Re: [PATCH V2] perf/x86/intel: New start period for the freq mode

["Liang, Kan" <kan.liang@linux.intel.com>]

Hi Peter,

Ping. Could you please let me know if you have any comments.

Thanks,
Kan

On 2024-10-22 9:04 a.m., kan.liang@linux.intel.com wrote:
> From: Kan Liang <kan.liang@linux.intel.com>
> 
> The freq mode is the current default mode of Linux perf. 1 period is
> used as a start period. The period is auto-adjusted in each tick or an
> overflow to meet the frequency target.
> 
> The start period 1 is too low and may trigger some issues.
> - Many HWs do not support period 1 well.
>   https://lore.kernel.org/lkml/875xs2oh69.ffs@tglx/
> - For an event that occurs frequently, period 1 is too far away from the
>   real period. Lots of the samples are generated at the beginning.
>   The distribution of samples may not be even.
> - A low start period for the frequently occurring event also challenges
>   virtualization, which has a longer path to handle a PMI.
> 
> The limit_period only checks the minimum acceptable value for HW.
> It cannot be used to set the start period. Because, some events may
> need a very low period. The limit_period cannot be set too high. It
> doesn't help with the events that occur frequently.
> 
> It's hard to find a universal start period for all events. The idea is
> only to give an estimate for the popular HW and HW cache events. For the
> rest of the events, start from the lowest possible recommended value.
> 
> Signed-off-by: Kan Liang <kan.liang@linux.intel.com>
> ---
> 
> Changes since V1:
> - Move to Intel specific code
> 
>  arch/x86/events/intel/core.c | 85 ++++++++++++++++++++++++++++++++++++
>  1 file changed, 85 insertions(+)
> 
> diff --git a/arch/x86/events/intel/core.c b/arch/x86/events/intel/core.c
> index 812b1a1cdf47..d3b133384361 100644
> --- a/arch/x86/events/intel/core.c
> +++ b/arch/x86/events/intel/core.c
> @@ -4076,6 +4076,85 @@ static void intel_pmu_set_acr_dyn_mask(struct perf_event *event, int idx,
>  		event->hw.dyn_mask &= mask;
>  }
>  
> +static u64 intel_pmu_freq_start_period(struct perf_event *event)
> +{
> +	int type = event->attr.type;
> +	u64 config, factor;
> +	s64 start;
> +
> +	/*
> +	 * The 127 is the lowest possible recommended SAV (sample after value)
> +	 * for a 4000 freq (default freq), according to the event list JSON file.
> +	 * Also, assume the workload is idle 50% time.
> +	 */
> +	factor = 64 * 4000;
> +	if (type != PERF_TYPE_HARDWARE && type != PERF_TYPE_HW_CACHE)
> +		goto end;
> +
> +	/*
> +	 * The estimation of the start period in the freq mode is
> +	 * based on the below assumption.
> +	 *
> +	 * For a cycles or an instructions event, 1GHZ of the
> +	 * underlying platform, 1 IPC. The workload is idle 50% time.
> +	 * The start period = 1,000,000,000 * 1 / freq / 2.
> +	 *		    = 500,000,000 / freq
> +	 *
> +	 * Usually, the branch-related events occur less than the
> +	 * instructions event. According to the Intel event list JSON
> +	 * file, the SAV (sample after value) of a branch-related event
> +	 * is usually 1/4 of an instruction event.
> +	 * The start period of branch-related events = 125,000,000 / freq.
> +	 *
> +	 * The cache-related events occurs even less. The SAV is usually
> +	 * 1/20 of an instruction event.
> +	 * The start period of cache-related events = 25,000,000 / freq.
> +	 */
> +	config = event->attr.config & PERF_HW_EVENT_MASK;
> +	if (type == PERF_TYPE_HARDWARE) {
> +		switch (config) {
> +		case PERF_COUNT_HW_CPU_CYCLES:
> +		case PERF_COUNT_HW_INSTRUCTIONS:
> +		case PERF_COUNT_HW_BUS_CYCLES:
> +		case PERF_COUNT_HW_STALLED_CYCLES_FRONTEND:
> +		case PERF_COUNT_HW_STALLED_CYCLES_BACKEND:
> +		case PERF_COUNT_HW_REF_CPU_CYCLES:
> +			factor = 500000000;
> +			break;
> +		case PERF_COUNT_HW_BRANCH_INSTRUCTIONS:
> +		case PERF_COUNT_HW_BRANCH_MISSES:
> +			factor = 125000000;
> +			break;
> +		case PERF_COUNT_HW_CACHE_REFERENCES:
> +		case PERF_COUNT_HW_CACHE_MISSES:
> +			factor = 25000000;
> +			break;
> +		default:
> +			goto end;
> +		}
> +	}
> +
> +	if (type == PERF_TYPE_HW_CACHE)
> +		factor = 25000000;
> +end:
> +	/*
> +	 * Usually, a prime or a number with less factors (close to prime)
> +	 * is chosen as an SAV, which makes it less likely that the sampling
> +	 * period synchronizes with some periodic event in the workload.
> +	 * Minus 1 to make it at least avoiding values near power of twos
> +	 * for the default freq.
> +	 */
> +	start = DIV_ROUND_UP_ULL(factor, event->attr.sample_freq) - 1;
> +
> +	if (start > x86_pmu.max_period)
> +		start = x86_pmu.max_period;
> +
> +	if (x86_pmu.limit_period)
> +		x86_pmu.limit_period(event, &start);
> +
> +	return start;
> +}
> +
>  static int intel_pmu_hw_config(struct perf_event *event)
>  {
>  	int ret = x86_pmu_hw_config(event);
> @@ -4087,6 +4166,12 @@ static int intel_pmu_hw_config(struct perf_event *event)
>  	if (ret)
>  		return ret;
>  
> +	if (event->attr.freq && event->attr.sample_freq) {
> +		event->hw.sample_period = intel_pmu_freq_start_period(event);
> +		event->hw.last_period = event->hw.sample_period;
> +		local64_set(&event->hw.period_left, event->hw.sample_period);
> +	}
> +
>  	if (event->attr.precise_ip) {
>  		if ((event->attr.config & INTEL_ARCH_EVENT_MASK) == INTEL_FIXED_VLBR_EVENT)
>  			return -EINVAL;