Greetings impatient user. Once again, note that this documentation focuses on the CHEMISTRY part of PelePhysics.
If you are familiar with PelePhysics, have it installed already and would simply like to know which chemistry-related keywords and/or environment variables to set in your various input files to perform a simulation with one of the codes available in the PeleSuite, then I invite you to directly skip to section Activating the different CVODE solver options via the input files.
If you are in a hurry but still would like more context, visit section Introduction to be referred to portions of this document that are of interest to you.
First, make sure that “Git” is installed on your machine—we recommend version 1.7.x or higher. Then…
Download the AMReX repository by typing:
git clone https://github.com/AMReX-Codes/amrex.git
This will create an
amrex/ folder on your machine. Next, set the environment variable
AMREX_HOME to point to the location where you have downloaded AMReX:
Clone the Pele repository:
git clone firstname.lastname@example.org:AMReX-Combustion/PelePhysics.git
This will create a
PelePhysics folder on your machine. Set the environment variable
PELE_PHYSICS_HOME to point to the location of this folder.
Periodically update both of these repositories by typing
git pullwithin each repository.
Install CVODE and SuiteSparse
*The user is in charge of installing the proper CVODE version, as well as installing and properly linking the KLU library if sparsity features are needed.
SuiteSparse is a suite of Sparse Matrix software that is very handy when dealing with big kinetic mechanisms (the Jacobian of which are usually very sparse). In such a case, CVODE can make use of the KLU library, which is part of SuiteSparse, to perform sparse linear algebra. Documentation and further information can be found on SuiteSparse website.
At the time this note is written, the recommended SuiteSparse version is 5.4.0. Follow these steps to set-up your working environment and build the required libraries:
Go to the SuiteSparse website and download the compressed file for the recommended version
Copy the tar file into
Untar (‘tar -zxvf’), cd into it and type ‘make’ into the following folders:
metis-5.1.0and type ‘make config shared=1’ followed by ‘make’
KLUand type ‘make’
Check that all dynamic libraries have correctly been generated and copied into the folder
It is recommended that you add the path
LD_LIBRARY_PATH, for precaution
Note that depending upon your compiler, the static
.aversions of the libraries might also be required. In such a case, you can copy them directly from each program folder into the
CVODE is a solver for stiff and nonstiff ordinary differential equation (ODE) systems. Documentation and further information can be found online. At the time this note is written, the recommended CVODE version is v5.0.0.
The CVODE sources are distributed as compressed archives, with names following the convention
cvode-x.y.z.tar.gz. They can be downloaded by following
this link. However, we have designed a simple script enabling to install the current version the correct way. Simply:
get_sundials_v5dev1_CUDA.shdepending on your application (GPU or not) and machine
SUNDIALS_LIB_DIRenvironment variable to point to the location where all CVODE libraries have been generated. If you followed these guidelines, it should be
It is recommended, here also, that you add the path
LD_LIBRARY_PATH, as paths can get lost in the build of external libraries
Note that if you do not want to use the KLU library, you can also disable the flags (
-DKLU_ENABLE) in the