19 - <br>
2 <br>
UNIX> <br>

</div>


<h3>6.1.3 &nbsp &nbsp Changing the Multiplication Technique </h3>
The following list describes the multiplication techinques that may be changed with "-m". We keep the description
here brief. Please refer to The Paper for detailed descriptions of these techniques.<br><br>


<li><b> "TABLE:" </b> Multiplication and division are implemented with tables. The tables consume quite a bit of memory
(2<sup>w</sup> &#215 2 <sup>w</sup> &#215  <sup>w</sup>/
8  bytes), so they are most useful when <em>w</em> is small. Please see <b>"SSE," "LAZY," "DOUBLE"</b> and

<b>"QUAD"</b> under region techniques below for further modifications to <b>"TABLE"</b> to perform <b>multiply_region()</b></li><br>


<li> <b>"LOG:"</b> This employs discrete (or "Zeph") logarithm <b>tables</b> to implement multiplication and division. The
memory usage is roughly (3 &#215 2<sup>w</sup> &#215 w /
8  bytes), so they are most useful when w is small, but they tolerate
larger <em>w</em> than <b>"TABLE."</b> If the polynomial is not primitive (see section 6.1.2), then you cannot use <b>"LOG"</b> as
an implementation. In that case,<b> gf_init_hard()</b> or <b>create_gf_from_argv()</b> will fail</li><br>


<li><b> "LOG ZERO:"</b> Discrete logarithm tables which include extra room for zero entries. This more than doubles
the memory consumption to remove an <b>if</b> statement (please see [GMS08] or The Paper for more description). It
doesn’t really make a huge deal of difference in performance</li><br>

<li> <b>"LOG ZERO EXT:"</b> This expends even more memory to remove another <b>if</b> statement. Again, please see The
Paper for an explanation. As with <b>"LOG ZERO,"</b> the performance difference is negligible</li><br>

<li> <b>"SHIFT:"</b> Implementation straight from the definition of Galois Field multiplication, by shifting and XOR-ing,
then reducing the product using the polynomial. This is <em>slooooooooow,</em> so we don’t recommend you use it</li><br>


<li> <b>"CARRY FREE:"</b> This is identical to <b>"SHIFT,"</b> however it leverages the SSE instruction PCLMUL to perform
carry-freemultiplications in single instructions. As such, it is the fastest way to perform multiplication for large
values of <em>w</em> when that instruction is available. Its performance depends on the polynomial used. See The Paper
for details, and see section 7.7 below for the speedups available when <em>w </em>= 16 and <em>w</em> = 32 if you use a different
polynomial than the default one</li><br>


<li> <b>"BYTWO p:"</b> This implements multiplication by successively multiplying the product by two and selectively
XOR-ing the multiplicand. See The Paper for more detail. It can leverage Anvin’s optimization that multiplies
64 and 128 bits of numbers in <em>GF(2<sup>w</sup>) </em> by two with just a few instructions. The SSE version requires SSE2</li><br>


<li> <b>"BYTWO b:"</b> This implements multiplication by successively multiplying the multiplicand by two and selectively
XOR-ing it into the product. It can also leverage Anvin's optimization, and it has the feature that when
you're multiplying a region by a very small constant (like 2), it can terminate the multiplication early. As such,
if you are multiplying regions of bytes by two (as in the Linux RAID-6 Reed-Solomon code [Anv09]), this is
the fastest of the techniques, regardless of the value of <em>w.</em> The SSE version requires SSE2</li><br>


<li> <b>"SPLIT:"</b> Split multiplication tables (like the LR tables in [GMS08], or the SIMD tables for w &#8804 8 in [LHy08,
Anv09, PGM13b]). This argument must be followed by two more arguments, w<sub>a</sub> and w<sub>b</sub>, which are the index
sizes of the sub-tables. This implementation reduces the size of the table from <b>"TABLE,"</b> but requires multiple
</li><br>






<br/>

6 &nbsp &nbsp  <em>  THE DEFAULTS     </em>   <span id="index_number">18 </span> <br><br><br>
<ul>
table lookups. For example, the following multiplies 100 and 200 in <em>GF(2<sup>8</sup>) </em> using two 4K tables, as opposed 
to one 64K table when you use <b>"TABLE:"</b><br><br>
<div id="number_spacing">
UNIX> ./gf_mult 100 200 8 -m SPLIT 8 4 - <br>
79<br>
UNIX><br><br>
</div>

See section 7.4 for additional information on the arguments to <b>"SPLIT."</b> The SSE version typically requires
SSSE3.<br><br>


<li> <b>"GROUP:"</b> This implements the "left-to-right comb" technique [LBOX12]. I'm afraid we don't like that name,
so we call it <b>"GROUP,"</b> because it performs table lookup on groups of bits for shifting (left) and reducing (right).
It takes two additional arguments - g<sub>s,</sub> which is the number of bits you use while shifting (left) and g<sub>r</sub>, which
is the number of bits you use while reducing (right). Increasing these arguments can you higher computational
speed, but requires more memory. SSE version exists only for <em> w </em> = 128 and it requires SSE4. For more
description on the arguments g<sub>s</sub> and g<sub>r</sub>, see section 7.5. For a full description of <b>"GROUP"</b> algorithm, please
see The Paper.
</li><br>

<li> <b>"COMPOSITE:"</b> This allows you to perform operations on a composite Galois Field, <em> GF((2<sup>l</sup>)<sup>k</sup>)</em> as described
in [GMS08], [LBOX12] and The Paper. The field size <em>w </em> is equal to <em>lk.</em> It takes one argument, which is <em>k,</em> and
then a specification of the base field. Currently, the only value of <em>k</em> that is supported is two. However, that may
change in a future revision of the library. </li><br>


In order to specify the base field, put appropriate flags after specifying <em>k.</em> The single dash ends the base field,
and after that, you may continue making specifications for the composite field. This process can be continued
for multiple layers of <b>"COMPOSITE."</b> As an example, the following multiplies 1000000 and 2000000
in <em>GF((2<sup>16</sup>)<sup>2</sup>),</em> where the base field uses <b>BYTWO_p</b> for multiplication: <br><br>
<center>./gf mult 1000000 2000000 32 -m COMPOSITE 2 <span style="color:red">-m BYTWO p - -</span> </center><br>

In the above example, the red text applies to the base field, and the black text applies to the composite field.
Composite fields have two defining polynomials - one for the composite field, and one for the base field. Thus, if
you want to change polynomials, you should change both. The polynomial for the composite field must be of the
form x<sup>2</sup>+sx+1, where s is an element of <em>GF(2<sup>k</sup>).</em> To change it, you specify s (in hexadecimal)with "-p." In the
example below, we multiply 20000 and 30000 in <em>GF((2<sup>8</sup>)<sup>2</sup>) </em>, setting s to three, and using x<sup>8</sup>+x<sup>4</sup>+x<sup>3</sup>+x<sup>2</sup>+1
as the polynomial for the base field: <br><br>

<center>./gf mult 20000 30000 16 -m COMPOSITE 2 <span style="color:red">-p 0x11d </span> - -p 0x3 - </center> <br><br>

If you use composite fields, you should consider using <b>"ALTMAP"</b> as well. The reason is that the region
operations will go much faster. Please see section 7.6.<br><br>
As with changing the polynomial, when you use a composite field, <em> GF((2<sup>l</sup>)<sup>k</sup>)</em>, you are using a different field
than the "standard" field for <em> GF((2<sup>l</sup>)<sup>k</sup>)</em>. All Galois Fields are isomorphic to each other, so they all have the
desired properties; however, the fields themselves change when you use composite fields.<br><br>
</ul>
<p>
With the exception of <b>"COMPOSITE"</b>, only one multiplication technique can be provided for a given Galois
Field instance. Composite fields may use composite fields as their base fields, in which case the specification will be
recursive. </p>








<br/>

6 &nbsp &nbsp  <em>  THE DEFAULTS     </em>   <span id="index_number">19 </span> <br><br><br>

<h3>6.1.4 &nbsp &nbsp &nbsp Changing the Division Technique </h3>

There are two techniques for division that may be set with "-d". If "-d" is not specified, then appropriate defaults
are employed. For example, when the multiplication technique is <b>"TABLE,"</b> a table is created for division as well as
multiplication. When <b>"LOG"</b> is specified, the logarithm tables are used for division. With <b>"COMPOSITE,"</b> a special
variant of Euclid's algorithm is employed that performs division using multiplication and division in the base field.
Otherwise, Euclid's algorithm is used. Please see The Paper for a description of Euclid's algorithm applied to Galois
Fields.

<p>If you use "-d", you must also specify the multiplication technique with "-m." </p>
<p>To force Euclid's algorithm instead of the defaults, you may specify it with "-d EUCLID." If instead, you would
rather convert elements of a Galois Field to a binary matrix and find an element's inverse by inverting the matrix,
then specify "-d MATRIX." In all of our tests, <b>"MATRIX"</b> is slower than <b>"EUCLID." "MATRIX" </b> is also not defined
for <em>w </em> > 32.
</p>


<h3>6.1.5  &nbsp&nbsp&nbsp Changing the Region Technique </h3>
The following are the region multiplication options ("-r"):
<ul>
<li>
<b>"SSE:"</b> Use SSE instructions. Initialization will fail if the instructions aren't supported. Table 2 details the
multiplication techniques which can leverage SSE instructions and which versions of SSE are required. </li><br>

<center>
<div id="data1">
<table cellpadding="6" cellspacing="0" style="text-align:center;font-size:19px">
<tr>
<th>Multiplication <br> Technique</th><th>multiply() </th><th>multiply_region() </th><th>SSE Version </th><th>Comments</th>

</tr>
<tr>
<td><b>"TABLE"</b></td><td >- </td><td>Yes</td><td>SSSE3</td><td>Only for <em>GF(2<sup>4</sup>). </em></td>

<tr>
<td><b>"SPLIT"</b></td><td>-</td><td>Yes</td><td>SSSE3</td><td>Only when the second argument equals 4.</td>

<tr>
<td><b>"SPLIt"</b></td><td>- </td><td>Yes</td><td>SSE4</td><td>When <em>w </em> = 64 and not using <b>"ALTMAP".</b></td>

<tr>
<td><b>"BYTWO p"</b></td><td>- </td><td>Yes</td><td>SSE2</td><td></td>

<tr>
<td><b>"BYTWO p"</b></td><td>- </td><td>Yes</td><td>SSE2</td><td></td>

</table></div> <br><br>
Table 2: Multiplication techniques which can leverage SSE instructions when they are available.
</center> <br><br>












<li> <b>"NOSSE:"</b> Force non-SSE version </li><br>

<li> <b> "DOUBLE:"</b> Use a table that is indexed on two words rather than one. This applies only to <em>w  </em> = 4, where
the table is indexed on bytes rather than 4-bit quantities, and to <em>w </em> = 8, where the table is indexed on shorts
rather than bytes. In each case, the table lookup performs two multiplications at a time, which makes region
multiplication faster. It doubles the size of the lookup table. </li><br>

<li> <b>"QUAD:"</b> Use a table that is indexed on four words rather than two or one. This only applies to <em>w </em> = 4, where
the table is indexed on shorts. The "Quad" table may be lazily created or created ahead of time (the default). If
the latter, then it consumes 2<sup>4</sup> &#215 2<sup>16</sup> &#215 2 = 2 MB of memory. </li><br>

<li> <b> "LAZY:"</b> Typically it's clear whether tables are constructed upon initialization or lazily when a region operation
is performed. There are two times where it is ambiguous: <b>"QUAD"</b> when <em>w </em> = 4 and <b>"DOUBLE"</b> when <em>w </em> = 8.
If you don't specify anything, these tables are created upon initialization, consuming a lot of memory. If you
specify <b>"LAZY,"</b> then the necessary row of the table is created lazily when you call <b>"multiply_region().</b>
</li>

</ul>











<br/>

6 &nbsp &nbsp  <em>  THE DEFAULTS     </em>   <span id="index_number">20 </span> <br><br><br>
<ul>

<li> <b>"ALTMAP:"</b> Use an alternate mapping, where words are split across different subregions of memory. There
are two places where this matters. The first is when implementing "<b>SPLIT</b> <em>w </em> 4" using SSE when <em>w </em> > 8. In
these cases, each byte of the word is stored in a different 128-bit vector, which allows the implementation to
better leverage 16-byte table lookups. See section 7.4 for examples, and The Paper or [PGM13b] for detailed
explanations.<br><br> </li>

The second place where it matters is when using <b>"COMPOSITE."</b> In this case, it is advantageous to split each
memory region into two chunks, and to store half of each word in a different chunk. This allows us to call
<b>region_multiply() </b> recursively on the base field, which is <em>much </em> faster than the alternative. See Section 7.6 for
examples, and The Paper for an explanation.<br><br>

It is important to note that with <b>"ALTMAP,"</b> the words are not "converted" from a standard mapping to an
alternate mapping and back again. They are assumed to always be in the alternate mapping. This typically
doesn't matter, so long as you always use the same <b>"ALTMAP"</b> calls. Please see section 7.9 for further details
on <b>"ALTMAP,"</b> especially with respect to alignment.<br><br>

<li> <b>"CAUCHY:"</b> Break memory into <em>w </em> subregions and perform only XOR's as in Cauchy Reed-Solomon coding
[BKK<sup>+</sup>95] (also described in The Paper). This works for <em>any</em> value of <em>w </em> &#8804 32, even those that are not
powers of two. If SSE2 is available, then XOR's work 128 bits at a time. For <b>"CAUCHY"</b> to work correctly,
<em>size</em> must be a multiple of <em>w </em>.</li> </ul>



<p>It is possible to combine region multiplication options. This is fully supported as long as <b>gf_methods</b> has the combination
listed. If multiple region options are required, they should be specified independently (as flags for <b>gf_init_hard()</b>
and independent options for command-line tools and <b>create_gf_from_argv()).</b> </p>


<h3>6.2  &nbsp&nbsp&nbspDetermining Supported Techniques with gf methods </h3>


The program <b>gf_methods</b> prints a list of supported methods on standard output. It is called as follows:<br><br>
<div id="number_spacing">
<center>./gf methods <em>w </em> -BADC -LUMDRB <br><br> </center> </div>

The first argument is <em>w </em>, which may be any legal value of <em>w </em>. The second argument has the following flags: <br><br>
<ul>

<li> <b>"B:"</b> This only prints out "basic" methods that are useful for the given value of <em>w </em>. It omits <b>"SHIFT"</b> and other
methods that are never really going to be useful.</li><br>

<li> <b> "A:"</b> In constrast, this specifies to print "all" methods. </li><br>

<li> <b>"D:"</b> This includes the <b>"EUCLID"</b> and <b>"MATRIX"</b> methods for division. By default, they are not included. </li><br>

<li> <b>"C:"</b> This includes the <b>"CAUCHY"</b> methods for region multiplication. By default, it is not included.</li> <br>
</ul>
<p>
You may specify multiple of these as the second argument. If you include both <b>"B"</b> and <b>"A,"</b> then it uses the last
one specified. </p>
<p>
The last argument determines the output format of <b>gf_methods.</b> If it is <b>"L,"</b> then it simply lists methods. If it
is <b>"U,"</b> then the output contains <b>gf_unit</b> commands for each of the methods. For the others, the output contains
<b>gf_time_tool.sh</b> commands for <b>M </b>ultiplication,<b>D</b>ivision,<b>R</b>egion multiplications with multiple buffer sizes, and the
<b>B</b>est region multiplication. </p>
<p>
<b>gf_methods</b> enumerates combinations of flags, and calls <b>create_gf_from_argv()</b> to see if the combinations are
supported. Although it enumerates a large number of combinations, it doesn't enumerate all possible parameters for
<b>"SPLIT," "GROUP"</b> or <b>"COMPOSITE."</b> </p>

