ensure context switch always happen

[dips48]

Linux kernel scheduler has changed a lot, which imported
wake_wide to determine whether to use wake_affine in
try_to_wake_up other tasks, since some cpus don't share LLC ,
Under this situation, the behaviour of this case differs a lot，
before using wake_wide, two threads will work on same cpu,
after this, two threads will work on different cpus,
which causes a really big score gap.
For example:
(no using wake_wide) kernel 3.10 score 977
(using wake_wide) kernel 4.9 score 227
which confused me, so i suggest to bind two threads
on different cpus while theads booting up.In this way, we can ensue there always
has context switches.

[gstrauss]

(General comment: I question the applicability of some of these overly simplistic benchmarks)

Should we care about hyperthread cores versus primary cores? What about cores on different CPU dies? What about P (performance) cores versus E (efficiency) cores?

Alternatively -- and also changing the test in such a way as to invalidate comparison with prior scores -- might consider sched_yield() on each loop to force context switch.