[mythtv] [mythtv-commits] mythtv commit: r10912 by danielk

[Daniel Kristjansson]

On Thu, 2006-08-24 at 19:09 +0200, Lukas Kasprowicz wrote:
> Am Donnerstag, 24. August 2006 15:13 schrieb Daniel Kristjansson:
> > Hmm, what about adding ".*" after "CPU"? I'd like to capture
> > all the CPU's in that family.
> Adding ".*" after CPU does not fetch it, but adding ".*" before CPU 
> ("Genuine.*") fetches it.
I'll replace it with "Genuine.*Intel.*CPU"

> > Sure, I couldn't tell from Intel's site if this was a pentium4
> > or pentium3 w/pentium4 features arch, if it is a pentium3-ish
> > arch we should use the pentium-m flag.
> On Intels Website it is described as .*Centrino Duo. As i am informed, it is a 
> pentium-m (which belongs to .*Centrino.*) as a dual core.
> On list i noticed people using gcc 3.3. pentium-m is availible since gcc 3.4. 
> i do not know how long backwards compatible you want to be.
Hmm, well I'll leave it as pentium4 then, we want this to work
with gcc 2.9x and later. Centrino is a chipset name, not a CPU
name, you can use these dual core CPU's with other chipsets.
pentium-m and pentium4 code will run on either architecture as
they have the same instruction sets. In practice, gcc doesn't
generate significantly different code for the two architectures.
The Pentium 4 type CPUs are designed to run very fast, while
Pentium M CPUs are designed to run on very cool; the Pentium M
does less optimization of the code in the CPU core and expects
you to unroll loops, arrange instructions to avoid stalls, etc.
gcc does this even on CPUs which don't need it, like the P4.

-- Daniel