yafu bugs

[bsquared]

Originally Posted by Kvasir

I noticed something strange this morning in the yafu output:

04/19/21 00:47:13, nfs: commencing msieve filtering
04/19/21 01:08:14, nfs: commencing msieve linear algebra
04/19/21 02:06:43, nfs: commencing msieve sqrt
04/19/21 02:38:55, prp66 = 192303762923047619052802638244647992855880578333545590122359339543
04/19/21 02:38:55, prp66 = 55752044095394727965374636253930753410057647404899226012311297057
04/19/21 02:38:55, NFS elapsed time = 100460.3539 seconds.

Those two primes are clearly not of equal length, in fact the last one is a p65.

Huh, guess I need to rtfm a little closer:

Originally Posted by gmplib.org

Function: size_t mpz_sizeinbase (const mpz_t op, int base)

Return the size of op measured in number of digits in the given base. base can vary
from 2 to 62. The sign of op is ignored, just the absolute value is used. The result will
be either exact or 1 too big. If base is a power of 2, the result is always exact. If op is
zero the return value is always 1.

This function can be used to determine the space required when converting op to a
string. The right amount of allocation is normally two more than the value returned by
mpz_sizeinbase, one extra for a minus sign and one for the null-terminator.

It will be noted that mpz_sizeinbase(op,2) can be used to locate the most significant 1
bit in op, counting from 1. (Unlike the bitwise functions which start from 0, See Logical
and Bit Manipulation Functions.)

Thanks, I learned something already today.

[bur]

Originally Posted by bsquared

Are those spaces in front of those options? Yafu skips lines that start with whitespace so just remove them and you'll hopefully be good.

That was it, thanks.

[wombatman]

Running into an issue while trying to do some testsieving (and gnfs/snfs in general) with YAFU.

Specifically, the tune() function runs just fine all the way through both the SIQS and NFS testing.

I run something like

./yafu "nfs(rsa(500))" -v

I get a segmentation fault:

polynomial degree: 5
max stage 1 norm: 8.05e+22
max stage 2 norm: 6.57e+19
min E-value: 4.11e-12
poly select deadline: 554355
time limit set to 153.99 CPU-hours
expecting poly E from 7.00e-12 to > 8.05e-12
searching leading coefficients from 8192 to 8442
deadline: 8640000 CPU-seconds per coefficient
coeff 8280 specialq 1 - 7293164 other 28289 - 67893
Segmentation fault

YAFU and all related programs (msieve, ecm, etc) are built in WSL2 without issue. Any idea what might be causing this?

[bsquared]

Originally Posted by wombatman

Running into an issue while trying to do some testsieving (and gnfs/snfs in general) with YAFU.

Specifically, the tune() function runs just fine all the way through both the SIQS and NFS testing.

I run something like

./yafu "nfs(rsa(500))" -v

I get a segmentation fault:

polynomial degree: 5
max stage 1 norm: 8.05e+22
max stage 2 norm: 6.57e+19
min E-value: 4.11e-12
poly select deadline: 554355
time limit set to 153.99 CPU-hours
expecting poly E from 7.00e-12 to > 8.05e-12
searching leading coefficients from 8192 to 8442
deadline: 8640000 CPU-seconds per coefficient
coeff 8280 specialq 1 - 7293164 other 28289 - 67893
Segmentation fault

YAFU and all related programs (msieve, ecm, etc) are built in WSL2 without issue. Any idea what might be causing this?

What version of msieve? I've seen this before, yes. It could be related to the gmp-6.2.0 change that was incompatible with some low-level conversions in msieve.

[wombatman]

Originally Posted by bsquared

What version of msieve? I've seen this before, yes. It could be related to the gmp-6.2.0 change that was incompatible with some low-level conversions in msieve.

It's Revision 1041. GMP version is 6.2.1.

[bsquared]

Originally Posted by wombatman

It's Revision 1041. GMP version is 6.2.1.

Ok. I'm not seeing the problem with linux or visual studio windows versions as of now, but I'll check a WSL build tonight.

[wombatman]

Originally Posted by bsquared

Ok. I'm not seeing the problem with linux or visual studio windows versions as of now, but I'll check a WSL build tonight.

Thanks. If there's something I can do to provide more information, please let me know.

[wombatman]

As a quick follow-up, I checked the WSL2-built msieve to make sure it was functioning properly. I used a 100-digit number for the sake of brevity, but msieve started both the polynomial search and sieve stages with no issue. I didn't bother with the linear algebra part, since that's not where I'm hitting the segmentation fault.

[bsquared]

I updated my wsl2 copy of msieve to 1041, rebuilt everything, and I'm still not seeing the crash. I have been making a few changes to yafu 2.0 lately, so if you haven't yet, update to get those. If it still crashes, please run the job in gdb and backtrace when it segfaults to help pinpoint where it's happening.

[wombatman]

Originally Posted by bsquared

I updated my wsl2 copy of msieve to 1041, rebuilt everything, and I'm still not seeing the crash. I have been making a few changes to yafu 2.0 lately, so if you haven't yet, update to get those. If it still crashes, please run the job in gdb and backtrace when it segfaults to help pinpoint where it's happening.

That did the trick! I had forgotten that because I modified yafu.ini, "git pull" wouldn't actually fetch anything new. Once I renamed that, I got the updated files, recompiled, and now it works as expected.

[unconnected]

How to set minimum number of relations when filtering should started in YAFU? Seems like min_rels estimation somehow broken for small numbers.

