Re: Mersenne: The recent popularity of Single-Checking
>Since when has this project become a competitive event ? > >This mailing list has gotten several messages like the one above, >which I interpret as comparing ANOTHER USER to standards set by >the writer, rather than acknowledging that *all* users are >contributors to the project. Will it mean the end of the world >if that other user had mis-stated the resources available to him ? No, however it reduces the chances of finding a prime for others and delays the milestones. I am not in the project solely to find a prime - I have already committed to spending this entire summer from late June to early September on a single exponent for QA, reducing my chance of finding a prime at least eightfold for that time. >I think it is intrusive to publicly comment about ANOTHER USER when >the writer spots something that does not meet his own expectations. >Can't we please let each participant remain responsible for his own >performance?"Why do you look down on your brother? For we will >all stand before God's judgment seat." (Romans 14:10) > >mikus I just have a problem with extreme cases like this one. It's one thing to cache three months of work, and quite another to cache the better part of a decade. Also, many of the exponents in question were down in the 8.1 M range, implying that the user had deliberately taken on smaller reissued exponents and abandoned them. This hurts the progress of GIMPS as a whole. Not that I have not myself abandoned exponents - my machine dropped several double-checks when I stopped running 24/7 due to problems sleeping. Nathan Get Your Private, Free E-mail from MSN Hotmail at http://www.hotmail.com _ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: The recent popularity of Single-Checking
> Nathan Russell wrote: > >There is a user, "sd70045", who has almost 100 single-checking > >assignments out on a single machine ID > > In any case, these exponents will expire in 16 days. As a > result, I'm not concerned about it. Within 3 weeks > they'll be re-circulated among other users. The only thing > that would cause concern, is if the user intentionally updates > these exponents in the next two weeks... > xactly! Last time I checked there was an infinate number of possible mersenne candidates and using the current version of prime95/NT/whatever, we are umlikely to run out any time soon,so unless things get REALLY serious, surely with things like this it's best to just wait till they expire and let the automatted system deal with it? Chris > > Eric > > > _ > Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm > Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers > _ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: The recent popularity of Single-Checking
Nathan Russell wrote: >There is a user, "sd70045", who has almost 100 single-checking >assignments out on a single machine ID. These would take a >state-of-the-art box well over two years to finish. Additionally, >these assignments have almost identical figures for time to >complete etc. The first exponent in this group is 8936071; the >others are directly below it. I examined this, and found out that there is actually 197 assignments checked out to this individual (188 to the same machine ID (7 dbl-chks, 5 factoring, 176 L-L tests)). By my estimates, this single machine ID has >5 yrs worth of work for even the faster state-of-the-art PC. While they have various run times, they all have 16 days to go and 16 days until expiration... They all were also updated on the same date and time (10-Feb-00 17:55). None appear to have had any work performed on them at all!! While I normally might think this might be a person switching over to use PrimeNet from previously not using it, and possibly using a large cluster (using the same ID for the entire cluster), there are a few indications this isn't the case. First, their ranking on PrimeNet is 8112 and 2323 for primality testing and factoring respectively, and their P-90 CPU hrs/day at 13.79. Second, their ranking on George's list is 6800 (with only one additional exponent tested above PrimeNet's count). Finally, they have 6 other machines that actually appear to be performing some kind of work. In any case, these exponents will expire in 16 days. As a result, I'm not concerned about it. Within 3 weeks they'll be re-circulated among other users. The only thing that would cause concern, is if the user intentionally updates these exponents in the next two weeks... Eric _ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: The recent popularity of Single-Checking
> There is a user, "___", who has almost 100 single-checkingassignments > out on a single machine ID. These would take a state-of-the-art box well > over two years to finish. Additionally, these assignments have almost > identical figures for time to complete etc. The first exponent in this > group is ___; the others are directly below it. Since when has this project become a competitive event ? This mailing list has gotten several messages like the one above, which I interpret as comparing ANOTHER USER to standards set by the writer, rather than acknowledging that *all* users are contributors to the project. Will it mean the end of the world if that other user had mis-stated the resources available to him ? I think it is intrusive to publicly comment about ANOTHER USER when the writer spots something that does not meet his own expectations. Can't we please let each participant remain responsible for his own performance?"Why do you look down on your brother? For we will all stand before God's judgment seat." (Romans 14:10) mikus _ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Mersenne Digest V1 #739
