20081110, 18:57  #485  
Just call me Henry
"David"
Sep 2007
Cambridge (GMT/BST)
2^{2}×1,433 Posts 
Quote:
if you want me to email you the other file i will do it is 6.6mb uncompressed Code:
1072468 1006 1606146 1008 1574578 1016 1436120 1027 1560474 1030 1115812 1032 1720724 1034 1487226 1036 1404010 1043 1053682 1047 1346606 1048 1978384 1051 1077094 1060 1116890 1060 1972326 1065 1612900 1069 1841014 1069 1803906 1072 1815456 1072 1521268 1073 1499650 1074 1712340 1077 1480092 1080 1117682 1089 1347010 1094 1026294 1095 1369448 1095 1766032 1095 1169910 1096 1120634 1100 1663490 1117 1962788 1124 1158502 1125 1416270 1128 1860952 1129 1975986 1131 1867308 1138 1608066 1141 1681612 1143 1322764 1144 1972512 1148 1564044 1155 1093986 1165 1197366 1172 1798426 1177 1812858 1180 1245874 1184 1583694 1184 1160416 1195 1882258 1197 1817282 1199 1438618 1202 1079700 1224 1713112 1226 1302742 1236 1897384 1246 1470584 1257 1075182 1260 1872842 1266 1125422 1274 1798116 1276 1763258 1283 1141028 1294 1858156 1295 1488052 1296 1947040 1301 1199456 1315 1025468 1325 1054948 1335 1812988 1342 1991906 1358 1939480 1361 1924970 1368 1132366 1373 1718832 1382 1434526 1384 1612138 1427 1989740 1441 1743476 1451 1293564 1453 1887070 1470 1855566 1480 1715222 1482 1754804 1492 1336570 1520 1175372 1521 1549344 1539 1813116 1578 1290216 1581 1763412 1581 1099662 1590 1928030 1595 1454214 1603 1930270 1607 1684492 1621 1891178 1635 1983490 1637 1839414 1644 1878400 1649 1778098 1655 1847550 1678 1163050 1679 1837428 1683 1991168 1690 1208134 1703 1904388 1705 1783478 1712 1728576 1725 1355924 1741 1615222 1763 1434478 1836 1902052 1875 1587598 1917 1960126 1930 1411736 1952 1903788 1983 1261892 1997 1965872 2001 1571372 2003 1593850 2082 1303160 2098 1754464 2105 1237252 2124 1665414 2124 1401502 2139 1411056 2159 1617984 2159 1163498 2185 1020308 2193 1095700 2202 1035678 2229 1704660 2244 1009860 2246 1159494 2283 1319712 2304 1389108 2339 1883208 2355 1529734 2386 1603160 2415 1079904 2441 1733744 2448 1125670 2482 1380426 2484 1504306 2487 1812242 2532 1643262 2557 1611864 2602 1759328 2605 1169576 2626 1243344 2650 1893444 2669 1525122 2685 1392152 2708 1931208 2742 1946158 2770 1097834 2799 1101552 2862 1741240 2877 1432568 2892 1364474 2904 1070370 2938 1716420 2995 1954284 3094 1508434 3141 1593210 3208 1605302 3215 1219208 3457 1466048 3628 1993850 3631 1859548 3648 1227664 3679 1245410 3778 1117176 3859 1105592 3893 1046944 3909 1961964 4051 1605386 4069 1639034 4204 1024490 4232 1386014 4404 1919064 4422 1131758 4549 1925294 4887 1135190 4936 1304132 4960 1538474 5050 1937250 5176 1493958 5395 1474060 5725 1974600 6093 1461744 6191 1936564 6242 1222984 6658 1824626 6661 1927162 6765 1982148 6953 1359472 7072 1629142 7198 1692630 7299 1532818 7387 1748198 7992 1844870 11022 1152044 11482 1700990 12354 1878582 12950 1588442 14715 if you find no errors in my base15 files i will use the same scripts 

