Phi.exe
 

When using phi.exe i get the following for 5,447-
[code]633431414877126006851067349030487179738737485483799943885361176411239387706027852446464243101591219648673368261514838572895614432258963458850260822278177593102842860846078284594333977551169453128691642515121
# 5^447-1, difficulty: 206.90, skewness: 0.45, alpha: 0.00
# Cost: 2.56025e+017, est. time: 121.92 GHz days (not accurate yet!)
X4 25
X3 0
X2 5
X1 0
X0 1
Y1 -1
Y0 5293955920339377119177015629247762262821197509765625
M 5293955920339377119177015629247762262821197509765625[/code] or [code]633431414877126006851067349030487179738737485483799943885361176411239387706027852446464243101591219648673368261514838572895614432258963458850260822278177593102842860846078284594333977551169453128691642515121
# 5^447-1, difficulty: 312.44, skewness: 0.45, alpha: 0.00
# Cost: 7.12402e+020, est. time: 339239.14 GHz days (not accurate yet!)
X6 125
X5 0
X4 0
X3 0
X2 0
X1 0
X0 -1
Y1 -1
Y0 5293955920339377119177015629247762262821197509765625
M 5293955920339377119177015629247762262821197509765625[/code]

Why doesn't phi.exe find:[code]633431414877126006851067349030487179738737485483799943885361176411239387706027852446464243101591219648673368261514838572895614432258963458850260822278177593102842860846078284594333977551169453128691642515121
Y0 -88817841970012523233890533447265625
Y1 1
X0 25
X1 0
X2 0
X3 5
X4 0
X5 0
X6 1[/code]
Which also has difficulty 206.90?

Also one request. Can the program be changed to also output msieve polynomials?

Which version is that? I get

[CODE]
kruppaal@cookie.loria.fr:~/ecm/ecm/trunk > phi -deg6 633431414877126006851067349030487179738737485483799943885361176411239387706027852446464243101591219648673368261514838572895614432258963458850260822278177593102842860846078284594333977551169453128691642515121
633431414877126006851067349030487179738737485483799943885361176411239387706027852446464243101591219648673368261514838572895614432258963458850260822278177593102842860846078284594333977551169453128691642515121
# 5^447-1, difficulty: 209.69 (*2/9=46.60), skewness: 1.71, alpha: 0.00
# Cost: 3.24875e+17, est. time: 154.702 GHz days (not accurate yet!)
# sq side = -r
X6 1
X5 0
X4 0
X3 5
X2 0
X1 0
X0 25
Y1 -1
Y0 88817841970012523233890533447265625
M 88817841970012523233890533447265625
[/CODE]

Alex

[QUOTE=akruppa;116686]Which version is that?[/QUOTE]I 'think' 0.1.2

Is the latest version available as a windows executable?

Hmm, version 0.1.2.1 does not find the good sextic. My binary might have been made from a modified source I have at home... maybe I added the n==3%6, degree==6 case there but did not release it yet. I'll see when I get home.

Alex

My code at home added cases for degree 6 with exponent divisible by 3. I copied that over into phi 0.1.2.1. (The other code contains too much unfinished stuff to be released.) I also added some output for msieve, but I'm not sure if it's correct this way... can you please test? If it works I'll make this version 0.1.2.2.

Alex

[QUOTE=akruppa;116803]My code at home added cases for degree 6 with exponent divisible by 3. I copied that over into phi 0.1.2.1. (The other code contains too much unfinished stuff to be released.) I also added some output for msieve, but I'm not sure if it's correct this way... can you please test? If it works I'll make this version 0.1.2.2.

Alex[/QUOTE]Can you provide an executable?

Gah, I forgot last evening when I got home. Sorry. I'll try to make one tonight.

Alex

Here it is.

Alex

i get an error saying it is not a valid win32 program

Hm, it worked in Wine... anyone else getting this error? Henry, what Windows version are you using?

Alex

I am getting this error too - I am using Windows XP professional.

windows vista ultimate

[QUOTE=Andi47;117034]I am getting this error too - I am using Windows XP professional.[/QUOTE]

When I try to run it in msys / mingw, I get "./phi: Permission denied."

[QUOTE=Andi47;117034]I am getting this error too - I am using Windows XP professional.[/QUOTE]I'm on Vista and get a popup "phi.exe is not a valid Win32 application".After that, i get a "Access is denied."

I have two cross-compilers installed, not sure what the difference is. There's a binary from the other one, and not stripped of symbols this time. Maybe this works.

Gah, upload limit is ~250k for .zip files. For .pdf it's larger, so I'm renaming the file... simply rename back to .zip.

Alex

Now it works under windows, I have not (yet?) tested if the poly works with msieve.

Just one question: phi.exe outputs SNFS polynomials, does it make sense to enter this into msieve, which does GNFS?

SNFS and GNFS differ only in how the polynomial is selected. The fact that msieve does GNFS means that it has code that can find reasonably good polynomials for numbers not of a simple algebraic form. However, if you supply a polynomial made by other means, i.e. with pencil and paper or with phi, msieve can skip the polynomial selection step, will sieve away just as happily and finish the factorization.

Alex

what limits are there on what n and x i can input it seems to always crash for me

[QUOTE=akruppa;117114]SNFS and GNFS differ only in how the polynomial is selected. The fact that msieve does GNFS means that it has code that can find reasonably good polynomials for numbers not of a simple algebraic form. However, if you supply a polynomial made by other means, i.e. with pencil and paper or with phi, msieve can skip the polynomial selection step, will sieve away just as happily and finish the factorization.

Alex[/QUOTE]

Thanks.

How can I feed the polynomial into msieve? (I must do something wrong or miss something, as it always says that it failed to read the poly)

[QUOTE=Andi47;117178]
How can I feed the polynomial into msieve? (I must do something wrong or miss something, as it always says that it failed to read the poly)[/QUOTE]
Can you give an example? The factor base file contains the number to be factored, and this has to match the number given on the command line (or from a file like worktodo.ini). This number should not include any small factors that are known, otherwise the library will find the small factors and give up.

Also, you should be using GGNFS for the sieving, it would finish about 5x faster.

[QUOTE=jasonp;117217]Can you give an example? The factor base file contains the number to be factored, and this has to match the number given on the command line (or from a file like worktodo.ini). This number should not include any small factors that are known, otherwise the library will find the small factors and give up.

Also, you should be using GGNFS for the sieving, it would finish about 5x faster.[/QUOTE]

An example, just for testing if it works, I will definitely NOT finish this factorization:

phi.exe output for 3,527-:
[code]N 5466315818570157744600674404470135554420684751405676880272078210385283404043543171940308633417556688973629087522011173645136493977169190093057301775731456787243
R0 125236737537878753441860054533045969266612127846243
R1 -1
A0 -1
A1 0
A2 0
A3 0
A4 0
A5 9
[/code]

Repeat, I will definitely NOT finish this or even run it any longer than a few minutes, I just wanted to test if it works.

GGNFS: I tried to install GGNFS on my computer last winter, but due to lack of Linux knowledge I did not succeed to get it to run in a windows / mingw (msys) environment.

[QUOTE=Andi47;117221]An example, just for testing if it works[/QUOTE]
Paste this output into msieve.fb, copy the first line to worktodo.ini (without the 'N ') and run 'msieve -v -ns'. It starts sieving for me, though the automatically chosen parameters are very suboptimal for such a larger SNFS job