Wednesday 25 May will be our round-table panel discussion
in CS 240A.  As preparation, you are to write a "position
statement" of about 50 words on one question to do with
languages / machines / systems for parallel computing,
and *email* it to me tonight (or at least by 7am tomorrow
so I can read it before class).

Then in class I'll give everyone a chance to state their
position and we'll discuss them.  I hope you will take and
defend controversial positions; if everybody agrees with
everybody else then I'll disagree with everyone just to
get a debate started.

Here are some sample questions; you can pick one to answer
or make up a better one yourself:

- Is MPI or UPC easier to use?  Which is better for
  performance tuning?

- How about UPC vs CAF (has anyone used both)?

- What does the parallel language of your dreams look like?
  What would it take to make it work?

- How do you debug parallel programs?  What tools would
  make it easier?

- How do you tune parallel performance?  What tools would
  make it easier?

- Will the Cray X1 and its descendants be viable in the
  marketplace?

- If you were DARPA, which of the three HPC vendors -- Sun,
  IBM, Cray -- would you fund to build prototype machines
  for round 3 of HPCS?

- For the non-computer scientists in class:  What, if
  anything, would make it worthwhile for somebody in your
  field to use one of the 10 biggest parallel machines in
  the world?

- For the computer scientists in the class:  Same question.

- If you had $10K to spend on a parallel machine would you
  buy a Beowulf cluster?  If you had $10M would you spend
  it on a really big cluster or on something else?

- What would it take to make Matlab*P a widely used
  parallel language?

- What is the most important open problem in parallel
  algorithms?

- What is the most important open problem in architecture
  of parallel machines?

- What is the most important open problem in parallel
  languages and tools?

- Is it a matrix or is it a graph?