## Non-tree routing for reliability and yield improvement (2002)

### Cached

### Download Links

- [vlsicad.ucsd.edu]
- [www.cs.york.ac.uk]
- [www.engr.uconn.edu]
- [mesl.ucsd.edu]
- DBLP

### Other Repositories/Bibliography

Venue: | in Proceedings of the IEEE/ACM International Conference on Computer-Aided Design |

Citations: | 9 - 0 self |

### BibTeX

@INPROCEEDINGS{Kahng02non-treerouting,

author = {Andrew B. Kahng and Bao Liu and Ion I. Măndoiu},

title = {Non-tree routing for reliability and yield improvement},

booktitle = {in Proceedings of the IEEE/ACM International Conference on Computer-Aided Design},

year = {2002},

pages = {260--266},

publisher = {ACM Press}

}

### OpenURL

### Abstract

We propose to introduce redundant interconnects for manufacturing yield and reliability improvement. By introducing redundant interconnects, the potential for open faults is reduced at the cost of increased potential for short faults; overall, manufacturing yield and fault tolerance can be improved. We focus on a post-processing, tree augmentation approach which can be easily integrated in current physical design flows. Our contributions are as follows: We formulate the problem as a variant of the classical 2-edge-connectivity augmentation problem in which we take into account such practical issues as wirelength increase budget, routing obstacles, and use of Steiner points. We show that an optimum solution can always be found on the Hanan grid defined by the terminals and the corners of the feasible routing region. We give a compact integer program formulation which is solved in practical runtime by the commercial optimization package CPLEX for nets with up to 100 terminals. We give a well-scaling greedy algorithm which has practical runtime up to 1,000 terminals, and comes on the average within 1-2 % of the optimum computed by CPLEX. We give a comprehensive experimental study comparing the solution quality and runtime of our methods with the best methods reported in the literature for the related 2-edge-connectivity augmentation problem, including a sophisticated heuristic based on minimum-weight branchings [11] and a recent genetic algorithm [17]. Experiments on randomly generated and industry testcases show that our greedy augmentation method achieves significant increase in reliability (as measured by the percentage of biconnected tree edges) with very small increase in wirelength. For example, on 1,000 terminal nets the average percentage of biconnected tree edges is 34 ¡ 19 % for a wirelength increase of only 1%, and 87 ¡ 73 % for a wirelength increase of 20%. SPICE simulations on industry routed nets show that non-tree routing has the additional benefit of reducing maximum sink delay by an average of 28 ¡ 26 % compared to Steiner routing, and by an average of 3 ¡ 72 % compared to timing optimized routing. SPICE simulations further imply that non-tree routing has smaller delay variation due to process variability. I.

### Citations

1139 |
Geometric Algorithms and Combinatorial Optimization
- Grötschel, Lovasz, et al.
- 1988
(Show Context)
Citation Context .... 4. The greedy MRTA algorithm. Remark. We note that, despite its exponential size, the fractional relaxation of ILP (5) can still be solved in polynomial time, e.g., by using the Ellipsoid algorithm =-=[6]-=- with a separation oracle that runs a min-cut algorithm for eachsu¡ v¢ ¢ T to check feasibility of any given solution. This can be combined with a standard branch-and-bound procedure for finding the o... |

354 |
Introduction to Algorithms, 2nd ed
- Cormen, Leiserson, et al.
- 2001
(Show Context)
Citation Context ...unnecessary since lasu¡ v¢ is given by the rectilinear distance between u and v. In this case the greedy algorithm runs in OsN 2 K¢ time. 3 For a Fibonacci heap implementation of Dijkstra’s algorithm =-=[3]-=-, D ¡ O �� P¤ F� 2 log� P¤ F�¡ . Since edges of HsP¤ F¡ have positive integer weights, it is possible to improve the asymptotic bound on the running time by replacing Dijkstra’s algorithm with Thorup’... |

107 |
Tarjan, “Augmentation problems
- Eswaran, E
- 1976
(Show Context)
Citation Context ... in which a given subgraph must be augmented at minimum cost into a 2-edge (or more generally k-edge) connected graph. 1 Finding a minimum cost k-edge-connected augmentation is NP-hard even for k ¢ 2 =-=[5]-=-, and much work has been devoted to finding good heuristics and approximation algorithms, see, e.g., [7] and the references therein. The MRTA formulation differs from the 2-edge-connectivity augmentat... |

