OMXResult omxVCCOMM_Copy16x16(
const OMX_U8 *pSrc,
OMX_U8 *pDst,
OMX_INT step)
{
/* Definitions and Initializations*/
OMX_INT count,index, x, y;
/* Argument error checks */
armRetArgErrIf(pSrc == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(pDst == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(!armIs16ByteAligned(pSrc), OMX_Sts_BadArgErr);
armRetArgErrIf(!armIs16ByteAligned(pDst), OMX_Sts_BadArgErr);
armRetArgErrIf(((step < 16) || (step % 16)), OMX_Sts_BadArgErr);
/* Copying the ref 16x16 blk to the curr blk */
for (y = 0, count = 0, index = 0; y < 16; y++, count = count + step - 16)
{
for (x = 0; x < 16; x++, count++, index++)
{
pDst[index] = pSrc[count];
}
}
return OMX_Sts_NoErr;
}
OMXResult omxVCM4P10_DeblockLuma_I(
OMX_U8* pSrcDst,
OMX_S32 srcdstStep,
const OMX_U8* pAlpha,
const OMX_U8* pBeta,
const OMX_U8* pThresholds,
const OMX_U8 *pBS
)
{
OMXResult errorCode;
armRetArgErrIf(pSrcDst == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(armNot8ByteAligned(pSrcDst), OMX_Sts_BadArgErr);
armRetArgErrIf(srcdstStep & 7, OMX_Sts_BadArgErr);
armRetArgErrIf(pAlpha == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(pBeta == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(pThresholds == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(armNot4ByteAligned(pThresholds), OMX_Sts_BadArgErr);
armRetArgErrIf(pBS == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(armNot4ByteAligned(pBS), OMX_Sts_BadArgErr);
errorCode = omxVCM4P10_FilterDeblockingLuma_VerEdge_I(
pSrcDst, srcdstStep, pAlpha, pBeta, pThresholds, pBS);
armRetArgErrIf(errorCode != OMX_Sts_NoErr, errorCode)
errorCode = omxVCM4P10_FilterDeblockingLuma_HorEdge_I(
pSrcDst, srcdstStep, pAlpha+2, pBeta+2, pThresholds+16, pBS+16);
return errorCode;
}
OMXResult omxSP_FFTGetBufSize_R_S16S32(
OMX_INT order,
OMX_INT *pSize
)
{
OMX_INT Nby2;
OMX_INT N;
/* Input parameter check */
armRetArgErrIf(pSize == NULL, OMX_Sts_BadArgErr)
armRetArgErrIf(order < 0, OMX_Sts_BadArgErr)
armRetArgErrIf(order > 12, OMX_Sts_BadArgErr)
/* Check for order zero */
if (order == 0)
{
*pSize = sizeof(ARMsFFTSpec_FC64);
return OMX_Sts_NoErr;
}
Nby2 = 1 << (order - 1);
N = 1 << order;
/* 2 pointers to store bitreversed array and twiddle factor array */
*pSize = sizeof(ARMsFFTSpec_FC64)
/* N bitreversed Numbers */
+ sizeof(OMX_U16) * Nby2
/* Twiddle factors */
+ sizeof(OMX_FC64) * Nby2
+ sizeof(OMX_F64) * (2 + N);
return OMX_Sts_NoErr;
}
OMXResult omxVCM4P2_QuantInvInter_I(
OMX_S16 * pSrcDst,
OMX_INT QP
)
{
OMX_INT coeffCount, Sign;
/* Argument error checks */
armRetArgErrIf(pSrcDst == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(((QP <= 0) || (QP >= 32)), OMX_Sts_BadArgErr);
/* Second Inverse quantisation method */
for (coeffCount = 0; coeffCount < 64; coeffCount++)
{
/* check sign */
Sign = armSignCheck (pSrcDst[coeffCount]);
/* Quantize the coeff */
if (QP & 0x1)
{
pSrcDst[coeffCount] = (2* armAbs(pSrcDst[coeffCount]) + 1) * QP;
pSrcDst[coeffCount] *= Sign;
}
else
{
pSrcDst[coeffCount] = (2* armAbs(pSrcDst[coeffCount]) + 1)
* QP - 1;
pSrcDst[coeffCount] *= Sign;
}
/* Saturate */
pSrcDst[coeffCount] = armClip (-2048, 2047, pSrcDst[coeffCount]);
}
return OMX_Sts_NoErr;
}
OMXResult omxVCM4P2_DCT8x8blk (const OMX_S16 *pSrc, OMX_S16 *pDst)
{
OMX_INT x, y, u, v;
/* Argument error checks */
armRetArgErrIf(pSrc == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(!armIs16ByteAligned(pSrc), OMX_Sts_BadArgErr);
armRetArgErrIf(pDst == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(!armIs16ByteAligned(pDst), OMX_Sts_BadArgErr);
for (u = 0; u < 8; u++)
{
for (v = 0; v < 8; v++)
{
OMX_F64 sum = 0.0;
for (x = 0; x < 8; x++)
{
for (y = 0; y < 8; y++)
{
sum += pSrc[(x * 8) + y] *
armVCM4P2_preCalcDCTCos[x][u] *
armVCM4P2_preCalcDCTCos[y][v];
}
}
pDst[(u * 8) + v]= armRoundFloatToS16 (sum);
}
}
return OMX_Sts_NoErr;
}
OMXResult omxVCM4P10_MEInit(
OMXVCM4P10MEMode MEMode,
const OMXVCM4P10MEParams *pMEParams,
void *pMESpec
)
{
ARMVCM4P10_MESpec *armMESpec = (ARMVCM4P10_MESpec *) pMESpec;
armRetArgErrIf(!pMEParams, OMX_Sts_BadArgErr);
armRetArgErrIf(!pMESpec, OMX_Sts_BadArgErr);
armRetArgErrIf((MEMode != OMX_VC_M4P10_FAST_SEARCH) &&
(MEMode != OMX_VC_M4P10_FULL_SEARCH), OMX_Sts_BadArgErr);
armRetArgErrIf((pMEParams->searchRange16x16 <= 0) ||
(pMEParams->searchRange8x8 <= 0) ||
(pMEParams->searchRange4x4 <= 0), OMX_Sts_BadArgErr);
armMESpec->MEParams.blockSplitEnable8x8 = pMEParams->blockSplitEnable8x8;
armMESpec->MEParams.blockSplitEnable4x4 = pMEParams->blockSplitEnable4x4;
armMESpec->MEParams.halfSearchEnable = pMEParams->halfSearchEnable;
armMESpec->MEParams.quarterSearchEnable = pMEParams->quarterSearchEnable;
armMESpec->MEParams.intraEnable4x4 = pMEParams->intraEnable4x4;
armMESpec->MEParams.searchRange16x16 = pMEParams->searchRange16x16;
armMESpec->MEParams.searchRange8x8 = pMEParams->searchRange8x8;
armMESpec->MEParams.searchRange4x4 = pMEParams->searchRange4x4;
armMESpec->MEMode = MEMode;
return OMX_Sts_NoErr;
}
OMXResult omxVCM4P2_QuantInter_I(
OMX_S16 * pSrcDst,
OMX_U8 QP,
OMX_INT shortVideoHeader
)
{
/* Definitions and Initializations*/
OMX_INT coeffCount;
OMX_INT fSign;
OMX_INT maxClpAC = 0, minClpAC = 0;
OMX_INT maxClpDC = 0, minClpDC = 0;
/* Argument error checks */
armRetArgErrIf(pSrcDst == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(((QP <= 0) || (QP >= 32)), OMX_Sts_BadArgErr);
/* One argument check is delayed until we have ascertained that */
/* pQMatrix is not NULL. */
/* Set the Clip Range based on SVH on/off */
if(shortVideoHeader == 1)
{
maxClpDC = 254;
minClpDC = 1;
maxClpAC = 127;
minClpAC = -127;
}
else
{
maxClpDC = 2047;
minClpDC = -2047;
maxClpAC = 2047;
minClpAC = -2047;
}
/* Second Inverse quantisation method */
for (coeffCount = 0; coeffCount < 64; coeffCount++)
{
fSign = armSignCheck (pSrcDst[coeffCount]);
pSrcDst[coeffCount] = (armAbs(pSrcDst[coeffCount])
- (QP/2))/(2 * QP);
pSrcDst[coeffCount] *= fSign;
/* Clip */
if (coeffCount == 0)
{
pSrcDst[coeffCount] =
(OMX_S16) armClip (minClpDC, maxClpDC, pSrcDst[coeffCount]);
}
else
{
pSrcDst[coeffCount] =
(OMX_S16) armClip (minClpAC, maxClpAC, pSrcDst[coeffCount]);
}
}
return OMX_Sts_NoErr;
}
OMXResult omxVCM4P2_BlockMatch_Integer_16x16(
const OMX_U8 *pSrcRefBuf,
OMX_INT refWidth,
const OMXRect *pRefRect,
const OMX_U8 *pSrcCurrBuf,
const OMXVCM4P2Coordinate *pCurrPointPos,
const OMXVCMotionVector *pSrcPreMV,
const OMX_INT *pSrcPreSAD,
void *pMESpec,
OMXVCMotionVector *pDstMV,
OMX_INT *pDstSAD
)
{
OMX_U8 BlockSize = 16;
/* Argument error checks */
armRetArgErrIf(!armIs16ByteAligned(pSrcCurrBuf), OMX_Sts_BadArgErr);
return ( armVCM4P2_BlockMatch_Integer(
pSrcRefBuf,
refWidth,
pRefRect,
pSrcCurrBuf,
pCurrPointPos,
pSrcPreMV,
pSrcPreSAD,
pMESpec,
pDstMV,
pDstSAD,
BlockSize)
);
}
/**
* Function: omxVCM4P2_EncodeVLCZigzag_Inter (6.2.4.5.3)
*
* Description:
* Performs classical zigzag scanning and VLC encoding for one inter block.
