|
For Full-Text PDF, please login, if you are a member of IEICE,
or go to Pay Per View on menu list, if you are a nonmember of IEICE.
|
Simultaneous Placement and Global Routing for Transport-Processing FPGA Layout
Nozumu TOGAWA Masao SATO Tatsuo OHTSUKI
Publication
IEICE TRANSACTIONS on Fundamentals of Electronics, Communications and Computer Sciences
Vol.E79-A
No.12
pp.2140-2150 Publication Date: 1996/12/25 Online ISSN:
DOI: Print ISSN: 0916-8508 Type of Manuscript: Special Section PAPER (Special Section on VLSI Design and CAD Algorithms) Category: Keyword: FPGA, transport processing, layout design, placement and routing network flow,
Full Text: PDF(878.9KB)>>
Summary:
Transport-processing FPGAs have been proposed for flexible telecommunication systems. Since those FPGAs have finer granularity of logic functions to implement circuits on them, the amount of routing resources tends to increase. In order to keep routing congstion small, it is necessary to execute placement and routing simultaneously. This paper proposes a simultaneous placement and global routing algorithm for transport-processing FPGAs whose primary objective is minimizing routing congestion. The algorithm is based on hierarchical bipartition of layout regions and sets of LUTs (Look Up Tables) to be placed. It achieves bipartitioning which leads to small routing congestion by applying a network flow technique to it and computing a maximum flow and a minimum cut. If there exist connections between bipartitioned LUT sets, pairs of pseudo-terminals are introduced to preserve the connections. A sequence of pseudo-terminals represents a global route of each net. As a result, both placement of LUTs and global routing are determined when hierarchical bipartitioning procedures are finished. The proposed algorithm has been implemented and applied to practical transport-processing circuits. The experimental results demonstrate that it decreases routing congestion by an average of 37% compared with a conventional algorithm and achieves 100% routing for the circuits for which the conventional algorithm causes unrouted nets.
|
|
|