Re: performance-oriented languages?

"John Max Skaller" <>
21 Oct 2004 22:27:21 -0400

          From comp.compilers

Related articles
[3 earlier articles]
Re: performance-oriented languages? (2004-10-04)
Re: performance-oriented languages? (Gioele Barabucci) (2004-10-04)
Re: performance-oriented languages? (Tony Finch) (2004-10-09)
Re: performance-oriented languages? (Philipp Klaus Krause) (2004-10-09)
Re: performance-oriented languages? (Stefano Lanzavecchia) (2004-10-12)
Re: performance-oriented languages? (2004-10-17)
Re: performance-oriented languages? (John Max Skaller) (2004-10-21)
Re: performance-oriented languages? (2004-10-21)
Re: performance-oriented languages? (Daniel Berlin) (2004-10-23)
Re: performance-oriented languages? (glen herrmannsfeldt) (2004-10-24)
| List of all articles for this month |

From: "John Max Skaller" <>
Newsgroups: comp.compilers
Date: 21 Oct 2004 22:27:21 -0400
Organization: Compilers Central
References: 04-10-015
Keywords: performance, practice
Posted-Date: 21 Oct 2004 22:27:21 EDT

On Sat, 02 Oct 2004 01:19:37 -0400, Brandon J. Van Every wrote:

> Can anyone comment on what languages or compilers have 'performance'
> as the primary goal of the project?
> An example might be Mlton, the whole program SML optimizing compiler.
> I think Intel C++ would count. VC++ wouldn't,
> it's oriented towards the IDE for .NET stuff, not best possible
> performance. I don't think g++ would count, as given its results in
> recent years it doesn't seem to be a performance-oriented project.

Actually, these two negatives aren't so. Microsoft takes performance
of the generated code more seriously than most. They have managers
for every processor whose job is to get the best possible performance
from the generated code for that processor (I know that because I was
interviewed for such a job).

Similarly, g++ has a lot of people working on optimisations, there are
a *lot* of available optimisations ;)

So don't think it is fair in either case to say these projects don't
have a performance orientation. Perhaps it is the case that the goals
haven't been met. Most language processors have to place Standards
compliance and correctness first, and most have a second goal of
portability to meet as well as performance.

As far as I can tell, g++ at least emits fairly reasonable code out of
the back end. It has much more problem with higher level
optimisations. That problem will never go away -- you can't write a
good compiler for an extremely complex and nasty language in an
language like C, so grossly unsuited to writing compilers.

OTOH when you have a high quality compiler for a nice high level
language -- you have the capability to write a good compiler even for
a nasty language like C++ -- but somehow the incentive to do so is
gone :)

Post a followup to this message

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