AbstractResearchGroup
From Abstract wiki
ABSTRACT @ CMU
The ABSTRACT research group at CMU is the research group of Brandon Lucia. Our group does research at the intersection of computer architecture, computer systems, and programming languages. Our research focuses on improving the programmability, reliability, and efficiency of computing devices and systems. Our work cuts across the layers of the system stack, from the microarchitecture to the application, often touching on everything in between.
Research
Our research focuses on three main themes, ecompassing (1) Intermittent and Energy-harvesting computing, (2) Concurrency and Parallelism, and (3) Heterogeneous and Emerging Computer Architectures.
Intermittent and Energy-harvesting Computing Devices
We are working to define the system stack for intermittent computing devices, like ones that harvest energy from their environment. Our goal is to make these devices programmable, reliable, low-power, and robust to common-case power failures. Making these devices useful requires us to rethink the whole system stack to deal with complexity and reliability, from programming abstractions tolerant of interruptions, down to super-energy-efficient circuits we use to implement microarchitectural features tolerant of byzantine failures.
An Energy-interference-free Hardware/Software Debugger for Intermittent Energy-harvesting Systems
Alexei Colin, Graham Harvey, Brandon Lucia, Alanson Sample ASPLOS 2016
Energy-interference-free System and Toolchain Support for Energy-harvesting Devices
Alexei Colin, Alanson Sample, Brandon Lucia CASES 2015 (short paper)
A Simpler, Safer Programming and Execution Model for Intermittent Systems
Brandon Lucia, Benjamin Ransford PLDI 2015
Concurrency and Parallelism
We are working to develop software techniques and hardware architectural features to make concurrent and parallel computer systems and software correct, reliable, and efficient. We research new architectural features that make programming easier, hardware-software solutions to making debugging automatic, and architecture, compiler, and runtime gadgets that automatically avoid failures in broken programs. Our work leverages approximation, statistical inference, symbolic execution, and dynamic analysis in novel ways to make systems behave better.
Valor: Efficient, Software-Only Region Conflict Exceptions
Swarnendu Biswas (OSU), Minjia Zhang (OSU), Michael D. Bond (OSU), Brandon Lucia OOPSLA 2015
OOPSLA'15 Distinguished Paper Award
OOPSLA'15 Distinguished Artifact Award
Concurrency Debugging with Differential Schedule Projections
Nuno Machado, Luis Rodrigues, Brandon Lucia PLDI 2015
Approximate and Heterogeneous Computing
We are working to define the programming and execution models for the increasingly heterogeneous computing systems of the future, like ones with a mixture of CPUs, GPUs FPGA, CGRAs, and ASICs. Especially exciting topics in this area are approximate computing techniques that trade imprecision for increased performance and programmability. We are investigating how to define a programming, execution, and memory consistency model for heterogeneous and approximate systems. Our research incorporates analytical and empirical modeling and automatic synthesis of software and hardware.
Systems Should Automatically Specialize Code and Data Brandon Lucia and Todd Mytkowicz Approx 2014
People @ CMU
Faculty
PhD Students
Nuno Machado (summer 2015 Visiting Scholar, from IST Lisboa)
Vignesh Balaji
Masters Students
Dhruva Tirumala
Preeti Upendra Murthy
Undergraduates
Marie Bremner
Mark McElwaine (now a junior at CMU)
Graham Harvey (now an Imagineer @ Disney)
High School Interns
Savi Medlang (Now a freshman at Wake Forest College)
External Collaborators
Alanson Sample (Disney Research Pittsburgh)