Re: [Myrinet] weird latency

On Tue, Jun 11, 2002 at 05:30:27PM -0700, Andrew Gaul wrote:
> I am measuring the latency between two hosts with Myrinet adapters and
> have some curious results(*).  I am using GM 1.5.1 on 33MHz 32bit
> machines and am getting 65-70 us round-trip for 1000 byte packets using
> a ping utility I wrote(**).  There are >70 us outliers, which I expect,
> but I also see <65 us outliers, clustered together in predictable
> intervals (~13 ms).  What could be causing these?  I am running Linux
> 2.4.18 and have scheduled the process with the soft real-time scheduler.
> I am measuring times by reading the cycle counter available on Pentium
> and newer processors.

Killing X brought "fixed" the erratic timings by increasing the number
of points where the negative outliers were.  I don't have a good
explanation for the behaviour, although I did notice that the period of
the outliers changed with the refresh rate.  It could be an artifact of
the on-board video that I'm using or simply PCI contention (although the
video should be on the dedicated AGP bus).

Thanks all,

-- 
Andrew Gaul
[email protected]