zmusic/thirdparty/timidity/mix.cpp
Christoph Oelckers a4eae42ec5 - initial commit
2020-01-02 17:52:30 +01:00

759 lines
14 KiB
C++

/*
TiMidity -- Experimental MIDI to WAVE converter
Copyright (C) 1995 Tuukka Toivonen <toivonen@clinet.fi>
This library is free software; you can redistribute it and/or
modify it under the terms of the GNU Lesser General Public
License as published by the Free Software Foundation; either
version 2.1 of the License, or (at your option) any later version.
This library 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
Lesser General Public License for more details.
You should have received a copy of the GNU Lesser General Public
License along with this library; if not, write to the Free Software
Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
mix.c
*/
#include <math.h>
#include <stdio.h>
#include <stdlib.h>
#include "timidity.h"
#include "common.h"
#include "instrum.h"
#include "playmidi.h"
namespace Timidity
{
static int convert_envelope_rate(Renderer *song, uint8_t rate)
{
int r;
r = 3 - ((rate>>6) & 0x3);
r *= 3;
r = (int)(rate & 0x3f) << r; /* 6.9 fixed point */
/* 15.15 fixed point. */
return int(((r * 44100) / song->rate) * song->control_ratio) << 9;
}
void Envelope::Init(Renderer *song, Voice *v)
{
Type = v->sample->type;
env.bUpdating = true;
if (Type == INST_GUS)
{
gf1.Init(song, v);
gf1.ApplyToAmp(v);
}
else
{
sf2.Init(song, v);
sf2.ApplyToAmp(v);
}
}
void GF1Envelope::Init(Renderer *song, Voice *v)
{
/* Ramp up from 0 */
stage = 0;
volume = 0;
for (int i = 0; i < 6; ++i)
{
offset[i] = v->sample->envelope.gf1.offset[i] << (7 + 15);
rate[i] = convert_envelope_rate(song, v->sample->envelope.gf1.rate[i]);
}
Recompute(v);
}
void GF1Envelope::Release(Voice *v)
{
if (!(v->sample->modes & PATCH_NO_SRELEASE) || (v->sample->modes & PATCH_FAST_REL))
{
/* ramp out to minimum volume with rate from final release stage */
stage = GF1_RELEASEC+1;
target = 0;
increment = -rate[GF1_RELEASEC];
}
else if (v->sample->modes & PATCH_SUSTAIN)
{
if (stage < GF1_RELEASE)
{
stage = GF1_RELEASE;
}
Recompute(v);
}
bUpdating = true;
}
/* Returns 1 if envelope runs out */
bool GF1Envelope::Recompute(Voice *v)
{
int newstage;
newstage = stage;
if (newstage > GF1_RELEASEC)
{
/* Envelope ran out. */
increment = 0;
bUpdating = false;
v->status &= ~(VOICE_SUSTAINING | VOICE_LPE);
v->status |= VOICE_RELEASING;
/* play sampled release */
return 0;
}
if (newstage == GF1_RELEASE && !(v->status & VOICE_RELEASING) && (v->sample->modes & PATCH_SUSTAIN))
{
v->status |= VOICE_SUSTAINING;
/* Freeze envelope until note turns off. Trumpets want this. */
increment = 0;
bUpdating = false;
}
else
{
stage = newstage + 1;
if (volume == offset[newstage])
{
return Recompute(v);
}
target = offset[newstage];
increment = rate[newstage];
if (target < volume)
increment = -increment;
}
return 0;
}
bool GF1Envelope::Update(Voice *v)
{
volume += increment;
if (((increment < 0) && (volume <= target)) || ((increment > 0) && (volume >= target)))
{
volume = target;
if (Recompute(v))
{
return 1;
}
}
return 0;
}
void GF1Envelope::ApplyToAmp(Voice *v)
{
double env_vol = v->attenuation;
double final_amp;
final_amp = FINAL_MIX_SCALE;
if (v->tremolo_phase_increment != 0)
{ // [RH] FIXME: This is wrong. Tremolo should offset the
// envelope volume, not scale it.
