> Actually, another problem with the implementation of this patch has
> just occurred to me: doesn't it prevent the user from even writing FP
> or NEON instructions even in inline assembly? That strikes me as a
> very bad thing, if so.

I think that should be an error. GCC throws an error if you try to use floating 
point types with -mgeneral-regs-only, so inline assembly wouldn't work anyway.

However I do think that the current implementation of disabling FP/NEON isn't 
great. Ideally we should also error on the use of floating point types, either 
in the front-end or a hard error in the backend. I'm not sure how to achieve 
that in the backend though, suggestions appreciated.

Amara




_______________________________________________
cfe-commits mailing list
[email protected]
http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits

Reply via email to