[Ranjan]

Hello

I have tried running the RAE2822 shape optimization case on multiple cores. The optimal solution differs with the number of cores. Looking deeper into the design process, it turns out the change in angle of attack computed differs with the number of cores while the direct flow solution is the same.

The initial difference is in the order of 1e-07, but this difference gets amplified as one proceeds to various optimization iterations.

I am using the shape_optimization.py script so I guess the mesh partitioning scheme is common among all cores.

I can always reduce the numerical precision but that will reduce my solution accuracy.

Any ideas ?

Thanks in advance.