Can NVDLA and GPU work in parallel?

Hi,

I am running multiple detection engines with TensorRT on Jetson Xavier

The plan is to have GPU and NVDLA running in different threads of the same process to reduce process time.

But what I got is GPU and NVDLA are working in turn, and execution time increased.

Is there anything that I should pay attention to?

Hi,

You can run the different model in DLAs and GPU at the same time.

Please refer below dev talk:
https://devtalk.nvidia.com/default/topic/1063351/jetson-agx-xavier/i-don-t-know-the-dla-/post/5385153/#5385153

Thanks

Thanks for the reply.

Issue resolved by adding
#define CUDA_API_PER_THREAD_DEFAULT_STREAM