UncleAl Posted November 3, 2009 Report Posted November 3, 2009 Do folks have a fast box (core duo) sitting idle or a bored cluster with nothing to do? Uncle Al could use a donation of, oh, 20,000 CPU-hrs in Linux. (Knoppix Live 5.3, not 6.x, is an excellent DvD boot for any box. Remember to start typing within 60 seconds at the first splash screen, at the bottom, knoppix lang=us dma [ENTER]) Uncle Al is again crunching a monstrous iterative problem, http://www.mazepath.com/uncleal/glydense.png Theory is -2 slope, 0.4633849 intercept. http://www.mazepath.com/uncleal/gly2dens.png Needs more crunching to 35K radius. PURSUING THE LIMITS OF FAILED SYMMETRY Linux static file, about 5 MB of output, excecutes in RAM, only uses a hard or flash drive to hold the program and results. If you want to compile the C++ source file yourself, that's also OK. My AMD FX-55 is crunching dense out to 20,000 A radius (finish around March). A Canadian core-duo Mac is doing 20K-25K (60 overnights). 25K-35K is open. Execution time increases as (radius)^2. Windoze is 40% slower in the same iron. Got crunch? Here or xenophage'usual connector'gmail(give it a dot)com Quote
alexander Posted November 3, 2009 Report Posted November 3, 2009 i have a bunch of boxes at home i could donate... they are mostly P4HT's though, but i mean i have 4HTs and probably 5 more normap P4s... my quad is still down for the time :( ... i could make my work quad do some work though... Quote
UncleAl Posted November 5, 2009 Author Report Posted November 5, 2009 Pentiums can be slow for matrix operations. We'll start with a timing benchmark to find out. First, the OS. If you have a native Linux boot, fantastic! Skip to bottom. If not, no problem... 1) Burn a DvD for Knoppix Live! 5.3.1 (NOT ver. 6.x - for the visually handicapped), Debian Linux, KNOPPIX - MirrorsKNOPPIX - Live Linux Filesystem On CDDistroWatch.com: Put the fun back into computing. Use Linux, BSD.Knoppix 5.3.1 flash drive install using Windows | USB Pen Drive Linux DvD on an 8 GB or larger flash drive 2) Boot your box with the Knoppix DvD in the drive. Within 60 seconds of the first splash screen appearing, begin to type in knoppix lang=us dma [ENTER] 3) A page of techno-drool will write, then the Knoppix OS (Debian) loads, then a splash screen as tools load. Kill the splash screen when it finishes. 4) Logout is click bottom leftmost icon to pop a menu, logout, choose logout. The screen reverts, more techo-drool writes, the DvD drive opens. Remove DvD and close drive. We could have an idle box or three feel the love of computation. First a timing benchmark, then the crunch. Knoppix default locks out the hard drive. Run my progam and store its output from a 1 GB or larger flash drive. Windoze cannot exact revenge on what it can never see. Quote
alexander Posted November 5, 2009 Report Posted November 5, 2009 i haves 3 boxes with gentoo :D where is your program? Quote
UncleAl Posted November 6, 2009 Author Report Posted November 6, 2009 http://www.mazepath.com/uncleal/unclechi.zip contains chigly - Linux executable static file. All necessary math libraries are contained.testser.bat - two lines. First fast line sets a timestamp. Second slow line will run overnight. Net runtime allows prediction of any radius-sampling interval. Program chigly quickly writes a 0-byte file, does all the calculations, then outputs data all at once to fill the file. Bring up a system prompt, change directory to the goodies, (chmod permissions for execution and write as necessary) then ./testser.bat [ENTER] ./executable (start radius) increment (end radius) flag > output file. Output looks like (your starting radius will be 20001.720 and about 175 lines), 20001.710 4259355482159 0.999999988117854958 20001.740 4259374639421 0.999999988601445702 20001.760 4259387421387 0.999999988247695298 20001.790 4259406581075 0.999999986435139091 20003.410 4260441657621 0.999999995685679421 radius, A atoms CHI Send me or post the net runtime in each box. We'll see what is reasonable, radius range and sampling density vs. total runtime. "Best efforts will not substitute for knowledge," W. Edwards Deming. Quote
