[Click] Kernel Click multithreading

[Beyers Cronje]

Hi Tom,

You need to tell Click which schedulable elements should be bound to which
thread.

You have two options:

Static binding - See the StaticThreadSched element
http://read.cs.ucla.edu/click/elements/staticthreadsched

Dynamic balanced binding - See the BalancedThreadSched element
http://read.cs.ucla.edu/click/elements/balancedthreadsched

Additionally, anyone that is even thinking about multithreading with Click
should as a start read the  "Flexible control of parallelism in a
multiprocessor PC router" publication, link over at
http://read.cs.ucla.edu/click/publications

Lastly, I recall a recent post on the list where a recent commit on the tree
broke thread scheduling, but was subsequently fixed by Eddie. So best would
be to work from the most recent GIT release.

Beyers


On Fri, Apr 10, 2009 at 1:22 AM, Tom Gibson <tom.gibson at ipvidnet.com> wrote:

> What would a simple click config look like to test utilizing all my
> available CPU cores with simulated data?
>
> I tried the following, but I'm not seeing more than 1 core utilized:
>
> InfiniteSource(\<0800>) -> rate1 :: Counter() -> Discard;
> InfiniteSource(\<0800>) -> rate2 :: Counter() -> Discard;
> InfiniteSource(\<0800>) -> rate3 :: Counter() -> Discard;
> InfiniteSource(\<0800>) -> rate4 :: Counter() -> Discard;
> InfiniteSource(\<0800>) -> rate5 :: Counter() -> Discard;
> InfiniteSource(\<0800>) -> rate6 :: Counter() -> Discard;
>
>
> I read that a single scheduled element only ever executes on one thread.  I
> figured making 6 InfiniteSource elements would cause 6 cores to be
> utilized.
>
> I configured click with --enable-multithread=8  and installed with -t 6.
> What am I missing?
>
> -Tom
> _______________________________________________
> click mailing list
> click at amsterdam.lcs.mit.edu
> https://amsterdam.lcs.mit.edu/mailman/listinfo/click
>