void RTMPPublisher::SendPacketForReal(BYTE *data, UINT size, DWORD timestamp, PacketType type)
{
//OSDebugOut (TEXT("%u: SendPacketForReal (%d bytes - %08x @ %u, type %d)\n"), OSGetTime(), size, quickHash(data,size), timestamp, type);
//Log(TEXT("packet| timestamp: %u, type: %u, bytes: %u"), timestamp, (UINT)type, size);
OSEnterMutex(hDataMutex);
if(bConnected)
{
ProcessPackets();
bool bSend = bSentFirstKeyframe;
if(!bSentFirstKeyframe)
{
if(type == PacketType_VideoHighest)
{
bSend = true;
OSDebugOut(TEXT("got keyframe: %u\r\n"), OSGetTime());
}
}
if(bSend)
{
if(!bSentFirstAudio && type == PacketType_Audio)
{
timestamp = 0;
bSentFirstAudio = true;
}
totalFrames++;
if(type != PacketType_Audio)
totalVideoFrames++;
bool bAddPacket = false;
if(type >= packetWaitType)
{
if(type != PacketType_Audio)
packetWaitType = PacketType_VideoDisposable;
bAddPacket = true;
}
if(bAddPacket)
{
List<BYTE> paddedData;
paddedData.SetSize(size+RTMP_MAX_HEADER_SIZE);
mcpy(paddedData.Array()+RTMP_MAX_HEADER_SIZE, data, size);
if(!bSentFirstKeyframe)
{
DataPacket sei;
App->GetVideoEncoder()->GetSEI(sei);
paddedData.InsertArray(RTMP_MAX_HEADER_SIZE+5, sei.lpPacket, sei.size);
bSentFirstKeyframe = true;
}
currentBufferSize += paddedData.Num();
UINT droppedFrameVal = queuedPackets.Num() ? queuedPackets.Last().distanceFromDroppedFrame+1 : 10000;
UINT id = FindClosestQueueIndex(timestamp);
NetworkPacket *queuedPacket = queuedPackets.InsertNew(id);
queuedPacket->distanceFromDroppedFrame = droppedFrameVal;
queuedPacket->data.TransferFrom(paddedData);
queuedPacket->timestamp = timestamp;
queuedPacket->type = type;
}
else
{
if(type < PacketType_VideoHigh)
numBFramesDumped++;
else
numPFramesDumped++;
}
}
}
OSLeaveMutex(hDataMutex);
}
//todo: this function is an abomination, this is just disgusting. fix it.
//...seriously, this is really, really horrible. I mean this is amazingly bad.
void OBS::MainCaptureLoop()
{
int curRenderTarget = 0, curYUVTexture = 0, curCopyTexture = 0;
int copyWait = NUM_RENDER_BUFFERS-1;
bSentHeaders = false;
bFirstAudioPacket = true;
bool bLogLongFramesProfile = GlobalConfig->GetInt(TEXT("General"), TEXT("LogLongFramesProfile"), LOGLONGFRAMESDEFAULT) != 0;
float logLongFramesProfilePercentage = GlobalConfig->GetFloat(TEXT("General"), TEXT("LogLongFramesProfilePercentage"), 10.f);
Vect2 baseSize = Vect2(float(baseCX), float(baseCY));
Vect2 outputSize = Vect2(float(outputCX), float(outputCY));
Vect2 scaleSize = Vect2(float(scaleCX), float(scaleCY));
HANDLE hMatrix = yuvScalePixelShader->GetParameterByName(TEXT("yuvMat"));
HANDLE hScaleVal = yuvScalePixelShader->GetParameterByName(TEXT("baseDimensionI"));
//----------------------------------------
// x264 input buffers
int curOutBuffer = 0;
bool bUsingQSV = videoEncoder->isQSV();//GlobalConfig->GetInt(TEXT("Video Encoding"), TEXT("UseQSV")) != 0;
bUsing444 = false;
EncoderPicture lastPic;
EncoderPicture outPics[NUM_OUT_BUFFERS];
for(int i=0; i<NUM_OUT_BUFFERS; i++)
{
if(bUsingQSV)
{
outPics[i].mfxOut = new mfxFrameSurface1;
memset(outPics[i].mfxOut, 0, sizeof(mfxFrameSurface1));
mfxFrameData& data = outPics[i].mfxOut->Data;
videoEncoder->RequestBuffers(&data);
}
else
{
outPics[i].picOut = new x264_picture_t;
x264_picture_init(outPics[i].picOut);
}
}
if(bUsing444)
{
for(int i=0; i<NUM_OUT_BUFFERS; i++)
{
outPics[i].picOut->img.i_csp = X264_CSP_BGRA; //although the x264 input says BGR, x264 actually will expect packed UYV
outPics[i].picOut->img.i_plane = 1;
}
}
else
{
if(!bUsingQSV)
for(int i=0; i<NUM_OUT_BUFFERS; i++)
x264_picture_alloc(outPics[i].picOut, X264_CSP_NV12, outputCX, outputCY);
}
int bCongestionControl = AppConfig->GetInt (TEXT("Video Encoding"), TEXT("CongestionControl"), 0);
bool bDynamicBitrateSupported = App->GetVideoEncoder()->DynamicBitrateSupported();
int defaultBitRate = AppConfig->GetInt(TEXT("Video Encoding"), TEXT("MaxBitrate"), 1000);
int currentBitRate = defaultBitRate;
QWORD lastAdjustmentTime = 0;
UINT adjustmentStreamId = 0;
//std::unique_ptr<ProfilerNode> encodeThreadProfiler;
//----------------------------------------
// time/timestamp stuff
bool bWasLaggedFrame = false;
totalStreamTime = 0;
lastAudioTimestamp = 0;
//----------------------------------------
// start audio capture streams
desktopAudio->StartCapture();
if(micAudio) micAudio->StartCapture();
//----------------------------------------
// status bar/statistics stuff
DWORD fpsCounter = 0;
int numLongFrames = 0;
int numTotalFrames = 0;
bytesPerSec = 0;
captureFPS = 0;
curFramesDropped = 0;
curStrain = 0.0;
PostMessage(hwndMain, OBS_UPDATESTATUSBAR, 0, 0);
QWORD lastBytesSent[3] = {0, 0, 0};
DWORD lastFramesDropped = 0;
double bpsTime = 0.0;
double lastStrain = 0.0f;
DWORD numSecondsWaited = 0;
//----------------------------------------
// 444->420 thread data
int numThreads = MAX(OSGetTotalCores()-2, 1);
HANDLE *h420Threads = (HANDLE*)Allocate(sizeof(HANDLE)*numThreads);
Convert444Data *convertInfo = (Convert444Data*)Allocate(sizeof(Convert444Data)*numThreads);
zero(h420Threads, sizeof(HANDLE)*numThreads);
zero(convertInfo, sizeof(Convert444Data)*numThreads);
for(int i=0; i<numThreads; i++)
{
convertInfo[i].width = outputCX;
convertInfo[i].height = outputCY;
convertInfo[i].hSignalConvert = CreateEvent(NULL, FALSE, FALSE, NULL);
convertInfo[i].hSignalComplete = CreateEvent(NULL, FALSE, FALSE, NULL);
convertInfo[i].bNV12 = bUsingQSV;
convertInfo[i].numThreads = numThreads;
if(i == 0)
convertInfo[i].startY = 0;
else
convertInfo[i].startY = convertInfo[i-1].endY;
if(i == (numThreads-1))
convertInfo[i].endY = outputCY;
else
convertInfo[i].endY = ((outputCY/numThreads)*(i+1)) & 0xFFFFFFFE;
}
bool bEncode;
bool bFirstFrame = true;
bool bFirstImage = true;
bool bFirstEncode = true;
bool bUseThreaded420 = bUseMultithreadedOptimizations && (OSGetTotalCores() > 1) && !bUsing444;
List<HANDLE> completeEvents;
if(bUseThreaded420)
{
for(int i=0; i<numThreads; i++)
{
h420Threads[i] = OSCreateThread((XTHREAD)Convert444Thread, convertInfo+i);
completeEvents << convertInfo[i].hSignalComplete;
}
}
//----------------------------------------
QWORD streamTimeStart = GetQPCTimeNS();
QWORD lastStreamTime = 0;
QWORD firstFrameTimeMS = streamTimeStart/1000000;
QWORD frameLengthNS = 1000000000/fps;
while(WaitForSingleObject(hVideoEvent, INFINITE) == WAIT_OBJECT_0)
{
if (bShutdownVideoThread)
break;
QWORD renderStartTime = GetQPCTimeNS();
totalStreamTime = DWORD((renderStartTime-streamTimeStart)/1000000);
bool bRenderView = !IsIconic(hwndMain) && bRenderViewEnabled;
QWORD renderStartTimeMS = renderStartTime/1000000;
