Re: rearranging code invalidates liveness info

Fergus Henderson <fjh@munta.cs.mu.OZ.AU> asks:
> anything but the most trivial rearrangements will invalidate the annotations
> on the code. ... How do compiler writers usually handle this problem?


mernst@research.microsoft.com writes:
|> ... At the other end of the spectrum (requiring much
|> more work and potentially saving more effort) is incremental maintenance of
|> semantic information, either directly or by rerunning analyses over a small
|> affected region in which the semantic information has been invalidated.


Warning: making a small change CAN result in the "affected region" being the
whole program. "Optimistic" analyses (such as constant propagation) suffer
this in particular. I can write a program such that making a small edit
(edit is size delta) up front will change the constants for the whole program
(solution change is size N). Worse, I can do this such that the solution
barely changes, but you are required to check the whole program to verify
correctness (i.e., edit is size delta, solution change is size delta, but
runtime to maintain correctness is size N). The incremental-analysis guys
are working on a *hard* problem.


--
cliffc@cs.rice.edu -- Massively Scalar Compiler Group, Rice University
--

Return to the comp.compilers page.
Search the comp.compilers archives again.