Re: Where would you like to spend that resource? (WAS: yup!)

pardo@gar.cs.washington.edu (David Keppel)
Thu, 22 Aug 91 18:05:24 GMT

          From comp.compilers

Related articles
yup! wulf@capa.cs.Virginia.EDU (1991-08-18)
Where would you like to spend that resource? (WAS: yup!) pardo@cs.washington.edu (1991-08-19)
Re: Where would you like to spend that resource? (WAS: yup!) clc5q@hemlock.cs.Virginia.EDU (1991-08-21)
Re: Where would you like to spend that resource? (WAS: yup!) moss@cs.umass.edu (1991-08-22)
Re: Where would you like to spend that resource? (WAS: yup!) pardo@gar.cs.washington.edu (1991-08-22)
Re: Where would you like to spend that resource? (WAS: yup!) meissner@osf.org (1991-08-31)
| List of all articles for this month |

Newsgroups: comp.compilers
From: pardo@gar.cs.washington.edu (David Keppel)
Keywords: optimize, design
Organization: Computer Science & Engineering, U. of Washington, Seattle
References: 91-08-082 91-08-092 91-08-107
Date: Thu, 22 Aug 91 18:05:24 GMT

Hm, perhaps I should give this speech from the protected area *underneath*
my soap box. Excuse me while I <clank> don <clank> my <bonk!> armor.


clc5q@hemlock.cs.Virginia.EDU writes:
>[The original discussion *was* about optimization;
> I got the impression you were worried about time spent optimizing
> during the software development cycle;
> doing array bounds checking is not redundant work, nobody said they
> were;
> this *is* `comp.compilers'.]


Yes, the original discussion was about optimization. When somebody said the
state of the practice for optimizers was bad, I stepped in and made the
opinionated statement that the quest for speed was (IMHO) overstated.


I am not worried about cycles spent running the optimizer during development;
I didn't mean to give that impression. And yes, a good optmizer can remove
redundant array bounds checks (see e.g., the PL.801 work which claimed only a
few percent overhead for optimized full array bounds checks).


I'm saying that while a lot of time is spent on optimizers (and optimizers are
a good thing) it is my contention that the performance of many programs is
less important than their correctness. I would rather have more work
improving that state of the practice than improving the optimizer state of the
practice.


Bill Wulf says that correctness-vs-performance is a favorite academic debate,
but in commercial systems they occur together or not at all; neither
performance nor correctness are ``add-on''. Optimizing compilers, he says,
are a tool that let designers and builders focus on larger issues of algorithm
and data structure. Right on.


I think perhaps my original posting was misplaced (or at least the
`Followup-To:' line). I don't want to roll back compiler technology, nor am I
arguing that optimizers are developing along the wrong lines. I am making
project goal, programming style and methodology, and language design
arguments. So rather than explain myself further I think I'll fall off.


Whew, I think this soapbox is flame-retardent!


;-D oN ( The goal of a new machine ) Pardo
--


Post a followup to this message

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