Parallel MIMD programming for global models of atmospheric flow

Jakob, Rüdiger; Hack, James J.

doi:10.1145/76263.76275

Cited by 2 publications

(1 citation statement)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Published performance results for the spectral transform focus on various climate and weather models on different architectures [1], [4], [5], [6], [7], [8], [9], [11], [12], [13], [16], [17], [19], [20], [21], [23], [26], [27], [28], [31], [32]. Such literature data are generally unsuitable for rigorous model performance comparisons because of the dissimilar factors from which the data were generated.…”

Section: Roy W Melton and Linda M Willsmentioning

confidence: 99%

An Analysis of the Spectral Transform Operations in Climate and Weather Models

Melton¹,

Wills²

2008

SIAM J. Sci. Comput.

View full text Add to dashboard Cite

The spectral transform method used in climate and weather models is known to be computationally intensive. Typically accounting for more than 90% of the execution time of a serial model, it is poised to benefit from computational parallelism. Since dimensionally global transforms impact parallel performance, it is important to establish the realizable parallel efficiency of the spectral transform. To this end, this paper quantitatively characterizes the parallel characteristics of the spectral transform within an atmospheric modeling context. It comprehensively characterizes and catalogs a baseline of operations required for the spectral transform. While previous investigations of the spectral transform method have offered highly idealized analyses that are abstract and simplified in terms of orders of computational magnitude, this research provides a detailed model of the computational complexity of the spectral transform, validated by empirical results. From this validated quantitative analysis, an operational closed-form expression characterizes spectral transform performance in terms of general processor parameters and atmospheric data dimensions. These generalized statements of the computational requirements for the spectral transform can serve as a basis for exploiting parallelism. Introduction.Given initial conditions that represent meteorological field values of the atmosphere at some point in time, climate and weather models predict future climate and weather based on numerical solutions of a set of time-dependent partial differential equations [24]. These models typically employ one of two solution methods: grid-point or spectral [25]. While most local and regional models are grid-point, most global models are spectral [3].The computational demands of current models tax even the most powerful computers available. As models evolve to reflect a more comprehensive understanding of atmospheric processes, these models require even more computations. The high costs of both model development and computing resources drive concerns about the feasibility of a desired solution for a given model on a particular hardware platform. The efficient, scalable mapping of climate and weather models to high-performance, massively parallel multiprocessor systems depends on characterizing the degree of parallelism inherent in the models. Even though data parallel computation is intrinsic to the modeled atmospheric processes, parallel characteristics (e.g., limitations, interactions, etc.) of the facilitating spectral transform have not been understood. This paper takes a first step toward addressing these concerns for spectral models by quantifying the number and types of operations required for the spectral transform.

show abstract