MATLAB: Can I use “mpiprofile” on a “parfor” loop

parallelParallel Computing Toolboxpctperformanceprofile

I would like to profile my "parfor" loop. When I try using "mpiprofile" on it in MATLAB R2019b, I get an error:

Error using mpiprofile (line 162)
MPIPROFILE on must run in pmode or as part of a parallel job.
See doc for MPIPROFILE in the Parallel Computing Toolbox.

Best Answer

Support for using "mpiprofile" on "parfor" loops was added in MATLAB R2020a:

https://www.mathworks.com/help/releases/R2020a/parallel-computing/profiling-parallel-code.html

In prior releases, "mpiprofile" only works on communicating jobs:

https://www.mathworks.com/help/releases/R2019b/parallel-computing/profiling-parallel-code.html

"The parallel profiler works on communicating jobs, including inside pmode. It does not work on parfor-loops."

Related Solutions

MATLAB: How to take advantage of useParallel option for a remote job running on workers in a MDCS cluster

1. In order to use parallel pool with the cluster, please consider using the 'batch' function procedure to submit the jobs into cluster:

https://www.mathworks.com/help/distcomp/batch.html

When using the 'batch' command, it has the option to specify the number of workers to be used in addition to the job itself. Please refer to the example below:

c=parcluster('profile');
j = batch(c,@scriptUseParallel, 'Pool', 2);
wait(j);
diary(j);
Y2=fetchOutputs(j);

2. In the cluster setup, please consider to configure each machine to have the same number of workers as the number of cores.

Therefore each core can be started with a MATLAB process and being utilized by the batch job.

MATLAB: How to save the mpiprofile report

One way I figured to do something similar is the following.

I am assuming you are using "pmode" because as far as my understanding goes, you need to run on "pmode" to run mpiprofile.

So this is how it goes.

Saving

1. Save the mpiprofile info inside each worker(or it is often called lab from my understanding).

mpiprofile mpistruct;

2. Then you transfer to the client using the following command

pmode lab2client mpistruct;

3. Then save the mpistruct into a .mat file

save('mpistruct');

When you are trying to view the mpiprofile result:

1. Load the mpi file

load('mpistruct.mat');

2. Run mpiprofile viewer

mpiprofile('viewer', mpistruct);

This should pop up the browser.

Note. Above code was tested with 2015b.

Best Answer

Related Solutions

MATLAB: How to take advantage of useParallel option for a remote job running on workers in a MDCS cluster

MATLAB: How to save the mpiprofile report

Related Question