mfxStatus IntelDecoder::FlushDecoderAndRender()
{
mfxStatus sts = MFX_ERR_NONE;
mfxGetTime(&tStart);
//
// Stage 2: Retrieve the buffered decoded frames
//
while (MFX_ERR_NONE <= sts || MFX_ERR_MORE_SURFACE == sts) {
if (MFX_WRN_DEVICE_BUSY == sts)
MSDK_SLEEP(1); // Wait if device is busy, then repeat the same call to DecodeFrameAsync
nIndex = GetFreeSurfaceIndex(pmfxSurfaces, numSurfaces); // Find free frame surface
MSDK_CHECK_ERROR(MFX_ERR_NOT_FOUND, nIndex, MFX_ERR_MEMORY_ALLOC);
// Decode a frame asychronously (returns immediately)
sts = mfxDEC->DecodeFrameAsync(NULL, pmfxSurfaces[nIndex], &pmfxOutSurface, &syncp);
// Ignore warnings if output is available,
// if no output and no action required just repeat the DecodeFrameAsync call
if (MFX_ERR_NONE < sts && syncp)
sts = MFX_ERR_NONE;
if (MFX_ERR_NONE == sts)
sts = pSession->SyncOperation(syncp, 60000); // Synchronize. Waits until decoded frame is ready
if (MFX_ERR_NONE == sts) {
++nFrame;
if (impl_type == MFX_IMPL_SOFTWARE)
{
outMan.Render(pmfxOutSurface);
}
else
{
// Surface locking required when read/write D3D surfaces
sts = pMfxAllocator->Lock(pMfxAllocator->pthis, pmfxOutSurface->Data.MemId, &(pmfxOutSurface->Data));
MSDK_BREAK_ON_ERROR(sts);
outMan.Render(pmfxOutSurface);
sts = pMfxAllocator->Unlock(pMfxAllocator->pthis, pmfxOutSurface->Data.MemId, &(pmfxOutSurface->Data));
}
printf("Frame number: %d\r", nFrame);
fflush(stdout);
}
}
// MFX_ERR_MORE_DATA indicates that all buffers has been fetched, exit in case of other errors
MSDK_IGNORE_MFX_STS(sts, MFX_ERR_MORE_DATA);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
mfxGetTime(&tEnd);
elapsed += TimeDiffMsec(tEnd, tStart) / 1000;
double fps = ((double)nFrame / elapsed);
printf("\nExecution time: %3.2f s (%3.2f fps)\n", elapsed, fps);
return sts;
}
INT CmSurfaceManager::CreateSurface2D(CmOsResource * pCmOsResource,
BOOL bIsCmCreated,
CmSurface2D * &pSurface2D)
{
UINT handle = 0;
UINT index = m_pCmDevice->ValidSurfaceIndexStart();
INT result = 0;
UINT width = 0;
UINT height = 0;
UINT pitch = 0;
CM_SURFACE_FORMAT format = CM_SURFACE_FORMAT_UNKNOWN;
if (pCmOsResource == NULL) {
return CM_INVALID_GENOS_RESOURCE_HANDLE;
}
pSurface2D = NULL;
result = GetSurfaceInfo(pCmOsResource, width, height, pitch, format);
if (result != CM_SUCCESS) {
CM_ASSERT(0);
return result;
}
result = Surface2DSanityCheck(width, height, format);
if (result != CM_SUCCESS) {
CM_ASSERT(0);
return result;
}
if (GetFreeSurfaceIndex(index) != CM_SUCCESS) {
CM_ASSERT(0);
return CM_EXCEED_SURFACE_AMOUNT;
}
if (m_2DSurfaceCount >= m_max2DSurfaceCount) {
CM_ASSERT(0);
return CM_EXCEED_SURFACE_AMOUNT;
}
result =
AllocateSurface2D(width, height, format, pCmOsResource, handle);
if (result != CM_SUCCESS) {
CM_ASSERT(0);
return result;
}
result =
CmSurface2D::Create(index, handle, width, height, pitch, format,
bIsCmCreated, this, pSurface2D);
if (result != CM_SUCCESS) {
FreeSurface2D(handle);
CM_ASSERT(0);
return result;
}
m_SurfaceArray[index] = pSurface2D;
UPDATE_PROFILE_FOR_2D_SURFACE(index, width, height, format, FALSE);
return CM_SUCCESS;
}
HRESULT CDXVADecoderMpeg2::DecodeFrameInternal (BYTE* pDataIn, UINT nSize, REFERENCE_TIME rtStart, REFERENCE_TIME rtStop)
{
HRESULT hr = S_FALSE;
int nSurfaceIndex = -1;
CComPtr<IMediaSample> pSampleToDeliver;
int nFieldType = -1;
int nSliceType = -1;
bool bIsField = false;
int bFrame_repeat_pict = 0;
CHECK_HR_FALSE (FFMpeg2DecodeFrame (&m_PictureParams, &m_QMatrixData, m_SliceInfo, &m_nSliceCount, m_pFilter->GetAVCtx(),
m_pFilter->GetFrame(), &m_nNextCodecIndex, &nFieldType, &nSliceType, pDataIn, nSize, &bIsField, &bFrame_repeat_pict));
// Wait I frame after a flush
if (m_bFlushed && (!m_PictureParams.bPicIntra || (bIsField && m_PictureParams.bSecondField))) {
TRACE_MPEG2 ("CDXVADecoderMpeg2::DecodeFrame() : Flush - wait I frame, %ws\n", FrameType(bIsField, m_PictureParams.bSecondField));
return S_FALSE;
}
CHECK_HR (GetFreeSurfaceIndex (nSurfaceIndex, &pSampleToDeliver, rtStart, rtStop));
if (!bIsField || (bIsField && !m_PictureParams.bSecondField)) {
UpdatePictureParams(nSurfaceIndex);
}
TRACE_MPEG2 ("CDXVADecoderMpeg2::DecodeFrame() : Surf = %d, PictureType = %d, %ws, m_nNextCodecIndex = %d, rtStart = [%I64d]\n",
nSurfaceIndex, nSliceType, FrameType(bIsField, m_PictureParams.bSecondField), m_nNextCodecIndex, rtStart);
{
CHECK_HR (BeginFrame(nSurfaceIndex, pSampleToDeliver));
// Send picture parameters
CHECK_HR (AddExecuteBuffer (DXVA2_PictureParametersBufferType, sizeof(m_PictureParams), &m_PictureParams));
// Add quantization matrix
CHECK_HR (AddExecuteBuffer (DXVA2_InverseQuantizationMatrixBufferType, sizeof(m_QMatrixData), &m_QMatrixData));
// Add slice control
CHECK_HR (AddExecuteBuffer (DXVA2_SliceControlBufferType, sizeof (DXVA_SliceInfo)*m_nSliceCount, &m_SliceInfo));
// Add bitstream
CHECK_HR (AddExecuteBuffer (DXVA2_BitStreamDateBufferType, nSize, pDataIn, &nSize));
// Decode frame
CHECK_HR (Execute());
CHECK_HR (EndFrame(nSurfaceIndex));
}
bool bAdded = AddToStore (nSurfaceIndex, pSampleToDeliver, (m_PictureParams.bPicBackwardPrediction != 1), rtStart, rtStop,
bIsField, (FF_FIELD_TYPE)nFieldType, (FF_SLICE_TYPE)nSliceType, FFGetCodedPicture(m_pFilter->GetAVCtx()));
if (bAdded) {
hr = DisplayNextFrame();
}
m_bFlushed = false;
return hr;
}
// === Public functions
HRESULT CDXVADecoderMpeg2::DecodeFrame (BYTE* pDataIn, UINT nSize, REFERENCE_TIME rtStart, REFERENCE_TIME rtStop)
{
HRESULT hr;
int nSurfaceIndex;
CComPtr<IMediaSample> pSampleToDeliver;
int nFieldType;
int nSliceType;
FFMpeg2DecodeFrame (&m_PictureParams, &m_QMatrixData, m_SliceInfo, &m_nSliceCount, m_pFilter->GetAVCtx(),
m_pFilter->GetFrame(), &m_nNextCodecIndex, &nFieldType, &nSliceType, pDataIn, nSize);
// Wait I frame after a flush
if (m_bFlushed && ! m_PictureParams.bPicIntra)
return S_FALSE;
hr = GetFreeSurfaceIndex (nSurfaceIndex, &pSampleToDeliver, rtStart, rtStop);
if (FAILED (hr))
{
ASSERT (hr == VFW_E_NOT_COMMITTED); // Normal when stop playing
return hr;
}
CHECK_HR (BeginFrame(nSurfaceIndex, pSampleToDeliver));
UpdatePictureParams(nSurfaceIndex);
TRACE_MPEG2 ("=> %s %I64d Surf=%d\n", GetFFMpegPictureType(nSliceType), rtStart, nSurfaceIndex);
TRACE_MPEG2("CDXVADecoderMpeg2 : Decode frame %i\n", m_PictureParams.bPicScanMethod);
CHECK_HR (AddExecuteBuffer (DXVA2_PictureParametersBufferType, sizeof(m_PictureParams), &m_PictureParams));
CHECK_HR (AddExecuteBuffer (DXVA2_InverseQuantizationMatrixBufferType, sizeof(m_QMatrixData), &m_QMatrixData));
// Send bitstream to accelerator
CHECK_HR (AddExecuteBuffer (DXVA2_SliceControlBufferType, sizeof (DXVA_SliceInfo)*m_nSliceCount, &m_SliceInfo));
CHECK_HR (AddExecuteBuffer (DXVA2_BitStreamDateBufferType, nSize, pDataIn, &nSize));
// Decode frame
CHECK_HR (Execute());
CHECK_HR (EndFrame(nSurfaceIndex));
AddToStore (nSurfaceIndex, pSampleToDeliver, (m_PictureParams.bPicBackwardPrediction != 1), rtStart, rtStop,
false,(FF_FIELD_TYPE)nFieldType, (FF_SLICE_TYPE)nSliceType, FFGetCodedPicture(m_pFilter->GetAVCtx()));
m_bFlushed = false;
return DisplayNextFrame();
}
INT CmSurfaceManager::CreateSurface2DUP(UINT width, UINT height,
CM_SURFACE_FORMAT format, void *pSysMem,
CmSurface2DUP * &pSurface2D)
{
pSurface2D = NULL;
UINT index = m_pCmDevice->ValidSurfaceIndexStart();
if (GetFreeSurfaceIndex(index) != CM_SUCCESS) {
return CM_EXCEED_SURFACE_AMOUNT;
}
if (m_2DUPSurfaceCount >= m_max2DUPSurfaceCount) {
CM_ASSERT(0);
return CM_EXCEED_SURFACE_AMOUNT;
}
UINT handle = 0;
INT result =
AllocateSurface2DUP(width, height, format, pSysMem, handle);
if (result != CM_SUCCESS) {
CM_ASSERT(0);
return result;
}
result =
CmSurface2DUP::Create(index, handle, width, height, format, this,
pSurface2D);
if (result != CM_SUCCESS) {
FreeSurface2DUP(handle);
CM_ASSERT(0);
return result;
}
m_SurfaceArray[index] = pSurface2D;
m_2DUPSurfaceCount++;
UINT sizeperpixel = 1;
GetFormatSize(format, sizeperpixel);
m_SurfaceSizes[index] = width * height * sizeperpixel;
return CM_SUCCESS;
}
extern "C" __declspec(dllexport) int encodeBitmap(IntelEncoderHandle *pHandle, void *pBitmap, void **ppBuffer)
{
mfxStatus sts = MFX_ERR_NONE;
mfxSyncPoint syncp;
int nEncSurfIdx = 0;
// Find free frame surface
nEncSurfIdx = GetFreeSurfaceIndex(pHandle->ppEncSurfaces,pHandle->nEncSurfNum);
// TODO:
//sts = LoadRawFrame(pEncSurfaces[nEncSurfIdx], fSource, true);
for (;;) {
sts = MFXVideoENCODE_EncodeFrameAsync(pHandle->session, NULL, pHandle->ppEncSurfaces[nEncSurfIdx], &pHandle->mfxBS, &syncp);
if (MFX_ERR_NONE < sts && !syncp) { // Repeat the call if warning and no output
if (MFX_WRN_DEVICE_BUSY == sts) Sleep(5);
} else if (MFX_ERR_NONE < sts && syncp) { // Ignore warnings if output is available
sts = MFX_ERR_NONE;
break;
} else if (MFX_ERR_NOT_ENOUGH_BUFFER == sts) {
return 0;
}
else break;
}
if (MFX_ERR_NONE == sts){
sts = MFXVideoCORE_SyncOperation(pHandle->session, syncp, 60000); // Synchronize. Wait until encoded frame is ready
if (MFX_ERR_NONE != sts) return 0;
*ppBuffer = pHandle->mfxBS.Data + pHandle->mfxBS.DataOffset;
int iResult = (int) pHandle->mfxBS.DataLength;
pHandle->mfxBS.DataLength = 0;
return iResult;
}
}
INT CmSurfaceManager::AllocateSurfaceIndex(UINT width, UINT height, UINT depth,
CM_SURFACE_FORMAT format,
UINT & freeIndex,
BOOL & useNewSurface, void *pSysMem)
{
UINT index = m_pCmDevice->ValidSurfaceIndexStart();
if ((m_bufferCount >= m_maxBufferCount && width && !height && !depth) ||
(m_2DSurfaceCount >= m_max2DSurfaceCount && width && height
&& !depth)) {
if (!TouchSurfaceInPoolForDestroy()) {
CM_ASSERT(0);
return CM_FAILURE;
}
}
if (m_pCmDevice->IsSurfaceReuseEnabled() && !pSysMem) {
index = GetReuseSurfaceIndex(width, height, depth, format);
if (index) {
useNewSurface = FALSE;
freeIndex = index;
m_SurfaceReleased[index] = FALSE;
UPDATE_STATE_FOR_SURFACE_REUSE(index);
return CM_SUCCESS;
}
}
if (GetFreeSurfaceIndex(index) != CM_SUCCESS) {
return CM_FAILURE;
}
useNewSurface = TRUE;
freeIndex = index;
m_SurfaceReleased[index] = FALSE;
return CM_SUCCESS;
}
// === Public functions
HRESULT CDXVADecoderMpeg2::DecodeFrame(BYTE* pDataIn, UINT nSize, REFERENCE_TIME rtStart, REFERENCE_TIME rtStop)
{
HRESULT hr;
int nFieldType;
int nSliceType;
FFMpeg2DecodeFrame(&m_PictureParams, &m_QMatrixData, m_SliceInfo, &m_nSliceCount, m_pFilter->GetAVCtx(),
m_pFilter->GetFrame(), &m_nNextCodecIndex, &nFieldType, &nSliceType, pDataIn, nSize);
if (m_PictureParams.bSecondField && !m_bSecondField) {
m_bSecondField = true;
}
// Wait I frame after a flush
if (m_bFlushed && (!m_PictureParams.bPicIntra || (m_bSecondField && m_PictureParams.bSecondField))) {
TRACE_MPEG2("CDXVADecoderMpeg2::DecodeFrame() : Flush - wait I frame\n");
return S_FALSE;
}
if (m_bSecondField) {
if (!m_PictureParams.bSecondField) {
m_rtStart = rtStart;
m_rtStop = rtStop;
m_pSampleToDeliver = NULL;
hr = GetFreeSurfaceIndex(m_nSurfaceIndex, &m_pSampleToDeliver, rtStart, rtStop);
if (FAILED(hr)) {
ASSERT(hr == VFW_E_NOT_COMMITTED); // Normal when stop playing
return hr;
}
}
} else {
m_rtStart = rtStart;
m_rtStop = rtStop;
m_pSampleToDeliver = NULL;
hr = GetFreeSurfaceIndex(m_nSurfaceIndex, &m_pSampleToDeliver, rtStart, rtStop);
if (FAILED(hr)) {
ASSERT(hr == VFW_E_NOT_COMMITTED); // Normal when stop playing
return hr;
}
}
if (m_pSampleToDeliver == NULL) {
return S_FALSE;
}
CHECK_HR_TRACE(BeginFrame(m_nSurfaceIndex, m_pSampleToDeliver));
if (m_bSecondField) {
if (!m_PictureParams.bSecondField) {
UpdatePictureParams(m_nSurfaceIndex);
}
} else {
UpdatePictureParams(m_nSurfaceIndex);
}
TRACE_MPEG2("CDXVADecoderMpeg2::DecodeFrame() : Surf = %d, PictureType = %d, SecondField = %d, m_nNextCodecIndex = %d, rtStart = [%I64d]\n",
m_nSurfaceIndex, nSliceType, m_PictureParams.bSecondField, m_nNextCodecIndex, rtStart);
CHECK_HR_TRACE(AddExecuteBuffer(DXVA2_PictureParametersBufferType, sizeof(m_PictureParams), &m_PictureParams));
CHECK_HR_TRACE(AddExecuteBuffer(DXVA2_InverseQuantizationMatrixBufferType, sizeof(m_QMatrixData), &m_QMatrixData));
// Send bitstream to accelerator
CHECK_HR_TRACE(AddExecuteBuffer(DXVA2_SliceControlBufferType, sizeof(DXVA_SliceInfo)*m_nSliceCount, &m_SliceInfo));
CHECK_HR_TRACE(AddExecuteBuffer(DXVA2_BitStreamDateBufferType, nSize, pDataIn, &nSize));
// Decode frame
CHECK_HR_TRACE(Execute());
CHECK_HR_TRACE(EndFrame(m_nSurfaceIndex));
if (m_bSecondField) {
if (m_PictureParams.bSecondField) {
AddToStore(m_nSurfaceIndex, m_pSampleToDeliver, (m_PictureParams.bPicBackwardPrediction != 1), m_rtStart, m_rtStop,
false, (FF_FIELD_TYPE)nFieldType, (FF_SLICE_TYPE)nSliceType, FFGetCodedPicture(m_pFilter->GetAVCtx()));
hr = DisplayNextFrame();
}
} else {
AddToStore(m_nSurfaceIndex, m_pSampleToDeliver, (m_PictureParams.bPicBackwardPrediction != 1), m_rtStart, m_rtStop,
false, (FF_FIELD_TYPE)nFieldType, (FF_SLICE_TYPE)nSliceType, FFGetCodedPicture(m_pFilter->GetAVCtx()));
hr = DisplayNextFrame();
}
m_bFlushed = false;
return hr;
}
mfxStatus QSV_Encoder_Internal::Encode(uint64_t ts, uint8_t *pDataY, uint8_t *pDataUV, uint32_t strideY, uint32_t strideUV, mfxBitstream **pBS)
{
mfxStatus sts = MFX_ERR_NONE;
*pBS = NULL;
int nSurfIdx = GetFreeSurfaceIndex(m_pmfxSurfaces, m_nSurfNum); // Find free input frame surface
int nTaskIdx = GetFreeTaskIndex(m_pTaskPool, m_nTaskPool);
if (MFX_ERR_NOT_FOUND == nTaskIdx || MFX_ERR_NOT_FOUND == nSurfIdx)
{
// No more free tasks or surfaces, need to sync
sts = m_session.SyncOperation(m_pTaskPool[m_nFirstSyncTask].syncp, 60000);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
mfxU8 *pTemp = m_outBitstream.Data;
memcpy(&m_outBitstream, &m_pTaskPool[m_nFirstSyncTask].mfxBS, sizeof(mfxBitstream));
m_pTaskPool[m_nFirstSyncTask].mfxBS.Data = pTemp;
m_pTaskPool[m_nFirstSyncTask].mfxBS.DataLength = 0;
m_pTaskPool[m_nFirstSyncTask].mfxBS.DataOffset = 0;
m_pTaskPool[m_nFirstSyncTask].syncp = NULL;
m_nFirstSyncTask = (m_nFirstSyncTask + 1) % m_nTaskPool;
*pBS = &m_outBitstream;
if (nTaskIdx == MFX_ERR_NOT_FOUND)
nTaskIdx = GetFreeTaskIndex(m_pTaskPool, m_nTaskPool);
if (nSurfIdx == MFX_ERR_NOT_FOUND)
nSurfIdx = GetFreeSurfaceIndex(m_pmfxSurfaces, m_nSurfNum);
}
mfxFrameSurface1 *pSurface = m_pmfxSurfaces[nSurfIdx];
sts = m_mfxAllocator.Lock(m_mfxAllocator.pthis, pSurface->Data.MemId, &(pSurface->Data));
sts = LoadNV12(pSurface, pDataY, pDataUV, strideY, strideUV);
pSurface->Data.TimeStamp = ts;
sts = m_mfxAllocator.Unlock(m_mfxAllocator.pthis, pSurface->Data.MemId, &(pSurface->Data));
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
for (;;)
{
// Encode a frame asychronously (returns immediately)
sts = m_pmfxENC->EncodeFrameAsync(NULL, pSurface, &m_pTaskPool[nTaskIdx].mfxBS, &m_pTaskPool[nTaskIdx].syncp);
if (MFX_ERR_NONE < sts && !m_pTaskPool[nTaskIdx].syncp)
{
// Repeat the call if warning and no output
if (MFX_WRN_DEVICE_BUSY == sts)
MSDK_SLEEP(1); // Wait if device is busy, then repeat the same call
}
else if (MFX_ERR_NONE < sts && m_pTaskPool[nTaskIdx].syncp)
{
sts = MFX_ERR_NONE; // Ignore warnings if output is available
break;
}
else if (MFX_ERR_NOT_ENOUGH_BUFFER == sts) {
// Allocate more bitstream buffer memory here if needed...
