Proton/jxrlib/image/decode/segdec.c
2020-09-29 14:29:06 -05:00

1206 lines
42 KiB
C

//*@@@+++@@@@******************************************************************
//
// Copyright © Microsoft Corp.
// All rights reserved.
//
// Redistribution and use in source and binary forms, with or without
// modification, are permitted provided that the following conditions are met:
//
// • Redistributions of source code must retain the above copyright notice,
// this list of conditions and the following disclaimer.
// • Redistributions in binary form must reproduce the above copyright notice,
// this list of conditions and the following disclaimer in the documentation
// and/or other materials provided with the distribution.
//
// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
// AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
// IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
// ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
// LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
// CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
// SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
// INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
// CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
// ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
// POSSIBILITY OF SUCH DAMAGE.
//
//*@@@---@@@@******************************************************************
#include "strcodec.h"
#include "decode.h"
#ifdef MEM_TRACE
#define TRACE_MALLOC 1
#define TRACE_NEW 0
#define TRACE_HEAP 0
#include "memtrace.h"
#endif
extern const int dctIndex[3][16];
extern const int blkOffset[16];
extern const int blkOffsetUV[4];
static Int DecodeSignificantAbsLevel (struct CAdaptiveHuffman *pAHexpt, BitIOInfo* pIO);
//#undef X86OPT_INLINE
#ifdef X86OPT_INLINE
#define _FORCEINLINE __forceinline
#else // X86OPT_INLINE
#define _FORCEINLINE
#endif // X86OPT_INLINE
//================================================================
// Memory access functions
//================================================================
static U32 _FORCEINLINE _load4(void* pv)
{
#ifdef _BIG__ENDIAN_
return (*(U32*)pv);
#else // _BIG__ENDIAN_
#if defined(_M_IA64) || defined(_ARM_)
U32 v;
v = ((U16 *) pv)[0];
v |= ((U32)((U16 *) pv)[1]) << 16;
return _byteswap_ulong(v);
#else // _M_IA64
return _byteswap_ulong(*(U32*)pv);
#endif // _M_IA64
#endif // _BIG__ENDIAN_
}
static _FORCEINLINE U32 _peekBit16(BitIOInfo* pIO, U32 cBits)
{
PEEKBIT16(pIO, cBits);
// masking is not needed here because shift of unsigned int is implemented as a logical shift (SHR)!
}
#define LOAD16 _load4
static _FORCEINLINE U32 _flushBit16(BitIOInfo* pIO, U32 cBits)
{
FLUSHBIT16(pIO, cBits);
}
static _FORCEINLINE U32 _getBit16(BitIOInfo* pIO, U32 cBits)
{
U32 uiRet = _peekBit16(pIO, cBits);
_flushBit16(pIO, cBits);
return uiRet;
}
#define SIGN_BIT(TypeOrValue) (((UInt) 1) << (8 * sizeof (TypeOrValue) - 1))
/***********************************************************************************************************
Huffman decode (input is a fully built Huffman table)
***********************************************************************************************************/
Int getHuff(const short *pDecodeTable, BitIOInfo* pIO)
{
Int iSymbol, iSymbolHuff;
iSymbol = pDecodeTable[peekBit16(pIO, HUFFMAN_DECODE_ROOT_BITS)];
flushBit16(pIO, iSymbol < 0 ? HUFFMAN_DECODE_ROOT_BITS : iSymbol & ((1 << HUFFMAN_DECODE_ROOT_BITS_LOG) - 1));
iSymbolHuff = iSymbol >> HUFFMAN_DECODE_ROOT_BITS_LOG;
if (iSymbolHuff < 0) {
iSymbolHuff = iSymbol;
while ((iSymbolHuff = pDecodeTable[iSymbolHuff + SIGN_BIT (pDecodeTable[0]) + getBit16(pIO, 1)]) < 0);
}
return (iSymbolHuff);
}
#if 1
static _FORCEINLINE U32 _getBool16(BitIOInfo* pIO)
{
U32 uiRet = pIO->uiAccumulator >> 31;//_peekBit16(pIO, 1);
//_flushBit16(pIO, 1);
pIO->cBitsUsed++;
if (pIO->cBitsUsed < 16) {
pIO->uiAccumulator <<= 1;
}
else {
pIO->pbCurrent = MASKPTR(pIO->pbCurrent + ((pIO->cBitsUsed >> 3)/* & 2*/), pIO->iMask);
