Difference between revisions of "Scale"
Line 1: | Line 1: | ||
== How to run Scale == | == How to run Scale == | ||
− | The first step is to ensure that you have permission to use the current version of Scale from [http://rsicc.ornl.gov/Default.aspx RSICC]. Once you have permission, let me know so I can add you to the '' | + | The first step is to ensure that you have permission to use the current version of Scale from [http://rsicc.ornl.gov/Default.aspx RSICC]. Once you have permission, please let me know, so I can add you to the ''scale63'' (or other version) group on the cluster. You can then load the Scale module with <code>module load scale</code>. |
Make sure you run Scale either in a PBS script (for Torque), or as an interactive job. Never run a job on the head node. | Make sure you run Scale either in a PBS script (for Torque), or as an interactive job. Never run a job on the head node. | ||
− | === Running Scale (6.1.1) === | + | Example scripts for SCALE are here: [https://necluster.ne.utk.edu/wiki/index.php?title=TORQUE/Maui#Scale Running SCALE] |
+ | |||
+ | === SCALE 6.3.1 === | ||
+ | SCALE 6.3.1 is available in 3 module versions on NEcluster. | ||
+ | Unless you have a specific reason otherwise, use the first one below. | ||
+ | |||
+ | * scale/6.3.1 - Serial binaries distributed by ORNL. This should be used by most users. | ||
+ | * scale/6.3.1-mpi - NEcluster-specific MPI build. Useful for parallel KENO, SHIFT, and DENOVO. You should know details about your calculation (memory and IO needs) before using these. | ||
+ | * scale/6.3.1-mpi-debug - MPI build with debugging hooks in. Useful if your case crashes, and you want to find out why. | ||
+ | |||
+ | THE REST OF THIS PAGE IS OBSOLETE | ||
+ | |||
+ | === Running Old Scale (6.1.1) === | ||
==== Basic Run ==== | ==== Basic Run ==== |
Latest revision as of 16:01, 17 November 2023
Contents
How to run Scale
The first step is to ensure that you have permission to use the current version of Scale from RSICC. Once you have permission, please let me know, so I can add you to the scale63 (or other version) group on the cluster. You can then load the Scale module with module load scale
.
Make sure you run Scale either in a PBS script (for Torque), or as an interactive job. Never run a job on the head node.
Example scripts for SCALE are here: Running SCALE
SCALE 6.3.1
SCALE 6.3.1 is available in 3 module versions on NEcluster. Unless you have a specific reason otherwise, use the first one below.
- scale/6.3.1 - Serial binaries distributed by ORNL. This should be used by most users.
- scale/6.3.1-mpi - NEcluster-specific MPI build. Useful for parallel KENO, SHIFT, and DENOVO. You should know details about your calculation (memory and IO needs) before using these.
- scale/6.3.1-mpi-debug - MPI build with debugging hooks in. Useful if your case crashes, and you want to find out why.
THE REST OF THIS PAGE IS OBSOLETE
Running Old Scale (6.1.1)
Basic Run
Running Scale in it's most basic form involves calling the Scale 6.1 script with several arguments. As an example, you can run Scale 6.1 with the input file called myInput.inp as such:
user@node:~/$ batch6.1 myInput.inp
This will run your input on the node that you're currently logged in on (don't use the head node!).
Running Scale (6.1.2) [default]
The only difference is that Scale 6.1.2 depends on the intel module and the openmpi/1.6.4 module. Ensure that these are loaded before you load the Scale 6.1.2 module. Use this version since it has bugs fixed.
Running Scale (dev)
Basic Run
You can run the Scale development version the same way as version 6.1 after loading the scale/dev module:
module load scale/dev
KENO MPI
Currently, KENO MPI doesn't work with the version of OpenMPI provided by Fedora. I have provided a version of OpenMPI (v1.4.3 OR 1.6.4 both work) that works with KENO MPI as a module for those users who are adventurous enough to compile Scale on their own. The following system third-party libraries will continue to work with the old version of OpenMPI (as they don't require MPI):
BLAS LAPACK Qt GSL Zlib
The following TPLs will *not* work as compiled with the old version of OpenMPI:
HDF5 SILO
The adventurous user can check out Scale, load the module openmpi/1.4.3 (or openmpi/1.6.4 if you used that) (after unloading the system OpenMPI implementation), and use the configuration script /opt/scale_dev/script/necluster-gnu-cmake-mpi to compile Scale with KENO MPI support.