mfxStatus CQuickSyncDecoder::InitFrameAllocator(mfxVideoParam* pVideoParams, mfxU32 nPitch)
{
MSDK_TRACE("QsDecoder: InitFrameAllocator\n");
// Already initialized
if (m_pFrameSurfaces)
{
return MFX_ERR_NONE;
}
MSDK_CHECK_POINTER(m_pmfxDEC, MFX_ERR_NOT_INITIALIZED);
mfxStatus sts = MFX_ERR_NONE;
// Initialize frame allocator (if needed)
sts = CreateAllocator();
MSDK_CHECK_NOT_EQUAL(sts, MFX_ERR_NONE, sts);
// Find how many surfaces are needed
mfxFrameAllocRequest allocRequest;
MSDK_ZERO_VAR(allocRequest);
sts = m_pmfxDEC->QueryIOSurf(pVideoParams, &allocRequest);
MSDK_IGNORE_MFX_STS(sts, MFX_WRN_PARTIAL_ACCELERATION);
MSDK_IGNORE_MFX_STS(sts, MFX_WRN_INCOMPATIBLE_VIDEO_PARAM);
MSDK_CHECK_RESULT_P_RET(sts, MFX_ERR_NONE);
allocRequest.NumFrameSuggested = (mfxU16)m_nAuxFrameCount + allocRequest.NumFrameSuggested;
allocRequest.NumFrameMin = allocRequest.NumFrameSuggested;
// Decide memory type
allocRequest.Type = MFX_MEMTYPE_EXTERNAL_FRAME | MFX_MEMTYPE_FROM_DECODE;
allocRequest.Type |= (m_bUseD3DAlloc) ?
MFX_MEMTYPE_VIDEO_MEMORY_DECODER_TARGET : MFX_MEMTYPE_SYSTEM_MEMORY;
memcpy(&allocRequest.Info, &pVideoParams->mfx.FrameInfo, sizeof(mfxFrameInfo));
// Allocate frames with H aligned at 32 for both progressive and interlaced content
allocRequest.Info.Height = MSDK_ALIGN32(allocRequest.Info.Height);
allocRequest.Info.Width = (mfxU16)nPitch;
// Perform allocation call. result is saved in m_AllocResponse
sts = m_pFrameAllocator->Alloc(m_pFrameAllocator->pthis, &allocRequest, &m_AllocResponse);
MSDK_CHECK_RESULT_P_RET(sts, MFX_ERR_NONE);
m_nRequiredFramesNum = m_AllocResponse.NumFrameActual;
ASSERT(m_nRequiredFramesNum == allocRequest.NumFrameSuggested);
m_pFrameSurfaces = new mfxFrameSurface1[m_nRequiredFramesNum];
MSDK_CHECK_POINTER(m_pFrameSurfaces, MFX_ERR_MEMORY_ALLOC);
MSDK_ZERO_MEMORY(m_pFrameSurfaces, sizeof(mfxFrameSurface1) * m_nRequiredFramesNum);
// Allocate decoder work & output surfaces
for (mfxU32 i = 0; i < m_nRequiredFramesNum; ++i)
{
// Copy frame info
memcpy(&(m_pFrameSurfaces[i].Info), &pVideoParams->mfx.FrameInfo, sizeof(mfxFrameInfo));
// Save pointer to allocator specific surface object (mid)
m_pFrameSurfaces[i].Data.MemId = m_AllocResponse.mids[i];
m_pFrameSurfaces[i].Data.Pitch = (mfxU16)nPitch;
}
return sts;
}
mfxStatus IntelDecoder::QueryAndAllocRequiredSurfacesForHW()
{
mfxStatus sts = MFX_ERR_NONE;
// Query number of required surfaces for decoder
mfxFrameAllocRequest Request;
memset(&Request, 0, sizeof(Request));
sts = mfxDEC->QueryIOSurf(&mfxVideoParams, &Request);
MSDK_IGNORE_MFX_STS(sts, MFX_WRN_PARTIAL_ACCELERATION);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
numSurfaces = Request.NumFrameSuggested;
Request.Type |= WILL_READ; // This line is only required for Windows DirectX11 to ensure that surfaces can be retrieved by the application
// Allocate surfaces for decoder
//mfxFrameAllocResponse mfxResponse;
sts = pMfxAllocator->Alloc(pMfxAllocator->pthis, &Request, &mfxResponse);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
// Allocate surface headers (mfxFrameSurface1) for decoder
pmfxSurfaces = new mfxFrameSurface1 *[numSurfaces];
MSDK_CHECK_POINTER(pmfxSurfaces, MFX_ERR_MEMORY_ALLOC);
for (int i = 0; i < numSurfaces; i++) {
pmfxSurfaces[i] = new mfxFrameSurface1;
memset(pmfxSurfaces[i], 0, sizeof(mfxFrameSurface1));
