Constraint-Driven IO Planning and Placement for Chip Package Co-Design

Constraint-Driven IO Planning and Placement for Chip Package Co-Design
paly

This paper discusses a new approach to IO planning and placement for chip package co-design. The authors introduce the concept of "chip package aware design constraints" and present

About Constraint-Driven IO Planning and Placement for Chip Package Co-Design

PowerPoint presentation about 'Constraint-Driven IO Planning and Placement for Chip Package Co-Design'. This presentation describes the topic on This paper discusses a new approach to IO planning and placement for chip package co-design. The authors introduce the concept of "chip package aware design constraints" and present. The key topics included in this slideshow are . Download this presentation absolutely free.

Presentation Transcript


Slide1Constraint Driven I/O Planning andPlacement for Chip-package Co-design Jinjun Xiong ,  Yiuchung Wong, Egino Sarto ,  Lei He University of California, Los Angeles Rio Design Automation, Inc.

Slide2Agenda Motivation  Overview of our approach  Chip-package aware design constraints  CIOP: constraint-driven I/O placement problem formulation  Multi-step CIOP algorithm  Experiment results  Conclusion

Slide3SiliconPackage Board (Cadence)

Slide4Chip-package Co-desgin Bridge the gap between chip and package designers • Better convergence • Quicker time-to-market • Cost reduction  I/O placement plays a critical role • The interface between chip and package designs

Slide5I/O Placement for Flip-chip Challenges • I/O cells placed anywhere on the die • Consider the bump locations on the package • Timing closure • Signal integrity (SI) • Power integrity  Complicated design constraints are generated in practice to guide the I/O placement

Slide6Major Contributions A formal definition of a set of design constraints  A new formulation of constraint-driven I/O placement  An effective multi-step design methodology for chip-package co-design

Slide7Co-design Methodology Global I/O and core co- placement  Bump array placement • Areas for bump pads  I/O site definition • Areas for I/O cells  Constraint driven detailed I/O placement  I/O placement consists of three essential sub-problems • Placement of bump arrays • Placement of I/O sites • Placement of I/O cells

Slide8Power Integrity Constraints Power domain constraint • I/O cell voltage specification • Cells from same domain prefer physically closer  Minimize power plane cut lines in the package • Provide proper power reference plane for traces • Depend on physical locations of I/O cells  Proper signal-power-ground (SPG) ratio • Primary and secondary P/G driver cells • Minimize voltage drop and Ldi/dt noise

Slide9Timing Constraints Substrate routes in package varies significantly • Length spans from 1mm to 21mm • Timing varies more than 70ps for SSTL_2  I/O cells with critical timing constraints shall take this into account • Differential pair prefer to escape in parallel

Slide10I/O Standard Related Constraints High-speed design    high-speed I/O  I/O standard requirements • Relative timing requirements on signals • Likely to be connected to the same interface at other chips, so prefer to keep relative order to ease routing  Closeness constraint  Bump assignment feasibility constraint

Slide11Floorplan Induced Region Constraints Top-down design flow • PCB floorplan  Bottom-up design • Chip floorplan  I/O cells have region preference • Which side? • What location?

Slide12CIOP Problem Formulation Given: a fixed die size, a net-list with I/O cells, a set of design constraints  Find: • Placement of bump arrays • Placement of I/O site • Legal placement of I/O cells  Such that: all design constraints are satisfied • Wire length is also minimized

Slide13Agenda Motivation  Overview of our approach  Chip-package aware design constraints  CIOP: constraint-driven I/O placement problem formulation  Multi-step CIOP algorithm • Constraint-driven global I/O planning • Constraint-driven detailed I/O placement  Experiment results  Conclusion

Slide14Global I/O and Core Co-placement Wire-length driven  Constraint driven • Minimize power domain slicing on the package planes  Grid-based • Uniformity  No restriction on a particular global placement engine • Force-directed • Partition-based • Analytic-based

Slide15Global I/O and Core Co-placement Additional components to cost function • Region constraint: quadratic penalty functions • SI constraints and escapability constraints   bin capacity constraints • High level abstraction for efficiency consideration  Power domain constraints: I/O cells from the same power domain closer to each other • Add a virtual net to connect I/O cells belonging to the same domain • Each bin is assigned to at most one power domain • Decided by the majority I/O cells’ power domain property • Adjacent bins of the same domain are merged • If one power domain is too fragmented, the corresponding virtual net will be given a higher weight in the next placement run

Slide16Global I/O and Core Co-placement Power domain definition • Majority I/O cells location • Modeled in global placement  Translated to region constraints for I/O cells for the following steps

Slide17Bump and Site Definition Regular bump pattern is preferred • Escapability analysis  Regular I/O site is preferred • I/O proximity • RDL planar routability analysis  I/O sites more than I/O cells • SPG ratio consideration • Flexibility for later bump assignment  I/O super site: a cluster of I/O sites

Slide18Agenda Motivation  Overview of our approach  Chip-package aware design constraints  CIOP: constraint-driven I/O placement problem formulation  Multi-step CIOP algorithm • Constraint-driven global I/O planning • Constraint-driven detailed I/O placement  Experiment results  Conclusion

Slide19ILP Feasibility Problem for SuperSite Assignment  One I/O cell to one I/O site  I/O site capacity const  Differential pair capacity  Region constr.  Differential pair const.

Slide20ILP Feasibility Problem for SuperSite Assignment  Captures clustering constraints (L i , C i L )

Slide21Legal Assignment of I/O Cells toI/O Sites  Solved on a per super site basis  Min-cost-max-flow problem • A bipartite graph G(V1,V2,E) • V1: the set of  I/O cells assigned to the super site • V2: the set of I/O cells within the super site • E: the feasible connection between V1 and V2 • Query bumps escape layer properties • Query substrate route characteristics: e.g., impedance, route length • Determine whether or not an I/O cell is allowed to be assigned to an I/O site • Cost of E: preference in assignment • RDL wire length from I/O cells to I/O sites • Constraint violation

Slide22Experiment Results Real industrial designs  Constraints not include the ones that are generated internally

Slide23Experiment Results CSR: Constraint Satisfaction Ratio  Our algorithm can satisfy all design constraints in one iteration  Runtime is very promising

Slide24Comparison Study Two base-line algorithms are studied • TIOP: conventional constraint-oblivious approach • TCIOP: Constraint-driven global I/O planning + conventional constraint-oblivious I/O placement • Both may not satisfy all design constraints in one iteration  Iterative local refinement procedure follows to further improve CSR • Swapping, shifting, relocating

Slide25Comparison on CSR X-axis: iteration number  Y-axis: CSR in percentage  Recall: our CIOP’s CSR = 100% using one iteration TIOP TCIOP

Slide26Comparison on Wire Length Normalize wire length w.r.t. that of TIOP’s zeroth iteration  Wire length increase in percentage

Slide27Agenda Motivation  Overview of our approach  Chip-package aware design constraints  CIOP: constraint-driven I/O placement problem formulation  Multi-step CIOP algorithm  Experiment results  Conclusion

Slide28Conclusion Formally defined a set of common design constraints for chip-package co- design  Formulated a detailed constraint-driven I/O placement problem (CIOP)  Solved CIOP via an effective multi-step algorithm