GPU to 72 status...

[axn]

Originally Posted by TheMawn

I grabbed some LL TF to 73 bits (everything default, asked for 100 assignments) and the preview showed 70,002,XXX from 72 to 73, but I got 66M 71 to 73.

Originally Posted by chalsall

With George's knowledge, I brought in everything in the 70M range not yet assigned and not yet TFed to at least 72 bits. This was to prevent anything below 72 bits from being assigned for P-1 work.

Chris, I think you missed the part where the preview window showed one thing, but the actual assignment ended up being different.

[chalsall]

Originally Posted by LaurV

In this case, would you please extend these tables vertically?

Done.

[chalsall]

Originally Posted by axn

Chris, I think you missed the part where the preview window showed one thing, but the actual assignment ended up being different.

I didn't miss it, but forgot to speak to it.

The 70M range was brought in, but not made available for assignment (with one user's exception, which George, Garo and Jerry was aware of). I made a SPE, and didn't change the preview code.

Sorry about that -- this was a precautionary action I undertook. Based on the recycling rules clearly documented at the Mersenne.org Assignment Rules page, the "starvation" of Category 4 candidates is about to end.

[kladner]

.....the "starvation" of Category 4 candidates is about to end.

Does this mean that we will return to trial factoring to 74 where appropriate?

[chalsall]

Originally Posted by kladner

Does this mean that we will return to trial factoring to 74 where appropriate?

Yes.

Probably in about two weeks time, as resources allow.

[Aramis Wyler]

I had my gall bladder removed a couple of weeks ago, and was in the hospital for a week before that getting tests done, and so I missed some things regarding what was being factored, and now I see my computer started pulling numbers in the 70M range because they were lowest factored. Sigh. Since we seem to be having some sort of crisis I don't honestly understand about numbers in the 'high 69M range' I grabbed 500 assignments > 69.5M to put to 73. 500 because that is about a week and a half worth of work, and am releasing any other assignments I have.

I am having trouble understanding the exact nature of the category 4 problem, but I hope this helps. When I tried to reserve 'let gpu72 decide' on the web site it gave me 66M numbers to factor from 73 to 74, and that was not the behavior I expected from the above posts.

[chalsall]

Originally Posted by Aramis Wyler

I am having trouble understanding the exact nature of the category 4 problem, but I hope this helps.

It will indeed! Thanks!

The "Cat 4" problem is simply that I didn't appropriately appreciate just how much "churn" occurred for a large amount of participants. I thought we had enough of a buffer for George to activate the new assignment rules, not realizing that for the first two months there would be ~1,000 assignments requested per day for "Cat 4", rather than 300 (the number of assignments actually completed).

The good news is as of tonight, we're now two months into the new assignment rules. Thus, the original recycling rule of "60 days without report, no assignment" will kick in, and a great many candidates will be recycled, and within a week or two the pressure will be off.

Originally Posted by Aramis Wyler

When I tried to reserve 'let gpu72 decide' on the web site it gave me 66M numbers to factor from 73 to 74, and that was not the behavior I expected from the above posts.

You should have only seen that if you changed the default pledge level from 63 to 64. And, by the way, "Let GPU72 Decide" is only available through MISFIT -- on the GPU72 manual assignment web page the option is "What Makes Sense" -- subtly different.

Lastly, the reason you were given some 70Ms to process is because you've configured MISFIT to use option #1 -- Lowest TF level. The 70M range was not actually intended to be exposed, but I made a mistake in not changing the preview code, which someone pointed out.

Sorry for the f' ups on my end. (And, as an aside, this is a "tough and exacting crowd"! Just the way I like it!!! )

[Aramis Wyler]

In fact, I do know the difference between the 'what makes sense' and 'Let Gpu72 decide' options are, and both are available on the get work form for lltf, and I stand by the statement you quoted.

[chalsall]

Originally Posted by Aramis Wyler

In fact, I do know the difference between the 'what makes sense' and 'Let Gpu72 decide' options are, and both are available on the get work form for lltf, and I stand by the statement you quoted.

Ah hell... Sorry; you are correct and I was wrong. It is sobering to be to told how your own software works...

Please let me know if you see any other issues.

[Aramis Wyler]

lol! fwiw it's pretty cool how the let gpu decide options so clearly disables the irrelevant field, nice work on that.

New info though, purely coincidental. While I picked up the new assignments and put them in the queue, exactly 1 70M number finished on each video card. The spider does not appear to have picked up these results, which makes sense if they weren't supposed to be assigned.

I'm not having a hate on challsall day, though, it's just I'm finally off all the narcotics and so I'm going through about a month's worth of data on my end.

[chalsall]

Originally Posted by Aramis Wyler

I'm not having a hate on challsall day, though, it's just I'm finally off all the narcotics and so I'm going through about a month's worth of data on my end.

Not a problem. I know I'm "Only Human", and so often make stupid mistakes. Feedback is always welcome and appreciated.

So you know, the GPU72 side "Observing Spider" *IS* watching the 70M range, so it should have picked up the submission to Primenet within 30 minutes. It does a "full scan" at 10 and 40 minutes after each hour.

Please PM me the candidates in question, and I'll drill down.