break;
}
else
break;
}
return sts;
}
int MSDKVpp::HandleProcess()
{
if (m_vppCfg.comp_num != m_mapMediaBuf.size())
{
VPP_TRACE_ERROR("[MSDKVpp]-----Something wrong here, all sinkpads detached\n");
return -1;
}
std::map<MSDKDecodeVpp*, MediaBuf>::iterator it = m_mapMediaBuf.begin();
mfxFrameSurface1* pInSurface = NULL;
mfxStatus sts = MFX_ERR_NONE;
int nIndex = 0;
m_tCompStc = 0;
while(!m_bWantToStop) {
usleep(1000);
if (VPP_COMP == m_mode) {
//composite case, break if no more data when preparing vpp frames
int prepare_result = PrepareVppCompFrames();
if (-1 == prepare_result)
break;
if (1 == prepare_result)
{
m_bEndOfStream = true;
break;
}
//printf("[MSDKVpp %p]-----Prepare composite frames over\n", this);
} else {
if (!m_bAccessNextElem)
continue;
if (it->second.pRingBuf->IsEmpty()) { //No data
if (it->first->GetDataEos()) //No more data in the furture
pInSurface = NULL;
else
continue;
} else {
it->second.pRingBuf->Get(pInSurface);
}
}
if (!m_bInit) {
m_mapMediaBuf.begin()->second.pRingBuf->Get(pInSurface);
Locker<Mutex> l(m_xMsdkInit);
sts = InitVpp(pInSurface);
if (MFX_ERR_NONE == sts) {
m_bInit = true;
VPP_TRACE_INFO("[MSDKVpp]VPP element %p init successfully\n", this);
} else {
VPP_TRACE_ERROR("[MSDKVpp]VPP create failed: %d\n", sts);
break;
}
}
if (m_bReinit) {
//Re-init VPP.
mfxU32 before_change = GetSysTime();
VPP_TRACE_INFO("[MSDKVpp]stop/init vpp\n");
m_pVpp->Close();
VPP_TRACE_INFO("Re-init VPP...\n");
m_mapMediaBuf.begin()->second.pRingBuf->Get(pInSurface);
{
Locker<Mutex> l(m_xMsdkReinit);
sts = InitVpp(pInSurface);
m_bReinit = false;
}
VPP_TRACE_INFO("[MSDKVpp]Re-Init VPP takes time %u(us)\n", GetSysTime()-before_change);
}
//malloc memory in InitVpp(pInSurface), then will get free surface slot
nIndex = GetFreeSurfaceIndex(m_pSurfacePool, m_nSurfaces);
if (MFX_ERR_NOT_FOUND == nIndex) {
//printf("[MSDKVpp]-----Cann't get free surface slot, check if not free\n");
continue;
}
if (VPP_COMP == m_mode) {
if (m_pMeasuremnt)
{
m_pMeasuremnt->GetLock();
m_pMeasuremnt->TimeStpStart(VPP_FRAME_TIME_STAMP, this);
m_pMeasuremnt->RelLock();
}
for (it = m_mapMediaBuf.begin(); it != m_mapMediaBuf.end(); ++it) {
it->second.pRingBuf->Pop(pInSurface);
//printf("[MSDKVpp %p]Get pInSurface %p succ\n", pInSurface);
sts = DoingVpp(pInSurface, m_pSurfacePool[nIndex]);
pInSurface->Data.Locked--;
}
//printf("[MSDKVpp %p]-----Composite multiple frames into one frame successfully\n", this);
} else {
sts = DoingVpp(pInSurface, m_pSurfacePool[nIndex]);
m_bAccessNextElem = false;
dynamic_cast<MSDKVpp*>(it->first)->ReleaseSurface();
if (!pInSurface) //resize case, break when be notified end of stream
break;
}
}
if (!m_bWantToStop) {
VPP_TRACE_INFO("[%p]This work flow finish video processing and will be stopped.\n", this);
m_bWantToStop = true;
}
m_rCallback.StopTrain();
return 0;
}
INT CmSurfaceManager::CreateSurface2D(UINT width, UINT height, UINT pitch,
BOOL bIsCmCreated,
CM_SURFACE_FORMAT format,
CmSurface2D * &pSurface2D)
{
UINT handle = 0;
UINT index = m_pCmDevice->ValidSurfaceIndexStart();
INT result = 0;
pSurface2D = NULL;
result = Surface2DSanityCheck(width, height, format);
if (result != CM_SUCCESS) {
CM_ASSERT(0);
return result;
}
if (bIsCmCreated) {
BOOL useNewSurface = TRUE;
if (AllocateSurfaceIndex
(width, height, 0, format, index, useNewSurface,
NULL) != CM_SUCCESS) {
return CM_EXCEED_SURFACE_AMOUNT;
}
if (!useNewSurface) {
CmSurface *pSurface = m_SurfaceArray[index];
if (pSurface
&& (pSurface->Type() ==
CM_ENUM_CLASS_TYPE_CMSURFACE2D)) {
pSurface2D =
static_cast < CmSurface2D * >(pSurface);
} else {
return CM_FAILURE;
}
UpdateSurface2D(pSurface2D, width, height, format);
UPDATE_PROFILE_FOR_2D_SURFACE(index, width, height,
format, TRUE);
return CM_SUCCESS;
}
} else {
if (GetFreeSurfaceIndex(index) != CM_SUCCESS) {
return CM_EXCEED_SURFACE_AMOUNT;
}
}
if (m_2DSurfaceCount >= m_max2DSurfaceCount) {
CM_ASSERT(0);
return CM_EXCEED_SURFACE_AMOUNT;
}
result = AllocateSurface2D(width, height, format, handle, pitch);
if (result != CM_SUCCESS) {
CM_ASSERT(0);
return result;
}
result =
CmSurface2D::Create(index, handle, width, height, pitch, format,
TRUE, this, pSurface2D);
if (result != CM_SUCCESS) {
FreeSurface2D(handle);
CM_ASSERT(0);
return result;
}
m_SurfaceArray[index] = pSurface2D;
UPDATE_PROFILE_FOR_2D_SURFACE(index, width, height, format, FALSE);
return CM_SUCCESS;
}
HRESULT CDXVADecoderH264::DecodeFrame (BYTE* pDataIn, UINT nSize, REFERENCE_TIME rtStart, REFERENCE_TIME rtStop)
{
HRESULT hr = S_FALSE;
CH264Nalu Nalu;
UINT nSlices = 0;
int nSurfaceIndex;
int nFieldType;
int nSliceType;
int nFramePOC;
IDirect3DSurface9* pSampleToDeliver;
int nDXIndex = 0;
UINT nNalOffset = 0;
int nOutPOC;
REFERENCE_TIME rtOutStart;
if(pDataIn == NULL || nSize == 0)
return S_FALSE;
Nalu.SetBuffer (pDataIn, nSize, m_nNALLength);
FFH264DecodeBuffer (m_pFilter->GetAVCtx(), pDataIn, nSize, &nFramePOC, &nOutPOC, &rtOutStart);
//CLog::Log(LOGDEBUG, "nFramePOC = %d nOutPOC %d rtOutStart%d", nFramePOC, nOutPOC, rtOutStart);
while (Nalu.ReadNext())
{
switch (Nalu.GetType())
{
case NALU_TYPE_SLICE:
case NALU_TYPE_IDR:
if(m_bUseLongSlice)
{
m_pSliceLong[nSlices].BSNALunitDataLocation = nNalOffset;
m_pSliceLong[nSlices].SliceBytesInBuffer = Nalu.GetDataLength()+3; //.GetRoundedDataLength();
m_pSliceLong[nSlices].slice_id = nSlices;
FF264UpdateRefFrameSliceLong(&m_DXVAPicParams, &m_pSliceLong[nSlices], m_pFilter->GetAVCtx());
if (nSlices>0)
m_pSliceLong[nSlices-1].NumMbsForSlice = m_pSliceLong[nSlices].NumMbsForSlice = m_pSliceLong[nSlices].first_mb_in_slice - m_pSliceLong[nSlices-1].first_mb_in_slice;
}
nSlices++;
nNalOffset += (UINT)(Nalu.GetDataLength() + 3);
if (nSlices > MAX_SLICES) break;
break;
}
}
if (nSlices == 0) return S_FALSE;
m_nMaxWaiting = min (max (m_DXVAPicParams.num_ref_frames, 3), 8);
// If parsing fail (probably no PPS/SPS), continue anyway it may arrived later (happen on truncated streams)
if (FAILED (FFH264BuildPicParams (&m_DXVAPicParams, &m_DXVAScalingMatrix, &nFieldType, &nSliceType, m_pFilter->GetAVCtx(), m_pFilter->GetPCIVendor())))
return S_FALSE;
// Wait I frame after a flush
if (m_bFlushed && !m_DXVAPicParams.IntraPicFlag)
return S_FALSE;
CHECK_HR (GetFreeSurfaceIndex (nSurfaceIndex, &pSampleToDeliver, rtStart, rtStop));
FFH264SetCurrentPicture (nSurfaceIndex, &m_DXVAPicParams, m_pFilter->GetAVCtx());
CHECK_HR (BeginFrame(pSampleToDeliver));
m_DXVAPicParams.StatusReportFeedbackNumber++;
// TRACE("CDXVADecoderH264 : Decode frame %u\n", m_DXVAPicParams.StatusReportFeedbackNumber);
// Send picture parameters
CHECK_HR (AddExecuteBuffer (DXVA2_PictureParametersBufferType, sizeof(m_DXVAPicParams), &m_DXVAPicParams));
CHECK_HR (Execute());
// Add bitstream, slice control and quantization matrix
CHECK_HR (AddExecuteBuffer (DXVA2_BitStreamDateBufferType, nSize, pDataIn, &nSize));
if (m_bUseLongSlice)
{
CHECK_HR(AddExecuteBuffer(DXVA2_SliceControlBufferType, sizeof(DXVA_Slice_H264_Long)*nSlices, m_pSliceLong));
}
else
{
CHECK_HR (AddExecuteBuffer (DXVA2_SliceControlBufferType, sizeof (DXVA_Slice_H264_Short)*nSlices, m_pSliceShort));
}
CHECK_HR (AddExecuteBuffer (DXVA2_InverseQuantizationMatrixBufferType, sizeof (DXVA_Qmatrix_H264), (void*)&m_DXVAScalingMatrix));
// Decode bitstream
CHECK_HR (Execute());
CHECK_HR (EndFrame(nSurfaceIndex));
#ifdef _DEBUG
//DisplayStatus();
#endif
bool bAdded = AddToStore (nSurfaceIndex, m_DXVAPicParams.RefPicFlag, rtStart, rtStop,
m_DXVAPicParams.field_pic_flag, (FF_FIELD_TYPE)nFieldType,
(FF_SLICE_TYPE)nSliceType, nFramePOC);
FFH264UpdateRefFramesList (&m_DXVAPicParams, m_pFilter->GetAVCtx());
ClearUnusedRefFrames();
if (bAdded)
{
hr = DisplayNextFrame();
if (nOutPOC != -1)
{
m_nOutPOC = nOutPOC;
m_rtOutStart = rtOutStart;
}
}
m_bFlushed = false;
return hr;
}
mfxStatus IntelDecoder::RunDecodeAndRender()
{
mfxStatus sts = MFX_ERR_NONE;
// ===============================================================
// Start decoding the frames from the stream
//
mfxGetTime(&tStart);
pmfxOutSurface = NULL;
pmfxOutSurface_sw = NULL;
nIndex = 0;
nIndex2 = 0;
nFrame = 0;
//
// Stage 1: Main decoding loop
//
while (MFX_ERR_NONE <= sts || MFX_ERR_MORE_DATA == sts || MFX_ERR_MORE_SURFACE == sts) {
if (MFX_WRN_DEVICE_BUSY == sts)
MSDK_SLEEP(1); // Wait if device is busy, then repeat the same call to DecodeFrameAsync
if (MFX_ERR_MORE_DATA == sts) {
sts = ReadBitStreamData(&mfxBS, fSource); // Read more data into input bit stream
MSDK_BREAK_ON_ERROR(sts);
}
if (MFX_ERR_MORE_SURFACE == sts || MFX_ERR_NONE == sts) {
nIndex = GetFreeSurfaceIndex(pmfxSurfaces, numSurfaces); // Find free frame surface
MSDK_CHECK_ERROR(MFX_ERR_NOT_FOUND, nIndex, MFX_ERR_MEMORY_ALLOC);
}
// Decode a frame asychronously (returns immediately)
// - If input bitstream contains multiple frames DecodeFrameAsync will start decoding multiple frames, and remove them from bitstream
sts = mfxDEC->DecodeFrameAsync(&mfxBS, pmfxSurfaces[nIndex], &pmfxOutSurface, &syncp);
// Ignore warnings if output is available,
// if no output and no action required just repeat the DecodeFrameAsync call
if (MFX_ERR_NONE < sts && syncp)
sts = MFX_ERR_NONE;
if (MFX_ERR_NONE == sts)
sts = pSession->SyncOperation(syncp, 60000); // Synchronize. Wait until decoded frame is ready
if (MFX_ERR_NONE == sts) {
++nFrame;
if (impl_type == MFX_IMPL_SOFTWARE)
{
outMan.Render(pmfxOutSurface);
}
else
{
// Surface locking required when read/write video surfaces
sts = pMfxAllocator->Lock(pMfxAllocator->pthis, pmfxOutSurface->Data.MemId, &(pmfxOutSurface->Data));
MSDK_BREAK_ON_ERROR(sts);
outMan.Render(pmfxOutSurface);
sts = pMfxAllocator->Unlock(pMfxAllocator->pthis, pmfxOutSurface->Data.MemId, &(pmfxOutSurface->Data));
MSDK_BREAK_ON_ERROR(sts);
}
printf("Frame number: %d\r", nFrame);
fflush(stdout);
}
}
// MFX_ERR_MORE_DATA means that file has ended, need to go to buffering loop, exit in case of other errors
MSDK_IGNORE_MFX_STS(sts, MFX_ERR_MORE_DATA);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
mfxGetTime(&tEnd);
elapsed = TimeDiffMsec(tEnd, tStart) / 1000;
double fps = ((double)nFrame / elapsed);
printf("\nExecution time: %3.2f s (%3.2f fps)\n", elapsed, fps);
return sts;
}
// === Public functions
HRESULT TDXVADecoderVC1::DecodeFrame(BYTE* pDataIn, UINT nSize, REFERENCE_TIME rtStart, REFERENCE_TIME rtStop)
{
HRESULT hr;
int nSurfaceIndex;
CComPtr<IMediaSample> pSampleToDeliver;
int nFieldType, nSliceType;
UINT nFrameSize, nSize_Result;
m_pCodec->libavcodec->FFVC1UpdatePictureParam(&m_PictureParams, m_pCodec->avctx, &nFieldType, &nSliceType, pDataIn, nSize, &nFrameSize, FALSE, &m_bFrame_repeat_pict);
if (m_pCodec->libavcodec->FFIsSkipped(m_pCodec->avctx)) {
return S_OK;
}
// Wait I frame after a flush
if (m_bFlushed && ! m_PictureParams.bPicIntra) {
return S_FALSE;
}
hr = GetFreeSurfaceIndex(nSurfaceIndex, &pSampleToDeliver, rtStart, rtStop);
if (FAILED(hr)) {
ASSERT(hr == VFW_E_NOT_COMMITTED); // Normal when stop playing
return hr;
}
CHECK_HR(BeginFrame(nSurfaceIndex, pSampleToDeliver));
DPRINTF(_l("TDXVADecoderVC1::DecodeFrame - PictureType = %s, rtStart = %I64d Surf=%d\n"), m_pCodec->libavcodec->GetFFMpegPictureType(nSliceType), rtStart, nSurfaceIndex);
m_PictureParams.wDecodedPictureIndex = nSurfaceIndex;
m_PictureParams.wDeblockedPictureIndex = m_PictureParams.wDecodedPictureIndex;
// Manage reference picture list
if (!m_PictureParams.bPicBackwardPrediction) {
if (m_wRefPictureIndex[0] != NO_REF_FRAME) {
RemoveRefFrame(m_wRefPictureIndex[0]);
}
m_wRefPictureIndex[0] = m_wRefPictureIndex[1];
m_wRefPictureIndex[1] = nSurfaceIndex;
}
m_PictureParams.wForwardRefPictureIndex = (m_PictureParams.bPicIntra == 0) ? m_wRefPictureIndex[0] : NO_REF_FRAME;
m_PictureParams.wBackwardRefPictureIndex = (m_PictureParams.bPicBackwardPrediction == 1) ? m_wRefPictureIndex[1] : NO_REF_FRAME;
m_PictureParams.bPic4MVallowed = (m_PictureParams.wBackwardRefPictureIndex == NO_REF_FRAME && m_PictureParams.bPicStructure == 3) ? 1 : 0;
m_PictureParams.bPicDeblockConfined |= (m_PictureParams.wBackwardRefPictureIndex == NO_REF_FRAME) ? 0x04 : 0;
m_PictureParams.bPicScanMethod++; // Use for status reporting sections 3.8.1 and 3.8.2
DPRINTF(_l("TDXVADecoderVC1::DecodeFrame - Decode frame %i\n"), m_PictureParams.bPicScanMethod);
// Send picture params to accelerator
CHECK_HR(AddExecuteBuffer(DXVA2_PictureParametersBufferType, sizeof(m_PictureParams), &m_PictureParams));
// Send bitstream to accelerator
CHECK_HR(AddExecuteBuffer(DXVA2_BitStreamDateBufferType, nFrameSize ? nFrameSize : nSize, pDataIn, &nSize_Result));
m_SliceInfo.wQuantizerScaleCode = 1; // TODO : 1->31 ???
