Abstract
Fast numerical libraries have been a cornerstone of scientific computing for decades, but this comes at a price. Programs may be tied to vendor specific software ecosystems resulting in polluted, non-portable code. As we enter an era of heterogeneous computing, there is an explosion in the number of accelerator libraries required to harness specialized hardware. We need a system that allows developers to exploit ever-changing accelerator libraries, without over-specializing their code.
As we cannot know the behavior of future libraries ahead of time, this paper develops a scheme that assists developers in matching their code to new libraries, without requiring the source code for these libraries. Furthermore, it can recover equivalent code from programs that use existing libraries and automatically port them to new interfaces. It first uses program synthesis to determine the meaning of a library, then maps the synthesized description into generalized constraints which are used to search the program for replacement opportunities to present to the developer.
We applied this approach to existing large applications from the scientific computing and deep learning domains. Using our approach, we show speedups ranging from 1.1× to over 10× on end to end performance when using accelerator libraries.
As we cannot know the behavior of future libraries ahead of time, this paper develops a scheme that assists developers in matching their code to new libraries, without requiring the source code for these libraries. Furthermore, it can recover equivalent code from programs that use existing libraries and automatically port them to new interfaces. It first uses program synthesis to determine the meaning of a library, then maps the synthesized description into generalized constraints which are used to search the program for replacement opportunities to present to the developer.
We applied this approach to existing large applications from the scientific computing and deep learning domains. Using our approach, we show speedups ranging from 1.1× to over 10× on end to end performance when using accelerator libraries.
Original language | English |
---|---|
Title of host publication | 2019 28th International Conference on Parallel Architectures and Compilation Techniques (PACT) |
Publisher | Institute of Electrical and Electronics Engineers (IEEE) |
Pages | 55-67 |
Number of pages | 13 |
ISBN (Electronic) | 978-1-7281-3613-4 |
ISBN (Print) | 978-1-7281-3614-1 |
DOIs | |
Publication status | Published - 7 Nov 2019 |
Event | 28th International Conference on Parallel Architectures and Compilation Techniques - Seattle, United States Duration: 21 Sep 2019 → 25 Sep 2019 https://pactconf.org/ |
Publication series
Name | |
---|---|
Publisher | Institute of Electrical and Electronics Engineers (IEEE) |
ISSN (Print) | 1089-795X |
ISSN (Electronic) | 2641-7936 |
Conference
Conference | 28th International Conference on Parallel Architectures and Compilation Techniques |
---|---|
Abbreviated title | PACT 2019 |
Country/Territory | United States |
City | Seattle |
Period | 21/09/19 → 25/09/19 |
Internet address |
Keywords
- C++ language
- learning (artificial intelligence)
- message passing
- parallel programming
- program compilers
- software architecture
- software libraries
- software portability
- scientific computing
- type-directed program synthesis
- constraint generation
- library portability
- vendor specific software ecosystems
- heterogeneous computing
- source code
- Libraries
- Informatics
- Synthesizers
- Semantics
- Software
- Ecosystems
- Deep learning
- program synthesis
- code rejuvenation
- constraint programming
- compilers