void OBS::DrawPreview(const Vect2 &renderFrameSize, const Vect2 &renderFrameOffset, const Vect2 &renderFrameCtrlSize, int curRenderTarget, PreviewDrawType type)
{
LoadVertexShader(mainVertexShader);
LoadPixelShader(mainPixelShader);
Ortho(0.0f, renderFrameCtrlSize.x, renderFrameCtrlSize.y, 0.0f, -100.0f, 100.0f);
if(type != Preview_Projector
&& (renderFrameCtrlSize.x != oldRenderFrameCtrlWidth
|| renderFrameCtrlSize.y != oldRenderFrameCtrlHeight))
{
// User is drag resizing the window. We don't recreate the swap chains so our coordinates are wrong
SetViewport(0.0f, 0.0f, (float)oldRenderFrameCtrlWidth, (float)oldRenderFrameCtrlHeight);
}
else
SetViewport(0.0f, 0.0f, renderFrameCtrlSize.x, renderFrameCtrlSize.y);
// Draw background (Black if fullscreen/projector, window colour otherwise)
if(type == Preview_Fullscreen || type == Preview_Projector)
ClearColorBuffer(0x000000);
else
ClearColorBuffer(GetSysColor(COLOR_BTNFACE));
if(bTransitioning)
{
BlendFunction(GS_BLEND_ONE, GS_BLEND_ZERO);
DrawSprite(transitionTexture, 0xFFFFFFFF,
renderFrameOffset.x, renderFrameOffset.y,
renderFrameOffset.x + renderFrameSize.x, renderFrameOffset.y + renderFrameSize.y);
BlendFunction(GS_BLEND_FACTOR, GS_BLEND_INVFACTOR, transitionAlpha);
}
DrawSprite(mainRenderTextures[curRenderTarget], 0xFFFFFFFF,
renderFrameOffset.x, renderFrameOffset.y,
renderFrameOffset.x + renderFrameSize.x, renderFrameOffset.y + renderFrameSize.y);
}
void BezierController::MakeSegment(float time,
float inc,
const PathPoint& p1,
const PathPoint& p2,
const PathPoint& p3,
const PathPoint& p4)
{
osg::Vec3 pos, tangent;
for (int i = 0; i < 3; ++i)
{
pos[i] = (BlendFunction(inc,0) * p1.GetPosition()[i]) +
(BlendFunction(inc,1) * p2.GetPosition()[i]) +
(BlendFunction(inc,2) * p3.GetPosition()[i]) +
(BlendFunction(inc,3) * p4.GetPosition()[i]);
//tangent[i] = (TangentFunction(inc,1) * p2.GetPosition()[i]) + (TangentFunction(inc,2) * p3.GetPosition()[i]) + (TangentFunction(inc,3) * p4.GetPosition()[i]);
}
osg::Quat quat;
quat.slerp(1.0 - inc, p4.GetOrientation(), p1.GetOrientation());
PathData pd;
pd.mTime = time;
pd.mPoint.SetPosition(pos);
pd.mPoint.SetOrientation(quat);
mPath.push_back(pd);
}
void TexturedLine::Render()
{
if(!lineTexture) return;
Vect lineDir = (pointB-pointA).Norm();
float halfWidth = width*0.5f;
Camera *viewCam = level->GetCurrentCamera();
Vect camPos = viewCam->GetWorldPos();
Vect crossLine;
DepthWriteEnable(FALSE);
EnableBlending(TRUE);
if(lineTexture->HasAlpha())
BlendFunction(GS_BLEND_SRCALPHA, GS_BLEND_INVSRCALPHA);
else
BlendFunction(GS_BLEND_ONE, GS_BLEND_ONE);
LoadTexture(lineTexture);
LoadDefault3DSampler();
MatrixPush();
MatrixTranslate(GetWorldPos());
MatrixRotate(GetWorldRot());
RenderStart();
Vect endAdjust = (lineDir*halfWidth);
Vect backEnd = (pointA-endAdjust);
Vect frontEnd = (pointB+endAdjust);
Color(color);
crossLine = (camPos-backEnd).Cross(lineDir).Norm() * halfWidth;
Vertex(backEnd+crossLine); TexCoord(1.0f, 0.0f);
Vertex(backEnd-crossLine); TexCoord(0.0f, 0.0f);
crossLine = (camPos-pointA).Cross(lineDir).Norm() * halfWidth;
Vertex(pointA+crossLine); TexCoord(1.0f, 0.5f);
Vertex(pointA-crossLine); TexCoord(0.0f, 0.5f);
crossLine = (camPos-pointB).Cross(lineDir).Norm() * halfWidth;
