Adventures in building and using yafu - Page 2

James Heinrich · 2020-09-04, 21:52

Quote:

Originally Posted by paulunderwood

It looks like make is not finding gmp. Maybe you should set the paths correctly. Is gmp installed locally? You might need gmp-dev.

GMP was installed per Ed's guide, the GMP section. The source is in ../gmp/ relative to the yafu directory source directory I'm trying to compile. I redid the gmp section just in case I missed something the first time (but I don't think I did). Same result as above with makeing yafu.

You mention "set the paths correctly" -- can you be more specific as to where I should be setting them?

ls -l /usr/local/lib

Code:

total 59600
 3618442 Sep  3 15:45 libecm.a
     874 Sep  3 15:45 libecm.la
 1318856 Sep  4 17:47 libgmp.a
     913 Sep  4 17:47 libgmp.la
      16 Sep  4 17:47 libgmp.so -> libgmp.so.10.4.0
      16 Sep  4 17:47 libgmp.so.10 -> libgmp.so.10.4.0
  535137 Sep  4 17:47 libgmp.so.10.4.0
 1333084 Nov 23  2018 liblzma.a
     944 Nov 23  2018 liblzma.la
      16 Nov 23  2018 liblzma.so -> liblzma.so.5.2.4
      16 Nov 23  2018 liblzma.so.5 -> liblzma.so.5.2.4
  708676 Nov 23  2018 liblzma.so.5.2.4
      16 Aug 28  2014 libpari.so -> libpari.so.2.7.1
 5694955 Aug 28  2014 libpari.so.2.7.1
      16 Aug 28  2014 libpari.so.4 -> libpari.so.2.7.1
      22 Mar  1  2016 libwkhtmltox.so -> libwkhtmltox.so.0.13.0
      22 Mar  1  2016 libwkhtmltox.so.0 -> libwkhtmltox.so.0.13.0
      22 Mar  1  2016 libwkhtmltox.so.0.13 -> libwkhtmltox.so.0.13.0
47782840 Apr 12  2015 libwkhtmltox.so.0.13.0
    4096 May 13  2019 p7zip
    4096 Aug 28  2014 pari
    4096 Sep  4 17:47 pkgconfig

paulunderwood · 2020-09-04, 22:06

In Makefile, try commenting out:

Code:

 
       INC += -I../gmp/include
        LIBS += -L../gmp/lib/linux/x86_64

and add in its place:

Code:

       INC += -I/usr/local/include
       LIBS += -L/usr/local/lib

Is gmp.h in /usr/local/include/ ?

James Heinrich · 2020-09-04, 22:15

Quote:

Originally Posted by paulunderwood

Is gmp.h in /usr/local/include/ ?

Yes.
ls -l /usr/local/include

Code:

-rw-r--r--  8159 Sep  3 15:45 ecm.h
-rw-r--r-- 83978 Sep  4 17:47 gmp.h
drwxr-xr-x  4096 Nov 23  2018 lzma
-rw-r--r--  9817 Nov 23  2018 lzma.h
drwxr-xr-x  4096 Aug 28  2014 pari
drwxr-xr-x  4096 Mar  1  2016 wkhtmltox

Quote:

Originally Posted by paulunderwood

In Makefile, try commenting out ... and add in its place

Oh my, I think that's done it!

At least, yafu "tune" is running right now! Thank you Paul!

paulunderwood · 2020-09-04, 22:22

My pleasure. I think the local include/link caught ecm too

EdH · 2020-09-05, 00:00

I do need to change something on my YAFU page. Thanks for pointing it out.

Only the "wip" Makefile has the 7.0.3 reference:

Code:

 CC = gcc-7.3.0

The "trunk" version does not.

Glad to read that you are up and running.

James Heinrich · 2020-09-05, 00:36

Quote:

Originally Posted by EdH

Glad to read that you are up and running.

Indeed, it appears that I am, thanks to your comprehensive guide, and some pointers from Paul & Chris.

