Add support for passing cupy arrays to "C" lang

[bouweandela]

Add support for passing cupy arrays to "C" lang compiler.

Closes #224

[bouweandela]

I'm not entirely sure where to update the documentation and examples. I think e.g. this example
https://github.com/bouweandela/kernel_tuner/blob/4719acdf4ff096a6e9586e0148c6a9753335393d/examples/cuda/convolution_streams.py#L22-L23
could be simplified by using cupyx.zeros_pinned to allocate the memory. Please let me know if I missed anything.

[benvanwerkhoven]

Thanks a lot for your work on this Bouwe!

I think actually the Point-in-Polygon example might be easier to adapt:
https://github.com/kerneltuner/kernel_tuner/blob/master/examples/cuda/pnpoly.py
The current example uses PyCUDA and indeed passes a GPU memory allocated created using PyCUDA to the Compiler backend. So you could create a variant of that using CuPy instead of PyCUDA, now that you've made this possible. The advantage of only changing the point-in-polygon example is that you don't need to change the CUDA code.

For the time being, I think it's fine to just copy the old example to a new file and call this pnpoly_cupy.py. Perhaps at some point we should consider grouping examples into directories by what backend they are using.

[sonarcloud]

Kudos, SonarCloud Quality Gate passed!   

0 Bugs
0 Vulnerabilities
0 Security Hotspots
0 Code Smells

No Coverage information
0.0% Duplication

[bouweandela]

For the time being, I think it's fine to just copy the old example to a new file and call this pnpoly_cupy.py.

Done in 303ef3a

[benvanwerkhoven]

Thanks a lot @bouweandela ! I've extensively tested everything in this pull request, it all seems to work perfectly!