Title: More Realistic Power Grid Verification Based on Hierarchical Current and Power constraints
1More Realistic Power Grid Verification Based on
Hierarchical Current and Power constraints
- 2Chung-Kuan Cheng, 2Peng Du, 2Andrew B. Kahng,
1Grantham K. H. Pang, 1Yuanzhe Wang, 1Ngai Wong - 1. The University of Hong Kong
- 2. University of California, San Diego
2Outline
- Background
- Problem formulation
- Efficient solver
- Experimental results
- Conclusion
3Background
Power grid -gt RCL network External voltage
sources -gt ideal voltage sources Transistors,
logic gates, etc -gt ideal current sources
4Background
External voltage source power grid
transistors, logic gates, etc
VoltageZs
Current
Logic error
Voltage drop
Timing error
Gates ? Current density ? Line width
? External voltage ?
Voltage drop ?
Noise margin ?
Power grid voltage drop verification is becoming
indispensable
5Background
1. Simulation-based power grid verification
capacitance inductance
admittanceimpedance
input distribution matrix
nodal voltages
current patterns
transient analysis
current patterns
voltage drops
6Background
2. Worst case power grid verification
max Voltage_Drop subject to Current_Constraints
Design experience Design requirements
- Early-stage verification current patterns
unknown - Uncertain working modes too many possible
current patterns
Check maxVoltage_Drop ltNoise_Margin
7Background
xk nodal voltage at k i current sources IL
local current bounds IG global current bounds c
relationship between voltage and current U
current distribution matrix
Worst-case voltage drop prediction via solving
linear programming problems
D. Kouroussis and F. N. Najm, A static
pattern-independent technique for power grid
voltage integrity verification, 2003
8Background
- Solving linear programs
- Simplex algorithm theoretically NP-hard O(n3)
in practice. - Ellipsoid algorithm O(n4)
- Interior-point algorithm O(n3.5)
- n is usually large (gt millions)
Existing work (for higher efficiency) Geometric
method -gt trade-off with accuracy(Ferzli, ICCAD
07) Dual algorithm -gt still large complexity
(convex optimization)(Xiong, DAC 07)
9Outline
- Background
- Problem formulation
- Efficient solver
- Experimental results
- Conclusion
10Problem formulation
Relationship between voltage drop and currents
Backward Euler
Numerically equivalent to transient analysis
11Problem formulation
Hierarchical current and power constraints
1 Local current constraints
2 Block-level current constraints
Different from previous work, U is a 0/1 matrix
with each column containing at most one 1.
This is the requirement of hierarchy
12Background
3 Block-level power constraints
i
i
t
t
current constraints gt peak value of current
waveform
power constraints gt area under current waveform
13Problem formulation
4 High level power constraints
are 0/1 matrices with each column containing at
most one 1
Hierarchical Constraints
14Problem formulation
Worst-case voltage drop occurs at the final time
step (see the paper for detailed proof). Thus the
linear programming problem reads
15Outline
- Background
- Problem formulation
- Efficient solver
- Experimental results
- Conclusion
16Efficient solver
Coefficient computation
- We do not have to solve ci,k for every i (those
i to be solved form a set O) - Solving those nodes with current sources attached
( current sources usually lt of nodes) - Solving those critical nodes which have great
influence on circuit performance
17Efficient solver
A parallel algorithm without matrix inversion is
desired.
transpose
- Requiring one sparse-LU and kt forward/backward
substitutions - Parallelizable
18Efficient solver
ci,k known now
Rename variables by treating each entry of each
u(k?t) as independent variables
The objective function can be rewritten as
19Efficient solver
Each constraint represents that the sum of some
variables belonging to a set is smaller than a
bound
20Efficient solver
The problem is rewritten as
The constraints here are hierarchical, which
follows that for any two sets , at
least one of the 3 equations holds
not hierarchical
hierarchical
21Efficient solver
Lemma The objective function reaches maximum
when all the variables associated with
negative are set to zero.
- Intuitive interpretation
- The objective function is smaller when variables
with negative variables are positive - Set these variables to zero will not decrease the
feasible set defined by constraints.
22Efficient Solver
Set all the variables associated with negative
coefficients as zero and sort the remaining
coefficients in the descending order
The problems becomes
Then it can be proven that a sorting-deletion
algorithm can give the optimal solution.
23Efficient solver
Intuitive interpretation Give the variable
associated with the largest coefficient the
largest possible value. Then delete this variable
from the problem and do the same procedure again.
24Efficient solver
- Complexity of the sorting-deletion algorithm
- Coefficient sorting (using the most efficient
sorting algorithm) - Deletion procedure (r is the of level in the
hierarchical structure, mkt is of variables) - Much lower than standard algorithms
25Outline
- Background
- Problem formulation
- Efficient solver
- Experimental results
- Conclusion
26Experimental results
Models used 3-D power grid structure with 4
metal layers
- Compare the voltage drop predictions with and
without power constraints - Compare the CPU time using sorting-deletion
algorithm and standard algorithms
27Experimental results
Worst-case current patterns with and without
power constraints (pcs). Introduction of power
constraints may reduce over-pessimism.
28Experimental results
Worst-case voltage drop predictions with and
without power constraints (pcs). Introduction of
power constraints may reduce over-pessimism.
29Experimental results
(1) Standard LP solver fails due to too many
iterations
CPU time comparison between standard algorithms
and sorting-deletion algorithm. Significant
speed-up is achieved.
30Conclusion
- Introduction of power constraints provide more
realistic current patterns and less pessimistic
voltage drops. - Efficient and parallelizable coefficient
computation is proposed. - Sorting-deletion algorithm significantly reduces
the CPU time to solve the linear programming
problems.