memcpy(&(pmfxSurfaces[i]->Info), &(mfxVideoParams.mfx.FrameInfo), sizeof(mfxFrameInfo));
pmfxSurfaces[i]->Data.MemId = mfxResponse.mids[i]; // MID (memory id) represents one video NV12 surface
}
return sts;
}
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;
}
mfxStatus QSV_Encoder_Internal::Drain()
{
mfxStatus sts = MFX_ERR_NONE;
//
// Drain the buffered encoded frames
//
while (MFX_ERR_NONE <= sts)
{
int nTaskIdx = GetFreeTaskIndex(m_pTaskPool, m_nTaskPool);
if (MFX_ERR_NOT_FOUND == nTaskIdx)
{
// No more free tasks, need to sync
sts = m_session.SyncOperation(m_pTaskPool[m_nFirstSyncTask].syncp, 60000);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
m_pTaskPool[m_nFirstSyncTask].syncp = NULL;
m_nFirstSyncTask = (m_nFirstSyncTask + 1) % m_nTaskPool;
}
else
{
for (;;)
{
// Encode a frame asychronously (returns immediately)
sts = m_pmfxENC->EncodeFrameAsync(NULL, NULL, &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
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);
//
// Sync all remaining tasks in task pool
//
while (m_pTaskPool[m_nFirstSyncTask].syncp)
{
sts = m_session.SyncOperation(m_pTaskPool[m_nFirstSyncTask].syncp, 60000);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
m_pTaskPool[m_nFirstSyncTask].syncp = NULL;
m_nFirstSyncTask = (m_nFirstSyncTask + 1) % m_nTaskPool;
}
return sts;
}
mfxStatus QSV_Encoder_Internal::Open(qsv_param_t * pParams)
{
mfxStatus sts = MFX_ERR_NONE;
if (m_bUseD3D11)
// Use D3D11 surface
sts = Initialize(m_impl, m_ver, &m_session, &m_mfxAllocator);
else
// Use system memory
sts = Initialize(m_impl, m_ver, &m_session, NULL);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
m_pmfxENC = new MFXVideoENCODE(m_session);
InitParams(pParams);
sts = m_pmfxENC->Query(&m_mfxEncParams, &m_mfxEncParams);
MSDK_IGNORE_MFX_STS(sts, MFX_WRN_INCOMPATIBLE_VIDEO_PARAM);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
sts = AllocateSurfaces();
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
sts = m_pmfxENC->Init(&m_mfxEncParams);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
sts = GetVideoParam();
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
sts = InitBitstream();
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
return sts;
}
mfxStatus CQuickSyncDecoder::DecodeHeader(mfxBitstream* bs, mfxVideoParam* par)
{
MSDK_CHECK_POINTER(m_pmfxDEC && bs && par, MFX_ERR_NULL_PTR);
mfxStatus sts = m_pmfxDEC->DecodeHeader(bs, par);
// Try again, marking the bitstream as complete
// This workaround should work on all driver versions. But it doesn't work on 15.28 & 15.31
if (MFX_ERR_MORE_DATA == sts)
{
mfxU16 oldFlag = bs->DataFlag;
bs->DataFlag = MFX_BITSTREAM_COMPLETE_FRAME;
sts = m_pmfxDEC->DecodeHeader(bs, par);
bs->DataFlag = oldFlag;
}
// Another workaround for 15.28 and 15.31 drivers
if (MFX_ERR_MORE_DATA == sts && par->mfx.CodecId == MFX_CODEC_AVC)
{
mfxBitstream bs2 = *bs;
bs2.Data = new mfxU8[bs->DataLength + 5];
memcpy(bs2.Data, bs->Data + bs->DataOffset, bs->DataLength - bs->DataOffset);
bs2.MaxLength = bs2.DataLength = bs->DataLength + 5 - bs->DataOffset;
// Write H264 start code + start of splice section
*((unsigned*)(bs2.Data + bs2.DataLength - 5)) = 0x01000000;
bs2.Data[bs2.DataLength -1] = 1; // write SPLICE NALU
sts = m_pmfxDEC->DecodeHeader(&bs2, par);
delete[] bs2.Data; // Cleanup
}
MSDK_IGNORE_MFX_STS(sts, MFX_WRN_PARTIAL_ACCELERATION);
