Message boards :
Number crunching :
Boinc 6.6.2 just released (CUDA)
Message board moderation
Author | Message |
---|---|
ML1 Send message Joined: 25 Nov 01 Posts: 21019 Credit: 7,508,002 RAC: 20 |
Folks, For all you people on the cutting edge and especially cutting CUDA... Boinc 6.2.2 has just hit the downloads page, and for all of Windows, Mac, and Linux. Happy fast crunchin', Martin OK: Usual health warning for the blind and unwary: Development version, make your own backup first, report back for any problems for debug. See new freedom: Mageia Linux Take a look for yourself: Linux Format The Future is what We all make IT (GPLv3) |
Fred W Send message Joined: 13 Jun 99 Posts: 2524 Credit: 11,954,210 RAC: 0 |
Folks, Did you mean to say 6.6.2?? F. [edit]If so, the title of the thread needs changing too :)[/edit] [edit2]And thanks for the heads-up[/edit2] |
Bukken Send message Joined: 3 Apr 99 Posts: 50 Credit: 3,007,776 RAC: 0 |
Needs fixing ! No text in tasks or messages pages in boinc manager. Reboot ect. doesn´t help. |
ML1 Send message Joined: 25 Nov 01 Posts: 21019 Credit: 7,508,002 RAC: 20 |
Folks, Ooooops! Indeed so... Boinc 6.6.2 I'll correct the title also. (That'll teach me to try downloading and typing at the same time!) Cheers, Martin See new freedom: Mageia Linux Take a look for yourself: Linux Format The Future is what We all make IT (GPLv3) |
ML1 Send message Joined: 25 Nov 01 Posts: 21019 Credit: 7,508,002 RAC: 20 |
Needs fixing ! Undoubtedly, as noted in the messages window: "[---] Starting BOINC client version 6.6.2 for x86_64-pc-linux-gnu [---] This a development version of BOINC and may not function properly" No text in tasks or messages pages in boinc manager. Are you attached to any projects? Have you installed successfully? What system are you running? Meanwhile, all looks good on this system for Boinc 6.6.2. Good luck, Martin See new freedom: Mageia Linux Take a look for yourself: Linux Format The Future is what We all make IT (GPLv3) |
Ivailo Bonev Send message Joined: 26 Jun 00 Posts: 247 Credit: 35,864,461 RAC: 2 |
Yep, I can see much more improvements in new version than 6.4.5! Thanks for fixing bug with columns sorting of my native language, it was irritating for me. 6.6.2 now runing on all PCs from 2 days without a problem, but I don't have CUDA ;) |
ML1 Send message Joined: 25 Nov 01 Posts: 21019 Credit: 7,508,002 RAC: 20 |
OK... So Boinc 6.4.5 didn't recognise my nVidia GPU and neither does Boinc 6.6.2 :-( [---] Starting BOINC client version 6.6.2 for x86_64-pc-linux-gnu [---] This a development version of BOINC and may not function properly [---] log flags: task, file_xfer, sched_ops [---] Libraries: libcurl/7.18.0 OpenSSL/0.9.8g zlib/1.2.3 c-ares/1.5.1 [...] [---] OS: Linux: 2.6.27.7-server-1mnb [---] Memory: 3.87 GB physical, 15.99 GB virtual [---] Disk: 64.00 GB total, 20.03 GB free [---] Local time is UTC +0 hours [---] No CUDA devices found [---] No coprocessors So what is it that Boinc looks/checks for? My Linux system reports: # more /proc/driver/nvidia/version NVRM version: NVIDIA UNIX x86_64 Kernel Module 180.22 Tue Jan 6 09:15:58 PST 2009 GCC version: gcc version 4.3.2 (GCC) # # more /proc/driver/nvidia/cards/0 Model: GeForce 8600 GT IRQ: 18 Video BIOS: 60.84.5e.00.00 Card Type: PCI-E DMA Size: 40 bits DMA Mask: 0xffffffffff Bus Location: 01.00.0 # #ldd boinc linux-vdso.so.1 => (0x00007fffdb1fe000) libdl.so.2 => /lib64/libdl.so.2 (0x00007f7ed2cdf000) libnsl.so.1 => /lib64/libnsl.so.1 (0x00007f7ed2ac7000) libz.so.1 => /lib64/libz.so.1 (0x00007f7ed28b2000) libstdc++.so.6 => /usr/lib64/libstdc++.so.6 (0x00007f7ed25a9000) libpthread.so.0 => /lib64/libpthread.so.0 (0x00007f7ed238e000) libm.so.6 => /lib64/libm.so.6 (0x00007f7ed210a000) libc.so.6 => /lib64/libc.so.6 (0x00007f7ed1da9000) libgcc_s.so.1 => /lib64/libgcc_s.so.1 (0x00007f7ed1b93000) /lib64/ld-linux-x86-64.so.2 (0x00007f7ed2ee3000) The GeForce 8600 GT has 256MBytes of video RAM and is listed as CUDA enabled: NVIDIA CUDA-enabled products Soooo... What's missing? Cheers, Martin See new freedom: Mageia Linux Take a look for yourself: Linux Format The Future is what We all make IT (GPLv3) |