QWORD curStreamTime = latestVideoTimeNS;
if (!lastStreamTime)
lastStreamTime = curStreamTime-frameLengthNS;
QWORD frameDelta = curStreamTime-lastStreamTime;
//if (!lastStreamTime)
// lastStreamTime = renderStartTime-frameLengthNS;
//QWORD frameDelta = renderStartTime-lastStreamTime;
double fSeconds = double(frameDelta)*0.000000001;
//lastStreamTime = renderStartTime;
bool bUpdateBPS = false;
profileIn("video thread frame");
//Log(TEXT("Stream Time: %llu"), curStreamTime);
//Log(TEXT("frameDelta: %lf"), fSeconds);
//------------------------------------
if(bRequestKeyframe && keyframeWait > 0)
{
keyframeWait -= int(frameDelta);
if(keyframeWait <= 0)
{
GetVideoEncoder()->RequestKeyframe();
bRequestKeyframe = false;
}
}
if(!bPushToTalkDown && pushToTalkTimeLeft > 0)
{
pushToTalkTimeLeft -= int(frameDelta);
OSDebugOut(TEXT("time left: %d\r\n"), pushToTalkTimeLeft);
if(pushToTalkTimeLeft <= 0)
{
pushToTalkTimeLeft = 0;
bPushToTalkOn = false;
}
}
//------------------------------------
OSEnterMutex(hSceneMutex);
if (bPleaseEnableProjector)
ActuallyEnableProjector();
else if(bPleaseDisableProjector)
DisableProjector();
if(bResizeRenderView)
{
GS->ResizeView();
bResizeRenderView = false;
}
//------------------------------------
if(scene)
{
profileIn("scene->Preprocess");
scene->Preprocess();
for(UINT i=0; i<globalSources.Num(); i++)
globalSources[i].source->Preprocess();
profileOut;
scene->Tick(float(fSeconds));
for(UINT i=0; i<globalSources.Num(); i++)
globalSources[i].source->Tick(float(fSeconds));
}
//------------------------------------
QWORD curBytesSent = network->GetCurrentSentBytes();
curFramesDropped = network->NumDroppedFrames();
bpsTime += fSeconds;
if(bpsTime > 1.0f)
{
if(numSecondsWaited < 3)
++numSecondsWaited;
//bytesPerSec = DWORD(curBytesSent - lastBytesSent);
bytesPerSec = DWORD(curBytesSent - lastBytesSent[0]) / numSecondsWaited;
if(bpsTime > 2.0)
bpsTime = 0.0f;
else
bpsTime -= 1.0;
if(numSecondsWaited == 3)
{
lastBytesSent[0] = lastBytesSent[1];
lastBytesSent[1] = lastBytesSent[2];
lastBytesSent[2] = curBytesSent;
}
else
lastBytesSent[numSecondsWaited] = curBytesSent;
captureFPS = fpsCounter;
fpsCounter = 0;
bUpdateBPS = true;
}
fpsCounter++;
curStrain = network->GetPacketStrain();
EnableBlending(TRUE);
BlendFunction(GS_BLEND_SRCALPHA, GS_BLEND_INVSRCALPHA);
//------------------------------------
// render the mini render texture
LoadVertexShader(mainVertexShader);
LoadPixelShader(mainPixelShader);
SetRenderTarget(mainRenderTextures[curRenderTarget]);
Ortho(0.0f, baseSize.x, baseSize.y, 0.0f, -100.0f, 100.0f);
SetViewport(0, 0, baseSize.x, baseSize.y);
if(scene)
scene->Render();
//------------------------------------
if(bTransitioning)
{
if(!transitionTexture)
{
transitionTexture = CreateTexture(baseCX, baseCY, GS_BGRA, NULL, FALSE, TRUE);
if(transitionTexture)
{
D3D10Texture *d3dTransitionTex = static_cast<D3D10Texture*>(transitionTexture);
D3D10Texture *d3dSceneTex = static_cast<D3D10Texture*>(mainRenderTextures[lastRenderTarget]);
GetD3D()->CopyResource(d3dTransitionTex->texture, d3dSceneTex->texture);
}
else
bTransitioning = false;
}
else if(transitionAlpha >= 1.0f)
{
delete transitionTexture;
transitionTexture = NULL;
bTransitioning = false;
}
}
if(bTransitioning)
{
EnableBlending(TRUE);
transitionAlpha += float(fSeconds)*5.0f;
if(transitionAlpha > 1.0f)
transitionAlpha = 1.0f;
}
else
EnableBlending(FALSE);
//------------------------------------
// render the mini view thingy
if (bProjector) {
SetRenderTarget(projectorTexture);
Vect2 renderFrameSize, renderFrameOffset;
Vect2 projectorSize = Vect2(float(projectorWidth), float(projectorHeight));
float projectorAspect = (projectorSize.x / projectorSize.y);
float baseAspect = (baseSize.x / baseSize.y);
if (projectorAspect < baseAspect) {
float fProjectorWidth = float(projectorWidth);
renderFrameSize = Vect2(fProjectorWidth, fProjectorWidth / baseAspect);
renderFrameOffset = Vect2(0.0f, (projectorSize.y-renderFrameSize.y) * 0.5f);
} else {
float fProjectorHeight = float(projectorHeight);
renderFrameSize = Vect2(fProjectorHeight * baseAspect, fProjectorHeight);
renderFrameOffset = Vect2((projectorSize.x-renderFrameSize.x) * 0.5f, 0.0f);
}
DrawPreview(renderFrameSize, renderFrameOffset, projectorSize, curRenderTarget, Preview_Projector);
SetRenderTarget(NULL);
}
if(bRenderView)
{
// Cache
const Vect2 renderFrameSize = GetRenderFrameSize();
const Vect2 renderFrameOffset = GetRenderFrameOffset();
const Vect2 renderFrameCtrlSize = GetRenderFrameControlSize();
SetRenderTarget(NULL);
DrawPreview(renderFrameSize, renderFrameOffset, renderFrameCtrlSize, curRenderTarget,
bFullscreenMode ? Preview_Fullscreen : Preview_Standard);
//draw selections if in edit mode
if(bEditMode && !bSizeChanging)
{
if(scene) {
LoadVertexShader(solidVertexShader);
LoadPixelShader(solidPixelShader);
solidPixelShader->SetColor(solidPixelShader->GetParameter(0), 0xFF0000);
scene->RenderSelections(solidPixelShader);
}
}
}
else if(bForceRenderViewErase)
{
InvalidateRect(hwndRenderFrame, NULL, TRUE);
UpdateWindow(hwndRenderFrame);
bForceRenderViewErase = false;
}
//------------------------------------
// actual stream output
LoadVertexShader(mainVertexShader);
LoadPixelShader(yuvScalePixelShader);
Texture *yuvRenderTexture = yuvRenderTextures[curRenderTarget];
SetRenderTarget(yuvRenderTexture);
switch(colorDesc.matrix)
{
case ColorMatrix_GBR:
yuvScalePixelShader->SetMatrix(hMatrix, colorDesc.fullRange ? (float*)yuvFullMat[0] : (float*)yuvMat[0]);
break;
case ColorMatrix_YCgCo:
yuvScalePixelShader->SetMatrix(hMatrix, colorDesc.fullRange ? (float*)yuvFullMat[1] : (float*)yuvMat[1]);
break;
case ColorMatrix_BT2020NCL:
yuvScalePixelShader->SetMatrix(hMatrix, colorDesc.fullRange ? (float*)yuvFullMat[2] : (float*)yuvMat[2]);
break;
case ColorMatrix_BT709:
yuvScalePixelShader->SetMatrix(hMatrix, colorDesc.fullRange ? (float*)yuvFullMat[3] : (float*)yuvMat[3]);
break;
case ColorMatrix_SMPTE240M:
yuvScalePixelShader->SetMatrix(hMatrix, colorDesc.fullRange ? (float*)yuvFullMat[4] : (float*)yuvMat[4]);
break;
default:
yuvScalePixelShader->SetMatrix(hMatrix, colorDesc.fullRange ? (float*)yuvFullMat[5] : (float*)yuvMat[5]);
}
if(downscale < 2.01)
yuvScalePixelShader->SetVector2(hScaleVal, 1.0f/baseSize);
else if(downscale < 3.01)
yuvScalePixelShader->SetVector2(hScaleVal, 1.0f/(outputSize*3.0f));
Ortho(0.0f, outputSize.x, outputSize.y, 0.0f, -100.0f, 100.0f);
SetViewport(0.0f, 0.0f, outputSize.x, outputSize.y);
//why am I using scaleSize instead of outputSize for the texture?
