DanielCChen wrote:

> > +1 to the engineering but I fundamentally disagree with having 
> > `-finit-local=` cover all types at once. Assigning some arbitrary value to 
> > integer and logical when `=snan` is specified doesn't make sense to me. I 
> > think starting with gfortran's options is the way to go - and if you want 
> > to add the ability to specify a hex value in addition, even better.
> 
> Thanks for the quick review!
> 
> I have no objection to start with only `zero` and `0x<hex>` to begin with and 
> leave `nan` and `snan` out as those two (`zero` and `hex`) are the most 
> common usage. The Gfortran's `-finit-local-zero` is already implemented in 
> this PR as an aliasing to `-finit-local=zero`. I think we should keep this 
> option format so it is easier to expand in the future.

@rscottmanley 
I generated a Type mapping table for `-finit-local=` to show how each type is 
implemented so that we can decide if we want to drop the support for `nan` and 
`snan` for this PR.

https://github.com/llvm/llvm-project/pull/216164
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Reply via email to