pIO->cBitsUsed &= 16 - 1;
pIO->uiAccumulator = LOAD16(pIO->pbCurrent) << pIO->cBitsUsed;
}
return uiRet;
}
static _FORCEINLINE I32 _getSign(BitIOInfo* pIO)
{
I32 uiRet = (int) pIO->uiAccumulator >> 31;//_peekBit16(pIO, 1);
//_flushBit16(pIO, 1);
pIO->cBitsUsed++;
if (pIO->cBitsUsed < 16) {
pIO->uiAccumulator <<= 1;
}
else {
pIO->pbCurrent = MASKPTR(pIO->pbCurrent + ((pIO->cBitsUsed >> 3)/* & 2*/), pIO->iMask);
pIO->cBitsUsed &= 16 - 1;
pIO->uiAccumulator = LOAD16(pIO->pbCurrent) << pIO->cBitsUsed;
}
return uiRet;
}
#else
#define _getBool16(x) _getBit16((x),1)
#define _getSign(x) (-_getBit16((x),1))
#endif
/** this function returns cBits if zero is read, or a signed value if first cBits are not all zero **/
static _FORCEINLINE I32 _getBit16s(BitIOInfo* pIO, U32 cBits)
{
I32 iRet = (I32)_peekBit16(pIO, cBits + 1);
iRet = ((iRet >> 1) ^ (-(iRet & 1))) + (iRet & 1);
_flushBit16(pIO, cBits + (iRet != 0));
return iRet;
}
/*************************************************************************
Huffman decoding with short tables
*************************************************************************/
static _FORCEINLINE Int _getHuffShort(const short *pDecodeTable, BitIOInfo* pIO)
{
Int iSymbol = pDecodeTable[_peekBit16(pIO, HUFFMAN_DECODE_ROOT_BITS)];
assert(iSymbol >= 0);
// for some strange reason, inlining flushBit doesn't work well
flushBit16(pIO, iSymbol & ((1 << HUFFMAN_DECODE_ROOT_BITS_LOG) - 1));
return (iSymbol >> HUFFMAN_DECODE_ROOT_BITS_LOG);
}
/*************************************************************************
Adapt + Huffman init
*************************************************************************/
static Int AdaptDecFixed (CAdaptiveHuffman *pAH)
{
AdaptDiscriminant (pAH);
return ICERR_OK;
}
/*************************************************************************
DecodeCBP
*************************************************************************/
static Void DecodeCBP(CWMImageStrCodec * pSC, CCodingContext *pContext)
{
BitIOInfo* pIO = pContext->m_pIOAC;
const COLORFORMAT cf = pSC->m_param.cfColorFormat;
const Int iChannel = (cf == NCOMPONENT || cf == CMYK) ? (Int) pSC->m_param.cNumChannels : 1;
Int iCBPCY, iCBPCU , iCBPCV;
Int k, iBlock, i;
Int iNumCBP;
Bool bIsChroma;
CAdaptiveHuffman *pAHCBP = pContext->m_pAdaptHuffCBPCY;
CAdaptiveHuffman *pAHCBP1 = pContext->m_pAdaptHuffCBPCY1;
CAdaptiveHuffman *pAHex1 = pContext->m_pAHexpt[1];
readIS_L1(pSC, pIO);
for (i = 0; i < iChannel; i++) {
iCBPCY = iCBPCU = iCBPCV = 0;
iNumCBP = _getHuffShort(pAHCBP1->m_hufDecTable, pIO);
pAHCBP1->m_iDiscriminant += pAHCBP1->m_pDelta[iNumCBP];
switch (iNumCBP) {
case 2:
iNumCBP = _getBit16(pIO, 2);
if (iNumCBP == 0)
iNumCBP = 3;
else if (iNumCBP == 1)
iNumCBP = 5;
else {
static const Int aTab[] = { 6, 9, 10, 12 };
iNumCBP = aTab[iNumCBP * 2 + _getBool16 (pIO) - 4];
}
break;
case 1:
iNumCBP = 1 << _getBit16(pIO, 2);
break;
case 3:
iNumCBP = 0xf ^ (1 << _getBit16(pIO, 2));
break;
case 4:
iNumCBP = 0xf;
}
for (iBlock = 0; iBlock < 4; iBlock++) {
if (iNumCBP & (1 << iBlock)) {
static const UInt gFLC0[] = { 0,2,1,2,2,0 };
static const UInt gOff0[] = { 0,4,2,8,12,1 };
static const UInt gOut0[] = { 0,15,3,12, 1,2,4,8, 5,6,9,10, 7,11,13,14 };
Int iNumBlockCBP = getHuff(pAHCBP->m_hufDecTable, pIO);
unsigned int val = (unsigned int) iNumBlockCBP + 1, iCode1;
pAHCBP->m_iDiscriminant += pAHCBP->m_pDelta[iNumBlockCBP];
iNumBlockCBP = 0;
if (val >= 6) { // chroma present
if (_getBool16 (pIO)) {
iNumBlockCBP = 0x10;
}
else if (_getBool16 (pIO)) {
iNumBlockCBP = 0x20;
}
else {
iNumBlockCBP = 0x30;
}
if (val == 9) {
if (_getBool16 (pIO)) {
// do nothing
}
else if (_getBool16 (pIO)) {
val = 10;
}
else {
val = 11;
}
}
val -= 6;
}
iCode1 = gOff0[val];
if (gFLC0[val]) {
iCode1 += _getBit16(pIO, gFLC0[val]);
}
iNumBlockCBP += gOut0[iCode1];
switch (cf) {
case YUV_444:
iCBPCY |= ((iNumBlockCBP & 0xf) << (iBlock * 4));
for (k = 0; k < 2; k++) {
bIsChroma = ((iNumBlockCBP>>(k+4)) & 0x01);
