void DeblockingBSCalc_c (SWelsFuncPtrList* pFunc, SMB* pCurMb, uint8_t uiBS[2][4][4], Mb_Type uiCurMbType,
int32_t iMbStride, int32_t iLeftFlag, int32_t iTopFlag) {
if (iLeftFlag) {
* (uint32_t*)uiBS[0][0] = IS_INTRA ((pCurMb - 1)->uiMbType) ? 0x04040404 : DeblockingBSMarginalMBAvcbase (pCurMb,
pCurMb - 1, 0);
} else {
* (uint32_t*)uiBS[0][0] = 0;
}
if (iTopFlag) {
* (uint32_t*)uiBS[1][0] = IS_INTRA ((pCurMb - iMbStride)->uiMbType) ? 0x04040404 : DeblockingBSMarginalMBAvcbase (
pCurMb, (pCurMb - iMbStride), 1);
} else {
* (uint32_t*)uiBS[1][0] = 0;
}
//SKIP MB_16x16 or others
if (uiCurMbType != MB_TYPE_SKIP) {
pFunc->pfSetNZCZero (pCurMb->pNonZeroCount); // set all none-zero nzc to 1; dbk can be opti!
if (uiCurMbType == MB_TYPE_16x16) {
DeblockingBSInsideMBAvsbase (pCurMb->pNonZeroCount, uiBS, 1);
} else {
DeblockingBSInsideMBNormal (pCurMb, uiBS, pCurMb->pNonZeroCount);
}
} else {
* (uint32_t*)uiBS[0][1] = * (uint32_t*)uiBS[0][2] = * (uint32_t*)uiBS[0][3] =
* (uint32_t*)uiBS[1][1] = * (uint32_t*)uiBS[1][2] = * (uint32_t*)uiBS[1][3] = 0;
}
}
void DeblockingMbAvcbase (SWelsFuncPtrList* pFunc, SMB* pCurMb, SDeblockingFilter* pFilter) {
uint8_t uiBS[2][4][4] = { 0 };
Mb_Type uiCurMbType = pCurMb->uiMbType;
int32_t iMbStride = pFilter->iMbStride;
int32_t iMbX = pCurMb->iMbX;
int32_t iMbY = pCurMb->iMbY;
bool bLeftBsValid[2] = { (iMbX > 0), ((iMbX > 0) && (pCurMb->uiSliceIdc == (pCurMb - 1)->uiSliceIdc))};
bool bTopBsValid[2] = { (iMbY > 0), ((iMbY > 0) && (pCurMb->uiSliceIdc == (pCurMb - iMbStride)->uiSliceIdc))};
int32_t iLeftFlag = bLeftBsValid[pFilter->uiFilterIdc];
int32_t iTopFlag = bTopBsValid[pFilter->uiFilterIdc];
switch (uiCurMbType) {
case MB_TYPE_INTRA4x4:
case MB_TYPE_INTRA16x16:
case MB_TYPE_INTRA_PCM:
DeblockingIntraMb (&pFunc->pfDeblocking, pCurMb, pFilter);
break;
default:
if (iLeftFlag) {
* (uint32_t*)uiBS[0][0] = IS_INTRA ((pCurMb - 1)->uiMbType) ? 0x04040404 : DeblockingBSMarginalMBAvcbase (pCurMb,
pCurMb - 1, 0);
} else {
* (uint32_t*)uiBS[0][0] = 0;
}
if (iTopFlag) {
* (uint32_t*)uiBS[1][0] = IS_INTRA ((pCurMb - iMbStride)->uiMbType) ? 0x04040404 : DeblockingBSMarginalMBAvcbase (
pCurMb, (pCurMb - iMbStride), 1);
} else {
* (uint32_t*)uiBS[1][0] = 0;
}
//SKIP MB_16x16 or others
if (uiCurMbType != MB_TYPE_SKIP) {
pFunc->pfSetNZCZero (pCurMb->pNonZeroCount); // set all none-zero nzc to 1; dbk can be opti!
if (uiCurMbType == MB_TYPE_16x16) {
DeblockingBSInsideMBAvsbase (pCurMb->pNonZeroCount, uiBS, 1);
} else {
DeblockingBSInsideMBNormal (pCurMb, uiBS, pCurMb->pNonZeroCount);
}
} else {
* (uint32_t*)uiBS[0][1] = * (uint32_t*)uiBS[0][2] = * (uint32_t*)uiBS[0][3] =
* (uint32_t*)uiBS[1][1] = * (uint32_t*)uiBS[1][2] = * (uint32_t*)uiBS[1][3] = 0;
}
DeblockingInterMb (&pFunc->pfDeblocking, pCurMb, pFilter, uiBS);
break;
}
}
int32_t CComplexityAnalysis::GetFrameSadExcludeBackground (SPixMap* pSrcPixMap, SPixMap* pRefPixMap) {
int32_t iWidth = pSrcPixMap->sRect.iRectWidth;
int32_t iHeight = pSrcPixMap->sRect.iRectHeight;
int32_t iMbWidth = iWidth >> 4;
int32_t iMbHeight = iHeight >> 4;
int32_t iMbNum = iMbWidth * iMbHeight;
int32_t iMbNumInGom = m_sComplexityAnalysisParam.iMbNumInGom;
int32_t iGomMbNum = (iMbNum + iMbNumInGom - 1) / iMbNumInGom;
int32_t iGomMbStartIndex = 0, iGomMbEndIndex = 0;
uint8_t* pBackgroundMbFlag = (uint8_t*)m_sComplexityAnalysisParam.pBackgroundMbFlag;
uint32_t* uiRefMbType = (uint32_t*)m_sComplexityAnalysisParam.uiRefMbType;
SVAACalcResult* pVaaCalcResults = m_sComplexityAnalysisParam.pCalcResult;
int32_t* pGomForegroundBlockNum = m_sComplexityAnalysisParam.pGomForegroundBlockNum;
uint32_t uiFrameSad = 0;
for (int32_t j = 0; j < iGomMbNum; j ++) {
iGomMbStartIndex = j * iMbNumInGom;
iGomMbEndIndex = WELS_MIN ((j + 1) * iMbNumInGom, iMbNum);
for (int32_t i = iGomMbStartIndex; i < iGomMbEndIndex; i ++) {
if (pBackgroundMbFlag[i] == 0 || IS_INTRA (uiRefMbType[i])) {
pGomForegroundBlockNum[j]++;
uiFrameSad += pVaaCalcResults->pSad8x8[i][0];
uiFrameSad += pVaaCalcResults->pSad8x8[i][1];
uiFrameSad += pVaaCalcResults->pSad8x8[i][2];
uiFrameSad += pVaaCalcResults->pSad8x8[i][3];
}
}
}
return (uiFrameSad);
}
void DeblockingBSCalc_neon (SWelsFuncPtrList* pFunc, SMB* pCurMb, uint8_t uiBS[2][4][4], Mb_Type uiCurMbType,
int32_t iMbStride, int32_t iLeftFlag, int32_t iTopFlag) {
DeblockingBSCalcEnc_neon (pCurMb->pNonZeroCount, pCurMb->sMv, pCurMb->uiNeighborAvail, iMbStride, uiBS);
if (iLeftFlag) {
if (IS_INTRA ((pCurMb - 1)->uiMbType)) {
* (uint32_t*)uiBS[0][0] = 0x04040404;
}
} else {
* (uint32_t*)uiBS[0][0] = 0;
}
if (iTopFlag) {
if (IS_INTRA ((pCurMb - iMbStride)->uiMbType)) {
* (uint32_t*)uiBS[1][0] = 0x04040404;
}
} else {
* (uint32_t*)uiBS[1][0] = 0;
}
}
void CComplexityAnalysis::AnalyzeGomComplexityViaSad (SPixMap* pSrcPixMap, SPixMap* pRefPixMap) {
int32_t iWidth = pSrcPixMap->sRect.iRectWidth;
int32_t iHeight = pSrcPixMap->sRect.iRectHeight;
int32_t iMbWidth = iWidth >> 4;
int32_t iMbHeight = iHeight >> 4;
int32_t iMbNum = iMbWidth * iMbHeight;
int32_t iMbNumInGom = m_sComplexityAnalysisParam.iMbNumInGom;
int32_t iGomMbNum = (iMbNum + iMbNumInGom - 1) / iMbNumInGom;
int32_t iGomMbStartIndex = 0, iGomMbEndIndex = 0, iGomMbRowNum = 0;
int32_t iMbStartIndex = 0, iMbEndIndex = 0;
uint8_t* pBackgroundMbFlag = (uint8_t*)m_sComplexityAnalysisParam.pBackgroundMbFlag;
uint32_t* uiRefMbType = (uint32_t*)m_sComplexityAnalysisParam.uiRefMbType;
SVAACalcResult* pVaaCalcResults = m_sComplexityAnalysisParam.pCalcResult;
int32_t* pGomForegroundBlockNum = (int32_t*)m_sComplexityAnalysisParam.pGomForegroundBlockNum;
int32_t* pGomComplexity = (int32_t*)m_sComplexityAnalysisParam.pGomComplexity;
uint32_t uiGomSad = 0, uiFrameSad = 0;
InitGomSadFunc (m_pfGomSad, m_sComplexityAnalysisParam.iCalcBgd);
for (int32_t j = 0; j < iGomMbNum; j ++) {
uiGomSad = 0;
iGomMbStartIndex = j * iMbNumInGom;
iGomMbEndIndex = WELS_MIN ((j + 1) * iMbNumInGom, iMbNum);
iGomMbRowNum = (iGomMbEndIndex + iMbWidth - 1) / iMbWidth - iGomMbStartIndex / iMbWidth;
iMbStartIndex = iGomMbStartIndex;
iMbEndIndex = WELS_MIN ((iMbStartIndex / iMbWidth + 1) * iMbWidth, iGomMbEndIndex);
do {
for (int32_t i = iMbStartIndex; i < iMbEndIndex; i ++) {
m_pfGomSad (&uiGomSad, pGomForegroundBlockNum + j, pVaaCalcResults->pSad8x8[i], pBackgroundMbFlag[i]
&& !IS_INTRA (uiRefMbType[i]));
}
iMbStartIndex = iMbEndIndex;
iMbEndIndex = WELS_MIN (iMbEndIndex + iMbWidth , iGomMbEndIndex);
} while (--iGomMbRowNum);
pGomComplexity[j] = uiGomSad;
uiFrameSad += pGomComplexity[j];
}
m_sComplexityAnalysisParam.iFrameComplexity = uiFrameSad;
}
static void x264_cabac_mb_type( x264_t *h, x264_cabac_t *cb )
{
const int i_mb_type = h->mb.i_type;
if( h->sh.b_mbaff &&
(!(h->mb.i_mb_y & 1) || IS_SKIP(h->mb.type[h->mb.i_mb_xy - h->mb.i_mb_stride])) )
{
x264_cabac_encode_decision_noup( cb, 70 + h->mb.cache.i_neighbour_interlaced, h->mb.b_interlaced );
}
if( h->sh.i_type == SLICE_TYPE_I )
{
int ctx = 0;
if( (h->mb.i_neighbour & MB_LEFT) && h->mb.i_mb_type_left != I_4x4 )
ctx++;
if( (h->mb.i_neighbour & MB_TOP) && h->mb.i_mb_type_top != I_4x4 )
ctx++;
x264_cabac_mb_type_intra( h, cb, i_mb_type, 3+ctx, 3+3, 3+4, 3+5, 3+6, 3+7 );
}
else if( h->sh.i_type == SLICE_TYPE_P )
{
/* prefix: 14, suffix: 17 */
if( i_mb_type == P_L0 )
{
x264_cabac_encode_decision_noup( cb, 14, 0 );
x264_cabac_encode_decision_noup( cb, 15, h->mb.i_partition != D_16x16 );
x264_cabac_encode_decision_noup( cb, 17-(h->mb.i_partition == D_16x16), h->mb.i_partition == D_16x8 );
}
else if( i_mb_type == P_8x8 )
{
x264_cabac_encode_decision_noup( cb, 14, 0 );
x264_cabac_encode_decision_noup( cb, 15, 0 );
x264_cabac_encode_decision_noup( cb, 16, 1 );
}
else /* intra */
{
/* prefix */
x264_cabac_encode_decision_noup( cb, 14, 1 );
/* suffix */
x264_cabac_mb_type_intra( h, cb, i_mb_type, 17+0, 17+1, 17+2, 17+2, 17+3, 17+3 );
}
}
else //if( h->sh.i_type == SLICE_TYPE_B )
{
int ctx = 0;
if( (h->mb.i_neighbour & MB_LEFT) && h->mb.i_mb_type_left != B_SKIP && h->mb.i_mb_type_left != B_DIRECT )
ctx++;
if( (h->mb.i_neighbour & MB_TOP) && h->mb.i_mb_type_top != B_SKIP && h->mb.i_mb_type_top != B_DIRECT )
ctx++;
if( i_mb_type == B_DIRECT )
{
x264_cabac_encode_decision_noup( cb, 27+ctx, 0 );
return;
}
x264_cabac_encode_decision_noup( cb, 27+ctx, 1 );
if( i_mb_type == B_8x8 )
{
x264_cabac_encode_decision_noup( cb, 27+3, 1 );
x264_cabac_encode_decision_noup( cb, 27+4, 1 );
x264_cabac_encode_decision( cb, 27+5, 1 );
x264_cabac_encode_decision( cb, 27+5, 1 );
x264_cabac_encode_decision_noup( cb, 27+5, 1 );
}
else if( IS_INTRA( i_mb_type ) )
{
/* prefix */
x264_cabac_encode_decision_noup( cb, 27+3, 1 );
x264_cabac_encode_decision_noup( cb, 27+4, 1 );
x264_cabac_encode_decision( cb, 27+5, 1 );
x264_cabac_encode_decision( cb, 27+5, 0 );
x264_cabac_encode_decision( cb, 27+5, 1 );
/* suffix */
x264_cabac_mb_type_intra( h, cb, i_mb_type, 32+0, 32+1, 32+2, 32+2, 32+3, 32+3 );
}
else
{
static const uint8_t i_mb_bits[9*3] =
{
0x31, 0x29, 0x4, /* L0 L0 */
0x35, 0x2d, 0, /* L0 L1 */
0x43, 0x63, 0, /* L0 BI */
0x3d, 0x2f, 0, /* L1 L0 */
0x39, 0x25, 0x6, /* L1 L1 */
0x53, 0x73, 0, /* L1 BI */
0x4b, 0x6b, 0, /* BI L0 */
0x5b, 0x7b, 0, /* BI L1 */
0x47, 0x67, 0x21 /* BI BI */
};
const int idx = (i_mb_type - B_L0_L0) * 3 + (h->mb.i_partition - D_16x8);
int bits = i_mb_bits[idx];
x264_cabac_encode_decision_noup( cb, 27+3, bits&1 );
x264_cabac_encode_decision( cb, 27+5-(bits&1), (bits>>1)&1 ); bits >>= 2;
if( bits != 1 )
{
x264_cabac_encode_decision( cb, 27+5, bits&1 ); bits >>= 1;
x264_cabac_encode_decision( cb, 27+5, bits&1 ); bits >>= 1;
x264_cabac_encode_decision( cb, 27+5, bits&1 ); bits >>= 1;
if( bits != 1 )
x264_cabac_encode_decision_noup( cb, 27+5, bits&1 );
}
}
}
static av_noinline void FUNC(hl_decode_mb)(H264Context *h)
{
MpegEncContext *const s = &h->s;
const int mb_x = s->mb_x;
const int mb_y = s->mb_y;
const int mb_xy = h->mb_xy;
const int mb_type = s->current_picture.f.mb_type[mb_xy];
uint8_t *dest_y, *dest_cb, *dest_cr;
int linesize, uvlinesize /*dct_offset*/;
int i, j;
int *block_offset = &h->block_offset[0];
const int transform_bypass = !SIMPLE && (s->qscale == 0 && h->sps.transform_bypass);
/* is_h264 should always be true if SVQ3 is disabled. */
const int is_h264 = !CONFIG_SVQ3_DECODER || SIMPLE || s->codec_id == AV_CODEC_ID_H264;
void (*idct_add)(uint8_t *dst, int16_t *block, int stride);
const int block_h = 16 >> s->chroma_y_shift;
const int chroma422 = CHROMA422;
dest_y = s->current_picture.f.data[0] + ((mb_x << PIXEL_SHIFT) + mb_y * s->linesize) * 16;
dest_cb = s->current_picture.f.data[1] + (mb_x << PIXEL_SHIFT) * 8 + mb_y * s->uvlinesize * block_h;
dest_cr = s->current_picture.f.data[2] + (mb_x << PIXEL_SHIFT) * 8 + mb_y * s->uvlinesize * block_h;
s->vdsp.prefetch(dest_y + (s->mb_x & 3) * 4 * s->linesize + (64 << PIXEL_SHIFT), s->linesize, 4);
s->vdsp.prefetch(dest_cb + (s->mb_x & 7) * s->uvlinesize + (64 << PIXEL_SHIFT), dest_cr - dest_cb, 2);
h->list_counts[mb_xy] = h->list_count;
if (!SIMPLE && MB_FIELD) {
linesize = h->mb_linesize = s->linesize * 2;
uvlinesize = h->mb_uvlinesize = s->uvlinesize * 2;
block_offset = &h->block_offset[48];
if (mb_y & 1) { // FIXME move out of this function?
dest_y -= s->linesize * 15;
dest_cb -= s->uvlinesize * (block_h - 1);
dest_cr -= s->uvlinesize * (block_h - 1);
}
if (FRAME_MBAFF) {
int list;
for (list = 0; list < h->list_count; list++) {
if (!USES_LIST(mb_type, list))
continue;
if (IS_16X16(mb_type)) {
int8_t *ref = &h->ref_cache[list][scan8[0]];
fill_rectangle(ref, 4, 4, 8, (16 + *ref) ^ (s->mb_y & 1), 1);
} else {
for (i = 0; i < 16; i += 4) {
int ref = h->ref_cache[list][scan8[i]];
if (ref >= 0)
fill_rectangle(&h->ref_cache[list][scan8[i]], 2, 2,
8, (16 + ref) ^ (s->mb_y & 1), 1);
}
}
}
}
} else {
linesize = h->mb_linesize = s->linesize;
uvlinesize = h->mb_uvlinesize = s->uvlinesize;
// dct_offset = s->linesize * 16;
}
if (!SIMPLE && IS_INTRA_PCM(mb_type)) {
const int bit_depth = h->sps.bit_depth_luma;
if (PIXEL_SHIFT) {
int j;
GetBitContext gb;
init_get_bits(&gb, (uint8_t *)h->mb,
ff_h264_mb_sizes[h->sps.chroma_format_idc] * bit_depth);
for (i = 0; i < 16; i++) {
uint16_t *tmp_y = (uint16_t *)(dest_y + i * linesize);
for (j = 0; j < 16; j++)
tmp_y[j] = get_bits(&gb, bit_depth);
}
if (SIMPLE || !CONFIG_GRAY || !(s->flags & CODEC_FLAG_GRAY)) {
if (!h->sps.chroma_format_idc) {
for (i = 0; i < block_h; i++) {
uint16_t *tmp_cb = (uint16_t *)(dest_cb + i * uvlinesize);
uint16_t *tmp_cr = (uint16_t *)(dest_cr + i * uvlinesize);
for (j = 0; j < 8; j++) {
tmp_cb[j] = tmp_cr[j] = 1 << (bit_depth - 1);
}
}
} else {
for (i = 0; i < block_h; i++) {
uint16_t *tmp_cb = (uint16_t *)(dest_cb + i * uvlinesize);
for (j = 0; j < 8; j++)
tmp_cb[j] = get_bits(&gb, bit_depth);
}
for (i = 0; i < block_h; i++) {
uint16_t *tmp_cr = (uint16_t *)(dest_cr + i * uvlinesize);
for (j = 0; j < 8; j++)
tmp_cr[j] = get_bits(&gb, bit_depth);
}
}
}
} else {
for (i = 0; i < 16; i++)
memcpy(dest_y + i * linesize, (uint8_t *)h->mb + i * 16, 16);
if (SIMPLE || !CONFIG_GRAY || !(s->flags & CODEC_FLAG_GRAY)) {
if (!h->sps.chroma_format_idc) {
for (i = 0; i < 8; i++) {
memset(dest_cb + i*uvlinesize, 1 << (bit_depth - 1), 8);
memset(dest_cr + i*uvlinesize, 1 << (bit_depth - 1), 8);
}
} else {
uint8_t *src_cb = (uint8_t *)h->mb + 256;
uint8_t *src_cr = (uint8_t *)h->mb + 256 + block_h * 8;
for (i = 0; i < block_h; i++) {
memcpy(dest_cb + i * uvlinesize, src_cb + i * 8, 8);
memcpy(dest_cr + i * uvlinesize, src_cr + i * 8, 8);
}
}
}
}
} else {
if (IS_INTRA(mb_type)) {
if (h->deblocking_filter)
xchg_mb_border(h, dest_y, dest_cb, dest_cr, linesize,
uvlinesize, 1, 0, SIMPLE, PIXEL_SHIFT);
if (SIMPLE || !CONFIG_GRAY || !(s->flags & CODEC_FLAG_GRAY)) {
h->hpc.pred8x8[h->chroma_pred_mode](dest_cb, uvlinesize);
h->hpc.pred8x8[h->chroma_pred_mode](dest_cr, uvlinesize);
}
hl_decode_mb_predict_luma(h, mb_type, is_h264, SIMPLE,
transform_bypass, PIXEL_SHIFT,
block_offset, linesize, dest_y, 0);
if (h->deblocking_filter)
xchg_mb_border(h, dest_y, dest_cb, dest_cr, linesize,
uvlinesize, 0, 0, SIMPLE, PIXEL_SHIFT);
} else if (is_h264) {
if (chroma422) {
FUNC(hl_motion_422)(h, dest_y, dest_cb, dest_cr,
s->me.qpel_put, s->dsp.put_h264_chroma_pixels_tab,
s->me.qpel_avg, s->dsp.avg_h264_chroma_pixels_tab,
h->h264dsp.weight_h264_pixels_tab,
h->h264dsp.biweight_h264_pixels_tab);
} else {
FUNC(hl_motion_420)(h, dest_y, dest_cb, dest_cr,
s->me.qpel_put, s->dsp.put_h264_chroma_pixels_tab,
s->me.qpel_avg, s->dsp.avg_h264_chroma_pixels_tab,
h->h264dsp.weight_h264_pixels_tab,
h->h264dsp.biweight_h264_pixels_tab);
}
}
hl_decode_mb_idct_luma(h, mb_type, is_h264, SIMPLE, transform_bypass,
PIXEL_SHIFT, block_offset, linesize, dest_y, 0);
if ((SIMPLE || !CONFIG_GRAY || !(s->flags & CODEC_FLAG_GRAY)) &&
(h->cbp & 0x30)) {
uint8_t *dest[2] = { dest_cb, dest_cr };
if (transform_bypass) {
if (IS_INTRA(mb_type) && h->sps.profile_idc == 244 &&
(h->chroma_pred_mode == VERT_PRED8x8 ||
h->chroma_pred_mode == HOR_PRED8x8)) {
h->hpc.pred8x8_add[h->chroma_pred_mode](dest[0],
block_offset + 16,
h->mb + (16 * 16 * 1 << PIXEL_SHIFT),
uvlinesize);
h->hpc.pred8x8_add[h->chroma_pred_mode](dest[1],
block_offset + 32,
h->mb + (16 * 16 * 2 << PIXEL_SHIFT),
uvlinesize);
} else {
idct_add = s->dsp.add_pixels4;
for (j = 1; j < 3; j++) {
for (i = j * 16; i < j * 16 + 4; i++)
if (h->non_zero_count_cache[scan8[i]] ||
dctcoef_get(h->mb, PIXEL_SHIFT, i * 16))
idct_add(dest[j - 1] + block_offset[i],
h->mb + (i * 16 << PIXEL_SHIFT),
uvlinesize);
if (chroma422) {
for (i = j * 16 + 4; i < j * 16 + 8; i++)
if (h->non_zero_count_cache[scan8[i + 4]] ||
dctcoef_get(h->mb, PIXEL_SHIFT, i * 16))
idct_add(dest[j - 1] + block_offset[i + 4],
h->mb + (i * 16 << PIXEL_SHIFT),
uvlinesize);
}
}
}
} else {
if (is_h264) {
int qp[2];
if (chroma422) {
qp[0] = h->chroma_qp[0] + 3;
qp[1] = h->chroma_qp[1] + 3;
} else {
qp[0] = h->chroma_qp[0];
qp[1] = h->chroma_qp[1];
}
if (h->non_zero_count_cache[scan8[CHROMA_DC_BLOCK_INDEX + 0]])
h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + (16 * 16 * 1 << PIXEL_SHIFT),
h->dequant4_coeff[IS_INTRA(mb_type) ? 1 : 4][qp[0]][0]);
if (h->non_zero_count_cache[scan8[CHROMA_DC_BLOCK_INDEX + 1]])
h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + (16 * 16 * 2 << PIXEL_SHIFT),
h->dequant4_coeff[IS_INTRA(mb_type) ? 2 : 5][qp[1]][0]);
h->h264dsp.h264_idct_add8(dest, block_offset,
h->mb, uvlinesize,
h->non_zero_count_cache);
} else if (CONFIG_SVQ3_DECODER) {
h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + 16 * 16 * 1,
h->dequant4_coeff[IS_INTRA(mb_type) ? 1 : 4][h->chroma_qp[0]][0]);
h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + 16 * 16 * 2,
h->dequant4_coeff[IS_INTRA(mb_type) ? 2 : 5][h->chroma_qp[1]][0]);
for (j = 1; j < 3; j++) {
for (i = j * 16; i < j * 16 + 4; i++)
if (h->non_zero_count_cache[scan8[i]] || h->mb[i * 16]) {
uint8_t *const ptr = dest[j - 1] + block_offset[i];
ff_svq3_add_idct_c(ptr, h->mb + i * 16,
uvlinesize,
ff_h264_chroma_qp[0][s->qscale + 12] - 12, 2);
}
}
}
}
}
}
if (h->cbp || IS_INTRA(mb_type)) {
s->dsp.clear_blocks(h->mb);
s->dsp.clear_blocks(h->mb + (24 * 16 << PIXEL_SHIFT));
}
}
static av_noinline void FUNC(hl_decode_mb_444)(H264Context *h)
{
MpegEncContext *const s = &h->s;
const int mb_x = s->mb_x;
const int mb_y = s->mb_y;
const int mb_xy = h->mb_xy;
const int mb_type = s->current_picture.f.mb_type[mb_xy];
uint8_t *dest[3];
int linesize;
int i, j, p;
int *block_offset = &h->block_offset[0];
const int transform_bypass = !SIMPLE && (s->qscale == 0 && h->sps.transform_bypass);
const int plane_count = (SIMPLE || !CONFIG_GRAY || !(s->flags & CODEC_FLAG_GRAY)) ? 3 : 1;
for (p = 0; p < plane_count; p++) {
dest[p] = s->current_picture.f.data[p] +
((mb_x << PIXEL_SHIFT) + mb_y * s->linesize) * 16;
s->vdsp.prefetch(dest[p] + (s->mb_x & 3) * 4 * s->linesize + (64 << PIXEL_SHIFT),
s->linesize, 4);
}
h->list_counts[mb_xy] = h->list_count;
if (!SIMPLE && MB_FIELD) {
linesize = h->mb_linesize = h->mb_uvlinesize = s->linesize * 2;
block_offset = &h->block_offset[48];
if (mb_y & 1) // FIXME move out of this function?
