In XSLTMark testing there are two test (of 40 aprox.) consuming about 90%
of overall time for the 0.44. Tom has added a match predicate optimization,
which makes the 0.50 about 80 times faster for these two particular tests.
It should change XSLTMark results significantly.
We asked XSLTMark organizers to update results for the 0.50 to get a kind
of "official data", but it hasn't been done yet. I understand they can't
rerun the benchmark for each new version of each processor again.
If you don't use predicates frequently in your transforms, do not expect
any performance improvement.
Regards,
Petr
David Hedbor wrote:
> I remember a while back that there were some performance testing done
> on Sablotron 0.44. Also I seem to remember that you said you'd fixed
> some of the really slow parts for 0.50. Is this correct? Have the
> benchmark been rerun? I'm just wondering out of curiosity, not because
> I'm having any particular performance issue.
>
> --
> [ Below is a random fortune, which is unrelated to the above message. ]
> A handful of patience is worth more than a bushel of brains.
--
Petr Cimprich
Ginger Alliance Ltd.
www.gingerall.com
