/*!
 *************************************************************************************
 * \file mv-search.c
 *
 * \brief
 *    Motion Vector Search, unified for B and P Pictures
 *
 * \author
 *    Main contributors (see contributors.h for copyright, address and affiliation details)
 *      - Stephan Wenger                  <stewe@cs.tu-berlin.de>
 *      - Inge Lille-Langoy               <inge.lille-langoy@telenor.com>
 *      - Rickard Sjoberg                 <rickard.sjoberg@era.ericsson.se>
 *      - Stephan Wenger                  <stewe@cs.tu-berlin.de>
 *      - Jani Lainema                    <jani.lainema@nokia.com>
 *      - Detlev Marpe                    <marpe@hhi.de>
 *      - Thomas Wedi                     <wedi@tnt.uni-hannover.de>
 *      - Heiko Schwarz                   <hschwarz@hhi.de>
 *      - Alexis Michael Tourapis         <alexismt@ieee.org>
 *
 *************************************************************************************
*/

#include "contributors.h"

#include <stdlib.h>
#include <math.h>
#include <limits.h>
#include <string.h>

#include "global.h"

#include "image.h"
#include "mv-search.h"
#include "refbuf.h"
#include "memalloc.h"
#include "mb_access.h"
#include "fast_me.h"

#include <time.h>
#include <sys/timeb.h>

// These procedure pointers are used by motion_search() and one_eigthpel()
static pel_t *(*PelYline_11) (pel_t *, int, int, int, int);
static pel_t *(*get_line) (pel_t**, int, int, int, int);
static pel_t *(*get_line_p1) (pel_t**, int, int, int, int);
static pel_t *(*get_line_p2) (pel_t**, int, int, int, int);

// Statistics, temporary
int     max_mvd;
short*  spiral_search_x;
short*  spiral_search_y;
short*  spiral_hpel_search_x;
short*  spiral_hpel_search_y;

int*    mvbits;
int*    refbits;
unsigned int*  byte_abs;
int**** motion_cost;


void SetMotionVectorPredictor (short  pmv[2],
                               char   **refPic,
                               short  ***tmp_mv,
                               short  ref_frame,
                               int    list,
                               int    block_x,
                               int    block_y,
                               int    blockshape_x,
                               int    blockshape_y);

extern ColocatedParams *Co_located;

#ifdef _FAST_FULL_ME_

/*****
 *****  static variables for fast integer motion estimation
 *****
 */
static int  **search_setup_done;  //!< flag if all block SAD's have been calculated yet
static int  **search_center_x;    //!< absolute search center for fast full motion search
static int  **search_center_y;    //!< absolute search center for fast full motion search
static int  **pos_00;             //!< position of (0,0) vector
static distpel  *****BlockSAD;        //!< SAD for all blocksize, ref. frames and motion vectors
static int  **max_search_range;


/*!
 ***********************************************************************
 * \brief
 *    function creating arrays for fast integer motion estimation
 ***********************************************************************
 */
void
InitializeFastFullIntegerSearch ()
{
  int  i, j, k, list;
  int  search_range = input->search_range;
  int  max_pos      = (2*search_range+1) * (2*search_range+1);
  
  if ((BlockSAD = (distpel*****)malloc (2 * sizeof(distpel****))) == NULL)
    no_mem_exit ("InitializeFastFullIntegerSearch: BlockSAD");
  
  for (list=0; list<2;list++)
  {
    if ((BlockSAD[list] = (distpel****)malloc ((img->max_num_references) * sizeof(distpel***))) == NULL)
      no_mem_exit ("InitializeFastFullIntegerSearch: BlockSAD");
    for (i = 0; i < img->max_num_references; i++)
    {
      if ((BlockSAD[list][i] = (distpel***)malloc (8 * sizeof(distpel**))) == NULL)
        no_mem_exit ("InitializeFastFullIntegerSearch: BlockSAD");
      for (j = 1; j < 8; j++)
      {
        if ((BlockSAD[list][i][j] = (distpel**)malloc (16 * sizeof(distpel*))) == NULL)
          no_mem_exit ("InitializeFastFullIntegerSearch: BlockSAD");
        for (k = 0; k < 16; k++)
        {
          if ((BlockSAD[list][i][j][k] = (distpel*)malloc (max_pos * sizeof(distpel))) == NULL)
            no_mem_exit ("InitializeFastFullIntegerSearch: BlockSAD");
        }        
      }
    }
  }
  
