It sounds like either the wmapro mdct has additional pre/post steps compared to codeclib (or alternatively is lacking codeclibs pre/post steps and doing those steps elsewhere), or else wmapro has different expectations/assumptions compared to codeclib (e.g. codeclib has strict requirements over input output buffers not overlapping etc). I'd be surprised if codeclib imdct has a bug, as such, that only affects decoding wmapro frames.
On 11 Jul 2010 18:33, "Mohamed Tarek" <[email protected]> wrote: On Sun, Jul 11, 2010 at 2:51 PM, Mohamed Tarek <[email protected]> wrote: > > > Yes I might need to... I tried that and got the exact same faulty output. I tried copying the fixed point coefficients into the floating point ones passing those to the float imdct and then copying back the float coeffs into the fixed coeffs just after, and the sound quality was perfect, so it seems it's actually a problem with the mdct. (both codeclib's and libwmapro's) > > Also, there seems to be something wrong with the patch, I'll prepare another one tonight. > I take that back. The patch applied and compiled cleanly against a fresh checkout. -- MT
