Hi all,
sorry I don't have time to update the 'mikecheng' version of the
8hz code. I'm full-on busy with teaching prac work.
I hope someone is taking all these modifications that people
mention and putting them into the 'mikecheng' version. [not that it's my
version any longer. but sort of like egcs, I'd like a progressive and
fast updated version happening.]
> The atan2() in fft.c computes the phase of each sample of the fft. The only place
> I re-formulated caluculations of the form cos(2*atan2(A,B)-atan2(C,D)) as functions
> of A, B, C, and D, which are just the real and imaginary parts of one bin of the
> DFT.
> Anyway, it seems rather complicated to explain now, but it does give a 4% speed
> improvement over version "2b-m11" of the code I downloaded. Maybe someone can take a
Interesting! I'll have to check this out when i get time. (not until the
school holidays: september 18th!)
> > >I've noticed that the sources still have <yuck> unnecessary trig
> > >functions </yuck> in fft.c and l3psy.c. They can be removed from the 8hz
> > >02b sources, and gain about 10% speed without trying hard. My question
this was someone else's post: Check out http://www.cryogen.com/mikecheng
for my modified sources which are heaps faster than the 8hz-2b sources,
and do away with heaps and heaps of the trig.
Any word from the 8hz people?
later
mike
