Talk GPU Computing

Target audience: Engineers and scientists using Python for scientific and numerical computing.
Prerequisites: Intermediate Python and Numpy experience. Some familiarity with a concurrent programming model, such as threads, multiprocessing, or MPI.
Software requirements: An OpenCL implementation for your GPU (NVIDIA or AMD) or the AMD CPU OpenCL implementation if you don't have a recent AMD or NVIDIA GPU; PyOpenCL.

Topics covered:

  • Hardware architecture overview
  • The basics:
    • Host<->device memory transfers
    • Compute kernel examples: Matrix multiply, random number generation.
  • Optimization techniques
  • PyOpenCL compared to PyCUDA
  • Does it live up to the hype: What problems are suited for the GPU.
tagged by
no related entity