Acceleration with Graphics Processing Units

A Graphics Processing Unit (GPU) is a massively parallel computer, capable of teraflop performance.

• A Massively Parallel Processor: the GPU
  • Introduction to General Purpose GPUs
  • Graphics Processors as Parallel Computers
  • Bibliography
  • Exercises
• Programming GPUs with PyCUDA and Julia
  • Data Parallelism
  • Matrix Matrix Multiplication
  • PyCUDA
  • Vendor Agnostic GPU Computing in Julia
  • Bibliography
  • Exercises
• Introduction to CUDA
  • Our first GPU Program
  • CUDA Program Structure
  • using CUDA.jl
  • Exercises
• Data Parallelism and Matrix Multiplication
  • Data Parallelism
  • Code for Matrix-Matrix Multiplication
  • Two Dimensional Arrays of Threads
  • Examining Performance
  • using CUDA.jl and Metal.jl
  • Exercises
• Device Memories and Matrix Multiplication
  • Device Memories
  • Matrix Multiplication
  • using shared memory with CUDA.jl
  • Bibliography
  • Exercises
• Thread Organization and Matrix Multiplication
  • Thread Organization
  • Matrix Matrix Multiplication
  • Submatrices with Threads in CUDA.jl
  • Thread Synchronization
  • Bibliography
  • Exercises
• Warps and Reduction Algorithms
  • More on Thread Execution
  • Parallel Reduction Algorithms
  • Julia Defined Kernels
  • Bibliography
  • Exercises