I'm working on my sliding reduce again.
To recap, it's like reduce, but:
- continuously updated output
- it uses just two delay lines and two operators for each doubling of
the number of samples you want to apply the operator to.
So for calculating the sum of the last 65536 samples, you just need
just 32 delay lines and 32 plus operators.
- You can still continuously vary the number of samples to compute at
runtime.
The current implementation works perfectly and is very cheap.
The downside is that you need to provide a value to signify 'disabled'.
I found a workaround for that, but it is using much more CPU then
expected.
To explain the problem, here is a much simplified version of what I'm
doing:
process(x) =
example(size,maxSize,x);
// example2(size,maxSize,x);
maxSize = 4; // for block diagram
// maxSize = 256; // for testing
size = (vslider("size", 0, 0, maxSize, 1));
example(size,maxSize,x) =
par(i, maxSize, x@i:useVal(size,i))
: combine(maxSize);
example2(size,maxSize,x) =
par(i, maxSize, x@i) <:
par(i, maxSize, combine(i+1),nothing(maxSize-i-1))
:par(i, maxSize, _*((i+1)==size)):>_
with {
nothing(0) = 0:!;
nothing(1) = !;
nothing(n) = !,nothing(n-1);
};
combine(1) = _;
combine(2) = +;
combine(n) = (combine(n-1),_) :+;
useVal(target,current) = _*(current<target);
It's much easier to understand visually:
example:
https://i.imgsafe.org/db6228a807.png
example2:
https://i.imgsafe.org/db63c1d8d7.png
The second strategy doesn't make sense for + or -, but to use the first
one a for min or max you need to replace the *1 and *0 with _ and
'disabled val'.
You cannot use strategy 1 with an arbitrary operator at all, you need
strategy 2.
So, I implemented it:
https://github.com/magnetophon/faustCompressors/blob/master/slidingReduce.lib#L95
There are a few problems though:
-I thought that parts of a faust program that don't have state can be
turned off at runtime when they are not needed for the result.
Afaik only one of the blocks of summers in example2 should be running at
one time. Going by CPU usage, it isn't.
- With a fixed number of samples, the CPU usage should be identical. It
isn't.
- With a fixed number of samples, the compiler should be able to figure
out that only one block of summers is needed, and not spend time
compiling the others. It doesn't, example2 takes ages to compile, even
with a fixed nr of samples.
Thanks for listening!
Bart.
------------------------------------------------------------------------------
What NetFlow Analyzer can do for you? Monitors network bandwidth and traffic
patterns at an interface-level. Reveals which users, apps, and protocols are
consuming the most bandwidth. Provides multi-vendor support for NetFlow,
J-Flow, sFlow and other flows. Make informed decisions using capacity
planning reports. https://ad.doubleclick.net/ddm/clk/305295220;132659582;e
_______________________________________________
Faudiostream-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/faudiostream-users