Programming Multicore Architectures 1 B RAMAMURTHY CSE 651

Programming Multi-core Architectures 1 B. RAMAMURTHY CSE 651 C, B. Ramamurthy 6/16/2017

Moore’s law 2 �Gordon Moore’s original paper. �Hmm. . What is Moore’ s law? �Moore's law is the observation that the number of transistors in a dense integrated circuit doubles approximately every two years. From Wikipedia. . �This also has been interpreted as performance doubling every 18 months. �See the chart in the next slide, again from https: //en. wikipedia. org/wiki/Moore%27 s_law CSE 651 C, B. Ramamurthy 6/16/2017

CSE 651 C, B. Ramamurthy 3 6/16/2017

Addressing the Power Issues 4 �Lower frequency. . Multi-core: divide single core into quad core and multiples of that. �Architectural change �Performance comes from software: make software parallel �Real parallelism for multi-threaded programs �Newer programming models �Difficult to program �APIs such as Open. MP appeared to ease the programming process. CSE 651 C, B. Ramamurthy 6/16/2017

Summary 5 �Single core �Multiple threads: Posix standard for threads: pthreads library �API for Multi-threaded programming for multi-core architectures: Open. MP �API for multiple processor clusters: message passing parallelism: Open. MPI CSE 651 C, B. Ramamurthy 6/16/2017

Activity 1: Automotive Domain 6 � Lets explore any opportunities and use cases for Multicore programming 2. Multithreaded programming 3. Multiple node computing (Cluster computing) 1. CSE 651 C, B. Ramamurthy 6/16/2017