Re: timing information with perf

[David Ahern <dsahern@gmail.com>]

On 11/16/15 5:12 PM, Arnaldo Carvalho de Melo wrote:
> Em Mon, Nov 16, 2015 at 01:08:15PM -0500, Frank Ch. Eigler escreveu:
>> Olga.Kornievskaia wrote:
>
>>> [...] say there are two tracepoints ?foobar_enter? and
>>> ?foobar_exit?. Each tracepoint logs a timestamp. I?d like to be able
>>> to say that on average it takes that many time-units between ?enter?
>>> and ?exit? tracepoints. [...]
>
>> For completeness, a complete systemtap implementation of this could look like:
>
>> # stap -e '
>>     global s, t%
>>     probe kernel.trace("foobar_enter") {
>>         t[tid()]=gettimeofday_us()
>>     }
>>     probe kernel.trace("foobar_exit") {
>>         if (tid() in t) { s <<< gettimeofday_us() - t[tid()] }
>>     }
>>     probe timer.s(5),end {
>>         printf("cumulative average us: %d\n", @avg(s))
>>     }
>> '
>
> Is that done syncronously?

Interesting solution. How does it scale with CPUs and tasks?