Request: In-depth description for migrating off of NumPy

[ClimbsRocks]

It seems like the biggest opportunity for this project is not to be used in new, small projects that are just getting off the ground, but rather to be used as a NumPy replacement for the many large projects that already rely on NumPy.

It would be really useful for those projects to have in-depth documentation on what they'll have to change to migrate from NumPy to CuPy. Ideally this would include:

• A list of NumPy functions for which CuPy is a drop-in replacement, and no other modification is needed
• A list of which NumPy operations are not supported yet
• Any other changes the codebase will have to make, beyond the obvious drop-in ndarray replacement, to support this
• Any changes or limitations the other project's users will have to make to take advantage of CuPy

Can't wait to see this gain wider adoption! GPU-speed machine learning algorithms sounds like a dream.