# Specifying and Executing Optimizations for Parallel Programs

**Authors:** William Mansky, Dennis Griffith, Elsa L. Gunter  
**arXiv:** [1407.7932](https://arxiv.org/abs/1407.7932)  
**Date:** 2014-07-30  
**Summary page:** https://www.summarizepaper.com/en/paper/1407.7932v1/

## Key points

- Authors William Mansky, Dennis Griffith, and Elsa L. Gunter focus on compiler optimizations for parallel programs
- Introduce PTRANS, a domain-specific language for formal specification of compiler transformations
- PTRANS syntax enables cleaner and more comprehensible specification of program optimizations
- Executable semantics of PTRANS allow for testing and refinement before integration into a compiler
- Showcase utility of PTRANS by specifying, testing, and refining a redundant store elimination optimization tailored for parallel programs

## Summary

, , , , 

In their paper titled "Specifying and Executing Optimizations for Parallel Programs," authors William Mansky, Dennis Griffith, and Elsa L. Gunter delve into the intricate world of compiler optimizations for parallel programs. Compiler optimizations are essential components of modern compilers, typically expressed as rewrites on program graphs. The authors introduce PTRANS, a domain-specific language designed for the formal specification of compiler transformations to address challenges in compiling parallel programs. They describe its executable semantics and emphasize its fundamental approach of representing program transformations as rewrites on control flow graphs with temporal logic side conditions. The syntax of PTRANS enables cleaner and more comprehensible specification of program optimizations, while its executable semantics allow for testing and refinement before integration into a compiler. The authors showcase the utility of PTRANS by using it to state, test, and refine the specification of a redundant store elimination optimization specifically tailored for parallel programs. Through their research presented in this paper, Mansky et al. provide valuable insights into enhancing compiler optimizations for parallel computing environments, ultimately contributing to improved efficiency and performance in parallel program compilation processes.