A Multi-Language Environment For Programmable Code Optimization and Empirical Tuning
PI: Qing Yi (UCCS)
Collaborators: Clint Whaley (LSU), Daniel Quinlan (Livermore)
Funding: Department of Energy
Period: 09/15/09 - 09/14/13
This project developed an integrated optimization environment for programmable code optimization and empirical tuning within the framework of existing languages. The environment uses ROSE, a source-to-source optimizing compiler at DOE/LLNL, and POET, an transformation scripting language at UTSA, to support the automated parameterization of source-to-source optimizations and the empirical tuning of applications in C, C++, and Fortran 2003. Our approach permits different levels of possible automation and programmer intervention, from fully-automated tuning of whole applications to semi-automated development of domain-specific libraries. Such an environment permits maximal impact on the performance optimization of existing and future software development, including both the optimization needs of computational kernels and the more general requirements of whole program optimizations. Our work is integrated as an external development mechanism for the widely-adopted ATLAS library and will be connected with existing empirical tuning research under DOE SciDAC PERI program.