Vertex(pointB+crossLine); TexCoord(1.0f, 0.5f);
Vertex(pointB-crossLine); TexCoord(0.0f, 0.5f);
crossLine = (camPos-frontEnd).Cross(lineDir).Norm() * halfWidth;
Vertex(frontEnd+crossLine); TexCoord(1.0f, 1.0f);
Vertex(frontEnd-crossLine); TexCoord(0.0f, 1.0f);
RenderStop(GS_TRIANGLESTRIP);
MatrixPop();
LoadTexture(NULL);
}
static void platformApply(unsigned char* sourcePixelA, unsigned char* sourcePixelB,
unsigned char* destinationPixel, unsigned pixelArrayLength)
{
unsigned len = pixelArrayLength / 4;
for (unsigned pixelOffset = 0; pixelOffset < len; pixelOffset++) {
unsigned char alphaA = sourcePixelA[3];
unsigned char alphaB = sourcePixelB[3];
destinationPixel[0] = BlendFunction(sourcePixelA[0], sourcePixelB[0], alphaA, alphaB);
destinationPixel[1] = BlendFunction(sourcePixelA[1], sourcePixelB[1], alphaA, alphaB);
destinationPixel[2] = BlendFunction(sourcePixelA[2], sourcePixelB[2], alphaA, alphaB);
destinationPixel[3] = 255 - fastDivideBy255((255 - alphaA) * (255 - alphaB));
sourcePixelA += 4;
sourcePixelB += 4;
destinationPixel += 4;
}
}
void GraphicsCaptureSource::Render(const Vect2 &pos, const Vect2 &size)
{
if(capture)
{
Shader *lastShader = GetCurrentPixelShader();
float fGamma = float(-(gamma-100) + 100) * 0.01f;
LoadPixelShader(drawShader);
HANDLE hGamma = drawShader->GetParameterByName(TEXT("gamma"));
if(hGamma)
drawShader->SetFloat(hGamma, fGamma);
//----------------------------------------------------------
// capture mouse
bMouseCaptured = false;
if(bCaptureMouse)
{
CURSORINFO ci;
zero(&ci, sizeof(ci));
ci.cbSize = sizeof(ci);
if(GetCursorInfo(&ci) && hwndCapture)
{
mcpy(&cursorPos, &ci.ptScreenPos, sizeof(cursorPos));
ScreenToClient(hwndCapture, &cursorPos);
if(ci.flags & CURSOR_SHOWING)
{
if(ci.hCursor == hCurrentCursor)
bMouseCaptured = true;
else
{
HICON hIcon = CopyIcon(ci.hCursor);
hCurrentCursor = ci.hCursor;
delete cursorTexture;
cursorTexture = NULL;
if(hIcon)
{
ICONINFO ii;
if(GetIconInfo(hIcon, &ii))
{
xHotspot = int(ii.xHotspot);
yHotspot = int(ii.yHotspot);
UINT width, height;
LPBYTE lpData = GetCursorData(hIcon, ii, width, height);
if(lpData)
{
cursorTexture = CreateTexture(width, height, GS_BGRA, lpData, FALSE);
if(cursorTexture)
bMouseCaptured = true;
Free(lpData);
}
DeleteObject(ii.hbmColor);
DeleteObject(ii.hbmMask);
}
DestroyIcon(hIcon);
}
}
}
}
}
//----------------------------------------------------------
// game texture
Texture *tex = capture->LockTexture();
Vect2 texPos = Vect2(0.0f, 0.0f);
Vect2 texStretch = Vect2(1.0f, 1.0f);
if(tex)
{
Vect2 texSize = Vect2(float(tex->Width()), float(tex->Height()));
Vect2 totalSize = API->GetBaseSize();
Vect2 center = totalSize*0.5f;
BlendFunction(GS_BLEND_ONE, GS_BLEND_ZERO);
if(bStretch)
{
if(bIgnoreAspect)
texStretch *= totalSize;
else
{
float multiplyVal = (texSize.y * (totalSize.x / texSize.x) > totalSize.y) ? totalSize.y / texSize.y : multiplyVal = totalSize.x / texSize.x;
texStretch *= texSize*multiplyVal;
texPos = center-(texStretch*0.5f);
}
}
else
{
texStretch *= texSize;
texPos = center-(texStretch*0.5f);
}
Vect2 sizeAdjust = size/totalSize;
texPos *= sizeAdjust;
texPos += pos;
texStretch *= sizeAdjust;
RoundVect2(texPos);
RoundVect2(texSize);