<p>Some examples of calling <b>gf_methods</b> are shown below in section 6.3.2. </p>







<br/>

6 &nbsp &nbsp  <em>  THE DEFAULTS     </em>   <span id="index_number">21 </span> <br><br><br>


<h3>6.3 Testing with <b>gf_unit </b>, <b>gf_time </b>, and time_tool.sh </h3>



<b>gf_unit </b> and <b>gf_time </b> may be used to verify that a combination of arguments works correctly and efficiently on your
platform. If you plan to stray from the defaults, it is probably best to run both tools to ensure there are no issues with
your environment. <b>gf_unit </b> will run a set of unit tests based on the arguments provided to the tool, and <b>gf_time </b> will
time Galois Field methods based on the provided arguments.<br>
The usage of gf_ unit is:<br><br>
<div id="number_spacing">
<b>gf_unit </b> w tests seed method<br><br> </div>
The usage of gf_ time is:<br><br>
<div id="number_spacing">
<b>gf_time </b> w tests seed buffer-size iterations method<br><br>
</div>



The seed is an integer- negative one uses the current time. The tests are specified by a listing of characters. The
following tests are supported (All are supported by <b>gf_time.</b> Only ', 'S' and 'R' are supported by <b>gf_unit</b>):<br><br>

<ul>
<li> <b>'M':</b> Single multiplications</li><br>
<li> <b> 'D':</b> Single divisions</li><br>
<li> <b> 'I':</b> Single inverses</li><br>
<li> <b>'G': </b> Region multiplication of a buffer by a random constant</li><br>
<li> <b>'0': </b> Region multiplication of a buffer by zero (does nothing and<b>bzero()</b>)</li><br>
<li> <b>'1': </b> Region multiplication of a buffer by one (does <b>memcpy()</b> and <b>XOR</b>)</li><br>
<li> <b>'2': </b> Region multiplication of a buffer by two – sometimes this is faster than general multiplication</li><br>
<li> <b>'S':</b> All three single tests</li><br>
<li> <b>'R':</b> All four region tests</li><br>
<li> <b>'A':</b> All seven tests</li><br>
</ul>





<p>Here are some examples of calling <b>gf_unit</b> and <b>gf_time</b> to verify that <b>"-m SPLIT 32 4 -r ALTMAP -"</b> works
in <em>GF(2<sup>32</sup>),</em> and to get a feel for its performance. First, we go to the test directory and call <b>gf_unit:</b> </p><br><br>


<div id="number_spacing">
UNIX> cd test <br>
UNIX> ./gf_unit 32 A -1 -m SPLIT 32 4 -r ALTMAP - <br>
Args: 32 A -1 -m SPLIT 32 4 -r ALTMAP - / size (bytes): 684 <br>
UNIX> <br><br>
</div>

<b>gf_unit</b> reports on the arguments and how may bytes the <b>gf_t</b> consumes. If it discovers any problems or inconsistencies
with multiplication, division or region multiplication, it will report them. Here, there are no problems.
Next, we move to the <b>tools</b> directory and run performance tests on a 10K buffer, with 10,000 iterations of each test:<br><br>


UNIX> cd ../tools <br>
UNIX> ./gf_time 32 A -1 10240 10000 -m SPLIT 32 4 -r ALTMAP -<br>
Seed: 1388435794 <br>
<div id="number_spacing">
<table cellpadding="0" cellspacing="25" style="font-size:19px,font-family: 'Roboto Condensed', sans-serif;
">

<tr>

<td>Multiply:</td> <td>4.090548 s</td> <td> Mops: </td> <td> 24.414 </td> <td>5.968 Mega-ops/s </td> </tr>
<tr><td>Divide:</td> <td> 37.794962 s </td> <td>Mops: </td> <td> 24.414 </td> <td>0.646 Mega-ops/s </td> </tr>
<tr><td>Inverse:</td> <td> 33.709875 s </td> <td> Mops: </td> <td> 24.414 </td> <td> 0.724 Mega-ops/s </td> </tr>
<tr><td>Region-Random: XOR: 0 </td> <td> 0.035210 s </td> <td> MB:</td> <td> 97.656 </td> <td> 2773.527 MB/s </td></tr>
<tr><td>Region-Random: XOR: 1 </td> <td> 0.036081 s</td> <td> MB:</td> <td> 97.656 </td> <td>2706.578 MB/s </td></tr>
<tr><td>Region-By-Zero:XOR: 0 </td> <td> 0.003199 s </tD> <td> MB: </td> <td>97.656 </td> <td> 30523.884 MB/s </td> </tr>
<tr><td>Region-By-Zero: XOR: 1 </td> <td> 0.000626 s  </td> <td>MB: </td> <td> 97.656 </td> <td> 156038.095 MB/s </td></tr>

</table>
</div>










<br/>

6 &nbsp &nbsp  <em>  THE DEFAULTS     </em>   <span id="index_number">22 </span> <br><br><br>

<div id="number_spacing">
<table cellpadding="0" cellspacing="10" style="font-family: 'Roboto Condensed', sans-serif;
">

<tr>
<td>Region-By-One: XOR: 0</td> <td> 0.003810 s</td> <td> MB:</td> <td> 97.656 </td> <td> 25628.832 MB/s </td>
<tr><td>Region-By-One: XOR: 1 </td> <td> 0.008363 s </td> <td> MB:</td> <td> 97.656 </tD> <td>11677.500 MB/s </td></tr>

<tr><td>Region-By-Two: XOR: 0 </td> <td>0.032942 s  </td> <td>MB: </td> <td> 97.656 </td> <td> 2964.486 MB/s </td> </tr>
<tr><td>Region-By-Two: XOR: 1 </td> <td> 0.033488 s </td> <td> MB: </td> <td> 97.656 </td> <td> 2916.153 MB/s </td> </tr> </table>
</div>
UNIX><br><br>

<p>The first column of output displays the name of the test performed. Region tests will test with and without the XOR
flag being set (see Section 4.3 for an example). The second column displays the total time the test took to complete
measured in seconds (s). The third column displays the size of the test measured in millions of operations (Mops) for
single tests and in Megabytes (MB) for the region tests. The final column displays the speed of the tests calculated
from the second and third columns, and is where you should look to get an idea of a method's performance.</p>
<p>
If the output of <b>gf_unit</b> and <b>gf_time</b> are to your satisfaction, you can incorporate the method into application code
using create <b>gf_from_argv()</b> or <b>gf_init hard().</b></p>
<p>
The performance of "Region-By-Zero" and "Region-By-One" will not change from test to test, as all methods make
the same calls for these. "Region-By-Zero" with "XOR: 1" does nothing except set up the tests. Therefore, you may
use it as a control.</p>

<h3>6.3.1 &nbsp &nbsp &nbsp time tool.sh </h3> 

Finally, the shell script <b>time_tool.sh</b> makes a bunch of calls to <b>gf_time</b> to give a rough estimate of performance. It is
called as follows:<br><br>
usage sh time_tool.sh M|D|R|B w method<br><br>


<p>The values for the first argument are <b>MDRB,</b> for <b>M</b>ultiplication, <b>D</b>ivision,<b>R</b>egion multiplications with multiple
buffer sizes, and the <b>B</b>est region multiplication. For the example above, let's call <b>time_tool.sh</b> to get a rough idea of
performance: </p><br><br>

<div id="number_spacing">
UNIX> sh time_tool.sh M 32 -m SPLIT 32 4 -r ALTMAP - <br>
M speed (MB/s): 6.03 W-Method: 32 -m SPLIT 32 4 -r ALTMAP - <br>
UNIX> sh time_tool.sh D 32 -m SPLIT 32 4 -r ALTMAP - <br>
D speed (MB/s): 0.65 W-Method: 32 -m SPLIT 32 4 -r ALTMAP - <br>
UNIX> sh time_tool.sh R 32 -m SPLIT 32 4 -r ALTMAP - <br>

<table cellpadding="0" cellspacing="10" style="font-family: 'Roboto Condensed', sans-serif;
">

<tr>
<td>Region Buffer-Size:</td> <td> 16K (MB/s):</td> <td>3082.91</td><td> W-Method: 32 </td> <td>-m SPLIT 32 4 </td> <td>-r ALTMAP -</td> </tr>
<tr><td>Region Buffer-Size:</td> <td>32K (MB/s): </td> <td>3529.07 </td><td> W-Method: 32 </td> <td>-m SPLIT 32 4 </td> <td>-r ALTMAP -</td> </tr>
<tr><td>Region Buffer-Size:</td> <td>64K (MB/s): </td> <td> 3749.94</td><td> W-Method: 32 </td> <td>-m SPLIT 32 4 </td> <td>-r ALTMAP -</td> </tr>
<tr><td>Region Buffer-Size:</td> <td>128K (MB/s):</td> <td>3861.27 </td> <td>W-Method: 32 </td> <td>-m SPLIT 32 4 </td> <td>-r ALTMAP -</td> </tr>
<tr><td>Region Buffer-Size:</td> <td>512K (MB/s):</td> <td>3820.82 </td><td> W-Method: 32 </td> <td>-m SPLIT 32 4 </td> <td>-r ALTMAP -</td> </tr>
<tr><td>Region Buffer-Size:</td> <td>1M (MB/s):</td> <td>3737.41 </td><td> W-Method: 32 </td> <td>-m SPLIT 32 4 </td> <td>-r ALTMAP -</td>  </tr>
<tr><td>Region Buffer-Size:</td> <td>2M (MB/s):</td> <td>3002.90 </td><td> W-Method: 32 </td> <td>-m SPLIT 32 4 </td> <td>-r ALTMAP -</td> </tr>
<tr><td>Region Buffer-Size:</td> <td>4M (MB/s): </td><td>2760.77</td><td> W-Method: 32 </td> <td>-m SPLIT 32 4 </td> <td>-r ALTMAP -</td> </tr>
<tr><td>Region Best (MB/s):</td><td> 3861.27</td><td> W-Method: 32 </td> <td>-m SPLIT 32 4 </td> <td>-r ALTMAP -</td> </tr>
</table>

UNIX> sh time_tool.sh B 32 -m SPLIT 32 4 -r ALTMAP - <br>
Region Best (MB/s): 3929.09  W-Method: 32  -m SPLIT 32 4 -r ALTMAP -</br>
UNIX><br><br>
</div>
<p>
We say that <b>time_tool.sh </b>is "rough" because it tries to limit each test to 5 ms or less. Thus, the time granularity
is fine, which means that the numbers may not be as precise as they could be were the time granularity to be course.
When in doubt, you should make your own calls to <b>gf_time</b> with a lot of iterations, so that startup costs and roundoff
error may be minimized. </p>








<br/>

6 &nbsp &nbsp  <em>  THE DEFAULTS     </em>   <span id="index_number">23 </span> <br><br><br>

<h3>6.3.2 &nbsp &nbsp &nbsp An example of gf methods and time tool.sh </h3><br><br>
Let's give an example of how some of these components fit together. Suppose we want to explore the basic techniques
in <em>GF(2<sup>32</sup>).</em> First, let's take a look at what <b>gf_methods</b> suggests as "basic" methods: <br><br>
<div id="number_spacing">
UNIX> gf_methods 32 -B -L <br>
w=32: - <br>
w=32: -m GROUP 4 8 - <br>
w=32: -m SPLIT 32 4 - <br>
w=32: -m SPLIT 32 4 -r ALTMAP - <br>
w=32: -m SPLIT 32 8 - <br>
w=32: -m SPLIT 8 8 - <br>
w=32: -m COMPOSITE 2 - - <br>
w=32: -m COMPOSITE 2 - -r ALTMAP - <br>
UNIX> <br><br>
</div>


<p>

You'll note, this is on my old Macbook Pro, which doesn't support (PCLMUL), so <b>"CARRY FREE"</b> is not included
as an option. Now, let's run the unit tester on these to make sure they work, and to see their memory consumption: </p><br><br>

<div id="number_spacing">
UNIX> gf_methods 32 -B -U <br>
../test/gf_unit 32 A -1 - <br>
../test/gf_unit 32 A -1 -m GROUP 4 8 - <br>
../test/gf_unit 32 A -1 -m SPLIT 32 4 - <br>
../test/gf_unit 32 A -1 -m SPLIT 32 4 -r ALTMAP - <br>
../test/gf_unit 32 A -1 -m SPLIT 32 8 - <br>
../test/gf_unit 32 A -1 -m SPLIT 8 8 - <br>
../test/gf_unit 32 A -1 -m COMPOSITE 2 - - <br>
../test/gf_unit 32 A -1 -m COMPOSITE 2 - -r ALTMAP - <br>
UNIX> gf_methods 32 -B -U | sh <br>
Args: 32 A -1 - / size (bytes): 684 <br>
Args: 32 A -1 -m GROUP 4 8 - / size (bytes): 1296 <br>
Args: 32 A -1 -m SPLIT 32 4 - / size (bytes): 684 <br>
Args: 32 A -1 -m SPLIT 32 4 -r ALTMAP - / size (bytes): 684 <br>
Args: 32 A -1 -m SPLIT 32 8 - / size (bytes): 4268 <br>
Args: 32 A -1 -m SPLIT 8 8 - / size (bytes): 1839276 <br>
Args: 32 A -1 -m COMPOSITE 2 - - / size (bytes): 524648 <br>
Args: 32 A -1 -m COMPOSITE 2 - -r ALTMAP - / size (bytes): 524648 <br>
UNIX> <br> <br>
</div>
<p>
As anticipated, <b>"SPLIT 8 8"</b> consumes quite a bit of memory! Now, let's see how well they perform with both
single multiplications and region multiplications: </p> <br><br>
<div id="number_spacing">
UNIX> gf_methods 32 -B -M <br>
sh time_tool.sh M 32 - <br>
sh time_tool.sh M 32 -m GROUP 4 8  - <br>
sh time_tool.sh M 32 -m SPLIT 32 4 - <br>
sh time_tool.sh M 32 -m SPLIT 32 4 -r ALTMAP -<br>
sh time_tool.sh M 32 -m SPLIT 32 8 - <br>
sh time_tool.sh M 32 -m SPLIT 8 8 - <br>

</div>








<br/>

6 &nbsp &nbsp  <em>  THE DEFAULTS     </em>   <span id="index_number">24 </span> <br><br><br>

<div id="number_spacing">
sh time_tool.sh M 32 -m COMPOSITE 2 - <br>
sh time_tool.sh M 32 -m COMPOSITE 2 - -r ALTMAP <br>
UNIX> gf_methods 32 -B -M | sh
M speed (MB/s): 5.90 W-Method: 32 <br>
M speed (MB/s): 14.09 W-Method: 32 -m GROUP 4 8 <br>
M speed (MB/s): 5.60 W-Method: 32 -m SPLIT 32 4 <br>
M speed (MB/s): 5.19 W-Method: 32 -m SPLIT 32 4 -r ALTMAP <br>
M speed (MB/s): 5.98 W-Method: 32 -m SPLIT 32 8 <br>
M speed (MB/s): 22.10 W-Method: 32 -m SPLIT 8 8 <br>
M speed (MB/s): 34.98 W-Method: 32 -m COMPOSITE 2 - <br>
M speed (MB/s): 34.16 W-Method: 32 -m COMPOSITE 2 - -r ALTMAP <br>
UNIX> gf_methods 32 -B -B | sh
Region Best (MB/s): 2746.76 W-Method: 32 <br>
Region Best (MB/s): 177.06 W-Method: 32 -m GROUP 4 8 <br>
Region Best (MB/s): 2818.75 W-Method: 32 -m SPLIT 32 4 <br>
Region Best (MB/s): 3818.21 W-Method: 32 -m SPLIT 32 4 -r ALTMAP <br>
Region Best (MB/s): 728.68 W-Method: 32 -m SPLIT 32 8 <br>
Region Best (MB/s): 730.97 W-Method: 32 -m SPLIT 8 8 <br>
Region Best (MB/s): 190.20 W-Method: 32 -m COMPOSITE 2 - <br>
Region Best (MB/s): 1837.99 W-Method: 32 -m COMPOSITE 2 - -r ALTMAP <br>
UNIX>
</div>
<p>
The default is quite a bit slower than the best performing methods for both single and region multiplication. So
why are the defaults the way that they are? As detailed at the beginning of this chapter, we strive for lower memory
consumption, so we don't use <b>"SPLIT 8 8,"</b> which consumes 1.75MB.We don't implement alternate fields by default,
which is why we don't use <b>"COMPOSITE."</b> Finally, we don't implement alternate mappings of memory by default,
which is why we don't use "<b>-m SPLIT 32 4 -r ALTMAP -.</b>"</p>

