Messages in this thread | | | Date | Fri, 5 Mar 2010 11:51:32 -0800 | Subject | Re: [PATCH 06/14] perf, x86: PEBS infrastructure | From | Stephane Eranian <> |
| |
On Fri, Mar 5, 2010 at 11:39 AM, Peter Zijlstra <peterz@infradead.org> wrote: > On Fri, 2010-03-05 at 11:11 -0800, Stephane Eranian wrote: >> By rewinding you get the IP of the instruction that caused the event. But the >> register state is STILL reflecting the situation at retirement of that >> instruction. > > I'm quite aware of that, but its near impossible to reverse interpret > instructions such as "mov rdx, $rdx", so until we get fault like PEBS > we'll just have to live with this. > I am sure you are, but that's not necessarily true of everybody. So this needs to be clearly spelled out in some documentation about precise + your MISC_EXACT bit.
| |