Yes, on a single core machine, running multiple instances will just make
every one run more slowly.  That's try for any compute-intensive task.

0.3 should benefit from running multiple instances in parallel since each
instance only uses a single core.

0.4 is parallelized internally for many operations, so it uses all cores and
you don't benefit as much from running multiple instances.

Tom

On Mon, May 11, 2009 at 18:05, ocr_n00b <[email protected]> wrote:

>
> Thanks for answering Tom. I think I was on a single core machine. I am
> gonna try it on dual core machine today and let you know the results.
> I am using 0.3.
> So it should parallelize when running multiple instances on a dual
> core box.
>
> Thanks
>
> On May 9, 5:41 am, Thomas Breuel <[email protected]> wrote:
> > Do you have a multicore machine?  Is it set up correctly?  How are the
> CPUs
> > utilized (have a look with "htop").
> >
> > OCRopus 0.3 is single threaded internally.  If you run multiple instances
> on
> > a multicore machine, it parallelizes.
> >
> > OCRopus 0.4 is multicore, which means that it's much faster if you run
> one
> > instance and you won't see any overall speedup if you run multiple
> > instances.
> >
> > Tom
> >
> > On Sat, May 9, 2009 at 00:05, ocr_n00b <[email protected]> wrote:
> >
> > > Hi All;
> >
> > > I have been using "ocroscript rec-ltess source.png out.txt" to convert
> > > my images to text. Everything works as expected.
> >
> > > Next up I launch 5 different instances of ocroscript rec-ltess on the
> > > same server via 5 different threads in my c++ application. I just
> > > launch a shell script which further launches the ocroscript command.
> >
> > > Now, suppose one image takes 30 seconds to be OCR'd, I expected that
> > > the 5 different copies of the same image would take much faster if
> > > processed simultaneously, as compared to serially processing each of
> > > them.
> >
> > > What I am seeing is that in case of 5 different instances of
> > > ocroscript, it takes 5*30=150 seconds for EACH image to be OCR'd.
> > > Which is same as serially processing each image.
> >
> > > Is ocroscript meant to be used in multi threaded setup? It looks like
> > > it is sharing some resources internally and hence not giving good
> > > preformance in case of multi threaded setup.
> >
> > > Thanks!
> >
>

--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups 
"ocropus" group.
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to 
[email protected]
For more options, visit this group at 
http://groups.google.com/group/ocropus?hl=en
-~----------~----~----~----~------~----~------~--~---

Reply via email to