Dear Adrian,
Thanks! I found that pmemd.cuda can run without a problem if there are only a few atoms in the restraints (I tried to restrain the COMs of several residues). However the system will blow up if there
are a lot of atoms in each atom group, even if I set rk to 0.0... The dist.RST file is:
# test of Jarzynski for a distance in DNA
&rst iat=-1,-1, r2=6.7783, rk2=0., r2a=7.7783,
igr1=2,5,6,7,9,11,13,17,19,20,21,23,25,27,31,33,34,35,37,39,42,45,46,47,48,49,50,52,54,57,60,61,62,63,64,65,67,69,72,75,76,77,78,79,80,82,84,87,89,93,97,98,99,101,103,106,108,112,116,117,118,120,122,125,128,131,133,134,137,140,141,142,144,146,149,152,155,158,162,163,164,166,168,171,173,177,181,182,183,185,187,190,191,192,195,196,197,199,201,204,207,208,209,210,211,212,214,216,219,222,223,224,227,228,229,231,233,236,239,242,244,245,248,251,252,253,255,257,260,261,262,263,264,265,267,269,271,275,278,282,283,284,286,288,291,293,297,301,302,303,305,307,311,312,313,315,317,320,322,326,330,331,332,334,336,339,342,343,347,348,349,351,353,356,359,360,361,362,363,364,366,368,370,374,378,379,380,382,384,387,390,393,396,400,401,402,404,406,409,412,413,417,418,419,421,423,426,429,432,435,439,440,441,443,445,448,451,452,453,454,455,456,458,460,463,466,467,471,472,473,475,477,480,483,486,489,493,494,495,497,500,501,502,504,506,509,511,512,513,515,517,519,523,526,530,531,532,534,536,539,542,545,547,548,551,554,555,556,558,
560,563,564,566,568,570,572,573,574,576,578,581,583,587,591,592,593,595,597,600,603,606,608,609,612,615,616,617,619,621,624,626,630,634,635,636,638,640,642,646,648,649,650,652,654,657,660,661,662,663,664,665,667,669,673,674,675,677,679,682,685,688,691,695,696,697,699,701,704,706,710,714,715,716,718,720,723,726,729,731,732,735,738,739,740,742,744,747,750,751,752,753,754,755,757,759,762,765,766,767,768,769,770,772,774,777,779,783,787,788,789,791,793,796,799,802,804,805,808,811,812,813,815,817,820,823,824,825,826,827,828,830,832,835,838,841,844,848,849,850,852,854,857,858,859,860,861,862,864,866,869,872,875,877,878,881,884,885,886,888,890,893,895,899,903,904,905,907,909,912,914,918,922,923,924,926,928,932,933,934,936,938,941,944,945,949,950,951,953,955,959,960,961,963,965,967,971,975,976,977,979,981,983,987,990,994,995,996,998,1000,1003,1006,1009,1011,1012,1015,1018,1019,1020,1022,1024,1027,1030,1033,1036,1040,1041,1042,1044,1046,1049,1052,1055,1058,1062,1063,1064,1066,1068,1071,1072,1074,1076,1078,1080,1081,10
82,1084,1087,1088,1089,1091,1093,1096,1099,1100,1104,1105,1106,1108,
igr2=1113,1116,1117,1118,1120,1123,1124,1125,1127,1129,1131,1135,1137,1138,1139,1141,1143,1145,1149,1151,1152,1153,1155,1157,1160,1163,1164,1165,1166,1167,1168,1170,1172,1175,1178,1179,1180,1181,1182,1183,1185,1187,1190,1193,1194,1195,1196,1197,1198,1200,1202,1205,1207,1211,1215,1216,1217,1219,1221,1224,1226,1230,1234,1235,1236,1238,1240,1243,1246,1249,1251,1252,1255,1258,1259,1260,1262,1264,1267,1270,1273,1276,1280,1281,1282,1284,1286,1289,1291,1295,1299,1300,1301,1303,1305,1308,1309,1310,1313,1314,1315,1317,1319,1322,1325,1326,1327,1328,1329,1330,1332,1334,1337,1340,1341,1342,1345,1346,1347,1349,1351,1354,1357,1360,1362,1363,1366,1369,1370,1371,1373,1375,1378,1379,1380,1381,1382,1383,1385,1387,1389,1393,1396,1400,1401,1402,1404,1406,1409,1411,1415,1419,1420,1421,1423,1425,1429,1430,1431,1433,1435,1438,1440,1444,1448,1449,1450,1452,1454,1457,1460,1461,1465,1466,1467,1469,1471,1474,1477,1478,1479,1480,1481,1482,1484,1486,1488,1492,1496,1497,1498,1500,1502,1505,1508,1511,1514,1518,1519,1520,1522,1524,1527,153
