Efficient Design and Analysis of Robust Power Distribution Meshes Puneet Gupta Blaze DFM Inc. Andrew B. Kahng.

Slides:



Advertisements
Similar presentations
THERMAL-AWARE BUS-DRIVEN FLOORPLANNING PO-HSUN WU & TSUNG-YI HO Department of Computer Science and Information Engineering, National Cheng Kung University.
Advertisements

Paul Falkenstern and Yuan Xie Yao-Wen Chang Yu Wang Three-Dimensional Integrated Circuits (3D IC) Floorplan and Power/Ground Network Co-synthesis ASPDAC’10.
Proportion Priors for Image Sequence Segmentation Claudia Nieuwenhuis, etc. ICCV 2013 Oral.
Noise Model for Multiple Segmented Coupled RC Interconnects Andrew B. Kahng, Sudhakar Muddu †, Niranjan A. Pol ‡ and Devendra Vidhani* UCSD CSE and ECE.
Post-Placement Voltage Island Generation for Timing-Speculative Circuits Rong Ye†, Feng Yuan†, Zelong Sun†, Wen-Ben Jone§ and Qiang Xu†‡
Geo479/579: Geostatistics Ch14. Search Strategies.
Visual Recognition Tutorial
Path Finding for 3D Power Distribution Networks A. B. Kahng and C. K. Cheng UC San Diego Feb 18, 2011.
UH Unit Hydrograph Model Response Functions of Linear Systems Basic operational rules:  Principle of Proportionality: f(cQ ) = c  f(Q)  Principle of.
Power-Aware Placement
Lecture 8: Clock Distribution, PLL & DLL
Design Sensitivities to Variability: Extrapolations and Assessments in Nanometer VLSI Y. Kevin Cao *, Puneet Gupta +, Andrew Kahng +, Dennis Sylvester.
Supply Voltage Degradation Aware Analytical Placement Andrew B. Kahng, Bao Liu and Qinke Wang UCSD CSE Department {abk, bliu,
April 16th, Photomask Japan 2008 Electrical Metrics for Lithographic Line-End Tapering Puneet Gupta 3,
From Compaq, ASP- DAC00. Power Consumption Power consumption is on the rise due to: - Higher integration levels (more devices & wires) - Rising clock.
Toward Performance-Driven Reduction of the Cost of RET-Based Lithography Control Dennis Sylvester Jie Yang (Univ. of Michigan,
MAE 552 Heuristic Optimization Instructor: John Eddy Lecture #18 3/6/02 Taguchi’s Orthogonal Arrays.
Fundamentals of Power Electronics 1 Chapter 19: Resonant Conversion Chapter 19 Resonant Conversion Introduction 19.1Sinusoidal analysis of resonant converters.
Accurate Pseudo-Constructive Wirelength and Congestion Estimation Andrew B. Kahng, UCSD CSE and ECE Depts., La Jolla Xu Xu, UCSD CSE Dept., La Jolla Supported.
A Proposal for Routing-Based Timing-Driven Scan Chain Ordering Puneet Gupta 1 Andrew B. Kahng 1 Stefanus Mantik 2
Detailed Placement for Leakage Reduction Using Systematic Through-Pitch Variation Andrew B. Kahng †‡ Swamy Muddu ‡ Puneet Sharma ‡ CSE † and ECE ‡ Departments,
ISPD 2000, San DiegoApr 10, Requirements for Models of Achievable Routing Andrew B. Kahng, UCLA Stefanus Mantik, UCLA Dirk Stroobandt, Ghent.
Topography-Aware OPC for Better DOF margin and CD control Puneet Gupta*, Andrew B. Kahng*†‡, Chul-Hong Park†, Kambiz Samadi†, and Xu Xu‡ * Blaze-DFM Inc.
Triple Patterning Aware Detailed Placement With Constrained Pattern Assignment Haitong Tian, Yuelin Du, Hongbo Zhang, Zigang Xiao, Martin D.F. Wong.
SLIP 2000April 9, Wiring Layer Assignments with Consistent Stage Delays Andrew B. Kahng (UCLA) Dirk Stroobandt (Ghent University) Supported.
Clustering Ram Akella Lecture 6 February 23, & 280I University of California Berkeley Silicon Valley Center/SC.
Exposure In Wireless Ad-Hoc Sensor Networks S. Megerian, F. Koushanfar, G. Qu, G. Veltri, M. Potkonjak ACM SIG MOBILE 2001 (Mobicom) Journal version: S.
UC San Diego Computer Engineering VLSI CAD Laboratory UC San Diego Computer Engineering VLSI CAD Laboratory UC San Diego Computer Engineering VLSI CAD.
Noise and Delay Uncertainty Studies for Coupled RC Interconnects Andrew B. Kahng, Sudhakar Muddu † and Devendra Vidhani ‡ UCLA Computer Science Department,
Radial Basis Function Networks
Non-Linear Simultaneous Equations
More Realistic Power Grid Verification Based on Hierarchical Current and Power constraints 2 Chung-Kuan Cheng, 2 Peng Du, 2 Andrew B. Kahng, 1 Grantham.
