Srikar Dronamraju <[email protected]> writes:
> PowerVM systems configured in shared processors mode have some unique
> challenges. Some device-tree properties will be missing on a shared
> processor. Hence some sched domains may not make sense for shared processor
> systems.
>
> Most shared processor systems are over-provisioned. Underlying PowerVM
> Hypervisor would schedule at a Big Core granularity. The most recent power
> processors support two almost independent cores. In a lightly loaded
> condition, it helps the overall system performance if we pack to lesser
> number of Big Cores.
>
Is this good to do if the systems are not over-provisioned? What will be
the performance impact in that case with and without the change?
-aneesh
* Aneesh Kumar K.V <[email protected]> [2023-11-15 11:24:59]:
> Srikar Dronamraju <[email protected]> writes:
>
> > PowerVM systems configured in shared processors mode have some unique
> > challenges. Some device-tree properties will be missing on a shared
> > processor. Hence some sched domains may not make sense for shared processor
> > systems.
> >
> > Most shared processor systems are over-provisioned. Underlying PowerVM
> > Hypervisor would schedule at a Big Core granularity. The most recent power
> > processors support two almost independent cores. In a lightly loaded
> > condition, it helps the overall system performance if we pack to lesser
> > number of Big Cores.
> >
>
> Is this good to do if the systems are not over-provisioned? What will be
> the performance impact in that case with and without the change?
>
We are consolidating 1 thread per thread group (aka SMT domain).
Since each thread-group is suppose to be independent including having a
private L1/L2/L3 cache, we expect minimal impact in non over provisioned
scenario.
In Over utilization scenario, the changes in this patchset will not even kick in.
--
Thanks and Regards
Srikar Dronamraju