• State Accepted Answer
  • Locked Locked
  • Replies 2 replies
  • Subscribers 21 subscribers
  • Views 1033 views
  • Users 0 members are here
2025 survey

We are running a survey to help us improve the experience for all of our members. If you see the survey appear, please take the time to tell us about your experience if you can.

Options
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

Lack of *cscmv functions in ARM performance library

Deng Shi

I did not find the related *cscmv related BLAS functions in ARM performance library like MKL_*cscmv functions.

Is there any plan to cover them in newer version? If yes, what will be the target release?

Parents
  • +1 Arm Employee Badge

    Hi,

    Yes, Arm Performance Libraries have a lot of sparse linear algebra functionality, including the sparse matrix-vector multiplication functionality you mention.  However the interface is slightly different to that in oneMKL.  Indeed the functions you reference are themselves deprecated.  All vendor libraries have moved towards an "inspector-executor" model of operation which means that there is much greater opportunity for optimization and hence increased performance for you.

    For a description of exactly how this works in Arm Performance Libraries have a read of https://developer.arm.com/documentation/101004/2410/Sparse-Linear-Algebra/Sparse-Linear-Algebra-Introduction and look at the example that follows.  The example is in CSR format, rather than CSC, but both are supported in ArmPL, along with other sparse matrix formats.

    Hope this helps.

    Chris

Reply
  • +1 Arm Employee Badge

    Hi,

    Yes, Arm Performance Libraries have a lot of sparse linear algebra functionality, including the sparse matrix-vector multiplication functionality you mention.  However the interface is slightly different to that in oneMKL.  Indeed the functions you reference are themselves deprecated.  All vendor libraries have moved towards an "inspector-executor" model of operation which means that there is much greater opportunity for optimization and hence increased performance for you.

    For a description of exactly how this works in Arm Performance Libraries have a read of https://developer.arm.com/documentation/101004/2410/Sparse-Linear-Algebra/Sparse-Linear-Algebra-Introduction and look at the example that follows.  The example is in CSR format, rather than CSC, but both are supported in ArmPL, along with other sparse matrix formats.

    Hope this helps.

    Chris

Children