 mersenneforum.org > Math about Lucas-Lehmer test and Prime 95
 Register FAQ Search Today's Posts Mark Forums Read  2003-07-22, 18:43 #1 Annunaki   Jul 2003 31 Posts about Lucas-Lehmer test and Prime 95 My question is about primarity test of M numbers.. I am runing a primarity test of M16328749 and about every 20 minutes or so i got about 0.06% of total job done.. However it calls some confusion in my mind.. Does anybody knows if the test is stopped if it suddenly finds some facor of the number.. or it is continiued befor the number is compleetly factored.. To my mind if would be logical if the test stopped to run after first factor is found, but how can the test estimitate then how much percent of total work is done if it is not known when any factor may or may not appear.. Does the % mean the job done in % in case it really is the prime.. And then does it means that if the i see more than 90% of job finished i should allready start to get nervous:)?   2003-07-22, 19:04 #2 eepiccolo   Dec 2002 Frederick County, MD 2·5·37 Posts The Lucas-Lehmer test doesn't factor a number, it just test for primality. Trial Factoring is performed on a number before The LL test is performed, and if no factor is found, the the LL test will start. An LL test always must complete to 100% to determine primality. Only then can the program determine if the number is prime or not, which it does by looking at the LL residue.   2003-07-22, 19:52 #3 Annunaki   Jul 2003 111112 Posts strange.. if trial factorign found no factors then this nuumber should already prime.. or trial factoring only does half job? and i thought primarity test is still searching for factors - if no factors found then nuber is prime.. But then could anyone explain how works this test..   2003-07-22, 20:08 #4 eepiccolo   Dec 2002 Frederick County, MD 5628 Posts Trial factoring only looks for small factors. If no factors are found, then the program goes on to the LL test, which only proves or disproves primality; it tells nothing about the factors. Here is a link to the math explaination on the mersenne.org site: http://www.mersenne.org/math.htm   2003-07-22, 20:10   #5
trif

Aug 2002

2·101 Posts Quote:
 Originally Posted by Annunaki strange.. if trial factorign found no factors then this nuumber should already prime.. or trial factoring only does half job? and i thought primarity test is still searching for factors - if no factors found then nuber is prime.. But then could anyone explain how works this test..
Trial factoring doesn't factor all the way, but only to a shallow bit depth. It takes out the "easy" ones in order to avoid doing an LL test on an exponent where a factor can be found with relatively little work. Completely trial factoring exponents of the size we are working on now would be impossible.   2003-07-23, 09:25 #6 Annunaki   Jul 2003 31 Posts yes that is what seemed so strange.. because if trial factoring is just dividing with all primes in a row we should wait some hundreds of years just to hope to see some factor:)   2003-07-23, 12:37   #7
smh

