Thanks to the good folks at Primegrid, I've now found a way to force P-1 testing, and to determine the B1 and B2 value by yourself, rather than letting Prime95 do it. Would anyone by chance know what B1 value is safe to use, to avoid untested gaps, when I have a candidate presieved to p=136T? Is there any place to see what bit depth a given kind of value of B1 compares to? Anyway I'll get back to you when I've done some testings on my Quad, such that it can be determined weather or not anything is to be gained on Pminus1 at current testlevel.

Take care

Kenneth

Pfactor worktype determines bounds for you. Pminus1 requires bounds as input. I can't help you with the actual calculations, but that's the difference between the two. [url=http://mersennewiki.org/index.php/Worktodo.txt]Syntax[/url]

I had already done some testing. Prim95 will refuse to run P-1 on these exponents given the amount of TF that has already been run. You can trust the program's calculations**

136T is approximately 47 bits. At this stage, there is really no point in doing P-1, since the expected probability of success will be very less. And as you sieve higher, this only becomes worse.

** Actually, you can't trust the programs calculation for bases other than 2, due to a bug. I had to use an equivalent base 2 test to get reliable probability calculation.

PS:- TF depth and P-1 bounds are not really directly comparable.

[QUOTE=axn;281676]I had already done some testing. Prim95 will refuse to run P-1 on these exponents given the amount of TF that has already been run. You can trust the program's calculations**

136T is approximately 47 bits. At this stage, there is really no point in doing P-1, since the expected probability of success will be very less. And as you sieve higher, this only becomes worse.

** Actually, you can't trust the programs calculation for bases other than 2, due to a bug. I had to use an equivalent base 2 test to get reliable probability calculation.

PS:- TF depth and P-1 bounds are not really directly comparable.[/QUOTE]

OK, thanks, I'll not do any testings then, instead I'll continue as planned using NewPGen until I reaches p=2,000T or optimal sievedepth if it should be lower than p=2P (~51 bit). Sievespeed is currently ~30.4M p/sec per core, wich is equivalent to ~10.2T/day on a Q6600 2.4GHz.

[QUOTE=axn;281676]
PS:- TF depth and P-1 bounds are not really directly comparable.[/QUOTE]
I apologize if this is out of context.

V4 used to admire itself on the way LL was decently ahead of DC, and TF was decently ahead of LL.

TF is only sensibly done on GPUs these days.
P-1 is no problem when a competent CPU
embarks on a 40-day LL.

Tell me a LL assignment that has been optimally TFed.
(Other than the ones I do that Eric kindly goes OTT on his modest GPU)

David

BTW P-1 and TF only overlap by ~1/3,
and pausing TF one bit short is the most
misguided idea ever proposed here*.
(Basic reason being that most of the time
neither will find a factor)
[B]*And that is saying something.[/B]

Sieving is complete to p=1000T, the removal rate is about 188 minutes/factor and there is a total of 81882 candidates remaining. Continuing to p=2000T. By the time I get to p=2000T hopefully something usefull and productive supporting AVX has been released (I'm here thinking in terms of LLR or PFGW).

Take care

KEP

Ps. If anyone feels like helping out with the search for borderprimes around the power of 10, go to [URL="http://www.worldofnumbers.com/borderprp.htm"]http://www.worldofnumbers.com/borderprp.htm[/URL] (or try to PM me) and see how to help out. If you care to join the MegaPRP search, then please send me a PM.

[QUOTE=KEP;291404]Sieving is complete to p=1000T, the removal rate is about 188 minutes/factor and there is a total of 81882 candidates remaining. Continuing to p=2000T. By the time I get to p=2000T hopefully something usefull and productive supporting AVX has been released (I'm here thinking in terms of LLR or PFGW).

Take care

KEP

Ps. If anyone feels like helping out with the search for borderprimes around the power of 10, go to [URL="http://www.worldofnumbers.com/borderprp.htm"]http://www.worldofnumbers.com/borderprp.htm[/URL] (or try to PM me) and see how to help out. If you care to join the MegaPRP search, then please send me a PM.[/QUOTE]Just interested: why do you use 1000T and 2000T instead of 1P and 2P respectively?

[QUOTE=xilman;291412]Just interested: why do you use 1000T and 2000T instead of 1P and 2P respectively?[/QUOTE]

Simple question, with a simple answer:

I use 1000T in stead of 1P and 2000T in stead of 2P, because I like to know myself what I'm talking about and since I'm not aware what "P" is short for, I prefer to use "T" for Trillion. But enlighten me and I might just change my habbit :smile:

Thanks for your question. Take care

KEP

That's a SI prefixe. Please study this: [url]http://physics.nist.gov/cuu/Units/prefixes.html[/url]

[QUOTE=pinhodecarlos;291420]That's a SI prefixe. Please study this: [url]http://physics.nist.gov/cuu/Units/prefixes.html[/url][/QUOTE]

OK thanks, so to sum up:

Sieve depth is at p=1P (1 Peta) and continues to p=2P (2 Peta).

Thanks for teaching me something new :smile:

KEP!

[QUOTE=KEP;291423]Thanks for teaching me something new :smile:[/QUOTE]You're welcome.


With luck, you're not the only one here to have learned something new. The opportunity to educate myself is one of the reasons I spend so much time on the forum