[XSL-LIST Mailing List Archive Home] [By Thread] [By Date]

Re: [xsl] XSLT 2.0 in a web browser


Subject: Re: [xsl] XSLT 2.0 in a web browser
From: "Colin Adams" <colinpauladams@xxxxxxxxxxxxxx>
Date: Wed, 26 Mar 2008 08:09:10 +0000

On 26/03/2008, James Fuller <james.fuller.2007@xxxxxxxxx> wrote:

> just wondering;  how would someone who downloads firefox source then
>  compile .... u can't expect people to depend on a pre-compiled binary,
>  on the other hand I can't see mozilla follks wanting to add an Eiffel
>  compile step as a dependency. making apologies now for my Eiffel
>  ignorance ;)

I was thinking of a Firefox extension. People do use these - I for one
(and are therefore depending upon a pre-compiled binary).

>  and to make my question a bit more XSLT related .... do u have any
>  formal XSLT performance benchmarks for gestalt ?

Not formal. I think it would be a good idea for someone to set up a
suite of XSLT 2.0 performance benchmarks.

Informally, Gestalt 1.0.x has quadratic behaviour in the size of the
source documents (approx.).
I have spent the last 3 weekends profiling (I finally decided this was
more important than regular expression compatibility), getting nowhere
(I has assumed all along that somewhere I had a poor algorithm, but I
could find no trace of it), until Monday night, when I switched from
profiling the Eiffel code to profiling the generated C code, using
gprof. Immediately it was apparent that nearly all the execution time
was due to the garbage collector.
So I have now started to reduce the number of objects allocated. The
first effort halved the runtime, so that looks promising. It will take
several weeks of effort though, as I have to inspect all the code, and
there's a lot of it!


Current Thread
Keywords