Detailed Description

API below is provided to resolve problem of CPU resource over-subscription by multiple thread pools from different multi-threading frameworks. This is common problem for cases when OpenCV compiled threading framework is different from the Users Applications framework.

Applications can replace OpenCV parallel_for() backend with own implementation (to reuse Application's thread pool).

Backend API usage examples

Intel TBB

include header with simple implementation of TBB backend:
#include "opencv2/core/parallel/backend/parallel_for.tbb.hpp"
execute backend replacement code:
cv::parallel::setParallelForBackend(std::make_shared<cv::parallel::tbb::ParallelForBackend>());
configuration of compiler/linker options is responsibility of Application's scripts

OpenMP

include header with simple implementation of OpenMP backend:
#include "opencv2/core/parallel/backend/parallel_for.openmp.hpp"
execute backend replacement code:
//omp_set_dynamic(1);

cv::parallel::setParallelForBackend(std::make_shared<cv::parallel::openmp::ParallelForBackend>());
Configuration of compiler/linker options is responsibility of Application's scripts

Plugins support

Runtime configuration options:

change backend priority: OPENCV_PARALLEL_PRIORITY_<backend>=9999
disable backend: OPENCV_PARALLEL_PRIORITY_<backend>=0
specify list of backends with high priority (>100000): OPENCV_PARALLEL_PRIORITY_LIST=TBB,OPENMP. Unknown backends are registered as new plugins.

Classes
class	cv::parallel::ParallelForAPI

class	cv::parallel::openmp::ParallelForBackend

class	cv::parallel::tbb::ParallelForBackend

Functions
void	cv::parallel::setParallelForBackend (const std::shared_ptr< ParallelForAPI > &api, bool propagateNumThreads=true)
	Replace OpenCV parallel_for backend.

bool	cv::parallel::setParallelForBackend (const std::string &backendName, bool propagateNumThreads=true)
	Change OpenCV parallel_for backend.

Function Documentation

◆ setParallelForBackend() [1/2]

void cv::parallel::setParallelForBackend	(	const std::shared_ptr< ParallelForAPI > &	api,
		bool	propagateNumThreads = true )

Python:
	cv.parallel.setParallelForBackend(	backendName[, propagateNumThreads]	) ->	retval

#include <opencv2/core/parallel/parallel_backend.hpp>

Replace OpenCV parallel_for backend.

Application can replace OpenCV parallel_for() backend with own implementation.

Note: This call is not thread-safe. Consider calling this function from the main() before any other OpenCV processing functions (and without any other created threads).

◆ setParallelForBackend() [2/2]

bool cv::parallel::setParallelForBackend	(	const std::string &	backendName,
		bool	propagateNumThreads = true )

Python:
	cv.parallel.setParallelForBackend(	backendName[, propagateNumThreads]	) ->	retval