although the fact that 0.94 is faster than 0.95beta-jeremias-tiff
would make that ....weird.
Not really. It would be much weirder IMO if FOP (or any software for
that matter) were to offer increased flexibility/functionality/
features without any performance penalty whatsoever. If that were
the case, Vista should run on a 386 without any noticeable slow-down
when compared to, say, Windows 3.11... errmmm.
Oh...well let's leave windows out of this .... :)
I'd think it pretty 'normal' that the new image-handling framework
adds /some/ overhead. A factor of six would be on the high side,
though, but nothing points in the direction of the image-handling
framework alone. A lot has been changed and added between 0.94 and
0.95, so even though this may seem 'weird', I somehow expect 0.95 to
be slightly slower than 0.94.
I understand that...no such thing as a free lunch. But as you said, a
factor 6 (assuming what I found is correct) irrespective of where in
fop/xmlgraphics/batik the slowdown lies, is...a lot.
Also, since (I presume) you are timing singular runs, there is no
accounting for things like static initialization. This only occurs
for the very first run in a given runtime session. If the newer code
moved some parts to those areas, then this would only slow down that
first run, but could speed up/save memory in all subsequent runs in
the same VM.
Well, the numbers come from single runs yes. Otoh, I started out with
a stylesheet with more tiff's (which is how I originally ran into the
slowdown). There the performance penalty was far worse ...to the
extent that it makes the (web) application unusable with 0.95beta. To
be complete - that was before the 'jeremias-tiff' version.
I'll try to find some time to ...well...not sure what I can do. But
it's either ignore post 0.94 versions or fix something (and I want to
use post 0.94 versions for different reasons)
I will try to profile the code I guess and figure out where the
problem might have been introduced (which will be quite an adventure
having no real idea how the code is structured nor what was changed
when...I know svn does help there).
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
