Will the arrays dynamically created by cuda in the kernel function be automatically released?

Will the arrays dynamically created by

cuda in kernel functions be released by themselves? If not, how should I free up the memory occupied by the array on the gpu?

Apr.09,2021
Menu