m_SliceInfo.dwSliceBitsInBuffer = nSize_Result * 8;
CHECK_HR(AddExecuteBuffer(DXVA2_SliceControlBufferType, sizeof(m_SliceInfo), &m_SliceInfo));
// Decode frame
CHECK_HR(Execute());
CHECK_HR(EndFrame(nSurfaceIndex));
// ***************
if (nFrameSize) { // Decoding Second Field
m_pCodec->libavcodec->FFVC1UpdatePictureParam(&m_PictureParams, m_pCodec->avctx, NULL, NULL, pDataIn, nSize, NULL, TRUE, &m_bFrame_repeat_pict);
CHECK_HR(BeginFrame(nSurfaceIndex, pSampleToDeliver));
DPRINTF(_l("TDXVADecoderVC1::DecodeFrame - PictureType = %s\n"), m_pCodec->libavcodec->GetFFMpegPictureType(nSliceType));
CHECK_HR(AddExecuteBuffer(DXVA2_PictureParametersBufferType, sizeof(m_PictureParams), &m_PictureParams));
// Send bitstream to accelerator
CHECK_HR(AddExecuteBuffer(DXVA2_BitStreamDateBufferType, nSize - nFrameSize, pDataIn + nFrameSize, &nSize_Result));
m_SliceInfo.wQuantizerScaleCode = 1; // TODO : 1->31 ???
m_SliceInfo.dwSliceBitsInBuffer = nSize_Result * 8;
CHECK_HR(AddExecuteBuffer(DXVA2_SliceControlBufferType, sizeof(m_SliceInfo), &m_SliceInfo));
// Decode frame
CHECK_HR(Execute());
CHECK_HR(EndFrame(nSurfaceIndex));
}
// ***************
#ifdef _DEBUG
DisplayStatus();
#endif
// Re-order B frames
if (m_pCodec->isReorderBFrame()) {
if (m_PictureParams.bPicBackwardPrediction == 1) {
SwapRT(rtStart, m_rtStartDelayed);
SwapRT(rtStop, m_rtStopDelayed);
} else {
// Save I or P reference time (swap later)
if (!m_bFlushed) {
if (m_nDelayedSurfaceIndex != -1) {
UpdateStore(m_nDelayedSurfaceIndex, m_rtStartDelayed, m_rtStopDelayed);
}
m_rtStartDelayed = m_rtStopDelayed = _I64_MAX;
SwapRT(rtStart, m_rtStartDelayed);
SwapRT(rtStop, m_rtStopDelayed);
m_nDelayedSurfaceIndex = nSurfaceIndex;
}
}
}
AddToStore(nSurfaceIndex, pSampleToDeliver, (m_PictureParams.bPicBackwardPrediction != 1), rtStart, rtStop,
false, (FF_FIELD_TYPE)nFieldType, (FF_SLICE_TYPE)nSliceType, 0);
m_bFlushed = false;
return DisplayNextFrame();
}
int main()
{
mfxStatus sts = MFX_ERR_NONE;
mfxU16 inputWidth = 1920;
mfxU16 inputHeight = 1080;
// =====================================================================
// Intel Media SDK encode pipeline setup
// - In this example we are encoding an AVC (H.264) stream
// - Video memory surfaces are used
// - Asynchronous operation by executing more than one encode operation simultaneously
//
// Open input YV12 YUV file
FILE* fSource;
fopen_s(&fSource, "bbb1920x1080.yuv", "rb");
MSDK_CHECK_POINTER(fSource, MFX_ERR_NULL_PTR);
// Create output elementary stream (ES) H.264 file
FILE* fSink;
fopen_s(&fSink, "test_d3d_async.264", "wb");
MSDK_CHECK_POINTER(fSink, MFX_ERR_NULL_PTR);
// Initialize Media SDK session
// - MFX_IMPL_AUTO_ANY selects HW accelaration if available (on any adapter)
// - Version 1.0 is selected for greatest backwards compatibility.
// If more recent API features are needed, change the version accordingly
mfxIMPL impl = MFX_IMPL_AUTO_ANY;
#ifdef DX11_D3D
impl |= MFX_IMPL_VIA_D3D11;
#endif
mfxVersion ver = {0, 1};
MFXVideoSession mfxSession;
sts = mfxSession.Init(impl, &ver);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
// Create DirectX device context
mfxHDL deviceHandle;
sts = CreateHWDevice(mfxSession, &deviceHandle, NULL);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
// Provide device manager to Media SDK
sts = mfxSession.SetHandle(DEVICE_MGR_TYPE, deviceHandle);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
mfxFrameAllocator mfxAllocator;
mfxAllocator.Alloc = simple_alloc;
mfxAllocator.Free = simple_free;
mfxAllocator.Lock = simple_lock;
mfxAllocator.Unlock = simple_unlock;
mfxAllocator.GetHDL = simple_gethdl;
// When using video memory we must provide Media SDK with an external allocator
sts = mfxSession.SetFrameAllocator(&mfxAllocator);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
// Initialize encoder parameters
mfxVideoParam mfxEncParams;
memset(&mfxEncParams, 0, sizeof(mfxEncParams));
mfxEncParams.mfx.CodecId = MFX_CODEC_AVC;
mfxEncParams.mfx.TargetUsage = MFX_TARGETUSAGE_BALANCED;
mfxEncParams.mfx.TargetKbps = 2000;
mfxEncParams.mfx.RateControlMethod = MFX_RATECONTROL_VBR;
mfxEncParams.mfx.FrameInfo.FrameRateExtN = 30;
mfxEncParams.mfx.FrameInfo.FrameRateExtD = 1;
mfxEncParams.mfx.FrameInfo.FourCC = MFX_FOURCC_NV12;
mfxEncParams.mfx.FrameInfo.ChromaFormat = MFX_CHROMAFORMAT_YUV420;
mfxEncParams.mfx.FrameInfo.PicStruct = MFX_PICSTRUCT_PROGRESSIVE;
mfxEncParams.mfx.FrameInfo.CropX = 0;
mfxEncParams.mfx.FrameInfo.CropY = 0;
mfxEncParams.mfx.FrameInfo.CropW = inputWidth;
mfxEncParams.mfx.FrameInfo.CropH = inputHeight;
// Width must be a multiple of 16
// Height must be a multiple of 16 in case of frame picture and a multiple of 32 in case of field picture
mfxEncParams.mfx.FrameInfo.Width = MSDK_ALIGN16(inputWidth);
mfxEncParams.mfx.FrameInfo.Height = (MFX_PICSTRUCT_PROGRESSIVE == mfxEncParams.mfx.FrameInfo.PicStruct)?
MSDK_ALIGN16(inputHeight) : MSDK_ALIGN32(inputHeight);
mfxEncParams.IOPattern = MFX_IOPATTERN_IN_VIDEO_MEMORY;
// Configure Media SDK to keep more operations in flight
// - AsyncDepth represents the number of tasks that can be submitted, before synchronizing is required
// - The choice of AsyncDepth = 4 is quite arbitrary but has proven to result in good performance
mfxEncParams.AsyncDepth = 4;
// Create Media SDK encoder
MFXVideoENCODE mfxENC(mfxSession);
// Validate video encode parameters (optional)
// - In this example the validation result is written to same structure
// - MFX_WRN_INCOMPATIBLE_VIDEO_PARAM is returned if some of the video parameters are not supported,
// instead the encoder will select suitable parameters closest matching the requested configuration
sts = mfxENC.Query(&mfxEncParams, &mfxEncParams);
MSDK_IGNORE_MFX_STS(sts, MFX_WRN_INCOMPATIBLE_VIDEO_PARAM);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
// Query number of required surfaces for encoder
mfxFrameAllocRequest EncRequest;
memset(&EncRequest, 0, sizeof(EncRequest));
sts = mfxENC.QueryIOSurf(&mfxEncParams, &EncRequest);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
#ifdef DX11_D3D
EncRequest.Type |= WILL_WRITE; // Hint to DX11 memory handler that application will write data to input surfaces
#endif
// Allocate required surfaces
mfxFrameAllocResponse mfxResponse;
sts = mfxAllocator.Alloc(mfxAllocator.pthis, &EncRequest, &mfxResponse);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
mfxU16 nEncSurfNum = mfxResponse.NumFrameActual;
// Allocate surface headers (mfxFrameSurface1) for decoder
mfxFrameSurface1** pmfxSurfaces = new mfxFrameSurface1*[nEncSurfNum];
MSDK_CHECK_POINTER(pmfxSurfaces, MFX_ERR_MEMORY_ALLOC);
for (int i = 0; i < nEncSurfNum; i++)
{
pmfxSurfaces[i] = new mfxFrameSurface1;
memset(pmfxSurfaces[i], 0, sizeof(mfxFrameSurface1));
memcpy(&(pmfxSurfaces[i]->Info), &(mfxEncParams.mfx.FrameInfo), sizeof(mfxFrameInfo));
pmfxSurfaces[i]->Data.MemId = mfxResponse.mids[i]; // MID (memory id) represent one D3D NV12 surface
#ifndef ENABLE_INPUT
// In case simulating direct access to frames we initialize the allocated surfaces with default pattern
// - For true benchmark comparisons to async workloads all surfaces must have the same data
#ifndef DX11_D3D
IDirect3DSurface9 *pSurface;
D3DSURFACE_DESC desc;
D3DLOCKED_RECT locked;
pSurface = (IDirect3DSurface9 *)mfxResponse.mids[i];
pSurface->GetDesc(&desc);
pSurface->LockRect(&locked, 0, D3DLOCK_NOSYSLOCK);
memset((mfxU8 *)locked.pBits, 100, desc.Height*locked.Pitch); // Y plane
memset((mfxU8 *)locked.pBits + desc.Height * locked.Pitch, 50, (desc.Height*locked.Pitch)/2); // UV plane
pSurface->UnlockRect();
#else
// For now, just leave D3D11 surface data uninitialized
#endif
#endif
}
// Initialize the Media SDK encoder
sts = mfxENC.Init(&mfxEncParams);
MSDK_IGNORE_MFX_STS(sts, MFX_WRN_PARTIAL_ACCELERATION);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
// Retrieve video parameters selected by encoder.
// - BufferSizeInKB parameter is required to set bit stream buffer size
mfxVideoParam par;
memset(&par, 0, sizeof(par));
sts = mfxENC.GetVideoParam(&par);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
// Create task pool to improve asynchronous performance (greater GPU utilization)
mfxU16 taskPoolSize = mfxEncParams.AsyncDepth; // number of tasks that can be submitted, before synchronizing is required
Task* pTasks = new Task[taskPoolSize];
memset(pTasks, 0, sizeof(Task) * taskPoolSize);
for(int i=0;i<taskPoolSize;i++)
{
// Prepare Media SDK bit stream buffer
pTasks[i].mfxBS.MaxLength = par.mfx.BufferSizeInKB * 1000;
pTasks[i].mfxBS.Data = new mfxU8[pTasks[i].mfxBS.MaxLength];
MSDK_CHECK_POINTER(pTasks[i].mfxBS.Data, MFX_ERR_MEMORY_ALLOC);
}
// ===================================
// Start encoding the frames
//
#ifdef ENABLE_BENCHMARK
LARGE_INTEGER tStart, tEnd;
QueryPerformanceFrequency(&tStart);
double freq = (double)tStart.QuadPart;
QueryPerformanceCounter(&tStart);
#endif
int nEncSurfIdx = 0;
int nTaskIdx = 0;
int nFirstSyncTask = 0;
mfxU32 nFrame = 0;
//
// Stage 1: Main encoding loop
//
while (MFX_ERR_NONE <= sts || MFX_ERR_MORE_DATA == sts)
{
nTaskIdx = GetFreeTaskIndex(pTasks, taskPoolSize); // Find free task
if(MFX_ERR_NOT_FOUND == nTaskIdx)
{
// No more free tasks, need to sync
sts = mfxSession.SyncOperation(pTasks[nFirstSyncTask].syncp, 60000);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
sts = WriteBitStreamFrame(&pTasks[nFirstSyncTask].mfxBS, fSink);
MSDK_BREAK_ON_ERROR(sts);
pTasks[nFirstSyncTask].syncp = NULL;
nFirstSyncTask = (nFirstSyncTask + 1) % taskPoolSize;
++nFrame;
#ifdef ENABLE_OUTPUT
printf("Frame number: %d\r", nFrame);
#endif
}
else
{
nEncSurfIdx = GetFreeSurfaceIndex(pmfxSurfaces, nEncSurfNum); // Find free frame surface
if (MFX_ERR_NOT_FOUND == nEncSurfIdx)
return MFX_ERR_MEMORY_ALLOC;
// Surface locking required when read/write D3D surfaces
sts = mfxAllocator.Lock(mfxAllocator.pthis, pmfxSurfaces[nEncSurfIdx]->Data.MemId, &(pmfxSurfaces[nEncSurfIdx]->Data));
MSDK_BREAK_ON_ERROR(sts);
sts = LoadRawFrame(pmfxSurfaces[nEncSurfIdx], fSource);
MSDK_BREAK_ON_ERROR(sts);
sts = mfxAllocator.Unlock(mfxAllocator.pthis, pmfxSurfaces[nEncSurfIdx]->Data.MemId, &(pmfxSurfaces[nEncSurfIdx]->Data));
MSDK_BREAK_ON_ERROR(sts);
for (;;)
{
// Encode a frame asychronously (returns immediately)
sts = mfxENC.EncodeFrameAsync(NULL, pmfxSurfaces[nEncSurfIdx], &pTasks[nTaskIdx].mfxBS, &pTasks[nTaskIdx].syncp);
if (MFX_ERR_NONE < sts && !pTasks[nTaskIdx].syncp) // Repeat the call if warning and no output
{
if (MFX_WRN_DEVICE_BUSY == sts)
Sleep(1); // Wait if device is busy, then repeat the same call
}
else if (MFX_ERR_NONE < sts && pTasks[nTaskIdx].syncp)
{
sts = MFX_ERR_NONE; // Ignore warnings if output is available
break;
}
else if (MFX_ERR_NOT_ENOUGH_BUFFER == sts)
{
// Allocate more bitstream buffer memory here if needed...
