Messages in this thread | | | Date | Fri, 14 Oct 2005 06:35:55 +0200 | From | Ingo Molnar <> | Subject | Re: sched_clock -> check_tsc_unstable -> tsc_read_c3_time ?!? |
| |
* Lee Revell <rlrevell@joe-job.com> wrote:
> On Thu, 2005-10-13 at 13:11 -0700, john stultz wrote: > > Yea, you're right about the inlining. Although I'm not sure why those > > functions should take microseconds to execute. That's very strange. > > Latency tracing overhead plus a slow (600MHz) machine?
yeah. The micro-timings of latency tracing can be misleading. Function calls are very fast on most CPUs (even on a 600MHz one), but with the latency tracer generating one trace entry per function call, there's considerable added overhead.
we could in theory calibrate the tracing overhead and subtract it from cycle readings [i've done this in a previous mcount() based tracer implementation, years ago], but that would make the latency trace timestamps less useful as a global time reference.
Ingo - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
| |