for (p = 0; p < 3; p++)
dest[p] -= s->linesize * 15;
if (FRAME_MBAFF) {
int list;
for (list = 0; list < h->list_count; list++) {
if (!USES_LIST(mb_type, list))
continue;
if (IS_16X16(mb_type)) {
int8_t *ref = &h->ref_cache[list][scan8[0]];
fill_rectangle(ref, 4, 4, 8, (16 + *ref) ^ (s->mb_y & 1), 1);
} else {
for (i = 0; i < 16; i += 4) {
int ref = h->ref_cache[list][scan8[i]];
if (ref >= 0)
fill_rectangle(&h->ref_cache[list][scan8[i]], 2, 2,
8, (16 + ref) ^ (s->mb_y & 1), 1);
}
}
}
}
} else {
linesize = h->mb_linesize = h->mb_uvlinesize = s->linesize;
}
if (!SIMPLE && IS_INTRA_PCM(mb_type)) {
if (PIXEL_SHIFT) {
const int bit_depth = h->sps.bit_depth_luma;
GetBitContext gb;
init_get_bits(&gb, (uint8_t *)h->mb, 768 * bit_depth);
for (p = 0; p < plane_count; p++)
for (i = 0; i < 16; i++) {
uint16_t *tmp = (uint16_t *)(dest[p] + i * linesize);
for (j = 0; j < 16; j++)
tmp[j] = get_bits(&gb, bit_depth);
}
} else {
for (p = 0; p < plane_count; p++)
for (i = 0; i < 16; i++)
memcpy(dest[p] + i * linesize,
(uint8_t *)h->mb + p * 256 + i * 16, 16);
}
} else {
if (IS_INTRA(mb_type)) {
if (h->deblocking_filter)
xchg_mb_border(h, dest[0], dest[1], dest[2], linesize,
linesize, 1, 1, SIMPLE, PIXEL_SHIFT);
for (p = 0; p < plane_count; p++)
hl_decode_mb_predict_luma(h, mb_type, 1, SIMPLE,
transform_bypass, PIXEL_SHIFT,
block_offset, linesize, dest[p], p);
if (h->deblocking_filter)
xchg_mb_border(h, dest[0], dest[1], dest[2], linesize,
linesize, 0, 1, SIMPLE, PIXEL_SHIFT);
} else {
FUNC(hl_motion_444)(h, dest[0], dest[1], dest[2],
s->me.qpel_put, s->dsp.put_h264_chroma_pixels_tab,
s->me.qpel_avg, s->dsp.avg_h264_chroma_pixels_tab,
h->h264dsp.weight_h264_pixels_tab,
h->h264dsp.biweight_h264_pixels_tab);
}
for (p = 0; p < plane_count; p++)
hl_decode_mb_idct_luma(h, mb_type, 1, SIMPLE, transform_bypass,
PIXEL_SHIFT, block_offset, linesize,
dest[p], p);
}
if (h->cbp || IS_INTRA(mb_type)) {
s->dsp.clear_blocks(h->mb);
s->dsp.clear_blocks(h->mb + (24 * 16 << PIXEL_SHIFT));
}
}
/*!
*************************************************************************************
* \brief
* Mode Decision for a macroblock
*************************************************************************************
*/
void encode_one_macroblock_low (Macroblock *currMB)
{
Slice *currSlice = currMB->p_slice;
RDOPTStructure *p_RDO = currSlice->p_RDO;
ImageParameters *p_Img = currMB->p_Img;
InputParameters *p_Inp = currMB->p_Inp;
PicMotionParams *motion = &p_Img->enc_picture->motion;
imgpel ***temp_img; // to temp store the Y data for 8x8 transform
int block, mode, i=0, j;
RD_PARAMS enc_mb;
int bmcost[5] = {INT_MAX};
double rd_cost = 0;
int cost = 0;
int min_cost = INT_MAX, cost_direct=0, have_direct=0;
int intra1 = 0;
int temp_cpb = 0;
Boolean best_transform_flag = FALSE;
int cost8x8_direct = 0;
short islice = (short) (currSlice->slice_type == I_SLICE);
short bslice = (short) (currSlice->slice_type == B_SLICE);
short pslice = (short) ((currSlice->slice_type == P_SLICE) || (currSlice->slice_type == SP_SLICE));
short intra = (short) (islice || (pslice && currMB->mb_y == p_Img->mb_y_upd && p_Img->mb_y_upd!=p_Img->mb_y_intra));
int lambda_mf[3];
Block8x8Info *b8x8info = p_Img->b8x8info;
//int mb_available[3] = { 1, 1, 1};
char **ipredmodes = p_Img->ipredmode;
short *allmvs = (currSlice->slice_type == I_SLICE) ? NULL: currSlice->all_mv[0][0][0][0][0];
int ****i4p; //for non-RD-opt. mode
imgpel **mb_pred = currSlice->mb_pred[0];
Boolean tmp_8x8_flag, tmp_no_mbpart;
BestMode md_best;
Info8x8 best;
init_md_best(&md_best);
// Init best (need to create simple function)
best.pdir = 0;
best.bipred = 0;
best.ref[LIST_0] = 0;
best.ref[LIST_1] = -1;
get_mem3Dpel(&temp_img, 3, MB_BLOCK_SIZE, MB_BLOCK_SIZE);
intra |= RandomIntra (p_Img, currMB->mbAddrX); // Forced Pseudo-Random Intra
//===== Setup Macroblock encoding parameters =====
init_enc_mb_params(currMB, &enc_mb, intra);
if (p_Inp->AdaptiveRounding)
{
reset_adaptive_rounding(p_Img);
}
if (currSlice->MbaffFrameFlag)
{
reset_mb_nz_coeff(p_Img, currMB->mbAddrX);
}
//===== S T O R E C O D I N G S T A T E =====
//---------------------------------------------------
currSlice->store_coding_state (currMB, currSlice->p_RDO->cs_cm);
if (!intra)
{
//===== set direct motion vectors =====
currMB->best_mode = 10; // let us set best_mode to an intra mode to avoid possible bug with RDOQ
if (bslice && enc_mb.valid[0])
{
currSlice->Get_Direct_Motion_Vectors (currMB);
}
if (p_Inp->CtxAdptLagrangeMult == 1)
{
get_initial_mb16x16_cost(currMB);
}
//===== MOTION ESTIMATION FOR 16x16, 16x8, 8x16 BLOCKS =====
for (mode = 1; mode < 4; mode++)
{
best.bipred = 0;
best.mode = (char) mode;
b8x8info->best[mode][0].bipred = 0;
if (enc_mb.valid[mode])
{
for (cost=0, block=0; block<(mode==1?1:2); block++)
{
update_lambda_costs(currMB, &enc_mb, lambda_mf);
PartitionMotionSearch (currMB, mode, block, lambda_mf);
//--- set 4x4 block indizes (for getting MV) ---
j = (block==1 && mode==2 ? 2 : 0);
i = (block==1 && mode==3 ? 2 : 0);
//--- get cost and reference frame for List 0 prediction ---
bmcost[LIST_0] = INT_MAX;
list_prediction_cost(currMB, LIST_0, block, mode, &enc_mb, bmcost, best.ref);
if (bslice)
{
//--- get cost and reference frame for List 1 prediction ---
bmcost[LIST_1] = INT_MAX;
list_prediction_cost(currMB, LIST_1, block, mode, &enc_mb, bmcost, best.ref);
// Compute bipredictive cost between best list 0 and best list 1 references
list_prediction_cost(currMB, BI_PRED, block, mode, &enc_mb, bmcost, best.ref);
// currently Bi predictive ME is only supported for modes 1, 2, 3 and ref 0
if (is_bipred_enabled(p_Inp, mode))
{
list_prediction_cost(currMB, BI_PRED_L0, block, mode, &enc_mb, bmcost, 0);
list_prediction_cost(currMB, BI_PRED_L1, block, mode, &enc_mb, bmcost, 0);
}
else
{
bmcost[BI_PRED_L0] = INT_MAX;
bmcost[BI_PRED_L1] = INT_MAX;
}
// Determine prediction list based on mode cost
determine_prediction_list(bmcost, &best, &cost);
}
else // if (bslice)
{
best.pdir = 0;
cost += bmcost[LIST_0];
}
assign_enc_picture_params(currMB, mode, &best, 2 * block);
//----- set reference frame and direction parameters -----
set_block8x8_info(b8x8info, mode, block, &best);
//--- set reference frames and motion vectors ---
if (mode>1 && block==0)
currSlice->set_ref_and_motion_vectors (currMB, motion, &best, block);
} // for (block=0; block<(mode==1?1:2); block++)
currMB->luma_transform_size_8x8_flag = FALSE;
if (p_Inp->Transform8x8Mode) //for inter rd-off, set 8x8 to do 8x8 transform
{
SetModesAndRefframeForBlocks(currMB, (short) mode);
currMB->luma_transform_size_8x8_flag = (byte) TransformDecision(currMB, -1, &cost);
}
if (cost < min_cost)
{
currMB->best_mode = (short) mode;
min_cost = cost;
best_transform_flag = currMB->luma_transform_size_8x8_flag;
if (p_Inp->CtxAdptLagrangeMult == 1)
{
adjust_mb16x16_cost(currMB, cost);
}
}
} // if (enc_mb.valid[mode])
} // for (mode=1; mode<4; mode++)
if (enc_mb.valid[P8x8])
{
//===== store coding state of macroblock =====
currSlice->store_coding_state (currMB, currSlice->p_RDO->cs_mb);
memset( currSlice->cofAC[0][0][0], 0, 2080 * sizeof(int)); // 4 * 4 * 2 * 65
currMB->valid_8x8 = FALSE;
if (p_Inp->Transform8x8Mode)
{
ResetRD8x8Data(p_Img, p_RDO->tr8x8);
//===========================================================
// Check 8x8 partition with transform size 8x8
//===========================================================
//===== LOOP OVER 8x8 SUB-PARTITIONS (Motion Estimation & Mode Decision) =====
for (cost_direct = 0, block = 0; block < 4; block++)
{
submacroblock_mode_decision_low(currMB, &enc_mb, p_RDO->tr8x8, p_RDO->cofAC8x8ts[block],
&have_direct, block, &cost_direct, &cost, &cost8x8_direct, 1);
set_subblock8x8_info(b8x8info, P8x8, block, p_RDO->tr8x8);
}
currMB->luma_transform_size_8x8_flag = FALSE; //switch to 4x4 transform size
}// if (p_Inp->Transform8x8Mode)
if (p_Inp->Transform8x8Mode != 2)
{
ResetRD8x8Data(p_Img, p_RDO->tr4x4);
//=================================================================
// Check 8x8, 8x4, 4x8 and 4x4 partitions with transform size 4x4
//=================================================================
//===== LOOP OVER 8x8 SUB-PARTITIONS (Motion Estimation & Mode Decision) =====
for (cost_direct = 0, block=0; block<4; block++)
{
submacroblock_mode_decision_low(currMB, &enc_mb, p_RDO->tr4x4, p_RDO->coefAC8x8[block],
&have_direct, block, &cost_direct, &cost, &cost8x8_direct, 0);
set_subblock8x8_info(b8x8info, P8x8, block, p_RDO->tr4x4);
}
}// if (p_Inp->Transform8x8Mode != 2)
if (p_Inp->RCEnable)
rc_store_diff(currSlice->diffy, &p_Img->pCurImg[currMB->opix_y], currMB->pix_x, mb_pred);
//check cost for P8x8 for non-rdopt mode
if (((p_Inp->Transform8x8Mode < 2) && (p_RDO->tr4x4->mb_p8x8_cost < min_cost)) ||
((p_Inp->Transform8x8Mode > 0) && (p_RDO->tr8x8->mb_p8x8_cost < min_cost)))
{
currMB->best_mode = P8x8;
if (p_Inp->Transform8x8Mode == 2)
{
min_cost = p_RDO->tr8x8->mb_p8x8_cost;
currMB->luma_transform_size_8x8_flag = TRUE;
}
else if (p_Inp->Transform8x8Mode)
{
if (p_RDO->tr8x8->mb_p8x8_cost < p_RDO->tr4x4->mb_p8x8_cost)
{
min_cost = p_RDO->tr8x8->mb_p8x8_cost;
currMB->luma_transform_size_8x8_flag = TRUE;
}
else if(p_RDO->tr4x4->mb_p8x8_cost < p_RDO->tr8x8->mb_p8x8_cost)
{
min_cost = p_RDO->tr4x4->mb_p8x8_cost;
currMB->luma_transform_size_8x8_flag = FALSE;
}
else
{
if (GetBestTransformP8x8(currMB) == 0)
{
min_cost = p_RDO->tr4x4->mb_p8x8_cost;
currMB->luma_transform_size_8x8_flag = FALSE;
}
else
{
min_cost = p_RDO->tr8x8->mb_p8x8_cost;
currMB->luma_transform_size_8x8_flag = TRUE;
}
}
}
else
{
min_cost = p_RDO->tr4x4->mb_p8x8_cost;
currMB->luma_transform_size_8x8_flag = FALSE;
}
}// if ((p_RDO->tr4x4->mb_p8x8_cost < min_cost || p_RDO->tr8x8->mb_p8x8_cost < min_cost))
p_Img->giRDOpt_B8OnlyFlag = FALSE;
}
// Find a motion vector for the Skip mode
if(pslice)
FindSkipModeMotionVector (currMB);
}
else // if (!intra)
{
min_cost = INT_MAX;
}
//========= C H O O S E B E S T M A C R O B L O C K M O D E =========
//-------------------------------------------------------------------------
tmp_8x8_flag = currMB->luma_transform_size_8x8_flag; //save 8x8_flag
tmp_no_mbpart = currMB->NoMbPartLessThan8x8Flag; //save no-part-less
if ((p_Img->yuv_format != YUV400) && (p_Img->yuv_format != YUV444))
// precompute all chroma intra prediction modes
intra_chroma_prediction(currMB, NULL, NULL, NULL);
if (enc_mb.valid[0] && bslice) // check DIRECT MODE
{
if(have_direct)
{
switch(p_Inp->Transform8x8Mode)
{
case 1: // Mixture of 8x8 & 4x4 transform
cost = ((cost8x8_direct < cost_direct) || !(enc_mb.valid[5] && enc_mb.valid[6] && enc_mb.valid[7]))
? cost8x8_direct : cost_direct;
break;
case 2: // 8x8 Transform only
cost = cost8x8_direct;
break;
default: // 4x4 Transform only
cost = cost_direct;
break;
}
}
else
{ //!have_direct
cost = GetDirectCostMB (currMB);
}
if (cost!=INT_MAX)
{
cost -= (int)floor(16 * enc_mb.lambda_md + 0.4999);
}
if (cost <= min_cost)
{
if(p_Img->active_sps->direct_8x8_inference_flag && p_Inp->Transform8x8Mode)
{
if(p_Inp->Transform8x8Mode==2)
currMB->luma_transform_size_8x8_flag = TRUE;
else
{
if(cost8x8_direct < cost_direct)
currMB->luma_transform_size_8x8_flag = TRUE;
else
currMB->luma_transform_size_8x8_flag = FALSE;
}
}
else
currMB->luma_transform_size_8x8_flag = FALSE;
//Rate control
if (p_Inp->RCEnable)
rc_store_diff(currSlice->diffy, &p_Img->pCurImg[currMB->opix_y], currMB->pix_x, mb_pred);
min_cost = cost;
currMB->best_mode = 0;
tmp_8x8_flag = currMB->luma_transform_size_8x8_flag;
}
else
{
currMB->luma_transform_size_8x8_flag = (byte) tmp_8x8_flag; // restore if not best
currMB->NoMbPartLessThan8x8Flag = (byte) tmp_no_mbpart; // restore if not best
}
}
currMB->min_rdcost = (double) min_cost;
if (enc_mb.valid[I8MB]) // check INTRA8x8
{
currMB->luma_transform_size_8x8_flag = TRUE; // at this point cost will ALWAYS be less than min_cost
currMB->mb_type = currMB->ar_mode = I8MB;
temp_cpb = Mode_Decision_for_Intra8x8Macroblock (currMB, enc_mb.lambda_md, &rd_cost);
if (rd_cost <= currMB->min_rdcost) //HYU_NOTE. bug fix. 08/15/07
{
currMB->cbp = temp_cpb;
if (p_Img->P444_joined)
{
currSlice->curr_cbp[0] = currSlice->cmp_cbp[1];
currSlice->curr_cbp[1] = currSlice->cmp_cbp[2];
}
if(enc_mb.valid[I4MB])
{
//coeffs
if (p_Inp->Transform8x8Mode != 2)
{
i4p = p_RDO->cofAC;
p_RDO->cofAC = currSlice->cofAC;
currSlice->cofAC = i4p;
}
}
copy_image_data_16x16(temp_img[0], &p_Img->enc_picture->imgY[currMB->pix_y], 0, currMB->pix_x);
if (p_Img->P444_joined)
{
copy_image_data_16x16(temp_img[1], &p_Img->enc_picture->imgUV[0][currMB->pix_y], 0, currMB->pix_x);
copy_image_data_16x16(temp_img[2], &p_Img->enc_picture->imgUV[1][currMB->pix_y], 0, currMB->pix_x);
}
//Rate control
if (p_Inp->RCEnable)
rc_store_diff(currSlice->diffy, &p_Img->pCurImg[currMB->opix_y], currMB->pix_x, mb_pred);
currMB->min_rdcost = rd_cost;
currMB->best_mode = I8MB;
tmp_8x8_flag = currMB->luma_transform_size_8x8_flag;
}
else
{
currMB->luma_transform_size_8x8_flag = (byte) tmp_8x8_flag; // restore if not best
if (p_Img->P444_joined)
{
currMB->cbp |= currSlice->curr_cbp[0];
currMB->cbp |= currSlice->curr_cbp[1];
currSlice->cmp_cbp[1] = currMB->cbp;
currSlice->cmp_cbp[2] = currMB->cbp;
}
}
}
if (enc_mb.valid[I4MB]) // check INTRA4x4
{
currMB->luma_transform_size_8x8_flag = FALSE;
currMB->mb_type = currMB->ar_mode = I4MB;
temp_cpb = Mode_Decision_for_Intra4x4Macroblock (currMB, enc_mb.lambda_md, &rd_cost);
if (rd_cost <= currMB->min_rdcost)
{
currMB->cbp = temp_cpb;
//Rate control
if (p_Inp->RCEnable)
rc_store_diff(currSlice->diffy, &p_Img->pCurImg[currMB->opix_y], currMB->pix_x, mb_pred);
currMB->min_rdcost = rd_cost;
currMB->best_mode = I4MB;
tmp_8x8_flag = currMB->luma_transform_size_8x8_flag;
}
else
{
currMB->luma_transform_size_8x8_flag = (byte) tmp_8x8_flag; // restore if not best
if (p_Img->P444_joined)
{
currMB->cbp |= currSlice->curr_cbp[0];
currMB->cbp |= currSlice->curr_cbp[1];
currSlice->cmp_cbp[1] = currMB->cbp;
currSlice->cmp_cbp[2] = currMB->cbp;
}
//coeffs
i4p = p_RDO->cofAC;
p_RDO->cofAC = currSlice->cofAC;
currSlice->cofAC=i4p;
}
}
if (enc_mb.valid[I16MB]) // check INTRA16x16
{
currMB->luma_transform_size_8x8_flag = FALSE;
intrapred_16x16 (currMB, PLANE_Y);
if (p_Img->P444_joined)
{
select_plane(p_Img, PLANE_U);
intrapred_16x16 (currMB, PLANE_U);
select_plane(p_Img, PLANE_V);
intrapred_16x16 (currMB, PLANE_V);
select_plane(p_Img, PLANE_Y);
}
rd_cost = currSlice->find_sad_16x16 (currMB);
if (rd_cost < currMB->min_rdcost)
{
//Rate control
if (p_Inp->RCEnable)
rc_store_diff(currSlice->diffy, &p_Img->pCurImg[currMB->opix_y], currMB->pix_x, currSlice->mpr_16x16[0][(short) currMB->i16mode]);
currMB->best_mode = I16MB;
currMB->min_rdcost = rd_cost;
currMB->cbp = currMB->trans_16x16 (currMB, PLANE_Y);
if (p_Img->P444_joined)
{
select_plane(p_Img, PLANE_U);
currSlice->cmp_cbp[1] = currMB->trans_16x16(currMB, PLANE_U);
select_plane(p_Img, PLANE_V);
currSlice->cmp_cbp[2] = currMB->trans_16x16(currMB, PLANE_V);
select_plane(p_Img, PLANE_Y);
currMB->cbp |= currSlice->cmp_cbp[1];
currMB->cbp |= currSlice->cmp_cbp[2];
currSlice->cmp_cbp[1] = currMB->cbp;
currSlice->cmp_cbp[2] = currMB->cbp;
}
}
else
{
currMB->luma_transform_size_8x8_flag = (byte) tmp_8x8_flag; // restore
currMB->NoMbPartLessThan8x8Flag = (byte) tmp_no_mbpart; // restore
}
}
intra1 = IS_INTRA(currMB);
//===== S E T F I N A L M A C R O B L O C K P A R A M E T E R S ======
//---------------------------------------------------------------------------
{
//===== set parameters for chosen mode =====
SetModesAndRefframeForBlocks (currMB, currMB->best_mode);
if (currMB->best_mode == P8x8)
{
if (currMB->luma_transform_size_8x8_flag && (p_RDO->tr8x8->cbp8x8 == 0) && p_Inp->Transform8x8Mode != 2)
currMB->luma_transform_size_8x8_flag = FALSE;
SetCoeffAndReconstruction8x8 (currMB);
memset(currMB->intra_pred_modes, DC_PRED, MB_BLOCK_PARTITIONS * sizeof(char));
for (j = currMB->block_y; j < currMB->block_y + BLOCK_MULTIPLE; j++)
memset(&ipredmodes[j][currMB->block_x], DC_PRED, BLOCK_MULTIPLE * sizeof(char));
}
else
{
//===== set parameters for chosen mode =====
if (currMB->best_mode == I8MB)
{
memcpy(currMB->intra_pred_modes,currMB->intra_pred_modes8x8, MB_BLOCK_PARTITIONS * sizeof(char));
for(j = currMB->block_y; j < currMB->block_y + BLOCK_MULTIPLE; j++)
memcpy(&p_Img->ipredmode[j][currMB->block_x],&p_Img->ipredmode8x8[j][currMB->block_x], BLOCK_MULTIPLE * sizeof(char));
//--- restore reconstruction for 8x8 transform ---
copy_image_data_16x16(&p_Img->enc_picture->imgY[currMB->pix_y], temp_img[0], currMB->pix_x, 0);
if (p_Img->P444_joined)
{
copy_image_data_16x16(&p_Img->enc_picture->imgUV[0][currMB->pix_y], temp_img[1], currMB->pix_x, 0);
copy_image_data_16x16(&p_Img->enc_picture->imgUV[1][currMB->pix_y], temp_img[2], currMB->pix_x, 0);
}
}
if ((currMB->best_mode!=I4MB)&&(currMB->best_mode != I8MB))
{
memset(currMB->intra_pred_modes,DC_PRED, MB_BLOCK_PARTITIONS * sizeof(char));
for(j = currMB->block_y; j < currMB->block_y + BLOCK_MULTIPLE; j++)
memset(&ipredmodes[j][currMB->block_x],DC_PRED, BLOCK_MULTIPLE * sizeof(char));
currMB->ar_mode = currMB->best_mode;
if (currMB->best_mode!=I16MB)
{
if((currMB->best_mode >= 1) && (currMB->best_mode <= 3))
currMB->luma_transform_size_8x8_flag = (byte) best_transform_flag;
if (currSlice->P444_joined)
{
luma_residual_coding_p444(currMB);
if((currMB->cbp==0 && currSlice->cmp_cbp[1] == 0 && currSlice->cmp_cbp[2] == 0) &&(currMB->best_mode == 0))
currMB->luma_transform_size_8x8_flag = FALSE;
}
else
{
luma_residual_coding(currMB);
if((currMB->cbp==0)&&(currMB->best_mode == 0))
currMB->luma_transform_size_8x8_flag = FALSE;
}
//Rate control
if (p_Inp->RCEnable)
rc_store_diff(currSlice->diffy, &p_Img->pCurImg[currMB->opix_y], currMB->pix_x, mb_pred);
}
}
}
//check luma cbp for transform size flag
if (((currMB->cbp&15) == 0) && currMB->mb_type != I4MB && currMB->mb_type != I8MB)
currMB->luma_transform_size_8x8_flag = FALSE;
// precompute all chroma intra prediction modes
if ((p_Img->yuv_format != YUV400) && (p_Img->yuv_format != YUV444))
intra_chroma_prediction(currMB, NULL, NULL, NULL);
currMB->i16offset = 0;
if ((p_Img->yuv_format != YUV400) && (p_Img->yuv_format != YUV444))
chroma_residual_coding (currMB);
if (currMB->best_mode == I16MB)
{
currMB->i16offset = I16Offset (currMB->cbp, currMB->i16mode);
}
currSlice->SetMotionVectorsMB (currMB, motion);
//===== check for SKIP mode =====
if(p_Img->P444_joined)
{
if ((pslice) && currMB->best_mode == 1 && currMB->cbp==0 && currSlice->cmp_cbp[1] == 0 && currSlice->cmp_cbp[2] == 0 &&
motion->ref_idx[LIST_0][currMB->block_y][currMB->block_x] == 0 &&
motion->mv [LIST_0][currMB->block_y][currMB->block_x][0] == allmvs[0] &&
motion->mv [LIST_0][currMB->block_y][currMB->block_x][1] == allmvs[1])
{
currMB->mb_type = currMB->b8x8[0].mode = currMB->b8x8[1].mode = currMB->b8x8[2].mode = currMB->b8x8[3].mode = 0;
currMB->luma_transform_size_8x8_flag = FALSE;
}
}
else if ((pslice) && currMB->best_mode == 1 && currMB->cbp==0 &&
motion->ref_idx[LIST_0][currMB->block_y][currMB->block_x] == 0 &&
motion->mv [LIST_0][currMB->block_y][currMB->block_x][0] == allmvs[0] &&
motion->mv [LIST_0][currMB->block_y][currMB->block_x][1] == allmvs[1])
{
currMB->mb_type = currMB->b8x8[0].mode = currMB->b8x8[1].mode = currMB->b8x8[2].mode = currMB->b8x8[3].mode = 0;
currMB->luma_transform_size_8x8_flag = FALSE;
}
if (currSlice->MbaffFrameFlag || (p_Inp->UseRDOQuant && currSlice->RDOQ_QP_Num > 1))
set_mbaff_parameters(currMB);
}
// Rate control
if(p_Inp->RCEnable && p_Inp->RCUpdateMode <= MAX_RC_MODE)
rc_store_mad(currMB);
//===== Decide if this MB will restrict the reference frames =====
if (p_Inp->RestrictRef)
update_refresh_map(currMB, intra, intra1);
/*update adaptive rounding offset p_Inp*/
if (p_Img->AdaptiveRounding)
{
update_offset_params(currMB, currMB->best_mode, currMB->luma_transform_size_8x8_flag);
}
free_mem3Dpel(temp_img);
}
/* FIXME: B-type MBs not handled yet properly */
void x264_visualize_show( x264_t *h )
{
int mb_xy;
static const stringlist_t mb_types[] = {
/* Block types marked as NULL will not be drawn */
{ I_4x4 , "red" },
{ I_8x8 , "#ff5640" },
{ I_16x16 , "#ff8060" },
{ I_PCM , "violet" },
{ P_L0 , "SlateBlue" },
{ P_8x8 , "blue" },
{ P_SKIP , "green" },
{ B_DIRECT, "yellow" },
{ B_L0_L0 , "yellow" },
{ B_L0_L1 , "yellow" },
{ B_L0_BI , "yellow" },
{ B_L1_L0 , "yellow" },
{ B_L1_L1 , "yellow" },
{ B_L1_BI , "yellow" },
{ B_BI_L0 , "yellow" },
{ B_BI_L1 , "yellow" },
{ B_BI_BI , "yellow" },
{ B_8x8 , "yellow" },
{ B_SKIP , "yellow" },
};
static const int waitkey = 1; /* Wait for enter after each frame */
static const int drawbox = 1; /* Draw box around each block */
static const int borders = 0; /* Display extrapolated borders outside frame */
static const int zoom = 2; /* Zoom factor */
static const int pad = 32;
uint8_t *const frame = h->fdec->plane[0];
const int width = h->param.i_width;
const int height = h->param.i_height;
const int stride = h->fdec->i_stride[0];
if (borders) {
disp_gray_zoom(0, frame - pad*stride - pad, width+2*pad, height+2*pad, stride, "fdec", zoom);
} else {
disp_gray_zoom(0, frame, width, height, stride, "fdec", zoom);
}
for( mb_xy = 0; mb_xy < h->sps->i_mb_width * h->sps->i_mb_height; mb_xy++ )
{
visualize_t *const v = (visualize_t*)h->visualize + mb_xy;
const int mb_y = mb_xy / h->sps->i_mb_width;
const int mb_x = mb_xy % h->sps->i_mb_width;
char *const col = GET_STRING(mb_types, v->i_type);
int x = mb_x*16*zoom;
int y = mb_y*16*zoom;
int l = 0;
unsigned int i, j;
if (col==NULL) continue;
if (borders) {
x += pad*zoom;
y += pad*zoom;
}
disp_setcolor(col);
if (drawbox) disp_rect(0, x, y, x+16*zoom-1, y+16*zoom-1);
if (v->i_type==P_L0 || v->i_type==P_8x8 || v->i_type==P_SKIP) {
/* Predicted (inter) mode, with motion vector */
if (v->i_partition==D_16x16 || v->i_type==P_SKIP) {
mv(x+8*zoom, y+8*zoom, v->mv[l][0][0], v->ref[l][0][0], zoom, col);
}
if (v->i_partition==D_16x8) {
if (drawbox) disp_rect(0, x, y, x+16*zoom, y+8*zoom);
mv(x+8*zoom, y+4*zoom, v->mv[l][0][0], v->ref[l][0][0], zoom, col);
if (drawbox) disp_rect(0, x, y+8*zoom, x+16*zoom, y+16*zoom);
mv(x+8*zoom, y+12*zoom, v->mv[l][2][0], v->ref[l][2][0], zoom, col);
}
if (v->i_partition==D_8x16) {
if (drawbox) disp_rect(0, x, y, x+8*zoom, y+16*zoom);
mv(x+4*zoom, y+8*zoom, v->mv[l][0][0], v->ref[l][0][0], zoom, col);
if (drawbox) disp_rect(0, x+8*zoom, y, x+16*zoom, y+16*zoom);
mv(x+12*zoom, y+8*zoom, v->mv[l][0][2], v->ref[l][0][2], zoom, col);
}
if (v->i_partition==D_8x8) {
for (i=0; i<2; i++) for (j=0; j<2; j++) {
int sp = v->i_sub_partition[i*2+j];
const int x0 = x + j*8*zoom;
const int y0 = y + i*8*zoom;
l = x264_mb_partition_listX_table[0][sp] ? 0 : 1; /* FIXME: not tested if this works */
if (IS_SUB8x8(sp)) {
if (drawbox) disp_rect(0, x0, y0, x0+8*zoom, y0+8*zoom);
mv(x0+4*zoom, y0+4*zoom, v->mv[l][2*i][2*j], v->ref[l][2*i][2*j], zoom, col);
}
if (IS_SUB8x4(sp)) {
if (drawbox) disp_rect(0, x0, y0, x0+8*zoom, y0+4*zoom);
if (drawbox) disp_rect(0, x0, y0+4*zoom, x0+8*zoom, y0+8*zoom);
mv(x0+4*zoom, y0+2*zoom, v->mv[l][2*i][2*j], v->ref[l][2*i][2*j], zoom, col);
mv(x0+4*zoom, y0+6*zoom, v->mv[l][2*i+1][2*j], v->ref[l][2*i+1][2*j], zoom, col);
}
if (IS_SUB4x8(sp)) {
if (drawbox) disp_rect(0, x0, y0, x0+4*zoom, y0+8*zoom);
if (drawbox) disp_rect(0, x0+4*zoom, y0, x0+8*zoom, y0+8*zoom);
mv(x0+2*zoom, y0+4*zoom, v->mv[l][2*i][2*j], v->ref[l][2*i][2*j], zoom, col);
mv(x0+6*zoom, y0+4*zoom, v->mv[l][2*i][2*j+1], v->ref[l][2*i][2*j+1], zoom, col);
}
if (IS_SUB4x4(sp)) {
if (drawbox) disp_rect(0, x0, y0, x0+4*zoom, y0+4*zoom);
if (drawbox) disp_rect(0, x0+4*zoom, y0, x0+8*zoom, y0+4*zoom);
if (drawbox) disp_rect(0, x0, y0+4*zoom, x0+4*zoom, y0+8*zoom);
if (drawbox) disp_rect(0, x0+4*zoom, y0+4*zoom, x0+8*zoom, y0+8*zoom);
mv(x0+2*zoom, y0+2*zoom, v->mv[l][2*i][2*j], v->ref[l][2*i][2*j], zoom, col);
mv(x0+6*zoom, y0+2*zoom, v->mv[l][2*i][2*j+1], v->ref[l][2*i][2*j+1], zoom, col);
mv(x0+2*zoom, y0+6*zoom, v->mv[l][2*i+1][2*j], v->ref[l][2*i+1][2*j], zoom, col);
mv(x0+6*zoom, y0+6*zoom, v->mv[l][2*i+1][2*j+1], v->ref[l][2*i+1][2*j+1], zoom, col);
}
}
}
}
if (IS_INTRA(v->i_type) || v->i_type==I_PCM) {
/* Intra coded */
if (v->i_type==I_16x16) {
switch (v->i_intra16x16_pred_mode) {
case I_PRED_16x16_V:
disp_line(0, x+2*zoom, y+2*zoom, x+14*zoom, y+2*zoom);
break;
case I_PRED_16x16_H:
disp_line(0, x+2*zoom, y+2*zoom, x+2*zoom, y+14*zoom);
break;
case I_PRED_16x16_DC:
case I_PRED_16x16_DC_LEFT:
case I_PRED_16x16_DC_TOP:
case I_PRED_16x16_DC_128:
disp_line(0, x+2*zoom, y+2*zoom, x+14*zoom, y+2*zoom);
disp_line(0, x+2*zoom, y+2*zoom, x+2*zoom, y+14*zoom);
break;
case I_PRED_16x16_P:
disp_line(0, x+2*zoom, y+2*zoom, x+8*zoom, y+8*zoom);
break;
}
}
if (v->i_type==I_4x4 || v->i_type==I_8x8) {
const int di = v->i_type==I_8x8 ? 2 : 1;
const int zoom2 = zoom * di;
for (i=0; i<4; i+=di) for (j=0; j<4; j+=di) {
const int x0 = x + j*4*zoom;
const int y0 = y + i*4*zoom;
if (drawbox) disp_rect(0, x0, y0, x0+4*zoom2, y0+4*zoom2);
switch (v->intra4x4_pred_mode[i][j]) {
case I_PRED_4x4_V: /* Vertical */
disp_line(0, x0+0*zoom2, y0+1*zoom2, x0+4*zoom2, y0+1*zoom2);
break;
case I_PRED_4x4_H: /* Horizontal */
disp_line(0, x0+1*zoom2, y0+0*zoom2, x0+1*zoom2, y0+4*zoom2);
break;
case I_PRED_4x4_DC: /* DC, average from top and left sides */
case I_PRED_4x4_DC_LEFT:
case I_PRED_4x4_DC_TOP:
case I_PRED_4x4_DC_128:
disp_line(0, x0+1*zoom2, y0+1*zoom2, x0+4*zoom2, y0+1*zoom2);
disp_line(0, x0+1*zoom2, y0+1*zoom2, x0+1*zoom2, y0+4*zoom2);
break;
case I_PRED_4x4_DDL: /* Topright-bottomleft */
disp_line(0, x0+0*zoom2, y0+0*zoom2, x0+4*zoom2, y0+4*zoom2);
break;
case I_PRED_4x4_DDR: /* Topleft-bottomright */
disp_line(0, x0+0*zoom2, y0+4*zoom2, x0+4*zoom2, y0+0*zoom2);
break;
case I_PRED_4x4_VR: /* Mix of topleft-bottomright and vertical */
disp_line(0, x0+0*zoom2, y0+2*zoom2, x0+4*zoom2, y0+1*zoom2);
break;
case I_PRED_4x4_HD: /* Mix of topleft-bottomright and horizontal */
disp_line(0, x0+2*zoom2, y0+0*zoom2, x0+1*zoom2, y0+4*zoom2);
break;
case I_PRED_4x4_VL: /* Mix of topright-bottomleft and vertical */
disp_line(0, x0+0*zoom2, y0+1*zoom2, x0+4*zoom2, y0+2*zoom2);
break;
case I_PRED_4x4_HU: /* Mix of topright-bottomleft and horizontal */
disp_line(0, x0+1*zoom2, y0+0*zoom2, x0+2*zoom2, y0+4*zoom2);
break;
}
}
}
}
}
disp_sync();
if (waitkey) getchar();
}
static void x264_cabac_mb_type( x264_t *h, x264_cabac_t *cb )
{
const int i_mb_type = h->mb.i_type;
if( h->sh.i_type == SLICE_TYPE_I )
{
int ctx = 0;
if( h->mb.i_mb_type_left >= 0 && h->mb.i_mb_type_left != I_4x4 )
{
ctx++;
}
if( h->mb.i_mb_type_top >= 0 && h->mb.i_mb_type_top != I_4x4 )
{
ctx++;
}
x264_cabac_mb_type_intra( h, cb, i_mb_type, 3+ctx, 3+3, 3+4, 3+5, 3+6, 3+7 );
}
else if( h->sh.i_type == SLICE_TYPE_P )
{
/* prefix: 14, suffix: 17 */
if( i_mb_type == P_L0 )
{
if( h->mb.i_partition == D_16x16 )
{
x264_cabac_encode_decision( cb, 14, 0 );
x264_cabac_encode_decision( cb, 15, 0 );
x264_cabac_encode_decision( cb, 16, 0 );
}
else if( h->mb.i_partition == D_16x8 )
{
x264_cabac_encode_decision( cb, 14, 0 );
x264_cabac_encode_decision( cb, 15, 1 );
x264_cabac_encode_decision( cb, 17, 1 );
}
else if( h->mb.i_partition == D_8x16 )
{
x264_cabac_encode_decision( cb, 14, 0 );
x264_cabac_encode_decision( cb, 15, 1 );
x264_cabac_encode_decision( cb, 17, 0 );
}
}
else if( i_mb_type == P_8x8 )
{
x264_cabac_encode_decision( cb, 14, 0 );
x264_cabac_encode_decision( cb, 15, 0 );
x264_cabac_encode_decision( cb, 16, 1 );
}
else /* intra */
{
/* prefix */
x264_cabac_encode_decision( cb, 14, 1 );
/* suffix */
x264_cabac_mb_type_intra( h, cb, i_mb_type, 17+0, 17+1, 17+2, 17+2, 17+3, 17+3 );
}
}
else if( h->sh.i_type == SLICE_TYPE_B )
{
int ctx = 0;
if( h->mb.i_mb_type_left >= 0 && h->mb.i_mb_type_left != B_SKIP && h->mb.i_mb_type_left != B_DIRECT )
{
ctx++;
}
if( h->mb.i_mb_type_top >= 0 && h->mb.i_mb_type_top != B_SKIP && h->mb.i_mb_type_top != B_DIRECT )
{
ctx++;
}
if( i_mb_type == B_DIRECT )
{
x264_cabac_encode_decision( cb, 27+ctx, 0 );
}
else if( i_mb_type == B_8x8 )
{
x264_cabac_encode_decision( cb, 27+ctx, 1 );
x264_cabac_encode_decision( cb, 27+3, 1 );
x264_cabac_encode_decision( cb, 27+4, 1 );
x264_cabac_encode_decision( cb, 27+5, 1 );
x264_cabac_encode_decision( cb, 27+5, 1 );
x264_cabac_encode_decision( cb, 27+5, 1 );
}
else if( IS_INTRA( i_mb_type ) )
{
/* prefix */
x264_cabac_encode_decision( cb, 27+ctx, 1 );
x264_cabac_encode_decision( cb, 27+3, 1 );
x264_cabac_encode_decision( cb, 27+4, 1 );
x264_cabac_encode_decision( cb, 27+5, 1 );
x264_cabac_encode_decision( cb, 27+5, 0 );
x264_cabac_encode_decision( cb, 27+5, 1 );
/* suffix */
x264_cabac_mb_type_intra( h, cb, i_mb_type, 32+0, 32+1, 32+2, 32+2, 32+3, 32+3 );
}
else
{
static const int i_mb_len[21] =
{
3, 6, 6, /* L0 L0 */
3, 6, 6, /* L1 L1 */
6, 7, 7, /* BI BI */
6, 6, /* L0 L1 */
6, 6, /* L1 L0 */
7, 7, /* L0 BI */
7, 7, /* L1 BI */
7, 7, /* BI L0 */
7, 7, /* BI L1 */
};
static const int i_mb_bits[21][7] =
{
{ 1, 0, 0, }, { 1, 1, 0, 0, 0, 1, }, { 1, 1, 0, 0, 1, 0, }, /* L0 L0 */
{ 1, 0, 1, }, { 1, 1, 0, 0, 1, 1, }, { 1, 1, 0, 1, 0, 0, }, /* L1 L1 */
{ 1, 1, 0, 0, 0, 0 ,}, { 1, 1, 1, 1, 0, 0 , 0 }, { 1, 1, 1, 1, 0, 0 , 1 },/* BI BI */
{ 1, 1, 0, 1, 0, 1, }, { 1, 1, 0, 1, 1, 0, }, /* L0 L1 */
{ 1, 1, 0, 1, 1, 1, }, { 1, 1, 1, 1, 1, 0, }, /* L1 L0 */
{ 1, 1, 1, 0, 0, 0, 0 }, { 1, 1, 1, 0, 0, 0, 1 }, /* L0 BI */
{ 1, 1, 1, 0, 0, 1, 0 }, { 1, 1, 1, 0, 0, 1, 1 }, /* L1 BI */
{ 1, 1, 1, 0, 1, 0, 0 }, { 1, 1, 1, 0, 1, 0, 1 }, /* BI L0 */
{ 1, 1, 1, 0, 1, 1, 0 }, { 1, 1, 1, 0, 1, 1, 1 } /* BI L1 */
};
const int i_partition = h->mb.i_partition;
int idx = 0;
int i;
switch( i_mb_type )
{
/* D_16x16, D_16x8, D_8x16 */
case B_BI_BI: idx += 3;
case B_L1_L1: idx += 3;
case B_L0_L0:
if( i_partition == D_16x8 )
idx += 1;
else if( i_partition == D_8x16 )
idx += 2;
break;
/* D_16x8, D_8x16 */
case B_BI_L1: idx += 2;
case B_BI_L0: idx += 2;
case B_L1_BI: idx += 2;
case B_L0_BI: idx += 2;
case B_L1_L0: idx += 2;
case B_L0_L1:
idx += 3*3;
if( i_partition == D_8x16 )
idx++;
break;
default:
x264_log(h, X264_LOG_ERROR, "error in B mb type\n" );
return;
}
x264_cabac_encode_decision( cb, 27+ctx, i_mb_bits[idx][0] );
x264_cabac_encode_decision( cb, 27+3, i_mb_bits[idx][1] );
x264_cabac_encode_decision( cb, 27+(i_mb_bits[idx][1] != 0 ? 4 : 5), i_mb_bits[idx][2] );
for( i = 3; i < i_mb_len[idx]; i++ )
{
x264_cabac_encode_decision( cb, 27+5, i_mb_bits[idx][i] );
}
}
}
else
{
x264_log(h, X264_LOG_ERROR, "unknown SLICE_TYPE unsupported in x264_macroblock_write_cabac\n" );
}
}
/*!
*************************************************************************************
* \brief
* Mode Decision for a macroblock with error resilience
*************************************************************************************
*/
void encode_one_macroblock_highloss (Macroblock *currMB)
{
Slice *currSlice = currMB->p_Slice;
RDOPTStructure *p_RDO = currSlice->p_RDO;
VideoParameters *p_Vid = currMB->p_Vid;
InputParameters *p_Inp = currMB->p_Inp;
PicMotionParams **motion = p_Vid->enc_picture->mv_info;
int max_index = 9;
int rerun, block, index, mode, i, j;
RD_PARAMS enc_mb;
distblk bmcost[5] = {DISTBLK_MAX};
distblk cost=0;
distblk min_cost = DISTBLK_MAX;
int intra1 = 0;
int mb_available[3];
short bslice = (short) (currSlice->slice_type == B_SLICE);
short pslice = (short) ((currSlice->slice_type == P_SLICE) || (currSlice->slice_type == SP_SLICE));
short intra = (short) ((currSlice->slice_type == I_SLICE) || (currSlice->slice_type == SI_SLICE) || (pslice && currMB->mb_y == p_Vid->mb_y_upd && p_Vid->mb_y_upd != p_Vid->mb_y_intra));
int lambda_mf[3];
short runs = (short) ((p_Inp->RestrictRef==1 && (pslice || (bslice && p_Vid->nal_reference_idc>0))) ? 2 : 1);
imgpel **mb_pred = currSlice->mb_pred[0];
Block8x8Info *b8x8info = p_Vid->b8x8info;
char chroma_pred_mode_range[2];
short inter_skip = 0;
BestMode md_best;
Info8x8 best;
init_md_best(&md_best);
// Init best (need to create simple function)
best.pdir = 0;
best.bipred = 0;
best.ref[LIST_0] = 0;
best.ref[LIST_1] = -1;
intra |= RandomIntra (p_Vid, currMB->mbAddrX); // Forced Pseudo-Random Intra
//===== Setup Macroblock encoding parameters =====
init_enc_mb_params(currMB, &enc_mb, intra);
// Perform multiple encodings if rdopt with losses is enabled
for (rerun=0; rerun<runs; rerun++)
{
if (runs==2)
p_Inp->rdopt= (rerun==0) ? 1 : 3;
if (p_Inp->AdaptiveRounding)
{
reset_adaptive_rounding(p_Vid);
}
if (currSlice->mb_aff_frame_flag)
{
reset_mb_nz_coeff(p_Vid, currMB->mbAddrX);
}
//===== S T O R E C O D I N G S T A T E =====
//---------------------------------------------------
currSlice->store_coding_state (currMB, currSlice->p_RDO->cs_cm);
if (!intra)
{
//===== set skip/direct motion vectors =====
if (enc_mb.valid[0])
{
if (bslice)
currSlice->Get_Direct_Motion_Vectors (currMB);
else
FindSkipModeMotionVector (currMB);
}
if (p_Inp->CtxAdptLagrangeMult == 1)
{
get_initial_mb16x16_cost(currMB);
}
//===== MOTION ESTIMATION FOR 16x16, 16x8, 8x16 BLOCKS =====
for (mode = 1; mode < 4; mode++)
{
best.mode = (char) mode;
best.bipred = 0;
b8x8info->best[mode][0].bipred = 0;
if (enc_mb.valid[mode])
{
for (cost=0, block=0; block<(mode==1?1:2); block++)
{
update_lambda_costs(currMB, &enc_mb, lambda_mf);
PartitionMotionSearch (currMB, mode, block, lambda_mf);
//--- set 4x4 block indices (for getting MV) ---
j = (block==1 && mode==2 ? 2 : 0);
i = (block==1 && mode==3 ? 2 : 0);
//--- get cost and reference frame for List 0 prediction ---
bmcost[LIST_0] = DISTBLK_MAX;
list_prediction_cost(currMB, LIST_0, block, mode, &enc_mb, bmcost, best.ref);
if (bslice)
{
//--- get cost and reference frame for List 1 prediction ---
bmcost[LIST_1] = DISTBLK_MAX;
list_prediction_cost(currMB, LIST_1, block, mode, &enc_mb, bmcost, best.ref);
// Compute bipredictive cost between best list 0 and best list 1 references
list_prediction_cost(currMB, BI_PRED, block, mode, &enc_mb, bmcost, best.ref);
// currently Bi predictive ME is only supported for modes 1, 2, 3 and ref 0
if (is_bipred_enabled(p_Vid, mode))
{
get_bipred_cost(currMB, mode, block, i, j, &best, &enc_mb, bmcost);
}
else
{
bmcost[BI_PRED_L0] = DISTBLK_MAX;
bmcost[BI_PRED_L1] = DISTBLK_MAX;
}
// Determine prediction list based on mode cost
determine_prediction_list(bmcost, &best, &cost);
}
else // if (bslice)
{
best.pdir = 0;
cost += bmcost[LIST_0];
}
assign_enc_picture_params(currMB, mode, &best, 2 * block);
//----- set reference frame and direction parameters -----
set_block8x8_info(b8x8info, mode, block, &best);
//--- set reference frames and motion vectors ---
if (mode>1 && block==0)
currSlice->set_ref_and_motion_vectors (currMB, motion, &best, block);
} // for (block=0; block<(mode==1?1:2); block++)
if (cost < min_cost)
{
md_best.mode = (byte) mode;
md_best.cost = cost;
currMB->best_mode = (short) mode;
min_cost = cost;
if (p_Inp->CtxAdptLagrangeMult == 1)
{
adjust_mb16x16_cost(currMB, cost);
}
}
} // if (enc_mb.valid[mode])
} // for (mode=1; mode<4; mode++)
if (enc_mb.valid[P8x8])
{
currMB->valid_8x8 = FALSE;
if (p_Inp->Transform8x8Mode)
{
ResetRD8x8Data(p_Vid, p_RDO->tr8x8);
currMB->luma_transform_size_8x8_flag = TRUE; //switch to 8x8 transform size
//===========================================================
// Check 8x8 partition with transform size 8x8
//===========================================================
//===== LOOP OVER 8x8 SUB-PARTITIONS (Motion Estimation & Mode Decision) =====
for (block = 0; block < 4; block++)
{
currSlice->submacroblock_mode_decision(currMB, &enc_mb, p_RDO->tr8x8, p_RDO->cofAC8x8ts[block], block, &cost);
if(!currMB->valid_8x8)
break;
set_subblock8x8_info(b8x8info, P8x8, block, p_RDO->tr8x8);
}
}// if (p_Inp->Transform8x8Mode)
currMB->valid_4x4 = FALSE;
if (p_Inp->Transform8x8Mode != 2)
{
currMB->luma_transform_size_8x8_flag = FALSE; //switch to 8x8 transform size
ResetRD8x8Data(p_Vid, p_RDO->tr4x4);
//=================================================================
// Check 8x8, 8x4, 4x8 and 4x4 partitions with transform size 4x4
//=================================================================
//===== LOOP OVER 8x8 SUB-PARTITIONS (Motion Estimation & Mode Decision) =====
for (block = 0; block < 4; block++)
{
currSlice->submacroblock_mode_decision(currMB, &enc_mb, p_RDO->tr4x4, p_RDO->coefAC8x8[block], block, &cost);
if(!currMB->valid_4x4)
break;
set_subblock8x8_info(b8x8info, P8x8, block, p_RDO->tr4x4);
}
}// if (p_Inp->Transform8x8Mode != 2)
if (p_Inp->RCEnable)
rc_store_diff(currSlice->diffy, &p_Vid->pCurImg[currMB->opix_y], currMB->pix_x, mb_pred);
p_Vid->giRDOpt_B8OnlyFlag = FALSE;
}
}
else // if (!intra)
{
min_cost = DISTBLK_MAX;
}
// Set Chroma mode
set_chroma_pred_mode(currMB, enc_mb, mb_available, chroma_pred_mode_range);
//========= C H O O S E B E S T M A C R O B L O C K M O D E =========
//-------------------------------------------------------------------------
for (currMB->c_ipred_mode = chroma_pred_mode_range[0]; currMB->c_ipred_mode<=chroma_pred_mode_range[1]; currMB->c_ipred_mode++)
{
// bypass if c_ipred_mode is not allowed
if ( (p_Vid->yuv_format != YUV400) &&
( ((!intra || !p_Inp->IntraDisableInterOnly) && p_Inp->ChromaIntraDisable == 1 && currMB->c_ipred_mode!=DC_PRED_8)
|| (currMB->c_ipred_mode == VERT_PRED_8 && !mb_available[0])
|| (currMB->c_ipred_mode == HOR_PRED_8 && !mb_available[1])
|| (currMB->c_ipred_mode == PLANE_8 && (!mb_available[1] || !mb_available[0] || !mb_available[2]))))
continue;
//===== GET BEST MACROBLOCK MODE =====
for (index=0; index < max_index; index++)
{
mode = mb_mode_table[index];
if (enc_mb.valid[mode])
{
if (p_Vid->yuv_format != YUV400)
{
currMB->i16mode = 0;
}
// Skip intra modes in inter slices if best mode is inter <P8x8 with cbp equal to 0
if (currSlice->P444_joined)
{
if (p_Inp->SkipIntraInInterSlices && !intra && mode >= I16MB
&& currMB->best_mode <=3 && currMB->best_cbp == 0 && currSlice->cmp_cbp[1] == 0 && currSlice->cmp_cbp[2] == 0 && (currMB->min_rdcost < weighted_cost(enc_mb.lambda_mdfp,5)))
continue;
}
else
{
if (p_Inp->SkipIntraInInterSlices)
{
if (!intra && mode >= I4MB)
{
if (currMB->best_mode <=3 && currMB->best_cbp == 0 && (currMB->min_rdcost < weighted_cost(enc_mb.lambda_mdfp, 5)))
{
continue;
}
else if (currMB->best_mode == 0 && (currMB->min_rdcost < weighted_cost(enc_mb.lambda_mdfp,6)))
{
continue;
}
}
}
}
compute_mode_RD_cost(currMB, &enc_mb, (short) mode, &inter_skip);
}
}// for (index=0; index<max_index; index++)
}// for (currMB->c_ipred_mode=DC_PRED_8; currMB->c_ipred_mode<=chroma_pred_mode_range[1]; currMB->c_ipred_mode++)
restore_nz_coeff(currMB);
if (rerun==0)
intra1 = IS_INTRA(currMB);
} // for (rerun=0; rerun<runs; rerun++)
//===== S E T F I N A L M A C R O B L O C K P A R A M E T E R S ======
//---------------------------------------------------------------------------
update_qp_cbp_tmp(currMB, p_RDO->cbp);
currSlice->set_stored_mb_parameters (currMB);
// Rate control
if(p_Inp->RCEnable && p_Inp->RCUpdateMode <= MAX_RC_MODE)
rc_store_mad(currMB);
//===== Decide if this MB will restrict the reference frames =====
if (p_Inp->RestrictRef)
update_refresh_map(currMB, intra, intra1);
}
/*****************************************************************************
* x264_macroblock_analyse:
*****************************************************************************/
void dull_macroblock_analyse_P_BEST( x264_t *h )
{
x264_mb_analysis_t analysis;
int i_cost = COST_MAX;
int i;
dull_mb_analyse_init_P( h, &analysis );
/*--------------------------- Do the analysis ---------------------------*/
//{ macroblock_analyse_P //{ macroblock_analyse_P //{ macroblock_analyse_P //{ macroblock_analyse_P
{
int b_skip = 0;
analysis.b_try_skip = 0;
if( b_skip )
{
h->mb.i_type = P_SKIP;
h->mb.i_partition = D_16x16;
assert( h->mb.cache.pskip_mv[1] <= h->mb.mv_max_spel[1] || h->i_thread_frames == 1 );
/* Set up MVs for future predictors */
for( i = 0; i < h->mb.pic.i_fref[0]; i++ )
M32( h->mb.mvr[0][i][h->mb.i_mb_xy] ) = 0;
}
else
{
const unsigned int flags = h->param.analyse.inter;
int i_type;
int i_partition;
x264_mb_analyse_load_costs( h, &analysis );
dull_mb_analyse_inter_p16x16_2( h, &analysis );
if( h->mb.i_type == P_SKIP )
{
for( i = 1; i < h->mb.pic.i_fref[0]; i++ )
M32( h->mb.mvr[0][i][h->mb.i_mb_xy] ) = 0;
return;
}
if( flags & X264_ANALYSE_PSUB16x16 )
{
if( h->param.analyse.b_mixed_references )
x264_mb_analyse_inter_p8x8_mixed_ref( h, &analysis );
else
dull_mb_analyse_inter_p8x8_2( h, &analysis );
}
/* Select best inter mode */
i_type = P_L0;
i_partition = D_16x16;
i_cost = analysis.l0.me16x16.cost;
if( ( flags & X264_ANALYSE_PSUB16x16 ) &&
analysis.l0.i_cost8x8 < analysis.l0.me16x16.cost )
{
i_type = P_8x8;
i_partition = D_8x8;
i_cost = analysis.l0.i_cost8x8;
/* Do sub 8x8 */
if( flags & X264_ANALYSE_PSUB8x8 )
{
for( i = 0; i < 4; i++ )
{
x264_mb_analyse_inter_p4x4( h, &analysis, i );
if( analysis.l0.i_cost4x4[i] < analysis.l0.me8x8[i].cost )
{
int i_cost8x8 = analysis.l0.i_cost4x4[i];
h->mb.i_sub_partition[i] = D_L0_4x4;
i_cost += i_cost8x8 - analysis.l0.me8x8[i].cost;
}
x264_mb_cache_mv_p8x8( h, &analysis, i );
}
analysis.l0.i_cost8x8 = i_cost;
}
}
h->mb.i_partition = i_partition;
/* refine qpel */
//FIXME mb_type costs?
if( analysis.i_mbrd || !h->mb.i_subpel_refine )
{
/* refine later */
}
else if( i_partition == D_16x16 )
{
x264_me_refine_qpel( h, &analysis.l0.me16x16 );
i_cost = analysis.l0.me16x16.cost;
}
else if( i_partition == D_16x8 )
{
x264_me_refine_qpel( h, &analysis.l0.me16x8[0] );
x264_me_refine_qpel( h, &analysis.l0.me16x8[1] );
i_cost = analysis.l0.me16x8[0].cost + analysis.l0.me16x8[1].cost;
}
else if( i_partition == D_8x16 )
{
x264_me_refine_qpel( h, &analysis.l0.me8x16[0] );
x264_me_refine_qpel( h, &analysis.l0.me8x16[1] );
i_cost = analysis.l0.me8x16[0].cost + analysis.l0.me8x16[1].cost;
}
else if( i_partition == D_8x8 )
{
int i8x8;
i_cost = 0;
for( i8x8 = 0; i8x8 < 4; i8x8++ )
{
switch( h->mb.i_sub_partition[i8x8] )
{
case D_L0_8x8:
x264_me_refine_qpel( h, &analysis.l0.me8x8[i8x8] );
i_cost += analysis.l0.me8x8[i8x8].cost;
break;
case D_L0_8x4:
x264_me_refine_qpel( h, &analysis.l0.me8x4[i8x8][0] );
x264_me_refine_qpel( h, &analysis.l0.me8x4[i8x8][1] );
i_cost += analysis.l0.me8x4[i8x8][0].cost +
analysis.l0.me8x4[i8x8][1].cost;
break;
case D_L0_4x8:
x264_me_refine_qpel( h, &analysis.l0.me4x8[i8x8][0] );
x264_me_refine_qpel( h, &analysis.l0.me4x8[i8x8][1] );
i_cost += analysis.l0.me4x8[i8x8][0].cost +
analysis.l0.me4x8[i8x8][1].cost;
break;
case D_L0_4x4:
x264_me_refine_qpel( h, &analysis.l0.me4x4[i8x8][0] );
x264_me_refine_qpel( h, &analysis.l0.me4x4[i8x8][1] );
x264_me_refine_qpel( h, &analysis.l0.me4x4[i8x8][2] );
x264_me_refine_qpel( h, &analysis.l0.me4x4[i8x8][3] );
i_cost += analysis.l0.me4x4[i8x8][0].cost +
analysis.l0.me4x4[i8x8][1].cost +
analysis.l0.me4x4[i8x8][2].cost +
analysis.l0.me4x4[i8x8][3].cost;
break;
default:
x264_log( h, X264_LOG_ERROR, "internal error (!8x8 && !4x4)\n" );
break;
}
}
}
if( h->mb.b_chroma_me )
{
x264_mb_analyse_intra_chroma( h, &analysis );
x264_mb_analyse_intra( h, &analysis, i_cost - analysis.i_satd_i8x8chroma );
analysis.i_satd_i16x16 += analysis.i_satd_i8x8chroma;
analysis.i_satd_i8x8 += analysis.i_satd_i8x8chroma;
analysis.i_satd_i4x4 += analysis.i_satd_i8x8chroma;
}
else
x264_mb_analyse_intra( h, &analysis, i_cost );
COPY2_IF_LT( i_cost, analysis.i_satd_i16x16, i_type, I_16x16 );
COPY2_IF_LT( i_cost, analysis.i_satd_i8x8, i_type, I_8x8 );
COPY2_IF_LT( i_cost, analysis.i_satd_i4x4, i_type, I_4x4 );
h->mb.i_type = i_type;
if( analysis.i_mbrd >= 2 && h->mb.i_type != I_PCM )
{
if( IS_INTRA( h->mb.i_type ) )
{
x264_intra_rd_refine( h, &analysis );
}
else if( i_partition == D_16x16 )
{
x264_macroblock_cache_ref( h, 0, 0, 4, 4, 0, analysis.l0.me16x16.i_ref );
analysis.l0.me16x16.cost = i_cost;
x264_me_refine_qpel_rd( h, &analysis.l0.me16x16, analysis.i_lambda2, 0, 0 );
}
else if( i_partition == D_16x8 )
{
h->mb.i_sub_partition[0] = h->mb.i_sub_partition[1] =
h->mb.i_sub_partition[2] = h->mb.i_sub_partition[3] = D_L0_8x8;
x264_macroblock_cache_ref( h, 0, 0, 4, 2, 0, analysis.l0.me16x8[0].i_ref );
x264_macroblock_cache_ref( h, 0, 2, 4, 2, 0, analysis.l0.me16x8[1].i_ref );
x264_me_refine_qpel_rd( h, &analysis.l0.me16x8[0], analysis.i_lambda2, 0, 0 );
x264_me_refine_qpel_rd( h, &analysis.l0.me16x8[1], analysis.i_lambda2, 8, 0 );
}
else if( i_partition == D_8x16 )
{
h->mb.i_sub_partition[0] = h->mb.i_sub_partition[1] =
h->mb.i_sub_partition[2] = h->mb.i_sub_partition[3] = D_L0_8x8;
x264_macroblock_cache_ref( h, 0, 0, 2, 4, 0, analysis.l0.me8x16[0].i_ref );
x264_macroblock_cache_ref( h, 2, 0, 2, 4, 0, analysis.l0.me8x16[1].i_ref );
x264_me_refine_qpel_rd( h, &analysis.l0.me8x16[0], analysis.i_lambda2, 0, 0 );
x264_me_refine_qpel_rd( h, &analysis.l0.me8x16[1], analysis.i_lambda2, 4, 0 );
}
else if( i_partition == D_8x8 )
{
int i8x8;
x264_analyse_update_cache( h, &analysis );
for( i8x8 = 0; i8x8 < 4; i8x8++ )
{
if( h->mb.i_sub_partition[i8x8] == D_L0_8x8 )
{
x264_me_refine_qpel_rd( h, &analysis.l0.me8x8[i8x8], analysis.i_lambda2, i8x8*4, 0 );
}
else if( h->mb.i_sub_partition[i8x8] == D_L0_8x4 )
{
x264_me_refine_qpel_rd( h, &analysis.l0.me8x4[i8x8][0], analysis.i_lambda2, i8x8*4+0, 0 );
x264_me_refine_qpel_rd( h, &analysis.l0.me8x4[i8x8][1], analysis.i_lambda2, i8x8*4+2, 0 );
}
else if( h->mb.i_sub_partition[i8x8] == D_L0_4x8 )
{
x264_me_refine_qpel_rd( h, &analysis.l0.me4x8[i8x8][0], analysis.i_lambda2, i8x8*4+0, 0 );
x264_me_refine_qpel_rd( h, &analysis.l0.me4x8[i8x8][1], analysis.i_lambda2, i8x8*4+1, 0 );
}
else if( h->mb.i_sub_partition[i8x8] == D_L0_4x4 )
{
x264_me_refine_qpel_rd( h, &analysis.l0.me4x4[i8x8][0], analysis.i_lambda2, i8x8*4+0, 0 );
x264_me_refine_qpel_rd( h, &analysis.l0.me4x4[i8x8][1], analysis.i_lambda2, i8x8*4+1, 0 );
x264_me_refine_qpel_rd( h, &analysis.l0.me4x4[i8x8][2], analysis.i_lambda2, i8x8*4+2, 0 );
x264_me_refine_qpel_rd( h, &analysis.l0.me4x4[i8x8][3], analysis.i_lambda2, i8x8*4+3, 0 );
}
}
}
}
}
}
//} macroblock_analyse_P //} macroblock_analyse_P //} macroblock_analyse_P //} macroblock_analyse_P
dull_analyse_update_cache_P( h, &analysis );
}
__tcsm1_main int main() {
int c;
S32I2M(xr16,3);
c = i_la(_gp);
int fifo_rp = 0;
unsigned int XCHGtmp;
H264_Frame_GlbARGs *dFRM = (H264_Frame_GlbARGs *)TCSM1_FRM_ARGS;
H264_MB_DecARGs *dMB_aux = (H264_MB_DecARGs *)TCSM1_MBARGS_BUF;
H264_MB_DecARGs *dMB2 = (H264_MB_DecARGs *)TCSM1_MBARGS_BUF2;
H264_AUX_T *AUX_T = (H264_AUX_T *)TCSM1_AUX_T;
MDMA_DesNode *MDMA1_TRAN = (MDMA_DesNode *)TCSM1_GP1_TRAN_CHAIN;
MDMA_DesNode *MDMA1_ARG = (MDMA_DesNode *)TCSM1_GP1_ARG_CHAIN;
H264_MB_DecARGs *dMBsrc;
fifo_wp = (int *)TCSM1_FIFO_WP;
dMBsrc = (H264_MB_DecARGs *)(dFRM->dMB_baseaddr_aux);
int mb_num;
int mb_start;
int total_mbs;
mb_start = dFRM->first_mb_in_frame;
total_mbs = dFRM->mb_width * dFRM->mb_height;
int i;
uint16_t *intra_pred4x4_top_ptr[2];
intra_pred4x4_top_ptr[0]=intra_pred4x4_top[0];
intra_pred4x4_top_ptr[1]=intra_pred4x4_top[1];
uint16_t *intra_pred4x4_left_ptr[2];
intra_pred4x4_left_ptr[0]=intra_pred4x4_left[0];
intra_pred4x4_left_ptr[1]=intra_pred4x4_left[1];
uint8_t mb_x_d1, mb_y_d1;
uint8_t mb_x_d2, mb_y_d2;
mb_x_d1 = mb_y_d1 = mb_x_d2 = mb_y_d2 =0;
AUX_PMON_CREAT(mc);
AUX_PMON_CREAT(idct);
AUX_PMON_CREAT(dblk);
AUX_PMON_CREAT(intra);
AUX_PMON_CREAT(mdma);
AUX_PMON_CREAT(sync);
do{
}while(fifo_rp == *fifo_wp);
fifo_rp++;
MDMA1_ARG->TSA = get_phy_addr_aux((uint32_t)dMBsrc);
MDMA1_ARG->TDA = (uint32_t)dMB_aux;
MDMA1_ARG->STRD = MDMA_STRD(64, 64);
MDMA1_ARG->UNIT = MDMA_UNIT(1,64,(sizeof(H264_MB_DecARGs)));
SET_MDMA1_DHA((uint32_t)TCSM1_GP1_ARG_CHAIN);
MDMA1_RUN();
dMBsrc++;
POLLING_MDMA1_END();
char bakup_x0 = dFRM->mb_width-1;
char bakup_x1 = 0;
int mb_type_last= 0;
volatile unsigned int *infar = (unsigned int *)TCSM1_H264_DBLK_INFAR;
uint8_t non_zero_count_cache_luma_last[16];
for ( mb_num = mb_start; (mb_num < total_mbs+2); mb_num ++ ) {
AUX_PMON_ON(sync);
do{
}while(fifo_rp == *fifo_wp);
fifo_rp++;
int gp1_tran_start = 0;
int gp1_tran_len = sizeof(H264_MB_DecARGs);
if (dMB_aux->next_mb_no_weight_flag) {
gp1_tran_start = (53 << 2);
gp1_tran_len -= (53 << 2);
}
if (dMB_aux->next_mb_skip_flag) {
gp1_tran_len -= (192 << 2);
}
AUX_PMON_OFF(sync);
AUX_PMON_ON(mdma);
AUX_PMON_OFF(mdma);
AUX_PMON_ON(intra);
MDMA1_ARG->TSA = (get_phy_addr_aux((uint32_t)dMBsrc) + gp1_tran_start);
MDMA1_ARG->TDA = ((uint32_t)dMB2 + gp1_tran_start);
MDMA1_ARG->STRD = MDMA_STRD(64, 64);
MDMA1_ARG->UNIT = MDMA_UNIT(1,64,gp1_tran_len);
dMBsrc++;
POLLING_MDMA1_END(); //ensure curr dblked MB has been trans out
SET_MDMA1_DHA((uint32_t)TCSM1_GP1_ARG_CHAIN);
MDMA1_RUN();
int mb_x= dMB_aux->mb_x;
int mb_y= dMB_aux->mb_y;
int mb_type= dMB_aux->mb_type;
if (dMB_aux->curr_mb_skip_flag)
for(i=0; i<24; i++)
dMB_aux->mb[i*16] = 0;
AUX_T->mc_des_dirty=0;
if(IS_INTRA(mb_type_last)){
// chroma predict
Intra_pred_chroma(dMB_aux->chroma_pred_mode, AUX_T->h264_urecon[1],
AUX_T->h264_urecon[0] + MB_CHROM_WIDTH, TCSM1_BOTTOM_U_13PIX+4);
Intra_pred_chroma(dMB_aux->chroma_pred_mode, AUX_T->h264_vrecon[1],
AUX_T->h264_vrecon[0] + MB_CHROM_WIDTH, TCSM1_BOTTOM_V_13PIX+4);
// luma predict
if(IS_INTRA4x4(mb_type_last)){
if(IS_8x8DCT(mb_type_last)){
for(i=0; i<16; i+=4){
uint8_t * src_left = (uint32_t)(intra_pred4x4_left_ptr[1][i]) | 0xF4000000;
uint8_t * src_top = (uint32_t)(intra_pred4x4_top_ptr[1][i]) | 0xF4000000;
uint8_t * src_topleft;
if ( i==8 )
src_topleft = src_left - RECON_BUF_STRIDE - 1;
else src_topleft = src_top - 1;
uint8_t * ptr= AUX_T->h264_yrecon[1] + MB_offset_recon[i];
int dir= dMB_aux->intra4x4_pred_mode_cache[ i ];
int nnz = non_zero_count_cache_luma_last[i];
Intra_pred_luma_8x8l(dir,ptr,(dMB_aux->topleft_samples_available<<i)&0x8000,
(dMB_aux->topright_samples_available<<i)&0x4000, src_left, src_top, src_topleft, RECON_BUF_STRIDE);
if(nnz){
if(nnz == 1 && dMB_aux->mb[i*16])
ff_h264_idct8_dc_add_c(ptr, dMB_aux->mb + i*16, RECON_BUF_STRIDE);
else
ff_h264_idct8_add_c(ptr, dMB_aux->mb + i*16, RECON_BUF_STRIDE);
}
}
}else {
for(i=0; i<16; i++){
uint8_t * src_left = (uint32_t)(intra_pred4x4_left_ptr[1][i]) | 0xf4000000;
uint8_t * src_top = (uint32_t)(intra_pred4x4_top_ptr[1][i]) | 0xf4000000;
uint8_t * src_topleft;
if ( (i==2) || (i==8) || (i==10) )
src_topleft = src_left - RECON_BUF_STRIDE;
else src_topleft = src_top;
uint8_t * ptr= AUX_T->h264_yrecon[1] + MB_offset_recon[i];
uint8_t *topright;
int dir= dMB_aux->intra4x4_pred_mode_cache[ i ];
int nnz, tr;
if(dir == DIAG_DOWN_LEFT_PRED || dir == VERT_LEFT_PRED){
int topright_avail= (dMB_aux->topright_samples_available<<i)&0x8000;
if(!topright_avail){
tr= src_top[3]*0x01010101;
topright= (uint8_t*) &tr;
}else{
topright= src_top + 4;
}
}
Intra_pred_luma_4x4(dir, ptr, src_left, topright, src_top, src_topleft);
nnz = non_zero_count_cache_luma_last[i];
if(nnz){
if(nnz == 1 && dMB_aux->mb[i*16])
ff_h264_idct_dc_add_c(ptr, dMB_aux->mb + i*16, RECON_BUF_STRIDE);
else
ff_h264_idct_add_c(ptr, dMB_aux->mb + i*16, RECON_BUF_STRIDE);
}
}
}
}else{
Intra_pred_luma_16x16(dMB_aux->intra16x16_pred_mode, AUX_T->h264_yrecon[1],
AUX_T->h264_yrecon[0] + MB_LUMA_WIDTH, TCSM1_BOTTOM_Y_25PIX+4);
luma_dc_dequant_idct_c(dMB_aux->mb, dMB_aux->dequant4_coeff[0]);
for(i=0; i<16; i++){
if(non_zero_count_cache_luma_last[i])
ff_h264_idct_add_c(AUX_T->h264_yrecon[1] + MB_offset_recon[i],
dMB_aux->mb + i*16, RECON_BUF_STRIDE);
else if(dMB_aux->mb[i*16])
ff_h264_idct_dc_add_c(AUX_T->h264_yrecon[1] + MB_offset_recon[i],
dMB_aux->mb + i*16, RECON_BUF_STRIDE);
}
}
}
AUX_PMON_OFF(intra);
if(IS_INTER(mb_type)){
hl_motion_hw(dFRM, dMB_aux, AUX_T);
}
AUX_PMON_ON(mc);
MC_POLLING_END();
AUX_PMON_OFF(mc);
AUX_PMON_ON(dblk);
while ((*infar)!= TCSM1_H264_DBLK_INFDA) {}
AUX_PMON_OFF(dblk);
if(AUX_T->mc_des_dirty){
H264_MC_DesNode *h264_mc = (H264_MC_DesNode *)(AUX_T->h264_mc_des_ptr[0]);
h264_mc[AUX_T->mc_des_dirty - 1].VNodeHead = H264_VHEAD_UNLINK;
SET_MC_DHA((uint32_t)h264_mc);
CLEAR_MC_TTEND();
SET_MC_DCS();
}
if(IS_INTRA(mb_type)){
uint32_t * bakup_src = AUX_T->BackupMBbottom_Y[bakup_x1]-4;
uint32_t * bakup_dst = TCSM1_BOTTOM_Y_25PIX;
bakup_dst[0] = bakup_src[0];
bakup_dst[1] = bakup_src[1];
bakup_dst[2] = bakup_src[2];
bakup_dst[3] = bakup_src[3];
bakup_dst[4] = bakup_src[4];
bakup_dst[5] = bakup_src[5];
bakup_dst[6] = bakup_src[6];
bakup_src = AUX_T->BackupMBbottom_U[bakup_x1]-4;
bakup_dst = TCSM1_BOTTOM_U_13PIX;
bakup_dst[0] = bakup_src[0];
bakup_dst[1] = bakup_src[1];
bakup_dst[2] = bakup_src[2];
bakup_dst[3] = bakup_src[3];
bakup_src = AUX_T->BackupMBbottom_V[bakup_x1]-4;
bakup_dst = TCSM1_BOTTOM_V_13PIX;
bakup_dst[0] = bakup_src[0];
bakup_dst[1] = bakup_src[1];
bakup_dst[2] = bakup_src[2];
bakup_dst[3] = bakup_src[3];
}
if(IS_INTER(mb_type_last))
hl_motion_hw_next(dMB_aux,AUX_T,mb_type_last);
*(uint32_t*)&AUX_T->sub_mb_type[0] = *(uint32_t*)&dMB_aux->sub_mb_type[0];
*(uint32_t*)&AUX_T->sub_mb_type[2] = *(uint32_t*)&dMB_aux->sub_mb_type[2];
AUX_PMON_ON(idct);
AUX_PMON_OFF(idct);
if (IS_INTRA_PCM(mb_type)) {
unsigned int x, y;
for(i=0; i<24; i++) {
for (y=0; y<4; y++) {
for (x=0; x<4; x++) {
*(AUX_T->h264_yrecon[0] + MB_offset_recon[i] + y*RECON_BUF_STRIDE + x) = dMB_aux->mb[i*16+y*4+x];
}
}
}
}
{
int linesize, uvlinesize;
uint8_t *dest_y_d2, *dest_u_d2, *dest_v_d2;
linesize = dFRM->linesize;
uvlinesize = dFRM->uvlinesize;
dest_y_d2 = dFRM->current_picture.y_ptr + (mb_y_d2 * 16* linesize ) + mb_x_d2 * 16;
dest_u_d2 = dFRM->current_picture.u_ptr + (mb_y_d2 * 8 * uvlinesize) + mb_x_d2 * 8;
dest_v_d2 = dFRM->current_picture.v_ptr + (mb_y_d2 * 8 * uvlinesize) + mb_x_d2 * 8;
//move dblked MB out
MDMA1_TRAN[0].TSA = AUX_T->h264_ydblk[1]-4;
MDMA1_TRAN[0].TDA = get_phy_addr_aux((uint32_t)dest_y_d2-4);
MDMA1_TRAN[1].TSA = AUX_T->h264_ydblk[1]-4+DBLK_U_OFST;
MDMA1_TRAN[1].TDA = get_phy_addr_aux((uint32_t)dest_u_d2-4);
//MDMA1_TRAN[1].UNIT = MDMA_UNIT(1,12,96);
MDMA1_TRAN[2].TSA = AUX_T->h264_ydblk[1]-4+DBLK_V_OFST;
MDMA1_TRAN[2].TDA = get_phy_addr_aux((uint32_t)dest_v_d2-4);
POLLING_MDMA1_END(); //ensure next MB's args has been received
SET_MDMA1_DHA((uint32_t)MDMA1_TRAN);
MDMA1_RUN();
}
//---------idct fo inter---------------
if(IS_INTER(mb_type_last)){
void (*idct_dc_add)(uint8_t *dst, DCTELEM *block, int stride);
void (*idct_add)(uint8_t *dst, DCTELEM *block, int stride);
int di;
if(IS_8x8DCT(mb_type_last)){
idct_dc_add = ff_h264_idct8_dc_add_c;
idct_add = ff_h264_idct8_add_c;
di = 4;
}else{
idct_dc_add = ff_h264_idct_dc_add_c;
idct_add = ff_h264_idct_add_c;
di = 1;
}
for(i=0; i<16; i+=di){
int nnz = non_zero_count_cache_luma_last[i];
if(nnz){
if(nnz==1 && dMB_aux->mb[i*16])
idct_dc_add(AUX_T->h264_yrecon[1] + MB_offset_recon[i],
dMB_aux->mb + i*16, RECON_BUF_STRIDE);
else
idct_add(AUX_T->h264_yrecon[1] + MB_offset_recon[i],
dMB_aux->mb + i*16, RECON_BUF_STRIDE);
}
}
}
//------------- chroma idct------------
if(mb_type_last){
chroma_dc_dequant_idct_c(dMB_aux->mb + 16*16, dMB_aux->dequant4_coeff[1]);
chroma_dc_dequant_idct_c(dMB_aux->mb + 16*16+4*16, dMB_aux->dequant4_coeff[2]);
for(i=16; i<16+8; i++){
if(dMB_aux->non_zero_count_cache_chroma[ i - 16 ]) {
ff_h264_idct_add_c(AUX_T->h264_yrecon[1] + MB_offset_recon[i],
dMB_aux->mb + i*16, RECON_BUF_STRIDE);
} else if (dMB_aux->mb[i*16]) {
ff_h264_idct_dc_add_c(AUX_T->h264_yrecon[1] + MB_offset_recon[i],
dMB_aux->mb + i*16, RECON_BUF_STRIDE);
}
}
}
if(!(mb_x==0 && mb_y==0)){
SET_DHA_DBLK((unsigned int)AUX_T->h264_dblk_des_ptr[1]);
*infar = 0;
SET_DCS_DBLK(0x1);
}
filter_mb_dblk(dFRM, dMB_aux, AUX_T);
{
uint32_t * bakup_src = AUX_T->h264_yrecon[1] + 15*RECON_BUF_STRIDE;
uint32_t * bakup_dst = AUX_T->BackupMBbottom_Y[bakup_x0];
bakup_dst[0] = bakup_src[0];
bakup_dst[1] = bakup_src[1];
bakup_dst[2] = bakup_src[2];
bakup_dst[3] = bakup_src[3];
bakup_src = AUX_T->h264_urecon[1] + 7*RECON_BUF_STRIDE;
bakup_dst = AUX_T->BackupMBbottom_U[bakup_x0];
bakup_dst[0] = bakup_src[0];
bakup_dst[1] = bakup_src[1];
bakup_src = AUX_T->h264_vrecon[1] + 7*RECON_BUF_STRIDE;
bakup_dst = AUX_T->BackupMBbottom_V[bakup_x0];
bakup_dst[0] = bakup_src[0];
bakup_dst[1] = bakup_src[1];
}
mb_x_d2 = mb_x_d1;
mb_y_d2 = mb_y_d1;
mb_x_d1 = mb_x;
mb_y_d1 = mb_y;
mb_type_last=mb_type;
for(i=0;i<16;i++)
non_zero_count_cache_luma_last[i]=dMB_aux->non_zero_count_cache_luma[ scan5[i] ];
// abnormal exit
if (dMB_aux->deblocking_filter & 0x8) break;
bakup_x0=(bakup_x0==((dFRM->mb_width)-1))?0:(bakup_x0+1);//hit second line's tail
bakup_x1=(bakup_x1==((dFRM->mb_width)-1))?0:(bakup_x1+1);
XCHG2(AUX_T->h264_yrecon[0],AUX_T->h264_yrecon[1],XCHGtmp);
XCHG2(AUX_T->h264_urecon[0],AUX_T->h264_urecon[1],XCHGtmp);
XCHG2(AUX_T->h264_vrecon[0],AUX_T->h264_vrecon[1],XCHGtmp);
XCHG2(AUX_T->h264_ydblk[0],AUX_T->h264_ydblk[1],XCHGtmp);
XCHG2(AUX_T->h264_dblk_des_ptr[0],AUX_T->h264_dblk_des_ptr[1],XCHGtmp);
XCHG2(AUX_T->h264_mc_des_ptr[0],AUX_T->h264_mc_des_ptr[1],XCHGtmp);
XCHG2(intra_pred4x4_top_ptr[0],intra_pred4x4_top_ptr[1],XCHGtmp);
XCHG2(intra_pred4x4_left_ptr[0],intra_pred4x4_left_ptr[1],XCHGtmp);
XCHG2(dMB_aux,dMB2,XCHGtmp);
}
AUX_PMON_TRAN(mc,PMON_MC_BUF);
AUX_PMON_TRAN(idct,PMON_IDCT_BUF);
AUX_PMON_TRAN(dblk,PMON_DBLK_BUF);
AUX_PMON_TRAN(intra,PMON_INTRA_BUF);
AUX_PMON_TRAN(mdma,PMON_MDMA_BUF);
AUX_PMON_TRAN(sync,PMON_SYNC_BUF);
/*
task_done: aux-cpu task done flag, only write by aux-cpu, only read by main-cpu
*/
MDMA1_ARG->TSA = (TCSM1_TASK_DONE);
MDMA1_ARG->TDA = (TCSM0_TASK_DONE);
MDMA1_ARG->STRD = MDMA_STRD(4,4);
MDMA1_ARG->UNIT = MDMA_UNIT(1,4,4);
SET_MDMA1_DHA((uint32_t)TCSM1_GP1_ARG_CHAIN);
MDMA1_RUN();
i_nop;
i_nop;
i_nop;
i_nop;
__asm__ __volatile__ ("wait");
}
/*!
*************************************************************************************
* \brief
* Mode Decision for a macroblock
*************************************************************************************
*/
void encode_one_macroblock_low (Macroblock *currMB)
{
int block, mode, i, j, k, dummy;
char best_pdir;
RD_PARAMS enc_mb;
char best_ref[2] = {0, -1};
int bmcost[5] = {INT_MAX};
double rd_cost = 0, min_rd_cost = 1e30;
int cost = 0;
int min_cost = INT_MAX, cost_direct=0, have_direct=0, i16mode=0;
int intra1 = 0;
int temp_cpb = 0;
int best_transform_flag = 0;
int cost8x8_direct = 0;
short islice = (short) (img->type==I_SLICE);
short bslice = (short) (img->type==B_SLICE);
short pslice = (short) ((img->type==P_SLICE) || (img->type==SP_SLICE));
short intra = (short) (islice || (pslice && img->mb_y==img->mb_y_upd && img->mb_y_upd!=img->mb_y_intra));
int lambda_mf[3];
int pix_x, pix_y;
int prev_mb_nr = FmoGetPreviousMBNr(img->current_mb_nr);
Macroblock* prevMB = (prev_mb_nr >= 0) ? &img->mb_data[prev_mb_nr]:NULL ;
char **ipredmodes = img->ipredmode;
short *allmvs = params->IntraProfile ? NULL: img->all_mv[0][0][0][0][0];
int ****i4p; //for non-RD-opt. mode
imgpel (*curr_mpr)[16] = img->mpr[0];
int tmp_8x8_flag, tmp_no_mbpart;
// Fast Mode Decision
short inter_skip = 0;
if(params->SearchMode == UM_HEX)
{
UMHEX_decide_intrabk_SAD();
}
else if (params->SearchMode == UM_HEX_SIMPLE)
{
smpUMHEX_decide_intrabk_SAD();
}
intra |= RandomIntra (img->current_mb_nr); // Forced Pseudo-Random Intra
//===== Setup Macroblock encoding parameters =====
init_enc_mb_params(currMB, &enc_mb, intra, bslice);
// reset chroma intra predictor to default
currMB->c_ipred_mode = DC_PRED_8;
//===== S T O R E C O D I N G S T A T E =====
//---------------------------------------------------
store_coding_state (currMB, cs_cm);
if (!intra)
{
//===== set direct motion vectors =====
best_mode = 1;
if (bslice)
{
Get_Direct_Motion_Vectors (currMB);
}
if (params->CtxAdptLagrangeMult == 1)
{
get_initial_mb16x16_cost(currMB);
}
//===== MOTION ESTIMATION FOR 16x16, 16x8, 8x16 BLOCKS =====
for (min_cost=INT_MAX, mode=1; mode<4; mode++)
{
bi_pred_me = 0;
img->bi_pred_me[mode]=0;
if (enc_mb.valid[mode] && !inter_skip)
{
for (cost=0, block=0; block<(mode==1?1:2); block++)
{
update_lambda_costs(&enc_mb, lambda_mf);
PartitionMotionSearch (currMB, mode, block, lambda_mf);
//--- set 4x4 block indizes (for getting MV) ---
j = (block==1 && mode==2 ? 2 : 0);
i = (block==1 && mode==3 ? 2 : 0);
//--- get cost and reference frame for List 0 prediction ---
bmcost[LIST_0] = INT_MAX;
list_prediction_cost(currMB, LIST_0, block, mode, enc_mb, bmcost, best_ref);
if (bslice)
{
//--- get cost and reference frame for List 1 prediction ---
bmcost[LIST_1] = INT_MAX;
list_prediction_cost(currMB, LIST_1, block, mode, enc_mb, bmcost, best_ref);
// Compute bipredictive cost between best list 0 and best list 1 references
list_prediction_cost(currMB, BI_PRED, block, mode, enc_mb, bmcost, best_ref);
// Finally, if mode 16x16, compute cost for bipredictive ME vectore
if (params->BiPredMotionEstimation && mode == 1)
{
list_prediction_cost(currMB, BI_PRED_L0, block, mode, enc_mb, bmcost, 0);
list_prediction_cost(currMB, BI_PRED_L1, block, mode, enc_mb, bmcost, 0);
}
else
{
bmcost[BI_PRED_L0] = INT_MAX;
bmcost[BI_PRED_L1] = INT_MAX;
}
// Determine prediction list based on mode cost
determine_prediction_list(mode, bmcost, best_ref, &best_pdir, &cost, &bi_pred_me);
}
else // if (bslice)
{
best_pdir = 0;
cost += bmcost[LIST_0];
}
assign_enc_picture_params(mode, best_pdir, block, enc_mb.list_offset[LIST_0], best_ref[LIST_0], best_ref[LIST_1], bslice);
//----- set reference frame and direction parameters -----
if (mode==3)
{
best8x8l0ref [3][block ] = best8x8l0ref [3][ block+2] = best_ref[LIST_0];
best8x8pdir [3][block ] = best8x8pdir [3][ block+2] = best_pdir;
best8x8l1ref [3][block ] = best8x8l1ref [3][ block+2] = best_ref[LIST_1];
}
else if (mode==2)
{
best8x8l0ref [2][2*block] = best8x8l0ref [2][2*block+1] = best_ref[LIST_0];
best8x8pdir [2][2*block] = best8x8pdir [2][2*block+1] = best_pdir;
best8x8l1ref [2][2*block] = best8x8l1ref [2][2*block+1] = best_ref[LIST_1];
}
else
{
memset(&best8x8l0ref [1][0], best_ref[LIST_0], 4 * sizeof(char));
memset(&best8x8l1ref [1][0], best_ref[LIST_1], 4 * sizeof(char));
best8x8pdir [1][0] = best8x8pdir [1][1] = best8x8pdir [1][2] = best8x8pdir [1][3] = best_pdir;
}
//--- set reference frames and motion vectors ---
if (mode>1 && block==0)
SetRefAndMotionVectors (currMB, block, mode, best_pdir, best_ref[LIST_0], best_ref[LIST_1]);
} // for (block=0; block<(mode==1?1:2); block++)
currMB->luma_transform_size_8x8_flag = 0;
if (params->Transform8x8Mode) //for inter rd-off, set 8x8 to do 8x8 transform
{
SetModesAndRefframeForBlocks(currMB, mode);
currMB->luma_transform_size_8x8_flag = TransformDecision(currMB, -1, &cost);
}
if ((!inter_skip) && (cost < min_cost))
{
best_mode = (short) mode;
min_cost = cost;
best_transform_flag = currMB->luma_transform_size_8x8_flag;
if (params->CtxAdptLagrangeMult == 1)
{
adjust_mb16x16_cost(cost);
}
}
} // if (enc_mb.valid[mode])
} // for (mode=1; mode<4; mode++)
if ((!inter_skip) && enc_mb.valid[P8x8])
{
giRDOpt_B8OnlyFlag = 1;
tr8x8.cost8x8 = INT_MAX;
tr4x4.cost8x8 = INT_MAX;
//===== store coding state of macroblock =====
store_coding_state (currMB, cs_mb);
currMB->all_blk_8x8 = -1;
if (params->Transform8x8Mode)
{
tr8x8.cost8x8 = 0;
//===========================================================
// Check 8x8 partition with transform size 8x8
//===========================================================
//===== LOOP OVER 8x8 SUB-PARTITIONS (Motion Estimation & Mode Decision) =====
for (cost_direct=cbp8x8=cbp_blk8x8=cnt_nonz_8x8=0, block = 0; block < 4; block++)
{
submacroblock_mode_decision(enc_mb, &tr8x8, currMB, cofAC8x8ts[0][block], cofAC8x8ts[1][block], cofAC8x8ts[2][block],
&have_direct, bslice, block, &cost_direct, &cost, &cost8x8_direct, 1);
best8x8mode [block] = tr8x8.part8x8mode [block];
best8x8pdir [P8x8][block] = tr8x8.part8x8pdir [block];
best8x8l0ref[P8x8][block] = tr8x8.part8x8l0ref[block];
best8x8l1ref[P8x8][block] = tr8x8.part8x8l1ref[block];
}
// following params could be added in RD_8x8DATA structure
cbp8_8x8ts = cbp8x8;
cbp_blk8_8x8ts = cbp_blk8x8;
cnt_nonz8_8x8ts = cnt_nonz_8x8;
currMB->luma_transform_size_8x8_flag = 0; //switch to 4x4 transform size
//--- re-set coding state (as it was before 8x8 block coding) ---
//reset_coding_state (currMB, cs_mb);
}// if (params->Transform8x8Mode)
if (params->Transform8x8Mode != 2)
{
tr4x4.cost8x8 = 0;
//=================================================================
// Check 8x8, 8x4, 4x8 and 4x4 partitions with transform size 4x4
//=================================================================
//===== LOOP OVER 8x8 SUB-PARTITIONS (Motion Estimation & Mode Decision) =====
for (cost_direct=cbp8x8=cbp_blk8x8=cnt_nonz_8x8=0, block=0; block<4; block++)
{
submacroblock_mode_decision(enc_mb, &tr4x4, currMB, cofAC8x8[block], cofAC8x8CbCr[0][block], cofAC8x8CbCr[1][block],
&have_direct, bslice, block, &cost_direct, &cost, &cost8x8_direct, 0);
best8x8mode [block] = tr4x4.part8x8mode [block];
best8x8pdir [P8x8][block] = tr4x4.part8x8pdir [block];
best8x8l0ref[P8x8][block] = tr4x4.part8x8l0ref[block];
best8x8l1ref[P8x8][block] = tr4x4.part8x8l1ref[block];
}
//--- re-set coding state (as it was before 8x8 block coding) ---
// reset_coding_state (currMB, cs_mb);
}// if (params->Transform8x8Mode != 2)
//--- re-set coding state (as it was before 8x8 block coding) ---
reset_coding_state (currMB, cs_mb);
// This is not enabled yet since mpr has reverse order.
if (params->RCEnable)
rc_store_diff(img->opix_x, img->opix_y, curr_mpr);
//check cost for P8x8 for non-rdopt mode
if (tr4x4.cost8x8 < min_cost || tr8x8.cost8x8 < min_cost)
{
best_mode = P8x8;
if (params->Transform8x8Mode == 2)
{
min_cost = tr8x8.cost8x8;
currMB->luma_transform_size_8x8_flag=1;
}
else if (params->Transform8x8Mode)
{
if (tr8x8.cost8x8 < tr4x4.cost8x8)
{
min_cost = tr8x8.cost8x8;
currMB->luma_transform_size_8x8_flag=1;
}
else if(tr4x4.cost8x8 < tr8x8.cost8x8)
{
min_cost = tr4x4.cost8x8;
currMB->luma_transform_size_8x8_flag=0;
}
else
{
if (GetBestTransformP8x8() == 0)
{
min_cost = tr4x4.cost8x8;
currMB->luma_transform_size_8x8_flag=0;
}
else
{
min_cost = tr8x8.cost8x8;
currMB->luma_transform_size_8x8_flag=1;
}
}
}
else
{
min_cost = tr4x4.cost8x8;
currMB->luma_transform_size_8x8_flag=0;
}
}// if ((tr4x4.cost8x8 < min_cost || tr8x8.cost8x8 < min_cost))
giRDOpt_B8OnlyFlag = 0;
}
else // if (enc_mb.valid[P8x8])
{
tr4x4.cost8x8 = INT_MAX;
}
// Find a motion vector for the Skip mode
if(pslice)
FindSkipModeMotionVector (currMB);
}
else // if (!intra)
{
min_cost = INT_MAX;
}
//========= C H O O S E B E S T M A C R O B L O C K M O D E =========
//-------------------------------------------------------------------------
tmp_8x8_flag = currMB->luma_transform_size_8x8_flag; //save 8x8_flag
tmp_no_mbpart = currMB->NoMbPartLessThan8x8Flag; //save no-part-less
if ((img->yuv_format != YUV400) && (img->yuv_format != YUV444))
// precompute all chroma intra prediction modes
IntraChromaPrediction(currMB, NULL, NULL, NULL);
if (enc_mb.valid[0] && bslice) // check DIRECT MODE
{
if(have_direct)
{
switch(params->Transform8x8Mode)
{
case 1: // Mixture of 8x8 & 4x4 transform
cost = ((cost8x8_direct < cost_direct) || !(enc_mb.valid[5] && enc_mb.valid[6] && enc_mb.valid[7]))
? cost8x8_direct : cost_direct;
break;
case 2: // 8x8 Transform only
cost = cost8x8_direct;
break;
default: // 4x4 Transform only
cost = cost_direct;
break;
}
}
else
{ //!have_direct
cost = GetDirectCostMB (currMB, bslice);
}
if (cost!=INT_MAX)
{
cost -= (int)floor(16*enc_mb.lambda_md+0.4999);
}
if (cost <= min_cost)
{
if(active_sps->direct_8x8_inference_flag && params->Transform8x8Mode)
{
if(params->Transform8x8Mode==2)
currMB->luma_transform_size_8x8_flag=1;
else
{
if(cost8x8_direct < cost_direct)
currMB->luma_transform_size_8x8_flag=1;
else
currMB->luma_transform_size_8x8_flag=0;
}
}
else
currMB->luma_transform_size_8x8_flag=0;
//Rate control
if (params->RCEnable)
rc_store_diff(img->opix_x, img->opix_y, curr_mpr);
min_cost = cost;
best_mode = 0;
tmp_8x8_flag = currMB->luma_transform_size_8x8_flag;
}
else
{
currMB->luma_transform_size_8x8_flag = tmp_8x8_flag; // restore if not best
currMB->NoMbPartLessThan8x8Flag = tmp_no_mbpart; // restore if not best
}
}
min_rd_cost = (double) min_cost;
if (enc_mb.valid[I8MB]) // check INTRA8x8
{
currMB->luma_transform_size_8x8_flag = 1; // at this point cost will ALWAYS be less than min_cost
currMB->mb_type = I8MB;
temp_cpb = Mode_Decision_for_new_Intra8x8Macroblock (currMB, enc_mb.lambda_md, &rd_cost);
if (rd_cost <= min_rd_cost) //HYU_NOTE. bug fix. 08/15/07
{
currMB->cbp = temp_cpb;
if (img->P444_joined)
{
curr_cbp[0] = cmp_cbp[1];
curr_cbp[1] = cmp_cbp[2];
}
if(enc_mb.valid[I4MB]) //KHHan. bug fix. Oct.15.2007
{
//coeffs
if (params->Transform8x8Mode != 2)
{
i4p=cofAC; cofAC=img->cofAC; img->cofAC=i4p;
}
}
for(j=0; j<MB_BLOCK_SIZE; j++)
{
pix_y = img->pix_y + j;
for(i=0; i<MB_BLOCK_SIZE; i++)
{
pix_x = img->pix_x + i;
temp_imgY[j][i] = enc_picture->imgY[pix_y][pix_x];
}
}
if (img->P444_joined)
{
for(j=0; j<MB_BLOCK_SIZE; j++)
{
pix_y = img->pix_y + j;
for(i=0; i<MB_BLOCK_SIZE; i++)
{
pix_x = img->pix_x + i;
temp_imgU[j][i] = enc_picture->imgUV[0][pix_y][pix_x];
temp_imgV[j][i] = enc_picture->imgUV[1][pix_y][pix_x];
}
}
}
//Rate control
if (params->RCEnable)
rc_store_diff(img->opix_x, img->opix_y, curr_mpr);
min_rd_cost = rd_cost;
best_mode = I8MB;
tmp_8x8_flag = currMB->luma_transform_size_8x8_flag;
}
else
{
currMB->luma_transform_size_8x8_flag = tmp_8x8_flag; // restore if not best
if (img->P444_joined)
{
cmp_cbp[1] = curr_cbp[0];
cmp_cbp[2] = curr_cbp[1];
currMB->cbp |= cmp_cbp[1];
currMB->cbp |= cmp_cbp[2];
cmp_cbp[1] = currMB->cbp;
cmp_cbp[2] = currMB->cbp;
}
}
}
if (enc_mb.valid[I4MB]) // check INTRA4x4
{
currMB->luma_transform_size_8x8_flag = 0;
currMB->mb_type = I4MB;
temp_cpb = Mode_Decision_for_Intra4x4Macroblock (currMB, enc_mb.lambda_md, &rd_cost);
if (rd_cost <= min_rd_cost)
{
currMB->cbp = temp_cpb;
//Rate control
if (params->RCEnable)
rc_store_diff(img->opix_x, img->opix_y, curr_mpr);
min_rd_cost = rd_cost;
best_mode = I4MB;
tmp_8x8_flag = currMB->luma_transform_size_8x8_flag;
}
else
{
currMB->luma_transform_size_8x8_flag = tmp_8x8_flag; // restore if not best
if (img->P444_joined)
{
cmp_cbp[1] = curr_cbp[0];
cmp_cbp[2] = curr_cbp[1];
currMB->cbp |= cmp_cbp[1];
currMB->cbp |= cmp_cbp[2];
cmp_cbp[1] = currMB->cbp;
cmp_cbp[2] = currMB->cbp;
}
//coeffs
i4p=cofAC; cofAC=img->cofAC; img->cofAC=i4p;
}
}
if (enc_mb.valid[I16MB]) // check INTRA16x16
{
currMB->luma_transform_size_8x8_flag = 0;
intrapred_16x16 (currMB, PLANE_Y);
if (img->P444_joined)
{
select_plane(PLANE_U);
intrapred_16x16 (currMB, PLANE_U);
select_plane(PLANE_V);
intrapred_16x16 (currMB, PLANE_V);
select_plane(PLANE_Y);
}
switch(params->FastIntra16x16)
{
case 0:
default:
find_sad_16x16 = find_sad_16x16_JM;
break;
}
rd_cost = find_sad_16x16 (currMB, &i16mode);
if (rd_cost < min_rd_cost)
{
//Rate control
if (params->RCEnable)
rc_store_diff(img->opix_x,img->opix_y,img->mpr_16x16[0][i16mode]);
best_mode = I16MB;
min_rd_cost = rd_cost;
currMB->cbp = pDCT_16x16 (currMB, PLANE_Y, i16mode);
if (img->P444_joined)
{
select_plane(PLANE_U);
cmp_cbp[1] = pDCT_16x16(currMB, PLANE_U, i16mode);
select_plane(PLANE_V);
cmp_cbp[2] = pDCT_16x16(currMB, PLANE_V, i16mode);
select_plane(PLANE_Y);
currMB->cbp |= cmp_cbp[1];
currMB->cbp |= cmp_cbp[2];
cmp_cbp[1] = currMB->cbp;
cmp_cbp[2] = currMB->cbp;
}
}
else
{
currMB->luma_transform_size_8x8_flag = tmp_8x8_flag; // restore
currMB->NoMbPartLessThan8x8Flag = tmp_no_mbpart; // restore
}
}
intra1 = IS_INTRA(currMB);
//===== S E T F I N A L M A C R O B L O C K P A R A M E T E R S ======
//---------------------------------------------------------------------------
{
//===== set parameters for chosen mode =====
SetModesAndRefframeForBlocks (currMB, best_mode);
if (best_mode==P8x8)
{
if (currMB->luma_transform_size_8x8_flag && (cbp8_8x8ts == 0) && params->Transform8x8Mode != 2)
currMB->luma_transform_size_8x8_flag = 0;
SetCoeffAndReconstruction8x8 (currMB);
memset(currMB->intra_pred_modes, DC_PRED, MB_BLOCK_PARTITIONS * sizeof(char));
for (k=0, j = img->block_y; j < img->block_y + BLOCK_MULTIPLE; j++)
memset(&ipredmodes[j][img->block_x], DC_PRED, BLOCK_MULTIPLE * sizeof(char));
}
else
{
//===== set parameters for chosen mode =====
if (best_mode == I8MB)
{
memcpy(currMB->intra_pred_modes,currMB->intra_pred_modes8x8, MB_BLOCK_PARTITIONS * sizeof(char));
for(j = img->block_y; j < img->block_y + BLOCK_MULTIPLE; j++)
memcpy(&img->ipredmode[j][img->block_x],&img->ipredmode8x8[j][img->block_x], BLOCK_MULTIPLE * sizeof(char));
//--- restore reconstruction for 8x8 transform ---
for(j=0; j<MB_BLOCK_SIZE; j++)
{
memcpy(&enc_picture->imgY[img->pix_y + j][img->pix_x],temp_imgY[j], MB_BLOCK_SIZE * sizeof(imgpel));
}
if (img->P444_joined)
{
for(j=0; j<MB_BLOCK_SIZE; j++)
{
memcpy(&enc_picture->imgUV[0][img->pix_y + j][img->pix_x],temp_imgU[j], MB_BLOCK_SIZE * sizeof(imgpel));
memcpy(&enc_picture->imgUV[1][img->pix_y + j][img->pix_x],temp_imgV[j], MB_BLOCK_SIZE * sizeof(imgpel));
}
}
}
if ((best_mode!=I4MB)&&(best_mode != I8MB))
{
memset(currMB->intra_pred_modes,DC_PRED, MB_BLOCK_PARTITIONS * sizeof(char));
for(j = img->block_y; j < img->block_y + BLOCK_MULTIPLE; j++)
memset(&ipredmodes[j][img->block_x],DC_PRED, BLOCK_MULTIPLE * sizeof(char));
if (best_mode!=I16MB)
{
if((best_mode>=1) && (best_mode<=3))
currMB->luma_transform_size_8x8_flag = best_transform_flag;
LumaResidualCoding (currMB);
if (img->P444_joined)
{
if((currMB->cbp==0 && cmp_cbp[1] == 0 && cmp_cbp[2] == 0) &&(best_mode==0))
currMB->luma_transform_size_8x8_flag = 0;
}
else if((currMB->cbp==0)&&(best_mode==0))
currMB->luma_transform_size_8x8_flag = 0;
//Rate control
if (params->RCEnable)
rc_store_diff(img->opix_x,img->opix_y,curr_mpr);
}
}
}
//check luma cbp for transform size flag
if (((currMB->cbp&15) == 0) && !(IS_OLDINTRA(currMB) || currMB->mb_type == I8MB))
currMB->luma_transform_size_8x8_flag = 0;
// precompute all chroma intra prediction modes
if ((img->yuv_format != YUV400) && (img->yuv_format != YUV444))
IntraChromaPrediction(currMB, NULL, NULL, NULL);
img->i16offset = 0;
dummy = 0;
if ((img->yuv_format != YUV400) && (img->yuv_format != YUV444))
ChromaResidualCoding (currMB);
if (best_mode==I16MB)
{
img->i16offset = I16Offset (currMB->cbp, i16mode);
}
SetMotionVectorsMB (currMB, bslice);
//===== check for SKIP mode =====
if(img->P444_joined)
{
if ((pslice) && best_mode==1 && currMB->cbp==0 && cmp_cbp[1] == 0 && cmp_cbp[2] == 0 &&
enc_picture->ref_idx[LIST_0][img->block_y][img->block_x] == 0 &&
enc_picture->mv [LIST_0][img->block_y][img->block_x][0] == allmvs[0] &&
enc_picture->mv [LIST_0][img->block_y][img->block_x][1] == allmvs[1])
{
currMB->mb_type = currMB->b8mode[0] = currMB->b8mode[1] = currMB->b8mode[2] = currMB->b8mode[3] = 0;
currMB->luma_transform_size_8x8_flag = 0;
}
}
else if ((pslice) && best_mode==1 && currMB->cbp==0 &&
enc_picture->ref_idx[LIST_0][img->block_y][img->block_x] == 0 &&
enc_picture->mv [LIST_0][img->block_y][img->block_x][0] == allmvs[0] &&
enc_picture->mv [LIST_0][img->block_y][img->block_x][1] == allmvs[1])
{
currMB->mb_type = currMB->b8mode[0] = currMB->b8mode[1] = currMB->b8mode[2] = currMB->b8mode[3] = 0;
currMB->luma_transform_size_8x8_flag = 0;
}
if (img->MbaffFrameFlag || (params->UseRDOQuant && params->RDOQ_QP_Num > 1))
set_mbaff_parameters(currMB);
}
// Rate control
if(params->RCEnable && params->RCUpdateMode <= MAX_RC_MODE)
rc_store_mad(currMB);
update_qp_cbp(currMB, best_mode);
rdopt->min_rdcost = min_rd_cost;
rdopt->min_dcost = min_rd_cost;
if ( (img->MbaffFrameFlag)
&& (img->current_mb_nr%2)
&& (currMB->mb_type ? 0:((bslice) ? !currMB->cbp:1)) // bottom is skip
&& (prevMB->mb_type ? 0:((bslice) ? !prevMB->cbp:1))
&& !(field_flag_inference(currMB) == enc_mb.curr_mb_field)) // top is skip
{
rdopt->min_rdcost = 1e30; // don't allow coding of a MB pair as skip if wrong inference
}
//===== Decide if this MB will restrict the reference frames =====
if (params->RestrictRef)
update_refresh_map(intra, intra1, currMB);
if(params->SearchMode == UM_HEX)
{
UMHEX_skip_intrabk_SAD(best_mode, listXsize[enc_mb.list_offset[LIST_0]]);
}
else if(params->SearchMode == UM_HEX_SIMPLE)
{
smpUMHEX_skip_intrabk_SAD(best_mode, listXsize[enc_mb.list_offset[LIST_0]]);
}
//--- constrain intra prediction ---
if(params->UseConstrainedIntraPred && (img->type==P_SLICE || img->type==B_SLICE))
{
img->intra_block[img->current_mb_nr] = IS_INTRA(currMB);
}
}
void readLumaCoeff_B8(int block8x8, struct inp_par *inp, struct img_par *img)
{
int i;
int mb_nr = img->current_mb_nr;
Macroblock *currMB = &mb_data[mb_nr];
const int cbp = currMB->cbp;
SyntaxElement currSE;
int intra;
int inumblk; /* number of blocks per CBP*/
int inumcoeff; /* number of coeffs per block */
int icoef; /* current coefficient */
int ipos;
int run, level;
int ii,jj;
int sbx, sby;
int boff_x, boff_y;
int any_coeff;
int vlc_numcoef;
int cbp_blk_mask;
int tablenum; //add by qwang
static const int incVlc_intra[7] = { 0,1,2,4,7,10,3000};
static const int incVlc_inter[7] = { 0,1,2,3,6,9,3000};
int buffer_level[65]; //add by qwang
int buffer_run[64]; //add by qwang
int EOB_Pos_intra[7] = { -1, 8, 8, 8, 6, 0, 0};
int EOB_Pos_inter[7] = {-1, 2, 2, 2, 2, 0, 0};
char (*AVS_VLC_table_intra)[64][2];
char (*AVS_VLC_table_inter)[64][2];
int symbol2D,Golomb_se_type;
const char (*table2D)[27];
//digipro_1
float QP99;
long sum;
int shift = 7;
//digipro_0
int val, QPI;
int qp, q_shift; // dequantization parameters
static const int blkmode2ctx [4] = {LUMA_8x8, LUMA_8x4, LUMA_4x8, LUMA_4x4};
// this has to be done for each subblock seperately
intra = IS_INTRA(currMB);
inumblk = 1;
inumcoeff = 65; // all positions + EOB
// ========= dequantization values ===========
qp = currMB->qp; // using old style qp.
q_shift = qp/QUANT_PERIOD;
QP99 =(float)(pow(2.0, (float)(qp-5.0)/8.0));
for (shift=1; shift<16; shift++)
{
QPI = (int)((1<<shift)*QP99+0.5);
if (QPI>(1<<16))
{
shift -=1;
QPI = (int)((1<<shift)*QP99+0.5);
break;
}
}
if (shift==16) shift=15;
QPI = (int)((1<<shift)*QP99+0.5);
//make decoder table for 2DVLC_INTRA code
if(AVS_2DVLC_INTRA_dec[0][0][1]<0) // Don't need to set this every time. rewrite later.
{
memset(AVS_2DVLC_INTRA_dec,-1,sizeof(AVS_2DVLC_INTRA_dec));
for(i=0;i<7;i++)
{
table2D=AVS_2DVLC_INTRA[i];
for(run=0;run<26;run++)
for(level=0;level<27;level++)
{
ipos=table2D[run][level];
assert(ipos<64);
if(ipos>=0)
{
if(i==0)
{
AVS_2DVLC_INTRA_dec[i][ipos][0]=level+1;
AVS_2DVLC_INTRA_dec[i][ipos][1]=run;
AVS_2DVLC_INTRA_dec[i][ipos+1][0]=-(level+1);
AVS_2DVLC_INTRA_dec[i][ipos+1][1]=run;
}
else
{
AVS_2DVLC_INTRA_dec[i][ipos][0]=level;
AVS_2DVLC_INTRA_dec[i][ipos][1]=run;
if(level)
{
AVS_2DVLC_INTRA_dec[i][ipos+1][0]=-(level);
AVS_2DVLC_INTRA_dec[i][ipos+1][1]=run;
}
}
}
}
}
assert(AVS_2DVLC_INTRA_dec[0][0][1]>=0); //otherwise, tables are bad.
}
//make decoder table for 2DVLC_INTER code
if(AVS_2DVLC_INTER_dec[0][0][1]<0) // Don't need to set this every time. rewrite later.
{
memset(AVS_2DVLC_INTER_dec,-1,sizeof(AVS_2DVLC_INTER_dec));
for(i=0;i<7;i++)
{
table2D=AVS_2DVLC_INTER[i];
for(run=0;run<26;run++)
for(level=0;level<27;level++)
{
ipos=table2D[run][level];
assert(ipos<64);
if(ipos>=0)
{
if(i==0)
{
AVS_2DVLC_INTER_dec[i][ipos][0]=level+1;
AVS_2DVLC_INTER_dec[i][ipos][1]=run;
AVS_2DVLC_INTER_dec[i][ipos+1][0]=-(level+1);
AVS_2DVLC_INTER_dec[i][ipos+1][1]=run;
}
else
{
AVS_2DVLC_INTER_dec[i][ipos][0]=level;
AVS_2DVLC_INTER_dec[i][ipos][1]=run;
if(level)
{
AVS_2DVLC_INTER_dec[i][ipos+1][0]=-(level);
AVS_2DVLC_INTER_dec[i][ipos+1][1]=run;
}
}
}
}
}
assert(AVS_2DVLC_INTER_dec[0][0][1]>=0); //otherwise, tables are bad.
}
//clear cbp_blk bits of thie 8x8 block (and not all 4!)
cbp_blk_mask = cbp_blk_masks[0] ;
if(block8x8&1)cbp_blk_mask<<=2;
if(block8x8&2)cbp_blk_mask<<=8;
currMB->cbp_blk&=~cbp_blk_mask;
vlc_numcoef=-1;
Golomb_se_type=SE_LUM_AC_INTER;
if( intra )
{
vlc_numcoef=0; //this means 'use numcoeffs symbol'.
Golomb_se_type=SE_LUM_AC_INTRA;
}
AVS_VLC_table_intra = AVS_2DVLC_INTRA_dec;
AVS_VLC_table_inter = AVS_2DVLC_INTER_dec;
// === decoding ===
if ( cbp & (1<<block8x8) )
{
// === set offset in current macroblock ===
boff_x = ( (block8x8%2)<<3 );
boff_y = ( (block8x8/2)<<3 );
img->subblock_x = boff_x>>2;
img->subblock_y = boff_y>>2;
ipos = -1;
any_coeff=1; //modified by qwang any_coeff=0
if(intra)
{
tablenum = 0;
for(i=0; i<inumcoeff; i++)
{
//read 2D symbol
currSE.type = Golomb_se_type;
//currSE.golomb_grad = 2;
//currSE.golomb_maxlevels=4;
currSE.golomb_grad = VLC_Golomb_Order[0][tablenum][0];
currSE.golomb_maxlevels = VLC_Golomb_Order[0][tablenum][1];
readSyntaxElement_GOLOMB(&currSE,img,inp);
symbol2D = currSE.value1;
//if(symbol2D == EOB_Pos_intra[tablenum])
if(symbol2D == EOB_Pos_intra[tablenum])
{
vlc_numcoef = i;
break;
}
if(symbol2D < CODE2D_ESCAPE_SYMBOL)
{
level = AVS_2DVLC_INTRA_dec[tablenum][symbol2D][0];
run = AVS_2DVLC_INTRA_dec[tablenum][symbol2D][1];
}
else
{
// changed by dj
run = (symbol2D-CODE2D_ESCAPE_SYMBOL)>>1;
//decode level
currSE.type=Golomb_se_type;
currSE.golomb_grad = 1;
currSE.golomb_maxlevels=11; //2007.05.09
readSyntaxElement_GOLOMB(&currSE,img,inp);
level = currSE.value1 + ((run>MaxRun[0][tablenum])?1:RefAbsLevel[tablenum][run]);
// if( (symbol2D-CODE2D_ESCAPE_SYMBOL) & 1 )
if(symbol2D & 1)
level=-level;
}
// 保存level,run到缓冲区
buffer_level[i] = level;
buffer_run[i] = run;
if(abs(level) > incVlc_intra[tablenum])
{
if(abs(level) <= 2)
tablenum = abs(level);
else if(abs(level) <= 4)
tablenum = 3;
else if(abs(level) <= 7)
tablenum = 4;
else if(abs(level) <= 10)
tablenum = 5;
else
tablenum = 6;
}
}//loop for icoef
//将解码的level,run写到img->m7[][];
for(i=(vlc_numcoef-1); i>=0; i--)
{
ipos += (buffer_run[i]+1);
ii = SCAN[img->picture_structure][ipos][0];
jj = SCAN[img->picture_structure][ipos][1];
shift = IQ_SHIFT[qp];
QPI = IQ_TAB[qp];
val = buffer_level[i];
sum = (val*QPI+(1<<(shift-2)) )>>(shift-1);
img->m7[boff_x + ii][boff_y + jj] = sum;
}
}//if (intra)
else
{
static void pred_spatial_direct_motion(H264Context * const h, int *mb_type){
MpegEncContext * const s = &h->s;
int b8_stride = 2;
int b4_stride = h->b_stride;
int mb_xy = h->mb_xy, mb_y = s->mb_y;
int mb_type_col[2];
const int16_t (*l1mv0)[2], (*l1mv1)[2];
const int8_t *l1ref0, *l1ref1;
const int is_b8x8 = IS_8X8(*mb_type);
unsigned int sub_mb_type= MB_TYPE_L0L1;
int i8, i4;
int ref[2];
int mv[2];
int list;
assert(h->ref_list[1][0].f.reference & 3);
await_reference_mb_row(h, &h->ref_list[1][0], s->mb_y + !!IS_INTERLACED(*mb_type));
#define MB_TYPE_16x16_OR_INTRA (MB_TYPE_16x16|MB_TYPE_INTRA4x4|MB_TYPE_INTRA16x16|MB_TYPE_INTRA_PCM)
/* ref = min(neighbors) */
for(list=0; list<2; list++){
int left_ref = h->ref_cache[list][scan8[0] - 1];
int top_ref = h->ref_cache[list][scan8[0] - 8];
int refc = h->ref_cache[list][scan8[0] - 8 + 4];
const int16_t *C= h->mv_cache[list][ scan8[0] - 8 + 4];
if(refc == PART_NOT_AVAILABLE){
refc = h->ref_cache[list][scan8[0] - 8 - 1];
C = h-> mv_cache[list][scan8[0] - 8 - 1];
}
ref[list] = FFMIN3((unsigned)left_ref, (unsigned)top_ref, (unsigned)refc);
if(ref[list] >= 0){
//this is just pred_motion() but with the cases removed that cannot happen for direct blocks
const int16_t * const A= h->mv_cache[list][ scan8[0] - 1 ];
const int16_t * const B= h->mv_cache[list][ scan8[0] - 8 ];
int match_count= (left_ref==ref[list]) + (top_ref==ref[list]) + (refc==ref[list]);
if(match_count > 1){ //most common
mv[list]= pack16to32(mid_pred(A[0], B[0], C[0]),
mid_pred(A[1], B[1], C[1]) );
}else {
assert(match_count==1);
if(left_ref==ref[list]){
mv[list]= AV_RN32A(A);
}else if(top_ref==ref[list]){
mv[list]= AV_RN32A(B);
}else{
mv[list]= AV_RN32A(C);
}
}
}else{
int mask= ~(MB_TYPE_L0 << (2*list));
mv[list] = 0;
ref[list] = -1;
if(!is_b8x8)
*mb_type &= mask;
sub_mb_type &= mask;
}
}
if(ref[0] < 0 && ref[1] < 0){
ref[0] = ref[1] = 0;
if(!is_b8x8)
*mb_type |= MB_TYPE_L0L1;
sub_mb_type |= MB_TYPE_L0L1;
}
if(!(is_b8x8|mv[0]|mv[1])){
fill_rectangle(&h->ref_cache[0][scan8[0]], 4, 4, 8, (uint8_t)ref[0], 1);
fill_rectangle(&h->ref_cache[1][scan8[0]], 4, 4, 8, (uint8_t)ref[1], 1);
fill_rectangle(&h->mv_cache[0][scan8[0]], 4, 4, 8, 0, 4);
fill_rectangle(&h->mv_cache[1][scan8[0]], 4, 4, 8, 0, 4);
*mb_type= (*mb_type & ~(MB_TYPE_8x8|MB_TYPE_16x8|MB_TYPE_8x16|MB_TYPE_P1L0|MB_TYPE_P1L1))|MB_TYPE_16x16|MB_TYPE_DIRECT2;
return;
}
if (IS_INTERLACED(h->ref_list[1][0].f.mb_type[mb_xy])) { // AFL/AFR/FR/FL -> AFL/FL
if (!IS_INTERLACED(*mb_type)) { // AFR/FR -> AFL/FL
mb_y = (s->mb_y&~1) + h->col_parity;
mb_xy= s->mb_x + ((s->mb_y&~1) + h->col_parity)*s->mb_stride;
b8_stride = 0;
}else{
mb_y += h->col_fieldoff;
mb_xy += s->mb_stride*h->col_fieldoff; // non zero for FL -> FL & differ parity
}
goto single_col;
}else{ // AFL/AFR/FR/FL -> AFR/FR
if(IS_INTERLACED(*mb_type)){ // AFL /FL -> AFR/FR
mb_y = s->mb_y&~1;
mb_xy= s->mb_x + (s->mb_y&~1)*s->mb_stride;
mb_type_col[0] = h->ref_list[1][0].f.mb_type[mb_xy];
mb_type_col[1] = h->ref_list[1][0].f.mb_type[mb_xy + s->mb_stride];
b8_stride = 2+4*s->mb_stride;
b4_stride *= 6;
if (IS_INTERLACED(mb_type_col[0]) != IS_INTERLACED(mb_type_col[1])) {
mb_type_col[0] &= ~MB_TYPE_INTERLACED;
mb_type_col[1] &= ~MB_TYPE_INTERLACED;
}
sub_mb_type |= MB_TYPE_16x16|MB_TYPE_DIRECT2; /* B_SUB_8x8 */
if( (mb_type_col[0] & MB_TYPE_16x16_OR_INTRA)
&& (mb_type_col[1] & MB_TYPE_16x16_OR_INTRA)
&& !is_b8x8){
*mb_type |= MB_TYPE_16x8 |MB_TYPE_DIRECT2; /* B_16x8 */
}else{
*mb_type |= MB_TYPE_8x8;
}
}else{ // AFR/FR -> AFR/FR
single_col:
mb_type_col[0] =
mb_type_col[1] = h->ref_list[1][0].f.mb_type[mb_xy];
sub_mb_type |= MB_TYPE_16x16|MB_TYPE_DIRECT2; /* B_SUB_8x8 */
if(!is_b8x8 && (mb_type_col[0] & MB_TYPE_16x16_OR_INTRA)){
*mb_type |= MB_TYPE_16x16|MB_TYPE_DIRECT2; /* B_16x16 */
}else if(!is_b8x8 && (mb_type_col[0] & (MB_TYPE_16x8|MB_TYPE_8x16))){
*mb_type |= MB_TYPE_DIRECT2 | (mb_type_col[0] & (MB_TYPE_16x8|MB_TYPE_8x16));
}else{
if(!h->sps.direct_8x8_inference_flag){
/* FIXME save sub mb types from previous frames (or derive from MVs)
* so we know exactly what block size to use */
sub_mb_type += (MB_TYPE_8x8-MB_TYPE_16x16); /* B_SUB_4x4 */
}
*mb_type |= MB_TYPE_8x8;
}
}
}
await_reference_mb_row(h, &h->ref_list[1][0], mb_y);
l1mv0 = &h->ref_list[1][0].f.motion_val[0][h->mb2b_xy [mb_xy]];
l1mv1 = &h->ref_list[1][0].f.motion_val[1][h->mb2b_xy [mb_xy]];
l1ref0 = &h->ref_list[1][0].f.ref_index [0][4 * mb_xy];
l1ref1 = &h->ref_list[1][0].f.ref_index [1][4 * mb_xy];
if(!b8_stride){
if(s->mb_y&1){
l1ref0 += 2;
l1ref1 += 2;
l1mv0 += 2*b4_stride;
l1mv1 += 2*b4_stride;
}
}
if(IS_INTERLACED(*mb_type) != IS_INTERLACED(mb_type_col[0])){
int n=0;
for(i8=0; i8<4; i8++){
int x8 = i8&1;
int y8 = i8>>1;
int xy8 = x8+y8*b8_stride;
int xy4 = 3*x8+y8*b4_stride;
int a,b;
if(is_b8x8 && !IS_DIRECT(h->sub_mb_type[i8]))
continue;
h->sub_mb_type[i8] = sub_mb_type;
fill_rectangle(&h->ref_cache[0][scan8[i8*4]], 2, 2, 8, (uint8_t)ref[0], 1);
fill_rectangle(&h->ref_cache[1][scan8[i8*4]], 2, 2, 8, (uint8_t)ref[1], 1);
if(!IS_INTRA(mb_type_col[y8]) && !h->ref_list[1][0].long_ref
&& ( (l1ref0[xy8] == 0 && FFABS(l1mv0[xy4][0]) <= 1 && FFABS(l1mv0[xy4][1]) <= 1)
|| (l1ref0[xy8] < 0 && l1ref1[xy8] == 0 && FFABS(l1mv1[xy4][0]) <= 1 && FFABS(l1mv1[xy4][1]) <= 1))){
a=b=0;
if(ref[0] > 0)
a= mv[0];
if(ref[1] > 0)
b= mv[1];
n++;
}else{
a= mv[0];
b= mv[1];
}
fill_rectangle(&h->mv_cache[0][scan8[i8*4]], 2, 2, 8, a, 4);
fill_rectangle(&h->mv_cache[1][scan8[i8*4]], 2, 2, 8, b, 4);
}
if(!is_b8x8 && !(n&3))
*mb_type= (*mb_type & ~(MB_TYPE_8x8|MB_TYPE_16x8|MB_TYPE_8x16|MB_TYPE_P1L0|MB_TYPE_P1L1))|MB_TYPE_16x16|MB_TYPE_DIRECT2;
}else if(IS_16X16(*mb_type)){
static void x264_cabac_mb_type( x264_t *h, x264_cabac_t *cb )
{
const int i_mb_type = h->mb.i_type;
if( h->sh.b_mbaff &&
(!(h->mb.i_mb_y & 1) || IS_SKIP(h->mb.type[h->mb.i_mb_xy - h->mb.i_mb_stride])) )
{
x264_cabac_encode_decision( cb, 70 + h->mb.cache.i_neighbour_interlaced, h->mb.b_interlaced );
}
if( h->sh.i_type == SLICE_TYPE_I )
{
int ctx = 0;
if( h->mb.i_mb_type_left >= 0 && h->mb.i_mb_type_left != I_4x4 )
{
ctx++;
}
if( h->mb.i_mb_type_top >= 0 && h->mb.i_mb_type_top != I_4x4 )
{
ctx++;
}
x264_cabac_mb_type_intra( h, cb, i_mb_type, 3+ctx, 3+3, 3+4, 3+5, 3+6, 3+7 );
}
else if( h->sh.i_type == SLICE_TYPE_P )
{
/* prefix: 14, suffix: 17 */
if( i_mb_type == P_L0 )
{
if( h->mb.i_partition == D_16x16 )
{
x264_cabac_encode_decision( cb, 14, 0 );
x264_cabac_encode_decision( cb, 15, 0 );
x264_cabac_encode_decision( cb, 16, 0 );
}
else if( h->mb.i_partition == D_16x8 )
{
x264_cabac_encode_decision( cb, 14, 0 );
x264_cabac_encode_decision( cb, 15, 1 );
x264_cabac_encode_decision( cb, 17, 1 );
}
else if( h->mb.i_partition == D_8x16 )
{
x264_cabac_encode_decision( cb, 14, 0 );
x264_cabac_encode_decision( cb, 15, 1 );
x264_cabac_encode_decision( cb, 17, 0 );
}
}
else if( i_mb_type == P_8x8 )
{
x264_cabac_encode_decision( cb, 14, 0 );
x264_cabac_encode_decision( cb, 15, 0 );
x264_cabac_encode_decision( cb, 16, 1 );
}
else /* intra */
{
/* prefix */
x264_cabac_encode_decision( cb, 14, 1 );
/* suffix */
x264_cabac_mb_type_intra( h, cb, i_mb_type, 17+0, 17+1, 17+2, 17+2, 17+3, 17+3 );
}
}
else if( h->sh.i_type == SLICE_TYPE_B )
{
int ctx = 0;
if( h->mb.i_mb_type_left >= 0 && h->mb.i_mb_type_left != B_SKIP && h->mb.i_mb_type_left != B_DIRECT )
{
ctx++;
}
if( h->mb.i_mb_type_top >= 0 && h->mb.i_mb_type_top != B_SKIP && h->mb.i_mb_type_top != B_DIRECT )
{
ctx++;
}
if( i_mb_type == B_DIRECT )
{
x264_cabac_encode_decision( cb, 27+ctx, 0 );
}
else if( i_mb_type == B_8x8 )
{
x264_cabac_encode_decision( cb, 27+ctx, 1 );
x264_cabac_encode_decision( cb, 27+3, 1 );
x264_cabac_encode_decision( cb, 27+4, 1 );
x264_cabac_encode_decision( cb, 27+5, 1 );
x264_cabac_encode_decision( cb, 27+5, 1 );
x264_cabac_encode_decision( cb, 27+5, 1 );
}
else if( IS_INTRA( i_mb_type ) )
{
/* prefix */
x264_cabac_encode_decision( cb, 27+ctx, 1 );
x264_cabac_encode_decision( cb, 27+3, 1 );
x264_cabac_encode_decision( cb, 27+4, 1 );
x264_cabac_encode_decision( cb, 27+5, 1 );
x264_cabac_encode_decision( cb, 27+5, 0 );
x264_cabac_encode_decision( cb, 27+5, 1 );
/* suffix */
x264_cabac_mb_type_intra( h, cb, i_mb_type, 32+0, 32+1, 32+2, 32+2, 32+3, 32+3 );
}
else
{
static const int i_mb_len[9*3] =
{
6, 6, 3, /* L0 L0 */
6, 6, 0, /* L0 L1 */
7, 7, 0, /* L0 BI */
6, 6, 0, /* L1 L0 */
6, 6, 3, /* L1 L1 */
7, 7, 0, /* L1 BI */
7, 7, 0, /* BI L0 */
7, 7, 0, /* BI L1 */
7, 7, 6, /* BI BI */
};
static const int i_mb_bits[9*3][7] =
{
{ 1,1,0,0,0,1 }, { 1,1,0,0,1,0, }, { 1,0,0 }, /* L0 L0 */
{ 1,1,0,1,0,1 }, { 1,1,0,1,1,0 }, {0}, /* L0 L1 */
{ 1,1,1,0,0,0,0 }, { 1,1,1,0,0,0,1 }, {0}, /* L0 BI */
{ 1,1,0,1,1,1 }, { 1,1,1,1,1,0 }, {0}, /* L1 L0 */
{ 1,1,0,0,1,1 }, { 1,1,0,1,0,0 }, { 1,0,1 }, /* L1 L1 */
{ 1,1,1,0,0,1,0 }, { 1,1,1,0,0,1,1 }, {0}, /* L1 BI */
{ 1,1,1,0,1,0,0 }, { 1,1,1,0,1,0,1 }, {0}, /* BI L0 */
{ 1,1,1,0,1,1,0 }, { 1,1,1,0,1,1,1 }, {0}, /* BI L1 */
{ 1,1,1,1,0,0,0 }, { 1,1,1,1,0,0,1 }, { 1,1,0,0,0,0 }, /* BI BI */
};
const int idx = (i_mb_type - B_L0_L0) * 3 + (h->mb.i_partition - D_16x8);
int i;
x264_cabac_encode_decision( cb, 27+ctx, i_mb_bits[idx][0] );
x264_cabac_encode_decision( cb, 27+3, i_mb_bits[idx][1] );
x264_cabac_encode_decision( cb, 27+5-i_mb_bits[idx][1], i_mb_bits[idx][2] );
for( i = 3; i < i_mb_len[idx]; i++ )
x264_cabac_encode_decision( cb, 27+5, i_mb_bits[idx][i] );
}
}
else
{
x264_log(h, X264_LOG_ERROR, "unknown SLICE_TYPE unsupported in x264_macroblock_write_cabac\n" );
}
}
static void FUNC(intra_pred)(HEVCContext *s, int x0, int y0, int log2_size, int c_idx)
{
#define PU(x) \
((x) >> s->sps->log2_min_pu_size)
#define MVF(x, y) \
(s->ref->tab_mvf[(x) + (y) * min_pu_width])
#define MVF_PU(x, y) \
MVF(PU(x0 + ((x) << hshift)), PU(y0 + ((y) << vshift)))
#define IS_INTRA(x, y) \
(MVF_PU(x, y).pred_flag == PF_INTRA)
#define MIN_TB_ADDR_ZS(x, y) \
s->pps->min_tb_addr_zs[(y) * s->sps->min_tb_width + (x)]
#define EXTEND_LEFT(ptr, start, length) \
for (i = (start); i > (start) - (length); i--) \
ptr[i - 1] = ptr[i]
#define EXTEND_RIGHT(ptr, start, length) \
for (i = (start); i < (start) + (length); i++) \
ptr[i] = ptr[i - 1]
#define EXTEND_UP(ptr, start, length) EXTEND_LEFT(ptr, start, length)
#define EXTEND_DOWN(ptr, start, length) EXTEND_RIGHT(ptr, start, length)
#define EXTEND_LEFT_CIP(ptr, start, length) \
for (i = (start); i > (start) - (length); i--) \
if (!IS_INTRA(i - 1, -1)) \
ptr[i - 1] = ptr[i]
#define EXTEND_RIGHT_CIP(ptr, start, length) \
for (i = (start); i < (start) + (length); i++) \
if (!IS_INTRA(i, -1)) \
ptr[i] = ptr[i - 1]
#define EXTEND_UP_CIP(ptr, start, length) \
for (i = (start); i > (start) - (length); i--) \
if (!IS_INTRA(-1, i - 1)) \
ptr[i - 1] = ptr[i]
#define EXTEND_UP_CIP_0(ptr, start, length) \
for (i = (start); i > (start) - (length); i--) \
ptr[i - 1] = ptr[i]
#define EXTEND_DOWN_CIP(ptr, start, length) \
for (i = (start); i < (start) + (length); i++) \
if (!IS_INTRA(-1, i)) \
ptr[i] = ptr[i - 1]
HEVCLocalContext *lc = s->HEVClc;
int i;
int hshift = s->sps->hshift[c_idx];
int vshift = s->sps->vshift[c_idx];
int size = (1 << log2_size);
int size_in_luma = size << hshift;
int size_in_tbs = size_in_luma >> s->sps->log2_min_tb_size;
int x = x0 >> hshift;
int y = y0 >> vshift;
int x_tb = x0 >> s->sps->log2_min_tb_size;
int y_tb = y0 >> s->sps->log2_min_tb_size;
int cur_tb_addr = MIN_TB_ADDR_ZS(x_tb, y_tb);
ptrdiff_t stride = s->frame->linesize[c_idx] / sizeof(pixel);
pixel *src = (pixel*)s->frame->data[c_idx] + x + y * stride;
int min_pu_width = s->sps->min_pu_width;
enum IntraPredMode mode = c_idx ? lc->pu.intra_pred_mode_c :
lc->tu.cur_intra_pred_mode;
pixel left_array[2 * MAX_TB_SIZE + 1];
pixel filtered_left_array[2 * MAX_TB_SIZE + 1];
pixel top_array[2 * MAX_TB_SIZE + 1];
pixel filtered_top_array[2 * MAX_TB_SIZE + 1];