break;
}
else
break;
}
}
}
// MFX_ERR_MORE_DATA means that the input file has ended, need to go to buffering loop, exit in case of other errors
MSDK_IGNORE_MFX_STS(sts, MFX_ERR_MORE_DATA);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
//
// Stage 2: Retrieve the buffered encoded frames
//
while (MFX_ERR_NONE <= sts)
{
nTaskIdx = GetFreeTaskIndex(pTasks, taskPoolSize); // Find free task
if(MFX_ERR_NOT_FOUND == nTaskIdx)
{
// No more free tasks, need to sync
sts = mfxSession.SyncOperation(pTasks[nFirstSyncTask].syncp, 60000);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
sts = WriteBitStreamFrame(&pTasks[nFirstSyncTask].mfxBS, fSink);
MSDK_BREAK_ON_ERROR(sts);
pTasks[nFirstSyncTask].syncp = NULL;
nFirstSyncTask = (nFirstSyncTask + 1) % taskPoolSize;
++nFrame;
#ifdef ENABLE_OUTPUT
printf("Frame number: %d\r", nFrame);
#endif
}
else
{
for (;;)
{
// Encode a frame asychronously (returns immediately)
sts = mfxENC.EncodeFrameAsync(NULL, NULL, &pTasks[nTaskIdx].mfxBS, &pTasks[nTaskIdx].syncp);
if (MFX_ERR_NONE < sts && !pTasks[nTaskIdx].syncp) // Repeat the call if warning and no output
{
if (MFX_WRN_DEVICE_BUSY == sts)
Sleep(1); // Wait if device is busy, then repeat the same call
}
else if (MFX_ERR_NONE < sts && pTasks[nTaskIdx].syncp)
{
sts = MFX_ERR_NONE; // Ignore warnings if output is available
break;
}
else
break;
}
}
}
// MFX_ERR_MORE_DATA indicates that there are no more buffered frames, exit in case of other errors
MSDK_IGNORE_MFX_STS(sts, MFX_ERR_MORE_DATA);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
//
// Stage 3: Sync all remaining tasks in task pool
//
while(pTasks[nFirstSyncTask].syncp)
{
sts = mfxSession.SyncOperation(pTasks[nFirstSyncTask].syncp, 60000);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
sts = WriteBitStreamFrame(&pTasks[nFirstSyncTask].mfxBS, fSink);
MSDK_BREAK_ON_ERROR(sts);
pTasks[nFirstSyncTask].syncp = NULL;
nFirstSyncTask = (nFirstSyncTask + 1) % taskPoolSize;
++nFrame;
#ifdef ENABLE_OUTPUT
printf("Frame number: %d\r", nFrame);
#endif
}
#ifdef ENABLE_BENCHMARK
QueryPerformanceCounter(&tEnd);
double duration = ((double)tEnd.QuadPart - (double)tStart.QuadPart) / freq;
printf("\nExecution time: %3.2fs (%3.2ffps)\n", duration, nFrame/duration);
#endif
// ===================================================================
// Clean up resources
// - It is recommended to close Media SDK components first, before releasing allocated surfaces, since
// some surfaces may still be locked by internal Media SDK resources.
mfxENC.Close();
// mfxSession closed automatically on destruction
for (int i = 0; i < nEncSurfNum; i++)
delete pmfxSurfaces[i];
MSDK_SAFE_DELETE_ARRAY(pmfxSurfaces);
for(int i=0;i<taskPoolSize;i++)
MSDK_SAFE_DELETE_ARRAY(pTasks[i].mfxBS.Data);
MSDK_SAFE_DELETE_ARRAY(pTasks);
fclose(fSource);
fclose(fSink);
CleanupHWDevice();
return 0;
}
int MSDKEncode::HandleProcess()
{
mfxStatus sts = MFX_ERR_NONE;
mfxFrameSurface1* pFrameSurface = NULL;
mfxSyncPoint syncpE;
mfxEncodeCtrl* pEncCtrl = NULL;
int startP = -1;
#ifndef CONFIG_READ_RAW_BUFFER
std::map<MSDKVpp*, RING_BUFFER*>::iterator it = m_mapRingBuf.begin();
#endif
while (!m_bWantToStop) {
usleep(1000);
//Check if need to generate key frame
if (m_bForceKeyFrame)
pEncCtrl = &m_encCtrl;
else
pEncCtrl = NULL;
#ifndef CONFIG_READ_RAW_BUFFER
if (it->second->IsEmpty()) //No data
{
//printf("[MSDKEncode]-----There's no more data, just continue or exit the main loop\n");
if (it->first->GetDataEos()) //No more data in the furture
{
pEncCtrl = NULL;
pFrameSurface = NULL;
}
else
continue;
}
else
{
if (MASTER == m_type)
{
if (!m_bAccessNextElem)
continue;
it->second->Get(pFrameSurface);
}
else
it->second->Pop(pFrameSurface);
}
//printf("[MSDKEncode]-----Get next frame surface successfully\n");
#endif
if (!m_bInit) {
#ifdef CONFIG_READ_RAW_BUFFER
sts = InitEncoder(NULL);
#else
sts = InitEncoder(pFrameSurface);
#endif
if (MFX_ERR_NONE == sts) {
if (m_pMeasuremnt)
{
m_pMeasuremnt->GetLock();
pipelineinfo einfo;
einfo.mElementType = m_type;
einfo.mChannelNum = MSDKBase::nEncChannels;
MSDKBase::nEncChannels++;
m_pMeasuremnt->SetElementInfo(ENC_ENDURATION_TIME_STAMP, this, &einfo);
m_pMeasuremnt->TimeStpStart(ENC_ENDURATION_TIME_STAMP, this);
m_pMeasuremnt->RelLock();
}
m_bInit = true;
H264E_TRACE_INFO("[MSDKEncode]Encoder %p init successfully\n", this);
} else {
H264E_TRACE_ERROR("Encode init failed: %d\n", sts);
return -1;
}
}
#ifdef CONFIG_READ_RAW_BUFFER
int nIndex = GetFreeSurfaceIndex(m_pSurfacePool, m_nSurfaces); //Find free frame surface slot
if (MFX_ERR_NOT_FOUND == nIndex)
continue;
else
pFrameSurface = m_pSurfacePool[nIndex];
sts = LoadRawFrame(m_pSurfacePool[nIndex]);
if (MFX_ERR_MORE_DATA == sts) {
if (m_pInputMem->GetDataEof()) { //No more data in the future
pEncCtrl = NULL;
pFrameSurface = NULL;
} else {
continue;
}
}
//printf("[MSDKEncode]-----Get the free surface slot and complete load of the raw frame\n");
#endif
if (m_pMeasuremnt)
{
m_pMeasuremnt->GetLock();
m_pMeasuremnt->TimeStpStart(ENC_FRAME_TIME_STAMP, this);
m_pMeasuremnt->RelLock();
}
for (;;) {
//Encode a frame asychronously(returns immediately)
sts = m_pEncode->EncodeFrameAsync(pEncCtrl, pFrameSurface, &m_outputBs, &syncpE);
//printf("[MSDKEncode]-----EncodeFrameAsync ret code: %d\n", sts);
if (sts > MFX_ERR_NONE && !syncpE) { //Repeat the call if warning and no output
if (MFX_WRN_DEVICE_BUSY == sts)
usleep(1000); //wait if device is busy, then repeat the same call
} else {
if (sts > MFX_ERR_NONE && syncpE)
sts = MFX_ERR_NONE; //ignore warnings if output is available
if (MFX_ERR_NOT_ENOUGH_BUFFER == sts)
H264E_TRACE_WARNI("[MSDKEncode]-----The size of buffer allocated for encoder is too small\n");
break;
}
}
if (MFX_ERR_NONE == sts) {
sts = m_pSession->SyncOperation(syncpE, 60000);
if (m_pMeasuremnt)
{
m_pMeasuremnt->GetLock();
m_pMeasuremnt->TimeStpFinish(ENC_FRAME_TIME_STAMP, this);
m_pMeasuremnt->RelLock();
}
//check if output is key frame: startP = 0-key frame otherwise not
//A group of pictures sizeof(GROUP_OF_PICTURE*2) have a key frame
startP = (++startP)%GROUP_OF_PICTURE;
//release surface pool slot
if (MASTER == m_type) {
m_bAccessNextElem = false;
it->first->ReleaseSurface();
} else {
pFrameSurface->Data.Locked--;
}
m_pNotify->OnGetMSDKCodecData(m_outputBs.Data+m_outputBs.DataOffset, m_outputBs.DataLength, (!startP || m_bForceKeyFrame), m_nLogicIndex);
m_outputBs.DataLength = 0;
if (m_bForceKeyFrame) {
m_bForceKeyFrame = false;
//printf("[MSDKEncode]-----Force key frame successfully\n");
startP = 0;
}
}
if (!pFrameSurface)
{
if (m_pMeasuremnt)
{
m_pMeasuremnt->GetLock();
m_pMeasuremnt->TimeStpFinish(ENC_ENDURATION_TIME_STAMP, this);
m_pMeasuremnt->RelLock();
}
break;
}
}
if (!m_bWantToStop) {
H264E_TRACE_INFO("[MSDKEncode]Got EOS in Encoder %p\n", this);
m_bWantToStop = true;
}
return 0;
}
DWORD WINAPI TranscodeThread(LPVOID arg)
{
ThreadData *pData = (ThreadData *)arg;
int id = pData->id;
mfxStatus sts = MFX_ERR_NONE;
// =====================================================================
// Intel Media SDK transcode opaque pipeline setup
// - Transcode H.264 to H.264, resizing the encoded stream to half the resolution using VPP
// - Multiple streams are transcoded concurrently
// - Same input stream is used for all concurrent threadcoding threads
//
// Open input H.264 elementary stream (ES) file
FILE* fSource;
char inFile[100] = "bbb640x480.264";
fopen_s(&fSource, inFile, "rb");
MSDK_CHECK_POINTER(fSource, MFX_ERR_NULL_PTR);
// Create output elementary stream (ES) H.264 file
FILE* fSink;
char outFile[100] = "bbb320x240_xx.264";
outFile[11] = '0' + (char)(id/10);
outFile[12] = '0' + (char)(id%10);
fopen_s(&fSink, outFile, "wb");
MSDK_CHECK_POINTER(fSink, MFX_ERR_NULL_PTR);
MFXVideoSession* pmfxSession = NULL;
// Initialize Media SDK session
// - MFX_IMPL_AUTO_ANY selects HW accelaration if available (on any adapter)
// - Version 1.3 is selected since the opaque memory feature was added in this API release
// If more recent API features are needed, change the version accordingly
mfxIMPL impl = MFX_IMPL_AUTO_ANY;
mfxVersion ver = {3, 1}; // Note: API 1.3 !
pmfxSession = new MFXVideoSession;
MSDK_CHECK_POINTER(pmfxSession, MFX_ERR_NULL_PTR);
sts = pmfxSession->Init(impl, &ver);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
// Create Media SDK decoder & encoder & VPP
MFXVideoDECODE* pmfxDEC = new MFXVideoDECODE(*pmfxSession);
MSDK_CHECK_POINTER(pmfxDEC, MFX_ERR_NULL_PTR);
MFXVideoENCODE* pmfxENC = new MFXVideoENCODE(*pmfxSession);
MSDK_CHECK_POINTER(pmfxENC, MFX_ERR_NULL_PTR);
MFXVideoVPP* pmfxVPP = new MFXVideoVPP(*pmfxSession);
MSDK_CHECK_POINTER(pmfxVPP, MFX_ERR_NULL_PTR);
// Set required video parameters for decode
mfxVideoParam mfxDecParams;
memset(&mfxDecParams, 0, sizeof(mfxDecParams));
mfxDecParams.mfx.CodecId = MFX_CODEC_AVC;
mfxDecParams.IOPattern = MFX_IOPATTERN_OUT_OPAQUE_MEMORY;
// Configure Media SDK to keep more operations in flight
// - AsyncDepth represents the number of tasks that can be submitted, before synchronizing is required
// - The choice of AsyncDepth = 3 is quite arbitrary but has proven to result in good performance
mfxDecParams.AsyncDepth = 3;
// Prepare Media SDK bit stream buffer for decoder
// - Arbitrary buffer size for this example
mfxBitstream mfxBS;
memset(&mfxBS, 0, sizeof(mfxBS));
mfxBS.MaxLength = 1024 * 1024;
mfxBS.Data = new mfxU8[mfxBS.MaxLength];
MSDK_CHECK_POINTER(mfxBS.Data, MFX_ERR_MEMORY_ALLOC);
// Read a chunk of data from stream file into bit stream buffer
// - Parse bit stream, searching for header and fill video parameters structure
// - Abort if bit stream header is not found in the first bit stream buffer chunk
sts = ReadBitStreamData(&mfxBS, fSource);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
sts = pmfxDEC->DecodeHeader(&mfxBS, &mfxDecParams);
MSDK_IGNORE_MFX_STS(sts, MFX_WRN_PARTIAL_ACCELERATION);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
// Initialize VPP parameters
mfxVideoParam VPPParams;
memset(&VPPParams, 0, sizeof(VPPParams));
// Input data
VPPParams.vpp.In.FourCC = MFX_FOURCC_NV12;
VPPParams.vpp.In.ChromaFormat = MFX_CHROMAFORMAT_YUV420;
VPPParams.vpp.In.CropX = 0;
VPPParams.vpp.In.CropY = 0;
VPPParams.vpp.In.CropW = mfxDecParams.mfx.FrameInfo.CropW;
VPPParams.vpp.In.CropH = mfxDecParams.mfx.FrameInfo.CropH;
VPPParams.vpp.In.PicStruct = MFX_PICSTRUCT_PROGRESSIVE;
VPPParams.vpp.In.FrameRateExtN = 30;
VPPParams.vpp.In.FrameRateExtD = 1;
// width must be a multiple of 16
// height must be a multiple of 16 in case of frame picture and a multiple of 32 in case of field picture
VPPParams.vpp.In.Width = MSDK_ALIGN16(VPPParams.vpp.In.CropW);
VPPParams.vpp.In.Height = (MFX_PICSTRUCT_PROGRESSIVE == VPPParams.vpp.In.PicStruct)?
MSDK_ALIGN16(VPPParams.vpp.In.CropH) : MSDK_ALIGN32(VPPParams.vpp.In.CropH);
// Output data
VPPParams.vpp.Out.FourCC = MFX_FOURCC_NV12;
VPPParams.vpp.Out.ChromaFormat = MFX_CHROMAFORMAT_YUV420;
VPPParams.vpp.Out.CropX = 0;
VPPParams.vpp.Out.CropY = 0;
VPPParams.vpp.Out.CropW = VPPParams.vpp.In.CropW/2; // Half the resolution of decode stream
VPPParams.vpp.Out.CropH = VPPParams.vpp.In.CropH/2;
VPPParams.vpp.Out.PicStruct = MFX_PICSTRUCT_PROGRESSIVE;
VPPParams.vpp.Out.FrameRateExtN = 30;
VPPParams.vpp.Out.FrameRateExtD = 1;
// width must be a multiple of 16
// height must be a multiple of 16 in case of frame picture and a multiple of 32 in case of field picture
VPPParams.vpp.Out.Width = MSDK_ALIGN16(VPPParams.vpp.Out.CropW);
VPPParams.vpp.Out.Height = (MFX_PICSTRUCT_PROGRESSIVE == VPPParams.vpp.Out.PicStruct)?
MSDK_ALIGN16(VPPParams.vpp.Out.CropH) : MSDK_ALIGN32(VPPParams.vpp.Out.CropH);
VPPParams.IOPattern = MFX_IOPATTERN_IN_OPAQUE_MEMORY | MFX_IOPATTERN_OUT_OPAQUE_MEMORY;
// Configure Media SDK to keep more operations in flight
// - AsyncDepth represents the number of tasks that can be submitted, before synchronizing is required
VPPParams.AsyncDepth = mfxDecParams.AsyncDepth;
// Initialize encoder parameters
mfxVideoParam mfxEncParams;
memset(&mfxEncParams, 0, sizeof(mfxEncParams));
mfxEncParams.mfx.CodecId = MFX_CODEC_AVC;
mfxEncParams.mfx.TargetUsage = MFX_TARGETUSAGE_BALANCED;
mfxEncParams.mfx.TargetKbps = 500;
mfxEncParams.mfx.RateControlMethod = MFX_RATECONTROL_VBR;
mfxEncParams.mfx.FrameInfo.FrameRateExtN = 30;
mfxEncParams.mfx.FrameInfo.FrameRateExtD = 1;
mfxEncParams.mfx.FrameInfo.FourCC = MFX_FOURCC_NV12;
mfxEncParams.mfx.FrameInfo.ChromaFormat = MFX_CHROMAFORMAT_YUV420;
mfxEncParams.mfx.FrameInfo.PicStruct = MFX_PICSTRUCT_PROGRESSIVE;
mfxEncParams.mfx.FrameInfo.CropX = 0;
mfxEncParams.mfx.FrameInfo.CropY = 0;
mfxEncParams.mfx.FrameInfo.CropW = VPPParams.vpp.Out.CropW; // Half the resolution of decode stream
mfxEncParams.mfx.FrameInfo.CropH = VPPParams.vpp.Out.CropH;
// width must be a multiple of 16
// height must be a multiple of 16 in case of frame picture and a multiple of 32 in case of field picture
mfxEncParams.mfx.FrameInfo.Width = MSDK_ALIGN16(mfxEncParams.mfx.FrameInfo.CropW);
mfxEncParams.mfx.FrameInfo.Height = (MFX_PICSTRUCT_PROGRESSIVE == mfxEncParams.mfx.FrameInfo.PicStruct)?
