Performance Analysis of SIMD Vectorization of High-Order Finite-Element Kernels

Sornet, Gauthier; Jubertie, Sylvain; Dupros, Fabrice; Martin, Florent De; Limet, Sébastien

doi:10.1109/hpcs.2018.00074

Cited by 3 publications

(2 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This is especially suitable for AVX512 vector machines. Other implementation schemes can be found in Sornet et al [72]. Compared to RegSEM, I/O (Input/Output) operation is effectively performed using Hierarchical Data Format HDF5 (https://support.hdfgroup.org/HDF5, accessed on 24 January 2022) library.…”

Section: Sem3dmentioning

confidence: 99%

SEM3D: A 3D High-Fidelity Numerical Earthquake Simulator for Broadband (0–10 Hz) Seismic Response Prediction at a Regional Scale

et al. 2022

View full text Add to dashboard Cite

In this paper, we present SEM3D: a 3D high-fidelity numerical earthquake simulator that is tailored to predict the seismic wave field of complex earthquake scenarios from the fault to the epicenter site. SEM3D solves the wave-propagation problem by means of the spectral element method (SEM). The presented demonstrative test case was a blind MW6.0 earthquake scenario at the European experimental site located in the sedimentary basin of Argostoli on the island of Kefalonia (Western Greece). A well-constrained geological model, obtained via geophysical inversion studies, and seismological model, given the large database of seismic traces recorded by the newly installed ARGONET network, of the site were considered. The domain of interest covered a region of 44 km × 44 km × 63 km, with the smallest grid size of 130 m × 130 m × 35 m. This allowed us to simulate the ground shaking in its entirety, from the seismic source to the epicenter site within a 0–10 Hz frequency band. Owing to the pseudo-spectral nature of the numerical method and given the high polynomial order (i.e., degree nine), the model featured 1.35·1010 DOFs (degrees of freedom). The variability of the synthetic wave field generated within the basin is assessed herein, exploring different random realizations of the mean velocity structure and heterogeneous rupture path.

show abstract

Section: Sem3dmentioning

confidence: 99%

SEM3D: A 3D High-Fidelity Numerical Earthquake Simulator for Broadband (0–10 Hz) Seismic Response Prediction at a Regional Scale

et al. 2022

View full text Add to dashboard Cite

show abstract

“…Its scalability on Shaheen II supercomputer is presented in Sochala et al (2020). Its most computationally intensive kernel -the computation of the internal forces -has been optimized for the single instruction multiple data (SIMD) vectorization (Jubertie et al 2018;Sornet et al 2018).…”

Section: O R I Gmentioning

confidence: 99%

Influential parameters on 3-D synthetic ground motions in a sedimentary basin derived from global sensitivity analysis

Martin

Chaljub

Thierry³

et al. 2021

Geophysical Journal International

View full text Add to dashboard Cite

Summary Which physical parameters are the most influential when predicting earthquake ground motions in a 3-D sedimentary basin? We answer quantitatively by doing a global sensitivity analysis of two quantities of interest: the peak ground motions (PGMs) and a time-frequency representation (the S transform) of ground motions resulting from the synthetic anelastic responses of the EUROSEISTEST. This domain of interest is modeled by two layers with uncertain depth-dependent mechanical properties and is illuminated by a plane S-wave propagating vertically upward in an uncertain homogeneous elastic bedrock. The global sensitivity analysis is conducted on 800+ physics-based simulations of the EUROSEISTEST requiring 8+ million core-hours (i.e., ≈ 900 years of mono-core computation). The analysis of the PGMs at the free surface displays the spatial influence of the uncertain input parameters over the entire basin scale, while the analysis of the time-frequency representation shows their influence at a specific location inside the basin. The global sensitivity analysis done on the PGMs points out that their most influential parameter in the middle of the basin is the quality factor QS (it controls up to 80 per cent of the PGMs in certain locations where the sediments thickness is larger than 200 m). On the other hand, the geological layering configuration (here represented by the depth of a gelological interface controlling the geological layering) strongly influences the PGMs close to the basin edges, up to 90 per cent. We also found that the shear-wave velocity at the free surface of the basin and the one of the bedrock underlying the basin are to be considered on an equal footing, both influencing the PGMs in the middle of the basin and close to its edges. We highlight that the bedrock to basin amplification of the PGMs shows a clear increase with respect to the thickness of the sediments, but this amplification saturates from 200 m of sediments around the value of three and is frequency dependent. This PGMs amplification starts from about one tenth of the mean S-wavelength propagating in the basin. The global sensitivity analysis done on the S transform of the ground motions shows that (i) the own effect of the parameters fully controls the first S-wave train and mostly controls the direct arrival of the basin-induced surfaces waves, (ii) the quality factor QS controls 40 to 60 per cent of the decay of amplitude of coda waves, the remaining part being mainly controlled by interaction effects due to the coupling effect of several parameters, and (iii) the interaction effects between the parameters increases with time, suggesting under the hypotheses of our study that the own effects control the ballistic wave propagation while the interaction effects control the diffusive wave propagation.

show abstract