Title: PiCAP: A Parallel and Incremental Capacitance Extraction Considering Stochastic Process Variation
1PiCAP A Parallel and Incremental Capacitance
Extraction Considering Stochastic Process
Variation
Fang Gong1, Hao Yu2, and Lei He1 1Electrical
Engineering Dept., UCLA 2Berkeley Design
Automation Presented by Fang Gong
2Outline
- Background and Motivation
- Algorithms
- Experimental Results
- Conclusion and Future Work
3Outline
- Background and Motivation
- Algorithms
- Experimental Results
- Conclusion and Future Work
4Process Variation and Cap Extraction
- Process variation leads to capacitance variation
- OPC lithography and CMP polishing
- Capacitance variation affects circuit performance
- Delay variation and analog mismatch
5Background of BEM Based Cap Extraction
Source Panel j
- Capacitance extraction in FastCap
- Procedures
- Discretize metal surface into panels
- Form linear system by collocation
- Results in dense potential coeffs
- Solve by iterative GMRES
- Fast Multipole method (FMM) to evaluate Matrix
Vector Product (MVP) - Preconditioned GMRES iteration with guided
convergence
Observe Panel i
- Difficulties for stochastic capacitance
extraction - How to consider variations in FMM?
- How to consider different variations in
precondition?
6Motivation of Our Work
- Existing works
- Stochastic integral by low-rank approximation
- Zhu, Z. and White, J. FastSies a fast
stochastic integral equation solver for modeling
the rough surface effect. In Proceedings of
IEEE/ACM ICCAD 2005. - Pros Rigorous formulation
- Cons Random integral is slow for full-chip
extraction - Stochastic orthogonal polynomial (SOP) expansion
- Cui, J., and etc. Variational capacitance
modeling using orthogonal polynomial method.
In Proceedings of the 18th ACM Great Lakes
Symposium on VLSI 2008. - Pros An efficient non-Monte-Carlo approach
- Cons SOP expansion results in an augmented and
dense linear system
- Objective of our work
- Fast multi-pole method (FMM) with nearly O(n)
performance with a further parallel improvement. - Pre-conditioner should be updated incrementally
for different variation.
7Outline
- Background and Motivation
- Algorithms
- Experimental Results
- Conclusion and Future Work
8Flow of piCAP
Incrementally update preconditioner
Potential Coefficient
- Represent Pij with stochastic geometric moments
- Use parallel FMM to evaluate MVP of Pxq
- Obtain capacitance (mean and variance) with
incrementally preconditioned GMRES
9Stochastic Geometric Moment
- Consider two independent variation sources panel
distance (d) and panel width (w)? - Multipole expansion along x-y-z coordinates
multipole moments and local moments - Mi and Li show an explicit dependence on geometry
parameters, and are called geometric moments.
10Stochastic Potential Coefficient Expansion
- Stochastic Potential Coefficient
- Relate geometric parameters to random variables
- Let be random variable for panel width w,
and be random variable for panel distance
d. - Geometric moments Mp and Lp are
- Now, the potential coefficient is
11Stochastic Potential Coefficient Expansion
- Stochastic Potential Coefficient
- Relate geometric parameters to random variables
- Let be random variable for panel width w,
and be random variable for panel distance
d. - Geometric moments Mp and Lp are
- Now, the potential coefficient is
- n-order stochastic orthogonal polynomial
expansion of P - Accordingly, m-th order (m 2n n(n - 1))
expansion of charge is
12Augmented System
- Recap SOP expansion leads to a large and dense
system equation
13Parallel Fast Multipole Method--upward
- Overview of Parallel fast-multipole method (FMM)?
- group panels in cubes, and build hierarchical
tree for cubes - We use 8-degree trees in implementation, but use
2-degree trees for illustration here. - A parallel FMM distributes cubes to different
processors - Upward Pass
Level 0
Level 1
Level 2
M2M
Level 3
- Update parents moments by summing the moments
of its childrencalled M2M operation
- starting from bottom level, it calculates
stochastic geometric moments
- M2M operations can be performed in parallel at
different nodes
14Parallel Fast Multipole Method--Downwards
M2L
Level 0
L2L
Level 1
Level 2
M2M
Level 3
M2M, L2L are local operations, while M2L is
global operation. How to reduce communication
traffic due to global operation?
- Sum L2L results with near-field potential for all
panels at bottom level and return Pxq
- At the top level, calculation of potential
between two cubescalled M2L operation.
- potential is further distributed down to
children from their parent in parallelcalled L2L
operation.
- Calculate near-field potential directly in
parallel
15Reduction of traffic between processors
- Global data dependence exists in M2L operation at
the top level - Pre-fetch moments distributes its moments to all
cubes on its dependency list before the
calculations. - As such, it can hide communication time.
16Flow of piCAP
- Use spectrum pre-conditioner to accelerate
convergence - Incrementally update the pre-conditioner for
different variation.
17Deflated Spectral Iteration
- Why need spectral preconditioner
- GMRES needs too many iterations to achieve
convergence. - Spectral preconditioner shifts the spectrum of
system matrix to improve the iteration
convergence - Deflated spectral iteration
- k (k1 power iteration) partial eigen-pairs
- Spectrum preconditioner
-
- Why need incremental precondition
- Variation can significantly change spectral
distribution - Building each pre-conditioner for different
variations is expensive - Simultaneously considering all variations
increases the complexity of our model.
18Incremental Precondition
- For updated system , the update for
the i-th eigen vector
is - is the subspace composed of
- is the updated spectrum
- Updated pre-conditioner W is
Inverse operation only involves diagonal matrix DK
Consider different variations by updating the
nominal preconditioner partially.
19Outline
- Background and Motivation
- Algorithm
- Experimental Results
- Conclusion and Future Work
20Accuracy Comparison
- Setup two panels with random variation for
distance d and width w - Result Stochastic Geometric Moments have high
accuracy with average error of 1.8, and can be
up to 1000X faster than MC
21Runtime for parallel FMM
- Setup
- Two-layer example with 20 conductors.
- Other 40, 80, 160 conductors
- Evaluate Pxq (MVP) with 10 perturbationon panel
distance - Result
- All examples can have about 3X speedup with 4
processors
22Efficiency of spectral preconditioner
- Setup Three test structures single plate, 2x2
bus, cubic - Result
- Compare diagonal precondition with spectrum
precondition - Spectrum precondition accelerates convergence of
GMRES (3X).
23Speedup by Incremental Precondition
- Setup
- Test on two-layer 20 conductor example
- Incremental update of nominal pre-conditioner for
different variation sources - Compare with non-incremental one
- Result Up to 15X speedup over non-incremental
results, and only incremental one can finish all
large examples.
24Conclusion and Future Work
- Introduce stochastic geometric moments
- Develop a parallel FMM to evaluate the
matrix-vector product with process variation - Develop a spectral pre-conditioner incrementally
to consider different variations - Future Work extend our parallel and incremental
solver to solve other IC-variation related
stochastic analysis
25Thanks
PiCAP A Parallel and Incremental Capacitance
Extraction Considering Stochastic Process
VariationFang Gong, Hao Yu and Lei He