Messages in this thread | | | Date | Tue, 3 Jan 2017 17:06:41 -0500 (EST) | From | Vince Weaver <> | Subject | Re: [PATCH] perf/x86: Reject non sampling events with precise_ip |
| |
On Tue, 3 Jan 2017, Jiri Olsa wrote: > On Tue, Jan 03, 2017 at 10:40:59AM +0100, Peter Zijlstra wrote: > > > > I think we should reject non sampling pebs events, as you say they make > > no sense what so ever. > > ook, attached >
you can use the PEBS events to gather aggregate stats though and they seem roughly right. Are they truly meaningless?
I had misremembered that they might not have the determinism problems of regular events (turns out that's wrong). They oddly seem to be worse in some limited tests I did.
So I guess nothing will be lost if they're disabled.
Vince
| |