CRUS LLRnet server discussion

[em99010pepe]

That'll also be a good test for the internet connection....I'm gonna add my quad-core to llrnet...lol

[mdettweiler]

Originally Posted by em99010pepe

Anon,

As tnerual said I need to stop the server before adding more work. At current speed the server will dry in less than 4 hours. (I added a fast machine)

Should I add your file? BTW, that file only has 30 numbers....I think it's a bit low....

Yes, please add the file. It only has 30 numbers, yes, but that still should be adequate for a quick little doublecheck.

[mdettweiler]

Originally Posted by kar_bon

found a test file in the old 15k-forum for testing llr version 3 which had problems with small k/n pairs. this testfile contains about 10.000 candidates from (204307539,3291) to (2232007,1490605)!
all candidates should be primes!

Carlos, can you add this, too? You can add it immediately below the 30 candidates I posted, when you edit the knpairs.txt file.

Thanks!

[mdettweiler]

Originally Posted by em99010pepe

That'll also be a good test for the internet connection....I'm gonna add my quad-core to llrnet...lol

With your quad-core, we should have the remaining numbers cleaned up in no time!

I'll reserve the Sierpinski work for you to add after the server's run dry (i.e. after it's completed both the original Riesel work and the doublecheck work).

[mdettweiler]

Okay, here's the HUGE file with all the Sierp. Base 16 candidates in it.

Carlos, after the server goes dry (i.e. all knpairs, including the doublechecks, are handed out and residuals returned), copy off the results.txt file, and mail it to me. (You already have my email address.) The results file will include both the Riesel candidates and the doublecheck work; I'll separate those myself, then generate normal-style lresults files (as well as some cool stats that I've got in mind!) from the Riesel candidates and submit the lresults to Gary. After you've copied off the results.txt file, insert the Sierpinski file, and start it up again!

[mdettweiler]

Originally Posted by tnerual

first close the server ... then add the numbers then restart the server ...

if you don't do that, the server will erase the new numbers ...

for info: i think that the residues for k*2^4n-1 and for k*16^n-1 are not the same ... even if llr (classic one) reduce k*16^n-1 to k*2^4n-1

Yes, that's because the old LLR doesn't reduce base 16 to base 2, and does a PRP test on it (which is not compatible with LLR residual-wise). The new LLR, however, reduces to base 2 and does an LLR test on the base 2 number. However, we've effectively mimiced that function for our LLRnet by pre-reducing the file to base 2--so LLRnet treats it as any old base 2 file. Thus, residuals for a base 16 number auto-reduced to base 2, and the same number manually reduced to base 2 before feeding it in to LLR, are compatible.

I wrote a little Perl script to do the job of converting a base 16 file to base 2--after I've made some refinements to it, I'll post it here.

[em99010pepe]

Sierp. Base 16 - PORT 444

Server: CRUS.dynip.telepac.pt
Port: 444

[mdettweiler]

Originally Posted by em99010pepe

Sierp. Base 16 - PORT 444

Server: CRUS.dynip.telepac.pt
Port: 444

Oh, okay. Now that I think about it, a separate server probably is the best way to ensure a smooth transition.

Okay, here's what you should do: After the Riesel server has run dry and you've shut it off, shut down the Sierp. server, set it to port 443, and start it up again. I don't know why I didn't think of that before!

[mdettweiler]

BTW Carlos, did you get to insert the doublecheck work for the Riesel server yet?

[em99010pepe]

Originally Posted by Anonymous

Oh, okay. Now that I think about it, a separate server probably is the best way to ensure a smooth transition.

Okay, here's what you should do: After the Riesel server has run dry and you've shut it off, shut down the Sierp. server, set it to port 443, and start it up again. I don't know why I didn't think of that before!

I prefer to have two, one for Riesel and one Sierp...

Originally Posted by Anonymous

BTW Carlos, did you get to insert the doublecheck work for the Riesel server yet?

Yep...

[mdettweiler]

Originally Posted by em99010pepe

I prefer to have two, one for Riesel and one Sierp...



Yep...

Well, I guess we could do that...but since at the moment, Sierp. is the higher priority (it's getting less attention from the manual LLRers), we'll want that one on port 443, since port 443 can be accessed through corporate firewalls usually. Then the Riesel one can go on port 444.

As for keeping the Riesel one busy: How about you let it run dry first, after inserting the doublecheck work, then after the residuals have been verified, you can insert more Riesel work that I'll be giving you. What do you think?