> On Feb 7, 2016, at 11:24 AM, Mahmood Naderan via dtrace-discuss > <[email protected]> wrote: > > Hello, > The first question relates to the Dtrace, but the second question relates to > the Solaris. So, you may find the second one off topic... I do appreciate if > you answer both. > > > 1) How can I measure the duration of a Dtrace script? I mean, a script starts > for some syscall duration measurements. As I press ^c, I want to know what > was the duration of system calls and whats was the duration of the script > itself. Then by a differentiation, I will be aware of non-syscalls (i.e. > normal user code).
I can’t speak for Solaris, but in illumos-derived distributions there are per-cpu kstats that measuring CPU-nanoseconds spent running dtrace, user, system, interrupt, and idle (cpu_nsec_dtrace). This is not subject to the measurement effects of typical profilers, but it also is not easily separable from other activity, especially for multi-threaded programs running on multiple CPUs. However, as a first order measurement, you can see whether dtrace is consuming a lot of time. Using this technique, we’ve been able to reduce the cost of running dtrace in some cases. — richard -- [email protected] +1-760-896-4422 ------------------------------------------- dtrace-discuss Archives: https://www.listbox.com/member/archive/184261/=now RSS Feed: https://www.listbox.com/member/archive/rss/184261/25769126-e243886f Modify Your Subscription: https://www.listbox.com/member/?member_id=25769126&id_secret=25769126-8d47a7b2 Powered by Listbox: http://www.listbox.com