john deneer Send message Joined: 16 Nov 06 Posts: 331 Credit: 20,996,606 RAC: 0 |
Needs fixing ! Are you by any chance using Raistmer's package, running cuda and an optimzed AstroPulse together? I had the same symptoms which you describe (and I use Raistmer's package) after upgrading from 6.6.0. After a restart BoincManager spontaneously 'died' and disappeared from Windows task manager. I had to remove Boinc completely (using control panel to remove software). After reinstalling Boinc 6.6.2 it worked okay. It is now running Cuda 6.08, Seti Enhanced 6.03 and AstroPulse concurrently on 2 cores and 2 graphic cards. So it appears now to be able to run MultiBeam on the cpu and the gpu alongside each other. No optimized versions installed with this 6.6.2 yet. The app_info.xml for this is a bit too complicated for me to construct, I don't have enough experience to tackle a setup like this. I'll just have to wait until someone comes up with a working version of app_info.xml for a setup like this. Regards, John. |
Byron S Goodgame Send message Joined: 16 Jan 06 Posts: 1145 Credit: 3,936,993 RAC: 0 |
Soooo... You do have gpu enabled on your SETI@home preferences for that pc (though it probably wouldn't keep Boinc from seeing the card)? Also I was under the impression the CUDA app was only for Windows. Edit: Also since I know nothing about Linux this may not apply, but in Vista CUDA doesn't work if Boinc is installed as a service. I'm not sure if it could be the same for Linux. Updating the drivers might have an effect too if this hasn't already been done. |
ML1 Send message Joined: 25 Nov 01 Posts: 21019 Credit: 7,508,002 RAC: 20 |
Soooo... OK, good guesses... I'm running Mandriva 2009.0 GNU/Linux and Boinc runs as a 'service' on a restricted account. I previously added access to the graphics but... "Use GPU" was already enabled in my account settings. Strangely, the upload and download limits were x1000 what they used to be! I hope the units are KBytes and not perhaps just Bytes as they might have used to be (otherwise my transfers are now going to be very slow!). I removed the app_info.xml for s@h. Restarting Boinc and I got a new message: NVIDIA: could not open the device file /dev/nvidiactl (Permission denied). Mmmm... Checking /dev showed that /dev/nvidiactl and /dev/nvidia0 are locked to give exclusive access to only the user currently in control of the display. Boinc runs as a different user and so was locked out. The fix is to add the Boinc user to the "video" group and then as root (or sudo) the commands: chgrp video /dev/nvidia* chmod g+rw /dev/nvidia* Restart Boinc and: [---] CUDA device: GeForce 8600 GT (256MB, est. 17GFLOPS) Yae... Boinc then pesters all attached projects for something CUDA to do! Note: If you run Boinc as the same user that you are logged on as, then there should be no problem. If run as a 'service', then you need to do the "video" group trick. For my setup, I also need to modify the udev rule to set: KERNEL=="nvidia*|nvidiactl*", GROUP="video", MODE="0660" Now to see if that does fix it and survives a logoff/logon and/or reboot... And then to add the app_info.xml back in there... Happy fast crunchin', Martin See new freedom: Mageia Linux Take a look for yourself: Linux Format The Future is what We all make IT (GPLv3) |
ML1 Send message Joined: 25 Nov 01 Posts: 21019 Credit: 7,508,002 RAC: 20 |