//because outputSize can be trimmed by up to three pixels due to 128-bit alignment.
//using the scale function with outputSize can cause slightly inaccurate scaled images
if(bTransitioning)
{
BlendFunction(GS_BLEND_ONE, GS_BLEND_ZERO);
DrawSpriteEx(transitionTexture, 0xFFFFFFFF, 0.0f, 0.0f, scaleSize.x, scaleSize.y, 0.0f, 0.0f, 1.0f, 1.0f);
BlendFunction(GS_BLEND_FACTOR, GS_BLEND_INVFACTOR, transitionAlpha);
}
DrawSpriteEx(mainRenderTextures[curRenderTarget], 0xFFFFFFFF, 0.0f, 0.0f, outputSize.x, outputSize.y, 0.0f, 0.0f, 1.0f, 1.0f);
//------------------------------------
if (bProjector && !copyWait)
projectorSwap->Present(0, 0);
if(bRenderView && !copyWait)
static_cast<D3D10System*>(GS)->swap->Present(0, 0);
OSLeaveMutex(hSceneMutex);
//------------------------------------
// present/upload
profileIn("GPU download and conversion");
bEncode = true;
if(copyWait)
{
copyWait--;
bEncode = false;
}
else
{
//audio sometimes takes a bit to start -- do not start processing frames until audio has started capturing
if(!bRecievedFirstAudioFrame)
{
static bool bWarnedAboutNoAudio = false;
if (renderStartTimeMS-firstFrameTimeMS > 10000 && !bWarnedAboutNoAudio)
{
bWarnedAboutNoAudio = true;
//AddStreamInfo (TEXT ("WARNING: OBS is not receiving audio frames. Please check your audio devices."), StreamInfoPriority_Critical);
}
bEncode = false;
}
else if(bFirstFrame)
{
firstFrameTimestamp = lastStreamTime/1000000;
bFirstFrame = false;
}
if(!bEncode)
{
if(curYUVTexture == (NUM_RENDER_BUFFERS-1))
curYUVTexture = 0;
else
curYUVTexture++;
}
}
lastStreamTime = curStreamTime;
if(bEncode)
{
UINT prevCopyTexture = (curCopyTexture == 0) ? NUM_RENDER_BUFFERS-1 : curCopyTexture-1;
ID3D10Texture2D *copyTexture = copyTextures[curCopyTexture];
profileIn("CopyResource");
if(!bFirstEncode && bUseThreaded420)
{
WaitForMultipleObjects(completeEvents.Num(), completeEvents.Array(), TRUE, INFINITE);
copyTexture->Unmap(0);
}
D3D10Texture *d3dYUV = static_cast<D3D10Texture*>(yuvRenderTextures[curYUVTexture]);
GetD3D()->CopyResource(copyTexture, d3dYUV->texture);
profileOut;
ID3D10Texture2D *prevTexture = copyTextures[prevCopyTexture];
if(bFirstImage) //ignore the first frame
bFirstImage = false;
else
{
HRESULT result;
D3D10_MAPPED_TEXTURE2D map;
if(SUCCEEDED(result = prevTexture->Map(0, D3D10_MAP_READ, 0, &map)))
{
int prevOutBuffer = (curOutBuffer == 0) ? NUM_OUT_BUFFERS-1 : curOutBuffer-1;
int nextOutBuffer = (curOutBuffer == NUM_OUT_BUFFERS-1) ? 0 : curOutBuffer+1;
EncoderPicture &prevPicOut = outPics[prevOutBuffer];
EncoderPicture &picOut = outPics[curOutBuffer];
EncoderPicture &nextPicOut = outPics[nextOutBuffer];
if(!bUsing444)
{
profileIn("conversion to 4:2:0");
if(bUseThreaded420)
{
for(int i=0; i<numThreads; i++)
{
convertInfo[i].input = (LPBYTE)map.pData;
convertInfo[i].inPitch = map.RowPitch;
if(bUsingQSV)
{
mfxFrameData& data = nextPicOut.mfxOut->Data;
videoEncoder->RequestBuffers(&data);
convertInfo[i].outPitch = data.Pitch;
convertInfo[i].output[0] = data.Y;
convertInfo[i].output[1] = data.UV;
}
else
{
convertInfo[i].output[0] = nextPicOut.picOut->img.plane[0];
convertInfo[i].output[1] = nextPicOut.picOut->img.plane[1];
convertInfo[i].output[2] = nextPicOut.picOut->img.plane[2];
}
SetEvent(convertInfo[i].hSignalConvert);
}
if(bFirstEncode)
bFirstEncode = bEncode = false;
}
else
{
if(bUsingQSV)
{
mfxFrameData& data = picOut.mfxOut->Data;
videoEncoder->RequestBuffers(&data);
LPBYTE output[] = {data.Y, data.UV};
Convert444toNV12((LPBYTE)map.pData, outputCX, map.RowPitch, data.Pitch, outputCY, 0, outputCY, output);
}
else
Convert444toNV12((LPBYTE)map.pData, outputCX, map.RowPitch, outputCX, outputCY, 0, outputCY, picOut.picOut->img.plane);
prevTexture->Unmap(0);
}
profileOut;
}
if(bEncode)
{
//encodeThreadProfiler.reset(::new ProfilerNode(TEXT("EncodeThread"), true));
//encodeThreadProfiler->MonitorThread(hEncodeThread);
curFramePic = &picOut;
}
curOutBuffer = nextOutBuffer;
}
else
{
//We have to crash, or we end up deadlocking the thread when the convert threads are never signalled
if (result == DXGI_ERROR_DEVICE_REMOVED)
{
String message;
HRESULT reason = GetD3D()->GetDeviceRemovedReason();
switch (reason)
{
case DXGI_ERROR_DEVICE_RESET:
case DXGI_ERROR_DEVICE_HUNG:
message = TEXT("Your video card or driver froze and was reset. Please check for possible hardware / driver issues.");
break;
case DXGI_ERROR_DEVICE_REMOVED:
message = TEXT("Your video card disappeared from the system. Please check for possible hardware / driver issues.");
break;
case DXGI_ERROR_DRIVER_INTERNAL_ERROR:
message = TEXT("Your video driver reported an internal error. Please check for possible hardware / driver issues.");
break;
case DXGI_ERROR_INVALID_CALL:
message = TEXT("Your video driver reported an invalid call. Please check for possible driver issues.");
break;
default:
message = TEXT("DXGI_ERROR_DEVICE_REMOVED");
break;
}
message << TEXT(" This error can also occur if you have enabled opencl in x264 custom settings.");
CrashError (TEXT("Texture->Map failed: 0x%08x 0x%08x\r\n\r\n%s"), result, reason, message.Array());
}
else
CrashError (TEXT("Texture->Map failed: 0x%08x"), result);
}
}
if(curCopyTexture == (NUM_RENDER_BUFFERS-1))
curCopyTexture = 0;
else
curCopyTexture++;
