U.C. Berkeley CS267: Comments on Assn 2&3

I'm well aware that this assignmnet was, at times, very frustrating because of cm5 capriciousness. This fact was considered when grading the assignments. Nevertheless, there are a few general comments I'd like to make.

There are at least two ways that you should have verified that your computations are correct. The first is to just visually inspect the results. The second much better method is to compare the numeric output from each of the programs (to within some precision). Although this is not feasible for all problems, for this assignment (and potentially future ones), it is an easy and accurate way to verify correctness. For the bouncing balls assignment, I strongly suggest that you implement a uni-processor version (with graphics) to verify your code. If anyone is interested, I have some simple X code in place that you can use (send me mail if interested).

Several groups figured out how to run CMF on a variable number of processors. Although I didn't penalize for not doing this, it was mentioned in the CMF Language Reference Manual.

The assignment clearly stated to try two cases, Case 1 and Case 2. Several groups didn't say anything about these cases. Please read the assignment carefully, and if there is something you do not do, mention why.

A few groups only turned a few pages of graphics. John Cage's 4'33" is nice, but in general, I would like a write-up that describes your solutions, algorithms, problems, experiences, gripes, etc. in detail. This will not only let me know what you did, but also give me a gauge as to how the assignment went. Also, since we're not turning in code, I'd like the algorithmic description (with computation/communication phases, communication schedule, etc) completely, clearly, accurately, and simply written. As is often said: "perspicacity and perspicuity should be the twin aims of writers: an acuteness of understanding coupled with lucidity of expression."

Tis true, a waste of paper is an abomination of the spirit (especially in Berkeley). So I'm not going to ask for code listings. I'd still, however, like to have access to the code. In each future assignment, please include some form of pointer (URL, ftp site, rodin path, etc) where I can find the code to check some things out if need be.

By the way, I'm perfectly happy to accept assignments as WWW pages. If you're interested, send me the URL rather than turning in hard-copy. In such case, I'll email the results back to each group member.