if (bIsChroma) { // U is present in block
Int iCode = _getHuffShort(pAHex1->m_hufDecTable, pIO);
switch (iCode) {
case 1:
iCode = _getBit16(pIO, 2);
if (iCode == 0)
iCode = 3;
else if (iCode == 1)
iCode = 5;
else {
static const Int aTab[] = { 6, 9, 10, 12 };
iCode = aTab[iCode * 2 + _getBool16 (pIO) - 4];
}
break;
case 0:
iCode = 1 << _getBit16(pIO, 2);
break;
case 2:
iCode = 0xf ^ (1 << _getBit16(pIO, 2));
break;
case 3:
iCode = 0xf;
}
if (k == 0)
iCBPCU |= (iCode << (iBlock * 4));
else
iCBPCV |= (iCode << (iBlock * 4));
}
}
break;
case YUV_420:
iCBPCY |= ((iNumBlockCBP & 0xf) << (iBlock * 4));
iCBPCU |= ((iNumBlockCBP >> 4) & 0x1) << (iBlock);
iCBPCV |= ((iNumBlockCBP >> 5) & 0x1) << (iBlock);
break;
case YUV_422:
iCBPCY |= ((iNumBlockCBP & 0xf) << (iBlock * 4));
for (k = 0; k < 2; k ++) {
Int iCode = 5;
const Int iShift[4] = {0, 1, 4, 5};
if((iNumBlockCBP >> (k + 4)) & 0x01) {
if(_getBool16(pIO)) {
iCode = 1;
}
else if(_getBool16(pIO)){
iCode = 4;
}
iCode <<= iShift[iBlock];
if(k == 0) iCBPCU |= iCode;
else iCBPCV |= iCode;
}
}
break;
default:
iCBPCY |= (iNumBlockCBP << (iBlock * 4));
}
}
}
pSC->MBInfo.iDiffCBP[i] = iCBPCY;
if (cf == YUV_420 || cf == YUV_444 || cf == YUV_422) {
pSC->MBInfo.iDiffCBP[1] = iCBPCU;
pSC->MBInfo.iDiffCBP[2] = iCBPCV;
}
}
}
/*************************************************************************
Experimental code -- decodeBlock
SR = <0 1 2> == <last, nonsignificant, significant run>
alphabet 12:
pAHexpt[0] == <SR', SL, SR | first symbol>
alphabet 6:
pAHexpt[1] == <SR', SL | continuous>
pAHexpt[2] == <SR', SL | continuous>
alphabet 4:
pAHexpt[3] == <SR', SL | 2 free slots> (SR may be last or insignificant only)
alphabet f(run) (this can be extended to 6 contexts - SL and SR')
pAHexpt[4] == <run | continuous>
alphabet f(lev) (this can be extended to 9 contexts)
pAHexpt[5-6] == <lev | continuous> first symbol
pAHexpt[7-8] == <lev | continuous> condition on SRn no use
*************************************************************************/
Int _FORCEINLINE DecodeSignificantRun (Int iMaxRun, struct CAdaptiveHuffman *pAHexpt, BitIOInfo* pIO)
{
Int iIndex;
static const Int aRemap[] = {1,2,3,5,7, 1,2,3,5,7, /*1,2,3,4,6, */1,2,3,4,5 };
Int iBin = gSignificantRunBin[iMaxRun];
Int iRun = 0, iFLC = 0;
if (iMaxRun < 5) {
if (iMaxRun == 1) {
return 1;
}
else if (_getBool16 (pIO)) {
return 1;
}
else if (iMaxRun == 2 || _getBool16 (pIO)) {
return 2;
}
else if (iMaxRun == 3 || _getBool16 (pIO)) {
return 3;
}
return 4;
}
iIndex = _getHuffShort (pAHexpt->m_hufDecTable, pIO);
iIndex += iBin * 5;
iRun = aRemap[iIndex];
iFLC = gSignificantRunFixedLength[iIndex];
if (iFLC) {
iRun += _getBit16 (pIO, iFLC);
}
return iRun;
}
#ifndef X86OPT_INLINE
static Void DecodeFirstIndex (Int *pIndex, struct CAdaptiveHuffman *pAHexpt,
BitIOInfo* pIO)
#else
static __forceinline Void DecodeFirstIndex (Int *pIndex, struct CAdaptiveHuffman *pAHexpt,
BitIOInfo* pIO)
#endif
{
Int iIndex;
iIndex = getHuff (pAHexpt->m_hufDecTable, pIO);
pAHexpt->m_iDiscriminant += pAHexpt->m_pDelta[iIndex];
pAHexpt->m_iDiscriminant1 += pAHexpt->m_pDelta1[iIndex];
*pIndex = iIndex;
}
#ifndef X86OPT_INLINE
static Void DecodeIndex (Int *pIndex, Int iLoc, struct CAdaptiveHuffman *pAHexpt,
BitIOInfo* pIO)
#else
static __forceinline Void DecodeIndex (Int *pIndex, Int iLoc,
struct CAdaptiveHuffman *pAHexpt, BitIOInfo* pIO)
#endif
{
Int iIndex;
if (iLoc < 15) {
iIndex = _getHuffShort (pAHexpt->m_hufDecTable, pIO);
pAHexpt->m_iDiscriminant += pAHexpt->m_pDelta[iIndex];
pAHexpt->m_iDiscriminant1 += pAHexpt->m_pDelta1[iIndex];
*pIndex = iIndex;
}
else if (iLoc == 15) {
if (_getBool16 (pIO) == 0) {
iIndex = 0;
}
else if (_getBool16 (pIO) == 0) {
iIndex = 2;
}
else {
iIndex = 1 + 2 * _getBool16 (pIO);
}
*pIndex = iIndex;
}
else { //if (iLoc == 16) { /* deterministic */
Int iSL = _getBit16 (pIO, 1/* + 1*/);
*pIndex = iSL;// >> 1;
}
}
static _FORCEINLINE Int DecodeBlock (Bool bChroma, Int *aLocalCoef, struct CAdaptiveHuffman **pAHexpt,
