mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   Software (https://www.mersenneforum.org/forumdisplay.php?f=10)
-   -   Prime95/PrimeNet Mersenne Probability bug (https://www.mersenneforum.org/showthread.php?t=24597)

wfgarnett3 2019-07-18 02:58

Prime95/PrimeNet Mersenne Probability bug
 
2 Attachment(s)
George,

I have the screenshots for you to explain this bug.

I got a new assignment and my worktodo.txt file is timestamped 7/17/2019 12:57AM when it got the assignment and has the below line (note the worktodo.txt file has not changed at all and is timestamped the same still to this moment)

[Worker #1]
PRP=72234AC90891F2306AE2D9C69C87A4B6,1,2,86060617,-1

At 10:39PM on 7/17 in Prime95 I went Test-->Status and it as shown in the attached screenshot1.png "The chance that the exponent you are testing will yield a Mersenne prime is about 1 in 662132".

I went Help -->About PrimeNet Service to send the 5 million residue.
I then exited Prime95.

I went back into Prime95 and went Test-->Status and now it states as shown in the attached screenshot2.png "The chance that the exponent you are testing will yield a Mersenne rpme is about 1 in 1601932".

That is the bug -- I am testing the same exponent, worktodo.txt did not get touched at all, yet a couple minutes apart Prime95/PrimeNet now reports a different probability. For the rest of the time I test this exponent it will report the different probabilty.

This happened for me for exponents in this new Prime95 (29.8 build 3) when I do my PRP first time tests -- it changes the probability and for the rest of the test until it is done will continue to report this incorrect probability. The previous 29.4 software I used for LL first time tests never did this -- it would always report the 1 in 662132 probability for instance.

Can you advise what is happening?

William

LaurV 2019-07-18 04:27

No idea if you found a bug or not (never looked for these probability numbers, they have no practical use, we just know from experience that the probability to find a prime is quite small :blush:)

But... The probabilities can change on the fly if somebody reports more TF or some P-1 with larger bounds (not the case here, tho).

Also... This exponent had a P-1 of about 5% and it was TF to 76 bits, therefore the probability of LL turning it prime is something like 1 in over a million. I think the first number is the one which is wrong, and the last one is the correct one. But this is just "in the head" approximation, I didn't take the pencil, so I could be wrong.

Edit: mind that your worktodo does not contain information of how far factored, and pm1 bounds, etc, those are only known after contacting the server. A typical LL line look like: "Test=blahblahkeykey,332199017,79,1", which for example, tells me the TF level, and give a hint about P-1. But for PRP, you need to contact the server to know these things. So, from this point of view, it also makes sense that the second number is right.

The chance of your number to be prime does not depend on the PRP or LL, it depends mostly of the exponent size and how much work was already done on it, except the PRP or LL work you are currently doing. Progressing into the LL or PRP test does not change the chances of the number being prime.

Therefore don't be so sad, your chances to find a prime it is just very-very-very small :razz::razz:

ixfd64 2019-07-18 21:11

I believe this is fixed in version 29.8 build 5: [url]https://mersenneforum.org/showpost.php?p=508842&postcount=2[/url]

wfgarnett3 2019-07-19 03:37

Thanks ixfd64 I see George states it is fixed in the later version

----
21) For PRP work, TF depth was not being written to worktodo.txt unless P-1 was required. This caused Test/Status to underestimate the chance that the PRP test would find a new Mersenne prime. Fixed in 29.8 build 5


All times are UTC. The time now is 18:26.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2020, Jelsoft Enterprises Ltd.