Inventor(s)

Abstract

Clock tree synthesis for a system on a chip can distribute a clock signal through an H-tree that feeds several tap regions.  Balanced timing calls for each rectilinear clock region to carry a comparable share of a load product formed from total distance and sink capacitance.  Reaching that balance over a non-binary grid has historically consumed hours of manual trial and error, in part because live queries against a physical design database respond slowly.  To address these limitations disclosed is a method that caches sink coordinates and capacitance values in memory, anchors a two-by-two base grid, and injects supplemental boundary lines to reach a non-binary tile grid.  Adaptive block coordinate descent then refines the shared seams over a cascading resolution schedule.  The method yields gap-free, overlap-free tiles with reduced load variance in seconds.

Creative Commons License

Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.

Share

COinS