Uday Khedker

Professor, Department of Computer Science & Engg., IIT Bombay

Research

• Areas of Interest

  Have been working in the area of compilers and program analysis since 1989. Current broad area of research is precise and scalable program analysis. topics of interest are all aspects of Data Flow Analysis (eg. interprocedural analysis, heap analysis, pointer analysis etc.).
  

• Research Group

  We have an active group of people working in the area of programming systems. Current list of active collaborators includes a couple of faculty members in CSE (Amitabha Sanyal , Supratim Biswas ), several students and project engineers. The names of the project engineers can be found on the GCC Resource Center web page.
  
  

• Publications and other technical writing

• Current Research
  

  (Last updated on 14 Dec 2022)
  There are three major strands in my current work on precise and scalable program analysis. Please see this presentation to get an overview.
   
  • The first strand explores precise pointer analysis of C/C++ programs in its full glory where precision is with respect to non-deterministic branching in programs. While the rest of the world tried to begin with scalable and imprecise methods and tries to increase their precision heuristically (and hence without precision guarantees), we begin with precise but non-scalable methods and try to increase their scalability heuristically (thereby preserving and hence guaranteeing, precision modulo non-deterministic branching).
  • The second strand explores precise and scalable interprocedural analysis where precision requires maintaining full flow- and context-sensitivity.
  • The third strand of my research proposes CoS-SSA (context-sensitive SSA) which extends the default intraprocedural SSA (which is restricted to local non-address taken variables) to global variables at the interprocedural level without any compromise on context-sensitivity even in presence pointers and recursion.
    

• Past Research

  A Generalised Theory of Exhaustive and Incremental Data Flow Analysis. This work has provided
  
  
  • First ever theoretical framework for handling bidirectional data flow problems.
    
  • A unified framework for handling unidirectional and bidirectional data flow problems.
  • A functional model for unidirectional and bidirectional incremental data flow analysis.
  • Efficient, simple, and generic algorithms for exhaustive and incremental analysis of unidirectional and bidirectional flows.
  • First ever complexity bounds for bidirectional data flows
  • More precise and more general unified complexity bounds for unidirectional and bidirectional data flows.