A Parallel Scan Algorithm in the Tensor Core Unit Model

Abstract summary

The paper expresses parallel prefix sums using fixed-size matrix multiplications in the Tensor Core Unit computational model. It derives bounds on algorithm depth, execution time and the number of matrix operations as functions of input size, matrix size, latency and available tensor cores.

Abstract on arXiv

DOI arXiv PDF

Cite this work

@misc{aparallelscanalgorithminthetensorcoreunitmodel,
  title = {A Parallel Scan Algorithm in the Tensor Core Unit Model},
  author = {Anastasios Zouzias and William F. McColl},
  year = {2023},
  note = {Euro-Par},
  url = {https://doi.org/10.1007/978-3-031-39698-4_33}
}

Download