env_vol *= v->tremolo_volume;
}
env_vol *= volume / float(1 << 30);
env_vol = calc_gf1_amp(env_vol);
env_vol *= final_amp;
v->left_mix = float(env_vol * v->left_offset);
v->right_mix = float(env_vol * v->right_offset);
}
void SF2Envelope::Init(Renderer *song, Voice *v)
{
stage = 0;
volume = 0;
DelayTime = v->sample->envelope.sf2.delay_vol;
AttackTime = v->sample->envelope.sf2.attack_vol;
HoldTime = v->sample->envelope.sf2.hold_vol;
DecayTime = v->sample->envelope.sf2.decay_vol;
SustainLevel = v->sample->envelope.sf2.sustain_vol;
ReleaseTime = v->sample->envelope.sf2.release_vol;
SampleRate = song->rate;
HoldStart = 0;
RateMul = song->control_ratio / song->rate;
RateMul_cB = RateMul * 960;
bUpdating = true;
}
void SF2Envelope::Release(Voice *v)
{
if (stage == SF2_ATTACK)
{
// The attack stage does not use an attenuation in cB like all the rest.
volume = float(log10(volume) * -200);
}
stage = SF2_RELEASE;
bUpdating = true;
}
static double timecent_to_sec(float timecent)
{
if (timecent == -32768)
return 0;
return pow(2.0, timecent / 1200.0);
}
static double calc_rate(double ratemul, double sec)
{
if (sec < 0.006)
sec = 0.006;
return ratemul / sec;
}
static void shutoff_voice(Voice *v)
{
v->status &= ~(VOICE_SUSTAINING | VOICE_LPE);
v->status |= VOICE_RELEASING | VOICE_STOPPING;
}
static bool check_release(double RateMul, double sec)
{
double rate = calc_rate(960 * RateMul, sec);
// Is release rate very fast? If so, don't do the release, but do
// the voice off ramp instead.
return (rate < 960/20);
}
/* Returns 1 if envelope runs out */
bool SF2Envelope::Update(Voice *v)
{
double sec;
double newvolume = 0;
// NOTE! The volume scale is different for different stages of the
// envelope generator:
// Attack stage goes from 0.0 -> 1.0, multiplied directly to the output.
// The following stages go from 0 -> -1000 cB (but recorded positively)
switch (stage)
{
case SF2_DELAY:
if (v->sample_count >= timecent_to_sec(DelayTime) * SampleRate)
{
stage = SF2_ATTACK;
return Update(v);
}
return 0;
case SF2_ATTACK:
sec = timecent_to_sec(AttackTime);
if (sec <= 0)
{ // instantaneous attack
newvolume = 1;
}
else
{
newvolume = volume + calc_rate(RateMul, sec);
}
if (newvolume >= 1)
{
volume = 0;
HoldStart = v->sample_count;
if (HoldTime <= -32768)
{ // hold time is 0, so skip right to decay
stage = SF2_DECAY;
}
else
{
stage = SF2_HOLD;
}
return Update(v);
}
break;
case SF2_HOLD:
if (v->sample_count - HoldStart >= timecent_to_sec(HoldTime) * SampleRate)
{
stage = SF2_DECAY;
return Update(v);
}
return 0;
case SF2_DECAY:
sec = timecent_to_sec(DecayTime);
if (sec <= 0)
{ // instantaneous decay
newvolume = SustainLevel;
}
else
{
newvolume = volume + calc_rate(RateMul_cB, sec);
}
if (newvolume >= SustainLevel)
{
newvolume = SustainLevel;
stage = SF2_SUSTAIN;
bUpdating = false;
if (!(v->status & VOICE_RELEASING))
{
v->status |= VOICE_SUSTAINING;
}
}
break;
case SF2_SUSTAIN:
// Stay here until released.
return 0;
case SF2_RELEASE:
sec = timecent_to_sec(ReleaseTime);
if (sec <= 0)
{ // instantaneous release
newvolume = 1000;
}
else
{
newvolume = volume + calc_rate(RateMul_cB, sec);
}
if (newvolume >= 960)
{
stage = SF2_FINISHED;
shutoff_voice(v);
bUpdating = false;
return 1;
}
break;
case SF2_FINISHED:
return 1;
}
volume = (float)newvolume;
return 0;
}
/* EMU 8k/10k don't follow spec in regards to volume attenuation.
* This factor is used in the equation pow (10.0, cb / FLUID_ATTEN_POWER_FACTOR).
