Ok, read that, thank you.

So getting everything in one shared partition and running distmake
over many machines is one method.  I have some questions about this:

Is there some part of the process (i.e. one of the commands in the
makefile) that require to have the entire image in memory?  And is
that on every machine?  How can I estimate how much memory each
machine would need on each machine?

Are there parts of the process that will only run on one machine that
will block it from using multiple machines?  For example the emblend?

There is also a parallel make where you start make with a --jobs flag
and it does something similar to distmake except locally.  Let's say
we used that but instead of running the commands like nona  -z LZW -r
ldr -m TIFF_m -o "IMG_5290-IMG_5440" -i 127 hug63C.tmp, it could
prefix it with something, let's call it rrun, that would package up
the necessary files off to another machine, run them, and return the
result.  This would be more grid-like.  It might enable us to create
some sort of pano server to process large panos on lots of machines
without actually needing to have shared disk.

One of the big questions in my mind though is if this could even work
because at some point, one of these processes on such another machine
might need access to everything, or it might need somehow to merge
it's result into a larger file.

Has anyone ever built a script to reserve some number of amazon ec2
instances, fire off a job, let it run, and return the result and shut
down the instances?  Such a run might only cost a few dollars (I
hope!) and run very very fast.

And by the way, this is the sort of thing one might do to stitch say
thousands of images, something that a single machine could take weeks
to do.

Michael

On Jan 12, 11:05 pm, Bruno Postle <[email protected]> wrote:
> On Wed 12-Jan-2011 at 13:28 -0800, michael.grant wrote:
>
>
>
> > Bruno, when you talk about a makefile model, where can I see that?  
> > I see, on windows, there's a make.exe, and I'm familiar with make.  
> > Where/how are these makefiles generated that make is run on?
>
> There is some documentation 
> here:http://wiki.panotools.org/Panorama_scripting_in_a_nutshell#Makefile_s...
>
> Basically Hugin constructs a list of all the temporary files and the
> rules to assemble them, then writes it all to a Makefile.  The
> stitching process is then managed by gnu make, you can
> close Hugin, or start a new project during stitching, or run
> stitching entirely on the command-line later or on another machine.
>
> > Would the other machines need access to all the images in the
> > pano?  Could only the overlapping parts of the images be sent to
> > the other server for stitching?
>
> The simplest way to do it is to use something like 'distmake', this
> would require that all the photos were on a shared filesystem
>
> --
> Bruno

-- 
You received this message because you are subscribed to the Google Groups 
"Hugin and other free panoramic software" group.
A list of frequently asked questions is available at: 
http://wiki.panotools.org/Hugin_FAQ
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/hugin-ptx

Reply via email to