20081111, 05:33  #486  
May 2007
Kansas; USA
2×19×269 Posts 
Quote:
This is a sufficient list of primes for my use. Everything looks great! There are officially 9 k's remaining for k=1M2M. Can you provide me with an updated test limit? The last that you stated was n=14.4K. Since you have a prime for n=14715, I'll show n=14.7K. I'll update the web pages shortly. BTW, you need to use a little punctuation. lol I can't tell if your 1st line is making a statement followed by asking a question or if it's just one big runon sentence that is making a statement with a couple of words left out. If it's a question, can you ask it again? Gary 

20081111, 06:23  #487  
May 2007
Kansas; USA
2×19×269 Posts 
Quote:
lol, you're right. It is 1/9th of the file! my bad Why can't you put more than one core on it and still find the lowest prime? Do what I do when I want to test a range and have no gaps while testing: Sort the file into multiple files using a 1, 2, 3, 4, 1, 2, 3, 4, etc. sequence. Here's what I mean: File one: k/n pair 1 k/n pair 5 k/n pair 9 etc. File 2: k/n pair 2 k/n pair 6 k/n pair 10 etc. File 3: k/n pair 3 k/n pair 7 k/n pair 11 etc. File 4: k/n pair 4 k/n pair 8 k/n pair 12 etc. This can be done by a cutandpaste into Excel column A, then add a column B with 1, 2, 3, 4, 1, 2, 3, 4, 1, 2, 3, 4, etc. in it, and then sort on column B. You don't need a secondary sort because even though all the 1's are the same, Excel doesn't change the position of rows unless the sorting requires it to. So it will keep the file in its original sequence within each occurrence of '1'. The same for each occurrence of '2', '3', and '4'. That way each of the 4 files is still in proper nvalue sequence. That way, you never have gaps in your testing unless one of the testing cores is significantly faster than another. In effect, it somewhat replicates what an LLRnet server does if you had 4 cores on one. For base 27, you could test n=100K200K that way. It'd be a lot of work for 1 core but on 4 cores running concurrently at the same nrange such as this, it wouldn't be too bad. On a related note: I'm kind of tired of my Sierp base 12 effort crawling along on 1 core at n=196K (going to n=250K). At its current rate, it will take ~5055 CPU days to get it up to 250K. In the next day or 2, I'm thinking of dividing it up on 3 quads with the files split up just like I am showing above. I'll just split it into 12 files using a 1,2,3,4,5,6,7,8,9,10,11,12,1,2,3,etc. sequence. That will knock it out in ~45 days if it doesn't find a prime but if it finds one, it will be the smallest. Then I can get back to other stuff and not have it tieing up a core. I run many of my multicore conjecture efforts this way. I'm currently running Riesel base 256 on 4 cores this way. It's currently at n=65K with no gaps below or above. If a prime is found, I stop all 4 cores and remove the k from each one of them, which are at the same nrange. The problem with dividing it up by nrange is that in the long run, it takes more CPU time because frequently you will have tested ranges much higher than the prime you find and those tests will have taken much longer. The point being: Those cores on higher ranges could have been used to find the smaller prime much more quickly by using the above tact. Gary P.S. BTW, I've actually done this on situations where one machine is significantly faster than another yet managed to keep them testing at the same nrange. The math gets a little tricky when deciding what numbers to put in column B but it can be done. If that is your situation and you want to do this, I could PM you with my method of doing it if you wanted. One thing that is a requirement: After you're done, it's important to merge and resort all of the results files by nvalue/kvalue to make sure you missed no tests. It's easy to miss or duplicate a result when messing around with resorting files like this. Last fiddled with by gd_barnes on 20081111 at 06:40 

