forked from len0rd/rockbox
Some shifting optimizations. Working code. 50% realtime.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@6323 a1c6a512-1295-4272-9138-f99709370657
This commit is contained in:
Stepan Moskovchenko
parent
c3d0a229cc
commit
1f5fb99819
6 changed files with 238 additions and 201 deletions
|
|
@ -16,8 +16,6 @@
|
|||
*
|
||||
****************************************************************************/
|
||||
|
||||
|
||||
|
||||
extern struct plugin_api * rb;
|
||||
|
||||
struct Event * getEvent(struct Track * tr, int evNum)
|
||||
|
|
@ -178,7 +176,20 @@ int initSynth(struct MIDIfile * mf, char * filename, char * drumConfig)
|
|||
|
||||
|
||||
|
||||
inline signed short int getSample(struct GWaveform * wf, unsigned int s)
|
||||
int currentVoice IDATA_ATTR;
|
||||
struct SynthObject * so IDATA_ATTR;
|
||||
struct GWaveform * wf IDATA_ATTR;
|
||||
int s IDATA_ATTR;
|
||||
short s1 IDATA_ATTR;
|
||||
short s2 IDATA_ATTR;
|
||||
short sample IDATA_ATTR; //For synthSample
|
||||
unsigned int cpShifted IDATA_ATTR;
|
||||
|
||||
unsigned char b1 IDATA_ATTR;
|
||||
unsigned char b2 IDATA_ATTR;
|
||||
|
||||
|
||||
inline int getSample(int s)
|
||||
{
|
||||
|
||||
//16 bit samples
|
||||
|
|
@ -186,24 +197,23 @@ inline signed short int getSample(struct GWaveform * wf, unsigned int s)
|
|||
{
|
||||
|
||||
if(s<<1 >= wf->wavSize)
|
||||
{
|
||||
printf("\n!!!!! %d \t %d", s<<1, wf->wavSize);
|
||||
return 0;
|
||||
|
||||
|
||||
/*
|
||||
* Probably put the signed/unsigned and and 8-16 bit conversion
|
||||
* into the patch loader and have it run there, once.
|
||||
*/
|
||||
|
||||
}
|
||||
// signed short a = ((short *)wf->data)[s];
|
||||
|
||||
//Sign conversion moved into guspat.c
|
||||
unsigned char b1=wf->data[s<<1]; //+((wf->mode & 2) << 6);
|
||||
unsigned char b2=wf->data[(s<<1)|1]; //+((wf->mode & 2) << 6);
|
||||
b1=wf->data[s<<1]+((wf->mode & 2) << 6);
|
||||
b2=wf->data[(s<<1)|1]+((wf->mode & 2) << 6);
|
||||
return (b1 | (b2<<8)) ;
|
||||
|
||||
//Get rid of this sometime
|
||||
}
|
||||
else
|
||||
{ //8-bit samples
|
||||
//Do we even have anything 8-bit in our set?
|
||||
unsigned char b1=wf->data[s]+((wf->mode & 2) << 6);
|
||||
int b1=wf->data[s]+((wf->mode & 2) << 6);
|
||||
return b1<<8;
|
||||
}
|
||||
}
|
||||
|
|
@ -251,6 +261,12 @@ inline void setPoint(struct SynthObject * so, int pt)
|
|||
*/
|
||||
so->curRate = r<<10;
|
||||
|
||||
//Do this here because the patches assume a 44100 sampling rate
|
||||
//We've halved our sampling rate, ergo the ADSR code will be
|
||||
//called half the time. Ergo, double the rate to keep stuff
|
||||
//sounding right.
