Re: [PATCH v2 1/4] x86, mwaitt: add monitorx and mwaitx instruction

[Huang Rui <ray.huang@amd.com>]

On Tue, Jun 09, 2015 at 07:06:14PM +0200, Borislav Petkov wrote:
> On Tue, Jun 09, 2015 at 09:44:59AM -0700, Andy Lutomirski wrote:
> > [1] For those who weren't bitten by this repeatedly, modern Intel CPUs
> > (at least Sandy Bridge, anyway) will, by default, detect when all
> > cores are in C1 or deeper, think to themselves "wow, the OS selected
> > C1 -- it must want a very deep sleep indeed", and put the whole
> > package into some kind of deep sleep state.  The subsequent wakeup
> > takes tens of milliseconds.  Doing this in udelay would be awful.
> 
> That's a good point. Reportedly, the current MWAITX enters something
> between C0 and C1 but the way I understood it, going forward, it will
> enter deeper sleep states.
> 
> So for shallow C-states, your idle enter/exit latency is low enough but
> I'd guess deeper states would be a problem.
> 

Andy, Boris, that's right. Thanks. :)
If MWAITX will enter C1 and restore back repeatedly. It will impact
the accuracy of delay. I will ask HW designer to check if it already
has configuration to control the target power state.

Thanks,
Rui