You copied the Doc URL to your clipboard.

Arm Performance Libraries Quick Reference

Compiler command options that control the use of Arm Performance Libraries (ArmPL).

Basic usage

To link Arm Performance Libraries, and provide a serial implementation with 32-bit integers that are optimized for the host CPU, use:

-mcpu=native -armpl

To tailor Arm Performance Libraries to your application and hardware, there are three decisions to make:

  • Decision 1: Which microarchitecture are you compling for?

    To compile for the host CPU, use:

    -mcpu=native
    

    To compile for a specific CPU microarchitecture, use:

    -mcpu=<target-processor>
    
  • Decision 2: Do you want an OpenMP-enabled build?

    To use serial Arm Performance Libraries:

    -armpl
    

    (defaults to no OpenMP)

    To use parallel Arm Performance Libraries

    -armpl=parallel
    

    Note

    In Fortran, it is equivalent to specify:

    -armpl -fopenmp
    
  • Decision 3: Do you need 32-bit or 64-bit integers?

    To use 32-bit integers (the default):

    -armpl
    

    To use 64-bit integers:

    -armpl=ilp64
    

    Note

    • In Fortran is is equivalent to specify

      -armpl -i8
      
    • Options can be combined, for example:

      -armpl=ilp64, parallel
      

Porting to Arm Performance Libraries

Most users port their codes without issues because all calls are the same as they have been on previous systems.

Warning

If applicable, ensure you include armpl.h rather than, for example, mkl.h.

Was this page helpful? Yes No