There were a couple of bugs at the outlet and viscous flux calculation. These have been fixed and now the differences
between CPU/GPU calculations after the first iteration are <1e-12.
After 400 iterations the relative differences, e.q. (p_gpu-p_cpu)/p_cpu, is still below 1e-6.
I tried to include the old reference files in a new directory but git would not see them for some reason. I have the files ready in case we want to include them here
There were a couple of bugs at the outlet and viscous flux calculation. These have been fixed and now the differences between CPU/GPU calculations after the first iteration are <1e-12. After 400 iterations the relative differences, e.q.
(p_gpu-p_cpu)/p_cpu,
is still below 1e-6.