what is behind the parallel computing toolbox?

Hello All,

I am wondering about the implementation behind the parallel computing toolbox. Is it using MPI (e.g. MPICH2, OpenMPI), Multiple Thread library(e.g. OpenMP, pthread), or a library developed by Mathworks for a multi-core PC? Thanks for clarifying this for me,

Hailiang

Answer 1

[Gaurav Sharma]

Hailiang

You can learn quite a bit from this journal paper (free download on Springer website):

MATLAB: A language for parallel computing

URL: http://dx.doi.org/10.1007/s10766-008-0082-5

[hailiang shen]

Thanks Gaurav,

This is a very comprehensive paper, discussing the framework of parallel computing, and the choices turning PCT the way it is.

On further going through the matlab document, I am confused by the difference between pmode and matlabpool. In pmode, I like the interactively display of the outputs of each lab, which facilitates the debugging greatly; however, I did not find a way for collective communication, such as functions as MPI_Scatter(), and MPI_Gather() (Is it corresponding to gcat function?). How can I scatter a big array in lab 1 to other labs?

Seems parfor and spmd are designed specifically for matlabpool. Is that correct? What is the difference between collective communication and distributed arrays?

Hailiang

[Ashish Uthama]

http://www.mathworks.com/products/parallel-computing/parallel/distarrays.html might help with your first question (and maybe your last).

yes, parfor and spmd work with a matlabpool.