Oh. It works only with ALL of the MT code removed, even Distributor().
Fair enough, however I'd thought "threads=1" might be OK.
I must have misinterpreted the manual "2) multiple instances of the filter can be created" and thought using MT0.7 multithreading would be OK.
In my case, opening and doing things inside DGMultiSource I'd expect to be comparatively quick, however anything in code after that, eg deblocking and limitedsharpenfaster and re-interlacing, would not benefit from multithreading.
Do you think it's a NV type bug or a HC type bug (Vdub playing up seems to suggest NV) or even a bug at all ? Happy to report it elsewhere, just let me know.
|