lec5 - CS575ParallelProcessing Lecturefive:Performance

Info iconThis preview shows pages 1–7. Sign up to view the full content.

View Full Document Right Arrow Icon
CS575 Parallel Processing Lecture five: Performance Wim Bohm, Colorado State University Material on Naïve model from: “Speedup vs Efficiency in Parallel Systems” - Eager, Zahorjan and Lazowska IEEE Transactions on Computers, Vol 38 No 3, March 1989 CSU has institution license for IEEE online library Goto  http://ieeexplore.ieee.org  using browser opened from CSU domain machine or with CSU VPN client to download Except as otherwise noted, the content of this presentation is licensed under  the Creative Commons Attribution 2.5 license.
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
CS575 lecture 5 2 Parallel Processing Divide a computation into sub tasks Execute sub tasks in parallel Can all sub tasks run in parallel? NO! Usually there is data dependence between the sub  tasks Benefit? Speedup Cost? More resources Processors, memories, network
Background image of page 2
CS575 lecture 5 3 Notation T i :   time to execute a program with i processors T : time to execute a program with unbounded # processors Speedup:   S(n) =  T 1  / T n Linear speedup: S(n) = k.n Often used in the stricter sense: S(n) = n     Efficiency:  E(n) =  S(n) / n Average utilization of n processors Range? What does E(n) = 1 signify?   Does E(n) = 1 happen a lot in practice? Data dependence, communication, contention
Background image of page 3

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
CS575 lecture 5 4 Bounds on Speedup Achievable bounds Amdahl’s law If fraction f of a program is inherently sequential, the bound on S(n): T 1  = 1 T n  = f + (1-f)/n S(n)   1 / (f + (1-f)/n) Even simpler: S(n) < 1 / f What does this assume, and thus totally ignore?
Background image of page 4
CS575 lecture 5 5 Zahorjan et. al. slightly less naïve Program : Acyclic Directed graph Nodes   tasks,  Edges   precedence relations Strict  A B: B cannot begin until A is finished Fixed set of tasks, no deadlock Machine : n identical processors  Execute each task in one time-step No communication overhead Scheduling : Work conserving Leaves no task idle when processor available
Background image of page 5

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
CS575 lecture 5 6 Program Parallelism How many steps does the  program take? finite resources
Background image of page 6
Image of page 7
This is the end of the preview. Sign up to access the rest of the document.

This note was uploaded on 09/24/2009 for the course CS 525 taught by Professor Rjyosy during the Winter '09 term at Central Mich..

Page1 / 22

lec5 - CS575ParallelProcessing Lecturefive:Performance

This preview shows document pages 1 - 7. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online