MSDK_ALIGN16(mfxEncParams.mfx.FrameInfo.CropH) : MSDK_ALIGN32(mfxEncParams.mfx.FrameInfo.CropH);
mfxEncParams.IOPattern = MFX_IOPATTERN_IN_OPAQUE_MEMORY;
// Configure Media SDK to keep more operations in flight
// - AsyncDepth represents the number of tasks that can be submitted, before synchronizing is required
mfxEncParams.AsyncDepth = mfxDecParams.AsyncDepth;
// Query number required surfaces for decoder
mfxFrameAllocRequest DecRequest;
memset(&DecRequest, 0, sizeof(DecRequest));
sts = pmfxDEC->QueryIOSurf(&mfxDecParams, &DecRequest);
MSDK_IGNORE_MFX_STS(sts, MFX_WRN_PARTIAL_ACCELERATION);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
// Query number required surfaces for encoder
mfxFrameAllocRequest EncRequest;
memset(&EncRequest, 0, sizeof(EncRequest));
sts = pmfxENC->QueryIOSurf(&mfxEncParams, &EncRequest);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
// Query number of required surfaces for VPP
mfxFrameAllocRequest VPPRequest[2];// [0] - in, [1] - out
memset(&VPPRequest, 0, sizeof(mfxFrameAllocRequest)*2);
sts = pmfxVPP->QueryIOSurf(&VPPParams, VPPRequest);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
// Determine the required number of surfaces for decoder output (VPP input) and for VPP output (encoder input)
mfxU16 nSurfNumDecVPP = DecRequest.NumFrameSuggested + VPPRequest[0].NumFrameSuggested + VPPParams.AsyncDepth;
mfxU16 nSurfNumVPPEnc = EncRequest.NumFrameSuggested + VPPRequest[1].NumFrameSuggested + VPPParams.AsyncDepth;
// Initialize shared surfaces for decoder, VPP and encode
// - Note that no buffer memory is allocated, for opaque memory this is handled by Media SDK internally
// - Frame surface array keeps reference to all surfaces
// - Opaque memory is configured with the mfxExtOpaqueSurfaceAlloc extended buffers
mfxFrameSurface1** pSurfaces = new mfxFrameSurface1*[nSurfNumDecVPP];
MSDK_CHECK_POINTER(pSurfaces, MFX_ERR_MEMORY_ALLOC);
for (int i = 0; i < nSurfNumDecVPP; i++)
{
pSurfaces[i] = new mfxFrameSurface1;
MSDK_CHECK_POINTER(pSurfaces[i], MFX_ERR_MEMORY_ALLOC);
memset(pSurfaces[i], 0, sizeof(mfxFrameSurface1));
memcpy(&(pSurfaces[i]->Info), &(DecRequest.Info), sizeof(mfxFrameInfo));
}
mfxFrameSurface1** pSurfaces2 = new mfxFrameSurface1*[nSurfNumVPPEnc];
MSDK_CHECK_POINTER(pSurfaces2, MFX_ERR_MEMORY_ALLOC);
for (int i = 0; i < nSurfNumVPPEnc; i++)
{
pSurfaces2[i] = new mfxFrameSurface1;
MSDK_CHECK_POINTER(pSurfaces2[i], MFX_ERR_MEMORY_ALLOC);
memset(pSurfaces2[i], 0, sizeof(mfxFrameSurface1));
memcpy(&(pSurfaces2[i]->Info), &(EncRequest.Info), sizeof(mfxFrameInfo));
}
mfxExtOpaqueSurfaceAlloc extOpaqueAllocDec;
memset(&extOpaqueAllocDec, 0, sizeof(extOpaqueAllocDec));
extOpaqueAllocDec.Header.BufferId = MFX_EXTBUFF_OPAQUE_SURFACE_ALLOCATION;
extOpaqueAllocDec.Header.BufferSz = sizeof(mfxExtOpaqueSurfaceAlloc);
mfxExtBuffer* pExtParamsDec = (mfxExtBuffer*)&extOpaqueAllocDec;
mfxExtOpaqueSurfaceAlloc extOpaqueAllocVPP;
memset(&extOpaqueAllocVPP, 0, sizeof(extOpaqueAllocVPP));
extOpaqueAllocVPP.Header.BufferId = MFX_EXTBUFF_OPAQUE_SURFACE_ALLOCATION;
extOpaqueAllocVPP.Header.BufferSz = sizeof(mfxExtOpaqueSurfaceAlloc);
mfxExtBuffer* pExtParamsVPP = (mfxExtBuffer*)&extOpaqueAllocVPP;
mfxExtOpaqueSurfaceAlloc extOpaqueAllocEnc;
memset(&extOpaqueAllocEnc, 0, sizeof(extOpaqueAllocEnc));
extOpaqueAllocEnc.Header.BufferId = MFX_EXTBUFF_OPAQUE_SURFACE_ALLOCATION;
extOpaqueAllocEnc.Header.BufferSz = sizeof(mfxExtOpaqueSurfaceAlloc);
mfxExtBuffer* pExtParamsENC = (mfxExtBuffer*)&extOpaqueAllocEnc;
extOpaqueAllocDec.Out.Surfaces = pSurfaces;
extOpaqueAllocDec.Out.NumSurface = nSurfNumDecVPP;
extOpaqueAllocDec.Out.Type = DecRequest.Type;
memcpy(&extOpaqueAllocVPP.In, &extOpaqueAllocDec.Out, sizeof(extOpaqueAllocDec.Out));
extOpaqueAllocVPP.Out.Surfaces = pSurfaces2;
extOpaqueAllocVPP.Out.NumSurface = nSurfNumVPPEnc;
extOpaqueAllocVPP.Out.Type = EncRequest.Type;
memcpy(&extOpaqueAllocEnc.In, &extOpaqueAllocVPP.Out, sizeof(extOpaqueAllocVPP.Out));
mfxDecParams.ExtParam = &pExtParamsDec;
mfxDecParams.NumExtParam = 1;
VPPParams.ExtParam = &pExtParamsVPP;
VPPParams.NumExtParam = 1;
mfxEncParams.ExtParam = &pExtParamsENC;
mfxEncParams.NumExtParam = 1;
// Initialize the Media SDK decoder
sts = pmfxDEC->Init(&mfxDecParams);
MSDK_IGNORE_MFX_STS(sts, MFX_WRN_PARTIAL_ACCELERATION);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
// Initialize the Media SDK encoder
sts = pmfxENC->Init(&mfxEncParams);
MSDK_IGNORE_MFX_STS(sts, MFX_WRN_PARTIAL_ACCELERATION);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
// Initialize Media SDK VPP
sts = pmfxVPP->Init(&VPPParams);
MSDK_IGNORE_MFX_STS(sts, MFX_WRN_PARTIAL_ACCELERATION);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
// Retrieve video parameters selected by encoder.
// - BufferSizeInKB parameter is required to set bit stream buffer size
mfxVideoParam par;
memset(&par, 0, sizeof(par));
sts = pmfxENC->GetVideoParam(&par);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
// Create task pool to improve asynchronous performance (greater GPU utilization)
mfxU16 taskPoolSize = mfxEncParams.AsyncDepth; // number of tasks that can be submitted, before synchronizing is required
Task* pTasks = new Task[taskPoolSize];
memset(pTasks, 0, sizeof(Task) * taskPoolSize);
for(int i=0;i<taskPoolSize;i++)
{
// Prepare Media SDK bit stream buffer
pTasks[i].mfxBS.MaxLength = par.mfx.BufferSizeInKB * 1000;
pTasks[i].mfxBS.Data = new mfxU8[pTasks[i].mfxBS.MaxLength];
MSDK_CHECK_POINTER(pTasks[i].mfxBS.Data, MFX_ERR_MEMORY_ALLOC);
}
// ===================================
// Start transcoding the frames
//
#ifdef ENABLE_BENCHMARK
LARGE_INTEGER tStart, tEnd;
QueryPerformanceFrequency(&tStart);
double freq = (double)tStart.QuadPart;
QueryPerformanceCounter(&tStart);
#endif
mfxSyncPoint syncpD, syncpV;
mfxFrameSurface1* pmfxOutSurface = NULL;
mfxU32 nFrame = 0;
int nIndex = 0;
int nIndex2 = 0;
int nFirstSyncTask = 0;
int nTaskIdx = 0;
//
// Stage 1: Main transcoding loop
//
while (MFX_ERR_NONE <= sts || MFX_ERR_MORE_DATA == sts || MFX_ERR_MORE_SURFACE == sts)
{
nTaskIdx = GetFreeTaskIndex(pTasks, taskPoolSize); // Find free task
if(MFX_ERR_NOT_FOUND == nTaskIdx)
{
// No more free tasks, need to sync
sts = pmfxSession->SyncOperation(pTasks[nFirstSyncTask].syncp, 60000);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
sts = WriteBitStreamFrame(&pTasks[nFirstSyncTask].mfxBS, fSink);
MSDK_BREAK_ON_ERROR(sts);
pTasks[nFirstSyncTask].syncp = NULL;
pTasks[nFirstSyncTask].mfxBS.DataLength = 0;
pTasks[nFirstSyncTask].mfxBS.DataOffset = 0;
nFirstSyncTask = (nFirstSyncTask + 1) % taskPoolSize;
++nFrame;
#ifdef ENABLE_OUTPUT
if((nFrame % 100) == 0)
printf("(%d) Frame number: %d\n", id, nFrame);
#endif
}
else
{
if (MFX_WRN_DEVICE_BUSY == sts)
Sleep(1); // just wait and then repeat the same call to DecodeFrameAsync
if (MFX_ERR_MORE_DATA == sts)
{
sts = ReadBitStreamData(&mfxBS, fSource); // Read more data to input bit stream
MSDK_BREAK_ON_ERROR(sts);
}
if (MFX_ERR_MORE_SURFACE == sts || MFX_ERR_NONE == sts)
{
nIndex = GetFreeSurfaceIndex(pSurfaces, nSurfNumDecVPP); // Find free frame surface
if (MFX_ERR_NOT_FOUND == nIndex)
return MFX_ERR_MEMORY_ALLOC;
}
// Decode a frame asychronously (returns immediately)
sts = pmfxDEC->DecodeFrameAsync(&mfxBS, pSurfaces[nIndex], &pmfxOutSurface, &syncpD);
// Ignore warnings if output is available,
// if no output and no action required just repeat the DecodeFrameAsync call
if (MFX_ERR_NONE < sts && syncpD)
sts = MFX_ERR_NONE;
if (MFX_ERR_NONE == sts)
{
nIndex2 = GetFreeSurfaceIndex(pSurfaces2, nSurfNumVPPEnc); // Find free frame surface
if (MFX_ERR_NOT_FOUND == nIndex)
return MFX_ERR_MEMORY_ALLOC;
for (;;)
{
// Process a frame asychronously (returns immediately)
sts = pmfxVPP->RunFrameVPPAsync(pmfxOutSurface, pSurfaces2[nIndex2], NULL, &syncpV);
if (MFX_ERR_NONE < sts && !syncpV) // repeat the call if warning and no output
{
if (MFX_WRN_DEVICE_BUSY == sts)
Sleep(1); // wait if device is busy
}
else if (MFX_ERR_NONE < sts && syncpV)
{
sts = MFX_ERR_NONE; // ignore warnings if output is available
break;
}
else
break; // not a warning
}
// VPP needs more data, let decoder decode another frame as input
if (MFX_ERR_MORE_DATA == sts)
{
continue;
}
else if (MFX_ERR_MORE_SURFACE == sts)
{
// Not relevant for the illustrated workload! Therefore not handled.
// Relevant for cases when VPP produces more frames at output than consumes at input. E.g. framerate conversion 30 fps -> 60 fps
break;
}
else
MSDK_BREAK_ON_ERROR(sts);
for (;;)
{
// Encode a frame asychronously (returns immediately)
sts = pmfxENC->EncodeFrameAsync(NULL, pSurfaces2[nIndex2], &pTasks[nTaskIdx].mfxBS, &pTasks[nTaskIdx].syncp);
if (MFX_ERR_NONE < sts && !pTasks[nTaskIdx].syncp) // repeat the call if warning and no output
{
if (MFX_WRN_DEVICE_BUSY == sts)
Sleep(1); // wait if device is busy
}
else if (MFX_ERR_NONE < sts && pTasks[nTaskIdx].syncp)
{
sts = MFX_ERR_NONE; // ignore warnings if output is available
break;
}
else if (MFX_ERR_NOT_ENOUGH_BUFFER == sts)
{
// Allocate more bitstream buffer memory here if needed...
break;
}
else
break;
}
}
}
}
// MFX_ERR_MORE_DATA means that file has ended, need to go to buffering loop, exit in case of other errors
MSDK_IGNORE_MFX_STS(sts, MFX_ERR_MORE_DATA);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
//
// Stage 2: Retrieve the buffered decoded frames
//
while (MFX_ERR_NONE <= sts || MFX_ERR_MORE_SURFACE == sts)
{
nTaskIdx = GetFreeTaskIndex(pTasks, taskPoolSize); // Find free task
if(MFX_ERR_NOT_FOUND == nTaskIdx)
{
// No more free tasks, need to sync
sts = pmfxSession->SyncOperation(pTasks[nFirstSyncTask].syncp, 60000);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
sts = WriteBitStreamFrame(&pTasks[nFirstSyncTask].mfxBS, fSink);
MSDK_BREAK_ON_ERROR(sts);
pTasks[nFirstSyncTask].syncp = NULL;
pTasks[nFirstSyncTask].mfxBS.DataLength = 0;
pTasks[nFirstSyncTask].mfxBS.DataOffset = 0;
nFirstSyncTask = (nFirstSyncTask + 1) % taskPoolSize;
++nFrame;
#ifdef ENABLE_OUTPUT
if((nFrame % 100) == 0)
printf("(%d) Frame number: %d\n", id, nFrame);
#endif
}
else
{
if (MFX_WRN_DEVICE_BUSY == sts)
Sleep(1);
nIndex = GetFreeSurfaceIndex(pSurfaces, nSurfNumDecVPP); // Find free frame surface
if (MFX_ERR_NOT_FOUND == nIndex)
return MFX_ERR_MEMORY_ALLOC;
// Decode a frame asychronously (returns immediately)
sts = pmfxDEC->DecodeFrameAsync(NULL, pSurfaces[nIndex], &pmfxOutSurface, &syncpD);
// Ignore warnings if output is available,
// if no output and no action required just repeat the DecodeFrameAsync call
if (MFX_ERR_NONE < sts && syncpD)
sts = MFX_ERR_NONE;
if (MFX_ERR_NONE == sts)
{
nIndex2 = GetFreeSurfaceIndex(pSurfaces2, nSurfNumVPPEnc); // Find free frame surface
if (MFX_ERR_NOT_FOUND == nIndex)
return MFX_ERR_MEMORY_ALLOC;
for (;;)
{
// Process a frame asychronously (returns immediately)
sts = pmfxVPP->RunFrameVPPAsync(pmfxOutSurface, pSurfaces2[nIndex2], NULL, &syncpV);
if (MFX_ERR_NONE < sts && !syncpV) // repeat the call if warning and no output
{
if (MFX_WRN_DEVICE_BUSY == sts)
Sleep(1); // wait if device is busy
}
else if (MFX_ERR_NONE < sts && syncpV)
{
sts = MFX_ERR_NONE; // ignore warnings if output is available
break;
}
else
break; // not a warning
}
// VPP needs more data, let decoder decode another frame as input
if (MFX_ERR_MORE_DATA == sts)
{
continue;
}
else if (MFX_ERR_MORE_SURFACE == sts)
{
// Not relevant for the illustrated workload! Therefore not handled.
// Relevant for cases when VPP produces more frames at output than consumes at input. E.g. framerate conversion 30 fps -> 60 fps
break;
}
else
MSDK_BREAK_ON_ERROR(sts);
for (;;)
{
// Encode a frame asychronously (returns immediately)
sts = pmfxENC->EncodeFrameAsync(NULL, pSurfaces2[nIndex2], &pTasks[nTaskIdx].mfxBS, &pTasks[nTaskIdx].syncp);
if (MFX_ERR_NONE < sts && !pTasks[nTaskIdx].syncp) // repeat the call if warning and no output
{
if (MFX_WRN_DEVICE_BUSY == sts)
Sleep(1); // wait if device is busy
}
else if (MFX_ERR_NONE < sts && pTasks[nTaskIdx].syncp)
{
sts = MFX_ERR_NONE; // ignore warnings if output is available
break;
}
else if (MFX_ERR_NOT_ENOUGH_BUFFER == sts)
{
// Allocate more bitstream buffer memory here if needed...
break;
}
else
break;
}
}
}
}
// MFX_ERR_MORE_DATA indicates that all decode buffers has been fetched, exit in case of other errors
MSDK_IGNORE_MFX_STS(sts, MFX_ERR_MORE_DATA);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
//
// Stage 3: Retrieve buffered frames from VPP
//
while (MFX_ERR_NONE <= sts || MFX_ERR_MORE_DATA == sts || MFX_ERR_MORE_SURFACE == sts)
{
nTaskIdx = GetFreeTaskIndex(pTasks, taskPoolSize); // Find free task
if(MFX_ERR_NOT_FOUND == nTaskIdx)
{
// No more free tasks, need to sync
sts = pmfxSession->SyncOperation(pTasks[nFirstSyncTask].syncp, 60000);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
sts = WriteBitStreamFrame(&pTasks[nFirstSyncTask].mfxBS, fSink);
MSDK_BREAK_ON_ERROR(sts);
pTasks[nFirstSyncTask].syncp = NULL;
pTasks[nFirstSyncTask].mfxBS.DataLength = 0;
pTasks[nFirstSyncTask].mfxBS.DataOffset = 0;
nFirstSyncTask = (nFirstSyncTask + 1) % taskPoolSize;
++nFrame;
#ifdef ENABLE_OUTPUT
if((nFrame % 100) == 0)
printf("(%d) Frame number: %d\n", id, nFrame);
#endif
}
else
{
nIndex2 = GetFreeSurfaceIndex(pSurfaces2, nSurfNumVPPEnc); // Find free frame surface
if (MFX_ERR_NOT_FOUND == nIndex)
return MFX_ERR_MEMORY_ALLOC;
for (;;)
{
// Process a frame asychronously (returns immediately)
sts = pmfxVPP->RunFrameVPPAsync(NULL, pSurfaces2[nIndex2], NULL, &syncpV);
if (MFX_ERR_NONE < sts && !syncpV) // repeat the call if warning and no output
{
if (MFX_WRN_DEVICE_BUSY == sts)
Sleep(1); // wait if device is busy
}
else if (MFX_ERR_NONE < sts && syncpV)
{
sts = MFX_ERR_NONE; // ignore warnings if output is available
break;
}
else
break; // not a warning
}
if (MFX_ERR_MORE_SURFACE == sts)
{
// Not relevant for the illustrated workload! Therefore not handled.