*
* Input Arguments:
*
* ppBitStream - pointer to the pointer to the current byte in the bit
* stream
* pBitOffset - pointer to the bit position in the byte pointed by
* *ppBitStream. Valid within 0 to 7
* pQDctBlkCoef - pointer to the quantized DCT coefficient
* pattern - block pattern which is used to decide whether this block is
* encoded
* shortVideoHeader - binary flag indicating presence of
* short_video_header; escape modes 0-3 are used if
* shortVideoHeader==0, and escape mode 4 is used when
* shortVideoHeader==1.
*
* Output Arguments:
*
* ppBitStream - *ppBitStream is updated after the block is encoded so that
* it points to the current byte in the bit stream buffer.
* pBitOffset - *pBitOffset is updated so that it points to the current bit
* position in the byte pointed by *ppBitStream.
*
* Return Value:
*
* OMX_Sts_NoErr - no error
* OMX_Sts_BadArgErr - Bad arguments
* - At least one of the pointers: is NULL: ppBitStream, *ppBitStream,
* pBitOffset, pQDctBlkCoef
* - *pBitOffset < 0, or *pBitOffset >7.
*
*/
OMXResult omxVCM4P2_EncodeVLCZigzag_Inter(
OMX_U8 **ppBitStream,
OMX_INT * pBitOffset,
const OMX_S16 *pQDctBlkCoef,
OMX_U8 pattern,
OMX_INT shortVideoHeader
)
{
OMX_U8 start = 0;
const OMX_U8 *pZigzagTable = armVCM4P2_aClassicalZigzagScan;
/* Argument error checks */
armRetArgErrIf(ppBitStream == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(*ppBitStream == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(pBitOffset == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(pQDctBlkCoef == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf((*pBitOffset < 0) || (*pBitOffset >7), OMX_Sts_BadArgErr);
if (pattern)
{
armVCM4P2_PutVLCBits (
ppBitStream,
pBitOffset,
pQDctBlkCoef,
shortVideoHeader,
start,
26,
40,
10,
1,
armVCM4P2_InterL0RunIdx,
armVCM4P2_InterVlcL0,
armVCM4P2_InterL1RunIdx,
armVCM4P2_InterVlcL1,
armVCM4P2_InterL0LMAX,
armVCM4P2_InterL1LMAX,
armVCM4P2_InterL0RMAX,
armVCM4P2_InterL1RMAX,
pZigzagTable
);
} /* Pattern check ends*/
return OMX_Sts_NoErr;
}
/**
* Function: omxVCCOMM_ExpandFrame_I (6.1.3.2.1)
*
* Description:
* This function expands a reconstructed frame in-place. The unexpanded
* source frame should be stored in a plane buffer with sufficient space
* pre-allocated for edge expansion, and the input frame should be located in
* the plane buffer center. This function executes the pixel expansion by
* replicating source frame edge pixel intensities in the empty pixel
* locations (expansion region) between the source frame edge and the plane
* buffer edge. The width/height of the expansion regions on the
* horizontal/vertical edges is controlled by the parameter iExpandPels.
*
* Input Arguments:
*
* pSrcDstPlane - pointer to the top-left corner of the frame to be
* expanded; must be aligned on an 8-byte boundary.
* iFrameWidth - frame width; must be a multiple of 8.
* iFrameHeight -frame height; must be a multiple of 8.
* iExpandPels - number of pixels to be expanded in the horizontal and
* vertical directions; must be a multiple of 8.
* iPlaneStep - distance, in bytes, between the start of consecutive lines
* in the plane buffer; must be larger than or equal to
* (iFrameWidth + 2 * iExpandPels).
*
* Output Arguments:
*
* pSrcDstPlane -Pointer to the top-left corner of the frame (NOT the
* top-left corner of the plane); must be aligned on an 8-byte
* boundary.
*
* Return Value:
*
* OMX_Sts_NoErr - no error
* OMX_Sts_BadArgErr - bad arguments; returned under any of the following
* conditions:
* - pSrcDstPlane is NULL.
* - pSrcDstPlane is not aligned on an 8-byte boundary.
* - one of the following parameters is either equal to zero or is a
* non-multiple of 8: iFrameHeight, iFrameWidth, iPlaneStep, or
* iExpandPels.
* - iPlaneStep < (iFrameWidth + 2 * iExpandPels).
*
*/
OMXResult omxVCCOMM_ExpandFrame_I(
OMX_U8* pSrcDstPlane,
OMX_U32 iFrameWidth,
OMX_U32 iFrameHeight,
OMX_U32 iExpandPels,
OMX_U32 iPlaneStep
)
{
OMX_INT x, y;
OMX_U8* pLeft;
OMX_U8* pRight;
OMX_U8* pTop;
OMX_U8* pBottom;
/* check for argument error */
armRetArgErrIf(pSrcDstPlane == NULL, OMX_Sts_BadArgErr)
armRetArgErrIf(armNot8ByteAligned(pSrcDstPlane), OMX_Sts_BadArgErr)
armRetArgErrIf(iFrameWidth == 0 || iFrameWidth & 7, OMX_Sts_BadArgErr)
armRetArgErrIf(iFrameHeight== 0 || iFrameHeight & 7, OMX_Sts_BadArgErr)
armRetArgErrIf(iExpandPels == 0 || iExpandPels & 7, OMX_Sts_BadArgErr)
armRetArgErrIf(iPlaneStep == 0 || iPlaneStep & 7, OMX_Sts_BadArgErr)
armRetArgErrIf(iPlaneStep < (iFrameWidth + 2 * iExpandPels),
OMX_Sts_BadArgErr)
/* Top and Bottom */
pTop = pSrcDstPlane - (iExpandPels * iPlaneStep);
pBottom = pSrcDstPlane + (iFrameHeight * iPlaneStep);
for (y = 0; y < (OMX_INT)iExpandPels; y++)
{
for (x = 0; x < (OMX_INT)iFrameWidth; x++)
{
pTop [y * iPlaneStep + x] =
pSrcDstPlane [x];
pBottom [y * iPlaneStep + x] =
pSrcDstPlane [(iFrameHeight - 1) * iPlaneStep + x];
}
}
/* Left, Right and Corners */
pLeft = pSrcDstPlane - iExpandPels;
pRight = pSrcDstPlane + iFrameWidth;
for (y = -(OMX_INT)iExpandPels; y < (OMX_INT)(iFrameHeight + iExpandPels); y++)
{
for (x = 0; x < (OMX_INT)iExpandPels; x++)
{
pLeft [y * iPlaneStep + x] =
pSrcDstPlane [y * iPlaneStep + 0];
pRight [y * iPlaneStep + x] =
pSrcDstPlane [y * iPlaneStep + (iFrameWidth - 1)];
}
}
return OMX_Sts_NoErr;
}
OMXResult omxVCM4P2_BlockMatch_Half_8x8(
const OMX_U8 *pSrcRefBuf,
OMX_INT refWidth,
const OMXRect *pRefRect,
const OMX_U8 *pSrcCurrBuf,
const OMXVCM4P2Coordinate *pSearchPointRefPos,
OMX_INT rndVal,
OMXVCMotionVector *pSrcDstMV,
OMX_INT *pDstSAD
)
{
/* For a blocksize of 8x8 */
OMX_U8 BlockSize = 8;
/* Argument error checks */
armRetArgErrIf(pSrcRefBuf == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(pRefRect == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(pSrcCurrBuf == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(pSearchPointRefPos == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(pSrcDstMV == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(!armIs8ByteAligned(pSrcCurrBuf), OMX_Sts_BadArgErr);
return (armVCM4P2_BlockMatch_Half(
pSrcRefBuf,
refWidth,
pRefRect,
pSrcCurrBuf,
pSearchPointRefPos,
rndVal,
pSrcDstMV,
pDstSAD,
BlockSize));
}
OMXResult omxVCM4P2_MEInit(
OMXVCM4P2MEMode MEMode,
const OMXVCM4P2MEParams *pMEParams,
void *pMESpec
)
{
ARMVCM4P2_MESpec *armMESpec = (ARMVCM4P2_MESpec *) pMESpec;
armRetArgErrIf(!pMEParams, OMX_Sts_BadArgErr);
armRetArgErrIf(!pMESpec, OMX_Sts_BadArgErr);
armRetArgErrIf((MEMode != OMX_VC_M4P2_FAST_SEARCH) &&
(MEMode != OMX_VC_M4P2_FULL_SEARCH), OMX_Sts_BadArgErr);
armRetArgErrIf(pMEParams->searchRange <= 0, OMX_Sts_BadArgErr);
armMESpec->MEParams.searchEnable8x8 = pMEParams->searchEnable8x8;
armMESpec->MEParams.halfPelSearchEnable = pMEParams->halfPelSearchEnable;
armMESpec->MEParams.searchRange = pMEParams->searchRange;
armMESpec->MEParams.rndVal = pMEParams->rndVal;
armMESpec->MEMode = MEMode;
return OMX_Sts_NoErr;
}
OMXResult armVCM4P10_InterpolateHalfHor_Luma(
const OMX_U8* pSrc,
OMX_U32 iSrcStep,
OMX_U8* pDst,
OMX_U32 iDstStep,
OMX_U32 iWidth,
OMX_U32 iHeight
)
{
OMX_INT x, y;
OMX_S32 HalfCoeff, pos;
/* check for argument error */
armRetArgErrIf(pSrc == NULL, OMX_Sts_BadArgErr)
armRetArgErrIf(pDst == NULL, OMX_Sts_BadArgErr)
for (y = 0; y < iHeight; y++)
{
for (x = 0; x < iWidth; x++)
{
pos = y * iSrcStep + x;
HalfCoeff =
pSrc [pos - 2] -
5 * pSrc [pos - 1] +
20 * pSrc [pos] +
20 * pSrc [pos + 1] -
5 * pSrc [pos + 2] +
pSrc [pos + 3];
HalfCoeff = (HalfCoeff + 16) >> 5;
HalfCoeff = armClip(0, 255, HalfCoeff);
pDst [y * iDstStep + x] = HalfCoeff;
} /* x */
} /* y */
return OMX_Sts_NoErr;
}
/**
* Functions: armSwapElem
*
* Description:
* These function swaps two elements at the specified pointer locations.
