Installation of GGNFS

LegionMammal978 · 2017-01-18, 01:24

Currently trying to set up GGNFS/Msieve on Ubuntu as according to this guide. I downloaded and unzipped the GGNFS repo here and tried to compile it as described by the INSTALL file. However, running make in the top-level directory didn't work:

Code:

$ make
Possible targets are:
    pentium2                   Intel Pentium 2
    pentium3                   Intel Pentium 3
    pentium4                   Intel Pentium 4
    prescott                 Intel Pentium 4 with SSE3
    pentium-m                   Intel Pentium M
    athlon             AMD Athlon (k7)
    athlon-xp         AMD Athlon XP (k7) with SSE
    x86_64                   AMD Opteron/Athlon64 (k8)
    nocona                   Intel 64-bit-capable Xeon/Pentium
    ppc_970                  PowerPC 970
    ppc_7450                  PowerPC 7450
    doc             Documentation
    snapshot                Sources snapshot
    install             Installation
    clean                  Clean up

As suggested, I moved into the src folder and tried again, but it still didn't compile:

Code:

$ make
cc -I. -I.. -I../include  -DNDEBUG -O3 -funroll-loops -finline-functions -ftracer -fomit-frame-pointer -W -Wall -march= -pipe -DMALLOC_REPORTING -DGMP_BUG -o getprimes.o -c getprimes.c
cc: error: missing argument to ‘-march=’
Makefile:91: recipe for target 'getprimes.o' failed
make: *** [getprimes.o] Error 1

What am I doing wrong?

jasonp · 2017-01-18, 02:48

You need to use 'make <target_from_that_list>' in the top level directory. But practically speaking nobody uses the GGNFS source distribution anymore, the entire codebase is there to host the Franke/Kleinjung sieving tool and a few utilities. You can find precompiled binaries for the siever plus instructions for use here.

LegionMammal978 · 2017-01-18, 22:00

Quote:

Originally Posted by jasonp

You can find precompiled binaries for the siever plus instructions for use here.

I've looked at that, and it seems to contain the gnfs-lasieve files. I've been able to compile the Msieve binary from source, but what about all the other files in the sample directory there? Do I just have to put factmsieve.py in the same folder as the msieve executable and the gnfs-lasieve files?

Also, unrelated question: I have GMP-ECM installed, but Msieve refuses to compile with the ECM=1 flag:

Code:

$ make all ECM=1
[...]
common/smallfact/gmp_ecm.o common/smallfact/gmp_ecm.c
common/smallfact/gmp_ecm.c:35:17: fatal error: ecm.h: No such file or directory
compilation terminated.
Makefile:290: recipe for target 'common/smallfact/gmp_ecm.o' failed
make: *** [common/smallfact/gmp_ecm.o] Error 1

Is this also obsolete?

jasonp · 2017-01-18, 22:19

ecm.h is the header that GMP-ECM provides for a basic application-level interface to the core of their library. It's possible that ecm.h is not considered suitable to be an installed output of the library, or the ECM package on your machine doesn't include it. For RedHat systems you probably would need an ecm-devel package or some such, if you didn't build it yourself.

Personally, I use msys2 on windows and build GMP-ECM, then manually move the compiled library and ecm.h to a directory that I know is in gcc's include path.

My understanding with GGNFS is that you don't need any of the binaries that are built except the siever. It has no dependencies on the other parts of the distribution.

LegionMammal978 · 2017-01-18, 23:40

Quote:

Originally Posted by jasonp

ecm.h is the header that GMP-ECM provides for a basic application-level interface to the core of their library. It's possible that ecm.h is not considered suitable to be an installed output of the library, or the ECM package on your machine doesn't include it. For RedHat systems you probably would need an ecm-devel package or some such, if you didn't build it yourself.

This has gotten crazy. I compiled GMP-ECM from source and put ecm.h in the include folder. However, now it's saying:

Code:

/usr/bin/ld: cannot find -lecm

Not sure what its problem is...

wombatman · 2017-01-19, 00:09

Did you do "make install"?

LegionMammal978 · 2017-01-19, 00:54

Quote:

Originally Posted by wombatman

Did you do "make install"?

Code:

$ make install ECM=1
make: *** No rule to make target 'install'.  Stop.

I used 'make all ECM=1'.

wombatman · 2017-01-19, 01:04

Sorry, I meant when you compiled GMP-ECM. Usually goes something like "make; make check; make install".

Edit: The other thing I would add is that there's not really much point to compiling MSieve with ECM=1. It only checks for very small factors (15 digits, maybe?) that you should be finding with GMP-ECM or YAFU before Msieve comes into the picture.

LegionMammal978 · 2017-01-19, 01:54

I finally got factmsieve.py working. Thanks!

wombatman · 2017-01-19, 02:19

Quote:

Originally Posted by LegionMammal978

I finally got factmsieve.py working. Thanks!

Excellent!

LegionMammal978 · 2017-01-19, 02:39

Also, just curious, the Python script describes Msieve GPU support, is that anything I should care about?

2017-01-18, 22:19	#4
jasonp Tribal Bullet Oct 2004 3579₁₀ Posts	ecm.h is the header that GMP-ECM provides for a basic application-level interface to the core of their library. It's possible that ecm.h is not considered suitable to be an installed output of the library, or the ECM package on your machine doesn't include it. For RedHat systems you probably would need an ecm-devel package or some such, if you didn't build it yourself. Personally, I use msys2 on windows and build GMP-ECM, then manually move the compiled library and ecm.h to a directory that I know is in gcc's include path. My understanding with GGNFS is that you don't need any of the binaries that are built except the siever. It has no dependencies on the other parts of the distribution. Last fiddled with by jasonp on 2017-01-18 at 22:24 Reason: remove wrong advice

2017-01-19, 01:04	#8
wombatman I moo ablest echo power! May 2013 3504₈ Posts	Sorry, I meant when you compiled GMP-ECM. Usually goes something like "make; make check; make install". Edit: The other thing I would add is that there's not really much point to compiling MSieve with ECM=1. It only checks for very small factors (15 digits, maybe?) that you should be finding with GMP-ECM or YAFU before Msieve comes into the picture. Last fiddled with by wombatman on 2017-01-19 at 01:06

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Installation and running of mkfactc	daxmick	GPU Computing	44	2018-03-28 02:04
GMP-ECM - installation - how to compile - Ubuntu 15	ThomRuley	GMP-ECM	2	2017-01-06 16:54
installation of OS trouble	wildrabbitt	Linux	5	2015-12-22 16:51
Don't know where installation folder went	jinydu	Information & Answers	10	2009-10-14 01:29
GMP-ECM 6.2-rc1 - installation under MinGW failed	Andi47	GMP-ECM	5	2008-04-30 17:46

2017-01-18, 02:48	#2
jasonp Tribal Bullet Oct 2004 3·1,193 Posts	You need to use 'make <target_from_that_list>' in the top level directory. But practically speaking nobody uses the GGNFS source distribution anymore, the entire codebase is there to host the Franke/Kleinjung sieving tool and a few utilities. You can find precompiled binaries for the siever plus instructions for use here.

2017-01-19, 00:09	#6
wombatman I moo ablest echo power! May 2013 2²·3·5·31 Posts	Did you do "make install"?

2017-01-19, 01:54	#9
LegionMammal978 Jan 2017 Glory to Arstotzka! 14₁₀ Posts	I finally got factmsieve.py working. Thanks!

2017-01-19, 02:39	#11
LegionMammal978 Jan 2017 Glory to Arstotzka! 2·7 Posts	Also, just curious, the Python script describes Msieve GPU support, is that anything I should care about?