Re: Writing fast compilers... RS/6000

[From comp.arch -John]


In article <1991Aug15.010352.1@watt.ccs.tuns.ca>,
macdonaldk@watt.ccs.tuns.ca writes:
|> An overstatement, I think. Optimizing compilers are not going to
|> produce as much improvement on well written code as they will on student
|> first attempts, because the common subexpressions have be done in
|> source, the invariant stuff is out of loops, the loops are unrolled, the
|> register type has been used here and there... all by a one pass code
|> generator called a programmer.


Well, it depends on what you mean by ``well written code''. I think there are
very good arguments that say that good code contains common subexpressions,
contains rolled loops, etc. The code is clearer, earier to maintain etc. Also,
a compiler can often do a better job if the code is in this form than a
programmer can do by putting it into a more obscure form. Indeed, this is so
much the case that at least one book on ``How to write programs in FORTRAN for
supercomputers'' explicitly says:-


                                don't eliminate common subexpressions


(and one or two other surprising things that I forget now).


I would rather have a compiler which let well-written (:-) straightforword
code run very fast than have to produce obscure (and often machine-dependant)
code in order to achieve the same performance.


Roger Shepherd, INMOS Ltd JANET: roger@uk.co.inmos
1000 Aztec West UUCP: ukc!inmos!roger or uunet!inmos-c!roger
Almondsbury INTERNET: roger@inmos.com
+44 454 616616 ROW: roger@inmos.com OR roger@inmos.co.uk
--

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