Schardl     
Tao B. Schardl
Ph.D. Candidate
Department of Electrical Engineering and Computer Science
MIT CSAIL
"Deterministic Parallel Random-Number Generation,
Science-Based Performance Engineering, and the End of Moore's Law"

Wednesday, March 9, 4:00 PM
Packard Lab room 508

Abstract:   Despite the monetary and human costs of slow code, few programmers today know how to write fast code, because writing fast code is notoriously hard. How, for example, can ordinary programmers cope with the nondeterminism inherent in concurrent execution that makes utilizing the parallel processing cores error prone and difficult? How can they grapple with the ad hoc and unprincipled nature of software performance engineering? Will writing fast code forever be too hard for ordinary programmers to bother doing?

This three-part talk offers insight into each of these questions. I will first dive into the specific problem of generating pseudorandom numbers deterministically in parallel, thereby illustrating some of the algorithm and system techniques that together can enable deterministic parallel programming. I will then discuss how the impending end of Moore's law will increase the value of software performance engineering as an essential skill for ordinary programmers. I will conclude by broadly characterizing my doctoral-thesis work on transforming software performance engineering into a principled and scientific process.

Bio:   Tao B. Schardl is a Ph.D. candidate working with Professor Charles Leiserson in the Supertech group in MIT CSAIL. His work draws from algorithms and systems to develop programming-language technologies, tools, and parallel algorithms that simplify the tasks of writing, reasoning about, and debugging fast code. Mr. Schardl received his Bachelor of Science in Computer Science and Electrical Engineering from MIT in 2009 and his Masters of Engineering in Computer Science and Electrical Engineering from MIT in 2010.

© 2014-2016 Computer Science and Engineering, P.C. Rossin College of Engineering & Applied Science, Lehigh University, Bethlehem PA 18015.