Performance Evaluation of Parallel Computers-Parallel Processing-Lecture Slides, Slides of Parallel Computing and Programming

Download Performance Evaluation of Parallel Computers-Parallel Processing-Lecture Slides and more Slides Parallel Computing and Programming in PDF only on Docsity!

Performance Evaluation of Parallel

Computers

Example: Adding numbers on Hypercube

Performance Evaluation of Parallel Computers

Parallel SpeedUp Performance

Laws

• Amdahl’s Law• Gustafson’s law• Sun & Li’s Law

Amdahl’s Law

Amdahl’s Law……cont’d

Gustafson’s law

Gustafson’s law…..cont’d

Performance

Measurement Tools

Performance Measurement Tools:

Utilization Displays

Performance Measurement Tools:

Communication Displays

Performance Measurement Tools:

Task Displays

From Previous Lecture:Algorithm for Prefix-Sum

on Hypercube

All-to-all broadcast on a Hypercube

All-to-all broadcast on a

d

-dimensional hypercube.

The Prefix-Sum Operation on Hypercube •^

The operation can be implemented using the all-to-allbroadcast kernel.

-^

We must account for the fact that in prefix sums thenode with label

k

uses information from only the

k

-node

subset whose labels are less than or equal to

k

•^

This is implemented using an additional result buffer.The content of an incoming message is added to theresult buffer only if the message comes from a node witha smaller label than the recipient node.

-^

The contents of the outgoing message (denoted byparentheses in the figure) are updated with everyincoming message.