Memory-efficient array redistribution through portable collective communication
This addresses the bottleneck of array redistribution in large-scale deep learning workloads, offering improved efficiency and memory safety for practitioners using SPMD parallelism.
The paper tackles the problem of efficiently redistributing multi-dimensional array data in SPMD computations for deep learning, presenting a type-directed synthesis approach that achieves a geometric mean speedup of 1.22x and up to 5.7x compared to an existing implementation while providing provable memory guarantees.
Modern large-scale deep learning workloads highlight the need for parallel execution across many devices in order to fit model data into hardware accelerator memories. In these settings, array redistribution may be required during a computation, but can also become a bottleneck if not done efficiently. In this paper we address the problem of redistributing multi-dimensional array data in SPMD computations, the most prevalent form of parallelism in deep learning. We present a type-directed approach to synthesizing array redistributions as sequences of MPI-style collective operations. We prove formally that our synthesized redistributions are memory-efficient and perform no excessive data transfers. Array redistribution for SPMD computations using collective operations has also been implemented in the context of the XLA SPMD partitioner, a production-grade tool for partitioning programs across accelerator systems. We evaluate our approach against the XLA implementation and find that our approach delivers a geometric mean speedup of $1.22\times$, with maximum speedups as a high as $5.7\times$, while offering provable memory guarantees, making our system particularly appealing for large-scale models.