The multiphase solver in the gpu code (compiled in cpu or gpu) doesn's provide the same results as horses3d-master. The only test case that is working in multiphase (same results in gpu-gpu, gpu-cpu and horses-master) is the EntropyConservingTest. However, if the density of the fluid 2 is increased (for example to 100):
fluid 2 density (kg/m^3) = 100.0
Then the difference between the codes keeps growing. After 2.000 iterations the error is around 10^-7. After 10.000 iterations, the difference is one order of magnitude for some residuals.
@Jerryntk any idea where the problem may come from? It could be anything. Also, it's weird we only see the issue when the density ratio is high.
The multiphase solver in the gpu code (compiled in cpu or gpu) doesn's provide the same results as horses3d-master. The only test case that is working in multiphase (same results in gpu-gpu, gpu-cpu and horses-master) is the EntropyConservingTest. However, if the density of the fluid 2 is increased (for example to 100):
fluid 2 density (kg/m^3) = 100.0Then the difference between the codes keeps growing. After 2.000 iterations the error is around 10^-7. After 10.000 iterations, the difference is one order of magnitude for some residuals.
@Jerryntk any idea where the problem may come from? It could be anything. Also, it's weird we only see the issue when the density ratio is high.