alexander Posted November 9, 2009 Report Posted November 9, 2009 methinks someone needs to thread that... its only running on one core on this machine.... you want to process faster, use all available cores :) you realize that it would run faster on my 3.4GHz P4 then on a core 2, this Phenom, Opteron x6 or the newest Xeons... Quote
alexander Posted November 9, 2009 Report Posted November 9, 2009 first thing's first, you can process both files at the same time by backgrounding the tasks TESTser.sh #!/bin/bash time ./chigly 20000.02 1.7 20003.5 1 > test1s.txt & time ./chigly 20000.02 1.7 20300.0 1 > test2s.txt & also this helps you with exact runtimes... at least that uses multiple cores to compute... Quote
alexander Posted November 9, 2009 Report Posted November 9, 2009 another thing that would be nice is to be able to stop those scripts and resume them later on (i.e. so they write intermediate data sets to those files by themselves...) Quote
alexander Posted November 10, 2009 Report Posted November 10, 2009 runtimes are as follows Total:real 555m46.706s First part completed in 9m33.946s results areTest 1 20001.720 4259361870862 0.999999988594252545 20003.420 4260448045281 0.999999994649642359 20003.500 4260499162979 0.999999993126468836 Test 2 20001.720 4259361870862 0.999999988594252545 20003.420 4260448045281 0.999999994649642359 20005.120 4261534363498 0.999999995179691807 20006.820 4262620854446 0.999999995201757996 20008.520 4263707501089 0.999999990024515506 20010.220 4264794439290 0.999999993445269768 20011.920 4265881483851 0.999999990910678439 20013.620 4266968718726 0.999999994762929436 20015.320 4268056121260 0.999999996047898969 20017.020 4269143791304 0.999999987695493508 20018.720 4270231589456 0.999999987554289916 20020.420 4271319505184 0.999999991830882253 20022.120 4272407699069 0.999999996288999695 20023.820 4273496085157 0.999999990639592077 20025.520 4274584606121 0.999999992618870754 20027.220 4275673306435 0.999999988131538419 20028.920 4276762242069 0.999999990513646065 20030.620 4277851341967 0.999999987242677472 20032.320 4278940590424 0.999999993414362768 20034.020 4280030045444 0.999999988962631154 20035.720 4281119722009 0.999999989955129668 20037.420 4282209563189 0.999999989509968786 20039.120 4283299547747 0.999999990108163289 20040.820 4284389747106 0.999999996048109095 20042.520 4285480169216 0.999999989886553838 20044.220 4286570714110 0.999999992840413433 20045.920 4287661479962 0.999999993352570419 20047.620 4288752409959 0.999999990939777028 20049.320 4289843572355 0.999999984129342799 20051.020 4290934827332 0.999999990174983774 20052.720 4292026353985 0.999999992846431543 20054.420 4293118031957 0.999999994137800261 20056.120 4294209926800 0.999999990182230622 20057.820 4295301932334 0.999999989827982889 20059.520 4296394211677 0.999999991288107340 20061.220 4297486608255 0.999999994718789965 20062.920 4298579218516 0.999999989419900789 20064.620 4299672012074 0.999999994918723926 20066.320 4300765000584 0.999999994377438840 20068.020 4301858143412 0.999999992522425562 20069.720 4302951493218 0.999999991474312880 20071.420 4304045040747 0.999999988002993316 20073.120 4305138759734 0.999999990870006052 20074.820 4306232653309 0.999999996667306982 20076.520 4307326751557 0.999999990780238210 20078.220 4308421024488 0.999999991340194848 20079.920 4309515478846 0.999999995009349498 20081.620 4310610112346 0.999999989122606470 20083.320 4311704958446 0.999999994639300035 