Data layout optimization for multi-valued containers in OpenCL

Strzodka, Robert

doi:10.1016/j.jpdc.2011.10.012

Cited by 11 publications

(6 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Several recent efforts [8,10,12,21,24] have taken advantage of the substructure of objects in arrays to change the layout of the data in memory to better match the different types of processors in a heterogeneous system, such as converting from an array-of-structs to a struct-of-arrays or transposing the contents of arrays. However, all require some changes to the application code and commit to a layout decision at compile time.…”

Section: Related Workmentioning

confidence: 99%

Structure Slicing: Extending Logical Regions with Fields

Bauer

Treichler

Slaughter

et al. 2014

SC14: International Conference for High Performance Computing, Networking, Storage and Analysis

View full text Add to dashboard Cite

Abstract-Applications on modern supercomputers are increasingly limited by the cost of data movement, but mainstream programming systems have few abstractions for describing the structure of a program's data. Consequently, the burden of managing data movement, placement, and layout currently falls primarily upon the programmer.To address this problem we previously proposed a data model based on logical regions and described Legion, a programming system incorporating logical regions. In this paper, we present structure slicing, which incorporates fields into the logical region data model. We show that structure slicing enables Legion to automatically infer task parallelism from field non-interference, decouple the specification of data usage from layout, and reduce the overall amount of data moved. We demonstrate that structure slicing enables both strong and weak scaling of three Legion applications including S3D, a production combustion simulation that uses logical regions with thousands of fields, with speedups of up to 3.68X over a vectorized CPU-only Fortran implementation and 1.88X over an independently hand-tuned OpenACC code.

show abstract

Section: Related Workmentioning

confidence: 99%

Structure Slicing: Extending Logical Regions with Fields

Bauer

Treichler

Slaughter

et al. 2014

SC14: International Conference for High Performance Computing, Networking, Storage and Analysis

View full text Add to dashboard Cite

show abstract

“…The selection of a proper data layout for a specific application depends on its underlying algorithm. In general, the usual language syntax and standard container types lead naturally to the AoS layout while SIMD units much prefer the SoA format (Strzodka 2012a ).…”

Section: Introductionmentioning

confidence: 99%

“…In order to improve the efficiency of accessing memories on the GPU, many studies have been performed to transform different types of layouts to others, e.g., from AoS to SoA, or vice versa (Mistry et al. 2011 ; Strzodka 2012a , b ; Sung et al. 2012 ).…”

Section: Introductionmentioning

confidence: 99%

Impact of data layouts on the efficiency of GPU-accelerated IDW interpolation

Mei

Tian

2016

SpringerPlus

View full text Add to dashboard Cite

This paper focuses on evaluating the impact of different data layouts on the computational efficiency of GPU-accelerated Inverse Distance Weighting (IDW) interpolation algorithm. First we redesign and improve our previous GPU implementation that was performed by exploiting the feature of CUDA dynamic parallelism (CDP). Then we implement three versions of GPU implementations, i.e., the naive version, the tiled version, and the improved CDP version, based upon five data layouts, including the Structure of Arrays (SoA), the Array of Structures (AoS), the Array of aligned Structures (AoaS), the Structure of Arrays of aligned Structures (SoAoS), and the Hybrid layout. We also carry out several groups of experimental tests to evaluate the impact. Experimental results show that: the layouts AoS and AoaS achieve better performance than the layout SoA for both the naive version and tiled version, while the layout SoA is the best choice for the improved CDP version. We also observe that: for the two combined data layouts (the SoAoS and the Hybrid), there are no notable performance gains when compared to other three basic layouts. We recommend that: in practical applications, the layout AoaS is the best choice since the tiled version is the fastest one among three versions. The source code of all implementations are publicly available.Electronic supplementary materialThe online version of this article (doi:10.1186/s40064-016-1731-6) contains supplementary material, which is available to authorized users.

show abstract

“…Two typical examples are: organizing data with similar access patterns in structures or rearranging array of structures (AoS) as structure of arrays (SoA). Recent work extends the classical SoA layout by introducing AoSoA (Array of Structure of Array) [16], also called ASA [14]. In this paper we prefer the expression tiled-AoS, but we remark that all approaches exploit the same idea: mixing AoS and SoA in a unique data layout.…”

Section: Introductionmentioning

confidence: 99%

Automatic Data Layout Optimizations for GPUs

Kofler

Cosenza

Fahringer

2015

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Memory optimizations have became increasingly important in order to fully exploit the computational power of modern GPUs. The data arrangement has a big impact on the performance, and it is very hard for GPU programmers to identify a well-suited data layout. Classical data layout transformations include grouping together data fields that have similar access patterns, or transforming Array-of-Structures (AoS) to Structure-of-Arrays (SoA). This paper presents an optimization infrastructure to automatically determine an improved data layout for OpenCL programs written in AoS layout. Our framework consists of two separate algorithms: The first one constructs a graph-based model, which is used to split the AoS input struct into several clusters of fields, based on hardware dependent parameters. The second algorithm selects a good per-cluster data layout (e.g., SoA, AoS or an intermediate layout) using a decision tree. Results show that the combination of both algorithms is able to deliver higher performance than the individual algorithms. The layouts proposed by our framework result in speedups of up to 2.22, 1.89 and 2.83 on an AMD FirePro S9000, NVIDIA GeForce GTX 480 and NVIDIA Tesla k20m, respectively, over different AoS sample programs, and up to 1.18 over a manually optimized program.

show abstract

Data layout optimization for multi-valued containers in OpenCL

Cited by 11 publications

References 9 publications

Structure Slicing: Extending Logical Regions with Fields

Structure Slicing: Extending Logical Regions with Fields

Impact of data layouts on the efficiency of GPU-accelerated IDW interpolation

Automatic Data Layout Optimizations for GPUs

Contact Info

Product

Resources

About