gzdoom-gles/libraries/timidity/resample.cpp
Christoph Oelckers 2ddc9ee1f3 - made the Timidity(GUS) device a separate library.
# Conflicts:
#	src/CMakeLists.txt
2020-01-05 01:22:14 +01:00

615 lines
12 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
resample.c
*/
#include <math.h>
#include <stdio.h>
#include <stdlib.h>
#include "timidity.h"
#include "common.h"
#include "instrum.h"
#include "playmidi.h"
namespace Timidity
{
#define RESAMPLATION {\
int o = ofs >> FRACTION_BITS, m = ofs & FRACTION_MASK; \
*dest++ = src[o] + (src[o + 1] - src[o]) * m / (1 << FRACTION_BITS);\
}
#define FINALINTERP if (ofs == le) *dest++ = src[ofs >> FRACTION_BITS];
/* So it isn't interpolation. At least it's final. */
/*************** resampling with fixed increment *****************/
static sample_t *rs_plain(sample_t *resample_buffer, Voice *v, int *countptr)
{
/* Play sample until end, then free the voice. */
const sample_t
*src = v->sample->data;
sample_t
*dest = resample_buffer;
int
ofs = v->sample_offset,
incr = v->sample_increment,
le = v->sample->data_length,
count = *countptr;
int i;
if (incr < 0) incr = -incr; /* In case we're coming out of a bidir loop */
/* Precalc how many times we should go through the loop.
NOTE: Assumes that incr > 0 and that ofs <= le */
i = (le - ofs) / incr + 1;
if (i > count)
{
i = count;
count = 0;
}
else
{
count -= i;
}
while (i--)
{
RESAMPLATION;
ofs += incr;
}
if (ofs >= le)
{
FINALINTERP;
v->status = 0;
*countptr -= count + 1;
}
v->sample_offset = ofs; /* Update offset */
return resample_buffer;
}
static sample_t *rs_loop(sample_t *resample_buffer, Voice *vp, int count)
{
/* Play sample until end-of-loop, skip back and continue. */
int
ofs = vp->sample_offset,
incr = vp->sample_increment,
le = vp->sample->loop_end,
ll = le - vp->sample->loop_start;
sample_t
*dest = resample_buffer;
const sample_t
*src = vp->sample->data;
int i;
while (count)
{
if (ofs >= le)
/* NOTE: Assumes that ll > incr and that incr > 0. */
ofs -= ll;
/* Precalc how many times we should go through the loop */
i = (le - ofs) / incr + 1;
if (i > count)
{
i = count;
count = 0;
}
else
{
count -= i;
}
while (i--)
{
RESAMPLATION;
ofs += incr;
}
}
vp->sample_offset=ofs; /* Update offset */
return resample_buffer;
}
static sample_t *rs_bidir(sample_t *resample_buffer, Voice *vp, int count)
{
int
ofs = vp->sample_offset,
incr = vp->sample_increment,
le = vp->sample->loop_end,
ls = vp->sample->loop_start;
sample_t
*dest = resample_buffer;
const sample_t
*src = vp->sample->data;
int
le2 = le << 1,
ls2 = ls << 1,
i;
/* Play normally until inside the loop region */
if (ofs <= ls)
{
/* NOTE: Assumes that incr > 0, which is NOT always the case
when doing bidirectional looping. I have yet to see a case
where both ofs <= ls AND incr < 0, however. */
i = (ls - ofs) / incr + 1;
if (i > count)
{
i = count;
count = 0;
}
else
{
count -= i;
}
while (i--)
{
RESAMPLATION;
ofs += incr;
}
}
/* Then do the bidirectional looping */
while(count)
{
/* Precalc how many times we should go through the loop */
i = ((incr > 0 ? le : ls) - ofs) / incr + 1;
if (i > count)
{
i = count;
count = 0;
}
else
{
count -= i;
}
while (i--)
{
RESAMPLATION;
ofs += incr;
}
if (ofs >= le)
{
/* fold the overshoot back in */
ofs = le2 - ofs;
incr *= -1;
}
else if (ofs <= ls)
{
ofs = ls2 - ofs;
incr *= -1;
}
}
vp->sample_increment = incr;
vp->sample_offset = ofs; /* Update offset */
return resample_buffer;
}
/*********************** vibrato versions ***************************/
/* We only need to compute one half of the vibrato sine cycle */
static int vib_phase_to_inc_ptr(int phase)
{
if (phase < VIBRATO_SAMPLE_INCREMENTS / 2)
return VIBRATO_SAMPLE_INCREMENTS / 2 - 1 - phase;
else if (phase >= VIBRATO_SAMPLE_INCREMENTS * 3 / 2)
