[PATCH] mmap_cache: Use prime muber for size of hash table.

[Lukas Slebodnik]

ehlo,

yet another patch to improve mmap_cache.

LS

[Simo Sorce]

On Mon, 2013-08-19 at 13:35 +0200, Lukas Slebodnik wrote:
> ehlo,
>
> yet another patch to improve mmap_cache.

Do you have any metrics that warrants this change ?

Simo.

[Stephen Gallagher]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 08/19/2013 08:20 AM, Simo Sorce wrote:
> On Mon, 2013-08-19 at 13:35 +0200, Lukas Slebodnik wrote:
> > ehlo,
> >
> > yet another patch to improve mmap_cache.
>
> Do you have any metrics that warrants this change ?

Please do not do this. Primes *rapidly* get further between as it gets
larger. While I don't expect that our memory cache is likely to get
larger than a few hundred entries, the processing time necessary to
identify a prime number above that will grow prohibitively more
difficult. The potential hash-collision avoidance is not worth the cost.
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.14 (GNU/Linux)
> Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/
>
> iEYEARECAAYFAlISGIwACgkQeiVVYja6o6NRGwCgobCv/u5jrFxX9SLdOxQ3V9o2
> jHcAnA+2WRVb2MX+HhuUVeHE/W1N1MAM
> =oz4B
> -----END PGP SIGNATURE-----

[Lukas Slebodnik]

On (19/08/13 08:20), Simo Sorce wrote:
> On Mon, 2013-08-19 at 13:35 +0200, Lukas Slebodnik wrote:
> > ehlo,
> >
> > yet another patch to improve mmap_cache.
>
> Do you have any metrics that warrants this change ?
>
> Simo.
>
> -- 
> Simo Sorce * Red Hat, Inc * New York

I will do some measurements later, because it is not high priority patch. It is
only possible improvement.

LS