Message boards :
Number crunching :
OpenCL MB v8.12 issues thread
Message board moderation
Author | Message |
---|---|
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
If you see lags that prevent normal PC operation or driver restart or invalids please post you config and other circumstances here. Credit issues go elsewhere. It's app's technical support thread so stay on topic - be warned. |
Keith Myers Send message Joined: 29 Apr 01 Posts: 13164 Credit: 1,160,866,277 RAC: 1,873 |
Raistmer, can you point me at the specific thread and messages that list the parameters needed to run the new SoG app. I need the recommended settings for both the MB command line file and the new sections for the app_config file to run multiple SoG tasks on two GPU's in each of my computers. I assume the aistub file is correct for the app_info entry for the 8.12 app and doesn't need editing. Is this correct? Seti@Home classic workunits:20,676 CPU time:74,226 hours A proud member of the OFA (Old Farts Association) |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
Raistmer, can you point me at the specific thread and messages that list the parameters needed to run the new SoG app. I need the recommended settings for both the MB command line file and the new sections for the app_config file to run multiple SoG tasks on two GPU's in each of my computers. I assume the aistub file is correct for the app_info entry for the 8.12 app and doesn't need editing. Is this correct? There is no "aistub" for stock app. Also, there is no app_info.xml for stock app. Those belong to anonymous platform and standalone app pack one can download separately. For stock app all tuning go into app_config.xml and app-specific mb_cmdline*.txt file if needed. There is one more app-specific config file handy in case of mixed capabilities multi-GPU host. Consult with app's ReadMe on usage. Try tune line recommended by Mike elsewhere and check that -sbs 256 mentioned if you want to speedup over defaults. I have no compatible hardware by myself so not the right person to ask about tuning line. |
Keith Myers Send message Joined: 29 Apr 01 Posts: 13164 Credit: 1,160,866,277 RAC: 1,873 |
I run the Anonymous platform and Lunatics. I either have your package from your repository or the one on Mike's World. I've looked at the packages and have tried to understand all the many, many choices of parameter options. Still not too clear and just how to set up my 970's. I remember seeing Mike's command line options in another message but I don't understand how to coordinate the necessary settings for both the command line file and the app_config entries. Zalster has offered help in another thread and I will wait till he chimes in. Thanks for the reply. Seti@Home classic workunits:20,676 CPU time:74,226 hours A proud member of the OFA (Old Farts Association) |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
I run the Anonymous platform and Lunatics. ... but I don't understand how to coordinate the necessary settings for both the command line file and the app_config entries. For what exact reason you use app_config.xml along with app_info.xml ? app_info.xml is enough in case of anonymous platform. |
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13736 Credit: 208,696,464 RAC: 304 |
For what exact reason you use app_config.xml along with app_info.xml ? It's much easier to edit 1 entry in app_config.xml to change the number of WUs on a GPU than to edit multiple entries in app_info.xml and not stuff something up. Grant Darwin NT |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14650 Credit: 200,643,578 RAC: 874 |
app_info.xml is enough in case of anonymous platform. App_config.xml is quicker, simpler, and safer for testing the results of minor changes - which can be deployed via a 'read config files', without restarting the client. I've been trying to shift people in that direction since it was first made available - though of course you have to be running a "new enough" client (v7.0.40, from memory). @ Raistmer, Please can you confirm exactly which application(s) has/have been deployed at Main as v8.12? You've had my comments on r3340 at Beta: I see mention of a r3347 this morning. |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
For what exact reason you use app_config.xml along with app_info.xml ? If Keith Myers use app_config.xml for that reason, then, accordingly to app's ReadMe -instances_per_device N :Sets allowed number of simultaneously executed GPU app instances per GPU device (shared with MultiBeam app instances). , next option should be provided: -instances_per_device N where N should be same number of tasks per GPU that supplied to app_config. In case of multi-vendor host worth to add, accordingly to ReadMe: -total_GPU_instances_num N : To use together with -cpu_lock on multi-vendor GPU hosts. Set N to total number of simultaneously running GPU -total_GPU_instances_num N where N is total number of OpenCL GPU tasks host will run simultaneously. Currently limited to 64. |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
