Posts by old

21) Message boards : Number crunching : Credits Should Show Up Soon for New Application (hadcm3i) (Message 28055) Posted 22 Apr 2007 by old_user1 Post: thanks for that information. I think I\'ve found the culprit -- the model id info (BOINC appid) wasn\'t put in the cpdn_trickle program. I\'ve now bumped it up to 20 slots and recompiled cpdn_trickle so that should last us for years (version 5.40 hadcm3i is #6). I should have originally just used a std::vector I guess for the data struct, oh well. and all this 24 hours after my great \"going away party\" last night! Smile
22) Message boards : Number crunching : Credits Should Show Up Soon for New Application (hadcm3i) (Message 27942) Posted 17 Apr 2007 by old_user1 Post: What about those running 5.41 on Linux? It\'s tied to the application (# 6, hadcm3i), rather than the version, so Linux 5.41 and Mac 5.39 would be fine.
23) Questions and Answers : Macintosh : No trickles being credited for new Intel Mac app? (Message 27937) Posted 17 Apr 2007 by old_user1 Post: Hi, I just posted in the Message Boards/Number Crunching about this -- basically I had forgotten to add a line in the database to reflect this new version (hadcm3i, workunits hadcm3inct_*) -- I am manually running the stats programme so the credits should be up shortly; sorry about that!
24) Message boards : Number crunching : Credits Should Show Up Soon for New Application (hadcm3i) (Message 27936) Posted 17 Apr 2007 by old_user1 Post: Hi, unfortunately I had forgotten to put a line in my credits database to show that we have released a new application (version 5.40, hadcm3i). This should now be fixed and the stats program is currently running so in an hour or so the new credits for those running 5.40 (on any platform) should be in place. Sorry for the inconvenience!
25) Message boards : Number crunching : Trojan boinc installation by rogue member (Message 27087) Posted 26 Feb 2007 by old_user1 Post: Are we supposed to \"police the world\" when people run unknown software from the Internet? There isn\'t much we can do other than turn off the account(s) in question. There were 6 other projects involved, do you similarly slam them for not calling Scotland Yard, the CIA, and the FBI?
26) Message boards : Number crunching : Status of plans for shorter wus? (Message 27076) Posted 25 Feb 2007 by old_user1 Post: Here is an update on developments are far as the modelling done via CPDN; in order of which they will happen over the next two years. Some are in the \"proposal\" stage so may not come to anything if the reviewers & funding bodies don\'t fund it, I\'ll put a (P) for those: 1) optimized (CPU & file I/O) HadCM3L model to come out of beta testing soon (1-2 months) 2) 80-year workunits to redo some of the more interesting results of the BBC experiment with different forcings (a 1920-2000 and 2000-2080). These workunits would use (1) and probably happen in 1-3 months. 3) I think we\'ll be doing another experiment or two with the HadAM3 model (9-times higher resolution model from the Seasonal Attribution Experiment) 4) regional model (UK MetOffice PRECIS Programme HadRM3 model, see http://www.precis.org.uk) embedded with the HadCM3L global model (which generates the boundary conditions for the region). We\'ll probably do regions of southern Africa and US Pacific Northwest and work with scientists from those area ((P) -- Microsoft is hopefully going to fund this, I\'m going to Seattle with Richard Jones of the MetOffice for meetings next week). I find this project very exciting (and I hope Microsoft does :-) as it will allow scientists from countries that don\'t have a lot of computer resources to take advantage of setting up experiments with CPDN, having the huge CPU power thanks to the participants, and then being able to use data they would otherwise never have been able to get. 5) high-resolution global modelling, with the latest generation MetOffice model (HadGAM1). this is at 4 times the resolution (i.e. \"half\" the hadcm3 resolution, but not quite as high as HadAM3) The experiment is led by William Ingram of the MetOffice (and Oxford too now) to do an in-depth study of cloud feedbacks within the model (i.e. the positive & negative feedbacks of clouds on the climate etc). The modelling code is quite a bit different, and we are keeping things in 64-bit for the first time, so it could take a gigabyte of RAM to run, as well as a 64-bit PC! 6) HadCM3 but use a higher resolution ocean with hopefully no flux corrections (P) 7) amongst the modelling, a big part of the project work will be getting the results online for scientists (Milo is doing a lot in this area as Tolu & I tend to get bogged down on the BOINC client/modelling side). So you can see (as ever) there are a lot of apps on CPDN!
