mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   Prime Sierpinski Project (https://www.mersenneforum.org/forumdisplay.php?f=48)
-   -   Sieve Vs PRP (https://www.mersenneforum.org/showthread.php?t=7608)

Chino112 2007-03-23 16:59

Sieve Vs PRP
 
As the range reservation today is at 536T and has rocketed in the last 3-6 months I was wondering what people thought about this. Is this affected the PRP ranges we are currently doing? As PRP seems to have slipped to only a brave few while the rest concentrate on Sieving.

I know that the more you sieve, the more you eliminate. But has the balance shifted too much? Is the sieving eliminating PRP ranges that we wont get to for a while etc.

what is everyone`s thoughts?

hhh 2007-03-23 17:30

Sieve vs PRP, one of my favorite topics!:grin:

Let me put it like this: Joe Averagefactor eliminates a candidate at about 25M. Such an LLR test takes like forever; at least several weeks.
So, as long as we are not waiting for a factor several weeks, we should sieve.

HA!, you say, but if I do a little LLR-test now, no harm is done. Let us LLR up to 5M and sieve then. Ok, I reply, but then you wouldn't have got all these factors for 3M-5M which would have saved considerable LLR-work.

HA! you say again, but I would have found a prime. And hence we could sieve faster. Ok, I say again, but not much. Sieve speed is logarithmic, and adding or removing 5% of the candidates results in 2-3% speed change only. And even if we sieve first, and half of the factors go out of the window and never save an LLR-test, well, then we should not sieve until we get a factor every 4 weeks, but every 2 weeks. And that's still far ahead.

In this manner, I'm not very sorry if we LLR less and sieve more. But I would be even more happy if we could maintain the LLR-rate and sieve more.

But then again I think that all this is random fluctuation, and that we have to be happy with what we get.

Bottomline: LLR is cool, sieve is cooler.

Yours H.

PS: My exhaustive description of my point of view isn't meant to prevent others to express their points of view. Especially opinions about how to raise interest (without going BOINC) would be interesting, I think. H.

Joe O 2007-03-24 13:48

Chino112,
Are you missing some PRP results?
[CODE]PRP Stats:
rank nickname amount last 30 days points last result teamname
53 Chin112 48 0 17752.719567 2007-02-08 [/CODE]

HHH,
Good post. Raising interest you say? How about two contests, running simultaneously. Who can find the most factors in April, and who can do the most LLR/PRP tests in April. It will be more fun than only one contest, or even two contests one after the other. April will be perfect because it only has 30 days and we can use the "last 30 days column" of the last update on the last day to decide the winners. The prize? Bragging rights. The winners name on the stats page in a prominent position?

Chino112 2007-03-25 14:29

huh? where is that ranking from?

I am currently 10th in PRP and 16th in Sieve. So I am not sure as to where I am 53rd and why I would be missing results.

Chino112 2007-03-25 14:33

oh wait I see. I have been put in twice in the rankings. One as Chino112 and one as Chin112 :lol:

48 results assigned to an o-less me.

ltd 2007-03-25 16:15

This is not my day, week, month, ....? :raman: :raman: :surrender
Again my fault. :whistle:

I will correct that in the stats when i import your latest results.

Lars

ltd 2007-03-28 19:15

Some additional numbers about sieving.
Since we found the last prime the following amount of factors were found:

Total factors returned: 41343

When we look at our next target to test all pairs below n=4M the following number of factors are found:

In total we found 3562 factor between 1006 and 4000000
679 factors removed a test before a first PRP test was done
2399 factors removed a pair after the first test but before a second test
484 factors were found for pairs which are already doublechecked.


All times are UTC. The time now is 11:00.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, Jelsoft Enterprises Ltd.