Manual GPU assignments
 

Would it be possible for an advanced version of:
[url]https://www.mersenne.org/manual_gpu_assignment/[/url]

Where you just input exact exponent range and bit level range and it only checks if it is unassigned and if the bit levels are the lowest available, and then gives us those assignments we requested that are matching those 2 requirements maybe up to a limit of X exponents or X Ghz-Days depending on what is safe for the server to handle at once.

[QUOTE=ATH;532737]Would it be possible for an advanced version of:
[url]https://www.mersenne.org/manual_gpu_assignment/[/url]

Where you just input exact exponent range and bit level range and it only checks if it is unassigned and if the bit levels are the lowest available, and then gives us those assignments we requested that are matching those 2 requirements maybe up to a limit of X exponents or X Ghz-Days depending on what is safe for the server to handle at once.[/QUOTE]

A bit like this page?

[url]https://www.mersenne.org/report_factoring_effort/?exp_lo=200000000&exp_hi=200099999[/url]

That is only the report, you will not get the actual assignments.

It should be like the current version: [url]https://www.mersenne.org/manual_gpu_assignment/[/url]
but more precise. Give you exactly what you want (if unassigned and lowest bitlevels).

[QUOTE=ATH;532778]That is only the report, you will not get the actual assignments.

It should be like the current version: [url]https://www.mersenne.org/manual_gpu_assignment/[/url]
but more precise. Give you exactly what you want (if unassigned and lowest bitlevels).[/QUOTE]

Sure you can :-)

See attached

[QUOTE=Gordon;532786]Sure you can :-)

See attached[/QUOTE]
It gives a list of stuff, yes, but does not assign them on the Primenet server. That's the point.

[QUOTE=nomead;532787]It gives a list of stuff, yes, but does not assign them on the Primenet server. That's the point.[/QUOTE]

Remove the "N/A," put the lines in your worktodo.txt in Prime95, do a manual communication session with the "Send new expected completion dates to server" checked. You will now have assignment ID's.

[QUOTE=Gordon;532786]Sure you can :-)

See attached[/QUOTE]

Yes, I know, that is exactly what I'm doing now, but the problem is they are not assigned to me on the server, so sometimes you manage to collide with someone else doing the same range.

I know I can do a manual communication in Prime95, but it is very slow if you want to do lots of low bit level assignments, but I guess that might also be too stressful anyway on an "updated" GPU assignment page. But another thing the current GPU page won't let me do is pick a single exponent and reserve it up to a high bit depth.


I just felt that there are room for improvement in the current manual GPU assignment page, but maybe it is too much effort for the reward.

[QUOTE=ATH;532798]But another thing the current GPU page won't let me do is pick a single exponent and reserve it up to a high bit depth.[/QUOTE]
1) reserve an exponent of list of exponents for TF to the next bit level.
2) edit your copy of the reservations to raise the stop bit level to the level you want to complete. (maybe by a perl or python script)
3) run all the bit levels of an exponent before reporting completion of any bit levels for it
4) The reservation will persist until you report a TF result, it expires, or someone poaches it.

Or,
use the optional bit level to factor to field of [URL]https://www.mersenne.org/manual_gpu_assignment/[/URL]

[QUOTE=nomead;532787]It gives a list of stuff, yes, but does not assign them on the Primenet server. That's the point.[/QUOTE]

Does that matter? You can check the results in anyway...

[QUOTE=ATH;532798]Yes, I know, that is exactly what I'm doing now, but the problem is they are not assigned to me on the server, so sometimes you manage to collide with someone else doing the same range.

I know I can do a manual communication in Prime95, but it is very slow if you want to do lots of low bit level assignments, but I guess that might also be too stressful anyway on an "updated" GPU assignment page. But another thing the current GPU page won't let me do is pick a single exponent and reserve it up to a high bit depth.


I just felt that there are room for improvement in the current manual GPU assignment page, but maybe it is too much effort for the reward.[/QUOTE]

So you sometimes get a collision, with a fast enough GPU it might mean you sometimes throw away 15-20 minutes of computation.

I'm currently plowing through the sub 1M exponents, in the 200k range it's about 24 mins per bit level...on a gtx-1660

[QUOTE=Gordon;532878]So you sometimes get a collision, with a fast enough GPU it might mean you sometimes throw away 15-20 minutes of computation.

I'm currently plowing through the sub 1M exponents, in the 200k range it's about 24 mins per bit level...on a gtx-1660[/QUOTE]For the other person whose work yours collides with, perhaps on a slower gpu or cpu, it may mean a day or more wasted.

[URL]https://www.mersenne.org/legal/#TCU[/URL] implies there's a sort of property right to an assignment, by using words like "your assignment". Performing work on tasks assigned to someone else is referred to as "poaching" or "stealing". It's generally discouraged.
It's of more concern at and above the GIMPS wavefront, where there may be a substantial prize award at stake.

Participation constitutes agreement to the terms of use.