P+1 suggested work units

[Uncwilly]

Until the PrimeNet server is set-up to assign P+1, can someone who understands what work is desirable to be down post suggested Worktodo lines? Especially if they can be grouped in some way, like "quick and easy for Prime95 testing", "in depth assignments", "insane assignments to test the system."

I would be willing to help keep the list updated for the adds and claims (like over in the strategic dc thread.) I just have no idea what exponents and bounds make sense.

[ATH]

We have been discussing this here, at least for <1M:
https://mersenneforum.org/showthread.php?t=22476

I'm getting the stats for >1M and I can generate worktodo lines for exponents with sensible bounds. But I'm not sure who wants to run these. The "<20M unfactored" group is already working their own chosen ranges.

[firejuggler]

more specifically, this post
https://mersenneforum.org/showpost.p...&postcount=443

[Uncwilly]

Originally Posted by ATH

But I'm not sure who wants to run these. The "<20M unfactored" group is already working their own chosen ranges.

I would like to run some. I ran some real rando tests. But I am throwing nearly random guesses at things. For example I ran 1010407 Start=2/7, B1=10000000, B2=4000000000
And I have the first 10M decimal digit with no known factor queued up with similar bounds. I would be will to run some in the <20M range. But, I would not know what is reasonable to try. Firejuggler's link does not help me either. Don't we have to specify a B2?

[masser]

Originally Posted by Uncwilly

I would like to run some. I ran some real rando tests. But I am throwing nearly random guesses at things. For example I ran 1010407 Start=2/7, B1=10000000, B2=4000000000
And I have the first 10M decimal digit with no known factor queued up with similar bounds. I would be will to run some in the <20M range. But, I would not know what is reasonable to try. Firejuggler's link does not help me either. Don't we have to specify a B2?

I have attached two worktodo files for work in the <20M range that will help my current effort. The first file uses the first starting seed and has B1,B2 specified. The second file uses the second starting seed and has B1,B2 specified, but because the sieve depth (in TF bits) is also included, B2 will be ignored. In this case, the code will select a B2 that makes sense given available memory.

[chalsall]

Originally Posted by masser

I have attached two worktodo files for work in the <20M range that will help my current effort.

I've put the first file's entries onto my six-core machine.

[masser]

Originally Posted by chalsall

I've put the first file's entries onto my six-core machine.

Each task is about 35 minutes on my four-core machine.

Let me know if you or anyone else wants more. I can't promise that they'll be perfectly optimal, but I think I can develop tasks that are useful for the <20M project.

[firejuggler]

Another datapoint
2.8811 GHz-day
A typical PM1 for me has a B1 or 1.2M and a B2 of 76.8M

[Uncwilly]

I took the second set.

The entry I had made up was estimated to finish in September.

[ATH]

You should specify the trial factor level in the worktodo line, then it ignores your B2 and calculates the optimal.

Instead of:
Pplus1=1,2,14090003,-1,500000,10000000,1

then

Pplus1=1,2,14090003,-1,500000,10000000,1,72

[masser]

Originally Posted by ATH

You should specify the trial factor level in the worktodo line, then it ignores your B2 and calculates the optimal.

Instead of:
Pplus1=1,2,14090003,-1,500000,10000000,1

then

Pplus1=1,2,14090003,-1,500000,10000000,1,72

I will definitely do this if there are requests for more files.

Is there demand for more files? I have a few ideas for shorter tasks that would still be useful for the <20M project.