Aligning intent and behavior in software systems: how programs communicate & their distribution and organization

Dietz, William Barrett

Permalink

https://hdl.handle.net/2142/107994 Copy

Description

Title

Aligning intent and behavior in software systems: how programs communicate & their distribution and organization

Author(s)

Dietz, William Barrett

Issue Date

2020-05-05

Director of Research (if dissertation) or Advisor (if thesis)

Adve, Vikram S

Doctoral Committee Chair(s)

Adve, Vikram S

Committee Member(s)

• Regehr, John D
• Xie, Tao
• Misailovic, Sasa

Department of Study

Computer Science

Discipline

Computer Science

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

Ph.D.

Degree Level

Dissertation

Date of Ingest

2020-08-26T21:54:54Z

Keyword(s)

• Compilers
• Link-Time Optimization
• Slipstream
• Software Multiplexing
• ALLVM
• LLVM
• LLVM IR
• Nix

Abstract

Managing the overwhelming complexity of software is a fundamental challenge because complexity is the root cause of problems regarding software performance, size, and security. Complexity is what makes software hard to understand, and our ability to understand software in whole or in part is essential to being able to address these problems effectively. Attacking this overwhelming complexity is the fundamental challenge I seek to address by simplifying how we write, organize, and think about programs. Within this dissertation I present a system of tools and a set of solutions for improving the nature of software by focusing on programmer’s desired outcome, i.e. their intent. At the program level, the conventional focus, it is impossible to identify complexity that, at the system level, is unnecessary. This “accidental complexity” includes everything from unused features to independent implementations of common algorithmic tasks. Software techniques driving innovation simultaneously increase the distance between what is intended by humans – developers, designers, and especially the users – and what the executing code does in practice. By preserving the declarative intent of the programmer, which is lost in the traditional process of compiling and linking and building software, it is easier to abstract away unnecessary details. The Slipstream, ALLVM, and software multiplexing methods presented here automatically reduce complexity of programs while retaining intended function of the program. This results in improved performance at both startup and run-time, as well as reduced disk and memory usage.

Graduation Semester

2020-05

Type of Resource

Thesis

Permalink

http://hdl.handle.net/2142/107994

Copyright and License Information

Copyright 2020 William B. Dietz

Owning Collections