PPoPP Reproducibility
Builds rigorous reproducibility packages and benchmark descriptions for parallel computing and PPoPP research papers.
PPoPP Reproducibility provides structured guidance and checklists for ensuring parallel performance experiments in PPoPP papers can be reproduced across diverse hardware architectures. It helps researchers systematically document hardware topologies, control sources of runtime non-determinism (thread pinning, NUMA policies, DVFS, warm-up runs), manage dataset provenance, and design scalable artifact evaluation packages so reviewers can reliably reproduce scaling trends and relative speedup claims.
주요 기능
01Structures detailed hardware, topology, and runtime environment specifications for parallel systems
02Guides artifact packaging and pre-submission validation passes for academic peer review
03Standardizes warm-up iterations, repeat counts, seed recording, and variance reporting
04Establishes scaled-down benchmark configurations to verify scaling trends across different machines
05Controls sources of parallel non-determinism like NUMA placement, thread affinity, and frequency scaling
061,090 GitHub stars
사용 사례
01Documenting multi-core, NUMA, GPU, and HPC cluster environments for reproducible research
02Auditing parallel benchmarking experiments to eliminate non-deterministic performance jitter
03Preparing artifact evaluation packages for ACM PPoPP and parallel computing conferences