pixel *left = left_array + 1;
pixel *top = top_array + 1;
pixel *filtered_left = filtered_left_array + 1;
pixel *filtered_top = filtered_top_array + 1;
int cand_bottom_left = lc->na.cand_bottom_left && cur_tb_addr > MIN_TB_ADDR_ZS(x_tb - 1, y_tb + size_in_tbs);
int cand_left = lc->na.cand_left;
int cand_up_left = lc->na.cand_up_left;
int cand_up = lc->na.cand_up;
int cand_up_right = lc->na.cand_up_right && cur_tb_addr > MIN_TB_ADDR_ZS(x_tb + size_in_tbs, y_tb - 1);
int bottom_left_size = (FFMIN(y0 + 2 * size_in_luma, s->sps->height) -
(y0 + size_in_luma)) >> vshift;
int top_right_size = (FFMIN(x0 + 2 * size_in_luma, s->sps->width) -
(x0 + size_in_luma)) >> hshift;
if (s->pps->constrained_intra_pred_flag == 1) {
int size_in_luma_pu = PU(size_in_luma);
int on_pu_edge_x = !(x0 & ((1 << s->sps->log2_min_pu_size) - 1));
int on_pu_edge_y = !(y0 & ((1 << s->sps->log2_min_pu_size) - 1));
if (!size_in_luma_pu)
size_in_luma_pu++;
if (cand_bottom_left == 1 && on_pu_edge_x) {
int x_left_pu = PU(x0 - 1);
int y_bottom_pu = PU(y0 + size_in_luma);
int max = FFMIN(size_in_luma_pu, s->sps->min_pu_height - y_bottom_pu);
cand_bottom_left = 0;
for (i = 0; i < max; i++)
cand_bottom_left |= (MVF(x_left_pu, y_bottom_pu + i).pred_flag == PF_INTRA);
}
if (cand_left == 1 && on_pu_edge_x) {
int x_left_pu = PU(x0 - 1);
int y_left_pu = PU(y0);
int max = FFMIN(size_in_luma_pu, s->sps->min_pu_height - y_left_pu);
cand_left = 0;
for (i = 0; i < max; i++)
cand_left |= (MVF(x_left_pu, y_left_pu + i).pred_flag == PF_INTRA);
}
if (cand_up_left == 1) {
int x_left_pu = PU(x0 - 1);
int y_top_pu = PU(y0 - 1);
cand_up_left = MVF(x_left_pu, y_top_pu).pred_flag == PF_INTRA;
}
if (cand_up == 1 && on_pu_edge_y) {
int x_top_pu = PU(x0);
int y_top_pu = PU(y0 - 1);
int max = FFMIN(size_in_luma_pu, s->sps->min_pu_width - x_top_pu);
cand_up = 0;
for (i = 0; i < max; i++)
cand_up |= (MVF(x_top_pu + i, y_top_pu).pred_flag == PF_INTRA);
}
if (cand_up_right == 1 && on_pu_edge_y) {
int y_top_pu = PU(y0 - 1);
int x_right_pu = PU(x0 + size_in_luma);
int max = FFMIN(size_in_luma_pu, s->sps->min_pu_width - x_right_pu);
cand_up_right = 0;
for (i = 0; i < max; i++)
cand_up_right |= (MVF(x_right_pu + i, y_top_pu).pred_flag == PF_INTRA);
}
for (i = 0; i < 2 * MAX_TB_SIZE; i++) {
left[i] = 128;
top[i] = 128;
}
top[-1] = 128;
}
if (cand_bottom_left) {
for (i = size + bottom_left_size; i < (size << 1); i++)
if (IS_INTRA(-1, size + bottom_left_size - 1) ||
!s->pps->constrained_intra_pred_flag)
left[i] = POS(-1, size + bottom_left_size - 1);
for (i = size + bottom_left_size - 1; i >= size; i--)
if (IS_INTRA(-1, i) || !s->pps->constrained_intra_pred_flag)
left[i] = POS(-1, i);
}
if (cand_left)
for (i = size - 1; i >= 0; i--)
if (IS_INTRA(-1, i) || !s->pps->constrained_intra_pred_flag)
left[i] = POS(-1, i);
if (cand_up_left)
if (IS_INTRA(-1, -1) || !s->pps->constrained_intra_pred_flag) {
left[-1] = POS(-1, -1);
top[-1] = left[-1];
}
if (cand_up)
for (i = size - 1; i >= 0; i--)
if (IS_INTRA(i, -1) || !s->pps->constrained_intra_pred_flag)
top[i] = POS(i, -1);
if (cand_up_right) {
for (i = size + top_right_size; i < (size << 1); i++)
if (IS_INTRA(size + top_right_size - 1, -1) ||
!s->pps->constrained_intra_pred_flag)
top[i] = POS(size + top_right_size - 1, -1);
for (i = size + top_right_size - 1; i >= size; i--)
if (IS_INTRA(i, -1) || !s->pps->constrained_intra_pred_flag)
top[i] = POS(i, -1);
}
if (s->pps->constrained_intra_pred_flag == 1) {
if (cand_bottom_left || cand_left || cand_up_left || cand_up || cand_up_right) {
int size_max_x = x0 + ((2 * size) << hshift) < s->sps->width ?
2 * size : (s->sps->width - x0) >> hshift;
int size_max_y = y0 + ((2 * size) << vshift) < s->sps->height ?
2 * size : (s->sps->height - y0) >> vshift;
int j = size + (cand_bottom_left? bottom_left_size: 0) -1;
if (!cand_up_right) {
size_max_x = x0 + ((size) << hshift) < s->sps->width ?
size : (s->sps->width - x0) >> hshift;
}
if (!cand_bottom_left) {
size_max_y = y0 + (( size) << vshift) < s->sps->height ?
size : (s->sps->height - y0) >> vshift;
}
/*!
*************************************************************************************
* \brief
* Mode Decision for a macroblock with error resilience
*************************************************************************************
*/
void encode_one_macroblock_highloss (Macroblock *currMB)
{
int max_index = 9;
int rerun, block, index, mode, i, j, ctr16x16;
char best_pdir;
RD_PARAMS enc_mb;
double min_rdcost = 1e30;
double min_dcost = 1e30;
char best_ref[2] = {0, -1};
int bmcost[5] = {INT_MAX};
int cost=0;
int min_cost = INT_MAX, cost_direct=0, have_direct=0, i16mode=0;
int intra1 = 0;
int cost8x8_direct = 0;
int mb_available_up;
int mb_available_left;
int mb_available_up_left;
int best8x8l0ref, best8x8l1ref;
int is_cavlc = (img->currentSlice->symbol_mode == CAVLC);
short islice = (short) (img->type==I_SLICE);
short bslice = (short) (img->type==B_SLICE);
short pslice = (short) ((img->type==P_SLICE) || (img->type==SP_SLICE));
short intra = (short) (islice || (pslice && img->mb_y==img->mb_y_upd && img->mb_y_upd!=img->mb_y_intra));
int lambda_mf[3];
short runs = (short) (params->RestrictRef==1 && (pslice || (bslice && img->nal_reference_idc>0)) ? 2 : 1);
int prev_mb_nr = FmoGetPreviousMBNr(img->current_mb_nr);
Macroblock* prevMB = (prev_mb_nr >= 0) ? &img->mb_data[prev_mb_nr]:NULL ;
imgpel (*mb_pred)[16] = img->mb_pred[0];
Block8x8Info *b8x8info = img->b8x8info;
short min_chroma_pred_mode, max_chroma_pred_mode;
short inter_skip = 0;
short bipred_me = 0;
double min_rate = 0;
if(params->SearchMode == UM_HEX)
{
UMHEX_decide_intrabk_SAD();
}
else if (params->SearchMode == UM_HEX_SIMPLE)
{
smpUMHEX_decide_intrabk_SAD();
}
intra |= RandomIntra (img->current_mb_nr); // Forced Pseudo-Random Intra
//===== Setup Macroblock encoding parameters =====
init_enc_mb_params(currMB, &enc_mb, intra, bslice);
// Perform multiple encodings if rdopt with losses is enabled
for (rerun=0; rerun<runs; rerun++)
{
if (runs==2)
params->rdopt= (rerun==0) ? 1 : 3;
// reset chroma intra predictor to default
currMB->c_ipred_mode = DC_PRED_8;
//===== S T O R E C O D I N G S T A T E =====
//---------------------------------------------------
store_coding_state (currMB, cs_cm);
if (!intra)
{
//===== set direct motion vectors =====
best_mode = 1;
if (bslice)
{
Get_Direct_Motion_Vectors (currMB);
}
if (params->CtxAdptLagrangeMult == 1)
{
get_initial_mb16x16_cost(currMB);
}
//===== MOTION ESTIMATION FOR 16x16, 16x8, 8x16 BLOCKS =====
for (min_cost=INT_MAX, mode=1; mode<4; mode++)
{
bipred_me = 0;
b8x8info->bipred8x8me[mode][0] = 0;
if (enc_mb.valid[mode])
{
for (cost=0, block=0; block<(mode==1?1:2); block++)
{
update_lambda_costs(&enc_mb, lambda_mf);
PartitionMotionSearch (currMB, mode, block, lambda_mf);
//--- set 4x4 block indizes (for getting MV) ---
j = (block==1 && mode==2 ? 2 : 0);
i = (block==1 && mode==3 ? 2 : 0);
//--- get cost and reference frame for List 0 prediction ---
bmcost[LIST_0] = INT_MAX;
list_prediction_cost(currMB, LIST_0, block, mode, &enc_mb, bmcost, best_ref);
if (bslice)
{
//--- get cost and reference frame for List 1 prediction ---
bmcost[LIST_1] = INT_MAX;
list_prediction_cost(currMB, LIST_1, block, mode, &enc_mb, bmcost, best_ref);
// Compute bipredictive cost between best list 0 and best list 1 references
list_prediction_cost(currMB, BI_PRED, block, mode, &enc_mb, bmcost, best_ref);
// currently Bi prediction ME is only supported for modes 1, 2, 3
if (is_bipred_enabled(mode))
{
list_prediction_cost(currMB, BI_PRED_L0, block, mode, &enc_mb, bmcost, 0);
list_prediction_cost(currMB, BI_PRED_L1, block, mode, &enc_mb, bmcost, 0);
}
else
{
bmcost[BI_PRED_L0] = INT_MAX;
bmcost[BI_PRED_L1] = INT_MAX;
}
// Determine prediction list based on mode cost
determine_prediction_list(mode, bmcost, best_ref, &best_pdir, &cost, &bipred_me);
}
else // if (bslice)
{
best_pdir = 0;
cost += bmcost[LIST_0];
}
assign_enc_picture_params(mode, best_pdir, block, enc_mb.list_offset[LIST_0], best_ref[LIST_0], best_ref[LIST_1], bslice, bipred_me);
//----- set reference frame and direction parameters -----
set_block8x8_info(b8x8info, mode, block, best_ref, best_pdir, bipred_me);
//--- set reference frames and motion vectors ---
if (mode>1 && block==0)
SetRefAndMotionVectors (currMB, block, mode, best_pdir, best_ref[LIST_0], best_ref[LIST_1], bipred_me);
} // for (block=0; block<(mode==1?1:2); block++)
if (cost < min_cost)
{
best_mode = (short) mode;
min_cost = cost;
if (params->CtxAdptLagrangeMult == 1)
{
adjust_mb16x16_cost(cost);
}
}
} // if (enc_mb.valid[mode])
} // for (mode=1; mode<4; mode++)
if (enc_mb.valid[P8x8])
{
giRDOpt_B8OnlyFlag = 1;
tr8x8.mb_p8x8_cost = INT_MAX;
tr4x4.mb_p8x8_cost = INT_MAX;
//===== store coding state of macroblock =====
store_coding_state (currMB, cs_mb);
currMB->all_blk_8x8 = -1;
if (params->Transform8x8Mode)
{
tr8x8.mb_p8x8_cost = 0;
//===========================================================
// Check 8x8 partition with transform size 8x8
//===========================================================
//===== LOOP OVER 8x8 SUB-PARTITIONS (Motion Estimation & Mode Decision) =====
for (cost_direct=cbp8x8=cbp_blk8x8=cnt_nonz_8x8=0, block=0; block<4; block++)
{
submacroblock_mode_decision(&enc_mb, &tr8x8, currMB, cofAC8x8ts[0][block], cofAC8x8ts[1][block], cofAC8x8ts[2][block],
&have_direct, bslice, block, &cost_direct, &cost, &cost8x8_direct, 1, is_cavlc);
set_subblock8x8_info(b8x8info, P8x8, block, &tr8x8);
}
// following params could be added in RD_8x8DATA structure
cbp8_8x8ts = cbp8x8;
cbp_blk8_8x8ts = cbp_blk8x8;
cnt_nonz8_8x8ts = cnt_nonz_8x8;
currMB->luma_transform_size_8x8_flag = 0; //switch to 4x4 transform size
//--- re-set coding state (as it was before 8x8 block coding) ---
//reset_coding_state (currMB, cs_mb);
}// if (params->Transform8x8Mode)
if (params->Transform8x8Mode != 2)
{
tr4x4.mb_p8x8_cost = 0;
//=================================================================
// Check 8x8, 8x4, 4x8 and 4x4 partitions with transform size 4x4
//=================================================================
//===== LOOP OVER 8x8 SUB-PARTITIONS (Motion Estimation & Mode Decision) =====
for (cost_direct=cbp8x8=cbp_blk8x8=cnt_nonz_8x8=0, block=0; block<4; block++)
{
submacroblock_mode_decision(&enc_mb, &tr4x4, currMB, cofAC8x8[block], cofAC8x8CbCr[0][block], cofAC8x8CbCr[1][block],
&have_direct, bslice, block, &cost_direct, &cost, &cost8x8_direct, 0, is_cavlc);
set_subblock8x8_info(b8x8info, P8x8, block, &tr4x4);
}
//--- re-set coding state (as it was before 8x8 block coding) ---
// reset_coding_state (currMB, cs_mb);
}// if (params->Transform8x8Mode != 2)
//--- re-set coding state (as it was before 8x8 block coding) ---
reset_coding_state (currMB, cs_mb);
// This is not enabled yet since mpr has reverse order.
if (params->RCEnable)
rc_store_diff(img->opix_x, img->opix_y, mb_pred);
//check cost for P8x8 for non-rdopt mode
giRDOpt_B8OnlyFlag = 0;
}
else // if (enc_mb.valid[P8x8])
{
tr4x4.mb_p8x8_cost = INT_MAX;
}
// Find a motion vector for the Skip mode
if(pslice)
FindSkipModeMotionVector (currMB);
}
else // if (!intra)
{
min_cost = INT_MAX;
}
//========= C H O O S E B E S T M A C R O B L O C K M O D E =========
//-------------------------------------------------------------------------
{
if ((img->yuv_format != YUV400) && !IS_INDEPENDENT(params))
{
// precompute all new chroma intra prediction modes
IntraChromaPrediction(currMB, &mb_available_up, &mb_available_left, &mb_available_up_left);
if (params->FastCrIntraDecision )
{
IntraChromaRDDecision(currMB, enc_mb);
min_chroma_pred_mode = (short) currMB->c_ipred_mode;
max_chroma_pred_mode = (short) currMB->c_ipred_mode;
}
else
{
min_chroma_pred_mode = DC_PRED_8;
max_chroma_pred_mode = PLANE_8;
}
}
else
{
min_chroma_pred_mode = DC_PRED_8;
max_chroma_pred_mode = DC_PRED_8;
}
for (currMB->c_ipred_mode=min_chroma_pred_mode; currMB->c_ipred_mode<=max_chroma_pred_mode; currMB->c_ipred_mode++)
{
// bypass if c_ipred_mode is not allowed
if ( (img->yuv_format != YUV400) &&
( ((!intra || !params->IntraDisableInterOnly) && params->ChromaIntraDisable == 1 && currMB->c_ipred_mode!=DC_PRED_8)
|| (currMB->c_ipred_mode == VERT_PRED_8 && !mb_available_up)
|| (currMB->c_ipred_mode == HOR_PRED_8 && !mb_available_left)
|| (currMB->c_ipred_mode == PLANE_8 && (!mb_available_left || !mb_available_up || !mb_available_up_left))))
continue;
//===== GET BEST MACROBLOCK MODE =====
for (ctr16x16=0, index=0; index < max_index; index++)
{
mode = mb_mode_table[index];
if (img->yuv_format != YUV400)
{
i16mode = 0;
}
//--- for INTER16x16 check all prediction directions ---
if (mode==1 && bslice)
{
update_prediction_for_mode16x16(b8x8info, ctr16x16, &index);
ctr16x16++;
}
// Skip intra modes in inter slices if best mode is inter <P8x8 with cbp equal to 0
if (params->SkipIntraInInterSlices && !intra && mode >= I4MB && best_mode <=3 && currMB->cbp == 0)
continue;
// check if weights are in valid range for biprediction.
if (bslice && active_pps->weighted_bipred_idc == 1 && mode < P8x8)
{
int cur_blk, cur_comp;
int weight_sum;
Boolean invalid_mode = FALSE;
for (cur_blk = 0; cur_blk < 4; cur_blk ++)
{
if (b8x8info->best8x8pdir[mode][cur_blk] == 2)
{
for (cur_comp = 0; cur_comp < (active_sps->chroma_format_idc == YUV400 ? 1 : 3) ; cur_comp ++)
{
best8x8l0ref = (int) b8x8info->best8x8l0ref[mode][cur_blk];
best8x8l1ref = (int) b8x8info->best8x8l1ref[mode][cur_blk];
weight_sum = wbp_weight[0][best8x8l0ref][best8x8l1ref][cur_comp] + wbp_weight[1][best8x8l0ref][best8x8l1ref][cur_comp];
if (weight_sum < -128 || weight_sum > 127)
{
invalid_mode = TRUE;
break;
}
}
if (invalid_mode == TRUE)
break;
}
}
if (invalid_mode == TRUE)
continue;
}
if (enc_mb.valid[mode])
compute_mode_RD_cost(mode, currMB, &enc_mb, &min_rdcost, &min_dcost, &min_rate, i16mode, bslice, &inter_skip, is_cavlc);
}// for (ctr16x16=0, index=0; index<max_index; index++)
}// for (currMB->c_ipred_mode=DC_PRED_8; currMB->c_ipred_mode<=max_chroma_pred_mode; currMB->c_ipred_mode++)
#ifdef BEST_NZ_COEFF
for (j=0;j<4;j++)
for (i=0; i<(4+img->num_blk8x8_uv); i++)
img->nz_coeff[img->current_mb_nr][j][i] = gaaiMBAFF_NZCoeff[j][i];
#endif
}
if (rerun==0)
intra1 = IS_INTRA(currMB);
} // for (rerun=0; rerun<runs; rerun++)
//===== S E T F I N A L M A C R O B L O C K P A R A M E T E R S ======
//---------------------------------------------------------------------------
update_qp_cbp_tmp(currMB, cbp, best_mode);
set_stored_macroblock_parameters (currMB);
// Rate control
if(params->RCEnable && params->RCUpdateMode <= MAX_RC_MODE)
rc_store_mad(currMB);
update_qp_cbp(currMB, best_mode);
rdopt->min_rdcost = min_rdcost;
rdopt->min_dcost = min_dcost;
if ( (img->MbaffFrameFlag)
&& (img->current_mb_nr%2)
&& (currMB->mb_type ? 0:((bslice) ? !currMB->cbp:1)) // bottom is skip
&& (prevMB->mb_type ? 0:((bslice) ? !prevMB->cbp:1))
&& !(field_flag_inference(currMB) == enc_mb.curr_mb_field)) // top is skip
{
rdopt->min_rdcost = 1e30; // don't allow coding of a MB pair as skip if wrong inference
}
//===== Decide if this MB will restrict the reference frames =====
if (params->RestrictRef)
update_refresh_map(intra, intra1, currMB);
if(params->SearchMode == UM_HEX)
{
UMHEX_skip_intrabk_SAD(best_mode, listXsize[enc_mb.list_offset[LIST_0]]);
}
else if(params->SearchMode == UM_HEX_SIMPLE)
{
smpUMHEX_skip_intrabk_SAD(best_mode, listXsize[enc_mb.list_offset[LIST_0]]);
}
//--- constrain intra prediction ---
if(params->UseConstrainedIntraPred && (img->type==P_SLICE || img->type==B_SLICE))
{
img->intra_block[img->current_mb_nr] = IS_INTRA(currMB);
}
}