Re: [xml] proposed patch to hash function

On Mon, 17 Jun 2002, Daniel Veillard wrote:

On Sun, Jun 16, 2002 at 11:02:03PM +0100, Sander Vesik wrote:

This is a proposed patch for the hash function in libxml2 - according to
my benchmarking, it should give ~ 1 - 2 % of speed advantage. I hoped to
also finish teh second stage of the patch, but that is not quite ready for
public consumption yet.

  Well, it doesn't seems to change much performance wise for me (I did
a very limited test), and since it should improve hashing properties
compared to my minimal checksum, I applied it and commited :-)

The benchmarking I used was:
        a) xmllint --noout --valid --timing --repeat $file
        b) xsltproc --nnout --timing --repeat $style $file

Where the file was the gedit xml manual and the stylesheet was the
gnome-customisations.xsl (which in turn includes & applies the docbook xsl
stylesheets). The metric was the average of 10 consecutive runs.
Unfortunately, the impact of hashing on stylesheet application is less
that what i expected.

If there are other realtively easy to run benchmarks to try (no, XSLTmark
is not in this category) then i can include the data for these aswell in
the future. 



Daniel Veillard      | Red Hat Network
veillard redhat com  | libxml GNOME XML XSLT toolkit | Rpmfind RPM search engine


        you'll rescue me right?
        in the exact same way that they never did
        i'll be happy right?
        when your healing powers kick in

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]