Generic global placement and floorplanning
Hans Eisenmann, Frank Johannes
Abstract
Hans Eisenmann, Frank Johannes
Abstract
We present a new force directed method for global placement. Besides the well-known wire length dependent forces we use additional forces to reduce cell overlaps and to consider the placement area. Compared to existing approaches, the main advantage is that the algorithm provides increased flexibility and enables a variety of demanding applications. Our algorithm is capable of addressing the problems of global placement, floorplanning, timing minimization and interaction to logic synthesis. Among the considered objective functions are area, timing, congestion and heat distribution. The iterative nature of the algorithm assures that timing requirements are precisely met. While showing similar CPU time requirements it outperforms Gordian by an average of 6 percent and TimberWolf by an average of 8 percent in wire length and yields significantly better timing results.
OpenAlex reports 396 citations for this work. Citation counts describe recorded attention and do not establish research quality.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
We present a new force directed method for global placement. Besides the well-known wire length dependent forces we use additional forces to reduce cell overlaps and to consider the placement area. Compared to existing approaches, the main advantage is that the algorithm provides increased flexibility and enables a variety of demanding applications. Our algorithm is capable of addressing the problems of global placement, floorplanning, timing minimization and interaction to logic synthesis. Among the considered objective functions are area, timing, congestion and heat distribution. The iterative nature of the algorithm assures that timing requirements are precisely met. While showing similar CPU time requirements it outperforms Gordian by an average of 6 percent and TimberWolf by an average of 8 percent in wire length and yields significantly better timing results.
Key concepts: Floorplan, Flexibility (engineering), Computer science, Minification, Placement, Parallel computing, Mathematical optimization, Algorithm