20085.020 4312799987189 0.999999994275341641 20086.720 4313895165808 0.999999993893411860 20088.420 4314990561778 0.999999991126426091 20090.120 4316086150989 0.999999988761057975 20091.820 4317181912360 0.999999987709944342 20093.520 4318277804587 0.999999992174968032 20095.220 4319373987760 0.999999985872259763 20096.920 4320470282167 0.999999996784338086 20098.620 4321566784555 0.999999993832270764 20100.320 4322663432085 0.999999993945012458 20102.020 4323760363674 0.999999981329827029 20103.720 4324857382303 0.999999993920746463 20105.420 4325954624574 0.999999996185738309 20107.120 4327052067515 0.999999993295895623 20108.820 4328149692884 0.999999988243463135 20110.520 4329247476809 0.999999992246309238 20112.220 4330345421518 0.999999990564056899 20113.920 4331443629615 0.999999989324330600 20115.620 4332541997004 0.999999990681602665 20117.320 4333640504907 0.999999991709455110 20119.020 4334739233720 0.999999990175830177 20120.720 4335838183525 0.999999988716490980 20122.420 4336937286002 0.999999991048408068 20124.120 4338036517470 0.999999987487067931 20125.820 4339136010966 0.999999993475943227 20127.520 4340235695508 0.999999989511507567 20129.220 4341335495552 0.999999995445637963 20130.920 4342435501654 0.999999992180161715 20132.620 4343535756895 0.999999993018001864 20134.320 4344636173603 0.999999985271416020 20136.020 4345736707651 0.999999994344388950 20137.720 4346837495449 0.999999993394085264 20139.420 4347938451059 0.999999991818489476 20141.120 4349039613416 0.999999985995493343 20142.820 4350140876743 0.999999988535410685 20144.520 4351242461096 0.999999989523248612 20146.220 4352344131518 0.999999991679746864 20147.920 4353446034207 0.999999989028021031 20149.620 4354548071760 0.999999990987522559 20151.320 4355650369289 0.999999992451520732 20153.020 4356752805907 0.999999988951668543 20154.720 4357855416378 0.999999997141202479 20156.420 4358958231356 0.999999992330278066 20158.120 4360061249192 0.999999991853812890 20159.820 4361164444000 0.999999993731339928 20161.520 4362267790079 0.999999993301936308 20163.220 4363371365109 0.999999989111628291 20164.920 4364475116785 0.999999991862276717 20166.620 4365579051000 0.999999990048595701 20168.320 4366683146974 0.999999995896386685 20170.020 4367787452148 0.999999994203436830 20171.720 4368891961296 0.999999989407606602 20173.420 4369996600948 0.999999993225798179 20175.120 4371101498601 0.999999993926361905 20176.820 4372206539188 0.999999995614990123 20178.520 4373311771091 0.999999996222156030 20180.220 4374417212244 0.999999993541282513 20181.920 4375522801951 0.999999994843845089 20183.620 4376628611850 0.999999992030259654 20185.320 4377734551500 0.999999992044445491 20187.020 4378840760951 0.999999991866954919 20188.720 4379947081464 0.999999990413534425 20190.420 4381053637745 0.999999993947556217 20192.120 4382160345273 0.999999991834810080 20193.820 4383267300005 0.999999986364032181 20195.520 4384374366981 0.999999994289219150 20197.220 4385481659469 0.999999991027384913 20198.920 4386589127903 0.999999994023599019 20200.620 4387696799490 0.999999989251394149 20202.320 4388804615666 0.999999990617714164 20204.020 4389912652158 0.999999990766145236 20205.720 4391020902064 0.999999985957336270 20207.420 4392129289756 0.999999993653136452 20209.120 4393237859249 0.999999993249723213 20210.820 4394346659546 0.999999994869016931 20212.520 4395455630857 0.999999984440528224 20214.220 4396564757221 0.999999992628782138 20215.920 4397674057803 0.999999991882179892 20217.620 4398783631697 0.999999990182567581 20219.320 4399893346024 0.999999990355702417 20221.020 4401003218600 0.999999988848367920 20222.720 4402113291179 0.999999991230513521 20224.420 4403223606209 0.999999987486440684 20226.120 4404334048305 0.999999987179093601 20227.820 4405444645004 0.999999992198148762 