[PATCH] D68578: [HIP] Fix device stub name

[John McCall via Phabricator via cfe-commits]

rjmccall added a comment.

There are a number of places in IRGen that pass around `GlobalDecl`s with the expectation that that's sufficient to uniquely identify a symbol.  The fact that IRGen breaks down the GD at the last second before passing it to the mangler, rather than passing it to the mangler and letting the mangler decide what to do with it, doesn't really change anything and is arguably poor code design anyway.  Inventing a second declaration, or trying to propagate a flag outside of GD, is just fighting the architecture for no good reason.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D68578/new/

https://reviews.llvm.org/D68578