if(bFlip)
DrawSprite(tex, 0xFFFFFFFF, texPos.x, texPos.y+texStretch.y, texPos.x+texStretch.x, texPos.y);
else
DrawSprite(tex, 0xFFFFFFFF, texPos.x, texPos.y, texPos.x+texStretch.x, texPos.y+texStretch.y);
capture->UnlockTexture();
BlendFunction(GS_BLEND_SRCALPHA, GS_BLEND_INVSRCALPHA);
//----------------------------------------------------------
// draw mouse
if(bMouseCaptured && cursorTexture && GetForegroundWindow() == hwndCapture)
{
Vect2 newCursorPos = Vect2(float(cursorPos.x-xHotspot), float(cursorPos.y-xHotspot));
Vect2 newCursorSize = Vect2(float(cursorTexture->Width()), float(cursorTexture->Height()));
newCursorPos /= texSize;
newCursorSize /= texSize;
newCursorPos *= texStretch;
newCursorPos += texPos;
newCursorSize *= texStretch;
bool bInvertCursor = false;
if(invertShader)
{
if(bInvertCursor = ((GetAsyncKeyState(VK_LBUTTON) & 0x8000) != 0 || (GetAsyncKeyState(VK_RBUTTON) & 0x8000) != 0))
LoadPixelShader(invertShader);
}
DrawSprite(cursorTexture, 0xFFFFFFFF, newCursorPos.x, newCursorPos.y, newCursorPos.x+newCursorSize.x, newCursorPos.y+newCursorSize.y);
}
}
if(lastShader)
LoadPixelShader(lastShader);
}
}
//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(!pushToTalkDown && 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 = 0;
if (network) {
curBytesSent = network->GetCurrentSentBytes();
curFramesDropped = network->NumDroppedFrames();
} else if (numSecondsWaited) {
//reset stats if the network disappears
bytesPerSec = 0;
bpsTime = 0;
numSecondsWaited = 0;
curBytesSent = 0;
zero(lastBytesSent, sizeof(lastBytesSent));
}
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++;
if(network) 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, (numTotalFrames > 0) ? (double(numLongFrames)/double(numTotalFrames))*100.0 : 0.0f);
}
//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 hScaleVal = yuvScalePixelShader->GetParameterByName(TEXT("baseDimensionI"));
//----------------------------------------
// x264 input buffers
int curOutBuffer = 0;
bool bUsingQSV = videoEncoder->isQSV();//GlobalConfig->GetInt(TEXT("Video Encoding"), TEXT("UseQSV")) != 0;
if(bUsingQSV)
bUsing444 = false;
EncoderPicture lastPic;
EncoderPicture outPics[NUM_OUT_BUFFERS];
DWORD outTimes[NUM_OUT_BUFFERS] = {0, 0, 0};
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;
//----------------------------------------
// time/timestamp stuff
bufferedTimes.Clear();
ctsOffsets.Clear();
int bufferedFrames = 1; //to avoid constantly polling number of frames
#ifdef USE_100NS_TIME
QWORD streamTimeStart = GetQPCTime100NS();
QWORD frameTime100ns = 10000000/fps;
QWORD sleepTargetTime = 0;
bool bWasLaggedFrame = false;
#else
DWORD streamTimeStart = OSGetTime();
DWORD fpsTimeAdjust = 0;
#endif
totalStreamTime = 0;
lastAudioTimestamp = 0;
latestVideoTime = firstSceneTimestamp = GetQPCTimeMS();
DWORD fpsTimeNumerator = 1000-(frameTime*fps);
DWORD fpsTimeDenominator = fps;
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);
convertInfo[i].bNV12 = bUsingQSV;
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();
#ifdef USE_100NS_TIME
lastStreamTime = GetQPCTime100NS()-frameTime100ns;
#else
lastStreamTime = firstFrameTime-frameTime;
#endif
//bool bFirstAudioPacket = true;
