DOWNLOAD¶
Here you can FREELY download AMITEX_FFTP for RESEARCH & EDUCATION purposes, according to the licence.
Since v8.17.2, you can also download documents used for TRAINING SESSIONS.
The usage of AMITEX_FFTP can be divided into two parts :
- The Standard usage is based on a user interface described in USER. It doesn’t require any code development.
- The Extended usage is mainly devoted to the implementation of various “couplings” (different physics, different codes). It requires code development. If interested, feel free to contact us (contacts).
Just keep in mind that AMITEX_FFTP is not a “big” code, there is not so many people behind it ;-)
VERSION 8.17.x + TRAINING SESSION + VALIDATION TESTS¶
The version 8.17.2 is now quite stable with a quite broad scope, the next 8.17.x versions will consist of bug corrections (if reported) or minor modifications.
However the development branch is still in progress ! Contact us if interested !
AMITEX_FFTP v8.17.13¶
Date : September 2023 (8.17.13)
- bug correction : non monotonous loading path + finite strain, option -coeff2print, + small bug corrections
AMITEX_FFTP v8.17.11¶
Date : March 2022 (8.17.11)
- small bug corrections
- compatibility with intel-oneapi compiler (version tested 2021.1)
AMITEX_FFTP v8.17.9¶
Date : May 2021 (8.17.9)
- modification for strain gradient loadings (i.e. torsion/flexion) => grad(grad(U)) components are now given independantly.
Date : april 2021 (8.17.8)
- possibility to differentiate 2decomp/amitex optimization options in file
install - bug corrections (output)
The first point is required to run the code on ‘ubuntu 20.04’ (with native gcc9.3.0) or ‘Opensuse Tumbleweed’ (with gcc10) :
in install, OPTIM_2decomp must be set to “-O2” instead of “-O3” otherwise the code (2decomp, in fact) fails at execution (see Installation guide)
AMITEX_FFTP v8.17.2¶
Date : july 2020
Important modifications have been introduced since the last update.
The most important are :
- introduction of composite voxels
- new command line functionalities (type
amitex_fftp -help) - constant stress direction loading together with an applied strain component (small and finite strains)
- beam and plate loadings (torsion and flexion)
- code reorganization to facilitate couplings
- various bug corrections and small improvements
- a new web site!
TRAINING SESSION¶
Download the last training session (10/10/2023)
The purpose of the training session is to help users to :
- install the code and launch simulations
- generate VTK input files, with matlab/octave scripts and functions
- become familiar with the different xml files (material, loading, algorithm)
Then, uncompressing the archive will give you one directory:
tar xvf amitex_formation_v1.0.3.tar
cd amitex_training
ls
Algorithms Behaviors Loadings_outputs matlab_octave Scripts
amitex_training_pdf env_amitex_training.sh Materials Microstructures update_solution
The document amitex_training_pdf/FORMATION_AMITEX_general_xxx.pdf will give you a general description of AMITEX_FFTP and its Standard usage.
Then follow the guide in amitex_training_pdf/FORMATION_AMITEX_practice_xxx.pdf.
Note that the directories in amitex_training contain a solutions directory from wich you can copy-paste the files in their parent directories to run the simulations proposed in Scripts.
Of course, prerequisites (see Installation) have to be fullfilled and the environment variables in env_amitex_training.sh have to be adjusted to you configuration (see HOW TO LAUNCH AMITEX_FFTP).
This will be the first part of the training session.
Note that this WEB SITE is the REFERENCE guide. If you observe any anomaly please contact us (contacts).
VALIDATION TESTS¶
The compuation time for validation tests depends of course on your hardware, compiler and MPI library. A few results are given below for comparison.
Simulation on 1 node (skylake 28 proc.: Bi-Proc Intel Xeon 14 cores 6132 2.60G) :
- gcc8.1.0 - openMPI3.1.0 - fftw3.3.2 : 371s (on 28 proc)
- intel15 - intelMPI5 - fftw3.3.2 : 203s (on 28 proc)
An important, and not understood difference is observed between intel and gnu configurations.
VERSION 4.0.0¶
Important modifications have been introduced with respect to version 2.3.5.
The most important for the users are :
- a simplified way to introduce user defined behaviours, still compatible with MFront (http://tfel.sourceforge.net/index.html)
- a new possiblity to take into account diffusion problems (stationary problems with a unique variable for the moment, like thermal diffusion)
- the dimensions of the voxels (SPACING in vtk files) are now taken into account
- bug corrections (especially the problem with the recent versions of openMPI)
Warning
Version 2.3.5 input xml files are no more compatible
These files have to be slightly adapted
VERSION 2.3.5¶
Only minor modifications have been introduced with respect to version 2.3.4. The most important for the users are :
- bug fixed : writing of messages on the standard output when using gcc compiler
- introduction of an anisotropic elastic behaviour (Law_Number : 3)
Validation tests have been run on the CEA cluster (maldives). Compared to version 2.3.4, validation tests on the other platforms should no be affected.
VERSION 2.3.4¶
Validation tests have been run on various clusters, or standalone PC, but only with intel XEON processors, with the following results:
| CCRT cluster (airain) | intel 14 | bullxmpi 1.2.8.4 | OK |
| MDS cluster (poincare) | gcc 4.9.0 | openMPI 1.8.4 | OK |
| gcc 4.7.2 | openMPI 1.6.3 | OK | |
| intel 13 | intelMPI 4 | OK | |
| intel 15 | intelMPI 5 | OK | |
| CEA cluster (maldives) | intel 15 | intelMPI 5 | OK |
| gcc 4.8.2 | openMPI 1.6.5 | OK | |
| gcc 4.8.2 | openMPI 1.8.1 | PB | |
| Standalone PC | gcc 4.9.2 | openMPI 1.6.5 | OK |
The problem with openMPI 1.8.1 was observed on the validation test beton_relax_65_old.
Warning
Performances obtained with [Intel compiler + IntelMPI] are significantly better than with [GCC compiler + openMPI].