Using Static Allocation Algorithms for Matrix Matrix Multiplication on Multicores and GPUs

Lionel Eyraud-Dubois 1, 2 Thomas Lambert 1, 3
1 Realopt - Reformulations based algorithms for Combinatorial Optimization
LaBRI - Laboratoire Bordelais de Recherche en Informatique, IMB - Institut de Mathématiques de Bordeaux, Inria Bordeaux - Sud-Ouest
Abstract : We consider the problem of data allocation when performing matrix multiplication on a heterogeneous node, with multicores and GPUs. Classical (cyclic) allocations designed for homogeneous settings are not appropriate, but the advent of task-based runtime systems makes it possible to use more general allocations. Previous theoretical work has proposed square and cube partitioning algorithms aimed at minimizing data movement for matrix multiplication. We propose techniques to adapt these continuous square partitionings to allocating discrete tiles of a matrix, and strategies to adapt the static allocation at run-time. We use these techniques in an implementation of Matrix Multiplication based on the StarPU runtime system, and we show through extensive experiments that this implementation allows to consistently obtain a lower communication volume while improving slightly the execution time, compared to standard state-of-the-art dynamic strategies.
Complete list of metadatas

Cited literature [20 references]  Display  Hide  Download

https://hal.inria.fr/hal-01670678
Contributor : Lionel Eyraud-Dubois <>
Submitted on : Thursday, May 31, 2018 - 3:36:08 PM
Last modification on : Tuesday, May 7, 2019 - 11:42:03 AM
Long-term archiving on : Saturday, September 1, 2018 - 2:34:29 PM

File

icpp.pdf
Files produced by the author(s)

Identifiers

Citation

Lionel Eyraud-Dubois, Thomas Lambert. Using Static Allocation Algorithms for Matrix Matrix Multiplication on Multicores and GPUs. ICPP 2018 - 47th International Conference on Parallel Processing, Aug 2018, Eugene, OR, United States. ⟨10.1145/3225058.3225066⟩. ⟨hal-01670678v2⟩

Share

Metrics

Record views

195

Files downloads

351