TNL::Benchmarks::BenchmarkResult Struct Reference

Container for benchmark measurement results. More...

#include <TNL/Benchmarks/BenchmarkResult.h>

Collaboration diagram for TNL::Benchmarks::BenchmarkResult:

Public Types
using	HeaderElements = Logging::HeaderElements
using	RowElements = Logging::RowElements

Public Member Functions
virtual RowElements	getRowElements () const
	Returns formatted row data for logging.
virtual HeaderElements	getTableHeader () const
	Returns table header column names.
virtual void	setDerivedResults (double datasetSize, double baseTime, std::size_t operationsPerLoop)
	Computes derived metrics from raw timing data.
virtual void	setTimeResults (std::size_t loops_, double time_, double time_median_, double time_stddev_, double cpu_cycles_, double cpu_cycles_median_, double cpu_cycles_stddev_)
	Stores raw timing results.

Public Attributes
double	bandwidth = std::numeric_limits< double >::quiet_NaN()
double	cpu_cycles = std::numeric_limits< double >::quiet_NaN()
double	cpu_cycles_median = std::numeric_limits< double >::quiet_NaN()
double	cpu_cycles_per_operation = 0
double	cpu_cycles_stddev = std::numeric_limits< double >::quiet_NaN()
std::size_t	loops = 0
std::size_t	operations_per_loop = 0
double	speedup = std::numeric_limits< double >::quiet_NaN()
double	time = std::numeric_limits< double >::quiet_NaN()
double	time_median = std::numeric_limits< double >::quiet_NaN()
double	time_stddev = std::numeric_limits< double >::quiet_NaN()

Detailed Description

Container for benchmark measurement results.

Stores timing data and derived metrics from benchmark runs. The class supports virtual methods to allow derived classes to customize result storage and formatting.

Measured values:

• loops: Number of iterations executed
• time: Mean execution time per iteration
• time_median: Median execution time per iteration
• time_stddev: Standard deviation of execution time
• cpu_cycles: Mean CPU cycles per iteration (host devices only)
• cpu_cycles_median: Median CPU cycles per iteration (host devices only)
• cpu_cycles_stddev: Standard deviation of CPU cycles

Derived values (computed by setDerivedResults):

• bandwidth: Dataset size divided by execution time (GB/s)
• speedup: Baseline time divided by execution time
• cpu_cycles_per_operation: CPU cycles divided by operations per loop

Member Function Documentation

getRowElements()

virtual RowElements TNL::Benchmarks::BenchmarkResult::getRowElements ( ) const

inlinenodiscardvirtual

Returns formatted row data for logging.

Formats all result values as strings using scientific notation for floating-point values. Missing or invalid values are shown as "N/A".

Returns

LoggingRowElements containing formatted string values

getTableHeader()

virtual HeaderElements TNL::Benchmarks::BenchmarkResult::getTableHeader ( ) const

inlinenodiscardvirtual

Returns table header column names.

Returns

Vector of column names in display order

setDerivedResults()

virtual void TNL::Benchmarks::BenchmarkResult::setDerivedResults ( double datasetSize, double baseTime, std::size_t operationsPerLoop )

inlinevirtual

Computes derived metrics from raw timing data.

Must be called after setTimeResults. Calculates bandwidth, speedup, and cycles per operation based on dataset size, baseline time, and operation count.

Parameters

datasetSize	Dataset size in gigabytes
baseTime	Baseline time for speedup calculation
operationsPerLoop	Number of operations performed per iteration

setTimeResults()

virtual void TNL::Benchmarks::BenchmarkResult::setTimeResults ( std::size_t loops_, double time_, double time_median_, double time_stddev_, double cpu_cycles_, double cpu_cycles_median_, double cpu_cycles_stddev_ )

inlinevirtual

Stores raw timing results.

Called by timeFunction after completing all iterations. Derived classes can override to customize how timing data is stored.

Parameters

loops_	Number of iterations executed
time_	Mean execution time per iteration
time_median_	Median execution time per iteration
time_stddev_	Standard deviation of execution time
cpu_cycles_	Mean CPU cycles per iteration
cpu_cycles_median_	Median CPU cycles per iteration
cpu_cycles_stddev_	Standard deviation of CPU cycles

---

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

• src/TNL/Benchmarks/BenchmarkResult.h