2019 MIT Center for Computational Engineering Symposium

Stay tuned for more details!

MIT Distinguished Seminar Series in Computational Science and Engineering
Thursday, December 6, 2018 | 12:00 PM* | 3-370

Making Sparse Fast
Saman Amarasinghe
Professor and Associate Department Head, Department of Electrical Engineering and Computer Science
Massachusetts Institute of Technology

Achieving high performance is no easy task. When it comes to program operating on sparse data, where there is very little hardware, language or compiler support, getting high performance is nearly impossible. As important modern applications such as data analytics and simulations operate on sparse data, lack of performance is becoming a critical issue. Achieving high performance was so important from the early days of computing, many researchers have spent their lifetime trying to extract more FLOPS out of critical codes. Hardcore performance engineers try to get to this performance nirvana single handedly without any help from languages, compilers or tools. In this talk, using three examples, TACO and GraphIt, I’ll argue that domain specific languages and compiler technology can take most of the performance optimization burden even in a very difficult domain such as sparse computations. TACO is an optimizing code generator for sparse linear and tensor algebra. TACO introduces a new technique for compiling compound tensor algebra expressions into efficient loops. TACO-generated code has competitive performance to best-in-class hand-written codes for tensor and matrix operations. GraphIt is a Domain Specific Language and compiler for high-performance graph computing. GraphIt separates algorithm, schedule and physical data layout, providing the programmer with the ultimate control over optimization. GraphIt outperforms the state-of-the-art libraries and DLSs up to 2.4× on scale-free graphs and 4.7× on road graphs.

Saman P. Amarasinghe is the Associate Department Head in the Department of Electrical Engineering and Computer Science at Massachusetts Institute of Technology and a member of the Computer Science and Artificial Intelligence Laboratory (CSAIL) where he leads the Commit compiler group. Under Saman's guidance, the Commit group developed the TACO, GraphIt, StreamIt, PetaBricks, StreamJIT, Halide, Simit, and MILK programming languages and compilers, DynamoRIO dynamic instrumentation system, Superword level parallelism for SIMD vectorization, Program Shepherding to protect programs against external attacks, the OpenTuner extendable autotuner, and the Kendo deterministic execution system. His research interests are in discovering novel approaches to improve the performance of modern computer systems without unduly increasing the complexity faced by the end users, application developers, compiler writers, or computer architects. Saman received his BS in Electrical Engineering and Computer Science from Cornell University in 1988, and his MSEE and Ph.D from Stanford University in 1990 and 1997, respectively.

MIT Distinguished Seminar Series in Computational Science and Engineering
Thursday, December 13, 2018 | 12:00 PM* | 3-370

Spectral POD and resolvent analysis of turbulent flows
Tim Colonius
Frank and Ora Lee Marble Professor of Mechanical Engineering
California Institute of Technology

Amongst many available data-driven modal decompositions of utility in fluid mechanics, the frequency-domain version of the proper orthogonal decomposition, which we call spectral POD (SPOD), plays a special role in the analysis of stationary turbulence. While typically requiring more data than the more popular space-only POD and dynamic mode decompositions (DMD), SPOD modes are optimal in expressing structures that evolve coherently in both space and time, and they can be regarded mathematically as optimally-averaged DMD modes. The SPOD spectrum is also related to the resolvent spectrum of the linearized dynamics (the linearized Navier-Stokes equations in this case) and examination of the relationships between the SPOD and resolvent modes yields information about how coherent structures are forced by nonlinear interactions amongst coherent and incoherent turbulence. The theory and algorithms will be demonstrated through examples from the literature as well as a detailed study of turbulence in high-speed jets.

Tim Colonius is the Frank and Ora Lee Marble Professor of Mechanical Engineering at the California Institute of Technology. He received his B.S. from the University of Michigan in 1987 and M.S and Ph.D. in Mechanical Engineering from Stanford University in 1988 and 1994, respectively. He and his research team use numerical simulations to study a range of problems in fluid dynamics, including aeroacoustics, flow control, instabilities, shock waves, and bubble dynamics. Prof. Colonius also investigates medical applications of ultrasound, and is a member of the Medical Engineering faculty at Caltech. He is a Fellow of the American Physical Society and the Acoustical Society of America, and he is Editor-in-Chief of the journal Theoretical and Computational Fluid Dynamics. He was the recipient of the 2018 AIAA Aeroacoustics Award.

*Lunch provided at 11:45

MIT Distinguished Seminar Series in Computational Science and Engineering
Thursday, February 21, 2019 | 12:00 PM | Location TBD

Lek-Heng Lim
Associate Professor, Department of Statistics and the College
University of Chicago

MIT Distinguished Seminar Series in Computational Science and Engineering
Thursday, May 2, 2019 | 12:00 PM | Location TBD

Arnaud Doucet
Professor of Statistics
Oxford University