Page 1 of 1

HotSpot Variation for each run

Posted: Tue Feb 07, 2006 5:51 pm
by dhruva

I am trying to evaluate the oxygen7.0 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 hotspots each time.
Any possible reason for this kind of behavior


Posted: Tue Feb 07, 2006 7:19 pm
by stefan
It is not uncommon to experience different hotspots results if you are running a fast transformation (something simple like personal.xml / personal.xsl from oXygen sample directory). These are the kind of transformations that do not need profiling ;-)

If the execution time is insignificant, the accuracy of the time reading and the related computations performed to determine a hotspot affects in a greater proportion the overall result. The operating system could also use some memory cache for the next runs and even the CPU load could differ between two runs.

Please check the below video presentation for a more suitable case. ... iling.html

If you use jdk 1.5 instead of jdk 1.4 you will be able to set your time scale to microseconds instead of milliseconds and the time readings should be more accurate. Please also check the other options from: ... s-profiler