When will it stop? - Page 2

EdH · 2010-05-18, 01:27

Quote:

Originally Posted by ThomRuley

Can't seem to find that kind of file. Is that going to be a problem?

I would not consider you to have a problem at this point. It is possible the log file is not written until the end of the filtering operation. Unfortunately, the filtering (-nc1) stage does not have a countdown timer like the linear algebra (-nc2) stage.

That many relations will also take a lot of memory. I wonder if that might be slowing things down. Have you noticed any memory messages, such as available and used amounts?

Since it isn't giving you error messages, I suppose you'll just need patience at this point. . .

ThomRuley · 2010-05-18, 03:07

Just to make sure I understand the process this program uses:

Step 1 is lattice sieving to find a large number of relations for q, whatever that is.

Step 2 is some sort of filtering. What is it filtering?

Step 3 is a linear algebra step. What sort of linear algebra does it do?

I've only taken math up through basic calculus so far, so feel free to refer me to any sources to help bring me up to speed.

Thom

jasonp · 2010-05-18, 03:33

Quote:

Originally Posted by ThomRuley

Step 2 is some sort of filtering. What is it filtering?

More detail on NFS filtering than you probably wanted
More detail on the linear algebra than anyone needs
(The CWI ftp site has a more legible copy of Peter's paper)

Andi47 · 2010-05-18, 05:04

Quote:

Originally Posted by ThomRuley

Can't seem to find that kind of file. Is that going to be a problem?

Just a thought: Have you correctly installed msieve.exe? *) Open your FactMsieve.pl file (with a text editor) and look for a line that starts with "$MSIEVE=". The default line reads:

$MSIEVE=$GGNFS_BIN_PATH."/msieve".$EXEC_SUFFIX;

If the line in your file is the same, the msieve binary should be in the same directory as your gnfs-lasieve4I1?e files.

*) If yor system is windows, the binary is named "msieve.exe". If it is linux, the binary is named "msieve" without .exe suffix.

Batalov · 2010-05-18, 05:27

Andi may be just right! Good thinking. If there's no log, then you may have no working msieve.

The old perl script always checked for all working binaries even before starting; I thought python script does, too. Apparently this can be disabled as far as I can see from the python-script thread, which is not a good thing by default.

Also, a minor issue: that infamous minimum estimate is a bit funky in the python script but this is not easy to fix; all estimates are funky... The commented out "pi(...)+pi(...)" estimate* had theoretical grounds; if I had my druthers, I'd rather spline/change only the 0.2 parameter in it and not touch the meat of the formula (think about it - it takes care of uneven lpba/r limits nicely), but that's just an idea - someone has to do the tests. That's not going to be me - I don't use this or the old script...
_____

*) well, ok, pi(...)+pi(...) looks like N/log(N) there, but you will recognize it with minimal effort.

Brian Gladman · 2010-05-18, 07:02

Quote:

Originally Posted by Batalov

Andi may be just right! Good thinking. If there's no log, then you may have no working msieve.

The old perl script always checked for all working binaries even before starting; I thought python script does, too. Apparently this can be disabled as far as I can see from the python-script thread, which is not a good thing by default.

The default in the Python script is that binary checks are turned on.

A bug in the original Perl script meant that msieve was sometimes called before its existence was checked. This was carried over into the Python script but I revamped the checking in the latest version so this bug has now been removed.

Brian

Andi47 · 2010-05-18, 07:37

Quote:

Originally Posted by Batalov

Andi may be just right! Good thinking. If there's no log, then you may have no working msieve.

The old perl script always checked for all working binaries even before starting; I thought python script does, too. Apparently this can be disabled as far as I can see from the python-script thread, which is not a good thing by default.
[/SIZE]

Oops, I did not recognize that he is using the python script.

I don't know what the corresponding line in FactMsieve.py is.

Does the python script by default expect the msieve binary in the same directory as the gnfs-lasieve* binaries too?

(edit: The essence of my posting stays the same: Make sure that the msieve binary is installed in the correct directory.)

Brian Gladman · 2010-05-18, 08:49

Quote:

Originally Posted by Andi47

Oops, I did not recognize that he is using the python script.

I don't know what the corresponding line in FactMsieve.py is.

Does the python script by default expect the msieve binary in the same directory as the gnfs-lasieve* binaries too?

The Python script has to be edited before it is used to set the ggnfs and msieve paths. The script sets these two paths individually so they don't have to be the same. The default is:

GGNFS_PATH = '../../bin/x64/Release/'
MSIEVE_PATH = '../../../msieve/build.vc9/x64/Release/'

which is where I put them

Brian

ThomRuley · 2010-05-19, 01:56

I'm afraid this whole conversation has confused me more than anything. Does the activity on my screen mean I need to do something different? The previous numbers ran without a hitch.

Batalov · 2010-05-19, 02:12

Well, it was you who said that there's no log anywhere in sight. And sent us all on a witch hunt.
If this is not the first number, then I revert to my very first message -- all is going fine!
...achem (here goes)... just step back and enjoy the blinkenlights.

/thassa joke, son, as jasonp says. Or Foghorn Leghorn, to be precise/

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Sieving: When to stop?	SaneMur	Riesel Prime Search	4	2011-07-18 17:04
Stop p95 or llr before restarting?	Joshua2	Software	6	2005-05-16 16:36
Should we stop at k=249?	jocelynl	15k Search	2	2004-07-10 13:31
When to stop sieving?	ebx	3*2^n-1 Search	9	2004-07-06 15:30
RMA - stop LLR bug	paulunderwood	3*2^n-1 Search	8	2004-05-19 07:10

2010-05-18, 03:07	#13
ThomRuley May 2003 2³·31 Posts	Just to make sure I understand the process this program uses: Step 1 is lattice sieving to find a large number of relations for q, whatever that is. Step 2 is some sort of filtering. What is it filtering? Step 3 is a linear algebra step. What sort of linear algebra does it do? I've only taken math up through basic calculus so far, so feel free to refer me to any sources to help bring me up to speed. Thom

2010-05-18, 05:27	#16
Batalov "Serge" Mar 2008 Phi(4,2^7658614+1)/2 3⁶·13 Posts	Andi may be just right! Good thinking. If there's no log, then you may have no working msieve. The old perl script always checked for all working binaries even before starting; I thought python script does, too. Apparently this can be disabled as far as I can see from the python-script thread, which is not a good thing by default. Also, a minor issue: that infamous minimum estimate is a bit funky in the python script but this is not easy to fix; all estimates are funky... The commented out "pi(...)+pi(...)" estimate* had theoretical grounds; if I had my druthers, I'd rather spline/change only the 0.2 parameter in it and not touch the meat of the formula (think about it - it takes care of uneven lpba/r limits nicely), but that's just an idea - someone has to do the tests. That's not going to be me - I don't use this or the old script... _____ *) well, ok, pi(...)+pi(...) looks like N/log(N) there, but you will recognize it with minimal effort.

2010-05-19, 01:56	#20
ThomRuley May 2003 2³×31 Posts	I'm afraid this whole conversation has confused me more than anything. Does the activity on my screen mean I need to do something different? The previous numbers ran without a hitch.

2010-05-19, 02:12	#21
Batalov "Serge" Mar 2008 Phi(4,2^7658614+1)/2 3⁶×13 Posts	Well, it was you who said that there's no log anywhere in sight. And sent us all on a witch hunt. If this is not the first number, then I revert to my very first message -- all is going fine! ...achem (here goes)... just step back and enjoy the blinkenlights. /thassa joke, son, as jasonp says. Or Foghorn Leghorn, to be precise/