C2/C3 states cause poor TCP performance with re(4)

[Aragon Gouveia]

On 03/13/11 09:36, Thor Ablestar wrote:
> I have some similar but unrelated problems and during googling saw the
> article that states that wakeup from C3 takes significant amount of time
> (57 uS). You cannot both have a zero response time and C3.
>> dev.cpu.1.cx_supported: C1/1 C2/1 C3/57
>
> To accelerate things you should keep 1 core in C1 and all others in C3.

Makes sense, and your suggestion works too.  Thanks!