Synthesizing Optimal Parallelism Placement and Reduction Strategies on Hierarchical Systems for Deep Learning

no code implementations • 20 Oct 2021 • Ningning Xie, Tamara Norman, Dominik Grewe, Dimitrios Vytiniotis

We present a novel characterization of the mapping of multiple parallelism forms (e. g. data and model parallelism) onto hierarchical accelerator systems that is hierarchy-aware and greatly reduces the space of software-to-hardware mapping.

Program Synthesis

Coherence of Type Class Resolution

no code implementations • 1 Jul 2019 • Gert-Jan Bottu, Ningning Xie, Koar Marntirosian, Tom Schrijvers

Inspired by the notion of full abstraction, we present a two-step strategy that first elaborates nondeterministically into an intermediate language that preserves contextual equivalence, and then deterministically elaborates from there into the target language.

Programming Languages 68N15, 68N18 D.3.1