-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hello,

On Dec 7, 2007, at 12:39, Joerg Sonnenberger wrote:

> On Fri, Dec 07, 2007 at 01:43:55PM +0100, Joerg Sonnenberger wrote:
>> No. There's some code that is supposed to balance interrupts over  
>> cpus,
>> but I think it is utterly broken.
>
> Just to clarify -- the interrupts are scheduled to arrive at a  
> specific
> CPU for the APIC case. The code is supposed to do a round-robin
> selection, but I don't think that code is working. You can enforce  
> that
> interrupts arrive at a specific CPU by manually changing the right  
> field
> when the interrupt is currently not routed. There is no API for that
> yet.

It would be nice to have such an API - OpenPIC for instance allows  
quite a lot of control over interrupt distribution, maybe we want to  
use some of that at some point. Right now there's a kernel option to  
select wether interrupts go to cpu0 or all CPUs but that might as  
well be a sysctl.

have fun
Michael
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.7 (Darwin)

iQEVAwUBR1mKrMpnzkX8Yg2nAQIEpQf/Vxe6I9w2UBj3hAz+ZfFyG88NvXeOHR7K
jUxD3kjf+JiIZ7Imi1L8Q/Y2wtAk4WD3qtoJdWQJe709M71QqX/yZpoz2RiWx2aY
PLrKLotgep6Z6i9uJ9izg2NYrEyr6G5FT7stwdpAvQhEw0gVk4e5h3UC3NIUGsQk
cSqaQ2yX5QFgL46ShOIfrp44ODeHcat+kDHT0Y42VtzgpkWnbkpgTnSzXM4pAqr0
wJGTrgwckJPkaDLpE2SoX0AsTfwm0iJ8TY+wTIW45lFjM4KhR44wrJu43AVtrpLp
8zyj6le4LWrLHtvggtB3FUrnd0uuAbOAOCsOODtBmbkKonkheioSog==
=EsBU
-----END PGP SIGNATURE-----