New TF assignment strategy

Prime95 · 2011-10-30, 01:40

Exponents for TF below 58M are only available from the manual testing pages.
TF exponents above 58M are available for automatic assignment - presumably prime95.

Newly released exponents below 58M will be factored 3 extra bits.

If you GPUers can stay ahead of all that, there are 32000 exponents in that range that are currently factored to 2^71 that I can turn into TF assignments

Dubslow · 2011-10-30, 04:32

Wooooooooooo!!!!
Does this include stuff between 45M and 50M?

davieddy · 2011-10-30, 07:19

Quote:

Originally Posted by Prime95

Exponents for TF below 58M are only available from the manual testing pages.
TF exponents above 58M are available for automatic assignment - presumably prime95.

Newly released exponents below 58M will be factored 3 extra bits.

If you GPUers can stay ahead of all that, there are 32000 exponents in that range that are currently factored to 2^71 that I can turn into TF assignments

Even without GPUs (and Oliver) the idea that any TF to date above
60M is what I previously referred to as "pissing in the wind".

I think one man (P-1) and his dog (Garo) can manage to keep 72 bits
ahead of the wavefront.

Expect some music soon.

Make that NOW

D

Christenson · 2011-10-30, 14:49

There's those of us that would like to add another bit or two to the DC range, too. Not that I'm not turning in regular and irregular DCs...

davieddy · 2011-10-30, 17:16

Quote:

Originally Posted by Christenson

There's those of us that would like to add another bit or two to the DC range, too. Not that I'm not turning in regular and irregular DCs...

Like the oh so subtly-named band "AC/DC".

How do I create a poll here?
I think "How far should we push our hardware up DCs?" would
just about get past the thought police.

Finbarr Saunders

KyleAskine · 2011-10-31, 22:05

What is the consensus on where to take these < 60000000 to?

72 or 73?

I have it set to 73 now, but should we just try to get everything to 72 first, and then we can get these reassigned to us again if we get everything to 72?

Christenson · 2011-10-31, 23:03

Take them all to 72 bits first...there will be just as many factors from 71-72 bits as from 72-73 bits, but 73 bits will take twice the effort.

petrw1 · 2011-11-02, 15:50

this could be related....

I just noticed the cutoff line between TF and ECM on the Primenet status page has dropped from 10M to 20M. http://www.mersenne.org/primenet/

...even though there has been very little ECM done beyond 6329203

LaurV · 2011-11-03, 04:25

Something is fishy with that table or I am not the right guy to understand it. For example what bothers me is the range 27M-36M. According with the table, all expos had one LL done already. So, all the expos should be either available for DC, or assigned for DC. Why the hack I can see plenty of exponents assigned for first-time LL too??
(not talking about TF, P1, ECM, other things, only LL and DC, and please remark that no LLERR present too, so either the table is wrong, or some guys do redundant work, they still have exponents assigned which were finished by other faster hunters, but in that case, their assignment should count as DC, and not as LL in the table. Quite confusing!)

Also, above 37M (and up to about 48M, and a bit upper) what appears as "Assigned for first LL" should NEVER be higher then the sum of "Status No-LL" and "Status LLERR". That would be logical...

The only possible explanation is that the LL and DC assignments were handed out in about the same time, and the DC finished faster then the LL, and the table still keep the status according with the assignment type, and not with the situation in the data base, or it is not actualized properly.

cheesehead · 2011-11-03, 05:25

Quote:

Originally Posted by LaurV

Something is fishy with that table or I am not the right guy to understand it.

Just to check: Are you referring to the Exponent Status Distribution table at http://www.mersenne.org/primenet/ ?

Quote:

For example what bothers me is the range 27M-36M. According with the table, all expos had one LL done already.

I don't quite see that in the Exponent Status Distribution table.

Quote:

So, all the expos should be either available for DC, or assigned for DC. Why the hack I can see plenty of exponents assigned for first-time LL too??

Suppose a first-time LL had errors. It can be reassigned for another first-time LL, before it's made available for DC.

Quote:

(not talking about TF, P1, ECM, other things, only LL and DC, and please remark that no LLERR present too, so either the table is wrong, or some guys do redundant work, they still have exponents assigned which were finished by other faster hunters, but in that case, their assignment should count as DC, and not as LL in the table.)

Also, above 37M (and up to about 48M, and a bit upper) what appears as "Assigned for first LL" should NEVER be higher then the sum of "Status No-LL" and "Status LLERR". That would be logical...

