BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Los_Angeles
X-LIC-LOCATION:America/Los_Angeles
BEGIN:DAYLIGHT
TZOFFSETFROM:-0800
TZOFFSETTO:-0700
TZNAME:PDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0700
TZOFFSETTO:-0800
TZNAME:PST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260730T152640Z
LOCATION:
DTSTART;TZID=America/Los_Angeles:20260728T173800
DTEND;TZID=America/Los_Angeles:20260728T173900
UID:dac_DAC 2026_sess317_RESEARCH1300@linklings.com
SUMMARY:ComPart: Community-Guided Post-Coarsening for High-Quality Hypergr
 aph Partitioning
DESCRIPTION:Yugao Zhu and Zhicheng Guo (Hong Kong University of Science an
 d Technology (HKUST)); Yuchao Wu (The Hong Kong University of Science and 
 Technology); and Mengming Li, Jing Wang, and Zhiyao Xie (Hong Kong Univers
 ity of Science and Technology (HKUST))\n\nHypergraph partitioning is a cri
 tical step in the design of complex embedded systems, essential for optimi
 zing task mapping on heterogeneous MPSoCs and enabling multi-FPGA prototyp
 ing. \nMany existing methods rely on community detection to identify modul
 es with dense internal and sparse external connections, typically utilizin
 g them to constrain the coarsening phase—a widely adopted paradigm. In thi
 s work, we propose ComPart, a generalized framework that integrates divers
 e community detection methods to uncover high-quality clusterings througho
 ut the post-coarsening stages (i.e., initial partitioning and uncoarsening
 ). These discovered clusterings serve as distinct structural guides, enabl
 ing the refinement process to identify superior partitioning solutions. Ou
 r framework offers two key advantages: (1) it establishes a new paradigm t
 hat leverages community structures detected during uncoarsening to escape 
 local optima and explore globally meaningful solution subspaces, transcend
 ing the limitations of standard local refinements; and (2) it flexibly acc
 ommodates both existing and future community detection methods. Furthermor
 e, we theoretically generalize locally-dense decomposition—originally from
  graphs—to the hypergraph domain. We provide the formal extension and nece
 ssary proofs to apply this technique to hypergraphs, marking its first app
 lication in hypergraph partitioning. Specifically, we utilize this rigorou
 sly derived decomposition to guide the initial partitioning phase toward s
 uperior starting points. Experimental results on standard benchmarks demon
 strate that our method consistently outperforms state-of-the-art methods i
 n solution quality.\n\nTrack: EDA7-II. Physical Design and Verification, D
 esign\n\nSession Chairs: Christian Pilato (Politecnico di Milano); Mohsen 
 Imani (University of California, Irvine); Flavio Ponzina (San Diego State 
 University); Amin Farshidi (Cadence Design Systems, Inc.); Yonggan Fu (Geo
 rgia Institute of Technology, NVIDIA); Dimitrios Tsaras (Huawei, Hong Kong
  University of Science and Technology (HKUST)); Anh Nguyen (Fujitsu Limite
 d); Ruiyang Qin (Villanova University); Jiliang Zhang (Hunan University); 
 Meng Li (Peking University); and Rassul Bairamkulov (Advanced Micro Device
 s (AMD))\n\n
END:VEVENT
END:VCALENDAR