You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@pdfbox.apache.org by Stefan Sauerer <St...@cas.de> on 2019/09/03 08:30:03 UTC

Hardware Acceleration

Dear Sir or Madam,

my name is Stefan and I'm working at CAS Software AG, a company in Karlsruhe. Currently, I'm looking for a topic for my master thesis.
As we are using your library PDFBox for document to image conversion, I had an idea for a topic. But first, I would like to ask you,
if there has been some investigation, or developement by your team before, regarding this topic.

The basic idea is, to increase performance of the document conversion by using hardware acceleration (GPU).
If I'm right, there is no computation done by the GPU at the moment. Which means, that everything during document conversion
gets processed by the CPU. My assumption is, that the GPU could be used for transforming elements, such as text/glyphs, images, etc. to their final
size and position. I assume, that this increases the performance.

The question is, if there has been any investigation with respect to hardware acceleration.

I very much look forward to hearing from you.

Best regards,

Stefan Sauerer

Re: Hardware Acceleration

Posted by Tilman Hausherr <TH...@t-online.de>.
Hello Stefan,

We didn't investigate that topic; all we know is that there's a rumor 
that hardware acceleration is used by the JVM for some operations when 
run on a physical system (not a virtual machine).

Most of our slow rendering files have poor performance in 
ColorConvertOp.filter(). This is in PDColorSpace.toRGBImageAWT().

Try the file here with a profiler:
https://issues.apache.org/jira/browse/PDFBOX-4392

Tilman


Am 03.09.2019 um 10:30 schrieb Stefan Sauerer:
> Dear Sir or Madam,
>
> my name is Stefan and I'm working at CAS Software AG, a company in Karlsruhe. Currently, I'm looking for a topic for my master thesis.
> As we are using your library PDFBox for document to image conversion, I had an idea for a topic. But first, I would like to ask you,
> if there has been some investigation, or developement by your team before, regarding this topic.
>
> The basic idea is, to increase performance of the document conversion by using hardware acceleration (GPU).
> If I'm right, there is no computation done by the GPU at the moment. Which means, that everything during document conversion
> gets processed by the CPU. My assumption is, that the GPU could be used for transforming elements, such as text/glyphs, images, etc. to their final
> size and position. I assume, that this increases the performance.
>
> The question is, if there has been any investigation with respect to hardware acceleration.
>
> I very much look forward to hearing from you.
>
> Best regards,
>
> Stefan Sauerer
>


Re: Hardware Acceleration

Posted by Tilman Hausherr <TH...@t-online.de>.
Some more ideas:
- I have about 30 files that are slow or were slow (but many still are), 
if you like to investigate for "opportunities"
- the one I mentioned yesterday renders in better speed with PDF.js. 
I've never heard that they use the GPU, usually they are faster than us
- video about PDF rendering in general by the PDF-Tools boss: 
https://www.youtube.com/watch?v=TiqDqd-1pwU&t   (very scary!)
- video about PDFBox https://www.youtube.com/watch?v=Sf8BfGaRUkk    and 
https://www.youtube.com/watch?v=agzqehWt-ko
- you're not the first person who would write a thesis related to 
PDFBox, this guy wrote about text extraction:
https://ad.informatik.uni-freiburg.de/files/icecite


Tilman

Am 03.09.2019 um 10:30 schrieb Stefan Sauerer:
> Dear Sir or Madam,
>
> my name is Stefan and I'm working at CAS Software AG, a company in Karlsruhe. Currently, I'm looking for a topic for my master thesis.
> As we are using your library PDFBox for document to image conversion, I had an idea for a topic. But first, I would like to ask you,
> if there has been some investigation, or developement by your team before, regarding this topic.
>
> The basic idea is, to increase performance of the document conversion by using hardware acceleration (GPU).
> If I'm right, there is no computation done by the GPU at the moment. Which means, that everything during document conversion
> gets processed by the CPU. My assumption is, that the GPU could be used for transforming elements, such as text/glyphs, images, etc. to their final
> size and position. I assume, that this increases the performance.
>
> The question is, if there has been any investigation with respect to hardware acceleration.
>
> I very much look forward to hearing from you.
>
> Best regards,
>
> Stefan Sauerer
>


---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@pdfbox.apache.org
For additional commands, e-mail: users-help@pdfbox.apache.org