  if ((search_setup_done = (int**)malloc (2*sizeof(int)))==NULL)
    no_mem_exit ("InitializeFastFullIntegerSearch: search_setup_done");
  if ((search_center_x = (int**)malloc (2*sizeof(int)))==NULL)
    no_mem_exit ("InitializeFastFullIntegerSearch: search_center_x");
  if ((search_center_y = (int**)malloc (2*sizeof(int)))==NULL)
    no_mem_exit ("InitializeFastFullIntegerSearch: search_center_y");
  if ((pos_00 = (int**)malloc (2*sizeof(int)))==NULL)
    no_mem_exit ("InitializeFastFullIntegerSearch: pos_00");
  if ((max_search_range = (int**)malloc (2*sizeof(int)))==NULL)
    no_mem_exit ("InitializeFastFullIntegerSearch: max_search_range");
  
  for (list=0; list<2; list++)
  {
    if ((search_setup_done[list] = (int*)malloc ((img->max_num_references)*sizeof(int)))==NULL)
      no_mem_exit ("InitializeFastFullIntegerSearch: search_setup_done");
    if ((search_center_x[list] = (int*)malloc ((img->max_num_references)*sizeof(int)))==NULL)
      no_mem_exit ("InitializeFastFullIntegerSearch: search_center_x");
    if ((search_center_y[list] = (int*)malloc ((img->max_num_references)*sizeof(int)))==NULL)
      no_mem_exit ("InitializeFastFullIntegerSearch: search_center_y");
    if ((pos_00[list] = (int*)malloc ((img->max_num_references)*sizeof(int)))==NULL)
      no_mem_exit ("InitializeFastFullIntegerSearch: pos_00");
    if ((max_search_range[list] = (int*)malloc ((img->max_num_references)*sizeof(int)))==NULL)
      no_mem_exit ("InitializeFastFullIntegerSearch: max_search_range");
  }
  
  // assign max search ranges for reference frames
  if (input->full_search == 2)
  {
    for (list=0;list<2;list++)
      for (i=0; i<img->max_num_references; i++)  
        max_search_range[list][i] = search_range;
  }
  else
  {
    for (list=0;list<2;list++)
    {
      max_search_range[list][0] = search_range;
      for (i=1; i< img->max_num_references; i++)  max_search_range[list][i] = search_range / 2;
    }
  }
  
}



/*!
 ***********************************************************************
 * \brief
 *    function for deleting the arrays for fast integer motion estimation
 ***********************************************************************
 */
void
ClearFastFullIntegerSearch ()
{
  int  i, j, k, list;
  
  for (list=0; list<2; list++)
  {
    for (i = 0; i < img->max_num_references; i++)
    {
      for (j = 1; j < 8; j++)
      {
        for (k = 0; k < 16; k++)
        {
          free (BlockSAD[list][i][j][k]);
        }
        free (BlockSAD[list][i][j]);
      }
      free (BlockSAD[list][i]);
    }
    free (BlockSAD[list]);
  }
  free (BlockSAD);
  
  for (list=0; list<2; list++)
  {
    free (search_setup_done[list]);
    free (search_center_x[list]);
    free (search_center_y[list]);
    free (pos_00[list]);
    free (max_search_range[list]);
  }
  free (search_setup_done);
  free (search_center_x);
  free (search_center_y);
  free (pos_00);
  free (max_search_range);
  
}


/*!
 ***********************************************************************
 * \brief
 *    function resetting flags for fast integer motion estimation
 *    (have to be called in start_macroblock())
 ***********************************************************************
 */
void
ResetFastFullIntegerSearch ()
{
  int i,list;
  
  for (list=0; list<2; list++)
    for (i = 0; i < img->max_num_references; i++)
      search_setup_done [list][i] = 0;
}

/*!
 ***********************************************************************
 * \brief
 *    calculation of SAD for larger blocks on the basis of 4x4 blocks
 ***********************************************************************
 */
void
SetupLargerBlocks (int list, int refindex, int max_pos)
{
#define ADD_UP_BLOCKS()   _o=*_bo; _i=*_bi; _j=*_bj; for(pos=0;pos<max_pos;pos++) _o[pos] = _i[pos] + _j[pos];
#define INCREMENT(inc)    _bo+=inc; _bi+=inc; _bj+=inc;
  
  distpel   pos, **_bo, **_bi, **_bj;
  register distpel *_o,   *_i,   *_j;
  
