PFGW 3.8.3 (with gwnum v28.7) Released

[rogue]

Originally Posted by rogue

I've released PFGW 3.7.3 over at sourceforge. A problem was found when using -f that caused it to crash. That problem was fixed.

Please do not use pfgw 3.7.3 with the -f switch. It appears to be missing factors. I'm getting some help in tracking down the problem.

[Batalov]

Mark,
I have recently run some -go and -gxo tests (with large primes) and I have a proposal which may be easy to implement. Both -go and -gxo tests proceed all the way until finishing testing base 11, before reporting, only at the very end (this is an example; not mine):

11*2^2897409+1 is a Factor of GF(2897408,3)!!!! (30574.520000 seconds)
11*2^2897409+1 is a Factor of xGF(2897408,7,4)!!!! (0.010000 seconds)

I am fairly sure that the first of these two statements could have been output already after finishing the test in base 3 (i.e. ~5 hours earlier). I can (and I will) delve into the code to try to propose a patch, but maybe this is entirely trivial (with some rearranged logic)?

Of course, this is going to be of purely hedonic value, and of course the user could wait for another 5 hours, but wouldn't it be nice?

[axn]

Originally Posted by Batalov

I am fairly sure that the first of these two statements could have been output already after finishing the test in base 3 (i.e. ~5 hours earlier).

I would have thought that all the bases have to be powered _simultaneously_ if you want to efficiently test xGFs?

[rogue]

Originally Posted by Batalov

Mark,
I have recently run some -go and -gxo tests (with large primes) and I have a proposal which may be easy to implement. Both -go and -gxo tests proceed all the way until finishing testing base 11, before reporting, only at the very end (this is an example; not mine):

11*2^2897409+1 is a Factor of GF(2897408,3)!!!! (30574.520000 seconds)
11*2^2897409+1 is a Factor of xGF(2897408,7,4)!!!! (0.010000 seconds)

I am fairly sure that the first of these two statements could have been output already after finishing the test in base 3 (i.e. ~5 hours earlier). I can (and I will) delve into the code to try to propose a patch, but maybe this is entirely trivial (with some rearranged logic)?

Of course, this is going to be of purely hedonic value, and of course the user could wait for another 5 hours, but wouldn't it be nice?

I didn't write the code, so I can't state that this is as easy as it appears. Feel free to take a look. The code is in the attached file.

[Batalov]

I've read it for an hour yesterday.
It is obviously non-trivial to rearrange the logic right away, but I'll muck with it.

I was also curious to see what some people change in the code to find consolation prizes like this or this. The "extended extended" searches are possible with "-gx{2,27}{2,28}" for example; it creates more "GF_Subs" etc etc. All of the eligible GF_sprime_%d tests are run first (less for -g, because some bases can never produce -1 (mod p_tested). It all makes sense. Haven't thought yet how to rearrange. Ugh, all those globals!

P.S. For the impatient ones, the FermatFactor-only code could be run as "-gos2"... Interesting.

P.P.S. There are funny side-effects that can be produced in the extended extended part. For example, while GF(8) is explicitly protected from over-reporting when the number is a Fermat factor, GF(32) is not. (E.g. try pfgw -gxo"{1,31}{2,32}" -l -q"189*2^90061+1"; the xGFs of course are found at 1/k rate and casting a wider net, everyone can find an xGF sooner or later.)

[rogue]

Originally Posted by Batalov

Haven't thought yet how to rearrange. Ugh, all those globals!

Preaching to the choir! globals aren't necessarily evil, but there are too many of them.

I would love to rewrite much of pfgw, but that is no small undertaking. Some of the functions are long and the execution path to access any of the pfoo functions is scary. I would also eliminate the use of overloaded operators. That is one of my least favorite things in C++ (with the exception of using '+' on string objects).

[pepi37]

Is it possible to add one "feature" to WinPFGW. If I need to do decimal expansion of some number, I must first create txt file with number, then start WinPFGW and enter pfgw -od -lprime.txt -q"inifile.txt"
It will be great that I can enter only pfgw -od -lprime.txt -q"number" without making inifile.txt and got result in prime.txt
I think is simple change in PFGW code...

[ET_]

A question about pfgw.

If I run (say)

./pfgw 64s 2046*(2^1257787-1)+1

I can test if the factor is composite or PRP.

If the number is PRP, may I then run the script

SCRIPT

DIM expo, 1257787
DIM base, 2
DIM k, 1023
DIM result, 0
DIM mers, (2*k*(2^expo-1))+1
DIMS rstr
OPENFILEAPP r_file,results.txt

POWMOD result,base,2^expo-1,mers
WRITE r_file,result

to check if the result is 1, and therefore the number checked is prime and divides the double Mersenne?

Luigi

[henryzz]

You would have to find the PRP in the script or modify the script to read PRPs from a file. It is quite possible that one of these options is quite viable. It would theoretically be possible to read an ABC file with a pfgw script.

[ET_]

Originally Posted by henryzz

You would have to find the PRP in the script or modify the script to read PRPs from a file. It is quite possible that one of these options is quite viable. It would theoretically be possible to read an ABC file with a pfgw script.

Well, I don't mind checking for compositenes by hand. A double Mersenne PRP factor is still quite rare at the actual level of sieving...

I just want to be ready when I will find one

Luigi

[rogue]

Originally Posted by pepi37

Is it possible to add one "feature" to WinPFGW. If I need to do decimal expansion of some number, I must first create txt file with number, then start WinPFGW and enter pfgw -od -lprime.txt -q"inifile.txt"
It will be great that I can enter only pfgw -od -lprime.txt -q"number" without making inifile.txt and got result in prime.txt
I think is simple change in PFGW code...

I don't understand. -q takes a number or expression as input, not a file. You would have pfgw -od -lprime.txt infile.txt or pfgw -od -lprime.txt -q"number"

Can you give me a specific example of what you are trying to do so that I can understand?