>
> > Another funny thing is that the latency for the gigabit
> > adapter (e1000) is also higher than fast ethernet (eepro100)
> > with small messages (<256 bytes) :
>
> You could try setting the RxIntDelay module parameter to 0, that should
> improve round trip latency. The balance between latency on the receive path
> and interrupt rate can be difficult to manage, hopefully a dynamic method
> like NAPI will result in Ethernet drivers that need less hand tuning.
>
Hmm the docs say that RxIntDelay is by default 0 but when I tried it, it
helped for sure (although the numbers are still a bit higher that with
eepro100).
Thanks,
-- Steffen Persvold | Scalable Linux Systems | Try out the world's best mailto:sp@scali.com | http://www.scali.com | performing MPI implementation: Tel: (+47) 2262 8950 | Olaf Helsets vei 6 | - ScaMPI 1.13.8 - Fax: (+47) 2262 8951 | N0621 Oslo, NORWAY | >320MBytes/s and <4uS latency- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/