As you have already seen from at least one of the responses to your OP, not revealing (in code or at least narrative form) the unsuccessful approaches you have already attempted to a problem is likely to result in frustration all around.

There are many monks lurking about the place who are vastly better qualified than I to respond to your OPed question were you to supply more basic info on what you want to achieve and what you have tried. I can say that the overhead of each Perl array element is many bytes (16? 32? More?) of RAM, so what is at first glance a 3 Gig array is, in reality, much larger. However, tools like PDL (see also PDL::FAQ) are designed to operate on large, multi-dimensional character and integer arrays of this kind as 'raw' data. Of course, if each chromosome of the genome can be represented as a string (resulting in an array of, e.g., 46 elements), then we are back again to something like 3 Gig. As always, more info will be helpful to all.