27) Questions and Answers : Windows : No \"finished file\" error (Message 27075) Posted 25 Feb 2007 by old_user1 Post: the \"no finished file\" message is a trivial warning as others have noticed. On the good side, the new cpdn we\'re beta testing uses a lot less file I/O and I think is a lot less prone to these errors. I was running cpdn and backing up my hard drive to a firewire drive (200GB), doing other stuff, running a virus & spyware scan, and the model was fine (the hard drive light was solid red for two hours).
28) Message boards : Number crunching : Status of plans for shorter wus? (Message 27048) Posted 24 Feb 2007 by old_user1 Post: just some quick points I\'d like to make: a) 64-bit does not necessarily mean \"faster\" b) \"faster\" does not always mean \"better\" (i.e. models can have the physics \"optimized away\" if people don\'t do the boring work of really checking what is going on, whether it\'s climate modelling or something else) --- maybe they won\'t admit it, but every BOINC project leader I talked with (i.e. all of \'em ;-) moans about crazy credit-mad overclockers who are just in it for speed. Does nobody realize that the first thing to go (based on data of thousands of workunits) is the FPU which is why \"redundant computing\" is so big (on climate models, these guys would just crash after a few days anyway)? c) if the 2-3 CPDN developers read the boards as much as some people wanted, we\'d never have time to do the 6 or 7 apps we\'ve done over the past few years. I don\'t think any other project, including pretty much every other one that has far more staff & equipment than us, has done as many apps as we have. And soon we\'ll be doing a 64-bit latest-and-greatest-state-of-the-art HadGAM1 model from the MetOffice (that should be out in a year or so). d) Thank god for moderators who answer 99.99% of the questions and help 99.999% of the users and just email me when I don\'t notice the servers are on fire! ;-) e) climate models are primarily 64-bit, but the number of 64-bit machines out there is still pretty tiny (although as I said we\'re doing a 64-bit model in a year). f) the labor in porting a 1-2 million line climate model is not a linear jump from the other apps out there that are maybe 10-100K lines! g) we\'re already beta-testing shorter workunits (40 yrs, the standard will probably be 80 years) which among other things have: h) I\'ve optimized the code quite a bit (compiler optimizations) and still get sensible results (people are reporting 1 and \"sub-1\" second per timestep) i) I don\'t think any other workunit will ever approach the 160 year one, except perhaps ocean spinups in the future (a side-project for the dedicated cruncher/enthusiast) j) I figured out a sensible way to do temp files so that file/disk I/O is reduced 92% (i.e. the dreaded tmp/cache2 and aswap/oswap files that are written to 5 times a second it seems) k) credits are OK but overrated IMHO --- doing what typically only clusters through supercomputers do on your home PC or laptop is far more interesting to me, not to mention running an entire global climate model (and soon -- an embedded regional model!) OK, now time to get some sleep as the lager is wearing off... ;-)
29) Message boards : Number crunching : Why this forum, if the project team doesn\'t bother to go here? (Message 27027) Posted 23 Feb 2007 by old_user1 Post: I agree, it is only possible because we have such great moderators across all the boards that keep us informed when something big comes up like a server down, credits messed up, etc.
30) Message boards : Number crunching : Why this forum, if the project team doesn\'t bother to go here? (Message 27021) Posted 23 Feb 2007 by old_user1 Post: Hi, I try to get around to all the boards, but as you\'ve seen, we have too many! Tolu & I were against this but the \"powers that be\" wanted separate boards. Plus we started as a non-BOINC project (before BOINC was publicly available), and had the old PHPbb boards to maintain, then this one, then they wanted a separate one for the BBC experiment. So I\'m afraid if I miss something I have to rely on one of the moderators to PM me (or Tolu or Milo) for our immediate attention.
