Prime gaps

[Ilya Zakharevich <ilya@math.ohio-state.edu>]

The current limit of primes 436e6 is related to a prime gap >255
appearing near this number.  When the gap >511 appears first?

Looking on the graph in "The first 50M primes" I see that it close to
1e12, but where exactly?

Is it possible to have a not backward-compatible primediff which will
keep gaps/2 ?  Then the theoretic limit will move to 4e9 on 32-bit
machines, and this may be lifted if initprimes() takes a double as a
limit... 

Ilya