return sts;
}
mfxStatus IntelDecoder::QueryAndAllocRequiredSurfacesForSW()
{
mfxStatus sts = MFX_ERR_NONE;
// 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);
numSurfaces = DecRequest.NumFrameSuggested;
//VPPRequest[0].Type |= WILL_WRITE; // This line is only required for Windows DirectX11 to ensure that surfaces can be written to by the application
//DecRequest.Type |= WILL_READ; // This line is only required for Windows DirectX11 to ensure that surfaces can be retrieved by the application
// Allocate surfaces for decoder
// - 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_ALIGN(DecRequest.Info.Width);
mfxU16 height = (mfxU16)MSDK_ALIGN16(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 * numSurfaces];
// Allocate surface headers (mfxFrameSurface1) for decoder
pmfxSurfaces = new mfxFrameSurface1 *[numSurfaces];
MSDK_CHECK_POINTER(pmfxSurfaces, MFX_ERR_MEMORY_ALLOC);
for (int i = 0; i < numSurfaces; 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;
}
return sts;
}
/* Methods required for integration with Media SDK */
mfxStatus Rotate::PluginInit(mfxCoreInterface *core)
{
MSDK_CHECK_POINTER(core, MFX_ERR_NULL_PTR);
mfxStatus sts = MFX_ERR_NONE;
MSDK_SAFE_DELETE(m_pmfxCore);
m_pmfxCore = new mfxCoreInterface;
MSDK_CHECK_POINTER(m_pmfxCore, MFX_ERR_MEMORY_ALLOC);
*m_pmfxCore = *core;
mfxCoreParam par = {0};
sts = m_pmfxCore->GetCoreParam(m_pmfxCore->pthis, &par);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
m_impl = par.Impl;
mfxHDL hdl = 0;
#if defined(_WIN32) || defined(_WIN64)
if (MFX_IMPL_VIA_MASK(m_impl) == MFX_IMPL_VIA_D3D9) {
sts = m_pmfxCore->GetHandle(m_pmfxCore->pthis, MFX_HANDLE_D3D9_DEVICE_MANAGER, &m_device);
}
else if (MFX_IMPL_VIA_MASK(m_impl) == MFX_IMPL_VIA_D3D11) {
sts = m_pmfxCore->GetHandle(m_pmfxCore->pthis, MFX_HANDLE_D3D11_DEVICE, &m_device);
} else {
hdl = 0;
}
#else
sts = m_pmfxCore->GetHandle(m_pmfxCore->pthis, MFX_HANDLE_VA_DISPLAY, &m_device);
#endif
// SW lib is used if GetHandle return MFX_ERR_NOT_FOUND
MSDK_IGNORE_MFX_STS(sts, MFX_ERR_NOT_FOUND);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
// if external allocator not set use the one from core interface
if (!m_pAlloc && m_pmfxCore->FrameAllocator.pthis)
m_pAlloc = &m_pmfxCore->FrameAllocator;
return MFX_ERR_NONE;
}
mfxStatus QSV_Encoder_Internal::Open(qsv_param_t * pParams)
{
mfxStatus sts = MFX_ERR_NONE;
InitParams(pParams);
sts = m_pmfxENC->Query(&m_mfxEncParams, &m_mfxEncParams);
MSDK_IGNORE_MFX_STS(sts, MFX_WRN_INCOMPATIBLE_VIDEO_PARAM);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
sts = AllocateSurfaces();
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
sts = m_pmfxENC->Init(&m_mfxEncParams);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
sts = GetVideoParam();
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
sts = InitBitstream();
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
return sts;
}
mfxStatus IntelDecoder::InitializeX(HWND hWnd)
{
if (SetDecodeOptions() == MFX_ERR_NULL_PTR)
{
fprintf_s(stdout, "Source file couldn't be found.");
return MFX_ERR_NULL_PTR;
}
// Open input H.264 elementary stream (ES) file
MSDK_FOPEN(fSource, options.SourceName, "rb");
MSDK_CHECK_POINTER(fSource, MFX_ERR_NULL_PTR);
mfxIMPL impl = options.impl;
//Version 1.3 is selected for Video Conference Mode compatibility.