94 |
Undirected single-source shortest paths with positive integer weights in linear time
- Thorup
- 1999
(Show Context)
Citation Context ...e positive integer weights, it is possible to improve the asymptotic bound on the running time by replacing Dijkstra’s algorithm with Thorup’s recent linear-time single-source shortest path algorithm =-=[19]-=-. For Thorup’s algorithm, D becomes proportional to the number of vertices and edges in HsP¤ F¡ , i.e., D ¡ O �� P¤ F� 2¡ . 9s10 IV. EXPERIMENTAL RESULTS We compare our integer program and the greedy ... |

69 | Design of survivable networks
- Grötschel, Monma, et al.
- 1995
(Show Context)
Citation Context ...nnected graph. 1 Finding a minimum cost k-edge-connected augmentation is NP-hard even for k ¢ 2 [5], and much work has been devoted to finding good heuristics and approximation algorithms, see, e.g., =-=[7]-=- and the references therein. The MRTA formulation differs from the 2-edge-connectivity augmentation (E2AUG) problem in several respects:sWhile E2AUG is typically formulated for graphs, MRTA has a stro... |

48 | New performance driven routing techniques with explicit area/delay tradeoff and simultaneous wire sizing
- Lillis, Cheng, et al.
- 1996
(Show Context)
Citation Context ...om a recent industry design and routed by Cadence WarpRouter using minimum-area optimization, while the second set consisted of randomly generated nets routed using the timing-driven P-Tree algorithm =-=[13]-=- with buffer insertion disabled and identical sink required-arrival times. Each interconnect was represented by a Π model and driven by a 1¤ 8V voltage source with a ramped input signal of 150ps slew ... |

35 |
Modeling of interconnect capacitance, delay & crosstalk in VLSI
- Wong, Lee, et al.
- 2000
(Show Context)
Citation Context ...under three scenarios: nominal wire width, wire width reduced by dw ¢ 6¤ 67%, and wire width increased by dw ¢ 6¤ 67%. For each scenario we computed unit length wire capacitance using the formulas in =-=[20]-=- for parallel lines between two planes, including area, fringe, and coupling capacitances. The maximum 50% sink delay and its variation in percents are reported for the two sets of test instances in T... |

29 |
Modeling of defects in integrated circuit photolithographic patterns
- Stapper
- 1984
(Show Context)
Citation Context ... routing (which have negligible probability of occurrence), and consider only defects that affect its edges.sStapper’s defect size distribution. We assume Stapper’s 1sx 3 distribution of defect sizes =-=[18]-=-, [4]. Under this assumption, the probability that a net N fails due to a short defect is proportional to the length of wires routed parallel to N at minimum spacing, and the probability that N fails ... |

20 |
YOR: a yield-optimizing routing algorithm by minimizing critical areas and vias
- Kuo
- 1993
(Show Context)
Citation Context ...size defects that induce IC faulty behavior. A typical figure of merit that measures the layout’s robustness is obtained as the ratio of the total critical area to the layout area [4]. Previous works =-=[12]-=-, [10], [2] have addressed manufacturing reliability of interconnect exclusively under the assumption that tree topologies are used for routing all nets. In this paper we propose to introduce layout l... |

19 |
Impact of systematic spatial intra-chip gate length variability on performance of high-speed digital circuits
- Orshansky, Milor, et al.
(Show Context)
Citation Context ...tacles (random nets with 20 sinks). For evaluating robustness to process variability we used an experiment which models systematic variation sources such as lens aberrations which cover 5-10mm ranges =-=[16]-=-, [15], i.e., ranges that are larger than those covered today by unbuffered interconnect. Assuming 100% wire width correlation, we ran our simulations under three scenarios: nominal wire width, wire w... |

17 | Graph augmentation and related problems: theory and practice
- Hsu
- 1993
(Show Context)
Citation Context ...es different tradeoffs than tree augmentation for delay or skew optimization. Our Manhattan Routing Tree Augmentation (MRTA) formulation resembles the classical edge connectivity augmentation problem =-=[8]-=-, in which a given subgraph must be augmented at minimum cost into a 2-edge (or more generally k-edge) connected graph. 1 Finding a minimum cost k-edge-connected augmentation is NP-hard even for k ¢ 2... |

16 | Post routing performance optimization via multilink insertion and non-uniform wiresizing
- Xue, Kuh
- 1995
(Show Context)
Citation Context ...by forming “functional fill” as opposed to the present “dummy fill” methodologies.sTree augmentation schemes have been previously proposed in the context of clock routing for delay and skew reduction =-=[21]-=- and critical net routing for delay optimization [14]. However, previous algorithms do not work well in our context, since tree augmentation for manufacturability and reliability improvement involves ... |

14 | A uniform framework for approximating weighted connectivity problems
- Khuller, Raghavachari, et al.
- 1999
(Show Context)
Citation Context ...untime of our methods with the best methods reported in the literature for the related 2-edge-connectivity augmentation problem, including a sophisticated heuristic based on minimum-weight branchings =-=[11]-=- and a recent genetic algorithm [17]. Experiments on randomly generated and industry testcases show that our greedy augmentation method achieves significant increase in reliability (as measured by the... |

