GravTet: A high-performance computing software for gravity forward modeling on unstructured tetrahedral meshes.

Mohamadhasan Sarvandani and Modeste Irakarama and Paul Cupillard and Jeremie Giraud and Guillaume Caumon. ( 2026 )
in: 2026 {RING} meeting, pages 335--351, ASGA

Abstract

Unstructured tetrahedral meshes represent complex topography and subsurface structure well, motivating their use in three-dimensional gravity forward modeling. GravTet is an open-source software for the vertical gravity field component gz using second-order quadrature on unstructured tetrahedral meshes. A high-performance computing implementation in C++ combines cache blocking, SIMD vectorization, and multithreading. A Python/Flask REST server and a web interface support model setup, execution, and visualization. On cube benchmarks, optimized modes reach up to ∼ 90× speedup versus the reference while matching the analytical closed-form solution at floating-point precision (∼ 10−10%). A valley–mountain half-space example illustrates topography-bearing geometries. Overall, GravTet makes tetrahedral gravity forward modeling both accurate and computationally practical for interpretation and inversion workflows.

Download / Links

BibTeX Reference

@inproceedings{Sarvandani2026RM,
 abstract = {Unstructured tetrahedral meshes represent complex topography and subsurface structure well, motivating their use in three-dimensional gravity forward modeling. GravTet is an open-source software for the vertical gravity field component gz using second-order quadrature on unstructured tetrahedral meshes. A high-performance computing implementation in C++ combines cache blocking, SIMD vectorization, and multithreading. A Python/Flask REST server and a web interface support model setup, execution, and visualization. On cube benchmarks, optimized modes reach up to ∼ 90× speedup versus the reference while matching the analytical closed-form solution at floating-point precision (∼ 10−10%). A valley–mountain half-space example illustrates topography-bearing geometries. Overall, GravTet makes tetrahedral gravity forward modeling both accurate and computationally practical for interpretation and inversion workflows.},
 author = {Sarvandani, Mohamadhasan and Irakarama, Modeste and Cupillard, Paul and Giraud, Jérémie and Caumon, Guillaume},
 booktitle = {2026 {RING} meeting},
 language = {en},
 pages = {335--351},
 publisher = {ASGA},
 title = {GravTet: A high-performance computing software for gravity forward modeling on unstructured tetrahedral meshes.},
 year = {2026}
}