Modeling black-box components with probabilistic synthesis

Abstract

This paper is concerned with synthesizing programs based on black-box oracles: we are interested in the case where there exists an executable implementation of a component or library, but its internal structure is unknown. We are provided with just an API or function signature, and aim to synthesize a program with equivalent behavior.To attack this problem, we detail Presyn: a program synthesizer designed for flexible interoperation with existing programs and compiler toolchains. Presyn uses high-level imperative control-flow structures and a pair of cooperating predictive models to efficiently narrow the space of potential programs. These models can be trained effectively on small corpora of synthesized examples.We evaluate Presyn against five leading program synthesizers on a collection of 112 synthesis benchmarks collated from previous studies and real-world software libraries. We show that Presyn is able to synthesize a wider range of programs than each of them with less human input. We demonstrate the application of our approach to real-world code and software engineering problems with two case studies: accelerator library porting and detection of duplicated library reimplementations.CCS Concepts: • Software and its engineering → General programming languages; Automatic programming; Programming by example; Genetic programming.

Learning C to x86 Translation: An Experiment in Neural Compilation

“…For this work, we base our dataset on a subset 5 of 64 functions extracted from the program synthesis benchmark collated in Collie et al [4]. 6 We then add a main function with the required input/output calls to execute them with randomly generated input/output pairs (referred as IO examples, from now on).…”

“…Specifically, subword-nmt [5] . 4 This has the benefit of decreasing the vocabulary size while making out-of-vocabulary tokens virtually impossible (since unknown tokens can be reconstructed from ASCII characters or other subwords present in the vocabulary).…”

“…The best model can only compile correctly 1 about 33% of the functions in a benchmark built from an existing program synthesis evaluation set [4]; it specially struggles to compile functions with numerous arguments and arrays. Given the complexity of the problem, we make all the resources and code generated in this work publicly available.…”

C2TACO: Lifting Tensor Code to TACO

“…This makes this verification impossible to embed into a synthesis loop where we check thousands of candidates. Our synthesis must use alternatives like observational equivalence [22] in order to achieve the necessary scalability.…”

Bind the gap: compiling real software to hardware FFT accelerators

“…Constraint matching [27,45,57] provides a way of matching and extracting interfaces from high-level code. Unfortunately, these approaches are brittle [45] -they do not scale beyond a single implementation/accelerator pair, and constraint patterns are extremely hard to write [39,40,58]. Rewrite-rule based compilers can be used to target accelerators [77], but these still rely on initial matching using constraints or similar.…”