if(curYUVTexture == (NUM_RENDER_BUFFERS-1))
curYUVTexture = 0;
else
curYUVTexture++;
if (bCongestionControl && bDynamicBitrateSupported && !bTestStream && totalStreamTime > 15000)
{
if (curStrain > 25)
{
if (renderStartTimeMS - lastAdjustmentTime > 1500)
{
if (currentBitRate > 100)
{
currentBitRate = (int)(currentBitRate * (1.0 - (curStrain / 400)));
App->GetVideoEncoder()->SetBitRate(currentBitRate, -1);
if (!adjustmentStreamId)
adjustmentStreamId = App->AddStreamInfo (FormattedString(TEXT("Congestion detected, dropping bitrate to %d kbps"), currentBitRate).Array(), StreamInfoPriority_Low);
else
App->SetStreamInfo(adjustmentStreamId, FormattedString(TEXT("Congestion detected, dropping bitrate to %d kbps"), currentBitRate).Array());
bUpdateBPS = true;
}
lastAdjustmentTime = renderStartTimeMS;
}
}
else if (currentBitRate < defaultBitRate && curStrain < 5 && lastStrain < 5)
{
if (renderStartTimeMS - lastAdjustmentTime > 5000)
{
if (currentBitRate < defaultBitRate)
{
currentBitRate += (int)(defaultBitRate * 0.05);
if (currentBitRate > defaultBitRate)
currentBitRate = defaultBitRate;
}
App->GetVideoEncoder()->SetBitRate(currentBitRate, -1);
/*if (!adjustmentStreamId)
App->AddStreamInfo (FormattedString(TEXT("Congestion clearing, raising bitrate to %d kbps"), currentBitRate).Array(), StreamInfoPriority_Low);
else
App->SetStreamInfo(adjustmentStreamId, FormattedString(TEXT("Congestion clearing, raising bitrate to %d kbps"), currentBitRate).Array());*/
App->RemoveStreamInfo(adjustmentStreamId);
adjustmentStreamId = 0;
bUpdateBPS = true;
lastAdjustmentTime = renderStartTimeMS;
}
}
}
}
lastRenderTarget = curRenderTarget;
if(curRenderTarget == (NUM_RENDER_BUFFERS-1))
curRenderTarget = 0;
else
curRenderTarget++;
if(bUpdateBPS || !CloseDouble(curStrain, lastStrain) || curFramesDropped != lastFramesDropped)
{
PostMessage(hwndMain, OBS_UPDATESTATUSBAR, 0, 0);
lastStrain = curStrain;
lastFramesDropped = curFramesDropped;
}
//------------------------------------
// we're about to sleep so we should flush the d3d command queue
profileIn("flush");
GetD3D()->Flush();
profileOut;
profileOut;
profileOut; //frame
//------------------------------------
// frame sync
QWORD renderStopTime = GetQPCTimeNS();
if(bWasLaggedFrame = (frameDelta > frameLengthNS))
{
numLongFrames++;
if(bLogLongFramesProfile && (numLongFrames/float(max(1, numTotalFrames)) * 100.) > logLongFramesProfilePercentage)
DumpLastProfileData();
}
//OSDebugOut(TEXT("Frame adjust time: %d, "), frameTimeAdjust-totalTime);
numTotalFrames++;
}
DisableProjector();
//encodeThreadProfiler.reset();
if(!bUsing444)
{
if(bUseThreaded420)
{
for(int i=0; i<numThreads; i++)
{
if(h420Threads[i])
{
convertInfo[i].bKillThread = true;
SetEvent(convertInfo[i].hSignalConvert);
OSTerminateThread(h420Threads[i], 10000);
h420Threads[i] = NULL;
}
if(convertInfo[i].hSignalConvert)
{
CloseHandle(convertInfo[i].hSignalConvert);
convertInfo[i].hSignalConvert = NULL;
}
if(convertInfo[i].hSignalComplete)
{
CloseHandle(convertInfo[i].hSignalComplete);
convertInfo[i].hSignalComplete = NULL;
}
}
if(!bFirstEncode)
{
ID3D10Texture2D *copyTexture = copyTextures[curCopyTexture];
copyTexture->Unmap(0);
}
}
if(bUsingQSV)
for(int i = 0; i < NUM_OUT_BUFFERS; i++)
delete outPics[i].mfxOut;
else
for(int i=0; i<NUM_OUT_BUFFERS; i++)
{
x264_picture_clean(outPics[i].picOut);
delete outPics[i].picOut;
}
}
Free(h420Threads);
Free(convertInfo);
Log(TEXT("Total frames rendered: %d, number of late frames: %d (%0.2f%%) (it's okay for some frames to be late)"), numTotalFrames, numLongFrames, (double(numLongFrames)/double(numTotalFrames))*100.0);
}
static void send_command(void *opaque, struct usb_msd_cbw *cbw, uint8_t *data, uint32_t len)
{
MSDState *s = (MSDState *)opaque;
DPRINTF("Command: lun=%d tag=0x%x len %zd data=0x%02x\n", cbw->lun, cbw->tag, cbw->data_len, cbw->cmd[0]);
uint32_t lba;
uint32_t xfer_len;
s->last_cmd = cbw->cmd[0];
switch(cbw->cmd[0])
{
case TEST_UNIT_READY:
//Do something?
s->result = GOOD;
set_sense(s, NO_SENSE, 0);
/* If error */
//s->result = CHECK_CONDITION;
//set_sense(s, NOT_READY, 0);
break;
case REQUEST_SENSE: //device shall keep old sense data
s->result = GOOD;
//memcpy_s(s->buf, s->data_len, s->sense_buf, sizeof(s->sense_buf)); //not on !WINDOWS
memcpy(s->buf, s->sense_buf,
/* TODO or error out instead? */
s->data_len < sizeof(s->sense_buf) ? s->data_len : sizeof(s->sense_buf));
break;
case INQUIRY:
set_sense(s, NO_SENSE, 0);
memset(s->buf, 0, sizeof(s->buf));
s->off = 0;
s->buf[0] = 0; //0x0 - direct access device, 0x1f - no fdd
s->buf[1] = 1 << 7; //removable
s->buf[3] = 1; //UFI response data format
//inq data len can be zero
strncpy((char*)&s->buf[8], "QEMU", 8); //8 bytes vendor
strncpy((char*)&s->buf[16], "USB Drive", 16); //16 bytes product
strncpy((char*)&s->buf[32], "1", 4); //4 bytes product revision
s->result = 0;
break;
case READ_CAPACITY:
long cur_tell, end_tell;
uint32_t *last_lba, *blk_len;
set_sense(s, NO_SENSE, 0);
memset(s->buf, 0, sizeof(s->buf));
s->off = 0;
cur_tell = ftell(s->hfile);
fseek(s->hfile, 0, SEEK_END);
end_tell = ftell(s->hfile);
fseek(s->hfile, cur_tell, SEEK_SET);
last_lba = (uint32_t*)&s->buf[0];
blk_len = (uint32_t*)&s->buf[4]; //in bytes
//right?
