I wonder if there isn't a simple communication problem going on there wrt 'recycled exponents' because I can't fathom any difference between numbers that have had a partial LL done and numbers than haven't been tried at all (LLs are worth exactly nothing if they aren't carried to completion). That being the case I wonder if David is talking about DCs when he says recycled?

Recycled exponents will include ones where someone reserved it for LL, but never even started the LL test. But that makes even less difference than not completing a LL test.

Are recycled exponents made available to the same list of people as first time tests? If not that would make a difference.

The other possibility is that even though whoever started the test isn't reporting progress they might actually be running it, and suddenly report a result. But has this ever happened in the past? And would a one in a million chance be worth worrying about?

Chris

[QUOTE=garo;362986]Catching recycled exponents makes complete sense especially since most exponents end up expiring 2-3 times before they are eventually completed. And it improves project throughput thus bringing the next prime find closer than if we had followed your virgin exponent obsession![/QUOTE]

Let us at least start from a point of concensus: we are aiming to find another Mersenne Prime as soon as possible. That said George (of all people) disputed this as the only way of justifying your antics.

We need to maximize the probability of finding one or more primes per day, which currently stands at 1/1500.5 (:wink: @ Axn and Phil).

This is achieved by maximizing the number of LL completions per day, the lower the mean expo the better.

Now we come to the crunch: this depends on the number of virgins assigned each day.
Currently this is decided by Chris, who has an ingenious ploy which ensures that the assignment of virgins does not exceed their production rate.: whenever the enthusiam for virgins gets too high, he assigns 68-69M expos to dampen their ardour. This is a foolproof strategy because these get recycled wjth neglible risk of being completed (<100 so far).

What is the assignment demand?
Well, just allow all expos TFed to 73 be available for a week or so, and we'll find out. This is what I would call an empirical[SUP]TM[/SUP] approach.

David

BTW Garo, I assume you intend to delegate your moderation here with the same frankness and humility as Chris.

Sigh.
Hasn't anyone heard of Kirchoff's Current Law?

@Aramis and chris2be8
A recycled exponent is one which is not at the leading edge of LL testing and has been assigned for LL testing before. The original assignee either didn't start the test or left it incomplete and has not updated the status of the exponent - i.e. not run Prime95 with the computer connected to the Internet - for 60 days. Such exponents are "expired" at 0000 UTC every day and thrown back in the LL pool for assignment. Anyone can get these exponents for testing, the list or potential assignees is the same as regular LL testers.

It doesn't really matter whether these exponents have had a partial test done since the new assignee will have to restart the LL test from scratch.

@chris2be8: Occasionally, people do continue testing but not communicate with Primenet. When they eventually report the result it is counted as a doublecheck. Not sure how often this happens but can't be more than 1 in every 10,000 first time tests.

[QUOTE=davieddy;362991]
If I were offered two 60M exponents, one at 72 and the other at 74 I would say "Why the **** were these not both TFed to 73?"
[/QUOTE]
Because this is a recycled exponent and when it was assigned for LL testing we were letting exponents go after TF to 72 bits. Now that we have more fire-power we will give it a more thorough rogering. Capisci?

[QUOTE=davieddy;362991]
Define sufficuent. If the bit level was deemed fit for a virgin, it should be fit for an old hag which has been round the block a few times.
[/QUOTE]
Sufficient is what the community decides. With George as primus inter pares. Currently, we feel that 74 bits is sufficient for 60-69M exponents even though James' site says that given infinite TF power we could take these to 75.

[QUOTE=davieddy;362991]
It is throwing good money after bad. It will get LLed eventually, and TFing a 54M (I ask you) from 72 to 73 does not add to the LL assignment pool. The effort would be MUCH better spent taking a 66M expo to 73[B].
[/B][/QUOTE]Why? And we will take the 66M to 74 bits anyway. Didn't you just agree that you would prefer a 54M over a 66M?


[QUOTE=davieddy;363056]Let us at least start from a point of concensus: ...
Now we come to the crunch: this depends on the number of virgins assigned each day.
[/QUOTE]
No it doesn't. Not even close. It depends on how many people complete an LL test each day. And that is it.

