f8db8a3faa
Applies the clang-format style to the 2.1 branch as done for master in
5dbf1809c6
.
387 lines
9 KiB
C++
387 lines
9 KiB
C++
/*************************************************************************/
|
|
/* audio_stream_resampled.cpp */
|
|
/*************************************************************************/
|
|
/* This file is part of: */
|
|
/* GODOT ENGINE */
|
|
/* http://www.godotengine.org */
|
|
/*************************************************************************/
|
|
/* Copyright (c) 2007-2017 Juan Linietsky, Ariel Manzur. */
|
|
/* */
|
|
/* Permission is hereby granted, free of charge, to any person obtaining */
|
|
/* a copy of this software and associated documentation files (the */
|
|
/* "Software"), to deal in the Software without restriction, including */
|
|
/* without limitation the rights to use, copy, modify, merge, publish, */
|
|
/* distribute, sublicense, and/or sell copies of the Software, and to */
|
|
/* permit persons to whom the Software is furnished to do so, subject to */
|
|
/* the following conditions: */
|
|
/* */
|
|
/* The above copyright notice and this permission notice shall be */
|
|
/* included in all copies or substantial portions of the Software. */
|
|
/* */
|
|
/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */
|
|
/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */
|
|
/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/
|
|
/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */
|
|
/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */
|
|
/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */
|
|
/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */
|
|
/*************************************************************************/
|
|
#include "audio_stream_resampled.h"
|
|
#include "globals.h"
|
|
|
|
#if 0
|
|
int AudioStreamResampled::get_channel_count() const {
|
|
|
|
if (!rb)
|
|
return 0;
|
|
|
|
return channels;
|
|
}
|
|
|
|
|
|
template<int C>
|
|
uint32_t AudioStreamResampled::_resample(int32_t *p_dest,int p_todo,int32_t p_increment) {
|
|
|
|
uint32_t read=offset&MIX_FRAC_MASK;
|
|
|
|
for (int i=0;i<p_todo;i++) {
|
|
|
|
offset = (offset + p_increment)&(((1<<(rb_bits+MIX_FRAC_BITS))-1));
|
|
read+=p_increment;
|
|
uint32_t pos = offset >> MIX_FRAC_BITS;
|
|
uint32_t frac = offset & MIX_FRAC_MASK;
|
|
#ifndef FAST_AUDIO
|
|
ERR_FAIL_COND_V(pos>=rb_len,0);
|
|
#endif
|
|
uint32_t pos_next = (pos+1)&rb_mask;
|
|
//printf("rb pos %i\n",pos);
|
|
|
|
// since this is a template with a known compile time value (C), conditionals go away when compiling.
|
|
if (C==1) {
|
|
|
|
int32_t v0 = rb[pos];
|
|
int32_t v0n=rb[pos_next];
|
|
#ifndef FAST_AUDIO
|
|
v0+=(v0n-v0)*(int32_t)frac >> MIX_FRAC_BITS;
|
|
#endif
|
|
v0<<=16;
|
|
p_dest[i]=v0;
|
|
|
|
}
|
|
if (C==2) {
|
|
|
|
int32_t v0 = rb[(pos<<1)+0];
|
|
int32_t v1 = rb[(pos<<1)+1];
|
|
int32_t v0n=rb[(pos_next<<1)+0];
|
|
int32_t v1n=rb[(pos_next<<1)+1];
|
|
|
|
#ifndef FAST_AUDIO
|
|
v0+=(v0n-v0)*(int32_t)frac >> MIX_FRAC_BITS;
|
|
v1+=(v1n-v1)*(int32_t)frac >> MIX_FRAC_BITS;
|
|
#endif
|
|
v0<<=16;
|
|
v1<<=16;
|
|
p_dest[(i<<1)+0]=v0;
|
|
p_dest[(i<<1)+1]=v1;
|
|
|
|
}
|
|
|
|
if (C==4) {
|
|
|
|
int32_t v0 = rb[(pos<<2)+0];
|
|
