Re: [fluid-dev] Re: Changes checked into CVS

[Mihail Zenkov]

On Sun, 10 Dec 2006 16:28:52 +0000
Josh Green <address@hidden> wrote:

> Your dithering patch at this point isn't complete, so I wanted to get it
> right before next release :)  Therefore, returning to this discussion..
> 
> On Fri, 2006-11-24 at 21:19 +0200, Mihail Zenkov wrote:
> > 
> > Yes you right, but in case without roundf we have two time bigger
> > amplitude and imho bigger THD (i see it when compare spectrum in
> > audacity). See attach to more undersend.
> 
> Sounds like it would be good to use roundf.  I'll need to add
> a ./configure detection for it though, since I imagine it wont be
> available on Windows and possibly OSX?  Not sure.  I was thinking of
> just making a macro like:
> 
> #define FLUID_ROUNDF   roundf
> 
> Which gets defined if roundf is truly available.  Any ideas of a
> semi-efficient equivalent that could be used in place of this macro if
> roundf() is not available?  I suppose it does not need to be as good as
> roundf, just as long as it is improving things in regards to dithering.

IMHO better replace roundf with this code in all case:

inline int roundi(float x)
{
        if (x >= 0.0f)
                return (int)(x+0.5f);
        else
                return (int)(x-0.5f);
}

It slightly faster then gcc roundf (1.5 time).