Update a-Amplifier, use Ardour::Amp

This fixes an issue with a-Amp interpolating the parameter in dB,
resulting in a double-exponential fade when the parameter changes.

Now fade is linear in dB, also using Ardour' Amp processor is more
efficient, since interpolation happens in C++.

share/scripts/_amp4.lua

@@ -0,0 +1,79 @@
+ardour {
+	["type"]    = "dsp",
+	name        = "Simple Amp IV",
+	category    = "Amplifier",
+	license     = "MIT",
+	author      = "Ardour Team",
+	description = [[Versatile +/- 20dB multichannel amplifier]]
+}
+
+function dsp_ioconfig ()
+	return
+	{
+		-- -1, -1 = any number of channels as long as input and output count matches
+		{ audio_in = -1, audio_out = -1},
+	}
+end
+
+
+function dsp_params ()
+	return
+	{
+		{ ["type"] = "input", name = "Gain", min = -20, max = 20, default = 0, unit="dB"},
+	}
+end
+
+local lpf = 0.02 -- parameter low-pass filter time-constant
+local cur_gain = 0 -- current smoothed gain (dB)
+
+-- called once when plugin is instantiated
+function dsp_init (rate)
+	lpf = 780 / rate -- interpolation time constant
+end
+
+function low_pass_filter_param (old, new, limit)
+	if math.abs (old - new) < limit  then
+		return new
+	else
+		return old + lpf * (new - old)
+	end
+end
+
+-- the DSP callback function
+-- "ins" and "outs" are http://manual.ardour.org/lua-scripting/class_reference/#C:FloatArray
+function dsp_run (ins, outs, n_samples)
+	local ctrl = CtrlPorts:array() -- get control port array (read/write)
+	local siz = n_samples -- samples remaining to process
+	local off = 0 -- already processed samples
+	local changed = false
+
+	-- if the gain parameter was changed, process at most 32 samples at a time
+	-- and interpolate gain until the current settings match the target values
+	if cur_gain ~= ctrl[1] then
+		changed = true
+		siz = 32
+	end
+
+	while n_samples > 0 do
+		if siz > n_samples then siz = n_samples end -- process at most "remaining samples"
+		if changed then
+			-- smooth gain changes above 0.02 dB difference
+			-- XXX this interpolates dB, resulting in a double-exponential fade XXX
+			cur_gain = low_pass_filter_param (cur_gain, ctrl[1], 0.02)
+		end
+
+		local gain = ARDOUR.DSP.dB_to_coefficient (cur_gain) -- 10 ^ (0.05 * cur_gain)
+
+		for c = 1,#ins do -- process all channels
+			-- check if output and input buffers for this channel are identical
+			-- http://manual.ardour.org/lua-scripting/class_reference/#C:FloatArray
+			if ins[c] ~= outs[c] then
+				-- http://manual.ardour.org/lua-scripting/class_reference/#ARDOUR:DSP
+				ARDOUR.DSP.copy_vector (outs[c]:offset (off), ins[c]:offset (off), siz)
+			end
+			ARDOUR.DSP.apply_gain_to_buffer (outs[c]:offset (off), siz, gain); -- apply-gain, process in-place
+		end
+		n_samples = n_samples - siz
+		off = off + siz
+	end
+end