Search papers, labs, and topics across Lattice.
This paper extends the ItoyoriFBC Asynchronous Many-Task (AMT) runtime by implementing a promise-future model that allows for dynamic task dependencies, overcoming the limitations of the traditional future-only model. This enhancement enables the expression of more complex algorithms, such as Hierarchical LU factorization (HLU), which were previously constrained by fixed task dependencies. Experimental results demonstrate significant performance improvements, achieving a 15.6x speedup on up to 16 nodes, showcasing the model's efficiency in handling dynamic workloads.
Dynamic task dependencies can now be expressed in AMT runtimes, leading to a staggering 15.6x speedup in complex algorithms like Hierarchical LU factorization.
Asynchronous Many-Task (AMT) runtimes use futures as placeholders for values produced by other tasks. In the ItoyoriFBC AMT runtime, the existing future-only model binds each future to its producer at creation time and requires the number of tasks that read each future to be fixed at compile time. This prevents directly expressing algorithms that create dependencies dynamically. We extend ItoyoriFBC with an implementation of a promise-future model that lifts these limitations. Thereby, our ItoyoriFBC variant supports dynamic algorithms such as Hierarchical LU factorization (HLU). We experimentally evaluated our implementation using HLU on up to 16 nodes and observed near-ideal scaling with a 15.6x speedup.
