tra added inline comments.
================
Comment at: clang/lib/CodeGen/CGBuiltin.cpp:18912
+ case NVPTX::BI__nvvm_ldu_h2: {
+ return MakeHalfType(BuiltinID, E, *this);
+ }
----------------
We seem to be checking builtin IDs twice. Once here and then in MakeHalfType
where we need to map them to intrinsics.
I think the approach used for atomics above would work better here, too -- just
pass appropriate Intrinsic as a parameter to MakeHalfType and remove the switch
there.
Another option you may consider is that if the availability is only dependent
on PTX or SM version, then we may get by with declaring the builtins as
TARGET_BUILTIN with appropriate constraints and let the standard enforcement
machinery to handle diags.
Repository:
rG LLVM Github Monorepo
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D146715/new/
https://reviews.llvm.org/D146715
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
