1 / 52

Nike’s Software Architecture and Infrastructure: Enabling Integrated Solutions for Gigahertz Designs

Nike’s Software Architecture and Infrastructure: Enabling Integrated Solutions for Gigahertz Designs V. Nagbhushan, Nike Development, DT, Intel Corp. Yehuda Shiran, Nike Development, DT, Intel Corp. Satish Venkatesan, Nike Development, DT, Intel Corp.

betty_james
Télécharger la présentation

Nike’s Software Architecture and Infrastructure: Enabling Integrated Solutions for Gigahertz Designs

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Nike’s Software Architecture and Infrastructure:Enabling Integrated Solutions for Gigahertz Designs V. Nagbhushan, Nike Development, DT, Intel Corp. Yehuda Shiran, Nike Development, DT, Intel Corp. Satish Venkatesan, Nike Development, DT, Intel Corp. Tamar Yehoshua, Nike Development, DT, Intel Corp.

  2. Agenda • Introduction • External & Internal Vectors • Drawback of existing architectures • Nike’s software architecture • LaMA, Principles, Data model • Development environment • Development cycle & Validation • Conclusion

  3. Introduction • Nike was developed at the Design and Test Technology (DT) of Intel® • intended to provide the future generation CAD tool suite • First release – Q3, 1999 VLSI CAD seminar 2004

  4. External Vectors • Industry technology increases exponentially towards chips with over 100M transistors, and beyond 10GHz frequencies • Feature sizes decrease, aggressive circuit styles becoming norm • Second order effects like noise and inductance become key factors • Need increased visibility into data for multiple domains (circuit, layout, etc.) VLSI CAD seminar 2004

  5. External Vectors – cont. • Marketing deadline demands, forces decrease in design time • Computing environment changes with the emergence of Windows NT* on Intel® architecture VLSI CAD seminar 2004

  6. Internal Vectors • Growth in coverage of tool suite & overall software size • Extensibility and maintainability • Persistence of core architecture • Add new feature without massive code changes • customizable VLSI CAD seminar 2004

  7. Internal Vectors – cont. • Tool quality and Development efficiency • Reduce design iterations • Well defined software development methodology • Avoid numerous cycles to fix defects VLSI CAD seminar 2004

  8. Drawbacks of existing archs. • Non-modular • Severely impacts development efficiency • Local implementations of similar/identical functionality causes inconsistent behavior • Difficult reuse • Existence of several data models VLSI CAD seminar 2004

  9. Drawbacks – cont. • Difficult data exchange across domains • Inconsistent terminology, interfaces & implementations • Difficult implementation of unified data mapping across domains (e.g. circuit  layout data translation) often results in data loss when ad-hoc mapping fashion is used VLSI CAD seminar 2004

  10. Drawbacks – cont. • Ad-hoc persistence mechanisms • ASCII files used to store data persistently • Inadequate use of industry standard formats • Proliferation of files • Multiple readers/writers • Semantic mismatches VLSI CAD seminar 2004

  11. Drawbacks – cont. • ASCII is not performance oriented • Capacity limitations • Intolerant of new software releases • Difficult implementation of incremental I/O • Inconsistent look & feel • Different interfaces for the same actions results in loss of productivity VLSI CAD seminar 2004

  12. Short summary - up to this point • Several vectors (internal/external) and drawbacks of existing technologies,requires a better technology to be able to develop the next generation of chips VLSI CAD seminar 2004

  13. Nike’s Software Architecture VLSI CAD seminar 2004

  14. LaMA • Layered Modular Architecture • Hierarchical decomposition of a complex tool suite • More than 100 tools • More than 1,000,000 code lines VLSI CAD seminar 2004

  15. LaMA – cont. VLSI CAD seminar 2004

  16. LaMA – cont. • The overall suite consists of several subsystems • Sub-systems • Each one is targeted at a particular user flow. • Comprised of several capabilities VLSI CAD seminar 2004

  17. LaMA – cont. • Capabilities • Represents a user-visible functionality • Implemented by one or more software modules • The modules and atoms are designed with well-defined interfaces to facilitate reuse by several capabilities. VLSI CAD seminar 2004

  18. LaMA – cont. VLSI CAD seminar 2004

  19. Software Architecture Principles • Set to meet or beat high-level specifications set forth • Integrated tool suites (cockpits) • Enables users to execute flow or perform similar operations • Improves user productivity VLSI CAD seminar 2004

  20. Software Arch. Principles – cont. • Modular design • Common, unified services • Well documented interface per component • Similar functional components should be implemented as common services VLSI CAD seminar 2004

  21. Software Arch. Principles – cont. • Common look & feel • Similar visual appearance for UI for objects performing similar operations • Feel (behavior) of similar operations and data objects should be similar • Improves predictability and productivity VLSI CAD seminar 2004

  22. Software Arch. Principles – cont. • Support incremental processing • Handle small delta changes to the input by incremental processing and produce results exhibiting small delta changes • Extensible – allow user to easily extend functionality (per capability or entire suite) • Support plug&play with external tools VLSI CAD seminar 2004

  23. Nike Layout Software Arch. VLSI CAD seminar 2004

  24. Nike Layout SW Arch. – cont. • Each module has a well defined procedural interface (API) • Data Model (DM) • In-memory repository for all primary, non-derivable data of the system • All modules access CAD data through its API VLSI CAD seminar 2004

