Related articles |
---|
Re: Intrinsically fast/slow languages diamond@tkou02.enet.dec.com (diamond@tkovoa) (1990-06-22) |
Newsgroups: | comp.compilers |
From: | "diamond@tkovoa" <diamond@tkou02.enet.dec.com> |
Date: | Fri, 22 Jun 90 17:27:02 GMT |
Organization: | Compilers Central |
Keywords: | design, optimize |
Bogus: | pardo@june.cs.washington.edu (David Keppel) writes: |
Bogus: | >Reducto ad perfect optimizer: no intrinsically inefficient languages. |
Bogus: | No. |
Bogus: | >Reducto ad current optimizers: yes intrinsically inefficient languages. |
Bogus: | Yes. |
Bogus: | >Reducto ad carefully-designed languages with good optimizers: |
Bogus: | >sophisticated features, and if you write ``static code'' programs you |
Bogus: | >get ``static code'' performance. |
Bogus: | Yes. |
Bogus: | Some languages will always be intrinsically inefficient. (They might have |
Bogus: | to be "carefully" designed, or carefully edited by examination of the |
Bogus: | typing of a few zillion monkeys. In the latter case, "most" of the |
Bogus: | languages designed by the monkeys will be inherently inefficient, and |
Bogus: | "most" of the time spent editing will be to discard things that aren't |
Bogus: | languages, rather than checking whether a language is intrinsically |
Bogus: | inefficient.) |
Bogus: | For example, suppose a language is supposed to contain representations of |
Bogus: | all regular languages. Our language is a meta-language. Now, if we |
Bogus: | design our meta-language to accept type-3 grammars and determine whether |
Bogus: | the input represents a regular language, then we have a task which can |
Bogus: | (optionally) be met with trivial efficiency. If we design our |
Bogus: | meta-language to accept type-0 grammars and determine whether the input |
Bogus: | represents a regular language, then we have an inherently very |
Bogus: | inefficient, indeed impossible, task. Some meta-languages are |
Bogus: | intermediate in inherent inefficiency. |
Bogus: | -- |
Bogus: | Norman Diamond, Nihon DEC diamond@tkou02.enet.dec.com |
Bogus: | This is me speaking. If you want to hear the company speak, you need DECtalk. |
Bogus: | -- |
Bogus: | Send compilers articles to compilers@esegue.segue.boston.ma.us |
Bogus: | {spdcc | ima | lotus| world}!esegue. Meta-mail to compilers-request@esegue. |
Return to the
comp.compilers page.
Search the
comp.compilers archives again.