Re: Why I've Dropped In

On 6/18/2025 2:31 AM, Anton Ertl wrote:

Stefan Monnier <monnier@iro.umontreal.ca> writes:

Ah, well, pairing is a different problem than the "incomplete register
specifiers" I'm talking about.  Indeed, it can be much more difficult to
adapt a Chaitin-style allocator to handle pairing because it can't be
expressed simply in the interference graph.

 I remember reading a paper about register allocation for register
pairs, but don't find that paper right now.  Anyway, what I read was
based on graph-colouring IIRC and it looked pretty plausible and not
too complicated.  But the devil is in the details.
 

IME, register allocation for pairs isn't much harder than for single registers.
Just in the pair case, both N+0 and N+1 need to be available, with generally N+0 being assigned the value, and N+1 a placeholder (signaling that the spot is taken up by a 2-part value in the adjacent spot).
Kinda, well in BGBCC it is more of a mess (decided not to go into how much of a trash fire BGBCC's register allocator is).

- anton