Quote:
Originally Posted by henryzz
Have just submitted 177 new gaps of which 23 are new gap lengths(assuming I counted correctly). This might take a while for the queue to clear.
How deep does Danaj's surround_primes function sieve for gaps of length 150000? Also is there a way of determining what % of time is spent sieving vs prp testing by this function? I am wondering where the crossover point is for switching to testing using pfgw.

I believe surround_primes uses
Code:
def sieve_depth_pg(log2n):
log2log2n = int(math.ceil(math.log2(log2n)))
return log2n * (log2n >> 5) * int(log2log2n * 1.5) >> 1
I've personally found the crossover point for pfgw to be around ~8000 bits but I haven't profiled the overhead from Danaj's script.
I believe % time in sieve is very small (0.1%  1%).