[ptx] x64 benchmarks

[pablo.dangelo at informatik.uni-ulm.de]

Zitat von Rich <rich at hq.vsaa.lv>:

> Hal V. Engel wrote:
> > On Sunday 15 January 2006 10:04 am, Marek Januszewski wrote:
> >> Marek Januszewski wrote:
> >>> Hello,
> ...
> > I think the real bench mark would be to setup a machine with more than 4
> gig
> > (say 8 or 16) and then compare stitching performance of a huge pano on the
> > order of say 100 or 150 images.   Under those conditions I am sure that the
> > 64 bit setup would blow the doors off of the 32 bit setup.
>
> cropped tiff output might help here, but i believe hugin would fail to
> produce any meaningful output with >100 images from a recent digital
> camera :)
>
> i had a pano of ~40 pics and without cropped tiff output hugin just kept
> hitting 2gb boundary for a single process on 32b platform (at lest
> that's how pablo explained this o me :) ).

It should work better with cropped tiff output. For full 360 deg panos nona
should now be able to process panoramas twice as big as without cropped tiff.

For panoramas with less than 360° deg, it should be possible to build really
large mosaics, since each remapped image will be not much larger than the
source image.

The next bottleneck will be enblend (which can handle the size, but it will
probably take a lot of time to render) and the 2 or 4 GB barrier of the tiff
files.

ciao
  Pablo

----------------------------------------------------------------
This message was sent using IMP, the Internet Messaging Program.