Gauss’ Law.
Logic Optimization Mohammad Sharifkhani. Reading Textbook II, Chapters 5 and 6 (parts related to power and speed.) Following Papers: –Nose, Sakurai, 2000.
Authors: Jia-Wei Fang,Chin-Hsiung Hsu,and Yao-Wen Chang DAC 2007 speaker: sheng yi An Integer Linear Programming Based Routing Algorithm for Flip-Chip.
Research on Analysis and Physical Synthesis Chung-Kuan Cheng CSE Department UC San Diego
Pattern Selection based co-design of Floorplan and Power/Ground Network with Wiring Resource Optimization L. Li, Y. Ma, N. Xu, Y. Wang and X. Hong WuHan.
On-chip power distribution in deep submicron technologies
UC San Diego / VLSI CAD Laboratory Incremental Multiple-Scan Chain Ordering for ECO Flip-Flop Insertion Andrew B. Kahng, Ilgweon Kang and Siddhartha Nath.
Module 1: Statistical Issues in Micro simulation Paul Sousa.
PAT328, Section 3, March 2001MAR120, Lecture 4, March 2001S16-1MAR120, Section 16, December 2001 SECTION 16 HEAT TRANSFER ANALYSIS.
ECE 8443 – Pattern Recognition ECE 8423 – Adaptive Signal Processing Objectives: Deterministic vs. Random Maximum A Posteriori Maximum Likelihood Minimum.
Copyright © 2009 Pearson Education, Inc. Ampère’s Law.
1 ELEC 3105 Basic EM and Power Engineering Start Solutions to Poisson’s and/or Laplace’s.
Chapter 22 Gauss’s Law Chapter 22 opener. Gauss’s law is an elegant relation between electric charge and electric field. It is more general than Coulomb’s.
Partition-Driven Standard Cell Thermal Placement Guoqiang Chen Synopsys Inc. Sachin Sapatnekar Univ of Minnesota For ISPD 2003.
I N V E N T I V EI N V E N T I V E A Morphing Approach To Address Placement Stability Philip Chong Christian Szegedy.
IIT-Madras, Momentum Transfer: July 2005-Dec 2005 Perturbation: Background n Algebraic n Differential Equations.
1 A Fast Algorithm for Power Grid Design Jaskirat Singh Sachin Sapatnekar Department of Electrical and Computer Engineering University of Minnesota.
MIDPOINT CIRCLE & ELLIPSE GENERARTING ALGORITHMS
Local Search and Optimization Presented by Collin Kanaley.
Quality of LP-based Approximations for Highly Combinatorial Problems Lucian Leahu and Carla Gomes Computer Science Department Cornell University.
Accelerating Dynamic Time Warping Clustering with a Novel Admissible Pruning Strategy Nurjahan BegumLiudmila Ulanova Jun Wang 1 Eamonn Keogh University.
Discretization Methods Chapter 2. Training Manual May 15, 2001 Inventory # Discretization Methods Topics Equations and The Goal Brief overview.
System in Package and Chip-Package-Board Co-Design
Modelling and Simulation of Passive Optical Devices João Geraldo P. T. dos Reis and Henrique J. A. da Silva Introduction Integrated Optics is a field of.
Video Coding Presented By: Dr. S. K. Singh Department of Computer Engineering, Indian Institute of Technology (B.H.U.) Varanasi
Lecture 8: Measurement Errors 1. Objectives List some sources of measurement errors. Classify measurement errors into systematic and random errors. Study.
Effective Linear Programming-Based Placement Techniques Sherief Reda UC San Diego Amit Chowdhary Intel Corporation.
Dept. of Electronics Engineering & Institute of Electronics National Chiao Tung University Hsinchu, Taiwan ISPD’16 Generating Routing-Driven Power Distribution.
1 Hardware Reliability Margining for the Dark Silicon Era Liangzhen Lai and Puneet Gupta Department of Electrical Engineering University of California,
Appendix A with Woodruff Edits Linear Programming Using the Excel Solver Copyright © 2010 by The McGraw-Hill Companies, Inc. All rights reserved.McGraw-Hill/Irwin.
EEE 431 Computational Methods in Electrodynamics
ELEC 3105 Basic EM and Power Engineering
Ampère’s Law Figure Arbitrary path enclosing a current, for Ampère’s law. The path is broken down into segments of equal length Δl.
ECEN 460 Power System Operation and Control
Revisiting and Bounding the Benefit From 3D Integration
Haihua Su, Sani R. Nassif IBM ARL
Lecture 10 Biot-Savart’s Law.
Presentation transcript:

