Home Thesis and sources → Thesis Propositions [Section index]

Thesis Propositions


Propositions

accompanying the thesis

Rule-based Compilation of Data-parallel Programs

Leo Breebaart

  1. A rule-based, user-programmable compiler opens doors that remain closed when more conventional ‘black box’ compilation techniques are used.
    ( this thesis )
  2. Full and direct access to the parse tree is indispensable for efficient rewrite rules.
    ( this thesis )
  3. David Gries' program translation maxim: “Never put off until run time what you can do at compile time” (Compiler Construction for Digital Computers, 1971) is more broadly applicable: “Never generate dynamically what you can determine beforehand”. This rule should for example also be heeded when building websites.
  4. The ever-increasing complexity of software systems will only remain manageable if the implementations of these systems themselves will, to a larger extent than is currently the case, be generated by software.
  5. As users rightly judge programming languages and systems not by their potential, but by their first implementations, the quality of these implementations is crucially important.
  6. The object-based aspects of C++ are too often in the way of the template-based aspects, and vice versa.
  7. For the quality and maintainability of software the active use and development of regression tests is even more important than writing documentation.
  8. Manufacturers of consumer electronics, PDAs, and even desktop computers underestimate the extent to which the consumer bonds with the software that runs on these devices, rather than with the devices themselves.
  9. Linux is only free if your time has no value.
    ( Jamie Zawinski, http://www.jwz.org/, 1998 )

These propositions are considered defendable and as such have been approved by the supervisor, Prof. dr. ir. H.J. Sips.


[Up one level]

Leo Breebaart (leo@lspace.org)
Last updated: 27 June 2016