MVDegrain1::MVDegrain1(PClip _child, PClip mvbw, PClip mvfw, int _thSAD, int _thSADC, int _YUVplanes, int _nLimit,
PClip _pelclip, int _nIdx, int _nSCD1, int _nSCD2, bool _mmx, bool _isse, int _MaxThreads,
int _PreFetch, int _SadMode, IScriptEnvironment* env) :
MVDegrainBase(_child, 1, _YUVplanes, _nLimit, _pelclip, _nIdx, _mmx, _isse, env, mvfw,
"MVDegrain1", -1, _MaxThreads, _PreFetch, _SadMode)
{
// PreFetch max 21 since 21*3=63 and 64 is max threads at one time
if (_PreFetch<1 || _PreFetch>21) env->ThrowError("MVDegrain1: PreFetch must be >=1 and <=21");
// initialize MVClip's
for (unsigned int PreFetchNum=0; PreFetchNum<static_cast<unsigned int>(_PreFetch); ++PreFetchNum) {
pmvClipF[PreFetchNum][0]=new MVClip(mvfw, _nSCD1, _nSCD2, env, true);
pmvClipB[PreFetchNum][0]=new MVClip(mvbw, _nSCD1, _nSCD2, env, true);
}
CheckSimilarity(*pmvClipF[0][0], "mvfw", env);
CheckSimilarity(*pmvClipB[0][0], "mvbw", env);
// normalize thSAD
thSAD = _thSAD*pmvClipB[0][0]->GetThSCD1()/_nSCD1; // normalize to block SAD
thSADC = _thSADC*pmvClipB[0][0]->GetThSCD1()/_nSCD1; // chroma
// find the maximum extent
unsigned int MaxDelta=static_cast<unsigned int>(pmvClipF[0][0]->GetDeltaFrame());
if (static_cast<unsigned int>(pmvClipB[0][0]->GetDeltaFrame())>MaxDelta)
MaxDelta=static_cast<unsigned int>(pmvClipB[0][0]->GetDeltaFrame());
// numframes 2*MaxDelta+1, i.e. to cover all possible frames in sliding window
mvCore->AddFrames(nIdx, (2*MaxDelta)*_PreFetch+1, pmvClipB[0][0]->GetLevelCount(), nWidth, nHeight, nPel, nHPadding, nVPadding,
YUVPLANES, _isse, yRatioUV);
}
MVDegrainMulti::MVDegrainMulti(PClip _child, PClip mvMulti, int _RefFrames, int _thSAD, int _thSADC, int _YUVplanes, int _nLimit,
PClip _pelclip, int _nIdx, int _nSCD1, int _nSCD2, bool _mmx, bool _isse, int _MaxThreads,
int _PreFetch, int _SadMode, IScriptEnvironment* env) :
MVDegrainBase(_child, _RefFrames, _YUVplanes, _nLimit, _pelclip, _nIdx, _mmx, _isse, env, mvMulti,
"MVDegrainMulti", 0, _MaxThreads, _PreFetch, _SadMode), RefFrames(_RefFrames)
{
if (RefFrames<1 || RefFrames>32) env->ThrowError("MVDegrainMulti: refframes must be >=1 and <=32");
// get the true number of reference frames
VideoInfo mvMultivi=mvMulti->GetVideoInfo();
unsigned int RefFramesAvailable=mvMultivi.height/2;
// if refframes is greater than MVAnalyseMulti height then limit to height
if (RefFramesAvailable<RefFrames) {
RefFrames=RefFramesAvailable;
UpdateNumRefFrames(RefFrames, env);
}
// PreFetch max 21 since 21*3=63 and 64 is max threads at one time
if (_PreFetch<1 || _PreFetch>21) env->ThrowError("MVDegrainMulti: PreFetch must be >=1 and <=21");
if (_PreFetch*RefFrames>32) env->ThrowError("MVDegrainMulti: PreFetch*RefFrames<=32");
// initialize MVClip's which are in order BX, ..., B3, B2, B1, F1, F2, F3, ..., FX in mvMulti
for (unsigned int PreFetchNum=0; PreFetchNum<static_cast<unsigned int>(_PreFetch); ++PreFetchNum) {
if (RefFrames<RefFramesAvailable) {
// we are taking a subset of the mvMulti clip
for(unsigned int RefNum=0; RefNum<RefFrames; ++RefNum) {
pmvClipF[PreFetchNum][RefNum]=new MVClip(mvMulti, _nSCD1, _nSCD2, env, true, RefFramesAvailable+RefNum);
pmvClipB[PreFetchNum][RefNum]=new MVClip(mvMulti, _nSCD1, _nSCD2, env, true, RefFramesAvailable-RefNum-1);
}
}
else {
// we are taking the full mvMulti clip
for(unsigned int RefNum=0; RefNum<RefFrames; ++RefNum) {
pmvClipF[PreFetchNum][RefNum]=new MVClip(mvMulti, _nSCD1, _nSCD2, env, true, RefFrames+RefNum);
pmvClipB[PreFetchNum][RefNum]=new MVClip(mvMulti, _nSCD1, _nSCD2, env, true, RefFrames-RefNum-1);
}
}
}
// check simularities
CheckSimilarity(*pmvClipF[0][0], "mvMulti", env); // only need to check one since they are grouped together
// normalize thSAD
thSAD = _thSAD*pmvClipB[0][0]->GetThSCD1()/_nSCD1; // normalize to block SAD
