Message boards :
Number crunching :
Linux x64 Cuda Multibeam (x41g)
Message board moderation
Previous · 1 · 2 · 3
Author | Message |
---|---|
Crunch3r Send message Joined: 15 Apr 99 Posts: 1546 Credit: 3,438,823 RAC: 0 |
Oh well, best of luck with your collaborations then. I certainly hope you have less communication difficulties at the projects you believe in helping more. It's funny, I never had a problem with you before, but you made one now. Let's call it a personality clash & move on then. Dude... I don't know why you think you need to start personal attacks, since i've never had any issues with you before this one.(which i haven't considered offensive nor as a personal attack while i posted it). My initial post was just sort of a 'hint' at the one compiling the code to consider building it on a not so recent linux distro... I don't see why you feel personally attacked by that!!!! And yes, i seem to have less communication difficulties with MilkyWay@Home, Collatz,PrimeGrid etc... it's kind of straight forward. I post what's on my mind and what needs to be done to get where those projects what to be.... No need to get personal nor start attacking someone,since we're on the same page and want to get things done...Efficiency is what gets the job done and that's what is asked for. Join BOINC United now! |
jason_gee Send message Joined: 24 Nov 06 Posts: 7489 Credit: 91,093,184 RAC: 0 |
Dude... I don't know why you think you need to start personal attacks, since i've never had any issues with you before this one.(which i haven't considered offensive nor as a personal attack while i posted it). My initial post was just sort of a 'hint' at the one compiling the code to consider building it on a not so recent linux distro... I don't see why you feel personally attacked by that!!!! It's got nothing to do with the technical suggestions about Linux builds, which I know you know what you are talking about. No question at all. Bringing up the GPL after your very public history with it was not a good idea, and certainly pushes my buttons. Let's leave it at 'personality clash' status & move on, so that I can interpret you drifting back here as a sign of positive interest generated by the work, rather than someone trying to threaten people with the GPL. Jason "Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions. |
Crunch3r Send message Joined: 15 Apr 99 Posts: 1546 Credit: 3,438,823 RAC: 0 |
My intention was not to threaten anyone using GPL... and i think you got the whole message totally wrong in the worst way you've could interpreted it. It was never intended as a threat in the first place and if you would have had any doubts about it, a simple PM about my post would have clarified it's intension. Anyway, you've taken it as a personal attack that it was not even close to what it was supposed to be.... Kind of strange, why you feel frightened by someone asking to consider older supporting 'older' linux OS versions as well or at least post a public source code along with the release to fix those issues... Anyway, i've gotten rid of the the SuSE 11.2 OS running on a ntel(R) Atom(TM) CPU 330 ... the ION GPU is now running the latest CUDA app on windows 7 x64... so i don't care about the linux dependency hell any more. Join BOINC United now! |
jason_gee Send message Joined: 24 Nov 06 Posts: 7489 Credit: 91,093,184 RAC: 0 |
Not strange if you think about it, like I said a personality clash. If you didn't mean to threaten that's great! thanks!. In future when you want to know something, simply ask, and don't mention the GPL. Believe me it'll work a lot better. "Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions. |
Crunch3r Send message Joined: 15 Apr 99 Posts: 1546 Credit: 3,438,823 RAC: 0 |
Not strange if you think about it, like I said a personality clash. If you didn't mean to threaten that's great! thanks!. In future when you want to know something, simply ask, and don't mention the GPL. Believe me it'll work a lot better. Yes, i think it was sort of a miscommunication, resulting in a bad response. I'd suggest for future releases to include a link to a public SVN/GIT repository to avoid such misinterpretations in the first place like Milkyway@home does. That'll get rid of misinformation like that in the first place.... Join BOINC United now! |
jason_gee Send message Joined: 24 Nov 06 Posts: 7489 Credit: 91,093,184 RAC: 0 |
