This is the mail archive of the xsl-list@mulberrytech.com mailing list .


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

RE: Updated Benchmark Available


>
> The benchmark results look interesting, but what do they really
> demonstrate?
>
> At my opinion, with the longest test running not more than 0.6 seconds,
> these results compare the time which XSLT engines need to warm-up and to
> cleanup, rather than the time they need to perform their main job.
> Furthermore, the warm-up and the cleanup time depends on the run-time
> system of the underlying platform (JVM, C++ RTL etc.), not on the
> quality of the XSLT processor itself.
>

There is some truth in this for the Java stuff. Sebastien Sahuc has pointed
out that with just 10 runs, Hotspot is not getting time to find the
'hotspots'.
I have updated the tests to perform 100 runs for warm up and 100 for timing.
In practice this has helped the Java engines on Test1 but made little to no
difference on Test 2 & 3.

> From the practical point of view, the result of this benchmark could be
> formulated as follows:
>
> "... All XSLT processors (except those crashed) demonstrated a good
> performance ..." (0.6 sec is a good response time - isn't it?)
>

While its true that all the processors demonstrated good performance times
there are significant differences between them. My main interest is in
using XSLT in high-transaction web sites. In this environment a bad
choice of processor can cost you 8 times more transformation time. In other
environments you may not care at all about this in which case the
benchmark is irrelevant to you.

> I also could not understand why the issue whether the XSLT processor is
> using the "pre-parsed" style sheets is so important. Some processors do
> it, some don't, but the efficiency of this feature in fact depends on
> the processor architecture and on the way how this pre-parsing is
> integrated with the other processor components.
>

This really becomes a issue when you are performing many transformations
using the same style sheet. It may be possible to write a quick engine
that does not use any form of pre-parsed style sheet but my guess would
be that it would perform poorly on this particular benchmark.


> Kind regards
>
> Alexey Gokhberg
> Developer of Unicorn XSLT Processor
> http://www.unicorn-enterprises.com/uxsl_1_02_15.zip
>
>
>  XSL-List info and archive:  http://www.mulberrytech.com/xsl/xsl-list

Regards,
Kev.


_________________________________________________________
Do You Yahoo!?
Get your free @yahoo.com address at http://mail.yahoo.com


 XSL-List info and archive:  http://www.mulberrytech.com/xsl/xsl-list

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]