Updated mythtv-fixes port

[Geoff Buckingham]

On Sat, Mar 24, 2007 at 12:18:33PM -0400, Joe Auty wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> 
> Would anybody care to clarify who stands to gain from this port, i.e.
> what sorts of hardware is unoptimized and covered here? I'm a little bit
> confused...
> 
> 
The reason I adapted the port Greg maintains to use the fixes branch of
MythtV-0.20 rather than the 0.20 release code was to get support for
protocol version 31, rather than version 30 provided in the release.
This allows me to use a FreeBSD MythTV front end with recent canned Linux
builds which use the fixes branch.

The upstream configure script, common to both ports, fails to detect the
CPU correctly under freebsd. THe best it can do is get i386 from uname.
(Not sure what happens in x86_64. Anybody know?)

As a result it uses -march=i386 during the ports build process. With my patch
to configure you should get -march=<your cpu family> and get -DMMX if your
cpu has mmx. On an Athlon 64 3200+ (i386 FreeBSD) i found this reduced CPU
utilisation during playback of a DVB-T mpeg2 from around 30% to around 10%.

I havn't looked at the impact on transcoding ad flagging etc as I currently use
a Linux backend.

Anyone who wants protocol 30 but feels the need for optimised binaries may
be able to copy patch-configureAA from this port to the files directory of
Greg's port.