<p>Of course, you may change these defaults if you please.</p>
<p>
<b>Test question:</b> Given the numbers above, it would appear that <b>"COMPOSITE"</b> yields the fastest performance of
single multiplication, while "SPLIT 32 4" yields the fastest performance of region multiplication. Should I use two
gf t's in my application – one for single multiplication that uses <b>"COMPOSITE,"</b> and one for region multiplication
that uses <b>"SPLIT 32 4?"</b></p>
<p>
The answer to this is "no." Why? Because composite fields are different from the "standard" fields, and if you mix
these two <b>gf_t</b>'s, then you are using different fields for single multiplication and region multiplication. Please read
section 7.2 for a little more information on this.</p>

<h3>6.4 &nbsp &nbsp &nbspCalling gf_init_hard()</h3>

We recommend that you use <b>create_gf_from_argv()</b> instead of <b>gf_init_hard().</b> However, there are extra things that
you can do with <b>gf_init_hard().</b> Here's the prototype:<br><br>
<div id="number_spacing">
int gf_init_hard(gf_t *gf<br>
<div style="padding-left:100px">
int w<br>
int mult_type<br>
int region_type<br>
int divide_type<br>
uint64_t prim_poly<br>
int arg1<br>
int arg2<br>
</div>
</div>








<br/>

6 &nbsp &nbsp  <em>  THE DEFAULTS     </em>   <span id="index_number">25 </span> <br><br><br>
<div id="number_spacing">
<div style="padding-left:100px">
GFP base_gf, <br>
void *scratch_memory); </div><br><br>


The arguments mult type, region type and divide type allow for the same specifications as above, except the
types are integer constants defined in gf complete.h: <br><br>
typedef enum {GF_MULT_DEFAULT,<br>
<div style="padding-left:124px">
GF_MULT_SHIFT<br>
GF_MULT_CARRY_FREE<br>
GF_MULT_GROUP<br>
GF_MULT_BYTWO_p<br>
GF_MULT_BYTWO_b<br>
GF_MULT_TABLE<br>
GF_MULT_LOG_TABLE<br>
GF_MULT_LOG_ZERO<br>
GF_MULT_LOG_ZERO_EXT<br>
GF_MULT_SPLIT_TABLE<br>
GF_MULT_COMPOSITE } gf_mult_type_t;<br><br>

</div>

#define GF_REGION_DEFAULT (0x0)<br>
#define GF_REGION_DOUBLE_TABLE (0x1) <br>
#define GF_REGION_QUAD_TABLE (0x2) <br>
#define GF_REGION_LAZY (0x4) <br>
#define GF_REGION_SSE (0x8) <br>
#define GF_REGION_NOSSE (0x10) <br>
#define GF_REGION_ALTMAP (0x20) <br>
#define GF_REGION_CAUCHY (0x40) <br><br>
typedef enum { GF_DIVIDE_DEFAULT<br>
<div style="padding-left:130px">GF_DIVIDE_MATRIX<br>
GF_DIVIDE_EUCLID } gf_division_type_t;<br><br>
</div>
</div>
<p>
You can mix the region types with bitwise or. The arguments to <b>GF_MULT_GROUP,GF_MULT_SPLIT_TABLE</b>
and <b>GF_MULT_COMPOSITE</b> are specified in arg1 and arg2. <b>GF_MULT_COMPOSITE</b> also takes a base field
in <b>base_gf.</b> The base field is itself a <b>gf_t,</b> which should have been created previously with <b>create_gf_fro_argv(),</b>
<b>gf_init_easy()</b> or <b>gf_init_hard().</b> Note that this <b>base_gf</b> has its own <b>base_gf</b> member and can be a composite field
itself.</p>
<p>
You can specify an alternate polynomial in <b>prim_poly.</b> For <em>w </em>&#8804 32, the leftmost one (the one in bit position <em>w</em>) is
optional. If you omit it, it will be added for you. For <em>w </em> = 64, there's no room for that one, so you have to leave it off.
For <em>w </em>= 128, your polynomial can only use the bottom-most 64 bits. Fortunately, the standard polynomial only uses
those bits. If you set <b>prim_poly</b> to zero, the library selects the "standard" polynomial.
</p>
<p>
Finally, <b>scratch_memory</b> is there in case you don't want <b>gf_init_hard()</b> to call <b>malloc()</b>. Youmay call <b>gf_scratch_size()</b>
to find out how much extra memory each technique uses, and then you may pass it a pointer for it to use in <b>scratc_memory.</b>
If you set scratch memory to NULL, then the extra memory is allocated for you with <b>malloc().</b> If you use <b>gf_init_easy()</b>
or <b>create_gf_from_argv(),</b> or you use <b>gf_init_hard()</b> and set <b>scratch_memory</b> to <b>NULL,</b> then you should call <b>gf_free()</b>
to free memory. If you use <b>gf_init_hard()</b> and use your own <b>scratch_memory</b> you can still call <b>gf_free(),</b> and it will
not do anything.</p>
<p>
Both <b>gf_init_hard()</b> and <b>gf_scratch_size()</b> return zero if the arguments don't specify a valid <b>gf_t.</b> When that happens,
you can call <b>gf_error()</b> to print why the call failed.</p>








<br/>


6 &nbsp &nbsp  <em>  FURTHER INFORMATION ON OPTIONS AND ALGORITHMS     </em>   <span id="index_number">26  </span> <br><br><br>


<p>We'll give you one example of calling <b>gf_ init_hard().</b> Suppose you want to make a <b>gf_ init_hard()</b> call to be
equivalent to "-m SPLIT 16 4 -r SSE -r ALTMAP -" and you want to allocate the scratch space yourself. Then you'd
do the following:</p><br><br>

<div id="number_spacing">
gf_t gf; <br>
void *scratch; <br>
int size; <br>
size = gf_scratch_size(16, GF_MULT_SPLIT_TABLE,<br>
GF_REGION_SSE | GF_REGION_ALTMAP,<br>
GF_DIVIDE_DEFAULT,<br>
16, 4); <br>
if (size == 0) { gf_error(); exit(1); } /* It failed. That shouldn’t happen */<br>
scratch = (void *) malloc(size); <br>
if (scratch == NULL) { perror("malloc"); exit(1); } <br>
if (!gf_init_hard(&gf, 16, GF_MULT_SPLIT_TABLE, <br>
GF_REGION_SSE | GF_REGION_ALTMAP, <br>
GF_DIVIDE_DEFAULT,<br>
0, 16, 4, NULL, scratch)) { <br>
gf_error(); <br>
exit(1); <br>
} <br>

</div>


<h3>6.5 &nbsp   &nbsp   gf_size() </h3>

You can call <b>gf_size(gf_t *gf)</b> to learn the memory consumption of the <b>gf_t.</b> It returns all memory consumed by the
<b>gf_t,</b> including the <b>gf_t</b> itself, any scratch memory required by the gf_ t, and the memory consumed by the sub-field
if the field is <b>"COMPOSITE."</b> If you provided your own memory to <b>gf_init_hard(),</b> it does not report the size of
this memory, but what the size should be, as determined by <b>gf_scratch size(). gf_ unit() </b> prints out the return value of
<b>gf_size()</b> on the given field.


<h2>7 &nbsp Further Information on Options and Algorithms </h2>
<h3>
7.1 &nbsp Inlining Single Multiplication and Division for Speed </h3>

Obviously, procedure calls are more expensive than single instructions, and the mechanics of multiplication in <b>"TABLE"</b>
and <b>"LOG"</b> are pretty simple. For that reason, we support inlining for <b>"TABLE"</b> when <em>w </em> = 4 and <em>w </em> = 8, and
for <b>"LOG"</b> when <em>w </em> = 16. We elaborate below.
<p>
When <em>w </em> = 4, you may inline multiplication and division as follows. The following procedures return pointers to
the multiplication and division tables respectively: </p> <br><br>

<div id="number_spacing">
uint8_t *gf_w4_get_mult_table(gf_t * gf);<br>
uint8_t *gf_w4_get_div_table(gf_t * gf);<br><br>
</div>
<p>The macro <b>Gf_W4_INLINE_MULTDIV </b>(<em>table, a, b</em>) then multiplies or divides <em>a </em> by <em>b</em> using the given table. This
of course only works if the multiplication technique is <b>"TABLE,"</b> which is the default for <em>w </em> = 4. If the multiplication
technique is not <b>"TABLE,"</b> then <b>gf_w4_get_mult_table()</b> will return <b>NULL.</b></p>








<br/>


6 &nbsp &nbsp  <em>  FURTHER INFORMATION ON OPTIONS AND ALGORITHMS     </em>   <span id="index_number">27  </span> <br><br><br>




<p>When <em>w </em> = 8, the procedures <b>gf_w8_et_mult_table()</b> and <b>gf_ w8_get_div_table(),</b> and the macro </p>

<b>GF_W8_INLINE_MULTDIV </b>(<em>table, a, b</em>) work identically to the <em>w </em> = 4 case.

<p>When <em>w </em> = 16, the following procedures return pointers to the logarithm table, and the two inverse logarithm tables
respectively: </p><br>

<div id="number_spacing">
uint16_t *gf_w16_get_log_table(gf_t * gf); <br>
uint16_t *gf_w16_get_mult_alog_table(gf_t * gf);<br>
uint16_t *gf_w16_get_div_alog_table(gf_t * gf);<br>

</div>
<br>
<p>
The first inverse logarithm table works for multiplication, and the second works for division. They actually point
to the same table, but to different places in the table. You may then use the macro <b>GF_W16_INLINE_MULT</b>(<em>log,
alog, a, b </em>) to multiply <em>a</em> and <em>b</em>, and the macro <b>GF_W16_INLINE_DIV </b>(<em>log, alog, a, b </em>) to divide a and b. Make
sure you use the <em>alog</em> table returned by <b>gf_w16_get_mult_alog_table()</b> for multiplication and the one returned by
<b>gf_w16_get_div_alog_table()</b> for division. Here are some timings: </p> <br><br>


UNIX> gf_time 4 M 0 10240 10240 - <br>
Seed: 0 <br>
Multiply: 0.228860 s Mops: 100.000 436.949 Mega-ops/s <br>
UNIX> gf_inline_time 4 0 10240 10240 <br>
Seed: 0 <br>
Inline mult: 0.096859 s Mops: 100.000 1032.424 Mega-ops/s <br>
UNIX> gf_time 8 M 0 10240 10240 - <br>
Seed: 0 <br>
Multiply: 0.228931 s Mops: 100.000 436.812 Mega-ops/s <br>
UNIX> gf_inline_time 8 0 10240 10240 <br>
Seed: 0 <br>
Inline mult: 0.114300 s Mops: 100.000 874.889 Mega-ops/s <br>
UNIX> gf_time 16 M 0 10240 10240 - <br>
Seed: 0 <br>
Multiply: 0.193626 s Mops: 50.000 258.229 Mega-ops/s <br>
UNIX> gf_inline_time 16 0 10240 10240 <br>
Seed: 0 <br>
Inline mult: 0.310229 s Mops: 100.000 322.342 Mega-ops/s <br>
UNIX> <br> <br>

<h3>
7.2 &nbsp &nbsp Using different techniques for single and region multiplication </h3>


You may want to "mix and match" the techniques. For example, suppose you'd like to use "-m SPLIT 8 8" for
<b>multiply()</b> in <em>GF(2<sup>32</sup>),</em> because it's fast, and you don't mind consuming all of that space for tables. However, for
<b>multiply_region(),</b> you'd like to use "-m SPLIT 32 4 -r ALTMAP," because that's the fastest way to implement
<b>multiply_region().</b> Unfortunately, There is no way to create a <b>gf_t</b> that does this combination. In this case, you should
simply create two <b>gf_t's,</b> and use one for <b>multiply()</b> and the other for <b>multiply_region().</b> All of the implementations
may be used interchangably with the following exceptions:

<ul>
<li>
<b>"COMPOSITE"</b> implements a different Galois Field. </li><br>

<li>If you change a field's polynomial, then the resulting Galois Field will be different. </li>

</ul>








<br/>


6 &nbsp &nbsp  <em>  FURTHER INFORMATION ON OPTIONS AND ALGORITHMS     </em>   <span id="index_number">28  </span> <br><br><br>

<ul>
<li>

If you are using <b>"ALTMAP"</b> to multiply regions, then the contents of the resulting regions of memory will
depend on the multiplication technique, the size of the region and its alignment. Please see section 7.9 for a
detailed explanation of this. </li>

<li>If you are using <b>"CAUCHY"</b> to multiply regions, then like <b>"ALTMAP,"</b> the contents of the result regions of
memory the multiplication technique and the size of the region. You don't have to worry about alignment. </li>

<h3>7.3 &nbsp &nbsp General <em>w </em>  </h3>
The library supports Galois Field arithmetic with 2 < <em>w </em> &#8804 32. Values of <em>w </em> which are not whole number powers of
2 are handled by the functions in <b>gf_wgen.c</b> . For these values of <em>w </em>, the available multiplication types are <b>"SHIFT,"
"BYT<em>w </em>O p," "BYT<em>w </em>O b," "GROUP," "TABLE"</b> and <b>"LOG." "LOG" </b> is only valid for <em>w </em> < 28 and <b>"TABLE"</b>

is only valid for <em>w </em> < 15. The defaults for these values of <em>w </em> are <b>"TABLE"</b> for <em>w </em> < 8, <b>"LOG"</b> for <em>w </em> < 16, and
<b>"BYT<em>w </em>O p"</b> for <em>w </em> < 32.<br><br>

<h3>7.4 Arguments to "SPLIT" </h3>

The "SPLIT" technique is based on the distributive property of multiplication and addition: <br><br>
<center>
a * (b + c) = (a * b) + (a * c). </center>
<br>
This property allo<em>w </em>s us to, for example, split an eight bit <em>w </em>ord into t<em>w </em>o four-bit components and calculate the product
by performing t<em>w </em>o table lookups in 16-element tables on each of the compoents, and adding the result. There is much
more information on <b>"SPLIT"</b> in The Paper. Here <em>w </em>e describe the version of <b>"SPLIT"</b> implemented in GF-Complete.

