BLASX
- 1 June 2016
- conference paper
- conference paper
- Published by Association for Computing Machinery (ACM)
Abstract
Basic Linear Algebra Subprograms (BLAS) are a set of low level linear algebra kernels widely adopted by applications involved with the deep learning and scientific computing. The massive and economic computing power brought forth by the emerging GPU architectures drives interest in implementation of compute-intensive level 3 BLAS on multi-GPU systems. In this paper, we investigate existing multi-GPU level 3 BLAS and present that 1) issues, such as the improper load balancing, inefficient communication, insufficient GPU stream level concurrency and data caching, impede current implementations from fully harnessing heterogeneous computing resources; 2) and the inter-GPU Peer-to-Peer(P2P) communication remains unexplored. We then present BLASX: a highly optimized multi-GPU level-3 BLAS. We adopt the concepts of algorithms-by-tiles treating a matrix tile as the basic data unit and operations on tiles as the basic task. Tasks are guided with a dynamic asynchronous runtime, which is cache and locality aware. The communication cost under BLASX becomes trivial as it perfectly overlaps communication and computation across multiple streams during asynchronous task progression. It also takes the current tile cache scheme one step further by proposing an innovative 2-level hierarchical tile cache, taking advantage of inter-GPU P2P communication. As a result, linear speedup is observable with BLASX under multi-GPU configurations; and the extensive benchmarks demonstrate that BLASX consistently outperforms the related leading industrial and academic implementations such as cuBLAS-XT, SuperMatrix, MAGMA.Keywords
This publication has 15 references indexed in Scilit:
- Hierarchical DAG Scheduling for Hybrid Distributed SystemsPublished by Institute of Electrical and Electronics Engineers (IEEE) ,2015
- Enabling and scaling matrix computations on heterogeneous multi-core and multi-GPU systemsPublished by Association for Computing Machinery (ACM) ,2012
- Optimizing symmetric dense matrix-vector multiplication on GPUsPublished by Association for Computing Machinery (ACM) ,2011
- High-performance implementation of the level-3 BLASACM Transactions on Mathematical Software, 2008
- SuperMatrixPublished by Association for Computing Machinery (ACM) ,2008
- Scheduling multithreaded computations by work stealingJournal of the ACM, 1999
- On the All-Pairs-Shortest-Path Problem in Unweighted Undirected GraphsJournal of Computer and System Sciences, 1995
- A set of level 3 basic linear algebra subprogramsACM Transactions on Mathematical Software, 1990
- A class of compatible cache consistency protocols and their support by the IEEE futurebusACM SIGARCH Computer Architecture News, 1986
- On the complexity of fixed-priority scheduling of periodic, real-time tasksPerformance Evaluation, 1982