rpa::range_fwrd< It > Class Template Reference

For forward_tag iterators if single-thread access. This is an output range. More...

#include <common.h>

List of all members.

Public Types
typedef It	iterator
typedef It	result_t
Public Member Functions
	range_fwrd (const result_t &aBeg)
const result_t &	result (void) const
	Always returns the same data type as the one given to the constructor.
void	leave_seq (const iterator &anIter)
	This is used to set, what will be the return value of the parallel algorithm.
void	leave_par (const iterator &anIter)
iterator	begout (void) const
slice	chop (size_t aN)
void	chop (size_t aN, slice &aSR)
	Naive implementation for reusing objects, because we assume they're cheap.
slice	tail (void)
size_t	ceiling (void) const
	Limits the maximum size of a 'chop' if the output is a fixed-size buffer.
Protected Attributes
It	_It
seqnum_t	_NbChops
Private Member Functions
	range_fwrd ()
Classes
class	slice

---

Detailed Description

template<class It>
class rpa::range_fwrd< It >

For forward_tag iterators if single-thread access. This is an output range.

What ranges are :

• Input ranges store an pair of input iterators (begin and end).
• Output ranges store the output iterator where the result must be written.
• Ranges store type information.
• Ranges store mutexes to synchronize threads accessing the same iterator.
• Ranges store sizing information too, such as the number of elements if it is known, the minimal number of elements to process in one-thread.
• Used as type discriminator to choose an algorithm at compile-time.

---

Member Typedef Documentation

template<class It>

typedef It rpa::range_fwrd< It >::iterator

template<class It>

typedef It rpa::range_fwrd< It >::result_t

---

Constructor & Destructor Documentation

template<class It>

rpa::range_fwrd< It >::range_fwrd

(

)

[private]

template<class It>

rpa::range_fwrd< It >::range_fwrd

(

const result_t &

aBeg

)

[inline, explicit]

The input parameter is an iterator that may, for example, given as output iterator to a STL algorithm.

---

Member Function Documentation

template<class It>

iterator rpa::range_fwrd< It >::begout

(

void

)

const [inline]

This is passed to an algorithm in the main thread, when executing in sequential context (For exemple, when there are no sub-threads).

template<class It>

size_t rpa::range_fwrd< It >::ceiling

(

void

)

const [inline]

Limits the maximum size of a 'chop' if the output is a fixed-size buffer.

In all other cases including this one, simply returns the step range_step.

template<class It>

void rpa::range_fwrd< It >::chop	(	size_t	aN,
		slice &	aSR
	)			[inline]

Naive implementation for reusing objects, because we assume they're cheap.

Needed only for 'heavy' slices (buffers). Must call 'chop(int)' first.

template<class It>

slice rpa::range_fwrd< It >::chop

(

size_t

aN

)

[inline]

Is called by '<input-sequence>.chop()' 'chop' may be called with aN=0 elements, because :

• In 'step', we cannot test if a sequence is empty before calling 'chop', for race-conditions reasons : We would otherwise have to lock the sequence too long, and not only during the 'chop'.
• The main-thread must make a 'chop' for all sub-threads, in order to have a valid sequence, even if for the last time. Could be optimised ? It is necessary to have a centralised incrementation. The individual algorithms do the same in several sub-threads, but the point is that it must be made at a single place. Otherwise, we should wait for the end of a thread, before starting a second one. NOTE: This must NOT be called when the output is a fixed-size buffer, because in such a case the output sequence must not change, as several loops of the same thread write at the same place.

template<class It>

void rpa::range_fwrd< It >::leave_par

(

const iterator &

anIter

)

[inline]

template<class It>

void rpa::range_fwrd< It >::leave_seq

(

const iterator &

anIter

)

[inline]

This is used to set, what will be the return value of the parallel algorithm.

Because the return value is homogeneous to the output iterator, this return value can be used for later insertions in other algorithms.

template<class It>

const result_t& rpa::range_fwrd< It >::result

(

void

)

const [inline]

Always returns the same data type as the one given to the constructor.

This way, if the range was initialised with a back_insert_iterator, we still get a back_insert_iterator as return type.

template<class It>

slice rpa::range_fwrd< It >::tail

(

void

)

[inline]

Is called by '<input-sequence>.tail()', when the input sequence is a range_size. Last call after zero or several 'chop' calls.

---

Member Data Documentation

template<class It>

It rpa::range_fwrd< It >::_It [protected]

Stores the output iterator when threads are allocated. It is read by several threads, and therefore must be volatile.

template<class It>

seqnum_t rpa::range_fwrd< It >::_NbChops [protected]

---

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

• common.h

---