OSDev.org

HEllo,
I am wondering if I can find anybody who worked with the barrelfish on man-ycore chip like intel scc, I understand the theoretical structure of barrelfish and how does it work, but what if I want to dynamically control the memory or even number of cpus in a zone if I want to use the zoning concept when running these multi-kernel OSes on such many-core chips .. any idea even from where to start thinking as I have lots of questions let me think I need to re-design the whole kernel, coz it is still not clear how to do such a thing with the barrelfish the only reliable multi-kernel OS I know !

Thanks,
Hegazi

ahegazy wrote:HEllo,
I am wondering if I can find anybody who worked with the barrelfish on man-ycore chip like intel scc, I understand the theoretical structure of barrelfish and how does it work, but what if I want to dynamically control the memory or even number of cpus in a zone if I want to use the zoning concept when running these multi-kernel OSes on such many-core chips .. any idea even from where to start thinking as I have lots of questions let me think I need to re-design the whole kernel, coz it is still not clear how to do such a thing with the barrelfish the only reliable multi-kernel OS I know !

Thanks,
Hegazi

1a) This is an English speaking forum, "coz" I had to google "man-ycore" to figure out you meant "many core."
1b) In the English language, the period (.) is used to separate sentences for readability purposes, and the question mark (?) is used in place of the period to denote a question.
1c) In English, run on sentences should be split into multiple separate sentences.
2) I've never heard of a "barrelfish" chip. If you give a chip number, we may be able to better help you.

I don't know the answer to your question, and I have no experience of Barrelfish, but thanks for the heads-up. I have great respect for other work from ETH, so I must look into Barrelfish.

Yo:

iansjack wrote:I don't know the answer to your question, and I have no experience of Barrelfish, but thanks for the heads-up. I have great respect for other work from ETH, so I must look into Barrelfish.

Don't get your hopes too high. Last I checked, it was nothing impressive, though that was a while back.

--Peace out,
gravaera

It's obviously not a production OS, but it looks to have some interesting ideas (at a first glance). I'm always open to new ideas.

Hi,

gravaera wrote:

iansjack wrote:I don't know the answer to your question, and I have no experience of Barrelfish, but thanks for the heads-up. I have great respect for other work from ETH, so I must look into Barrelfish.

Don't get your hopes too high. Last I checked, it was nothing impressive, though that was a while back.

When I looked at into it last (back when it was first announced publicly) it made little sense to me - I saw massive performance problems maintaining any type of global state. I had another look yesterday and it seems they solved/avoided the performance problems by using shared memory like a normal kernel.

With this in mind I think the only real difference between barrelfish and a more traditional kernel is merely word games. It doesn't matter if you say "per CPU kernel with a lot of shared stuff" or "shared kernel with a lot of per-CPU stuff", it adds up to the same end result (a mixture of shared and per-CPU).


Cheers,

Brendan