Abstract
The effects of optimisation on debugging were characterised, first as metric-based distances along paths, and then within an algebraic framework [6, 7, 8]. In them, we observed algorithmic ambiguities which may seriously compromise the efficiency-though not the semantic definition-of a debugger for optimised programs. Informal observations appear to indicate enhanced performance if displaced codes were favoured for further displacements, and unmoved codes were preferentially left alone. This reluctance to change the "state of code motion" seems much like the physical quantity known as inertia. This paper highlights points to support the "inertial argument", and raises questions to be explored in relation to it.
Publisher
Association for Computing Machinery (ACM)