Paul Davis
b35518e212
This is mostly a simple lexical search+replace but the absence of operator< for std::weak_ptr<T> leads to some complications, particularly with Evoral::Sequence and ExportPortChannel.
466 lines
12 KiB
C++
466 lines
12 KiB
C++
/*
|
|
* Copyright (C) 2011-2017 Paul Davis <paul@linuxaudiosystems.com>
|
|
* Copyright (C) 2011 Carl Hetherington <carl@carlh.net>
|
|
* Copyright (C) 2013 John Emmas <john@creativepost.co.uk>
|
|
* Copyright (C) 2014 Robin Gareus <robin@gareus.org>
|
|
*
|
|
* This program is free software; you can redistribute it and/or modify
|
|
* it under the terms of the GNU General Public License as published by
|
|
* the Free Software Foundation; either version 2 of the License, or
|
|
* (at your option) any later version.
|
|
*
|
|
* This program is distributed in the hope that it will be useful,
|
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
* GNU General Public License for more details.
|
|
*
|
|
* You should have received a copy of the GNU General Public License along
|
|
* with this program; if not, write to the Free Software Foundation, Inc.,
|
|
* 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
|
|
*/
|
|
|
|
#include <cmath>
|
|
#include <cstdio>
|
|
#include <cstdlib>
|
|
#include <cstring>
|
|
|
|
#include <iostream>
|
|
#include <string>
|
|
|
|
#ifdef COMPILER_MSVC
|
|
#include <malloc.h>
|
|
#endif
|
|
|
|
#include "pbd/cartesian.h"
|
|
#include "pbd/compose.h"
|
|
|
|
#include "ardour/amp.h"
|
|
#include "ardour/audio_buffer.h"
|
|
#include "ardour/buffer_set.h"
|
|
#include "ardour/pan_controllable.h"
|
|
#include "ardour/pannable.h"
|
|
#include "ardour/speakers.h"
|
|
|
|
#include "vbap.h"
|
|
#include "vbap_speakers.h"
|
|
|
|
#include "pbd/i18n.h"
|
|
|
|
using namespace PBD;
|
|
using namespace ARDOUR;
|
|
using namespace std;
|
|
|
|
static PanPluginDescriptor _descriptor = {
|
|
"VBAP 2D panner",
|
|
"http://ardour.org/plugin/panner_vbap",
|
|
"http://ardour.org/plugin/panner_vbap#ui",
|
|
-1, -1,
|
|
10,
|
|
VBAPanner::factory
|
|
};
|
|
|
|
extern "C" ARDOURPANNER_API PanPluginDescriptor*
|
|
panner_descriptor ()
|
|
{
|
|
return &_descriptor;
|
|
}
|
|
|
|
VBAPanner::Signal::Signal (VBAPanner&, uint32_t, uint32_t n_speakers)
|
|
{
|
|
resize_gains (n_speakers);
|
|
|
|
desired_gains[0] = desired_gains[1] = desired_gains[2] = 0;
|
|
outputs[0] = outputs[1] = outputs[2] = -1;
|
|
desired_outputs[0] = desired_outputs[1] = desired_outputs[2] = -1;
|
|
}
|
|
|
|
void
|
|
VBAPanner::Signal::resize_gains (uint32_t n)
|
|
{
|
|
gains.assign (n, 0.0);
|
|
}
|
|
|
|
VBAPanner::VBAPanner (std::shared_ptr<Pannable> p, std::shared_ptr<Speakers> s)
|
|
: Panner (p)
|
|
, _speakers (new VBAPSpeakers (s))
|
|
{
|
|
_pannable->pan_azimuth_control->Changed.connect_same_thread (*this, boost::bind (&VBAPanner::update, this));
|
|
_pannable->pan_elevation_control->Changed.connect_same_thread (*this, boost::bind (&VBAPanner::update, this));
|
|
_pannable->pan_width_control->Changed.connect_same_thread (*this, boost::bind (&VBAPanner::update, this));
|
|
if (!_pannable->has_state ()) {
|
|
reset ();
|
|
}
|
|
|
|
update ();
|
|
}
|
|
|
|
VBAPanner::~VBAPanner ()
|
|
{
|
|
clear_signals ();
|
|
}
|
|
|
|
void
|
|
VBAPanner::clear_signals ()
|
|
{
|
|
for (vector<Signal*>::iterator i = _signals.begin (); i != _signals.end (); ++i) {
|
|
delete *i;
|
|
}
|
|
_signals.clear ();
|
|
}
|
|
|
|
