ompi-x/moc

MPI Open MP Coordination library

moc's Issues

Copyright statements are not up-to-date

Copyright statements are missing in src/moc.c and outdated in test/simple_test.c

Support Case12 - Ignoring GPUs

Support Case10 - Ignoring GPUs

Support Case4 - Ignoring GPUs

Support Case11 - Ignoring GPUs

support rank-based OMP places with "vanilla" OpenMP

Add support to MOC to use generic (unmodified) OpenMP runtime, but define the OMP_PLACES value as a function of the MPI ranks (i.e., OMPI_PLACES defined in MOC_Init() before OpenMP initialized).

This should be enabled/disabled at compile time to avoid any unexpected interactions when using a modified OMP runtime (e.g., libomp).

NOTE: Focus on OpenMP threads (places) on "cores" use case first. Also, avoid NUMA case at first because we do not have hwloc/fine-grained info.

Target use cases (in order of interest/priority):

• 1.a) 1 rank per physical node -> OpenMP threads (places) on cores
• 2.a) 1 rank per core -> OpenMP threads (places) on cores
• 3.a) 1 rank per NUMA node -> OpenMP threads (places) on cores
• 1.b) 1 rank per physical node -> OpenMP threads (places) on hwthreads
• 2.b) 1 rank per core -> OpenMP threads (places) on hwthreads
• 3.b) 1 rank per NUMA node -> OpenMP threads (places) on hwthreads
• 4.x) Future - 1 rank per GPU -> OpenMP threads (places) ???

Fix test compilation

For now, I am using a handwritten Makefile to compile mpi_omp_hello.c

CC=clang
CFLAGS=-fopenmp
MPI_COMPILE_FLAGS=$(shell mpicc --showme:compile)
MPI_LINK_FLAGS=$(shell mpicc --showme:link)

all: mpi_omp_hello

mpi_omp_hello: mpi_omp_hello.c
    $(CC) -I/home/gvh/projects/OMPIX/codes/install/llvm/include \
        -I/home/gvh/projects/OMPIX/install/pmix/include \
        -I/home/gvh/projects/OMPIX/install/moc/include \
        -L/home/gvh/projects/OMPIX/install/pmix/lib \
        -L/home/gvh/projects/OMPIX/install/moc/lib \
        $(MPI_COMPILE_FLAGS) $(CFLAGS) $(MPI_LINK_FLAGS) \
        -o mpi_omp_hello mpi_omp_hello.c -lpmix -lmoc

clean:
    rm -f mpi_omp_hello

Support Case8 - Ignoring GPUs

Support Case2 - Ignoring GPUs

Support Case5 - Ignoring GPUs

Support Case1 - Ignoring GPUs

1 MPI rank per core.

Basically for that one, we just need to find the best way to do this on Summitdev/summit and document it.

Put in place mechanism to switch between MOC modes

We want to be able to support two different contexts:

1. The Open MP compiler is not modified and MOC performs static configuration of the environment through Open MP environment variables. Static mode.
2. The Open MP compiler is modified and MOC coordinates MPI and Open MP. Dynamic mode.

Since these modes are not compatible, a configure/compile-time option is acceptable. Documentation is required.

Support Case3 - Ignoring GPUs

Support Case7 - Ignoring GPUs

tjn: setup build env on summitdev for "vanilla" omp case

Setup the build environment on Summitdev for the "vanilla" OMP (no modifications to OpenMP) case. This will require OpenMPI (libevent, pmix, ompi) and MOC (ompi, pmix) and a proper LLVM/Clang.

Note: Install into stf010 area.

• libevent
• pmix
• ompi
• moc (branch: tjn-myrank-place-vanilla)
• llvm/clang (module load? or from source, see scripts)

Support Case6 - Ignoring GPUs

Support Case9 - Ignoring GPUs