*blk_len = LBA_BLOCK_SIZE;//descriptor is currently max 64 bytes for bulk though
*last_lba = end_tell / *blk_len;
DPRINTF("read capacity lba=0x%x, block=0x%x\n", *last_lba, *blk_len);
*last_lba = bswap32(*last_lba);
*blk_len = bswap32(*blk_len);
s->result = GOOD;
break;
case READ_12:
case READ_10:
s->result = GOOD;
s->off = 0;
set_sense(s, NO_SENSE, 0);
lba = bswap32(*(uint32_t *)&cbw->cmd[2]);
if(cbw->cmd[0] == READ_10)
xfer_len = bswap16(*(uint16_t *)&cbw->cmd[7]);
else
xfer_len = bswap32(*(uint32_t *)&cbw->cmd[6]);
DPRINTF("read lba=0x%x, len=0x%x\n", lba, xfer_len * LBA_BLOCK_SIZE);
if(xfer_len == 0) //TODO nothing to do
break;
if(fseek(s->hfile, lba * LBA_BLOCK_SIZE, SEEK_SET)) {
s->result = 0x2;//PHASE_ERROR
set_sense(s, MEDIUM_ERROR, 0);
return;
}
memset(s->buf, 0, sizeof(s->buf));
//Or do actual reading in USB_MSDM_DATAIN?
//TODO probably dont set data_len to read length
if(!(s->data_len = fread(s->buf, 1, /*s->data_len*/ xfer_len * LBA_BLOCK_SIZE, s->hfile))) {
s->result = 0x2;//PHASE_ERROR
set_sense(s, MEDIUM_ERROR, 0);
}
break;
case WRITE_12:
case WRITE_10:
s->result = GOOD;//everything is fine
s->off = 0;
set_sense(s, NO_SENSE, 0);
lba = bswap32(*(uint32_t *)&cbw->cmd[2]);
if(cbw->cmd[0] == WRITE_10)
xfer_len = bswap16(*(uint16_t *)&cbw->cmd[7]);
else
xfer_len = bswap32(*(uint32_t *)&cbw->cmd[6]);
DPRINTF("write lba=0x%x, len=0x%x\n", lba, xfer_len * LBA_BLOCK_SIZE);
//if(xfer_len == 0) //nothing to do
// break;
if(fseek(s->hfile, lba * LBA_BLOCK_SIZE, SEEK_SET)) {
s->result = 0x2;//PHASE_ERROR
set_sense(s, MEDIUM_ERROR, 0);
return;
}
s->data_len = xfer_len * LBA_BLOCK_SIZE;
//Actual write comes with next command in USB_MSDM_DATAOUT
break;
default:
OSDebugOut(TEXT("usb-msd: invalid command %d\n"), cbw->cmd[0]);
s->result = 0x1; //COMMAND_FAILED
set_sense(s, ILLEGAL_REQUEST, INVALID_COMMAND_OPERATION);
break;
}
}
void PulseAudioSource::stream_read_cb (pa_stream *p, size_t nbytes, void *userdata)
{
PulseAudioSource *src = (PulseAudioSource *) userdata;
const void* padata = NULL;
if (src->mQuit)
return;
OSDebugOut("stream_read_callback %d bytes\n", nbytes);
int ret = pa_stream_peek(p, &padata, &nbytes);
OSDebugOut("pa_stream_peek %zu %s\n", nbytes, pa_strerror(ret));
if (ret != PA_OK)
return;
auto dur = std::chrono::duration_cast<ms>(hrc::now() - src->mLastGetBuffer).count();
if (src->mPaused || dur > 50000) {
ret = pa_stream_drop(p);
if (ret != PA_OK)
OSDebugOut("pa_stream_drop %s\n", pa_strerror(ret));
return;
}
{
size_t old_size = src->mQBuffer.size();
size_t nfloats = nbytes / sizeof(float);
src->mQBuffer.resize(old_size + nfloats);
memcpy(&src->mQBuffer[old_size], padata, nbytes);
//if copy succeeded, drop samples at pulse's side
ret = pa_stream_drop(p);
if (ret != PA_OK)
OSDebugOut("pa_stream_drop %s\n", pa_strerror(ret));
}
size_t resampled = static_cast<size_t>(src->mQBuffer.size() * src->mResampleRatio * src->mTimeAdjust);// * src->mSSpec.channels;
if (resampled == 0)
resampled = src->mQBuffer.size();
std::vector<float> rebuf(resampled);
SRC_DATA data;
memset(&data, 0, sizeof(SRC_DATA));
data.data_in = &src->mQBuffer[0];
data.input_frames = src->mQBuffer.size() / src->mSSpec.channels;
data.data_out = &rebuf[0];
data.output_frames = resampled / src->mSSpec.channels;
data.src_ratio = src->mResampleRatio * src->mTimeAdjust;
src_process(src->mResampler, &data);
std::lock_guard<std::mutex> lock(src->mMutex);
uint32_t len = data.output_frames_gen * src->mSSpec.channels;
size_t size = src->mResampledBuffer.size();
if (len > 0)
{
//too long, drop samples, caused by saving/loading savestates and random stutters
int sizeInMS = (((src->mResampledBuffer.size() + len) * 1000 / src->mSSpec.channels) / src->mOutputSamplesPerSec);
int threshold = src->mBuffering > 25 ? src->mBuffering : 25;
if (sizeInMS > threshold)
{
size = 0;
src->mResampledBuffer.resize(len);
}
else
src->mResampledBuffer.resize(size + len);
src_float_to_short_array(&rebuf[0], &(src->mResampledBuffer[size]), len);
}
//#if _DEBUG
// if (file && len)
// fwrite(&(src->mResampledBuffer[size]), sizeof(short), len, file);
//#endif
auto remSize = data.input_frames_used * src->mSSpec.channels;
src->mQBuffer.erase(src->mQBuffer.begin(), src->mQBuffer.begin() + remSize);
OSDebugOut("Resampler: in %ld out %ld used %ld gen %ld, rb: %zd, qb: %zd\n",
data.input_frames, data.output_frames,
data.input_frames_used, data.output_frames_gen,
src->mResampledBuffer.size(), src->mQBuffer.size());
}
bool Init()
{
int ret = 0;
mPMainLoop = pa_threaded_mainloop_new();
pa_mainloop_api *mlapi = pa_threaded_mainloop_get_api(mPMainLoop);
mPContext = pa_context_new (mlapi, "USBqemu-pulse");
ret = pa_context_connect (mPContext,
mServer,
PA_CONTEXT_NOFLAGS,
NULL
);
OSDebugOut("pa_context_connect %s\n", pa_strerror(ret));
if (ret != PA_OK)
goto error;
pa_context_set_state_callback(mPContext,
pa_context_state_cb,
&mPAready
);
pa_threaded_mainloop_start(mPMainLoop);
// wait for pa_context_state_cb
for(;;)
{
if(mPAready == 1) break;
if(mPAready == 2 || mQuit) goto error;
}
mStream = pa_stream_new(mPContext,
"USBqemu-pulse",
&mSSpec,
NULL
);
pa_stream_set_read_callback(mStream,
stream_read_cb,
this
);
// Sets individual read callback fragsize but recording itself
// still "lags" ~1sec (read_cb is called in bursts) without
// PA_STREAM_ADJUST_LATENCY
pa_buffer_attr buffer_attr;
buffer_attr.maxlength = (uint32_t) -1;
buffer_attr.tlength = (uint32_t) -1;
buffer_attr.prebuf = (uint32_t) -1;
buffer_attr.minreq = (uint32_t) -1;
buffer_attr.fragsize = pa_usec_to_bytes(mBuffering * 1000, &mSSpec);
OSDebugOut("usec_to_bytes %zu\n", buffer_attr.fragsize);
ret = pa_stream_connect_record(mStream,
mDevice.c_str(),
&buffer_attr,
PA_STREAM_ADJUST_LATENCY
);
OSDebugOut("pa_stream_connect_record %s\n", pa_strerror(ret));
if (ret != PA_OK)
goto error;
// Setup resampler
if (mResampler)
mResampler = src_delete(mResampler);
mResampler = src_new(SRC_SINC_FASTEST, mSSpec.channels, &ret);
if (!mResampler)
{
OSDebugOut("Failed to create resampler: error %08X\n", ret);
goto error;
}
mLastGetBuffer = hrc::now();
return true;
error:
Uninit();
return false;
}
//todo: this function is an abomination, this is just disgusting. fix it.
