[mythtv-users] Do you build mythtv from source, and why?

Raymond Wagner raymond at wagnerrp.com
Tue Mar 29 16:42:56 UTC 2011


On 3/29/2011 10:27, Frans Grotepass wrote:
> On Tue, 29 Mar 2011 10:47:58 +1100
> Jean-Yves Avenard<jyavenard at gmail.com>  wrote:
>> As for compilation flags; in these days and age; I doubt you would
>> notice the difference performance wise following *tweaks* on the
>> compiler
> Is that so? The Atom that is often used in MythTV, a light weight
> processor with a rich instruction set with many extensions for
> multimedia processing should benefit by it. It is specifically in this
> day and age with more extensive instruction sets that sticking with
> just the i386 instruction set wastes the functionality.

The Atom is a piece of garbage.  When Intel designed the Atom, they took 
a modern processor design, and then ripped out the last 15 years of 
research into branch prediction and instruction dispatch, leaving 
something that is every bit as cycle efficient as the old Netburst 
architecture, but half the clock rate.  Those 'rich multimedia 
extensions' you speak of are just SIMD units, 
single-instruction-multiple-data, vector instructions that have nothing 
to do with multimedia, but can merely be used with the type of 
processing often used for graphics.  These same vector instructions are 
similarly available in every other modern processor design, meaning they 
are nothing special to the Atom.

The only saving grace for the Atom is that nVidia released VDPAU for 
Linux, and then paired one of their mobile chipsets to an Atom.  No one 
uses an Atom for MythTV, they use the nVidia 9400M.  Without that 
graphics processor to offload anything intensive to, the Atom would be 
incapable of handling anything more than standard definition content, 
even with any and all compile optimizations you could think of.


More information about the mythtv-users mailing list