Computational costs associated with different con2prim schemes in terms of number of EOS interpolations required. We plot the histograms of frequency versus number of EOS calls using 20 bins between 0 and 230 (280, 4000) for 3D Newton–Raphson (2D Newton–Raphson, 1D Ridders). We also show the mean and median numbers of EOS calls using dashed red and green lines, respectively, on top of the histograms. We compare the costs of the 3D Newton–Raphson, 2D Newton–Raphson, and 1D Ridders’​​​​​​​ methods for Θ = 2.29 in the top panels, Θ = 3.47 in the middle panels, and Θ = 10.26 in the bottom panels. We find that the 1D Ridders’​​​​​​​ method, even though the most accurate and robust, is also the most expensive among all schemes, with median EOS calls ≳650. On the other hand, 3D Newton–Raphson and 2D Newton–Raphson, though less robust, are much cheaper, with median EOS calls ≲15.