Linux newpgen pfgw llr

[jasong]

My topic title is more for searchers than for people just browsing the forum. Just thought I'd mention that. :)

I'm using snewpgen to sieve a file, and there are a couple things that annoy me. First, why can't we pick the prime-finding program the output gets sent to? And second, snewpgen only prints the time per factor when a factor is actually found, which sucks when there's 20 or more minutes between factors. It's basically a crap shoot figuring out when to stop.

Now that I've vented, it's time to be productive. :) Has anyone ever tried changing the name of the Linux llr program to pfgw to see if things would automatically switch over that way?

[paulunderwood]

First, why can't we pick the prime-finding program the output gets sent to?

Jean's LLR program did not exist when NewPGen was written by Paul Jobling.

And second, snewpgen only prints the time per factor when a factor is actually found, which sucks when there's 20 or more minutes between factors.

If you are really looking at "20+ mins" you ought to keep a daily manual log of your sieving progress. (For 321 our finishing depth was about 1GHz Athlon day.) I recall NewPGen does its running average based on the last 20(?) factors found. So you would have to take into account that this average might not be sensitive to fluctuations in the time to find a factor; there might be reboots too and also other processes running on your computer, slowing NewPgen down. Finally different chips have different performances; It was well understood that Athlons were great for sieving and Pentium4s were better at LLR'ing.

It's basically a crap shoot figuring out when to stop.

Somewhere in this forum is a program written by Thomas Ritschel that calculates the optimal cutoff for NewPGen for one-pass LLR/PRP.

Now that I've vented, it's time to be productive. :) Has anyone ever tried changing the name of the Linux llr program to pfgw to see if things would automatically switch over that way?

My guess is that this would do. Have you tried it?

[jasong]

Hmmmm, I thought I responded to this thread. I must not have ever actually submitted the message.

Anyway, thanks for the response, and I've decided to use the logging feature, like you recommended.

[pacionet]

I tried to run pfgw (both builded by myself both the executable) but I got an error : it block and after about 40 minutes it output a pfgw_err.log saying contact ... for bugs , try the -a1 flag .

Also with -a1 flag the program failed at the first iteration....

[ET_]

Originally Posted by pacionet

I tried to run pfgw (both builded by myself both the executable) but I got an error : it block and after about 40 minutes it output a pfgw_err.log saying contact ... for bugs , try the -a1 flag .

Also with -a1 flag the program failed at the first iteration....

Would you mind telling us the kind of search you are doing, so that we may help you track the problem?

Luigi

[pacionet]

I ran pfgw this way:

> ./pfgw factorial.out

where factorial.out contains output of newpgen:

ABC $a!3$b // MF Sieved to: 4741603759
[....]

On Windows no problems at all.
On Linux (Red Hat) I got an error like the one reported here

http://tech.groups.yahoo.com/group/o...w/message/1722

after about 40 minutes (no progress is shown) (The
following output is not my error but my error is the same)


*** WARNING *** : An error was detected in the reciprocal routine.
If this occurred during normal use, contact bugs@...,
including the number you were testing to produce this error,
and disregard the results of tests on this number.
If this error occurs in optimization mode, you may ignore it.
PRP: 17*2^(429318+1)-197*2^202534-1 1/429323 mro=0.5 sum=5507.75/0.00
Iteration: 1/429323 ERROR: SUM(INPUTS) != SUM(OUTPUTS),
9 != 9
(Diff=0 max allowed=5508)
(Test aborted, try again using the -a1 switch)

With the -a1 switch I got a similar behaviour and a similar error.

Is a critical bug ?

[paulunderwood]

Does the solution given in the recent thread http://tech.groups.yahoo.com/group/o...w/message/1882 solve your problem?

Here is another link:
http://tech.groups.yahoo.com/group/p...m/message/8452

(Out of my depth.)

[pacionet]

Disabling NX or XD in the BIOS ?

I have no root privileges to try this solution...

[paulunderwood]

The only work-arounds I can see are: crunch the (few) problematic numbers on a box that copes with them; or try compiling PFGW (from the available source) statically, and maybe using an older version of GMP.

[pacionet]

Static compiling didn't solve the issue.
Maybe I'll try with an older version of GMP.

[paulunderwood]

Try compiling some "older versions" from
http://tech.groups.yahoo.com/group/openpfgw/files/
(membership required)