Global Sources
EE Times-Asia
Stay in touch with EE Times Asia
EE Times-Asia > Controls/MCUs

IMEC rethinks reconfigurability approach

Posted: 07 Apr 2004 ?? ?Print Version ?Bookmark and Share

Keywords:imec? long instruction word? vliw? mobile?

The IMEC research center is blending concepts in very long instruction word processing with emerging ideas in reconfigurable computing arrays to form a processing cluster that is at once powerful, configurable, and low enough in power for use in mobile applications.

Along with great energy efficiency - a trademark of the Interuniversity Microelectronics Center -the architecture is designed to be fully programmable in C, eliminating the need to hand-code tasks for the processing array.

In IMEC's novel, hybrid architecture, each of the functional units of a relatively conventional very long instruction word (VLIW) processor becomes, in effect, an element in the top row of a heterogeneous array of processing elements. In this way the array is tightly coupled to the VLIW machine, sharing instruction-launch logic, register files and even the function units.

"In practice, at least 90 percent of the code in an application will run on the VLIW engine. It will extract what instruction-level parallelism exists in that code," said Rudy Lauwereins, VP for design technology at IMEC and a professor at the Catholic University of Leuven. "The critical 10 percent of the code - which usually accounts for 70 to 90 percent of the execution time - will be compiled for the coarse-grained hardware array, where it can be accelerated by an order of magnitude." The result can be a huge reduction in the number of machine cycles needed to execute an algorithm. That, in turn, can translate into lower clock frequency and lower power, into higher bandwidth, or into some felicitous combination of the two.

In a simulated implementation of the MPEG-2 decoder in Mediabench, the IMEC flow separated the roughly 10,000 lines of C into two groups: about 97 percent of the code that was executed after optimization on the VLIW engine, and 16 loops that were compiled to the coarse-grained array.

- Ron Wilson

EE Times

Article Comments - IMEC rethinks reconfigurability appr...
*? You can enter [0] more charecters.
*Verify code:


Visit Asia Webinars to learn about the latest in technology and get practical design tips.

Back to Top