Why does my work not appear in CPUs, but does on Assignments?

[drkirkby]

I'm doing PRP testing of a large exponent

https://www.mersenne.org/report_expo...2646233&full=1

This shows up on my Exponent Status as 2.2% done, and updated 2021-02-21 (i.e. today). I just done a manual upload of latest expected completion data. My prime.log file shows an estimated completion of 25th May.

However, this is not seen in my CPUs. There's nothing to indicate computer "jackdaw" is working on this - see screen shots attached. I can't understand why one part of the Primenet server is showing the information, and another is not.

My prime.log file shows the data is being uploaded to the server. I've not dropped the CPU at any time.

[Comm thread Feb 20 21:16] Sending expected completion date for M332646233: May 25 2021
[Comm thread Feb 20 21:16] URL: http://v5.mersenne.org/v5server/?v=0...ACFD29ABCD0B02
[Comm thread Feb 20 21:16] RESPONSE:
[Comm thread Feb 20 21:16] pnErrorResult=0
[Comm thread Feb 20 21:16] pnErrorDetail=SUCCESS
[Comm thread Feb 20 21:16] ==END==

[Viliam Furik]

If you look at the list again, for example now, you should see it there again. It was not showing, because the CPU was doing some other task at the moment - PRP certification. These are randomly assigned to users who opt-in (or more specifically do not opt-out) for doing the certification and are prioritized. As it was a 10M exponent cofactor result, it took a very short amount of time.

[Uncwilly]

Try going to: https://www.mersenne.org/workload/
Do you see the exponent in question there?

[drkirkby]

Originally Posted by Uncwilly

Try going to: https://www.mersenne.org/workload/
Do you see the exponent in question there?

Yes, it shows at



- see attached.



It still does not show at https://www.mersenne.org/cpus/ but what I notice is that https://www.mersenne.org/cpus/ shows a certificate of 10309619 having used over 400 GHz days, and 0% completed. That seems a bit unlikely.



Dave

[Uncwilly]

Originally Posted by drkirkby

It still does not show at https://www.mersenne.org/cpus/ but what I notice is that https://www.mersenne.org/cpus/ shows a certificate of 10309619 having used over 400 GHz days, and 0% completed. That seems a bit unlikely.

That is the total output of the CPU, not for that assignment.

I edited your attachment to hid your assignment ID's,

[drkirkby]

Originally Posted by Uncwilly

Try going to: https://www.mersenne.org/workload/
Do you see the exponent in question there?

As I wrote before, I do see it there, but it does not seem to be updating. It still says 0.9% done, estimated 94 days to go, finishing on 2021-05-27. Yet the output of mprime shows something quite different.

[Work thread Feb 22 10:03] Setting affinity to run helper thread 9 on CPU core #11
[Work thread Feb 22 10:03] Resuming Gerbicz error-checking PRP test of M332646233 using AVX-512 FFT length 18M, Pass1=1152, Pass2=16K, clm=2, 25 threads
[Work thread Feb 22 10:03] PRP proof using power=10 and 64-bit hash size.
[Work thread Feb 22 10:03] Proof requires 42.6GB of temporary disk space and uploading a 457MB proof file.
[Work thread Feb 22 10:03] Iteration: 12413044 / 332646233 [3.73%].
[Work thread Feb 22 10:05] Iteration: 12420000 / 332646233 [3.73%], ms/iter: 18.029, ETA: 66d 19:42
[Work thread Feb 22 10:08] Iteration: 12430000 / 332646233 [3.73%], ms/iter: 18.020, ETA: 66d 18:52
[Work thread Feb 22 10:11] Iteration: 12440000 / 332646233 [3.73%], ms/iter: 17.960, ETA: 66d 13:29
[Work thread Feb 22 10:14] Iteration: 12450000 / 332646233 [3.74%], ms/iter: 18.000, ETA: 66d 16:59
[Work thread Feb 22 10:17] Iteration: 12460000 / 332646233 [3.74%], ms/iter: 17.830, ETA: 66d 01:46
[Work thread Feb 22 10:20] Iteration: 12470000 / 332646233 [3.74%], ms/iter: 17.793, ETA: 65d 22:25
[Work thread Feb 22 10:23] Iteration: 12480000 / 332646233 [3.75%], ms/iter: 17.793, ETA: 65d 22:26
[Work thread Feb 22 10:26] Iteration: 12490000 / 332646233 [3.75%], ms/iter: 17.786, ETA: 65d 21:42
[Work thread Feb 22 10:29] Iteration: 12500000 / 332646233 [3.75%], ms/iter: 17.765, ETA: 65d 19:47
[Work thread Feb 22 10:32] Iteration: 12510000 / 332646233 [3.76%], ms/iter: 17.738, ETA: 65d 17:20
[Work thread Feb 22 10:35] Iteration: 12520000 / 332646233 [3.76%], ms/iter: 17.894, ETA: 66d 07:13
[Work thread Feb 22 10:38] Iteration: 12530000 / 332646233 [3.76%], ms/iter: 17.988, ETA: 66d 15:30
[Work thread Feb 22 10:41] Iteration: 12540000 / 332646233 [3.76%], ms/iter: 17.825, ETA: 66d 00:56
[Work thread Feb 22 10:44] Iteration: 12550000 / 332646233 [3.77%], ms/iter: 17.777, ETA: 65d 20:39
[Work thread Feb 22 10:47] Iteration: 12560000 / 332646233 [3.77%], ms/iter: 17.936, ETA: 66d 10:44

There's obviously a fairly big difference between 0.9%/94 days and 3.76%/66 days.

So there seems two issues

1) Not updating at https://www.mersenne.org/workload/
2) Not shown at https://www.mersenne.org/cpus/

worktodo.txt had two lines in it - the big PRP test followed by a certificate. I stopped mprime, swapped the order of those, and re-started it. The certificate was quickly shown as completed.

https://www.mersenne.org/report_expo...0309619&full=1

but when mprime finished that, and started back on the more CPU intensive task, it still does not update on the GIMPS website.

I may have "confused" the server, as I did unassign one exponent from mprime, which still showed as assigned to me on the website. I then unassigned it on the website.

I'm obviously not keen to waste 64 days computing something, if it is no use to man nor beast.


If the worst comes to the worst, could this manually be updated later? Or would I just be better to stop the process and start something else? If I do that latter, I think I will take on something less ambitious!.


Dave

[drkirkby]

Okay, things have changed. I decided to start a smaller exponent



https://www.mersenne.org/report_expo...9841707&full=1


which will take about 8 days, so 10% of the time. Now



https://www.mersenne.org/workload/


shows sensible data for the large exponent. Now it is showing it is 3.80% done, which is about right. It also has extended the end data a bit, as I'm working on the smaller exponent. But https://www.mersenne.org/workload/
looks a lot more sensible than it did an hour or two ago.

[Uncwilly]

Originally Posted by drkirkby

As I wrote before, I do see it there, but it does not seem to be updating. It still says 0.9% done, estimated 94 days to go, finishing on 2021-05-27.

How often is your machine set to send estimated completion dates to PrimeNet. It should only do so once a day or so. Even less is ok if you are working on numbers up near 332M.

1) Not updating at https://www.mersenne.org/workload/

That is what I addressed in the lines above.

2) Not shown at https://www.mersenne.org/cpus/

Not really the place to check for your current assignment progress.

worktodo.txt had two lines in it - the big PRP test followed by a certificate. I stopped mprime, swapped the order of those, and re-started it. The certificate was quickly shown as completed.

That had no actual effect. The Cert assignment is a "priority" assignment. Once mprime notices it in the queue, it starts that and will finish it ahead of other assignments.

I may have "confused" the server, as I did unassign one exponent from mprime, which still showed as assigned to me on the website. I then unassigned it on the website.

If you manually pulled it out of the worktodo file, mprime won't know to tell PrimeNet about it.

I'm obviously not keen to waste 64 days computing something, if it is no use to man nor beast.

Think of it as "cycles". There is a more or less set amount of compute that is needed to test that number. If you don't test that number, someone else will. If it is up in the 332.19M range, then it is a 100M decimal digit number. If it is prime, then it is an award winner.

If the worst comes to the worst, could this manually be updated later? Or would I just be better to stop the process and start something else?

If you have the save files on your machine still, you could resume it. What you do is your own choice.

I would suggest that you try to ask more of your questions in a single thread. That will help the people that answer you see the full scope of your issues.