[FFmpeg-devel] [RFC] fast audio interleaving

[Rich Felker]

On Tue, Sep 25, 2007 at 08:54:51PM +0200, Aurelien Jacobs wrote:
> Benjamin Larsson wrote:
> 
> > Hi, lots of audio codecs duplicate the interleaving of left and right
> > samples when outputting stereo. Wouldn't it make sense to dsp optimize
> > just this operation ? I'm sure there are some SIMD operations that do
> > exactly what we need.
> 
> One nice solution (IIRC already proposed by Michael long ago) would be
> to allow audio codecs to output a planar format. Then if the soundcard
> don't support planar format (most probable), it should be converted
> to packed format using some independent code (IOW a libswscale for audio).
> And indeed that looks like something that can be optimized using SIMD.

I agree strongly with this proposed approach. "Planar" format is
preferred anyway for most filtering operations, and makes everything
audio-channel-order-agnostic just like planar video makes things
color-channel-order-agnostic.

Rich