//...seriously, this is really, really horrible. I mean this is amazingly bad.
void OBS::MainCaptureLoop()
{
int curRenderTarget = 0, curYUVTexture = 0, curCopyTexture = 0;
int copyWait = NUM_RENDER_BUFFERS-1;
bSentHeaders = false;
bFirstAudioPacket = true;
Vect2 baseSize = Vect2(float(baseCX), float(baseCY));
Vect2 outputSize = Vect2(float(outputCX), float(outputCY));
Vect2 scaleSize = Vect2(float(scaleCX), float(scaleCY));
HANDLE hScaleVal = yuvScalePixelShader->GetParameterByName(TEXT("baseDimensionI"));
LPVOID nullBuff = NULL;
//----------------------------------------
// x264 input buffers
int curOutBuffer = 0;
x264_picture_t *lastPic = NULL;
x264_picture_t outPics[NUM_OUT_BUFFERS];
DWORD outTimes[NUM_OUT_BUFFERS] = {0, 0, 0};
for(int i=0; i<NUM_OUT_BUFFERS; i++)
x264_picture_init(&outPics[i]);
if(bUsing444)
{
for(int i=0; i<NUM_OUT_BUFFERS; i++)
{
outPics[i].img.i_csp = X264_CSP_BGRA; //although the x264 input says BGR, x264 actually will expect packed UYV
outPics[i].img.i_plane = 1;
}
}
else
{
for(int i=0; i<NUM_OUT_BUFFERS; i++)
x264_picture_alloc(&outPics[i], X264_CSP_I420, outputCX, outputCY);
}
//----------------------------------------
// time/timestamp stuff
LARGE_INTEGER clockFreq;
QueryPerformanceFrequency(&clockFreq);
bufferedTimes.Clear();
ctsOffsets.Clear();
#ifdef USE_100NS_TIME
QWORD streamTimeStart = GetQPCTime100NS(clockFreq.QuadPart);
QWORD frameTime100ns = 10000000/fps;
QWORD sleepTargetTime = 0;
bool bWasLaggedFrame = false;
#else
DWORD streamTimeStart = OSGetTime();
#endif
totalStreamTime = 0;
lastAudioTimestamp = 0;
latestVideoTime = firstSceneTimestamp = GetQPCTimeMS(clockFreq.QuadPart);
DWORD fpsTimeNumerator = 1000-(frameTime*fps);
DWORD fpsTimeDenominator = fps;
DWORD fpsTimeAdjust = 0;
DWORD cfrTime = 0;
DWORD cfrTimeAdjust = 0;
//----------------------------------------
// start audio capture streams
desktopAudio->StartCapture();
if(micAudio) micAudio->StartCapture();
//----------------------------------------
// status bar/statistics stuff
DWORD fpsCounter = 0;
int numLongFrames = 0;
int numTotalFrames = 0;
int numTotalDuplicatedFrames = 0;
bytesPerSec = 0;
captureFPS = 0;
curFramesDropped = 0;
curStrain = 0.0;
PostMessage(hwndMain, OBS_UPDATESTATUSBAR, 0, 0);
QWORD lastBytesSent[3] = {0, 0, 0};
DWORD lastFramesDropped = 0;
#ifdef USE_100NS_TIME
double bpsTime = 0.0;
#else
float bpsTime = 0.0f;
#endif
double lastStrain = 0.0f;
DWORD numSecondsWaited = 0;
//----------------------------------------
// 444->420 thread data
int numThreads = MAX(OSGetTotalCores()-2, 1);
HANDLE *h420Threads = (HANDLE*)Allocate(sizeof(HANDLE)*numThreads);
Convert444Data *convertInfo = (Convert444Data*)Allocate(sizeof(Convert444Data)*numThreads);
zero(h420Threads, sizeof(HANDLE)*numThreads);
zero(convertInfo, sizeof(Convert444Data)*numThreads);
for(int i=0; i<numThreads; i++)
{
convertInfo[i].width = outputCX;
convertInfo[i].height = outputCY;
convertInfo[i].hSignalConvert = CreateEvent(NULL, FALSE, FALSE, NULL);
convertInfo[i].hSignalComplete = CreateEvent(NULL, FALSE, FALSE, NULL);
if(i == 0)
convertInfo[i].startY = 0;
else
convertInfo[i].startY = convertInfo[i-1].endY;
if(i == (numThreads-1))
convertInfo[i].endY = outputCY;
else
convertInfo[i].endY = ((outputCY/numThreads)*(i+1)) & 0xFFFFFFFE;
}
bool bFirstFrame = true;
bool bFirstImage = true;
bool bFirst420Encode = true;
bool bUseThreaded420 = bUseMultithreadedOptimizations && (OSGetTotalCores() > 1) && !bUsing444;
List<HANDLE> completeEvents;
if(bUseThreaded420)
{
for(int i=0; i<numThreads; i++)
{
h420Threads[i] = OSCreateThread((XTHREAD)Convert444Thread, convertInfo+i);
completeEvents << convertInfo[i].hSignalComplete;
}
}
//----------------------------------------
QWORD curStreamTime = 0, lastStreamTime, firstFrameTime = GetQPCTimeMS(clockFreq.QuadPart);
lastStreamTime = firstFrameTime-frameTime;
bool bFirstAudioPacket = true;
while(bRunning)
{
#ifdef USE_100NS_TIME
QWORD renderStartTime = GetQPCTime100NS(clockFreq.QuadPart);
if(sleepTargetTime == 0 || bWasLaggedFrame)
sleepTargetTime = renderStartTime;
#else
DWORD renderStartTime = OSGetTime();
totalStreamTime = renderStartTime-streamTimeStart;
DWORD frameTimeAdjust = frameTime;
fpsTimeAdjust += fpsTimeNumerator;
if(fpsTimeAdjust > fpsTimeDenominator)
{
fpsTimeAdjust -= fpsTimeDenominator;
++frameTimeAdjust;
}
#endif
bool bRenderView = !IsIconic(hwndMain) && bRenderViewEnabled;
profileIn("frame");
#ifdef USE_100NS_TIME
QWORD qwTime = renderStartTime/10000;
latestVideoTime = qwTime;
QWORD frameDelta = renderStartTime-lastStreamTime;
double fSeconds = double(frameDelta)*0.0000001;
//Log(TEXT("frameDelta: %f"), fSeconds);
lastStreamTime = renderStartTime;
#else
QWORD qwTime = GetQPCTimeMS(clockFreq.QuadPart);
latestVideoTime = qwTime;
QWORD frameDelta = qwTime-lastStreamTime;
float fSeconds = float(frameDelta)*0.001f;
//Log(TEXT("frameDelta: %llu"), frameDelta);
lastStreamTime = qwTime;
#endif
//------------------------------------
