[MLton-devel] cvs commit: backend and x86-codegen cleanup

[Matthew Fluet]

> >   (Note, the x86-codegen
> >   still needs to know that the offset is from the gcState in order to
> >   correctly assign those memory locations to the GCState class
>
> This is for aliasing purposes only, right?

Correct.

> I also wanted to get rid of Machine.Operand.GCState, instead using
> Machine.Operand.Label (Label.fromString gcState).
>
> Would it be possible to handle that, creating a separate class for
> offsets from labels?  You can then at least assume that different
> offsets from the same label don't alias and that offsets from a label
> don't alias the heap or the stack.

That would be fine.  The mayAlias function right now assumes that if
two memory locations with different immediate expressions (e.g., a label)
for the base, then the two locations don't alias.  I.e., everything in
offsets from GCState is guaranteed not to alias anything in offsets from
globalPointerNonRoot.



-------------------------------------------------------
Enterprise Linux Forum Conference & Expo, June 4-6, 2003, Santa Clara
The only event dedicated to issues related to Linux enterprise solutions
www.enterpriselinuxforum.com

_______________________________________________
MLton-devel mailing list
MLton-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/mlton-devel