[linux-mm-cc] Fwd: [RFC][PATCH 0/6] compcache: Compressed Caching

[Anderson Briglia]

On Sat, Apr 12, 2008 at 8:50 AM, Nai Xia <nai.xia at gmail.com> wrote:
> On Thu, Apr 10, 2008 at 12:04 AM, Nitin Gupta <nitingupta910 at gmail.com> wrote:
>  > ---------- Forwarded message ----------
>  >  From: Andrew Morton <akpm at linux-foundation.org>
>  >  Date: Wed, Apr 9, 2008 at 8:17 AM
>  >  Subject: Re: [RFC][PATCH 0/6] compcache: Compressed Caching
>  >  To: nitingupta910 at gmail.com
>  >  Cc: linux-mm at kvack.org
>  >
>  >  On Fri, 21 Mar 2008 01:29:58 +0530 Nitin Gupta <nitingupta910 at gmail.com> wrote:
>  >   > For general desktop use, this is giving *significant* performance gain
>  >   > under memory pressure. For now, it has been tested only on x86.
>  >
>  >
>  >  <snip>
>  >
>  >   The values of "*significant*" should be exhaustively documented in the
>  >   patch changelogs. That is 100%-the-entire-whole-point of the patchset!
>  >   Omitting that information tends to reduce the number of C's.
>  >
>  >  </snip>
>  >
>  >  <End>
>  >
>  >
>  >
>  >  So, I will require performance numbers to justify inclusion of compcache
>  >  in mainline. We already have separate perf numbers for lzo(compressor) and
>  >  tlsf(allocator) but none that shows that using compcache is good for
>  >  perf under memory pressure.
>  >
>  >  If anyone is willing to take up this task, you can refer this paper:
>  >  http://ols.108.redhat.com/2007/Reprints/briglia-Reprint.pdf
>
>  I'd like to help, and
>  1. do you think the benchmark from memtest is enough convincing?
>  shall we introduce some additional tests?

Additional tests should be great. I believe memtest is not enough. For
embedded system some developers use Mibench[1] as benchmarking.

>  2. I plan to use a x86 destop, is it ok? or does the Compressed Caching
>  focus mainly on embedded systems?

Nitin, correct me if I'm wrong but this ccache version is focused on
embedded systems, right? The version used on that paper[2] was focused
on "swapless" devices, not on embedded systems. Tests should include
embedded system scenario.

[1] http://www.eecs.umich.edu/mibench/
[2] http://ols.108.redhat.com/2007/Reprints/briglia-Reprint.pdf

>
>
>  >
>  >  This paper shows benchmarks of previous implementation of memory compression.
>  >  It will give good idea on what to measure.
>  >
>  >
>  >  Thanks,
>  >  Nitin
>  >  _______________________________________________
>  >  linux-mm-cc mailing list
>  >  linux-mm-cc at lists.laptop.org
>  >  http://lists.laptop.org/listinfo/linux-mm-cc
>  >
>
>  Best Regards,
>  Nai
>
>
> _______________________________________________
>  linux-mm-cc mailing list
>  linux-mm-cc at lists.laptop.org
>  http://lists.laptop.org/listinfo/linux-mm-cc
>