Adjust Block Size for 2D Image Processing

[CherifMZ]

@StRigaud

Description

Currently, the block size for processing 2D images is set to (16, 16) (see here).
https://github.com/clEsperanto/CLIc_prototype/blob/d453ae9802e1e1768455a642c8de09cee875a704/clic/src/cudabackend.cpp#L848
However, I believe that this block size may not be optimal for 2D images since we are working with a block size of 512 and not 256.

Solution

I propose that the block size should be adjusted to (32, 16) when the x-dimension is larger than the y-dimension, and to (16, 32) when the y-dimension is larger than the x-dimension.
Additionally, consider adjusting the block size for images with a z-dimension.