Cause v8.12 currently deployed only for nVidia hosts and I have no compatible hardware I can't confirm which app deployed even moving from anonymous platform to stock. |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14650 Credit: 200,643,578 RAC: 874 |
@ Raistmer, Oh, s**t. Can you even advise me which applications you have submitted for consideration? And link me up with whoever has been doing your testing? (OK, so I can read it from stderr when the results start coming in) If anyone is already receiving opencl_nvidia_sah and/or opencl_nvidia_SoG tasks, could they dig into the project folder, please, and dig out the exact executable filename(s), please? (I can sort out the rest of the download url myself). I'll try and throw an installer together - the old one will be broken by these new version numbers (cached sah/SoG tasks will be trashed if it's run as it stands) - but this has come at a really, really bad time for me. |
tullio Send message Joined: 9 Apr 04 Posts: 8797 Credit: 2,930,782 RAC: 1 |
setiathome_8.12_windows_intelx86_opencl_nvidia_SoG This on my Windows 10 PC Tullio |
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13736 Credit: 208,696,464 RAC: 304 |
setiathome_8.12_windows_intelx86_opencl_nvidia_SoG setiathome_8.12_windows_intelx86_opencl_nvidia_SoG.exe? EDIT- Is case important (upper/lower)? Grant Darwin NT |
tullio Send message Joined: 9 Apr 04 Posts: 8797 Credit: 2,930,782 RAC: 1 |
This is the string I found in the projects/setiathome directory, defined "Application". Tullio |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14650 Credit: 200,643,578 RAC: 874 |
It's OK, I've got it. Windows (by default) hides the .exe extension, so I added that automatically. More importantly, the actual filename is setiathome_8.12_windows_intelx86__opencl_nvidia_SoG.exe - Eric is very consistent about putting a double underscore between the 'platform' and 'coprocessor' sections of the filename. Now, where's my hex editor? :-D Edit - b****r. "OpenCL version by Raistmer, r%d". Plan B... |
Mike Send message Joined: 17 Feb 01 Posts: 34258 Credit: 79,922,639 RAC: 80 |
You could simply download r3430_SOG from my website. With each crime and every kindness we birth our future. |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14650 Credit: 200,643,578 RAC: 874 |
You could simply download r3430_SOG from my website. I've got that already - I reported non-linear progress indicator and a 'finish file present too long' error. I had hoped those would be attended to before release, but I can confirm the stock app is... OpenCL version by Raistmer, r3430 Edit - do you happen to know what the current 'minimum driver' restriction is? And is it affected by Jacob Klein's OpenCL bug in 364.72? |
Mike Send message Joined: 17 Feb 01 Posts: 34258 Credit: 79,922,639 RAC: 80 |
You could simply download r3430_SOG from my website. Not exactly, but i know that 358.87 works fine. With each crime and every kindness we birth our future. |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
So you confirmed that all goes accordingly to plan. Good.
No issues were reported with much prolonged than needed beta test. So, I assume answer is NO. EDIT: minimum driver restriction was determined by my testing in separate thread. What is used in plan class by Eric - no idea. |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14650 Credit: 200,643,578 RAC: 874 |
EDIT: minimum driver restriction was determined by my testing in separate thread. What is used in plan class by Eric - no idea. Originally, the minimum driver was enforced by the application itself - I remember testing a deliberately 'unrestricted' r3431. Eric's plan_class won't affect anonymous platform deployment via an installer - but I need to place a "driver upgrade may be needed" warning on the face of the installer, because many CUDA users will be happily running the older, more efficient, driver versions: the CUDA apps don't have so many driver compatibility issues. For example, my GTX 470 is running driver: 335.23, and my GTX 670 is at driver: 344.11. From memory, I think both of those will fail? |
tullio Send message Joined: 9 Apr 04 Posts: 8797 Credit: 2,930,782 RAC: 1 |
I am using Nvidia driver 365.19. Tullio |
©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.