mirror of https://github.com/PCSX2/pcsx2.git
SPU2-X: Stretcher: Better behavior - can now maintain stable 1:1 with 30ms latency on my system if the game is light enough, and much better resilience to infinite underruns (e.g. when going from turbo to 60fps, or from turbo to 10fps, even with under 50ms latency): 1. Improved averaging on init (and after reset) 2. Added reset after few underruns. 3. automatically slightly increase target latency when pcsx2 runs slow. 4. un-staled dynamic tuning (still disabled by default, but now can be enabled and works reasonably well). 5. Lowered minimum latency from 50ms to 30ms.
Next commit will include some cleanups, and I'll probably also remove the dynamic-tunning and related support code (not a lot of code though, but.. it's currently unused). I'm also considering completely removing the old stretcher from the code, such that both dynamic-tunning and the old stretcher will only live in SVN history. Meanwhile, you can test this patch and report back here if you notice a difference. BTW, on my system I use the portaudio output since IIRC it offers the lowest latency. As before, to display stretcher messages at the console, enable underrun/overrun messages at the SPU2X debug UI (which is also available on normal builds). git-svn-id: http://pcsx2.googlecode.com/svn/trunk@5729 96395faa-99c1-11dd-bbfe-3dabce05a288
This commit is contained in:
avihal@gmail.com
parent
c2acd80236
commit
49d417b60e
|
|
@ -18,6 +18,7 @@
|
|||
#include "Global.h"
|
||||
#include "soundtouch/SoundTouch.h"
|
||||
#include <wx/datetime.h>
|
||||
#include <algorithm>
|
||||
|
||||
//Uncomment the next line to use the old time stretcher
|
||||
//#define SPU2X_USE_OLD_STRETCHER
|
||||
|
|
@ -75,12 +76,12 @@ float SndBuffer::GetStatusPct()
|
|||
//Keeping the buffer at required latency:
|
||||
// This algorithm stabilises when the actual latency is <speed>*<required_latency>. While this is just fine at 100% speed,
|
||||
// it's problematic especially for slow speeds, as the number of actual samples at the buffer gets very small on that case,
|
||||
// which may lead to underruns (or just too much latency when running very fast fast).
|
||||
// which may lead to underruns (or just too much latency when running very fast).
|
||||
//To compensate for that, the algorithm has a slowly moving compensation factor which will eventually bring the actual latency to the required one.
|
||||
//compensationDivider defines how slow this compensation changes. By default it's set to 100,
|
||||
// which will finalize the compensation after about 200 iterations.
|
||||
//
|
||||
// Note, this algorithm is intentionally simplified by not taking extreme actions at extreme scenarios (mostly underruns when speed dtops sharply),
|
||||
// Note, this algorithm is intentionally simplified by not taking extreme actions at extreme scenarios (mostly underruns when speed drops sharply),
|
||||
// and let's the overrun/underrun protections do what they should (doesn't happen much though in practice, even at big FPS variations).
|
||||
//
|
||||
// These params were tested to show good respond and stability, on all audio systems (dsound, wav, port audio, xaudio2),
|
||||
|
|
@ -89,7 +90,7 @@ float SndBuffer::GetStatusPct()
|
|||
int targetIPS=750;
|
||||
|
||||
//Dynamic tuning changes the values of the base algorithm parameters (derived from targetIPS) to adapt, in real time, to
|
||||
// diferent number of invocations/sec (mostly affects number of iterations to average).
|
||||
// different number of invocations/sec (mostly affects number of iterations to average).
|
||||
// Dynamic tuning can have a slight negative effect on the behavior of the algorithm, so it's preferred to have it off.
|
||||
//Currently it looks like it's around 750/sec on all systems when playing at 100% speed (50/60fps),
|
||||
// and proportional to that speed otherwise.
|
||||
|
|
@ -98,25 +99,45 @@ int targetIPS=750;
|
|||
//#define NEWSTRETCHER_USE_DYNAMIC_TUNING
|
||||
|
||||
|
||||
//running average can be implemented in O(1) time.
|
||||
//For the sake of simplicity, this average is calculated in O(<buffer-size>). Possibly improve later.
|
||||
//
|
||||
// Use a power of 2 numbers so the compiler can optimize 'modulo' as 'and'
|
||||
#define MAX_STRETCH_AVERAGE_LEN 128
|
||||
int STRETCH_AVERAGE_LEN=50.0 *targetIPS/750;
|
||||
//adds a value to the running average buffer, and return new running average.
|
||||
//Additional performance note: since MAX_STRETCH_AVERAGE_LEN = 128 (or any power of 2), the '%' below
|
||||
//could be replaced with a faster '&'. The compiler is highly likely to do it since all the values are unsigned.
|
||||
#define AVERAGING_BUFFER_SIZE 256U
|
||||
unsigned int AVERAGING_WINDOW = 50.0 * targetIPS/750;
|
||||
|
||||
|
||||
#define STRETCHER_RESET_THRESHOLD 5
|
||||
int gRequestStretcherReset = STRETCHER_RESET_THRESHOLD;
|
||||
//Adds a value to the running average buffer, and return the new running average.