int32_t v1 = rb[(pos<<2)+1];
|
|
int32_t v2 = rb[(pos<<2)+2];
|
|
int32_t v3 = rb[(pos<<2)+3];
|
|
int32_t v0n = rb[(pos_next<<2)+0];
|
|
int32_t v1n=rb[(pos_next<<2)+1];
|
|
int32_t v2n=rb[(pos_next<<2)+2];
|
|
int32_t v3n=rb[(pos_next<<2)+3];
|
|
|
|
#ifndef FAST_AUDIO
|
|
v0+=(v0n-v0)*(int32_t)frac >> MIX_FRAC_BITS;
|
|
v1+=(v1n-v1)*(int32_t)frac >> MIX_FRAC_BITS;
|
|
v2+=(v2n-v2)*(int32_t)frac >> MIX_FRAC_BITS;
|
|
v3+=(v3n-v3)*(int32_t)frac >> MIX_FRAC_BITS;
|
|
#endif
|
|
v0<<=16;
|
|
v1<<=16;
|
|
v2<<=16;
|
|
v3<<=16;
|
|
p_dest[(i<<2)+0]=v0;
|
|
p_dest[(i<<2)+1]=v1;
|
|
p_dest[(i<<2)+2]=v2;
|
|
p_dest[(i<<2)+3]=v3;
|
|
|
|
}
|
|
|
|
if (C==6) {
|
|
|
|
int32_t v0 = rb[(pos*6)+0];
|
|
int32_t v1 = rb[(pos*6)+1];
|
|
int32_t v2 = rb[(pos*6)+2];
|
|
int32_t v3 = rb[(pos*6)+3];
|
|
int32_t v4 = rb[(pos*6)+4];
|
|
int32_t v5 = rb[(pos*6)+5];
|
|
int32_t v0n = rb[(pos_next*6)+0];
|
|
int32_t v1n=rb[(pos_next*6)+1];
|
|
int32_t v2n=rb[(pos_next*6)+2];
|
|
int32_t v3n=rb[(pos_next*6)+3];
|
|
int32_t v4n=rb[(pos_next*6)+4];
|
|
int32_t v5n=rb[(pos_next*6)+5];
|
|
|
|
#ifndef FAST_AUDIO
|
|
v0+=(v0n-v0)*(int32_t)frac >> MIX_FRAC_BITS;
|
|
v1+=(v1n-v1)*(int32_t)frac >> MIX_FRAC_BITS;
|
|
v2+=(v2n-v2)*(int32_t)frac >> MIX_FRAC_BITS;
|
|
v3+=(v3n-v3)*(int32_t)frac >> MIX_FRAC_BITS;
|
|
v4+=(v4n-v4)*(int32_t)frac >> MIX_FRAC_BITS;
|
|
v5+=(v5n-v5)*(int32_t)frac >> MIX_FRAC_BITS;
|
|
#endif
|
|
v0<<=16;
|
|
v1<<=16;
|
|
v2<<=16;
|
|
v3<<=16;
|
|
v4<<=16;
|
|
v5<<=16;
|
|
p_dest[(i*6)+0]=v0;
|
|
p_dest[(i*6)+1]=v1;
|
|
p_dest[(i*6)+2]=v2;
|
|
p_dest[(i*6)+3]=v3;
|
|
p_dest[(i*6)+4]=v4;
|
|
p_dest[(i*6)+5]=v5;
|
|
|
|
}
|
|
|
|
|
|
}
|
|
|
|
|
|
return read>>MIX_FRAC_BITS;//rb_read_pos=offset>>MIX_FRAC_BITS;
|
|
|
|
}
|
|
|
|
|
|
bool AudioStreamResampled::mix(int32_t *p_dest, int p_frames) {
|
|
|
|
|
|
if (!rb || !_can_mix())
|
|
return false;
|
|
|
|
int write_pos_cache=rb_write_pos;
|
|
|
|
int32_t increment=(mix_rate*MIX_FRAC_LEN)/get_mix_rate();
|
|
|
|
int rb_todo;
|
|
|
|
if (write_pos_cache==rb_read_pos) {
|
|
return false; //out of buffer
|
|
|
|
} else if (rb_read_pos<write_pos_cache) {
|
|
|
|
rb_todo=write_pos_cache-rb_read_pos; //-1?
|
|
} else {
|
|
|
|
rb_todo=(rb_len-rb_read_pos)+write_pos_cache; //-1?
|
|
}
|
|
|
|
int todo = MIN( ((int64_t(rb_todo)<<MIX_FRAC_BITS)/increment)+1, p_frames );
|
|
#if 0
|
|
if (int(mix_rate)==get_mix_rate()) {
|
|
|
|
|
|
if (channels==6) {
|
|
|
|
for(int i=0;i<p_frames;i++) {
|
|
|
|
int from = ((rb_read_pos+i)&rb_mask)*6;
|
|
int to = i*6;
|
|
|
|
p_dest[from+0]=int32_t(rb[to+0])<<16;
|
|
p_dest[from+1]=int32_t(rb[to+1])<<16;
|
|
p_dest[from+2]=int32_t(rb[to+2])<<16;
|
|
p_dest[from+3]=int32_t(rb[to+3])<<16;
|
|
p_dest[from+4]=int32_t(rb[to+4])<<16;
|
|
p_dest[from+5]=int32_t(rb[to+5])<<16;
|
|
}
|
|
|
|
} else {
|
|
int len=p_frames*channels;
|
|
int from=rb_read_pos*channels;
|
|
int mask=0;
|
|
switch(channels) {
|
|
case 1: mask=rb_len-1; break;
|
|
case 2: mask=(rb_len*2)-1; break;
|
|
case 4: mask=(rb_len*4)-1; break;
|
|
}
|
|
|
|
for(int i=0;i<len;i++) {
|
|
|
|
p_dest[i]=int32_t(rb[(from+i)&mask])<<16;
|
|
}
|
|
}
|
|
|
|
rb_read_pos = (rb_read_pos+p_frames)&rb_mask;
|
|
} else
|
|
#endif
|
|
{
|
|
|
|
uint32_t read=0;
|
|
switch(channels) {
