For students in an undergraduate computer science course-who most likely have no “real world working experience” in their chosen field of study-it is extremely difficult for them to understand why what they are learning is important. They are left wondering how they can apply their skill-sets outside of the classroom. I personally remember that some of my favorite classes at UC Berkeley were rich with examples and stories about how concepts in the lesson plan mapped to the world outside my college environment.
Michael Wrinn and Tim Mattson of Intel Corporation kicked off SC10 on Sunday with a full-day tutorial “A Hands-on Introduction to OpenMP”.
The tutorial was taught with an active learning methodology, which Tim noted is a much more challenging teaching style than simply lecturing. At the start of the event, the attendees connected to the 32-core/64 thread Intel® Manycore Testing Lab (used throughout the full-day tutorial).
The tutorial covered the following topics:
The speaker started with a brief summary of why power and thermal modeling is important to HPC.
- #1 top 500 system consumes 4MW power
- The cost of a megawatt of power for one year ranging from $590 to $1.85 million
Well, it has not come to this, but there is still quite a bit of heat in the discussion as to how academia should respond to the near universal move by CPU manufacturers to a manycore architecture. I think that the response must be that parallel programming is just the way computer and computational sciences are taught. Of course that is easier said than done.
- Page 1