clDES

Implementation

clDES uses a graph based model to represent discrete-event systems, which are implemented with adjacency sparse matrices. It has a ViennaCL OpenCL back-end to enable a gpgpu approach. The CPU processing is made using the Eigen3 library and it can exploit multiple cores present in your hardware by enabling OpenMP on your compiler.

Only CPU operations are available now. The OpenCL code is being refactored. ViennaCL is going to be replaced by VexCL + custom kernels implementation.

Operations

Member Operations

Operation	Implementation
Accessible part	cldes::DESystem<NEvents, StorageIndex>::accessiblePart()
Coaccessible part	cldes::DESystem<NEvents, StorageIndex>::coaccessiblePart()
Trim	cldes::DESystemBase<NEvents, StorageIndex>::trim()
Verify Observer Property	cldes::DESystem<NEvents, StorageIndex>::verifyObsProp()

Non-member Operations

Operation	Implementation
Projection	cldes::op::proj<NEvents, StorageIndex>()
Inverse Projection	cldes::op::invproj()
Synchronization: parallel composition	cldes::op::synchronize()
Monolithic Supervisor Synthesis	cldes::op::supC()
Observer Property Searcher	cldes::op::searchObsProp()

Compiling

Follow the instructions bellow to build the project.

Requirements

It is necessary to make sure that CMake 3.11 is installed, such as LLVM Clang 6 or GCC/G++ 8 developer tools.

System support to Boost Library version >= 1.58 and Eigen3 Library are required. OpenCL version >= 1.2, and OpenMP version >= 5 are optional.

clDES embedded libraries so far:

• ViennaCL 1.7.1: Efficient linear algebra library for computations on many-core architectures.
• Sparsepp: The faster sparse hash sets/maps C++14 implementation I know.

It may compile and run smoothly on older versions. These tools mentioned above compose the tool chain used to develop and to test so far, ArchLinux and Fedora 28.

Build project

clDES is a template library, which means you only need to copy it to a directory on your compiler search path. However, you may want to compile it to run its benchmarks and tests. After checking if all requirements are satisfied, compile the project:

$ export CC=/usr/bin/clang
$ export CXX=/usr/bin/clang++
$ export CLDES_INCLUDE_PATH=<clDES_root>/include/
$ cd  <clDES_root>
$ mkdir build; cd build
$ cmake ..
$ make -j4

Optional features

To enable OpenMP or OpenCL, it is necessary to set the corresponding flags:

OpenMP:

# [...]
$ cmake -DCLDES_OPENMP_ENABLED=true ..
$ make -j4

OpenCL:

# [...]
$ cmake -DCLDES_OPENCL_ENABLED=true ..
$ make -j4

It is always necessary to set CLDES_INCLUDE_PATH to run clDES applications which execute operations that use custom OpenCL kernels, such as cldes::op::Syncronization(). Custom Kernels Currently disabled.

Dev build

If you wanna contribute to clDES, you may want a debug build by setting CMake's flag BUILD_TYPE to Debug:

$ export CC=/usr/bin/clang
$ export CXX=/usr/bin/clang++
$ export CLDES_INCLUDE_PATH=<clDES_root>/include/
$ cd  <clDES_root>
$ mkdir build; cd build
$ cmake -DCMAKE_BUILD_TYPE=Debug ..
$ make -j4

Running tests

After inserting code, or if you just want to test the library, execute the tests using CTest:

$ cd <clDES_root>/build
$ make test