Search results
Alexia Massalin (formerly Henry Massalin) is an American computer scientist and programmer. She pioneered the concept of superoptimization, and designed the Synthesis kernel, a small kernel with a Unix compatibility layer that makes heavy use of self-modifying code for efficiency.
Jul 2, 2019 · Alexia Massalin’s 1992 PhD thesis has long been one of my favorites. It promotes the view that operating systems can be much more efficient than then-current operating systems via runtime code generation, lock-free synchronization, and fine-grained scheduling.
Jul 3, 2019 · "Alexia Massalin’s 1992 PhD thesis has long been one of my favorites. It promotes the view that operating systems can be much more efficient than then-current operating systems via runtime code generation, lock-free synchronization, and fine-grained scheduling.
- The Art of Specification
- The First Wave of Synthesis: Deductive Reasoning
- The Second Wave of Synthesis: Inductive Reasoning
- The Third Wave of Synthesis: Statistical Reasoning
- What’s Next?
Before the computer can automatically generate a program, we need to give it a specification of what the program should do. The specification needs to describe the program’s desired behavior in enough detail to ensure the program does what we want, but ideally, it should be much simpler than the program itself, since otherwise we may as well just w...
The earliest examples of program synthesis used formal specifications to automatically deduce a correct program. The idea is to take a formal specification and work backwards, searching for a program that provably satisfies the specification. Some of these early synthesizers built on the successes of automated theorem provers, using them to automat...
The disadvantage of deductive synthesis is that it requires both a complete axiomatization of the target language and a complete formal specification. The second wave of synthesis research focuses on inductivesynthesis to try to relax these requirements. The idea of inductive reasoning is to find programs that work on a subset of inputs and general...
The recent surge in the effectiveness of machine learning has driven interest in program synthesis using statistical techniques. One approach is to train a model to guide existing synthesis techniques more effectively—for example, DeepCoderuses a learned model to guide the brute-force enumeration of candidate programs. These learned models replace ...
Synthesis techniques are notorious for poor scalability—working on small examples (e.g., Excel macros), but not on many of the types of programs we care about. Tackling the scalability challenge is one of the most important directions for synthesis research. There’s been recent work on profiling synthesis tools and on automatically inferring good s...
Mar 10, 2021 · Alexia Massalin coined the term “ superoptimization ” in her 1987 paper Superoptimizer – A look at the Smallest Program. Massalin used a (pruned) exhaustive search to find the shortest implementation of various straight line computations in the 68000 instruction set.
The Synthesis kernel presented in Alexia Massalin's Ph.D. thesis [17] [18] is a tiny Unix kernel that takes a structured, or even object oriented, approach to self-modifying code, where code is created for individual quajects, like filehandles.
Affiliations: [MicroUnity Systems Engineering, Inc., Santa Clara, CA, USA].
