#include "../mednafen.h"
#include "../general.h"
#include "../state.h"

#include "surface.h"

extern "C" uint8_t psx_gpu_upscale_shift;

#include "Deinterlacer.h"
Deinterlacer::Deinterlacer() : FieldBuffer(NULL), StateValid(false), DeintType(DEINT_WEAVE)
{
 PrevDRect.x = 0;
 PrevDRect.y = 0;

 PrevDRect.w = 0;
 PrevDRect.h = 0;
}

Deinterlacer::~Deinterlacer()
{
 if(FieldBuffer)
 {
  delete FieldBuffer;
  FieldBuffer = NULL;
 }
}

void Deinterlacer::SetType(unsigned dt)
{
 if(DeintType != dt)
 {
  DeintType = dt;

  LWBuffer.resize(0);
  if(FieldBuffer)
  {
   delete FieldBuffer;
   FieldBuffer = NULL;
  }
  StateValid = false;
 }
}

template<typename T>
void Deinterlacer::InternalProcess(MDFN_Surface *surface, MDFN_Rect &DisplayRect, int32 *LineWidths, const bool field)
{
 //
 // We need to output with LineWidths as always being valid to handle the case of horizontal resolution change between fields
 // while in interlace mode, so clear the first LineWidths entry if it's == ~0, and
 // [...]
 const bool LineWidths_In_Valid = (LineWidths[0] != ~0);
 const bool WeaveGood = (StateValid && PrevDRect.h == DisplayRect.h && DeintType == DEINT_WEAVE);
 //
 // XReposition stuff is to prevent exceeding the dimensions of the video surface under certain conditions(weave deinterlacer, previous field has higher
 // horizontal resolution than current field, and current field's rectangle has an x offset that's too large when taking into consideration the previous field's
 // width; for simplicity, we don't check widths, but just assume that the previous field's maximum width is >= than the current field's maximum width).
 //
 const int32 XReposition = ((WeaveGood && DisplayRect.x > PrevDRect.x) ? DisplayRect.x : 0);

 //printf("%2d %2d, %d\n", DisplayRect.x, PrevDRect.x, XReposition);

 if(XReposition)
  DisplayRect.x = 0;

 if(surface->h && !LineWidths_In_Valid)
 {
  LineWidths[0] = 0;
 }

 for(int y = 0; y < DisplayRect.h / 2; y++)
 {
  // [...]
  // set all relevant source line widths to the contents of DisplayRect(also simplifies the src_lw and related pointer calculation code
  // farther below.
  if(!LineWidths_In_Valid)
   LineWidths[(y * 2) + field + DisplayRect.y] = DisplayRect.w;

  if(XReposition)
  {
    memmove(surface->pixels + ((y * 2) + field + DisplayRect.y) * surface->pitchinpix,
	    surface->pixels + ((y * 2) + field + DisplayRect.y) * surface->pitchinpix + XReposition,
	    LineWidths[(y * 2) + field + DisplayRect.y] * sizeof(T));
  }

  if(WeaveGood)
  {
   const T* src = FieldBuffer->pixels + y * FieldBuffer->pitchinpix;
   T* dest = surface->pixels + ((y * 2) + (field ^ 1) + DisplayRect.y) * surface->pitchinpix + DisplayRect.x;
   int32 *dest_lw = &LineWidths[(y * 2) + (field ^ 1) + DisplayRect.y];

   *dest_lw = LWBuffer[y];

   if (psx_gpu_upscale_shift == 0)
      memcpy(dest, src, LWBuffer[y] * sizeof(T));
  }
  else if(DeintType == DEINT_BOB)
  {
   const T* src = surface->pixels + ((y * 2) + field + DisplayRect.y) * surface->pitchinpix + DisplayRect.x;
   T* dest = surface->pixels + ((y * 2) + (field ^ 1) + DisplayRect.y) * surface->pitchinpix + DisplayRect.x;
   const int32 *src_lw = &LineWidths[(y * 2) + field + DisplayRect.y];
   int32 *dest_lw = &LineWidths[(y * 2) + (field ^ 1) + DisplayRect.y];

   *dest_lw = *src_lw;

   memcpy(dest, src, *src_lw * sizeof(T));
  }
  else
  {
   const int32 *src_lw = &LineWidths[(y * 2) + field + DisplayRect.y];
   const T* src = surface->pixels + ((y * 2) + field + DisplayRect.y) * surface->pitchinpix + DisplayRect.x;
   const int32 dly = ((y * 2) + (field + 1) + DisplayRect.y);
   T* dest = surface->pixels + dly * surface->pitchinpix + DisplayRect.x;

   if(y == 0 && field)
   {
    T black = MAKECOLOR(0, 0, 0, 0);
    T* dm2 = surface->pixels + (dly - 2) * surface->pitchinpix;

    LineWidths[dly - 2] = *src_lw;

    for(int x = 0; x < *src_lw; x++)
     dm2[x] = black;
   }

   if(dly < (DisplayRect.y + DisplayRect.h))
   {
    LineWidths[dly] = *src_lw;
    memcpy(dest, src, *src_lw * sizeof(T));
   }
  }

  //
  //
  //
  //
  //
  //
  if(DeintType == DEINT_WEAVE)
  {
   const int32 *src_lw = &LineWidths[(y * 2) + field + DisplayRect.y];
   const T* src = surface->pixels + ((y * 2) + field + DisplayRect.y) * surface->pitchinpix + DisplayRect.x;
   T* dest = FieldBuffer->pixels + y * FieldBuffer->pitchinpix;

   memcpy(dest, src, *src_lw * sizeof(uint32));
   LWBuffer[y] = *src_lw;

   StateValid = true;
  }
 }
}

void Deinterlacer::Process(MDFN_Surface *surface, MDFN_Rect &DisplayRect, int32 *LineWidths, const bool field)
{
 const MDFN_Rect DisplayRect_Original = DisplayRect;

 if(DeintType == DEINT_WEAVE)
 {
  if(!FieldBuffer || FieldBuffer->w < surface->w || FieldBuffer->h < (surface->h / 2))
  {
   if(FieldBuffer)
    delete FieldBuffer;

   FieldBuffer = new MDFN_Surface(NULL, surface->w, surface->h / 2, surface->w, surface->format);
   LWBuffer.resize(FieldBuffer->h);
  }
  else if(memcmp(&surface->format, &FieldBuffer->format, sizeof(MDFN_PixelFormat)))
  {
   FieldBuffer->SetFormat(surface->format, StateValid && PrevDRect.h == DisplayRect.h);
  }
 }

#if defined(WANT_32BPP)
 InternalProcess<uint32>(surface, DisplayRect, LineWidths, field);
#elif defined(WANT_16BPP)
 InternalProcess<uint16>(surface, DisplayRect, LineWidths, field);
#endif

 PrevDRect = DisplayRect_Original;
}

void Deinterlacer::ClearState(void)
{
 StateValid = false;

 PrevDRect.x = 0;
 PrevDRect.y = 0;

 PrevDRect.w = 0;
 PrevDRect.h = 0;
}