Please rank Assignments by time/importance

[Yura]

Could somebody please rank below type of work based on:
- time required to finish single assignment
- importance of assignment for the entire project

World record tests
Smallest available first-time tests
Double-check tests
P-1 factoring
Trial factoring
ECM factoring
100 million digit tests

[petrw1]

Originally Posted by Yura

Could somebody please rank below type of work based on:
- time required to finish single assignment
- importance of assignment for the entire project

World record tests
Smallest available first-time tests
Double-check tests
P-1 factoring
Trial factoring
ECM factoring
100 million digit tests

Time required: (I shall assume a mainstream PC like a dual core 3 Ghz with adequate RAM for P1 and ECM)

TF: anywhere from minutes for high exponents at low bit levels to 1 day for the lower exponents to higher bit levels
ECM: Hours for ECM on small Mersennes; days to a couple weeks for larger Fermat numbers.
P-1: about 2 days in the current assigned ranges.
Double Check: about 7-10 days in the current ranges
Smallest LL and World Record LL are both in about the same ranges now: If you time it right and with a little luck you can still get exponents as low as 39M for Smallest and they must be at least 43M for Record but most LL assignment now are in the low 50 Millions: 40M would take about 3 weeks; 50M about 5 weeks.
100M digit: more than 1 year to complete.

[petrw1]

Originally Posted by Yura

Could somebody please rank below type of work based on:
- time required to finish single assignment
- importance of assignment for the entire project

World record tests
Smallest available first-time tests
Double-check tests
P-1 factoring
Trial factoring
ECM factoring
100 million digit tests

Importance is harder to define because there are several opinions on this one:
Some say "LL": The ultimate goal of this project is to find Mersenne Primes and only LL tests can do that.

Some say "DC": A MP needs to be double checked (usually twice) to be sure it is a prime and to be sure what sequential position a MP holds ALL smaller exponents must be LL checked and double checked. DC tests are now several years behind the current LL range (i.e the 21st prime found in 2004 still has not been verified to be the 21st)

You could make a case for TF or P-1: The more exponents that can be eliminated with a factor found the less that have to undergo the relatively much long LL and DC test and the quicker we can move up the list.

And, if money is important the next big prize is at the 100M digit level.

So all work types have value. Therefore, in my humble opinion the most important is that you contribute in a way that you enjoy and learn.

[Brian-E]

Originally Posted by petrw1

You could make a case for TF or P-1: The more exponents that can be eliminated with a factor found the less that have to undergo the relatively much long LL and DC test and the quicker we can move up the list.

Concerning the importance of TF and P-1, this also raises the point that these work types are necessary before LL tests are carried out - since we do not waste our effort by carrying out expensive LL tests on numbers which could have been quickly proven composite by discovering a factor. Therefore TF and P-1 can be considered particularly important because they can form a bottle neck in that LL testing is held up if TF and P-1 fall behind. TF is not a problem because plenty of testers are doing that, but the P-1 factorisation is apparently in danger of failing to stay ahead of the wave of first-time LL testing.

With that in mind, you could possibly put "P-1 (large)" at the top of the importance list.

Beyond that, the question of importance is completely subjective as petrw1 explains, depending on how you rank the importance of breaking records, winning prizes, discovering a new Mersenne Prime, discovering the largest known prime, looking for a 100M digit prime, increasing the threshold below which the primeness or compositeness of all mersenne numbers is known, and (fully) factorising low mersenne numbers.

[Yura]

Thank you petrw1.

I just wanted to find out how to be the most useful to this project and how best to employ all those wasted CPU cycles...

[petrw1]

Originally Posted by Yura

Thank you petrw1.

I just wanted to find out how to be the most useful to this project and how best to employ all those wasted CPU cycles...

Then I would agree with Brian. If you can make the necessary RAM (at least 500MB per core) available the most help is still required in P-1. Or if you are low on RAM consider DC. OR one core of each.

[Yura]

Thank you petrw1 and Brian-E for your advices.

I released few assigned LL exponents (it looks like everybody wants to be famous ;) ) and manually assigned few P-1. As per 'Status' from Prime95 it will take about 1.5 days to complete one P-1.
If that is The most helpful task to the project I will be doing a lot of those :))


Where I can suggest new feature for Prime95?
I would like to see expected completion date (and time) on each line in Worker window. That way I do not have to do Test, Status every time I am wandering when test will be completed for current exponent. Or to have a check box in Options, Preferences to have it displayed or not if that takes too much CPU cycles.

[ckdo]

Originally Posted by Yura

I would like to see expected completion date (and time) on each line in Worker window.

[Like]

[petrw1]

Originally Posted by Yura

Where I can suggest new feature for Prime95?
I would like to see expected completion date (and time) on each line in Worker window.

Or in the title bar of each worker window since each worker only works on one assignment at a time.

[Christenson]

Originally Posted by petrw1

Or in the title bar of each worker window since each worker only works on one assignment at a time.

mfaktc 0.17 gives the time to completion of each TF bit level on each reporting line; the amount of CPU time involved is pretty trivial as long as the status lines aren't appearing many times per second.

Menu titlebars are nice for the Windows version, but I have P95 running in text mode only on Linux, and also want the "expected time to complete workunit" there.