Batched operations for Python bindings?

[ethanhs]

Hi! Really neat stuff here, I'm curious as to whether it would be possible to expose the batched gemm/axpy operations to Python. I'm hoping to use them in a project I'm working on.

Thanks for the awesome project!

[CNugteren]

I checked and indeed all special (non standard BLAS) routines are not in pyclblast. I see no reason why they wouldn't be, except for a little bit of work. I'll look at this in the coming weeks when I have time.

[ethanhs]

Thank you! I realized that I probably will need to use the opencl kernels directly after all for my particular use case, so don't feel any pressure.

[CNugteren]

Code is updated (see #386), and an example is included for batched-AXPY.

But I'm having issues updating the official PyCLBlast package, but of an issue with dependencies and the way I've build them locally: following this doesn't work, since it results in the legacy linux_x86_64 tag, but I believe we need a manylinux tag, see e.g. here. If someone is willing to do that for me, please go ahead. In the meantime I have attached a Python 3.7 linux wheel here, or you can of course check out the source and build it yourself if it is not compatible.