<p>
<b>"SPLIT"</b> takes t<em>w </em>o arguments, <em>w </em>hich are the number of bits in each component of a, <em>w </em>hich <em>w </em>e call <em>w </em><sub>a</sub>, and the
number of bits in each component of b, <em>w </em>hich <em>w </em>e call <em>w </em><sub>b.</sub> If the t<em>w </em>o differ, it does not matter <em>w </em>hich is bigger - the
library recognizes this and performs the correct implementation. The legal values of <em>w </em><sub>a</sub> and <em>w </em><sub>b</sub> fall into five categories:
</p><br>


<ol>
<li>
 <em>w </em><sub>a</sub> is equal to <em>w </em> and <em>w </em><sub>b</sub> is equal to four. In this case, b is broken up into <em>w </em>/4
four-bit <em>w </em>ords <em>w </em>hich are used
in 16-element lookup tables. The tables are created on demand in <b>multiply_region()</b> and the SSSE3 instruction

<b>mm_shuffle_epi8()</b> is leveraged to perform 16 lookups in parallel. Thus, these are very fast implementations.
<em>w </em>hen <em>w </em> &#8805 16, you should combine this <em>w </em>ith <b>"ALTMAP"</b> to get the best performance (see The Paper
or [PGM13b] for explanation). If you do this please see section 7.9 for information about <b>"ALTMAP"</b> and
alignment.<br><br>


If you don't use <b>"ALTMAP,"</b> the implementations for <em>w </em> &#8712 {16, 32, 64} convert the standard representation into
<b>"ALTMAP,"</b> perform the multiplication <em>w </em>ith <b>"ALTMAP"</b> and then convert back to the standard representation.
The performance difference using <b>"ALTMAP"</b> can be significant: <br><br><br>

<div id="number_spacing">
<center>
<div id="table_page28">
<table cellpadding="6" cellspacing="0" style="text-align:center;font-size:19px">
<tr>
<td> gf time 16 G 0 1048576 100 -m SPLIT 16 4 -</td> <td>Speed = 8,389 MB/s </td> 
</tr>
<tr>
<td>gf time 16 G 0 1048576 100 -m SPLIT 16 4 -r ALTMAP - </td> <td>Speed = 8,389 MB/s </td> 
</tr>

<tr>
<td>gf time 32 G 0 1048576 100 -m SPLIT 32 4 -</td> <td> Speed = 5,304 MB/s</td> 
</tr>
<tr>
<td>gf time 32 G 0 1048576 100 -m SPLIT 32 4 -r ALTMAP -</td> <td> Speed = 7,146 MB/s</td> 
</tr>


<tr>
<td>gf time 64 G 0 1048576 100 -m SPLIT 64 4 - </td> <td>Speed = 2,595 MB/s </td> 
</tr>

<tr>
<td>gf time 64 G 0 1048576 100 -m SPLIT 64 4 -r ALTMAP - </td> <td>Speed = 3,436 MB/s </td> 
</tr>
</div>



</table>
</div>









<br/>


6 &nbsp &nbsp  <em>  FURTHER INFORMATION ON OPTIONS AND ALGORITHMS     </em>   <span id="index_number">29  </span> <br><br><br>

<ol style="list-style-type:none">


<li>2. &nbsp w<sub>a</sub> is equal to <em>w </em> and w<sub>b</sub> is equal to eight. Now, b is broken into bytes, each of these is used in its own 256-element
lookup table. This is typically the best w<sub>a</sub>y to perform <b>multiply_region()</b> without SSE.</li> 
Because this is a region optimization, when you specify these options, you get a default <b>multiply()</b>  see
Table 1 for a listing of the defaults. See section 7.2 for using a different <b>multiply()</b> than the defaults.<br><br>


<li>
3. &nbsp w<sub>a</sub> is equal to <em>w </em> and <em>w </em><sub>b</sub> is equal to 16. This is only valid for <em>w </em> = 32 and <em>w </em> = 64. No<em>w </em>, b is broken into shorts,
each of these is used in its own 64K-element lookup table. This is typically slower than when <em>w </em><sub>b</suB> equals 8, and
requires more amortization (larger buffer sizes) to be effective. </li><br>


<li>4. &nbsp <em>w </em><sub>a</sub> and <em>w </em><sub>b</sub> are both equal to eight. Now  both <em>a</em> and <em>b</em> are broken into bytes, 
and the products of the various bytes
are looked up in multiple 256 &#215 256 tables. In <em>GF(2<sup>16</sup>),</em> there are three of these tables. In <em>GF(232),</em> there are
seven, and in <em>GF(2<sup>64</sup>)</em> there are fifteen. Thus, this implementation can be a space hog. How ever, for <em>w </em> = 32,
this is the fastest way to perform <b>multiply()</b> on some machines.
when this option is employed, <b>multiply_region()</b> is implemented in an identical fashion to when <em>w </em><sub>a</sub> = <em>w </em>
and <em>w </em><sub>b</sub> = 8. </li><br>

<li>5.&nbsp w<sub>a</sub> = 32 and w<sub>b</sub> = 2. (<em>w</em> = 32 only). I was playing with a different way to use <b>mm_shuffle_epi8().</b> It works,
but it's slower than when w<sub>b</sub> = 4.
</li>

</ul>



<h2>7.5 &nbsp&nbsp Arguments to "GROUP" </h3>

The <b>"GROUP"</b> multiplication option takes t<em>w </em>o arguments, g<sub>s</sub> and g<sub>r</sub>. It implements multiplication in the same manner
as <b>"SHIFT,"</b> except it uses a table of size 2<sup>gs</sup> to perform g<sup>s</sup> shifts at a time, and a table of size 2<sup>gr</sup> to perform g<sup>r</sup>
reductions at at time. The program <b>gf_methods</b> only prints the options 4 4 and 4 8 as arguments for <b>"GROUP."</b>
However, other values of g<sub>s</sub> and g<sub>r</sub> are legal and sometimes desirable: <br><br>

<ol>
<li>
 For <em>w </em> &#8804 32 and <em>w </em> = 64, any values of g<sub>s</sub> and g<sub>r</sub> may be used, so long as they are less than or equal to <em>w </em> and so
long as the tables fit into memory. There are four exceptions to this, listed belo<em>w </em>. </li><br>
<li> For <em>w </em> = 4, <b>"GROUP"</b> is not supported. </li><br>
<li> For <em>w </em> = 8, <b>"GROUP"</b> is not supported. </li><br>
<li> For <em>w </em> = 16, <b>"GROUP"</b> is only supported for gs = gr = 4. </li><br>
<li> For <em>w </em> = 128 <b>"GROUP"</b> only supports <em>g<sub>s</sub></em> = 4 and <em> g<sub>r</b> </em> &#8712 {4, 8, 16}.</li><br>
</ol>
<p>
The way that gs and gr impact performance is as follows. The <b>"SHIFT"</b> implementation works by performing a
carry-free multiplication in <em>w </em> steps, and then performing reduction in <em>w </em> steps. In "GROUP," the carry-free multiplication
is reduced to  <em>w /</em>g<sub>s</sub>steps, and the reduction is reduced to <em>w /</em>g<sub>r</sub>

. Both require tables. The table for the carry-free
multiplication must be created at the beginning of each <b>multiply()</b> or <b>multiply_region(),</b> while the table for reduction
is created when the <b>gf_t</b> is initialized. For that reason, it makes sense for g<sub>r</sub> to be bigger than g<sub>s.</sub></p>

<p>
To give a flavor for the impact of these arguments, Figure 3 show </em>s the performance of varying g<sub>s</sub> and g<sub>r</sub> for
single multiplication and region multiplication respectively, in <em> GF(2<sup>32</sup>)</em> and <em>GF(2<sup>64</sup>).</em> As the graphs demonstrate,
<b>multiply()</b> performs better <em>w </em>ith smaller values of gs, <em>w </em>hile multiply region() amortizes the creation of the shifting
table, and can tolerate larger values of g<sub>s.</sub> <em>w </em>hen g<sub>s</sub> equals g<sub>r,</sub> there are some optimizations that we hand-encode.
These can be seen clearly in the <b>multiply_region()</b> graphs.
</p>








<br/>
7 &nbsp &nbsp  <em>   FURTHER INFORMATION ON OPTIONS AND ALGORITHMS     </em>   <span id="index_number">30 </span> 


<div id="box_1"> 
 
<div class="image-cell_3"> </div>

<div class="image-cell_4"> </div>
</div>
Figure 3: The performance of <b>multiply()</b> and <b>multiply_region()</b> using <b>"GROUP,"</b> and varying the arguments <br> g<sub>s</sub>
and g<sub>r.</sub> All graphs are heat maps with black equaling zero. The region size is 100KB.

<h3>7.6 &nbspConsiderations with "COMPOSITE" </h3>


As mentioned above, using <b>"ALTMAP"</b> with <b>"COMPOSITE"</b> allows <b>multiply_region()</b> to recursively call <b>multiply_
region(),</b> rather than simply calling <b>multiply()</b> on every word in the region. The difference can be pronounced:<br><br>

<div id="table_page28"><center>

<table cellpadding="6" cellspacing="0" style="text-align:center;font-size:19px"><tr>
<td>
gf time 32 G 0 10240 10240 -m COMPOSITE 2 - -
Speed = 322 MB/s </td> </tr>
<tr>
<td>gf time 32 G 0 10240 10240 -m COMPOSITE 2 - -r ALTMAP -
Speed = 3,368 MB/s </td> </tr>

<tr>
<td>
gf time 32 G 0 10240 10240 -m COMPOSITE 2 -m SPLIT 16 4 -r ALTMAP - -r ALTMAP -
Speed = 3,925 MB/s </td> </tr>
</center>
</table>
</div>


<br><br>
<p>
There is support for performing <b>multiply()</b> inline for the <b>"TABLE"</b> implementations for w &#8712 {4, 8} and for the
"LOG" implementation for <em>w</em> = 16 (see section 7.1). These are leveraged by <b>multiply()</b> in <b>"COMPOSITE,"</b> and
by <b>multiply_region()</b> if you are not using <b>"ALTMAP."</b> To demonstrate this, in the table below, you can see that the
performance of <b>multiply()</b> with <b>"SPLIT 8 4"</b> is 88 percent as fast than the default in <em>w</em> = 8 (which is <b>"TABLE"</b>).
When you use each as a base field for <b>"COMPOSITE"</b> with <em>w</em> = 16, the one with <b>"SPLIT 8 4"</b> is now just 37 percent
as fast. The difference is the inlining of multiplication in the base field when <b>"TABLE"</b> is employed:</p><br><br>

<div id="table_page28" border="0"><center>

    <table cellpadding="6" cellspacing="0" style="text-align:center;font-size:19px">

      <tr><td>gf time 8 M 0 1048576 100 - Speed = 501 Mega-ops/s</td> </tr>
      <tr><td>gf time 8 M 0 1048576 100 -m SPLIT 8 4 - Speed = 439 Mega-ops/s </td> </tr>
      <tr><td>gf time 8 M 0 1048576 100 -m COMPOSITE 2 - - Speed = 207 Mega-ops/s </td> </tr>
      <tr><td>gf time 8 M 0 1048576 100 -m COMPOSITE 2 -m SPLIT 8 4 - - Speed = 77 Mega-ops/s </td> </tr>

    </table> 
    </center>
<br><br>
</div>

You can keep making recursive definitions of composites field if you want. For example, this one's not too slow for
region operations (641 MB/s):








<br/>
<br/>


6 &nbsp &nbsp  <em>  FURTHER INFORMATION ON OPTIONS AND ALGORITHMS     </em>   <span id="index_number">31  </span> <br><br><br>

<div id="number_spacing">
<center>
gf time 128 G 0 1048576 100 -m COMPOSITE 2 <span style="color:red">-m COMPOSITE 2 </span> <span style="color:blue">-m COMPOSITE 2 </span> <br>
<span style="color:rgb(250, 149, 167)">-m SPLIT 16 4 -r ALTMAP -</span> <span style="color:blue">-r ALTMAP -</span> <span style="color:red"> -r ALTMAP -</span> -r ALTMAP -
</center>
</div><br>

<p>Please see section 7.8.1 for a discussion of polynomials in composite fields.</p>

<h2>7.7 &nbsp &nbsp &nbsp "CARRY FREE" and the Primitive Polynomial </h2>


If your machine supports the PCLMUL instruction, then we leverage that in <b>"CARRY FREE."</b> This implementation
first performs a carry free multiplication of two <em>w</em>-bit numbers, which yields a 2<em>w</em>-bit number. It does this with
one PCLMUL instruction. To reduce the 2<em>w</em>-bit number back to a <em>w</em>-bit number requires some manipulation of the
polynomial. As it turns out, if the polynomial has a lot of contiguous zeroes following its leftmost one, the number of
reduction steps may be minimized. For example, with <em>w </em> = 32, we employ the polynomial 0x100400007, because that
is what other libraries employ. This only has 9 contiguous zeros following the one, which means that the reduction
takes four steps. If we instead use 0x1000000c5, which has 24 contiguous zeros, the reduction takes just two steps.
You can see the difference in performance:
<br><br>
<center>
<div id="table_page28">

<table cellpadding="6" cellspacing="0" style="text-align:center;font-size:19px">
<tr>

<td>gf time 32 M 0 1048576 100 -m CARRY FREE - </td> <td> Speed = 48 Mega-ops/s</td> </tr>

<tr><td>gf time 32 M 0 1048576 100 -m CARRY FREE -p 0xc5 -</td> <td> Speed = 81 Mega-ops/s </td> </tr>

</table></center>
</div>
<br><br>

<p>
This is relevant for <em>w </em> = 16 and <em>w </em> = 32, where the "standard" polynomials are sub-optimal with respect to
<b>"CARRY FREE."</b> For <em>w </em> = 16, the polynomial 0x1002d has the desired property. It’s less important, of course,
with <em>w </em> = 16, because <b>"LOG"</b> is so much faster than <b>CARRY FREE.</b> </p>

<h2>7.8 &nbsp  More on Primitive Polynomials </h3>

<h3>7.8.1 &nbsp Primitive Polynomials that are not Primitive </h4>

The library is willing to work with most polynomials, even if they are not primitive or irreducible. For example, the
polynomial x<sup>4</sup> + x<sup>3</sup> +x<sup>2</sup> +x+1 is irreducible, and therefore generates a valid Galois Field for <em>GF(2<sup>4</sup>).</em> However, it
is not primitive, because 2<sup>5</sup> = 1. For that reason, if you use this polynomial, you cannot use the <b>"LOG"</b> method. The
other methods will work fine: <br><br>

<div id="number_spacing">

UNIX> gf_mult 2 2 4 -p 0xf -  <br>
4 <br>
UNIX> gf_mult 4 2 4 -p 0xf - <br>
8 <br>
UNIX> gf_mult 8 2 4 -p 0xf - <br>
15 <br>
UNIX> gf_mult 15  2 4 -p 0xf - <br>
1 <br>
UNIX> gf_div 1 15 4 -p 0xf - <br>
2 <br>
UNIX> gf_div 1 15 4 -p 0xf -m LOG - <br>
usage: gf_div a b w [method] - does division of a and b in GF(2&#710;w) <br>
Bad Method Specification: Cannot use Log tables because the polynomial is not primitive. <br>
UNIX>  <br>
</div>
<p>
If a polynomial is reducible, then it does not define a Galois Field, but instead a ring. GF-Complete attempts to
work here where it can; however certain parts of the library will not work:
</p>






<br/>


6 &nbsp &nbsp  <em>  FURTHER INFORMATION ON OPTIONS AND ALGORITHMS     </em>   <span id="index_number">32  </span> <br><br><br>
<ol>
<li>
Division is a best effort service. The problemis that often quotients are not unique. If <b>divide()</b> returns a non-zero
number, then that number will be a valid quotient, but it may be one of many. If the multiplication technique is
<b>"TABLE,"</b> then if a quotient exists, one is returned. Otherwise, zero is returned. Here are some examples - the
polynomial x<sup>4</sup> + 1 is reducible, and therefore produces a ring. Below, we see that with this polynomal, 1*6 = 6
and 14*6 = 6. Therefore, 6/6 has two valid quotients: 1 and 14. GF-Complete returns 14 as the quotient:</li><br>

