mersenneforum.org  

Go Back   mersenneforum.org > Great Internet Mersenne Prime Search > PrimeNet

Reply
 
Thread Tools
Old 2021-10-21, 02:00   #1
techn1ciaN
 
techn1ciaN's Avatar
 
Oct 2021
U.S. / Maine

9310 Posts
Default Checked out exponents for DC-via-PRP; PrimeNet thinks they're cat 0 and will unreserve in a week

I recently discovered that GPUOwl works on my laptop's integrated graphics and decided to get it doing a bit of DC. I recalled reading on here that DC via PRP is slightly preferable to LL DC because it theoretically avoids an occasional triple-check, so I went to manual assignments and checked out two exponents with the "Double check using PRP with proof" preference.


According to https://www.mersenne.org/thresholds/ the exponents I checked out (M63608191 & M63608201) are solidly in DC cat 4. But, my https://www.mersenne.org/workload/ reports that they're both in cat 0 and going to expire in six days. GPUOwl gives an ETA of 10 days and change for one exponent, so I couldn't meet that date even if I only checked out one at a time and left my laptop running 24/7.


Obviously, PrimeNet seems to think I've checked out for first-time PRP and is applying the category thresholds for that. But — is there any known fix for this, or do I just have to unreserve these exponents, download an older GPUOwl, and do old-fashioned LL DC?
techn1ciaN is offline   Reply With Quote
Old 2021-10-21, 02:40   #2
slandrum
 
Jan 2021
California

24610 Posts
Default

You reserved them as PRP, not as PRP-DC (which is a recently added type).

The 6 day timeout is false. You did a manual reservation, and the "work must be started in X days" rule doesn't apply. Because you aren't supposed to get a manual reservation for cat-0 or cat-1 the server glitches on reporting that expiration, but once it passes you'll suddenly find you have 23 more days before the real expiration. That's just a bug in the server reporting that timeout.

Once you see the real expiration, you can extend it.

You can also reserve the exponent for one type of work, edit your worktodo and actually do a different type of work on it. There are a number of people who do that.

Last fiddled with by slandrum on 2021-10-21 at 02:44
slandrum is offline   Reply With Quote
Old 2021-10-21, 03:08   #3
techn1ciaN
 
techn1ciaN's Avatar
 
Oct 2021
U.S. / Maine

3·31 Posts
Default

Thank you for the information. Good to know it's just a reporting bug and the exponents will not actually be unreserved. I'll resume running GPUOwl.



Quote:
You reserved them as PRP, not as PRP-DC

To be clear, I selected "Double-check using PRP with proof"; both of the exponents I checked out have LL first tests, which would not be possible if I had selected PRP-DC. I elected not to use PRP-DC because I figured that DC-via-PRP would give me relatively smaller exponents that my IGP could work in a more reasonable time.
techn1ciaN is offline   Reply With Quote
Old 2021-10-21, 13:13   #4
kriesel
 
kriesel's Avatar
 
"TF79LL86GIMPS96gpu17"
Mar 2017
US midwest

10111000111102 Posts
Default

Welcome to the forum.
Sounds like a server problem. There's an existing thread for that.
Previous posts there about that long-known issue include:
https://mersenneforum.org/showpost.p...postcount=2249
https://mersenneforum.org/showpost.p...postcount=2179
https://mersenneforum.org/showpost.p...postcount=2184
https://mersenneforum.org/showpost.p...postcount=2164
https://mersenneforum.org/showpost.p...postcount=2159 (including list of previous posts about the issue, back to March 2019) and some discussion near some of those.
Please use existing threads when applicable.

See also the reference info collection.

Last fiddled with by kriesel on 2021-10-21 at 13:17
kriesel is offline   Reply With Quote
Old 2021-10-21, 13:34   #5
techn1ciaN
 
techn1ciaN's Avatar
 
Oct 2021
U.S. / Maine

5D16 Posts
Default

Hello kriesel.


I'm confused; I read the posts you linked and they all seem to deal specifically with PRP-DC, not DC-via-PRP. In fact, https://mersenneforum.org/showpost.p...postcount=2249 even says that "only exponents checked out as PRP DC for exponents with LL first tests already completed ... get correctly indicated as PRP-D currently" — doesn't that mean I should not be having this problem, since I checked out that work type?



I attempted to search for an existing thread regarding this issue, but all of my search queries pulled up only PRP DC–related content and I was (incorrectly, so apologies) assuming that this issue was limited to the PRP-via-DC preference, since it reports under https://www.mersenne.org/workload/ as PRP and not PRP-D.
techn1ciaN is offline   Reply With Quote
Old 2021-10-21, 13:49   #6
Uncwilly
6809 > 6502
 
Uncwilly's Avatar
 
"""""""""""""""""""
Aug 2003
101×103 Posts

2×3×5×337 Posts
Default

Quote:
Originally Posted by techn1ciaN View Post
I'm confused; I read the posts you linked and they all seem to deal specifically with PRP-DC, not DC-via-PRP.
It doesn't make a difference. PRP-CF get treated like Cat 0 too. The server sees all PRP assignments below 106521590 (as of the moment) as Cat 0. It does not see the PRP-CF as a different class of PRP, nor does it see any DC via PRP as anything but Cat 0. There will have to be some reworking of the code to fix this.

Since you got it as a manual assignment, you might try to manually extend it. The other thing to try is to use a copy of Prime95 to get the assignments as regular LL-DC and then move them over to the GPU. Then just turn in the result.
Uncwilly is offline   Reply With Quote
Old 2021-10-21, 13:57   #7
techn1ciaN
 
techn1ciaN's Avatar
 
Oct 2021
U.S. / Maine

1358 Posts
Default

Quote:
There will have to be some reworking of the code to fix this.

I see; interesting. Do you know where specifically the inability to differentiate arises? I'm assuming it's to do in some way with the fact that GIMPS just didn't have PRP tests for much of its history, but don't have a more informed guess than that.


Quote:
...you might try to manually extend it.

I did go to my extensions page and do this to see what would happen. Interestingly, on the extensions page, both exponents correctly reported as 179 days to go (before I extended them), which seems to track with slandrum's answer that this is just a reporting problem and will not get the exponent unreserved. They changed to 239 days post-extension, on that page, but remained 6 days on https://www.mersenne.org/workload/.
techn1ciaN is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Not-Checked or -DoubleChecked Exponents Mario23 PrimeNet 6 2014-03-19 03:04
How so I unreserve orphaned exponents SSteve PrimeNet 11 2011-03-02 22:04
factored or double checked exponents assigned S485122 PrimeNet 1 2007-11-19 12:59
All exponents below 13,000,000 double-checked. Aillas Lounge 5 2006-06-30 07:27
22.01M's and 66 bit non-PrimeNet exponents trif Lone Mersenne Hunters 2 2003-08-02 18:25

All times are UTC. The time now is 10:21.


Wed Dec 1 10:21:06 UTC 2021 up 131 days, 4:50, 1 user, load averages: 1.12, 1.12, 1.15

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, Jelsoft Enterprises Ltd.

This forum has received and complied with 0 (zero) government requests for information.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.
A copy of the license is included in the FAQ.