20229.520 4406555515070 0.999999996792673342 20231.220 4407666547800 0.999999991682079619 20232.920 4408777728779 0.999999991282926835 20234.620 4409889084766 0.999999990036615959 20236.320 4411000724504 0.999999988750289820 20238.020 4412112461515 0.999999995655655206 20239.720 4413224407430 0.999999991299199146 20241.420 4414336508609 0.999999989171639188 20243.120 4415448893033 0.999999989816916041 20244.820 4416561376122 0.999999991873625252 20246.520 4417674046161 0.999999993859419806 20248.220 4418786945508 0.999999996433797655 20249.920 4419900053573 0.999999985750925973 20251.620 4421013273750 0.999999995539414920 20253.320 4422126713519 0.999999993178130324 20255.020 4423240347904 0.999999992479451276 20256.720 4424354204803 0.999999985555412015 20258.420 4425468169413 0.999999994528048460 20260.120 4426582373398 0.999999994060182938 20261.820 4427696758932 0.999999992691456991 20263.520 4428811332418 0.999999991218541227 20265.220 4429926069390 0.999999994519562783 20266.920 4431041028846 0.999999991670847747 20268.620 4432156171064 0.999999989891235307 20270.320 4433271456592 0.999999995253883805 20272.020 4434386960885 0.999999986870606067 20273.720 4435502650512 0.999999989443990067 20275.420 4436618539717 0.999999995007389934 20277.120 4437734568533 0.999999991797665039 20278.820 4438850848106 0.999999994728046116 20280.520 4439967284683 0.999999996611792830 20282.220 4441083923149 0.999999986536439640 20283.920 4442200711513 0.999999994096468717 20285.620 4443317725595 0.999999992428347469 20287.320 4444434896412 0.999999994754612986 20289.020 4445552272436 0.999999990874859525 20290.720 4446669827983 0.999999993781437944 20292.420 4447787601717 0.999999992383838327 20294.120 4448905507756 0.999999994097877431 20295.820 4450023649890 0.999999994551920307 20297.520 4451141954126 0.999999995307585510 20299.220 4452260464481 0.999999995271881209 20300.000 4452773711454 0.999999989129201273 Processor: Phenom X4 Quote
UncleAl Posted November 11, 2009 Author Report Posted November 11, 2009 Thank you. Good news, everyone! We've got results for 334,311 computed points over the first 5000 A of radius. The trick is density of sampling to fill in the curve, not radius span of sparse sampling. 5000 A radius contains 66.5 billion atoms. I can't handle 66 billion points. We're going for extreme sampling density at smaller radii. Another 250K unique points threatens to saturate my least squares fitter. That should do it. We're adding eight plies of 0.03 A increments, 20 A to 1000 A with 0.001 A staggered starting points. 0.005 is done. The small radius halves are running in two machines. Cute tactic for using all cores! Why should hardware sit idle? At your leisure, load what is convenient of ./chigly 600.001 0.03 1000.0 1 > 02.zz./chigly 600.002 0.03 1000.0 1 > 04.zz./chigly 600.003 0.03 1000.0 1 > 06.zz./chigly 600.004 0.03 1000.0 1 > 08.zz./chigly 600.006 0.03 1000.0 1 > 10.zz./chigly 600.007 0.03 1000.0 1 > 12.zz./chigly 600.008 0.03 1000.0 1 > 14.zz./chigly 600.009 0.03 1000.0 1 > 16.zz Post what you are crunching. I'll run the rest as hardware frees. ZIP the accumulated output, send to xenophage'usual connector'gmail(give it a dot)com (gmail is intolerant of executables in a ZIP. Data only) Symmetry: Culture and Science (ISSN 0865-4824) - Contentspages 233-247 and 307-316 Corrected galley proofs are submitted. It really is the 2008 issue - they've got a backlog. Get in on paper, then assault physics academics again with fatter data. Eat the elephant one bite at a time. Quote
alexander Posted November 11, 2009 Report Posted November 11, 2009 uh you are forgetting the & that backgrounds that process... Quote
alexander Posted November 11, 2009 Report Posted November 11, 2009 Also can you post the source for that util? i would like to see it optimized for multiple architectures (within 64 bit), makes it run faster, you know... Quote
