Re: Efficient bytecode design and interpretation

Michal Gajda <mg169780@zodiac.mimuw.edu.pl> writes:
> [For example - why GForth VM, with all vmgen'ed optimizations, performs so
> badly at "Programming Language Shootout" page? Note, that I chosen 0
> weights for all test that were _not_ implemented for Forth, and regarded
> only bytecode implementations-of which OCaml clearly shines]


I see Anton has already replied, but I kind of feel like my shootout
site is being misleading here, so I will make a quick comment.


Not all Gforth solutions have been optimized on my site yet. In
particular the solutions where other languages use optimized hash
tables, Gforth uses wordlists, and would probably do better if
rewritten with a custom hash table implementation.


I find that Gforth is astonishingly fast at other tests, like the
Sieve.


My shootout page is still quite flawed (as all benchmarks are :-) I
advise all my readers to please read the caveats on the methodology
page, and be very careful about trying to make generalizations.


Cheers,
Doug

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