GPU to 72 status...

[chalsall]

Originally Posted by kladner

Am I missing something?

Remember that GPU72 doesn't check as frequently for LL/DC completions; it can now take a couple of hours or so.

[nucleon]

Just a heads up...

I've shifted to doing "to 2^73". Currently about 60% of my TF capacity is on 2^73, 40% is still doing 2^72. That should migrate to 100% over the week.

If Xyzzy keeps doing "to 2^72" (or less), I'll expect Xyzzy to get ahead of me in the saved stat.

-- Craig

[Xyzzy]

Why the jump to 73?

[kracker]

[LaurV]

Warning:

• 2 of your Double Check assignments have been completed by another worker. These are highlighted in red below with a double-dagger (‡).It would be best if you stopped working on these and unreserved them, as the work is no longer needed.
  

That was freaking me forgetting to check if I am logged in when I reported the results ... I think that this would always be the case when the "reportee" is "anonymous". Maybe you can teach your spidy about that.

[kracker]

...
Me too.

Warning:
1 of your Lucas-Lehmer assignments have been tested by another worker. These are highlighted in red below with a dagger (†).Please understand when you turn in the results these will be credited as a Double Check by PrimeNet.

I DID log in. I tried submitting it again and it said I already submitted it...

[nucleon]

Originally Posted by Xyzzy

Why the jump to 73?

Couple of reasons.

1) 2^73 work is less micro-management. Less result lines submitted, less time I need to distribute work etc..

2) It's more efficient to do 2^71-2^73, in one hit than 2^71-2^72, then 2^72-2^73. We're coming fast to exhaust most of the work to 2^72. I'd thought I'd get in and get some efficiency gains (although pretty minor).

-- Craig

[chalsall]

Originally Posted by LaurV

That was freaking me forgetting to check if I am logged in when I reported the results ... I think that this would always be the case when the "reportee" is "anonymous". Maybe you can teach your spidy about that.

Should I not warn people if Anonymous completes an assignment? I think it's better to warn people, and if they mistakingly submitted the result without being logged in, at least they would know. And if they were "poached", they could stop work which wasn't needed.

Originally Posted by kracker

I DID log in. I tried submitting it again and it said I already submitted it...

Sorry -- stupid Spider Trainer error. I didn't tell Spidy your correct PrimeNet display name.

[LaurV]

Second thought, ye right...

[kracker]

Originally Posted by chalsall

Sorry -- stupid Spider Trainer error. I didn't tell Spidy your correct PrimeNet display name.

You're not the only one. "cracker? kwacker? Oh, sorry, it's kracker?"
I should have chosen something else....

[davieddy]

Originally Posted by nucleon

Just a heads up...

I've shifted to doing "to 2^73". Currently about 60% of my TF capacity is on 2^73, 40% is still doing 2^72. That should migrate to 100% over the week.

If Xyzzy keeps doing "to 2^72" (or less), I'll expect Xyzzy to get ahead of me in the saved stat.

-- Craig

Originally Posted by Xyzzy

Why the jump to 73?

Originally Posted by nucleon

Couple of reasons.


1) 2^73 work is less micro-management. Less result lines submitted, less time I need to distribute work etc..

2) It's more efficient to do 2^71-2^73, in one hit than 2^71-2^72, then 2^72-2^73. We're coming fast to exhaust most of the work to 2^72. I'd thought I'd get in and get some efficiency gains (although pretty minor).

-- Craig

Assuming you are TFing "virgin" exponents, this is precisely what I
have been advocating for a couple of years.

To keep up with the rate of completion of LL tests,
the wavefront needs to advance by at least that.

When it was at 53M, we judged (accurately it is now clear)
that the GPU capacity could manage 72 bits for 200 NEW
LL assignments per day, hence the name "GPUto72".

About 800 LL tests are abandoned per day, so spidy could
TF them on a "breadth first" basis to 2 bits fewer.

But the sprawling breadth first approach ahead of the wavefront
just poaches the low-hanging fruit.
One would rather do 69 to 73 than 72 to 73.

The "work saved" metric is deceptive.
If that was the criterion, everyone would be TFing
exponents around a billion.

The purpose is to ensure that no new LLs are assigned
inadequately TFed.

David