const Int iContextOffset, BitIOInfo* pIO, Int iLocation)
{
Int iSR, iSRn, iIndex, iNumNonzero = 1, iCont, iSign;
struct CAdaptiveHuffman **pAH1 = pAHexpt + iContextOffset + bChroma * 3;
/** first symbol **/
DecodeFirstIndex (&iIndex, /*&iSign, */pAH1[0], pIO);
iSR = (iIndex & 1);
iSRn = iIndex >> 2;
iCont = iSR & iSRn;
iSign = _getSign(pIO);
if (iIndex & 2 /* iSL */) {
aLocalCoef[1] = (DecodeSignificantAbsLevel (pAHexpt[6 + iContextOffset + iCont], pIO) ^ iSign) - iSign;
}
else {
aLocalCoef[1] = (1 | iSign); // 0 -> 1; -1 -> -1
}
aLocalCoef[0] = 0;
if (iSR == 0) {
aLocalCoef[0] = DecodeSignificantRun (15 - iLocation, pAHexpt[0], pIO);
}
iLocation += aLocalCoef[0] + 1;
while (iSRn != 0) {
iSR = iSRn & 1;
aLocalCoef[iNumNonzero * 2] = 0;
if (iSR == 0) {
aLocalCoef[iNumNonzero * 2] = DecodeSignificantRun (15 - iLocation, pAHexpt[0], pIO);
}
iLocation += aLocalCoef[iNumNonzero * 2] + 1;
DecodeIndex (&iIndex, /*&iSign, */iLocation, pAH1[iCont + 1], pIO);
iSRn = iIndex >> 1;
assert (iSRn >= 0 && iSRn < 3);
iCont &= iSRn; /** huge difference! **/
iSign = _getSign(pIO);
if (iIndex & 1 /* iSL */) {
aLocalCoef[iNumNonzero * 2 + 1] =
(DecodeSignificantAbsLevel (pAHexpt[6 + iContextOffset + iCont], pIO) ^ iSign) - iSign;
}
else {
aLocalCoef[iNumNonzero * 2 + 1] = (1 | iSign); // 0 -> 1; -1 -> -1 (was 1 + (iSign * 2))
}
iNumNonzero++;
}
return iNumNonzero;
}
/*************************************************************************
DecodeBlockHighpass :
*************************************************************************/
static _FORCEINLINE Int DecodeBlockHighpass (const Bool bChroma, struct CAdaptiveHuffman **pAHexpt,
BitIOInfo* pIO, const Int iQP, Int *pCoef, CAdaptiveScan *pScan)
{
const Int iContextOffset = CTDC + CONTEXTX;
UInt iLoc = 1;
Int iSR, iSRn, iIndex, iNumNonzero = 1, iCont, iSign, iLevel;
struct CAdaptiveHuffman **pAH1 = pAHexpt + iContextOffset + bChroma * 3;
const CAdaptiveScan *pConstScan = (const CAdaptiveScan *) pScan;
/** first symbol **/
DecodeFirstIndex (&iIndex, /*&iSign, */pAH1[0], pIO);
iSR = (iIndex & 1);
iSRn = iIndex >> 2;
iCont = iSR & iSRn;
iSign = _getSign(pIO);
iLevel = (iQP ^ iSign) - iSign;
if (iIndex & 2 /* iSL */) {
iLevel *= DecodeSignificantAbsLevel (pAHexpt[6 + iContextOffset + iCont], pIO);// ^ iSign) - iSign;
}
//else {
// iLevel = (1 | iSign); // 0 -> 1; -1 -> -1
//}
if (iSR == 0) {
iLoc += DecodeSignificantRun (15 - iLoc, pAHexpt[0], pIO);
}
iLoc &= 0xf;
pCoef[pConstScan[iLoc].uScan] = (PixelI) iLevel;//(PixelI)(iQP * iLevel);
pScan[iLoc].uTotal++;
if (iLoc && pScan[iLoc].uTotal > pScan[iLoc - 1].uTotal) {
CAdaptiveScan cTemp = pScan[iLoc];
pScan[iLoc] = pScan[iLoc - 1];
pScan[iLoc - 1] = cTemp;
}
iLoc = (iLoc + 1) & 0xf;
//iLoc++;
while (iSRn != 0) {
iSR = iSRn & 1;
if (iSR == 0) {
iLoc += DecodeSignificantRun (15 - iLoc, pAHexpt[0], pIO);
if (iLoc >= 16)
return 16;
}
DecodeIndex (&iIndex, /*&iSign, */iLoc + 1, pAH1[iCont + 1], pIO);
iSRn = iIndex >> 1;
assert (iSRn >= 0 && iSRn < 3);
iCont &= iSRn; /** huge difference! **/
iSign = _getSign(pIO);
iLevel = (iQP ^ iSign) - iSign;
if (iIndex & 1 /* iSL */) {
iLevel *= DecodeSignificantAbsLevel (pAHexpt[6 + iContextOffset + iCont], pIO);// ^ iSign) - iSign;
//iLevel = (DecodeSignificantAbsLevel (pAHexpt[6 + iContextOffset + iCont], pIO) ^ iSign) - iSign;
}
//else {
// iLevel = (1 | iSign); // 0 -> 1; -1 -> -1 (was 1 + (iSign * 2))
//}
pCoef[pConstScan[iLoc].uScan] = (PixelI) iLevel;//(PixelI)(iQP * iLevel);
pScan[iLoc].uTotal++;
if (iLoc && pScan[iLoc].uTotal > pScan[iLoc - 1].uTotal) {
CAdaptiveScan cTemp = pScan[iLoc];
pScan[iLoc] = pScan[iLoc - 1];
pScan[iLoc - 1] = cTemp;
}
iLoc = (iLoc + 1) & 0xf;
iNumNonzero++;
}
return iNumNonzero;
}
/*************************************************************************
DecodeBlockAdaptive
*************************************************************************/
static _FORCEINLINE Int DecodeBlockAdaptive (Bool bNoSkip, Bool bChroma, CAdaptiveHuffman **pAdHuff,