void
|
|
VBAPanner::configure_io (ChanCount in, ChanCount /* ignored - we use Speakers */)
|
|
{
|
|
uint32_t n = in.n_audio ();
|
|
|
|
clear_signals ();
|
|
|
|
for (uint32_t i = 0; i < n; ++i) {
|
|
Signal* s = new Signal (*this, i, _speakers->n_speakers ());
|
|
_signals.push_back (s);
|
|
}
|
|
|
|
update ();
|
|
}
|
|
|
|
void
|
|
VBAPanner::update ()
|
|
{
|
|
_can_automate_list.clear ();
|
|
_can_automate_list.insert (Evoral::Parameter (PanAzimuthAutomation));
|
|
if (_signals.size () > 1) {
|
|
_can_automate_list.insert (Evoral::Parameter (PanWidthAutomation));
|
|
}
|
|
if (_speakers->dimension () == 3) {
|
|
_can_automate_list.insert (Evoral::Parameter (PanElevationAutomation));
|
|
}
|
|
|
|
/* recompute signal directions based on panner azimuth and, if relevant, width (diffusion) and elevation parameters */
|
|
double elevation = _pannable->pan_elevation_control->get_value () * 90.0;
|
|
|
|
if (_signals.size () > 1) {
|
|
double w = -(_pannable->pan_width_control->get_value ());
|
|
double signal_direction = 1.0 - (_pannable->pan_azimuth_control->get_value () + (w / 2));
|
|
double grd_step_per_signal = w / (_signals.size () - 1);
|
|
for (vector<Signal*>::iterator s = _signals.begin (); s != _signals.end (); ++s) {
|
|
Signal* signal = *s;
|
|
|
|
int over = signal_direction;
|
|
over -= (signal_direction >= 0) ? 0 : 1;
|
|
signal_direction -= (double)over;
|
|
|
|
signal->direction = AngularVector (signal_direction * 360.0, elevation);
|
|
compute_gains (signal->desired_gains, signal->desired_outputs, signal->direction.azi, signal->direction.ele);
|
|
signal_direction += grd_step_per_signal;
|
|
}
|
|
} else if (_signals.size () == 1) {
|
|
double center = (1.0 - _pannable->pan_azimuth_control->get_value ()) * 360.0;
|
|
|
|
/* width has no role to play if there is only 1 signal: VBAP does not do "diffusion" of a single channel */
|
|
|
|
Signal* s = _signals.front ();
|
|
s->direction = AngularVector (center, elevation);
|
|
compute_gains (s->desired_gains, s->desired_outputs, s->direction.azi, s->direction.ele);
|
|
}
|
|
|
|
SignalPositionChanged (); /* emit */
|
|
}
|
|
|
|
void
|
|
VBAPanner::compute_gains (double gains[3], int speaker_ids[3], int azi, int ele)
|
|
{
|
|
/* calculates gain factors using loudspeaker setup and given direction */
|
|
double cartdir[3];
|
|
double power;
|
|
int i, j, k;
|
|
double small_g;
|
|
double big_sm_g, gtmp[3];
|
|
const int dimension = _speakers->dimension ();
|
|
assert (dimension == 2 || dimension == 3);
|
|
|
|
spherical_to_cartesian (azi, ele, 1.0, cartdir[0], cartdir[1], cartdir[2]);
|
|
big_sm_g = -100000.0;
|
|
|
|
gains[0] = gains[1] = gains[2] = 0;
|
|
speaker_ids[0] = speaker_ids[1] = speaker_ids[2] = 0;
|
|
|
|
for (i = 0; i < _speakers->n_tuples (); i++) {
|
|
small_g = 10000000.0;
|
|
|
|
for (j = 0; j < dimension; j++) {
|
|
gtmp[j] = 0.0;
|
|
|
|
for (k = 0; k < dimension; k++) {
|
|
gtmp[j] += cartdir[k] * _speakers->matrix (i)[j * dimension + k];
|
|
}
|
|
|
|
if (gtmp[j] < small_g) {
|
|
small_g = gtmp[j];
|
|
}
|
|
}
|
|
|
|
if (small_g > big_sm_g) {
|
|
big_sm_g = small_g;
|
|
|
|
gains[0] = gtmp[0];
|
|
gains[1] = gtmp[1];
|
|
|
|
speaker_ids[0] = _speakers->speaker_for_tuple (i, 0);
|
|
speaker_ids[1] = _speakers->speaker_for_tuple (i, 1);
|
|
|
|
if (_speakers->dimension () == 3) {
|
|
gains[2] = gtmp[2];
|
|
speaker_ids[2] = _speakers->speaker_for_tuple (i, 2);
