Bug 719878

Summary:	GPU not detected until boinc service is restarted
Product:	[Fedora] Fedora	Reporter:	Germano Massullo (Thetra) <germano.massullo>
Component:	boinc-client	Assignee:	Milos Jakubicek <xjakub>
Status:	CLOSED RAWHIDE	QA Contact:	Fedora Extras Quality Assurance <extras-qa>
Severity:	unspecified	Docs Contact:
Priority:	unspecified
Version:	15	CC:	cheekyboinc, g.laurent, mmahut, xjakub
Target Milestone:	---
Target Release:	---
Hardware:	Unspecified
OS:	Unspecified
Whiteboard:
Fixed In Version:		Doc Type:	Bug Fix
Doc Text:		Story Points:	---
Clone Of:
Clones:	879776 (view as bug list)		Environment:
Last Closed:	2012-02-08 14:23:38 UTC	Type:	---
Regression:	---	Mount Type:	---
Documentation:	---	CRM:
Verified Versions:		Category:	---
oVirt Team:	---	RHEL 7.3 requirements from Atomic Host:
Cloudforms Team:	---	Target Upstream Version:
Embargoed:

Description Germano Massullo (Thetra) 2011-07-08 09:20:18 UTC

I have a nVidia Geforce GTX 260 with 275.09.07 drivers.
When I start the computer, BOINC tolds me that there is not a GPU for GPU computing.
If I restart the BOINC service, BOINC reconnize correcyl the GPU and does all the calculus.
I think this problem is caused by the fact that BOINC server is loaded before nVidia drivers during boot sequence. Is it possible?

Comment 1 Milos Jakubicek 2011-08-20 12:22:28 UTC

By GPU computing you mean that you manually installed the CUDA library?

Comment 2 Germano Massullo (Thetra) 2011-08-20 13:31:56 UTC

I installed akmod-nvidia from rpm fusion months ago when I setted up the computer

Comment 3 Milos Jakubicek 2011-08-20 16:29:51 UTC

Interesting, looks like this is now enough to enable GPU computing. Unfortunately I have no access to a machine where I could test this (no Nvidia/ATI). In another forum I've found some suspect that X server must be already running in order to recognize the GPU, that's sort of a problem: we do not want to start boinc after X server (for CPU computing it does not make sense, it should be enough that you power on the machine). So probably the best solution would be to REstart boinc when the X starts, it is certainly not optimal but I cannot come up with anything better than that. I'll try to get to a machine where I could test this.

Comment 4 Germano Massullo (Thetra) 2011-08-23 10:37:54 UTC

I suggest you to talk with other BOINC developers, they may have the same our problem on other distros.

Comment 5 GuL 2011-12-23 10:14:36 UTC

(In reply to comment #3)
> In another forum I've found some suspect that X server must be
> already running in order to recognize the GPU

I confirm. And it is not possible using sysV or rc.local as in F16 X server is started by systemd and explicity waiting for rc.local to finish (/lib/systemd/system/prefdm.service), even if we add a sleep 30 line inside rc.local.

That's the reason why I am trying to launch boinc using systemd. See https://bugzilla.redhat.com/show_bug.cgi?id=754600

> that's sort of a problem: we do
> not want to start boinc after X server (for CPU computing it does not make
> sense, it should be enough that you power on the machine).

I can understand that on a machine without X. But if you've got X and the default runlevel is graphical, what is the problem ? What I can suggest is to have two different policies, one in the package boinc-client for the cpu only and one in the package boinc-manager for the cpu+gpu.

> So probably the best
> solution would be to REstart boinc when the X starts, it is certainly not
> optimal but I cannot come up with anything better than that. I'll try to get to
> a machine where I could test this.

If I can test something, tell me

Comment 6 Milos Jakubicek 2012-02-08 14:23:38 UTC

Hopefully this should be fixed with the upcoming switch to systemd that I just built for F17/F18, please reopen otherwise.