• State
  • Locked Locked
  • Replies 1 reply
  • Subscribers 97 subscribers
  • Views 264 views
  • Users 0 members are here
Support feedback
Options
Options
Related

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

SK-AM62: Low level API for GEMM

Aid Vllasaliu
Prodigy 80 points
Part Number: SK-AM62
Other Parts Discussed in Thread: AM68A

Hi,

Do you have any C/C++ examples or documentation for low level GEMM AI accelerators?

We're trying to get our custom models to work, as TF/TFLite and other solutions are not sufficient for or NN models.

  • 0
    TI__Genius 16006 points

    Hello,

    By GEMM do you mean general matrix multiply? Please clarify.

    The AM62X processor does not have an accelerator for matrix multiply. The AM62A does feature this and has a library, which we have documentation for. I do not know of any examples for GEMM on AM62X since it does not have acceleration for these operations outside what you might get from code optimization on Arm A53 cores.

    We support ONNX and TFlite on the AM62X. Other processors in the AM6xA family (e.g. AM62A, AM68A) also have TVM support. edgeai-tidl-tools repository on Github has examples and documentation for AM6xA processors and may be more relevant to your request on GEMM

    Best,

    Reese