sihuan wrote:

> Why are the built in’s using a different naming than the spec? The spec has 
> i32x2 but the builtins use v2i32?
> 
> I hope we can use operator+ and operator- on vector types so we don’t need 
> any builtins for padd/psub.

I apologize for the confusion caused by the naming. The use of `v2i32` was a 
result of my over-complicated design of mapping spec names to internal 
builtins. I will switch to using `operator+` and `operator-` for `padd` and 
`psub`, and remove these builtins.


> This spec is one of two proposals I’m aware of for intrinsics. The mailing 
> list has not made any decision about which will move forward.

Thank you for the heads-up. I’ve been following the mailing list and am aware 
of the ongoing discussions regarding the two proposals. My original intention  
was just to set up the initial path for P-ext intrinsics, but now  I realize 
that `padd`/`psub` was not the best choice for this purpose.
I’ve switched this PR to a Draft for now and will continue to monitor the 
mailing list.

https://github.com/llvm/llvm-project/pull/174068
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Reply via email to