Hi Ernst, At 02:11 PM 10/25/2001 -0400, [EMAIL PROTECTED] wrote: >Last night, one of the 2 mprime jobs I run on my Linux >PC at work died. It apprently died due to an illegal >sumout error. Now these are quite common, an normally >appear in my results.txt file in the form > >Iteration: 1019208/10199069, ERROR: ILLEGAL SUMOUT >Possible hardware failure, consult the readme file. >Continuing from last save file. > >The one last night, OTOH, was of the form > >ERROR: ILLEGAL SUMOUT >Possible hardware failure, consult the readme file.
This is weird. FYI: ILLEGAL SUMOUT means the checksum value was an invalid floating point value, such as infinity, NaN, etc. This can be caused by a cpu problem, memory problem, and (at least in Windows 9x) an OS driver not saving the FPU state properly. I would guess Linux would only have the first two causes. >i.e. no iteration number, and no "Continuing from last >save file." message following it - it just died at this >point. Maybe it was doing P-1 factoring??? Have you tried running mprime -d v21 mprime should have finished off the LL test before doing any other work - so I'm a bit baffled here. > There was also a file write error about 2 hours >before the crash, this turned out to be due to a full >user partition on my hard drive (which I've since fixed) >and I don't know if it has anything to do with the >sumout errors (it seemes they should not be related.) I agree they should not be related. >Has one of my CPUs gone flaky on me? My guess is yes (or memory is bad). >Is it possible >that both of the 2 savefiles (they're both there, and >both of the proper size) are corrupt? Yes, but you would see an error message (I think in both results.txt and prime.log) that the pXXXXXXX is bad, renaming qXXXXXXX to pXXXXXXX, and again pXXXXXXX is bad. >Iteration: 7752549/12962641, ERROR: SUM(INPUTS) != SUM(OUTPUTS), >1524171845291614 != 283103064441664.6 This error is real bad - almost always an indicator of a hardware problem. Time for some torture tests. Sorry I cannot be more helpful, George _________________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
