>From https://ccrma.stanford.edu/~jos/aspf/Naming_Output_Signals.html
Suppose you have a block diagram bd that outputs two signals, and you
would like to give them names. You cannot simply say ``x,y = bd'' as
you would in matlab. Instead, use the signal blocking operator `!':
x = bd : _,!;
y = bd : !,_;
This does not result in two instances of bd. The optimization recognizes
the common subexpression bd and computes it only once.
And this is what compiler/boxes/boxes.cpp:makeRecProjectionsList() does.
Unfortunately, this is not that simple when "bd" has inputs, in this case "This
does not result in two instances of bd" is not true in some sense; and that is
why letrec is buggy imo.
So, iiuc boxWithRecDef() translates this code
process = x, y letrec {
'x = expr_x;
'y = expr_y;
};
into something like
process = x,y with {
// buildRecursiveBodyDef()
LETRECBODY = \(x, y).(expr_x, expr_y) ~ (_,_);
// makeRecProjectionsList()
x = LETRECBODY : _,!;
y = LETRECBODY : !,_;
};
just like the doc above suggests. However this means that
inputs(process) == 2 * (inputs(expr_x) + inputs(expr_y))
and the halve of these inputs will be ignored. This looks wrong to me.
For example,
process = _ : bug;
bug = x,y letrec {
'x = _;
'y = 0;
};
doesn't compile because inputs(bug) == 2.
This code
process = _,_ : bug;
bug = x,y letrec {
'x = _;
'y = 0;
};
compiles to
virtual void compute(int count, FAUSTFLOAT** inputs, FAUSTFLOAT**
outputs) {
FAUSTFLOAT* input0 = inputs[0];
FAUSTFLOAT* input1 = inputs[1];
FAUSTFLOAT* output0 = outputs[0];
FAUSTFLOAT* output1 = outputs[1];
for (int i = 0; (i < count); i = (i + 1)) {
output0[i] = FAUSTFLOAT(float(input0[i]));
output1[i] = FAUSTFLOAT(0);
}
}
note that input1 is not used.
Oleg.
_______________________________________________
Faudiostream-users mailing list
Faudiostream-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/faudiostream-users