BitIOInfo *pIO, BitIOInfo *pIOFL,
PixelI *pCoeffs, CAdaptiveScan *pScan,
const Int iModelBits, const Int iTrim, const Int iQP,
const Int *pOrder, const Bool bSkipFlexbits)
{
// const Int iLocation = 1;
// const Int iContextOffset = CTDC + CONTEXTX;
Int kk, iNumNonzero = 0, iFlex = iModelBits - iTrim;
if (iFlex < 0 || bSkipFlexbits)
iFlex = 0;
if (bNoSkip) {
const Int iQP1 = (iQP << iModelBits);
iNumNonzero = DecodeBlockHighpass (bChroma, pAdHuff, pIO, iQP1, pCoeffs, pScan);
}
if (iFlex) {
UInt k;
if (iQP + iTrim == 1) { // only iTrim = 0, iQP = 1 is legal
assert (iTrim == 0);
assert (iQP == 1);
for (k = 1; k < 16; k++) {
PixelI *pk = pCoeffs + pOrder[k];
if (*pk < 0) {
Int fine = _getBit16(pIOFL, iFlex);
*pk -= (PixelI)(fine);
}
else if (*pk > 0) {
Int fine = _getBit16(pIOFL, iFlex);
*pk += (PixelI)(fine);
}
else {
*pk = (PixelI)(_getBit16s(pIOFL, iFlex));
}
}
}
else {
const Int iQP1 = iQP << iTrim;
for (k = 1; k < 16; k++) {
kk = pCoeffs[pOrder[k]];
if (kk < 0) {
Int fine = _getBit16(pIOFL, iFlex);
pCoeffs[pOrder[k]] -= (PixelI)(iQP1 * fine);
}
else if (kk > 0) {
Int fine = _getBit16(pIOFL, iFlex);
pCoeffs[pOrder[k]] += (PixelI)(iQP1 * fine);
}
else {
pCoeffs[pOrder[k]] = (PixelI)(iQP1 * _getBit16s(pIOFL, iFlex));
}
}
}
}
return iNumNonzero;
}
/*************************************************************************
GetCoeffs
*************************************************************************/
static _FORCEINLINE Int DecodeCoeffs (CWMImageStrCodec * pSC, CCodingContext *pContext,
Int iMBX, Int iMBY,
BitIOInfo* pIO, BitIOInfo *pIOFL)
{
CWMITile * pTile = pSC->pTile + pSC->cTileColumn;
const COLORFORMAT cf = pSC->m_param.cfColorFormat;
const Int iChannels = (Int) pSC->m_param.cNumChannels;
const Int iPlanes = (cf == YUV_420 || cf == YUV_422) ? 1 : iChannels;
Int iQP;
CAdaptiveScan *pScan;
PixelI *pCoeffs;
Int i, iBlock, iSubblock, iNBlocks = 4;
Int iModelBits = pContext->m_aModelAC.m_iFlcBits[0];
Int aLaplacianMean[2] = { 0, 0}, *pLM = aLaplacianMean + 0;
const Int *pOrder = dctIndex[0];
const Int iOrient = pSC->MBInfo.iOrientation;
Bool bChroma = FALSE;
Int iCBPCU = pSC->MBInfo.iCBP[1];
Int iCBPCV = pSC->MBInfo.iCBP[2];
Int iCBPCY = pSC->MBInfo.iCBP[0];
UNREFERENCED_PARAMETER( iMBX );
UNREFERENCED_PARAMETER( iMBY );
/** set scan arrays and other MB level constants **/
if (iOrient == 1) {
pScan = pContext->m_aScanVert;
}
else {
pScan = pContext->m_aScanHoriz;
}
if (cf == YUV_420) {
iNBlocks = 6;
iCBPCY += (iCBPCU << 16) + (iCBPCV << 20);
}
else if (cf == YUV_422) {
iNBlocks = 8;
iCBPCY += (iCBPCU << 16) + (iCBPCV << 24);
}
for (i = 0; i < iPlanes; i++) {
Int iIndex = 0, iNumNonZero;
if(pSC->WMISCP.sbSubband != SB_NO_FLEXBITS)
readIS_L1(pSC, pIOFL);
for (iBlock = 0; iBlock < iNBlocks; iBlock++) {
readIS_L2(pSC, pIO);
if (pIO != pIOFL)
readIS_L2(pSC, pIOFL);
iQP = (pSC->m_param.bTranscode ? 1 : pTile->pQuantizerHP[iPlanes > 1 ? i : (iBlock > 3 ? (cf == YUV_420 ? iBlock - 3 : iBlock / 2 - 1) : 0)][pSC->MBInfo.iQIndexHP].iQP);
for (iSubblock = 0; iSubblock < 4; iSubblock++, iIndex++, iCBPCY >>= 1) {
pCoeffs = pSC->p1MBbuffer[i] + blkOffset[iIndex & 0xf];
//if (iBlock < 4) {//(cf == YUV_444) {
//bBlockNoSkip = ((iTempCBPC & (1 << iIndex1)) != 0);
//pCoeffs = pSC->p1MBbuffer[iBlock >> 2] + blkOffset[iIndex & 0xf];
//}
//else {
if (iBlock >= 4) {
if(cf == YUV_420) {
pCoeffs = pSC->p1MBbuffer[iBlock - 3] + blkOffsetUV[iSubblock];
}
else { // YUV_422
pCoeffs = pSC->p1MBbuffer[1 + (1 & (iBlock >> 1))] + ((iBlock & 1) * 32) + blkOffsetUV_422[iSubblock];
}
}
/** read AC values **/
assert (pSC->m_Dparam->bSkipFlexbits == 0 || pSC->WMISCP.bfBitstreamFormat == FREQUENCY || pSC->WMISCP.sbSubband == SB_NO_FLEXBITS);
iNumNonZero = DecodeBlockAdaptive ((iCBPCY & 1), bChroma, pContext->m_pAHexpt,
pIO, pIOFL, pCoeffs, pScan, iModelBits, pContext->m_iTrimFlexBits,
iQP, pOrder, pSC->m_Dparam->bSkipFlexbits);
if(iNumNonZero > 16) // something is wrong!
return ICERR_ERROR;
// shouldn't this be > 15?
