On Tue, Apr 17, 2007 at 04:03:41PM +1000, Peter Williams wrote:
quoted text
> There's a lot of ugly code in the load balancer that is only there to 
> overcome the side effects of SMT and dual core.  A lot of it was put 
> there by Intel employees trying to make load balancing more friendly to 
> their systems.  What I'm suggesting is that an N CPUs per runqueue is a 
> better way of achieving that end.  I may (of course) be wrong but I 
> think that the idea deserves more consideration than you're willing to 
> give it.

This may be a good one to ask Ingo about, as he did significant
performance work on per-core runqueues for SMT. While I did write
per-node runqueue code for NUMA at some point in the past, I did no
tuning or other performance work on it, only functionality.

I've actually dealt with kernels using elder versions of Ingo's code
for per-core runqueues on SMT, but was never called upon to examine
that particular code for either performance or stability, so I'm
largely ignorant of what the perceived outcome of it was.


-- wli
-
unsubscribe notice
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/