What's New

Hybrid Precision

In cases with large differences in scale, small changes in variables may be masked, impacting the behavior of the simulation. Running these cases in double precision has a significant runtime penalty and restricts the simulation to flagship GPUs. Hybrid precision offers a solution by making it possible to account for the most important phenomena at a small overhead compared to the equivalent single precision simulation.

Sparse addressing infrastructure

Addressing infrastructure can now be dynamically adjusted to cover only those regions in computational domain with fluid particle presence. This reduces memory consumption in cases with low ratio of fluid volume to overall computational domain volume, allowing for handling larger domain sizes.

64-bit addressing infrastructure

A larger address space to support sparsely populated computational domains with a high ratio of domain size to particle spacing is now used. In practice, this extends the solver capability to handle water management cases with very large domain sizes or very small particle spacings. Due to inherent nature of these cases requiring double precision arithmetic, larger addressing infrastructure is only available in hybrid and double precision executables.

Inquiry probe for regulating inlet region velocity

Relative volume flowrate reading of a probe can now be used to adjust the inlet region velocity. This facilitates linking inflow and outflow rates at the boundaries of the computational domain and maintaining a predetermined volume of fluid within the simulation.

Supporting ParaView 5.13 series

All output and state files generated by the solver and auxiliaries are adjusted to support ParaView versions 5.9.1 to 5.13.2.

Planetary motion day rotation in carrier arm reference frame

Day rotation frequency of a planet gear in a planetary gear system can now be defined in either inertial or carrier arm reference frame. Non-parallel planet day and year rotation axes are also supported when carrier arm reference frame is used to define planet motion.

Extended position file motion treatment

Alternative treatment methods are included in the position file motion definition to calculate velocity and acceleration at position file resolution and improve simulation response to high frequency position files.

Two new tutorials for water management

Two new tutorials have been added for windscreen – cowl water management and cab roof rainwater runoff.

Operation modes

Operation modes are deprecated and will be removed in a future release.

• Angle limit constraint in passive rigid body motion now follows the specified rotation range in certain configurations.
• Missing synchronized surface extractor output at restart/continue output times resulted in VTK errors in nFX[c] jobs with vtkp OutputFormat. The issue has now been resolved.

• Up to 1.2x performance improvement in Tartakovsky F1 surface tension model.

Optional reference velocity in body frame

Specifying reference velocity is now optional when body frame is active. This simplifies case setup where high acceleration events are more dominant than high speed events, for example in predominantly vibrational tank sloshing cases.

Independent interpolation domain size in nFX[c]

Interpolation domain size of nFX[c] can now be set independently from nanoFluidX. This simplifies narrowing down the post-processing data to the volume of interest.

Performance improvement in Tartakovsky F1 surface tension model

Tartakovsky F1 surface tension model implementation has been modified, offering up to 1.2x speedup compared to prior implementation.

Load balancing adjustment for cases with large wetting area

Load balancing method in Weighted interaction scheme has been adjusted to account for cases with larger wetted surface to fluid bulk ratio, offering performance improvements for some water management and water wading cases.

Memory usage summary

In addition to the running memory usage, a summary of peak memory usage of host and device can now be shown at the end of the simulation log. The summary is activated via specifying relevant options in the configuration file.

Availability of cell mesh information prior to simulation

Domain addressing and cell mesh data can now be viewed when running nFX[p], nFX[c] in --nfxp mode or Data Check in SimLab.

• Total number of interpolation points exceeding 2147483647 led to unexpected behavior in nFX[c].