Week 2

During week two:
  • I read more about CUDA programming.
    • I read chapters 1-3 of an online book.
    • I watched the slides and listened to online lectures.
    • Some of the material was new, some old, it helped me become more familiar with CUDA.
    • The materials can be found here.
  • I had multiple meetings with Dr. Taufer and Philip.
    • We had a discussion about the basics of GPU programming and CUDA.
    • We discussed other portions of our project (in particular, of good methodologies for measuring performance).
    • We decided to apply an N-factor analysis for the study of our performance.
    • To this end, we re-engineered the code to allow for measurements of time spent in the program (i.e.: host to/from device communication, kernel execution).
  • I signed up for two nvidia webinars.
    • Philip and I have participated in one.
    • This webinar covered the basics of using CUDA for GPGPU programming.
  • We also participated in the second with Dr. Taufer.
    • This webinar covered performance considerations and ways to optimize performance.
  • I added a way in our makefile for the matrix multiplication code to compile at either double or single precision.
  • I began working on a script to automate the n factor analysis for the matrix multiplication and its performance.
  • With Dr. Taufer and others in the group, I joined Dr. Pollock and some of her students in a picnic.
    • It was fun!