LOOPer_artifact_PACT25

LOOPer: PACT 2025 Artifact Evaluation This artifact contains the source code, pre-trained cost model, evaluation scripts, and a containerized environment to reproduce the key results presented in the PACT 2025 paper: "LOOPer: A Learned Automatic Code Optimizer For Polyhedral Compilers". LOOPer is a deep learning-based polyhedral autoscheduler that expands the scope of optimizable programs and the space of explored transformations. This artifact provides all the necessary components to:1. Run the LOOPer autoscheduler with its deep learning cost model.2. Generate performance results for the 30 benchmarks in the PolyBench suite.3. Compare LOOPer's performance against the Pluto compiler, a state-of-the-art baseline.4. Verify the core claims of the paper.