[Mplayer-cvslog] CVS: main/libaf af_comp.c,NONE,1.1 af_gate.c,NONE,1.1 af_pan.c,NONE,1.1 af_resample.h,NONE,1.1 af_tools.c,NONE,1.1 Makefile,1.6,1.7 af.c,1.19,1.20 af.h,1.12,1.13 af_channels.c,1.8,1.9 af_delay.c,1.8,1.9 af_equalizer.c,1.2,1.3 af_format.c,1.9,1.10 af_resample.c,1.15,1.16 af_volume.c,1.5,1.6 control.h,1.3,1.4

[Anders Johansson]

> On Sat, Jan 04, 2003 at 04:19:38PM +0800, Anders Johansson wrote:
> > > On Sat, Dec 28, 2002 at 02:59:56PM +0100, Anders Johansson CVS wrote:
> > > > Update of /cvsroot/mplayer/main/libaf
> > > > In directory mail:/var/tmp.root/cvs-serv18447
> > > > 
> > > > Modified Files:
> > > > 	Makefile af.c af.h af_channels.c af_delay.c af_equalizer.c 
> > > > 	af_format.c af_resample.c af_volume.c control.h 
> > > > Added Files:
> > > > 	af_comp.c af_gate.c af_pan.c af_resample.h af_tools.c 
> > > > Log Message:
> > > > Changes includes:
> > > > - Improved runtime control system 
> > > > - 3 New filter panning, compressor/limiter and a noise gate
> > > >   - The compressor/limiter and the noise gate are not yet finished
> > > >   - The panning filter does combined mixing and channel routing and
> > > >     can be used to down-mix from stereo to mono (for example)
> > > > - Improvements to volume and channel
> > > >   - volume now has a very good soft clipping using sin()
> > > >   - channel can handle generic routing of audio data 
> > > > - Conversion of all filters to handle floating point data
> > > > - Cleanup of message printing
> > > > - Fix for the sig 11 bug reported by Denes
> > > 
> > > This change broke resampling entirely on my system (K6). Resampling
> > > from 48000 down to 44100 or anything else I try produces nothing but
> > > noise! I'll look into it myself, but if you have any ideas right off
> > > go ahead and fix asap!
> > 
> > Hi, can not reproduce pls send your cmdline switches and output from -v
> > switch and I'll have a better chance of figuring out what's wrong.
> 
> Well, -v helped make it clear that something is totally broken, but I
> still have no idea how to fix it. Apparently all sorts of
> int->float->int nonsense is taking place which would make it unusably
> slow even if it worked. Take a look:
> 

It is the way it is supposed to work. libaf detects that your machine
is fast enough to use floating point processing and converts the audio
data to float before it is processed. Unfortunately your sound card
driver does not accept float so it is converted back to int. I have
tested this on a AMD K6/300 (my devel machine) and it uses very little
CPU (about the same as the volume control). Anyway something goes
wrong in your machine (not sure what) to work around you could use
-af-adv force=1 and it will use fixpoint processing instead see
sound.html for more info. I'd like to solve this problem, not sure
what to continue with try -af format=4:float,format=2:int with a lo
sample frequency file, and see what happens. I'll read mail agin on monday.

Cheers,
//Anders