<div id="number_spacing">
UNIX> gf_mult 1 6 4 -p 0x1 -<br>
6 <br>
UNIX> gf_mult 14 6 4 -p 0x1 - <br>
6 <br>
UNIX> gf_div 6 6 4 -p 0x1 - <br>
14 <br>
UNIX> <br><br>
</div>


<li>When <b>"EUCLID"</b> is employed for division, it uses the extended Euclidean algorithm for GCD to find a number's
inverse, and then it multiplies by the inverse. The problem is that not all numbers in a ring have inverses. For
example, in the above ring, there is no number <em>a</em> such that 6a = 1. Thus, 6 has no inverse. This means that even
though 6/6 has quotients in this ring, <b>"EUCLID"</b> will fail on it because it is unable to find the inverse of 6. It will
return 0:
</li><br>
<div id="number_spacing">
UNIX> gf_div 6 6 4 -p 0x1 -m TABLE -d EUCLID -<br>
0<br>
UNIX><br>
</div><br>

<li> Inverses only work if a number has an inverse. Inverses may not be unique. </li><br>

<li> <b>"LOG"</b> will not work. In cases where the default would be <b>"LOG,"</b> <b>"SHIFT"</b> is used instead. </li>
</ol>

<p>
Due to problems with division, <b>gf_unit</b> may fail on a reducible polynomial. If you are determined to use such a
polynomial, don't let this error discourage you.
</p>

<h3>7.8.2 Default Polynomials for Composite Fields </h3>

GF-Complete will successfully select a default polynomial in the following composite fields:
<ul>
<li> <em>w </em> = 8 and the default polynomial (0x13) is employed for <em>GF(2<sup>4</sup>)</em></li><br>
<li> w = 16 and the default polynomial (0x11d) is employed for <em>GF(2<sup>8</sup>)</em></li><br>
<li> <em>w </em> = 32 and the default polynomial (0x1100b) is employed for <em>GF(2<sup>16</sup>) </em></li><br>
<li> <em>w </em> = 32 and 0x1002d is employed for <em>GF(2<sup>16</sup>) </em></li><br>
<li> <em>w </em> = 32 and the base field for <em>GF(w<em>16</em>) </em> is a composite field that uses a default polynomial</li><br>
<li> <em>w </em> = 64 and the default polynomial (0x100400007) is employed for <em>GF(2<sup>32</sup>)</em></li><br>
<li> <em>w </em> = 64 and 0x1000000c5 is employed for <em>GF(2<sup>32</sup>) </em></li><br>
<li> <em>w </em> = 64 and the base field for <em>GF(w<sup>32</sup>) </em> is a composite field that uses a default polynomial</li><br>
<li> <em>w </em> = 128 and the default polynomial (0x1b) is employed for <em>GF(2<sup>64</sup>) </em></li><br>
<li> <em>w </em> = 128 and the base field for <em> GF(w<sup>64 </sup>) </em> is a composite field that uses a default polynomial</li><br>
</ul>








<br/>


6 &nbsp &nbsp  <em>  FURTHER INFORMATION ON OPTIONS AND ALGORITHMS     </em>   <span id="index_number">33  </span> <br><br><br>


<h3>7.8.3 The Program gf poly for Verifying Irreducibility of Polynomials </h3>

The program <b>gf_poly</b> uses the Ben-Or algorithm[GP97] to determine whether a polynomial with coefficients in <em> GF(2<sup>w </sup>) </em>
is reducible. Its syntax is:<br><br>
<div id="number_spacing">
gf_poly w method power:coef power:coef ... 
</div>

<br>
<p>You can use it to test for irreducible polynomials with binary coefficients by specifying w = 1. For example, from
the discussion above, we know that x<sup>4</sup> +x+1 and x<sup>4</sup> +x<sup>3</sup> +x<sup>2</sup> +x+1 are both irreducible, but x<sup>4</sup> +1 is reducible.
<b>gf_poly</b> confirms:<p><br>

<div id="number_spacing">
UNIX> gf_poly 1 - 4:1 1:1 0:1 <br>
Poly: x&#710;4 + x + 1 <br>
Irreducible. <br>
UNIX> gf_poly 1 - 4:1 3:1 2:1 1:1 0:1 <rb>
Poly: x&#710;4 + x&#710;3 + x&#710;2 + x + 1 <br>
Irreducible. <br>
UNIX> gf_poly 1 - 4:1 0:1 r<br>
Poly: x&#710;4 + 1 <br>
Reducible. <br>
UNIX> <br>

</div>


<p>
For composite fields <em>GF((2<sup>l</sup>)<sup>2</sup>),</em> we are looking for a value s such that x<sup>2</sup> + sx + 1 is irreducible. That value
depends on the base field. For example, for the default field <em>GF(2<sup>32</sup>),</em> a value of <em>s</em> = 2 makes the polynomial
irreducible. However, if the polynomial 0xc5 is used (so that PCLMUL is fast - see section 7.7), then <em>s</em> = 2 yields a
reducible polynomial, but <em>s</em> = 3 yields an irreducible one. You can use <b>gf_poly</b> to help verify these things, and to help
define s if you need to stray from the defaults:</p> <br>

<div id="number_spacing">
UNIX> gf_poly 32 - 2:1 1:2 0:1<br>
Poly: x&#710;2 + (0x2)x + 1 <br>
Irreducible. <br>
UNIX> gf_poly 32 -p 0xc5 - 2:1 1:2 0:1 <br>
Poly: x&#710;2 + (0x2)x + 1 <br>
Reducible. <br>
UNIX> gf_poly 32 -p 0xc5 - 2:1 1:3 0:1 <br>
Poly: x&#710;2 + (0x3)x + 1 <br>
Irreducible. <br>
UNIX> <br>
</div>

<p>
<b>gf_unit</b> does random sampling to test for problems. In particular, it chooses a random a and a random b, multiplies
them, and then tests the result by dividing it by a and b. When w is large, this sampling does not come close to
providing complete coverage to check for problems. In particular, if the polynomial is reducible, there is a good
chance that <b>gf_unit</b> won't discover any problems. For example, the following <b>gf_unit</b> call does not flag any problems,
even though the polynomial is reducible.</p>
<br>
<div id="number_spacing">
UNIX> gf_unit 64 A 0 -m COMPOSITE 2 -p 0xc5 - -p 2 -<br>
UNIX>
</div>

<p>
How can we demonstrate that this particular field has a problem? Well, when the polynomial is 0xc5, we can factor
x<sup>2</sup> + 2x + 1 as (x + 0x7f6f95f9)(x + 0x7f6f95fb). Thus, in the composite field, when we multiply 0x17f6f95f9 by
0x17f6f95fb, we get zero. That's the problem:
</p>








<br/>


6 &nbsp &nbsp  <em>  FURTHER INFORMATION ON OPTIONS AND ALGORITHMS     </em>   <span id="index_number">34  </span> <br><br><br>

<div id="number_spacing">

UNIX> gf_mult 7f6f95f9 7f6f95fb 32h -p 0xc5 - <br>
1 <br>
UNIX> gf_mult 17f6f95f9 17f6f95fb 64h -m COMPOSITE 2 -p 0xc5 - -p 2 - <br>
0 <br>
UNIX> <br>

</div>

<h2>7.9 "ALTMAP" considerations and extract_word() </h2>

There are two times when you may employ alternate memory mappings:
<ol>
<li> When using <b>"SPLIT"</b> and w<sub>b</sub> = 4. </li>
<li> When using <b>"COMPOSITE."</b> </li>
</ol>

Additionally, by default, the <b>"CAUCHY"</b> region option also employs an alternate memory mapping.

<p>When you use alternate memory mappings, the exact mapping of words in <em> GF(2<sup>w </sup>) </em> to memory depends on the
situation, the size of the region, and the alignment of the pointers. To help you figure things out, we have included the
procedures <b>extract_word.wxx()</b> as part of the <b>gf_t</b> struct. This procedure takes four parameters: </p>
<ul>
<li>A pointer to the <b>gf_t.</b> </li>
<li> The beginning of the memory region. </li>
<li>The number of bytes in the memory region. </li>
<li>The desired word number: <em>n.</em> </li>
</ul>

<p>
It then returns the <em>n</em>-th word in memory. When the standard mapping is employed, this simply returns the <em>n</em>-
th contiguous word in memory. With alternate mappings, each word may be split over several memory regions, so
<b>extract_word()</b> grabs the relevant parts of each memory region to extract the word. Below, we go over each of the
above situations in detail. Please refer to Figure 2 in Section 5 for reference. </p>


<h3>7.9.1 Alternate mappings with "SPLIT" </h3>

The alternate mapping with <b>"SPLIT"</b> is employed so that we can best leverage <b>mm_shuffle_epi8().</b> Please read [PGM13b]
for details as to why. Consider an example when <em>w</em> = 16. In the main region of memory (the middle region in Figure
2), multiplication proceeds in units of 32 bytes, which are each broken into two 16-byte regions. The first region
holds the high bytes of each word in <em>GF(2<sup>16</sup>),</em> and the second region holds the low bytes.
Let's look at a very detailed example, from <b>gf_example_5.c.</b> This program makes the following call, where <b>gf</b> has

been initialized for <em>w</em> = 16, using <b>"SPLIT"</b> and <b>"ALTMAP:"</b><br><br>
<div id="number_spacing">
gf.multiply_region.w32(&gf, a, b, 0x1234, 30*2, 0);
</div><br>


<p>In other words, it is multiplying a region a of 60 bytes (30 words) by the constant 0x1234 in <em> GF(2<sup>16</sup>),</em> and placing
the result into <em>b.</em> The pointers <em>a</em> and <em>b</em> have been set up so that they are not multiples of 16. The first line of output
prints <em>a</em> and <em>b:</em></p><br>

a: 0x10010008c b: 0x10010015c <br><br>

As described in Section 5, the regions of memory are split into three parts:








<br/>


6 &nbsp &nbsp  <em>  FURTHER INFORMATION ON OPTIONS AND ALGORITHMS     </em>   <span id="index_number">35  </span> <br><br><br>


<ol>
<li> 4 bytes starting at 0x1001008c / 0x10010015c. </li>
<li> 32 bytes starting at 0x10010090 / 0x100100160. </li>
<li> 24 bytes starting at 0x100100b0 / 0x100100180. </li>

</ol>


<p>In the first and third parts, the bytes are laid out according to the standard mapping. However, the second part is
split into two 16-byte regions- one that holds the high bytes of each word and one that holds the low bytes. To help
illustrate, the remainder of the output prints the 30 words of <em>a</em> and <em>b</em> as they appear in memory, and then the 30 return
values of <b>extract_word.w32():</b> </p><br>

<div id="number_spacing">
<table cellspacing="6" style="text-align:right">

<tr>
<td></td> <td> 1</td> <td> 2 </td> <td> 3 </td> <td> 4</td> <td> 5 </td> <td> 6 </td> <td> 7</td> <td> 8 </td> <td> 9</td> </tr>
<tr>
<td>a:</td><td> 640b</td> <td> 07e5</td> <td> 2fba </td> <td> ce5d </td> <td> f1f9</td> <td> 3ab8</td> <td> c518 </td> <td> 1d97</td> <td> 45a7</td>
 <td> 0160</td> </tr>
 
<tr><td>b:</td> <td>1ba3</td><td> 644e</td> <td> 84f8</td> <td> be3c</td> <td> 4318</td> <td> 4905</td> <td> b2fb </td> <td> 46eb </td> <td> ef01 </td>
 <td>a503</td> 
</tr>
</table> 
 <br><br>
<table cellspacing="6" style="text-align:right">

<tr>
<td> 10</td> <td> 11 </td> <td> 12</td> <td> 13</td> <td> 14 </td> <td> 15 </td> <td> 16</td> <td> 17</td> <td>18</td> <td> 19 </td></tr>
<tr>
<td>a:</td><td> 3759</td> <td> b107</td> <td> 9660 </td> <td> 3fde </td> <td> b3ea</td> <td> 8a53</td> <td> 75ff </td> <td> 46dc</td> <td> c504</td>
 <td> 72c2</td> </tr>
 
<tr><td>b:</td> <td>da27</td><td> e166</td> <td> a0d2</td> <td> b3a2</td> <td> 1699</td> <td> 3a3e</td> <td> 47fb </td> <td> 39af </td> <td> 1314 </td>
 <td>8e76</td> 
</tr>
</table> 

<table cellspacing="6" style="text-align:right">
<br><br>
<tr>
<td> 20</td> <td> 21 </td> <td> 22</td> <td> 23</td> <td> 24 </td> <td> 25 </td> <td> 26</td> <td> 27</td> <td>28</td> <td> 29 </td></tr>
<tr>
<td>a:</td><td> b469</td> <td> 1b97</td> <td> e91d </td> <td> 1dbc </td> <td> 131e</td> <td> 47e0</td> <td> c11a </td> <td> 7f07</td> <td> 76e0</td>
 <td> fe86</td> </tr>
 
<tr><td>b:</td> <td>937c</td><td> a5db</td> <td> 01b7</td> <td> 7f5f</td> <td> 8974</td> <td> 05e1</td> <td> cff3 </td> <td> a09c </td> <td> de3c </td>
 <td>4ac0</td> 
</tr>
</table> 
<br><br>
<table cellspacing="6">


<tr><td>Word</td><td> 0:</td> <td>0x640b </td><td>*</td> <td>0x1234</td> <td>=</td> <td>0x1ba3 Word 15:</td> <td>0x4575 </td><td>*</td> <td>0x1234</td> <td>=</td> <td>0xef47</td></tr>     
<tr><td>Word</td> <td> 1:</td> <td>0x07e5 </td><td>*</td> <td>0x1234</td> <td>=</td> <td>0x644e Word 16:</td> <td>0x60dc </td><td>*</td> <td>0x1234</td> <td>=</td> <td>0x03af</td></tr>
<tr><td>Word</td> <td> 2:</td> <td>0xba59 </td><td>*</td> <td>0x1234</td> <td>=</td> <td>0xf827 Word 17:</td> <td>0x0146 </td><td>*</td> <td>0x1234</td> <td>=</td> <td>0xa539 </td> </tr>
<tr><td>Word</td> <td>3:</td> <td>0x2f37 </td><td>*</td> <td>0x1234</td> <td>=</td> <td>0x84da Word 18:</td> <td>0xc504 </td><td>*</td> <td>0x1234</td> <td>=</td> <td>0x1314 </td> </tr>
<tr><td>Word</td> <td>4:</td> <td>0x5d07 </td><td>*</td> <td>0x1234</td> <td>=</td> <td>0x3c66 Word 19:</td> <td>0x72c2 </td><td>*</td> <td>0x1234</td> <td>=</td> <td>0x8e76 </td> </tr>
<tr><td>Word</td> <td>5:</td> <td>0xceb1 </td><td>*</td> <td>0x1234</td> <td>=</td> <td>0xbee1 Word 20:</td> <td>0xb469 </td><td>*</td> <td>0x1234</td> <td>=</td> <td>0x937c </td> </tr>
<tr><td>Word</td> <td>6:</td> <td>0xf960 </td><td>*</td> <td>0x1234</td> <td>=</td> <td>0x18d2 Word 21:</td> <td>0x1b97 </td><td>*</td> <td>0x1234</td> <td>=</td> <td>0xa5db </td> </tr>
<tr><td>Word</td> <td>7:</td> <td>0xf196 </td><td>*</td> <td>0x1234</td> <td>=</td> <td>0x43a0 Word 22:</td> <td>0xe91d </td><td>*</td> <td>0x1234</td> <td>=</td> <td>0x01b7 </td> </tr>
<tr><td>Word</td> <td>8:</td> <td>0xb8de </td><td>*</td> <td>0x1234</td> <td>=</td> <td>0x05a2 Word 23:</td> <td>0x1dbc </td><td>*</td> <td>0x1234</td> <td>=</td> <td>0x7f5f </td> </tr>
<tr><td>Word</td> <td>9:</td> <td>0x3a3f </td><td>*</td> <td>0x1234</td> <td>=</td> <td>0x49b3 Word 24:</td> <td>0x131e </td><td>*</td> <td>0x1234</td> <td>=</td> <td>0x8974 </td> </tr>
<tr><td>Word</td> <td>10:</td> <td>0x18ea </td><td>*</td> <td>0x1234</td> <td>=</td> <td>0xfb99 Word 25:</td> <td>0x47e0 </td><td>*</td> <td>0x1234</td> <td>=</td> <td>0x05e1 </td> </tr>
<tr><td>Word</td> <td>11:</td> <td>0xc5b3 </td><td>*</td> <td>0x1234</td> <td>=</td> <td>0xb216 Word 26:</td> <td>0xc11a </td><td>*</td> <td>0x1234</td> <td>=</td> <td>0xcff3  </td> </tr>
<tr><td>Word</td> <td>12:</td> <td>0x9753 </td><td>*</td> <td>0x1234</td> <td>=</td> <td>0xeb3e Word 27:</td> <td>0x7f07 </td><td>*</td> <td>0x1234</td> <td>=</td> <td>0xa09c  </td> </tr>
<tr><td>Word</td> <td>13:</td> <td>0x1d8a </td><td>*</td> <td>0x1234</td> <td>=</td> <td>0x463a Word 28:</td> <td>0x76e0 </td><td>*</td> <td>0x1234</td> <td>=</td> <td>0xde3c  </td> </tr>
<tr><td>Word</td> <td>14:</td> <td>0xa7ff </td><td>*</td> <td>0x1234</td> <td>=</td> <td>0x01fb Word 29:</td> <td>0xfe86 <td>*</td> <td>0x1234</td> <td>=</td> <td>0x4ac0 </td> </tr>

