[cfe-dev] [CUDA] Why "Disallow 'extern __shared__' variables"?

[Arpith C Jacob via cfe-dev]

Hi,

I was wondering why Clang-cuda now disallows adding 'extern' on __shared__
variables with this patch:
https://reviews.llvm.org/D25125?id=73133

I was using it successfully, and I believe correctly, in both nvcc and
clang-cuda to build multiple relocatable objects with extern shared memory
references that linked to these globals with nvlink.  Please let me know if
you would like specific examples and I'll be happy to describe it.

Thanks,
Arpith
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-dev/attachments/20161028/40dc8c36/attachment.html>