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.

Linux/AM5728: How to retain memalign from Linux (A15) side to OpenCL Buffers

Part Number: AM5728

Tool/software: Linux

Hi,

We are trying to get VLIB libraries to get called from within OpenCL programs running on the latest Processor SDK on the AM5728 EVM. Since there aren't any reference/example codes to try out, we are figuring stuff out as we go ahead.

How to we make sure the buffers being passed into OpenCL kernels are memory aligned as required by various VLIB function calls?

Also, in case there is even a single, unsupported, unvalidated VLIB example program running on OpenCL, can someone from TI please share that as a starting point?

  • Hi Anup,

    I've notified the design team to elaborate on this. Feedback will be posted here.

    Best Regards,
    Yordan
  • Hi Anup,

    Regarding buffer alignment, enclosed is in our documentation git repo waiting to be pushed to external documentation website.

    - Yuan

    Buffer Alignment
    =======================

    In TI's implementation, global buffers (both DDR and TI extended MSMC)
    and local buffers are always aligned at 128-byte memory boundary. Sub-buffers
    are aligned accordingly to their origin/offset in the existing global buffers.
  • Hi Yuan,

    Thanks a lot. This is very helpful.

    Also, any guidance on VLIB?

    "Also, in case there is even a single, unsupported, unvalidated VLIB example program running on OpenCL, can someone from TI please share that as a starting point?"