// Relevant for cases when VPP produces more frames at output than consumes at input. E.g. framerate conversion 30 fps -> 60 fps
break;
}
else
MSDK_BREAK_ON_ERROR(sts);
for (;;)
{
// Encode a frame asychronously (returns immediately)
sts = pmfxENC->EncodeFrameAsync(NULL, pSurfaces2[nIndex2], &pTasks[nTaskIdx].mfxBS, &pTasks[nTaskIdx].syncp);
if (MFX_ERR_NONE < sts && !pTasks[nTaskIdx].syncp) // repeat the call if warning and no output
{
if (MFX_WRN_DEVICE_BUSY == sts)
Sleep(1); // wait if device is busy
}
else if (MFX_ERR_NONE < sts && pTasks[nTaskIdx].syncp)
{
sts = MFX_ERR_NONE; // ignore warnings if output is available
break;
}
else if (MFX_ERR_NOT_ENOUGH_BUFFER == sts)
{
// Allocate more bitstream buffer memory here if needed...
break;
}
else
break;
}
}
}
// MFX_ERR_MORE_DATA indicates that all VPP buffers has been fetched, exit in case of other errors
MSDK_IGNORE_MFX_STS(sts, MFX_ERR_MORE_DATA);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
//
// Stage 4: Retrieve the buffered encoded frames
//
while (MFX_ERR_NONE <= sts)
{
nTaskIdx = GetFreeTaskIndex(pTasks, taskPoolSize); // Find free task
if(MFX_ERR_NOT_FOUND == nTaskIdx)
{
// No more free tasks, need to sync
sts = pmfxSession->SyncOperation(pTasks[nFirstSyncTask].syncp, 60000);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
sts = WriteBitStreamFrame(&pTasks[nFirstSyncTask].mfxBS, fSink);
MSDK_BREAK_ON_ERROR(sts);
pTasks[nFirstSyncTask].syncp = NULL;
pTasks[nFirstSyncTask].mfxBS.DataLength = 0;
pTasks[nFirstSyncTask].mfxBS.DataOffset = 0;
nFirstSyncTask = (nFirstSyncTask + 1) % taskPoolSize;
++nFrame;
#ifdef ENABLE_OUTPUT
if((nFrame % 100) == 0)
printf("(%d) Frame number: %d\n", id, nFrame);
#endif
}
else
{
for (;;)
{
// Encode a frame asychronously (returns immediately)
sts = pmfxENC->EncodeFrameAsync(NULL, NULL, &pTasks[nTaskIdx].mfxBS, &pTasks[nTaskIdx].syncp);
if (MFX_ERR_NONE < sts && !pTasks[nTaskIdx].syncp) // repeat the call if warning and no output
{
if (MFX_WRN_DEVICE_BUSY == sts)
Sleep(1); // wait if device is busy
}
else if (MFX_ERR_NONE < sts && pTasks[nTaskIdx].syncp)
{
sts = MFX_ERR_NONE; // ignore warnings if output is available
break;
}
else
break;
}
}
}
// MFX_ERR_MORE_DATA indicates that there are no more buffered frames, exit in case of other errors
MSDK_IGNORE_MFX_STS(sts, MFX_ERR_MORE_DATA);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
//
// Stage 5: Sync all remaining tasks in task pool
//
while(pTasks[nFirstSyncTask].syncp)
{
sts = pmfxSession->SyncOperation(pTasks[nFirstSyncTask].syncp, 60000);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
sts = WriteBitStreamFrame(&pTasks[nFirstSyncTask].mfxBS, fSink);
MSDK_BREAK_ON_ERROR(sts);
pTasks[nFirstSyncTask].syncp = NULL;
pTasks[nFirstSyncTask].mfxBS.DataLength = 0;
pTasks[nFirstSyncTask].mfxBS.DataOffset = 0;
nFirstSyncTask = (nFirstSyncTask + 1) % taskPoolSize;
++nFrame;
#ifdef ENABLE_OUTPUT
if((nFrame % 100) == 0)
printf("(%d) Frame number: %d\n", id, nFrame);
#endif
}
#ifdef ENABLE_BENCHMARK
QueryPerformanceCounter(&tEnd);
double duration = ((double)tEnd.QuadPart - (double)tStart.QuadPart) / freq;
printf("\n[%d] Execution time: %3.2fs (%3.2ffps)\n", pData->id, duration, nFrame/duration);
#endif
// ===================================================================
// Clean up resources
// - It is recommended to close Media SDK components first, before releasing allocated surfaces, since
// some surfaces may still be locked by internal Media SDK resources.
pmfxENC->Close();
pmfxDEC->Close();
pmfxVPP->Close();
delete pmfxENC;
delete pmfxDEC;
delete pmfxVPP;
pmfxSession->Close();
delete pmfxSession;
for (int i = 0; i < nSurfNumDecVPP; i++)
delete pSurfaces[i];
for (int i = 0; i < nSurfNumVPPEnc; i++)
delete pSurfaces2[i];
MSDK_SAFE_DELETE_ARRAY(pSurfaces);
MSDK_SAFE_DELETE_ARRAY(pSurfaces2);
MSDK_SAFE_DELETE_ARRAY(mfxBS.Data);
for(int i=0;i<taskPoolSize;i++)
MSDK_SAFE_DELETE_ARRAY(pTasks[i].mfxBS.Data);
MSDK_SAFE_DELETE_ARRAY(pTasks);
fclose(fSource);
fclose(fSink);
return 0;
}
int main()
{
mfxStatus sts = MFX_ERR_NONE;
mfxU16 inputWidth = 1920;
mfxU16 inputHeight = 1080;
// =====================================================================
// Intel Media SDK Video Pre/Post Processing (VPP) pipeline setup
// - Showcasing two VPP features
// - Resize (frame width and height is halved)
// - ProcAmp: Increase brightness
// - Video memory surfaces are used
//
// Open input YV12 YUV file
FILE* fSource;
fopen_s(&fSource, "bbb1920x1080.yuv", "rb");
MSDK_CHECK_POINTER(fSource, MFX_ERR_NULL_PTR);
// Create output YUV file
FILE* fSink;
fopen_s(&fSink, "bbb960x540_vpp_bright_d3d.yuv", "wb");
MSDK_CHECK_POINTER(fSink, MFX_ERR_NULL_PTR);
// Initialize Media SDK session
// - MFX_IMPL_AUTO_ANY selects HW accelaration if available (on any adapter)
// - Version 1.0 is selected for greatest backwards compatibility.
// If more recent API features are needed, change the version accordingly
mfxIMPL impl = MFX_IMPL_AUTO_ANY;
#ifdef DX11_D3D
impl |= MFX_IMPL_VIA_D3D11;
#endif
mfxVersion ver = {0, 1};
MFXVideoSession mfxSession;
sts = mfxSession.Init(impl, &ver);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
// Initialize VPP parameters
mfxVideoParam VPPParams;
memset(&VPPParams, 0, sizeof(VPPParams));
// Input data
VPPParams.vpp.In.FourCC = MFX_FOURCC_NV12;
VPPParams.vpp.In.ChromaFormat = MFX_CHROMAFORMAT_YUV420;
VPPParams.vpp.In.CropX = 0;
VPPParams.vpp.In.CropY = 0;
VPPParams.vpp.In.CropW = inputWidth;
VPPParams.vpp.In.CropH = inputHeight;
VPPParams.vpp.In.PicStruct = MFX_PICSTRUCT_PROGRESSIVE;
VPPParams.vpp.In.FrameRateExtN = 30;
VPPParams.vpp.In.FrameRateExtD = 1;
// width must be a multiple of 16
// height must be a multiple of 16 in case of frame picture and a multiple of 32 in case of field picture
VPPParams.vpp.In.Width = MSDK_ALIGN16(inputWidth);
VPPParams.vpp.In.Height = (MFX_PICSTRUCT_PROGRESSIVE == VPPParams.vpp.In.PicStruct)?
MSDK_ALIGN16(inputHeight) : MSDK_ALIGN32(inputHeight);
// Output data
VPPParams.vpp.Out.FourCC = MFX_FOURCC_NV12;
VPPParams.vpp.Out.ChromaFormat = MFX_CHROMAFORMAT_YUV420;
VPPParams.vpp.Out.CropX = 0;
VPPParams.vpp.Out.CropY = 0;
VPPParams.vpp.Out.CropW = inputWidth/2;
VPPParams.vpp.Out.CropH = inputHeight/2;
VPPParams.vpp.Out.PicStruct = MFX_PICSTRUCT_PROGRESSIVE;
VPPParams.vpp.Out.FrameRateExtN = 30;
VPPParams.vpp.Out.FrameRateExtD = 1;
// width must be a multiple of 16
// height must be a multiple of 16 in case of frame picture and a multiple of 32 in case of field picture
VPPParams.vpp.Out.Width = MSDK_ALIGN16(VPPParams.vpp.Out.CropW);
VPPParams.vpp.Out.Height = (MFX_PICSTRUCT_PROGRESSIVE == VPPParams.vpp.Out.PicStruct)?
MSDK_ALIGN16(VPPParams.vpp.Out.CropH) : MSDK_ALIGN32(VPPParams.vpp.Out.CropH);
VPPParams.IOPattern = MFX_IOPATTERN_IN_VIDEO_MEMORY | MFX_IOPATTERN_OUT_VIDEO_MEMORY;
// Create Media SDK VPP component
MFXVideoVPP mfxVPP(mfxSession);
// Create DirectX device context
mfxHDL deviceHandle;
sts = CreateHWDevice(mfxSession, &deviceHandle, NULL);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
// Provide device manager to Media SDK
sts = mfxSession.SetHandle(DEVICE_MGR_TYPE, deviceHandle);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
mfxFrameAllocator mfxAllocator;
mfxAllocator.Alloc = simple_alloc;
mfxAllocator.Free = simple_free;
mfxAllocator.Lock = simple_lock;
mfxAllocator.Unlock = simple_unlock;
mfxAllocator.GetHDL = simple_gethdl;
// When using video memory we must provide Media SDK with an external allocator
sts = mfxSession.SetFrameAllocator(&mfxAllocator);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
// Query number of required surfaces for VPP
mfxFrameAllocRequest VPPRequest[2];// [0] - in, [1] - out
memset(&VPPRequest, 0, sizeof(mfxFrameAllocRequest)*2);
sts = mfxVPP.QueryIOSurf(&VPPParams, VPPRequest);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
#ifdef DX11_D3D
VPPRequest[0].Type |= WILL_WRITE; // Hint to DX11 memory handler that application will write data to input surfaces
VPPRequest[1].Type |= WILL_READ; // Hint to DX11 memory handler that application will read data from output surfaces
#endif
// Allocate required surfaces
mfxFrameAllocResponse mfxResponseIn;
mfxFrameAllocResponse mfxResponseOut;
sts = mfxAllocator.Alloc(mfxAllocator.pthis, &VPPRequest[0], &mfxResponseIn);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
sts = mfxAllocator.Alloc(mfxAllocator.pthis, &VPPRequest[1], &mfxResponseOut);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
mfxU16 nVPPSurfNumIn = mfxResponseIn.NumFrameActual;
mfxU16 nVPPSurfNumOut = mfxResponseOut.NumFrameActual;
// Allocate surface headers (mfxFrameSurface1) for VPP
mfxFrameSurface1** pVPPSurfacesIn = new mfxFrameSurface1*[nVPPSurfNumIn];
MSDK_CHECK_POINTER(pVPPSurfacesIn, MFX_ERR_MEMORY_ALLOC);
for (int i = 0; i < nVPPSurfNumIn; i++)
{
pVPPSurfacesIn[i] = new mfxFrameSurface1;
memset(pVPPSurfacesIn[i], 0, sizeof(mfxFrameSurface1));
memcpy(&(pVPPSurfacesIn[i]->Info), &(VPPParams.vpp.In), sizeof(mfxFrameInfo));
pVPPSurfacesIn[i]->Data.MemId = mfxResponseIn.mids[i]; // MID (memory id) represent one D3D NV12 surface
#ifndef ENABLE_INPUT
// In case simulating direct access to frames we initialize the allocated surfaces with default pattern
// - For true benchmark comparisons to async workloads all surfaces must have the same data
#ifndef DX11_D3D
IDirect3DSurface9 *pSurface;
D3DSURFACE_DESC desc;
D3DLOCKED_RECT locked;
pSurface = (IDirect3DSurface9 *)mfxResponseIn.mids[i];
pSurface->GetDesc(&desc);
pSurface->LockRect(&locked, 0, D3DLOCK_NOSYSLOCK);
memset((mfxU8 *)locked.pBits, 100, desc.Height*locked.Pitch); // Y plane
memset((mfxU8 *)locked.pBits + desc.Height * locked.Pitch, 50, (desc.Height*locked.Pitch)/2); // UV plane
pSurface->UnlockRect();
#else
// For now, just leave D3D11 surface data uninitialized
#endif
#endif
}
mfxFrameSurface1** pVPPSurfacesOut = new mfxFrameSurface1*[nVPPSurfNumOut];
MSDK_CHECK_POINTER(pVPPSurfacesOut, MFX_ERR_MEMORY_ALLOC);
for (int i = 0; i < nVPPSurfNumOut; i++)
{
pVPPSurfacesOut[i] = new mfxFrameSurface1;
memset(pVPPSurfacesOut[i], 0, sizeof(mfxFrameSurface1));
memcpy(&(pVPPSurfacesOut[i]->Info), &(VPPParams.vpp.Out), sizeof(mfxFrameInfo));
pVPPSurfacesOut[i]->Data.MemId = mfxResponseOut.mids[i]; // MID (memory id) represent one D3D NV12 surface
}
// Initialize extended buffer for frame processing
// - Process amplifier (ProcAmp) used to control brightness
// - mfxExtVPPDoUse: Define the processing algorithm to be used
// - mfxExtVPPProcAmp: ProcAmp configuration
// - mfxExtBuffer: Add extended buffers to VPP parameter configuration
mfxExtVPPDoUse extDoUse;
mfxU32 tabDoUseAlg[1];
extDoUse.Header.BufferId = MFX_EXTBUFF_VPP_DOUSE;
extDoUse.Header.BufferSz = sizeof(mfxExtVPPDoUse);
extDoUse.NumAlg = 1;
extDoUse.AlgList = tabDoUseAlg;
tabDoUseAlg[0] = MFX_EXTBUFF_VPP_PROCAMP;
mfxExtVPPProcAmp procampConfig;
procampConfig.Header.BufferId = MFX_EXTBUFF_VPP_PROCAMP;
procampConfig.Header.BufferSz = sizeof(mfxExtVPPProcAmp);
procampConfig.Hue = 0.0f; // Default
procampConfig.Saturation = 1.0f; // Default
procampConfig.Contrast = 1.0; // Default
procampConfig.Brightness = 40.0; // Adjust brightness
mfxExtBuffer* ExtBuffer[2];
ExtBuffer[0] = (mfxExtBuffer*)&extDoUse;
ExtBuffer[1] = (mfxExtBuffer*)&procampConfig;
VPPParams.NumExtParam = 2;
VPPParams.ExtParam = (mfxExtBuffer**)&ExtBuffer[0];
// Initialize Media SDK VPP
sts = mfxVPP.Init(&VPPParams);
MSDK_IGNORE_MFX_STS(sts, MFX_WRN_PARTIAL_ACCELERATION);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
// ===================================
// Start processing the frames
//
#ifdef ENABLE_BENCHMARK
LARGE_INTEGER tStart, tEnd;
QueryPerformanceFrequency(&tStart);
double freq = (double)tStart.QuadPart;
QueryPerformanceCounter(&tStart);
#endif
int nSurfIdxIn = 0, nSurfIdxOut = 0;
mfxSyncPoint syncp;
mfxU32 nFrame = 0;
//
// Stage 1: Main processing loop
//
while (MFX_ERR_NONE <= sts || MFX_ERR_MORE_DATA == sts)
{
nSurfIdxIn = GetFreeSurfaceIndex(pVPPSurfacesIn, nVPPSurfNumIn); // Find free input frame surface
if (MFX_ERR_NOT_FOUND == nSurfIdxIn)
return MFX_ERR_MEMORY_ALLOC;
// Surface locking required when read/write D3D surfaces
sts = mfxAllocator.Lock(mfxAllocator.pthis, pVPPSurfacesIn[nSurfIdxIn]->Data.MemId, &(pVPPSurfacesIn[nSurfIdxIn]->Data));
MSDK_BREAK_ON_ERROR(sts);
sts = LoadRawFrame(pVPPSurfacesIn[nSurfIdxIn], fSource); // Load frame from file into surface
MSDK_BREAK_ON_ERROR(sts);
sts = mfxAllocator.Unlock(mfxAllocator.pthis, pVPPSurfacesIn[nSurfIdxIn]->Data.MemId, &(pVPPSurfacesIn[nSurfIdxIn]->Data));
MSDK_BREAK_ON_ERROR(sts);
nSurfIdxOut = GetFreeSurfaceIndex(pVPPSurfacesOut, nVPPSurfNumOut); // Find free output frame surface
if (MFX_ERR_NOT_FOUND == nSurfIdxOut)
return MFX_ERR_MEMORY_ALLOC;
// Process a frame asychronously (returns immediately)
sts = mfxVPP.RunFrameVPPAsync(pVPPSurfacesIn[nSurfIdxIn], pVPPSurfacesOut[nSurfIdxOut], NULL, &syncp);
if (MFX_ERR_MORE_DATA == sts)
continue;
// MFX_ERR_MORE_SURFACE means output is ready but need more surface (example: Frame Rate Conversion 30->60)
// * Not handled in this example!
