sched: fair: Reduce runtime allocated for tasks constrained by CFS bandwidth
A bunch of kernels for desktop Linux have been reducing this value to improve interactivity. From Zen[1] to CachyOS[2]. There have been attempts to reduce it on Android as well. Experiment with reducing the CFS bandwidth slice to 4 msec, 1 less from the default. This is something I honestly don't want userspace to touch so keep it out from sysfs and modify it from the kernel directly instead. I honestly think that the 'interactivity' benefits (if it does hold water) of this change should be reflected on all performance modes on FreshROMs. Test for performance and battery life. [1]: https://github.com/zen-kernel/zen-kernel/commit/7de2596b35ac1db [2]: https://github.com/CachyOS/linux/blob/base-5.18/kernel/sched/fair.c Signed-off-by: John Vincent <git@tenseventyseven.cf>
This commit is contained in:
parent
9341ba576e
commit
fff787f527
1 changed files with 1 additions and 1 deletions
|
@ -132,7 +132,7 @@ int __weak arch_asym_cpu_priority(int cpu)
|
|||
*
|
||||
* (default: 5 msec, units: microseconds)
|
||||
*/
|
||||
unsigned int sysctl_sched_cfs_bandwidth_slice = 5000UL;
|
||||
unsigned int sysctl_sched_cfs_bandwidth_slice = 4000UL;
|
||||
#endif
|
||||
|
||||
static inline void update_load_add(struct load_weight *lw, unsigned long inc)
|
||||
|
|
Loading…
Reference in a new issue