TDA3: Improve the operation speed of the vision algorithm

Hi,

For which target you are talking about?

On TDA3x, we have EVE and DSP both programmable processors to accelerate vision algorithms. Using EDMA and cache appropriately is very important to get the best performance but apart from that one can optimize the code for the given processor (DSP/EVE) by using the specialized instructions offered by the processor. You can make loops in your code, restructure the code to make it more friendly for vector cores and other core-specific optimizations can be applied. I would suggest you to go through below training material

• C66x DSP: https://training.ti.com/c6000-embedded-design-workshop
• EVE: Programmer’s guide of EVE and other training material – Available in CDDS from TI and would need you to have specific NDA with TI

Thanks,
with Regards,
Pramod