1 |
On Wed August 3 2005 08:15 am, Mike S wrote: |
2 |
> <snipped from another thread> |
3 |
> |
4 |
> On Mon August 1 2005 07:26 am, Mike S wrote: |
5 |
> >> > Is kernel preemption a bad thing? |
6 |
> >> > |
7 |
> >> > --Mike S |
8 |
> > > |
9 |
> > > No, just broken. |
10 |
> > |
11 |
> > Unless you enable SMP in conjunction with it (yes, even if you _only have |
12 |
> > one CPU_!). |
13 |
> |
14 |
> I opted to use an upstream kernel. Is this also broken here, or just in |
15 |
> the gentoo kernel? And I do have SMP enabled. Basically whatever |
16 |
> speed-ups I can get I am trying to use. |
17 |
> |
18 |
> --Mike S |
19 |
|
20 |
As I understand, it preemption is not exactly a speedup. It is supposed to |
21 |
improve the responsiveness of the kernel to user input. The kernel can |
22 |
switch contexts in more places inside the kernel in response to other inputs. |
23 |
This is great for desktops, but probably not for servers/number crunchers. |
24 |
|
25 |
I have no idea if it's fixed or not. Try disabling SMP and enabling |
26 |
preemption followed by an emerge of something large (openoffice, kde, you |
27 |
pick). If it succeeds, maybe it's fixed. If not, then it's still broken. |
28 |
|
29 |
Erik |
30 |
|
31 |
P.S. I'm no expert on the kernel (I'm a laser physicist.), so you might be |
32 |
better off Googling for an answer. |
33 |
-- |
34 |
******************************************* |
35 |
Dr. Erik Zeek |
36 |
Postdoctoral Research Associate |
37 |
Department of Physics and Astronomy |
38 |
The University of Georgia |
39 |
Athens, GA 30602-2451 |
40 |
|
41 |
Tel: 706-542-7293 |
42 |
Email: mailto:zeekec@×××××××××××××.com |
43 |
Html: http://www.physast.uga.edu/~zeekec |
44 |
******************************************* |
45 |
Against stupidity the very gods |
46 |
Themselves contend in vain. |
47 |
- Johann Christoph Friedrich von Schiller (1801) |
48 |
******************************************* |