[Pkg-exppsy-pynifti] [Nipy-devel] Image, design, usecases

[Souheil Inati]

Hi Matthew,

> git clone git://git.debian.org/pkg-exppsy/pynifti.git
> cd pynifti/
> git checkout -b souheil/brifti origin/mb/brifti
>

I can't believe that I just got suckered into installing git :-)
I'll look into the code and see if I can figure out what's going on.

>
> Yes, that's the thing that's waiting to happen...  If an example
> script comes to you, then send it - that would help. For now, have a
> look again at:
>
> https://cirl.berkeley.edu/mb312/pyniftidocs/design.html
>
> - I've tried to add something to describe your case at the end...

Looks good.  Thanks.

It's going to take me quite a while to understand how to do multi- 
processor or cluster computation in python.  I'm hoping to make a big  
push this summer.  I'll let you know if I come up with something useful.

For now though, I believe that the simplest thing would be to have  
something for memory mapped writing that is similar to the way that  
memory mapped reading works now.  Schematically maybe something like  
this (I'm thinking of stupidly parallel apps where all parameters are  
known ahead of time):
1) set up the header like in the empty file case
2) initialize the shared memory file with zeros
3) Then as long as whatever dispatches the computation pieces knows  
what it is doing, it can do random access writes to the array using  
the memory mapped numpy array.

Someone who has thought about memory mapping files and/or parallel  
computing in python might have something more intelligent to say about  
this.

Cheers,
Souheil
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.alioth.debian.org/pipermail/pkg-exppsy-pynifti/attachments/20090514/50a7f8ac/attachment.htm>