</table>
</div>
<br>
In the first region are words 0 and 1, which are identical to how they appear in memory: 0x640b and 0x07e5. In
the second region are words 2 through 17. These words are split among the two sixteen-byte regions. For example,
word 2, which <b>extract_word()</b> reports is 0xba59, is constructed from the low byte in word 2 (0xba) and the low byte
in word 10 (0x59). Since 0xba59 * 0x1234 = 0xf827, we see that the low byte in word 2 of <em> b </em> is 0xf8, and the low byte
in word 10 is 0x27.
<p>When we reach word 22, we are in the third region of memory, and words are once again identical to how they
appear in memory.</p>

<p>While this is confusing, we stress that that so long as you call <b>multiply_region()</b> with pointers of the same alignment
and regions of the same size, your results with <b>ALTMAP</b> will be consistent. If you call it with pointers of </p>






<br/>


7 &nbsp &nbsp  <em>  FURTHER INFORMATION ON OPTIONS AND ALGORITHMS     </em>   <span id="index_number">36  </span> <br><br><br>

different alignments, or with different region sizes, then the results will not be consistent. To reiterate, if you don't use
<b>ALTMAP,</b> you don't have to worry about any of this - words will always be laid out contiguously in memory.
<p>
When <em>w</em> = 32, the middle region is a multiple of 64, and each word in the middle region is broken into bytes, each
of which is in a different 16-byte region. When <em>w</em> = 64, the middle region is a multiple of 128, and each word is
stored in eight 16-byte regions. And finally, when<em>w</em> = 128, the middle region is a multiple of 128, and each word is
stored in 16 16-byte regions.</p><br>

<h3>7.9.2 &nbsp Alternate mappings with "COMPOSITE" </h3>

With <b>"COMPOSITE,"</b> the alternate mapping divides the middle region in half. The lower half of each word is stored
in the first half of the middle region, and the higher half is stored in the second half. To illustrate, gf example 6
performs the same example as gf example 5, except it is using <b>"COMPOSITE"</b> in GF((2<sup>16</sup>)<sup>2</sup>), and it is multiplying
a region of 120 bytes rather than 60. As before, the pointers are not aligned on 16-bit quantities, so the region is broken
into three regions of 4 bytes, 96 bytes, and 20 bytes. In the first and third region, each consecutive four byte word is a
word in <em>GF(2<sup>32</sup>).</em> For example, word 0 is 0x562c640b, and word 25 is 0x46bc47e0. In the middle region, the low two
bytes of each word come from the first half, and the high two bytes come from the second half. For example, word 1
as reported by <b>extract_word()</b> is composed of the lower two bytes of word 1 of memory (0x07e5), and the lower two
bytes of word 13 (0x3fde). The product of 0x3fde07e5 and 0x12345678 is 0x211c880d, which is stored in the lower
two bytes of words 1 and 13 of <em>b.</em><br><br>

a: 0x10010011c b: 0x1001001ec

<br><br>

<div id="number_spacing">
<table cellspacing="6" style="text-align:right">

<tr>
<td></td> <td> 1</td> <td> 2 </td> <td> 3 </td> <td> 4</td> <td> 5 </td> <td> 6 </td> <td> 7</td> <td> 8 </td> <td> 9</td> </tr>
<tr>
<td>a:</td><td> 562c640b</td> <td> 959407e5</td> <td> 56592fba </td> <td> cbadce5d </td> <td> 1d1cf1f9</td> <td> 35d73ab8</td> <td> 6493c518 </td> <td> b37c1d97</td> 
<td> 8e4545a7</td>
 <td> c0d80160</td> </tr>
 
<tr><td>b:</td> <td>f589f36c</td><td> f146880d</td> <td> 74f7b349</td> <td> 7ea7c5c6</td> <td> 34827c1a</td> <td> 93cc3746</td> <td> bfd9288b </td>
 <td> 763941d1 </td> 
<td> bcd33a5d </td>
 <td>da695e64</td> 
</tr>
</table> 


<br><br>
<table cellspacing="6" style="text-align:right">

<tr>
<td> 10</td> <td> 11 </td> <td> 12</td> <td> 13</td> <td> 14 </td> <td> 15 </td> <td> 16</td> <td> 17</td> <td>18</td> <td> 19 </td></tr>
<tr>
<td>a:</td><td> 965b3759</td> <td> cb3eb107</td> <td> 1b129660 </td> <td> 95a33fde </td> <td> 95a7b3ea</td> <td> d16c8a53</td> <td> 153375ff </td> 
<td> f74646dc</td> <td> 35aac504</td>
 <td> 98f972c2</td> </tr>
 
<tr><td>b:</td> <td>fd70f125</td><td> 3274fa8f</td> <td> d9dd34ee</td> <td> c01a211c</td> <td> d4402403</td> <td> 8b55c08b</td> <td> da45f0ad </td> 
<td> 90992e18 </td> <td> b65e0902 </td>
 <td>d91069b5</td> 
</tr>
</table> 


<table cellspacing="6" style="text-align:right">
<br><br>
<tr>
<td> 20</td> <td> 21 </td> <td> 22</td> <td> 23</td> <td> 24 </td> <td> 25 </td> <td> 26</td> <td> 27</td> <td>28</td> <td> 29 </td></tr>
<tr>
<td>a:</td><td> 5509b469</td> <td> 7f8a1b97</td> <td> 3472e91d </td> <td> 9ee71dbc </td> <td> de4e131e</td> <td> 46bc47e0</td> <td> 5bc9c11a </td>
 <td> 931d7f07</td> <td> c85cfe86</td>
 <td> fe86</td> </tr>
 
<tr><td>b:</td> <td>fc92b8f5</td><td> edd59668</td> <td> b4bc0d90</td> <td> a679e4ce</td> <td> 1a98f7d0</td> <td> 6038765f</td> <td> b2ff333f </td> <td> e7937e49 </td> 
<td> fa5a5867 </td>
 <td>79c00ea2</td> 
</tr>
</table> 
<br><br>


<table cellspacing="6" style="text-align:right">


<tr><td>Word</td><td> 0:</td> <td>0x562c640b </td><td>*</td> <td>0x12345678</td> <td>=</td> <td>0xf589f36c Word 15:</td> <td>0xb46945a7 </td><td>*</td> <td>0x12345678</td> <td>=</td> <td>0xb8f53a5d</td></tr>     
<tr><td>Word</td> <td> 1:</td> <td>0x3fde07e5 </td><td>*</td> <td>0x12345678</td> <td>=</td> <td>0x211c880d Word 16:</td> <td>0x55098e45 </td><td>*</td> <td>0x12345678</td> <td>=</td> <td>0xfc92bcd3</td></tr>
<tr><td>Word</td> <td> 2:</td> <td>0x95a39594 </td><td>*</td> <td>0x12345678</td> <td>=</td> <td>0xc01af146 Word 17:</td> <td>0x1b970160 </td><td>*</td> <td>0x12345678</td> <td>=</td> <td>0x96685e64 </td> </tr>
<tr><td>Word</td> <td>3:</td> <td>0xb3ea2fba </td><td>*</td> <td>0x12345678</td> <td>=</td> <td>0x2403b349 Word 18:</td> <td>0x7f8ac0d8 </td><td>*</td> <td>0x12345678</td> <td>=</td> <td>0xedd5da69 </td> </tr>
<tr><td>Word</td> <td>4:</td> <td>0x95a75659 </td><td>*</td> <td>0x12345678</td> <td>=</td> <td>0xd44074f7 Word 19:</td> <td>0xe91d3759 </td><td>*</td> <td>0x12345678</td> <td>=</td> <td>0x0d90f125 </td> </tr>
<tr><td>Word</td> <td>5:</td> <td>0x8a53ce5d </td><td>*</td> <td>0x12345678</td> <td>=</td> <td>0xc08bc5c6 Word 20:</td> <td>0x3472965b </td><td>*</td> <td>0x12345678</td> <td>=</td> <td>0xb4bcfd70 </td> </tr>
<tr><td>Word</td> <td>6:</td> <td>0xd16ccbad </td><td>*</td> <td>0x12345678</td> <td>=</td> <td>0x8b557ea7 Word 21:</td> <td>0x1dbcb107 </td><td>*</td> <td>0x12345678</td> <td>=</td> <td>0xe4cefa8f </td> </tr>
<tr><td>Word</td> <td>7:</td> <td>0x75fff1f9 </td><td>*</td> <td>0x12345678</td> <td>=</td> <td>0xf0ad7c1a Word 22:</td> <td>0x9ee7cb3e </td><td>*</td> <td>0x12345678</td> <td>=</td> <td>0xa6793274 </td> </tr>
<tr><td>Word</td> <td>8:</td> <td>0x15331d1c </td><td>*</td> <td>0x12345678</td> <td>=</td> <td>0xda453482 Word 23:</td> <td>0x131e9660 </td><td>*</td> <td>0x12345678</td> <td>=</td> <td>0xf7d034ee </td> </tr>
<tr><td>Word</td> <td>9:</td> <td>0x46dc3ab8 </td><td>*</td> <td>0x12345678</td> <td>=</td> <td>0x2e183746 Word 24:</td> <td>0xde4e1b12 </td><td>*</td> <td>0x12345678</td> <td>=</td> <td>0x1a98d9dd </td> </tr>
<tr><td>Word</td> <td>10:</td> <td>0xf74635d7 </td><td>*</td> <td>0x12345678</td> <td>=</td> <td>0x909993cc Word 25:</td> <td>0x46bc47e0 </td><td>*</td> <td>0x12345678</td> <td>=</td> <td>0x6038765f </td> </tr>
<tr><td>Word</td> <td>11:</td> <td>0xc504c518 </td><td>*</td> <td>0x12345678</td> <td>=</td> <td>0x0902288b Word 26:</td> <td>0x5bc9c11a </td><td>*</td> <td>0x12345678</td> <td>=</td> <td>0xb2ff333f  </td> </tr>
<tr><td>Word</td> <td>12:</td> <td>0x35aa6493 </td><td>*</td> <td>0x12345678</td> <td>=</td> <td>0xb65ebfd9 Word 27:</td> <td>0x931d7f07 </td><td>*</td> <td>0x12345678</td> <td>=</td> <td>0xe7937e49  </td> </tr>

</table>
</div>








<br/>


8 &nbsp &nbsp  <em>  THREAD SAFETY     </em>   <span id="index_number">37  </span> <br><br><br>
<div id="number_spacing">
<table cellpadding="6" cellspacing="0">
<tr>
<td>Word 13:</td> <td> 0x72c21d97</td> <td> *</td> <td> 0x12345678</td> <td> =</td> <td> 0x69b541d1</td> <td> Word 28:</tD>

<td> 0xd40676e0 </td> <td> * </td> <td> 0x12345678 </td> <td> = </td> <td> 0xfa5a5867 </td> </tr>

<tr><td>Word 14:</td> <td> 0x98f9b37c</td> <td> * </td> <td> 0x12345678 </td> <td> = </td> <td> 0xd9107639</td> <td> Word 29:</td>
<td> 0xc85cfe86</td> <td>*</td> <td> 0x12345678</td> <td> =</td> <td> 0x79c00ea2</td></tr>

</table>
</div><br>


<p>
As with <b>"SPLIT,"</b> using <b>multiply_region()</b> with <b>"COMPOSITE"</b> and <b>"ALTMAP"</b> will be consistent only if the
alignment of pointers and region sizes are identical. </p>


<h3>7.9.3 The mapping of "CAUCHY" </h3>

With <b>"CAUCHY,"</b> the region is partitioned into <em>w</em> subregions, and each word in the region is broken into <em>w</em> bits,
each of which is stored in a different subregion. To illustrate, <b>gf_example_7</b> multiplies a region of three bytes by 5
in <em>GF(2<sup>3</sup>)</em> using <b>"CAUCHY:"</b><br><br>

<div id="number_spacing">

UNIX> gf_example_7 <br>
a: 0x100100190 b: 0x1001001a0 <br><br>
a: 0x0b 0xe5 0xba <br>
b: 0xee 0xba 0x0b <br><br>
a bits: 00001011 11100101 10111010 <br>
b bits: 11101110 10111010 00001011<br><br>
Word 0: 3 * 5 = 4 <br>
Word 1: 5 * 5 = 7 <br>
Word 2: 2 * 5 = 1 <br>
Word 3: 5 * 5 = 7 <br>  
Word 4: 4 * 5 = 2 <br>
Word 5: 6 * 5 = 3 <br>
Word 6: 2 * 5 = 1 <br>
Word 7: 6 * 5 = 3 <br>
UNIX><br><br> </div>
<p>

The program prints the three bytes of a and b in hexadecimal and in binary. To see how words are broken up,
consider word 0, which is the lowest bit of each of the three bytes of a (and b). These are the bits 1, 1 and 0 in a, and
0, 0, and 1 in b. Accordingly, the word is 3 in a, and 3*5 = 4 in b. Similarly, word 7 is the high bit in each byte: 0, 1, 1
(6) in a, and 1, 1, 0 (3) in b.</p>
<p>With <b>"CAUCHY," multiply_region()</b>may be implemented exclusively with XOR operations. Please see [BKK<sup>+</sup>95]
for more information on the motivation behind <b>"CAUCHY."</b> </p>

<h2>8 &nbsp Thread Safety </h2>

Once you initialize a <b>gf_t,</b> you may use it wontonly in multiple threads for all operations except for the ones below.
With the implementations listed below, the scratch space in the <b>gf_t</b> is used for temporary tables, and therefore you
cannot call <b>region_multiply,</b> and in some cases <b>multiply</b> from multiple threads because they will overwrite each
others' tables. In these cases, if you want to call the procedures from multiple threads, you should allocate a separate
gf_t for each thread:
<ul>
<li>
 All "GROUP" implementations are not thread safe for either <b>region_multiply()</b> or <b> multiply().</b> Other than
<b>"GROUP," multiply() </b> is always thread-safe.

