[Bug tapsets/11209] New: Provide an alternative time tapset

[Bug tapsets/11209] New: Provide an alternative time tapset

[jistone at redhat dot com]

For most scripts, I think the Epoch time provided by gettimeofday_s() is not
that relevant.  Most of the time, all that's needed are relative times from one
event to another, and any time source should be fine.

I suggest creating a new walltime tapset that uses whatever time source is most
convenient/efficient.  On newer kernels, an implementation backed by the
kernel's cpu_clock() may be a good choice.  Where no better option exists, we
can always fall back to gettimeofday automatically.

-- 
           Summary: Provide an alternative time tapset
           Product: systemtap
           Version: unspecified
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: tapsets
        AssignedTo: systemtap at sources dot redhat dot com
        ReportedBy: jistone at redhat dot com


http://sourceware.org/bugzilla/show_bug.cgi?id=11209

------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.

[Bug tapsets/11209] Provide an alternative time tapset

[fche at redhat dot com]

------- Additional Comments From fche at redhat dot com  2010-01-26 17:12 -------
Is cpu_clock() synchronized across cpus?

-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=11209

------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.

[Bug tapsets/11209] Provide an alternative time tapset

[jistone at redhat dot com]

------- Additional Comments From jistone at redhat dot com  2010-01-26 22:55 -------
(In reply to comment #1)
> Is cpu_clock() synchronized across cpus?

At the top of sched_clock.c, it says:

 * The clock: sched_clock_cpu() is monotonic per cpu, and should be somewhat
 * consistent between cpus (never more than 2 jiffies difference).

We should probably experiment to estimate the practical difference of being
"somewhat consistent".

(And FWIW, I'm not even sure how close we can assert our current gtod is.)

-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=11209

------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.