Hello,
I know that since capability 2.0, CUDA has hardware support for 3D grid where z >= 1. The JCuda API, however, only has methods such as, KernelLauncher.setGrid(x,y) and I did not see any support for the z dimension. Is there anyway around this or will this be added later?
In the ‘cuLaunchKernel’ you can specify the 3D grid size, of course. Specifically regarding the KernelLauncher: The latest version (0.0.4) has a method “setGridSize(int x, int y, int z)” (and the old one for backward compatibility). Maybe you have an outdated version?