VLSI Design

Research Article

CONTANGO: Integrated Optimization of SoC Clock Networks

Table 5

Results on the ISPD’09 Contest benchmark suite. CLR is reported in ps, capacitance in % of the limit specified in benchmarks, and CPU time in . Best results from the ISPD’09 contest and best results overall are shown in bold. Runtime is dominated by SPICE runs. It was not used for scoring at the ISPD’09 contest and can be improved by using FastSPICE, Arnoldi approximation, and so forth.


	Contango (this work)			NTU			NCTU			U. of Michigan
Benchmark	9/10/2009			3/30/2009			3/30/2009			3/30/2009
	CLR	Cap.		CLR	Cap.	⌚	CLR	Cap.	⌚	CLR	Cap.	⌚

ispd09f11	13.36	99.61	6488	26.71	85.53	14764	22.31	89.90	23358	32.29	73.86	3892
ispd09f12	15.27	99.99	6564	25.73	84.72	13934	22.18	87.86	14992	32.17	73.45	3944
ispd09f21	17.40	96.74	6673	30.54	80.79	14978	19.61	86.65	26420	34.31	74.30	4587
ispd09f22	12.36	97.43	3618	24.51	81.82	7189	16.38	85.01	9432	30.45	70.01	2005
ispd09f31	12.81	98.29	21379	45.07	73.49	40088	212.0	92.38	1.29	51.34	81.53	17333
ispd09f32	17.92	99.24	12895	36.90	80.14	3566	fail	—	—	40.32	77.39	10599
ispd09fnb1	13.40	78.38	778	fail	—	—	fail	—	—	19.84	63.10	477

Average	14.65	95.66	8342	31.57	81.08	15753	58.49	88.36	14841	34.39	73.38	6120
Relative	1.0	1.0	1.0	2.15	0.85	1.89	3.99	0.92	1.78	2.35	0.77	0.73