OK... Adding the app_info.xml back in for the Linux CUDA app and... CUDA is running. With lot's more being downloaded! Another problem now is that the boincmgr is unusable! $ ./boincmgr connect: Operation now in progress SIGSEGV: segmentation violation Stack trace (28 frames): ./boincmgr(boinc_catch_signal+0x43)[0x52d883] /lib64/libpthread.so.0[0x7fba21ea4d20] ./boincmgr[0x51d5da] ./boincmgr[0x51fc67] ./boincmgr[0x443eb4] ./boincmgr[0x443a49] ./boincmgr[0x42cc39] ./boincmgr[0x6e032c] ./boincmgr[0x6e0a9f] ./boincmgr[0x6e0d93] ./boincmgr[0x43b6ba] ./boincmgr[0x6e032c] ./boincmgr[0x6e0a9f] ./boincmgr[0x6e0d93] ./boincmgr[0x6e112c] ./boincmgr[0x66f254] ./boincmgr[0x5ae1e6] ./boincmgr[0x55795b] /usr/lib64/libglib-2.0.so.0(g_main_context_dispatch+0x232)[0x7fba22e2c8d2] /usr/lib64/libglib-2.0.so.0[0x7fba22e3005d] /usr/lib64/libglib-2.0.so.0(g_main_loop_run+0x1cd)[0x7fba22e3058d] /usr/lib64/libgtk-x11-2.0.so.0(gtk_main+0xa7)[0x7fba259043d7] ./boincmgr[0x6184a0] ./boincmgr[0x5ae0e4] ./boincmgr[0x6974f0] ./boincmgr[0x44c362] /lib64/libc.so.6(__libc_start_main+0xe6)[0x7fba218cf316] ./boincmgr[0x41fcc9] Exiting... ???! Quite bemusing... All good fun! Martin [edit] That's completely repeatable over a few attempts. The Boinc Manager window gets drawn and populated with the menu titles and side buttons. The crash seems to be at the point that the tasks list should update to list the tasks for the first time. [/edit] See new freedom: Mageia Linux Take a look for yourself: Linux Format The Future is what We all make IT (GPLv3) |
Fred W Send message Joined: 13 Jun 99 Posts: 2524 Credit: 11,954,210 RAC: 0 |
OK... Adding the app_info.xml back in for the Linux CUDA app and... CUDA is running. With lot's more being downloaded! Not just a Linux problem. I've downloaded 6.6.2 onto my Vista x64 box. All was running hunky-dory until I had reason to re-boot and then exactly the same symptoms - a couple of lines in the Tasks tab and nothing at all in the Messages tab. Well - we were warned; "Expect the unexpected". F. [edit]And it just won't stop downloading tasks. I've got over 400 since I set NNT!![/edit] |
Fred W Send message Joined: 13 Jun 99 Posts: 2524 Credit: 11,954,210 RAC: 0 |
One further point on the "blind" Boinc Manager is that when this happens, a check of Windows Task Manager shows multiple instances of BoincManager running and non of them will shut down the connected clients. F. |
MadMaC Send message Joined: 4 Apr 01 Posts: 201 Credit: 47,158,217 RAC: 0 |
Yeah, Ive noticed that, but otherwise BM seems to be running OK, the only bit that is not working is the 'task' view, I get messgaes and can see work being uploaded, but I have no visibility on task process/progress. As soon as I click tasks, the manager crashes... |
Jord Send message Joined: 9 Jun 99 Posts: 15184 Credit: 4,362,181 RAC: 3 |
What screen resolution and colour depth do you run on when BM crashes? Do you use CUDA? If yes, on what kind of card? |
Fred W Send message Joined: 13 Jun 99 Posts: 2524 Credit: 11,954,210 RAC: 0 |
What screen resolution and colour depth do you run on when BM crashes? I'm running CUDA on a GTX295 @ 1440x900 32bit colour. Shouldn't be a vid memory problem on this beast. The problem seemed to disappear when I unloaded Raistmer's r103 AP (i.e. deleted the app_info) but I had only 10 mins to check before I had to leave for work. Another 2 or 3 hours before I can check again. F. |
BMaytum Send message Joined: 3 Apr 99 Posts: 104 Credit: 4,382,041 RAC: 2 |