thSADC = _thSADC*pmvClipB[0][0]->GetThSCD1()/_nSCD1; // chroma
// find the maximum extent
unsigned int MaxDelta=static_cast<unsigned int>(pmvClipF[0][RefFrames-1]->GetDeltaFrame());
if (static_cast<unsigned int>(pmvClipB[0][RefFrames-1]->GetDeltaFrame())>MaxDelta)
MaxDelta=static_cast<unsigned int>(pmvClipB[0][RefFrames-1]->GetDeltaFrame());
// numframes 2*MaxDelta+1, i.e. to cover all possible frames in sliding window
mvCore->AddFrames(nIdx, (2*MaxDelta)*_PreFetch+1, pmvClipB[0][0]->GetLevelCount(), nWidth, nHeight, nPel, nHPadding, nVPadding,
YUVPLANES, _isse, yRatioUV);
}
void Conversions::Copy(Image& Source, Image& Dest)
{
CheckSimilarity(Source, Dest);
if (Source.Step() == Dest.Step())
{
Source.SendIfNeeded();
m_CL->GetQueue().enqueueCopyBuffer(Source, Dest, 0, 0, Source.Step() * Source.Height());
Dest.SetInDevice();
return;
}
Kernel(copy, Source, Dest, Source.Step(), Dest.Step());
}
MVFlowBlur::MVFlowBlur(PClip _child, PClip super, PClip _mvbw, PClip _mvfw, int _blur256, int _prec,
int nSCD1, int nSCD2, bool _isse, bool _planar, IScriptEnvironment* env) :
GenericVideoFilter(_child),
MVFilter(_mvfw, "MFlowBlur", env, 1, 0),
mvClipB(_mvbw, nSCD1, nSCD2, env, 1, 0),
mvClipF(_mvfw, nSCD1, nSCD2, env, 1, 0)
{
blur256 = _blur256;
prec = _prec;
isse = _isse;
planar = _planar;
CheckSimilarity(mvClipB, "mvbw", env);
CheckSimilarity(mvClipF, "mvfw", env);
SuperParams64Bits params;
memcpy(¶ms, &super->GetVideoInfo().num_audio_samples, 8);
int nHeightS = params.nHeight;
int nSuperHPad = params.nHPad;
int nSuperVPad = params.nVPad;
int nSuperPel = params.nPel;
int nSuperModeYUV = params.nModeYUV;
int nSuperLevels = params.nLevels;
int nSuperWidth = super->GetVideoInfo().width; // really super
int nSuperHeight = super->GetVideoInfo().height;
if ( nHeight != nHeightS
|| nWidth != nSuperWidth - nSuperHPad * 2
|| nPel != nSuperPel)
{
env->ThrowError("MFlowBlur : wrong super frame clip");
}
if (nPel==1)
finest = super; // v2.0.9.1
else
{
finest = new MVFinest(super, isse, env);
AVSValue cache_args[1] = { finest };
finest = env->Invoke("InternalCache", AVSValue(cache_args,1)).AsClip(); // add cache for speed
}
// if ( nWidth != vi.width || (nWidth + nHPadding*2)*nPel != finest->GetVideoInfo().width
// || nHeight != vi.height || (nHeight + nVPadding*2)*nPel != finest->GetVideoInfo().height)
// env->ThrowError("MVFlowBlur: wrong source of finest frame size");
nHeightUV = nHeight/yRatioUV;
nWidthUV = nWidth/2;// for YV12
nHPaddingUV = nHPadding/2;
nVPaddingUV = nHPadding/yRatioUV;
VPitchY = nWidth;
VPitchUV= nWidthUV;
VXFullYB = new BYTE [nHeight*VPitchY];
VXFullUVB = new BYTE [nHeightUV*VPitchUV];
VYFullYB = new BYTE [nHeight*VPitchY];
VYFullUVB = new BYTE [nHeightUV*VPitchUV];
VXFullYF = new BYTE [nHeight*VPitchY];
VXFullUVF = new BYTE [nHeightUV*VPitchUV];
VYFullYF = new BYTE [nHeight*VPitchY];
VYFullUVF = new BYTE [nHeightUV*VPitchUV];
VXSmallYB = new BYTE [nBlkX*nBlkY];
VYSmallYB = new BYTE [nBlkX*nBlkY];
VXSmallUVB = new BYTE [nBlkX*nBlkY];
VYSmallUVB = new BYTE [nBlkX*nBlkY];
VXSmallYF = new BYTE [nBlkX*nBlkY];
VYSmallYF = new BYTE [nBlkX*nBlkY];
VXSmallUVF = new BYTE [nBlkX*nBlkY];
VYSmallUVF = new BYTE [nBlkX*nBlkY];
MaskSmallB = new BYTE [nBlkX*nBlkY];
MaskFullYB = new BYTE [nHeight*VPitchY];
MaskFullUVB = new BYTE [nHeightUV*VPitchUV];
MaskSmallF = new BYTE [nBlkX*nBlkY];
MaskFullYF = new BYTE [nHeight*VPitchY];
MaskFullUVF = new BYTE [nHeightUV*VPitchUV];
int CPUF_Resize = env->GetCPUFlags();
if (!isse) CPUF_Resize = (CPUF_Resize & !CPUF_INTEGER_SSE) & !CPUF_SSE2;
upsizer = new SimpleResize(nWidth, nHeight, nBlkX, nBlkY, CPUF_Resize);
upsizerUV = new SimpleResize(nWidthUV, nHeightUV, nBlkX, nBlkY, CPUF_Resize);
