The smooth vortex test is a classical test for MHD codes. It is described

for instance in \cite{dumbserJCP2016}. Because this is a exact solution,

for instance in \cite{dumbserJCP2016}. Because this is an exact solution,

it allows to assess the accuracy of the solver. Here we also used

this test to evaluate the efficiency of the parallel implementation.

The test case is built upon a single vortex, which is a stationary

...

...

@@ -938,7 +938,7 @@

\end{tabular}\caption{Convergence and performance study\label{tab:Convergence-and-performance}.

Some tests are done in single precision (float32) and others in double

precision (float64). The ``efficiency'' is a comparison for N=1024

with the slowest device. "CU" means "Compute Units": it is the number of activated cores in a CPU computation or of OpenCL processing elements for a GPU computation. }

with the slowest device. "CU" means "Compute Units": it is the number of activated cores in a CPU computation or of OpenCL compute units for a GPU computation. }