Manual GPU assigments for Trial Factoring
 

Hello all my name is erick84mm, I have been crunching numbers as anonymous since 2012 and in 2017 I decided to join with my tag erick84mm. This is my second post ("Hello!")

Right now I'm trying to collaborate with Trial Factoring for 100M digit Mersenne numbers since I saw GIMPS is collecting funds and this can reduce the number of LL tests. I have a GPU to do that. It takes around 30 min per Mp in the ranges of 75-76.

I tried to obtain 1000 thousand of those exponents since is the work for the month.

However, 2 things happened:

1. Instead of getting the exponents for 100M digits I got 1000 ECM exponents. (I did this twice), I manually deleted all those ECM exponents since I'm running mfaktc and I cannot do ECM.

2. I decreased the number of requested exponents to 10 and I was able to get 5 at first and 2 afterward. Now I cannot get any exponent anymore. The error is the following.

Error code: 40
Error text: No assignment available for GPU trial factoring, cpu_id: #####, user_id #####



So I was wondering, am I doing something wrong? How can I obtain the exponents in that range? Should I just focus on doing the exponents of lower ranges as 180.000.000?

Thank you very much for your answers in advance
erick84mm

[QUOTE=erick84mm;517347]Right now I'm trying to collaborate with Trial Factoring for 100M digit Mersenne numbers since I saw GIMPS is collecting funds and this can reduce the number of LL tests. I have a GPU to do that. It takes around 30 min per Mp in the ranges of 75-76.

I tried to obtain 1000 thousand of those exponents since is the work for the month. [/QUOTE]
Welcome to the forum. You should use GPUto72 ( GPU72.com ). It will allow you to select exponents up in that range. Set-up over there and if you select LMH from the assignment screen you can get exponents up there. I would suggest that with a GPU you take the exponents up to higher levels than 76. Ideally you take the exponents up to 80 or 81. (see this chart [url]https://www.mersenne.ca/cudalucas.php?model=9[/url]). You can do what you want, but that would get the exponents ready for the P-1 and primality testing.

[QUOTE=erick84mm;517347]
I tried to obtain 1000 thousand of those exponents since is the work for the month. [/QUOTE]

Try [url]https://www.mersenne.org/manual_gpu_assignment/[/url]

[QUOTE=Uncwilly;517353]You should use GPUto72 ( GPU72.com ). It will allow you to select exponents up in that range.[/QUOTE]

Actually, no. Use the link Prime95 provided to get the assignments directly from Primenet.

There has been little uptake on the assignments from GPU72. While current assignments will be observed and honored, no further assignments in 332M are available.

[QUOTE=Prime95;517415]Try [url]https://www.mersenne.org/manual_gpu_assignment/[/url][/QUOTE]

That is the part that I mention is not working.

When I request exponents in the GPU form this come up
Error code: 40
Error text: No assignment available for GPU trial factoring, cpu_id: xxxxx, user_id = xxxxx

I have attached the images of the error.

[QUOTE=erick84mm;517432]That is the part that I mention is not working.

When I request exponents in the GPU form this come up
Error code: 40
Error text: No assignment available for GPU trial factoring, cpu_id: xxxxx, user_id = xxxxx

I have attached the images of the error.[/QUOTE]

You'll need to enter an optional exponent range -- something like 332400000 to 332600000

Now is working! Thank you! My confusion was due to the fact that I always get the normal CPU exponents just by selecting trial factoring and it provides an answer even when I don't set any interval. However, here is not the case! Thank you very much! Now I'm crunching numbers in the 100M interval! :D

[QUOTE=erick84mm;517461]Now is working! Thank you! My confusion was due to the fact that I always get the normal CPU exponents just by selecting trial factoring and it provides an answer even when I don't set any interval. However, here is not the case! Thank you very much! Now I'm crunching numbers in the 100M interval! :D[/QUOTE]
Yes, that is a confusing aspect of the gpu_assignment page.
Either some default values, or explanatory text, would be a helpful addition on that page.
It does not seem very optional, if the page generates an error without a minimum and maximum exponent specified manually.
Why not have the page provide defaults, that correspond to the category selected, eg

332192807 for minimum and ~360000000 for maximum for 100Mdigit,
93000000 for minimum and ~190000000 for first time LLs?
Including the header link on [url]https://www.mersenne.org/manual_assignment/[/url] to the gpu_assignment page and vice versa was a welcome recent improvement.

[QUOTE=kriesel;517485]Why not have the page provide defaults, that correspond to the category selected[/QUOTE]It's already roughed in, waiting for George's approval.

[QUOTE=kriesel;517330]That's interesting.

This is a PRP DC delivered by the Primenet API;
PRP=(aid),1,2,79335979,-1,75,0,3,[B]1[/B] in the worktodo file,
PRP=(aid),1,2,79335979,-1,75,0,3,[B]4[/B] in the All cpus and system-specific listings on my assignments page

PRP first test assignments do not list a residue type in the worktodo line on the assignments page, ending in a comma.
PRP=(aid),1,2,exponent,-1,75,0,3,[/QUOTE]

As best I can tell, the PRP worktodo line looks like:
PRP=[assignment id],k,b,exponent,c,TF bits,prp base,prp type,[# of known factors]

The assignment id is optional, as is the # of known factors on the end (for cofactor PRP testing)

The base will either match any previous test, or 3 if no other PRP results are recorded. The type will also match any previous test, but if no previous test will done, the /workload page actually leaves it empty. I don't know if that's by design (so the client will decide what's appropriate based on what version of the program it is) or if we actually should be defaulting it to type=4.

At the time this was all setup not too long ago, gpuowl and prime95 were, I think, using different bases and types so maybe that's why no default was picked there? But now, don't they both use base=3 and type=4 ? Or something like that?

EDIT: by the way, for an exponent like the one mentioned that has *two* previous results with different types/bases, it pulls the values from the most recent one.

[QUOTE=Madpoo;517491]As best I can tell, the PRP worktodo line looks like:
PRP=[assignment id],k,b,exponent,c,TF bits,prp base,prp type,[# of known factors][/QUOTE]

The [c],75,0,3,[/c] has a zero in the middle, which is the number of tests saved. This parameter only makes sense for P−1 assignments, and is just a placeholder here, so it can always be set to zero.


Also, I'm not sure if I'm correctly interpreting what kriesel wrote, but it seems like the worktodo file and the All CPUs section of [url]https://www.mersenne.org/workload/[/url] had conflicting information for the residue type for the [I]same[/I] assignment. So the All CPUs section should instead accurately reflect the actual residue type for the assignment.