if ( (pixelType & VideoInfo::CS_YUY2) == VideoInfo::CS_YUY2 && !planar)
{
DstPlanes = new YUY2Planes(nWidth, nHeight);
}
}
MVFlowBlur::MVFlowBlur(PClip _child, PClip _mvbw, PClip _mvfw, int _blur256, int _prec, PClip _pelclip, int _nIdx,
int nSCD1, int nSCD2, bool _mmx, bool _isse, IScriptEnvironment* env) :
GenericVideoFilter(_child),
MVFilter(_mvfw, "MVFlowBlur", env),
mvClipB(_mvbw, nSCD1, nSCD2, env, true),
mvClipF(_mvfw, nSCD1, nSCD2, env, true),
pelclip(_pelclip)
{
blur256 = _blur256;
prec = _prec;
nIdx = _nIdx;
mmx = _mmx;
isse = _isse;
CheckSimilarity(mvClipB, "mvbw", env);
CheckSimilarity(mvClipF, "mvfw", env);
mvCore->AddFrames(nIdx, 3, mvClipB.GetLevelCount(), nWidth, nHeight,
nPel, nHPadding, nVPadding, YUVPLANES, isse, yRatioUV);
// mvCore->AddFrames(nIdx, MV_BUFFER_FRAMES, mvClipF.GetLevelCount(), nWidth, nHeight,
// nPel, nBlkSize, nBlkSize, YUVPLANES, isse, yRatioUV);
usePelClipHere = false;
if (pelclip && (nPel > 1)) {
if (pelclip->GetVideoInfo().width != nWidth*nPel || pelclip->GetVideoInfo().height != nHeight*nPel)
env->ThrowError("MVFlowBlur: pelclip frame size must be 2X of source!");
else
usePelClipHere = true;
}
nHeightUV = nHeight/yRatioUV;
nWidthUV = nWidth/2;// for YV12
nHPaddingUV = nHPadding/2;
nVPaddingUV = nHPadding/yRatioUV;
VPitchY = nWidth;
VPitchUV= nWidthUV;
VXFullYB = new BYTE [nHeight*VPitchY];
VXFullUVB = new BYTE [nHeightUV*VPitchUV];
VYFullYB = new BYTE [nHeight*VPitchY];
VYFullUVB = new BYTE [nHeightUV*VPitchUV];
VXFullYF = new BYTE [nHeight*VPitchY];
VXFullUVF = new BYTE [nHeightUV*VPitchUV];
VYFullYF = new BYTE [nHeight*VPitchY];
VYFullUVF = new BYTE [nHeightUV*VPitchUV];
VXSmallYB = new BYTE [nBlkX*nBlkY];
VYSmallYB = new BYTE [nBlkX*nBlkY];
VXSmallUVB = new BYTE [nBlkX*nBlkY];
VYSmallUVB = new BYTE [nBlkX*nBlkY];
VXSmallYF = new BYTE [nBlkX*nBlkY];
VYSmallYF = new BYTE [nBlkX*nBlkY];
VXSmallUVF = new BYTE [nBlkX*nBlkY];
VYSmallUVF = new BYTE [nBlkX*nBlkY];
MaskSmallB = new BYTE [nBlkX*nBlkY];
MaskFullYB = new BYTE [nHeight*VPitchY];
MaskFullUVB = new BYTE [nHeightUV*VPitchUV];
MaskSmallF = new BYTE [nBlkX*nBlkY];
MaskFullYF = new BYTE [nHeight*VPitchY];
MaskFullUVF = new BYTE [nHeightUV*VPitchUV];
int pel2WidthY = (nWidth + 2*nHPadding)*nPel;
pel2HeightY = (nHeight + 2*nVPadding)*nPel;
int pel2WidthUV = (nWidthUV + 2*nHPaddingUV)*nPel;
pel2HeightUV = (nHeightUV + 2*nVPaddingUV)*nPel;
pel2PitchY = (pel2WidthY + 15) & (~15);
pel2PitchUV = (pel2WidthUV + 15) & (~15);
pel2OffsetY = pel2PitchY * nVPadding*nPel + nHPadding*nPel;
pel2OffsetUV = pel2PitchUV * nVPaddingUV*nPel + nHPaddingUV*nPel;
if (nPel>1) {
pel2PlaneYB = new BYTE [pel2PitchY*pel2HeightY];
pel2PlaneUB = new BYTE [pel2PitchUV*pel2HeightUV];
pel2PlaneVB = new BYTE [pel2PitchUV*pel2HeightUV];
}
int CPUF_Resize = env->GetCPUFlags();
if (!isse) CPUF_Resize = (CPUF_Resize & !CPUF_INTEGER_SSE) & !CPUF_SSE2;
upsizer = new SimpleResize(nWidth, nHeight, nBlkX, nBlkY, CPUF_Resize);
upsizerUV = new SimpleResize(nWidthUV, nHeightUV, nBlkX, nBlkY, CPUF_Resize);
DstPlanes = new YUY2Planes(nWidth, nHeight, vi.pixel_type, isse);
}
MVDegrain1::MVDegrain1(
PClip _child, PClip _super, PClip mvbw, PClip mvfw,
int _thSAD, int _thSADC, int _YUVplanes, int _nLimit, int _nLimitC,
int _nSCD1, int _nSCD2, bool _isse, bool _planar, bool _lsb_flag,
bool mt_flag, IScriptEnvironment* env
)
: GenericVideoFilter(_child)
, MVFilter ((! mvfw) ? mvbw : mvfw, "MDegrain1", env, (! mvfw) ? 2 : 1, (! mvfw) ? 1 : 0)
, mvClipF ((! mvfw) ? mvbw : mvfw, _nSCD1, _nSCD2, env, (! mvfw) ? 2 : 1, (! mvfw) ? 1 : 0)
, mvClipB ((! mvfw) ? mvbw : mvbw, _nSCD1, _nSCD2, env, (! mvfw) ? 2 : 1, (! mvfw) ? 0 : 0)
, super (_super)
, lsb_flag (_lsb_flag)
, height_lsb_mul ((_lsb_flag) ? 2 : 1)
, DstShort (0)
, DstInt (0)
{
thSAD = _thSAD*mvClipB.GetThSCD1()/_nSCD1; // normalize to block SAD
