Bian, Kun (2013-12). Efficient Path Delay Test Generation with Boolean Satisfiability. Doctoral Dissertation. Thesis uri icon

abstract

  • This dissertation focuses on improving the accuracy and efficiency of path delay test generation using a Boolean satisfiability (SAT) solver. As part of this research, one of the most commonly used SAT solvers, MiniSat, was integrated into the path delay test generator CodGen. A mixed structural-functional approach was implemented in CodGen where longest paths were detected using the K Longest Path Per Gate (KLPG) algorithm and path justification and dynamic compaction were handled with the SAT solver. Advanced techniques were implemented in CodGen to further speed up the performance of SAT based path delay test generation using the knowledge of the circuit structure. SAT solvers are inherently circuit structure unaware, and significant speedup can be availed if structure information of the circuit is provided to the SAT solver. The advanced techniques explored include: Dynamic SAT Solving (DSS), Circuit Observability Don't Care (Cir-ODC), SAT based static learning, dynamic learnt clause management and Approximate Observability Don't Care (ACODC). Both ISCAS 89 and ITC 99 benchmarks as well as industrial circuits were used to demonstrate that the performance of CodGen was significantly improved with MiniSat and the use of circuit structure.
  • This dissertation focuses on improving the accuracy and efficiency of path delay test generation using a Boolean satisfiability (SAT) solver. As part of this research, one of the most commonly used SAT solvers, MiniSat, was integrated into the path delay test generator CodGen. A mixed structural-functional approach was implemented in CodGen where longest paths were detected using the K Longest Path Per Gate (KLPG) algorithm and path justification and dynamic compaction were handled with the SAT solver.

    Advanced techniques were implemented in CodGen to further speed up the performance of SAT based path delay test generation using the knowledge of the circuit structure. SAT solvers are inherently circuit structure unaware, and significant speedup can be availed if structure information of the circuit is provided to the SAT solver. The advanced techniques explored include: Dynamic SAT Solving (DSS), Circuit Observability Don't Care (Cir-ODC), SAT based static learning, dynamic learnt clause management and Approximate Observability Don't Care (ACODC). Both ISCAS 89 and ITC 99 benchmarks as well as industrial circuits were used to demonstrate that the performance of CodGen was significantly improved with MiniSat and the use of circuit structure.

publication date

  • December 2013
  • December 2013