Subject: Re: [linux-audio-dev] MMX, SSE, SSE2, 3DNOW
From: Richard Guenther (rguenth_AT_tat.physik.uni-tuebingen.de)
Date: Mon Feb 18 2002 - 11:06:22 EET
On Sun, 17 Feb 2002, Billy Biggs wrote:
> On 18 Feb 2002, Bob Ham wrote:
>
> > Not really. How are plugins different from programs in this respect? I
> > don't think any OS says to programs "do you use SSE instructions? no?
> > I'll not run you then." Instead, programs (or, more usually, build
> > systems) say "does this machine support SSE instructions? no? I'll not
> > execute/build them then."
>
> It really should be the program that detects and not the build
> system. :) It would be very bad if a package maintainer didn't also build
> the SSE2 version of a function because their machine was a P3.
>
> There's lots of code floating around to detect cpu type in the
> code. See mm_accel() and friends, and ask omega_AT_gstreamer.net for his
> work on speciallib.
>
> Plugins shouldn't be any different than programs, they can just as
> easily determine CPU type.
In fact, LADSPA plugins even have a proper place for the check:
the ladspa_descriptor() function, where it can set up the descriptor
according to the host CPU type.
Btw. one point to raise is that a plugin in _no_ case should use
other than plain i386 instructions if no detection is available, and
in the latter case always includes a i386 version for fallback (all
of course assuming x86 CPU type). Otherwise a binary distributed
plugin may cause crashes on low-end machines.
Richard.
-- Richard Guenther <richard.guenther_AT_uni-tuebingen.de> WWW: http://www.tat.physik.uni-tuebingen.de/~rguenth/ The GLAME Project: http://www.glame.de/
This archive was generated by hypermail 2b28 : Mon Feb 18 2002 - 10:55:59 EET