7 Março 2018, 15:26 - Luís Revez
Talk by Prof. Maurício Breternitz
Date: March 14 13:30pm
Location: Room 0.19 (informática II building)
TALK TITLE: Microarchitecture, Computing Systems, High Performance Computing and End-to-End Deep Neural Networks
ABSTRACT: In this three-part presentation I intend to discuss highlights from my experience as a researcher in industrial labs in the USA (at IBM, Motorola, Intel Labs and IBM Research), discussing the microarchitecture computing systems, and related compilation issues.
Then I intend to discuss current and future experience in high performance computing systems leading to Exascale, from system organization to performance analysis and big data applications. Finally I describe current investigation topics which focus on novel applications of end-to-end deep neural networks, focusing on system organization and research opportunities. I am also available to share experiences related to industry and university interaction and collaboration.
BIO: Mauricio is a Research Fellow and Invited Auxiliary Professor in DLS group of ISTAR at ISCTE.
He received a Ph.D. in Computer Engineering from Carnegie-Mellon University, is a Senior Member of the IEEE, holds 48 U.S. patents and has 52 more pending in areas related to compilation, code optimization, binary translation, processor, cache and memory system organization and cryptography, and his publications have been cited more than 1016 times.
Mauricio's academic service include advising multiple Ph.D. internships and thesis committees, organizing ACM/IEEE conferences; service on multiple program committees as well as participation in successful grant submissions to the U.S. Dept. of Energy and European Horizon 2020. Mauricio's research interest focuses on real-life solutions with practical impact, developed at IBM (T.J. Watson and Austin), Motorola, Intel Labs, TimesN Systems (an Austin startup) and most recently at AMD Research in Austin, Texas where he worked on Exascale computing.
His research areas include end-to-end applications of machine learning, and architecture and performance of computing systems. He worked on parallelizing compilers for a research multiprocessor, and for instruction-level parallel VLIW architectures; on binary translation of x86 codes; on IP telephony libraries and on parallelizing database server programs. At Intel, Mauricio conceived and pushed through product deployment innovative microcode compression inventions, now in every product, that had significant impact (estimated upwards of US$18M savings). His current research focuses on novel understanding and application of end-to-end deep neural networks to achieve efficient computation systems.