mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   Conjectures 'R Us (https://www.mersenneforum.org/forumdisplay.php?f=81)
-   -   Looks like PrimeGrid's noticed us.... (https://www.mersenneforum.org/showthread.php?t=10056)

mdettweiler 2008-03-05 18:23

Looks like PrimeGrid's noticed us....
 
Check this out:

[url]http://www.primegrid.com/forum_thread.php?id=862[/url]

Look at the bottom of the first post. :smile:

em99010pepe 2008-03-05 18:46

I have one prime found with CRUS and if I find more I won't share with PrimeGrid, just in case they decide to jump in.

mdettweiler 2008-03-05 18:50

[quote=em99010pepe;127849]I have one prime found with CRUS and if I find more I won't share with PrimeGrid, just in case they decide to jump in.[/quote]
Yes, of course--when they help out projects, they only require that they share credit for primes found through their system. Primes found through manual reservations or any other non-PrimeGrid system (such as LLRnet) are still reported without PrimeGrid. :smile:

gd_barnes 2008-03-06 03:25

I could potentially welcome PrimeGrid in the future when we get past n=~2M base 2 on SOME powers-of-2 bases and past ~250000 digits on many other bases. Sierp base 4 is a possibility now but I wouldn't consider giving it up with so few k's remaining. Since the Sierp base 4 project put the Sierp base 4 k's near n=2 million, I'd like to see us take it to at least to 3 million before considering outside help.

The proof of some of the conjectures, even for bases <= 32 will last 100's/1000's of years even assuming doubling of computer speeds every 18 months-2 years, which I think is unlikely to continue indefinitely. Some may have a prime for their final k far in excess of n=1 billion; some may be in excess of n=1 trillion. Heck, Riesel Sieve may last 100's of years all by itself. Hence monsterous DC efforts will be needed on them.

That is unless we can come up with some grand new mathematical method for testing them or narrowing down where primes might be much more likely.


Gary

mdettweiler 2008-03-06 06:04

[quote=gd_barnes;127883]I could potentially welcome PrimeGrid in the future when we get past n=~2M base 2 on SOME powers-of-2 bases and past ~250000 digits on many other bases. Sierp base 4 is a possibility now but I wouldn't consider giving it up with so few k's remaining. Since the Sierp base 4 project put the Sierp base 4 k's near n=2 million, I'd like to see us take it to at least to 3 million before considering outside help.

The proof of some of the conjectures, even for bases <= 32 will last 100's/1000's of years even assuming doubling of computer speeds every 18 months-2 years, which I think is unlikely to continue indefinitely. Some may have a prime for their final k far in excess of n=1 billion; some may be in excess of n=1 trillion. Heck, Riesel Sieve may last 100's of years all by itself. Hence monsterous DC efforts will be needed on them.

That is unless we can come up with some grand new mathematical method for testing them or narrowing down where primes might be much more likely.


Gary[/quote]
Yeah, I see what you mean. Hey you know, maybe PrimeGrid could be really helpful for the bases with huge conjectures, such as base 3 and base 7. :smile:

tnerual 2008-03-06 10:10

[QUOTE=Anonymous;127894]Yeah, I see what you mean. Hey you know, maybe PrimeGrid could be really helpful for the bases with huge conjectures, such as base 3 and base 7. :smile:[/QUOTE]

not really for those base ... the problem with those base is that you have to remove a lot of k with small n ... and it's time consuming ... (not per test but due to to the large amount of test)

for my k*31^n+1 i'm at n=5000 and one test take only 9 sec on my centrino duo 1600 ...
but i only have 3000 k left (starting with 6M) ...

with boinc, you can't send workunit that take less than 10 seconds (if you don't want to crash your server ... and to spare bandwith) so we have to "manually" test every base up to n=5000 before sending them to boinc (and that's with k=31, not with k=3) ...

boinc processing is there to help kill high n primes (one test evry 2 day is not a problem with a lot of contributors, you keep a good pace of testing) or to do boring task as double check ('cause people don't care about primes, they care about boinc points)

when i will reach n= 10000 on my k, i will release it and send all files to Gary, and i will start killing those stupid base by 1 million k chunks ... (will plan that with Gary when reaching my goal)

mdettweiler 2008-03-06 15:53

[quote=tnerual;127910]not really for those base ... the problem with those base is that you have to remove a lot of k with small n ... and it's time consuming ... (not per test but due to to the large amount of test)

for my k*31^n+1 i'm at n=5000 and one test take only 9 sec on my centrino duo 1600 ...
but i only have 3000 k left (starting with 6M) ...

with boinc, you can't send workunit that take less than 10 seconds (if you don't want to crash your server ... and to spare bandwith) so we have to "manually" test every base up to n=5000 before sending them to boinc (and that's with k=31, not with k=3) ...

boinc processing is there to help kill high n primes (one test evry 2 day is not a problem with a lot of contributors, you keep a good pace of testing) or to do boring task as double check ('cause people don't care about primes, they care about boinc points)

when i will reach n= 10000 on my k, i will release it and send all files to Gary, and i will start killing those stupid base by 1 million k chunks ... (will plan that with Gary when reaching my goal)[/quote]
Ah, I see. I was thinking about when those pesky bases got to a higher n-level, but then, you're right, most of the work on those probably will be at lower n-levels where lots of k's are being eliminated. :smile:

gd_barnes 2008-03-06 17:33

[quote=tnerual;127910]and i will start killing those stupid base by 1 million k chunks ... (will plan that with Gary when reaching my goal)[/quote]


I have a little 'secret' project I've been working on for ~2-3 weeks on a single core of my work laptop.

I have already been searching Sierp base 3. Base 3 is a VERY prolific prime base, which is one reason that the conjectures are so high. k's are dropping in rapid fashion at low n-levels. But with more than 10^15 of them, it might take a little while! :smile:

Bases 7 and 15, I have no desire to tackle those for now. They take far longer to test at same n-level. You can have those if you want them. I agree that they must be tackled in k=1M chunks. Although base 3 on 2-3 high-speed cores could easily be tackled in k=10M chunks.

You know what I think the most difficult base to eliminate k's on is? Base 25! Look at how tough base 5 for that project! The conjecture on Riesel base 5 and 25 is the same and for Sierp, it's much higher on base 25! Base 25 will be UGLY!! I think k's will drop much faster on bases 3, 7, and 15 and even with more than 10^14 k's on all, I think they will be more fun to hack away at.

I'll report my status on Sierp base 3 when I reach k=2M. I'll be there this weekend. You'll be amazed by the few # of k's remaining. I'll then update the web pages to reflect the info.


Gary


All times are UTC. The time now is 23:28.

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