I threw my pet number (I came across it last week in an automated too-big-to-factor-quickly report from mersenne.ca, I think it's a k from a composite factor, ~~don't remember which exponent~~ from one 350-bit 8of9 composite factor of M9532331) at my new yafu and it did its thing with ecm and nfs and came up with the answer for me:

Code:

>> factor(122516985909244517881828509074071415559791900903255048509694468186063012576476804164857444126360193)
fac: factoring 122516985909244517881828509074071415559791900903255048509694468186063012576476804164857444126360193
fac: using pretesting plan: normal
fac: no tune info: using qs/gnfs crossover of 95 digits
div: primes less than 10000
fmt: 1000000 iterations
rho: x^2 + 3, starting 1000 iterations on C98
rho: x^2 + 2, starting 1000 iterations on C98
rho: x^2 + 1, starting 1000 iterations on C98
nfs: searching for brent special forms...
nfs: searching for homogeneous cunningham special forms...
nfs: searching for XYYXF special forms...
nfs: couldn't find special form
pm1: starting B1 = 150K, B2 = gmp-ecm default on C98
ecm: 32/32 curves on C98, B1=2K, B2=gmp-ecm default
ecm: 80/80 curves on C98, B1=11K, B2=gmp-ecm default
ecm: 216/216 curves on C98, B1=50K, B2=gmp-ecm default, ETA: 1 sec
pm1: starting B1 = 3750K, B2 = gmp-ecm default on C98
ecm: 408/408 curves on C98, B1=250K, B2=gmp-ecm default, ETA: 3 sec
nfs: commencing nfs on c98: 40838995303081505960609503024690471853263966967751682836564822728687670858825601388285814708786731
...
linear algebra completed 190209 of 190470 dimensions (99.9%, ETA 0h 0m)
nfs: commencing msieve sqrt
NFS elapsed time = 2501.1627 seconds.
Total factoring time = 2658.3809 seconds
***factors found***
P1 = 3
P59 = 56778227132612889065677008632219351134855595667964207526657
P39 = 719272111256604635831656054993527440683

So I'm happy!

EdH · 2020-09-05, 02:05

Quote:

Originally Posted by James Heinrich

. . .
So I'm happy!

Excellent! Glad I could be of help. (And, thanks for letting me know I needed to make an edit.)

Actually, I needed to make several edits today. The msieve source code has been updated to incorporate the GMP-6.2.0 change also, so I had to edit that note out of my instructions.

LaurV · 2020-09-09, 08:21

Quote:

Originally Posted by James Heinrich

***factors found***
P1 = 3
So I'm happy!

Haha!

James Heinrich · 2020-11-12, 22:00

Revisiting this again, this time on my new CentOS-8 server, I thought I'd pass on a few notes and observations in case someone else as inexperienced as I am tries this:

Package names:
Before anything else I needed to run yum --enablerepo=extras install epel-release or I wouldn't be able to find any of these packages.
apt-get replaced with yum mostly works fine, but a few package names need translating:
g++ -> gcc-c++
zlib1g-dev -> zlib-devel
I also had to install xz since it wasn't installed by default.

With those minor changes everything went fine until I got to the ggnfs section, and found that Ben's archive is no longer downloadable from https://sites.google.com/site/bbuhrow/lasieve4_64.7z

I had previously downloaded these sievers on my old server so I copied them over.
The instructions also talk about how to build these from source, but the "apply the following patches" section was a bit intimidating (at least using vi, perhaps I should be using some tool to apply those patches?) so I skipped that.

I finally built yafu with make x86_64 NFS=1 USE_SSE41=1 USE_AVX2=1 since my CPU (Xeon(R) Gold 5120) apparently supports both SSE4.1 and AVX2 but about 5 seconds into yafu "tune" I get a segfault. I rebuilt it without AVX2 and it completed tune without error.

I threw my pet number at it and it found the factors I expected. On my previous 8-core Atom C2750 it took 2658 seconds, on my new 6-core (VM) Xeon Gold 5120 it took 1368 seconds.

bsquared · 2020-11-12, 22:08

Quote:

Originally Posted by James Heinrich

With those minor changes everything went fine until I got to the ggnfs section, and found that Ben's archive is no longer downloadable from https://sites.google.com/site/bbuhrow/lasieve4_64.7z

Google forced me to migrate that page a while back, looks like the linked files didn't make it through that process. I'll see if I can fix it.

Quote:

Originally Posted by James Heinrich

I finally built yafu with make x86_64 NFS=1 USE_SSE41=1 USE_AVX2=1 since my CPU (Xeon(R) Gold 5120) apparently supports both SSE4.1 and AVX2 but about 5 seconds into yafu "tune" I get a segfault. I rebuilt it without AVX2 and it completed tune without error.

