v5 work types
 

I have some machines which are low on RAM set to LL-NF to avoid P-1 work. LL-NF is supposed to be "LL test with no factoring". I expected fully factored assignments ready for LL testing, but what I actually get is the exact opposite: factoring assignments! Are there no more double check assignments left?

Bump!

I'll look at it when I get back

I have this line in my prime.txt:
[code]WorkPreference=6[/code]I'm running mprime 25.7 on that machine. One core is now wasting time with no assignments, and the other has received ECM assignments.

6 used to be the designation for LL-only tests. Is that still the case?

6 seems to be ECM Fermat :

[quote]Use the following values to select a work type:
0 - Whatever makes the most sense
2 - Trial factoring
100 - First time primality tests
101 - Double-checking
102 - World record primality tests
4 - P-1 factoring
104 - 100 million digit primality tests
1 - Trial factoring to low limits
5 - ECM on small Mersenne numbers
6 - ECM on Fermat numbers[/quote]

What are 3 and 103 then? Inquiring minds want to know.

[QUOTE=Uncwilly;149451]What are 3 and 103 then? Inquiring minds want to know.[/QUOTE]

According to the v5 spec: 3 = P-1 factoring on small Mersenne numbers. 103 = LL testing on 10 million digit numbers.

I'm still not very sure on the difference between [I]first time primality[/I], [I]world record primality[/I] and [I]100 million digit primality[/I].

I know the difference between [I]first time[/I] and [I]100 million digit[/I], but what is [I]world record[/I]? Anything bigger than the current record? Wouldn't that be the same as first time primality? (Just want to clarify).

[QUOTE=hockmeng;149616]I'm still not very sure on the difference between [I]first time primality[/I], [I]world record primality[/I] and [I]100 million digit primality[/I].

I know the difference between [I]first time[/I] and [I]100 million digit[/I], but what is [I]world record[/I]? Anything bigger than the current record? Wouldn't that be the same as first time primality? (Just want to clarify).[/QUOTE]

There's still a lot of exponents below the current world record that haven't been tested. Most of the time they're all reserved, so the two work options end up drawing assignments from the same pool. But when those lower first time tests expire and go back into the work queue, they'll only be assigned to people with first time LL as their work option.

[quote=hockmeng;149616]Anything bigger than the current record? Wouldn't that be the same as first time primality?[/quote]
Expansion on Kevin's explanation:

From a particular set of available exponents, PrimeNet does assign first-time primality tests in increasing order of exponent. However, different users perform LL tests at different rates. Thus, LL tests are not necessarily completed in increasing order of exponent. Some folks may be using systems that are slower than those used by the most recent world-record finders, so may still be in the midst of tests on exponents lower than those that had been assigned later to the record-finders.

Now, complicate that with the fact that many test assignments are abandoned (or deliberately released) by their assignees. PrimeNet waits a certain amount of time after getting the latest progress report before deciding to classify an assignment as abandoned. Thus, some folks may be assigned first-time LL tests on exponents which had already been assigned one or more times, months or years earlier, but are without a first-time completion yet. In some cases, an exponent may be years "behind" exponents just above or below it. Even if the later assignees have fast systems, they may now just be in the midst of testing exponents which are far lower than the world record.

That's why there's a difference between assigning

the lowest available untested exponent

and assigning

the lowest available untested exponent that's above the current world record.

Well, I stopped mprime, and changed 6 to 102. Where are the work types documented? I don't see it in readme.txt or undoc.txt.

I started it again, and received:
[code]
$ mprime -d
[Main thread Nov 19 16:09] Mersenne number primality test program version 25.7
[Main thread Nov 19 16:09] Starting workers.
[Worker #1 Nov 19 16:09] Worker starting
[Worker #1 Nov 19 16:09] Setting affinity to run worker on logical CPU #0
[Worker #2 Nov 19 16:09] Worker starting
[Worker #2 Nov 19 16:09] Setting affinity to run worker on logical CPU #1
[Worker #1 Nov 19 16:09] No work to do at the present time. Waiting.
[/code]Of course, it doesn't explain why it is "waiting" instead of "getting more work". It just says "waiting".

A little more investigation, and something really strange has happened. My prime.log timestamp is Oct 27, so mprime hasn't even communicated to the server since then! Checking results.txt, I noticed that the last two results it had (Nov 12, Nov 18) were still not credited on my account. So, I submitted them with the manual submission page, and they were accepted just fine.

So, I think that something about my configuration is bad. When I upgraded to v25, I kept my previous prime.ini, but renamed it as prime.txt and added in the "V5UserID=uigrad" line. Is there anything else that I needed to do?

Are you sure it has UsePrimenet=1 ?

The worktype specifications are on the server : PrimeNet Summary / Web API Specification [URL=http://v5.mersenne.org/v5design/v5webAPI_0.97.html#7.0]Assignment Work Types[/URL].

Jacob

Thanks CADavis, that was the issue. It's back to doing LL tests again now :smile:
Thanks Jacob. That page is very useful, I wish I had found it before!

[QUOTE=S00113;146747]I have some machines which are low on RAM set to LL-NF to avoid P-1 work. LL-NF is supposed to be "LL test with no factoring". I expected fully factored assignments ready for LL testing, but what I actually get is the exact opposite: factoring assignments! Are there no more double check assignments left?[/QUOTE]

Fixed. If an exponent is within 10 million of an exponent that requires TF or P-1 then you will get the one not needing factoring.