Re: [PATCH] gpio: add trace events for setting direction and value

["Uwe Kleine-König" <u.kleine-koenig@pengutronix.de>]

Hi Grant,

On Wed, Mar 02, 2011 at 03:03:05PM -0700, Grant Likely wrote:
> On Fri, Feb 18, 2011 at 10:48:06AM -0500, Steven Rostedt wrote:
> > On Fri, 2011-02-18 at 10:58 +0100, Uwe Kleine-König wrote:
> > > Hi Steven, hi Grant,
> > > 
> > > On Wed, Feb 16, 2011 at 12:12:51PM -0500, Steven Rostedt wrote:
> > > > Note: to save the memory footprint of these tracepoints, you can use
> > > > DEFINE_EVENT_PRINT(). You can see the usage for this in the
> > > > include/trace/events/kmem.h.
> > > > But to do this, you will need to have a single TP_STRUCT__entry() for
> > > > both. Not sure if this is what you want.
> > > > 
> > > > 	TP_STRUCT__entry(
> > > > 		__field(unsigned, gpiq)
> > > > 		__field(int, get_in)
> > > > 		__field(int, value_err)
> > > > 
> > > > ??
> > > > 
> > > > Just a suggestion, but may not be worth it.
> > > Yeah, I saw that, still I think it's sane to keep them seperated.
> > > Or how much would we save?  Can you estimate that?
> > 
> > You can do it :)   Especially since it can vary by archs.
> > 
> > Just compile the kernel once this way, and then try it with
> > DEFINE_EVENT_PRINT(), compile the kernel and run size on the two.
> > 
> > Then you can see the difference it makes. It may end up not being worth
> > the difference. But as embedded uses gpio the most, I'll leave that up
> > to you.
> 
> Uwe, any update on this?  Are you going to spin a new patch, or should
> I take this one?
I think it's OK to take as is.

Best regards
Uwe

-- 
Pengutronix e.K.                           | Uwe Kleine-König            |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |