class MWStatistics

This class serves as a holder for worker classes before they go to die

Public Methods

void gather( void* )

Literally, take the stats from this worker and store them in the stats class

void get_stats(double *average_bench, double *equivalent_bench, double *min_bench, double *max_bench, double *av_present_workers, double *av_nonsusp_workers, double *av_active_workers, double *equi_pool_performance, double *equi_run_time, double *parallel_performance, double *wall_time, MWList *workerlist )

This function is just like makestats(), except it return the relative statistics information to the caller

void makestats()

This method is called at the end of a run

MWStatistics()

Default constructor

void read_checkpoint( FILE *cfp )

Read in stats information across the checkpoint

void update_best_bench_results( double bres )

Everytime we get a new bench factor, we update the max and min ones if necessary

void write_checkpoint( FILE *cfp, MWList *wlist )

We find that we should save stats across checkpoints

~MWStatistics()

Destructor...

Private Fields

double* cputimes

Array of times holding the CPU times of the workerIDs

double* susptimes

Array of times that the workerIDs have been suspended

double* uptimes

Array of times that the workerIDs have been "up"

double* workingtimes

Array of times that the workerIDs have been working

Private Methods

int get_max_vid()

What is the maximum "virtuid ID" that we ever assigned to a worker

double mean( double *array, int length )

Computes the mean of an array of number.

double var( double *array, int length, int mean )

Computes the variance of an array of numbers (given the mean, to avoid "double processing" the list -- which you don't need to do anyway, but Mike wrote this function)

Private

Normalized Statistics. MW computes "normalized" statistics

double max_bench_result

Min benchmark result -- the "slowest" machine we ever had

double min_bench_result

Max benchmark result -- the "faster" machine we ever had

double* sumbenchmark

The sum of benchmark factors for this vid

int* numbenchmark

The number of benchmar factors for this vid

MWStatistics()

Default constructor

~MWStatistics()

Destructor...

void gather( void* )

Literally, take the stats from this worker and store them in the stats class. This is done when the worker dies and during a checkpoint.

void makestats()

This method is called at the end of a run. It prints out some statistical information on that run.

void get_stats(double *average_bench, double *equivalent_bench, double *min_bench, double *max_bench, double *av_present_workers, double *av_nonsusp_workers, double *av_active_workers, double *equi_pool_performance, double *equi_run_time, double *parallel_performance, double *wall_time, MWList *workerlist )

This function is just like makestats(), except it return the relative statistics information to the caller

void update_best_bench_results( double bres )

Everytime we get a new bench factor, we update the max and min ones if necessary

void write_checkpoint( FILE *cfp, MWList *wlist )

We find that we should save stats across checkpoints. Here we're given the ckpt FILE* and the list of running workers. Basically, we write out the sum of the old stats we've got plus the stats of the running workers.

void read_checkpoint( FILE *cfp )

Read in stats information across the checkpoint

double* uptimes

Array of times that the workerIDs have been "up"

double* workingtimes

Array of times that the workerIDs have been working

double* cputimes

Array of times holding the CPU times of the workerIDs

double* susptimes

Array of times that the workerIDs have been suspended

Normalized Statistics. MW computes "normalized" statistics

about the runs performance based on the relative speed of the machines that the user obtained during the course of the run.

double max_bench_result

Min benchmark result -- the "slowest" machine we ever had

double min_bench_result

Max benchmark result -- the "faster" machine we ever had

double* sumbenchmark

The sum of benchmark factors for this vid

int* numbenchmark

The number of benchmar factors for this vid

double mean( double *array, int length )

Computes the mean of an array of number.

double var( double *array, int length, int mean )

Computes the variance of an array of numbers (given the mean, to avoid "double processing" the list -- which you don't need to do anyway, but Mike wrote this function)

int get_max_vid()

What is the maximum "virtuid ID" that we ever assigned to a worker. This amounts to asking how many workers did we ever have in ALL states at any one time during the course of the run.

alphabetic index hierarchy of classes

this page has been generated automatically by doc++

(c)opyright by Malte Zöckler, Roland Wunderling
contact: doc++@zib.de