BUG IN VERSION 17 OF GIMPS SOFTWARE From: "Clayton Smith" To: Subject: Important notice for GIMPS members Date: Mon, 5 Apr 1999 17:24:51 Hi everyone: A few days ago, a bug was discovered in version 17 of the GIMPS client software. This bug produces bad results for exponents >4.2 million as described in the newsletter below. Fortunately, there hasn't been too much work lost. To get things back on track, please download the version 18 software from the GIMPS website as described below. Be sure to overwrite only the executable(s), leaving the configuration files intact. As always, there is information about the project, and preconfigured clients available on the class website (http://www.eng.uwaterloo.ca/classes/comp03a, under Misc. Stuff section). This site now contains the new version of the program. Thanks for your patience. - Clayton ----- Original Message ----- From: George Woltman To: Sent: Sunday, April 04, 1999 5:13 AM Subject: The Mersenne Newsletter, issue #15 > To: Great Internet Mersenne Prime Search Registered Recipient > > Note: This will be the last automatically emailed newsletter. You > will not receive anymore newsletters unless you explicitly sign up > as described below. > > VERY IMPORTANT: Please read this if you are still a GIMPS participant!!!! > > The Mersenne Newsletter, issue #15 April 3, 1999 > > Contents > 1. Serious Bug Found in Version 17 > 2. EFF Announces $50,000 Prize For Finding Million Digit Prime!!! > 3. Future Newsletters > 4. Welcome New Members > 5. PrimeNet Server News > 6. ECM Factoring News > 7. George's New Email Address > 8. GIMPS Status Update > > > 1. Serious Bug Found in Version 17 > ---------------------------------- > Version 17 was released a few months ago to allow double-checking > previously tested exponents. Since about 1 in 100 previous tests are > incorrect, you can still find a Mersenne prime by double-checking. > This is the perfect chore for slower Pentiums. You don't have to wait > several months for each Lucas-Lehmer test to complete. > > Unfortunately, a bug was introduced that affected tests on exponents > above 4,194,304. Download version 18 immediately at > http://www.mersenne.org/freesoft.htm or ftp://entropia.com/gimps > > This bug affected about 11,000 out of 59,000 reported results above 4,194,304. > All results produced by version 16 are OK. All results that were started > by version 16 and completed by version 17 are OK. All factoring and > double-checking work is OK. > > After upgrading to version 18, you may get error messages that look like this: > Error reading intermediate file: p6180331 > Renaming intermediate file q6180331 to p6180331. > Error reading intermediate file: p6180331 > This is normal. Prime95 must discard incorrect version 17 save files. > > To those that have been using version 17 to test exponents above 4,194,304 > I offer my sincerest apologies. I know that upgrading will be very > inconvenient and it will be disheartening to lose months of work. Sorry. > > On the bright side, perhaps one of the 11,000 affected exponents will > turn out to be prime allowing someone to win a big prize described below! > > > 2. EFF Announces $50,000 Prize For Finding Million Digit Prime!!! > ----------------------------------------------------------------- > On March 31, the Electronic Frontier Foundation announced a $50,000 > prize to be awarded to the first person to find a million digit > prime number! The details are available at http://www.eff.org/coop-awards/ > > Since GIMPS is now testing numbers well in excess of 1 million digits, > some lucky GIMPS member could very well win this prize! > > If you are using business computers, a friend's computer, or a > co-worker's computer, I recommend you discuss how the $50,000 prize > would be divided should you be the lucky winner. Note that according > to the contest rules, you could be denied the prize if you have not > legally obtained permission to use the computer that finds the prime number. > > In light of the EFF anouncement, the $1,500 prize fund for discovering > the 38th Mersenne prime has been discontinued. > > > 3. Future Newsletters > --------------------- > I will no longer email the Mersenne newsletter to every GIMPS user. Instead, > there will be three ways to get future newsletters: > 1) Online at http://www.mersenne.org/prime.htm. Check periodically for > new editions. > 2) Subscribe to the Mersenne mailing list. I will send future newsletters > to the mailing list. > 3) Use the web form at http://entropia.com/ips/register.html to request > having future newsletters emailed to you. A link to the web form