Message boards :
Number crunching :
Lunatics_x41g_linux64_cuda32.7z
Message board moderation
Author | Message |
---|---|
OTS Send message Joined: 6 Jan 08 Posts: 369 Credit: 20,533,537 RAC: 0 |
I finally upgraded my Linux server from 32 bit Slackware to 64 bit and installed an nvidia 750ti video card. In preparing to do battle in making the video card process SETI work, I download the “Lunatics_x41g_linux64_cuda32.7z†application and extracted the contents. I then looked at the “app_info.xml†examples and noted the following two lines <app_name>setiathome_enhanced</app_name> <version_num>611</version_num> Based on those lines it appears that is for the older MB work units and perhaps will not work with the newer “SETI@home V7†work units. If that is correct, is there a newer version somewhere? I poked around and Googled pretty extensively and could not find anything newer. |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14653 Credit: 200,643,578 RAC: 874 |
Where did you download it from? You're correct, it's an outdated version (by about 2 years), and should be deprecated. It should not be used for the new setiathome_v7 workunits (unless a developer can correct me). I'll leave it to the Linux gurus to point you in the right direction for a replacement (if available). |
OTS Send message Joined: 6 Jan 08 Posts: 369 Credit: 20,533,537 RAC: 0 |
Where did you download it from? http://lunatics.kwsn.net/index.php?module=Downloads;catd=1 The first listing under Multibeam for Linux. http://lunatics.kwsn.net/index.php?module=Downloads;sa=dlview;id=311 It was the only nvidia GPU app I could find for MB or for AP. |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14653 Credit: 200,643,578 RAC: 874 |
OK, thanks. I can tidy that up once we have a definitive answer from Urs or Jason. |
jason_gee Send message Joined: 24 Nov 06 Posts: 7489 Credit: 91,093,184 RAC: 0 |
x41g IIRC (it's a bit past 3am ;)) did have some rudimentary v7 support (though I'm not definite whether Aaron Haviland had that enabled, he probably did, but it would have been my initial slow implementation. However, there is an issue using Cuda 3.2 applications with Maxwell (specifics unknown at this time, but likely translates to the 750ti on Linux too) Best customised Linux build at the moment would be Petri33's I beleive, though most Linux users here seem to have rolled their own straight from the x41zc sources. If more general availability of a build becomes an issue, I can have another crack at it in a week or so, as I've been slowly gearing up for cross platform simultaneous releases. "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. |
Josef W. Segur Send message Joined: 30 Oct 99 Posts: 4504 Credit: 1,414,761 RAC: 0 |
That x41g build is OK for setiathome_v7 tasks, see for instance Valid SETI@home v7 tasks for computer 5940343. For the download from Crunchers Anonymous, arkayn has updated the included app_info. Joe |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14653 Credit: 200,643,578 RAC: 874 |
That x41g build is OK for setiathome_v7 tasks, see for instance Valid SETI@home v7 tasks for computer 5940343. That host has a GTX 570. Has anyone checked whether Jason's concern over cuda32/Maxwell extends to Linux too? |
OTS Send message Joined: 6 Jan 08 Posts: 369 Credit: 20,533,537 RAC: 0 |
That x41g build is OK for setiathome_v7 tasks, see for instance Valid SETI@home v7 tasks for computer 5940343. Thanks for saving me from a possible heartache. I was about to dive into it. Now I will wait to see what Jason says. |
jason_gee Send message Joined: 24 Nov 06 Posts: 7489 Credit: 91,093,184 RAC: 0 |
That x41g build is OK for setiathome_v7 tasks, see for instance Valid SETI@home v7 tasks for computer 5940343. Relooking over the theory ( Maxwell compatibility guide ) suggests that it should be OK. I still have reservations due to the unexplained issue with Windows 32 bit on 970's & 980's. That may not apply because we're dealing with a 64 bit binary and libraries. If you can do an ldd on the executable before allowing it to run, and it finds suitable Cuda3.2 libraries, then that'd be a good sign. I recall adding origin to the linker configuration after Aaron made that binary, so it might find system ones first if they exist (which wouldn't matter). On the other hand it might want the libraries put in via ld config. "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. |
Josef W. Segur Send message Joined: 30 Oct 99 Posts: 4504 Credit: 1,414,761 RAC: 0 |
That x41g build is OK for setiathome_v7 tasks, see for instance Valid SETI@home v7 tasks for computer 5940343. Agreed that needs checking. I've stepped back to 1200 in the Top Hosts list without finding a Linux/Maxwell combination, will continue. I presume the issue could be checked with an offline run using a suitable test WU, but haven't searched for discussions about it. Joe |
OTS Send message Joined: 6 Jan 08 Posts: 369 Credit: 20,533,537 RAC: 0 |
If you can do an ldd on the executable before allowing it to run, and it finds suitable Cuda3.2 libraries, then that'd be a good sign. I recall adding origin to the linker configuration after Aaron made that binary, so it might find system ones first if they exist (which wouldn't matter). On the other hand it might want the libraries put in via ld config. Does not look good - and I do not know much (as in not a thing) about linking. ldd setiathome_x41g_x86_64-pc-linux-gnu_cuda32 linux-vdso.so.1 (0x00007fff21976000) libpthread.so.0 => /lib64/libpthread.so.0 (0x00007ff9c0c22000) libcudart.so.3 => not found libcufft.so.3 => not found libstdc++.so.6 => /usr/lib64/libstdc++.so.6 (0x00007ff9c091f000) libm.so.6 => /lib64/libm.so.6 (0x00007ff9c0618000) libgcc_s.so.1 => /usr/lib64/libgcc_s.so.1 (0x00007ff9c0401000) libc.so.6 => /lib64/libc.so.6 (0x00007ff9c0041000) /lib64/ld-linux-x86-64.so.2 (0x00007ff9c0e74000) |