* The size of each element could be anything as specified by <elemSize>
*
* Return Value:
* OMXResult -- Error status from the function
*/
OMXResult armSwapElem(
OMX_U8 *pBuf1,
OMX_U8 *pBuf2,
OMX_INT elemSize
)
{
OMX_INT i;
OMX_U8 temp;
armRetArgErrIf(!pBuf1 || !pBuf2, OMX_Sts_NullPtrErr);
for(i = 0; i < elemSize; i++)
{
temp = *(pBuf1 + i);
*(pBuf1 + i) = *(pBuf2 + i);
*(pBuf2 + i) = temp;
}
return OMX_Sts_NoErr;
}
OMXResult omxVCM4P10_DecodeChromaDcCoeffsToPairCAVLC (
const OMX_U8** ppBitStream,
OMX_S32* pOffset,
OMX_U8* pNumCoeff,
OMX_U8** ppPosCoefbuf
)
{
armRetArgErrIf(ppBitStream==NULL , OMX_Sts_BadArgErr);
armRetArgErrIf(*ppBitStream==NULL , OMX_Sts_BadArgErr);
armRetArgErrIf(pOffset==NULL , OMX_Sts_BadArgErr);
armRetArgErrIf(*pOffset<0 , OMX_Sts_BadArgErr);
armRetArgErrIf(*pOffset>7 , OMX_Sts_BadArgErr);
armRetArgErrIf(pNumCoeff==NULL , OMX_Sts_BadArgErr);
armRetArgErrIf(ppPosCoefbuf==NULL , OMX_Sts_BadArgErr);
armRetArgErrIf(*ppPosCoefbuf==NULL , OMX_Sts_BadArgErr);
return armVCM4P10_DecodeCoeffsToPair(ppBitStream, pOffset, pNumCoeff,
ppPosCoefbuf, 4, 4);
}
OMXResult omxVCM4P2_TransRecBlockCoef_inter(
const OMX_S16 *pSrc,
OMX_S16 * pDst,
OMX_S16 * pRec,
OMX_U8 QP,
OMX_INT shortVideoHeader
)
{
/* 64 elements are needed but to align it to 16 bytes need
8 more elements of padding */
OMX_S16 tempBuffer[72];
OMX_S16 *pTempBuffer;
OMX_INT i;
/* Aligning the local buffers */
pTempBuffer = armAlignTo16Bytes(tempBuffer);
/* Argument error checks */
armRetArgErrIf(pSrc == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(pRec == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(pDst == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(!armIs16ByteAligned(pSrc), OMX_Sts_BadArgErr);
armRetArgErrIf(!armIs16ByteAligned(pRec), OMX_Sts_BadArgErr);
armRetArgErrIf(!armIs16ByteAligned(pDst), OMX_Sts_BadArgErr);
armRetArgErrIf(((QP <= 0) || (QP >= 32)), OMX_Sts_BadArgErr);
omxVCM4P2_DCT8x8blk (pSrc, pDst);
omxVCM4P2_QuantInter_I(
pDst,
QP,
shortVideoHeader);
for (i = 0; i < 64; i++)
{
pTempBuffer[i] = pDst[i];
}
omxVCM4P2_QuantInvInter_I(
pTempBuffer,
QP);
omxVCM4P2_IDCT8x8blk (pTempBuffer, pRec);
return OMX_Sts_NoErr;
}
OMXResult omxSP_FIROne_Direct_S16(
OMX_S16 val,
OMX_S16 * pResult,
const OMX_S16 * pTapsQ15,
OMX_INT tapsLen,
OMX_S16 * pDelayLine,
OMX_INT * pDelayLineIndex
)
{
OMX_U32 index;
OMX_S32 accum;
OMX_S16 *pDelayCurrent;
/* Input parameter check */
armRetArgErrIf((pResult == NULL), OMX_Sts_BadArgErr)
armRetArgErrIf((pTapsQ15 == NULL), OMX_Sts_BadArgErr)
armRetArgErrIf((tapsLen <= 0), OMX_Sts_BadArgErr)
armRetArgErrIf((pDelayLine == NULL), OMX_Sts_BadArgErr)
armRetArgErrIf((pDelayLineIndex == NULL), OMX_Sts_BadArgErr)
armRetArgErrIf((*pDelayLineIndex < 0), OMX_Sts_BadArgErr)
armRetArgErrIf((*pDelayLineIndex >= tapsLen), OMX_Sts_BadArgErr)
/* Update the delay state */
pDelayCurrent = &pDelayLine [*pDelayLineIndex];
/* Copy input to current delay line position */
pDelayCurrent [0] = pDelayCurrent [tapsLen] = val;
accum = 0;
for (index = 0; index < tapsLen; index++)
{
accum += (OMX_S32)pTapsQ15 [index] *
(OMX_S32)pDelayCurrent [index];
}
if (--(*pDelayLineIndex) < 0)
{
*pDelayLineIndex = tapsLen - 1;
}
/* Store the result */
*pResult = armSatRoundLeftShift_S32(accum, -15);
return OMX_Sts_NoErr;
}
OMXResult omxSP_IIROne_BiQuadDirect_S16_I(
OMX_S16 * pValResult,
const OMX_S16 * pTaps,
OMX_INT numBiquad,
OMX_S32 * pDelayLine
)
{
OMXResult errorCode;
armRetArgErrIf(pValResult == NULL, OMX_Sts_BadArgErr)
errorCode = omxSP_IIROne_BiQuadDirect_S16(
*pValResult,
pValResult,
pTaps,
numBiquad,
pDelayLine);
return errorCode;
}
OMXResult omxVCM4P10_TransformDequantChromaDCFromPair(
const OMX_U8 **ppSrc,
OMX_S16* pDst,
OMX_INT QP
)
{
armRetArgErrIf(ppSrc == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(*ppSrc == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(pDst == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(armNot4ByteAligned(pDst), OMX_Sts_BadArgErr);
armRetArgErrIf(QP<0, OMX_Sts_BadArgErr);
armRetArgErrIf(QP>51, OMX_Sts_BadArgErr);
armVCM4P10_UnpackBlock2x2(ppSrc, pDst);
InvTransformDC2x2(pDst);
DequantChromaDC2x2(pDst, QP);
return OMX_Sts_NoErr;
}
OMXResult armVCM4P2_DecodeVLCZigzag_Intra(
const OMX_U8 ** ppBitStream,
OMX_INT * pBitOffset,
OMX_S16 * pDst,
OMX_U8 predDir,
OMX_INT shortVideoHeader,
OMX_U8 start
)
{
OMX_U8 last = 0;
const OMX_U8 *pZigzagTable = armVCM4P2_aClassicalZigzagScan;
OMXResult errorCode;
/* Argument error checks */
armRetArgErrIf(ppBitStream == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(*ppBitStream == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(pBitOffset == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(pDst == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(!armIs4ByteAligned(pDst), OMX_Sts_BadArgErr);
armRetArgErrIf((*pBitOffset < 0) || (*pBitOffset >7), OMX_Sts_BadArgErr);
armRetArgErrIf((predDir > 2), OMX_Sts_BadArgErr);
switch (predDir)
{
case OMX_VC_NONE:
{
pZigzagTable = armVCM4P2_aClassicalZigzagScan;
break;
}
case OMX_VC_HORIZONTAL:
{
pZigzagTable = armVCM4P2_aVerticalZigzagScan;
break;
}
case OMX_VC_VERTICAL:
{
pZigzagTable = armVCM4P2_aHorizontalZigzagScan;
break;
}
}
errorCode = armVCM4P2_GetVLCBits (
ppBitStream,
pBitOffset,
pDst,
shortVideoHeader,
start,
&last,
10,
62,
7,
21,
armVCM4P2_IntraL0RunIdx,
armVCM4P2_IntraVlcL0,
armVCM4P2_IntraL1RunIdx,
armVCM4P2_IntraVlcL1,
armVCM4P2_IntraL0LMAX,
armVCM4P2_IntraL1LMAX,
armVCM4P2_IntraL0RMAX,
armVCM4P2_IntraL1RMAX,
pZigzagTable );
armRetDataErrIf((errorCode != OMX_Sts_NoErr), errorCode);
if (last == 0)
{
return OMX_Sts_Err;
}
return OMX_Sts_NoErr;
}
OMXResult omxVCM4P10_DequantTransformResidualFromPairAndAdd(
const OMX_U8 **ppSrc,
const OMX_U8 *pPred,
const OMX_S16 *pDC,
OMX_U8 *pDst,
OMX_INT predStep,
OMX_INT dstStep,
OMX_INT QP,
OMX_INT AC
)
{
OMX_S16 pBuffer[16+4];
OMX_S16 *pDelta;
int i,x,y;
armRetArgErrIf(pPred == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(armNot4ByteAligned(pPred),OMX_Sts_BadArgErr);
armRetArgErrIf(pDst == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(armNot4ByteAligned(pDst), OMX_Sts_BadArgErr);
armRetArgErrIf(predStep & 3, OMX_Sts_BadArgErr);
armRetArgErrIf(dstStep & 3, OMX_Sts_BadArgErr);
armRetArgErrIf(AC!=0 && (QP<0), OMX_Sts_BadArgErr);
armRetArgErrIf(AC!=0 && (QP>51), OMX_Sts_BadArgErr);
armRetArgErrIf(AC!=0 && ppSrc==NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(AC!=0 && *ppSrc==NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(AC==0 && pDC==NULL, OMX_Sts_BadArgErr);
pDelta = armAlignTo8Bytes(pBuffer);
for (i=0; i<16; i++)
{
pDelta[i] = 0;
}
if (AC)
{
armVCM4P10_UnpackBlock4x4(ppSrc, pDelta);
DequantLumaAC4x4(pDelta, QP);
}
if (pDC)
{
pDelta[0] = pDC[0];
}
armVCM4P10_TransformResidual4x4(pDelta,pDelta);
for (y=0; y<4; y++)
{
for (x=0; x<4; x++)
{
pDst[y*dstStep+x] = (OMX_U8)armClip(0,255,pPred[y*predStep+x] + pDelta[4*y+x]);
}
}
return OMX_Sts_NoErr;
}
/**
* Function: omxVCM4P10_InvTransformDequant_LumaDC (6.3.5.6.3)
*
* Description:
* This function performs inverse 4x4 Hadamard transform and then dequantizes
* the coefficients.
*
* Input Arguments:
*
* pSrc - Pointer to the 4x4 array of the 4x4 Hadamard-transformed and
* quantized coefficients. 16 byte alignment required.
* iQP - Quantization parameter; must be in the range [0,51].
*
* Output Arguments:
*
* pDst - Pointer to inverse-transformed and dequantized coefficients.
* 16-byte alignment required.
*
* Return Value:
*
* OMX_Sts_NoErr - no error
* OMX_Sts_BadArgErr - bad arguments; returned if any of the following
* conditions are true:
* - at least one of the following pointers is NULL: pSrc
* - pSrc or pDst is not aligned on a 16-byte boundary
*
*/
OMXResult omxVCM4P10_InvTransformDequant_LumaDC(
const OMX_S16* pSrc,
OMX_S16* pDst,
OMX_U32 iQP
)
{
OMX_INT i, j;
OMX_S32 m1[4][4], m2[4][4], Value;
OMX_S32 QPer, V;
/* check for argument error */
armRetArgErrIf(pSrc == NULL, OMX_Sts_BadArgErr)
armRetArgErrIf(pDst == NULL, OMX_Sts_BadArgErr)
armRetArgErrIf(iQP > 51, OMX_Sts_BadArgErr)
armRetArgErrIf(armNot16ByteAligned(pSrc), OMX_Sts_BadArgErr)
armRetArgErrIf(armNot16ByteAligned(pDst), OMX_Sts_BadArgErr)
/* Inv Hadamard Transform for DC Luma 4x4 block */
/* Horizontal */
for (i = 0; i < 4; i++)
{
j = i * 4;
m1[i][0] = pSrc[j + 0] + pSrc[j + 2]; /* a+c */
m1[i][1] = pSrc[j + 1] + pSrc[j + 3]; /* b+d */
m1[i][2] = pSrc[j + 0] - pSrc[j + 2]; /* a-c */
m1[i][3] = pSrc[j + 1] - pSrc[j + 3]; /* b-d */
m2[i][0] = m1[i][0] + m1[i][1]; /* a+b+c+d */
m2[i][1] = m1[i][2] + m1[i][3]; /* a+b-c-d */
m2[i][2] = m1[i][2] - m1[i][3]; /* a-b-c+d */
m2[i][3] = m1[i][0] - m1[i][1]; /* a-b+c-d */
}
/* Vertical */
for (i = 0; i < 4; i++)
{
m1[0][i] = m2[0][i] + m2[2][i];
m1[1][i] = m2[1][i] + m2[3][i];
m1[2][i] = m2[0][i] - m2[2][i];
m1[3][i] = m2[1][i] - m2[3][i];
m2[0][i] = m1[0][i] + m1[1][i];
m2[1][i] = m1[2][i] + m1[3][i];
m2[2][i] = m1[2][i] - m1[3][i];
m2[3][i] = m1[0][i] - m1[1][i];
}
/* Scaling */
QPer = iQP / 6;
V = armVCM4P10_VMatrix [iQP % 6][0];
for (j = 0; j < 4; j++)
{
for (i = 0; i < 4; i++)
{
if (QPer < 2)
{
Value = (m2[j][i] * V + (1 << (1 - QPer))) >> (2 - QPer);
}
else
{
Value = m2[j][i] * V * (1 << (QPer - 2));
}
pDst[j * 4 + i] = (OMX_S16) Value;
}
/**
* Function: omxVCM4P10_GetVLCInfo (6.3.5.9.1)
*
* Description:
* This function extracts run-length encoding (RLE) information from the
* coefficient matrix. The results are returned in an OMXVCM4P10VLCInfo
* structure.
*
* Input Arguments:
*
* pSrcCoeff - pointer to the transform coefficient matrix. 8-byte
* alignment required.
* pScanMatrix - pointer to the scan order definition matrix. For a luma
* block the scan matrix should follow [ISO14496-10] section 8.5.4,
* and should contain the values 0, 1, 4, 8, 5, 2, 3, 6, 9, 12, 13,
* 10, 7, 11, 14, 15. For a chroma block, the scan matrix should
* contain the values 0, 1, 2, 3.
* bAC - indicates presence of a DC coefficient; 0 = DC coefficient
* present, 1= DC coefficient absent.
* MaxNumCoef - specifies the number of coefficients contained in the
* transform coefficient matrix, pSrcCoeff. The value should be 16
* for blocks of type LUMADC, LUMAAC, LUMALEVEL, and CHROMAAC. The
* value should be 4 for blocks of type CHROMADC.
*
* Output Arguments:
*
* pDstVLCInfo - pointer to structure that stores information for
* run-length coding.
