Week 3

  • I went to the group picnic held for the visiting research students at UDel.
    • The food was great!
  • I finished an awk script that automates the running of our code and the collection of the results.
    • Now, we are using it to collect execution times for our matrix multiplication code.
    • For the matrix multiplication, we automatically generated different test cases that have either single or double precision and different matrix sizes.
    • We use the script on different machines with different cards. The tests are executed without user intervention.
    • The script can be adapted for other programs.
  • I reviewed Philip's presentation with Dr. Taufer.
    • We discussed general tips for effective scientific communication.
  • I created Matlab programs that reads in the data files created by the awk script, and visualizes the data contained in these files.
    • Right now, we have four different GPU types, four different CPU's, and matrix sizes ranging from 10 to 1800 in increments of 10.
    • I plan to use the program to analyze the data produced by these scripts.
  • I created a Matlab program used to explore the feasibility of using an iterative Newton-Raphson method for calculating the square root of a number.
    • This method converges very quickly, but makes use of many (potentially expensive) division operations.
  • I began working on a synthetic code to demonstrate the effects of drifting on GPU's.
  • I attended, along with Philip and Dr. Taufer, another Nvidia webinar
    • This webinar covered advanced performance optimizations in CUDA.
  • I attended the GCL weekly group meeting.
    • Philip gave his presentation.