Well the second benchmark was performed with the 0.40 release, which was
still alpha, so I won't bother to respond to that one.
The XSLTMark benchmark is another matter. I've already responded to this
both on the Xalan list and the Mulberry XSL list. Basically, this was a
flawed benchmark. Their drivers for Xalan-J, Xalan-C, and Saxon included
the overhead of parsing the XML documents in the timings. Since they were
parsing documents up to 100 times, you can imagine what this would do to
the results.
In addition, their Xalan-C driver was hopelessly flawed, including a really
outrageous memory leak.
You should also know that Datamark is an XML vendor, so you should consider
that when evaluating their "benchmark" suite.
In the end, you should do your own testing for performance, since results
will vary depending on the input document and the stylesheet.
Dave
"Weber, Heiko"
<heiko.weber@individua To: <[EMAIL PROTECTED]>
l-web.net> cc: (bcc: David N
Bertoni/CAM/Lotus)
Subject: RE: Why is Xalan for
C++ slower than the Java version?
06/19/2001 02:26 PM
Please respond to
xalan-dev
Hello Dave,
I found the following two benchmarks:
http://www.xml.com/pub/a/2001/03/28/xsltmark/results.html
http://www.tfi-technology.com/xml/xslbench.html
Both claim the Java version to be faster than the C++ version,
but both used older C++ versions than Java versions. How did
the C++ version perform in your tests? I had expected the
C++ version to be much faster than the Java version, so I was
really surprised by these results.
thanks,
Heiko
-----Original Message-----
From: [EMAIL PROTECTED]
Sent: Tue 19.06.2001 17:11
To: [EMAIL PROTECTED]
Cc:
Subject: Re: Why is Xalan for C++ slower than the Java version?
To which benchmarks are you referring? Our testing shows Xalan-C++ to
be
generally faster than the Java version.
Dave
(See attached file: winmail.dat)
winmail.dat
