Discretization

Solve a PDE (heat,solid mech) with different discretisation methods from low to high order

Quick Facts

Type

Extended Mini App

Status

Benchmark Ready

Work Packages

WP1, WP3

Frameworks

Feel++, PETSc

This application is benchmark-ready and can be used for performance evaluation.

1. Overview

Solve a PDE (heat,solid mech) with different discretisation methods from low to high order

1.1. Repository

github.com/numpex/apps-feelpp

2. Technical Stack

2.1. Frameworks Used

  • Feel++

  • PETSc

2.2. Parallel Frameworks

  • MPI

  • Kokkos

3. Methods & Algorithms

3.1. WP1: Discretization

  • unstructured mesh

  • finite element

  • dG/hdG

  • cG

  • spectral element

3.2. WP3: Solvers

  • domain decomposition methods

  • krylov solver

  • preconditioning

4. Data Flow

4.1. Inputs

  • Gmsh

  • JSON config

4.2. Outputs

  • JSON small-data

  • JSON reports

  • in-house

  • Markdown or Asciidoc

5. Benchmarking

5.1. Metrics

  • Benchmark Verification

  • Strong Scalability

  • Weak Scalability

  • Io Scaling

5.2. Benchmark Scope

  • Full System

  • Method Verification

  • Solver Scaling

6. Timeline

Milestone Date

Specification Due

2025-06-15

Prototype Due

2025-07-01

7. Team

Partners: Unistra

Responsible: V. Chabannes; C. Prud’homme

WP7 Engineer: Javier Cladellas (UNISTRA)