Logical CPU -> physical CPU conversion

John Baldwin jhb at FreeBSD.org
Fri Feb 18 13:33:06 PST 2005


On Thursday 17 February 2005 10:26 am, Petri Helenius wrote:
> John Baldwin wrote:
> > On Feb 17, 2005, at 3:27 AM, Joseph Koshy wrote:
> >
> >
> > For current Intel HTT CPUs, yes.  Note that this might get tricky with
> > the dual core CPUs coming from both AMD and Intel in the not too
> > distant future as we don't know yet how they will enumerate the cores.
> > It may be that the cores "look" like HTT cores when they are actually
> > full cores, so I would try to abstract this in a way that you can use
> > different implementations for different CPUs to determine when a
> > logical CPU is a HTT CPU vs. a full core.  Also, I imagine that even
> > with dual core Intel CPUs there is a chance that they might do HTT
> > inside each core, so that a single physical processor would have 2
> > cores each with 2 HTT so a total of 4 logical CPUs from one piece of
> > silicon.
>
> It would be logical to follow the documented convention;
> -quote-
> In future IA-32 processors supporting Hyper-Threading Technology that
> implement more than
> two logical processors per physical processor, the logical processor bit
> shown in Figure 7-5 will
> be expanded to a 2- or 3-bit field to allow each of the logical
> processors to be identified. The
> package ID and cluster ID fields will be shifted to the left
> accordingly. Also, the package ID may
> be expanded to more than 2 bits, requiring the cluster ID field to be
> shifted to the left.
> -quote-

That's what Intel says, yes.  We don't know how AMD is going to enumerate 
cores yet.

-- 
John Baldwin <jhb at FreeBSD.org>  <><  http://www.FreeBSD.org/~jhb/
"Power Users Use the Power to Serve"  =  http://www.FreeBSD.org


More information about the freebsd-smp mailing list