f1_seek = f1.seek did not change the performance at all. As it turns out each call is only 646 nanoseconds slower than 'C'. However, that is still 80% of the time to perform a file seek, which I would think is a relatively slow operation compared to just making a system call.