[ VIGRA Homepage | Function Index | Class Index | Namespaces | File List | Main Page ]

details ParallelOptions Class Reference VIGRA

Option base class for parallel algorithms. More...

#include <vigra/threadpool.hxx>

Inheritance diagram for ParallelOptions:
BlockwiseOptions BlockwiseConvolutionOptions< N > BlockwiseLabelOptions

Public Types

enum  { Auto = -1, Nice = -2, NoThreads = 0 }
 

Public Member Functions

int getActualNumThreads () const
 Get desired number of threads. More...
 
int getNumThreads () const
 Get desired number of threads. More...
 
ParallelOptions & numThreads (const int n)
 Set the number of threads or one of the constants Auto, Nice and NoThreads. More...
 

Detailed Description

Option base class for parallel algorithms.

#include <vigra/threadpool.hxx>
Namespace: vigra

Member Enumeration Documentation

anonymous enum

Constants for special settings.

Enumerator
Auto 

Determine number of threads automatically (from threading::thread::hardware_concurrency())

Nice 

Use half as many threads as Auto would.

NoThreads 

Switch off multi-threading (i.e. execute tasks sequentially)

Member Function Documentation

int getNumThreads ( ) const

Get desired number of threads.

Note: This function may return 0, which means that multi-threading shall be switched off entirely. If an algorithm receives this value, it should revert to a sequential implementation. In contrast, if numThread() == 1, the parallel algorithm version shall be executed with a single thread.

int getActualNumThreads ( ) const

Get desired number of threads.

In contrast to numThread(), this will always return a value >=1.

ParallelOptions& numThreads ( const int  n)

Set the number of threads or one of the constants Auto, Nice and NoThreads.

Default: ParallelOptions::Auto (use system default)

This setting is ignored if the preprocessor flag VIGRA_SINGLE_THREADED is defined. Then, the number of threads is set to 0 and all tasks revert to sequential algorithm implementations. The same can be achieved at runtime by passing n = 0 to this function. In contrast, passing n = 1 causes the parallel algorithm versions to be executed with a single thread. Both possibilities are mainly useful for debugging.


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

© Ullrich Köthe (ullrich.koethe@iwr.uni-heidelberg.de)
Heidelberg Collaboratory for Image Processing, University of Heidelberg, Germany

html generated using doxygen and Python
vigra 1.11.1 (Fri May 19 2017)