================
@@ -225,7 +225,9 @@ ABIArgInfo AMDGPUABIInfo::classifyArgumentType(QualType Ty,
bool Variadic,
// Records with non-trivial destructors/copy-constructors should not be
// passed by value.
if (auto RAA = getRecordArgABI(Ty, getCXXABI()))
- return getNaturalAlignIndirect(Ty, RAA == CGCXXABI::RAA_DirectInMemory);
+ return getNaturalAlignIndirect(
+ Ty, getContext().getTargetAddressSpace(LangAS::Default),
+ RAA == CGCXXABI::RAA_DirectInMemory);
----------------
AlexVlx wrote:
Apologies if my reply came off as defensive (and for the delay), it wasn't
intended as such, merely wanted to convey trepidation around doing too much in
one change. I am in agreement with the points you made / the way forward. In
what regards the proposed rule, as I mentioned elsewhere, I'm of the opinion
that we should leave indirect returns pointing to the AllocaAS (since this is
the only possibility at the moment), with a note that this would require
reworking in the future (where the base ABI would be to return a pointer to the
default AS, with targets potentially altering this).
https://github.com/llvm/llvm-project/pull/114062
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
