[x265] [PATCH 2 of 2] pass TLD into class FrameFilter

Deepthi Nandakumar deepthi at multicorewareinc.com
Thu Jun 26 07:47:31 CEST 2014


On Tue, Jun 24, 2014 at 5:36 AM, Min Chen <chenm003 at 163.com> wrote:

> # HG changeset patch
> # User Min Chen <chenm003 at 163.com>
> # Date 1403568362 25200
> # Node ID efa48bc0245bded1418db3c42b042acb9969146c
> # Parent  12c1d8aaa8f56a8f2de74c8ff1451d99d04c817d
> pass TLD into class FrameFilter
>
> diff -r 12c1d8aaa8f5 -r efa48bc0245b source/encoder/cturow.h
> --- a/source/encoder/cturow.h   Mon Jun 23 17:03:49 2014 -0700
> +++ b/source/encoder/cturow.h   Mon Jun 23 17:06:02 2014 -0700
> @@ -47,6 +47,10 @@
>      RDCost      m_rdCost;
>      TComTrQuant m_trQuant;
>
> +    // NOTE: the maximum LCU 64x64 have 256 partitions
> +    bool        m_edgeFilter[256];
> +    uint8_t     m_blockingStrength[256];
> +
>      void init(Encoder&);
>      ~ThreadLocalData();
>  };
> diff -r 12c1d8aaa8f5 -r efa48bc0245b source/encoder/encoder.cpp
> --- a/source/encoder/encoder.cpp        Mon Jun 23 17:03:49 2014 -0700
> +++ b/source/encoder/encoder.cpp        Mon Jun 23 17:06:02 2014 -0700
> @@ -42,6 +42,7 @@
>  #include "x265.h"
>
>  using namespace x265;
> +ThreadLocalData* Encoder::m_threadLocalData;
>
>  Encoder::Encoder()
>  {
> @@ -194,9 +195,10 @@
>      if (m_frameEncoder)
>      {
>          int numRows = (m_param->sourceHeight + g_maxCUSize - 1) /
> g_maxCUSize;
> +        int numCols = (m_param->sourceWidth  + g_maxCUSize - 1) /
> g_maxCUSize;
>          for (int i = 0; i < m_param->frameNumThreads; i++)
>          {
> -            if (!m_frameEncoder[i].init(this, numRows))
> +            if (!m_frameEncoder[i].init(this, numRows, numCols))
>              {
>                  x265_log(m_param, X265_LOG_ERROR, "Unable to initialize
> frame encoder, aborting\n");
>                  m_aborted = true;
> diff -r 12c1d8aaa8f5 -r efa48bc0245b source/encoder/encoder.h
> --- a/source/encoder/encoder.h  Mon Jun 23 17:03:49 2014 -0700
> +++ b/source/encoder/encoder.h  Mon Jun 23 17:06:02 2014 -0700
> @@ -175,7 +175,7 @@
>
>      x265_param*        m_param;
>      RateControl*       m_rateControl;
> -    ThreadLocalData*   m_threadLocalData;
> +    static ThreadLocalData*   m_threadLocalData;
>
>      bool               m_bEnableRDOQ;
>
> diff -r 12c1d8aaa8f5 -r efa48bc0245b source/encoder/frameencoder.cpp
> --- a/source/encoder/frameencoder.cpp   Mon Jun 23 17:03:49 2014 -0700
> +++ b/source/encoder/frameencoder.cpp   Mon Jun 23 17:06:02 2014 -0700
> @@ -80,15 +80,17 @@
>      stop();
>  }
>
> -bool FrameEncoder::init(Encoder *top, int numRows)
> +bool FrameEncoder::init(Encoder *top, int numRows, int numCols)
>  {
>      bool ok = true;
>
>      m_top = top;
>      m_param = top->m_param;
>      m_numRows = numRows;
> +    m_numCols = numCols;
>      m_filterRowDelay = (m_param->saoLcuBasedOptimization &&
> m_param->saoLcuBoundary) ?
>          2 : (m_param->bEnableSAO || m_param->bEnableLoopFilter ? 1 : 0);
> +    m_filterRowDelayCus = m_filterRowDelay * numCols;
>
>      m_rows = new CTURow[m_numRows];
>      for (int i = 0; i < m_numRows; ++i)
> @@ -505,7 +507,7 @@
>              // Extend border after whole-frame SAO is finished
>              for (int row = 0; row < m_numRows; row++)
>              {
> -                m_frameFilter.processRowPost(row);
> +                m_frameFilter.processRowPost(row, 0);
