RE: [docbook-apps] Re: Support for callout extensions in xsltproc

[Jeff Beal <jeff.beal@ansys.com>]

> though.  It seems
> > that Saxon is able to write files faster on Windows then 
> XSLTProc, so even
> > though xsltproc is "transforming" faster, Saxon is able to 
> get the file onto
> > the hard drive faster.  When I'm going to FO (where there's 
> only one file
> > write operation), xsltproc is considerably faster.
> 
>   Okay, do a Request For Enhancement on libxslt bugzilla about this.
> Maybe I can find why writing is slower than expected, I don't remember
> doing any performance analysis for chunking i.e. 
> exslt:document extension
> maybe there is something wrong.

I'll go ahead and file that.  If you need a large test-doc, I'll figure
something out.  Our doc is proprietary, so I can't just send it to you
as-is, but I can run it through an XSLT transform to obfuscate the text.
 
> > I should also point out that my doc set is abnormally huge. 
>  On my dual PIII
> > 1.4 GHz machine, either processor takes almost five hours 
> to do a complete
> > build.  We build 26 manuals simultaneously, which are about 
> 10,000 printed
> > pages and about 6,000 HTML files.  There may just be something about
> > scalability where Saxon wins out.
> 
>   Yes in that case the Java load + warmup of the JIT is lost in the
> time spent on processing. But for a 5hours processing my take is that
> you're likely to swap like hell on the box, measuring purely the speed
> of your I/O subsystem. I doubt the CPU load it 100% of the CPU (or
> 2x50% depending on the scheduling and affinity processing of your OS
> since you're wunning an SMP). 5 hours of a 1.4 GHZ CPU is way too much
> IMHO even for 6000 resulting files. I bet it's totally I/O bound, the
> swap processing competing with the write I/O of the chunking code.

It doesn't seem to be doing any processor scheduling with either xsltproc or
saxon, but either of them will get full usage of a single processor for the
duration of the build.  When I do two builds at once -- one for HTML Help,
one for Oracle Help (JavaHelp variant) -- I get full usage out of both
processors.

>   Definitely not "common usage" ... :-)

Absolutely not.  That's why I hadn't bothered complaining about performance
in the past.  (I'm still not complaining.  I build overnight.  It gives my
computer something to do while I'm gone.)

Jeff