Msieve number field sieve polynomial selection failing

[BigNumberGuy]

Hello, when I attempt to run the command "./msieve <integer> -v -np", it briefly prints info before terminating with "bus error: ./msieve <integer> -v -np" with no further info.

[swellman]

Best practice is to place <integer> in a flat text file called worktodo.ini in the msieve directory.

See https://www.mersenneforum.org/showpo...0&postcount=12 for advice on improving efficiency and results of poly searching with msieve. I use stage1_norm of 5.5% of default value and stage2_norm of 6% of default value but YMMV.

If you are searching for a GNFS poly of a larger composite, say > c180, you should consider using CADO.

[chris2be8]

That's not enough information for us to tell what's wrong.

Please post all the output you did get, especially the msieve version which should be on the first line of output.

Then post whatever ended up in msieve.log and msieve.fb (I am assuming you can write to wherever you are running msieve).

And what OS are you running and what sort of system is it?


Edit, crossposted with swellman.
What he did should work, I just tested it on a c93.

[BigNumberGuy]

Originally Posted by swellman

Best practice is to place <integer> in a flat text file called worktodo.ini in the msieve directory.

See https://www.mersenneforum.org/showpo...0&postcount=12 for advice on improving efficiency and results of poly searching with msieve. I use stage1_norm of 5.5% of default value and stage2_norm of 6% of default value but YMMV.

If you are searching for a GNFS poly of a larger composite, say > c180, you should consider using CADO.

I have tried that as well, exits with the same error.

[BigNumberGuy]

Originally Posted by chris2be8

That's not enough information for us to tell what's wrong.

Please post all the output you did get, especially the msieve version which should be on the first line of output.

Then post whatever ended up in msieve.log and msieve.fb (I am assuming you can write to wherever you are running msieve).

And what OS are you running and what sort of system is it?


Edit, crossposted with swellman.
What he did should work, I just tested it on a c93.

Msieve v. 1.51 (SVN unknown)
Fri Mar 17 08:47:07 2023
random seeds: 7c422c0b 9b5008c6
factoring 278928736249208979543689230453796583940253796839402679583485674358995909409509483756324351678942875362738490985478637875828757264343234325617829811000909487652 (159 digits)
searching for 15-digit factors
P-1 stage 2 factor found
commencing number field sieve (134-digit input)
commencing number field sieve polynomial selection
polynomial degree: 5
max stage 1 norm: 1.88e+20
max stage 2 norm: 4.06e+18
min E-value: 3.76e-11
poly select deadline: 91784
time limit set to 25.50 CPU-hours
expecting poly E from 4.50e-11 to > 5.17e-11
searching leading coefficients from 1 to 60
deadline: 200 CPU-seconds per coefficient
coeff 12 specialq 1 - 47493 other 21110 - 50664
zsh: bus error  ./msieve  -v -np "1,60"

(ignore the number, that is a random number for testing)

I attached msieve.log here.
I don't see a msieve.fb in the directory.

I am on MacOS with Apple M1, ARM64

[chris2be8]

Thanks, that gives us something to go on.

What you are doing looks reasonable.

The thing I suspect is "Apple M1, ARM64". I'm using x86_64 (Intel and AMD CPUs) so you are not running the same code as me.

Where did you get the msieve program from? Did you compile it yourself?

But this will need help from someone who understands the code. Can you get a core dump or run it under a debugger?

[jasonp]

v1.51 is pretty old. If you are linking against a modern version of GMP there is a critical bugfix in svn r1033 needed to avoid crashes, this may be what is happening.

[BigNumberGuy]

msieve postprocessing seems to exit with the same errors

[BigNumberGuy]

Originally Posted by chris2be8

Thanks, that gives us something to go on.

What you are doing looks reasonable.

The thing I suspect is "Apple M1, ARM64". I'm using x86_64 (Intel and AMD CPUs) so you are not running the same code as me.

Where did you get the msieve program from? Did you compile it yourself?

But this will need help from someone who understands the code. Can you get a core dump or run it under a debugger?

Compiled it myself, removed the -march flag because my compiler does not want to compile it for some reason