Ken Kennedy (1945–2007) was an American computer scientist who pioneered optimizing compilers and parallel programming languages. His work enabled scientists and engineers to harness supercomputer power without mastering low-level parallel programming.
High Performance Fortran
Kennedy led the High Performance Fortran Forum, creating HPF as a standard for data-parallel programming. HPF aimed to let programmers write simple code while compilers handled the complexity of distributing computation across processors.
Compiler Research
Kennedy’s research at Rice University advanced:
- Dependence analysis: Understanding which computations can run in parallel
- Loop transformations: Optimizing loops for cache and parallelism
- Interprocedural analysis: Optimizing across function boundaries
- Automatic parallelization: Converting sequential code to parallel
Rice University
Kennedy founded and built Rice’s computer science department into a leading center for compiler and parallel computing research. Many influential compiler researchers trained under his guidance.
Industry Impact
Kennedy’s compiler techniques were adopted by:
- Intel, IBM, and other processor vendors
- High-performance computing centers
- Scientific computing applications
- Modern compiler frameworks
Recognition
Kennedy received the ACM-IEEE Ken Kennedy Award (named after him posthumously), the ACM Software System Award, and numerous other honors for his contributions to high-performance computing.
Legacy
Kennedy’s vision was that parallel programming should be accessible to domain scientists, not just parallel computing experts. His research continues to influence how we compile code for modern multi-core and distributed systems.