82xx CPU features

82xx CPU features

[Guennadi Liakhovetski]

Hi all

wandering around the idle code, noticed the following:

#define CPU_FTRS_603	(CPU_FTR_COMMON | CPU_FTR_SPLIT_ID_CACHE | \
	    CPU_FTR_MAYBE_CAN_DOZE | CPU_FTR_USE_TB | \
	    CPU_FTR_MAYBE_CAN_NAP | CPU_FTR_PPC_LE)

...

#define CPU_FTRS_82XX	(CPU_FTR_COMMON | CPU_FTR_SPLIT_ID_CACHE | \
	    CPU_FTR_MAYBE_CAN_DOZE | CPU_FTR_USE_TB)

Whereas even in

static struct cpu_spec cpu_specs[] = {
...
	{	/* 82xx (8240, 8245, 8260 are all 603e cores) */
...

So, wanted to ask - is there a specific reason why CPU_FTR_MAYBE_CAN_NAP 
is not set for 82xx (as well as CPU_FTR_PPC_LE for that matter...)

Thanks
Guennadi
---
Guennadi Liakhovetski

Re: 82xx CPU features

[Kumar Gala]

On Jan 25, 2007, at 4:39 PM, Guennadi Liakhovetski wrote:

> Hi all
>
> wandering around the idle code, noticed the following:
>
> #define CPU_FTRS_603	(CPU_FTR_COMMON | CPU_FTR_SPLIT_ID_CACHE | \
> 	    CPU_FTR_MAYBE_CAN_DOZE | CPU_FTR_USE_TB | \
> 	    CPU_FTR_MAYBE_CAN_NAP | CPU_FTR_PPC_LE)
>
> ...
>
> #define CPU_FTRS_82XX	(CPU_FTR_COMMON | CPU_FTR_SPLIT_ID_CACHE | \
> 	    CPU_FTR_MAYBE_CAN_DOZE | CPU_FTR_USE_TB)
>
> Whereas even in
>
> static struct cpu_spec cpu_specs[] = {
> ...
> 	{	/* 82xx (8240, 8245, 8260 are all 603e cores) */
> ...
>
> So, wanted to ask - is there a specific reason why  
> CPU_FTR_MAYBE_CAN_NAP
> is not set for 82xx (as well as CPU_FTR_PPC_LE for that matter...)

For _CAN_NAP, I'm guessing its historic, maybe Dan knows why it  
wasn't set.  What happens if you try to set it?  Are there  
implications at the system level with regards to _CAN_NAP?

I'm looking at the CPU_FTR_PPC_LE it varies if we should use _PPC_LE  
vs _TRUE_LE (or whatever the other one is called).

- k

Re: 82xx CPU features

[Guennadi Liakhovetski]

On Fri, 26 Jan 2007, Kumar Gala wrote:

> For _CAN_NAP, I'm guessing its historic, maybe Dan knows why it wasn't set.
> What happens if you try to set it?  Are there implications at the system level
> with regards to _CAN_NAP?

Just tested: set CPU_FTR_MAYBE_CAN_DOZE for 82xx, CONFIG_BDI_SWITCH is not 
defined, echoed 1 to /proc/sys/kernel/powersave-nap - and a 8241 still 
runs... I'll try to measure if there's any difference in power 
consumption:-))

> I'm looking at the CPU_FTR_PPC_LE it varies if we should use _PPC_LE vs
> _TRUE_LE (or whatever the other one is called).

Thanks
Guennadi
---
Guennadi Liakhovetski

Re: 82xx CPU features

[Kumar Gala]

On Jan 30, 2007, at 4:31 PM, Guennadi Liakhovetski wrote:

> On Fri, 26 Jan 2007, Kumar Gala wrote:
>
>> For _CAN_NAP, I'm guessing its historic, maybe Dan knows why it  
>> wasn't set.
>> What happens if you try to set it?  Are there implications at the  
>> system level
>> with regards to _CAN_NAP?
>
> Just tested: set CPU_FTR_MAYBE_CAN_DOZE for 82xx, CONFIG_BDI_SWITCH  
> is not
> defined, echoed 1 to /proc/sys/kernel/powersave-nap - and a 8241 still
> runs... I'll try to measure if there's any difference in power
> consumption:-))
>
>> I'm looking at the CPU_FTR_PPC_LE it varies if we should use  
>> _PPC_LE vs
>> _TRUE_LE (or whatever the other one is called).

It would be good for someone to test an 8260 class (826x, 8272,  
etc..) to see if they have any issues.

- k

Re: 82xx CPU features

[Vitaly Bordug]

[-- Attachment #1: Type: text/plain, Size: 978 bytes --]

On Tue, 30 Jan 2007 16:35:03 -0600
Kumar Gala wrote:

> 
> On Jan 30, 2007, at 4:31 PM, Guennadi Liakhovetski wrote:
> 
> > On Fri, 26 Jan 2007, Kumar Gala wrote:
> >
> >> For _CAN_NAP, I'm guessing its historic, maybe Dan knows why it  
> >> wasn't set.
> >> What happens if you try to set it?  Are there implications at the  
> >> system level
> >> with regards to _CAN_NAP?
> >
> > Just tested: set CPU_FTR_MAYBE_CAN_DOZE for 82xx,
> > CONFIG_BDI_SWITCH is not
> > defined, echoed 1 to /proc/sys/kernel/powersave-nap - and a 8241
> > still runs... I'll try to measure if there's any difference in power
> > consumption:-))
> >
> >> I'm looking at the CPU_FTR_PPC_LE it varies if we should use  
> >> _PPC_LE vs
> >> _TRUE_LE (or whatever the other one is called).
> 
> It would be good for someone to test an 8260 class (826x, 8272,  
> etc..) to see if they have any issues.

I can see with 8272 if that makes any difference.
--
Thanks, Vitaly

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 189 bytes --]