thSADC = _thSADC*mvClipB.GetThSCD1()/_nSCD1; // chroma threshold, normalized to block SAD
YUVplanes = _YUVplanes;
nLimit = _nLimit;
nLimitC = _nLimitC;
isse = _isse;
planar = _planar;
CheckSimilarity(mvClipF, "mvfw", env);
CheckSimilarity(mvClipB, "mvbw", env);
const ::VideoInfo & vi_super = _super->GetVideoInfo ();
// get parameters of prepared super clip - v2.0
SuperParams64Bits params;
memcpy(¶ms, &vi_super.num_audio_samples, 8);
int nHeightS = params.nHeight;
int nSuperHPad = params.nHPad;
int nSuperVPad = params.nVPad;
int nSuperPel = params.nPel;
nSuperModeYUV = params.nModeYUV;
int nSuperLevels = params.nLevels;
pRefBGOF = new MVGroupOfFrames(nSuperLevels, nWidth, nHeight, nSuperPel, nSuperHPad, nSuperVPad, nSuperModeYUV, isse, yRatioUV, mt_flag);
pRefFGOF = new MVGroupOfFrames(nSuperLevels, nWidth, nHeight, nSuperPel, nSuperHPad, nSuperVPad, nSuperModeYUV, isse, yRatioUV, mt_flag);
int nSuperWidth = vi_super.width;
int nSuperHeight = vi_super.height;
if ( nHeight != nHeightS
|| nHeight != vi.height
|| nWidth != nSuperWidth-nSuperHPad*2
|| nWidth != vi.width
|| nPel != nSuperPel)
{
env->ThrowError("MDegrain1 : wrong source or super frame size");
}
if ( (pixelType & VideoInfo::CS_YUY2) == VideoInfo::CS_YUY2 && !planar)
{
DstPlanes = new YUY2Planes(nWidth, nHeight * height_lsb_mul);
SrcPlanes = new YUY2Planes(nWidth, nHeight);
}
dstShortPitch = ((nWidth + 15)/16)*16;
dstIntPitch = dstShortPitch;
if (nOverlapX >0 || nOverlapY>0)
{
OverWins = new OverlapWindows(nBlkSizeX, nBlkSizeY, nOverlapX, nOverlapY);
OverWinsUV = new OverlapWindows(nBlkSizeX/2, nBlkSizeY/yRatioUV, nOverlapX/2, nOverlapY/yRatioUV);
if (lsb_flag)
{
DstInt = new int [dstIntPitch * nHeight];
}
else
{
DstShort = new unsigned short[dstShortPitch*nHeight];
}
}
switch (nBlkSizeX)
{
case 32:
if (nBlkSizeY==16) { OVERSLUMALSB = OverlapsLsb_C<32,16>;
if (yRatioUV==2) { OVERSCHROMALSB = OverlapsLsb_C<16,8>; }
else { OVERSCHROMALSB = OverlapsLsb_C<16,16>; }
} else if (nBlkSizeY==32) { OVERSLUMALSB = OverlapsLsb_C<32,32>;
if (yRatioUV==2) { OVERSCHROMALSB = OverlapsLsb_C<16,16>; }
else { OVERSCHROMALSB = OverlapsLsb_C<16,32>; }
} break;
case 16:
if (nBlkSizeY==16) { OVERSLUMALSB = OverlapsLsb_C<16,16>;
if (yRatioUV==2) { OVERSCHROMALSB = OverlapsLsb_C<8,8>; }
else { OVERSCHROMALSB = OverlapsLsb_C<8,16>; }
} else if (nBlkSizeY==8) { OVERSLUMALSB = OverlapsLsb_C<16,8>;
if (yRatioUV==2) { OVERSCHROMALSB = OverlapsLsb_C<8,4>; }
else { OVERSCHROMALSB = OverlapsLsb_C<8,8>; }
} else if (nBlkSizeY==2) { OVERSLUMALSB = OverlapsLsb_C<16,2>;
if (yRatioUV==2) { OVERSCHROMALSB = OverlapsLsb_C<8,1>; }
else { OVERSCHROMALSB = OverlapsLsb_C<8,2>; }
}
break;
case 4:
OVERSLUMALSB = OverlapsLsb_C<4,4>;
if (yRatioUV==2) { OVERSCHROMALSB = OverlapsLsb_C<2,2>; }
else { OVERSCHROMALSB = OverlapsLsb_C<2,4>; }
break;
case 8:
default:
if (nBlkSizeY==8) { OVERSLUMALSB = OverlapsLsb_C<8,8>;
if (yRatioUV==2) { OVERSCHROMALSB = OverlapsLsb_C<4,4>; }
else { OVERSCHROMALSB = OverlapsLsb_C<4,8>; }
}else if (nBlkSizeY==4) { OVERSLUMALSB = OverlapsLsb_C<8,4>;
if (yRatioUV==2) { OVERSCHROMALSB = OverlapsLsb_C<4,2>; }
else { OVERSCHROMALSB = OverlapsLsb_C<4,4>; }
}
}
if ( ((env->GetCPUFlags() & CPUF_SSE2) != 0) & isse)
{
switch (nBlkSizeX)
{
case 32:
if (nBlkSizeY==16) { OVERSLUMA = Overlaps32x16_sse2; DEGRAINLUMA = Degrain1_sse2<32,16>;
if (yRatioUV==2) { OVERSCHROMA = Overlaps16x8_sse2; DEGRAINCHROMA = Degrain1_sse2<16,8>; }
else { OVERSCHROMA = Overlaps16x16_sse2;DEGRAINCHROMA = Degrain1_sse2<16,16>; }
} else if (nBlkSizeY==32) { OVERSLUMA = Overlaps32x32_sse2; DEGRAINLUMA = Degrain1_sse2<32,32>;
if (yRatioUV==2) { OVERSCHROMA = Overlaps16x16_sse2; DEGRAINCHROMA = Degrain1_sse2<16,16>; }