* By the standard this should be -200.0. */
#define FLUID_ATTEN_POWER_FACTOR (-531.509)
#define atten2amp(x) pow(10.0, (x) / FLUID_ATTEN_POWER_FACTOR)
static double cb_to_amp(double x) // centibels to amp
{
return pow(10, x / -200.f);
}
void SF2Envelope::ApplyToAmp(Voice *v)
{
double amp;
if (stage == SF2_DELAY)
{
v->left_mix = 0;
v->right_mix = 0;
return;
}
amp = v->sample->type == INST_SF2 ? atten2amp(v->attenuation) : cb_to_amp(v->attenuation);
switch (stage)
{
case SF2_ATTACK:
amp *= volume;
break;
case SF2_HOLD:
break;
default:
amp *= cb_to_amp(volume);
break;
}
amp *= FINAL_MIX_SCALE * 0.5;
v->left_mix = float(amp * v->left_offset);
v->right_mix = float(amp * v->right_offset);
}
void apply_envelope_to_amp(Voice *v)
{
v->eg1.ApplyToAmp(v);
}
static void update_tremolo(Voice *v)
{
int depth = v->sample->tremolo_depth << 7;
if (v->tremolo_sweep != 0)
{
/* Update sweep position */
v->tremolo_sweep_position += v->tremolo_sweep;
if (v->tremolo_sweep_position >= (1 << SWEEP_SHIFT))
{
/* Swept to max amplitude */
v->tremolo_sweep = 0;
}
else
{
/* Need to adjust depth */
depth *= v->tremolo_sweep_position;
depth >>= SWEEP_SHIFT;
}
}
v->tremolo_phase += v->tremolo_phase_increment;
v->tremolo_volume = (float)
(1.0 - FSCALENEG((sine(v->tremolo_phase >> RATE_SHIFT) + 1.0)
* depth * TREMOLO_AMPLITUDE_TUNING,
17));
/* I'm not sure about the +1.0 there -- it makes tremoloed voices'
volumes on average the lower the higher the tremolo amplitude. */
}
/* Returns 1 if the note died */
static int update_signal(Voice *v)
{
if (v->eg1.env.bUpdating && v->eg1.Update(v))
{
return 1;
}
if (v->tremolo_phase_increment != 0)
{
update_tremolo(v);
}
apply_envelope_to_amp(v);
return 0;
}
static void mix_mystery_signal(int32_t control_ratio, const sample_t *sp, float *lp, Voice *v, int count)
{
final_volume_t
left = v->left_mix,
right = v->right_mix;
int cc;
sample_t s;
if (!(cc = v->control_counter))
{
cc = control_ratio;
if (update_signal(v))
return; /* Envelope ran out */
left = v->left_mix;
right = v->right_mix;
}
while (count)
{
if (cc < count)
{
count -= cc;
while (cc--)
{
s = *sp++;
lp[0] += left * s;
lp[1] += right * s;
lp += 2;
}
cc = control_ratio;
if (update_signal(v))
return; /* Envelope ran out */
left = v->left_mix;
right = v->right_mix;
}
else
{
v->control_counter = cc - count;
while (count--)
{
s = *sp++;
lp[0] += left * s;
lp[1] += right * s;
lp += 2;
}
return;
}
}
}
static void mix_single_signal(int32_t control_ratio, const sample_t *sp, float *lp, Voice *v, float *ampat, int count)
{
final_volume_t amp;
int cc;
if (0 == (cc = v->control_counter))
{
cc = control_ratio;
if (update_signal(v))
return; /* Envelope ran out */
}
amp = *ampat;
while (count)
{
if (cc < count)
{
count -= cc;
while (cc--)
{
lp[0] += *sp++ * amp;
lp += 2;
}
cc = control_ratio;
if (update_signal(v))
return; /* Envelope ran out */
amp = *ampat;
}
else
{
v->control_counter = cc - count;
while (count--)
{
lp[0] += *sp++ * amp;
lp += 2;
}
return;
}
}
}
static void mix_single_left_signal(int32_t control_ratio, const sample_t *sp, float *lp, Voice *v, int count)
{
mix_single_signal(control_ratio, sp, lp, v, &v->left_mix, count);
}
static void mix_single_right_signal(int32_t control_ratio, const sample_t *sp, float *lp, Voice *v, int count)
{
mix_single_signal(control_ratio, sp, lp + 1, v, &v->right_mix, count);
}
static void mix_mono_signal(int32_t control_ratio, const sample_t *sp, float *lp, Voice *v, int count)
{
final_volume_t