06/23/21 05:06:01 v1.34.5 @ workstation, ****************************
06/23/21 05:06:01 v1.34.5 @ workstation, Starting factorization of 4769680008811953517130252531389050839921028010533834008817232296514476227210306415512533832918083
06/23/21 05:06:01 v1.34.5 @ workstation, using pretesting plan: normal
06/23/21 05:06:01 v1.34.5 @ workstation, no tune info: using qs/gnfs crossover of 95 digits
06/23/21 05:06:01 v1.34.5 @ workstation, input indicated to have been pretested to t25.00
06/23/21 05:06:01 v1.34.5 @ workstation, ****************************
06/23/21 05:06:01 v1.34.5 @ workstation, rho: x^2 + 3, starting 1000 iterations on C97
06/23/21 05:06:01 v1.34.5 @ workstation, rho: x^2 + 2, starting 1000 iterations on C97
06/23/21 05:06:01 v1.34.5 @ workstation, rho: x^2 + 1, starting 1000 iterations on C97
06/23/21 05:06:01 v1.34.5 @ workstation, pm1: starting B1 = 3750K, B2 = gmp-ecm default on C97
06/23/21 05:06:06 v1.34.5 @ workstation, current ECM pretesting depth: 25.33
06/23/21 05:06:06 v1.34.5 @ workstation, scheduled 390 curves at B1=250000 toward target pretesting depth of 29.85
06/23/21 05:16:16 v1.34.5 @ workstation, Finished 390 curves using Lenstra ECM method on C97 input, B1=250K, B2=gmp-ecm default
06/23/21 05:16:16 v1.34.5 @ workstation, final ECM pretested depth: 29.86
06/23/21 05:16:16 v1.34.5 @ workstation, scheduler: switching to sieve method
06/23/21 05:16:16 v1.34.5 @ workstation, nfs: commencing nfs on c97: 4769680008811953517130252531389050839921028010533834008817232296514476227210306415512533832918083
06/23/21 05:16:16 v1.34.5 @ workstation, nfs: commencing poly selection with 1 threads
06/23/21 05:16:16 v1.34.5 @ workstation, nfs: setting deadline of 964 seconds
06/23/21 05:32:20 v1.34.5 @ workstation, nfs: completed 12 ranges of size 250 in 963.7039 seconds
06/23/21 05:32:20 v1.34.5 @ workstation, nfs: best poly = # norm 3.118081e-13 alpha -5.832394 e 1.825e-08 rroots 2
06/23/21 05:32:20 v1.34.5 @ workstation, nfs: commencing lattice sieving with 1 threads
06/23/21 05:41:37 v1.34.5 @ workstation, nfs: commencing lattice sieving with 1 threads
06/23/21 05:51:18 v1.34.5 @ workstation, nfs: commencing lattice sieving with 1 threads
06/23/21 06:00:53 v1.34.5 @ workstation, nfs: commencing lattice sieving with 1 threads
06/23/21 06:09:52 v1.34.5 @ workstation, nfs: commencing lattice sieving with 1 threads
06/23/21 06:19:38 v1.34.5 @ workstation, nfs: commencing lattice sieving with 1 threads
06/23/21 06:28:50 v1.34.5 @ workstation, nfs: commencing lattice sieving with 1 threads
06/23/21 06:38:44 v1.34.5 @ workstation, nfs: commencing lattice sieving with 1 threads
06/23/21 06:49:01 v1.34.5 @ workstation, nfs: commencing lattice sieving with 1 threads
06/23/21 06:58:44 v1.34.5 @ workstation, nfs: commencing lattice sieving with 1 threads
06/23/21 07:09:12 v1.34.5 @ workstation, nfs: commencing lattice sieving with 1 threads
06/23/21 07:20:04 v1.34.5 @ workstation, nfs: commencing msieve filtering
06/23/21 07:21:16 v1.34.5 @ workstation, nfs: raising min_rels by 5.00 percent to 2216153
06/23/21 07:21:16 v1.34.5 @ workstation, nfs: commencing lattice sieving with 1 threads
06/23/21 07:32:04 v1.34.5 @ workstation, nfs: commencing msieve filtering
06/23/21 07:33:26 v1.34.5 @ workstation, nfs: raising min_rels by 5.00 percent to 2420636
06/23/21 07:33:26 v1.34.5 @ workstation, nfs: commencing lattice sieving with 1 threads
06/23/21 07:43:52 v1.34.5 @ workstation, nfs: commencing msieve filtering
06/23/21 07:45:15 v1.34.5 @ workstation, nfs: raising min_rels by 5.00 percent to 2629918
06/23/21 07:45:15 v1.34.5 @ workstation, nfs: commencing lattice sieving with 1 threads
06/23/21 07:55:55 v1.34.5 @ workstation, nfs: commencing msieve filtering
06/23/21 07:57:31 v1.34.5 @ workstation, nfs: raising min_rels by 5.00 percent to 2845238
06/23/21 07:57:31 v1.34.5 @ workstation, nfs: commencing lattice sieving with 1 threads
06/23/21 08:08:03 v1.34.5 @ workstation, nfs: commencing msieve filtering
06/23/21 08:10:10 v1.34.5 @ workstation, nfs: commencing msieve linear algebra
06/23/21 08:20:27 v1.34.5 @ workstation, nfs: commencing msieve sqrt
06/23/21 08:21:21 v1.34.5 @ workstation, prp46 = 7412512799061243486864906593531917503138920627
06/23/21 08:21:21 v1.34.5 @ workstation, prp51 = 643463308342079205421383456862481650555873117703729
06/23/21 08:21:21 v1.34.5 @ workstation, NFS elapsed time = 11104.6899 seconds.