|
||||
float addToAvg(float val){
|
||||
static float avg_fullness[MAX_STRETCH_AVERAGE_LEN]={0};
|
||||
static uint nextAvgPos=0;
|
||||
static float avg_fullness[AVERAGING_BUFFER_SIZE];
|
||||
static unsigned int nextAvgPos = 0;
|
||||
static unsigned int available = 0; // Make sure we're not averaging AVERAGING_WINDOW items if we inserted less.
|
||||
if (gRequestStretcherReset >= STRETCHER_RESET_THRESHOLD)
|
||||
available = 0;
|
||||
|
||||
avg_fullness[nextAvgPos]=val;
|
||||
nextAvgPos=(nextAvgPos+1)%MAX_STRETCH_AVERAGE_LEN;
|
||||
float sum=0;
|
||||
for(int c=0, i=(nextAvgPos+MAX_STRETCH_AVERAGE_LEN-1)%MAX_STRETCH_AVERAGE_LEN; c<STRETCH_AVERAGE_LEN; c++, i=(i+MAX_STRETCH_AVERAGE_LEN-1)%MAX_STRETCH_AVERAGE_LEN)
|
||||
sum+=avg_fullness[i];
|
||||
if (available < AVERAGING_BUFFER_SIZE)
|
||||
available++;
|
||||
|
||||
sum= (float)sum/(float)STRETCH_AVERAGE_LEN;
|
||||
return sum?sum:1;
|
||||
avg_fullness[nextAvgPos] = val;
|
||||
nextAvgPos = (nextAvgPos + 1U) % AVERAGING_BUFFER_SIZE;
|
||||
|
||||
unsigned int actualWindow = std::min(available, AVERAGING_WINDOW);
|
||||
unsigned int first = (nextAvgPos - actualWindow + AVERAGING_BUFFER_SIZE)
|
||||
% AVERAGING_BUFFER_SIZE;
|
||||
|
||||
// Possible optimization: if we know that actualWindow hasn't changed since
|
||||
// last invocation, we could calculate the running average in O(1) instead of O(N)
|
||||
// by keeping a running sum between invocations, and then
|
||||
// do "runningSum = runningSum + val - avg_fullness[(first-1)%...]" instead of the following loop.
|
||||
// Few gotchas: val overwrites first-1, handling actualWindow changes, etc.
|
||||
// However, this isn't hot code, so unless proven otherwise, we can live with unoptimized code.
|
||||
float sum = 0;
|
||||
for(unsigned int i = first; i < first + actualWindow; i++) {
|
||||
sum += avg_fullness[i % AVERAGING_BUFFER_SIZE];
|
||||
}
|
||||
sum = sum / actualWindow;
|
||||
|
||||
return sum ? sum : 1; // 1 because that's the 100% perfect speed value
|
||||
}
|
||||
|
||||
template <class T>
|
||||
|
|
@ -134,9 +155,15 @@ void SndBuffer::UpdateTempoChangeSoundTouch2()
|
|||
float baseTargetFullness=(double)targetSamplesReservoir;///(double)m_size;//0.05;
|
||||
|
||||
//state vars
|
||||
static bool inside_hysteresis=false;
|
||||
static int hys_ok_count=0;
|
||||
static float dynamicTargetFullness=baseTargetFullness;
|
||||
static bool inside_hysteresis;//=false;
|
||||
static int hys_ok_count;//=0;
|
||||
static float dynamicTargetFullness;//=baseTargetFullness;
|
||||
if (gRequestStretcherReset >= STRETCHER_RESET_THRESHOLD) {
|
||||
ConLog("______> stretch: Reset.\n");
|
||||
inside_hysteresis=false;
|
||||
hys_ok_count=0;
|
||||
dynamicTargetFullness=baseTargetFullness;
|
||||
}
|
||||
|
||||
int data = _GetApproximateDataInBuffer();
|
||||
float bufferFullness=(float)data;///(float)m_size;
|
||||
|
|
@ -149,10 +176,10 @@ void SndBuffer::UpdateTempoChangeSoundTouch2()
|
|||
wxTimeSpan delta = unow.Subtract(last);
|
||||
if( delta.GetMilliseconds()>500 ){
|
||||
int pot_targetIPS=1000.0/delta.GetMilliseconds().ToDouble()*iters;
|
||||
if(pot_targetIPS != clamp(pot_targetIPS, int((float)targetIPS/1.3f), int((float)targetIPS*1.3f)) ){
|
||||
if(!IsInRange(pot_targetIPS, int((float)targetIPS/1.3f), int((float)targetIPS*1.3f)) ){
|
||||
if(MsgOverruns()) ConLog("Stretcher: setting iters/sec from %d to %d\n", targetIPS, pot_targetIPS);
|
||||
targetIPS=pot_targetIPS;
|
||||
STRETCH_AVERAGE_LEN=clamp((int)(50.0f *(float)targetIPS/750.0f), 3, MAX_STRETCH_AVERAGE_LEN);
|
||||
AVERAGING_WINDOW=GetClamped((int)(50.0f *(float)targetIPS/750.0f), 3, (int)AVERAGING_BUFFER_SIZE);
|
||||
}
|
||||
last=unow;
|
||||
iters=0;
|
||||
|
|
@ -171,8 +198,10 @@ void SndBuffer::UpdateTempoChangeSoundTouch2()
|
|||
float avgerage = addToAvg(tempoAdjust);
|
||||
tempoAdjust = avgerage;
|
||||
tempoAdjust = GetClamped( tempoAdjust, 0.05f, 10.0f);
|
||||
|
||||
|
||||
|
||||
if (tempoAdjust < 1)
|
||||
baseTargetFullness /= sqrt(tempoAdjust); // slightly increase latency when running slow.