>              }
>          }
>
> @@ -845,7 +847,7 @@
>      }
>
>      // setup thread-local data
> -    ThreadLocalData& tld = threadId >= 0 ?
> m_top->m_threadLocalData[threadId] : m_tld;
> +    ThreadLocalData& tld = threadId >= 0 ?
> Encoder::m_threadLocalData[threadId] : m_tld;
>      tld.m_trQuant.m_nr = &m_nr;
>      tld.m_search.m_mref = m_mref;
>      codeRow.setThreadLocalData(tld);
> @@ -856,7 +858,8 @@
>      tld.m_cuCoder.m_log =
> &tld.m_cuCoder.m_sliceTypeLog[m_frame->getSlice()->getSliceType()];
>
>      int64_t startTime = x265_mdate();
> -    const uint32_t numCols = m_frame->getPicSym()->getFrameWidthInCU();
> +    assert(m_frame->getPicSym()->getFrameWidthInCU() == m_numCols);
> +    const uint32_t numCols = m_numCols;
>      const uint32_t lineStartCUAddr = row * numCols;
>      bool bIsVbv = m_param->rc.vbvBufferSize > 0 &&
> m_param->rc.vbvMaxBitrate > 0;
>
> diff -r 12c1d8aaa8f5 -r efa48bc0245b source/encoder/frameencoder.h
> --- a/source/encoder/frameencoder.h     Mon Jun 23 17:03:49 2014 -0700
> +++ b/source/encoder/frameencoder.h     Mon Jun 23 17:06:02 2014 -0700
> @@ -65,15 +65,15 @@
>
>      void setThreadPool(ThreadPool *p);
>
> -    bool init(Encoder *top, int numRows);
> +    bool init(Encoder *top, int numRows, int numCols);
>
>      void destroy();
>
>      void processRowEncoder(int row, const int threadId);
>
> -    void processRowFilter(int row)
> +    void processRowFilter(int row, const int threadId)
>      {
> -        m_frameFilter.processRow(row);
> +        m_frameFilter.processRow(row, threadId);
>      }
>
>      void enqueueRowEncoder(int row)
> @@ -108,7 +108,7 @@
>          }
>          else
>          {
> -            processRowFilter(realRow);
> +            processRowFilter(realRow, threadId);
>
>              // NOTE: Active next row
>              if (realRow != m_numRows - 1)
> @@ -154,6 +154,7 @@
>      bool                     m_threadActive;
>
>      int                      m_numRows;
> +    uint32_t                 m_numCols;
>      CTURow*                  m_rows;
>      TComSPS                  m_sps;
>      TComPPS                  m_pps;
> @@ -195,6 +196,7 @@
>      Frame*                   m_frame;
>
>      int                      m_filterRowDelay;
> +    int                      m_filterRowDelayCus;
>      Event                    m_completionEvent;
>      int64_t                  m_totalTime;
>      bool                     m_isReferenced;
> diff -r 12c1d8aaa8f5 -r efa48bc0245b source/encoder/framefilter.cpp
> --- a/source/encoder/framefilter.cpp    Mon Jun 23 17:03:49 2014 -0700
> +++ b/source/encoder/framefilter.cpp    Mon Jun 23 17:06:02 2014 -0700
> @@ -121,13 +121,15 @@
>      }
>  }
>
> -void FrameFilter::processRow(int row)
> +void FrameFilter::processRow(int row, const int threadId)
>
 {
>      PPAScopeEvent(Thread_filterCU);
> +    assert(threadId >= 0);
> +    ThreadLocalData& tld = Encoder::m_threadLocalData[threadId];
>
>      if (!m_param->bEnableLoopFilter && !m_param->bEnableSAO)
>      {
> -        processRowPost(row);
> +        processRowPost(row, threadId);
>          return;
>      }
>
> @@ -146,26 +148,23 @@
>
>      if (m_param->bEnableLoopFilter)
>      {
> -        bool edgeFilter[256];    // NOTE: the maximum LCU 64x64 have 256
> partitions
> -        uint8_t blockingStrength[256];
> -
>          for (uint32_t col = 0; col < numCols; col++)
