c31 lecture 16

E shaw research david shaw nice success story 16

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: period Designed and built by D.E. Shaw Research David Shaw: nice success story! 16 Physics (source: SciDAC) 17 Simulation method (source: SciDAC) 18 Simulation vs. prediction (source: SciDAC) 19 Exascale power (source: SciDAC) 20 Top 500 supercomputers (source: SciDAC) 21 Exascale roadmap (source: SciDAC) 22 A “standard” computer Grabbed from arstechnica.com 23 The CPU-Memory Gap http://blogs.sun.com/toddjobson/entry/processors_and_performance_ 24 Hardware prospective (source: SciDAC 2010) 25 Exponential growth in proc. count (SciDAC) 26 The Energy challenge (source: SciDAC) 27 Great Reality #1 Computers do more than execute programs They need to get data in and out I/O system critical to program reliability and performance They communicate with each other over networks Many system-level issues arise in presence of network 28 AMD Bulldozer Architecture Grabbed from wikipedia 29 Great Reality #2 Parallelism is key to future performance Power-Efficient Performance comes from CMPs Chip Multiprocessors (CMPs) Multiple processor cores integrated onto a single silicon die or multichip module Examples? • Intel’s Core iX • Sun’s Niagara 2 • IBM/Sony’s Cell • nVidia’s Tesla and Fermi Programmers need to learn how to exploit parallelism in their applications 30 Memory Hierarchy Grabbed from http://ktutorialspure-c.blogspot.com 31 Great Reality #4 Memory Matters Memory is not unbounded It must be allocated and managed Many applications are memory dominated Memory performance is not uniform Cache and virtual memory effects can greatly affect program performance Adapting program to characteristics of memory system can lead to major speed improvements Memory referencing bugs especially pernicious Effects are distant in both time and space 32 Ariane 5 Exploded 37 seconds after liftoff Cargo worth $500 million Why Computed horizontal velocity as floating point number Converted to 16-bit integer Worked OK for Ariane 4 33 Reality check #3: Floating Point The magic of infinity, 1: Code: double val1 = tan(MATH...
View Full Document

Ask a homework question - tutors are online