Re: [CM] Performance of S7

[Massimiliano Gubinelli]

Dear Bil, On 31. Dec 2020, at 15:13, b...@ccrma.stanford.edu <b...@ccrma.stanford.edu> wrote: I think the problem here is with-let -- it is blocking the normal optimizations that would otherwise speed up fib.  Here is an ugly substitute: (let ((temp-value #f))    (define *texmacs-module* (curlet))    (define-macro (tm-define head . body)      `(eval (list 'define ',head ',@body) *texmacs-module*))    (let ()        (tm-define fib (lambda (n) (if (< n 2) n (+ (fib (- n 1)) (fib (- n 2)))))))    (let ((n 40))        (format *stdout* "(fib ~A) : ~A \n" n (fib n)))) There has to be a prettier way to do it! with-let should not be so paranoid...  thanks, indeed it works for the small example. However the *real thing* still run as slow as before even with this replacement. I've extracted the "real" tm-define and I attach a test case which indeed still show the slowness even using the "eval" trick. I hope there is something to do about it. Even a trick would do, this is the only point in all the sources where I do some trickery with lets. Max

fib4.scm
Description: Binary data

_______________________________________________
Cmdist mailing list
Cmdist@ccrma.stanford.edu
https://cm-mail.stanford.edu/mailman/listinfo/cmdist