(*pLM) += iNumNonZero;
}
if (iBlock == 3) {
iModelBits = pContext->m_aModelAC.m_iFlcBits[1];
pLM = aLaplacianMean + 1;
bChroma = TRUE;
}
}
iCBPCY = pSC->MBInfo.iCBP[(i + 1) & 0xf];
assert (MAX_CHANNELS == 16);
}
/** update model at end of MB **/
UpdateModelMB (cf, iChannels, aLaplacianMean, &(pContext->m_aModelAC));
return ICERR_OK;
}
/*************************************************************************
DecodeSignificantAbsLevel
*************************************************************************/
#ifndef X86OPT_INLINE
static Int DecodeSignificantAbsLevel (struct CAdaptiveHuffman *pAHexpt, BitIOInfo* pIO)
#else
static __forceinline Int DecodeSignificantAbsLevel (struct CAdaptiveHuffman *pAHexpt, BitIOInfo* pIO)
#endif
{
UInt iIndex;
Int iFixed, iLevel;
static const Int aRemap[] = { 2, 3, 4, 6, 10, 14 };
static const Int aFixedLength[] = { 0, 0, 1, 2, 2, 2 };
iIndex = (UInt)getHuff (pAHexpt->m_hufDecTable, pIO);
assert(iIndex <= 6);
pAHexpt->m_iDiscriminant += pAHexpt->m_pDelta[iIndex];
if (iIndex < 2) {
iLevel = iIndex + 2; // = aRemap[iIndex]
}
else if (iIndex < 6) {
iFixed = aFixedLength[iIndex];
iLevel = aRemap[iIndex] + _getBit16 (pIO, iFixed);
}
else{
iFixed = _getBit16 (pIO, 4) + 4;
if (iFixed == 19) {
iFixed += _getBit16 (pIO, 2);
if (iFixed == 22) {
iFixed += _getBit16 (pIO, 3);
}
}
iLevel = 2 + (1 << iFixed);
iIndex = getBit32 (pIO, iFixed);
iLevel += iIndex;
}
return iLevel;
}
U8 decodeQPIndex(BitIOInfo* pIO,U8 cBits)
{
if(_getBit16(pIO, 1) == 0)
return 0;
return (U8)(_getBit16(pIO, cBits) + 1);
}
/*************************************************************************
DecodeSecondStageCoeff
*************************************************************************/
Int DecodeMacroblockLowpass (CWMImageStrCodec * pSC, CCodingContext *pContext,
Int iMBX, Int iMBYdummy)
{
const COLORFORMAT cf = pSC->m_param.cfColorFormat;
const Int iChannels = (Int) pSC->m_param.cNumChannels;
const Int iFullPlanes = (cf == YUV_420 || cf == YUV_422) ? 2 : iChannels;
Int k;
CAdaptiveScan *pScan = pContext->m_aScanLowpass;
BitIOInfo* pIO = pContext->m_pIOLP;
Int iModelBits = pContext->m_aModelLP.m_iFlcBits[0];
Int aRLCoeffs[32], iNumNonzero = 0, iIndex = 0;
Int aLaplacianMean[2] = { 0, 0}, *pLM = aLaplacianMean;
Int iChannel, iCBP = 0;
#ifndef ARMOPT_BITIO // ARM opt always uses 32-bit version of getBits
U32 (*getBits)(BitIOInfo* pIO, U32 cBits) = _getBit16;
#endif
CWMIMBInfo * pMBInfo = &pSC->MBInfo;
I32 *aDC[MAX_CHANNELS];
UNREFERENCED_PARAMETER( iMBX );
UNREFERENCED_PARAMETER( iMBYdummy );
readIS_L1(pSC, pIO);
if((pSC->WMISCP.bfBitstreamFormat != SPATIAL) && (pSC->pTile[pSC->cTileColumn].cBitsLP > 0)) // MB-based LP QP index
pMBInfo->iQIndexLP = decodeQPIndex(pIO, pSC->pTile[pSC->cTileColumn].cBitsLP);
// set arrays
for (k = 0; k < (Int) pSC->m_param.cNumChannels; k++) {
aDC[k & 15] = pMBInfo->iBlockDC[k];
}
/** reset adaptive scan totals **/
if (pSC->m_bResetRGITotals) {
int iScale = 2;
int iWeight = iScale * 16;
pScan[0].uTotal = MAXTOTAL;
for (k = 1; k < 16; k++) {
pScan[k].uTotal = iWeight;
iWeight -= iScale;
}
}
/** in raw mode, this can take 6% of the bits in the extreme low rate case!!! **/
if (cf == YUV_420 || cf == YUV_422 || cf == YUV_444) {
int iCountM = pContext->m_iCBPCountMax, iCountZ = pContext->m_iCBPCountZero;
int iMax = iFullPlanes * 4 - 5; /* actually (1 << iNChannels) - 1 **/
if (iCountZ <= 0 || iCountM < 0) {
iCBP = 0;
if (_getBool16 (pIO)) {
iCBP = 1;
k = _getBit16 (pIO, iFullPlanes - 1);
if (k) {
iCBP = k * 2 + _getBit16(pIO, 1);
}
}
if (iCountM < iCountZ)
iCBP = iMax - iCBP;
}
else {
iCBP = _getBit16(pIO, iFullPlanes);
}
iCountM += 1 - 4 * (iCBP == iMax);//(b + c - 2*a);
iCountZ += 1 - 4 * (iCBP == 0);//(a + b - 2*c);
if (iCountM < -8)
iCountM = -8;
else if (iCountM > 7)
iCountM = 7;
pContext->m_iCBPCountMax = iCountM;
if (iCountZ < -8)
iCountZ = -8;
else if (iCountZ > 7)
iCountZ = 7;
pContext->m_iCBPCountZero = iCountZ;
}
else { /** 1 or N channel **/
for (iChannel = 0; iChannel < iChannels; iChannel++)
iCBP |= (getBits (pIO, 1) << iChannel);
}
#ifndef ARMOPT_BITIO // ARM opt always uses 32-bit version of getBits
if (pContext->m_aModelLP.m_iFlcBits[0] > 14 || pContext->m_aModelLP.m_iFlcBits[1] > 14) {
getBits = getBit32;
}
#endif
for (iChannel = 0; iChannel < iFullPlanes; iChannel++) {
