Re: [RFC 1/3] sched/fair: add util_est on top of PELT

[Pavan Kondeti]

On Fri, Aug 25, 2017 at 3:50 PM, Patrick Bellasi
<patrick.bellasi@xxxxxxx> wrote:
> The util_avg signal computed by PELT is too variable for some use-cases.
> For example, a big task waking up after a long sleep period will have its
> utilization almost completely decayed. This introduces some latency before
> schedutil will be able to pick the best frequency to run a task.
>

<snip>

> diff --git a/include/linux/sched.h b/include/linux/sched.h
> index c28b182c9833..8d7bc55f68d5 100644
> --- a/include/linux/sched.h
> +++ b/include/linux/sched.h
> @@ -26,6 +26,7 @@
>  #include <linux/signal_types.h>
>  #include <linux/mm_types_task.h>
>  #include <linux/task_io_accounting.h>
> +#include <linux/average.h>
>
>  /* task_struct member predeclarations (sorted alphabetically): */
>  struct audit_context;
> @@ -277,6 +278,16 @@ struct load_weight {
>         u32                             inv_weight;
>  };
>
> +/**
> + * Utilizaton's Exponential Weighted Moving Average (EWMA)
> + *
> + * Support functions to track an EWMA for the utilization of SEs and RQs. New
> + * samples will be added to the moving average each time a task completes an
> + * activation. Thus the weight is chosen so that the EWMA wil be relatively
> + * insensitive to transient changes to the task's workload.
> + */
> +DECLARE_EWMA(util, 0, 4);
> +
>  /*

Should the factor be 1 instead of 0? i.e 25% contribution from the
recent sample.

Thanks,
Pavan


-- 
Qualcomm India Private Limited, on behalf of Qualcomm Innovation Center, Inc.
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a
Linux Foundation Collaborative Project