Re: [Lse-tech] zerocopy pipe, new version

Hubertus Franke (frankeh@watson.ibm.com)
Wed, 16 Jan 2002 10:29:54 -0500


On Tue, Jan 15, 2002 at 03:20:38PM -0500, Benjamin LaHaise wrote:
> On Tue, Jan 15, 2002 at 09:08:45PM +0100, Manfred Spraul wrote:
> > My patch is definitively WIP - right now I again broke the -ENOMEM and
> > -EFAULT handling.
>
> I am aware of that, but the lse-tech posting made it sound as if things are
> great now since the SMP numbers improved. Please folks, remember that UP
> numbers are important too.
>
> -ben
> --
> Fish.

Ben, yes you are right, the lse-posting in a second reading is misleading.
As reported previously http://lse.sourceforge.net/pipe/pipe-report
UP numbers see degradations for LM-Bench, other benchmarks are OK.

This is not solved either by an integration of zero-copy with large pipes.
We have however shown that the for SMP systems adding larger pipes to
zero-copy pipes makes sense and for UP and 1-way sticking with a 1-page
pipe does not degradate Manfred's patch.

It hence boils down to a proper parameterization of the pipe dependent on
the configuration.

-- Hubertus

-
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/