================
@@ -37,8 +37,8 @@ static const unsigned SPIRDefIsPrivMap[] = {
0, // cuda_device
0, // cuda_constant
0, // cuda_shared
- // SYCL address space values for this map are dummy
- 0, // sycl_global
+ // Most SYCL address space values for this map are dummy
----------------
AlexVlx wrote:
Yes, not only should it be OCL only, but it probably should only be OCL with no
generic AS support only (please check out the PR I linked where we switched
that over for AMDGPU). PrivateAsDefault was/is a bad hack. Note that you might
want to fork the AS map switch into a different PR, as the fallout might end up
a bit of a slog. E.g., this is still stuck in limbo #113930, and it will bite.
https://github.com/llvm/llvm-project/pull/134399
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
