Proposed LL assignment and recycle rules - Page 7

NBtarheel_33 · 2014-02-20, 02:09

Quote:

Originally Posted by Prime95

1)The only downside I see is that some slow computers could get 63M exponents that could slow down milestones.

What milestone would a 63M exponent affect? Or do you know something about the next prime that you're keeping to yourself? Hmm...wink wink...

garo · 2014-02-20, 20:02

Sounds like we have a plan. Lowering the Cat 3/4 cut-off and TF to 73 only for Cat 4 exponents is the best way to go about this. And as NB_Tarheel points out, it is unlikely that 63M exponents will be holding up any milestones in the next 18 months. The first-time tail is too long for that. The new rules will shorten the tail but I think it will be a while before 63M is holding up progress.

GPU WMS bug reported here: http://www.mersenneforum.org/showpos...postcount=2697

Primeinator · 2014-02-20, 21:17

Quote:

Originally Posted by garo

Sounds like we have a plan. Lowering the Cat 3/4 cut-off and TF to 73 only for Cat 4 exponents is the best way to go about this. And as NB_Tarheel points out, it is unlikely that 63M exponents will be holding up any milestones in the next 18 months. The first-time tail is too long for that. The new rules will shorten the tail but I think it will be a while before 63M is holding up progress.

GPU WMS bug reported here: http://www.mersenneforum.org/showpos...postcount=2697

Just out of curiosity where is wavefront/first-time trail? I currently do WR-LLs and I get a 10M difference in exponents. This doesn't bother me, I am just curious why my computer will be doing a 69M candidate at the same time it is doing a 61M candidate (for example).

garo · 2014-02-20, 22:56

Quote:

Originally Posted by Primeinator

Just out of curiosity where is wavefront/first-time trail? I currently do WR-LLs and I get a 10M difference in exponents. This doesn't bother me, I am just curious why my computer will be doing a 69M candidate at the same time it is doing a 61M candidate (for example).

According to this page: http://www.mersenne.org/primenet you cold be working on anything from 48M to 69M.

TheMawn · 2014-02-21, 00:36

What happens is you pretty much grab the smallest available exponent at whatever time you request an exponent. Now, the reason there is a tail is some assignments never get finished. They expire and are recycled. As old assignments further down the list are recycled, people pick them up.

Some people will "hoard" assignments. If you checked every six hours or whatever and whenever you noticed an assignment below 50M for example, you quickly went and requested one assignment, you would pick it up and keep it for later. This way, you went through more smaller assignments than your average bear. This is fine, as long as people don't hold up the line in the process.

The idea of the new assignment and recycling rules is to SERIOUSLY shorten the tail. Check out the Primenet Summary. At this point in time, I see about 1000 <60M assignments available and 20,000 assigned (of which only about 1000 are smaller than 54M). By giving the "trailing edge" assignments to fast computers, they will be finished instead of recycled, and this will shorten the tail and tidy things up.

To be honest, it'll probably reduce the occurrence of finding new primes smaller than the record holder. When I was looking up Mersenne Primes before I joined this project, I thought it was silly that the primes were not found in order. Now that I know how it works, I see that it isn't silly at all.

petrw1 · 2014-04-28, 20:31

I still get duplicate CPUs a couple times a year ... no big deal, I just merge them.

Except with the new preferred assignment rules ... because it merges old into new ... after the merge it shows up as a new CPU without the necessary rating to allow it to get preferred assignments ... until each core completes 2 relevant assignments.

Could it instead merge new into old.... or keep the rating of old?

Thanks

manfred4 · 2014-04-28, 21:16

I had the same problem as you have and to work around that I just manually said prime95 to test some numbers, that weren't assignet ever - with the statistics I was able to find new ones for each core in the 54M Range and PrimeNet accepted me doing that work without having to reached those thresholds yet.

lycorn · 2014-04-29, 21:41

Quote:

Originally Posted by petrw1

I still get duplicate CPUs a couple times a year ... no big deal, I just merge them.

Except with the new preferred assignment rules ... because it merges old into new ... after the merge it shows up as a new CPU without the necessary rating to allow it to get preferred assignments ... until each core completes 2 relevant assignments.

Could it instead merge new into old.... or keep the rating of old?

Thanks

+1
Same here.

TheMawn · 2014-05-30, 18:36

How have the categories been working out? The trailing edge was certainly sliced clean off, but what about in other respects?

sonjohan · 2014-07-30, 16:12

I normally have all my computers running 24/7 (including mfackc).
However, due to the extreme heat in my country, I have decided to shut them down as long as the heatwave is there.

