[Removed Suresh and Venki from discussion, they switched their companies
probably]

On 7 January 2013 20:34, Tejun Heo [off-list ref] wrote:

The latter part "not using idle cpu just for processing work" does
apply to homogeneous systems too but as I wrote earlier work items
don't spontaneously happen on an idle core.  Something, including
timer, needs to kick it.  So, by definition, a CPU already isn't idle
when a work item starts execution on it.  What am I missing here?

We are talking about a core being idle from schedulers perspective :)

quoted

I have another idea that we can try:

queue_work_on_any_cpu().

With this we would not break any existing code and can try to migrate
old users to
this new infrastructure (atleast the ones which are rearming works from their
work_handlers). What do you say?

Yeah, this could be a better solution, I think.  Plus, it's not like
finding the optimal cpu is free.

Thanks for the first part (When i shared this idea with Vincent and Amit, i
wasn't sure at all about the feedback i will get from you and others, but i
am very happy now :) ).

I couldn't understand the second part. We still need to search for a free cpu
for this new routine. And the implementation would almost be same as the
implementation of queue_work() in my initial patch

quoted

To take care of the cache locality issue, we can pass an argument to
this routine,
that can provide
- the mask of cpus to schedule this work on
  OR
- Sched Level (SD_LEVEL) of cpus to run it.

Let's start simple for now.  If we really need it, we can always add
more later.

:)
Agreed. But i liked the idea from steven, we can have two routines:
queue_work_on_any_cpu() and queue_work_on_cpus()