Active Topics
-
Facebook chat on Nokia N9 - back to life :) (41)
to MeeGo / Harmattan by shakolan - 19 hrs, 28 mins ago -
Any chance to get back mirror.thecust.net ? (3)
to MeeGo / Harmattan by dredlok706 - 19 hrs, 38 mins ago -
TLS 1.2 for N9 is ready (17)
to MeeGo / Harmattan by shakolan - 21 hrs, 9 mins ago -
Pinephone and Pinephone Pro on Leste vs Mobian (0)
to Maemo 7 / Leste by biketool - 1 day, 3 hrs ago -
need .bin for update nokia 770 tablet (17)
to Nokia 770 by amigo157 - 6 days ago -
I have found alien dalvik rpm (3)
to SailfishOS by HtheB - 6 days, 9 hrs ago - more...
|
2011-03-22
, 11:41
|
|
Posts: 27 |
Thanked: 3 times |
Joined on Jan 2008
|
#272
|
I've been following kernel development for a long time. An enhanced, bug fixed kernel is _very_ welcome. Thanks.
> 256 HZ timer: The diablo kernel has a slow 128 HZ timer,
> strongly related to the number of task switches the task
> scheduler can do. This kernel is configured with a 256 HZ
> timer, and additionally the SD scheduler allows a task switch
> within a much more shorter time (defined by the rr_interval
> tunable). This means that the device will be much more
> interactive.
Have you checked what changing to a higher HZ value does to power usage? I suspect it increases it a fair amount. If I remember correctly diablo uses the dyntick patchset to reduce this effect, but there are programs that disable dyntick and cause power to be eaten.
Its probably worth testing with the lower HZ value - suspect it will not make a great difference to response but should help power...
On the scheduler. Have you queried Con about SD vs BFS? A patchset porting a version of BFS to 2.6.21 exists. Depending on what Con suggests, it might be worth trying.
There is also a recent patch set from intel that improves the ondemand policy. On my desktop it gives a noticeable reduction in power usage (my UPS is good for longer). If this can be back ported it would be nice.
Would you happen to have a git tree with the kernel source you are using available?
Thanks Again,
Ed
> 256 HZ timer: The diablo kernel has a slow 128 HZ timer,
> strongly related to the number of task switches the task
> scheduler can do. This kernel is configured with a 256 HZ
> timer, and additionally the SD scheduler allows a task switch
> within a much more shorter time (defined by the rr_interval
> tunable). This means that the device will be much more
> interactive.
Have you checked what changing to a higher HZ value does to power usage? I suspect it increases it a fair amount. If I remember correctly diablo uses the dyntick patchset to reduce this effect, but there are programs that disable dyntick and cause power to be eaten.
Its probably worth testing with the lower HZ value - suspect it will not make a great difference to response but should help power...
On the scheduler. Have you queried Con about SD vs BFS? A patchset porting a version of BFS to 2.6.21 exists. Depending on what Con suggests, it might be worth trying.
There is also a recent patch set from intel that improves the ondemand policy. On my desktop it gives a noticeable reduction in power usage (my UPS is good for longer). If this can be back ported it would be nice.
Would you happen to have a git tree with the kernel source you are using available?
Thanks Again,
Ed
|
|
2011-03-22
, 14:03
|
|
Posts: 1,101 |
Thanked: 1,184 times |
Joined on Aug 2008
@ Spain
|
#273
|
Originally Posted by edt
Yes, the current kernel flavour has been tested since early january. You are a bit late to the party...
Have you checked what changing to a higher HZ value does to power usage?
I suspect it increases it a fair amount.
If I remember correctly diablo uses the dyntick patchset to reduce this effect, but there are programs that disable dyntick and cause power to be eaten.
The timer just provides a timeframe for the scheduler, and the number of tasks switches (within a certain order) only increases the work done very slightly, so when the cpu is at work, the impact is negligible.
When the cpu is at rest, the timer generates no interrupts, so no work is done. If a program disables dyntic, it doesn't matter much 128 vs 256, the battery will drain quite fast anyway, the problem is such program.
On the scheduler. Have you queried Con about SD vs BFS? A patchset porting a version of BFS to 2.6.21 exists. Depending on what Con suggests, it might be worth trying.
There is also a recent patch set from intel that improves the ondemand policy. On my desktop it gives a noticeable reduction in power usage (my UPS is good for longer). If this can be back ported it would be nice.
If it is for current kernels, I doubt it can be ported, at least with a reasonable amount of work.
Would you happen to have a git tree with the kernel source you are using available?
Last edited by maacruz; 2011-03-22 at 14:10.
 |
|
2011-03-22
, 15:20
|
|
Posts: 708 |
Thanked: 125 times |
Joined on Jan 2007
@ Too Close To D.C
|
#274
|
Question about swap memory size. My swap memory size goes to the limit that was set via the control panel but nevers goes back to 0 when no apps are being run. Should the swap memory size go towards 0 when the unit is not running any apps?
__________________
N810-WiMAX
N810-WiMAX
|
|
2011-03-22
, 16:12
|
|
Posts: 1,101 |
Thanked: 1,184 times |
Joined on Aug 2008
@ Spain
|
#275
|
Originally Posted by heavyt
No, once something goes to swap, it stays there till is needed again or it is freed, so swap never goes to 0.
Question about swap memory size. My swap memory size goes to the limit that was set via the control panel but nevers goes back to 0 when no apps are being run. Should the swap memory size go towards 0 when the unit is not running any apps?
| The Following User Says Thank You to maacruz For This Useful Post: | ||
 |
|
2011-03-22
, 17:18
|
|
Posts: 218 |
Thanked: 59 times |
Joined on Feb 2010
@ spain
|
#276
|
Hi, maacruz.
Yesterday the omelette was done and the picnic at the country was successful. Thanks
I have installed all the packages the old and the upgraded ones.
With the new bluez-utils package I can pair with my BH-214 headset and it plays for 2-3 seconds aprox. Then it suddenly stops playing. Pairing is still established but no sound trough headset.
I'm still testing.
Yesterday the omelette was done and the picnic at the country was successful. Thanks
I have installed all the packages the old and the upgraded ones.
With the new bluez-utils package I can pair with my BH-214 headset and it plays for 2-3 seconds aprox. Then it suddenly stops playing. Pairing is still established but no sound trough headset.
I'm still testing.
| The Following User Says Thank You to Straycat For This Useful Post: | ||
|
|
2011-03-22
, 19:36
|
|
Posts: 35 |
Thanked: 14 times |
Joined on Oct 2010
@ Durham, NC
|
#277
|
I installed the original DT along with the first update. Then I installed the zimage kernel from 3/9. So, to install the 2nd update do I simply do the following?
1. Download only the "update 2" deb packages to "/media/mmc1/dt"
2. Open a xterm and become root
3. Change to the directory where the packages are (cd /media/mmc1/dt)
4. Install all debs ("dpkg -i *.deb")
I just want to make sure before I actually go ahead with it. Please let me know if I am missing any steps. Thanks!
1. Download only the "update 2" deb packages to "/media/mmc1/dt"
2. Open a xterm and become root
3. Change to the directory where the packages are (cd /media/mmc1/dt)
4. Install all debs ("dpkg -i *.deb")
I just want to make sure before I actually go ahead with it. Please let me know if I am missing any steps. Thanks!
|
|
2011-03-22
, 22:12
|
|
Posts: 1,101 |
Thanked: 1,184 times |
Joined on Aug 2008
@ Spain
|
#278
|
Originally Posted by jmart
You need to install only the update2 packages. Of course, installing everything will work too.
I installed the original DT along with the first update. Then I installed the zimage kernel from 3/9. So, to install the 2nd update do I simply do the following?
1. Download only the "update 2" deb packages to "/media/mmc1/dt"
2. Open a xterm and become root
3. Change to the directory where the packages are (cd /media/mmc1/dt)
4. Install all debs ("dpkg -i *.deb")
I just want to make sure before I actually go ahead with it. Please let me know if I am missing any steps. Thanks!
| The Following User Says Thank You to maacruz For This Useful Post: | ||
 |
|
2011-03-22
, 22:58
|
|
Posts: 1,137 |
Thanked: 402 times |
Joined on Sep 2007
@ Catalunya
|
#279
|
Originally Posted by luca
I doubt it I'm running the old kernel:
~$ uname -a
Linux noname 2.6.21-omap1#6 PREEMPT Sat Feb 19 10:59:14 CET 2011 armv6l unknown
(if for old kernel you mean the community update kernel, that was the first time I installed diablo turbo).
Originally Posted by maacruz
uname -a
Linux noname 2.6.21-omap1 #2 PREEMPT Sun Mar 13 23:29:39 CET 2011 armv6l unknown
Ok, but the "Sat Feb 19" one cannot be from the community update (I checked the repository and that was in august 2010), according to the date it must be one of the diablo turbo one.
The only diablo turbo kernel package I installed is the "broken" one, and you said that, due to a bug, it didn't really install the kernel.
I'm confused now, (more than usual
)
|
|
2011-03-23
, 01:32
|
|
Posts: 27 |
Thanked: 3 times |
Joined on Jan 2008
|
#280
|
> I haven't contacted Con in any way. At the time I couln't find any
> port of BFS to 2.6.21, so if you know one, please provide the link.
My error here. The back port was actually from 2.6.31 to 2.6.29. The message, but not the actual patch, said .21....
> What is the effect of that patch? Links?
> if it is for current kernels, I doubt it can be ported, at least with a > reasonable amount of work.
I allows the ondemand policy to use intermediate frequencies more often instead of just the highest and lowest.
> No, just the set of patches already published and the diablo
> kernel tar.gz. I play on a kernel tree, and when I'm done with
> whatever I'm trying to do I extract a pach and destroy the tree to
> start afresh again.
Does this have all the patches (nokia's too) for n8x0 included. If not where are the patches from nokia? I'll see if the ondemand stuff can be backported.
TIA
Ed
> port of BFS to 2.6.21, so if you know one, please provide the link.
My error here. The back port was actually from 2.6.31 to 2.6.29. The message, but not the actual patch, said .21....
> What is the effect of that patch? Links?
> if it is for current kernels, I doubt it can be ported, at least with a > reasonable amount of work.
I allows the ondemand policy to use intermediate frequencies more often instead of just the highest and lowest.
> No, just the set of patches already published and the diablo
> kernel tar.gz. I play on a kernel tree, and when I'm done with
> whatever I'm trying to do I extract a pach and destroy the tree to
> start afresh again.
Does this have all the patches (nokia's too) for n8x0 included. If not where are the patches from nokia? I'll see if the ondemand stuff can be backported.
TIA
Ed
Linux noname 2.6.21-omap1 #2 PREEMPT Sun Mar 13 23:29:39 CET 2011 armv6l unknown