|
|
} else {
|
|
gains[2] = 0.0;
|
|
speaker_ids[2] = -1;
|
|
}
|
|
}
|
|
}
|
|
|
|
power = sqrt (gains[0] * gains[0] + gains[1] * gains[1] + gains[2] * gains[2]);
|
|
|
|
if (power > 0) {
|
|
gains[0] /= power;
|
|
gains[1] /= power;
|
|
gains[2] /= power;
|
|
}
|
|
}
|
|
|
|
void
|
|
VBAPanner::distribute (BufferSet& inbufs, BufferSet& obufs, gain_t gain_coefficient, pframes_t nframes)
|
|
{
|
|
uint32_t n;
|
|
vector<Signal*>::iterator s;
|
|
|
|
assert (inbufs.count ().n_audio () == _signals.size ());
|
|
|
|
for (s = _signals.begin (), n = 0; s != _signals.end (); ++s, ++n) {
|
|
Signal* signal (*s);
|
|
|
|
distribute_one (inbufs.get_audio (n), obufs, gain_coefficient, nframes, n);
|
|
|
|
memcpy (signal->outputs, signal->desired_outputs, sizeof (signal->outputs));
|
|
}
|
|
}
|
|
|
|
void
|
|
VBAPanner::distribute_one (AudioBuffer& srcbuf, BufferSet& obufs, gain_t gain_coefficient, pframes_t nframes, uint32_t which)
|
|
{
|
|
Sample* const src = srcbuf.data ();
|
|
Signal* signal (_signals[which]);
|
|
|
|
/* VBAP may distribute the signal across up to 3 speakers depending on
|
|
* the configuration of the speakers.
|
|
*
|
|
* But the set of speakers in use "this time" may be different from
|
|
* the set of speakers "the last time". So we have up to 6 speakers
|
|
* involved, and we have to interpolate so that those no longer
|
|
* in use are rapidly faded to silence and those newly in use
|
|
* are rapidly faded to their correct level. This prevents clicks
|
|
* as we change the set of speakers used to put the signal in
|
|
* a given position.
|
|
*
|
|
* However, the speakers are represented by output buffers, and other
|
|
* speakers may write to the same buffers, so we cannot use
|
|
* anything here that will simply assign new (sample) values
|
|
* to the output buffers - everything must be done via mixing
|
|
* functions and not assignment/copying.
|
|
*/
|
|
|
|
vector<double>::size_type sz = signal->gains.size ();
|
|
|
|
assert (sz == obufs.count ().n_audio ());
|
|
|
|
int8_t* outputs = (int8_t*)alloca (sz); // on the stack, no malloc
|
|
|
|
/* set initial state of each output "record"
|
|
*/
|
|
|
|
for (uint32_t o = 0; o < sz; ++o) {
|
|
outputs[o] = 0;
|
|
}
|
|
|
|
/* for all outputs used this time and last time,
|
|
* change the output record to show what has
|
|
* happened.
|
|
*/
|
|
|
|
for (int o = 0; o < 3; ++o) {
|
|
if (signal->outputs[o] != -1) {
|
|
/* used last time */
|
|
outputs[signal->outputs[o]] |= 1;
|
|
}
|
|
|
|
if (signal->desired_outputs[o] != -1) {
|
|
/* used this time */
|
|
outputs[signal->desired_outputs[o]] |= 1 << 1;
|
|
}
|
|
}
|
|
|
|
/* at this point, we can test a speaker's status:
|
|
*
|
|
* (*outputs[o] & 1) <= in use before
|
|
* (*outputs[o] & 2) <= in use this time
|
|
* (*outputs[o] & 3) == 3 <= in use both times
|
|
* *outputs[o] == 0 <= not in use either time
|
|
*
|
|
*/
|
|
|
|
for (int o = 0; o < 3; ++o) {
|
|
pan_t pan;
|
|
int output = signal->desired_outputs[o];
|
|
|
|
if (output == -1) {
|
|
continue;
|
|
}
|
|
|
|
pan = gain_coefficient * signal->desired_gains[o];
|
|
|
|
if (pan == 0.0 && signal->gains[output] == 0.0) {
|
|
/* nothing deing delivered to this output */
|
|
|
|
signal->gains[output] = 0.0;
|
|
|
|
} else if (fabs (pan - signal->gains[output]) > 0.00001) {
|
|
/* signal to this output but the gain coefficient has changed, so
|
|
* interpolate between them.
|
|
*/
|
|
|
|
AudioBuffer& buf (obufs.get_audio (output));
|
|
buf.accumulate_with_ramped_gain_from (srcbuf.data (), nframes, signal->gains[output], pan, 0);
|
|
signal->gains[output] = pan;
|
|
|
|
} else {
|
|
/* signal to this output, same gain as before so just copy with gain */
|
|
|
|
mix_buffers_with_gain (obufs.get_audio (output).data (), src, nframes, pan);
|
|
signal->gains[output] = pan;
|
|
}
|
|
}
|
|
|
|
/* clean up the outputs that were used last time but not this time */
|
|
|
|
for (uint32_t o = 0; o < sz; ++o) {
|
|
if (outputs[o] == 1) {
|
|
/* take signal and deliver with a rapid fade out */
|
|
AudioBuffer& buf (obufs.get_audio (o));
|
|
buf.accumulate_with_ramped_gain_from (srcbuf.data (), nframes, signal->gains[o], 0.0, 0);
|
|
signal->gains[o] = 0.0;
|
|
}
|
|
}
|
|
|
|
/* note that the output buffers were all silenced at some point
|
|
* so anything we didn't write to with this signal (or any others)
|
|
* is just as it should be.
|
|
*/
|
|
}
|
|
|
|
void
|
|
VBAPanner::distribute_one_automated (AudioBuffer& /*src*/, BufferSet& /*obufs*/,
|
|
samplepos_t /*start*/, samplepos_t /*end*/,
|
|
pframes_t /*nframes*/, pan_t** /*buffers*/, uint32_t /*which*/)
|
|
{
|
|
/* XXX to be implemented */
|
|
}
|
|
|
|
XMLNode&
|
|
VBAPanner::get_state () const
|
|
{
|
|
XMLNode& node (Panner::get_state ());
|
|
node.set_property (X_ ("uri"), _descriptor.panner_uri);
|
|
/* this is needed to allow new sessions to load with old Ardour: */
|
|
node.set_property (X_ ("type"), _descriptor.name);
|
|
return node;
|
|
}
|
|
|
|
Panner*
|
|
VBAPanner::factory (std::shared_ptr<Pannable> p, std::shared_ptr<Speakers> s)
|
|
{
|
|
return new VBAPanner (p, s);
|
|
}
|
|
|
|
ChanCount
|
|
VBAPanner::in () const
|
|
{
|
|
return ChanCount (DataType::AUDIO, _signals.size ());
|
|
}
|
|
|
|
ChanCount
|
|
VBAPanner::out () const
|
|
{
|
|
return ChanCount (DataType::AUDIO, _speakers->n_speakers ());
|
|
}
|
|
|
|
string
|
|
VBAPanner::value_as_string (std::shared_ptr<const AutomationControl> ac) const
|
|
{
|
|
double val = ac->get_value ();
|
|
|
|
switch (ac->parameter ().type ()) {
|
|
case PanAzimuthAutomation: /* direction */
|
|
return string_compose (_ (u8"%1\u00B0"), (int(rint (val * 360.0)) + 180) % 360);
|
|
|
|
case PanWidthAutomation: /* diffusion */
|
|
return string_compose (_ ("%1%%"), (int)floor (100.0 * fabs (val)));
|
|
|
|
case PanElevationAutomation: /* elevation */
|
|
return string_compose (_ (u8"%1\u00B0"), (int)floor (90.0 * fabs (val)));
|
|
|
|
default:
|
|
return _ ("unused");
|
|
}
|
|
}
|
|
|
|
AngularVector
|
|
VBAPanner::signal_position (uint32_t n) const
|
|
{
|
|
if (n < _signals.size ()) {
|
|
return _signals[n]->direction;
|
|
}
|
|
|
|
return AngularVector ();
|
|
}
|
|
|
|
std::shared_ptr<Speakers>
|
|
VBAPanner::get_speakers () const
|
|
{
|
|
return _speakers->parent ();
|
|
}
|
|
|
|
void
|
|
VBAPanner::set_position (double p)
|
|
{
|
|
/* map into 0..1 range */
|
|
int over = p;
|
|
over -= (p >= 0) ? 0 : 1;
|
|
p -= (double)over;
|
|
_pannable->pan_azimuth_control->set_value (p, Controllable::NoGroup);
|
|
}
|
|
|
|
void
|
|
VBAPanner::set_width (double w)
|
|
{
|
|
_pannable->pan_width_control->set_value (min (1.0, max (-1.0, w)), Controllable::NoGroup);
|
|
}
|
|
|
|
void
|
|
VBAPanner::set_elevation (double e)
|
|
{
|
|
_pannable->pan_elevation_control->set_value (min (1.0, max (0.0, e)), Controllable::NoGroup);
|
|
}
|
|
|
|
void
|
|
VBAPanner::reset ()
|
|
{
|
|
set_position (.5);
|
|
if (_signals.size () > 1) {
|
|
set_width (1.0 - (1.0 / (double)_signals.size ()));
|
|
} else {
|
|
set_width (1.0);
|
|
}
|
|
set_elevation (0);
|
|
|
|
update ();
|
|
}
|