Hi,
I'd like to do some tests with cpuidle on my panda board but i'm facing some issues:
Originally, i'd like to use the Linaro developer image which uses the landing team branch tilt-linux-linaro-3.1 but the omap4 cpuidle driver only works when cpu1 is off whereas i'd like to do some test with both cpus. The android kernel has got a cpuidle driver which can enter C-states when both cpus are running and I was wondering if there is a plan to merge this driver into the tilt-linux-linaro-x.x branch ? maybe in 3.2 ?
Then, when I'm testing the landing-panda-11.12-release android image on my panda board revA1, the cpuidle is using only C0 state (WFI) according to cpuidle stats.
My test sequence: prevent suspend with: echo qwerty > /sys/power/wake_lock let android enter the early_suspend mode get cpuidle statistics on cpu0 et cpu1 in cpuidle/stateX/time : only state0 time is different from 0
Is it a normal behavior ?
Regards, Vincent
On 01/13/2012 05:59 PM, Somebody in the thread at some point said:
Hi,
I'd like to do some tests with cpuidle on my panda board but i'm facing some issues:
Originally, i'd like to use the Linaro developer image which uses the landing team branch tilt-linux-linaro-3.1 but the omap4 cpuidle driver only works when cpu1 is off whereas i'd like to do some test with both cpus. The android kernel has got a cpuidle driver which can enter C-states when both cpus are running and I was wondering if there is a plan to merge this driver into the tilt-linux-linaro-x.x branch ? maybe in 3.2 ?
Then, when I'm testing the landing-panda-11.12-release android image on my panda board revA1, the cpuidle is using only C0 state (WFI) according to cpuidle stats.
My test sequence: prevent suspend with: echo qwerty> /sys/power/wake_lock let android enter the early_suspend mode get cpuidle statistics on cpu0 et cpu1 in cpuidle/stateX/time : only state0 time is different from 0
Is it a normal behavior ?
No I think cpuidle is broken with what we have in tilt-3.1.
It's worth trying tilt-tracking which, last time Jassi looked at it was working and visiting all the C states.
-Andy
On 13 January 2012 16:13, Andy Green andy.green@linaro.org wrote:
On 01/13/2012 05:59 PM, Somebody in the thread at some point said:
Hi,
I'd like to do some tests with cpuidle on my panda board but i'm facing some issues:
Originally, i'd like to use the Linaro developer image which uses the landing team branch tilt-linux-linaro-3.1 but the omap4 cpuidle driver only works when cpu1 is off whereas i'd like to do some test with both cpus. The android kernel has got a cpuidle driver which can enter C-states when both cpus are running and I was wondering if there is a plan to merge this driver into the tilt-linux-linaro-x.x branch ? maybe in 3.2 ?
Then, when I'm testing the landing-panda-11.12-release android image on my panda board revA1, the cpuidle is using only C0 state (WFI) according to cpuidle stats.
My test sequence: prevent suspend with: echo qwerty> /sys/power/wake_lock let android enter the early_suspend mode get cpuidle statistics on cpu0 et cpu1 in cpuidle/stateX/time : only state0 time is different from 0
Is it a normal behavior ?
No I think cpuidle is broken with what we have in tilt-3.1.
It's worth trying tilt-tracking which, last time Jassi looked at it was working and visiting all the C states.
IIRC, CPU_0 entered C1, C2 states only when the CPU_1 is offline, when I tested.
Vincent, would you remember exactly which android branch has it working with both cpus online?
Cheers, Jassi
On 13 January 2012 12:12, Jassi Brar jaswinder.singh@linaro.org wrote:
On 13 January 2012 16:13, Andy Green andy.green@linaro.org wrote:
On 01/13/2012 05:59 PM, Somebody in the thread at some point said:
Hi,
I'd like to do some tests with cpuidle on my panda board but i'm facing some issues:
Originally, i'd like to use the Linaro developer image which uses the landing team branch tilt-linux-linaro-3.1 but the omap4 cpuidle driver only works when cpu1 is off whereas i'd like to do some test with both cpus. The android kernel has got a cpuidle driver which can enter C-states when both cpus are running and I was wondering if there is a plan to merge this driver into the tilt-linux-linaro-x.x branch ? maybe in 3.2 ?
Then, when I'm testing the landing-panda-11.12-release android image on my panda board revA1, the cpuidle is using only C0 state (WFI) according to cpuidle stats.
My test sequence: prevent suspend with: echo qwerty> /sys/power/wake_lock let android enter the early_suspend mode get cpuidle statistics on cpu0 et cpu1 in cpuidle/stateX/time : only state0 time is different from 0
Is it a normal behavior ?
No I think cpuidle is broken with what we have in tilt-3.1.
It's worth trying tilt-tracking which, last time Jassi looked at it was working and visiting all the C states.