  //--- blocktype 6 ---
  _bo = BlockSAD[list][refindex][6];
  _bi = BlockSAD[list][refindex][7];
  _bj = _bi + 4;
  ADD_UP_BLOCKS(); INCREMENT(1);
  ADD_UP_BLOCKS(); INCREMENT(1);
  ADD_UP_BLOCKS(); INCREMENT(1);
  ADD_UP_BLOCKS(); INCREMENT(5);
  ADD_UP_BLOCKS(); INCREMENT(1);
  ADD_UP_BLOCKS(); INCREMENT(1);
  ADD_UP_BLOCKS(); INCREMENT(1);
  ADD_UP_BLOCKS();
  
  //--- blocktype 5 ---
  _bo = BlockSAD[list][refindex][5];
  _bi = BlockSAD[list][refindex][7];
  _bj = _bi + 1;
  ADD_UP_BLOCKS(); INCREMENT(2);
  ADD_UP_BLOCKS(); INCREMENT(2);
  ADD_UP_BLOCKS(); INCREMENT(2);
  ADD_UP_BLOCKS(); INCREMENT(2);
  ADD_UP_BLOCKS(); INCREMENT(2);
  ADD_UP_BLOCKS(); INCREMENT(2);
  ADD_UP_BLOCKS(); INCREMENT(2);
  ADD_UP_BLOCKS();
  
  //--- blocktype 4 ---
  _bo = BlockSAD[list][refindex][4];
  _bi = BlockSAD[list][refindex][6];
  _bj = _bi + 1;
  ADD_UP_BLOCKS(); INCREMENT(2);
  ADD_UP_BLOCKS(); INCREMENT(6);
  ADD_UP_BLOCKS(); INCREMENT(2);
  ADD_UP_BLOCKS();
  
  //--- blocktype 3 ---
  _bo = BlockSAD[list][refindex][3];
  _bi = BlockSAD[list][refindex][4];
  _bj = _bi + 8;
  ADD_UP_BLOCKS(); INCREMENT(2);
  ADD_UP_BLOCKS();
  
  //--- blocktype 2 ---
  _bo = BlockSAD[list][refindex][2];
  _bi = BlockSAD[list][refindex][4];
  _bj = _bi + 2;
  ADD_UP_BLOCKS(); INCREMENT(8);
  ADD_UP_BLOCKS();
  
  //--- blocktype 1 ---
  _bo = BlockSAD[list][refindex][1];
  _bi = BlockSAD[list][refindex][3];
  _bj = _bi + 2;
  ADD_UP_BLOCKS();
}


/*!
 ***********************************************************************
 * \brief
 *    Setup the fast search for an macroblock
 ***********************************************************************
 */
void SetupFastFullPelSearch (short ref, int list)  // <--  reference frame parameter, list0 or 1
{
  short   pmv[2];
  pel_t   orig_blocks[256], *orgptr=orig_blocks, *refptr;
  int     offset_x, offset_y, x, y, range_partly_outside, ref_x, ref_y, pos, abs_x, abs_y, bindex, blky;
  int     LineSadBlk0, LineSadBlk1, LineSadBlk2, LineSadBlk3;
  int     max_width, max_height;
  int     img_width, img_height;
  
  StorablePicture *ref_picture;
  pel_t   *ref_pic;
  
  distpel**   block_sad = BlockSAD[list][ref][7];
  int     search_range  = max_search_range[list][ref];
  int     max_pos       = (2*search_range+1) * (2*search_range+1);
  
  int     list_offset   = img->mb_data[img->current_mb_nr].list_offset; 
  
  int     apply_weights = ( (active_pps->weighted_pred_flag && (img->type == P_SLICE || img->type == SP_SLICE)) ||
    (active_pps->weighted_bipred_idc && (img->type == B_SLICE)));
  
  
  ref_picture     = listX[list+list_offset][ref];
  
  //===== Use weighted Reference for ME ====
  if (apply_weights && input->UseWeightedReferenceME)
    ref_pic       = ref_picture->imgY_11_w;
  else
    ref_pic       = ref_picture->imgY_11;
  
  max_width     = ref_picture->size_x - 17;
  max_height    = ref_picture->size_y - 17;
  
  img_width     = ref_picture->size_x;
  img_height    = ref_picture->size_y;
  
  //===== get search center: predictor of 16x16 block =====
  SetMotionVectorPredictor (pmv, enc_picture->ref_idx[list], enc_picture->mv[list], ref, list, 0, 0, 16, 16);
  
  search_center_x[list][ref] = pmv[0] / 4;