Builds lightweight, continuous software monitoring and analysis to improve program data reference locality, scalability, and power consumption.
|
The Daedalus project has two main goals: (1) to produce tools for quantifying, analyzing, and improving a program's data reference locality and scalability, (2) to build an optimization framework for automatically improving a program's data reference locality and scalability. |
|
|
PPP is a novel scheme for efficiently profiling paths in programs. PPP enables path profiling at extremely low overheads by observing that most consumers of path profiles are usually interested in profiling a small number of paths that are known a priori, even though the number of potential paths in functions can be quite large. |
|
|
Gargoyle is a research project that targets energy-efficient computing by redesigning software and hardware. On the software side we are exploring a framework called Green that enables programmers to take advantage of domain-specific approximation opportunities in a systematic manner while still providing statistical QoS guarantees. On the hardware side we are investigating architectural designs that deliver the performance-per-Joule efficiency of small cores without impacting application quality-of-service robustness, flexibility, and reliability for emerging Internet-scale applications, which increasingly invoke computationally intensive kernels. |
|
|
Software performance is traditionally measured using profiling, which is often too little (only certain inputs are profiled) or too late (to make requisite changes to address the root cause before shipping). The SPEED project attempts to address these limitations by static estimation of symbolic computational complexity of programs. It builds over recent advances in static program analysis, which has traditionally been used for checking correctness as opposed to measuring performance. |
|
|
Doloto is a system that analyzes application workloads and automatically performs code splitting of existing large Web 2.0 applications. After being processed by Doloto, an application will initia lly transfer only the portion of code necessary for application initialization. The rest of the application's code is replaced by short stu bs -- their actual function code is transferred lazily in the background or, at the latest, on-demand on first execution. Since code dow nload is interleaved with application execution, users can start interacting with the Web application much s ooner, without waiting for the code that implements extra, unused features. |
|
|
NinjaWare is a project investigating lightweight continuous monitoring and analysis of software. Our implementation allows us to gather fine-grain temporal information about high frequency events such as program data accesses with very low overhead (< 5%). We are exploring leveraging this infrastructure to build a wide variety of always-on runtime tools ranging from memory leak and data race detectors to program specification/invariant checkers and security monitors. |
|
