I would like to use the AMGX library in my CFD calculations.
Currently, I am investigating to find a suitable combination among the various settings of the library for my CFD solver.
However, although it works with certain settings, it fails frequently when I change the settings.
In particular, I would like to use the FGMRES method, but it does not work well.
So, I would like advices on whether there is a problem with my matrix data or if I should change some settings in FGMRES.
Furthermore, I would like to know if there are any recommended settings in AMGX to solve matrices like this.
As an example, I upload the matrix data I want to solve in my CFD solver: “rae2822_2d_sample.mtx”.
This is a matrix data for compressible flow analysis around a two-dimensional airfoil.
The dimension of the matrix in terms of block-units is 21,739. The size of block is 5.
I made this file by AMGX_write_system() of the library.
This matrix data can be solved with “config_sample.json”, but it fails with “FGMRES_AGGREGATION.json”.
Here, the latter one is used in QuickStart of README.
For example, if you run with the following command as in QuickStart, the calculation will fail.
I would like to know why it fails and how to avoid it.
amgx_capi.exe -m rae2822_2d_sample.mtx -c FGMRES_AGGREGATION.json
Thank you in advance.
samples.7z (25.6 MB)