On Mon, 2017-06-26 at 18:12 +0200, Peter Zijlstra wrote:

> Also note that your use of task_h_load() in the new numa thing
> suffers
> from exactly the problem effective_load() is trying to solve.

Thinking about this some more, I suspect that
using effective_load() in combination with
select_idle_sibling(), which will often place
the task on a different CPU than the one specified,
may not lead to entirely useful results...

-- 
All rights reversed