mfxVersion ver = { { 3, 1 } };
pSession = new MFXVideoSession();
pMfxAllocator = (mfxFrameAllocator*)malloc(sizeof(mfxFrameAllocator));
memset(pMfxAllocator, 0, sizeof(mfxFrameAllocator));
mfxStatus sts = Initialize(impl, ver, pSession, pMfxAllocator);
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
sts = pSession->QueryIMPL(&impl_type);
if (impl_type == MFX_IMPL_SOFTWARE)
{
printf("Implementation type is : SOFTWARE\n");
}
else
{
printf("Implementation type is : HARDWARE\n");
}
//impl_type = 2;
// Create Media SDK decoder
mfxDEC = new MFXVideoDECODE(*pSession);
SetDecParameters();
// Prepare Media SDK bit stream buffer
memset(&mfxBS, 0, sizeof(mfxBS));
mfxBS.DataFlag = MFX_BITSTREAM_COMPLETE_FRAME;
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);
mfxVideoParams.AsyncDepth = 1;
outMan.InitD3D(hWnd, mfxVideoParams.mfx.FrameInfo.CropW, mfxVideoParams.mfx.FrameInfo.CropH);
// Query selected implementation and version
if (impl_type == MFX_IMPL_SOFTWARE)
{
sts = QueryAndAllocRequiredSurfacesForSW();
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
}
else
{
sts = QueryAndAllocRequiredSurfacesForHW();
MSDK_CHECK_RESULT(sts, MFX_ERR_NONE, sts);
}
// 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);
}
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;
}
extern "C" __declspec(dllexport) void *openEncoder(int *pErrorCode, int width, int height, int bitRate, int gop)
{
*pErrorCode = 0;
IntelEncoderHandle *pHandle = (IntelEncoderHandle *) malloc(sizeof(IntelEncoderHandle));
mfxStatus sts = MFX_ERR_NONE;
mfxIMPL impl = MFX_IMPL_AUTO_ANY;
mfxVersion ver;
ver.Major = 1;
ver.Minor = 0;
sts = MFXInit(impl, &ver, &pHandle->session);
if (MFX_ERR_NONE != sts) {
// TODO:
*pErrorCode = -1;
}
MFXQueryIMPL(pHandle->session, &impl);
mfxVersion verTemp;
MFXQueryVersion(pHandle->session, &verTemp);
mfxVideoParam mfxEncParams;
memset(&mfxEncParams, 0, sizeof(mfxEncParams));
mfxEncParams.mfx.CodecId = MFX_CODEC_AVC;
// mfxEncParams.mfx.CodecProfile = MFX_PROFILE_AVC_CONSTRAINED_BASELINE;
mfxEncParams.mfx.TargetUsage = MFX_TARGETUSAGE_BALANCED;
if (0 == bitRate) bitRate = 128 * 8;
mfxEncParams.mfx.TargetKbps = bitRate;
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 = width;
mfxEncParams.mfx.FrameInfo.CropH = height;
// 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(width);
mfxEncParams.mfx.FrameInfo.Height =
(MFX_PICSTRUCT_PROGRESSIVE == mfxEncParams.mfx.FrameInfo.PicStruct) ? MSDK_ALIGN16(width) : MSDK_ALIGN32(height);
mfxEncParams.IOPattern = MFX_IOPATTERN_IN_SYSTEM_MEMORY;
sts = MFXVideoENCODE_Query(pHandle->session, &mfxEncParams, &mfxEncParams);
if (MFX_ERR_NONE != sts) {
// TODO:
*pErrorCode = -2;
}
mfxFrameAllocRequest EncRequest;
memset(&EncRequest, 0, sizeof(EncRequest));
sts = MFXVideoENCODE_QueryIOSurf(pHandle->session, &mfxEncParams, &EncRequest);
if (MFX_ERR_NONE != sts) {
// TODO:
*pErrorCode = -3;
}
pHandle->nEncSurfNum = EncRequest.NumFrameSuggested;
mfxU16 w = (mfxU16)MSDK_ALIGN32(EncRequest.Info.Width);
mfxU16 h = (mfxU16)MSDK_ALIGN32(EncRequest.Info.Height);
mfxU8 bitsPerPixel = 12; // NV12 format is a 12 bits per pixel format
mfxU32 surfaceSize = w * h * bitsPerPixel / 8;
