[petsc-dev] CUDA sm_13 for double

[Filippo Spiga]

Dear Satish,
    my workstation has 3 TESLA C1060 and 1 GeForce 9500 GT.

On Thu, Jan 13, 2011 at 3:22 AM, Satish Balay <balay at mcs.anl.gov> wrote:
>
> Potential issues:
> - the machine has 2 nvidia cards with different version
> - the card is locked by a different user - running cuda code [at configure
> time]
> - the compile node doesn't have the nvidia cards [like on a cluster..]
>

I solved the first two issues in this way
1. I disabled the GeForce card setting the 'compute mode' as 'Prohibited'.
In this way, no CUDA kernel can run on that card
2. I set the 'compute mode' of every TESLA card as 'Exclusive'. In this way,
only one kernel per time can run on the "selected" device.

Disabling the GeForce, when I run a CUDA kernel only the 3 TESLA card are
potentially selected to run my code and, as soon as the driver assign me a
specific card, that card will be only mine until the kernel is finished.

About the 3 issue, the Makefile can check if the compiler exists...

I never used directly CUDA on my macbook, but I am quite sure that on a
Linux system these tricks work well.

Cheers

--
Filippo SPIGA, MSc Computer Science
~ homepage: http://tinyurl.com/fspiga ~

«Nobody will drive us out of Cantor's paradise.»
    -- David Hilbert

*****
Disclaimer: "Please note this message and any attachment are CONFIDENTIAL an
may be privileged or otherwise protected from disclosure. The contents are
not to be disclosed to anyone other than the addressee. Unauthorized
recipients are requested to preserve this confidentiality and to advise the
sender immediately of any error in transmission."
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20110113/6bbc866e/attachment.html>