Project 3 Clarifications
Last Updated 2013-03-13
- Question 6 - For question 6 you need to examine cut-offs
in two (really 3) places. First you are looking at sequential
vs. parallel versions of corner finding. Looking at V1 and V2 works
here, vary the cutoff for V2. Second you are looking at cut-offs in
the grid-building step. Looking at V3 and V4 works here. There are
actually two places to vary the cut-off for V4 - the the summing of
census data and in the combining of grids. You don't have to worry
about finding the optimal combination of cut-offs or anything but we
do expect data on varying cut-offs for both corner finding and both
phases of grid building. Be sure to answer the question posed: If
you make the cut-off high enough, does the parallel performance
approach sequential performance?
Last Updated 2013-03-12
- Version 4 - combining must be done in parallel - Just to
be clear, for version 4, when combining the results of two
subproblems, you need to do this in parallel.
- What To Turn in For Phase B - For phase B please turn in
ALL files needed to run versions 1-5. Please submit any extra java
files you may have created for testing or otherwise. We will be
re-examining your version 1 and 2 code that you submit with phase B,
so this is your opportunity to fix any bugs, improve the style of
that code, etc. You DO need to submit testing code with phase
B. You do NOT need to submit timing code with phase B - you will
submit this on Thurs with your writeup - not to be graded, just to
help us understand how you did your timing if there are any
questions. You DO need to submit any above and beyond items with
phase B.
Last Updated 2013-03-05
- What To Turn in For Phase A - For phase A please turn in
ALL files needed to run your version 1 and version 2. Please submit
any extra java files you may have created for testing or otherwise.
Phase A turn-in will be graded pimarily on correctness. You will be
resubmitting you "final" version of version 1 and version 2 with
phase B.
- Cutoffs - Strictly speaking, for functional correctness
(which is the main thing we are looking for in phase-A turn-in) it
will be fine to NOT have a cutoff at this point. For phase B turn-in,
you should put in a cutoffs. Note that question 6 in the writeup asks
you to perform timing experiments involving different cutoffs.
- Invalid User Inputs - The point of specifying that "Any
illegal input (i.e., not 4 integers on one line) indicates the
user is done and the program should end." was to avoid specifying
(and you having to implement) a complex spec of what to do when
they user enters X. It will be fine if you want to be more
forgiving and give the user feedback on what was wrong
(e.g. "query not in valid range") and another chance then this
will also be o.k. Although please do be sure that it IS easy to
actually QUIT the program (say by typing a q).