PFGW 3.2.0 has been Released

[rogue]

Originally Posted by rogue

I have posted Windows binaries here, http://sourceforge.net/projects/openpfgw/files/. MacIntel and Linux binaries will follow in the next day or so.

The major change in this release is that it has been updated to use gwnum 25.14, which addresses the following issues:

• Gwnum would lose the /d in some (k*b^n+c)/d cases. This happened primarily in cases where the number was between 300 and 350 bits.
• The final conversion from gwnum to binary could fail in cases where generic reduction was used. The larger the number the less likely the bug would occur.
• Zero padded FFTs of length 80 and 112 could fail.
• More conservative in selecting an FFT length for non-base-2 cases.

This release also goes back to using rational FFTs, which are slightly faster than irrational FFTs.

With these changes to gwnum, there will be fewer roundoff errors with PFGW. These changes also address primality tests that silently fail. "Silently fail" refers to tests that do not trigger roundoff errors, yet have an incorrect result. It is highly recommended to update your clients to this release.

MacIntel and Linux binaries are now available.

[rogue]

I have posted all binaries here, http://sourceforge.net/projects/openpfgw/files/.

There is only one change to this release. GFN testing will output a string "GFN testing complete" after it has completed. This message will be used as an affirmation to the PRPNet client (3.2.4, when that is released) that GFN testing completed and was not cancelled.

[ET_]

Originally Posted by rogue

I have posted all binaries here, http://sourceforge.net/projects/openpfgw/files/.

There is only one change to this release. GFN testing will output a string "GFN testing complete" after it has completed. This message will be used as an affirmation to the PRPNet client (3.2.4, when that is released) that GFN testing completed and was not cancelled.

Is there another distributed project working with PRPnet on GFN?

Luigi

[rogue]

Originally Posted by ET_

Is there another distributed project working with PRPnet on GFN?

Luigi

Let me rephrase my statement as there appears to be some confusion. This change refers to GFN divisibility testing of Proth primes, i.e. using the -gx switch of PFGW.

A search for GFN primes is underway at PrimeGrid. Right now they have servers for n=2^15 an n=2^16.

[kar_bon]

Found this while testing some small cases:

pfgw -q"9238*619^619+1"
PFGW Version 3.3.4.20100405.Win_Stable [GWNUM 25.14]
9238*619^619+1 is composite: RES64: [AF5E2DAE777932BE] (0.1337s+0.0003s)


pfgw -a1 -q"9238*619^619+1"
PFGW Version 3.3.4.20100405.Win_Stable [GWNUM 25.14]
9238*619^619+1 is 3-PRP! (0.1342s+0.0002s)

Same results on WinPFGW.

LLRV3.8.1 says:
9238*619^619+1 is not prime. RES64: 724FCB9240CF6A8B. OLD64: 56EF62B6C26E3F9E Time : 113.720 ms.

Note different RES, too.

I got also many errors like

-----------------------------------------------------------------------
Error occuring in PFGW at Wed Sep 01 14:46:26 2010
Expr = 3366*308^308+1
Detected in MAXERR>0.45 (round off check) in prp_using_gwnum
Iteration: 199/2557 ERROR: ROUND OFF 0.5>0.45
PFGW will automatically rerun the test with -a1
-----------------------------------------------------------------------
Error occuring in PFGW at Wed Sep 01 14:46:26 2010
Expr = 3370*308^308+1
Detected in MAXERR>0.45 (round off check) in prp_using_gwnum
Iteration: 164/2557 ERROR: ROUND OFF 0.5>0.45
PFGW will automatically rerun the test with -a1
-----------------------------------------------------------------------
Error occuring in PFGW at Wed Sep 01 14:46:26 2010
Expr = 3372*308^308+1
Detected in MAXERR>0.45 (round off check) in prp_using_gwnum
Iteration: 225/2557 ERROR: ROUND OFF 1>0.45
PFGW will automatically rerun the test with -a1

but running again, lonely, they're not reproduceable!

Why there're these differences?

[rogue]

Originally Posted by kar_bon

Found this while testing some small cases:

pfgw -q"9238*619^619+1"
PFGW Version 3.3.4.20100405.Win_Stable [GWNUM 25.14]
9238*619^619+1 is composite: RES64: [AF5E2DAE777932BE] (0.1337s+0.0003s)


pfgw -a1 -q"9238*619^619+1"
PFGW Version 3.3.4.20100405.Win_Stable [GWNUM 25.14]
9238*619^619+1 is 3-PRP! (0.1342s+0.0002s)

Same results on WinPFGW.

LLRV3.8.1 says:
9238*619^619+1 is not prime. RES64: 724FCB9240CF6A8B. OLD64: 56EF62B6C26E3F9E Time : 113.720 ms.

Note different RES, too.

I got also many errors like

-----------------------------------------------------------------------
Error occuring in PFGW at Wed Sep 01 14:46:26 2010
Expr = 3366*308^308+1
Detected in MAXERR>0.45 (round off check) in prp_using_gwnum
Iteration: 199/2557 ERROR: ROUND OFF 0.5>0.45
PFGW will automatically rerun the test with -a1
-----------------------------------------------------------------------
Error occuring in PFGW at Wed Sep 01 14:46:26 2010
Expr = 3370*308^308+1
Detected in MAXERR>0.45 (round off check) in prp_using_gwnum
Iteration: 164/2557 ERROR: ROUND OFF 0.5>0.45
PFGW will automatically rerun the test with -a1
-----------------------------------------------------------------------
Error occuring in PFGW at Wed Sep 01 14:46:26 2010
Expr = 3372*308^308+1
Detected in MAXERR>0.45 (round off check) in prp_using_gwnum
Iteration: 225/2557 ERROR: ROUND OFF 1>0.45
PFGW will automatically rerun the test with -a1

but running again, lonely, they're not reproduceable!

Why there're these differences?

I'll investigate and forward on to George if necessary.

[kar_bon]

Alpertron ECM-applet gives 9238*619^619+1 prime, LLR not!

[3.14159]

Alperton + PARI + N-1 + Proth.exe says prime! This is a program bug! For the tests: See here.

Also: Karsten, you have the same problem as I did? I knew it was not just me!

Hey: I think we have found a real example of the elusive pseudocomposite! (It has passed Miller-Rabin for about 30-40 bases, for bases of 8 digits, 11 digits, and 12 digits)

LLR and PFGW's normal mode say composite, apparently.

[rogue]

Originally Posted by 3.14159

Alperton + PARI + N-1 + Proth.exe says prime! This is a program bug! For the tests: See here.

Also: Karsten, you have the same problem as I did? I knew it was not just me!

Hey: I think we have found a real example of the elusive pseudocomposite! (It has passed Miller-Rabin for about 30-40 bases, for bases of 8 digits, 11 digits, and 12 digits)

LLR and PFGW's normal mode say composite, apparently.

This is a bug in gwnum. I've passed the information on to George.

[3.14159]

Originally Posted by Rogue

This is a bug in gwnum. I've passed the information on to George.

Is he the program developer?

Also: Notify me whenever bugs are fixed.

[kar_bon]

Originally Posted by 3.14159

Is he the program developer?

Also: Notify me whenever bugs are fixed.

Keep cool!

George Woltman is the developer of the gwnum-library which is used by Prime95, PFGW and LLR.

He (or someone else) will notify here in the forum, so everyone will be aware of a new version!