Fix vp10_accumulate_frame_counts once and for all.
This ensures the multi-threaded and single-threaded encoder/decoder always uses the same probability contexts. Change-Id: I6f1e7c6bd8808c390c1dc0a628ae97db3acedf6d
This commit is contained in:
@@ -3491,7 +3491,7 @@ static const uint8_t *decode_tiles_mt(VP10Decoder *pbi,
|
||||
if (cm->refresh_frame_context == REFRESH_FRAME_CONTEXT_BACKWARD) {
|
||||
for (i = 0; i < num_workers; ++i) {
|
||||
TileWorkerData *const twd = (TileWorkerData*)pbi->tile_workers[i].data1;
|
||||
vp10_accumulate_frame_counts(cm, &twd->counts, 1);
|
||||
vp10_accumulate_frame_counts(cm, &twd->counts);
|
||||
}
|
||||
}
|
||||
|
||||
|
Reference in New Issue
Block a user