else { OVERSCHROMA = Overlaps16x32_sse2; DEGRAINCHROMA = Degrain1_sse2<16,32>; }
} break;
case 16:
if (nBlkSizeY==16) { OVERSLUMA = Overlaps16x16_sse2; DEGRAINLUMA = Degrain1_sse2<16,16>;
if (yRatioUV==2) { OVERSCHROMA = Overlaps8x8_sse2; DEGRAINCHROMA = Degrain1_sse2<8,8>; }
else { OVERSCHROMA = Overlaps8x16_sse2;DEGRAINCHROMA = Degrain1_sse2<8,16>; }
} else if (nBlkSizeY==8) { OVERSLUMA = Overlaps16x8_sse2; DEGRAINLUMA = Degrain1_sse2<16,8>;
if (yRatioUV==2) { OVERSCHROMA = Overlaps8x4_sse2; DEGRAINCHROMA = Degrain1_sse2<8,4>; }
else { OVERSCHROMA = Overlaps8x8_sse2; DEGRAINCHROMA = Degrain1_sse2<8,8>; }
} else if (nBlkSizeY==2) { OVERSLUMA = Overlaps16x2_sse2; DEGRAINLUMA = Degrain1_sse2<16,2>;
if (yRatioUV==2) { OVERSCHROMA = Overlaps8x1_sse2; DEGRAINCHROMA = Degrain1_sse2<8,1>; }
else { OVERSCHROMA = Overlaps8x2_sse2; DEGRAINCHROMA = Degrain1_sse2<8,2>; }
}
break;
case 4:
OVERSLUMA = Overlaps4x4_sse2; DEGRAINLUMA = Degrain1_mmx<4,4>;
if (yRatioUV==2) { OVERSCHROMA = Overlaps_C<2,2>; DEGRAINCHROMA = Degrain1_C<2,2>; }
else { OVERSCHROMA = Overlaps_C<2,4>; DEGRAINCHROMA = Degrain1_C<2,4>; }
break;
case 8:
default:
if (nBlkSizeY==8) { OVERSLUMA = Overlaps8x8_sse2; DEGRAINLUMA = Degrain1_sse2<8,8>;
if (yRatioUV==2) { OVERSCHROMA = Overlaps4x4_sse2; DEGRAINCHROMA = Degrain1_mmx<4,4>; }
else { OVERSCHROMA = Overlaps4x8_sse2; DEGRAINCHROMA = Degrain1_mmx<4,8>; }
}else if (nBlkSizeY==4) { OVERSLUMA = Overlaps8x4_sse2; DEGRAINLUMA = Degrain1_sse2<8,4>;
if (yRatioUV==2) { OVERSCHROMA = Overlaps4x2_sse2; DEGRAINCHROMA = Degrain1_mmx<4,2>; }
else { OVERSCHROMA = Overlaps4x4_sse2; DEGRAINCHROMA = Degrain1_mmx<4,4>; }
}
}
}
else if ( isse )
{
switch (nBlkSizeX)
{
case 32:
if (nBlkSizeY==16) { OVERSLUMA = Overlaps32x16_sse2; DEGRAINLUMA = Degrain1_mmx<32,16>;
if (yRatioUV==2) { OVERSCHROMA = Overlaps16x8_sse2; DEGRAINCHROMA = Degrain1_mmx<16,8>; }
else { OVERSCHROMA = Overlaps16x16_sse2;DEGRAINCHROMA = Degrain1_mmx<16,16>; }
} else if (nBlkSizeY==32) { OVERSLUMA = Overlaps32x32_sse2; DEGRAINLUMA = Degrain1_mmx<32,32>;
if (yRatioUV==2) { OVERSCHROMA = Overlaps16x16_sse2; DEGRAINCHROMA = Degrain1_mmx<16,16>; }
else { OVERSCHROMA = Overlaps16x32_sse2; DEGRAINCHROMA = Degrain1_mmx<16,32>; }
} break;
case 16:
if (nBlkSizeY==16) { OVERSLUMA = Overlaps16x16_sse2; DEGRAINLUMA = Degrain1_mmx<16,16>;
if (yRatioUV==2) { OVERSCHROMA = Overlaps8x8_sse2; DEGRAINCHROMA = Degrain1_mmx<8,8>; }
else { OVERSCHROMA = Overlaps8x16_sse2;DEGRAINCHROMA = Degrain1_mmx<8,16>; }
} else if (nBlkSizeY==8) { OVERSLUMA = Overlaps16x8_sse2; DEGRAINLUMA = Degrain1_mmx<16,8>;
if (yRatioUV==2) { OVERSCHROMA = Overlaps8x4_sse2; DEGRAINCHROMA = Degrain1_mmx<8,4>; }
else { OVERSCHROMA = Overlaps8x8_sse2; DEGRAINCHROMA = Degrain1_mmx<8,8>; }
} else if (nBlkSizeY==2) { OVERSLUMA = Overlaps16x2_sse2; DEGRAINLUMA = Degrain1_mmx<16,2>;
if (yRatioUV==2) { OVERSCHROMA = Overlaps8x1_sse2; DEGRAINCHROMA = Degrain1_mmx<8,1>; }
else { OVERSCHROMA = Overlaps8x2_sse2; DEGRAINCHROMA = Degrain1_mmx<8,2>; }
}
break;
case 4:
OVERSLUMA = Overlaps4x4_sse2; DEGRAINLUMA = Degrain1_mmx<4,4>;
if (yRatioUV==2) { OVERSCHROMA = Overlaps_C<2,2>; DEGRAINCHROMA = Degrain1_C<2,2>; }
else { OVERSCHROMA = Overlaps_C<2,4>; DEGRAINCHROMA = Degrain1_C<2,4>; }
break;
case 8:
default:
if (nBlkSizeY==8) { OVERSLUMA = Overlaps8x8_sse2; DEGRAINLUMA = Degrain1_mmx<8,8>;
if (yRatioUV==2) { OVERSCHROMA = Overlaps4x4_sse2; DEGRAINCHROMA = Degrain1_mmx<4,4>; }
else { OVERSCHROMA = Overlaps4x8_sse2; DEGRAINCHROMA = Degrain1_mmx<4,8>; }
}else if (nBlkSizeY==4) { OVERSLUMA = Overlaps8x4_sse2; DEGRAINLUMA = Degrain1_mmx<8,4>;
if (yRatioUV==2) { OVERSCHROMA = Overlaps4x2_sse2; DEGRAINCHROMA = Degrain1_mmx<4,2>; }
else { OVERSCHROMA = Overlaps4x4_sse2; DEGRAINCHROMA = Degrain1_mmx<4,4>; }
}
}
}
