vecpar

This is a header-only library for enabling single source code (C++) to target heterogeneous platforms (CPU, GPU). The project is still in early R&D phase.

Supported backends

• (CPU) OpenMP
• (GPU) CUDA

Dependencies

The project requires LLVM/Clang to build the sources. Recommendation:

spack install llvm@14.0.0 +all_targets +cuda cuda_arch=<XY>

Dependency	OpenMP backend	CUDA backend	Tests
vecmem library	x	x	x
OpenMP 5.0 (enabled by default with LLVM14)	x
CUDA 11.5.0		x
GoogleTest			x

vecpar uses nestoroprysk/FunctionComposition for supporting the algorithm chaining functionality.

Installation

Get the code

git clone --recurse-submodules https://github.com/wr-hamburg/vecpar.git

To build the code

cmake -S <source_dir> -B <build_directory>

cmake --build <build_directory> \
       -DVECPAR_BUILD_OMP_BACKEND=On \ 
       -DVECPAR_BUILD_CUDA_BACKEND=On 

To enable the automated tests, set also -DVECPAR_BUILD_TESTS=On.

By default, all build options are enabled.

To compile for aarch64, set CC/CXX environment variables to appropriate aarch64 compilers.

To install the library

cmake --install <build_directory>

Collection types

vecpar supports vecmem::vectors as input and/or output for all operations. vecmem::jagged_vectors are supported for a restricted subset (marked with x):

Abstraction	Jagged_vector as input(s)	Jagged_vector as output
map	x	x
filter
reduce
map-filter	x
map-reduce	x