Mini-drive for k=3010-3200 n=420259-600000

[gd_barnes]

Originally Posted by mdettweiler

Gary currently verifies all results on all drives in periodic blocks--when he posts in the drive threads saying "all results and primes have been verified up to n=x" that means he's done it for that range.

No!! That is incorrect. In a PM exchange with what I thought to be both you and Karsten, Karsten is now verifying all server results vs. the sieve files. I was duplicating his results matchup and it took me too long anyway. I only verify primes in the various places and have found no problems in a very long time. I suspect Karsten verifies both results vs. the original sieve file and found primes in post 1 of the drive threads, on top 5000, his own rieselprime page, and his own drive page. But since he is doing the last two items solely on his own, it's best to have another person doublecheck them. That's the only reason I do any kind of verification at all.

Here is what I think we should do based on what you guys are saying:

1. Let the server continue to offload them to the results page.

2. Max, plop all manual results into the results page within 1-2 days after they are done.

3. Karsten, when LLRnet/PRPnet -AND- all manual ranges have been completed up to some set limit, then process all of the results from the results page up to that limit and verify the results against the original sieve file. Since all server and manually-done results should be there, there should be no missing pairs for any range. If some manual results are missing, then it will be up to Max to find them and load them to the page.

Does that sound like the way to go?

Max, this page you're talking about where the sieve is automatically matched up against some processed results sounds like a lot of work and would require a LOT Of testing to make sure it is correct 100% of the time. I highly recommend against it until you are able to become less busy in your personal life. What we are talking about is right now so let's stick with talking about that.


Gary

[kar_bon]

As input for my processing/verifying script I use the normal LLRnet result file (3 lines for one pair). I think I can easily edit this to input normal LLR-resultsfiles, too (one line forn one pair).
So I only need all resultfiles available like those from the servers!

[mdettweiler]

Originally Posted by gd_barnes

No!! That is incorrect. In a PM exchange with what I thought to be both you and Karsten, Karsten is now verifying all server results vs. the sieve files. I was duplicating his results matchup and it took me too long anyway. I only verify primes in the various places and have found no problems in a very long time. I suspect Karsten verifies both results vs. the original sieve file and found primes in post 1 of the drive threads, on top 5000, his own rieselprime page, and his own drive page. But since he is doing the last two items solely on his own, it's best to have another person doublecheck them. That's the only reason I do any kind of verification at all.

Ah, I see. I don't believe I ever got a PM to that effect, but that sounds fine. As long as somebody's doing the verification, we're OK.

Here is what I think we should do based on what you guys are saying:

1. Let the server continue to offload them to the results page.

2. Max, plop all manual results into the results page within 1-2 days after they are done.

3. Karsten, when LLRnet/PRPnet -AND- all manual ranges have been completed up to some set limit, then process all of the results from the results page up to that limit and verify the results against the original sieve file. Since all server and manually-done results should be there, there should be no missing pairs for any range. If some manual results are missing, then it will be up to Max to find them and load them to the page.

Does that sound like the way to go?

Okay, that should work. Sounds like a plan. Most of the active drives are being done exclusively via servers at this point, so manual ranges should be far and few enough that the plan you outlined would indeed make things somewhat simpler.

Max, this page you're talking about where the sieve is automatically matched up against some processed results sounds like a lot of work and would require a LOT Of testing to make sure it is correct 100% of the time. I highly recommend against it until you are able to become less busy in your personal life. What we are talking about is right now so let's stick with talking about that.

Yeah, I definitely wasn't thinking of doing that right away. My mention of that was just me rambling about possibly overambitious ideas for the future.

[gamer007]

519K-520K is complete. No primes.

Reserving 520K-521K.

[Svenie25]

516K-517K is done, no primes, rsults emailed to Gary. The lower file will be done today or until tomorrow morning.

Just in time before going to take a cure for at least 3 weeks. ;)

[kar_bon]

@Jonathan (gamer007)
Please check your result-file for the range n=512k-513k:
There're 5 pairs missing (from end of sieve-file), namely

3091 512999
3123 512999
3129 512999
3105 513000
3119 513000

@Max
The resultfile for range n=506k-507k contains twice the pair 3117*2^506005-1!
Please delete one of them.

@Max,Gary
By now I've not checked the manually results, because I got not all of them.
The above error was very obvious and I don't know if there're more errors!
Please check those result-files.

@all
I've updated the Mini Drive #3 page at rieselprime.de.

Karsten

[mdettweiler]

Originally Posted by kar_bon

@Max
The resultfile for range n=506k-507k contains twice the pair 3117*2^506005-1!
Please delete one of them.

Hmm...strange. I don't see that in my copy. Here's what I have for n=506005:

3019*2^506005-1 is not prime.  LLR Res64: 318C7157BE6FB2AB  Time : 426.101 sec.
3025*2^506005-1 is not prime.  LLR Res64: 14C6F5CF76B5653B  Time : 430.757 sec.
3099*2^506005-1 is not prime.  LLR Res64: A7D070F2E8ED25C1  Time : 421.261 sec.
3117*2^506005-1 is not prime.  LLR Res64: A894E68E2E5A83EA  Time : 407.875 sec.
3129*2^506005-1 is not prime.  LLR Res64: 1BF46C44FC32AC12  Time : 404.264 sec.
3139*2^506005-1 is not prime.  LLR Res64: 85AEC41B034E4C72  Time : 387.427 sec.

Thus, it would seem that somewhere along the line (Gary?) the duplicate was removed prior to my receiving the results.

[kar_bon]

Originally Posted by mdettweiler

Thus, it would seem that somewhere along the line (Gary?) the duplicate was removed prior to my receiving the results.

Download from post #222!

[mdettweiler]

Originally Posted by kar_bon

Download from post #222!

Yes, I tried that and verified that the duplicate is present in the file attached there. However, the file I have (which was provided to me by Gary as an aggregate 500K-510K file) does not have the duplicate, so I must assume it was removed somewhere along the line.

[gamer007]

Originally Posted by kar_bon

@Jonathan (gamer007)
Please check your result-file for the range n=312k-313k:
There're 5 pairs missing (from end of sieve-file), namely

3091 512999
3123 512999
3129 512999
3105 513000
3119 513000

Hmm.. not sure what happened there. Probably missed them while I was dividing the work between my cores. Here's the updated results.

[kar_bon]

Thanks! I've updated post #222 with the correct result-file (doubled result for pair 3117*2^506005-1).