[openib-general] Re: [PATCH] ipoib tune rx ring size

[Michael S. Tsirkin]

Quoting r. William Jordan <bjordan.ics at gmail.com>:
> Subject: Re: [PATCH] ipoib tune rx ring size
> 
> On 6/20/05, Grant Grundler <iod00d at hp.com> wrote:
> > On Mon, Jun 20, 2005 at 06:14:04PM +0300, Michael S. Tsirkin wrote:
> > > Reducing the rx ring size gives me bandwidth win of about 1%.
> > 
> > Michael,
> > Doesn't this sound counter-intuitive to you?

The explanation could be RX livelock 4x HCA is already fast enough
to saturate the CPU.

> > Can you share which configuration (chipset, CPU, IO bus, HCA)
> > you measured this on?
> > 
> > I've also seen this to be true with gige - but only in one odd configuration.
> 
> I'd be skeptical also. I've found this to be true in only one very
> contrived instance.
> 

Well, I see it in practice, and its easy enough to reproduce even with
simplest netperf benchmark.
I have two back-to-back systems.
Each has two xeons 3.4G , 1Mb cache in x86_64 mode.
The systems are intel nocona, HT disabled, with arbel in tavor mode.

Grant, are you willing to test the patch and post what effect if any this has
on netperf?

-- 
MST