Scribd
Upload
265 views7 pages

Principles of Scalable Performance: CSE539: Advanced Computer Architecture

This chapter discusses principles of scalable performance in parallel processing. It covers parallel processing applications such as massive parallelism for grand challenges in fields like magnetic recording, drug design, and image processing. The chapter also discusses sources of parallel overhead and application models for parallel processing like instruction parallelism vs data parallelism. It analyzes scalability using models like fixed load, fixed time, and fixed memory and how efficiency decreases as machine size and overhead increase.

Uploaded by

Shilpa Shridhar
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
265 views7 pages

Principles of Scalable Performance: CSE539: Advanced Computer Architecture

This chapter discusses principles of scalable performance in parallel processing. It covers parallel processing applications such as massive parallelism for grand challenges in fields like magnetic recording, drug design, and image processing. The chapter also discusses sources of parallel overhead and application models for parallel processing like instruction parallelism vs data parallelism. It analyzes scalability using models like fixed load, fixed time, and fixed memory and how efficiency decreases as machine size and overhead increase.

Uploaded by

Shilpa Shridhar
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 7

CSE539: Advanced Computer Architecture

Chapter 3
Principles of Scalable Performance
Book: “Advanced Computer Architecture – Parallelism, Scalability, Programmability”, Hwang & Jotwani

Sumit Mittu
Assistant Professor, CSE/IT
Lovely Professional University
[email protected]
In this chapter…

• Parallel Processing Applications


• Sources of Parallel Overhead*
• Application Models for Parallel Processing
• Speedup Performance Laws

Sumit Mittu, Assistant Professor, CSE/IT, Lovely Professional University 2


PARALLEL PROCESSING APPLICATIONS
Massive Parallelism for Grand Challenges
• Massively Parallel Processing
• High-Performance Computing and Communication (HPCC) Program
• Grand Challenges
o Magnetic Recording Industry
o Rational Drug Design
o Design of High-speed Transport Aircraft
o Catalyst for Chemical Reactions
o Ocean Modelling and Environment Modelling
o Digital Anatomy in:
• Medical Diagnosis, Pollution reduction, Image Processing, Education Research

Sumit Mittu, Assistant Professor, CSE/IT, Lovely Professional University 3


PARALLEL PROCESSING APPLICATIONS
Massive Parallelism for Grand Challenges

Sumit Mittu, Assistant Professor, CSE/IT, Lovely Professional University 4


PARALLEL PROCESSING APPLICATIONS
Massive Parallelism for Grand Challenges
• Exploiting Massive Parallelism
o Instruction Parallelism v/s Data Parallelism

• The Past and The Future


o Early Representative Massively Parallel Processing Systems
• Kendall Square KSR-1: All-cache ring multiprocessor; 43 Gflops peak performance
• IBM MPP Model: 1024 IBM RISC/6000 processors; 50 Gflops peak performance
• Cray MPP Model: 1024 DEC Alpha Processor chips; 150 Gflops peak performance
• Intel Paragon: 2D mesh-connected multicomputer; 300 Gflops peak performance
• Fujistu VPP-500: 222-PE MIMD vector system; 355 Gflops peak performance
• TMC CM-5: 16K nodes of SPARC PEs, 2Tflops peak performance

Sumit Mittu, Assistant Professor, CSE/IT, Lovely Professional University 5


PARALLEL PROCESSING APPLICATIONS
Application Models for Parallel Processing
•  Keeping the workload as constant
o Efficiency E decreases rapidly as Machine Size n increases
• Because: Overhead h increases faster than machine size
• Scalable Computer
• Workload Growth and Efficiency Curves
• Application Models
o Fixed Load Model
o Fixed Time Model
o Fixed Memory Model

• Trade-offs in Scalability Analysis

Sumit Mittu, Assistant Professor, CSE/IT, Lovely Professional University 6


PARALLEL PROCESSING APPLICATIONS
Application Models for Parallel Processing

Sumit Mittu, Assistant Professor, CSE/IT, Lovely Professional University 7

You might also like