MTToolBox::AlgorithmPartialBitPattern< U, V, bit_len, param_num, try_bit_len, step, lsb > Class Template Reference

Algorithm that search tempering parameters to improve dimension of equi-distribution of output of pseudo random number generator. More...

#include <AlgorithmPartialBitPattern.hpp>

Inheritance diagram for MTToolBox::AlgorithmPartialBitPattern< U, V, bit_len, param_num, try_bit_len, step, lsb >:

Public Member Functions
int	operator() (TemperingCalculatable< U, V > &rand, bool verbose=false)
	Search tempering parameters. More...
bool	isLSBTempering () const
	Shows if searching tempering parameters is from LSBs. More...
Public Member Functions inherited from MTToolBox::AlgorithmTempering< U, V >
virtual	~AlgorithmTempering ()
	Virtual destructor (always required) More...

Detailed Description

template<typename U, typename V, int bit_len, int param_num, int try_bit_len, int step = 5, bool lsb = false>
class MTToolBox::AlgorithmPartialBitPattern< U, V, bit_len, param_num, try_bit_len, step, lsb >

Algorithm that search tempering parameters to improve dimension of equi-distribution of output of pseudo random number generator.

This algorithm was developed when we design Mersenne Twister for Graphic Processors (MTGP). This algorithm also is used in Tiny Mersenne Twister (TinyMT).

Warning

If you could not get high dimension of equi-distribution after tempering using parameters got by this algorithm, you should consider changing the design of random number generation algorithm.

Template Parameters

U	type of output of pseudo random number generator, for example, uint32_t. Only unsigned numbers are allowed.
V	type of output of parameter generator.
bit_len	bit length of variable for tempering parameter
param_num	number of tempering parameters, for instance, four for MTGP and one for TinyMT.
try_bit_len	length of tempering part of output.
step	specify how many bits are generated as bit pattern. If large number is specified, it will consume huge CPU time. For example, five is specified in MTGP parameter search.
lsb	When lsb is true, meaning of MSBs and LSBs are changed. This program calculate equi-distribution from LSB not MSB.

Member Function Documentation

template<typename U , typename V , int bit_len, int param_num, int try_bit_len, int step = 5, bool lsb = false>

bool MTToolBox::AlgorithmPartialBitPattern< U, V, bit_len, param_num, try_bit_len, step, lsb >::isLSBTempering ( ) const

inlinevirtual

Shows if searching tempering parameters is from LSBs.

Returns

true if searching is from LSBs.

Reimplemented from MTToolBox::AlgorithmTempering< U, V >.

template<typename U , typename V , int bit_len, int param_num, int try_bit_len, int step = 5, bool lsb = false>

int MTToolBox::AlgorithmPartialBitPattern< U, V, bit_len, param_num, try_bit_len, step, lsb >::operator() ( TemperingCalculatable< U, V > &  rand, bool  verbose = false  )

inlinevirtual

Search tempering parameters.

Searched parameters are set to rand. This process may consume large CPU time.

Parameters

rand	pseudo random number generator
verbose	if true output redundant messages.

Returns

always zero.

Implements MTToolBox::AlgorithmTempering< U, V >.

References MTToolBox::TemperingCalculatable< U, V >::resetReverseOutput(), and MTToolBox::TemperingCalculatable< U, V >::setReverseOutput().

---

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

• AlgorithmPartialBitPattern.hpp