Bug 67105 - pdftoppm taking hours to raster a single page PDF
Summary: pdftoppm taking hours to raster a single page PDF
Status: RESOLVED FIXED
Alias: None
Product: poppler
Classification: Unclassified
Component: utils (show other bugs)
Version: unspecified
Hardware: x86-64 (AMD64) Linux (All)
: medium normal
Assignee: poppler-bugs
QA Contact:
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2013-07-19 22:12 UTC by Cory
Modified: 2013-08-20 17:51 UTC (History)
0 users

See Also:
i915 platform:
i915 features:


Attachments
don't copy not needed bitmap (655 bytes, patch)
2013-08-19 10:17 UTC, Thomas Freitag
Details | Splinter Review

Description Cory 2013-07-19 22:12:24 UTC
I have a number of PDFs which are taking between 10 min to 2 hours to raster with pdftoppm at a reasonable 150 DPI resolution. We process a lot of files like these through our system and typical times are in the 1-10 sec range. 

Unfortunately, our PDFs are sensitive, so please e-mail me directly for a Dropbox link to some example PDFs. 

Thanks a lot for any help that you can provide!

Sincerely,
Cory
Comment 1 Albert Astals Cid 2013-07-19 22:27:56 UTC
Which poppler version are you using?

You can send them to me.
Comment 2 Albert Astals Cid 2013-08-16 23:48:42 UTC
FWIW I got the files, and yes they are slow :D
Comment 3 Thomas Freitag 2013-08-19 10:17:01 UTC
Created attachment 84241 [details] [review]
don't copy not needed bitmap

The problem here was that we copied 37.059 times the complete hugh bitmap (37.059 isolated transparency groups !!!) and then never used it because the transparency groups were isolated :-)
BTW, there is enough room for performance increasements even for non isolated transparency groups: 
1. The complete bitmap is copied but only the alpha channel is used and
2. even from the alpha channel only a part is needed!
(s. splash->setInNonIsolatedGroup())
Comment 4 Albert Astals Cid 2013-08-20 17:51:46 UTC
Pushed


Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.