• State Accepted Answer
  • Locked Locked
  • Replies 3 replies
  • Subscribers 20 subscribers
  • Views 2775 views
  • Users 0 members are here
Options
Related
How was your experience today?
This discussion has been locked.
You can no longer post new replies to this discussion. If you have a question you can start a new discussion

__pgmath_abort:Math dispatch table is either misconfigured or corrupted.

ndvn

Hello,

I am getting the following runtime error:

Error during math dispatch processing...

__pgmath_abort:Math dispatch table is either misconfigured or corrupted.

It occurs when my Fortran code is built with nvfortran and libamath.so is loaded into the process. After testing some debug flags, I came to the following:

**** div/zs/precise ****Entry point not defined for CPU target=armv8.

Maybe libamath.so lacks implementations of some math functions?

  • +1 Arm Employee Badge

    Hi ndvn

    What's likely happening is that nvfortran is asking for a function that either libamath does not provide or provides with a different naming signature. libamath is not designed to be compatible with nvfortran. You should use the math runtime library that comes with the Nvidia toolchain.

    Ta

    Rich

  • 0

    Yes, libamath.so contains BLAS and LAPACK routines. MLSDev software development company specializes in creating custom software solutions tailored to the individual needs and requirements of businesses or customers. These companies typically work with specialized programming languages and frameworks to deliver solutions that meet the customer's desired goals. They are usually knowledgeable in the most up-to-date technology trends and tools, offering their clients the ability to stay ahead of the competition.

  • 0 Arm Employee Badge in reply to VivianT

    Hi.

    I think a little clarification of nomenclature is useful here.

    • Arm Performance Libraries is the product that contains BLAS, LAPACK, FFTs and sparse functionality.  These are contained in objects with names like 'libarmpl_lp64.so'.
    • Optimized math.h calls, including vectorized versions, are provided through, e.g., 'libamath.so'.

    As Rich says, the versions of Arm Performance Libraries that are out today are compatible with both GCC and Arm Compiler for Linux toolchains. The NVHPC toolchain, through its 'pgmath' interface may have different functionality for certain math.h functions and this will be what is causing the error messages seen.

    If a version compatible with NVHPC was of interest that would be good input for future work.

    Hope this helps.

    Chris