>          {
>              const uint32_t cuAddr = lineStartCUAddr + col;
>              TComDataCU* cu = m_pic->getCU(cuAddr);
>
> -            m_loopFilter.loopFilterCU(cu, EDGE_VER, edgeFilter,
> blockingStrength);
> +            m_loopFilter.loopFilterCU(cu, EDGE_VER, tld.m_edgeFilter,
> tld.m_blockingStrength);
>
>              if (col > 0)
>              {
>                  TComDataCU* cu_prev = m_pic->getCU(cuAddr - 1);
> -                m_loopFilter.loopFilterCU(cu_prev, EDGE_HOR, edgeFilter,
> blockingStrength);
> +                m_loopFilter.loopFilterCU(cu_prev, EDGE_HOR,
> tld.m_edgeFilter, tld.m_blockingStrength);
>              }
>          }
>
>          {
>              TComDataCU* cu_prev = m_pic->getCU(lineStartCUAddr + numCols
> - 1);
> -            m_loopFilter.loopFilterCU(cu_prev, EDGE_HOR, edgeFilter,
> blockingStrength);
> +            m_loopFilter.loopFilterCU(cu_prev, EDGE_HOR,
> tld.m_edgeFilter, tld.m_blockingStrength);
>          }
>      }
>
> @@ -178,7 +177,7 @@
>          // NOTE: Delay a row because SAO decide need top row pixels at
> next row, is it HM's bug?
>          if (row >= m_saoRowDelay)
>          {
> -            processSao(row - m_saoRowDelay);
> +            processSao(row - m_saoRowDelay, threadId);
>          }
>      }
>
> @@ -190,7 +189,7 @@
>
>      if (row > 0)
>      {
> -        processRowPost(row - 1);
> +        processRowPost(row - 1, threadId);
>      }
>
>      if (row == m_numRows - 1)
> @@ -201,15 +200,15 @@
>
>              for (int i = m_numRows - m_saoRowDelay; i < m_numRows; i++)
>              {
> -                processSao(i);
> +                processSao(i, threadId);
>              }
>          }
>
> -        processRowPost(row);
> +        processRowPost(row, threadId);
>      }
>  }
>
> -void FrameFilter::processRowPost(int row)
> +void FrameFilter::processRowPost(int row, const int /*threadId*/)
>  {
>      const uint32_t numCols = m_pic->getPicSym()->getFrameWidthInCU();
>      const uint32_t lineStartCUAddr = row * numCols;
> @@ -502,7 +501,7 @@
>      return ssim;
>  }
>
> -void FrameFilter::processSao(int row)
> +void FrameFilter::processSao(int row, const int threadId)
>

Internal tests throw warning here:
D:\repo-staging\source\encoder\framefilter.cpp(504): warning C4100:
'threadId' : unreferenced formal parameter
Can be eliminated?


>  {
>      const uint32_t numCols = m_pic->getPicSym()->getFrameWidthInCU();
>      const uint32_t lineStartCUAddr = row * numCols;
> diff -r 12c1d8aaa8f5 -r efa48bc0245b source/encoder/framefilter.h
> --- a/source/encoder/framefilter.h      Mon Jun 23 17:03:49 2014 -0700
> +++ b/source/encoder/framefilter.h      Mon Jun 23 17:06:02 2014 -0700
> @@ -50,9 +50,9 @@
>
>      void start(Frame *pic);
>
> -    void processRow(int row);
> -    void processRowPost(int row);
> -    void processSao(int row);
> +    void processRow(int row, const int threadId);
> +    void processRowPost(int row, const int threadId);
> +    void processSao(int row, const int threadId);
>
>  protected:
>
>
> _______________________________________________
> x265-devel mailing list
> x265-devel at videolan.org
> https://mailman.videolan.org/listinfo/x265-devel
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.videolan.org/pipermail/x265-devel/attachments/20140626/8de69807/attachment-0001.html>


More information about the x265-devel mailing list