[9fans] cores a go go

[9fans] cores a go go

[erik quanstrom]

i rerolled the atom cd today with some noticable changes
for mp systems.

1.  apic handling has changed.  the mp code can now handle
apic ids > 7.  this can affect machines with fewer than 8 cores,
especially intel 2xxx-series and xeon 56xx-series processors.

2.  the apic code has a hack to find intel ht processors, even
when there is no mp table entry.  (as per usual.)  you might
see twice as many cores.

3.  msi interrupts are now the default.  see apic(3).
http://www.quanstro.net/magic/man2html/3/apic

this has been tested on a variety of phenom ii, opteron c32, intel 5[56]xx.
intel core i7, intel atom, and pentium iv systems from 2-24 cores.
but i'd be interested in hearing of any success/failure.

- erik

[9fans] cores a go go

[David Leimbach]

Nice!

Sent from my iPhone

On Mar 30, 2011, at 9:39 PM, erik quanstrom <quanstro at quanstro.net> wrote:

> i rerolled the atom cd today with some noticable changes
> for mp systems.
> 
> 1.  apic handling has changed.  the mp code can now handle
> apic ids > 7.  this can affect machines with fewer than 8 cores,
> especially intel 2xxx-series and xeon 56xx-series processors.
> 
> 2.  the apic code has a hack to find intel ht processors, even
> when there is no mp table entry.  (as per usual.)  you might
> see twice as many cores.
> 
> 3.  msi interrupts are now the default.  see apic(3).
> http://www.quanstro.net/magic/man2html/3/apic
> 
> this has been tested on a variety of phenom ii, opteron c32, intel 5[56]xx.
> intel core i7, intel atom, and pentium iv systems from 2-24 cores.
> but i'd be interested in hearing of any success/failure.
> 
> - erik
>