iterlin and iterget

Linearize nested loops for smarter dispatch by parfor

MATLAB does not allow nested parfor loops, resulting in idle time for workers at loop boundaries. In the following example, no loop is a good choice for conversion to parfor:

for plate = 1 : 3
for row = 2 : 7
for col = 4 : 9
% time-consuming computations here %
end
end
end

A solution for having workloads from within nested loops be dispatched to workers without idling is to linearize, but that approach becomes cumbersome if ranges do not start from one. Now, iterlin and iterget provide a simple pattern:

rng = [1, 3; 2, 7; 4, 9];
parfor k = iterlin(rng)
[plate, row, col] = iterget(rng, k);
% time-consuming computations here %
end

With the above pattern, you reduce idle time for workers, and no longer need to think about which loop to convert to parfor.