Virtual Pipelines by Agrawal and Sherwood is nominated for best paper at Micro
In thier paper, an architectural technique that efficiently supports high-bandwidth, uniform latency memory accesses, and high-confidence throughput even under adversarial conditions is proposed. The technique, Virtual pipelining. provides a simple to analyze programing model of a deep pipeline (deterministic latencies) with a completely different physical implementation (a memory system with banks and probabilistic mapping). This allows designers to effectively decouple the analysis of their algorithms and data structures from the analysis of the memory buses and banks. Unlike specialized hardware customized for a specific data-plane algorithm, the system makes no assumption about the memory access patterns. In thier paper, the authors present a mathematical argument for the system's ability to provably provide bandwidth with high confidence and demonstrate its functionality and area overhead through a synthesizable design. Even though the scheme is general purpose to support new applications such as packet reassembly, it outperforms the state of the art in specialized packet buffering architectures.
Introspective 3D Chips by Shashi Mysore, Banit Agrawal, and Prof. Sherwood is IEEE Micro Top Pick
Each year, a panel of 30 senior computer architects chooses 10 of the years most significant research publications for publication in a special issue of IEEE Micro. For the 3rd Year in a row, a paper from UCSB Computer Science is present: Introspective 3D Chips by Shashi Mysore, Banit Agrawal, and Sheng-Chih Lin, Navin Srivastava, Kaustav Banerjee, and Timothy Sherwood from ASPLOS 2006. To deal with the complexity of modern systems, software developers are increasingly dependent on specialized development tools such as security profilers, memory leak identifiers, data flight recorders, and dynamic type analysis. In their paper, the authors argue that a new way to attack this problem is with the addition of specialized analysis hardware, literally stacked on top of the processor die using 3D-integration technology. This provides a modular snap-on functionality that could be included with developer systems, while keeping the cost impact on consumer systems to minimum.
Paper by Mysore, Agrawal, Sherwood, Shrivastava, and Suri wins best paper award
The paper titled Profiling over Adaptive Ranges received the best paper award at CGO 06 (4th Annual ACM International Symposium on Code Generation and Optimization), which was held in New York during March 26-29. The paper describes a new geometry-based scheme to summarize the huge number of events processed by a modern computer system. The compact summary, called RAP, adaptively and dynamically zooms onto event ranges of interest, thus creating a profile of the program behavior which can then be used for processor optimization.
Tim Sherwood receives early Career award from the National Science Foundation
Tim Sherwood, an Assistant Professor in Computer Science, received the early Career award from the National Science Foundation to fund his research on high speed architectures for online security analysis. The research focus is in building specialized computer processors that are engineered to sort through suspicious packets, and developing new algorithms for hardware string matching.
BitSplit Architecture for Intrusion Detection is IEEE Micro Top Pick
Each year, a panel of 30 senior computer architects chooses 10 of the years most significant research publications for publication in a special issue of IEEE Micro. In this work, a novel high-speed tile-based packet scan technique is described that can search for strings and perform restricted pattern matching. To build an efficient system, the technique converts the large database of search strings into a language, and then rips the language into a set of sub-languages each of which describes a bit-slice of the original language. A recognizer for each of these sub-languages can then be loaded into a simple memory tile and run in parallel. While this sounds somewhat theoretical, a prototype has been implemented, demonstrated it's usefulness through the development of a custom rule compiler, and formally proved both the correctness and efficiency of our approach.