image placeholder

To deal with the complexities of modern software, computer scientists are
increasingly reliant on special software tools such as data-race detectors,
memory bounds checkers, and data flight recorders. However, these tools can
make the programs being tested run hundreds to thousands of times slower
than normal. In their paper on “Quantifying the Potential for Program
Analysis Peripherals”, UCSB students Mohit Tiwari and Shashidhar Mysore,
along with their advisor Professor Tim Sherwood, show that significant
performance improvements can be had through a combination of novel software
optimizations and computer peripherals designed specifically to help analyze
complex programs. To flush out the many non-trivial technical issues in
taking such an approach the authors developed a prototype system that maps a
DMA-based analysis engine, sitting on a PCI-mounted FPGA, into the Valgrind
instrumentation framework. Their efforts were recognized as the “Best
Paper” at the 18th Annual International Conference on Parallel Architecture
and Compilation Techniques (PACT 2009). The paper can be found on
http://www.cs.ucsb.edu/~sherwood/publications.html