IIRC, CPU_0 entered C1, C2 states only when the CPU_1 is offline, when I tested.
yes, but I need to keep both cpu online for my tests with sched_mc
Vincent, would you remember exactly which android branch has it working with both cpus online?
I'm using the the compiled baseline : https://android-build.linaro.org/builds/~linaro-android/landing-panda-11.12-... For this baseline, a cpuidle device is registered for each CPU (0,1) unlike the tilt tracking which register a device only for cpu0. This baseline uses an other cpuidle driver for omap4 which manages both CPU
Vincent
Cheers, Jassi
On 01/13/2012 08:34 PM, Somebody in the thread at some point said:
(adding AmitK)
Amit, do you know of a good place to raid for better cpuidle, closer to upstream than OZ?
-Andy
On 13 January 2012 12:12, Jassi Brarjaswinder.singh@linaro.org wrote:
On 13 January 2012 16:13, Andy Greenandy.green@linaro.org wrote:
On 01/13/2012 05:59 PM, Somebody in the thread at some point said:
Hi,
I'd like to do some tests with cpuidle on my panda board but i'm facing some issues:
Originally, i'd like to use the Linaro developer image which uses the landing team branch tilt-linux-linaro-3.1 but the omap4 cpuidle driver only works when cpu1 is off whereas i'd like to do some test with both cpus. The android kernel has got a cpuidle driver which can enter C-states when both cpus are running and I was wondering if there is a plan to merge this driver into the tilt-linux-linaro-x.x branch ? maybe in 3.2 ?
Then, when I'm testing the landing-panda-11.12-release android image on my panda board revA1, the cpuidle is using only C0 state (WFI) according to cpuidle stats.
My test sequence: prevent suspend with: echo qwerty> /sys/power/wake_lock let android enter the early_suspend mode get cpuidle statistics on cpu0 et cpu1 in cpuidle/stateX/time : only state0 time is different from 0
Is it a normal behavior ?
No I think cpuidle is broken with what we have in tilt-3.1.
It's worth trying tilt-tracking which, last time Jassi looked at it was working and visiting all the C states.
IIRC, CPU_0 entered C1, C2 states only when the CPU_1 is offline, when I tested.
yes, but I need to keep both cpu online for my tests with sched_mc
Vincent, would you remember exactly which android branch has it working with both cpus online?
I'm using the the compiled baseline : https://android-build.linaro.org/builds/~linaro-android/landing-panda-11.12-... For this baseline, a cpuidle device is registered for each CPU (0,1) unlike the tilt tracking which register a device only for cpu0. This baseline uses an other cpuidle driver for omap4 which manages both CPU
Vincent
Cheers, Jassi
On Fri, Jan 13, 2012 at 5:59 PM, Andy Green andy.green@linaro.org wrote:
Amit, do you know of a good place to raid for better cpuidle, closer to upstream than OZ?
wasn't that merged in 3.2? are you looking for that : http://git.kernel.org/?p=linux/kernel/git/torvalds/linux.git%3Ba=commit%3Bh=...
On 13 January 2012 22:44, Dechesne, Nicolas n-dechesne@ti.com wrote:
On Fri, Jan 13, 2012 at 5:59 PM, Andy Green andy.green@linaro.org wrote:
Amit, do you know of a good place to raid for better cpuidle, closer to upstream than OZ?
wasn't that merged in 3.2? are you looking for that : http://git.kernel.org/?p=linux/kernel/git/torvalds/linux.git%3Ba=commit%3Bh=...
We have this already. AIUI, Vincent wants CPU0 to switch between C1-2-3 even when CPU1 is online and apparently he has seen it possible (with the android link he provided).
-j
On Fri, Jan 13, 2012 at 7:14 PM, Dechesne, Nicolas n-dechesne@ti.com wrote:
On Fri, Jan 13, 2012 at 5:59 PM, Andy Green andy.green@linaro.org wrote:
Amit, do you know of a good place to raid for better cpuidle, closer to upstream than OZ?
wasn't that merged in 3.2? are you looking for that : http://git.kernel.org/?p=linux/kernel/git/torvalds/linux.git%3Ba=commit%3Bh=...
What was merged depends on hotplug, it isn't SMP cpuidle. At this point, I'm quite confused about the history on what happened to the SMP cpuidle driver and why it wasn't merged. Mike or Kevin should know.
/Amit
On Fri, Jan 13, 2012 at 2:41 PM, Amit Kucheria amit.kucheria@linaro.org wrote:
On Fri, Jan 13, 2012 at 7:14 PM, Dechesne, Nicolas n-dechesne@ti.com wrote:
On Fri, Jan 13, 2012 at 5:59 PM, Andy Green andy.green@linaro.org wrote:
Amit, do you know of a good place to raid for better cpuidle, closer to upstream than OZ?
wasn't that merged in 3.2? are you looking for that : http://git.kernel.org/?p=linux/kernel/git/torvalds/linux.git%3Ba=commit%3Bh=...
