Lecture 5/6 - Design and Analysis of Parallel Programs Flashcards

1
Q

PRAM Model

A

parallele random access machine

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
2
Q

MIMD

A

multiple instructions multiple data

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
3
Q

Parallel Performance Models

A
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
4
Q

Parallel execution time

A

Counted from the start time of the earliest task to the finishing time of the latest task

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
5
Q

Work

A

the total number of performed elementary operations

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
6
Q

Cost

A

the product of parallel execution time and #processors

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
7
Q

Speed-up

A

the factor by how much faster we can solve a problem with p processors than with 1 processor, usually in range (0…p)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
8
Q

Parallel efficiency

A

Speed-up / #processors, usually in (0…1)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
9
Q

Throughput

A

operations finished per second

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
10
Q

Scalability

A

does speedup keep growing well also when #processors grows large?

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
11
Q

Asymptotic Analysis

A

Estimation based on a cost model and algorithm idea (pseudocode operations) n Discuss behavior for large problem sizes, large #processors

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
12
Q

Empirical Analysis

A

n Implement in a concrete parallel programming langauge n Measure time on a concrete parallell computer l Vary number of processors used, as far as possible n More precise n More work, and fixing bad designs at this stage is expensive

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
13
Q

Greedy scheduling

A

(also known as ASAP, as soon as possible) Dispatch each task as soon as - it is data-ready (its predecessors have finished) - and a free processor is available

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
14
Q

Critical-path scheduling

A

Schedule tasks on critical path first, then insert remaining tasks where dependences allow, inserting new time steps if no appropriate free slot available

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
15
Q

Layer-wise scheduling

A

Decompose the task graph into layers of independent tasks Schedule all tasks in a layer before proceeding to the next

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
16
Q

Brent’s Theorem

A
17
Q

Amdahl’s Law

A
18
Q

Speedup Anomalies

A

superlinear speedup

possible causes: cache effects, search anomalies

19
Q

parallel prefix sum algorithms

A
20
Q

parallel list ranking algorithms

A