High-Performance Mesh-Quality Assessment for Diamond-Lattice Metamaterials

This repository contains the code developed for my COMPE 596 final project: an end-to-end pipeline that generates a tetrahedral mesh of a diamond-lattice metamaterial and computes geometric quality metrics on every element, using serial C++, MPI, and CUDA implementations to compare performance.

Features

1. Point-Cloud Generation

   • Defines an N x N x N diamond lattice (FCC sites + two-atom basis) with OpenMP-parallelized C++ code to emit a TetGen .node file.

2. Mesh Construction

   • Invokes TetGen on the generated .node file to produce .ele (tet connectivity) and .face (surface triangles).

3. Quality Metric Computation

   • Serial C++: Loads .node/.ele, computes per-tet aspect-ratio, aggregates min/avg/max.
   • MPI: Distributes tets across ranks, times local computations with MPI_Wtime(), then reduces global metrics and timing.
   • CUDA: Launches one thread per tet to compute aspect-ratio on the GPU, times the kernel with CUDA events, then aggregates results on the host.

4. Benchmarking Suite

   • Shell scripts to sweep lattice size (N) and number of MPI ranks, generating timing tables and speed-up plots.

Prerequisites

• C++17 compiler with OpenMP support (e.g., g++)
• MPI library & wrapper (e.g., mpicxx)
• CUDA Toolkit (nvcc)
• TetGen installed for mesh generation
• make or cmake