The Circuit Partitioning Page

Previously, the purpose of this page has been solely to make widely used partitioning benchmarks available in a (relatively) simple format. We are now expanding this page to include partitioning codes, results, and even partitioning solutions for this format. Any contributions or suggestions should be sent to Chuck Alpert.

Partitioning Codes

Fiduccia-Mattheyses Code

Algorithms for Partitioning Multi-dimensional Data Sets

Algorithms for Ordering and Partitioning Hypergraphs

Ali Dasdan's Iterative and Annealing Partitioning Package

Karypis and Kumar's Multilevel Metis Packages

Partitioning Slot in the MARCO/GSRC bookshelf

Partitioning Benchmarks in Hypergraph Format

Benchmark File Format

The following are small benchmarks or test cases. They do not represent real circuits and should be used for testing and debugging purposes only.

Clique (16)
qgr8 (12)
3k5 (15)
IC67 (67)
IC116 (116)
IC151 (151)

The following benchmarks were obtained from Professor C.K. Cheng at UCSD:

primary1 (833)
primary2 (3014)
test02 (1663)
test03 (1607)
test04 (1515)
test05 (2595)
test06 (1752)
bm1 (882)
19ks (2844)

The following netlists were derived using the vp2pr translator (obtained from TU Munich) that takes a vpnr file and outputs a proud file. The proud file was then translated into this netlist format; for these netlists all pads are treated as cells, so there are no "p"'s in these files.

fract (125)
balu (801)
struct (1952)
biomed (6514)
s9234 (5866)
s13207 (8772)
s15850 (10470)
industry2 (12637)
s35932 (18148)
s38584 (20995)
s38417 (23849)
golem3 (103048)

The following large benchmarks were obtained from Lars Hagen (; I believe they were directly derived from either .yal or .vpnr format.

industry3 (15433)
avq.small (21918)
avq.large (25178)

Benchmark Pointers and Other Partitioning Info

Partitioning Slot in the MARCO/GSRC bookshelf
CAD Benchmarking Laboratory
Partitioning Benchmarks in PROUD Format (TU Munich) ,
Berkeley Partitioning Lectures (part of CS267)
UCLA VLSI CAD Laboratory

For questions or comments, contact Charles Alpert at

Charles Alpert's Home Page