[Click] User-level Click and Multithreaded support

[Eddie Kohler]

Hi Tom,

Tom Gibson wrote:
> Hi all,
> 
> Its been a little while since I worked with Click and I just recently
> started again.  With my previous usage I limited myself to Kernel mode Click
> mainly because it looked to be the only way to utilize my multi-core (8
> total) servers processing capabilities.  The other day I ran ./configure
> --help and noticed an option to enable experimental user-mode multithreaded
> support.
> 
> Is this coming along well and is it expected to be fully mature / stable
> later?

Hard to answer either question, but I'd say that I hope that it will continue 
maturing, and I don't know of any terrible outstanding bugs at the moment.

> How would its performance compare to Kernel mode Click?  I have only a
> generalized concept that it will be slower in because of the extra overhead
> changing contexts to/from Kernel mode, but I don't know how often this
> happens and what other factors are involved.

Can't give you numbers.  Depends on number of CPUs, obviously, and lots of 
other stuff.  But yes, kernel/user crossings will make it slower for 
forwarding, significantly so, especially compared to polling drivers.

E


> -Tom
> _______________________________________________
> click mailing list
> click at amsterdam.lcs.mit.edu
> https://amsterdam.lcs.mit.edu/mailman/listinfo/click