</li>
</ul>









<br/>


9 &nbsp &nbsp  <em>  LISTING OF PROCEDURES     </em>   <span id="index_number">38  </span> <br><br><br>
<ul>
<li>

For <em>w </em> = 4, <b>region_multiply.w32()</b> is unsafe in in "-m TABLE -r QUAD -r LAZY." </li><br>
<li> For <em>w </em> = 8, <b> region_multiply.w32()</b> is unsafe in in "-m TABLE -r DOUBLE -r LAZY."</li><br>
<li> For <em>w </em> = 16, <b>region_multiply.w32() </b> is unsafe in in "-m TABLE."</li><br>
<li> For <em>w </em> &#8712 {32, 64, 128}, all <b>"SPLIT"</b> implementations are unsafe for <b>region_multiply().</b> This means that if the
default uses <b>"SPLIT"</b> (see Table 1 for when that occurs), then <b>region_multiply()</b> is not thread safe.</li><br>
<li> The <b>"COMPOSITE"</b> operations are only safe if the implementations of the underlying fields are safe.</li>
</ul>

<h2>9 &nbspListing of Procedures </h2>

The following is an alphabetical listing of the procedures, data types and global variables for users to employ in
GF-complete.<br>

<ul>
<li> <b>GF_W16_INLINE_DIV()</b> in <b>gf_complete.h:</b> This is a macro for inline division when <em>w </em> = 16. See section 7.1.</li><br>
<li> <b>GF_W16_INLINE_MULT()</b> in <b>gf_complete.h:</b> This is a macro for inline multiplication when <em>w </em> = 16. See
section 7.1.</li><br>
<li> <b>GF_W4_INLINE_MULTDIV()</b> in <b>gf_complete.h:</b> This is a macro for inline multiplication/division when <em>w </em> =
4. See section 7.1.</li><br>

<li> <b>GF_W8_INLINE_MULTDIV()</b> in <b>gf_complete.h:</b> This is a macro for inline multiplication/division when <em>w </em> =
8. See section 7.1.</li><br>
<li> <b>MOA_Fill_Random_Region()</b> in <b>gf_rand.h:</b> Fills a region with random numbers.</li><br>
<li> <b>MOA_Random_128()</b> in <b>gf_rand.h:</b> Creates a random 128-bit number.</li><br>
<li> <b>MOA_Random_32()</b> in <b>gf_rand.h:</b> Creates a random 32-bit number. </li><br>
<li> <b>MOA_Random_64()</b> in <b>gf_rand.h:</b> Creates a random 64-bit number. </li><br>
<li> <b>MOA_Random_W()</b> in <b>gf_rand.h:</b> Creates a random w-bit number, where <em>w </em> &#8804 32. </li><br>
<li> <b>MOA_Seed()</b> in <b>gf_rand.h:</b> Sets the seed for the random number generator. </li><br>
<li> <b>gf_errno</b> in <b>gf_complete.h:</b> This is to help figure out why an initialization call failed. See section 6.1.</li><br>
<li> <b>gf_create_gf_from_argv()</b> in <b>gf method.h:</b> Creates a gf t using C style argc/argv. See section 6.1.1. </li><br>
<li> <b>gf_division_type_t</b> in <b>gf_complete.h:</b> the different ways to specify division when using <b>gf_init_hard().</b> See 
section 6.4. </li><br>
<li> <b>gf_error()</b> in <b>gf_complete.h:</b> This prints out why an initialization call failed. See section 6.1. </li><br>

<li> <b>gf_extract</b> in <b>gf_complete.h:</b> This is the data type of <b>extract_word()</b> in a gf t. See section 7.9 for an example
of how to use extract word().</li>






<br/>


9 &nbsp &nbsp  <em>  LISTING OF PROCEDURES     </em>   <span id="index_number">39  </span> <br><br><br>
<ul>
<li>
<b>gf_free()</b> in <b>gf_complete.h:</b> If <b>gf_init easy(), gf_init hard()</b> or <b>create_gf_from_argv()</b> allocated memory, this
frees it. See section 6.4. </li>

<li> <b>gf_func_a_b</b> in <b>gf_complete.h:</b> This is the data type of <b>multiply()</b> and <b>divide()</b> in a gf_t. See section 4.2 for
examples of how to use <b>multiply()</b> and <b>divide()</b></li><br>

<li> <b>gf_func_a_b</b> in <b>gf_complete.h:</b> This is the data type of <b>multiply()</b> and <b>divide()</b> in a <b>gf_t.</b> See section 4.2 for
examples of how to use <b>multiply()</b> and <b>divide()</b></li><br>

<li> <b>gf_func_a</b> in <b>gf_complete.h:</b> This is the data type of <b>inverse()</b> in a <b>gf_t</b></li><br>

<li> <b>gf_general_add()</b> in <b>gf_general.h:</b> This adds two <b>gf_general_t's </b></li><br>

<li> <b>gf_general_divide()</b> in <b>gf_general.h:</b> This divides two <b>gf_general t's </b></li><br>

<li> <b>gf_general_do_region_check() </b> in <b>gf_general.h:</b> This checks a region multiply of <b>gf_general_t's </b></li><br>

<li> <b>gf_general_do_region_multiply() </b> in <b>gf_general.h:</b> This does a region multiply of <b>gf_general_t's </b></li><br>

<li> <b>gf_general_do_single_timing_test()</b> in <b>gf_general.h:</b> Used in <b>gf_time.c </b></li><br>

<li> <b>gf_general_inverse() </b> in <b>gf_general.h:</b> This takes the inverse of a <b>gf_general_t </b></li><br>

<li> <b>gf_general_is_one() </b> in <b>gf_general.h:</b> This tests whether a <b>gf_general_t </b> is one</li><br>

<li> <b>gf_general_is_two() </b> in <b>gf_general.h:</b> This tests whether a <b>gf_general_t  </b>is two</li><br>

<li> <b>gf_general_is_zero() </b> in <b>gf_general.h:</b> This tests whether a <b>gf_general_t </b> is zero</li><br>

<li> <b>gf_general_multiply() </b> in <b>gf_general.h:</b> This multiplies two <b>gf_general_t's.</b> See the implementation of gf_mult.c

for an example</li><br>
<li> <b>gf_general_s_to_val() </b> in <b>gf_general.h:</b> This converts a string to a <b>gf_general t.</b> See the implementation of
gf_mult.c for an example</li><br>
<li> <b>gf_general_set_one() </b> in <b>gf_general.h:</b> This sets a <b>gf_general_t</b> to one</li><br>
<li> <b>gf_general_set_random()</b> in <b>gf_general.h:</b> This sets a <b>gf_general_t </b> to a random number</li><br>
<li> <b>gf_general_set_two() in </b><b>gf_general.h:</b> This sets a <b>gf_general_t </b> to two</li><br>
<li> <b>gf_general_set_up_single_timing_test() </b> in <b>gf_general.h:</b> Used in <b>gf_time.c</b></li><br>
<li> <b>gf_general_set_zero() in </b><b>gf_general.h:</b> This sets a <b>gf_general_t_to_zero</b></li><br>
<li> <b>gf_general_t_in .</b><b>gf_general.h:</b> This is a general data type for all values of w. See the implementation of gf_mult.c
for examples of using these</li><br>
<li> <b>gf_general_val_to_s()</b> in<b>gf_general.h:</b> This converts a <b>gf_general_t </b> to a string. See the implementation of
<b>gf_mult.c</b> for an example</li><br>

<li> <b>gf_init_easy()</b> in <b>gf_complete.h:</b> This is how you initialize a default <b>gf_t.</b> See 4.2 through 4.5 for examples of
calling <b>gf_init_easy()</b></li><br>
</ul>







<br/>


9 &nbsp &nbsp  <em>  LISTING OF PROCEDURES     </em>   <span id="index_number">40  </span> <br><br><br>

<ul>

<li><b>gf_init hard()</b> in <b>gf_complete.h: </b> This allows you to initialize a <b>gf_t</b> without using the defaults. See 6.4. We
recommend calling create <b>gf_from argv()</b> when you can, instead of <b>gf_ init_hard()</b></li><br>

<li> <b>gf_ mult_type_t </b> in <b>gf_complete.h: </b> the different ways to specify multiplication when using <b>gf_init hard()</b>. See
section 6.4</li><br>

<li> <b>gf_region_type_t</b> in <b>gf_complete.h: </b> the different ways to specify region multiplication when using <b>gf_init_hard()</b>.
See section 6.4</li><br>

<li> <b>gf_region_in</b> <b>gf_complete.h: </b> This is the data type of <b>multiply_region()</b> in a <b>gf_t.</b> See section 4.3 for an example
of how to use <b>multiply_region()</b></li><br>

<li> <b>gf_scratch_size()</b> in <b>gf_complete.h: </b> This is how you calculate how much memory a <b>gf_t</b> needs. See section 6.4.</li><br>

<li> <b>gf_size()</b> in <b>gf_complete.h: </b> Returns the memory consumption of a <b>gf_t.</b> See section 6.5.</li><br>

<li> <b>gf_ val_128_t</b> in <b>gf_complete.h: </b> This is how you store a value where <em>w </em> &#8804 128. It is a pointer to two 64-bit
unsigned integers. See section 4.4</li><br>


<li> <b>gf_val_32_t</b> in <b>gf_ complete.h: </b> This is how you store a value where <em>w </em> &#8804 32. It is equivalent to a 32-bit unsigned
integer. See section 4.2</li><br>

<li> <b>gf_ val_64_t</b> in <b>gf_complete.h: </b> This is how you store a value where <em>w </em> &#8804 64. It is equivalent to a 64-bit unsigned
integer. See section 4.5</li><br>

<li> <b>gf_w16_get_div_alog_table()</b> in <b>gf_ complete.h: </b> This returns a pointer to an inverse logarithm table that can be
used for inlining division when <em>w </em> = 16. See section 7.1</li><br>


<li> <b>gf_w16_get_log_table()</b> in <b>gf_complete.h: </b> This returns a pointer to a logarithm table that can be used for inlining
when <em>w </em> = 16. See section 7.1</li><br>


<li> <b>gf_w16_get_mult_alog_table()</b> in <b>gf_complete.h: </b> This returns a pointer to an inverse logarithm table that can be
used for inlining multiplication when <em>w </em> = 16. See section 7.1</li><br>


<li> <b>gf_ w4 get div table()</b> in <b>gf_complete.h: </b> This returns a pointer to a division table that can be used for inlining
when <em>w </em> = 4. See section 7.1</li><br>


<li> <b>gf_w4_get_mult_table()</b> in <b>gf_complete.h: </b> This returns a pointer to a multiplication table that can be used for
inlining when <em>w </em> = 4. See section 7.1</li><br>

<li> <b>gf_w8_get_div_table()</b> in <b>gf_complete.h: </b> This returns a pointer to a division table that can be used for inlining
when <em>w </em> = 8. See section 7.1</li><br>

<li> <b>gf_w8_get_mult_table()</b> in <b>gf_complete.h: </b> This returns a pointer to a multiplication table that can be used for
inlining when <em>w </em> = 8. See section 7.1</li><br>

</ul>









<br/>
10 &nbsp &nbsp  <em>TROUBLESHOOTING </em>   <span id="index_number">41  </span> <br><br><br>

<ul>
<li><b> SSE support.</b> Leveraging SSE instructions requires processor support as well as compiler support. For example,
the Mac OS 10.8.4 (and possibly earlier versions) default compile environment fails to properly compile
PCLMUL instructions. This issue can be fixed by installing an alternative compiler; see Section 3 for details</li><br>

<li> <b>Initialization segfaults.</b> You have to already have allocated your <b>gf_t</b> before you pass a pointer to it in
<b>bgf_init_easy()</b>, <b>create_gf_ from_argv()</b>, or <b>bgf_ini_hard()</b></li><br>


<li> <b>GF-Complete is slower than it should be.</b> Perhaps your machine has SSE, but you haven't specified the SSE
compilation flags. See section 3 for how to compile using the proper flags</li><br>


<li> <b>Bad alignment.</b> If you get alignment errors, see Section 5</li><br>

<li> <b>Mutually exclusive region types.</b> Some combinations of region types are invalid. All valid and implemented
combinations are printed by <b>bgf_methods.c </b></li><br>

<li><b>Incompatible division types.</b> Some choices of multiplication type constrain choice of divide type. For example,
<b>"COMPOSITE"</b> methods only allow the default division type, which divides by finding inverses (i.e.,
neither <b>"EUCLID"</b> nor <b>"MATRIX"</b> are allowed). For each multiplication method printed by <b>gf_methods.c,</b> the
corresponding valid division types are also printed</li><br>


<li><b> Arbitrary "GROUP" arguments.</b> The legal arguments to <b>"GROUP"</b> are specified in section 7.5</li><br>

<li> <b> Arbitrary "SPLIt" arguments.</b> The legal arguments to <b>"SPLIt"</b> are specified in section 7.4</li><br>

<li> <b>Threading problems.</b> For threading questions, see Section 8</li><br>

<li> <b>No default polynomial.</b> If you change the polynomial in a base field using <b>"COMPOSITE,"</b> then unless it is
a special case for which GF-Complete finds a default polynomial, you'll need to specify the polynomial of the
composite field too. See 7.8.2 for the fields where GF-Complete will support default polynomials</li><br>
<li> Encoding/decoding with different fields. Certain fields are not compatible. Please see section 7.2 for an
explanation</li><br>


<li> <b>"ALTMAP" is confusing.</b> We agree. Please see section 7.9 for more explanation.

<li> <b>I used "ALTMAP" and it doesn't appear to be functioning correctly.</b> With 7.9, the size of the region and
its alignment both matter in terms of how <b>"ALTMAP"</b> performs <b>multiply_region()</b>. Please see section 7.9 for
detailed explanation</li><br>

<li><b>Where are the erasure codes?.</b> This library only implements Galois Field arithmetic, which is an underlying
component for erasure coding. Jerasure will eventually be ported to this library, so that you can have fast erasure
coding</li><br>
</ul>
<h2>11 &nbsp &nbsp Timings </h2>

We don't want to get too detailed with timing, because it is quite machine specific. However, here are the timings on
an Intel Core i7-3770 CPU running at 3.40 GHz, with 4 &#215 256 KB L2 caches and an 8MB L3 cache. All timings are
obtained with <b>gf_time</b> or <b>gf_inline_time,</b> in user mode with the machine dedicated solely to running these jobs.









<br/>
10 &nbsp &nbsp  <em>TROUBLESHOOTING </em>   <span id="index_number">41  </span> <br><br><br>

<div class="image-cell_5"> </div>
<center>Figure 4: Speed of doing single multiplications for w &#8712 {4, 8, 16}. </center>
<h2>11.1 &nbsp Multiply() </h2>

The performance of <b>multiply()</b> is displayed in Figures 4 for w &#8712 {4, 8, 16} and 5 for w &#8712 {32, 64, 128}. These
numbers were obtained by calling <b>gf_time</b> with the size and iterations both set to 10240. We plot the speed in megaops
per second.

