Performance Analysis for Stencil-Based 3D MPDATA Algorithm on GPU Architecture

Rojek, Krzysztof; Szustak, Łukasz; Wyrzykowski, Roman

doi:10.1007/978-3-642-55224-3_15

Cited by 9 publications

(14 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…MPDATA requires loading 5 input matrices and returns only one. We assume that the size of the computational grid is n × m × l. In previous work [11], it has been shown that MPDATA is strongly memory-bounded. The main bottleneck in adaptation of MPDATA to GPU architecture resides in memory traffic between global and local GPU memory.…”

Section: Adaptation Of Advection Transport Algorithm Mpdata To Gmentioning

confidence: 99%

Porting Multiscale Fluid Model EULAG to Modern Heterogeneous Architectures

Rosa¹,

Ciznicki²,

Rojek³

et al. 2014

IJAPM

Self Cite

View full text Add to dashboard Cite

Section: Adaptation Of Advection Transport Algorithm Mpdata To Gmentioning

confidence: 99%

Porting Multiscale Fluid Model EULAG to Modern Heterogeneous Architectures

Rosa¹,

Ciznicki²,

Rojek³

et al. 2014

IJAPM

Self Cite

View full text Add to dashboard Cite

“…It required to develop a different approach [14] than for the 2D version. In paper [13], we presented an analysis of resources usage in GPU, and its influence on the resulting performance. We detected the bottlenecks and developed a method for the efficient distribution of computation across CUDA kernels.…”

Section: Introductionmentioning

confidence: 99%

“…Currently, the model is being implemented as the new dynamical core of the COSMO weather prediction framework. The multidimensional positive definite advection transport algorithm (MPDATA) is among the most time-consuming calculations of the EULAG solver [5][6][7][8][9][10][11][12][13][14][15][16].…”

Section: Introductionmentioning

confidence: 99%

Performance modeling of 3D MPDATA simulations on GPU cluster

Rojek

Wyrzykowski

2016

J Supercomput

Self Cite

View full text Add to dashboard Cite

The goal of this study is to parallelize the multidimensional positive definite advection transport algorithm (MPDATA) across a computational cluster equipped with GPUs. Our approach permits us to provide an extensive overlapping GPU computations and data transfers, both between computational nodes, as well as between the GPU accelerator and CPU host within a node. For this aim, we decompose a computational domain into two unequal parts which correspond to either data dependent or data independent parts. Then, data transfers can be performed simultaneously with computations corresponding to the second part. Our approach allows for achieving 16.372 Tflop/s using 136 GPUs. To estimate the scalability of the proposed approach, a performance model dedicated to MPDATA simulations is developed. We focus on the analysis of computation and communication execution times, as well as the influence of overlapping data transfers and GPU computations, with regard to the number of nodes.

show abstract

“…Such transformation is widely described in the literature. For the full description of the most important aspects of the second-order-accurate formulation of MPDATA, the reader is referred to [5,11].…”

Section: Outline Of Mpdatamentioning

confidence: 99%

“…We currently focus on the use of MPDATA in NWP, where the size of grids is limited by ≤ 2048, ≤ 1024, and = [64,128]. In our previous work [11], it has been shown that MPDATA is strongly memory-bounded.…”

Section: Introductionmentioning

confidence: 99%

Adaptation of MPDATA Heterogeneous Stencil Computation to Intel Xeon Phi Coprocessor

Szustak

Rojek

Olas

et al. 2015

Scientific Programming

Self Cite

View full text Add to dashboard Cite

The multidimensional positive definite advection transport algorithm (MPDATA) belongs to the group of nonoscillatory forwardin-time algorithms and performs a sequence of stencil computations. MPDATA is one of the major parts of the dynamic core of the EULAG geophysical model. In this work, we outline an approach to adaptation of the 3D MPDATA algorithm to the Intel MIC architecture. In order to utilize available computing resources, we propose the (3 + 1)D decomposition of MPDATA heterogeneous stencil computations. This approach is based on combination of the loop tiling and fusion techniques. It allows us to ease memory/communication bounds and better exploit the theoretical floating point efficiency of target computing platforms. An important method of improving the efficiency of the (3 + 1)D decomposition is partitioning of available cores/threads into work teams. It permits for reducing inter-cache communication overheads. This method also increases opportunities for the efficient distribution of MPDATA computation onto available resources of the Intel MIC architecture, as well as Intel CPUs. We discuss preliminary performance results obtained on two hybrid platforms, containing two CPUs and Intel Xeon Phi. The top-of-the-line Intel Xeon Phi 7120P gives the best performance results, and executes MPDATA almost 2 times faster than two Intel Xeon E5-2697v2 CPUs.

show abstract

Performance Analysis for Stencil-Based 3D MPDATA Algorithm on GPU Architecture

Cited by 9 publications

References 6 publications

Porting Multiscale Fluid Model EULAG to Modern Heterogeneous Architectures

Porting Multiscale Fluid Model EULAG to Modern Heterogeneous Architectures

Performance modeling of 3D MPDATA simulations on GPU cluster

Adaptation of MPDATA Heterogeneous Stencil Computation to Intel Xeon Phi Coprocessor

Contact Info

Product

Resources

About