Home
last modified time | relevance | path

Searched refs:util_avg (Results 1 - 4 of 4) sorted by relevance

/device/soc/rockchip/common/sdk_linux/kernel/sched/
H A Dfair.c814 * based on the cfs_rq's current util_avg:
816 * util_avg = cfs_rq->util_avg / (cfs_rq->load_avg + 1) * se.load.weight
818 * However, in many cases, the above util_avg does not give a desired
822 * To solve this problem, we also cap the util_avg of successive tasks to
825 * util_avg_cap = (cpu_scale - cfs_rq->avg.util_avg) / 2^n
832 * task util_avg: 512, 256, 128, 64, 32, 16, 8, ...
833 * cfs_rq util_avg: 512, 768, 896, 960, 992, 1008, 1016, ...
835 * Finally, that extrapolated util_avg is clamped to the cap (util_avg_cap)
836 * if util_avg > util_avg_ca
[all...]
H A Dpelt.c308 WRITE_ONCE(sa->util_avg, sa->util_sum / divider); in ___update_load_avg()
422 * util_avg and runnable_load_avg are not supported and meaningless.
425 * running a rt/dl task through util_avg, the average thermal pressure is
427 * time weighted "delta" capacity unlike util_avg which is binary.
H A Dsched.h629 unsigned long util_avg; member
2658 return READ_ONCE(rq->avg_dl.util_avg); in cpu_util_dl()
2663 unsigned long util = READ_ONCE(rq->cfs.avg.util_avg); in cpu_util_cfs()
2674 return READ_ONCE(rq->avg_rt.util_avg); in cpu_util_rt()
2687 return rq->avg_irq.util_avg; in cpu_util_irq()
/device/soc/rockchip/common/sdk_linux/include/linux/
H A Dsched.h360 * has a few: load, load_avg, util_avg, freq, and capacity.
390 * - task: the task's util_avg at last task dequeue time
399 * The UTIL_AVG_UNCHANGED flag is used to synchronize util_est with util_avg
401 * util_avg has not been updated in the meantime.
403 * time. Since max value of util_est.enqueued for a task is 1024 (PELT util_avg
414 * The load/runnable/util_avg accumulates an infinite geometric series
425 * [util_avg definition]
427 * util_avg = running% * SCHED_CAPACITY_SCALE
435 * The load/runnable/util_avg doesn't directly factor frequency scaling and CPU
442 * util_avg'
466 unsigned long util_avg; global() member
[all...]

Completed in 16 milliseconds