Yes, i think it was sort of a miscommunication, resulting in a bad response. Alright, glad that's resolved & that's much better constructive suggestions. For now, what I'll do is change the policy from 'available on request' to an export in modified sources downloads every new release, like x41g source is there now. As it happens, the X-branch is meant to be going into Seti's Public repository anyway as next stock, so it will indeed have public SVN through Berkeley. It isn't possible/practical to host a public SVN at Lunatics at this time, since the hosting arrangements are provided for gratis, and there have been problems with spamming etc. There's still a lot to work out for the changes to open up for Linux development, as described in the first post, and isn't an instant move, but hopefully things will get to the stage registrations can be opened back up wider public testing done there. Jason "Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions. |
Crunch3r Send message Joined: 15 Apr 99 Posts: 1546 Credit: 3,438,823 RAC: 0 |
Yes, i think it was sort of a miscommunication, resulting in a bad response. Thanks that things got finally sorted and we're on the same page. GIT is free for open source and won't cost a dime.MilkyWay is using it for hosting the GPL source code. (Give it a try!) As long as the source code is GPL and is made public using GIT HUB, it's free of charge and the best way to to! Check it out ;) Join BOINC United now! |
jason_gee Send message Joined: 24 Nov 06 Posts: 7489 Credit: 91,093,184 RAC: 0 |
Will do, Thanks! Got a number of other issues to sort out, since Beta is already going to 6.98, but other than that the codebase is nearly at stable condition ready for a new optimisation pass, which will be interesting. Jason "Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions. |
aaronh Send message Joined: 27 Oct 99 Posts: 169 Credit: 1,442,686 RAC: 0 |
My initial post was just sort of a 'hint' at the one compiling the code to consider building it on a not so recent linux distro... I got that point clearly, as you can see with my discussions with ivan. As I stated to him, I checked major distributions for what I thought would be reasonable compatibility (mostly using the charts at distrowatch). Obviously, I was mistaken. SuSe 11.2 reached its end-of-life in May 2011. Should it still be supported? The question I think becomes, at what point, where do you draw the line? Compatiblity with *how old* of a distribution is desired/needed? |
Bernd Noessler Send message Joined: 15 Nov 09 Posts: 99 Credit: 52,635,434 RAC: 0 |
With the possibility to download the sources that is not a real problem. |
ivan Send message Joined: 5 Mar 01 Posts: 783 Credit: 348,560,338 RAC: 223 |
Situation Report: CUDA 4.1 on Scientific Linux CERN 6 It compiled. Any GPU results after 1100 GMT today on http://setiathome.berkeley.edu/show_host_detail.php?hostid=5370638 will have been done with CUDA 4.1. Silly me -- I had to download the Red Hat Enterprise 6 versions of the libraries, as my colleague had downloaded the Ubuntu version, which had conflicts with RHEL6 (I was under the impression that you had to be registered with nVidia to download; if this was once true it isn't any longer!). I may have to revisit the SLC5 issue, downloading the RHEL5.5 version of cudatoolkit. Aaron, I don't know if you're seeing it too, but I get lots of warnings for unused variables, code that doesn't do anything, and pointers converted to smaller integers. [Edit] First complete job validated -- http://setiathome.berkeley.edu/workunit.php?wuid=905355545 -- I don't see any obvious performance change with it. |
jason_gee Send message Joined: 24 Nov 06 Posts: 7489 Credit: 91,093,184 RAC: 0 |
Yep 4.1rc2 moved to publicly accessbile at some recent point. With 8800 GTS you won't see a noticeable difference yet with 4.1rc2. If a similar situation as on Windows, this is mostly because for Pre-Fermi nVidia hard wired in Cuda binaries now into CUFFT and left out PTX code. Attempting to force in Pre-Fermi PTX only results in breakage due to that part missing in CUFFT. I've reported that Fermi's receive benefit (about 10% or so) from the new LLVM based compiler where pre-Fermi don't, In part due to our dependance on CUFFT & receiving some gain from that new compiler for Fermis. Apparently they've managed to reproduce the issue (On Windows , anyway) and assigned it for further investigation. That'll hopefully mean newly compiled CUFFT binaries for 4.1 release. Jason "Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions. |
