Hello,
I am trying to parallelize a program. I am experiencing issues with a parallel for and nested functions. My main function contains a big parfor loop, and several nested functions. These nested function access (both read and write) the variables declared in the main function. Without parallelism, everything works fine: my nested function do their job and update the variables in my main function (as it is supposed to be). This is true whether I call the nested function directly or using a handle. However, when I switch to a parfor (and thus use handles, as it should be), and make so that no error appear, the nested function no longer updates the variables from the main function. Said otherwise, MATLAB does not detect any error, but my program doesn't really do anything!
Below is a representative example:
function garbage accumulator = []; handleFoo = @foo; for i=1:10 aux = i*5; handleFoo(aux); end accumulator function foo(n) accumulator(end+1) = n; end end
Running this program as-is will result "accumulator" equal to 1:10. But simply changing the for loop to a parfor will resut in an empty array.
Is this the expected behavior of MATLAB? Is there away to work around this? I could not find help on this specific question anywhere.
Thank you Antoine
Best Answer