Type Inference & SVA

[JTrenerry]

• SVA
  • Intra pass
  • IDE pass
• Type inferencing
  • Constraint generation (Relies on SVA)
  • Type coalescing
  • Type Automata
  • Type lowering
  • Transform itself

(run-transforms "ssa")
(run-transforms "cf-expressions")
(run-transforms "type-check")
(run-transforms "type-inference")

Note: SVA does not net very good results in most cases

[JTrenerry]

Is testing analysis just by doing the value soundness enough? The lattices I am using should be tested already I believe as it is just wrapped and map lattice from Hayden and bpal.

[JTrenerry]

I think it would be best if this is merged once fully complete as it would allow for more complete testing.

If SVA is needed elsewhere I can cherry pick those commits, but I also would need some better testing for SVA as well.

[JTrenerry]

Globals are not done in SVA, will probably have to confirm with Nick but he did not want me to just say things are globals and then the offsets goes out of that global area and now I have to widen.

I believe what he wanted (although I do not remember), after I get all the offsets I can then check to see if it lines up maybe, but even that seems iffy, I remember him saying, keep it is a constant unless I need it to be a global, but like when would I ever need it to be global?