Abstract | ||
---|---|---|
With the increased popularity of multi-GPU nodes in modern HPC clusters, it is imperative to develop matching programming paradigms for their efficient utilization. In order to take advantage of the local GPUs and the low-latency high-throughput interconnects that link them, programmers need to meticulously adapt parallel applications with respect to load balancing, boundary conditions and device synchronization. This paper presents MAPS-Multi, an automatic multi-GPU partitioning framework that distributes the workload based on the underlying memory access patterns. The framework consists of host- and device-level APIs that allow programs to efficiently run on a variety of GPU and multi-GPU architectures. The framework implements several layers of code optimization, device abstraction, and automatic inference of inter-GPU memory exchanges. The paper demonstrates that the performance of MAPS-Multi achieves near-linear scaling on fundamental computational operations, as well as real-world applications in deep learning and multivariate analysis. |
Year | DOI | Venue |
---|---|---|
2015 | 10.1145/2807591.2807611 | International Conference for High Performance Computing, Networking, Storage, and Analysis |
Keywords | Field | DocType |
Multi-GPU Programming, Memory Access Patterns | Program optimization,Uniform memory access,Programming paradigm,Computer science,Load balancing (computing),Instruction set,Parallel computing,Memory management,Memory map,CUDA Pinned memory,Distributed computing | Conference |
ISBN | Citations | PageRank |
978-1-5090-0273-3 | 14 | 0.61 |
References | Authors | |
15 | 4 |
Name | Order | Citations | PageRank |
---|---|---|---|
Tal Ben-Nun | 1 | 116 | 14.21 |
E. Levy | 2 | 22 | 2.24 |
Amnon Barak | 3 | 590 | 119.00 |
Eri Rubin | 4 | 20 | 1.18 |