On 07/12/2018 22:35, Steven Sistare wrote:
[...]
>>> + if (!sds->cfs_overload_cpus) {
>>> + mask = sparsemask_alloc_node(nr_cpu_ids, 3, flags, nid);
>> ^^^^^^^^^^ ^^^
>> (1) (2)
>>
>> (1): Is this necessary? Wouldn't cpumask_weight(span) suffice?
>
> weight does not work because the ids are not consecutive. As a future
> optimization I have a patch that computes a local CPU id within the LLC
> and uses that as the bitmap index, and that uses weight. For this first
> series I am keeping things simple.
>
Oh, duh, non-contiguous spans strike again. Fair enough!
[...]