MSDK_BREAK_ON_ERROR(sts);
sts = mfxSession.SyncOperation(syncp, 60000); // Synchronize. Wait until frame processing is ready
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
++nFrame;
#ifdef ENABLE_OUTPUT
// Surface locking required when read/write D3D surfaces
sts = mfxAllocator.Lock(mfxAllocator.pthis, pVPPSurfacesOut[nSurfIdxOut]->Data.MemId, &(pVPPSurfacesOut[nSurfIdxOut]->Data));
MSDK_BREAK_ON_ERROR(sts);
sts = WriteRawFrame(pVPPSurfacesOut[nSurfIdxOut], fSink);
MSDK_BREAK_ON_ERROR(sts);
sts = mfxAllocator.Unlock(mfxAllocator.pthis, pVPPSurfacesOut[nSurfIdxOut]->Data.MemId, &(pVPPSurfacesOut[nSurfIdxOut]->Data));
MSDK_BREAK_ON_ERROR(sts);
printf("Frame number: %d\r", nFrame);
#endif
}
// MFX_ERR_MORE_DATA means that the input file has ended, need to go to buffering loop, exit in case of other errors
MSDK_IGNORE_MFX_STS(sts, MFX_ERR_MORE_DATA);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
//
// Stage 2: Retrieve the buffered VPP frames
//
while (MFX_ERR_NONE <= sts)
{
nSurfIdxOut = GetFreeSurfaceIndex(pVPPSurfacesOut, nVPPSurfNumOut); // Find free frame surface
if (MFX_ERR_NOT_FOUND == nSurfIdxOut)
return MFX_ERR_MEMORY_ALLOC;
// Process a frame asychronously (returns immediately)
sts = mfxVPP.RunFrameVPPAsync(NULL, pVPPSurfacesOut[nSurfIdxOut], NULL, &syncp);
MSDK_IGNORE_MFX_STS(sts, MFX_ERR_MORE_SURFACE);
MSDK_BREAK_ON_ERROR(sts);
sts = mfxSession.SyncOperation(syncp, 60000); // Synchronize. Wait until frame processing is ready
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
++nFrame;
#ifdef ENABLE_OUTPUT
// Surface locking required when read/write D3D surfaces
sts = mfxAllocator.Lock(mfxAllocator.pthis, pVPPSurfacesOut[nSurfIdxOut]->Data.MemId, &(pVPPSurfacesOut[nSurfIdxOut]->Data));
MSDK_BREAK_ON_ERROR(sts);
sts = WriteRawFrame(pVPPSurfacesOut[nSurfIdxOut], fSink);
MSDK_BREAK_ON_ERROR(sts);
sts = mfxAllocator.Unlock(mfxAllocator.pthis, pVPPSurfacesOut[nSurfIdxOut]->Data.MemId, &(pVPPSurfacesOut[nSurfIdxOut]->Data));
MSDK_BREAK_ON_ERROR(sts);
printf("Frame number: %d\r", nFrame);
#endif
}
// MFX_ERR_MORE_DATA indicates that there are no more buffered frames, exit in case of other errors
MSDK_IGNORE_MFX_STS(sts, MFX_ERR_MORE_DATA);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
#ifdef ENABLE_BENCHMARK
QueryPerformanceCounter(&tEnd);
double duration = ((double)tEnd.QuadPart - (double)tStart.QuadPart) / freq;
printf("\nExecution time: %3.2fs (%3.2ffps)\n", duration, nFrame/duration);
#endif
// ===================================================================
// Clean up resources
// - It is recommended to close Media SDK components first, before releasing allocated surfaces, since
// some surfaces may still be locked by internal Media SDK resources.
mfxVPP.Close();
//mfxSession closed automatically on destruction
for (int i = 0; i < nVPPSurfNumIn; i++)
delete pVPPSurfacesIn[i];
MSDK_SAFE_DELETE_ARRAY(pVPPSurfacesIn);
for (int i = 0; i < nVPPSurfNumOut; i++)
delete pVPPSurfacesOut[i];
MSDK_SAFE_DELETE_ARRAY(pVPPSurfacesOut);
fclose(fSource);
fclose(fSink);
CleanupHWDevice();
return 0;
}
HRESULT CDXVADecoderH264_DXVA1::DecodeFrame(BYTE* pDataIn, UINT nSize, REFERENCE_TIME rtStart, REFERENCE_TIME rtStop)
{
HRESULT hr = S_FALSE;
int nSurfaceIndex = -1;
int nFramePOC = INT_MIN;
int nOutPOC = INT_MIN;
REFERENCE_TIME rtOutStart = INVALID_TIME;
CH264Nalu Nalu;
CComPtr<IMediaSample> pSampleToDeliver;
CHECK_HR_FALSE (FFH264DecodeFrame(m_pFilter->GetAVCtx(), m_pFilter->GetFrame(), pDataIn, nSize, rtStart,
&nFramePOC, &nOutPOC, &rtOutStart, &m_nNALLength));
// If parsing fail (probably no PPS/SPS), continue anyway it may arrived later (happen on truncated streams)
CHECK_HR_FALSE (FFH264BuildPicParams(m_pFilter->GetAVCtx(), &m_DXVAPicParams, &m_DXVAScalingMatrix, m_IsATIUVD));
TRACE_H264 ("CDXVADecoderH264_DXVA1::DecodeFrame() : nFramePOC = %11d, nOutPOC = %11d[%11d], [%d - %d], rtOutStart = [%20I64d]\n", nFramePOC, nOutPOC, m_nOutPOC, m_DXVAPicParams.field_pic_flag, m_DXVAPicParams.RefPicFlag, rtOutStart);
// Wait I frame after a flush
if (m_bFlushed && !m_DXVAPicParams.IntraPicFlag) {
TRACE_H264 ("CDXVADecoderH264_DXVA1::DecodeFrame() : Flush - wait I frame\n");
return S_FALSE;
}
CHECK_HR_FALSE (GetFreeSurfaceIndex(nSurfaceIndex, &pSampleToDeliver, rtStart, rtStop));
FFH264SetCurrentPicture(nSurfaceIndex, &m_DXVAPicParams, m_pFilter->GetAVCtx());
{
m_DXVAPicParams.StatusReportFeedbackNumber++;
CHECK_HR_FALSE (BeginFrame(nSurfaceIndex, pSampleToDeliver));
// Send picture parameters
CHECK_HR_FALSE (AddExecuteBuffer(DXVA2_PictureParametersBufferType, sizeof(m_DXVAPicParams), &m_DXVAPicParams));
// Add bitstream
CHECK_HR_FALSE (AddExecuteBuffer(DXVA2_BitStreamDateBufferType, nSize, pDataIn));
// Add quantization matrix
CHECK_HR_FALSE (AddExecuteBuffer(DXVA2_InverseQuantizationMatrixBufferType, sizeof(DXVA_Qmatrix_H264), &m_DXVAScalingMatrix));
// Add slice control
CHECK_HR_FALSE (AddExecuteBuffer(DXVA2_SliceControlBufferType, sizeof(DXVA_Slice_H264_Short) * m_nSlices, m_pSliceShort));
// Decode frame
CHECK_HR_FALSE (Execute());
CHECK_HR_FALSE (EndFrame(nSurfaceIndex));
}
bool bAdded = AddToStore(nSurfaceIndex, pSampleToDeliver, m_DXVAPicParams.RefPicFlag, rtStart, rtStop,
m_DXVAPicParams.field_pic_flag, nFramePOC);
FFH264UpdateRefFramesList(&m_DXVAPicParams, m_pFilter->GetAVCtx());
ClearUnusedRefFrames();
if (bAdded) {
hr = DisplayNextFrame();
}
if (nOutPOC != INT_MIN) {
m_nOutPOC = nOutPOC;
m_rtOutStart = rtOutStart;
}
m_bFlushed = false;
return hr;
}
INT CmSurfaceManager::CreateBuffer(UINT size, CM_BUFFER_TYPE type,
CmBuffer_RT * &pSurface1D,
CmOsResource * pCmOsResource, void *&pSysMem)
{
UINT index = m_pCmDevice->ValidSurfaceIndexStart();
pSurface1D = NULL;
if (pCmOsResource) {
if (GetFreeSurfaceIndex(index) != CM_SUCCESS) {
return CM_EXCEED_SURFACE_AMOUNT;
}
} else {
BOOL useNewSurface = TRUE;
if (AllocateSurfaceIndex
(size, 0, 0, CM_SURFACE_FORMAT_UNKNOWN, index,
useNewSurface, pSysMem) != CM_SUCCESS) {
return CM_EXCEED_SURFACE_AMOUNT;
}
if (!useNewSurface) {
CmSurface *pSurface = m_SurfaceArray[index];
if (pSurface
&& (pSurface->Type() ==
CM_ENUM_CLASS_TYPE_CMBUFFER_RT)) {
pSurface1D =
static_cast < CmBuffer_RT * >(pSurface);
} else {
return CM_FAILURE;
}
UpdateBuffer(pSurface1D, size);
UPDATE_PROFILE_FOR_1D_SURFACE(index, size, TRUE);
return CM_SUCCESS;
}
}
if (m_bufferCount >= m_maxBufferCount) {
CM_ASSERT(0);
return CM_EXCEED_SURFACE_AMOUNT;
}
UINT handle = 0;
INT result = AllocateBuffer(size, type, handle, pCmOsResource, pSysMem);
if (result != CM_SUCCESS) {
CM_ASSERT(0);
return result;
}
result =
CmBuffer_RT::Create(index, handle, size, pCmOsResource == NULL,
this, type, pSysMem, pSurface1D);
if (result != CM_SUCCESS) {
FreeBuffer(handle);
CM_ASSERT(0);
return result;
}
m_SurfaceArray[index] = pSurface1D;
UPDATE_PROFILE_FOR_1D_SURFACE(index, size, FALSE);
return CM_SUCCESS;
}
HRESULT CDXVADecoderH264::DecodeFrame(BYTE* pDataIn, UINT nSize, REFERENCE_TIME rtStart, REFERENCE_TIME rtStop)
{
HRESULT hr = S_FALSE;
UINT nSlices = 0;
int nSurfaceIndex = -1;
int nFieldType = -1;
int nSliceType = -1;
int nFramePOC = INT_MIN;
int nOutPOC = INT_MIN;
REFERENCE_TIME rtOutStart = _I64_MIN;
CH264Nalu Nalu;
UINT nNalOffset = 0;
CComPtr<IMediaSample> pSampleToDeliver;
CComQIPtr<IMPCDXVA2Sample> pDXVA2Sample;
int slice_step = 1;
if (FFH264DecodeBuffer(m_pFilter->GetAVCtx(), pDataIn, nSize, &nFramePOC, &nOutPOC, &rtOutStart) == -1) {
return S_FALSE;
}
while (!nSlices && slice_step <= 2) {
Nalu.SetBuffer(pDataIn, nSize, slice_step == 1 ? m_nNALLength : 0);
while (Nalu.ReadNext()) {
switch (Nalu.GetType()) {
case NALU_TYPE_SLICE:
case NALU_TYPE_IDR:
if (m_bUseLongSlice) {
m_pSliceLong[nSlices].BSNALunitDataLocation = nNalOffset;
m_pSliceLong[nSlices].SliceBytesInBuffer = (UINT)Nalu.GetDataLength() + 3; //.GetRoundedDataLength();
m_pSliceLong[nSlices].slice_id = nSlices;
FF264UpdateRefFrameSliceLong(&m_DXVAPicParams, &m_pSliceLong[nSlices], m_pFilter->GetAVCtx());
if (nSlices > 0) {
m_pSliceLong[nSlices - 1].NumMbsForSlice = m_pSliceLong[nSlices].NumMbsForSlice = m_pSliceLong[nSlices].first_mb_in_slice - m_pSliceLong[nSlices - 1].first_mb_in_slice;
}
}
nSlices++;
nNalOffset += (UINT)(Nalu.GetDataLength() + 3);
if (nSlices > MAX_SLICES) {
break;
}
break;
}
}
slice_step++;
}
if (!nSlices) {
return S_FALSE;
}
m_nMaxWaiting = min(max(m_DXVAPicParams.num_ref_frames, 3), 8);
// If parsing fail (probably no PPS/SPS), continue anyway it may arrived later (happen on truncated streams)
if (FAILED(FFH264BuildPicParams(&m_DXVAPicParams, &m_DXVAScalingMatrix, &nFieldType, &nSliceType, m_pFilter->GetAVCtx(), m_pFilter->GetPCIVendor()))) {
return S_FALSE;
}
TRACE_H264("CDXVADecoderH264::DecodeFrame() : nFramePOC = %11d, nOutPOC = %11d[%11d], [%d - %d], rtOutStart = [%20I64d]\n", nFramePOC, nOutPOC, m_nOutPOC, m_DXVAPicParams.field_pic_flag, m_DXVAPicParams.RefPicFlag, rtOutStart);
// Wait I frame after a flush
if (m_bFlushed && !m_DXVAPicParams.IntraPicFlag) {
TRACE_H264("CDXVADecoderH264::DecodeFrame() : Flush - wait I frame\n");
m_nBrokenFramesFlag = 0;
m_nBrokenFramesFlag_POC = 0;
m_nfield_pic_flag = m_DXVAPicParams.field_pic_flag;
m_nRefPicFlag = m_DXVAPicParams.RefPicFlag;
m_nPrevOutPOC = INT_MIN;
return S_FALSE;
}
/* Disabled, because that causes serious problems.
// Some magic code for detecting the incorrect decoding of interlaced frames ...
// TODO : necessary to make it better, and preferably on the side of ffmpeg ...
if (m_nfield_pic_flag && m_nfield_pic_flag == m_DXVAPicParams.field_pic_flag && m_nRefPicFlag == m_DXVAPicParams.RefPicFlag) {
if (m_nPrevOutPOC == m_nOutPOC && m_nOutPOC == INT_MIN) {
m_nBrokenFramesFlag_POC++;
}
m_nBrokenFramesFlag++;
} else {
m_nBrokenFramesFlag = 0;
m_nBrokenFramesFlag_POC = 0;
}
m_nfield_pic_flag = m_DXVAPicParams.field_pic_flag;
m_nRefPicFlag = m_DXVAPicParams.RefPicFlag;
m_nPrevOutPOC = m_nOutPOC;
if (m_nBrokenFramesFlag > 4) {
m_nBrokenFramesFlag = 0;
if (m_nBrokenFramesFlag_POC > 1) {
TRACE_H264("CDXVADecoderH264::DecodeFrame() : Detected broken frames ... flush data\n");
m_nBrokenFramesFlag_POC = 0;
Flush();
return S_FALSE;
}
}
//
*/
CHECK_HR_TRACE(GetFreeSurfaceIndex(nSurfaceIndex, &pSampleToDeliver, rtStart, rtStop));
FFH264SetCurrentPicture(nSurfaceIndex, &m_DXVAPicParams, m_pFilter->GetAVCtx());
CHECK_HR_TRACE(BeginFrame(nSurfaceIndex, pSampleToDeliver));
m_DXVAPicParams.StatusReportFeedbackNumber++;
// Send picture parameters
CHECK_HR_TRACE(AddExecuteBuffer(DXVA2_PictureParametersBufferType, sizeof(m_DXVAPicParams), &m_DXVAPicParams));
CHECK_HR_TRACE(Execute());
// Add bitstream, slice control and quantization matrix
CHECK_HR_TRACE(AddExecuteBuffer(DXVA2_BitStreamDateBufferType, nSize, pDataIn, &nSize));
if (m_bUseLongSlice) {
CHECK_HR_TRACE(AddExecuteBuffer(DXVA2_SliceControlBufferType, sizeof(DXVA_Slice_H264_Long)*nSlices, m_pSliceLong));
} else {
CHECK_HR_TRACE(AddExecuteBuffer(DXVA2_SliceControlBufferType, sizeof(DXVA_Slice_H264_Short)*nSlices, m_pSliceShort));
}
CHECK_HR_TRACE(AddExecuteBuffer(DXVA2_InverseQuantizationMatrixBufferType, sizeof(DXVA_Qmatrix_H264), (void*)&m_DXVAScalingMatrix));
// Decode bitstream
CHECK_HR_TRACE(Execute());
CHECK_HR_TRACE(EndFrame(nSurfaceIndex));
#if defined(_DEBUG) && 0
DisplayStatus();
#endif
bool bAdded = AddToStore(nSurfaceIndex, pSampleToDeliver, m_DXVAPicParams.RefPicFlag, rtStart, rtStop,
m_DXVAPicParams.field_pic_flag, (FF_FIELD_TYPE)nFieldType,
(FF_SLICE_TYPE)nSliceType, nFramePOC);
FFH264UpdateRefFramesList(&m_DXVAPicParams, m_pFilter->GetAVCtx());
ClearUnusedRefFrames();
if (bAdded) {
hr = DisplayNextFrame();
}
if (nOutPOC != INT_MIN) {
m_nOutPOC = nOutPOC;
m_rtOutStart = rtOutStart;
}
m_bFlushed = false;
return hr;
}
int main()
{
mfxStatus sts = MFX_ERR_NONE;
// =====================================================================
// Intel Media SDK decode pipeline setup
// - In this example we are decoding an AVC (H.264) stream
// - For simplistic memory management, system memory surfaces are used to store the decoded frames
// (Note that when using HW acceleration D3D surfaces are prefered, for better performance)
//
// - VPP used to post process (resize) the frame
//
// Open input H.264 elementary stream (ES) file
FILE* fSource;
fopen_s(&fSource, "bbb1920x1080.264", "rb");
MSDK_CHECK_POINTER(fSource, MFX_ERR_NULL_PTR);
// Create output YUV file
FILE* fSink;
fopen_s(&fSink, "dectest_960x540.yuv", "wb");
MSDK_CHECK_POINTER(fSink, MFX_ERR_NULL_PTR);
// Initialize Media SDK session
// - MFX_IMPL_AUTO_ANY selects HW accelaration if available (on any adapter)
// - Version 1.0 is selected for greatest backwards compatibility.
