Re: And speaking of fast compilers...

pardo@cs.washington.edu (David Keppel)
Mon, 23 Nov 1992 21:49:42 GMT

          From comp.compilers

Related articles
And speaking of fast compilers... pardo@cs.washington.edu (1992-11-12)
Re: And speaking of fast compilers... sasdrf@unx.sas.com (1992-11-16)
Re: And speaking of fast compilers... preston@dawn.cs.rice.edu (1992-11-17)
Re: And speaking of fast compilers... cheryl@gallant.apple.com (1992-11-17)
Re: And speaking of fast compilers... pardo@cs.washington.edu (1992-11-17)
Re: And speaking of fast compilers... pardo@cs.washington.edu (1992-11-23)
Re: And speaking of fast compilers... macrakis@osf.org (1992-11-24)
Re: And speaking of fast compilers... preston@miranda.cs.rice.edu (1992-12-03)
| List of all articles for this month |

Newsgroups: comp.compilers
From: pardo@cs.washington.edu (David Keppel)
Organization: Computer Science & Engineering, U. of Washington, Seattle
Date: Mon, 23 Nov 1992 21:49:42 GMT
References: 92-11-057 92-11-096
Keywords: performance, design, testing, comment

A while back I wrote about how the Brand X compiler took 6 minutes to
compile a 7,000-line program made up of one-page functions. And how Brand
Y took 40 minutes; over 1.5 hours with optimization turned on.


Well, according to my conversational friend, the latest Brand Y is down to
10 minutes and 25 minutes with optimization on. He's also had a chance to
test it on the latest and greatest from some other vendors and he found
(among other things):


    Brand I: 10 minutes with optimization off, fails with optimiation on.
    Brand J: Crashed the brand J hardware.
    Brand K: Based on the brand I front-end, fails during code generation.


All are validated Ada compilers.


I also heard from Vinod Grover <grover@Eng.Sun.COM> that he knows of an
Ada compiler front end that took 14 hours on a VAX 11/780 to parse and
semanticise a 3-line Ada program. The problem was that the overload
resolution phase of the front end was a two-pass backtracking algorithm
and the 3-line program had 6 levels of overloading of every symbol.




To summarize: language design is tricky and correctness doubly so.


"A program that produces wrong answers twice as fast is infinitely
slower." -- John Ousterhout


;-D on ( Slowing the way to success ) Pardo
[Back when I was writing an F77 compiler, I also found that I could pass
a lot of validation tests with amazingly broken versions of the compiler.
-John]
--


Post a followup to this message

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