rust_ml_experiments

Practical Rust & CUDA ML experiments

Projects: 🗒️

• Matrix Multiplication Benchmark (CPU vs GPU) rust_gemm.md
• GPU-Accelerated Image Blur (Box Filter) box_filter.md
• Mandelbrot Set Rendering mandelbrot.md
• N-Body Physics Simulation physics.md
• Burn & Cuda: Resnet-18 on CIFAR-10 restnet-18.md
• Custom CUDA Kenernel with rustacuda (Parallel Reduction) parrallel_reduction.md
• Monte Carlo Pi Estimation monte_carlo.md

Key Optimisations 📈

• Matrix Multiply (GEMM) : Shared Memory & tiling. 50 -100x expected speedup (vs CPU).
• Image Blur : Shared memory for 3x3 neighborhood. 20-50x expected speedup (vs CPU).
• Mandelbrot : Coalesced memory access. 10-30x expected speedup (vs CPU).
• N-Body : Shared memory & tree methods. 30-80x expected speedup (vs CPU).
• Parallel Reduction : Bank conflict avoidence. 50-150x expected speedup (vs CPU).
• Monte Carlo Pi : Minimize atomic ops. 40-100x expected speedup (vs CPU).

Build and Run it (Generic) 🛠️

• Run it Locally:

```bash
      cargo run --release
 ```  

• Build and Run it Docker:

```bash
      docker build -t ${PROJECT_TITLE} .
      docker run -it ${PROJECT_TITLE}
 ```