Index  | Recent Threads  | Unanswered Threads  | Who's Active  | Guidelines  | Search
 
World Community Grid Forums
  Category: Support
  Forum: GPU Support Forum
  Thread: GPU Work Units - Post Your Tech Support Questions Here

Quick Go ยป
No member browsing this thread
Thread Status: Active
Thread Type: Sticky Thread
Total posts in this thread: 290
Posts: 290   Pages: 29   [ Previous Page | 7 8 9 10 11 12 13 14 15 16 | Next Page ]
[ Jump to Last Post ]
Post new Thread
Author
Previous Thread This topic has been viewed 679067 times and has 289 replies Next Thread
goben_2003
Advanced Cruncher
Joined: Jun 16, 2006
Post Count: 145
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: GPU Work Units - Post Your Tech Support Questions Here
how would I go about making this app configuration file I am good with computers but not great with coding
I have this question too.

Hi,

The exact numbers to put in will depend on your specific hardware. Higher powered GPUs can run more of these opng units simultaneously.

To create the file open up a text editor such as notepad++ or the default windows notepad. Put the below code into it, adjust to your hardware, and save it as app_config.xml in the project directory. I think the default on windows is "C:\ProgramData\BOINC\projects\www.worldcommunitygrid.org". Then open up boinc manager and go to options->read local prefs files. If that does not exist, you are in simple view and need to switch to advanced view: View->Advanced view (or press Ctrl+Shift+a).

This example is for 2 work units to be processed at the same time, reserving 1 thread per gpu workunit. You can try that and see what the average gpu utilization is and then increase as necessary. To increase replace the .5 in gpu_usage with 1/Number desired. So 3=.33, 4=.25, etc.

If you have a more complex setup, like running a discrete gpu and an integrated gpu of a different brand, they can be setup differently. If that is the case, let me know.

I hope this helps! smile 

<app_config>
	<app>
		<name>opng</name>
		<gpu_versions>
			<gpu_usage>.5</gpu_usage>
			<cpu_usage>1</cpu_usage>
		</gpu_versions>
	</app>
</app_config>

----------------------------------------

----------------------------------------
[Edit 1 times, last edit by goben_2003 at Apr 29, 2021 8:43:46 PM]
[Apr 29, 2021 8:42:53 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Sagittarius Lupus
Cruncher
United States
Joined: Jan 5, 2008
Post Count: 3
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
confused Re: GPU Work Units - Post Your Tech Support Questions Here
Seeing fast failures of GPU work units on an AMD device (gfx900/Vega) using OpenCL on ROCm 4.0.0 with a consistent pattern that is repeatable if I run the task standalone:

../../projects/www.worldcommunitygrid.org/wcgrid_opng_autodockgpu_7.28_x86_64-pc-linux-gnu__opencl_ati_102 -jobs OPNG_0021702_00043.job -input OPNG_0021702_00043.zip -seed 579781518 -wcgruns 4100 -wcgdpf 82
INFO: Using gpu device from app init data 0
INFO:[01:44:42] Start AutoGrid...

autogrid4: Successful Completion.
INFO:[01:46:46] End AutoGrid...
INFO:[01:46:46] Start AutoDock for OB3ZINC000358915657--7jji_002_mgltools--TYR380_inert.dpf(Job #0)...
OpenCL device:                           gfx900
:0:rocdevice.cpp            :2302: 163984884746 us: Device::callbackQueue aborting with status: 0x29
SIGABRT: abort called
Stack trace (10 frames):
../../projects/www.worldcommunitygrid.org/wcgrid_opng_autodockgpu_7.28_x86_64-pc-linux-gnu__opencl_ati_102[0x4532b2]
/lib64/libpthread.so.0(+0x13140)[0x7f5d455d2140]
/lib64/libc.so.6(gsignal+0x10a)[0x7f5d4543402a]
/lib64/libc.so.6(abort+0x119)[0x7f5d4541d53a]
/usr/lib64/libamdocl64.so(+0xcbd9a)[0x7f5d45302d9a]
/usr/lib64/libhsa-runtime64.so.1(+0x38674)[0x7f5d44723674]
/usr/lib64/libhsa-runtime64.so.1(+0x5c162)[0x7f5d44747162]
/usr/lib64/libhsa-runtime64.so.1(+0x21238)[0x7f5d4470c238]
/lib64/libpthread.so.0(+0x8084)[0x7f5d455c7084]
/lib64/libc.so.6(clone+0x3f)[0x7f5d454fdb7f]

I had a few of these run successfully on a different version of the ROCm OpenCL runtime, so I'm going to revert to that and try again, but I'm looking for anyone else who might have encountered this.
----------------------------------------
PSA: this wolf eats guns, muscle cars, rebel battle flags, their owners, bigots, fascists, viral vectors, Republicans, and anything associated with the number 45.
[Apr 30, 2021 6:48:45 AM]   Link   Report threatening or abusive post: please login first  Go to top 
Sagittarius Lupus
Cruncher
United States
Joined: Jan 5, 2008
Post Count: 3
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
smile Re: GPU Work Units - Post Your Tech Support Questions Here
The OpenCL runtime for ROCm 4.1.0 does work, although it is not clear to me what precisely the difference is.
----------------------------------------
PSA: this wolf eats guns, muscle cars, rebel battle flags, their owners, bigots, fascists, viral vectors, Republicans, and anything associated with the number 45.
[Apr 30, 2021 7:26:10 AM]   Link   Report threatening or abusive post: please login first  Go to top 
Michael E.@ team Carl Sagan
Cruncher
U.S.A.
Joined: Jul 18, 2010
Post Count: 28
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: GPU Work Units - Post Your Tech Support Questions Here
My PC has an NVIDIA GPU (GeForce GTX 1050ti) as well as an Intel GPU (UHD Graphics 630). I am trying to run GPU tasks on both simultaneously. The NVIDIA one works as expected with completion in about 14 minutes. I suspect that is the active GPU.