[QUOTE=davieddy;363056]
Currently this is decided by Chris, who has an ingenious ploy which ensures that the assignment of virgins does not exceed their production rate.: [/QUOTE]
I think you misunderstand what Chris is doing.

[QUOTE=davieddy;363056]
BTW Garo, I assume you intend to delegate your moderation here with the same frankness and humility as Chris.[/QUOTE]
Agreed. I usually don't moderate posts in discussions I am an active part of. I do release your posts from the moderation queue when I see them. Except when I find them pointless in which case I let another mod make the call.

[QUOTE=davieddy;363063]Sigh.
Hasn't anyone heard of Kirchoff's Current Law?[/QUOTE]
And pray tell what relevance does that have to our present discussion?

30 day account for LLs in progress
 

>63M:
8000 virgins in.
Big number out and same big number straight back in.

<63M
0 virgins in.
Big number out. 1500 opt for extra rogering and 16 [STRIKE]die of ecstacy[/STRIKE] are factored. Big number - 16 back in.

8500 LLs completed.

Who is winning this debate?

:davieddy::max:

This discussion seems to have gone into territory far more pointless than usual. I even made a couple of bizarre logical leaps there just to try and tie down an anchor point because the basic premise (virgin assignments are better/worse to LL/TF than recycled ones) is so ridiculous on it's face.

We need to TF the smallest ones we can to whatever our firepower can support (not exceeding James' line) and release them. The level we TF them to needs to be based on the amount of assignments going out vs the amount we can complete. Whether or not a number has been partially LL'd before doesn't even remotely matter - the number hasn't gotten any more or less likely to be prime.

We should always aim high, because if we fail to get numbers to that 74th bit and they release at 73 then it's the same as if we'd aimed for 73. If we start releasing them 2 bitlevels low that would be a problem, but we're not even releasing them 1 bitlevel low right now. I mean really, saying omg we should release at 73 because someday if we might fail to release at 74 and be forced to release at 73 is farcical. It's like preemptive failure so as to say that we didn't fail.

Sorry about any confusion Bill, but it's NMFP[SUP]MT[/SUP].
The delay to my posts causes mystification and embarrassment to everyone except me.

I posted the 30 day profit/loss accounts some hours ago. I expect they are being audited.

D

[QUOTE=davieddy;363117]Who is winning this debate?[/QUOTE]

Clearly not you.

(Sorry for the delay in the reply, but by your own rules I was not allowed to approve your post.)

[QUOTE=davieddy;363117]Who is winning this debate?[/QUOTE]

Based on the number of people you've been able to convince that your theories are correct, I'd say you are losing the debate badly.

[quote=Aramis]I mean really, saying omg we should release at 73 because someday if we might fail to release at 74 and be forced to release at 73 is farcical.[/Quote]

A superb and concise analysis of this whole mess.

[QUOTE=davieddy;363131]The delay to my posts causes mystification and embarrassment to everyone except me.[/QUOTE]

The delay to your posts is no one's fault but your own.

Indeed.

[QUOTE=kracker;363139]Indeed.[/QUOTE]
That's three times I told you to piss off and yet you are still here.

[QUOTE=Aramis Wyler;363127]I mean really, saying omg we should release at 73 because someday if we might fail to release at 74 and be forced to release at 73 is farcical.[/QUOTE]
[QUOTE=Prime95;363134]A superb and concise analysis of this whole mess.[/QUOTE]Good posting from the two people who come anywhere near to seeing their way through this mess[SUP]TM[/SUP] as clearly as I do.

The mess[SUP]TM[/SUP] arises entirely through George permitting Chris to put double-speak into practice: TFing 63M exponents while assigning 69M exponents for LL goes under the Orwellian "TF is just keeping ahead of the LL wavefront".

George knows precisely what I am aiming for, and how to achieve it easily and quickly. It is what Garo, Aramis and I refer to as "Breathing Space".

David

[QUOTE=davieddy;363156]It is what Garo, Aramis and I refer to as "Breathing Space"[/QUOTE]

Enjoy your breathing space, sir.

[QUOTE=chalsall;363158]Enjoy your breathing space, sir.[/QUOTE]

...and don't hold your breath for an early release this time.

[QUOTE=chalsall;363158]Enjoy your breathing space, sir.[/QUOTE]

For the record: A moderator gave davieddy a 6-month ban, not for the "breathing space" post, but rather a useless and crass response to kracker that was not approved.

