making PrimeNet accept duplicate TF results?

ixfd64 · 2012-12-05, 20:47

I don't want to hijack the "GPU to 72 status" thread, so I thought I'd make a new one here.

There has been some discussion on whether to do double checking for TF results. The consensus is that it wouldn't be very productive as the cost of finding a missed factor far outweighs the extra LL testing.

That having been said, would it be a good idea to configure PrimeNet to accept duplicate TF results as long as they're from different users? Alternatively, should we let trusted users submit results for ranges that have already been tested? This would allow double checking for missed factors but also prevent people from submitting results twice.

Also, I've seen a couple of exponents with duplicate ranges listed in their history. Is there a trick to force PrimeNet to accept these results, or is this something only George/Scott can do?

The reason I ask is that 30 of my GPU to 72 assignments were accidentally poached. Chalsall has fixed GPU to 72 to prevent this from happening again, but PrimeNet is rejecting my results as duplicate. It's not that big of a deal, but I'd rather not let my efforts to go to waste. I PM'd George about it, but so far nothing has happened. Is there anything else I could do?

flashjh · 2012-12-05, 23:55

I wonder, has anyone ever found a factor missed by original TF? James has a site for P-1 missed factors. I know that duplication of effort, especially for TF & P-1 doesn't make much sense, but really, we're all still doing this stuff.

Just the other day I was wondering when anyone would re-do all our TF & P-1 looking for factors.

For the time being, what sample (quantity and depth) of TF rechecking could someone do with no factor before everyone would be comfortable that the original TF was good? What happens if someone finds a factor? Should we then really consider TFing and possibly P-1ing again?

Uncwilly · 2012-12-06, 00:16

Quote:

Originally Posted by flashjh

Just the other day I was wondering when anyone would re-do all our TF & P-1 looking for factors.

I believe that someone has been doing that for all of the numbers with no known factors. Starting at the very earliest of number.

dabaichi · 2012-12-06, 00:24

Quote:

Originally Posted by flashjh

I wonder, has anyone ever found a factor missed by original TF?

Yes. Actually Mersenne @ Home found some factors missed by original TF. For example 81096871, 83420257, 137041361, and here

LaurV · 2012-12-06, 03:39

See my opinion in the GPU72 thread (@chalsall, can you please move the posts related to TF, mine included, from that thread to here? I have not so high mod rights, thanks).

Generally, I believe allowing duplicate TF reports will not benefit too much, as long as we don't have a way to check if that task was really fulfilled, people will do it for credit, without effectively doing the work. One solution would be to give no credit for it, other solution would be to implement w1 keys in all reports (mfaktX, CuLu, talking about LL too, CuLu would have shifts also), but this effort may be too large for the benefit it brings. At the end, LL tests will anyhow be done if a factor is missing. If this happened 6 times in 60 millions exponents, we just lose an insignificant fraction of time doing additional LL. Poaching wastes MUCH more time... If up to me, I would invest time in finding ways to kill the poachers. Like hanging them in the city market by their guts, so the people can see them would be a good start

, but someone can come with a better and less aggressive idea.

markr · 2012-12-08, 07:04

Quote:

Originally Posted by flashjh

I wonder, has anyone ever found a factor missed by original TF?

Yes, for example this thread http://www.mersenneforum.org/showthread.php?t=1425 from several years ago.

2012-12-05, 20:47	#1
ixfd64 Bemusing Prompter "Danny" Dec 2002 California 5×479 Posts	making PrimeNet accept duplicate TF results? I don't want to hijack the "GPU to 72 status" thread, so I thought I'd make a new one here. There has been some discussion on whether to do double checking for TF results. The consensus is that it wouldn't be very productive as the cost of finding a missed factor far outweighs the extra LL testing. That having been said, would it be a good idea to configure PrimeNet to accept duplicate TF results as long as they're from different users? Alternatively, should we let trusted users submit results for ranges that have already been tested? This would allow double checking for missed factors but also prevent people from submitting results twice. Also, I've seen a couple of exponents with duplicate ranges listed in their history. Is there a trick to force PrimeNet to accept these results, or is this something only George/Scott can do? The reason I ask is that 30 of my GPU to 72 assignments were accidentally poached. Chalsall has fixed GPU to 72 to prevent this from happening again, but PrimeNet is rejecting my results as duplicate. It's not that big of a deal, but I'd rather not let my efforts to go to waste. I PM'd George about it, but so far nothing has happened. Is there anything else I could do?

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Primenet Recent Results Request	garo	PrimeNet	7	2014-02-03 02:53
When did PrimeNet begin rubberstamping results?	Chuck	PrimeNet	64	2014-01-06 01:22
How to get Wine/Yafu-Win32.exe to accept this expression?	Stargate38	YAFU	8	2012-09-17 19:44
V5 not reporting results to PrimeNet	edron1011	Software	8	2009-01-26 16:59
Problem with Reporting Results to Primenet	RMAC9.5	Software	3	2003-11-02 23:06

2012-12-05, 23:55	#2
flashjh "Jerry" Nov 2011 Vancouver, WA 1,123 Posts	I wonder, has anyone ever found a factor missed by original TF? James has a site for P-1 missed factors. I know that duplication of effort, especially for TF & P-1 doesn't make much sense, but really, we're all still doing this stuff. Just the other day I was wondering when anyone would re-do all our TF & P-1 looking for factors. For the time being, what sample (quantity and depth) of TF rechecking could someone do with no factor before everyone would be comfortable that the original TF was good? What happens if someone finds a factor? Should we then really consider TFing and possibly P-1ing again?

2012-12-06, 03:39	#5
LaurV Romulan Interpreter Jun 2011 Thailand 7²·197 Posts	See my opinion in the GPU72 thread (@chalsall, can you please move the posts related to TF, mine included, from that thread to here? I have not so high mod rights, thanks). Generally, I believe allowing duplicate TF reports will not benefit too much, as long as we don't have a way to check if that task was really fulfilled, people will do it for credit, without effectively doing the work. One solution would be to give no credit for it, other solution would be to implement w1 keys in all reports (mfaktX, CuLu, talking about LL too, CuLu would have shifts also), but this effort may be too large for the benefit it brings. At the end, LL tests will anyhow be done if a factor is missing. If this happened 6 times in 60 millions exponents, we just lose an insignificant fraction of time doing additional LL. Poaching wastes MUCH more time... If up to me, I would invest time in finding ways to kill the poachers. Like hanging them in the city market by their guts, so the people can see them would be a good start , but someone can come with a better and less aggressive idea.