High-Performance Computing (HPC) applications require extensive testing with thousands of tests executed on supercomputers and distributed clusters. While traditional software testing is relatively cheap, HPC testing costs are orders of magnitude higher due to greater computational demands and longer execution times. Existing test prioritization techniques target traditional software systems using execution time as a proxy for energy consumption, but remain unexplored for HPC applications where this correlation breaks down due to parallel efficiency, communication overhead, and accelerator usage. We propose a novel multi-objective optimization approach that treats energy consumption as a first-class objective alongside code coverage for HPC test prioritization. Using LAMMPS, a widely-used molecular dynamics simulator, we demonstrate 97.5% average energy savings compared to exhaustive testing while maintaining 95% average statement coverage of modified code, suggesting a promising direction for sustainable HPC testing.

@inproceedings{AnklesariaM26,
  title = {Energy-Aware Test Prioritization for High-Performance Computing: A Multi-Objective Approach},
  author = {Ninad Anklesaria and Manish Motwani},
  booktitle = {FSE},
  year = {2026},
  doi = {10.1145/3803437.3805582}
}