"The sight is dismal;
And our affairs from England come too late:
The ears are senseless that should give us hearing,
To tell him his commandment is fulfill'd,
That Rosencrantz and Guildenstern are dead"

[QUOTE=Prime95;363165]For the record: A moderator gave davieddy a 6-month ban, not for the "breathing space" post, but rather a useless and crass response to kracker that was not approved.[/QUOTE]

I am curious to what he said, since I only said one word in my post regarding him.

[QUOTE=kracker;363183]I am curious to what he said, since I only said one word in my post regarding him.[/QUOTE]

#208....

[QUOTE=chalsall;363184]#208....[/QUOTE]

Ah I see. Well, I would have said the same to him... except someone already did it for me.

It is called the [URL="http://en.wikipedia.org/wiki/Golden_Rule"]Golden Rule[/URL]:
"One should treat others as one would like others to treat oneself."

[QUOTE=Aramis Wyler;363127]I mean really, saying omg we should release at 73 because someday if we might fail to release at 74 and be forced to release at 73 is farcical.[/QUOTE]

[QUOTE=Aramis Wyler;363127]It's like preemptive failure so as to say that we didn't fail.[/QUOTE]

[QUOTE=Prime95;363134]A superb and concise analysis of this whole mess.[/QUOTE]

Good god I wish I had thought of that way to say it. Very nicely said.

I always took the approach of "We can TF 63M - 999M to 73 before ever contemplating TF to 74" vs "We can TF the work being done TOMORROW to 74 and worry about bringing the rest up to 73 and then 74 later when we're there"


David: Your predictions of us being knee-deep in shit by the end of the month have been wrong. Repeatedly. Whatever your theory is must be completely wrong. I'm a bit too busy at the moment but I will go dig up these predictions if you deny having made them.

Aramis kind of brings up an old point of mine. How long do you want us to TF to 73 for until we bring them to 74? We're obviously close-ish to having the firepower to do 74, at the very least. This means we're almost capable of TF to 73 twice as many exponents as we can LL. How much of a buffer do you want?

At worst, we release a few 73's (cue Chris to tell us about Spidy's ripcords). At best we TF extremely efficiently, using every bit of our firepower not a week too soon.

In your lovely situation of having a 50,000 exponent buffer, we put loads of time into TF that only gets LL'ed six months down the line. Inefficient.

[QUOTE=TheMawn;363519]David: Your predictions of us being knee-deep in shit by the end of the month have been wrong. Repeatedly. Whatever your theory is must be completely wrong. I'm a bit too busy at the moment but I will go dig up these predictions if you deny having made them.[/QUOTE]

Thanks for your support, but please let it go.

Nothing to see here; everything is fine....

GPU from 72
 

It appears George is letting standard factor assignments go to 72 bits in the 70M range.

[QUOTE=chalsall;362501] We're currently taking anything available below 60M to 73, and anything above 62M to 74.[/QUOTE]
I'm pretty happy with how my assignments have been coming down lately. I have a batch file that checks my queues (2 queues) and if there is room in them it requests numbers from 50M to 62M to TF to 73, then if there is still room in the queues it requests numbers greater than 62M to take to 74.

The quote above ignores the 60M to 62M block, maybe because they're already all to 73 and we're not taking them to 74. Fine with me.

I'm posting though because I see in [URL="http://http://www.gpu72.com/reports/available/"]available assignments[/URL] that there are a couple thousand assignments sitting there in the 60 to 62M range at 73, and the yellow parts of the graph say those aren't released till 74. So what is it, should those be being release at 73, or should we be factoring them to 74?

At the time of this posting, there are only 371 assignments left to take from 73 to 74 between 62M and 65M, so soon we'll be taking the 65M+ range from 71 to 74.

[QUOTE=Aramis Wyler;363821]So what is it, should those be being release at 73, or should we be factoring them to 74?[/QUOTE]

Things change as our situation (including available firepower) changes.

Currently those between 60M and 63M only TFed to 73 which become available from Primenet are held to take to 74.

These also form Spidy's "rip-cord" / "safety valve" -- if any candidates are at risk of being released at below 74 in the 69M range, several below 69M are released at 73.