3 ms·
Their process was extremely elaborate. They wrote a Haskell prototype which was then automatically translated to their specification/verification language (Isab
by pron 11y ago
Their process was extremely elaborate. They wrote a Haskell prototype which was then automatically translated to their specification/verification language (Isabelle/HOL). They then proved all the theorems they wanted (which, I can only guess, was very hard), and hand-translated the spec into C. Then they proved that the C translation was a refinement of the spec (i.e. doesn't deviate from its boundaries) using a theorem prover and an SMT solver.
That last part (translation validation[1]) is what I found most interesting, or, at least, the most novel. Here's the paper: http://www.cl.cam.ac.uk/~mom22/pldi13.pdf http://www.cl.cam.ac.uk/~mom22/pldi13.pdf
[1]: http://download.springer.com/static/pdf/209/chp%253A10.1007%252FBFb0054170.pdf?originUrl=http%3A%2F%2Flink.springer.com%2Fchapter%2F10.1007%2FBFb0054170&token2=exp=1442599845~acl=%2Fstatic%2Fpdf%2F209%2Fchp%25253A10.1007%25252FBFb0054170.pdf%3ForiginUrl%3Dhttp%253A%252F%252Flink.springer.com%252Fchapter%252F10.1007%252FBFb0054170*~hmac=4077ecacf62fe551914032ebe74604b7dbfbede85014bb5127269be688562d29 http://download.springer.com/static/pdf/209/chp%253A10.1007%...