MDE based FPGA physical Design Fast prototyping with Smalltalk

MDE based FPGA physical Design Fast prototyping with Smalltalk

• 1.
  MDE based FPGAphysical Design Fast prototyping with Smalltalk Ciprian Teodorov, Loïc Lagadec Loic.lagadec@univ-brest.fr Lab-STICC MOCS UMR 3192
• 2.
  FPGAs “Flexible” hardware Hard to program Time to market Hard to debug i1 Compute node LUT LUT LUT I1-i2 i2 E/S µP i1 LUT LUT i2 I1+i2 LUT Programmable LUT LUT LUT interconnection
• 3.
  FPGAs “Flexible” hardware Hard to program Time to market Hard to debug EDA required ! •  to circuit C •  ebug D •  enchmarking B
• 4.
  Our Smalltalk-based EDAlegacy
• 5.
  Legacy backfires "   Early developments (MADEO) started in 1996 "   Fast evolving domain (Moore + Murfy) "   Refactoring is not enough to keep in the race "   We have to re-design our framework
• 6.
  New direction "   We need to shift from "   a generic solution to be tailored on demand "   To "   a repository of model, algorithms, components "   In order to deliver "   Performances "   Scalability "   Flexibility "   Durability
• 7.
  LEGACY
• 8.
  Front end C code "   High level synthesis (compilation) "   Ressources allocation (logic synthesis) Circuit
• 9.
  Programming an FPGAin 4 steps
• 10.
  ADL Based EDAgenerators Spécification Architecture Architecture specification ADL Description Exploration Application Synthesizer Bitstream HW HW Configuration Compiler P&R Testbenches Testbenches generator Prototype Prototype Controller Compilation Synthesis/Compilation Simulation Simulation Validation Validation 10
• 11.
  Our flow Context Resources Zone Zone ADL Reconfigurable Description Zone zones description Behavioral code Bitstream Resource Configuration model model model Architecture Configuration Bistream VHDL controller Prototype 11 Simulation & synthesis
• 12.
  Some examples 12
• 13.
  RE-DESIGN
• 14.
  Goal oriented viewextraction
• 15.
  Tool engine
• 16.
  Models as commonvocabulary
• 17.
  Combinational circuit modeling
• 18.
  Target modeling
• 19.
  Re-design / copydown
• 20.
  CONCLUSION
• 21.
  Let’s try tosummarize "   Succes: target, tool flow
• 22.
  Conclusion "   Future work: "   Tools integration (eg Mondrian integration) "   Performances improvement "   Test coverage "   Algorithm pick and play GUI Thank you for your attention