Back to Research

Indian Institute of Technology Madras   Indian Institute of Technology Mandi  

Project 1157 - ValFly: Program analysis to make value types fly


Running from 2022 to present

ValFly: Program analysis to make value types fly


Research team:

  • IBM Project Lead (RCL): Tobi Ajila, IBM
  • IBM Manager (RCM): Tony Renaud, IBM
  • IBM Sponsor (RCS): Vijay Sundaresan, IBM
  • IBM Contributor (RCC): Daryl Maier, IBM
  • IBM Contributor (RCC): Hang Shao, IBM

Programmers often wish to have the performance of primitives for user defined datatypes. The major reason why ADTs have a larger cost when compared to primitives is that objects are allocated in the heap hence requiring an object header they must be referred to via a memory indirection. Value types (Project Valhalla) solve this issue by flattening eligible objects into their container, thus reducing the cost of indirection. The primary objective of our project is to selectively decide which value types are good candidates for object flattening in the JVM, along with expanding the optimizations possible in this scope. Our proposed solution can be used to perform a combination of static and dynamic analyses (promulgated by the PYE framework, TOPLAS 2019) to

  • Identify various access patterns under which fields of value types should be flattened in their containers.
  • Build an appropriate flattening strategy for such objects in Eclipse OpenJ9.
  • Perform various related optimizations in the scope of value types.
  • Learn More about the Research Team.  


Indian Institute of Technology Madras    Indian Institute of Technology Mandi