Menu

#335 big matrix crash (symm but not posdef)

v1.0_(example)
open
nobody
None
5
5 days ago
6 days ago
No

int-32 overflow in a rather special case (symm matrix, but not pos def, and very big);
this line in real_invert_symmetric_matrix seems to overflow :

int bytes = n * n * sizeof *aval; ``

1 Attachments

Discussion

  • Allin Cottrell

    Allin Cottrell - 5 days ago

    The line referenced above should be size_t bytes = n * n * sizeof *aval; and it's now corrected in git.

    A note on M = upper(M) + upper(M)' : given that M is 23171 x 23171, this line calls for more than 12GB of storage., and so is likely to bring some computers to their knees.

     
    • Sven Schreiber

      Sven Schreiber - 5 days ago

      Am 05.09.2026 um 14:51 schrieb Allin Cottrell:

      The line referenced above should be size_t bytes = n * n * sizeof *aval; and it's now corrected in git.

      A note on M = upper(M) + upper(M)' : given that M is 23171 x 23171, this line calls for more than 12GB of storage., and so is likely to bring some computers to their knees.

      Right - this was just a quick proof, not a refined test script. I'm not
      used to having symmetric-but-indefinite matrices in my scripts.

      thanks, sven

       

Log in to post a comment.