FYI: compcache is in mainline (scheduled for 2.6.33)

[Nelson Castillo]

On Wed, Jan 6, 2010 at 11:47 AM, Nicolas Dufresne
<nicolas.dufresne at gmail.com> wrote:
> Le mercredi 06 janvier 2010 à 16:24 +0100, Patryk Benderz a écrit :
>
> overhead to compress pages might actually result in decreasing FR's

The LZO compression algorithm is very fast. I've used LZO it a few times.
http://en.wikipedia.org/wiki/LZO

> Note that this is all about in RAM swap. If we saturate the RAM on FR and
> use flash for SWAP, you end up with something just unusable. Considering
> this feature, you reduce RAM space which will cause swapping, but the swap
> access is a lot faster. With this feature we could augment caching within
> software and kernel and reduce IOs to the SD cards (which are very slow on
> this phone).

I had a good experience with compcache about one and half years ago on
a IA-32 laptop with 256MB of RAM. By then it was unstable on the ARM
so I didn't try it.

I would try a small swap (10M?) in RAM with higher priority and the
usual SD card swap space.

But later... right now upstream is crashing and I still haven't
found/looked for the cause :-/