pawosm-arm wrote:

> no one bothered to do it and preferred magic builtins

If no one bothered, would it be accepted if someone try extending C++ splat 
init to all kind of vectors?

I presume a magic builtin would be some equivalent of GCC's 
`__builtin_vec_splat` (although in GCC it applies to altivec only). In clang it 
could be available to all kind of vectors. Technically speaking, for immediate 
param values it could be turned into `splat` IR instruction, and for the param 
values known only at the execution time it could be turned into `shullfevector` 
(basically, what the splat init for the `ext_` vector currently does). 
Considering what it takes to introduce a new builtin like that into clang, it 
would have to be supported with strong argumentation and ruling out all other 
options (like C++ splat init extended to all kind of vectors). I have serious 
doubts it will happen.


https://github.com/llvm/llvm-project/pull/205432
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Reply via email to