Re: [Xen-devel] does cpu_down makes brings core to deepest sleep state?

2015-10-26 Thread Hamed Rs
On Mon, Oct 26, 2015 at 1:14 PM, Jan Beulich wrote: > ing maintained by (and hence re I only turned off the cpu which has ID = 1 from 80th second to 100th second. As you can see in the result of "xenpm get-cpuidle-state" in time 100.7, the residency in C0 of cpu1 is equal to

Re: [Xen-devel] does cpu_down makes brings core to deepest sleep state?

2015-10-26 Thread Jan Beulich
>>> On 25.10.15 at 09:59, wrote: > I modified Xen kernel to turn off the core number 1 from 80th second to > 100th. > In 80th second I use cpu_down which is implemented in /xen/common/cpu.c to > turn off the 1st core. But, as you can see in below, the core does not be >

Re: [Xen-devel] does cpu_down makes brings core to deepest sleep state?

2015-10-26 Thread Jan Beulich
>>> On 26.10.15 at 12:13, wrote: >>> >>> >>> Did you look at the code generating the stats? >> >> No. Then may I suggest you do so, since ... >> C0 gets accounted >> to everything not explicitly accounted to any other C state (i.e. only >> C1...Cn get explicit stats

Re: [Xen-devel] does cpu_down makes brings core to deepest sleep state?

2015-10-26 Thread Jan Beulich
>>> On 26.10.15 at 11:40, wrote: > On Mon, Oct 26, 2015 at 1:14 PM, Jan Beulich wrote: >> ing maintained by (and hence re This is rather too little context you left in place. > I only turned off the cpu which has ID = 1 from 80th second to 100th >

Re: [Xen-devel] does cpu_down makes brings core to deepest sleep state?

2015-10-26 Thread Hamed Rs
> > > Did you look at the code generating the stats? No. > C0 gets accounted > to everything not explicitly accounted to any other C state (i.e. only > C1...Cn get explicit stats maintained). > > Jan > > I dont think so. C0 only accounted when all parts of cpu being on. I was wondering if

[Xen-devel] does cpu_down makes brings core to deepest sleep state?

2015-10-25 Thread Hamed Rs
Hi, I modified Xen kernel to turn off the core number 1 from 80th second to 100th. In 80th second I use cpu_down which is implemented in /xen/common/cpu.c to turn off the 1st core. But, as you can see in below, the core does not be in C4 idle state shile have almost 20 second CC6 state residency.