Bridging the Gap Between General-Purpose and Domain-Specific Compilers with Synthesis
Name
Bridging the gap.pdf
Size
502.38 KB
Format
Adobe PDF
Checksum (MD5)
d91c42ff0c92c4ca24d0647a54d41e8e
Author(s)
Cheung, Alvin
Kamil, Shoaib
Solar Lezama, Armando
Date Issued
2015
Journal
Leibniz International Proceedings in Informatics (LIPIcs)
Publisher
Schloss Dagstuhl--Leibniz-Zentrum fuer Informatik
Citation
Cheung, Alvin, Shoaib Kamil, and Armando Solar-Lezama. "Bridging the Gap Between General-Purpose and Domain-Specific Compilers with Synthesis." Leibniz International Proceedings in Informatics (LIPIcs), vol. 32, 2015, pp. 51-61.
Version
Final published version
Abstract
This paper describes a new approach to program optimization that allows general purpose code to benefit from the optimization power of domain-specific compilers. The key to this approach is a synthesis-based technique to raise the level of abstraction of general-purpose code to enable aggressive domain-specific optimizations. We have been implementing this approach in an extensible system called Herd. The system is designed around a collection of parameterized kernel translators. Each kernel translator is associated with a domain-specific compiler, and the role of each kernel translator is to scan the input code in search of code fragments that can be optimized by the domain-specific compiler embedded within each kernel translator. By leveraging general synthesis technology, it is possible to have a generic kernel translator that can be specialized by compiler developers for each domain-specific compiler, making it easy to build new domain knowledge into the overall system. We illustrate this new approach to build optimizing compilers in two different domains, and highlight research challenges that need to be addressed in order to achieve the ultimate vision.
MIT Department
Massachusetts Institute of Technology. Computer Science and Artificial Intelligence Laboratory
Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science
Terms of Use
Creative Commons Attribution 4.0 International License
Persistent DSpace Link
DOI of Published Version
https://doi.org/10.4230/LIPIcs.SNAPL.2015.51
