Kisan Call Centre - To harness potential of ICT in Agriculture by answer farm...
Plenary Session: application drive design alberto sv
1. Application Driven Design – New Directions Require New Tools! Alberto Sangiovanni-Vincentelli The Edgar L. and Harold H. Buttner Chair of EECSUniversity of California at Berkeley Co-Founder, CTA and Member of the BoardCadence Design Systems
26. 6 Challenges System Debugging is a Difficult Problem Early Software Development is the limiting factor in TTM Everybody has growing software content Software never works the first time, debugging complex system interactions is very difficult There is a gap between hardware and software engineers Debugging tools and techniques are very different We are in the middle of an Industry Inflection Point Shift from proprietary, vendor controlled Virtual Platform models to open, interoperable, standards based models The SystemC TLM2 standard has resulted in large scale migration towards a standards based approach to Virtual Platform creation Desire to avoid vendor lock-in and maintain tool independence Need to maintain control of source code for important IP models Virtual Platform Solution needed with extensive system-level debug and analysis to help find the root cause of difficult functional, performance, and power issues 6 6
34. Easy to distribute to many usersControllability, Observability, Repeatability May 3, 2011 Cadence Confidential: Shared under NDA only 7
35. 8 Tool Limitations Today Do not solve System Complexity or Debugging Crisis None have enough knowledge of both hardware and software stack Most do not have enough performance to simulate full system Outdated technology with poor performance Processor models take too long to create and are not fast enough Vendors asking as much as $1M for models that do not have state-of-the-art performance Missing Model Syndrome Proprietary systems put all the pressure on the vendor to provide all needed models, vendors did not allow users to change models Users are less interested to do model creation for closed system, limited benefit for the work Limits 3rd party ecosystem of model providers Unconnected to implementation, disconnected from actual design flow Simulator does not support connections to other models such as RTL simulation Understaffed side projects producing throw away models Adhoc / Limited Scope Solutions Architectural trade-off market is too small Software Engineering-only solution also is too small Must address System Debugging and Verification Problem to provide real value
44. Sensing based user generated content3.3 billion mobile device subscriptions worldwide in 2007 [Courtesy J. Shen, Nokia Research Center Palo Alto] 12
60. Platform-Based Design Application Space Architectural Space Application Instance Platform Instance Platform Design-Space Export Platform Mapping Platform: library of resources defining an abstraction layer Resources do contain virtual components i.e., place holders that will be customized in the implementation phase to meet constraints Very important resources are interconnections and communication protocols
62. The Way Forward Electronic and System Industry facing an array of complex problems from design to manufacturing involving complexity, power, reliability, re-configurability, integration, embedded software Design Methods and Tools lacking: active research field Innovation of this magnitude is difficult to achieve Collaboration is needed between governments, IC, and system industry, otherwise the situation is bound to become more critical. Not an issue of languages or point tools!
63. The Way Forward Electronic Industry facing an array of complex problems from design to manufacturing involving complexity, power, reliability, reconfigurability, integration, embedded software Design Methods and Tools lacking: active research field EDA vendors have to extend their reach into the system space Innovation of this magnitude is difficult to achieve Collaboration is needed between government, IC, and system industry, otherwise the situation is bound to become more critical. Not an issue of languages or point tools! Collaborate to Innovate