The reason is, that the when I got home, it sounded like a fighter jet turbine was in the room. I don't know when I'll be reactivating the Prime95 (shouldn't take that long...).
It would be a pity if I get downgraded just because I don't want to fry my PC's.

chalsall · 2014-07-30, 16:38

Quote:

Originally Posted by sonjohan

It would be a pity if I get downgraded just because I don't want to fry my PC's.

At the very worst case, you'd be "downgraded" from getting "Cat1" assignments to "Cat3". Probably only down to "Cat2". And only for two assignments.

If you're really concerned about this, and you are downgraded, pump out a couple of quick DC's, and you'll be back in the "good books".

Or, alternatively, you could ask George for an exemption. But an algorithmic way of taking into account such special cases would probably be more work than its worth.

2014-02-20, 20:02	#68
garo Aug 2002 Termonfeckin, IE 5314₈ Posts	Sounds like we have a plan. Lowering the Cat 3/4 cut-off and TF to 73 only for Cat 4 exponents is the best way to go about this. And as NB_Tarheel points out, it is unlikely that 63M exponents will be holding up any milestones in the next 18 months. The first-time tail is too long for that. The new rules will shorten the tail but I think it will be a while before 63M is holding up progress. GPU WMS bug reported here: http://www.mersenneforum.org/showpos...postcount=2697 Last fiddled with by garo on 2014-02-20 at 20:07

2014-04-28, 20:31	#72
petrw1 1976 Toyota Corona years forever! "Wayne" Nov 2006 Saskatchewan, Canada 2²·7·167 Posts	issue with merged CPUs I still get duplicate CPUs a couple times a year ... no big deal, I just merge them. Except with the new preferred assignment rules ... because it merges old into new ... after the merge it shows up as a new CPU without the necessary rating to allow it to get preferred assignments ... until each core completes 2 relevant assignments. Could it instead merge new into old.... or keep the rating of old? Thanks

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
PrimeNet Assignment Rules	S485122	PrimeNet	11	2021-05-20 14:54
Modifications to DC assignment rules	Prime95	PrimeNet	74	2017-01-18 18:36
Understanding assignment rules	Fred	PrimeNet	3	2016-05-19 13:40
Proposed DC assignment and recycle rules	Prime95	Data	110	2015-01-15 04:44
Proposed TF, P-1, ECM assignment and recycle rules	Prime95	Data	9	2014-02-27 23:52

2014-02-21, 00:36	#71
TheMawn May 2013 East. Always East. 11·157 Posts	What happens is you pretty much grab the smallest available exponent at whatever time you request an exponent. Now, the reason there is a tail is some assignments never get finished. They expire and are recycled. As old assignments further down the list are recycled, people pick them up. Some people will "hoard" assignments. If you checked every six hours or whatever and whenever you noticed an assignment below 50M for example, you quickly went and requested one assignment, you would pick it up and keep it for later. This way, you went through more smaller assignments than your average bear. This is fine, as long as people don't hold up the line in the process. The idea of the new assignment and recycling rules is to SERIOUSLY shorten the tail. Check out the Primenet Summary. At this point in time, I see about 1000 <60M assignments available and 20,000 assigned (of which only about 1000 are smaller than 54M). By giving the "trailing edge" assignments to fast computers, they will be finished instead of recycled, and this will shorten the tail and tidy things up. To be honest, it'll probably reduce the occurrence of finding new primes smaller than the record holder. When I was looking up Mersenne Primes before I joined this project, I thought it was silly that the primes were not found in order. Now that I know how it works, I see that it isn't silly at all.

2014-04-28, 21:16	#73
manfred4 Mar 2014 Germany 2³×3×5 Posts	I had the same problem as you have and to work around that I just manually said prime95 to test some numbers, that weren't assignet ever - with the statistics I was able to find new ones for each core in the 54M Range and PrimeNet accepted me doing that work without having to reached those thresholds yet.

2014-05-30, 18:36	#75
TheMawn May 2013 East. Always East. 11010111111₂ Posts	How have the categories been working out? The trailing edge was certainly sliced clean off, but what about in other respects?

2014-07-30, 16:12	#76
sonjohan May 2003 Belgium 2·139 Posts	I normally have all my computers running 24/7 (including mfackc). However, due to the extreme heat in my country, I have decided to shut them down as long as the heatwave is there. The reason is, that the when I got home, it sounded like a fighter jet turbine was in the room. I don't know when I'll be reactivating the Prime95 (shouldn't take that long...). It would be a pity if I get downgraded just because I don't want to fry my PC's.