Abstract
Legalizing a mixed‑size layout that holds large hard macro blocks together with millions of standard cells becomes difficult at advanced technology nodes, where mismatched technology grids and zero‑slack boundaries leave narrow room for movement. Prior approaches bound macro movement with one static displacement limit, so a tight limit can leave macros out of alignment and a loose limit can deadlock a discrete engine. Disclosed is a two‑phase method that separates legalization from alignment. A first phase carries hard feasibility rules without an alignment objective and returns a legal coordinate set. A bounds scaling module then sizes a search window for each macro from the misalignment that remains. A second phase prunes non‑overlap constraints between macro pairs that can no longer collide, then aligns macro edges onto shared routing tracks. The method can shorten convergence time and can recover alignment that a fixed limit may leave unresolved.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
S, Vijayavelu; Saluja, Amit; Barik, Surya; and Hazra, Arijit, "Two‑Phase Trust‑Region Constraint Programming with Dynamic Bounds Scaling for Macro Legalization and Alignment", Technical Disclosure Commons, (September 16, 2026)
https://www.tdcommons.org/dpubs_series/11740