What's Been Going On With CPUFreq & The Scheduler
As we've been covering the past few kernel cycles, a lot of low-level improvements have been happening to CPUFreq with going through a redesign and more plus the introduction of a new CPUFreq governor. If you're behind on this subject matter, here's some slides from this week's LinuxCon event that covers the changes.
Rafael J. Wysocki, the ACPI/PM subsystem maintainer for the Linux kernel, is once again presenting at LinuxCon. For LinuxCon NA 2016 he's talking about these CPUFreq changes that have been going on.
In case you missed it, the CPUFreq scheduler is now scheduler-driven since Linux 4.6 via callbacks and no longer relies upon timers. With Linux 4.7 was the new schedutil governor that makes use of scheduler data for making CPU frequency selection / power state decisions. These low-level changes though are really just the beginning with more improvements expected -- including making greater use of integration with the scheduler -- in future kernel updates.
You can see Rafael's PDF slides for those not in Toronto for LinuxCon.
Rafael J. Wysocki, the ACPI/PM subsystem maintainer for the Linux kernel, is once again presenting at LinuxCon. For LinuxCon NA 2016 he's talking about these CPUFreq changes that have been going on.
In case you missed it, the CPUFreq scheduler is now scheduler-driven since Linux 4.6 via callbacks and no longer relies upon timers. With Linux 4.7 was the new schedutil governor that makes use of scheduler data for making CPU frequency selection / power state decisions. These low-level changes though are really just the beginning with more improvements expected -- including making greater use of integration with the scheduler -- in future kernel updates.
You can see Rafael's PDF slides for those not in Toronto for LinuxCon.
23 Comments