mfaktc: a CUDA program for Mersenne prefactoring

[flashjh]

Originally Posted by James Heinrich

That's a server-side issue that is as-yet unaddressed. Doesn't matter what format you submit for manual results, Primenet will (currently) ignore the differences and process it however it sees fit. Which is usually right, but not always.

As for the GHz days, it's wrong, but it's not that big of a deal. However, it messes up the statistical results on your website which i think are really important for historical purposes. Do you think the data will get fixed in the future or is it permanent? I have kept an archive of all my results files, maybe one day they can be used to fix the data in PrimeNet?

Edit: I was just reading here that you have to upload results to James' site. I did not know it was separtate. I'll have to spend some time uploading my files.

I still wonder if it's possible to ever fix PrimeNet's data for all out GPU TFs that show as P-1?

[James Heinrich]

Originally Posted by flashjh

However, it messes up the statistical results on your website which i think are really important for historical purposes. Do you think the data will get fixed in the future or is it permanent? I have kept an archive of all my results files, maybe one day they can be used to fix the data in PrimeNet?

Any data that is uploaded directly to my site should be accurately credited for both work type and GHz-days. If it's not, let me know.

Primenet's interpretation of manual results remains an issue for now, but one that I can hopefully help address within the next few weeks/months. Where that affects my data is when I spider new results from PrimeNet; if these are mis-reported then I will import bad data. If you also upload your real results.txt data to my site then it's much easier to find invalid TF-PM1 confused results and potentially fix them.

I still wonder if it's possible to ever fix PrimeNet's data for all out GPU TFs that show as P-1?

Unfortunately I'm not convinced that Primenet retains detailed logs for factoring assignments (just increments factor-type counter for the user, and records the factor for the exponent). But it still should be possible to later go through the data and look for improbable P-1s (where it would be hard to find said factor with P-1, but relatively easy with TF). But I need to fix Primenet's poor interpretation of manual results first.

I'll have to spend some time uploading my files.

Shouldn't take that long to upload. Bundle up all your results into one or two files and upload away. Results files with PrimeNet-style user/computer IDs at the beginning of the line are automatically handled correctly; for mfakt* results which have no user/computer data you should specify your username and "computer" ID in the appropriate boxes. If you work with multiple GPUs, upload data for one GPU at a time to avoid confusion. Other than that there's no real limit on upload size (my server parses results much faster than primenet ).

[James Heinrich]

Feature request for TheJudger / Bdot:
Can you put two optional fields in mfaktc.ini for username and computerid, so that if the user fills them out then it puts the Primenet-style "UID: username/computername, " at the beginning of results lines? Makes parsing results much easier and less subject to confusion.

[flashjh]

I have quite a few TF and TF DC results that were reported to PrimeNet but they are still showing up in my assignments page on GPU to 72. It's been over 4 hours since the report to PrimeNet. Will they eventually get detected or is something wrong?

[kladner]

I have gotten all my Windows Prime95-64bit and mfaktc results submitted. EDIT: To http://mersenne-aries.sili.net

I'm not sure how to submit results from a defunct 32bit Prime95 setup.

Then there's a headless Linux box. I'm still really inept in this area, in that I don't know any easy way to copy that results.txt to Windows for submission. I attempted to extract the results from the PrimeNet Results page, but that is not correctly formatted and so was not recognized.

I'm accessing the Linux system via putty. Any guidance in getting those results would be appreciated.

EDIT2: Perhaps this query would have been better off posted here: http://www.mersenneforum.org/showthr...=11101&page=42
but at this point I'll leave well enough alone.

[Dubslow]

Originally Posted by kladner

Then there's a headless Linux box. I'm still really inept in this area, in that I don't know any easy way to copy that results.txt to Windows for submission.

I'm accessing the Linux system via putty. Any guidance in getting those results would be appreciated.

It's possible something like scp might work. I don't know exactly how to use it, but it should get you started. 'man scp'

Originally Posted by flashjh

I have quite a few TF and TF DC results that were reported to PrimeNet but they are still showing up in my assignments page on GPU to 72. It's been over 4 hours since the report to PrimeNet. Will they eventually get detected or is something wrong?

Put this in the GPU272 forum, in the 'Assignments discrepancy' thread.

[kladner]

Originally Posted by Dubslow

It's possible something like scp might work. I don't know exactly how to use it, but it should get you started. 'man scp'

Thanks, Dubslow. I'll have a look. Just having a name is a big step forward.

[chalsall]

Originally Posted by flashjh

I have quite a few TF and TF DC results that were reported to PrimeNet but they are still showing up in my assignments page on GPU to 72. It's been over 4 hours since the report to PrimeNet. Will they eventually get detected or is something wrong?

Sorry...

I had turned off the Observation spider while I drilled down on the assignment dependency problem. I figured out what was going on at about 0500 my time, but was so tired I forgot to turn it back on...

I'll have a full report over in the G72 sub-forum in a few hours (or maybe tomorrow -- it is Old Years Night, after all... ).

[Bdot]

Originally Posted by James Heinrich

Feature request for TheJudger / Bdot:
Can you put two optional fields in mfaktc.ini for username and computerid, so that if the user fills them out then it puts the Primenet-style "UID: username/computername, " at the beginning of results lines? Makes parsing results much easier and less subject to confusion.

I've added it to my list ...

[James Heinrich]

Originally Posted by James Heinrich

Feature request for TheJudger / Bdot:
Can you put two optional fields in mfaktc.ini for username and computerid, so that if the user fills them out then it puts the Primenet-style "UID: username/computername, " at the beginning of results lines? Makes parsing results much easier and less subject to confusion.

Second similar feature request:
Can you please also add a .ini option to output datestamp lines in results.txt, the same way Prime95 does? e.g.:

[Sat Dec 31 21:29:40 2011]
no factor for M50370833 from 2^70 to 2^71 [mfaktc 0.18 barrett79_mul32]

[TheJudger]

Originally Posted by James Heinrich

Feature request for TheJudger / Bdot:
Can you put two optional fields in mfaktc.ini for username and computerid, so that if the user fills them out then it puts the Primenet-style "UID: username/computername, " at the beginning of results lines? Makes parsing results much easier and less subject to confusion.

Shouldn't be a problem. We'll need those two entries in mfaktc.ini for automated primenet interaction, too.

Originally Posted by James Heinrich

Second similar feature request:
Can you please also add a .ini option to output datestamp lines in results.txt, the same way Prime95 does? e.g.:

Do we need this as an option or should we enable this allways?

Oliver