static int h261_decode_mb(H261Context *h){
MpegEncContext * const s = &h->s;
int i, cbp, xy;
cbp = 63;
// Read mba
do{
h->mba_diff = get_vlc2(&s->gb, h261_mba_vlc.table, H261_MBA_VLC_BITS, 2);
/* Check for slice end */
/* NOTE: GOB can be empty (no MB data) or exist only of MBA_stuffing */
if (h->mba_diff == MBA_STARTCODE){ // start code
h->gob_start_code_skipped = 1;
return SLICE_END;
}
}
while( h->mba_diff == MBA_STUFFING ); // stuffing
if ( h->mba_diff < 0 ){
if ( get_bits_count(&s->gb) + 7 >= s->gb.size_in_bits )
return SLICE_END;
av_log(s->avctx, AV_LOG_ERROR, "illegal mba at %d %d\n", s->mb_x, s->mb_y);
return SLICE_ERROR;
}
h->mba_diff += 1;
h->current_mba += h->mba_diff;
if ( h->current_mba > MBA_STUFFING )
return SLICE_ERROR;
s->mb_x= ((h->gob_number-1) % 2) * 11 + ((h->current_mba-1) % 11);
s->mb_y= ((h->gob_number-1) / 2) * 3 + ((h->current_mba-1) / 11);
xy = s->mb_x + s->mb_y * s->mb_stride;
ff_init_block_index(s);
ff_update_block_index(s);
// Read mtype
h->mtype = get_vlc2(&s->gb, h261_mtype_vlc.table, H261_MTYPE_VLC_BITS, 2);
h->mtype = h261_mtype_map[h->mtype];
// Read mquant
if ( IS_QUANT ( h->mtype ) ){
ff_set_qscale(s, get_bits(&s->gb, 5));
}
s->mb_intra = IS_INTRA4x4(h->mtype);
// Read mv
if ( IS_16X16 ( h->mtype ) ){
// Motion vector data is included for all MC macroblocks. MVD is obtained from the macroblock vector by subtracting the
// vector of the preceding macroblock. For this calculation the vector of the preceding macroblock is regarded as zero in the
// following three situations:
// 1) evaluating MVD for macroblocks 1, 12 and 23;
// 2) evaluating MVD for macroblocks in which MBA does not represent a difference of 1;
// 3) MTYPE of the previous macroblock was not MC.
if ( ( h->current_mba == 1 ) || ( h->current_mba == 12 ) || ( h->current_mba == 23 ) ||
( h->mba_diff != 1))
{
h->current_mv_x = 0;
h->current_mv_y = 0;
}
h->current_mv_x= decode_mv_component(&s->gb, h->current_mv_x);
h->current_mv_y= decode_mv_component(&s->gb, h->current_mv_y);
}else{
h->current_mv_x = 0;
h->current_mv_y = 0;
}
// Read cbp
if ( HAS_CBP( h->mtype ) ){
cbp = get_vlc2(&s->gb, h261_cbp_vlc.table, H261_CBP_VLC_BITS, 2) + 1;
}
if(s->mb_intra){
s->current_picture.mb_type[xy]= MB_TYPE_INTRA;
goto intra;
}
//set motion vectors
s->mv_dir = MV_DIR_FORWARD;
s->mv_type = MV_TYPE_16X16;
s->current_picture.mb_type[xy]= MB_TYPE_16x16 | MB_TYPE_L0;
s->mv[0][0][0] = h->current_mv_x * 2;//gets divided by 2 in motion compensation
s->mv[0][0][1] = h->current_mv_y * 2;
intra:
/* decode each block */
if(s->mb_intra || HAS_CBP(h->mtype)){
s->dsp.clear_blocks(s->block[0]);
for (i = 0; i < 6; i++) {
if (h261_decode_block(h, s->block[i], i, cbp&32) < 0){
return SLICE_ERROR;
}
cbp+=cbp;
}
}else{
for (i = 0; i < 6; i++)
s->block_last_index[i]= -1;
}
MPV_decode_mb(s, s->block);
return SLICE_OK;
}
void render_mbs()
{
H264Context *h = g_h;
GPUH264Context * const g = &h->gpu;
MpegEncContext * const s = &h->s;
H264mb* blockStore = g->block_buffer;
int i, l;
int lists = (h->slice_type==FF_B_TYPE)?2:1;
int dpb_pos = s->current_picture.gpu_dpb;
printf("Attempting to motion compensate %d blocks\n", (g->end-g->start+1));
glNewList(dispList, GL_COMPILE);
for(l=0; l < lists; l++)
{
glBegin(GL_QUADS);
for(i= g->start; i <= g->end; i++)
{
const int mb_x = blockStore[i].mb_x;
const int mb_y = blockStore[i].mb_y;
const int mb_xy = mb_x + mb_y*s->mb_stride;
const int mb_type = s->current_picture.mb_type[mb_xy];
int mv_x, mv_y, j;
//RUDD TODO ignoring Intra blocks for now
if(IS_INTER(mb_type))
{
if(IS_16X16(mb_type) && IS_DIR(mb_type, 0, l))
{
mv_x = blockStore[i].mv_cache[0][ scan8[0]][0];
mv_y = blockStore[i].mv_cache[0][ scan8[0]][1];
render_one_block(mb_x, mb_y, mv_x, mv_y, 16, 16, 0, 0,
h->ref_list[l][h->ref_cache[l][ scan8[0] ]].gpu_dpb, dpb_pos);
}
else if(IS_16X8(mb_type))
{
if(IS_DIR(mb_type, 0, l))
{
mv_x = blockStore[i].mv_cache[l][ scan8[0]][0];
mv_y = blockStore[i].mv_cache[l][ scan8[0]][1];
render_one_block(mb_x, mb_y, mv_x, mv_y, 16, 8, 0, 0,
h->ref_list[l][h->ref_cache[l][ scan8[0] ]].gpu_dpb, dpb_pos);
}
if(IS_DIR(mb_type, 1, l))
{
mv_x = blockStore[i].mv_cache[0][ scan8[8]][0];
mv_y = blockStore[i].mv_cache[0][ scan8[8]][1];
render_one_block(mb_x, mb_y, mv_x, mv_y, 16, 8, 0, -8,
h->ref_list[l][h->ref_cache[l][ scan8[8] ]].gpu_dpb, dpb_pos);
}
}
else if(IS_8X16(mb_type))
{
if(IS_DIR(mb_type, 0, l))
{
mv_x = blockStore[i].mv_cache[0][ scan8[0]][0];
mv_y = blockStore[i].mv_cache[0][ scan8[0]][1];
render_one_block(mb_x, mb_y, mv_x, mv_y, 8, 16, 0, 0,
h->ref_list[l][h->ref_cache[l][ scan8[0] ]].gpu_dpb, dpb_pos);
}
if(IS_DIR(mb_type, 1, l))
{
mv_x = blockStore[i].mv_cache[0][ scan8[4]][0];
mv_y = blockStore[i].mv_cache[0][ scan8[4]][1];
render_one_block(mb_x, mb_y, mv_x, mv_y, 8, 16, 8, 0,
h->ref_list[l][h->ref_cache[l][ scan8[4] ]].gpu_dpb, dpb_pos);
}
}
else
{
assert(IS_8X8(mb_type));
int j;
for(j=0;j<4;j++)
{
const int sub_mb_type= h->sub_mb_type[j];
const int n= 4*j;
int x_offset= (j&1);
int y_offset= (j&2)>>1;
if(!IS_DIR(sub_mb_type, 0, l))
continue;
if(IS_SUB_8X8(sub_mb_type))
{
mv_x = blockStore[i].mv_cache[0][ scan8[n]][0];
mv_y = blockStore[i].mv_cache[0][ scan8[n]][1];
render_one_block(mb_x, mb_y, mv_x, mv_y, 8, 8, 8*x_offset,-8*y_offset,
h->ref_list[l][h->ref_cache[l][ scan8[n] ]].gpu_dpb, dpb_pos);
}
else if(IS_SUB_8X4(sub_mb_type))
{
mv_x = blockStore[i].mv_cache[0][ scan8[n]][0];
mv_y = blockStore[i].mv_cache[0][ scan8[n]][1];
render_one_block(mb_x, mb_y, mv_x, mv_y, 8, 4, 8*x_offset,-8*y_offset,
h->ref_list[l][h->ref_cache[l][ scan8[n] ]].gpu_dpb, dpb_pos);
mv_x = blockStore[i].mv_cache[0][ scan8[n+2]][0];
mv_y = blockStore[i].mv_cache[0][ scan8[n+2]][1];
render_one_block(mb_x, mb_y, mv_x, mv_y, 8, 4, 8*x_offset,-8*y_offset-4,
h->ref_list[l][h->ref_cache[l][ scan8[n+2] ]].gpu_dpb, dpb_pos);
}
else if(IS_SUB_4X8(sub_mb_type))
{
mv_x = blockStore[i].mv_cache[0][ scan8[n]][0];
mv_y = blockStore[i].mv_cache[0][ scan8[n]][1];
render_one_block(mb_x, mb_y, mv_x, mv_y, 4, 8, 8*x_offset,-8*y_offset,
h->ref_list[l][h->ref_cache[l][ scan8[n] ]].gpu_dpb, dpb_pos);
mv_x = blockStore[i].mv_cache[0][ scan8[n+1]][0];
mv_y = blockStore[i].mv_cache[0][ scan8[n+1]][1];
render_one_block(mb_x, mb_y, mv_x, mv_y, 4, 8, 8*x_offset+4,-8*y_offset,
h->ref_list[l][h->ref_cache[l][ scan8[n+1] ]].gpu_dpb, dpb_pos);
}
else
{
mv_x = blockStore[i].mv_cache[0][ scan8[n]][0];
mv_y = blockStore[i].mv_cache[0][ scan8[n]][1];
render_one_block(mb_x, mb_y, mv_x, mv_y, 4, 4, 8*x_offset,-8*y_offset,
h->ref_list[l][h->ref_cache[l][ scan8[n] ]].gpu_dpb, dpb_pos);
mv_x = blockStore[i].mv_cache[0][ scan8[n+1]][0];
mv_y = blockStore[i].mv_cache[0][ scan8[n+1]][1];
render_one_block(mb_x, mb_y, mv_x, mv_y, 4, 4, 8*x_offset+4,-8*y_offset,
h->ref_list[l][h->ref_cache[l][ scan8[n+1] ]].gpu_dpb, dpb_pos);
mv_x = blockStore[i].mv_cache[0][ scan8[n+2]][0];
mv_y = blockStore[i].mv_cache[0][ scan8[n+2]][1];
render_one_block(mb_x, mb_y, mv_x, mv_y, 4, 4, 8*x_offset,-8*y_offset-4,
h->ref_list[l][h->ref_cache[l][ scan8[n+2] ]].gpu_dpb, dpb_pos);
mv_x = blockStore[i].mv_cache[0][ scan8[n+3]][0];
mv_y = blockStore[i].mv_cache[0][ scan8[n+3]][1];
render_one_block(mb_x, mb_y, mv_x, mv_y, 4, 4, 8*x_offset+4,-8*y_offset-4,
h->ref_list[l][h->ref_cache[l][ scan8[n+3] ]].gpu_dpb, dpb_pos);
}
}
}
}
}
}
glEnd();
glEndList();
}
static void MCFUNC(hl_motion)(const H264Context *h, H264SliceContext *sl,
uint8_t *dest_y,
uint8_t *dest_cb, uint8_t *dest_cr,
qpel_mc_func(*qpix_put)[16],
const h264_chroma_mc_func(*chroma_put),
qpel_mc_func(*qpix_avg)[16],
const h264_chroma_mc_func(*chroma_avg),
const h264_weight_func *weight_op,
const h264_biweight_func *weight_avg)
{
const int mb_xy = sl->mb_xy;
const int mb_type = h->cur_pic.mb_type[mb_xy];
av_assert2(IS_INTER(mb_type));
if (HAVE_THREADS && (h->avctx->active_thread_type & FF_THREAD_FRAME))
await_references(h, sl);
prefetch_motion(h, sl, 0, PIXEL_SHIFT, CHROMA_IDC);
if (IS_16X16(mb_type)) {
mc_part(h, sl, 0, 1, 16, 0, dest_y, dest_cb, dest_cr, 0, 0,
qpix_put[0], chroma_put[0], qpix_avg[0], chroma_avg[0],
weight_op, weight_avg,
IS_DIR(mb_type, 0, 0), IS_DIR(mb_type, 0, 1));
} else if (IS_16X8(mb_type)) {
mc_part(h, sl, 0, 0, 8, 8 << PIXEL_SHIFT, dest_y, dest_cb, dest_cr, 0, 0,
qpix_put[1], chroma_put[0], qpix_avg[1], chroma_avg[0],
weight_op, weight_avg,
IS_DIR(mb_type, 0, 0), IS_DIR(mb_type, 0, 1));
mc_part(h, sl, 8, 0, 8, 8 << PIXEL_SHIFT, dest_y, dest_cb, dest_cr, 0, 4,
qpix_put[1], chroma_put[0], qpix_avg[1], chroma_avg[0],
weight_op, weight_avg,
IS_DIR(mb_type, 1, 0), IS_DIR(mb_type, 1, 1));
} else if (IS_8X16(mb_type)) {
mc_part(h, sl, 0, 0, 16, 8 * sl->mb_linesize, dest_y, dest_cb, dest_cr, 0, 0,
qpix_put[1], chroma_put[1], qpix_avg[1], chroma_avg[1],
&weight_op[1], &weight_avg[1],
IS_DIR(mb_type, 0, 0), IS_DIR(mb_type, 0, 1));
mc_part(h, sl, 4, 0, 16, 8 * sl->mb_linesize, dest_y, dest_cb, dest_cr, 4, 0,
qpix_put[1], chroma_put[1], qpix_avg[1], chroma_avg[1],
&weight_op[1], &weight_avg[1],
IS_DIR(mb_type, 1, 0), IS_DIR(mb_type, 1, 1));
} else {
int i;
av_assert2(IS_8X8(mb_type));
for (i = 0; i < 4; i++) {
const int sub_mb_type = sl->sub_mb_type[i];
const int n = 4 * i;
int x_offset = (i & 1) << 2;
int y_offset = (i & 2) << 1;
if (IS_SUB_8X8(sub_mb_type)) {
mc_part(h, sl, n, 1, 8, 0, dest_y, dest_cb, dest_cr,
x_offset, y_offset,
qpix_put[1], chroma_put[1], qpix_avg[1], chroma_avg[1],
&weight_op[1], &weight_avg[1],
IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1));
} else if (IS_SUB_8X4(sub_mb_type)) {
mc_part(h, sl, n, 0, 4, 4 << PIXEL_SHIFT, dest_y, dest_cb, dest_cr,
x_offset, y_offset,
qpix_put[2], chroma_put[1], qpix_avg[2], chroma_avg[1],
&weight_op[1], &weight_avg[1],
IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1));
mc_part(h, sl, n + 2, 0, 4, 4 << PIXEL_SHIFT,
dest_y, dest_cb, dest_cr, x_offset, y_offset + 2,
qpix_put[2], chroma_put[1], qpix_avg[2], chroma_avg[1],
&weight_op[1], &weight_avg[1],
IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1));
} else if (IS_SUB_4X8(sub_mb_type)) {
mc_part(h, sl, n, 0, 8, 4 * sl->mb_linesize,
dest_y, dest_cb, dest_cr, x_offset, y_offset,
qpix_put[2], chroma_put[2], qpix_avg[2], chroma_avg[2],
&weight_op[2], &weight_avg[2],
IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1));
mc_part(h, sl, n + 1, 0, 8, 4 * sl->mb_linesize,
dest_y, dest_cb, dest_cr, x_offset + 2, y_offset,
qpix_put[2], chroma_put[2], qpix_avg[2], chroma_avg[2],
&weight_op[2], &weight_avg[2],
IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1));
} else {
int j;
av_assert2(IS_SUB_4X4(sub_mb_type));
for (j = 0; j < 4; j++) {
int sub_x_offset = x_offset + 2 * (j & 1);
int sub_y_offset = y_offset + (j & 2);
mc_part(h, sl, n + j, 1, 4, 0,
dest_y, dest_cb, dest_cr, sub_x_offset, sub_y_offset,
qpix_put[2], chroma_put[2], qpix_avg[2], chroma_avg[2],
&weight_op[2], &weight_avg[2],
IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1));
}
}
}
}
if (USES_LIST(mb_type, 1))
prefetch_motion(h, sl, 1, PIXEL_SHIFT, CHROMA_IDC);
}
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));
}
}
void motion_execute(H264_Slice_GlbARGs *SLICE_T, H264_MB_Ctrl_DecARGs *dmb, uint8_t *recon_buf,
uint8_t *motion_dha)
{
uint8_t *motion_douty, *motion_doutc;
uint8_t *motion_dsa = motion_dha + 0x108;
