Reputation: 21
I want to compute the optimal gridsize and blocksize in order to have a high rate of occupancy for a kernel launch. I have seen that there is a function in CUDA to do so, which is this one:
cudaOccupancyMaxPotentialBlockSize()
I am using PyCUDA and I would like to know if an equivalent function exists in PyCUDA.
Upvotes: 1
Views: 1322
Reputation: 72349
As best as I can tell, the driver API occupancy functions, including cuOccupancyMaxPotentialBlockSize
, have not been implemented in PyCUDA, as of version 2015.1.3, based on a search of the Github mirror.
Upvotes: 1