0,1531,1535,1536,1537,1539,1541,1544,1547,1550,1553,1557,1558,1559,1561,1563,1566,1569,1570,1571,1572,1573,1574,1576,1578,1581,1584,1585,1589,1590,1591,1593,1595,1598,1601,1604,1607,1611,1612,1613,1615,1618,1619,1620,1622,1624,1627,1629,1630,1631,1633,1635,1637,1641,1644,1648,1649,1650,1652,1654,1657,1660,1663,1665,1666,1669,1672,1673,1674,1676,1678,1681,1682,1684,1686,1688,1690,1691,1692,1694,1696,1699,1701,1705,1709,1710,1711,1713,1715,1718,1721,1724,1726,1727,1730,1733,1734,1735,1737,1739,1742,1744,1748,1752,1753,1754,1756,1758,1760,1764,1766,1767,1768,1770,1772,1775,1778,1779,1780,1781,1782,1783,1785,1787,1791,1792,1793,1795,1797,1800,1803,1806,1809,1813,1814,1815,1817,1819,1822,1824,1828,1832,1833,1834,1836,1838,1841,1844,1847,1849,1850,1853,1856,1857,1858,1860,1862,1865,1868,1869,1870,1871,1872,1873,1875,1877,1880,1883,1884,1885,1886,1887,1888,1890,1892,1895,1897,1901,1905,1906,1907,1909,1911,1914,1917,1920,1922,1923,1926,1929,1930,1931,1933,1935,1938,1941,1942,1943,1944,1945,1946,1948,1950,1953,1956,1
959,1962,1966,1967,1968,1970,1972,1975,1976,1977,1978,1979,1980,1982,1984,1987,1990,1993,1995,1996,1999,2002,2003,2004,2006,2008,2011,2013,2017,2021,2022,2023,2025,2027,2030,2032,2036,2040,2041,2042,2044,2046,2050,2051,2052,2054,2056,2059,2062,2063,2067,2068,2069,2071,2073,2077,2078,2079,2081,2083,2085,2089,2093,2094,2095,2097,2099,2101,2105,2108,2112,2113,2114,2116,2118,2121,2124,2127,2129,2130,2133,2136,2137,2138,2140,2142,2145,2148,2151,2154,2158,2159,2160,2162,2164,2167,2170,2173,2176,2180,2181,2182,2184,2186,2189,2190,2192,2194,2196,2198,2199,2200,2202,2205,2206,2207,2209,2211,2214,2217,2218,2222,2223,2224,2226
/
and the dist_vs_t is:
6.77830 6.77827 0.00000 0.00000
6.77840 6.78889 0.00000 0.00000
6.77850 8.28004 0.00000 0.00000
6.77860 16.78308 0.00000 0.00000
6.77870 29.20869 0.00000 0.00000
6.77880 37.50655 0.00000 0.00000
6.77890 52.50088 0.00000 0.00000
6.77900 94.50975 0.00000 0.00000
6.77910 161.99106 0.00000 0.00000
6.77920 228.18595 0.00000 0.00000
6.77930 306.24484 0.00000 0.00000
6.77940 366.32227 0.00000 0.00000
6.77950 401.46970 0.00000 0.00000
Maybe the limit of the number of atoms in igr is different for pmemd.cuda?
Sincerely yours
Xing
On 08/16/2017 05:55 PM, Adrian Roitberg wrote:
> Unless you are doing some VERY particular calculations, there is almost
> no gain to be had from using more than one GPU.
>
> Replica exchange for instance is one of the situations where you use
> more than one GPU.
>
> For SMD, I cannot imagine a reason to use multiple gpus.
>
> Try pmemd.cuda without MPI and it will work, and it will almost the same
> speed.
>
> BTW: the same applies to regular MD. Using pmemd.cuda.MPI for regular MD
> has very little gain. You are better off running 2 simulations with
> different seeds, one in each gpu.
>
> adrian
>
>
> On 8/16/17 3:16 PM, Xing wrote:
>> Dear all,
>>
>> I'm trying to use pmemd.cuda.MPI to run Steered MD (jar=1) on two GPUs. However the system will blow up within 10 steps while the pmemd.MPI will run without a problem on CPUs. I used pmemd.cuda.MPI
>> for "normal" molecular dynamics and it worked fine. Could this be some bug caused by the GPU precision limit? Is there a workaround if it is a bug?
>>
>>
>> Thanks and best wishes,
>>
>> Xing
>>
>>
>> _______________________________________________
>> AMBER mailing list
>> AMBER.ambermd.org
>> http://lists.ambermd.org/mailman/listinfo/amber
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Aug 17 2017 - 14:00:04 PDT
