From: Yong Shen <redacted>
Add cpuidle parameters to make cpuidle driver workable, but these
parameters need further tuning
Signed-off-by: Yong Shen <redacted>
---
arch/arm/mach-mx5/board-mx51_babbage.c | 114 ++++++++++++++++++++++++++++++++
arch/arm/mach-mx5/devices.c | 4 +
arch/arm/mach-mx5/devices.h | 1 +
3 files changed, 119 insertions(+), 0 deletions(-)
I didn't look at how it's working nor did a review but from a very quick
look, I'm wondering why it's in board-mx51_babbage.c. Can you explain to
me what makes this code specific to babbage ?
Arnaud
Hi Arnaud,
I also took a while to think about this before posting patches. I prefer to
put it in board related code since the various PMIC used on each boards may
have influence on cpuidle latency or other charactors, although it could be
minor.
Yong
On Tue, Feb 8, 2011 at 4:29 PM, Arnaud Patard [off-list ref]wrote:
yong.shen at linaro.org writes:
Hi,
quoted
From: Yong Shen <redacted>
Add cpuidle parameters to make cpuidle driver workable, but these
parameters need further tuning
Signed-off-by: Yong Shen <redacted>
---
arch/arm/mach-mx5/board-mx51_babbage.c | 114
I didn't look at how it's working nor did a review but from a very quick
look, I'm wondering why it's in board-mx51_babbage.c. Can you explain to
me what makes this code specific to babbage ?
Arnaud
From: Amit Kucheria <hidden> Date: 2011-02-08 17:16:35
On 11 Feb 08, Yong Shen wrote:
Hi Arnaud,
I also took a while to think about this before posting patches. I prefer to
put it in board related code since the various PMIC used on each boards may
have influence on cpuidle latency or other charactors, although it could be
minor.
Then those minor differences can be addressed in the common driver by using #ifdef
conditionals.
The cpuidle driver certainly shouldn't be board-specific.
/Amit
From: Rob Herring <hidden> Date: 2011-02-08 17:50:27
On 02/08/2011 09:51 AM, Yong Shen wrote:
Hi Arnaud,
I also took a while to think about this before posting patches. I prefer
to put it in board related code since the various PMIC used on each
boards may have influence on cpuidle latency or other charactors,
although it could be minor.
But you are not going to be doing voltage scaling in idle. Is it even
possible to do sleeping operations like accessing a PMIC in idle?
The core is powergated, so lowering voltage would not help. Doing bus
scaling or DDR self-refresh are the only likely additional operations.
Rob
On Tue, Feb 8, 2011 at 6:50 PM, Rob Herring [off-list ref] wrote:
On 02/08/2011 09:51 AM, Yong Shen wrote:
quoted
Hi Arnaud,
I also took a while to think about this before posting patches. I prefer
to put it in board related code since the various PMIC used on each
boards may have influence on cpuidle latency or other charactors,
although it could be minor.
But you are not going to be doing voltage scaling in idle. Is it even
possible to do sleeping operations like accessing a PMIC in idle?
I guess you are asking if modes like 'state retention' are possible for idle
state. so far there is no official mapping between c-states and various arm
idle states, which is something linaro power management group is working on.
Therefore I also treat 'state retention' as a c-state.
ps. remove lpdk in cc list to avoid auto-reply messages.
Yong
The core is powergated, so lowering voltage would not help. Doing bus
scaling or DDR self-refresh are the only likely additional operations.
Rob
Yong,
On Tue, Feb 8, 2011 at 9:21 PM, Yong Shen [off-list ref] wrote:
Hi Arnaud,
I also took a while to think about this before posting patches. I prefer to
put it in board related code since the various PMIC used on each boards may
have influence on cpuidle latency or other charactors, although it could be
minor.
WHy don't you define cpuidle_latency table per board file and use the
right one inside the common cpuidle driver. Then you can take care of
board related latencies cleanly.
Vishwa
Yong
On Tue, Feb 8, 2011 at 4:29 PM, Arnaud Patard [off-list ref]
wrote:
quoted
yong.shen at linaro.org writes:
Hi,
quoted
From: Yong Shen <redacted>
Add cpuidle parameters to make cpuidle driver workable, but these
parameters need further tuning
Signed-off-by: Yong Shen <redacted>
---
?arch/arm/mach-mx5/board-mx51_babbage.c | ?114
++++++++++++++++++++++++++++++++
?arch/arm/mach-mx5/devices.c ? ? ? ? ? ?| ? ?4 +
?arch/arm/mach-mx5/devices.h ? ? ? ? ? ?| ? ?1 +
?3 files changed, 119 insertions(+), 0 deletions(-)
I didn't look at how it's working nor did a review but from a very quick
look, I'm wondering why it's in board-mx51_babbage.c. Can you explain to
me what makes this code specific to babbage ?
Arnaud
Hi Vishwanath,
Thanks. I will take it as a reference.
Yong
On Wed, Feb 9, 2011 at 11:52 AM, Vishwanath Sripathy <
vishwanath.sripathy@linaro.org> wrote:
Yong,
On Tue, Feb 8, 2011 at 9:21 PM, Yong Shen [off-list ref] wrote:
quoted
Hi Arnaud,
I also took a while to think about this before posting patches. I prefer
to
quoted
put it in board related code since the various PMIC used on each boards
may
quoted
have influence on cpuidle latency or other charactors, although it could
be
quoted
minor.
WHy don't you define cpuidle_latency table per board file and use the
right one inside the common cpuidle driver. Then you can take care of
board related latencies cleanly.
Vishwa
quoted
Yong
On Tue, Feb 8, 2011 at 4:29 PM, Arnaud Patard <arnaud.patard@rtp-net.org
wrote:
quoted
yong.shen at linaro.org writes:
Hi,
quoted
From: Yong Shen <redacted>
Add cpuidle parameters to make cpuidle driver workable, but these
parameters need further tuning
Signed-off-by: Yong Shen <redacted>
---
arch/arm/mach-mx5/board-mx51_babbage.c | 114
++++++++++++++++++++++++++++++++
arch/arm/mach-mx5/devices.c | 4 +
arch/arm/mach-mx5/devices.h | 1 +
3 files changed, 119 insertions(+), 0 deletions(-)
I didn't look at how it's working nor did a review but from a very quick
look, I'm wondering why it's in board-mx51_babbage.c. Can you explain to
me what makes this code specific to babbage ?
Arnaud