Re: [PATCH V4 08/14] perf/x86/intel: Support per thread RDPMC TopDown metrics

[Peter Zijlstra]

On Mon, Sep 16, 2019 at 06:41:22AM -0700, kan.liang@xxxxxxxxxxxxxxx wrote:
> diff --git a/arch/x86/events/intel/core.c b/arch/x86/events/intel/core.c
> index 71f3086a8adc..7ec0f350d2ac 100644
> --- a/arch/x86/events/intel/core.c
> +++ b/arch/x86/events/intel/core.c
> @@ -2262,6 +2262,11 @@ static int icl_set_topdown_event_period(struct perf_event *event)
>  		local64_set(&hwc->period_left, 0);
>  	}
>  
> +	if ((hwc->saved_slots) && is_first_topdown_event_in_group(event)) {
> +		wrmsrl(MSR_CORE_PERF_FIXED_CTR3, hwc->saved_slots);
> +		wrmsrl(MSR_PERF_METRICS, hwc->saved_metric);
> +	}

> diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h
> index 61448c19a132..c125068f2e16 100644
> --- a/include/linux/perf_event.h
> +++ b/include/linux/perf_event.h
> @@ -133,6 +133,9 @@ struct hw_perf_event {
>  
>  			struct hw_perf_event_extra extra_reg;
>  			struct hw_perf_event_extra branch_reg;
> +
> +			u64		saved_slots;
> +			u64		saved_metric;
>  		};
>  		struct { /* software */
>  			struct hrtimer	hrtimer;

Normal counters save their counter value in hwc->period_left, why does
slots need a new word for that?

And since using METRIC means non-sampling, why can't we stick that
saved_metric field in one of the unused sampling fields?

ISTR asking this before...