if(bRequestKeyframe && keyframeWait > 0)
{
keyframeWait -= int(frameDelta);
if(keyframeWait <= 0)
{
GetVideoEncoder()->RequestKeyframe();
bRequestKeyframe = false;
}
}
if(!bPushToTalkDown && pushToTalkTimeLeft > 0)
{
pushToTalkTimeLeft -= int(frameDelta);
OSDebugOut(TEXT("time left: %d\r\n"), pushToTalkTimeLeft);
if(pushToTalkTimeLeft <= 0)
{
pushToTalkTimeLeft = 0;
bPushToTalkOn = false;
}
}
//------------------------------------
OSEnterMutex(hSceneMutex);
if(bResizeRenderView)
{
GS->ResizeView();
bResizeRenderView = false;
}
//------------------------------------
if(scene)
{
profileIn("scene->Preprocess");
scene->Preprocess();
for(UINT i=0; i<globalSources.Num(); i++)
globalSources[i].source->Preprocess();
profileOut;
scene->Tick(float(fSeconds));
for(UINT i=0; i<globalSources.Num(); i++)
globalSources[i].source->Tick(float(fSeconds));
}
//------------------------------------
QWORD curBytesSent = network->GetCurrentSentBytes();
curFramesDropped = network->NumDroppedFrames();
bool bUpdateBPS = false;
bpsTime += fSeconds;
if(bpsTime > 1.0f)
{
if(numSecondsWaited < 3)
++numSecondsWaited;
//bytesPerSec = DWORD(curBytesSent - lastBytesSent);
bytesPerSec = DWORD(curBytesSent - lastBytesSent[0]) / numSecondsWaited;
if(bpsTime > 2.0)
bpsTime = 0.0f;
else
bpsTime -= 1.0;
if(numSecondsWaited == 3)
{
lastBytesSent[0] = lastBytesSent[1];
lastBytesSent[1] = lastBytesSent[2];
lastBytesSent[2] = curBytesSent;
}
else
lastBytesSent[numSecondsWaited] = curBytesSent;
captureFPS = fpsCounter;
fpsCounter = 0;
bUpdateBPS = true;
}
fpsCounter++;
curStrain = network->GetPacketStrain();
EnableBlending(TRUE);
BlendFunction(GS_BLEND_SRCALPHA, GS_BLEND_INVSRCALPHA);
//------------------------------------
// render the mini render texture
LoadVertexShader(mainVertexShader);
LoadPixelShader(mainPixelShader);
SetRenderTarget(mainRenderTextures[curRenderTarget]);
Ortho(0.0f, baseSize.x, baseSize.y, 0.0f, -100.0f, 100.0f);
SetViewport(0, 0, baseSize.x, baseSize.y);
if(scene)
scene->Render();
//------------------------------------
if(bTransitioning)
{
if(!transitionTexture)
{
transitionTexture = CreateTexture(baseCX, baseCY, GS_BGRA, NULL, FALSE, TRUE);
if(transitionTexture)
{
D3D10Texture *d3dTransitionTex = static_cast<D3D10Texture*>(transitionTexture);
D3D10Texture *d3dSceneTex = static_cast<D3D10Texture*>(mainRenderTextures[lastRenderTarget]);
GetD3D()->CopyResource(d3dTransitionTex->texture, d3dSceneTex->texture);
}
else
bTransitioning = false;
}
else if(transitionAlpha >= 1.0f)
{
delete transitionTexture;
transitionTexture = NULL;
bTransitioning = false;
}
}
if(bTransitioning)
{
EnableBlending(TRUE);
transitionAlpha += float(fSeconds)*5.0f;
if(transitionAlpha > 1.0f)
transitionAlpha = 1.0f;
}
else
EnableBlending(FALSE);
//------------------------------------
// render the mini view thingy
if(bRenderView)
{
Vect2 renderFrameSize = Vect2(float(renderFrameWidth), float(renderFrameHeight));
SetRenderTarget(NULL);
LoadVertexShader(mainVertexShader);
LoadPixelShader(mainPixelShader);
Ortho(0.0f, renderFrameSize.x, renderFrameSize.y, 0.0f, -100.0f, 100.0f);
SetViewport(0.0f, 0.0f, renderFrameSize.x, renderFrameSize.y);
if(bTransitioning)
{
BlendFunction(GS_BLEND_ONE, GS_BLEND_ZERO);
if(renderFrameIn1To1Mode)
DrawSprite(transitionTexture, 0xFFFFFFFF, 0.0f, 0.0f, outputSize.x, outputSize.y);
else
DrawSprite(transitionTexture, 0xFFFFFFFF, 0.0f, 0.0f, renderFrameSize.x, renderFrameSize.y);
BlendFunction(GS_BLEND_FACTOR, GS_BLEND_INVFACTOR, transitionAlpha);
}
if(renderFrameIn1To1Mode)
DrawSprite(mainRenderTextures[curRenderTarget], 0xFFFFFFFF, 0.0f, 0.0f, outputSize.x, outputSize.y);
else
DrawSprite(mainRenderTextures[curRenderTarget], 0xFFFFFFFF, 0.0f, 0.0f, renderFrameSize.x, renderFrameSize.y);
Ortho(0.0f, renderFrameSize.x, renderFrameSize.y, 0.0f, -100.0f, 100.0f);
//draw selections if in edit mode
if(bEditMode && !bSizeChanging)
{
LoadVertexShader(solidVertexShader);
LoadPixelShader(solidPixelShader);
solidPixelShader->SetColor(solidPixelShader->GetParameter(0), 0xFFFF0000);
if(renderFrameIn1To1Mode)
Ortho(0.0f, renderFrameSize.x * downscale, renderFrameSize.y * downscale, 0.0f, -100.0f, 100.0f);
else
Ortho(0.0f, baseSize.x, baseSize.y, 0.0f, -100.0f, 100.0f);
if(scene)
scene->RenderSelections();
}
}
else if(bForceRenderViewErase)
{
InvalidateRect(hwndRenderFrame, NULL, TRUE);
UpdateWindow(hwndRenderFrame);
bForceRenderViewErase = false;
}
//------------------------------------
// actual stream output
LoadVertexShader(mainVertexShader);
LoadPixelShader(yuvScalePixelShader);
Texture *yuvRenderTexture = yuvRenderTextures[curRenderTarget];
SetRenderTarget(yuvRenderTexture);
yuvScalePixelShader->SetVector2(hScaleVal, 1.0f/baseSize);
Ortho(0.0f, outputSize.x, outputSize.y, 0.0f, -100.0f, 100.0f);
SetViewport(0.0f, 0.0f, outputSize.x, outputSize.y);
//why am I using scaleSize instead of outputSize for the texture?
//because outputSize can be trimmed by up to three pixels due to 128-bit alignment.
