correct CONFIG_HZ option
Hi All, I am using ARM multicore board and by default CONFIG_HZ=250 Query: - 1- How to decide HZ for particular hardware 2- Which is the best open source tool to test system performance from given HZ. Thanks, John
Hi... :) On Tue, Jun 26, 2012 at 12:10 PM, solmac john <johnsolmac@gmail.com> wrote:
I am using ARM multicore board and by default
okay, I am answering it from what I know about HZ impact on x86...
CONFIG_HZ=250
looks good... a middle safe number, not too high not too low...
Query: - 1- How to decide HZ for particular hardware
it's you who decide...do you want finer grained timer? or coarse one? the impact is usually toward latency and responsiveness....together with preemption model you choose actually.
2- Which is the best open source tool to test system performance from given HZ.
run your application in that platform and see if it gives you impact. from my experience, unless you need application that is sensitive in timing such as MIDI sequencer, you won't notice the difference. -- regards, Mulyadi Santosa Freelance Linux trainer and consultant blog: the-hydra.blogspot.com training: mulyaditraining.blogspot.com
Thanks for reply what is CONFIG_NO_HZ can we enable CONFIG_NO_HZ and CONFIG_HZ both in our config and what the drawback of dynamic CONFIG_NO_HZ tick. On Tue, Jun 26, 2012 at 10:47 AM, Mulyadi Santosa <mulyadi.santosa@gmail.com
wrote:
Hi... :)
On Tue, Jun 26, 2012 at 12:10 PM, solmac john <johnsolmac@gmail.com> wrote:
I am using ARM multicore board and by default
okay, I am answering it from what I know about HZ impact on x86...
CONFIG_HZ=250
looks good... a middle safe number, not too high not too low...
Query: - 1- How to decide HZ for particular hardware
it's you who decide...do you want finer grained timer? or coarse one?
the impact is usually toward latency and responsiveness....together with preemption model you choose actually.
2- Which is the best open source tool to test system performance from given HZ.
run your application in that platform and see if it gives you impact. from my experience, unless you need application that is sensitive in timing such as MIDI sequencer, you won't notice the difference.
-- regards,
Mulyadi Santosa Freelance Linux trainer and consultant
blog: the-hydra.blogspot.com training: mulyaditraining.blogspot.com
Hi... On Thu, Jun 28, 2012 at 6:21 PM, solmac john <johnsolmac@gmail.com> wrote:
Thanks for reply
what is CONFIG_NO_HZ
it enables so called dynamic tick....
can we enable CONFIG_NO_HZ and CONFIG_HZ both in our config
yup...that's fine...even better, it's recommended nowadays to reduce power consumption at certain degree.
and what the drawback of dynamic CONFIG_NO_HZ tick.
uhm, some people said to me that in certain combination, config_no_hz will introduce higher latency. This is due to the cost of occasional timer rearming, instead of classic approach where timer is programmed to shot regularly. But, if your setup is not very latency sensitive, the overhead is most likely negligible. -- regards, Mulyadi Santosa Freelance Linux trainer and consultant blog: the-hydra.blogspot.com training: mulyaditraining.blogspot.com
Hi, AFAIK. If you enable CONFIG_NO_HZ, then there is no meaning of HZ. If CPU is heavy loaded it run with 10ms, and it idle then timer tick is off. 2. drawback of dynamic CONFIG_NO_HZ tick a) slow response (less interactive system) On Fri, Jun 29, 2012 at 9:36 AM, Mulyadi Santosa <mulyadi.santosa@gmail.com> wrote:
Hi...
On Thu, Jun 28, 2012 at 6:21 PM, solmac john <johnsolmac@gmail.com> wrote:
Thanks for reply
what is CONFIG_NO_HZ
it enables so called dynamic tick....
can we enable CONFIG_NO_HZ and CONFIG_HZ both in our config
yup...that's fine...even better, it's recommended nowadays to reduce power consumption at certain degree.
and what the drawback of dynamic CONFIG_NO_HZ tick.
uhm, some people said to me that in certain combination, config_no_hz will introduce higher latency. This is due to the cost of occasional timer rearming, instead of classic approach where timer is programmed to shot regularly. But, if your setup is not very latency sensitive, the overhead is most likely negligible.
-- regards,
Mulyadi Santosa Freelance Linux trainer and consultant
blog: the-hydra.blogspot.com training: mulyaditraining.blogspot.com
_______________________________________________ Kernelnewbies mailing list Kernelnewbies@kernelnewbies.org http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
On 06/28/2012 01:21 PM, solmac john wrote:
Thanks for reply what is CONFIG_NO_HZ Hi!
If you have no idea what a config entry means, use the help in menuconfig (or whatever you are using)... e.g. for CONFIG_NO_HZ tells you ----------<snip>-------------- CONFIG_NO_HZ: This option enables a tickless system: timer interrupts will only trigger on an as-needed basis both when the system is busy and when the system is idle. Symbol: NO_HZ [=n] Type : boolean Prompt: Tickless System (Dynamic Ticks) Defined at kernel/time/Kconfig:7 Depends on: !ARCH_USES_GETTIMEOFFSET && GENERIC_CLOCKEVENTS [=y] Location: -> Processor type and features Selects: TICK_ONESHOT [=n] ----------<snap>------------- So this already gives you a general idea what the config entry does, and some other useful info like dependency on other stuff that has to be turn on/off.
can we enable CONFIG_NO_HZ and CONFIG_HZ both in our config This question is also answered here, as "Depends on:" would include !CONFIG_HZ_250 if it had to be switched or sth.
and what the drawback of dynamic CONFIG_NO_HZ tick.
I think the problems of longer latencies only apply on idle systems. Not 100% sure, but I think when the cpu goes into idle, the periodic timer is stopped and longer intervals are done using one-shot timer to avoid forcing the cpu out of idle just to see that nothing is to do and go back into idle. But I think if the CPU is not in idle the periodic timer. regards, andi
On Tue, Jun 26, 2012 at 10:47 AM, Mulyadi Santosa <mulyadi.santosa@gmail.com <mailto:mulyadi.santosa@gmail.com>> wrote:
Hi... :)
On Tue, Jun 26, 2012 at 12:10 PM, solmac john <johnsolmac@gmail.com <mailto:johnsolmac@gmail.com>> wrote: > I am using ARM multicore board and by default
okay, I am answering it from what I know about HZ impact on x86...
> CONFIG_HZ=250
looks good... a middle safe number, not too high not too low...
> Query: - 1- How to decide HZ for particular hardware
it's you who decide...do you want finer grained timer? or coarse one?
the impact is usually toward latency and responsiveness....together with preemption model you choose actually.
> 2- Which is the best open source tool to test system > performance from given HZ.
run your application in that platform and see if it gives you impact. from my experience, unless you need application that is sensitive in timing such as MIDI sequencer, you won't notice the difference.
-- regards,
Mulyadi Santosa Freelance Linux trainer and consultant
blog: the-hydra.blogspot.com <http://the-hydra.blogspot.com/> training: mulyaditraining.blogspot.com <http://mulyaditraining.blogspot.com/>
_______________________________________________ Kernelnewbies mailing list Kernelnewbies@kernelnewbies.org http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
participants (4)
-
Andreas Platschek -
Mulyadi Santosa -
naveen yadav -
solmac john