Hardware STG?

[Simon Peyton-Jones]

| advantages.  The cost of course is always an issue, but my=20
| point is that the=20
| difference in performance may in many cases be insignificant.
|=20
| I see a definite place for this hardware, if it can be engineered=20
| successfully.  For example, I could build a firewall that is provably=20
| inpenetrable.  I could build embedded medical devices that I=20
| can prove will not fail mysteriously.

Yes, you are right.  One possible goal is performance, but another, as
you point out, is reliability and correctness.  The issues are quite
different and I'm sure Keean will want to be clear which objective he is
shooting for.

If you could really build a device where you had a proof of correctness
that stretched from the original source code right through to hardware
execution, that would be a real tour de force.=20

Simon