1
begin
2
Parallel ComputingWhat is it good for?

• William M. Jones, Ph.D.
• Assistant Professor
• Computer Science Department
• Coastal Carolina University

3
What is Parallel Computing?(and why do we need
it?)

• Concurrently using multiple compute resources
• Solve problems faster
• Solve larger problems
• Solve problems with greater realism and fidelity
• Target science and engineering needs

4
The (new) Scientific Method
5
Simulation-based Design and Analysis
6
Weather Prediction
7
Astrophysics
8
Entertainment
9
So how do we start?Lots of well-designed
HARDWAREWell-written SOFTWAREA Good Start
10
Weve Come A Long Way
11
Progress
12
(No Transcript)
13
Basic ArchitectureSMP Shared Memory
14
Distributed Memory
15
Hybrid Model
16
Multi-core ILP VLIW Pipelining Vector
Units
Lots of power at multiple hierarchical levels
? Many levels of potential parallelism
17
But What About Software? applications
compilers debuggers performance analyzers
How do we even start to write a program that
could run on machines like these?
18
First you must unlearn that which you have
learned.(at least partially)
the sequential paradigm
19
New (better) Way of Thinking
the parallel paradigm
20
Domain Decomposition
21
Functional Decomposition
22
So, how much faster can we make it?
What are the limits to the performance
improvement?
23
(No Transcript)
24
(No Transcript)
25
(No Transcript)
26
Questions?