> On Mon, Oct 30, 2006 at 11:06:51AM +0100, Steinar H. Gunderson wrote: > > I have no idea why you think these ranges are the only true ones and drop > > automatically from nowhere. Prime95 ran with completely different ranges, > > “completely” is, of course, an overstatement -- but they were different, even > for non-SSE2 code.
Assuming they were calculated according to a fixed probability of error at the boundary. I expect different versions of the code were responsible for this. > The point was: They're not clear-cut and obvious, and if you step a bit > outside them, you'll still survive. If you read my last post, you will "have an idea why I think these ranges are clear cut". The probability of error varies sharply with exponent size. If you take an exponent near the bottom of a range and the probability of error, using the FFT size below, is less than ~10% you will do better than survive by using the smaller FFT (as I've pointed out before). David _________________________________________________________________ Be one of the first to try Windows Live Mail. http://ideas.live.com/programpage.aspx?versionId=5d21c51a-b161-4314-9b0e-4911fb2b2e6d _______________________________________________ Prime mailing list [email protected] http://hogranch.com/mailman/listinfo/prime