else
{
switch (nBlkSizeX)
{
case 32:
if (nBlkSizeY==16) { OVERSLUMA = Overlaps_C<32,16>; DEGRAINLUMA = Degrain1_C<32,16>;
if (yRatioUV==2) { OVERSCHROMA = Overlaps_C<16,8>; DEGRAINCHROMA = Degrain1_C<16,8>; }
else { OVERSCHROMA = Overlaps_C<16,16>;DEGRAINCHROMA = Degrain1_C<16,16>; }
} else if (nBlkSizeY==32) { OVERSLUMA = Overlaps_C<32,32>; DEGRAINLUMA = Degrain1_C<32,32>;
if (yRatioUV==2) { OVERSCHROMA = Overlaps_C<16,16>; DEGRAINCHROMA = Degrain1_C<16,16>; }
else { OVERSCHROMA = Overlaps_C<16,32>; DEGRAINCHROMA = Degrain1_C<16,32>; }
} break;
case 16:
if (nBlkSizeY==16) { OVERSLUMA = Overlaps_C<16,16>; DEGRAINLUMA = Degrain1_C<16,16>;
if (yRatioUV==2) { OVERSCHROMA = Overlaps_C<8,8>; DEGRAINCHROMA = Degrain1_C<8,8>; }
else { OVERSCHROMA = Overlaps_C<8,16>; DEGRAINCHROMA = Degrain1_C<8,16>; }
} else if (nBlkSizeY==8) { OVERSLUMA = Overlaps_C<16,8>; DEGRAINLUMA = Degrain1_C<16,8>;
if (yRatioUV==2) { OVERSCHROMA = Overlaps_C<8,4>; DEGRAINCHROMA = Degrain1_C<8,4>; }
else { OVERSCHROMA = Overlaps_C<8,8>; DEGRAINCHROMA = Degrain1_C<8,8>; }
} else if (nBlkSizeY==2) { OVERSLUMA = Overlaps_C<16,2>; DEGRAINLUMA = Degrain1_C<16,2>;
if (yRatioUV==2) { OVERSCHROMA = Overlaps_C<8,1>; DEGRAINCHROMA = Degrain1_C<8,1>; }
else { OVERSCHROMA = Overlaps_C<8,2>; DEGRAINCHROMA = Degrain1_C<8,2>; }
}
break;
case 4:
OVERSLUMA = Overlaps_C<4,4>; DEGRAINLUMA = Degrain1_C<4,4>;
if (yRatioUV==2) { OVERSCHROMA = Overlaps_C<2,2>; DEGRAINCHROMA = Degrain1_C<2,2>; }
else { OVERSCHROMA = Overlaps_C<2,4>; DEGRAINCHROMA = Degrain1_C<2,4>; }
break;
case 8:
default:
if (nBlkSizeY==8) { OVERSLUMA = Overlaps_C<8,8>; DEGRAINLUMA = Degrain1_C<8,8>;
if (yRatioUV==2) { OVERSCHROMA = Overlaps_C<4,4>; DEGRAINCHROMA = Degrain1_C<4,4>; }
else { OVERSCHROMA = Overlaps_C<4,8>; DEGRAINCHROMA = Degrain1_C<4,8>; }
}else if (nBlkSizeY==4) { OVERSLUMA = Overlaps_C<8,4>; DEGRAINLUMA = Degrain1_C<8,4>;
if (yRatioUV==2) { OVERSCHROMA = Overlaps_C<4,2>; DEGRAINCHROMA = Degrain1_C<4,2>; }
else { OVERSCHROMA = Overlaps_C<4,4>; DEGRAINCHROMA = Degrain1_C<4,4>; }
}
}
}
const int tmp_size = 32 * 32;
tmpBlock = new BYTE[tmp_size * height_lsb_mul];
tmpBlockLsb = (lsb_flag) ? (tmpBlock + tmp_size) : 0;
if (lsb_flag)
{
vi.height <<= 1;
}
}
MVFlowInter::MVFlowInter(PClip _child, PClip super, PClip _mvbw, PClip _mvfw, int _time256, double _ml,
bool _blend, int nSCD1, int nSCD2, bool _isse, bool _planar, IScriptEnvironment* env) :
GenericVideoFilter(_child),
MVFilter(_mvfw, "MFlowInter", env),
mvClipB(_mvbw, nSCD1, nSCD2, env),
mvClipF(_mvfw, nSCD1, nSCD2, env)
{
time256 = _time256;
ml = _ml;
isse = _isse;
planar = planar;
blend = _blend;
if (!mvClipB.IsBackward())
env->ThrowError("MFlowInter: wrong backward vectors");
if (mvClipF.IsBackward())
env->ThrowError("MFlowInter: wrong forward vectors");
CheckSimilarity(mvClipB, "mvbw", env);
CheckSimilarity(mvClipF, "mvfw", env);
SuperParams64Bits params;
memcpy(¶ms, &super->GetVideoInfo().num_audio_samples, 8);
int nHeightS = params.nHeight;
int nSuperHPad = params.nHPad;
int nSuperVPad = params.nVPad;
int nSuperPel = params.nPel;
int nSuperModeYUV = params.nModeYUV;
int nSuperLevels = params.nLevels;
int nSuperWidth = super->GetVideoInfo().width; // really super
int nSuperHeight = super->GetVideoInfo().height;
if (nHeight != nHeightS || nWidth != nSuperWidth-nSuperHPad*2)
env->ThrowError("MFlowInter : wrong super frame clip");
if (nPel==1)
finest = super; // v2.0.9.1
else
{
finest = new MVFinest(super, isse, env);
AVSValue cache_args[1] = { finest };
finest = env->Invoke("InternalCache", AVSValue(cache_args,1)).AsClip(); // add cache for speed
}
// if (nWidth != vi.width || (nWidth + nHPadding*2)*nPel != finest->GetVideoInfo().width ||
// nHeight != vi.height || (nHeight + nVPadding*2)*nPel != finest->GetVideoInfo().height )
