GGNFS problems on Windows

Jeff Gilchrist · 2011-01-11, 16:34

Quote:

Originally Posted by Andi47

I just tested with your windows binary (32 bit) of svn-400:
gnfs-lasieve4I16e -M 1 -a 2_956+.poly -o test11.out -f 43505239 -c 100

still crashes. :(
edit: the 64 bit binary crashes too on this input.

Darn. I guess it must be something else then that wasn't updated with the latest SVN code. That windows build was made with the new VS2010 compiler and a newer MPIR as well.

Jeff.

Brian Gladman · 2011-01-12, 09:48

Quote:

Originally Posted by Andi47

I just tested with your windows binary (32 bit) of svn-400:

gnfs-lasieve4I16e -M 1 -a 2_956+.poly -o test11.out -f 43505239 -c 100

still crashes. :(

edit: the 64 bit binary crashes too on this input.

In order to look at this I need to be able to run gnfs-lasieve4I16e and the command line alone is not sufficient for this. I am not quite sure how much more is needed (the poly file at least) but I will take a look at what is going on if I can obtain all that is needed to set up a context in which the siever will run.

Brian

Andi47 · 2011-01-12, 10:31

Quote:

Originally Posted by Brian Gladman

In order to look at this I need to be able to run gnfs-lasieve4I16e and the command line alone is not sufficient for this. I am not quite sure how much more is needed (the poly file at least) but I will take a look at what is going on if I can obtain all that is needed to set up a context in which the siever will run.

Brian

AFIK the siever and the polyfile should be sufficient to do the testing.

The polyfile is (name it "2_956+.poly" to be consistent with my command line):

Code:

n: 1332203900608590809938480645779084229713406802212647383719972524986911760145731133592887032687895490700549453758482415452796437240091528083264268919532584204888670812466960924609380670193
type: gnfs
# norm 2.589534e-18 alpha -8.321781 e 2.991e-14 rroots 5
skew: 320811150.54
c0:  405063155184101117008321502924944026097244672704
c1:  3339583019941934259279259002396544515176
c2: -47189249406761789351128506210198
c3: -265660627688613695069141
c4:  586429385543294
c5:  182280
Y0: -1488559874631410107110585286834796715
Y1:  149978515624994424157
alim: 150000000
rlim: 150000000
lpba: 32
lpbr: 31
mfba: 94
mfbr: 62
alambda: 3.6
rlambda: 2.6

P.S.: How much RAM does your PC have?

Brian Gladman · 2011-01-12, 11:48

I can run it on either of two machines, one with 4GB and one with 24GB.

I can see the crash so I will see if I can figure out what is going on.

Edit: bug found and corrected in the GGNFS SVN.

The code was picking up and using an outdated local version of gmp.h stored within the ggnfs directory structure but then linking with MPIR in MPIR's own root directory where the correct gmp.h is stored.

I've now changed ggnfs to pick up both the gmp header and the MPIR library from MPIR's root directory.

Brian

Andi47 · 2011-01-12, 12:10

Quote:

Originally Posted by Brian Gladman

I can run it on either of two machines, one with 4GB and one with 24GB.

I can see the crash so I will see if I can figure out what is going on.

Edit: bug found and corrected in the GGNFS SVN.

The code was picking up and using an outdated local version of gmp.h stored within the ggnfs directory structure but then linking with MPIR in MPIR's own root directory where the correct gmp.h is stored.

I've now changed ggnfs to pick up both the gmp header and the MPIR library from MPIR's root directory.

Brian

Does the newest SVN which your changes now run smoothly without crash?

@Jeff: can you please create windows binaries (32 and 64 bit) of SVN401, so that I can run some tests?

Brian Gladman · 2011-01-12, 12:36

The x64 version no longer crashes for your specific example. I think that the bug I found would show up in other sievers and in win32 as well. But binaries would also sometimes work if the two versions of gmp.h involved just happened to be compatible.

Brian

PS I don't like filling up other people's systems with attachments but if the 'powers that be' don't mind this I could upload a specific siever binary here.

Jeff is 'sort of' running a binary repository via his web page but I wonder whether we should actually run such a repository with the programs that we use around here so that I and others can upload corrected Windows binaries as soon as we have them.

Andi47 · 2011-01-12, 12:43

Quote:

Originally Posted by Brian Gladman

The x64 version no longer crashes for your specific example. I think that the bug I found would show up in other sievers and in win32 as well. But binaries would also sometimes work if the two versions of gmp.h involved just happened to be compatible.

