-
If a parallel program to process images is organized so that each process
works on separate images, it is:
A. data parallel
B. functionally parallel
C. both data and functional parallel
D. neither data nor functional parallel
-
Given an 8-way parallel program to make a movie, if it is organized so that
half the processes run a simulation algorithm and the other half of the
processes generate images from the simulation results, it is:
A. data parallel
B. functionally parallel
C. both data and functional parallel
D. neither data nor functional parallel
-
For the audio signal processing example, how many time units would it take to
process 4 chunks of data through all three filters in serial compared to parallel?
(One unit equals the time it takes to process one chunk of data through one filter.)
A. both would take 4 time units
B. serial would take 12 time units and parallel would take 4 time units
C. serial would take 12 time units and parallel would take 6 time units
D. serial would take 14 time units and parallel would take 8 time units
-
If a 3-D image processing code uses planes of pixel values instead of
nearest-neighbor pixel values in its algorithm, which domain decomposition
would work best?
A. None - the algorithm can't be parallelized
B. Decomposition by cutting the space along one dimension (slabs)
C. Decomposition by cutting the space in two dimensions (pencils)
D. Decomposition by cutting the space in three dimensions (blocks)
-
In the pollution example, what approaches might one use to improve the
speed of the code (time to get an answer at least as good as the one from
the example as given)? Assume the number of trees is constant.
A. Use twice the number of slices (each half as wide) to increase the
probability that each process has some work to do
B. Use half the number of slices (each twice as wide) so there are fewer
slices to deal with
C. Use a cyclic pattern of slices in the other dimension (i.e., oriented at
90 degrees to the pattern shown)
D. Either A or C
-
The chess code is an example of an optimization problem. Such problems often
lend themselves to master/worker parallelization because each instance can be
computed independently of all the others, yet the time to explore any single
instance may vary widely. Which of these is also an optimization problem?
A. Molecular Dynamics of a fluid
B. Protein folding (finding the minimum energy configuration for a protein)
C. Querying a database
D. Ray tracing
-
If the wave problem were solved on twice as many processors, the time to
an answer would
A. increase
B. stay the same
C. decrease to approximately 1/2
D. decrease to exactly 1/2 of the original time
-
If the wave problem were solved for a string of twice the length using
the same point spacing and twice as many processors, the time to an answer
would approximately
A. double
B. increase by 1.5 times
C. stay the same
D. decrease by 1/2
