[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

64 bit and precision? - was Re: [sc-users] DiskIn loop trig ?



Hi Scott,

On 11/29/08, Scott Wilson <s.d.wilson.1@xxxxxxxxxx> wrote:
> A while back someone did some tests of whether a full 64 bit audio chain
> would really be worth it. Those involved please chime in with details, but
> IIRC the conclusion was that for the most part audible improvements were
> only seen in cases with feedback, which makes sense. Internally, there's
> already a fair number of doubles used in UGens, where appropriate.
>
>  S.
>
>
>  On 29 Nov 2008, at 11:18, blackrain wrote:
>
>
> >
> > On 11/26/08, stefan kersten <sk@xxxxxxxxxx> wrote:
> >
> > > Julian Rohrhuber wrote:
> > >
> > > > I think it would be fine as a first step to have the server run on 64
> bit.
> > > > things like this:
> > > >
> > > > { x = Sweep.ar; sin(x * 2pi * 300) * 0.1 }.play;
> > > >
> > > > would play correctly.
> > > >
> > >
> > >
> > > to reiterate: there are two unrelated issues, a 64 bit address space vs
> > > the use of double precision for accumulators, coefficients or even audio
> > > samples.
> > >
> > >
> > > <sk>
> > >
> >
> > stefan, I'm still puzzled.
> > I do undestand the difference regards today;s implementation to what a
> > 64 bit address space implies.
> > the mappings of a 64 bit float internal resolution vs 32 whatever we
> > deal with now, I dont follow tho.
> > I cant see why the approach would have to be difft. All I see is loads
> > of work porting todays UGens code to comply - Is that what you mean?
> >
> > and yeah indeed 10.4+ deals w/o a chroot with a 64 bit attempt as you
> > are saying.
> >
> > Julian, it is kinda obvious (at least as I can see) why the UGens wont do.
> > perhaps a full 64 bit branch is really in need?
> >
> >
> > cheers,
> >
> > x
> >
> >

I dont quite agree.
I do undestand the overhead issues.
I know the lang is not made to deal with this.
I can understand to a point how wider words wont render an 'instant'
change of perception to most ppl and can see why changes to sample
rate will hands down beat a fine grain resolution (again, an instant
appreciation); Yet, it is the sum of some blocks of calcs of these
words that makes things seem (sound? =) fluid or gainy.
smoother cycles, a bit less jumpy gens.

ps, flipping this to a new subject now think we should have a bit ago.

cheers,

x

_______________________________________________
sc-users mailing list

info (subscription, etc.): http://www.beast.bham.ac.uk/research/sc_mailing_lists.shtml
archive: https://listarc.bham.ac.uk/marchives/sc-users/
search: https://listarc.bham.ac.uk/lists/sc-users/search/