build problem on x86_64

Prime95 · 2007-04-14, 00:44

Debian OS

I downloaded latest GNU GMP and did
./configure
make
make install

Now in the GMP-ECM directory I do a ./configure and get:

checking if gmp.h version and libgmp version are the same... no
configure: error: 'gmp.h' and 'libgmp' have different versions! you have to pro
erly reinstall GMP.

Ideas?

gribozavr · 2007-04-14, 01:00

I'd install GMP in, say, ~/opt/gmp, and not the system-wide directory. Or at least, uninstall system GMP first (use aptitude and '_' key).
./configure --prefix=/home/user/opt/gmp

And then build GMP-ECM with it:
./configure --prefix=/home/user/opt/gmp-ecm --with-gmp=/home/user/opt/gmp

And you'll probably need to add /home/user/opt/gmp/lib to your $LD_LIBRARY_PATH.

Maybe there is a simpler way, but that's the way I compile gmp-ecm, and it works for me (I'm use x86_64 debian sarge).

akruppa · 2007-04-15, 15:32

George,

do you have the Debian GMP package installed, and/or GMP compiled and installed from sources? If both are installed, or were installed and one wasn't cleanly uninstalled, GMP-ECM frequently gets mixed up about where to get the header files and where to get the library files from, resulting in the error you experienced.

Alex

Prime95 · 2007-04-15, 18:19

Alex, I believe Debian installed GMP by default. I downloaded and built the latest version. Gribozavr's trick worked and it is merrily attacking M1061 now.

My "mistake" was assuming "make install" would install the new version over the current version.

Andi47 · 2007-04-15, 18:24

Quote:

Originally Posted by Prime95

Alex, I believe Debian installed GMP by default. I downloaded and built the latest version. Gribozavr's trick worked and it is merrily attacking M1061 now.

What are you doing about M1061? ECM? P-1?

If you want to do P-1, I have P-1'ed M1061 up to B1=150e9 and B2=739,720,038,949,720, so if you want to continue for higher B1, I can send you the savefile of step 1 so that you can -resume.

Prime95 · 2007-04-15, 20:47

Quote:

Originally Posted by Andi47

What are you doing about M1061?

I'm running ECM.

I'm working on a port of prime95 (v25) to linux x86-64. While I'm doing that I want to keep the Opteron busy

akruppa · 2007-04-16, 08:03

Quote:

Originally Posted by Prime95

Alex, I believe Debian installed GMP by default. I downloaded and built the latest version. Gribozavr's trick worked and it is merrily attacking M1061 now.

My "mistake" was assuming "make install" would install the new version over the current version.

Packages installed by the distribution go in the /usr/{bin,lib,include,...} tree, whereas packages installed by the admin from other sources should go in /usr/local/{bin,lib,include,...} tree, so the distribution and other stuff can be kept separate. I guess configure found the header file in /usr/include and the library in /usr/local/lib.

Alex

R.D. Silverman · 2007-04-16, 14:48

Quote:

Originally Posted by Andi47

What are you doing about M1061? ECM? P-1?

If you want to do P-1, I have P-1'ed M1061 up to B1=150e9 and B2=739,720,038,949,720, so if you want to continue for higher B1, I can send you the savefile of step 1 so that you can -resume.

I would recommend against such further computation. It is unlikely to
yield any result.

bdodson · 2007-04-16, 17:01

Quote:

Originally Posted by R.D. Silverman

I would recommend against such further computation. It is unlikely to yield any result.

Count me in total agreement with Bob on this one. Even worse than
"unlikely"; an ecm factorization at this point runs a danger of being
unwelcome and/or unfortunate. One of the other C3xx Mp's would be
much better (or perhaps one of the Pn's). Here are the largest.
-Bruce

307 2,1039- 115942057...
307 2,1091+ 211720879...
309 2,1159- 620852682...
310 2,1051- 674820544...
311 2,1124+ 436637073...
313 2,1139- 277819073...
314 2,1199- 155076532...
317 2,1147+ 635461255...
318 2,1163- 438168120...
318 2,1159+ 165413132...
318 2,1084+ 216089190...
320 2,1061- (removed)
321 2,1123- 146589520...
326 2,1168+ 115035024...
327 2,1129- 258211855...
330 2,1129+ 458892513...
331 2,1153- 539631937...
332 2,1117- 332049358...
334 2,1187+ 111826808...
338 2,1123+ 379807796...
342 2,1171- 208792137...
355 2,1193- 110545391...

(reported acronym for nsa: "never said anything"; as in
"I heard that ..." "ah, we ..." --- reported by one of my former
students, from before nsa's more recent warm/fuzzy image.)

