Well.. we have http://squid.sourceforge.net/compactsentry/
Some progress has been made, but it is not yet stable and there is too
many other things requiring my attention at the moment.. if this gets
combined with chunked memory pools then there might be a measurable
difference (both reduce the memory overhead).
On the other hand we also have the plans on making index-less stores
where the index is integral in the on-disk structures. This would
eleminate most of Squid's memory use, instead moving it to disk
buffering requirements (which is quite fixed in size per load, not
linear to the disk size as today)
/Henrik
Chemolli Francesco (USI) wrote:
> It could be an interesting project to try and compact
> frequently-used data-stuctures in order to increase processor-cache
> performance.
Received on Mon Apr 23 2001 - 16:22:42 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:13:48 MST