//using the scale function with outputSize can cause slightly inaccurate scaled images
if(bTransitioning)
{
BlendFunction(GS_BLEND_ONE, GS_BLEND_ZERO);
DrawSpriteEx(transitionTexture, 0xFFFFFFFF, 0.0f, 0.0f, scaleSize.x, scaleSize.y, 0.0f, 0.0f, scaleSize.x, scaleSize.y);
BlendFunction(GS_BLEND_FACTOR, GS_BLEND_INVFACTOR, transitionAlpha);
}
DrawSpriteEx(mainRenderTextures[curRenderTarget], 0xFFFFFFFF, 0.0f, 0.0f, outputSize.x, outputSize.y, 0.0f, 0.0f, outputSize.x, outputSize.y);
//------------------------------------
if(bRenderView && !copyWait)
static_cast<D3D10System*>(GS)->swap->Present(0, 0);
OSLeaveMutex(hSceneMutex);
//------------------------------------
// present/upload
profileIn("video encoding and uploading");
bool bEncode = true;
if(copyWait)
{
copyWait--;
bEncode = false;
}
else
{
//audio sometimes takes a bit to start -- do not start processing frames until audio has started capturing
if(!bRecievedFirstAudioFrame)
bEncode = false;
else if(bFirstFrame)
{
firstFrameTime = qwTime;
bFirstFrame = false;
}
if(!bEncode)
{
if(curYUVTexture == (NUM_RENDER_BUFFERS-1))
curYUVTexture = 0;
else
curYUVTexture++;
}
}
if(bEncode)
{
curStreamTime = qwTime-firstFrameTime;
UINT prevCopyTexture = (curCopyTexture == 0) ? NUM_RENDER_BUFFERS-1 : curCopyTexture-1;
ID3D10Texture2D *copyTexture = copyTextures[curCopyTexture];
profileIn("CopyResource");
if(!bFirst420Encode && bUseThreaded420)
{
WaitForMultipleObjects(completeEvents.Num(), completeEvents.Array(), TRUE, INFINITE);
copyTexture->Unmap(0);
}
D3D10Texture *d3dYUV = static_cast<D3D10Texture*>(yuvRenderTextures[curYUVTexture]);
GetD3D()->CopyResource(copyTexture, d3dYUV->texture);
profileOut;
ID3D10Texture2D *prevTexture = copyTextures[prevCopyTexture];
if(bFirstImage) //ignore the first frame
bFirstImage = false;
else
{
HRESULT result;
D3D10_MAPPED_TEXTURE2D map;
if(SUCCEEDED(result = prevTexture->Map(0, D3D10_MAP_READ, 0, &map)))
{
int prevOutBuffer = (curOutBuffer == 0) ? NUM_OUT_BUFFERS-1 : curOutBuffer-1;
int nextOutBuffer = (curOutBuffer == NUM_OUT_BUFFERS-1) ? 0 : curOutBuffer+1;
x264_picture_t &prevPicOut = outPics[prevOutBuffer];
x264_picture_t &picOut = outPics[curOutBuffer];
x264_picture_t &nextPicOut = outPics[nextOutBuffer];
if(!bUsing444)
{
profileIn("conversion to 4:2:0");
if(bUseThreaded420)
{
outTimes[nextOutBuffer] = (DWORD)curStreamTime;
for(int i=0; i<numThreads; i++)
{
convertInfo[i].input = (LPBYTE)map.pData;
convertInfo[i].pitch = map.RowPitch;
convertInfo[i].output[0] = nextPicOut.img.plane[0];
convertInfo[i].output[1] = nextPicOut.img.plane[1];
convertInfo[i].output[2] = nextPicOut.img.plane[2];
SetEvent(convertInfo[i].hSignalConvert);
}
if(bFirst420Encode)
bFirst420Encode = bEncode = false;
}
else
{
outTimes[curOutBuffer] = (DWORD)curStreamTime;
Convert444to420((LPBYTE)map.pData, outputCX, map.RowPitch, outputCY, 0, outputCY, picOut.img.plane, SSE2Available());
prevTexture->Unmap(0);
}
profileOut;
}
else
{
outTimes[curOutBuffer] = (DWORD)curStreamTime;
picOut.img.i_stride[0] = map.RowPitch;
picOut.img.plane[0] = (uint8_t*)map.pData;
}
if(bEncode)
{
DWORD curFrameTimestamp = outTimes[prevOutBuffer];
//Log(TEXT("curFrameTimestamp: %u"), curFrameTimestamp);
//------------------------------------
FrameProcessInfo frameInfo;
frameInfo.firstFrameTime = firstFrameTime;
frameInfo.prevTexture = prevTexture;
if(bUseCFR)
{
while(cfrTime < curFrameTimestamp)
{
DWORD frameTimeAdjust = frameTime;
cfrTimeAdjust += fpsTimeNumerator;
if(cfrTimeAdjust > fpsTimeDenominator)
{
cfrTimeAdjust -= fpsTimeDenominator;
++frameTimeAdjust;
}
DWORD halfTime = (frameTimeAdjust+1)/2;
x264_picture_t *nextPic = (curFrameTimestamp-cfrTime <= halfTime) ? &picOut : &prevPicOut;
//Log(TEXT("cfrTime: %u, time: %u"), cfrTime, curFrameTimestamp);
//these lines are just for counting duped frames
if(nextPic == lastPic)
++numTotalDuplicatedFrames;
else
lastPic = nextPic;
frameInfo.picOut = nextPic;
frameInfo.picOut->i_pts = cfrTime;
frameInfo.frameTimestamp = cfrTime;
ProcessFrame(frameInfo);
cfrTime += frameTimeAdjust;
//Log(TEXT("cfrTime: %u, chi frame: %u"), cfrTime, (curFrameTimestamp-cfrTime <= halfTime));
}
}
else
{
picOut.i_pts = curFrameTimestamp;
frameInfo.picOut = &picOut;
frameInfo.frameTimestamp = curFrameTimestamp;
ProcessFrame(frameInfo);
}
}
curOutBuffer = nextOutBuffer;
}
else
{
//We have to crash, or we end up deadlocking the thread when the convert threads are never signalled
CrashError (TEXT("Texture->Map failed: 0x%08x"), result);
}
}
if(curCopyTexture == (NUM_RENDER_BUFFERS-1))
curCopyTexture = 0;
else
curCopyTexture++;
if(curYUVTexture == (NUM_RENDER_BUFFERS-1))
curYUVTexture = 0;
else
curYUVTexture++;
}
lastRenderTarget = curRenderTarget;
if(curRenderTarget == (NUM_RENDER_BUFFERS-1))
curRenderTarget = 0;
else
curRenderTarget++;
if(bUpdateBPS || !CloseDouble(curStrain, lastStrain) || curFramesDropped != lastFramesDropped)
{
PostMessage(hwndMain, OBS_UPDATESTATUSBAR, 0, 0);
lastStrain = curStrain;
lastFramesDropped = curFramesDropped;
}
profileOut;
profileOut;
//------------------------------------
// get audio while sleeping or capturing
//ReleaseSemaphore(hRequestAudioEvent, 1, NULL);
//------------------------------------
// frame sync
#ifdef USE_100NS_TIME
QWORD renderStopTime = GetQPCTime100NS(clockFreq.QuadPart);
sleepTargetTime += frameTime100ns;
if(bWasLaggedFrame = (sleepTargetTime <= renderStopTime))
numLongFrames++;
else
SleepTo(clockFreq.QuadPart, sleepTargetTime);
#else
DWORD renderStopTime = OSGetTime();
DWORD totalTime = renderStopTime-renderStartTime;
if(totalTime > frameTimeAdjust)
numLongFrames++;
else if(totalTime < frameTimeAdjust)
OSSleep(frameTimeAdjust-totalTime);
#endif
//OSDebugOut(TEXT("Frame adjust time: %d, "), frameTimeAdjust-totalTime);
numTotalFrames++;
}
if(!bUsing444)
{
if(bUseThreaded420)
{
for(int i=0; i<numThreads; i++)
{
if(h420Threads[i])
{
convertInfo[i].bKillThread = true;
SetEvent(convertInfo[i].hSignalConvert);
OSTerminateThread(h420Threads[i], 10000);
h420Threads[i] = NULL;
}
if(convertInfo[i].hSignalConvert)
{
CloseHandle(convertInfo[i].hSignalConvert);
convertInfo[i].hSignalConvert = NULL;
}
if(convertInfo[i].hSignalComplete)
{
CloseHandle(convertInfo[i].hSignalComplete);
convertInfo[i].hSignalComplete = NULL;
}
}
if(!bFirst420Encode)
{
ID3D10Texture2D *copyTexture = copyTextures[curCopyTexture];
copyTexture->Unmap(0);
}
}
for(int i=0; i<NUM_OUT_BUFFERS; i++)
x264_picture_clean(&outPics[i]);
}
Free(h420Threads);
Free(convertInfo);
Log(TEXT("Total frames rendered: %d, number of frames that lagged: %d (%0.2f%%) (it's okay for some frames to lag)"), numTotalFrames, numLongFrames, (double(numLongFrames)/double(numTotalFrames))*100.0);
if(bUseCFR)
Log(TEXT("Total duplicated CFR frames: %d"), numTotalDuplicatedFrames);
}
