Councilman, Aaron2021-05-272021-05-272021https://hdl.handle.net/11299/220246Modern processors are multicore and this trend is only likely to increase in the future. To truly exploit the power of modern computers, programs need to take advantage of multiple cores by exploiting parallelism. Writing parallel programs is difficult not only because of the inherent difficulties in ensuring correctness but also because many languages, especially low-level languages like C, lack good abstractions and rather rely on function calls. Because low-level imperative languages like C remain dominant in systems programming, and espe- cially in high-performance applications, developing parallel programs in C is important, but its reliance on function calls results in boiler-plate heavy code. This work intends to reduce the need for boiler-plate by introducing higher-level syntax for parallelism, and it does so in such a manner so as to decouple the implementation of the parallelism from its semantics, allowing programmers to reason about the semantics of their program and separately tune the implementation to fi nd the best performance possible. Furthermore, this work does so in an extensible manner, allowing new implementations of parallelism and synchronization to be developed independently and allowing programmers to use any selection of these imple- mentations that they wish. Finally, this system is flexible and allows new abstractions for parallel programming to be built on top of it and bene t from the varied implementations while also providing programmers higher-level abstractions. This system can also be used to combine different parallel programming implementations in manners that would be difficult without it, and does so while still providing reasonable runtime performance.enSumma Cum LaudeComputer ScienceCollege of Science and EngineeringThesis or Dissertation