pHandle->pSurfaceBuffers = (mfxU8 *)malloc(surfaceSize * pHandle->nEncSurfNum * sizeof(mfxU8));
pHandle->ppEncSurfaces = (mfxFrameSurface1 **)malloc(sizeof(mfxFrameSurface1*)* pHandle->nEncSurfNum);
for (int i = 0; i < pHandle->nEncSurfNum; i++)
{
pHandle->ppEncSurfaces[i] = (mfxFrameSurface1 *)malloc(sizeof(mfxFrameSurface1));
memset(pHandle->ppEncSurfaces[i], 0, sizeof(mfxFrameSurface1));
memcpy(&(pHandle->ppEncSurfaces[i]->Info), &(mfxEncParams.mfx.FrameInfo), sizeof(mfxFrameInfo));
pHandle->ppEncSurfaces[i]->Data.Y = &pHandle->pSurfaceBuffers[surfaceSize * i];
pHandle->ppEncSurfaces[i]->Data.U = pHandle->ppEncSurfaces[i]->Data.Y + w * h;
pHandle->ppEncSurfaces[i]->Data.V = pHandle->ppEncSurfaces[i]->Data.U + 1;
pHandle->ppEncSurfaces[i]->Data.Pitch = w;
// 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
memset(pHandle->ppEncSurfaces[i]->Data.Y, 100, w * h); // Y plane
memset(pHandle->ppEncSurfaces[i]->Data.U, 50, (w * h) / 2); // UV plane
}
sts = MFXVideoENCODE_Init(pHandle->session, &mfxEncParams);
MSDK_IGNORE_MFX_STS(sts, MFX_WRN_PARTIAL_ACCELERATION);
if (MFX_ERR_NONE != sts) {
// TODO:
*pErrorCode = -4;
}
mfxVideoParam par;
memset(&par, 0, sizeof(par));
sts = MFXVideoENCODE_GetVideoParam(pHandle->session, &par);
if (MFX_ERR_NONE != sts) {
// TODO:
*pErrorCode = -5;
}
memset(&pHandle->mfxBS, 0, sizeof(pHandle->mfxBS));
pHandle->mfxBS.MaxLength = par.mfx.BufferSizeInKB * 1024;
pHandle->mfxBS.Data = (mfxU8 *)malloc(sizeof(mfxU8)* pHandle->mfxBS.MaxLength);
return pHandle;
}
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;
}
mfxStatus CQuickSyncDecoder::InternalReset(mfxVideoParam* pVideoParams, mfxU32 nPitch, bool bInited)
{
MSDK_CHECK_POINTER(pVideoParams, MFX_ERR_NULL_PTR);
MSDK_CHECK_POINTER(m_pmfxDEC, MFX_ERR_NOT_INITIALIZED);
mfxStatus sts = MFX_ERR_NONE;
m_pVideoParams = pVideoParams;
if (NULL == m_pFrameAllocator)
{
bInited = false;
}
// Reset decoder
if (bInited)
{
sts = m_pmfxDEC->Reset(pVideoParams);
// Need to reset the frame allocator
if (MSDK_FAILED(sts))
{
m_pmfxDEC->Close();
FreeFrameAllocator();
bInited = false;
}
if (m_pFrameSurfaces != NULL)
{
// Another VC1 decoder + VPP bug workaround
for (int i = 0; i < m_nRequiredFramesNum; ++i)
{
m_pFrameSurfaces[i].Data.Locked = 0;
m_LockedSurfaces[i] = 0;
}
}
}
// Full init
if (!bInited)
{
// Setup allocator - will initialize D3D if needed
sts = InitFrameAllocator(pVideoParams, nPitch);
MSDK_CHECK_RESULT_P_RET(sts, MFX_ERR_NONE);
// Init MSDK decoder
sts = m_pmfxDEC->Init(pVideoParams);
switch (sts)
{
case MFX_ERR_NONE:
MSDK_TRACE("QsDecoder: decoder Init is successful\n");
break;
case MFX_WRN_PARTIAL_ACCELERATION:
MSDK_TRACE("QsDecoder: decoder Init is successful w/o HW acceleration\n");
m_bHwAcceleration = false;
break;
case MFX_WRN_INCOMPATIBLE_VIDEO_PARAM:
MSDK_TRACE("QsDecoder: decoder Init is successful - wrong video parameters\n");
break;
default:
MSDK_TRACE("QsDecoder: decoder Init has failed!\n");
break;
}
}
MSDK_IGNORE_MFX_STS(sts, MFX_WRN_INCOMPATIBLE_VIDEO_PARAM);
MSDK_IGNORE_MFX_STS(sts, MFX_WRN_PARTIAL_ACCELERATION);
return sts;
}
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;
}
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;
}
