Re: [linux-audio-dev] New release of my plugins

New Message Reply About this list Date view Thread view Subject view Author view Other groups

Subject: Re: [linux-audio-dev] New release of my plugins
From: David Olofson (david_AT_gardena.net)
Date: Sun Dec 03 2000 - 19:44:21 EET


On Sunday 03 December 2000 15:24, Benno Senoner wrote:
> I'm not an expert in pitchshifters, but it seems that there are a
> few one which are really good and work in realtime.
> For example the Roland VP9000 (a hw box) sounds really fantastic
> judging from the demo sounds I've heard on the CD included with the
> german Keyboards magazine. (I don't think that the VP9000 has THAT
> powerful DSPs on board ... )

It doesn't need a very powerfull DSP to beat the sh*t out of a
current P-III class workstation CPU. First, it's easier to optimize
code for a specific chip, and second, the only way workstation CPUs
can achieve DSP like performance is by fully utilizing the SIMD
extensions.

Just look at the cycle counts; while DSP have been doing (at least)
single cycle MUL + ADD and similar operations for a good while, P-III
class CPUs still take several cycles for similar operations, without
the ability to pipeline them. This seems to be very hard to fix one
workstation style cores.

However, the SIMD extensions in general *can* pieline operations, in
addition to running multiple operations in parallel (ex: MMX integer
MUL; 4 x 16 bit MUL, 3 cycles total, can start a now one every
cycle), and most of them have composite operations of the kind DSP
chips have, so there's still hope. Not with pure C code yet for a
while, though, as current compilers can't think parallel.

> As Paul said, commercial companies consider audio algorithms as
> their "big secret", but let's see how long this lasts until
> researchers at some university (like the SCRIME guys) produce a
> comparable opensource version.

In the long run, I think Free/Open algorithms will beat the
proprietary ones, simply because the former ones will be analyzed and
improved by lots of hackers, while the latter ones depend on the
competence and creativity of one or a few persons. I also find it
more likely to be a tendency of Free/Open efforts being driven by a
desire to achieve perfection, while proprietary solutions are
affected by other factors.

//David

.- M u C o S -------------------------. .- David Olofson --------.
| A Free/Open Source | | Audio Hacker |
| Plugin and Integration Standard | | Linux Advocate |
| for | | Open Source Advocate |
| Professional and Consumer | | Singer |
| Multimedia | | Songwriter |
`-----> http://www.linuxdj.com/mucos -' `---> david_AT_linuxdj.com -'


New Message Reply About this list Date view Thread view Subject view Author view Other groups

This archive was generated by hypermail 2b28 : Sun Dec 03 2000 - 20:14:13 EET