|
||||
so->curRate = so->curRate << 1;
|
||||
|
||||
|
||||
so->targetOffset = so->wf->envOffset[pt]<<(20);
|
||||
if(pt==0)
|
||||
|
|
@ -258,8 +274,6 @@ inline void setPoint(struct SynthObject * so, int pt)
|
|||
}
|
||||
|
||||
|
||||
long msi=0;
|
||||
|
||||
inline void stopVoice(struct SynthObject * so)
|
||||
{
|
||||
if(so->state == STATE_RAMPDOWN)
|
||||
|
|
@ -269,51 +283,37 @@ inline void stopVoice(struct SynthObject * so)
|
|||
|
||||
}
|
||||
|
||||
int rampDown = 0;
|
||||
|
||||
inline signed short int synthVoice(int v)
|
||||
inline signed short int synthVoice()
|
||||
{
|
||||
//Probably can combine these 2 lines into one..
|
||||
//But for now, this looks more readable
|
||||
// struct GPatch * pat = patchSet[chPat[voices[v].ch]];
|
||||
// struct GWaveform * wf = pat->waveforms[pat->noteTable[voices[v].note]];
|
||||
struct SynthObject * so = &voices[v];
|
||||
struct GWaveform * wf = so->wf;
|
||||
so = &voices[currentVoice];
|
||||
wf = so->wf;
|
||||
|
||||
signed int s;
|
||||
|
||||
if(so->state != STATE_RAMPDOWN)
|
||||
{
|
||||
if(so->loopDir==LOOPDIR_FORWARD)
|
||||
{
|
||||
so->cp += so->delta;
|
||||
}
|
||||
else
|
||||
{
|
||||
so->cp -= so->delta;
|
||||
}
|
||||
so->cp += so->delta;
|
||||
}
|
||||
|
||||
if( (so->cp>>9 >= (wf->wavSize)) && (so->state != STATE_RAMPDOWN))
|
||||
cpShifted = so->cp >> 10;
|
||||
|
||||
if( (cpShifted >= (wf->wavSize>>1)) && (so->state != STATE_RAMPDOWN))
|
||||
stopVoice(so);
|
||||
|
||||
/*
|
||||
//Original, working, no interpolation
|
||||
s=getSample(wf, (so->cp>>10));
|
||||
*/
|
||||
s2 = getSample((cpShifted)+1);
|
||||
|
||||
|
||||
|
||||
int s2=getSample(wf, (so->cp>>10)+1);
|
||||
|
||||
if((wf->mode & (LOOP_REVERSE|LOOP_PINGPONG)) && so->loopState == STATE_LOOPING && (so->cp>>10 <= (wf->startLoop>>1)))
|
||||
if((wf->mode & (LOOP_REVERSE|LOOP_PINGPONG)) && so->loopState == STATE_LOOPING && (cpShifted <= (wf->startLoop>>1)))
|
||||
{
|
||||
if(wf->mode & LOOP_REVERSE)
|
||||
{
|
||||
so->cp = (wf->endLoop)<<9;
|
||||
s2=getSample(wf, (so->cp>>10));
|
||||
cpShifted = so->cp >> 10;
|
||||
s2=getSample((cpShifted));
|
||||
} else
|
||||
{
|
||||
so->delta = -so->delta;
|
||||
so->loopDir = LOOPDIR_FORWARD;
|
||||
}
|
||||
}
|
||||
|
||||
if((wf->mode & 28) && (so->cp>>10 >= wf->endLoop>>1))
|
||||
|
|
@ -322,14 +322,21 @@ inline signed short int synthVoice(int v)
|
|||
if((wf->mode & (24)) == 0)
|
||||
{
|
||||
so->cp = (wf->startLoop)<<9;
|
||||
s2=getSample(wf, (so->cp>>10));
|
||||
cpShifted = so->cp >> 10;
|
||||
s2=getSample((cpShifted));
|
||||
} else
|
||||
{
|
||||
so->delta = -so->delta;
|
||||
so->loopDir = LOOPDIR_REVERSE;
|
||||
}
|
||||
}
|
||||
|
||||
//Better, working, linear interpolation
|
||||
int s1=getSample(wf, (so->cp>>10));
|
||||
s = s1 + ((long)((s2 - s1) * (so->cp & 1023))>>10);
|
||||
s1=getSample((cpShifted));
|
||||
s = s1 + ((signed)((s2 - s1) * (so->cp & 1023))>>10);
|
||||
|
||||
|
||||
//ADSR COMMENT WOULD GO FROM HERE.........
