[racket-dev] expected timing difference between fft-complex-forward and fft-complex-radix2-forward?

Yes it has. The main thing is that it has accelerators for 6 and 7 as
factors - the old accelerators stopped at 5. 8192 doesn't have either of
those factors (since it is a power of 2). But, there may have been bug fixes
that I don't remember. I will zip up the latest code and send them to you
this afternoon.

On Tue, Oct 19, 2010 at 9:15 AM, John Clements <clements at brinckerhoff.org>wrote:

>
> On Oct 19, 2010, at 5:17 AM, Doug Williams wrote:
>
> > On my laptop:
> >
> > #lang racket
> >
> > (require "fft.rkt")
> >
> > (define v (build-vector 16384 (lambda (i) (random))))
> >
> > (define v1 (vector-copy v))
> >
> > (collect-garbage)
> > (collect-garbage)
> > (collect-garbage)
> > (time (fft-complex-radix2-forward v1))
> >
> > (define v2 (vector-copy v))
> >
> > (collect-garbage)
> > (collect-garbage)
> > (collect-garbage)
> > (time (fft-complex-forward v2))
> >
> > (for/and ([i (in-range (vector-length v))])
> >  (< (magnitude (- (vector-ref v1 i) (vector-ref v2 i))) 1e-4))
> >
> > ==>
> >
> > cpu time: 94 real time: 94 gc time: 0
> > cpu time: 79 real time: 78 gc time: 0
> > #t
> >
> > I'll look at the rsound/fft implementation and see if I see the problem.
>
> It certainly sounds as though I broke it, or it changed after I got it.
>  Has the source changed since you sent it to me last year?
>
> John
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.racket-lang.org/dev/archive/attachments/20101019/52faafeb/attachment.html>