Bug description
I'm running the AeroDyn standalone version calling a model using OLAF. I performed a sensitivity analysis using an instance without parallelization and two instances with parallelization (OpenMP) using different number of threads: 8 and 128. For reference, the desired number of threads can be defined in Windows cmd as follows: set OMP_NUM_THREADS=8
During the first 25 seconds of the simulation, the outputs are relatively similar although not the same.
When looking at the wind speed in the longitudinal direction (turbulent inflow wind), I can see:
Although the input looks very similar, they behave differently after 6.66 seconds when using parallelization. The differences are not large (around 0.5%), but noticeable.
Actually, when looking at the input wind, it can be observed that the runs with parallelization include more oscillations.
The same phenomena is observed in the 3 components of the wind:
For context, the grid used in TurbSim did not use an odd number of points. Therefore, tthe wind at the hub location is internally interpolated.
Differences can also be observed in the aerodynamic loads. Although in this case, the differences start before the 6.66 s mark. See below for example:
OpenFAST Version
I'm using OpenFAST v5.0.0:
Bug description
I'm running the AeroDyn standalone version calling a model using OLAF. I performed a sensitivity analysis using an instance without parallelization and two instances with parallelization (OpenMP) using different number of threads: 8 and 128. For reference, the desired number of threads can be defined in Windows cmd as follows:
set OMP_NUM_THREADS=8During the first 25 seconds of the simulation, the outputs are relatively similar although not the same.
When looking at the wind speed in the longitudinal direction (turbulent inflow wind), I can see:
Although the input looks very similar, they behave differently after 6.66 seconds when using parallelization. The differences are not large (around 0.5%), but noticeable.
Actually, when looking at the input wind, it can be observed that the runs with parallelization include more oscillations.
The same phenomena is observed in the 3 components of the wind:
For context, the grid used in TurbSim did not use an odd number of points. Therefore, tthe wind at the hub location is internally interpolated.
Differences can also be observed in the aerodynamic loads. Although in this case, the differences start before the 6.66 s mark. See below for example:
OpenFAST Version
I'm using OpenFAST v5.0.0: