mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   Computer Science & Computational Number Theory (https://www.mersenneforum.org/forumdisplay.php?f=116)
-   -   Wilson-prime search practicalities (https://www.mersenneforum.org/showthread.php?t=16028)

MrRepunit 2011-10-07 22:27

I finished my range 300e9-340e9, did not find any new near Wilson prime. Reserving now the smaller range 340e9-350e9, hoping that the newer version will be ready soon. By the way, is there an update of the expected release date?

MrRepunit 2011-10-12 09:23

I finished my range 340e9 to 350e9, no Near-Wilson prime was found.
Also reserving 350e9 to 360e9.

Still no update on the expected release of the new version?

R. Gerbicz 2011-10-12 10:58

[QUOTE=MrRepunit;274220]
Still no update on the expected release of the new version?[/QUOTE]

There is no update. Though there are working seperate new parts of the code in different languages (Pari-gp, Sage), but these will be in c.

MrRepunit 2011-10-16 18:43

Finished the range 350e9 to 360e9, nothing found. I will continue the search on the range 360e9 to 400e9.

Francisco 2011-10-28 14:55

Hi. Let me introduce myself.

My name is Francisco, I'm working in the Ibercivis project, a BOINC volunteer computing platform. Danilo (hi Danilo) told me about your project.

As maybe you know, there is a subproject in Ibercivis searching wilson primes, but as Danilo knows our program is slower (a lot slower) than yours.

Danilo proposed me to combine our efforts, and that sounds very good to me, so...what do you think...it's possible to use your code in the Ibercivis project? If we do that, I can make you an account in order to send to the Ibercivis platforms the interval to search.

Best Regards.

Francisco.

jasonp 2011-10-29 12:44

The code used here is much faster because it uses sophisticated batch methods, and for the larger ranges requires many gigabytes of memory. The experience I've had with NFS@Home says that your participants may not want to run a distributed project that requires all the memory on their computers.

Not to speak for anyone else, just something to think about.

MrRepunit 2011-10-29 14:53

[QUOTE=jasonp;276230]The code used here is much faster because it uses sophisticated batch methods, and for the larger ranges requires many gigabytes of memory. The experience I've had with NFS@Home says that your participants may not want to run a distributed project that requires all the memory on their computers.

Not to speak for anyone else, just something to think about.[/QUOTE]

Well, the high memory usage is optional, even if one would use less it would be much faster than the normal iteration as done in the current Boinc project.

R. Gerbicz 2011-10-30 11:14

[QUOTE=Francisco;276115]Hi. Let me introduce myself.

My name is Francisco, I'm working in the Ibercivis project, a BOINC volunteer computing platform. Danilo (hi Danilo) told me about your project.

As maybe you know, there is a subproject in Ibercivis searching wilson primes, but as Danilo knows our program is slower (a lot slower) than yours.

Danilo proposed me to combine our efforts, and that sounds very good to me, so...what do you think...it's possible to use your code in the Ibercivis project? If we do that, I can make you an account in order to send to the Ibercivis platforms the interval to search.

Best Regards.

Francisco.[/QUOTE]

For the new code (still not finished) without asking some questions from users it would be very hard to make it efficient and userfreindly. For a modification to Boinc we need to know:

1. How many cores to run the code? Note that all cores are working on the same interval, and it will be much more efficient than running interval/cores on each core. In fact as I have see from different Boinc searches that they use all available cores, and I can do this also. But it is not needed.

2. How much memory would you give for the code? If you give it,say in MBytes, then the code would compute the size of "interval" that still fits in that.

(for non boinc code I'll skip the step 2, it is much better to give directly the interval)

R.D. Silverman 2011-10-31 13:14

[QUOTE=R. Gerbicz;276306]For the new code (still not finished) without asking some questions from users it would be very hard to make it efficient and userfreindly. For a modification to Boinc we need to know:

1. How many cores to run the code? Note that all cores are working on the same interval, and it will be much more efficient than running interval/cores on each core. In fact as I have see from different Boinc searches that they use all available cores, and I can do this also. But it is not needed.

2. How much memory would you give for the code? If you give it,say in MBytes, then the code would compute the size of "interval" that still fits in that.

(for non boinc code I'll skip the step 2, it is much better to give directly the interval)[/QUOTE]

May I suggest to everyone that this project should not expect to find
any new Wilson primes. I will not label the search as pointless, but everyone
needs to realize how [b][i]extremely[/i][/b] rare such primes will be.
Up to 10^100 we should expect only to find about 5 such primes, and
up to 10^20 there should only be 3 to 4. Up to N, we should only
expect ~loglog N such primes. This function grows very slowly.

If people want to do something [i]productive[/i], I suggest joining the
Sierpinski project. It, at least, has a definite conclusion that can be reached.

wblipp 2011-10-31 18:53

[QUOTE=R.D. Silverman;276453]If people want to do something [i]productive[/i], I suggest joining the Sierpinski project. It, at least, has a definite conclusion that can be reached.[/QUOTE]

:surprised Who are you and what have done with the real RDS? The one that thinks everybody should be working on the Cunningham project, but only with factoring code they have written themselves?

xilman 2011-10-31 20:39

[QUOTE=R.D. Silverman;276453]May I suggest to everyone that this project should not expect to find any new Wilson primes. I will not label the search as pointless, but everyone needs to realize how [b][i]extremely[/i][/b] rare such primes will be. Up to 10^100 we should expect only to find about 5 such primes, and up to 10^20 there should only be 3 to 4. Up to N, we should only expect ~loglog N such primes. This function grows very slowly.

If people want to do something [i]productive[/i], I suggest joining the Sierpinski project. It, at least, has a definite conclusion that can be reached.[/QUOTE]If you don't look, you won't find.

However, if you do look, you probably still won't find. C'est la vie.

Paul


All times are UTC. The time now is 11:37.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2022, Jelsoft Enterprises Ltd.