Publications of Torsten Hoefler

Yves Baumann, Tal Ben-Nun, Maciej Besta, Lukas Gianinazzi, Torsten Hoefler, Piotr Luczynski:

Low-Depth Spatial Tree Algorithms

(In Proceedings of the 38th IEEE International Parallel and Distributed Processing Symposium (IPDPS'24), presented in San Francisco, CA, USA, IEEE, May 2024)

Abstract

Contemporary accelerator designs exhibit a high degree of spatial localization, wherein two-dimensional physical distance determines communication costs between processing elements. This situation presents considerable algorithmic challenges, particularly when managing sparse data, a pivotal component in progressing data science. The spatial computer model quantifies communication locality by weighting processor communication costs by distance, introducing a term named energy. Moreover, it integrates depth, a widely-utilized metric, to promote high parallelism. We propose and analyze a framework for efficient spatial tree algorithms within the spatial computer model. Our primary method constructs a spatial tree layout that optimizes the locality of the neighbors in the compute grid. This approach thereby enables locality-optimized messaging within the tree. Our layout achieves a polynomial factor improvement in energy compared to utilizing a PRAM approach. Using this layout, we develop energy-efficient treefix sum and lowest common ancestor algorithms, which are both fundamental building blocks for other graph algorithms. With high probability, our algorithms exhibit near-linear energy and poly-logarithmic depth. Our contributions augment a growing body of work demonstrating that computations can have both high spatial locality and low depth. Moreover, our work constitutes an advancement in the spatial layout of irregular and sparse computations.

Documents

download slides:

BibTeX

@inproceedings{,
  author={Yves Baumann and Tal Ben-Nun and Maciej Besta and Lukas Gianinazzi and Torsten Hoefler and Piotr Luczynski},
  title={{Low-Depth Spatial Tree Algorithms}},
  year={2024},
  month={May},
  booktitle={Proceedings of the 38th IEEE International Parallel and Distributed Processing Symposium (IPDPS'24)},
  location={San Francisco, CA, USA},
  publisher={IEEE},
  source={http://www.unixer.de/~htor/publications/},
}


serving: 3.21.105.209:34378	© Torsten Hoefler