Running multiple scripts on separate cores in parallel

I have four scripts - all running a MILP problem using intlinprog. I want to run the four scripts simulataneously to produce their results in the workspace. I cannot run it as parfor problems as there exisits a 'break' in the intlinprog loop.
I have tried :
parpool(4)
parfor K = 1 : 4
if K == 1; test1; end
if K == 2; test2; end
if K == 3; test3; end
if K == 4; test4; end
end
However, it informs me:
Error using test2
Transparency violation error.
See Workspace Transparency in MATLAB Statements.
Error in runCodeScript (line 2)
parfor K = 1 : 2

2 Kommentare

Jan
Jan am 11 Aug. 2022
Bearbeitet: Jan am 11 Aug. 2022
Scripts share the worksape with the caller.
Convert the scripts to functions instead.
Quick fix thank you!

Melden Sie sich an, um zu kommentieren.

 Akzeptierte Antwort

Matt J
Matt J am 11 Aug. 2022
fcn=@(i) run("test"+i);
for K=1:n
F(K)=parfeval(fcn,0,K);
end;
wait(F)

5 Kommentare

How does the first line work? A loop?
The first line is defining an anonymous function that takes one parameter. It does not do any looping itself.
The for K line starts the looping, and inside the loop asks parfeval to start executing fcn(K) on a worker.
The first line is just an anonymous function that runs the i-th script.
Thank you Walter - so I would need to create a vector of the fucntion names to implement as fcn so all 4 run simulanteously on different workers?
No,
parfeval(fcn,0,K)
means to ask a worker to execute function fcn on a worker, and on the worker it will be passed the parameter K . Looking at the function source:
fcn=@(i) run("test"+i);
that would take the numeric value passed in, convert it to character vector, append it to test and ask to run() the result. So when parfeval(fcn,0,3) was executed, run("test3") would be executed, thus asking the script test3 to execute.
You would only need a vector (cell array) of function handles if the names of the scripts to execute were not predictable.

Melden Sie sich an, um zu kommentieren.

Weitere Antworten (0)

Kategorien

Mehr zu Parallel Computing Toolbox finden Sie in Hilfe-Center und File Exchange

Produkte

Version

R2022a

Community Treasure Hunt

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

Start Hunting!

Translated by