Hi Karl, On Linux you can get the number of cores with

my $no_of_cores = system ('/usr/bin/nproc');
[download]

I also tested your approach to split the load onto each core. It was slower than McA suggegestion and the serial processing of files but faster than the original parallel processing.

real 0m0.531s, user 0m0.449s, sys 0m0.208s

But thanks for sharing your wisdom. I learnt a lot about making use of a multi core cpu and splitting workloads.