On Wed, Mar 24, 2021 at 8:51 PM Steven Rostedt <rostedt@xxxxxxxxxxx> wrote:
>
> On Wed, 24 Mar 2021 15:04:10 +0200
> "Tzvetomir Stoyanov (VMware)" <tz.stoyanov@xxxxxxxxx> wrote:
>
> > @@ -1296,15 +1287,23 @@ static unsigned long long timestamp_correct(unsigned long long ts,
> > static unsigned long long timestamp_calc(unsigned long long ts,
> > struct tracecmd_input *handle)
> > {
> > - ts = timestamp_correct(ts, handle);
> > + /* Guest trace file, sync with host timestamps */
> > + if (handle->host.sync_enable)
> > + ts = timestamp_host_sync(ts, handle);
> >
> > - if (handle->ts2secs)
> > + if (handle->ts2secs) {
> > + /* user specified clock frequency */
> > ts *= handle->ts2secs;
> > - else if (handle->tsc_calc.mult) {
> > + } else if (handle->tsc_calc.mult) {
> > + /* auto calculated TSC clock frequency */
> > ts -= handle->tsc_calc.offset;
> > ts = mul_u64_u32_shr(ts, handle->tsc_calc.mult, handle->tsc_calc.shift);
> > }
> >
> > + /* User specified time offset with --ts-offset or --date options */
> > + if (handle->ts_offset)
> > + ts += handle->ts_offset;
> > +
> > return ts;
> > }
>
>
> Now that I'm playing with this, I find I want to see what the result is
> without the tsc_calc.offset, but can't do it. I'm thinking that the
> --ts-offset should modify that instead:
>
> bool ts_offset_set = false;
> [..]
>
> } else if (handel->tsc_calc.mult) {
> ts_offset_set = true;
> if (handle->ts_offset)
> ts += handle->ts_offset;
This will apply user specified offset instead of the autocalculated ?
I think it is better to have an option to disable the whole tsc2sec
adjustment when displaying the file, or just set a custom ts2sec:
trace-cmd report --ts2secs 1000000000 ... <- set ts2sec to 1,
this will be applied instead of tsc2nsec.
Or using a combination of "trace-cmd report --ts2secs ... --ts-offset
...", it replaces autocalculated tsc2sec with custom ones.
I think the existing "trace-cmd report" options are flexible enough.
> else
> ts -= handle->tsc_calc.offset;
> [..]
> }
>
> if (handle->ts_offset && !ts_offset_set)
> ts += handle->ts_offset;
>
> Or should we add another option to modify the ts_calc.offset?
>
> Thoughts?
>
> -- Steve
--
Tzvetomir (Ceco) Stoyanov
VMware Open Source Technology Center
|