WTF, GPU! computing twitter's who-to-follow on the GPU

Geil, Afton; Wang, Yangzihao; Owens, John D.

doi:10.1145/2660460.2660481

“…The state-of-the-art CUDA implementation of BFS [10] uses best-effort synchronization approaches, that avoid the use of atomic operations but provide a degree of filtering of duplicated elements. SSSP [16] and PR [4] use atomic operations to avoid duplication, yet incur in significant overheads. Taking BFS as an example, its exploration consists of two main phases or kernels: Expansion and Contraction.…”

Section: Gpgpu Graph Processingmentioning

confidence: 99%

“…G RAPH-BASED applications are ubiquitous in important domains such as data analytics [1] or machine learning [2] among many other examples. Road navigation and self-driving cars [3], recommendation systems [4] and speech recognition [5] are paradigmatic examples of graph processing workloads. Current trends towards increased data gathering [6] and knowledge-based applications result in an increased importance of graph-based applications and, at the same time, a demand for higher data processing capabilities, which motivates high-throughput graph processing on GPGPU architectures.…”

Section: Introductionmentioning

confidence: 99%

Energy-Efficient Stream Compaction Through Filtering and Coalescing Accesses in GPGPU Memory Partitions

Segura

¹

,

Arnau

²

,

González

³

2022

IEEE Trans. Comput.

View full text Add to dashboard Cite

Graph-based applications are essential in emerging domains such as data analytics or machine learning. Data gathering in a knowledge-based society requires great data processing efficiency. High-throughput GPGPU architectures are key to enable efficient graph processing. Nonetheless, irregular and sparse memory access patterns present in graph-based applications induce high memory divergence and contention, which result in poor GPGPU efficiency for graph processing. Recent work has pointed out the importance of stream compaction operations, and has proposed a Stream Compaction Unit (SCU) to offload them to a specialized hardware. On the other hand, memory contention caused by high divergence has been tackled with the Irregular accesses Reorder Unit (IRU), delivering improved memory coalescing. In this paper, we propose a new unit, the IRU-enhanced SCU (ISCU), that leverages the strengths of both approaches. The ISCU employs the efficient mechanisms of the IRU to improve SCU stream compaction efficiency and throughput limitations, achieving a synergistic effect for graph processing. We evaluate the ISCU for a wide variety of state-of-the-art graph-based algorithms and applications. Results show that the ISCU achieves a performance speedup of 2.2x and 90% energy savings derived from a high reduction of 78% memory accesses, while incurring in 8.5% area overhead.

show abstract

“…In Gunrock, we begin with a frontier that contains all vertices in the graph and end when all vertices have converged. Each iteration contains one advance operator to compute the PageRank value on the frontier of vertices, and one filter operator to remove the Bipartite graphs Geil et al [9] used Gunrock to implement Twitter's who-to-follow algorithm ("Money" [11]), which incorporated three node-ranking algorithms based on bipartite graphs (Personalized PageRank, Stochastic Approach for Link-Structure Analysis (SALSA), and Hyperlink-Induced Topic Search (HITS)). Their implementation, the first to use a programmable framework for bipartite graphs, demonstrated that Gunrock's advance operator is flexible enough to encompass all three node-ranking algorithms, including a 2-hop traversal in a bipartite graph.…”

Section: Pagerank and Other Node Ranking Algorithmsmentioning

confidence: 99%

Gunrock

Wang

¹

,

Davidson

²

,

Pan

³

et al. 2016

SIGPLAN Not.

Self Cite

62

0

View full text Add to dashboard Cite

For large-scale graph analytics on the GPU, the irregularity of data access/control flow and the complexity of programming GPUs have been two significant challenges for developing a programmable high-performance graph library. "Gunrock," our high-level bulksynchronous graph-processing system targeting the GPU, takes a new approach to abstracting GPU graph analytics: rather than designing an abstraction around computation, Gunrock instead implements a novel data-centric abstraction centered on operations on a vertex or edge frontier. Gunrock achieves a balance between performance and expressiveness by coupling high-performance GPU computing primitives and optimization strategies with a highlevel programming model that allows programmers to quickly develop new graph primitives with small code size and minimal GPU programming knowledge. We evaluate Gunrock on five graph primitives (BFS, BC, SSSP, CC, and PageRank) and show that Gunrock has on average at least an order of magnitude speedup over Boost and PowerGraph, comparable performance to the fastest GPU hardwired primitives, and better performance than any other GPU high-level graph library. * Currently an employee at Google. † Currently an employee at IBM.

show abstract

“…Bipartite graphs Geil et al [9] used Gunrock to implement Twitter's who-to-follow algorithm ("Money" [11]), which incorporated three node-ranking algorithms based on bipartite graphs (Personalized PageRank, Stochastic Approach for Link-Structure Analysis (SALSA), and Hyperlink-Induced Topic Search (HITS)). Their implementation, the first to use a programmable framework for bipartite graphs, demonstrated that Gunrock's advance operator is flexible enough to encompass all three node-ranking algorithms, including a 2-hop traversal in a bipartite graph.…”

Section: Pagerank and Other Node Ranking Algorithmsmentioning

confidence: 99%

Gunrock: a high-performance graph processing library on the GPU

Wang

¹

,

Davidson

²

,

Pan

³

et al. 2015

Self Cite

View full text Add to dashboard Cite

For large-scale graph analytics on the GPU, the irregularity of data access/control flow and the complexity of programming GPUs have been two significant challenges for developing a programmable high-performance graph library. "Gunrock," our high-level bulksynchronous graph-processing system targeting the GPU, takes a new approach to abstracting GPU graph analytics: rather than designing an abstraction around computation, Gunrock instead implements a novel data-centric abstraction centered on operations on a vertex or edge frontier. Gunrock achieves a balance between performance and expressiveness by coupling high-performance GPU computing primitives and optimization strategies with a highlevel programming model that allows programmers to quickly develop new graph primitives with small code size and minimal GPU programming knowledge. We evaluate Gunrock on five graph primitives (BFS, BC, SSSP, CC, and PageRank) and show that Gunrock has on average at least an order of magnitude speedup over Boost and PowerGraph, comparable performance to the fastest GPU hardwired primitives, and better performance than any other GPU high-level graph library.

show abstract

WTF, GPU! computing twitter's who-to-follow on the GPU

Cited by 12 publications

References 12 publications

Energy-Efficient Stream Compaction Through Filtering and Coalescing Accesses in GPGPU Memory Partitions

Energy-Efficient Stream Compaction Through Filtering and Coalescing Accesses in GPGPU Memory Partitions

Gunrock

Gunrock: a high-performance graph processing library on the GPU

Contact Info

Product

Resources

About