High-Energy Particle Storms Simulation

Multicore Programming Exam 2025/2026

Project Description

This project simulates the effects of high-energy particle bombardment on an exposed surface, such as the hull of a space vessel. The simulation considers a cross-section of the surface represented by a discrete number of control points.

The simulation follows a three-stage process for each wave of particles:

1. Bombardment: Particles impact the surface, transferring energy to the impact point and its neighborhood with distance-based attenuation.
2. Relaxation: The material redistributes its charge by averaging the energy value of each point with its immediate neighbors.
3. Analysis: The program identifies and reports the point with the highest accumulated energy for each wave.

---

Compilation and Build System

The project uses a Makefile to manage different parallel implementations. Use the following commands to build the targets:

General Commands

• make help: Shows the help menu with all available options.
• make all: Builds all versions (Sequential, MPI+OpenMP, and CUDA).
• make clean: Removes all compiled binaries and temporary files.
• make debug: Builds all versions with the -DDEBUG flag, enabling graphical output for small arrays (size $\le 35$).

Specific Targets

• make energy_storms_seq: Builds the Sequential baseline version.
• make energy_storms_mpi_omp: Builds the hybrid MPI + OpenMP version.
• make energy_storms_cuda: Builds the CUDA (GPU) version.

---

Implementations

1. Sequential (energy_storms_seq)

The baseline implementation where all phases (bombardment, relaxation, and maximum search) are executed on a single core.

2. Hybrid MPI + OpenMP (energy_storms_mpi_omp)

Designed for distributed computing:

• MPI: Used for domain decomposition, partitioning the control point array across different nodes.
• OpenMP: Used within each MPI process to parallelize the bombardment and relaxation loops across multiple CPU cores.
• Synchronization: Ensures data consistency during the relaxation phase, where neighboring values are required for averaging.

3. CUDA (energy_storms_cuda)

Leverages GPU acceleration:

• Parallel Kernels: High-throughput computation of energy attenuation for thousands of particles.
• Memory Management: Efficient data transfer between Host and Device and use of fast on-chip memory for relaxation stencils.

---

Usage

To run the simulation, use the following syntax:

./energy_storms_<version> <size> <wave_file1> [wave_file2 ...]