If you are willing, try the AVX2 build again by adding USE_BMI2=1 on the make line and see if that helps.

James Heinrich · 2020-11-12, 22:21

Quote:

Originally Posted by bsquared

If you are willing, try the AVX2 build again by adding USE_BMI2=1 on the make line and see if that helps.

Nope, still dies at the same point:

Code:

./yafu "tune"


Setting THREADS = 1 for tuning

starting SIQS on c60: 349594255864176572614071853194924838158088864370890996447417

==== sieving in progress (1 thread):    3568 relations needed ====
====           Press ctrl-c to abort and save state           ====
944 rels found: 353 full + 591 from 5250 partial, (1464.55 rels/sec)
Max specified relations found

sieve time = 0.0000, relation time = 0.0000, poly_time = 0.0000
trial division touched 558762 sieve locations out of 6085410816


elapsed time for ~10k relations of c60 = 7.4636 seconds.
extrapolated time for complete factorization = 11.8057 seconds

starting SIQS on c65: 34053408309992030649212497354061832056920539397279047809781589871

==== sieving in progress (1 thread):    6448 relations needed ====
====           Press ctrl-c to abort and save state           ====
Segmentation fault (core dumped)

2020-09-04, 22:06	#13
paulunderwood Sep 2002 Database er0rr 3,533 Posts	In Makefile, try commenting out: Code: INC += -I../gmp/include LIBS += -L../gmp/lib/linux/x86_64 and add in its place: Code: INC += -I/usr/local/include LIBS += -L/usr/local/lib Is gmp.h in /usr/local/include/ ? Last fiddled with by paulunderwood on 2020-09-04 at 22:08

2020-09-05, 00:00	#16
EdH "Ed Hall" Dec 2009 Adirondack Mtns 5×709 Posts	I do need to change something on my YAFU page. Thanks for pointing it out. Only the "wip" Makefile has the 7.0.3 reference: Code: CC = gcc-7.3.0 The "trunk" version does not. Glad to read that you are up and running.

2020-11-12, 22:00	#20
James Heinrich "James Heinrich" May 2004 ex-Northern Ontario 2²·811 Posts	Revisiting this again, this time on my new CentOS-8 server, I thought I'd pass on a few notes and observations in case someone else as inexperienced as I am tries this: Package names: Before anything else I needed to run yum --enablerepo=extras install epel-release or I wouldn't be able to find any of these packages. apt-get replaced with yum mostly works fine, but a few package names need translating: g++ -> gcc-c++ zlib1g-dev -> zlib-devel I also had to install xz since it wasn't installed by default. With those minor changes everything went fine until I got to the ggnfs section, and found that Ben's archive is no longer downloadable from https://sites.google.com/site/bbuhrow/lasieve4_64.7z I had previously downloaded these sievers on my old server so I copied them over. The instructions also talk about how to build these from source, but the "apply the following patches" section was a bit intimidating (at least using vi, perhaps I should be using some tool to apply those patches?) so I skipped that. I finally built yafu with make x86_64 NFS=1 USE_SSE41=1 USE_AVX2=1 since my CPU (Xeon(R) Gold 5120) apparently supports both SSE4.1 and AVX2 but about 5 seconds into yafu "tune" I get a segfault. I rebuilt it without AVX2 and it completed tune without error. I threw my pet number at it and it found the factors I expected. On my previous 8-core Atom C2750 it took 2658 seconds, on my new 6-core (VM) Xeon Gold 5120 it took 1368 seconds. Last fiddled with by James Heinrich on 2020-11-12 at 22:27

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Running YAFU via Aliqueit doesn't find yafu.ini	EdH	YAFU	8	2018-03-14 17:22
Adventures with 16f siever	VBCurtis	Factoring	6	2018-01-24 11:06
Building yafu on windows/linux	2147483647	YAFU	19	2016-12-09 07:59
The Adventures of a Donkey	a1call	Puzzles	9	2016-05-27 16:50
Building gcc 4.4.0	CRGreathouse	Software	1	2009-07-07 22:25

2020-09-04, 22:22	#15
paulunderwood Sep 2002 Database er0rr DCD₁₆ Posts	My pleasure. I think the local include/link caught ecm too