Re: Enabling SSA for alpha?

[Brad Lucier <lucier at math dot purdue dot edu>]

> No it means that there is no way for the *processor* to implement the 
> SSA form.  Read a little about SSA, particularly the PHI expressions that
> must be inserted at dominance frontiers.  There's no processor in the
> world that I'm aware of that implements the semantics of a PHI expression.
> Nor do any of GCC's optimizers know about PHI expressions.

OK, I've read Muchnick's short section on transforming to SSA form; he
doesn't give an algorithm for transforming *from* SSA form, and this
might be the critical part I don't understand---if you go from gcc's
intermediate representation to SSA and back again, do you lose the property
that no destination registers overlap source registers?  It would seem
that the algorithm to go from SSA to gcc's intermediate form is crucial here.

Brad