Re: [RFC PATCH 1/3] Unified trace buffer

[Ingo Molnar]

* Steven Rostedt <rostedt@xxxxxxxxxxx> wrote:

> 
> On Sat, 27 Sep 2008, Ingo Molnar wrote:
> > 
> > If all you do is to trace high-freq events on all CPUs and you are _not_ 
> > interested in the precise interactions, the overhead of global 
> > synchronization can hurt a lot.
> > 
> > In any case, SMP coherency of trace events is an independent property of 
> > the tracer, and preferably something that can be turned on/off.
> 
> Just a note.  The current ring buffering system that I'm proposing 
> keeps its own time stamp counter (currently sched_clock) that will 
> most likely be updated later. I'm trying to keep this ring buffer 
> system as dumb as possible. It does not even implement the merge sort. 
> That's up to the tracer to handle. There's nothing stopping the trace 
> from adding some atomic counter to each event to help it sort.

correct. The price is all the notifier/callback overhead and the loss of 
type checking of the record contents. But that's an unavoidable price of 
abstraction, at least in C.

> So yes, the tracer can implement anything it wants on top of the ring 
> buffer ;-)

yes, very nice! :)

	Ingo
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/