UncleAl Posted November 11, 2009 Author Report Posted November 11, 2009 "8^> We're done. I had 323K points to start. 300K more are coming in from my box and Canada, 300K more from Down Under. Add your pile less duplicate points overall (radius increments that did not add atoms) and I am at the limits of my least squares fitter and plotting software, even with kludges. Everytime we do this we are certain it is the last crystal lattice. Never deny a code poet his muse!By tomorrow I'll mount http://www.mazepath.com/uncleal/source.zip It will contain the serial and mpi/parallellized versions, readme files, and some example outputs. There are three variable areas that change with crystal lattice. It would be useful to have generic compiled executable and static-executable for which the three data areas could be exogenously entered to create specific executables. Don't give this any kind of priority. You've earned your footnote if we publish the results. IMPORTANT: BigCHI and CHIpir do *not* validate an input crystal lattice. They both assume the most extreme case of calculated parity divergence. Table salt would generate data, albeit meaningless. A crystal lattice must be qualified by passing Petitjean's QCM at angstrom radius incremenets from about 2 A radius to about 1200 atoms contained - a hugely different ball of wax, Petitjean / softwares / freewaresqcm.readme Atom coordinates must be input to QCM. Results are independent of input format, but runtime is not (graph theory subtleties). *.xyz format of 10 points would require the age of the universe. HyperChem *.hin format works fine through 1200+ atoms, but it can be massaged to drop runtime about 15% and avoid a potential exitless loop in the calculation. Complicated. Petitejean and I started playing in late 1999. He sought a molecule with CHI greater than 0.68, the best his Chem Dept. could volunteer. Every two or three days I'd e-mail some clevernesses I contrived, he'd send back the QCM outputs CHI, COR, DSI for each molecule. 18 months later... I hit CHI=0.70, COR=1, DSI=0. Complicated. 30 days after that I hit CHI=1 and a small error on his end needed revision. Then three classess of CHI=1 molecules - and one that should have been but wasn't. Complicated. Then we went from discrete molecules to unlimited crystal lattices. This has bounced through chemists, crystallographers (needed a Profoundly Gifted one), mathematicians, physicists, Severely Gifted programmers, a dozen hardware volunteers... over 10 years. Everybody should have a hobby. Now we've got to get the massively validated experiment off paper into the physics apparatus. It is no longer a question of intelligence or endeavor. It is the politics of grant funding and who gets credit for success if and when. Complicated. Quote
UncleAl Posted November 11, 2009 Author Report Posted November 11, 2009 1) http://www.mazepath.com/uncleal/source.zipis uploaded. 259,818 bytes 2) I can see your ZIP, I can click on it, but I do not have Hypography permission granted to open it or download it. As it is rather large it would be locally polite and operationally convenient to remove it then e-mail it to xenophage'usual connector'gmail(give it a dot)com Quote
C1ay Posted November 11, 2009 Report Posted November 11, 2009 I can see your ZIP, I can click on it, but I do not have Hypography permission granted to open it or download it. As it is rather large it would be locally polite and operationally convenient to remove it then e-mail it to xenophage'usual connector'gmail(give it a dot)com You shouldn't need any permissions to save it to your drive but I have emailed it to you anyhow... Quote
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.