Andi47 · 2007-04-16, 17:19

Quote:

Originally Posted by bdodson

Count me in total agreement with Bob on this one. Even worse than
"unlikely"; an ecm factorization at this point runs a danger of being
unwelcome and/or unfortunate. One of the other C3xx Mp's would be
much better (or perhaps one of the Pn's). Here are the largest.
-Bruce

307 2,1039- 115942057...
307 2,1091+ 211720879...
309 2,1159- 620852682...
310 2,1051- 674820544...
311 2,1124+ 436637073...
313 2,1139- 277819073...
314 2,1199- 155076532...
317 2,1147+ 635461255...
318 2,1163- 438168120...
318 2,1159+ 165413132...
318 2,1084+ 216089190...
320 2,1061- (removed)
321 2,1123- 146589520...
326 2,1168+ 115035024...
327 2,1129- 258211855...
330 2,1129+ 458892513...
331 2,1153- 539631937...
332 2,1117- 332049358...
334 2,1187+ 111826808...
338 2,1123+ 379807796...
342 2,1171- 208792137...
355 2,1193- 110545391...

(reported acronym for nsa: "never said anything"; as in
"I heard that ..." "ah, we ..." --- reported by one of my former
students, from before nsa's more recent warm/fuzzy image.)

Is it known how much ecm and p-1 work has been done on these?

philmoore · 2007-04-16, 19:35

Quote:

Originally Posted by bdodson

Count me in total agreement with Bob on this one. Even worse than "unlikely"; an ecm factorization at this point runs a danger of being unwelcome and/or unfortunate.

Why unwelcome or unfortunate? Has someone already started an NFS factorization on M1061? I interpreted Bob's post as saying that P-1 was unlikely to yield a factor, but he didn't appear to be directly referring to ECM.

On the other hand, if M1061 were to become an NFS target, how much ECM should be done beforehand? I would guess at least to 65 digits, maybe even 70, but this will depend on an estimate of the total effort needed for NFS. I would be interested in a very rough figure from any of you knowledgeable NFS people, realizing, of course, that without the actual software for the task in hand yet, estimates will be pretty tentative.

2007-04-14, 00:44	#1
Prime95 P90 years forever! Aug 2002 Yeehaw, FL 2·5·7·113 Posts	build problem on x86_64 Debian OS I downloaded latest GNU GMP and did ./configure make make install Now in the GMP-ECM directory I do a ./configure and get: checking if gmp.h version and libgmp version are the same... no configure: error: 'gmp.h' and 'libgmp' have different versions! you have to pro erly reinstall GMP. Ideas?

2007-04-14, 01:00	#2
gribozavr Mar 2005 Internet; Ukraine, Kiev 11×37 Posts	I'd install GMP in, say, ~/opt/gmp, and not the system-wide directory. Or at least, uninstall system GMP first (use aptitude and '_' key). ./configure --prefix=/home/user/opt/gmp And then build GMP-ECM with it: ./configure --prefix=/home/user/opt/gmp-ecm --with-gmp=/home/user/opt/gmp And you'll probably need to add /home/user/opt/gmp/lib to your $LD_LIBRARY_PATH. Maybe there is a simpler way, but that's the way I compile gmp-ecm, and it works for me (I'm use x86_64 debian sarge). Last fiddled with by gribozavr on 2007-04-14 at 01:01

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
build with MSVC 2010 problem	Geek2600	Msieve	0	2018-02-12 13:37
[Patch] CPU affinity prompt problem in mprime Linux / OS X build	Explorer09	Software	1	2017-03-01 02:34
Need to squeeze a few more cycles from x86_64 asm	ewmayer	Programming	36	2012-07-28 19:38
mprime on FC4 on x86_64 : "cannot execute binary file"	T.Rex	Software	9	2006-09-01 21:21
namespace problem during build	bad_source_bios	Software	7	2003-12-03 20:38

2007-04-15, 15:32	#3
akruppa "Nancy" Aug 2002 Alexandria 4643₈ Posts	George, do you have the Debian GMP package installed, and/or GMP compiled and installed from sources? If both are installed, or were installed and one wasn't cleanly uninstalled, GMP-ECM frequently gets mixed up about where to get the header files and where to get the library files from, resulting in the error you experienced. Alex

2007-04-15, 18:19	#4
Prime95 P90 years forever! Aug 2002 Yeehaw, FL 2·5·7·113 Posts	Alex, I believe Debian installed GMP by default. I downloaded and built the latest version. Gribozavr's trick worked and it is merrily attacking M1061 now. My "mistake" was assuming "make install" would install the new version over the current version.