Msieve with MPI block Lanczos - Page 10

debrouxl · 2013-05-27, 17:41

-fomit-frame-pointer and -march=native are options suitable for GCC, not for MSC.

wombatman · 2013-05-27, 17:42

Ok. Looks like I'll have to alter the makefile then.

poily · 2013-06-13, 10:12

I'm working on a new parallel version of msieve's LA and already made some progress. But currently all data I can test it on is 50K matrix from a very small factorization. In order to see the expected speedup I need matrices of size at least 500K or better 1-5M. Unfortunately I lost my data from RSA180, RSA190 and other factoring experiments. Can anyone here help me with the testing? All I need is .fb, .idx and .mat of size >500K.

xilman · 2013-06-13, 11:01

Quote:

Originally Posted by poily

I'm working on a new parallel version of msieve's LA and already made some progress. But currently all data I can test it on is 50K matrix from a very small factorization. In order to see the expected speedup I need matrices of size at least 500K or better 1-5M. Unfortunately I lost my data from RSA180, RSA190 and other factoring experiments. Can anyone here help me with the testing? All I need is .fb, .idx and .mat of size >500K.

I produce such things every few days on average and there's a fair chance that one or more are on backups even if none are on disk right now. Even if not, another one will be a long shortly.

Let me know your address and we'll see if we can arrange something.

Paul

jasonp · 2013-06-13, 11:44

fivemack has > 200GB of such factorizations, including presumably some very large ones. Note that the .idx file is only create by the code when compiled for MPI, so you may have to at least build the matrix.

poily · 2013-06-17, 11:28

Paul, check pm, please.

2013-06-13, 11:44	#104
jasonp Tribal Bullet Oct 2004 DD5₁₆ Posts	fivemack has > 200GB of such factorizations, including presumably some very large ones. Note that the .idx file is only create by the code when compiled for MPI, so you may have to at least build the matrix. Last fiddled with by jasonp on 2013-06-13 at 11:45

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
block wiedemann and block lanczos	ravlyuchenko	Msieve	5	2011-05-09 13:16
Why is lanczos hard to distribute?	Christenson	Factoring	39	2011-04-08 09:44
Block Lanczos with a reordering pass	jasonp	Msieve	18	2010-02-07 08:33
Lanczos error	Andi47	Msieve	7	2009-01-11 19:33
Msieve Lanczos scalability	Jeff Gilchrist	Msieve	1	2009-01-02 09:32

2013-05-27, 17:41	#100
debrouxl Sep 2009 977 Posts	-fomit-frame-pointer and -march=native are options suitable for GCC, not for MSC.

2013-05-27, 17:42	#101
wombatman I moo ablest echo power! May 2013 29×61 Posts	Ok. Looks like I'll have to alter the makefile then.

2013-06-13, 10:12	#102
poily Nov 2010 2·5² Posts	I'm working on a new parallel version of msieve's LA and already made some progress. But currently all data I can test it on is 50K matrix from a very small factorization. In order to see the expected speedup I need matrices of size at least 500K or better 1-5M. Unfortunately I lost my data from RSA180, RSA190 and other factoring experiments. Can anyone here help me with the testing? All I need is .fb, .idx and .mat of size >500K.

2013-06-17, 11:28	#105
poily Nov 2010 2·5² Posts	Paul, check pm, please.