================
@@ -128,12 +128,13 @@ enum class CudaArch {
GFX12_GENERIC,
GFX1200,
GFX1201,
+ AMDGCNSPIRV,
Generic, // A processor model named 'generic' if the target backend defines a
// public one.
LAST,
CudaDefault = CudaArch::SM_52,
- HIPDefault = CudaArch::GFX906,
+ HIPDefault = CudaArch::AMDGCNSPIRV,
----------------
AlexVlx wrote:
In the long run it should not (the BE is almost baked, for example); in the
interim, it will need the translator, and I (think I) do see what you are
saying, that it might be onerous to have HIPAMD depend on that - I can unflip
the default pending the BE coming along. The only cost would be slightly more
dubious ergonomics, but those are pre-existing.
https://github.com/llvm/llvm-project/pull/95061
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
