How tesraga1/parallel_programming_matrix_multiplication works
This repository is a collection of parallel matrix-multiplication implementations, not a single application. The principal workflow is to launch an MPI program with a matrix-size argument; a selected algorithm distributes matrix blocks, exchanges or broadcasts them among ranks, computes local products, and reports timing or a result check. The tree also shows CUDA and hybrid variants. Source excerpts confirm the MPI flows and some host-to-CUDA calls; implementations not excerpted are included as components where the tree supports them, but their internal relationships are left unspecified.
Main entry points: fox_mpi_cuda.c Fox MPI CUDA · MPI_cannon.c MPI Cannon GPU · summa_mpi_cuda.c SUMMA MPI CUDA · Cannon_MPI.c Cannon MPI
Components 14
Cannon variants
- Cannon MPI
src/Cannon/Cannon_MPI.c - Cannon CUDA
src/Cannon/Cannon_CUDA.cu - Cannon hybrid
src/Cannon/Cannon_Hybrid.cu - MPI Cannon GPU
src/Cannon/MPI_cannon.c - Cannon GPU multiply
src/Cannon/Cuda_Cannon.cu
Fox variants
- Fox MPI
src/Fox/MPI_only/fox_mpi.c - Fox MPI CUDA
src/Fox/MPI_and_CUDA/fox_mpi_cuda.c - Fox CUDA support
src/Fox/MPI_and_CUDA/fox_mpi_cuda.c - Fox CUDA multiply
src/Fox/MPI_and_CUDA/fox_mpi_cuda.cu
SUMMA variants
- SUMMA MPI
src/SUMMA/summa_mpi.c - SUMMA MPI CUDA
src/SUMMA/summa_mpi_cuda.c - SUMMA CUDA
src/SUMMA/summa_cuda.cu - SUMMA CUDA kernel
src/SUMMA/summa_mpi_cuda_kernel.cu
Other components
- MPI operator
Connections 10
- MPI operator → Cannon MPI: launches
src/Cannon/Cannon_MPI.c - MPI operator → MPI Cannon GPU: launches
src/Cannon/MPI_cannon.c - MPI operator → Fox MPI: launches
src/Fox/MPI_only/fox_mpi.c - MPI operator → Fox MPI CUDA: launches
src/Fox/MPI_and_CUDA/fox_mpi_cuda.c - MPI operator → SUMMA MPI: launches
src/SUMMA/summa_mpi.c - MPI operator → SUMMA MPI CUDA: launches
src/SUMMA/summa_mpi_cuda.c - MPI Cannon GPU → Cannon GPU multiply: calls GPU multiply
- Fox MPI CUDA → Fox CUDA support: calls CUDA support
src/Fox/MPI_and_CUDA/fox_mpi_cuda.c - Fox MPI CUDA → Fox CUDA multiply: launches multiply
src/Fox/MPI_and_CUDA/fox_mpi_cuda.c - SUMMA MPI CUDA → SUMMA CUDA kernel: calls CUDA multiply