[PATCH] D18110: [OpenMP] Fix SEMA bug in the capture of global variables in template functions.

[Samuel Antao via cfe-commits]

sfantao added a comment.

I just wanted to add to what Carlo just said, that the feature to capture by value is already used in the offloading upstreamed code. This is not a new feature we are proposing , it is already there used in the code and tested in relevant regression tests. This patch is just about fixing a bug that we identified related to that.

Changing back to capture everything by reference will require reverting/refactor code that is already upstreamed and functional. On top of that, as per the spec and also because of the aforementioned performance issues, we would have to go to capture things by value anyway. So, it looks to me that what you are suggesting is to move backwards, so I'd rather discuss what are the specific concerns with this patch or the by-value captures, so that we can properly address them instead of just delaying the discussion and have to move backwards in the meantime.

Thanks!
Samuel


http://reviews.llvm.org/D18110