All publications
Optimising Darwinian Data Structures on Google Guava
2017·SSBSE 2017
Figure 1: Multi-objective search on Guava. Candidate substitutions evaluate in parallel (a) and settle onto a time–memory Pareto front (b); dominated variants are discarded.∎
Where the line began: applying multi-objective Darwinian data-structure optimization to Google Guava, one of the most widely used Java libraries. Candidate structure substitutions were evaluated in parallel on cloud infrastructure, scored on execution time, CPU, and memory simultaneously.
The study demonstrated measurable gains on production-grade library code and, as importantly, established the practical machinery — parallel evaluation, multi-objective scoring, automated validation — that the decade of work after it would refine.
Key results
- Measurable time, CPU, and memory gains on Google Guava
- Parallel cloud-based evaluation of candidate variants
- First paper in the Darwinian optimization line (SSBSE 2017)