14 | Non-tree routing
- McCoy, Robins
- 1994
(Show Context)
Citation Context ...t “dummy fill” methodologies.sTree augmentation schemes have been previously proposed in the context of clock routing for delay and skew reduction [21] and critical net routing for delay optimization =-=[14]-=-. However, previous algorithms do not work well in our context, since tree augmentation for manufacturability and reliability improvement involves different tradeoffs than tree augmentation for delay ... |

12 |
A fast and simple Steiner routing heuristic
- Borah, Owens, et al.
- 1999
(Show Context)
Citation Context ...tance size n ¢ � 5¡ 10¡ 50¡ 100¡ 500¡ 1000� we generated 100 instances uniformly at random from a 10¡ 000s10¡ 000 grid. For each instance, a Steiner minimum tree is constructed using the ER heuristic =-=[1]-=- and then augmented (assuming no routing obstacles) by the four compared algorithms. Table I shows the statistics on the number of nodes (including wire turns), number of leaves, and total wirelength ... |

12 | Routing for Reliable Manufacturing
- Huijbregtz, Xue, et al.
- 1995
(Show Context)
Citation Context ...efects that induce IC faulty behavior. A typical figure of merit that measures the layout’s robustness is obtained as the ratio of the total critical area to the layout area [4]. Previous works [12], =-=[10]-=-, [2] have addressed manufacturing reliability of interconnect exclusively under the assumption that tree topologies are used for routing all nets. In this paper we propose to introduce layout level r... |

10 | Evolutionary local search for the edge-biconnectivity augmentation problem
- Raidl, Ljubic
- 2002
(Show Context)
Citation Context ...methods reported in the literature for the related 2-edge-connectivity augmentation problem, including a sophisticated heuristic based on minimum-weight branchings [11] and a recent genetic algorithm =-=[17]-=-. Experiments on randomly generated and industry testcases show that our greedy augmentation method achieves significant increase in reliability (as measured by the percentage of biconnected tree edge... |

4 |
Yield optimization in physical design: a review
- Chiluvuri
- 1996
(Show Context)
Citation Context ...ce optimization (sizing, etc.) design flows. New design techniques must be applied to improve manufacturing yield of large-area chips, as yield becomes an ever-greater determinant of design viability =-=[2]-=-. In nanometer-scale manufacturing technologies, likelihood of back-end-of-line (BEOL) defects (i.e., high-resistance via or interconnect defects) has increased relative to that of front-end-of-line (... |

4 |
de Gyvez, “Yield modeling and beol fundamentals
- P
(Show Context)
Citation Context ...ll centers of fixed size defects that induce IC faulty behavior. A typical figure of merit that measures the layout’s robustness is obtained as the ratio of the total critical area to the layout area =-=[4]-=-. Previous works [12], [10], [2] have addressed manufacturing reliability of interconnect exclusively under the assumption that tree topologies are used for routing all nets. In this paper we propose ... |

3 |
Predictive Technology Model. Available at http://wwwdevice.eecs.berkeley.edu/~ptm
- Berkeley
(Show Context)
Citation Context ...l and driven by a 1.8-V voltagesourcewith a ramped input signal of 150 ps slew time. 50% delay from the source to each sink was simulated based on 180 nm ITRS predictive technology model beta version =-=[9]-=- with the following parameters: unit wire resistance r = 0:040 = m, unit wirecapacitancec = 0:259 fF= m, sink capacitance c t = 63:358 fF and sourcedriving resistance R b = 139:434 . For evaluating ro... |

1 |
Personal Communication
- Orshansky
- 2002
(Show Context)
Citation Context ...riving resistance R b = 139:434 . For evaluating robustness to process variability, we used an experiment which models systematic variation sources such as lens aberrations which cover 5–10 mm ranges =-=[15]-=-, [16], i.e., ranges that are larger than those covered today by unbuffered interconnect. Assuming 100% wire-width correlation, we ran our simulations under three scenarios: nominal wirewidth, wirewid... |

1 |
Post routing performanceoptimization via multi-link insertion and nonuniform wiresizing
- Kuh
- 1995
(Show Context)
Citation Context ... forming “functional fill” as opposed to the present “dummy fill” methodologies. • Tree-augmentation schemes have been previously proposed in the context of clock routing for delay and skew reduction =-=[21]-=- and critical net routing for delay optimization [14]. However, previous algorithms do not work well in our context, since tree augmentation for manufacturability and reliability improvement involves ... |