[bitc-dev] Rationale for unboxed mutability in BitC?

[Sam Mason]

On Fri, Sep 15, 2006 at 10:03:05PM +0100, David Hopwood wrote:
> What I was proposing would only depend on one specific optimization: ref cells
> in specifically marked structure and array types must be inlined. Overall, the
> constraints on optimization of BitC would be reduced.

I'm probably not the best person to comment, but I think once you do this
then you need all the machinery that is being worked out by Swaroop to
determine if you actually can perform the optimisation safely.


  Sam