return VIBRATO_SAMPLE_INCREMENTS * 5 / 2 - 1 - phase;
else
return phase - VIBRATO_SAMPLE_INCREMENTS / 2;
}
static int update_vibrato(float output_rate, Voice *vp, int sign)
{
int depth;
int phase;
double a, pb;
if (vp->vibrato_phase++ >= 2 * VIBRATO_SAMPLE_INCREMENTS - 1)
vp->vibrato_phase = 0;
phase = vib_phase_to_inc_ptr(vp->vibrato_phase);
if (vp->vibrato_sample_increment[phase])
{
if (sign)
return -vp->vibrato_sample_increment[phase];
else
return vp->vibrato_sample_increment[phase];
}
/* Need to compute this sample increment. */
depth = vp->sample->vibrato_depth << 7;
if (vp->vibrato_sweep != 0)
{
/* Need to update sweep */
vp->vibrato_sweep_position += vp->vibrato_sweep;
if (vp->vibrato_sweep_position >= (1<<SWEEP_SHIFT))
vp->vibrato_sweep=0;
else
{
/* Adjust depth */
depth *= vp->vibrato_sweep_position;
depth >>= SWEEP_SHIFT;
}
}
a = FSCALE(((double)(vp->sample->sample_rate) * vp->frequency) /
((double)(vp->sample->root_freq) * output_rate),
FRACTION_BITS);
pb = (sine(vp->vibrato_phase * (1.0/(2*VIBRATO_SAMPLE_INCREMENTS)))
* (double)(depth) * VIBRATO_AMPLITUDE_TUNING);
a *= pow(2.0, pb / (8192 * 12.f));
/* If the sweep's over, we can store the newly computed sample_increment */
if (!vp->vibrato_sweep)
vp->vibrato_sample_increment[phase] = (int) a;
if (sign)
a = -a; /* need to preserve the loop direction */
return (int) a;
}
static sample_t *rs_vib_plain(sample_t *resample_buffer, float rate, Voice *vp, int *countptr)
{
/* Play sample until end, then free the voice. */
sample_t
*dest = resample_buffer;
const sample_t
*src = vp->sample->data;
int
le = vp->sample->data_length,
ofs = vp->sample_offset,
incr = vp->sample_increment,
count = *countptr;
int
cc = vp->vibrato_control_counter;
/* This has never been tested */
if (incr < 0) incr = -incr; /* In case we're coming out of a bidir loop */
while (count--)
{
if (!cc--)
{
cc = vp->vibrato_control_ratio;
incr = update_vibrato(rate, vp, 0);
}
RESAMPLATION;
ofs += incr;
if (ofs >= le)
{
FINALINTERP;
vp->status = 0;
*countptr -= count+1;
break;
}
}
vp->vibrato_control_counter = cc;
vp->sample_increment = incr;
vp->sample_offset = ofs; /* Update offset */
return resample_buffer;
}
static sample_t *rs_vib_loop(sample_t *resample_buffer, float rate, Voice *vp, int count)
{
/* Play sample until end-of-loop, skip back and continue. */
int
ofs = vp->sample_offset,
incr = vp->sample_increment,
le = vp->sample->loop_end,
ll = le - vp->sample->loop_start;
sample_t
*dest = resample_buffer;
const sample_t
*src = vp->sample->data;
int
cc = vp->vibrato_control_counter;
int i;
int
vibflag=0;
while (count)
{
/* Hopefully the loop is longer than an increment */
if (ofs >= le)
ofs -= ll;
/* Precalc how many times to go through the loop, taking
the vibrato control ratio into account this time. */
i = (le - ofs) / incr + 1;
if (i > count) i = count;
if (i > cc)
{
i = cc;
vibflag = 1;
}
else
{
cc -= i;
}
count -= i;
while (i--)
{
RESAMPLATION;
ofs += incr;
}
if (vibflag)
{
cc = vp->vibrato_control_ratio;
incr = update_vibrato(rate, vp, 0);
vibflag = 0;
}
}
vp->vibrato_control_counter = cc;
vp->sample_increment = incr;
vp->sample_offset = ofs; /* Update offset */
return resample_buffer;
}
static sample_t *rs_vib_bidir(sample_t *resample_buffer, float rate, Voice *vp, int count)
{
int
ofs = vp->sample_offset,
incr = vp->sample_increment,
le = vp->sample->loop_end,
ls = vp->sample->loop_start;
sample_t
*dest = resample_buffer;
const sample_t
*src = vp->sample->data;
int
cc = vp->vibrato_control_counter;
int
le2 = le << 1,
ls2 = ls << 1,
i;
int
vibflag = 0;
/* Play normally until inside the loop region */
while (count && (ofs <= ls))
{
i = (ls - ofs) / incr + 1;
if (i > count)
{
i = count;
}
if (i > cc)
{
i = cc;
vibflag = 1;
}
else
