[sldev] Cache politics: performance vs obfuscation

Laurent Laborde kerdezixe at gmail.com
Tue Jun 10 11:57:49 PDT 2008


2008/6/10 Dahlia Trimble <dahliatrimble at gmail.com>:
>
> Further, noone is going to be inducted into the programmer hall of fame for
> writing a binary file to a disk. If you want to do something that is a
> benefit to the community, develop a method that improves cache performance,
> saves network bandwidth, and doesnt alienate a large portion of the user
> base. If you want to make a cache browser/exporter, then call it that. I see
> no reason that a standardized image format would necessarily be the highest
> performance choice for a cache.

There is one good reason to use a standardized image format and it's
directly related to performance.
I widely used open format have tons of coder working on the libraries
to use this format and work on optimisation/performance.

A library used by millions of users, including large corporate
company, is most likely highly optimized, stable, reliable and secure.
Just take a look at the discussion about "commercial jpeg2000
librarie" vs "openjpeg2k librarie".

that's why we don't even talk about trying to improve the decoding
libraries performance anymore : it was done already, and by external
coders.
If LL develop a custom image format, it will cost a lot of ressource
to work on optimization, etc ...

-- 
F4FQM
Kerunix Flan
Laurent Laborde


More information about the SLDev mailing list