At 06:42 PM 9/12/2002 +0000, Brian J. Beesley wrote: >I haven't checked the source from the latest version but the TF limits should >surely be linked in some way to the LL/DC FFT run length crossovers. Many of >these _have_ been lowered. Slightly, and especially for P4 systems.
I have not changed the trial factoring limits in version 22. Actually, they haven't changed in years. >I would have thought that, since (ignoring runs with errors - which is a >reasonable first approximation) factoring before DC saves only one LL test, >whereas factoring before first LL test saves two. So trial factoring (TF) >depth for DC assignments should be one bit less than for LL assignments - This is actually a very reasonable and easy to implement idea. It doesn't fix all our problems, but it makes things a little bit better. Optimal trial factoring depth vs. P-1 vs. LL varies from machine to machine. To further complicate matters, the factoring, first LL test and second LL test is very likely to be done on different hardware. In essence, I've just "given up" on analyzing and implementing any improved solution. The good news is that implementing the absolutely optimal solution - whatever that is - would improve GIMPS throughput a very, very tiny amount. The original point of this thread, that P-1 should be run before running the last one or two bits of trial factoring is true. Unfortunately, it cannot be implemented without server changes. If we ever make P-1 a separate work type (also desirable), that would be the time to implement this. _________________________________________________________________________ Unsubscribe & list info -- http://www.ndatech.com/mersenne/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