*
* Return Value:
*
* OMX_Sts_NoErr - no error
* OMX_Sts_BadArgErr - bad arguments; returned if any of the following
* conditions are true:
* - at least one of the following pointers is NULL:
* pSrcCoeff, pScanMatrix, pDstVLCInfo
* - pSrcCoeff is not aligned on an 8-byte boundary
*
*/
OMXResult omxVCM4P10_GetVLCInfo (
const OMX_S16* pSrcCoeff,
const OMX_U8* pScanMatrix,
OMX_U8 bAC,
OMX_U32 MaxNumCoef,
OMXVCM4P10VLCInfo* pDstVLCInfo
)
{
OMX_INT i, MinIndex;
OMX_S32 Value;
OMX_U32 Mask = 4, RunBefore;
OMX_S16 *pLevel;
OMX_U8 *pRun;
OMX_S16 Buf [16];
/* check for argument error */
armRetArgErrIf(pSrcCoeff == NULL, OMX_Sts_BadArgErr)
armRetArgErrIf(armNot8ByteAligned(pSrcCoeff), OMX_Sts_BadArgErr)
armRetArgErrIf(pScanMatrix == NULL, OMX_Sts_BadArgErr)
armRetArgErrIf(pDstVLCInfo == NULL, OMX_Sts_BadArgErr)
armRetArgErrIf(bAC > 1, OMX_Sts_BadArgErr)
armRetArgErrIf(MaxNumCoef > 16, OMX_Sts_BadArgErr)
/* Initialize RLE Info structure */
pDstVLCInfo->uTrailing_Ones = 0;
pDstVLCInfo->uTrailing_One_Signs = 0;
pDstVLCInfo->uNumCoeffs = 0;
pDstVLCInfo->uTotalZeros = 0;
for (i = 0; i < 16; i++)
{
pDstVLCInfo->iLevels [i] = 0;
pDstVLCInfo->uRuns [i] = 0;
}
MinIndex = (bAC == 0 && MaxNumCoef == 15) ? 1 : 0;
for (i = MinIndex; i < (MaxNumCoef + MinIndex); i++)
{
/* Scan */
Buf [i - MinIndex] = pSrcCoeff [pScanMatrix [i]];
}
/* skip zeros at the end */
i = MaxNumCoef - 1;
while (!Buf [i] && i >= 0)
{
i--;
}
if (i < 0)
{
return OMX_Sts_NoErr;
}
/* Fill RLE Info structure */
pLevel = pDstVLCInfo->iLevels;
pRun = pDstVLCInfo->uRuns;
RunBefore = 0;
/* Handle first non zero separate */
pDstVLCInfo->uNumCoeffs++;
Value = Buf [i];
if (Value == 1 || Value == -1)
{
pDstVLCInfo->uTrailing_Ones++;
pDstVLCInfo->uTrailing_One_Signs |=
Value == -1 ? Mask : 0;
Mask >>= 1;
}
OMXResult omxVCM4P2_QuantIntra_I(
OMX_S16 * pSrcDst,
OMX_U8 QP,
OMX_INT blockIndex,
OMX_INT shortVideoHeader
)
{
/* Definitions and Initializations*/
/* Initialized to remove compilation error */
OMX_INT dcScaler = 0, coeffCount,fSign;
OMX_INT maxClpAC, minClpAC;
/* Argument error checks */
armRetArgErrIf(pSrcDst == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(((blockIndex < 0) || (blockIndex >= 10)), OMX_Sts_BadArgErr);
armRetArgErrIf(((QP <= 0) || (QP >= 32)), OMX_Sts_BadArgErr);
/* One argument check is delayed until we have ascertained that */
/* pQMatrix is not NULL. */
/* Set the Clip Range based on SVH on/off */
if(shortVideoHeader == 1)
{
maxClpAC = 127;
minClpAC = -127;
dcScaler = 8;
/* Dequant the DC value, this applies to both the methods */
pSrcDst[0] = armIntDivAwayFromZero (pSrcDst[0], dcScaler);
/* Clip between 1 and 254 */
pSrcDst[0] = (OMX_S16) armClip (1, 254, pSrcDst[0]);
}
else
{
maxClpAC = 2047;
minClpAC = -2047;
/* Calculate the DC scaler value */
if ((blockIndex < 4) || (blockIndex > 5))
{
if (QP >= 1 && QP <= 4)
{
dcScaler = 8;
}
else if (QP >= 5 && QP <= 8)
{
dcScaler = 2 * QP;
}
else if (QP >= 9 && QP <= 24)
{
dcScaler = QP + 8;
}
else
{
dcScaler = (2 * QP) - 16;
}
}
else if (blockIndex < 6)
{
if (QP >= 1 && QP <= 4)
{
dcScaler = 8;
}
else if (QP >= 5 && QP <= 24)
{
dcScaler = (QP + 13)/2;
}
else
{
dcScaler = QP - 6;
}
}
/* Dequant the DC value, this applies to both the methods */
pSrcDst[0] = armIntDivAwayFromZero (pSrcDst[0], dcScaler);
}
/* Second Inverse quantisation method */
for (coeffCount = 1; coeffCount < 64; coeffCount++)
{
fSign = armSignCheck (pSrcDst[coeffCount]);
pSrcDst[coeffCount] = armAbs(pSrcDst[coeffCount])/(2 * QP);
pSrcDst[coeffCount] *= fSign;
/* Clip */
pSrcDst[coeffCount] =
(OMX_S16) armClip (minClpAC, maxClpAC, pSrcDst[coeffCount]);
}
return OMX_Sts_NoErr;
}
OMXResult omxSP_FFTFwd_CToC_SC16_Sfs(
const OMX_SC16 *pSrc,
OMX_SC16 *pDst,
const OMXFFTSpec_C_SC16 *pFFTSpec,
OMX_INT scaleFactor
)
{
OMX_INT block, point;
OMX_INT i, j, N, NBy2;
OMX_U16 *pRevIndex;
OMX_FC64 *out;
OMX_FC64 *pT1, *pT2, *pT, *pTw, T;
ARMsFFTSpec_FC64 *pFFTStruct;
/* Input parameter check */
armRetArgErrIf(pSrc == NULL, OMX_Sts_BadArgErr)
armRetArgErrIf(armNot32ByteAligned(pSrc), OMX_Sts_BadArgErr)
armRetArgErrIf(pDst == NULL, OMX_Sts_BadArgErr)
armRetArgErrIf(armNot32ByteAligned(pDst), OMX_Sts_BadArgErr)
armRetArgErrIf(pFFTSpec == NULL, OMX_Sts_BadArgErr)
armRetArgErrIf(scaleFactor < 0, OMX_Sts_BadArgErr)
armRetArgErrIf(scaleFactor > 16, OMX_Sts_BadArgErr)
/* Order range check */
pFFTStruct = (ARMsFFTSpec_FC64 *) pFFTSpec;
N = pFFTStruct->N;
armRetArgErrIf(N < 1, OMX_Sts_BadArgErr)
armRetArgErrIf(N > (1 << 12), OMX_Sts_BadArgErr)
/* Handle order zero case separate */
if (N == 1)
{
pDst [0].Re = armSatRoundRightShift_S32_S16 (pSrc[0].Re, scaleFactor);
pDst [0].Im = armSatRoundRightShift_S32_S16 (pSrc[0].Im, scaleFactor);
return OMX_Sts_NoErr;
}
/* Do fft in float */
out = pFFTStruct->pBuf;
/* bit reversal */
pRevIndex = pFFTStruct->pBitRev;
for (i = 0; i < N; i++)
{
out [pRevIndex [i]].Re = (OMX_F64) pSrc [i]. Re;
out [pRevIndex [i]].Im = (OMX_F64) pSrc [i]. Im;
}
NBy2 = N >> 1;
pT = &T;
point = 2;
for (block = NBy2; block > 0; block >>= 1)
{
pTw = pFFTStruct->pTwiddle;
for (i = 0; i < point / 2; i++)
{
pT1 = out + i;
pT2 = pT1 + (point / 2);
for (j = 0; j < block; j++)
{
armSP_CPLX_MUL (pT, pTw, pT2);
armSP_CPLX_SUB (pT2, pT1, pT);
armSP_CPLX_ADD (pT1, pT1, pT);
pT1 += point;
pT2 += point;
}
pTw += block;
}
point <<= 1;
}
/* revert back from float */
for (i = 0; i < N; i++)
{
out [i].Re /= (1 << scaleFactor);
out [i].Im /= (1 << scaleFactor);
pDst [i]. Re = armSatRoundFloatToS16 (out [i].Re);
pDst [i]. Im = armSatRoundFloatToS16 (out [i].Im);
}
return OMX_Sts_NoErr;
}
OMXResult omxVCM4P2_MotionEstimationMB (
const OMX_U8 *pSrcCurrBuf,
OMX_S32 srcCurrStep,
const OMX_U8 *pSrcRefBuf,
OMX_S32 srcRefStep,
const OMXRect*pRefRect,
const OMXVCM4P2Coordinate *pCurrPointPos,
void *pMESpec,
const OMXVCM4P2MBInfoPtr *pMBInfo,
OMXVCM4P2MBInfo *pSrcDstMBCurr,
OMX_U16 *pDstSAD,
OMX_U16 *pDstBlockSAD
)
{
