* Peter Zijlstra (peterz@infradead.org) wrote:
[...]
quoted text
> > What would suit us would be to have an arch callback that is called
> > after all the transforms for a particular jump label key have been made.
> > That way we could optimise the individual patches, and do a sync step at
> > the end, ie. when we want the effect of the patching to be globally
> > visible. 
> 
> I think such a sync-barrier is desired (possibly only on the enable
> path) so we can actually say the tracepoints are on.
> 
> Which would mean sending IPIs to all CPUs and waiting for them to
> acknowledge them. Which, while not quite as expensive as stop_machine,
> its not really cheap either.

Yep, although this can be batched when enabling many tracepoints en
masse. May I suggest that you guys benchmark the two approaches so we
can figure out at how many tracepoints we start hitting a latency wall ?
100, 1000 and 10000 tracepoints should give interesting measurement
points. If we are still below 2 seconds on common hardware when enabling
10000 tracepoints, then the binary search might be fine.

Please note that HPA recommended the use of a perfect hash. It would
make sense, although there seems to be a non-null probability that the
perfect hash cannot be generated. There are techniques that will retry
with a different seed, but the kernel build time then becomes slightly
harder to predict (for very, very rare occurences, so maybe we don't
care).

Thanks,

Mathieu

-- 
Mathieu Desnoyers
Operating System Efficiency R&D Consultant
EfficiOS Inc.
http://www.efficios.com
--
unsubscribe notice
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/