SOWER Manual

Version 1.1

Edition 0.1/25 August 2023



















Farhat Research Group (FRG)
Stanford University



This manual was prepared with Texinfo (http://www.gnu.org/software/texinfo).

Short Contents

Table of Contents


Up: (dir)

SOWER


Next: 

1 Introduction

SOWER is a pre/post-processing software which was originally developed at the Center for Aerospace Structures at the University of Colorado to support the parallel execution of the AERO-S and AERO-F3D codes. This User's Manual documents the usage of SOWER.

SOWER can be used to perform the following tasks

Both the serial and parallel executions of the AERO-F code require the generation of appropriate binary input files using SOWER. The parallel execution of AERO-S for aeroelastic and other coupled AERO-F/AERO-S simulations requires the conversion of the matcher output files to binary format using SOWER (see also AERO-S's User's Manual for other cases where generating appropriate binary input files for this code is required).


Next: , Previous: Introduction

2 Input Files

This sections describes the input files used by SOWER for pre/post-processing the AERO-S and AERO-F codes. The input file names do not follow any naming convention.


Next: , Up: InputFiles

2.1 Input Files for AERO-F


Previous: Input Files for AERO-F, Up: InputFiles

2.2 Input Files for FEM


Next: , Previous: InputFiles

3 Output Files

For the purpose of parallel processing, AERO-F and AERO-S organize their computations and corresponding output around the concepts of subdomains and “clusters”, respectively. A subdomain is a collection cells, and/or dual cells, and/or elements and grid points (among others) that is assigned, possibly with other subdomains, to a single processor. An output cluster pertains to a collection of subdomains that are assigned to a single processor. Hence, the concepts of processors, subdomains, and clusters are related but the number of processors, number of subdomains, and number of clusters can be chosen independently with the constraint that each of the number of clusters and number of processors is less or equal to the number of subdomains. If one or multiple output clusters are assigned per processor, AERO-F and AERO-S write one or multiple binary output files per result and per processor containing each the trace of the global result on the subdomains defining the cluster. This strategy delivers optimal parallel I/O performance on distributed memory machines. Similarly, if one or multiple output clusters are assigned per “box” of processors, AERO-F and AERO-S write one or multiple binary output files per result and per box machine containing each the trace of the global result on the subdomains defining the cluster. This strategy delivers optimal parallel I/O performance on hybrid memory machines. Finally, if a single output cluster is assigned to all processors, AERO-F and AERO-S write a single binary output file per result for the entire mesh. This strategy is convenient for shared memory machines. In pre-processing mode, SOWER supports this approach to parallel execution and parallel I/O adopted by AERO-F and AERO-S by generating for each of them the required subdomain and cluster information based on the standard ASCII input to these codes. In post-processing mode, SOWER can be used to assemble all distributed binary files associated with the same result into a single ASCII file.

The following binary output files are generated by SOWER. They follow a suffix naming convention for quick identification of their purpose. Similar output files are produced for both the AERO-F and AERO-S codes.



Warning: existing files with same names are overwritten!








Next: , Up: OutputFiles

3.1 Pre-Processing Data Files

The following output files are generated by SOWER when used to generate the binary input files for the AERO-S and AERO-F codes.


Previous: Pre-Processing Data Files, Up: OutputFiles

3.2 Post-Processing Data Files

After executing the AERO-S or AERO-F distributed code, SOWER can be used to merge all distributed binary output files associated with the same result into a single ASCII output file readable by XPost as explained in the next section.


Previous: OutputFiles

4 Line Command Input

The different line-commands for performing the various tasks described in the introduction section of this manual are as follows.


Next: , Up: LineCommandInput

4.1 Pre-Processing Commands

This section describes the commands for producing the binary inputs for the AERO-F and AERO-S codes. The options are the same for either code.


Next: , Previous: Pre-Processing Commands, Up: LineCommandInput

4.2 Post-Processing Commands

This section describes the commands for post-processing the output files from the AERO-S and AERO-F codes. Post-processing of the AERO-F data files is necessary for visualization with the XPost software. However, post-processing of the structure data files is only necessary if the input files were in binary format (i.e. SOWER was used to generate the input files for a simulation in distributed mode).


Previous: Post-Processing Commands, Up: LineCommandInput

4.3 Other Processing Commands

This section describes other miscellaneous commands for pre/post-processing the input/output files from the AERO-S and AERO-F codes.

In particular, this command can be used to convert a distance-to-the-wall ASCII file filename.dwall generated by CD2TET (see the CD2TET User's Reference Manual) to a binary distributed file.

Furthermore, since the MESHTOOLS software — which can be used to perform planar cuts in geometry and result files — operates only on binary AERO-F output files, the above command can also be useful when wanting to “cut” a result file available in the ASCII XPost format.

In the context of a sensitivity analysis, it is also useful for splitting and transforming the ASCII file containing the derivatives of the mesh position at a wall boundary with respect to a number of shape design variables.