Yes, through the “noalias” attribute. Rustc uses it, well, when it’s not buggy at least. It’s currently turned off pending an upstream fix; we’ll have it back on when it’s fixed.

Is that the last thing blocking "fortran-parity" in rust? Is there a working group hpc would fall under?

Last I checked there was no fast-math on stable, and #[repr(align)] was recently added.

I don't know enough about fortran to say, to be honest.

> there was no fast-math on stable

We don't do global flags to tweak things like this, as a rule. You can do it on indivudal operators, or use wrapper types, to get these behaviors.

Oh, and there’s probably just be a new working group for hpc; we’ll have a process for spinning up groups early next year.