GPU to 72 status...

[Dubslow]

Originally Posted by chalsall

@petrw1: "This is curious?

Not terribly.

I haven't yet finished the code required to predict P-1 factors found.

It's much more complicated than predicting simple TFing.

But I am amused that someone finally noticed (or, perhaps, noticed and brought it forward).

Okay, now I'm "Not sure".

Predicted    Found
Not sure        27

Code sill in testing?

Edit: Also, I found the attached (right side). Perhaps a line break?

[James Heinrich]

http://www.gpu72.com/reports/workers/
I'm in 14th position in 13th place.

edit: it seems to be fixed now.

[chalsall]

Originally Posted by James Heinrich

http://www.gpu72.com/reports/workers/
I'm in 14th position in 13th place.

edit: it seems to be fixed now.

Yeah. The Rankings are in a seperate table, so there's a short temporal window where this can happen when someone's ranking changes before the script responsible finishes updating that table.

[flashjh]

I've had this happen quite a few times, but I finally am asking since I have the time.

In the attached capture, there is a 45M exponent that needs P-1, but I'm unable to get it. This happens with TF and P-1 exponents.

Does anyone know why?

[chalsall]

Originally Posted by flashjh

In the attached capture, there is a 45M exponent that needs P-1, but I'm unable to get it. This happens with TF and P-1 exponents.

Does anyone know why?

In order to clear out the low candidates, I had a conditional for P-1 such that if you had a history of less than 5 GHz Days / Day (not your case) or a candidate older than 15 days old (your case), you were limited to candidates at 52M or above. I have just removed this conditional.

There was a similar conditional for LLTF at 21 days, but this has not been active for about a week or so.

[flashjh]

Originally Posted by chalsall

In order to clear out the low candidates, I had a conditional for P-1 such that if you had a history of less than 5 GHz Days / Day (not your case) or a candidate older than 15 days old (your case), you were limited to candidates at 52M or above. I have just removed this conditional.

There was a similar conditional for LLTF at 21 days, but this has not been active for about a week or so.

I see. Thanks for the info -- that makes sense now. I actually don't mind that conditional, and it probably makes sense. If the assignment page had said that instead of that there weren't any exponents that met my conditions, I would have understood that I need to clear some older stuff out first ;)

Thanks again

[petrw1]

Less than 6 days of work left according to this:
http://www.gpu72.com/reports/estimated_completion/

Plus whatever new spidey gets in the next 6 days.

[chalsall]

Originally Posted by petrw1

Less than 6 days of work left according to this:
http://www.gpu72.com/reports/estimated_completion/

Not really...

That report shows the estimated completion across both ranges based on the 30 day average of all TF work the system is coordinating. Thus, we would only clear out what we currently hold in the DC range in six days if everyone did only DC work.

But, please don't do that!!! So everyone knows, we are currently approximately 220 days ahead of the DC wavefront. This is why "bcp19" AKA "Pete" has moved most of his fire power over to the LL range.

[petrw1]

Originally Posted by chalsall

That report shows the estimated completion across both ranges based on the 30 day average of all TF work the system is coordinating. Thus, we would only clear out what we currently hold in the DC range in six days if everyone did only DC work.

OK, so that explains why, even though the estimated completion is dropping is drops at a rate less than 1 day per day.

[bcp19]

Originally Posted by chalsall

Not really...

That report shows the estimated completion across both ranges based on the 30 day average of all TF work the system is coordinating. Thus, we would only clear out what we currently hold in the DC range in six days if everyone did only DC work.

But, please don't do that!!! So everyone knows, we are currently approximately 220 days ahead of the DC wavefront. This is why "bcp19" AKA "Pete" has moved most of his fire power over to the LL range.

Actually, I switched most of my effort to LL due to the fact that my heavyweight systems are less efficient for factor finding at DC ranges than LL. The one machine would only break even at 37M-38M for 70 bits, compared to 32M-33M for the ones still crunching. (and the HD5770 is technically efficient to 72 bits at 30M, since it cannot do LL yet)

[Bdot]

Originally Posted by bcp19

(and the HD5770 is technically efficient to 72 bits at 30M, since it cannot do LL yet)

One more reason to release my next version of mfakto soon

Chalsall, that's an interesting way to get localized web pages, cool! Sometimes it is hard to pick the correct one of several possible translations if you only have the phrase without context. Such as "Saved" (GHz-Days) that is now "Gespeichert" (which means you saved it to hard disk, for instance) instead of "Gespart" in German. Is there a way to object to some of the translations? Also, translating "GPU to 72" at all is questionable, but "GPU hoch 72" means "GPU to the power of 72", which is kind of funny (was that intended?).

For me personally, localization has a rather low priority (I have US-English keyboards, and sometimes have difficulties understanding German error messages of the Operating system at friend's PCs.) Therefore, if you're really looking for ideas what to implement, I have one:

I'd find it helpful if I had a single comments field on the "Get Assignments" pages that would save the entered comment for each of the assigned tasks, and display it on the "View Assignments" pages. I would use that field to enter the machine names that I requested the assignment for, making it much easier to check why there are still some old ones hanging around. If the "View Assignments" were sortable by the comments, this feature could also be used to check if machines are running dry without logging in to them.

And maybe, once your submission spider reserves work for mfakto , it could automatically insert the machine name (ComputerID) into the comments ...

BTW, for portability reasons I now decided to implement the file locking using an empty .lck file (worktodo.txt.lck and results.txt.lck). As long as it exists, mfakto will not touch the corresponding txt file.