xsl-list
[Top] [All Lists]

[xsl] Transformation time reporting problem with Saxon (Was: Re: [xsl] A new Sudoku xslt implementation (Was: Re: [xsl] Sudoku - A solution in XSLT 2))

2006-03-13 15:11:05
The mystery of Andrew and me getting conflicting times for the running
of his Sudoku xslt implementation has been solved.

Andrew's stylesheet included a large number of xsl:message instructions.

The time for these was obviously added to the total reported time in
my tests, performed from the XSelerator, but was not added in his
tests, performed from the command-line. When Andrew sent me a version
with the xsl:message instructions suppressed I got time results
consistent with his.

So, the problem is that when run from the command line Saxon does not
include the time for xsl:message instructions in the time it reports
when the -t option is specified. However, if Saxon is invoked by
another application (the XSelerator in my case), although the same
command-line string is used for the invocation, the time reported by
Saxon includes the time for the xsl:message instructions.

The problem is not in any one of these two ways of reporting time. The
problem is that there are two ways. What is needed is only a single
way of time reporting.


Let me use this message to confirm that Andrew's Sudoku xslt
implementation has been superior with most of the board configurations
being tested and to congratulate him for speeding it up (on these
configurations) almost 20 times just in a couple of days. For anyone
interested in this area it would be a challenge to produce a
significantly more efficient implementation.


--
Cheers,
Dimitre Novatchev
---------------------------------------
A writer is a person for whom writing is more difficult than it is for
other people.



On 3/13/06, andrew welch <andrew(_dot_)j(_dot_)welch(_at_)gmail(_dot_)com> 
wrote:
On 3/12/06, Dimitre Novatchev <dnovatchev(_at_)gmail(_dot_)com> wrote:
Yes, your new version runs now like this:

    3687ms, 25.9MB RAM

mine is still the same :o):

    6785ms, 10MB RAM


So, the cause of the incorrect timing was that when Saxon is run not
from the command line (although the XSelerator runs it with a
command-line invocation???) it reports much bigger time -- takes into
account the xsl:message times, while when run from the DOS prompt it
ignores any xsl:message times


Probably good to report this as a strange thing in the Saxon Help mailing 
list.

Yes - Mike might be able to shed some light on it.

Congratulations for speeding up the program more than 20 times!

Thanks!  My maching runs it at around 5.2 seconds... so you can see
how much faster your laptop is than my desktop :)


Cheers,
Dimitre.

P.S. I owe you an apology for reporting the incorrect timings to the
xsl-list. Would you like that I report the latest findings to the list
now ?

Yes that would be good, thanks.

cheers
andrew


--~------------------------------------------------------------------
XSL-List info and archive:  http://www.mulberrytech.com/xsl/xsl-list
To unsubscribe, go to: http://lists.mulberrytech.com/xsl-list/
or e-mail: <mailto:xsl-list-unsubscribe(_at_)lists(_dot_)mulberrytech(_dot_)com>
--~--

<Prev in Thread] Current Thread [Next in Thread>