Altair ultraFluidX 2022 Release Notes
Highlights
- Improved triangle splitting based on relative edge length; includes optional settings per part
- Automatic Overset Mesh refinement corrections
New Features
- Improved triangle splitting
- Surface triangles of the input geometry are now split based on a maximum edge length relative to the local voxel size. The old default (splitting based on a global reference length) can be reactivated via the input deck. If necessary, the settings for the triangle splitting can also be modified on a per-part basis.
- Automatic Overset Mesh refinement corrections
- Multiple refinement levels inside overset mesh (OSM) regions are supported. The interface that connects the OSM to the static background mesh must be uniformly refined.
Enhancements
- Improved seeding feature
- The seeding functionality in ultraFluidX 2022 now has the option to read pressure fields for the seeding in addition to the velocity fields. Moreover, a non-zero fallback pressure can be specified in case that no source information is found in the input data set. The ramp-up behavior for seeded runs is also improved and further improves the convergence behavior.
- Improved probe output
- The default output format for probes is now scientific notation with eight significant digits. For surface probes, the output of pressure is now a default, in addition to the non-dimensional cp value.
- Improved sectional aero coefficients
- The sectional aero coefficient output is revised. A new file (uFX_coefficientsPerSection <X/Y/Z>_location.txt) includes the global x/y/z positions of the sections in the global reference frame. The section index is also added to the mesh preview.
- Efficient alternative to mesh preview
- The mesh data, written into the separate ultraFluidX mesh preview files upon request, is now also available per rank in the ultraFluidX full data files by default.
- Library updates (CUDA, Open MPI, UCX)
- ultraFluidX 2022 utilizes more recent versions of CUDA (version 11.4.3), Open MPI (version 4.1.2), and the Unified Communication X (UCX) framework (version 1.12.0). This does not affect the minimum required CUDA driver version (450.80.02 or higher is required).
Resolved Issues
- Report on GPU memory consumption was based on the total amount of allocated memory on the device. Now the reported values are associated only with the current compute process.
- Precision issues in the computation of the OSM interface location that might lead to local artifacts in the flow field are now resolved.
- The Moment Reference System is now normalized upon solver deck reading.