home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!europa.asd.contel.com!emory!swrinde!zaphod.mps.ohio-state.edu!uwm.edu!spool.mu.edu!sgiblab!munnari.oz.au!comp.vuw.ac.nz!waikato.ac.nz!aukuni.ac.nz!cs18.cs.aukuni.ac.nz!jeremy
- Newsgroups: comp.theory
- Subject: Re: Synthesizing loop invariants -- unsolvable?
- Message-ID: <1992Nov6.053549.1408@cs.aukuni.ac.nz>
- From: jeremy@cs.aukuni.ac.nz (Jeremy Gibbons)
- Date: Fri, 6 Nov 1992 05:35:49 GMT
- References: <1992Nov5.230648.23466@babbage.ece.uc.edu>
- Organization: Computer Science Dept. University of Auckland
- Cc: jeremy@cs.aukuni.ac.nz
- Lines: 28
-
- In <1992Nov5.230648.23466@babbage.ece.uc.edu> dsims@don.ece.uc.edu (David Sims) writes:
-
- >I've been thrashing about with the problem of automatically generating
- >loop invariants for the purposes of program verification. Is the general
- >case of automatically synthesizing a loop invariant for any given loop
- >unsolvable?
- >
- >I suspect it is but cannot come up with a proof. Any help would be
- >appreciated.
-
- Of course it's solvable; "true" is an invariant of every loop. :-)
-
- Seriously, the halting problem reduces to finding a *variant*, so you can't
- write a program to find variants. Presumably the question you meant to ask
- was, is the problem of finding an invariant sufficiently strong to
- establish a given postcondition, assuming termination, computable? (This is
- equivalent to checking partial correctness of while-programs, right?) My
- intuition is also that it is uncomputable, but I too can't come up with a
- proof off the top of my head.
-
- Jeremy
-
- ---
- Jeremy Gibbons <jeremy@cs.aukuni.ac.nz> tel: +64 9 373 7599
- Department of Computer Science, fax: +64 9 373 7453
- University of Auckland, Private Bag 92019, Auckland, New Zealand.
-
-
-