will > also be available from http://www.mersenne.org/prime.htm > > > 4. Welcome > ---------- > Since the last newsletter 10 months ago, many thousands have joined GIMPS! > There are now 9,000 people and teams involved in the search. Welcome to all > the new searchers! I'm sorry your first newsletter had to announce bad > news. This is the first serious setback for GIMPS in its three and a half > year history. > > > 5. PrimeNet Server news > ----------------------- > Link your Top Producer's Page account name to your home page! > Check out the people, teams, businesses and schools behind the > top CPU time accounts at http://entropia.com/ips/tops.shtml, > and add your own page link! > > If you are using the RPC protocol, please use the faster HTTP > protocol instead. You can select this option in the version 18 > Windows software in the Test/PrimeNet dialog box. MPrime and > PrimeOS2 already use HTTP. > > To get your full PrimeNet account report without a password > (for example, to use in a web page), email primenet@entropia.com > with your request. > > If PrimeNet emails you a confirmation that it received > a Mersenne prime from your computer, please remember worldwide > academic research discovery rules apply. Please contact me > immediately at woltman@alum.mit.edu, and do not communicate the > discovery until we have verified it on a conventional Cray > supercomputer. I will help you work with EFF to collect your > prize. > > For other network, server or PrimeNet account related questions, > please email Scott Kurowski's team at primenet@entropia.com. > > Entropia.com upgraded the GIMPS Internet PrimeNet Server on > 9 February to PrimeNet 4.0, a powerful new system built on their > new v4 Internet computing technology. The 4.0 server offers > substantial reliability improvements and an order of magnitude > better performance and feature integration flexibility. > > The thoughtful Entropia.com community of GIMPS participants > continues to grow. The PrimeNet virtual supercomputer is > currently producing over 550 billion floating point operations > per second, which puts it among the top 10 most powerful systems > in the world. This amazing strength arises from the spare CPU > time from your computer and some 15,500 others on 9,000 > individual, team and business accounts. For more on > Entropia.com's PrimeNet system and related projects, see > http://entropia.com/ips and http://entropia.com, or email > info@entropia.com. > > The GIMPS PrimeNet Server is provided by Entropia.com, Inc. > > > > 6. ECM factoring > ---------------- > The program now lets you find factors of small Mersenne numbers using the > Elliptic Curve Method. Mathematicians are interested in completely factoring > Mersenne numbers as well as finding new Mersenne primes. If this interests > you, visit http:\\www.mersenne.org\ecm.htm for more details. To date we have > discovered 16 new Cunningham factors, including Conrad Curry's largest factor > ever found using ECM. > > > 7. New email address > -------------------- > My new email address is woltman@alum.mit.edu. My old address, > woltman@magicnet.net, still works. In fact the new address simply forwards > mail to the magicnet address. However, should I ever change ISPs the new > address will forward my mail to the appropriate place. > > > 8. Status > --------- > In the ten months since the last newsletter GIMPS as accomplished > the following milestones: > > * Double-checking proves M(1398269) is the 35th Mersenne prime. > * All exponents below M(2976221) tested at least once. > * All exponents below M(3021377) tested at least once. > * All Mersenne numbers less than a million digits tested at least once. > > We have completed first-time Lucas-Lehmer tests on about 45,000 Mersenne > numbers. Of these 45,000 numbers tested, we expected to find one > Mersenne prime. We've been a bit unlucky. > > However, there are still 396,543 exponents left to test below 20,500,000. > It will only take us 534,493 P-90 CPU years to test exponents and we expect > to find 3.71 Mersenne primes. At Primenet's current rate, we expect to > find one Mersenne Prime this year. > > You can watch the hourly progress of the search on the server's > status page at http://entropia.com/primenet. > > > Conclusion > ---------- > Once again, I'm sorry about the version 17 bug. Best wishes and good luck > in the future. Hopefully someone will soon find the 38th Mersenne prime > and collect a nice prize! And I can write a more upbeat newsletter! > > > Sincerely, > > George Woltman > woltman@alum.mit.edu > > and > > Scott Kurowski > primenet@entropia.com > >