GPU Preemption on TX2


Does anyone know if GPU preemption is enabled on TX2? If not, is there any kernel patch that can be applied?


Hi Cookie007, yes GPU preemption is enabled on TX2. You can individually verify it by checking the cudaDevAttrComputePreemptionSupported flag using the cudaDeviceGetAttribute() function.

Thank you dusty_nv!
Since I am new to TX2, wanted to know how to set the priority in the code. Is there any flag that I can set manually?

It would be done the same way as on discrete GPUs, by using the cudaStreamCreateWithPriority() function.