[asterisk-users] The High Performance Echo Canceller (HPEC)

[shadowym]

I generally agree with this.  I guess I just did not explain it very well. 

-----Original Message-----
From: Andrew Kohlsmith [mailto:akohlsmith-asterisk at benshaw.com] 
Sent: Wednesday, February 14, 2007 2:12 PM
To: asterisk-users at lists.digium.com
Subject: Re: [asterisk-users] The High Performance Echo Canceller (HPEC)

On Wednesday 14 February 2007 4:12 pm, shadowym wrote:
> The algorithms may be similar but EC is an infinitely variable
> non-linear(analog) process.  A CPU cannot do that.  You can fake it by 
> performing cpu intensive rapid calculations one after another but it 
> is fundamentally not an analog processor.  HWEC is designed to deal 
> with the analog process on an instant by instant basis performing 
> parallel computations.  A CPU cannot do that at ANY clock speed.

I think you are very sorely mistaken.  I've done DSP work on general-purpose
CPUs for many years.  All current processors have SIMD, which, until the
i586 (for Intel), was more or less only in DSPs.  Steven Critchfield has
been doing DSP work (spandsp) for much longer than I have, and is much
better at it than I will ever be.  :-)

Anything a DSP can do, a general-purpose CPU can do, but very likely slower.

There is no magic.  There is nothing particularly "special" about ASICs or
DSPs that general-purpose CPUs can't do; it's all a matter of how quickly it
can do it and how much you're willing to consume in system resources.

-A.