Messages in this thread | | | Date | Fri, 25 Jul 2008 17:30:57 -0700 | From | Jeremy Fitzhardinge <> | Subject | Re: [PATCH 0/4] x86_64: Optimize percpu accesses |
| |
Mike Travis wrote: > Yes, you're right, I wrote that quickly without really reading it back. > My point is that now that x86_read_percpu() and x86_write_percpu() do > evaluate to a single instruction (by definition atomic), then it doesn't > need to be surrounded by the preempt_disable()/preempt_enable() calls. >
Yep, correct.
> It appears as if I'm implying that's the case for get/put_cpu_var(). >
Right.
J
| |