Shared-Memory Hierarchical Process Mapping

Christian Schulz, Henning Woydt · 2025

Read Shared-Memory Hierarchical Process Mapping by Christian Schulz, Henning Woydt (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-2504-01726.txt.

Description

Schulz and Woydt present a shared-memory algorithm for hierarchical process mapping on supercomputers, in a paper from twenty twenty-five. Large scientific codes split into thousands of communicating tasks. Mapping tasks to hierarchical hardware must balance load and cut communication cost. Their method partitions the communication model by hierarchical multisection in parallel.

The algorithm wins best solution quality on ninety-five percent of instances. It also runs marginally faster than the next best rival. Even in serial mode, it beats prior serial methods on both quality and speed. Placing chatty task groups on nearby processors drives the savings. The work targets islands, racks, nodes, and cores as mapping levels.

Catalog record

CREATORChristian Schulz, Henning Woydt
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE58 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2504-01726

Browse the full library · About the Tower of Babel