The Intel GPU task is still running after 4 hours. It got to 99% completion an hour ago. It has been at 100% completion and 00:00:00 remaining time for about 30 minutes. Will it complete or should I abort it?

A general GUI question: I do not know to view the details of each task in WCG. It is possible?
----------------------------------------
[Edit 1 times, last edit by Michael E. at Apr 30, 2021 8:17:02 PM]
[Apr 30, 2021 7:42:43 PM]   Link   Report threatening or abusive post: please login first  Go to top 
dsalt2
Cruncher
Joined: Aug 17, 2008
Post Count: 1
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: GPU Work Units - Post Your Tech Support Questions Here
I'm getting a lot of OpenPandemics GPU units FAILING part way through with "Computation error", i mean like the majority of them are failing. I've no idea why, can anyone advise?
[Apr 30, 2021 9:45:06 PM]   Link   Report threatening or abusive post: please login first  Go to top 
goben_2003
Advanced Cruncher
Joined: Jun 16, 2006
Post Count: 145
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: GPU Work Units - Post Your Tech Support Questions Here
My PC has an NVIDIA GPU (GeForce GTX 1050ti) as well as an Intel GPU (UHD Graphics 630). I am trying to run GPU tasks on both simultaneously. The NVIDIA one works as expected with completion in about 14 minutes. I suspect that is the active GPU.

The Intel GPU task is still running after 4 hours. It got to 99% completion an hour ago. It has been at 100% completion and 00:00:00 remaining time for about 30 minutes. Will it complete or should I abort it?

A general GUI question: I do not know to view the details of each task in WCG. It is possible?

This is in Advanced View:
Go to the Tasks Tab
click on the task
click Properties
The line "Directory" will give you the slot number. Something like "slots/3"
Go to the slots folder inside the BOINC Data folder. In windows this is "C:\ProgramData\BOINC\slots"
Go to the folder number for the slot for the work unit. In this example that is the folder called "3"
Open the file called "stderr.txt"
In this file you should be able to see if it is making progress or if it stopped doing work about when it started saying 100% complete.

Also, if this is in Windows, you may want to look on the performance tab of task manager to see if it looks like the intel gpu is in use. I have the same intel gpu, it shows up properly in task manager by default(It uses the 3D graph for me).
----------------------------------------
[Apr 30, 2021 9:59:55 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Michael E.@ team Carl Sagan
Cruncher
U.S.A.
Joined: Jul 18, 2010
Post Count: 28
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: GPU Work Units - Post Your Tech Support Questions Here
@goben_2003 - Thank you! Great tips for my Windows toolbox! Yes this is on a Windows PC. As a BOINC user for decades, I never found the Properties button.

The task did eventually finish while I was away, but it took about 5+ hours. I will avoid the Windows GPU tasks for now, but am willing retest at a later date.

So is there any way to view the time and points for each complete task using WCG?
[Apr 30, 2021 11:55:31 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Eugene Zenzen
Veteran Cruncher
USA
Joined: Mar 31, 2006
Post Count: 889
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: GPU Work Units - Post Your Tech Support Questions Here
Thank you goben_2003 - I'll give more than one gpu work units at a time a try!
----------------------------------------
[May 1, 2021 1:13:26 AM]   Link   Report threatening or abusive post: please login first  Go to top 
goben_2003
Advanced Cruncher
Joined: Jun 16, 2006
Post Count: 145
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: GPU Work Units - Post Your Tech Support Questions Here
@goben_2003 - Thank you! Great tips for my Windows toolbox! Yes this is on a Windows PC. As a BOINC user for decades, I never found the Properties button.

The task did eventually finish while I was away, but it took about 5+ hours. I will avoid the Windows GPU tasks for now, but am willing retest at a later date.

So is there any way to view the time and points for each complete task using WCG?

You're welcome!

Just a note that the GPU tasks are quite a bit larger than they were to start. They started out as 20x the size of a GPU unit, and they seem to take several times as long now. I am not sure what the multiplier is now, but I suspect it is at least 40x. So it is like running at least 40 CPU units in that time. Feel free to not run them on the iGPU though, my UHD 620/630s take approximately 30-90 minutes depending on the specific work unit. The rtx 2060m is 15-24 minutes with running 10 at a time. smile

You can see the time and points for each task on the Results Status page. You can filter by device, project, and work unit status, such as Valid to only see the successfully completed work units. Note that they only stay on the page for a limited amount of time, so the old ones disappear. I use BoincTasks, where you can set the history to be longer. However, it does not know the points generated per task as that is not recorded locally. You do get some other information though, along with a few graphs.
----------------------------------------
[May 1, 2021 10:17:36 AM]   Link   Report threatening or abusive post: please login first  Go to top 
goben_2003
Advanced Cruncher
Joined: Jun 16, 2006
Post Count: 145
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: GPU Work Units - Post Your Tech Support Questions Here
Thank you goben_2003 - I'll give more than one gpu work units at a time a try!

You're welcome! smile
----------------------------------------
[May 1, 2021 10:18:18 AM]   Link   Report threatening or abusive post: please login first  Go to top 
Posts: 290   Pages: 29   [ Previous Page | 7 8 9 10 11 12 13 14 15 16 | Next Page ]
[ Jump to Last Post ]
Post new Thread