What was merged depends on hotplug, it isn't SMP cpuidle. At this point, I'm quite confused about the history on what happened to the SMP cpuidle driver and why it wasn't merged. Mike or Kevin should know.
Amit is correct. The hotplug-dependent cpuidle got merged.
The SMP CPUidle patches have been put on the list and should work patch into a recent kernel: http://article.gmane.org/gmane.linux.ports.tegra/2649
However the OMAP4 support seems to be missing: http://article.gmane.org/gmane.linux.ports.arm.omap/69186
So this isn't straightforward, but at least it's headed in the right direction.
Regards, Mike
On 01/14/2012 12:52 PM, Somebody in the thread at some point said:
On Fri, Jan 13, 2012 at 2:41 PM, Amit Kucheriaamit.kucheria@linaro.org wrote:
On Fri, Jan 13, 2012 at 7:14 PM, Dechesne, Nicolasn-dechesne@ti.com wrote:
On Fri, Jan 13, 2012 at 5:59 PM, Andy Greenandy.green@linaro.org wrote:
Amit, do you know of a good place to raid for better cpuidle, closer to upstream than OZ?
wasn't that merged in 3.2? are you looking for that : http://git.kernel.org/?p=linux/kernel/git/torvalds/linux.git%3Ba=commit%3Bh=...
What was merged depends on hotplug, it isn't SMP cpuidle. At this point, I'm quite confused about the history on what happened to the SMP cpuidle driver and why it wasn't merged. Mike or Kevin should know.
Amit is correct. The hotplug-dependent cpuidle got merged.
The SMP CPUidle patches have been put on the list and should work patch into a recent kernel: http://article.gmane.org/gmane.linux.ports.tegra/2649
However the OMAP4 support seems to be missing: http://article.gmane.org/gmane.linux.ports.arm.omap/69186
So this isn't straightforward, but at least it's headed in the right direction.
Thanks for the info... it sounds like we should try to add the SMP CPUidle from upstream and ask Santosh for his patches and see what happens.
Unfortunately we've been trying to use Andrey's new linux-linaro-tracking as a basis and despite fixing some of the problems the resulting tree still blows chunks midway through boot. So I think we'll have to give up on that for now and retry later.
-Andy
Hi All,
Thanks for these information, My understand is that I should wait for the adaptation of the SMP cpuidle proposal for omap4.
Andy, could you warn me when you will have a running smp cpu idle driver for panda ?
Regards, Vincent
On 14 January 2012 08:55, Andy Green andy.green@linaro.org wrote:
On 01/14/2012 12:52 PM, Somebody in the thread at some point said:
On Fri, Jan 13, 2012 at 2:41 PM, Amit Kucheriaamit.kucheria@linaro.org wrote:
On Fri, Jan 13, 2012 at 7:14 PM, Dechesne, Nicolasn-dechesne@ti.com wrote:
On Fri, Jan 13, 2012 at 5:59 PM, Andy Greenandy.green@linaro.org wrote:
Amit, do you know of a good place to raid for better cpuidle, closer to upstream than OZ?
wasn't that merged in 3.2? are you looking for that :
http://git.kernel.org/?p=linux/kernel/git/torvalds/linux.git%3Ba=commit%3Bh=...
What was merged depends on hotplug, it isn't SMP cpuidle. At this point, I'm quite confused about the history on what happened to the SMP cpuidle driver and why it wasn't merged. Mike or Kevin should know.
Amit is correct. The hotplug-dependent cpuidle got merged.
The SMP CPUidle patches have been put on the list and should work patch into a recent kernel: http://article.gmane.org/gmane.linux.ports.tegra/2649
However the OMAP4 support seems to be missing: http://article.gmane.org/gmane.linux.ports.arm.omap/69186
So this isn't straightforward, but at least it's headed in the right direction.
Thanks for the info... it sounds like we should try to add the SMP CPUidle from upstream and ask Santosh for his patches and see what happens.
Unfortunately we've been trying to use Andrey's new linux-linaro-tracking as a basis and despite fixing some of the problems the resulting tree still blows chunks midway through boot. So I think we'll have to give up on that for now and retry later.
-Andy
-- Andy Green | TI Landing Team Leader Linaro.org │ Open source software for ARM SoCs | Follow Linaro http://facebook.com/pages/Linaro/155974581091106 - http://twitter.com/#%21/linaroorg - http://linaro.org/linaro-blog
linaro-dev mailing list linaro-dev@lists.linaro.org http://lists.linaro.org/mailman/listinfo/linaro-dev
-
Amit Kucheria -
Andy Green -
Dechesne, Nicolas -
Jassi Brar -
Mike Turquette -
Vincent Guittot