A UGen for partitioned convolution. Its advantage over non-partitioning UGens such as Convolution2 is that the impulse response can be arbitrarily large amortization is used to spread processing and avoid CPU spikes.

The impulse response buffer must be specially prepared, using a /b_gen command that transforms an existing regularly formatted buffer to a new partitioned convolution ready buffer.

Examples

// Dan Stowell's reverb
// synthesize impulse response
val ir = (1f +: Vector.fill(100)(0f)) ++ (1f to 0f by -0.00002f).map { f =>
  if (math.random < 0.5)
    0f
  else
    f.pow(8) * (math.random - 0.5).signum * 0.1f
}

// ir.plot()

// send the IR to a regular buffer
val irBuf = Buffer(s)
irBuf.alloc(ir.size)
irBuf.setData(ir)

// calculate the partitioning parameters
val fftSize  = 2048
val numPart  = (ir.size * 2.0 / fftSize).ceil.toInt  // 49
val partSize = fftSize * numPart  // 100352

// create the specially formatted partitioned buffer
val partBuf  = Buffer(s)
partBuf.alloc(partSize)
// currently no predefined method for this command!
s ! osc.Message("/b_gen", partBuf.id, "PreparePartConv", irBuf.id, fftSize)

// now we can forget about the input buffer
irBuf.free()

val x = play {
  // trigger IR every 4 seconds
  val in = Impulse.ar(0.25) * 0.5
  PartConv.ar(in, fftSize, partBuf.id)
}

// do not forget to free the buffer eventually
x.free(); partBuf.free()