Illegal Instruction with example from docs

[votroto]

File in.ms contains the following:

z1, z2, z3
1073741827
7*z1*z2+5*z2*z3+z3^2+z1+5*z3+10,
7*z3^2-27*z1^2-15*z2^2+59*z3+3*z1,
8*z1^2+13*z1*z3+10*z3^2+z2+z1

Now, typing the following command ./msolve -g 2 -f in.ms -o out.ms results in Illegal instruction (core dumped).

versions

Same result on AMD epic and on Ryzen 7, same with the 0.6.5 release on the website and the 0.6.5-1 in the Arch linux packages.

the coredump info

$ coredumpctl info 19647
           PID: 19647 (msolve)
           UID: 1000 (votroto)
           GID: 1000 (votroto)
        Signal: 4 (ILL)
     Timestamp: Wed 2024-04-03 14:30:49 CEST (2h 31min ago)
  Command Line: msolve -g 1 -f cona.ms -o conagb.ms
    Executable: /usr/bin/msolve
 Control Group: /user.slice/user-1000.slice/session-2.scope
          Unit: session-2.scope
         Slice: user-1000.slice
       Session: 2
     Owner UID: 1000 (votroto)
       Boot ID: 8f23e8e8a7a344b1a1c4774af2e2ad34
    Machine ID: 0a28cbc9d0c445bba276547b812bd8b2
      Hostname: votrotot14
       Storage: /var/lib/systemd/coredump/core.msolve.1000.8f23e8e8a7a344b1a1c4774af2e2ad34.19647.1712147449000000.zst (present)
  Size on Disk: 297.3K
       Message: Process 19647 (msolve) of user 1000 dumped core.

                Stack trace of thread 19647:
                #0  0x00007b7425ef2816 initialize_basis_hash_table (libneogb-0.6.5.so + 0x8816)
                #1  0x00007b7425ef478b initialize_basis (libneogb-0.6.5.so + 0xa78b)
                #2  0x00005a5ffe79f6c2 n/a (msolve + 0x286c2)
                #3  0x00005a5ffe781dc6 n/a (msolve + 0xadc6)
                #4  0x00005a5ffe77c5b8 n/a (msolve + 0x55b8)
                #5  0x00007b7424c3bcd0 n/a (libc.so.6 + 0x25cd0)
                #6  0x00007b7424c3bd8a __libc_start_main (libc.so.6 + 0x25d8a)
                #7  0x00005a5ffe77cde5 n/a (msolve + 0x5de5)
                ELF object binary architecture: AMD x86-64

[ederc]

I tested msolve v0.6.5 from the website on AMD EPYC 9554 on gentoo with your above given example and I have no problems to compute the Gröbner basis in out.ms.

On the other hand, looking at the coredump, it seems that you use a different call and a different example:

Command Line: msolve -g 1 -f cona.ms -o conagb.ms

Can you provide me the cona.ms input.

[votroto]

Same contents, just different filename.

Whoops, just noticed the "intel" part of the name! (It works fine on an intel server...)

Beg my pardon. Couldn't get the dependencies on the server and downloaded the (wrong) binary. Then, it's just a case of a blind user in one instance, and probably an arch linux packaging error in the other.