[BopngMa]

Hello, everyone

I installed SU2 4.0.1 with MPI and CGNS support correctly, and I had tried a 2d airfoil of Nasa0714, it ran well. But when I ran a simulation of 3d mesh, I found that when I use a CGNS mesh, SU2 cannot load mixed mesh, then I tested tetrahedral mesh, it ran well, so my first question is that is it right that SU2 can load tetrahedral or hexahedral mesh only for now? My second question came from the same case, when I used the CGNS mesh, I can only use at most 4 threads for parallel, but when I used SU2_DEF command to transfer the mesh to .SU2 format, I can use as many cores as I can(I tested 16). So is this also all right?

Another question is about the version of SU2. I'm interested in Discrete Adjoint feature which Prof. Gauger's group has been working on. I noticed from the maillist that the discrete adjoint feature has been added to SU2 and I want to test it. I downloaded(and tested) su2_develop(and su2_v4.0.1, su2_feature_ad also) from github, and found that the option of --enable-codi=** has been added to the configuration lists, but the testcases has not changed. So did I miss anything? Please help me.

Thanks.

[BopngMa]

With the newer version of SU2_develop, the second problem partly solved.

cheers

[economon]

Hi,

Thanks for your questions. I'll give some short responses here:

1. For CGNS, we do support mixed elements meshes, but they should still be a single zone mesh. Check that your meshing program outputs the mesh as a single block/zone. Currently, we know that mixed meshes from Pointwise and ICEM are working.

2. The number of ranks for MPI is independent of the modules. You should be able to run as many as your hardware will support for SU2_CFD or SU2_DEF. Check that you are not running out of memory or having another issue with the compilation.

3. The AD work is currently in verification/testing mode, so we are still ironing out issues and making lots of changes. There will be more information coming soon, including a tutorial.

Hope this helps,
Tom