[PATCH 04/15] powerpc/time: Prepare to stop elapsing in dynticks-idle

[Shrikanth Hegde]

Hi Christophe.

>> No. With this patch there will not be any performance difference.
>> All it does is, instead of using mftb uses sched_clock at those places.
>>
> 
> For the record, I did some benchmark test with tools/testing/selftests/ 
> powerpc/benchmarks/null_syscall on powerpc 885 microcontroller:
> 
> Without your proposed patch:
> 
> root at vgoip:~# ./null_syscall
>     2729.98 ns     360.36 cycles
> 
> With your proposed patch below:
> 
> root at vgoip:~# ./null_syscall
>     3370.80 ns     444.95 cycles
> 
> So as expected it is a huge regression, almost 25% more time to run the 
> syscall.
> 
> Christophe
> 
> 
>>

Got it. My bad  in assuming it may not happen multiple before vtime_flush.

Btw, can you try latest vtime_reset patch? It shouldn't make any difference.
I tried perf bench syscall basic, and ./null_syscall.
I don't any difference for vtime_reset patch.