On Wed, Apr 04, 2007 at 09:27:12PM +0200, Zdenek Dvorak wrote: > 3) making this offset into an index, i.e, having > base: &a, indices: (step:1, value: n) > > Out of these, I like 3) the most, although it might be fairly expensive > memory-wise (any idea how common the variable offsets are?)
Hmm. I suppose that would work, and since one can have as many indicies as needed, one can continue to represent nested variable-displacement component_refs. r~