Mersenne DigestWednesday, May 24 2000Volume 01 : Number 739 -- Date: Mon, 22 May 2000 09:45:32 -0400 From: Sandy Harris <[EMAIL PROTECTED]> Subject: Mersenne: [Fwd: P1363: Primality prover] Is this of use here? - Original Message Subject: P1363: Primality prover Date: Mon, 22 May 2000 06:27:24 +0200 From: [EMAIL PROTECTED] Reply-To: [EMAIL PROTECTED] To: [EMAIL PROTECTED] - - This is a stds-p1363 broadcast. See the IEEE P1363 web page (http://grouper.ieee.org/groups/1363/) for more information. For list info, see http://grouper.ieee.org/groups/1363/maillist.html - - A beta release of CERTIFIX (a primality proving program I am writing) is available. It is based on the Goldwasser, Kilian & Atkin algorithm. CERTIFIX is an executable for Win95, Win98, NT (hardware Intel compatible). It is a freeware. Currently, it can certify a 1024 bit integer in less than 10 mn (AMD K6-2/450 processor). Download link: http://www.znz.freesurf.fr/files/certifix.zip (300 Kb) The package contains the 5 following files certifix.exe certifix.hlp readme.txt todo.txt changes.txt There is no "install" program. Just copy the files in the folder you want and click on CERTIFIX.EXE. Marcel Martin _ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers -- Date: Mon, 22 May 2000 13:43:02 EDT From: [EMAIL PROTECTED] Subject: Mersenne: Re: reliability of double-checking Stefan Struiker wrote (re. the reliability of double-checking): > How many DDs does it take to "reasonably establish" primality, given the > apparent slight error in the L-L test? Is there an extra-precision > version of the test to nail things down? A few years ago, you would likely have heard quite a few people say that such a result could only be acceptable if confirmed via an all- integer test, i.e. one devoid of roundoff error. Now that the reliability of carefully written floating-point-based codes for primality testing has been established to most everyone's satisfaction, the currently accepted standard is two independent tests (floating-point or not) running on separate hardware, preferably using different programs. George has pretty much obviated the latter criterion by using a binary-shifted start value for the LL test, which means that even running the same code on the same exponent and same hardware twice in a row will involve completely different FFT vectors (but which have digits of similar magnitudes) p-1 out of p times; in case one really needs to convince the remaining critics (e.g. if the first test indicates primality), one can simply make sure the double-check is done using an independent program. Nathan Russell wrote: > IIRC, there is no real difference in the double-check and the > first-time test. On average 1 in p times there is not, for the reasons I cited above. Note that this also applies to cases where Prime95 was used for one run and a different code for the other. An interesting sidelight to the all-integer vs. floating-point debate: all-integer is generally thought to be more reliable because it suffers no roundoff (RO) errors, but it's become quite clear that RO errors are not the major culprit in erroneous runs, and moreover can be detected quite reliably. It's the other system components (especially on PC-class hardware, where profit margins are thin to nonexistent and corners must be cut) that more often cause trouble; non-ECC memory, CPU overheating, unexpected (well, perhaps not under Windows :) interactions between various applications, actual flaws in the CPU or motherboard, etc. All of these can ruin a calculation, whether that be floating-point or all-integer. The other thing that is often overlooked is that even an all-integer run can suffer a non-hardware-related kind of error, namely of the overflow variety. Unlike RO error, this can be difficult (or at the very least expensive) to detect, especially for things like a modern Mersenne-testing algorithm, which never explicitly finds the unmodded square of the input number, i.e. does not permit a casting-out-nines- style checksum. Some will say, "why not just enable overflow trapping?", but most performance-oriented CPUs have no hardware support for such trapping, because it is inimical to speed. Instead, one must trap in software, using something like the following pseudocode sequence: unsigned x, y, z !* x, y, z are register-length ints logical of_trap z = x + y !* the sum may overflow... of_trap = (z < x) !* ...in which case we detect it via
Mersenne: The recent popularity of Single-Checking
Hi all, There is a user, "sd70045", who has almost 100 single-checkingassignments out on a single machine ID. These would take a state-of-the-art box well over two years to finish. Additionally, these assignments have almost identical figures for time to complete etc. The first exponent in this group is 8936071; the others are directly below it. Puzzled, Nathan Get Your Private, Free E-mail from MSN Hotmail at http://www.hotmail.com _ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers