Custom Query (1442 matches)
Results (19 - 21 of 1442)
| Ticket | Resolution | Summary | Owner | Reporter |
|---|---|---|---|---|
| #1369 | fixed | validity check for cb_qp_offset_list [ i ] and cr_qp_offset_list[ i ] | kolya | |
| Description |
assert should be added for check if values are inside the range, pls see the patch attached |
|||
| #138 | fixed | valgrind failures in HM-2.1, HM-2.2 and HM-3.0-dev | davidf | thomasd |
| Description |
There are numerous valgrind errors when running these versions, based on values not being initialised, apparently in EncIntraHeader (TEncSearch.cpp). HM-2.0 is free of such errors. HM-2.0-dev-mediatek seems to have introduced at least some of these errors. |
|||
| #249 | fixed | valgrind error in ALF decoding: | thomasd | |
| Description |
Applies to version HM-5.0 and HM-5.0rc1. When decoding a sequence with ALF, valgrind reports the use of uninitialised values in TComAdaptiveLoopFilter::calcVar, line 1378. No MD5 mismatches have been found yet. Encoder command line: -c encoder_lowdelay.cfg -c RaceHorsesC.cfg -i RaceHorses_832x480_30.yuv -b ./test.bit -o ./test.yuv --FramesToBeEncoded=10 --ALF=1 --SAO=0 This is the new version of calcVar introduced as a result of G609. The issue goes away if macro G609_NEW_BA_SUB is set to 0. It's possibly due to the use of a local array with padding not properly initialised. Setting the loop parameters to m_img_{height,width} - 3 instead of m_img_{height,width} + 3 removes the problem, but produces a different bit stream. It's unclear to me what the correct bounds should be. |
|||
