On Mon, 31 Jul 2000, you wrote:
> Hi,
>
> I work for a neuroscience lab at Caltech
> (http://broccoli.caltech.edu/~pinelab/PotterGroup.htm), and we are trying
> to gather data using a quad-processor SGI Visual Workstation 540. However,
> we are having trouble that we have diagnosed as a timing issue.
> gettimeofday() is non-monotonic, which is causing problems with some of
> the software we use, problems we can't really work around. We are using
> the 2.2.10 kernel, and we really have no choice as all of the hacks to
> make linux run on the Visual Workstation 540 are only available in a patch
> for kernel version 2.2.10 and none of us are very skilled kernel hackers.
>
> Can anyone point me to a patchfor 2.2.10 or a way to quickly modify the
> 2.2.10 code to fix this problem. Something somewhat odd is that the
> gettimeofday() seems to pretty consistently jump back about 3-4 seconds.
> It's also only present when running Xwindows and is highly correlated
> with changes in window focus.
This sounds almost exactly like the problems I was having when I
used to have an SMP machine with different speed CPUs. Is this
true here? Or are all of your processors running at the same
speed?
I wrote a couple of patches which are available at
http://hypocrite.org/linux/ (the ...new... patch is most likely
to help). These are against 2.3.99-pre if memory serves but
should port easily to earlier kernels.
Alternatively, try compiling and telling the kernel that you are
using a 586-possibly-without-TSC which *may* help.
I am most interested in hearing the results...
-- Christopher Thompson http://hypocrite.org/ All spam is reported - The average chocolate bar has 8 insects' legs in it. - The average human eats 8 spiders in their lifetime at night.- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Mon Jul 31 2000 - 21:00:35 EST