{
cc -= i;
}
count -= i;
while (i--)
{
RESAMPLATION;
ofs += incr;
}
if (vibflag)
{
cc = vp->vibrato_control_ratio;
incr = update_vibrato(rate, vp, 0);
vibflag = 0;
}
}
/* Then do the bidirectional looping */
while (count)
{
/* Precalc how many times we should go through the loop */
i = ((incr > 0 ? le : ls) - ofs) / incr + 1;
if(i > count)
{
i = count;
}
if(i > cc)
{
i = cc;
vibflag = 1;
}
else
{
cc -= i;
}
count -= i;
while (i--)
{
RESAMPLATION;
ofs += incr;
}
if (vibflag)
{
cc = vp->vibrato_control_ratio;
incr = update_vibrato(rate, vp, (incr < 0));
vibflag = 0;
}
if (ofs >= le)
{
/* fold the overshoot back in */
ofs = le2 - ofs;
incr *= -1;
}
else if (ofs <= ls)
{
ofs = ls2 - ofs;
incr *= -1;
}
}
vp->vibrato_control_counter = cc;
vp->sample_increment = incr;
vp->sample_offset = ofs; /* Update offset */
return resample_buffer;
}
sample_t *resample_voice(Renderer *song, Voice *vp, int *countptr)
{
int ofs;
uint16_t modes;
if (vp->sample->sample_rate == 0)
{
/* Pre-resampled data -- just update the offset and check if
we're out of data. */
ofs = vp->sample_offset >> FRACTION_BITS; /* Kind of silly to use FRACTION_BITS here... */
if (*countptr >= (vp->sample->data_length >> FRACTION_BITS) - ofs)
{
/* Note finished. Free the voice. */
vp->status = 0;
/* Let the caller know how much data we had left */
*countptr = (vp->sample->data_length >> FRACTION_BITS) - ofs;
}
else
{
vp->sample_offset += *countptr << FRACTION_BITS;
}
return vp->sample->data + ofs;
}
/* Need to resample. Use the proper function. */
modes = vp->sample->modes;
if (vp->status & VOICE_LPE)
{
if (vp->sample->loop_end - vp->sample->loop_start < 2)
{ // Loop is too short; turn it off.
vp->status &= ~VOICE_LPE;
}
}
if (vp->vibrato_control_ratio)
{
if (vp->status & VOICE_LPE)
{
if (modes & PATCH_BIDIR)
return rs_vib_bidir(song->resample_buffer, song->rate, vp, *countptr);
else
return rs_vib_loop(song->resample_buffer, song->rate, vp, *countptr);
}
else
{
return rs_vib_plain(song->resample_buffer, song->rate, vp, countptr);
}
}
else
{
if (vp->status & VOICE_LPE)
{
if (modes & PATCH_BIDIR)
return rs_bidir(song->resample_buffer, vp, *countptr);
else
return rs_loop(song->resample_buffer, vp, *countptr);
}
else
{
return rs_plain(song->resample_buffer, vp, countptr);
}
}
}
void pre_resample(Renderer *song, Sample *sp)
{
double a, xdiff;
int incr, ofs, newlen, count;
sample_t *newdata, *dest, *src = sp->data;
sample_t v1, v2, v3, v4, *vptr;
static const char note_name[12][3] =
{
"C", "C#", "D", "D#", "E", "F", "F#", "G", "G#", "A", "A#", "B"
};
if (sp->scale_factor != 0)
return;
a = (sp->sample_rate * note_to_freq(sp->scale_note)) / (sp->root_freq * song->rate);
if (a <= 0)
return;
newlen = (int)(sp->data_length / a);
if (newlen < 0 || (newlen >> FRACTION_BITS) > MAX_SAMPLE_SIZE)
return;
count = newlen >> FRACTION_BITS;
dest = newdata = (sample_t *)safe_malloc(count * sizeof(float));
ofs = incr = (sp->data_length - (1 << FRACTION_BITS)) / count;
if (--count)
*dest++ = src[0];
/* Since we're pre-processing and this doesn't have to be done in
real-time, we go ahead and do the full sliding cubic interpolation. */
while (--count)
{
vptr = src + (ofs >> FRACTION_BITS);
v1 = (vptr == src) ? *vptr : *(vptr - 1);
v2 = *vptr;
v3 = *(vptr + 1);
v4 = *(vptr + 2);
xdiff = FSCALENEG(ofs & FRACTION_MASK, FRACTION_BITS);
*dest++ = sample_t(v2 + (xdiff / 6.0) * (-2 * v1 - 3 * v2 + 6 * v3 - v4 +
xdiff * (3 * (v1 - 2 * v2 + v3) + xdiff * (-v1 + 3 * (v2 - v3) + v4))));
ofs += incr;
}
if (ofs & FRACTION_MASK)
{
RESAMPLATION
}
else
{
*dest++ = src[ofs >> FRACTION_BITS];
}
sp->data_length = newlen;
sp->loop_start = int(sp->loop_start / a);
sp->loop_end = int(sp->loop_end / a);
free(sp->data);
sp->data = newdata;
sp->sample_rate = 0;
}
}