Towards Domain-Specific Computing for Stencil Codes in HPC

Membarth, Richard; Hannig, Frank; Teich, Jürgen; Köstler, Harald

doi:10.1109/sc.companion.2012.136

Cited by 31 publications

(28 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Membarth at al. [17] discuss with HIPAcc a source to source compiler for image processing kernels that can use abstract hardware models to generate GPU code that addresses the different performance characteristics of various GPU models. Similarly to the previous systems, the last two approaches do not consider time tiling.…”

Section: Related Workmentioning

confidence: 99%

The Relation Between Diamond Tiling and Hexagonal Tiling

Grosser

Verdoolaege

Cohen

et al. 2014

Parallel Process. Lett.

View full text Add to dashboard Cite

Iterative stencil computations are important in scientific computing and more also in the embedded and mobile domain. Recent publications have shown that tiling schemes that ensure concurrent start provide efficient ways to execute these kernels. Diamond tiling and hybrid-hexagonal tiling are two tiling schemes that enable concurrent start. Both have different advantages: diamond tiling has been integrated in a general purpose optimization framework and uses a cost function to choose among tiling hyperplanes, whereas the greater flexibility with tile sizes for hybrid-hexagonal tiling has been exploited for effective generation of GPU code.In this paper we undertake a comparative study of these two tiling approaches and propose a hybrid approach that combines them. We analyze the effects of tile size and wavefront choices on tile-level parallelism, and formulate constraints for optimal diamond tile shapes. We then extend, for the case of two dimensions, the diamond tiling formulation into a hexagonal tiling one, which offers both the flexibility of hexagonal tiling and the generality of the original diamond tiling implementation. We also show how to compute tile sizes that maximize the compute-to-communication ratio, and apply this result to compare the best achievable ratio and the associated synchronization overhead for diamond and hexagonal tiling.

show abstract

Section: Related Workmentioning

confidence: 99%

The Relation Between Diamond Tiling and Hexagonal Tiling

Grosser

Verdoolaege

Cohen

et al. 2014

Parallel Process. Lett.

View full text Add to dashboard Cite

show abstract

“…Currently, we optimize the generated code [4,5], add more multigrid components, improve the DSLs and the transformations framework in Scala [6], add distributed memory parallelization for larger applications [7], and apply domain-specific optimizations [8].…”

Section: Future Workmentioning

confidence: 99%

Platform‐independent Description of Imaging Algorithms

Köstler¹

2014

Proc Appl Math and Mech

View full text Add to dashboard Cite

In the last years real-time imaging and more complex models for the various imaging applications became feasible mostly because of the progress made in parallel computer architectures like found in GPUs and modern multi-core CPUs. However, the implementation effort increases, if one wants to achieve good performance. A solution to this problem is to formulate the imaging algorithms in an abstract way in a domain-specific language (DSL) and then automatically generate efficient C++ or CUDA code. A multi-layered approach is sketched that allows users to describe applications in a natural way from the mathematical model down to the program specification.

show abstract

“…Examples include HIPA cc [28,29] and Halide [33], both are DSLss embedded into C ++ , allow the generation of code for different targets, among them CUDA and OpenCL. SPIRAL [32,31] generates platformtuned implementations of digital signal processing (DSP) algorithms.…”

Section: Introductionmentioning

confidence: 99%

A Scala prototype to generate multigrid solver implementations for different problems and target multi-core platforms

Köstler

Schmitt

Kuckuk

et al. 2017

IJCSE

Self Cite

View full text Add to dashboard Cite

the date of receipt and acceptance should be inserted later Abstract Many problems in computational science and engineering involve partial differential equations and thus require the numerical solution of large, sparse (non)linear systems of equations. Multigrid is known to be one of the most efficient methods for this purpose. However, the concrete multigrid algorithm and its implementation highly depend on the underlying problem and hardware. Therefore, changes in the code or many different variants are necessary to cover all relevant cases. In this article we provide a prototype implementation in Scala for a framework that allows abstract descriptions of PDEs, their discretization, and their numerical solution via multigrid algorithms. From these, one is able to generate data structures and implementations of multigrid components required to solve elliptic PDEs on structured grids. Two different test problems showcase our proposed automatic generation of multigrid solvers for both CPU and GPU target platforms.

show abstract

Towards Domain-Specific Computing for Stencil Codes in HPC

Cited by 31 publications

References 13 publications

The Relation Between Diamond Tiling and Hexagonal Tiling

The Relation Between Diamond Tiling and Hexagonal Tiling

Platform‐independent Description of Imaging Algorithms

A Scala prototype to generate multigrid solver implementations for different problems and target multi-core platforms

Contact Info

Product

Resources

About