Advanced Parallel Grid Processing with MPI

A C++ MPI-based parallel computing project demonstrating core parallel algorithms and communication patterns.

Features

Algorithms

• Heat Diffusion -- Parallel 2D Jacobi stencil with row-wise decomposition across MPI ranks. Supports both blocking and non-blocking halo exchange.
• Matrix Multiplication -- Parallel row-partitioned matrix multiplication using MPI_Scatterv/MPI_Gatherv with matrix broadcast.

Communication Demos

• Blocking Exchange -- Parity-ordered MPI_Send/MPI_Recv with deadlock prevention
• Non-blocking Exchange -- MPI_Isend/MPI_Irecv with MPI_Waitall
• Neighbor Exchange -- 1D and 2D exchange via MPI_Sendrecv
• Ring Communication -- Token passing around a ring of processes
• Pipeline Communication -- Value flowing through a chain of processes
• Deadlock Demo -- Demonstrates deadlock scenario and its fix
• Process Groups -- Splitting communicators via MPI_Comm_split

Prerequisites

• MPI Implementation -- MS-MPI (Windows), MPICH, or OpenMPI
• CMake >= 3.16
• C++17 compatible compiler (MSVC, GCC, Clang)
• Ninja (optional, recommended)

Building

Option 1: Command Line (CMake + Ninja)

cmake -B build -G Ninja
cmake --build build

Option 2: Command Line (CMake + default generator)

cmake -B build
cmake --build build --config Release

Running

Basic Usage

mpirun -np <num_processes> ./parallel_grid [algorithm] [options]

Note: On Windows with MS-MPI, use mpiexec instead of mpirun.

Interactive Mode

Run without arguments to get an interactive menu:

mpirun -np 4 ./parallel_grid

Command-Line Options

Algorithm	Command	Options
Heat Diffusion	heat	--grid <N> grid size (default 512), --iter <I> iterations (default 200), --nonblocking use non-blocking exchange
Matrix Multiplication	matmul	--size <N> random NxN size (default 256), --verify verify against sequential
Both	both (default)	Combines heat + matmul settings
Deadlock Demo	deadlock_demo	No additional options
Ring Demo	ring	No additional options
Pipeline Demo	pipeline	No additional options

Examples

# Heat diffusion on 256x256 grid, 100 iterations, 8 processes
mpirun -np 8 ./parallel_grid heat --grid 256 --iter 100

# Matrix multiplication 512x512 with verification
mpirun -np 4 ./parallel_grid matmul --size 512 --verify

# Non-blocking halo exchange
mpirun -np 4 ./parallel_grid heat --grid 512 --iter 200 --nonblocking

# Communication demo
mpirun -np 6 ./parallel_grid ring

## Output

- **Heat Diffusion** produces `heat_output.csv` containing the final temperature grid, suitable for visualization in Excel or Python.
- **Matrix Multiplication** prints timing results and optionally verifies correctness against a sequential implementation.
- All algorithms report wall-clock timing via the built-in `Timer` class.

## MPI Concepts Demonstrated

| Concept | MPI Functions Used |
|---|---|
| Data Decomposition | Row-wise partitioning of grids and matrices |
| Broadcast | `MPI_Bcast` |
| Scatter/Gather | `MPI_Scatterv`, `MPI_Gatherv` |
| Reduction | `MPI_Reduce` |
| Blocking Communication | `MPI_Send`, `MPI_Recv` |
| Non-blocking Communication | `MPI_Isend`, `MPI_Irecv`, `MPI_Waitall` |
| Send-Receive | `MPI_Sendrecv` |
| Process Groups | `MPI_Comm_split` |
| Synchronization | `MPI_Barrier` |