TF while assigned for LL?

cuBerBruce · 2014-09-26, 21:13

I currently have three "first LL" assignments for exponents that actually have a prior LL result that is suspected to be bad. I have noticed, though, that Jerry Hallett has reported trial factoring these exponents from 73 to 74 while they were assigned to me. As no factors were found, this doesn't affect me as I still need to complete these LL tests. But it was surprising to me that someone else apparently did some trial factoring on these exponents while I had LL assignments for them. I'm wondering if there is a GPU72 bug that caused this.

The exponents are:
58,380,271
58,381,049
58,464,473

Mark Rose · 2014-09-26, 23:30

It's possible that they were assigned for TF at some point then the assignments expired and the work was still completed.

TheMawn · 2014-09-27, 00:19

Mark Rose is probably right though it surprises me because Jerry is a pretty big producer if memory serves...

GPU72 is a big tool for us and it gets used a lot so it's natural for it to gather the occasional bug. There have been some updates to how GPU72 contacts Primenet which are very welcome but something might have gotten skewered in the process.

I know Chris is fairly busy at the moment so this might not get addressed right away. It's good that you've been watching this however. If a factor was found it would be good of you to catch the bug and cancel your LL.

cuBerBruce · 2014-09-27, 01:41

Looking at other exponents in the same region, I've found a few other exponents also recently reported as TF'ed by Jerry, and while someone had been assigned the exponent for LL somewhere from 5 to 9 months. These other exponents did not have any prior "suspect" result, so the fact that mine all had "suspect" LLs previously is probably just coincidental. So I wouldn't be surprised if Mark is right rather than some issue between PrimeNet and GPU72.

Since the new assignment rules were put in place, I think there is a fair amount of assignments being completed by people after getting recycled. I've noticed a few cases of triple checks having occurred on the DC wave because of this. This is more or less why I've been paying more attention to the history/status of my exponents lately.

Mark Rose · 2014-09-27, 02:02

Perhaps a future version of mprime/prime95 should do period checks and cancel any work that would be for a triple check?

Prime95 · 2014-09-27, 02:45

Quote:

Originally Posted by Mark Rose

Perhaps a future version of mprime/prime95 should do period checks and cancel any work that would be for a triple check?

Well, users did complain (in the distant past) about not getting their precious CPU credit for an abandoned test.

I understand the desire to eliminate all inefficiencies. However, my quad-core Haswell is now churning out 1 DC a day -- each DC takes 4 core-days to complete. So such a feature would (in this fairly uncommon occurrence) save on average 2 Haswell core-days. IMO, not a big deal.

I think most everyone was aware that the new assignment rules would result in a significant increase in the number of triple-checks and first-time-LL-tests-converted-to-doublechecks-due-to-an-expired-assignment-reporting-in-belatedly. The consensus was this price was worth paying to advance milestones in a more regular fashion.

Mark Rose · 2014-09-27, 03:05

I, too, agree that timely milestones are more worth-while than ultimate efficiency.

If such a feature is ever implemented, I feel it should be optional.

LaurV · 2014-09-27, 08:36

Quote:

Originally Posted by Prime95

Well...

Totally agree with George here. Very good posting.

chalsall · 2014-09-27, 22:14

Quote:

Originally Posted by TheMawn

I know Chris is fairly busy at the moment so this might not get addressed right away. It's good that you've been watching this however. If a factor was found it would be good of you to catch the bug and cancel your LL.

Please don't blame Jerry for this; his machines gets all their work from GPU72.

I probably made a mistake somewhere in my spiders which resulted in this.

IMHO, at the end of the day, not the end of the world.

TheMawn · 2014-09-28, 05:41

Quote:

Originally Posted by chalsall

IMHO, at the end of the day, not the end of the world.

Agreed. Small potatoes.

2014-09-26, 21:13	#1
cuBerBruce Aug 2012 Mass., USA 2·3·53 Posts	TF while assigned for LL? I currently have three "first LL" assignments for exponents that actually have a prior LL result that is suspected to be bad. I have noticed, though, that Jerry Hallett has reported trial factoring these exponents from 73 to 74 while they were assigned to me. As no factors were found, this doesn't affect me as I still need to complete these LL tests. But it was surprising to me that someone else apparently did some trial factoring on these exponents while I had LL assignments for them. I'm wondering if there is a GPU72 bug that caused this. The exponents are: 58,380,271 58,381,049 58,464,473

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Assigned exponents	ATH	PrimeNet	1	2017-12-15 06:35
CPUs assigned	houding	Hardware	10	2015-03-25 00:33
suspect LL assigned again to me	rudi_m	PrimeNet	10	2009-02-12 09:56
max exponents assigned.	JuanTutors	PrimeNet	2	2004-07-16 23:26
Well, I just got M40 assigned to myself...	GP2	PrimeNet	2	2003-12-03 03:32

2014-09-26, 23:30	#2
Mark Rose "/X\(‘-‘)/X\" Jan 2013 101101110001₂ Posts	It's possible that they were assigned for TF at some point then the assignments expired and the work was still completed.

2014-09-27, 00:19	#3
TheMawn May 2013 East. Always East. 11·157 Posts	Mark Rose is probably right though it surprises me because Jerry is a pretty big producer if memory serves... GPU72 is a big tool for us and it gets used a lot so it's natural for it to gather the occasional bug. There have been some updates to how GPU72 contacts Primenet which are very welcome but something might have gotten skewered in the process. I know Chris is fairly busy at the moment so this might not get addressed right away. It's good that you've been watching this however. If a factor was found it would be good of you to catch the bug and cancel your LL.

2014-09-27, 01:41	#4
cuBerBruce Aug 2012 Mass., USA 476₈ Posts	Looking at other exponents in the same region, I've found a few other exponents also recently reported as TF'ed by Jerry, and while someone had been assigned the exponent for LL somewhere from 5 to 9 months. These other exponents did not have any prior "suspect" result, so the fact that mine all had "suspect" LLs previously is probably just coincidental. So I wouldn't be surprised if Mark is right rather than some issue between PrimeNet and GPU72. Since the new assignment rules were put in place, I think there is a fair amount of assignments being completed by people after getting recycled. I've noticed a few cases of triple checks having occurred on the DC wave because of this. This is more or less why I've been paying more attention to the history/status of my exponents lately.

2014-09-27, 02:02	#5
Mark Rose "/X\(‘-‘)/X\" Jan 2013 29·101 Posts	Perhaps a future version of mprime/prime95 should do period checks and cancel any work that would be for a triple check?

2014-09-27, 03:05	#7
Mark Rose "/X\(‘-‘)/X\" Jan 2013 2929₁₀ Posts	I, too, agree that timely milestones are more worth-while than ultimate efficiency. If such a feature is ever implemented, I feel it should be optional.