Hi Oleg,
Sorry for the late answer.
I think this is an excellent suggestion! Since it is a typical "reduce"
operation, I suggest having "reduce" in the name. We can imagine having
three different versions depending if we start the reduction from the top
outputs, the bottom outputs, or in parallel.
import("stdfaust.lib");
// parallel reduce: starts reductions in parallel on both halves
parReduce(op,1) = si.bus(outputs(op));
parReduce(op,2) = op;
parReduce(op,n) = parReduce(op,int(n/2)), parReduce(op,n-int(n/2)) : op;
// top reduce, starts reductions from the top outputs
topReduce(op,1) = si.bus(outputs(op));
topReduce(op,2) = op;
topReduce(op,n) = topReduce(op,n-1), si.bus(inputs(op)-outputs(op)) : op;
// bottom reduce: starts reductions from the bottom outputs
botReduce(op,1) = si.bus(outputs(op));
botReduce(op,2) = op;
botReduce(op,n) = si.bus(inputs(op)-outputs(op)), botReduce(op,n-1) : op;
// examples
pmax(n) = par(i, n, (i,_)) : parReduce(op,n) with {
op(i1,v1,i2,v2) = select2(v2>v1,i1,i2), max(v1,v2);
};
tmax(n) = par(i, n, (i,_)) : topReduce(op,n) with {
op(i1,v1,i2,v2) = select2(v2>v1,i1,i2), max(v1,v2);
};
bmax(n) = par(i, n, (i,_)) : botReduce(op,n) with {
op(i1,v1,i2,v2) = select2(v2>v1,i1,i2), max(v1,v2);
};
process = pmax(16);
Cheers,
Yann
Le jeu. 25 mai 2023 à 15:04, Oleg Nesterov <o...@redhat.com> a écrit :
> Hello,
>
> I am wondering if it makes any sense to generalize ba.parallelOp,
> something like
>
> parallelOp(op,1) = si.bus(outputs(op));
> parallelOp(op,2) = op;
> parallelOp(op,n) = parallelOp(op,n-1),
> si.bus(inputs(op)-outputs(op)) : op;
>
> then we can do
>
> i_max(n) = par(i,n, (i,_)) : parallelOp(op,n) with {
> op(i1,v1,i2,v2) = select2(v2>v1,i1,i2), max(v1,v2);
> };
>
> for example
>
> process = 3,2,4,6,2 : i_max(5);
>
> outputs
>
> output0[i0] = FAUSTFLOAT(3); // index
> output1[i0] = FAUSTFLOAT(6); // maximum
>
> probably not.
>
> Oleg.
>
> On 05/25, Yann Orlarey wrote:
> >
> > Hi Daz,
> >
> > You did nothing wrong! I could compile it with the command line compiler,
> > but I should have also tested it with the IDE.
> >
> > Here is a better and faster solution because, instead of a tree of depth
> N,
> > it creates a more balanced tree of depth log(N):
> >
> > foo(n) = f(n,n-1) with {
> > f(1,c) = c,_;
> > f(2,c) = c-1,_,c,_ : major;
> > f(n,c) = f(n1,c-n2), f(n2,c) : major with {
> > n1 = int(n/2);
> > n2 = n-n1;
> > };
> > major(c1,v1,c2,v2) = select2(v1>v2, c2, c1), max(v1,v2);
> > };
> >
> > Cheers
> >
> > Yann
> >
> >
> >
> > Le jeu. 25 mai 2023 à 02:29, Daz Man <dazd...@hotmail.com> a écrit :
> >
> > > Hi Yann,
> > >
> > > Easier for some than others... 🙂 But...
> > >
> > > That code produces errors from "too much recursion" or "stack overflow"
> > > when run in the Faust online editor or in FaustLive. Did I perhaps use
> it
> > > wrongly?
> > >
> > > I did receive another reply from Bart, pointing to this slightly
> different
> > > function that I appear to have working, after changing my input and
> output
> > > formats to suit:
> > >
> > >
> https://github.com/magnetophon/faustExperiments/blob/0d2735bc4900dc4e51da24fb8ee05e49d44eb73c/acor.dsp#L76
> > >
> > > Thanks,
> > > Daz
> > >
> > > ------------------------------
> > > *From:* Yann Orlarey <orla...@grame.fr>
> > > *Sent:* Thursday, 25 May 2023 8:45 AM
> > > *To:* Daz Man <dazd...@hotmail.com>
> > > *Subject:* Re: [Faudiostream-users] Find N of greatest input
> > >
> > > Hi Daz,
> > >
> > > Such a function is not difficult to write:
> > >
> > > foo(1) = 0,_;
> > > foo(n) = foo(n-1), n-1,_ : major with {
> > > major(c1,v1,c2,v2) = ba.if(v1>v2, c1, c2), max(v1,v2);
> > > };
> > >
> > > Cheers
> > >
> > > Yann
> > >
> > >
> > > Le mar. 23 mai 2023 à 19:29, Daz Man <dazd...@hotmail.com> a écrit :
> > >
> > > Hi all,
> > >
> > > I need a function that compares 257 parallel inputs and returns the
> input
> > > N with the greatest positive value. (ie: which input has the biggest
> > > amplitude)
> > > What's the best way to do this?
> > > I looked at parallelMax, but it doesn't appear to return which of the N
> > > inputs was used for the max.
> > >
> > > Thanks,
> > > Daz
> > > _______________________________________________
> > > Faudiostream-users mailing list
> > > Faudiostream-users@lists.sourceforge.net
> > > https://lists.sourceforge.net/lists/listinfo/faudiostream-users
> > >
> > >
>
>
> > _______________________________________________
> > Faudiostream-users mailing list
> > Faudiostream-users@lists.sourceforge.net
> > https://lists.sourceforge.net/lists/listinfo/faudiostream-users
>
>
>
_______________________________________________
Faudiostream-users mailing list
Faudiostream-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/faudiostream-users