|
|
case 1: read=_resample<1>(p_dest,todo,increment); break;
|
|
case 2: read=_resample<2>(p_dest,todo,increment); break;
|
|
case 4: read=_resample<4>(p_dest,todo,increment); break;
|
|
case 6: read=_resample<6>(p_dest,todo,increment); break;
|
|
}
|
|
#if 1
|
|
//end of stream, fadeout
|
|
int remaining = p_frames-todo;
|
|
if (remaining && todo>0) {
|
|
|
|
//print_line("fadeout");
|
|
for(int c=0;c<channels;c++) {
|
|
|
|
for(int i=0;i<todo;i++) {
|
|
|
|
int32_t samp = p_dest[i*channels+c]>>8;
|
|
uint32_t mul = (todo-i) * 256 /todo;
|
|
//print_line("mul: "+itos(i)+" "+itos(mul));
|
|
p_dest[i*channels+c]=samp*mul;
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
#else
|
|
int remaining = p_frames-todo;
|
|
if (remaining && todo>0) {
|
|
|
|
|
|
for(int c=0;c<channels;c++) {
|
|
|
|
int32_t from = p_dest[(todo-1)*channels+c]>>8;
|
|
|
|
for(int i=0;i<remaining;i++) {
|
|
|
|
uint32_t mul = (remaining-i) * 256 /remaining;
|
|
p_dest[(todo+i)*channels+c]=from*mul;
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
#endif
|
|
|
|
//zero out what remains there to avoid glitches
|
|
for(int i=todo*channels;i<int(p_frames)*channels;i++) {
|
|
|
|
p_dest[i]=0;
|
|
}
|
|
|
|
if (read>rb_todo)
|
|
read=rb_todo;
|
|
|
|
rb_read_pos = (rb_read_pos+read)&rb_mask;
|
|
|
|
|
|
|
|
|
|
}
|
|
|
|
return true;
|
|
}
|
|
|
|
|
|
Error AudioStreamResampled::_setup(int p_channels,int p_mix_rate,int p_minbuff_needed) {
|
|
|
|
ERR_FAIL_COND_V(p_channels!=1 && p_channels!=2 && p_channels!=4 && p_channels!=6,ERR_INVALID_PARAMETER);
|
|
|
|
|
|
float buffering_sec = int(GLOBAL_DEF("audio/stream_buffering_ms",500))/1000.0;
|
|
int desired_rb_bits =nearest_shift(MAX(buffering_sec*p_mix_rate,p_minbuff_needed));
|
|
|
|
bool recreate=!rb;
|
|
|
|
if (rb && (uint32_t(desired_rb_bits)!=rb_bits || channels!=uint32_t(p_channels))) {
|
|
//recreate
|
|
|
|
memdelete_arr(rb);
|
|
memdelete_arr(read_buf);
|
|
recreate=true;
|
|
|
|
}
|
|
|
|
if (recreate) {
|
|
|
|
channels=p_channels;
|
|
rb_bits=desired_rb_bits;
|
|
rb_len=(1<<rb_bits);
|
|
rb_mask=rb_len-1;
|
|
rb = memnew_arr( int16_t, rb_len * p_channels );
|
|
read_buf = memnew_arr( int16_t, rb_len * p_channels );
|
|
|
|
}
|
|
|
|
mix_rate=p_mix_rate;
|
|
offset=0;
|
|
rb_read_pos=0;
|
|
rb_write_pos=0;
|
|
|
|
//avoid maybe strange noises upon load
|
|
for (int i=0;i<(rb_len*channels);i++) {
|
|
|
|
rb[i]=0;
|
|
read_buf[i]=0;
|
|
}
|
|
|
|
return OK;
|
|
|
|
}
|
|
|
|
void AudioStreamResampled::_clear() {
|
|
|
|
if (!rb)
|
|
return;
|
|
|
|
AudioServer::get_singleton()->lock();
|
|
//should be stopped at this point but just in case
|
|
if (rb) {
|
|
memdelete_arr(rb);
|
|
memdelete_arr(read_buf);
|
|
}
|
|
rb=NULL;
|
|
offset=0;
|
|
rb_read_pos=0;
|
|
rb_write_pos=0;
|
|
read_buf=NULL;
|
|
AudioServer::get_singleton()->unlock();
|
|
|
|
}
|
|
|
|
AudioStreamResampled::AudioStreamResampled() {
|
|
|
|
rb=NULL;
|
|
offset=0;
|
|
read_buf=NULL;
|
|
rb_read_pos=0;
|
|
rb_write_pos=0;
|
|
|
|
rb_bits=0;
|
|
rb_len=0;
|
|
rb_mask=0;
|
|
read_buff_len=0;
|
|
channels=0;
|
|
mix_rate=0;
|
|
|
|
}
|
|
|
|
AudioStreamResampled::~AudioStreamResampled() {
|
|
|
|
if (rb) {
|
|
memdelete_arr(rb);
|
|
memdelete_arr(read_buf);
|
|
}
|
|
|
|
}
|
|
|
|
#endif
|