Liquid Circuits: Automated Dynamic Hardware Acceleration of Compute-Intensive Applications
Lead Research Organisation:
Imperial College London
Department Name: Computing
Abstract
Imagine PCs could morph their hardware to match the programs that are running. After a while, new programs running on a machine would suddenly speed up by up to 10 times because the hardware has adapted to the new task. When running an application for the first time, the microprocessor reads the instructions of the new program and executes them. While executing the program the PC looks for ways to run the program faster. The next time the program is used, the computer uses the experience from the previous runs to do a better job. The process of utilising experience is called dynamic optimisation. In this project, we take a bold step forward and claim that it is possible to build a computer which learns how to execute programs faster by changing it's own hardware to adapt to the problem being solved. The result can be more than 10 times faster computers at the price of a graphics acceleration card such as nVidia or ATI.
Organisations
People |
ORCID iD |
Oskar Mencer (Principal Investigator) |
Publications
Atasu K
(2008)
CHIPS: Custom Hardware Instruction Processor Synthesis
in IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems
Fu H
(2009)
Accelerating Seismic Computations Using Customized Number Representations on FPGAs
in EURASIP Journal on Embedded Systems
Wu Q
(2009)
Architecture of Computing Systems - ARCS 2009