left = v->left_mix;
int cc;
if (!(cc = v->control_counter))
{
cc = control_ratio;
if (update_signal(v))
return; /* Envelope ran out */
left = v->left_mix;
}
while (count)
{
if (cc < count)
{
count -= cc;
while (cc--)
{
*lp++ += *sp++ * left;
}
cc = control_ratio;
if (update_signal(v))
return; /* Envelope ran out */
left = v->left_mix;
}
else
{
v->control_counter = cc - count;
while (count--)
{
*lp++ += *sp++ * left;
}
return;
}
}
}
static void mix_mystery(int32_t control_ratio, const sample_t *sp, float *lp, Voice *v, int count)
{
final_volume_t
left = v->left_mix,
right = v->right_mix;
sample_t s;
while (count--)
{
s = *sp++;
lp[0] += s * left;
lp[1] += s * right;
lp += 2;
}
}
static void mix_single(const sample_t *sp, float *lp, final_volume_t amp, int count)
{
while (count--)
{
lp[0] += *sp++ * amp;
lp += 2;
}
}
static void mix_single_left(const sample_t *sp, float *lp, Voice *v, int count)
{
mix_single(sp, lp, v->left_mix, count);
}
static void mix_single_right(const sample_t *sp, float *lp, Voice *v, int count)
{
mix_single(sp, lp + 1, v->right_mix, count);
}
static void mix_mono(const sample_t *sp, float *lp, Voice *v, int count)
{
final_volume_t
left = v->left_mix;
while (count--)
{
*lp++ += *sp++ * left;
}
}
/* Ramp a note out in c samples */
static void ramp_out(const sample_t *sp, float *lp, Voice *v, int c)
{
final_volume_t left, right, li, ri;
sample_t s = 0; /* silly warning about uninitialized s */
/* Fix by James Caldwell */
if ( c == 0 ) c = 1;
/* printf("Ramping out: left=%d, c=%d, li=%d\n", left, c, li); */
if (v->right_mix == 0) // All the way to the left
{
left = v->left_mix;
li = -(left/c);
if (li == 0) li = -1;
while (c--)
{
left += li;
if (left < 0)
return;
lp[0] += *sp++ * left;
lp += 2;
}
}
else if (v->left_mix == 0) // All the way to the right
{
right = v->right_mix;
ri = -(right/c);
if (ri == 0) ri = -1;
while (c--)
{
right += ri;
if (right < 0)
return;
s = *sp++;
lp[1] += *sp++ * right;
lp += 2;
}
}
else // Somewhere in the middle
{
left = v->left_mix;
li = -(left/c);
if (li == 0) li = -1;
right = v->right_mix;
ri = -(right/c);
if (ri == 0) ri = -1;
right = v->right_mix;
ri = -(right/c);
while (c--)
{
left += li;
right += ri;
if (left < 0)
{
if (right < 0)
{
return;
}
left = 0;
}
else if (right < 0)
{
right = 0;
}
s = *sp++;
lp[0] += s * left;
lp[1] += s * right;
lp += 2;
}
}
}
/**************** interface function ******************/
void mix_voice(Renderer *song, float *buf, Voice *v, int c)
{
int count = c;
sample_t *sp;
if (c < 0)
{
return;
}
if (v->status & VOICE_STOPPING)
{
if (count >= MAX_DIE_TIME)
count = MAX_DIE_TIME;
sp = resample_voice(song, v, &count);
ramp_out(sp, buf, v, count);
v->status = 0;
}
else
{
sp = resample_voice(song, v, &count);
if (count < 0)
{
return;
}
if (v->right_mix == 0) // All the way to the left
{
if (v->eg1.env.bUpdating || v->tremolo_phase_increment != 0)
{
mix_single_left_signal(song->control_ratio, sp, buf, v, count);
}
else
{
mix_single_left(sp, buf, v, count);
}
}
else if (v->left_mix == 0) // All the way to the right
{
if (v->eg1.env.bUpdating || v->tremolo_phase_increment != 0)
{
mix_single_right_signal(song->control_ratio, sp, buf, v, count);
}
else
{
mix_single_right(sp, buf, v, count);
}
}
else // Somewhere in the middle
{
if (v->eg1.env.bUpdating || v->tremolo_phase_increment)
{
mix_mystery_signal(song->control_ratio, sp, buf, v, count);
}
else
{
mix_mystery(song->control_ratio, sp, buf, v, count);
}
}
v->sample_count += count;
}
}
}