GPU id/name for manual results

[preda]

Would it be useful to include a GPU name (id) with manual results? If the server would like that, I could add it to gpuOwL result format.

The rationale is that some user can have multiple GPUs. If some results later prove bad, it may be easier to pin-point the "bad GPU" if an id is attached. (Similar to the CPU-name for mprime)

[LaurV]

We like it.

You can PM James with a link to this thread. He is a busy guy and doesn't read all forums, contrarily do kibitzes like ourselves (royal we), but he is the guy who has to make the parser, so if he likes it, it is good to go.

[James Heinrich]

I just happened upon this thread by chance.
The manual results parser should already accept user/computer as long as it's in the standard format:

UID: <username>(/<compname>),
e.g.:
UID: JamesHeinrich/GTX1080, <result line here as normal>

"UID: " and username are required, slash and gpuname are optional, prepended to your normal result line separated by one comma and one space as shown.

[preda]

Cool, I'll try this, thanks!

[Madpoo]

Originally Posted by preda

Cool, I'll try this, thanks!

As a side note, they'll still be credited to the "manual computing" cpu in the user's account, but the text info will be logged to help in identification later. And maybe at some point the manual result check in will use this info and credit the correct computer for that user.

[storm5510]

Originally Posted by Madpoo

... And maybe at some point the manual result check in will use this info and credit the correct computer for that user.

I would really like to see this happen. As you know, mfaktc has this option, but CUDALucas does not. Having a bit of programming background myself, I would not think it would be terribly difficult to add the userid/computerid and have it output in the results.txt file.

[kriesel]

Originally Posted by James Heinrich

I just happened upon this thread by chance.
The manual results parser should already accept user/computer as long as it's in the standard format:"UID: " and username are required, slash and gpuname are optional, prepended to your normal result line separated by one comma and one space as shown.

Some gpu model names contain embedded spaces. Examples: GTX1050 Ti; GTX1060 3GB. Those spaces can be stripped if they interfere with parsing. Do they?

Since more than one gpu per model and per system are possible, and being able to trace errors back to an individual gpu card will be useful, I suggest some sort of concatenation of system and gpu allowing unique identification be supported.
Something along the lines of
UID:username/systemname-gpumodelletter so that in the case where one username has multiple GTX1080 for example per box,
UID:JamesHeinrich/systema-GTX1080A
UID:JamesHeinrich/systema-GTX1080B
UID:JamesHeinrich/systema-GTX1080C
UID:JamesHeinrich/systemb-GTX1080D
etc.
(I think that might be more useful than username/GTX1080D for systemb, because it might be the system motherboard memory not the gpu that makes one case unreliable.Moving the gpu to different system hardware is a different configuration that may have different reliability. Maybe best practice is to uniquely label the system and gpu separately)

Being able to get results reports with a selector for a specific system name from PrimeNet, with separate counts for unverified, verified, and known-bad and %bad of total known-status (verified plus known-bad) for each would be very useful.

Are there limits in PrimeNet to the incoming results record length, or individual fields of it, that this might exceed? Any limits on separator choice between system and gpu?

I'm adding UID/systemid-gpuid to the bug and feature lists I'm keeping, beginning with cudalucas and cudapm1. Cudapm1 is derived from cudalucas so doing both should be straightforward.

[GP2]

Originally Posted by kriesel

Some gpu model names contain embedded spaces. Examples: GTX1050 Ti; GTX1060 3GB. Those spaces can be stripped if they interfere with parsing. Do they?

In the database there are numerous examples of the Computer name field containing embedded spaces, for instance:

https://mersenne.org/report_ll/?exp_...exfactor=1&B1=

Same goes for the User name field.

[preda]

gpuOwL now has -uid, which takes a string given by the user that will be put as UID: in results. Thus nothing automatic, the user decides the content.

[James Heinrich]

Originally Posted by kriesel

Some gpu model names contain embedded spaces. Examples: GTX1050 Ti; GTX1060 3GB. Those spaces can be stripped if they interfere with parsing. Do they?

Are there limits in PrimeNet to the incoming results record length, or individual fields of it, that this might exceed? Any limits on separator choice between system and gpu?

UID:<user>/<machine> parsing for all manual results looks like this:

^(UID: ([a-z0-9_\.\+\- ]+)(/([a-z0-9_\.\+\- @'/]+))?, )?

So you can have alphanumeric plus _ . + - [space] for username, and the same plus @ ' / for machinename.

There is of course a maximum field length for both user and machine names, but I'm not sure what they are. You can ask George or Aaron if it's important.

[kriesel]

Originally Posted by James Heinrich

UID:<user>/<machine> parsing for all manual results looks like this:

^(UID: ([a-z0-9_\.\+\- ]+)(/([a-z0-9_\.\+\- @'/]+))?, )?

So you can have alphanumeric plus _ . + - [space] for username, and the same plus @ ' / for machinename.

There is of course a maximum field length for both user and machine names, but I'm not sure what they are. You can ask George or Aaron if it's important.

Excellent succinct detailed reply. Thanks! :]
It even answers questions I hadn't thought to ask then, such as "could we wrap a system and gpu id combination in quotes." That would cause what looks like a perl pattern match to fail on the system and gpu id combo. There may be very good reasons for excluding some characters from the permitted pattern.

Will inquire about lengths, since I consider it important not to cause PrimeNet problems, or cause problems with my result submissions, and for any coder to know what the constraints are.