time unit in bench-skeleton

[Carlos O'Donell]

On 06/27/2017 07:59 PM, Victor Rodriguez wrote:
> I am trying to port the bench test to a most stable and usable
> platform like https://openbenchmarking.org/ in order to give others
> the capability to measure the performance of new glibc releases

That sounds very interesting. Please keep us updated on how that goes.

It would be interesting to see how such a tool handles the data over time.

> I am trying to understand the output of bench.out:
> 
>   "ffsll": {
>    "": {
>     "duration": 2.3834e+10,
>     "iterations": 5.5869e+09,
>     "max": 54.589,
>     "min": 4.246,
>     "mean": 4.26605
>    }
> 
> duration according to benchtests/bench-timing.h
> 
>  42 /* Measure the resolution of the clock so we can scale the number of
>  43    benchmark iterations by this value.  */
>  44 # define TIMING_INIT(res) \
>  45 ({                                        \
>  46   struct timespec start;                              \
>  47   clock_getres (CLOCK_PROCESS_CPUTIME_ID, &start);                \
>  48   (res) = start.tv_nsec;                          \
>  49 })
> 
> is in nanoseconds right?

Correct. This is the resolution in nanoseconds of the clock source.

-- 
Cheers,
Carlos.