OFFICIAL "SERVER PROBLEMS" THREAD

[nitf]

Originally Posted by James Heinrich

I assume this is a fairly large proof file. Can you please PM or email me a link to the proof file and I'll investigate further.

It was fairly large - 462MB - but apparently it managed to make it through with repeated automated retries. I should have waited one more iteration.

[James Heinrich]

Originally Posted by nitf

It was fairly large - 462MB - but apparently it managed to make it through with repeated automated retries. I should have waited one more iteration.

This is excellent -- you have tested (unintentionally) that the error-checking upload process is working correctly, and even though you had multiple upload errors it detected and retried them and finally succeeded.

It does seem quite a slow upload though -- is your internet connection slow in general, or do you have upload speed limited in Prime95?

[nitf]

Originally Posted by James Heinrich

It does seem quite a slow upload though -- is your internet connection slow in general, or do you have upload speed limited in Prime95?

Unbeknownst to me until now, it was throttling upload speed under Options/Resource Limits.

[kriesel]

Originally Posted by nitf

It was fairly large - 462MB

473041738 bytes per your screen capture, or ~461,955 KiB. The difference between that and what I posted earlier (473,041,680) is for a handful of scalars in addition to the 10 full size residues. The header info is typically ~60 bytes.

[slandrum]

This time I did the PRP, then the proof was automatically uploaded (about an hour later) and the CERT was run successfully on it. However the server says that it's PRP w/o proof running on "old software". While it's not the latest greatest version of Prime95, it's 30.7 which should not have any difference at all in PRP proof generation.

[James Heinrich]

Originally Posted by slandrum

the server says that it's PRP w/o proof running on "old software"

Thanks, fixed.

[masser]

Does the server accept P+1, stage-1-only, results? I tried to submit a file containing lines like this:

{"status":"NF", "exponent":581227, "worktype":"P+1", "b1":4000000, "start":"127/61", "fft-length":28672, "security-code":"XXXXXXXX", "program":{"name":"Prime95", "version":"30.9", "build":1, "port":8}, "timestamp":"2022-11-28 23:43:38"}

but they weren't accepted.

EDIT: The actual message from the manual submission page was "Did not understand 52 lines." with a box fill of the lines that were unknown.

[James Heinrich]

Originally Posted by masser

Does the server accept P+1, stage-1-only, results?

No. B1-only results are only accepted if a factor is found in stage-1.

[chalsall]

Originally Posted by James Heinrich

No. B1-only results are only accepted if a factor is found in stage-1.

Dare I ask why this knowledge is not accepted? DoS seems unlikely.

[James Heinrich]

Originally Posted by chalsall

Dare I ask why this knowledge is not accepted? DoS seems unlikely.

I'm not sure if it's a decision or an omission. I'll ask George for clarification when he gets back from vacation.
Is there a good reason why someone choosing to do such a specific worktype would choose to stop halfway and not do stage-2?

[masser]

On my 4-core cpus (haswell to coffee lake), stage 1 is faster with 4 workers (1 core per worker). Stage 2 is more convenient with 1 worker (dedicating most of system memory to stage 2). That's the reason for splitting. I don't have to report the intermediate stage 1 results, but I like to be thorough. When the server didn't accept the result, I thought perhaps a bug (or unconsidered case) had been found. (I think results from similar workflows are accepted for P-1, but I'll have to doublecheck. Yes, see here.)