GitHub - hegner/concurrent_framework: a demonstrator for a concurrent framework

Branches Tags

Name		Name	Last commit message	Last commit date
Latest commit History 35 Commits
cmake		cmake
.gitignore		.gitignore
Algo.h		Algo.h
AlgoPool.cpp		AlgoPool.cpp
AlgoPool.h		AlgoPool.h
CMSExample.h		CMSExample.h
CMakeLists.txt		CMakeLists.txt
ConcurrentTypes.h		ConcurrentTypes.h
Context.cpp		Context.cpp
Context.h		Context.h
EventLoopManager.cpp		EventLoopManager.cpp
EventLoopManager.h		EventLoopManager.h
ExampleChains.h		ExampleChains.h
Helpers.h		Helpers.h
LHCbGraph.h		LHCbGraph.h
README		README
Scheduler.cpp		Scheduler.cpp
Scheduler.h		Scheduler.h
Whiteboard.cpp		Whiteboard.cpp
Whiteboard.h		Whiteboard.h
taskbits_example.cpp		taskbits_example.cpp
thisCF4Hep.csh		thisCF4Hep.csh

Repository files navigation

To Compile:
===========

cmake -G Xcode -DTBB_INSTALL_DIR=/opt/intel/tbb/tbb40_297oss
...fire up XCode

OR

cmake -DTBB_INSTALL_DIR=/opt/intel/tbb/tbb40_297oss
make

About:
======

This project provides a concurrent whiteboard and a few implementations of parallel algorithm/module scheduling for HEP on top of Intel's TBB:

 1) dependency_graph.cpp uses tbb::flow::graph. A few additions to TBB were necessary
 2) taskgraph_example.cpp uses tbb:task directly and uses N copies of the algorithm dependency DAG to handle N events in parallel
 3) taskbits_example.cpp uses tbb:task directly and uses each one bit per algorithm to descripe the state of the event. 
 
The implementations 2) and 3) provide a feature to deal with limited resources/ non-reentrant algorithms.