// env->ThrowError("MVFlowInter: wrong source or finest frame size");
// may be padded for full frame cover
nBlkXP = (nBlkX*(nBlkSizeX - nOverlapX) + nOverlapX < nWidth) ? nBlkX+1 : nBlkX;
nBlkYP = (nBlkY*(nBlkSizeY - nOverlapY) + nOverlapY < nHeight) ? nBlkY+1 : nBlkY;
nWidthP = nBlkXP*(nBlkSizeX - nOverlapX) + nOverlapX;
nHeightP = nBlkYP*(nBlkSizeY - nOverlapY) + nOverlapY;
// for YV12
nWidthPUV = nWidthP/2;
nHeightPUV = nHeightP/yRatioUV;
nHeightUV = nHeight/yRatioUV;
nWidthUV = nWidth/2;
nHPaddingUV = nHPadding/2;
nVPaddingUV = nVPadding/yRatioUV;
VPitchY = (nWidthP + 15) & (~15);
VPitchUV = (nWidthPUV + 15) & (~15);
VXFullYB = new BYTE [nHeightP*VPitchY];
VXFullUVB = new BYTE [nHeightPUV*VPitchUV];
VYFullYB = new BYTE [nHeightP*VPitchY];
VYFullUVB = new BYTE [nHeightPUV*VPitchUV];
VXFullYF = new BYTE [nHeightP*VPitchY];
VXFullUVF = new BYTE [nHeightPUV*VPitchUV];
VYFullYF = new BYTE [nHeightP*VPitchY];
VYFullUVF = new BYTE [nHeightPUV*VPitchUV];
VXSmallYB = new BYTE [nBlkXP*nBlkYP];
VYSmallYB = new BYTE [nBlkXP*nBlkYP];
VXSmallUVB = new BYTE [nBlkXP*nBlkYP];
VYSmallUVB = new BYTE [nBlkXP*nBlkYP];
VXSmallYF = new BYTE [nBlkXP*nBlkYP];
VYSmallYF = new BYTE [nBlkXP*nBlkYP];
VXSmallUVF = new BYTE [nBlkXP*nBlkYP];
VYSmallUVF = new BYTE [nBlkXP*nBlkYP];
VXFullYBB = new BYTE [nHeightP*VPitchY];
VXFullUVBB = new BYTE [nHeightPUV*VPitchUV];
VYFullYBB = new BYTE [nHeightP*VPitchY];
VYFullUVBB = new BYTE [nHeightPUV*VPitchUV];
VXFullYFF = new BYTE [nHeightP*VPitchY];
VXFullUVFF = new BYTE [nHeightPUV*VPitchUV];
VYFullYFF = new BYTE [nHeightP*VPitchY];
VYFullUVFF = new BYTE [nHeightPUV*VPitchUV];
VXSmallYBB = new BYTE [nBlkXP*nBlkYP];
VYSmallYBB = new BYTE [nBlkXP*nBlkYP];
VXSmallUVBB = new BYTE [nBlkXP*nBlkYP];
VYSmallUVBB = new BYTE [nBlkXP*nBlkYP];
VXSmallYFF = new BYTE [nBlkXP*nBlkYP];
VYSmallYFF = new BYTE [nBlkXP*nBlkYP];
VXSmallUVFF = new BYTE [nBlkXP*nBlkYP];
VYSmallUVFF = new BYTE [nBlkXP*nBlkYP];
MaskSmallB = new BYTE [nBlkXP*nBlkYP];
MaskFullYB = new BYTE [nHeightP*VPitchY];
MaskFullUVB = new BYTE [nHeightPUV*VPitchUV];
MaskSmallF = new BYTE [nBlkXP*nBlkYP];
MaskFullYF = new BYTE [nHeightP*VPitchY];
MaskFullUVF = new BYTE [nHeightPUV*VPitchUV];
SADMaskSmallB = new BYTE [nBlkXP*nBlkYP];
SADMaskSmallF = new BYTE [nBlkXP*nBlkYP];
int CPUF_Resize = env->GetCPUFlags();
if (!isse) CPUF_Resize = (CPUF_Resize & !CPUF_INTEGER_SSE) & !CPUF_SSE2;
upsizer = new SimpleResize(nWidthP, nHeightP, nBlkXP, nBlkYP, CPUF_Resize);
upsizerUV = new SimpleResize(nWidthPUV, nHeightPUV, nBlkXP, nBlkYP, CPUF_Resize);
LUTVB = new int[256];
LUTVF = new int[256];
Create_LUTV(time256, LUTVB, LUTVF);
if ( (pixelType & VideoInfo::CS_YUY2) == VideoInfo::CS_YUY2 && !planar)
{
DstPlanes = new YUY2Planes(nWidth, nHeight);
}
}
MVFlowInter::MVFlowInter(PClip _child, PClip _mvbw, PClip _mvfw, int _time256, double _ml, double _mSAD,
double _mSADC, PClip _pelclip, int _nIdx, int nSCD1, int nSCD2, bool _mmx, bool _isse,
IScriptEnvironment* env) :
GenericVideoFilter(_child),
MVFilter(_mvfw, "MVFlowInter", env),
mvClipB(_mvbw, nSCD1, nSCD2, env, true),
mvClipF(_mvfw, nSCD1, nSCD2, env, true),
pelclip(_pelclip)
{
time256 = _time256;
ml = _ml;
mSAD = _mSAD;
mSADC = _mSADC;
nIdx = _nIdx;
mmx = _mmx;
isse = _isse;
normSAD1024 = unsigned int( 1024 * 255 / (mSAD * nBlkSizeX*nBlkSizeY) ); //normalize
normSADC1024 = unsigned int( 1024 * 255 / (mSADC * nBlkSizeX*nBlkSizeY) ); //normalize
if (!mvClipB.IsBackward())
env->ThrowError("MVFlowInter: wrong backward vectors");
if (mvClipF.IsBackward())
env->ThrowError("MVFlowInter: wrong forward vectors");
CheckSimilarity(mvClipB, "mvbw", env);
CheckSimilarity(mvClipF, "mvfw", env);
mvCore->AddFrames(nIdx, 3, mvClipB.GetLevelCount(), nWidth, nHeight,
