Optimal reliable crosstalk-driven interconnect optimization

Abstract

As technology advances apace, crosstalk becomes a design metric of comparable importance to area and delay. This article focuses mainly on the crosstalk issue, specifically on the impacts of physical design and process variation on crosstalk. While the feature size shrinks below 0.25 μm, the impact of process variation on crosstalk increases rapidly. Hence, a crosstalk insensitive design is desirable in the deep submicron regime. In this article, crosstalk sensitivity is referred to as the influence of process variation on crosstalk in a circuit. We show that the lower bound of crosstalk sensitivity grows quadratically, while that of crosstalk increases linearly. Therefore, designers should also consider crosstalk sensitivity, when optimizing other design objectives such as crosstalk, area, and delay. According to our modeling, these objectives are all in posynomial forms, and thus the multiobjective optimization problem can optimally be solved by Lagrangian relaxation. Experimental results show that our method is effective and efficient. For instance, a circuit of 2856 gates and 5272 wires is optimized using 13-minute runtime and 2.8-MB memory on a Pentium III 1.0 GHz PC with 256-MB memory. In particular, by relaxing Lagrange multipliers to the critical paths, it takes only two iterations for all solutions to converge to the global optimal, which is much more efficient than related previous work. This relaxation scheme provides a key insight into the rapid convergence in Lagrangian relaxation.

Shield count minimization in congested regions

“…These shielding requirements are used to drive our shield sharing optimization algorithm as follows. Critical signal nets are ordered by the degree of freedom in their placement 5 and by the number of shields required by them, and then routed gridlessly with the most . Each individual net is placed so as to reuse previously routed shields (including any prerouted power lines) as much as possible.…”

Shield count minimization in congested regions

“…These shielding requirements are used to drive our shield sharing optimization algorithm as follows. Critical signal nets are ordered by the degree of freedom in their placement 5 constrained nets being routed first (along with their associated shields). Each individual net is placed so as to reuse previously routed shields (including any prerouted power lines) as much as possible.…”

“…Note that the minimum width of a power line is usually greater than that of a generic signal line 5. The net placement constraints can arise from diverse sources such as the bounding box of a driver and its critical sinks, or the need to avoid the RC penalty of vias close to the driver(s).…”

Technology mapping with crosstalk noise avoidance

“…Generally, at the postrouting stage, crosstalk reduction can be achieved by transistor sizing, gate sizing or wire sizing for the accurate noise analysis based on the RC extraction of layout [24], [4], [8], [9], [23]. However, the flexibility of adjusting netlists might not be enough to fix hundreds or even thousands of interconnect noise problems, so eventually those unsolved noise problems would only rely on the rip-up and reroute step.…”