[LITMUS^RT] trace overheads

[shijunjie92]

There are only two places where I can put the 'LOCK' and 'UNLOCK': 'SCHED_EVENTS' and 'IPI_EVENTS'. Neither one of them can trace the lock-related overheads for there are both 0 of the LOCK and UNLOCK in the counts.csv. That means I haven't got the trace data of the LOCK and UNLOCK.

So when I using the DPCP, how can I trace the lock and unlock overheads?
1. Add these events inside the 'ft-trace-overheads' (how can I decide add them to SCHED_EVENTS or IPI_EVENTS)
2. What the meaning of 'BE_EVENTS' in 'ft-extract-samples'

Best Wishes!




You Sincerely!
SHI Junjie

2016-09-19 

shijunjie92 



发件人：Björn Brandenburg <bbb at mpi-sws.org>
发送时间：2016-09-19 11:13
主题：Re: [LITMUS^RT] trace overheads
收件人："litmus-dev"<litmus-dev at lists.litmus-rt.org>
抄送：


> On 19 Sep 2016, at 12:05, shijunjie92 <shijunjie92 at gmail.com> wrote: 
>  
> Hello all: 
>   
> When I trying to trace the 'LOCK and UNLOCK' overheads, I add 'LOCK' and 'UNLOCK' into the 'IPI_EVENTS' of 'ft-trace-overheads’. 

Why add them to IPI_EVENTS? LOCK and UNLOCK are not IPI events, no? 

> But when I do the step 'ft-select-samples counts.csv combined-overheads_*.float32 2>&1 | tee -a overhead-processing.log', errors occurs '[EE] Cutoff for UNLOCK is zero.[EE] Cutoff for LOCK is zero.’.  

That means that you have some trace files that don’t contain any LOCK/UNLOCK samples; counts.csv lists 0 for these events. Please consult the source code of ft-select-samples.  

The shuffling is not needed during debugging, only once you start collecting valid data. You should debug your implementation and workflow before shuffling/subsampling traces. 

- Björn 



_______________________________________________ 
litmus-dev mailing list 
litmus-dev at lists.litmus-rt.org 
https://lists.litmus-rt.org/listinfo/litmus-dev 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.litmus-rt.org/pipermail/litmus-dev/attachments/20160919/08aa1716/attachment.html>