View Single Post
Old 2007-05-16, 16:16   #5
(loop (#_fork))
fivemack's Avatar
Feb 2006
Cambridge, England

645410 Posts

That's a case which runs into Brent's rule, whose intent is that using ECM for longer than the expected runtime of a guaranteed-success method is a waste of computrons not to be rewarded.

[ie, if you have a C120=P60*P60, you factor it by GNFS and it takes 80 hours; you don't bother running the thousands of hours of ECM needed to get a sixty-digit factor. If you have a C180 which is susceptible to SNFS, it's daft to run ECM for longer than the 80 or so hours that the SNFS run would take]
fivemack is offline   Reply With Quote