A Transducers-based Programming Framework for Efficient Data Transformation

Many data analytics and scientific applications rely on data transformation tasks, such as encoding, decoding, parsing of structured and unstructured data, and conversions between data formats and layouts. Previous work has shown that data transformation can represent a performance bottleneck for da...

Celý popis

Uložené v:
Podrobná bibliografia
Vydané v:2024 33rd International Conference on Parallel Architectures and Compilation Techniques (PACT) s. 66 - 77
Hlavní autori: Nguyen, Tri, Becchi, Michela
Médium: Konferenčný príspevok..
Jazyk:English
Vydavateľské údaje: ACM 13.10.2024
Predmet:
Codes
Computational modeling
Computer languages
Data analysis
Data models
Engines
Graphics processing units
Optimization
Programming
Transducers
On-line prístup:Získať plný text
Tagy: Pridať tag
Žiadne tagy, Buďte prvý, kto otaguje tento záznam!
Popis
Shrnutí:Many data analytics and scientific applications rely on data transformation tasks, such as encoding, decoding, parsing of structured and unstructured data, and conversions between data formats and layouts. Previous work has shown that data transformation can represent a performance bottleneck for data analytics workloads. The transducers computational abstraction can be used to express a wide range of data transformations, and recent efforts have proposed configurable engines implementing various transducer models (from finite state transducers, to pushdown transducers, to extended models). This line of research, however, is still at an early stage. Notably, expressing data transformation using transducers requires a paradigm shift, impacting programmability. To address this problem, we propose a programming framework to map data transformation tasks onto a variety of transducer models. Our framework includes: (1) a platform agnostic programming language (xPTLang) to code transducer programs using intuitive programming constructs, and (2) a compiler that, given an xPTLang program, generates efficient transducer processing engines for CPU and GPU. Our compiler includes a set of optimizations to improve code efficiency. We demonstrate our framework on a diverse set of data transformation tasks on an Intel CPU and an Nvidia GPU.
DOI:10.1145/3656019.3676891