Brian

PS I don't like filling up other people's systems with attachments but if the 'powers that be' don't mind this I could upload a specific siever binary here.

Jeff is 'sort of' running a binary repository via his web page but I wonder whether we should actually run such a repository with the programs that we use around here so that I and others can upload corrected Windows binaries as soon as we have them.

Do you also have a x32 binary?

P.S.: If you don't mind, you can also upload the binaries at www.sendspace.com and post the link(s) here.

Brian Gladman · 2011-01-12, 13:14

I am uneasy about doing this with GPL'd code since it is important to maintain a clear link between binaries and the source code from which they have been derived.

I can build win32 and x64 binaries of a specific siever so if you PM me with an email address I can put them in a ZIP file and send them to you privately.

Brian

Andi47 · 2011-01-12, 14:58

Quote:

Originally Posted by Brian Gladman

I am uneasy about doing this with GPL'd code since it is important to maintain a clear link between binaries and the source code from which they have been derived.

I can build win32 and x64 binaries of a specific siever so if you PM me with an email address I can put them in a ZIP file and send them to you privately.

Brian

Thanks, I have sent you a PM with my email address.

Andi47 · 2011-01-15, 07:59

Quote:

Originally Posted by Brian Gladman

I am uneasy about doing this with GPL'd code since it is important to maintain a clear link between binaries and the source code from which they have been derived.

I can build win32 and x64 binaries of a specific siever so if you PM me with an email address I can put them in a ZIP file and send them to you privately.

Brian

I just tested:

gnfs-lasieve4I16e -M 1 -a 2_956+.poly -o test11.out -f 43505239 -c 100

still crashes with both the win32 and x64 binaries. (platform is windows 7 professional)

Brian Gladman · 2011-01-15, 09:17

Hi Andi,

I didn't see this crash earlier for some reason, but I do now. This turns out to be a different bug and, this time, one in the actual lattice sieve code. I hope that I have now corrected this one in the GGNFS SVN. I have sent you updated binaries for test.

Brian

2011-01-12, 11:48	#4
Brian Gladman May 2008 Worcester, United Kingdom 2²·7·19 Posts	I can run it on either of two machines, one with 4GB and one with 24GB. I can see the crash so I will see if I can figure out what is going on. Edit: bug found and corrected in the GGNFS SVN. The code was picking up and using an outdated local version of gmp.h stored within the ggnfs directory structure but then linking with MPIR in MPIR's own root directory where the correct gmp.h is stored. I've now changed ggnfs to pick up both the gmp header and the MPIR library from MPIR's root directory. Brian Last fiddled with by Brian Gladman on 2011-01-12 at 12:05 Reason: update

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Windows 7 Windows Update & Prime95 issue!!!	Unregistered	Information & Answers	14	2010-04-10 21:47
Windows 7 Problems	spaz	Information & Answers	6	2009-12-22 04:58
Problems building ggnfs	xilman	Factoring	7	2009-10-07 18:24
Problems with sr2sieve on x86-64 Windows	geoff	Prime Sierpinski Project	4	2007-12-16 00:25
Distributed GGNFS on Windows	ValerieVonck	Factoring	12	2006-02-02 07:25

2011-01-12, 12:36	#6
Brian Gladman May 2008 Worcester, United Kingdom 2²·7·19 Posts	The x64 version no longer crashes for your specific example. I think that the bug I found would show up in other sievers and in win32 as well. But binaries would also sometimes work if the two versions of gmp.h involved just happened to be compatible. Brian PS I don't like filling up other people's systems with attachments but if the 'powers that be' don't mind this I could upload a specific siever binary here. Jeff is 'sort of' running a binary repository via his web page but I wonder whether we should actually run such a repository with the programs that we use around here so that I and others can upload corrected Windows binaries as soon as we have them.

2011-01-12, 13:14	#8
Brian Gladman May 2008 Worcester, United Kingdom 2²×7×19 Posts	I am uneasy about doing this with GPL'd code since it is important to maintain a clear link between binaries and the source code from which they have been derived. I can build win32 and x64 binaries of a specific siever so if you PM me with an email address I can put them in a ZIP file and send them to you privately. Brian

2011-01-15, 09:17	#11
Brian Gladman May 2008 Worcester, United Kingdom 2²·7·19 Posts	Hi Andi, I didn't see this crash earlier for some reason, but I do now. This turns out to be a different bug and, this time, one in the actual lattice sieve code. I hope that I have now corrected this one in the GGNFS SVN. I have sent you updated binaries for test. Brian