jmmartinez wrote:

Thanks a lot for the comments!

In the last version of this patch I've addressed several mistakes that I'd made:
* Originally I put the fix in the `_cuda_` version of the header instead of its 
`_hip_` counterpart. In the last version I fixed this.
* I reduced the scope of the builtins to cover only these 6 builtins: 
isunordered, isgreater, isgreaterequal, isless, islessequal, islessgreater. It 
seems that only these got the `constexpr` while the other builtins use the 
`_CONSTEXPR23` macro. I'm going to create an issue on the STL to ask for 
clarification seems the unconditional constexpr on these builtins seems to go 
against the standard (we still have to address the `__cpp_lib_constexpr_cmath`).
* Used `__has_include(<version>)`
* Added a test

https://github.com/llvm/llvm-project/pull/200395
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Reply via email to