You are viewing a plain text version of this content. The canonical link for it is here.
Posted to fop-users@xmlgraphics.apache.org by Donald Mackinnon <do...@empowered-systems.com> on 2006/08/02 19:37:59 UTC
FOP 0.20.5 versus FOP 0.92 performance
Hi
I have generated large PDF reports on a Solaris 10 server running Java
1.5.0_01 for FOP release 0.20.5 and 0.92 Beta and findings are as
follows:-
1) FOP 0.92 Beta appears to be freeing up memory rapidly after .pdf
generation whereas FOP 0.20.5 does not.
2) FOP 0.92 Beta appears to use up significantly more memory in the
generation of the .pdf (circa 60% in the case of a big report)
3) The .pdf reports are taking significantly longer to create (between
50% and 100%) with the FOP 0.92.
4) The resultant .pdf file size created with FOP .92 Beta is
significantly larger than the FOP 0.20.5 as it has increased from
6.42MB to 12.5MB.
5) There are no forward-references in the XSLT file to improve
performance.
Are these results to be expected?
This message has been scanned for viruses by BlackSpider MailControl - www.blackspider.com
Re: FOP 0.20.5 versus FOP 0.92 performance
Posted by Jeremias Maerki <de...@jeremias-maerki.ch>.
Thanks for sharing these results with us. When I did some benchmarks
about 11 months ago the performance was slightly in favor of the new
code. In the meantime, new functionality has come in and it could be
that some of the changes may have had a negative impact on performance.
However, performance is always largely dependent on the kind of document
you're processing. So, it would be great if you could give some info
about the documents you've tested with (number and size of images, usage
of markers, page-number-citations etc. etc.).
In case your test documents contain a number of images, I'd be very
grateful if you could retest with the latest dev code (FOP Trunk). Fixes
in memory handling could have a larger impact on speed and memory
consumption. There's also a bug in 0.92beta concerning color profiles
which blew up the generated PDFs unnecessarily. That's the only way I
can explain the large increase of the PDF size. The PDFs generated by
the new versions are expected to be a little larger due to certain new
features but this should be in the area of a few KB max per document.
On 02.08.2006 19:37:59 Donald Mackinnon wrote:
> Hi
>
> I have generated large PDF reports on a Solaris 10 server running Java
> 1.5.0_01 for FOP release 0.20.5 and 0.92 Beta and findings are as
> follows:-
>
> 1) FOP 0.92 Beta appears to be freeing up memory rapidly after .pdf
> generation whereas FOP 0.20.5 does not.
>
> 2) FOP 0.92 Beta appears to use up significantly more memory in the
> generation of the .pdf (circa 60% in the case of a big report)
>
> 3) The .pdf reports are taking significantly longer to create (between
> 50% and 100%) with the FOP 0.92.
>
> 4) The resultant .pdf file size created with FOP .92 Beta is
> significantly larger than the FOP 0.20.5 as it has increased from
> 6.42MB to 12.5MB.
>
> 5) There are no forward-references in the XSLT file to improve
> performance.
>
> Are these results to be expected?
Jeremias Maerki
---------------------------------------------------------------------
To unsubscribe, e-mail: fop-users-unsubscribe@xmlgraphics.apache.org
For additional commands, e-mail: fop-users-help@xmlgraphics.apache.org
Re: FOP 0.20.5 versus FOP 0.92 performance
Posted by Andreas L Delmelle <a_...@pandora.be>.
On Aug 2, 2006, at 19:37, Donald Mackinnon wrote:
Hi,
> I have generated large PDF reports on a Solaris 10 server running
> Java 1.5.0_01 for FOP release 0.20.5 and 0.92 Beta and findings are
> as follows:
First of all: thanks for sharing your findings. Always interesting to
get feedback on performance/efficiency.
> 1) FOP 0.92 Beta appears to be freeing up memory rapidly
> after .pdf generation whereas FOP 0.20.5 does not.
Partly so, yes. Mainly because the layout-logic (and construction of
the area-tree) has been separated from constructing the tree of FOs.
Depending on whether you use relatively small-sized page-sequences,
this could indeed mean a drastic increase in efficiency there.
Mind that the FO tree is kept entirely in memory --if I esteemed
correctly-- until the Root is no longer referenced. For very large
input, but I mean HUGE --thousands of pages--, this can still
eventually lead to the heap being completely filled up with FOs, so
the layoutengine gets less-and-less space to work with... :(
> 2) FOP 0.92 Beta appears to use up significantly more memory in
> the generation of the .pdf (circa 60% in the case of a big report)
> 3) The .pdf reports are taking significantly longer to create
> (between 50% and 100%) with the FOP 0.92
Hmm. Real thorough profiling hasn't been performed so far. If you
know how to handle a profiler and know what to look out for, and if
you're volunteering... ;)
OTOH, increased functionality doesn't come for free --ultimately 0.92
does offer more features.
It would be interesting to make comparisons between certain types of
FO structure. For instance: how do tables/lists/images etc. compare
from one version to the other?
Cheers,
Andreas
---------------------------------------------------------------------
To unsubscribe, e-mail: fop-users-unsubscribe@xmlgraphics.apache.org
For additional commands, e-mail: fop-users-help@xmlgraphics.apache.org