gprofng-gui-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: When is TSTAMP==0?


From: Ruud van der Pas
Subject: Re: When is TSTAMP==0?
Date: Thu, 8 Jun 2023 14:48:58 +0000

Hi Peter,

Thanks. Perhaps some of the info in log.xml is useful? This file is in
the experiment directory and contains lines like this:

<event kind="run" tstamp="2232293.749759000" time="1681298196" tm_zone="0"/>
<event kind="sample" tstamp="0.000038840" id="0" 
label="collector_open_experiment"/>
<event kind="sample" tstamp="1.000922960" id="1" label="periodic"/>
<event kind="sample" tstamp="1.037127040" id="2" 
label="collector_close_experiment"/>
<event kind="run" tstamp="2232293.749759000" time="1681298196" tm_zone="0"/>
<event kind="exit" tstamp="1.036830040"/>

Kind regards, Ruud

> On 3 Jun 2023, at 07:26, Peter Kessler OS via Gprofng-gui-devel 
> <gprofng-gui-devel@gnu.org> wrote:
> 
> I understand that TSTAMP==0 is the beginning of the experiment.  But how do I 
> relate that to time stamps I might get from my application.  E.g., if my 
> application logging says that an operation took anomalously long at "Wed May 
> 31 12:27:02.123 2023" (down to milliseconds, e.g., in Unix Epoch time), how 
> do I relate that to a gprofng TSTAMP value so I can filter that section of my 
> experiment to look at in more detail? 
> 
> If the experiment overview gave a human-readable timestamp down to 
> milliseconds for TSTAMP==0 I could make the conversion between wall-time and 
> TSTAMP time.
> 
>                                                ... peter
> 
> From: Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
> Date: Thursday, June 1, 2023 at 16:45
> To: Peter Kessler OS <peter.kessler@os.amperecomputing.com>, 
> gprofng-gui-devel@gnu.org <gprofng-gui-devel@gnu.org>
> Subject: Re: When is TSTAMP==0?
> 
> 
> 
> On 6/1/23 14:19, Peter Kessler OS via Gprofng-gui-devel wrote:
> When does gprofng start counting time for TSTAMP filters?
> 
> 0 — the beginning of the experiment.
> 
> 
> 
> I have a program that reports the times at which certain program phases start 
> and stop.  Is there a way to coordinate those program times with the TSTAMP 
> filters available in gprofng? 
> 
> TSTAMP is in nanoseconds.
> For example, to see experiment between 1 and 2 seconds:
>   gprofng display text  -filter '(TSTAMP > 100000000) && (TSTAMP < 
> 200000000)' -func  test.1.er
> 
> 
> If you use gprofng-gui:
>  go to the 'Timeline' view.
>  Select time range
>  Right click
>  Select filter.
> 
> 
> Point mouse on the filter (lower left corner). You will see the current 
> filter. This filter you can use in `gprofng display text -filter`
> 
> 
> -Vladimir
> 
> 
> 
> 
> I see, for example, "Start Time    : Wed May 31 12:27:02 2023" in the 
> experiment overview, but that doesn't have nearly the resolution I think I 
> need to write accurate TSTAMP filters.  Maybe this is a feature request: 
> Could the overview print out the Unix epoch time (e.g., milliseconds since of 
> TSTAMP==0?  Then I could use Unix epoch time available in my application to 
> find values to put in gprofng filters.
> 
> Thanks for considering this request.
> 
>                                                ... peter


reply via email to

[Prev in Thread] Current Thread [Next in Thread]