<div dir="ltr"><div>Dear All,<br></div>I do like to know the measurement time units of the following IDs<br><div><div><ul><li><p class=""><tt class="">SCHED_START</tt>, <tt class="">SCHED_END</tt>: Used to measure the time spent to make a scheduling decision. <span class="" id="line-61"></span></p>
</li><li><p class=""><tt class="">CXS_START</tt>, <tt class="">CXS_END</tt>: Used to record the time spent to make a context switch. <span class="" id="line-62"></span></p></li><li><p class=""><tt class="">SCHED2_START</tt>, <tt class="">SCHED2_END</tt>:
Used to measure the time spent to perform post-context-switch cleanup
and management activities. This is part of the scheduling overhead but
for technical reasons cannot be measured as part of the interval [<tt class="">SCHED_START</tt>, <tt class="">SCHED_END</tt>]. <span class="" id="line-63"></span></p></li><li><p class=""><tt class="">TICK_START</tt>, <tt class="">TICK_END</tt>: Used to measure the overhead incurred due to Linux's periodic system tick. As of LITMUS<sup>RT</sup> 2014.2, no LITMUS<sup>RT</sup>-specific activity takes place during a system tick. <span class="" id="line-64"></span></p>
</li><li><p class=""><tt class="">QUANTUM_BOUDARY_START</tt>,
QUANTUM_BOUNDARY_END`: Used to measure the overhead incurred in
quantum-driven plugins (e.g., PFAIR) at the start of each quantum.
Available since LITMUS<sup>RT</sup> 2014.2. <span class="" id="line-65"></span></p></li><li><p class=""><tt class="">PLUGIN_SCHED_START</tt>, <tt class="">PLUGIN_SCHED_END</tt>: Like [<tt class="">SCHED_START</tt>, <tt class="">SCHED_END</tt>], but only measures the time spent by the active scheduling plugin. There is no equivalent <tt class="">SCHED2</tt> counterpart because the scheduling plugins do not directly contribute to the <tt class="">SCHED2</tt> overhead. <span class="" id="line-66"></span></p>
</li><li><p class=""><tt class="">RELEASE_START</tt>, <tt class="">RELEASE_END</tt>: Used to measure the time spent to enqueue a newly-released job in a ready queue. <span class="" id="line-67"></span></p></li><li><p class="">
<tt class="">RELEASE_LATENCY</tt>: Used to measure the difference between when a timer <em>should</em> have fired, and when it actually <em>did</em>
fire. In contrast to all other time stamps, <b>this overhead is directly
measured in nanoseconds</b> (and not in processor cycles as the other
overheads). </p></li></ul><p><br></p><p>Except <tt class="">RELEASE_LATENCY, are all the others measured in milliseconds ?</tt></p><p><tt class="">Or how can I determine that .</tt></p><p><tt class=""><br></tt></p><p><tt class="">Thanks in advance <br>
</tt></p><p><tt class="">Sanjib<br></tt></p></div></div></div>