Acceleration of NOISEtte Code for Scale-Resolving Supercomputer Simulations of Turbulent Flows

Gorobets, A.; Bakhvalov, Pavel Alexeevisch; Duben, A. P.; Rodionov, P. V.

doi:10.1134/s1995080220080077

Cited by 9 publications

(6 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In summary, the "business lunch" version with combined functions and mixed accuracy is about twice as fast on CPUs as the base version. The use of mixed accuracy speeds up the code about 1.6 -1.7 times and saves memory nearly twice with no effect on accuracy of results, as shown in [12]. The OpenCL version of the code gives significant acceleration, one GPU performs like 7-8 modern multicore CPUs, which is in good agreement with the memory bandwidth ratio between CPU and GPU.…”

Section: Discussionmentioning

confidence: 60%

“…Furthermore, since GPU memory is very limited, we developed a simplified version of the viscous fluxes calculation method by reducing the number of coefficients and using mixed single and double precision floating point formats. Details about this new method to compute viscous fluxes much cheaper can be found in [12]. In resource-intensive applications, this "business-lunch" configuration works about 15-20% faster than the baseline full version, called "playground".…”

Section: Simplification and Improvement Of Performancementioning

confidence: 99%

“…The OpenCL version includes the set of numerical schemes EBR-LO, EBR-3, EBR-5, EBR-WENO-LO, EBR-WENO-3, EBR-WENO-5, Riemann solvers, including Roe [13] with the Turkel preconditioner [14] for low-Mach flows, our simplified method for viscous fluxes [12], the set of RANS, LES models and hybrid approaches of the DES family (including IDDES [10]), explicit Runge-Kutta and implicit BDF1, BDF2 schemes, the preconditioned BiCGSTAB solver [15].…”

Section: Implementation For Gpu Computingmentioning

confidence: 99%

See 2 more Smart Citations

Adapting Complex and Clumsy CFD Code to Rapidly Changing Supercomputing Realities

Gorobets¹

2021

14th WCCM-ECCOMAS Congress

View full text Add to dashboard Cite

This work is devoted to acceleration and upgrade of the CFD code NOISEtte for scale-resolving simulations of compressible turbulent flows using edge-based high-accuracy methods on unstructured hybrid meshes. Attempts to extend the baseline multilevel MPI+OpenMP parallelization towards GPU-based hybrid systems have faced the problem: the code is too complex. It is an in-house research code with plenty of numerical methods, schemes, models, most of which are experimental and are not used in practical simulations. This chaotic zoo leads to excessive conditional branches, switches, redundant functional calls that slow down computations. Although the parallel algorithm is fully adapted to the stream processing paradigm, such an immense amount of code is too difficult to port efficiently to OpenCL or CUDA and maintain it in consistency with the CPU version. An approach to survive in the process of adaptation to hybrid systems has been elaborated. It consists of various components, such as creation of a simplified configurations, combining different stages of the algorithm in order to reduce memory traffic, collapsing multiple functions in one function without branches and switches, mixing single and double precision, etc. As a result, the upgraded code is about twice as fast on CPUs and can use GPUs from different manufacturers-AMD, NVIDIA, Intel through the OpenCL standard.

show abstract

Section: Discussionmentioning

confidence: 60%

Section: Simplification and Improvement Of Performancementioning

confidence: 99%

Section: Implementation For Gpu Computingmentioning

confidence: 99%

See 1 more Smart Citation

Adapting Complex and Clumsy CFD Code to Rapidly Changing Supercomputing Realities

Gorobets¹

2021

14th WCCM-ECCOMAS Congress

View full text Add to dashboard Cite

show abstract

“…The heterogeneous parallel algorithm is implemented in the NOISEtte code [8]. Further details on parallel algorithm, adaptation of the numerical algorithm and software implementation to GPU computing can be found in [7][8][9]. Examples of parallel speedups are shown in Fig.…”

Section: Parallel Computingmentioning

confidence: 99%

Technology for Supercomputer Simulation of Turbulent Flows in the Good New Days of Exascale Computing

2021

JSFI

View full text Add to dashboard Cite

“…minimization of work-item tasks to increase occupancy of compute units; mixed single and double floating point precision (single precision in some heavy arrays of discrete operator coefficients and in the linear solver) to reduce memory consumption and memory traffic, of course without affecting the accuracy of the results [18]; reordering of mesh objects (block Cuthill -McKee, lexicographical sorting) to improve memory access locality; new numerical algorithms with reduced memory consumption [18,19].…”

Section: Parallel Implementationmentioning

confidence: 99%

Supercomputer Simulation of Turbulent Flow Around Isolated UAV Rotor and Associated Acoustic Fields

Bobkov

Gorobets

Kozubskaya

et al. 2021

Communications in Computer and Information Science

View full text Add to dashboard Cite

The paper presents the results of supercomputer simulations of turbulent flow around a rotor of a small-scale unmanned aerial vehicle. Rotor aerodynamics and near-field acoustics are modeled using both RANS and scale-resolving hybrid RANS-LES approaches. The underlying numerical method uses an edge-based reconstruction scheme for higher-accuracy discretization on unstructured mixed-element meshes. The far-field acoustics are evaluated using the integral Ffowcs-Williams and Hawkings method. The numerical results are compared with the available experimental data. The simulations were carried out on a hybrid supercomputer using several dozen GPUs. The parallel efficiency of the simulation code and comparison of performance on different types of computing devices are also presented.

show abstract

Acceleration of NOISEtte Code for Scale-Resolving Supercomputer Simulations of Turbulent Flows

Cited by 9 publications

References 31 publications

Adapting Complex and Clumsy CFD Code to Rapidly Changing Supercomputing Realities

Adapting Complex and Clumsy CFD Code to Rapidly Changing Supercomputing Realities

Technology for Supercomputer Simulation of Turbulent Flows in the Good New Days of Exascale Computing

Supercomputer Simulation of Turbulent Flow Around Isolated UAV Rotor and Associated Acoustic Fields

Contact Info

Product

Resources

About