• State Resolved
  • Locked Locked
  • Replies 2 replies
  • Subscribers 97 subscribers
  • Views 190 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.

TDA2SX: How do EVE and DSP collaborate in ssdJacintoNetV2 network?

dolby Yu
Intellectual 310 points
Part Number: TDA2SX

Hi

As suggested by TIDL model import step, layersGroupId are set as '0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 2 2 2 0' in ssdJacintoNetV2 network.

Which mean DSP and EVE must run alternately in one frame.

However, the buffer in tidl_OD usecase can only be passed from EVE to DSP.

Can you explain some priciples in this collabration between DSP and EVE?

Thanks

Dolby

  • 0
    TI__Genius 17580 points

    Hi Dolby,

    >> Which mean DSP and EVE must run alternately in one frame.

    No, EVE can complete all the layers with layersGroupId as " 1 " first and then can pass the buffer to DSP to process the layers with layersGroupId as "2". Because all the layers that are running on EVE are mostly convolution layers so those can be completed on the EVE first and then on the DSP the concat, softMax and DetectionLayers will be executed later.

    You can refer to FAQ 21 and 22 in the TIDL user guide for more details about this.

    Thanks,

    Praveen 

  • 0 in reply to Praveen Eppa1
    Intellectual 310 points

    Thanks for your reply.

    Dolby