Introduction to CuPy

This repository provides a comprehensive guide to understanding and utilizing CuPy, a GPU-accelerated library for Python that is designed to scale large-scale numerical computations and data analysis workflows.

Note

This project is under active development.

Contents

Topic Covered

Topics	Duration
ARE Setup	30 Minutes
GPU architecture	30 Minutes
CuPy Basics	60 Minutes
User Kernels	30 Minutes
CUDA Events	15 Minutes
CUDA Stream	15 Minutes

• Prerequisite
• Learning Outcomes
• Modules
• Python Virtual Environment
• CuPy
  • GPU Parallelism
    • GPU Execution Model
      • Thread Indexing
      • How do we assign a unique thread id to each thread?
      • Warps and Warp Schedulers
    • Data Movement in GPUs
    • Streams
    • Exercise
  • CuPy Function Overview: Key Operations and Utilities
    • Device Count
    • Device Properties
    • Current Device
    • CuPy ndarray
      • Comparison with NumPy ndarray
      • Allocating ndarray on a Specific GPU
      • Finding the GPU where the ndarray is located
    • Data Movement in CuPy
    • Transferring Data from GPU to Host
    • Building Device Agnostic Code
    • Explicit data transferDifferences
    • Automatic Data Transfer
  • User defined Kernels
    • Elementwise Kernel
    • Reduction Kernel
      • Raw Kernel
  • CUDA Events
  • CUDA Streams
• Reference
• Contributers