--- branches/dev-api-4/xvidcore/TODO 2003/11/24 22:06:19 1224 +++ branches/dev-api-4/xvidcore/TODO 2003/12/05 14:44:35 1242 @@ -16,18 +16,19 @@ dshow are included. * update/fix CBR plugin - misses target bitrate, bitrate burst in static motion/high motion - transistions -* bug hunting (ed.gomez/sysKin) + transitions +* bug hunting (ed.gomez/syskin) - xvidcore is still a place where bugs feel confortable -* quant matrix usage (not thread nor instance safe) - - i (ed.gomez) started an API change to matrix related functions - (quant_mpeg_xxx). But the remaining problem is that matrices should - be put in FRAMEINFO structures, and matrices should be pushed down - through many function calls. I had no idea how to do that w/o - hacking nastly. -* docbook writing (ed.gomez) +* docbook writing (every developer should contribute) - New API has to be documented. * two pass code should not introduce quality regressions (every tester) + - first feedback proved kfthresholding was hurting quality, a + decision on its behavior has to be settled. +* YV12/I420/USER clarification + - item submited by Ivan Kalvachev + Planes are swaped +* MMX MPEG4 quantization is unprecise. This is probably an inherent + problem in some MMX opcode precision. Already completed items: @@ -44,12 +45,14 @@ * rawdec (use xvid_decraw instead) (ed.gomez) * Support for GMC 3 warp points (christoph) * New Qpel code (michael) -* ME splitting and ME improvements (sysKin) +* ME splitting and ME improvements (syskin) * New unix build process (ed.gomez) * New thread/instance safe sse2 code (p.massimino) * INSTALL guide for Unix and Win32 (ed.gomez) * dshow static link to libxvidcore.lib (peter) -* update/fix Lumimasking (sysKin) +* update/fix Lumimasking (syskin) * trellis for mpeg and relaxed optimization for big levels (skal) +* thread safe mpeg quantizing (michael) +* Interlacing for bvop and svop (syskin) -Last edited: $Date: 2003-11-24 22:06:19 $ +Last edited: $Date: 2003-12-05 14:44:35 $