|
Abstract
Automated Placement and Routing of Cell Matrix Circuits
Dimitri Yatsenko,
Department of Computer Science,
Utah State University, Logan, Utah, USA
|
Cell Matrix is a parallel self-configurable architecture. It is similar to
field programmable gate arrays (FPGAs) in that Cell Matrix elements can be
programmed. Cell Matrix circuits differ from FPGAs because groups of their
elements can dynamically reprogram adjacent elements. Advanced behaviors
arise from this ability, setting Cell Matrix apart from architectures
lacking self-configurability.
This thesis specifies the first Cell Matrix compiler to automatically
generate Cell Matrix circuits from logic specifications. Given a schematic
design, a set of available cells, and an optional seed placement, the Cell
Matrix compiler generates a complete and efficient code to load into Cell
Matrix cells.
The placement algorithm combines self-organizing maps and force-directed
optimization. An efficient gate placement emerges from the collective
action of individual gates following simple parameterized rules. The wire
routing uses an A* shortest-path algorithm. The overall approach is shown
to be highly parallelizable, flexible, and versatile.
|
|