Messages in this thread |  | | | Date | Fri, 3 Jul 2009 23:40:38 +0200 | | From | Ingo Molnar <> | | Subject | Re: [PATCH -tip] x86: atomic64: inline atomic64_read() |
* Linus Torvalds <torvalds@linux-foundation.org> wrote:
> Do we actually _have_ any performance-critical 64-bit counters > that have monotonicity guarantees? I have no idea. I'm just > throwing out the notion.
Due to the 'counter value flipping' we do on perfcounters for certain workloads (attr.inherit_stat=1 counters - inherited counters that neverheless provide precise per thread readouts - not just precise summaries), the monotonicity assumption is not generally true anymore.
But it would be nice if Eric could measure your suggested primitive, just that we see how many cycles we are talking about. I suspect the biggest difference would be a many-readers testcase - but that wont normally happen on perfcounters as the counters typically get read only from a single thread of executin.
Ingo
|  |