% cuda-parallel-patterns

An implementation portfolio of core CUDA parallel algorithms, memory optimizations, and performance profiles inspired by Shane Cook's "CUDA Programming".

Overview

• Purpose: learn and demonstrate CUDA patterns with runnable examples and performance notes.
• Layout:
  • common/ - shared helpers (timers, error checks)
  • 01_fundamentals/ - Vector Add, Matrix Multiply
  • 02_memory_patterns/ - Coalescing, Shared Memory, Constant Memory (examples)
  • 03_parallel_algorithms/ - Reduction, Scan, Bitonic Sort
  • 04_capstone_mini_project/ - Image convolution example
  • docs/ - performance notes and graphs

Quickstart Prerequisites: CUDA toolkit, CMake >= 3.18, a CUDA-capable GPU.

Build and run examples:

cd Embedded Programming/Cuda/cuda-parallel-patterns
./build.sh
# run a sample executable
./build/01_fundamentals/vector_add

Create & push a GitHub repo (helper script):

cd Embedded Programming/Cuda/cuda-parallel-patterns
./scripts/create_remote_repo.sh --name cuda-parallel-patterns --public

See docs/README.md for profile guidelines.