Cc: LKML <linux-kernel@...>, Ingo Molnar <mingo@...>, Linus Torvalds <torvalds@...>, Andrew Morton <akpm@...>, Peter Zijlstra <a.p.zijlstra@...>, Christoph Hellwig <hch@...>, Mathieu Desnoyers <mathieu.desnoyers@...>, Gregory Haskins <ghaskins@...>, Arnaldo Carvalho de Melo <acme@...>, Thomas Gleixner <tglx@...>, Tim Bird <tim.bird@...>, Sam Ravnborg <sam@...>, Jan Kiszka <jan.kiszka@...>, John Stultz <johnstul@...>, Arjan van de Ven <arjan@...>, Steven Rostedt <srostedt@...>
Well, I just found out about latencytop a couple of days ago when Arjan
announced it. I took a quick look and it seems to be more of a general
collection of information. This tracer pinpoints out specific maxes.
Although I may be wrong (I have to admit, I haven't had the time to really
look at what LatencyTop does, but I have been using this tracer in -rt for
a number of years).
As for using markers, I just piggied back on top of lockdep, since it has
the same hooks that I needed. If we were to replace them with markers,
then I would say lockdep should use the markers too.
One thing about markers, only one probe may be allowed at a time. Perhaps
we should fix this by using something like I did with the mcount hook.
-- Steve
--
