Icohedron wrote: In my opinion, we should just define `lit` without templates, thereby following Clang's overload resolution rules and allowing implicit vector truncation. Overload resolution and implicit vector truncation are behaviors that should be consistent across the language. To me, it feels wrong to have special-case behavior on language-defining features implemented/overridden on a per-function basis.
Implicit vector truncation is also an area that needs addressing as a problem with the language, rather than being a special-case behavior handled by each function's implementation. Again, a compiler warning would go a long way to addressing problems where implicit truncation may not be intended by the programmer. ``` warning: implicit truncation of vector type [-Wconversion] ``` Or maybe implicit truncation should be removed from the language entirely (for 202y perhaps)? If you can make the behavior consistent with DXC under all cases with some special implementation in hlsl_compat_overloads for 202x, then you can do that, but I don't think we should be writing an implementation that matches DXC now and then rewrite it entirely again when we swap to 202y that makes the language more consistent (which is one of our goals for HLSL's future, right?) https://github.com/llvm/llvm-project/pull/134171 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
