Mersenne Digest Tuesday, November 7 2000 Volume 01 : Number 790 ---------------------------------------------------------------------- Date: Thu, 2 Nov 2000 06:41:09 -0000 From: "Brian J. Beesley" <[EMAIL PROTECTED]> Subject: Re: Mersenne: ECM update -- M727 finished up to 50 digits On 1 Nov 00, at 18:17, David A. Miller wrote: > The last machine that I had working on M727 has finished its 1000 curves > at B1=44M. This is enough to finish the recommended number of curves at > that bound. Thus there are probably no factors below 10^50, and it won't > be practical to find the factors with ECM. The question arises as to whether or not it is economical to continue beyond B1=44000000, B2=4290000000. This may be a function of the native integer size - 64-bit systems shouldn't have any problems (given a suitable implementation) I guess the special number field people will take an interest as soon as they've cleared whatever they're working on at the moment. > > Have any other numbers received as much ECM effort as this? I'm betting > that there aren't many. Possibly not - even if the curves for B1=44000000 B2=4290000000 have been done, they've probably been done on a smaller exponent, therefore with less expenditure of time. Although the curves completed don't look as good, it is, however, possible that more CPU time has been expended trying to factor some of the Fermat numbers using ECM. There's still plenty of work for you to do! Regards Brian Beesley _________________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.exu.ilstu.edu/mersenne/faq-mers.txt ------------------------------ Date: Thu, 2 Nov 2000 10:18:52 +0000 From: Alexander Kruppa <[EMAIL PROTECTED]> Subject: Re: Mersenne: ECM update -- M727 finished up to 50 digits "Brian J. Beesley" wrote: > > The question arises as to whether or not it is economical to continue > beyond B1=44000000, B2=4290000000. This may be a function of the > native integer size - 64-bit systems shouldn't have any problems > (given a suitable implementation) I have a hacked version of mprime that uses another prime sieve (the primegen library), it will let you go all the way to B1=2^32, B2~=2^60. It seems to work alright but I havent given it too much testing yet - I've had little time recently. If there is any demand out there from truely desparate ECM factorers, I'll try to make a releasable version. > > Have any other numbers received as much ECM effort as this? I'm betting > > that there aren't many. > [...] > Although the curves completed don't look as good, it is, however, > possible that more CPU time has been expended trying to factor some > of the Fermat numbers using ECM. I think F14 beats even M727. I've done a lot of curves at B1=3M and B1=11M, if I recall correctly, one B1=11M took about 11 hours on a PII-400, so the 5200*3M + 1270*11M plus smaller curves should add up to something like 30.000 cpu hours, while M727 (1 hour per 44M curve) has around 25.000 hours accumulated. Ciao, Alex. _________________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.exu.ilstu.edu/mersenne/faq-mers.txt ------------------------------ Date: Tue, 07 Nov 2000 22:32:47 EST From: [EMAIL PROTECTED] Subject: Mersenne: New Mlucas binary for Sparc Thanks to Bill Rea, we now have a new, improved Mlucas 2.7a binary for Sparc Solaris (2.6 and above; I've also run it successfully under SunOS). On a 333 MHz Ultra II I have at work, my 2000-iteration timings for an exponent around 5.9M (320K FFT length) dropped from 9 to 7 minutes. Thanks, Bill! ftp://hogranch.com/pub/mayer/README.html Happy hunting, - -Ernst _________________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.exu.ilstu.edu/mersenne/faq-mers.txt ------------------------------ End of Mersenne Digest V1 #790 ******************************
