Largest LL tests ever completed

[Madpoo]

Originally Posted by kriesel

gpuowl hasn't offered nonzero shifts since its early LL incarnations around v0.5, ~3 years ago. https://www.mersenneforum.org/showpo...83&postcount=7

Oh... thought it did. LaurV meanwhile has still continued to poach my assignments. 31 of them now.

He's going through and double-checking his previous first-time work, even if the DC is assigned to someone else. Boo...

[kriesel]

Originally Posted by Madpoo

Oh... thought it did. LaurV meanwhile has still continued to poach my assignments. 31 of them now.

He's going through and double-checking his previous first-time work, even if the DC is assigned to someone else. Boo...

Cudalucas yes
cllucas no
prime95 / mprime yes
mlucas yes
gpuowl not for a long time, and some of us are lobbying Mihai for its return.

Re self-DC, I thought that was regarded as bad form.
Likely to earn a TC.
Only accepted in rare cases with extenuating circumstances. Like poaching.. Doing both seems like pushing one's luck.

[LaurV]

Originally Posted by Madpoo

Oh... thought it did. LaurV meanwhile has still continued to poach my assignments. 31 of them now.

He's going through and double-checking his previous first-time work, even if the DC is assigned to someone else. Boo...

There is no poaching. First, you reserved them AFTER you have seen (as an admin of PrimeNet) what I am doing, and you were informed (on PM) of what I am doing, probably hoping to stop me/scare me off, moreover I told you I am going to continue, so you may start "from the end" if you don't want to be stepped on toes, which you ignored, and second, due to the fact that they are "self-DC", one TC will anyhow be needed in the future, so you could just let them run. It is me who wasted the time and resources, not you, because your TC (or other's) will be considered the right one, and my "self DC" will be considered craziness, but what I do with my time and resources, is my problem. I do not consider the TC as important for this particular work, as I did the work honestly with different machines, different softwares, with 4-6 years difference between tests, and I know I didn't cheat, but you and some others still can consider me a cheater if you like, and TC them all. It is hurting a bit not being a "trusted" contributor after so many years, but no harm done, it will pass... and actually, I would be happy if you TC all of them, and find one mistake, to prove me a cheater. As a proof that you only intended to "scare me off", you stopped after I reported the results, and then you came here to complain what a bad guy I am, hehe
I still have 7 of them running, to be the guy with "no unverified results".

[masser]

Originally Posted by LaurV

I know I didn't cheat, but you and some others still can consider me a cheater if you like, and TC them all. It is hurting a bit not being a "trusted" contributor after so many years, but no harm done, it will pass...

I don't think anyone here thinks you are a cheater. I think having "trusted" self-double-checkers is a model that doesn't scale well. There could be a lot of unintended consequences to a model that allows "trusted" users.

[S485122]

Originally Posted by LaurV

...
the guy with "no unverified results".

Some "guys" limit themselves to double-checks and are thus with "no unverified results.

Jacob

[LaurV]

Originally Posted by masser

I think having "trusted" self-double-checkers is a model that doesn't scale well. There could be a lot of unintended consequences to a model that allows "trusted" users.

Yes, you are totally right, and this was discussed in the past. One situation in which somebody could "take advantage" of such thing, would be to fake a LL result and report it in my name (which is possible, the server "allows it" if you know how to do it), and that, being "trusted", will never be discovered as a fake, until a third party will accidentally TC it, or until some admin will accidentally find out that the IP address from where the result was reported is not in my range (Madpoo also knew about this, from PM discussion). Such action would hit the project, but also the "trusted" user ("Hey, look, you self-DCed this and both the first test and the DC are wrong, so you are a thief"). So, I am not exactly for "being a trusted user", and I am aware that such thing would not be good for the project. It is just that bitter taste, you know, when you are called out to the blackboard...

OTOH, we used to "trade credit" in the past, i.e. I always had too much GPU power and too less CPU power, and I was doing (and reporting) TF for different people, while they were doing (and reporting) LL/P-1 for me (ex: Kracker, Chris, etc). I was honest in that "trade", but I still don't know if they were too. (Well, now I know that they were, because all the self-DC results matched till today, and there are only few left, whose exponents are larger than the "trading period", so they were worked later by myself).

[retina]

Or another option is to not waste effort by self-DCing, and instead work on new exponents, or DCing someone else's results. And let someone else check on your results.

[Prime95]

Originally Posted by retina

Or another option is to not waste effort by self-DCing, and instead work on new exponents, or DCing someone else's results. And let someone else check on your results.

Or, if immediate DC is your goal, submit your exponents to the strategic DC and TC thread. I'm sure you'll find people willing to DC your work so as to avoid the "wasted" effort of self-DC.

[LaurV]

That's not good, it doesn't cure my paranoia and it doesn't piss anybody else off, either....
And the time is not really wasted, I have two R7s which can't be used for anything else until the "shift" issue in gpuOwl is addressed (so I could continue to do first tests and DCs in parallel, hehehe, right now is not fun, because I can report only one of them).

By now I am switching to "normal DC" for a while, till I will have the time to go back to the other cards and TF. I could have done "normal DCs" from start, but assigning them is a bit pain in the butt, however, and the list with own old work was ready available. Fun part (undiscovered yet by the masses here) is that my local copy was a bit old, and I ended up doing the work and reporting few exponents after they were already DCed by "fast" third parties (like RyanP) (but this was rare, actually I found only one case).

[kriesel]

Originally Posted by LaurV

I have two R7s which can't be used for anything else until the "shift" issue in gpuOwl is addressed.

R7's would be powerful for mass LLDC of existing low prime95 first tests. Help show https://www.mersenne.org/report_expo...xp_lo=57885161 is actually the 48th Mersenne prime.

[Madpoo]

Originally Posted by LaurV

There is no poaching.

M55582313
M56784851
M59626937

Just 3 of the top list of exponents you poached from people *besides me* that were assigned the exponent before you turned yours in, and whom, by the way, are still checking in their progress because they don't know you effectively expired their work in progress.

Besides the ones you poached from me, there are 27 you poached from others.

Oh, and don't take it personally that I'm doing TC's of your self-verified work. As you and others know, I started doing that years ago with the help of many others to clear a HUGE backlog of such things. Couldn't have done that on my own.

And with the system in place on the server to try and avoid handing those out automatically, I've been able to keep up with the stray self-verified checks that filtered in, with the exception of people who deliberately do it. Like I said, it's not personal.

And I'm still turning in the TC results for the ones assigned to me, plus I'm monitoring those other ones you poached that are being actively worked on by others. At some point I'll end up TC'ing those as well. I don't have the same massive server horsepower I used to, but with a few machines I've been able to keep up with it so far, it just takes me longer. Especially, as I mentioned, when someone is deliberately doing it on dozens of exponents.

For what it's worth... the reason I say "it's not personal" is because although we here on the forum, or in general, may know and trust other people's results, and trust them to do the right thing, this project is also about posterity. Envision some researchers years from now, looking back at the data, and having no idea who LaurV or Madpoo are, and wondering why someone was self-verifying their own work. It's not the scientific method to confirm your own results... you *need* peer review or the results just can't be generally trusted. Know what I'm saying? It's as simple as that... not casting aspersions on anyone, but it's totally in regards to the reputation of the project itself.