OMX_INT intraSAD, average, count, index, x, y;
OMXVCMotionVector dstMV16x16;
OMX_INT dstSAD16x16;
OMX_INT dstSAD8x8;
OMXVCM4P2MEParams *pMEParams;
OMXVCM4P2Coordinate TempCurrPointPos;
OMXVCM4P2Coordinate *pTempCurrPointPos;
OMX_U8 aTempSrcCurrBuf[271];
OMX_U8 *pTempSrcCurrBuf;
OMX_U8 *pDst;
OMX_U8 aDst[71];
OMX_S32 dstStep = 8;
OMX_INT predictType;
OMX_S32 Sad;
const OMX_U8 *pTempSrcRefBuf;
OMXVCMotionVector* pSrcCandMV1[4];
OMXVCMotionVector* pSrcCandMV2[4];
OMXVCMotionVector* pSrcCandMV3[4];
/* Argument error checks */
armRetArgErrIf(!armIs16ByteAligned(pSrcCurrBuf), OMX_Sts_BadArgErr);
armRetArgErrIf(!armIs16ByteAligned(pSrcRefBuf), OMX_Sts_BadArgErr);
armRetArgErrIf(((srcCurrStep % 16) || (srcRefStep % 16)), OMX_Sts_BadArgErr);
armRetArgErrIf(pSrcCurrBuf == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(pSrcRefBuf == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(pRefRect == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(pCurrPointPos == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(pSrcDstMBCurr == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(pDstSAD == NULL, OMX_Sts_BadArgErr);
pTempCurrPointPos = &(TempCurrPointPos);
pTempSrcCurrBuf = armAlignTo16Bytes(aTempSrcCurrBuf);
pMEParams = (OMXVCM4P2MEParams *)pMESpec;
pTempCurrPointPos->x = pCurrPointPos->x;
pTempCurrPointPos->y = pCurrPointPos->y;
pSrcDstMBCurr->mbType = OMX_VC_INTER;
/* Preparing a linear buffer for block match */
for (y = 0, index = count = 0; y < 16; y++, index += srcCurrStep - 16)
{
for(x = 0; x < 16; x++, count++, index++)
{
pTempSrcCurrBuf[count] = pSrcCurrBuf[index];
}
}
for(y = 0, index = 0; y < 2; y++)
{
for(x = 0; x < 2; x++,index++)
{
if((pMBInfo[0] != NULL) && (pMBInfo[0]->mbType != OMX_VC_INTRA))
{
pSrcCandMV1[index] = &(pMBInfo[0]->pMV0[y][x]);
}
else
{
pSrcCandMV1[index] = NULL;
}
if((pMBInfo[1] != NULL) && (pMBInfo[1]->mbType != OMX_VC_INTRA))
{
pSrcCandMV2[index] = &(pMBInfo[1]->pMV0[y][x]);
}
else
{
pSrcCandMV2[index] = NULL;
}
if((pMBInfo[3] != NULL) && (pMBInfo[3]->mbType != OMX_VC_INTRA))
{
pSrcCandMV3[index] = &(pMBInfo[3]->pMV0[y][x]);
}
else
{
pSrcCandMV3[index] = NULL;
}
}
}
/* Calculating SAD at MV(0,0) */
armVCCOMM_SAD(pTempSrcCurrBuf,
16,
pSrcRefBuf,
srcRefStep,
&Sad,
16,
16);
*pDstSAD = Sad;
/* Mode decision for NOT_CODED MB */
if(*pDstSAD == 0)
{
pSrcDstMBCurr->pMV0[0][0].dx = 0;
pSrcDstMBCurr->pMV0[0][0].dy = 0;
*pDstSAD = 0;
return OMX_Sts_NoErr;
}
omxVCM4P2_FindMVpred(
&(pSrcDstMBCurr->pMV0[0][0]),
pSrcCandMV1[0],
pSrcCandMV2[0],
pSrcCandMV3[0],
&(pSrcDstMBCurr->pMVPred[0][0]),
NULL,
0);
/* Inter 1 MV */
armVCM4P2_BlockMatch_16x16(
pSrcRefBuf,
srcRefStep,
pRefRect,
pTempSrcCurrBuf,
pCurrPointPos,
&(pSrcDstMBCurr->pMVPred[0][0]),
NULL,
pMEParams,
&dstMV16x16,
&dstSAD16x16);
/* Initialize all with 1 MV values */
pSrcDstMBCurr->pMV0[0][0].dx = dstMV16x16.dx;
pSrcDstMBCurr->pMV0[0][0].dy = dstMV16x16.dy;
pSrcDstMBCurr->pMV0[0][1].dx = dstMV16x16.dx;
pSrcDstMBCurr->pMV0[0][1].dy = dstMV16x16.dy;
pSrcDstMBCurr->pMV0[1][0].dx = dstMV16x16.dx;
pSrcDstMBCurr->pMV0[1][0].dy = dstMV16x16.dy;
pSrcDstMBCurr->pMV0[1][1].dx = dstMV16x16.dx;
pSrcDstMBCurr->pMV0[1][1].dy = dstMV16x16.dy;
*pDstSAD = dstSAD16x16;
if (pMEParams->searchEnable8x8)
{
/* Inter 4MV */
armVCM4P2_BlockMatch_8x8 (pSrcRefBuf,
srcRefStep, pRefRect,
pTempSrcCurrBuf, pTempCurrPointPos,
&(pSrcDstMBCurr->pMVPred[0][0]), NULL,
pMEParams, &(pSrcDstMBCurr->pMV0[0][0]),
&dstSAD8x8
);
pDstBlockSAD[0] = dstSAD8x8;
*pDstSAD = dstSAD8x8;
pTempCurrPointPos->x += 8;
pSrcRefBuf += 8;
omxVCM4P2_FindMVpred(
&(pSrcDstMBCurr->pMV0[0][1]),
pSrcCandMV1[1],
pSrcCandMV2[1],
pSrcCandMV3[1],
&(pSrcDstMBCurr->pMVPred[0][1]),
NULL,
1);
armVCM4P2_BlockMatch_8x8 (pSrcRefBuf,
srcRefStep, pRefRect,
pTempSrcCurrBuf, pTempCurrPointPos,
&(pSrcDstMBCurr->pMVPred[0][1]), NULL,
pMEParams, &(pSrcDstMBCurr->pMV0[0][1]),
&dstSAD8x8
);
pDstBlockSAD[1] = dstSAD8x8;
*pDstSAD += dstSAD8x8;
pTempCurrPointPos->x -= 8;
pTempCurrPointPos->y += 8;
pSrcRefBuf += (srcRefStep * 8) - 8;
omxVCM4P2_FindMVpred(
&(pSrcDstMBCurr->pMV0[1][0]),
pSrcCandMV1[2],
pSrcCandMV2[2],
pSrcCandMV3[2],
&(pSrcDstMBCurr->pMVPred[1][0]),
NULL,
2);
armVCM4P2_BlockMatch_8x8 (pSrcRefBuf,
srcRefStep, pRefRect,
pTempSrcCurrBuf, pTempCurrPointPos,
&(pSrcDstMBCurr->pMVPred[1][0]), NULL,
pMEParams, &(pSrcDstMBCurr->pMV0[1][0]),
&dstSAD8x8
);
pDstBlockSAD[2] = dstSAD8x8;
*pDstSAD += dstSAD8x8;
pTempCurrPointPos->x += 8;
pSrcRefBuf += 8;
omxVCM4P2_FindMVpred(
&(pSrcDstMBCurr->pMV0[1][1]),
pSrcCandMV1[3],
pSrcCandMV2[3],
pSrcCandMV3[3],
&(pSrcDstMBCurr->pMVPred[1][1]),
NULL,
3);
armVCM4P2_BlockMatch_8x8 (pSrcRefBuf,
srcRefStep, pRefRect,
pTempSrcCurrBuf, pTempCurrPointPos,
&(pSrcDstMBCurr->pMVPred[1][1]), NULL,
pMEParams, &(pSrcDstMBCurr->pMV0[1][1]),
&dstSAD8x8
);
pDstBlockSAD[3] = dstSAD8x8;
*pDstSAD += dstSAD8x8;
/* Checking if 4MV is equal to 1MV */
if (
(pSrcDstMBCurr->pMV0[0][0].dx != dstMV16x16.dx) ||
(pSrcDstMBCurr->pMV0[0][0].dy != dstMV16x16.dy) ||
(pSrcDstMBCurr->pMV0[0][1].dx != dstMV16x16.dx) ||
(pSrcDstMBCurr->pMV0[0][1].dy != dstMV16x16.dy) ||
(pSrcDstMBCurr->pMV0[1][0].dx != dstMV16x16.dx) ||
(pSrcDstMBCurr->pMV0[1][0].dy != dstMV16x16.dy) ||
(pSrcDstMBCurr->pMV0[1][1].dx != dstMV16x16.dx) ||
(pSrcDstMBCurr->pMV0[1][1].dy != dstMV16x16.dy)
)
{
/* select the 4 MV */
pSrcDstMBCurr->mbType = OMX_VC_INTER4V;
}
}
/* finding the error in intra mode */
for (count = 0, average = 0; count < 256 ; count++)
{
average = average + pTempSrcCurrBuf[count];
}
average = average/256;
intraSAD = 0;
/* Intra SAD calculation */
for (count = 0; count < 256 ; count++)
{
intraSAD += armAbs ((pTempSrcCurrBuf[count]) - (average));
}
/* Using the MPEG4 VM formula for intra/inter mode decision
Var < (SAD - 2*NB) where NB = N^2 is the number of pixels
of the macroblock.*/
if (intraSAD <= (*pDstSAD - 512))
{
pSrcDstMBCurr->mbType = OMX_VC_INTRA;
pSrcDstMBCurr->pMV0[0][0].dx = 0;