31) Message boards : Number crunching : No credit for 2 days (Message 26673) Posted 9 Feb 2007 by old_user1 Post: it looks like the stats process has toppled over for the past two nights. I ran it manually today and edited the script so that hopefully things will go through OK over the weekend.
32) Message boards : Number crunching : BOINC Version 5.8.8 Released (Message 26510) Posted 2 Feb 2007 by old_user1 Post: I have updated the download link here for the new BOINC version for Linux & Windows. Note that we do not have Mac yet but we hope to have a Mac Intel OS X version up within a month (contact tolu at atm dot ox dot ac dot uk to join the Mac Intel beta test). Version 5.8.8 should be safe to upgrade to if you are still using the original BBC/CCE/BOINC software from last February (I believe it renames the directory to \"BOINC\" so that this and future upgrades are correct). Some of the new features (from the BOINC website): * The BOINC Manager offers two different interfaces: \'Simple\' (small, graphical, skinnable) and \'Advanced\' (maximum information). The Simple GUI allows local editing of preferences. You can switch back and forth between Simple and Advanced. * CPU throttling: you can reduce CPU heat by limiting the fraction of time that BOINC computes. * Improved scheduling policies avoid missing job deadlines even on a slow computer with lots of projects. * You can limit the amount of memory used by BOINC; this lets you compute all the time without losing performance. * Mac OS X version runs applications in an unpriviledged account, increasing security. * Snooze button (in the system tray icon) lets you stop computing for one hour. We recommend that all BOINC users upgrade to 5.8.
33) Questions and Answers : Unix/Linux : New BOINC Version 5.8.8 Released (Message 26509) Posted 2 Feb 2007 by old_user1 Post: I have updated the download link here for the new BOINC version for Linux & Windows. Note that we do not have Mac yet but we hope to have a Mac Intel OS X version up within a month (contact tolu at atm dot ox dot ac dot uk to join the Mac Intel beta test). Version 5.8.8 should be safe to upgrade to if you are still using the original BBC/CCE/BOINC software from last February (I believe it renames the directory to \"BOINC\" so that this and future upgrades are correct). Some of the new features (from the BOINC website): * The BOINC Manager offers two different interfaces: \'Simple\' (small, graphical, skinnable) and \'Advanced\' (maximum information). The Simple GUI allows local editing of preferences. You can switch back and forth between Simple and Advanced. * CPU throttling: you can reduce CPU heat by limiting the fraction of time that BOINC computes. * Improved scheduling policies avoid missing job deadlines even on a slow computer with lots of projects. * You can limit the amount of memory used by BOINC; this lets you compute all the time without losing performance. * Mac OS X version runs applications in an unpriviledged account, increasing security. * Snooze button (in the system tray icon) lets you stop computing for one hour. We recommend that all BOINC users upgrade to 5.8.
34) Questions and Answers : Windows : BOINC Version 5.8.8 Released (Message 26508) Posted 2 Feb 2007 by old_user1 Post: I have updated the download link here for the new BOINC version for Linux & Windows. Note that we do not have Mac yet but we hope to have a Mac Intel OS X version up within a month (contact tolu at atm dot ox dot ac dot uk to join the Mac Intel beta test). Version 5.8.8 should be safe to upgrade to if you are still using the original BBC/CCE/BOINC software from last February (I believe it renames the directory to \"BOINC\" so that this and future upgrades are correct). Some of the new features (from the BOINC website): * The BOINC Manager offers two different interfaces: \'Simple\' (small, graphical, skinnable) and \'Advanced\' (maximum information). The Simple GUI allows local editing of preferences. You can switch back and forth between Simple and Advanced. * CPU throttling: you can reduce CPU heat by limiting the fraction of time that BOINC computes. * Improved scheduling policies avoid missing job deadlines even on a slow computer with lots of projects. * You can limit the amount of memory used by BOINC; this lets you compute all the time without losing performance. * Mac OS X version runs applications in an unpriviledged account, increasing security. * Snooze button (in the system tray icon) lets you stop computing for one hour. We recommend that all BOINC users upgrade to 5.8.
