Templatized storage and interface for a histogram stat. More...

#include <storage.hh>

Classes
struct	Params
	The parameters for a distribution stat. More...

Public Member Functions
	HistStor (Info *info)

void	add (HistStor *other)
	Adds the contents of the given storage to this storage. More...

void	sample (Counter val, int number)
	Add a value to the distribution for the given number of times. More...

size_type	size () const
	Return the number of buckets in this distribution. More...

bool	zero () const
	Returns true if any calls to sample have been made. More...

void	prepare (Info *info, DistData &data)

void	reset (Info *info)
	Reset stat value to default. More...

Private Member Functions
void	growUp ()
	Given a bucket size B, and a range of values [0, N], this function doubles the bucket size to double the range of values towards the positive infinite; that is, double the upper range of this storage so that the range becomes [0, 2*N]. More...

void	growOut ()
	Given a bucket size B, and a range of values [M, N], where M < 0, this function doubles the bucket size to double the range of values towards both positive and negative infinites; that is, it doubles both the lower and the upper range of this storage so that the range becomes [2M, 2N]. More...

void	growDown ()
	Given a bucket size B, and a range of values [0, N], this function doubles the bucket size to double the range of values towards the negative infinite; that is, it doubles the lower range of this storage so that the middle buckes contaihs zero as a lower bound. More...

Private Attributes
Counter	min_bucket
	Lower bound of the first bucket's range. More...

Counter	max_bucket
	Lower bound of the last bucket's range. More...

Counter	bucket_size
	The number of entries in each bucket. More...

Counter	sum
	The current sum. More...

Counter	logs
	The sum of logarithm of each sample, used to compute geometric mean. More...

Counter	squares
	The sum of squares. More...

Counter	samples
	The number of samples. More...

VCounter	cvec
	Counter for each bucket. More...

Detailed Description

Templatized storage and interface for a histogram stat.

The number of buckets is fixed on initialization; however, the bucket size isn't. That means that when samples that are outside the current range are seen, the bucket size will be increased so that each bucket can hold a bigger range of values. When that happens, the bucket's contents are re- located.

The min and max bucket values can only be, respectively, decreased and increased when sampling. If this wasn't true, samples that were previously within the buclet range could not be anymore within the valid range, making the storage's state incoherent. These values are set back to their initial states on reset().

The bucket range always is zero-centric. While the storage does not contain negative values the bucket range will keep its lower bound at zero, doubling the upper bound when needed; However, as soon a negative value is sampled, zero becomes the lower bound of the middle (rounded up) bucket. Although this means that the histogram will not be symmetric if negative values are sampled, it makes it possible to grow the buckets without keeping track of the individual values.

This happens because if zero was not a lower or upper bound, when its value was doubled, the lower and upper bound of the bucket containing zero would intersect with middle values of the previous and next buckets. For example, if the bucket containing zero has range [-2,2[, therefore its neighbor buckets would have ranges at [-6,-2[ and [2,6[. When the buckets are grown, the zero bucket would grow its range to [-4,4[, which cannot be easily extracted from the neighor buckets.

Definition at line 395 of file storage.hh.

Constructor & Destructor Documentation

◆ HistStor()

Stats::HistStor::HistStor ( Info * info )

inline

Definition at line 470 of file storage.hh.

References reset().

Member Function Documentation

◆ add()

void Stats::HistStor::add ( HistStor * other )

Adds the contents of the given storage to this storage.

Parameters

other The other storage to be added.

Definition at line 205 of file storage.cc.

References bucket_size, cvec, growUp(), ArmISA::i, logs, min_bucket, samples, size(), squares, and sum.

Referenced by TEST().

◆ growDown()

void Stats::HistStor::growDown ( )

private

Given a bucket size B, and a range of values [0, N], this function doubles the bucket size to double the range of values towards the negative infinite; that is, it doubles the lower range of this storage so that the middle buckes contaihs zero as a lower bound.

As such, the storage range becomes [-N, N+B] if there is an odd number of buckets, and [-N-B, N+B] if there is an even number of buckets.

Because the bucket size is doubled, the buckets contents are rearranged, and the original range of values are redistributed to free buckets for the newly appended ranges.

Definition at line 111 of file storage.cc.

References bucket_size, cvec, ArmISA::i, max_bucket, min_bucket, size(), and zero().

Referenced by sample().

◆ growOut()

void Stats::HistStor::growOut ( )

private

Given a bucket size B, and a range of values [M, N], where M < 0, this function doubles the bucket size to double the range of values towards both positive and negative infinites; that is, it doubles both the lower and the upper range of this storage so that the range becomes [2*M, 2*N].

Because the bucket size is doubled, the buckets contents are rearranged, and the original range of values are redistributed to free buckets for the newly appended ranges.

Definition at line 72 of file storage.cc.

