Re: Separating algorithms from implementations (long)

Dara Gallagher <dara_gallagher@my-deja.com> wrote:
>Also, I remember hearing (from at least 10 years ago) claims to the
>effect that the high-level and abstract nature of modern functional
>programming languages (i.e. after ML) would allow compilers to
>perform highly extensive optimization which would be impossible
>with imperative languages. Unfortunately these "wonder compilers"
>still haven't appeared.


True, but functional languages are still plenty fast. At the most
recent contest, the problem was to write a ray tracer, which you would
think C programs would be good at. But the fastest ray tracer (which
was 3x as fast as #2 and 9x as fast as #3) was written in the
Objective Caml dialect of ML. None of the C or C++ entries (9 of 39)
finished in the top 4 for speed. So there are pretty good ML
compilers out there.




Norman

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