// If more recent API features are needed, change the version accordingly
mfxIMPL impl = MFX_IMPL_AUTO_ANY;
mfxVersion ver = {0, 1};
MFXVideoSession mfxSession;
sts = mfxSession.Init(impl, &ver);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
// Create Media SDK decoder
MFXVideoDECODE mfxDEC(mfxSession);
// Create Media SDK VPP component
MFXVideoVPP mfxVPP(mfxSession);
// Set required video parameters for decode
// - In this example we are decoding an AVC (H.264) stream
// - For simplistic memory management, system memory surfaces are used to store the decoded frames
// (Note that when using HW acceleration D3D surfaces are prefered, for better performance)
mfxVideoParam mfxVideoParams;
memset(&mfxVideoParams, 0, sizeof(mfxVideoParams));
mfxVideoParams.mfx.CodecId = MFX_CODEC_AVC;
mfxVideoParams.IOPattern = MFX_IOPATTERN_OUT_SYSTEM_MEMORY;
// Prepare Media SDK bit stream buffer
// - Arbitrary buffer size for this example
mfxBitstream mfxBS;
memset(&mfxBS, 0, sizeof(mfxBS));
mfxBS.MaxLength = 1024 * 1024;
mfxBS.Data = new mfxU8[mfxBS.MaxLength];
MSDK_CHECK_POINTER(mfxBS.Data, MFX_ERR_MEMORY_ALLOC);
// Read a chunk of data from stream file into bit stream buffer
// - Parse bit stream, searching for header and fill video parameters structure
// - Abort if bit stream header is not found in the first bit stream buffer chunk
sts = ReadBitStreamData(&mfxBS, fSource);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
sts = mfxDEC.DecodeHeader(&mfxBS, &mfxVideoParams);
MSDK_IGNORE_MFX_STS(sts, MFX_WRN_PARTIAL_ACCELERATION);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
// Initialize VPP parameters
// - For simplistic memory management, system memory surfaces are used to store the raw frames
// (Note that when using HW acceleration D3D surfaces are prefered, for better performance)
mfxVideoParam VPPParams;
memset(&VPPParams, 0, sizeof(VPPParams));
// Input data
VPPParams.vpp.In.FourCC = MFX_FOURCC_NV12;
VPPParams.vpp.In.ChromaFormat = MFX_CHROMAFORMAT_YUV420;
VPPParams.vpp.In.CropX = 0;
VPPParams.vpp.In.CropY = 0;
VPPParams.vpp.In.CropW = mfxVideoParams.mfx.FrameInfo.CropW;
VPPParams.vpp.In.CropH = mfxVideoParams.mfx.FrameInfo.CropH;
VPPParams.vpp.In.PicStruct = MFX_PICSTRUCT_PROGRESSIVE;
VPPParams.vpp.In.FrameRateExtN = 30;
VPPParams.vpp.In.FrameRateExtD = 1;
// width must be a multiple of 16
// height must be a multiple of 16 in case of frame picture and a multiple of 32 in case of field picture
VPPParams.vpp.In.Width = MSDK_ALIGN16(VPPParams.vpp.In.CropW);
VPPParams.vpp.In.Height = (MFX_PICSTRUCT_PROGRESSIVE == VPPParams.vpp.In.PicStruct)?
MSDK_ALIGN16(VPPParams.vpp.In.CropH) : MSDK_ALIGN32(VPPParams.vpp.In.CropH);
// Output data
VPPParams.vpp.Out.FourCC = MFX_FOURCC_NV12;
VPPParams.vpp.Out.ChromaFormat = MFX_CHROMAFORMAT_YUV420;
VPPParams.vpp.Out.CropX = 0;
VPPParams.vpp.Out.CropY = 0;
VPPParams.vpp.Out.CropW = VPPParams.vpp.In.CropW/2; // Resize to half size resolution
VPPParams.vpp.Out.CropH = VPPParams.vpp.In.CropH/2;
VPPParams.vpp.Out.PicStruct = MFX_PICSTRUCT_PROGRESSIVE;
VPPParams.vpp.Out.FrameRateExtN = 30;
VPPParams.vpp.Out.FrameRateExtD = 1;
// width must be a multiple of 16
// height must be a multiple of 16 in case of frame picture and a multiple of 32 in case of field picture
VPPParams.vpp.Out.Width = MSDK_ALIGN16(VPPParams.vpp.Out.CropW);
VPPParams.vpp.Out.Height = (MFX_PICSTRUCT_PROGRESSIVE == VPPParams.vpp.Out.PicStruct)?
MSDK_ALIGN16(VPPParams.vpp.Out.CropH) : MSDK_ALIGN32(VPPParams.vpp.Out.CropH);
VPPParams.IOPattern = MFX_IOPATTERN_IN_SYSTEM_MEMORY | MFX_IOPATTERN_OUT_SYSTEM_MEMORY;
// Query number of required surfaces for decoder
mfxFrameAllocRequest DecRequest;
memset(&DecRequest, 0, sizeof(DecRequest));
sts = mfxDEC.QueryIOSurf(&mfxVideoParams, &DecRequest);
MSDK_IGNORE_MFX_STS(sts, MFX_WRN_PARTIAL_ACCELERATION);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
// Query number of required surfaces for VPP
mfxFrameAllocRequest VPPRequest[2];// [0] - in, [1] - out
memset(&VPPRequest, 0, sizeof(mfxFrameAllocRequest)*2);
sts = mfxVPP.QueryIOSurf(&VPPParams, VPPRequest);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
// Determine the required number of surfaces for decoder output (VPP input) and for VPP output
mfxU16 nSurfNumDecVPP = DecRequest.NumFrameSuggested + VPPRequest[0].NumFrameSuggested;
mfxU16 nSurfNumVPPOut = VPPRequest[1].NumFrameSuggested;
// Allocate surfaces for decoder and VPP In
// - Width and height of buffer must be aligned, a multiple of 32
// - Frame surface array keeps pointers all surface planes and general frame info
mfxU16 width = (mfxU16)MSDK_ALIGN32(DecRequest.Info.Width);
mfxU16 height = (mfxU16)MSDK_ALIGN32(DecRequest.Info.Height);
mfxU8 bitsPerPixel = 12; // NV12 format is a 12 bits per pixel format
mfxU32 surfaceSize = width * height * bitsPerPixel / 8;
mfxU8* surfaceBuffers = (mfxU8 *)new mfxU8[surfaceSize * nSurfNumDecVPP];
mfxFrameSurface1** pmfxSurfaces = new mfxFrameSurface1*[nSurfNumDecVPP];
MSDK_CHECK_POINTER(pmfxSurfaces, MFX_ERR_MEMORY_ALLOC);
for (int i = 0; i < nSurfNumDecVPP; i++)
{
pmfxSurfaces[i] = new mfxFrameSurface1;
memset(pmfxSurfaces[i], 0, sizeof(mfxFrameSurface1));
memcpy(&(pmfxSurfaces[i]->Info), &(mfxVideoParams.mfx.FrameInfo), sizeof(mfxFrameInfo));
pmfxSurfaces[i]->Data.Y = &surfaceBuffers[surfaceSize * i];
pmfxSurfaces[i]->Data.U = pmfxSurfaces[i]->Data.Y + width * height;
pmfxSurfaces[i]->Data.V = pmfxSurfaces[i]->Data.U + 1;
pmfxSurfaces[i]->Data.Pitch = width;
}
// Allocate surfaces for VPP Out
// - Width and height of buffer must be aligned, a multiple of 32
// - Frame surface array keeps pointers all surface planes and general frame info
width = (mfxU16)MSDK_ALIGN32(VPPRequest[1].Info.Width);
height = (mfxU16)MSDK_ALIGN32(VPPRequest[1].Info.Height);
bitsPerPixel = 12; // NV12 format is a 12 bits per pixel format
surfaceSize = width * height * bitsPerPixel / 8;
mfxU8* surfaceBuffers2 = (mfxU8 *)new mfxU8[surfaceSize * nSurfNumVPPOut];
mfxFrameSurface1** pmfxSurfaces2 = new mfxFrameSurface1*[nSurfNumVPPOut];
MSDK_CHECK_POINTER(pmfxSurfaces2, MFX_ERR_MEMORY_ALLOC);
for (int i = 0; i < nSurfNumVPPOut; i++)
{
pmfxSurfaces2[i] = new mfxFrameSurface1;
memset(pmfxSurfaces2[i], 0, sizeof(mfxFrameSurface1));
memcpy(&(pmfxSurfaces2[i]->Info), &(VPPParams.vpp.Out), sizeof(mfxFrameInfo));
pmfxSurfaces2[i]->Data.Y = &surfaceBuffers[surfaceSize * i];
pmfxSurfaces2[i]->Data.U = pmfxSurfaces2[i]->Data.Y + width * height;
pmfxSurfaces2[i]->Data.V = pmfxSurfaces2[i]->Data.U + 1;
pmfxSurfaces2[i]->Data.Pitch = width;
}
// Initialize the Media SDK decoder
sts = mfxDEC.Init(&mfxVideoParams);
MSDK_IGNORE_MFX_STS(sts, MFX_WRN_PARTIAL_ACCELERATION);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
// Initialize Media SDK VPP
sts = mfxVPP.Init(&VPPParams);
MSDK_IGNORE_MFX_STS(sts, MFX_WRN_PARTIAL_ACCELERATION);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
// ===============================================================
// Start decoding the frames from the stream
//
#ifdef ENABLE_BENCHMARK
LARGE_INTEGER tStart, tEnd;
QueryPerformanceFrequency(&tStart);
double freq = (double)tStart.QuadPart;
QueryPerformanceCounter(&tStart);
#endif
mfxSyncPoint syncpD;
mfxSyncPoint syncpV;
mfxFrameSurface1* pmfxOutSurface = NULL;
int nIndex = 0;
int nIndex2 = 0;
mfxU32 nFrame = 0;
//
// Stage 1: Main decoding loop
//
while (MFX_ERR_NONE <= sts || MFX_ERR_MORE_DATA == sts || MFX_ERR_MORE_SURFACE == sts)
{
if (MFX_WRN_DEVICE_BUSY == sts)
Sleep(1); // Wait if device is busy, then repeat the same call to DecodeFrameAsync
if (MFX_ERR_MORE_DATA == sts)
{
sts = ReadBitStreamData(&mfxBS, fSource); // Read more data into input bit stream
MSDK_BREAK_ON_ERROR(sts);
}
if (MFX_ERR_MORE_SURFACE == sts || MFX_ERR_NONE == sts)
{
nIndex = GetFreeSurfaceIndex(pmfxSurfaces, nSurfNumDecVPP); // Find free frame surface
if (MFX_ERR_NOT_FOUND == nIndex)
return MFX_ERR_MEMORY_ALLOC;
}
// Decode a frame asychronously (returns immediately)
sts = mfxDEC.DecodeFrameAsync(&mfxBS, pmfxSurfaces[nIndex], &pmfxOutSurface, &syncpD);
// Ignore warnings if output is available,
// if no output and no action required just repeat the DecodeFrameAsync call
if (MFX_ERR_NONE < sts && syncpD)
sts = MFX_ERR_NONE;
if (MFX_ERR_NONE == sts)
{
nIndex2 = GetFreeSurfaceIndex(pmfxSurfaces2, nSurfNumVPPOut); // Find free frame surface
if (MFX_ERR_NOT_FOUND == nIndex)
return MFX_ERR_MEMORY_ALLOC;
for (;;)
{
// Process a frame asychronously (returns immediately)
sts = mfxVPP.RunFrameVPPAsync(pmfxOutSurface, pmfxSurfaces2[nIndex2], NULL, &syncpV);
if (MFX_ERR_NONE < sts && !syncpV) // repeat the call if warning and no output
{
if (MFX_WRN_DEVICE_BUSY == sts)
Sleep(1); // wait if device is busy
}
else if (MFX_ERR_NONE < sts && syncpV)
{
sts = MFX_ERR_NONE; // ignore warnings if output is available
break;
}
else
break; // not a warning
}
// VPP needs more data, let decoder decode another frame as input
if (MFX_ERR_MORE_DATA == sts)
{
continue;
}
else if (MFX_ERR_MORE_SURFACE == sts)
{
// Not relevant for the illustrated workload! Therefore not handled.
// Relevant for cases when VPP produces more frames at output than consumes at input. E.g. framerate conversion 30 fps -> 60 fps
break;
}
else
MSDK_BREAK_ON_ERROR(sts);
}
if (MFX_ERR_NONE == sts)
sts = mfxSession.SyncOperation(syncpV, 60000); // Synchronize. Wait until decoded frame is ready
if (MFX_ERR_NONE == sts)
{
++nFrame;
#ifdef ENABLE_OUTPUT
sts = WriteRawFrame(pmfxSurfaces2[nIndex2], fSink);
MSDK_BREAK_ON_ERROR(sts);
printf("Frame number: %d\r", nFrame);
#endif
}
}
// MFX_ERR_MORE_DATA means that file has ended, need to go to buffering loop, exit in case of other errors
MSDK_IGNORE_MFX_STS(sts, MFX_ERR_MORE_DATA);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
//
// Stage 2: Retrieve the buffered decoded frames
//
while (MFX_ERR_NONE <= sts || MFX_ERR_MORE_SURFACE == sts)
{
if (MFX_WRN_DEVICE_BUSY == sts)
Sleep(1); // Wait if device is busy, then repeat the same call to DecodeFrameAsync
nIndex = GetFreeSurfaceIndex(pmfxSurfaces, nSurfNumDecVPP); // Find free frame surface
if (MFX_ERR_NOT_FOUND == nIndex)
return MFX_ERR_MEMORY_ALLOC;
// Decode a frame asychronously (returns immediately)
sts = mfxDEC.DecodeFrameAsync(NULL, pmfxSurfaces[nIndex], &pmfxOutSurface, &syncpD);
// Ignore warnings if output is available,
// if no output and no action required just repeat the DecodeFrameAsync call
if (MFX_ERR_NONE < sts && syncpD)
sts = MFX_ERR_NONE;
if (MFX_ERR_NONE == sts)
{
nIndex2 = GetFreeSurfaceIndex(pmfxSurfaces2, nSurfNumVPPOut); // Find free frame surface
if (MFX_ERR_NOT_FOUND == nIndex)
return MFX_ERR_MEMORY_ALLOC;
for (;;)
{
// Process a frame asychronously (returns immediately)
sts = mfxVPP.RunFrameVPPAsync(pmfxOutSurface, pmfxSurfaces2[nIndex2], NULL, &syncpV);
if (MFX_ERR_NONE < sts && !syncpV) // repeat the call if warning and no output
{
if (MFX_WRN_DEVICE_BUSY == sts)
Sleep(1); // wait if device is busy
}
else if (MFX_ERR_NONE < sts && syncpV)
{
sts = MFX_ERR_NONE; // ignore warnings if output is available
break;
}
else
break; // not a warning
}
// VPP needs more data, let decoder decode another frame as input
if (MFX_ERR_MORE_DATA == sts)
{
continue;
}
else if (MFX_ERR_MORE_SURFACE == sts)
{
// Not relevant for the illustrated workload! Therefore not handled.
// Relevant for cases when VPP produces more frames at output than consumes at input. E.g. framerate conversion 30 fps -> 60 fps
break;
}
else
MSDK_BREAK_ON_ERROR(sts);
}
if (MFX_ERR_NONE == sts)
sts = mfxSession.SyncOperation(syncpV, 60000); // Synchronize. Waits until decoded frame is ready
if (MFX_ERR_NONE == sts)
{
++nFrame;
#ifdef ENABLE_OUTPUT
sts = WriteRawFrame(pmfxSurfaces2[nIndex2], fSink);
MSDK_BREAK_ON_ERROR(sts);
printf("Frame number: %d\r", nFrame);
#endif
}
}
// MFX_ERR_MORE_DATA means that decoder is done with buffered frames, need to go to VPP buffering loop, exit in case of other errors
MSDK_IGNORE_MFX_STS(sts, MFX_ERR_MORE_DATA);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
//
// Stage 3: Retrieve the buffered VPP frames
//
while (MFX_ERR_NONE <= sts)
{
nIndex2 = GetFreeSurfaceIndex(pmfxSurfaces2, nSurfNumVPPOut); // Find free frame surface
if (MFX_ERR_NOT_FOUND == nIndex2)
return MFX_ERR_MEMORY_ALLOC;
// Process a frame asychronously (returns immediately)
sts = mfxVPP.RunFrameVPPAsync(NULL, pmfxSurfaces2[nIndex2], NULL, &syncpV);
MSDK_IGNORE_MFX_STS(sts, MFX_ERR_MORE_SURFACE);
MSDK_BREAK_ON_ERROR(sts);
sts = mfxSession.SyncOperation(syncpV, 60000); // Synchronize. Wait until frame processing is ready
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
++nFrame;
#ifdef ENABLE_OUTPUT
sts = WriteRawFrame(pmfxSurfaces2[nIndex2], fSink);
MSDK_BREAK_ON_ERROR(sts);
printf("Frame number: %d\r", nFrame);
#endif
}
// MFX_ERR_MORE_DATA indicates that all buffers has been fetched, exit in case of other errors
MSDK_IGNORE_MFX_STS(sts, MFX_ERR_MORE_DATA);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
#ifdef ENABLE_BENCHMARK
QueryPerformanceCounter(&tEnd);
double duration = ((double)tEnd.QuadPart - (double)tStart.QuadPart) / freq;
printf("\nExecution time: %3.2fs (%3.2ffps)\n", duration, nFrame/duration);
#endif
// ===================================================================
// Clean up resources
// - It is recommended to close Media SDK components first, before releasing allocated surfaces, since
// some surfaces may still be locked by internal Media SDK resources.
mfxDEC.Close();
mfxVPP.Close();
// mfxSession closed automatically on destruction
for (int i = 0; i < nSurfNumDecVPP; i++)
delete pmfxSurfaces[i];
for (int i = 0; i < nSurfNumVPPOut; i++)
delete pmfxSurfaces2[i];
MSDK_SAFE_DELETE_ARRAY(pmfxSurfaces);
MSDK_SAFE_DELETE_ARRAY(pmfxSurfaces2);
MSDK_SAFE_DELETE_ARRAY(surfaceBuffers);
MSDK_SAFE_DELETE_ARRAY(surfaceBuffers2);
MSDK_SAFE_DELETE_ARRAY(mfxBS.Data);
fclose(fSource);
fclose(fSink);
return 0;
}