jason_gee Send message Joined: 24 Nov 06 Posts: 7489 Credit: 91,093,184 RAC: 0 |
OK, you need a newer build (easiest option). I recall the other option of installing Cuda 3.2 toolkit for those libraries as being rather lengthy/challenging/unwieldy, and I'm not sure Aaron would have put forward compatible PTX in the binary either. The 980 goes into my Linux/Win10TP machine tonight. Not a quick turnaround solution unfortunately, as it's still a pretty bare install needing all the development stuff, so if others have a suitable build on hand that'd be best. Builds from current sources should find their cuda so's when planted in the directory alongside the executable, making life a fair bit easier. "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. |
Josef W. Segur Send message Joined: 30 Oct 99 Posts: 4504 Credit: 1,414,761 RAC: 0 |
FWIW at this point, Computer 7410039 is successfully running the Linux x41g build with a GTX 750. Joe |
jason_gee Send message Joined: 24 Nov 06 Posts: 7489 Credit: 91,093,184 RAC: 0 |
Might have come across a Cuda 5 build in my archive, uploading to my website and will provide a link shortly. "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. |
OTS Send message Joined: 6 Jan 08 Posts: 369 Credit: 20,533,537 RAC: 0 |
OK, you need a newer build (easiest option). I recall the other option of installing Cuda 3.2 toolkit for those libraries as being rather lengthy/challenging/unwieldy, and I'm not sure Aaron would have put forward compatible PTX in the binary either. The 980 goes into my Linux/Win10TP machine tonight. Not a quick turnaround solution unfortunately, as it's still a pretty bare install needing all the development stuff, so if others have a suitable build on hand that'd be best. Builds from current sources should find their cuda so's when planted in the directory alongside the executable, making life a fair bit easier. Perhaps someone will have a build as you suggest. If not, when someone is doing you a favor with a slim to no chance of reciprocating, you can learn to be very patient. Thanks. |
OTS Send message Joined: 6 Jan 08 Posts: 369 Credit: 20,533,537 RAC: 0 |
Might have come across a Cuda 5 build in my archive, uploading to my website and will provide a link shortly. Must have come in while I was typing my last. I will stick around. |
jason_gee Send message Joined: 24 Nov 06 Posts: 7489 Credit: 91,093,184 RAC: 0 |
Bottom of download page at: http://jgopt.org/download.html couple of hasty notes. ldd on that binary should see the libraries properly if in the same directory. I don;t recall whether 7zip format preserves the executable permission. "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. |
OTS Send message Joined: 6 Jan 08 Posts: 369 Credit: 20,533,537 RAC: 0 |
Bottom of download page at: I will give it a try and will change the permissions if need be. If it was a little after 3 AM when you first posted, it must be around 6 AM now. Either you're a real early riser or you have definitely reached the end of your day and won't be around much longer. Thanks for taking the time, I appreciate it. |
jason_gee Send message Joined: 24 Nov 06 Posts: 7489 Credit: 91,093,184 RAC: 0 |
just on 6:30. happens at times, and luckily no work today :D. Well good luck, and if it works it could save me a bit of juggling, so fingers crossed ;) "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. |
OTS Send message Joined: 6 Jan 08 Posts: 369 Credit: 20,533,537 RAC: 0 |
just on 6:30. happens at times, and luckily no work today :D. Well good luck, and if it works it could save me a bit of juggling, so fingers crossed ;) I extracted it and did the ldd check and it looks good. ldd setiathome_x41zc_x86_64-pc-linux-gnu_cuda50 linux-vdso.so.1 (0x00007fff9cbba000) libpthread.so.0 => /lib64/libpthread.so.0 (0x00007f773ea1c000) libcudart.so.5.0 => /usr/sbin/BOINC/./libcudart.so.5.0 (0x00007f773e7c1000) libcufft.so.5.0 => /usr/sbin/BOINC/./libcufft.so.5.0 (0x00007f773c80b000) libstdc++.so.6 => /usr/lib64/libstdc++.so.6 (0x00007f773c509000) libm.so.6 => /lib64/libm.so.6 (0x00007f773c201000) libgcc_s.so.1 => /usr/lib64/libgcc_s.so.1 (0x00007f773bfeb000) libc.so.6 => /lib64/libc.so.6 (0x00007f773bc2b000) /lib64/ld-linux-x86-64.so.2 (0x00007f773ec6e000) libdl.so.2 => /lib64/libdl.so.2 (0x00007f773ba26000) I then altered my app_info.xml to include it and restarted boinc. and it appears it found the 750 NVIDIA GPU 0: GeForce GTX 750 Ti (driver version unknown, CUDA version 7000, compute capability 5.0, 2048MB, 364 GFLOPS peak) but I suspect the "driver version unknown" could be a problem. If it is I will have to do some more tinkering. It is a headless server so I have been trying install the driver via command line only. Thanks again for your time. Note: To those who might also try this in the future. When you extract the files from the .7z, do not do so in your working directory as you might overwrite your working app_info.xml file. |
©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.