XML Editor
Sign up for a WebBoard account Sign Up Keyword Search Search More Options... Options
Chat Rooms Chat Help Help News News Log in to WebBoard Log in Not Logged in
Show tree view Topic
Topic Page 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Go to previous topicPrev TopicGo to next topicNext Topic
Postnext
Dhruva NagallaSubject: Execution Time Variation for node
Author: Dhruva Nagalla
Date: 07 Feb 2006 10:54 AM
Hi,

I am trying to evaluate the Stylus Studio 2006 on Winxp servicepack2 and j2sdk1.4.2_09
The odd behavior I noticed is when the profiler is ran on same set of xsl and xml files its giving different execution times each time for a particular node.
Any possible reason for this kind of behavior?
We are really in a hurry to buy a XSLT profiling product.

Hope you get back to us soon.

Thanks
Dhruva

Postnext
(Deleted User) Subject: Execution Time Variation for node
Author: (Deleted User)
Date: 07 Feb 2006 11:35 AM
Hi, Dhruva. A number of factors can contribute to time differences from one processing run to the next. Java garbage collection, for example, isn't a scheduled task -- Java does this when it thinks it's appropriate, which can be in the middle of XSLT processing. Also, initial runs will typically take longer than subsequent runs, so consider looking at the times over a number of runs.

Can you tell us a few things to help us better understand your issue?
- What engine are you using to process the XSLT?
- How much does the total time vary from one run to the next?
- How many runs did you use to gather your benchmark data?

David Foster
Stylus Studio Team

Postnext
Dhruva NagallaSubject: Execution Time Variation for node
Author: Dhruva Nagalla
Date: 07 Feb 2006 12:04 PM
I am using Saxon 8.6.1
The difference is around .03 to .07 secs for node (which takes an average of .25 secs)
Regarding your other question, what is the best count of runs we should use to gather data?
I basically checked for an average of two runs

Postnext
(Deleted User) Subject: Execution Time Variation for node
Author: (Deleted User)
Date: 07 Feb 2006 12:36 PM
Well, to get a meaningful sample, I'd consider looping the transformation and running it 100 or so times. At your run rate, this shouldn't be too time-consuming and might provide you with more meaningful numbers for comparison.

As I mentioned previously, the first run can often take longer than subsequent runs (allowing for the engine startup time and initialization), so coupling the statistics from that run with those from just one other isn't going to be especially telling.

David Foster
Stylus Studio Team

Postnext
Dhruva NagallaSubject: Execution Time Variation for node
Author: Dhruva Nagalla
Date: 08 Feb 2006 12:41 PM
I was wondering how the time column(call tree)for a node is showing microseconds, because JDK 1.4 only allows timings by millisecond.
Is that something manipulated by the profiler.

Posttop
(Deleted User) Subject: Execution Time Variation for node
Author: (Deleted User)
Date: 08 Feb 2006 12:55 PM
Hi, Dhruva. Timing is managed by C++ code inside Stylus Studio, which permits timing to the microsecond.

David Foster
Stylus Studio Team

 
Topic Page 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Go to previous topicPrev TopicGo to next topicNext Topic
Download A Free Trial of Stylus Studio 6 XML Professional Edition Today! Powered by Stylus Studio, the world's leading XML IDE for XML, XSLT, XQuery, XML Schema, DTD, XPath, WSDL, XHTML, SQL/XML, and XML Mapping!  
go

Log In Options

Site Map | Privacy Policy | Terms of Use | Trademarks
Stylus Scoop XML Newsletter:
W3C Member
Stylus Studio® and DataDirect XQuery ™are from DataDirect Technologies, is a registered trademark of Progress Software Corporation, in the U.S. and other countries. © 2004-2016 All Rights Reserved.