[kaffe] Benchmarking kaffe using XSLTMark

[Dalibor Topic]

Hi Jim,

--- Jim Pick <jim at kaffe.org> wrote:
> On Thu, 2002-05-30 at 09:41, Dalibor Topic wrote:
> 
> > entries. That should reduce the cost of
> class/resource
> > lookup from
> > linear (scanning the classpath) to constant
> (single
> > classpath entry)
> > for most resources.
> 
> I recently read on the gcj list that Mark Mitchell
> had posted a patch
> for gcj that did something similar, which
> dramatically sped things up
> (especially on some platforms, where file/directory
> accesses are very
> expensive).

Do you have an URL ?

> I guess my only concern would be increased
> complexity and memory
> requirements for the cache...

I think that most lists of classpath entries will have
only one element, like those for "java/lang/ref/",
"kaffe/io/", etc. (i.e. rt.jar). there shouldn't be
that many packages anyway in a typical classpath. But
talk is cheap and code can be benchmarked, so I'll try
to implement something over the weekend.

cheers,

dalibor topic

__________________________________________________
Do You Yahoo!?
Yahoo! - Official partner of 2002 FIFA World Cup
http://fifaworldcup.yahoo.com