  25. Nike Layout SW Arch. – cont. • Data Model (DM) – cont. • Example: layout data (UDM – Unified Data Model), contains data such as cells, wires, nets, transistors etc. • API guarantees consistency of data structures and semantics • I/O module – specialized engine, enablesfast I/O VLSI CAD seminar 2004

  26. Nike Layout SW Arch. – cont. • Engines • Well defined functionality • e.g. Algorithmic engine, core editing engine • Work off data from DM but may create temporary derived data for efficiency VLSI CAD seminar 2004

  27. Nike Layout SW Arch. – cont. • Capabilities • Environments addressing the traditional sub-domains such as full-chip layout, block layout and leaf cell • Provides functionality and customization appropriate to the capability • Can directly expose an engine functionality to the user (through UI) or hide or modify it VLSI CAD seminar 2004

  28. Nike Layout SW Arch. – cont. • User interface • Visibility of functionality and data • Common look • Instantiation of GUI items by capabilities and engines • Extension system enables customization for easy access to data and functionality VLSI CAD seminar 2004

  29. Nike Layout SW Arch. – cont. • User interface – cont. • Tcl interpreter and Windows* automation interfaces for DM, engine and capability modules • VBA for GUI customization on Windows NT* VLSI CAD seminar 2004

  30. Data Modeling • Design cycle involves many iterations, both within a phase and between phases VLSI CAD seminar 2004

  31. Data Modeling – Motivation • Data models should facilitate smoother automation/integration between different design tools • Should provide users with better capabilities to make tradeoffs/optimizations across design domains and between tools within a domain VLSI CAD seminar 2004

  32. Data Modeling – Motivation • Nike System architecture goals • Enable a flexible configuration that facilitates a plug&play system architecture • Achieve semantic consistency in data representation across layout, circuit & logic domains • Promote reuse of common components • Insulate individual CAD capabilities from persistent storage issues VLSI CAD seminar 2004

  33. Data Model Architecture • On representation multiple CAD tools act as producers and consumers • Layout editor is a design producer • Simulators and verification tools are consumers • Data flow between producers and consumers may transcend domain boundaries VLSI CAD seminar 2004

  34. Data Model Architecture – cont. • Diversity of elements and need for multi-way data sharing complicates Data modeling for CAD • A modeling framework that can accommodate an assortment of data types and be extensible to new types is required VLSI CAD seminar 2004

  35. Data Model Architecture – cont. • The solution • Framework that allows multiple levels of data models VLSI CAD seminar 2004

  36. Data Model Architecture – cont. • Core (UCM) – serves as common vector between logic, circuit and layout domains • Each domain extend the UCM with the domain’s specific data • CAD apps. can directly operate on UCM, a domain model or an application view • Unified domain models allows reuse of engines and minimizes I/O overhead VLSI CAD seminar 2004

  37. Data Model Architecture – cont. VLSI CAD seminar 2004

  38. Data Model Architecture – cont. • UCM serves as basis for enabling data-driven interactions between capabilities • Mapper modules – translates design objects between domains • incremental data exchange (by modules and methodologies that can provide this service) VLSI CAD seminar 2004

  39. Data Model Architecture – cont. • Summary – two key architectural goals have been addressed here • Facilitate interactive interactions across CAD tool boundaries • Enable incremental iterations through the design cycle VLSI CAD seminar 2004

  40. Short summary - up to this point • LaMA model for setting modularity • SW development principles and how they were implemented • The Data model and it’s unique significance VLSI CAD seminar 2004

  41. Software Infrastructure • Nike architecture requires strong focus on software quality • Common data model, high-level of module reuse magnify software defects • Complexity of multi-site development VLSI CAD seminar 2004

  42. Standard Development Env. (SDE) • Set Plan of Record (POR) • SDE POR lists all development tools and their versions that should be installed for each developer • Critical to ensure all code libraries are compatible and common methodologies can be implemented VLSI CAD seminar 2004

  43. Configuration Management • Code management tool – Rational’s ClearCase • Dynamic view • Parallel development • One virtual repository • Methodology of usage was set • Directory structure for the code and libraries for tagging of versions • naming conventions for branches • All Nike developers use identical setup, and can share & navigate the same source code (including multi-sited – Rational’s Multisite) VLSI CAD seminar 2004

  44. Software Development Cycle • Iterative software development process includes design, iterate and Test (DIT) • Each project breaks developments to < 12 weeks DIT cycles • Enables frequent points of synchronization for validation • Once a quarter - full synchronization of the tool to the same version on all sites VLSI CAD seminar 2004

  45. SW Development Cycle – cont. • Goal – as common modules mature and stabilize, to get to weekly synchronization and eventually daily • Bug-tracking system is used to track down defects to achieve the required stability VLSI CAD seminar 2004

  46. SW Development Cycle – cont. • Require all Nike tools to write a Market Requirements Document, Product Proposal, External Product Spec, Internal Product Spec and Test Plan • Each document has to be approved by the key customers, the system architects and the software architects VLSI CAD seminar 2004

  47. VLSI CAD seminar 2004

  48. Validation • Nike was going for a breakthrough in both quality and productivity • External vendors did not meet the specific demands for a test management system that will insure achieving these goals • Olympus – an internal test management tool developed by DT VLSI CAD seminar 2004

  49. Conclusion • Nike was introduced with the intension to be a breakthrough in the CAD world by setting new standards of robustness and modular software architecture that are essential for developing the next generations of modern chips VLSI CAD seminar 2004

  50. Conclusion – cont. • We have seen how the need for this architecture evolved • We have seen how the architecture was designed and implemented • We have seen how the development process was set and executed VLSI CAD seminar 2004

More Related