Got it, thanks!
On Fri, Jun 20, 2014 at 3:20 PM, Sandeep Tripathy
wrote:
> Hi Meraj,
> As mentioned earlier, we are counting only the interrupts which caused
> the core to wake up from one of the c-states.
>This count can vary from the total number of interrupts to that core.
> When the
Hi Meraj,
As mentioned earlier, we are counting only the interrupts which
caused the core to wake up from one of the c-states.
This count can vary from the total number of interrupts to that core.
When the core is already running we do not care
about number of interrupts its getting.
T
Hello Sandeep,
Thanks for reply.
I meant to say I counted "irq_handler_entry: irq=40 name=eth0" and
"irq_handler_entry: irq=18 name=ata_piix" instances in trace file (not
just grep irq=40/irq=18 as I mentioned before, sorry for the
confusion) and that count does not match with the output by idles
Hi Meraj,
We only want to count the wakeup sources (eg: irq/ipi etc *). Hence
the total count for an irq /ipi can differ.
* Note there can be spurious wakeups in some systems. ie: core exiting idle
without an irq.
Thanks
Sandeep
On 20 June 2014 13:58, Mohammad Merajul Islam Molla
wrote:
Hello Amit/Daniel,
I think there is a bug in idlestat due to which it reports wrong irq
counts in output.
Log is 9.998478 secs long with 1518 events
clusterA@state hits total(us) avg(us) min(us) max(us)
C6-SNB 455 9921571.0021805.65 0.00176303.00
