[PD-dev] Non-branching clipping algo...

[cgc at humboldtblvd.com]

Check out the SIMD instructions (SSE,MMX,Altivec,etc) for your CPU.  I know that
Altivec has specific bounds checking operations and also has bitwise comparisons
to make a mask for the vector.  For example, I've used the latter to eliminate
branching in video keying functions with good results.

cgc

Quoting Thomas Grill <t.grill at gmx.net>:

> Hi Luigi,
> i'm not sure if it's really necessary to avoid branching in any case.
> All of the new CPUs have a branch prediction that keeps the
> instruction
> queues filled, so it's maybe only about hinting the CPU in the right
> direction.
> There are general rules how branching is predicted (like jumps forward
> are
> not suspected to happen, while jumps backwards are) - have a look at the
> CPU
> manuals!
> 
> best greetings,
> Thomas
> 
> ----- Original Message ----- 
> From: "Luigi Castelli" <superbigio at yahoo.com>
> To: <pd-dev at iem.at>
> Sent: Monday, October 13, 2003 4:40 AM
> Subject: [PD-dev] Non-branching clipping algo...
> 
> 
> > Hi,
> >
> > I am trying to find a way to clip incoming values to a
> > range from 0.0 to 1.0 without using branching.
> >
> > I found a bunch of algorithms on musicdsp.org to
> > achieve clipping without branching but none were
> > satisfactory.
> >
> > What I am really aiming for is to find a way to modify
> > the tabfudge technique used in d_osc.c to achieve
> > clipping instead of wrapping. Is that possible ?
> >
> > Thanks for any help or suggestion.
> >
> > - Luigi
> >
> > =====
> > Luigi Castelli
> >
> > superbigio at yahoo.com
> >
> > __________________________________
> > Do you Yahoo!?
> > The New Yahoo! Shopping - with improved product search
> > http://shopping.yahoo.com
> >
> > _______________________________________________
> > PD-dev mailing list
> > PD-dev at iem.at
> > http://iem.at/cgi-bin/mailman/listinfo/pd-dev
> >
> >
> 
> 
> _______________________________________________
> PD-dev mailing list
> PD-dev at iem.at
> http://iem.at/cgi-bin/mailman/listinfo/pd-dev
>