Double Check Server
 

Just thinking about a double check server. People can point their computers to it and help eliminate some candidates. Who knows that maybe we missed a prime. Also it will be helpful for the future. Is anyone willing to host it? Is having it a good idea/ bad idea? What do you all think?

Double check for the low ranges is a problem to be done with llrnet as the residues are not comparable.

There is an effort for double checking with mystwalker and me participating.
At the moment the error rates are so low that i don't see the need for a bigger effort.
Most of the errors are found in the ranges that were run with the llr version that had an error. But even there most of the residues are correct.

Lars

How big is the error rate? How many residues did not match etc?

Here is a breakdown of the different numbers of tests done so far.

We have:
118165 first pass PRP test
22797 second pass test
59 third pass tests due to not fitting results

These 59 tests split into the following groups:

3 tests come from double reservations where one of the tests is run with
PRP and the other with llr version > 3.0 Therefore there were incompatible residues.

50 tests are from the range of tests where results came back from a buggy llr client

6 tests are real errors.

Lars

Does someone have the Linux version of PRP handy? Seems like mersenne.org is down again. :sad:

Lars:
Which doublechecks do you prefer?
<400k or the 700k - 900k?

btw.:
Is there a problem when I use a newer PRP version (currently still 3.1.0)? If no: "prp2414.zip" (resp. .tar.gz) is the newest one, right?

I had no problem to reach: [url]http://www.mersenne.org/gimps/[/url]

If you still have problems to get access tell me and i will download the file for you.

All PRP versions produce compatible residues so there is no problem with using prp2414.

If it is no problem for you please test the 700k-900k block first as this is the block with the residues that have the possible llr bug.

Lars

[QUOTE=ltd]I had no problem to reach: [url]http://www.mersenne.org/gimps/[/url][/quote]
It works again now...

Unfortunately, the Linux systems don't have libstdc++ 6 installed. Do you know if there's a statically linked version available?

[quote]If it is no problem for you please test the 700k-900k block first as this is the block with the residues that have the possible llr bug.[/QUOTE]

No problem.

Try sprp2414.tar.gz from the same side.

This should be the statical linked version.

Lars

Could you post the real 6 errors you mentioned and the residues, and who did them. I just want to play and see why the error was generated.

edit: How many numbers are left in the double check range from 700K to 950K due to llr bug? If there is a LLR double check server set up then I can put some computers on this server (instead of first check) and speed up the double check process. We don't want to miss a prime like SOB.

For 5 of these errors i allready talked to the contributors and it turned out to be two errors due to temperature problmes and three to a RAM problem.
The error i did not really investigate was k=90527, n=143351.
And guess who did send the bad result in. His username starts with a C and he is the founder of PSP. :wink:

The double checks for the bad llr client are all oin the file mw719559.txt which is reserved by Mystwalker. Remember that the llr client with the error was llr version 2.xx which had PRP compatible residues. So again there is the problem with incompatible residues if we would hand those WU out via llrnet.

Lars

ok
His username starts with a C and he is the founder of PSP. Could I have the residues for this number? :smile:

Secondly, are we only double checking till 900K or till where do you estimate the bug might have affected the numbers.


Citrix