"Sander"
Oct 2002
52.345322,5.52471

29·41 Posts Quote:
 Originally Posted by Annunaki yes that is what seemed so strange.. because if trial factoring is just dividing with all primes in a row we should wait some hundreds of years just to hope to see some factor:)
It would take much longer than a couple of hundred years to factor current numbers to the square root. The numbers we are testing have millions of digits. Trail factoring of numbers of 40 digits is almost impossible and very inefficient.   2003-07-23, 13:09 #8 Annunaki   Jul 2003 31 Posts tes have heard this allready.. just thought that some hundreds of years could be enough to see some factor.. if that number is prime then.. i feareven to guess how much time it would ask.. Of course all is dependent how and on what you run the program.. It is almost imposible only to our computers, but i think it is posible to human..   2003-07-28, 02:49 #9 cheesehead   "Richard B. Woods" Aug 2002 Wisconsin USA 22·3·641 Posts A while ago, while writing a posting illustrate the magnitudes of the numbers we work with, I looked up current estimates of the numbers of particles in the known universe, size of the known universe, and related stuff. Without giving links or specific citations, here is a rough calculation to demonstrate the impossibility of trial-factoring to the square root of the size of Mersenne number with which GIMPS is now working: The known universe could hold (far) less than 10^200 neutrons (the most compact elementary particle) if it were packed full, with no empty space. (In reality, the universe is more than 99.99% empty.) The "Planck time", which is smaller than any time in which any conceivable useful computation could take place, is greater than 10^-44 second. Let's make that 10^-100 second, just so there's no quibbling. So, no computation could be performed in less than 10^-100 second. Suppose the known universe were packed full of neutrons, and suppose each neutron were a computer capable of performing one trial-factoring division in 10^-100 second. So, altogether the universe could perform 10^300 trial-factoring divisions per second. Now, 10^300 is less than 2^1200, so all the computers in the entire known universe can perform no more than 2^1200 trial-factoring divisions per second. The estimated age of the known universe is 13 billion years. There are about 31 million seconds in a year. So the universe is about 400 million billion seconds old. That's 4 * 10^17 seconds, which is (much) less than 2^100 seconds. So all the computers in the universe, operating at the fastest possible speed for longer than the age of the universe, could perform no more than 2^1300 trial-factoring divisions. What is the size of a number than has 2^1300 primes below its square root? Let N be the number, and Q be its square root. Pi(Q) = 2^1300 Pi(Q) = approx Q / ln Q {Edit: Here I'm making a WAG ("educated guess":) that Q is about 2^1400 --} The natural log of 2^1400 would be less than 1400 (because e^1400 > 2^1400). So if Q were 2^1400, then Pi(Q) would be greater than (2^1400)/1400 > 2^1380. So, we know Q < 2^1400, and so N < 2^2800. In other words, if the entire known universe were packed full of neutrons and each neutron were a computer operating at maximum possible speed, and all the little computers ran for the entire age (so far) of the universe, it could completely trial-factor a number no larger than 2^2800. WAIT! We left out the optimizations -- like each factor has to be 2kp+1 and +-1 mod 8, and so on. Suppose our optimizations allow us to skip 999,999,999,999 out of every 1,000,000,000,000 primes below the square root of the number we're trying to factor. That means we can TF a trillion (~2^40) times as many potential factors. So we want Pi(Q) = 2^1300 * 2^40 = 2^1340 instead of 2^1300. Hmmm ... looking back, we find that "So if Q were 2^1400, then Pi(Q) would be greater than (2^1400)/1400 > 2^1380" still is valid. We don't have to change our previous answer -- The entire universe could TF a number no larger than 2^2800. (Using our current trial-factoring methods and optimizations, that is.) {EDIT: Let me restate that conclusion so that it is clearer when quoted out of context -- Even if the entire known universe were one solid computer operating at maximum speed for the entire time since the Big Bang, it could not yet have trial-factored a number larger than 2^2800 all the way to its square root.} Let's see ... how big are the numbers GIMPS is working on now? Current PrimeNet trial-factoring assignments are greater than M21000000 = 2^21000000 - 1 which is far, far larger than 2^2800.   2003-07-28, 03:34   #10
ColdFury

Aug 2002

32010 Posts Quote:
 it could completely trial-factor a number no larger than 2^2800.
And the sad thing is, this is more or less the method that the Neo Project is using to try to crack the X-Box's 2048 bit public key.   2003-07-28, 07:09 #11 Xyzzy   Aug 2002 22×2,161 Posts I've always wondered how the bit-depth of a RSA key affects its "crackability"... For example, I use a 2048-bit key for my sshd session... If I'm thinking about this right that key should be impossible to crack, right? I mean, RC5 took forever to crack 64 bits, right? Or am I looking at this wrong?   Thread Tools Show Printable Version Email this Page Similar Threads Thread Thread Starter Forum Replies Last Post Trilo Miscellaneous Math 25 2018-03-11 23:20 Mathsgirl Information & Answers 23 2014-12-10 16:25 MrRepunit Math 9 2012-05-10 03:50 science_man_88 Miscellaneous Math 48 2010-07-14 23:33 storm5510 Math 22 2009-09-24 22:32

All times are UTC. The time now is 11:35.

Fri Jun 9 11:35:27 UTC 2023 up 295 days, 9:04, 0 users, load averages: 0.88, 0.90, 0.85