References bucket_size, cvec, ArmISA::i, max_bucket, min_bucket, size(), and zero().

Referenced by sample().

◆ growUp()

void Stats::HistStor::growUp ( )

private

Given a bucket size B, and a range of values [0, N], this function doubles the bucket size to double the range of values towards the positive infinite; that is, double the upper range of this storage so that the range becomes [0, 2*N].

Because the bucket size is doubled, the buckets contents are rearranged, since the original range of values is mapped to the lower half buckets.

Definition at line 150 of file storage.cc.

References bucket_size, cvec, ArmISA::i, max_bucket, and size().

Referenced by add(), and sample().

◆ prepare()

void Stats::HistStor::prepare	(	Info *	info,
		DistData &	data
	)

inline

Definition at line 506 of file storage.hh.

References bucket_size, Stats::HistStor::Params::buckets, cvec, data, Stats::Hist, ArmISA::i, logs, max_bucket, min_bucket, samples, squares, Stats::Info::storageParams, sum, and Stats::DistParams::type.

Referenced by prepareCheckHistStor(), and TEST().

◆ reset()

void Stats::HistStor::reset ( Info * info )

inline

Reset stat value to default.

Definition at line 534 of file storage.hh.

References bucket_size, Stats::HistStor::Params::buckets, cvec, ArmISA::i, logs, max_bucket, min_bucket, samples, size(), squares, Stats::Info::storageParams, and sum.

Referenced by HistStor(), and TEST().

◆ sample()

void Stats::HistStor::sample	(	Counter	val,
		int	number
	)

Add a value to the distribution for the given number of times.

Parameters

val	The value to add.
number	The number of times to add the value.

Definition at line 172 of file storage.cc.

References bucket_size, cvec, growDown(), growOut(), growUp(), MipsISA::index, logs, max_bucket, min_bucket, samples, size(), squares, sum, and X86ISA::val.

Referenced by prepareCheckHistStor(), and TEST().

◆ size()

size_type Stats::HistStor::size ( ) const

inline

Return the number of buckets in this distribution.

Returns: the number of buckets.

Definition at line 493 of file storage.hh.

References cvec.

Referenced by add(), growDown(), growOut(), growUp(), reset(), sample(), and TEST().

◆ zero()

bool Stats::HistStor::zero ( ) const

inline

Returns true if any calls to sample have been made.

Returns: True if any values have been sampled.

Definition at line 500 of file storage.hh.

References samples.

Referenced by growDown(), growOut(), and TEST().

Member Data Documentation

◆ bucket_size

Counter Stats::HistStor::bucket_size

private

The number of entries in each bucket.

Definition at line 403 of file storage.hh.

Referenced by add(), growDown(), growOut(), growUp(), prepare(), reset(), and sample().

◆ cvec

VCounter Stats::HistStor::cvec

private

Counter for each bucket.

Definition at line 414 of file storage.hh.

Referenced by add(), growDown(), growOut(), growUp(), prepare(), reset(), sample(), and size().

◆ logs

Counter Stats::HistStor::logs

private

The sum of logarithm of each sample, used to compute geometric mean.

Definition at line 408 of file storage.hh.

Referenced by add(), prepare(), reset(), and sample().

◆ max_bucket

Counter Stats::HistStor::max_bucket

private

Lower bound of the last bucket's range.

Definition at line 401 of file storage.hh.

Referenced by growDown(), growOut(), growUp(), prepare(), reset(), and sample().

◆ min_bucket

Counter Stats::HistStor::min_bucket

private

Lower bound of the first bucket's range.

Definition at line 399 of file storage.hh.

Referenced by add(), growDown(), growOut(), prepare(), reset(), and sample().

◆ samples

Counter Stats::HistStor::samples

private

The number of samples.

Definition at line 412 of file storage.hh.

Referenced by add(), prepare(), reset(), sample(), and zero().

◆ squares

Counter Stats::HistStor::squares

private

The sum of squares.

Definition at line 410 of file storage.hh.

Referenced by add(), prepare(), reset(), and sample().

◆ sum

Counter Stats::HistStor::sum

private

The current sum.

Definition at line 406 of file storage.hh.

Referenced by add(), prepare(), reset(), and sample().

The documentation for this class was generated from the following files:

base/stats/storage.hh
base/stats/storage.cc

Classes

Public Member Functions

Private Member Functions

Private Attributes

Detailed Description

Constructor & Destructor Documentation

◆ HistStor()

Member Function Documentation

◆ add()

◆ growDown()

◆ growOut()

◆ growUp()

◆ prepare()

◆ reset()

◆ sample()

◆ size()

◆ zero()

Member Data Documentation

◆ bucket_size

◆ cvec

◆ logs

◆ max_bucket

◆ min_bucket

◆ samples

◆ squares

◆ sum