Parallel Computing Question

2 views (last 30 days)
Manoj
Manoj on 18 Feb 2012
Hi, I am a noob in matlab anf I am trying to parallelize loading of an array file.
So suppose the array size is X by Y. I would like to load this file to "n" matlab workers giving X/n by Y to 1st worker and X/n by Y to another and so on.... Does anyone know how can I achieve that since the file is too large and load of that file in a distributed way can save me loads of time.
I tried pmode but since it loads the entire data on each worker and as my data file is huge, it takes a lot of time. So i want to load only X/n by Y on each worker where n is the total number of workers.
So I am trying something like this for my application:
1. Start n worker nodes 2. load X/n by Y data on each worker node 3. Output = function (X/n, Y) 4. Final Output = Gather (Output) 5. Stop all worker nodes.
Do let me know if you can help me in any possible way.
Thanks

Answers (2)

the cyclist
the cyclist on 19 Feb 2012
I don't have experience with the parallel computing toolbox, but I speculate that the relatively new matfile() command, which allows loading of partial arrays, maybe be helpful to you.

Jill Reese
Jill Reese on 21 Feb 2012
You might take a look at the DLOAD functionality that is part of the Parallel Computing Toolbox. This function loads distributed arrays and composites from disk, but if you also supply the -scatter option, then DLOAD will also distribute non-distributed data where possible.

Categories

Find more on MATLAB Parallel Server in Help Center and File Exchange

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!