A high-performance shared memory library containing data structures and synchronization primitives compatible with shared memory, CUDA, and ROCm.

• CMake >= 3.10
• C++17 compatible compiler
• Optional: CUDA toolkit (for GPU support)
• Optional: ROCm (for AMD GPU support)
• Optional: MPI, ZeroMQ, Thallium (for distributed computing)

For those installing this component (rather than all of iowarp):

git clone https://github.com/grc-iit/grc-repo.git
spack repo add grc-repo
spack install cte-hermes-shm

This will install dependencies of cte-hermes-shm:

git clone https://github.com/grc-iit/grc-repo.git
spack repo add grc-repo
spack install cte-hermes-shm +nocompile
spack load cte-hermes-shm +nocompile

NOTE: spack load needs to be done for each new terminal.

This will compile:

git clone https://github.com/grc-iit/cte-hermes-shm.git
cd cte-hermes-shm
mkdir build
cd build
cmake ../ -DHSHM_ENABLE_CUDA=OFF -DHSHM_ENABLE_ROCM=OFF
make -j8

find_package(HermesShm CONFIG REQUIRED)
message(STATUS "found cte-hermes-shm.h at ${HermesShm_INCLUDE_DIRS}")
target_link_libraries(hshm::cxx)

find_package(HermesShm CONFIG REQUIRED)
message(STATUS "found cte-hermes-shm.h at ${HermesShm_INCLUDE_DIRS}")
target_link_libraries(hshm::cudacxx)

find_package(HermesShm CONFIG REQUIRED)
message(STATUS "found cte-hermes-shm.h at ${HermesShm_INCLUDE_DIRS}")
target_link_libraries(hshm::rocmcxx_gpu)

To run the tests, do the following:

ctest

To run the MPSC queue tests, do the following:

ctest -VV -R test_mpsc_queue_mpi

• include/hermes_shm/ - Core library headers
  • data_structures/ - Shared memory data structures
  • memory/ - Memory management and allocators
  • thread/ - Threading and synchronization primitives
  • util/ - Utility functions and helpers
• src/ - Core library implementation
• test/ - Unit tests and benchmarks
• benchmark/ - Performance benchmarks
• CMake/ - CMake configuration files