PAR Class 1, Thurs 2019-01-10

Source

1   Material

  1. Read the syllabus.
  2. Read https://computing.llnl.gov/tutorials/parallel_comp/ for an intro to parallel computing.
  3. Some points:
    1. Parallel computing is decades old; there were commercial machines in the 1980s. However, then, clocks speeds were increasing so serial machines were more interesting.
    2. Now: physics limits to processor speed.
    3. History of Nvidia.
    4. Intel CPUs vs Nvidia CUDA cores.
    5. Advantages and disadvantages of shared memory.
    6. OpenMP vs CUDA.
    7. Rate-limiting cost usually I/O not computation.
  4. Think about these questions for next week.
    1. Answer the following questions.
      1. Why have machine cycle speeds stopped increasing?
      2. What architectures do the top 3 machines on the Top 500 list use?
      3. Which one of the following 4 choices are most GPUs: SISD, SIMD, MISD, MIMD.
      4. Which one of the following 4 choices are most current multicore CPUs: SISD, SIMD, MISD, MIMD.
      5. Per Amdahl's law, if a program is 10% sequential and 90% parallelizable, what is the max speed up that can be obtained with an infinite number of parallel processors?

2   Computer

  1. parallel.ecse accounts
    1. I'll try to set yours up in class today.
    2. Play with them. I recomend connecting with ssh.
  2. Intel Xeon Phi coprocessor
    1. Parallel.ecse has one.
    2. This tech was dropped by Intel, so I don't intend to cover it in class.
    3. If anyone is interested, tell me today. Otherwise it may stop working when I upgrade the system from Ubuntu 16.04 to 18.xx

3   My research

I do parallel geometry algorithms on large problems for CAD and GIS. See my home page. If this is interesting, talk to me.