Re: [Xen-devel] [PATCH 0/4] xen: sched: optimize exclusive pinning and soft-affinity checking

[George Dunlap <george.dunlap@xxxxxxxxxx>]

On 09/15/2017 06:35 PM, Dario Faggioli wrote:
> Hello,
> 
> This series is a rework of a patch that was, originally, sent as part of a
> series (which went in already):
> 
>  https://lists.xen.org/archives/html/xen-devel/2017-07/msg02167.html
> 
> As it can be seen in the message above, George suggested doing things a little
> bit differently, and I agreed. however, that:
> - require a bit more of rework than expected, but here we are;
> - opened the possibility for even more optimization. :-)
> 
> Basically, the effect of the series is that:
> 1) when a vCPU is exclusively pinned to a pCPU, a lot of checks, while making
>    scheduling decisions in Credit1 and Credit2, are skipped (patch 2);
> 2) the check to see whether or not a vCPU has a soft-affinity that actually
>    matters for the scheduling of the vCPU itself, and should be considered is
>    optimized and made faster (patch 4).
> 
> So, the important bits are in patches 2 and 4, but both patches 1 and 3 are
> necessary to make the other twos possible.

I like it -- thanks, Dario!

 -George

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
https://lists.xen.org/xen-devel