>Yup, indeed. I presume xmms does indeed use floating-point maths in
>its decoding. Would a quick antidote to be to compile with the soft-
>float library, or are the ramifications going to be too horrible?
That's unlikely to gain you the kind of performance you need, unfortunately.
The soft-flo
>It's possible whatever Frank Smith fixed (`linker segfault') here:
>
>is one and the same, if this helps.
It's hard to tell. There's no obvious sign of any change at all in the source
code, and I can't find a copy of the original 2.9.5.0.37 to compare against.
p.
>Has anyone worked on any optimised ARM routines for MP3 decoding in xmms?
>I notice it's not fast enough as it stands on a 200MHz StrongARM; this seems
>pretty bad to me.
The time-consuming part of MPEG decode is usually the DCT. There is a
fixed-point implementation of this in Nicolas Pitre's
3 matches
Mail list logo