Efficient Design and Analysis of Robust Power Distribution Meshes Puneet Gupta Blaze DFM Inc. Andrew B. Kahng ECE & CSE UCSD Presented by Swamy Muddu

Motivation Current density and hence IR drop is increasing with every technology generation IR drop Static IR drop: resistive, driven by peak currents Transients: E.g., IR drop analysis and optimization is very slow Optimization is usually formulated as Non-Linear Program Analysis too slow to be used during Place & Route Typical global power distribution Horizonatal/Vertical stripes Power meshes Peripheral i/o or flipchip This work: Static IR drop, peripheral i/o

The 1D Case Consider a power stripe with single power source at one end. Order power tap-points 1..N starting from the one closest to the power source IR drop increases away from the power source Wire segments between current tap-points closer to Vdd carry higher current  A tapered power stripe is more bang for buck Closed form solution for optimal sizing can be calculated using Lagrangian multipliers. Assume area of a wire segment is proportional to its conductance V dd Power tap points r 45 i 23

The 1D Case contd.. Solution to the minimum area IR drop (MAIC) constrained stripe sizing: (r=resistance, i=current) Solution to the minimum IR drop area constrained (MIAC) problem is given by: (G=total conductance constraint)

IR Drop in Power Grids IR Drop increases going in from the power ring “Radial” nature of IR drop variation in a power grid  notice the similarity to the 1D case Equipotential contours take the form of “rings” : diamond shaped at the center of layout and square shaped at the periphery (shape of power ring) Bull’s Eye Equipotential ring Outer Power Ring An example IR drop map for a uniform 25 X 25 power mesh with uniform current requirements

Analysis of the 2D Case Assume square equipotential rings Divide power mesh into radial and tangential segments nXn mesh  rings Intuition: Current flows only in the radial segments

Radial vs. Tangential Segments Variation of peak IR drop with varying widths of radial and tangential segments is shown Radial segments impact IR drop much more than tangential segments  close to zero current flow in the tangential segments Assume radial current flow Peak IR Drop vs. conductance Of wire segments in a 25x25 power grid

Power Grid Optimization Three phase optimization of power grids: Radial Sizing. Size radial segments assuming uniform distribution of currents around rings. Tangential Sizing. Account for nonuniformity of current distribution. Divide rings into sectors and redistribute metal among sectors. Circumference Correction. Relax square ring assumption. Account for diamond to square ring shape transition All phases are based on closed form sizing expressions

Radial Sizing Assume no current flow along the equipotential ring  tangential segments sized to minimum width MAIC Sizing Solution i(p,p+1): current from ring p to p+1 MIAC Sizing Solution G R : total allocated radial conductance All radial segments originating from an equipotential ring are sized equally

Tangential Sizing Redistribute metal between the radial segments along the equipotential ring to account for non-uniform current distribution Divide power grid into quadrants Assume tap-points within a quadrant draw current from segments contained in the quadrant Enforce the radial-sizing IR drop from ring-to-ring Tangential Sizing Solution G P : total conductance of ring p i q p : current in quadrant q of ring p

Circumference Correction Equipotential contour progresses from a diamond at the center of chip to square at the power ring Heuristically size tangential segments r(x,y): resistance of the segment at location (x,y) r: resistance of the corresponding radial segment α: constant. nα=10 gives good tradeoff between area and IR drop Total Conductance after three phases G R : Allocated total radial conductance

Experiments Testcases: T1, T2: uniform current requirements T3: derived from an industry flip-chip testcase T4: derived from T3 MATLAB used as numerical solver to compute exact IR drop.

Results Upto 33% peak IR drop reduction with same area Upto 32% area reduction with same IR drop Results of MAIC Sizing Results of MIAC Sizing

Zero Time IR Drop Analysis VQ q p : IR drop in quadrant q of ring p Simple measure with high correlation to actual IR drop at any point on the mesh useful for optimization and quick, early prediction

Perturbations and Robustness Variation in IR drop can occur due to current and/or resistance variation Perturbation result: ∞ norm used for peak IR drop G: power distribution network conductance matrix I: Current matrix E: Perturbation in G e: Perturbation in I ||G||X||G -1 || ∞ : condition number = measure of robustness Uniform meshes tend to have lower condition numbers  more robust

Conclusions and Future Work Contributions: Closed form power mesh optimization achieving up to 33% peak IR drop reduction Closed form IR drop analyses Proposed a measure of robustness of power distribution networks Simple incremental power mesh optimization technique (not discussed here) Future Work Extensions to flipchips, hard macros Use in IR-drop aware placement  place high power cells closer to power source (power ring)