pSrcDstMBCurr->pMV0[0][0].dy = 0;
*pDstSAD = intraSAD;
pDstBlockSAD[0] = 0xFFFF;
pDstBlockSAD[1] = 0xFFFF;
pDstBlockSAD[2] = 0xFFFF;
pDstBlockSAD[3] = 0xFFFF;
}
if(pSrcDstMBCurr->mbType == OMX_VC_INTER)
{
pTempSrcRefBuf = pSrcRefBuf + (srcRefStep * dstMV16x16.dy) + dstMV16x16.dx;
if((dstMV16x16.dx & 0x1) && (dstMV16x16.dy & 0x1))
{
predictType = OMX_VC_HALF_PIXEL_XY;
}
else if(dstMV16x16.dx & 0x1)
{
predictType = OMX_VC_HALF_PIXEL_X;
}
else if(dstMV16x16.dy & 0x1)
{
predictType = OMX_VC_HALF_PIXEL_Y;
}
else
{
predictType = OMX_VC_INTEGER_PIXEL;
}
pDst = armAlignTo8Bytes(&(aDst[0]));
/* Calculating Block SAD at MV(dstMV16x16.dx,dstMV16x16.dy) */
/* Block 0 */
omxVCM4P2_MCReconBlock(pTempSrcRefBuf,
srcRefStep,
NULL,
pDst,
dstStep,
predictType,
pMEParams->rndVal);
armVCCOMM_SAD(pTempSrcCurrBuf,
16,
pDst,
dstStep,
&Sad,
8,
8);
pDstBlockSAD[0] = Sad;
/* Block 1 */
omxVCM4P2_MCReconBlock(pTempSrcRefBuf + 8,
srcRefStep,
NULL,
pDst,
dstStep,
predictType,
pMEParams->rndVal);
armVCCOMM_SAD(pTempSrcCurrBuf + 8,
16,
pDst,
dstStep,
&Sad,
8,
8);
pDstBlockSAD[1] = Sad;
/* Block 2 */
omxVCM4P2_MCReconBlock(pTempSrcRefBuf + (srcRefStep*8),
srcRefStep,
NULL,
pDst,
dstStep,
predictType,
pMEParams->rndVal);
armVCCOMM_SAD(pTempSrcCurrBuf + (16*8),
16,
pDst,
dstStep,
&Sad,
8,
8);
pDstBlockSAD[2] = Sad;
/* Block 3 */
omxVCM4P2_MCReconBlock(pTempSrcRefBuf + (srcRefStep*8) + 8,
srcRefStep,
NULL,
pDst,
dstStep,
predictType,
pMEParams->rndVal);
armVCCOMM_SAD(pTempSrcCurrBuf + (16*8) + 8,
16,
pDst,
dstStep,
&Sad,
8,
8);
pDstBlockSAD[3] = Sad;
}
return OMX_Sts_NoErr;
}
OMXResult armVCM4P2_BlockMatch_Integer(
const OMX_U8 *pSrcRefBuf,
OMX_INT refWidth,
const OMXRect *pRefRect,
const OMX_U8 *pSrcCurrBuf,
const OMXVCM4P2Coordinate *pCurrPointPos,
const OMXVCMotionVector *pSrcPreMV,
const OMX_INT *pSrcPreSAD,
void *pMESpec,
OMXVCMotionVector *pDstMV,
OMX_INT *pDstSAD,
OMX_U8 BlockSize
)
{
/* Definitions and Initializations*/
OMX_INT outer, inner, count,index;
OMX_INT candSAD;
/*(256*256 +1) this is to make the SAD max initially*/
OMX_INT minSAD = 0x10001, fromX, toX, fromY, toY;
/* Offset to the reference at the begining of the bounding box */
const OMX_U8 *pTempSrcRefBuf;
OMX_S16 x, y;
OMX_INT searchRange;
/* Argument error checks */
armRetArgErrIf(pSrcRefBuf == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(pRefRect == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(pSrcCurrBuf == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(pCurrPointPos == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(pMESpec == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(pDstMV == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(pDstSAD == NULL, OMX_Sts_BadArgErr);
searchRange = ((OMXVCM4P2MEParams *)pMESpec)->searchRange;
/* Check for valid region */
fromX = searchRange;
toX = searchRange;
fromY = searchRange;
toY = searchRange;
if ((pCurrPointPos->x - searchRange) < pRefRect->x)
{
fromX = pCurrPointPos->x - pRefRect->x;
}
if ((pCurrPointPos->x + BlockSize + searchRange) > (pRefRect->x + pRefRect->width))
{
toX = pRefRect->width - (pCurrPointPos->x - pRefRect->x) - BlockSize;
}
if ((pCurrPointPos->y - searchRange) < pRefRect->y)
{
fromY = pCurrPointPos->y - pRefRect->y;
}
if ((pCurrPointPos->y + BlockSize + searchRange) > (pRefRect->y + pRefRect->height))
{
toY = pRefRect->width - (pCurrPointPos->y - pRefRect->y) - BlockSize;
}
pDstMV->dx = -fromX;
pDstMV->dy = -fromY;
/* Looping on y- axis */
for (y = -fromY; y <= toY; y++)
{
/* Looping on x- axis */
for (x = -fromX; x <= toX; x++)
{
/* Positioning the pointer */
pTempSrcRefBuf = pSrcRefBuf + (refWidth * y) + x;
/* Calculate the SAD */
for (outer = 0, count = 0, index = 0, candSAD = 0;
outer < BlockSize;
outer++, index += refWidth - BlockSize)
{
for (inner = 0; inner < BlockSize; inner++, count++, index++)
{
candSAD += armAbs (pTempSrcRefBuf[index] - pSrcCurrBuf[count]);
}
}
/* Result calculations */
if (armVCM4P2_CompareMV (x, y, candSAD, pDstMV->dx/2, pDstMV->dy/2, minSAD))
{
*pDstSAD = candSAD;
minSAD = candSAD;
pDstMV->dx = x*2;
pDstMV->dy = y*2;
}
} /* End of x- axis */
} /* End of y-axis */
return OMX_Sts_NoErr;
}
OMXResult omxVCM4P2_TransRecBlockCoef_intra(
const OMX_U8 *pSrc,
OMX_S16 * pDst,
OMX_U8 * pRec,
OMX_S16 *pPredBufRow,
OMX_S16 *pPredBufCol,
OMX_S16 * pPreACPredict,
OMX_INT *pSumErr,
OMX_INT blockIndex,
OMX_U8 curQp,
const OMX_U8 *pQpBuf,
OMX_INT srcStep,
OMX_INT dstStep,
OMX_INT shortVideoHeader
)
{
/* 64 elements are needed but to align it to 16 bytes need
8 more elements of padding */
OMX_S16 tempBuf1[79], tempBuf2[79];
OMX_S16 tempBuf3[79];
OMX_S16 *pTempBuf1, *pTempBuf2,*pTempBuf3;
OMXVCM4P2VideoComponent videoComp;
OMX_U8 flag;
OMX_INT x, y, count, predDir;
OMX_INT predQP, ACPredFlag;
/* Aligning the local buffers */
pTempBuf1 = armAlignTo16Bytes(tempBuf1);
pTempBuf2 = armAlignTo16Bytes(tempBuf2);
pTempBuf3 = armAlignTo16Bytes(tempBuf3);
/* Argument error checks */
armRetArgErrIf(pSrc == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(pRec == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(pDst == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(!armIs8ByteAligned(pSrc), OMX_Sts_BadArgErr);
armRetArgErrIf(!armIs8ByteAligned(pRec), OMX_Sts_BadArgErr);
armRetArgErrIf(!armIs16ByteAligned(pDst), OMX_Sts_BadArgErr);
armRetArgErrIf(pPredBufRow == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(pPredBufCol == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(pPreACPredict == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(pSumErr == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(pQpBuf == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf((srcStep <= 0) || (dstStep <= 0) ||
(dstStep & 7) || (srcStep & 7)
, OMX_Sts_BadArgErr);
armRetArgErrIf((blockIndex < 0) || (blockIndex > 9), OMX_Sts_BadArgErr);
armRetArgErrIf((curQp <= 0) || (curQp >=32), OMX_Sts_BadArgErr);
