Implementing and Optimizing a Simple, Dependently-Typed Language
Michael Blaguszewski
Abstract
Michael Blaguszewski
Abstract
This thesis presents a compiler for the simple functional programming language LambdaPi, which includes dependent types. The compiler is written in Haskell and uses LLVM, a framework for building optimizing compiler backends. It can compile the complete standard library provided by LambdaPi’s authors into native machine code. It is not much of an optimizing compiler, but several obvious opportunities for improvement exist. First I discuss the theoretical background of project: the principles of dependent types and the languages which include them. I also give a brief overview of the LLVM system and of related work. The second section describes the process of implementation, which was done in stages from a trivial calculator language up to full LambdaPi. And finally we consider opportunities for optimization. Some of these stem from Edwin Brady’s [2005] analysis of Epigram, while others are lower-level and can be performed for us by LLVM.
A significance statement is not available in the OpenAlex record.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
This thesis presents a compiler for the simple functional programming language LambdaPi, which includes dependent types. The compiler is written in Haskell and uses LLVM, a framework for building optimizing compiler backends. It can compile the complete standard library provided by LambdaPi’s authors into native machine code. It is not much of an optimizing compiler, but several obvious opportunities for improvement exist. First I discuss the theoretical background of project: the principles of dependent types and the languages which include them. I also give a brief overview of the LLVM system and of related work. The second section describes the process of implementation, which was done in stages from a trivial calculator language up to full LambdaPi. And finally we consider opportunities for optimization. Some of these stem from Edwin Brady’s [2005] analysis of Epigram, while others are lower-level and can be performed for us by LLVM.
Key concepts: Compiler, Haskell, Computer science, Programming language, Functional programming, Simple (philosophy), Calculator, Optimizing compiler