U.S. Department of Energy

Pacific Northwest National Laboratory

Field Programmable Gate Array

From DIC

Jump to: navigation, search

Research at PNNL on how to address scientific streaming applications using hybrid CPU/FPGA architectures demonstrates that these architectures can deliver much higher performance than traditional computer systems, enabling the solution of larger scientific problems.

Many fields in experimental science are moving towards high-throughput methodologies where scientific instruments can process large numbers of samples and produce corresponding massive amounts of data. It is becoming increasingly important to perform online processing of the data as it is being produced by the instrument, without having to wait for data to be captured and then processed offline. An effective approach to perform this kind of streaming, online processing is to use FPGA boards directly attached to the instrument's digital data capture ports.

Developing and testing FPGA designs that will be processing live streaming data from an instrument requires a significant design, development and testing effort. It is also quite difficult to predict how much area a particular processing design will occupy and what its performance will be. Hybrid CPU/FPGA HPC systems are a suitable platform to develop and prototype processing algorithms for FPGAs that can be tested under more flexible software control while still having to deal with most of the limitations (and benefits) of a realistic hardware platform.

File:Simulating data processing for an Advanced Ion Mobility.pdf

Article Title: Field Programmable Gate Array (FPGA)

Article Added: 2010/08/17

Category(s): Hybrid Architectures

Last Update: 13 July 2011 | Pacific Northwest National Laboratory