In assignment2 handout it was mentioned that recursion is not supported in CUDA. What is the reason behind that? My wild guess is, its due to limited amount of memory available for each thread?
Faust
I did a little research, and it seems that CUDA does actually support recursion, but needs a Fermi card. It appears that CUDA's previous lack of recursion support is that the overhead of recursively calling the function would cause poor performance on the GPU.
In assignment2 handout it was mentioned that recursion is not supported in CUDA. What is the reason behind that? My wild guess is, its due to limited amount of memory available for each thread?
I did a little research, and it seems that CUDA does actually support recursion, but needs a Fermi card. It appears that CUDA's previous lack of recursion support is that the overhead of recursively calling the function would cause poor performance on the GPU.
https://developer.nvidia.com/cuda-toolkit-31-downloads