Posts by Lee Carre

41) Message boards : Number crunching : Rosetta gets only 60% of cpu power (AMD 64, 64bit linux) (Message 11803) Posted 9 Mar 2006 by Lee Carre Post: ah, sorry, thought you ment boinc when you said The score for rosetta are only based on the variables time and cpu score. So, the problem lies with the compiler. Maybe source code needs te become public to build a faster version? because the benchmarks are from boinc, not the science apps i'm guessing you know the source for boinc is completely seperate from the apps, and you can use a custom compiled boinc client, witht the standard apps (and custom apps with the standard client) the app_info.xml mechanism is to be able to use your own apps and get around the fact that they won't be signed with the correct key so i suppose it depends which you want to make faster? Indeed. I want to make the BOINC benchmark faster. Not the rosetta client. But there is the problem, i already have an 64Bit linux boinc version, wich is much faster then the 32 bit one, but cant get the app_info.xml working. I want to figure out why! well the app_info.xml is for optimised applications, not optimised clients what are you trying to do exactly?
42) Message boards : Number crunching : Running Boinc as a Service (Message 11753) Posted 7 Mar 2006 by Lee Carre Post: the manager should still work as normal just use it as you normally would you don't need to bother with the command line client or looking in the log files for messages if boincview can connect then the manager surely can and yes, as a service it's completely independant of users, it'll run when no-one, anyone and everyone is logged in it makes no difference
43) Message boards : Number crunching : Miscellaneous Work Unit Errors (Message 11179) Posted 22 Feb 2006 by Lee Carre Post: Result 11924681 returned an "Incorrect function. (0x1) - exit code 1 (0x1)" message
44) Message boards : Number crunching : New Rosetta 4.82 (Message 10974) Posted 19 Feb 2006 by Lee Carre Post: Easy answer.....don't run graphics, it steals valuable crunching time! oh, i don't, i only look every so often (like once a week) for aesthetic appeal lol i was just pointing out that there's a bug with them
45) Message boards : Number crunching : New Rosetta 4.82 (Message 10950) Posted 19 Feb 2006 by Lee Carre Post: There's an issue with the graphics, the graphs for energy and RMSD don't seem to update properly, it's like their stuck, hard to describe/explain this affects all my units
46) Message boards : Number crunching : Internet traffic and necessary data (Message 10740) Posted 13 Feb 2006 by Lee Carre Post: A simple code to add into rosetta, to verify wich instruction set each cpu can so u can be sure that app will no crash when executing SSE U will need to adapt it ... eg: If cpu sse capable goto sse_crunch routine Else goto non_sse_crunch routine and may be u will need to have a sse_crunch and a sse2_crunch routines ... *sure, no pc will get any app crash -:) i'm no programmer, so forgive the newbie question i understand the instruction set selection method, but how hard would it be to have different versions of the routines in the same app, would there be a lot of work involved or is it quite simple?
47) Message boards : Number crunching : Internet traffic and necessary data (Message 10739) Posted 13 Feb 2006 by Lee Carre Post: different compression methods will only be adopted if they work across all platforms, if they don't, then they're not appropriate for BOINC use Well, that's why I suggested bzip2 as it's an open-source, plug-in replacement for gzip. Works in all platforms. great stuff, prehaps suggest it on the boinc dev mailing list if it achieves consistently greater compression ratios, it'll help everyone :) Agreed, but as you correct said more processing power, NOT necessarily more HOSTS. Have a look at CPU stats true, but if you compare processing rate (using something like TeraFLOPS) against number of hosts, you'll get a positive correlation (more hosts = more processing) Personally, I'd be happy with offering a beta-SSE-enabled Rosetta executable, as optional install, like many people install optimised BOINC app. now that's an idea, but obviously to get the most benifit for the cost then you might as well deploy an app that will do it automatically, that's the best cost:benifit ratio, but as a half-way thing then yea, a seperate app would probably help, but you'd need quite a lot of people using the optimised version to notice an improvement
48) Message boards : Number crunching : Internet traffic and necessary data (Message 10706) Posted 13 Feb 2006 by Lee Carre Post: 2/ The next step to minimize donor overheads would be using 7zip/bzip2/etc instead of GZip, but the impact in overall traffic reduction will be small compared to #1 imo. different compression methods will only be adopted if they work across all platforms, if they don't, then they're not appropriate for BOINC use 3/ And as long as R is using mostly single precision floats, they could consider compiling with SSE for a 3-4 times speedup. or do as SETI Beta are doing, and re-write the app so that it uses a basic method by default, but uses SSE when it detects that the processor is capable of handling that instruction set, best of both world then, because i'm sure you'd get a lot of people complaining that rosetta no longer runs on their older computers, and besides, rosetta is still seeking more processing power last time i checked, so excluding hosts is a bad thing, especially if the app is just going to error out on a non-compatible host
49) Message boards : Number crunching : Internet traffic and necessary data (Message 10705) Posted 13 Feb 2006 by Lee Carre Post: If it was for app symbol table debug, it needs be sent only when the app changes eg: 4.81 -> 4.82 and then no need to send to the cruncher a new pdb at each new WU Thus, I still believe that it is for graphics usage "native fold" ah, you're right then, i must admit, i don't monitor what's downloaded and network usage, because bandwidth isn't a problem for me, so i don't really pay too much attention to what kind of files are downloaded
50) Message boards : Number crunching : Internet traffic and necessary data (Message 10694) Posted 12 Feb 2006 by Lee Carre Post: I verifyed that from all files that rosetta download for each WU that the largest file is the xxxx.pdb. (2 to 7 mb each pdb) and this large file is not* used by the crunching app I am believing that this file is used only to show the "native" protein fold, on the screen_saver -or- when we click "show graphics" If so, here a suggestion for rosetta options ops preferences to skip downloading this file at users choice Ofcourse who opt to not d/l it will not see the "native" fold on screen. as far as i know .pdb files are so that when/if the accociated app crashes, it produces a meaningful error rather than just an error code and a memory address (which isn't that helpful) yes, strictly pdb files aren't needed, but i'm sure most users will appreciate having an error message rather than a hex code however i may be wrong, and it could be for the screensaver and not a symbol file (but in my experience pdb files are generally symbol files for app crashes)
51) Message boards : Number crunching : Rosetta gets only 60% of cpu power (AMD 64, 64bit linux) (Message 10693) Posted 12 Feb 2006 by Lee Carre Post: Maybe source code needs te become public to build a faster version? the source code is avaliable, see the BOINC web site The source for BOINC is indeed avaliable. But according to this: url: http://boinc.berkeley.edu/anonymous_platform.php When you want to use the app_info.xml file: "This model is possible only with projects that make their application source code available." With applecation, they ment Rosetta. And as far is i know, the source of seti is public, and the source of rosetta is not. ah, sorry, thought you ment boinc when you said The score for rosetta are only based on the variables time and cpu score. So, the problem lies with the compiler. Maybe source code needs te become public to build a faster version? because the benchmarks are from boinc, not the science apps i'm guessing you know the source for boinc is completely seperate from the apps, and you can use a custom compiled boinc client, witht the standard apps (and custom apps with the standard client) the app_info.xml mechanism is to be able to use your own apps and get around the fact that they won't be signed with the correct key so i suppose it depends which you want to make faster?
