CS 6143 COMPUTER ARCHITECTURE II SPRING 2014 ACM Principles and Practice of Parallel Programming, PPoPP, 2006 Panel Presentations Parallel Processing is Mainstream : Multi-core Processor Chips
Handout 11CS 6143 Spring 2014 Page 2 ACM PPoPP 2006 Panel Presentations
Handout 11CS 6143 Spring 2014 Page 3 ACM PPoPP 2006 Panel Presentations
Handout 11CS 6143 Spring 2014 Page 4 ACM PPoPP 2006 Panel Presentations
Handout 11CS 6143 Spring 2014 Page 5 ACM PPoPP 2006 Panel Presentations
Handout 11CS 6143 Spring 2014 Page 6 ACM PPoPP 2006 Panel Presentations
Handout 11CS 6143 Spring 2014 Page 7 ACM PPoPP 2006 Panel Presentations
Handout 11CS 6143 Spring 2014 Page 8 ACM PPoPP 2006 Panel Presentations
Handout 11CS 6143 Spring 2014 Page 9 ACM PPoPP 2006 Panel Presentations
Handout 11CS 6143 Spring 2014 Page 10 ACM PPoPP 2006 Panel Presentations
Handout 11CS 6143 Spring 2014 Page 11 ACM PPoPP 2006 Panel Presentations
Handout 11CS 6143 Spring 2014 Page 12 ACM PPoPP 2006 Panel Presentations
Handout 11CS 6143 Spring 2014 Page 13 ACM PPoPP 2006 Panel Presentations
Handout 11CS 6143 Spring 2014 Page 14 ACM PPoPP 2006 Panel Presentations
Handout 11CS 6143 Spring 2014 Page 15 ACM PPoPP 2006 Panel Presentations
Handout 11CS 6143 Spring 2014 Page 16 ACM PPoPP 2006 Panel Presentations
Handout 11CS 6143 Spring 2014 Page 17 ACM PPoPP 2006 Panel Presentations
Handout 11CS 6143 Spring 2014 Page 18 ACM PPoPP 2006 Panel Presentations
Handout 11CS 6143 Spring 2014 Page 19 ACM PPoPP 2006 Panel Presentations
Handout 11CS 6143 Spring 2014 Page 20 ACM PPoPP 2006 Panel Presentations
Handout 11CS 6143 Spring 2014 Page 21 ACM PPoPP 2006 Panel Presentations
Handout 11CS 6143 Spring 2014 Page 22 ACM PPoPP 2006 Panel Presentations
Handout 11CS 6143 Spring 2014 Page 23 ACM PPoPP 2006 Panel Presentations
Handout 11CS 6143 Spring 2014 Page 24 ACM PPoPP 2006 Panel Presentations
Handout 11CS 6143 Spring 2014 Page 25 ACM PPoPP 2006 Panel Presentations
Handout 11CS 6143 Spring 2014 Page 26 ACM PPoPP 2006 Panel Presentations
Handout 11CS 6143 Spring 2014 Page 27 ACM PPoPP 2006 Panel Presentations
Handout 11CS 6143 Spring 2014 Page 28 ACM PPoPP 2006 Panel Presentations
Handout 11CS 6143 Spring 2014 Page 29 ACM PPoPP 2006 Panel Presentations
Handout 11CS 6143 Spring 2014 Page 30 ACM PPoPP 2006 Panel Presentations
Handout 11CS 6143 Spring 2014 Page 31 ACM PPoPP 2006 Panel Presentations
Handout 11CS 6143 Spring 2014 Page 32 ACM PPoPP 2006 Panel Presentations
Handout 11CS 6143 Spring 2014 Page 33 ACM PPoPP 2006 Panel Presentations
Handout 11CS 6143 Spring 2014 Page 34 ACM PPoPP 2006 Panel Presentations
Handout 11CS 6143 Spring 2014 Page 35 ACM PPoPP 2006 Panel Presentations
Handout 11CS 6143 Spring 2014 Page 36 ACM PPoPP 2006 Panel Presentations
Handout 11CS 6143 Spring 2014 Page 37 ACM PPoPP 2006 Panel Presentations
Handout 11CS 6143 Spring 2014 Page 38 ACM PPoPP 2006 Panel Presentations
Handout 11CS 6143 Spring 2014 Page 39 ACM PPoPP 2006 Panel Presentations
Handout 11CS 6143 Spring 2014 Page 40 ACM PPoPP 2006 Panel Presentations
Handout 11CS 6143 Spring 2014 Page 41 ACM PPoPP 2006 Panel Presentations
Handout 11CS 6143 Spring 2014 Page 42 ACM PPoPP 2006 Panel Presentations
Handout 11CS 6143 Spring 2014 Page 43 ACM PPoPP 2006 Panel Presentations
Handout 11CS 6143 Spring 2014 Page 44 ACM PPoPP 2006 Panel Presentations
Handout 11CS 6143 Spring 2014 Page 45 ACM PPoPP 2006 Panel Presentations
Handout 11CS 6143 Spring 2014 Page 46 ACM PPoPP 2006 Panel Presentations
Handout 11CS 6143 Spring 2014 Page 47 ACM PPoPP 2006 Panel Presentations
Handout 11CS 6143 Spring 2014 Page 48 How can parallel processing be mainstream ? Slide 19 by Luddy Harrison Rules of composition are needed so that There is independence between software components and their deployment Composing software components ? In hardware this is done now ! How can composing be implemented in software ? Communication computation ? A new computational method ? A new hardware technology leading to a new idea ? All of the above ? Is composing enough ? Parallel processing is relatively simple now with 4 to 10 cores ! But, memory wall, soft errors, defective chips and other unforeseen problems can make parallel processing harder On the other hand, programmable chips, nanotechnology, bio- inspired computing paradigms and other unforeseen new technologies can help parallel processing Professors Comments