mersenneforum.org (https://www.mersenneforum.org/index.php)
-   PrimeNet (https://www.mersenneforum.org/forumdisplay.php?f=11)

 afjewkes 2016-07-14 20:25

Assignment credits?

Hi, I'm new to prime95... I've tried to find information on this but I've been unable to, or I've read over it...

I'm currently running a LL first test for an exponent - if this was to find a prime, would I get credit at the end of this test (which is up to 2^79 for this exponent) and then a potential award once it's been checked by over users?

Another question, I'm running an exponent on a slower PC that seems to be taking 53 times longer? And in prime95 it says stage 1 is x% complete, how many stages are there?

 Uncwilly 2016-07-15 04:40

First and foremost, welcome to the search for Mersenne Primes! Hope you enjoy yourself.

Let's see if we can get you the answers are looking for:
[QUOTE=afjewkes;438120]I'm currently running a LL first test for an exponent - if this was to find a prime, would I get credit at the end of this test (which is up to 2^79 for this exponent) and then a potential award once it's been checked by over users?[/quote]
So, you have selected to do a first time test on an exponent. You say that it has been taken (trial factored) to 2^79. Based upon this, I am guessing that you are working on an exponent in the 332M range (the 100 million decimal digit range.) You are hoping to be the one to find the first known 100 million digit prime. These tests take a long time on many machines. Credit for any assignment is only awarded when the results of the task are turned in. For LL tests, the whole test has to be completed to known the result and credit can not be properly awarded for incomplete results. For Trial Factoring, credit can be award for each bit level result that gets turned in or when a factor is found. For P-1, the credit can only be awarded when the test is complete or the factor is found.

If it is found to be prime on your machine, it will be double and triple checked on different kinds of machines using different programs written by different people. If all of these show that it is prime, then process of pursuing the EFF award starts.

[quote]Another question, I'm running an exponent on a slower PC that seems to be taking 53 times longer? And in prime95 it says stage 1 is x% complete, how many stages are there?[/QUOTE]53 times longer than what? If you are working on an exponent in the 332M range, these take ~5,000 GHz-days worth of work to do an LL test on.

Stage 1: This makes me think that your machine is doing P-1 factoring. This and TF are done before LL testing to check for factors. With these test we can eliminate over 60% of the exponents by finding factors. It is a wise investment of time to take the numbers to certain bit levels and P-1 bounds before LL testing. P-1 has 2 stage with each of these having 2 phases. Once P-1 is complete, then the LL test will start. If an exponent has previously had enough TF and P-1 done on it, Prime95 will start the LL directly.

 LaurV 2016-07-15 06:26

[QUOTE=afjewkes;438120]
I'm currently running a LL first test for an exponent - if this was to find a prime, would I get credit at the end of this test (which is up to 2^79 for this exponent) and then a potential award once it's been checked by over users?
[/QUOTE]
There is no "award" beside of the credit (which can be used for nothing, except boasting to your friends that "look, I do math research!" hehe), except for the case when your exponent turns out to generate a prime number. In that case there may be a money reward, like \$3000 or so, and you will be famous for a week :smile:. The chances you find a prime are extremely slim. If you work in 332M range and find a prime, you can get the [URL="http://www.mersenne.org/legal/"]EFF award[/URL], and be famous for a little longer.

[QUOTE]
Another question, I'm running an exponent on a slower PC that seems to be taking 53 times longer? And in prime95 it says stage 1 is x% complete, how many stages are there?[/QUOTE]"Stages" are only when talking about "P-1 factoring". There are 2 stages, each taking the same amount of time, about. This P-1 is a "helping" phase, which tries to factor that big number, and in case it succeed, you will avoid running the [B][U]_LONGER_[/U][/B] (in time) LL test. You can not find a prime running the P-1, but you may be lucky and avoiding wasting even more time trying to prove a prime (by LL) which in fact is composite, having a small factor. P-1 can find this factor and avoid time consuming. Now, on the other hand, you should [B][U]NOT[/U][/B] get P-1 work, unless you expressed explicitly that option, or unless you selected a insanely huge exponent for which P-1 was not done enough. That is because there are dedicated software tools that work on different hardware (not on your CPU) which are more efficient in doing P-1, and there are few guys here who like to find large factors, so they do a lot of P-1, using "special" hardware, "filtering" in this way the exponents. What is left after these guy's work should be "ready for LL", and no P-1 should be necessary. This of course, unless you chosen deliberately an exponent which is "too big", and these guys here didn't reach it with their P-1 machines... Which I recommend you to cancel, and start a normal "LL front" exponent (the "front" is now around 79-80 millions, and slowly advancing).

Another idea is that you start first with double-checking smaller exponents, which will be good for you as a beginner, from three points of view: first, they finish faster (because the exponents are smaller) giving you a small satisfaction (and some fast "credits"), second, you can get a residue which you compare with the other guy who did initial test, and in this way you know that your hardware is working properly, and it is apt to do the longer LL tests, and third, during this testing period, you learn how the things are going here around. More info about P-1, LL, DC, etc., you can find on [URL="http://www.mersenne.org/various/math.php"]gimps' math page[/URL].

Edit: crosspost with Unc, I had this page open since morning (lunch break here now) but no time to write, and didn't see you posted, sorry.

 All times are UTC. The time now is 12:05.