sampling from a beta distribution

Hi All, I need to fit a model where the objective function calculation requires an irritative sampling from two arbitrary beta distributions (each time a different distribution). I got the thing to work but it is painfully slow (in terms of days and even weeks). I tried working with mex files, but it didn't really help. I also tried to create a matrix with 'handmade' samples (then searching the matrix instead of running betarnd), but the matrix got so big it didn't really make a difference.

Any thoughts?? I'm desperate here :)

Thank you, Nitzan.