• [^] # Re: pourquoi le lip

    Posté par . En réponse à la dépêche Sortie de SBCL 1.0. Évalué à 2.

    Je rajoute un lien sur le sujet :
    http://bc.tech.coop/blog/040308.html

    Le code vainqueur d'un benchmark n'utilise pas le lisp comme un lisper l'utilise.


    > So what have we learned? We confirmed what we pretty much knew: you
    > can write a C program in CL, at which point the relative speed of your
    > C and CL versions will depend on the relative quality of the code
    > generation. It's worthwhile to know the specifics of your
    > implementation (for instance, it seems that a substantial amount of
    > the last-mile speed up for this benchmark on CMUCL came from figuring
    > out ftruncate could be fast inside a macro, but not inside an inline
    > function).
    > What would be even more interesting to me would be an example more
    > like what I feel I experience anecdotally --- that I write programs
    > that are a bit slower than C (maybe 25% to 50%), but they're much more
    > flexible, more abstract, cleaner, easily modifiable, and so on. I
    > don't really feel like the CL version of almabench we currently have
    > shows any of the benefits of CL.