I'd love to answer, but not when I suspect we're referring to two different tables.

Quote:

The only possible explanation is that the LL and DC assignments were handed out in about the same time

No, that doesn't happen. No overlapping assignments on the same exponent, except for ECM where it's okay for more than one user to work on it at a time because of the randomized curves.

Dubslow · 2011-11-03, 05:27

No. If an LL assignment in poached, then the original assignment is not changed to a DC assignment. When it is reported, the exponent will be cleared, but the assignment will still register as LL. Same for the higher range; if it's been poached, the old assignment is still there.

2011-10-30, 01:40	#1
Prime95 P90 years forever! Aug 2002 Yeehaw, FL 1110101100110₂ Posts	New TF assignment strategy Exponents for TF below 58M are only available from the manual testing pages. TF exponents above 58M are available for automatic assignment - presumably prime95. Newly released exponents below 58M will be factored 3 extra bits. If you GPUers can stay ahead of all that, there are 32000 exponents in that range that are currently factored to 2^71 that I can turn into TF assignments

2011-11-02, 15:50	#8
petrw1 1976 Toyota Corona years forever! "Wayne" Nov 2006 Saskatchewan, Canada 4676₁₀ Posts	Trying not to start yet another thread this could be related.... I just noticed the cutoff line between TF and ECM on the Primenet status page has dropped from 10M to 20M. http://www.mersenne.org/primenet/ ...even though there has been very little ECM done beyond 6329203 Last fiddled with by petrw1 on 2011-11-02 at 15:53

2011-11-03, 04:25	#9
LaurV Romulan Interpreter Jun 2011 Thailand 2×5×31² Posts	Something is fishy with that table or I am not the right guy to understand it. For example what bothers me is the range 27M-36M. According with the table, all expos had one LL done already. So, all the expos should be either available for DC, or assigned for DC. Why the hack I can see plenty of exponents assigned for first-time LL too?? (not talking about TF, P1, ECM, other things, only LL and DC, and please remark that no LLERR present too, so either the table is wrong, or some guys do redundant work, they still have exponents assigned which were finished by other faster hunters, but in that case, their assignment should count as DC, and not as LL in the table. Quite confusing!) Also, above 37M (and up to about 48M, and a bit upper) what appears as "Assigned for first LL" should NEVER be higher then the sum of "Status No-LL" and "Status LLERR". That would be logical... The only possible explanation is that the LL and DC assignments were handed out in about the same time, and the DC finished faster then the LL, and the table still keep the status according with the assignment type, and not with the situation in the data base, or it is not actualized properly. Last fiddled with by LaurV on 2011-11-03 at 04:33

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
TF strategy	davieddy	Lounge	34	2012-03-17 02:03
Assignment Strategy	diamonddave	GPU to 72	18	2011-12-06 19:56
P95 trial division strategy	SPWorley	Math	8	2009-08-24 23:26
Strategy game	Kees	Puzzles	4	2006-04-07 07:17
New Strategy	Citrix	Prime Sierpinski Project	5	2004-10-31 12:25

2011-10-30, 04:32	#2
Dubslow Basketry That Evening! "Bunslow the Bold" Jun 2011 40<A<43 -89<O<-88 1C35₁₆ Posts	Wooooooooooo!!!! Does this include stuff between 45M and 50M?

2011-10-30, 14:49	#4
Christenson Dec 2010 Monticello 5·359 Posts	There's those of us that would like to add another bit or two to the DC range, too. Not that I'm not turning in regular and irregular DCs...

2011-10-31, 22:05	#6
KyleAskine Oct 2011 Maryland 2×5×29 Posts	What is the consensus on where to take these < 60000000 to? 72 or 73? I have it set to 73 now, but should we just try to get everything to 72 first, and then we can get these reassigned to us again if we get everything to 72?

2011-10-31, 23:03	#7
Christenson Dec 2010 Monticello 5×359 Posts	Take them all to 72 bits first...there will be just as many factors from 71-72 bits as from 72-73 bits, but 73 bits will take twice the effort.

2011-11-03, 05:27	#11
Dubslow Basketry That Evening! "Bunslow the Bold" Jun 2011 40<A<43 -89<O<-88 3·29·83 Posts	No. If an LL assignment in poached, then the original assignment is not changed to a DC assignment. When it is reported, the exponent will be cleared, but the assignment will still register as LL. Same for the higher range; if it's been poached, the old assignment is still there.