Exponent expiration

[Mr. P-1]

Originally Posted by Christenson

75M range is fairly high... you might want to join Mr P-1, me, and some others in getting a little better TF job done on exponents that are about to be assigned to LL tests.

Clarification: We're TFing exponents which are assigned to LL testing, but which we are catching when they expire. See the discussion in this thread from post #22 onward.

[Dubslow]

Semantically speaking, what Christenson said was true, he just didn't mention where they're coming from. But it doesn't really matter.

[davieddy]

Originally Posted by Mr. P-1

Clarification: We're TFing exponents which are assigned to LL testing, but which we are catching when they expire. See the discussion in this thread from post #22 onward.

It is proving to be an instructive/entertaining exercise!

We are going nowhere near poaching. OTOH if you were LL testing
a number which could have been TFed further, at what % stage would
someone finding a factor make you thank/curse him/her?

David

PS Could someone make this a poll?

[Mr. P-1]

Originally Posted by davieddy

We are going nowhere near poaching. OTOH if you were LL testing a number which could have been TFed further, at what % stage would someone finding a factor make you thank/curse him/her?

That's an interesting question. I'd probably thank them any time up to 50% and curse them from about 80% onward. But then I'd thank them if My LL was complete, and the exponent was coming up for a doublecheck.

But that's an emotional response and extremely irrational. Factorising an exponent assigned for first-time testing will save between 1 and 2 LLs on average, and closer to 2 than 1 because many exponents are assigned but not yet started. Taking that analysis to its logical conclusion we should prioritise factoring first-time-tests assigned to other people to 69 bits over the same exponent unassigned (or assigned to us) to 70 bits.

I don't propose we do this. Sometimes irrational emotions prevail over logic.

How do you think Rosalind Franklin felt about Watson and Crick beating her to the structure of DNA having "poached" her data?

[davieddy]

Originally Posted by Mr. P-1

How do you think Rosalind Franklin felt about Watson and Crick beating her to the structure of DNA having "poached" her data?

That's another interesting question.
At least her name is now well-known, and one would hope that
her deepest satifaction derives from her contribution to the greatest
biological discovery since Darwin or Adam.

Newton said something about standing on the shoulders of giants.
Relativity wasn't "Out of the blue".

OTOH "How can 60 carbon atoms be neatly arranged into a molecule?"
would not be a problem for anyone who had watched football since the
world cup in 1970, made a truncated icosahedron out of cardboard
or approximated a sphere for a 3D engine.

David

[Christenson]

But Mr P-1...
#1: At least one poster on this forum has complained that his first time check completed, but someone had TF'ed his exponent. Finding out about a TF in the middle of working on an exponent could only happen to Davieddy, who asks me to TF his LL assignments at the same time he LL's them, and therefore has a result (so far, all negative) show up in a way he knows about during the middle of his LL check.

#2: Now, Ckdo and crew have been doing TF in the LL-D range, and periodically finding factors AFTER (a year or three after) the first LL check. We don't seem to get complaints.

#3: The eventual result of grabbing all the 68-bit factored LL assignments and giving them a TF to 69 bits, is that we will run out of 68-bit TF'ed LL assignments, either because the ones that come back incomplete will all get touched, or because we will get the server to give us assignments just in front of the wavefront, or the wavefront will catch up with the 53-60M TF'ers that Mr Davie is so upset about.

I need to follow xyzzy to Newegg for a UPS so the home machine doesn't need rebooting quite so often...

[davieddy]

Originally Posted by Christenson

Finding out about a TF in the middle of working on an exponent could only happen to Davieddy, who asks me to TF his LL assignments at the same time he LL's them, and therefore has a result (so far, all negative) show up in a way he knows about during the middle of his LL check.

You know perfectly well that my request preceded the start of my LL.
My typo delayed you, but had you found a 74 bit factor in a couple
of days I would have been extremely grateful to hear about it.
Mystic Meg told me you wouldn't!

David

35% through ATM

[Mr. P-1]

Originally Posted by Christenson

#3: The eventual result of grabbing all the 68-bit factored LL assignments and giving them a TF to 69 bits, is that we will run out of 68-bit TF'ed LL assignments, either because the ones that come back incomplete will all get touched, or because we will get the server to give us assignments just in front of the wavefront, or the wavefront will catch up with the 53-60M TF'ers that Mr Davie is so upset about.

I had a look at what the server is handing out toward the end of the day (21:30 - 21:45 UTC) by anonymously requesting manual assignments.

Test= assignments are in the low 55Ms. All appear to have been P-1ed. Most are TFed to 71 bits, but a few are only at 69.

Pfactor= assignments are in the low 58Ms. All appear to have been TFed to 71 bits.

Factor= assignments are strange. When I asked for a single one, I got one in the low 55Ms, 69-70 bits. When I asked for a batch, the first one in the batch was like this, and all the others were in the high 59Ms 70-71 bits. The next batch was the same. I tried two more batchs, but these were all 59Ms 70-71 bits, no 55Ms.

I then unreserved these assignments.

Over the past 25 minutes (from 22:15 - 22:40 UTC), I tried repeating the experiment with Factor= assignments, alternating between requesting a single one, and a batch. The first one - a single - was a 57M 69-70 bits. All others, both singles and batches have been 59Ms 70-71 bits.

Hypothesis: Even when logged out, the server tracks what an individual is doing, perhaps by IP. If the user only asks for a small number of Factor assignments, the server assumes that these are destined for a CPU and so hands out an exponent in the Test-range 69-70. If the user starts requesting heaps and heaps of them, then the server assumes that they're destined for a GPU, and hands out a 59M 70-71 exponent.

[Dubslow]

It seems that assuming it's a GPU would fast-track it for lagging assignments, like low 50's that are only at 69 bits, or 40's at 68.

Edit: That's what the GPU should be assigned, rather than 59M.

[davieddy]

Originally Posted by Mr. P-1

I'd thank them if My LL was complete, and the exponent was coming up for a doublecheck.

Originally Posted by Christenson

Ckdo and crew have been doing TF in the LL-D range, and periodically finding factors AFTER (a year or three after) the first LL check. We don't seem to get complaints.

I would like to have seen my residue confirmed.
This is one of several incentives to do DCs.

David

[Mr. P-1]

Originally Posted by davieddy

That's another interesting question.
At least her name is now well-known,...

Not as well known as Watson and Crick. And she missed out of the Nobel Prize.

and one would hope that her deepest satifaction derives from her contribution to the greatest biological discovery since Darwin or Adam.

One would hope so. But she's still human.

Newton said something about standing on the shoulders of giants.

That was a oblique swipe at Robert Hooke, who was a short man.

Relativity wasn't "Out of the blue".

No, but Einstein didn't swipe anyone's unpublished data. And to the extent that Einstein's work was allegedly plagiarised, that is viewed as counting against Einstein.

OTOH "How can 60 carbon atoms be neatly arranged into a molecule?"
would not be a problem for anyone who had watched football since the
world cup in 1970,

You think? An awful lot of people have watch football, but none of them solved the problem.

made a truncated icosahedron out of cardboard
or approximated a sphere for a 3D engine.

Perhaps, but there are fewer such people than there are fans of football. (That's soccer, for those in the forum who think football is a game in which the ball is carried by the players' hands.). The geometers would have the best chance, I would think, of connecting their work to the C60 problem: "What shape has 60 vertexes, each with four adjoining edges?". The computer modelers are thinking about the problem from the wrong end, i.e., their starting point is the sphere, and their end point is the net that approximates it. The C60 problem starts with the number and connectivity of the vertexes.

The football fans have probably never given a moment's thought to the geometry of the pattern on the ball.