Question is the title.
I was reading this article:
https://stackoverflow.com/questions/4391162/cuda-determining-threads-per-block-blocks-per-grid
and it said something about there being 16 CUDA blocks maximum and if you used more you would be wasting time.
What is the limit on number of blocks for the Jetson Nano?
Maybe you can find the desired information here: CUDA Version features andspecifications
Jetson Nano has CUDA compute capability (version) 5.3.
There is a line:
“Maximum number of resident blocks per multiprocessor” which is 32. I assume that number refers to currently running blocks (is there such a thing as a non-running block?).
So I think my question is answered. Thanks.