ivan Send message Joined: 5 Mar 01 Posts: 783 Credit: 348,560,338 RAC: 223 |
OK, thanks for the info, JG. Now that I have a reliable Linux GPU version of MB, I'll start to look at retiring some of the old machines I have scattered around the labs and concentrating on more modern hardware. I've been promised a new Kepler GPU when they come out, and an nVidia-equipped laptop to replace the old Dell that's finally expired after eight years (partly in conjunction with a newly-arrived artist-in-residence who wants to use my hologram-reconstruction code in her projects). When a decent Kepler card is available I'll upgrade server01 and install SLC6 on it, then also put one in the Ubuntu machine quad-server. That should keep things rocking for a while! :-) |
ivan Send message Joined: 5 Mar 01 Posts: 783 Credit: 348,560,338 RAC: 223 |
I may have to revisit the SLC5 issue, downloading the RHEL5.5 version of cudatoolkit. ...which I have done, and this time it all compiled and linked nyet prahblem! So now pion2 is running CUDA 4.1 code as well. It'll be interesting to see if it changes the pecking order between the C1060 and the GTX 460. Until now the 460 seems to have had the edge on shorties (2'10" vs. 2'50", but with rather increased temperatures) while on longer WUs (~15 minutes) the C1060 was about 30" faster. [Edit] Running neck-and-neck, the 460 was 20 secs faster on a 14'30" task. |
aaronh Send message Joined: 27 Oct 99 Posts: 169 Credit: 1,442,686 RAC: 0 |
Aaron, I don't know if you're seeing it too, but I get lots of warnings for unused variables, code that doesn't do anything, and pointers converted to smaller integers. Yes, I've seen those. At the time, I was holding off with an "If it ain't broke, don't fix it" view. Slowly, they've been getting fixed. |
Francis Noel Send message Joined: 30 Aug 05 Posts: 452 Credit: 142,832,523 RAC: 94 |
Hello optimizers. Should I be watching this thread for further annoucements and developements until the "official" build is released ? If this thread was to be considered closed sorry for bumping it. Anyways if it is of interest to you to see how the app behaves 'in the wild' I have this guy running x41g along with AKV8: http://setiathome.berkeley.edu/show_host_detail.php?hostid=6376472 So far everything has validated. CentOS 6.2, GTX460, C2D. No library issues with the app. Using BOINC 6.10.58 because 6.12.34 complains about missing version info in libcurl. mambo |
Dave Mickey Send message Joined: 19 Oct 99 Posts: 178 Credit: 11,122,965 RAC: 0 |
Looking for help here. Trying to run a GTS240 here on Ubu 11.10, server. Yes, I decided to punish my self with commandline practice. This GPU has moved from a Win7 box where it has run a lot seti. I put it in the Ubu box, and got the nvidia drivers, and did that install along with blacklisting other drivers (like nouveau, riva, etc). After installing and updating some things like gcc, and others, got the driver install to run and rebuild the kernel, or its headers, or what ever it does. now lshw ID's it correctly as a GeForce, and driver = nvidia. This seems good. But boinc-client start complains xhost: unable to open display "" xhost error ignored, GPU computing may not be possible indeed, the boinc stdout startup says no usable gpu found. Since I installed server, there is no X running. Is that required for Ubu boinc to recognize a gpu as usable for crunching? If so, what's the least visible way to make boinc happy about the gpu? I've read some about xconfig, but haven't figured it out. I'd like to keep the machine booting to CLI, not GUI, if I can, just for the heck of it. Oh, and the 41g package stuff is in place, and CPU is crunching seti. starting with the xml for gpu and stock cpu. The driver from nvidia calls itself ...x86_64-295.33 what are the magic words here? thx Dave |
Khangollo Send message Joined: 1 Aug 00 Posts: 245 Credit: 36,410,524 RAC: 0 |
|
Dave Mickey Send message Joined: 19 Oct 99 Posts: 178 Credit: 11,122,965 RAC: 0 |
Wow, even tho I said magic words, I fully expected some long arcane procedure. But this was better than abracadabra - the cave of treasure opened right up (or whatever abracadabra did when uttered...sprouted a genie ? IDK..) Magic words, indeed! thx Dave |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.