uweigand wrote:

> Could we not use the same approach that the existing `LOAD_STACK_GUARD` uses 
> in order to avoid the canary being spilled? I.e. could we not mark 
> `LOAD_STACK_GUARD_ADDRESS` `isReMaterializable`, and do some further nudging 
> with kill flags to ensure that the value isn't reused?

If that works reliably, I guess I'd be fine with it.  We should have test cases 
to demonstrate that it is isn't spilled, however.

https://github.com/llvm/llvm-project/pull/169317
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Reply via email to