GraphIt: a high-performance graph DSL
Open accessThe DOI matches its Crossref record; 48 of the 48 references it cites were found in a registry or an index.
Abstract
The performance bottlenecks of graph applications depend not only on the algorithm and the underlying hardware, but also on the size and structure of the input graph. As a result, programmers must try different combinations of a large set of techniques, which make tradeoffs among locality, work-efficiency, and parallelism, to develop the best implementation for a specific algorithm and type of graph. Existing graph frameworks and domain specific languages (DSLs) lack flexibility, supporting only a limited set of optimizations. This paper introduces GraphIt, a new DSL for graph computations that generates fast implementations for algorithms with different performance characteristics running on graphs with different sizes and structures. GraphIt separates what is computed (algorithm) from how it is computed (schedule). Programmers specify the algorithm using an algorithm language, and performance optimizations are specified using a separate scheduling language. The algorithm language simplifies expressing the algorithms, while exposing opportunities for optimizations. We formulate graph optimizations, including edge traversal direction, data layout, parallelization, cache, NUMA, and kernel fusion optimizations, as tradeoffs among locality, parallelism, and work-efficiency. The scheduling language enables programmers to easily search through this complicated tradeoff space by composing together a large set of edge traversal, vertex data layout, and program structure optimizations. The separation of algorithm and schedule also enables us to build an autotuner on top of GraphIt to automatically find high-performance schedules. The compiler uses a new scheduling representation, the graph iteration space, to model, compose, and ensure the validity of the large number of optimizations. We evaluate GraphIt’s performance with seven algorithms on graphs with different structures and sizes. GraphIt outperforms the next fastest of six state-of-the-art shared-memory frameworks (Ligra, Green-Marl, GraphMat, Galois, Gemini, and Grazelle) on 24 out of 32 experiments by up to 4.8×, and is never more than 43% slower than the fastest framework on the other experiments. GraphIt also reduces the lines of code by up to an order of magnitude compared to the next fastest framework.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
(1999) · Cited by 12,854
What is Twitter, a social network or a news media?
Kwak, H., Lee, C., Park, H., & Moon, S. (2010) · Proceedings of the 19th international conference on World wide web · doi:10.1145/1772690.1772751 · Cited by 6,652 · cited 1×
“Edges LiveJournal (LJ) [Davis and Hu 2011] 5 M 69 M Twitter (TW) [Kwak et al. 2010] 41 M 1469 M WebGraph (WB) [Meusel et al.”
BackgroundThe university of Florida sparse matrix collection
Davis, T. A., & Hu, Y. (2011) · ACM Transactions on Mathematical Software · doi:10.1145/2049662.2049663 · Cited by 3,630
Bennett, J. R., & Lanning, S. (2007) · Cited by 1,466
PowerGraph: distributed graph-parallel computation on natural graphs
Gonzalez, J. E., Low, Y., Gu, H., Bickson, D., & Guestrin, C. (2012) · doi:10.5555/2387880.2387883 · Cited by 1,451
GraphChi: Large-Scale Graph Computation on Just a PC
Kyrola, A., Blelloch, G. E., & Guestrin, C. (2018) · Figshare · doi:10.1184/r1/6605999 · Cited by 868
Ligra: a lightweight graph processing framework for shared memory
Shun, J., & Blelloch, G. E. (2013) · Proceedings of the 18th ACM SIGPLAN symposium on Principles and practice of parallel programming · doi:10.1145/2442516.2442530 · Cited by 798 · cited 7×
“Hybrid traversal modes (DensePull-SparsePush and DensePush-SparsePush) use different directions (SparsePush, DensePull, and DensePush) in different iterations based on the size of the active set to improve work-efficiency [Beamer et al. 2012; Shun and Blelloch 2013].”
BackgroundMethodInfluentialSelf-citationAdvanced compiler optimizations for supercomputers
Padua, D. A., & Wolfe, M. J. (1986) · Communications of the ACM · doi:10.1145/7902.7904 · Cited by 690 · cited 2×
“Inspired by iteration space theory for dense loops [Padua and Wolfe 1986; Wolf and Lam 1991], we introduce an abstract graph iteration space model to represent, compose, and ensure the validity of edge traversal optimizations.”
MethodX-Stream: edge-centric graph processing using streaming partitions
Roy, A., Mihailovic, I., & Zwaenepoel, W. (2013) · Proceedings of the Twenty-Fourth ACM Symposium on Operating Systems Principles · doi:10.1145/2517349.2522740 · Cited by 673 · cited 2×
“When the graphs do fit in memory, out-of-core systems, such as X-Stream [Roy et al. 2013], have shown to be much slower than shared-memory frameworks [Grossman et al. 2018; Zhang et al. 2017].”
BackgroundA loop transformation theory and an algorithm to maximize parallelism
Wolf, M. E., & Lam, M. S. (1991) · IEEE Transactions on Parallel and Distributed Systems · doi:10.1109/71.97902 · Cited by 643 · cited 2×
“Inspired by iteration space theory for dense loops [Padua and Wolfe 1986; Wolf and Lam 1991], we introduce an abstract graph iteration space model to represent, compose, and ensure the validity of edge traversal optimizations.”
Method
Cited by
Taichi: a language for high-performance computation on spatially sparse data structures
Hu, Y., Li, T.-M., Anderson, L., Ragan-Kelley, J., & Durand, F. (2019) · ACM Transactions on Graphics · doi:10.1145/3355089.3356506 · Cited by 257
Performance Optimization using Multimodal Modeling and Heterogeneous GNN
Dutta, A., Alcaraz, J., TehraniJamsaz, A., Cesar, E., Sikora, A., & Jannesari, A. (2023) · Proceedings of the 32nd International Symposium on High-Performance Parallel and Distributed Computing · doi:10.1145/3588195.3592984 · Cited by 16
“OpenTuner [8] is a search-based autotuner which employs various search techniques such as AUC Bandit, Nelder-Mead, Torczon hill-climbers, etc. ytopt and OpenTuner have been previously used for a variety of tuning tasks [32, 35, 43, 81, 84] and were hence chosen as baselines for this paper.”
Method
Metadata
- Authors
- Zhang, Y., Yang, M., Baghdadi, R., Kamil, S., Shun, J., & Amarasinghe, S.
- Venue
- Proceedings of the ACM on Programming Languages
- Published
- 2018-10-24
- Volume
- 2
- Issue
- OOPSLA
- Pages
- 1-30
- ISSN
- 2475-1421 (ISSN-L)
- OpenAlex
- W2898231337
- Cited by
- 169
- References
- 48
- FWCI
- 8.85
- Type
- journal article
- Language
- en
- Open access
- diamond
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.