35) Questions and Answers : Macintosh : What it all a WASTE of TIME? (Message 26507) Posted 2 Feb 2007 by old_user1 Post: Hi, I\'ve been upgrading the server software, please try \"Update\" again
36) Message boards : Number crunching : First phase of migration of BBC/CCE to cpdn and standard version of boinc (Message 26400) Posted 27 Jan 2007 by old_user1 Post: Sorry megac, it was I who deleted your previous post as it was a bit nasty for the forum. Anyway, the BBC gave us some much needed publicity/participants and Ã‚Â£10K for a server for the results. So it wasn\'t as if we got millions of the licence fee etc (although we have been taxpayer funded via NERC & DTI in the past). In fact, Microsoft has helped keep us afloat with a fellowship to keep me on the project for two years (well another 1.5 years by now). So remember that the next time you hear someone cursing about the \"blue screen of death\" or \"evil\" Bill Gates! :-) The \"Meltdown\" documentary (the one that \"launched\" the BBC experiment) had quite a bit (well a few minutes) on the distributed computing nature, CPDN, SETI@home, but the recent \"results programme\" quickly \"refreshed\" this. I imagine they figured the general public would be bored by a \"computer-oriented\" segment, as exciting as it would be for us working & participating on CPDN and other projects. We should probably have also stressed that it wasn\'t really something for \"general consumption\", i.e. if we were as easy to run as SETI or Einstein or Folding@home that would be possible, but these climate models need to be run for a long time and on a pretty modern PC. The next version of the hadcm3 model will be easier on PCs I think (90% less disk i/o, a little more RAM, i.e. 100MB free required, and faster as it will require a minimum of a Pentium IV for SSE2 optimizations I turned on). Future models will be a bit tough -- the regional project, if it goes through, will require running the hadcm3 as above, as well as about 300MB for the regional model will run separately and be \"coupled\" to hadcm3 global model. Then the HadGAM model will be done and be a 64-bit model that needs probably 1GB of RAM minimum! So the next hadcm3 will be the \"minimum spec\", and we have lots of ideas for good experiments with it (more IPCC scenarios, airplane exhausts experiments, etc). Luckily with \"Moore\'s Law\" PCs will continue to grow and the average PC will be fine for running CPDN, whereas we\'ve always been far ahead of what the average PC can run (hence our lower numbers for users staying on, finishing runs, etc).
37) Message boards : Number crunching : First phase of migration of BBC/CCE to cpdn and standard version of boinc (Message 26314) Posted 23 Jan 2007 by old_user1 Post: the current xml stats on this server are corrected, so when the various external stats site pick them up, things should be back to normal (by \"normal\" I mean that classic & spinup credit will be added in, and BBC & Seasonal credit will not be in but of course tracked by their particular sites).
38) Questions and Answers : Windows : Advertising crawl in Graphics (Message 26308) Posted 23 Jan 2007 by old_user1 Post: yesterday I put in another message, it gets updated upon a trickle, so it could take a few days to update
39) Message boards : Number crunching : First phase of migration of BBC/CCE to cpdn and standard version of boinc (Message 26254) Posted 21 Jan 2007 by old_user1 Post: the stats should be back to normal now
40) Message boards : Number crunching : First phase of migration of BBC/CCE to cpdn and standard version of boinc (Message 26249) Posted 21 Jan 2007 by old_user1 Post: whoops, looks like I went to far the other way! It certainly doesn\'t make sense to weed out the classic & spinup #\'s. It was just an artifact of the new way I\'m tracking these \"extraneous\" credits. So spinups & classic CPDN will be counted back in here, and also the current & future beta tests (whenever I figure out an appropriate beta credit scheme). This makes sense because no other stats sites will have the spinup & classic #\'s, so they should go here. The BBC & Seasonal credits will show up for \"display purposes only\" here, and the stats sites of course will carry this site & the BBC & Seasonal projects for intercomparisons as well.

Previous 20 · Next 20

Posts by old_user1