Merge branch 'pr45/swesterfeld-freeverb-mix' - closes #23

* pr45/swesterfeld-freeverb-mix:
  devices/freeverb/freeverb.cc: support sample-accurate parameter changes
  devices/freeverb/freeverb.cc: optimize processing for constant mix case
  devices/freeverb/freeverb.cc: merge dry/wet parameters into mix parameter
	See #23. Also add parameter smoothing for mix to avoid clicks when
	changing/automating the parameter.

Signed-off-by: Tim Janik <timj@gnu.org>

devices/freeverb/freeverb.cc

@@ -1,6 +1,7 @@
 // This Source Code Form is licensed MPL-2.0: http://mozilla.org/MPL/2.0
 #include "ase/processor.hh"
 #include "ase/internal.hh"
+#include "devices/blepsynth/linearsmooth.hh"
 
 namespace {
 
@@ -15,6 +16,8 @@ class Freeverb : public AudioProcessor {
   IBusId stereoin;
   OBusId stereout;
   revmodel model;
+  LinearSmooth mix_smooth_;
+  bool mix_smooth_reset_ = false;
 public:
   Freeverb (const ProcessorSetup &psetup) :
     AudioProcessor (psetup)
@@ -28,7 +31,7 @@ class Freeverb : public AudioProcessor {
     info.website_url = "https://beast.testbit.eu";
     info.creator_name = "Jezar at Dreampoint";
   }
-  enum Params { MODE = 1, DRY, WET, ROOMSIZE, DAMPING, WIDTH };
+  enum Params { MODE = 1, MIX, ROOMSIZE, DAMPING, WIDTH };
   void
   initialize (SpeakerArrangement busses) override
   {
@@ -40,8 +43,7 @@ class Freeverb : public AudioProcessor {
     ParameterMap pmap;
 
     pmap.group = _("Reverb Settings");
-    pmap[DRY] = Param ("drylevel", _("Dry level"), _("Dry"), scaledry * initialdry, "dB", { 0, scaledry });
-    pmap[WET] = Param ("wetlevel", _("Wet level"), _("Wet"), scalewet * initialwet, "dB", { 0, scalewet });
+    pmap[MIX] = Param { "mix",   "Mix dry/wet", "Mix", 30, "%", { 0, 100 } };
 
     ChoiceS centries;
     centries += { "Signflip 2000",  _("Preserve May 2000 Freeverb damping sign flip") };
@@ -62,8 +64,9 @@ class Freeverb : public AudioProcessor {
   {
     switch (Params (paramid))
       {
-      case WET:         return model.setwet (get_param (paramid) / scalewet);
-      case DRY:         return model.setdry (get_param (paramid) / scaledry);
+      case MIX:         mix_smooth_.set (get_param (paramid) * 0.01, mix_smooth_reset_);
+                        mix_smooth_reset_ = false;
+                        break;
       case ROOMSIZE:    return model.setroomsize ((get_param (paramid) - offsetroom) / scaleroom);
       case WIDTH:       return model.setwidth (0.01 * get_param (paramid));
       case MODE:
@@ -75,18 +78,68 @@ class Freeverb : public AudioProcessor {
   reset (uint64 target_stamp) override
   {
     model.setmode (0);          // no-freeze, allow mute
+    model.setdry (0);           // no dry, we mix the output signal during render
+    model.setwet (1);           // only need reverb wet
     model.mute();               // silence internal buffers
+    mix_smooth_.reset (sample_rate(), 0.020);
+    mix_smooth_reset_ = true;
     adjust_all_params();
   }
   void
+  render_audio (float *input0, float *input1, float *output0, float *output1, uint n_frames)
+  {
+    if (!n_frames)
+      return;
+
+    // this only generates the wet signal
+    model.processreplace (input0, input1, output0, output1, n_frames, 1);
+
+    if (mix_smooth_.is_constant()) // faster version: mix parameter is constant during the block
+      {
+        float mix = mix_smooth_.get_next();
+        for (uint i = 0; i < n_frames; i++)
+          {
+            output0[i] = input0[i] * (1 - mix) + output0[i] * mix;
+            output1[i] = input1[i] * (1 - mix) + output1[i] * mix;
+          }
+      }
+    else
+      {
+        for (uint i = 0; i < n_frames; i++)
+          {
+            float mix = mix_smooth_.get_next();
+            output0[i] = input0[i] * (1 - mix) + output0[i] * mix;
+            output1[i] = input1[i] * (1 - mix) + output1[i] * mix;
+          }
+      }
+  }
+  void
   render (uint n_frames) override
   {
-    apply_input_events();
     float *input0 = const_cast<float*> (ifloats (stereoin, 0));
     float *input1 = const_cast<float*> (ifloats (stereoin, 1));
     float *output0 = oblock (stereout, 0);
     float *output1 = oblock (stereout, 1);
-    model.processreplace (input0, input1, output0, output1, n_frames, 1);
+
+    uint offset = 0;
+    MidiEventInput evinput = midi_event_input();
+    for (const auto &ev : evinput)
+      {
+        // process any audio that is before the event
+        render_audio (input0 + offset, input1 + offset, output0 + offset, output1 + offset, ev.frame - offset);
+        offset = ev.frame;
+
+        switch (ev.message())
+          {
+          case MidiMessage::PARAM_VALUE:
+            apply_event (ev);
+            adjust_param (ev.param);
+            break;
+          default: ;
+          }
+      }
+    // process frames after last event
+    render_audio (input0 + offset, input1 + offset, output0 + offset, output1 + offset, n_frames - offset);
   }
 };
 static auto freeverb = register_audio_processor<Freeverb> ("Ase::Devices::Freeverb");