const int mb_type= dmb->mb_type;
motion_douty = recon_buf;
motion_doutc = recon_buf + PREVIOUS_OFFSET_U;
SET_REG1_DSTA(TCSM1_PADDR((int)motion_douty));
SET_REG1_DSA(TCSM1_PADDR((int)motion_dsa));
SET_REG2_DSTA(TCSM1_PADDR((int)motion_doutc));
SET_REG2_DSA(TCSM1_PADDR((int)motion_dsa));
volatile int *tdd = (int *)motion_dha;
int tkn = 0;
motion_dsa[0] = 0x0;
tdd++;
if(IS_16X16(mb_type)){
motion_task(SLICE_T, dmb, 0, 0, tdd, &tkn,
IS_DIR(mb_type, 0, 0), IS_DIR(mb_type, 0, 1),
3/*blkh*/, 3/*blkw*/, 0/*boy*/, 0/*box*/);
}else if(IS_16X8(mb_type)){
motion_task(SLICE_T, dmb, 0, 0, tdd, &tkn,
IS_DIR(mb_type, 0, 0), IS_DIR(mb_type, 0, 1),
2/*blkh*/, 3/*blkw*/, 0/*boy*/, 0/*box*/);
motion_task(SLICE_T, dmb, 1, 2, tdd, &tkn,
IS_DIR(mb_type, 1, 0), IS_DIR(mb_type, 1, 1),
2/*blkh*/, 3/*blkw*/, 2/*boy*/, 0/*box*/);
}else if(IS_8X16(mb_type)){
motion_task(SLICE_T, dmb, 0, 0, tdd, &tkn,
IS_DIR(mb_type, 0, 0), IS_DIR(mb_type, 0, 1),
3/*blkh*/, 2/*blkw*/, 0/*boy*/, 0/*box*/);
motion_task(SLICE_T, dmb, 1, 1, tdd, &tkn,
IS_DIR(mb_type, 1, 0), IS_DIR(mb_type, 1, 1),
3/*blkh*/, 2/*blkw*/, 0/*boy*/, 2/*box*/);
}else{
int i;
int mv_n=0;
for(i=0; i<4; i++){
const int sub_mb_type= dmb->sub_mb_type[i];
if(IS_SUB_8X8(sub_mb_type)){
motion_task(SLICE_T, dmb, mv_n, i, tdd, &tkn,
IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1),
2/*blkh*/, 2/*blkw*/, (i & 0x2)/*boy*/, (i & 0x1)*2/*box*/);
mv_n++;
}else if(IS_SUB_8X4(sub_mb_type)){
motion_task(SLICE_T, dmb, mv_n, i, tdd, &tkn,
IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1),
1/*blkh*/, 2/*blkw*/, (i & 0x2)/*boy*/, (i & 0x1)*2/*box*/);
mv_n++;
motion_task(SLICE_T, dmb, mv_n, i, tdd, &tkn,
IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1),
1/*blkh*/, 2/*blkw*/, (i & 0x2)+1/*boy*/, (i & 0x1)*2/*box*/);
mv_n++;
}else if(IS_SUB_4X8(sub_mb_type)){
motion_task(SLICE_T, dmb, mv_n, i, tdd, &tkn,
IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1),
2/*blkh*/, 1/*blkw*/, (i & 0x2)/*boy*/, (i & 0x1)*2/*box*/);
mv_n++;
motion_task(SLICE_T, dmb, mv_n, i, tdd, &tkn,
IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1),
2/*blkh*/, 1/*blkw*/, (i & 0x2)/*boy*/, (i & 0x1)*2+1/*box*/);
mv_n++;
}else{
int j;
for(j=0; j<4; j++){
motion_task(SLICE_T, dmb, mv_n, i, tdd, &tkn,
IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1),
1/*blkh*/, 1/*blkw*/,
(i & 0x2) + (j & 0x2)/2/*boy*/, (i & 0x1)*2 + (j & 0x1)/*box*/);
mv_n++;
} //j
} //BLK4X4
} //i
} //BLK8X8
tdd[2*tkn-1] |= 0x1<<TDD_DOE_SFT;
tdd[-1] = TDD_HEAD(1,/*vld*/
1,/*lk*/
0,/*sync*/
1,/*ch1pel*/
2,/*ch2pel*/
TDD_POS_SPEC,/*posmd*/
TDD_MV_AUTO,/*mvmd*/
1,/*ch2en*/
tkn,/*tkn*/
dmb->mb_y,/*mby*/
dmb->mb_x/*mbx*/);
tdd[2*tkn] = TDD_HEAD(1,/*vld*/
0,/*lk*/
1,/*sync*/
1,/*ch1pel*/
2,/*ch2pel*/
TDD_POS_SPEC,/*posmd*/
TDD_MV_AUTO,/*mvmd*/
1,/*ch2en*/
0,/*tkn*/
0xFF,/*mby*/
0xFF/*mbx*/);
}