Multiparadigm Optimizing Retargetable Transdisciplinary Abstraction Language

Abstract

Scientists and engineers require ever more powerful software and hardware to analyze data and build models. Unfortunately, current solutions to the problem are often hard to use for scientists that are not software engineers. And software engineers often do not have the mathematical background to understand the scientific problem to solve.

This thesis describes MORTAL, a new general-purpose programming language and compiler for high-performance applications, which aims to bridge this gap by offering a multiparadigm programming environment that allows, for example, the mathematical formulae written by the scientist (perhaps using declarative programming) to be connected to the algorithms implemented by the software engineer (perhaps using object-oriented or functional programming) in a natural way, understood by both. The language will apply modern compiler and static analysis technology, along with contract programming, in new ways to both prevent bugs and improve runtime performance.

The implemented compiler is self-hosting and able to compile itself, showing that the language and its compiler, though not fully implemented yet, is already usable. The performance of MORTAL programs is also on par with the performance of C programs.

We believe MORTAL has the potential to become a useful language for solving many of the more demanding tasks of modern science.