52) Message boards : Number crunching : Rosetta gets only 60% of cpu power (AMD 64, 64bit linux) (Message 10651) Posted 11 Feb 2006 by Lee Carre Post: okay, i just figured out why are the scores are lower. Its for the very simple reason: The benchmark scores are lower than on equal windows systems. Our 3500+ scores 3300/5300, while on a windows system a system like this wil score 3300/9000. The score for rosetta are only based on the variables time and cpu score. So, the problem lies with the compiler. different benchmarks for the same host using different operating systems is a known problem Maybe source code needs te become public to build a faster version? the source code is avaliable, see the BOINC web site
53) Message boards : Number crunching : Internet traffic and necessary data (Message 10257) Posted 31 Jan 2006 by Lee Carre Post: useing a bunch of them for playing online games gaming doesn't use that much in the way of bandwidth (but i'm guessing newer games use more, due to more in-game data to account for, the main thing with games is response times, how long it takes data to get from one end to the other, that's why you always see "latency" numbers, so show the "speed" of your connection to the host
54) Message boards : Number crunching : Internet traffic and necessary data (Message 10196) Posted 29 Jan 2006 by Lee Carre Post: 40:1 home.... yup
55) Message boards : Number crunching : Internet traffic and necessary data (Message 10193) Posted 29 Jan 2006 by Lee Carre Post:
56) Message boards : Number crunching : Internet traffic and necessary data (Message 10186) Posted 29 Jan 2006 by Lee Carre Post:
57) Message boards : Number crunching : Internet traffic and necessary data (Message 10181) Posted 29 Jan 2006 by Lee Carre Post: Something that might interest stats fans here I live in the Channel Islands (Jersey to be precise, no, old Jersey, between England and France) we only have 2 true ISPs (a 3rd gets it's bandwidth from another) the main one is the telecoms provider (the only one in jersey, it's only 45 square miles) we've been able to get 2 meg service from day 1 (with 512k and 1 meg as well) and on the cheapest service, there's a 20GB/month limit jersey connects to london, and paris anyone have any ideas why jersey, being a small island, dependant on it's connections to london and paris, can offer such high bandwidth capacity?
58) Message boards : Number crunching : Internet traffic and necessary data (Message 10168) Posted 29 Jan 2006 by Lee Carre Post: Also, the CPDN team will send out a kill command (on your next connection) if the model goes way off track. ...they don't want to be wasting thousands of hours of CPU time on models that are worthless! great :) always nice when things are done properly ;)
59) Message boards : Number crunching : Internet traffic and necessary data (Message 10057) Posted 27 Jan 2006 by Lee Carre Post: And then there is the 300 MByte upload at the end... The classic CPDN used to ask you if it was OK to do this, to allow you to defer the upload to a better time. you can do it with boinc as well, just disable network access untill a more suitable time I understand that the sulphurs upload at the end of each of five phases - anyone know any better on this? If I am right, any idea of sizes of each of these megatrickles? yup, the results of each phase are uploaded at the end of the phase, rather than the end of the model run i'm not certain of the size, something's telling me it's 3 megs, but that doesn't seem right If I remember right this change was to offset the impact of the fact that those wu generate even more data that needs to be sent back. from what i've read, it's also so that the scientists get pliminary results back sooner, so they know if a model is stable or not, if it's not, might as well ditch it (set server state to "over" so it's not sent out again) it was also so that if a model failed before it completed, the science team at least have partial results for it's run, rather than having to wait for someone to fully complete the model to have any idea how it did
60) Message boards : Number crunching : Internet traffic and necessary data (Message 9985) Posted 27 Jan 2006 by Lee Carre Post: CPDN is actually not too bad, bandwitdh may look large in the first instance, but since that is pretty much it apart from some small trickles for half a year or so, it's overall negligable ;-) And then there is the 300 MByte upload at the end... well, over a couple of months that's not bad, i easilly download over 300 megs from rosetta in a week, and i don't exactly have a lot of hosts

Previous 20 · Next 20