ESPIRA: Improving Iterative Compilation 

Iterative optimization is the process of exploring a large space of compiler optimizations to find the best sequence of optimizations for a program based on some metric, such as improved speed, energy, size, etc. or a combination of metrics.  It has been shown that iterative optimization can outperform static approaches to compilation.  However, iterative optimization can be costly since testing many different compiler optimizations is typically required to obtain the best performance for a program. The ESPIRA project looks at techniques to improve search that can greatly diminish the cost of iterative optimization.

Selected Publications: 

[1] Iterative Optimization in the Polyhedral Model: Part II, Multidimensional Time.
Louis-Noel Pouchet, Cedric Bastoul, Albert Cohen, and John Cavazos. Acceptance: 34/184 (18%)
PLDI 2008 [gzip'd PS]  [PDF]

[2] MiDataSets: Creating The Conditions For A More Realistic Evaluation of Iterative Optimization.
Grigori Fursin, John Cavazos, Michael F.P. O'Boyle, and Olivier Temam. Acceptance: (29%)
HiPEAC 2007 [gzip'd PS]  [PDF]

[3] Using Machine Learning to Focus Iterative Optimization.
Felix Agakov, Edwin Bonilla, John Cavazos, Bjoern Franke, Grigori Fursin, Michael F. P. O'Boyle, John Thomson, Marc Toussaint, and Christopher K. I. Williams. Acceptance: 29/80 (36%)
CGO 2006 [gzip'd PS] [PDF] [Slides] Best Presentation Award

[4] Using Statistical Simulation for Studying Compiler-Microarchitecture Interactions.
Lieven Eeckhout and John Cavazos. Acceptance: 8/14 (57%)
INTERACT 2006 [gzip'd PS]  [PDF]