[lttng-dev] Removing timing overhead with tracepoint() <-- query dropped by mailing list

[Mathieu Desnoyers]

----- On Nov 17, 2016, at 12:21 PM, Mark E. Dawson, Jr. <medawsonjr at yahoo.com> wrote: 

> Mathieu,

> My query appears to be getting dropped by the mailing list, forcing me to ask
> you directly which I understand is both inefficient and a nuisance. We met when
> I spoke (along with Christoph Lameter) at the Tracing Summit a couple years
> back.

Hi Mark, 

Not sure why the mailing list would have dropped your email. Everything looks fine in the 
admin interface. Adding it in CC. (Michael, can you check if lttng-dev has issues with emails 
from yahoo domains ?) 

> Anyway, my question is as follows:

> If I want to remove a ton of compiled-in calls to an "rdtsc-based" C timing
> function and make them all conditionally-enabled based on whether or not I
> enable a tracing session, I can do so by adding the aforementioned "rdtsc" call
> as one of the arguments to tracepoint() along with our defined provider and
> event name, correct?

You can take a few approaches there: 

1) Use if (tracepoint_enabled()) { ..read tsc... do_tracepoint(...); } 

2) do this function call or static inline invocation as an argument of tracepoint(). 

3) do the function call/static inline invocation within the tracepoint probe provider 
(TRACEPOINT_EVENT() code). 

Let me know if I somehow misunderstood the intent of your question. 

Thanks, 

Mathieu 

-- 
Mathieu Desnoyers 
EfficiOS Inc. 
http://www.efficios.com 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.lttng.org/pipermail/lttng-dev/attachments/20161117/38db5ba5/attachment.html>