Results 1 to 5 of 5

Thread: A Power Saving Schema For The Linux Kernel Scheduler

Hybrid View

  1. #1
    Join Date
    Jan 2007
    Posts
    15,138

    Default A Power Saving Schema For The Linux Kernel Scheduler

    Phoronix: A Power Saving Schema For The Linux Kernel Scheduler

    An Intel engineer has proposed introducing a power saving schema for CFS, the Linux kernel's default scheduler. Code hasn't been presented yet, but there's lots of discussion about this topic to improve the power efficiency of the Linux kernel scheduler...

    http://www.phoronix.com/vr.php?view=MTE2MjE

  2. #2
    Join Date
    Apr 2011
    Posts
    38

    Default

    Intel's work here, if it's merged, will benefit a lot of downstream projects that uses the Linux kernel such as Andorid, tizen, Open WebOS, meego (or whatever it's called now) etc. This is the beauty of open source, developers of products will indirectly help competeing products by contributing upstream.

  3. #3
    Join Date
    Feb 2012
    Posts
    70

    Default no turbo boost

    but doing this is against the Turbo boost that Intel have implemented in their CPU's. So this will cause performance degradation.

  4. #4
    Join Date
    Jan 2011
    Posts
    398

    Default

    This seems to be in conflict with the sched_mc_power_savings and sched_smt_power_savings parameters that (from what I gathered) tell the kernel to group processes on a single core in order to put the others to sleep and reduce power consumptionů?

  5. #5
    Join Date
    Sep 2009
    Posts
    116

    Default

    Quote Originally Posted by mayankleoboy1 View Post
    but doing this is against the Turbo boost that Intel have implemented in their CPU's. So this will cause performance degradation.
    I'm sure this is primarily targeting power constrained platforms like Android. Remember that not all Intel CPUs have turbo and additionally I would think this power saving schema would be optional like many other parameters in CFS.

    Quote Originally Posted by stqn
    This seems to be in conflict with the sched_mc_power_savings and sched_smt_power_savings parameters that (from what I gathered) tell the kernel to group processes on a single core in order to put the others to sleep and reduce power consumptionů?
    In a world with DVFS I don't think keeping one core at 100% load is as efficient as two cores at even 75% load. Remember that Power = Voltage^2 / Resistance. Keeping voltage lower can make a huge difference on power draw even if it is at the cost of keeping additional cores active.

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •