Re: [xml] xmlbench Benchark results...

[Sander Vesik <sander_traveling yahoo co uk>]

Luca Padovani wrote:
> On Wed, 19 Mar 2003, Daniel Veillard wrote:
>
> > On Tue, Mar 18, 2003 at 04:23:45PM -0800, Sean Chittenden wrote:
> >
> > > I apologize if others have seen this, but:
> > >
> > > http://xmlbench.sourceforge.net/results/benchmark/index.html
> > >
> > > libxml performs quite well in the stack up.  Congrats to the libxml
>
> Very interesting page, thanks for posting.
>
>
> > I think I could speed up parsing speed and a number of small 
> > operations by using alloca for some tasks and possibly interning 
> > some strings at the parser level. I will probably go over another
> > profiling work if I have some time after Relax NG and Schemas work.
>
>
> Resuscitating an old thread about string sharing, is it true that if all
> strings in libxml2 were shared then string comparison (which, I presume,
> is extremely common operation in libxslt and xpath), could be improved a
> lot (constant time operation versus non constant)?

if all (and i do mean absolutely all, except for a small amount of very 
trivial and limited cases) used ref counted strings, then you could replace
string compares with pointer compares. Which may or may not give an overall 
speedup as you now will have to manage the ref counts which means locking 
and / or atomic inc/dec ops.

the larger speedup may well come from string dups and similar now being 
lighter 8-)

> -- luca