Newsgroups: | comp.compilers |
From: | larus@primost.cs.wisc.edu (James Larus) |
References: | <1990Jun13.143951.2129@esegue.segue.boston.ma.us> <1990Jun12.163959.2593@esegue.segue.boston.ma.us> |
Date: | Thu, 14 Jun 90 15:29:39 GMT |
Organization: | University of Wisconsin--Madison |
Keywords: | code, optimize |
In article <1990Jun13.143951.2129@esegue.segue.boston.ma.us>,
moss@cs.umass.edu (Eliot Moss) writes:
> I would also like to point out that optimizations techniques are known that
> can eliminate most array bounds checks, if bounds checking is one of the
> safety measures that concerns you.
Until recently, I would have agreed with this statement. However, in
next week's SIGPLAN PLDI Conference, there is a paper entitled "A Fresh
Look at Optimizing Array Bounds Checking," by Rajiv Gupta. It presents
an elaborate dataflow scheme for optimizing array bounds checking by
eliminating redundancies and moving tests out of loops. However, what
caught my eye is the table of results. After all optimizations were
applied, programs with bounds checkings ran 0-46% slower than programs
without bounds checking (down from 78-325% slower). That's a pretty
large performance degredation, considering it also had a large cost in
compiler time and complexity.
/Jim
[How does this compare with IBM's results? The papers I've seen on PL.8
suggest that they didn't think bounds checking was very expensive. -John]
--
Return to the
comp.compilers page.
Search the
comp.compilers archives again.