Re: [Swarm-Support] coupling swarm with matlab

[Marcus G. Daniels]

Gu, Zhaotao wrote:
> Thus the boundary conditions have to be regenerated
> throughout the whole simulation many times, and exchange input/output
> with MAS frequently. 
>   
If it is gigabytes of I/O a minute, then you probably should find a way 
to create a C function to call to get the boundary conditions. 
One idea would be to use this product 
(http://www.mathworks.com/products/compiler/) to compile your Matlab 
into object modules such that Swarm could call them from some `step' method.

Or, you might try the new xml-rpc module written this summer by Gennady 
Telegin.  Here the idea would be have Matlab act as a client or server 
to a Swarm simulation acting as a server or a client.

http://www.mathworks.com/access/helpdesk/help/techdoc/index.html?/access/helpdesk/help/techdoc/matlab_external/f6010.html&http://www.google.com/search?hl=en&q=matlab+web+services&btnG=Search

Another idea would be to represent the data in HDF5 and have one write 
to the HDF5 file and another read from it.  HDF5 is designed for this 
sort of thing.  And Matlab has what appears to be a primitive HDF5 interface

http://www.mathworks.com/access/helpdesk/help/techdoc/index.html?/access/helpdesk/help/techdoc/ref/hdf5.html&http://www.google.com/search?hl=en&q=matlab+hdf5&btnG=Google+Search

The main issue here would be understanding what Swarm writes and expects 
for object serialization and duplicating that using the primitive HDF5 
calls it has.

On the other hand, if it is megabytes a minute or less, then you might 
just read and write to ad-hoc text files.

Marcus