/* Setting the videoComp */
if (blockIndex <= 3)
{
videoComp = OMX_VC_LUMINANCE;
}
else
{
videoComp = OMX_VC_CHROMINANCE;
}
/* Converting from 2-d to 1-d buffer */
for (y = 0, count = 0; y < 8; y++)
{
for(x= 0; x < 8; x++, count++)
{
pTempBuf1[count] = pSrc[(y*srcStep) + x];
}
}
omxVCM4P2_DCT8x8blk (pTempBuf1, pTempBuf2);
omxVCM4P2_QuantIntra_I(
pTempBuf2,
curQp,
blockIndex,
shortVideoHeader);
/* Converting from 1-D to 2-D buffer */
for (y = 0, count = 0; y < 8; y++)
{
for(x = 0; x < 8; x++, count++)
{
/* storing tempbuf2 to tempbuf1 */
pTempBuf1[count] = pTempBuf2[count];
pDst[(y*dstStep) + x] = pTempBuf2[count];
}
}
/* AC and DC prediction */
armVCM4P2_SetPredDir(
blockIndex,
pPredBufRow,
pPredBufCol,
&predDir,
&predQP,
pQpBuf);
armRetDataErrIf(((predQP <= 0) || (predQP >= 32)), OMX_Sts_BadArgErr);
flag = 1;
if (*pSumErr < 0)
{
ACPredFlag = 0;
}
else
{
ACPredFlag = 1;
}
armVCM4P2_ACDCPredict(
pTempBuf2,
pPreACPredict,
pPredBufRow,
pPredBufCol,
curQp,
predQP,
predDir,
ACPredFlag,
videoComp,
flag,
pSumErr);
/* Reconstructing the texture data */
omxVCM4P2_QuantInvIntra_I(
pTempBuf1,
curQp,
videoComp,
shortVideoHeader);
omxVCM4P2_IDCT8x8blk (pTempBuf1, pTempBuf3);
for(count = 0; count < 64; count++)
{
pRec[count] = armMax(0,pTempBuf3[count]);
}
return OMX_Sts_NoErr;
}
OMXResult omxVCM4P2_EncodeMV(
OMX_U8 **ppBitStream,
OMX_INT *pBitOffset,
const OMXVCMotionVector * pMVCurMB,
const OMXVCMotionVector * pSrcMVLeftMB,
const OMXVCMotionVector * pSrcMVUpperMB,
const OMXVCMotionVector * pSrcMVUpperRightMB,
OMX_INT fcodeForward,
OMXVCM4P2MacroblockType MBType
)
{
OMXVCMotionVector dstMVPred, diffMV;
OMXVCMotionVector dstMVPredME[12];
/* Initialized to remove compilation warning */
OMX_INT iBlk, i, count = 1;
OMX_S32 mvHorResidual, mvVerResidual, mvHorData, mvVerData;
OMX_U8 scaleFactor, index;
/* Argument error checks */
armRetArgErrIf(ppBitStream == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(*ppBitStream == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(pBitOffset == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(pMVCurMB == NULL, OMX_Sts_BadArgErr);
armRetArgErrIf(((*pBitOffset < 0) || (*pBitOffset > 7)), OMX_Sts_BadArgErr);
armRetArgErrIf(((fcodeForward < 1) || (fcodeForward > 7)), \
OMX_Sts_BadArgErr);
if ((MBType == OMX_VC_INTRA) ||
(MBType == OMX_VC_INTRA_Q)
)
{
/* No candidate vectors hence make them zero */
for (i = 0; i < 12; i++)
{
dstMVPredME[i].dx = 0;
dstMVPredME[i].dy = 0;
}
return OMX_Sts_NoErr;
}
if ((MBType == OMX_VC_INTER4V) || (MBType == OMX_VC_INTER4V_Q))
{
count = 4;
}
else if ((MBType == OMX_VC_INTER) || (MBType == OMX_VC_INTER_Q))
{
count = 1;
}
/* Calculating the scale factor */
scaleFactor = 1 << (fcodeForward -1);
for (iBlk = 0; iBlk < count; iBlk++)
{
/* Find the predicted vector */
omxVCM4P2_FindMVpred (
pMVCurMB,
pSrcMVLeftMB,
pSrcMVUpperMB,
pSrcMVUpperRightMB,
&dstMVPred,
dstMVPredME,
iBlk );
/* Calculating the differential motion vector (diffMV) */
diffMV.dx = pMVCurMB[iBlk].dx - dstMVPred.dx;
diffMV.dy = pMVCurMB[iBlk].dy - dstMVPred.dy;
/* Calculating the mv_data and mv_residual for Horizantal MV */
if (diffMV.dx == 0)
{
mvHorResidual = 0;
mvHorData = 0;
}
else
{
mvHorResidual = ( armAbs(diffMV.dx) - 1) % scaleFactor;
mvHorData = (armAbs(diffMV.dx) - mvHorResidual + (scaleFactor - 1))
/ scaleFactor;
if (diffMV.dx < 0)
{
mvHorData = -mvHorData;
}
}
/* Calculating the mv_data and mv_residual for Vertical MV */
if (diffMV.dy == 0)
{
mvVerResidual = 0;
mvVerData = 0;
}
else
{
mvVerResidual = ( armAbs(diffMV.dy) - 1) % scaleFactor;
mvVerData = (armAbs(diffMV.dy) - mvVerResidual + (scaleFactor - 1))
/ scaleFactor;
if (diffMV.dy < 0)
{
mvVerData = -mvVerData;
}
}
/* Huffman encoding */
/* The index is actually calculate as
index = ((float) (mvHorData/2) + 16) * 2,
meaning the MV data is halfed and then normalized
to begin with zero and then doubled to take care of indexing
the fractional part included */
index = mvHorData + 32;
armPackVLC32 (ppBitStream, pBitOffset, armVCM4P2_aVlcMVD[index]);
if ((fcodeForward > 1) && (diffMV.dx != 0))
{
armPackBits (ppBitStream, pBitOffset, mvHorResidual, (fcodeForward -1));
}
/* The index is actually calculate as
index = ((float) (mvVerData/2) + 16) * 2,
meaning the MV data is halfed and then normalized
to begin with zero and then doubled to take care of indexing
the fractional part included */
index = mvVerData + 32;
armPackVLC32 (ppBitStream, pBitOffset, armVCM4P2_aVlcMVD[index]);
if ((fcodeForward > 1) && (diffMV.dy != 0))
{
armPackBits (ppBitStream, pBitOffset, mvVerResidual, (fcodeForward -1));
}
}
return OMX_Sts_NoErr;
}
OMXResult armVCM4P10_InterpolateHalfDiag_Luma(
const OMX_U8* pSrc,
OMX_U32 iSrcStep,
OMX_U8* pDst,
OMX_U32 iDstStep,
OMX_U32 iWidth,
OMX_U32 iHeight
)
{
OMX_S32 HalfCoeff, pos;
OMX_S16 Buf [21 * 16]; /* 21 rows by 16 pixels per row */
OMX_U32 y, x;
/* check for argument error */
armRetArgErrIf(pSrc == NULL, OMX_Sts_BadArgErr)
armRetArgErrIf(pDst == NULL, OMX_Sts_BadArgErr)
/*
* Intermediate values will be 1/2 pel at Horizontal direction
* Starting at (0.5, -2) at top extending to (0.5, height + 3) at bottom
* Buf contains a 2D array of size (iWidth)X(iHeight + 5)
*/
for (y = 0; y < iHeight + 5; y++)
{
for (x = 0; x < iWidth; x++)
{
pos = (y-2) * iSrcStep + x;
HalfCoeff =
pSrc [pos - 2] -
5 * pSrc [pos - 1] +
20 * pSrc [pos] +
20 * pSrc [pos + 1] -
5 * pSrc [pos + 2] +
pSrc [pos + 3];
Buf [y * iWidth + x] = (OMX_S16)HalfCoeff;
} /* x */
} /* y */
/* Vertical interpolate */
for (y = 0; y < iHeight; y++)
{
for (x = 0; x < iWidth; x++)
{
pos = y * iWidth + x;
HalfCoeff =
Buf [pos] -
5 * Buf [pos + 1 * iWidth] +
20 * Buf [pos + 2 * iWidth] +
20 * Buf [pos + 3 * iWidth] -
5 * Buf [pos + 4 * iWidth] +
Buf [pos + 5 * iWidth];
HalfCoeff = (HalfCoeff + 512) >> 10;
HalfCoeff = armClip(0, 255, HalfCoeff);
pDst [y * iDstStep + x] = (OMX_U8) HalfCoeff;
}
}
return OMX_Sts_NoErr;
}
