mprime moves to another number and sits on the results?

[henstepl]

I played with settings and I might have gotten myself a pickle. My 8-core computer has two workers running, and they've completed a few PRP proofs, so I know what PRP completion looks like.


But most recently, Worker #2 went all the way to 100% and then... sat on its results, and began another? Did I have an error? Did I do something wrong?


Here is my mprime -s

[Main thread Nov 25 14:17] Mersenne number primality test program version 30.3
[Main thread Nov 25 14:17] Optimizing for CPU architecture: AMD Bulldozer, L2 cache size: 4x2 MB, L3 cache size: 8 MB
Below is a report on the work you have queued and any expected completion
dates.
[Worker thread #1]
M114952363, PRP, Sat Jan 22 10:20 2022
[Worker thread #2]
M111513377, PRP, Fri Nov 26 00:20 2021
M113287753, PRP, Sun Jan 23 23:09 2022
The chance that one of the 3 exponents you are testing will yield a
Mersenne prime is about 1 in 289094.

Here is my worktodo.txt

[Worker #1]
PRP=AssignmentID_redacted_by_moderator,1,2,114952363,-1,76,2

[Worker #2]
PRP=AssignmentID_redacted_by_moderator,1,2,111513377,-1,77,2
PRP=AssignmentID_redacted_by_moderator,1,2,113287753,-1,76,0

The one ending in 3377 ought to have done something by now. But the exponent check https://www.mersenne.org/report_exponent/?exp_lo=111513377&full=1 seems like "no dice", and mprime moves on to 7753, and nothing seems to have happened with my valuable work. The estimated completion date in mprime -s (Fri Nov 26 00:20 2021) just keeps moving back.


Someone tell me what to do! I put a lot of CPU days into this.

[Uncwilly]

111513377 shows as being in Stage2 of the P-1 test. P-1 is a way of finding factors. This is done before the PRP to hopefully avoid it. And by what you posted, it shows that it is predicted to complete on the 26th.

What settings did you change recently? If you changed upload permissions it may take longer to submit the file. If you changed the "proof power" it can change the size of the file to upload.

I removed your Assignment ID's (they should not be shared).
The [M][/M] tags will link to the exponent page like I did above. And the [QUOTE][/QUOTE] tags make for nice formatting of text blocks.

[kruoli]

Originally Posted by Uncwilly

And the [­QUOTE][­/QUOTE] tags make for nice formatting of text blocks.

You might be confusing this with [C] or [CODE].

[henstepl]

Originally Posted by Uncwilly

111513377 shows as being in Stage2 of the P-1 test. P-1 is a way of finding factors. This is done before the PRP to hopefully avoid it. And by what you posted, it shows that it is predicted to complete on the 26th.

What settings did you change recently? If you changed upload permissions it may take longer to submit the file. If you changed the "proof power" it can change the size of the file to upload.

I removed your Assignment ID's (they should not be shared).
The [M][/M] tags will link to the exponent page like I did above. And the tags make for nice formatting of text blocks.

Thank you, and I have formatted my post better now. I've seen P-1s occur but I hope you can believe me: it did a whoooooole PRP test, and I was watching it go to 100%, but it never did what it did at other tests' 100% marks. The predicted time just keeps pushing back...


Forgive me if I lose track of what settings I've changed. I have a forgetful brain. This is my current prime.txt

V24OptionsConverted=1
WorkPreference=0
V30OptionsConverted=1
WGUID_version=2
StressTester=0
UsePrimenet=1
DialUp=0
V5UserID=ANONYMOUS
Priority=1
DaysOfWork=3
RunOnBattery=1
PRPGerbiczCompareIntervalAdj=1


[PrimeNet]
Debug=2
ProxyHost=
UploadRateLimit=0.25
UploadStartTime=00:00
UploadEndTime=24:00
DownloadDailyLimit=40

And when I ctrl-C to stop and restart, this is an abridgement of what happens:

[Main thread Nov 25 15:21] Starting workers.
...
[Worker #1 Nov 25 15:21] Optimal P-1 factoring of M114952363 using up to 307MB of memory.
...
[Worker #2 Nov 25 15:21] Optimal P-1 factoring of M111513377 using up to 307MB of memory.
...
[Worker #1 Nov 25 15:21] Other workers are using lots of memory now.
[Worker #1 Nov 25 15:21] Looking for work that uses less memory.
[Worker #2 Nov 25 15:21] Other workers are using lots of memory now.
[Worker #2 Nov 25 15:22] Looking for work that uses less memory.
[Worker #1 Nov 25 15:22] Resuming Gerbicz error-checking PRP test of M114952363 using FFT length 6M, Pass1=1K, Pass2=6K, clm=4, 4 threads
[Worker #1 Nov 25 15:22] PRP proof using power=8 and 64-bit hash size.
[Worker #1 Nov 25 15:22] Proof requires 3.7GB of temporary disk space and uploading a 129MB proof file.
[Worker #1 Nov 25 15:22] Iteration: 12759549 / 114952363 [11.09%].
[Worker #2 Nov 25 15:22] Resuming Gerbicz error-checking PRP test of M113287753 using FFT length 6M, Pass1=1K, Pass2=6K, clm=4, 4 threads
[Worker #2 Nov 25 15:22] PRP proof using power=8 and 64-bit hash size.
[Worker #2 Nov 25 15:22] Proof requires 3.6GB of temporary disk space and uploading a 127MB proof file.
[Worker #2 Nov 25 15:22] Iteration: 9123294 / 113287753 [8.05%].

So it starts two P-1s and immediately says something about memory, and then goes into two PRPs (only one of which matches the P-1)... I have 16GB DDR3 and I rebooted to close everything, isn't that enough?

[Happy5214]

Originally Posted by henstepl

So it starts two P-1s and immediately says something about memory, and then goes into two PRPs (only one of which matches the P-1)... I have 16GB DDR3 and I rebooted to close everything, isn't that enough?

mprime sets its own maximum RAM limit to a uselessly low level to avoid any surprises. Run mprime with the -m flag, select option 14 (Options/Resource Limits), and raise the disk (for the PRP proofs) and RAM limits to suitable values for your computer.

[Uncwilly]

Originally Posted by kruoli

You might be confusing this with [C] or [CODE].

Yes, slip of the mouse. Tags are ignored inside of them. Example: [STRIKE][/STRIKE]

[kruoli]

You can also "escape" tags by adding a non-printable character (Windows) after the opening bracket: hold Alt, then press 2, then 4, then 0, necessarily on the number block and not the number row, then release Alt. E.g. [­QUOTE][­/QUOTE].

[S485122]

Originally Posted by kruoli

You can also "escape" tags by adding a non-printable character (Windows) after the opening bracket: hold Alt, then press 2, then 4, then 0, necessarily on the number block and not the number row, then release Alt. E.g. [­QUOTE][­/QUOTE].

If you want tags not to be parsed enclose them in the [noparse] and [/noparse] tags.

[henstepl]

Originally Posted by Happy5214

mprime sets its own maximum RAM limit to a uselessly low level to avoid any surprises. Run mprime with the -m flag, select option 14 (Options/Resource Limits), and raise the disk (for the PRP proofs) and RAM limits to suitable values for your computer.

Yes, I will probably do that. As it is, I managed to force it back into play by reducing my number of workers to one. It did its P-1 and now it's at 99.61% on the PRP. Thanks friends!