20081111, 07:26  #488 
Just call me Henry
"David"
Sep 2007
Cambridge (GMT/BST)
1011001100100_{2} Posts 
my current test limit is n=21.5k
i presume it is the text after the primes that is the problem so i will restate that: I am thinking of doing some base 3 sieving soon. How much faster will it be per 1M ks than base 15 to take them to n=25k? 
20081111, 09:35  #489  
May 2007
Kansas; USA
2·19·269 Posts 
Quote:
I assume no primes since n=14.7K. Ah, OK. Makes sense now...lol...I can't quite answer your question fully because its dependent on a # of factors such as krange chosen, how you choose to test it, etc. I will say this: I think I remember that it generally took me around 1 CPU day to test every 1M krange of base 3 to n=25K. That would be only 500,000 k's. So it'd be 2 CPU days or so to do 1M k's. Based on my testing of k=22M on base 15 to n=5K, I think that took just a little over 1 CPU day. I'd guess another 23 CPU days to get it up to n=25K. Based on that I would estimate that it would take you about 4 times as long to test base 15 as it does to test base 3, both because base 3 is more prime and because its a lower base giving it even more opportunities for small primes. Less k's remaining, less testing time for those k's remaining at the same nrange means a big difference in total testing time. Trying doing a k=1M range like you did for base 15 and see if that's close. Base 3 has its pros and cons: Pro: A lot of primes so few k's remaining. Con: A lot of k's that are powers of 3 times k's that are already remaining. It can be very tricky to weed out the correct k's and remove them. Base 15 is easier in that regard because there are less powers of 15 vs. powers of 3 in any given krange, i.e. 15, 225, 3375, 50625, etc. vs. 3, 9, 27, 81, 243, 729, 2187, 6561, 19384, 59049, etc. I noticed that both you and Karsten had some k's remaining for a while in your base 15 testing that you didn't need to have but a prime was eventually found for all of them and so you ended up matching what I had remaining. It's not a big deal to be effectively doubletesting k's below n=25K. Above that and it's wasting a lot of CPU resources. Edit: I just now noticed this...Were you referring to sieving or primality testing on base 3 vs. base 15? Sieving would be only a little longer for base 15. Because base 15 is a higher base, you'll need to sieve it a little deeper to get to the optimal depth. Also, there will be more k's remaining within the same krange. I'd say base 15 would take, perhaps, 3035% longer than base 3 to sieve the same krange. Gary Last fiddled with by gd_barnes on 20081111 at 09:42 

20081111, 13:46  #490 
I quite division it
"Chris"
Feb 2005
England
2077_{10} Posts 
Ok. I'll look into doing that on OpenOffice Calc.
(Or Office on the Kids' PC.) At the moment I'm just sieving because I've already tested 4 unnecessary candidates. Tested to >111k, sieving at >9T, >300 factors found. It's difficult to know what is the best strategy with sieving/testing because we are just looking for the first prime. Hence, the uneasy feeling in my stomach that I've just stopped testing right before 'the' prime. lol 
20081111, 17:15  #491 
Just call me Henry
"David"
Sep 2007
Cambridge (GMT/BST)
2^{2}·1,433 Posts 
i dont know why i said sieving i meant primality proving
once my base 15 effort is finished to n=25k i will do 500000 ks from base 3 to n=25k i have had a rather large gap between primes almost 1/3 of the range tested 
20081111, 18:02  #492 
Sep 2005
Raleigh, North Carolina
337 Posts 

20081111, 20:57  #493  
Just call me Henry
"David"
Sep 2007
Cambridge (GMT/BST)
2^{2}·1,433 Posts 
Quote:
i only have four cores so i tend to not use more than one occasionally two cores per type of work i have just found another prime 1570340 21918 

20081114, 04:53  #494 
May 2007
Kansas; USA
2×19×269 Posts 

20081114, 04:55  #495 
May 2007
Kansas; USA
2×19×269 Posts 
Sierp base 12 is finally at n=200K...nothing to report; continuing on to n=250K.
Thanks to Max for speedy Phrot! My tests are ~40% faster: 3480 vs. 2090 secs. per test at n=195K!! Gary Last fiddled with by gd_barnes on 20081114 at 04:57 
Thread Tools  
Similar Threads  
Thread  Thread Starter  Forum  Replies  Last Post 
Bases 251500 reservations/statuses/primes  gd_barnes  Conjectures 'R Us  2180  20201022 16:42 
Riesel base 3 reservations/statuses/primes  KEP  Conjectures 'R Us  1049  20201021 20:40 
Bases 5011030 reservations/statuses/primes  KEP  Conjectures 'R Us  3740  20201021 18:24 
Bases 101250 reservations/statuses/primes  gd_barnes  Conjectures 'R Us  851  20201020 10:27 
Bases 33100 reservations/statuses/primes  Siemelink  Conjectures 'R Us  1671  20201010 06:28 