[PATCH] D56411: [CUDA][HIP][Sema] Fix template kernel with function as template parameter

[John McCall via Phabricator via cfe-commits]

rjmccall added a comment.

In D56411#1398291 <https://reviews.llvm.org/D56411#1398291>, @tra wrote:

> >> That said, does CUDA have a general rule resolving `__host__` vs. `__device__` overloads based on context?  And does it allow overloading based solely on `__host__` vs. `__device__`?
>
> NVCC does not. Clang does. See https://goo.gl/EXnymm for the details.
>
> AFAICT, NVIDIA is starting to consider adopting Clang's approach:
>  http://lists.llvm.org/pipermail/cfe-dev/2018-November/060070.html (original message from Bryce apparently didn't make it to the cfe-dev archive)


Okay.  Probably the template-argument rule ought to be the same as the address-of-function rule, which I assume means that there's a final pass that resolves ambiguities in favor of functions that can be used from the current context, to the extent that that's meaningful.  It's hard to tell because that document does not appear to include a formal specification.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D56411/new/

https://reviews.llvm.org/D56411