View Single Post
Old 2005-07-20, 19:32   #5
cheesehead's Avatar
"Richard B. Woods"
Aug 2002
Wisconsin USA

22·3·641 Posts

Those messages are normal ever since version 22 iintroduced "soft FFT crossovers".

New features in Version 22.8 of prime95.exe
1) Soft FFT crossovers have been implemented. If you test an exponent that
	is within 0.2% of the old hard FFT crossover point, then 1000 test
	iterations are run to determine if the smaller or larger FFT is
	appropriate for the exponent.
So, for certain small ranges of exponents near the FFT crossover points (only 0.4% -- 0.2% on each side), Prime95 runs a test to determine whether the smaller (faster) FFT size yields an acceptable average roundoff error. If not, Prime 95 will use the larger (slower, but more accurate) FFT size. What you saw were the messages from that procedure. The ranges in which Prime95 does this include only 1/250 of all exponents, which is why you usually don't encounter this.
cheesehead is offline