nPel, nHPadding, nVPadding, YUVPLANES, isse, yRatioUV);
usePelClipHere = false;
if (pelclip && (nPel > 1)) {
if (pelclip->GetVideoInfo().width != nWidth*nPel || pelclip->GetVideoInfo().height != nHeight*nPel)
env->ThrowError("MVFlowInter: pelclip frame size must be Pel of source!");
else
usePelClipHere = true;
}
// may be padded for full frame cover
nBlkXP = (nBlkX*(nBlkSizeX - nOverlapX) + nOverlapX < nWidth) ? nBlkX+1 : nBlkX;
nBlkYP = (nBlkY*(nBlkSizeY - nOverlapY) + nOverlapY < nHeight) ? nBlkY+1 : nBlkY;
nWidthP = nBlkXP*(nBlkSizeX - nOverlapX) + nOverlapX;
nHeightP = nBlkYP*(nBlkSizeY - nOverlapY) + nOverlapY;
// for YV12
nWidthPUV = nWidthP/2;
nHeightPUV = nHeightP/yRatioUV;
nHeightUV = nHeight/yRatioUV;
nWidthUV = nWidth/2;
nHPaddingUV = nHPadding/2;
nVPaddingUV = nVPadding/yRatioUV;
VPitchY = (nWidthP + 15) & (~15);
VPitchUV = (nWidthPUV + 15) & (~15);
VXFullYB = new BYTE [nHeightP*VPitchY];
VXFullUVB = new BYTE [nHeightPUV*VPitchUV];
VYFullYB = new BYTE [nHeightP*VPitchY];
VYFullUVB = new BYTE [nHeightPUV*VPitchUV];
VXFullYF = new BYTE [nHeightP*VPitchY];
VXFullUVF = new BYTE [nHeightPUV*VPitchUV];
VYFullYF = new BYTE [nHeightP*VPitchY];
VYFullUVF = new BYTE [nHeightPUV*VPitchUV];
VXSmallYB = new BYTE [nBlkXP*nBlkYP];
VYSmallYB = new BYTE [nBlkXP*nBlkYP];
VXSmallUVB = new BYTE [nBlkXP*nBlkYP];
VYSmallUVB = new BYTE [nBlkXP*nBlkYP];
VXSmallYF = new BYTE [nBlkXP*nBlkYP];
VYSmallYF = new BYTE [nBlkXP*nBlkYP];
VXSmallUVF = new BYTE [nBlkXP*nBlkYP];
VYSmallUVF = new BYTE [nBlkXP*nBlkYP];
VXFullYBB = new BYTE [nHeightP*VPitchY];
VXFullUVBB = new BYTE [nHeightPUV*VPitchUV];
VYFullYBB = new BYTE [nHeightP*VPitchY];
VYFullUVBB = new BYTE [nHeightPUV*VPitchUV];
VXFullYFF = new BYTE [nHeightP*VPitchY];
VXFullUVFF = new BYTE [nHeightPUV*VPitchUV];
VYFullYFF = new BYTE [nHeightP*VPitchY];
VYFullUVFF = new BYTE [nHeightPUV*VPitchUV];
VXSmallYBB = new BYTE [nBlkXP*nBlkYP];
VYSmallYBB = new BYTE [nBlkXP*nBlkYP];
VXSmallUVBB = new BYTE [nBlkXP*nBlkYP];
VYSmallUVBB = new BYTE [nBlkXP*nBlkYP];
VXSmallYFF = new BYTE [nBlkXP*nBlkYP];
VYSmallYFF = new BYTE [nBlkXP*nBlkYP];
VXSmallUVFF = new BYTE [nBlkXP*nBlkYP];
VYSmallUVFF = new BYTE [nBlkXP*nBlkYP];
MaskSmallB = new BYTE [nBlkXP*nBlkYP];
MaskFullYB = new BYTE [nHeightP*VPitchY];
MaskFullUVB = new BYTE [nHeightPUV*VPitchUV];
MaskSmallF = new BYTE [nBlkXP*nBlkYP];
MaskFullYF = new BYTE [nHeightP*VPitchY];
MaskFullUVF = new BYTE [nHeightPUV*VPitchUV];
SADMaskSmallB = new BYTE [nBlkXP*nBlkYP];
SADMaskSmallF = new BYTE [nBlkXP*nBlkYP];
int pel2WidthY = (nWidth + 2*nHPadding)*nPel;
pel2HeightY = (nHeight + 2*nVPadding)*nPel;
int pel2WidthUV = (nWidthUV + 2*nHPaddingUV)*nPel;
pel2HeightUV = (nHeightUV + 2*nVPaddingUV)*nPel;
pel2PitchY = (pel2WidthY + 15) & (~15);
pel2PitchUV = (pel2WidthUV + 15) & (~15);
pel2OffsetY = pel2PitchY * nVPadding*nPel + nHPadding*nPel;
pel2OffsetUV = pel2PitchUV * nVPaddingUV*nPel + nHPaddingUV*nPel;
if (nPel>1) {
pel2PlaneYB = new BYTE [pel2PitchY*pel2HeightY];
pel2PlaneUB = new BYTE [pel2PitchUV*pel2HeightUV];
pel2PlaneVB = new BYTE [pel2PitchUV*pel2HeightUV];
pel2PlaneYF = new BYTE [pel2PitchY*pel2HeightY];
pel2PlaneUF = new BYTE [pel2PitchUV*pel2HeightUV];
pel2PlaneVF = new BYTE [pel2PitchUV*pel2HeightUV];
}
int CPUF_Resize = env->GetCPUFlags();
if (!isse) CPUF_Resize = (CPUF_Resize & !CPUF_INTEGER_SSE) & !CPUF_SSE2;
upsizer = new SimpleResize(nWidthP, nHeightP, nBlkXP, nBlkYP, CPUF_Resize);
upsizerUV = new SimpleResize(nWidthPUV, nHeightPUV, nBlkXP, nBlkYP, CPUF_Resize);
Create_LUTV(time256, LUTVB, LUTVF);
SrcPlanes = new YUY2Planes(nWidth, nHeight, vi.pixel_type, isse);
RefPlanes = new YUY2Planes(nWidth, nHeight, vi.pixel_type, isse);
DstPlanes = new YUY2Planes(nWidth, nHeight, vi.pixel_type, isse);
}
