LOOPerSet

A large-scale dataset for data-driven polyhedral compiler optimization

LOOPerSet: A Large-Scale Dataset for Data-Driven Polyhedral Compiler Optimization

An open-source dataset of 28 million datapoints for program speed-up prediction.

Accelerated data generation by 5× through pruning and other data engineering techniques. Released as an open-source resource for the polyhedral compilation community.

  • Role: Research Assistant at NYU Abu Dhabi
  • Status: Released
  • Paper: arXiv:2510.10209

References