Parallel Programming

orbitz <>
Sat Aug 13 21:18:33 CEST 2005


Interesting.  I am taking an independent study on Parallel Programming 
shortly and I would like to use erlang to create some sort of useful 
framework.  I think I could use erlang to split up the data and then 
reconstruct it and handle communication, like you just described.


On Aug 13, 2005, at 2:50 PM, Serge Aleynikov wrote:

> We've done exactily this approach in one of the projects where a large 
> computational task was multiplexed between several nodes running C 
> ports doing mathematical optimization.  Erlang was used only for 
> splitting work load, distribution and gathering results.
>
> I am sure other libraries (like MPI) could have been used, but it took 
> very little effort to do parallelization in Erlang/OTP, and C-based 
> code was only focused on solving the actual optimization problem.  
> This approach also helped to use different resources concurrently for 
> solving mathematical, distributed and integrational problems, with 
> very little overlap of responsibilities, which allowed to implement 
> the solution impressively quickly.
>
> Regards,
>
> Serge
>
>  wrote:
>> Obviously Erlang is not made to do computationally intensive work 
>> such as computing primes or what not.  However I was thinking that 
>> Erlang obviously makes an excellent framework for doing that kind of 
>> work.
>> I was thinking one could use erlang to communicate between nodes and 
>> then use a port to a fortran program or what not.
>> Obviously I doubt I'm the first person to think of this, so has 
>> anyone else attempted this or have any suggestions as it if it is a 
>> good idea or not?
>> Thanks
>




More information about the erlang-questions mailing list