PixelI *pCoeffs = aDC[iChannel];
if (iCBP & 1) {
iNumNonzero = DecodeBlock (iChannel > 0, aRLCoeffs, pContext->m_pAHexpt,
CTDC, pIO, 1 + 9 * ((cf == YUV_420) && (iChannel == 1))
+ ((cf == YUV_422) && (iChannel == 1)));
if ((cf == YUV_420 || cf == YUV_422) && iChannel) {
Int aTemp[16]; //14 required, 16 for security
static const Int aRemap[] = { 4, 1,2,3, 5,6,7 };
const Int *pRemap = aRemap + (cf == YUV_420);
const Int iCount = (cf == YUV_420) ? 6 : 14;
(*pLM) += iNumNonzero;
iIndex = 0;
memset (aTemp, 0, sizeof(aTemp));
for (k = 0; k < iNumNonzero; k++) {
iIndex += aRLCoeffs[k * 2];
aTemp[iIndex & 0xf] = aRLCoeffs[k * 2 + 1];
iIndex++;
}
for (k = 0; k < iCount; k++) {
aDC[(k & 1) + 1][pRemap[k >> 1]] = aTemp[k];
}
}
else {
(*pLM) += iNumNonzero;
iIndex = 1;
for (k = 0; k < iNumNonzero; k++) {
iIndex += aRLCoeffs[k * 2];
pCoeffs[pScan[iIndex].uScan] = aRLCoeffs[k * 2 + 1];
pScan[iIndex].uTotal++;
if (pScan[iIndex].uTotal > pScan[iIndex - 1].uTotal) {
CAdaptiveScan cTemp = pScan[iIndex];
pScan[iIndex] = pScan[iIndex - 1];
pScan[iIndex - 1] = cTemp;
}
iIndex++;
}
}
}
if (iModelBits) {
if ((cf == YUV_420 || cf == YUV_422) && iChannel) {
for (k = 1; k < (cf == YUV_420 ? 4 : 8); k++) {
if (aDC[1][k] > 0) {
aDC[1][k] <<= iModelBits;
aDC[1][k] += getBits (pIO, iModelBits);
}
else if (aDC[1][k] < 0) {
aDC[1][k] <<= iModelBits;
aDC[1][k] -= getBits (pIO, iModelBits);
}
else {
aDC[1][k] = getBits (pIO, iModelBits);
if (aDC[1][k] && _getBool16 (pIO))
aDC[1][k] = -aDC[1][k];
}
if (aDC[2][k] > 0) {
aDC[2][k] <<= iModelBits;
aDC[2][k] += getBits (pIO, iModelBits);
}
else if (aDC[2][k] < 0) {
aDC[2][k] <<= iModelBits;
aDC[2][k] -= getBits (pIO, iModelBits);
}
else {
aDC[2][k] = getBits (pIO, iModelBits);
if (aDC[2][k] && _getBool16 (pIO))
aDC[2][k] = -aDC[2][k];
}
}
}
else {
#ifdef WIN32
const Int iMask = (1 << iModelBits) - 1;
#endif // WIN32
for (k = 1; k < 16; k++) {
#ifdef WIN32
if (pCoeffs[k]) {
Int r1 = _rotl(pCoeffs[k], iModelBits);
pCoeffs[k] = (r1 ^ getBits(pIO, iModelBits)) - (r1 & iMask);
}
#else // WIN32
if (pCoeffs[k] > 0) {
pCoeffs[k] <<= iModelBits;
pCoeffs[k] += getBits (pIO, iModelBits);
}
else if (pCoeffs[k] < 0) {
pCoeffs[k] <<= iModelBits;
pCoeffs[k] -= getBits (pIO, iModelBits);
}
#endif // WIN32
else {
//pCoeffs[k] = getBits (pIO, iModelBits);
//if (pCoeffs[k] && _getBool16 (pIO))
// pCoeffs[k] = -pCoeffs[k];
Int r1 = _peekBit16 (pIO, iModelBits + 1);
pCoeffs[k] = ((r1 >> 1) ^ (-(r1 & 1))) + (r1 & 1);
_flushBit16 (pIO, iModelBits + (pCoeffs[k] != 0));
}
}
}
}
pLM = aLaplacianMean + 1;
iModelBits = pContext->m_aModelLP.m_iFlcBits[1];
iCBP >>= 1;
}
UpdateModelMB (cf, iChannels, aLaplacianMean, &(pContext->m_aModelLP));
if (pSC->m_bResetContext) {
AdaptLowpassDec(pContext);
}
return ICERR_OK;
}
/*************************************************************************
8 bit YUV 420 macroblock decode function with 4x4 transform
Index order is as follows:
Y: U: V:
0 1 4 5 16 17 20 21
2 3 6 7 18 19 22 23
8 9 12 13
10 11 14 15
DCAC coefficients stored for 4x4 - offsets (x == no storage)
Y:
x x x [0..3]
x x x [4..7]
x x x [8..11]
[16..19] [20..23] [24..27] [28..31,12..15]
U, V:
x [0..3]
[8..11] [4..7,12..15]
*************************************************************************/
Int DecodeMacroblockDC(CWMImageStrCodec * pSC, CCodingContext *pContext, Int iMBX, Int iMBY)
{
CWMITile * pTile = pSC->pTile + pSC->cTileColumn;
CWMIMBInfo * pMBInfo = &pSC->MBInfo;
const COLORFORMAT cf = pSC->m_param.cfColorFormat;
const Int iChannels = (Int) pSC->m_param.cNumChannels;
BitIOInfo* pIO = pContext->m_pIODC;
Int iIndex, i;
Int aLaplacianMean[2] = { 0, 0}, *pLM = aLaplacianMean;
Int iModelBits = pContext->m_aModelDC.m_iFlcBits[0];
struct CAdaptiveHuffman *pAH;
Int iQDCY, iQDCU, iQDCV;
// const Int iChromaElements = (cf == YUV_420) ? 8 * 8 : ((cf == YUV_422) ? 8 * 16 : 16 * 16);
UNREFERENCED_PARAMETER( iMBX );
UNREFERENCED_PARAMETER( iMBY );
for (i = 0; i < iChannels; i++)
memset (pMBInfo->iBlockDC[i], 0, 16 * sizeof (I32));
readIS_L1(pSC, pIO);
pMBInfo->iQIndexLP = pMBInfo->iQIndexHP = 0;
if(pSC->WMISCP.bfBitstreamFormat == SPATIAL && pSC->WMISCP.sbSubband != SB_DC_ONLY){
if(pTile->cBitsLP > 0) // MB-based LP QP index
pMBInfo->iQIndexLP = decodeQPIndex(pIO, pTile->cBitsLP);
if( pSC->WMISCP.sbSubband != SB_NO_HIGHPASS && pTile->cBitsHP > 0) // MB-based HP QP index
