paperchalice wrote:

> I don't remember the example off-hand, but there was a fast-math transform I 
> wanted at one point that needed a flag on `fptoui`.
> 
> > I don't think we should do this until we have freed up the necessary flag 
> > space to also support uitofp as well.
> 
> I can see an argument for doing `fptoui` and to `fptosi` without the 
> inverses, but agreed that `sitofp` without `uitofp` is a bad idea.

PowerPC also need `sitofp/uitofp` with `afn` to allow double rounding.

https://github.com/llvm/llvm-project/pull/160475
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Reply via email to