Add note in CUDA docs about signature in @cuda.jit causing immediate compilation

[stuartarchibald]

Feature request

As title. Follow up to #4052

[joesinghh]

I'll workon this

[stuartarchibald]

Great, thanks @Joe-Sin7h

[joesinghh]

hey @stuartarchibald i'm new to open source.
Could you help me understand this issue because i'm unable to.

[p3732]

@Joe-Sin7h let me try to fill in some info. As the description states this is a follow-up to #4052 which was closed with an intention of adding information to the device-functions documentation.

I read through the thread again and from what I figured it is the case that device functions, so anything with @cuda.jit(device=True), are not compiled via the JIT compiler, but immediately with the python compiler.

Now, I am not sure if this is completely correct, but I thought it would be easier for someone who knows to just confirm this, than to write it themselves.