lkml.org 
[lkml]   [2016]   [Jan]   [28]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH v4] perf/x86/amd/power: Add AMD accumulated power reporting mechanism
On Thu, Jan 28, 2016 at 06:01:43PM +0800, Huang Rui wrote:
> For example: Carrizo has four CPU cores and two compute units (CUs).
> CPU0 and CPU1 belongs to CU0, CPU2 and CPU3 belongs to CU1.
>
> At normal initialization, cpu_mask should be "0,2". That means OS
> choose CPU0 in CU0 and CPU2 in CU1 to measure the CU0 and CU1's power
> consumption. If we make the CPU2 offline at runtime, OS need try to
> find another CPU in same compute unit (Here is CU1, only CPU3 can be
> picked). Then OS will move on to the CPU3 to measure CU1's power
> consumption instead of CPU2.

So basically you want to simply say:

"Find another CPU on the same compute unit and set it in the mask of
CPUs on which we do the measurements."

Which reminds me: that cpu_mask thing is insufficiently named - it
should be called measuring_cpus_mask or so.

Btw, the kbuild robot errors come from the fact that there are changes
to cpufeature.h which I didn't mention when applying your patches. So
I've pushed the whole pile here:

http://git.kernel.org/cgit/linux/kernel/git/bp/bp.git/log/?h=tip-perf

Please use that branch instead.

Thanks.

--
Regards/Gruss,
Boris.

ECO tip #101: Trim your mails when you reply.

\
 
 \ /
  Last update: 2016-01-28 14:21    [W:0.069 / U:0.356 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site