pMBInfo->iQIndexHP = decodeQPIndex(pIO, pTile->cBitsHP);
}
if(pTile->cBitsHP == 0 && pTile->cNumQPHP > 1) // use LP QP
pMBInfo->iQIndexHP = pMBInfo->iQIndexLP;
if (pMBInfo->iQIndexLP >= pTile->cNumQPLP || pMBInfo->iQIndexHP >= pTile->cNumQPHP)
return ICERR_ERROR;
if(cf == Y_ONLY || cf == CMYK || cf == NCOMPONENT) {
for (i = 0; i < iChannels; i++) {
iQDCY = 0;
/** get luminance DC **/
if (_getBool16 (pIO)) {
iQDCY = DecodeSignificantAbsLevel(pContext->m_pAHexpt[3], pIO) - 1;
*pLM += 1;
}
if (iModelBits) {
iQDCY = (iQDCY << iModelBits) | _getBit16(pIO, iModelBits);
}
if (iQDCY && _getBool16 (pIO))
iQDCY = -iQDCY;
pMBInfo->iBlockDC[i][0] = iQDCY;
pLM = aLaplacianMean + 1;
iModelBits = pContext->m_aModelDC.m_iFlcBits[1];
}
}
else {
/** find significant level in 3D **/
pAH = pContext->m_pAHexpt[2];
iIndex = getHuff (pAH->m_hufDecTable, pIO);
iQDCY = iIndex >> 2;
iQDCU = (iIndex >> 1) & 1;
iQDCV = iIndex & 1;
/** get luminance DC **/
if (iQDCY) {
iQDCY = DecodeSignificantAbsLevel(pContext->m_pAHexpt[3], pIO) - 1;
*pLM += 1;
}
if (iModelBits) {
iQDCY = (iQDCY << iModelBits) | _getBit16(pIO, iModelBits);
}
if (iQDCY && _getBool16 (pIO))
iQDCY = -iQDCY;
pMBInfo->iBlockDC[0][0] = iQDCY;
/** get chrominance DC **/
pLM = aLaplacianMean + 1;
iModelBits = pContext->m_aModelDC.m_iFlcBits[1];
if (iQDCU) {
iQDCU = DecodeSignificantAbsLevel(pContext->m_pAHexpt[4], pIO) - 1;
*pLM += 1;
}
if (iModelBits) {
iQDCU = (iQDCU << iModelBits) | _getBit16(pIO, iModelBits);
}
if (iQDCU && _getBool16 (pIO))
iQDCU = -iQDCU;
pMBInfo->iBlockDC[1][0] = iQDCU;
if (iQDCV) {
iQDCV = DecodeSignificantAbsLevel(pContext->m_pAHexpt[4], pIO) - 1;
*pLM += 1;
}
if (iModelBits) {
iQDCV = (iQDCV << iModelBits) | _getBit16(pIO, iModelBits);
}
if (iQDCV && _getBool16 (pIO))
iQDCV = -iQDCV;
pMBInfo->iBlockDC[2][0] = iQDCV;
}
UpdateModelMB (cf, iChannels, aLaplacianMean, &(pContext->m_aModelDC));
if(((!(pSC->WMISCP.bfBitstreamFormat != FREQUENCY || pSC->m_Dparam->cThumbnailScale < 16)) || pSC->WMISCP.sbSubband == SB_DC_ONLY) && pSC->m_bResetContext){
Int kk;
for (kk = 2; kk < 5; kk++) {
if (ICERR_OK != AdaptDecFixed (pContext->m_pAHexpt[kk])) {
return ICERR_ERROR;
}
}
}
return ICERR_OK;
}
/*************************************************************************
DecodeMacroblockHighpass
*************************************************************************/
Int DecodeMacroblockHighpass (CWMImageStrCodec *pSC, CCodingContext *pContext,
Int iMBX, Int iMBY)
{
/** reset adaptive scan totals **/
if (pSC->m_bResetRGITotals) {
int iScale = 2, k;
int iWeight = iScale * 16;
pContext->m_aScanHoriz[0].uTotal = pContext->m_aScanVert[0].uTotal = MAXTOTAL;
for (k = 1; k < 16; k++) {
pContext->m_aScanHoriz[k].uTotal = pContext->m_aScanVert[k].uTotal = iWeight;
iWeight -= iScale;
}
}
if((pSC->WMISCP.bfBitstreamFormat != SPATIAL) && (pSC->pTile[pSC->cTileColumn].cBitsHP > 0)) { // MB-based HP QP index
pSC->MBInfo.iQIndexHP = decodeQPIndex(pContext->m_pIOAC, pSC->pTile[pSC->cTileColumn].cBitsHP);
if (pSC->MBInfo.iQIndexHP >= pSC->pTile[pSC->cTileColumn].cNumQPHP)
goto ErrorExit;
}
else if(pSC->pTile[pSC->cTileColumn].cBitsHP == 0 && pSC->pTile[pSC->cTileColumn].cNumQPHP > 1) // use LP QP
pSC->MBInfo.iQIndexHP = pSC->MBInfo.iQIndexLP;
DecodeCBP (pSC, pContext);
predCBPDec(pSC, pContext);
if (DecodeCoeffs (pSC, pContext, iMBX, iMBY,
pContext->m_pIOAC, pContext->m_pIOFL) != ICERR_OK)
goto ErrorExit;
if (pSC->m_bResetContext) {
AdaptHighpassDec(pContext);
}
return ICERR_OK;
ErrorExit:
return ICERR_ERROR;
}
/*************************************************************************
Adapt
*************************************************************************/
Int AdaptLowpassDec(CCodingContext * pSC)
{
Int kk;
for (kk = 0; kk < CONTEXTX + CTDC; kk++) {
if (ICERR_OK != AdaptDecFixed (pSC->m_pAHexpt[kk])) {
goto ErrorExit;
}
}
return ICERR_OK;
ErrorExit:
return ICERR_ERROR;
}
Int AdaptHighpassDec(CCodingContext * pSC)
{
Int kk;
if (ICERR_OK != AdaptDecFixed (pSC->m_pAdaptHuffCBPCY)) {
goto ErrorExit;
}
if (ICERR_OK != AdaptDecFixed (pSC->m_pAdaptHuffCBPCY1)) {
goto ErrorExit;
}
for (kk = 0; kk < CONTEXTX; kk++) {
if (ICERR_OK != AdaptDecFixed (pSC->m_pAHexpt[kk + CONTEXTX + CTDC])) {
goto ErrorExit;
}
}
return ICERR_OK;
ErrorExit:
return ICERR_ERROR;
}