[mythtv-users] Re: HDTV performance in 0.18

[Axel Thimm]

On Mon, Apr 18, 2005 at 11:27:02PM -0400, Preet Khalsa wrote:
> On Tue, 2005-19-04 at 02:55 +0200, Axel Thimm wrote:
> > On Mon, Apr 18, 2005 at 03:38:53PM -0700, Big Wave Dave wrote:
> > > On 4/18/05, Brandon Beattie <brandon+myth at linuxis.us> wrote:
> > > > MythTV is often compiled with no support for MMX or anything
> > > > more than pentium pro optimization.  If you didn't enable
> > > > processor specific optimization and mmx, and opengl when doing
> > > > a ./configure then this could have the bad effect.  I haven't
> > > > noticed much if any CPU change between .17 and .18 (Granted I
> > > > run out of CVS) so I'd check that first.
> > > > 
> > > 
> > > Does anyone know how the ATRpms are compiled, in this respect?
> > 
> > Depends on the distribution, for FC3 they are tuned for pentium4.
> 
> Does this mean I should compile Myth for the x86_64 (AMD64) instead of
> just using the binaries?  (yes I did get the x86_664 version)

Oh no, sorry. The x86_64 packages are of course not optimized for
pentium4, gcc would bail out if told so. Here are the typical
optimization flags for x86 archs and compatibles on FC3 packages:

%__global_cflags     -O2 -g -pipe -Wp,-D_FORTIFY_SOURCE=2 -fexceptions

optflags: i386 %{__global_cflags} -m32 -march=i386 -mtune=pentium4 -fasynchronous-unwind-tables
optflags: i486 %{__global_cflags} -m32 -march=i486 -fasynchronous-unwind-tables
optflags: i586 %{__global_cflags} -m32 -march=i586 -fasynchronous-unwind-tables
optflags: i686 %{__global_cflags} -m32 -march=i686 -mtune=pentium4 -fasynchronous-unwind-tables
optflags: athlon %{__global_cflags} -m32 -march=athlon -fasynchronous-unwind-tables
optflags: ia64 %{__global_cflags}
optflags: x86_64 %{__global_cflags} -m64 -mtune=nocona

You can rebuild the packages with different optflags, if you like, to
see whether there is any difference. Note that on x86_64 MMX in mythtv
does not compile, so you will see a considerable performance drop wrt
i386. Perhaps trying the i386 packages would not be a bad idea.
-- 
Axel.Thimm at ATrpms.net
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
Url : http://mythtv.org/pipermail/mythtv-users/attachments/20050419/e38e93f0/attachment.pgp