... Another problem now is that the boincmgr is unusable! Boinc 6.6.2 is NOT working correctly for me either. Using WinXP Pro SP3 32bit, with Raistmer's V7 optimized MB_6.08_Mod_VLAR_kill_CUDA client application to crunch MB workunits on my 8800GT/512MB (G92) video card with nVidia 181.20 drivers. I've tried installing Boinc 6.6.2 over 6.6.0 (Boinc 6.6.2 crashes when started) and subsequently Uninstalling Boinc (Add/Remove in Control Panel) then installing 6.6.2. In both attempts, when 6.6.2 is started the Boinc GUI appears for several seconds, but blank in Tasks and Projects window, then Boinc GUI closes. Task Manager shows boinc.exe and the V7 Cuda client are still running. Moreover, Boinc 6.6.2 is "blind" as others noted to limits on WU downloads- after the Boinc GUI closed (but still running per Task Manager), my 6.6.2 downloaded about 50+ WUs in the 10 seconds it took for me to manually kill it in Task Manager. I've reverted to Boinc 6.6.0 for now, and am grinding through that pile of 50+ WUs using Raistmer's V7 opt Cuda client (fortunately, most of the 50+ WUs are shorties that crunch at a rate of 6-10 per hour). Assuming my difficulties aren't due to Raistmer's V7 Cuda client, I really do hope these apparent 6.6.2 problems get resolved because I've been anxiously looking forward to concurrently crunching MB units on GPU Cuda and on CPU (with optimized AK_V8 client). Sabertooth Z77, i7-3770K@4.2GHz, GTX680, W8.1Pro x64 P5N32-E SLI, C2D E8400@3Ghz, GTX580, Win7SP1Pro x64 & PCLinuxOS2015 x64 |
ML1 Send message Joined: 25 Nov 01 Posts: 21019 Credit: 7,508,002 RAC: 20 |
OK... Adding the app_info.xml back in for the Linux CUDA app and... CUDA is running. With lot's more being downloaded!... Looks like a fair few are rattling through... The first was 15dc08ae.16890.31750.8.8.210: 15dc08ae.16890.31750.8.8.210_1 The latest few are: 10dc08af.16856.3344.11.8.122_1 10dc08ae.17251.2113.16.8.0_1 10dc08ae.17251.2113.16.8.32_1 10dc08af.16856.3344.11.8.156_1 10dc08ae.17251.2113.16.8.18_1 10dc08af.16856.3344.11.8.178_0 10dc08ae.17251.2113.16.8.25_0 10dc08af.16856.3344.11.8.145_0 All have validated ok. More examples are pending. They all also show: [...] SETI@home MB CUDA 608 Linux 64bit SM 1.0 - r06 by Crunch3r :p [...] Cuda error 'cudaMalloc((void**) &dev_GaussFitResults' in file './cudaAcceleration.cu' in line 317 : out of memory. setiathome_CUDA: CUDA runtime ERROR in device memory allocation (Step 1 of 3). Falling back to HOST CPU processing... GPU memory leak? Or are all the WUs too big to process in what is left of the video 256MBytes after my few MBytes of display space? Further debug? Cheers, Martin See new freedom: Mageia Linux Take a look for yourself: Linux Format The Future is what We all make IT (GPLv3) |
Kokomiko Send message Joined: 15 Aug 04 Posts: 203 Credit: 157,712 RAC: 0 |
I had a strange behavior last night with the 6.6.2. SETI was set to No New Work and the Boinc Manager and the work cache to 1 day. Even so I set SETI to NNW the Boinc Manager was loading over 160 Seti extended WUs and 20 Astropuls WUs, thats much ure then the PC can do in one day. As a result he SETU WUs are all running in high prio mode. While I also have other projects running, I had to cancel the most of them. After a restart of the Boinc Manager the permanent loading of new work stops for a while. Then I restarted the whole PC and the Boinc Manager load again new work although SETI was set to NNW. What's wrong? Bug in the 6.6.2? I don't will downgrade to 6.6.0 while this is the first version with separate calculation for GPU and CPU. |
Fred W Send message Joined: 13 Jun 99 Posts: 2524 Credit: 11,954,210 RAC: 0 |
I had a strange behavior last night with the 6.6.2. SETI was set to No New Work and the Boinc Manager and the work cache to 1 day. Even so I set SETI to NNW the Boinc Manager was loading over 160 Seti extended WUs and 20 Astropuls WUs, thats much ure then the PC can do in one day. As a result he SETU WUs are all running in high prio mode. While I also have other projects running, I had to cancel the most of them. That exact behaviour has caused me to go back to 6.6.0 after I had received 1800 WU's in 2 days with NNT set for most of the time. F. |
©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.