|
||||
|
||||
if(so->curRate == 0)
|
||||
stopVoice(so);
|
||||
|
|
@ -366,8 +373,9 @@ inline signed short int synthVoice(int v)
|
|||
so->isUsed=0; //This is OK
|
||||
|
||||
|
||||
s = s * (so->curOffset >> 22);
|
||||
s = s>>6;
|
||||
s = (s * (so->curOffset >> 22) >> 6);
|
||||
|
||||
// ............. TO HERE
|
||||
|
||||
|
||||
if(so->state == STATE_RAMPDOWN)
|
||||
|
|
@ -377,64 +385,28 @@ inline signed short int synthVoice(int v)
|
|||
so->isUsed=0;
|
||||
}
|
||||
|
||||
s = s * so->decay; s = s >> 9;
|
||||
s = s * so->decay; s = s >> 10;
|
||||
|
||||
return s*((signed short int)so->vol*(signed short int)chVol[so->ch])>>14;
|
||||
}
|
||||
|
||||
|
||||
|
||||
int mhL[16];
|
||||
int mhR[16];
|
||||
int mp=0; //Mix position, for circular array
|
||||
// Was stuff for Ghetto Lowpass Filter, now deprecated.
|
||||
|
||||
|
||||
inline void synthSample(int * mixL, int * mixR)
|
||||
{
|
||||
int a=0;
|
||||
signed long int leftMix=0, rightMix=0, sample=0;
|
||||
for(a=0; a<MAX_VOICES; a++)
|
||||
// signed int leftMix=0, rightMix=0,
|
||||
*mixL = 0;
|
||||
*mixR = 0;
|
||||
for(currentVoice=0; currentVoice<MAX_VOICES; currentVoice++)
|
||||
{
|
||||
if(voices[a].isUsed==1)
|
||||
if(voices[currentVoice].isUsed==1)
|
||||
{
|
||||
sample = synthVoice(a);
|
||||
|
||||
leftMix += (sample*chPanLeft[voices[a].ch])>>7;
|
||||
rightMix += (sample*chPanRight[voices[a].ch])>>7;
|
||||
sample = synthVoice(currentVoice);
|
||||
*mixL += (sample*chPanLeft[voices[currentVoice].ch])>>7;
|
||||
*mixR += (sample*chPanRight[voices[currentVoice].ch])>>7;
|
||||
}
|
||||
}
|
||||
|
||||
//TODO: Automatic Gain Control, anyone?
|
||||
//Or, should this be implemented on the DSP's output volume instead?
|
||||
*mixL = leftMix;
|
||||
*mixR = rightMix;
|
||||
|
||||
return; //No more ghetto lowpass filter.. linear intrpolation works well.
|
||||
|
||||
|
||||
// HACK HACK HACK
|
||||
// This is the infamous Ghetto Lowpass Filter
|
||||
// Now that I have linear interpolation, it should not be needed anymore.
|
||||
/*
|
||||
mp++;
|
||||
if(mp==4)
|
||||
mp=0;
|
||||
|
||||
mhL[mp]=leftMix;
|
||||
mhR[mp]=rightMix;
|
||||
|
||||
*mixL = 0;
|
||||
*mixR = 0;
|
||||
|
||||
|
||||
for(a=0; a<4; a++)
|
||||
{
|
||||
*mixL += mhL[a];
|
||||
*mixR += mhR[a];
|
||||
}
|
||||
*mixL = *mixL>>4;
|
||||
*mixR = *mixR>>4;
|
||||
*/
|
||||
// END HACK END HACK END HACK
|
||||
}
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue