systems-lunch

UMass systems lunch

The UMass CS Systems Lunch talk series (organized by Emery Berger) is Fridays at 12:00pm Eastern in room A311 (LGRC) on your laptop anywhere on the planet! Everyone interested in systems is welcome to attend. Systems Lunch is a great opportunity to hear exciting talks by visitors as well as to learn about projects going on in the College.

All talks will be streamed on YouTube Live but participation in Q&A requires registration (which will get you on the Systems Lunch Slack). To register, fill out this Google form.

Fall 2020 - Spring 2021

October 2, 2020

Adrian Sampson (Cornell)
Languages and Compilers for Accelerator Design
video
We need to make it easier to design custom accelerators, especially for reconfigurable hardware (i.e., FPGAs). The current mainstream options are hardware description languages (HDLs), which are low-level languages that make it feel like you’re wiring up a circuit by hand, and high-level synthesis (HLS) tools, which compile legacy software languages like C or C++ to an HDL. The thesis of this talk is that better alternatives are possible. We find empirically (and unsurprisingly) that the semantic chasm between C++ and hardware circuits comes with myriad correctness and performance pitfalls. We advocate instead for more research on accelerator design languages: programming models that maintain computational semantics while not attempting to shield developers from thinking about hardware. Our lab is designing a programming language, Dahlia, that uses a type system to restrict HLS programs to a subset with predictable semantics and performance. Beyond Dahlia, we are designing an open-source compiler infrastructure to support the construction of new accelerator design languages.
Bio: Adrian Sampson is an assistant professor in the computer science department at Cornell. He works on programming languages, computer architecture, and the abstractions that separate them. He previously worked on approximate computing, the idea that we should allow machines to expose errors to some kinds of applications as a trade-off for computational efficiency. He sees approximate computing as an instance of a broader breakdown of airtight distinctions between hardware and software concerns.

October 9, 2020

Hila Peleg (UCSD)
Programmer Tools with Program Synthesis
video
Program synthesis is the problem of generating a program to satisfy a specification of user intent. Since these specifications are usually partial, this means searching a space of candidate programs for one that exhibits the desired behavior. A lion’s share of the work on program synthesis focuses on new ways to perform the search, but allowing the user to communicate intent remains a major challenge. Synthesis tools are often designed either with no particular group of users in mind, or for domain-specific users such as Excel users or data scientists. We wish to leverage synthesis to create better tools for developers, which means we should narrow our focus and our understanding of our users. Focusing on any subgroup of users allows making assumptions on both the input the user can generate for the synthesizer and the output they can consume. In the case of developers, concepts that are part of the programmer’s life such as code review and read-eval-print loops (REPL) can be leveraged for interactions with a synthesizer. This talk describes two synthesis-based tools that both leverage and cater to programmers as their users. However, designing for the user can put the interface of the synthesizer at odds with state of the art synthesis techniques. Synthesis is, at best, a computationally hard problem, and many existing program synthesis tools and techniques were designed around the synthesizer and its internals, their the user interface dictated by the needs of the synthesizer. We therefore demonstrate the process of concurrently building both the synthesizer and its intended user-facing tool as a solution to this problem.
Bio: Hila Peleg is a postdoctoral researcher in the CSE Department at University of California, San Diego, advised by Nadia Polikarpova. She recieved her PhD at the Technion, Israel. She also holds a degree in literature. Her research explores the way program synthesis can be transformed into tools for developers.

October 16, 2020

Jonathan Ragan-Kelley (MIT)
Organizing Computation for High-Performance Graphics and Visual Computing
video
In the face of declining returns to Moore’s law, future visual computing applications—from photorealistic real-time rendering, to 4D light field cameras, to pervasive sensing with deep learning—still demand orders of magnitude more computation than we currently have. From data centers to mobile devices, performance and energy scaling is limited by locality (the distance over which data has to move, e.g., from nearby caches, far away main memory, or across networks) and parallelism. Because of this, I argue that we should think of the performance and efficiency of an application as determined not just by the algorithm and the hardware on which it runs, but critically also by the organization of its computations and data. For algorithms with the same complexity—even the exact same set of arithmetic operations—the order and granularity of execution and placement of data can easily change performance by an order of magnitude because of locality and parallelism. To extract the full potential of our machines, we must treat the organization of computation as a first-class concern, while working across all levels, from algorithms and data structures, to programming languages, to hardware.
This talk will present facets of this philosophy in systems I have built for image processing, 3D graphics, and machine learning. I will show that, for the data-parallel pipelines common in these data-intensive applications, the possible organizations of computations and data, and the effect they have on performance, are driven by the fundamental dependencies in a given problem. Then I will show how, by exploiting domain knowledge to define structured spaces of possible organizations and dependencies, we can enable radically simpler high-performance programs, smarter compilers, and more efficient hardware. Finally, I will show how we use these structured spaces to unlock the power of machine learning for optimizing systems.
Bio: Jonathan Ragan-Kelley is Esther and Harold E. Edgerton Assistant Professor of Electrical Engineering & Computer Science at MIT, and Assistant Professor of Electrical Engineering & Computer Science at UC Berkeley. He was previously a postdoc in computer science at Stanford, where he worked with Pat Hanrahan, and spent a year as a visiting researcher at Google. His work focuses on high-efficiency computer graphics, at the intersection of graphics with systems, architecture, and compilers. He did his PhD with Frédo Durand and Saman Amarasinghe at MIT CSAIL, developing the Halide language together with Andrew Adams, has worked at three major GPU vendors in architecture, compilers, and research, and built a real-time preview system for the special effects industry in collaboration with Industrial Light & Magic.

October 23, 2020

Mike Hicks (Maryland)
A Verified Optimizer for Quantum Circuits
video
(No specific background on quantum computing will be needed to understand this talk.)
Programming quantum computers will be challenging, at least in the near term. Qubits will be scarce, and gate pipelines will need to be short to prevent decoherence. Fortunately, optimizing compilers can transform a source algorithm to work with fewer resources. Unfortunately, the very factors that motivate optimizing quantum compilers make it difficult to test their correctness. Formal methods can help. I will present VOQC, the first fully verified compiler for quantum circuits, written using the Coq proof assistant. Quantum circuits are expressed as programs in a simple, low-level language called SQIR, which is deeply embedded in Coq. Optimizations and other transformations are expressed as Coq functions, which are proved correct with respect to a semantics of SQIR programs. SQIR uses a semantics of matrices of complex numbers, which is the standard for quantum computation, but treats matrices symbolically in order to reason about programs that use an arbitrary number of quantum bits. SQIR’s careful design and our provided automation make it easy to write and verify a broad range of optimizations in VOQC, and even allow us to verify correctness properties of interesting source SQIR programs. This is joint work with Kesha Hietala, Robert Rand, Shih-Han Hung, Xiaodi Wu, and Michael Hicks, all at UMD.
Bio: Michael W. Hicks is a Professor in the Computer Science Department at the University of Maryland, where he has been since 2002, and the CTO of Correct Computation, Inc. (CCI), a role he has held since late 2018. His research interests range over topics in programming languages, computer security, systems, and software engineering. He was the first Director of the University of Maryland’s Cybersecurity Center (MC2), and was the elected Chair of ACM SIGPLAN, the Special Interest Group on Programming Languages; he now edits its blog, PL Perspectives, at https://blog.sigplan.org. In ongoing projects he is collaborating with the Checked C project to develop techniques for securing legacy C code; exploring synergies between cryptography and programming languages; developing techniques for better random (fuzz) testing and probabilistic reasoning; and designing high-assurance tools and languages for quantum computing.

October 30, 2020

James Mickens (Harvard)
Oblique: Accelerating Page Loads Using Symbolic Execution
video
Mobile devices are often stuck behind high-latency links. Unfortunately for mobile browsers, latency (not bandwidth) is often the key influence on page load time. Proxy-based web accelerators hide last-mile latency by analyzing a page’s content, and informing clients about useful objects to prefetch. However, most accelerators require content providers to divulge cleartext HTTPS data to third-party analysis servers. Acceleration systems can be installed on first-party web servers, avoiding the violation of end-to-end TLS security; however, due to the administrative overhead (and additional VM costs) associated with running an accelerator, many first-party content providers would prefer to outsource the acceleration work—if outsourcing were secure. In this talk, I introduce Oblique, a third-party web accelerator which enables secure outsourcing of page analysis. Oblique symbolically executes the client-side of a page load, generating a prefetch list of symbolic URLs. Each symbolic URL describes a URL that a client browser would fetch, given user-specific values for cookies, the User-Agent string, and other sensitive variables. Those sensitive values are never revealed to Oblique’s analysis server. Instead, during a real page load, the user’s browser concretizes URLs by reading sensitive local state; the browser can then prefetch the associated objects. Experiments involving real sites demonstrate that Oblique preserves TLS integrity while providing better performance than state-of-the-art accelerators. For popular sites, Oblique is also financially cheaper in terms of VM costs.
Bio: James Mickens is a professor of computer science at Harvard University. In the modern era, he spends most of his time trying to avoid unnecessary Zoom meetings. In the pre-modern era, he spent most of his time trying to avoid unnecessary in-person meetings. James Mickens’ spirit animal is a Pokemon that cannot attend meetings. Society will eventually realize his genius and give him a Nobel Peace Prize. James Mickens will not attend the award meeting.

November 13, 2020

Tianqi Chen (CMU)
TVM: An automated deep learning compiler
video
Data, models, and computing are the three pillars that enable machine learning to solve real-world problems at scale. Making progress on these three domains requires not only disruptive algorithmic advances but also systems innovations that can continue to squeeze more efficiency out of modern hardware. Learning systems are in the center of every intelligent application nowadays. However, the ever-growing demand for applications and hardware specialization creates a huge engineering burden for these systems, most of which rely on heuristics or manual optimization. In this talk, I will present a new approach that uses machine learning to automate system optimizations. I will describe our approach in the context of deep learning deployment problems. I will first discuss how to design invariant representations that can lead to transferable statistical cost models, and apply these representations to optimize tensor programs used in deep learning applications. I will then describe the system improvements we made to enable diverse hardware backends. TVM, our end-to-end system, delivers performance across hardware back-ends that are competitive with state-of-the-art, hand-tuned deep learning frameworks.
Bio: Tianqi Chen is currently an Assistant Professor at the Machine Learning Department and Computer Science Department of Carnegie Mellon University. He received his PhD. from the Paul G. Allen School of Computer Science & Engineering at the University of Washington, working with Carlos Guestrin on the intersection of machine learning and systems. He has created three major learning systems that are widely adopted: XGBoost, TVM, and MXNet (co-creator).

November 20, 2020

Emin Gün Sirer, Cornell / Ava Labs
Open and Programmable Finance with Avalanche
video
Avalanche is an open-source platform for launching decentralized applications and enterprise blockchain deployments in one interoperable, highly-scalable ecosystem. This talk will focus on the Avalanche network and the potential for decentralized and institutional finance applications built upon it.
Bio: Emin Gun Sirer is a professor of computer science at Cornell University, founder of Ava Labs, co-founder of bloXroute Inc, and co-director of the Initiative for Cryptocurrencies and Smart Contracts (IC3). Among other things, Sirer is known for having implemented the first currency that used proof of work to mint coins, for selfish mining, for characterizing the scale and centralization of existing cryptocurrencies, as well as having proposed the leading protocols for on-chain and off-chain scaling. Of all his collaborations, he is proudest of his contributions to the John Oliver Show’s segment on cryptocurrencies.

Previous systems lunch talks

Fall 2019

Fall 2017 - Spring 2018

Fall 2016

Fall 2015

Fall 2014 - Spring 2015

Fall 2013 - Spring 2014

Fall 2012 - Spring 2013

Fall 2011 - Spring 2012

Fall 2010 - Spring 2011

Fall 2009 - Spring 2010

Spring 2007