List<ProfilerNode> threadedProfilers;
bool bUsingThreadedProfilers = false;
while(bRunning || bufferedFrames)
{
#ifdef USE_100NS_TIME
QWORD renderStartTime = GetQPCTime100NS();
totalStreamTime = DWORD((renderStartTime-streamTimeStart)/10000);
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();
latestVideoTime = qwTime;
QWORD frameDelta = qwTime-lastStreamTime;
float fSeconds = float(frameDelta)*0.001f;
//Log(TEXT("frameDelta: %llu"), frameDelta);
lastStreamTime = qwTime;
#endif
bool bUpdateBPS = false;
profileIn("frame preprocessing and rendering");
//------------------------------------
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();
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)
{
// Cache
const Vect2 renderFrameSize = GetRenderFrameSize();
const Vect2 renderFrameOffset = GetRenderFrameOffset();
const Vect2 renderFrameCtrlSize = GetRenderFrameControlSize();
SetRenderTarget(NULL);
LoadVertexShader(mainVertexShader);
LoadPixelShader(mainPixelShader);
Ortho(0.0f, renderFrameCtrlSize.x, renderFrameCtrlSize.y, 0.0f, -100.0f, 100.0f);
if(renderFrameCtrlSize.x != oldRenderFrameCtrlWidth || renderFrameCtrlSize.y != oldRenderFrameCtrlHeight)
{
// User is drag resizing the window. We don't recreate the swap chains so our coordinates are wrong
SetViewport(0.0f, 0.0f, (float)oldRenderFrameCtrlWidth, (float)oldRenderFrameCtrlHeight);
}
else
SetViewport(0.0f, 0.0f, renderFrameCtrlSize.x, renderFrameCtrlSize.y);
// Draw background (Black if fullscreen, window colour otherwise)
if(bFullscreenMode)
ClearColorBuffer(0x000000);
else
ClearColorBuffer(GetSysColor(COLOR_BTNFACE));
if(bTransitioning)
{
BlendFunction(GS_BLEND_ONE, GS_BLEND_ZERO);
DrawSprite(transitionTexture, 0xFFFFFFFF,
renderFrameOffset.x, renderFrameOffset.y,
renderFrameOffset.x + renderFrameSize.x, renderFrameOffset.y + renderFrameSize.y);
BlendFunction(GS_BLEND_FACTOR, GS_BLEND_INVFACTOR, transitionAlpha);
}
DrawSprite(mainRenderTextures[curRenderTarget], 0xFFFFFFFF,
renderFrameOffset.x, renderFrameOffset.y,
renderFrameOffset.x + renderFrameSize.x, renderFrameOffset.y + renderFrameSize.y);
//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);
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(bRenderView && !copyWait)
static_cast<D3D10System*>(GS)->swap->Present(0, 0);
OSLeaveMutex(hSceneMutex);
profileOut;
//------------------------------------
// 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)
{
static bool bWarnedAboutNoAudio = false;
if (qwTime-firstFrameTime > 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)
{
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;
EncoderPicture &prevPicOut = outPics[prevOutBuffer];
EncoderPicture &picOut = outPics[curOutBuffer];
EncoderPicture &nextPicOut = outPics[nextOutBuffer];
if(!bUsing444)
{
profileIn("conversion to 4:2:0");
if(bUseThreaded420)
{
outTimes[nextOutBuffer] = (DWORD)curStreamTime;
bool firstRun = threadedProfilers.Num() == 0;
if(firstRun)
threadedProfilers.SetSize(numThreads);
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];
}
if(!firstRun)
threadedProfilers[i].~ProfilerNode();
::new (&threadedProfilers[i]) ProfilerNode(TEXT("Convert444Threads"), true);
threadedProfilers[i].MonitorThread(h420Threads[i]);