<p>As would be anticipated, the inlined operations (see section 7.1) outperform the others. Additionally, in all
cases with the exception of <em>w</em> = 32, the defaults are the fastest performing implementations. With w = 32,
"CARRY FREE" is the fastest with an alternate polynomial (see section 7.7). Because we require the defaults to
use a "standard" polynomial, we cannot use this implementation as the default. </p>

<h2>11.2 &nbsp Divide() </h2>

For the  <b>"TABLE"</b> and <b>"LOG"</b> implementations, the performance of division is the same as multiplication. This means
that for w &#8712 {4, 8, 16}, it is very fast indeed. For the other implementations, division is implemented with Euclid's
method, and is several factors slower than multiplication.
In Figure 6, we plot the speed of a few implementations of the larger word sizes. Compared to the <b>"TABLE"</b> and
<b>"LOG"</b> implemenations for the smaller word sizes, where the speeds are in the hundreds of mega-ops per second,
these are very slow. Of note is the <b>"COMPOSITE"</b> implementation for <em>w</em> = 32, which is much faster than the others








<br/>
10 &nbsp &nbsp  <em>TROUBLESHOOTING </em>   <span id="index_number">43  </span> <br><br><br>

<div class="image-cell_6"> </div>

<center>Figure 5: Speed of doing single multiplications for w &#8712 {32, 64, 128}. </center><br>

because it uses a special application of Euclid's method, which relies on division in <em>GF(2<sup>16</sup>),</em> which is very fast.<br><br>

<h3>11.3 &nbsp Multiply_Region() </h2>

Tables 3 through 8 show the performance of the various region operations. It should be noted that for <em>GF(2<sup>16 </sup>) </em>
through <em>GF(2<sup>128</sup>),</em> the default is not the fastest implementation of <b>multiply_region().</b> The reasons for this are outlined
in section 6
<p>
For these tables, we performed 1GB worth of <b>multiply_region()</b> calls for all regions of size 2i bytes for 10 &#8804 i &#8804
30. In the table, we plot the fastest speed obtained.</p>
<p>We note that the performance of <b>"CAUCHY"</b> can be improved with techniques from [LSXP13] and [PSR12].</p>









<br/>
<em>REFERENCES </em>   <span id="index_number">44  </span> <br><br><br>

<div class="image-cell_7"> </div>

<center>Figure 6: Speed of doing single divisions for w &#8712 {32, 64, 128}. </center><br>

<center>
<div id="data2">
<table cellpadding="2" cellspacing="0" style="text-align:center;font-size:19px">

<tr><th>Method</td> <th>Speed (MB/s)</td> </tr>

<tr><td>-m TABLE (Default) -</td> <td>11879.909</td> </tr>
<tr><td>-m TABLE -r CAUCHY -</td> <td>9079.712</td> </tr>
<tr><td>-m BYTWO b -</td> <td>5242.400</td> </tr>
<tr><td>-m BYTWO p -</td> <td>4078.431</td> </tr>
<tr><td>-m BYTWO b -r NOSSE -</td> <td>3799.699</td> </tr>
<tr><td>-m TABLE -r QUAD -</td> <td>3014.315</td> </tr>

<tr><td>-m TABLE -r DOUBLE -</td> <td>2253.627</td> </tr>
<tr><td>-m TABLE -r NOSSE -</td> <td>2021.237</td> </tr>
<tr><td>-m TABLE -r NOSSE -</td> <td>1061.497</td> </tr>
<tr><td>-m LOG -</td> <td>503.310</td> </tr>


<tr><td>m SHIFT -</td> <td>157.749</td> </tr>
<tr><td>-m CARRY FREE -</td> <td>86.202</td> </tr>
</div>
</table> <br><br>
</div> </center>
<center>Table 3: Speed of various calls to <b>multiply_region()</b> for <em>w</em> = 4. </center>

<h3>References </h3>

[Anv09] H. P. Anvin. The mathematics of RAID-6.<a href=""> http://kernel.org/pub/linux/kernel/people/hpa/
raid6.pdf,</a> 2009.<br><br>

[BKK<sup>+</sup>95] J. Blomer, M. Kalfane, M. Karpinski, R. Karp, M. Luby, and D. Zuckerman. An XOR-based erasureresilient
coding scheme. Technical Report TR-95-048, International Computer Science Institute, August
1995. <br><br>

[GMS08] K. Greenan, E. Miller, and T. J. Schwartz. Optimizing Galois Field arithmetic for diverse processor
architectures and applications. In MASCOTS 2008: <em>16th IEEE Symposium on Modeling, Analysis and
Simulation of Computer and Telecommunication Systems,</em> Baltimore, MD, September 2008.<br><br>


[GP97] S. Gao and D. Panario. Tests and constructions of irreducible polynomials over finite fields. In <em> Foundations
of Computational Mathematics,</em> pages 346–361. Springer Verlag, 1997.
















<br/>
<em>REFERENCES </em>   <span id="index_number">45  </span> <br><br><br>


<center>
<div id="data2">
<table cellpadding="2" cellspacing="0" style="text-align:center;font-size:19px">

<tr><th>Method</td> <th>Speed (MB/s)</td> </tr>
<tr><td>-m SPLIT 8 4 (Default)</td> <td>13279.146</td> </tr>
<tr><td>-m COMPOSITE 2 - -r ALTMAP -</td> <td>5516.588</td> </tr>
<tr><td>-m TABLE -r CAUCHY -</td> <td>4968.721</td> </tr>
<tr><td>-m BYTWO b -</td> <td>2656.463</td> </tr>
<tr><td>-m TABLE -r DOUBLE -</td> <td>2561.225</td> </tr>
<tr><td>-m TABLE -</td> <td>1408.577</td> </tr>

<tr><td>-m BYTWO b -r NOSSE -</td> <td>1382.409</td> </tr>
<tr><td>-m BYTWO p -</td> <td>1376.661</td> </tr>
<tr><td>-m LOG ZERO EXT -</td> <td>1175.739</td> </tr>
<tr><td>-m LOG ZERO -</td> <td>1174.694</td> </tr>


<tr><td>-m LOG -</td> <td>997.838</td> </tr>
<tr><td>-m SPLIT 8 4 -r NOSSE -</td> <td>885.897</td> </tr>


<tr><td>-m BYTWO p -r NOSSE -</td> <td>589.520</td> </tr>
<tr><td>-m COMPOSITE 2 - -</td> <td>327.039</td> </tr>


<tr><td>-m SHIFT -</td> <td>106.115</td> </tr>

<tr><td>-m CARRY FREE -</td> <td>104.299</td> </tr>


</div>
</table> <br><br>
</div> </center>
<center>Table 4: Speed of various calls to multiply region() for <em>w</em> = 4. </center><br><br>

[LBOX12] J. Luo, K. D. Bowers, A. Oprea, and L. Xu. Efficient software implementations of large finite fields
<em>GF(2<sup>n</sup>) </em> for secure storage applications.<em> ACM Transactions on Storage, 8(2),</em> February 2012.<br><br>

[LD00] J. Lopez and R. Dahab. High-speed software multiplication in f<sub>2<sup>m</sup></sub>. In <em>Annual International Conference
on Cryptology in India,</em> 2000.<br><br>

[LHy08] H. Li and Q. Huan-yan. Parallelized network coding with SIMD instruction sets. In <em>International Symposium
on Computer Science and Computational Technology,</em> pages 364-369. IEEE, December 2008.<br><br>

[LSXP13] J. Luo, M. Shrestha, L. Xu, and J. S. Plank. Efficient encoding schedules for XOR-based erasure codes.
<em>IEEE Transactions on Computing,</em>May 2013.<br><br>

[Mar94] G. Marsaglia. The mother of all random generators.<a href=""> ftp://ftp.taygeta.com/pub/c/mother.
c,</a> October 1994.<br>

[PGM13a] J. S. Plank, K. M. Greenan, and E. L. Miller. A complete treatment of software implementations of
finite field arithmetic for erasure coding applications. Technical Report UT-CS-13-717, University of
Tennessee, September 2013.<br><br>

[PGM13b] J. S. Plank, K. M. Greenan, and E. L. Miller. Screaming fast Galois Field arithmetic using Intel SIMD
instructions. In FAST-2013: <em>11th Usenix Conference on File and Storage Technologies,</em> San Jose, February
2013.<br><br>

[Pla97] J. S. Plank. A tutorial on Reed-Solomon coding for fault-tolerance in RAID-like systems.<em> Software -
Practice & Experience,</em> 27(9):995-1012, September 1997.












<br/>
<em>REFERENCES </em>   <span id="index_number">46  </span> <br><br><br>


<center>
<div id="data2">
<table cellpadding="2" cellspacing="0" style="text-align:center;font-size:19px">

<tr><th>Method</td> <th>Speed (MB/s)</td> </tr>
<tr><td>-m SPLIT 16 4 -r ALTMAP -</td> <td>10460.834</td> </tr>
<tr><td>-m SPLIT 16 4 -r SSE (Default) - </td> <td>8473.793</td> </tr>
<tr><td>-m COMPOSITE 2 - -r ALTMAP -</td> <td>5215.073</td> </tr>
<tr><td>-m LOG -r CAUCHY -</td> <td>2428.824</td> </tr>
<tr><td>-m TABLE -</td> <td>2319.129</td> </tr>
<tr><td>-m SPLIT 16 8 -</td> <td>2164.111</td> </tr>

<tr><td>-m SPLIT 8 8 -</td> <td>2163.993</td> </tr>
<tr><td>-m SPLIT 16 4 -r NOSSE -</td> <td>1148.810</td> </tr>
<tr><td>-m LOG -</td> <td>1019.896</td> </tr>
<tr><td>-m LOG ZERO -</td> <td>1016.814</td> </tr>
<tr><td>-m BYTWO b -</td> <td>738.879</td> </tr>
<tr><td>-m COMPOSITE 2 - -</td> <td>596.819</td> </tr>
<tr><td>-m BYTWO p -</td> <td>560.972</td> </tr>
<tr><td>-m GROUP 4 4 -</td> <td>450.815</td> </tr>
<tr><td>-m BYTWO b -r NOSSE -</td> <td>332.967</td> </tr>
<tr><td>-m BYTWO p -r NOSSE -</td> <td>249.849</td> </tr>
<tr><td>-m CARRY FREE -</td> <td>111.582</td> </tr>
<tr><td>-m SHIFT -</td> <td>95.813</td> </tr>


</div>
</table> <br><br>
</div> </center>
<center>Table 5: Speed of various calls to multiply region()  for <em>w</em> = 4. </center><br><br>

[PMG<sup>+</sup>13] J. S. Plank, E. L. Miller, K. M. Greenan, B. A. Arnold, J. A. Burnum, A. W. Disney, and A. C. McBride.
GF-Complete: A comprehensive open source library for Galois Field arithmetic. version 1.0. Technical
Report UT-CS-13-716, University of Tennessee, September 2013.<br><br>

[PSR12] J. S. Plank, C. D. Schuman, and B. D. Robison. Heuristics for optimizing matrix-based erasure codes for
fault-tolerant storage systems. In DSN-2012:<em> The International Conference on Dependable Systems and
Networks,</em> Boston, MA, June 2012. IEEE.<br><br>

[Rab89] M. O. Rabin. Efficient dispersal of information for security, load balancing, and fault tolerance. <em>Journal
of the Association for Computing Machinery,</em> 36(2):335-348, April 1989.









<br/>
<em>REFERENCES </em>   <span id="index_number">47  </span> <br><br><br>
<center>
<div id="data2">
<table cellpadding="2" cellspacing="0" style="text-align:center;font-size:19px">
<tr><th>Method</td> <th>Speed (MB/s)</td> </tr>
<tr>
 
<td>

-m SPLIT 32 4 -r SSE -r ALTMAP - <br>
-m SPLIT 32 4 (Default)  <br>
-m COMPOSITE 2 -m SPLIT 16 4 -r ALTMAP - -r ALTMAP - <br>
-m COMPOSITE 2 - -r ALTMAP -  <br>
-m SPLIT 8 8 <br> 
-m SPLIT 32 8 <br> 
-m SPLIT 32 16 <br> 
-m SPLIT 8 8 -r CAUCHY <br> 
-m SPLIT 32 4 -r NOSSE <br> 
-m CARRY FREE -p 0xc5 <br> 
-m COMPOSITE 2 - <br> 
-m BYTWO b <br> 
-m BYTWO p <br> 
-m GROUP 4 8 <br> 
-m GROUP 4 4 <br> 
-m CARRY FREE <br> 
-m BYTWO b -r NOSSE <br> 
-m BYTWO p -r NOSSE <br>
-m SHIFT <br> 

</td>

<td>
7185.440 <br>
5063.966 <br>
 4176.440 <br>
3360.860 <br>
1345.678 <br>
1340.656 <br>
1262.676 <br>
1143.263  <br>
 480.859 <br>
393.185 <br>
332.964 <br>
309.971 <br>
258.623 <br>
242.076 <br>
227.399 <br>
226.785 <br>
143.403 <br>
111.956 <br>
52.295 <br>
</td>


</tr>

</div>
</table> <br><br>
</div> </center>
<center>Table 6: Speed of various calls to multiply region() <em>w</em> = 4. </center><br><br>

<center>
<div id="data2">
<table cellpadding="2" cellspacing="0" style="text-align:center;font-size:19px">
<tr><th>Method</td> <th>Speed (MB/s)</td> </tr>
<tr>
 
<td>
-m SPLIT 64 4 -r ALTMAP - <br>
-m SPLIT 64 4 -r SSE (Default) - <br>
-m COMPOSITE 2 -m SPLIT 32 4 -r ALTMAP - -r ALTMAP - <br>
-m COMPOSITE 2 - -r ALTMAP -  <br>
-m SPLIT 64 16 - <br>
-m SPLIT 64 8 -  <br>
-m CARRY FREE -  <br>
-m SPLIT 64 4 -r NOSSE - <br>
-m GROUP 4 4 -  <br>
-m GROUP 4 8 -  <br>
-m BYTWO b -  <br>
-m BYTWO p -  <br>
-m SPLIT 8 8 - <br>
-m BYTWO p -r NOSSE - <br>
-m COMPOSITE 2 - - <br>
-m BYTWO b -r NOSSE - <br>
-m SHIFT - <br>

</td>

<td>3522.798 <br>
 2647.862 <br>
2461.572 <br>
1860.921 <br>
1066.490 <br>
998.461 <br>
975.290 <br>
545.479 <br>
230.137 <br>
153.947 <br>
144.052 <br>
124.538 <br>
98.892 <br>
77.912 <br>
77.522 <br>
36.391 <br>
25.282 <br>
</td>


</tr>

</div>
</table> <br><br>
</div> </center>
<center>Table 7: Speed of various calls to multiply region() for  <em>w</em> = 4. </center><br><br>













<br/>
<em>REFERENCES </em>   <span id="index_number">48  </span> <br><br><br>

<center>
<div id="data2">
<table cellpadding="2" cellspacing="0" style="text-align:center;font-size:19px">
<tr><th>Method</td> <th>Speed (MB/s)</td> </tr>
<tr>
 
<td>

-m SPLIT 128 4 -r ALTMAP- <br>
-m COMPOSITE 2 -m SPLIT 64 4 -r ALTMAP - -r ALTMAP- <br> 
-m COMPOSITE 2 - -r ALTMAP- <br> 
-m SPLIT 128 8 (Default)- <br>
-m CARRY FREE -<br> 
-m SPLIT 128 4 -<br> 
-m COMPOSITE 2 - <br>
-m GROUP 4 8 -<br> 
-m GROUP 4 4 -<br> 
-m BYTWO p -<br> 
-m BYTWO b -<br> 
-m SHIFT -<br> 
</td>

<td>
1727.683 <br>
1385.693 <br>
1041.456 <br>
872.619 <br>
814.030 <br>
500.133  <br>
289.207 <br>
133.583 <br>
116.187 <br>
25.162 <br>
25.157 <br>
14.183 <br>
</td>


</tr>

</div>
</table> <br><br>
</div> </center>
<center>Table 8: Speed of various calls to multiply region() for <em>w</em> = 4. </center><br><br>