Skip to content
Andrea Lani edited this page Jul 16, 2017 · 177 revisions

COOLFluiD logo

Welcome to the new COOLFluiD wiki!

Flag Counter


COOLFluiD (Computational Object-Oriented Libraries for Fluid Dynamics) is a component- based framework for scientific high-performance computing, CFD and multi-physics applications, originally developed at the Von Karman Institute for Fluid Dynamics.

Watch 2014 presentation at AMS seminar @NASA Advanced Supercomputing Division


COOLFluiD logo

COOLFluiD provides a powerful infrastructure for enabling concurrent multi-physics simulations, possibly exploiting multiple MPI communicators, heterogeneous CPU/GPU computing, massively parallel I/O capabilities, loosely coupled multi-domain simulations.


NEWS:

[16/07/2017] Within the PRACE preparatory access project OPTIMAS, the time-accurate, implicit, multi-fluid/Maxwell plasma Finite Volume solver has been successfully run on up to 60,000 cores on the XC40 Hazel Hen system at HLRS and up to 16,000 cores on the Blue Gene/Q JUQUEEN at JSC. Those runs also include parallel I/O. In order to achieve this, many code improvements have been implemented, including:

  • a more straightforward and efficient parallel synchronization algorithm for non-updatable DOFs;
  • a significant memory optimization for the reading algorithm;
  • a fix in the parallel reading algorithm affecting the collective reading of boundary connectivity data;
  • an algorithm to select a user-defined number of writing processes (aggregators) in each node;
  • a fix for the parallel mesh extruder, now able to create arbitrarily large 3D grids (tested up to 3 billion cells so far);
  • a mechanism to bypass the self-registration of polymorphic objects for compilers (like IBM's) that seem not to support it.

[29/06/2017] A new unstructured solver based on the Flux Reconstruction method has been developed in collaboration with KU Leuven. The code is steady, fully implicit, parallel, works for 2D and 3D (only quads and hexahedra are supported for the moment), Euler and Navier-Stokes, P1-P2-P3-P4-P5 (solution), P1-P2 (geometry) and can export the solution in both TECPLOT and Paraview. This represents the basic kernel for massive future developments, particularly targeted towards steady and unsteady high-order high-speed flows and plasma simulations.

[23/02/2017] COOLFluiD has been awarded 1/2 million CPU-hours through a PRACE HPC Preparatory Access (Type C) call. The goal will be to port the code on both CRAY (Hazel Hen, Gauss/HLRS, 185,088 CPU-cores #14 in Top500) and IBM Blue Gene/Q (JUQUEEN, Gauss/JSC, 458,752 CPU-cores #19 in Top500) systems, to optimize its parallel partitioning and I/O in order to scale on >100k cores.

[04/01/2017] Happy New Year picture: the first unsteady simulation of the complex thermochemical nonequilibrium plasma flow (with shock speed up to 10 Km/s) inside the NASA Electric Arc Shock Tube (EAST) facility with COOLFluiD aerothermodynamic solver and the PLATO thermochemical library (by Alessandro Munafo' at UIUC).

Full video available here (courtesy of Khalil Bensassi and many thanks to the NAS team at NASA Ames for the visualization).

EAST

[12/12/2016] COOLFluiD will benefit from Performance Optimisation & Productivity (POP), A Centre of Excellence in Computing Applications support for optimizing its massively parallel radiation code based on Finite Volume/Discrete Ordinate Method.

[10/12/2016] Three speakers from the COOLFluiD team (Alejandro Alvarez Laguna, Yana Maneva and Nataly Ozak Munoz) will present novel solar plasma simulation results at the AGU next week:

[04/12/2016] COOLFluiD latest news are now posted on Twitter.

[06/11/2016] COOLFluiD works with PARALUTION 1.1.0 using the CUDA back-end (basic single-core version). The implicit Multi-fluid / Maxwell solver has been partially (i.e. w/o diffusive terms) ported to GPU: either GPU-enabled PETSc or PARALUTION linear system solvers can be used with it. More testing and optimization are ongoing.

[01/10/2016] COOLFluiD works now also with PETSc 3.7.3, including CUDA bindings.

[07/06/2016] COOLFluiD contributes to NASA Ames Comparative Heliophysics Summer Program with the project "Multi-fluid global simulations of the Earth’s magnetosphere with multiple ions".

[25/05/2016] COOLFluiD scalability tested on NASA Pleiades up to 40k CPU-cores.

[25/02/2016] COOLFluiD works now also with PETSc 3.6.3, including CUDA bindings.

[01/11/2015] COOLFluiD Solar is on Slack.

[22/10/2014] COOLFluiD MHD magnetospheric model has been recently integrated into the first European Virtual Space Weather Modeling Center (VSWMC) together with many other models from all around Europe within a ESA GSTP program.


COOLFluiD is not:

  • "just" a CFD solver;
  • a fancy infrastructure promising multi-physics but only tested on toy problems;
  • a commercial code that will always give you an answer (right or not);
  • a monolithic piece of code with limited possibilities of evolution;
  • a free lunch for anybody (users or developers);
  • perfect, well documented and fully accomplished ... but we are working on it :)

COOLFluiD is a open end collaborative platform providing a powerful set of tools for:

  • solving toy or complex applications with existing numerical solvers;
  • building totally new or customized models/solvers with arbitrary data-structures;
  • easily defining reusable components (e.g. algorithms, models, BCs, wrappers);
  • building virtual prototypes and easily test new algorithms or models;
  • interfacing or coupling other libraries/solvers to tackle more complex problems;
  • large scale simulations.

COOLFluiD is developed open-source. Suggestions and contributions are welcome.


Gallery of applications

Available solvers and capabilities

HOWTO

Documentation

Development team

Citing COOLFluiD

Contacts

Home

Gallery

HOWTO

2019 NASA Ames presentation

2014 NASA Ames presentation


parallel computations of complex problems

Parallel mesh decomposition


Scalability in large scale simulation

High-performance computing (strong scaling on NASA Pleiades for 1/2 billion-cells 3D grid)


Modeling of high-speed reacting flows and plasma

Chemically reacting flows and plasma


Numerical Schlieren of turbulent flow on wing computed with RDS-LES

Complex all-speed flow simulations

Clone this wiki locally