PICO: Performance Insights for Collective Operations DOI: XXXXXXX.XXXXXXX ISBN: 978-1-4503-XXXX-X/2018/06 CCS: Networks Network performance evaluation CCS: Hardware Testing with distributed and parallel systems CCS: Computing methodologies Parallel computing methodologies CCS: Software and its engin
Saverio Pasqualoni, Lorenzo Piarulli, Daniele De Sensi · 2025
Read PICO: Performance Insights for Collective Operations DOI: XXXXXXX.XXXXXXX ISBN: 978-1-4503-XXXX-X/2018/06 CCS: Networks Network performance evaluation CCS: Hardware Testing with distributed and parallel systems CCS: Computing methodologies Parallel computing methodologies CCS: Software and its engin by Saverio Pasqualoni, Lorenzo Piarulli, Daniele De Sensi (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2508-16809.txt.
Description
Saverio Pasqualoni and co-authors present a framework for benchmarking shared operations in large computing systems in a 2025 paper. Collective operations, where many processors exchange data at once, sit at the heart of supercomputing and large AI training. Testing them fairly is hard because modern hardware and software stacks are so complex.
Their open source framework, PICO, splits experiment setup from the run itself, so tests move across platforms. It adapts settings for two common communication libraries, gives plain reference versions of the operations, and records the system setup for repeatable comparisons. On three major supercomputers, default settings ran up to five times slower than the best choice. Tuning guided by the tool cut simulated AI training times by up to 44 percent.
Catalog record
| CREATOR | Saverio Pasqualoni, Lorenzo Piarulli, Daniele De Sensi |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 46 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2508-16809