bUsingThreadedProfilers = true;
SetEvent(convertInfo[i].hSignalConvert);
}
if(bFirst420Encode)
bFirst420Encode = bEncode = false;
}
else
{
outTimes[curOutBuffer] = (DWORD)curStreamTime;
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;
}
else
{
outTimes[curOutBuffer] = (DWORD)curStreamTime;
picOut.picOut->img.i_stride[0] = map.RowPitch;
picOut.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(bDupeFrames)
{
while(cfrTime < curFrameTimestamp)
{
DWORD frameTimeAdjust = frameTime;
cfrTimeAdjust += fpsTimeNumerator;
if(cfrTimeAdjust > fpsTimeDenominator)
{
cfrTimeAdjust -= fpsTimeDenominator;
++frameTimeAdjust;
}
DWORD halfTime = (frameTimeAdjust+1)/2;
EncoderPicture &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.pic = &nextPic;
if(bUsingQSV)
frameInfo.pic->mfxOut->Data.TimeStamp = cfrTime;
else
frameInfo.pic->picOut->i_pts = cfrTime;
frameInfo.frameTimestamp = cfrTime;
ProcessFrame(frameInfo);
cfrTime += frameTimeAdjust;
//Log(TEXT("cfrTime: %u, chi frame: %u"), cfrTime, (curFrameTimestamp-cfrTime <= halfTime));
}
}
else
{
if(bUsingQSV)
picOut.mfxOut->Data.TimeStamp = curFrameTimestamp;
else
picOut.picOut->i_pts = curFrameTimestamp;
frameInfo.pic = &picOut;
frameInfo.frameTimestamp = curFrameTimestamp;
ProcessFrame(frameInfo);
}
if (!bRunning)
bufferedFrames = videoEncoder->GetBufferedFrames ();
}
if(bUsing444)
{
prevTexture->Unmap(0);
}
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;
}
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)
{
if (curStrain > 25)
{
if (qwTime - 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 = qwTime;
}
}
else if (currentBitRate < defaultBitRate && curStrain < 5 && lastStrain < 5)
{
if (qwTime - 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());*/
bUpdateBPS = true;
lastAdjustmentTime = qwTime;
}
}
}
}
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; //video encoding and uploading
profileOut; //frame
//------------------------------------
// frame sync
#ifdef USE_100NS_TIME
QWORD renderStopTime = GetQPCTime100NS();
sleepTargetTime += frameTime100ns;
if(bWasLaggedFrame = (sleepTargetTime <= renderStopTime))
{
numLongFrames++;
if(bLogLongFramesProfile && (numLongFrames/float(max(1, numTotalFrames)) * 100.) > logLongFramesProfilePercentage)
DumpLastProfileData();
}
else
SleepTo(sleepTargetTime);
#else
DWORD renderStopTime = OSGetTime();
DWORD totalTime = renderStopTime-renderStartTime;
if(totalTime > frameTimeAdjust)
{
numLongFrames++;
if(bLogLongFramesProfile && (numLongFrames/float(max(1, numTotalFrames)) * 100.) > logLongFramesProfilePercentage)
DumpLastProfileData();
}
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);
if(bUsingThreadedProfilers)
threadedProfilers[i].~ProfilerNode();
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);
}
}
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 frames that lagged: %d (%0.2f%%) (it's okay for some frames to lag)"), numTotalFrames, numLongFrames, (double(numLongFrames)/double(numTotalFrames))*100.0);
if(bDupeFrames)
Log(TEXT("Total duplicated frames: %d (%0.2f%%)"), numTotalDuplicatedFrames, (double(numTotalDuplicatedFrames)/double(numTotalFrames))*100.0);
}
//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);
}
