ktt::PreciseMeasurementParameters Struct Reference

#include <PreciseMeasurementParameters.h>

Public Member Functions
	PreciseMeasurementParameters ()
	PreciseMeasurementParameters (uint64_t minTimeMs, uint64_t maxTimeMs, double maxPowerDiff, DurationCalculationMethod durationCalculationMethod=DurationCalculationMethod::Minimum)
bool	IsValid () const

Static Public Member Functions
static Nanoseconds	CalculateDuration (const std::vector< Nanoseconds > &samples, DurationCalculationMethod method)
static double	CalculateStandardDeviation (const std::vector< Nanoseconds > &samples)
static DurationMeasurementResult	ComputeDurationAndStdev (const std::vector< Nanoseconds > &samples, DurationCalculationMethod method)

Public Attributes
uint64_t	minTimeMs
uint64_t	maxTimeMs
double	maxPowerDiff
DurationCalculationMethod	durationCalculationMethod

Detailed Description

Struct which holds parameters for precise measurement during kernel execution. When power measurement is enabled (KTT built with –power-usage), the kernel is executed multiple times until power readings stabilize within the specified tolerance. When power measurement is not available, the kernel is still executed multiple times for the specified minimum time to compute a more stable duration according to the selected calculation method.

Constructor & Destructor Documentation

PreciseMeasurementParameters() [1/2]

ktt::PreciseMeasurementParameters::PreciseMeasurementParameters

(

)

Default constructor with conservative default values. Defaults: minTime=2000ms, maxTime=20000ms (10x minTime), maxPowerDiff=0.005 (0.5%), durationCalculationMethod=Minimum

PreciseMeasurementParameters() [2/2]

ktt::PreciseMeasurementParameters::PreciseMeasurementParameters	(	uint64_t	minTimeMs,
		uint64_t	maxTimeMs,
		double	maxPowerDiff,
		DurationCalculationMethod	durationCalculationMethod = DurationCalculationMethod::Minimum
	)

Constructor with custom values.

Parameters

minTimeMs	Minimum measurement time in milliseconds.
maxTimeMs	Maximum measurement time in milliseconds.
maxPowerDiff	Maximum tolerated relative power difference (0.0 - 1.0).
durationCalculationMethod	Method to calculate duration from samples (default: Minimum).

Member Function Documentation

CalculateDuration()

static Nanoseconds ktt::PreciseMeasurementParameters::CalculateDuration ( const std::vector< Nanoseconds > &  samples, DurationCalculationMethod  method  )

static

Calculates kernel duration from multiple samples based on the selected method.

Parameters

samples	Vector of duration samples in nanoseconds.
method	Method to use for duration calculation.

Returns

Calculated duration in nanoseconds.

CalculateStandardDeviation()

static double ktt::PreciseMeasurementParameters::CalculateStandardDeviation ( const std::vector< Nanoseconds > &  samples )

static

Calculates standard deviation of duration samples.

Parameters

samples

Vector of duration samples in nanoseconds.

Returns

Standard deviation of the samples.

ComputeDurationAndStdev()

static DurationMeasurementResult ktt::PreciseMeasurementParameters::ComputeDurationAndStdev ( const std::vector< Nanoseconds > &  samples, DurationCalculationMethod  method  )

static

Calculates both duration and standard deviation from samples.

Parameters

samples	Vector of duration samples in nanoseconds.
method	Method to use for duration calculation.

Returns

Result containing duration, standard deviation, and sample count.

IsValid()

bool ktt::PreciseMeasurementParameters::IsValid

(

)

const

Checks if the parameters are valid.

Returns

True if parameters are valid (minTime > 0, maxTime >= minTime, 0 < maxPowerDiff <= 1.0).

Member Data Documentation

durationCalculationMethod

DurationCalculationMethod ktt::PreciseMeasurementParameters::durationCalculationMethod

Method used to calculate kernel duration from multiple samples. Defaults to Minimum for backward compatibility.

maxPowerDiff

double ktt::PreciseMeasurementParameters::maxPowerDiff

Maximum tolerated relative difference in power readings (0.0 - 1.0). When all recent samples are within this relative difference from the average, measurement is considered stable. Only used when power measurement is enabled (KTT built with –power-usage). When power measurement is not available, this parameter is ignored.

maxTimeMs

uint64_t ktt::PreciseMeasurementParameters::maxTimeMs

Maximum measurement time in milliseconds. When power measurement is enabled, measurement stops when this time is exceeded, even if stability has not been reached. When power measurement is not available, this parameter is ignored (see maxPowerDiff).

minTimeMs

uint64_t ktt::PreciseMeasurementParameters::minTimeMs

Minimum measurement time in milliseconds. Measurement continues until at least this much time has elapsed since the first measurement.

---

The documentation for this struct was generated from the following file:

• Source/Api/Configuration/PreciseMeasurementParameters.h