Problems
1. In scientific visualization, use is made of multiprocessor computers.
A major issue with multiprocessor computers is the amount of speedup that can be achieved compared to a uniprocessor computer. There are two laws covering speedup: Amdahl and Gustafson.
Do either of these have a formal proof? If so, present the formal proof and explain what it means (i.e., do not just use formalism, also use ordinary English sentences).
Do either of these have a basis in data? If so, explain the data and explain how the data is used to justify the law.
Which of these laws are more likely to be applicable to most problems in scientific visualization? Explain your answer with some examples.
Explain the various steps (or functionalities) in scientific visualization. Which steps are most easily implemented on a multiprocessor machine? Why?
In most cases, are truly parallel algorithms implemented for scientific visualization at the present epoch? If not, how is parallelization actu- ally implemented? (Hint: consider the implementation of Paraview and how one does scientific visualization within Paraview.)
2. We have examined several papers on some important aspects of visualization and pre-attentive processing (or cognition).
What is meant by pre-attentive?
Referring to the journal articles we examined, when does shape versus color come into play? How do these affect pre-attentive observations on data sets?
A major issue for visualization is compliance with the Americans with Disability Act (ADA) and the amendments to ADA in 2008.
Which sorts of disabilities most affect use of typical visualized data or simulations?
For visual disabilities (ones that cannot be mitigated by readily prescribed corrective lenses), which ones are most typical and what are some things that can be done by design to mitigate these?
What methods would you suggest for pre-attentive processing that could still be ADA compliant?
3. Consider the data in Table 1, presented as a matrix, for ac- tual run times in seconds of a particular executable program. Repeated experiments were made, and the uncertainty in each datum is 0.01 of the value of the datum. The PEs has a clock rate of 2.0 GHz. The format is (compute time)/(systems time excluding communication)/(communications time).
Display a set of contour plots or other appropriate plots for the mea- sured running times. N.B.: You may have to interpolate values within the table to produce these plots.
Given the nominal clock rate for the processor, report the data in clock ticks rather than seconds, and display a set of contour plots or other appropriate plots for the measured times in clock ticks.
Does the data display isoefficiency for any portion or portions of the reported measurements? Justify and explain your conclusions.
|
Amount of Data
|
10
|
20
|
40
|
80
|
|
Number of P Es
|
|
|
|
|
|
2
|
11/12/10
|
19/25/20
|
40/48/41
|
100/50/42
|
|
4
|
19/26/18
|
41/52/23
|
78/75/50
|
150/150/90
|
|
8
|
41/49/36
|
50/100/120
|
200/150/240
|
500/300/520
|
|
16
|
80/100/70
|
158/201/143
|
323/399/281
|
700/500/300
|
|
32
|
90/120/70
|
200/238/145
|
405/501/190
|
1000/1005/350
|
|
64
|
113/155/185
|
231/200/205
|
460/400/410
|
919/801/818
|
|
128
|
110/150/180
|
223/305/359
|
441/599/720
|
883/1200/1453
|
|
Amount of Data
|
160
|
320
|
640
|
|
Number of P Es
|
|
|
|
|
2
|
300/53/41
|
800/55/44
|
1700/65/50
|
|
4
|
398/520/200
|
1200/1500/600
|
3000/1500/700
|
|
8
|
1110/700/600
|
3000/1500/1200
|
8000/1800/2100
|
|
16
|
2000/1000/700
|
4000/2000/800
|
7995/4003/1599
|
|
32
|
2000/2000/800
|
6000/5010/1550
|
14000/2000/2010
|
|
64
|
1915/1700/950
|
5510/2100/1100
|
12000/2300/1400
|
|
128
|
1761/1400/1800
|
3518/1510/1930
|
7590/3035/2100
|
|
Amount of Data
|
1280
|
2560
|
|
Number of P Es
|
|
|
|
2
|
2100/70/55
|
6000/160/70
|
|
4
|
8000/1520/1000
|
12000/2000/1650
|
|
8
|
21000/3000/5000
|
100000/9000/6000
|
|
16
|
16005/8010/3200
|
80000/17000/5000
|
|
32
|
30000/4000/3000
|
75000/20000/8000
|
|
64
|
26000/2500/1800
|
60000/2800/2355
|
|
128
|
15000/6215/4375
|
30000/14000/7135
|