 |
Rodolfo's Project Page - RISE Summer 2006 |
 |
Intern: Rodolfo Neuber, Computer Engineering,
UC Santa Barbara
Mentor: Tim Sherwood
Faculty Supervisor: Tim Sherwood
Department: Computer Science |
BUILDING A RANGE ADAPTIVE PROFILING PROCESSOR
Current methods of programming analysis are limited to value-prediction, code
optimization, and behavioral analysis, among other things; however, very little
analysis focuses on identifying performance bottlenecks. Earlier this year, a
new method of programming analysis was created to indicate where a program text
can be modified to directly increase its performance and possibly remove
bottlenecks. This new method of analysis profiles entire sections of the
program text’s assembly code by keeping track of how many times an instruction
within those sections has been executed. The profiling works by grouping less
frequently executed instructions into larger sections, and by separating more
frequently executed instructions into smaller sections. Currently this analysis
is implemented in software and is limited to analyzing files after a program has
been executed. This paper details how this new method of programming analysis
can be implemented as a processor using a Field Programmable Gate Array (FPGA).
To ensure that the processor design is accurate, a comparison of its ability to
profile will be made against its software counterpart. Results of the profiling
will also show that the processor can handle the profiling in real time without
using resources on the machine. Nafion membranes recast from water-based
solution had higher potential voltages than commercially available Nafion of
comparable thickness.
Return to the RISE 2006 project list
