Possible extention to the "GPU to 72 Tool" project?

[bcp19]

Is there a reason some exponents do not show low level TF's done? I was checking on an assignment of mine and I see:

50374127 No factors below2^71
Assigned LL testing to "GPU Factoring" on 2011-11-01
History no factor for M50374127 from 2^69 to 2^70 [mfaktc 0.17-Win barrett79_mul32] by "bcp19" on 2011-11-11
History no factor for M50374127 from 2^70 to 2^71 [mfaktc 0.17-Win barrett79_mul32] by "bcp19" on 2011-11-11

Where others have things like:

50374229 No factors below2^69P-1B1=600000, B2=17550000
Unverified LL948ED3435F9883__ by "exhopp" on 2011-02-16
History no factor from 2^63 to 2^64 by "Joel McBeth" on 2009-04-08
History no factor from 2^64 to 2^65 by "Joel McBeth" on 2009-04-09
History no factor from 2^65 to 2^66 by "Joel McBeth" on 2009-04-10
History no factor from 2^66 to 2^67 by "Joel McBeth" on 2009-04-11
History B1=600000, B2=17550000 by "S485122" on 2009-04-15
History no factor from 2^67 to 2^68 by "ANONYMOUS" on 2009-04-20
History no factor from 2^68 to 2^69 by "Sturle Sunde" on 2009-04-23

Is there a reason why so many have nothing shown prior to ^69?

[petrw1]

Originally Posted by bcp19

Is there a reason some exponents do not show low level TF's done? I was checking on an assignment of mine and I see:

50374127 No factors below2^71
Assigned LL testing to "GPU Factoring" on 2011-11-01
History no factor for M50374127 from 2^69 to 2^70 [mfaktc 0.17-Win barrett79_mul32] by "bcp19" on 2011-11-11
History no factor for M50374127 from 2^70 to 2^71 [mfaktc 0.17-Win barrett79_mul32] by "bcp19" on 2011-11-11

Where others have things like:

50374229 No factors below2^69P-1B1=600000, B2=17550000
Unverified LL948ED3435F9883__ by "exhopp" on 2011-02-16
History no factor from 2^63 to 2^64 by "Joel McBeth" on 2009-04-08
History no factor from 2^64 to 2^65 by "Joel McBeth" on 2009-04-09
History no factor from 2^65 to 2^66 by "Joel McBeth" on 2009-04-10
History no factor from 2^66 to 2^67 by "Joel McBeth" on 2009-04-11
History B1=600000, B2=17550000 by "S485122" on 2009-04-15
History no factor from 2^67 to 2^68 by "ANONYMOUS" on 2009-04-20
History no factor from 2^68 to 2^69 by "Sturle Sunde" on 2009-04-23

Is there a reason why so many have nothing shown prior to ^69?

One likely explanation is that history was only stored and available since the new V5 server went it in Oct 2008

[petrw1]

Originally Posted by KyleAskine

Also, I wonder when more are going to be available. Is it just as exponents are TF'ed to 72, or is there a manual step in there?

I get nervous when I have under a couple day's worth queued up.

As I understand it from the first post by Chalsall; he is nabbing all exponets in the 45-55M range as LL tests expire (without completion) and making them available to more TF bits (71 or 72) before they are released back for LL.
Based on that more and more should appear each day.

I would further assume that if the day comes that this range is close to exhausted that it would be extended beyond 55M

[KyleAskine]

I meant P-1's. There are still tons to TF, as far as I can tell.

[chalsall]

Originally Posted by KyleAskine

I meant P-1's. There are still tons to TF, as far as I can tell.

The exponents available for P-1 work (by default) are those who have been TFed by GPUs up to 72 (or above) "bits". Therefore, the number available is a function of the work being done by GPUs (plus any expiring exponents my spider happens to grab which have already be TFed to 72 or higher with no P-1 work already done).

However, P-1 Workers can ask for exponents TFed to lower bit levels with the caveat that after the P-1 work has been completed the exponent will be re-assigned to a GPU Worker to take it up to 72 (unless, of course, a factor is found).

Mr. P-1 said he did not want to take any P-1 work where additional TF work would be done after, but other workers may not have that criteria.

[chalsall]

Originally Posted by KyleAskine

It looks like you might not be picking up when P-1s are finishing. I finished M47895521 a while ago, and it isn't showing up on the website.

My spider is picking up when the P-1 work has completed. However, there is a second step where a separate script updates the "Assigned" SQL table and calculates the GHz Days credit.

I had hoped to get this done tonight, but the code to calculate GHz Days (thanks again to James) for P-1 is quite complex, and my meetings today lasted a lot longer than expected.

Also, "She who Must be OK'ed" demands Friday nights to be "together time"...

Originally Posted by KyleAskine

It also took it to lower bounds than I would have liked. B2 was under 10M, when I am actually around 20M a lot. Is this because of the higher than usual TF?

I really cannot speak to that. The "PFactor=..." line(s) is such that Prime95/mprime should make the decision on the optimal B1/B2 bounds as a function of how far the exponent has been TFed.

Perhaps someone with more P-1 experience can speak to this?

[Dubslow]

I'm pretty sure Prime95 weighs the overall chance of finding a factor for both TF and P-1 when calculating the bounds. So if it's TF deeper, it has to do less P-1 to get the same chance of finding a factor. That's the way James' website is laid out, so that's my guess.

[petrw1]

Originally Posted by chalsall

My spider is picking up when the P-1 work has completed. However, there is a second step where a separate script updates the "Assigned" SQL table and calculates the GHz Days credit.

I apologize if this sounds like a dumb question but can't your spider also "spider" for the GHz Days that the Prime95 Server calculated?

[KyleAskine]

Originally Posted by chalsall

My spider is picking up when the P-1 work has completed. However, there is a second step where a separate script updates the "Assigned" SQL table and calculates the GHz Days credit.

I had hoped to get this done tonight, but the code to calculate GHz Days (thanks again to James) for P-1 is quite complex, and my meetings today lasted a lot longer than expected.

Also, "She who Must be OK'ed" demands Friday nights to be "together time"...



I really cannot speak to that. The "PFactor=..." line(s) is such that Prime95/mprime should make the decision on the optimal B1/B2 bounds as a function of how far the exponent has been TFed.

Perhaps someone with more P-1 experience can speak to this?

I want to thank you for everything you are doing for this project. It is really great! I hope I didn't come off as ungrateful!

[KyleAskine]

Originally Posted by Dubslow

I'm pretty sure Prime95 weighs the overall chance of finding a factor for both TF and P-1 when calculating the bounds. So if it's TF deeper, it has to do less P-1 to get the same chance of finding a factor. That's the way James' website is laid out, so that's my guess.

This is my assumption as well. If this is the case, then I think that we might need to change something, since the stated goal of this project is to try to cull out more exponents so we have less to LL. Of course, if we ever get a huge glut of P-1s and we are not keeping up, then perhaps we should change this.

Then again, if we can simply give every exponent any P-1, I guess we are making a difference! So maybe what we are doing is good!

[chalsall]

Originally Posted by KyleAskine

I want to thank you for everything you are doing for this project. It is really great!

Thank you for the kind words. And thank you (and all the other workers) for using it.

Originally Posted by KyleAskine

I hope I didn't come off as ungrateful!

Not at all.