|
||||
|
||||
dynamicTargetFullness += (baseTargetFullness/tempoAdjust - dynamicTargetFullness)/(double)compensationDivider;
|
||||
if( IsInRange(tempoAdjust, 0.9f, 1.1f) && IsInRange( dynamicTargetFullness, baseTargetFullness*0.9f, baseTargetFullness*1.1f) )
|
||||
dynamicTargetFullness=baseTargetFullness;
|
||||
|
|
@ -206,9 +235,9 @@ void SndBuffer::UpdateTempoChangeSoundTouch2()
|
|||
wxTimeSpan delta = unow.Subtract(last);
|
||||
|
||||
if(delta.GetMilliseconds()>1000){//report buffers state and tempo adjust every second
|
||||
ConLog("buffers: %4d ms (%3.0f%%), tempo: %f, comp: %2.3f, iters: %d, (N-IPS:%d -> avg:%d, minokc:%d, div:%d)\n",
|
||||
ConLog("buffers: %4d ms (%3.0f%%), tempo: %f, comp: %2.3f, iters: %d, (N-IPS:%d -> avg:%d, minokc:%d, div:%d) reset:%d\n",
|
||||
(int)(data/48), (double)(100.0*bufferFullness/baseTargetFullness), (double)tempoAdjust, (double)(dynamicTargetFullness/baseTargetFullness), iters, (int)targetIPS
|
||||
, STRETCH_AVERAGE_LEN, hys_min_ok_count, compensationDivider
|
||||
, AVERAGING_WINDOW, hys_min_ok_count, compensationDivider, gRequestStretcherReset
|
||||
);
|
||||
last=unow;
|
||||
iters=0;
|
||||
|
|
@ -217,6 +246,8 @@ void SndBuffer::UpdateTempoChangeSoundTouch2()
|
|||
}
|
||||
|
||||
pSoundTouch->setTempo(tempoAdjust);
|
||||
if (gRequestStretcherReset >= STRETCHER_RESET_THRESHOLD)
|
||||
gRequestStretcherReset = 0;
|
||||
|
||||
return;
|
||||
}
|
||||
|
|
@ -383,12 +414,14 @@ void SndBuffer::UpdateTempoChangeAsyncMixing()
|
|||
|
||||
void SndBuffer::timeStretchUnderrun()
|
||||
{
|
||||
gRequestStretcherReset++;
|
||||
// timeStretcher failed it's job. We need to slow down the audio some.
|
||||
|
||||
cTempo -= (cTempo * 0.12f);
|
||||
eTempo -= (eTempo * 0.30f);
|
||||
if( eTempo < 0.1f ) eTempo = 0.1f;
|
||||
pSoundTouch->setTempo( eTempo );
|
||||
// pSoundTouch->setTempo( eTempo );
|
||||
//pSoundTouch->setTempoChange(-30); // temporary (until stretcher is called) slow down
|
||||
}
|
||||
|
||||
s32 SndBuffer::timeStretchOverrun()
|
||||
|
|
@ -398,11 +431,12 @@ s32 SndBuffer::timeStretchOverrun()
|
|||
cTempo += cTempo * 0.12f;
|
||||
eTempo += eTempo * 0.40f;
|
||||
if( eTempo > 7.5f ) eTempo = 7.5f;
|
||||
pSoundTouch->setTempo( eTempo );
|
||||
//pSoundTouch->setTempo( eTempo );
|
||||
//pSoundTouch->setTempoChange(30);// temporary (until stretcher is called) speed up
|
||||
|
||||
// Throw out just a little bit (two packets worth) to help
|
||||
// give the TS some room to work:
|
||||
|
||||
gRequestStretcherReset++;
|
||||
return SndOutPacketSize*2;
|
||||
}
|
||||
|
||||
|
|
|
|||
|
|
@ -25,7 +25,7 @@ static const int LATENCY_MAX = 750;
|
|||
#endif
|
||||
|
||||
static const int LATENCY_MIN = 3;
|
||||
static const int LATENCY_MIN_TS = 50;
|
||||
static const int LATENCY_MIN_TS = 30;
|
||||
|
||||
// MIXING
|
||||
int Interpolation = 4;
|
||||
|
|
|
|||
Loading…
Reference in New Issue